aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/md/bcache/btree.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/md/bcache/btree.c')
-rw-r--r--drivers/md/bcache/btree.c2503
1 files changed, 2503 insertions, 0 deletions
diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c
new file mode 100644
index 000000000000..7a5658f04e62
--- /dev/null
+++ b/drivers/md/bcache/btree.c
@@ -0,0 +1,2503 @@
1/*
2 * Copyright (C) 2010 Kent Overstreet <kent.overstreet@gmail.com>
3 *
4 * Uses a block device as cache for other block devices; optimized for SSDs.
5 * All allocation is done in buckets, which should match the erase block size
6 * of the device.
7 *
8 * Buckets containing cached data are kept on a heap sorted by priority;
9 * bucket priority is increased on cache hit, and periodically all the buckets
10 * on the heap have their priority scaled down. This currently is just used as
11 * an LRU but in the future should allow for more intelligent heuristics.
12 *
13 * Buckets have an 8 bit counter; freeing is accomplished by incrementing the
14 * counter. Garbage collection is used to remove stale pointers.
15 *
16 * Indexing is done via a btree; nodes are not necessarily fully sorted, rather
17 * as keys are inserted we only sort the pages that have not yet been written.
18 * When garbage collection is run, we resort the entire node.
19 *
20 * All configuration is done via sysfs; see Documentation/bcache.txt.
21 */
22
23#include "bcache.h"
24#include "btree.h"
25#include "debug.h"
26#include "request.h"
27
28#include <linux/slab.h>
29#include <linux/bitops.h>
30#include <linux/hash.h>
31#include <linux/prefetch.h>
32#include <linux/random.h>
33#include <linux/rcupdate.h>
34#include <trace/events/bcache.h>
35
36/*
37 * Todo:
38 * register_bcache: Return errors out to userspace correctly
39 *
40 * Writeback: don't undirty key until after a cache flush
41 *
42 * Create an iterator for key pointers
43 *
44 * On btree write error, mark bucket such that it won't be freed from the cache
45 *
46 * Journalling:
47 * Check for bad keys in replay
48 * Propagate barriers
49 * Refcount journal entries in journal_replay
50 *
51 * Garbage collection:
52 * Finish incremental gc
53 * Gc should free old UUIDs, data for invalid UUIDs
54 *
55 * Provide a way to list backing device UUIDs we have data cached for, and
56 * probably how long it's been since we've seen them, and a way to invalidate
57 * dirty data for devices that will never be attached again
58 *
59 * Keep 1 min/5 min/15 min statistics of how busy a block device has been, so
60 * that based on that and how much dirty data we have we can keep writeback
61 * from being starved
62 *
63 * Add a tracepoint or somesuch to watch for writeback starvation
64 *
65 * When btree depth > 1 and splitting an interior node, we have to make sure
66 * alloc_bucket() cannot fail. This should be true but is not completely
67 * obvious.
68 *
69 * Make sure all allocations get charged to the root cgroup
70 *
71 * Plugging?
72 *
73 * If data write is less than hard sector size of ssd, round up offset in open
74 * bucket to the next whole sector
75 *
76 * Also lookup by cgroup in get_open_bucket()
77 *
78 * Superblock needs to be fleshed out for multiple cache devices
79 *
80 * Add a sysfs tunable for the number of writeback IOs in flight
81 *
82 * Add a sysfs tunable for the number of open data buckets
83 *
84 * IO tracking: Can we track when one process is doing io on behalf of another?
85 * IO tracking: Don't use just an average, weigh more recent stuff higher
86 *
87 * Test module load/unload
88 */
89
90static const char * const op_types[] = {
91 "insert", "replace"
92};
93
94static const char *op_type(struct btree_op *op)
95{
96 return op_types[op->type];
97}
98
99#define MAX_NEED_GC 64
100#define MAX_SAVE_PRIO 72
101
102#define PTR_DIRTY_BIT (((uint64_t) 1 << 36))
103
104#define PTR_HASH(c, k) \
105 (((k)->ptr[0] >> c->bucket_bits) | PTR_GEN(k, 0))
106
107struct workqueue_struct *bch_gc_wq;
108static struct workqueue_struct *btree_io_wq;
109
110void bch_btree_op_init_stack(struct btree_op *op)
111{
112 memset(op, 0, sizeof(struct btree_op));
113 closure_init_stack(&op->cl);
114 op->lock = -1;
115 bch_keylist_init(&op->keys);
116}
117
118/* Btree key manipulation */
119
120static void bkey_put(struct cache_set *c, struct bkey *k, int level)
121{
122 if ((level && KEY_OFFSET(k)) || !level)
123 __bkey_put(c, k);
124}
125
126/* Btree IO */
127
128static uint64_t btree_csum_set(struct btree *b, struct bset *i)
129{
130 uint64_t crc = b->key.ptr[0];
131 void *data = (void *) i + 8, *end = end(i);
132
133 crc = bch_crc64_update(crc, data, end - data);
134 return crc ^ 0xffffffffffffffffULL;
135}
136
137static void btree_bio_endio(struct bio *bio, int error)
138{
139 struct closure *cl = bio->bi_private;
140 struct btree *b = container_of(cl, struct btree, io.cl);
141
142 if (error)
143 set_btree_node_io_error(b);
144
145 bch_bbio_count_io_errors(b->c, bio, error, (bio->bi_rw & WRITE)
146 ? "writing btree" : "reading btree");
147 closure_put(cl);
148}
149
150static void btree_bio_init(struct btree *b)
151{
152 BUG_ON(b->bio);
153 b->bio = bch_bbio_alloc(b->c);
154
155 b->bio->bi_end_io = btree_bio_endio;
156 b->bio->bi_private = &b->io.cl;
157}
158
159void bch_btree_read_done(struct closure *cl)
160{
161 struct btree *b = container_of(cl, struct btree, io.cl);
162 struct bset *i = b->sets[0].data;
163 struct btree_iter *iter = b->c->fill_iter;
164 const char *err = "bad btree header";
165 BUG_ON(b->nsets || b->written);
166
167 bch_bbio_free(b->bio, b->c);
168 b->bio = NULL;
169
170 mutex_lock(&b->c->fill_lock);
171 iter->used = 0;
172
173 if (btree_node_io_error(b) ||
174 !i->seq)
175 goto err;
176
177 for (;
178 b->written < btree_blocks(b) && i->seq == b->sets[0].data->seq;
179 i = write_block(b)) {
180 err = "unsupported bset version";
181 if (i->version > BCACHE_BSET_VERSION)
182 goto err;
183
184 err = "bad btree header";
185 if (b->written + set_blocks(i, b->c) > btree_blocks(b))
186 goto err;
187
188 err = "bad magic";
189 if (i->magic != bset_magic(b->c))
190 goto err;
191
192 err = "bad checksum";
193 switch (i->version) {
194 case 0:
195 if (i->csum != csum_set(i))
196 goto err;
197 break;
198 case BCACHE_BSET_VERSION:
199 if (i->csum != btree_csum_set(b, i))
200 goto err;
201 break;
202 }
203
204 err = "empty set";
205 if (i != b->sets[0].data && !i->keys)
206 goto err;
207
208 bch_btree_iter_push(iter, i->start, end(i));
209
210 b->written += set_blocks(i, b->c);
211 }
212
213 err = "corrupted btree";
214 for (i = write_block(b);
215 index(i, b) < btree_blocks(b);
216 i = ((void *) i) + block_bytes(b->c))
217 if (i->seq == b->sets[0].data->seq)
218 goto err;
219
220 bch_btree_sort_and_fix_extents(b, iter);
221
222 i = b->sets[0].data;
223 err = "short btree key";
224 if (b->sets[0].size &&
225 bkey_cmp(&b->key, &b->sets[0].end) < 0)
226 goto err;
227
228 if (b->written < btree_blocks(b))
229 bch_bset_init_next(b);
230out:
231
232 mutex_unlock(&b->c->fill_lock);
233
234 spin_lock(&b->c->btree_read_time_lock);
235 bch_time_stats_update(&b->c->btree_read_time, b->io_start_time);
236 spin_unlock(&b->c->btree_read_time_lock);
237
238 smp_wmb(); /* read_done is our write lock */
239 set_btree_node_read_done(b);
240
241 closure_return(cl);
242err:
243 set_btree_node_io_error(b);
244 bch_cache_set_error(b->c, "%s at bucket %zu, block %zu, %u keys",
245 err, PTR_BUCKET_NR(b->c, &b->key, 0),
246 index(i, b), i->keys);
247 goto out;
248}
249
250void bch_btree_read(struct btree *b)
251{
252 BUG_ON(b->nsets || b->written);
253
254 if (!closure_trylock(&b->io.cl, &b->c->cl))
255 BUG();
256
257 b->io_start_time = local_clock();
258
259 btree_bio_init(b);
260 b->bio->bi_rw = REQ_META|READ_SYNC;
261 b->bio->bi_size = KEY_SIZE(&b->key) << 9;
262
263 bch_bio_map(b->bio, b->sets[0].data);
264
265 pr_debug("%s", pbtree(b));
266 trace_bcache_btree_read(b->bio);
267 bch_submit_bbio(b->bio, b->c, &b->key, 0);
268
269 continue_at(&b->io.cl, bch_btree_read_done, system_wq);
270}
271
272static void btree_complete_write(struct btree *b, struct btree_write *w)
273{
274 if (w->prio_blocked &&
275 !atomic_sub_return(w->prio_blocked, &b->c->prio_blocked))
276 wake_up(&b->c->alloc_wait);
