projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
bcache: A block layer cache
[deliverable/linux.git] / drivers / md / bcache / journal.c
1 /*
2 * bcache journalling code, for btree insertions
3 *
4 * Copyright 2012 Google, Inc.
5 */
6
7 #include "bcache.h"
8 #include "btree.h"
9 #include "debug.h"
10 #include "request.h"
11
12 /*
13 * Journal replay/recovery:
14 *
15 * This code is all driven from run_cache_set(); we first read the journal
16 * entries, do some other stuff, then we mark all the keys in the journal
17 * entries (same as garbage collection would), then we replay them - reinserting
18 * them into the cache in precisely the same order as they appear in the
19 * journal.
20 *
21 * We only journal keys that go in leaf nodes, which simplifies things quite a
22 * bit.
23 */
24
25 static void journal_read_endio(struct bio *bio, int error)
26 {
27 struct closure *cl = bio->bi_private;
28 closure_put(cl);
29 }
30
31 static int journal_read_bucket(struct cache *ca, struct list_head *list,
32 struct btree_op *op, unsigned bucket_index)
33 {
34 struct journal_device *ja = &ca->journal;
35 struct bio *bio = &ja->bio;
36
37 struct journal_replay *i;
38 struct jset *j, *data = ca->set->journal.w[0].data;
39 unsigned len, left, offset = 0;
40 int ret = 0;
41 sector_t bucket = bucket_to_sector(ca->set, ca->sb.d[bucket_index]);
42
43 pr_debug("reading %llu", (uint64_t) bucket);
44
45 while (offset < ca->sb.bucket_size) {
46 reread: left = ca->sb.bucket_size - offset;
47 len = min_t(unsigned, left, PAGE_SECTORS * 8);
48
49 bio_reset(bio);
50 bio->bi_sector = bucket + offset;
51 bio->bi_bdev = ca->bdev;
52 bio->bi_rw = READ;
53 bio->bi_size = len << 9;
54
55 bio->bi_end_io = journal_read_endio;
56 bio->bi_private = &op->cl;
57 bio_map(bio, data);
58
59 closure_bio_submit(bio, &op->cl, ca);
60 closure_sync(&op->cl);
61
62 /* This function could be simpler now since we no longer write
63 * journal entries that overlap bucket boundaries; this means
64 * the start of a bucket will always have a valid journal entry
65 * if it has any journal entries at all.
66 */
67
68 j = data;
69 while (len) {
70 struct list_head *where;
71 size_t blocks, bytes = set_bytes(j);
72
73 if (j->magic != jset_magic(ca->set))
74 return ret;
75
76 if (bytes > left << 9)
77 return ret;
78
79 if (bytes > len << 9)
80 goto reread;
81
82 if (j->csum != csum_set(j))
83 return ret;
84
85 blocks = set_blocks(j, ca->set);
86
87 while (!list_empty(list)) {
88 i = list_first_entry(list,
89 struct journal_replay, list);
90 if (i->j.seq >= j->last_seq)
91 break;
92 list_del(&i->list);
93 kfree(i);
94 }
95
96 list_for_each_entry_reverse(i, list, list) {
97 if (j->seq == i->j.seq)
98 goto next_set;
99
100 if (j->seq < i->j.last_seq)
101 goto next_set;
102
103 if (j->seq > i->j.seq) {
104 where = &i->list;
105 goto add;
106 }
107 }
108
109 where = list;
110 add:
111 i = kmalloc(offsetof(struct journal_replay, j) +
112 bytes, GFP_KERNEL);
113 if (!i)
114 return -ENOMEM;
