xref: /linux/fs/bcachefs/btree_key_cache.c (revision d642ef71)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include "bcachefs.h"
4 #include "btree_cache.h"
5 #include "btree_iter.h"
6 #include "btree_key_cache.h"
7 #include "btree_locking.h"
8 #include "btree_update.h"
9 #include "errcode.h"
10 #include "error.h"
11 #include "journal.h"
12 #include "journal_reclaim.h"
13 #include "trace.h"
14 
15 #include <linux/sched/mm.h>
16 
17 static inline bool btree_uses_pcpu_readers(enum btree_id id)
18 {
19 	return id == BTREE_ID_subvolumes;
20 }
21 
22 static struct kmem_cache *bch2_key_cache;
23 
24 static int bch2_btree_key_cache_cmp_fn(struct rhashtable_compare_arg *arg,
25 				       const void *obj)
26 {
27 	const struct bkey_cached *ck = obj;
28 	const struct bkey_cached_key *key = arg->key;
29 
30 	return ck->key.btree_id != key->btree_id ||
31 		!bpos_eq(ck->key.pos, key->pos);
32 }
33 
34 static const struct rhashtable_params bch2_btree_key_cache_params = {
35 	.head_offset	= offsetof(struct bkey_cached, hash),
36 	.key_offset	= offsetof(struct bkey_cached, key),
37 	.key_len	= sizeof(struct bkey_cached_key),
38 	.obj_cmpfn	= bch2_btree_key_cache_cmp_fn,
39 };
40 
41 __flatten
42 inline struct bkey_cached *
43 bch2_btree_key_cache_find(struct bch_fs *c, enum btree_id btree_id, struct bpos pos)
44 {
45 	struct bkey_cached_key key = {
46 		.btree_id	= btree_id,
47 		.pos		= pos,
48 	};
49 
50 	return rhashtable_lookup_fast(&c->btree_key_cache.table, &key,
51 				      bch2_btree_key_cache_params);
52 }
53 
54 static bool bkey_cached_lock_for_evict(struct bkey_cached *ck)
55 {
56 	if (!six_trylock_intent(&ck->c.lock))
57 		return false;
58 
59 	if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
60 		six_unlock_intent(&ck->c.lock);
61 		return false;
62 	}
63 
64 	if (!six_trylock_write(&ck->c.lock)) {
65 		six_unlock_intent(&ck->c.lock);
66 		return false;
67 	}
68 
69 	return true;
70 }
71 
72 static void bkey_cached_evict(struct btree_key_cache *c,
73 			      struct bkey_cached *ck)
74 {
75 	BUG_ON(rhashtable_remove_fast(&c->table, &ck->hash,
76 				      bch2_btree_key_cache_params));
77 	memset(&ck->key, ~0, sizeof(ck->key));
78 
79 	atomic_long_dec(&c->nr_keys);
80 }
81 
82 static void bkey_cached_free(struct btree_key_cache *bc,
83 			     struct bkey_cached *ck)
84 {
85 	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
86 
87 	BUG_ON(test_bit(BKEY_CACHED_DIRTY, &ck->flags));
88 
89 	ck->btree_trans_barrier_seq =
90 		start_poll_synchronize_srcu(&c->btree_trans_barrier);
91 
92 	if (ck->c.lock.readers) {
93 		list_move_tail(&ck->list, &bc->freed_pcpu);
94 		bc->nr_freed_pcpu++;
95 	} else {
96 		list_move_tail(&ck->list, &bc->freed_nonpcpu);
97 		bc->nr_freed_nonpcpu++;
98 	}
99 	atomic_long_inc(&bc->nr_freed);
100 
101 	kfree(ck->k);
102 	ck->k		= NULL;
103 	ck->u64s	= 0;
104 
105 	six_unlock_write(&ck->c.lock);
106 	six_unlock_intent(&ck->c.lock);
107 }
108 
109 #ifdef __KERNEL__
110 static void __bkey_cached_move_to_freelist_ordered(struct btree_key_cache *bc,
111 						   struct bkey_cached *ck)
112 {
113 	struct bkey_cached *pos;
114 
115 	bc->nr_freed_nonpcpu++;
116 
117 	list_for_each_entry_reverse(pos, &bc->freed_nonpcpu, list) {
118 		if (ULONG_CMP_GE(ck->btree_trans_barrier_seq,
119 				 pos->btree_trans_barrier_seq)) {
120 			list_move(&ck->list, &pos->list);
121 			return;
122 		}
123 	}
124 
125 	list_move(&ck->list, &bc->freed_nonpcpu);
126 }
127 #endif
128 
129 static void bkey_cached_move_to_freelist(struct btree_key_cache *bc,
130 					 struct bkey_cached *ck)
131 {
132 	BUG_ON(test_bit(BKEY_CACHED_DIRTY, &ck->flags));
133 
134 	if (!ck->c.lock.readers) {
135 #ifdef __KERNEL__
136 		struct btree_key_cache_freelist *f;
137 		bool freed = false;
138 
139 		preempt_disable();
140 		f = this_cpu_ptr(bc->pcpu_freed);
141 
142 		if (f->nr < ARRAY_SIZE(f->objs)) {