277
278 if (w->journal) {
279 atomic_dec_bug(w->journal);
280 __closure_wake_up(&b->c->journal.wait);
281 }
282
283 if (w->owner)
284 closure_put(w->owner);
285
286 w->prio_blocked = 0;
287 w->journal = NULL;
288 w->owner = NULL;
289}
290
291static void __btree_write_done(struct closure *cl)
292{
293 struct btree *b = container_of(cl, struct btree, io.cl);
294 struct btree_write *w = btree_prev_write(b);
295
296 bch_bbio_free(b->bio, b->c);
297 b->bio = NULL;
298 btree_complete_write(b, w);
299
300 if (btree_node_dirty(b))
301 queue_delayed_work(btree_io_wq, &b->work,
302 msecs_to_jiffies(30000));
303
304 closure_return(cl);
305}
306
307static void btree_write_done(struct closure *cl)
308{
309 struct btree *b = container_of(cl, struct btree, io.cl);
310 struct bio_vec *bv;
311 int n;
312
313 __bio_for_each_segment(bv, b->bio, n, 0)
314 __free_page(bv->bv_page);
315
316 __btree_write_done(cl);
317}
318
319static void do_btree_write(struct btree *b)
320{
321 struct closure *cl = &b->io.cl;
322 struct bset *i = b->sets[b->nsets].data;
323 BKEY_PADDED(key) k;
324
325 i->version = BCACHE_BSET_VERSION;
326 i->csum = btree_csum_set(b, i);
327
328 btree_bio_init(b);
329 b->bio->bi_rw = REQ_META|WRITE_SYNC;
330 b->bio->bi_size = set_blocks(i, b->c) * block_bytes(b->c);
331 bch_bio_map(b->bio, i);
332
333 bkey_copy(&k.key, &b->key);
334 SET_PTR_OFFSET(&k.key, 0, PTR_OFFSET(&k.key, 0) + bset_offset(b, i));
335
336 if (!bch_bio_alloc_pages(b->bio, GFP_NOIO)) {
337 int j;
338 struct bio_vec *bv;
339 void *base = (void *) ((unsigned long) i & ~(PAGE_SIZE - 1));
340
341 bio_for_each_segment(bv, b->bio, j)
342 memcpy(page_address(bv->bv_page),
343 base + j * PAGE_SIZE, PAGE_SIZE);
344
345 trace_bcache_btree_write(b->bio);
346 bch_submit_bbio(b->bio, b->c, &k.key, 0);
347
348 continue_at(cl, btree_write_done, NULL);
349 } else {
350 b->bio->bi_vcnt = 0;
351 bch_bio_map(b->bio, i);
352
353 trace_bcache_btree_write(b->bio);
354 bch_submit_bbio(b->bio, b->c, &k.key, 0);
355
356 closure_sync(cl);
357 __btree_write_done(cl);
358 }
359}
360
361static void __btree_write(struct btree *b)
362{
363 struct bset *i = b->sets[b->nsets].data;
364
365 BUG_ON(current->bio_list);
366
367 closure_lock(&b->io, &b->c->cl);
368 cancel_delayed_work(&b->work);
369
370 clear_bit(BTREE_NODE_dirty, &b->flags);
371 change_bit(BTREE_NODE_write_idx, &b->flags);
372
373 bch_check_key_order(b, i);
374 BUG_ON(b->written && !i->keys);
375
376 do_btree_write(b);
377
378 pr_debug("%s block %i keys %i", pbtree(b), b->written, i->keys);
379
380 b->written += set_blocks(i, b->c);
381 atomic_long_add(set_blocks(i, b->c) * b->c->sb.block_size,
382 &PTR_CACHE(b->c, &b->key, 0)->btree_sectors_written);
383
384 bch_btree_sort_lazy(b);
385
386 if (b->written < btree_blocks(b))
387 bch_bset_init_next(b);
388}
389
390static void btree_write_work(struct work_struct *w)
391{
392 struct btree *b = container_of(to_delayed_work(w), struct btree, work);
393
394 down_write(&b->lock);
395
396 if (btree_node_dirty(b))
397 __btree_write(b);
398 up_write(&b->lock);
399}
400
401void bch_btree_write(struct btree *b, bool now, struct btree_op *op)
402{
403 struct bset *i = b->sets[b->nsets].data;
404 struct btree_write *w = btree_current_write(b);
405
406 BUG_ON(b->written &&
407 (b->written >= btree_blocks(b) ||
408 i->seq != b->sets[0].data->seq ||
409 !i->keys));
410
411 if (!btree_node_dirty(b)) {
412 set_btree_node_dirty(b);
413 queue_delayed_work(btree_io_wq, &b->work,
414 msecs_to_jiffies(30000));
415 }
416
417 w->prio_blocked += b->prio_blocked;
418 b->prio_blocked = 0;
419
420 if (op && op->journal && !b->level) {
421 if (w->journal &&
422 journal_pin_cmp(b->c, w, op)) {
423 atomic_dec_bug(w->journal);
424 w->journal = NULL;
425 }
426
427 if (!w->journal) {
428 w->journal = op->journal;
429 atomic_inc(w->journal);
430 }
431 }
432
433 if (current->bio_list)
434 return;
435
436 /* Force write if set is too big */
437 if (now ||
438 b->level ||
439 set_bytes(i) > PAGE_SIZE - 48) {
440 if (op && now) {
441 /* Must wait on multiple writes */
442 BUG_ON(w->owner);
443 w->owner = &op->cl;
444 closure_get(&op->cl);
445 }
446
447 __btree_write(b);
448 }
449 BUG_ON(!b->written);
450}
451
452/*
453 * Btree in memory cache - allocation/freeing
454 * mca -> memory cache
455 */
456
457static void mca_reinit(struct btree *b)
458{
459 unsigned i;
460
461 b->flags = 0;
462 b->written = 0;
463 b->nsets = 0;
464
465 for (i = 0; i < MAX_BSETS; i++)
466 b->sets[i].size = 0;
467 /*
468 * Second loop starts at 1 because b->sets[0]->data is the memory we
469 * allocated
470 */
471 for (i = 1; i < MAX_BSETS; i++)
472 b->sets[i].data = NULL;
473}
474
475#define mca_reserve(c) (((c->root && c->root->level) \
476 ? c->root->level : 1) * 8 + 16)
477#define mca_can_free(c) \
478 max_t(int, 0, c->bucket_cache_used - mca_reserve(c))
479
480static void mca_data_free(struct btree *b)
481{
482 struct bset_tree *t = b->sets;
483 BUG_ON(!closure_is_unlocked(&b->io.cl));
484
485 if (bset_prev_bytes(b) < PAGE_SIZE)
486 kfree(t->prev);
487 else
488 free_pages((unsigned long) t->prev,
489 get_order(bset_prev_bytes(b)));
490
491 if (bset_tree_bytes(b) < PAGE_SIZE)
492 kfree(t->tree);
493 else
494 free_pages((unsigned long) t->tree,
495 get_order(bset_tree_bytes(b)));
496
497 free_pages((unsigned long) t->data, b->page_order);
498
499 t->prev = NULL;
500 t->tree = NULL;
501 t->data = NULL;
502 list_move(&b->list, &b->c->btree_cache_freed);
503 b->c->bucket_cache_used--;
504}
505
506static void mca_bucket_free(struct btree *b)
507{
508 BUG_ON(btree_node_dirty(b));
509
510 b->key.ptr[0] = 0;
511 hlist_del_init_rcu(&b->hash);
512 list_move(&b->list, &b->c->btree_cache_freeable);
513}
514
515static unsigned btree_order(struct bkey *k)
516{
517 return ilog2(KEY_SIZE(k) / PAGE_SECTORS ?: 1);
518}
519
520static void mca_data_alloc(struct btree *b, struct bkey *k, gfp_t gfp)
521{
522 struct bset_tree *t = b->sets;
523 BUG_ON(t->data);
524
525 b->page_order = max_t(unsigned,
526 ilog2(b->c->btree_pages),
527 btree_order(k));
528
529 t->data = (void *) __get_free_pages(gfp, b->page_order);
530 if (!t->data)
531 goto err;
532
533 t->tree = bset_tree_bytes(b) < PAGE_SIZE
534 ? kmalloc(bset_tree_bytes(b), gfp)
535 : (void *) __get_free_pages(gfp, get_order(bset_tree_bytes(b)));
536 if (!t->tree)
537 goto err;
538
539 t->prev = bset_prev_bytes(b) < PAGE_SIZE
540 ? kmalloc(bset_prev_bytes(b), gfp)
541 : (void *) __get_free_pages(gfp, get_order(bset_prev_bytes(b)));
542 if (!t->prev)
543 goto err;
544
545 list_move(&b->list, &b->c->btree_cache);
546 b->c->bucket_cache_used++;
547 return;
548err:
549 mca_data_free(b);
550}
551
552static struct btree *mca_bucket_alloc(struct cache_set *c,
553 struct bkey *k, gfp_t gfp)
554{
555 struct btree *b = kzalloc(sizeof(struct btree), gfp);
556 if (!b)
557 return NULL;
558
559 init_rwsem(&b->lock);
560 lockdep_set_novalidate_class(&b->lock);
561 INIT_LIST_HEAD(&b->list);
562 INIT_DELAYED_WORK(&b->work, btree_write_work);
563 b->c = c;
564 closure_init_unlocked(&b->io);
565
566 mca_data_alloc(b, k, gfp);
567 return b;
568}
569
570static int mca_reap(struct btree *b, struct closure *cl, unsigned min_order)
571{
572 lockdep_assert_held(&b->c->bucket_lock);
573
574 if (!down_write_trylock(&b->lock))
575 return -ENOMEM;
576
577 if (b->page_order < min_order) {
578 rw_unlock(true, b);
579 return -ENOMEM;
580 }
581
582 BUG_ON(btree_node_dirty(b) && !b->sets[0].data);
583
584 if (cl && btree_node_dirty(b))
585 bch_btree_write(b, true, NULL);
586
587 if (cl)
588 closure_wait_event_async(&b->io.wait, cl,
589 atomic_read(&b->io.cl.remaining) == -1);
590
591 if (btree_node_dirty(b) ||
592 !closure_is_unlocked(&b->io.cl) ||
593 work_pending(&b->work.work)) {
594 rw_unlock(true, b);
595 return -EAGAIN;
596 }
597
598 return 0;
599}
600
601static int bch_mca_shrink(struct shrinker *shrink, struct shrink_control *sc)
602{
603 struct cache_set *c = container_of(shrink, struct cache_set, shrink);
604 struct btree *b, *t;
605 unsigned long i, nr = sc->nr_to_scan;
606
607 if (c->shrinker_disabled)
608 return 0;
609
610 if (c->try_harder)
611 return 0;
612
613 /*
614 * If nr == 0, we're supposed to return the number of items we have
615 * cached. Not allowed to return -1.
