xref: /linux/fs/bcachefs/bcachefs.h (revision c4101e55974cc7d835fbd2d8e01553a3f61e9e75)
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 "errcode.h"
209 #include "fifo.h"
210 #include "nocow_locking_types.h"
211 #include "opts.h"
212 #include "recovery_types.h"
213 #include "sb-errors_types.h"
214 #include "seqmutex.h"
215 #include "util.h"
216 
217 #ifdef CONFIG_BCACHEFS_DEBUG
218 #define BCH_WRITE_REF_DEBUG
219 #endif
220 
221 #ifndef dynamic_fault
222 #define dynamic_fault(...)		0
223 #endif
224 
225 #define race_fault(...)			dynamic_fault("bcachefs:race")
226 
227 #define count_event(_c, _name)	this_cpu_inc((_c)->counters[BCH_COUNTER_##_name])
228 
229 #define trace_and_count(_c, _name, ...)					\
230 do {									\
231 	count_event(_c, _name);						\
232 	trace_##_name(__VA_ARGS__);					\
233 } while (0)
234 
235 #define bch2_fs_init_fault(name)					\
236 	dynamic_fault("bcachefs:bch_fs_init:" name)
237 #define bch2_meta_read_fault(name)					\
238 	 dynamic_fault("bcachefs:meta:read:" name)
239 #define bch2_meta_write_fault(name)					\
240 	 dynamic_fault("bcachefs:meta:write:" name)
241 
242 #ifdef __KERNEL__
243 #define BCACHEFS_LOG_PREFIX
244 #endif
245 
246 #ifdef BCACHEFS_LOG_PREFIX
247 
248 #define bch2_log_msg(_c, fmt)			"bcachefs (%s): " fmt, ((_c)->name)
249 #define bch2_fmt_dev(_ca, fmt)			"bcachefs (%s): " fmt "\n", ((_ca)->name)
250 #define bch2_fmt_dev_offset(_ca, _offset, fmt)	"bcachefs (%s sector %llu): " fmt "\n", ((_ca)->name), (_offset)
251 #define bch2_fmt_inum(_c, _inum, fmt)		"bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
252 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt)			\
253 	 "bcachefs (%s inum %llu offset %llu): " fmt "\n", ((_c)->name), (_inum), (_offset)
254 
255 #else
256 
257 #define bch2_log_msg(_c, fmt)			fmt
258 #define bch2_fmt_dev(_ca, fmt)			"%s: " fmt "\n", ((_ca)->name)
259 #define bch2_fmt_dev_offset(_ca, _offset, fmt)	"%s sector %llu: " fmt "\n", ((_ca)->name), (_offset)
260 #define bch2_fmt_inum(_c, _inum, fmt)		"inum %llu: " fmt "\n", (_inum)
261 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt)				\
262 	 "inum %llu offset %llu: " fmt "\n", (_inum), (_offset)
263 
264 #endif
265 
266 #define bch2_fmt(_c, fmt)		bch2_log_msg(_c, fmt "\n")
267 
268 __printf(2, 3)
269 void __bch2_print(struct bch_fs *c, const char *fmt, ...);
270 
271 #define maybe_dev_to_fs(_c)	_Generic((_c),				\
272 	struct bch_dev *:	((struct bch_dev *) (_c))->fs,		\
273 	struct bch_fs *:	(_c))
274 
275 #define bch2_print(_c, ...) __bch2_print(maybe_dev_to_fs(_c), __VA_ARGS__)
276 
277 #define bch2_print_ratelimited(_c, ...)					\
278 do {									\
279 	static DEFINE_RATELIMIT_STATE(_rs,				\
280 				      DEFAULT_RATELIMIT_INTERVAL,	\
281 				      DEFAULT_RATELIMIT_BURST);		\
282 									\
283 	if (__ratelimit(&_rs))						\
284 		bch2_print(_c, __VA_ARGS__);				\
285 } while (0)
286 
287 #define bch_info(c, fmt, ...) \
288 	bch2_print(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
289 #define bch_notice(c, fmt, ...) \
290 	bch2_print(c, KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
291 #define bch_warn(c, fmt, ...) \
292 	bch2_print(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
293 #define bch_warn_ratelimited(c, fmt, ...) \
294 	bch2_print_ratelimited(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
295 
296 #define bch_err(c, fmt, ...) \
297 	bch2_print(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
298 #define bch_err_dev(ca, fmt, ...) \
299 	bch2_print(c, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
300 #define bch_err_dev_offset(ca, _offset, fmt, ...) \
301 	bch2_print(c, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
302 #define bch_err_inum(c, _inum, fmt, ...) \
303 	bch2_print(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
304 #define bch_err_inum_offset(c, _inum, _offset, fmt, ...) \
