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/smp_lock.h>
13 #include <linux/fsnotify.h>
14 #include <linux/security.h>
15 #include <linux/module.h>
16 #include <linux/syscalls.h>
18 #include <asm/uaccess.h>
19 #include <asm/unistd.h>
21 struct file_operations generic_ro_fops
= {
22 .llseek
= generic_file_llseek
,
23 .read
= generic_file_read
,
24 .mmap
= generic_file_readonly_mmap
,
25 .sendfile
= generic_file_sendfile
,
28 EXPORT_SYMBOL(generic_ro_fops
);
30 loff_t
generic_file_llseek(struct file
*file
, loff_t offset
, int origin
)
33 struct inode
*inode
= file
->f_mapping
->host
;
38 offset
+= inode
->i_size
;
41 offset
+= file
->f_pos
;
44 if (offset
>=0 && offset
<=inode
->i_sb
->s_maxbytes
) {
45 if (offset
!= file
->f_pos
) {
55 EXPORT_SYMBOL(generic_file_llseek
);
57 loff_t
remote_llseek(struct file
*file
, loff_t offset
, int origin
)
64 offset
+= i_size_read(file
->f_dentry
->d_inode
);
67 offset
+= file
->f_pos
;
70 if (offset
>=0 && offset
<=file
->f_dentry
->d_inode
->i_sb
->s_maxbytes
) {
71 if (offset
!= file
->f_pos
) {
80 EXPORT_SYMBOL(remote_llseek
);
82 loff_t
no_llseek(struct file
*file
, loff_t offset
, int origin
)
86 EXPORT_SYMBOL(no_llseek
);
88 loff_t
default_llseek(struct file
*file
, loff_t offset
, int origin
)
95 offset
+= i_size_read(file
->f_dentry
->d_inode
);
98 offset
+= file
->f_pos
;
102 if (offset
!= file
->f_pos
) {
103 file
->f_pos
= offset
;
111 EXPORT_SYMBOL(default_llseek
);
113 loff_t
vfs_llseek(struct file
*file
, loff_t offset
, int origin
)
115 loff_t (*fn
)(struct file
*, loff_t
, int);
118 if (file
->f_mode
& FMODE_LSEEK
) {
120 if (file
->f_op
&& file
->f_op
->llseek
)
121 fn
= file
->f_op
->llseek
;
123 return fn(file
, offset
, origin
);
125 EXPORT_SYMBOL(vfs_llseek
);
127 asmlinkage off_t
sys_lseek(unsigned int fd
, off_t offset
, unsigned int origin
)
134 file
= fget_light(fd
, &fput_needed
);
140 loff_t res
= vfs_llseek(file
, offset
, origin
);
142 if (res
!= (loff_t
)retval
)
143 retval
= -EOVERFLOW
; /* LFS: should only happen on 32 bit platforms */
145 fput_light(file
, fput_needed
);
150 #ifdef __ARCH_WANT_SYS_LLSEEK
151 asmlinkage
long sys_llseek(unsigned int fd
, unsigned long offset_high
,
152 unsigned long offset_low
, loff_t __user
* result
,
161 file
= fget_light(fd
, &fput_needed
);
169 offset
= vfs_llseek(file
, ((loff_t
) offset_high
<< 32) | offset_low
,
172 retval
= (int)offset
;
175 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
179 fput_light(file
, fput_needed
);
186 int rw_verify_area(int read_write
, struct file
*file
, loff_t
*ppos
, size_t count
)
191 if (unlikely(count
> INT_MAX
))
194 if (unlikely((pos
< 0) || (loff_t
) (pos
+ count
) < 0))
197 inode
= file
->f_dentry
->d_inode
;
198 if (inode
->i_flock
&& MANDATORY_LOCK(inode
))
199 return locks_mandatory_area(read_write
== READ
? FLOCK_VERIFY_READ
: FLOCK_VERIFY_WRITE
, inode
, file
, pos
, count
);
206 static void wait_on_retry_sync_kiocb(struct kiocb
*iocb
)
208 set_current_state(TASK_UNINTERRUPTIBLE
);
209 if (!kiocbIsKicked(iocb
))
212 kiocbClearKicked(iocb
);
