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#ifndef __KERNEL__
81#include "jfs_user.h"
82#else
83#include <linux/time.h>
84#include <linux/fs.h>
85#include <linux/jbd2.h>
86#include <linux/errno.h>
87#include <linux/slab.h>
88#include <linux/list.h>
89#include <linux/init.h>
90#include <linux/bio.h>
91#include <linux/log2.h>
92#include <linux/hash.h>
93#endif
94
95static struct kmem_cache *jbd2_revoke_record_cache;
96static struct kmem_cache *jbd2_revoke_table_cache;
97
98
99
100
101
102struct jbd2_revoke_record_s
103{
104 struct list_head hash;
105 tid_t sequence;
106 unsigned long long blocknr;
107};
108
109
110
111struct jbd2_revoke_table_s
112{
113
114
115 int hash_size;
116 int hash_shift;
117 struct list_head *hash_table;
118};
119
120
121#ifdef __KERNEL__
122static void write_one_revoke_record(transaction_t *,
123 struct list_head *,
124 struct buffer_head **, int *,
125 struct jbd2_revoke_record_s *);
126static void flush_descriptor(journal_t *, struct buffer_head *, int);
127#endif
128
129
130
131static inline int hash(journal_t *journal, unsigned long long block)
132{
133 return hash_64(block, journal->j_revoke->hash_shift);
134}
135
136static int insert_revoke_hash(journal_t *journal, unsigned long long blocknr,
137 tid_t seq)
138{
139 struct list_head *hash_list;
140 struct jbd2_revoke_record_s *record;
141 gfp_t gfp_mask = GFP_NOFS;
142
143 if (journal_oom_retry)
144 gfp_mask |= __GFP_NOFAIL;
145 record = kmem_cache_alloc(jbd2_revoke_record_cache, gfp_mask);
146 if (!record)
147 return -ENOMEM;
148
149 record->sequence = seq;
150 record->blocknr = blocknr;
151 hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
152 spin_lock(&journal->j_revoke_lock);
153 list_add(&record->hash, hash_list);
154 spin_unlock(&journal->j_revoke_lock);
155 return 0;
156}
157
158
159
160static struct jbd2_revoke_record_s *find_revoke_record(journal_t *journal,
161 unsigned long long blocknr)
162{
163 struct list_head *hash_list;
164 struct jbd2_revoke_record_s *record;
165
166 hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
167
168 spin_lock(&journal->j_revoke_lock);
169 record = (struct jbd2_revoke_record_s *) hash_list->next;
170 while (&(record->hash) != hash_list) {
171 if (record->blocknr == blocknr) {
172 spin_unlock(&journal->j_revoke_lock);
173 return record;
174 }
175 record = (struct jbd2_revoke_record_s *) record->hash.next;
176 }
177 spin_unlock(&journal->j_revoke_lock);
178 return NULL;
179}
180
181void jbd2_journal_destroy_revoke_record_cache(void)
182{
183 kmem_cache_destroy(jbd2_revoke_record_cache);
184 jbd2_revoke_record_cache = NULL;
185}
186
187void jbd2_journal_destroy_revoke_table_cache(void)
188{
189 kmem_cache_destroy(jbd2_revoke_table_cache);
190 jbd2_revoke_table_cache = NULL;
191}
192
193int __init jbd2_journal_init_revoke_record_cache(void)
194{
195 J_ASSERT(!jbd2_revoke_record_cache);
196 jbd2_revoke_record_cache = KMEM_CACHE(jbd2_revoke_record_s,
197 SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY);
198
199 if (!jbd2_revoke_record_cache) {
200 pr_emerg("JBD2: failed to create revoke_record cache\n");
201 return -ENOMEM;
202 }
203 return 0;
204}
205
206int __init jbd2_journal_init_revoke_table_cache(void)
207{
208 J_ASSERT(!jbd2_revoke_table_cache);
209 jbd2_revoke_table_cache = KMEM_CACHE(jbd2_revoke_table_s,
210 SLAB_TEMPORARY);
211 if (!jbd2_revoke_table_cache) {
212 pr_emerg("JBD2: failed to create revoke_table cache\n");
213 return -ENOMEM;
