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
|
// SPDX-License-Identifier: GPL-2.0
/*
* Primary bucket allocation code
*
* Copyright 2012 Google, Inc.
*
* Allocation in bcache is done in terms of buckets:
*
* Each bucket has associated an 8 bit gen; this gen corresponds to the gen in
* btree pointers - they must match for the pointer to be considered valid.
*
* Thus (assuming a bucket has no dirty data or metadata in it) we can reuse a
* bucket simply by incrementing its gen.
*
* The gens (along with the priorities; it's really the gens are important but
* the code is named as if it's the priorities) are written in an arbitrary list
* of buckets on disk, with a pointer to them in the journal header.
*
* When we invalidate a bucket, we have to write its new gen to disk and wait
* for that write to complete before we use it - otherwise after a crash we
* could have pointers that appeared to be good but pointed to data that had
* been overwritten.
*
* Since the gens and priorities are all stored contiguously on disk, we can
* batch this up: We fill up the free_inc list with freshly invalidated buckets,
* call prio_write(), and when prio_write() finishes we pull buckets off the
* free_inc list and optionally discard them.
*
* free_inc isn't the only freelist - if it was, we'd often have to sleep while
* priorities and gens were being written before we could allocate. c->free is a
* smaller freelist, and buckets on that list are always ready to be used.
*
* If we've got discards enabled, that happens when a bucket moves from the
* free_inc list to the free list.
*
* It's important to ensure that gens don't wrap around - with respect to
* either the oldest gen in the btree or the gen on disk. This is quite
* difficult to do in practice, but we explicitly guard against it anyways - if
* a bucket is in danger of wrapping around we simply skip invalidating it that
* time around, and we garbage collect or rewrite the priorities sooner than we
* would have otherwise.
*
* bch2_bucket_alloc() allocates a single bucket from a specific device.
*
* bch2_bucket_alloc_set() allocates one or more buckets from different devices
* in a given filesystem.
*
* invalidate_buckets() drives all the processes described above. It's called
* from bch2_bucket_alloc() and a few other places that need to make sure free
* buckets are ready.
*
* invalidate_buckets_(lru|fifo)() find buckets that are available to be
* invalidated, and then invalidate them and stick them on the free_inc list -
* in either lru or fifo order.
*/
#include "bcachefs.h"
#include "alloc_background.h"
#include "alloc_foreground.h"
#include "btree_gc.h"
#include "buckets.h"
#include "clock.h"
#include "debug.h"
#include "disk_groups.h"
#include "ec.h"
#include "io.h"
#include "trace.h"
#include <linux/math64.h>
#include <linux/rculist.h>
#include <linux/rcupdate.h>
enum bucket_alloc_ret {
ALLOC_SUCCESS,
OPEN_BUCKETS_EMPTY,
FREELIST_EMPTY, /* Allocator thread not keeping up */
};
/*
* Open buckets represent a bucket that's currently being allocated from. They
* serve two purposes:
*
* - They track buckets that have been partially allocated, allowing for
* sub-bucket sized allocations - they're used by the sector allocator below
*
* - They provide a reference to the buckets they own that mark and sweep GC
* can find, until the new allocation has a pointer to it inserted into the
* btree
*
* When allocating some space with the sector allocator, the allocation comes
* with a reference to an open bucket - the caller is required to put that
* reference _after_ doing the index update that makes its allocation reachable.