305 	bch2_print(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
306 
307 #define bch_err_ratelimited(c, fmt, ...) \
308 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
309 #define bch_err_dev_ratelimited(ca, fmt, ...) \
310 	bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
311 #define bch_err_dev_offset_ratelimited(ca, _offset, fmt, ...) \
312 	bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
313 #define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
314 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
315 #define bch_err_inum_offset_ratelimited(c, _inum, _offset, fmt, ...) \
316 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
317 
318 static inline bool should_print_err(int err)
319 {
320 	return err && !bch2_err_matches(err, BCH_ERR_transaction_restart);
321 }
322 
323 #define bch_err_fn(_c, _ret)						\
324 do {									\
325 	if (should_print_err(_ret))					\
326 		bch_err(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
327 } while (0)
328 
329 #define bch_err_fn_ratelimited(_c, _ret)				\
330 do {									\
331 	if (should_print_err(_ret))					\
332 		bch_err_ratelimited(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
333 } while (0)
334 
335 #define bch_err_msg(_c, _ret, _msg, ...)				\
336 do {									\
337 	if (should_print_err(_ret))					\
338 		bch_err(_c, "%s(): error " _msg " %s", __func__,	\
339 			##__VA_ARGS__, bch2_err_str(_ret));		\
340 } while (0)
341 
342 #define bch_verbose(c, fmt, ...)					\
343 do {									\
344 	if ((c)->opts.verbose)						\
345 		bch_info(c, fmt, ##__VA_ARGS__);			\
346 } while (0)
347 
348 #define pr_verbose_init(opts, fmt, ...)					\
349 do {									\
350 	if (opt_get(opts, verbose))					\
351 		pr_info(fmt, ##__VA_ARGS__);				\
352 } while (0)
353 
354 /* Parameters that are useful for debugging, but should always be compiled in: */
355 #define BCH_DEBUG_PARAMS_ALWAYS()					\
356 	BCH_DEBUG_PARAM(key_merging_disabled,				\
357 		"Disables merging of extents")				\
358 	BCH_DEBUG_PARAM(btree_gc_always_rewrite,			\
359 		"Causes mark and sweep to compact and rewrite every "	\
360 		"btree node it traverses")				\
361 	BCH_DEBUG_PARAM(btree_gc_rewrite_disabled,			\
362 		"Disables rewriting of btree nodes during mark and sweep")\
363 	BCH_DEBUG_PARAM(btree_shrinker_disabled,			\
364 		"Disables the shrinker callback for the btree node cache")\
365 	BCH_DEBUG_PARAM(verify_btree_ondisk,				\
366 		"Reread btree nodes at various points to verify the "	\
367 		"mergesort in the read path against modifications "	\
368 		"done in memory")					\
369 	BCH_DEBUG_PARAM(verify_all_btree_replicas,			\
370 		"When reading btree nodes, read all replicas and "	\
371 		"compare them")						\
372 	BCH_DEBUG_PARAM(backpointers_no_use_write_buffer,		\
373 		"Don't use the write buffer for backpointers, enabling "\
374 		"extra runtime checks")
375 
376 /* Parameters that should only be compiled in debug mode: */
377 #define BCH_DEBUG_PARAMS_DEBUG()					\
378 	BCH_DEBUG_PARAM(expensive_debug_checks,				\
379 		"Enables various runtime debugging checks that "	\
380 		"significantly affect performance")			\
381 	BCH_DEBUG_PARAM(debug_check_iterators,				\
382 		"Enables extra verification for btree iterators")	\
383 	BCH_DEBUG_PARAM(debug_check_btree_accounting,			\
384 		"Verify btree accounting for keys within a node")	\
385 	BCH_DEBUG_PARAM(journal_seq_verify,				\
386 		"Store the journal sequence number in the version "	\
387 		"number of every btree key, and verify that btree "	\
388 		"update ordering is preserved during recovery")		\
389 	BCH_DEBUG_PARAM(inject_invalid_keys,				\
390 		"Store the journal sequence number in the version "	\
391 		"number of every btree key, and verify that btree "	\
392 		"update ordering is preserved during recovery")		\
393 	BCH_DEBUG_PARAM(test_alloc_startup,				\
394 		"Force allocator startup to use the slowpath where it"	\
