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()).
181 #define pr_fmt(fmt) "bcachefs: %s() " fmt "\n", __func__
183 #define pr_fmt(fmt) "%s() " fmt "\n", __func__
186 #include <linux/backing-dev-defs.h>
187 #include <linux/bug.h>
188 #include <linux/bio.h>
189 #include <linux/closure.h>
190 #include <linux/kobject.h>
191 #include <linux/list.h>
192 #include <linux/math64.h>
193 #include <linux/mutex.h>
194 #include <linux/percpu-refcount.h>
195 #include <linux/percpu-rwsem.h>
196 #include <linux/rhashtable.h>
197 #include <linux/rwsem.h>
198 #include <linux/semaphore.h>
199 #include <linux/seqlock.h>
200 #include <linux/shrinker.h>
201 #include <linux/srcu.h>
202 #include <linux/types.h>
203 #include <linux/workqueue.h>
204 #include <linux/zstd.h>
206 #include "bcachefs_format.h"
212 #define dynamic_fault(...) 0
213 #define race_fault(...) 0
215 #define bch2_fs_init_fault(name) \
216 dynamic_fault("bcachefs:bch_fs_init:" name)
217 #define bch2_meta_read_fault(name) \
218 dynamic_fault("bcachefs:meta:read:" name)
219 #define bch2_meta_write_fault(name) \
220 dynamic_fault("bcachefs:meta:write:" name)
223 #define bch2_fmt(_c, fmt) "bcachefs (%s): " fmt "\n", ((_c)->name)
224 #define bch2_fmt_inum(_c, _inum, fmt) "bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
226 #define bch2_fmt(_c, fmt) fmt "\n"
227 #define bch2_fmt_inum(_c, _inum, fmt) "inum %llu: " fmt "\n", (_inum)
230 #define bch_info(c, fmt, ...) \
231 printk(KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
232 #define bch_notice(c, fmt, ...) \
233 printk(KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
234 #define bch_warn(c, fmt, ...) \
235 printk(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
236 #define bch_warn_ratelimited(c, fmt, ...) \
237 printk_ratelimited(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
238 #define bch_err(c, fmt, ...) \
239 printk(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
241 #define bch_err_ratelimited(c, fmt, ...) \
242 printk_ratelimited(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
243 #define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
244 printk_ratelimited(KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
246 #define bch_verbose(c, fmt, ...) \
248 if ((c)->opts.verbose) \
249 bch_info(c, fmt, ##__VA_ARGS__); \
252 #define pr_verbose_init(opts, fmt, ...) \
254 if (opt_get(opts, verbose)) \
255 pr_info(fmt, ##__VA_ARGS__); \
258 /* Parameters that are useful for debugging, but should always be compiled in: */
259 #define BCH_DEBUG_PARAMS_ALWAYS() \
260 BCH_DEBUG_PARAM(key_merging_disabled, \
261 "Disables merging of extents") \
262 BCH_DEBUG_PARAM(btree_gc_always_rewrite, \
263 "Causes mark and sweep to compact and rewrite every " \
264 "btree node it traverses") \
265 BCH_DEBUG_PARAM(btree_gc_rewrite_disabled, \
266 "Disables rewriting of btree nodes during mark and sweep")\
267 BCH_DEBUG_PARAM(btree_shrinker_disabled, \
268 "Disables the shrinker callback for the btree node cache")\
269 BCH_DEBUG_PARAM(verify_btree_ondisk, \
