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
,
101 struct bio
*bio
, enum page_type type
)
103 if (!is_read_io(rw
)) {
104 atomic_inc(&sbi
->nr_wb_bios
);
105 if (current
->plug
&& (type
== DATA
|| type
== NODE
))
106 blk_finish_plug(current
->plug
);
111 static void __submit_merged_bio(struct f2fs_bio_info
*io
)
113 struct f2fs_io_info
*fio
= &io
->fio
;
118 if (is_read_io(fio
->rw
))
119 trace_f2fs_submit_read_bio(io
->sbi
->sb
, fio
, io
->bio
);
121 trace_f2fs_submit_write_bio(io
->sbi
->sb
, fio
, io
->bio
);
123 __submit_bio(io
->sbi
, fio
->rw
, io
->bio
, fio
->type
);
127 static bool __has_merged_page(struct f2fs_bio_info
*io
, struct inode
*inode
,
128 struct page
*page
, nid_t ino
)
130 struct bio_vec
*bvec
;
137 if (!inode
&& !page
&& !ino
)
140 bio_for_each_segment_all(bvec
, io
->bio
, i
) {
142 if (bvec
->bv_page
->mapping
)
143 target
= bvec
->bv_page
;
145 target
= fscrypt_control_page(bvec
->bv_page
);
147 if (inode
&& inode
== target
->mapping
->host
)
149 if (page
&& page
== target
)
151 if (ino
&& ino
== ino_of_node(target
))
158 static bool has_merged_page(struct f2fs_sb_info
*sbi
, struct inode
*inode
,
159 struct page
*page
, nid_t ino
,
162 enum page_type btype
= PAGE_TYPE_OF_BIO(type
);
163 struct f2fs_bio_info
*io
= &sbi
->write_io
[btype
];
166 down_read(&io
->io_rwsem
);
167 ret
= __has_merged_page(io
, inode
, page
, ino
);
168 up_read(&io
->io_rwsem
);
172 static void __f2fs_submit_merged_bio(struct f2fs_sb_info
*sbi
,
173 struct inode
*inode
, struct page
*page
,
174 nid_t ino
, enum page_type type
, int rw
)
176 enum page_type btype
= PAGE_TYPE_OF_BIO(type
);
177 struct f2fs_bio_info
*io
;
179 io
= is_read_io(rw
) ? &sbi
->read_io
: &sbi
->write_io
[btype
];
181 down_write(&io
->io_rwsem
);
183 if (!__has_merged_page(io
, inode
, page
, ino
))
186 /* change META to META_FLUSH in the checkpoint procedure */
187 if (type
>= META_FLUSH
) {
188 io
->fio
.type
= META_FLUSH
;
189 if (test_opt(sbi
, NOBARRIER
))
190 io
->fio
.rw
= WRITE_FLUSH
| REQ_META
| REQ_PRIO
;
192 io
->fio
.rw
= WRITE_FLUSH_FUA
| REQ_META
| REQ_PRIO
;
194 __submit_merged_bio(io
);
196 up_write(&io
->io_rwsem
);
199 void f2fs_submit_merged_bio(struct f2fs_sb_info
*sbi
, enum page_type type
,
202 __f2fs_submit_merged_bio(sbi
, NULL
, NULL
, 0, type
, rw
);
205 void f2fs_submit_merged_bio_cond(struct f2fs_sb_info
*sbi
,
206 struct inode
*inode
, struct page
*page
,
207 nid_t ino
, enum page_type type
, int rw
)
209 if (has_merged_page(sbi
, inode
, page
, ino
, type
))
210 __f2fs_submit_merged_bio(sbi
, inode
, page
, ino
, type
, rw
);
213 void f2fs_flush_merged_bios(struct f2fs_sb_info
*sbi
)
215 f2fs_submit_merged_bio(sbi
, DATA
, WRITE
);
216 f2fs_submit_merged_bio(sbi
, NODE
, WRITE
);
217 f2fs_submit_merged_bio(sbi
, META
, WRITE
);
221 * Fill the locked page with data located in the block address.
222 * Return unlocked page.
224 int f2fs_submit_page_bio(struct f2fs_io_info
*fio
)
227 struct page
*page
= fio
->encrypted_page
?