143 			f->objs[f->nr++] = ck;
144 			freed = true;
145 		}
146 		preempt_enable();
147 
148 		if (!freed) {
149 			mutex_lock(&bc->lock);
150 			preempt_disable();
151 			f = this_cpu_ptr(bc->pcpu_freed);
152 
153 			while (f->nr > ARRAY_SIZE(f->objs) / 2) {
154 				struct bkey_cached *ck2 = f->objs[--f->nr];
155 
156 				__bkey_cached_move_to_freelist_ordered(bc, ck2);
157 			}
158 			preempt_enable();
159 
160 			__bkey_cached_move_to_freelist_ordered(bc, ck);
161 			mutex_unlock(&bc->lock);
162 		}
163 #else
164 		mutex_lock(&bc->lock);
165 		list_move_tail(&ck->list, &bc->freed_nonpcpu);
166 		bc->nr_freed_nonpcpu++;
167 		mutex_unlock(&bc->lock);
168 #endif
169 	} else {
170 		mutex_lock(&bc->lock);
171 		list_move_tail(&ck->list, &bc->freed_pcpu);
172 		mutex_unlock(&bc->lock);
173 	}
174 }
175 
176 static void bkey_cached_free_fast(struct btree_key_cache *bc,
177 				  struct bkey_cached *ck)
178 {
179 	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
180 
181 	ck->btree_trans_barrier_seq =
182 		start_poll_synchronize_srcu(&c->btree_trans_barrier);
183 
184 	list_del_init(&ck->list);
185 	atomic_long_inc(&bc->nr_freed);
186 
187 	kfree(ck->k);
188 	ck->k		= NULL;
189 	ck->u64s	= 0;
190 
191 	bkey_cached_move_to_freelist(bc, ck);
192 
193 	six_unlock_write(&ck->c.lock);
194 	six_unlock_intent(&ck->c.lock);
195 }
196 
197 static struct bkey_cached *
198 bkey_cached_alloc(struct btree_trans *trans, struct btree_path *path,
199 		  bool *was_new)
200 {
201 	struct bch_fs *c = trans->c;
202 	struct btree_key_cache *bc = &c->btree_key_cache;
203 	struct bkey_cached *ck = NULL;
204 	bool pcpu_readers = btree_uses_pcpu_readers(path->btree_id);
205 	int ret;
206 
207 	if (!pcpu_readers) {
208 #ifdef __KERNEL__
209 		struct btree_key_cache_freelist *f;
210 
211 		preempt_disable();
212 		f = this_cpu_ptr(bc->pcpu_freed);
213 		if (f->nr)
214 			ck = f->objs[--f->nr];
215 		preempt_enable();
216 
217 		if (!ck) {
218 			mutex_lock(&bc->lock);
219 			preempt_disable();
220 			f = this_cpu_ptr(bc->pcpu_freed);
221 
222 			while (!list_empty(&bc->freed_nonpcpu) &&
223 			       f->nr < ARRAY_SIZE(f->objs) / 2) {
224 				ck = list_last_entry(&bc->freed_nonpcpu, struct bkey_cached, list);
225 				list_del_init(&ck->list);
226 				bc->nr_freed_nonpcpu--;
227 				f->objs[f->nr++] = ck;
228 			}
229 
230 			ck = f->nr ? f->objs[--f->nr] : NULL;
231 			preempt_enable();
232 			mutex_unlock(&bc->lock);
233 		}
234 #else
235 		mutex_lock(&bc->lock);
236 		if (!list_empty(&bc->freed_nonpcpu)) {
237 			ck = list_last_entry(&bc->freed_nonpcpu, struct bkey_cached, list);
238 			list_del_init(&ck->list);
239 			bc->nr_freed_nonpcpu--;
240 		}
241 		mutex_unlock(&bc->lock);
242 #endif
243 	} else {
244 		mutex_lock(&bc->lock);
245 		if (!list_empty(&bc->freed_pcpu)) {
246 			ck = list_last_entry(&bc->freed_pcpu, struct bkey_cached, list);
247 			list_del_init(&ck->list);
248 		}
249 		mutex_unlock(&bc->lock);
250 	}
251 
252 	if (ck) {
253 		ret = btree_node_lock_nopath(trans, &ck->c, SIX_LOCK_intent, _THIS_IP_);
254 		if (unlikely(ret)) {
255 			bkey_cached_move_to_freelist(bc, ck);
256 			return ERR_PTR(ret);
257 		}
258 
259 		path->l[0].b = (void *) ck;
260 		path->l[0].lock_seq = six_lock_seq(&ck->c.lock);
261 		mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
262 
263 		ret = bch2_btree_node_lock_write(trans, path, &ck->c);
264 		if (unlikely(ret)) {
265 			btree_node_unlock(trans, path, 0);
266 			bkey_cached_move_to_freelist(bc, ck);
267 			return ERR_PTR(ret);
268 		}
269 
270 		return ck;
271 	}
272 
273 	ck = allocate_dropping_locks(trans, ret,
274 			kmem_cache_zalloc(bch2_key_cache, _gfp));
275 	if (ret) {
276 		kmem_cache_free(bch2_key_cache, ck);
277 		return ERR_PTR(ret);
278 	}
279 
280 	if (!ck)
281 		return NULL;
282 
283 	INIT_LIST_HEAD(&ck->list);
284 	bch2_btree_lock_init(&ck->c, pcpu_readers ? SIX_LOCK_INIT_PCPU : 0);
