xref: /linux/fs/bcachefs/bcachefs.h (revision 4a4b30ea80d8cb5e8c4c62bb86201f4ea0d9b030)
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 #include <linux/unicode.h>
207 
208 #include "bcachefs_format.h"
209 #include "btree_journal_iter_types.h"
210 #include "disk_accounting_types.h"
211 #include "errcode.h"
212 #include "fifo.h"
213 #include "nocow_locking_types.h"
214 #include "opts.h"
215 #include "recovery_passes_types.h"
216 #include "sb-errors_types.h"
217 #include "seqmutex.h"
218 #include "time_stats.h"
219 #include "util.h"
220 
221 #ifdef CONFIG_BCACHEFS_DEBUG
222 #define BCH_WRITE_REF_DEBUG
223 #endif
224 
225 #ifndef dynamic_fault
226 #define dynamic_fault(...)		0
227 #endif
228 
229 #define race_fault(...)			dynamic_fault("bcachefs:race")
230 
231 #define count_event(_c, _name)	this_cpu_inc((_c)->counters[BCH_COUNTER_##_name])
232 
233 #define trace_and_count(_c, _name, ...)					\
234 do {									\
235 	count_event(_c, _name);						\
236 	trace_##_name(__VA_ARGS__);					\
237 } while (0)
238 
239 #define bch2_fs_init_fault(name)					\
240 	dynamic_fault("bcachefs:bch_fs_init:" name)
241 #define bch2_meta_read_fault(name)					\
242 	 dynamic_fault("bcachefs:meta:read:" name)
243 #define bch2_meta_write_fault(name)					\
244 	 dynamic_fault("bcachefs:meta:write:" name)
245 
246 #ifdef __KERNEL__
247 #define BCACHEFS_LOG_PREFIX
248 #endif
249 
250 #ifdef BCACHEFS_LOG_PREFIX
251 
252 #define bch2_log_msg(_c, fmt)			"bcachefs (%s): " fmt, ((_c)->name)
253 #define bch2_fmt_dev(_ca, fmt)			"bcachefs (%s): " fmt "\n", ((_ca)->name)
254 #define bch2_fmt_dev_offset(_ca, _offset, fmt)	"bcachefs (%s sector %llu): " fmt "\n", ((_ca)->name), (_offset)
255 #define bch2_fmt_inum(_c, _inum, fmt)		"bcachefs (%s inum %llu): " fmt "\n", ((_c)->name), (_inum)
256 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt)			\
257 	 "bcachefs (%s inum %llu offset %llu): " fmt "\n", ((_c)->name), (_inum), (_offset)
258 
259 #else
260 
261 #define bch2_log_msg(_c, fmt)			fmt
262 #define bch2_fmt_dev(_ca, fmt)			"%s: " fmt "\n", ((_ca)->name)
263 #define bch2_fmt_dev_offset(_ca, _offset, fmt)	"%s sector %llu: " fmt "\n", ((_ca)->name), (_offset)
264 #define bch2_fmt_inum(_c, _inum, fmt)		"inum %llu: " fmt "\n", (_inum)
265 #define bch2_fmt_inum_offset(_c, _inum, _offset, fmt)				\
266 	 "inum %llu offset %llu: " fmt "\n", (_inum), (_offset)
267 
268 #endif
269 
270 #define bch2_fmt(_c, fmt)		bch2_log_msg(_c, fmt "\n")
271 
272 void bch2_print_str(struct bch_fs *, const char *);
273 
274 __printf(2, 3)
275 void bch2_print_opts(struct bch_opts *, const char *, ...);
276 
277 __printf(2, 3)
278 void __bch2_print(struct bch_fs *c, const char *fmt, ...);
279 
280 #define maybe_dev_to_fs(_c)	_Generic((_c),				\
281 	struct bch_dev *:	((struct bch_dev *) (_c))->fs,		\
282 	struct bch_fs *:	(_c))
283 
284 #define bch2_print(_c, ...) __bch2_print(maybe_dev_to_fs(_c), __VA_ARGS__)
285 
286 #define bch2_print_ratelimited(_c, ...)					\
287 do {									\
288 	static DEFINE_RATELIMIT_STATE(_rs,				\
289 				      DEFAULT_RATELIMIT_INTERVAL,	\
290 				      DEFAULT_RATELIMIT_BURST);		\
291 									\
292 	if (__ratelimit(&_rs))						\
293 		bch2_print(_c, __VA_ARGS__);				\
294 } while (0)
295 
296 #define bch_info(c, fmt, ...) \
297 	bch2_print(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
298 #define bch_info_ratelimited(c, fmt, ...) \
299 	bch2_print_ratelimited(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
300 #define bch_notice(c, fmt, ...) \
301 	bch2_print(c, KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
302 #define bch_warn(c, fmt, ...) \
303 	bch2_print(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
304 #define bch_warn_ratelimited(c, fmt, ...) \
305 	bch2_print_ratelimited(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
306 
307 #define bch_err(c, fmt, ...) \
308 	bch2_print(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
309 #define bch_err_dev(ca, fmt, ...) \
310 	bch2_print(c, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
311 #define bch_err_dev_offset(ca, _offset, fmt, ...) \
312 	bch2_print(c, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
313 #define bch_err_inum(c, _inum, fmt, ...) \
314 	bch2_print(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
