4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
12 #include <linux/f2fs_fs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/mpage.h>
15 #include <linux/writeback.h>
16 #include <linux/backing-dev.h>
17 #include <linux/pagevec.h>
18 #include <linux/blkdev.h>
19 #include <linux/bio.h>
20 #include <linux/prefetch.h>
21 #include <linux/uio.h>
22 #include <linux/cleancache.h>
28 #include <trace/events/f2fs.h>
30 static void f2fs_read_end_io(struct bio
*bio
)
35 if (f2fs_bio_encrypted(bio
)) {
37 fscrypt_release_ctx(bio
->bi_private
);
39 fscrypt_decrypt_bio_pages(bio
->bi_private
, bio
);
44 bio_for_each_segment_all(bvec
, bio
, i
) {
45 struct page
*page
= bvec
->bv_page
;
48 SetPageUptodate(page
);
50 ClearPageUptodate(page
);
58 static void f2fs_write_end_io(struct bio
*bio
)
60 struct f2fs_sb_info
*sbi
= bio
->bi_private
;
64 bio_for_each_segment_all(bvec
, bio
, i
) {
65 struct page
*page
= bvec
->bv_page
;
67 fscrypt_pullback_bio_page(&page
, true);
69 if (unlikely(bio
->bi_error
)) {
70 set_bit(AS_EIO
, &page
->mapping
->flags
);
71 f2fs_stop_checkpoint(sbi
, true);
73 end_page_writeback(page
);
75 if (atomic_dec_and_test(&sbi
->nr_wb_bios
) &&
76 wq_has_sleeper(&sbi
->cp_wait
))
77 wake_up(&sbi
->cp_wait
);
83 * Low-level block read/write IO operations.
85 static struct bio
*__bio_alloc(struct f2fs_sb_info
*sbi
, block_t blk_addr
,
86 int npages
, bool is_read
)
90 bio
= f2fs_bio_alloc(npages
);
92 bio
->bi_bdev
= sbi
->sb
->s_bdev
;
93 bio
->bi_iter
.bi_sector
= SECTOR_FROM_BLOCK(blk_addr
);
94 bio
->bi_end_io
= is_read
? f2fs_read_end_io
: f2fs_write_end_io
;
95 bio
->bi_private
= is_read
? NULL
: sbi
;
100 static inline void __submit_bio(struct f2fs_sb_info
*sbi
, int rw
,
104 atomic_inc(&sbi
->nr_wb_bios
);
108 static void __submit_merged_bio(struct f2fs_bio_info
*io
)
110 struct f2fs_io_info
*fio
= &io
->fio
;
115 if (is_read_io(fio
->rw
))
116 trace_f2fs_submit_read_bio(io
->sbi
->sb
, fio
, io
->bio
);
118 trace_f2fs_submit_write_bio(io
->sbi
->sb
, fio
, io
->bio
);
120 __submit_bio(io
->sbi
, fio
->rw
, io
->bio
);
124 static bool __has_merged_page(struct f2fs_bio_info
*io
, struct inode
*inode
,
125 struct page
*page
, nid_t ino
)
127 struct bio_vec
*bvec
;
134 if (!inode
&& !page
&& !ino
)
137 bio_for_each_segment_all(bvec
, io
->bio
, i
) {
139 if (bvec
->bv_page
->mapping
)
140 target
= bvec
->bv_page
;
142 target
= fscrypt_control_page(bvec
->bv_page
);
144 if (inode
&& inode
== target
->mapping
->host
)
146 if (page
&& page
== target
)
148 if (ino
&& ino
== ino_of_node(target
))
155 static bool has_merged_page(struct f2fs_sb_info
*sbi
, struct inode
*inode
,
156 struct page
*page
, nid_t ino
,
159 enum page_type btype
= PAGE_TYPE_OF_BIO(type
);
160 struct f2fs_bio_info
*io
= &sbi
->write_io
[btype
];
163 down_read(&io
->io_rwsem
);
164 ret
= __has_merged_page(io
, inode
, page
, ino
);
165 up_read(&io
->io_rwsem
);
169 static void __f2fs_submit_merged_bio(struct f2fs_sb_info
*sbi
,
170 struct inode
*inode
, struct page
*page
,
171 nid_t ino
, enum page_type type
, int rw
)
173 enum page_type btype
= PAGE_TYPE_OF_BIO(type
);
174 struct f2fs_bio_info
*io
;
176 io
= is_read_io(rw
) ? &sbi
->read_io
: &sbi
->write_io
[btype
];
178 down_write(&io
->io_rwsem
);
180 if (!__has_merged_page(io
, inode
, page
, ino
))
183 /* change META to META_FLUSH in the checkpoint procedure */
184 if (type
>= META_FLUSH
) {
185 io
->fio
.type
= META_FLUSH
;
186 if (test_opt(sbi
, NOBARRIER
))
187 io
->fio
.rw
= WRITE_FLUSH
| REQ_META
| REQ_PRIO
;
189 io
->fio
.rw
= WRITE_FLUSH_FUA
| REQ_META
| REQ_PRIO
;
191 __submit_merged_bio(io
);
193 up_write(&io
->io_rwsem
);
196 void f2fs_submit_merged_bio(struct f2fs_sb_info
*sbi
, enum page_type type
,
199 __f2fs_submit_merged_bio(sbi
, NULL
, NULL
, 0, type
, rw
);
202 void f2fs_submit_merged_bio_cond(struct f2fs_sb_info
*sbi
,
203 struct inode
*inode
, struct page
*page
,
204 nid_t ino
, enum page_type type
, int rw
)
206 if (has_merged_page(sbi
, inode
, page
, ino
, type
))
207 __f2fs_submit_merged_bio(sbi
, inode
, page
, ino
, type
, rw
);
210 void f2fs_flush_merged_bios(struct f2fs_sb_info
*sbi
)
212 f2fs_submit_merged_bio(sbi
, DATA
, WRITE
);
213 f2fs_submit_merged_bio(sbi
, NODE
, WRITE
);
214 f2fs_submit_merged_bio(sbi
, META
, WRITE
);
218 * Fill the locked page with data located in the block address.
219 * Return unlocked page.
221 int f2fs_submit_page_bio(struct f2fs_io_info
*fio
)
224 struct page
*page
= fio
->encrypted_page
?