285 
286 	ck->c.cached = true;
287 	BUG_ON(!six_trylock_intent(&ck->c.lock));
288 	BUG_ON(!six_trylock_write(&ck->c.lock));
289 	*was_new = true;
290 	return ck;
291 }
292 
293 static struct bkey_cached *
294 bkey_cached_reuse(struct btree_key_cache *c)
295 {
296 	struct bucket_table *tbl;
297 	struct rhash_head *pos;
298 	struct bkey_cached *ck;
299 	unsigned i;
300 
301 	mutex_lock(&c->lock);
302 	rcu_read_lock();
303 	tbl = rht_dereference_rcu(c->table.tbl, &c->table);
304 	for (i = 0; i < tbl->size; i++)
305 		rht_for_each_entry_rcu(ck, pos, tbl, i, hash) {
306 			if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
307 			    bkey_cached_lock_for_evict(ck)) {
308 				bkey_cached_evict(c, ck);
309 				goto out;
310 			}
311 		}
312 	ck = NULL;
313 out:
314 	rcu_read_unlock();
315 	mutex_unlock(&c->lock);
316 	return ck;
317 }
318 
319 static struct bkey_cached *
320 btree_key_cache_create(struct btree_trans *trans, struct btree_path *path)
321 {
322 	struct bch_fs *c = trans->c;
323 	struct btree_key_cache *bc = &c->btree_key_cache;
324 	struct bkey_cached *ck;
325 	bool was_new = false;
326 
327 	ck = bkey_cached_alloc(trans, path, &was_new);
328 	if (IS_ERR(ck))
329 		return ck;
330 
331 	if (unlikely(!ck)) {
332 		ck = bkey_cached_reuse(bc);
333 		if (unlikely(!ck)) {
334 			bch_err(c, "error allocating memory for key cache item, btree %s",
335 				bch2_btree_id_str(path->btree_id));
336 			return ERR_PTR(-BCH_ERR_ENOMEM_btree_key_cache_create);
337 		}
338 
339 		mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
340 	}
341 
342 	ck->c.level		= 0;
343 	ck->c.btree_id		= path->btree_id;
344 	ck->key.btree_id	= path->btree_id;
345 	ck->key.pos		= path->pos;
346 	ck->valid		= false;
347 	ck->flags		= 1U << BKEY_CACHED_ACCESSED;
348 
349 	if (unlikely(rhashtable_lookup_insert_fast(&bc->table,
350 					  &ck->hash,
351 					  bch2_btree_key_cache_params))) {
352 		/* We raced with another fill: */
353 
354 		if (likely(was_new)) {
355 			six_unlock_write(&ck->c.lock);
356 			six_unlock_intent(&ck->c.lock);
357 			kfree(ck);
358 		} else {
359 			bkey_cached_free_fast(bc, ck);
360 		}
361 
362 		mark_btree_node_locked(trans, path, 0, BTREE_NODE_UNLOCKED);
363 		return NULL;
364 	}
365 
366 	atomic_long_inc(&bc->nr_keys);
367 
368 	six_unlock_write(&ck->c.lock);
369 
370 	return ck;
371 }
372 
373 static int btree_key_cache_fill(struct btree_trans *trans,
374 				struct btree_path *ck_path,
375 				struct bkey_cached *ck)
376 {
377 	struct btree_iter iter;
378 	struct bkey_s_c k;
379 	unsigned new_u64s = 0;
380 	struct bkey_i *new_k = NULL;
381 	int ret;
382 
383 	k = bch2_bkey_get_iter(trans, &iter, ck->key.btree_id, ck->key.pos,
384 			       BTREE_ITER_KEY_CACHE_FILL|
385 			       BTREE_ITER_CACHED_NOFILL);
386 	ret = bkey_err(k);
387 	if (ret)
388 		goto err;
389 
390 	if (!bch2_btree_node_relock(trans, ck_path, 0)) {
391 		trace_and_count(trans->c, trans_restart_relock_key_cache_fill, trans, _THIS_IP_, ck_path);
392 		ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_fill);
393 		goto err;
394 	}
395 
396 	/*
397 	 * bch2_varint_decode can read past the end of the buffer by at
398 	 * most 7 bytes (it won't be used):
399 	 */
400 	new_u64s = k.k->u64s + 1;
401 
402 	/*
403 	 * Allocate some extra space so that the transaction commit path is less
404 	 * likely to have to reallocate, since that requires a transaction
405 	 * restart:
406 	 */
407 	new_u64s = min(256U, (new_u64s * 3) / 2);
408 
409 	if (new_u64s > ck->u64s) {
410 		new_u64s = roundup_pow_of_two(new_u64s);
411 		new_k = kmalloc(new_u64s * sizeof(u64), GFP_NOWAIT|__GFP_NOWARN);
412 		if (!new_k) {
413 			bch2_trans_unlock(trans);
414 
415 			new_k = kmalloc(new_u64s * sizeof(u64), GFP_KERNEL);
416 			if (!new_k) {
417 				bch_err(trans->c, "error allocating memory for key cache key, btree %s u64s %u",
