4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/file.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * ext4 fs regular file handling primitives
17 * 64-bit file support on 64-bit platforms by Jakub Jelinek
18 * (jj@sunsite.ms.mff.cuni.cz)
21 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/mount.h>
25 #include <linux/path.h>
26 #include <linux/quotaops.h>
27 #include <linux/pagevec.h>
29 #include "ext4_jbd2.h"
34 * Called when an inode is released. Note that this is different
35 * from ext4_file_open: open gets called at every open, but release
36 * gets called only when /all/ the files are closed.
38 static int ext4_release_file(struct inode
*inode
, struct file
*filp
)
40 if (ext4_test_inode_state(inode
, EXT4_STATE_DA_ALLOC_CLOSE
)) {
41 ext4_alloc_da_blocks(inode
);
42 ext4_clear_inode_state(inode
, EXT4_STATE_DA_ALLOC_CLOSE
);
44 /* if we are the last writer on the inode, drop the block reservation */
45 if ((filp
->f_mode
& FMODE_WRITE
) &&
46 (atomic_read(&inode
->i_writecount
) == 1) &&
47 !EXT4_I(inode
)->i_reserved_data_blocks
)
49 down_write(&EXT4_I(inode
)->i_data_sem
);
50 ext4_discard_preallocations(inode
);
51 up_write(&EXT4_I(inode
)->i_data_sem
);
53 if (is_dx(inode
) && filp
->private_data
)
54 ext4_htree_free_dir_info(filp
->private_data
);
59 void ext4_unwritten_wait(struct inode
*inode
)
61 wait_queue_head_t
*wq
= ext4_ioend_wq(inode
);
63 wait_event(*wq
, (atomic_read(&EXT4_I(inode
)->i_unwritten
) == 0));
67 * This tests whether the IO in question is block-aligned or not.
68 * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they
69 * are converted to written only after the IO is complete. Until they are
70 * mapped, these blocks appear as holes, so dio_zero_block() will assume that
71 * it needs to zero out portions of the start and/or end block. If 2 AIO
72 * threads are at work on the same unwritten block, they must be synchronized
73 * or one thread will zero the other's data, causing corruption.
76 ext4_unaligned_aio(struct inode
*inode
, const struct iovec
*iov
,
77 unsigned long nr_segs
, loff_t pos
)
79 struct super_block
*sb
= inode
->i_sb
;
80 int blockmask
= sb
->s_blocksize
- 1;
81 size_t count
= iov_length(iov
, nr_segs
);
82 loff_t final_size
= pos
+ count
;
84 if (pos
>= inode
->i_size
)
87 if ((pos
& blockmask
) || (final_size
& blockmask
))
94 ext4_file_dio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
95 unsigned long nr_segs
, loff_t pos
)
97 struct file
*file
= iocb
->ki_filp
;
98 struct inode
*inode
= file
->f_mapping
->host
;
100 int unaligned_aio
= 0;
103 size_t length
= iov_length(iov
, nr_segs
);
105 if (ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
) &&
106 !is_sync_kiocb(iocb
))
107 unaligned_aio
= ext4_unaligned_aio(inode
, iov
, nr_segs
, pos
);
109 /* Unaligned direct AIO must be serialized; see comment above */
111 mutex_lock(ext4_aio_mutex(inode
));
112 ext4_unwritten_wait(inode
);
115 BUG_ON(iocb
->ki_pos
!= pos
);
117 mutex_lock(&inode
->i_mutex
);
118 blk_start_plug(&plug
);
120 iocb
->private = &overwrite
;
122 /* check whether we do a DIO overwrite or not */
123 if (ext4_should_dioread_nolock(inode
) && !unaligned_aio
&&
124 !file
->f_mapping
->nrpages
&& pos
+ length
<= i_size_read(inode
)) {
125 struct ext4_map_blocks map
;
126 unsigned int blkbits
= inode
->i_blkbits
;
129 map
.m_lblk
= pos
>> blkbits
;
130 map
.m_len
= (EXT4_BLOCK_ALIGN(pos
+ length
, blkbits
) >> blkbits
)
134 err
= ext4_map_blocks(NULL
, inode
, &map
, 0);
136 * 'err==len' means that all of blocks has been preallocated no
137 * matter they are initialized or not. For excluding
138 * uninitialized extents, we need to check m_flags. There are
139 * two conditions that indicate for initialized extents.
140 * 1) If we hit extent cache, EXT4_MAP_MAPPED flag is returned;
141 * 2) If we do a real lookup, non-flags are returned.
142 * So we should check these two conditions.
