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 SYSCALL_DEFINE3(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 */
323 COMPAT_SYSCALL_DEFINE3(lseek
, unsigned int, fd
, compat_off_t
, offset
, unsigned int, whence
)
325 return sys_lseek(fd
, offset
, whence
);
329 #ifdef __ARCH_WANT_SYS_LLSEEK
330 SYSCALL_DEFINE5(llseek
, unsigned int, fd
, unsigned long, offset_high
,
331 unsigned long, offset_low
, loff_t __user
*, result
,
332 unsigned int, whence
)
335 struct fd f
= fdget_pos(fd
);
342 if (whence
> SEEK_MAX
)
345 offset
= vfs_llseek(f
.file
, ((loff_t
) offset_high
<< 32) | offset_low
,
348 retval
= (int)offset
;
351 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
360 int rw_verify_area(int read_write
, struct file
*file
, const loff_t
*ppos
, size_t count
)
364 int retval
= -EINVAL
;
366 inode
= file_inode(file
);
367 if (unlikely((ssize_t
) count
< 0))
370 if (unlikely(pos
< 0)) {
371 if (!unsigned_offsets(file
))
373 if (count
>= -pos
) /* both values are in 0..LLONG_MAX */
375 } else if (unlikely((loff_t
) (pos
+ count
) < 0)) {
376 if (!unsigned_offsets(file
))
380 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
381 retval
= locks_mandatory_area(inode
, file
, pos
, pos
+ count
- 1,
382 read_write
== READ
? F_RDLCK
: F_WRLCK
);
386 return security_file_permission(file
,
387 read_write
== READ
? MAY_READ
: MAY_WRITE
);
390 static ssize_t
new_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
392 struct iovec iov
= { .iov_base
= buf
, .iov_len
= len
};
394 struct iov_iter iter
;
397 init_sync_kiocb(&kiocb
, filp
);
398 kiocb
.ki_pos
= *ppos
;
399 iov_iter_init(&iter
, READ
, &iov
, 1, len
);
401 ret
= call_read_iter(filp
, &kiocb
, &iter
);
402 BUG_ON(ret
== -EIOCBQUEUED
);
403 *ppos
= kiocb
.ki_pos
;
407 ssize_t
__vfs_read(struct file
*file
, char __user
*buf
, size_t count
,
410 if (file
->f_op
->read
)
411 return file
->f_op
->read(file
, buf
, count
, pos
);
412 else if (file
->f_op
->read_iter
)
413 return new_sync_read(file
, buf
, count
, pos
);
418 ssize_t
kernel_read(struct file
*file
, void *buf
, size_t count
, loff_t
*pos
)
425 /* The cast to a user pointer is valid due to the set_fs() */
426 result
= vfs_read(file
, (void __user
*)buf
, count
, pos
);
430 EXPORT_SYMBOL(kernel_read
);
432 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
436 if (!(file
->f_mode
& FMODE_READ
))
438 if (!(file
->f_mode
& FMODE_CAN_READ
))
440 if (unlikely(!access_ok(VERIFY_WRITE
, buf
, count
)))
443 ret
= rw_verify_area(READ
, file
, pos
, count
);
445 if (count
> MAX_RW_COUNT
)
446 count
= MAX_RW_COUNT
;
447 ret
= __vfs_read(file
, buf
, count
, pos
);
449 fsnotify_access(file
);
450 add_rchar(current
, ret
);
457 EXPORT_SYMBOL_GPL(vfs_read
);
459 static ssize_t
new_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
461 struct iovec iov
= { .iov_base
= (void __user
*)buf
, .iov_len
= len
};
463 struct iov_iter iter
;
466 init_sync_kiocb(&kiocb
, filp
);
467 kiocb
.ki_pos
= *ppos
;
468 iov_iter_init(&iter
, WRITE
, &iov
, 1, len
);
470 ret
= call_write_iter(filp
, &kiocb
, &iter
);
471 BUG_ON(ret
== -EIOCBQUEUED
);
473 *ppos
= kiocb
.ki_pos
;
477 ssize_t
__vfs_write(struct file
*file
, const char __user
*p
, size_t count
,
480 if (file
->f_op
->write
)
481 return file
->f_op
->write(file
, p
, count
, pos
);
482 else if (file
->f_op
->write_iter
)
483 return new_sync_write(file
, p
, count
, pos
);
488 vfs_readf_t
vfs_readf(struct file
*file
)
490 const struct file_operations
*fop
= file
->f_op
;
495 return new_sync_read
;
496 return ERR_PTR(-ENOSYS
);
498 EXPORT_SYMBOL_GPL(vfs_readf
);
500 vfs_writef_t
vfs_writef(struct file
*file
)
502 const struct file_operations
*fop
= file
->f_op
;
507 return new_sync_write
;
508 return ERR_PTR(-ENOSYS
);
510 EXPORT_SYMBOL_GPL(vfs_writef
);
512 ssize_t
__kernel_write(struct file
*file
, const void *buf
, size_t count
, loff_t
*pos
)
515 const char __user
*p
;
518 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
523 p
= (__force
const char __user
*)buf
;
524 if (count
> MAX_RW_COUNT
)
525 count
= MAX_RW_COUNT