*/
void __bch2_open_bucket_put(struct bch_fs *c, struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->ptr.dev);
if (ob->ec) {
bch2_ec_bucket_written(c, ob);
return;
}
percpu_down_read(&c->usage_lock);
spin_lock(&ob->lock);
bch2_mark_alloc_bucket(c, ca, PTR_BUCKET_NR(ca, &ob->ptr),
false, gc_pos_alloc(c, ob), 0);
ob->valid = false;
spin_unlock(&ob->lock);
percpu_up_read(&c->usage_lock);
spin_lock(&c->freelist_lock);
ob->freelist = c->open_buckets_freelist;
c->open_buckets_freelist = ob - c->open_buckets;
c->open_buckets_nr_free++;
spin_unlock(&c->freelist_lock);
closure_wake_up(&c->open_buckets_wait);
}
void bch2_open_bucket_write_error(struct bch_fs *c,
struct open_buckets *obs,
unsigned dev)
{
struct open_bucket *ob;
unsigned i;
open_bucket_for_each(c, obs, ob, i)
if (ob->ptr.dev == dev &&
ob->ec)
bch2_ec_bucket_cancel(c, ob);
}
static struct open_bucket *bch2_open_bucket_alloc(struct bch_fs *c)
{
struct open_bucket *ob;
BUG_ON(!c->open_buckets_freelist || !c->open_buckets_nr_free);
ob = c->open_buckets + c->open_buckets_freelist;
c->open_buckets_freelist = ob->freelist;
atomic_set(&ob->pin, 1);
c->open_buckets_nr_free--;
return ob;
}
static void open_bucket_free_unused(struct bch_fs *c,
struct open_bucket *ob,
bool may_realloc)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->ptr.dev);
BUG_ON(ca->open_buckets_partial_nr >=
ARRAY_SIZE(ca->open_buckets_partial));
if (ca->open_buckets_partial_nr <
ARRAY_SIZE(ca->open_buckets_partial) &&
may_realloc) {
spin_lock(&c->freelist_lock);
ob->on_partial_list = true;
ca->open_buckets_partial[ca->open_buckets_partial_nr++] =
ob - c->open_buckets;
spin_unlock(&c->freelist_lock);
closure_wake_up(&c->open_buckets_wait);
closure_wake_up(&c->freelist_wait);
} else {
bch2_open_bucket_put(c, ob);
}
}
static void verify_not_stale(struct bch_fs *c, const struct open_buckets *obs)
{
#ifdef CONFIG_BCACHEFS_DEBUG
struct open_bucket *ob;
unsigned i;
open_bucket_for_each(c, obs, ob, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->ptr.dev);
BUG_ON(ptr_stale(ca, &ob->ptr));
}
#endif
}
/* _only_ for allocating the journal on a new device: */
long bch2_bucket_alloc_new_fs(struct bch_dev *ca)
{
struct bucket_array *buckets;
ssize_t b;
rcu_read_lock();
buckets = bucket_array(ca);
for (b = ca->mi.first_bucket; b < ca->mi.nbuckets; b++)
if (is_available_bucket(buckets->b[b].mark))
goto success;
b = -1;
success:
rcu_read_unlock();
return b;
}
static inline unsigned open_buckets_reserved(enum alloc_reserve reserve)
{
switch (reserve) {
case RESERVE_ALLOC:
return 0;
case RESERVE_BTREE:
return BTREE_NODE_RESERVE / 2;
default:
return BTREE_NODE_RESERVE;
}
}
/**
* bch_bucket_alloc - allocate a single bucket from a specific device
*
* Returns index of bucket on success, 0 on failure
* */
struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
enum alloc_reserve reserve,
bool may_alloc_partial,
struct closure *cl)
{
struct bucket_array *buckets;
struct open_bucket *ob;
long bucket = 0;
spin_lock(&c->freelist_lock);
if (may_alloc_partial &&
ca->open_buckets_partial_nr) {
ob = c->open_buckets +
ca->open_buckets_partial[--ca->open_buckets_partial_nr];
ob->on_partial_list = false;
spin_unlock(&c->freelist_lock);
return ob;
}
if (unlikely(c->open_buckets_nr_free <= open_buckets_reserved(reserve))) {
if (cl)
closure_wait(&c->open_buckets_wait, cl);