115 memcpy(&i->j, j, bytes);
116 list_add(&i->list, where);
117 ret = 1;
118
119 ja->seq[bucket_index] = j->seq;
120 next_set:
121 offset += blocks * ca->sb.block_size;
122 len -= blocks * ca->sb.block_size;
123 j = ((void *) j) + blocks * block_bytes(ca);
124 }
125 }
126
127 return ret;
128 }
129
130 int bch_journal_read(struct cache_set *c, struct list_head *list,
131 struct btree_op *op)
132 {
133 #define read_bucket(b) \
134 ({ \
135 int ret = journal_read_bucket(ca, list, op, b); \
136 __set_bit(b, bitmap); \
137 if (ret < 0) \
138 return ret; \
139 ret; \
140 })
141
142 struct cache *ca;
143 unsigned iter;
144
145 for_each_cache(ca, c, iter) {
146 struct journal_device *ja = &ca->journal;
147 unsigned long bitmap[SB_JOURNAL_BUCKETS / BITS_PER_LONG];
148 unsigned i, l, r, m;
149 uint64_t seq;
150
151 bitmap_zero(bitmap, SB_JOURNAL_BUCKETS);
152 pr_debug("%u journal buckets", ca->sb.njournal_buckets);
153
154 /* Read journal buckets ordered by golden ratio hash to quickly
155 * find a sequence of buckets with valid journal entries
156 */
157 for (i = 0; i < ca->sb.njournal_buckets; i++) {
158 l = (i * 2654435769U) % ca->sb.njournal_buckets;
159
160 if (test_bit(l, bitmap))
161 break;
162
163 if (read_bucket(l))
164 goto bsearch;
165 }
166
167 /* If that fails, check all the buckets we haven't checked
168 * already
169 */
170 pr_debug("falling back to linear search");
171
172 for (l = 0; l < ca->sb.njournal_buckets; l++) {
173 if (test_bit(l, bitmap))
174 continue;
175
176 if (read_bucket(l))
177 goto bsearch;
178 }
179 bsearch:
180 /* Binary search */
181 m = r = find_next_bit(bitmap, ca->sb.njournal_buckets, l + 1);
182 pr_debug("starting binary search, l %u r %u", l, r);
183
184 while (l + 1 < r) {
185 m = (l + r) >> 1;
186
187 if (read_bucket(m))
188 l = m;
189 else
190 r = m;
191 }
192
193 /* Read buckets in reverse order until we stop finding more
194 * journal entries
195 */
196 pr_debug("finishing up");
197 l = m;
198
199 while (1) {
200 if (!l--)
201 l = ca->sb.njournal_buckets - 1;
202
203 if (l == m)
204 break;
205
206 if (test_bit(l, bitmap))
207 continue;
208
209 if (!read_bucket(l))
210 break;
211 }
212
213 seq = 0;
214
215 for (i = 0; i < ca->sb.njournal_buckets; i++)
216 if (ja->seq[i] > seq) {
217 seq = ja->seq[i];
218 ja->cur_idx = ja->discard_idx =
219 ja->last_idx = i;
220
221 }
222 }
223
224 c->journal.seq = list_entry(list->prev,
225 struct journal_replay,
226 list)->j.seq;
227
228 return 0;
229 #undef read_bucket
230 }
231
232 void bch_journal_mark(struct cache_set *c, struct list_head *list)
233 {
234 atomic_t p = { 0 };
235 struct bkey *k;
236 struct journal_replay *i;
237 struct journal *j = &c->journal;
238 uint64_t last = j->seq;
239
240 /*
241 * journal.pin should never fill up - we never write a journal
242 * entry when it would fill up. But if for some reason it does, we
243 * iterate over the list in reverse order so that we can just skip that
244 * refcount instead of bugging.