;
526 ret
= __vfs_write(file
, p
, count
, pos
);
529 fsnotify_modify(file
);
530 add_wchar(current
, ret
);
535 EXPORT_SYMBOL(__kernel_write
);
537 ssize_t
kernel_write(struct file
*file
, const void *buf
, size_t count
,
545 /* The cast to a user pointer is valid due to the set_fs() */
546 res
= vfs_write(file
, (__force
const char __user
*)buf
, count
, pos
);
551 EXPORT_SYMBOL(kernel_write
);
553 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
557 if (!(file
->f_mode
& FMODE_WRITE
))
559 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
561 if (unlikely(!access_ok(VERIFY_READ
, buf
, count
)))
564 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
566 if (count
> MAX_RW_COUNT
)
567 count
= MAX_RW_COUNT
;
568 file_start_write(file
);
569 ret
= __vfs_write(file
, buf
, count
, pos
);
571 fsnotify_modify(file
);
572 add_wchar(current
, ret
);
575 file_end_write(file
);
580 EXPORT_SYMBOL_GPL(vfs_write
);
582 static inline loff_t
file_pos_read(struct file
*file
)
587 static inline void file_pos_write(struct file
*file
, loff_t pos
)
592 SYSCALL_DEFINE3(read
, unsigned int, fd
, 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_read(f
.file
, buf
, count
, &pos
);
601 file_pos_write(f
.file
, pos
);
607 SYSCALL_DEFINE3(write
, unsigned int, fd
, const char __user
*, buf
,
610 struct fd f
= fdget_pos(fd
);
611 ssize_t ret
= -EBADF
;
614 loff_t pos
= file_pos_read(f
.file
);
615 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
617 file_pos_write(f
.file
, pos
);
624 SYSCALL_DEFINE4(pread64
, unsigned int, fd
, char __user
*, buf
,
625 size_t, count
, loff_t
, pos
)
628 ssize_t ret
= -EBADF
;
636 if (f
.file
->f_mode
& FMODE_PREAD
)
637 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
644 SYSCALL_DEFINE4(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
645 size_t, count
, loff_t
, pos
)
648 ssize_t ret
= -EBADF
;
656 if (f
.file
->f_mode
& FMODE_PWRITE
)
657 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
664 static ssize_t
do_iter_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
665 loff_t
*ppos
, int type
, rwf_t flags
)
670 init_sync_kiocb(&kiocb
, filp
);
671 ret
= kiocb_set_rw_flags(&kiocb
, flags
);
674 kiocb
.ki_pos
= *ppos
;
677 ret
= call_read_iter(filp
, &kiocb
, iter
);
679 ret
= call_write_iter(filp
, &kiocb
, iter
);
680 BUG_ON(ret
== -EIOCBQUEUED
);
681 *ppos
= kiocb
.ki_pos
;
685 /* Do it by hand, with file-ops */
686 static ssize_t
do_loop_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
687 loff_t
*ppos
, int type
, rwf_t flags
)
691 if (flags
& ~RWF_HIPRI
)
694 while (iov_iter_count(iter
)) {
695 struct iovec iovec
= iov_iter_iovec(iter
);
699 nr
= filp
->f_op
->read(filp
, iovec
.iov_base
,
700 iovec
.iov_len
, ppos
);
702 nr
= filp
->f_op
->write(filp
, iovec
.iov_base
,
703 iovec
.iov_len
, ppos
);
712 if (nr
!= iovec
.iov_len
)
714 iov_iter_advance(iter
, nr
);
720 /* A write operation does a read from user space and vice versa */
721 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
724 * rw_copy_check_uvector() - Copy an array of &struct iovec from userspace
725 * into the kernel and check that it is valid.
727 * @type: One of %CHECK_IOVEC_ONLY, %READ, or %WRITE.
728 * @uvector: Pointer to the userspace array.
729 * @nr_segs: Number of elements in userspace array.
730 * @fast_segs: Number of elements in @fast_pointer.
731 * @fast_pointer: Pointer to (usually small on-stack) kernel array.
732 * @ret_pointer: (output parameter) Pointer to a variable that will point to
733 * either @fast_pointer, a newly allocated kernel array, or NULL,
734 * depending on which array was used.
736 * This function copies an array of &struct iovec of @nr_segs from
737 * userspace into the kernel and checks that each element is valid (e.g.
738 * it does not point to a kernel address or cause overflow by being too
741 * As an optimization, the caller may provide a pointer to a small
742 * on-stack array in @fast_pointer, typically %UIO_FASTIOV elements long
743 * (the size of this array, or 0 if unused, should be given in @fast_segs).
