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#include <linux/time.h>
38#include <linux/semaphore.h>
39#include <linux/vmalloc.h>
40#include "reiserfs.h"
41#include <linux/kernel.h>
42#include <linux/errno.h>
43#include <linux/fcntl.h>
44#include <linux/stat.h>
45#include <linux/string.h>
46#include <linux/buffer_head.h>
47#include <linux/workqueue.h>
48#include <linux/writeback.h>
49#include <linux/blkdev.h>
50#include <linux/backing-dev.h>
51#include <linux/uaccess.h>
52#include <linux/slab.h>
53
54
55
56#define JOURNAL_LIST_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
57 j_list))
58#define JOURNAL_WORK_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
59 j_working_list))
60
61
62
63
64static int reiserfs_mounted_fs_count;
65
66static struct workqueue_struct *commit_wq;
67
68#define JOURNAL_TRANS_HALF 1018
69
70#define BUFNR 64
71
72
73
74#define BLOCK_FREED 2
75#define BLOCK_FREED_HOLDER 3
76
77#define BLOCK_NEEDS_FLUSH 4
78#define BLOCK_DIRTIED 5
79
80
81#define LIST_TOUCHED 1
82#define LIST_DIRTY 2
83#define LIST_COMMIT_PENDING 4
84
85
86#define FLUSH_ALL 1
87#define COMMIT_NOW 2
88#define WAIT 4
89
90static int do_journal_end(struct reiserfs_transaction_handle *,
91 struct super_block *, unsigned long nblocks,
92 int flags);
93static int flush_journal_list(struct super_block *s,
94 struct reiserfs_journal_list *jl, int flushall);
95static int flush_commit_list(struct super_block *s,
96 struct reiserfs_journal_list *jl, int flushall);
97static int can_dirty(struct reiserfs_journal_cnode *cn);
98static int journal_join(struct reiserfs_transaction_handle *th,
99 struct super_block *sb, unsigned long nblocks);
100static void release_journal_dev(struct super_block *super,
101 struct reiserfs_journal *journal);
102static int dirty_one_transaction(struct super_block *s,
103 struct reiserfs_journal_list *jl);
104static void flush_async_commits(struct work_struct *work);
105static void queue_log_writer(struct super_block *s);
106
107
108enum {
109 JBEGIN_REG = 0,
110 JBEGIN_JOIN = 1,
111 JBEGIN_ABORT = 2,
112};
113
114static int do_journal_begin_r(struct reiserfs_transaction_handle *th,
115 struct super_block *sb,
116 unsigned long nblocks, int join);
117
118static void init_journal_hash(struct super_block *sb)
119{
120 struct reiserfs_journal *journal = SB_JOURNAL(sb);
121 memset(journal->j_hash_table, 0,
122 JOURNAL_HASH_SIZE * sizeof(struct reiserfs_journal_cnode *));
123}
124
125
126
127
128
129
130static int reiserfs_clean_and_file_buffer(struct buffer_head *bh)
131{
132 if (bh) {
133 clear_buffer_dirty(bh);
134 clear_buffer_journal_test(bh);
135 }
136 return 0;
137}
138
139static struct reiserfs_bitmap_node *allocate_bitmap_node(struct super_block
140 *sb)
141{
142 struct reiserfs_bitmap_node *bn;
143 static int id;
144
145 bn = kmalloc(sizeof(struct reiserfs_bitmap_node), GFP_NOFS);
146 if (!bn) {
147 return NULL;
148 }
149 bn->data = kzalloc(sb->s_blocksize, GFP_NOFS);
150 if (!bn->data) {
151 kfree(bn);
152 return NULL;
153 }
154 bn->id = id++;
155 INIT_LIST_HEAD(&bn->list);
156 return bn;
157}
158
159static struct reiserfs_bitmap_node *get_bitmap_node(struct super_block *sb)
160{
161 struct reiserfs_journal *journal = SB_JOURNAL(sb);
162 struct reiserfs_bitmap_node *bn = NULL;
163 struct list_head *entry = journal->j_bitmap_nodes.next;
164
165 journal->j_used_bitmap_nodes++;
166 repeat:
167
168 if (entry != &journal->j_bitmap_nodes) {
169 bn = list_entry(entry, struct reiserfs_bitmap_node, list);
170 list_del(entry);
171 memset(bn->data, 0, sb->s_blocksize);
172 journal->j_free_bitmap_nodes--;
173 return bn;
174 }
175 bn = allocate_bitmap_node(sb);
176 if (!bn) {
177 yield();
178 goto repeat;
179 }
180 return bn;
181}
182static inline void free_bitmap_node(struct super_block *sb,
183 struct reiserfs_bitmap_node *bn)
184{
185 struct reiserfs_journal *journal = SB_JOURNAL(sb);
186 journal->j_used_bitmap_nodes--;
187 if (journal->j_free_bitmap_nodes > REISERFS_MAX_BITMAP_NODES) {
188 kfree(bn->data);
189 kfree(bn);
190 } else {
191 list_add(&bn->list, &journal->j_bitmap_nodes);
192 journal->j_free_bitmap_nodes++;
193 }
194}
195
196static void allocate_bitmap_nodes(struct super_block *sb)
197{
198 int i;
199 struct reiserfs_journal *journal = SB_JOURNAL(sb);
200 struct reiserfs_bitmap_node *bn = NULL;
201 for (i = 0; i < REISERFS_MIN_BITMAP_NODES; i++) {
202 bn = allocate_bitmap_node(sb);
203 if (bn) {
204 list_add(&bn->list, &journal->j_bitmap_nodes);
205 journal->j_free_bitmap_nodes++;
206 } else {
207 break;
208 }
209 }
210}
211
212static int set_bit_in_list_bitmap(struct super_block *sb,
213 b_blocknr_t block,
214 struct reiserfs_list_bitmap *jb)
215{
216 unsigned int bmap_nr = block / (sb->s_blocksize << 3);
217 unsigned int bit_nr = block % (sb->s_blocksize << 3);
218
219 if (!jb->bitmaps[bmap_nr]) {
220 jb->bitmaps[bmap_nr] = get_bitmap_node(sb);
221 }
222 set_bit(bit_nr, (unsigned long *)jb->bitmaps[bmap_nr]->data);
223 return 0;
224}
225
226static void cleanup_bitmap_list(struct super_block *sb,
227 struct reiserfs_list_bitmap *jb)
228{
229 int i;
230 if (jb->bitmaps == NULL)
231 return;
232
233 for (i = 0; i < reiserfs_bmap_count(sb); i++) {
234 if (jb->bitmaps[i]) {
235 free_bitmap_node(sb, jb->bitmaps[i]);
236 jb->bitmaps[i] = NULL;
237 }
238 }
239}
240
241
242
243
244static int free_list_bitmaps(struct super_block *sb,
245 struct reiserfs_list_bitmap *jb_array)
246{
247 int i;
248 struct reiserfs_list_bitmap *jb;
249 for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
250 jb = jb_array + i;
251 jb->journal_list = NULL;
252 cleanup_bitmap_list(sb, jb);
253 vfree(jb->bitmaps);
254 jb->bitmaps = NULL;
255 }
256 return 0;
257}
258
259static int free_bitmap_nodes(struct super_block *sb)
260{
261 struct reiserfs_journal *journal = SB_JOURNAL(sb);
262 struct list_head *next = journal->j_bitmap_nodes.next;
263 struct reiserfs_bitmap_node *bn;
264
265 while (next != &journal->j_bitmap_nodes) {
266 bn = list_entry(next, struct reiserfs_bitmap_node, list);
267 list_del(next);
268 kfree(bn->data);
269 kfree(bn);
270 next = journal->j_bitmap_nodes.next;
271 journal->j_free_bitmap_nodes--;
272 }
273
274 return 0;
275}
276
277
278
279
280
281int reiserfs_allocate_list_bitmaps(struct super_block *sb,
282 struct reiserfs_list_bitmap *jb_array,
283 unsigned int bmap_nr)
284{
285 int i;
286 int failed = 0;
287 struct reiserfs_list_bitmap *jb;
288 int mem = bmap_nr * sizeof(struct reiserfs_bitmap_node *);
289
290 for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
291 jb = jb_array + i;
292 jb->journal_list = NULL;
293 jb->bitmaps = vzalloc(mem);
294 if (!jb->bitmaps) {
295 reiserfs_warning(sb, "clm-2000", "unable to "
296 "allocate bitmaps for journal lists");
297 failed = 1;
298 break;
299 }
300 }
301 if (failed) {
302 free_list_bitmaps(sb, jb_array);
303 return -1;
304 }
305 return 0;
306}
307
308
309
310
311
312static struct reiserfs_list_bitmap *get_list_bitmap(struct super_block *sb,
313 struct reiserfs_journal_list
314 *jl)
315{
316 int i, j;
317 struct reiserfs_journal *journal = SB_JOURNAL(sb);
318 struct reiserfs_list_bitmap *jb = NULL;
319
320 for (j = 0; j < (JOURNAL_NUM_BITMAPS * 3); j++) {
321 i = journal->j_list_bitmap_index;
322 journal->j_list_bitmap_index = (i + 1) % JOURNAL_NUM_BITMAPS;
323 jb = journal->j_list_bitmap + i;
324 if (journal->j_list_bitmap[i].journal_list) {
325 flush_commit_list(sb,
326 journal->j_list_bitmap[i].
327 journal_list, 1);
328 if (!journal->j_list_bitmap[i].journal_list) {
329 break;
330 }
331 } else {
332 break;
333 }
334 }
335 if (jb->journal_list) {
336 return NULL;
337 }
338 jb->journal_list = jl;
339 return jb;
340}
341
342
343
344
345
346
347static struct reiserfs_journal_cnode *allocate_cnodes(int num_cnodes)
348{
349 struct reiserfs_journal_cnode *head;
350 int i;
351 if (num_cnodes <= 0) {
352 return NULL;
353 }
354 head = vzalloc(num_cnodes * sizeof(struct reiserfs_journal_cnode));
355 if (!head) {
356 return NULL;
357 }
358 head[0].prev = NULL;
359 head[0].next = head + 1;
360 for (i = 1; i < num_cnodes; i++) {
361 head[i].prev = head + (i - 1);
362 head[i].next = head + (i + 1);
363 }
364 head[num_cnodes - 1].next = NULL;
365 return head;
366}
367
368
369
370
371static struct reiserfs_journal_cnode *get_cnode(struct super_block *sb)
372{
373 struct reiserfs_journal_cnode *cn;
374 struct reiserfs_journal *journal = SB_JOURNAL(sb);
375
376 reiserfs_check_lock_depth(sb, "get_cnode");
377
378 if (journal->j_cnode_free <= 0) {
379 return NULL;
380 }
381 journal->j_cnode_used++;
382 journal->j_cnode_free--;
383 cn = journal->j_cnode_free_list;
384 if (!cn) {
385 return cn;
386 }
387 if (cn->next) {
388 cn->next->prev = NULL;
389 }
390 journal->j_cnode_free_list = cn->next;
391 memset(cn, 0, sizeof(struct reiserfs_journal_cnode));
392 return cn;
393}
394
395
396
397
398static void free_cnode(struct super_block *sb,
399 struct reiserfs_journal_cnode *cn)
400{
401 struct reiserfs_journal *journal = SB_JOURNAL(sb);
402
403 reiserfs_check_lock_depth(sb, "free_cnode");
404
405 journal->j_cnode_used--;
406 journal->j_cnode_free++;
407
408 cn->next = journal->j_cnode_free_list;
409 if (journal->j_cnode_free_list) {
410 journal->j_cnode_free_list->prev = cn;
411 }
412 cn->prev = NULL;
413 journal->j_cnode_free_list = cn;
414}
415
416static void clear_prepared_bits(struct buffer_head *bh)
417{
418 clear_buffer_journal_prepared(bh);
419 clear_buffer_journal_restore_dirty(bh);
420}
421
422
423static inline struct reiserfs_journal_cnode *get_journal_hash_dev(struct
424 super_block
425 *sb,
426 struct
427 reiserfs_journal_cnode
428 **table,
429 long bl)
430{
431 struct reiserfs_journal_cnode *cn;
432 cn = journal_hash(table, sb, bl);
433 while (cn) {
434 if (cn->blocknr == bl && cn->sb == sb)
435 return cn;
436 cn = cn->hnext;
437 }
438 return (struct reiserfs_journal_cnode *)0;
439}
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459int reiserfs_in_journal(struct super_block *sb,
460 unsigned int bmap_nr, int bit_nr, int search_all,
461 b_blocknr_t * next_zero_bit)
462{
463 struct reiserfs_journal *journal = SB_JOURNAL(sb);
464 struct reiserfs_journal_cnode *cn;
465 struct reiserfs_list_bitmap *jb;
466 int i;
467 unsigned long bl;
468
469 *next_zero_bit = 0;
470
471 PROC_INFO_INC(sb, journal.in_journal);
472
473
474
475
476 if (search_all) {
477 for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
478 PROC_INFO_INC(sb, journal.in_journal_bitmap);
479 jb = journal->j_list_bitmap + i;
480 if (jb->journal_list && jb->bitmaps[bmap_nr] &&
481 test_bit(bit_nr,
482 (unsigned long *)jb->bitmaps[bmap_nr]->
483 data)) {
484 *next_zero_bit =
485 find_next_zero_bit((unsigned long *)
486 (jb->bitmaps[bmap_nr]->
487 data),
488 sb->s_blocksize << 3,
489 bit_nr + 1);
490 return 1;
491 }
492 }
493 }
494
495 bl = bmap_nr * (sb->s_blocksize << 3) + bit_nr;
496
497 if (search_all
498 && (cn =
499 get_journal_hash_dev(sb, journal->j_list_hash_table, bl))) {
500 return 1;
501 }
502
503
504 if ((cn = get_journal_hash_dev(sb, journal->j_hash_table, bl))) {
505 BUG();
506 return 1;
507 }
508
509 PROC_INFO_INC(sb, journal.in_journal_reusable);
510
511 return 0;
512}
513
514
515
516static inline void insert_journal_hash(struct reiserfs_journal_cnode **table,
517 struct reiserfs_journal_cnode *cn)
518{
519 struct reiserfs_journal_cnode *cn_orig;
520
521 cn_orig = journal_hash(table, cn->sb, cn->blocknr);
522 cn->hnext = cn_orig;
523 cn->hprev = NULL;
524 if (cn_orig) {
525 cn_orig->hprev = cn;
526 }
527 journal_hash(table, cn->sb, cn->blocknr) = cn;
528}
529
530
531static inline void lock_journal(struct super_block *sb)
532{
533 PROC_INFO_INC(sb, journal.lock_journal);
534
535 reiserfs_mutex_lock_safe(&SB_JOURNAL(sb)->j_mutex, sb);
536}
537
538
539static inline void unlock_journal(struct super_block *sb)
540{
541 mutex_unlock(&SB_JOURNAL(sb)->j_mutex);
542}
543
544static inline void get_journal_list(struct reiserfs_journal_list *jl)
545{
546 jl->j_refcount++;
547}
548
549static inline void put_journal_list(struct super_block *s,
550 struct reiserfs_journal_list *jl)
551{
552 if (jl->j_refcount < 1) {
553 reiserfs_panic(s, "journal-2", "trans id %u, refcount at %d",
554 jl->j_trans_id, jl->j_refcount);
555 }
556 if (--jl->j_refcount == 0)
557 kfree(jl);
558}
559
560
561
562
563
564
565static void cleanup_freed_for_journal_list(struct super_block *sb,
566 struct reiserfs_journal_list *jl)
567{
568
569 struct reiserfs_list_bitmap *jb = jl->j_list_bitmap;
570 if (jb) {
571 cleanup_bitmap_list(sb, jb);
572 }
573 jl->j_list_bitmap->journal_list = NULL;
574 jl->j_list_bitmap = NULL;
575}
576
577static int journal_list_still_alive(struct super_block *s,
578 unsigned int trans_id)
579{
580 struct reiserfs_journal *journal = SB_JOURNAL(s);
581 struct list_head *entry = &journal->j_journal_list;
582 struct reiserfs_journal_list *jl;
583
584 if (!list_empty(entry)) {
585 jl = JOURNAL_LIST_ENTRY(entry->next);
