2 * linux/fs/read_write.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/slab.h>
8 #include <linux/stat.h>
9 #include <linux/fcntl.h>
10 #include <linux/file.h>
11 #include <linux/uio.h>
12 #include <linux/fsnotify.h>
13 #include <linux/security.h>
14 #include <linux/export.h>
15 #include <linux/syscalls.h>
16 #include <linux/pagemap.h>
17 #include <linux/splice.h>
18 #include <linux/compat.h>
19 #include <linux/mount.h>
23 #include <asm/uaccess.h>
24 #include <asm/unistd.h>
26 typedef ssize_t (*io_fn_t
)(struct file
*, char __user
*, size_t, loff_t
*);
27 typedef ssize_t (*iter_fn_t
)(struct kiocb
*, struct iov_iter
*);
29 const struct file_operations generic_ro_fops
= {
30 .llseek
= generic_file_llseek
,
31 .read_iter
= generic_file_read_iter
,
32 .mmap
= generic_file_readonly_mmap
,
33 .splice_read
= generic_file_splice_read
,
36 EXPORT_SYMBOL(generic_ro_fops
);
38 static inline int unsigned_offsets(struct file
*file
)
40 return file
->f_mode
& FMODE_UNSIGNED_OFFSET
;
44 * vfs_setpos - update the file offset for lseek
45 * @file: file structure in question
46 * @offset: file offset to seek to
47 * @maxsize: maximum file size
49 * This is a low-level filesystem helper for updating the file offset to
50 * the value specified by @offset if the given offset is valid and it is
51 * not equal to the current file offset.
53 * Return the specified offset on success and -EINVAL on invalid offset.
55 loff_t
vfs_setpos(struct file
*file
, loff_t offset
, loff_t maxsize
)
57 if (offset
< 0 && !unsigned_offsets(file
))
62 if (offset
!= file
->f_pos
) {
68 EXPORT_SYMBOL(vfs_setpos
);
71 * generic_file_llseek_size - generic llseek implementation for regular files
72 * @file: file structure to seek on
73 * @offset: file offset to seek to
74 * @whence: type of seek
75 * @size: max size of this file in file system
76 * @eof: offset used for SEEK_END position
78 * This is a variant of generic_file_llseek that allows passing in a custom
79 * maximum file size and a custom EOF position, for e.g. hashed directories
82 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
83 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
84 * read/writes behave like SEEK_SET against seeks.
87 generic_file_llseek_size(struct file
*file
, loff_t offset
, int whence
,
88 loff_t maxsize
, loff_t eof
)
96 * Here we special-case the lseek(fd, 0, SEEK_CUR)
97 * position-querying operation. Avoid rewriting the "same"
98 * f_pos value back to the file because a concurrent read(),
99 * write() or lseek() might have altered it
104 * f_lock protects against read/modify/write race with other
105 * SEEK_CURs. Note that parallel writes and reads behave
108 spin_lock(&file
->f_lock
);
109 offset
= vfs_setpos(file
, file
->f_pos
+ offset
, maxsize
);
110 spin_unlock(&file
->f_lock
);
114 * In the generic case the entire file is data, so as long as
115 * offset isn't at the end of the file then the offset is data.
117 if ((unsigned long long)offset
>= eof
)
122 * There is a virtual hole at the end of the file, so as long as
123 * offset isn't i_size or larger, return i_size.
125 if ((unsigned long long)offset
>= eof
)
131 return vfs_setpos(file
, offset
, maxsize
);
133 EXPORT_SYMBOL(generic_file_llseek_size
);
136 * generic_file_llseek - generic llseek implementation for regular files
137 * @file: file structure to seek on
138 * @offset: file offset to seek to
139 * @whence: type of seek
141 * This is a generic implemenation of ->llseek useable for all normal local
142 * filesystems. It just updates the file offset to the value specified by
143 * @offset and @whence.
145 loff_t
generic_file_llseek(struct file
*file
, loff_t offset
, int whence
)
147 struct inode
*inode
= file
->f_mapping
->host
;
149 return generic_file_llseek_size(file
, offset
, whence
,
150 inode
->i_sb
->s_maxbytes
,
153 EXPORT_SYMBOL(generic_file_llseek
);
156 * fixed_size_llseek - llseek implementation for fixed-sized devices
157 * @file: file structure to seek on
158 * @offset: file offset to seek to
159 * @whence: type of seek
160 * @size: size of the file
163 loff_t
fixed_size_llseek(struct file
*file
, loff_t offset
, int whence
, loff_t size
)
166 case SEEK_SET
: case SEEK_CUR
: case SEEK_END
:
167 return generic_file_llseek_size(file
, offset
, whence
,
173 EXPORT_SYMBOL(fixed_size_llseek
);
176 * no_seek_end_llseek - llseek implementation for fixed-sized devices
177 * @file: file structure to seek on
178 * @offset: file offset to seek to
179 * @whence: type of seek
182 loff_t
no_seek_end_llseek(struct file
*file
, loff_t offset
, int whence
)
185 case SEEK_SET
: case SEEK_CUR
:
186 return generic_file_llseek_size(file
, offset
, whence
,
192 EXPORT_SYMBOL(no_seek_end_llseek
);
195 * no_seek_end_llseek_size - llseek implementation for fixed-sized devices
196 * @file: file structure to seek on
197 * @offset: file offset to seek to
198 * @whence: type of seek
199 * @size: maximal offset allowed
202 loff_t
no_seek_end_llseek_size(struct file
*file
, loff_t offset
, int whence
, loff_t size
)
205 case SEEK_SET
: case SEEK_CUR
:
206 return generic_file_llseek_size(file
, offset
, whence
,
212 EXPORT_SYMBOL(no_seek_end_llseek_size
);
215 * noop_llseek - No Operation Performed llseek implementation
216 * @file: file structure to seek on
217 * @offset: file offset to seek to
218 * @whence: type of seek
220 * This is an implementation of ->llseek useable for the rare special case when
221 * userspace expects the seek to succeed but the (device) file is actually not
222 * able to perform the seek. In this case you use noop_llseek() instead of
223 * falling back to the default implementation of ->llseek.