395 		"can't find enough free buckets without invalidating"	\
396 		"cached data")						\
397 	BCH_DEBUG_PARAM(force_reconstruct_read,				\
398 		"Force reads to use the reconstruct path, when reading"	\
399 		"from erasure coded extents")				\
400 	BCH_DEBUG_PARAM(test_restart_gc,				\
401 		"Test restarting mark and sweep gc when bucket gens change")
402 
403 #define BCH_DEBUG_PARAMS_ALL() BCH_DEBUG_PARAMS_ALWAYS() BCH_DEBUG_PARAMS_DEBUG()
404 
405 #ifdef CONFIG_BCACHEFS_DEBUG
406 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALL()
407 #else
408 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALWAYS()
409 #endif
410 
411 #define BCH_DEBUG_PARAM(name, description) extern bool bch2_##name;
412 BCH_DEBUG_PARAMS()
413 #undef BCH_DEBUG_PARAM
414 
415 #ifndef CONFIG_BCACHEFS_DEBUG
416 #define BCH_DEBUG_PARAM(name, description) static const __maybe_unused bool bch2_##name;
417 BCH_DEBUG_PARAMS_DEBUG()
418 #undef BCH_DEBUG_PARAM
419 #endif
420 
421 #define BCH_TIME_STATS()			\
422 	x(btree_node_mem_alloc)			\
423 	x(btree_node_split)			\
424 	x(btree_node_compact)			\
425 	x(btree_node_merge)			\
426 	x(btree_node_sort)			\
427 	x(btree_node_read)			\
428 	x(btree_node_read_done)			\
429 	x(btree_interior_update_foreground)	\
430 	x(btree_interior_update_total)		\
431 	x(btree_gc)				\
432 	x(data_write)				\
433 	x(data_read)				\
434 	x(data_promote)				\
435 	x(journal_flush_write)			\
436 	x(journal_noflush_write)		\
437 	x(journal_flush_seq)			\
438 	x(blocked_journal_low_on_space)		\
439 	x(blocked_journal_low_on_pin)		\
440 	x(blocked_journal_max_in_flight)	\
441 	x(blocked_allocate)			\
442 	x(blocked_allocate_open_bucket)		\
443 	x(blocked_write_buffer_full)		\
444 	x(nocow_lock_contended)
445 
446 enum bch_time_stats {
447 #define x(name) BCH_TIME_##name,
448 	BCH_TIME_STATS()
449 #undef x
450 	BCH_TIME_STAT_NR
451 };
452 
453 #include "alloc_types.h"
454 #include "btree_types.h"
455 #include "btree_write_buffer_types.h"
456 #include "buckets_types.h"
457 #include "buckets_waiting_for_journal_types.h"
458 #include "clock_types.h"
459 #include "disk_groups_types.h"
460 #include "ec_types.h"
461 #include "journal_types.h"
462 #include "keylist_types.h"
463 #include "quota_types.h"
464 #include "rebalance_types.h"
465 #include "replicas_types.h"
466 #include "subvolume_types.h"
467 #include "super_types.h"
468 #include "thread_with_file_types.h"
469 
470 /* Number of nodes btree coalesce will try to coalesce at once */
471 #define GC_MERGE_NODES		4U
472 
473 /* Maximum number of nodes we might need to allocate atomically: */
474 #define BTREE_RESERVE_MAX	(BTREE_MAX_DEPTH + (BTREE_MAX_DEPTH - 1))
475 
476 /* Size of the freelist we allocate btree nodes from: */
477 #define BTREE_NODE_RESERVE	(BTREE_RESERVE_MAX * 4)
478 
479 #define BTREE_NODE_OPEN_BUCKET_RESERVE	(BTREE_RESERVE_MAX * BCH_REPLICAS_MAX)
480 
481 struct btree;
482 
483 enum gc_phase {
484 	GC_PHASE_NOT_RUNNING,
485 	GC_PHASE_START,
486 	GC_PHASE_SB,
487 
488 	GC_PHASE_BTREE_stripes,
489 	GC_PHASE_BTREE_extents,
490 	GC_PHASE_BTREE_inodes,
491 	GC_PHASE_BTREE_dirents,
492 	GC_PHASE_BTREE_xattrs,
493 	GC_PHASE_BTREE_alloc,
494 	GC_PHASE_BTREE_quotas,
495 	GC_PHASE_BTREE_reflink,
496 	GC_PHASE_BTREE_subvolumes,
497 	GC_PHASE_BTREE_snapshots,
498 	GC_PHASE_BTREE_lru,
499 	GC_PHASE_BTREE_freespace,
500 	GC_PHASE_BTREE_need_discard,
501 	GC_PHASE_BTREE_backpointers,
502 	GC_PHASE_BTREE_bucket_gens,
503 	GC_PHASE_BTREE_snapshot_trees,
504 	GC_PHASE_BTREE_deleted_inodes,
505 	GC_PHASE_BTREE_logged_ops,
506 	GC_PHASE_BTREE_rebalance_work,
507 
508 	GC_PHASE_PENDING_DELETE,
509 };
510 
511 struct gc_pos {
512 	enum gc_phase		phase;
513 	struct bpos		pos;
514 	unsigned		level;
515 };
516 
517 struct reflink_gc {
518 	u64		offset;
519 	u32		size;
520 	u32		refcount;
521 };
522 
523 typedef GENRADIX(struct reflink_gc) reflink_gc_table;
524 
525 struct io_count {
