forked from torvalds/linux
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfile.c
1635 lines (1450 loc) · 45.9 KB
/
file.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* This file is part of UBIFS.
*
* Copyright (C) 2006-2008 Nokia Corporation.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 as published by
* the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program; if not, write to the Free Software Foundation, Inc., 51
* Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*
* Authors: Artem Bityutskiy (Битюцкий Артём)
* Adrian Hunter
*/
/*
* This file implements VFS file and inode operations for regular files, device
* nodes and symlinks as well as address space operations.
*
* UBIFS uses 2 page flags: @PG_private and @PG_checked. @PG_private is set if
* the page is dirty and is used for optimization purposes - dirty pages are
* not budgeted so the flag shows that 'ubifs_write_end()' should not release
* the budget for this page. The @PG_checked flag is set if full budgeting is
* required for the page e.g., when it corresponds to a file hole or it is
* beyond the file size. The budgeting is done in 'ubifs_write_begin()', because
* it is OK to fail in this function, and the budget is released in
* 'ubifs_write_end()'. So the @PG_private and @PG_checked flags carry
* information about how the page was budgeted, to make it possible to release
* the budget properly.
*
* A thing to keep in mind: inode @i_mutex is locked in most VFS operations we
* implement. However, this is not true for 'ubifs_writepage()', which may be
* called with @i_mutex unlocked. For example, when flusher thread is doing
* background write-back, it calls 'ubifs_writepage()' with unlocked @i_mutex.
* At "normal" work-paths the @i_mutex is locked in 'ubifs_writepage()', e.g.
* in the "sys_write -> alloc_pages -> direct reclaim path". So, in
* 'ubifs_writepage()' we are only guaranteed that the page is locked.
*
* Similarly, @i_mutex is not always locked in 'ubifs_readpage()', e.g., the
* read-ahead path does not lock it ("sys_read -> generic_file_aio_read ->
* ondemand_readahead -> readpage"). In case of readahead, @I_SYNC flag is not
* set as well. However, UBIFS disables readahead.
*/
#include "ubifs.h"
#include <linux/mount.h>
#include <linux/slab.h>
static int read_block(struct inode *inode, void *addr, unsigned int block,
struct ubifs_data_node *dn)
{
struct ubifs_info *c = inode->i_sb->s_fs_info;
int err, len, out_len;
union ubifs_key key;
unsigned int dlen;
data_key_init(c, &key, inode->i_ino, block);
err = ubifs_tnc_lookup(c, &key, dn);
if (err) {
if (err == -ENOENT)
/* Not found, so it must be a hole */
memset(addr, 0, UBIFS_BLOCK_SIZE);
return err;
}
ubifs_assert(le64_to_cpu(dn->ch.sqnum) >
ubifs_inode(inode)->creat_sqnum);
len = le32_to_cpu(dn->size);
if (len <= 0 || len > UBIFS_BLOCK_SIZE)
goto dump;
dlen = le32_to_cpu(dn->ch.len) - UBIFS_DATA_NODE_SZ;
out_len = UBIFS_BLOCK_SIZE;
err = ubifs_decompress(c, &dn->data, dlen, addr, &out_len,
le16_to_cpu(dn->compr_type));
if (err || len != out_len)
goto dump;
/*
* Data length can be less than a full block, even for blocks that are
* not the last in the file (e.g., as a result of making a hole and
* appending data). Ensure that the remainder is zeroed out.
