summaryrefslogtreecommitdiff
path: root/drivers/md/bcache/journal.c
blob: 8dce022a2b5777831fa1f9a9472404700193e609 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
/*
 * bcache journalling code, for btree insertions
 *
 * Copyright 2012 Google, Inc.
 */

#include "bcache.h"
#include "buckets.h"
#include "btree.h"
#include "debug.h"
#include "extents.h"
#include "gc.h"
#include "io.h"
#include "journal.h"
#include "super.h"

#include <trace/events/bcache.h>

/* Sequence number of oldest dirty journal entry */
#define last_seq(j)	((j)->seq - fifo_used(&(j)->pin) + 1)

#define for_each_jset_key(k, jkeys, jset)			\
	for_each_jset_jkeys(jkeys, jset)			\
		if (JKEYS_TYPE(jkeys) == JKEYS_BTREE_KEYS)	\
			for (k = (jkeys)->start;		\
			     k < bset_bkey_last(jkeys);		\
			     k = bkey_next(k))

static inline void bch_journal_add_entry(struct jset *j, const void *data,
					 size_t u64s, unsigned type,
					 enum btree_id id, unsigned level)
{
	struct jset_entry *jkeys = (struct jset_entry *) bset_bkey_last(j);

	jkeys->u64s = u64s;
	jkeys->btree_id = id;
	jkeys->level = level;
	jkeys->flags = 0;
	SET_JKEYS_TYPE(jkeys, type);

	memcpy(jkeys->_data, data, u64s * sizeof(u64));
	j->u64s += jset_u64s(u64s);
}

static struct jset_entry *bch_journal_find_entry(struct jset *j, unsigned type,
						enum btree_id id)
{
	struct jset_entry *jkeys;

	for_each_jset_jkeys(jkeys, j)
		if (JKEYS_TYPE(jkeys) == type && jkeys->btree_id == id)
			return jkeys;

	return NULL;
}

struct bkey *bch_journal_find_btree_root(struct cache_set *c, struct jset *j,
					 enum btree_id id, unsigned *level)
{
	struct bkey *k;
	struct jset_entry *jkeys =
		bch_journal_find_entry(j, JKEYS_BTREE_ROOT, id);

	if (!jkeys)
		return NULL;

	k = jkeys->start;
	*level = jkeys->level;

	if (!jkeys->u64s || jkeys->u64s != k->u64s ||
	    bkey_invalid(c, BKEY_TYPE_BTREE, k)) {
		bch_cache_set_error(c, "invalid btree root in journal");
		return NULL;
	}

	*level = jkeys->level;
	return k;
}

static void bch_journal_add_btree_root(struct jset *j, enum btree_id id,
				       struct bkey *k, unsigned level)
{
	bch_journal_add_entry(j, k, k->u64s, JKEYS_BTREE_ROOT, id, level);
}

static inline void bch_journal_add_prios(struct cache_set *c, struct jset *j)
{
	bch_journal_add_entry(j, c->journal.prio_buckets, c->sb.nr_in_set,
			      JKEYS_PRIO_PTRS, 0, 0);
}

/*
 * Journal replay/recovery:
 *
 * This code is all driven from run_cache_set(); we first read the journal
 * entries, do some other stuff, then we mark all the keys in the journal
 * entries (same as garbage collection would), then we replay them - reinserting
 * them into the cache in precisely the same order as they appear in the
 * journal.
 *
 * We only journal keys that go in leaf nodes, which simplifies things quite a
 * bit.
 */

struct journal_list {
	struct closure		cl;
	struct mutex		lock;
	struct mutex		cache_set_buffer_lock;
	struct list_head	*head;
	int			ret;
};

static int journal_add_entry(struct journal_list *jlist, struct jset *j)
{
	struct journal_replay *i, *pos;
	struct list_head *where;
	size_t bytes = set_bytes(j);
	int ret = 0;

	mutex_lock(&jlist->lock);

	/* This entry too old? */
	if (!list_empty(jlist->head)) {
		i = list_last_entry(jlist->head, struct journal_replay, list);
		if (j->seq < i->j.last_seq) {
			pr_debug("j->seq %llu i->j.seq %llu",
				 j->seq, i->j.seq);
			goto out;
		}
	}

	ret = 1;

	/* Drop entries we don't need anymore */
	list_for_each_entry_safe(i, pos, jlist->head, list) {
		if (i->j.seq >= j->last_seq)
			break;
		list_del(&i->list);
		kfree(i);
	}

	list_for_each_entry_reverse(i, jlist->head, list) {
		if (j->seq == i->j.seq) {
			pr_debug("j->seq %llu i->j.seq %llu",
				 j->seq, i->j.seq);
			goto out;
		}

		if (j->seq > i->j.seq) {
			where = &i->list;
			goto add;
		}
	}

	where = jlist->head;
add:
	i = kmalloc(offsetof(struct journal_replay, j) + bytes, GFP_KERNEL);
	if (!i) {
		ret = -ENOMEM;
		goto out;
	}

	memcpy(&i->j, j, bytes);
	list_add(&i->list, where);

	pr_debug("seq %llu", j->seq);
out:
	mutex_unlock(&jlist->lock);
	return ret;
}

static int journal_read_bucket(struct cache *ca, struct journal_list *jlist,
			       unsigned bucket_index, u64 *seq)
{
	struct cache_set *c = ca->set;
	struct journal_device *ja = &ca->journal;
	struct bio *bio = &ja->bio;
	struct jset *j, *data;
	unsigned len, left, offset = 0;
	sector_t bucket = bucket_to_sector(ca,
				journal_bucket(ca, bucket_index));
	bool entries_found = false;
	int ret = 0;