spin_unlock(&c->freelist_lock);
trace_open_bucket_alloc_fail(ca, reserve);
return ERR_PTR(-OPEN_BUCKETS_EMPTY);
}
if (likely(fifo_pop(&ca->free[RESERVE_NONE], bucket)))
goto out;
switch (reserve) {
case RESERVE_ALLOC:
if (fifo_pop(&ca->free[RESERVE_BTREE], bucket))
goto out;
break;
case RESERVE_BTREE:
if (fifo_used(&ca->free[RESERVE_BTREE]) * 2 >=
ca->free[RESERVE_BTREE].size &&
fifo_pop(&ca->free[RESERVE_BTREE], bucket))
goto out;
break;
case RESERVE_MOVINGGC:
if (fifo_pop(&ca->free[RESERVE_MOVINGGC], bucket))
goto out;
break;
default:
break;
}
if (cl)
closure_wait(&c->freelist_wait, cl);
spin_unlock(&c->freelist_lock);
trace_bucket_alloc_fail(ca, reserve);
return ERR_PTR(-FREELIST_EMPTY);
out:
verify_not_on_freelist(c, ca, bucket);
ob = bch2_open_bucket_alloc(c);
spin_lock(&ob->lock);
buckets = bucket_array(ca);
ob->valid = true;
ob->sectors_free = ca->mi.bucket_size;
ob->ptr = (struct bch_extent_ptr) {
.gen = buckets->b[bucket].mark.gen,
.offset = bucket_to_sector(ca, bucket),
.dev = ca->dev_idx,
};
bucket_io_clock_reset(c, ca, bucket, READ);
bucket_io_clock_reset(c, ca, bucket, WRITE);
spin_unlock(&ob->lock);
spin_unlock(&c->freelist_lock);
bch2_wake_allocator(ca);
trace_bucket_alloc(ca, reserve);
return ob;
}
static int __dev_stripe_cmp(struct dev_stripe_state *stripe,
unsigned l, unsigned r)
{
return ((stripe->next_alloc[l] > stripe->next_alloc[r]) -
(stripe->next_alloc[l] < stripe->next_alloc[r]));
}
#define dev_stripe_cmp(l, r) __dev_stripe_cmp(stripe, l, r)
struct dev_alloc_list bch2_dev_alloc_list(struct bch_fs *c,
struct dev_stripe_state *stripe,
struct bch_devs_mask *devs)
{
struct dev_alloc_list ret = { .nr = 0 };
struct bch_dev *ca;
unsigned i;
for_each_member_device_rcu(ca, c, i, devs)
ret.devs[ret.nr++] = i;
bubble_sort(ret.devs, ret.nr, dev_stripe_cmp);
return ret;
}
void bch2_dev_stripe_increment(struct bch_fs *c, struct bch_dev *ca,
struct dev_stripe_state *stripe)
{
u64 *v = stripe->next_alloc + ca->dev_idx;
u64 free_space = dev_buckets_free(c, ca);
u64 free_space_inv = free_space
? div64_u64(1ULL << 48, free_space)
: 1ULL << 48;
u64 scale = *v / 4;
if (*v + free_space_inv >= *v)
*v += free_space_inv;
else
*v = U64_MAX;
for (v = stripe->next_alloc;
v < stripe->next_alloc + ARRAY_SIZE(stripe->next_alloc); v++)
*v = *v < scale ? 0 : *v - scale;
}
#define BUCKET_MAY_ALLOC_PARTIAL (1 << 0)
#define BUCKET_ALLOC_USE_DURABILITY (1 << 1)
static int bch2_bucket_alloc_set(struct bch_fs *c,
struct open_buckets *ptrs,
struct dev_stripe_state *stripe,
struct bch_devs_mask *devs_may_alloc,
unsigned nr_replicas,
unsigned *nr_effective,
bool *have_cache,
enum alloc_reserve reserve,
unsigned flags,
struct closure *cl)
{
struct dev_alloc_list devs_sorted =
bch2_dev_alloc_list(c, stripe, devs_may_alloc);
struct bch_dev *ca;
bool alloc_failure = false;
unsigned i, durability;
BUG_ON(*nr_effective >= nr_replicas);
for (i = 0; i < devs_sorted.nr; i++) {
struct open_bucket *ob;
ca = rcu_dereference(c->devs[devs_sorted.devs[i]]);
if (!ca)
continue;
if (!ca->mi.durability && *have_cache)
continue;
ob = bch2_bucket_alloc(c, ca, reserve,
flags & BUCKET_MAY_ALLOC_PARTIAL, cl);
if (IS_ERR(ob)) {
enum bucket_alloc_ret ret = -PTR_ERR(ob);
WARN_ON(reserve == RESERVE_MOVINGGC &&
ret != OPEN_BUCKETS_EMPTY);
if (cl)
return -EAGAIN;
if (ret == OPEN_BUCKETS_EMPTY)
return -ENOSPC;
alloc_failure = true;