144 if (err
== len
&& (map
.m_flags
& EXT4_MAP_MAPPED
))
148 ret
= __generic_file_aio_write(iocb
, iov
, nr_segs
, &iocb
->ki_pos
);
149 mutex_unlock(&inode
->i_mutex
);
151 if (ret
> 0 || ret
== -EIOCBQUEUED
) {
154 err
= generic_write_sync(file
, pos
, ret
);
155 if (err
< 0 && ret
> 0)
158 blk_finish_plug(&plug
);
161 mutex_unlock(ext4_aio_mutex(inode
));
167 ext4_file_write(struct kiocb
*iocb
, const struct iovec
*iov
,
168 unsigned long nr_segs
, loff_t pos
)
170 struct inode
*inode
= file_inode(iocb
->ki_filp
);
174 * If we have encountered a bitmap-format file, the size limit
175 * is smaller than s_maxbytes, which is for extent-mapped files.
178 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
))) {
179 struct ext4_sb_info
*sbi
= EXT4_SB(inode
->i_sb
);
180 size_t length
= iov_length(iov
, nr_segs
);
182 if ((pos
> sbi
->s_bitmap_maxbytes
||
183 (pos
== sbi
->s_bitmap_maxbytes
&& length
> 0)))
186 if (pos
+ length
> sbi
->s_bitmap_maxbytes
) {
187 nr_segs
= iov_shorten((struct iovec
*)iov
, nr_segs
,
188 sbi
->s_bitmap_maxbytes
- pos
);
192 if (unlikely(iocb
->ki_filp
->f_flags
& O_DIRECT
))
193 ret
= ext4_file_dio_write(iocb
, iov
, nr_segs
, pos
);
195 ret
= generic_file_aio_write(iocb
, iov
, nr_segs
, pos
);
200 static const struct vm_operations_struct ext4_file_vm_ops
= {
201 .fault
= filemap_fault
,
202 .page_mkwrite
= ext4_page_mkwrite
,
203 .remap_pages
= generic_file_remap_pages
,
206 static int ext4_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
208 struct address_space
*mapping
= file
->f_mapping
;
210 if (!mapping
->a_ops
->readpage
)
213 vma
->vm_ops
= &ext4_file_vm_ops
;
217 static int ext4_file_open(struct inode
* inode
, struct file
* filp
)
219 struct super_block
*sb
= inode
->i_sb
;
220 struct ext4_sb_info
*sbi
= EXT4_SB(inode
->i_sb
);
221 struct ext4_inode_info
*ei
= EXT4_I(inode
);
222 struct vfsmount
*mnt
= filp
->f_path
.mnt
;
226 if (unlikely(!(sbi
->s_mount_flags
& EXT4_MF_MNTDIR_SAMPLED
) &&
227 !(sb
->s_flags
& MS_RDONLY
))) {
228 sbi
->s_mount_flags
|= EXT4_MF_MNTDIR_SAMPLED
;
230 * Sample where the filesystem has been mounted and
231 * store it in the superblock for sysadmin convenience
232 * when trying to sort through large numbers of block
233 * devices or filesystem images.
235 memset(buf
, 0, sizeof(buf
));
237 path
.dentry
= mnt
->mnt_root
;
238 cp
= d_path(&path
, buf
, sizeof(buf
));
243 handle
= ext4_journal_start_sb(sb
, EXT4_HT_MISC
, 1);
245 return PTR_ERR(handle
);
246 err
= ext4_journal_get_write_access(handle
, sbi
->s_sbh
);
248 ext4_journal_stop(handle
);
251 strlcpy(sbi
->s_es
->s_last_mounted
, cp
,
252 sizeof(sbi
->s_es
->s_last_mounted
));
253 ext4_handle_dirty_super(handle
, sb
);
254 ext4_journal_stop(handle
);
258 * Set up the jbd2_inode if we are opening the inode for
259 * writing and the journal is present
261 if (sbi
->s_journal
&& !ei
->jinode
&& (filp
->f_mode
& FMODE_WRITE
)) {
262 struct jbd2_inode
*jinode
= jbd2_alloc_inode(GFP_KERNEL
);
264 spin_lock(&inode
->i_lock
);
267 spin_unlock(&inode
->i_lock
);
271 jbd2_journal_init_jbd_inode(ei
->jinode
, inode
);
274 spin_unlock(&inode
->i_lock
);
275 if (unlikely(jinode
!= NULL
))
276 jbd2_free_inode(jinode
);
278 return dquot_file_open(inode
, filp
);
282 * Here we use ext4_map_blocks() to get a block mapping for a extent-based
283 * file rather than ext4_ext_walk_space() because we can introduce
284 * SEEK_DATA/SEEK_HOLE for block-mapped and extent-mapped file at the same
285 * function. When extent status tree has been fully implemented, it will
286 * track all extent status for a file and we can directly use it to
287 * retrieve the offset for SEEK_DATA/SEEK_HOLE.
