1 // SPDX-License-Identifier: GPL-2.0
5 #include "btree_cache.h"
7 #include "btree_iter.h"
8 #include "btree_locking.h"
14 #include <linux/prefetch.h>
15 #include <linux/sched/mm.h>
16 #include <linux/seq_buf.h>
18 #define BTREE_CACHE_NOT_FREED_INCREMENT(counter) \
20 if (shrinker_counter) \
21 bc->not_freed_##counter++; \
24 const char * const bch2_btree_node_flags[] = {
31 void bch2_recalc_btree_reserve(struct bch_fs *c)
33 unsigned i, reserve = 16;
35 if (!c->btree_roots[0].b)
38 for (i = 0; i < BTREE_ID_NR; i++)
39 if (c->btree_roots[i].b)
40 reserve += min_t(unsigned, 1,
41 c->btree_roots[i].b->c.level) * 8;
43 c->btree_cache.reserve = reserve;
46 static inline unsigned btree_cache_can_free(struct btree_cache *bc)
48 return max_t(int, 0, bc->used - bc->reserve);
51 static void btree_node_to_freedlist(struct btree_cache *bc, struct btree *b)
53 if (b->c.lock.readers)
54 list_move(&b->list, &bc->freed_pcpu);
56 list_move(&b->list, &bc->freed_nonpcpu);
59 static void btree_node_data_free(struct bch_fs *c, struct btree *b)
61 struct btree_cache *bc = &c->btree_cache;
63 EBUG_ON(btree_node_write_in_flight(b));
65 clear_btree_node_just_written(b);
67 kvpfree(b->data, btree_bytes(c));
72 munmap(b->aux_data, btree_aux_data_bytes(b));
78 btree_node_to_freedlist(bc, b);
81 static int bch2_btree_cache_cmp_fn(struct rhashtable_compare_arg *arg,
84 const struct btree *b = obj;
85 const u64 *v = arg->key;
87 return b->hash_val == *v ? 0 : 1;
90 static const struct rhashtable_params bch_btree_cache_params = {
91 .head_offset = offsetof(struct btree, hash),
92 .key_offset = offsetof(struct btree, hash_val),
93 .key_len = sizeof(u64),
94 .obj_cmpfn = bch2_btree_cache_cmp_fn,
97 static int btree_node_data_alloc(struct bch_fs *c, struct btree *b, gfp_t gfp)
99 BUG_ON(b->data || b->aux_data);
101 b->data = kvpmalloc(btree_bytes(c), gfp);
103 return -BCH_ERR_ENOMEM_btree_node_mem_alloc;
105 b->aux_data = kvmalloc(btree_aux_data_bytes(b), gfp);
107 b->aux_data = mmap(NULL, btree_aux_data_bytes(b),
108 PROT_READ|PROT_WRITE|PROT_EXEC,
109 MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);
110 if (b->aux_data == MAP_FAILED)
114 kvpfree(b->data, btree_bytes(c));
116 return -BCH_ERR_ENOMEM_btree_node_mem_alloc;
122 static struct btree *__btree_node_mem_alloc(struct bch_fs *c, gfp_t gfp)
126 b = kzalloc(sizeof(struct btree), gfp);
130 bkey_btree_ptr_init(&b->key);
131 #ifdef CONFIG_DEBUG_LOCK_ALLOC
132 lockdep_set_no_check_recursion(&b->c.lock.dep_map);
134 INIT_LIST_HEAD(&b->list);
135 INIT_LIST_HEAD(&b->write_blocked);
136 b->byte_order = ilog2(btree_bytes(c));
140 struct btree *__bch2_btree_node_mem_alloc(struct bch_fs *c)
142 struct btree_cache *bc = &c->btree_cache;
145 b = __btree_node_mem_alloc(c, GFP_KERNEL);
149 if (btree_node_data_alloc(c, b, GFP_KERNEL)) {
154 bch2_btree_lock_init(&b->c, 0);
157 list_add(&b->list, &bc->freeable);
161 /* Btree in memory cache - hash table */
163 void bch2_btree_node_hash_remove(struct btree_cache *bc, struct btree *b)
165 int ret = rhashtable_remove_fast(&bc->table, &b->hash, bch_btree_cache_params);
169 /* Cause future lookups for this node to fail: */
173 int __bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b)
176 b->hash_val = btree_ptr_hash_val(&b->key);
178 return rhashtable_lookup_insert_fast(&bc->table, &b->hash,
179 bch_btree_cache_params);