228 fio
->encrypted_page
: fio
->page
;
230 trace_f2fs_submit_page_bio(page
, fio
);
231 f2fs_trace_ios(fio
, 0);
233 /* Allocate a new bio */
234 bio
= __bio_alloc(fio
->sbi
, fio
->new_blkaddr
, 1, is_read_io(fio
->rw
));
236 if (bio_add_page(bio
, page
, PAGE_SIZE
, 0) < PAGE_SIZE
) {
241 __submit_bio(fio
->sbi
, fio
->rw
, bio
, fio
->type
);
245 void f2fs_submit_page_mbio(struct f2fs_io_info
*fio
)
247 struct f2fs_sb_info
*sbi
= fio
->sbi
;
248 enum page_type btype
= PAGE_TYPE_OF_BIO(fio
->type
);
249 struct f2fs_bio_info
*io
;
250 bool is_read
= is_read_io(fio
->rw
);
251 struct page
*bio_page
;
253 io
= is_read
? &sbi
->read_io
: &sbi
->write_io
[btype
];
255 if (fio
->old_blkaddr
!= NEW_ADDR
)
256 verify_block_addr(sbi
, fio
->old_blkaddr
);
257 verify_block_addr(sbi
, fio
->new_blkaddr
);
259 down_write(&io
->io_rwsem
);
261 if (io
->bio
&& (io
->last_block_in_bio
!= fio
->new_blkaddr
- 1 ||
262 io
->fio
.rw
!= fio
->rw
))
263 __submit_merged_bio(io
);
265 if (io
->bio
== NULL
) {
266 int bio_blocks
= MAX_BIO_BLOCKS(sbi
);
268 io
->bio
= __bio_alloc(sbi
, fio
->new_blkaddr
,
269 bio_blocks
, is_read
);
273 bio_page
= fio
->encrypted_page
? fio
->encrypted_page
: fio
->page
;
275 if (bio_add_page(io
->bio
, bio_page
, PAGE_SIZE
, 0) <
277 __submit_merged_bio(io
);
281 io
->last_block_in_bio
= fio
->new_blkaddr
;
282 f2fs_trace_ios(fio
, 0);
284 up_write(&io
->io_rwsem
);
285 trace_f2fs_submit_page_mbio(fio
->page
, fio
);
288 static void __set_data_blkaddr(struct dnode_of_data
*dn
)
290 struct f2fs_node
*rn
= F2FS_NODE(dn
->node_page
);
293 /* Get physical address of data block */
294 addr_array
= blkaddr_in_node(rn
);
295 addr_array
[dn
->ofs_in_node
] = cpu_to_le32(dn
->data_blkaddr
);
299 * Lock ordering for the change of data block address:
302 * update block addresses in the node page
304 void set_data_blkaddr(struct dnode_of_data
*dn
)
306 f2fs_wait_on_page_writeback(dn
->node_page
, NODE
, true);
307 __set_data_blkaddr(dn
);
308 if (set_page_dirty(dn
->node_page
))
309 dn
->node_changed
= true;
312 void f2fs_update_data_blkaddr(struct dnode_of_data
*dn
, block_t blkaddr
)
314 dn
->data_blkaddr
= blkaddr
;
315 set_data_blkaddr(dn
);
316 f2fs_update_extent_cache(dn
);
319 /* dn->ofs_in_node will be returned with up-to-date last block pointer */
320 int reserve_new_blocks(struct dnode_of_data
*dn
, blkcnt_t count
)
322 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
327 if (unlikely(is_inode_flag_set(dn
->inode
, FI_NO_ALLOC
)))
329 if (unlikely(!inc_valid_block_count(sbi
, dn
->inode
, &count
)))
332 trace_f2fs_reserve_new_blocks(dn
->inode
, dn
->nid
,
333 dn
->ofs_in_node
, count
);
335 f2fs_wait_on_page_writeback(dn
->node_page
, NODE
, true);
337 for (; count
> 0; dn
->ofs_in_node
++) {
339 datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
340 if (blkaddr
== NULL_ADDR
) {
341 dn
->data_blkaddr
= NEW_ADDR
;
342 __set_data_blkaddr(dn
);
347 if (set_page_dirty(dn
->node_page
))
348 dn
->node_changed
= true;
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 f2fs_i_size_write(inode
, ((loff_t
)(index
+ 1) << PAGE_SHIFT
));
571 static int __allocate_data_block(struct dnode_of_data
*dn
)
573 struct f2fs_sb_info
*sbi
= F2FS_I_SB(dn
->inode
);
574 struct f2fs_summary sum
;
576 int seg
= CURSEG_WARM_DATA
;
580 if (unlikely(is_inode_flag_set(dn
->inode
, FI_NO_ALLOC
)))
583 dn
->data_blkaddr
= datablock_addr(dn
->node_page
, dn
->ofs_in_node
);
584 if (dn
->data_blkaddr
== NEW_ADDR
)
587 if (unlikely(!inc_valid_block_count(sbi
, dn
->inode
, &count
)))
591 get_node_info(sbi
, dn
->nid
, &ni
);
592 set_summary(&sum
, dn
->nid
, dn
->ofs_in_node
, ni
.version
);
594 if (dn
->ofs_in_node
== 0 && dn
->inode_page
== dn
->node_page
)
595 seg
= CURSEG_DIRECT_IO
;
597 allocate_data_block(sbi
, NULL
, dn
->data_blkaddr
, &dn
->data_blkaddr
,
599 set_data_blkaddr(dn
);
602 fofs
= start_bidx_of_node(ofs_of_node(dn
->node_page
), dn
->inode
) +
604 if (i_size_read(dn
->inode
) < ((loff_t
)(fofs
+ 1) << PAGE_SHIFT
))
605 f2fs_i_size_write(dn
->inode
,
606 ((loff_t
)(fofs
+ 1) << PAGE_SHIFT
));
610 ssize_t
f2fs_preallocate_blocks(struct kiocb
*iocb
, struct iov_iter
*from
)
612 struct inode
*inode
= file_inode(iocb
->ki_filp
);
613 struct f2fs_map_blocks map
;
616 map
.m_lblk
= F2FS_BLK_ALIGN(iocb
->ki_pos
);
617 map
.m_len
= F2FS_BYTES_TO_BLK(iov_iter_count(from
));
618 map
.m_next_pgofs
= NULL
;
620 if (f2fs_encrypted_inode(inode
))
623 if (iocb
->ki_flags
& IOCB_DIRECT
) {
624 ret
= f2fs_convert_inline_inode(inode
);
627 return f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_DIO
);
629 if (iocb
->ki_pos
+ iov_iter_count(from
) > MAX_INLINE_DATA
) {
630 ret
= f2fs_convert_inline_inode(inode
);
634 if (!f2fs_has_inline_data(inode
))
635 return f2fs_map_blocks(inode
, &map
, 1, F2FS_GET_BLOCK_PRE_AIO
);
640 * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with
641 * f2fs_map_blocks structure.
642 * If original data blocks are allocated, then give them to blockdev.