213 __set_current_state(TASK_RUNNING
);
216 ssize_t
do_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
221 init_sync_kiocb(&kiocb
, filp
);
222 kiocb
.ki_pos
= *ppos
;
223 while (-EIOCBRETRY
==
224 (ret
= filp
->f_op
->aio_read(&kiocb
, buf
, len
, kiocb
.ki_pos
)))
225 wait_on_retry_sync_kiocb(&kiocb
);
227 if (-EIOCBQUEUED
== ret
)
228 ret
= wait_on_sync_kiocb(&kiocb
);
229 *ppos
= kiocb
.ki_pos
;
233 EXPORT_SYMBOL(do_sync_read
);
235 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
239 if (!(file
->f_mode
& FMODE_READ
))
241 if (!file
->f_op
|| (!file
->f_op
->read
&& !file
->f_op
->aio_read
))
243 if (unlikely(!access_ok(VERIFY_WRITE
, buf
, count
)))
246 ret
= rw_verify_area(READ
, file
, pos
, count
);
248 ret
= security_file_permission (file
, MAY_READ
);
250 if (file
->f_op
->read
)
251 ret
= file
->f_op
->read(file
, buf
, count
, pos
);
253 ret
= do_sync_read(file
, buf
, count
, pos
);
255 fsnotify_access(file
->f_dentry
);
256 current
->rchar
+= ret
;
265 EXPORT_SYMBOL(vfs_read
);
267 ssize_t
do_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
272 init_sync_kiocb(&kiocb
, filp
);
273 kiocb
.ki_pos
= *ppos
;
274 while (-EIOCBRETRY
==
275 (ret
= filp
->f_op
->aio_write(&kiocb
, buf
, len
, kiocb
.ki_pos
)))
276 wait_on_retry_sync_kiocb(&kiocb
);
278 if (-EIOCBQUEUED
== ret
)
279 ret
= wait_on_sync_kiocb(&kiocb
);
280 *ppos
= kiocb
.ki_pos
;
284 EXPORT_SYMBOL(do_sync_write
);
286 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
290 if (!(file
->f_mode
& FMODE_WRITE
))
292 if (!file
->f_op
|| (!file
->f_op
->write
&& !file
->f_op
->aio_write
))
294 if (unlikely(!access_ok(VERIFY_READ
, buf
, count
)))
297 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
299 ret
= security_file_permission (file
, MAY_WRITE
);
301 if (file
->f_op
->write
)
302 ret
= file
->f_op
->write(file
, buf
, count
, pos
);
304 ret
= do_sync_write(file
, buf
, count
, pos
);
306 fsnotify_modify(file
->f_dentry
);
307 current
->wchar
+= ret
;
316 EXPORT_SYMBOL(vfs_write
);
318 static inline loff_t
file_pos_read(struct file
*file
)
323 static inline void file_pos_write(struct file
*file
, loff_t pos
)
328 asmlinkage ssize_t
sys_read(unsigned int fd
, char __user
* buf
, size_t count
)
331 ssize_t ret
= -EBADF
;
334 file
= fget_light(fd
, &fput_needed
);
336 loff_t pos
= file_pos_read(file
);
337 ret
= vfs_read(file
, buf
, count
, &pos
);
338 file_pos_write(file
, pos
);
339 fput_light(file
, fput_needed
);
344 EXPORT_SYMBOL_GPL(sys_read
);
346 asmlinkage ssize_t
sys_write(unsigned int fd
, const char __user
* buf
, size_t count
)
349 ssize_t ret
= -EBADF
;
352 file
= fget_light(fd
, &fput_needed
);
354 loff_t pos
= file_pos_read(file
);
355 ret
= vfs_write(file
, buf
, count
, &pos
);
356 file_pos_write(file
, pos
);
357 fput_light(file
, fput_needed
);
363 asmlinkage ssize_t
sys_pread64(unsigned int fd
, char __user
*buf
,
364 size_t count
, loff_t pos
)
367 ssize_t ret
= -EBADF
;
373 file
= fget_light(fd
, &fput_needed
);
376 if (file
->f_mode
& FMODE_PREAD
)
377 ret
= vfs_read(file
, buf
, count
, &pos
);
378 fput_light(file
