1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/read_write.c
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/stat.h>
10 #include <linux/sched/xacct.h>
11 #include <linux/fcntl.h>
12 #include <linux/file.h>
13 #include <linux/uio.h>
14 #include <linux/fsnotify.h>
15 #include <linux/security.h>
16 #include <linux/export.h>
17 #include <linux/syscalls.h>
18 #include <linux/pagemap.h>
19 #include <linux/splice.h>
20 #include <linux/compat.h>
21 #include <linux/mount.h>
25 #include <linux/uaccess.h>
26 #include <asm/unistd.h>
28 const struct file_operations generic_ro_fops
= {
29 .llseek
= generic_file_llseek
,
30 .read_iter
= generic_file_read_iter
,
31 .mmap
= generic_file_readonly_mmap
,
32 .splice_read
= generic_file_splice_read
,
35 EXPORT_SYMBOL(generic_ro_fops
);
37 static inline bool unsigned_offsets(struct file
*file
)
39 return file
->f_mode
& FMODE_UNSIGNED_OFFSET
;
43 * vfs_setpos - update the file offset for lseek
44 * @file: file structure in question
45 * @offset: file offset to seek to
46 * @maxsize: maximum file size
48 * This is a low-level filesystem helper for updating the file offset to
49 * the value specified by @offset if the given offset is valid and it is
50 * not equal to the current file offset.
52 * Return the specified offset on success and -EINVAL on invalid offset.
54 loff_t
vfs_setpos(struct file
*file
, loff_t offset
, loff_t maxsize
)
56 if (offset
< 0 && !unsigned_offsets(file
))
61 if (offset
!= file
->f_pos
) {
67 EXPORT_SYMBOL(vfs_setpos
);
70 * generic_file_llseek_size - generic llseek implementation for regular files
71 * @file: file structure to seek on
72 * @offset: file offset to seek to
73 * @whence: type of seek
74 * @size: max size of this file in file system
75 * @eof: offset used for SEEK_END position
77 * This is a variant of generic_file_llseek that allows passing in a custom
78 * maximum file size and a custom EOF position, for e.g. hashed directories
81 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
82 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
83 * read/writes behave like SEEK_SET against seeks.
86 generic_file_llseek_size(struct file
*file
, loff_t offset
, int whence
,
87 loff_t maxsize
, loff_t eof
)
95 * Here we special-case the lseek(fd, 0, SEEK_CUR)
96 * position-querying operation. Avoid rewriting the "same"
97 * f_pos value back to the file because a concurrent read(),
98 * write() or lseek() might have altered it
103 * f_lock protects against read/modify/write race with other
104 * SEEK_CURs. Note that parallel writes and reads behave
107 spin_lock(&file
->f_lock
);
108 offset
= vfs_setpos(file
, file
->f_pos
+ offset
, maxsize
);
109 spin_unlock(&file
->f_lock
);
113 * In the generic case the entire file is data, so as long as
114 * offset isn't at the end of the file then the offset is data.
116 if ((unsigned long long)offset
>= eof
)
121 * There is a virtual hole at the end of the file, so as long as
122 * offset isn't i_size or larger, return i_size.
124 if ((unsigned long long)offset
>= eof
)
130 return vfs_setpos(file
, offset
, maxsize
);
132 EXPORT_SYMBOL(generic_file_llseek_size
);
135 * generic_file_llseek - generic llseek implementation for regular files
136 * @file: file structure to seek on
137 * @offset: file offset to seek to
138 * @whence: type of seek
140 * This is a generic implemenation of ->llseek useable for all normal local
141 * filesystems. It just updates the file offset to the value specified by
142 * @offset and @whence.
144 loff_t
generic_file_llseek(struct file
*file
, loff_t offset
, int whence
)
146 struct inode
*inode
= file
->f_mapping
->host
;
148 return generic_file_llseek_size(file
, offset
, whence
,
149 inode
->i_sb
->s_maxbytes
,
152 EXPORT_SYMBOL(generic_file_llseek
);
155 * fixed_size_llseek - llseek implementation for fixed-sized devices
156 * @file: file structure to seek on
157 * @offset: file offset to seek to
158 * @whence: type of seek
159 * @size: size of the file
162 loff_t
fixed_size_llseek(struct file
*file
, loff_t offset
, int whence
, loff_t size
)
165 case SEEK_SET
: case SEEK_CUR
: case SEEK_END
:
166 return generic_file_llseek_size(file
, offset
, whence
,
172 EXPORT_SYMBOL(fixed_size_llseek
);
175 * no_seek_end_llseek - llseek implementation for fixed-sized devices
176 * @file: file structure to seek on
177 * @offset: file offset to seek to
178 * @whence: type of seek
181 loff_t
no_seek_end_llseek(struct file
*file
, loff_t offset
, int whence
)
184 case SEEK_SET
: case SEEK_CUR
:
185 return generic_file_llseek_size(file
, offset
, whence
,
191 EXPORT_SYMBOL(no_seek_end_llseek
);
194 * no_seek_end_llseek_size - llseek implementation for fixed-sized devices
195 * @file: file structure to seek on
196 * @offset: file offset to seek to
197 * @whence: type of seek
198 * @size: maximal offset allowed
201 loff_t
no_seek_end_llseek_size(struct file
*file
, loff_t offset
, int whence
, loff_t size
)
204 case SEEK_SET
: case SEEK_CUR
:
205 return generic_file_llseek_size(file
, offset
, whence
,
211 EXPORT_SYMBOL(no_seek_end_llseek_size
);
214 * noop_llseek - No Operation Performed llseek implementation
215 * @file: file structure to seek on
216 * @offset: file offset to seek to
217 * @whence: type of seek
219 * This is an implementation of ->llseek useable for the rare special case when
220 * userspace expects the seek to succeed but the (device) file is actually not
221 * able to perform the seek. In this case you use noop_llseek() instead of
222 * falling back to the default implementation of ->llseek.