644 * a. preallocate requested block addresses
645 * b. do not use extent cache for better performance
646 * c. give the block addresses to blockdev
648 int f2fs_map_blocks(struct inode
*inode
, struct f2fs_map_blocks
*map
,
649 int create
, int flag
)
651 unsigned int maxblocks
= map
->m_len
;
652 struct dnode_of_data dn
;
653 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
654 int mode
= create
? ALLOC_NODE
: LOOKUP_NODE_RA
;
655 pgoff_t pgofs
, end_offset
, end
;
656 int err
= 0, ofs
= 1;
657 unsigned int ofs_in_node
, last_ofs_in_node
;
659 struct extent_info ei
;
660 bool allocated
= false;
666 /* it only supports block size == page size */
667 pgofs
= (pgoff_t
)map
->m_lblk
;
668 end
= pgofs
+ maxblocks
;
670 if (!create
&& f2fs_lookup_extent_cache(inode
, pgofs
, &ei
)) {
671 map
->m_pblk
= ei
.blk
+ pgofs
- ei
.fofs
;
672 map
->m_len
= min((pgoff_t
)maxblocks
, ei
.fofs
+ ei
.len
- pgofs
);
673 map
->m_flags
= F2FS_MAP_MAPPED
;
681 /* When reading holes, we need its node page */
682 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
683 err
= get_dnode_of_data(&dn
, pgofs
, mode
);
685 if (flag
== F2FS_GET_BLOCK_BMAP
)
687 if (err
== -ENOENT
) {
689 if (map
->m_next_pgofs
)
691 get_next_page_offset(&dn
, pgofs
);
697 ofs_in_node
= dn
.ofs_in_node
;
698 end_offset
= ADDRS_PER_PAGE(dn
.node_page
, inode
);
701 blkaddr
= datablock_addr(dn
.node_page
, dn
.ofs_in_node
);
703 if (blkaddr
== NEW_ADDR
|| blkaddr
== NULL_ADDR
) {
705 if (unlikely(f2fs_cp_error(sbi
))) {
709 if (flag
== F2FS_GET_BLOCK_PRE_AIO
) {
710 if (blkaddr
== NULL_ADDR
) {
712 last_ofs_in_node
= dn
.ofs_in_node
;
715 err
= __allocate_data_block(&dn
);
717 set_inode_flag(inode
, FI_APPEND_WRITE
);
723 map
->m_flags
= F2FS_MAP_NEW
;
724 blkaddr
= dn
.data_blkaddr
;
726 if (flag
== F2FS_GET_BLOCK_BMAP
) {
730 if (flag
== F2FS_GET_BLOCK_FIEMAP
&&
731 blkaddr
== NULL_ADDR
) {
732 if (map
->m_next_pgofs
)
733 *map
->m_next_pgofs
= pgofs
+ 1;
735 if (flag
!= F2FS_GET_BLOCK_FIEMAP
||
741 if (flag
== F2FS_GET_BLOCK_PRE_AIO
)
744 if (map
->m_len
== 0) {
745 /* preallocated unwritten block should be mapped for fiemap. */
746 if (blkaddr
== NEW_ADDR
)
747 map
->m_flags
|= F2FS_MAP_UNWRITTEN
;
748 map
->m_flags
|= F2FS_MAP_MAPPED
;
750 map
->m_pblk
= blkaddr
;
752 } else if ((map
->m_pblk
!= NEW_ADDR
&&
753 blkaddr
== (map
->m_pblk
+ ofs
)) ||
754 (map
->m_pblk
== NEW_ADDR
&& blkaddr
== NEW_ADDR
) ||
755 flag
== F2FS_GET_BLOCK_PRE_DIO
) {
766 /* preallocate blocks in batch for one dnode page */
767 if (flag
== F2FS_GET_BLOCK_PRE_AIO
&&
768 (pgofs
== end
|| dn
.ofs_in_node
== end_offset
)) {
770 dn
.ofs_in_node
= ofs_in_node
;
771 err
= reserve_new_blocks(&dn
, prealloc
);
775 map
->m_len
+= dn
.ofs_in_node
- ofs_in_node
;
776 if (prealloc
&& dn
.ofs_in_node
!= last_ofs_in_node
+ 1) {
780 dn
.ofs_in_node
= end_offset
;
785 else if (dn
.ofs_in_node
< end_offset
)
792 f2fs_balance_fs(sbi
, allocated
);
802 f2fs_balance_fs(sbi
, allocated
);
805 trace_f2fs_map_blocks(inode
, map
, err
);
809 static int __get_data_block(struct inode
*inode
, sector_t iblock
,
810 struct buffer_head
*bh
, int create
, int flag
,
813 struct f2fs_map_blocks map
;
817 map
.m_len
= bh
->b_size
>> inode
->i_blkbits
;
818 map
.m_next_pgofs
= next_pgofs
;
820 ret
= f2fs_map_blocks(inode
, &map
, create
, flag
);
822 map_bh(bh
, inode
->i_sb
, map
.m_pblk
);
823 bh
->b_state
= (bh
->b_state
& ~F2FS_MAP_FLAGS
) | map
.m_flags
;
824 bh
->b_size
= map
.m_len
<< inode
->i_blkbits
;
829 static int get_data_block(struct inode
*inode
, sector_t iblock
,
830 struct buffer_head
*bh_result
, int create
, int flag
,
833 return __get_data_block(inode
, iblock
, bh_result
, create
,
837 static int get_data_block_dio(struct inode
*inode
, sector_t iblock
,
838 struct buffer_head
*bh_result
, int create
)
840 return __get_data_block(inode
, iblock
, bh_result
, create
,
841 F2FS_GET_BLOCK_DIO
, NULL
);
844 static int get_data_block_bmap(struct inode
*inode
, sector_t iblock
,
845 struct buffer_head
*bh_result
, int create
)
847 /* Block number less than F2FS MAX BLOCKS */
848 if (unlikely(iblock
>= F2FS_I_SB(inode
)->max_file_blocks
))
851 return __get_data_block(inode
, iblock
, bh_result
, create
,
852 F2FS_GET_BLOCK_BMAP
, NULL
);
855 static inline sector_t
logical_to_blk(struct inode
*inode
, loff_t offset
)
857 return (offset
>> inode
->i_blkbits
);
860 static inline loff_t
blk_to_logical(struct inode
*inode
, sector_t blk
)
862 return (blk
<< inode
->i_blkbits
);
865 int f2fs_fiemap(struct inode
*inode
, struct fiemap_extent_info
*fieinfo
,
868 struct buffer_head map_bh
;
869 sector_t start_blk
, last_blk
;
872 u64 logical
= 0, phys
= 0, size
= 0;
876 ret
= fiemap_check_flags(fieinfo
, FIEMAP_FLAG_SYNC
);
880 if (f2fs_has_inline_data(inode
)) {
881 ret
= f2fs_inline_data_fiemap(inode
, fieinfo
, start
, len
);
888 isize
= i_size_read(inode
);
892 if (start
+ len
> isize
)
895 if (logical_to_blk(inode
, len
) == 0)
896 len
= blk_to_logical(inode
, 1);
898 start_blk
= logical_to_blk(inode
, start
);
899 last_blk
= logical_to_blk(inode
, start
+ len
- 1);
902 memset(&map_bh
, 0, sizeof(struct buffer_head
));
905 ret
= get_data_block(inode
, start_blk
, &map_bh
, 0,
906 F2FS_GET_BLOCK_FIEMAP
, &next_pgofs
);
911 if (!buffer_mapped(&map_bh
)) {
912 start_blk
= next_pgofs
;
913 /* Go through holes util pass the EOF */
914 if (blk_to_logical(inode
, start_blk
) < isize
)
916 /* Found a hole beyond isize means no more extents.