182 int bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b,
183 unsigned level, enum btree_id id)
190 mutex_lock(&bc->lock);
191 ret = __bch2_btree_node_hash_insert(bc, b);
193 list_add_tail(&b->list, &bc->live);
194 mutex_unlock(&bc->lock);
200 static inline struct btree *btree_cache_find(struct btree_cache *bc,
201 const struct bkey_i *k)
203 u64 v = btree_ptr_hash_val(k);
205 return rhashtable_lookup_fast(&bc->table, &v, bch_btree_cache_params);
209 * this version is for btree nodes that have already been freed (we're not
210 * reaping a real btree node)
212 static int __btree_node_reclaim(struct bch_fs *c, struct btree *b, bool flush, bool shrinker_counter)
214 struct btree_cache *bc = &c->btree_cache;
217 lockdep_assert_held(&bc->lock);
219 if (b->flags & ((1U << BTREE_NODE_dirty)|
220 (1U << BTREE_NODE_read_in_flight)|
221 (1U << BTREE_NODE_write_in_flight))) {
223 if (btree_node_dirty(b))
224 BTREE_CACHE_NOT_FREED_INCREMENT(dirty);
225 else if (btree_node_read_in_flight(b))
226 BTREE_CACHE_NOT_FREED_INCREMENT(read_in_flight);
227 else if (btree_node_write_in_flight(b))
228 BTREE_CACHE_NOT_FREED_INCREMENT(write_in_flight);
229 return -BCH_ERR_ENOMEM_btree_node_reclaim;
232 /* XXX: waiting on IO with btree cache lock held */
233 bch2_btree_node_wait_on_read(b);
234 bch2_btree_node_wait_on_write(b);
237 if (!six_trylock_intent(&b->c.lock)) {
238 BTREE_CACHE_NOT_FREED_INCREMENT(lock_intent);
239 return -BCH_ERR_ENOMEM_btree_node_reclaim;
242 if (!six_trylock_write(&b->c.lock)) {
243 BTREE_CACHE_NOT_FREED_INCREMENT(lock_write);
244 goto out_unlock_intent;
247 /* recheck under lock */
248 if (b->flags & ((1U << BTREE_NODE_read_in_flight)|
249 (1U << BTREE_NODE_write_in_flight))) {
251 if (btree_node_read_in_flight(b))
252 BTREE_CACHE_NOT_FREED_INCREMENT(read_in_flight);
253 else if (btree_node_write_in_flight(b))
254 BTREE_CACHE_NOT_FREED_INCREMENT(write_in_flight);
257 six_unlock_write(&b->c.lock);
258 six_unlock_intent(&b->c.lock);
262 if (btree_node_noevict(b)) {
263 BTREE_CACHE_NOT_FREED_INCREMENT(noevict);
266 if (btree_node_write_blocked(b)) {
267 BTREE_CACHE_NOT_FREED_INCREMENT(write_blocked);
270 if (btree_node_will_make_reachable(b)) {
271 BTREE_CACHE_NOT_FREED_INCREMENT(will_make_reachable);
275 if (btree_node_dirty(b)) {
277 BTREE_CACHE_NOT_FREED_INCREMENT(dirty);
281 * Using the underscore version because we don't want to compact
282 * bsets after the write, since this node is about to be evicted
283 * - unless btree verify mode is enabled, since it runs out of
284 * the post write cleanup:
286 if (bch2_verify_btree_ondisk)
287 bch2_btree_node_write(c, b, SIX_LOCK_intent,
288 BTREE_WRITE_cache_reclaim);
290 __bch2_btree_node_write(c, b,
291 BTREE_WRITE_cache_reclaim);
293 six_unlock_write(&b->c.lock);
294 six_unlock_intent(&b->c.lock);
298 if (b->hash_val && !ret)
299 trace_and_count(c, btree_cache_reap, c, b);
302 six_unlock_write(&b->c.lock);
304 six_unlock_intent(&b->c.lock);
305 ret = -BCH_ERR_ENOMEM_btree_node_reclaim;
309 static int btree_node_reclaim(struct bch_fs *c, struct btree *b, bool shrinker_counter)
311 return __btree_node_reclaim(c, b, false, shrinker_counter);
314 static int btree_node_write_and_reclaim(struct bch_fs *c, struct btree *b)
316 return __btree_node_reclaim(c, b, true, false);
319 static unsigned long bch2_btree_cache_scan(struct shrinker *shrink,