745 * @ret_pointer will always point to the array that was used, so the
746 * caller must take care not to call kfree() on it e.g. in case the
747 * @fast_pointer array was used and it was allocated on the stack.
749 * Return: The total number of bytes covered by the iovec array on success
750 * or a negative error code on error.
752 ssize_t
rw_copy_check_uvector(int type
, const struct iovec __user
* uvector
,
753 unsigned long nr_segs
, unsigned long fast_segs
,
754 struct iovec
*fast_pointer
,
755 struct iovec
**ret_pointer
)
759 struct iovec
*iov
= fast_pointer
;
762 * SuS says "The readv() function *may* fail if the iovcnt argument
763 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
764 * traditionally returned zero for zero segments, so...
772 * First get the "struct iovec" from user memory and
773 * verify all the pointers
775 if (nr_segs
> UIO_MAXIOV
) {
779 if (nr_segs
> fast_segs
) {
780 iov
= kmalloc(nr_segs
*sizeof(struct iovec
), GFP_KERNEL
);
786 if (copy_from_user(iov
, uvector
, nr_segs
*sizeof(*uvector
))) {
792 * According to the Single Unix Specification we should return EINVAL
793 * if an element length is < 0 when cast to ssize_t or if the
794 * total length would overflow the ssize_t return value of the
797 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
801 for (seg
= 0; seg
< nr_segs
; seg
++) {
802 void __user
*buf
= iov
[seg
].iov_base
;
803 ssize_t len
= (ssize_t
)iov
[seg
].iov_len
;
805 /* see if we we're about to use an invalid len or if
806 * it's about to overflow ssize_t */
812 && unlikely(!access_ok(vrfy_dir(type
), buf
, len
))) {
816 if (len
> MAX_RW_COUNT
- ret
) {
817 len
= MAX_RW_COUNT
- ret
;
818 iov
[seg
].iov_len
= len
;
828 ssize_t
compat_rw_copy_check_uvector(int type
,
829 const struct compat_iovec __user
*uvector
, unsigned long nr_segs
,
830 unsigned long fast_segs
, struct iovec
*fast_pointer
,
831 struct iovec
**ret_pointer
)
833 compat_ssize_t tot_len
;
834 struct iovec
*iov
= *ret_pointer
= fast_pointer
;
839 * SuS says "The readv() function *may* fail if the iovcnt argument
840 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
841 * traditionally returned zero for zero segments, so...
847 if (nr_segs
> UIO_MAXIOV
)
849 if (nr_segs
> fast_segs
) {
851 iov
= kmalloc(nr_segs
*sizeof(struct iovec
), GFP_KERNEL
);
858 if (!access_ok(VERIFY_READ
, uvector
, nr_segs
*sizeof(*uvector
)))
862 * Single unix specification:
863 * We should -EINVAL if an element length is not >= 0 and fitting an
866 * In Linux, the total length is limited to MAX_RW_COUNT, there is
867 * no overflow possibility.
871 for (seg
= 0; seg
< nr_segs
; seg
++) {
875 if (__get_user(len
, &uvector
->iov_len
) ||
876 __get_user(buf
, &uvector
->iov_base
)) {
880 if (len
< 0) /* size_t not fitting in compat_ssize_t .. */
883 !access_ok(vrfy_dir(type
), compat_ptr(buf
), len
)) {
887 if (len
> MAX_RW_COUNT
- tot_len
)
888 len
= MAX_RW_COUNT
- tot_len
;
890 iov
->iov_base
= compat_ptr(buf
);
891 iov
->iov_len
= (compat_size_t
) len
;
902 static ssize_t
do_iter_read(struct file
*file
, struct iov_iter
*iter
,
903 loff_t
*pos
, rwf_t flags
)
908 if (!(file
->f_mode
& FMODE_READ
))
910 if (!(file
->f_mode
& FMODE_CAN_READ
))
913 tot_len
= iov_iter_count(iter
);
916 ret
= rw_verify_area(READ
, file
, pos
, tot_len
);
920 if (file
->f_op
->read_iter
)
921 ret
= do_iter_readv_writev(file
, iter
, pos
, READ
, flags
);
923 ret
= do_loop_readv_writev(file
, iter
, pos
, READ
, flags
);
926 fsnotify_access(file
);
930 ssize_t
vfs_iter_read(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
933 if (!file
->f_op
->read_iter
)
935 return do_iter_read(file
, iter
, ppos
, flags
);
937 EXPORT_SYMBOL(vfs_iter_read
);
939 static ssize_t
do_iter_write(struct file
*file
, struct iov_iter
*iter
,
940 loff_t
*pos
, rwf_t flags
)
945 if (!(file
->f_mode
& FMODE_WRITE
))
947 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
950 tot_len
= iov_iter_count(iter
);
953 ret
= rw_verify_area(WRITE
, file
, pos
, tot_len
);
957 if (file
->f_op
->write_iter
)
958 ret
= do_iter_readv_writev(file
, iter
, pos
, WRITE
, flags
);
960 ret
= do_loop_readv_writev(file
, iter
, pos
, WRITE
, flags
);
962 fsnotify_modify(file
);
966 ssize_t
vfs_iter_write(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
969 if (!file
->f_op
->write_iter
)
971 return do_iter_write(file
, iter
, ppos
, flags
);
973 EXPORT_SYMBOL(vfs_iter_write
);
975 ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
976 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
978 struct iovec iovstack
[UIO_FASTIOV
];
979 struct iovec
*iov
= iovstack
;
980 struct iov_iter iter
;
983 ret
= import_iovec(READ