270 "Reread btree nodes at various points to verify the " \
271 "mergesort in the read path against modifications " \
273 BCH_DEBUG_PARAM(verify_all_btree_replicas, \
274 "When reading btree nodes, read all replicas and " \
277 /* Parameters that should only be compiled in in debug mode: */
278 #define BCH_DEBUG_PARAMS_DEBUG() \
279 BCH_DEBUG_PARAM(expensive_debug_checks, \
280 "Enables various runtime debugging checks that " \
281 "significantly affect performance") \
282 BCH_DEBUG_PARAM(debug_check_iterators, \
283 "Enables extra verification for btree iterators") \
284 BCH_DEBUG_PARAM(debug_check_bkeys, \
285 "Run bkey_debugcheck (primarily checking GC/allocation "\
286 "information) when iterating over keys") \
287 BCH_DEBUG_PARAM(debug_check_btree_accounting, \
288 "Verify btree accounting for keys within a node") \
289 BCH_DEBUG_PARAM(journal_seq_verify, \
290 "Store the journal sequence number in the version " \
291 "number of every btree key, and verify that btree " \
292 "update ordering is preserved during recovery") \
293 BCH_DEBUG_PARAM(inject_invalid_keys, \
294 "Store the journal sequence number in the version " \
295 "number of every btree key, and verify that btree " \
296 "update ordering is preserved during recovery") \
297 BCH_DEBUG_PARAM(test_alloc_startup, \
298 "Force allocator startup to use the slowpath where it" \
299 "can't find enough free buckets without invalidating" \
301 BCH_DEBUG_PARAM(force_reconstruct_read, \
302 "Force reads to use the reconstruct path, when reading" \
303 "from erasure coded extents") \
304 BCH_DEBUG_PARAM(test_restart_gc, \
305 "Test restarting mark and sweep gc when bucket gens change")
307 #define BCH_DEBUG_PARAMS_ALL() BCH_DEBUG_PARAMS_ALWAYS() BCH_DEBUG_PARAMS_DEBUG()
309 #ifdef CONFIG_BCACHEFS_DEBUG
310 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALL()
312 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALWAYS()
315 #define BCH_DEBUG_PARAM(name, description) extern bool bch2_##name;
317 #undef BCH_DEBUG_PARAM
319 #ifndef CONFIG_BCACHEFS_DEBUG
320 #define BCH_DEBUG_PARAM(name, description) static const bool bch2_##name;
321 BCH_DEBUG_PARAMS_DEBUG()
322 #undef BCH_DEBUG_PARAM
325 #define BCH_TIME_STATS() \
326 x(btree_node_mem_alloc) \
327 x(btree_node_split) \
328 x(btree_node_compact) \
329 x(btree_node_merge) \
332 x(btree_interior_update_foreground) \
333 x(btree_interior_update_total) \
335 x(btree_lock_contended_read) \
336 x(btree_lock_contended_intent) \
337 x(btree_lock_contended_write) \
341 x(journal_flush_write) \
342 x(journal_noflush_write) \
343 x(journal_flush_seq) \
345 x(blocked_allocate) \
346 x(blocked_allocate_open_bucket)
348 enum bch_time_stats {
349 #define x(name) BCH_TIME_##name,
355 #include "alloc_types.h"
356 #include "btree_types.h"
357 #include "buckets_types.h"
358 #include "clock_types.h"
359 #include "ec_types.h"
360 #include "journal_types.h"
361 #include "keylist_types.h"
362 #include "quota_types.h"
363 #include "rebalance_types.h"
364 #include "replicas_types.h"
365 #include "subvolume_types.h"
366 #include "super_types.h"