586 if (jl->j_trans_id <= trans_id) {
587 return 1;
588 }
589 }
590 return 0;
591}
592
593
594
595
596
597
598
599
600
601
602
603static void release_buffer_page(struct buffer_head *bh)
604{
605 struct page *page = bh->b_page;
606 if (!page->mapping && trylock_page(page)) {
607 page_cache_get(page);
608 put_bh(bh);
609 if (!page->mapping)
610 try_to_free_buffers(page);
611 unlock_page(page);
612 page_cache_release(page);
613 } else {
614 put_bh(bh);
615 }
616}
617
618static void reiserfs_end_buffer_io_sync(struct buffer_head *bh, int uptodate)
619{
620 char b[BDEVNAME_SIZE];
621
622 if (buffer_journaled(bh)) {
623 reiserfs_warning(NULL, "clm-2084",
624 "pinned buffer %lu:%s sent to disk",
625 bh->b_blocknr, bdevname(bh->b_bdev, b));
626 }
627 if (uptodate)
628 set_buffer_uptodate(bh);
629 else
630 clear_buffer_uptodate(bh);
631
632 unlock_buffer(bh);
633 release_buffer_page(bh);
634}
635
636static void reiserfs_end_ordered_io(struct buffer_head *bh, int uptodate)
637{
638 if (uptodate)
639 set_buffer_uptodate(bh);
640 else
641 clear_buffer_uptodate(bh);
642 unlock_buffer(bh);
643 put_bh(bh);
644}
645
646static void submit_logged_buffer(struct buffer_head *bh)
647{
648 get_bh(bh);
649 bh->b_end_io = reiserfs_end_buffer_io_sync;
650 clear_buffer_journal_new(bh);
651 clear_buffer_dirty(bh);
652 if (!test_clear_buffer_journal_test(bh))
653 BUG();
654 if (!buffer_uptodate(bh))
655 BUG();
656 submit_bh(WRITE, bh);
657}
658
659static void submit_ordered_buffer(struct buffer_head *bh)
660{
661 get_bh(bh);
662 bh->b_end_io = reiserfs_end_ordered_io;
663 clear_buffer_dirty(bh);
664 if (!buffer_uptodate(bh))
665 BUG();
666 submit_bh(WRITE, bh);
667}
668
669#define CHUNK_SIZE 32
670struct buffer_chunk {
671 struct buffer_head *bh[CHUNK_SIZE];
672 int nr;
673};
674
675static void write_chunk(struct buffer_chunk *chunk)
676{
677 int i;
678 for (i = 0; i < chunk->nr; i++) {
679 submit_logged_buffer(chunk->bh[i]);
680 }
681 chunk->nr = 0;
682}
683
684static void write_ordered_chunk(struct buffer_chunk *chunk)
685{
686 int i;
687 for (i = 0; i < chunk->nr; i++) {
688 submit_ordered_buffer(chunk->bh[i]);
689 }
690 chunk->nr = 0;
691}
692
693static int add_to_chunk(struct buffer_chunk *chunk, struct buffer_head *bh,
694 spinlock_t * lock, void (fn) (struct buffer_chunk *))
695{
696 int ret = 0;
697 BUG_ON(chunk->nr >= CHUNK_SIZE);
698 chunk->bh[chunk->nr++] = bh;
699 if (chunk->nr >= CHUNK_SIZE) {
700 ret = 1;
701 if (lock)
702 spin_unlock(lock);
703 fn(chunk);
704 if (lock)
705 spin_lock(lock);
706 }
707 return ret;
708}
709
710static atomic_t nr_reiserfs_jh = ATOMIC_INIT(0);
711static struct reiserfs_jh *alloc_jh(void)
712{
713 struct reiserfs_jh *jh;
714 while (1) {
715 jh = kmalloc(sizeof(*jh), GFP_NOFS);
716 if (jh) {
717 atomic_inc(&nr_reiserfs_jh);
718 return jh;
719 }
720 yield();
721 }
722}
723
724
725
726
727
728void reiserfs_free_jh(struct buffer_head *bh)
729{
730 struct reiserfs_jh *jh;
731
732 jh = bh->b_private;
733 if (jh) {
734 bh->b_private = NULL;
735 jh->bh = NULL;
736 list_del_init(&jh->list);
737 kfree(jh);
738 if (atomic_read(&nr_reiserfs_jh) <= 0)
739 BUG();
740 atomic_dec(&nr_reiserfs_jh);
741 put_bh(bh);
742 }
743}
744
745static inline int __add_jh(struct reiserfs_journal *j, struct buffer_head *bh,
746 int tail)
747{
748 struct reiserfs_jh *jh;
749
750 if (bh->b_private) {
751 spin_lock(&j->j_dirty_buffers_lock);
752 if (!bh->b_private) {
753 spin_unlock(&j->j_dirty_buffers_lock);
754 goto no_jh;
755 }
756 jh = bh->b_private;
757 list_del_init(&jh->list);
758 } else {
759 no_jh:
760 get_bh(bh);
761 jh = alloc_jh();
762 spin_lock(&j->j_dirty_buffers_lock);
763
764
765
766 BUG_ON(bh->b_private);
767 jh->bh = bh;
768 bh->b_private = jh;
769 }
770 jh->jl = j->j_current_jl;
771 if (tail)
772 list_add_tail(&jh->list, &jh->jl->j_tail_bh_list);
773 else {
774 list_add_tail(&jh->list, &jh->jl->j_bh_list);
775 }
776 spin_unlock(&j->j_dirty_buffers_lock);
777 return 0;
778}
779
780int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh)
781{
782 return __add_jh(SB_JOURNAL(inode->i_sb), bh, 1);
783}
784int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh)
785{
786 return __add_jh(SB_JOURNAL(inode->i_sb), bh, 0);
787}
788
789#define JH_ENTRY(l) list_entry((l), struct reiserfs_jh, list)
790static int write_ordered_buffers(spinlock_t * lock,
791 struct reiserfs_journal *j,
792 struct reiserfs_journal_list *jl,
793 struct list_head *list)
794{
795 struct buffer_head *bh;
796 struct reiserfs_jh *jh;
797 int ret = j->j_errno;
798 struct buffer_chunk chunk;
799 struct list_head tmp;
800 INIT_LIST_HEAD(&tmp);
801
802 chunk.nr = 0;
803 spin_lock(lock);
804 while (!list_empty(list)) {
805 jh = JH_ENTRY(list->next);
806 bh = jh->bh;
807 get_bh(bh);
808 if (!trylock_buffer(bh)) {
809 if (!buffer_dirty(bh)) {
810 list_move(&jh->list, &tmp);
811 goto loop_next;
812 }
813 spin_unlock(lock);
814 if (chunk.nr)
815 write_ordered_chunk(&chunk);
816 wait_on_buffer(bh);
817 cond_resched();
818 spin_lock(lock);
819 goto loop_next;
820 }
821
822
823
824
825 if (!buffer_uptodate(bh) && buffer_dirty(bh)) {
826 clear_buffer_dirty(bh);
827 ret = -EIO;
828 }
829 if (buffer_dirty(bh)) {
830 list_move(&jh->list, &tmp);
831 add_to_chunk(&chunk, bh, lock, write_ordered_chunk);
832 } else {
833 reiserfs_free_jh(bh);
834 unlock_buffer(bh);
835 }
836 loop_next:
837 put_bh(bh);
838 cond_resched_lock(lock);
839 }
840 if (chunk.nr) {
841 spin_unlock(lock);
842 write_ordered_chunk(&chunk);
843 spin_lock(lock);
844 }
845 while (!list_empty(&tmp)) {
846 jh = JH_ENTRY(tmp.prev);
847 bh = jh->bh;
848 get_bh(bh);
849 reiserfs_free_jh(bh);
850
851 if (buffer_locked(bh)) {
852 spin_unlock(lock);
853 wait_on_buffer(bh);
854 spin_lock(lock);
855 }
856 if (!buffer_uptodate(bh)) {
857 ret = -EIO;
858 }
859
860
861
862
863
864
865
866
867 if (buffer_dirty(bh) && unlikely(bh->b_page->mapping == NULL)) {
868 spin_unlock(lock);
869 ll_rw_block(WRITE, 1, &bh);
870 spin_lock(lock);
871 }
872 put_bh(bh);
873 cond_resched_lock(lock);
874 }
875 spin_unlock(lock);
876 return ret;
877}
878
879static int flush_older_commits(struct super_block *s,
880 struct reiserfs_journal_list *jl)
881{
882 struct reiserfs_journal *journal = SB_JOURNAL(s);
883 struct reiserfs_journal_list *other_jl;
884 struct reiserfs_journal_list *first_jl;
885 struct list_head *entry;
886 unsigned int trans_id = jl->j_trans_id;
887 unsigned int other_trans_id;
888 unsigned int first_trans_id;
889
890 find_first:
891
892
893
894 first_jl = jl;
895 entry = jl->j_list.prev;
896 while (1) {
897 other_jl = JOURNAL_LIST_ENTRY(entry);
898 if (entry == &journal->j_journal_list ||
899 atomic_read(&other_jl->j_older_commits_done))
900 break;
901
902 first_jl = other_jl;
903 entry = other_jl->j_list.prev;
904 }
905
906
907 if (first_jl == jl) {
908 return 0;
909 }
910
911 first_trans_id = first_jl->j_trans_id;
912
913 entry = &first_jl->j_list;
914 while (1) {
915 other_jl = JOURNAL_LIST_ENTRY(entry);
916 other_trans_id = other_jl->j_trans_id;
917
918 if (other_trans_id < trans_id) {
919 if (atomic_read(&other_jl->j_commit_left) != 0) {
920 flush_commit_list(s, other_jl, 0);
921
922
923 if (!journal_list_still_alive(s, trans_id))
924 return 1;
925
926
927
928
929
930 if (!journal_list_still_alive
931 (s, other_trans_id)) {
932 goto find_first;
933 }
934 }
935 entry = entry->next;
936 if (entry == &journal->j_journal_list)
937 return 0;
938 } else {
939 return 0;
940 }
941 }
942 return 0;
943}
944
945static int reiserfs_async_progress_wait(struct super_block *s)
946{
947 struct reiserfs_journal *j = SB_JOURNAL(s);
948
949 if (atomic_read(&j->j_async_throttle)) {
950 reiserfs_write_unlock(s);
951 congestion_wait(BLK_RW_ASYNC, HZ / 10);
952 reiserfs_write_lock(s);
953 }
954
955 return 0;
956}
957
958
959
960
961
962
963
964
965static int flush_commit_list(struct super_block *s,
966 struct reiserfs_journal_list *jl, int flushall)
967{
968 int i;
969 b_blocknr_t bn;
970 struct buffer_head *tbh = NULL;
971 unsigned int trans_id = jl->j_trans_id;
972 struct reiserfs_journal *journal = SB_JOURNAL(s);
973 int retval = 0;
974 int write_len;
975
976 reiserfs_check_lock_depth(s, "flush_commit_list");
977
978 if (atomic_read(&jl->j_older_commits_done)) {
979 return 0;
980 }
981
982
983
984
985 BUG_ON(jl->j_len <= 0);
986 BUG_ON(trans_id == journal->j_trans_id);
987
988 get_journal_list(jl);
989 if (flushall) {
990 if (flush_older_commits(s, jl) == 1) {
991
992 goto put_jl;
993 }
994 }
995
996
997 reiserfs_mutex_lock_safe(&jl->j_commit_mutex, s);
998
999 if (!journal_list_still_alive(s, trans_id)) {
1000 mutex_unlock(&jl->j_commit_mutex);
1001 goto put_jl;
1002 }
1003 BUG_ON(jl->j_trans_id == 0);
1004
1005
1006 if (atomic_read(&(jl->j_commit_left)) <= 0) {
1007 if (flushall) {
1008 atomic_set(&(jl->j_older_commits_done), 1);
1009 }
1010 mutex_unlock(&jl->j_commit_mutex);
1011 goto put_jl;
1012 }
1013
1014 if (!list_empty(&jl->j_bh_list)) {
1015 int ret;
1016
1017
1018
1019
1020
1021 reiserfs_write_unlock(s);
1022 ret = write_ordered_buffers(&journal->j_dirty_buffers_lock,
1023 journal, jl, &jl->j_bh_list);
1024 if (ret < 0 && retval == 0)
1025 retval = ret;
1026 reiserfs_write_lock(s);
1027 }
1028 BUG_ON(!list_empty(&jl->j_bh_list));
1029
1030
1031
1032
1033
1034
1035
1036 atomic_inc(&journal->j_async_throttle);
1037 write_len = jl->j_len + 1;
1038 if (write_len < 256)
1039 write_len = 256;
1040 for (i = 0 ; i < write_len ; i++) {
1041 bn = SB_ONDISK_JOURNAL_1st_BLOCK(s) + (jl->j_start + i) %
1042 SB_ONDISK_JOURNAL_SIZE(s);
1043 tbh = journal_find_get_block(s, bn);
1044 if (tbh) {
1045 if (buffer_dirty(tbh)) {
1046 reiserfs_write_unlock(s);
1047 ll_rw_block(WRITE, 1, &tbh);
1048 reiserfs_write_lock(s);
1049 }
1050 put_bh(tbh) ;
1051 }
1052 }
1053 atomic_dec(&journal->j_async_throttle);
1054
1055 for (i = 0; i < (jl->j_len + 1); i++) {
1056 bn = SB_ONDISK_JOURNAL_1st_BLOCK(s) +
1057 (jl->j_start + i) % SB_ONDISK_JOURNAL_SIZE(s);
1058 tbh = journal_find_get_block(s, bn);
1059
1060 reiserfs_write_unlock(s);
1061 wait_on_buffer(tbh);
1062 reiserfs_write_lock(s);
1063
1064
1065
1066
1067 if (buffer_dirty(tbh)) {
1068 reiserfs_write_unlock(s);
1069 sync_dirty_buffer(tbh);
1070 reiserfs_write_lock(s);
1071 }
1072 if (unlikely(!buffer_uptodate(tbh))) {
1073#ifdef CONFIG_REISERFS_CHECK
1074 reiserfs_warning(s, "journal-601",
1075 "buffer write failed");
1076#endif
1077 retval = -EIO;
1078 }
1079 put_bh(tbh);
1080 put_bh(tbh);
1081 atomic_dec(&(jl->j_commit_left));
1082 }
1083
1084 BUG_ON(atomic_read(&(jl->j_commit_left)) != 1);
1085
1086
1087
1088
1089
1090 if (likely(!retval && !reiserfs_is_journal_aborted (journal))) {
1091 if (buffer_dirty(jl->j_commit_bh))
1092 BUG();
1093 mark_buffer_dirty(jl->j_commit_bh) ;
1094 reiserfs_write_unlock(s);
1095 if (reiserfs_barrier_flush(s))
1096 __sync_dirty_buffer(jl->j_commit_bh, WRITE_FLUSH_FUA);
1097 else
1098 sync_dirty_buffer(jl->j_commit_bh);
1099 reiserfs_write_lock(s);
1100 }
1101
1102
1103
1104
1105 if (unlikely(!buffer_uptodate(jl->j_commit_bh))) {
1106#ifdef CONFIG_REISERFS_CHECK
1107 reiserfs_warning(s, "journal-615", "buffer write failed");
1108#endif
1109 retval = -EIO;
1110 }
1111 bforget(jl->j_commit_bh);
1112 if (journal->j_last_commit_id != 0 &&
1113 (jl->j_trans_id - journal->j_last_commit_id) != 1) {
1114 reiserfs_warning(s, "clm-2200", "last commit %lu, current %lu",
1115 journal->j_last_commit_id, jl->j_trans_id);
1116 }
1117 journal->j_last_commit_id = jl->j_trans_id;
1118
1119
1120 cleanup_freed_for_journal_list(s, jl);
1121
1122 retval = retval ? retval : journal->j_errno;
1123
1124
1125 if (!retval)
1126 dirty_one_transaction(s, jl);
1127 atomic_dec(&(jl->j_commit_left));
1128
1129 if (flushall) {
1130 atomic_set(&(jl->j_older_commits_done), 1);
1131 }
1132 mutex_unlock(&jl->j_commit_mutex);
1133 put_jl:
1134 put_journal_list(s, jl);
1135
1136 if (retval)
1137 reiserfs_abort(s, retval, "Journal write error in %s",
1138 __func__);
1139 return retval;
1140}
1141
1142
1143
1144
1145
1146static struct reiserfs_journal_list *find_newer_jl_for_cn(struct
1147 reiserfs_journal_cnode
1148 *cn)
1149{
1150 struct super_block *sb = cn->sb;
1151 b_blocknr_t blocknr = cn->blocknr;
1152
1153 cn = cn->hprev;
1154 while (cn) {
1155 if (cn->sb == sb && cn->blocknr == blocknr && cn->jlist) {
1156 return cn->jlist;
1157 }
1158 cn = cn->hprev;
1159 }
1160 return NULL;
1161}
1162
1163static int newer_jl_done(struct reiserfs_journal_cnode *cn)
1164{
1165 struct super_block *sb = cn->sb;
1166 b_blocknr_t blocknr = cn->blocknr;
1167
1168 cn = cn->hprev;
1169 while (cn) {
1170 if (cn->sb == sb && cn->blocknr == blocknr && cn->jlist &&
1171 atomic_read(&cn->jlist->j_commit_left) != 0)
1172 return 0;
1173 cn = cn->hprev;
1174 }
1175 return 1;
1176}
1177
1178static void remove_journal_hash(struct super_block *,
1179 struct reiserfs_journal_cnode **,
1180 struct reiserfs_journal_list *, unsigned long,
1181 int);
1182
1183