418 					bch2_btree_id_str(ck->key.btree_id), new_u64s);
419 				ret = -BCH_ERR_ENOMEM_btree_key_cache_fill;
420 				goto err;
421 			}
422 
423 			if (!bch2_btree_node_relock(trans, ck_path, 0)) {
424 				kfree(new_k);
425 				trace_and_count(trans->c, trans_restart_relock_key_cache_fill, trans, _THIS_IP_, ck_path);
426 				ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_fill);
427 				goto err;
428 			}
429 
430 			ret = bch2_trans_relock(trans);
431 			if (ret) {
432 				kfree(new_k);
433 				goto err;
434 			}
435 		}
436 	}
437 
438 	ret = bch2_btree_node_lock_write(trans, ck_path, &ck_path->l[0].b->c);
439 	if (ret) {
440 		kfree(new_k);
441 		goto err;
442 	}
443 
444 	if (new_k) {
445 		kfree(ck->k);
446 		ck->u64s = new_u64s;
447 		ck->k = new_k;
448 	}
449 
450 	bkey_reassemble(ck->k, k);
451 	ck->valid = true;
452 	bch2_btree_node_unlock_write(trans, ck_path, ck_path->l[0].b);
453 
454 	/* We're not likely to need this iterator again: */
455 	set_btree_iter_dontneed(&iter);
456 err:
457 	bch2_trans_iter_exit(trans, &iter);
458 	return ret;
459 }
460 
461 static noinline int
462 bch2_btree_path_traverse_cached_slowpath(struct btree_trans *trans, struct btree_path *path,
463 					 unsigned flags)
464 {
465 	struct bch_fs *c = trans->c;
466 	struct bkey_cached *ck;
467 	int ret = 0;
468 
469 	BUG_ON(path->level);
470 
471 	path->l[1].b = NULL;
472 
473 	if (bch2_btree_node_relock_notrace(trans, path, 0)) {
474 		ck = (void *) path->l[0].b;
475 		goto fill;
476 	}
477 retry:
478 	ck = bch2_btree_key_cache_find(c, path->btree_id, path->pos);
479 	if (!ck) {
480 		ck = btree_key_cache_create(trans, path);
481 		ret = PTR_ERR_OR_ZERO(ck);
482 		if (ret)
483 			goto err;
484 		if (!ck)
485 			goto retry;
486 
487 		mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
488 		path->locks_want = 1;
489 	} else {
490 		enum six_lock_type lock_want = __btree_lock_want(path, 0);
491 
492 		ret = btree_node_lock(trans, path, (void *) ck, 0,
493 				      lock_want, _THIS_IP_);
494 		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
495 			goto err;
496 
497 		BUG_ON(ret);
498 
499 		if (ck->key.btree_id != path->btree_id ||
500 		    !bpos_eq(ck->key.pos, path->pos)) {
501 			six_unlock_type(&ck->c.lock, lock_want);
502 			goto retry;
503 		}
504 
505 		mark_btree_node_locked(trans, path, 0,
506 				       (enum btree_node_locked_type) lock_want);
507 	}
508 
509 	path->l[0].lock_seq	= six_lock_seq(&ck->c.lock);
510 	path->l[0].b		= (void *) ck;
511 fill:
512 	path->uptodate = BTREE_ITER_UPTODATE;
513 
514 	if (!ck->valid && !(flags & BTREE_ITER_CACHED_NOFILL)) {
515 		/*
516 		 * Using the underscore version because we haven't set
517 		 * path->uptodate yet:
518 		 */
519 		if (!path->locks_want &&
520 		    !__bch2_btree_path_upgrade(trans, path, 1, NULL)) {
521 			trace_and_count(trans->c, trans_restart_key_cache_upgrade, trans, _THIS_IP_);
522 			ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_upgrade);
523 			goto err;
524 		}
525 
526 		ret = btree_key_cache_fill(trans, path, ck);
527 		if (ret)
528 			goto err;
529 
530 		ret = bch2_btree_path_relock(trans, path, _THIS_IP_);
531 		if (ret)
532 			goto err;
533 
534 		path->uptodate = BTREE_ITER_UPTODATE;
535 	}
536 
537 	if (!test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
538 		set_bit(BKEY_CACHED_ACCESSED, &ck->flags);
539 
540 	BUG_ON(btree_node_locked_type(path, 0) != btree_lock_want(path, 0));
541 	BUG_ON(path->uptodate);
542 
543 	return ret;
544 err:
545 	path->uptodate = BTREE_ITER_NEED_TRAVERSE;
546 	if (!bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
547 		btree_node_unlock(trans, path, 0);
548 		path->l[0].b = ERR_PTR(ret);
549 	}
550 	return ret;
551 }
552 
553 int bch2_btree_path_traverse_cached(struct btree_trans *trans, struct btree_path *path,
554 				    unsigned flags)
555 {
556 	struct bch_fs *c = trans->c;