continue;
}
durability = (flags & BUCKET_ALLOC_USE_DURABILITY)
? ca->mi.durability : 1;
__clear_bit(ca->dev_idx, devs_may_alloc->d);
*nr_effective += durability;
*have_cache |= !durability;
ob_push(c, ptrs, ob);
bch2_dev_stripe_increment(c, ca, stripe);
if (*nr_effective >= nr_replicas)
return 0;
}
return alloc_failure ? -ENOSPC : -EROFS;
}
/* Allocate from stripes: */
/*
* XXX: use a higher watermark for allocating open buckets here:
*/
static int ec_stripe_alloc(struct bch_fs *c, struct ec_stripe_head *h)
{
struct bch_devs_mask devs;
struct open_bucket *ob;
unsigned i, nr_have = 0, nr_data =
min_t(unsigned, h->nr_active_devs,
EC_STRIPE_MAX) - h->redundancy;
bool have_cache = true;
int ret = 0;
BUG_ON(h->blocks.nr > nr_data);
BUG_ON(h->parity.nr > h->redundancy);
devs = h->devs;
open_bucket_for_each(c, &h->parity, ob, i)
__clear_bit(ob->ptr.dev, devs.d);
open_bucket_for_each(c, &h->blocks, ob, i)
__clear_bit(ob->ptr.dev, devs.d);
percpu_down_read(&c->usage_lock);
rcu_read_lock();
if (h->parity.nr < h->redundancy) {
nr_have = h->parity.nr;
ret = bch2_bucket_alloc_set(c, &h->parity,
&h->parity_stripe,
&devs,
h->redundancy,
&nr_have,
&have_cache,
RESERVE_NONE,
0,
NULL);
if (ret)
goto err;
}
if (h->blocks.nr < nr_data) {
nr_have = h->blocks.nr;
ret = bch2_bucket_alloc_set(c, &h->blocks,
&h->block_stripe,
&devs,
nr_data,
&nr_have,
&have_cache,
RESERVE_NONE,
0,
NULL);
if (ret)
goto err;
}
rcu_read_unlock();
percpu_up_read(&c->usage_lock);
return bch2_ec_stripe_new_alloc(c, h);
err:
rcu_read_unlock();
percpu_up_read(&c->usage_lock);
return -1;
}
/*
* if we can't allocate a new stripe because there are already too many
* partially filled stripes, force allocating from an existing stripe even when
* it's to a device we don't want:
*/
static void bucket_alloc_from_stripe(struct bch_fs *c,
struct open_buckets *ptrs,
struct write_point *wp,
struct bch_devs_mask *devs_may_alloc,
u16 target,
unsigned erasure_code,
unsigned nr_replicas,
unsigned *nr_effective,
bool *have_cache)
{
struct dev_alloc_list devs_sorted;
struct ec_stripe_head *h;
struct open_bucket *ob;
struct bch_dev *ca;
unsigned i, ec_idx;
if (!erasure_code)
return;
if (nr_replicas < 2)
return;
if (ec_open_bucket(c, ptrs))
return;
h = bch2_ec_stripe_head_get(c, target, erasure_code, nr_replicas - 1);
if (!h)
return;
if (!h->s && ec_stripe_alloc(c, h))
goto out_put_head;
rcu_read_lock();
devs_sorted = bch2_dev_alloc_list(c, &wp->stripe, devs_may_alloc);
rcu_read_unlock();
for (i = 0; i < devs_sorted.nr; i++)
open_bucket_for_each(c, &h->s->blocks, ob, ec_idx)
if (ob->ptr.dev == devs_sorted.devs[i] &&
!test_and_set_bit(ec_idx, h->s->blocks_allocated))
goto got_bucket;
goto out_put_head;
got_bucket:
ca = bch_dev_bkey_exists(c, ob->ptr.dev);
ob->ec_idx = ec_idx;
ob->ec = h->s;
__clear_bit(ob->ptr.dev, devs_may_alloc->d);
*nr_effective += ca->mi.durability;
*have_cache |= !ca->mi.durability;
ob_push(c, ptrs, ob);
atomic_inc(&h->s->pin);
out_put_head:
bch2_ec_stripe_head_put(h);
}
/* Sector allocator */
static void get_buckets_from_writepoint(struct bch_fs *c,
struct open_buckets *ptrs,
struct write_point *wp,
struct bch_devs_mask *devs_may_alloc,
unsigned nr_replicas,
unsigned *nr_effective,
bool *have_cache,
bool need_ec)
{
struct open_buckets ptrs_skip = { .nr = 0 };
struct open_bucket *ob;
unsigned i;