1184
1185
1186
1187
1188static void remove_all_from_journal_list(struct super_block *sb,
1189 struct reiserfs_journal_list *jl,
1190 int debug)
1191{
1192 struct reiserfs_journal *journal = SB_JOURNAL(sb);
1193 struct reiserfs_journal_cnode *cn, *last;
1194 cn = jl->j_realblock;
1195
1196
1197
1198
1199 while (cn) {
1200 if (cn->blocknr != 0) {
1201 if (debug) {
1202 reiserfs_warning(sb, "reiserfs-2201",
1203 "block %u, bh is %d, state %ld",
1204 cn->blocknr, cn->bh ? 1 : 0,
1205 cn->state);
1206 }
1207 cn->state = 0;
1208 remove_journal_hash(sb, journal->j_list_hash_table,
1209 jl, cn->blocknr, 1);
1210 }
1211 last = cn;
1212 cn = cn->next;
1213 free_cnode(sb, last);
1214 }
1215 jl->j_realblock = NULL;
1216}
1217
1218
1219
1220
1221
1222
1223
1224
1225static int _update_journal_header_block(struct super_block *sb,
1226 unsigned long offset,
1227 unsigned int trans_id)
1228{
1229 struct reiserfs_journal_header *jh;
1230 struct reiserfs_journal *journal = SB_JOURNAL(sb);
1231
1232 if (reiserfs_is_journal_aborted(journal))
1233 return -EIO;
1234
1235 if (trans_id >= journal->j_last_flush_trans_id) {
1236 if (buffer_locked((journal->j_header_bh))) {
1237 reiserfs_write_unlock(sb);
1238 wait_on_buffer((journal->j_header_bh));
1239 reiserfs_write_lock(sb);
1240 if (unlikely(!buffer_uptodate(journal->j_header_bh))) {
1241#ifdef CONFIG_REISERFS_CHECK
1242 reiserfs_warning(sb, "journal-699",
1243 "buffer write failed");
1244#endif
1245 return -EIO;
1246 }
1247 }
1248 journal->j_last_flush_trans_id = trans_id;
1249 journal->j_first_unflushed_offset = offset;
1250 jh = (struct reiserfs_journal_header *)(journal->j_header_bh->
1251 b_data);
1252 jh->j_last_flush_trans_id = cpu_to_le32(trans_id);
1253 jh->j_first_unflushed_offset = cpu_to_le32(offset);
1254 jh->j_mount_id = cpu_to_le32(journal->j_mount_id);
1255
1256 set_buffer_dirty(journal->j_header_bh);
1257 reiserfs_write_unlock(sb);
1258
1259 if (reiserfs_barrier_flush(sb))
1260 __sync_dirty_buffer(journal->j_header_bh, WRITE_FLUSH_FUA);
1261 else
1262 sync_dirty_buffer(journal->j_header_bh);
1263
1264 reiserfs_write_lock(sb);
1265 if (!buffer_uptodate(journal->j_header_bh)) {
1266 reiserfs_warning(sb, "journal-837",
1267 "IO error during journal replay");
1268 return -EIO;
1269 }
1270 }
1271 return 0;
1272}
1273
1274static int update_journal_header_block(struct super_block *sb,
1275 unsigned long offset,
1276 unsigned int trans_id)
1277{
1278 return _update_journal_header_block(sb, offset, trans_id);
1279}
1280
1281
1282
1283
1284
1285static int flush_older_journal_lists(struct super_block *sb,
1286 struct reiserfs_journal_list *jl)
1287{
1288 struct list_head *entry;
1289 struct reiserfs_journal_list *other_jl;
1290 struct reiserfs_journal *journal = SB_JOURNAL(sb);
1291 unsigned int trans_id = jl->j_trans_id;
1292
1293
1294
1295
1296 restart:
1297 entry = journal->j_journal_list.next;
1298
1299 if (entry == &journal->j_journal_list)
1300 return 0;
1301 other_jl = JOURNAL_LIST_ENTRY(entry);
1302 if (other_jl->j_trans_id < trans_id) {
1303 BUG_ON(other_jl->j_refcount <= 0);
1304
1305 flush_journal_list(sb, other_jl, 0);
1306
1307
1308 goto restart;
1309 }
1310 return 0;
1311}
1312
1313static void del_from_work_list(struct super_block *s,
1314 struct reiserfs_journal_list *jl)
1315{
1316 struct reiserfs_journal *journal = SB_JOURNAL(s);
1317 if (!list_empty(&jl->j_working_list)) {
1318 list_del_init(&jl->j_working_list);
1319 journal->j_num_work_lists--;
1320 }
1321}
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332static int flush_journal_list(struct super_block *s,
1333 struct reiserfs_journal_list *jl, int flushall)
1334{
1335 struct reiserfs_journal_list *pjl;
1336 struct reiserfs_journal_cnode *cn, *last;
1337 int count;
1338 int was_jwait = 0;
1339 int was_dirty = 0;
1340 struct buffer_head *saved_bh;
1341 unsigned long j_len_saved = jl->j_len;
1342 struct reiserfs_journal *journal = SB_JOURNAL(s);
1343 int err = 0;
1344
1345 BUG_ON(j_len_saved <= 0);
1346
1347 if (atomic_read(&journal->j_wcount) != 0) {
1348 reiserfs_warning(s, "clm-2048", "called with wcount %d",
1349 atomic_read(&journal->j_wcount));
1350 }
1351 BUG_ON(jl->j_trans_id == 0);
1352
1353
1354 if (flushall) {
1355 reiserfs_mutex_lock_safe(&journal->j_flush_mutex, s);
1356 } else if (mutex_trylock(&journal->j_flush_mutex)) {
1357 BUG();
1358 }
1359
1360 count = 0;
1361 if (j_len_saved > journal->j_trans_max) {
1362 reiserfs_panic(s, "journal-715", "length is %lu, trans id %lu",
1363 j_len_saved, jl->j_trans_id);
1364 return 0;
1365 }
1366
1367
1368 if (atomic_read(&(jl->j_nonzerolen)) <= 0 &&
1369 atomic_read(&(jl->j_commit_left)) <= 0) {
1370 goto flush_older_and_return;
1371 }
1372
1373
1374
1375
1376 flush_commit_list(s, jl, 1);
1377
1378 if (!(jl->j_state & LIST_DIRTY)
1379 && !reiserfs_is_journal_aborted(journal))
1380 BUG();
1381
1382
1383 if (atomic_read(&(jl->j_nonzerolen)) <= 0 &&
1384 atomic_read(&(jl->j_commit_left)) <= 0) {
1385 goto flush_older_and_return;
1386 }
1387
1388
1389
1390
1391 if (atomic_read(&(journal->j_wcount)) != 0) {
1392 reiserfs_panic(s, "journal-844", "journal list is flushing, "
1393 "wcount is not 0");
1394 }
1395 cn = jl->j_realblock;
1396 while (cn) {
1397 was_jwait = 0;
1398 was_dirty = 0;
1399 saved_bh = NULL;
1400
1401 if (cn->blocknr == 0) {
1402 goto free_cnode;
1403 }
1404
1405
1406 if (!(jl->j_state & LIST_DIRTY))
1407 goto free_cnode;
1408
1409 pjl = find_newer_jl_for_cn(cn);
1410
1411
1412
1413
1414 if (!pjl && cn->bh) {
1415 saved_bh = cn->bh;
1416
1417
1418
1419
1420 get_bh(saved_bh);
1421
1422 if (buffer_journal_dirty(saved_bh)) {
1423 BUG_ON(!can_dirty(cn));
1424 was_jwait = 1;
1425 was_dirty = 1;
1426 } else if (can_dirty(cn)) {
1427
1428 BUG();
1429 }
1430 }
1431
1432
1433
1434
1435 if (pjl) {
1436 if (atomic_read(&pjl->j_commit_left))
1437 flush_commit_list(s, pjl, 1);
1438 goto free_cnode;
1439 }
1440
1441
1442
1443
1444 if (saved_bh == NULL) {
1445 goto free_cnode;
1446 }
1447
1448
1449
1450
1451
1452 if ((!was_jwait) && !buffer_locked(saved_bh)) {
1453 reiserfs_warning(s, "journal-813",
1454 "BAD! buffer %llu %cdirty %cjwait, "
1455 "not in a newer tranasction",
1456 (unsigned long long)saved_bh->
1457 b_blocknr, was_dirty ? ' ' : '!',
1458 was_jwait ? ' ' : '!');
1459 }
1460 if (was_dirty) {
1461
1462 get_bh(saved_bh);
1463 set_bit(BLOCK_NEEDS_FLUSH, &cn->state);
1464 lock_buffer(saved_bh);
1465 BUG_ON(cn->blocknr != saved_bh->b_blocknr);
1466 if (buffer_dirty(saved_bh))
1467 submit_logged_buffer(saved_bh);
1468 else
1469 unlock_buffer(saved_bh);
1470 count++;
1471 } else {
1472 reiserfs_warning(s, "clm-2082",
1473 "Unable to flush buffer %llu in %s",
1474 (unsigned long long)saved_bh->
1475 b_blocknr, __func__);
1476 }
1477 free_cnode:
1478 last = cn;
1479 cn = cn->next;
1480 if (saved_bh) {
1481
1482 put_bh(saved_bh);
1483 if (atomic_read(&(saved_bh->b_count)) < 0) {
1484 reiserfs_warning(s, "journal-945",
1485 "saved_bh->b_count < 0");
1486 }
1487 }
1488 }
1489 if (count > 0) {
1490 cn = jl->j_realblock;
1491 while (cn) {
1492 if (test_bit(BLOCK_NEEDS_FLUSH, &cn->state)) {
1493 if (!cn->bh) {
1494 reiserfs_panic(s, "journal-1011",
1495 "cn->bh is NULL");
1496 }
1497
1498 reiserfs_write_unlock(s);
1499 wait_on_buffer(cn->bh);
1500 reiserfs_write_lock(s);
1501
1502 if (!cn->bh) {
1503 reiserfs_panic(s, "journal-1012",
1504 "cn->bh is NULL");
1505 }
1506 if (unlikely(!buffer_uptodate(cn->bh))) {
1507#ifdef CONFIG_REISERFS_CHECK
1508 reiserfs_warning(s, "journal-949",
1509 "buffer write failed");
1510#endif
1511 err = -EIO;
1512 }
1513
1514
1515
1516 BUG_ON(!test_clear_buffer_journal_dirty
1517 (cn->bh));
1518
1519
1520 put_bh(cn->bh);
1521
1522 release_buffer_page(cn->bh);
1523 }
1524 cn = cn->next;
1525 }
1526 }
1527
1528 if (err)
1529 reiserfs_abort(s, -EIO,
1530 "Write error while pushing transaction to disk in %s",
1531 __func__);
1532 flush_older_and_return:
1533
1534
1535
1536
1537
1538
1539 if (flushall) {
1540 flush_older_journal_lists(s, jl);
1541 }
1542
1543 err = journal->j_errno;
1544
1545
1546
1547
1548
1549
1550
1551
1552 if (!err && flushall) {
1553 err =
1554 update_journal_header_block(s,
1555 (jl->j_start + jl->j_len +
1556 2) % SB_ONDISK_JOURNAL_SIZE(s),
1557 jl->j_trans_id);
1558 if (err)
1559 reiserfs_abort(s, -EIO,
1560 "Write error while updating journal header in %s",
1561 __func__);
1562 }
1563 remove_all_from_journal_list(s, jl, 0);
1564 list_del_init(&jl->j_list);
1565 journal->j_num_lists--;
1566 del_from_work_list(s, jl);
1567
1568 if (journal->j_last_flush_id != 0 &&
1569 (jl->j_trans_id - journal->j_last_flush_id) != 1) {
1570 reiserfs_warning(s, "clm-2201", "last flush %lu, current %lu",
1571 journal->j_last_flush_id, jl->j_trans_id);
1572 }
1573 journal->j_last_flush_id = jl->j_trans_id;
1574
1575
1576
1577
1578 jl->j_len = 0;
1579 atomic_set(&(jl->j_nonzerolen), 0);
1580 jl->j_start = 0;
1581 jl->j_realblock = NULL;
1582 jl->j_commit_bh = NULL;
1583 jl->j_trans_id = 0;
1584 jl->j_state = 0;
1585 put_journal_list(s, jl);
1586 if (flushall)
1587 mutex_unlock(&journal->j_flush_mutex);
1588 return err;
1589}
1590
1591static int test_transaction(struct super_block *s,
1592 struct reiserfs_journal_list *jl)
1593{
1594 struct reiserfs_journal_cnode *cn;
1595
1596 if (jl->j_len == 0 || atomic_read(&jl->j_nonzerolen) == 0)
1597 return 1;
1598
1599 cn = jl->j_realblock;
1600 while (cn) {
1601
1602
1603
1604 if (cn->blocknr == 0) {
1605 goto next;
1606 }
1607 if (cn->bh && !newer_jl_done(cn))
1608 return 0;
1609 next:
1610 cn = cn->next;
1611 cond_resched();
1612 }
1613 return 0;
1614}
1615
1616static int write_one_transaction(struct super_block *s,
1617 struct reiserfs_journal_list *jl,
1618 struct buffer_chunk *chunk)
1619{
1620 struct reiserfs_journal_cnode *cn;
1621 int ret = 0;
1622
1623 jl->j_state |= LIST_TOUCHED;
1624 del_from_work_list(s, jl);
1625 if (jl->j_len == 0 || atomic_read(&jl->j_nonzerolen) == 0) {
1626 return 0;
1627 }
1628
1629 cn = jl->j_realblock;
1630 while (cn) {
1631
1632
1633
1634 if (cn->blocknr == 0) {
1635 goto next;
1636 }
1637 if (cn->bh && can_dirty(cn) && buffer_dirty(cn->bh)) {
1638 struct buffer_head *tmp_bh;
1639
1640
1641
1642
1643 tmp_bh = cn->bh;
1644 get_bh(tmp_bh);
1645 lock_buffer(tmp_bh);
1646 if (cn->bh && can_dirty(cn) && buffer_dirty(tmp_bh)) {
1647 if (!buffer_journal_dirty(tmp_bh) ||
1648 buffer_journal_prepared(tmp_bh))
1649 BUG();
1650 add_to_chunk(chunk, tmp_bh, NULL, write_chunk);
1651 ret++;
1652 } else {
1653
1654 unlock_buffer(tmp_bh);
1655 }
1656 put_bh(tmp_bh);
1657 }
1658 next:
1659 cn = cn->next;
1660 cond_resched();
1661 }
1662 return ret;
1663}
1664
1665
1666static int dirty_one_transaction(struct super_block *s,
1667 struct reiserfs_journal_list *jl)
1668{
1669 struct reiserfs_journal_cnode *cn;
1670 struct reiserfs_journal_list *pjl;
1671 int ret = 0;
1672
1673 jl->j_state |= LIST_DIRTY;
1674 cn = jl->j_realblock;
1675 while (cn) {
1676
1677
1678
1679
1680 pjl = find_newer_jl_for_cn(cn);
1681 if (!pjl && cn->blocknr && cn->bh
1682 && buffer_journal_dirty(cn->bh)) {
1683 BUG_ON(!can_dirty(cn));
1684
1685
1686
1687
1688 clear_buffer_journal_new(cn->bh);
1689 if (buffer_journal_prepared(cn->bh)) {
1690 set_buffer_journal_restore_dirty(cn->bh);
1691 } else {
1692 set_buffer_journal_test(cn->bh);
1693 mark_buffer_dirty(cn->bh);
1694 }
1695 }
1696 cn = cn->next;
1697 }
1698 return ret;
1699}
1700
1701static int kupdate_transactions(struct super_block *s,
1702 struct reiserfs_journal_list *jl,
1703 struct reiserfs_journal_list **next_jl,
1704 unsigned int *next_trans_id,
1705 int num_blocks, int num_trans)
1706{
1707 int ret = 0;
1708 int written = 0;
1709 int transactions_flushed = 0;
1710 unsigned int orig_trans_id = jl->j_trans_id;
1711 struct buffer_chunk chunk;
1712 struct list_head *entry;
1713 struct reiserfs_journal *journal = SB_JOURNAL(s);
1714 chunk.nr = 0;
1715
1716 reiserfs_mutex_lock_safe(&journal->j_flush_mutex, s);
1717 if (!journal_list_still_alive(s, orig_trans_id)) {
1718 goto done;
1719 }
1720
1721
1722
1723
1724 while ((num_trans && transactions_flushed < num_trans) ||
1725 (!num_trans && written < num_blocks)) {
1726
1727 if (jl->j_len == 0 || (jl->j_state & LIST_TOUCHED) ||
1728 atomic_read(&jl->j_commit_left)
1729 || !(jl->j_state & LIST_DIRTY)) {
1730 del_from_work_list(s, jl);
1731 break;
1732 }
1733 ret = write_one_transaction(s, jl, &chunk);
1734
1735 if (ret < 0)
1736 goto done;
1737 transactions_flushed++;
1738 written += ret;
1739 entry = jl->j_list.next;
1740
1741
1742 if (entry == &journal->j_journal_list) {
1743 break;
1744 }
1745 jl = JOURNAL_LIST_ENTRY(entry);
1746
1747
1748 if (jl->j_trans_id <= orig_trans_id)
1749 break;
1750 }
1751 if (chunk.nr) {
1752 write_chunk(&chunk);
1753 }
1754
1755 done:
1756 mutex_unlock(&journal->j_flush_mutex);
1757 return ret;
1758}
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769static int flush_used_journal_lists(struct super_block *s,
