1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _BCACHEFS_H
3 #define _BCACHEFS_H
4
5 /*
6 * SOME HIGH LEVEL CODE DOCUMENTATION:
7 *
8 * Bcache mostly works with cache sets, cache devices, and backing devices.
9 *
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.
14 *
15 * Multiple cache devices is intended to give us the ability to mirror dirty
16 * cached data and metadata, without mirroring clean cached data.
17 *
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.
23 *
24 * A cache set can have multiple (many) backing devices attached to it.
25 *
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.
31 *
32 * Flash only volumes work but they're not production ready because the moving
33 * garbage collector needs more work. More on that later.
34 *
35 * BUCKETS/ALLOCATION:
36 *
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
40 * it.
41 *
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+
44 * works efficiently.
45 *
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).
50 *
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.
55 *
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
60 * this up).
61 *
62 * Bcache is entirely COW - we never write twice to a bucket, even buckets that
63 * contain metadata (including btree nodes).
64 *
65 * THE BTREE:
66 *
67 * Bcache is in large part design around the btree.
68 *
69 * At a high level, the btree is just an index of key -> ptr tuples.
70 *
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).
74 *
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.
79 *
80 * Pointers contain the cache device id, the offset on that device, and an 8 bit
81 * generation number. More on the gen later.
82 *
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
85 * direction.
86 *
87 * Updates are fairly well abstracted, though. There are two different ways of
88 * updating the btree; insert and replace.
89 *
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.
93 *
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.
98 *
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.
103 *
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.
107 *
108 * BTREE NODES:
109 *
110 * Our unit of allocation is a bucket, and we can't arbitrarily allocate and
111 * free smaller than a bucket - so, that's how big our btree nodes are.
112 *
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.)
117 *
118 * Anyways, btree nodes are big - big enough to be inefficient with a textbook
119 * btree implementation.
120 *
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.
124 *
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
131 * smaller).
132 *
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.
136 *
137 * GARBAGE COLLECTION:
138 *
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.
142 *
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.
146 *
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.
151 *
152 * THE JOURNAL:
153 *
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
158 * implemented.
159 *
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.
167 *
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
171 * writing them out.
172 *
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()).
177 */
178
179 #undef pr_fmt
180 #ifdef __KERNEL__
181 #define pr_fmt(fmt) "bcachefs: %s() " fmt "\n", __func__
182 #else
183 #define pr_fmt(fmt) "%s() " fmt "\n", __func__
184 #endif
185
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/refcount.h>
197 #include <linux/rhashtable.h>
198 #include <linux/rwsem.h>
199 #include <linux/semaphore.h>
200 #include <linux/seqlock.h>
201 #include <linux/shrinker.h>
202 #include <linux/srcu.h>
203 #include <linux/types.h>
204 #include <linux/workqueue.h>
205 #include <linux/zstd.h>
206
207 #include "bcachefs_format.h"
208 #include "btree_journal_iter_types.h"
209 #include "disk_accounting_types.h"
210 #include "errcode.h"
211 #include "fifo.h"
212 #include "nocow_locking_types.h"
213 #include "opts.h"
214 #include "recovery_passes_types.h"
215 #include "sb-errors_types.h"
216 #include "seqmutex.h"
217 #include "time_stats.h"
218 #include "util.h"
219
220 #ifdef CONFIG_BCACHEFS_DEBUG
221 #define BCH_WRITE_REF_DEBUG
222 #endif
223
224 #ifndef dynamic_fault
225 #define dynamic_fault(...) 0
226 #endif
227
228 #define race_fault(...) dynamic_fault("bcachefs:race")
229
230 #define count_event(_c, _name) this_cpu_inc((_c)->counters[BCH_COUNTER_##_name])
231
232 #define trace_and_count(_c, _name, ...) \
233 do { \
234 count_event(_c, _name); \
235 trace_##_name(__VA_ARGS__); \
236 } while (0)
237
238 #define bch2_fs_init_fault(name) \
239 dynamic_fault("bcachefs:bch_fs_init:" name)
240 #define bch2_meta_read_fault(name) \
241 dynamic_fault("bcachefs:meta:read:" name)
242 #define bch2_meta_write_fault(name) \
243 dynamic_fault("bcachefs:meta:write:" name)
244
245 #ifdef __KERNEL__
246 #define BCACHEFS_LOG_PREFIX
247 #endif
248
249 #ifdef BCACHEFS_LOG_PREFIX
250
251 #define bch2_log_msg(_c, fmt) "bcachefs (%s): " fmt, ((_c)->name)
252 #define bch2_fmt_dev(_ca, fmt) "bcachefs (%s): " fmt "\n", ((_ca)->name)
253 #define bch2_fmt_dev_offset(_ca, _offset, fmt) "bcachefs (%s sector %llu): " fmt "\n", ((_ca)->name), (_offset)