225 loff_t
noop_llseek(struct file
*file
, loff_t offset
, int whence
)
229 EXPORT_SYMBOL(noop_llseek
);
231 loff_t
no_llseek(struct file
*file
, loff_t offset
, int whence
)
235 EXPORT_SYMBOL(no_llseek
);
237 loff_t
default_llseek(struct file
*file
, loff_t offset
, int whence
)
239 struct inode
*inode
= file_inode(file
);
245 offset
+= i_size_read(inode
);
249 retval
= file
->f_pos
;
252 offset
+= file
->f_pos
;
256 * In the generic case the entire file is data, so as
257 * long as offset isn't at the end of the file then the
260 if (offset
>= inode
->i_size
) {
267 * There is a virtual hole at the end of the file, so
268 * as long as offset isn't i_size or larger, return
271 if (offset
>= inode
->i_size
) {
275 offset
= inode
->i_size
;
279 if (offset
>= 0 || unsigned_offsets(file
)) {
280 if (offset
!= file
->f_pos
) {
281 file
->f_pos
= offset
;
290 EXPORT_SYMBOL(default_llseek
);
292 loff_t
vfs_llseek(struct file
*file
, loff_t offset
, int whence
)
294 loff_t (*fn
)(struct file
*, loff_t
, int);
297 if (file
->f_mode
& FMODE_LSEEK
) {
298 if (file
->f_op
->llseek
)
299 fn
= file
->f_op
->llseek
;
301 return fn(file
, offset
, whence
);
303 EXPORT_SYMBOL(vfs_llseek
);
305 SYSCALL_DEFINE3(lseek
, unsigned int, fd
, off_t
, offset
, unsigned int, whence
)
308 struct fd f
= fdget_pos(fd
);
313 if (whence
<= SEEK_MAX
) {
314 loff_t res
= vfs_llseek(f
.file
, offset
, whence
);
316 if (res
!= (loff_t
)retval
)
317 retval
= -EOVERFLOW
; /* LFS: should only happen on 32 bit platforms */
324 COMPAT_SYSCALL_DEFINE3(lseek
, unsigned int, fd
, compat_off_t
, offset
, unsigned int, whence
)
326 return sys_lseek(fd
, offset
, whence
);
330 #ifdef __ARCH_WANT_SYS_LLSEEK
331 SYSCALL_DEFINE5(llseek
, unsigned int, fd
, unsigned long, offset_high
,
332 unsigned long, offset_low
, loff_t __user
*, result
,
333 unsigned int, whence
)
336 struct fd f
= fdget_pos(fd
);
343 if (whence
> SEEK_MAX
)
346 offset
= vfs_llseek(f
.file
, ((loff_t
) offset_high
<< 32) | offset_low
,
349 retval
= (int)offset
;
352 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
361 ssize_t
vfs_iter_read(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
)
366 if (!file
->f_op
->read_iter
)
369 init_sync_kiocb(&kiocb
, file
);
370 kiocb
.ki_pos
= *ppos
;
373 ret
= file
->f_op
->read_iter(&kiocb
, iter
);
374 BUG_ON(ret
== -EIOCBQUEUED
);
376 *ppos
= kiocb
.ki_pos
;
379 EXPORT_SYMBOL(vfs_iter_read
);
381 ssize_t
vfs_iter_write(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
)
386 if (!file
->f_op
->write_iter
)
389 init_sync_kiocb(&kiocb
, file
);
390 kiocb
.ki_pos
= *ppos
;
393 ret
= file
->f_op
->write_iter(&kiocb
, iter
);
394 BUG_ON(ret
== -EIOCBQUEUED
);
396 *ppos
= kiocb
.ki_pos
;
397 fsnotify_modify(file
);
401 EXPORT_SYMBOL(vfs_iter_write
);
403 int rw_verify_area(int read_write
, struct file
*file
, const loff_t
*ppos
, size_t count
)
407 int retval
= -EINVAL
;
409 inode
= file_inode(file
);
410 if (unlikely((ssize_t
) count
< 0))
413 if (unlikely(pos
< 0)) {
414 if (!unsigned_offsets(file
))
416 if (count
>= -pos
) /* both values are in 0..LLONG_MAX */
418 } else if (unlikely((loff_t
) (pos
+ count
) < 0)) {
419 if (!unsigned_offsets(file
))
423 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
424 retval
= locks_mandatory_area(inode
, file
, pos
, pos
+ count
- 1,
425 read_write
== READ
? F_RDLCK
: F_WRLCK
);
429 return security_file_permission(file
,
430 read_write
== READ
? MAY_READ
: MAY_WRITE
);
433 static ssize_t
new_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