368 /* Number of nodes btree coalesce will try to coalesce at once */
369 #define GC_MERGE_NODES 4U
371 /* Maximum number of nodes we might need to allocate atomically: */
372 #define BTREE_RESERVE_MAX (BTREE_MAX_DEPTH + (BTREE_MAX_DEPTH - 1))
374 /* Size of the freelist we allocate btree nodes from: */
375 #define BTREE_NODE_RESERVE (BTREE_RESERVE_MAX * 4)
377 #define BTREE_NODE_OPEN_BUCKET_RESERVE (BTREE_RESERVE_MAX * BCH_REPLICAS_MAX)
382 GC_PHASE_NOT_RUNNING,
386 GC_PHASE_BTREE_stripes,
387 GC_PHASE_BTREE_extents,
388 GC_PHASE_BTREE_inodes,
389 GC_PHASE_BTREE_dirents,
390 GC_PHASE_BTREE_xattrs,
391 GC_PHASE_BTREE_alloc,
392 GC_PHASE_BTREE_quotas,
393 GC_PHASE_BTREE_reflink,
394 GC_PHASE_BTREE_subvolumes,
395 GC_PHASE_BTREE_snapshots,
397 GC_PHASE_PENDING_DELETE,
412 typedef GENRADIX(struct reflink_gc) reflink_gc_table;
415 u64 sectors[2][BCH_DATA_NR];
420 struct percpu_ref ref;
421 struct completion ref_completion;
422 struct percpu_ref io_ref;
423 struct completion io_ref_completion;
429 * Cached version of this device's member info from superblock
430 * Committed by bch2_write_super() -> bch_fs_mi_update()
432 struct bch_member_cpu mi;
434 char name[BDEVNAME_SIZE];
436 struct bch_sb_handle disk_sb;
437 struct bch_sb *sb_read_scratch;
441 struct bch_devs_mask self;
443 /* biosets used in cloned bios for writing multiple replicas */
444 struct bio_set replica_set;
448 * Per-bucket arrays are protected by c->mark_lock, bucket_lock and
449 * gc_lock, for device resize - holding any is sufficient for access:
450 * Or rcu_read_lock(), but only for ptr_stale():
452 struct bucket_array __rcu *buckets[2];
453 struct bucket_gens *bucket_gens;
454 unsigned long *buckets_nouse;
455 struct rw_semaphore bucket_lock;
457 struct bch_dev_usage *usage_base;
458 struct bch_dev_usage __percpu *usage[JOURNAL_BUF_NR];
459 struct bch_dev_usage __percpu *usage_gc;
462 u64 new_fs_bucket_idx;
463 struct task_struct __rcu *alloc_thread;
466 * free: Buckets that are ready to be used
468 * free_inc: Incoming buckets - these are buckets that currently have
469 * cached data in them, and we can't reuse them until after we write
470 * their new gen to disk. After prio_write() finishes writing the new
471 * gens/prios, they'll be moved to the free list (and possibly discarded
474 alloc_fifo free[RESERVE_NR];
476 unsigned nr_open_buckets;
478 open_bucket_idx_t open_buckets_partial[OPEN_BUCKETS_COUNT];
479 open_bucket_idx_t open_buckets_partial_nr;
481 size_t fifo_last_bucket;
483 size_t inc_gen_needs_gc;
484 size_t inc_gen_really_needs_gc;
486 enum allocator_states allocator_state;
488 alloc_heap alloc_heap;
490 atomic64_t rebalance_work;
492 struct journal_device journal;
493 u64 prev_journal_sector;
495 struct work_struct io_error_work;
497 /* The rest of this all shows up in sysfs */
498 atomic64_t cur_latency[2];
499 struct time_stats io_latency[2];
501 #define CONGESTED_MAX 1024
505 struct io_count __percpu *io_done;
511 BCH_FS_ALLOC_READ_DONE,
513 BCH_FS_ALLOCATOR_RUNNING,
514 BCH_FS_ALLOCATOR_STOPPING,