, fput_needed
);
384 asmlinkage ssize_t
sys_pwrite64(unsigned int fd
, const char __user
*buf
,
385 size_t count
, loff_t pos
)
388 ssize_t ret
= -EBADF
;
394 file
= fget_light(fd
, &fput_needed
);
397 if (file
->f_mode
& FMODE_PWRITE
)
398 ret
= vfs_write(file
, buf
, count
, &pos
);
399 fput_light(file
, fput_needed
);
406 * Reduce an iovec's length in-place. Return the resulting number of segments
408 unsigned long iov_shorten(struct iovec
*iov
, unsigned long nr_segs
, size_t to
)
410 unsigned long seg
= 0;
413 while (seg
< nr_segs
) {
415 if (len
+ iov
->iov_len
>= to
) {
416 iov
->iov_len
= to
- len
;
425 EXPORT_SYMBOL(iov_shorten
);
427 /* A write operation does a read from user space and vice versa */
428 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
430 static ssize_t
do_readv_writev(int type
, struct file
*file
,
431 const struct iovec __user
* uvector
,
432 unsigned long nr_segs
, loff_t
*pos
)
434 typedef ssize_t (*io_fn_t
)(struct file
*, char __user
*, size_t, loff_t
*);
435 typedef ssize_t (*iov_fn_t
)(struct file
*, const struct iovec
*, unsigned long, loff_t
*);
438 struct iovec iovstack
[UIO_FASTIOV
];
439 struct iovec
*iov
=iovstack
, *vector
;
446 * SuS says "The readv() function *may* fail if the iovcnt argument
447 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
448 * traditionally returned zero for zero segments, so...
455 * First get the "struct iovec" from user memory and
456 * verify all the pointers
459 if ((nr_segs
> UIO_MAXIOV
) || (nr_segs
<= 0))
463 if (nr_segs
> UIO_FASTIOV
) {
465 iov
= kmalloc(nr_segs
*sizeof(struct iovec
), GFP_KERNEL
);
470 if (copy_from_user(iov
, uvector
, nr_segs
*sizeof(*uvector
)))
474 * Single unix specification:
475 * We should -EINVAL if an element length is not >= 0 and fitting an
476 * ssize_t. The total length is fitting an ssize_t
478 * Be careful here because iov_len is a size_t not an ssize_t
482 for (seg
= 0; seg
< nr_segs
; seg
++) {
483 void __user
*buf
= iov
[seg
].iov_base
;
484 ssize_t len
= (ssize_t
)iov
[seg
].iov_len
;
486 if (len
< 0) /* size_t not fitting an ssize_t .. */
488 if (unlikely(!access_ok(vrfy_dir(type
), buf
, len
)))
491 if ((ssize_t
)tot_len
< 0) /* maths overflow on the ssize_t */
499 ret
= rw_verify_area(type
, file
, pos
, tot_len
);
502 ret
= security_file_permission(file
, type
== READ
? MAY_READ
: MAY_WRITE
);
508 fn
= file
->f_op
->read
;
509 fnv
= file
->f_op
->readv
;
511 fn
= (io_fn_t
)file
->f_op
->write
;
512 fnv
= file
->f_op
->writev
;
515 ret
= fnv(file
, iov
, nr_segs
, pos
);
519 /* Do it by hand, with file-ops */
522 while (nr_segs
> 0) {
527 base
= vector
->iov_base
;
528 len
= vector
->iov_len
;
532 nr
= fn(file
, base
, len
, pos
);
545 if ((ret
+ (type
== READ
)) > 0) {
547 fsnotify_access(file
->f_dentry
);
549 fsnotify_modify(file
->f_dentry
);
557 ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
558 unsigned long vlen
, loff_t
*pos
)
560 if (!(file
->f_mode
& FMODE_READ
))
562 if (!file
->f_op
|| (!file
->f_op
->readv
&& !file
->f_op
->read
))
565 return do_readv_writev(READ
, file
, vec
, vlen