225 fio
->encrypted_page
: fio
->page
;
227 trace_f2fs_submit_page_bio(page
, fio
);
228 f2fs_trace_ios(fio
, 0);
230 /* Allocate a new bio */
231 bio
= __bio_alloc(fio
->sbi
, fio
->new_blkaddr
, 1, is_read_io(fio
->rw
));
233 if (bio_add_page(bio
, page
, PAGE_SIZE
, 0) < PAGE_SIZE
) {
238 __submit_bio(fio
->sbi
, fio
->rw
, bio
);
242 void f2fs_submit_page_mbio(struct f2fs_io_info
*fio
)
244 struct f2fs_sb_info
*sbi
= fio
->sbi
;
245 enum page_type btype
= PAGE_TYPE_OF_BIO(fio
->type
);
246 struct f2fs_bio_info
*io
;
247 bool is_read
= is_read_io(fio
->rw
);
248 struct page
*bio_page
;
250 io
= is_read
? &sbi
->read_io
: &sbi
->write_io
[btype
];
252 if (fio
->old_blkaddr
!= NEW_ADDR
)
253 verify_block_addr(sbi
, fio
->old_blkaddr
);
254 verify_block_addr(sbi
, fio
->new_blkaddr
);
256 down_write(&io
->io_rwsem
);
258 if (io
->bio
&& (io
->last_block_in_bio
!= fio
->new_blkaddr
- 1 ||
259 io
->fio
.rw
!= fio
->rw
))
260 __submit_merged_bio(io
);
262 if (io
->bio
== NULL
) {
263 int bio_blocks
= MAX_BIO_BLOCKS(sbi
);
265 io
->bio
= __bio_alloc(sbi
, fio
->new_blkaddr
,
266 bio_blocks
, is_read
);
270 bio_page
= fio
->encrypted_page
? fio
->encrypted_page
: fio
->page
;
272 if (bio_add_page(io
->bio
, bio_page
, PAGE_SIZE
, 0) <
274 __submit_merged_bio(io
);
278 io
->last_block_in_bio
= fio
->new_blkaddr
;
279 f2fs_trace_ios(fio
, 0);
281 up_write(&io
->io_rwsem
);
282 trace_f2fs_submit_page_mbio(fio
->page
, fio
);
285 static void __set_data_blkaddr(struct dnode_of_data
*dn
)
287 struct f2fs_node
*rn
= F2FS_NODE(dn
->node_page
);
290 /* Get physical address of data block */
291 addr_array
= blkaddr_in_node(rn
);
292 addr_array
[dn
->ofs_in_node
] = cpu_to_le32(dn
->data_blkaddr
);
296 * Lock ordering for the change of data block address:
299 * update block addresses in the node page
301 void set_data_blkaddr(struct dnode_of_data
*dn
)
303 f2fs_wait_on_page_writeback(dn
->node_page
, NODE
, true);
304 __set_data_blkaddr(dn
);
305 if (set_page_dirty(dn
->node_page
))
306 dn
->node_changed
= true;
309 void f2fs_update_data_blkaddr(struct dnode_of_data
*dn
, block_t blkaddr
)
311 dn
->data_blkaddr
= blkaddr
;
312 set_data_blkaddr(dn
);
313 f2fs_update_extent_cache(dn
);
316 /* dn->ofs_in_node will be returned with up-to-date last block pointer */
317 int reserve_new_blocks(struct dnode_of_data
*dn
, blkcnt_t count
)
319 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
324 if (unlikely(is_inode_flag_set(F2FS_I(dn
->inode
), FI_NO_ALLOC
)))
326 if (unlikely(!inc_valid_block_count(sbi
, dn
->inode
, &count
)))
329 trace_f2fs_reserve_new_blocks(dn
->inode
, dn
->nid
,
330 dn
->ofs_in_node
, count
);
332 f2fs_wait_on_page_writeback(dn
->node_page
, NODE
, true);
334 for (; count
> 0; dn
->ofs_in_node
++) {
336 datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
337 if (blkaddr
== NULL_ADDR
) {
338 dn
->data_blkaddr
= NEW_ADDR
;
339 __set_data_blkaddr(dn
);
344 if (set_page_dirty(dn
->node_page
))
345 dn
->node_changed
= true;
347 mark_inode_dirty(dn
->inode
);
352 /* Should keep dn->ofs_in_node unchanged */
353 int reserve_new_block(struct dnode_of_data
*dn
)
355 unsigned int ofs_in_node
= dn
->ofs_in_node
;
358 ret
= reserve_new_blocks(dn
, 1);
359 dn
->ofs_in_node
= ofs_in_node
;
363 int f2fs_reserve_block(struct dnode_of_data
*dn
, pgoff_t index
)
365 bool need_put
= dn
->inode_page
? false : true;
368 err
= get_dnode_of_data(dn
, index
, ALLOC_NODE
);
372 if (dn
->data_blkaddr
== NULL_ADDR
)
373 err
= reserve_new_block(dn
);
379 int f2fs_get_block(struct dnode_of_data
*dn
, pgoff_t index
)
381 struct extent_info ei
;
382 struct inode
*inode
= dn
->inode
;
384 if (f2fs_lookup_extent_cache(inode
, index
, &ei
)) {
385 dn
->data_blkaddr
= ei
.blk
+ index
- ei
.fofs
;
389 return f2fs_reserve_block(dn
, index
);
392 struct page
*get_read_data_page(struct inode
*inode
, pgoff_t index
,
393 int rw
, bool for_write
)
395 struct address_space
*mapping
= inode
->i_mapping
;
396 struct dnode_of_data dn
;
398 struct extent_info ei
;
400 struct f2fs_io_info fio
= {
401 .sbi
= F2FS_I_SB(inode
),
404 .encrypted_page
= NULL
,
407 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
408 return read_mapping_page(mapping
, index
, NULL
);
410 page
= f2fs_grab_cache_page(mapping
, index
, for_write
);
412 return ERR_PTR(-ENOMEM
);
414 if (f2fs_lookup_extent_cache(inode
, index
, &ei
)) {
415 dn
.data_blkaddr
= ei
.blk
+ index
- ei
.fofs
;
419 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
420 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
425 if (unlikely(dn
.data_blkaddr
== NULL_ADDR
)) {
430 if (PageUptodate(page
)) {
436 * A new dentry page is allocated but not able to be written, since its
437 * new inode page couldn't be allocated due to -ENOSPC.
438 * In such the case, its blkaddr can be remained as NEW_ADDR.
439 * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata.
441 if (dn
.data_blkaddr
== NEW_ADDR
) {
442 zero_user_segment(page
, 0, PAGE_SIZE
);
443 SetPageUptodate(page
);
448 fio
.new_blkaddr
= fio
.old_blkaddr
= dn
.data_blkaddr
;
450 err
= f2fs_submit_page_bio(&fio
);
456 f2fs_put_page(page
, 1);
460 struct page
*find_data_page(struct inode
*inode
, pgoff_t index
)
462 struct address_space
*mapping
= inode
->i_mapping
;
465 page
= find_get_page(mapping
, index
);
466 if (page
&& PageUptodate(page
))
468 f2fs_put_page(page
, 0);
470 page
= get_read_data_page(inode
, index
, READ_SYNC
, false);
474 if (PageUptodate(page
))
477 wait_on_page_locked(page
);
478 if (unlikely(!PageUptodate(page
))) {
479 f2fs_put_page(page
, 0);
480 return ERR_PTR(-EIO
);
486 * If it tries to access a hole, return an error.