224 loff_t
noop_llseek(struct file
*file
, loff_t offset
, int whence
)
228 EXPORT_SYMBOL(noop_llseek
);
230 loff_t
no_llseek(struct file
*file
, loff_t offset
, int whence
)
234 EXPORT_SYMBOL(no_llseek
);
236 loff_t
default_llseek(struct file
*file
, loff_t offset
, int whence
)
238 struct inode
*inode
= file_inode(file
);
244 offset
+= i_size_read(inode
);
248 retval
= file
->f_pos
;
251 offset
+= file
->f_pos
;
255 * In the generic case the entire file is data, so as
256 * long as offset isn't at the end of the file then the
259 if (offset
>= inode
->i_size
) {
266 * There is a virtual hole at the end of the file, so
267 * as long as offset isn't i_size or larger, return
270 if (offset
>= inode
->i_size
) {
274 offset
= inode
->i_size
;
278 if (offset
>= 0 || unsigned_offsets(file
)) {
279 if (offset
!= file
->f_pos
) {
280 file
->f_pos
= offset
;
289 EXPORT_SYMBOL(default_llseek
);
291 loff_t
vfs_llseek(struct file
*file
, loff_t offset
, int whence
)
293 loff_t (*fn
)(struct file
*, loff_t
, int);
296 if (file
->f_mode
& FMODE_LSEEK
) {
297 if (file
->f_op
->llseek
)
298 fn
= file
->f_op
->llseek
;
300 return fn(file
, offset
, whence
);
302 EXPORT_SYMBOL(vfs_llseek
);
304 off_t
ksys_lseek(unsigned int fd
, off_t offset
, unsigned int whence
)
307 struct fd f
= fdget_pos(fd
);
312 if (whence
<= SEEK_MAX
) {
313 loff_t res
= vfs_llseek(f
.file
, offset
, whence
);
315 if (res
!= (loff_t
)retval
)
316 retval
= -EOVERFLOW
; /* LFS: should only happen on 32 bit platforms */
322 SYSCALL_DEFINE3(lseek
, unsigned int, fd
, off_t
, offset
, unsigned int, whence
)
324 return ksys_lseek(fd
, offset
, whence
);
328 COMPAT_SYSCALL_DEFINE3(lseek
, unsigned int, fd
, compat_off_t
, offset
, unsigned int, whence
)
330 return ksys_lseek(fd
, offset
, whence
);
334 #if !defined(CONFIG_64BIT) || defined(CONFIG_COMPAT)
335 SYSCALL_DEFINE5(llseek
, unsigned int, fd
, unsigned long, offset_high
,
336 unsigned long, offset_low
, loff_t __user
*, result
,
337 unsigned int, whence
)
340 struct fd f
= fdget_pos(fd
);
347 if (whence
> SEEK_MAX
)
350 offset
= vfs_llseek(f
.file
, ((loff_t
) offset_high
<< 32) | offset_low
,
353 retval
= (int)offset
;
356 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
365 int rw_verify_area(int read_write
, struct file
*file
, const loff_t
*ppos
, size_t count
)
369 int retval
= -EINVAL
;
371 inode
= file_inode(file
);
372 if (unlikely((ssize_t
) count
< 0))
375 if (unlikely(pos
< 0)) {
376 if (!unsigned_offsets(file
))
378 if (count
>= -pos
) /* both values are in 0..LLONG_MAX */
380 } else if (unlikely((loff_t
) (pos
+ count
) < 0)) {
381 if (!unsigned_offsets(file
))
385 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
386 retval
= locks_mandatory_area(inode
, file
, pos
, pos
+ count
- 1,
387 read_write
== READ
? F_RDLCK
: F_WRLCK
);
391 return security_file_permission(file
,
392 read_write
== READ
? MAY_READ
: MAY_WRITE
);
395 static ssize_t
new_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
397 struct iovec iov
= { .iov_base
= buf
, .iov_len
= len
};
399 struct iov_iter iter
;
402 init_sync_kiocb(&kiocb
, filp
);
403 kiocb
.ki_pos
= *ppos
;
404 iov_iter_init(&iter
, READ
, &iov
, 1, len
);
406 ret
= call_read_iter(filp
, &kiocb
, &iter
);
407 BUG_ON(ret
== -EIOCBQUEUED
);
408 *ppos
= kiocb
.ki_pos
;
412 ssize_t
__vfs_read(struct file
*file
, char __user
*buf
, size_t count
,
415 if (file
->f_op
->read
)
416 return file
->f_op
->read(file
, buf
, count
, pos
);
417 else if (file
->f_op
->read_iter
)
418 return new_sync_read(file
, buf
, count
, pos
);
423 ssize_t
kernel_read(struct file
*file
, void *buf
, size_t count
, loff_t
*pos
)
430 /* The cast to a user pointer is valid due to the set_fs() */
431 result
= vfs_read(file
, (void __user
*)buf
, count
, pos
);
435 EXPORT_SYMBOL(kernel_read
);
437 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
441 if (!(file
->f_mode
& FMODE_READ
))
443 if (!(file
->f_mode
& FMODE_CAN_READ
))
445 if (unlikely(!access_ok(buf
, count
)))
448 ret
= rw_verify_area(READ
, file
, pos
, count
);
450 if (count
> MAX_RW_COUNT
)
451 count
= MAX_RW_COUNT
;
452 ret
= __vfs_read(file
, buf
, count
, pos
);
454 fsnotify_access(file
);
455 add_rchar(current
, ret
);
463 static ssize_t
new_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
465 struct iovec iov
= { .iov_base
= (void __user
*)buf
, .iov_len
= len
};
467 struct iov_iter iter
;
470 init_sync_kiocb(&kiocb
, filp
);
471 kiocb
.ki_pos
= *ppos
;
472 iov_iter_init(&iter
, WRITE
, &iov
, 1, len
);
474 ret
= call_write_iter(filp
, &kiocb
, &iter
);
475 BUG_ON(ret
== -EIOCBQUEUED
);
477 *ppos
= kiocb
.ki_pos
;
481 static ssize_t
__vfs_write(struct file
*file
, const char __user
*p
,
482 size_t count
, loff_t
*pos
)
484 if (file
->f_op
->write
)
485 return file
->f_op
->write(file
, p
, count
, pos
);
486 else if (file
->f_op
->write_iter
)
487 return new_sync_write(file
, p
, count
, pos
);
492 ssize_t
__kernel_write(struct file
*file
, const void *buf
, size_t count
, loff_t
*pos
)
495 const char __user
*p
;
498 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
503 p
= (__force
const char __user
*)buf
;
504 if (count
> MAX_RW_COUNT
)
505 count
= MAX_RW_COUNT
;
506 ret
= __vfs_write(file
, p
, count
, pos
);
509 fsnotify_modify(file
);
510 add_wchar(current