245 */
246
247 list_for_each_entry_reverse(i, list, list) {
248 BUG_ON(last < i->j.seq);
249 i->pin = NULL;
250
251 while (last-- != i->j.seq)
252 if (fifo_free(&j->pin) > 1) {
253 fifo_push_front(&j->pin, p);
254 atomic_set(&fifo_front(&j->pin), 0);
255 }
256
257 if (fifo_free(&j->pin) > 1) {
258 fifo_push_front(&j->pin, p);
259 i->pin = &fifo_front(&j->pin);
260 atomic_set(i->pin, 1);
261 }
262
263 for (k = i->j.start;
264 k < end(&i->j);
265 k = bkey_next(k)) {
266 unsigned j;
267
268 for (j = 0; j < KEY_PTRS(k); j++) {
269 struct bucket *g = PTR_BUCKET(c, k, j);
270 atomic_inc(&g->pin);
271
272 if (g->prio == BTREE_PRIO &&
273 !ptr_stale(c, k, j))
274 g->prio = INITIAL_PRIO;
275 }
276
277 __bch_btree_mark_key(c, 0, k);
278 }
279 }
280 }
281
282 int bch_journal_replay(struct cache_set *s, struct list_head *list,
283 struct btree_op *op)
284 {
285 int ret = 0, keys = 0, entries = 0;
286 struct bkey *k;
287 struct journal_replay *i =
288 list_entry(list->prev, struct journal_replay, list);
289
290 uint64_t start = i->j.last_seq, end = i->j.seq, n = start;
291
292 list_for_each_entry(i, list, list) {
293 BUG_ON(i->pin && atomic_read(i->pin) != 1);
294
295 if (n != i->j.seq)
296 pr_err("journal entries %llu-%llu "
297 "missing! (replaying %llu-%llu)\n",
298 n, i->j.seq - 1, start, end);
299
300 for (k = i->j.start;
301 k < end(&i->j);
302 k = bkey_next(k)) {
303 pr_debug("%s", pkey(k));
304 bkey_copy(op->keys.top, k);
305 bch_keylist_push(&op->keys);
306
307 op->journal = i->pin;
308 atomic_inc(op->journal);
309
310 ret = bch_btree_insert(op, s);
311 if (ret)
312 goto err;
313
314 BUG_ON(!bch_keylist_empty(&op->keys));
315 keys++;
316
317 cond_resched();
318 }
319
320 if (i->pin)
321 atomic_dec(i->pin);
322 n = i->j.seq + 1;
323 entries++;
324 }
325
326 pr_info("journal replay done, %i keys in %i entries, seq %llu",
327 keys, entries, end);
328
329 while (!list_empty(list)) {
330 i = list_first_entry(list, struct journal_replay, list);
331 list_del(&i->list);
332 kfree(i);
333 }
334 err:
335 closure_sync(&op->cl);
336 return ret;
337 }
338
339 /* Journalling */
340
341 static void btree_flush_write(struct cache_set *c)
342 {
343 /*
344 * Try to find the btree node with that references the oldest journal
345 * entry, best is our current candidate and is locked if non NULL:
346 */
347 struct btree *b, *best = NULL;
348 unsigned iter;
349
350 for_each_cached_btree(b, c, iter) {
351 if (!down_write_trylock(&b->lock))
352 continue;
353
354 if (!btree_node_dirty(b) ||
355 !btree_current_write(b)->journal) {
356 rw_unlock(true, b);
357 continue;
358 }
359
360 if (!best)
361 best = b;
362 else if (journal_pin_cmp(c,
363 btree_current_write(best),
364 btree_current_write(b))) {
365 rw_unlock(true, best);
366 best = b;
367 } else
368 rw_unlock(true, b);
369 }
370
371 if (best)
372 goto out;
373
374 /* We can't find the best btree node, just pick the first */
375 list_for_each_entry(b, &c->btree_cache, list)
376 if (!b->level && btree_node_dirty(b)) {
377 best = b;
378 rw_lock(true, best, best->level);
379 goto found;
380 }
381
382 out:
383 if (!best)
384 return;
385 found:
386 if (btree_node_dirty(best))
387 bch_btree_write(best, true, NULL);
388 rw_unlock(true, best);
389 }
390
391 #define last_seq(j) ((j)->seq - fifo_used(&(j)->pin) + 1)
392
393 static void journal_discard_endio(struct bio *bio, int error)
394 {
395 struct journal_device *ja =
396 container_of(bio, struct journal_device, discard_bio);