, pos
);
568 EXPORT_SYMBOL(vfs_readv
);
570 ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
571 unsigned long vlen
, loff_t
*pos
)
573 if (!(file
->f_mode
& FMODE_WRITE
))
575 if (!file
->f_op
|| (!file
->f_op
->writev
&& !file
->f_op
->write
))
578 return do_readv_writev(WRITE
, file
, vec
, vlen
, pos
);
581 EXPORT_SYMBOL(vfs_writev
);
584 sys_readv(unsigned long fd
, const struct iovec __user
*vec
, unsigned long vlen
)
587 ssize_t ret
= -EBADF
;
590 file
= fget_light(fd
, &fput_needed
);
592 loff_t pos
= file_pos_read(file
);
593 ret
= vfs_readv(file
, vec
, vlen
, &pos
);
594 file_pos_write(file
, pos
);
595 fput_light(file
, fput_needed
);
599 current
->rchar
+= ret
;
605 sys_writev(unsigned long fd
, const struct iovec __user
*vec
, unsigned long vlen
)
608 ssize_t ret
= -EBADF
;
611 file
= fget_light(fd
, &fput_needed
);
613 loff_t pos
= file_pos_read(file
);
614 ret
= vfs_writev(file
, vec
, vlen
, &pos
);
615 file_pos_write(file
, pos
);
616 fput_light(file
, fput_needed
);
620 current
->wchar
+= ret
;
625 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
626 size_t count
, loff_t max
)
628 struct file
* in_file
, * out_file
;
629 struct inode
* in_inode
, * out_inode
;
632 int fput_needed_in
, fput_needed_out
;
635 * Get input file, and verify that it is ok..
638 in_file
= fget_light(in_fd
, &fput_needed_in
);
641 if (!(in_file
->f_mode
& FMODE_READ
))
644 in_inode
= in_file
->f_dentry
->d_inode
;
647 if (!in_file
->f_op
|| !in_file
->f_op
->sendfile
)
651 ppos
= &in_file
->f_pos
;
653 if (!(in_file
->f_mode
& FMODE_PREAD
))
655 retval
= rw_verify_area(READ
, in_file
, ppos
, count
);
659 retval
= security_file_permission (in_file
, MAY_READ
);
664 * Get output file, and verify that it is ok..
667 out_file
= fget_light(out_fd
, &fput_needed_out
);
670 if (!(out_file
->f_mode
& FMODE_WRITE
))
673 if (!out_file
->f_op
|| !out_file
->f_op
->sendpage
)
675 out_inode
= out_file
->f_dentry
->d_inode
;
676 retval
= rw_verify_area(WRITE
, out_file
, &out_file
->f_pos
, count
);
680 retval
= security_file_permission (out_file
, MAY_WRITE
);
685 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
689 if (unlikely(pos
< 0))
691 if (unlikely(pos
+ count
> max
)) {
698 retval
= in_file
->f_op
->sendfile(in_file
, ppos
, count
, file_send_actor
, out_file
);
701 current
->rchar
+= retval
;
702 current
->wchar
+= retval
;
711 fput_light(out_file
, fput_needed_out
);
713 fput_light(in_file
, fput_needed_in
);
718 asmlinkage ssize_t
sys_sendfile(int out_fd
, int in_fd
, off_t __user
*offset
, size_t count
)
725 if (unlikely(get_user(off
, offset
)))
728 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
729 if (unlikely(put_user(pos
, offset
)))
734 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
737 asmlinkage ssize_t
sys_sendfile64(int out_fd
, int in_fd
, loff_t __user
*offset
, size_t count
)
743 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
745 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
746 if (unlikely(put_user(pos
, offset
)))
751 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);