*/
if (len < UBIFS_BLOCK_SIZE)
memset(addr + len, 0, UBIFS_BLOCK_SIZE - len);
return 0;
dump:
ubifs_err(c, "bad data node (block %u, inode %lu)",
block, inode->i_ino);
ubifs_dump_node(c, dn);
return -EINVAL;
}
static int do_readpage(struct page *page)
{
void *addr;
int err = 0, i;
unsigned int block, beyond;
struct ubifs_data_node *dn;
struct inode *inode = page->mapping->host;
loff_t i_size = i_size_read(inode);
dbg_gen("ino %lu, pg %lu, i_size %lld, flags %#lx",
inode->i_ino, page->index, i_size, page->flags);
ubifs_assert(!PageChecked(page));
ubifs_assert(!PagePrivate(page));
addr = kmap(page);
block = page->index << UBIFS_BLOCKS_PER_PAGE_SHIFT;
beyond = (i_size + UBIFS_BLOCK_SIZE - 1) >> UBIFS_BLOCK_SHIFT;
if (block >= beyond) {
/* Reading beyond inode */
SetPageChecked(page);
memset(addr, 0, PAGE_SIZE);
goto out;
}
dn = kmalloc(UBIFS_MAX_DATA_NODE_SZ, GFP_NOFS);
if (!dn) {
err = -ENOMEM;
goto error;
}
i = 0;
while (1) {
int ret;
if (block >= beyond) {
/* Reading beyond inode */
err = -ENOENT;
memset(addr, 0, UBIFS_BLOCK_SIZE);
} else {
ret = read_block(inode, addr, block, dn);
if (ret) {
err = ret;
if (err != -ENOENT)
break;
} else if (block + 1 == beyond) {
int dlen = le32_to_cpu(dn->size);
int ilen = i_size & (UBIFS_BLOCK_SIZE - 1);
if (ilen && ilen < dlen)
memset(addr + ilen, 0, dlen - ilen);
}
}
if (++i >= UBIFS_BLOCKS_PER_PAGE)
break;
block += 1;
addr += UBIFS_BLOCK_SIZE;
}
if (err) {
struct ubifs_info *c = inode->i_sb->s_fs_info;
if (err == -ENOENT) {
/* Not found, so it must be a hole */
SetPageChecked(page);
dbg_gen("hole");
goto out_free;
}
ubifs_err(c, "cannot read page %lu of inode %lu, error %d",
page->index, inode->i_ino, err);
goto error;
}
out_free:
kfree(dn);
out:
SetPageUptodate(page);
ClearPageError(page);
flush_dcache_page(page);
kunmap(page);
return 0;
error:
kfree(dn);
ClearPageUptodate(page);
SetPageError(page);
flush_dcache_page(page);
kunmap(page);
return err;
}
/**
* release_new_page_budget - release budget of a new page.
* @c: UBIFS file-system description object
*
* This is a helper function which releases budget corresponding to the budget
* of one new page of data.
*/
static void release_new_page_budget(struct ubifs_info *c)
{
struct ubifs_budget_req req = { .recalculate = 1, .new_page = 1 };
ubifs_release_budget(c, &req);
}
/**
* release_existing_page_budget - release budget of an existing page.
* @c: UBIFS file-system description object
*
* This is a helper function which releases budget corresponding to the budget
* of changing one one page of data which already exists on the flash media.
*/
static void release_existing_page_budget(struct ubifs_info *c)
{
struct ubifs_budget_req req = { .dd_growth = c->bi.page_budget};
ubifs_release_budget(c, &req);
}
static int write_begin_slow(struct address_space *mapping,
loff_t pos, unsigned len, struct page **pagep,
unsigned flags)
{
struct inode *inode = mapping->host;
struct ubifs_info *c = inode->i_sb->s_fs_info;
pgoff_t index = pos >> PAGE_SHIFT;
struct ubifs_budget_req req = { .new_page = 1 };
int uninitialized_var(err), appending = !!(pos + len > inode->i_size);
struct page *page;
dbg_gen("ino %lu, pos %llu, len %u, i_size %lld",
inode->i_ino, pos, len, inode->i_size);
/*
* At the slow path we have to budget before locking the page, because
* budgeting may force write-back, which would wait on locked pages and
* deadlock if we had the page locked. At this point we do not know
* anything about the page, so assume that this is a new page which is
* written to a hole. This corresponds to largest budget. Later the
* budget will be amended if this is not true.