, ret
);
515 EXPORT_SYMBOL(__kernel_write
);
517 ssize_t
kernel_write(struct file
*file
, const void *buf
, size_t count
,
525 /* The cast to a user pointer is valid due to the set_fs() */
526 res
= vfs_write(file
, (__force
const char __user
*)buf
, count
, pos
);
531 EXPORT_SYMBOL(kernel_write
);
533 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
537 if (!(file
->f_mode
& FMODE_WRITE
))
539 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
541 if (unlikely(!access_ok(buf
, count
)))
544 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
546 if (count
> MAX_RW_COUNT
)
547 count
= MAX_RW_COUNT
;
548 file_start_write(file
);
549 ret
= __vfs_write(file
, buf
, count
, pos
);
551 fsnotify_modify(file
);
552 add_wchar(current
, ret
);
555 file_end_write(file
);
561 static inline loff_t
file_pos_read(struct file
*file
)
563 return file
->f_mode
& FMODE_STREAM
? 0 : file
->f_pos
;
566 static inline void file_pos_write(struct file
*file
, loff_t pos
)
568 if ((file
->f_mode
& FMODE_STREAM
) == 0)
572 ssize_t
ksys_read(unsigned int fd
, char __user
*buf
, size_t count
)
574 struct fd f
= fdget_pos(fd
);
575 ssize_t ret
= -EBADF
;
578 loff_t pos
= file_pos_read(f
.file
);
579 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
581 file_pos_write(f
.file
, pos
);
587 SYSCALL_DEFINE3(read
, unsigned int, fd
, char __user
*, buf
, size_t, count
)
589 return ksys_read(fd
, buf
, count
);
592 ssize_t
ksys_write(unsigned int fd
, const char __user
*buf
, size_t count
)
594 struct fd f
= fdget_pos(fd
);
595 ssize_t ret
= -EBADF
;
598 loff_t pos
= file_pos_read(f
.file
);
599 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
601 file_pos_write(f
.file
, pos
);
608 SYSCALL_DEFINE3(write
, unsigned int, fd
, const char __user
*, buf
,
611 return ksys_write(fd
, buf
, count
);
614 ssize_t
ksys_pread64(unsigned int fd
, char __user
*buf
, size_t count
,
618 ssize_t ret
= -EBADF
;
626 if (f
.file
->f_mode
& FMODE_PREAD
)
627 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
634 SYSCALL_DEFINE4(pread64
, unsigned int, fd
, char __user
*, buf
,
635 size_t, count
, loff_t
, pos
)
637 return ksys_pread64(fd
, buf
, count
, pos
);
640 ssize_t
ksys_pwrite64(unsigned int fd
, const char __user
*buf
,
641 size_t count
, loff_t pos
)
644 ssize_t ret
= -EBADF
;
652 if (f
.file
->f_mode
& FMODE_PWRITE
)
653 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
660 SYSCALL_DEFINE4(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
661 size_t, count
, loff_t
, pos
)
663 return ksys_pwrite64(fd
, buf
, count
, pos
);
666 static ssize_t
do_iter_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
667 loff_t
*ppos
, int type
, rwf_t flags
)
672 init_sync_kiocb(&kiocb
, filp
);
673 ret
= kiocb_set_rw_flags(&kiocb
, flags
);
676 kiocb
.ki_pos
= *ppos
;
679 ret
= call_read_iter(filp
, &kiocb
, iter
);
681 ret
= call_write_iter(filp
, &kiocb
, iter
);
682 BUG_ON(ret
== -EIOCBQUEUED
);
683 *ppos
= kiocb
.ki_pos
;
687 /* Do it by hand, with file-ops */
688 static ssize_t
do_loop_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
689 loff_t
*ppos
, int type
, rwf_t flags
)
693 if (flags
& ~RWF_HIPRI
)
696 while (iov_iter_count(iter
)) {
697 struct iovec iovec
= iov_iter_iovec(iter
);
701 nr
= filp
->f_op
->read(filp
, iovec
.iov_base
,
702 iovec
.iov_len
, ppos
);
704 nr
= filp
->f_op
->write(filp
, iovec
.iov_base
,
705 iovec
.iov_len
, ppos
);
714 if (nr
!= iovec
.iov_len
)
716 iov_iter_advance(iter
, nr
);
723 * rw_copy_check_uvector() - Copy an array of &struct iovec from userspace
724 * into the kernel and check that it is valid.
726 * @type: One of %CHECK_IOVEC_ONLY, %READ, or %WRITE.
727 * @uvector: Pointer to the userspace array.
728 * @nr_segs: Number of elements in userspace array.
729 * @fast_segs: Number of elements in @fast_pointer.
730 * @fast_pointer: Pointer to (usually small on-stack) kernel array.
731 * @ret_pointer: (output parameter) Pointer to a variable that will point to
732 * either @fast_pointer, a newly allocated kernel array, or NULL,
733 * depending on which array was used.
735 * This function copies an array of &struct iovec of @nr_segs from
736 * userspace into the kernel and checks that each element is valid (e.g.
737 * it does not point to a kernel address or cause overflow by being too
740 * As an optimization, the caller may provide a pointer to a small
741 * on-stack array in @fast_pointer, typically %UIO_FASTIOV elements long
742 * (the size of this array, or 0 if unused, should be given in @fast_segs).
744 * @ret_pointer will always point to the array that was used, so the
745 * caller must take care not to call kfree() on it e.g. in case the
746 * @fast_pointer array was used and it was allocated on the stack.
748 * Return: The total number of bytes covered by the iovec array on success
749 * or a negative error code on error.
751 ssize_t
rw_copy_check_uvector(int type
, const struct iovec __user
* uvector
,
752 unsigned long nr_segs
, unsigned long fast_segs
,
753 struct iovec
*fast_pointer
,
754 struct iovec
**ret_pointer
)
758 struct iovec
*iov
= fast_pointer
;
761 * SuS says "The readv() function *may* fail if the iovcnt argument
762 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
763 * traditionally returned zero for zero segments, so...
