1
2
3
4
5
6
7
8
9
10
11
12#include <linux/module.h>
13#include <linux/compiler.h>
14#include <linux/fs.h>
15#include <linux/uaccess.h>
16#include <linux/aio.h>
17#include <linux/capability.h>
18#include <linux/kernel_stat.h>
19#include <linux/gfp.h>
20#include <linux/mm.h>
21#include <linux/swap.h>
22#include <linux/mman.h>
23#include <linux/pagemap.h>
24#include <linux/file.h>
25#include <linux/uio.h>
26#include <linux/hash.h>
27#include <linux/writeback.h>
28#include <linux/backing-dev.h>
29#include <linux/pagevec.h>
30#include <linux/blkdev.h>
31#include <linux/security.h>
32#include <linux/syscalls.h>
33#include <linux/cpuset.h>
34#include <linux/hardirq.h>
35#include <linux/memcontrol.h>
36#include <linux/mm_inline.h>
37#include "internal.h"
38
39
40
41
42#include <linux/buffer_head.h>
43
44#include <asm/mman.h>
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
117void __delete_from_page_cache(struct page *page)
118{
119 struct address_space *mapping = page->mapping;
120
121 radix_tree_delete(&mapping->page_tree, page->index);
122 page->mapping = NULL;
123 mapping->nrpages--;
124 __dec_zone_page_state(page, NR_FILE_PAGES);
125 if (PageSwapBacked(page))
126 __dec_zone_page_state(page, NR_SHMEM);
127 BUG_ON(page_mapped(page));
128
129
130
131
132
133
134
135
136 if (PageDirty(page) && mapping_cap_account_dirty(mapping)) {
137 dec_zone_page_state(page, NR_FILE_DIRTY);
138 dec_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE);
139 }
140}
141
142
143
144
145
146
147
148
149
150void delete_from_page_cache(struct page *page)
151{
152 struct address_space *mapping = page->mapping;
153 void (*freepage)(struct page *);
154
155 BUG_ON(!PageLocked(page));
156
157 freepage = mapping->a_ops->freepage;
158 spin_lock_irq(&mapping->tree_lock);
159 __delete_from_page_cache(page);
160 spin_unlock_irq(&mapping->tree_lock);
161 mem_cgroup_uncharge_cache_page(page);
162
163 if (freepage)
164 freepage(page);
165 page_cache_release(page);
166}
167EXPORT_SYMBOL(delete_from_page_cache);
168
169static int sleep_on_page(void *word)
170{
171 io_schedule();
172 return 0;
173}
174
175static int sleep_on_page_killable(void *word)
176{
177 sleep_on_page(word);
178 return fatal_signal_pending(current) ? -EINTR : 0;
179}
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196int __filemap_fdatawrite_range(struct address_space *mapping, loff_t start,
197 loff_t end, int sync_mode)
198{
199 int ret;
200 struct writeback_control wbc = {
201 .sync_mode = sync_mode,
202 .nr_to_write = LONG_MAX,
203 .range_start = start,
204 .range_end = end,
205 };
206
207 if (!mapping_cap_writeback_dirty(mapping))
208 return 0;
209
210 ret = do_writepages(mapping, &wbc);
211 return ret;
212}
213
214static inline int __filemap_fdatawrite(struct address_space *mapping,
215 int sync_mode)
216{
217 return __filemap_fdatawrite_range(mapping, 0, LLONG_MAX, sync_mode);
218}
219
220int filemap_fdatawrite(struct address_space *mapping)
221{
222 return __filemap_fdatawrite(mapping, WB_SYNC_ALL);
223}
224EXPORT_SYMBOL(filemap_fdatawrite);
225
226int filemap_fdatawrite_range(struct address_space *mapping, loff_t start,
227 loff_t end)
228{
229 return __filemap_fdatawrite_range(mapping, start, end, WB_SYNC_ALL);
230}
231EXPORT_SYMBOL(filemap_fdatawrite_range);
232
233
234
235
236
237
238
239
240int filemap_flush(struct address_space *mapping)
241{
242 return __filemap_fdatawrite(mapping, WB_SYNC_NONE);
243}
244EXPORT_SYMBOL(filemap_flush);
245
246
247
248
249
250
251
252
253
254
255int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte,
256 loff_t end_byte)
257{
258 pgoff_t index = start_byte >> PAGE_CACHE_SHIFT;
259 pgoff_t end = end_byte >> PAGE_CACHE_SHIFT;
260 struct pagevec pvec;
261 int nr_pages;
262 int ret = 0;
263
264 if (end_byte < start_byte)
265 return 0;
266
267 pagevec_init(&pvec, 0);
268 while ((index <= end) &&
269 (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
270 PAGECACHE_TAG_WRITEBACK,
271 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) {
272 unsigned i;
273
274 for (i = 0; i < nr_pages; i++) {
275 struct page *page = pvec.pages[i];
276
277
278 if (page->index > end)
279 continue;
280
281 wait_on_page_writeback(page);
282 if (TestClearPageError(page))
283 ret = -EIO;
284 }
285 pagevec_release(&pvec);
286 cond_resched();
287 }
288
289
290 if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
291 ret = -ENOSPC;
292 if (test_and_clear_bit(AS_EIO, &mapping->flags))
293 ret = -EIO;
294
295 return ret;
296}
297EXPORT_SYMBOL(filemap_fdatawait_range);
298
299
300
301
302
303
304
305
306int filemap_fdatawait(struct address_space *mapping)
307{
308 loff_t i_size = i_size_read(mapping->host);
309
310 if (i_size == 0)
311 return 0;
312
313 return filemap_fdatawait_range(mapping, 0, i_size - 1);
314}
315EXPORT_SYMBOL(filemap_fdatawait);
316
317int filemap_write_and_wait(struct address_space *mapping)
318{
319 int err = 0;
320
321 if (mapping->nrpages) {
322 err = filemap_fdatawrite(mapping);
323
324
325
326
327
328
329 if (err != -EIO) {
330 int err2 = filemap_fdatawait(mapping);
331 if (!err)
332 err = err2;
333 }
334 }
335 return err;
336}
337EXPORT_SYMBOL(filemap_write_and_wait);
338
339
340
341
342
343
344
345
346
347
348
349
350int filemap_write_and_wait_range(struct address_space *mapping,
351 loff_t lstart, loff_t lend)
352{
353 int err = 0;
354
355 if (mapping->nrpages) {
356 err = __filemap_fdatawrite_range(mapping, lstart, lend,
357 WB_SYNC_ALL);
358
359 if (err != -EIO) {
360 int err2 = filemap_fdatawait_range(mapping,
361 lstart, lend);
362 if (!err)
363 err = err2;
364 }
365 }
366 return err;
367}
368EXPORT_SYMBOL(filemap_write_and_wait_range);
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask)
386{
387 int error;
388 struct mem_cgroup *memcg = NULL;
389
390 VM_BUG_ON(!PageLocked(old));
391 VM_BUG_ON(!PageLocked(new));
392 VM_BUG_ON(new->mapping);
393
394
395
396
397
398
399
400
401
402 error = mem_cgroup_prepare_migration(old, new, &memcg, gfp_mask);
403 if (error)
404 return error;
405
406 error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
407 if (!error) {
408 struct address_space *mapping = old->mapping;
409 void (*freepage)(struct page *);
410
411 pgoff_t offset = old->index;