214 }
215 return 0;
216}
217
218static struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size)
219{
220 int shift = 0;
221 int tmp = hash_size;
222 struct jbd2_revoke_table_s *table;
223
224 table = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL);
225 if (!table)
226 goto out;
227
228 while((tmp >>= 1UL) != 0UL)
229 shift++;
230
231 table->hash_size = hash_size;
232 table->hash_shift = shift;
233 table->hash_table =
234 kmalloc_array(hash_size, sizeof(struct list_head), GFP_KERNEL);
235 if (!table->hash_table) {
236 kmem_cache_free(jbd2_revoke_table_cache, table);
237 table = NULL;
238 goto out;
239 }
240
241 for (tmp = 0; tmp < hash_size; tmp++)
242 INIT_LIST_HEAD(&table->hash_table[tmp]);
243
244out:
245 return table;
246}
247
248static void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table)
249{
250 int i;
251 struct list_head *hash_list;
252
253 for (i = 0; i < table->hash_size; i++) {
254 hash_list = &table->hash_table[i];
255 J_ASSERT(list_empty(hash_list));
256 }
257
258 kfree(table->hash_table);
259 kmem_cache_free(jbd2_revoke_table_cache, table);
260}
261
262
263int jbd2_journal_init_revoke(journal_t *journal, int hash_size)
264{
265 J_ASSERT(journal->j_revoke_table[0] == NULL);
266 J_ASSERT(is_power_of_2(hash_size));
267
268 journal->j_revoke_table[0] = jbd2_journal_init_revoke_table(hash_size);
269 if (!journal->j_revoke_table[0])
270 goto fail0;
271
272 journal->j_revoke_table[1] = jbd2_journal_init_revoke_table(hash_size);
273 if (!journal->j_revoke_table[1])
274 goto fail1;
275
276 journal->j_revoke = journal->j_revoke_table[1];
277
278 spin_lock_init(&journal->j_revoke_lock);
279
280 return 0;
281
282fail1:
283 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
284 journal->j_revoke_table[0] = NULL;
285fail0:
286 return -ENOMEM;
287}
288
289
290void jbd2_journal_destroy_revoke(journal_t *journal)
291{
292 journal->j_revoke = NULL;
293 if (journal->j_revoke_table[0])
294 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
295 if (journal->j_revoke_table[1])
296 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[1]);
297}
298
299
300#ifdef __KERNEL__
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
326int jbd2_journal_revoke(handle_t *handle, unsigned long long blocknr,
327 struct buffer_head *bh_in)
328{
329 struct buffer_head *bh = NULL;
330 journal_t *journal;
331 struct block_device *bdev;
332 int err;
333
334 might_sleep();
335 if (bh_in)
336 BUFFER_TRACE(bh_in, "enter");
337
338 journal = handle->h_transaction->t_journal;
339 if (!jbd2_journal_set_features(journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)){
340 J_ASSERT (!"Cannot set revoke feature!");
341 return -EINVAL;
342 }
343
344 bdev = journal->j_fs_dev;
345 bh = bh_in;
346
347 if (!bh) {
348 bh = __find_get_block(bdev, blocknr, journal->j_blocksize);
349 if (bh)
350 BUFFER_TRACE(bh, "found on hash");
351 }
352#ifdef JBD2_EXPENSIVE_CHECKING
353 else {
354 struct buffer_head *bh2;
355
356
357
358 bh2 = __find_get_block(bdev, blocknr, journal->j_blocksize);
359 if (bh2) {
360
361 if (bh2 != bh && buffer_revokevalid(bh2))
362
363
364
365
366
367
368 J_ASSERT_BH(bh2, buffer_revoked(bh2));
369 put_bh(bh2);
370 }
371 }
372#endif
373
374 if (WARN_ON_ONCE(handle->h_revoke_credits <= 0)) {
375 if (!bh_in)
376 brelse(bh);
377 return -EIO;
378 }
379
380
381
382 if (bh) {
383 if (!J_EXPECT_BH(bh, !buffer_revoked(bh),
384 "inconsistent data on disk")) {
385 if (!bh_in)
