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>
21 #include <asm/uaccess.h>
22 #include <asm/unistd.h>
24 typedef ssize_t (*io_fn_t
)(struct file
*, char __user
*, size_t, loff_t
*);
25 typedef ssize_t (*iter_fn_t
)(struct kiocb
*, struct iov_iter
*);
27 const struct file_operations generic_ro_fops
= {
28 .llseek
= generic_file_llseek
,
29 .read_iter
= generic_file_read_iter
,
30 .mmap
= generic_file_readonly_mmap
,
31 .splice_read
= generic_file_splice_read
,
34 EXPORT_SYMBOL(generic_ro_fops
);
36 static inline int unsigned_offsets(struct file
*file
)
38 return file
->f_mode
& FMODE_UNSIGNED_OFFSET
;
42 * vfs_setpos - update the file offset for lseek
43 * @file: file structure in question
44 * @offset: file offset to seek to
45 * @maxsize: maximum file size
47 * This is a low-level filesystem helper for updating the file offset to
48 * the value specified by @offset if the given offset is valid and it is
49 * not equal to the current file offset.
51 * Return the specified offset on success and -EINVAL on invalid offset.
53 loff_t
vfs_setpos(struct file
*file
, loff_t offset
, loff_t maxsize
)
55 if (offset
< 0 && !unsigned_offsets(file
))
60 if (offset
!= file
->f_pos
) {
66 EXPORT_SYMBOL(vfs_setpos
);
69 * generic_file_llseek_size - generic llseek implementation for regular files
70 * @file: file structure to seek on
71 * @offset: file offset to seek to
72 * @whence: type of seek
73 * @size: max size of this file in file system
74 * @eof: offset used for SEEK_END position
76 * This is a variant of generic_file_llseek that allows passing in a custom
77 * maximum file size and a custom EOF position, for e.g. hashed directories
80 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
81 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
82 * read/writes behave like SEEK_SET against seeks.
85 generic_file_llseek_size(struct file
*file
, loff_t offset
, int whence
,
86 loff_t maxsize
, loff_t eof
)
94 * Here we special-case the lseek(fd, 0, SEEK_CUR)
95 * position-querying operation. Avoid rewriting the "same"
96 * f_pos value back to the file because a concurrent read(),
97 * write() or lseek() might have altered it
102 * f_lock protects against read/modify/write race with other
103 * SEEK_CURs. Note that parallel writes and reads behave
106 spin_lock(&file
->f_lock
);
107 offset
= vfs_setpos(file
, file
->f_pos
+ offset
, maxsize
);
108 spin_unlock(&file
->f_lock
);
112 * In the generic case the entire file is data, so as long as
113 * offset isn't at the end of the file then the offset is data.
120 * There is a virtual hole at the end of the file, so as long as
121 * offset isn't i_size or larger, return i_size.
129 return vfs_setpos(file
, offset
, maxsize
);
131 EXPORT_SYMBOL(generic_file_llseek_size
);
134 * generic_file_llseek - generic llseek implementation for regular files
135 * @file: file structure to seek on
136 * @offset: file offset to seek to
137 * @whence: type of seek
139 * This is a generic implemenation of ->llseek useable for all normal local
140 * filesystems. It just updates the file offset to the value specified by
141 * @offset and @whence.
143 loff_t
generic_file_llseek(struct file
*file
, loff_t offset
, int whence
)
145 struct inode
*inode
= file
->f_mapping
->host
;
147 return generic_file_llseek_size(file
, offset
, whence
,
148 inode
->i_sb
->s_maxbytes
,
151 EXPORT_SYMBOL(generic_file_llseek
);
154 * fixed_size_llseek - llseek implementation for fixed-sized devices
155 * @file: file structure to seek on
156 * @offset: file offset to seek to
157 * @whence: type of seek
158 * @size: size of the file
161 loff_t
fixed_size_llseek(struct file
*file
, loff_t offset
, int whence
, loff_t size
)
164 case SEEK_SET
: case SEEK_CUR
: case SEEK_END
:
165 return generic_file_llseek_size(file
, offset
, whence
,
171 EXPORT_SYMBOL(fixed_size_llseek
);
174 * noop_llseek - No Operation Performed llseek implementation
175 * @file: file structure to seek on
176 * @offset: file offset to seek to
177 * @whence: type of seek
179 * This is an implementation of ->llseek useable for the rare special case when
180 * userspace expects the seek to succeed but the (device) file is actually not
181 * able to perform the seek. In this case you use noop_llseek() instead of
182 * falling back to the default implementation of ->llseek.