487 * Because, the callers, functions in dir.c and GC, should be able to know
488 * whether this page exists or not.
490 struct page
*get_lock_data_page(struct inode
*inode
, pgoff_t index
,
493 struct address_space
*mapping
= inode
->i_mapping
;
496 page
= get_read_data_page(inode
, index
, READ_SYNC
, for_write
);
500 /* wait for read completion */
502 if (unlikely(!PageUptodate(page
))) {
503 f2fs_put_page(page
, 1);
504 return ERR_PTR(-EIO
);
506 if (unlikely(page
->mapping
!= mapping
)) {
507 f2fs_put_page(page
, 1);
514 * Caller ensures that this data page is never allocated.
515 * A new zero-filled data page is allocated in the page cache.
517 * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
519 * Note that, ipage is set only by make_empty_dir, and if any error occur,
520 * ipage should be released by this function.
522 struct page
*get_new_data_page(struct inode
*inode
,
523 struct page
*ipage
, pgoff_t index
, bool new_i_size
)
525 struct address_space
*mapping
= inode
->i_mapping
;
527 struct dnode_of_data dn
;
530 page
= f2fs_grab_cache_page(mapping
, index
, true);
533 * before exiting, we should make sure ipage will be released
534 * if any error occur.
536 f2fs_put_page(ipage
, 1);
537 return ERR_PTR(-ENOMEM
);
540 set_new_dnode(&dn
, inode
, ipage
, NULL
, 0);
541 err
= f2fs_reserve_block(&dn
, index
);
543 f2fs_put_page(page
, 1);
549 if (PageUptodate(page
))
552 if (dn
.data_blkaddr
== NEW_ADDR
) {
553 zero_user_segment(page
, 0, PAGE_SIZE
);
554 SetPageUptodate(page
);
556 f2fs_put_page(page
, 1);
558 /* if ipage exists, blkaddr should be NEW_ADDR */
559 f2fs_bug_on(F2FS_I_SB(inode
), ipage
);
560 page
= get_lock_data_page(inode
, index
, true);
565 if (new_i_size
&& i_size_read(inode
) <
566 ((loff_t
)(index
+ 1) << PAGE_SHIFT
)) {
567 i_size_write(inode
, ((loff_t
)(index
+ 1) << PAGE_SHIFT
));
568 /* Only the directory inode sets new_i_size */
569 set_inode_flag(F2FS_I(inode
), FI_UPDATE_DIR
);
574 static int __allocate_data_block(struct dnode_of_data
*dn
)
576 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
577 struct f2fs_summary sum
;
579 int seg
= CURSEG_WARM_DATA
;
583 if (unlikely(is_inode_flag_set(F2FS_I(dn
->inode
), FI_NO_ALLOC
)))
586 dn
->data_blkaddr
= datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
587 if (dn
->data_blkaddr
== NEW_ADDR
)
590 if (unlikely(!inc_valid_block_count(sbi
, dn
->inode
, &count
)))
594 get_node_info(sbi
, dn
->nid
, &ni
);
595 set_summary(&sum
, dn
->nid
, dn
->ofs_in_node
, ni
.version
);
597 if (dn
->ofs_in_node
== 0 && dn
->inode_page
== dn
->node_page
)
598 seg
= CURSEG_DIRECT_IO
;
600 allocate_data_block(sbi
, NULL
, dn
->data_blkaddr
, &dn
->data_blkaddr
,
602 set_data_blkaddr(dn
);
605 fofs
= start_bidx_of_node(ofs_of_node(dn
->node_page
), dn
->inode
) +
607 if (i_size_read(dn
->inode
) < ((loff_t
)(fofs
+ 1) << PAGE_SHIFT
))
608 i_size_write(dn
->inode
,
609 ((loff_t
)(fofs
+ 1) << PAGE_SHIFT
));
613 ssize_t
f2fs_preallocate_blocks(struct kiocb
*iocb
, struct iov_iter
*from
)
615 struct inode
*inode
= file_inode(iocb
->ki_filp
);
616 struct f2fs_map_blocks map
;
619 map
.m_lblk
= F2FS_BLK_ALIGN(iocb
->ki_pos
);
620 map
.m_len
= F2FS_BYTES_TO_BLK(iov_iter_count(from
));
621 map
.m_next_pgofs
= NULL
;
623 if (f2fs_encrypted_inode(inode
))
626 if (iocb
->ki_flags
& IOCB_DIRECT
) {
627 ret
= f2fs_convert_inline_inode(inode
);
630 return f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_DIO
);
632 if (iocb
->ki_pos
+ iov_iter_count(from
) > MAX_INLINE_DATA
) {
633 ret
= f2fs_convert_inline_inode(inode
);
637 if (!f2fs_has_inline_data(inode
))
638 return f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_AIO
);
643 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
644 * f2fs_map_blocks structure.
645 * If original data blocks are allocated, then give them to blockdev.