open_bucket_for_each(c, &wp->ptrs, ob, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->ptr.dev);
if (*nr_effective < nr_replicas &&
test_bit(ob->ptr.dev, devs_may_alloc->d) &&
(ca->mi.durability ||
(wp->type == BCH_DATA_USER && !*have_cache)) &&
(ob->ec || !need_ec)) {
__clear_bit(ob->ptr.dev, devs_may_alloc->d);
*nr_effective += ca->mi.durability;
*have_cache |= !ca->mi.durability;
ob_push(c, ptrs, ob);
} else {
ob_push(c, &ptrs_skip, ob);
}
}
wp->ptrs = ptrs_skip;
}
static int open_bucket_add_buckets(struct bch_fs *c,
struct open_buckets *ptrs,
struct write_point *wp,
struct bch_devs_list *devs_have,
u16 target,
unsigned erasure_code,
unsigned nr_replicas,
unsigned *nr_effective,
bool *have_cache,
enum alloc_reserve reserve,
struct closure *_cl)
{
struct bch_devs_mask devs;
struct open_bucket *ob;
struct closure *cl = NULL;
unsigned i, flags = BUCKET_ALLOC_USE_DURABILITY;
int ret;
if (wp->type == BCH_DATA_USER)
flags |= BUCKET_MAY_ALLOC_PARTIAL;
rcu_read_lock();
devs = target_rw_devs(c, wp->type, target);
rcu_read_unlock();
/* Don't allocate from devices we already have pointers to: */
for (i = 0; i < devs_have->nr; i++)
__clear_bit(devs_have->devs[i], devs.d);
open_bucket_for_each(c, ptrs, ob, i)
__clear_bit(ob->ptr.dev, devs.d);
if (erasure_code) {
get_buckets_from_writepoint(c, ptrs, wp, &devs,
nr_replicas, nr_effective,
have_cache, true);
if (*nr_effective >= nr_replicas)
return 0;
bucket_alloc_from_stripe(c, ptrs, wp, &devs,
target, erasure_code,
nr_replicas, nr_effective,
have_cache);
if (*nr_effective >= nr_replicas)
return 0;
}
get_buckets_from_writepoint(c, ptrs, wp, &devs,
nr_replicas, nr_effective,
have_cache, false);
if (*nr_effective >= nr_replicas)
return 0;
percpu_down_read(&c->usage_lock);
rcu_read_lock();
retry_blocking:
/*
* Try nonblocking first, so that if one device is full we'll try from
* other devices:
*/
ret = bch2_bucket_alloc_set(c, ptrs, &wp->stripe, &devs,
nr_replicas, nr_effective, have_cache,
reserve, flags, cl);
if (ret && ret != -EROFS && !cl && _cl) {
cl = _cl;
goto retry_blocking;
}
rcu_read_unlock();
percpu_up_read(&c->usage_lock);
return ret;
}
void bch2_open_buckets_stop_dev(struct bch_fs *c, struct bch_dev *ca,
struct open_buckets *obs,
enum bch_data_type data_type)
{
struct open_buckets ptrs = { .nr = 0 };
struct open_bucket *ob, *ob2;
unsigned i, j;
open_bucket_for_each(c, obs, ob, i) {
bool drop = !ca || ob->ptr.dev == ca->dev_idx;
if (!drop && ob->ec) {
mutex_lock(&ob->ec->lock);
open_bucket_for_each(c, &ob->ec->blocks, ob2, j)
drop |= ob2->ptr.dev == ca->dev_idx;
open_bucket_for_each(c, &ob->ec->parity, ob2, j)
drop |= ob2->ptr.dev == ca->dev_idx;
mutex_unlock(&ob->ec->lock);
}
if (drop)
bch2_open_bucket_put(c, ob);
else
ob_push(c, &ptrs, ob);
}
*obs = ptrs;
}
void bch2_writepoint_stop(struct bch_fs *c, struct bch_dev *ca,
struct write_point *wp)
{
mutex_lock(&wp->lock);
bch2_open_buckets_stop_dev(c, ca, &wp->ptrs, wp->type);
mutex_unlock(&wp->lock);
}
static inline struct hlist_head *writepoint_hash(struct bch_fs *c,
unsigned long write_point)
{
unsigned hash =
hash_long(write_point, ilog2(ARRAY_SIZE(c->write_points_hash)));
return &c->write_points_hash[hash];
}
static struct write_point *__writepoint_find(struct hlist_head *head,
unsigned long write_point)
{
struct write_point *wp;
hlist_for_each_entry_rcu(wp, head, node)
if (wp->write_point == write_point)
return wp;