526 	u64			sectors[2][BCH_DATA_NR];
527 };
528 
529 struct bch_dev {
530 	struct kobject		kobj;
531 	struct percpu_ref	ref;
532 	struct completion	ref_completion;
533 	struct percpu_ref	io_ref;
534 	struct completion	io_ref_completion;
535 
536 	struct bch_fs		*fs;
537 
538 	u8			dev_idx;
539 	/*
540 	 * Cached version of this device's member info from superblock
541 	 * Committed by bch2_write_super() -> bch_fs_mi_update()
542 	 */
543 	struct bch_member_cpu	mi;
544 	atomic64_t		errors[BCH_MEMBER_ERROR_NR];
545 
546 	__uuid_t		uuid;
547 	char			name[BDEVNAME_SIZE];
548 
549 	struct bch_sb_handle	disk_sb;
550 	struct bch_sb		*sb_read_scratch;
551 	int			sb_write_error;
552 	dev_t			dev;
553 	atomic_t		flush_seq;
554 
555 	struct bch_devs_mask	self;
556 
557 	/* biosets used in cloned bios for writing multiple replicas */
558 	struct bio_set		replica_set;
559 
560 	/*
561 	 * Buckets:
562 	 * Per-bucket arrays are protected by c->mark_lock, bucket_lock and
563 	 * gc_lock, for device resize - holding any is sufficient for access:
564 	 * Or rcu_read_lock(), but only for ptr_stale():
565 	 */
566 	struct bucket_array __rcu *buckets_gc;
567 	struct bucket_gens __rcu *bucket_gens;
568 	u8			*oldest_gen;
569 	unsigned long		*buckets_nouse;
570 	struct rw_semaphore	bucket_lock;
571 
572 	struct bch_dev_usage		*usage_base;
573 	struct bch_dev_usage __percpu	*usage[JOURNAL_BUF_NR];
574 	struct bch_dev_usage __percpu	*usage_gc;
575 
576 	/* Allocator: */
577 	u64			new_fs_bucket_idx;
578 	u64			alloc_cursor;
579 
580 	unsigned		nr_open_buckets;
581 	unsigned		nr_btree_reserve;
582 
583 	size_t			inc_gen_needs_gc;
584 	size_t			inc_gen_really_needs_gc;
585 	size_t			buckets_waiting_on_journal;
586 
587 	atomic64_t		rebalance_work;
588 
589 	struct journal_device	journal;
590 	u64			prev_journal_sector;
591 
592 	struct work_struct	io_error_work;
593 
594 	/* The rest of this all shows up in sysfs */
595 	atomic64_t		cur_latency[2];
596 	struct bch2_time_stats	io_latency[2];
597 
598 #define CONGESTED_MAX		1024
599 	atomic_t		congested;
600 	u64			congested_last;
601 
602 	struct io_count __percpu *io_done;
603 };
604 
605 /*
606  * initial_gc_unfixed
607  * error
608  * topology error
609  */
610 
611 #define BCH_FS_FLAGS()			\
612 	x(started)			\
613 	x(may_go_rw)			\
614 	x(rw)				\
615 	x(was_rw)			\
616 	x(stopping)			\
617 	x(emergency_ro)			\
618 	x(going_ro)			\
619 	x(write_disable_complete)	\
620 	x(clean_shutdown)		\
621 	x(fsck_running)			\
622 	x(initial_gc_unfixed)		\
623 	x(need_another_gc)		\
624 	x(need_delete_dead_snapshots)	\
625 	x(error)			\
626 	x(topology_error)		\
627 	x(errors_fixed)			\
628 	x(errors_not_fixed)
629 
630 enum bch_fs_flags {
631 #define x(n)		BCH_FS_##n,
632 	BCH_FS_FLAGS()
633 #undef x
634 };
635 
636 struct btree_debug {
637 	unsigned		id;
638 };
639 
640 #define BCH_TRANSACTIONS_NR 128
641 
642 struct btree_transaction_stats {
643 	struct bch2_time_stats	duration;
644 	struct bch2_time_stats	lock_hold_times;
645 	struct mutex		lock;
646 	unsigned		nr_max_paths;
647 	unsigned		journal_entries_size;
648 	unsigned		max_mem;
649 	char			*max_paths_text;
650 };
651 
652 struct bch_fs_pcpu {
653 	u64			sectors_available;
654 };
655 
656 struct journal_seq_blacklist_table {
657 	size_t			nr;
658 	struct journal_seq_blacklist_table_entry {
659 		u64		start;
660 		u64		end;
661 		bool		dirty;
662 	}			entries[];
663 };
664 
665 struct journal_keys {
666 	struct journal_key {
667 		u64		journal_seq;
668 		u32		journal_offset;
669 		enum btree_id	btree_id:8;
670 		unsigned	level:8;
671 		bool		allocated;
672 		bool		overwritten;
673 		struct bkey_i	*k;
674 	}			*d;
675 	/*
676 	 * Gap buffer: instead of all the empty space in the array being at the
677 	 * end of the buffer - from @nr to @size - the empty space is at @gap.
678 	 * This means that sequential insertions are O(n) instead of O(n^2).