771 * First get the "struct iovec" from user memory and
772 * verify all the pointers
774 if (nr_segs
> UIO_MAXIOV
) {
778 if (nr_segs
> fast_segs
) {
779 iov
= kmalloc_array(nr_segs
, sizeof(struct iovec
), GFP_KERNEL
);
785 if (copy_from_user(iov
, uvector
, nr_segs
*sizeof(*uvector
))) {
791 * According to the Single Unix Specification we should return EINVAL
792 * if an element length is < 0 when cast to ssize_t or if the
793 * total length would overflow the ssize_t return value of the
796 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
800 for (seg
= 0; seg
< nr_segs
; seg
++) {
801 void __user
*buf
= iov
[seg
].iov_base
;
802 ssize_t len
= (ssize_t
)iov
[seg
].iov_len
;
804 /* see if we we're about to use an invalid len or if
805 * it's about to overflow ssize_t */
811 && unlikely(!access_ok(buf
, len
))) {
815 if (len
> MAX_RW_COUNT
- ret
) {
816 len
= MAX_RW_COUNT
- ret
;
817 iov
[seg
].iov_len
= len
;
827 ssize_t
compat_rw_copy_check_uvector(int type
,
828 const struct compat_iovec __user
*uvector
, unsigned long nr_segs
,
829 unsigned long fast_segs
, struct iovec
*fast_pointer
,
830 struct iovec
**ret_pointer
)
832 compat_ssize_t tot_len
;
833 struct iovec
*iov
= *ret_pointer
= fast_pointer
;
838 * SuS says "The readv() function *may* fail if the iovcnt argument
839 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
840 * traditionally returned zero for zero segments, so...
846 if (nr_segs
> UIO_MAXIOV
)
848 if (nr_segs
> fast_segs
) {
850 iov
= kmalloc_array(nr_segs
, sizeof(struct iovec
), GFP_KERNEL
);
857 if (!access_ok(uvector
, nr_segs
*sizeof(*uvector
)))
861 * Single unix specification:
862 * We should -EINVAL if an element length is not >= 0 and fitting an
865 * In Linux, the total length is limited to MAX_RW_COUNT, there is
866 * no overflow possibility.
870 for (seg
= 0; seg
< nr_segs
; seg
++) {
874 if (__get_user(len
, &uvector
->iov_len
) ||
875 __get_user(buf
, &uvector
->iov_base
)) {
879 if (len
< 0) /* size_t not fitting in compat_ssize_t .. */
882 !access_ok(compat_ptr(buf
), len
)) {
886 if (len
> MAX_RW_COUNT
- tot_len
)
887 len
= MAX_RW_COUNT
- tot_len
;
889 iov
->iov_base
= compat_ptr(buf
);
890 iov
->iov_len
= (compat_size_t
) len
;
901 static ssize_t
do_iter_read(struct file
*file
, struct iov_iter
*iter
,
902 loff_t
*pos
, rwf_t flags
)
907 if (!(file
->f_mode
& FMODE_READ
))
909 if (!(file
->f_mode
& FMODE_CAN_READ
))
912 tot_len
= iov_iter_count(iter
);
915 ret
= rw_verify_area(READ
, file
, pos
, tot_len
);
919 if (file
->f_op
->read_iter
)
920 ret
= do_iter_readv_writev(file
, iter
, pos
, READ
, flags
);
922 ret
= do_loop_readv_writev(file
, iter
, pos
, READ
, flags
);
925 fsnotify_access(file
);
929 ssize_t
vfs_iter_read(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
932 if (!file
->f_op
->read_iter
)
934 return do_iter_read(file
, iter
, ppos
, flags
);
936 EXPORT_SYMBOL(vfs_iter_read
);
938 static ssize_t
do_iter_write(struct file
*file
, struct iov_iter
*iter
,
939 loff_t
*pos
, rwf_t flags
)
944 if (!(file
->f_mode
& FMODE_WRITE
))
946 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
949 tot_len
= iov_iter_count(iter
);
952 ret
= rw_verify_area(WRITE
, file
, pos
, tot_len
);
956 if (file
->f_op
->write_iter
)
957 ret
= do_iter_readv_writev(file
, iter
, pos
, WRITE
, flags
);
959 ret
= do_loop_readv_writev(file
, iter
, pos
, WRITE
, flags
);
961 fsnotify_modify(file
);
965 ssize_t
vfs_iter_write(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
968 if (!file
->f_op
->write_iter
)
970 return do_iter_write(file
, iter
, ppos
, flags
);
972 EXPORT_SYMBOL(vfs_iter_write
);
974 ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
975 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
977 struct iovec iovstack
[UIO_FASTIOV
];
978 struct iovec
*iov
= iovstack
;
979 struct iov_iter iter
;
982 ret
= import_iovec(READ
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
, &iter
);
984 ret
= do_iter_read(file
, &iter
, pos
, flags
);
991 static ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
992 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
994 struct iovec iovstack
[UIO_FASTIOV
];
995 struct iovec
*iov
= iovstack
;
996 struct iov_iter iter
;
999 ret
= import_iovec(WRITE
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
, &iter
);
1001 file_start_write(file
);
1002 ret
= do_iter_write(file
, &iter
, pos
, flags
);
1003 file_end_write(file
);
1009 static ssize_t
do_readv(unsigned long fd
, const struct iovec __user
*vec
,
1010 unsigned long vlen
, rwf_t flags
)
1012 struct fd f
= fdget_pos(fd
);
1013 ssize_t ret
= -EBADF
;
1016 loff_t pos
= file_pos_read(f
.file
);
1017 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1019 file_pos_write(f
.file
, pos
);
1024 add_rchar(current
, ret
);
1029 static ssize_t
do_writev(unsigned long fd
, const struct iovec __user
*vec
,
1030 unsigned long vlen
, rwf_t flags
)
1032 struct fd f
= fdget_pos(fd
);
1033 ssize_t ret
= -EBADF
;
1036 loff_t pos
= file_pos_read(f
.file
);
1037 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1039 file_pos_write(f
.file
, pos
);
1044 add_wchar(current
, ret
);
1049 static inline loff_t
pos_from_hilo(unsigned long high
, unsigned long low
)
1051 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
1052 return (((loff_t
)high
<< HALF_LONG_BITS
) << HALF_LONG_BITS
) | low
;
1055 static ssize_t
do_preadv(unsigned long fd
, const struct iovec __user
*vec
,
1056 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1059 ssize_t ret
= -EBADF
;
1067 if (f
.file
->f_mode
& FMODE_PREAD
)
1068 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1073 add_rchar(current
, ret
);
1078 static ssize_t
do_pwritev(unsigned long fd
, const struct iovec __user
*vec
,
1079 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1082 ssize_t ret
= -EBADF
;
1090 if (f
.file
->f_mode
& FMODE_PWRITE
)
1091 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1096 add_wchar(current
, ret
);
1101 SYSCALL_DEFINE3(readv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1102 unsigned long, vlen
)
1104 return do_readv(fd
, vec
, vlen
, 0);
1107 SYSCALL_DEFINE3(writev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1108 unsigned long, vlen
)
1110 return do_writev(fd