	data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS);
	if (!data) {
		mutex_lock(&jlist->cache_set_buffer_lock);
		data = c->journal.w[0].data;
	}

	pr_debug("reading %u", bucket_index);

	while (offset < ca->mi.bucket_size) {
reread:		left = ca->mi.bucket_size - offset;
		len = min_t(unsigned, left, PAGE_SECTORS << JSET_BITS);

		bio_reset(bio);
		bio->bi_bdev		= ca->disk_sb.bdev;
		bio->bi_iter.bi_sector	= bucket + offset;
		bio->bi_iter.bi_size	= len << 9;
		bch_bio_map(bio, data);

		ret = submit_bio_wait(READ_SYNC, bio);
		if (bch_meta_read_fault("journal"))
			ret = -EIO;
		if (ret) {
			__bch_cache_error(ca,
					  "IO error %d reading journal "
					  "from offset %zu",
					  ret, bucket + offset);
			goto err;
		}

		/* This function could be simpler now since we no longer write
		 * journal entries that overlap bucket boundaries; this means
		 * the start of a bucket will always have a valid journal entry
		 * if it has any journal entries at all.
		 */

		j = data;
		while (len) {
			size_t blocks, bytes = set_bytes(j);
			u64 got, expect;

			if (bch_meta_read_fault("journal"))
				goto err;

			if (j->magic != jset_magic(&c->sb)) {
				pr_debug("%u: bad magic", bucket_index);
				goto err;
			}

			got = j->version;
			expect = BCACHE_JSET_VERSION;
			if (got != expect) {
				__bch_cache_error(ca,
					"bad version (got %llu expect %llu) "
					"while reading journal from offset %zu",
						  got, expect, bucket + offset);
				goto err;
			}

			if (bytes > left << 9 ||
			    bytes > PAGE_SIZE << JSET_BITS) {
				__bch_cache_error(ca,
					"too big (%zu bytes) "
					"while reading journal from offset %zu",
						  bytes, bucket + offset);
				goto err;
			}

			if (bytes > len << 9)
				goto reread;

			got = j->csum;
			expect = csum_set(j, JSET_CSUM_TYPE(j));
			if (got != expect) {
				__bch_cache_error(ca,
					"bad checksum (got %llu expect %llu) "
					"while reading journal from offset %zu",
						  got, expect, bucket + offset);
				goto err;
			}

			ret = journal_add_entry(jlist, j);
			if (ret < 0)
				goto err;
			if (ret) {
				ja->seq[bucket_index] = j->seq;
				entries_found = true;
			}

			if (j->seq > *seq)
				*seq = j->seq;

			blocks = set_blocks(j, block_bytes(c));

			pr_debug("next");
			offset	+= blocks * ca->sb.block_size;
			len	-= blocks * ca->sb.block_size;
			j = ((void *) j) + blocks * block_bytes(ca);
		}
	}

	ret = entries_found;
err:
	if (data == c->journal.w[0].data)
		mutex_unlock(&jlist->cache_set_buffer_lock);
	else
		free_pages((unsigned long) data, JSET_BITS);

	return ret;
}

static void bch_journal_read_device(struct closure *cl)
{
#define read_bucket(b)							\
	({								\
		int ret = journal_read_bucket(ca, jlist, b, &seq);	\
		__set_bit(b, bitmap);					\
		if (ret < 0) {						\
			mutex_lock(&jlist->lock);			\
			jlist->ret = ret;				\
			mutex_unlock(&jlist->lock);			\
			closure_return(cl);				\
		}							\
		ret;							\
	 })

	struct journal_device *ja =
		container_of(cl, struct journal_device, read);
	struct cache *ca = container_of(ja, struct cache, journal);
	struct journal_list *jlist =
		container_of(cl->parent, struct journal_list, cl);

	unsigned nr_buckets = bch_nr_journal_buckets(&ca->sb);
	unsigned long bitmap[BITS_TO_LONGS(nr_buckets)];
	unsigned i, l, r, m;
	u64 seq = 0;

	bitmap_zero(bitmap, nr_buckets);
	pr_debug("%u journal buckets", nr_buckets);

	if (!blk_queue_nonrot(bdev_get_queue(ca->disk_sb.bdev)))
		goto linear_scan;

	/*
	 * Read journal buckets ordered by golden ratio hash to quickly
	 * find a sequence of buckets with valid journal entries
	 */
	for (i = 0; i < nr_buckets; i++) {
		l = (i * 2654435769U) % nr_buckets;

		if (test_bit(l, bitmap))
			break;

		if (read_bucket(l))
			goto bsearch;
	}

	/*
	 * If that fails, check all the buckets we haven't checked
	 * already
	 */
	pr_debug("falling back to linear search");
linear_scan:
	for (l = find_first_zero_bit(bitmap, nr_buckets);
	     l < nr_buckets;
	     l = find_next_zero_bit(bitmap, nr_buckets, l + 1))
		if (read_bucket(l))
			goto bsearch;