315 #define bch_err_inum_offset(c, _inum, _offset, fmt, ...) \
316 	bch2_print(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
317 
318 #define bch_err_ratelimited(c, fmt, ...) \
319 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
320 #define bch_err_dev_ratelimited(ca, fmt, ...) \
321 	bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
322 #define bch_err_dev_offset_ratelimited(ca, _offset, fmt, ...) \
323 	bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
324 #define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
325 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
326 #define bch_err_inum_offset_ratelimited(c, _inum, _offset, fmt, ...) \
327 	bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
328 
should_print_err(int err)329 static inline bool should_print_err(int err)
330 {
331 	return err && !bch2_err_matches(err, BCH_ERR_transaction_restart);
332 }
333 
334 #define bch_err_fn(_c, _ret)						\
335 do {									\
336 	if (should_print_err(_ret))					\
337 		bch_err(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
338 } while (0)
339 
340 #define bch_err_fn_ratelimited(_c, _ret)				\
341 do {									\
342 	if (should_print_err(_ret))					\
343 		bch_err_ratelimited(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
344 } while (0)
345 
346 #define bch_err_msg(_c, _ret, _msg, ...)				\
347 do {									\
348 	if (should_print_err(_ret))					\
349 		bch_err(_c, "%s(): error " _msg " %s", __func__,	\
350 			##__VA_ARGS__, bch2_err_str(_ret));		\
351 } while (0)
352 
353 #define bch_verbose(c, fmt, ...)					\
354 do {									\
355 	if ((c)->opts.verbose)						\
356 		bch_info(c, fmt, ##__VA_ARGS__);			\
357 } while (0)
358 
359 #define bch_verbose_ratelimited(c, fmt, ...)				\
360 do {									\
361 	if ((c)->opts.verbose)						\
362 		bch_info_ratelimited(c, fmt, ##__VA_ARGS__);		\
363 } while (0)
364 
365 #define pr_verbose_init(opts, fmt, ...)					\
366 do {									\
367 	if (opt_get(opts, verbose))					\
368 		pr_info(fmt, ##__VA_ARGS__);				\
369 } while (0)
370 
371 /* Parameters that are useful for debugging, but should always be compiled in: */
372 #define BCH_DEBUG_PARAMS_ALWAYS()					\
373 	BCH_DEBUG_PARAM(key_merging_disabled,				\
374 		"Disables merging of extents")				\
375 	BCH_DEBUG_PARAM(btree_node_merging_disabled,			\
376 		"Disables merging of btree nodes")			\
377 	BCH_DEBUG_PARAM(btree_gc_always_rewrite,			\
378 		"Causes mark and sweep to compact and rewrite every "	\
379 		"btree node it traverses")				\
380 	BCH_DEBUG_PARAM(btree_gc_rewrite_disabled,			\
381 		"Disables rewriting of btree nodes during mark and sweep")\
382 	BCH_DEBUG_PARAM(btree_shrinker_disabled,			\
383 		"Disables the shrinker callback for the btree node cache")\
384 	BCH_DEBUG_PARAM(verify_btree_ondisk,				\
385 		"Reread btree nodes at various points to verify the "	\
386 		"mergesort in the read path against modifications "	\
387 		"done in memory")					\
388 	BCH_DEBUG_PARAM(verify_all_btree_replicas,			\
389 		"When reading btree nodes, read all replicas and "	\
390 		"compare them")						\
391 	BCH_DEBUG_PARAM(backpointers_no_use_write_buffer,		\
392 		"Don't use the write buffer for backpointers, enabling "\
393 		"extra runtime checks")
394 
395 /* Parameters that should only be compiled in debug mode: */
396 #define BCH_DEBUG_PARAMS_DEBUG()					\
397 	BCH_DEBUG_PARAM(expensive_debug_checks,				\
398 		"Enables various runtime debugging checks that "	\
399 		"significantly affect performance")			\
400 	BCH_DEBUG_PARAM(debug_check_iterators,				\
401 		"Enables extra verification for btree iterators")	\
402 	BCH_DEBUG_PARAM(debug_check_btree_accounting,			\
403 		"Verify btree accounting for keys within a node")	\
404 	BCH_DEBUG_PARAM(journal_seq_verify,				\
405 		"Store the journal sequence number in the version "	\
406 		"number of every btree key, and verify that btree "	\
407 		"update ordering is preserved during recovery")		\
408 	BCH_DEBUG_PARAM(inject_invalid_keys,				\
409 		"Store the journal sequence number in the version "	\
410 		"number of every btree key, and verify that btree "	\
411 		"update ordering is preserved during recovery")		\
412 	BCH_DEBUG_PARAM(test_alloc_startup,				\
413 		"Force allocator startup to use the slowpath where it"	\