386 brelse(bh);
387 return -EIO;
388 }
389 set_buffer_revoked(bh);
390 set_buffer_revokevalid(bh);
391 if (bh_in) {
392 BUFFER_TRACE(bh_in, "call jbd2_journal_forget");
393 jbd2_journal_forget(handle, bh_in);
394 } else {
395 BUFFER_TRACE(bh, "call brelse");
396 __brelse(bh);
397 }
398 }
399 handle->h_revoke_credits--;
400
401 jbd_debug(2, "insert revoke for block %llu, bh_in=%p\n",blocknr, bh_in);
402 err = insert_revoke_hash(journal, blocknr,
403 handle->h_transaction->t_tid);
404 BUFFER_TRACE(bh_in, "exit");
405 return err;
406}
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423int jbd2_journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
424{
425 struct jbd2_revoke_record_s *record;
426 journal_t *journal = handle->h_transaction->t_journal;
427 int need_cancel;
428 int did_revoke = 0;
429 struct buffer_head *bh = jh2bh(jh);
430
431 jbd_debug(4, "journal_head %p, cancelling revoke\n", jh);
432
433
434
435
436
437 if (test_set_buffer_revokevalid(bh)) {
438 need_cancel = test_clear_buffer_revoked(bh);
439 } else {
440 need_cancel = 1;
441 clear_buffer_revoked(bh);
442 }
443
444 if (need_cancel) {
445 record = find_revoke_record(journal, bh->b_blocknr);
446 if (record) {
447 jbd_debug(4, "cancelled existing revoke on "
448 "blocknr %llu\n", (unsigned long long)bh->b_blocknr);
449 spin_lock(&journal->j_revoke_lock);
450 list_del(&record->hash);
451 spin_unlock(&journal->j_revoke_lock);
452 kmem_cache_free(jbd2_revoke_record_cache, record);
453 did_revoke = 1;
454 }
455 }
456
457#ifdef JBD2_EXPENSIVE_CHECKING
458
459 record = find_revoke_record(journal, bh->b_blocknr);
460 J_ASSERT_JH(jh, record == NULL);
461#endif
462
463
464
465
466
467 if (need_cancel) {
468 struct buffer_head *bh2;
469 bh2 = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
470 if (bh2) {
471 if (bh2 != bh)
472 clear_buffer_revoked(bh2);
473 __brelse(bh2);
474 }
475 }
476 return did_revoke;
477}
478
479
480
481
482
483
484void jbd2_clear_buffer_revoked_flags(journal_t *journal)
485{
486 struct jbd2_revoke_table_s *revoke = journal->j_revoke;
487 int i = 0;
488
489 for (i = 0; i < revoke->hash_size; i++) {
490 struct list_head *hash_list;
491 struct list_head *list_entry;
492 hash_list = &revoke->hash_table[i];
493
494 list_for_each(list_entry, hash_list) {
495 struct jbd2_revoke_record_s *record;
496 struct buffer_head *bh;
497 record = (struct jbd2_revoke_record_s *)list_entry;
498 bh = __find_get_block(journal->j_fs_dev,
499 record->blocknr,
500 journal->j_blocksize);
501 if (bh) {
502 clear_buffer_revoked(bh);
503 __brelse(bh);
504 }
505 }
506 }
507}
508
509
510
511
512
513void jbd2_journal_switch_revoke_table(journal_t *journal)
514{
515 int i;
516
517 if (journal->j_revoke == journal->j_revoke_table[0])
518 journal->j_revoke = journal->j_revoke_table[1];
519 else
520 journal->j_revoke = journal->j_revoke_table[0];
521
522 for (i = 0; i < journal->j_revoke->hash_size; i++)
523 INIT_LIST_HEAD(&journal->j_revoke->hash_table[i]);
524}
525
526
527
528
529
530void jbd2_journal_write_revoke_records(transaction_t *transaction,
531 struct list_head *log_bufs)
532{
533 journal_t *journal = transaction->t_journal;
534 struct buffer_head *descriptor;
535 struct jbd2_revoke_record_s *record;
536 struct jbd2_revoke_table_s *revoke;
537 struct list_head *hash_list;
538 int i, offset, count;
539
540 descriptor = NULL;
541 offset = 0;
542 count = 0;
543
544
545 revoke = journal->j_revoke == journal->j_revoke_table[0] ?