	/* no journal entries on this device? */
	if (l == nr_buckets)
		closure_return(cl);
bsearch:
	/* Binary search */
	m = l;
	r = find_next_bit(bitmap, nr_buckets, l + 1);
	pr_debug("starting binary search, l %u r %u", l, r);

	while (l + 1 < r) {
		u64 cur_seq = seq;

		m = (l + r) >> 1;
		read_bucket(m);

		if (cur_seq != seq)
			l = m;
		else
			r = m;
	}

	/*
	 * Read buckets in reverse order until we stop finding more
	 * journal entries
	 */
	pr_debug("finishing up: m %u njournal_buckets %u",
		 m, nr_buckets);
	l = m;

	while (1) {
		if (!l--)
			l = nr_buckets - 1;

		if (l == m)
			break;

		if (test_bit(l, bitmap))
			continue;

		if (!read_bucket(l))
			break;
	}

	seq = 0;

	for (i = 0; i < nr_buckets; i++)
		if (ja->seq[i] > seq) {
			seq = ja->seq[i];
			/*
			 * When journal_next_bucket() goes to allocate for
			 * the first time, it'll use the bucket after
			 * ja->cur_idx
			 */
			ja->cur_idx = i;
			ja->last_idx = ja->discard_idx = (i + 1) %
				nr_buckets;
			pr_debug("cur_idx %d last_idx %d",
				 ja->cur_idx, ja->last_idx);
		}

	closure_return(cl);
#undef read_bucket
}

static void journal_entries_free(struct list_head *list)
{

	while (!list_empty(list)) {
		struct journal_replay *i =
			list_first_entry(list, struct journal_replay, list);
		list_del(&i->list);
		kfree(i);
	}
}

const char *bch_journal_read(struct cache_set *c, struct list_head *list)
{
	struct jset_entry *prio_ptrs;
	struct journal_list jlist;
	struct jset *j;
	struct list_head *l;
	struct cache *ca;
	unsigned iter;

	closure_init_stack(&jlist.cl);
	mutex_init(&jlist.lock);
	mutex_init(&jlist.cache_set_buffer_lock);
	jlist.head = list;
	jlist.ret = 0;

	for_each_cache(ca, c, iter)
		closure_call(&ca->journal.read,
			     bch_journal_read_device,
			     system_unbound_wq,
			     &jlist.cl);

	closure_sync(&jlist.cl);

	if (jlist.ret) {
		journal_entries_free(list);

		return jlist.ret == -ENOMEM
			? "cannot allocate memory for journal"
			: "error reading journal";
	}

	if (list_empty(list))
		return "no journal entries found";

	/* XXX: this can't tolerate missing journal entries */
	list_for_each(l, list) {
		atomic_t p = { 1 };

		BUG_ON(!fifo_push(&c->journal.pin, p));
		atomic_set(&fifo_back(&c->journal.pin), 1);
	}

	j = &list_entry(list->prev, struct journal_replay, list)->j;

	c->journal.seq = j->seq;

	BUG_ON(last_seq(&c->journal) != j->last_seq);

	prio_ptrs = bch_journal_find_entry(j, JKEYS_PRIO_PTRS, 0);
	if (!prio_ptrs)
		return "prio bucket ptrs not found";

	memcpy(c->journal.prio_buckets,
	       prio_ptrs->_data,
	       prio_ptrs->u64s * sizeof(u64));

	return NULL;
}

void bch_journal_mark(struct cache_set *c, struct list_head *list)
{
	struct bkey *k;
	struct jset_entry *j;
	struct journal_replay *r;

	list_for_each_entry(r, list, list)
		for_each_jset_key(k, j, &r->j) {
			if ((j->level || j->btree_id == BTREE_ID_EXTENTS) &&
			    !bkey_invalid(c, j->level
					  ? BKEY_TYPE_BTREE : j->btree_id, k))
				__bch_btree_mark_key(c, j->level, k);
		}
}

static int bch_journal_replay_key(struct cache_set *c, enum btree_id id,
				  struct bkey *k)
{
	int ret;
	BKEY_PADDED(key) temp;
	bool do_subtract = id == BTREE_ID_EXTENTS && k->type == BCH_EXTENT;

	trace_bcache_journal_replay_key(k);

	if (do_subtract)
		bkey_copy(&temp.key, k);

	ret = bch_btree_insert(c, id, &keylist_single(k), NULL, NULL);
	if (ret)
		return ret;

	/*
	 * Subtract sectors after replay since bch_btree_insert() added
	 * them again
	 */
	if (do_subtract)
		__bch_add_sectors(c, NULL, bkey_i_to_extent_c(&temp.key),
				  bkey_start_offset(&temp.key),
				  -temp.key.size, false);

	return 0;
}

int bch_journal_replay(struct cache_set *c, struct list_head *list)
{
	int ret = 0, keys = 0, entries = 0;
	struct bkey *k;
	struct jset_entry *jkeys;
	struct journal_replay *i =
		list_entry(list->prev, struct journal_replay, list);

	uint64_t start = i->j.last_seq, end = i->j.seq, n = start;

	list_for_each_entry(i, list, list) {
		cache_set_err_on(n != i->j.seq, c,
"bcache: journal entries %llu-%llu missing! (replaying %llu-%llu)",
				 n, i->j.seq - 1, start, end);

		c->journal.cur_pin =
			&c->journal.pin.data[((c->journal.pin.back - 1 -
					       (c->journal.seq - i->j.seq)) &
					      c->journal.pin.mask)];

		BUG_ON(atomic_read(c->journal.cur_pin) != 1);

		for_each_jset_key(k, jkeys, &i->j) {
			cond_resched();
			ret = bch_journal_replay_key(c, jkeys->btree_id, k);
			if (ret)
				goto err;

			keys++;
		}

		if (atomic_dec_and_test(c->journal.cur_pin))
			wake_up(&c->journal.wait);

		n = i->j.seq + 1;
		entries++;
	}

	pr_info("journal replay done, %i keys in %i entries, seq %llu",
		keys, entries, end);
err:
	if (ret)
		pr_err("journal replay error: %d", ret);

	journal_entries_free(list);

	return ret;
}

static int bch_set_nr_journal_buckets(struct cache *ca, unsigned nr)
{
	unsigned u64s = bch_journal_buckets_offset(&ca->sb) + nr;
	u64 *p;
	int ret;

	ret = bch_super_realloc(&ca->disk_sb, u64s);
	if (ret)
		return ret;

	p = krealloc(ca->journal.seq, nr * sizeof(u64), GFP_KERNEL|__GFP_ZERO);
	if (!p)
		return -ENOMEM;

	ca->journal.seq = p;
	ca->sb.u64s = u64s;

	return 0;
}

int bch_cache_journal_alloc(struct cache *ca)
{
	int ret;
	unsigned i;

	if (CACHE_TIER(&ca->mi) != 0)
		return 0;