414 		"can't find enough free buckets without invalidating"	\
415 		"cached data")						\
416 	BCH_DEBUG_PARAM(force_reconstruct_read,				\
417 		"Force reads to use the reconstruct path, when reading"	\
418 		"from erasure coded extents")				\
419 	BCH_DEBUG_PARAM(test_restart_gc,				\
420 		"Test restarting mark and sweep gc when bucket gens change")
421 
422 #define BCH_DEBUG_PARAMS_ALL() BCH_DEBUG_PARAMS_ALWAYS() BCH_DEBUG_PARAMS_DEBUG()
423 
424 #ifdef CONFIG_BCACHEFS_DEBUG
425 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALL()
426 #else
427 #define BCH_DEBUG_PARAMS() BCH_DEBUG_PARAMS_ALWAYS()
428 #endif
429 
430 #define BCH_DEBUG_PARAM(name, description) extern bool bch2_##name;
431 BCH_DEBUG_PARAMS()
432 #undef BCH_DEBUG_PARAM
433 
434 #ifndef CONFIG_BCACHEFS_DEBUG
435 #define BCH_DEBUG_PARAM(name, description) static const __maybe_unused bool bch2_##name;
436 BCH_DEBUG_PARAMS_DEBUG()
437 #undef BCH_DEBUG_PARAM
438 #endif
439 
440 #define BCH_TIME_STATS()			\
441 	x(btree_node_mem_alloc)			\
442 	x(btree_node_split)			\
443 	x(btree_node_compact)			\
444 	x(btree_node_merge)			\
445 	x(btree_node_sort)			\
446 	x(btree_node_read)			\
447 	x(btree_node_read_done)			\
448 	x(btree_node_write)			\
449 	x(btree_interior_update_foreground)	\
450 	x(btree_interior_update_total)		\
451 	x(btree_gc)				\
452 	x(data_write)				\
453 	x(data_read)				\
454 	x(data_promote)				\
455 	x(journal_flush_write)			\
456 	x(journal_noflush_write)		\
457 	x(journal_flush_seq)			\
458 	x(blocked_journal_low_on_space)		\
459 	x(blocked_journal_low_on_pin)		\
460 	x(blocked_journal_max_in_flight)	\
461 	x(blocked_journal_max_open)		\
462 	x(blocked_key_cache_flush)		\
463 	x(blocked_allocate)			\
464 	x(blocked_allocate_open_bucket)		\
465 	x(blocked_write_buffer_full)		\
466 	x(nocow_lock_contended)
467 
468 enum bch_time_stats {
469 #define x(name) BCH_TIME_##name,
470 	BCH_TIME_STATS()
471 #undef x
472 	BCH_TIME_STAT_NR
473 };
474 
475 #include "alloc_types.h"
476 #include "btree_gc_types.h"
477 #include "btree_types.h"
478 #include "btree_node_scan_types.h"
479 #include "btree_write_buffer_types.h"
480 #include "buckets_types.h"
481 #include "buckets_waiting_for_journal_types.h"
482 #include "clock_types.h"
483 #include "disk_groups_types.h"
484 #include "ec_types.h"
485 #include "journal_types.h"
486 #include "keylist_types.h"
487 #include "quota_types.h"
488 #include "rebalance_types.h"
489 #include "replicas_types.h"
490 #include "sb-members_types.h"
491 #include "subvolume_types.h"
492 #include "super_types.h"
493 #include "thread_with_file_types.h"
494 
495 /* Number of nodes btree coalesce will try to coalesce at once */
496 #define GC_MERGE_NODES		4U
497 
498 /* Maximum number of nodes we might need to allocate atomically: */
499 #define BTREE_RESERVE_MAX	(BTREE_MAX_DEPTH + (BTREE_MAX_DEPTH - 1))
500 
501 /* Size of the freelist we allocate btree nodes from: */
502 #define BTREE_NODE_RESERVE	(BTREE_RESERVE_MAX * 4)
503 
504 #define BTREE_NODE_OPEN_BUCKET_RESERVE	(BTREE_RESERVE_MAX * BCH_REPLICAS_MAX)
505 
506 struct btree;
507 
508 struct io_count {
509 	u64			sectors[2][BCH_DATA_NR];
510 };
511 
512 struct discard_in_flight {
513 	bool			in_progress:1;
514 	u64			bucket:63;
515 };
516 
517 struct bch_dev {
518 	struct kobject		kobj;
519 #ifdef CONFIG_BCACHEFS_DEBUG
520 	atomic_long_t		ref;
521 	bool			dying;
522 	unsigned long		last_put;
523 #else
524 	struct percpu_ref	ref;
525 #endif
526 	struct completion	ref_completion;
527 	struct percpu_ref	io_ref;
528 	struct completion	io_ref_completion;
529 
530 	struct bch_fs		*fs;
531 
532 	u8			dev_idx;
533 	/*
534 	 * Cached version of this device's member info from superblock
535 	 * Committed by bch2_write_super() -> bch_fs_mi_update()
536 	 */
537 	struct bch_member_cpu	mi;
538 	atomic64_t		errors[BCH_MEMBER_ERROR_NR];
539 	unsigned long		write_errors_start;
540 
541 	__uuid_t		uuid;
542 	char			name[BDEVNAME_SIZE];
543 
544 	struct bch_sb_handle	disk_sb;
545 	struct bch_sb		*sb_read_scratch;
546 	int			sb_write_error;
547 	dev_t			dev;
548 	atomic_t		flush_seq;
549 
550 	struct bch_devs_mask	self;
551 
552 	/*
553 	 * Buckets:
554 	 * Per-bucket arrays are protected by either rcu_read_lock or
555 	 * state_lock, for device resize.