435 struct iovec iov
= { .iov_base
= buf
, .iov_len
= len
};
437 struct iov_iter iter
;
440 init_sync_kiocb(&kiocb
, filp
);
441 kiocb
.ki_pos
= *ppos
;
442 iov_iter_init(&iter
, READ
, &iov
, 1, len
);
444 ret
= filp
->f_op
->read_iter(&kiocb
, &iter
);
445 BUG_ON(ret
== -EIOCBQUEUED
);
446 *ppos
= kiocb
.ki_pos
;
450 ssize_t
__vfs_read(struct file
*file
, char __user
*buf
, size_t count
,
453 if (file
->f_op
->read
)
454 return file
->f_op
->read(file
, buf
, count
, pos
);
455 else if (file
->f_op
->read_iter
)
456 return new_sync_read(file
, buf
, count
, pos
);
460 EXPORT_SYMBOL(__vfs_read
);
462 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
466 if (!(file
->f_mode
& FMODE_READ
))
468 if (!(file
->f_mode
& FMODE_CAN_READ
))
470 if (unlikely(!access_ok(VERIFY_WRITE
, buf
, count
)))
473 ret
= rw_verify_area(READ
, file
, pos
, count
);
475 if (count
> MAX_RW_COUNT
)
476 count
= MAX_RW_COUNT
;
477 ret
= __vfs_read(file
, buf
, count
, pos
);
479 fsnotify_access(file
);
480 add_rchar(current
, ret
);
488 EXPORT_SYMBOL(vfs_read
);
490 static ssize_t
new_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
492 struct iovec iov
= { .iov_base
= (void __user
*)buf
, .iov_len
= len
};
494 struct iov_iter iter
;
497 init_sync_kiocb(&kiocb
, filp
);
498 kiocb
.ki_pos
= *ppos
;
499 iov_iter_init(&iter
, WRITE
, &iov
, 1, len
);
501 ret
= filp
->f_op
->write_iter(&kiocb
, &iter
);
502 BUG_ON(ret
== -EIOCBQUEUED
);
504 *ppos
= kiocb
.ki_pos
;
508 ssize_t
__vfs_write(struct file
*file
, const char __user
*p
, size_t count
,
511 if (file
->f_op
->write
)
512 return file
->f_op
->write(file
, p
, count
, pos
);
513 else if (file
->f_op
->write_iter
)
514 return new_sync_write(file
, p
, count
, pos
);
518 EXPORT_SYMBOL(__vfs_write
);
520 ssize_t
__kernel_write(struct file
*file
, const char *buf
, size_t count
, loff_t
*pos
)
523 const char __user
*p
;
526 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
531 p
= (__force
const char __user
*)buf
;
532 if (count
> MAX_RW_COUNT
)
533 count
= MAX_RW_COUNT
;
534 ret
= __vfs_write(file
, p
, count
, pos
);
537 fsnotify_modify(file
);
538 add_wchar(current
, ret
);
544 EXPORT_SYMBOL(__kernel_write
);
546 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
550 if (!(file
->f_mode
& FMODE_WRITE
))
552 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
554 if (unlikely(!access_ok(VERIFY_READ
, buf
, count
)))
557 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
559 if (count
> MAX_RW_COUNT
)
560 count
= MAX_RW_COUNT
;
561 file_start_write(file
);
562 ret
= __vfs_write(file
, buf
, count
, pos
);
564 fsnotify_modify(file
);
565 add_wchar(current
, ret
);
568 file_end_write(file
);
574 EXPORT_SYMBOL(vfs_write
);
576 static inline loff_t
file_pos_read(struct file
*file
)
581 static inline void file_pos_write(struct file
*file
, loff_t pos
)
586 SYSCALL_DEFINE3(read
, unsigned int, fd
, char __user
*, buf
, size_t, count
)
588 struct fd f
= fdget_pos(fd
);
589 ssize_t ret
= -EBADF
;
592 loff_t pos
= file_pos_read(f
.file
);
593 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
595 file_pos_write(f
.file
, pos
);
601 SYSCALL_DEFINE3(write
, unsigned int, fd
, const char __user
*, buf
,
604 struct fd f
= fdget_pos(fd
);
605 ssize_t ret
= -EBADF
;
608 loff_t pos
= file_pos_read(f
.file
);
609 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
611 file_pos_write(f
.file
, pos
);
618 SYSCALL_DEFINE4(pread64
, unsigned int, fd
, char __user
*, buf
,
619 size_t, count
, loff_t
, pos
)
622 ssize_t ret
= -EBADF
;
630 if (f
.file