679 	 */
680 	size_t			gap;
681 	size_t			nr;
682 	size_t			size;
683 	atomic_t		ref;
684 	bool			initial_ref_held;
685 };
686 
687 struct btree_trans_buf {
688 	struct btree_trans	*trans;
689 };
690 
691 #define REPLICAS_DELTA_LIST_MAX	(1U << 16)
692 
693 #define BCACHEFS_ROOT_SUBVOL_INUM					\
694 	((subvol_inum) { BCACHEFS_ROOT_SUBVOL,	BCACHEFS_ROOT_INO })
695 
696 #define BCH_WRITE_REFS()						\
697 	x(trans)							\
698 	x(write)							\
699 	x(promote)							\
700 	x(node_rewrite)							\
701 	x(stripe_create)						\
702 	x(stripe_delete)						\
703 	x(reflink)							\
704 	x(fallocate)							\
705 	x(discard)							\
706 	x(invalidate)							\
707 	x(delete_dead_snapshots)					\
708 	x(snapshot_delete_pagecache)					\
709 	x(sysfs)							\
710 	x(btree_write_buffer)
711 
712 enum bch_write_ref {
713 #define x(n) BCH_WRITE_REF_##n,
714 	BCH_WRITE_REFS()
715 #undef x
716 	BCH_WRITE_REF_NR,
717 };
718 
719 struct bch_fs {
720 	struct closure		cl;
721 
722 	struct list_head	list;
723 	struct kobject		kobj;
724 	struct kobject		counters_kobj;
725 	struct kobject		internal;
726 	struct kobject		opts_dir;
727 	struct kobject		time_stats;
728 	unsigned long		flags;
729 
730 	int			minor;
731 	struct device		*chardev;
732 	struct super_block	*vfs_sb;
733 	dev_t			dev;
734 	char			name[40];
735 	struct stdio_redirect	*stdio;
736 	struct task_struct	*stdio_filter;
737 
738 	/* ro/rw, add/remove/resize devices: */
739 	struct rw_semaphore	state_lock;
740 
741 	/* Counts outstanding writes, for clean transition to read-only */
742 #ifdef BCH_WRITE_REF_DEBUG
743 	atomic_long_t		writes[BCH_WRITE_REF_NR];
744 #else
745 	struct percpu_ref	writes;
746 #endif
747 	/*
748 	 * Analagous to c->writes, for asynchronous ops that don't necessarily
749 	 * need fs to be read-write
750 	 */
751 	refcount_t		ro_ref;
752 	wait_queue_head_t	ro_ref_wait;
753 
754 	struct work_struct	read_only_work;
755 
756 	struct bch_dev __rcu	*devs[BCH_SB_MEMBERS_MAX];
757 
758 	struct bch_replicas_cpu replicas;
759 	struct bch_replicas_cpu replicas_gc;
760 	struct mutex		replicas_gc_lock;
761 	mempool_t		replicas_delta_pool;
762 
763 	struct journal_entry_res btree_root_journal_res;
764 	struct journal_entry_res replicas_journal_res;
765 	struct journal_entry_res clock_journal_res;
766 	struct journal_entry_res dev_usage_journal_res;
767 
768 	struct bch_disk_groups_cpu __rcu *disk_groups;
769 
770 	struct bch_opts		opts;
771 
772 	/* Updated by bch2_sb_update():*/
773 	struct {
774 		__uuid_t	uuid;
775 		__uuid_t	user_uuid;
776 
777 		u16		version;
778 		u16		version_min;
779 		u16		version_upgrade_complete;
780 
781 		u8		nr_devices;
782 		u8		clean;
783 
784 		u8		encryption_type;
785 
786 		u64		time_base_lo;
787 		u32		time_base_hi;
788 		unsigned	time_units_per_sec;
789 		unsigned	nsec_per_time_unit;
790 		u64		features;
791 		u64		compat;
792 		unsigned long	errors_silent[BITS_TO_LONGS(BCH_SB_ERR_MAX)];
793 	}			sb;
794 
795 
796 	struct bch_sb_handle	disk_sb;
797 
798 	unsigned short		block_bits;	/* ilog2(block_size) */
799 
800 	u16			btree_foreground_merge_threshold;
801 
802 	struct closure		sb_write;
803 	struct mutex		sb_lock;
804 
805 	/* snapshot.c: */
806 	struct snapshot_table __rcu *snapshots;
807 	size_t			snapshot_table_size;
808 	struct mutex		snapshot_table_lock;
809 	struct rw_semaphore	snapshot_create_lock;
810 
811 	struct work_struct	snapshot_delete_work;
812 	struct work_struct	snapshot_wait_for_pagecache_and_delete_work;
813 	snapshot_id_list	snapshots_unlinked;
814 	struct mutex		snapshots_unlinked_lock;
815 
816 	/* BTREE CACHE */
817 	struct bio_set		btree_bio;
818 	struct workqueue_struct	*io_complete_wq;
819 
820 	struct btree_root	btree_roots_known[BTREE_ID_NR];
821 	DARRAY(struct btree_root) btree_roots_extra;
822 	struct mutex		btree_root_lock;
823 
824 	struct btree_cache	btree_cache;
825 
826 	/*
827 	 * Cache of allocated btree nodes - if we allocate a btree node and
828 	 * don't use it, if we free it that space can't be reused until going
829 	 * _all_ the way through the allocator (which exposes us to a livelock
830 	 * when allocating btree reserves fail halfway through) - instead, we
831 	 * can stick them here:
832 	 */
833 	struct btree_alloc	btree_reserve_cache[BTREE_NODE_RESERVE * 2];
834 	unsigned		btree_reserve_cache_nr;
835 	struct mutex		btree_reserve_cache_lock;
836 
837 	mempool_t		btree_interior_update_pool;
838 	struct list_head	btree_interior_update_list;
839 	struct list_head	btree_interior_updates_unwritten;
840 	struct mutex		btree_interior_update_lock;
841 	struct closure_waitlist	btree_interior_update_wait;
842 
843 	struct workqueue_struct	*btree_interior_update_worker;
844 	struct work_struct	btree_interior_update_work;
845 
846 	struct list_head	pending_node_rewrites;
847 	struct mutex		pending_node_rewrites_lock;
848 
849 	/* btree_io.c: */
850 	spinlock_t		btree_write_error_lock;
851 	struct btree_write_stats {
852 		atomic64_t	nr;
853 		atomic64_t	bytes;
854 	}			btree_write_stats[BTREE_WRITE_TYPE_NR];
855 
856 	/* btree_iter.c: */
857 	struct seqmutex		btree_trans_lock;
858 	struct list_head	btree_trans_list;
859 	mempool_t		btree_trans_pool;
860 	mempool_t		btree_trans_mem_pool;
861 	struct btree_trans_buf  __percpu	*btree_trans_bufs;
862 
863 	struct srcu_struct	btree_trans_barrier;
864 	bool			btree_trans_barrier_initialized;
865 
866 	struct btree_key_cache	btree_key_cache;
867 	unsigned		btree_key_cache_btrees;
868 
869 	struct btree_write_buffer btree_write_buffer;
870 
871 	struct workqueue_struct	*btree_update_wq;
872 	struct workqueue_struct	*btree_io_complete_wq;
873 	/* copygc needs its own workqueue for index updates.. */
874 	struct workqueue_struct	*copygc_wq;
875 	/*
876 	 * Use a dedicated wq for write ref holder tasks. Required to avoid
877 	 * dependency problems with other wq tasks that can block on ref
878 	 * draining, such as read-only transition.