	/* clamp journal size to 512MB (in sectors) */

	ret = bch_set_nr_journal_buckets(ca,
			clamp_t(unsigned, ca->mi.nbuckets >> 8,
				8, (1 << 20) / ca->mi.bucket_size));
	if (ret)
		return ret;

	for (i = 0; i < bch_nr_journal_buckets(&ca->sb); i++) {
		unsigned long r = ca->mi.first_bucket + i;
		bch_mark_metadata_bucket(ca, &ca->buckets[r], true);
		set_journal_bucket(ca, i, r);
	}

	return 0;
}

/* Journalling */

static void journal_discard_endio(struct bio *bio, int error)
{
	struct journal_device *ja =
		container_of(bio, struct journal_device, discard_bio);
	struct cache *ca = container_of(ja, struct cache, journal);

	atomic_set(&ja->discard_in_flight, DISCARD_DONE);

	wake_up(&ca->set->journal.wait);
	closure_put(&ca->set->cl);
}

static void journal_discard_work(struct work_struct *work)
{
	struct journal_device *ja =
		container_of(work, struct journal_device, discard_work);

	submit_bio(0, &ja->discard_bio);
}

static bool do_journal_discard(struct cache *ca)
{
	struct journal_device *ja = &ca->journal;
	struct bio *bio = &ja->discard_bio;

	if (!CACHE_DISCARD(&ca->mi) ||
	    !blk_queue_discard(bdev_get_queue(ca->disk_sb.bdev))) {
		ja->discard_idx = ja->last_idx;
		return true;
	}

	switch (atomic_read(&ja->discard_in_flight)) {
	case DISCARD_IN_FLIGHT:
		return false;

	case DISCARD_DONE:
		ja->discard_idx = (ja->discard_idx + 1) %
			bch_nr_journal_buckets(&ca->sb);

		atomic_set(&ja->discard_in_flight, DISCARD_READY);
		/* fallthrough */

	case DISCARD_READY:
		if (ja->discard_idx == ja->last_idx)
			return true;

		atomic_set(&ja->discard_in_flight, DISCARD_IN_FLIGHT);

		bio_init(bio);
		bio->bi_iter.bi_sector	= bucket_to_sector(ca,
						journal_bucket(ca, ja->discard_idx));
		bio->bi_bdev		= ca->disk_sb.bdev;
		bio->bi_rw		= REQ_WRITE|REQ_DISCARD;
		bio->bi_max_vecs	= 1;
		bio->bi_io_vec		= bio->bi_inline_vecs;
		bio->bi_iter.bi_size	= bucket_bytes(ca);
		bio->bi_end_io		= journal_discard_endio;

		closure_get(&ca->set->cl);
		INIT_WORK(&ja->discard_work, journal_discard_work);
		schedule_work(&ja->discard_work);

		return false;

	default:
		BUG();
		return true;
	}
}

/* Number of journal buckets which may be discarded and reused */
static inline unsigned journal_free_buckets(struct cache *ca)
{
	struct journal_device *ja = &ca->journal;
	unsigned nr = bch_nr_journal_buckets(&ca->sb);
	return (ja->last_idx - ja->cur_idx + nr - 1) % nr;
}

#define JSET_SECTORS (PAGE_SECTORS << JSET_BITS)

/* Number of u64s we can write to the current journal bucket */
static size_t journal_write_u64s_remaining(struct cache_set *c)
{
	ssize_t u64s = (min_t(size_t,
			      c->journal.sectors_free,
			      JSET_SECTORS) << 9) / sizeof(u64);

	/* Subtract off some for the btree roots */
	u64s -= BTREE_ID_NR * (JSET_KEYS_U64s + BKEY_EXTENT_MAX_U64s);

	/* And for the prio pointers */
	u64s -= JSET_KEYS_U64s + c->sb.nr_in_set;

	return max_t(ssize_t, 0L, u64s);
}

static void journal_entry_no_room(struct cache_set *c)
{
	struct bkey_i_extent *e = bkey_i_to_extent(&c->journal.key);
	struct bch_extent_ptr *ptr;
	struct cache *ca;

	extent_for_each_ptr_backwards(e, ptr)
		if (!(ca = PTR_CACHE(c, ptr)) ||
		    ca->journal.sectors_free <= c->journal.sectors_free)
			bch_extent_drop_ptr(&e->k, ptr - e->v.ptr);

	c->journal.sectors_free = 0;
	c->journal.u64s_remaining = 0;
}

/**
 * journal_reclaim_work - free up journal buckets
 *
 * Background journal reclaim writes out btree nodes. It should be run
 * early enough so that we never completely run out of journal buckets.
 *
 * High watermarks for triggering background reclaim:
 * - FIFO has fewer than 512 entries left
 * - fewer than 25% journal buckets free
 *
 * Background reclaim runs until low watermarks are reached:
 * - FIFO has more than 1024 entries left
 * - more than 50% journal buckets free
 *
 * As long as a reclaim can complete in the time it takes to fill up
 * 512 journal entries or 25% of all journal buckets, then
 * journal_next_bucket() should not stall.
 */
static void journal_reclaim_work(struct work_struct *work)
{
	struct cache_set *c = container_of(work, struct cache_set,
					   journal.reclaim_work);
	struct cache *ca;
	u64 oldest_seq = 0;
	unsigned iter;
	bool flush = false;
	size_t fifo_cutoff;

	pr_debug("started");
	spin_lock(&c->journal.lock);

	rcu_read_lock();

	for_each_cache_rcu(ca, c, iter) {
		struct journal_device *ja = &ca->journal;
		unsigned nr = bch_nr_journal_buckets(&ca->sb);
		unsigned next = (ja->cur_idx + (nr >> 1)) % nr;

		if ((CACHE_TIER(&ca->mi) != 0)
		    || (CACHE_STATE(&ca->mi) != CACHE_ACTIVE))
			continue;