1770 struct reiserfs_journal_list *jl)
1771{
1772 unsigned long len = 0;
1773 unsigned long cur_len;
1774 int ret;
1775 int i;
1776 int limit = 256;
1777 struct reiserfs_journal_list *tjl;
1778 struct reiserfs_journal_list *flush_jl;
1779 unsigned int trans_id;
1780 struct reiserfs_journal *journal = SB_JOURNAL(s);
1781
1782 flush_jl = tjl = jl;
1783
1784
1785 if (reiserfs_data_log(s))
1786 limit = 1024;
1787
1788 for (i = 0; i < 256 && len < limit; i++) {
1789 if (atomic_read(&tjl->j_commit_left) ||
1790 tjl->j_trans_id < jl->j_trans_id) {
1791 break;
1792 }
1793 cur_len = atomic_read(&tjl->j_nonzerolen);
1794 if (cur_len > 0) {
1795 tjl->j_state &= ~LIST_TOUCHED;
1796 }
1797 len += cur_len;
1798 flush_jl = tjl;
1799 if (tjl->j_list.next == &journal->j_journal_list)
1800 break;
1801 tjl = JOURNAL_LIST_ENTRY(tjl->j_list.next);
1802 }
1803
1804
1805
1806
1807 if (flush_jl != jl) {
1808 ret = kupdate_transactions(s, jl, &tjl, &trans_id, len, i);
1809 }
1810 flush_journal_list(s, flush_jl, 1);
1811 return 0;
1812}
1813
1814
1815
1816
1817
1818void remove_journal_hash(struct super_block *sb,
1819 struct reiserfs_journal_cnode **table,
1820 struct reiserfs_journal_list *jl,
1821 unsigned long block, int remove_freed)
1822{
1823 struct reiserfs_journal_cnode *cur;
1824 struct reiserfs_journal_cnode **head;
1825
1826 head = &(journal_hash(table, sb, block));
1827 if (!head) {
1828 return;
1829 }
1830 cur = *head;
1831 while (cur) {
1832 if (cur->blocknr == block && cur->sb == sb
1833 && (jl == NULL || jl == cur->jlist)
1834 && (!test_bit(BLOCK_FREED, &cur->state) || remove_freed)) {
1835 if (cur->hnext) {
1836 cur->hnext->hprev = cur->hprev;
1837 }
1838 if (cur->hprev) {
1839 cur->hprev->hnext = cur->hnext;
1840 } else {
1841 *head = cur->hnext;
1842 }
1843 cur->blocknr = 0;
1844 cur->sb = NULL;
1845 cur->state = 0;
1846 if (cur->bh && cur->jlist)
1847 atomic_dec(&(cur->jlist->j_nonzerolen));
1848 cur->bh = NULL;
1849 cur->jlist = NULL;
1850 }
1851 cur = cur->hnext;
1852 }
1853}
1854
1855static void free_journal_ram(struct super_block *sb)
1856{
1857 struct reiserfs_journal *journal = SB_JOURNAL(sb);
1858 kfree(journal->j_current_jl);
1859 journal->j_num_lists--;
1860
1861 vfree(journal->j_cnode_free_orig);
1862 free_list_bitmaps(sb, journal->j_list_bitmap);
1863 free_bitmap_nodes(sb);
1864 if (journal->j_header_bh) {
1865 brelse(journal->j_header_bh);
1866 }
1867
1868
1869
1870 release_journal_dev(sb, journal);
1871 vfree(journal);
1872}
1873
1874
1875
1876
1877
1878static int do_journal_release(struct reiserfs_transaction_handle *th,
1879 struct super_block *sb, int error)
1880{
1881 struct reiserfs_transaction_handle myth;
1882 int flushed = 0;
1883 struct reiserfs_journal *journal = SB_JOURNAL(sb);
1884
1885
1886
1887 if (!error && !(sb->s_flags & MS_RDONLY)) {
1888
1889 BUG_ON(!th->t_trans_id);
1890 do_journal_end(th, sb, 10, FLUSH_ALL);
1891
1892
1893 if (!journal_join(&myth, sb, 1)) {
1894 reiserfs_prepare_for_journal(sb,
1895 SB_BUFFER_WITH_SB(sb),
1896 1);
1897 journal_mark_dirty(&myth, sb,
1898 SB_BUFFER_WITH_SB(sb));
1899 do_journal_end(&myth, sb, 1, FLUSH_ALL);
1900 flushed = 1;
1901 }
1902 }
1903
1904
1905 if (!error && reiserfs_is_journal_aborted(journal)) {
1906 memset(&myth, 0, sizeof(myth));
1907 if (!journal_join_abort(&myth, sb, 1)) {
1908 reiserfs_prepare_for_journal(sb,
1909 SB_BUFFER_WITH_SB(sb),
1910 1);
1911 journal_mark_dirty(&myth, sb,
1912 SB_BUFFER_WITH_SB(sb));
1913 do_journal_end(&myth, sb, 1, FLUSH_ALL);
1914 }
1915 }
1916
1917 reiserfs_mounted_fs_count--;
1918
1919 cancel_delayed_work(&SB_JOURNAL(sb)->j_work);
1920
1921
1922
1923
1924
1925 reiserfs_write_unlock(sb);
1926
1927 cancel_delayed_work_sync(&REISERFS_SB(sb)->old_work);
1928 flush_workqueue(commit_wq);
1929
1930 if (!reiserfs_mounted_fs_count) {
1931 destroy_workqueue(commit_wq);
1932 commit_wq = NULL;
1933 }
1934
1935 free_journal_ram(sb);
1936
1937 reiserfs_write_lock(sb);
1938
1939 return 0;
1940}
1941
1942
1943
1944
1945int journal_release(struct reiserfs_transaction_handle *th,
1946 struct super_block *sb)
1947{
1948 return do_journal_release(th, sb, 0);
1949}
1950
1951
1952
1953
1954int journal_release_error(struct reiserfs_transaction_handle *th,
1955 struct super_block *sb)
1956{
1957 return do_journal_release(th, sb, 1);
1958}
1959
1960
1961static int journal_compare_desc_commit(struct super_block *sb,
1962 struct reiserfs_journal_desc *desc,
1963 struct reiserfs_journal_commit *commit)
1964{
1965 if (get_commit_trans_id(commit) != get_desc_trans_id(desc) ||
1966 get_commit_trans_len(commit) != get_desc_trans_len(desc) ||
1967 get_commit_trans_len(commit) > SB_JOURNAL(sb)->j_trans_max ||
1968 get_commit_trans_len(commit) <= 0) {
1969 return 1;
1970 }
1971 return 0;
1972}
1973
1974
1975
1976
1977
1978static int journal_transaction_is_valid(struct super_block *sb,
1979 struct buffer_head *d_bh,
1980 unsigned int *oldest_invalid_trans_id,
1981 unsigned long *newest_mount_id)
1982{
1983 struct reiserfs_journal_desc *desc;
1984 struct reiserfs_journal_commit *commit;
1985 struct buffer_head *c_bh;
1986 unsigned long offset;
1987
1988 if (!d_bh)
1989 return 0;
1990
1991 desc = (struct reiserfs_journal_desc *)d_bh->b_data;
1992 if (get_desc_trans_len(desc) > 0
1993 && !memcmp(get_journal_desc_magic(d_bh), JOURNAL_DESC_MAGIC, 8)) {
1994 if (oldest_invalid_trans_id && *oldest_invalid_trans_id
1995 && get_desc_trans_id(desc) > *oldest_invalid_trans_id) {
1996 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
1997 "journal-986: transaction "
1998 "is valid returning because trans_id %d is greater than "
1999 "oldest_invalid %lu",
2000 get_desc_trans_id(desc),
2001 *oldest_invalid_trans_id);
2002 return 0;
2003 }
2004 if (newest_mount_id
2005 && *newest_mount_id > get_desc_mount_id(desc)) {
2006 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2007 "journal-1087: transaction "
2008 "is valid returning because mount_id %d is less than "
2009 "newest_mount_id %lu",
2010 get_desc_mount_id(desc),
2011 *newest_mount_id);
2012 return -1;
2013 }
2014 if (get_desc_trans_len(desc) > SB_JOURNAL(sb)->j_trans_max) {
2015 reiserfs_warning(sb, "journal-2018",
2016 "Bad transaction length %d "
2017 "encountered, ignoring transaction",
2018 get_desc_trans_len(desc));
2019 return -1;
2020 }
2021 offset = d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(sb);
2022
2023
2024 c_bh =
2025 journal_bread(sb,
2026 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2027 ((offset + get_desc_trans_len(desc) +
2028 1) % SB_ONDISK_JOURNAL_SIZE(sb)));
2029 if (!c_bh)
2030 return 0;
2031 commit = (struct reiserfs_journal_commit *)c_bh->b_data;
2032 if (journal_compare_desc_commit(sb, desc, commit)) {
2033 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2034 "journal_transaction_is_valid, commit offset %ld had bad "
2035 "time %d or length %d",
2036 c_bh->b_blocknr -
2037 SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2038 get_commit_trans_id(commit),
2039 get_commit_trans_len(commit));
2040 brelse(c_bh);
2041 if (oldest_invalid_trans_id) {
2042 *oldest_invalid_trans_id =
2043 get_desc_trans_id(desc);
2044 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2045 "journal-1004: "
2046 "transaction_is_valid setting oldest invalid trans_id "
2047 "to %d",
2048 get_desc_trans_id(desc));
2049 }
2050 return -1;
2051 }
2052 brelse(c_bh);
2053 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2054 "journal-1006: found valid "
2055 "transaction start offset %llu, len %d id %d",
2056 d_bh->b_blocknr -
2057 SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2058 get_desc_trans_len(desc),
2059 get_desc_trans_id(desc));
2060 return 1;
2061 } else {
2062 return 0;
2063 }
2064}
2065
2066static void brelse_array(struct buffer_head **heads, int num)
2067{
2068 int i;
2069 for (i = 0; i < num; i++) {
2070 brelse(heads[i]);
2071 }
2072}
2073
2074
2075
2076
2077
2078
2079static int journal_read_transaction(struct super_block *sb,
2080 unsigned long cur_dblock,
2081 unsigned long oldest_start,
2082 unsigned int oldest_trans_id,
2083 unsigned long newest_mount_id)
2084{
2085 struct reiserfs_journal *journal = SB_JOURNAL(sb);
2086 struct reiserfs_journal_desc *desc;
2087 struct reiserfs_journal_commit *commit;
2088 unsigned int trans_id = 0;
2089 struct buffer_head *c_bh;
2090 struct buffer_head *d_bh;
2091 struct buffer_head **log_blocks = NULL;
2092 struct buffer_head **real_blocks = NULL;
2093 unsigned int trans_offset;
2094 int i;
2095 int trans_half;
2096
2097 d_bh = journal_bread(sb, cur_dblock);
2098 if (!d_bh)
2099 return 1;
2100 desc = (struct reiserfs_journal_desc *)d_bh->b_data;
2101 trans_offset = d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(sb);
2102 reiserfs_debug(sb, REISERFS_DEBUG_CODE, "journal-1037: "
2103 "journal_read_transaction, offset %llu, len %d mount_id %d",
2104 d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2105 get_desc_trans_len(desc), get_desc_mount_id(desc));
2106 if (get_desc_trans_id(desc) < oldest_trans_id) {
2107 reiserfs_debug(sb, REISERFS_DEBUG_CODE, "journal-1039: "
2108 "journal_read_trans skipping because %lu is too old",
2109 cur_dblock -
2110 SB_ONDISK_JOURNAL_1st_BLOCK(sb));
2111 brelse(d_bh);
2112 return 1;
2113 }
2114 if (get_desc_mount_id(desc) != newest_mount_id) {
2115 reiserfs_debug(sb, REISERFS_DEBUG_CODE, "journal-1146: "
2116 "journal_read_trans skipping because %d is != "
2117 "newest_mount_id %lu", get_desc_mount_id(desc),
2118 newest_mount_id);
2119 brelse(d_bh);
2120 return 1;
2121 }
2122 c_bh = journal_bread(sb, SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2123 ((trans_offset + get_desc_trans_len(desc) + 1) %
2124 SB_ONDISK_JOURNAL_SIZE(sb)));
2125 if (!c_bh) {
2126 brelse(d_bh);
2127 return 1;
2128 }
2129 commit = (struct reiserfs_journal_commit *)c_bh->b_data;
2130 if (journal_compare_desc_commit(sb, desc, commit)) {
2131 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2132 "journal_read_transaction, "
2133 "commit offset %llu had bad time %d or length %d",
2134 c_bh->b_blocknr -
2135 SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2136 get_commit_trans_id(commit),
2137 get_commit_trans_len(commit));
2138 brelse(c_bh);
2139 brelse(d_bh);
2140 return 1;
2141 }
2142
2143 if (bdev_read_only(sb->s_bdev)) {
2144 reiserfs_warning(sb, "clm-2076",
2145 "device is readonly, unable to replay log");
2146 brelse(c_bh);
2147 brelse(d_bh);
2148 return -EROFS;
2149 }
2150
2151 trans_id = get_desc_trans_id(desc);
2152
2153 log_blocks = kmalloc(get_desc_trans_len(desc) *
2154 sizeof(struct buffer_head *), GFP_NOFS);
2155 real_blocks = kmalloc(get_desc_trans_len(desc) *
2156 sizeof(struct buffer_head *), GFP_NOFS);
2157 if (!log_blocks || !real_blocks) {
2158 brelse(c_bh);
2159 brelse(d_bh);
2160 kfree(log_blocks);
2161 kfree(real_blocks);
2162 reiserfs_warning(sb, "journal-1169",
2163 "kmalloc failed, unable to mount FS");
2164 return -1;
2165 }
2166
2167 trans_half = journal_trans_half(sb->s_blocksize);
2168 for (i = 0; i < get_desc_trans_len(desc); i++) {
2169 log_blocks[i] =
2170 journal_getblk(sb,
2171 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2172 (trans_offset + 1 +
2173 i) % SB_ONDISK_JOURNAL_SIZE(sb));
2174 if (i < trans_half) {
2175 real_blocks[i] =
2176 sb_getblk(sb,
2177 le32_to_cpu(desc->j_realblock[i]));
2178 } else {
2179 real_blocks[i] =
2180 sb_getblk(sb,
2181 le32_to_cpu(commit->
2182 j_realblock[i - trans_half]));
2183 }
2184 if (real_blocks[i]->b_blocknr > SB_BLOCK_COUNT(sb)) {
2185 reiserfs_warning(sb, "journal-1207",
2186 "REPLAY FAILURE fsck required! "
2187 "Block to replay is outside of "
2188 "filesystem");
2189 goto abort_replay;
2190 }
2191
2192 if (is_block_in_log_or_reserved_area
2193 (sb, real_blocks[i]->b_blocknr)) {
2194 reiserfs_warning(sb, "journal-1204",
2195 "REPLAY FAILURE fsck required! "
2196 "Trying to replay onto a log block");
2197 abort_replay:
2198 brelse_array(log_blocks, i);
2199 brelse_array(real_blocks, i);
2200 brelse(c_bh);
2201 brelse(d_bh);
2202 kfree(log_blocks);
2203 kfree(real_blocks);
2204 return -1;
2205 }
2206 }
2207
2208 ll_rw_block(READ, get_desc_trans_len(desc), log_blocks);
2209 for (i = 0; i < get_desc_trans_len(desc); i++) {
2210
2211 reiserfs_write_unlock(sb);
2212 wait_on_buffer(log_blocks[i]);
2213 reiserfs_write_lock(sb);
2214
2215 if (!buffer_uptodate(log_blocks[i])) {
2216 reiserfs_warning(sb, "journal-1212",
2217 "REPLAY FAILURE fsck required! "
2218 "buffer write failed");
2219 brelse_array(log_blocks + i,
2220 get_desc_trans_len(desc) - i);
2221 brelse_array(real_blocks, get_desc_trans_len(desc));
2222 brelse(c_bh);
2223 brelse(d_bh);
2224 kfree(log_blocks);
2225 kfree(real_blocks);
2226 return -1;
2227 }
2228 memcpy(real_blocks[i]->b_data, log_blocks[i]->b_data,
2229 real_blocks[i]->b_size);
2230 set_buffer_uptodate(real_blocks[i]);
2231 brelse(log_blocks[i]);
2232 }
2233
