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
= filemap_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 * @maxsize: 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
default_llseek(struct file
*file
, loff_t offset
, int whence
)
232 struct inode
*inode
= file_inode(file
);
238 offset
+= i_size_read(inode
);
242 retval
= file
->f_pos
;
245 offset
+= file
->f_pos
;
249 * In the generic case the entire file is data, so as
250 * long as offset isn't at the end of the file then the
253 if (offset
>= inode
->i_size
) {
260 * There is a virtual hole at the end of the file, so
261 * as long as offset isn't i_size or larger, return
264 if (offset
>= inode
->i_size
) {
268 offset
= inode
->i_size
;
272 if (offset
>= 0 || unsigned_offsets(file
)) {
273 if (offset
!= file
->f_pos
) {
274 file
->f_pos
= offset
;
283 EXPORT_SYMBOL(default_llseek
);
285 loff_t
vfs_llseek(struct file
*file
, loff_t offset
, int whence
)
287 if (!(file
->f_mode
& FMODE_LSEEK
))
289 return file
->f_op
->llseek(file
, offset
, whence
);
291 EXPORT_SYMBOL(vfs_llseek
);
293 static off_t
ksys_lseek(unsigned int fd
, off_t offset
, unsigned int whence
)
296 struct fd f
= fdget_pos(fd
);
301 if (whence
<= SEEK_MAX
) {
302 loff_t res
= vfs_llseek(f
.file
, offset
, whence
);
304 if (res
!= (loff_t
)retval
)
305 retval
= -EOVERFLOW
; /* LFS: should only happen on 32 bit platforms */
311 SYSCALL_DEFINE3(lseek
, unsigned int, fd
, off_t
, offset
, unsigned int, whence
)
313 return ksys_lseek(fd
, offset
, whence
);
317 COMPAT_SYSCALL_DEFINE3(lseek
, unsigned int, fd
, compat_off_t
, offset
, unsigned int, whence
)
319 return ksys_lseek(fd
, offset
, whence
);
323 #if !defined(CONFIG_64BIT) || defined(CONFIG_COMPAT) || \
324 defined(__ARCH_WANT_SYS_LLSEEK)
325 SYSCALL_DEFINE5(llseek
, unsigned int, fd
, unsigned long, offset_high
,
326 unsigned long, offset_low
, loff_t __user
*, result
,
327 unsigned int, whence
)
330 struct fd f
= fdget_pos(fd
);
337 if (whence
> SEEK_MAX
)
340 offset
= vfs_llseek(f
.file
, ((loff_t
) offset_high
<< 32) | offset_low
,
343 retval
= (int)offset
;
346 if (!copy_to_user(result
, &offset
, sizeof(offset
)))
355 int rw_verify_area(int read_write
, struct file
*file
, const loff_t
*ppos
, size_t count
)
357 if (unlikely((ssize_t
) count
< 0))
363 if (unlikely(pos
< 0)) {
364 if (!unsigned_offsets(file
))
366 if (count
>= -pos
) /* both values are in 0..LLONG_MAX */
368 } else if (unlikely((loff_t
) (pos
+ count
) < 0)) {
369 if (!unsigned_offsets(file
))
374 return security_file_permission(file
,
375 read_write
== READ
? MAY_READ
: MAY_WRITE
);
377 EXPORT_SYMBOL(rw_verify_area
);
379 static ssize_t
new_sync_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
382 struct iov_iter iter
;
385 init_sync_kiocb(&kiocb
, filp
);
386 kiocb
.ki_pos
= (ppos
? *ppos
: 0);
387 iov_iter_ubuf(&iter
, ITER_DEST
, buf
, len
);
389 ret
= call_read_iter(filp
, &kiocb
, &iter
);
390 BUG_ON(ret
== -EIOCBQUEUED
);
392 *ppos
= kiocb
.ki_pos
;
396 static int warn_unsupported(struct file
*file
, const char *op
)
399 "kernel %s not supported for file %pD4 (pid: %d comm: %.20s)\n",
400 op
, file
, current
->pid
, current
->comm
);
404 ssize_t
__kernel_read(struct file
*file
, void *buf
, size_t count
, loff_t
*pos
)
408 .iov_len
= min_t(size_t, count
, MAX_RW_COUNT
),
411 struct iov_iter iter
;
414 if (WARN_ON_ONCE(!(file
->f_mode
& FMODE_READ
)))
416 if (!(file
->f_mode
& FMODE_CAN_READ
))
419 * Also fail if ->read_iter and ->read are both wired up as that
420 * implies very convoluted semantics.