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
, &iter
);
985 ret
= do_iter_read(file
, &iter
, pos
, flags
);
992 static ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
993 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
995 struct iovec iovstack
[UIO_FASTIOV
];
996 struct iovec
*iov
= iovstack
;
997 struct iov_iter iter
;
1000 ret
= import_iovec(WRITE
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
, &iter
);
1002 file_start_write(file
);
1003 ret
= do_iter_write(file
, &iter
, pos
, flags
);
1004 file_end_write(file
);
1010 static ssize_t
do_readv(unsigned long fd
, const struct iovec __user
*vec
,
1011 unsigned long vlen
, rwf_t flags
)
1013 struct fd f
= fdget_pos(fd
);
1014 ssize_t ret
= -EBADF
;
1017 loff_t pos
= file_pos_read(f
.file
);
1018 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1020 file_pos_write(f
.file
, pos
);
1025 add_rchar(current
, ret
);
1030 static ssize_t
do_writev(unsigned long fd
, const struct iovec __user
*vec
,
1031 unsigned long vlen
, rwf_t flags
)
1033 struct fd f
= fdget_pos(fd
);
1034 ssize_t ret
= -EBADF
;
1037 loff_t pos
= file_pos_read(f
.file
);
1038 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1040 file_pos_write(f
.file
, pos
);
1045 add_wchar(current
, ret
);
1050 static inline loff_t
pos_from_hilo(unsigned long high
, unsigned long low
)
1052 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
1053 return (((loff_t
)high
<< HALF_LONG_BITS
) << HALF_LONG_BITS
) | low
;
1056 static ssize_t
do_preadv(unsigned long fd
, const struct iovec __user
*vec
,
1057 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1060 ssize_t ret
= -EBADF
;
1068 if (f
.file
->f_mode
& FMODE_PREAD
)
1069 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1074 add_rchar(current
, ret
);
1079 static ssize_t
do_pwritev(unsigned long fd
, const struct iovec __user
*vec
,
1080 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1083 ssize_t ret
= -EBADF
;
1091 if (f
.file
->f_mode
& FMODE_PWRITE
)
1092 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1097 add_wchar(current
, ret
);
1102 SYSCALL_DEFINE3(readv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1103 unsigned long, vlen
)
1105 return do_readv(fd
, vec
, vlen
, 0);
1108 SYSCALL_DEFINE3(writev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1109 unsigned long, vlen
)
1111 return do_writev(fd
, vec
, vlen
, 0);
1114 SYSCALL_DEFINE5(preadv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1115 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1117 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1119 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1122 SYSCALL_DEFINE6(preadv2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1123 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1126 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1129 return do_readv(fd
, vec
, vlen
, flags
);
1131 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1134 SYSCALL_DEFINE5(pwritev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1135 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1137 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1139 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1142 SYSCALL_DEFINE6(pwritev2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1143 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1146 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1149 return do_writev(fd
, vec
, vlen
, flags
);
1151 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1154 #ifdef CONFIG_COMPAT
1155 static size_t compat_readv(struct file
*file
,
1156 const struct compat_iovec __user
*vec
,
1157 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
1159 struct iovec iovstack
[UIO_FASTIOV
];
1160 struct iovec
*iov
= iovstack
;
1161 struct iov_iter iter
;
1164 ret
= compat_import_iovec(READ
, vec
, vlen
, UIO_FASTIOV
, &iov
, &iter
);
1166 ret
= do_iter_read(file
, &iter
, pos
, flags
);
1170 add_rchar(current
, ret
);
1175 static size_t do_compat_readv(compat_ulong_t fd
,
1176 const struct compat_iovec __user
*vec
,
1177 compat_ulong_t vlen
, rwf_t flags
)
1179 struct fd f
= fdget_pos(fd
);
1185 pos
= f
.file
->f_pos
;
1186 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1188 f
.file
->f_pos
= pos
;
1194 COMPAT_SYSCALL_DEFINE3(readv
, compat_ulong_t
, fd
,
1195 const struct compat_iovec __user
*,vec
,
1196 compat_ulong_t
, vlen
)
1198 return do_compat_readv(fd
, vec
, vlen
, 0);
1201 static long do_compat_preadv64(unsigned long fd
,
1202 const struct compat_iovec __user
*vec
,
1203 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1214 if (f
.file
->f_mode
& FMODE_PREAD
)