647 * a. preallocate requested block addresses
648 * b. do not use extent cache for better performance
649 * c. give the block addresses to blockdev
651 int f2fs_map_blocks(struct inode
*inode
, struct f2fs_map_blocks
*map
,
652 int create
, int flag
)
654 unsigned int maxblocks
= map
->m_len
;
655 struct dnode_of_data dn
;
656 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
657 int mode
= create
? ALLOC_NODE
: LOOKUP_NODE_RA
;
658 pgoff_t pgofs
, end_offset
, end
;
659 int err
= 0, ofs
= 1;
660 unsigned int ofs_in_node
, last_ofs_in_node
;
662 struct extent_info ei
;
663 bool allocated
= false;
669 /* it only supports block size == page size */
670 pgofs
= (pgoff_t
)map
->m_lblk
;
671 end
= pgofs
+ maxblocks
;
673 if (!create
&& f2fs_lookup_extent_cache(inode
, pgofs
, &ei
)) {
674 map
->m_pblk
= ei
.blk
+ pgofs
- ei
.fofs
;
675 map
->m_len
= min((pgoff_t
)maxblocks
, ei
.fofs
+ ei
.len
- pgofs
);
676 map
->m_flags
= F2FS_MAP_MAPPED
;
684 /* When reading holes, we need its node page */
685 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
686 err
= get_dnode_of_data(&dn
, pgofs
, mode
);
688 if (flag
== F2FS_GET_BLOCK_BMAP
)
690 if (err
== -ENOENT
) {
692 if (map
->m_next_pgofs
)
694 get_next_page_offset(&dn
, pgofs
);
700 ofs_in_node
= dn
.ofs_in_node
;
701 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
704 blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
706 if (blkaddr
== NEW_ADDR
|| blkaddr
== NULL_ADDR
) {
708 if (unlikely(f2fs_cp_error(sbi
))) {
712 if (flag
== F2FS_GET_BLOCK_PRE_AIO
) {
713 if (blkaddr
== NULL_ADDR
) {
715 last_ofs_in_node
= dn
.ofs_in_node
;
718 err
= __allocate_data_block(&dn
);
720 set_inode_flag(F2FS_I(inode
),
727 map
->m_flags
= F2FS_MAP_NEW
;
728 blkaddr
= dn
.data_blkaddr
;
730 if (flag
== F2FS_GET_BLOCK_BMAP
) {
734 if (flag
== F2FS_GET_BLOCK_FIEMAP
&&
735 blkaddr
== NULL_ADDR
) {
736 if (map
->m_next_pgofs
)
737 *map
->m_next_pgofs
= pgofs
+ 1;
739 if (flag
!= F2FS_GET_BLOCK_FIEMAP
||
745 if (flag
== F2FS_GET_BLOCK_PRE_AIO
)
748 if (map
->m_len
== 0) {
749 /* preallocated unwritten block should be mapped for fiemap. */
750 if (blkaddr
== NEW_ADDR
)
751 map
->m_flags
|= F2FS_MAP_UNWRITTEN
;
752 map
->m_flags
|= F2FS_MAP_MAPPED
;
754 map
->m_pblk
= blkaddr
;
756 } else if ((map
->m_pblk
!= NEW_ADDR
&&
757 blkaddr
== (map
->m_pblk
+ ofs
)) ||
758 (map
->m_pblk
== NEW_ADDR
&& blkaddr
== NEW_ADDR
) ||
759 flag
== F2FS_GET_BLOCK_PRE_DIO
) {
770 /* preallocate blocks in batch for one dnode page */
771 if (flag
== F2FS_GET_BLOCK_PRE_AIO
&&
772 (pgofs
== end
|| dn
.ofs_in_node
== end_offset
)) {
774 dn
.ofs_in_node
= ofs_in_node
;
775 err
= reserve_new_blocks(&dn
, prealloc
);
779 map
->m_len
+= dn
.ofs_in_node
- ofs_in_node
;
780 if (prealloc
&& dn
.ofs_in_node
!= last_ofs_in_node
+ 1) {
784 dn
.ofs_in_node
= end_offset
;
789 else if (dn
.ofs_in_node
< end_offset
)
793 sync_inode_page(&dn
);
798 f2fs_balance_fs(sbi
, allocated
);
805 sync_inode_page(&dn
);
810 f2fs_balance_fs(sbi
, allocated
);
813 trace_f2fs_map_blocks(inode
, map
, err
);
817 static int __get_data_block(struct inode
*inode
, sector_t iblock
,
818 struct buffer_head
*bh
, int create
, int flag
,
821 struct f2fs_map_blocks map
;
825 map
.m_len
= bh
->b_size
>> inode
->i_blkbits
;
826 map
.m_next_pgofs
= next_pgofs
;
828 ret
= f2fs_map_blocks(inode
, &map
, create
, flag
);
830 map_bh(bh
, inode
->i_sb
, map
.m_pblk
);
831 bh
->b_state
= (bh
->b_state
& ~F2FS_MAP_FLAGS
) | map
.m_flags
;
832 bh
->b_size
= map
.m_len
<< inode
->i_blkbits
;
837 static int get_data_block(struct inode
*inode
, sector_t iblock
,
838 struct buffer_head
*bh_result
, int create
, int flag
,
841 return __get_data_block(inode
, iblock
, bh_result
, create
,
845 static int get_data_block_dio(struct inode
*inode
, sector_t iblock
,
846 struct buffer_head
*bh_result
, int create
)
848 return __get_data_block(inode
, iblock
, bh_result
, create
,
849 F2FS_GET_BLOCK_DIO
, NULL
);
852 static int get_data_block_bmap(struct inode
*inode
, sector_t iblock
,
853 struct buffer_head
*bh_result
, int create
)
855 /* Block number less than F2FS MAX BLOCKS */
856 if (unlikely(iblock
>= F2FS_I_SB(inode
)->max_file_blocks
))
859 return __get_data_block(inode
, iblock
, bh_result
, create
,
860 F2FS_GET_BLOCK_BMAP
, NULL
);
863 static inline sector_t
logical_to_blk(struct inode
*inode
, loff_t offset
)
865 return (offset
>> inode
->i_blkbits
);
868 static inline loff_t
blk_to_logical(struct inode
*inode
, sector_t blk
)
870 return (blk
<< inode
->i_blkbits
);
873 int f2fs_fiemap(struct inode
*inode
, struct fiemap_extent_info
*fieinfo
,
876 struct buffer_head map_bh
;
877 sector_t start_blk
, last_blk
;
880 u64 logical
= 0, phys
= 0, size
= 0;
884 ret
= fiemap_check_flags(fieinfo
, FIEMAP_FLAG_SYNC
);
888 if (f2fs_has_inline_data(inode
)) {
889 ret
= f2fs_inline_data_fiemap(inode
, fieinfo
, start
, len
);
896 isize
= i_size_read(inode
);
900 if (start
+ len
> isize
)
903 if (logical_to_blk(inode
, len
) == 0)
904 len
= blk_to_logical(inode
, 1);
906 start_blk
= logical_to_blk(inode
, start
);
907 last_blk
= logical_to_blk(inode
, start
+ len
- 1);
910 memset(&map_bh
, 0, sizeof(struct buffer_head
));
913 ret
= get_data_block(inode
, start_blk
, &map_bh
, 0,
914 F2FS_GET_BLOCK_FIEMAP
, &next_pgofs
);
919 if (!buffer_mapped(&map_bh
)) {
920 start_blk
= next_pgofs
;
921 /* Go through holes util pass the EOF */
922 if (blk_to_logical(inode
, start_blk
) < isize
)
924 /* Found a hole beyond isize means no more extents.
925 * Note that the premise is that filesystems don't
926 * punch holes beyond isize and keep size unchanged.
928 flags
|= FIEMAP_EXTENT_LAST
;
932 if (f2fs_encrypted_inode(inode
))
933 flags
|= FIEMAP_EXTENT_DATA_ENCRYPTED
;
935 ret
= fiemap_fill_next_extent(fieinfo
, logical
,
939 if (start_blk
> last_blk
|| ret
)
942 logical
= blk_to_logical(inode
, start_blk
);
943 phys
= blk_to_logical(inode
, map_bh
.b_blocknr
);
944 size
= map_bh
.b_size
;
946 if (buffer_unwritten(&map_bh
))
947 flags
= FIEMAP_EXTENT_UNWRITTEN
;
949 start_blk
+= logical_to_blk(inode
, size
);
953 if (fatal_signal_pending(current
))
966 * This function was originally taken from fs/mpage.c, and customized for f2fs.
