1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/file.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * ext4 fs regular file handling primitives
18 * 64-bit file support on 64-bit platforms by Jakub Jelinek
19 * (jj@sunsite.ms.mff.cuni.cz)
22 #include <linux/time.h>
24 #include <linux/iomap.h>
25 #include <linux/mount.h>
26 #include <linux/path.h>
27 #include <linux/dax.h>
28 #include <linux/quotaops.h>
29 #include <linux/pagevec.h>
30 #include <linux/uio.h>
31 #include <linux/mman.h>
32 #include <linux/backing-dev.h>
34 #include "ext4_jbd2.h"
40 * Returns %true if the given DIO request should be attempted with DIO, or
41 * %false if it should fall back to buffered I/O.
43 * DIO isn't well specified; when it's unsupported (either due to the request
44 * being misaligned, or due to the file not supporting DIO at all), filesystems
45 * either fall back to buffered I/O or return EINVAL. For files that don't use
46 * any special features like encryption or verity, ext4 has traditionally
47 * returned EINVAL for misaligned DIO. iomap_dio_rw() uses this convention too.
48 * In this case, we should attempt the DIO, *not* fall back to buffered I/O.
50 * In contrast, in cases where DIO is unsupported due to ext4 features, ext4
51 * traditionally falls back to buffered I/O.
53 * This function implements the traditional ext4 behavior in all these cases.
55 static bool ext4_should_use_dio(struct kiocb
*iocb
, struct iov_iter
*iter
)
57 struct inode
*inode
= file_inode(iocb
->ki_filp
);
58 u32 dio_align
= ext4_dio_alignment(inode
);
66 return IS_ALIGNED(iocb
->ki_pos
| iov_iter_alignment(iter
), dio_align
);
69 static ssize_t
ext4_dio_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
72 struct inode
*inode
= file_inode(iocb
->ki_filp
);
74 if (iocb
->ki_flags
& IOCB_NOWAIT
) {
75 if (!inode_trylock_shared(inode
))
78 inode_lock_shared(inode
);
81 if (!ext4_should_use_dio(iocb
, to
)) {
82 inode_unlock_shared(inode
);
84 * Fallback to buffered I/O if the operation being performed on
85 * the inode is not supported by direct I/O. The IOCB_DIRECT
86 * flag needs to be cleared here in order to ensure that the
87 * direct I/O path within generic_file_read_iter() is not
90 iocb
->ki_flags
&= ~IOCB_DIRECT
;
91 return generic_file_read_iter(iocb
, to
);
94 ret
= iomap_dio_rw(iocb
, to
, &ext4_iomap_ops
, NULL
, 0, NULL
, 0);
95 inode_unlock_shared(inode
);
97 file_accessed(iocb
->ki_filp
);
102 static ssize_t
ext4_dax_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
104 struct inode
*inode
= file_inode(iocb
->ki_filp
);
107 if (iocb
->ki_flags
& IOCB_NOWAIT
) {
108 if (!inode_trylock_shared(inode
))
111 inode_lock_shared(inode
);
114 * Recheck under inode lock - at this point we are sure it cannot
117 if (!IS_DAX(inode
)) {
118 inode_unlock_shared(inode
);
119 /* Fallback to buffered IO in case we cannot support DAX */
120 return generic_file_read_iter(iocb
, to
);
122 ret
= dax_iomap_rw(iocb
, to
, &ext4_iomap_ops
);
123 inode_unlock_shared(inode
);
125 file_accessed(iocb
->ki_filp
);
130 static ssize_t
ext4_file_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
132 struct inode
*inode
= file_inode(iocb
->ki_filp
);
134 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
137 if (!iov_iter_count(to
))
138 return 0; /* skip atime */
142 return ext4_dax_read_iter(iocb
, to
);
144 if (iocb
->ki_flags
& IOCB_DIRECT
)
145 return ext4_dio_read_iter(iocb
, to
);
147 return generic_file_read_iter(iocb
, to
);
151 * Called when an inode is released. Note that this is different
152 * from ext4_file_open: open gets called at every open, but release
153 * gets called only when /all/ the files are closed.