184 loff_t
noop_llseek(struct file
*file
, loff_t offset
, int whence
)
188 EXPORT_SYMBOL(noop_llseek
);
190 loff_t
no_llseek(struct file
*file
, loff_t offset
, int whence
)
194 EXPORT_SYMBOL(no_llseek
);
196 loff_t
default_llseek(struct file
*file
, loff_t offset
, int whence
)
198 struct inode
*inode
= file_inode(file
);
201 mutex_lock(&inode
->i_mutex
);
204 offset
+= i_size_read(inode
);
208 retval
= file
->f_pos
;
211 offset
+= file
->f_pos
;
215 * In the generic case the entire file is data, so as
216 * long as offset isn't at the end of the file then the
219 if (offset
>= inode
->i_size
) {
226 * There is a virtual hole at the end of the file, so
227 * as long as offset isn't i_size or larger, return
230 if (offset
>= inode
->i_size
) {
234 offset
= inode
->i_size
;
238 if (offset
>= 0 || unsigned_offsets(file
)) {
239 if (offset
!= file
->f_pos
) {
240 file
->f_pos
= offset
;
246 mutex_unlock(&inode
->i_mutex
);
249 EXPORT_SYMBOL(default_llseek
);
251 loff_t
vfs_llseek(struct file
*file
, loff_t offset
, int whence
)
253 loff_t (*fn
)(struct file
*, loff_t
, int);
256 if (file
->f_mode
& FMODE_LSEEK
) {
257 if (file
->f_op
->llseek
)
258 fn
= file
->f_op
->llseek
;
260 return fn(file
, offset
, whence
);
262 EXPORT_SYMBOL(vfs_llseek
);
264 static inline struct fd
fdget_pos(int fd
)
266 return __to_fd(__fdget_pos(fd
));
269 static inline void fdput_pos(struct fd f
)
271 if (f
.flags
& FDPUT_POS_UNLOCK
)
272 mutex_unlock(&f
.file
->f_pos_lock
);
276 SYSCALL_DEFINE3(lseek
, unsigned int, fd
, off_t
, offset
, unsigned int, whence
)
279 struct fd f
= fdget_pos(fd
);
284 if (whence
<= SEEK_MAX
) {
285 loff_t res
= vfs_llseek(f
.file
, offset
, whence
);
287 if (res
!= (loff_t
)retval
)
288 retval
= -EOVERFLOW
; /* LFS: should only happen on 32 bit platforms */
295 COMPAT_SYSCALL_DEFINE3(lseek
, unsigned int, fd
, compat_off_t
, offset
, unsigned int, whence
)
297 return sys_lseek(fd
, offset
, whence
);
301 #ifdef __ARCH_WANT_SYS_LLSEEK
302 SYSCALL_DEFINE5(llseek
, unsigned int, fd
, unsigned long, offset_high
,
303 unsigned long, offset_low
, loff_t __user
*, result
,
304 unsigned int, whence
)
307 struct fd f
= fdget_pos(fd
);
314 if (whence
> SEEK_MAX
)
317 offset
= vfs_llseek(f
.file
, ((loff_t
) offset_high
<< 32) | offset_low
,
320 retval
= (int)offset
;
323 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
332 ssize_t
vfs_iter_read(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
)
337 if (!file
->f_op
->read_iter
)
340 init_sync_kiocb(&kiocb
, file
);
341 kiocb
.ki_pos
= *ppos
;
344 ret
= file
->f_op
->read_iter(&kiocb
, iter
);
345 BUG_ON(ret
== -EIOCBQUEUED
);
347 *ppos
= kiocb
.ki_pos
;
350 EXPORT_SYMBOL(vfs_iter_read
);
352 ssize_t
vfs_iter_write(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
)
357 if (!file
->f_op
->write_iter
)
360 init_sync_kiocb(&kiocb
, file
);
361 kiocb
.ki_pos
= *ppos
;
364 ret
= file
->f_op
->write_iter(&kiocb
, iter
);
365 BUG_ON(ret
== -EIOCBQUEUED
);
367 *ppos
= kiocb
.ki_pos
;
370 EXPORT_SYMBOL(vfs_iter_write
);
373 * rw_verify_area doesn't like huge counts. We limit
374 * them to something that fits in "int" so that others
375 * won't have to do range checks all the time.