616 */
617 if (!nr)
618 return mca_can_free(c) * c->btree_pages;
619
620 /* Return -1 if we can't do anything right now */
621 if (sc->gfp_mask & __GFP_WAIT)
622 mutex_lock(&c->bucket_lock);
623 else if (!mutex_trylock(&c->bucket_lock))
624 return -1;
625
626 nr /= c->btree_pages;
627 nr = min_t(unsigned long, nr, mca_can_free(c));
628
629 i = 0;
630 list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) {
631 if (!nr)
632 break;
633
634 if (++i > 3 &&
635 !mca_reap(b, NULL, 0)) {
636 mca_data_free(b);
637 rw_unlock(true, b);
638 --nr;
639 }
640 }
641
642 /*
643 * Can happen right when we first start up, before we've read in any
644 * btree nodes
645 */
646 if (list_empty(&c->btree_cache))
647 goto out;
648
649 for (i = 0; nr && i < c->bucket_cache_used; i++) {
650 b = list_first_entry(&c->btree_cache, struct btree, list);
651 list_rotate_left(&c->btree_cache);
652
653 if (!b->accessed &&
654 !mca_reap(b, NULL, 0)) {
655 mca_bucket_free(b);
656 mca_data_free(b);
657 rw_unlock(true, b);
658 --nr;
659 } else
660 b->accessed = 0;
661 }
662out:
663 nr = mca_can_free(c) * c->btree_pages;
664 mutex_unlock(&c->bucket_lock);
665 return nr;
666}
667
668void bch_btree_cache_free(struct cache_set *c)
669{
670 struct btree *b;
671 struct closure cl;
672 closure_init_stack(&cl);
673
674 if (c->shrink.list.next)
675 unregister_shrinker(&c->shrink);
676
677 mutex_lock(&c->bucket_lock);
678
679#ifdef CONFIG_BCACHE_DEBUG
680 if (c->verify_data)
681 list_move(&c->verify_data->list, &c->btree_cache);
682#endif
683
684 list_splice(&c->btree_cache_freeable,
685 &c->btree_cache);
686
687 while (!list_empty(&c->btree_cache)) {
688 b = list_first_entry(&c->btree_cache, struct btree, list);
689
690 if (btree_node_dirty(b))
691 btree_complete_write(b, btree_current_write(b));
692 clear_bit(BTREE_NODE_dirty, &b->flags);
693
694 mca_data_free(b);
695 }
696
697 while (!list_empty(&c->btree_cache_freed)) {
698 b = list_first_entry(&c->btree_cache_freed,
699 struct btree, list);
700 list_del(&b->list);
701 cancel_delayed_work_sync(&b->work);
702 kfree(b);
703 }
704
705 mutex_unlock(&c->bucket_lock);
706}
707
708int bch_btree_cache_alloc(struct cache_set *c)
709{
710 unsigned i;
711
712 /* XXX: doesn't check for errors */
713
714 closure_init_unlocked(&c->gc);
715
716 for (i = 0; i < mca_reserve(c); i++)
717 mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL);
718
719 list_splice_init(&c->btree_cache,
720 &c->btree_cache_freeable);
721
722#ifdef CONFIG_BCACHE_DEBUG
723 mutex_init(&c->verify_lock);
724
725 c->verify_data = mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL);
726
727 if (c->verify_data &&
728 c->verify_data->sets[0].data)
729 list_del_init(&c->verify_data->list);
730 else
731 c->verify_data = NULL;
732#endif
733
734 c->shrink.shrink = bch_mca_shrink;
735 c->shrink.seeks = 4;
736 c->shrink.batch = c->btree_pages * 2;
737 register_shrinker(&c->shrink);
738
739 return 0;
740}
741
742/* Btree in memory cache - hash table */
743
744static struct hlist_head *mca_hash(struct cache_set *c, struct bkey *k)
745{
746 return &c->bucket_hash[hash_32(PTR_HASH(c, k), BUCKET_HASH_BITS)];
747}
748
749static struct btree *mca_find(struct cache_set *c, struct bkey *k)
750{
751 struct btree *b;
752
753 rcu_read_lock();
754 hlist_for_each_entry_rcu(b, mca_hash(c, k), hash)
755 if (PTR_HASH(c, &b->key) == PTR_HASH(c, k))
756 goto out;
757 b = NULL;
758out:
759 rcu_read_unlock();
760 return b;
761}
762
763static struct btree *mca_cannibalize(struct cache_set *c, struct bkey *k,
764 int level, struct closure *cl)
765{
766 int ret = -ENOMEM;
767 struct btree *i;
768
769 if (!cl)
770 return ERR_PTR(-ENOMEM);
771
772 /*
773 * Trying to free up some memory - i.e. reuse some btree nodes - may
774 * require initiating IO to flush the dirty part of the node. If we're
775 * running under generic_make_request(), that IO will never finish and
776 * we would deadlock. Returning -EAGAIN causes the cache lookup code to
777 * punt to workqueue and retry.
778 */
779 if (current->bio_list)
780 return ERR_PTR(-EAGAIN);
781
782 if (c->try_harder && c->try_harder != cl) {
783 closure_wait_event_async(&c->try_wait, cl, !c->try_harder);
784 return ERR_PTR(-EAGAIN);
785 }
786
787 /* XXX: tracepoint */
788 c->try_harder = cl;
789 c->try_harder_start = local_clock();
790retry:
791 list_for_each_entry_reverse(i, &c->btree_cache, list) {
792 int r = mca_reap(i, cl, btree_order(k));
793 if (!r)
794 return i;
795 if (r != -ENOMEM)
796 ret = r;
797 }
798
799 if (ret == -EAGAIN &&
800 closure_blocking(cl)) {
801 mutex_unlock(&c->bucket_lock);
802 closure_sync(cl);
803 mutex_lock(&c->bucket_lock);
804 goto retry;
805 }
806
807 return ERR_PTR(ret);
808}
809
810/*
811 * We can only have one thread cannibalizing other cached btree nodes at a time,
812 * or we'll deadlock. We use an open coded mutex to ensure that, which a
813 * cannibalize_bucket() will take. This means every time we unlock the root of
814 * the btree, we need to release this lock if we have it held.
815 */
816void bch_cannibalize_unlock(struct cache_set *c, struct closure *cl)
817{
818 if (c->try_harder == cl) {
819 bch_time_stats_update(&c->try_harder_time, c->try_harder_start);
820 c->try_harder = NULL;
821 __closure_wake_up(&c->try_wait);
822 }
823}
824
825static struct btree *mca_alloc(struct cache_set *c, struct bkey *k,
826 int level, struct closure *cl)
827{
828 struct btree *b;
829
830 lockdep_assert_held(&c->bucket_lock);
831
832 if (mca_find(c, k))
833 return NULL;
834
835 /* btree_free() doesn't free memory; it sticks the node on the end of
836 * the list. Check if there's any freed nodes there:
837 */
838 list_for_each_entry(b, &c->btree_cache_freeable, list)
839 if (!mca_reap(b, NULL, btree_order(k)))
840 goto out;
841
842 /* We never free struct btree itself, just the memory that holds the on
843 * disk node. Check the freed list before allocating a new one:
844 */
845 list_for_each_entry(b, &c->btree_cache_freed, list)
846 if (!mca_reap(b, NULL, 0)) {
847 mca_data_alloc(b, k, __GFP_NOWARN|GFP_NOIO);
848 if (!b->sets[0].data)
849 goto err;
850 else
851 goto out;
852 }
853
854 b = mca_bucket_alloc(c, k, __GFP_NOWARN|GFP_NOIO);
855 if (!b)
856 goto err;
857
858 BUG_ON(!down_write_trylock(&b->lock));
859 if (!b->sets->data)
860 goto err;
861out:
862 BUG_ON(!closure_is_unlocked(&b->io.cl));
863
864 bkey_copy(&b->key, k);
865 list_move(&b->list, &c->btree_cache);
866 hlist_del_init_rcu(&b->hash);
867 hlist_add_head_rcu(&b->hash, mca_hash(c, k));
868
869 lock_set_subclass(&b->lock.dep_map, level + 1, _THIS_IP_);
870 b->level = level;
871
872 mca_reinit(b);
873
874 return b;
875err:
876 if (b)
877 rw_unlock(true, b);
878
879 b = mca_cannibalize(c, k, level, cl);
880 if (!IS_ERR(b))
881 goto out;
882
883 return b;
884}
885
886/**
887 * bch_btree_node_get - find a btree node in the cache and lock it, reading it
888 * in from disk if necessary.
889 *
890 * If IO is necessary, it uses the closure embedded in struct btree_op to wait;
891 * if that closure is in non blocking mode, will return -EAGAIN.
892 *
893 * The btree node will have either a read or a write lock held, depending on
894 * level and op->lock.