412 freepage = mapping->a_ops->freepage;
413
414 page_cache_get(new);
415 new->mapping = mapping;
416 new->index = offset;
417
418 spin_lock_irq(&mapping->tree_lock);
419 __delete_from_page_cache(old);
420 error = radix_tree_insert(&mapping->page_tree, offset, new);
421 BUG_ON(error);
422 mapping->nrpages++;
423 __inc_zone_page_state(new, NR_FILE_PAGES);
424 if (PageSwapBacked(new))
425 __inc_zone_page_state(new, NR_SHMEM);
426 spin_unlock_irq(&mapping->tree_lock);
427 radix_tree_preload_end();
428 if (freepage)
429 freepage(old);
430 page_cache_release(old);
431 mem_cgroup_end_migration(memcg, old, new, true);
432 } else {
433 mem_cgroup_end_migration(memcg, old, new, false);
434 }
435
436 return error;
437}
438EXPORT_SYMBOL_GPL(replace_page_cache_page);
439
440
441
442
443
444
445
446
447
448
449
450int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
451 pgoff_t offset, gfp_t gfp_mask)
452{
453 int error;
454
455 VM_BUG_ON(!PageLocked(page));
456
457 error = mem_cgroup_cache_charge(page, current->mm,
458 gfp_mask & GFP_RECLAIM_MASK);
459 if (error)
460 goto out;
461
462 error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
463 if (error == 0) {
464 page_cache_get(page);
465 page->mapping = mapping;
466 page->index = offset;
467
468 spin_lock_irq(&mapping->tree_lock);
469 error = radix_tree_insert(&mapping->page_tree, offset, page);
470 if (likely(!error)) {
471 mapping->nrpages++;
472 __inc_zone_page_state(page, NR_FILE_PAGES);
473 if (PageSwapBacked(page))
474 __inc_zone_page_state(page, NR_SHMEM);
475 spin_unlock_irq(&mapping->tree_lock);
476 } else {
477 page->mapping = NULL;
478 spin_unlock_irq(&mapping->tree_lock);
479 mem_cgroup_uncharge_cache_page(page);
480 page_cache_release(page);
481 }
482 radix_tree_preload_end();
483 } else
484 mem_cgroup_uncharge_cache_page(page);
485out:
486 return error;
487}
488EXPORT_SYMBOL(add_to_page_cache_locked);
489
490int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
491 pgoff_t offset, gfp_t gfp_mask)
492{
493 int ret;
494
495
496
497
498
499
500
501 if (mapping_cap_swap_backed(mapping))
502 SetPageSwapBacked(page);
503
504 ret = add_to_page_cache(page, mapping, offset, gfp_mask);
505 if (ret == 0) {
506 if (page_is_file_cache(page))
507 lru_cache_add_file(page);
508 else
509 lru_cache_add_anon(page);
510 }
511 return ret;
512}
513EXPORT_SYMBOL_GPL(add_to_page_cache_lru);
514
515#ifdef CONFIG_NUMA
516struct page *__page_cache_alloc(gfp_t gfp)
517{
518 int n;
519 struct page *page;
520
521 if (cpuset_do_page_mem_spread()) {
522 get_mems_allowed();
523 n = cpuset_mem_spread_node();
524 page = alloc_pages_exact_node(n, gfp, 0);
525 put_mems_allowed();
526 return page;
527 }
528 return alloc_pages(gfp, 0);
529}
530EXPORT_SYMBOL(__page_cache_alloc);
531#endif
532
533
534
535
536
537
538
539
540
541
542
543static wait_queue_head_t *page_waitqueue(struct page *page)
544{
545 const struct zone *zone = page_zone(page);
546
547 return &zone->wait_table[hash_ptr(page, zone->wait_table_bits)];
548}
549
550static inline void wake_up_page(struct page *page, int bit)
551{
552 __wake_up_bit(page_waitqueue(page), &page->flags, bit);
553}
554
555void wait_on_page_bit(struct page *page, int bit_nr)
556{
557 DEFINE_WAIT_BIT(wait, &page->flags, bit_nr);
558
559 if (test_bit(bit_nr, &page->flags))
560 __wait_on_bit(page_waitqueue(page), &wait, sleep_on_page,
561 TASK_UNINTERRUPTIBLE);
562}
563EXPORT_SYMBOL(wait_on_page_bit);
564
565
566
567
568
569
570
571
572void add_page_wait_queue(struct page *page, wait_queue_t *waiter)
573{
574 wait_queue_head_t *q = page_waitqueue(page);
575 unsigned long flags;
576
577 spin_lock_irqsave(&q->lock, flags);
578 __add_wait_queue(q, waiter);
579 spin_unlock_irqrestore(&q->lock, flags);
580}
581EXPORT_SYMBOL_GPL(add_page_wait_queue);
582
583
584
585
586
587
588
589
590
591
592
593
594
595void unlock_page(struct page *page)
596{
597 VM_BUG_ON(!PageLocked(page));
598 clear_bit_unlock(PG_locked, &page->flags);
599 smp_mb__after_clear_bit();
600 wake_up_page(page, PG_locked);
601}
602EXPORT_SYMBOL(unlock_page);
603
604
605
606
607
608void end_page_writeback(struct page *page)
609{
610 if (TestClearPageReclaim(page))
611 rotate_reclaimable_page(page);
612
613 if (!test_clear_page_writeback(page))
614 BUG();
615
616 smp_mb__after_clear_bit();
617 wake_up_page(page, PG_writeback);
618}
619EXPORT_SYMBOL(end_page_writeback);
620
621
622
623
624
625void __lock_page(struct page *page)
626{
627 DEFINE_WAIT_BIT(wait, &page->flags, PG_locked);
628
629 __wait_on_bit_lock(page_waitqueue(page), &wait, sleep_on_page,
630 TASK_UNINTERRUPTIBLE);
631}
632EXPORT_SYMBOL(__lock_page);
633
634int __lock_page_killable(struct page *page)
635{
636 DEFINE_WAIT_BIT(wait, &page->flags, PG_locked);
637
638 return __wait_on_bit_lock(page_waitqueue(page), &wait,
639 sleep_on_page_killable, TASK_KILLABLE);
640}
641EXPORT_SYMBOL_GPL(__lock_page_killable);
642
643int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
644 unsigned int flags)
645{
646 if (!(flags & FAULT_FLAG_ALLOW_RETRY)) {
647 __lock_page(page);
648 return 1;
649 } else {
650 if (!(flags & FAULT_FLAG_RETRY_NOWAIT)) {
651 up_read(&mm->mmap_sem);
652 wait_on_page_locked(page);
653 }
654 return 0;
655 }
656}
657
658
659
660
661
662
663
664
665
666struct page *find_get_page(struct address_space *mapping, pgoff_t offset)
667{
668 void **pagep;
669 struct page *page;
670
671 rcu_read_lock();
672repeat:
673 page = NULL;
674 pagep = radix_tree_lookup_slot(&mapping->page_tree, offset);
675 if (pagep) {
676 page = radix_tree_deref_slot(pagep);
677 if (unlikely(!page))
678 goto out;
679 if (radix_tree_deref_retry(page))
680 goto repeat;
681
682 if (!page_cache_get_speculative(page))
683 goto repeat;
684
685
686
687
688
689
690 if (unlikely(page != *pagep)) {
691 page_cache_release(page);
692 goto repeat;
693 }
694 }
695out:
696 rcu_read_unlock();
697
698 return page;
699}
700EXPORT_SYMBOL(find_get_page);
701
702
703
704
705
706
707
708
709
710
711
712struct page *find_lock_page(struct address_space *mapping, pgoff_t offset)
713{
714 struct page *page;
715
716repeat:
717 page = find_get_page(mapping, offset);
718 if (page) {
719 lock_page(page);
720
721 if (unlikely(page->mapping != mapping)) {