*/
if (appending)
/* We are appending data, budget for inode change */
req.dirtied_ino = 1;
err = ubifs_budget_space(c, &req);
if (unlikely(err))
return err;
page = grab_cache_page_write_begin(mapping, index, flags);
if (unlikely(!page)) {
ubifs_release_budget(c, &req);
return -ENOMEM;
}
if (!PageUptodate(page)) {
if (!(pos & ~PAGE_MASK) && len == PAGE_SIZE)
SetPageChecked(page);
else {
err = do_readpage(page);
if (err) {
unlock_page(page);
put_page(page);
ubifs_release_budget(c, &req);
return err;
}
}
SetPageUptodate(page);
ClearPageError(page);
}
if (PagePrivate(page))
/*
* The page is dirty, which means it was budgeted twice:
* o first time the budget was allocated by the task which
* made the page dirty and set the PG_private flag;
* o and then we budgeted for it for the second time at the
* very beginning of this function.
*
* So what we have to do is to release the page budget we
* allocated.
*/
release_new_page_budget(c);
else if (!PageChecked(page))
/*
* We are changing a page which already exists on the media.
* This means that changing the page does not make the amount
* of indexing information larger, and this part of the budget
* which we have already acquired may be released.
*/
ubifs_convert_page_budget(c);
if (appending) {
struct ubifs_inode *ui = ubifs_inode(inode);
/*
* 'ubifs_write_end()' is optimized from the fast-path part of
* 'ubifs_write_begin()' and expects the @ui_mutex to be locked
* if data is appended.
*/
mutex_lock(&ui->ui_mutex);
if (ui->dirty)
/*
* The inode is dirty already, so we may free the
* budget we allocated.
*/
ubifs_release_dirty_inode_budget(c, ui);
}
*pagep = page;
return 0;
}
/**
* allocate_budget - allocate budget for 'ubifs_write_begin()'.
* @c: UBIFS file-system description object
* @page: page to allocate budget for
* @ui: UBIFS inode object the page belongs to
* @appending: non-zero if the page is appended
*
* This is a helper function for 'ubifs_write_begin()' which allocates budget
* for the operation. The budget is allocated differently depending on whether
* this is appending, whether the page is dirty or not, and so on. This
* function leaves the @ui->ui_mutex locked in case of appending. Returns zero
* in case of success and %-ENOSPC in case of failure.
*/
static int allocate_budget(struct ubifs_info *c, struct page *page,
struct ubifs_inode *ui, int appending)
{
struct ubifs_budget_req req = { .fast = 1 };
if (PagePrivate(page)) {
if (!appending)
/*
* The page is dirty and we are not appending, which
* means no budget is needed at all.
*/
return 0;
mutex_lock(&ui->ui_mutex);
if (ui->dirty)
/*
* The page is dirty and we are appending, so the inode
* has to be marked as dirty. However, it is already
* dirty, so we do not need any budget. We may return,
* but @ui->ui_mutex hast to be left locked because we
* should prevent write-back from flushing the inode
* and freeing the budget. The lock will be released in
* 'ubifs_write_end()'.
*/
return 0;
/*
* The page is dirty, we are appending, the inode is clean, so
* we need to budget the inode change.
*/
req.dirtied_ino = 1;
} else {
if (PageChecked(page))
/*
* The page corresponds to a hole and does not
* exist on the media. So changing it makes
* make the amount of indexing information
* larger, and we have to budget for a new
* page.
*/
req.new_page = 1;
else
/*
* Not a hole, the change will not add any new
* indexing information, budget for page
* change.
*/
req.dirtied_page = 1;
if (appending) {
mutex_lock(&ui->ui_mutex);
if (!ui->dirty)
/*
* The inode is clean but we will have to mark
* it as dirty because we are appending. This
* needs a budget.