895 */
896struct btree *bch_btree_node_get(struct cache_set *c, struct bkey *k,
897 int level, struct btree_op *op)
898{
899 int i = 0;
900 bool write = level <= op->lock;
901 struct btree *b;
902
903 BUG_ON(level < 0);
904retry:
905 b = mca_find(c, k);
906
907 if (!b) {
908 mutex_lock(&c->bucket_lock);
909 b = mca_alloc(c, k, level, &op->cl);
910 mutex_unlock(&c->bucket_lock);
911
912 if (!b)
913 goto retry;
914 if (IS_ERR(b))
915 return b;
916
917 bch_btree_read(b);
918
919 if (!write)
920 downgrade_write(&b->lock);
921 } else {
922 rw_lock(write, b, level);
923 if (PTR_HASH(c, &b->key) != PTR_HASH(c, k)) {
924 rw_unlock(write, b);
925 goto retry;
926 }
927 BUG_ON(b->level != level);
928 }
929
930 b->accessed = 1;
931
932 for (; i <= b->nsets && b->sets[i].size; i++) {
933 prefetch(b->sets[i].tree);
934 prefetch(b->sets[i].data);
935 }
936
937 for (; i <= b->nsets; i++)
938 prefetch(b->sets[i].data);
939
940 if (!closure_wait_event(&b->io.wait, &op->cl,
941 btree_node_read_done(b))) {
942 rw_unlock(write, b);
943 b = ERR_PTR(-EAGAIN);
944 } else if (btree_node_io_error(b)) {
945 rw_unlock(write, b);
946 b = ERR_PTR(-EIO);
947 } else
948 BUG_ON(!b->written);
949
950 return b;
951}
952
953static void btree_node_prefetch(struct cache_set *c, struct bkey *k, int level)
954{
955 struct btree *b;
956
957 mutex_lock(&c->bucket_lock);
958 b = mca_alloc(c, k, level, NULL);
959 mutex_unlock(&c->bucket_lock);
960
961 if (!IS_ERR_OR_NULL(b)) {
962 bch_btree_read(b);
963 rw_unlock(true, b);
964 }
965}
966
967/* Btree alloc */
968
969static void btree_node_free(struct btree *b, struct btree_op *op)
970{
971 unsigned i;
972
973 /*
974 * The BUG_ON() in btree_node_get() implies that we must have a write
975 * lock on parent to free or even invalidate a node
976 */
977 BUG_ON(op->lock <= b->level);
978 BUG_ON(b == b->c->root);
979 pr_debug("bucket %s", pbtree(b));
980
981 if (btree_node_dirty(b))
982 btree_complete_write(b, btree_current_write(b));
983 clear_bit(BTREE_NODE_dirty, &b->flags);
984
985 if (b->prio_blocked &&
986 !atomic_sub_return(b->prio_blocked, &b->c->prio_blocked))
987 wake_up(&b->c->alloc_wait);
988
989 b->prio_blocked = 0;
990
991 cancel_delayed_work(&b->work);
992
993 mutex_lock(&b->c->bucket_lock);
994
995 for (i = 0; i < KEY_PTRS(&b->key); i++) {
996 BUG_ON(atomic_read(&PTR_BUCKET(b->c, &b->key, i)->pin));
997
998 bch_inc_gen(PTR_CACHE(b->c, &b->key, i),
999 PTR_BUCKET(b->c, &b->key, i));
1000 }
1001
1002 bch_bucket_free(b->c, &b->key);
1003 mca_bucket_free(b);
1004 mutex_unlock(&b->c->bucket_lock);
1005}
1006
1007struct btree *bch_btree_node_alloc(struct cache_set *c, int level,
1008 struct closure *cl)
1009{
1010 BKEY_PADDED(key) k;
1011 struct btree *b = ERR_PTR(-EAGAIN);
1012
1013 mutex_lock(&c->bucket_lock);
1014retry:
1015 if (__bch_bucket_alloc_set(c, WATERMARK_METADATA, &k.key, 1, cl))
1016 goto err;
1017
1018 SET_KEY_SIZE(&k.key, c->btree_pages * PAGE_SECTORS);
1019
1020 b = mca_alloc(c, &k.key, level, cl);
1021 if (IS_ERR(b))
1022 goto err_free;
1023
1024 if (!b) {
1025 cache_bug(c,
1026 "Tried to allocate bucket that was in btree cache");
1027 __bkey_put(c, &k.key);
1028 goto retry;
1029 }
1030
1031 set_btree_node_read_done(b);
1032 b->accessed = 1;
1033 bch_bset_init_next(b);
1034
1035 mutex_unlock(&c->bucket_lock);
1036 return b;
1037err_free:
1038 bch_bucket_free(c, &k.key);
1039 __bkey_put(c, &k.key);
1040err:
1041 mutex_unlock(&c->bucket_lock);
1042 return b;
1043}
1044
1045static struct btree *btree_node_alloc_replacement(struct btree *b,
1046 struct closure *cl)
1047{
1048 struct btree *n = bch_btree_node_alloc(b->c, b->level, cl);
1049 if (!IS_ERR_OR_NULL(n))
1050 bch_btree_sort_into(b, n);
1051
1052 return n;
1053}
1054
1055/* Garbage collection */
1056
1057uint8_t __bch_btree_mark_key(struct cache_set *c, int level, struct bkey *k)
1058{
1059 uint8_t stale = 0;
1060 unsigned i;
1061 struct bucket *g;
1062
1063 /*
1064 * ptr_invalid() can't return true for the keys that mark btree nodes as
1065 * freed, but since ptr_bad() returns true we'll never actually use them
1066 * for anything and thus we don't want mark their pointers here
1067 */
1068 if (!bkey_cmp(k, &ZERO_KEY))
1069 return stale;
1070
1071 for (i = 0; i < KEY_PTRS(k); i++) {
1072 if (!ptr_available(c, k, i))
1073 continue;
1074
1075 g = PTR_BUCKET(c, k, i);
1076
1077 if (gen_after(g->gc_gen, PTR_GEN(k, i)))
1078 g->gc_gen = PTR_GEN(k, i);
1079
1080 if (ptr_stale(c, k, i)) {
1081 stale = max(stale, ptr_stale(c, k, i));
1082 continue;
1083 }
1084
1085 cache_bug_on(GC_MARK(g) &&
1086 (GC_MARK(g) == GC_MARK_METADATA) != (level != 0),
1087 c, "inconsistent ptrs: mark = %llu, level = %i",
1088 GC_MARK(g), level);
1089
1090 if (level)
1091 SET_GC_MARK(g, GC_MARK_METADATA);
1092 else if (KEY_DIRTY(k))
1093 SET_GC_MARK(g, GC_MARK_DIRTY);
1094
1095 /* guard against overflow */
1096 SET_GC_SECTORS_USED(g, min_t(unsigned,
1097 GC_SECTORS_USED(g) + KEY_SIZE(k),
1098 (1 << 14) - 1));
1099
1100 BUG_ON(!GC_SECTORS_USED(g));
1101 }
1102
1103 return stale;
1104}
1105
1106#define btree_mark_key(b, k) __bch_btree_mark_key(b->c, b->level, k)
1107
1108static int btree_gc_mark_node(struct btree *b, unsigned *keys,
1109 struct gc_stat *gc)
1110{
1111 uint8_t stale = 0;
1112 unsigned last_dev = -1;
1113 struct bcache_device *d = NULL;
1114 struct bkey *k;
1115 struct btree_iter iter;
1116 struct bset_tree *t;
1117
1118 gc->nodes++;
1119
1120 for_each_key_filter(b, k, &iter, bch_ptr_invalid) {
1121 if (last_dev != KEY_INODE(k)) {
1122 last_dev = KEY_INODE(k);
1123
1124 d = KEY_INODE(k) < b->c->nr_uuids
1125 ? b->c->devices[last_dev]
1126 : NULL;
1127 }
1128
1129 stale = max(stale, btree_mark_key(b, k));
1130
1131 if (bch_ptr_bad(b, k))
1132 continue;
1133
1134 *keys += bkey_u64s(k);
1135
1136 gc->key_bytes += bkey_u64s(k);
1137 gc->nkeys++;
1138
1139 gc->data += KEY_SIZE(k);
1140 if (KEY_DIRTY(k)) {
1141 gc->dirty += KEY_SIZE(k);
1142 if (d)
1143 d->sectors_dirty_gc += KEY_SIZE(k);
1144 }
1145 }
1146
1147 for (t = b->sets; t <= &b->sets[b->nsets]; t++)
1148 btree_bug_on(t->size &&
1149 bset_written(b, t) &&
1150 bkey_cmp(&b->key, &t->end) < 0,
1151 b, "found short btree key in gc");
1152
1153 return stale;
1154}
1155
1156static struct btree *btree_gc_alloc(struct btree *b, struct bkey *k,
1157 struct btree_op *op)
1158{
1159 /*
1160 * We block priorities from being written for the duration of garbage
1161 * collection, so we can't sleep in btree_alloc() ->
1162 * bch_bucket_alloc_set(), or we'd risk deadlock - so we don't pass it
1163 * our closure.
1164 */
1165 struct btree *n = btree_node_alloc_replacement(b, NULL);
1166
1167 if (!IS_ERR_OR_NULL(n)) {
1168 swap(b, n);
1169
1170 memcpy(k->ptr, b->key.ptr,
1171 sizeof(uint64_t) * KEY_PTRS(&b->key));
1172
1173 __bkey_put(b->c, &b->key);
1174 atomic_inc(&b->c->prio_blocked);
1175 b->prio_blocked++;
1176
1177 btree_node_free(n, op);
1178 up_write(&n->lock);
1179 }
1180
1181 return b;
1182}
1183
1184/*
1185 * Leaving this at 2 until we've got incremental garbage collection done; it
1186 * could be higher (and has been tested with 4) except that garbage collection
1187 * could take much longer, adversely affecting latency.
1188 */
1189#define GC_MERGE_NODES 2U
1190
1191struct gc_merge_info {
1192 struct btree *b;
1193 struct bkey *k;
1194 unsigned keys;
1195};
1196
1197static void btree_gc_coalesce(struct btree *b, struct btree_op *op,
1198 struct gc_stat *gc, struct gc_merge_info *r)
1199{
1200 unsigned nodes = 0, keys = 0, blocks;
1201 int i;
1202
1203 while (nodes < GC_MERGE_NODES && r[nodes].b)
1204 keys += r[nodes++].keys;
1205
1206 blocks = btree_default_blocks(b->c) * 2 / 3;
1207
1208 if (nodes < 2 ||
1209 __set_blocks(b->sets[0].data, keys, b->c) > blocks * (nodes - 1))
1210 return;
1211
1212 for (i = nodes - 1; i >= 0; --i) {
1213 if (r[i].b->written)
1214 r[i].b = btree_gc_alloc(r[i].b, r[i].k, op);
1215
1216 if (r[i].b->written)
1217 return;
1218 }
1219
1220 for (i = nodes - 1; i > 0; --i) {
1221 struct bset *n1 = r[i].b->sets->data;
1222 struct bset *n2 = r[i - 1].b->sets->data;
1223 struct bkey *k, *last = NULL;
1224
1225 keys = 0;
1226
1227 if (i == 1) {
1228 /*
1229 * Last node we're not getting rid of - we're getting
1230 * rid of the node at r[0]. Have to try and fit all of
1231 * the remaining keys into this node; we can't ensure
1232 * they will always fit due to rounding and variable
1233 * length keys (shouldn't be possible in practice,
1234 * though)
1235 */
1236 if (__set_blocks(n1, n1->keys + r->keys,
1237 b->c) > btree_blocks(r[i].b))
1238 return;
1239
1240 keys = n2->keys;
1241 last = &r->b->key;
1242 } else
1243 for (k = n2->start;
1244 k < end(n2);
1245 k = bkey_next(k)) {
1246 if (__set_blocks(n1, n1->keys + keys +
1247 bkey_u64s(k), b->c) > blocks)
1248 break;
1249
1250 last = k;
1251 keys += bkey_u64s(k);
1252 }
1253
1254 BUG_ON(__set_blocks(n1, n1->keys + keys,
1255 b->c) > btree_blocks(r[i].b));
1256
1257 if (last) {
1258 bkey_copy_key(&r[i].b->key, last);
1259 bkey_copy_key(r[i].k, last);
1260 }
1261
1262 memcpy(end(n1),
1263 n2->start,
1264 (void *) node(n2, keys) - (void *) n2->start);
1265
1266 n1->keys += keys;
1267
1268 memmove(n2->start,
1269 node(n2, keys),
1270 (void *) end(n2) - (void *) node(n2, keys));
1271
1272 n2->keys -= keys;
1273
1274 r[i].keys = n1->keys;
1275 r[i - 1].keys = n2->keys;
1276 }
1277
1278 btree_node_free(r->b, op);
1279 up_write(&r->b->lock);
1280
1281 pr_debug("coalesced %u nodes", nodes);
1282
1283 gc->nodes--;
1284 nodes--;
1285
1286 memmove(&r[0], &r[1], sizeof(struct gc_merge_info) * nodes);
1287 memset(&r[nodes], 0, sizeof(struct gc_merge_info));
1288}
1289
1290static int btree_gc_recurse(struct btree *b, struct btree_op *op,