722 unlock_page(page);
723 page_cache_release(page);
724 goto repeat;
725 }
726 VM_BUG_ON(page->index != offset);
727 }
728 return page;
729}
730EXPORT_SYMBOL(find_lock_page);
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749struct page *find_or_create_page(struct address_space *mapping,
750 pgoff_t index, gfp_t gfp_mask)
751{
752 struct page *page;
753 int err;
754repeat:
755 page = find_lock_page(mapping, index);
756 if (!page) {
757 page = __page_cache_alloc(gfp_mask);
758 if (!page)
759 return NULL;
760
761
762
763
764
765
766 err = add_to_page_cache_lru(page, mapping, index,
767 (gfp_mask & GFP_RECLAIM_MASK));
768 if (unlikely(err)) {
769 page_cache_release(page);
770 page = NULL;
771 if (err == -EEXIST)
772 goto repeat;
773 }
774 }
775 return page;
776}
777EXPORT_SYMBOL(find_or_create_page);
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795unsigned find_get_pages(struct address_space *mapping, pgoff_t start,
796 unsigned int nr_pages, struct page **pages)
797{
798 unsigned int i;
799 unsigned int ret;
800 unsigned int nr_found;
801
802 rcu_read_lock();
803restart:
804 nr_found = radix_tree_gang_lookup_slot(&mapping->page_tree,
805 (void ***)pages, start, nr_pages);
806 ret = 0;
807 for (i = 0; i < nr_found; i++) {
808 struct page *page;
809repeat:
810 page = radix_tree_deref_slot((void **)pages[i]);
811 if (unlikely(!page))
812 continue;
813
814
815
816
817
818 if (radix_tree_deref_retry(page)) {
819 WARN_ON(start | i);
820 goto restart;
821 }
822
823 if (!page_cache_get_speculative(page))
824 goto repeat;
825
826
827 if (unlikely(page != *((void **)pages[i]))) {
828 page_cache_release(page);
829 goto repeat;
830 }
831
832 pages[ret] = page;
833 ret++;
834 }
835
836
837
838
839
840 if (unlikely(!ret && nr_found))
841 goto restart;
842 rcu_read_unlock();
843 return ret;
844}
845
846
847
848
849
850
851
852
853
854
855
856
857
858unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t index,
859 unsigned int nr_pages, struct page **pages)
860{
861 unsigned int i;
862 unsigned int ret;
863 unsigned int nr_found;
864
865 rcu_read_lock();
866restart:
867 nr_found = radix_tree_gang_lookup_slot(&mapping->page_tree,
868 (void ***)pages, index, nr_pages);
869 ret = 0;
870 for (i = 0; i < nr_found; i++) {
871 struct page *page;
872repeat:
873 page = radix_tree_deref_slot((void **)pages[i]);
874 if (unlikely(!page))
875 continue;
876
877
878
879
880
881 if (radix_tree_deref_retry(page))
882 goto restart;
883
884 if (!page_cache_get_speculative(page))
885 goto repeat;
886
887
888 if (unlikely(page != *((void **)pages[i]))) {
889 page_cache_release(page);
890 goto repeat;
891 }
892
893
894
895
896
897
898 if (page->mapping == NULL || page->index != index) {
899 page_cache_release(page);
900 break;
901 }
902
903 pages[ret] = page;
904 ret++;
905 index++;
906 }
907 rcu_read_unlock();
908 return ret;
909}
910EXPORT_SYMBOL(find_get_pages_contig);
911
912
913
914
915
916
917
918
919
920
921
922
923unsigned find_get_pages_tag(struct address_space *mapping, pgoff_t *index,
924 int tag, unsigned int nr_pages, struct page **pages)
925{
926 unsigned int i;
927 unsigned int ret;
928 unsigned int nr_found;
929
930 rcu_read_lock();
931restart:
932 nr_found = radix_tree_gang_lookup_tag_slot(&mapping->page_tree,
933 (void ***)pages, *index, nr_pages, tag);
934 ret = 0;
935 for (i = 0; i < nr_found; i++) {
936 struct page *page;
937repeat:
938 page = radix_tree_deref_slot((void **)pages[i]);
939 if (unlikely(!page))
940 continue;
941
942
943
944
945
946 if (radix_tree_deref_retry(page))
947 goto restart;
948
949 if (!page_cache_get_speculative(page))
950 goto repeat;
951
952
953 if (unlikely(page != *((void **)pages[i]))) {
954 page_cache_release(page);
955 goto repeat;
956 }
957
958 pages[ret] = page;
959 ret++;
960 }
961
962
963
964
965
966 if (unlikely(!ret && nr_found))
967 goto restart;
968 rcu_read_unlock();
969
970 if (ret)
971 *index = pages[ret - 1]->index + 1;
972
973 return ret;
974}
975EXPORT_SYMBOL(find_get_pages_tag);
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990struct page *
991grab_cache_page_nowait(struct address_space *mapping, pgoff_t index)
992{
993 struct page *page = find_get_page(mapping, index);
994
995 if (page) {
996 if (trylock_page(page))
997 return page;
998 page_cache_release(page);
999 return NULL;
1000 }
1001 page = __page_cache_alloc(mapping_gfp_mask(mapping) & ~__GFP_FS);
1002 if (page && add_to_page_cache_lru(page, mapping, index, GFP_NOFS)) {
1003 page_cache_release(page);
1004 page = NULL;
1005 }
1006 return page;
1007}
1008EXPORT_SYMBOL(grab_cache_page_nowait);
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025static void shrink_readahead_size_eio(struct file *filp,
1026 struct file_ra_state *ra)
1027{
1028 ra->ra_pages /= 4;
1029}
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044static void do_generic_file_read(struct file *filp, loff_t *ppos,
1045 read_descriptor_t *desc, read_actor_t actor)
1046{
1047 struct address_space *mapping = filp->f_mapping;
1048 struct inode *inode = mapping->host;
1049 struct file_ra_state *ra = &filp->f_ra;
1050 pgoff_t index;
1051 pgoff_t last_index;
1052 pgoff_t prev_index;
1053 unsigned long offset;
1054 unsigned int prev_offset;
1055 int error;
1056
1057 index = *ppos >> PAGE_CACHE_SHIFT;
1058 prev_index = ra->prev_pos >> PAGE_CACHE_SHIFT;
1059 prev_offset = ra->prev_pos & (PAGE_CACHE_SIZE-1);
1060 last_index = (*ppos + desc->count + PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT;
1061 offset = *ppos & ~PAGE_CACHE_MASK;
1062
1063 for (;;) {
1064 struct page *page;
1065 pgoff_t end_index;
1066 loff_t isize;
1067 unsigned long nr, ret;
1068
1069 cond_resched();
1070find_page:
1071 page = find_get_page(mapping, index);
1072 if (!page) {
1073 page_cache_sync_readahead(mapping,
1074 ra, filp,
1075 index, last_index - index);
1076 page = find_get_page(mapping, index);
1077 if (unlikely(page == NULL))
1078 goto no_cached_page;
1079 }
1080 if (PageReadahead(page)) {
1081 page_cache_async_readahead(mapping,
1082 ra, filp, page,
1083 index, last_index - index);
1084 }
1085 if (!PageUptodate(page)) {
1086 if (inode->i_blkbits == PAGE_CACHE_SHIFT ||
1087 !mapping->a_ops->is_partially_uptodate)
1088 goto page_not_up_to_date;