2234 for (i = 0; i < get_desc_trans_len(desc); i++) {
2235 set_buffer_dirty(real_blocks[i]);
2236 write_dirty_buffer(real_blocks[i], WRITE);
2237 }
2238 for (i = 0; i < get_desc_trans_len(desc); i++) {
2239 wait_on_buffer(real_blocks[i]);
2240 if (!buffer_uptodate(real_blocks[i])) {
2241 reiserfs_warning(sb, "journal-1226",
2242 "REPLAY FAILURE, fsck required! "
2243 "buffer write failed");
2244 brelse_array(real_blocks + i,
2245 get_desc_trans_len(desc) - i);
2246 brelse(c_bh);
2247 brelse(d_bh);
2248 kfree(log_blocks);
2249 kfree(real_blocks);
2250 return -1;
2251 }
2252 brelse(real_blocks[i]);
2253 }
2254 cur_dblock =
2255 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2256 ((trans_offset + get_desc_trans_len(desc) +
2257 2) % SB_ONDISK_JOURNAL_SIZE(sb));
2258 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2259 "journal-1095: setting journal " "start to offset %ld",
2260 cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(sb));
2261
2262
2263 journal->j_start = cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(sb);
2264 journal->j_last_flush_trans_id = trans_id;
2265 journal->j_trans_id = trans_id + 1;
2266
2267 if (journal->j_trans_id == 0)
2268 journal->j_trans_id = 10;
2269 brelse(c_bh);
2270 brelse(d_bh);
2271 kfree(log_blocks);
2272 kfree(real_blocks);
2273 return 0;
2274}
2275
2276
2277
2278
2279
2280
2281
2282static struct buffer_head *reiserfs_breada(struct block_device *dev,
2283 b_blocknr_t block, int bufsize,
2284 b_blocknr_t max_block)
2285{
2286 struct buffer_head *bhlist[BUFNR];
2287 unsigned int blocks = BUFNR;
2288 struct buffer_head *bh;
2289 int i, j;
2290
2291 bh = __getblk(dev, block, bufsize);
2292 if (buffer_uptodate(bh))
2293 return (bh);
2294
2295 if (block + BUFNR > max_block) {
2296 blocks = max_block - block;
2297 }
2298 bhlist[0] = bh;
2299 j = 1;
2300 for (i = 1; i < blocks; i++) {
2301 bh = __getblk(dev, block + i, bufsize);
2302 if (buffer_uptodate(bh)) {
2303 brelse(bh);
2304 break;
2305 } else
2306 bhlist[j++] = bh;
2307 }
2308 ll_rw_block(READ, j, bhlist);
2309 for (i = 1; i < j; i++)
2310 brelse(bhlist[i]);
2311 bh = bhlist[0];
2312 wait_on_buffer(bh);
2313 if (buffer_uptodate(bh))
2314 return bh;
2315 brelse(bh);
2316 return NULL;
2317}
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328static int journal_read(struct super_block *sb)
2329{
2330 struct reiserfs_journal *journal = SB_JOURNAL(sb);
2331 struct reiserfs_journal_desc *desc;
2332 unsigned int oldest_trans_id = 0;
2333 unsigned int oldest_invalid_trans_id = 0;
2334 time_t start;
2335 unsigned long oldest_start = 0;
2336 unsigned long cur_dblock = 0;
2337 unsigned long newest_mount_id = 9;
2338 struct buffer_head *d_bh;
2339 struct reiserfs_journal_header *jh;
2340 int valid_journal_header = 0;
2341 int replay_count = 0;
2342 int continue_replay = 1;
2343 int ret;
2344 char b[BDEVNAME_SIZE];
2345
2346 cur_dblock = SB_ONDISK_JOURNAL_1st_BLOCK(sb);
2347 reiserfs_info(sb, "checking transaction log (%s)\n",
2348 bdevname(journal->j_dev_bd, b));
2349 start = get_seconds();
2350
2351
2352
2353
2354
2355 journal->j_header_bh = journal_bread(sb,
2356 SB_ONDISK_JOURNAL_1st_BLOCK(sb)
2357 + SB_ONDISK_JOURNAL_SIZE(sb));
2358 if (!journal->j_header_bh) {
2359 return 1;
2360 }
2361 jh = (struct reiserfs_journal_header *)(journal->j_header_bh->b_data);
2362 if (le32_to_cpu(jh->j_first_unflushed_offset) <
2363 SB_ONDISK_JOURNAL_SIZE(sb)
2364 && le32_to_cpu(jh->j_last_flush_trans_id) > 0) {
2365 oldest_start =
2366 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2367 le32_to_cpu(jh->j_first_unflushed_offset);
2368 oldest_trans_id = le32_to_cpu(jh->j_last_flush_trans_id) + 1;
2369 newest_mount_id = le32_to_cpu(jh->j_mount_id);
2370 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2371 "journal-1153: found in "
2372 "header: first_unflushed_offset %d, last_flushed_trans_id "
2373 "%lu", le32_to_cpu(jh->j_first_unflushed_offset),
2374 le32_to_cpu(jh->j_last_flush_trans_id));
2375 valid_journal_header = 1;
2376
2377
2378
2379
2380
2381 d_bh =
2382 journal_bread(sb,
2383 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2384 le32_to_cpu(jh->j_first_unflushed_offset));
2385 ret = journal_transaction_is_valid(sb, d_bh, NULL, NULL);
2386 if (!ret) {
2387 continue_replay = 0;
2388 }
2389 brelse(d_bh);
2390 goto start_log_replay;
2391 }
2392
2393
2394
2395
2396 while (continue_replay
2397 && cur_dblock <
2398 (SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2399 SB_ONDISK_JOURNAL_SIZE(sb))) {
2400
2401
2402 d_bh =
2403 reiserfs_breada(journal->j_dev_bd, cur_dblock,
2404 sb->s_blocksize,
2405 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2406 SB_ONDISK_JOURNAL_SIZE(sb));
2407 ret =
2408 journal_transaction_is_valid(sb, d_bh,
2409 &oldest_invalid_trans_id,
2410 &newest_mount_id);
2411 if (ret == 1) {
2412 desc = (struct reiserfs_journal_desc *)d_bh->b_data;
2413 if (oldest_start == 0) {
2414 oldest_trans_id = get_desc_trans_id(desc);
2415 oldest_start = d_bh->b_blocknr;
2416 newest_mount_id = get_desc_mount_id(desc);
2417 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2418 "journal-1179: Setting "
2419 "oldest_start to offset %llu, trans_id %lu",
2420 oldest_start -
2421 SB_ONDISK_JOURNAL_1st_BLOCK
2422 (sb), oldest_trans_id);
2423 } else if (oldest_trans_id > get_desc_trans_id(desc)) {
2424
2425 oldest_trans_id = get_desc_trans_id(desc);
2426 oldest_start = d_bh->b_blocknr;
2427 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2428 "journal-1180: Resetting "
2429 "oldest_start to offset %lu, trans_id %lu",
2430 oldest_start -
2431 SB_ONDISK_JOURNAL_1st_BLOCK
2432 (sb), oldest_trans_id);
2433 }
2434 if (newest_mount_id < get_desc_mount_id(desc)) {
2435 newest_mount_id = get_desc_mount_id(desc);
2436 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2437 "journal-1299: Setting "
2438 "newest_mount_id to %d",
2439 get_desc_mount_id(desc));
2440 }
2441 cur_dblock += get_desc_trans_len(desc) + 2;
2442 } else {
2443 cur_dblock++;
2444 }
2445 brelse(d_bh);
2446 }
2447
2448 start_log_replay:
2449 cur_dblock = oldest_start;
2450 if (oldest_trans_id) {
2451 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2452 "journal-1206: Starting replay "
2453 "from offset %llu, trans_id %lu",
2454 cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2455 oldest_trans_id);
2456
2457 }
2458 replay_count = 0;
2459 while (continue_replay && oldest_trans_id > 0) {
2460 ret =
2461 journal_read_transaction(sb, cur_dblock, oldest_start,
2462 oldest_trans_id, newest_mount_id);
2463 if (ret < 0) {
2464 return ret;
2465 } else if (ret != 0) {
2466 break;
2467 }
2468 cur_dblock =
2469 SB_ONDISK_JOURNAL_1st_BLOCK(sb) + journal->j_start;
2470 replay_count++;
2471 if (cur_dblock == oldest_start)
2472 break;
2473 }
2474
2475 if (oldest_trans_id == 0) {
2476 reiserfs_debug(sb, REISERFS_DEBUG_CODE,
2477 "journal-1225: No valid " "transactions found");
2478 }
2479
2480
2481
2482
2483 if (valid_journal_header && replay_count == 0) {
2484 journal->j_start = le32_to_cpu(jh->j_first_unflushed_offset);
2485 journal->j_trans_id =
2486 le32_to_cpu(jh->j_last_flush_trans_id) + 1;
2487
2488 if (journal->j_trans_id == 0)
2489 journal->j_trans_id = 10;
2490 journal->j_last_flush_trans_id =
2491 le32_to_cpu(jh->j_last_flush_trans_id);
2492 journal->j_mount_id = le32_to_cpu(jh->j_mount_id) + 1;
2493 } else {
2494 journal->j_mount_id = newest_mount_id + 1;
2495 }
2496 reiserfs_debug(sb, REISERFS_DEBUG_CODE, "journal-1299: Setting "
2497 "newest_mount_id to %lu", journal->j_mount_id);
2498 journal->j_first_unflushed_offset = journal->j_start;
2499 if (replay_count > 0) {
2500 reiserfs_info(sb,
2501 "replayed %d transactions in %lu seconds\n",
2502 replay_count, get_seconds() - start);
2503 }
2504 if (!bdev_read_only(sb->s_bdev) &&
2505 _update_journal_header_block(sb, journal->j_start,
2506 journal->j_last_flush_trans_id)) {
2507
2508
2509
2510 return -1;
2511 }
2512 return 0;
2513}
2514
2515static struct reiserfs_journal_list *alloc_journal_list(struct super_block *s)
2516{
2517 struct reiserfs_journal_list *jl;
2518 jl = kzalloc(sizeof(struct reiserfs_journal_list),
2519 GFP_NOFS | __GFP_NOFAIL);
2520 INIT_LIST_HEAD(&jl->j_list);
2521 INIT_LIST_HEAD(&jl->j_working_list);
2522 INIT_LIST_HEAD(&jl->j_tail_bh_list);
2523 INIT_LIST_HEAD(&jl->j_bh_list);
2524 mutex_init(&jl->j_commit_mutex);
2525 SB_JOURNAL(s)->j_num_lists++;
2526 get_journal_list(jl);
2527 return jl;
2528}
2529
2530static void journal_list_init(struct super_block *sb)
2531{
2532 SB_JOURNAL(sb)->j_current_jl = alloc_journal_list(sb);
2533}
2534
2535static void release_journal_dev(struct super_block *super,
2536 struct reiserfs_journal *journal)
2537{
2538 if (journal->j_dev_bd != NULL) {
2539 blkdev_put(journal->j_dev_bd, journal->j_dev_mode);
2540 journal->j_dev_bd = NULL;
2541 }
2542}
2543
2544static int journal_init_dev(struct super_block *super,
2545 struct reiserfs_journal *journal,
2546 const char *jdev_name)
2547{
2548 int result;
2549 dev_t jdev;
2550 fmode_t blkdev_mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL;
2551 char b[BDEVNAME_SIZE];
2552
2553 result = 0;
2554
2555 journal->j_dev_bd = NULL;
2556 jdev = SB_ONDISK_JOURNAL_DEVICE(super) ?
2557 new_decode_dev(SB_ONDISK_JOURNAL_DEVICE(super)) : super->s_dev;
2558
2559 if (bdev_read_only(super->s_bdev))
2560 blkdev_mode = FMODE_READ;
2561
2562
2563 if ((!jdev_name || !jdev_name[0])) {
2564 if (jdev == super->s_dev)
2565 blkdev_mode &= ~FMODE_EXCL;
2566 journal->j_dev_bd = blkdev_get_by_dev(jdev, blkdev_mode,
2567 journal);
2568 journal->j_dev_mode = blkdev_mode;
2569 if (IS_ERR(journal->j_dev_bd)) {
2570 result = PTR_ERR(journal->j_dev_bd);
2571 journal->j_dev_bd = NULL;
2572 reiserfs_warning(super, "sh-458",
2573 "cannot init journal device '%s': %i",
2574 __bdevname(jdev, b), result);
2575 return result;
2576 } else if (jdev != super->s_dev)
2577 set_blocksize(journal->j_dev_bd, super->s_blocksize);
2578
2579 return 0;
2580 }
2581
2582 journal->j_dev_mode = blkdev_mode;
2583 journal->j_dev_bd = blkdev_get_by_path(jdev_name, blkdev_mode, journal);
2584 if (IS_ERR(journal->j_dev_bd)) {
2585 result = PTR_ERR(journal->j_dev_bd);
2586 journal->j_dev_bd = NULL;
2587 reiserfs_warning(super,
2588 "journal_init_dev: Cannot open '%s': %i",
2589 jdev_name, result);
2590 return result;
2591 }
2592
2593 set_blocksize(journal->j_dev_bd, super->s_blocksize);
2594 reiserfs_info(super,
2595 "journal_init_dev: journal device: %s\n",
2596 bdevname(journal->j_dev_bd, b));
2597 return 0;
2598}
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612#define REISERFS_STANDARD_BLKSIZE (4096)
2613
2614static int check_advise_trans_params(struct super_block *sb,
2615 struct reiserfs_journal *journal)
2616{
2617 if (journal->j_trans_max) {
2618
2619
2620 int ratio = 1;
2621 if (sb->s_blocksize < REISERFS_STANDARD_BLKSIZE)
2622 ratio = REISERFS_STANDARD_BLKSIZE / sb->s_blocksize;
2623
2624 if (journal->j_trans_max > JOURNAL_TRANS_MAX_DEFAULT / ratio ||
2625 journal->j_trans_max < JOURNAL_TRANS_MIN_DEFAULT / ratio ||
2626 SB_ONDISK_JOURNAL_SIZE(sb) / journal->j_trans_max <
2627 JOURNAL_MIN_RATIO) {
2628 reiserfs_warning(sb, "sh-462",
2629 "bad transaction max size (%u). "
2630 "FSCK?", journal->j_trans_max);
2631 return 1;
2632 }
2633 if (journal->j_max_batch != (journal->j_trans_max) *
2634 JOURNAL_MAX_BATCH_DEFAULT/JOURNAL_TRANS_MAX_DEFAULT) {
2635 reiserfs_warning(sb, "sh-463",
2636 "bad transaction max batch (%u). "
2637 "FSCK?", journal->j_max_batch);
2638 return 1;
2639 }
2640 } else {
2641
2642
2643
2644
2645 if (sb->s_blocksize != REISERFS_STANDARD_BLKSIZE) {
2646 reiserfs_warning(sb, "sh-464", "bad blocksize (%u)",
2647 sb->s_blocksize);
2648 return 1;
2649 }
2650 journal->j_trans_max = JOURNAL_TRANS_MAX_DEFAULT;
2651 journal->j_max_batch = JOURNAL_MAX_BATCH_DEFAULT;
2652 journal->j_max_commit_age = JOURNAL_MAX_COMMIT_AGE;
2653 }
2654 return 0;
2655}
2656
2657
2658
2659
2660int journal_init(struct super_block *sb, const char *j_dev_name,
2661 int old_format, unsigned int commit_max_age)
2662{
2663 int num_cnodes = SB_ONDISK_JOURNAL_SIZE(sb) * 2;
2664 struct buffer_head *bhjh;
2665 struct reiserfs_super_block *rs;
2666 struct reiserfs_journal_header *jh;
2667 struct reiserfs_journal *journal;
2668 struct reiserfs_journal_list *jl;
2669 char b[BDEVNAME_SIZE];
2670 int ret;
2671
2672 journal = SB_JOURNAL(sb) = vzalloc(sizeof(struct reiserfs_journal));
2673 if (!journal) {
2674 reiserfs_warning(sb, "journal-1256",
2675 "unable to get memory for journal structure");
2676 return 1;
2677 }
2678 INIT_LIST_HEAD(&journal->j_bitmap_nodes);
2679 INIT_LIST_HEAD(&journal->j_prealloc_list);
2680 INIT_LIST_HEAD(&journal->j_working_list);
2681 INIT_LIST_HEAD(&journal->j_journal_list);
2682 journal->j_persistent_trans = 0;
2683 if (reiserfs_allocate_list_bitmaps(sb, journal->j_list_bitmap,
2684 reiserfs_bmap_count(sb)))
2685 goto free_and_return;
2686
2687 allocate_bitmap_nodes(sb);