155 static int ext4_release_file(struct inode
*inode
, struct file
*filp
)
157 if (ext4_test_inode_state(inode
, EXT4_STATE_DA_ALLOC_CLOSE
)) {
158 ext4_alloc_da_blocks(inode
);
159 ext4_clear_inode_state(inode
, EXT4_STATE_DA_ALLOC_CLOSE
);
161 /* if we are the last writer on the inode, drop the block reservation */
162 if ((filp
->f_mode
& FMODE_WRITE
) &&
163 (atomic_read(&inode
->i_writecount
) == 1) &&
164 !EXT4_I(inode
)->i_reserved_data_blocks
) {
165 down_write(&EXT4_I(inode
)->i_data_sem
);
166 ext4_discard_preallocations(inode
, 0);
167 up_write(&EXT4_I(inode
)->i_data_sem
);
169 if (is_dx(inode
) && filp
->private_data
)
170 ext4_htree_free_dir_info(filp
->private_data
);
176 * This tests whether the IO in question is block-aligned or not.
177 * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they
178 * are converted to written only after the IO is complete. Until they are
179 * mapped, these blocks appear as holes, so dio_zero_block() will assume that
180 * it needs to zero out portions of the start and/or end block. If 2 AIO
181 * threads are at work on the same unwritten block, they must be synchronized
182 * or one thread will zero the other's data, causing corruption.
185 ext4_unaligned_io(struct inode
*inode
, struct iov_iter
*from
, loff_t pos
)
187 struct super_block
*sb
= inode
->i_sb
;
188 unsigned long blockmask
= sb
->s_blocksize
- 1;
190 if ((pos
| iov_iter_alignment(from
)) & blockmask
)
197 ext4_extending_io(struct inode
*inode
, loff_t offset
, size_t len
)
199 if (offset
+ len
> i_size_read(inode
) ||
200 offset
+ len
> EXT4_I(inode
)->i_disksize
)
205 /* Is IO overwriting allocated and initialized blocks? */
206 static bool ext4_overwrite_io(struct inode
*inode
, loff_t pos
, loff_t len
)
208 struct ext4_map_blocks map
;
209 unsigned int blkbits
= inode
->i_blkbits
;
212 if (pos
+ len
> i_size_read(inode
))
215 map
.m_lblk
= pos
>> blkbits
;
216 map
.m_len
= EXT4_MAX_BLOCKS(len
, pos
, blkbits
);
219 err
= ext4_map_blocks(NULL
, inode
, &map
, 0);
221 * 'err==len' means that all of the blocks have been preallocated,
222 * regardless of whether they have been initialized or not. To exclude
223 * unwritten extents, we need to check m_flags.
225 return err
== blklen
&& (map
.m_flags
& EXT4_MAP_MAPPED
);
228 static ssize_t
ext4_generic_write_checks(struct kiocb
*iocb
,
229 struct iov_iter
*from
)
231 struct inode
*inode
= file_inode(iocb
->ki_filp
);
234 if (unlikely(IS_IMMUTABLE(inode
)))
237 ret
= generic_write_checks(iocb
, from
);
242 * If we have encountered a bitmap-format file, the size limit
243 * is smaller than s_maxbytes, which is for extent-mapped files.