917 * Note that the premise is that filesystems don't
918 * punch holes beyond isize and keep size unchanged.
920 flags
|= FIEMAP_EXTENT_LAST
;
924 if (f2fs_encrypted_inode(inode
))
925 flags
|= FIEMAP_EXTENT_DATA_ENCRYPTED
;
927 ret
= fiemap_fill_next_extent(fieinfo
, logical
,
931 if (start_blk
> last_blk
|| ret
)
934 logical
= blk_to_logical(inode
, start_blk
);
935 phys
= blk_to_logical(inode
, map_bh
.b_blocknr
);
936 size
= map_bh
.b_size
;
938 if (buffer_unwritten(&map_bh
))
939 flags
= FIEMAP_EXTENT_UNWRITTEN
;
941 start_blk
+= logical_to_blk(inode
, size
);
945 if (fatal_signal_pending(current
))
958 * This function was originally taken from fs/mpage.c, and customized for f2fs.
959 * Major change was from block_size == page_size in f2fs by default.
961 static int f2fs_mpage_readpages(struct address_space
*mapping
,
962 struct list_head
*pages
, struct page
*page
,
965 struct bio
*bio
= NULL
;
967 sector_t last_block_in_bio
= 0;
968 struct inode
*inode
= mapping
->host
;
969 const unsigned blkbits
= inode
->i_blkbits
;
970 const unsigned blocksize
= 1 << blkbits
;
971 sector_t block_in_file
;
973 sector_t last_block_in_file
;
975 struct block_device
*bdev
= inode
->i_sb
->s_bdev
;
976 struct f2fs_map_blocks map
;
982 map
.m_next_pgofs
= NULL
;
984 for (page_idx
= 0; nr_pages
; page_idx
++, nr_pages
--) {
986 prefetchw(&page
->flags
);
988 page
= list_entry(pages
->prev
, struct page
, lru
);
989 list_del(&page
->lru
);
990 if (add_to_page_cache_lru(page
, mapping
,
991 page
->index
, GFP_KERNEL
))
995 block_in_file
= (sector_t
)page
->index
;
996 last_block
= block_in_file
+ nr_pages
;
997 last_block_in_file
= (i_size_read(inode
) + blocksize
- 1) >>
999 if (last_block
> last_block_in_file
)
1000 last_block
= last_block_in_file
;
1003 * Map blocks using the previous result first.
1005 if ((map
.m_flags
& F2FS_MAP_MAPPED
) &&
1006 block_in_file
> map
.m_lblk
&&
1007 block_in_file
< (map
.m_lblk
+ map
.m_len
))
1011 * Then do more f2fs_map_blocks() calls until we are
1012 * done with this page.