254 #define bch2_fmt_inum(_c, _inum, fmt) "bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
255 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt) \
256 "bcachefs (%s inum %llu offset %llu): " fmt "\n", ((_c)->name), (_inum), (_offset)
257
258 #else
259
260 #define bch2_log_msg(_c, fmt) fmt
261 #define bch2_fmt_dev(_ca, fmt) "%s: " fmt "\n", ((_ca)->name)
262 #define bch2_fmt_dev_offset(_ca, _offset, fmt) "%s sector %llu: " fmt "\n", ((_ca)->name), (_offset)
263 #define bch2_fmt_inum(_c, _inum, fmt) "inum %llu: " fmt "\n", (_inum)
264 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt) \
265 "inum %llu offset %llu: " fmt "\n", (_inum), (_offset)
266
267 #endif
268
269 #define bch2_fmt(_c, fmt) bch2_log_msg(_c, fmt "\n")
270
271 void bch2_print_str(struct bch_fs *, const char *);
272
273 __printf(2, 3)
274 void bch2_print_opts(struct bch_opts *, const char *, ...);
275
276 __printf(2, 3)
277 void __bch2_print(struct bch_fs *c, const char *fmt, ...);
278
279 #define maybe_dev_to_fs(_c) _Generic((_c), \
280 struct bch_dev *: ((struct bch_dev *) (_c))->fs, \
281 struct bch_fs *: (_c))
282
283 #define bch2_print(_c, ...) __bch2_print(maybe_dev_to_fs(_c), __VA_ARGS__)
284
285 #define bch2_print_ratelimited(_c, ...) \
286 do { \
287 static DEFINE_RATELIMIT_STATE(_rs, \
288 DEFAULT_RATELIMIT_INTERVAL, \
289 DEFAULT_RATELIMIT_BURST); \
290 \
291 if (__ratelimit(&_rs)) \
292 bch2_print(_c, __VA_ARGS__); \
293 } while (0)
294
295 #define bch_info(c, fmt, ...) \
296 bch2_print(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
297 #define bch_info_ratelimited(c, fmt, ...) \
298 bch2_print_ratelimited(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
299 #define bch_notice(c, fmt, ...) \
300 bch2_print(c, KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
301 #define bch_warn(c, fmt, ...) \
302 bch2_print(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
303 #define bch_warn_ratelimited(c, fmt, ...) \
304 bch2_print_ratelimited(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
305
306 #define bch_err(c, fmt, ...) \
307 bch2_print(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
308 #define bch_err_dev(ca, fmt, ...) \
309 bch2_print(c, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
310 #define bch_err_dev_offset(ca, _offset, fmt, ...) \
311 bch2_print(c, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
312 #define bch_err_inum(c, _inum, fmt, ...) \
313 bch2_print(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
314 #define bch_err_inum_offset(c, _inum, _offset, fmt, ...) \
315 bch2_print(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
316
317 #define bch_err_ratelimited(c, fmt, ...) \
318 bch2_print_ratelimited(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
319 #define bch_err_dev_ratelimited(ca, fmt, ...) \
320 bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
321 #define bch_err_dev_offset_ratelimited(ca, _offset, fmt, ...) \
322 bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
323 #define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
324 bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
325 #define bch_err_inum_offset_ratelimited(c, _inum, _offset, fmt, ...) \
326 bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
327
should_print_err(int err)328 static inline bool should_print_err(int err)
329 {
330 return err && !bch2_err_matches(err, BCH_ERR_transaction_restart);
331 }
332
333 #define bch_err_fn(_c, _ret) \
334 do { \
335 if (should_print_err(_ret)) \
336 bch_err(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
337 } while (0)
338
339 #define bch_err_fn_ratelimited(_c, _ret) \
340 do { \
341 if (should_print_err(_ret)) \
342 bch_err_ratelimited(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
343 } while (0)
344
345 #define bch_err_msg(_c, _ret, _msg, ...) \
346 do { \
347 if (should_print_err(_ret)) \
348 bch_err(_c, "%s(): error " _msg " %s", __func__, \
349 ##__VA_ARGS__, bch2_err_str(_ret)); \
350 } while (0)
351
352 #define bch_verbose(c, fmt, ...) \
353 do { \
354 if ((c)->opts.verbose) \
355 bch_info(c, fmt, ##__VA_ARGS__); \
356 } while (0)
357
358 #define bch_verbose_ratelimited(c, fmt, ...) \
359 do { \
360 if ((c)->opts.verbose) \
361 bch_info_ratelimited(c, fmt, ##__VA_ARGS__); \
362 } while (0)
363
364 #define pr_verbose_init(opts, fmt, ...) \
365 do { \
366 if (opt_get(opts, verbose)) \
367 pr_info(fmt, ##__VA_ARGS__); \
368 } while (0)
369
370 /* Parameters that are useful for debugging, but should always be compiled in: */
371 #define BCH_DEBUG_PARAMS_ALWAYS() \
372 BCH_DEBUG_PARAM(key_merging_disabled, \
373 "Disables merging of extents") \
374 BCH_DEBUG_PARAM(btree_node_merging_disabled, \
375 "Disables merging of btree nodes") \
376 BCH_DEBUG_PARAM(btree_gc_always_rewrite, \
377 "Causes mark and sweep to compact and rewrite every " \
378 "btree node it traverses") \
379 BCH_DEBUG_PARAM(btree_gc_rewrite_disabled, \
380 "Disables rewriting of btree nodes during mark and sweep")\
381 BCH_DEBUG_PARAM(btree_shrinker_disabled, \
382 "Disables the shrinker callback for the btree node cache")\
383 BCH_DEBUG_PARAM(verify_btree_ondisk, \
384 "Reread btree nodes at various points to verify the " \
385 "mergesort in the read path against modifications " \
386 "done in memory") \