1215 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1220 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1221 COMPAT_SYSCALL_DEFINE4(preadv64
, unsigned long, fd
,
1222 const struct compat_iovec __user
*,vec
,
1223 unsigned long, vlen
, loff_t
, pos
)
1225 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1229 COMPAT_SYSCALL_DEFINE5(preadv
, compat_ulong_t
, fd
,
1230 const struct compat_iovec __user
*,vec
,
1231 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1233 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1235 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1238 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1239 COMPAT_SYSCALL_DEFINE5(preadv64v2
, unsigned long, fd
,
1240 const struct compat_iovec __user
*,vec
,
1241 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1243 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1247 COMPAT_SYSCALL_DEFINE6(preadv2
, compat_ulong_t
, fd
,
1248 const struct compat_iovec __user
*,vec
,
1249 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
,
1252 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1255 return do_compat_readv(fd
, vec
, vlen
, flags
);
1257 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1260 static size_t compat_writev(struct file
*file
,
1261 const struct compat_iovec __user
*vec
,
1262 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
1264 struct iovec iovstack
[UIO_FASTIOV
];
1265 struct iovec
*iov
= iovstack
;
1266 struct iov_iter iter
;
1269 ret
= compat_import_iovec(WRITE
, vec
, vlen
, UIO_FASTIOV
, &iov
, &iter
);
1271 file_start_write(file
);
1272 ret
= do_iter_write(file
, &iter
, pos
, flags
);
1273 file_end_write(file
);
1277 add_wchar(current
, ret
);
1282 static size_t do_compat_writev(compat_ulong_t fd
,
1283 const struct compat_iovec __user
* vec
,
1284 compat_ulong_t vlen
, rwf_t flags
)
1286 struct fd f
= fdget_pos(fd
);
1292 pos
= f
.file
->f_pos
;
1293 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1295 f
.file
->f_pos
= pos
;
1300 COMPAT_SYSCALL_DEFINE3(writev
, compat_ulong_t
, fd
,
1301 const struct compat_iovec __user
*, vec
,
1302 compat_ulong_t
, vlen
)
1304 return do_compat_writev(fd
, vec
, vlen
, 0);
1307 static long do_compat_pwritev64(unsigned long fd
,
1308 const struct compat_iovec __user
*vec
,
1309 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1320 if (f
.file
->f_mode
& FMODE_PWRITE
)
1321 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1326 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1327 COMPAT_SYSCALL_DEFINE4(pwritev64
, unsigned long, fd
,
1328 const struct compat_iovec __user
*,vec
,
1329 unsigned long, vlen
, loff_t
, pos
)
1331 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1335 COMPAT_SYSCALL_DEFINE5(pwritev
, compat_ulong_t
, fd
,
1336 const struct compat_iovec __user
*,vec
,
1337 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1339 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1341 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1344 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1345 COMPAT_SYSCALL_DEFINE5(pwritev64v2
, unsigned long, fd
,
1346 const struct compat_iovec __user
*,vec
,
1347 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1349 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1353 COMPAT_SYSCALL_DEFINE6(pwritev2
, compat_ulong_t
, fd
,
1354 const struct compat_iovec __user
*,vec
,
1355 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
, rwf_t
, flags
)
1357 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1360 return do_compat_writev(fd
, vec
, vlen
, flags
);
1362 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1367 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
1368 size_t count
, loff_t max
)
1371 struct inode
*in_inode
, *out_inode
;
1378 * Get input file, and verify that it is ok..
1384 if (!(in
.file
->f_mode
& FMODE_READ
))
1388 pos
= in
.file
->f_pos
;
1391 if (!(in
.file
->f_mode
& FMODE_PREAD
))
1394 retval
= rw_verify_area(READ
, in
.file
, &pos
, count
);
1397 if (count
> MAX_RW_COUNT
)
1398 count
= MAX_RW_COUNT
;
1401 * Get output file, and verify that it is ok..
1404 out
= fdget(out_fd
);
1407 if (!(out
.file
->f_mode
& FMODE_WRITE
))
1410 in_inode
= file_inode(in
.file
);
1411 out_inode
= file_inode(out
.file
);
1412 out_pos
= out
.file
->f_pos
;
1413 retval
= rw_verify_area(WRITE
, out
.file
, &out_pos
, count
);
1418 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
1420 if (unlikely(pos
+ count
> max
)) {
1421 retval
= -EOVERFLOW
;
1430 * We need to debate whether we can enable this or not. The
1431 * man page documents EAGAIN return for the output at least,
1432 * and the application is arguably buggy if it doesn't expect
1433 * EAGAIN on a non-blocking file descriptor.