557 	struct bkey_cached *ck;
558 	int ret = 0;
559 
560 	EBUG_ON(path->level);
561 
562 	path->l[1].b = NULL;
563 
564 	if (bch2_btree_node_relock_notrace(trans, path, 0)) {
565 		ck = (void *) path->l[0].b;
566 		goto fill;
567 	}
568 retry:
569 	ck = bch2_btree_key_cache_find(c, path->btree_id, path->pos);
570 	if (!ck) {
571 		return bch2_btree_path_traverse_cached_slowpath(trans, path, flags);
572 	} else {
573 		enum six_lock_type lock_want = __btree_lock_want(path, 0);
574 
575 		ret = btree_node_lock(trans, path, (void *) ck, 0,
576 				      lock_want, _THIS_IP_);
577 		EBUG_ON(ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart));
578 
579 		if (ret)
580 			return ret;
581 
582 		if (ck->key.btree_id != path->btree_id ||
583 		    !bpos_eq(ck->key.pos, path->pos)) {
584 			six_unlock_type(&ck->c.lock, lock_want);
585 			goto retry;
586 		}
587 
588 		mark_btree_node_locked(trans, path, 0,
589 				       (enum btree_node_locked_type) lock_want);
590 	}
591 
592 	path->l[0].lock_seq	= six_lock_seq(&ck->c.lock);
593 	path->l[0].b		= (void *) ck;
594 fill:
595 	if (!ck->valid)
596 		return bch2_btree_path_traverse_cached_slowpath(trans, path, flags);
597 
598 	if (!test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
599 		set_bit(BKEY_CACHED_ACCESSED, &ck->flags);
600 
601 	path->uptodate = BTREE_ITER_UPTODATE;
602 	EBUG_ON(!ck->valid);
603 	EBUG_ON(btree_node_locked_type(path, 0) != btree_lock_want(path, 0));
604 
605 	return ret;
606 }
607 
608 static int btree_key_cache_flush_pos(struct btree_trans *trans,
609 				     struct bkey_cached_key key,
610 				     u64 journal_seq,
611 				     unsigned commit_flags,
612 				     bool evict)
613 {
614 	struct bch_fs *c = trans->c;
615 	struct journal *j = &c->journal;
616 	struct btree_iter c_iter, b_iter;
617 	struct bkey_cached *ck = NULL;
618 	int ret;
619 
620 	bch2_trans_iter_init(trans, &b_iter, key.btree_id, key.pos,
621 			     BTREE_ITER_SLOTS|
622 			     BTREE_ITER_INTENT|
623 			     BTREE_ITER_ALL_SNAPSHOTS);
624 	bch2_trans_iter_init(trans, &c_iter, key.btree_id, key.pos,
625 			     BTREE_ITER_CACHED|
626 			     BTREE_ITER_INTENT);
627 	b_iter.flags &= ~BTREE_ITER_WITH_KEY_CACHE;
628 
629 	ret = bch2_btree_iter_traverse(&c_iter);
630 	if (ret)
631 		goto out;
632 
633 	ck = (void *) c_iter.path->l[0].b;
634 	if (!ck)
635 		goto out;
636 
637 	if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
638 		if (evict)
639 			goto evict;
640 		goto out;
641 	}
642 
643 	BUG_ON(!ck->valid);
644 
645 	if (journal_seq && ck->journal.seq != journal_seq)
646 		goto out;
647 
648 	/*
649 	 * Since journal reclaim depends on us making progress here, and the
650 	 * allocator/copygc depend on journal reclaim making progress, we need
651 	 * to be using alloc reserves:
652 	 */
653 	ret   = bch2_btree_iter_traverse(&b_iter) ?:
654 		bch2_trans_update(trans, &b_iter, ck->k,
655 				  BTREE_UPDATE_KEY_CACHE_RECLAIM|
656 				  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
657 				  BTREE_TRIGGER_NORUN) ?:
658 		bch2_trans_commit(trans, NULL, NULL,
659 				  BTREE_INSERT_NOCHECK_RW|
660 				  BTREE_INSERT_NOFAIL|
661 				  (ck->journal.seq == journal_last_seq(j)
662 				   ? BCH_WATERMARK_reclaim
663 				   : 0)|
664 				  commit_flags);
665 
666 	bch2_fs_fatal_err_on(ret &&
667 			     !bch2_err_matches(ret, BCH_ERR_transaction_restart) &&
668 			     !bch2_err_matches(ret, BCH_ERR_journal_reclaim_would_deadlock) &&
669 			     !bch2_journal_error(j), c,
670 			     "error flushing key cache: %s", bch2_err_str(ret));
671 	if (ret)
672 		goto out;
673 
674 	bch2_journal_pin_drop(j, &ck->journal);
675 
676 	BUG_ON(!btree_node_locked(c_iter.path, 0));
677 
678 	if (!evict) {
679 		if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
680 			clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