387 BCH_DEBUG_PARAM(verify_all_btree_replicas, \
388 "When reading btree nodes, read all replicas and " \
389 "compare them") \
390 BCH_DEBUG_PARAM(backpointers_no_use_write_buffer, \
391 "Don't use the write buffer for backpointers, enabling "\
392 "extra runtime checks")
393
394 /* Parameters that should only be compiled in debug mode: */
395 #define BCH_DEBUG_PARAMS_DEBUG() \
396 BCH_DEBUG_PARAM(expensive_debug_checks, \
397 "Enables various runtime debugging checks that " \
398 "significantly affect performance") \
399 BCH_DEBUG_PARAM(debug_check_iterators, \
400 "Enables extra verification for btree iterators") \
401 BCH_DEBUG_PARAM(debug_check_btree_accounting, \
402 "Verify btree accounting for keys within a node") \
403 BCH_DEBUG_PARAM(journal_seq_verify, \
404 "Store the journal sequence number in the version " \
405 "number of every btree key, and verify that btree " \
406 "update ordering is preserved during recovery") \
407 BCH_DEBUG_PARAM(inject_invalid_keys, \
408 "Store the journal sequence number in the version " \
409 "number of every btree key, and verify that btree " \
410 "update ordering is preserved during recovery") \
411 BCH_DEBUG_PARAM(test_alloc_startup, \
412 "Force allocator startup to use the slowpath where it" \
413 "can't find enough free buckets without invalidating" \
414 "cached data") \
415 BCH_DEBUG_PARAM(force_reconstruct_read, \
416 "Force reads to use the reconstruct path, when reading" \
417 "from erasure coded extents") \
418 BCH_DEBUG_PARAM(test_restart_gc, \
419 "Test restarting mark and sweep gc when bucket gens change")
420
421 #define BCH_DEBUG_PARAMS_ALL() BCH_DEBUG_PARAMS_ALWAYS() BCH_DEBUG_PARAMS_DEBUG()
422
423 #ifdef CONFIG_BCACHEFS_DEBUG
424 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALL()
425 #else
426 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALWAYS()
427 #endif
428
429 #define BCH_DEBUG_PARAM(name, description) extern bool bch2_##name;
430 BCH_DEBUG_PARAMS()
431 #undef BCH_DEBUG_PARAM
432
433 #ifndef CONFIG_BCACHEFS_DEBUG
434 #define BCH_DEBUG_PARAM(name, description) static const __maybe_unused bool bch2_##name;
435 BCH_DEBUG_PARAMS_DEBUG()
436 #undef BCH_DEBUG_PARAM
437 #endif
438
439 #define BCH_TIME_STATS() \
440 x(btree_node_mem_alloc) \
441 x(btree_node_split) \
442 x(btree_node_compact) \
443 x(btree_node_merge) \
444 x(btree_node_sort) \
445 x(btree_node_read) \
446 x(btree_node_read_done) \
447 x(btree_interior_update_foreground) \
448 x(btree_interior_update_total) \
449 x(btree_gc) \
450 x(data_write) \
451 x(data_read) \
452 x(data_promote) \
453 x(journal_flush_write) \
454 x(journal_noflush_write) \
455 x(journal_flush_seq) \
456 x(blocked_journal_low_on_space) \
457 x(blocked_journal_low_on_pin) \
458 x(blocked_journal_max_in_flight) \
459 x(blocked_key_cache_flush) \
460 x(blocked_allocate) \
461 x(blocked_allocate_open_bucket) \
462 x(blocked_write_buffer_full) \
463 x(nocow_lock_contended)
464
465 enum bch_time_stats {
466 #define x(name) BCH_TIME_##name,
467 BCH_TIME_STATS()
468 #undef x
469 BCH_TIME_STAT_NR
470 };
471
472 #include "alloc_types.h"
473 #include "btree_gc_types.h"
474 #include "btree_types.h"
475 #include "btree_node_scan_types.h"
476 #include "btree_write_buffer_types.h"
477 #include "buckets_types.h"
478 #include "buckets_waiting_for_journal_types.h"
479 #include "clock_types.h"
480 #include "disk_groups_types.h"
481 #include "ec_types.h"
482 #include "journal_types.h"
483 #include "keylist_types.h"
484 #include "quota_types.h"
485 #include "rebalance_types.h"
486 #include "replicas_types.h"
487 #include "sb-members_types.h"
488 #include "subvolume_types.h"
489 #include "super_types.h"
490 #include "thread_with_file_types.h"
491
492 /* Number of nodes btree coalesce will try to coalesce at once */
493 #define GC_MERGE_NODES 4U
494
495 /* Maximum number of nodes we might need to allocate atomically: */
496 #define BTREE_RESERVE_MAX (BTREE_MAX_DEPTH + (BTREE_MAX_DEPTH - 1))
497
498 /* Size of the freelist we allocate btree nodes from: */
499 #define BTREE_NODE_RESERVE (BTREE_RESERVE_MAX * 4)
500
501 #define BTREE_NODE_OPEN_BUCKET_RESERVE (BTREE_RESERVE_MAX * BCH_REPLICAS_MAX)
502
503 struct btree;
504
505 struct io_count {
506 u64 sectors[2][BCH_DATA_NR];
507 };
508
509 struct discard_in_flight {
510 bool in_progress:1;
511 u64 bucket:63;
512 };
513
514 struct bch_dev {
515 struct kobject kobj;
516 #ifdef CONFIG_BCACHEFS_DEBUG
517 atomic_long_t ref;
518 bool dying;
519 unsigned long last_put;
520 #else
521 struct percpu_ref ref;
522 #endif
523 struct completion ref_completion;
524 struct percpu_ref io_ref;
525 struct completion io_ref_completion;
526
527 struct bch_fs *fs;
528
529 u8 dev_idx;
530 /*
531 * Cached version of this device's member info from superblock
532 * Committed by bch2_write_super() -> bch_fs_mi_update()
533 */
534 struct bch_member_cpu mi;
535 atomic64_t errors[BCH_MEMBER_ERROR_NR];
536
537 __uuid_t uuid;
538 char name[BDEVNAME_SIZE];
539
540 struct bch_sb_handle disk_sb;
541 struct bch_sb *sb_read_scratch;
542 int sb_write_error;
543 dev_t dev;
544 atomic_t flush_seq;
545
546 struct bch_devs_mask self;
547
548 /*
549 * Buckets:
550 * Per-bucket arrays are protected by either rcu_read_lock or
551 * state_lock, for device resize.