1435 if (in
.file
->f_flags
& O_NONBLOCK
)
1436 fl
= SPLICE_F_NONBLOCK
;
1438 file_start_write(out
.file
);
1439 retval
= do_splice_direct(in
.file
, &pos
, out
.file
, &out_pos
, count
, fl
);
1440 file_end_write(out
.file
);
1443 add_rchar(current
, retval
);
1444 add_wchar(current
, retval
);
1445 fsnotify_access(in
.file
);
1446 fsnotify_modify(out
.file
);
1447 out
.file
->f_pos
= out_pos
;
1451 in
.file
->f_pos
= pos
;
1457 retval
= -EOVERFLOW
;
1467 SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
, off_t __user
*, offset
, size_t, count
)
1474 if (unlikely(get_user(off
, offset
)))
1477 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1478 if (unlikely(put_user(pos
, offset
)))
1483 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1486 SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
, loff_t __user
*, offset
, size_t, count
)
1492 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1494 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1495 if (unlikely(put_user(pos
, offset
)))
1500 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1503 #ifdef CONFIG_COMPAT
1504 COMPAT_SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
,
1505 compat_off_t __user
*, offset
, compat_size_t
, count
)
1512 if (unlikely(get_user(off
, offset
)))
1515 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1516 if (unlikely(put_user(pos
, offset
)))
1521 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1524 COMPAT_SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
,
1525 compat_loff_t __user
*, offset
, compat_size_t
, count
)
1531 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1533 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1534 if (unlikely(put_user(pos
, offset
)))
1539 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1544 * copy_file_range() differs from regular file read and write in that it
1545 * specifically allows return partial success. When it does so is up to
1546 * the copy_file_range method.
1548 ssize_t
vfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
1549 struct file
*file_out
, loff_t pos_out
,
1550 size_t len
, unsigned int flags
)
1552 struct inode
*inode_in
= file_inode(file_in
);
1553 struct inode
*inode_out
= file_inode(file_out
);
1559 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1561 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1564 ret
= rw_verify_area(READ
, file_in
, &pos_in
, len
);
1568 ret
= rw_verify_area(WRITE
, file_out
, &pos_out
, len
);
1572 if (!(file_in
->f_mode
& FMODE_READ
) ||
1573 !(file_out
->f_mode
& FMODE_WRITE
) ||
1574 (file_out
->f_flags
& O_APPEND
))
1577 /* this could be relaxed once a method supports cross-fs copies */
1578 if (inode_in
->i_sb
!= inode_out
->i_sb
)
1584 file_start_write(file_out
);
1587 * Try cloning first, this is supported by more file systems, and
1588 * more efficient if both clone and copy are supported (e.g. NFS).
1590 if (file_in
->f_op
->clone_file_range
) {
1591 ret
= file_in
->f_op
->clone_file_range(file_in
, pos_in
,
1592 file_out
, pos_out
, len
);
1599 if (file_out
->f_op
->copy_file_range
) {
1600 ret
= file_out
->f_op
->copy_file_range(file_in
, pos_in
, file_out
,
1601 pos_out
, len
, flags
);
1602 if (ret
!= -EOPNOTSUPP
)
1606 ret
= do_splice_direct(file_in
, &pos_in
, file_out
, &pos_out
,
1607 len
> MAX_RW_COUNT
? MAX_RW_COUNT
: len
, 0);
1611 fsnotify_access(file_in
);
1612 add_rchar(current
, ret
);
1613 fsnotify_modify(file_out
);
1614 add_wchar(current
, ret
);
1620 file_end_write(file_out
);
1624 EXPORT_SYMBOL(vfs_copy_file_range
);
1626 SYSCALL_DEFINE6(copy_file_range
, int, fd_in
, loff_t __user
*, off_in
,
1627 int, fd_out
, loff_t __user
*, off_out
,
1628 size_t, len
, unsigned int, flags
)
1634 ssize_t ret
= -EBADF
;
1636 f_in
= fdget(fd_in
);
1640 f_out
= fdget(fd_out
);
1646 if (copy_from_user(&pos_in
, off_in
, sizeof(loff_t
)))
1649 pos_in
= f_in
.file
->f_pos
;
1653 if (copy_from_user(&pos_out
, off_out
, sizeof(loff_t
)))
1656 pos_out
= f_out
.file
->f_pos
;
1659 ret
= vfs_copy_file_range(f_in
.file
, pos_in
, f_out
.file
, pos_out
, len
,
1666 if (copy_to_user(off_in
, &pos_in
, sizeof(loff_t
)))
1669 f_in
.file
->f_pos
= pos_in
;
1673 if (copy_to_user(off_out
, &pos_out
, sizeof(loff_t
)))
1676 f_out
.file
->f_pos
= pos_out
;
1688 static int clone_verify_area(struct file
*file
, loff_t pos
, u64 len
, bool write
)
1690 struct inode
*inode
= file_inode(file
);
1692 if (unlikely(pos
< 0))
1695 if (unlikely((loff_t
) (pos
+ len
) < 0))
1698 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
1699 loff_t end
= len
? pos
+ len
- 1 : OFFSET_MAX
;
1702 retval
= locks_mandatory_area(inode
, file
, pos
, end
,
1703 write
? F_WRLCK
: F_RDLCK
);
1708 return security_file_permission(file
, write
? MAY_WRITE
: MAY_READ
);
1712 * Check that the two inodes are eligible for cloning, the ranges make
1713 * sense, and then flush all dirty data. Caller must ensure that the
1714 * inodes have been locked against any other modifications.