, vec
, vlen
, 0);
1113 SYSCALL_DEFINE5(preadv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1114 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1116 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1118 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1121 SYSCALL_DEFINE6(preadv2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1122 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1125 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1128 return do_readv(fd
, vec
, vlen
, flags
);
1130 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1133 SYSCALL_DEFINE5(pwritev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1134 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1136 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1138 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1141 SYSCALL_DEFINE6(pwritev2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1142 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1145 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1148 return do_writev(fd
, vec
, vlen
, flags
);
1150 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1153 #ifdef CONFIG_COMPAT
1154 static size_t compat_readv(struct file
*file
,
1155 const struct compat_iovec __user
*vec
,
1156 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
1158 struct iovec iovstack
[UIO_FASTIOV
];
1159 struct iovec
*iov
= iovstack
;
1160 struct iov_iter iter
;
1163 ret
= compat_import_iovec(READ
, vec
, vlen
, UIO_FASTIOV
, &iov
, &iter
);
1165 ret
= do_iter_read(file
, &iter
, pos
, flags
);
1169 add_rchar(current
, ret
);
1174 static size_t do_compat_readv(compat_ulong_t fd
,
1175 const struct compat_iovec __user
*vec
,
1176 compat_ulong_t vlen
, rwf_t flags
)
1178 struct fd f
= fdget_pos(fd
);
1184 pos
= f
.file
->f_pos
;
1185 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1187 f
.file
->f_pos
= pos
;
1193 COMPAT_SYSCALL_DEFINE3(readv
, compat_ulong_t
, fd
,
1194 const struct compat_iovec __user
*,vec
,
1195 compat_ulong_t
, vlen
)
1197 return do_compat_readv(fd
, vec
, vlen
, 0);
1200 static long do_compat_preadv64(unsigned long fd
,
1201 const struct compat_iovec __user
*vec
,
1202 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1213 if (f
.file
->f_mode
& FMODE_PREAD
)
1214 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1219 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1220 COMPAT_SYSCALL_DEFINE4(preadv64
, unsigned long, fd
,
1221 const struct compat_iovec __user
*,vec
,
1222 unsigned long, vlen
, loff_t
, pos
)
1224 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1228 COMPAT_SYSCALL_DEFINE5(preadv
, compat_ulong_t
, fd
,
1229 const struct compat_iovec __user
*,vec
,
1230 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1232 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1234 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1237 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1238 COMPAT_SYSCALL_DEFINE5(preadv64v2
, unsigned long, fd
,
1239 const struct compat_iovec __user
*,vec
,
1240 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1243 return do_compat_readv(fd
, vec
, vlen
, flags
);
1245 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1249 COMPAT_SYSCALL_DEFINE6(preadv2
, compat_ulong_t
, fd
,
1250 const struct compat_iovec __user
*,vec
,
1251 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
,
1254 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1257 return do_compat_readv(fd
, vec
, vlen
, flags
);
1259 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1262 static size_t compat_writev(struct file
*file
,
1263 const struct compat_iovec __user
*vec
,
1264 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
1266 struct iovec iovstack
[UIO_FASTIOV
];
1267 struct iovec
*iov
= iovstack
;
1268 struct iov_iter iter
;
1271 ret
= compat_import_iovec(WRITE
, vec
, vlen
, UIO_FASTIOV
, &iov
, &iter
);
1273 file_start_write(file
);
1274 ret
= do_iter_write(file
, &iter
, pos
, flags
);
1275 file_end_write(file
);
1279 add_wchar(current
, ret
);
1284 static size_t do_compat_writev(compat_ulong_t fd
,
1285 const struct compat_iovec __user
* vec
,
1286 compat_ulong_t vlen
, rwf_t flags
)
1288 struct fd f
= fdget_pos(fd
);
1294 pos
= f
.file
->f_pos
;
1295 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1297 f
.file
->f_pos
= pos
;
1302 COMPAT_SYSCALL_DEFINE3(writev
, compat_ulong_t
, fd
,
1303 const struct compat_iovec __user
*, vec
,
1304 compat_ulong_t
, vlen
)
1306 return do_compat_writev(fd
, vec
, vlen
, 0);
1309 static long do_compat_pwritev64(unsigned long fd
,
1310 const struct compat_iovec __user
*vec
,
1311 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1322 if (f
.file
->f_mode
& FMODE_PWRITE
)
1323 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1328 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1329 COMPAT_SYSCALL_DEFINE4(pwritev64
, unsigned long, fd
,
1330 const struct compat_iovec __user
*,vec
,
1331 unsigned long, vlen
, loff_t
, pos
)
1333 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1337 COMPAT_SYSCALL_DEFINE5(pwritev
, compat_ulong_t
, fd
,
1338 const struct compat_iovec __user
*,vec
,
1339 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1341 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1343 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1346 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1347 COMPAT_SYSCALL_DEFINE5(pwritev64v2
, unsigned long, fd
,
1348 const struct compat_iovec __user
*,vec
,
1349 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1352 return do_compat_writev(fd
, vec
, vlen
, flags
);
1354 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1358 COMPAT_SYSCALL_DEFINE6(pwritev2
, compat_ulong_t
, fd
,
1359 const struct compat_iovec __user
*,vec
,
1360 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
, rwf_t
, flags
)
1362 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1365 return do_compat_writev(fd
, vec
, vlen
, flags
);
1367 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1372 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
1373 size_t count
, loff_t max
)
1376 struct inode
*in_inode
, *out_inode
;
1383 * Get input file, and verify that it is ok..