		/*
		 * Write out enough btree nodes to free up 50% journal
		 * buckets
		 */
		if (journal_free_buckets(ca) < (nr >> 1)) {
			oldest_seq = max_t(u64, oldest_seq,
					   ja->seq[next]);
			flush = true;
		}
	}

	rcu_read_unlock();

	fifo_cutoff = (c->journal.pin.size >> 5);
	if (fifo_free(&c->journal.pin) <= fifo_cutoff) {
		/* Write out enough btree nodes to free up 33% of the FIFO */
		oldest_seq = max_t(u64, oldest_seq,
				   last_seq(&c->journal)
				   + fifo_cutoff);
		flush = true;
	}

	if (!flush) {
		spin_unlock(&c->journal.lock);
		return;
	}

	BUG_ON(oldest_seq == 0);
	oldest_seq -= last_seq(&c->journal);
	spin_unlock(&c->journal.lock);
	bch_btree_write_oldest(c, oldest_seq);
}

/**
 * journal_reclaim_fast - do the fast part of journal reclaim
 *
 * Called from IO submission context, does not block. Cleans up after btree
 * write completions by advancing the journal pin and each cache's last_idx,
 * and kicking off discards.
 *
 * Returns true if background reclaim needs to run.
 */
static bool journal_reclaim_fast(struct cache_set *c)
{
	struct cache *ca;
	u64 last_seq;
	atomic_t p;
	unsigned iter;
	bool need_reclaim = false;

	lockdep_assert_held(&c->journal.lock);

	/*
	 * Unpin journal entries whose reference counts reached zero, meaning
	 * all btree nodes got written out
	 */
	while (!atomic_read(&fifo_front(&c->journal.pin)))
		fifo_pop(&c->journal.pin, p);

	/* If less than 15% space free, kick off background reclaim */
	if (fifo_free(&c->journal.pin) <= (c->journal.pin.size >> 6))
		need_reclaim = true;

	last_seq = last_seq(&c->journal);

	rcu_read_lock();

	/*
	 * Advance last_idx to point to the oldest journal entry containing
	 * btree node updates that have not yet been written out
	 */
	for_each_cache_rcu(ca, c, iter) {
		struct journal_device *ja = &ca->journal;
		unsigned nr = bch_nr_journal_buckets(&ca->sb);

		if ((CACHE_TIER(&ca->mi) != 0)
		    || (CACHE_STATE(&ca->mi) != CACHE_ACTIVE))
			continue;

		while (ja->last_idx != ja->cur_idx &&
		       ja->seq[ja->last_idx] < last_seq)
			ja->last_idx = (ja->last_idx + 1) % nr;

		/*
		 * If less than 25% journal buckets free, kick off
		 * background reclaim
		 */
		if (journal_free_buckets(ca) < (nr >> 2))
			need_reclaim = true;

		do_journal_discard(ca);
	}

	rcu_read_unlock();

	return need_reclaim;
}

/**
 * journal_next_bucket - move on to the next journal bucket if possible
 */
static void journal_next_bucket(struct cache_set *c)
{
	struct bkey_i_extent *e = bkey_i_to_extent(&c->journal.key);
	struct bch_extent_ptr *ptr;
	struct cache *ca;
	unsigned iter;

	pr_debug("started");
	lockdep_assert_held(&c->journal.lock);

	/*
	 * only supposed to be called when we're out of space/haven't started a
	 * new journal entry
	 */
	BUG_ON(c->journal.u64s_remaining);
	BUG_ON(c->journal.cur->data->u64s);

	if (journal_reclaim_fast(c))
		queue_work(system_long_wq, &c->journal.reclaim_work);

	rcu_read_lock();

	/*
	 * Drop any pointers to devices that have been removed, are no longer
	 * empty, or filled up their current journal bucket:
	 */
	extent_for_each_ptr_backwards(e, ptr)
		if (!(ca = PTR_CACHE(c, ptr)) ||
		    !ca->journal.sectors_free ||
		    CACHE_STATE(&ca->mi) != CACHE_ACTIVE)
			bch_extent_drop_ptr(&e->k, ptr - e->v.ptr);

	/*
	 * Determine location of the next journal write:
	 * XXX: sort caches by free journal space
	 */
	for_each_cache_rcu(ca, c, iter) {
		struct journal_device *ja = &ca->journal;
		unsigned next = (ja->cur_idx + 1) %
			bch_nr_journal_buckets(&ca->sb);

		if (bch_extent_ptrs(e) == CACHE_SET_META_REPLICAS_WANT(&c->sb))
			break;

		/*
		 * Check that we can use this device, and aren't already using
		 * it:
		 */
		if ((CACHE_TIER(&ca->mi) != 0) ||
		    (CACHE_STATE(&ca->mi) != CACHE_ACTIVE) ||
		    bch_extent_has_device(e, ca->sb.nr_this_dev))
			continue;

		/* No journal buckets available for writing on this device */
		if (next == ja->discard_idx)
			continue;