681 			atomic_long_dec(&c->btree_key_cache.nr_dirty);
682 		}
683 	} else {
684 		struct btree_path *path2;
685 evict:
686 		trans_for_each_path(trans, path2)
687 			if (path2 != c_iter.path)
688 				__bch2_btree_path_unlock(trans, path2);
689 
690 		bch2_btree_node_lock_write_nofail(trans, c_iter.path, &ck->c);
691 
692 		if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
693 			clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
694 			atomic_long_dec(&c->btree_key_cache.nr_dirty);
695 		}
696 
697 		mark_btree_node_locked_noreset(c_iter.path, 0, BTREE_NODE_UNLOCKED);
698 		bkey_cached_evict(&c->btree_key_cache, ck);
699 		bkey_cached_free_fast(&c->btree_key_cache, ck);
700 	}
701 out:
702 	bch2_trans_iter_exit(trans, &b_iter);
703 	bch2_trans_iter_exit(trans, &c_iter);
704 	return ret;
705 }
706 
707 int bch2_btree_key_cache_journal_flush(struct journal *j,
708 				struct journal_entry_pin *pin, u64 seq)
709 {
710 	struct bch_fs *c = container_of(j, struct bch_fs, journal);
711 	struct bkey_cached *ck =
712 		container_of(pin, struct bkey_cached, journal);
713 	struct bkey_cached_key key;
714 	struct btree_trans *trans = bch2_trans_get(c);
715 	int srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
716 	int ret = 0;
717 
718 	btree_node_lock_nopath_nofail(trans, &ck->c, SIX_LOCK_read);
719 	key = ck->key;
720 
721 	if (ck->journal.seq != seq ||
722 	    !test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
723 		six_unlock_read(&ck->c.lock);
724 		goto unlock;
725 	}
726 
727 	if (ck->seq != seq) {
728 		bch2_journal_pin_update(&c->journal, ck->seq, &ck->journal,
729 					bch2_btree_key_cache_journal_flush);
730 		six_unlock_read(&ck->c.lock);
731 		goto unlock;
732 	}
733 	six_unlock_read(&ck->c.lock);
734 
735 	ret = commit_do(trans, NULL, NULL, 0,
736 		btree_key_cache_flush_pos(trans, key, seq,
737 				BTREE_INSERT_JOURNAL_RECLAIM, false));
738 unlock:
739 	srcu_read_unlock(&c->btree_trans_barrier, srcu_idx);
740 
741 	bch2_trans_put(trans);
742 	return ret;
743 }
744 
745 /*
746  * Flush and evict a key from the key cache:
747  */
748 int bch2_btree_key_cache_flush(struct btree_trans *trans,
749 			       enum btree_id id, struct bpos pos)
750 {
751 	struct bch_fs *c = trans->c;
752 	struct bkey_cached_key key = { id, pos };
753 
754 	/* Fastpath - assume it won't be found: */
755 	if (!bch2_btree_key_cache_find(c, id, pos))
756 		return 0;
757 
758 	return btree_key_cache_flush_pos(trans, key, 0, 0, true);
759 }
760 
761 bool bch2_btree_insert_key_cached(struct btree_trans *trans,
762 				  unsigned flags,
763 				  struct btree_insert_entry *insert_entry)
764 {
765 	struct bch_fs *c = trans->c;
766 	struct bkey_cached *ck = (void *) insert_entry->path->l[0].b;
767 	struct bkey_i *insert = insert_entry->k;
768 	bool kick_reclaim = false;
769 
770 	BUG_ON(insert->k.u64s > ck->u64s);
771 
772 	bkey_copy(ck->k, insert);
773 	ck->valid = true;
774 
775 	if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
776 		EBUG_ON(test_bit(BCH_FS_CLEAN_SHUTDOWN, &c->flags));
777 		set_bit(BKEY_CACHED_DIRTY, &ck->flags);
778 		atomic_long_inc(&c->btree_key_cache.nr_dirty);
779 
780 		if (bch2_nr_btree_keys_need_flush(c))
781 			kick_reclaim = true;
782 	}
783 
784 	/*
785 	 * To minimize lock contention, we only add the journal pin here and
786 	 * defer pin updates to the flush callback via ->seq. Be careful not to
787 	 * update ->seq on nojournal commits because we don't want to update the
788 	 * pin to a seq that doesn't include journal updates on disk. Otherwise
789 	 * we risk losing the update after a crash.
790 	 *
791 	 * The only exception is if the pin is not active in the first place. We
792 	 * have to add the pin because journal reclaim drives key cache
793 	 * flushing. The flush callback will not proceed unless ->seq matches
794 	 * the latest pin, so make sure it starts with a consistent value.