320 struct shrink_control *sc)
322 struct bch_fs *c = container_of(shrink, struct bch_fs,
324 struct btree_cache *bc = &c->btree_cache;
326 unsigned long nr = sc->nr_to_scan;
327 unsigned long can_free = 0;
328 unsigned long freed = 0;
329 unsigned long touched = 0;
331 unsigned long ret = SHRINK_STOP;
332 bool trigger_writes = atomic_read(&bc->dirty) + nr >=
335 if (bch2_btree_shrinker_disabled)
338 mutex_lock(&bc->lock);
339 flags = memalloc_nofs_save();
342 * It's _really_ critical that we don't free too many btree nodes - we
343 * have to always leave ourselves a reserve. The reserve is how we
344 * guarantee that allocating memory for a new btree node can always
345 * succeed, so that inserting keys into the btree can always succeed and
346 * IO can always make forward progress:
348 can_free = btree_cache_can_free(bc);
349 nr = min_t(unsigned long, nr, can_free);
352 list_for_each_entry_safe(b, t, &bc->freeable, list) {
354 * Leave a few nodes on the freeable list, so that a btree split
355 * won't have to hit the system allocator:
365 if (!btree_node_reclaim(c, b, true)) {
366 btree_node_data_free(c, b);
367 six_unlock_write(&b->c.lock);
368 six_unlock_intent(&b->c.lock);
374 list_for_each_entry_safe(b, t, &bc->live, list) {
377 if (btree_node_accessed(b)) {
378 clear_btree_node_accessed(b);
379 bc->not_freed_access_bit++;
380 } else if (!btree_node_reclaim(c, b, true)) {
382 btree_node_data_free(c, b);
385 bch2_btree_node_hash_remove(bc, b);
386 six_unlock_write(&b->c.lock);
387 six_unlock_intent(&b->c.lock);
391 } else if (trigger_writes &&
392 btree_node_dirty(b) &&
393 !btree_node_will_make_reachable(b) &&
394 !btree_node_write_blocked(b) &&
395 six_trylock_read(&b->c.lock)) {
396 list_move(&bc->live, &b->list);
397 mutex_unlock(&bc->lock);
398 __bch2_btree_node_write(c, b, BTREE_WRITE_cache_reclaim);
399 six_unlock_read(&b->c.lock);
402 mutex_lock(&bc->lock);
410 if (&t->list != &bc->live)
411 list_move_tail(&bc->live, &t->list);
413 mutex_unlock(&bc->lock);
416 memalloc_nofs_restore(flags);
417 trace_and_count(c, btree_cache_scan, sc->nr_to_scan, can_free, ret);
421 static unsigned long bch2_btree_cache_count(struct shrinker *shrink,
422 struct shrink_control *sc)
424 struct bch_fs *c = container_of(shrink, struct bch_fs,
426 struct btree_cache *bc = &c->btree_cache;
428 if (bch2_btree_shrinker_disabled)
431 return btree_cache_can_free(bc);
434 static void bch2_btree_cache_shrinker_to_text(struct seq_buf *s, struct shrinker *shrink)
436 struct bch_fs *c = container_of(shrink, struct bch_fs,
439 size_t buflen = seq_buf_get_buf(s, &cbuf);
440 struct printbuf out = PRINTBUF_EXTERN(cbuf, buflen);
442 bch2_btree_cache_to_text(&out, &c->btree_cache);
443 seq_buf_commit(s, out.pos);
446 void bch2_fs_btree_cache_exit(struct bch_fs *c)
448 struct btree_cache *bc = &c->btree_cache;
452 if (bc->shrink.list.next)
453 unregister_shrinker(&bc->shrink);
455 /* vfree() can allocate memory: */
456 flags = memalloc_nofs_save();
457 mutex_lock(&bc->lock);
460 list_move(&c->verify_data->list, &bc->live);
462 kvpfree(c->verify_ondisk, btree_bytes(c));
464 for (i = 0; i < BTREE_ID_NR; i++)
465 if (c->btree_roots[i].b)
466 list_add(&c->btree_roots[i].b->list, &bc->live);
468 list_splice(&bc->freeable, &bc->live);
470 while (!list_empty(&bc->live)) {
471 b = list_first_entry(&bc->live, struct btree, list);