967 * Major change was from block_size == page_size in f2fs by default.
969 static int f2fs_mpage_readpages(struct address_space
*mapping
,
970 struct list_head
*pages
, struct page
*page
,
973 struct bio
*bio
= NULL
;
975 sector_t last_block_in_bio
= 0;
976 struct inode
*inode
= mapping
->host
;
977 const unsigned blkbits
= inode
->i_blkbits
;
978 const unsigned blocksize
= 1 << blkbits
;
979 sector_t block_in_file
;
981 sector_t last_block_in_file
;
983 struct block_device
*bdev
= inode
->i_sb
->s_bdev
;
984 struct f2fs_map_blocks map
;
990 map
.m_next_pgofs
= NULL
;
992 for (page_idx
= 0; nr_pages
; page_idx
++, nr_pages
--) {
994 prefetchw(&page
->flags
);
996 page
= list_entry(pages
->prev
, struct page
, lru
);
997 list_del(&page
->lru
);
998 if (add_to_page_cache_lru(page
, mapping
,
999 page
->index
, GFP_KERNEL
))
1003 block_in_file
= (sector_t
)page
->index
;
1004 last_block
= block_in_file
+ nr_pages
;
1005 last_block_in_file
= (i_size_read(inode
) + blocksize
- 1) >>
1007 if (last_block
> last_block_in_file
)
1008 last_block
= last_block_in_file
;
1011 * Map blocks using the previous result first.
1013 if ((map
.m_flags
& F2FS_MAP_MAPPED
) &&
1014 block_in_file
> map
.m_lblk
&&
1015 block_in_file
< (map
.m_lblk
+ map
.m_len
))
1019 * Then do more f2fs_map_blocks() calls until we are
1020 * done with this page.
1024 if (block_in_file
< last_block
) {
1025 map
.m_lblk
= block_in_file
;
1026 map
.m_len
= last_block
- block_in_file
;
1028 if (f2fs_map_blocks(inode
, &map
, 0,
1029 F2FS_GET_BLOCK_READ
))
1030 goto set_error_page
;
1033 if ((map
.m_flags
& F2FS_MAP_MAPPED
)) {
1034 block_nr
= map
.m_pblk
+ block_in_file
- map
.m_lblk
;
1035 SetPageMappedToDisk(page
);
1037 if (!PageUptodate(page
) && !cleancache_get_page(page
)) {
1038 SetPageUptodate(page
);
1042 zero_user_segment(page
, 0, PAGE_SIZE
);
1043 SetPageUptodate(page
);
1049 * This page will go to BIO. Do we need to send this
1052 if (bio
&& (last_block_in_bio
!= block_nr
- 1)) {
1054 __submit_bio(F2FS_I_SB(inode
), READ
, bio
);
1058 struct fscrypt_ctx
*ctx
= NULL
;
1060 if (f2fs_encrypted_inode(inode
) &&
1061 S_ISREG(inode
->i_mode
)) {
1063 ctx
= fscrypt_get_ctx(inode
, GFP_NOFS
);
1065 goto set_error_page
;
1067 /* wait the page to be moved by cleaning */
1068 f2fs_wait_on_encrypted_page_writeback(
1069 F2FS_I_SB(inode
), block_nr
);
1072 bio
= bio_alloc(GFP_KERNEL
,
1073 min_t(int, nr_pages
, BIO_MAX_PAGES
));
1076 fscrypt_release_ctx(ctx
);
1077 goto set_error_page
;
1079 bio
->bi_bdev
= bdev
;
1080 bio
->bi_iter
.bi_sector
= SECTOR_FROM_BLOCK(block_nr
);
1081 bio
->bi_end_io
= f2fs_read_end_io
;
1082 bio
->bi_private
= ctx
;
1085 if (bio_add_page(bio
, page
, blocksize
, 0) < blocksize
)
1086 goto submit_and_realloc
;
1088 last_block_in_bio
= block_nr
;
1092 zero_user_segment(page
, 0, PAGE_SIZE
);
1097 __submit_bio(F2FS_I_SB(inode
), READ
, bio
);
1105 BUG_ON(pages
&& !list_empty(pages
));
1107 __submit_bio(F2FS_I_SB(inode
), READ
, bio
);
1111 static int f2fs_read_data_page(struct file
*file
, struct page
*page
)
1113 struct inode
*inode
= page
->mapping
->host
;
1116 trace_f2fs_readpage(page
, DATA
);
1118 /* If the file has inline data, try to read it directly */
1119 if (f2fs_has_inline_data(inode
))
1120 ret
= f2fs_read_inline_data(inode
, page
);
1122 ret
= f2fs_mpage_readpages(page
->mapping
, NULL
, page
, 1);
1126 static int f2fs_read_data_pages(struct file
*file
,
1127 struct address_space
*mapping
,
1128 struct list_head
*pages
, unsigned nr_pages
)
1130 struct inode
*inode
= file
->f_mapping
->host
;
1131 struct page
*page
= list_entry(pages
->prev
, struct page
, lru
);
1133 trace_f2fs_readpages(inode
, page
, nr_pages
);
1135 /* If the file has inline data, skip readpages */
1136 if (f2fs_has_inline_data(inode
))
1139 return f2fs_mpage_readpages(mapping
, pages
, NULL
, nr_pages
);
1142 int do_write_data_page(struct f2fs_io_info
*fio
)
1144 struct page
*page
= fio
->page
;
1145 struct inode
*inode
= page
->mapping
->host
;
1146 struct dnode_of_data dn
;
1149 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
1150 err
= get_dnode_of_data(&dn
, page
->index
, LOOKUP_NODE
);
1154 fio
->old_blkaddr
= dn
.data_blkaddr
;
1156 /* This page is already truncated */
1157 if (fio
->old_blkaddr
== NULL_ADDR
) {
1158 ClearPageUptodate(page
);
1162 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
)) {
1163 gfp_t gfp_flags
= GFP_NOFS
;
1165 /* wait for GCed encrypted page writeback */
1166 f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode
),
1169 fio
->encrypted_page
= fscrypt_encrypt_page(inode
, fio
->page
,
1171 if (IS_ERR(fio
->encrypted_page
)) {
1172 err
= PTR_ERR(fio
->encrypted_page
);
1173 if (err
== -ENOMEM
) {
1174 /* flush pending ios and wait for a while */
1175 f2fs_flush_merged_bios(F2FS_I_SB(inode
));
1176 congestion_wait(BLK_RW_ASYNC
, HZ
/50);
1177 gfp_flags
|= __GFP_NOFAIL
;
1185 set_page_writeback(page
);
1188 * If current allocation needs SSR,
1189 * it had better in-place writes for updated data.