->f_mode
& FMODE_PREAD
)
631 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
638 SYSCALL_DEFINE4(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
639 size_t, count
, loff_t
, pos
)
642 ssize_t ret
= -EBADF
;
650 if (f
.file
->f_mode
& FMODE_PWRITE
)
651 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
659 * Reduce an iovec's length in-place. Return the resulting number of segments
661 unsigned long iov_shorten(struct iovec
*iov
, unsigned long nr_segs
, size_t to
)
663 unsigned long seg
= 0;
666 while (seg
< nr_segs
) {
668 if (len
+ iov
->iov_len
>= to
) {
669 iov
->iov_len
= to
- len
;
677 EXPORT_SYMBOL(iov_shorten
);
679 static ssize_t
do_iter_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
680 loff_t
*ppos
, iter_fn_t fn
, int flags
)
685 if (flags
& ~(RWF_HIPRI
| RWF_DSYNC
| RWF_SYNC
))
688 init_sync_kiocb(&kiocb
, filp
);
689 if (flags
& RWF_HIPRI
)
690 kiocb
.ki_flags
|= IOCB_HIPRI
;
691 if (flags
& RWF_DSYNC
)
692 kiocb
.ki_flags
|= IOCB_DSYNC
;
693 if (flags
& RWF_SYNC
)
694 kiocb
.ki_flags
|= (IOCB_DSYNC
| IOCB_SYNC
);
695 kiocb
.ki_pos
= *ppos
;
697 ret
= fn(&kiocb
, iter
);
698 BUG_ON(ret
== -EIOCBQUEUED
);
699 *ppos
= kiocb
.ki_pos
;
703 /* Do it by hand, with file-ops */
704 static ssize_t
do_loop_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
705 loff_t
*ppos
, io_fn_t fn
, int flags
)
709 if (flags
& ~RWF_HIPRI
)
712 while (iov_iter_count(iter
)) {
713 struct iovec iovec
= iov_iter_iovec(iter
);
716 nr
= fn(filp
, iovec
.iov_base
, iovec
.iov_len
, ppos
);
724 if (nr
!= iovec
.iov_len
)
726 iov_iter_advance(iter
, nr
);
732 /* A write operation does a read from user space and vice versa */
733 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
736 * rw_copy_check_uvector() - Copy an array of &struct iovec from userspace
737 * into the kernel and check that it is valid.
739 * @type: One of %CHECK_IOVEC_ONLY, %READ, or %WRITE.
740 * @uvector: Pointer to the userspace array.
741 * @nr_segs: Number of elements in userspace array.
742 * @fast_segs: Number of elements in @fast_pointer.
743 * @fast_pointer: Pointer to (usually small on-stack) kernel array.
744 * @ret_pointer: (output parameter) Pointer to a variable that will point to
745 * either @fast_pointer, a newly allocated kernel array, or NULL,
746 * depending on which array was used.
748 * This function copies an array of &struct iovec of @nr_segs from
749 * userspace into the kernel and checks that each element is valid (e.g.
750 * it does not point to a kernel address or cause overflow by being too
753 * As an optimization, the caller may provide a pointer to a small
754 * on-stack array in @fast_pointer, typically %UIO_FASTIOV elements long
755 * (the size of this array, or 0 if unused, should be given in @fast_segs).
757 * @ret_pointer will always point to the array that was used, so the
758 * caller must take care not to call kfree() on it e.g. in case the
759 * @fast_pointer array was used and it was allocated on the stack.
761 * Return: The total number of bytes covered by the iovec array on success
762 * or a negative error code on error.
764 ssize_t
rw_copy_check_uvector(int type
, const struct iovec __user
* uvector
,
765 unsigned long nr_segs
, unsigned long fast_segs
,
766 struct iovec
*fast_pointer
,
767 struct iovec
**ret_pointer
)
771 struct iovec
*iov
= fast_pointer
;
774 * SuS says "The readv() function *may* fail if the iovcnt argument
775 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
776 * traditionally returned zero for zero segments, so...