552 */
553 GENRADIX(struct bucket) buckets_gc;
554 struct bucket_gens __rcu *bucket_gens;
555 u8 *oldest_gen;
556 unsigned long *buckets_nouse;
557
558 unsigned long *bucket_backpointer_mismatches;
559 unsigned long *bucket_backpointer_empty;
560
561 struct bch_dev_usage __percpu *usage;
562
563 /* Allocator: */
564 u64 alloc_cursor[3];
565
566 unsigned nr_open_buckets;
567 unsigned nr_partial_buckets;
568 unsigned nr_btree_reserve;
569
570 size_t inc_gen_needs_gc;
571 size_t inc_gen_really_needs_gc;
572 size_t buckets_waiting_on_journal;
573
574 struct work_struct invalidate_work;
575 struct work_struct discard_work;
576 struct mutex discard_buckets_in_flight_lock;
577 DARRAY(struct discard_in_flight) discard_buckets_in_flight;
578 struct work_struct discard_fast_work;
579
580 atomic64_t rebalance_work;
581
582 struct journal_device journal;
583 u64 prev_journal_sector;
584
585 struct work_struct io_error_work;
586
587 /* The rest of this all shows up in sysfs */
588 atomic64_t cur_latency[2];
589 struct bch2_time_stats_quantiles io_latency[2];
590
591 #define CONGESTED_MAX 1024
592 atomic_t congested;
593 u64 congested_last;
594
595 struct io_count __percpu *io_done;
596 };
597
598 /*
599 * initial_gc_unfixed
600 * error
601 * topology error
602 */
603
604 #define BCH_FS_FLAGS() \
605 x(new_fs) \
606 x(started) \
607 x(clean_recovery) \
608 x(btree_running) \
609 x(accounting_replay_done) \
610 x(may_go_rw) \
611 x(rw) \
612 x(was_rw) \
613 x(stopping) \
614 x(emergency_ro) \
615 x(going_ro) \
616 x(write_disable_complete) \
617 x(clean_shutdown) \
618 x(recovery_running) \
619 x(fsck_running) \
620 x(initial_gc_unfixed) \
621 x(need_delete_dead_snapshots) \
622 x(error) \
623 x(topology_error) \
624 x(errors_fixed) \
625 x(errors_not_fixed) \
626 x(no_invalid_checks)
627
628 enum bch_fs_flags {
629 #define x(n) BCH_FS_##n,
630 BCH_FS_FLAGS()
631 #undef x
632 };
633
634 struct btree_debug {
635 unsigned id;
636 };
637
638 #define BCH_TRANSACTIONS_NR 128
639
640 struct btree_transaction_stats {
641 struct bch2_time_stats duration;
642 struct bch2_time_stats lock_hold_times;
643 struct mutex lock;
644 unsigned nr_max_paths;
645 unsigned journal_entries_size;
646 unsigned max_mem;
647 char *max_paths_text;
648 };
649
650 struct bch_fs_pcpu {
651 u64 sectors_available;
652 };
653
654 struct journal_seq_blacklist_table {
655 size_t nr;
656 struct journal_seq_blacklist_table_entry {
657 u64 start;
658 u64 end;
659 bool dirty;
660 } entries[];
661 };
662
663 struct btree_trans_buf {
664 struct btree_trans *trans;
665 };
666
667 #define BCACHEFS_ROOT_SUBVOL_INUM \
668 ((subvol_inum) { BCACHEFS_ROOT_SUBVOL, BCACHEFS_ROOT_INO })
669
670 #define BCH_WRITE_REFS() \
671 x(journal) \
672 x(trans) \
673 x(write) \
674 x(promote) \
675 x(node_rewrite) \
676 x(stripe_create) \
677 x(stripe_delete) \
678 x(reflink) \
679 x(fallocate) \
680 x(fsync) \
681 x(dio_write) \
682 x(discard) \
683 x(discard_fast) \
684 x(check_discard_freespace_key) \
685 x(invalidate) \
686 x(delete_dead_snapshots) \
687 x(gc_gens) \
688 x(snapshot_delete_pagecache) \
689 x(sysfs) \
690 x(btree_write_buffer)
691
692 enum bch_write_ref {
693 #define x(n) BCH_WRITE_REF_##n,
694 BCH_WRITE_REFS()
695 #undef x
696 BCH_WRITE_REF_NR,
697 };
698
699 struct bch_fs {
700 struct closure cl;
701
702 struct list_head list;
703 struct kobject kobj;
704 struct kobject counters_kobj;
705 struct kobject internal;
706 struct kobject opts_dir;
707 struct kobject time_stats;
708 unsigned long flags;
709
710 int minor;
711 struct device *chardev;
712 struct super_block *vfs_sb;
713 dev_t dev;
714 char name[40];
715 struct stdio_redirect *stdio;
716 struct task_struct *stdio_filter;
717
718 /* ro/rw, add/remove/resize devices: */
719 struct rw_semaphore state_lock;
720
721 /* Counts outstanding writes, for clean transition to read-only */
722 #ifdef BCH_WRITE_REF_DEBUG
723 atomic_long_t writes[BCH_WRITE_REF_NR];
724 #else
725 struct percpu_ref writes;
726 #endif
727 /*
728 * Certain operations are only allowed in single threaded mode, during
729 * recovery, and we want to assert that this is the case:
730 */
731 struct task_struct *recovery_task;
732
733 /*