473 BUG_ON(btree_node_read_in_flight(b) ||
474 btree_node_write_in_flight(b));
476 if (btree_node_dirty(b))
477 bch2_btree_complete_write(c, b, btree_current_write(b));
478 clear_btree_node_dirty_acct(c, b);
480 btree_node_data_free(c, b);
483 BUG_ON(atomic_read(&c->btree_cache.dirty));
485 list_splice(&bc->freed_pcpu, &bc->freed_nonpcpu);
487 while (!list_empty(&bc->freed_nonpcpu)) {
488 b = list_first_entry(&bc->freed_nonpcpu, struct btree, list);
490 six_lock_exit(&b->c.lock);
494 mutex_unlock(&bc->lock);
495 memalloc_nofs_restore(flags);
497 if (bc->table_init_done)
498 rhashtable_destroy(&bc->table);
501 int bch2_fs_btree_cache_init(struct bch_fs *c)
503 struct btree_cache *bc = &c->btree_cache;
507 pr_verbose_init(c->opts, "");
509 ret = rhashtable_init(&bc->table, &bch_btree_cache_params);
513 bc->table_init_done = true;
515 bch2_recalc_btree_reserve(c);
517 for (i = 0; i < bc->reserve; i++)
518 if (!__bch2_btree_node_mem_alloc(c)) {
519 ret = -BCH_ERR_ENOMEM_fs_btree_cache_init;
523 list_splice_init(&bc->live, &bc->freeable);
525 mutex_init(&c->verify_lock);
527 bc->shrink.count_objects = bch2_btree_cache_count;
528 bc->shrink.scan_objects = bch2_btree_cache_scan;
529 bc->shrink.to_text = bch2_btree_cache_shrinker_to_text;
530 bc->shrink.seeks = 4;
531 ret = register_shrinker(&bc->shrink, "%s/btree_cache", c->name);
533 pr_verbose_init(c->opts, "ret %i", ret);
537 void bch2_fs_btree_cache_init_early(struct btree_cache *bc)
539 mutex_init(&bc->lock);
540 INIT_LIST_HEAD(&bc->live);
541 INIT_LIST_HEAD(&bc->freeable);
542 INIT_LIST_HEAD(&bc->freed_pcpu);
543 INIT_LIST_HEAD(&bc->freed_nonpcpu);
547 * We can only have one thread cannibalizing other cached btree nodes at a time,
548 * or we'll deadlock. We use an open coded mutex to ensure that, which a
549 * cannibalize_bucket() will take. This means every time we unlock the root of
550 * the btree, we need to release this lock if we have it held.
552 void bch2_btree_cache_cannibalize_unlock(struct bch_fs *c)
554 struct btree_cache *bc = &c->btree_cache;
556 if (bc->alloc_lock == current) {
557 trace_and_count(c, btree_cache_cannibalize_unlock, c);
558 bc->alloc_lock = NULL;
559 closure_wake_up(&bc->alloc_wait);
563 int bch2_btree_cache_cannibalize_lock(struct bch_fs *c, struct closure *cl)
565 struct btree_cache *bc = &c->btree_cache;
566 struct task_struct *old;
568 old = cmpxchg(&bc->alloc_lock, NULL, current);
569 if (old == NULL || old == current)
573 trace_and_count(c, btree_cache_cannibalize_lock_fail, c);
574 return -BCH_ERR_ENOMEM_btree_cache_cannibalize_lock;
577 closure_wait(&bc->alloc_wait, cl);
579 /* Try again, after adding ourselves to waitlist */
580 old = cmpxchg(&bc->alloc_lock, NULL, current);
581 if (old == NULL || old == current) {
583 closure_wake_up(&bc->alloc_wait);
587 trace_and_count(c, btree_cache_cannibalize_lock_fail, c);
588 return -BCH_ERR_btree_cache_cannibalize_lock_blocked;
591 trace_and_count(c, btree_cache_cannibalize_lock, c);
595 static struct btree *btree_node_cannibalize(struct bch_fs *c)
597 struct btree_cache *bc = &c->btree_cache;
600 list_for_each_entry_reverse(b, &bc->live, list)
601 if (!btree_node_reclaim(c, b, false))
605 list_for_each_entry_reverse(b, &bc->live, list)
606 if (!btree_node_write_and_reclaim(c, b))
610 * Rare case: all nodes were intent-locked.