		BUG_ON(bch_extent_ptrs(e) >= BKEY_EXTENT_PTRS_MAX);

		ja->sectors_free = ca->mi.bucket_size;

		ja->cur_idx = next;
		e->v.ptr[bch_extent_ptrs(e)] =
			PTR(0, bucket_to_sector(ca, journal_bucket(ca,
							ja->cur_idx)),
			    ca->sb.nr_this_dev);

		trace_bcache_journal_next_bucket(ca, ja->cur_idx,
				ja->last_idx, ja->discard_idx);
		bch_set_extent_ptrs(e, bch_extent_ptrs(e) + 1);
	}

	/* set c->journal.sectors_free to the min of any device */
	c->journal.sectors_free = UINT_MAX;

	extent_for_each_online_device(c, e, ptr, ca)
		c->journal.sectors_free = min(c->journal.sectors_free,
					     ca->journal.sectors_free);
	if (c->journal.sectors_free == UINT_MAX)
		c->journal.sectors_free = 0;

	rcu_read_unlock();
}

void bch_journal_next(struct journal *j)
{
	atomic_t p = { 1 };

	j->cur = (j->cur == j->w)
		? &j->w[1]
		: &j->w[0];

	/*
	 * The fifo_push() needs to happen at the same time as j->seq is
	 * incremented for last_seq() to be calculated correctly
	 */
	BUG_ON(!fifo_push(&j->pin, p));
	atomic_set(&fifo_back(&j->pin), 1);

	if (test_bit(JOURNAL_REPLAY_DONE, &j->flags))
		j->cur_pin = &fifo_back(&j->pin);

	j->cur->data->seq	= ++j->seq;
	j->cur->data->u64s	= 0;
	j->u64s_remaining	= 0;
}

static void journal_write_endio(struct bio *bio, int error)
{
	struct cache *ca = container_of(bio, struct cache, journal.bio);
	struct journal_write *w = bio->bi_private;

	if (error || bch_meta_write_fault("journal"))
		bch_cache_error(ca, "IO error %d writing journal", error);

	closure_put(&w->c->journal.io);
	percpu_ref_put(&ca->ref);
}

static void journal_write_done(struct closure *cl)
{
	struct journal *j = container_of(cl, struct journal, io);
	struct journal_write *w = (j->cur == j->w)
		? &j->w[1]
		: &j->w[0];

	__closure_wake_up(&w->wait);

	atomic_set(&j->in_flight, 0);
	wake_up(&j->wait);

	if (test_bit(JOURNAL_NEED_WRITE, &j->flags))
		mod_delayed_work(system_wq, &j->write_work, 0);
}

static void journal_write_locked(struct closure *cl)
	__releases(c->journal.lock)
{
	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
	struct cache *ca;
	struct journal_write *w = c->journal.cur;
	struct bkey_i_extent *e = bkey_i_to_extent(&c->journal.key);
	struct bch_extent_ptr *ptr;
	BKEY_PADDED(k) tmp;
	unsigned i, sectors;

	struct bio *bio;
	struct bio_list list;
	bio_list_init(&list);

	BUG_ON(c->journal.res_count);
	BUG_ON(journal_full(&c->journal));

	clear_bit(JOURNAL_NEED_WRITE, &c->journal.flags);
	clear_bit(JOURNAL_DIRTY, &c->journal.flags);
	cancel_delayed_work(&c->journal.write_work);

	spin_lock(&c->btree_root_lock);

	for (i = 0; i < BTREE_ID_NR; i++) {
		struct btree *b = c->btree_roots[i];

		if (b)
			bch_journal_add_btree_root(w->data, i,
						   &b->key, b->level);
	}

	spin_unlock(&c->btree_root_lock);

	bch_journal_add_prios(c, w->data);

	w->data->read_clock	= c->prio_clock[READ].hand;
	w->data->write_clock	= c->prio_clock[WRITE].hand;
	w->data->magic		= jset_magic(&c->sb);
	w->data->version	= BCACHE_JSET_VERSION;
	w->data->last_seq	= last_seq(&c->journal);

	SET_JSET_CSUM_TYPE(w->data, CACHE_PREFERRED_CSUM_TYPE(&c->sb));
	w->data->csum		= csum_set(w->data, JSET_CSUM_TYPE(w->data));

	sectors = set_blocks(w->data, block_bytes(c)) * c->sb.block_size;

	BUG_ON(sectors > c->journal.sectors_free);
	c->journal.sectors_free -= sectors;

	extent_for_each_ptr(e, ptr) {
		rcu_read_lock();
		ca = PTR_CACHE(c, ptr);
		if (ca)
			percpu_ref_get(&ca->ref);
		rcu_read_unlock();

		if (!ca) {
			/* XXX: fix this */
			pr_err("missing journal write\n");
			continue;
		}

		BUG_ON(sectors > ca->journal.sectors_free);
		ca->journal.sectors_free -= sectors;

		/*
		 * If we don't have enough space for a full-sized journal
		 * entry, go to the next bucket. We do this check after
		 * the write, so that if our bucket size is really small
		 * we don't get stuck forever.
		 */
		if (ca->journal.sectors_free < JSET_SECTORS) {
			c->journal.sectors_free = 0;
			ca->journal.sectors_free = 0;
		}

		bio = &ca->journal.bio;

		atomic_long_add(sectors, &ca->meta_sectors_written);

		bio_reset(bio);
		bio->bi_iter.bi_sector	= PTR_OFFSET(ptr);
		bio->bi_bdev		= ca->disk_sb.bdev;
		bio->bi_rw		= (REQ_WRITE|REQ_SYNC|REQ_META|
					   REQ_FLUSH|REQ_FUA);
		bio->bi_iter.bi_size	= sectors << 9;

		bio->bi_end_io		= journal_write_endio;
		bio->bi_private		= w;
		bch_bio_map(bio, w->data);

		trace_bcache_journal_write(bio);
		bio_list_add(&list, bio);