734 * Analagous to c->writes, for asynchronous ops that don't necessarily
735 * need fs to be read-write
736 */
737 refcount_t ro_ref;
738 wait_queue_head_t ro_ref_wait;
739
740 struct work_struct read_only_work;
741
742 struct bch_dev __rcu *devs[BCH_SB_MEMBERS_MAX];
743
744 struct bch_accounting_mem accounting;
745
746 struct bch_replicas_cpu replicas;
747 struct bch_replicas_cpu replicas_gc;
748 struct mutex replicas_gc_lock;
749
750 struct journal_entry_res btree_root_journal_res;
751 struct journal_entry_res clock_journal_res;
752
753 struct bch_disk_groups_cpu __rcu *disk_groups;
754
755 struct bch_opts opts;
756
757 /* Updated by bch2_sb_update():*/
758 struct {
759 __uuid_t uuid;
760 __uuid_t user_uuid;
761
762 u16 version;
763 u16 version_incompat;
764 u16 version_incompat_allowed;
765 u16 version_min;
766 u16 version_upgrade_complete;
767
768 u8 nr_devices;
769 u8 clean;
770
771 u8 encryption_type;
772
773 u64 time_base_lo;
774 u32 time_base_hi;
775 unsigned time_units_per_sec;
776 unsigned nsec_per_time_unit;
777 u64 features;
778 u64 compat;
779 unsigned long errors_silent[BITS_TO_LONGS(BCH_FSCK_ERR_MAX)];
780 u64 btrees_lost_data;
781 } sb;
782
783
784 struct bch_sb_handle disk_sb;
785
786 unsigned short block_bits; /* ilog2(block_size) */
787
788 u16 btree_foreground_merge_threshold;
789
790 struct closure sb_write;
791 struct mutex sb_lock;
792
793 /* snapshot.c: */
794 struct snapshot_table __rcu *snapshots;
795 struct mutex snapshot_table_lock;
796 struct rw_semaphore snapshot_create_lock;
797
798 struct work_struct snapshot_delete_work;
799 struct work_struct snapshot_wait_for_pagecache_and_delete_work;
800 snapshot_id_list snapshots_unlinked;
801 struct mutex snapshots_unlinked_lock;
802
803 /* BTREE CACHE */
804 struct bio_set btree_bio;
805 struct workqueue_struct *btree_read_complete_wq;
806 struct workqueue_struct *btree_write_submit_wq;
807
808 struct btree_root btree_roots_known[BTREE_ID_NR];
809 DARRAY(struct btree_root) btree_roots_extra;
810 struct mutex btree_root_lock;
811
812 struct btree_cache btree_cache;
813
814 /*
815 * Cache of allocated btree nodes - if we allocate a btree node and
816 * don't use it, if we free it that space can't be reused until going
817 * _all_ the way through the allocator (which exposes us to a livelock
818 * when allocating btree reserves fail halfway through) - instead, we
819 * can stick them here:
820 */
821 struct btree_alloc btree_reserve_cache[BTREE_NODE_RESERVE * 2];
822 unsigned btree_reserve_cache_nr;
823 struct mutex btree_reserve_cache_lock;
824
825 mempool_t btree_interior_update_pool;
826 struct list_head btree_interior_update_list;
827 struct list_head btree_interior_updates_unwritten;
828 struct mutex btree_interior_update_lock;
829 struct closure_waitlist btree_interior_update_wait;
830
831 struct workqueue_struct *btree_interior_update_worker;
832 struct work_struct btree_interior_update_work;
833
834 struct workqueue_struct *btree_node_rewrite_worker;
835 struct list_head btree_node_rewrites;
836 struct list_head btree_node_rewrites_pending;
837 spinlock_t btree_node_rewrites_lock;
838 struct closure_waitlist btree_node_rewrites_wait;
839
840 /* btree_io.c: */
841 spinlock_t btree_write_error_lock;
842 struct btree_write_stats {
843 atomic64_t nr;
844 atomic64_t bytes;
845 } btree_write_stats[BTREE_WRITE_TYPE_NR];
846
847 /* btree_iter.c: */
848 struct seqmutex btree_trans_lock;
849 struct list_head btree_trans_list;
850 mempool_t btree_trans_pool;
851 mempool_t btree_trans_mem_pool;
852 struct btree_trans_buf __percpu *btree_trans_bufs;
853
854 struct srcu_struct btree_trans_barrier;
855 bool btree_trans_barrier_initialized;
856
857 struct btree_key_cache btree_key_cache;
858 unsigned btree_key_cache_btrees;
859
860 struct btree_write_buffer btree_write_buffer;
861
862 struct workqueue_struct *btree_update_wq;
863 struct workqueue_struct *btree_io_complete_wq;
864 /* copygc needs its own workqueue for index updates.. */
865 struct workqueue_struct *copygc_wq;
866 /*
867 * Use a dedicated wq for write ref holder tasks. Required to avoid
868 * dependency problems with other wq tasks that can block on ref
869 * draining, such as read-only transition.