1291 struct closure *writes, struct gc_stat *gc)
1292{
1293 void write(struct btree *r)
1294 {
1295 if (!r->written)
1296 bch_btree_write(r, true, op);
1297 else if (btree_node_dirty(r)) {
1298 BUG_ON(btree_current_write(r)->owner);
1299 btree_current_write(r)->owner = writes;
1300 closure_get(writes);
1301
1302 bch_btree_write(r, true, NULL);
1303 }
1304
1305 up_write(&r->lock);
1306 }
1307
1308 int ret = 0, stale;
1309 unsigned i;
1310 struct gc_merge_info r[GC_MERGE_NODES];
1311
1312 memset(r, 0, sizeof(r));
1313
1314 while ((r->k = bch_next_recurse_key(b, &b->c->gc_done))) {
1315 r->b = bch_btree_node_get(b->c, r->k, b->level - 1, op);
1316
1317 if (IS_ERR(r->b)) {
1318 ret = PTR_ERR(r->b);
1319 break;
1320 }
1321
1322 r->keys = 0;
1323 stale = btree_gc_mark_node(r->b, &r->keys, gc);
1324
1325 if (!b->written &&
1326 (r->b->level || stale > 10 ||
1327 b->c->gc_always_rewrite))
1328 r->b = btree_gc_alloc(r->b, r->k, op);
1329
1330 if (r->b->level)
1331 ret = btree_gc_recurse(r->b, op, writes, gc);
1332
1333 if (ret) {
1334 write(r->b);
1335 break;
1336 }
1337
1338 bkey_copy_key(&b->c->gc_done, r->k);
1339
1340 if (!b->written)
1341 btree_gc_coalesce(b, op, gc, r);
1342
1343 if (r[GC_MERGE_NODES - 1].b)
1344 write(r[GC_MERGE_NODES - 1].b);
1345
1346 memmove(&r[1], &r[0],
1347 sizeof(struct gc_merge_info) * (GC_MERGE_NODES - 1));
1348
1349 /* When we've got incremental GC working, we'll want to do
1350 * if (should_resched())
1351 * return -EAGAIN;
1352 */
1353 cond_resched();
1354#if 0
1355 if (need_resched()) {
1356 ret = -EAGAIN;
1357 break;
1358 }
1359#endif
1360 }
1361
1362 for (i = 1; i < GC_MERGE_NODES && r[i].b; i++)
1363 write(r[i].b);
1364
1365 /* Might have freed some children, must remove their keys */
1366 if (!b->written)
1367 bch_btree_sort(b);
1368
1369 return ret;
1370}
1371
1372static int bch_btree_gc_root(struct btree *b, struct btree_op *op,
1373 struct closure *writes, struct gc_stat *gc)
1374{
1375 struct btree *n = NULL;
1376 unsigned keys = 0;
1377 int ret = 0, stale = btree_gc_mark_node(b, &keys, gc);
1378
1379 if (b->level || stale > 10)
1380 n = btree_node_alloc_replacement(b, NULL);
1381
1382 if (!IS_ERR_OR_NULL(n))
1383 swap(b, n);
1384
1385 if (b->level)
1386 ret = btree_gc_recurse(b, op, writes, gc);
1387
1388 if (!b->written || btree_node_dirty(b)) {
1389 atomic_inc(&b->c->prio_blocked);
1390 b->prio_blocked++;
1391 bch_btree_write(b, true, n ? op : NULL);
1392 }
1393
1394 if (!IS_ERR_OR_NULL(n)) {
1395 closure_sync(&op->cl);
1396 bch_btree_set_root(b);
1397 btree_node_free(n, op);
1398 rw_unlock(true, b);
1399 }
1400
1401 return ret;
1402}
1403
1404static void btree_gc_start(struct cache_set *c)
1405{
1406 struct cache *ca;
1407 struct bucket *b;
1408 struct bcache_device **d;
1409 unsigned i;
1410
1411 if (!c->gc_mark_valid)
1412 return;
1413
1414 mutex_lock(&c->bucket_lock);
1415
1416 c->gc_mark_valid = 0;
1417 c->gc_done = ZERO_KEY;
1418
1419 for_each_cache(ca, c, i)
1420 for_each_bucket(b, ca) {
1421 b->gc_gen = b->gen;
1422 if (!atomic_read(&b->pin))
1423 SET_GC_MARK(b, GC_MARK_RECLAIMABLE);
1424 }
1425
1426 for (d = c->devices;
1427 d < c->devices + c->nr_uuids;
1428 d++)
1429 if (*d)
1430 (*d)->sectors_dirty_gc = 0;
1431
1432 mutex_unlock(&c->bucket_lock);
1433}
1434
1435size_t bch_btree_gc_finish(struct cache_set *c)
1436{
1437 size_t available = 0;
1438 struct bucket *b;
1439 struct cache *ca;
1440 struct bcache_device **d;
1441 unsigned i;
1442
1443 mutex_lock(&c->bucket_lock);
1444
1445 set_gc_sectors(c);
1446 c->gc_mark_valid = 1;
1447 c->need_gc = 0;
1448
1449 if (c->root)
1450 for (i = 0; i < KEY_PTRS(&c->root->key); i++)
1451 SET_GC_MARK(PTR_BUCKET(c, &c->root->key, i),
1452 GC_MARK_METADATA);
1453
1454 for (i = 0; i < KEY_PTRS(&c->uuid_bucket); i++)
1455 SET_GC_MARK(PTR_BUCKET(c, &c->uuid_bucket, i),
1456 GC_MARK_METADATA);
1457
1458 for_each_cache(ca, c, i) {
1459 uint64_t *i;
1460
1461 ca->invalidate_needs_gc = 0;
1462
1463 for (i = ca->sb.d; i < ca->sb.d + ca->sb.keys; i++)
1464 SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
1465
1466 for (i = ca->prio_buckets;
1467 i < ca->prio_buckets + prio_buckets(ca) * 2; i++)
1468 SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
1469
1470 for_each_bucket(b, ca) {
1471 b->last_gc = b->gc_gen;
1472 c->need_gc = max(c->need_gc, bucket_gc_gen(b));
1473
1474 if (!atomic_read(&b->pin) &&
1475 GC_MARK(b) == GC_MARK_RECLAIMABLE) {
1476 available++;
1477 if (!GC_SECTORS_USED(b))
1478 bch_bucket_add_unused(ca, b);
1479 }
1480 }
1481 }
1482
1483 for (d = c->devices;
1484 d < c->devices + c->nr_uuids;
1485 d++)
1486 if (*d) {
1487 unsigned long last =
1488 atomic_long_read(&((*d)->sectors_dirty));
1489 long difference = (*d)->sectors_dirty_gc - last;
1490
1491 pr_debug("sectors dirty off by %li", difference);
1492
1493 (*d)->sectors_dirty_last += difference;
1494
1495 atomic_long_set(&((*d)->sectors_dirty),
1496 (*d)->sectors_dirty_gc);
1497 }
1498
1499 mutex_unlock(&c->bucket_lock);
1500 return available;
1501}
1502
1503static void bch_btree_gc(struct closure *cl)
1504{
1505 struct cache_set *c = container_of(cl, struct cache_set, gc.cl);
1506 int ret;
1507 unsigned long available;
1508 struct gc_stat stats;
1509 struct closure writes;
1510 struct btree_op op;
1511
1512 uint64_t start_time = local_clock();
1513 trace_bcache_gc_start(c->sb.set_uuid);
1514 blktrace_msg_all(c, "Starting gc");
1515
1516 memset(&stats, 0, sizeof(struct gc_stat));
1517 closure_init_stack(&writes);
1518 bch_btree_op_init_stack(&op);
1519 op.lock = SHRT_MAX;
1520
1521 btree_gc_start(c);
1522
1523 ret = btree_root(gc_root, c, &op, &writes, &stats);
1524 closure_sync(&op.cl);
1525 closure_sync(&writes);
1526
1527 if (ret) {
1528 blktrace_msg_all(c, "Stopped gc");
1529 pr_warn("gc failed!");
1530
1531 continue_at(cl, bch_btree_gc, bch_gc_wq);
1532 }
1533
1534 /* Possibly wait for new UUIDs or whatever to hit disk */
1535 bch_journal_meta(c, &op.cl);
1536 closure_sync(&op.cl);
1537
1538 available = bch_btree_gc_finish(c);
1539
1540 bch_time_stats_update(&c->btree_gc_time, start_time);
1541
1542 stats.key_bytes *= sizeof(uint64_t);
1543 stats.dirty <<= 9;
1544 stats.data <<= 9;
1545 stats.in_use = (c->nbuckets - available) * 100 / c->nbuckets;
1546 memcpy(&c->gc_stats, &stats, sizeof(struct gc_stat));
1547 blktrace_msg_all(c, "Finished gc");
1548
1549 trace_bcache_gc_end(c->sb.set_uuid);
1550 wake_up(&c->alloc_wait);
1551
1552 continue_at(cl, bch_moving_gc, bch_gc_wq);
1553}
1554
1555void bch_queue_gc(struct cache_set *c)
1556{
1557 closure_trylock_call(&c->gc.cl, bch_btree_gc, bch_gc_wq, &c->cl);
1558}
1559
1560/* Initial partial gc */
1561
1562static int bch_btree_check_recurse(struct btree *b, struct btree_op *op,
1563 unsigned long **seen)
1564{
1565 int ret;
1566 unsigned i;
1567 struct bkey *k;
1568 struct bucket *g;
1569 struct btree_iter iter;
1570
1571 for_each_key_filter(b, k, &iter, bch_ptr_invalid) {
1572 for (i = 0; i < KEY_PTRS(k); i++) {
1573 if (!ptr_available(b->c, k, i))
1574 continue;
1575
1576 g = PTR_BUCKET(b->c, k, i);
1577
1578 if (!__test_and_set_bit(PTR_BUCKET_NR(b->c, k, i),
1579 seen[PTR_DEV(k, i)]) ||
1580 !ptr_stale(b->c, k, i)) {
1581 g->gen = PTR_GEN(k, i);
1582
1583 if (b->level)
1584 g->prio = BTREE_PRIO;
1585 else if (g->prio == BTREE_PRIO)
1586 g->prio = INITIAL_PRIO;
1587 }
1588 }
1589
1590 btree_mark_key(b, k);
1591 }
1592
1593 if (b->level) {
1594 k = bch_next_recurse_key(b, &ZERO_KEY);
1595
1596 while (k) {
1597 struct bkey *p = bch_next_recurse_key(b, k);
1598 if (p)
1599 btree_node_prefetch(b->c, p, b->level - 1);
1600
1601 ret = btree(check_recurse, k, b, op, seen);
1602 if (ret)
1603 return ret;
1604
1605 k = p;
1606 }
1607 }
1608
1609 return 0;
1610}
1611
1612int bch_btree_check(struct cache_set *c, struct btree_op *op)
1613{
1614 int ret = -ENOMEM;
1615 unsigned i;
1616 unsigned long *seen[MAX_CACHES_PER_SET];
1617
1618 memset(seen, 0, sizeof(seen));
1619
1620 for (i = 0; c->cache[i]; i++) {
1621 size_t n = DIV_ROUND_UP(c->cache[i]->sb.nbuckets, 8);
1622 seen[i] = kmalloc(n, GFP_KERNEL);
1623 if (!seen[i])
1624 goto err;
1625
1626 /* Disables the seen array until prio_read() uses it too */
1627 memset(seen[i], 0xFF, n);
1628 }
1629
1630 ret = btree_root(check_recurse, c, op, seen);
1631err:
1632 for (i = 0; i < MAX_CACHES_PER_SET; i++)
1633 kfree(seen[i]);
1634 return ret;
1635}
1636
1637/* Btree insertion */
1638
1639static void shift_keys(struct btree *b, struct bkey *where, struct bkey *insert)
1640{
1641 struct bset *i = b->sets[b->nsets].data;
1642
1643 memmove((uint64_t *) where + bkey_u64s(insert),
1644 where,
1645 (void *) end(i) - (void *) where);
1646
1647 i->keys += bkey_u64s(insert);
1648 bkey_copy(where, insert);
1649 bch_bset_fix_lookup_table(b, where);
1650}
1651
1652static bool fix_overlapping_extents(struct btree *b,
1653 struct bkey *insert,
1654 struct btree_iter *iter,
1655 struct btree_op *op)
1656{
1657 void subtract_dirty(struct bkey *k, int sectors)
1658 {
1659 struct bcache_device *d = b->c->devices[KEY_INODE(k)];
1660
1661 if (KEY_DIRTY(k) && d)
1662 atomic_long_sub(sectors, &d->sectors_dirty);
1663 }
1664
1665 unsigned old_size, sectors_found = 0;
1666
1667 while (1) {
1668 struct bkey *k = bch_btree_iter_next(iter);
1669 if (!k ||
1670 bkey_cmp(&START_KEY(k), insert) >= 0)
1671 break;
1672
1673 if (bkey_cmp(k, &START_KEY(insert)) <= 0)
1674 continue;
1675
1676 old_size = KEY_SIZE(k);
1677
1678 /*
1679 * We might overlap with 0 size extents; we can't skip these
1680 * because if they're in the set we're inserting to we have to
1681 * adjust them so they don't overlap with the key we're
1682 * inserting. But we don't want to check them for BTREE_REPLACE
1683 * operations.