1089 if (!trylock_page(page))
1090 goto page_not_up_to_date;
1091
1092 if (!page->mapping)
1093 goto page_not_up_to_date_locked;
1094 if (!mapping->a_ops->is_partially_uptodate(page,
1095 desc, offset))
1096 goto page_not_up_to_date_locked;
1097 unlock_page(page);
1098 }
1099page_ok:
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109 isize = i_size_read(inode);
1110 end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1111 if (unlikely(!isize || index > end_index)) {
1112 page_cache_release(page);
1113 goto out;
1114 }
1115
1116
1117 nr = PAGE_CACHE_SIZE;
1118 if (index == end_index) {
1119 nr = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
1120 if (nr <= offset) {
1121 page_cache_release(page);
1122 goto out;
1123 }
1124 }
1125 nr = nr - offset;
1126
1127
1128
1129
1130
1131 if (mapping_writably_mapped(mapping))
1132 flush_dcache_page(page);
1133
1134
1135
1136
1137
1138 if (prev_index != index || offset != prev_offset)
1139 mark_page_accessed(page);
1140 prev_index = index;
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152 ret = actor(desc, page, offset, nr);
1153 offset += ret;
1154 index += offset >> PAGE_CACHE_SHIFT;
1155 offset &= ~PAGE_CACHE_MASK;
1156 prev_offset = offset;
1157
1158 page_cache_release(page);
1159 if (ret == nr && desc->count)
1160 continue;
1161 goto out;
1162
1163page_not_up_to_date:
1164
1165 error = lock_page_killable(page);
1166 if (unlikely(error))
1167 goto readpage_error;
1168
1169page_not_up_to_date_locked:
1170
1171 if (!page->mapping) {
1172 unlock_page(page);
1173 page_cache_release(page);
1174 continue;
1175 }
1176
1177
1178 if (PageUptodate(page)) {
1179 unlock_page(page);
1180 goto page_ok;
1181 }
1182
1183readpage:
1184
1185
1186
1187
1188
1189 ClearPageError(page);
1190
1191 error = mapping->a_ops->readpage(filp, page);
1192
1193 if (unlikely(error)) {
1194 if (error == AOP_TRUNCATED_PAGE) {
1195 page_cache_release(page);
1196 goto find_page;
1197 }
1198 goto readpage_error;
1199 }
1200
1201 if (!PageUptodate(page)) {
1202 error = lock_page_killable(page);
1203 if (unlikely(error))
1204 goto readpage_error;
1205 if (!PageUptodate(page)) {
1206 if (page->mapping == NULL) {
1207
1208
1209
1210 unlock_page(page);
1211 page_cache_release(page);
1212 goto find_page;
1213 }
1214 unlock_page(page);
1215 shrink_readahead_size_eio(filp, ra);
1216 error = -EIO;
1217 goto readpage_error;
1218 }
1219 unlock_page(page);
1220 }
1221
1222 goto page_ok;
1223
1224readpage_error:
1225
1226 desc->error = error;
1227 page_cache_release(page);
1228 goto out;
1229
1230no_cached_page:
1231
1232
1233
1234
1235 page = page_cache_alloc_cold(mapping);
1236 if (!page) {
1237 desc->error = -ENOMEM;
1238 goto out;
1239 }
1240 error = add_to_page_cache_lru(page, mapping,
1241 index, GFP_KERNEL);
1242 if (error) {
1243 page_cache_release(page);
1244 if (error == -EEXIST)
1245 goto find_page;
1246 desc->error = error;
1247 goto out;
1248 }
1249 goto readpage;
1250 }
1251
1252out:
1253 ra->prev_pos = prev_index;
1254 ra->prev_pos <<= PAGE_CACHE_SHIFT;
1255 ra->prev_pos |= prev_offset;
1256
1257 *ppos = ((loff_t)index << PAGE_CACHE_SHIFT) + offset;
1258 file_accessed(filp);
1259}
1260
1261int file_read_actor(read_descriptor_t *desc, struct page *page,
1262 unsigned long offset, unsigned long size)
1263{
1264 char *kaddr;
1265 unsigned long left, count = desc->count;
1266
1267 if (size > count)
1268 size = count;
1269
1270
1271
1272
1273
1274 if (!fault_in_pages_writeable(desc->arg.buf, size)) {
1275 kaddr = kmap_atomic(page, KM_USER0);
1276 left = __copy_to_user_inatomic(desc->arg.buf,
1277 kaddr + offset, size);
1278 kunmap_atomic(kaddr, KM_USER0);
1279 if (left == 0)
1280 goto success;
1281 }
1282
1283
1284 kaddr = kmap(page);
1285 left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
1286 kunmap(page);
1287
1288 if (left) {
1289 size -= left;
1290 desc->error = -EFAULT;
1291 }
1292success:
1293 desc->count = count - size;
1294 desc->written += size;
1295 desc->arg.buf += size;
1296 return size;
1297}
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310int generic_segment_checks(const struct iovec *iov,
1311 unsigned long *nr_segs, size_t *count, int access_flags)
1312{
1313 unsigned long seg;
1314 size_t cnt = 0;
1315 for (seg = 0; seg < *nr_segs; seg++) {
1316 const struct iovec *iv = &iov[seg];
1317
1318
1319
1320
1321
1322 cnt += iv->iov_len;
1323 if (unlikely((ssize_t)(cnt|iv->iov_len) < 0))
1324 return -EINVAL;
1325 if (access_ok(access_flags, iv->iov_base, iv->iov_len))
1326 continue;
1327 if (seg == 0)
1328 return -EFAULT;
1329 *nr_segs = seg;
1330 cnt -= iv->iov_len;
1331 break;
1332 }
1333 *count = cnt;
1334 return 0;
1335}
1336EXPORT_SYMBOL(generic_segment_checks);
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348ssize_t
1349generic_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
1350 unsigned long nr_segs, loff_t pos)
1351{
1352 struct file *filp = iocb->ki_filp;
1353 ssize_t retval;
1354 unsigned long seg = 0;
1355 size_t count;
1356 loff_t *ppos = &iocb->ki_pos;
1357 struct blk_plug plug;
1358
1359 count = 0;
1360 retval = generic_segment_checks(iov, &nr_segs, &count, VERIFY_WRITE);
1361 if (retval)
1362 return retval;
1363
1364 blk_start_plug(&plug);
1365
1366
1367 if (filp->f_flags & O_DIRECT) {
1368 loff_t size;
1369 struct address_space *mapping;
1370 struct inode *inode;
1371
1372 mapping = filp->f_mapping;
1373 inode = mapping->host;
1374 if (!count)
1375 goto out;
1376 size = i_size_read(inode);
1377 if (pos < size) {
1378 retval = filemap_write_and_wait_range(mapping, pos,
1379 pos + iov_length(iov, nr_segs) - 1);
1380 if (!retval) {
1381 retval = mapping->a_ops->direct_IO(READ, iocb,
1382 iov, pos, nr_segs);
1383 }
1384 if (retval > 0) {
1385 *ppos = pos + retval;
1386 count -= retval;
1387 }
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397 if (retval < 0 || !count || *ppos >= size) {
1398 file_accessed(filp);
1399 goto out;
1400 }
1401 }
1402 }
1403
1404 count = retval;
1405 for (seg = 0; seg < nr_segs; seg++) {
1406 read_descriptor_t desc;
1407 loff_t offset = 0;
1408
1409
1410
1411
1412
1413 if (count) {
1414 if (count > iov[seg].iov_len) {
1415 count -= iov[seg].iov_len;
1416 continue;
1417 }