870 */
871 struct workqueue_struct *write_ref_wq;
872
873 /* ALLOCATION */
874 struct bch_devs_mask rw_devs[BCH_DATA_NR];
875 unsigned long rw_devs_change_count;
876
877 u64 capacity; /* sectors */
878 u64 reserved; /* sectors */
879
880 /*
881 * When capacity _decreases_ (due to a disk being removed), we
882 * increment capacity_gen - this invalidates outstanding reservations
883 * and forces them to be revalidated
884 */
885 u32 capacity_gen;
886 unsigned bucket_size_max;
887
888 atomic64_t sectors_available;
889 struct mutex sectors_available_lock;
890
891 struct bch_fs_pcpu __percpu *pcpu;
892
893 struct percpu_rw_semaphore mark_lock;
894
895 seqcount_t usage_lock;
896 struct bch_fs_usage_base __percpu *usage;
897 u64 __percpu *online_reserved;
898
899 unsigned long allocator_last_stuck;
900
901 struct io_clock io_clock[2];
902
903 /* JOURNAL SEQ BLACKLIST */
904 struct journal_seq_blacklist_table *
905 journal_seq_blacklist_table;
906
907 /* ALLOCATOR */
908 spinlock_t freelist_lock;
909 struct closure_waitlist freelist_wait;
910
911 open_bucket_idx_t open_buckets_freelist;
912 open_bucket_idx_t open_buckets_nr_free;
913 struct closure_waitlist open_buckets_wait;
914 struct open_bucket open_buckets[OPEN_BUCKETS_COUNT];
915 open_bucket_idx_t open_buckets_hash[OPEN_BUCKETS_COUNT];
916
917 open_bucket_idx_t open_buckets_partial[OPEN_BUCKETS_COUNT];
918 open_bucket_idx_t open_buckets_partial_nr;
919
920 struct write_point btree_write_point;
921 struct write_point rebalance_write_point;
922
923 struct write_point write_points[WRITE_POINT_MAX];
924 struct hlist_head write_points_hash[WRITE_POINT_HASH_NR];
925 struct mutex write_points_hash_lock;
926 unsigned write_points_nr;
927
928 struct buckets_waiting_for_journal buckets_waiting_for_journal;
929
930 /* GARBAGE COLLECTION */
931 struct work_struct gc_gens_work;
932 unsigned long gc_count;
933
934 enum btree_id gc_gens_btree;
935 struct bpos gc_gens_pos;
936
937 /*
938 * Tracks GC's progress - everything in the range [ZERO_KEY..gc_cur_pos]
939 * has been marked by GC.
940 *
941 * gc_cur_phase is a superset of btree_ids (BTREE_ID_extents etc.)
942 *
943 * Protected by gc_pos_lock. Only written to by GC thread, so GC thread
944 * can read without a lock.
945 */
946 seqcount_t gc_pos_lock;
947 struct gc_pos gc_pos;
948
949 /*
950 * The allocation code needs gc_mark in struct bucket to be correct, but
951 * it's not while a gc is in progress.
