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
|
// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "bkey_buf.h"
#include "bset.h"
#include "btree_cache.h"
#include "btree_journal_iter.h"
#include "journal_io.h"
#include <linux/sort.h>
/*
* For managing keys we read from the journal: until journal replay works normal
* btree lookups need to be able to find and return keys from the journal where
* they overwrite what's in the btree, so we have a special iterator and
* operations for the regular btree iter code to use:
*/
static int __journal_key_cmp(enum btree_id l_btree_id,
unsigned l_level,
struct bpos l_pos,
const struct journal_key *r)
{
return (cmp_int(l_btree_id, r->btree_id) ?:
cmp_int(l_level, r->level) ?:
bpos_cmp(l_pos, r->k->k.p));
}
static int journal_key_cmp(const struct journal_key *l, const struct journal_key *r)
{
return __journal_key_cmp(l->btree_id, l->level, l->k->k.p, r);
}
static inline size_t idx_to_pos(struct journal_keys *keys, size_t idx)
{
size_t gap_size = keys->size - keys->nr;
if (idx >= keys->gap)
idx += gap_size;
return idx;
}
static inline struct journal_key *idx_to_key(struct journal_keys *keys, size_t idx)
{
return keys->d + idx_to_pos(keys, idx);
}
static size_t __bch2_journal_key_search(struct journal_keys *keys,
enum btree_id id, unsigned level,
struct bpos pos)
{
size_t l = 0, r = keys->nr, m;
while (l < r) {
m = l + ((r - l) >> 1);
if (__journal_key_cmp(id, level, pos, idx_to_key(keys, m)) > 0)
l = m + 1;
else
r = m;
}
BUG_ON(l < keys->nr &&
__journal_key_cmp(id, level, pos, idx_to_key(keys, l)) > 0);
BUG_ON(l &&
__journal_key_cmp(id, level, pos, idx_to_key(keys, l - 1)) <= 0);
return l;
}
static size_t bch2_journal_key_search(struct journal_keys *keys,
enum btree_id id, unsigned level,
struct bpos pos)
{
return idx_to_pos(keys, __bch2_journal_key_search(keys, id, level, pos));
}
/* Returns first non-overwritten key >= search key: */
struct bkey_i *bch2_journal_keys_peek_upto(struct bch_fs *c, enum btree_id btree_id,
unsigned level, struct bpos pos,
struct bpos end_pos, size_t *idx)
{
struct journal_keys *keys = &c->journal_keys;
unsigned iters = 0;
struct journal_key *k;
BUG_ON(*idx > keys->nr);
search:
if (!*idx)
*idx = __bch2_journal_key_search(keys, btree_id, level, pos);
while (*idx &&
__journal_key_cmp(btree_id, level, end_pos, idx_to_key(keys, *idx - 1)) <= 0) {
--(*idx);
iters++;
if (iters == 10) {
*idx = 0;
goto search;
}
}
while ((k = *idx < keys->nr ? idx_to_key(keys, *idx) : NULL)) {
if (__journal_key_cmp(btree_id, level, end_pos, k) < 0)
return NULL;
if (k->overwritten) {
(*idx)++;
continue;
}
if (__journal_key_cmp(btree_id, level, pos, k) <= 0)
return k->k;
(*idx)++;
iters++;
if (iters == 10) {
*idx = 0;
goto search;
}
}
return NULL;
}
struct bkey_i *bch2_journal_keys_peek_slot(struct bch_fs *c, enum btree_id btree_id,
unsigned level, struct bpos pos)
{
size_t idx = 0;
return bch2_journal_keys_peek_upto(c, btree_id, level, pos, pos, &idx);
}
static void journal_iters_fix(struct bch_fs *c)
{
struct journal_keys *keys = &c->journal_keys;
/* The key we just inserted is immediately before the gap: */
size_t gap_end = keys->gap + (keys->size - keys->nr);
struct btree_and_journal_iter *iter;
/*
* If an iterator points one after the key we just inserted, decrement
* the iterator so it points at the key we just inserted - if the