1418 offset = count;
1419 count = 0;
1420 }
1421
1422 desc.written = 0;
1423 desc.arg.buf = iov[seg].iov_base + offset;
1424 desc.count = iov[seg].iov_len - offset;
1425 if (desc.count == 0)
1426 continue;
1427 desc.error = 0;
1428 do_generic_file_read(filp, ppos, &desc, file_read_actor);
1429 retval += desc.written;
1430 if (desc.error) {
1431 retval = retval ?: desc.error;
1432 break;
1433 }
1434 if (desc.count > 0)
1435 break;
1436 }
1437out:
1438 blk_finish_plug(&plug);
1439 return retval;
1440}
1441EXPORT_SYMBOL(generic_file_aio_read);
1442
1443static ssize_t
1444do_readahead(struct address_space *mapping, struct file *filp,
1445 pgoff_t index, unsigned long nr)
1446{
1447 if (!mapping || !mapping->a_ops || !mapping->a_ops->readpage)
1448 return -EINVAL;
1449
1450 force_page_cache_readahead(mapping, filp, index, nr);
1451 return 0;
1452}
1453
1454SYSCALL_DEFINE(readahead)(int fd, loff_t offset, size_t count)
1455{
1456 ssize_t ret;
1457 struct file *file;
1458
1459 ret = -EBADF;
1460 file = fget(fd);
1461 if (file) {
1462 if (file->f_mode & FMODE_READ) {
1463 struct address_space *mapping = file->f_mapping;
1464 pgoff_t start = offset >> PAGE_CACHE_SHIFT;
1465 pgoff_t end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
1466 unsigned long len = end - start + 1;
1467 ret = do_readahead(mapping, file, start, len);
1468 }
1469 fput(file);
1470 }
1471 return ret;
1472}
1473#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
1474asmlinkage long SyS_readahead(long fd, loff_t offset, long count)
1475{
1476 return SYSC_readahead((int) fd, offset, (size_t) count);
1477}
1478SYSCALL_ALIAS(sys_readahead, SyS_readahead);
1479#endif
1480
1481#ifdef CONFIG_MMU
1482
1483
1484
1485
1486
1487
1488
1489
1490static int page_cache_read(struct file *file, pgoff_t offset)
1491{
1492 struct address_space *mapping = file->f_mapping;
1493 struct page *page;
1494 int ret;
1495
1496 do {
1497 page = page_cache_alloc_cold(mapping);
1498 if (!page)
1499 return -ENOMEM;
1500
1501 ret = add_to_page_cache_lru(page, mapping, offset, GFP_KERNEL);
1502 if (ret == 0)
1503 ret = mapping->a_ops->readpage(file, page);
1504 else if (ret == -EEXIST)
1505 ret = 0;
1506
1507 page_cache_release(page);
1508
1509 } while (ret == AOP_TRUNCATED_PAGE);
1510
1511 return ret;
1512}
1513
1514#define MMAP_LOTSAMISS (100)
1515
1516
1517
1518
1519
1520static void do_sync_mmap_readahead(struct vm_area_struct *vma,
1521 struct file_ra_state *ra,
1522 struct file *file,
1523 pgoff_t offset)
1524{
1525 unsigned long ra_pages;
1526 struct address_space *mapping = file->f_mapping;
1527
1528
1529 if (VM_RandomReadHint(vma))
1530 return;
1531
1532 if (VM_SequentialReadHint(vma) ||
1533 offset - 1 == (ra->prev_pos >> PAGE_CACHE_SHIFT)) {
1534 page_cache_sync_readahead(mapping, ra, file, offset,
1535 ra->ra_pages);
1536 return;
1537 }
1538
1539 if (ra->mmap_miss < INT_MAX)
1540 ra->mmap_miss++;
1541
1542
1543
1544
1545
1546 if (ra->mmap_miss > MMAP_LOTSAMISS)
1547 return;
1548
1549
1550
1551
1552 ra_pages = max_sane_readahead(ra->ra_pages);
1553 if (ra_pages) {
1554 ra->start = max_t(long, 0, offset - ra_pages/2);
1555 ra->size = ra_pages;
1556 ra->async_size = 0;
1557 ra_submit(ra, mapping, file);
1558 }
1559}
1560
1561
1562
1563
1564
1565static void do_async_mmap_readahead(struct vm_area_struct *vma,
1566 struct file_ra_state *ra,
1567 struct file *file,
1568 struct page *page,
1569 pgoff_t offset)
1570{
1571 struct address_space *mapping = file->f_mapping;
1572
1573
1574 if (VM_RandomReadHint(vma))
1575 return;
1576 if (ra->mmap_miss > 0)
1577 ra->mmap_miss--;
1578 if (PageReadahead(page))
1579 page_cache_async_readahead(mapping, ra, file,
1580 page, offset, ra->ra_pages);
1581}
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595int filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
1596{
1597 int error;
1598 struct file *file = vma->vm_file;
1599 struct address_space *mapping = file->f_mapping;
1600 struct file_ra_state *ra = &file->f_ra;
1601 struct inode *inode = mapping->host;
1602 pgoff_t offset = vmf->pgoff;
1603 struct page *page;
1604 pgoff_t size;
1605 int ret = 0;
1606
1607 size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1608 if (offset >= size)
1609 return VM_FAULT_SIGBUS;
1610
1611
1612
1613
1614 page = find_get_page(mapping, offset);
1615 if (likely(page)) {
1616
1617
1618
1619
1620 do_async_mmap_readahead(vma, ra, file, page, offset);
1621 } else {
1622
1623 do_sync_mmap_readahead(vma, ra, file, offset);
1624 count_vm_event(PGMAJFAULT);
1625 ret = VM_FAULT_MAJOR;
1626retry_find:
1627 page = find_get_page(mapping, offset);
1628 if (!page)
1629 goto no_cached_page;
1630 }
1631
1632 if (!lock_page_or_retry(page, vma->vm_mm, vmf->flags)) {
1633 page_cache_release(page);
1634 return ret | VM_FAULT_RETRY;
1635 }
1636
1637
1638 if (unlikely(page->mapping != mapping)) {
1639 unlock_page(page);
1640 put_page(page);
1641 goto retry_find;
1642 }
1643 VM_BUG_ON(page->index != offset);
1644
1645
1646
1647
1648
1649 if (unlikely(!PageUptodate(page)))
1650 goto page_not_uptodate;
1651
1652
1653
1654
1655
1656 size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1657 if (unlikely(offset >= size)) {
1658 unlock_page(page);
1659 page_cache_release(page);
1660 return VM_FAULT_SIGBUS;
1661 }
1662
1663 ra->prev_pos = (loff_t)offset << PAGE_CACHE_SHIFT;
1664 vmf->page = page;
1665 return ret | VM_FAULT_LOCKED;
1666
1667no_cached_page:
1668
1669
1670
1671
1672 error = page_cache_read(file, offset);
1673
1674
1675
1676
1677
1678
1679 if (error >= 0)
1680 goto retry_find;
1681
1682
1683
1684
1685
1686
1687 if (error == -ENOMEM)
1688 return VM_FAULT_OOM;
1689 return VM_FAULT_SIGBUS;
1690
1691page_not_uptodate:
1692
1693
1694
1695
1696
1697
1698 ClearPageError(page);
1699 error = mapping->a_ops->readpage(file, page);
1700 if (!error) {
1701 wait_on_page_locked(page);
1702 if (!PageUptodate(page))
1703 error = -EIO;
1704 }
1705 page_cache_release(page);
1706
1707 if (!error || error == AOP_TRUNCATED_PAGE)
1708 goto retry_find;
1709
1710
1711 shrink_readahead_size_eio(file, ra);
1712 return VM_FAULT_SIGBUS;
1713}
1714EXPORT_SYMBOL(filemap_fault);
1715
1716const struct vm_operations_struct generic_file_vm_ops = {
1717 .fault = filemap_fault,