245 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
))) {
246 struct ext4_sb_info
*sbi
= EXT4_SB(inode
->i_sb
);
248 if (iocb
->ki_pos
>= sbi
->s_bitmap_maxbytes
)
250 iov_iter_truncate(from
, sbi
->s_bitmap_maxbytes
- iocb
->ki_pos
);
253 return iov_iter_count(from
);
256 static ssize_t
ext4_write_checks(struct kiocb
*iocb
, struct iov_iter
*from
)
260 count
= ext4_generic_write_checks(iocb
, from
);
264 ret
= file_modified(iocb
->ki_filp
);
270 static ssize_t
ext4_buffered_write_iter(struct kiocb
*iocb
,
271 struct iov_iter
*from
)
274 struct inode
*inode
= file_inode(iocb
->ki_filp
);
276 if (iocb
->ki_flags
& IOCB_NOWAIT
)
280 ret
= ext4_write_checks(iocb
, from
);
284 current
->backing_dev_info
= inode_to_bdi(inode
);
285 ret
= generic_perform_write(iocb
, from
);
286 current
->backing_dev_info
= NULL
;
290 if (likely(ret
> 0)) {
292 ret
= generic_write_sync(iocb
, ret
);
298 static ssize_t
ext4_handle_inode_extension(struct inode
*inode
, loff_t offset
,
299 ssize_t written
, size_t count
)
302 bool truncate
= false;
303 u8 blkbits
= inode
->i_blkbits
;
304 ext4_lblk_t written_blk
, end_blk
;
308 * Note that EXT4_I(inode)->i_disksize can get extended up to
309 * inode->i_size while the I/O was running due to writeback of delalloc
310 * blocks. But, the code in ext4_iomap_alloc() is careful to use
311 * zeroed/unwritten extents if this is possible; thus we won't leave
312 * uninitialized blocks in a file even if we didn't succeed in writing
313 * as much as we intended.
315 WARN_ON_ONCE(i_size_read(inode
) < EXT4_I(inode
)->i_disksize
);
316 if (offset
+ count
<= EXT4_I(inode
)->i_disksize
) {
318 * We need to ensure that the inode is removed from the orphan
319 * list if it has been added prematurely, due to writeback of
322 if (!list_empty(&EXT4_I(inode
)->i_orphan
) && inode
->i_nlink
) {
323 handle
= ext4_journal_start(inode
, EXT4_HT_INODE
, 2);
325 if (IS_ERR(handle
)) {
326 ext4_orphan_del(NULL
, inode
);
327 return PTR_ERR(handle
);
330 ext4_orphan_del(handle
, inode
);
331 ext4_journal_stop(handle
);
340 handle
= ext4_journal_start(inode
, EXT4_HT_INODE
, 2);
341 if (IS_ERR(handle
)) {
342 written
= PTR_ERR(handle
);
346 if (ext4_update_inode_size(inode
, offset
+ written
)) {
347 ret
= ext4_mark_inode_dirty(handle
, inode
);
350 ext4_journal_stop(handle
);
356 * We may need to truncate allocated but not written blocks beyond EOF.
358 written_blk
= ALIGN(offset
+ written
, 1 << blkbits
);
359 end_blk
= ALIGN(offset
+ count
, 1 << blkbits
);
360 if (written_blk
< end_blk
&& ext4_can_truncate(inode
))
364 * Remove the inode from the orphan list if it has been extended and
365 * everything went OK.
367 if (!truncate
&& inode
->i_nlink
)
368 ext4_orphan_del(handle
, inode
);
369 ext4_journal_stop(handle
);
373 ext4_truncate_failed_write(inode
);
375 * If the truncate operation failed early, then the inode may
376 * still be on the orphan list. In that case, we need to try
377 * remove the inode from the in-memory linked list.
380 ext4_orphan_del(NULL
, inode
);
386 static int ext4_dio_write_end_io(struct kiocb
*iocb
, ssize_t size
,
387 int error
, unsigned int flags
)
389 loff_t pos
= iocb
->ki_pos
;
390 struct inode
*inode
= file_inode(iocb
->ki_filp
);
395 if (size
&& flags
& IOMAP_DIO_UNWRITTEN
) {
396 error
= ext4_convert_unwritten_extents(NULL
, inode
, pos
, size
);
401 * If we are extending the file, we have to update i_size here before
402 * page cache gets invalidated in iomap_dio_rw(). Otherwise racing
403 * buffered reads could zero out too much from page cache pages. Update
404 * of on-disk size will happen later in ext4_dio_write_iter() where
405 * we have enough information to also perform orphan list handling etc.
406 * Note that we perform all extending writes synchronously under
407 * i_rwsem held exclusively so i_size update is safe here in that case.