1389 if (!(in
.file
->f_mode
& FMODE_READ
))
1393 pos
= in
.file
->f_pos
;
1396 if (!(in
.file
->f_mode
& FMODE_PREAD
))
1399 retval
= rw_verify_area(READ
, in
.file
, &pos
, count
);
1402 if (count
> MAX_RW_COUNT
)
1403 count
= MAX_RW_COUNT
;
1406 * Get output file, and verify that it is ok..
1409 out
= fdget(out_fd
);
1412 if (!(out
.file
->f_mode
& FMODE_WRITE
))
1414 in_inode
= file_inode(in
.file
);
1415 out_inode
= file_inode(out
.file
);
1416 out_pos
= out
.file
->f_pos
;
1417 retval
= rw_verify_area(WRITE
, out
.file
, &out_pos
, count
);
1422 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
1424 if (unlikely(pos
+ count
> max
)) {
1425 retval
= -EOVERFLOW
;
1434 * We need to debate whether we can enable this or not. The
1435 * man page documents EAGAIN return for the output at least,
1436 * and the application is arguably buggy if it doesn't expect
1437 * EAGAIN on a non-blocking file descriptor.
1439 if (in
.file
->f_flags
& O_NONBLOCK
)
1440 fl
= SPLICE_F_NONBLOCK
;
1442 file_start_write(out
.file
);
1443 retval
= do_splice_direct(in
.file
, &pos
, out
.file
, &out_pos
, count
, fl
);
1444 file_end_write(out
.file
);
1447 add_rchar(current
, retval
);
1448 add_wchar(current
, retval
);
1449 fsnotify_access(in
.file
);
1450 fsnotify_modify(out
.file
);
1451 out
.file
->f_pos
= out_pos
;
1455 in
.file
->f_pos
= pos
;
1461 retval
= -EOVERFLOW
;
1471 SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
, off_t __user
*, offset
, size_t, count
)
1478 if (unlikely(get_user(off
, offset
)))
1481 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1482 if (unlikely(put_user(pos
, offset
)))
1487 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1490 SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
, loff_t __user
*, offset
, size_t, count
)
1496 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1498 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1499 if (unlikely(put_user(pos
, offset
)))
1504 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1507 #ifdef CONFIG_COMPAT
1508 COMPAT_SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
,
1509 compat_off_t __user
*, offset
, compat_size_t
, count
)
1516 if (unlikely(get_user(off
, offset
)))
1519 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1520 if (unlikely(put_user(pos
, offset
)))
1525 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1528 COMPAT_SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
,
1529 compat_loff_t __user
*, offset
, compat_size_t
, count
)
1535 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1537 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1538 if (unlikely(put_user(pos
, offset
)))
1543 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1548 * copy_file_range() differs from regular file read and write in that it
1549 * specifically allows return partial success. When it does so is up to
1550 * the copy_file_range method.
1552 ssize_t
vfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
1553 struct file
*file_out
, loff_t pos_out
,
1554 size_t len
, unsigned int flags
)
1556 struct inode
*inode_in
= file_inode(file_in
);
1557 struct inode
*inode_out
= file_inode(file_out
);
1563 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1565 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1568 ret
= rw_verify_area(READ
, file_in
, &pos_in
, len
);
1572 ret
= rw_verify_area(WRITE
, file_out
, &pos_out
, len
);
1576 if (!(file_in
->f_mode
& FMODE_READ
) ||
1577 !(file_out
->f_mode
& FMODE_WRITE
) ||
1578 (file_out
->f_flags
& O_APPEND
))
1581 /* this could be relaxed once a method supports cross-fs copies */
1582 if (inode_in
->i_sb
!= inode_out
->i_sb
)
1588 file_start_write(file_out
);
1591 * Try cloning first, this is supported by more file systems, and
1592 * more efficient if both clone and copy are supported (e.g. NFS).
1594 if (file_in
->f_op
->remap_file_range
) {
1597 cloned
= file_in
->f_op
->remap_file_range(file_in
, pos_in
,
1599 min_t(loff_t
, MAX_RW_COUNT
, len
),
1600 REMAP_FILE_CAN_SHORTEN
);
1607 if (file_out
->f_op
->copy_file_range
) {
1608 ret
= file_out
->f_op
->copy_file_range(file_in
, pos_in
, file_out
,
1609 pos_out
, len
, flags
);
1610 if (ret
!= -EOPNOTSUPP
)
1614 ret
= do_splice_direct(file_in
, &pos_in
, file_out
, &pos_out
,
1615 len
> MAX_RW_COUNT
? MAX_RW_COUNT
: len
, 0);
1619 fsnotify_access(file_in
);
1620 add_rchar(current
, ret
);
1621 fsnotify_modify(file_out
);
1622 add_wchar(current
, ret
);
1628 file_end_write(file_out
);
1632 EXPORT_SYMBOL(vfs_copy_file_range
);
1634 SYSCALL_DEFINE6(copy_file_range
, int, fd_in
, loff_t __user
*, off_in
,
1635 int, fd_out
, loff_t __user
*, off_out
,
1636 size_t, len
, unsigned int, flags
)
1642 ssize_t ret
= -EBADF
;
1644 f_in
= fdget(fd_in
);
1648 f_out
= fdget(fd_out
);
1654 if (copy_from_user(&pos_in
, off_in
, sizeof(loff_t
)))
1657 pos_in
= f_in
.file
->f_pos
;
1661 if (copy_from_user(&pos_out
, off_out
, sizeof(loff_t
)))
1664 pos_out
= f_out
.file
->f_pos
;
1667 ret
= vfs_copy_file_range(f_in
.file
, pos_in
, f_out
.file
, pos_out
, len
,
1674 if (copy_to_user(off_in
, &pos_in
, sizeof(loff_t
)))
1677 f_in
.file
->f_pos
= pos_in
;
1681 if (copy_to_user(off_out
, &pos_out
, sizeof(loff_t
)))
1684 f_out
.file
->f_pos
= pos_out
;
1696 static int remap_verify_area(struct file
*file
, loff_t pos
, loff_t len
,
1699 struct inode
*inode
= file_inode(file
);
1701 if (unlikely(pos
< 0 || len
< 0))
1704 if (unlikely((loff_t
) (pos
+ len
) < 0))
1707 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
1708 loff_t end
= len
? pos
+ len
- 1 : OFFSET_MAX
;
1711 retval
= locks_mandatory_area(inode
, file
, pos
, end
,
1712 write
? F_WRLCK
: F_RDLCK
);
1717 return security_file_permission(file
, write
? MAY_WRITE
: MAY_READ
);
1720 * Ensure that we don't remap a partial EOF block in the middle of something
1721 * else. Assume that the offsets have already been checked for block
1724 * For deduplication we always scale down to the previous block because we
1725 * can't meaningfully compare post-EOF contents.