2688
2689
2690 SB_JOURNAL_1st_RESERVED_BLOCK(sb) = (old_format ?
2691 REISERFS_OLD_DISK_OFFSET_IN_BYTES
2692 / sb->s_blocksize +
2693 reiserfs_bmap_count(sb) +
2694 1 :
2695 REISERFS_DISK_OFFSET_IN_BYTES /
2696 sb->s_blocksize + 2);
2697
2698
2699
2700 if (!SB_ONDISK_JOURNAL_DEVICE(sb) &&
2701 (SB_JOURNAL_1st_RESERVED_BLOCK(sb) +
2702 SB_ONDISK_JOURNAL_SIZE(sb) > sb->s_blocksize * 8)) {
2703 reiserfs_warning(sb, "journal-1393",
2704 "journal does not fit for area addressed "
2705 "by first of bitmap blocks. It starts at "
2706 "%u and its size is %u. Block size %ld",
2707 SB_JOURNAL_1st_RESERVED_BLOCK(sb),
2708 SB_ONDISK_JOURNAL_SIZE(sb),
2709 sb->s_blocksize);
2710 goto free_and_return;
2711 }
2712
2713 if (journal_init_dev(sb, journal, j_dev_name) != 0) {
2714 reiserfs_warning(sb, "sh-462",
2715 "unable to initialize jornal device");
2716 goto free_and_return;
2717 }
2718
2719 rs = SB_DISK_SUPER_BLOCK(sb);
2720
2721
2722 bhjh = journal_bread(sb,
2723 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
2724 SB_ONDISK_JOURNAL_SIZE(sb));
2725 if (!bhjh) {
2726 reiserfs_warning(sb, "sh-459",
2727 "unable to read journal header");
2728 goto free_and_return;
2729 }
2730 jh = (struct reiserfs_journal_header *)(bhjh->b_data);
2731
2732
2733 if (is_reiserfs_jr(rs)
2734 && (le32_to_cpu(jh->jh_journal.jp_journal_magic) !=
2735 sb_jp_journal_magic(rs))) {
2736 reiserfs_warning(sb, "sh-460",
2737 "journal header magic %x (device %s) does "
2738 "not match to magic found in super block %x",
2739 jh->jh_journal.jp_journal_magic,
2740 bdevname(journal->j_dev_bd, b),
2741 sb_jp_journal_magic(rs));
2742 brelse(bhjh);
2743 goto free_and_return;
2744 }
2745
2746 journal->j_trans_max = le32_to_cpu(jh->jh_journal.jp_journal_trans_max);
2747 journal->j_max_batch = le32_to_cpu(jh->jh_journal.jp_journal_max_batch);
2748 journal->j_max_commit_age =
2749 le32_to_cpu(jh->jh_journal.jp_journal_max_commit_age);
2750 journal->j_max_trans_age = JOURNAL_MAX_TRANS_AGE;
2751
2752 if (check_advise_trans_params(sb, journal) != 0)
2753 goto free_and_return;
2754 journal->j_default_max_commit_age = journal->j_max_commit_age;
2755
2756 if (commit_max_age != 0) {
2757 journal->j_max_commit_age = commit_max_age;
2758 journal->j_max_trans_age = commit_max_age;
2759 }
2760
2761 reiserfs_info(sb, "journal params: device %s, size %u, "
2762 "journal first block %u, max trans len %u, max batch %u, "
2763 "max commit age %u, max trans age %u\n",
2764 bdevname(journal->j_dev_bd, b),
2765 SB_ONDISK_JOURNAL_SIZE(sb),
2766 SB_ONDISK_JOURNAL_1st_BLOCK(sb),
2767 journal->j_trans_max,
2768 journal->j_max_batch,
2769 journal->j_max_commit_age, journal->j_max_trans_age);
2770
2771 brelse(bhjh);
2772
2773 journal->j_list_bitmap_index = 0;
2774 journal_list_init(sb);
2775
2776 memset(journal->j_list_hash_table, 0,
2777 JOURNAL_HASH_SIZE * sizeof(struct reiserfs_journal_cnode *));
2778
2779 INIT_LIST_HEAD(&journal->j_dirty_buffers);
2780 spin_lock_init(&journal->j_dirty_buffers_lock);
2781
2782 journal->j_start = 0;
2783 journal->j_len = 0;
2784 journal->j_len_alloc = 0;
2785 atomic_set(&(journal->j_wcount), 0);
2786 atomic_set(&(journal->j_async_throttle), 0);
2787 journal->j_bcount = 0;
2788 journal->j_trans_start_time = 0;
2789 journal->j_last = NULL;
2790 journal->j_first = NULL;
2791 init_waitqueue_head(&(journal->j_join_wait));
2792 mutex_init(&journal->j_mutex);
2793 mutex_init(&journal->j_flush_mutex);
2794
2795 journal->j_trans_id = 10;
2796 journal->j_mount_id = 10;
2797 journal->j_state = 0;
2798 atomic_set(&(journal->j_jlock), 0);
2799 journal->j_cnode_free_list = allocate_cnodes(num_cnodes);
2800 journal->j_cnode_free_orig = journal->j_cnode_free_list;
2801 journal->j_cnode_free = journal->j_cnode_free_list ? num_cnodes : 0;
2802 journal->j_cnode_used = 0;
2803 journal->j_must_wait = 0;
2804
2805 if (journal->j_cnode_free == 0) {
2806 reiserfs_warning(sb, "journal-2004", "Journal cnode memory "
2807 "allocation failed (%ld bytes). Journal is "
2808 "too large for available memory. Usually "
2809 "this is due to a journal that is too large.",
2810 sizeof (struct reiserfs_journal_cnode) * num_cnodes);
2811 goto free_and_return;
2812 }
2813
2814 init_journal_hash(sb);
2815 jl = journal->j_current_jl;
2816
2817
2818
2819
2820
2821
2822 reiserfs_write_lock(sb);
2823 jl->j_list_bitmap = get_list_bitmap(sb, jl);
2824 reiserfs_write_unlock(sb);
2825 if (!jl->j_list_bitmap) {
2826 reiserfs_warning(sb, "journal-2005",
2827 "get_list_bitmap failed for journal list 0");
2828 goto free_and_return;
2829 }
2830
2831
2832
2833
2834
2835 reiserfs_write_lock(sb);
2836 ret = journal_read(sb);
2837 reiserfs_write_unlock(sb);
2838 if (ret < 0) {
2839 reiserfs_warning(sb, "reiserfs-2006",
2840 "Replay Failure, unable to mount");
2841 goto free_and_return;
2842 }
2843
2844 reiserfs_mounted_fs_count++;
2845 if (reiserfs_mounted_fs_count <= 1)
2846 commit_wq = alloc_workqueue("reiserfs", WQ_MEM_RECLAIM, 0);
2847
2848 INIT_DELAYED_WORK(&journal->j_work, flush_async_commits);
2849 journal->j_work_sb = sb;
2850 return 0;
2851 free_and_return:
2852 free_journal_ram(sb);
2853 return 1;
2854}
2855
2856
2857
2858
2859
2860
2861int journal_transaction_should_end(struct reiserfs_transaction_handle *th,
2862 int new_alloc)
2863{
2864 struct reiserfs_journal *journal = SB_JOURNAL(th->t_super);
2865 time_t now = get_seconds();
2866
2867 BUG_ON(!th->t_trans_id);
2868 if (th->t_refcount > 1)
2869 return 0;
2870 if (journal->j_must_wait > 0 ||
2871 (journal->j_len_alloc + new_alloc) >= journal->j_max_batch ||
2872 atomic_read(&(journal->j_jlock)) ||
2873 (now - journal->j_trans_start_time) > journal->j_max_trans_age ||
2874 journal->j_cnode_free < (journal->j_trans_max * 3)) {
2875 return 1;
2876 }
2877
2878 journal->j_len_alloc += new_alloc;
2879 th->t_blocks_allocated += new_alloc ;
2880 return 0;
2881}
2882
2883
2884
2885void reiserfs_block_writes(struct reiserfs_transaction_handle *th)
2886{
2887 struct reiserfs_journal *journal = SB_JOURNAL(th->t_super);
2888 BUG_ON(!th->t_trans_id);
2889 journal->j_must_wait = 1;
2890 set_bit(J_WRITERS_BLOCKED, &journal->j_state);
2891 return;
2892}
2893
2894
2895
2896void reiserfs_allow_writes(struct super_block *s)
2897{
2898 struct reiserfs_journal *journal = SB_JOURNAL(s);
2899 clear_bit(J_WRITERS_BLOCKED, &journal->j_state);
2900 wake_up(&journal->j_join_wait);
2901}
2902
2903
2904
2905void reiserfs_wait_on_write_block(struct super_block *s)
2906{
2907 struct reiserfs_journal *journal = SB_JOURNAL(s);
2908 wait_event(journal->j_join_wait,
2909 !test_bit(J_WRITERS_BLOCKED, &journal->j_state));
2910}
2911
2912static void queue_log_writer(struct super_block *s)
2913{
2914 wait_queue_t wait;
2915 struct reiserfs_journal *journal = SB_JOURNAL(s);
2916 set_bit(J_WRITERS_QUEUED, &journal->j_state);
2917
2918
2919
2920
2921
2922 init_waitqueue_entry(&wait, current);
2923 add_wait_queue(&journal->j_join_wait, &wait);
2924 set_current_state(TASK_UNINTERRUPTIBLE);
2925 if (test_bit(J_WRITERS_QUEUED, &journal->j_state)) {
2926 reiserfs_write_unlock(s);
2927 schedule();
2928 reiserfs_write_lock(s);
2929 }
2930 __set_current_state(TASK_RUNNING);
2931 remove_wait_queue(&journal->j_join_wait, &wait);
2932}
2933
2934static void wake_queued_writers(struct super_block *s)
2935{
2936 struct reiserfs_journal *journal = SB_JOURNAL(s);
2937 if (test_and_clear_bit(J_WRITERS_QUEUED, &journal->j_state))
2938 wake_up(&journal->j_join_wait);
2939}
2940
2941static void let_transaction_grow(struct super_block *sb, unsigned int trans_id)
2942{
2943 struct reiserfs_journal *journal = SB_JOURNAL(sb);
2944 unsigned long bcount = journal->j_bcount;
2945 while (1) {
2946 reiserfs_write_unlock(sb);
2947 schedule_timeout_uninterruptible(1);
2948 reiserfs_write_lock(sb);
2949 journal->j_current_jl->j_state |= LIST_COMMIT_PENDING;
2950 while ((atomic_read(&journal->j_wcount) > 0 ||
2951 atomic_read(&journal->j_jlock)) &&
2952 journal->j_trans_id == trans_id) {
2953 queue_log_writer(sb);
2954 }
2955 if (journal->j_trans_id != trans_id)
2956 break;
2957 if (bcount == journal->j_bcount)
2958 break;
2959 bcount = journal->j_bcount;
2960 }
2961}
2962
2963
2964
2965
2966
2967
2968
2969static int do_journal_begin_r(struct reiserfs_transaction_handle *th,
2970 struct super_block *sb, unsigned long nblocks,
2971 int join)
2972{
2973 time_t now = get_seconds();
2974 unsigned int old_trans_id;
2975 struct reiserfs_journal *journal = SB_JOURNAL(sb);
2976 struct reiserfs_transaction_handle myth;
2977 int sched_count = 0;
2978 int retval;
2979
2980 reiserfs_check_lock_depth(sb, "journal_begin");
2981 BUG_ON(nblocks > journal->j_trans_max);
2982
2983 PROC_INFO_INC(sb, journal.journal_being);
2984
2985 th->t_refcount = 1;
2986 th->t_super = sb;
2987
2988 relock:
2989 lock_journal(sb);
2990 if (join != JBEGIN_ABORT && reiserfs_is_journal_aborted(journal)) {
2991 unlock_journal(sb);
2992 retval = journal->j_errno;
2993 goto out_fail;
2994 }
2995 journal->j_bcount++;
2996
2997 if (test_bit(J_WRITERS_BLOCKED, &journal->j_state)) {
2998 unlock_journal(sb);
2999 reiserfs_write_unlock(sb);
3000 reiserfs_wait_on_write_block(sb);
3001 reiserfs_write_lock(sb);
3002 PROC_INFO_INC(sb, journal.journal_relock_writers);
3003 goto relock;
3004 }
3005 now = get_seconds();
3006
3007
3008
3009
3010
3011
3012 if ((!join && journal->j_must_wait > 0) ||
3013 (!join
3014 && (journal->j_len_alloc + nblocks + 2) >= journal->j_max_batch)
3015 || (!join && atomic_read(&journal->j_wcount) > 0
3016 && journal->j_trans_start_time > 0
3017 && (now - journal->j_trans_start_time) >
3018 journal->j_max_trans_age) || (!join
3019 && atomic_read(&journal->j_jlock))
3020 || (!join && journal->j_cnode_free < (journal->j_trans_max * 3))) {
3021
3022 old_trans_id = journal->j_trans_id;
3023 unlock_journal(sb);
3024
3025 if (!join && (journal->j_len_alloc + nblocks + 2) >=
3026 journal->j_max_batch &&
3027 ((journal->j_len + nblocks + 2) * 100) <
3028 (journal->j_len_alloc * 75)) {
3029 if (atomic_read(&journal->j_wcount) > 10) {
3030 sched_count++;
3031 queue_log_writer(sb);
3032 goto relock;
3033 }
3034 }
3035
3036
3037
3038 if (atomic_read(&journal->j_jlock)) {
3039 while (journal->j_trans_id == old_trans_id &&
3040 atomic_read(&journal->j_jlock)) {
3041 queue_log_writer(sb);
3042 }
3043 goto relock;
3044 }
3045 retval = journal_join(&myth, sb, 1);
3046 if (retval)
3047 goto out_fail;
3048
3049
3050 if (old_trans_id != journal->j_trans_id) {
3051 retval = do_journal_end(&myth, sb, 1, 0);
3052 } else {
3053 retval = do_journal_end(&myth, sb, 1, COMMIT_NOW);
3054 }
3055
3056 if (retval)
3057 goto out_fail;
3058
3059 PROC_INFO_INC(sb, journal.journal_relock_wcount);
3060 goto relock;
3061 }
3062
3063 if (journal->j_trans_start_time == 0) {
3064 journal->j_trans_start_time = get_seconds();
3065 }
3066 atomic_inc(&(journal->j_wcount));
3067 journal->j_len_alloc += nblocks;
3068 th->t_blocks_logged = 0;
3069 th->t_blocks_allocated = nblocks;
3070 th->t_trans_id = journal->j_trans_id;
3071 unlock_journal(sb);
3072 INIT_LIST_HEAD(&th->t_list);
3073 return 0;
3074
3075 out_fail:
3076 memset(th, 0, sizeof(*th));
3077
3078
3079
3080 th->t_super = sb;
3081 return retval;
3082}
3083
3084struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
3085 super_block
3086 *s,
3087 int nblocks)
3088{
3089 int ret;
3090 struct reiserfs_transaction_handle *th;
3091
3092
3093
3094
3095 if (reiserfs_transaction_running(s)) {
3096 th = current->journal_info;
3097 th->t_refcount++;
3098 BUG_ON(th->t_refcount < 2);
3099
3100 return th;
3101 }
3102 th = kmalloc(sizeof(struct reiserfs_transaction_handle), GFP_NOFS);
3103 if (!th)
3104 return NULL;
3105 ret = journal_begin(th, s, nblocks);
3106 if (ret) {
3107 kfree(th);
3108 return NULL;
3109 }
3110
3111 SB_JOURNAL(s)->j_persistent_trans++;
3112 return th;
3113}
3114
3115int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *th)
3116{
3117 struct super_block *s = th->t_super;
3118 int ret = 0;
3119 if (th->t_trans_id)
3120 ret = journal_end(th, th->t_super, th->t_blocks_allocated);
3121 else
3122 ret = -EIO;
3123 if (th->t_refcount == 0) {
3124 SB_JOURNAL(s)->j_persistent_trans--;
3125 kfree(th);
3126 }
3127 return ret;
3128}
3129
3130static int journal_join(struct reiserfs_transaction_handle *th,
3131 struct super_block *sb, unsigned long nblocks)
3132{
3133 struct reiserfs_transaction_handle *cur_th = current->journal_info;
3134
3135
3136
3137
3138 th->t_handle_save = cur_th;
3139 BUG_ON(cur_th && cur_th->t_refcount > 1);
3140 return do_journal_begin_r(th, sb, nblocks, JBEGIN_JOIN);
3141}
3142
3143int journal_join_abort(struct reiserfs_transaction_handle *th,
3144 struct super_block *sb, unsigned long nblocks)
3145{
3146 struct reiserfs_transaction_handle *cur_th = current->journal_info;
3147
3148
3149
3150
3151 th->t_handle_save = cur_th;
3152 BUG_ON(cur_th && cur_th->t_refcount > 1);
3153 return do_journal_begin_r(th, sb, nblocks, JBEGIN_ABORT);
3154}
3155
3156int journal_begin(struct reiserfs_transaction_handle *th,
3157 struct super_block *sb, unsigned long nblocks)