291 * When we retrieve the offset for SEEK_DATA/SEEK_HOLE, we would need to
292 * lookup page cache to check whether or not there has some data between
293 * [startoff, endoff] because, if this range contains an unwritten extent,
294 * we determine this extent as a data or a hole according to whether the
295 * page cache has data or not.
297 static int ext4_find_unwritten_pgoff(struct inode
*inode
,
299 struct ext4_map_blocks
*map
,
303 unsigned int blkbits
;
311 blkbits
= inode
->i_sb
->s_blocksize_bits
;
314 endoff
= (map
->m_lblk
+ map
->m_len
) << blkbits
;
316 index
= startoff
>> PAGE_CACHE_SHIFT
;
317 end
= endoff
>> PAGE_CACHE_SHIFT
;
319 pagevec_init(&pvec
, 0);
322 unsigned long nr_pages
;
324 num
= min_t(pgoff_t
, end
- index
, PAGEVEC_SIZE
);
325 nr_pages
= pagevec_lookup(&pvec
, inode
->i_mapping
, index
,
328 if (whence
== SEEK_DATA
)
331 BUG_ON(whence
!= SEEK_HOLE
);
333 * If this is the first time to go into the loop and
334 * offset is not beyond the end offset, it will be a
335 * hole at this offset
337 if (lastoff
== startoff
|| lastoff
< endoff
)
343 * If this is the first time to go into the loop and
344 * offset is smaller than the first page offset, it will be a
345 * hole at this offset.
347 if (lastoff
== startoff
&& whence
== SEEK_HOLE
&&
348 lastoff
< page_offset(pvec
.pages
[0])) {
353 for (i
= 0; i
< nr_pages
; i
++) {
354 struct page
*page
= pvec
.pages
[i
];
355 struct buffer_head
*bh
, *head
;
358 * If the current offset is not beyond the end of given
359 * range, it will be a hole.
361 if (lastoff
< endoff
&& whence
== SEEK_HOLE
&&
370 if (unlikely(page
->mapping
!= inode
->i_mapping
)) {
375 if (!page_has_buffers(page
)) {
380 if (page_has_buffers(page
)) {
381 lastoff
= page_offset(page
);
382 bh
= head
= page_buffers(page
);
384 if (buffer_uptodate(bh
) ||
385 buffer_unwritten(bh
)) {
386 if (whence
== SEEK_DATA
)
389 if (whence
== SEEK_HOLE
)
393 *offset
= max_t(loff_t
,
398 lastoff
+= bh
->b_size
;
399 bh
= bh
->b_this_page
;
400 } while (bh
!= head
);
403 lastoff
= page_offset(page
) + PAGE_SIZE
;
408 * The no. of pages is less than our desired, that would be a
411 if (nr_pages
< num
&& whence
== SEEK_HOLE
) {
417 index
= pvec
.pages
[i
- 1]->index
+ 1;
418 pagevec_release(&pvec
);
419 } while (index
<= end
);
422 pagevec_release(&pvec
);
427 * ext4_seek_data() retrieves the offset for SEEK_DATA.
429 static loff_t
ext4_seek_data(struct file
*file
, loff_t offset
, loff_t maxsize
)
431 struct inode
*inode
= file
->f_mapping
->host
;
432 struct ext4_map_blocks map
;
433 struct extent_status es
;
434 ext4_lblk_t start
, last
, end
;
435 loff_t dataoff
, isize
;
439 mutex_lock(&inode
->i_mutex
);
441 isize
= i_size_read(inode
);
442 if (offset
>= isize
) {
443 mutex_unlock(&inode
->i_mutex
);
447 blkbits
= inode
->i_sb
->s_blocksize_bits
;
448 start
= offset
>> blkbits
;
450 end
= isize
>> blkbits
;
455 map
.m_len
= end
- last
+ 1;
456 ret
= ext4_map_blocks(NULL
, inode
, &map
, 0);
457 if (ret
> 0 && !(map
.m_flags
& EXT4_MAP_UNWRITTEN
)) {
459 dataoff
= last
<< blkbits
;
464 * If there is a delay extent at this offset,
465 * it will be as a data.
467 ext4_es_find_delayed_extent(inode
, last
, &es
);
468 if (es
.es_len
!= 0 && in_range(last
, es
.es_lblk
, es
.es_len
)) {
470 dataoff
= last
<< blkbits
;
475 * If there is a unwritten extent at this offset,
476 * it will be as a data or a hole according to page
477 * cache that has data or not.