613 WARN_ONCE(1, "btree cache cannibalize failed\n");
618 struct btree *bch2_btree_node_mem_alloc(struct btree_trans *trans, bool pcpu_read_locks)
620 struct bch_fs *c = trans->c;
621 struct btree_cache *bc = &c->btree_cache;
622 struct list_head *freed = pcpu_read_locks
624 : &bc->freed_nonpcpu;
625 struct btree *b, *b2;
626 u64 start_time = local_clock();
629 flags = memalloc_nofs_save();
630 mutex_lock(&bc->lock);
633 * We never free struct btree itself, just the memory that holds the on
634 * disk node. Check the freed list before allocating a new one:
636 list_for_each_entry(b, freed, list)
637 if (!btree_node_reclaim(c, b, false)) {
638 list_del_init(&b->list);
642 b = __btree_node_mem_alloc(c, __GFP_NOWARN);
644 mutex_unlock(&bc->lock);
645 b = __btree_node_mem_alloc(c, GFP_KERNEL);
648 mutex_lock(&bc->lock);
651 bch2_btree_lock_init(&b->c, pcpu_read_locks ? SIX_LOCK_INIT_PCPU : 0);
653 BUG_ON(!six_trylock_intent(&b->c.lock));
654 BUG_ON(!six_trylock_write(&b->c.lock));
658 * btree_free() doesn't free memory; it sticks the node on the end of
659 * the list. Check if there's any freed nodes there:
661 list_for_each_entry(b2, &bc->freeable, list)
662 if (!btree_node_reclaim(c, b2, false)) {
663 swap(b->data, b2->data);
664 swap(b->aux_data, b2->aux_data);
665 btree_node_to_freedlist(bc, b2);
666 six_unlock_write(&b2->c.lock);
667 six_unlock_intent(&b2->c.lock);
671 mutex_unlock(&bc->lock);
673 if (btree_node_data_alloc(c, b, __GFP_NOWARN|GFP_KERNEL))
676 mutex_lock(&bc->lock);
679 mutex_unlock(&bc->lock);
681 BUG_ON(btree_node_hashed(b));
682 BUG_ON(btree_node_dirty(b));
683 BUG_ON(btree_node_write_in_flight(b));
690 b->whiteout_u64s = 0;
691 bch2_btree_keys_init(b);
692 set_btree_node_accessed(b);
694 bch2_time_stats_update(&c->times[BCH_TIME_btree_node_mem_alloc],
697 memalloc_nofs_restore(flags);
700 mutex_lock(&bc->lock);
702 /* Try to cannibalize another cached btree node: */
703 if (bc->alloc_lock == current) {
704 b2 = btree_node_cannibalize(c);
705 clear_btree_node_just_written(b2);
706 bch2_btree_node_hash_remove(bc, b2);
709 swap(b->data, b2->data);
710 swap(b->aux_data, b2->aux_data);
711 btree_node_to_freedlist(bc, b2);
712 six_unlock_write(&b2->c.lock);
713 six_unlock_intent(&b2->c.lock);
716 list_del_init(&b->list);
719 mutex_unlock(&bc->lock);
721 trace_and_count(c, btree_cache_cannibalize, c);
725 mutex_unlock(&bc->lock);
726 memalloc_nofs_restore(flags);
727 return ERR_PTR(-BCH_ERR_ENOMEM_btree_node_mem_alloc);
730 /* Slowpath, don't want it inlined into btree_iter_traverse() */
731 static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
732 struct btree_path *path,
733 const struct bkey_i *k,
734 enum btree_id btree_id,
736 enum six_lock_type lock_type,
739 struct bch_fs *c = trans->c;
740 struct btree_cache *bc = &c->btree_cache;
744 BUG_ON(level + 1 >= BTREE_MAX_DEPTH);
746 * Parent node must be locked, else we could read in a btree node that's
749 if (path && !bch2_btree_node_relock(trans, path, level + 1)) {
750 trace_and_count(c, trans_restart_relock_parent_for_fill, trans, _THIS_IP_, path);
751 return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_fill_relock));
754 b = bch2_btree_node_mem_alloc(trans, level != 0);
756 if (bch2_err_matches(PTR_ERR_OR_ZERO(b), ENOMEM)) {
757 trans->memory_allocation_failure = true;
758 trace_and_count(c, trans_restart_memory_allocation_failure, trans, _THIS_IP_, path);
759 return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_fill_mem_alloc_fail));
766 * Btree nodes read in from disk should not have the accessed bit set
767 * initially, so that linear scans don't thrash the cache:
769 clear_btree_node_accessed(b);
771 bkey_copy(&b->key, k);
772 if (bch2_btree_node_hash_insert(bc, b, level, btree_id)) {
773 /* raced with another fill: */
775 /* mark as unhashed... */
778 mutex_lock(&bc->lock);
779 list_add(&b->list, &bc->freeable);
780 mutex_unlock(&bc->lock);
782 six_unlock_write(&b->c.lock);
783 six_unlock_intent(&b->c.lock);
787 set_btree_node_read_in_flight(b);
789 six_unlock_write(&b->c.lock);