1016 if (block_in_file
< last_block
) {
1017 map
.m_lblk
= block_in_file
;
1018 map
.m_len
= last_block
- block_in_file
;
1020 if (f2fs_map_blocks(inode
, &map
, 0,
1021 F2FS_GET_BLOCK_READ
))
1022 goto set_error_page
;
1025 if ((map
.m_flags
& F2FS_MAP_MAPPED
)) {
1026 block_nr
= map
.m_pblk
+ block_in_file
- map
.m_lblk
;
1027 SetPageMappedToDisk(page
);
1029 if (!PageUptodate(page
) && !cleancache_get_page(page
)) {
1030 SetPageUptodate(page
);
1034 zero_user_segment(page
, 0, PAGE_SIZE
);
1035 SetPageUptodate(page
);
1041 * This page will go to BIO. Do we need to send this
1044 if (bio
&& (last_block_in_bio
!= block_nr
- 1)) {
1046 __submit_bio(F2FS_I_SB(inode
), READ
, bio
, DATA
);
1050 struct fscrypt_ctx
*ctx
= NULL
;
1052 if (f2fs_encrypted_inode(inode
) &&
1053 S_ISREG(inode
->i_mode
)) {
1055 ctx
= fscrypt_get_ctx(inode
, GFP_NOFS
);
1057 goto set_error_page
;
1059 /* wait the page to be moved by cleaning */
1060 f2fs_wait_on_encrypted_page_writeback(
1061 F2FS_I_SB(inode
), block_nr
);
1064 bio
= bio_alloc(GFP_KERNEL
,
1065 min_t(int, nr_pages
, BIO_MAX_PAGES
));
1068 fscrypt_release_ctx(ctx
);
1069 goto set_error_page
;
1071 bio
->bi_bdev
= bdev
;
1072 bio
->bi_iter
.bi_sector
= SECTOR_FROM_BLOCK(block_nr
);
1073 bio
->bi_end_io
= f2fs_read_end_io
;
1074 bio
->bi_private
= ctx
;
1077 if (bio_add_page(bio
, page
, blocksize
, 0) < blocksize
)
1078 goto submit_and_realloc
;
1080 last_block_in_bio
= block_nr
;
1084 zero_user_segment(page
, 0, PAGE_SIZE
);
1089 __submit_bio(F2FS_I_SB(inode
), READ
, bio
, DATA
);
1097 BUG_ON(pages
&& !list_empty(pages
));
1099 __submit_bio(F2FS_I_SB(inode
), READ
, bio
, DATA
);
1103 static int f2fs_read_data_page(struct file
*file
, struct page
*page
)
1105 struct inode
*inode
= page
->mapping
->host
;
1108 trace_f2fs_readpage(page
, DATA
);
1110 /* If the file has inline data, try to read it directly */
1111 if (f2fs_has_inline_data(inode
))
1112 ret
= f2fs_read_inline_data(inode
, page
);
1114 ret
= f2fs_mpage_readpages(page
->mapping
, NULL
, page
, 1);
1118 static int f2fs_read_data_pages(struct file
*file
,
1119 struct address_space
*mapping
,
1120 struct list_head
*pages
, unsigned nr_pages
)
1122 struct inode
*inode
= file
->f_mapping
->host
;
1123 struct page
*page
= list_entry(pages
->prev
, struct page
, lru
);
1125 trace_f2fs_readpages(inode
, page
, nr_pages
);
1127 /* If the file has inline data, skip readpages */
1128 if (f2fs_has_inline_data(inode
))
1131 return f2fs_mpage_readpages(mapping
, pages
, NULL
, nr_pages
);
1134 int do_write_data_page(struct f2fs_io_info
*fio
)
1136 struct page
*page
= fio
->page
;
1137 struct inode
*inode
= page
->mapping
->host
;
1138 struct dnode_of_data dn
;
1141 set_new_dnode(&dn
, inode
, NULL
, NULL
, 0);
1142 err
= get_dnode_of_data(&dn
, page
->index
, LOOKUP_NODE
);
1146 fio
->old_blkaddr
= dn
.data_blkaddr
;
1148 /* This page is already truncated */
1149 if (fio
->old_blkaddr
== NULL_ADDR
) {
1150 ClearPageUptodate(page
);
1154 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
)) {
1155 gfp_t gfp_flags
= GFP_NOFS
;
1157 /* wait for GCed encrypted page writeback */
1158 f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode
),
1161 fio
->encrypted_page
= fscrypt_encrypt_page(inode
, fio
->page
,
1163 if (IS_ERR(fio
->encrypted_page
)) {
1164 err
= PTR_ERR(fio
->encrypted_page
);
1165 if (err
== -ENOMEM
) {
1166 /* flush pending ios and wait for a while */
1167 f2fs_flush_merged_bios(F2FS_I_SB(inode
));
1168 congestion_wait(BLK_RW_ASYNC
, HZ
/50);
1169 gfp_flags
|= __GFP_NOFAIL
;
1177 set_page_writeback(page
);
1180 * If current allocation needs SSR,
1181 * it had better in-place writes for updated data.
1183 if (unlikely(fio
->old_blkaddr
!= NEW_ADDR
&&
1184 !is_cold_data(page
) &&
1185 !IS_ATOMIC_WRITTEN_PAGE(page
) &&
1186 need_inplace_update(inode
))) {
1187 rewrite_data_page(fio
);
1188 set_inode_flag(inode
, FI_UPDATE_WRITE
);
1189 trace_f2fs_do_write_data_page(page
, IPU
);
1191 write_data_page(&dn
, fio
);
1192 trace_f2fs_do_write_data_page(page
, OPU
);
1193 set_inode_flag(inode
, FI_APPEND_WRITE
);
1194 if (page
->index
== 0)
1195 set_inode_flag(inode
, FI_FIRST_BLOCK_WRITTEN
);
1198 f2fs_put_dnode(&dn
);
1202 static int f2fs_write_data_page(struct page
*page
,
1203 struct writeback_control
*wbc
)
1205 struct inode
*inode
= page
->mapping
->host
;
1206 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1207 loff_t i_size
= i_size_read(inode
);
1208 const pgoff_t end_index
= ((unsigned long long) i_size
)
1210 loff_t psize
= (page
->index
+ 1) << PAGE_SHIFT
;
1211 unsigned offset
= 0;
1212 bool need_balance_fs
= false;
1214 struct f2fs_io_info fio
= {
1217 .rw
= (wbc
->sync_mode
== WB_SYNC_ALL
) ? WRITE_SYNC
: WRITE
,
1219 .encrypted_page
= NULL
,
1222 trace_f2fs_writepage(page
, DATA
);
1224 if (page
->index
< end_index
)
1228 * If the offset is out-of-range of file size,
1229 * this page does not have to be written to disk.