397 struct cache *ca = container_of(ja, struct cache, journal);
398
399 atomic_set(&ja->discard_in_flight, DISCARD_DONE);
400
401 closure_wake_up(&ca->set->journal.wait);
402 closure_put(&ca->set->cl);
403 }
404
405 static void journal_discard_work(struct work_struct *work)
406 {
407 struct journal_device *ja =
408 container_of(work, struct journal_device, discard_work);
409
410 submit_bio(0, &ja->discard_bio);
411 }
412
413 static void do_journal_discard(struct cache *ca)
414 {
415 struct journal_device *ja = &ca->journal;
416 struct bio *bio = &ja->discard_bio;
417
418 if (!ca->discard) {
419 ja->discard_idx = ja->last_idx;
420 return;
421 }
422
423 switch (atomic_read(&ja->discard_in_flight) == DISCARD_IN_FLIGHT) {
424 case DISCARD_IN_FLIGHT:
425 return;
426
427 case DISCARD_DONE:
428 ja->discard_idx = (ja->discard_idx + 1) %
429 ca->sb.njournal_buckets;
430
431 atomic_set(&ja->discard_in_flight, DISCARD_READY);
432 /* fallthrough */
433
434 case DISCARD_READY:
435 if (ja->discard_idx == ja->last_idx)
436 return;
437
438 atomic_set(&ja->discard_in_flight, DISCARD_IN_FLIGHT);
439
440 bio_init(bio);
441 bio->bi_sector = bucket_to_sector(ca->set,
442 ca->sb.d[ja->discard_idx]);
443 bio->bi_bdev = ca->bdev;
444 bio->bi_rw = REQ_WRITE|REQ_DISCARD;
445 bio->bi_max_vecs = 1;
446 bio->bi_io_vec = bio->bi_inline_vecs;
447 bio->bi_size = bucket_bytes(ca);
448 bio->bi_end_io = journal_discard_endio;
449
450 closure_get(&ca->set->cl);
451 INIT_WORK(&ja->discard_work, journal_discard_work);
452 schedule_work(&ja->discard_work);
453 }
454 }
455
456 static void journal_reclaim(struct cache_set *c)
457 {
458 struct bkey *k = &c->journal.key;
459 struct cache *ca;
460 uint64_t last_seq;
461 unsigned iter, n = 0;
462 atomic_t p;
463
464 while (!atomic_read(&fifo_front(&c->journal.pin)))
465 fifo_pop(&c->journal.pin, p);
466
467 last_seq = last_seq(&c->journal);
468
469 /* Update last_idx */
470
471 for_each_cache(ca, c, iter) {
472 struct journal_device *ja = &ca->journal;
473
474 while (ja->last_idx != ja->cur_idx &&
475 ja->seq[ja->last_idx] < last_seq)
476 ja->last_idx = (ja->last_idx + 1) %
477 ca->sb.njournal_buckets;
478 }
479
480 for_each_cache(ca, c, iter)
481 do_journal_discard(ca);
482
483 if (c->journal.blocks_free)
484 return;
485
486 /*
487 * Allocate:
488 * XXX: Sort by free journal space
489 */
490
491 for_each_cache(ca, c, iter) {
492 struct journal_device *ja = &ca->journal;
493 unsigned next = (ja->cur_idx + 1) % ca->sb.njournal_buckets;
494
495 /* No space available on this device */
496 if (next == ja->discard_idx)
497 continue;
498
499 ja->cur_idx = next;
500 k->ptr[n++] = PTR(0,
501 bucket_to_sector(c, ca->sb.d[ja->cur_idx]),
502 ca->sb.nr_this_dev);
503 }
504
505 bkey_init(k);
506 SET_KEY_PTRS(k, n);
507
508 if (n)
509 c->journal.blocks_free = c->sb.bucket_size >> c->block_bits;
510
511 if (!journal_full(&c->journal))
512 __closure_wake_up(&c->journal.wait);
513 }
514
515 void bch_journal_next(struct journal *j)
516 {
517 atomic_t p = { 1 };
518
519 j->cur = (j->cur == j->w)
520 ? &j->w[1]
521 : &j->w[0];
522
523 /*
524 * The fifo_push() needs to happen at the same time as j->seq is
525 * incremented for last_seq() to be calculated correctly
526 */
527 BUG_ON(!fifo_push(&j->pin, p));
528 atomic_set(&fifo_back(&j->pin), 1);
529
530 j->cur->data->seq = ++j->seq;
531 j->cur->need_write = false;
532 j->cur->data->keys = 0;
533
534 if (fifo_full(&j->pin))
535 pr_debug("journal_pin full (%zu)", fifo_used(&j->pin));
536 }
537