790 seq = six_lock_seq(&b->c.lock);
791 six_unlock_intent(&b->c.lock);
793 /* Unlock before doing IO: */
795 bch2_trans_unlock(trans);
797 bch2_btree_node_read(c, b, sync);
803 int ret = bch2_trans_relock(trans) ?:
804 bch2_btree_path_relock_intent(trans, path);
806 BUG_ON(!trans->restarted);
811 if (!six_relock_type(&b->c.lock, lock_type, seq)) {
813 trace_and_count(c, trans_restart_relock_after_fill, trans, _THIS_IP_, path);
814 return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_relock_after_fill));
820 static noinline void btree_bad_header(struct bch_fs *c, struct btree *b)
822 struct printbuf buf = PRINTBUF;
824 if (!test_bit(BCH_FS_INITIAL_GC_DONE, &c->flags))
828 "btree node header doesn't match ptr\n"
829 "btree %s level %u\n"
831 bch2_btree_ids[b->c.btree_id], b->c.level);
832 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
834 prt_printf(&buf, "\nheader: btree %s level %llu\n"
836 bch2_btree_ids[BTREE_NODE_ID(b->data)],
837 BTREE_NODE_LEVEL(b->data));
838 bch2_bpos_to_text(&buf, b->data->min_key);
840 prt_printf(&buf, "\nmax ");
841 bch2_bpos_to_text(&buf, b->data->max_key);
843 bch2_fs_inconsistent(c, "%s", buf.buf);
847 static inline void btree_check_header(struct bch_fs *c, struct btree *b)
849 if (b->c.btree_id != BTREE_NODE_ID(b->data) ||
850 b->c.level != BTREE_NODE_LEVEL(b->data) ||
851 !bpos_eq(b->data->max_key, b->key.k.p) ||
852 (b->key.k.type == KEY_TYPE_btree_ptr_v2 &&
853 !bpos_eq(b->data->min_key,
854 bkey_i_to_btree_ptr_v2(&b->key)->v.min_key)))
855 btree_bad_header(c, b);
858 static struct btree *__bch2_btree_node_get(struct btree_trans *trans, struct btree_path *path,
859 const struct bkey_i *k, unsigned level,
860 enum six_lock_type lock_type,
861 unsigned long trace_ip)
863 struct bch_fs *c = trans->c;
864 struct btree_cache *bc = &c->btree_cache;
869 EBUG_ON(level >= BTREE_MAX_DEPTH);
871 b = btree_cache_find(bc, k);
874 * We must have the parent locked to call bch2_btree_node_fill(),
875 * else we could read in a btree node from disk that's been
878 b = bch2_btree_node_fill(trans, path, k, path->btree_id,
879 level, lock_type, true);
881 /* We raced and found the btree node in the cache */
888 if (btree_node_read_locked(path, level + 1))
889 btree_node_unlock(trans, path, level + 1);
891 ret = btree_node_lock(trans, path, &b->c, level, lock_type, trace_ip);
892 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
897 if (unlikely(b->hash_val != btree_ptr_hash_val(k) ||
898 b->c.level != level ||
900 six_unlock_type(&b->c.lock, lock_type);
901 if (bch2_btree_node_relock(trans, path, level + 1))
904 trace_and_count(c, trans_restart_btree_node_reused, trans, trace_ip, path);
905 return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_lock_node_reused));
908 /* avoid atomic set bit if it's not needed: */
909 if (!btree_node_accessed(b))
910 set_btree_node_accessed(b);
913 if (unlikely(btree_node_read_in_flight(b))) {
914 u32 seq = six_lock_seq(&b->c.lock);
916 six_unlock_type(&b->c.lock, lock_type);
917 bch2_trans_unlock(trans);
919 bch2_btree_node_wait_on_read(b);
922 * should_be_locked is not set on this path yet, so we need to
923 * relock it specifically:
926 int ret = bch2_trans_relock(trans) ?:
927 bch2_btree_path_relock_intent(trans, path);
929 BUG_ON(!trans->restarted);
934 if (!six_relock_type(&b->c.lock, lock_type, seq))
938 prefetch(b->aux_data);
940 for_each_bset(b, t) {
941 void *p = (u64 *) b->aux_data + t->aux_data_offset;
943 prefetch(p + L1_CACHE_BYTES * 0);
944 prefetch(p + L1_CACHE_BYTES * 1);
945 prefetch(p + L1_CACHE_BYTES * 2);
948 if (unlikely(btree_node_read_error(b))) {
949 six_unlock_type(&b->c.lock, lock_type);
950 return ERR_PTR(-EIO);
953 EBUG_ON(b->c.btree_id != path->btree_id);
954 EBUG_ON(BTREE_NODE_LEVEL(b->data) != level);
955 btree_check_header(c, b);
961 * bch_btree_node_get - find a btree node in the cache and lock it, reading it
962 * in from disk if necessary.
964 * The btree node will have either a read or a write lock held, depending on
965 * the @write parameter.