return NULL;
}
static inline bool too_many_writepoints(struct bch_fs *c, unsigned factor)
{
u64 stranded = c->write_points_nr * c->bucket_size_max;
u64 free = bch2_fs_sectors_free(c, bch2_fs_usage_read(c));
return stranded * factor > free;
}
static bool try_increase_writepoints(struct bch_fs *c)
{
struct write_point *wp;
if (c->write_points_nr == ARRAY_SIZE(c->write_points) ||
too_many_writepoints(c, 32))
return false;
wp = c->write_points + c->write_points_nr++;
hlist_add_head_rcu(&wp->node, writepoint_hash(c, wp->write_point));
return true;
}
static bool try_decrease_writepoints(struct bch_fs *c,
unsigned old_nr)
{
struct write_point *wp;
mutex_lock(&c->write_points_hash_lock);
if (c->write_points_nr < old_nr) {
mutex_unlock(&c->write_points_hash_lock);
return true;
}
if (c->write_points_nr == 1 ||
!too_many_writepoints(c, 8)) {
mutex_unlock(&c->write_points_hash_lock);
return false;
}
wp = c->write_points + --c->write_points_nr;
hlist_del_rcu(&wp->node);
mutex_unlock(&c->write_points_hash_lock);
bch2_writepoint_stop(c, NULL, wp);
return true;
}
static struct write_point *writepoint_find(struct bch_fs *c,
unsigned long write_point)
{
struct write_point *wp, *oldest;
struct hlist_head *head;
if (!(write_point & 1UL)) {
wp = (struct write_point *) write_point;
mutex_lock(&wp->lock);
return wp;
}
head = writepoint_hash(c, write_point);
restart_find:
wp = __writepoint_find(head, write_point);
if (wp) {
lock_wp:
mutex_lock(&wp->lock);
if (wp->write_point == write_point)
goto out;
mutex_unlock(&wp->lock);
goto restart_find;
}
restart_find_oldest:
oldest = NULL;
for (wp = c->write_points;
wp < c->write_points + c->write_points_nr; wp++)
if (!oldest || time_before64(wp->last_used, oldest->last_used))
oldest = wp;
mutex_lock(&oldest->lock);
mutex_lock(&c->write_points_hash_lock);
if (oldest >= c->write_points + c->write_points_nr ||
try_increase_writepoints(c)) {
mutex_unlock(&c->write_points_hash_lock);
mutex_unlock(&oldest->lock);
goto restart_find_oldest;
}
wp = __writepoint_find(head, write_point);
if (wp && wp != oldest) {
mutex_unlock(&c->write_points_hash_lock);
mutex_unlock(&oldest->lock);
goto lock_wp;
}
wp = oldest;
hlist_del_rcu(&wp->node);
wp->write_point = write_point;
hlist_add_head_rcu(&wp->node, head);
mutex_unlock(&c->write_points_hash_lock);
out:
wp->last_used = sched_clock();
return wp;
}
/*
* Get us an open_bucket we can allocate from, return with it locked:
*/
struct write_point *bch2_alloc_sectors_start(struct bch_fs *c,
unsigned target,
unsigned erasure_code,
struct write_point_specifier write_point,
struct bch_devs_list *devs_have,
unsigned nr_replicas,
unsigned nr_replicas_required,
enum alloc_reserve reserve,
unsigned flags,
struct closure *cl)
{
struct write_point *wp;
struct open_bucket *ob;
unsigned nr_effective = 0;
struct open_buckets ptrs = { .nr = 0 };
bool have_cache = false;
unsigned write_points_nr;
int ret = 0, i;
BUG_ON(!nr_replicas || !nr_replicas_required);
retry:
write_points_nr = c->write_points_nr;
wp = writepoint_find(c, write_point.v);
/* metadata may not allocate on cache devices: */
if (wp->type != BCH_DATA_USER)
have_cache = true;
if (!target || (flags & BCH_WRITE_ONLY_SPECIFIED_DEVS)) {
ret = open_bucket_add_buckets(c, &ptrs, wp, devs_have,
target, erasure_code,
nr_replicas, &nr_effective,
&have_cache, reserve, cl);
} else {