556 	 */
557 	GENRADIX(struct bucket)	buckets_gc;
558 	struct bucket_gens __rcu *bucket_gens;
559 	u8			*oldest_gen;
560 	unsigned long		*buckets_nouse;
561 
562 	unsigned long		*bucket_backpointer_mismatches;
563 	unsigned long		*bucket_backpointer_empty;
564 
565 	struct bch_dev_usage __percpu	*usage;
566 
567 	/* Allocator: */
568 	u64			alloc_cursor[3];
569 
570 	unsigned		nr_open_buckets;
571 	unsigned		nr_partial_buckets;
572 	unsigned		nr_btree_reserve;
573 
574 	size_t			inc_gen_needs_gc;
575 	size_t			inc_gen_really_needs_gc;
576 	size_t			buckets_waiting_on_journal;
577 
578 	struct work_struct	invalidate_work;
579 	struct work_struct	discard_work;
580 	struct mutex		discard_buckets_in_flight_lock;
581 	DARRAY(struct discard_in_flight)	discard_buckets_in_flight;
582 	struct work_struct	discard_fast_work;
583 
584 	atomic64_t		rebalance_work;
585 
586 	struct journal_device	journal;
587 	u64			prev_journal_sector;
588 
589 	struct work_struct	io_error_work;
590 
591 	/* The rest of this all shows up in sysfs */
592 	atomic64_t		cur_latency[2];
593 	struct bch2_time_stats_quantiles io_latency[2];
594 
595 #define CONGESTED_MAX		1024
596 	atomic_t		congested;
597 	u64			congested_last;
598 
599 	struct io_count __percpu *io_done;
600 };
601 
602 /*
603  * initial_gc_unfixed
604  * error
605  * topology error
606  */
607 
608 #define BCH_FS_FLAGS()			\
609 	x(new_fs)			\
610 	x(started)			\
611 	x(clean_recovery)		\
612 	x(btree_running)		\
613 	x(accounting_replay_done)	\
614 	x(may_go_rw)			\
615 	x(rw)				\
616 	x(was_rw)			\
617 	x(stopping)			\
618 	x(emergency_ro)			\
619 	x(going_ro)			\
620 	x(write_disable_complete)	\
621 	x(clean_shutdown)		\
622 	x(recovery_running)		\
623 	x(fsck_running)			\
624 	x(initial_gc_unfixed)		\
625 	x(need_delete_dead_snapshots)	\
626 	x(error)			\
627 	x(topology_error)		\
628 	x(errors_fixed)			\
629 	x(errors_not_fixed)		\
630 	x(no_invalid_checks)		\
631 	x(discard_mount_opt_set)	\
632 
633 enum bch_fs_flags {
634 #define x(n)		BCH_FS_##n,
635 	BCH_FS_FLAGS()
636 #undef x
637 };
638 
639 struct btree_debug {
640 	unsigned		id;
641 };
642 
643 #define BCH_TRANSACTIONS_NR 128
644 
645 struct btree_transaction_stats {
646 	struct bch2_time_stats	duration;
647 	struct bch2_time_stats	lock_hold_times;
648 	struct mutex		lock;
649 	unsigned		nr_max_paths;
650 	unsigned		journal_entries_size;
651 	unsigned		max_mem;
652 	char			*max_paths_text;
653 };
654 
655 struct bch_fs_pcpu {
656 	u64			sectors_available;
657 };
658 
659 struct journal_seq_blacklist_table {
660 	size_t			nr;
661 	struct journal_seq_blacklist_table_entry {
662 		u64		start;
663 		u64		end;
664 		bool		dirty;
665 	}			entries[];
666 };
667 
668 struct btree_trans_buf {
669 	struct btree_trans	*trans;
670 };
671 
672 #define BCACHEFS_ROOT_SUBVOL_INUM					\
673 	((subvol_inum) { BCACHEFS_ROOT_SUBVOL,	BCACHEFS_ROOT_INO })
674 
675 #define BCH_WRITE_REFS()						\
676 	x(journal)							\
677 	x(trans)							\
678 	x(write)							\
679 	x(promote)							\
680 	x(node_rewrite)							\
681 	x(stripe_create)						\
682 	x(stripe_delete)						\
683 	x(reflink)							\
684 	x(fallocate)							\
685 	x(fsync)							\
686 	x(dio_write)							\
687 	x(discard)							\
688 	x(discard_fast)							\
689 	x(check_discard_freespace_key)					\
690 	x(invalidate)							\
691 	x(delete_dead_snapshots)					\
692 	x(gc_gens)							\
693 	x(snapshot_delete_pagecache)					\
694 	x(sysfs)							\
695 	x(btree_write_buffer)						\
696 	x(btree_node_scrub)
697 
698 enum bch_write_ref {
699 #define x(n) BCH_WRITE_REF_##n,
700 	BCH_WRITE_REFS()
701 #undef x
702 	BCH_WRITE_REF_NR,
703 };
704 
705 #define BCH_FS_DEFAULT_UTF8_ENCODING UNICODE_AGE(12, 1, 0)
706 
707 struct bch_fs {
708 	struct closure		cl;
709 
710 	struct list_head	list;
711 	struct kobject		kobj;
712 	struct kobject		counters_kobj;
713 	struct kobject		internal;
714 	struct kobject		opts_dir;
715 	struct kobject		time_stats;
716 	unsigned long		flags;
717 
718 	int			minor;
719 	struct device		*chardev;
720 	struct super_block	*vfs_sb;
721 	dev_t			dev;
722 	char			name[40];
723 	struct stdio_redirect	*stdio;
724 	struct task_struct	*stdio_filter;
725 
726 	/* ro/rw, add/remove/resize devices: */
727 	struct rw_semaphore	state_lock;
728 
729 	/* Counts outstanding writes, for clean transition to read-only */
730 #ifdef BCH_WRITE_REF_DEBUG
731 	atomic_long_t		writes[BCH_WRITE_REF_NR];
732 #else
733 	struct percpu_ref	writes;
734 #endif
735 	/*
736 	 * Certain operations are only allowed in single threaded mode, during
737 	 * recovery, and we want to assert that this is the case:
738 	 */
739 	struct task_struct	*recovery_task;
740 
741 	/*
742 	 * Analagous to c->writes, for asynchronous ops that don't necessarily
743 	 * need fs to be read-write
744 	 */
745 	refcount_t		ro_ref;
746 	wait_queue_head_t	ro_ref_wait;
747 
748 	struct work_struct	read_only_work;
749 
750 	struct bch_dev __rcu	*devs[BCH_SB_MEMBERS_MAX];
751 
752 	struct bch_accounting_mem accounting;
753 
754 	struct bch_replicas_cpu replicas;
755 	struct bch_replicas_cpu replicas_gc;
756 	struct mutex		replicas_gc_lock;
757 
758 	struct journal_entry_res btree_root_journal_res;
759 	struct journal_entry_res clock_journal_res;
760 
761 	struct bch_disk_groups_cpu __rcu *disk_groups;
762 
763 	struct bch_opts		opts;
764 
765 	/* Updated by bch2_sb_update():*/
766 	struct {
767 		__uuid_t	uuid;
768 		__uuid_t	user_uuid;
769 
770 		u16		version;
771 		u16		version_incompat;
772 		u16		version_incompat_allowed;
773 		u16		version_min;
774 		u16		version_upgrade_complete;
775 
776 		u8		nr_devices;
777 		u8		clean;
778 
779 		u8		encryption_type;
780 
781 		u64		time_base_lo;
782 		u32		time_base_hi;
783 		unsigned	time_units_per_sec;
784 		unsigned	nsec_per_time_unit;
785 		u64		features;
786 		u64		compat;
787 		unsigned long	errors_silent[BITS_TO_LONGS(BCH_FSCK_ERR_MAX)];
788 		u64		btrees_lost_data;
789 	}			sb;
790 
791 #ifdef CONFIG_UNICODE
792 	struct unicode_map	*cf_encoding;
793 #endif
794 
795 	struct bch_sb_handle	disk_sb;
796 
797 	unsigned short		block_bits;	/* ilog2(block_size) */
798 
799 	u16			btree_foreground_merge_threshold;
800 
801 	struct closure		sb_write;
802 	struct mutex		sb_lock;
803 
804 	/* snapshot.c: */
805 	struct snapshot_table __rcu *snapshots;
806 	struct mutex		snapshot_table_lock;
807 	struct rw_semaphore	snapshot_create_lock;
808 
809 	struct work_struct	snapshot_delete_work;
810 	struct work_struct	snapshot_wait_for_pagecache_and_delete_work;
811 	snapshot_id_list	snapshots_unlinked;
812 	struct mutex		snapshots_unlinked_lock;
813 
814 	/* BTREE CACHE */
815 	struct bio_set		btree_bio;
816 	struct workqueue_struct	*btree_read_complete_wq;
817 	struct workqueue_struct	*btree_write_submit_wq;
818 
819 	struct btree_root	btree_roots_known[BTREE_ID_NR];
820 	DARRAY(struct btree_root) btree_roots_extra;
821 	struct mutex		btree_root_lock;
822 
823 	struct btree_cache	btree_cache;
824 
825 	/*
826 	 * Cache of allocated btree nodes - if we allocate a btree node and
827 	 * don't use it, if we free it that space can't be reused until going
828 	 * _all_ the way through the allocator (which exposes us to a livelock
829 	 * when allocating btree reserves fail halfway through) - instead, we
830 	 * can stick them here:
831 	 */
832 	struct btree_alloc	btree_reserve_cache[BTREE_NODE_RESERVE * 2];
833 	unsigned		btree_reserve_cache_nr;
834 	struct mutex		btree_reserve_cache_lock;
835 
836 	mempool_t		btree_interior_update_pool;
837 	struct list_head	btree_interior_update_list;
838 	struct list_head	btree_interior_updates_unwritten;
839 	struct mutex		btree_interior_update_lock;
840 	struct closure_waitlist	btree_interior_update_wait;
841 
842 	struct workqueue_struct	*btree_interior_update_worker;
843 	struct work_struct	btree_interior_update_work;
844 
845 	struct workqueue_struct	*btree_node_rewrite_worker;
846 	struct list_head	btree_node_rewrites;
847 	struct list_head	btree_node_rewrites_pending;
848 	spinlock_t		btree_node_rewrites_lock;
849 	struct closure_waitlist	btree_node_rewrites_wait;
850 
851 	/* btree_io.c: */
852 	spinlock_t		btree_write_error_lock;
853 	struct btree_write_stats {
854 		atomic64_t	nr;
855 		atomic64_t	bytes;
856 	}			btree_write_stats[BTREE_WRITE_TYPE_NR];
857 
858 	/* btree_iter.c: */
859 	struct seqmutex		btree_trans_lock;
860 	struct list_head	btree_trans_list;
861 	mempool_t		btree_trans_pool;
862 	mempool_t		btree_trans_mem_pool;
863 	struct btree_trans_buf  __percpu	*btree_trans_bufs;
864 
865 	struct srcu_struct	btree_trans_barrier;
866 	bool			btree_trans_barrier_initialized;
867 
868 	struct btree_key_cache	btree_key_cache;
869 	unsigned		btree_key_cache_btrees;
870 
871 	struct btree_write_buffer btree_write_buffer;
872 
873 	struct workqueue_struct	*btree_update_wq;
874 	struct workqueue_struct	*btree_io_complete_wq;
875 	/* copygc needs its own workqueue for index updates.. */
876 	struct workqueue_struct	*copygc_wq;
877 	/*
878 	 * Use a dedicated wq for write ref holder tasks. Required to avoid
879 	 * dependency problems with other wq tasks that can block on ref
880 	 * draining, such as read-only transition.