967 struct btree *bch2_btree_node_get(struct btree_trans *trans, struct btree_path *path,
968 const struct bkey_i *k, unsigned level,
969 enum six_lock_type lock_type,
970 unsigned long trace_ip)
972 struct bch_fs *c = trans->c;
977 EBUG_ON(level >= BTREE_MAX_DEPTH);
979 b = btree_node_mem_ptr(k);
982 * Check b->hash_val _before_ calling btree_node_lock() - this might not
983 * be the node we want anymore, and trying to lock the wrong node could
984 * cause an unneccessary transaction restart:
986 if (unlikely(!c->opts.btree_node_mem_ptr_optimization ||
988 b->hash_val != btree_ptr_hash_val(k)))
989 return __bch2_btree_node_get(trans, path, k, level, lock_type, trace_ip);
991 if (btree_node_read_locked(path, level + 1))
992 btree_node_unlock(trans, path, level + 1);
994 ret = btree_node_lock(trans, path, &b->c, level, lock_type, trace_ip);
995 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
1000 if (unlikely(b->hash_val != btree_ptr_hash_val(k) ||
1001 b->c.level != level ||
1003 six_unlock_type(&b->c.lock, lock_type);
1004 if (bch2_btree_node_relock(trans, path, level + 1))
1005 return __bch2_btree_node_get(trans, path, k, level, lock_type, trace_ip);
1007 trace_and_count(c, trans_restart_btree_node_reused, trans, trace_ip, path);
1008 return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_lock_node_reused));
1011 if (unlikely(btree_node_read_in_flight(b))) {
1012 u32 seq = six_lock_seq(&b->c.lock);
1014 six_unlock_type(&b->c.lock, lock_type);
1015 bch2_trans_unlock(trans);
1017 bch2_btree_node_wait_on_read(b);
1020 * should_be_locked is not set on this path yet, so we need to
1021 * relock it specifically:
1024 int ret = bch2_trans_relock(trans) ?:
1025 bch2_btree_path_relock_intent(trans, path);
1027 BUG_ON(!trans->restarted);
1028 return ERR_PTR(ret);
1032 if (!six_relock_type(&b->c.lock, lock_type, seq))
1033 return __bch2_btree_node_get(trans, path, k, level, lock_type, trace_ip);
1036 prefetch(b->aux_data);
1038 for_each_bset(b, t) {
1039 void *p = (u64 *) b->aux_data + t->aux_data_offset;
1041 prefetch(p + L1_CACHE_BYTES * 0);
1042 prefetch(p + L1_CACHE_BYTES * 1);
1043 prefetch(p + L1_CACHE_BYTES * 2);
1046 /* avoid atomic set bit if it's not needed: */
1047 if (!btree_node_accessed(b))
1048 set_btree_node_accessed(b);
1050 if (unlikely(btree_node_read_error(b))) {
1051 six_unlock_type(&b->c.lock, lock_type);
1052 return ERR_PTR(-EIO);
1055 EBUG_ON(b->c.btree_id != path->btree_id);
1056 EBUG_ON(BTREE_NODE_LEVEL(b->data) != level);
1057 btree_check_header(c, b);
1062 struct btree *bch2_btree_node_get_noiter(struct btree_trans *trans,
1063 const struct bkey_i *k,
1064 enum btree_id btree_id,
1068 struct bch_fs *c = trans->c;
1069 struct btree_cache *bc = &c->btree_cache;
1071 struct bset_tree *t;
1074 EBUG_ON(level >= BTREE_MAX_DEPTH);
1076 if (c->opts.btree_node_mem_ptr_optimization) {
1077 b = btree_node_mem_ptr(k);
1082 b = btree_cache_find(bc, k);
1087 b = bch2_btree_node_fill(trans, NULL, k, btree_id,
1088 level, SIX_LOCK_read, true);
1090 /* We raced and found the btree node in the cache */
1095 !bch2_btree_cache_cannibalize_lock(c, NULL))
1102 ret = btree_node_lock_nopath(trans, &b->c, SIX_LOCK_read, _THIS_IP_);
1103 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
1104 return ERR_PTR(ret);
1108 if (unlikely(b->hash_val != btree_ptr_hash_val(k) ||
1109 b->c.btree_id != btree_id ||
1110 b->c.level != level)) {
1111 six_unlock_read(&b->c.lock);
1116 /* XXX: waiting on IO with btree locks held: */
1117 __bch2_btree_node_wait_on_read(b);
1119 prefetch(b->aux_data);
1121 for_each_bset(b, t) {
1122 void *p = (u64 *) b->aux_data + t->aux_data_offset;
1124 prefetch(p + L1_CACHE_BYTES * 0);
1125 prefetch(p + L1_CACHE_BYTES * 1);
1126 prefetch(p + L1_CACHE_BYTES * 2);
1129 /* avoid atomic set bit if it's not needed: */
1130 if (!btree_node_accessed(b))
1131 set_btree_node_accessed(b);