879 	 */
880 	struct workqueue_struct *write_ref_wq;
881 
882 	/* ALLOCATION */
883 	struct bch_devs_mask	rw_devs[BCH_DATA_NR];
884 
885 	u64			capacity; /* sectors */
886 
887 	/*
888 	 * When capacity _decreases_ (due to a disk being removed), we
889 	 * increment capacity_gen - this invalidates outstanding reservations
890 	 * and forces them to be revalidated
891 	 */
892 	u32			capacity_gen;
893 	unsigned		bucket_size_max;
894 
895 	atomic64_t		sectors_available;
896 	struct mutex		sectors_available_lock;
897 
898 	struct bch_fs_pcpu __percpu	*pcpu;
899 
900 	struct percpu_rw_semaphore	mark_lock;
901 
902 	seqcount_t			usage_lock;
903 	struct bch_fs_usage		*usage_base;
904 	struct bch_fs_usage __percpu	*usage[JOURNAL_BUF_NR];
905 	struct bch_fs_usage __percpu	*usage_gc;
906 	u64 __percpu		*online_reserved;
907 
908 	/* single element mempool: */
909 	struct mutex		usage_scratch_lock;
910 	struct bch_fs_usage_online *usage_scratch;
911 
912 	struct io_clock		io_clock[2];
913 
914 	/* JOURNAL SEQ BLACKLIST */
915 	struct journal_seq_blacklist_table *
916 				journal_seq_blacklist_table;
917 	struct work_struct	journal_seq_blacklist_gc_work;
918 
919 	/* ALLOCATOR */
920 	spinlock_t		freelist_lock;
921 	struct closure_waitlist	freelist_wait;
922 	u64			blocked_allocate;
923 	u64			blocked_allocate_open_bucket;
924 
925 	open_bucket_idx_t	open_buckets_freelist;
926 	open_bucket_idx_t	open_buckets_nr_free;
927 	struct closure_waitlist	open_buckets_wait;
928 	struct open_bucket	open_buckets[OPEN_BUCKETS_COUNT];
929 	open_bucket_idx_t	open_buckets_hash[OPEN_BUCKETS_COUNT];
930 
931 	open_bucket_idx_t	open_buckets_partial[OPEN_BUCKETS_COUNT];
932 	open_bucket_idx_t	open_buckets_partial_nr;
933 
934 	struct write_point	btree_write_point;
935 	struct write_point	rebalance_write_point;
936 
937 	struct write_point	write_points[WRITE_POINT_MAX];
938 	struct hlist_head	write_points_hash[WRITE_POINT_HASH_NR];
939 	struct mutex		write_points_hash_lock;
940 	unsigned		write_points_nr;
941 
942 	struct buckets_waiting_for_journal buckets_waiting_for_journal;
943 	struct work_struct	discard_work;
944 	struct work_struct	invalidate_work;
945 
946 	/* GARBAGE COLLECTION */
947 	struct task_struct	*gc_thread;
948 	atomic_t		kick_gc;
949 	unsigned long		gc_count;
950 
951 	enum btree_id		gc_gens_btree;
952 	struct bpos		gc_gens_pos;
953 
954 	/*
955 	 * Tracks GC's progress - everything in the range [ZERO_KEY..gc_cur_pos]
956 	 * has been marked by GC.
957 	 *
958 	 * gc_cur_phase is a superset of btree_ids (BTREE_ID_extents etc.)
959 	 *
960 	 * Protected by gc_pos_lock. Only written to by GC thread, so GC thread
961 	 * can read without a lock.
962 	 */
963 	seqcount_t		gc_pos_lock;
964 	struct gc_pos		gc_pos;
965 
966 	/*
967 	 * The allocation code needs gc_mark in struct bucket to be correct, but
968 	 * it's not while a gc is in progress.