1727 * For clone we only link a partial EOF block above the destination file's EOF.
1729 * Shorten the request if possible.
1731 static int generic_remap_check_len(struct inode
*inode_in
,
1732 struct inode
*inode_out
,
1735 unsigned int remap_flags
)
1737 u64 blkmask
= i_blocksize(inode_in
) - 1;
1738 loff_t new_len
= *len
;
1740 if ((*len
& blkmask
) == 0)
1743 if ((remap_flags
& REMAP_FILE_DEDUP
) ||
1744 pos_out
+ *len
< i_size_read(inode_out
))
1745 new_len
&= ~blkmask
;
1747 if (new_len
== *len
)
1750 if (remap_flags
& REMAP_FILE_CAN_SHORTEN
) {
1755 return (remap_flags
& REMAP_FILE_DEDUP
) ? -EBADE
: -EINVAL
;
1759 * Read a page's worth of file data into the page cache. Return the page
1762 static struct page
*vfs_dedupe_get_page(struct inode
*inode
, loff_t offset
)
1766 page
= read_mapping_page(inode
->i_mapping
, offset
>> PAGE_SHIFT
, NULL
);
1769 if (!PageUptodate(page
)) {
1771 return ERR_PTR(-EIO
);
1778 * Compare extents of two files to see if they are the same.
1779 * Caller must have locked both inodes to prevent write races.
1781 static int vfs_dedupe_file_range_compare(struct inode
*src
, loff_t srcoff
,
1782 struct inode
*dest
, loff_t destoff
,
1783 loff_t len
, bool *is_same
)
1789 struct page
*src_page
;
1790 struct page
*dest_page
;
1798 src_poff
= srcoff
& (PAGE_SIZE
- 1);
1799 dest_poff
= destoff
& (PAGE_SIZE
- 1);
1800 cmp_len
= min(PAGE_SIZE
- src_poff
,
1801 PAGE_SIZE
- dest_poff
);
1802 cmp_len
= min(cmp_len
, len
);
1806 src_page
= vfs_dedupe_get_page(src
, srcoff
);
1807 if (IS_ERR(src_page
)) {
1808 error
= PTR_ERR(src_page
);
1811 dest_page
= vfs_dedupe_get_page(dest
, destoff
);
1812 if (IS_ERR(dest_page
)) {
1813 error
= PTR_ERR(dest_page
);
1814 unlock_page(src_page
);
1818 src_addr
= kmap_atomic(src_page
);
1819 dest_addr
= kmap_atomic(dest_page
);
1821 flush_dcache_page(src_page
);
1822 flush_dcache_page(dest_page
);
1824 if (memcmp(src_addr
+ src_poff
, dest_addr
+ dest_poff
, cmp_len
))
1827 kunmap_atomic(dest_addr
);
1828 kunmap_atomic(src_addr
);
1829 unlock_page(dest_page
);
1830 unlock_page(src_page
);
1831 put_page(dest_page
);
1850 * Check that the two inodes are eligible for cloning, the ranges make
1851 * sense, and then flush all dirty data. Caller must ensure that the
1852 * inodes have been locked against any other modifications.
1854 * If there's an error, then the usual negative error code is returned.
1855 * Otherwise returns 0 with *len set to the request length.
1857 int generic_remap_file_range_prep(struct file
*file_in
, loff_t pos_in
,
1858 struct file
*file_out
, loff_t pos_out
,
1859 loff_t
*len
, unsigned int remap_flags
)
1861 struct inode
*inode_in
= file_inode(file_in
);
1862 struct inode
*inode_out
= file_inode(file_out
);
1863 bool same_inode
= (inode_in
== inode_out
);
1866 /* Don't touch certain kinds of inodes */
1867 if (IS_IMMUTABLE(inode_out
))
1870 if (IS_SWAPFILE(inode_in
) || IS_SWAPFILE(inode_out
))
1873 /* Don't reflink dirs, pipes, sockets... */
1874 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1876 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1879 /* Zero length dedupe exits immediately; reflink goes to EOF. */
1881 loff_t isize
= i_size_read(inode_in
);
1883 if ((remap_flags
& REMAP_FILE_DEDUP
) || pos_in
== isize
)
1887 *len
= isize
- pos_in
;
1892 /* Check that we don't violate system file offset limits. */
1893 ret
= generic_remap_checks(file_in
, pos_in
, file_out
, pos_out
, len
,
1898 /* Wait for the completion of any pending IOs on both files */
1899 inode_dio_wait(inode_in
);
1901 inode_dio_wait(inode_out
);
1903 ret
= filemap_write_and_wait_range(inode_in
->i_mapping
,
1904 pos_in
, pos_in
+ *len
- 1);
1908 ret
= filemap_write_and_wait_range(inode_out
->i_mapping
,
1909 pos_out
, pos_out
+ *len
- 1);
1914 * Check that the extents are the same.