408 * If the write was not extending, we cannot see pos > i_size here
409 * because operations reducing i_size like truncate wait for all
410 * outstanding DIO before updating i_size.
413 if (pos
> i_size_read(inode
))
414 i_size_write(inode
, pos
);
419 static const struct iomap_dio_ops ext4_dio_write_ops
= {
420 .end_io
= ext4_dio_write_end_io
,
424 * The intention here is to start with shared lock acquired then see if any
425 * condition requires an exclusive inode lock. If yes, then we restart the
426 * whole operation by releasing the shared lock and acquiring exclusive lock.
428 * - For unaligned_io we never take shared lock as it may cause data corruption
429 * when two unaligned IO tries to modify the same block e.g. while zeroing.
431 * - For extending writes case we don't take the shared lock, since it requires
432 * updating inode i_disksize and/or orphan handling with exclusive lock.
434 * - shared locking will only be true mostly with overwrites. Otherwise we will
435 * switch to exclusive i_rwsem lock.
437 static ssize_t
ext4_dio_write_checks(struct kiocb
*iocb
, struct iov_iter
*from
,
438 bool *ilock_shared
, bool *extend
)
440 struct file
*file
= iocb
->ki_filp
;
441 struct inode
*inode
= file_inode(file
);
447 ret
= ext4_generic_write_checks(iocb
, from
);
451 offset
= iocb
->ki_pos
;
453 if (ext4_extending_io(inode
, offset
, count
))
456 * Determine whether the IO operation will overwrite allocated
457 * and initialized blocks.
458 * We need exclusive i_rwsem for changing security info
459 * in file_modified().
461 if (*ilock_shared
&& (!IS_NOSEC(inode
) || *extend
||
462 !ext4_overwrite_io(inode
, offset
, count
))) {
463 if (iocb
->ki_flags
& IOCB_NOWAIT
) {
467 inode_unlock_shared(inode
);
468 *ilock_shared
= false;
473 ret
= file_modified(file
);
480 inode_unlock_shared(inode
);
486 static ssize_t
ext4_dio_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
490 struct inode
*inode
= file_inode(iocb
->ki_filp
);
491 loff_t offset
= iocb
->ki_pos
;
492 size_t count
= iov_iter_count(from
);
493 const struct iomap_ops
*iomap_ops
= &ext4_iomap_ops
;
494 bool extend
= false, unaligned_io
= false;
495 bool ilock_shared
= true;
498 * We initially start with shared inode lock unless it is
499 * unaligned IO which needs exclusive lock anyways.
501 if (ext4_unaligned_io(inode
, from
, offset
)) {
503 ilock_shared
= false;
506 * Quick check here without any i_rwsem lock to see if it is extending
507 * IO. A more reliable check is done in ext4_dio_write_checks() with
508 * proper locking in place.
510 if (offset
+ count
> i_size_read(inode
))
511 ilock_shared
= false;
513 if (iocb
->ki_flags
& IOCB_NOWAIT
) {
515 if (!inode_trylock_shared(inode
))
518 if (!inode_trylock(inode
))
523 inode_lock_shared(inode
);
528 /* Fallback to buffered I/O if the inode does not support direct I/O. */
529 if (!ext4_should_use_dio(iocb
, from
)) {
531 inode_unlock_shared(inode
);
534 return ext4_buffered_write_iter(iocb
, from
);
537 ret
= ext4_dio_write_checks(iocb
, from
, &ilock_shared
, &extend
);
541 /* if we're going to block and IOCB_NOWAIT is set, return -EAGAIN */
542 if ((iocb
->ki_flags
& IOCB_NOWAIT
) && (unaligned_io
|| extend
)) {
547 * Make sure inline data cannot be created anymore since we are going
548 * to allocate blocks for DIO. We know the inode does not have any
549 * inline data now because ext4_dio_supported() checked for that.