969 	 */
970 	struct rw_semaphore	gc_lock;
971 	struct mutex		gc_gens_lock;
972 
973 	/* IO PATH */
974 	struct semaphore	io_in_flight;
975 	struct bio_set		bio_read;
976 	struct bio_set		bio_read_split;
977 	struct bio_set		bio_write;
978 	struct mutex		bio_bounce_pages_lock;
979 	mempool_t		bio_bounce_pages;
980 	struct bucket_nocow_lock_table
981 				nocow_locks;
982 	struct rhashtable	promote_table;
983 
984 	mempool_t		compression_bounce[2];
985 	mempool_t		compress_workspace[BCH_COMPRESSION_TYPE_NR];
986 	mempool_t		decompress_workspace;
987 	size_t			zstd_workspace_size;
988 
989 	struct crypto_shash	*sha256;
990 	struct crypto_sync_skcipher *chacha20;
991 	struct crypto_shash	*poly1305;
992 
993 	atomic64_t		key_version;
994 
995 	mempool_t		large_bkey_pool;
996 
997 	/* MOVE.C */
998 	struct list_head	moving_context_list;
999 	struct mutex		moving_context_lock;
1000 
1001 	/* REBALANCE */
1002 	struct bch_fs_rebalance	rebalance;
1003 
1004 	/* COPYGC */
1005 	struct task_struct	*copygc_thread;
1006 	struct write_point	copygc_write_point;
1007 	s64			copygc_wait_at;
1008 	s64			copygc_wait;
1009 	bool			copygc_running;
1010 	wait_queue_head_t	copygc_running_wq;
1011 
1012 	/* STRIPES: */
1013 	GENRADIX(struct stripe) stripes;
1014 	GENRADIX(struct gc_stripe) gc_stripes;
1015 
1016 	struct hlist_head	ec_stripes_new[32];
1017 	spinlock_t		ec_stripes_new_lock;
1018 
1019 	ec_stripes_heap		ec_stripes_heap;
1020 	struct mutex		ec_stripes_heap_lock;
1021 
1022 	/* ERASURE CODING */
1023 	struct list_head	ec_stripe_head_list;
1024 	struct mutex		ec_stripe_head_lock;
1025 
1026 	struct list_head	ec_stripe_new_list;
1027 	struct mutex		ec_stripe_new_lock;
1028 	wait_queue_head_t	ec_stripe_new_wait;
1029 
1030 	struct work_struct	ec_stripe_create_work;
1031 	u64			ec_stripe_hint;
1032 
1033 	struct work_struct	ec_stripe_delete_work;
1034 
1035 	struct bio_set		ec_bioset;
1036 
1037 	/* REFLINK */
1038 	reflink_gc_table	reflink_gc_table;
1039 	size_t			reflink_gc_nr;
1040 
1041 	/* fs.c */
1042 	struct list_head	vfs_inodes_list;
1043 	struct mutex		vfs_inodes_lock;
1044 
1045 	/* VFS IO PATH - fs-io.c */
1046 	struct bio_set		writepage_bioset;
1047 	struct bio_set		dio_write_bioset;
1048 	struct bio_set		dio_read_bioset;
1049 	struct bio_set		nocow_flush_bioset;
1050 
1051 	/* QUOTAS */
1052 	struct bch_memquota_type quotas[QTYP_NR];
1053 
1054 	/* RECOVERY */
1055 	u64			journal_replay_seq_start;
1056 	u64			journal_replay_seq_end;
1057 	/*
1058 	 * Two different uses:
1059 	 * "Has this fsck pass?" - i.e. should this type of error be an
1060 	 * emergency read-only
1061 	 * And, in certain situations fsck will rewind to an earlier pass: used
1062 	 * for signaling to the toplevel code which pass we want to run now.
1063 	 */
1064 	enum bch_recovery_pass	curr_recovery_pass;
1065 	/* bitmap of explicitly enabled recovery passes: */
1066 	u64			recovery_passes_explicit;
1067 	/* bitmask of recovery passes that we actually ran */
1068 	u64			recovery_passes_complete;
1069 	/* never rewinds version of curr_recovery_pass */
1070 	enum bch_recovery_pass	recovery_pass_done;
1071 	struct semaphore	online_fsck_mutex;
1072 
1073 	/* DEBUG JUNK */
1074 	struct dentry		*fs_debug_dir;
1075 	struct dentry		*btree_debug_dir;
1076 	struct btree_debug	btree_debug[BTREE_ID_NR];
1077 	struct btree		*verify_data;
1078 	struct btree_node	*verify_ondisk;
1079 	struct mutex		verify_lock;
1080 
1081 	u64			*unused_inode_hints;
1082 	unsigned		inode_shard_bits;
1083 
1084 	/*
1085 	 * A btree node on disk could have too many bsets for an iterator to fit
1086 	 * on the stack - have to dynamically allocate them
1087 	 */
1088 	mempool_t		fill_iter;
1089 
1090 	mempool_t		btree_bounce_pool;
1091 
1092 	struct journal		journal;
1093 	GENRADIX(struct journal_replay *) journal_entries;
1094 	u64			journal_entries_base_seq;
1095 	struct journal_keys	journal_keys;
1096 	struct list_head	journal_iters;
1097 
1098 	u64			last_bucket_seq_cleanup;
1099 
1100 	u64			counters_on_mount[BCH_COUNTER_NR];
1101 	u64 __percpu		*counters;
1102 
1103 	unsigned		btree_gc_periodic:1;
1104 	unsigned		copy_gc_enabled:1;
1105 	bool			promote_whole_extents;
1106 
1107 	struct bch2_time_stats	times[BCH_TIME_STAT_NR];
1108 