546 journal->j_revoke_table[1] : journal->j_revoke_table[0];
547
548 for (i = 0; i < revoke->hash_size; i++) {
549 hash_list = &revoke->hash_table[i];
550
551 while (!list_empty(hash_list)) {
552 record = (struct jbd2_revoke_record_s *)
553 hash_list->next;
554 write_one_revoke_record(transaction, log_bufs,
555 &descriptor, &offset, record);
556 count++;
557 list_del(&record->hash);
558 kmem_cache_free(jbd2_revoke_record_cache, record);
559 }
560 }
561 if (descriptor)
562 flush_descriptor(journal, descriptor, offset);
563 jbd_debug(1, "Wrote %d revoke records\n", count);
564}
565
566
567
568
569
570
571static void write_one_revoke_record(transaction_t *transaction,
572 struct list_head *log_bufs,
573 struct buffer_head **descriptorp,
574 int *offsetp,
575 struct jbd2_revoke_record_s *record)
576{
577 journal_t *journal = transaction->t_journal;
578 int csum_size = 0;
579 struct buffer_head *descriptor;
580 int sz, offset;
581
582
583
584
585
586 if (is_journal_aborted(journal))
587 return;
588
589 descriptor = *descriptorp;
590 offset = *offsetp;
591
592
593 if (jbd2_journal_has_csum_v2or3(journal))
594 csum_size = sizeof(struct jbd2_journal_block_tail);
595
596 if (jbd2_has_feature_64bit(journal))
597 sz = 8;
598 else
599 sz = 4;
600
601
602 if (descriptor) {
603 if (offset + sz > journal->j_blocksize - csum_size) {
604 flush_descriptor(journal, descriptor, offset);
605 descriptor = NULL;
606 }
607 }
608
609 if (!descriptor) {
610 descriptor = jbd2_journal_get_descriptor_buffer(transaction,
611 JBD2_REVOKE_BLOCK);
612 if (!descriptor)
613 return;
614
615
616 BUFFER_TRACE(descriptor, "file in log_bufs");
617 jbd2_file_log_bh(log_bufs, descriptor);
618
619 offset = sizeof(jbd2_journal_revoke_header_t);
620 *descriptorp = descriptor;
621 }
622
623 if (jbd2_has_feature_64bit(journal))
624 * ((__be64 *)(&descriptor->b_data[offset])) =
625 cpu_to_be64(record->blocknr);
626 else
627 * ((__be32 *)(&descriptor->b_data[offset])) =
628 cpu_to_be32(record->blocknr);
629 offset += sz;
630
631 *offsetp = offset;
632}
633
634
635
636
637
638
639
640
641static void flush_descriptor(journal_t *journal,
642 struct buffer_head *descriptor,
643 int offset)
644{
645 jbd2_journal_revoke_header_t *header;
646
647 if (is_journal_aborted(journal))
648 return;
649
650 header = (jbd2_journal_revoke_header_t *)descriptor->b_data;
651 header->r_count = cpu_to_be32(offset);
652 jbd2_descriptor_block_csum_set(journal, descriptor);
653
654 set_buffer_jwrite(descriptor);
655 BUFFER_TRACE(descriptor, "write");
656 set_buffer_dirty(descriptor);
657 write_dirty_buffer(descriptor, REQ_SYNC);
658}
659#endif
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683int jbd2_journal_set_revoke(journal_t *journal,
684 unsigned long long blocknr,
685 tid_t sequence)
686{
687 struct jbd2_revoke_record_s *record;
688
689 record = find_revoke_record(journal, blocknr);
690 if (record) {
691
692
693 if (tid_gt(sequence, record->sequence))
694 record->sequence = sequence;
695 return 0;
696 }
697 return insert_revoke_hash(journal, blocknr, sequence);
698}
699
700
701
702
703
704
705
706
707int jbd2_journal_test_revoke(journal_t *journal,
708 unsigned long long blocknr,
709 tid_t sequence)
710{
711 struct jbd2_revoke_record_s *record;
712
713 record = find_revoke_record(journal, blocknr);
714 if (!record)
715 return 0;
716 if (tid_gt(sequence, record->sequence))
717 return 0;
718 return 1;
719}
720
721
722
723
724
725
726void jbd2_journal_clear_revoke(journal_t *journal)
727{
728 int i;
729 struct list_head *hash_list;
730 struct jbd2_revoke_record_s *record;
731 struct jbd2_revoke_table_s *revoke;
732
733 revoke = journal->j_revoke;
734
735 for (i = 0; i < revoke->hash_size; i++) {
736 hash_list = &revoke->hash_table[i];
737 while (!list_empty(hash_list)) {
738 record = (struct jbd2_revoke_record_s*) hash_list->next;
739 list_del(&record->hash);
740 kmem_cache_free(jbd2_revoke_record_cache, record);
741 }
742 }
743}
744