784 * First get the "struct iovec" from user memory and
785 * verify all the pointers
787 if (nr_segs
> UIO_MAXIOV
) {
791 if (nr_segs
> fast_segs
) {
792 iov
= kmalloc(nr_segs
*sizeof(struct iovec
), GFP_KERNEL
);
798 if (copy_from_user(iov
, uvector
, nr_segs
*sizeof(*uvector
))) {
804 * According to the Single Unix Specification we should return EINVAL
805 * if an element length is < 0 when cast to ssize_t or if the
806 * total length would overflow the ssize_t return value of the
809 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
813 for (seg
= 0; seg
< nr_segs
; seg
++) {
814 void __user
*buf
= iov
[seg
].iov_base
;
815 ssize_t len
= (ssize_t
)iov
[seg
].iov_len
;
817 /* see if we we're about to use an invalid len or if
818 * it's about to overflow ssize_t */
824 && unlikely(!access_ok(vrfy_dir(type
), buf
, len
))) {
828 if (len
> MAX_RW_COUNT
- ret
) {
829 len
= MAX_RW_COUNT
- ret
;
830 iov
[seg
].iov_len
= len
;
839 static ssize_t
do_readv_writev(int type
, struct file
*file
,
840 const struct iovec __user
* uvector
,
841 unsigned long nr_segs
, loff_t
*pos
,
845 struct iovec iovstack
[UIO_FASTIOV
];
846 struct iovec
*iov
= iovstack
;
847 struct iov_iter iter
;
852 ret
= import_iovec(type
, uvector
, nr_segs
,
853 ARRAY_SIZE(iovstack
), &iov
, &iter
);
857 tot_len
= iov_iter_count(&iter
);
860 ret
= rw_verify_area(type
, file
, pos
, tot_len
);
865 fn
= file
->f_op
->read
;
866 iter_fn
= file
->f_op
->read_iter
;
868 fn
= (io_fn_t
)file
->f_op
->write
;
869 iter_fn
= file
->f_op
->write_iter
;
870 file_start_write(file
);
874 ret
= do_iter_readv_writev(file
, &iter
, pos
, iter_fn
, flags
);
876 ret
= do_loop_readv_writev(file
, &iter
, pos
, fn
, flags
);
879 file_end_write(file
);
883 if ((ret
+ (type
== READ
)) > 0) {
885 fsnotify_access(file
);
887 fsnotify_modify(file
);
892 ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
893 unsigned long vlen
, loff_t
*pos
, int flags
)
895 if (!(file
->f_mode
& FMODE_READ
))
897 if (!(file
->f_mode
& FMODE_CAN_READ
))
900 return do_readv_writev(READ
, file
, vec
, vlen
, pos
, flags
);
903 EXPORT_SYMBOL(vfs_readv
);
905 ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
906 unsigned long vlen
, loff_t
*pos
, int flags
)
908 if (!(file
->f_mode
& FMODE_WRITE
))
910 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
913 return do_readv_writev(WRITE
, file
, vec
, vlen
, pos
, flags
);
916 EXPORT_SYMBOL(vfs_writev
);
918 static ssize_t
do_readv(unsigned long fd
, const struct iovec __user
*vec
,
919 unsigned long vlen
, int flags
)
921 struct fd f
= fdget_pos(fd
);
922 ssize_t ret
= -EBADF
;
925 loff_t pos
= file_pos_read(f
.file
);
926 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
928 file_pos_write(f
.file
, pos
);
933 add_rchar(current
, ret
);
938 static ssize_t
do_writev(unsigned long fd
, const struct iovec __user
*vec
,
939 unsigned long vlen
, int flags
)
941 struct fd f
= fdget_pos(fd
);
942 ssize_t ret
= -EBADF
;
945 loff_t pos
= file_pos_read(f
.file
);
946 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
948 file_pos_write(f
.file
, pos
);
953 add_wchar(current
, ret
);
958 static inline loff_t
pos_from_hilo(unsigned long high
, unsigned long low
)
960 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
961 return (((loff_t
)high
<< HALF_LONG_BITS
) << HALF_LONG_BITS
) | low
;
964 static ssize_t
do_preadv(unsigned long fd
, const struct iovec __user
*vec
,
965 unsigned long vlen
, loff_t pos
, int flags
)
968 ssize_t ret
= -EBADF
;
976 if (f
.file
->f_mode
& FMODE_PREAD
)
977 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
982 add_rchar(current
, ret
);
987 static ssize_t
do_pwritev(unsigned long fd
, const struct iovec __user
*vec
,
988 unsigned long vlen
, loff_t pos
, int flags
)
991 ssize_t ret
= -EBADF
;
999 if (f
.file
->f_mode
& FMODE_PWRITE
)
1000 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1005 add_wchar(current
, ret
);
1010 SYSCALL_DEFINE3(readv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1011 unsigned long, vlen
)
1013 return do_readv(fd
, vec
, vlen
, 0);
1016 SYSCALL_DEFINE3(writev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1017 unsigned long, vlen
)
1019 return do_writev(fd
, vec
, vlen
, 0);
1022 SYSCALL_DEFINE5(preadv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1023 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1025 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1027 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1030 SYSCALL_DEFINE6(preadv2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1031 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1034 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1037 return do_readv(fd
, vec
, vlen
, flags
);
1039 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1042 SYSCALL_DEFINE5(pwritev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1043 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1045 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1047 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1050 SYSCALL_DEFINE6(pwritev2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1051 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1054 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1057 return do_writev(fd
, vec
, vlen
, flags
);
1059 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1062 #ifdef CONFIG_COMPAT
1064 static ssize_t
compat_do_readv_writev(int type
, struct file
*file
,
1065 const struct compat_iovec __user
*uvector
,
1066 unsigned long nr_segs
, loff_t
*pos
,
1069 compat_ssize_t tot_len
;
1070 struct iovec iovstack