ret = open_bucket_add_buckets(c, &ptrs, wp, devs_have,
target, erasure_code,
nr_replicas, &nr_effective,
&have_cache, reserve, NULL);
if (!ret)
goto alloc_done;
ret = open_bucket_add_buckets(c, &ptrs, wp, devs_have,
0, erasure_code,
nr_replicas, &nr_effective,
&have_cache, reserve, cl);
}
alloc_done:
BUG_ON(!ret && nr_effective < nr_replicas);
if (erasure_code && !ec_open_bucket(c, &ptrs))
pr_debug("failed to get ec bucket: ret %u", ret);
if (ret == -EROFS &&
nr_effective >= nr_replicas_required)
ret = 0;
if (ret)
goto err;
/* Free buckets we didn't use: */
open_bucket_for_each(c, &wp->ptrs, ob, i)
open_bucket_free_unused(c, ob, wp->type == BCH_DATA_USER);
wp->ptrs = ptrs;
wp->sectors_free = UINT_MAX;
open_bucket_for_each(c, &wp->ptrs, ob, i)
wp->sectors_free = min(wp->sectors_free, ob->sectors_free);
BUG_ON(!wp->sectors_free || wp->sectors_free == UINT_MAX);
verify_not_stale(c, &wp->ptrs);
return wp;
err:
open_bucket_for_each(c, &wp->ptrs, ob, i)
if (ptrs.nr < ARRAY_SIZE(ptrs.v))
ob_push(c, &ptrs, ob);
else
open_bucket_free_unused(c, ob,
wp->type == BCH_DATA_USER);
wp->ptrs = ptrs;
mutex_unlock(&wp->lock);
if (ret == -ENOSPC &&
try_decrease_writepoints(c, write_points_nr))
goto retry;
return ERR_PTR(ret);
}
/*
* Append pointers to the space we just allocated to @k, and mark @sectors space
* as allocated out of @ob
*/
void bch2_alloc_sectors_append_ptrs(struct bch_fs *c, struct write_point *wp,
struct bkey_i_extent *e, unsigned sectors)
{
struct open_bucket *ob;
unsigned i;
BUG_ON(sectors > wp->sectors_free);
wp->sectors_free -= sectors;
open_bucket_for_each(c, &wp->ptrs, ob, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->ptr.dev);
struct bch_extent_ptr tmp = ob->ptr;
EBUG_ON(bch2_extent_has_device(extent_i_to_s_c(e), ob->ptr.dev));
tmp.cached = bkey_extent_is_cached(&e->k) ||
(!ca->mi.durability && wp->type == BCH_DATA_USER);
tmp.offset += ca->mi.bucket_size - ob->sectors_free;
extent_ptr_append(e, tmp);
BUG_ON(sectors > ob->sectors_free);
ob->sectors_free -= sectors;
}
}
/*
* Append pointers to the space we just allocated to @k, and mark @sectors space
* as allocated out of @ob
*/
void bch2_alloc_sectors_done(struct bch_fs *c, struct write_point *wp)
{
struct open_buckets ptrs = { .nr = 0 }, keep = { .nr = 0 };
struct open_bucket *ob;
unsigned i;
open_bucket_for_each(c, &wp->ptrs, ob, i)
ob_push(c, !ob->sectors_free ? &ptrs : &keep, ob);
wp->ptrs = keep;
mutex_unlock(&wp->lock);
bch2_open_buckets_put(c, &ptrs);
}
void bch2_fs_allocator_foreground_init(struct bch_fs *c)
{
struct open_bucket *ob;
struct write_point *wp;
mutex_init(&c->write_points_hash_lock);
c->write_points_nr = ARRAY_SIZE(c->write_points);
/* open bucket 0 is a sentinal NULL: */
spin_lock_init(&c->open_buckets[0].lock);
for (ob = c->open_buckets + 1;
ob < c->open_buckets + ARRAY_SIZE(c->open_buckets); ob++) {
spin_lock_init(&ob->lock);
c->open_buckets_nr_free++;
ob->freelist = c->open_buckets_freelist;
c->open_buckets_freelist = ob - c->open_buckets;
}
writepoint_init(&c->btree_write_point, BCH_DATA_BTREE);
writepoint_init(&c->rebalance_write_point, BCH_DATA_USER);
for (wp = c->write_points;
wp < c->write_points + c->write_points_nr; wp++) {
writepoint_init(wp, BCH_DATA_USER);
wp->last_used = sched_clock();
wp->write_point = (unsigned long) wp;
hlist_add_head_rcu(&wp->node,
writepoint_hash(c, wp->write_point));
}
}
|