795 	 */
796 	if (!(insert_entry->flags & BTREE_UPDATE_NOJOURNAL) ||
797 	    !journal_pin_active(&ck->journal)) {
798 		ck->seq = trans->journal_res.seq;
799 	}
800 	bch2_journal_pin_add(&c->journal, trans->journal_res.seq,
801 			     &ck->journal, bch2_btree_key_cache_journal_flush);
802 
803 	if (kick_reclaim)
804 		journal_reclaim_kick(&c->journal);
805 	return true;
806 }
807 
808 void bch2_btree_key_cache_drop(struct btree_trans *trans,
809 			       struct btree_path *path)
810 {
811 	struct bch_fs *c = trans->c;
812 	struct bkey_cached *ck = (void *) path->l[0].b;
813 
814 	BUG_ON(!ck->valid);
815 
816 	/*
817 	 * We just did an update to the btree, bypassing the key cache: the key
818 	 * cache key is now stale and must be dropped, even if dirty:
819 	 */
820 	if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
821 		clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
822 		atomic_long_dec(&c->btree_key_cache.nr_dirty);
823 		bch2_journal_pin_drop(&c->journal, &ck->journal);
824 	}
825 
826 	ck->valid = false;
827 }
828 
829 static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
830 					   struct shrink_control *sc)
831 {
832 	struct bch_fs *c = shrink->private_data;
833 	struct btree_key_cache *bc = &c->btree_key_cache;
834 	struct bucket_table *tbl;
835 	struct bkey_cached *ck, *t;
836 	size_t scanned = 0, freed = 0, nr = sc->nr_to_scan;
837 	unsigned start, flags;
838 	int srcu_idx;
839 
840 	mutex_lock(&bc->lock);
841 	srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
842 	flags = memalloc_nofs_save();
843 
844 	/*
845 	 * Newest freed entries are at the end of the list - once we hit one
846 	 * that's too new to be freed, we can bail out:
847 	 */
848 	scanned += bc->nr_freed_nonpcpu;
849 
850 	list_for_each_entry_safe(ck, t, &bc->freed_nonpcpu, list) {
851 		if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
852 						 ck->btree_trans_barrier_seq))
853 			break;
854 
855 		list_del(&ck->list);
856 		six_lock_exit(&ck->c.lock);
857 		kmem_cache_free(bch2_key_cache, ck);
858 		atomic_long_dec(&bc->nr_freed);
859 		freed++;
860 		bc->nr_freed_nonpcpu--;
861 	}
862 
863 	if (scanned >= nr)
864 		goto out;
865 
866 	scanned += bc->nr_freed_pcpu;
867 
868 	list_for_each_entry_safe(ck, t, &bc->freed_pcpu, list) {
869 		if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
870 						 ck->btree_trans_barrier_seq))
871 			break;
872 
873 		list_del(&ck->list);
874 		six_lock_exit(&ck->c.lock);
875 		kmem_cache_free(bch2_key_cache, ck);
876 		atomic_long_dec(&bc->nr_freed);
877 		freed++;
878 		bc->nr_freed_pcpu--;
879 	}
880 
881 	if (scanned >= nr)
882 		goto out;
883 
884 	rcu_read_lock();
885 	tbl = rht_dereference_rcu(bc->table.tbl, &bc->table);
886 	if (bc->shrink_iter >= tbl->size)
887 		bc->shrink_iter = 0;
888 	start = bc->shrink_iter;
889 
890 	do {
891 		struct rhash_head *pos, *next;
892 
893 		pos = rht_ptr_rcu(rht_bucket(tbl, bc->shrink_iter));
894 
895 		while (!rht_is_a_nulls(pos)) {
896 			next = rht_dereference_bucket_rcu(pos->next, tbl, bc->shrink_iter);
897 			ck = container_of(pos, struct bkey_cached, hash);
898 
899 			if (test_bit(BKEY_CACHED_DIRTY, &ck->flags))
900 				goto next;
901 
902 			if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
903 				clear_bit(BKEY_CACHED_ACCESSED, &ck->flags);
904 			else if (bkey_cached_lock_for_evict(ck)) {
905 				bkey_cached_evict(bc, ck);
906 				bkey_cached_free(bc, ck);
907 			}
908 
909 			scanned++;
910 			if (scanned >= nr)
911 				break;
912 next:
913 			pos = next;
914 		}
915 
916 		bc->shrink_iter++;
917 		if (bc->shrink_iter >= tbl->size)
918 			bc->shrink_iter = 0;
919 	} while (scanned < nr && bc->shrink_iter != start);
920 
921 	rcu_read_unlock();
922 out:
923 	memalloc_nofs_restore(flags);
924 	srcu_read_unlock(&c->btree_trans_barrier, srcu_idx);
925 	mutex_unlock(&bc->lock);
926 
927 	return freed;
928 }
929 
930 static unsigned long bch2_btree_key_cache_count(struct shrinker *shrink,
931 					    struct shrink_control *sc)
932 {
933 	struct bch_fs *c = shrink->private_data;
934 	struct btree_key_cache *bc = &c->btree_key_cache;
935 	long nr = atomic_long_read(&bc->nr_keys) -