1231 offset
= i_size
& (PAGE_SIZE
- 1);
1232 if ((page
->index
>= end_index
+ 1) || !offset
)
1235 zero_user_segment(page
, offset
, PAGE_SIZE
);
1237 if (unlikely(is_sbi_flag_set(sbi
, SBI_POR_DOING
)))
1239 if (f2fs_is_drop_cache(inode
))
1241 /* we should not write 0'th page having journal header */
1242 if (f2fs_is_volatile_file(inode
) && (!page
->index
||
1243 (!wbc
->for_reclaim
&&
1244 available_free_memory(sbi
, BASE_CHECK
))))
1247 /* we should bypass data pages to proceed the kworkder jobs */
1248 if (unlikely(f2fs_cp_error(sbi
))) {
1249 mapping_set_error(page
->mapping
, -EIO
);
1253 /* Dentry blocks are controlled by checkpoint */
1254 if (S_ISDIR(inode
->i_mode
)) {
1255 err
= do_write_data_page(&fio
);
1259 if (!wbc
->for_reclaim
)
1260 need_balance_fs
= true;
1261 else if (has_not_enough_free_secs(sbi
, 0))
1266 if (f2fs_has_inline_data(inode
))
1267 err
= f2fs_write_inline_data(inode
, page
);
1269 err
= do_write_data_page(&fio
);
1270 if (F2FS_I(inode
)->last_disk_size
< psize
)
1271 F2FS_I(inode
)->last_disk_size
= psize
;
1272 f2fs_unlock_op(sbi
);
1274 if (err
&& err
!= -ENOENT
)
1277 clear_cold_data(page
);
1279 inode_dec_dirty_pages(inode
);
1281 ClearPageUptodate(page
);
1283 if (wbc
->for_reclaim
) {
1284 f2fs_submit_merged_bio_cond(sbi
, NULL
, page
, 0, DATA
, WRITE
);
1285 remove_dirty_inode(inode
);
1289 f2fs_balance_fs(sbi
, need_balance_fs
);
1291 if (unlikely(f2fs_cp_error(sbi
)))
1292 f2fs_submit_merged_bio(sbi
, DATA
, WRITE
);
1297 redirty_page_for_writepage(wbc
, page
);
1303 * This function was copied from write_cche_pages from mm/page-writeback.c.
1304 * The major change is making write step of cold data page separately from
1305 * warm/hot data page.
1307 static int f2fs_write_cache_pages(struct address_space
*mapping
,
1308 struct writeback_control
*wbc
)
1312 struct pagevec pvec
;
1314 pgoff_t
uninitialized_var(writeback_index
);
1316 pgoff_t end
; /* Inclusive */
1319 int range_whole
= 0;
1322 pagevec_init(&pvec
, 0);
1324 if (wbc
->range_cyclic
) {
1325 writeback_index
= mapping
->writeback_index
; /* prev offset */
1326 index
= writeback_index
;
1333 index
= wbc
->range_start
>> PAGE_SHIFT
;
1334 end
= wbc
->range_end
>> PAGE_SHIFT
;
1335 if (wbc
->range_start
== 0 && wbc
->range_end
== LLONG_MAX
)
1337 cycled
= 1; /* ignore range_cyclic tests */
1339 if (wbc
->sync_mode
== WB_SYNC_ALL
|| wbc
->tagged_writepages
)
1340 tag
= PAGECACHE_TAG_TOWRITE
;
1342 tag
= PAGECACHE_TAG_DIRTY
;
1344 if (wbc
->sync_mode
== WB_SYNC_ALL
|| wbc
->tagged_writepages
)
1345 tag_pages_for_writeback(mapping
, index
, end
);
1347 while (!done
&& (index
<= end
)) {
1350 nr_pages
= pagevec_lookup_tag(&pvec
, mapping
, &index
, tag
,
1351 min(end
- index
, (pgoff_t
)PAGEVEC_SIZE
- 1) + 1);
1355 for (i
= 0; i
< nr_pages
; i
++) {
1356 struct page
*page
= pvec
.pages
[i
];
1358 if (page
->index
> end
) {
1363 done_index
= page
->index
;
1367 if (unlikely(page
->mapping
!= mapping
)) {
1373 if (!PageDirty(page
)) {
1374 /* someone wrote it for us */
1375 goto continue_unlock
;
1378 if (PageWriteback(page
)) {
1379 if (wbc
->sync_mode
!= WB_SYNC_NONE
)
1380 f2fs_wait_on_page_writeback(page
,
1383 goto continue_unlock
;
1386 BUG_ON(PageWriteback(page
));
1387 if (!clear_page_dirty_for_io(page
))
1388 goto continue_unlock
;
1390 ret
= mapping
->a_ops
->writepage(page
, wbc
);
1391 if (unlikely(ret
)) {
1392 done_index
= page
->index
+ 1;
1397 if (--wbc
->nr_to_write
<= 0 &&
1398 wbc
->sync_mode
== WB_SYNC_NONE
) {
1403 pagevec_release(&pvec
);
1407 if (!cycled
&& !done
) {
1410 end
= writeback_index
- 1;
1413 if (wbc
->range_cyclic
|| (range_whole
&& wbc
->nr_to_write
> 0))
1414 mapping
->writeback_index
= done_index
;
1419 static int f2fs_write_data_pages(struct address_space
*mapping
,
1420 struct writeback_control
*wbc
)
1422 struct inode
*inode
= mapping
->host
;
1423 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1426 /* deal with chardevs and other special file */
1427 if (!mapping
->a_ops
->writepage
)
1430 /* skip writing if there is no dirty page in this inode */
1431 if (!get_dirty_pages(inode
) && wbc
->sync_mode
== WB_SYNC_NONE
)
1434 if (S_ISDIR(inode
->i_mode
) && wbc
->sync_mode
== WB_SYNC_NONE
&&
1435 get_dirty_pages(inode
) < nr_pages_to_skip(sbi
, DATA
) &&
1436 available_free_memory(sbi
, DIRTY_DENTS
))
1439 /* skip writing during file defragment */
1440 if (is_inode_flag_set(inode
, FI_DO_DEFRAG
))
1443 /* during POR, we don't need to trigger writepage at all. */
1444 if (unlikely(is_sbi_flag_set(sbi
, SBI_POR_DOING
)))
1447 trace_f2fs_writepages(mapping
->host
, wbc
, DATA
);
1449 ret
= f2fs_write_cache_pages(mapping
, wbc
);
1451 * if some pages were truncated, we cannot guarantee its mapping->host
1452 * to detect pending bios.