1718};
1719
1720
1721
1722int generic_file_mmap(struct file * file, struct vm_area_struct * vma)
1723{
1724 struct address_space *mapping = file->f_mapping;
1725
1726 if (!mapping->a_ops->readpage)
1727 return -ENOEXEC;
1728 file_accessed(file);
1729 vma->vm_ops = &generic_file_vm_ops;
1730 vma->vm_flags |= VM_CAN_NONLINEAR;
1731 return 0;
1732}
1733
1734
1735
1736
1737int generic_file_readonly_mmap(struct file *file, struct vm_area_struct *vma)
1738{
1739 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
1740 return -EINVAL;
1741 return generic_file_mmap(file, vma);
1742}
1743#else
1744int generic_file_mmap(struct file * file, struct vm_area_struct * vma)
1745{
1746 return -ENOSYS;
1747}
1748int generic_file_readonly_mmap(struct file * file, struct vm_area_struct * vma)
1749{
1750 return -ENOSYS;
1751}
1752#endif
1753
1754EXPORT_SYMBOL(generic_file_mmap);
1755EXPORT_SYMBOL(generic_file_readonly_mmap);
1756
1757static struct page *__read_cache_page(struct address_space *mapping,
1758 pgoff_t index,
1759 int (*filler)(void *,struct page*),
1760 void *data,
1761 gfp_t gfp)
1762{
1763 struct page *page;
1764 int err;
1765repeat:
1766 page = find_get_page(mapping, index);
1767 if (!page) {
1768 page = __page_cache_alloc(gfp | __GFP_COLD);
1769 if (!page)
1770 return ERR_PTR(-ENOMEM);
1771 err = add_to_page_cache_lru(page, mapping, index, GFP_KERNEL);
1772 if (unlikely(err)) {
1773 page_cache_release(page);
1774 if (err == -EEXIST)
1775 goto repeat;
1776
1777 return ERR_PTR(err);
1778 }
1779 err = filler(data, page);
1780 if (err < 0) {
1781 page_cache_release(page);
1782 page = ERR_PTR(err);
1783 }
1784 }
1785 return page;
1786}
1787
1788static struct page *do_read_cache_page(struct address_space *mapping,
1789 pgoff_t index,
1790 int (*filler)(void *,struct page*),
1791 void *data,
1792 gfp_t gfp)
1793
1794{
1795 struct page *page;
1796 int err;
1797
1798retry:
1799 page = __read_cache_page(mapping, index, filler, data, gfp);
1800 if (IS_ERR(page))
1801 return page;
1802 if (PageUptodate(page))
1803 goto out;
1804
1805 lock_page(page);
1806 if (!page->mapping) {
1807 unlock_page(page);
1808 page_cache_release(page);
1809 goto retry;
1810 }
1811 if (PageUptodate(page)) {
1812 unlock_page(page);
1813 goto out;
1814 }
1815 err = filler(data, page);
1816 if (err < 0) {
1817 page_cache_release(page);
1818 return ERR_PTR(err);
1819 }
1820out:
1821 mark_page_accessed(page);
1822 return page;
1823}
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840struct page *read_cache_page_async(struct address_space *mapping,
1841 pgoff_t index,
1842 int (*filler)(void *,struct page*),
1843 void *data)
1844{
1845 return do_read_cache_page(mapping, index, filler, data, mapping_gfp_mask(mapping));
1846}
1847EXPORT_SYMBOL(read_cache_page_async);
1848
1849static struct page *wait_on_page_read(struct page *page)
1850{
1851 if (!IS_ERR(page)) {
1852 wait_on_page_locked(page);
1853 if (!PageUptodate(page)) {
1854 page_cache_release(page);
1855 page = ERR_PTR(-EIO);
1856 }
1857 }
1858 return page;
1859}
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875struct page *read_cache_page_gfp(struct address_space *mapping,
1876 pgoff_t index,
1877 gfp_t gfp)
1878{
1879 filler_t *filler = (filler_t *)mapping->a_ops->readpage;
1880
1881 return wait_on_page_read(do_read_cache_page(mapping, index, filler, NULL, gfp));
1882}
1883EXPORT_SYMBOL(read_cache_page_gfp);
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897struct page *read_cache_page(struct address_space *mapping,
1898 pgoff_t index,
1899 int (*filler)(void *,struct page*),
1900 void *data)
1901{
1902 return wait_on_page_read(read_cache_page_async(mapping, index, filler, data));
1903}
1904EXPORT_SYMBOL(read_cache_page);
1905
1906
1907
1908
1909
1910
1911
1912int should_remove_suid(struct dentry *dentry)
1913{
1914 mode_t mode = dentry->d_inode->i_mode;
1915 int kill = 0;
1916
1917
1918 if (unlikely(mode & S_ISUID))
1919 kill = ATTR_KILL_SUID;
1920
1921
1922
1923
1924
1925 if (unlikely((mode & S_ISGID) && (mode & S_IXGRP)))
1926 kill |= ATTR_KILL_SGID;
1927
1928 if (unlikely(kill && !capable(CAP_FSETID) && S_ISREG(mode)))
1929 return kill;
1930
1931 return 0;
1932}
1933EXPORT_SYMBOL(should_remove_suid);
1934
1935static int __remove_suid(struct dentry *dentry, int kill)
1936{
1937 struct iattr newattrs;
1938
1939 newattrs.ia_valid = ATTR_FORCE | kill;
1940 return notify_change(dentry, &newattrs);
1941}
1942
1943int file_remove_suid(struct file *file)
1944{
1945 struct dentry *dentry = file->f_path.dentry;
1946 int killsuid = should_remove_suid(dentry);
1947 int killpriv = security_inode_need_killpriv(dentry);
1948 int error = 0;
1949
1950 if (killpriv < 0)
1951 return killpriv;
1952 if (killpriv)
1953 error = security_inode_killpriv(dentry);
1954 if (!error && killsuid)
1955 error = __remove_suid(dentry, killsuid);
1956
1957 return error;
1958}
1959EXPORT_SYMBOL(file_remove_suid);
1960
1961static size_t __iovec_copy_from_user_inatomic(char *vaddr,
1962 const struct iovec *iov, size_t base, size_t bytes)
1963{
1964 size_t copied = 0, left = 0;
1965
1966 while (bytes) {
1967 char __user *buf = iov->iov_base + base;
1968 int copy = min(bytes, iov->iov_len - base);
1969
1970 base = 0;
1971 left = __copy_from_user_inatomic(vaddr, buf, copy);
1972 copied += copy;
1973 bytes -= copy;
1974 vaddr += copy;
1975 iov++;
1976
1977 if (unlikely(left))
1978 break;
1979 }
1980 return copied - left;
1981}
1982
1983
1984
1985
1986
1987
1988size_t iov_iter_copy_from_user_atomic(struct page *page,
1989 struct iov_iter *i, unsigned long offset, size_t bytes)
1990{
1991 char *kaddr;
1992 size_t copied;
1993
1994 BUG_ON(!in_atomic());
1995 kaddr = kmap_atomic(page, KM_USER0);
1996 if (likely(i->nr_segs == 1)) {
1997 int left;
1998 char __user *buf = i->iov->iov_base + i->iov_offset;
1999 left = __copy_from_user_inatomic(kaddr + offset, buf, bytes);
2000 copied = bytes - left;
2001 } else {
2002 copied = __iovec_copy_from_user_inatomic(kaddr + offset,
2003 i->iov, i->iov_offset, bytes);
2004 }
2005 kunmap_atomic(kaddr, KM_USER0);
2006
2007 return copied;
2008}
2009EXPORT_SYMBOL(iov_iter_copy_from_user_atomic);
2010
2011
2012
2013
2014
2015
2016