[UIO_FASTIOV
];
1071 struct iovec
*iov
= iovstack
;
1072 struct iov_iter iter
;
1077 ret
= compat_import_iovec(type
, uvector
, nr_segs
,
1078 UIO_FASTIOV
, &iov
, &iter
);
1082 tot_len
= iov_iter_count(&iter
);
1085 ret
= rw_verify_area(type
, file
, pos
, tot_len
);
1090 fn
= file
->f_op
->read
;
1091 iter_fn
= file
->f_op
->read_iter
;
1093 fn
= (io_fn_t
)file
->f_op
->write
;
1094 iter_fn
= file
->f_op
->write_iter
;
1095 file_start_write(file
);
1099 ret
= do_iter_readv_writev(file
, &iter
, pos
, iter_fn
, flags
);
1101 ret
= do_loop_readv_writev(file
, &iter
, pos
, fn
, flags
);
1104 file_end_write(file
);
1108 if ((ret
+ (type
== READ
)) > 0) {
1110 fsnotify_access(file
);
1112 fsnotify_modify(file
);
1117 static size_t compat_readv(struct file
*file
,
1118 const struct compat_iovec __user
*vec
,
1119 unsigned long vlen
, loff_t
*pos
, int flags
)
1121 ssize_t ret
= -EBADF
;
1123 if (!(file
->f_mode
& FMODE_READ
))
1127 if (!(file
->f_mode
& FMODE_CAN_READ
))
1130 ret
= compat_do_readv_writev(READ
, file
, vec
, vlen
, pos
, flags
);
1134 add_rchar(current
, ret
);
1139 static size_t do_compat_readv(compat_ulong_t fd
,
1140 const struct compat_iovec __user
*vec
,
1141 compat_ulong_t vlen
, int flags
)
1143 struct fd f
= fdget_pos(fd
);
1149 pos
= f
.file
->f_pos
;
1150 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1152 f
.file
->f_pos
= pos
;
1158 COMPAT_SYSCALL_DEFINE3(readv
, compat_ulong_t
, fd
,
1159 const struct compat_iovec __user
*,vec
,
1160 compat_ulong_t
, vlen
)
1162 return do_compat_readv(fd
, vec
, vlen
, 0);
1165 static long do_compat_preadv64(unsigned long fd
,
1166 const struct compat_iovec __user
*vec
,
1167 unsigned long vlen
, loff_t pos
, int flags
)
1178 if (f
.file
->f_mode
& FMODE_PREAD
)
1179 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1184 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1185 COMPAT_SYSCALL_DEFINE4(preadv64
, unsigned long, fd
,
1186 const struct compat_iovec __user
*,vec
,
1187 unsigned long, vlen
, loff_t
, pos
)
1189 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1193 COMPAT_SYSCALL_DEFINE5(preadv
, compat_ulong_t
, fd
,
1194 const struct compat_iovec __user
*,vec
,
1195 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1197 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1199 return do_compat_preadv64(fd
, vec
, vlen
, pos
, 0);
1202 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1203 COMPAT_SYSCALL_DEFINE5(preadv64v2
, unsigned long, fd
,
1204 const struct compat_iovec __user
*,vec
,
1205 unsigned long, vlen
, loff_t
, pos
, int, flags
)
1208 return do_compat_readv(fd
, vec
, vlen
, flags
);
1210 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1214 COMPAT_SYSCALL_DEFINE6(preadv2
, compat_ulong_t
, fd
,
1215 const struct compat_iovec __user
*,vec
,
1216 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
,
1219 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1222 return do_compat_readv(fd
, vec
, vlen
, flags
);
1224 return do_compat_preadv64(fd
, vec
, vlen
, pos
, flags
);
1227 static size_t compat_writev(struct file
*file
,
1228 const struct compat_iovec __user
*vec
,
1229 unsigned long vlen
, loff_t
*pos
, int flags
)
1231 ssize_t ret
= -EBADF
;
1233 if (!(file
->f_mode
& FMODE_WRITE
))
1237 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
1240 ret
= compat_do_readv_writev(WRITE
, file
, vec
, vlen
, pos
, flags
);
1244 add_wchar(current
, ret
);
1249 static size_t do_compat_writev(compat_ulong_t fd
,
1250 const struct compat_iovec __user
* vec
,
1251 compat_ulong_t vlen
, int flags
)
1253 struct fd f
= fdget_pos(fd
);
1259 pos
= f
.file
->f_pos
;
1260 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1262 f
.file
->f_pos
= pos
;
1267 COMPAT_SYSCALL_DEFINE3(writev
, compat_ulong_t
, fd
,
1268 const struct compat_iovec __user
*, vec
,
1269 compat_ulong_t
, vlen
)
1271 return do_compat_writev(fd
, vec
, vlen
, 0);
1274 static long do_compat_pwritev64(unsigned long fd
,
1275 const struct compat_iovec __user
*vec
,
1276 unsigned long vlen
, loff_t pos
, int flags
)
1287 if (f
.file
->f_mode
& FMODE_PWRITE
)
1288 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1293 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1294 COMPAT_SYSCALL_DEFINE4(pwritev64
, unsigned long, fd
,
1295 const struct compat_iovec __user
*,vec
,
1296 unsigned long, vlen
, loff_t
, pos
)
1298 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1302 COMPAT_SYSCALL_DEFINE5(pwritev
, compat_ulong_t
, fd
,
1303 const struct compat_iovec __user
*,vec
,
1304 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1306 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1308 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, 0);
1311 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1312 COMPAT_SYSCALL_DEFINE5(pwritev64v2
, unsigned long, fd
,
1313 const struct compat_iovec __user
*,vec
,
1314 unsigned long, vlen
, loff_t
, pos
, int, flags
)
1317 return do_compat_writev(fd
, vec
, vlen
, flags
);
1319 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1323 COMPAT_SYSCALL_DEFINE6(pwritev2
, compat_ulong_t
, fd
,
1324 const struct compat_iovec __user
*,vec
,
1325 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
, int, flags
)
1327 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1330 return do_compat_writev(fd
, vec
, vlen
, flags
);
1332 return do_compat_pwritev64(fd
, vec
, vlen
, pos
, flags
);
1337 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
1338 size_t count
, loff_t max
)
1341 struct inode
*in_inode
, *out_inode
;
1348 * Get input file, and verify that it is ok..