1454 f2fs_submit_merged_bio(sbi
, DATA
, WRITE
);
1456 remove_dirty_inode(inode
);
1460 wbc
->pages_skipped
+= get_dirty_pages(inode
);
1461 trace_f2fs_writepages(mapping
->host
, wbc
, DATA
);
1465 static void f2fs_write_failed(struct address_space
*mapping
, loff_t to
)
1467 struct inode
*inode
= mapping
->host
;
1468 loff_t i_size
= i_size_read(inode
);
1471 truncate_pagecache(inode
, i_size
);
1472 truncate_blocks(inode
, i_size
, true);
1476 static int prepare_write_begin(struct f2fs_sb_info
*sbi
,
1477 struct page
*page
, loff_t pos
, unsigned len
,
1478 block_t
*blk_addr
, bool *node_changed
)
1480 struct inode
*inode
= page
->mapping
->host
;
1481 pgoff_t index
= page
->index
;
1482 struct dnode_of_data dn
;
1484 bool locked
= false;
1485 struct extent_info ei
;
1489 * we already allocated all the blocks, so we don't need to get
1490 * the block addresses when there is no need to fill the page.
1492 if (!f2fs_has_inline_data(inode
) && !f2fs_encrypted_inode(inode
) &&
1496 if (f2fs_has_inline_data(inode
) ||
1497 (pos
& PAGE_MASK
) >= i_size_read(inode
)) {
1502 /* check inline_data */
1503 ipage
= get_node_page(sbi
, inode
->i_ino
);
1504 if (IS_ERR(ipage
)) {
1505 err
= PTR_ERR(ipage
);
1509 set_new_dnode(&dn
, inode
, ipage
, ipage
, 0);
1511 if (f2fs_has_inline_data(inode
)) {
1512 if (pos
+ len
<= MAX_INLINE_DATA
) {
1513 read_inline_data(page
, ipage
);
1514 set_inode_flag(inode
, FI_DATA_EXIST
);
1516 set_inline_node(ipage
);
1518 err
= f2fs_convert_inline_page(&dn
, page
);
1521 if (dn
.data_blkaddr
== NULL_ADDR
)
1522 err
= f2fs_get_block(&dn
, index
);
1524 } else if (locked
) {
1525 err
= f2fs_get_block(&dn
, index
);
1527 if (f2fs_lookup_extent_cache(inode
, index
, &ei
)) {
1528 dn
.data_blkaddr
= ei
.blk
+ index
- ei
.fofs
;
1531 err
= get_dnode_of_data(&dn
, index
, LOOKUP_NODE
);
1532 if (err
|| dn
.data_blkaddr
== NULL_ADDR
) {
1533 f2fs_put_dnode(&dn
);
1541 /* convert_inline_page can make node_changed */
1542 *blk_addr
= dn
.data_blkaddr
;
1543 *node_changed
= dn
.node_changed
;
1545 f2fs_put_dnode(&dn
);
1548 f2fs_unlock_op(sbi
);
1552 static int f2fs_write_begin(struct file
*file
, struct address_space
*mapping
,
1553 loff_t pos
, unsigned len
, unsigned flags
,
1554 struct page
**pagep
, void **fsdata
)
1556 struct inode
*inode
= mapping
->host
;
1557 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1558 struct page
*page
= NULL
;
1559 pgoff_t index
= ((unsigned long long) pos
) >> PAGE_SHIFT
;
1560 bool need_balance
= false;
1561 block_t blkaddr
= NULL_ADDR
;
1564 trace_f2fs_write_begin(inode
, pos
, len
, flags
);
1567 * We should check this at this moment to avoid deadlock on inode page
1568 * and #0 page. The locking rule for inline_data conversion should be:
1569 * lock_page(page #0) -> lock_page(inode_page)
1572 err
= f2fs_convert_inline_inode(inode
);
1577 page
= grab_cache_page_write_begin(mapping
, index
, flags
);
1585 err
= prepare_write_begin(sbi
, page
, pos
, len
,
1586 &blkaddr
, &need_balance
);
1590 if (need_balance
&& has_not_enough_free_secs(sbi
, 0)) {
1592 f2fs_balance_fs(sbi
, true);
1594 if (page
->mapping
!= mapping
) {
1595 /* The page got truncated from under us */
1596 f2fs_put_page(page
, 1);
1601 f2fs_wait_on_page_writeback(page
, DATA
, false);
1603 /* wait for GCed encrypted page writeback */
1604 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
1605 f2fs_wait_on_encrypted_page_writeback(sbi
, blkaddr
);
1607 if (len
== PAGE_SIZE
)
1609 if (PageUptodate(page
))
1612 if ((pos
& PAGE_MASK
) >= i_size_read(inode
)) {
1613 unsigned start
= pos
& (PAGE_SIZE
- 1);
1614 unsigned end
= start
+ len
;
1616 /* Reading beyond i_size is simple: memset to zero */
1617 zero_user_segments(page
, 0, start
, end
, PAGE_SIZE
);
1621 if (blkaddr
== NEW_ADDR
) {
1622 zero_user_segment(page
, 0, PAGE_SIZE
);
1624 struct f2fs_io_info fio
= {
1628 .old_blkaddr
= blkaddr
,
1629 .new_blkaddr
= blkaddr
,
1631 .encrypted_page
= NULL
,
1633 err
= f2fs_submit_page_bio(&fio
);
1638 if (unlikely(!PageUptodate(page
))) {
1642 if (unlikely(page
->mapping
!= mapping
)) {
1643 f2fs_put_page(page
, 1);