*/
req.dirtied_ino = 1;
}
}
return ubifs_budget_space(c, &req);
}
/*
* This function is called when a page of data is going to be written. Since
* the page of data will not necessarily go to the flash straight away, UBIFS
* has to reserve space on the media for it, which is done by means of
* budgeting.
*
* This is the hot-path of the file-system and we are trying to optimize it as
* much as possible. For this reasons it is split on 2 parts - slow and fast.
*
* There many budgeting cases:
* o a new page is appended - we have to budget for a new page and for
* changing the inode; however, if the inode is already dirty, there is
* no need to budget for it;
* o an existing clean page is changed - we have budget for it; if the page
* does not exist on the media (a hole), we have to budget for a new
* page; otherwise, we may budget for changing an existing page; the
* difference between these cases is that changing an existing page does
* not introduce anything new to the FS indexing information, so it does
* not grow, and smaller budget is acquired in this case;
* o an existing dirty page is changed - no need to budget at all, because
* the page budget has been acquired by earlier, when the page has been
* marked dirty.
*
* UBIFS budgeting sub-system may force write-back if it thinks there is no
* space to reserve. This imposes some locking restrictions and makes it
* impossible to take into account the above cases, and makes it impossible to
* optimize budgeting.
*
* The solution for this is that the fast path of 'ubifs_write_begin()' assumes
* there is a plenty of flash space and the budget will be acquired quickly,
* without forcing write-back. The slow path does not make this assumption.
*/
static int ubifs_write_begin(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned flags,
struct page **pagep, void **fsdata)
{
struct inode *inode = mapping->host;
struct ubifs_info *c = inode->i_sb->s_fs_info;
struct ubifs_inode *ui = ubifs_inode(inode);
pgoff_t index = pos >> PAGE_SHIFT;
int uninitialized_var(err), appending = !!(pos + len > inode->i_size);
int skipped_read = 0;
struct page *page;
ubifs_assert(ubifs_inode(inode)->ui_size == inode->i_size);
ubifs_assert(!c->ro_media && !c->ro_mount);
if (unlikely(c->ro_error))
return -EROFS;
/* Try out the fast-path part first */
page = grab_cache_page_write_begin(mapping, index, flags);
if (unlikely(!page))
return -ENOMEM;
if (!PageUptodate(page)) {
/* The page is not loaded from the flash */
if (!(pos & ~PAGE_MASK) && len == PAGE_SIZE) {
/*
* We change whole page so no need to load it. But we
* do not know whether this page exists on the media or
* not, so we assume the latter because it requires
* larger budget. The assumption is that it is better
* to budget a bit more than to read the page from the
* media. Thus, we are setting the @PG_checked flag
* here.
*/
SetPageChecked(page);
skipped_read = 1;
} else {
err = do_readpage(page);
if (err) {
unlock_page(page);
put_page(page);
return err;
}
}
SetPageUptodate(page);
ClearPageError(page);
}
err = allocate_budget(c, page, ui, appending);
if (unlikely(err)) {
ubifs_assert(err == -ENOSPC);
/*
* If we skipped reading the page because we were going to
* write all of it, then it is not up to date.
*/
if (skipped_read) {
ClearPageChecked(page);
ClearPageUptodate(page);
}
/*
* Budgeting failed which means it would have to force
* write-back but didn't, because we set the @fast flag in the
* request. Write-back cannot be done now, while we have the
* page locked, because it would deadlock. Unlock and free
* everything and fall-back to slow-path.
*/
if (appending) {
ubifs_assert(mutex_is_locked(&ui->ui_mutex));
mutex_unlock(&ui->ui_mutex);
}
unlock_page(page);
put_page(page);
return write_begin_slow(mapping, pos, len, pagep, flags);
}
/*
* Whee, we acquired budgeting quickly - without involving
* garbage-collection, committing or forcing write-back. We return
* with @ui->ui_mutex locked if we are appending pages, and unlocked
* otherwise. This is an optimization (slightly hacky though).
*/
*pagep = page;
return 0;
}
/**
* cancel_budget - cancel budget.