1354 if (!(in
.file
->f_mode
& FMODE_READ
))
1358 pos
= in
.file
->f_pos
;
1361 if (!(in
.file
->f_mode
& FMODE_PREAD
))
1364 retval
= rw_verify_area(READ
, in
.file
, &pos
, count
);
1367 if (count
> MAX_RW_COUNT
)
1368 count
= MAX_RW_COUNT
;
1371 * Get output file, and verify that it is ok..
1374 out
= fdget(out_fd
);
1377 if (!(out
.file
->f_mode
& FMODE_WRITE
))
1380 in_inode
= file_inode(in
.file
);
1381 out_inode
= file_inode(out
.file
);
1382 out_pos
= out
.file
->f_pos
;
1383 retval
= rw_verify_area(WRITE
, out
.file
, &out_pos
, count
);
1388 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
1390 if (unlikely(pos
+ count
> max
)) {
1391 retval
= -EOVERFLOW
;
1400 * We need to debate whether we can enable this or not. The
1401 * man page documents EAGAIN return for the output at least,
1402 * and the application is arguably buggy if it doesn't expect
1403 * EAGAIN on a non-blocking file descriptor.
1405 if (in
.file
->f_flags
& O_NONBLOCK
)
1406 fl
= SPLICE_F_NONBLOCK
;
1408 file_start_write(out
.file
);
1409 retval
= do_splice_direct(in
.file
, &pos
, out
.file
, &out_pos
, count
, fl
);
1410 file_end_write(out
.file
);
1413 add_rchar(current
, retval
);
1414 add_wchar(current
, retval
);
1415 fsnotify_access(in
.file
);
1416 fsnotify_modify(out
.file
);
1417 out
.file
->f_pos
= out_pos
;
1421 in
.file
->f_pos
= pos
;
1427 retval
= -EOVERFLOW
;
1437 SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
, off_t __user
*, offset
, size_t, count
)
1444 if (unlikely(get_user(off
, offset
)))
1447 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1448 if (unlikely(put_user(pos
, offset
)))
1453 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1456 SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
, loff_t __user
*, offset
, size_t, count
)
1462 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1464 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1465 if (unlikely(put_user(pos
, offset
)))
1470 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1473 #ifdef CONFIG_COMPAT
1474 COMPAT_SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
,
1475 compat_off_t __user
*, offset
, compat_size_t
, count
)
1482 if (unlikely(get_user(off
, offset
)))
1485 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1486 if (unlikely(put_user(pos
, offset
)))
1491 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1494 COMPAT_SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
,
1495 compat_loff_t __user
*, offset
, compat_size_t
, count
)
1501 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1503 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1504 if (unlikely(put_user(pos
, offset
)))
1509 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1514 * copy_file_range() differs from regular file read and write in that it
1515 * specifically allows return partial success. When it does so is up to
1516 * the copy_file_range method.
1518 ssize_t
vfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
1519 struct file
*file_out
, loff_t pos_out
,
1520 size_t len
, unsigned int flags
)
1522 struct inode
*inode_in
= file_inode(file_in
);
1523 struct inode
*inode_out
= file_inode(file_out
);
1529 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1531 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1534 ret
= rw_verify_area(READ
, file_in
, &pos_in
, len
);
1538 ret
= rw_verify_area(WRITE
, file_out
, &pos_out
, len
);
1542 if (!(file_in
->f_mode
& FMODE_READ
) ||
1543 !(file_out
->f_mode
& FMODE_WRITE
) ||
1544 (file_out
->f_flags
& O_APPEND
))
1547 /* this could be relaxed once a method supports cross-fs copies */
1548 if (inode_in
->i_sb
!= inode_out
->i_sb
)
1554 ret
= mnt_want_write_file(file_out
);
1559 if (file_out
->f_op
->copy_file_range
)
1560 ret
= file_out
->f_op
->copy_file_range(file_in
, pos_in
, file_out
,
1561 pos_out
, len
, flags
);
1562 if (ret
== -EOPNOTSUPP
)
1563 ret
= do_splice_direct(file_in
, &pos_in
, file_out
, &pos_out
,
1564 len
> MAX_RW_COUNT
? MAX_RW_COUNT
: len
, 0);
1567 fsnotify_access(file_in
);
1568 add_rchar(current
, ret
);
1569 fsnotify_modify(file_out
);
1570 add_wchar(current
, ret
);
1575 mnt_drop_write_file(file_out
);
1579 EXPORT_SYMBOL(vfs_copy_file_range