* decrement was unnecessary, bch2_btree_and_journal_iter_peek() will
* handle that:
*/
list_for_each_entry(iter, &c->journal_iters, journal.list)
if (iter->journal.idx == gap_end)
iter->journal.idx = keys->gap - 1;
}
static void journal_iters_move_gap(struct bch_fs *c, size_t old_gap, size_t new_gap)
{
struct journal_keys *keys = &c->journal_keys;
struct journal_iter *iter;
size_t gap_size = keys->size - keys->nr;
list_for_each_entry(iter, &c->journal_iters, list) {
if (iter->idx > old_gap)
iter->idx -= gap_size;
if (iter->idx >= new_gap)
iter->idx += gap_size;
}
}
int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
unsigned level, struct bkey_i *k)
{
struct journal_key n = {
.btree_id = id,
.level = level,
.k = k,
.allocated = true,
/*
* Ensure these keys are done last by journal replay, to unblock
* journal reclaim:
*/
.journal_seq = U32_MAX,
};
struct journal_keys *keys = &c->journal_keys;
size_t idx = bch2_journal_key_search(keys, id, level, k->k.p);
BUG_ON(test_bit(BCH_FS_rw, &c->flags));
if (idx < keys->size &&
journal_key_cmp(&n, &keys->d[idx]) == 0) {
if (keys->d[idx].allocated)
kfree(keys->d[idx].k);
keys->d[idx] = n;
return 0;
}
if (idx > keys->gap)
idx -= keys->size - keys->nr;
if (keys->nr == keys->size) {
struct journal_keys new_keys = {
.nr = keys->nr,
.size = max_t(size_t, keys->size, 8) * 2,
};
new_keys.d = kvmalloc_array(new_keys.size, sizeof(new_keys.d[0]), GFP_KERNEL);
if (!new_keys.d) {
bch_err(c, "%s: error allocating new key array (size %zu)",
__func__, new_keys.size);
return -BCH_ERR_ENOMEM_journal_key_insert;
}
/* Since @keys was full, there was no gap: */
memcpy(new_keys.d, keys->d, sizeof(keys->d[0]) * keys->nr);
kvfree(keys->d);
keys->d = new_keys.d;
keys->nr = new_keys.nr;
keys->size = new_keys.size;
/* And now the gap is at the end: */
keys->gap = keys->nr;
}
journal_iters_move_gap(c, keys->gap, idx);
move_gap(keys->d, keys->nr, keys->size, keys->gap, idx);
keys->gap = idx;
keys->nr++;
keys->d[keys->gap++] = n;
journal_iters_fix(c);
return 0;
}
/*
* Can only be used from the recovery thread while we're still RO - can't be
* used once we've got RW, as journal_keys is at that point used by multiple
* threads:
*/
int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id,
unsigned level, struct bkey_i *k)
{
struct bkey_i *n;
int ret;
n = kmalloc(bkey_bytes(&k->k), GFP_KERNEL);
if (!n)
return -BCH_ERR_ENOMEM_journal_key_insert;
bkey_copy(n, k);
ret = bch2_journal_key_insert_take(c, id, level, n);
if (ret)
kfree(n);
return ret;
}
int bch2_journal_key_delete(struct bch_fs *c, enum btree_id id,
unsigned level, struct bpos pos)
{
struct bkey_i whiteout;
bkey_init(&whiteout.k);
whiteout.k.p = pos;
return bch2_journal_key_insert(c, id, level, &whiteout);
}
void bch2_journal_key_overwritten(struct bch_fs *c, enum btree_id btree,
unsigned level, struct bpos pos)
{
struct journal_keys *keys = &c->journal_keys;
size_t idx = bch2_journal_key_search(keys, btree, level, pos);
if (idx < keys->size &&
keys->d[idx].btree_id == btree &&
keys->d[idx].level == level &&
bpos_eq(keys->d[idx].k->k.p, pos))
keys->d[idx].overwritten = true;
}
static void bch2_journal_iter_advance(struct journal_iter *iter)
{
if (iter->idx < iter->keys->size) {
iter->idx++;
if (iter->idx == iter->keys->gap)
iter->idx += iter->keys->size - iter->keys->nr;
}
}
static struct bkey_s_c bch2_journal_iter_peek(struct journal_iter *iter)
{
struct journal_key *k = iter->keys->d + iter->idx;