* @c: UBIFS file-system description object
* @page: page to cancel budget for
* @ui: UBIFS inode object the page belongs to
* @appending: non-zero if the page is appended
*
* This is a helper function for a page write operation. It unlocks the
* @ui->ui_mutex in case of appending.
*/
static void cancel_budget(struct ubifs_info *c, struct page *page,
struct ubifs_inode *ui, int appending)
{
if (appending) {
if (!ui->dirty)
ubifs_release_dirty_inode_budget(c, ui);
mutex_unlock(&ui->ui_mutex);
}
if (!PagePrivate(page)) {
if (PageChecked(page))
release_new_page_budget(c);
else
release_existing_page_budget(c);
}
}
static int ubifs_write_end(struct file *file, struct address_space *mapping,
loff_t pos, unsigned len, unsigned copied,
struct page *page, void *fsdata)
{
struct inode *inode = mapping->host;
struct ubifs_inode *ui = ubifs_inode(inode);
struct ubifs_info *c = inode->i_sb->s_fs_info;
loff_t end_pos = pos + len;
int appending = !!(end_pos > inode->i_size);
dbg_gen("ino %lu, pos %llu, pg %lu, len %u, copied %d, i_size %lld",
inode->i_ino, pos, page->index, len, copied, inode->i_size);
if (unlikely(copied < len && len == PAGE_SIZE)) {
/*
* VFS copied less data to the page that it intended and
* declared in its '->write_begin()' call via the @len
* argument. If the page was not up-to-date, and @len was
* @PAGE_SIZE, the 'ubifs_write_begin()' function did
* not load it from the media (for optimization reasons). This
* means that part of the page contains garbage. So read the
* page now.
*/
dbg_gen("copied %d instead of %d, read page and repeat",
copied, len);
cancel_budget(c, page, ui, appending);
ClearPageChecked(page);
/*
* Return 0 to force VFS to repeat the whole operation, or the
* error code if 'do_readpage()' fails.
*/
copied = do_readpage(page);
goto out;
}
if (!PagePrivate(page)) {
SetPagePrivate(page);
atomic_long_inc(&c->dirty_pg_cnt);
__set_page_dirty_nobuffers(page);
}
if (appending) {
i_size_write(inode, end_pos);
ui->ui_size = end_pos;
/*
* Note, we do not set @I_DIRTY_PAGES (which means that the
* inode has dirty pages), this has been done in
* '__set_page_dirty_nobuffers()'.
*/
__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
ubifs_assert(mutex_is_locked(&ui->ui_mutex));
mutex_unlock(&ui->ui_mutex);
}
out:
unlock_page(page);
put_page(page);
return copied;
}
/**
* populate_page - copy data nodes into a page for bulk-read.
* @c: UBIFS file-system description object
* @page: page
* @bu: bulk-read information
* @n: next zbranch slot
*
* This function returns %0 on success and a negative error code on failure.