422 if (unlikely(!file
->f_op
->read_iter
|| file
->f_op
->read
))
423 return warn_unsupported(file
, "read");
425 init_sync_kiocb(&kiocb
, file
);
426 kiocb
.ki_pos
= pos
? *pos
: 0;
427 iov_iter_kvec(&iter
, ITER_DEST
, &iov
, 1, iov
.iov_len
);
428 ret
= file
->f_op
->read_iter(&kiocb
, &iter
);
432 fsnotify_access(file
);
433 add_rchar(current
, ret
);
439 ssize_t
kernel_read(struct file
*file
, void *buf
, size_t count
, loff_t
*pos
)
443 ret
= rw_verify_area(READ
, file
, pos
, count
);
446 return __kernel_read(file
, buf
, count
, pos
);
448 EXPORT_SYMBOL(kernel_read
);
450 ssize_t
vfs_read(struct file
*file
, char __user
*buf
, size_t count
, loff_t
*pos
)
454 if (!(file
->f_mode
& FMODE_READ
))
456 if (!(file
->f_mode
& FMODE_CAN_READ
))
458 if (unlikely(!access_ok(buf
, count
)))
461 ret
= rw_verify_area(READ
, file
, pos
, count
);
464 if (count
> MAX_RW_COUNT
)
465 count
= MAX_RW_COUNT
;
467 if (file
->f_op
->read
)
468 ret
= file
->f_op
->read(file
, buf
, count
, pos
);
469 else if (file
->f_op
->read_iter
)
470 ret
= new_sync_read(file
, buf
, count
, pos
);
474 fsnotify_access(file
);
475 add_rchar(current
, ret
);
481 static ssize_t
new_sync_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
484 struct iov_iter iter
;
487 init_sync_kiocb(&kiocb
, filp
);
488 kiocb
.ki_pos
= (ppos
? *ppos
: 0);
489 iov_iter_ubuf(&iter
, ITER_SOURCE
, (void __user
*)buf
, len
);
491 ret
= call_write_iter(filp
, &kiocb
, &iter
);
492 BUG_ON(ret
== -EIOCBQUEUED
);
494 *ppos
= kiocb
.ki_pos
;
498 /* caller is responsible for file_start_write/file_end_write */
499 ssize_t
__kernel_write_iter(struct file
*file
, struct iov_iter
*from
, loff_t
*pos
)
504 if (WARN_ON_ONCE(!(file
->f_mode
& FMODE_WRITE
)))
506 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
509 * Also fail if ->write_iter and ->write are both wired up as that
510 * implies very convoluted semantics.
512 if (unlikely(!file
->f_op
->write_iter
|| file
->f_op
->write
))
513 return warn_unsupported(file
, "write");
515 init_sync_kiocb(&kiocb
, file
);
516 kiocb
.ki_pos
= pos
? *pos
: 0;
517 ret
= file
->f_op
->write_iter(&kiocb
, from
);
521 fsnotify_modify(file
);
522 add_wchar(current
, ret
);
528 /* caller is responsible for file_start_write/file_end_write */
529 ssize_t
__kernel_write(struct file
*file
, const void *buf
, size_t count
, loff_t
*pos
)
532 .iov_base
= (void *)buf
,
533 .iov_len
= min_t(size_t, count
, MAX_RW_COUNT
),
535 struct iov_iter iter
;
536 iov_iter_kvec(&iter
, ITER_SOURCE
, &iov
, 1, iov
.iov_len
);
537 return __kernel_write_iter(file
, &iter
, pos
);
540 * This "EXPORT_SYMBOL_GPL()" is more of a "EXPORT_SYMBOL_DONTUSE()",
541 * but autofs is one of the few internal kernel users that actually
542 * wants this _and_ can be built as a module. So we need to export
543 * this symbol for autofs, even though it really isn't appropriate
544 * for any other kernel modules.