515 BCH_FS_INITIAL_GC_DONE,
516 BCH_FS_INITIAL_GC_UNFIXED,
517 BCH_FS_TOPOLOGY_REPAIR_DONE,
526 BCH_FS_WRITE_DISABLE_COMPLETE,
530 BCH_FS_TOPOLOGY_ERROR,
532 BCH_FS_ERRORS_NOT_FIXED,
535 BCH_FS_NEED_ANOTHER_GC,
536 BCH_FS_DELETED_NODES,
537 BCH_FS_REBUILD_REPLICAS,
538 BCH_FS_HOLD_BTREE_WRITES,
543 struct dentry *btree;
544 struct dentry *btree_format;
545 struct dentry *failed;
549 u64 sectors_available;
552 struct journal_seq_blacklist_table {
554 struct journal_seq_blacklist_table_entry {
561 struct journal_keys {
563 enum btree_id btree_id:8;
573 u64 journal_seq_base;
576 struct btree_path_buf {
577 struct btree_path *path;
580 #define REPLICAS_DELTA_LIST_MAX (1U << 16)
585 u32 subvol; /* Nonzero only if a subvolume points to this node: */
594 #define BCACHEFS_ROOT_SUBVOL_INUM \
595 ((subvol_inum) { BCACHEFS_ROOT_SUBVOL, BCACHEFS_ROOT_INO })
600 struct list_head list;
602 struct kobject internal;
603 struct kobject opts_dir;
604 struct kobject time_stats;
608 struct device *chardev;
609 struct super_block *vfs_sb;
613 /* ro/rw, add/remove/resize devices: */
614 struct rw_semaphore state_lock;
616 /* Counts outstanding writes, for clean transition to read-only */
617 struct percpu_ref writes;
618 struct work_struct read_only_work;
620 struct bch_dev __rcu *devs[BCH_SB_MEMBERS_MAX];
622 struct bch_replicas_cpu replicas;
623 struct bch_replicas_cpu replicas_gc;
624 struct mutex replicas_gc_lock;
625 mempool_t replicas_delta_pool;
627 struct journal_entry_res btree_root_journal_res;
628 struct journal_entry_res replicas_journal_res;
629 struct journal_entry_res clock_journal_res;
630 struct journal_entry_res dev_usage_journal_res;
632 struct bch_disk_groups_cpu __rcu *disk_groups;
634 struct bch_opts opts;
636 /* Updated by bch2_sb_update():*/
651 unsigned time_units_per_sec;
652 unsigned nsec_per_time_unit;
658 struct bch_sb_handle disk_sb;
660 unsigned short block_bits; /* ilog2(block_size) */
662 u16 btree_foreground_merge_threshold;
664 struct closure sb_write;
665 struct mutex sb_lock;
668 GENRADIX(struct snapshot_t) snapshots;
669 struct bch_snapshot_table __rcu *snapshot_table;
670 struct mutex snapshot_table_lock;
671 struct work_struct snapshot_delete_work;
672 struct work_struct snapshot_wait_for_pagecache_and_delete_work;
673 struct snapshot_id_list snapshots_unlinked;
674 struct mutex snapshots_unlinked_lock;
677 struct bio_set btree_bio;
678 struct workqueue_struct *io_complete_wq;
680 struct btree_root btree_roots[BTREE_ID_NR];
681 struct mutex btree_root_lock;
683 struct btree_cache btree_cache;
686 * Cache of allocated btree nodes - if we allocate a btree node and
687 * don't use it, if we free it that space can't be reused until going
688 * _all_ the way through the allocator (which exposes us to a livelock
689 * when allocating btree reserves fail halfway through) - instead, we
690 * can stick them here:
692 struct btree_alloc btree_reserve_cache[BTREE_NODE_RESERVE * 2];
693 unsigned btree_reserve_cache_nr;
694 struct mutex btree_reserve_cache_lock;
696 mempool_t btree_interior_update_pool;