1109 	struct btree_transaction_stats btree_transaction_stats[BCH_TRANSACTIONS_NR];
1110 
1111 	/* ERRORS */
1112 	struct list_head	fsck_error_msgs;
1113 	struct mutex		fsck_error_msgs_lock;
1114 	bool			fsck_alloc_msgs_err;
1115 
1116 	bch_sb_errors_cpu	fsck_error_counts;
1117 	struct mutex		fsck_error_counts_lock;
1118 };
1119 
1120 extern struct wait_queue_head bch2_read_only_wait;
1121 
1122 static inline void bch2_write_ref_get(struct bch_fs *c, enum bch_write_ref ref)
1123 {
1124 #ifdef BCH_WRITE_REF_DEBUG
1125 	atomic_long_inc(&c->writes[ref]);
1126 #else
1127 	percpu_ref_get(&c->writes);
1128 #endif
1129 }
1130 
1131 static inline bool __bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1132 {
1133 #ifdef BCH_WRITE_REF_DEBUG
1134 	return !test_bit(BCH_FS_going_ro, &c->flags) &&
1135 		atomic_long_inc_not_zero(&c->writes[ref]);
1136 #else
1137 	return percpu_ref_tryget(&c->writes);
1138 #endif
1139 }
1140 
1141 static inline bool bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1142 {
1143 #ifdef BCH_WRITE_REF_DEBUG
1144 	return !test_bit(BCH_FS_going_ro, &c->flags) &&
1145 		atomic_long_inc_not_zero(&c->writes[ref]);
1146 #else
1147 	return percpu_ref_tryget_live(&c->writes);
1148 #endif
1149 }
1150 
1151 static inline void bch2_write_ref_put(struct bch_fs *c, enum bch_write_ref ref)
1152 {
1153 #ifdef BCH_WRITE_REF_DEBUG
1154 	long v = atomic_long_dec_return(&c->writes[ref]);
1155 
1156 	BUG_ON(v < 0);
1157 	if (v)
1158 		return;
1159 	for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++)
1160 		if (atomic_long_read(&c->writes[i]))
1161 			return;
1162 
1163 	set_bit(BCH_FS_write_disable_complete, &c->flags);
1164 	wake_up(&bch2_read_only_wait);
1165 #else
1166 	percpu_ref_put(&c->writes);
1167 #endif
1168 }
1169 
1170 static inline bool bch2_ro_ref_tryget(struct bch_fs *c)
1171 {
1172 	if (test_bit(BCH_FS_stopping, &c->flags))
1173 		return false;
1174 
1175 	return refcount_inc_not_zero(&c->ro_ref);
1176 }
1177 
1178 static inline void bch2_ro_ref_put(struct bch_fs *c)
1179 {
1180 	if (refcount_dec_and_test(&c->ro_ref))
1181 		wake_up(&c->ro_ref_wait);
1182 }
1183 
1184 static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
1185 {
1186 #ifndef NO_BCACHEFS_FS
1187 	if (c->vfs_sb)
1188 		c->vfs_sb->s_bdi->ra_pages = ra_pages;
1189 #endif
1190 }
1191 
1192 static inline unsigned bucket_bytes(const struct bch_dev *ca)
1193 {
1194 	return ca->mi.bucket_size << 9;
1195 }
1196 
1197 static inline unsigned block_bytes(const struct bch_fs *c)
1198 {
1199 	return c->opts.block_size;
1200 }
1201 
1202 static inline unsigned block_sectors(const struct bch_fs *c)
1203 {
1204 	return c->opts.block_size >> 9;
1205 }
1206 
1207 static inline size_t btree_sectors(const struct bch_fs *c)
1208 {
1209 	return c->opts.btree_node_size >> 9;
1210 }
1211 
1212 static inline bool btree_id_cached(const struct bch_fs *c, enum btree_id btree)
1213 {
1214 	return c->btree_key_cache_btrees & (1U << btree);
1215 }
1216 
1217 static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
1218 {
1219 	struct timespec64 t;
1220 	s32 rem;
1221 
1222 	time += c->sb.time_base_lo;
1223 
1224 	t.tv_sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
1225 	t.tv_nsec = rem * c->sb.nsec_per_time_unit;
1226 	return t;
1227 }
1228 
1229 static inline s64 timespec_to_bch2_time(const struct bch_fs *c, struct timespec64 ts)
1230 {
1231 	return (ts.tv_sec * c->sb.time_units_per_sec +
1232 		(int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
1233 }
1234 
1235 static inline s64 bch2_current_time(const struct bch_fs *c)
1236 {
1237 	struct timespec64 now;
1238 
1239 	ktime_get_coarse_real_ts64(&now);
1240 	return timespec_to_bch2_time(c, now);
1241 }
1242 
1243 static inline bool bch2_dev_exists2(const struct bch_fs *c, unsigned dev)
1244 {
1245 	return dev < c->sb.nr_devices && c->devs[dev];
1246 }
1247 
1248 static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
1249 {
1250 	struct stdio_redirect *stdio = c->stdio;
1251 
1252 	if (c->stdio_filter && c->stdio_filter != current)
1253 		stdio = NULL;
1254 	return stdio;
1255 }
1256 
1257 #define BKEY_PADDED_ONSTACK(key, pad)				\
1258 	struct { struct bkey_i key; __u64 key ## _pad[pad]; }
1259 
1260 #endif /* _BCACHEFS_H */
1261