*/
static int populate_page(struct ubifs_info *c, struct page *page,
struct bu_info *bu, int *n)
{
int i = 0, nn = *n, offs = bu->zbranch[0].offs, hole = 0, read = 0;
struct inode *inode = page->mapping->host;
loff_t i_size = i_size_read(inode);
unsigned int page_block;
void *addr, *zaddr;
pgoff_t end_index;
dbg_gen("ino %lu, pg %lu, i_size %lld, flags %#lx",
inode->i_ino, page->index, i_size, page->flags);
addr = zaddr = kmap(page);
end_index = (i_size - 1) >> PAGE_SHIFT;
if (!i_size || page->index > end_index) {
hole = 1;
memset(addr, 0, PAGE_SIZE);
goto out_hole;
}
page_block = page->index << UBIFS_BLOCKS_PER_PAGE_SHIFT;
while (1) {
int err, len, out_len, dlen;
if (nn >= bu->cnt) {
hole = 1;
memset(addr, 0, UBIFS_BLOCK_SIZE);
} else if (key_block(c, &bu->zbranch[nn].key) == page_block) {
struct ubifs_data_node *dn;
dn = bu->buf + (bu->zbranch[nn].offs - offs);
ubifs_assert(le64_to_cpu(dn->ch.sqnum) >
ubifs_inode(inode)->creat_sqnum);
len = le32_to_cpu(dn->size);
if (len <= 0 || len > UBIFS_BLOCK_SIZE)
goto out_err;
dlen = le32_to_cpu(dn->ch.len) - UBIFS_DATA_NODE_SZ;
out_len = UBIFS_BLOCK_SIZE;
err = ubifs_decompress(c, &dn->data, dlen, addr, &out_len,
le16_to_cpu(dn->compr_type));
if (err || len != out_len)
goto out_err;
if (len < UBIFS_BLOCK_SIZE)
memset(addr + len, 0, UBIFS_BLOCK_SIZE - len);
nn += 1;
read = (i << UBIFS_BLOCK_SHIFT) + len;
} else if (key_block(c, &bu->zbranch[nn].key) < page_block) {
nn += 1;
continue;
} else {
hole = 1;
memset(addr, 0, UBIFS_BLOCK_SIZE);
}
if (++i >= UBIFS_BLOCKS_PER_PAGE)
break;
addr += UBIFS_BLOCK_SIZE;
page_block += 1;
}
if (end_index == page->index) {
int len = i_size & (PAGE_SIZE - 1);
if (len && len < read)
memset(zaddr + len, 0, read - len);
}
out_hole:
if (hole) {
SetPageChecked(page);
dbg_gen("hole");
}
SetPageUptodate(page);
ClearPageError(page);
flush_dcache_page(page);
kunmap(page);
*n = nn;
return 0;
out_err:
ClearPageUptodate(page);
SetPageError(page);
flush_dcache_page(page);
kunmap(page);
ubifs_err(c, "bad data node (block %u, inode %lu)",
page_block, inode->i_ino);
return -EINVAL;
}
/**
* ubifs_do_bulk_read - do bulk-read.
* @c: UBIFS file-system description object
* @bu: bulk-read information
* @page1: first page to read
*
* This function returns %1 if the bulk-read is done, otherwise %0 is returned.
*/
static int ubifs_do_bulk_read(struct ubifs_info *c, struct bu_info *bu,
struct page *page1)
{
pgoff_t offset = page1->index, end_index;
struct address_space *mapping = page1->mapping;
struct inode *inode = mapping->host;
struct ubifs_inode *ui = ubifs_inode(inode);
int err, page_idx, page_cnt, ret = 0, n = 0;
int allocate = bu->buf ? 0 : 1;
loff_t isize;
err = ubifs_tnc_get_bu_keys(c, bu);
if (err)
goto out_warn;
if (bu->eof) {
/* Turn off bulk-read at the end of the file */
ui->read_in_a_row = 1;
ui->bulk_read = 0;
}
page_cnt = bu->blk_cnt >> UBIFS_BLOCKS_PER_PAGE_SHIFT;
if (!page_cnt) {
/*
* This happens when there are multiple blocks per page and the
* blocks for the first page we are looking for, are not
* together. If all the pages were like this, bulk-read would
* reduce performance, so we turn it off for a while.
*/
goto out_bu_off;
}
if (bu->cnt) {
if (allocate) {
/*
* Allocate bulk-read buffer depending on how many data
* nodes we are going to read.