697 struct list_head btree_interior_update_list;
698 struct list_head btree_interior_updates_unwritten;
699 struct mutex btree_interior_update_lock;
700 struct closure_waitlist btree_interior_update_wait;
702 struct workqueue_struct *btree_interior_update_worker;
703 struct work_struct btree_interior_update_work;
706 struct mutex btree_trans_lock;
707 struct list_head btree_trans_list;
708 mempool_t btree_paths_pool;
709 mempool_t btree_trans_mem_pool;
710 struct btree_path_buf __percpu *btree_paths_bufs;
712 struct srcu_struct btree_trans_barrier;
713 bool btree_trans_barrier_initialized;
715 struct btree_key_cache btree_key_cache;
717 struct workqueue_struct *btree_update_wq;
718 struct workqueue_struct *btree_io_complete_wq;
719 /* copygc needs its own workqueue for index updates.. */
720 struct workqueue_struct *copygc_wq;
723 struct bch_devs_mask rw_devs[BCH_DATA_NR];
725 u64 capacity; /* sectors */
728 * When capacity _decreases_ (due to a disk being removed), we
729 * increment capacity_gen - this invalidates outstanding reservations
730 * and forces them to be revalidated
733 unsigned bucket_size_max;
735 atomic64_t sectors_available;
736 struct mutex sectors_available_lock;
738 struct bch_fs_pcpu __percpu *pcpu;
740 struct percpu_rw_semaphore mark_lock;
742 seqcount_t usage_lock;
743 struct bch_fs_usage *usage_base;
744 struct bch_fs_usage __percpu *usage[JOURNAL_BUF_NR];
745 struct bch_fs_usage __percpu *usage_gc;
746 u64 __percpu *online_reserved;
748 /* single element mempool: */
749 struct mutex usage_scratch_lock;
750 struct bch_fs_usage_online *usage_scratch;
752 struct io_clock io_clock[2];
754 /* JOURNAL SEQ BLACKLIST */
755 struct journal_seq_blacklist_table *
756 journal_seq_blacklist_table;
757 struct work_struct journal_seq_blacklist_gc_work;
760 spinlock_t freelist_lock;
761 struct closure_waitlist freelist_wait;
762 u64 blocked_allocate;
763 u64 blocked_allocate_open_bucket;
765 open_bucket_idx_t open_buckets_freelist;
766 open_bucket_idx_t open_buckets_nr_free;
767 struct closure_waitlist open_buckets_wait;
768 struct open_bucket open_buckets[OPEN_BUCKETS_COUNT];
769 open_bucket_idx_t open_buckets_hash[OPEN_BUCKETS_COUNT];
771 struct write_point btree_write_point;
772 struct write_point rebalance_write_point;
774 struct write_point write_points[WRITE_POINT_MAX];
775 struct hlist_head write_points_hash[WRITE_POINT_HASH_NR];
776 struct mutex write_points_hash_lock;
777 unsigned write_points_nr;
779 /* GARBAGE COLLECTION */
780 struct task_struct *gc_thread;
782 unsigned long gc_count;
784 enum btree_id gc_gens_btree;
785 struct bpos gc_gens_pos;
788 * Tracks GC's progress - everything in the range [ZERO_KEY..gc_cur_pos]
789 * has been marked by GC.
791 * gc_cur_phase is a superset of btree_ids (BTREE_ID_extents etc.)
793 * Protected by gc_pos_lock. Only written to by GC thread, so GC thread
794 * can read without a lock.
796 seqcount_t gc_pos_lock;
797 struct gc_pos gc_pos;
800 * The allocation code needs gc_mark in struct bucket to be correct, but
801 * it's not while a gc is in progress.