1133 if (unlikely(btree_node_read_error(b))) {
1134 six_unlock_read(&b->c.lock);
1139 EBUG_ON(b->c.btree_id != btree_id);
1140 EBUG_ON(BTREE_NODE_LEVEL(b->data) != level);
1141 btree_check_header(c, b);
1143 bch2_btree_cache_cannibalize_unlock(c);
1147 int bch2_btree_node_prefetch(struct btree_trans *trans,
1148 struct btree_path *path,
1149 const struct bkey_i *k,
1150 enum btree_id btree_id, unsigned level)
1152 struct bch_fs *c = trans->c;
1153 struct btree_cache *bc = &c->btree_cache;
1156 BUG_ON(trans && !btree_node_locked(path, level + 1));
1157 BUG_ON(level >= BTREE_MAX_DEPTH);
1159 b = btree_cache_find(bc, k);
1163 b = bch2_btree_node_fill(trans, path, k, btree_id,
1164 level, SIX_LOCK_read, false);
1165 return PTR_ERR_OR_ZERO(b);
1168 void bch2_btree_node_evict(struct btree_trans *trans, const struct bkey_i *k)
1170 struct bch_fs *c = trans->c;
1171 struct btree_cache *bc = &c->btree_cache;
1174 b = btree_cache_find(bc, k);
1178 /* not allowed to wait on io with btree locks held: */
1180 /* XXX we're called from btree_gc which will be holding other btree
1183 __bch2_btree_node_wait_on_read(b);
1184 __bch2_btree_node_wait_on_write(b);
1186 btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
1187 btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_write);
1189 if (btree_node_dirty(b)) {
1190 __bch2_btree_node_write(c, b, BTREE_WRITE_cache_reclaim);
1191 six_unlock_write(&b->c.lock);
1192 six_unlock_intent(&b->c.lock);
1196 BUG_ON(btree_node_dirty(b));
1198 mutex_lock(&bc->lock);
1199 btree_node_data_free(c, b);
1200 bch2_btree_node_hash_remove(bc, b);
1201 mutex_unlock(&bc->lock);
1203 six_unlock_write(&b->c.lock);
1204 six_unlock_intent(&b->c.lock);
1207 void bch2_btree_node_to_text(struct printbuf *out, struct bch_fs *c,
1208 const struct btree *b)
1210 const struct bkey_format *f = &b->format;
1211 struct bset_stats stats;
1213 memset(&stats, 0, sizeof(stats));
1215 bch2_btree_keys_stats(b, &stats);
1217 prt_printf(out, "l %u ", b->c.level);
1218 bch2_bpos_to_text(out, b->data->min_key);
1219 prt_printf(out, " - ");
1220 bch2_bpos_to_text(out, b->data->max_key);
1221 prt_printf(out, ":\n"
1223 bch2_val_to_text(out, c, bkey_i_to_s_c(&b->key));
1225 prt_printf(out, "\n"
1226 " format: u64s %u fields %u %u %u %u %u\n"
1227 " unpack fn len: %u\n"
1228 " bytes used %zu/%zu (%zu%% full)\n"
1229 " sib u64s: %u, %u (merge threshold %u)\n"
1230 " nr packed keys %u\n"
1231 " nr unpacked keys %u\n"
1233 " failed unpacked %zu\n",
1235 f->bits_per_field[0],
1236 f->bits_per_field[1],
1237 f->bits_per_field[2],
1238 f->bits_per_field[3],
1239 f->bits_per_field[4],
1241 b->nr.live_u64s * sizeof(u64),
1242 btree_bytes(c) - sizeof(struct btree_node),
1243 b->nr.live_u64s * 100 / btree_max_u64s(c),
1246 c->btree_foreground_merge_threshold,
1248 b->nr.unpacked_keys,
1253 void bch2_btree_cache_to_text(struct printbuf *out, const struct btree_cache *bc)
1255 prt_printf(out, "nr nodes:\t\t%u\n", bc->used);
1256 prt_printf(out, "nr dirty:\t\t%u\n", atomic_read(&bc->dirty));
1257 prt_printf(out, "cannibalize lock:\t%p\n", bc->alloc_lock);
1259 prt_printf(out, "freed:\t\t\t\t%u\n", bc->freed);
1260 prt_printf(out, "not freed, dirty:\t\t%u\n", bc->not_freed_dirty);
1261 prt_printf(out, "not freed, write in flight:\t%u\n", bc->not_freed_write_in_flight);
1262 prt_printf(out, "not freed, read in flight:\t%u\n", bc->not_freed_read_in_flight);
1263 prt_printf(out, "not freed, lock intent failed:\t%u\n", bc->not_freed_lock_intent);
1264 prt_printf(out, "not freed, lock write failed:\t%u\n", bc->not_freed_lock_write);
1265 prt_printf(out, "not freed, access bit:\t\t%u\n", bc->not_freed_access_bit);
1266 prt_printf(out, "not freed, no evict failed:\t%u\n", bc->not_freed_noevict);
1267 prt_printf(out, "not freed, write blocked:\t%u\n", bc->not_freed_write_blocked);
1268 prt_printf(out, "not freed, will make reachable:\t%u\n", bc->not_freed_will_make_reachable);