546 EXPORT_SYMBOL_GPL(__kernel_write
);
548 ssize_t
kernel_write(struct file
*file
, const void *buf
, size_t count
,
553 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
557 file_start_write(file
);
558 ret
= __kernel_write(file
, buf
, count
, pos
);
559 file_end_write(file
);
562 EXPORT_SYMBOL(kernel_write
);
564 ssize_t
vfs_write(struct file
*file
, const char __user
*buf
, size_t count
, loff_t
*pos
)
568 if (!(file
->f_mode
& FMODE_WRITE
))
570 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
572 if (unlikely(!access_ok(buf
, count
)))
575 ret
= rw_verify_area(WRITE
, file
, pos
, count
);
578 if (count
> MAX_RW_COUNT
)
579 count
= MAX_RW_COUNT
;
580 file_start_write(file
);
581 if (file
->f_op
->write
)
582 ret
= file
->f_op
->write(file
, buf
, count
, pos
);
583 else if (file
->f_op
->write_iter
)
584 ret
= new_sync_write(file
, buf
, count
, pos
);
588 fsnotify_modify(file
);
589 add_wchar(current
, ret
);
592 file_end_write(file
);
596 /* file_ppos returns &file->f_pos or NULL if file is stream */
597 static inline loff_t
*file_ppos(struct file
*file
)
599 return file
->f_mode
& FMODE_STREAM
? NULL
: &file
->f_pos
;
602 ssize_t
ksys_read(unsigned int fd
, char __user
*buf
, size_t count
)
604 struct fd f
= fdget_pos(fd
);
605 ssize_t ret
= -EBADF
;
608 loff_t pos
, *ppos
= file_ppos(f
.file
);
613 ret
= vfs_read(f
.file
, buf
, count
, ppos
);
614 if (ret
>= 0 && ppos
)
621 SYSCALL_DEFINE3(read
, unsigned int, fd
, char __user
*, buf
, size_t, count
)
623 return ksys_read(fd
, buf
, count
);
626 ssize_t
ksys_write(unsigned int fd
, const char __user
*buf
, size_t count
)
628 struct fd f
= fdget_pos(fd
);
629 ssize_t ret
= -EBADF
;
632 loff_t pos
, *ppos
= file_ppos(f
.file
);
637 ret
= vfs_write(f
.file
, buf
, count
, ppos
);
638 if (ret
>= 0 && ppos
)
646 SYSCALL_DEFINE3(write
, unsigned int, fd
, const char __user
*, buf
,
649 return ksys_write(fd
, buf
, count
);
652 ssize_t
ksys_pread64(unsigned int fd
, char __user
*buf
, size_t count
,
656 ssize_t ret
= -EBADF
;
664 if (f
.file
->f_mode
& FMODE_PREAD
)
665 ret
= vfs_read(f
.file
, buf
, count
, &pos
);
672 SYSCALL_DEFINE4(pread64
, unsigned int, fd
, char __user
*, buf
,
673 size_t, count
, loff_t
, pos
)
675 return ksys_pread64(fd
, buf
, count
, pos
);
678 #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PREAD64)
679 COMPAT_SYSCALL_DEFINE5(pread64
, unsigned int, fd
, char __user
*, buf
,
680 size_t, count
, compat_arg_u64_dual(pos
))
682 return ksys_pread64(fd
, buf
, count
, compat_arg_u64_glue(pos
));
686 ssize_t
ksys_pwrite64(unsigned int fd
, const char __user
*buf
,
687 size_t count
, loff_t pos
)
690 ssize_t ret
= -EBADF
;
698 if (f
.file
->f_mode
& FMODE_PWRITE
)
699 ret
= vfs_write(f
.file
, buf
, count
, &pos
);
706 SYSCALL_DEFINE4(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
707 size_t, count
, loff_t
, pos
)
709 return ksys_pwrite64(fd
, buf
, count
, pos
);
712 #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PWRITE64)
713 COMPAT_SYSCALL_DEFINE5(pwrite64