1684 */
1685
1686 if (op->type == BTREE_REPLACE &&
1687 KEY_SIZE(k)) {
1688 /*
1689 * k might have been split since we inserted/found the
1690 * key we're replacing
1691 */
1692 unsigned i;
1693 uint64_t offset = KEY_START(k) -
1694 KEY_START(&op->replace);
1695
1696 /* But it must be a subset of the replace key */
1697 if (KEY_START(k) < KEY_START(&op->replace) ||
1698 KEY_OFFSET(k) > KEY_OFFSET(&op->replace))
1699 goto check_failed;
1700
1701 /* We didn't find a key that we were supposed to */
1702 if (KEY_START(k) > KEY_START(insert) + sectors_found)
1703 goto check_failed;
1704
1705 if (KEY_PTRS(&op->replace) != KEY_PTRS(k))
1706 goto check_failed;
1707
1708 /* skip past gen */
1709 offset <<= 8;
1710
1711 BUG_ON(!KEY_PTRS(&op->replace));
1712
1713 for (i = 0; i < KEY_PTRS(&op->replace); i++)
1714 if (k->ptr[i] != op->replace.ptr[i] + offset)
1715 goto check_failed;
1716
1717 sectors_found = KEY_OFFSET(k) - KEY_START(insert);
1718 }
1719
1720 if (bkey_cmp(insert, k) < 0 &&
1721 bkey_cmp(&START_KEY(insert), &START_KEY(k)) > 0) {
1722 /*
1723 * We overlapped in the middle of an existing key: that
1724 * means we have to split the old key. But we have to do
1725 * slightly different things depending on whether the
1726 * old key has been written out yet.
1727 */
1728
1729 struct bkey *top;
1730
1731 subtract_dirty(k, KEY_SIZE(insert));
1732
1733 if (bkey_written(b, k)) {
1734 /*
1735 * We insert a new key to cover the top of the
1736 * old key, and the old key is modified in place
1737 * to represent the bottom split.
1738 *
1739 * It's completely arbitrary whether the new key
1740 * is the top or the bottom, but it has to match
1741 * up with what btree_sort_fixup() does - it
1742 * doesn't check for this kind of overlap, it
1743 * depends on us inserting a new key for the top
1744 * here.
1745 */
1746 top = bch_bset_search(b, &b->sets[b->nsets],
1747 insert);
1748 shift_keys(b, top, k);
1749 } else {
1750 BKEY_PADDED(key) temp;
1751 bkey_copy(&temp.key, k);
1752 shift_keys(b, k, &temp.key);
1753 top = bkey_next(k);
1754 }
1755
1756 bch_cut_front(insert, top);
1757 bch_cut_back(&START_KEY(insert), k);
1758 bch_bset_fix_invalidated_key(b, k);
1759 return false;
1760 }
1761
1762 if (bkey_cmp(insert, k) < 0) {
1763 bch_cut_front(insert, k);
1764 } else {
1765 if (bkey_written(b, k) &&
1766 bkey_cmp(&START_KEY(insert), &START_KEY(k)) <= 0) {
1767 /*
1768 * Completely overwrote, so we don't have to
1769 * invalidate the binary search tree
1770 */
1771 bch_cut_front(k, k);
1772 } else {
1773 __bch_cut_back(&START_KEY(insert), k);
1774 bch_bset_fix_invalidated_key(b, k);
1775 }
1776 }
1777
1778 subtract_dirty(k, old_size - KEY_SIZE(k));
1779 }
1780
1781check_failed:
1782 if (op->type == BTREE_REPLACE) {
1783 if (!sectors_found) {
1784 op->insert_collision = true;
1785 return true;
1786 } else if (sectors_found < KEY_SIZE(insert)) {
1787 SET_KEY_OFFSET(insert, KEY_OFFSET(insert) -
1788 (KEY_SIZE(insert) - sectors_found));
1789 SET_KEY_SIZE(insert, sectors_found);
1790 }
1791 }
1792
1793 return false;
1794}
1795
1796static bool btree_insert_key(struct btree *b, struct btree_op *op,
1797 struct bkey *k)
1798{
1799 struct bset *i = b->sets[b->nsets].data;
1800 struct bkey *m, *prev;
1801 const char *status = "insert";
1802
1803 BUG_ON(bkey_cmp(k, &b->key) > 0);
1804 BUG_ON(b->level && !KEY_PTRS(k));
1805 BUG_ON(!b->level && !KEY_OFFSET(k));
1806
1807 if (!b->level) {
1808 struct btree_iter iter;
1809 struct bkey search = KEY(KEY_INODE(k), KEY_START(k), 0);
1810
1811 /*
1812 * bset_search() returns the first key that is strictly greater
1813 * than the search key - but for back merging, we want to find
1814 * the first key that is greater than or equal to KEY_START(k) -
1815 * unless KEY_START(k) is 0.