*/
bu->buf_len = bu->zbranch[bu->cnt - 1].offs +
bu->zbranch[bu->cnt - 1].len -
bu->zbranch[0].offs;
ubifs_assert(bu->buf_len > 0);
ubifs_assert(bu->buf_len <= c->leb_size);
bu->buf = kmalloc(bu->buf_len, GFP_NOFS | __GFP_NOWARN);
if (!bu->buf)
goto out_bu_off;
}
err = ubifs_tnc_bulk_read(c, bu);
if (err)
goto out_warn;
}
err = populate_page(c, page1, bu, &n);
if (err)
goto out_warn;
unlock_page(page1);
ret = 1;
isize = i_size_read(inode);
if (isize == 0)
goto out_free;
end_index = ((isize - 1) >> PAGE_SHIFT);
for (page_idx = 1; page_idx < page_cnt; page_idx++) {
pgoff_t page_offset = offset + page_idx;
struct page *page;
if (page_offset > end_index)
break;
page = find_or_create_page(mapping, page_offset,
GFP_NOFS | __GFP_COLD);
if (!page)
break;
if (!PageUptodate(page))
err = populate_page(c, page, bu, &n);
unlock_page(page);
put_page(page);
if (err)
break;
}
ui->last_page_read = offset + page_idx - 1;
out_free:
if (allocate)
kfree(bu->buf);
return ret;
out_warn:
ubifs_warn(c, "ignoring error %d and skipping bulk-read", err);
goto out_free;
out_bu_off:
ui->read_in_a_row = ui->bulk_read = 0;
goto out_free;
}
/**
* ubifs_bulk_read - determine whether to bulk-read and, if so, do it.
* @page: page from which to start bulk-read.
*
* Some flash media are capable of reading sequentially at faster rates. UBIFS
* bulk-read facility is designed to take advantage of that, by reading in one
* go consecutive data nodes that are also located consecutively in the same
* LEB. This function returns %1 if a bulk-read is done and %0 otherwise.
*/
static int ubifs_bulk_read(struct page *page)
{
struct inode *inode = page->mapping->host;
struct ubifs_info *c = inode->i_sb->s_fs_info;
struct ubifs_inode *ui = ubifs_inode(inode);
pgoff_t index = page->index, last_page_read = ui->last_page_read;
struct bu_info *bu;
int err = 0, allocated = 0;
ui->last_page_read = index;
if (!c->bulk_read)
return 0;
/*
* Bulk-read is protected by @ui->ui_mutex, but it is an optimization,
* so don't bother if we cannot lock the mutex.
*/
if (!mutex_trylock(&ui->ui_mutex))
return 0;
if (index != last_page_read + 1) {
/* Turn off bulk-read if we stop reading sequentially */
ui->read_in_a_row = 1;
if (ui->bulk_read)
ui->bulk_read = 0;
goto out_unlock;
}
if (!ui->bulk_read) {
ui->read_in_a_row += 1;
if (ui->read_in_a_row < 3)
goto out_unlock;
/* Three reads in a row, so switch on bulk-read */
ui->bulk_read = 1;
}
/*
* If possible, try to use pre-allocated bulk-read information, which
* is protected by @c->bu_mutex.
*/
if (mutex_trylock(&c->bu_mutex))
bu = &c->bu;
else {
bu = kmalloc(sizeof(struct bu_info), GFP_NOFS | __GFP_NOWARN);
if (!bu)
goto out_unlock;
bu->buf = NULL;
allocated = 1;
}
bu->buf_len = c->max_bu_buf_len;
data_key_init(c, &bu->key, inode->i_ino,
page->index << UBIFS_BLOCKS_PER_PAGE_SHIFT);
err = ubifs_do_bulk_read(c, bu, page);
if (!allocated)
mutex_unlock(&c->bu_mutex);
else
kfree(bu);
out_unlock:
mutex_unlock(&ui->ui_mutex);
return err;
}
static int ubifs_readpage(struct file *file, struct page *page)
{
if (ubifs_bulk_read(page))
return 0;
do_readpage(page);
unlock_page(page);
return 0;
}
static int do_writepage(struct page *page, int len)
{
int err = 0, i, blen;
unsigned int block;
void *addr;
union ubifs_key key;
struct inode *inode = page->mapping->host;
struct ubifs_info *c = inode->i_sb->s_fs_info;
#ifdef UBIFS_DEBUG
struct ubifs_inode *ui = ubifs_inode(inode);
spin_lock(&ui->ui_lock);
ubifs_assert(page->index <= ui->synced_i_size >> PAGE_SHIFT);
spin_unlock(&ui->ui_lock);
#endif
/* Update radix tree tags */
set_page_writeback(page);
addr = kmap(page);