2017size_t iov_iter_copy_from_user(struct page *page,
2018 struct iov_iter *i, unsigned long offset, size_t bytes)
2019{
2020 char *kaddr;
2021 size_t copied;
2022
2023 kaddr = kmap(page);
2024 if (likely(i->nr_segs == 1)) {
2025 int left;
2026 char __user *buf = i->iov->iov_base + i->iov_offset;
2027 left = __copy_from_user(kaddr + offset, buf, bytes);
2028 copied = bytes - left;
2029 } else {
2030 copied = __iovec_copy_from_user_inatomic(kaddr + offset,
2031 i->iov, i->iov_offset, bytes);
2032 }
2033 kunmap(page);
2034 return copied;
2035}
2036EXPORT_SYMBOL(iov_iter_copy_from_user);
2037
2038void iov_iter_advance(struct iov_iter *i, size_t bytes)
2039{
2040 BUG_ON(i->count < bytes);
2041
2042 if (likely(i->nr_segs == 1)) {
2043 i->iov_offset += bytes;
2044 i->count -= bytes;
2045 } else {
2046 const struct iovec *iov = i->iov;
2047 size_t base = i->iov_offset;
2048
2049
2050
2051
2052
2053 while (bytes || unlikely(i->count && !iov->iov_len)) {
2054 int copy;
2055
2056 copy = min(bytes, iov->iov_len - base);
2057 BUG_ON(!i->count || i->count < copy);
2058 i->count -= copy;
2059 bytes -= copy;
2060 base += copy;
2061 if (iov->iov_len == base) {
2062 iov++;
2063 base = 0;
2064 }
2065 }
2066 i->iov = iov;
2067 i->iov_offset = base;
2068 }
2069}
2070EXPORT_SYMBOL(iov_iter_advance);
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081int iov_iter_fault_in_readable(struct iov_iter *i, size_t bytes)
2082{
2083 char __user *buf = i->iov->iov_base + i->iov_offset;
2084 bytes = min(bytes, i->iov->iov_len - i->iov_offset);
2085 return fault_in_pages_readable(buf, bytes);
2086}
2087EXPORT_SYMBOL(iov_iter_fault_in_readable);
2088
2089
2090
2091
2092size_t iov_iter_single_seg_count(struct iov_iter *i)
2093{
2094 const struct iovec *iov = i->iov;
2095 if (i->nr_segs == 1)
2096 return i->count;
2097 else
2098 return min(i->count, iov->iov_len - i->iov_offset);
2099}
2100EXPORT_SYMBOL(iov_iter_single_seg_count);
2101
2102
2103
2104
2105
2106
2107
2108
2109inline int generic_write_checks(struct file *file, loff_t *pos, size_t *count, int isblk)
2110{
2111 struct inode *inode = file->f_mapping->host;
2112 unsigned long limit = rlimit(RLIMIT_FSIZE);
2113
2114 if (unlikely(*pos < 0))
2115 return -EINVAL;
2116
2117 if (!isblk) {
2118
2119 if (file->f_flags & O_APPEND)
2120 *pos = i_size_read(inode);
2121
2122 if (limit != RLIM_INFINITY) {
2123 if (*pos >= limit) {
2124 send_sig(SIGXFSZ, current, 0);
2125 return -EFBIG;
2126 }
2127 if (*count > limit - (typeof(limit))*pos) {
2128 *count = limit - (typeof(limit))*pos;
2129 }
2130 }
2131 }
2132
2133
2134
2135
2136 if (unlikely(*pos + *count > MAX_NON_LFS &&
2137 !(file->f_flags & O_LARGEFILE))) {
2138 if (*pos >= MAX_NON_LFS) {
2139 return -EFBIG;
2140 }
2141 if (*count > MAX_NON_LFS - (unsigned long)*pos) {
2142 *count = MAX_NON_LFS - (unsigned long)*pos;
2143 }
2144 }
2145
2146
2147
2148
2149
2150
2151
2152
2153 if (likely(!isblk)) {
2154 if (unlikely(*pos >= inode->i_sb->s_maxbytes)) {
2155 if (*count || *pos > inode->i_sb->s_maxbytes) {
2156 return -EFBIG;
2157 }
2158
2159 }
2160
2161 if (unlikely(*pos + *count > inode->i_sb->s_maxbytes))
2162 *count = inode->i_sb->s_maxbytes - *pos;
2163 } else {
2164#ifdef CONFIG_BLOCK
2165 loff_t isize;
2166 if (bdev_read_only(I_BDEV(inode)))
2167 return -EPERM;
2168 isize = i_size_read(inode);
2169 if (*pos >= isize) {
2170 if (*count || *pos > isize)
2171 return -ENOSPC;
2172 }
2173
2174 if (*pos + *count > isize)
2175 *count = isize - *pos;
2176#else
2177 return -EPERM;
2178#endif
2179 }
2180 return 0;
2181}
2182EXPORT_SYMBOL(generic_write_checks);
2183
2184int pagecache_write_begin(struct file *file, struct address_space *mapping,
2185 loff_t pos, unsigned len, unsigned flags,
2186 struct page **pagep, void **fsdata)
2187{
2188 const struct address_space_operations *aops = mapping->a_ops;
2189
2190 return aops->write_begin(file, mapping, pos, len, flags,
2191 pagep, fsdata);
2192}
2193EXPORT_SYMBOL(pagecache_write_begin);
2194
2195int pagecache_write_end(struct file *file, struct address_space *mapping,
2196 loff_t pos, unsigned len, unsigned copied,
2197 struct page *page, void *fsdata)
2198{
2199 const struct address_space_operations *aops = mapping->a_ops;
2200
2201 mark_page_accessed(page);
2202 return aops->write_end(file, mapping, pos, len, copied, page, fsdata);
2203}
2204EXPORT_SYMBOL(pagecache_write_end);
2205
2206ssize_t
2207generic_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
2208 unsigned long *nr_segs, loff_t pos, loff_t *ppos,
2209 size_t count, size_t ocount)
2210{
2211 struct file *file = iocb->ki_filp;
2212 struct address_space *mapping = file->f_mapping;
2213 struct inode *inode = mapping->host;
2214 ssize_t written;
2215 size_t write_len;
2216 pgoff_t end;
2217
2218 if (count != ocount)
2219 *nr_segs = iov_shorten((struct iovec *)iov, *nr_segs, count);
2220
2221 write_len = iov_length(iov, *nr_segs);
2222 end = (pos + write_len - 1) >> PAGE_CACHE_SHIFT;
2223
2224 written = filemap_write_and_wait_range(mapping, pos, pos + write_len - 1);
2225 if (written)
2226 goto out;
2227
2228
2229
2230
2231
2232
2233
2234 if (mapping->nrpages) {
2235 written = invalidate_inode_pages2_range(mapping,
2236 pos >> PAGE_CACHE_SHIFT, end);
2237
2238
2239
2240
2241 if (written) {
2242 if (written == -EBUSY)
2243 return 0;
2244 goto out;
2245 }
2246 }
2247
2248 written = mapping->a_ops->direct_IO(WRITE, iocb, iov, pos, *nr_segs);
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258 if (mapping->nrpages) {
2259 invalidate_inode_pages2_range(mapping,
2260 pos >> PAGE_CACHE_SHIFT, end);
2261 }
2262
2263 if (written > 0) {
2264 pos += written;
2265 if (pos > i_size_read(inode) && !S_ISBLK(inode->i_mode)) {
2266 i_size_write(inode, pos);
2267 mark_inode_dirty(inode);
2268 }
2269 *ppos = pos;
2270 }
2271out:
2272 return written;
2273}
2274EXPORT_SYMBOL(generic_file_direct_write);
2275
2276
2277
2278
2279
2280struct page *grab_cache_page_write_begin(struct address_space *mapping,
2281 pgoff_t index, unsigned flags)
2282{
2283 int status;
2284 struct page *page;
2285 gfp_t gfp_notmask = 0;
2286 if (flags & AOP_FLAG_NOFS)
2287 gfp_notmask = __GFP_FS;
2288repeat:
2289 page = find_lock_page(mapping, index);
2290 if (page)
2291 return page;