1716 * Returns: 0 for "nothing to clone", 1 for "something to clone", or
1717 * the usual negative error code.
1719 int vfs_clone_file_prep_inodes(struct inode
*inode_in
, loff_t pos_in
,
1720 struct inode
*inode_out
, loff_t pos_out
,
1721 u64
*len
, bool is_dedupe
)
1723 loff_t bs
= inode_out
->i_sb
->s_blocksize
;
1726 bool same_inode
= (inode_in
== inode_out
);
1729 /* Don't touch certain kinds of inodes */
1730 if (IS_IMMUTABLE(inode_out
))
1733 if (IS_SWAPFILE(inode_in
) || IS_SWAPFILE(inode_out
))
1736 /* Don't reflink dirs, pipes, sockets... */
1737 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1739 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1742 /* Are we going all the way to the end? */
1743 isize
= i_size_read(inode_in
);
1747 /* Zero length dedupe exits immediately; reflink goes to EOF. */
1749 if (is_dedupe
|| pos_in
== isize
)
1753 *len
= isize
- pos_in
;
1756 /* Ensure offsets don't wrap and the input is inside i_size */
1757 if (pos_in
+ *len
< pos_in
|| pos_out
+ *len
< pos_out
||
1758 pos_in
+ *len
> isize
)
1761 /* Don't allow dedupe past EOF in the dest file */
1765 disize
= i_size_read(inode_out
);
1766 if (pos_out
>= disize
|| pos_out
+ *len
> disize
)
1770 /* If we're linking to EOF, continue to the block boundary. */
1771 if (pos_in
+ *len
== isize
)
1772 blen
= ALIGN(isize
, bs
) - pos_in
;
1776 /* Only reflink if we're aligned to block boundaries */
1777 if (!IS_ALIGNED(pos_in
, bs
) || !IS_ALIGNED(pos_in
+ blen
, bs
) ||
1778 !IS_ALIGNED(pos_out
, bs
) || !IS_ALIGNED(pos_out
+ blen
, bs
))
1781 /* Don't allow overlapped reflink within the same file */
1783 if (pos_out
+ blen
> pos_in
&& pos_out
< pos_in
+ blen
)
1787 /* Wait for the completion of any pending IOs on both files */
1788 inode_dio_wait(inode_in
);
1790 inode_dio_wait(inode_out
);
1792 ret
= filemap_write_and_wait_range(inode_in
->i_mapping
,
1793 pos_in
, pos_in
+ *len
- 1);
1797 ret
= filemap_write_and_wait_range(inode_out
->i_mapping
,
1798 pos_out
, pos_out
+ *len
- 1);
1803 * Check that the extents are the same.
1806 bool is_same
= false;
1808 ret
= vfs_dedupe_file_range_compare(inode_in
, pos_in
,
1809 inode_out
, pos_out
, *len
, &is_same
);
1818 EXPORT_SYMBOL(vfs_clone_file_prep_inodes
);
1820 int vfs_clone_file_range(struct file
*file_in
, loff_t pos_in
,
1821 struct file
*file_out
, loff_t pos_out
, u64 len
)
1823 struct inode
*inode_in
= file_inode(file_in
);
1824 struct inode
*inode_out
= file_inode(file_out
);
1827 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1829 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1833 * FICLONE/FICLONERANGE ioctls enforce that src and dest files are on
1834 * the same mount. Practically, they only need to be on the same file
1837 if (inode_in
->i_sb
!= inode_out
->i_sb
)
1840 if (!(file_in
->f_mode
& FMODE_READ
) ||
1841 !(file_out
->f_mode
& FMODE_WRITE
) ||
1842 (file_out
->f_flags
& O_APPEND
))
1845 if (!file_in
->f_op
->clone_file_range
)
1848 ret
= clone_verify_area(file_in
, pos_in
, len
, false);
1852 ret
= clone_verify_area(file_out
, pos_out
, len
, true);
1856 if (pos_in
+ len
> i_size_read(inode_in
))
1859 ret
= file_in
->f_op
->clone_file_range(file_in
, pos_in
,
1860 file_out
, pos_out
, len
);
1862 fsnotify_access(file_in
);
1863 fsnotify_modify(file_out
);
1868 EXPORT_SYMBOL(vfs_clone_file_range
);
1871 * Read a page's worth of file data into the page cache. Return the page
1874 static struct page
*vfs_dedupe_get_page(struct inode
*inode
, loff_t offset
)
1876 struct address_space
*mapping
;
1880 n
= offset
>> PAGE_SHIFT
;
1881 mapping
= inode
->i_mapping
;
1882 page
= read_mapping_page(mapping
, n
, NULL
);
1885 if (!PageUptodate(page
)) {
1887 return ERR_PTR(-EIO
);
1894 * Compare extents of two files to see if they are the same.