538 static void journal_write_endio(struct bio *bio, int error)
539 {
540 struct journal_write *w = bio->bi_private;
541
542 cache_set_err_on(error, w->c, "journal io error");
543 closure_put(&w->c->journal.io.cl);
544 }
545
546 static void journal_write(struct closure *);
547
548 static void journal_write_done(struct closure *cl)
549 {
550 struct journal *j = container_of(cl, struct journal, io.cl);
551 struct cache_set *c = container_of(j, struct cache_set, journal);
552
553 struct journal_write *w = (j->cur == j->w)
554 ? &j->w[1]
555 : &j->w[0];
556
557 __closure_wake_up(&w->wait);
558
559 if (c->journal_delay_ms)
560 closure_delay(&j->io, msecs_to_jiffies(c->journal_delay_ms));
561
562 continue_at(cl, journal_write, system_wq);
563 }
564
565 static void journal_write_unlocked(struct closure *cl)
566 {
567 struct cache_set *c = container_of(cl, struct cache_set, journal.io.cl);
568 struct cache *ca;
569 struct journal_write *w = c->journal.cur;
570 struct bkey *k = &c->journal.key;
571 unsigned i, sectors = set_blocks(w->data, c) * c->sb.block_size;
572
573 struct bio *bio;
574 struct bio_list list;
575 bio_list_init(&list);
576
577 if (!w->need_write) {
578 /*
579 * XXX: have to unlock closure before we unlock journal lock,
580 * else we race with bch_journal(). But this way we race
581 * against cache set unregister. Doh.
582 */
583 set_closure_fn(cl, NULL, NULL);
584 closure_sub(cl, CLOSURE_RUNNING + 1);
585 spin_unlock(&c->journal.lock);
586 return;
587 } else if (journal_full(&c->journal)) {
588 journal_reclaim(c);
589 spin_unlock(&c->journal.lock);
590
591 btree_flush_write(c);
592 continue_at(cl, journal_write, system_wq);
593 }
594
595 c->journal.blocks_free -= set_blocks(w->data, c);
596
597 w->data->btree_level = c->root->level;
598
599 bkey_copy(&w->data->btree_root, &c->root->key);
600 bkey_copy(&w->data->uuid_bucket, &c->uuid_bucket);
601
602 for_each_cache(ca, c, i)
603 w->data->prio_bucket[ca->sb.nr_this_dev] = ca->prio_buckets[0];
604
605 w->data->magic = jset_magic(c);
606 w->data->version = BCACHE_JSET_VERSION;
607 w->data->last_seq = last_seq(&c->journal);
608 w->data->csum = csum_set(w->data);
609
610 for (i = 0; i < KEY_PTRS(k); i++) {
611 ca = PTR_CACHE(c, k, i);
612 bio = &ca->journal.bio;
613
614 atomic_long_add(sectors, &ca->meta_sectors_written);
615
616 bio_reset(bio);
617 bio->bi_sector = PTR_OFFSET(k, i);
618 bio->bi_bdev = ca->bdev;
619 bio->bi_rw = REQ_WRITE|REQ_SYNC|REQ_META|REQ_FLUSH;
620 bio->bi_size = sectors << 9;
621
622 bio->bi_end_io = journal_write_endio;
623 bio->bi_private = w;
624 bio_map(bio, w->data);
625
626 trace_bcache_journal_write(bio);
627 bio_list_add(&list, bio);
628
629 SET_PTR_OFFSET(k, i, PTR_OFFSET(k, i) + sectors);
630
631 ca->journal.seq[ca->journal.cur_idx] = w->data->seq;
632 }
633
634 atomic_dec_bug(&fifo_back(&c->journal.pin));
635 bch_journal_next(&c->journal);
636 journal_reclaim(c);
637
638 spin_unlock(&c->journal.lock);
639
640 while ((bio = bio_list_pop(&list)))
641 closure_bio_submit(bio, cl, c->cache[0]);
642
643 continue_at(cl, journal_write_done, NULL);
644 }
645
646 static void journal_write(struct closure *cl)
647 {
648 struct cache_set *c = container_of(cl, struct cache_set, journal.io.cl);
649
650 spin_lock(&c->journal.lock);
651 journal_write_unlocked(cl);
652 }
653
654 static void __journal_try_write(struct cache_set *c, bool noflush)
655 {
656 struct closure *cl = &c->journal.io.cl;
657
658 if (!closure_trylock(cl, &c->cl))