377 int rw_verify_area(int read_write
, struct file
*file
, const loff_t
*ppos
, size_t count
)
381 int retval
= -EINVAL
;
383 inode
= file_inode(file
);
384 if (unlikely((ssize_t
) count
< 0))
387 if (unlikely(pos
< 0)) {
388 if (!unsigned_offsets(file
))
390 if (count
>= -pos
) /* both values are in 0..LLONG_MAX */
392 } else if (unlikely((loff_t
) (pos
+ count
) < 0)) {
393 if (!unsigned_offsets(file
))
397 if (unlikely(inode
->i_flctx
&& mandatory_lock(inode
))) {
398 retval
= locks_mandatory_area(
399 read_write
== READ
? FLOCK_VERIFY_READ
: FLOCK_VERIFY_WRITE
,
400 inode
, file
, pos
, count
);
404 retval
= security_file_permission(file
,
405 read_write
== READ
? MAY_READ
: MAY_WRITE
);
408 return count
> MAX_RW_COUNT
? MAX_RW_COUNT
: count
;
411 static ssize_t
new_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
413 struct iovec iov
= { .iov_base
= buf
, .iov_len
= len
};
415 struct iov_iter iter
;
418 init_sync_kiocb(&kiocb
, filp
);
419 kiocb
.ki_pos
= *ppos
;
420 iov_iter_init(&iter
, READ
, &iov
, 1, len
);
422 ret
= filp
->f_op
->read_iter(&kiocb
, &iter
);
423 BUG_ON(ret
== -EIOCBQUEUED
);
424 *ppos
= kiocb
.ki_pos
;
428 ssize_t
__vfs_read(struct file
*file
, char __user
*buf
, size_t count
,
431 if (file
->f_op
->read
)
432 return file
->f_op
->read(file
, buf
, count
, pos
);
433 else if (file
->f_op
->read_iter
)
434 return new_sync_read(file
, buf
, count
, pos
);
438 EXPORT_SYMBOL(__vfs_read
);
440 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
444 if (!(file
->f_mode
& FMODE_READ
))
446 if (!(file
->f_mode
& FMODE_CAN_READ
))
448 if (unlikely(!access_ok(VERIFY_WRITE
, buf
, count
)))
451 ret
= rw_verify_area(READ
, file
, pos
, count
);
454 ret
= __vfs_read(file
, buf
, count
, pos
);
456 fsnotify_access(file
);
457 add_rchar(current
, ret
);
465 EXPORT_SYMBOL(vfs_read
);
467 static ssize_t
new_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
469 struct iovec iov
= { .iov_base
= (void __user
*)buf
, .iov_len
= len
};
471 struct iov_iter iter
;
474 init_sync_kiocb(&kiocb
, filp
);
475 kiocb
.ki_pos
= *ppos
;
476 iov_iter_init(&iter
, WRITE
, &iov
, 1, len
);
478 ret
= filp
->f_op
->write_iter(&kiocb
, &iter
);
479 BUG_ON(ret
== -EIOCBQUEUED
);
480 *ppos
= kiocb
.ki_pos
;
484 ssize_t
__vfs_write(struct file
*file
, const char __user
*p
, size_t count
,
487 if (file
->f_op
->write
)
488 return file
->f_op
->write(file
, p
, count
, pos
);
489 else if (file
->f_op
->write_iter
)
490 return new_sync_write(file
, p
, count
, pos
);
494 EXPORT_SYMBOL(__vfs_write
);
496 ssize_t
__kernel_write(struct file
*file
, const char *buf
, size_t count
, loff_t
*pos
)
499 const char __user
*p
;
502 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
507 p
= (__force
const char __user
*)buf
;
508 if (count
> MAX_RW_COUNT
)
509 count
= MAX_RW_COUNT
;
510 ret
= __vfs_write(file
, p
, count
, pos
);
513 fsnotify_modify(file
);
514 add_wchar(current
, ret
);
520 EXPORT_SYMBOL(__kernel_write
);
522 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
526 if (!(file
->f_mode