1191 if (unlikely(fio
->old_blkaddr
!= NEW_ADDR
&&
1192 !is_cold_data(page
) &&
1193 !IS_ATOMIC_WRITTEN_PAGE(page
) &&
1194 need_inplace_update(inode
))) {
1195 rewrite_data_page(fio
);
1196 set_inode_flag(F2FS_I(inode
), FI_UPDATE_WRITE
);
1197 trace_f2fs_do_write_data_page(page
, IPU
);
1199 write_data_page(&dn
, fio
);
1200 trace_f2fs_do_write_data_page(page
, OPU
);
1201 set_inode_flag(F2FS_I(inode
), FI_APPEND_WRITE
);
1202 if (page
->index
== 0)
1203 set_inode_flag(F2FS_I(inode
), FI_FIRST_BLOCK_WRITTEN
);
1206 f2fs_put_dnode(&dn
);
1210 static int f2fs_write_data_page(struct page
*page
,
1211 struct writeback_control
*wbc
)
1213 struct inode
*inode
= page
->mapping
->host
;
1214 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1215 loff_t i_size
= i_size_read(inode
);
1216 const pgoff_t end_index
= ((unsigned long long) i_size
)
1218 unsigned offset
= 0;
1219 bool need_balance_fs
= false;
1221 struct f2fs_io_info fio
= {
1224 .rw
= (wbc
->sync_mode
== WB_SYNC_ALL
) ? WRITE_SYNC
: WRITE
,
1226 .encrypted_page
= NULL
,
1229 trace_f2fs_writepage(page
, DATA
);
1231 if (page
->index
< end_index
)
1235 * If the offset is out-of-range of file size,
1236 * this page does not have to be written to disk.
1238 offset
= i_size
& (PAGE_SIZE
- 1);
1239 if ((page
->index
>= end_index
+ 1) || !offset
)
1242 zero_user_segment(page
, offset
, PAGE_SIZE
);
1244 if (unlikely(is_sbi_flag_set(sbi
, SBI_POR_DOING
)))
1246 if (f2fs_is_drop_cache(inode
))
1248 /* we should not write 0'th page having journal header */
1249 if (f2fs_is_volatile_file(inode
) && (!page
->index
||
1250 (!wbc
->for_reclaim
&&
1251 available_free_memory(sbi
, BASE_CHECK
))))
1254 /* Dentry blocks are controlled by checkpoint */
1255 if (S_ISDIR(inode
->i_mode
)) {
1256 if (unlikely(f2fs_cp_error(sbi
)))
1258 err
= do_write_data_page(&fio
);
1262 /* we should bypass data pages to proceed the kworkder jobs */
1263 if (unlikely(f2fs_cp_error(sbi
))) {
1268 if (!wbc
->for_reclaim
)
1269 need_balance_fs
= true;
1270 else if (has_not_enough_free_secs(sbi
, 0))
1275 if (f2fs_has_inline_data(inode
))
1276 err
= f2fs_write_inline_data(inode
, page
);
1278 err
= do_write_data_page(&fio
);
1279 f2fs_unlock_op(sbi
);
1281 if (err
&& err
!= -ENOENT
)
1284 clear_cold_data(page
);
1286 inode_dec_dirty_pages(inode
);
1288 ClearPageUptodate(page
);
1290 if (wbc
->for_reclaim
) {
1291 f2fs_submit_merged_bio_cond(sbi
, NULL
, page
, 0, DATA
, WRITE
);
1292 remove_dirty_inode(inode
);
1296 f2fs_balance_fs(sbi
, need_balance_fs
);
1298 if (unlikely(f2fs_cp_error(sbi
)))
1299 f2fs_submit_merged_bio(sbi
, DATA
, WRITE
);
1304 redirty_page_for_writepage(wbc
, page
);
1305 return AOP_WRITEPAGE_ACTIVATE
;
1308 static int __f2fs_writepage(struct page
*page
, struct writeback_control
*wbc
,
1311 struct address_space
*mapping
= data
;
1312 int ret
= mapping
->a_ops
->writepage(page
, wbc
);
1313 mapping_set_error(mapping
, ret
);
1318 * This function was copied from write_cche_pages from mm/page-writeback.c.
1319 * The major change is making write step of cold data page separately from
1320 * warm/hot data page.
1322 static int f2fs_write_cache_pages(struct address_space
*mapping
,
1323 struct writeback_control
*wbc
, writepage_t writepage
,
1328 struct pagevec pvec
;
1330 pgoff_t
uninitialized_var(writeback_index
);
1332 pgoff_t end
; /* Inclusive */
1335 int range_whole
= 0;
1339 pagevec_init(&pvec
, 0);
1341 if (wbc
->range_cyclic
) {
1342 writeback_index
= mapping
->writeback_index
; /* prev offset */
1343 index
= writeback_index
;
1350 index
= wbc
->range_start
>> PAGE_SHIFT
;
1351 end
= wbc
->range_end
>> PAGE_SHIFT
;
1352 if (wbc
->range_start
== 0 && wbc
->range_end
== LLONG_MAX
)
1354 cycled
= 1; /* ignore range_cyclic tests */
1356 if (wbc
->sync_mode
== WB_SYNC_ALL
|| wbc
->tagged_writepages
)
1357 tag
= PAGECACHE_TAG_TOWRITE
;
1359 tag
= PAGECACHE_TAG_DIRTY
;
1361 if (wbc
->sync_mode
== WB_SYNC_ALL
|| wbc
->tagged_writepages
)
1362 tag_pages_for_writeback(mapping
, index
, end
);
1364 while (!done
&& (index
<= end
)) {
1367 nr_pages
= pagevec_lookup_tag(&pvec
, mapping
, &index
, tag
,
1368 min(end
- index
, (pgoff_t
)PAGEVEC_SIZE
- 1) + 1);
1372 for (i
= 0; i
< nr_pages
; i
++) {
1373 struct page
*page
= pvec
.pages
[i
];
1375 if (page
->index
> end
) {
1380 done_index
= page
->index
;
1384 if (unlikely(page
->mapping
!= mapping
)) {
1390 if (!PageDirty(page
)) {
1391 /* someone wrote it for us */
1392 goto continue_unlock
;
1395 if (step
== is_cold_data(page
))
1396 goto continue_unlock
;
1398 if (PageWriteback(page
)) {
1399 if (wbc
->sync_mode
!= WB_SYNC_NONE
)
1400 f2fs_wait_on_page_writeback(page
,
1403 goto continue_unlock
;
1406 BUG_ON(PageWriteback(page
));
1407 if (!clear_page_dirty_for_io(page
))
1408 goto continue_unlock
;
1410 ret
= (*writepage
)(page
, wbc
, data
);
1411 if (unlikely(ret
)) {
1412 if (ret
== AOP_WRITEPAGE_ACTIVATE
) {
1416 done_index
= page
->index
+ 1;
1422 if (--wbc