659 spin_unlock(&c->journal.lock);
660 else if (noflush && journal_full(&c->journal)) {
661 spin_unlock(&c->journal.lock);
662 continue_at(cl, journal_write, system_wq);
663 } else
664 journal_write_unlocked(cl);
665 }
666
667 #define journal_try_write(c) __journal_try_write(c, false)
668
669 void bch_journal_meta(struct cache_set *c, struct closure *cl)
670 {
671 struct journal_write *w;
672
673 if (CACHE_SYNC(&c->sb)) {
674 spin_lock(&c->journal.lock);
675
676 w = c->journal.cur;
677 w->need_write = true;
678
679 if (cl)
680 BUG_ON(!closure_wait(&w->wait, cl));
681
682 __journal_try_write(c, true);
683 }
684 }
685
686 /*
687 * Entry point to the journalling code - bio_insert() and btree_invalidate()
688 * pass bch_journal() a list of keys to be journalled, and then
689 * bch_journal() hands those same keys off to btree_insert_async()
690 */
691
692 void bch_journal(struct closure *cl)
693 {
694 struct btree_op *op = container_of(cl, struct btree_op, cl);
695 struct cache_set *c = op->c;
696 struct journal_write *w;
697 size_t b, n = ((uint64_t *) op->keys.top) - op->keys.list;
698
699 if (op->type != BTREE_INSERT ||
700 !CACHE_SYNC(&c->sb))
701 goto out;
702
703 /*
704 * If we're looping because we errored, might already be waiting on
705 * another journal write:
706 */
707 while (atomic_read(&cl->parent->remaining) & CLOSURE_WAITING)
708 closure_sync(cl->parent);
709
710 spin_lock(&c->journal.lock);
711
712 if (journal_full(&c->journal)) {
713 /* XXX: tracepoint */
714 closure_wait(&c->journal.wait, cl);
715
716 journal_reclaim(c);
717 spin_unlock(&c->journal.lock);
718
719 btree_flush_write(c);
720 continue_at(cl, bch_journal, bcache_wq);
721 }
722
723 w = c->journal.cur;
724 w->need_write = true;
725 b = __set_blocks(w->data, w->data->keys + n, c);
726
727 if (b * c->sb.block_size > PAGE_SECTORS << JSET_BITS ||
728 b > c->journal.blocks_free) {
729 /* XXX: If we were inserting so many keys that they won't fit in
730 * an _empty_ journal write, we'll deadlock. For now, handle
731 * this in bch_keylist_realloc() - but something to think about.
732 */
733 BUG_ON(!w->data->keys);
734
735 /* XXX: tracepoint */
736 BUG_ON(!closure_wait(&w->wait, cl));
737
738 closure_flush(&c->journal.io);
739
740 journal_try_write(c);
741 continue_at(cl, bch_journal, bcache_wq);
742 }
743
744 memcpy(end(w->data), op->keys.list, n * sizeof(uint64_t));
745 w->data->keys += n;
746
747 op->journal = &fifo_back(&c->journal.pin);
748 atomic_inc(op->journal);
749
750 if (op->flush_journal) {
751 closure_flush(&c->journal.io);
752 closure_wait(&w->wait, cl->parent);
753 }
754
755 journal_try_write(c);
756 out:
757 bch_btree_insert_async(cl);
758 }
759
760 void bch_journal_free(struct cache_set *c)
761 {
762 free_pages((unsigned long) c->journal.w[1].data, JSET_BITS);
763 free_pages((unsigned long) c->journal.w[0].data, JSET_BITS);
764 free_fifo(&c->journal.pin);
765 }
766
767 int bch_journal_alloc(struct cache_set *c)
768 {
769 struct journal *j = &c->journal;
770
771 closure_init_unlocked(&j->io);
772 spin_lock_init(&j->lock);
773
774 c->journal_delay_ms = 100;
775
776 j->w[0].c = c;
777 j->w[1].c = c;
778
779 if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)) ||
780 !(j->w[0].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)) ||
781 !(j->w[1].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)))
782 return -ENOMEM;
783
784 return 0;
785 }
Linux kernel - Deliverables
This page took 0.047575 seconds and 5 git commands to generate.