& FMODE_WRITE
))
528 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
530 if (unlikely(!access_ok(VERIFY_READ
, buf
, count
)))
533 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
536 file_start_write(file
);
537 ret
= __vfs_write(file
, buf
, count
, pos
);
539 fsnotify_modify(file
);
540 add_wchar(current
, ret
);
543 file_end_write(file
);
549 EXPORT_SYMBOL(vfs_write
);
551 static inline loff_t
file_pos_read(struct file
*file
)
556 static inline void file_pos_write(struct file
*file
, loff_t pos
)
561 SYSCALL_DEFINE3(read
, unsigned int, fd
, char __user
*, buf
, size_t, count
)
563 struct fd f
= fdget_pos(fd
);
564 ssize_t ret
= -EBADF
;
567 loff_t pos
= file_pos_read(f
.file
);
568 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
570 file_pos_write(f
.file
, pos
);
576 SYSCALL_DEFINE3(write
, unsigned int, fd
, const char __user
*, buf
,
579 struct fd f
= fdget_pos(fd
);
580 ssize_t ret
= -EBADF
;
583 loff_t pos
= file_pos_read(f
.file
);
584 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
586 file_pos_write(f
.file
, pos
);
593 SYSCALL_DEFINE4(pread64
, unsigned int, fd
, char __user
*, buf
,
594 size_t, count
, loff_t
, pos
)
597 ssize_t ret
= -EBADF
;
605 if (f
.file
->f_mode
& FMODE_PREAD
)
606 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
613 SYSCALL_DEFINE4(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
614 size_t, count
, loff_t
, pos
)
617 ssize_t ret
= -EBADF
;
625 if (f
.file
->f_mode
& FMODE_PWRITE
)
626 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
634 * Reduce an iovec's length in-place. Return the resulting number of segments
636 unsigned long iov_shorten(struct iovec
*iov
, unsigned long nr_segs
, size_t to
)
638 unsigned long seg
= 0;
641 while (seg
< nr_segs
) {
643 if (len
+ iov
->iov_len
>= to
) {
644 iov
->iov_len
= to
- len
;
652 EXPORT_SYMBOL(iov_shorten
);
654 static ssize_t
do_iter_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
655 loff_t
*ppos
, iter_fn_t fn
)
660 init_sync_kiocb(&kiocb
, filp
);
661 kiocb
.ki_pos
= *ppos
;
663 ret
= fn(&kiocb
, iter
);
664 BUG_ON(ret
== -EIOCBQUEUED
);
665 *ppos
= kiocb
.ki_pos
;
669 /* Do it by hand, with file-ops */
670 static ssize_t
do_loop_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
671 loff_t
*ppos
, io_fn_t fn
)
675 while (iov_iter_count(iter
)) {
676 struct iovec iovec
= iov_iter_iovec(iter
);
679 nr
= fn(filp
, iovec
.iov_base
, iovec
.iov_len
, ppos
);
687 if (nr
!= iovec
.iov_len
)
689 iov_iter_advance(iter
, nr
);
695 /* A write operation does a read from user space and vice versa */
696 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
698 ssize_t
rw_copy_check_uvector(int type
, const struct iovec __user
* uvector
,
699 unsigned long nr_segs
, unsigned long fast_segs
,
700 struct iovec
*fast_pointer
,
701 struct iovec
**ret_pointer
)
705 struct iovec
*iov
= fast_pointer
;
708 * SuS says "The readv() function *may* fail if the iovcnt argument
709 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
710 * traditionally returned zero for zero segments, so...