2292
2293 page = __page_cache_alloc(mapping_gfp_mask(mapping) & ~gfp_notmask);
2294 if (!page)
2295 return NULL;
2296 status = add_to_page_cache_lru(page, mapping, index,
2297 GFP_KERNEL & ~gfp_notmask);
2298 if (unlikely(status)) {
2299 page_cache_release(page);
2300 if (status == -EEXIST)
2301 goto repeat;
2302 return NULL;
2303 }
2304 return page;
2305}
2306EXPORT_SYMBOL(grab_cache_page_write_begin);
2307
2308static ssize_t generic_perform_write(struct file *file,
2309 struct iov_iter *i, loff_t pos)
2310{
2311 struct address_space *mapping = file->f_mapping;
2312 const struct address_space_operations *a_ops = mapping->a_ops;
2313 long status = 0;
2314 ssize_t written = 0;
2315 unsigned int flags = 0;
2316
2317
2318
2319
2320 if (segment_eq(get_fs(), KERNEL_DS))
2321 flags |= AOP_FLAG_UNINTERRUPTIBLE;
2322
2323 do {
2324 struct page *page;
2325 unsigned long offset;
2326 unsigned long bytes;
2327 size_t copied;
2328 void *fsdata;
2329
2330 offset = (pos & (PAGE_CACHE_SIZE - 1));
2331 bytes = min_t(unsigned long, PAGE_CACHE_SIZE - offset,
2332 iov_iter_count(i));
2333
2334again:
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346 if (unlikely(iov_iter_fault_in_readable(i, bytes))) {
2347 status = -EFAULT;
2348 break;
2349 }
2350
2351 status = a_ops->write_begin(file, mapping, pos, bytes, flags,
2352 &page, &fsdata);
2353 if (unlikely(status))
2354 break;
2355
2356 if (mapping_writably_mapped(mapping))
2357 flush_dcache_page(page);
2358
2359 pagefault_disable();
2360 copied = iov_iter_copy_from_user_atomic(page, i, offset, bytes);
2361 pagefault_enable();
2362 flush_dcache_page(page);
2363
2364 mark_page_accessed(page);
2365 status = a_ops->write_end(file, mapping, pos, bytes, copied,
2366 page, fsdata);
2367 if (unlikely(status < 0))
2368 break;
2369 copied = status;
2370
2371 cond_resched();
2372
2373 iov_iter_advance(i, copied);
2374 if (unlikely(copied == 0)) {
2375
2376
2377
2378
2379
2380
2381
2382
2383 bytes = min_t(unsigned long, PAGE_CACHE_SIZE - offset,
2384 iov_iter_single_seg_count(i));
2385 goto again;
2386 }
2387 pos += copied;
2388 written += copied;
2389
2390 balance_dirty_pages_ratelimited(mapping);
2391
2392 } while (iov_iter_count(i));
2393
2394 return written ? written : status;
2395}
2396
2397ssize_t
2398generic_file_buffered_write(struct kiocb *iocb, const struct iovec *iov,
2399 unsigned long nr_segs, loff_t pos, loff_t *ppos,
2400 size_t count, ssize_t written)
2401{
2402 struct file *file = iocb->ki_filp;
2403 ssize_t status;
2404 struct iov_iter i;
2405
2406 iov_iter_init(&i, iov, nr_segs, count, written);
2407 status = generic_perform_write(file, &i, pos);
2408
2409 if (likely(status >= 0)) {
2410 written += status;
2411 *ppos = pos + status;
2412 }
2413
2414 return written ? written : status;
2415}
2416EXPORT_SYMBOL(generic_file_buffered_write);
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437ssize_t __generic_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
2438 unsigned long nr_segs, loff_t *ppos)
2439{
2440 struct file *file = iocb->ki_filp;
2441 struct address_space * mapping = file->f_mapping;
2442 size_t ocount;
2443 size_t count;
2444 struct inode *inode = mapping->host;
2445 loff_t pos;
2446 ssize_t written;
2447 ssize_t err;
2448
2449 ocount = 0;
2450 err = generic_segment_checks(iov, &nr_segs, &ocount, VERIFY_READ);
2451 if (err)
2452 return err;
2453
2454 count = ocount;
2455 pos = *ppos;
2456
2457 vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
2458
2459
2460 current->backing_dev_info = mapping->backing_dev_info;
2461 written = 0;
2462
2463 err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
2464 if (err)
2465 goto out;
2466
2467 if (count == 0)
2468 goto out;
2469
2470 err = file_remove_suid(file);
2471 if (err)
2472 goto out;
2473
2474 file_update_time(file);
2475
2476
2477 if (unlikely(file->f_flags & O_DIRECT)) {
2478 loff_t endbyte;
2479 ssize_t written_buffered;
2480
2481 written = generic_file_direct_write(iocb, iov, &nr_segs, pos,
2482 ppos, count, ocount);
2483 if (written < 0 || written == count)
2484 goto out;
2485
2486
2487
2488
2489 pos += written;
2490 count -= written;
2491 written_buffered = generic_file_buffered_write(iocb, iov,
2492 nr_segs, pos, ppos, count,
2493 written);
2494
2495
2496
2497
2498
2499
2500
2501 if (written_buffered < 0) {
2502 err = written_buffered;
2503 goto out;
2504 }
2505
2506
2507
2508
2509
2510
2511 endbyte = pos + written_buffered - written - 1;
2512 err = filemap_write_and_wait_range(file->f_mapping, pos, endbyte);
2513 if (err == 0) {
2514 written = written_buffered;
2515 invalidate_mapping_pages(mapping,
2516 pos >> PAGE_CACHE_SHIFT,
2517 endbyte >> PAGE_CACHE_SHIFT);
2518 } else {
2519
2520
2521
2522
2523 }
2524 } else {
2525 written = generic_file_buffered_write(iocb, iov, nr_segs,
2526 pos, ppos, count, written);
2527 }
2528out:
2529 current->backing_dev_info = NULL;
2530 return written ? written : err;
2531}
2532EXPORT_SYMBOL(__generic_file_aio_write);
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545ssize_t generic_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
2546 unsigned long nr_segs, loff_t pos)
2547{
2548 struct file *file = iocb->ki_filp;
2549 struct inode *inode = file->f_mapping->host;
2550 struct blk_plug plug;
2551 ssize_t ret;
2552
2553 BUG_ON(iocb->ki_pos != pos);
2554
2555 mutex_lock(&inode->i_mutex);
2556 blk_start_plug(&plug);
2557 ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
2558 mutex_unlock(&inode->i_mutex);
2559
2560 if (ret > 0 || ret == -EIOCBQUEUED) {
2561 ssize_t err;
2562
2563 err = generic_write_sync(file, pos, ret);
2564 if (err < 0 && ret > 0)
2565 ret = err;
2566 }
2567 blk_finish_plug(&plug);
2568 return ret;
2569}
2570EXPORT_SYMBOL(generic_file_aio_write);
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589int try_to_release_page(struct page *page, gfp_t gfp_mask)
2590{
2591 struct address_space * const mapping = page->mapping;
2592
2593 BUG_ON(!PageLocked(page));
2594 if (PageWriteback(page))
2595 return 0;
2596
2597 if (mapping && mapping->a_ops->releasepage)
2598 return mapping->a_ops->releasepage(page, gfp_mask);
2599 return try_to_free_buffers(page);
2600}
2601
2602EXPORT_SYMBOL(try_to_release_page);
2603