1816 */
1817 if (KEY_OFFSET(&search))
1818 SET_KEY_OFFSET(&search, KEY_OFFSET(&search) - 1);
1819
1820 prev = NULL;
1821 m = bch_btree_iter_init(b, &iter, &search);
1822
1823 if (fix_overlapping_extents(b, k, &iter, op))
1824 return false;
1825
1826 while (m != end(i) &&
1827 bkey_cmp(k, &START_KEY(m)) > 0)
1828 prev = m, m = bkey_next(m);
1829
1830 if (key_merging_disabled(b->c))
1831 goto insert;
1832
1833 /* prev is in the tree, if we merge we're done */
1834 status = "back merging";
1835 if (prev &&
1836 bch_bkey_try_merge(b, prev, k))
1837 goto merged;
1838
1839 status = "overwrote front";
1840 if (m != end(i) &&
1841 KEY_PTRS(m) == KEY_PTRS(k) && !KEY_SIZE(m))
1842 goto copy;
1843
1844 status = "front merge";
1845 if (m != end(i) &&
1846 bch_bkey_try_merge(b, k, m))
1847 goto copy;
1848 } else
1849 m = bch_bset_search(b, &b->sets[b->nsets], k);
1850
1851insert: shift_keys(b, m, k);
1852copy: bkey_copy(m, k);
1853merged:
1854 bch_check_keys(b, "%s for %s at %s: %s", status,
1855 op_type(op), pbtree(b), pkey(k));
1856 bch_check_key_order_msg(b, i, "%s for %s at %s: %s", status,
1857 op_type(op), pbtree(b), pkey(k));
1858
1859 if (b->level && !KEY_OFFSET(k))
1860 b->prio_blocked++;
1861
1862 pr_debug("%s for %s at %s: %s", status,
1863 op_type(op), pbtree(b), pkey(k));
1864
1865 return true;
1866}
1867
1868bool bch_btree_insert_keys(struct btree *b, struct btree_op *op)
1869{
1870 bool ret = false;
1871 struct bkey *k;
1872 unsigned oldsize = bch_count_data(b);
1873
1874 while ((k = bch_keylist_pop(&op->keys))) {
1875 bkey_put(b->c, k, b->level);
1876 ret |= btree_insert_key(b, op, k);
1877 }
1878
1879 BUG_ON(bch_count_data(b) < oldsize);
1880 return ret;
1881}
1882
1883bool bch_btree_insert_check_key(struct btree *b, struct btree_op *op,
1884 struct bio *bio)
1885{
1886 bool ret = false;
1887 uint64_t btree_ptr = b->key.ptr[0];
1888 unsigned long seq = b->seq;
1889 BKEY_PADDED(k) tmp;
1890
1891 rw_unlock(false, b);
1892 rw_lock(true, b, b->level);
1893
1894 if (b->key.ptr[0] != btree_ptr ||
1895 b->seq != seq + 1 ||
1896 should_split(b))
1897 goto out;
1898
1899 op->replace = KEY(op->inode, bio_end(bio), bio_sectors(bio));
1900
1901 SET_KEY_PTRS(&op->replace, 1);
1902 get_random_bytes(&op->replace.ptr[0], sizeof(uint64_t));
1903
1904 SET_PTR_DEV(&op->replace, 0, PTR_CHECK_DEV);
1905
1906 bkey_copy(&tmp.k, &op->replace);
1907
1908 BUG_ON(op->type != BTREE_INSERT);
1909 BUG_ON(!btree_insert_key(b, op, &tmp.k));
1910 bch_btree_write(b, false, NULL);
1911 ret = true;
1912out:
1913 downgrade_write(&b->lock);
1914 return ret;
1915}
1916
1917static int btree_split(struct btree *b, struct btree_op *op)
1918{
1919 bool split, root = b == b->c->root;
1920 struct btree *n1, *n2 = NULL, *n3 = NULL;
1921 uint64_t start_time = local_clock();
1922
1923 if (b->level)
1924 set_closure_blocking(&op->cl);
1925
1926 n1 = btree_node_alloc_replacement(b, &op->cl);
1927 if (IS_ERR(n1))
1928 goto err;
1929
1930 split = set_blocks(n1->sets[0].data, n1->c) > (btree_blocks(b) * 4) / 5;
1931
1932 pr_debug("%ssplitting at %s keys %i", split ? "" : "not ",
1933 pbtree(b), n1->sets[0].data->keys);
1934
1935 if (split) {
1936 unsigned keys = 0;
1937
1938 n2 = bch_btree_node_alloc(b->c, b->level, &op->cl);
1939 if (IS_ERR(n2))
1940 goto err_free1;
1941
1942 if (root) {
1943 n3 = bch_btree_node_alloc(b->c, b->level + 1, &op->cl);
1944 if (IS_ERR(n3))
1945 goto err_free2;
1946 }
1947
1948 bch_btree_insert_keys(n1, op);
1949
1950 /* Has to be a linear search because we don't have an auxiliary
1951 * search tree yet
1952 */
1953
1954 while (keys < (n1->sets[0].data->keys * 3) / 5)
1955 keys += bkey_u64s(node(n1->sets[0].data, keys));
1956
1957 bkey_copy_key(&n1->key, node(n1->sets[0].data, keys));
1958 keys += bkey_u64s(node(n1->sets[0].data, keys));
1959
1960 n2->sets[0].data->keys = n1->sets[0].data->keys - keys;
1961 n1->sets[0].data->keys = keys;
1962
1963 memcpy(n2->sets[0].data->start,
1964 end(n1->sets[0].data),
1965 n2->sets[0].data->keys * sizeof(uint64_t));
1966
1967 bkey_copy_key(&n2->key, &b->key);
1968
1969 bch_keylist_add(&op->keys, &n2->key);
1970 bch_btree_write(n2, true, op);
1971 rw_unlock(true, n2);
1972 } else
1973 bch_btree_insert_keys(n1, op);
1974
1975 bch_keylist_add(&op->keys, &n1->key);
1976 bch_btree_write(n1, true, op);
1977
1978 if (n3) {
1979 bkey_copy_key(&n3->key, &MAX_KEY);
1980 bch_btree_insert_keys(n3, op);
1981 bch_btree_write(n3, true, op);
1982
1983 closure_sync(&op->cl);
1984 bch_btree_set_root(n3);
1985 rw_unlock(true, n3);
1986 } else if (root) {
1987 op->keys.top = op->keys.bottom;
1988 closure_sync(&op->cl);
1989 bch_btree_set_root(n1);
1990 } else {
1991 unsigned i;
1992
1993 bkey_copy(op->keys.top, &b->key);
1994 bkey_copy_key(op->keys.top, &ZERO_KEY);
1995
1996 for (i = 0; i < KEY_PTRS(&b->key); i++) {
1997 uint8_t g = PTR_BUCKET(b->c, &b->key, i)->gen + 1;
1998
1999 SET_PTR_GEN(op->keys.top, i, g);
2000 }
2001
2002 bch_keylist_push(&op->keys);
2003 closure_sync(&op->cl);
2004 atomic_inc(&b->c->prio_blocked);
2005 }
2006
2007 rw_unlock(true, n1);
2008 btree_node_free(b, op);
2009
2010 bch_time_stats_update(&b->c->btree_split_time, start_time);
2011
2012 return 0;
2013err_free2:
2014 __bkey_put(n2->c, &n2->key);
2015 btree_node_free(n2, op);
2016 rw_unlock(true, n2);
2017err_free1:
2018 __bkey_put(n1->c, &n1->key);
2019 btree_node_free(n1, op);
2020 rw_unlock(true, n1);
2021err:
2022 if (n3 == ERR_PTR(-EAGAIN) ||
2023 n2 == ERR_PTR(-EAGAIN) ||
2024 n1 == ERR_PTR(-EAGAIN))
2025 return -EAGAIN;
2026
2027 pr_warn("couldn't split");
2028 return -ENOMEM;
2029}
2030
2031static int bch_btree_insert_recurse(struct btree *b, struct btree_op *op,
2032 struct keylist *stack_keys)
2033{
2034 if (b->level) {
2035 int ret;
2036 struct bkey *insert = op->keys.bottom;
2037 struct bkey *k = bch_next_recurse_key(b, &START_KEY(insert));
2038
2039 if (!k) {
2040 btree_bug(b, "no key to recurse on at level %i/%i",
2041 b->level, b->c->root->level);
2042
2043 op->keys.top = op->keys.bottom;
2044 return -EIO;
2045 }
2046
2047 if (bkey_cmp(insert, k) > 0) {
2048 unsigned i;
2049
2050 if (op->type == BTREE_REPLACE) {
2051 __bkey_put(b->c, insert);
2052 op->keys.top = op->keys.bottom;
2053 op->insert_collision = true;
2054 return 0;
2055 }
2056
2057 for (i = 0; i < KEY_PTRS(insert); i++)
2058 atomic_inc(&PTR_BUCKET(b->c, insert, i)->pin);
2059
2060 bkey_copy(stack_keys->top, insert);
2061
2062 bch_cut_back(k, insert);
2063 bch_cut_front(k, stack_keys->top);
2064
2065 bch_keylist_push(stack_keys);
2066 }
2067
2068 ret = btree(insert_recurse, k, b, op, stack_keys);
2069 if (ret)
2070 return ret;
2071 }
2072
2073 if (!bch_keylist_empty(&op->keys)) {
2074 if (should_split(b)) {
2075 if (op->lock <= b->c->root->level) {
2076 BUG_ON(b->level);
2077 op->lock = b->c->root->level + 1;
2078 return -EINTR;
2079 }
2080 return btree_split(b, op);
2081 }
2082
2083 BUG_ON(write_block(b) != b->sets[b->nsets].data);
2084
2085 if (bch_btree_insert_keys(b, op))
2086 bch_btree_write(b, false, op);
2087 }
2088
2089 return 0;
2090}
2091
2092int bch_btree_insert(struct btree_op *op, struct cache_set *c)
2093{
2094 int ret = 0;
2095 struct keylist stack_keys;
2096
2097 /*
2098 * Don't want to block with the btree locked unless we have to,
2099 * otherwise we get deadlocks with try_harder and between split/gc
2100 */
2101 clear_closure_blocking(&op->cl);
2102
2103 BUG_ON(bch_keylist_empty(&op->keys));
2104 bch_keylist_copy(&stack_keys, &op->keys);
2105 bch_keylist_init(&op->keys);
2106
2107 while (!bch_keylist_empty(&stack_keys) ||
2108 !bch_keylist_empty(&op->keys)) {
2109 if (bch_keylist_empty(&op->keys)) {
2110 bch_keylist_add(&op->keys,
2111 bch_keylist_pop(&stack_keys));
2112 op->lock = 0;
2113 }
2114
2115 ret = btree_root(insert_recurse, c, op, &stack_keys);
2116
2117 if (ret == -EAGAIN) {
2118 ret = 0;
2119 closure_sync(&op->cl);
2120 } else if (ret) {
2121 struct bkey *k;
2122
2123 pr_err("error %i trying to insert key for %s",
2124 ret, op_type(op));
2125
2126 while ((k = bch_keylist_pop(&stack_keys) ?:
2127 bch_keylist_pop(&op->keys)))
2128 bkey_put(c, k, 0);
2129 }
2130 }
2131
2132 bch_keylist_free(&stack_keys);
2133
2134 if (op->journal)
2135 atomic_dec_bug(op->journal);
2136 op->journal = NULL;
2137 return ret;
2138}
2139
2140void bch_btree_set_root(struct btree *b)
2141{
2142 unsigned i;
2143
2144 BUG_ON(!b->written);
2145
2146 for (i = 0; i < KEY_PTRS(&b->key); i++)
2147 BUG_ON(PTR_BUCKET(b->c, &b->key, i)->prio != BTREE_PRIO);
2148
2149 mutex_lock(&b->c->bucket_lock);
2150 list_del_init(&b->list);
2151 mutex_unlock(&b->c->bucket_lock);
2152
2153 b->c->root = b;
2154 __bkey_put(b->c, &b->key);
2155
2156 bch_journal_meta(b->c, NULL);
2157 pr_debug("%s for %pf", pbtree(b), __builtin_return_address(0));
2158}
2159
2160/* Cache lookup */
2161
2162static int submit_partial_cache_miss(struct btree *b, struct btree_op *op,
2163 struct bkey *k)
2164{
2165 struct search *s = container_of(op, struct search, op);
2166 struct bio *bio = &s->bio.bio;
2167 int ret = 0;
2168
2169 while (!ret &&
2170 !op->lookup_done) {
2171 unsigned sectors = INT_MAX;
2172
2173 if (KEY_INODE(k) == op->inode) {
2174 if (KEY_START(k) <= bio->bi_sector)
2175 break;
2176
2177 sectors = min_t(uint64_t, sectors,
2178 KEY_START(k) - bio->bi_sector);
2179 }
2180
2181 ret = s->d->cache_miss(b, s, bio, sectors);
2182 }
2183
2184 return ret;
2185}
2186
2187/*
2188 * Read from a single key, handling the initial cache miss if the key starts in
2189 * the middle of the bio
2190 */
2191static int submit_partial_cache_hit(struct btree *b, struct btree_op *op,
2192 struct bkey *k)
2193{
2194 struct search *s = container_of(op, struct search, op);
2195 struct bio *bio = &s->bio.bio;
2196 unsigned ptr;
2197 struct bio *n;
2198
2199 int ret = submit_partial_cache_miss(b, op, k);
2200 if (ret || op->lookup_done)
2201 return ret;
2202
2203 /* XXX: figure out best pointer - for multiple cache devices */
2204 ptr = 0;
2205
2206 PTR_BUCKET(b->c, k, ptr)->prio = INITIAL_PRIO;
2207
2208 while (!op->lookup_done &&
2209 KEY_INODE(k) == op->inode &&
2210 bio->bi_sector < KEY_OFFSET(k)) {
2211 struct bkey *bio_key;
2212 sector_t sector = PTR_OFFSET(k, ptr) +
2213 (bio->bi_sector - KEY_START(k));
2214 unsigned sectors = min_t(uint64_t, INT_MAX,
2215 KEY_OFFSET(k) - bio->bi_sector);
2216
2217 n = bch_bio_split(bio, sectors, GFP_NOIO, s->d->bio_split);
2218 if (!n)
2219 return -EAGAIN;
2220
2221 if (n == bio)
2222 op->lookup_done = true;
2223
2224 bio_key = &container_of(n, struct bbio, bio)->key;
2225
2226 /*
2227 * The bucket we're reading from might be reused while our bio
2228 * is in flight, and we could then end up reading the wrong
2229 * data.