718 * First get the "struct iovec" from user memory and
719 * verify all the pointers
721 if (nr_segs
> UIO_MAXIOV
) {
725 if (nr_segs
> fast_segs
) {
726 iov
= kmalloc(nr_segs
*sizeof(struct iovec
), GFP_KERNEL
);
732 if (copy_from_user(iov
, uvector
, nr_segs
*sizeof(*uvector
))) {
738 * According to the Single Unix Specification we should return EINVAL
739 * if an element length is < 0 when cast to ssize_t or if the
740 * total length would overflow the ssize_t return value of the
743 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
747 for (seg
= 0; seg
< nr_segs
; seg
++) {
748 void __user
*buf
= iov
[seg
].iov_base
;
749 ssize_t len
= (ssize_t
)iov
[seg
].iov_len
;
751 /* see if we we're about to use an invalid len or if
752 * it's about to overflow ssize_t */
758 && unlikely(!access_ok(vrfy_dir(type
), buf
, len
))) {
762 if (len
> MAX_RW_COUNT
- ret
) {
763 len
= MAX_RW_COUNT
- ret
;
764 iov
[seg
].iov_len
= len
;
773 static ssize_t
do_readv_writev(int type
, struct file
*file
,
774 const struct iovec __user
* uvector
,
775 unsigned long nr_segs
, loff_t
*pos
)
778 struct iovec iovstack
[UIO_FASTIOV
];
779 struct iovec
*iov
= iovstack
;
780 struct iov_iter iter
;
785 ret
= import_iovec(type
, uvector
, nr_segs
,
786 ARRAY_SIZE(iovstack
), &iov
, &iter
);
790 tot_len
= iov_iter_count(&iter
);
793 ret
= rw_verify_area(type
, file
, pos
, tot_len
);
798 fn
= file
->f_op
->read
;
799 iter_fn
= file
->f_op
->read_iter
;
801 fn
= (io_fn_t
)file
->f_op
->write
;
802 iter_fn
= file
->f_op
->write_iter
;
803 file_start_write(file
);
807 ret
= do_iter_readv_writev(file
, &iter
, pos
, iter_fn
);
809 ret
= do_loop_readv_writev(file
, &iter
, pos
, fn
);
812 file_end_write(file
);
816 if ((ret
+ (type
== READ
)) > 0) {
818 fsnotify_access(file
);
820 fsnotify_modify(file
);
825 ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
826 unsigned long vlen
, loff_t
*pos
)
828 if (!(file
->f_mode
& FMODE_READ
))
830 if (!(file
->f_mode
& FMODE_CAN_READ
))
833 return do_readv_writev(READ
, file
, vec
, vlen
, pos
);
836 EXPORT_SYMBOL(vfs_readv
);
838 ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
839 unsigned long vlen
, loff_t
*pos
)
841 if (!(file
->f_mode
& FMODE_WRITE
))
843 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
846 return do_readv_writev(WRITE
, file
, vec
, vlen
, pos
);
849 EXPORT_SYMBOL(vfs_writev
);
851 SYSCALL_DEFINE3(readv
, unsigned long, fd
, const struct iovec __user
*, vec
,
854 struct fd f
= fdget_pos(fd
);
855 ssize_t ret
= -EBADF
;
858 loff_t pos
= file_pos_read(f
.file
);
859 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
);
861 file_pos_write(f
.file
, pos
);
866 add_rchar(current
, ret
);
871 SYSCALL_DEFINE3(writev
, unsigned long, fd
, const struct iovec __user
*, vec
,
874 struct fd f
= fdget_pos(fd
);
875 ssize_t ret
= -EBADF
;
878 loff_t pos
= file_pos_read(f
.file
);
879 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
);
881 file_pos_write(f
.file
, pos
);
886 add_wchar(current
, ret
);
891 static inline loff_t
pos_from_hilo(unsigned long high
, unsigned long low
)
893 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
894 return (((loff_t
)high
<< HALF_LONG_BITS
) << HALF_LONG_BITS
) | low
;
897 SYSCALL_DEFINE5(preadv
, unsigned long, fd
, const struct iovec __user
*, vec
,
898 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
900 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
902 ssize_t ret
= -EBADF
;
910 if (f
.file
->f_mode
& FMODE_PREAD
)
911 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