551 ext4_clear_inode_state(inode
, EXT4_STATE_MAY_INLINE_DATA
);
553 offset
= iocb
->ki_pos
;
557 * Unaligned direct IO must be serialized among each other as zeroing
558 * of partial blocks of two competing unaligned IOs can result in data
561 * So we make sure we don't allow any unaligned IO in flight.
562 * For IOs where we need not wait (like unaligned non-AIO DIO),
563 * below inode_dio_wait() may anyway become a no-op, since we start
564 * with exclusive lock.
567 inode_dio_wait(inode
);
570 handle
= ext4_journal_start(inode
, EXT4_HT_INODE
, 2);
571 if (IS_ERR(handle
)) {
572 ret
= PTR_ERR(handle
);
576 ret
= ext4_orphan_add(handle
, inode
);
578 ext4_journal_stop(handle
);
582 ext4_journal_stop(handle
);
586 iomap_ops
= &ext4_iomap_overwrite_ops
;
587 ret
= iomap_dio_rw(iocb
, from
, iomap_ops
, &ext4_dio_write_ops
,
588 (unaligned_io
|| extend
) ? IOMAP_DIO_FORCE_WAIT
: 0,
594 ret
= ext4_handle_inode_extension(inode
, offset
, ret
, count
);
598 inode_unlock_shared(inode
);
602 if (ret
>= 0 && iov_iter_count(from
)) {
606 offset
= iocb
->ki_pos
;
607 err
= ext4_buffered_write_iter(iocb
, from
);
612 * We need to ensure that the pages within the page cache for
613 * the range covered by this I/O are written to disk and
614 * invalidated. This is in attempt to preserve the expected
615 * direct I/O semantics in the case we fallback to buffered I/O
616 * to complete off the I/O request.
619 endbyte
= offset
+ err
- 1;
620 err
= filemap_write_and_wait_range(iocb
->ki_filp
->f_mapping
,
623 invalidate_mapping_pages(iocb
->ki_filp
->f_mapping
,
624 offset
>> PAGE_SHIFT
,
625 endbyte
>> PAGE_SHIFT
);
633 ext4_dax_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
640 struct inode
*inode
= file_inode(iocb
->ki_filp
);
642 if (iocb
->ki_flags
& IOCB_NOWAIT
) {
643 if (!inode_trylock(inode
))
649 ret
= ext4_write_checks(iocb
, from
);
653 offset
= iocb
->ki_pos
;
654 count
= iov_iter_count(from
);
656 if (offset
+ count
> EXT4_I(inode
)->i_disksize
) {
657 handle
= ext4_journal_start(inode
, EXT4_HT_INODE
, 2);
658 if (IS_ERR(handle
)) {
659 ret
= PTR_ERR(handle
);
663 ret
= ext4_orphan_add(handle
, inode
);
665 ext4_journal_stop(handle
);
670 ext4_journal_stop(handle
);
673 ret
= dax_iomap_rw(iocb
, from
, &ext4_iomap_ops
);
676 ret
= ext4_handle_inode_extension(inode
, offset
, ret
, count
);
680 ret
= generic_write_sync(iocb
, ret
);
686 ext4_file_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
688 struct inode
*inode
= file_inode(iocb
->ki_filp
);
690 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
695 return ext4_dax_write_iter(iocb
, from
);
697 if (iocb
->ki_flags
& IOCB_DIRECT
)
698 return ext4_dio_write_iter(iocb
, from
);
700 return ext4_buffered_write_iter(iocb
, from
);
704 static vm_fault_t
ext4_dax_huge_fault(struct vm_fault
*vmf
,
705 enum page_entry_size pe_size
)
710 handle_t
*handle
= NULL
;
711 struct inode
*inode
= file_inode(vmf
->vma
->vm_file
);
712 struct super_block
*sb
= inode
->i_sb
;
715 * We have to distinguish real writes from writes which will result in a
716 * COW page; COW writes should *not* poke the journal (the file will not
717 * be changed). Doing so would cause unintended failures when mounted
720 * We check for VM_SHARED rather than vmf->cow_page since the latter is
721 * unset for pe_size != PE_SIZE_PTE (i.e. only in do_cow_fault); for
722 * other sizes, dax_iomap_fault will handle splitting / fallback so that
723 * we eventually come back with a COW page.