block = page->index << UBIFS_BLOCKS_PER_PAGE_SHIFT;
i = 0;
while (len) {
blen = min_t(int, len, UBIFS_BLOCK_SIZE);
data_key_init(c, &key, inode->i_ino, block);
err = ubifs_jnl_write_data(c, inode, &key, addr, blen);
if (err)
break;
if (++i >= UBIFS_BLOCKS_PER_PAGE)
break;
block += 1;
addr += blen;
len -= blen;
}
if (err) {
SetPageError(page);
ubifs_err(c, "cannot write page %lu of inode %lu, error %d",
page->index, inode->i_ino, err);
ubifs_ro_mode(c, err);
}
ubifs_assert(PagePrivate(page));
if (PageChecked(page))
release_new_page_budget(c);
else
release_existing_page_budget(c);
atomic_long_dec(&c->dirty_pg_cnt);
ClearPagePrivate(page);
ClearPageChecked(page);
kunmap(page);
unlock_page(page);
end_page_writeback(page);
return err;
}
/*
* When writing-back dirty inodes, VFS first writes-back pages belonging to the
* inode, then the inode itself. For UBIFS this may cause a problem. Consider a
* situation when a we have an inode with size 0, then a megabyte of data is
* appended to the inode, then write-back starts and flushes some amount of the
* dirty pages, the journal becomes full, commit happens and finishes, and then
* an unclean reboot happens. When the file system is mounted next time, the
* inode size would still be 0, but there would be many pages which are beyond
* the inode size, they would be indexed and consume flash space. Because the
* journal has been committed, the replay would not be able to detect this
* situation and correct the inode size. This means UBIFS would have to scan
* whole index and correct all inode sizes, which is long an unacceptable.
*
* To prevent situations like this, UBIFS writes pages back only if they are
* within the last synchronized inode size, i.e. the size which has been
* written to the flash media last time. Otherwise, UBIFS forces inode
* write-back, thus making sure the on-flash inode contains current inode size,
* and then keeps writing pages back.
*
* Some locking issues explanation. 'ubifs_writepage()' first is called with
* the page locked, and it locks @ui_mutex. However, write-back does take inode
* @i_mutex, which means other VFS operations may be run on this inode at the
* same time. And the problematic one is truncation to smaller size, from where
* we have to call 'truncate_setsize()', which first changes @inode->i_size,
* then drops the truncated pages. And while dropping the pages, it takes the
* page lock. This means that 'do_truncation()' cannot call 'truncate_setsize()'
* with @ui_mutex locked, because it would deadlock with 'ubifs_writepage()'.
* This means that @inode->i_size is changed while @ui_mutex is unlocked.
*
* XXX(truncate): with the new truncate sequence this is not true anymore,
* and the calls to truncate_setsize can be move around freely. They should
* be moved to the very end of the truncate sequence.
*
* But in 'ubifs_writepage()' we have to guarantee that we do not write beyond
* inode size. How do we do this if @inode->i_size may became smaller while we
* are in the middle of 'ubifs_writepage()'? The UBIFS solution is the
* @ui->ui_isize "shadow" field which UBIFS uses instead of @inode->i_size
* internally and updates it under @ui_mutex.
*
* Q: why we do not worry that if we race with truncation, we may end up with a
* situation when the inode is truncated while we are in the middle of
* 'do_writepage()', so we do write beyond inode size?
* A: If we are in the middle of 'do_writepage()', truncation would be locked
* on the page lock and it would not write the truncated inode node to the
* journal before we have finished.
*/
static int ubifs_writepage(struct page *page, struct writeback_control *wbc)
{