		SET_PTR_OFFSET(ptr, PTR_OFFSET(ptr) + sectors);

		ca->journal.seq[ca->journal.cur_idx] = w->data->seq;
	}

	/*
	 * Make a copy of the key we're writing to for check_mark_super, since
	 * journal_next_bucket will change it
	 */
	bkey_copy(&tmp.k, &e->k);

	atomic_dec_bug(&fifo_back(&c->journal.pin));
	bch_journal_next(&c->journal);
	wake_up(&c->journal.wait);

	spin_unlock(&c->journal.lock);

	bch_check_mark_super(c, &tmp.k, true);

	while ((bio = bio_list_pop(&list)))
		closure_bio_submit_punt(bio, cl, c);

	closure_return_with_destructor(cl, journal_write_done);
}

static bool __journal_write(struct cache_set *c)
	__releases(c->journal.lock)
{
	BUG_ON(!c->journal.res_count &&
	       !test_bit(JOURNAL_DIRTY, &c->journal.flags));

	if (!c->journal.res_count &&
	    !atomic_xchg(&c->journal.in_flight, 1)) {
		closure_call(&c->journal.io, journal_write_locked,
			     NULL, &c->cl);
		return true;
	} else {
		spin_unlock(&c->journal.lock);
		return false;
	}
}

static bool journal_try_write(struct cache_set *c)
{
	set_bit(JOURNAL_NEED_WRITE, &c->journal.flags);
	return __journal_write(c);
}

static void journal_unlock(struct cache_set *c)
{
	if (test_bit(JOURNAL_NEED_WRITE, &c->journal.flags))
		__journal_write(c);
	else
		spin_unlock(&c->journal.lock);
}

static void journal_write_work(struct work_struct *work)
{
	struct cache_set *c = container_of(to_delayed_work(work),
					   struct cache_set,
					   journal.write_work);
	spin_lock(&c->journal.lock);
	if (test_bit(JOURNAL_DIRTY, &c->journal.flags))
		set_bit(JOURNAL_NEED_WRITE, &c->journal.flags);
	journal_unlock(c);
}

/*
 * This function releases the journal write structure so other threads can
 * then proceed to add their keys as well.
 */
void __bch_journal_res_put(struct cache_set *c,
			   struct journal_res *res,
			   struct closure *parent)
{
	lockdep_assert_held(&c->journal.lock);

	BUG_ON(!res->ref);

	c->journal.u64s_remaining += res->nkeys;
	--c->journal.res_count;
	res->nkeys = 0;
	res->ref = 0;

	if (parent && test_bit(JOURNAL_DIRTY, &c->journal.flags)) {
		BUG_ON(!closure_wait(&c->journal.cur->wait, parent));
		set_bit(JOURNAL_NEED_WRITE, &c->journal.flags);
	}

	journal_unlock(c);
}

static inline bool journal_bucket_has_room(struct cache_set *c)
{
	return (c->journal.sectors_free && fifo_free(&c->journal.pin) > 1);
}

static bool __journal_res_get(struct cache_set *c, struct journal_res *res,
			      unsigned nkeys_min, unsigned nkeys_max,
			      uint64_t *start_time)
{
	BUG_ON(res->ref);
	BUG_ON(nkeys_max < nkeys_min);

	spin_lock(&c->journal.lock);

	while (1) {
		/* Check if there is still room in the current journal entry */
		if (nkeys_min < c->journal.u64s_remaining) {
			res->nkeys = min_t(unsigned, nkeys_max,
					   c->journal.u64s_remaining - 1);
			res->ref = 1;

			c->journal.u64s_remaining -= res->nkeys;
			c->journal.res_count++;
			spin_unlock(&c->journal.lock);

			if (*start_time)
				bch_time_stats_update(&c->journal_full_time,
						      *start_time);

			return true;
		}

		/* local_clock() can of course be 0 but we don't care */
		if (*start_time == 0)
			*start_time = local_clock();

		/* If the current journal entry is not empty, write it */
		if (c->journal.cur->data->u64s) {
			/*
			 * If the previous journal entry is still being written, we
			 * have to wait
			 */
			if (!journal_try_write(c)) {
				trace_bcache_journal_entry_full(c);
				return false;
			}

			spin_lock(&c->journal.lock);
			continue;
		}

		/*
		 * There's room in the entry, but not enough for our reservation --
		 * since we haven't added any keys yet, we're at the end of a bucket,
		 * so skip to the end of the bucket
		 */
		if (c->journal.u64s_remaining) {
			BUG_ON(test_bit(JOURNAL_DIRTY, &c->journal.flags)
			       && !test_bit(JOURNAL_NEED_WRITE, &c->journal.flags));

			journal_entry_no_room(c);
		}

		/* If there's room in the journal bucket, start a new entry */
		if (journal_bucket_has_room(c)) {
			c->journal.u64s_remaining =
				journal_write_u64s_remaining(c);
			BUG_ON(!c->journal.u64s_remaining);
			pr_debug("done: %d", c->journal.u64s_remaining);
			continue;
		}

		/* Try to get a new journal bucket */
		journal_next_bucket(c);

		if (!journal_bucket_has_room(c)) {
			/* Still no room, we have to wait */
			spin_unlock(&c->journal.lock);
			trace_bcache_journal_full(c);
			return false;
		}
	}
}