952 */
953 struct rw_semaphore gc_lock;
954 struct mutex gc_gens_lock;
955
956 /* IO PATH */
957 struct semaphore io_in_flight;
958 struct bio_set bio_read;
959 struct bio_set bio_read_split;
960 struct bio_set bio_write;
961 struct bio_set replica_set;
962 struct mutex bio_bounce_pages_lock;
963 mempool_t bio_bounce_pages;
964 struct bucket_nocow_lock_table
965 nocow_locks;
966 struct rhashtable promote_table;
967
968 mempool_t compression_bounce[2];
969 mempool_t compress_workspace[BCH_COMPRESSION_OPT_NR];
970 size_t zstd_workspace_size;
971
972 struct crypto_shash *sha256;
973 struct crypto_sync_skcipher *chacha20;
974 struct crypto_shash *poly1305;
975
976 atomic64_t key_version;
977
978 mempool_t large_bkey_pool;
979
980 /* MOVE.C */
981 struct list_head moving_context_list;
982 struct mutex moving_context_lock;
983
984 /* REBALANCE */
985 struct bch_fs_rebalance rebalance;
986
987 /* COPYGC */
988 struct task_struct *copygc_thread;
989 struct write_point copygc_write_point;
990 s64 copygc_wait_at;
991 s64 copygc_wait;
992 bool copygc_running;
993 wait_queue_head_t copygc_running_wq;
994
995 /* STRIPES: */
996 GENRADIX(struct stripe) stripes;
997 GENRADIX(struct gc_stripe) gc_stripes;
998
999 struct hlist_head ec_stripes_new[32];
1000 spinlock_t ec_stripes_new_lock;
1001
1002 ec_stripes_heap ec_stripes_heap;
1003 struct mutex ec_stripes_heap_lock;
1004
1005 /* ERASURE CODING */
1006 struct list_head ec_stripe_head_list;
1007 struct mutex ec_stripe_head_lock;
1008
1009 struct list_head ec_stripe_new_list;
1010 struct mutex ec_stripe_new_lock;
1011 wait_queue_head_t ec_stripe_new_wait;
1012
1013 struct work_struct ec_stripe_create_work;
1014 u64 ec_stripe_hint;
1015
1016 struct work_struct ec_stripe_delete_work;
1017
1018 struct bio_set ec_bioset;
1019
1020 /* REFLINK */
1021 reflink_gc_table reflink_gc_table;
1022 size_t reflink_gc_nr;
1023
1024 /* fs.c */
1025 struct list_head vfs_inodes_list;
1026 struct mutex vfs_inodes_lock;
1027 struct rhashtable vfs_inodes_table;
1028 struct rhltable vfs_inodes_by_inum_table;
1029
1030 /* VFS IO PATH - fs-io.c */
1031 struct bio_set writepage_bioset;
1032 struct bio_set dio_write_bioset;
1033 struct bio_set dio_read_bioset;
1034 struct bio_set nocow_flush_bioset;
1035
1036 /* QUOTAS */
1037 struct bch_memquota_type quotas[QTYP_NR];
1038
1039 /* RECOVERY */
1040 u64 journal_replay_seq_start;
1041 u64 journal_replay_seq_end;
1042 /*
1043 * Two different uses:
1044 * "Has this fsck pass?" - i.e. should this type of error be an
1045 * emergency read-only
1046 * And, in certain situations fsck will rewind to an earlier pass: used
1047 * for signaling to the toplevel code which pass we want to run now.
1048 */
1049 enum bch_recovery_pass curr_recovery_pass;
1050 enum bch_recovery_pass next_recovery_pass;
1051 /* bitmask of recovery passes that we actually ran */
1052 u64 recovery_passes_complete;
1053 /* never rewinds version of curr_recovery_pass */
1054 enum bch_recovery_pass recovery_pass_done;
1055 spinlock_t recovery_pass_lock;
1056 struct semaphore online_fsck_mutex;
1057
1058 /* DEBUG JUNK */
1059 struct dentry *fs_debug_dir;
1060 struct dentry *btree_debug_dir;
1061 struct btree_debug btree_debug[BTREE_ID_NR];
1062 struct btree *verify_data;
1063 struct btree_node *verify_ondisk;
1064 struct mutex verify_lock;
1065
1066 /*
1067 * A btree node on disk could have too many bsets for an iterator to fit
1068 * on the stack - have to dynamically allocate them
1069 */
1070 mempool_t fill_iter;
1071
1072 mempool_t btree_bounce_pool;
1073
1074 struct journal journal;
1075 GENRADIX(struct journal_replay *) journal_entries;
1076 u64 journal_entries_base_seq;
1077 struct journal_keys journal_keys;
1078 struct list_head journal_iters;
1079
1080 struct find_btree_nodes found_btree_nodes;
1081
1082 u64 last_bucket_seq_cleanup;
1083
1084 u64 counters_on_mount[BCH_COUNTER_NR];
1085 u64 __percpu *counters;
1086
1087 struct bch2_time_stats times[BCH_TIME_STAT_NR];
1088
1089 struct btree_transaction_stats btree_transaction_stats[BCH_TRANSACTIONS_NR];
1090
1091 /* ERRORS */
1092 struct list_head fsck_error_msgs;
1093 struct mutex fsck_error_msgs_lock;
1094 bool fsck_alloc_msgs_err;
1095
1096 bch_sb_errors_cpu fsck_error_counts;
1097 struct mutex fsck_error_counts_lock;
1098 };
1099
1100 extern struct wait_queue_head bch2_read_only_wait;
1101
bch2_write_ref_get(struct bch_fs * c,enum bch_write_ref ref)1102 static inline void bch2_write_ref_get(struct bch_fs *c, enum bch_write_ref ref)
1103 {
1104 #ifdef BCH_WRITE_REF_DEBUG
1105 atomic_long_inc(&c->writes[ref]);
1106 #else
1107 percpu_ref_get(&c->writes);
1108 #endif
1109 }
1110