while (k < iter->keys->d + iter->keys->size &&
k->btree_id == iter->btree_id &&
k->level == iter->level) {
if (!k->overwritten)
return bkey_i_to_s_c(k->k);
bch2_journal_iter_advance(iter);
k = iter->keys->d + iter->idx;
}
return bkey_s_c_null;
}
static void bch2_journal_iter_exit(struct journal_iter *iter)
{
list_del(&iter->list);
}
static void bch2_journal_iter_init(struct bch_fs *c,
struct journal_iter *iter,
enum btree_id id, unsigned level,
struct bpos pos)
{
iter->btree_id = id;
iter->level = level;
iter->keys = &c->journal_keys;
iter->idx = bch2_journal_key_search(&c->journal_keys, id, level, pos);
}
static struct bkey_s_c bch2_journal_iter_peek_btree(struct btree_and_journal_iter *iter)
{
return bch2_btree_node_iter_peek_unpack(&iter->node_iter,
iter->b, &iter->unpacked);
}
static void bch2_journal_iter_advance_btree(struct btree_and_journal_iter *iter)
{
bch2_btree_node_iter_advance(&iter->node_iter, iter->b);
}
void bch2_btree_and_journal_iter_advance(struct btree_and_journal_iter *iter)
{
if (bpos_eq(iter->pos, SPOS_MAX))
iter->at_end = true;
else
iter->pos = bpos_successor(iter->pos);
}
static void btree_and_journal_iter_prefetch(struct btree_and_journal_iter *_iter)
{
struct btree_and_journal_iter iter = *_iter;
struct bch_fs *c = iter.trans->c;
unsigned level = iter.journal.level;
struct bkey_buf tmp;
unsigned nr = test_bit(BCH_FS_started, &c->flags)
? (level > 1 ? 0 : 2)
: (level > 1 ? 1 : 16);
iter.prefetch = false;
bch2_bkey_buf_init(&tmp);
while (nr--) {
bch2_btree_and_journal_iter_advance(&iter);
struct bkey_s_c k = bch2_btree_and_journal_iter_peek(&iter);
if (!k.k)
break;
bch2_bkey_buf_reassemble(&tmp, c, k);
bch2_btree_node_prefetch(iter.trans, NULL, tmp.k, iter.journal.btree_id, level - 1);
}
bch2_bkey_buf_exit(&tmp, c);
}
struct bkey_s_c bch2_btree_and_journal_iter_peek(struct btree_and_journal_iter *iter)
{
struct bkey_s_c btree_k, journal_k, ret;
if (iter->prefetch && iter->journal.level)
btree_and_journal_iter_prefetch(iter);
again:
if (iter->at_end)
return bkey_s_c_null;
while ((btree_k = bch2_journal_iter_peek_btree(iter)).k &&
bpos_lt(btree_k.k->p, iter->pos))
bch2_journal_iter_advance_btree(iter);
while ((journal_k = bch2_journal_iter_peek(&iter->journal)).k &&
bpos_lt(journal_k.k->p, iter->pos))
bch2_journal_iter_advance(&iter->journal);
ret = journal_k.k &&
(!btree_k.k || bpos_le(journal_k.k->p, btree_k.k->p))
? journal_k
: btree_k;
if (ret.k && iter->b && bpos_gt(ret.k->p, iter->b->data->max_key))
ret = bkey_s_c_null;
if (ret.k) {
iter->pos = ret.k->p;
if (bkey_deleted(ret.k)) {
bch2_btree_and_journal_iter_advance(iter);
goto again;
}
} else {
iter->pos = SPOS_MAX;
iter->at_end = true;
}
return ret;
}
void bch2_btree_and_journal_iter_exit(struct btree_and_journal_iter *iter)
{
bch2_journal_iter_exit(&iter->journal);
}
void __bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans,
struct btree_and_journal_iter *iter,
struct btree *b,
struct btree_node_iter node_iter,
struct bpos pos)
{
memset(iter, 0, sizeof(*iter));
iter->trans = trans;
iter->b = b;
iter->node_iter = node_iter;
bch2_journal_iter_init(trans->c, &iter->journal, b->c.btree_id, b->c.level, pos);
INIT_LIST_HEAD(&iter->journal.list);
iter->pos = b->data->min_key;
iter->at_end = false;
}
/*
* this version is used by btree_gc before filesystem has gone RW and
* multithreaded, so uses the journal_iters list:
*/
void bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans,
struct btree_and_journal_iter *iter,
struct btree *b)
{
struct btree_node_iter node_iter;
bch2_btree_node_iter_init_from_start(&node_iter, b);
__bch2_btree_and_journal_iter_init_node_iter(trans, iter, b, node_iter, b->data->min_key);
list_add(&iter->journal.list, &trans->c->journal_iters);
}
/* sort and dedup all keys in the journal: */
void bch2_journal_entries_free(struct bch_fs *c)
{
struct journal_replay **i;
struct genradix_iter iter;
genradix_for_each(&c->journal_entries, iter, i)
kvfree(*i);
genradix_free(&c->journal_entries);
}
/*
* When keys compare equal, oldest compares first:
*/
static int journal_sort_key_cmp(const void *_l, const void *_r)
{
const struct journal_key *l = _l;
const struct journal_key *r = _r;
return journal_key_cmp(l, r) ?:
cmp_int(l->journal_seq, r->journal_seq) ?:
cmp_int(l->journal_offset, r->journal_offset);
}
void bch2_journal_keys_put(struct bch_fs *c)
{
struct journal_keys *keys = &c->journal_keys;
struct journal_key *i;
BUG_ON(atomic_read(&keys->ref) <= 0);
if (!atomic_dec_and_test(&keys->ref))
return;
move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr);
keys->gap = keys->nr;
for (i = keys->d; i < keys->d + keys->nr; i++)
if (i->allocated)
kfree(i->k);
kvfree(keys->d);
keys->d = NULL;
keys->nr = keys->gap = keys->size = 0;
bch2_journal_entries_free(c);
}
static void __journal_keys_sort(struct journal_keys *keys)
{
struct journal_key *src, *dst;
sort(keys->d, keys->nr, sizeof(keys->d[0]), journal_sort_key_cmp, NULL);
src = dst = keys->d;
while (src < keys->d + keys->nr) {
while (src + 1 < keys->d + keys->nr &&
!journal_key_cmp(src, src + 1))
src++;
*dst++ = *src++;
}
keys->nr = dst - keys->d;
}
int bch2_journal_keys_sort(struct bch_fs *c)
{
struct genradix_iter iter;
struct journal_replay *i, **_i;
struct journal_keys *keys = &c->journal_keys;
size_t nr_keys = 0, nr_read = 0;
genradix_for_each(&c->journal_entries, iter, _i) {
i = *_i;
if (!i || i->ignore)
continue;
for_each_jset_key(k, entry, &i->j)
nr_keys++;
}
if (!nr_keys)
return 0;
keys->size = roundup_pow_of_two(nr_keys);
keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
if (!keys->d) {
bch_err(c, "Failed to allocate buffer for sorted journal keys (%zu keys); trying slowpath",
nr_keys);
do {
keys->size >>= 1;
keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
} while (!keys->d && keys->size > nr_keys / 8);
if (!keys->d) {
bch_err(c, "Failed to allocate %zu size buffer for sorted journal keys; exiting",
keys->size);
return -BCH_ERR_ENOMEM_journal_keys_sort;
}
}
genradix_for_each(&c->journal_entries, iter, _i) {
i = *_i;
if (!i || i->ignore)
continue;
cond_resched();
for_each_jset_key(k, entry, &i->j) {
if (keys->nr == keys->size) {
__journal_keys_sort(keys);
if (keys->nr > keys->size * 7 / 8) {
bch_err(c, "Too many journal keys for slowpath; have %zu compacted, buf size %zu, processed %zu/%zu",
keys->nr, keys->size, nr_read, nr_keys);
return -BCH_ERR_ENOMEM_journal_keys_sort;
}
}
keys->d[keys->nr++] = (struct journal_key) {
.btree_id = entry->btree_id,
.level = entry->level,
.k = k,
.journal_seq = le64_to_cpu(i->j.seq),
.journal_offset = k->_data - i->j._data,
};
nr_read++;
}
}
__journal_keys_sort(keys);
keys->gap = keys->nr;
bch_verbose(c, "Journal keys: %zu read, %zu after sorting and compacting", nr_keys, keys->nr);
return 0;
}
|