);
1581 SYSCALL_DEFINE6(copy_file_range
, int, fd_in
, loff_t __user
*, off_in
,
1582 int, fd_out
, loff_t __user
*, off_out
,
1583 size_t, len
, unsigned int, flags
)
1589 ssize_t ret
= -EBADF
;
1591 f_in
= fdget(fd_in
);
1595 f_out
= fdget(fd_out
);
1601 if (copy_from_user(&pos_in
, off_in
, sizeof(loff_t
)))
1604 pos_in
= f_in
.file
->f_pos
;
1608 if (copy_from_user(&pos_out
, off_out
, sizeof(loff_t
)))
1611 pos_out
= f_out
.file
->f_pos
;
1614 ret
= vfs_copy_file_range(f_in
.file
, pos_in
, f_out
.file
, pos_out
, len
,
1621 if (copy_to_user(off_in
, &pos_in
, sizeof(loff_t
)))
1624 f_in
.file
->f_pos
= pos_in
;
1628 if (copy_to_user(off_out
, &pos_out
, sizeof(loff_t
)))
1631 f_out
.file
->f_pos
= pos_out
;
1643 static int clone_verify_area(struct file
*file
, loff_t pos
, u64 len
, bool write
)
1645 struct inode
*inode
= file_inode(file
);
1647 if (unlikely(pos
< 0))
1650 if (unlikely((loff_t
) (pos
+ len
) < 0))
1653 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
1654 loff_t end
= len
? pos
+ len
- 1 : OFFSET_MAX
;
1657 retval
= locks_mandatory_area(inode
, file
, pos
, end
,
1658 write
? F_WRLCK
: F_RDLCK
);
1663 return security_file_permission(file
, write
? MAY_WRITE
: MAY_READ
);
1666 int vfs_clone_file_range(struct file
*file_in
, loff_t pos_in
,
1667 struct file
*file_out
, loff_t pos_out
, u64 len
)
1669 struct inode
*inode_in
= file_inode(file_in
);
1670 struct inode
*inode_out
= file_inode(file_out
);
1673 if (inode_in
->i_sb
!= inode_out
->i_sb
||
1674 file_in
->f_path
.mnt
!= file_out
->f_path
.mnt
)
1677 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1679 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1682 if (!(file_in
->f_mode
& FMODE_READ
) ||
1683 !(file_out
->f_mode
& FMODE_WRITE
) ||
1684 (file_out
->f_flags
& O_APPEND
))
1687 if (!file_in
->f_op
->clone_file_range
)
1690 ret
= clone_verify_area(file_in
, pos_in
, len
, false);
1694 ret
= clone_verify_area(file_out
, pos_out
, len
, true);
1698 if (pos_in
+ len
> i_size_read(inode_in
))
1701 ret
= mnt_want_write_file(file_out
);
1705 ret
= file_in
->f_op
->clone_file_range(file_in
, pos_in
,
1706 file_out
, pos_out
, len
);
1708 fsnotify_access(file_in
);
1709 fsnotify_modify(file_out
);
1712 mnt_drop_write_file(file_out
);
1715 EXPORT_SYMBOL(vfs_clone_file_range
);
1717 int vfs_dedupe_file_range(struct file
*file
, struct file_dedupe_range
*same
)
1719 struct file_dedupe_range_info
*info
;
1720 struct inode
*src
= file_inode(file
);
1725 bool is_admin
= capable(CAP_SYS_ADMIN
);
1726 u16 count
= same
->dest_count
;
1727 struct file
*dst_file
;
1731 if (!(file
->f_mode
& FMODE_READ
))
1734 if (same
->reserved1
|| same
->reserved2
)
1737 off
= same
->src_offset
;
1738 len
= same
->src_length
;
1741 if (S_ISDIR(src
->i_mode
))
1745 if (!S_ISREG(src
->i_mode
))
1748 ret
= clone_verify_area(file
, off
, len
, false);
1753 /* pre-format output fields to sane values */
1754 for (i
= 0; i
< count
; i
++) {
1755 same
->info
[i
].bytes_deduped
= 0ULL;
1756 same
->info
[i
].status
= FILE_DEDUPE_RANGE_SAME
;
1759 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
1761 struct fd dst_fd
= fdget(info
->dest_fd
);
1763 dst_file
= dst_fd
.file
;
1765 info
->status
= -EBADF
;
1768 dst
= file_inode(dst_file
);
1770 ret
= mnt_want_write_file(dst_file
);
1776 dst_off
= info
->dest_offset
;
1777 ret
= clone_verify_area(dst_file
, dst_off
, len
, true);
1784 if (info
->reserved
) {
1785 info
->status
= -EINVAL
;
1786 } else if (!(is_admin
|| (dst_file
->f_mode
& FMODE_WRITE
))) {
1787 info
->status
= -EINVAL
;
1788 } else if (file
->f_path
.mnt
!= dst_file
->f_path
.mnt
) {
1789 info
->status
= -EXDEV
;
1790 } else if (S_ISDIR(dst
->i_mode
)) {
1791 info
->status
= -EISDIR
;
1792 } else if (dst_file
->f_op
->dedupe_file_range
== NULL
) {
1793 info
->status
= -EINVAL
;
1795 deduped
= dst_file
->f_op
->dedupe_file_range(file
, off
,
1798 if (deduped
== -EBADE
)
1799 info
->status
= FILE_DEDUPE_RANGE_DIFFERS
;
1800 else if (deduped
< 0)
1801 info
->status
= deduped
;
1803 info
->bytes_deduped
+= deduped
;
1807 mnt_drop_write_file(dst_file
);
1811 if (fatal_signal_pending(current
))
1818 EXPORT_SYMBOL(vfs_dedupe_file_range
);