, unsigned int, fd
, const char __user
*, buf
,
714 size_t, count
, compat_arg_u64_dual(pos
))
716 return ksys_pwrite64(fd
, buf
, count
, compat_arg_u64_glue(pos
));
720 static ssize_t
do_iter_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
721 loff_t
*ppos
, int type
, rwf_t flags
)
726 init_sync_kiocb(&kiocb
, filp
);
727 ret
= kiocb_set_rw_flags(&kiocb
, flags
);
730 kiocb
.ki_pos
= (ppos
? *ppos
: 0);
733 ret
= call_read_iter(filp
, &kiocb
, iter
);
735 ret
= call_write_iter(filp
, &kiocb
, iter
);
736 BUG_ON(ret
== -EIOCBQUEUED
);
738 *ppos
= kiocb
.ki_pos
;
742 /* Do it by hand, with file-ops */
743 static ssize_t
do_loop_readv_writev(struct file
*filp
, struct iov_iter
*iter
,
744 loff_t
*ppos
, int type
, rwf_t flags
)
748 if (flags
& ~RWF_HIPRI
)
751 while (iov_iter_count(iter
)) {
755 nr
= filp
->f_op
->read(filp
, iter_iov_addr(iter
),
756 iter_iov_len(iter
), ppos
);
758 nr
= filp
->f_op
->write(filp
, iter_iov_addr(iter
),
759 iter_iov_len(iter
), ppos
);
768 if (nr
!= iter_iov_len(iter
))
770 iov_iter_advance(iter
, nr
);
776 ssize_t
vfs_iocb_iter_read(struct file
*file
, struct kiocb
*iocb
,
777 struct iov_iter
*iter
)
782 if (!file
->f_op
->read_iter
)
784 if (!(file
->f_mode
& FMODE_READ
))
786 if (!(file
->f_mode
& FMODE_CAN_READ
))
789 tot_len
= iov_iter_count(iter
);
792 ret
= rw_verify_area(READ
, file
, &iocb
->ki_pos
, tot_len
);
796 ret
= call_read_iter(file
, iocb
, iter
);
799 fsnotify_access(file
);
802 EXPORT_SYMBOL(vfs_iocb_iter_read
);
804 ssize_t
vfs_iter_read(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
810 if (!file
->f_op
->read_iter
)
812 if (!(file
->f_mode
& FMODE_READ
))
814 if (!(file
->f_mode
& FMODE_CAN_READ
))
817 tot_len
= iov_iter_count(iter
);
820 ret
= rw_verify_area(READ
, file
, ppos
, tot_len
);
824 ret
= do_iter_readv_writev(file
, iter
, ppos
, READ
, flags
);
827 fsnotify_access(file
);
830 EXPORT_SYMBOL(vfs_iter_read
);
832 ssize_t
vfs_iocb_iter_write(struct file
*file
, struct kiocb
*iocb
,
833 struct iov_iter
*iter
)
838 if (!file
->f_op
->write_iter
)
840 if (!(file
->f_mode
& FMODE_WRITE
))
842 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
845 tot_len
= iov_iter_count(iter
);
848 ret
= rw_verify_area(WRITE
, file
, &iocb
->ki_pos
, tot_len
);
852 ret
= call_write_iter(file
, iocb
, iter
);
854 fsnotify_modify(file
);
858 EXPORT_SYMBOL(vfs_iocb_iter_write
);
860 ssize_t
vfs_iter_write(struct file
*file
, struct iov_iter
*iter
, loff_t
*ppos
,
866 if (!(file
->f_mode
& FMODE_WRITE
))
868 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
870 if (!file
->f_op
->write_iter
)
873 tot_len
= iov_iter_count(iter
);
877 ret
= rw_verify_area(WRITE
, file
, ppos
, tot_len
);
881 file_start_write(file
);
882 ret
= do_iter_readv_writev(file
, iter
, ppos
, WRITE
, flags
);
884 fsnotify_modify(file
);
885 file_end_write(file
);
889 EXPORT_SYMBOL(vfs_iter_write
);
891 static ssize_t
vfs_readv(struct file
*file
, const struct iovec __user
*vec
,
892 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