);
916 add_rchar(current
, ret
);
921 SYSCALL_DEFINE5(pwritev
, unsigned long, fd
, const struct iovec __user
*, vec
,
922 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
924 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
926 ssize_t ret
= -EBADF
;
934 if (f
.file
->f_mode
& FMODE_PWRITE
)
935 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
);
940 add_wchar(current
, ret
);
947 static ssize_t
compat_do_readv_writev(int type
, struct file
*file
,
948 const struct compat_iovec __user
*uvector
,
949 unsigned long nr_segs
, loff_t
*pos
)
951 compat_ssize_t tot_len
;
952 struct iovec iovstack
[UIO_FASTIOV
];
953 struct iovec
*iov
= iovstack
;
954 struct iov_iter iter
;
959 ret
= compat_import_iovec(type
, uvector
, nr_segs
,
960 UIO_FASTIOV
, &iov
, &iter
);
964 tot_len
= iov_iter_count(&iter
);
967 ret
= rw_verify_area(type
, file
, pos
, tot_len
);
972 fn
= file
->f_op
->read
;
973 iter_fn
= file
->f_op
->read_iter
;
975 fn
= (io_fn_t
)file
->f_op
->write
;
976 iter_fn
= file
->f_op
->write_iter
;
977 file_start_write(file
);
981 ret
= do_iter_readv_writev(file
, &iter
, pos
, iter_fn
);
983 ret
= do_loop_readv_writev(file
, &iter
, pos
, fn
);
986 file_end_write(file
);
990 if ((ret
+ (type
== READ
)) > 0) {
992 fsnotify_access(file
);
994 fsnotify_modify(file
);
999 static size_t compat_readv(struct file
*file
,
1000 const struct compat_iovec __user
*vec
,
1001 unsigned long vlen
, loff_t
*pos
)
1003 ssize_t ret
= -EBADF
;
1005 if (!(file
->f_mode
& FMODE_READ
))
1009 if (!(file
->f_mode
& FMODE_CAN_READ
))
1012 ret
= compat_do_readv_writev(READ
, file
, vec
, vlen
, pos
);
1016 add_rchar(current
, ret
);
1021 COMPAT_SYSCALL_DEFINE3(readv
, compat_ulong_t
, fd
,
1022 const struct compat_iovec __user
*,vec
,
1023 compat_ulong_t
, vlen
)
1025 struct fd f
= fdget_pos(fd
);
1031 pos
= f
.file
->f_pos
;
1032 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
);
1034 f
.file
->f_pos
= pos
;
1039 static long __compat_sys_preadv64(unsigned long fd
,
1040 const struct compat_iovec __user
*vec
,
1041 unsigned long vlen
, loff_t pos
)
1052 if (f
.file
->f_mode
& FMODE_PREAD
)
1053 ret
= compat_readv(f
.file
, vec
, vlen
, &pos
);
1058 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1059 COMPAT_SYSCALL_DEFINE4(preadv64
, unsigned long, fd
,
1060 const struct compat_iovec __user
*,vec
,
1061 unsigned long, vlen
, loff_t
, pos
)
1063 return __compat_sys_preadv64(fd
, vec
, vlen
, pos
);
1067 COMPAT_SYSCALL_DEFINE5(preadv
, compat_ulong_t
, fd
,
1068 const struct compat_iovec __user
*,vec
,
1069 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1071 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1073 return __compat_sys_preadv64(fd
, vec
, vlen
, pos
);
1076 static size_t compat_writev(struct file
*file
,
1077 const struct compat_iovec __user
*vec
,
1078 unsigned long vlen
, loff_t
*pos
)
1080 ssize_t ret
= -EBADF
;
1082 if (!(file
->f_mode
& FMODE_WRITE
))
1086 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
1089 ret
= compat_do_readv_writev(WRITE
, file
, vec
, vlen
, pos
);
1093 add_wchar(current
, ret
);
1098 COMPAT_SYSCALL_DEFINE3(writev
, compat_ulong_t
, fd
,
1099 const struct compat_iovec __user
*, vec
,
1100 compat_ulong_t
, vlen
)
1102 struct fd f
= fdget_pos(fd
);
1108 pos
= f
.file
->f_pos
;
1109 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
);
1111 f
.file
->f_pos
= pos
;
1116 static long __compat_sys_pwritev64(unsigned long fd
,
1117 const struct compat_iovec __user
*vec
,
1118 unsigned long vlen