936 		atomic_long_read(&bc->nr_dirty);
937 
938 	return max(0L, nr);
939 }
940 
941 void bch2_fs_btree_key_cache_exit(struct btree_key_cache *bc)
942 {
943 	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
944 	struct bucket_table *tbl;
945 	struct bkey_cached *ck, *n;
946 	struct rhash_head *pos;
947 	LIST_HEAD(items);
948 	unsigned i;
949 #ifdef __KERNEL__
950 	int cpu;
951 #endif
952 
953 	shrinker_free(bc->shrink);
954 
955 	mutex_lock(&bc->lock);
956 
957 	/*
958 	 * The loop is needed to guard against racing with rehash:
959 	 */
960 	while (atomic_long_read(&bc->nr_keys)) {
961 		rcu_read_lock();
962 		tbl = rht_dereference_rcu(bc->table.tbl, &bc->table);
963 		if (tbl)
964 			for (i = 0; i < tbl->size; i++)
965 				rht_for_each_entry_rcu(ck, pos, tbl, i, hash) {
966 					bkey_cached_evict(bc, ck);
967 					list_add(&ck->list, &items);
968 				}
969 		rcu_read_unlock();
970 	}
971 
972 #ifdef __KERNEL__
973 	for_each_possible_cpu(cpu) {
974 		struct btree_key_cache_freelist *f =
975 			per_cpu_ptr(bc->pcpu_freed, cpu);
976 
977 		for (i = 0; i < f->nr; i++) {
978 			ck = f->objs[i];
979 			list_add(&ck->list, &items);
980 		}
981 	}
982 #endif
983 
984 	BUG_ON(list_count_nodes(&bc->freed_pcpu) != bc->nr_freed_pcpu);
985 	BUG_ON(list_count_nodes(&bc->freed_nonpcpu) != bc->nr_freed_nonpcpu);
986 
987 	list_splice(&bc->freed_pcpu,	&items);
988 	list_splice(&bc->freed_nonpcpu,	&items);
989 
990 	mutex_unlock(&bc->lock);
991 
992 	list_for_each_entry_safe(ck, n, &items, list) {
993 		cond_resched();
994 
995 		bch2_journal_pin_drop(&c->journal, &ck->journal);
996 
997 		list_del(&ck->list);
998 		kfree(ck->k);
999 		six_lock_exit(&ck->c.lock);
1000 		kmem_cache_free(bch2_key_cache, ck);
1001 	}
1002 
1003 	if (atomic_long_read(&bc->nr_dirty) &&
1004 	    !bch2_journal_error(&c->journal) &&
1005 	    test_bit(BCH_FS_WAS_RW, &c->flags))
1006 		panic("btree key cache shutdown error: nr_dirty nonzero (%li)\n",
1007 		      atomic_long_read(&bc->nr_dirty));
1008 
1009 	if (atomic_long_read(&bc->nr_keys))
1010 		panic("btree key cache shutdown error: nr_keys nonzero (%li)\n",
1011 		      atomic_long_read(&bc->nr_keys));
1012 
1013 	if (bc->table_init_done)
1014 		rhashtable_destroy(&bc->table);
1015 
1016 	free_percpu(bc->pcpu_freed);
1017 }
1018 
1019 void bch2_fs_btree_key_cache_init_early(struct btree_key_cache *c)
1020 {
1021 	mutex_init(&c->lock);
1022 	INIT_LIST_HEAD(&c->freed_pcpu);
1023 	INIT_LIST_HEAD(&c->freed_nonpcpu);
1024 }
1025 
1026 int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
1027 {
1028 	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
1029 	struct shrinker *shrink;
1030 
1031 #ifdef __KERNEL__
1032 	bc->pcpu_freed = alloc_percpu(struct btree_key_cache_freelist);
1033 	if (!bc->pcpu_freed)
1034 		return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1035 #endif
1036 
1037 	if (rhashtable_init(&bc->table, &bch2_btree_key_cache_params))
1038 		return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1039 
1040 	bc->table_init_done = true;
1041 
1042 	shrink = shrinker_alloc(0, "%s-btree_key_cache", c->name);
1043 	if (!shrink)
1044 		return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1045 	bc->shrink = shrink;
1046 	shrink->seeks		= 0;
1047 	shrink->count_objects	= bch2_btree_key_cache_count;
1048 	shrink->scan_objects	= bch2_btree_key_cache_scan;
1049 	shrink->private_data	= c;
1050 	shrinker_register(shrink);
1051 	return 0;
1052 }
1053 
1054 void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *c)
1055 {
1056 	prt_printf(out, "nr_freed:\t%lu",	atomic_long_read(&c->nr_freed));
1057 	prt_newline(out);
1058 	prt_printf(out, "nr_keys:\t%lu",	atomic_long_read(&c->nr_keys));
1059 	prt_newline(out);
1060 	prt_printf(out, "nr_dirty:\t%lu",	atomic_long_read(&c->nr_dirty));
1061 	prt_newline(out);
1062 }
1063 
1064 void bch2_btree_key_cache_exit(void)
1065 {
1066 	kmem_cache_destroy(bch2_key_cache);
1067 }
1068 
1069 int __init bch2_btree_key_cache_init(void)
1070 {
1071 	bch2_key_cache = KMEM_CACHE(bkey_cached, SLAB_RECLAIM_ACCOUNT);
1072 	if (!bch2_key_cache)
1073 		return -ENOMEM;
1074 
1075 	return 0;
1076 }
1077