2230 *
2231 * We guard against this by checking (in cache_read_endio()) if
2232 * the pointer is stale again; if so, we treat it as an error
2233 * and reread from the backing device (but we don't pass that
2234 * error up anywhere).
2235 */
2236
2237 bch_bkey_copy_single_ptr(bio_key, k, ptr);
2238 SET_PTR_OFFSET(bio_key, 0, sector);
2239
2240 n->bi_end_io = bch_cache_read_endio;
2241 n->bi_private = &s->cl;
2242
2243 trace_bcache_cache_hit(n);
2244 __bch_submit_bbio(n, b->c);
2245 }
2246
2247 return 0;
2248}
2249
2250int bch_btree_search_recurse(struct btree *b, struct btree_op *op)
2251{
2252 struct search *s = container_of(op, struct search, op);
2253 struct bio *bio = &s->bio.bio;
2254
2255 int ret = 0;
2256 struct bkey *k;
2257 struct btree_iter iter;
2258 bch_btree_iter_init(b, &iter, &KEY(op->inode, bio->bi_sector, 0));
2259
2260 pr_debug("at %s searching for %u:%llu", pbtree(b), op->inode,
2261 (uint64_t) bio->bi_sector);
2262
2263 do {
2264 k = bch_btree_iter_next_filter(&iter, b, bch_ptr_bad);
2265 if (!k) {
2266 /*
2267 * b->key would be exactly what we want, except that
2268 * pointers to btree nodes have nonzero size - we
2269 * wouldn't go far enough
2270 */
2271
2272 ret = submit_partial_cache_miss(b, op,
2273 &KEY(KEY_INODE(&b->key),
2274 KEY_OFFSET(&b->key), 0));
2275 break;
2276 }
2277
2278 ret = b->level
2279 ? btree(search_recurse, k, b, op)
2280 : submit_partial_cache_hit(b, op, k);
2281 } while (!ret &&
2282 !op->lookup_done);
2283
2284 return ret;
2285}
2286
2287/* Keybuf code */
2288
2289static inline int keybuf_cmp(struct keybuf_key *l, struct keybuf_key *r)
2290{
2291 /* Overlapping keys compare equal */
2292 if (bkey_cmp(&l->key, &START_KEY(&r->key)) <= 0)
2293 return -1;
2294 if (bkey_cmp(&START_KEY(&l->key), &r->key) >= 0)
2295 return 1;
2296 return 0;
2297}
2298
2299static inline int keybuf_nonoverlapping_cmp(struct keybuf_key *l,
2300 struct keybuf_key *r)
2301{
2302 return clamp_t(int64_t, bkey_cmp(&l->key, &r->key), -1, 1);
2303}
2304
2305static int bch_btree_refill_keybuf(struct btree *b, struct btree_op *op,
2306 struct keybuf *buf, struct bkey *end)
2307{
2308 struct btree_iter iter;
2309 bch_btree_iter_init(b, &iter, &buf->last_scanned);
2310
2311 while (!array_freelist_empty(&buf->freelist)) {
2312 struct bkey *k = bch_btree_iter_next_filter(&iter, b,
2313 bch_ptr_bad);
2314
2315 if (!b->level) {
2316 if (!k) {
2317 buf->last_scanned = b->key;
2318 break;
2319 }
2320
2321 buf->last_scanned = *k;
2322 if (bkey_cmp(&buf->last_scanned, end) >= 0)
2323 break;
2324
2325 if (buf->key_predicate(buf, k)) {
2326 struct keybuf_key *w;
2327
2328 pr_debug("%s", pkey(k));
2329
2330 spin_lock(&buf->lock);
2331
2332 w = array_alloc(&buf->freelist);
2333
2334 w->private = NULL;
2335 bkey_copy(&w->key, k);
2336
2337 if (RB_INSERT(&buf->keys, w, node, keybuf_cmp))
2338 array_free(&buf->freelist, w);
2339
2340 spin_unlock(&buf->lock);
2341 }
2342 } else {
2343 if (!k)
2344 break;
2345
2346 btree(refill_keybuf, k, b, op, buf, end);
2347 /*
2348 * Might get an error here, but can't really do anything
2349 * and it'll get logged elsewhere. Just read what we
2350 * can.
2351 */
2352
2353 if (bkey_cmp(&buf->last_scanned, end) >= 0)
2354 break;
2355
2356 cond_resched();
2357 }
2358 }
2359
2360 return 0;
2361}
2362
2363void bch_refill_keybuf(struct cache_set *c, struct keybuf *buf,
2364 struct bkey *end)
2365{
2366 struct bkey start = buf->last_scanned;
2367 struct btree_op op;
2368 bch_btree_op_init_stack(&op);
2369
2370 cond_resched();
2371
2372 btree_root(refill_keybuf, c, &op, buf, end);
2373 closure_sync(&op.cl);
2374
2375 pr_debug("found %s keys from %llu:%llu to %llu:%llu",
2376 RB_EMPTY_ROOT(&buf->keys) ? "no" :
2377 array_freelist_empty(&buf->freelist) ? "some" : "a few",
2378 KEY_INODE(&start), KEY_OFFSET(&start),
2379 KEY_INODE(&buf->last_scanned), KEY_OFFSET(&buf->last_scanned));
2380
2381 spin_lock(&buf->lock);
2382
2383 if (!RB_EMPTY_ROOT(&buf->keys)) {
2384 struct keybuf_key *w;
2385 w = RB_FIRST(&buf->keys, struct keybuf_key, node);
2386 buf->start = START_KEY(&w->key);
2387
2388 w = RB_LAST(&buf->keys, struct keybuf_key, node);
2389 buf->end = w->key;
2390 } else {
2391 buf->start = MAX_KEY;
2392 buf->end = MAX_KEY;
2393 }
2394
2395 spin_unlock(&buf->lock);
2396}
2397
2398static void __bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
2399{
2400 rb_erase(&w->node, &buf->keys);
2401 array_free(&buf->freelist, w);
2402}
2403
2404void bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
2405{
2406 spin_lock(&buf->lock);
2407 __bch_keybuf_del(buf, w);
2408 spin_unlock(&buf->lock);
2409}
2410
2411bool bch_keybuf_check_overlapping(struct keybuf *buf, struct bkey *start,
2412 struct bkey *end)
2413{
2414 bool ret = false;
2415 struct keybuf_key *p, *w, s;
2416 s.key = *start;
2417
2418 if (bkey_cmp(end, &buf->start) <= 0 ||
2419 bkey_cmp(start, &buf->end) >= 0)
2420 return false;
2421
2422 spin_lock(&buf->lock);
2423 w = RB_GREATER(&buf->keys, s, node, keybuf_nonoverlapping_cmp);
2424
2425 while (w && bkey_cmp(&START_KEY(&w->key), end) < 0) {
2426 p = w;
2427 w = RB_NEXT(w, node);
2428
2429 if (p->private)
2430 ret = true;
2431 else
2432 __bch_keybuf_del(buf, p);
2433 }
2434
2435 spin_unlock(&buf->lock);
2436 return ret;
2437}
2438
2439struct keybuf_key *bch_keybuf_next(struct keybuf *buf)
2440{
2441 struct keybuf_key *w;
2442 spin_lock(&buf->lock);
2443
2444 w = RB_FIRST(&buf->keys, struct keybuf_key, node);
2445
2446 while (w && w->private)
2447 w = RB_NEXT(w, node);
2448
2449 if (w)
2450 w->private = ERR_PTR(-EINTR);
2451
2452 spin_unlock(&buf->lock);
2453 return w;
2454}
2455
2456struct keybuf_key *bch_keybuf_next_rescan(struct cache_set *c,
2457 struct keybuf *buf,
2458 struct bkey *end)
2459{
2460 struct keybuf_key *ret;
2461
2462 while (1) {
2463 ret = bch_keybuf_next(buf);
2464 if (ret)
2465 break;
2466
2467 if (bkey_cmp(&buf->last_scanned, end) >= 0) {
2468 pr_debug("scan finished");
2469 break;
2470 }
2471
2472 bch_refill_keybuf(c, buf, end);
2473 }
2474
2475 return ret;
2476}
2477
2478void bch_keybuf_init(struct keybuf *buf, keybuf_pred_fn *fn)
2479{
2480 buf->key_predicate = fn;
2481 buf->last_scanned = MAX_KEY;
2482 buf->keys = RB_ROOT;
2483
2484 spin_lock_init(&buf->lock);
2485 array_allocator_init(&buf->freelist);
2486}
2487
2488void bch_btree_exit(void)
2489{
2490 if (btree_io_wq)
2491 destroy_workqueue(btree_io_wq);
2492 if (bch_gc_wq)
2493 destroy_workqueue(bch_gc_wq);
2494}
2495
2496int __init bch_btree_init(void)
2497{
2498 if (!(bch_gc_wq = create_singlethread_workqueue("bch_btree_gc")) ||
2499 !(btree_io_wq = create_singlethread_workqueue("bch_btree_io")))
2500 return -ENOMEM;
2501
2502 return 0;
2503}
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-18 04:55:04 -0500