->nr_to_write
<= 0 &&
1423 wbc
->sync_mode
== WB_SYNC_NONE
) {
1428 pagevec_release(&pvec
);
1437 if (!cycled
&& !done
) {
1440 end
= writeback_index
- 1;
1443 if (wbc
->range_cyclic
|| (range_whole
&& wbc
->nr_to_write
> 0))
1444 mapping
->writeback_index
= done_index
;
1449 static int f2fs_write_data_pages(struct address_space
*mapping
,
1450 struct writeback_control
*wbc
)
1452 struct inode
*inode
= mapping
->host
;
1453 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1454 bool locked
= false;
1458 /* deal with chardevs and other special file */
1459 if (!mapping
->a_ops
->writepage
)
1462 /* skip writing if there is no dirty page in this inode */
1463 if (!get_dirty_pages(inode
) && wbc
->sync_mode
== WB_SYNC_NONE
)
1466 if (S_ISDIR(inode
->i_mode
) && wbc
->sync_mode
== WB_SYNC_NONE
&&
1467 get_dirty_pages(inode
) < nr_pages_to_skip(sbi
, DATA
) &&
1468 available_free_memory(sbi
, DIRTY_DENTS
))
1471 /* skip writing during file defragment */
1472 if (is_inode_flag_set(F2FS_I(inode
), FI_DO_DEFRAG
))
1475 /* during POR, we don't need to trigger writepage at all. */
1476 if (unlikely(is_sbi_flag_set(sbi
, SBI_POR_DOING
)))
1479 trace_f2fs_writepages(mapping
->host
, wbc
, DATA
);
1481 diff
= nr_pages_to_write(sbi
, DATA
, wbc
);
1483 if (!S_ISDIR(inode
->i_mode
) && wbc
->sync_mode
== WB_SYNC_ALL
) {
1484 mutex_lock(&sbi
->writepages
);
1487 ret
= f2fs_write_cache_pages(mapping
, wbc
, __f2fs_writepage
, mapping
);
1488 f2fs_submit_merged_bio_cond(sbi
, inode
, NULL
, 0, DATA
, WRITE
);
1490 mutex_unlock(&sbi
->writepages
);
1492 remove_dirty_inode(inode
);
1494 wbc
->nr_to_write
= max((long)0, wbc
->nr_to_write
- diff
);
1498 wbc
->pages_skipped
+= get_dirty_pages(inode
);
1499 trace_f2fs_writepages(mapping
->host
, wbc
, DATA
);
1503 static void f2fs_write_failed(struct address_space
*mapping
, loff_t to
)
1505 struct inode
*inode
= mapping
->host
;
1506 loff_t i_size
= i_size_read(inode
);
1509 truncate_pagecache(inode
, i_size
);
1510 truncate_blocks(inode
, i_size
, true);
1514 static int prepare_write_begin(struct f2fs_sb_info
*sbi
,
1515 struct page
*page
, loff_t pos
, unsigned len
,
1516 block_t
*blk_addr
, bool *node_changed
)
1518 struct inode
*inode
= page
->mapping
->host
;
1519 pgoff_t index
= page
->index
;
1520 struct dnode_of_data dn
;
1522 bool locked
= false;
1523 struct extent_info ei
;
1527 * we already allocated all the blocks, so we don't need to get
1528 * the block addresses when there is no need to fill the page.
1530 if (!f2fs_has_inline_data(inode
) && !f2fs_encrypted_inode(inode
) &&
1534 if (f2fs_has_inline_data(inode
) ||
1535 (pos
& PAGE_MASK
) >= i_size_read(inode
)) {
1540 /* check inline_data */
1541 ipage
= get_node_page(sbi
, inode
->i_ino
);
1542 if (IS_ERR(ipage
)) {
1543 err
= PTR_ERR(ipage
);
1547 set_new_dnode(&dn
, inode
, ipage
, ipage
, 0);
1549 if (f2fs_has_inline_data(inode
)) {
1550 if (pos
+ len
<= MAX_INLINE_DATA
) {
1551 read_inline_data(page
, ipage
);
1552 set_inode_flag(F2FS_I(inode
), FI_DATA_EXIST
);
1554 set_inline_node(ipage
);
1556 err
= f2fs_convert_inline_page(&dn
, page
);
1559 if (dn
.data_blkaddr
== NULL_ADDR
)
1560 err
= f2fs_get_block(&dn
, index
);
1562 } else if (locked
) {
1563 err
= f2fs_get_block(&dn
, index
);
1565 if (f2fs_lookup_extent_cache(inode
, index
, &ei
)) {
1566 dn
.data_blkaddr
= ei
.blk
+ index
- ei
.fofs
;
1569 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
1570 if (err
|| dn
.data_blkaddr
== NULL_ADDR
) {
1571 f2fs_put_dnode(&dn
);
1579 /* convert_inline_page can make node_changed */
1580 *blk_addr
= dn
.data_blkaddr
;
1581 *node_changed
= dn
.node_changed
;
1583 f2fs_put_dnode(&dn
);
1586 f2fs_unlock_op(sbi
);
1590 static int f2fs_write_begin(struct file
*file
, struct address_space
*mapping
,
1591 loff_t pos
, unsigned len
, unsigned flags
,
1592 struct page
**pagep
, void **fsdata
)
1594 struct inode
*inode
= mapping
->host
;
1595 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1596 struct page
*page
= NULL
;
1597 pgoff_t index
= ((unsigned long long) pos
) >> PAGE_SHIFT
;
1598 bool need_balance
= false;
1599 block_t blkaddr
= NULL_ADDR
;
1602 trace_f2fs_write_begin(inode
, pos
, len
, flags
);
1605 * We should check this at this moment to avoid deadlock on inode page
1606 * and #0 page. The locking rule for inline_data conversion should be:
1607 * lock_page(page #0) -> lock_page(inode_page)
1610 err
= f2fs_convert_inline_inode(inode
);
1615 page
= grab_cache_page_write_begin(mapping
, index
, flags
);
1623 err
= prepare_write_begin(sbi
, page
, pos
, len
,
1624 &blkaddr
, &need_balance
);
1628 if (need_balance
&& has_not_enough_free_secs(sbi
, 0)) {
1630 f2fs_balance_fs(sbi
, true);
1632 if (page
->mapping
!= mapping
) {
1633 /* The page got truncated from under us */
1634 f2fs_put_page(page
, 1);
1639 f2fs_wait_on_page_writeback(page
, DATA
, false);
1641 /* wait for GCed encrypted page writeback */