725 bool write
= (vmf
->flags
& FAULT_FLAG_WRITE
) &&
726 (vmf
->vma
->vm_flags
& VM_SHARED
);
727 struct address_space
*mapping
= vmf
->vma
->vm_file
->f_mapping
;
731 sb_start_pagefault(sb
);
732 file_update_time(vmf
->vma
->vm_file
);
733 filemap_invalidate_lock_shared(mapping
);
735 handle
= ext4_journal_start_sb(sb
, EXT4_HT_WRITE_PAGE
,
736 EXT4_DATA_TRANS_BLOCKS(sb
));
737 if (IS_ERR(handle
)) {
738 filemap_invalidate_unlock_shared(mapping
);
739 sb_end_pagefault(sb
);
740 return VM_FAULT_SIGBUS
;
743 filemap_invalidate_lock_shared(mapping
);
745 result
= dax_iomap_fault(vmf
, pe_size
, &pfn
, &error
, &ext4_iomap_ops
);
747 ext4_journal_stop(handle
);
749 if ((result
& VM_FAULT_ERROR
) && error
== -ENOSPC
&&
750 ext4_should_retry_alloc(sb
, &retries
))
752 /* Handling synchronous page fault? */
753 if (result
& VM_FAULT_NEEDDSYNC
)
754 result
= dax_finish_sync_fault(vmf
, pe_size
, pfn
);
755 filemap_invalidate_unlock_shared(mapping
);
756 sb_end_pagefault(sb
);
758 filemap_invalidate_unlock_shared(mapping
);
764 static vm_fault_t
ext4_dax_fault(struct vm_fault
*vmf
)
766 return ext4_dax_huge_fault(vmf
, PE_SIZE_PTE
);
769 static const struct vm_operations_struct ext4_dax_vm_ops
= {
770 .fault
= ext4_dax_fault
,
771 .huge_fault
= ext4_dax_huge_fault
,
772 .page_mkwrite
= ext4_dax_fault
,
773 .pfn_mkwrite
= ext4_dax_fault
,
776 #define ext4_dax_vm_ops ext4_file_vm_ops
779 static const struct vm_operations_struct ext4_file_vm_ops
= {
780 .fault
= filemap_fault
,
781 .map_pages
= filemap_map_pages
,
782 .page_mkwrite
= ext4_page_mkwrite
,
785 static int ext4_file_mmap(struct file
*file
, struct vm_area_struct
*vma
)
787 struct inode
*inode
= file
->f_mapping
->host
;
788 struct ext4_sb_info
*sbi
= EXT4_SB(inode
->i_sb
);
789 struct dax_device
*dax_dev
= sbi
->s_daxdev
;
791 if (unlikely(ext4_forced_shutdown(sbi
)))
795 * We don't support synchronous mappings for non-DAX files and
796 * for DAX files if underneath dax_device is not synchronous.
798 if (!daxdev_mapping_supported(vma
, dax_dev
))
802 if (IS_DAX(file_inode(file
))) {
803 vma
->vm_ops
= &ext4_dax_vm_ops
;
804 vma
->vm_flags
|= VM_HUGEPAGE
;
806 vma
->vm_ops
= &ext4_file_vm_ops
;
811 static int ext4_sample_last_mounted(struct super_block
*sb
,
812 struct vfsmount
*mnt
)
814 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
820 if (likely(ext4_test_mount_flag(sb
, EXT4_MF_MNTDIR_SAMPLED
)))
823 if (sb_rdonly(sb
) || !sb_start_intwrite_trylock(sb
))
826 ext4_set_mount_flag(sb
, EXT4_MF_MNTDIR_SAMPLED
);
828 * Sample where the filesystem has been mounted and
829 * store it in the superblock for sysadmin convenience
830 * when trying to sort through large numbers of block
831 * devices or filesystem images.