1895 * Caller must have locked both inodes to prevent write races.
1897 int vfs_dedupe_file_range_compare(struct inode
*src
, loff_t srcoff
,
1898 struct inode
*dest
, loff_t destoff
,
1899 loff_t len
, bool *is_same
)
1905 struct page
*src_page
;
1906 struct page
*dest_page
;
1914 src_poff
= srcoff
& (PAGE_SIZE
- 1);
1915 dest_poff
= destoff
& (PAGE_SIZE
- 1);
1916 cmp_len
= min(PAGE_SIZE
- src_poff
,
1917 PAGE_SIZE
- dest_poff
);
1918 cmp_len
= min(cmp_len
, len
);
1922 src_page
= vfs_dedupe_get_page(src
, srcoff
);
1923 if (IS_ERR(src_page
)) {
1924 error
= PTR_ERR(src_page
);
1927 dest_page
= vfs_dedupe_get_page(dest
, destoff
);
1928 if (IS_ERR(dest_page
)) {
1929 error
= PTR_ERR(dest_page
);
1930 unlock_page(src_page
);
1934 src_addr
= kmap_atomic(src_page
);
1935 dest_addr
= kmap_atomic(dest_page
);
1937 flush_dcache_page(src_page
);
1938 flush_dcache_page(dest_page
);
1940 if (memcmp(src_addr
+ src_poff
, dest_addr
+ dest_poff
, cmp_len
))
1943 kunmap_atomic(dest_addr
);
1944 kunmap_atomic(src_addr
);
1945 unlock_page(dest_page
);
1946 unlock_page(src_page
);
1947 put_page(dest_page
);
1964 EXPORT_SYMBOL(vfs_dedupe_file_range_compare
);
1966 int vfs_dedupe_file_range(struct file
*file
, struct file_dedupe_range
*same
)
1968 struct file_dedupe_range_info
*info
;
1969 struct inode
*src
= file_inode(file
);
1974 bool is_admin
= capable(CAP_SYS_ADMIN
);
1975 u16 count
= same
->dest_count
;
1976 struct file
*dst_file
;
1980 if (!(file
->f_mode
& FMODE_READ
))
1983 if (same
->reserved1
|| same
->reserved2
)
1986 off
= same
->src_offset
;
1987 len
= same
->src_length
;
1990 if (S_ISDIR(src
->i_mode
))
1994 if (!S_ISREG(src
->i_mode
))
1997 ret
= clone_verify_area(file
, off
, len
, false);
2002 if (off
+ len
> i_size_read(src
))
2005 /* pre-format output fields to sane values */
2006 for (i
= 0; i
< count
; i
++) {
2007 same
->info
[i
].bytes_deduped
= 0ULL;
2008 same
->info
[i
].status
= FILE_DEDUPE_RANGE_SAME
;
2011 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
2013 struct fd dst_fd
= fdget(info
->dest_fd
);
2015 dst_file
= dst_fd
.file
;
2017 info
->status
= -EBADF
;
2020 dst
= file_inode(dst_file
);
2022 ret
= mnt_want_write_file(dst_file
);
2028 dst_off
= info
->dest_offset
;
2029 ret
= clone_verify_area(dst_file
, dst_off
, len
, true);
2036 if (info
->reserved
) {
2037 info
->status
= -EINVAL
;
2038 } else if (!(is_admin
|| (dst_file
->f_mode
& FMODE_WRITE
))) {
2039 info
->status
= -EINVAL
;
2040 } else if (file
->f_path
.mnt
!= dst_file
->f_path
.mnt
) {
2041 info
->status
= -EXDEV
;
2042 } else if (S_ISDIR(dst
->i_mode
)) {
2043 info
->status
= -EISDIR
;
2044 } else if (dst_file
->f_op
->dedupe_file_range
== NULL
) {
2045 info
->status
= -EINVAL
;
2047 deduped
= dst_file
->f_op
->dedupe_file_range(file
, off
,
2050 if (deduped
== -EBADE
)
2051 info
->status
= FILE_DEDUPE_RANGE_DIFFERS
;
2052 else if (deduped
< 0)
2053 info
->status
= deduped
;
2055 info
->bytes_deduped
+= deduped
;
2059 mnt_drop_write_file(dst_file
);
2063 if (fatal_signal_pending(current
))
2070 EXPORT_SYMBOL(vfs_dedupe_file_range
);