881 	 */
882 	struct workqueue_struct *write_ref_wq;
883 
884 	/* ALLOCATION */
885 	struct bch_devs_mask	rw_devs[BCH_DATA_NR];
886 	unsigned long		rw_devs_change_count;
887 
888 	u64			capacity; /* sectors */
889 	u64			reserved; /* sectors */
890 
891 	/*
892 	 * When capacity _decreases_ (due to a disk being removed), we
893 	 * increment capacity_gen - this invalidates outstanding reservations
894 	 * and forces them to be revalidated
895 	 */
896 	u32			capacity_gen;
897 	unsigned		bucket_size_max;
898 
899 	atomic64_t		sectors_available;
900 	struct mutex		sectors_available_lock;
901 
902 	struct bch_fs_pcpu __percpu	*pcpu;
903 
904 	struct percpu_rw_semaphore	mark_lock;
905 
906 	seqcount_t			usage_lock;
907 	struct bch_fs_usage_base __percpu *usage;
908 	u64 __percpu		*online_reserved;
909 
910 	unsigned long		allocator_last_stuck;
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 
918 	/* ALLOCATOR */
919 	spinlock_t		freelist_lock;
920 	struct closure_waitlist	freelist_wait;
921 
922 	open_bucket_idx_t	open_buckets_freelist;
923 	open_bucket_idx_t	open_buckets_nr_free;
924 	struct closure_waitlist	open_buckets_wait;
925 	struct open_bucket	open_buckets[OPEN_BUCKETS_COUNT];
926 	open_bucket_idx_t	open_buckets_hash[OPEN_BUCKETS_COUNT];
927 
928 	open_bucket_idx_t	open_buckets_partial[OPEN_BUCKETS_COUNT];
929 	open_bucket_idx_t	open_buckets_partial_nr;
930 
931 	struct write_point	btree_write_point;
932 	struct write_point	rebalance_write_point;
933 
934 	struct write_point	write_points[WRITE_POINT_MAX];
935 	struct hlist_head	write_points_hash[WRITE_POINT_HASH_NR];
936 	struct mutex		write_points_hash_lock;
937 	unsigned		write_points_nr;
938 
939 	struct buckets_waiting_for_journal buckets_waiting_for_journal;
940 
941 	/* GARBAGE COLLECTION */
942 	struct work_struct	gc_gens_work;
943 	unsigned long		gc_count;
944 
945 	enum btree_id		gc_gens_btree;
946 	struct bpos		gc_gens_pos;
947 
948 	/*
949 	 * Tracks GC's progress - everything in the range [ZERO_KEY..gc_cur_pos]
950 	 * has been marked by GC.
951 	 *
952 	 * gc_cur_phase is a superset of btree_ids (BTREE_ID_extents etc.)
953 	 *
954 	 * Protected by gc_pos_lock. Only written to by GC thread, so GC thread
955 	 * can read without a lock.
956 	 */
957 	seqcount_t		gc_pos_lock;
958 	struct gc_pos		gc_pos;
959 
960 	/*
961 	 * The allocation code needs gc_mark in struct bucket to be correct, but
962 	 * it's not while a gc is in progress.
963 	 */
964 	struct rw_semaphore	gc_lock;
965 	struct mutex		gc_gens_lock;
966 
967 	/* IO PATH */
968 	struct semaphore	io_in_flight;
969 	struct bio_set		bio_read;
970 	struct bio_set		bio_read_split;
971 	struct bio_set		bio_write;
972 	struct bio_set		replica_set;
973 	struct mutex		bio_bounce_pages_lock;
974 	mempool_t		bio_bounce_pages;
975 	struct bucket_nocow_lock_table
976 				nocow_locks;
977 	struct rhashtable	promote_table;
978 
979 	mempool_t		compression_bounce[2];
980 	mempool_t		compress_workspace[BCH_COMPRESSION_OPT_NR];
981 	size_t			zstd_workspace_size;
982 
983 	struct crypto_sync_skcipher *chacha20;
984 	struct crypto_shash	*poly1305;
985 
986 	atomic64_t		key_version;
987 
988 	mempool_t		large_bkey_pool;
989 
990 	/* MOVE.C */
991 	struct list_head	moving_context_list;
992 	struct mutex		moving_context_lock;
993 
994 	/* REBALANCE */
995 	struct bch_fs_rebalance	rebalance;
996 
997 	/* COPYGC */
998 	struct task_struct	*copygc_thread;
999 	struct write_point	copygc_write_point;
1000 	s64			copygc_wait_at;
1001 	s64			copygc_wait;
1002 	bool			copygc_running;
1003 	wait_queue_head_t	copygc_running_wq;
1004 
1005 	/* STRIPES: */
1006 	GENRADIX(struct gc_stripe) gc_stripes;
1007 
1008 	struct hlist_head	ec_stripes_new[32];
1009 	spinlock_t		ec_stripes_new_lock;
1010 
1011 	/* ERASURE CODING */
1012 	struct list_head	ec_stripe_head_list;
1013 	struct mutex		ec_stripe_head_lock;
1014 
1015 	struct list_head	ec_stripe_new_list;
1016 	struct mutex		ec_stripe_new_lock;