833 memset(buf
, 0, sizeof(buf
));
835 path
.dentry
= mnt
->mnt_root
;
836 cp
= d_path(&path
, buf
, sizeof(buf
));
841 handle
= ext4_journal_start_sb(sb
, EXT4_HT_MISC
, 1);
842 err
= PTR_ERR(handle
);
845 BUFFER_TRACE(sbi
->s_sbh
, "get_write_access");
846 err
= ext4_journal_get_write_access(handle
, sb
, sbi
->s_sbh
,
850 lock_buffer(sbi
->s_sbh
);
851 strncpy(sbi
->s_es
->s_last_mounted
, cp
,
852 sizeof(sbi
->s_es
->s_last_mounted
));
853 ext4_superblock_csum_set(sb
);
854 unlock_buffer(sbi
->s_sbh
);
855 ext4_handle_dirty_metadata(handle
, NULL
, sbi
->s_sbh
);
857 ext4_journal_stop(handle
);
863 static int ext4_file_open(struct inode
*inode
, struct file
*filp
)
867 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
870 ret
= ext4_sample_last_mounted(inode
->i_sb
, filp
->f_path
.mnt
);
874 ret
= fscrypt_file_open(inode
, filp
);
878 ret
= fsverity_file_open(inode
, filp
);
883 * Set up the jbd2_inode if we are opening the inode for
884 * writing and the journal is present
886 if (filp
->f_mode
& FMODE_WRITE
) {
887 ret
= ext4_inode_attach_jinode(inode
);
892 filp
->f_mode
|= FMODE_NOWAIT
| FMODE_BUF_RASYNC
;
893 return dquot_file_open(inode
, filp
);
897 * ext4_llseek() handles both block-mapped and extent-mapped maxbytes values
898 * by calling generic_file_llseek_size() with the appropriate maxbytes
901 loff_t
ext4_llseek(struct file
*file
, loff_t offset
, int whence
)
903 struct inode
*inode
= file
->f_mapping
->host
;
906 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
907 maxbytes
= EXT4_SB(inode
->i_sb
)->s_bitmap_maxbytes
;
909 maxbytes
= inode
->i_sb
->s_maxbytes
;
913 return generic_file_llseek_size(file
, offset
, whence
,
914 maxbytes
, i_size_read(inode
));
916 inode_lock_shared(inode
);
917 offset
= iomap_seek_hole(inode
, offset
,
918 &ext4_iomap_report_ops
);
919 inode_unlock_shared(inode
);
922 inode_lock_shared(inode
);
923 offset
= iomap_seek_data(inode
, offset
,
924 &ext4_iomap_report_ops
);
925 inode_unlock_shared(inode
);
931 return vfs_setpos(file
, offset
, maxbytes
);
934 const struct file_operations ext4_file_operations
= {
935 .llseek
= ext4_llseek
,
936 .read_iter
= ext4_file_read_iter
,
937 .write_iter
= ext4_file_write_iter
,
938 .iopoll
= iocb_bio_iopoll
,
939 .unlocked_ioctl
= ext4_ioctl
,
941 .compat_ioctl
= ext4_compat_ioctl
,
943 .mmap
= ext4_file_mmap
,
944 .mmap_supported_flags
= MAP_SYNC
,
945 .open
= ext4_file_open
,
946 .release
= ext4_release_file
,
947 .fsync
= ext4_sync_file
,
948 .get_unmapped_area
= thp_get_unmapped_area
,
949 .splice_read
= generic_file_splice_read
,
950 .splice_write
= iter_file_splice_write
,
951 .fallocate
= ext4_fallocate
,
954 const struct inode_operations ext4_file_inode_operations
= {
955 .setattr
= ext4_setattr
,
956 .getattr
= ext4_file_getattr
,
957 .listxattr
= ext4_listxattr
,
958 .get_inode_acl
= ext4_get_acl
,
959 .set_acl
= ext4_set_acl
,
960 .fiemap
= ext4_fiemap
,
961 .fileattr_get
= ext4_fileattr_get
,
962 .fileattr_set
= ext4_fileattr_set
,