803 struct rw_semaphore gc_lock;
806 struct semaphore io_in_flight;
807 struct bio_set bio_read;
808 struct bio_set bio_read_split;
809 struct bio_set bio_write;
810 struct mutex bio_bounce_pages_lock;
811 mempool_t bio_bounce_pages;
812 struct rhashtable promote_table;
814 mempool_t compression_bounce[2];
815 mempool_t compress_workspace[BCH_COMPRESSION_TYPE_NR];
816 mempool_t decompress_workspace;
817 ZSTD_parameters zstd_params;
819 struct crypto_shash *sha256;
820 struct crypto_sync_skcipher *chacha20;
821 struct crypto_shash *poly1305;
823 atomic64_t key_version;
825 mempool_t large_bkey_pool;
828 struct bch_fs_rebalance rebalance;
831 struct task_struct *copygc_thread;
832 copygc_heap copygc_heap;
833 struct write_point copygc_write_point;
836 /* DATA PROGRESS STATS */
837 struct list_head data_progress_list;
838 struct mutex data_progress_lock;
841 GENRADIX(struct stripe) stripes;
842 GENRADIX(struct gc_stripe) gc_stripes;
844 ec_stripes_heap ec_stripes_heap;
845 spinlock_t ec_stripes_heap_lock;
848 struct list_head ec_stripe_head_list;
849 struct mutex ec_stripe_head_lock;
851 struct list_head ec_stripe_new_list;
852 struct mutex ec_stripe_new_lock;
854 struct work_struct ec_stripe_create_work;
857 struct bio_set ec_bioset;
859 struct work_struct ec_stripe_delete_work;
860 struct llist_head ec_stripe_delete_list;
864 reflink_gc_table reflink_gc_table;
865 size_t reflink_gc_nr;
867 /* VFS IO PATH - fs-io.c */
868 struct bio_set writepage_bioset;
869 struct bio_set dio_write_bioset;
870 struct bio_set dio_read_bioset;
873 atomic64_t btree_writes_nr;
874 atomic64_t btree_writes_sectors;
875 spinlock_t btree_write_error_lock;
878 struct list_head fsck_errors;
879 struct mutex fsck_error_lock;
883 struct bch_memquota_type quotas[QTYP_NR];
886 struct dentry *debug;
887 struct btree_debug btree_debug[BTREE_ID_NR];
888 struct btree *verify_data;
889 struct btree_node *verify_ondisk;
890 struct mutex verify_lock;
892 u64 *unused_inode_hints;
893 unsigned inode_shard_bits;
896 * A btree node on disk could have too many bsets for an iterator to fit
897 * on the stack - have to dynamically allocate them
901 mempool_t btree_bounce_pool;
903 struct journal journal;
904 struct list_head journal_entries;
905 struct journal_keys journal_keys;
906 struct list_head journal_iters;
908 u64 last_bucket_seq_cleanup;
910 /* The rest of this all shows up in sysfs */
911 atomic_long_t read_realloc_races;
912 atomic_long_t extent_migrate_done;
913 atomic_long_t extent_migrate_raced;
915 unsigned btree_gc_periodic:1;
916 unsigned copy_gc_enabled:1;
917 bool promote_whole_extents;
919 struct time_stats times[BCH_TIME_STAT_NR];
922 static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
924 #ifndef NO_BCACHEFS_FS
926 c->vfs_sb->s_bdi->ra_pages = ra_pages;
930 static inline unsigned bucket_bytes(const struct bch_dev *ca)
932 return ca->mi.bucket_size << 9;
935 static inline unsigned block_bytes(const struct bch_fs *c)
937 return c->opts.block_size;
940 static inline unsigned block_sectors(const struct bch_fs *c)
942 return c->opts.block_size >> 9;
945 static inline size_t btree_sectors(const struct bch_fs *c)
947 return c->opts.btree_node_size >> 9;
950 static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
955 time += c->sb.time_base_lo;
957 t.tv_sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
958 t.tv_nsec = rem * c->sb.nsec_per_time_unit;
962 static inline s64 timespec_to_bch2_time(const struct bch_fs *c, struct timespec64 ts)
964 return (ts.tv_sec * c->sb.time_units_per_sec +
965 (int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
968 static inline s64 bch2_current_time(const struct bch_fs *c)
970 struct timespec64 now;
972 ktime_get_coarse_real_ts64(&now);
973 return timespec_to_bch2_time(c, now);
976 static inline bool bch2_dev_exists2(const struct bch_fs *c, unsigned dev)
978 return dev < c->sb.nr_devices && c->devs[dev];
981 #endif /* _BCACHEFS_H */