__bch2_write_ref_tryget(struct bch_fs * c,enum bch_write_ref ref)1111 static inline bool __bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1112 {
1113 #ifdef BCH_WRITE_REF_DEBUG
1114 return !test_bit(BCH_FS_going_ro, &c->flags) &&
1115 atomic_long_inc_not_zero(&c->writes[ref]);
1116 #else
1117 return percpu_ref_tryget(&c->writes);
1118 #endif
1119 }
1120
bch2_write_ref_tryget(struct bch_fs * c,enum bch_write_ref ref)1121 static inline bool bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1122 {
1123 #ifdef BCH_WRITE_REF_DEBUG
1124 return !test_bit(BCH_FS_going_ro, &c->flags) &&
1125 atomic_long_inc_not_zero(&c->writes[ref]);
1126 #else
1127 return percpu_ref_tryget_live(&c->writes);
1128 #endif
1129 }
1130
bch2_write_ref_put(struct bch_fs * c,enum bch_write_ref ref)1131 static inline void bch2_write_ref_put(struct bch_fs *c, enum bch_write_ref ref)
1132 {
1133 #ifdef BCH_WRITE_REF_DEBUG
1134 long v = atomic_long_dec_return(&c->writes[ref]);
1135
1136 BUG_ON(v < 0);
1137 if (v)
1138 return;
1139 for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++)
1140 if (atomic_long_read(&c->writes[i]))
1141 return;
1142
1143 set_bit(BCH_FS_write_disable_complete, &c->flags);
1144 wake_up(&bch2_read_only_wait);
1145 #else
1146 percpu_ref_put(&c->writes);
1147 #endif
1148 }
1149
bch2_ro_ref_tryget(struct bch_fs * c)1150 static inline bool bch2_ro_ref_tryget(struct bch_fs *c)
1151 {
1152 if (test_bit(BCH_FS_stopping, &c->flags))
1153 return false;
1154
1155 return refcount_inc_not_zero(&c->ro_ref);
1156 }
1157
bch2_ro_ref_put(struct bch_fs * c)1158 static inline void bch2_ro_ref_put(struct bch_fs *c)
1159 {
1160 if (refcount_dec_and_test(&c->ro_ref))
1161 wake_up(&c->ro_ref_wait);
1162 }
1163
bch2_set_ra_pages(struct bch_fs * c,unsigned ra_pages)1164 static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
1165 {
1166 #ifndef NO_BCACHEFS_FS
1167 if (c->vfs_sb)
1168 c->vfs_sb->s_bdi->ra_pages = ra_pages;
1169 #endif
1170 }
1171
bucket_bytes(const struct bch_dev * ca)1172 static inline unsigned bucket_bytes(const struct bch_dev *ca)
1173 {
1174 return ca->mi.bucket_size << 9;
1175 }
1176
block_bytes(const struct bch_fs * c)1177 static inline unsigned block_bytes(const struct bch_fs *c)
1178 {
1179 return c->opts.block_size;
1180 }
1181
block_sectors(const struct bch_fs * c)1182 static inline unsigned block_sectors(const struct bch_fs *c)
1183 {
1184 return c->opts.block_size >> 9;
1185 }
1186
btree_id_cached(const struct bch_fs * c,enum btree_id btree)1187 static inline bool btree_id_cached(const struct bch_fs *c, enum btree_id btree)
1188 {
1189 return c->btree_key_cache_btrees & (1U << btree);
1190 }
1191
bch2_time_to_timespec(const struct bch_fs * c,s64 time)1192 static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
1193 {
1194 struct timespec64 t;
1195 s64 sec;
1196 s32 rem;
1197
1198 time += c->sb.time_base_lo;
1199
1200 sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
1201
1202 set_normalized_timespec64(&t, sec, rem * (s64)c->sb.nsec_per_time_unit);
1203
1204 return t;
1205 }
1206
timespec_to_bch2_time(const struct bch_fs * c,struct timespec64 ts)1207 static inline s64 timespec_to_bch2_time(const struct bch_fs *c, struct timespec64 ts)
1208 {
1209 return (ts.tv_sec * c->sb.time_units_per_sec +
1210 (int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
1211 }
1212
bch2_current_time(const struct bch_fs * c)1213 static inline s64 bch2_current_time(const struct bch_fs *c)
1214 {
1215 struct timespec64 now;
1216
1217 ktime_get_coarse_real_ts64(&now);
1218 return timespec_to_bch2_time(c, now);
1219 }
1220
bch2_current_io_time(const struct bch_fs * c,int rw)1221 static inline u64 bch2_current_io_time(const struct bch_fs *c, int rw)
1222 {
1223 return max(1ULL, (u64) atomic64_read(&c->io_clock[rw].now) & LRU_TIME_MAX);
1224 }
1225
bch2_fs_stdio_redirect(struct bch_fs * c)1226 static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
1227 {
1228 struct stdio_redirect *stdio = c->stdio;
1229
1230 if (c->stdio_filter && c->stdio_filter != current)
1231 stdio = NULL;
1232 return stdio;
1233 }
1234
metadata_replicas_required(struct bch_fs * c)1235 static inline unsigned metadata_replicas_required(struct bch_fs *c)
1236 {
1237 return min(c->opts.metadata_replicas,
1238 c->opts.metadata_replicas_required);
1239 }
1240
data_replicas_required(struct bch_fs * c)1241 static inline unsigned data_replicas_required(struct bch_fs *c)
1242 {
1243 return min(c->opts.data_replicas,
1244 c->opts.data_replicas_required);
1245 }
1246
1247 #define BKEY_PADDED_ONSTACK(key, pad) \
1248 struct { struct bkey_i key; __u64 key ## _pad[pad]; }
1249
1250 #endif /* _BCACHEFS_H */
1251