1647 /* avoid symlink page */
1648 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
)) {
1649 err
= fscrypt_decrypt_page(page
);
1655 SetPageUptodate(page
);
1657 clear_cold_data(page
);
1661 f2fs_put_page(page
, 1);
1662 f2fs_write_failed(mapping
, pos
+ len
);
1666 static int f2fs_write_end(struct file
*file
,
1667 struct address_space
*mapping
,
1668 loff_t pos
, unsigned len
, unsigned copied
,
1669 struct page
*page
, void *fsdata
)
1671 struct inode
*inode
= page
->mapping
->host
;
1673 trace_f2fs_write_end(inode
, pos
, len
, copied
);
1675 set_page_dirty(page
);
1677 if (pos
+ copied
> i_size_read(inode
))
1678 f2fs_i_size_write(inode
, pos
+ copied
);
1680 f2fs_put_page(page
, 1);
1681 f2fs_update_time(F2FS_I_SB(inode
), REQ_TIME
);
1685 static int check_direct_IO(struct inode
*inode
, struct iov_iter
*iter
,
1688 unsigned blocksize_mask
= inode
->i_sb
->s_blocksize
- 1;
1690 if (offset
& blocksize_mask
)
1693 if (iov_iter_alignment(iter
) & blocksize_mask
)
1699 static ssize_t
f2fs_direct_IO(struct kiocb
*iocb
, struct iov_iter
*iter
)
1701 struct address_space
*mapping
= iocb
->ki_filp
->f_mapping
;
1702 struct inode
*inode
= mapping
->host
;
1703 size_t count
= iov_iter_count(iter
);
1704 loff_t offset
= iocb
->ki_pos
;
1707 err
= check_direct_IO(inode
, iter
, offset
);
1711 if (f2fs_encrypted_inode(inode
) && S_ISREG(inode
->i_mode
))
1714 trace_f2fs_direct_IO_enter(inode
, offset
, count
, iov_iter_rw(iter
));
1716 err
= blockdev_direct_IO(iocb
, inode
, iter
, get_data_block_dio
);
1717 if (iov_iter_rw(iter
) == WRITE
) {
1719 set_inode_flag(inode
, FI_UPDATE_WRITE
);
1721 f2fs_write_failed(mapping
, offset
+ count
);
1724 trace_f2fs_direct_IO_exit(inode
, offset
, count
, iov_iter_rw(iter
), err
);
1729 void f2fs_invalidate_page(struct page
*page
, unsigned int offset
,
1730 unsigned int length
)
1732 struct inode
*inode
= page
->mapping
->host
;
1733 struct f2fs_sb_info
*sbi
= F2FS_I_SB(inode
);
1735 if (inode
->i_ino
>= F2FS_ROOT_INO(sbi
) &&
1736 (offset
% PAGE_SIZE
|| length
!= PAGE_SIZE
))
1739 if (PageDirty(page
)) {
1740 if (inode
->i_ino
== F2FS_META_INO(sbi
))
1741 dec_page_count(sbi
, F2FS_DIRTY_META
);
1742 else if (inode
->i_ino
== F2FS_NODE_INO(sbi
))
1743 dec_page_count(sbi
, F2FS_DIRTY_NODES
);
1745 inode_dec_dirty_pages(inode
);
1748 /* This is atomic written page, keep Private */
1749 if (IS_ATOMIC_WRITTEN_PAGE(page
))
1752 set_page_private(page
, 0);
1753 ClearPagePrivate(page
);
1756 int f2fs_release_page(struct page
*page
, gfp_t wait
)
1758 /* If this is dirty page, keep PagePrivate */
1759 if (PageDirty(page
))
1762 /* This is atomic written page, keep Private */
1763 if (IS_ATOMIC_WRITTEN_PAGE(page
))
1766 set_page_private(page
, 0);
1767 ClearPagePrivate(page
);
1771 static int f2fs_set_data_page_dirty(struct page
*page
)
1773 struct address_space
*mapping
= page
->mapping
;
1774 struct inode
*inode
= mapping
->host
;
1776 trace_f2fs_set_page_dirty(page
, DATA
);
1778 SetPageUptodate(page
);
1780 if (f2fs_is_atomic_file(inode
)) {
1781 if (!IS_ATOMIC_WRITTEN_PAGE(page
)) {
1782 register_inmem_page(inode
, page
);
1786 * Previously, this page has been registered, we just
1792 if (!PageDirty(page
)) {
1793 __set_page_dirty_nobuffers(page
);
1794 update_dirty_page(inode
, page
);
1800 static sector_t
f2fs_bmap(struct address_space
*mapping
, sector_t block
)
1802 struct inode
*inode
= mapping
->host
;
1804 if (f2fs_has_inline_data(inode
))
1807 /* make sure allocating whole blocks */
1808 if (mapping_tagged(mapping
, PAGECACHE_TAG_DIRTY
))
1809 filemap_write_and_wait(mapping
);
1811 return generic_block_bmap(mapping
, block
, get_data_block_bmap
);
1814 const struct address_space_operations f2fs_dblock_aops
= {
1815 .readpage
= f2fs_read_data_page
,
1816 .readpages
= f2fs_read_data_pages
,
1817 .writepage
= f2fs_write_data_page
,
1818 .writepages
= f2fs_write_data_pages
,
1819 .write_begin
= f2fs_write_begin
,
1820 .write_end
= f2fs_write_end
,
1821 .set_page_dirty
= f2fs_set_data_page_dirty
,
1822 .invalidatepage
= f2fs_invalidate_page
,
1823 .releasepage
= f2fs_release_page
,
1824 .direct_IO
= f2fs_direct_IO
,