, loff_t pos
)
1129 if (f
.file
->f_mode
& FMODE_PWRITE
)
1130 ret
= compat_writev(f
.file
, vec
, vlen
, &pos
);
1135 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1136 COMPAT_SYSCALL_DEFINE4(pwritev64
, unsigned long, fd
,
1137 const struct compat_iovec __user
*,vec
,
1138 unsigned long, vlen
, loff_t
, pos
)
1140 return __compat_sys_pwritev64(fd
, vec
, vlen
, pos
);
1144 COMPAT_SYSCALL_DEFINE5(pwritev
, compat_ulong_t
, fd
,
1145 const struct compat_iovec __user
*,vec
,
1146 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1148 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1150 return __compat_sys_pwritev64(fd
, vec
, vlen
, pos
);
1154 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
1155 size_t count
, loff_t max
)
1158 struct inode
*in_inode
, *out_inode
;
1165 * Get input file, and verify that it is ok..
1171 if (!(in
.file
->f_mode
& FMODE_READ
))
1175 pos
= in
.file
->f_pos
;
1178 if (!(in
.file
->f_mode
& FMODE_PREAD
))
1181 retval
= rw_verify_area(READ
, in
.file
, &pos
, count
);
1187 * Get output file, and verify that it is ok..
1190 out
= fdget(out_fd
);
1193 if (!(out
.file
->f_mode
& FMODE_WRITE
))
1196 in_inode
= file_inode(in
.file
);
1197 out_inode
= file_inode(out
.file
);
1198 out_pos
= out
.file
->f_pos
;
1199 retval
= rw_verify_area(WRITE
, out
.file
, &out_pos
, count
);
1205 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
1207 if (unlikely(pos
+ count
> max
)) {
1208 retval
= -EOVERFLOW
;
1217 * We need to debate whether we can enable this or not. The
1218 * man page documents EAGAIN return for the output at least,
1219 * and the application is arguably buggy if it doesn't expect
1220 * EAGAIN on a non-blocking file descriptor.
1222 if (in
.file
->f_flags
& O_NONBLOCK
)
1223 fl
= SPLICE_F_NONBLOCK
;
1225 file_start_write(out
.file
);
1226 retval
= do_splice_direct(in
.file
, &pos
, out
.file
, &out_pos
, count
, fl
);
1227 file_end_write(out
.file
);
1230 add_rchar(current
, retval
);
1231 add_wchar(current
, retval
);
1232 fsnotify_access(in
.file
);
1233 fsnotify_modify(out
.file
);
1234 out
.file
->f_pos
= out_pos
;
1238 in
.file
->f_pos
= pos
;
1244 retval
= -EOVERFLOW
;
1254 SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
, off_t __user
*, offset
, size_t, count
)
1261 if (unlikely(get_user(off
, offset
)))
1264 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1265 if (unlikely(put_user(pos
, offset
)))
1270 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1273 SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
, loff_t __user
*, offset
, size_t, count
)
1279 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1281 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1282 if (unlikely(put_user(pos
, offset
)))
1287 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1290 #ifdef CONFIG_COMPAT
1291 COMPAT_SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
,
1292 compat_off_t __user
*, offset
, compat_size_t
, count
)
1299 if (unlikely(get_user(off
, offset
)))
1302 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1303 if (unlikely(put_user(pos
, offset
)))
1308 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1311 COMPAT_SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
,
1312 compat_loff_t __user
*, offset
, compat_size_t
, count
)
1318 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1320 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1321 if (unlikely(put_user(pos
, offset
)))
1326 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);