1916 if (remap_flags
& REMAP_FILE_DEDUP
) {
1917 bool is_same
= false;
1919 ret
= vfs_dedupe_file_range_compare(inode_in
, pos_in
,
1920 inode_out
, pos_out
, *len
, &is_same
);
1927 ret
= generic_remap_check_len(inode_in
, inode_out
, pos_out
, len
,
1932 /* If can't alter the file contents, we're done. */
1933 if (!(remap_flags
& REMAP_FILE_DEDUP
)) {
1934 /* Update the timestamps, since we can alter file contents. */
1935 if (!(file_out
->f_mode
& FMODE_NOCMTIME
)) {
1936 ret
= file_update_time(file_out
);
1942 * Clear the security bits if the process is not being run by
1943 * root. This keeps people from modifying setuid and setgid
1946 ret
= file_remove_privs(file_out
);
1953 EXPORT_SYMBOL(generic_remap_file_range_prep
);
1955 loff_t
do_clone_file_range(struct file
*file_in
, loff_t pos_in
,
1956 struct file
*file_out
, loff_t pos_out
,
1957 loff_t len
, unsigned int remap_flags
)
1959 struct inode
*inode_in
= file_inode(file_in
);
1960 struct inode
*inode_out
= file_inode(file_out
);
1963 WARN_ON_ONCE(remap_flags
& REMAP_FILE_DEDUP
);
1965 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1967 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1971 * FICLONE/FICLONERANGE ioctls enforce that src and dest files are on
1972 * the same mount. Practically, they only need to be on the same file
1975 if (inode_in
->i_sb
!= inode_out
->i_sb
)
1978 if (!(file_in
->f_mode
& FMODE_READ
) ||
1979 !(file_out
->f_mode
& FMODE_WRITE
) ||
1980 (file_out
->f_flags
& O_APPEND
))
1983 if (!file_in
->f_op
->remap_file_range
)
1986 ret
= remap_verify_area(file_in
, pos_in
, len
, false);
1990 ret
= remap_verify_area(file_out
, pos_out
, len
, true);
1994 ret
= file_in
->f_op
->remap_file_range(file_in
, pos_in
,
1995 file_out
, pos_out
, len
, remap_flags
);
1999 fsnotify_access(file_in
);
2000 fsnotify_modify(file_out
);
2003 EXPORT_SYMBOL(do_clone_file_range
);
2005 loff_t
vfs_clone_file_range(struct file
*file_in
, loff_t pos_in
,
2006 struct file
*file_out
, loff_t pos_out
,
2007 loff_t len
, unsigned int remap_flags
)
2011 file_start_write(file_out
);
2012 ret
= do_clone_file_range(file_in
, pos_in
, file_out
, pos_out
, len
,
2014 file_end_write(file_out
);
2018 EXPORT_SYMBOL(vfs_clone_file_range
);
2020 /* Check whether we are allowed to dedupe the destination file */
2021 static bool allow_file_dedupe(struct file
*file
)
2023 if (capable(CAP_SYS_ADMIN
))
2025 if (file
->f_mode
& FMODE_WRITE
)
2027 if (uid_eq(current_fsuid(), file_inode(file
)->i_uid
))
2029 if (!inode_permission(file_inode(file
), MAY_WRITE
))
2034 loff_t
vfs_dedupe_file_range_one(struct file
*src_file
, loff_t src_pos
,
2035 struct file
*dst_file
, loff_t dst_pos
,
2036 loff_t len
, unsigned int remap_flags
)
2040 WARN_ON_ONCE(remap_flags
& ~(REMAP_FILE_DEDUP
|
2041 REMAP_FILE_CAN_SHORTEN
));
2043 ret
= mnt_want_write_file(dst_file
);
2047 ret
= remap_verify_area(dst_file
, dst_pos
, len
, true);
2049 goto out_drop_write
;
2052 if (!allow_file_dedupe(dst_file
))
2053 goto out_drop_write
;
2056 if (src_file
->f_path
.mnt
!= dst_file
->f_path
.mnt
)
2057 goto out_drop_write
;
2060 if (S_ISDIR(file_inode(dst_file
)->i_mode
))
2061 goto out_drop_write
;
2064 if (!dst_file
->f_op
->remap_file_range
)
2065 goto out_drop_write
;
2069 goto out_drop_write
;
2072 ret
= dst_file
->f_op
->remap_file_range(src_file
, src_pos
, dst_file
,
2073 dst_pos
, len
, remap_flags
| REMAP_FILE_DEDUP
);
2075 mnt_drop_write_file(dst_file
);
2079 EXPORT_SYMBOL(vfs_dedupe_file_range_one
);
2081 int vfs_dedupe_file_range(struct file
*file
, struct file_dedupe_range
*same
)
2083 struct file_dedupe_range_info
*info
;
2084 struct inode
*src
= file_inode(file
);
2089 u16 count
= same
->dest_count
;
2092 if (!(file
->f_mode
& FMODE_READ
))
2095 if (same
->reserved1
|| same
->reserved2
)
2098 off
= same
->src_offset
;
2099 len
= same
->src_length
;
2101 if (S_ISDIR(src
->i_mode
))
2104 if (!S_ISREG(src
->i_mode
))
2107 if (!file
->f_op
->remap_file_range
)
2110 ret
= remap_verify_area(file
, off
, len
, false);
2115 if (off
+ len
> i_size_read(src
))
2118 /* Arbitrary 1G limit on a single dedupe request, can be raised. */
2119 len
= min_t(u64
, len
, 1 << 30);
2121 /* pre-format output fields to sane values */
2122 for (i
= 0; i
< count
; i
++) {
2123 same
->info
[i
].bytes_deduped
= 0ULL;
2124 same
->info
[i
].status
= FILE_DEDUPE_RANGE_SAME
;
2127 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
2128 struct fd dst_fd
= fdget(info
->dest_fd
);
2129 struct file
*dst_file
= dst_fd
.file
;
2132 info
->status
= -EBADF
;
2136 if (info
->reserved
) {
2137 info
->status
= -EINVAL
;
2141 deduped
= vfs_dedupe_file_range_one(file
, off
, dst_file
,
2142 info
->dest_offset
, len
,
2143 REMAP_FILE_CAN_SHORTEN
);
2144 if (deduped
== -EBADE
)
2145 info
->status
= FILE_DEDUPE_RANGE_DIFFERS
;
2146 else if (deduped
< 0)
2147 info
->status
= deduped
;
2149 info
->bytes_deduped
= len
;
2154 if (fatal_signal_pending(current
))
2159 EXPORT_SYMBOL(vfs_dedupe_file_range
);