3158{
3159 struct reiserfs_transaction_handle *cur_th = current->journal_info;
3160 int ret;
3161
3162 th->t_handle_save = NULL;
3163 if (cur_th) {
3164
3165 if (cur_th->t_super == sb) {
3166 BUG_ON(!cur_th->t_refcount);
3167 cur_th->t_refcount++;
3168 memcpy(th, cur_th, sizeof(*th));
3169 if (th->t_refcount <= 1)
3170 reiserfs_warning(sb, "reiserfs-2005",
3171 "BAD: refcount <= 1, but "
3172 "journal_info != 0");
3173 return 0;
3174 } else {
3175
3176
3177
3178
3179 reiserfs_warning(sb, "clm-2100",
3180 "nesting info a different FS");
3181 th->t_handle_save = current->journal_info;
3182 current->journal_info = th;
3183 }
3184 } else {
3185 current->journal_info = th;
3186 }
3187 ret = do_journal_begin_r(th, sb, nblocks, JBEGIN_REG);
3188 BUG_ON(current->journal_info != th);
3189
3190
3191
3192
3193 if (ret)
3194 current->journal_info = th->t_handle_save;
3195 else
3196 BUG_ON(!th->t_refcount);
3197
3198 return ret;
3199}
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210int journal_mark_dirty(struct reiserfs_transaction_handle *th,
3211 struct super_block *sb, struct buffer_head *bh)
3212{
3213 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3214 struct reiserfs_journal_cnode *cn = NULL;
3215 int count_already_incd = 0;
3216 int prepared = 0;
3217 BUG_ON(!th->t_trans_id);
3218
3219 PROC_INFO_INC(sb, journal.mark_dirty);
3220 if (th->t_trans_id != journal->j_trans_id) {
3221 reiserfs_panic(th->t_super, "journal-1577",
3222 "handle trans id %ld != current trans id %ld",
3223 th->t_trans_id, journal->j_trans_id);
3224 }
3225
3226 prepared = test_clear_buffer_journal_prepared(bh);
3227 clear_buffer_journal_restore_dirty(bh);
3228
3229 if (buffer_journaled(bh)) {
3230 PROC_INFO_INC(sb, journal.mark_dirty_already);
3231 return 0;
3232 }
3233
3234
3235
3236
3237
3238 if (!prepared || buffer_dirty(bh)) {
3239 reiserfs_warning(sb, "journal-1777",
3240 "buffer %llu bad state "
3241 "%cPREPARED %cLOCKED %cDIRTY %cJDIRTY_WAIT",
3242 (unsigned long long)bh->b_blocknr,
3243 prepared ? ' ' : '!',
3244 buffer_locked(bh) ? ' ' : '!',
3245 buffer_dirty(bh) ? ' ' : '!',
3246 buffer_journal_dirty(bh) ? ' ' : '!');
3247 }
3248
3249 if (atomic_read(&(journal->j_wcount)) <= 0) {
3250 reiserfs_warning(sb, "journal-1409",
3251 "returning because j_wcount was %d",
3252 atomic_read(&(journal->j_wcount)));
3253 return 1;
3254 }
3255
3256
3257
3258 if (journal->j_len >= journal->j_trans_max) {
3259 reiserfs_panic(th->t_super, "journal-1413",
3260 "j_len (%lu) is too big",
3261 journal->j_len);
3262 }
3263
3264 if (buffer_journal_dirty(bh)) {
3265 count_already_incd = 1;
3266 PROC_INFO_INC(sb, journal.mark_dirty_notjournal);
3267 clear_buffer_journal_dirty(bh);
3268 }
3269
3270 if (journal->j_len > journal->j_len_alloc) {
3271 journal->j_len_alloc = journal->j_len + JOURNAL_PER_BALANCE_CNT;
3272 }
3273
3274 set_buffer_journaled(bh);
3275
3276
3277 if (!cn) {
3278 cn = get_cnode(sb);
3279 if (!cn) {
3280 reiserfs_panic(sb, "journal-4", "get_cnode failed!");
3281 }
3282
3283 if (th->t_blocks_logged == th->t_blocks_allocated) {
3284 th->t_blocks_allocated += JOURNAL_PER_BALANCE_CNT;
3285 journal->j_len_alloc += JOURNAL_PER_BALANCE_CNT;
3286 }
3287 th->t_blocks_logged++;
3288 journal->j_len++;
3289
3290 cn->bh = bh;
3291 cn->blocknr = bh->b_blocknr;
3292 cn->sb = sb;
3293 cn->jlist = NULL;
3294 insert_journal_hash(journal->j_hash_table, cn);
3295 if (!count_already_incd) {
3296 get_bh(bh);
3297 }
3298 }
3299 cn->next = NULL;
3300 cn->prev = journal->j_last;
3301 cn->bh = bh;
3302 if (journal->j_last) {
3303 journal->j_last->next = cn;
3304 journal->j_last = cn;
3305 } else {
3306 journal->j_first = cn;
3307 journal->j_last = cn;
3308 }
3309 reiserfs_schedule_old_flush(sb);
3310 return 0;
3311}
3312
3313int journal_end(struct reiserfs_transaction_handle *th,
3314 struct super_block *sb, unsigned long nblocks)
3315{
3316 if (!current->journal_info && th->t_refcount > 1)
3317 reiserfs_warning(sb, "REISER-NESTING",
3318 "th NULL, refcount %d", th->t_refcount);
3319
3320 if (!th->t_trans_id) {
3321 WARN_ON(1);
3322 return -EIO;
3323 }
3324
3325 th->t_refcount--;
3326 if (th->t_refcount > 0) {
3327 struct reiserfs_transaction_handle *cur_th =
3328 current->journal_info;
3329
3330
3331
3332
3333 BUG_ON(cur_th->t_super != th->t_super);
3334
3335 if (th != cur_th) {
3336 memcpy(current->journal_info, th, sizeof(*th));
3337 th->t_trans_id = 0;
3338 }
3339 return 0;
3340 } else {
3341 return do_journal_end(th, sb, nblocks, 0);
3342 }
3343}
3344
3345
3346
3347
3348
3349
3350
3351
3352static int remove_from_transaction(struct super_block *sb,
3353 b_blocknr_t blocknr, int already_cleaned)
3354{
3355 struct buffer_head *bh;
3356 struct reiserfs_journal_cnode *cn;
3357 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3358 int ret = 0;
3359
3360 cn = get_journal_hash_dev(sb, journal->j_hash_table, blocknr);
3361 if (!cn || !cn->bh) {
3362 return ret;
3363 }
3364 bh = cn->bh;
3365 if (cn->prev) {
3366 cn->prev->next = cn->next;
3367 }
3368 if (cn->next) {
3369 cn->next->prev = cn->prev;
3370 }
3371 if (cn == journal->j_first) {
3372 journal->j_first = cn->next;
3373 }
3374 if (cn == journal->j_last) {
3375 journal->j_last = cn->prev;
3376 }
3377 if (bh)
3378 remove_journal_hash(sb, journal->j_hash_table, NULL,
3379 bh->b_blocknr, 0);
3380 clear_buffer_journaled(bh);
3381
3382 if (!already_cleaned) {
3383 clear_buffer_journal_dirty(bh);
3384 clear_buffer_dirty(bh);
3385 clear_buffer_journal_test(bh);
3386 put_bh(bh);
3387 if (atomic_read(&(bh->b_count)) < 0) {
3388 reiserfs_warning(sb, "journal-1752",
3389 "b_count < 0");
3390 }
3391 ret = 1;
3392 }
3393 journal->j_len--;
3394 journal->j_len_alloc--;
3395 free_cnode(sb, cn);
3396 return ret;
3397}
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409static int can_dirty(struct reiserfs_journal_cnode *cn)
3410{
3411 struct super_block *sb = cn->sb;
3412 b_blocknr_t blocknr = cn->blocknr;
3413 struct reiserfs_journal_cnode *cur = cn->hprev;
3414 int can_dirty = 1;
3415
3416
3417
3418
3419
3420 while (cur && can_dirty) {
3421 if (cur->jlist && cur->bh && cur->blocknr && cur->sb == sb &&
3422 cur->blocknr == blocknr) {
3423 can_dirty = 0;
3424 }
3425 cur = cur->hprev;
3426 }
3427
3428
3429
3430 cur = cn->hnext;
3431 while (cur && can_dirty) {
3432 if (cur->jlist && cur->jlist->j_len > 0 &&
3433 atomic_read(&(cur->jlist->j_commit_left)) > 0 && cur->bh &&
3434 cur->blocknr && cur->sb == sb && cur->blocknr == blocknr) {
3435 can_dirty = 0;
3436 }
3437 cur = cur->hnext;
3438 }
3439 return can_dirty;
3440}
3441
3442
3443
3444
3445int journal_end_sync(struct reiserfs_transaction_handle *th,
3446 struct super_block *sb, unsigned long nblocks)
3447{
3448 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3449
3450 BUG_ON(!th->t_trans_id);
3451
3452 BUG_ON(th->t_refcount > 1);
3453 if (journal->j_len == 0) {
3454 reiserfs_prepare_for_journal(sb, SB_BUFFER_WITH_SB(sb),
3455 1);
3456 journal_mark_dirty(th, sb, SB_BUFFER_WITH_SB(sb));
3457 }
3458 return do_journal_end(th, sb, nblocks, COMMIT_NOW | WAIT);
3459}
3460
3461
3462
3463
3464static void flush_async_commits(struct work_struct *work)
3465{
3466 struct reiserfs_journal *journal =
3467 container_of(work, struct reiserfs_journal, j_work.work);
3468 struct super_block *sb = journal->j_work_sb;
3469 struct reiserfs_journal_list *jl;
3470 struct list_head *entry;
3471
3472 reiserfs_write_lock(sb);
3473 if (!list_empty(&journal->j_journal_list)) {
3474
3475 entry = journal->j_journal_list.prev;
3476 jl = JOURNAL_LIST_ENTRY(entry);
3477 flush_commit_list(sb, jl, 1);
3478 }
3479 reiserfs_write_unlock(sb);
3480}
3481
3482
3483
3484
3485
3486void reiserfs_flush_old_commits(struct super_block *sb)
3487{
3488 time_t now;
3489 struct reiserfs_transaction_handle th;
3490 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3491
3492 now = get_seconds();
3493
3494
3495
3496 if (list_empty(&journal->j_journal_list))
3497 return;
3498
3499
3500
3501
3502 if (atomic_read(&journal->j_wcount) <= 0 &&
3503 journal->j_trans_start_time > 0 &&
3504 journal->j_len > 0 &&
3505 (now - journal->j_trans_start_time) > journal->j_max_trans_age) {
3506 if (!journal_join(&th, sb, 1)) {
3507 reiserfs_prepare_for_journal(sb,
3508 SB_BUFFER_WITH_SB(sb),
3509 1);
3510 journal_mark_dirty(&th, sb,
3511 SB_BUFFER_WITH_SB(sb));
3512
3513
3514
3515
3516 do_journal_end(&th, sb, 1, COMMIT_NOW | WAIT);
3517 }
3518 }
3519}
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532static int check_journal_end(struct reiserfs_transaction_handle *th,
3533 struct super_block *sb, unsigned long nblocks,
3534 int flags)
3535{
3536
3537 time_t now;
3538 int flush = flags & FLUSH_ALL;
3539 int commit_now = flags & COMMIT_NOW;
3540 int wait_on_commit = flags & WAIT;
3541 struct reiserfs_journal_list *jl;
3542 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3543
3544 BUG_ON(!th->t_trans_id);
3545
3546 if (th->t_trans_id != journal->j_trans_id) {
3547 reiserfs_panic(th->t_super, "journal-1577",
3548 "handle trans id %ld != current trans id %ld",
3549 th->t_trans_id, journal->j_trans_id);
3550 }
3551
3552 journal->j_len_alloc -= (th->t_blocks_allocated - th->t_blocks_logged);
3553 if (atomic_read(&(journal->j_wcount)) > 0) {
3554 atomic_dec(&(journal->j_wcount));
3555 }
3556
3557
3558
3559
3560
3561 BUG_ON(journal->j_len == 0);
3562
3563
3564
3565
3566
3567
3568
3569 if (atomic_read(&(journal->j_wcount)) > 0) {
3570 if (flush || commit_now) {
3571 unsigned trans_id;
3572
3573 jl = journal->j_current_jl;
3574 trans_id = jl->j_trans_id;
3575 if (wait_on_commit)
3576 jl->j_state |= LIST_COMMIT_PENDING;
3577 atomic_set(&(journal->j_jlock), 1);
3578 if (flush) {
3579 journal->j_next_full_flush = 1;
3580 }
3581 unlock_journal(sb);
3582
3583
3584 while (journal->j_trans_id == trans_id) {
3585 if (atomic_read(&journal->j_jlock)) {
3586 queue_log_writer(sb);
3587 } else {
3588 lock_journal(sb);
3589 if (journal->j_trans_id == trans_id) {
3590 atomic_set(&(journal->j_jlock),
3591 1);
3592 }
3593 unlock_journal(sb);
3594 }
3595 }
3596 BUG_ON(journal->j_trans_id == trans_id);
3597
3598 if (commit_now
3599 && journal_list_still_alive(sb, trans_id)
3600 && wait_on_commit) {
3601 flush_commit_list(sb, jl, 1);
3602 }
3603 return 0;
3604 }
3605 unlock_journal(sb);
3606 return 0;
3607 }
3608
3609
3610 now = get_seconds();
3611 if ((now - journal->j_trans_start_time) > journal->j_max_trans_age) {
3612 commit_now = 1;
3613 journal->j_next_async_flush = 1;
3614 }
3615
3616
3617 if (!(journal->j_must_wait > 0) && !(atomic_read(&(journal->j_jlock)))
3618 && !flush && !commit_now && (journal->j_len < journal->j_max_batch)
3619 && journal->j_len_alloc < journal->j_max_batch
3620 && journal->j_cnode_free > (journal->j_trans_max * 3)) {
3621 journal->j_bcount++;
3622 unlock_journal(sb);
3623 return 0;
3624 }
3625
3626 if (journal->j_start > SB_ONDISK_JOURNAL_SIZE(sb)) {
3627 reiserfs_panic(sb, "journal-003",
3628 "j_start (%ld) is too high",
3629 journal->j_start);
3630 }
3631 return 1;
3632}
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648int journal_mark_freed(struct reiserfs_transaction_handle *th,
3649 struct super_block *sb, b_blocknr_t blocknr)
3650{
3651 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3652 struct reiserfs_journal_cnode *cn = NULL;
3653 struct buffer_head *bh = NULL;
3654 struct reiserfs_list_bitmap *jb = NULL;
3655 int cleaned = 0;
3656 BUG_ON(!th->t_trans_id);
3657
3658 cn = get_journal_hash_dev(sb, journal->j_hash_table, blocknr);
3659 if (cn && cn->bh) {
3660 bh = cn->bh;
3661 get_bh(bh);
3662 }
3663
3664 if (bh && buffer_journal_new(bh)) {
3665 clear_buffer_journal_new(bh);
3666 clear_prepared_bits(bh);
3667 reiserfs_clean_and_file_buffer(bh);
3668 cleaned = remove_from_transaction(sb, blocknr, cleaned);
3669 } else {
3670
3671 jb = journal->j_current_jl->j_list_bitmap;
3672 if (!jb) {
3673 reiserfs_panic(sb, "journal-1702",
3674 "journal_list_bitmap is NULL");
3675 }
3676 set_bit_in_list_bitmap(sb, blocknr, jb);
3677
3678
3679
3680 if (bh) {
3681 clear_prepared_bits(bh);
3682 reiserfs_clean_and_file_buffer(bh);
3683 }
3684 cleaned = remove_from_transaction(sb, blocknr, cleaned);
3685
3686
3687 cn = get_journal_hash_dev(sb, journal->j_list_hash_table,
3688 blocknr);
3689 while (cn) {
3690 if (sb == cn->sb && blocknr == cn->blocknr) {
3691 set_bit(BLOCK_FREED, &cn->state);
3692 if (cn->bh) {
3693 if (!cleaned) {
3694
3695
3696
3697 clear_buffer_journal_dirty(cn->
3698 bh);
3699 clear_buffer_dirty(cn->bh);
3700 clear_buffer_journal_test(cn->
3701 bh);
3702 cleaned = 1;
3703 put_bh(cn->bh);
3704 if (atomic_read
3705 (&(cn->bh->b_count)) < 0) {
3706 reiserfs_warning(sb,
3707 "journal-2138",
3708 "cn->bh->b_count < 0");
3709 }
3710 }
3711 if (cn->jlist) {
3712 atomic_dec(&
3713 (cn->jlist->
3714 j_nonzerolen));
3715 }
3716 cn->bh = NULL;
3717 }
3718 }
3719 cn = cn->hnext;
3720 }
3721 }
3722
3723 if (bh)
3724 release_buffer_page(bh);
3725 return 0;
3726}
3727
3728void reiserfs_update_inode_transaction(struct inode *inode)