1642 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
1643 f2fs_wait_on_encrypted_page_writeback(sbi
, blkaddr
);
1645 if (len
== PAGE_SIZE
)
1647 if (PageUptodate(page
))
1650 if ((pos
& PAGE_MASK
) >= i_size_read(inode
)) {
1651 unsigned start
= pos
& (PAGE_SIZE
- 1);
1652 unsigned end
= start
+ len
;
1654 /* Reading beyond i_size is simple: memset to zero */
1655 zero_user_segments(page
, 0, start
, end
, PAGE_SIZE
);
1659 if (blkaddr
== NEW_ADDR
) {
1660 zero_user_segment(page
, 0, PAGE_SIZE
);
1662 struct f2fs_io_info fio
= {
1666 .old_blkaddr
= blkaddr
,
1667 .new_blkaddr
= blkaddr
,
1669 .encrypted_page
= NULL
,
1671 err
= f2fs_submit_page_bio(&fio
);
1676 if (unlikely(!PageUptodate(page
))) {
1680 if (unlikely(page
->mapping
!= mapping
)) {
1681 f2fs_put_page(page
, 1);
1685 /* avoid symlink page */
1686 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
)) {
1687 err
= fscrypt_decrypt_page(page
);
1693 SetPageUptodate(page
);
1695 clear_cold_data(page
);
1699 f2fs_put_page(page
, 1);
1700 f2fs_write_failed(mapping
, pos
+ len
);
1704 static int f2fs_write_end(struct file
*file
,
1705 struct address_space
*mapping
,
1706 loff_t pos
, unsigned len
, unsigned copied
,
1707 struct page
*page
, void *fsdata
)
1709 struct inode
*inode
= page
->mapping
->host
;
1711 trace_f2fs_write_end(inode
, pos
, len
, copied
);
1713 set_page_dirty(page
);
1715 if (pos
+ copied
> i_size_read(inode
)) {
1716 i_size_write(inode
, pos
+ copied
);
1717 mark_inode_dirty(inode
);
1720 f2fs_put_page(page
, 1);
1721 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1725 static int check_direct_IO(struct inode
*inode
, struct iov_iter
*iter
,
1728 unsigned blocksize_mask
= inode
->i_sb
->s_blocksize
- 1;
1730 if (offset
& blocksize_mask
)
1733 if (iov_iter_alignment(iter
) & blocksize_mask
)
1739 static ssize_t
f2fs_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1741 struct address_space
*mapping
= iocb
->ki_filp
->f_mapping
;
1742 struct inode
*inode
= mapping
->host
;
1743 size_t count
= iov_iter_count(iter
);
1744 loff_t offset
= iocb
->ki_pos
;
1747 err
= check_direct_IO(inode
, iter
, offset
);
1751 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
1754 trace_f2fs_direct_IO_enter(inode
, offset
, count
, iov_iter_rw(iter
));
1756 err
= blockdev_direct_IO(iocb
, inode
, iter
, get_data_block_dio
);
1757 if (iov_iter_rw(iter
) == WRITE
) {
1759 set_inode_flag(F2FS_I(inode
), FI_UPDATE_WRITE
);
1761 f2fs_write_failed(mapping
, offset
+ count
);
1764 trace_f2fs_direct_IO_exit(inode
, offset
, count
, iov_iter_rw(iter
), err
);
1769 void f2fs_invalidate_page(struct page
*page
, unsigned int offset
,
1770 unsigned int length
)
1772 struct inode
*inode
= page
->mapping
->host
;
1773 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1775 if (inode
->i_ino
>= F2FS_ROOT_INO(sbi
) &&
1776 (offset
% PAGE_SIZE
|| length
!= PAGE_SIZE
))
1779 if (PageDirty(page
)) {
1780 if (inode
->i_ino
== F2FS_META_INO(sbi
))
1781 dec_page_count(sbi
, F2FS_DIRTY_META
);
1782 else if (inode
->i_ino
== F2FS_NODE_INO(sbi
))
1783 dec_page_count(sbi
, F2FS_DIRTY_NODES
);
1785 inode_dec_dirty_pages(inode
);
1788 /* This is atomic written page, keep Private */
1789 if (IS_ATOMIC_WRITTEN_PAGE(page
))
1792 set_page_private(page
, 0);
1793 ClearPagePrivate(page
);
1796 int f2fs_release_page(struct page
*page
, gfp_t wait
)
1798 /* If this is dirty page, keep PagePrivate */
1799 if (PageDirty(page
))
1802 /* This is atomic written page, keep Private */
1803 if (IS_ATOMIC_WRITTEN_PAGE(page
))
1806 set_page_private(page
, 0);
1807 ClearPagePrivate(page
);
1811 static int f2fs_set_data_page_dirty(struct page
*page
)
1813 struct address_space
*mapping
= page
->mapping
;
1814 struct inode
*inode
= mapping
->host
;
1816 trace_f2fs_set_page_dirty(page
, DATA
);
1818 SetPageUptodate(page
);
1820 if (f2fs_is_atomic_file(inode
)) {
1821 if (!IS_ATOMIC_WRITTEN_PAGE(page
)) {
1822 register_inmem_page(inode
, page
);
1826 * Previously, this page has been registered, we just
1832 if (!PageDirty(page
)) {
1833 __set_page_dirty_nobuffers(page
);
1834 update_dirty_page(inode
, page
);
1840 static sector_t
f2fs_bmap(struct address_space
*mapping
, sector_t block
)
1842 struct inode
*inode
= mapping
->host
;
1844 if (f2fs_has_inline_data(inode
))
1847 /* make sure allocating whole blocks */
1848 if (mapping_tagged(mapping
, PAGECACHE_TAG_DIRTY
))
1849 filemap_write_and_wait(mapping
);
1851 return generic_block_bmap(mapping
, block
, get_data_block_bmap
);
1854 const struct address_space_operations f2fs_dblock_aops
= {
1855 .readpage
= f2fs_read_data_page
,
1856 .readpages
= f2fs_read_data_pages
,
1857 .writepage
= f2fs_write_data_page
,
1858 .writepages
= f2fs_write_data_pages
,
1859 .write_begin
= f2fs_write_begin
,
1860 .write_end
= f2fs_write_end
,
1861 .set_page_dirty
= f2fs_set_data_page_dirty
,
1862 .invalidatepage
= f2fs_invalidate_page
,
1863 .releasepage
= f2fs_release_page
,
1864 .direct_IO
= f2fs_direct_IO
,