1017 	wait_queue_head_t	ec_stripe_new_wait;
1018 
1019 	struct work_struct	ec_stripe_create_work;
1020 	u64			ec_stripe_hint;
1021 
1022 	struct work_struct	ec_stripe_delete_work;
1023 
1024 	struct bio_set		ec_bioset;
1025 
1026 	/* REFLINK */
1027 	reflink_gc_table	reflink_gc_table;
1028 	size_t			reflink_gc_nr;
1029 
1030 	/* fs.c */
1031 	struct list_head	vfs_inodes_list;
1032 	struct mutex		vfs_inodes_lock;
1033 	struct rhashtable	vfs_inodes_table;
1034 	struct rhltable		vfs_inodes_by_inum_table;
1035 
1036 	/* VFS IO PATH - fs-io.c */
1037 	struct bio_set		writepage_bioset;
1038 	struct bio_set		dio_write_bioset;
1039 	struct bio_set		dio_read_bioset;
1040 	struct bio_set		nocow_flush_bioset;
1041 
1042 	/* QUOTAS */
1043 	struct bch_memquota_type quotas[QTYP_NR];
1044 
1045 	/* RECOVERY */
1046 	u64			journal_replay_seq_start;
1047 	u64			journal_replay_seq_end;
1048 	/*
1049 	 * Two different uses:
1050 	 * "Has this fsck pass?" - i.e. should this type of error be an
1051 	 * emergency read-only
1052 	 * And, in certain situations fsck will rewind to an earlier pass: used
1053 	 * for signaling to the toplevel code which pass we want to run now.
1054 	 */
1055 	enum bch_recovery_pass	curr_recovery_pass;
1056 	enum bch_recovery_pass	next_recovery_pass;
1057 	/* bitmask of recovery passes that we actually ran */
1058 	u64			recovery_passes_complete;
1059 	/* never rewinds version of curr_recovery_pass */
1060 	enum bch_recovery_pass	recovery_pass_done;
1061 	spinlock_t		recovery_pass_lock;
1062 	struct semaphore	online_fsck_mutex;
1063 
1064 	/* DEBUG JUNK */
1065 	struct dentry		*fs_debug_dir;
1066 	struct dentry		*btree_debug_dir;
1067 	struct btree_debug	btree_debug[BTREE_ID_NR];
1068 	struct btree		*verify_data;
1069 	struct btree_node	*verify_ondisk;
1070 	struct mutex		verify_lock;
1071 
1072 	/*
1073 	 * A btree node on disk could have too many bsets for an iterator to fit
1074 	 * on the stack - have to dynamically allocate them
1075 	 */
1076 	mempool_t		fill_iter;
1077 
1078 	mempool_t		btree_bounce_pool;
1079 
1080 	struct journal		journal;
1081 	GENRADIX(struct journal_replay *) journal_entries;
1082 	u64			journal_entries_base_seq;
1083 	struct journal_keys	journal_keys;
1084 	struct list_head	journal_iters;
1085 
1086 	struct find_btree_nodes	found_btree_nodes;
1087 
1088 	u64			last_bucket_seq_cleanup;
1089 
1090 	u64			counters_on_mount[BCH_COUNTER_NR];
1091 	u64 __percpu		*counters;
1092 
1093 	struct bch2_time_stats	times[BCH_TIME_STAT_NR];
1094 
1095 	struct btree_transaction_stats btree_transaction_stats[BCH_TRANSACTIONS_NR];
1096 
1097 	/* ERRORS */
1098 	struct list_head	fsck_error_msgs;
1099 	struct mutex		fsck_error_msgs_lock;
1100 	bool			fsck_alloc_msgs_err;
1101 
1102 	bch_sb_errors_cpu	fsck_error_counts;
1103 	struct mutex		fsck_error_counts_lock;
1104 };
1105 
1106 extern struct wait_queue_head bch2_read_only_wait;
1107 
bch2_write_ref_get(struct bch_fs * c,enum bch_write_ref ref)1108 static inline void bch2_write_ref_get(struct bch_fs *c, enum bch_write_ref ref)
1109 {
1110 #ifdef BCH_WRITE_REF_DEBUG
1111 	atomic_long_inc(&c->writes[ref]);
1112 #else
1113 	percpu_ref_get(&c->writes);
1114 #endif
1115 }
1116 
__bch2_write_ref_tryget(struct bch_fs * c,enum bch_write_ref ref)1117 static inline bool __bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1118 {
1119 #ifdef BCH_WRITE_REF_DEBUG
1120 	return !test_bit(BCH_FS_going_ro, &c->flags) &&
1121 		atomic_long_inc_not_zero(&c->writes[ref]);
1122 #else
1123 	return percpu_ref_tryget(&c->writes);
1124 #endif
1125 }
1126 
bch2_write_ref_tryget(struct bch_fs * c,enum bch_write_ref ref)1127 static inline bool bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
1128 {
1129 #ifdef BCH_WRITE_REF_DEBUG
1130 	return !test_bit(BCH_FS_going_ro, &c->flags) &&
1131 		atomic_long_inc_not_zero(&c->writes[ref]);
1132 #else
1133 	return percpu_ref_tryget_live(&c->writes);
1134 #endif
1135 }
1136 