3729{
3730 struct reiserfs_journal *journal = SB_JOURNAL(inode->i_sb);
3731 REISERFS_I(inode)->i_jl = journal->j_current_jl;
3732 REISERFS_I(inode)->i_trans_id = journal->j_trans_id;
3733}
3734
3735
3736
3737
3738
3739static int __commit_trans_jl(struct inode *inode, unsigned long id,
3740 struct reiserfs_journal_list *jl)
3741{
3742 struct reiserfs_transaction_handle th;
3743 struct super_block *sb = inode->i_sb;
3744 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3745 int ret = 0;
3746
3747
3748 if (id == journal->j_trans_id) {
3749 jl = journal->j_current_jl;
3750
3751 let_transaction_grow(sb, id);
3752 if (journal->j_trans_id != id) {
3753 goto flush_commit_only;
3754 }
3755
3756 ret = journal_begin(&th, sb, 1);
3757 if (ret)
3758 return ret;
3759
3760
3761 if (journal->j_trans_id != id) {
3762 reiserfs_prepare_for_journal(sb, SB_BUFFER_WITH_SB(sb),
3763 1);
3764 journal_mark_dirty(&th, sb, SB_BUFFER_WITH_SB(sb));
3765 ret = journal_end(&th, sb, 1);
3766 goto flush_commit_only;
3767 }
3768
3769 ret = journal_end_sync(&th, sb, 1);
3770 if (!ret)
3771 ret = 1;
3772
3773 } else {
3774
3775
3776
3777
3778 flush_commit_only:
3779 if (journal_list_still_alive(inode->i_sb, id)) {
3780
3781
3782
3783
3784
3785 if (atomic_read(&jl->j_commit_left) > 1)
3786 ret = 1;
3787 flush_commit_list(sb, jl, 1);
3788 if (journal->j_errno)
3789 ret = journal->j_errno;
3790 }
3791 }
3792
3793 return ret;
3794}
3795
3796int reiserfs_commit_for_inode(struct inode *inode)
3797{
3798 unsigned int id = REISERFS_I(inode)->i_trans_id;
3799 struct reiserfs_journal_list *jl = REISERFS_I(inode)->i_jl;
3800
3801
3802
3803
3804 if (!id || !jl) {
3805 reiserfs_update_inode_transaction(inode);
3806 id = REISERFS_I(inode)->i_trans_id;
3807
3808 }
3809
3810 return __commit_trans_jl(inode, id, jl);
3811}
3812
3813void reiserfs_restore_prepared_buffer(struct super_block *sb,
3814 struct buffer_head *bh)
3815{
3816 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3817 PROC_INFO_INC(sb, journal.restore_prepared);
3818 if (!bh) {
3819 return;
3820 }
3821 if (test_clear_buffer_journal_restore_dirty(bh) &&
3822 buffer_journal_dirty(bh)) {
3823 struct reiserfs_journal_cnode *cn;
3824 cn = get_journal_hash_dev(sb,
3825 journal->j_list_hash_table,
3826 bh->b_blocknr);
3827 if (cn && can_dirty(cn)) {
3828 set_buffer_journal_test(bh);
3829 mark_buffer_dirty(bh);
3830 }
3831 }
3832 clear_buffer_journal_prepared(bh);
3833}
3834
3835extern struct tree_balance *cur_tb;
3836
3837
3838
3839
3840
3841
3842
3843int reiserfs_prepare_for_journal(struct super_block *sb,
3844 struct buffer_head *bh, int wait)
3845{
3846 PROC_INFO_INC(sb, journal.prepare);
3847
3848 if (!trylock_buffer(bh)) {
3849 if (!wait)
3850 return 0;
3851 lock_buffer(bh);
3852 }
3853 set_buffer_journal_prepared(bh);
3854 if (test_clear_buffer_dirty(bh) && buffer_journal_dirty(bh)) {
3855 clear_buffer_journal_test(bh);
3856 set_buffer_journal_restore_dirty(bh);
3857 }
3858 unlock_buffer(bh);
3859 return 1;
3860}
3861
3862static void flush_old_journal_lists(struct super_block *s)
3863{
3864 struct reiserfs_journal *journal = SB_JOURNAL(s);
3865 struct reiserfs_journal_list *jl;
3866 struct list_head *entry;
3867 time_t now = get_seconds();
3868
3869 while (!list_empty(&journal->j_journal_list)) {
3870 entry = journal->j_journal_list.next;
3871 jl = JOURNAL_LIST_ENTRY(entry);
3872
3873 if (jl->j_timestamp < (now - (JOURNAL_MAX_TRANS_AGE * 4)) &&
3874 atomic_read(&jl->j_commit_left) == 0 &&
3875 test_transaction(s, jl)) {
3876 flush_used_journal_lists(s, jl);
3877 } else {
3878 break;
3879 }
3880 }
3881}
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893static int do_journal_end(struct reiserfs_transaction_handle *th,
3894 struct super_block *sb, unsigned long nblocks,
3895 int flags)
3896{
3897 struct reiserfs_journal *journal = SB_JOURNAL(sb);
3898 struct reiserfs_journal_cnode *cn, *next, *jl_cn;
3899 struct reiserfs_journal_cnode *last_cn = NULL;
3900 struct reiserfs_journal_desc *desc;
3901 struct reiserfs_journal_commit *commit;
3902 struct buffer_head *c_bh;
3903 struct buffer_head *d_bh;
3904 int cur_write_start = 0;
3905 int old_start;
3906 int i;
3907 int flush;
3908 int wait_on_commit;
3909 struct reiserfs_journal_list *jl, *temp_jl;
3910 struct list_head *entry, *safe;
3911 unsigned long jindex;
3912 unsigned int commit_trans_id;
3913 int trans_half;
3914
3915 BUG_ON(th->t_refcount > 1);
3916 BUG_ON(!th->t_trans_id);
3917
3918
3919
3920 if (th->t_trans_id == ~0U)
3921 flags |= FLUSH_ALL | COMMIT_NOW | WAIT;
3922 flush = flags & FLUSH_ALL;
3923 wait_on_commit = flags & WAIT;
3924
3925 current->journal_info = th->t_handle_save;
3926 reiserfs_check_lock_depth(sb, "journal end");
3927 if (journal->j_len == 0) {
3928 reiserfs_prepare_for_journal(sb, SB_BUFFER_WITH_SB(sb),
3929 1);
3930 journal_mark_dirty(th, sb, SB_BUFFER_WITH_SB(sb));
3931 }
3932
3933 lock_journal(sb);
3934 if (journal->j_next_full_flush) {
3935 flags |= FLUSH_ALL;
3936 flush = 1;
3937 }
3938 if (journal->j_next_async_flush) {
3939 flags |= COMMIT_NOW | WAIT;
3940 wait_on_commit = 1;
3941 }
3942
3943
3944
3945
3946 if (!check_journal_end(th, sb, nblocks, flags)) {
3947 reiserfs_schedule_old_flush(sb);
3948 wake_queued_writers(sb);
3949 reiserfs_async_progress_wait(sb);
3950 goto out;
3951 }
3952
3953
3954 if (journal->j_next_full_flush) {
3955 flush = 1;
3956 }
3957
3958
3959
3960
3961
3962 if (journal->j_must_wait > 0) {
3963 flush = 1;
3964 }
3965#ifdef REISERFS_PREALLOCATE
3966
3967
3968 current->journal_info = th;
3969 th->t_refcount++;
3970 reiserfs_discard_all_prealloc(th);
3971
3972 th->t_refcount--;
3973 current->journal_info = th->t_handle_save;
3974#endif
3975
3976
3977 d_bh =
3978 journal_getblk(sb,
3979 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
3980 journal->j_start);
3981 set_buffer_uptodate(d_bh);
3982 desc = (struct reiserfs_journal_desc *)(d_bh)->b_data;
3983 memset(d_bh->b_data, 0, d_bh->b_size);
3984 memcpy(get_journal_desc_magic(d_bh), JOURNAL_DESC_MAGIC, 8);
3985 set_desc_trans_id(desc, journal->j_trans_id);
3986
3987
3988 c_bh = journal_getblk(sb, SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
3989 ((journal->j_start + journal->j_len +
3990 1) % SB_ONDISK_JOURNAL_SIZE(sb)));
3991 commit = (struct reiserfs_journal_commit *)c_bh->b_data;
3992 memset(c_bh->b_data, 0, c_bh->b_size);
3993 set_commit_trans_id(commit, journal->j_trans_id);
3994 set_buffer_uptodate(c_bh);
3995
3996
3997 jl = journal->j_current_jl;
3998
3999
4000
4001
4002
4003
4004 reiserfs_mutex_lock_safe(&jl->j_commit_mutex, sb);
4005
4006
4007 commit_trans_id = jl->j_trans_id;
4008
4009 atomic_set(&jl->j_older_commits_done, 0);
4010 jl->j_trans_id = journal->j_trans_id;
4011 jl->j_timestamp = journal->j_trans_start_time;
4012 jl->j_commit_bh = c_bh;
4013 jl->j_start = journal->j_start;
4014 jl->j_len = journal->j_len;
4015 atomic_set(&jl->j_nonzerolen, journal->j_len);
4016 atomic_set(&jl->j_commit_left, journal->j_len + 2);
4017 jl->j_realblock = NULL;
4018
4019
4020
4021
4022
4023 trans_half = journal_trans_half(sb->s_blocksize);
4024 for (i = 0, cn = journal->j_first; cn; cn = cn->next, i++) {
4025 if (buffer_journaled(cn->bh)) {
4026 jl_cn = get_cnode(sb);
4027 if (!jl_cn) {
4028 reiserfs_panic(sb, "journal-1676",
4029 "get_cnode returned NULL");
4030 }
4031 if (i == 0) {
4032 jl->j_realblock = jl_cn;
4033 }
4034 jl_cn->prev = last_cn;
4035 jl_cn->next = NULL;
4036 if (last_cn) {
4037 last_cn->next = jl_cn;
4038 }
4039 last_cn = jl_cn;
4040
4041
4042
4043 if (is_block_in_log_or_reserved_area
4044 (sb, cn->bh->b_blocknr)) {
4045 reiserfs_panic(sb, "journal-2332",
4046 "Trying to log block %lu, "
4047 "which is a log block",
4048 cn->bh->b_blocknr);
4049 }
4050 jl_cn->blocknr = cn->bh->b_blocknr;
4051 jl_cn->state = 0;
4052 jl_cn->sb = sb;
4053 jl_cn->bh = cn->bh;
4054 jl_cn->jlist = jl;
4055 insert_journal_hash(journal->j_list_hash_table, jl_cn);
4056 if (i < trans_half) {
4057 desc->j_realblock[i] =
4058 cpu_to_le32(cn->bh->b_blocknr);
4059 } else {
4060 commit->j_realblock[i - trans_half] =
4061 cpu_to_le32(cn->bh->b_blocknr);
4062 }
4063 } else {
4064 i--;
4065 }
4066 }
4067 set_desc_trans_len(desc, journal->j_len);
4068 set_desc_mount_id(desc, journal->j_mount_id);
4069 set_desc_trans_id(desc, journal->j_trans_id);
4070 set_commit_trans_len(commit, journal->j_len);
4071
4072
4073 BUG_ON(journal->j_len == 0);
4074
4075
4076
4077
4078
4079 mark_buffer_dirty(d_bh);
4080
4081
4082 cur_write_start = journal->j_start;
4083 cn = journal->j_first;
4084 jindex = 1;
4085 while (cn) {
4086 clear_buffer_journal_new(cn->bh);
4087
4088 if (buffer_journaled(cn->bh)) {
4089 struct buffer_head *tmp_bh;
4090 char *addr;
4091 struct page *page;
4092 tmp_bh =
4093 journal_getblk(sb,
4094 SB_ONDISK_JOURNAL_1st_BLOCK(sb) +
4095 ((cur_write_start +
4096 jindex) %
4097 SB_ONDISK_JOURNAL_SIZE(sb)));
4098 set_buffer_uptodate(tmp_bh);
4099 page = cn->bh->b_page;
4100 addr = kmap(page);
4101 memcpy(tmp_bh->b_data,
4102 addr + offset_in_page(cn->bh->b_data),
4103 cn->bh->b_size);
4104 kunmap(page);
4105 mark_buffer_dirty(tmp_bh);
4106 jindex++;
4107 set_buffer_journal_dirty(cn->bh);
4108 clear_buffer_journaled(cn->bh);
4109 } else {
4110
4111 reiserfs_warning(sb, "journal-2048",
4112 "BAD, buffer in journal hash, "
4113 "but not JDirty!");
4114 brelse(cn->bh);
4115 }
4116 next = cn->next;
4117 free_cnode(sb, cn);
4118 cn = next;
4119 reiserfs_write_unlock(sb);
4120 cond_resched();
4121 reiserfs_write_lock(sb);
4122 }
4123
4124
4125
4126
4127
4128
4129 journal->j_current_jl = alloc_journal_list(sb);
4130
4131
4132 list_add_tail(&jl->j_list, &journal->j_journal_list);
4133 list_add_tail(&jl->j_working_list, &journal->j_working_list);
4134 journal->j_num_work_lists++;
4135
4136
4137 old_start = journal->j_start;
4138 journal->j_start =
4139 (journal->j_start + journal->j_len +
4140 2) % SB_ONDISK_JOURNAL_SIZE(sb);
4141 atomic_set(&(journal->j_wcount), 0);
4142 journal->j_bcount = 0;
4143 journal->j_last = NULL;
4144 journal->j_first = NULL;
4145 journal->j_len = 0;
4146 journal->j_trans_start_time = 0;
4147
4148 if (++journal->j_trans_id == 0)
4149 journal->j_trans_id = 10;
4150 journal->j_current_jl->j_trans_id = journal->j_trans_id;
4151 journal->j_must_wait = 0;
4152 journal->j_len_alloc = 0;
4153 journal->j_next_full_flush = 0;
4154 journal->j_next_async_flush = 0;
4155 init_journal_hash(sb);
4156
4157
4158
4159 smp_mb();
4160
4161
4162
4163
4164
4165
4166
4167 if (!list_empty(&jl->j_tail_bh_list)) {
4168 reiserfs_write_unlock(sb);
4169 write_ordered_buffers(&journal->j_dirty_buffers_lock,
4170 journal, jl, &jl->j_tail_bh_list);
4171 reiserfs_write_lock(sb);
4172 }
4173 BUG_ON(!list_empty(&jl->j_tail_bh_list));
4174 mutex_unlock(&jl->j_commit_mutex);
4175
4176
4177
4178
4179
4180
4181
4182
4183 if (flush) {
4184 flush_commit_list(sb, jl, 1);
4185 flush_journal_list(sb, jl, 1);
4186 } else if (!(jl->j_state & LIST_COMMIT_PENDING))
4187 queue_delayed_work(commit_wq, &journal->j_work, HZ / 10);
4188
4189
4190
4191
4192
4193 first_jl:
4194 list_for_each_safe(entry, safe, &journal->j_journal_list) {
4195 temp_jl = JOURNAL_LIST_ENTRY(entry);
4196 if (journal->j_start <= temp_jl->j_start) {
4197 if ((journal->j_start + journal->j_trans_max + 1) >=
4198 temp_jl->j_start) {
4199 flush_used_journal_lists(sb, temp_jl);
4200 goto first_jl;
4201 } else if ((journal->j_start +
4202 journal->j_trans_max + 1) <
4203 SB_ONDISK_JOURNAL_SIZE(sb)) {
4204
4205
4206
4207
4208 break;
4209 }
4210 } else if ((journal->j_start +
4211 journal->j_trans_max + 1) >
4212 SB_ONDISK_JOURNAL_SIZE(sb)) {
4213 if (((journal->j_start + journal->j_trans_max + 1) %
4214 SB_ONDISK_JOURNAL_SIZE(sb)) >=
4215 temp_jl->j_start) {
4216 flush_used_journal_lists(sb, temp_jl);
4217 goto first_jl;
4218 } else {
4219
4220
4221
4222
4223 break;
4224 }
4225 }
4226 }
4227 flush_old_journal_lists(sb);
4228
4229 journal->j_current_jl->j_list_bitmap =
4230 get_list_bitmap(sb, journal->j_current_jl);
4231
4232 if (!(journal->j_current_jl->j_list_bitmap)) {
4233 reiserfs_panic(sb, "journal-1996",
4234 "could not get a list bitmap");
4235 }
4236
4237 atomic_set(&(journal->j_jlock), 0);
4238 unlock_journal(sb);
4239
4240 clear_bit(J_WRITERS_QUEUED, &journal->j_state);
4241 wake_up(&(journal->j_join_wait));
4242
4243 if (!flush && wait_on_commit &&
4244 journal_list_still_alive(sb, commit_trans_id)) {
4245 flush_commit_list(sb, jl, 1);
4246 }
4247 out:
4248 reiserfs_check_lock_depth(sb, "journal end2");
4249
4250 memset(th, 0, sizeof(*th));
4251
4252
4253
4254 th->t_super = sb;
4255
4256 return journal->j_errno;
4257}
4258
4259
4260void reiserfs_abort_journal(struct super_block *sb, int errno)
4261{
4262 struct reiserfs_journal *journal = SB_JOURNAL(sb);
4263 if (test_bit(J_ABORTED, &journal->j_state))
4264 return;
4265
4266 if (!journal->j_errno)
4267 journal->j_errno = errno;
4268
4269 sb->s_flags |= MS_RDONLY;
4270 set_bit(J_ABORTED, &journal->j_state);
4271
4272#ifdef CONFIG_REISERFS_CHECK
4273 dump_stack();
4274#endif
4275}
4276