479 if (map
.m_flags
& EXT4_MAP_UNWRITTEN
) {
481 unwritten
= ext4_find_unwritten_pgoff(inode
, SEEK_DATA
,
488 dataoff
= last
<< blkbits
;
489 } while (last
<= end
);
491 mutex_unlock(&inode
->i_mutex
);
496 if (dataoff
< 0 && !(file
->f_mode
& FMODE_UNSIGNED_OFFSET
))
498 if (dataoff
> maxsize
)
501 if (dataoff
!= file
->f_pos
) {
502 file
->f_pos
= dataoff
;
510 * ext4_seek_hole() retrieves the offset for SEEK_HOLE.
512 static loff_t
ext4_seek_hole(struct file
*file
, loff_t offset
, loff_t maxsize
)
514 struct inode
*inode
= file
->f_mapping
->host
;
515 struct ext4_map_blocks map
;
516 struct extent_status es
;
517 ext4_lblk_t start
, last
, end
;
518 loff_t holeoff
, isize
;
522 mutex_lock(&inode
->i_mutex
);
524 isize
= i_size_read(inode
);
525 if (offset
>= isize
) {
526 mutex_unlock(&inode
->i_mutex
);
530 blkbits
= inode
->i_sb
->s_blocksize_bits
;
531 start
= offset
>> blkbits
;
533 end
= isize
>> blkbits
;
538 map
.m_len
= end
- last
+ 1;
539 ret
= ext4_map_blocks(NULL
, inode
, &map
, 0);
540 if (ret
> 0 && !(map
.m_flags
& EXT4_MAP_UNWRITTEN
)) {
542 holeoff
= last
<< blkbits
;
547 * If there is a delay extent at this offset,
548 * we will skip this extent.
550 ext4_es_find_delayed_extent(inode
, last
, &es
);
551 if (es
.es_len
!= 0 && in_range(last
, es
.es_lblk
, es
.es_len
)) {
552 last
= es
.es_lblk
+ es
.es_len
;
553 holeoff
= last
<< blkbits
;
558 * If there is a unwritten extent at this offset,
559 * it will be as a data or a hole according to page
560 * cache that has data or not.
562 if (map
.m_flags
& EXT4_MAP_UNWRITTEN
) {
564 unwritten
= ext4_find_unwritten_pgoff(inode
, SEEK_HOLE
,
568 holeoff
= last
<< blkbits
;
575 } while (last
<= end
);
577 mutex_unlock(&inode
->i_mutex
);
582 if (holeoff
< 0 && !(file
->f_mode
& FMODE_UNSIGNED_OFFSET
))
584 if (holeoff
> maxsize
)
587 if (holeoff
!= file
->f_pos
) {
588 file
->f_pos
= holeoff
;
596 * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values
597 * by calling generic_file_llseek_size() with the appropriate maxbytes
600 loff_t
ext4_llseek(struct file
*file
, loff_t offset
, int whence
)
602 struct inode
*inode
= file
->f_mapping
->host
;
605 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
606 maxbytes
= EXT4_SB(inode
->i_sb
)->s_bitmap_maxbytes
;
608 maxbytes
= inode
->i_sb
->s_maxbytes
;
614 return generic_file_llseek_size(file
, offset
, whence
,
615 maxbytes
, i_size_read(inode
));
617 return ext4_seek_data(file
, offset
, maxbytes
);
619 return ext4_seek_hole(file
, offset
, maxbytes
);
625 const struct file_operations ext4_file_operations
= {
626 .llseek
= ext4_llseek
,
627 .read
= do_sync_read
,
628 .write
= do_sync_write
,
629 .aio_read
= generic_file_aio_read
,
630 .aio_write
= ext4_file_write
,
631 .unlocked_ioctl
= ext4_ioctl
,
633 .compat_ioctl
= ext4_compat_ioctl
,
635 .mmap
= ext4_file_mmap
,
636 .open
= ext4_file_open
,
637 .release
= ext4_release_file
,
638 .fsync
= ext4_sync_file
,
639 .splice_read
= generic_file_splice_read
,
640 .splice_write
= generic_file_splice_write
,
641 .fallocate
= ext4_fallocate
,
644 const struct inode_operations ext4_file_inode_operations
= {
645 .setattr
= ext4_setattr
,
646 .getattr
= ext4_getattr
,
647 .setxattr
= generic_setxattr
,
648 .getxattr
= generic_getxattr
,
649 .listxattr
= ext4_listxattr
,
650 .removexattr
= generic_removexattr
,
651 .get_acl
= ext4_get_acl
,
652 .fiemap
= ext4_fiemap
,