/*
 * Essentially the entry function to the journaling code. When bcache is doing
 * a btree insert, it calls this function to get the current journal write.
 * Journal write is the structure used set up journal writes. The calling
 * function will then add its keys to the structure, queuing them for the
 * next write.
 *
 * To ensure forward progress, the current task must not be holding any
 * btree node write locks.
 */
void bch_journal_res_get(struct cache_set *c, struct journal_res *res,
			 unsigned nkeys_min, unsigned nkeys_max)
{
	uint64_t start_time = 0;

	wait_event(c->journal.wait,
		   __journal_res_get(c, res, nkeys_min, nkeys_max,
				     &start_time));
}

void bch_journal_set_dirty(struct cache_set *c)
{
	if (!test_and_set_bit(JOURNAL_DIRTY, &c->journal.flags))
		schedule_delayed_work(&c->journal.write_work,
				      msecs_to_jiffies(c->journal.delay_ms));
}

void bch_journal_add_keys(struct cache_set *c, struct journal_res *res,
			  enum btree_id id, const struct bkey *k,
			  unsigned level)
{
	unsigned actual = jset_u64s(k->u64s);

	BUG_ON(!res->ref);
	BUG_ON(actual > res->nkeys);
	res->nkeys -= actual;

	spin_lock(&c->journal.lock);
	bch_journal_add_entry(c->journal.cur->data, k, k->u64s,
			      JKEYS_BTREE_KEYS, id, level);
	bch_journal_set_dirty(c);
	spin_unlock(&c->journal.lock);
}

void bch_journal_meta(struct cache_set *c, struct closure *parent)
{
	struct journal_res res;
	unsigned u64s = jset_u64s(0);

	memset(&res, 0, sizeof(res));

	if (!CACHE_SYNC(&c->sb))
		return;

	bch_journal_res_get(c, &res, u64s, u64s);
	if (res.ref) {
		bch_journal_set_dirty(c);
		bch_journal_res_put(c, &res, parent);
	}
}

void bch_journal_free(struct cache_set *c)
{
	free_pages((unsigned long) c->journal.w[1].data, JSET_BITS);
	free_pages((unsigned long) c->journal.w[0].data, JSET_BITS);
	free_fifo(&c->journal.pin);
}

int bch_journal_alloc(struct cache_set *c)
{
	struct journal *j = &c->journal;

	spin_lock_init(&j->lock);
	init_waitqueue_head(&j->wait);
	INIT_DELAYED_WORK(&j->write_work, journal_write_work);
	INIT_WORK(&j->reclaim_work, journal_reclaim_work);

	c->journal.delay_ms = 10;

	bkey_extent_init(&j->key);

	j->w[0].c = c;
	j->w[1].c = c;

	if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)) ||
	    !(j->w[0].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)) ||
	    !(j->w[1].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)))
		return -ENOMEM;

	return 0;
}

ssize_t bch_journal_print_debug(struct journal *j, char *buf)
{
	return snprintf(buf, PAGE_SIZE,
			"active journal entries:\t%zu\n"
			"seq:\t\t\t%llu\n"
			"reservation count:\t%u\n"
			"io in flight:\t\t%i\n"
			"need write:\t\t%i\n"
			"dirty:\t\t\t%i\n"
			"replay done:\t\t%i\n",
			fifo_used(&j->pin),
			j->seq,
			j->res_count,
			atomic_read(&j->in_flight),
			test_bit(JOURNAL_NEED_WRITE,	&j->flags),
			test_bit(JOURNAL_DIRTY,		&j->flags),
			test_bit(JOURNAL_REPLAY_DONE,	&j->flags));
}

static bool bch_journal_writing_to_device(struct cache *ca)
{
	struct cache_set *c = ca->set;
	bool ret;

	spin_lock(&c->journal.lock);
	ret = bch_extent_has_device(bkey_i_to_extent(&c->journal.key),
				    ca->sb.nr_this_dev);
	spin_unlock(&c->journal.lock);

	return ret;
}

/*
 * This asumes that ca has already been marked read-only so that
 * journal_next_bucket won't pick buckets out of ca any more.
 * Hence, if the journal is not currently pointing to ca, there
 * will be no new writes to journal entries in ca after all the
 * pending ones have been flushed to disk.
 *
 * If the journal is being written to ca, write a new record, and
 * journal_next_bucket will notice that the device is no longer
 * writeable and pick a new set of devices to write to.
 */

int bch_journal_move(struct cache *ca)
{
	struct closure cl;
	unsigned i, nr_buckets;
	uint64_t last_flushed_seq;
	struct cache_set *c = ca->set;
	int ret = 0;		/* Success */

	closure_init_stack(&cl);

	if (bch_journal_writing_to_device(ca)) {
		/*
		 * bch_journal_meta will write a record and we'll wait
		 * for the write to complete.
		 * Actually writing the journal (journal_write_locked)
		 * will call journal_next_bucket which notices that the
		 * device is no longer writeable, and picks a new one.
		 */
		bch_journal_meta(c, &cl);
		/* Wait for the meta-data write */
		closure_sync(&cl);
		BUG_ON(bch_journal_writing_to_device(ca));
	}

	/*
	 * Flush all btree updates to backing store so that any
	 * journal entries written to ca become stale and are no
	 * longer needed.
	 */
	bch_btree_flush(c);

	/*
	 * Force a meta-data journal entry to be written so that
	 * we have newer journal entries in devices other than ca,
	 * and wait for the meta data write to complete.
	 */
	bch_journal_meta(c, &cl);
	closure_sync(&cl);

	/*
	 * Verify that we no longer need any of the journal entries in
	 * the device
	 */
	spin_lock(&c->journal.lock);
	last_flushed_seq = last_seq(&c->journal);
	spin_unlock(&c->journal.lock);

	nr_buckets = bch_nr_journal_buckets(&ca->sb);

	for (i = 0; i < nr_buckets; i += 1)
		BUG_ON(ca->journal.seq[i] > last_flushed_seq);

	return ret;
}