bch2_write_ref_put(struct bch_fs * c,enum bch_write_ref ref)1137 static inline void bch2_write_ref_put(struct bch_fs *c, enum bch_write_ref ref)
1138 {
1139 #ifdef BCH_WRITE_REF_DEBUG
1140 	long v = atomic_long_dec_return(&c->writes[ref]);
1141 
1142 	BUG_ON(v < 0);
1143 	if (v)
1144 		return;
1145 	for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++)
1146 		if (atomic_long_read(&c->writes[i]))
1147 			return;
1148 
1149 	set_bit(BCH_FS_write_disable_complete, &c->flags);
1150 	wake_up(&bch2_read_only_wait);
1151 #else
1152 	percpu_ref_put(&c->writes);
1153 #endif
1154 }
1155 
bch2_ro_ref_tryget(struct bch_fs * c)1156 static inline bool bch2_ro_ref_tryget(struct bch_fs *c)
1157 {
1158 	if (test_bit(BCH_FS_stopping, &c->flags))
1159 		return false;
1160 
1161 	return refcount_inc_not_zero(&c->ro_ref);
1162 }
1163 
bch2_ro_ref_put(struct bch_fs * c)1164 static inline void bch2_ro_ref_put(struct bch_fs *c)
1165 {
1166 	if (refcount_dec_and_test(&c->ro_ref))
1167 		wake_up(&c->ro_ref_wait);
1168 }
1169 
bch2_set_ra_pages(struct bch_fs * c,unsigned ra_pages)1170 static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
1171 {
1172 #ifndef NO_BCACHEFS_FS
1173 	if (c->vfs_sb)
1174 		c->vfs_sb->s_bdi->ra_pages = ra_pages;
1175 #endif
1176 }
1177 
bucket_bytes(const struct bch_dev * ca)1178 static inline unsigned bucket_bytes(const struct bch_dev *ca)
1179 {
1180 	return ca->mi.bucket_size << 9;
1181 }
1182 
block_bytes(const struct bch_fs * c)1183 static inline unsigned block_bytes(const struct bch_fs *c)
1184 {
1185 	return c->opts.block_size;
1186 }
1187 
block_sectors(const struct bch_fs * c)1188 static inline unsigned block_sectors(const struct bch_fs *c)
1189 {
1190 	return c->opts.block_size >> 9;
1191 }
1192 
btree_id_cached(const struct bch_fs * c,enum btree_id btree)1193 static inline bool btree_id_cached(const struct bch_fs *c, enum btree_id btree)
1194 {
1195 	return c->btree_key_cache_btrees & (1U << btree);
1196 }
1197 
bch2_time_to_timespec(const struct bch_fs * c,s64 time)1198 static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
1199 {
1200 	struct timespec64 t;
1201 	s64 sec;
1202 	s32 rem;
1203 
1204 	time += c->sb.time_base_lo;
1205 
1206 	sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
1207 
1208 	set_normalized_timespec64(&t, sec, rem * (s64)c->sb.nsec_per_time_unit);
1209 
1210 	return t;
1211 }
1212 
timespec_to_bch2_time(const struct bch_fs * c,struct timespec64 ts)1213 static inline s64 timespec_to_bch2_time(const struct bch_fs *c, struct timespec64 ts)
1214 {
1215 	return (ts.tv_sec * c->sb.time_units_per_sec +
1216 		(int) ts.tv_nsec / c->sb.nsec_per_time_unit) - c->sb.time_base_lo;
1217 }
1218 
bch2_current_time(const struct bch_fs * c)1219 static inline s64 bch2_current_time(const struct bch_fs *c)
1220 {
1221 	struct timespec64 now;
1222 
1223 	ktime_get_coarse_real_ts64(&now);
1224 	return timespec_to_bch2_time(c, now);
1225 }
1226 
bch2_current_io_time(const struct bch_fs * c,int rw)1227 static inline u64 bch2_current_io_time(const struct bch_fs *c, int rw)
1228 {
1229 	return max(1ULL, (u64) atomic64_read(&c->io_clock[rw].now) & LRU_TIME_MAX);
1230 }
1231 
bch2_fs_stdio_redirect(struct bch_fs * c)1232 static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
1233 {
1234 	struct stdio_redirect *stdio = c->stdio;
1235 
1236 	if (c->stdio_filter && c->stdio_filter != current)
1237 		stdio = NULL;
1238 	return stdio;
1239 }
1240 
metadata_replicas_required(struct bch_fs * c)1241 static inline unsigned metadata_replicas_required(struct bch_fs *c)
1242 {
1243 	return min(c->opts.metadata_replicas,
1244 		   c->opts.metadata_replicas_required);
1245 }
1246 
data_replicas_required(struct bch_fs * c)1247 static inline unsigned data_replicas_required(struct bch_fs *c)
1248 {
1249 	return min(c->opts.data_replicas,
1250 		   c->opts.data_replicas_required);
1251 }
1252 
1253 #define BKEY_PADDED_ONSTACK(key, pad)				\
1254 	struct { struct bkey_i key; __u64 key ## _pad[pad]; }
1255 
1256 #endif /* _BCACHEFS_H */
1257