894 struct iovec iovstack
[UIO_FASTIOV
];
895 struct iovec
*iov
= iovstack
;
896 struct iov_iter iter
;
900 if (!(file
->f_mode
& FMODE_READ
))
902 if (!(file
->f_mode
& FMODE_CAN_READ
))
905 ret
= import_iovec(ITER_DEST
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
,
910 tot_len
= iov_iter_count(&iter
);
914 ret
= rw_verify_area(READ
, file
, pos
, tot_len
);
918 if (file
->f_op
->read_iter
)
919 ret
= do_iter_readv_writev(file
, &iter
, pos
, READ
, flags
);
921 ret
= do_loop_readv_writev(file
, &iter
, pos
, READ
, flags
);
924 fsnotify_access(file
);
929 static ssize_t
vfs_writev(struct file
*file
, const struct iovec __user
*vec
,
930 unsigned long vlen
, loff_t
*pos
, rwf_t flags
)
932 struct iovec iovstack
[UIO_FASTIOV
];
933 struct iovec
*iov
= iovstack
;
934 struct iov_iter iter
;
938 if (!(file
->f_mode
& FMODE_WRITE
))
940 if (!(file
->f_mode
& FMODE_CAN_WRITE
))
943 ret
= import_iovec(ITER_SOURCE
, vec
, vlen
, ARRAY_SIZE(iovstack
), &iov
,
948 tot_len
= iov_iter_count(&iter
);
952 ret
= rw_verify_area(WRITE
, file
, pos
, tot_len
);
956 file_start_write(file
);
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
);
963 file_end_write(file
);
969 static ssize_t
do_readv(unsigned long fd
, const struct iovec __user
*vec
,
970 unsigned long vlen
, rwf_t flags
)
972 struct fd f
= fdget_pos(fd
);
973 ssize_t ret
= -EBADF
;
976 loff_t pos
, *ppos
= file_ppos(f
.file
);
981 ret
= vfs_readv(f
.file
, vec
, vlen
, ppos
, flags
);
982 if (ret
>= 0 && ppos
)
988 add_rchar(current
, ret
);
993 static ssize_t
do_writev(unsigned long fd
, const struct iovec __user
*vec
,
994 unsigned long vlen
, rwf_t flags
)
996 struct fd f
= fdget_pos(fd
);
997 ssize_t ret
= -EBADF
;
1000 loff_t pos
, *ppos
= file_ppos(f
.file
);
1005 ret
= vfs_writev(f
.file
, vec
, vlen
, ppos
, flags
);
1006 if (ret
>= 0 && ppos
)
1007 f
.file
->f_pos
= pos
;
1012 add_wchar(current
, ret
);
1017 static inline loff_t
pos_from_hilo(unsigned long high
, unsigned long low
)
1019 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
1020 return (((loff_t
)high
<< HALF_LONG_BITS
) << HALF_LONG_BITS
) | low
;
1023 static ssize_t
do_preadv(unsigned long fd
, const struct iovec __user
*vec
,
1024 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1027 ssize_t ret
= -EBADF
;
1035 if (f
.file
->f_mode
& FMODE_PREAD
)
1036 ret
= vfs_readv(f
.file
, vec
, vlen
, &pos
, flags
);
1041 add_rchar(current
, ret
);
1046 static ssize_t
do_pwritev(unsigned long fd
, const struct iovec __user
*vec
,
1047 unsigned long vlen
, loff_t pos
, rwf_t flags
)
1050 ssize_t ret
= -EBADF
;
1058 if (f
.file
->f_mode
& FMODE_PWRITE
)
1059 ret
= vfs_writev(f
.file
, vec
, vlen
, &pos
, flags
);
1064 add_wchar(current
, ret
);
1069 SYSCALL_DEFINE3(readv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1070 unsigned long, vlen
)
1072 return do_readv(fd
, vec
, vlen
, 0);
1075 SYSCALL_DEFINE3(writev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1076 unsigned long, vlen
)
1078 return do_writev(fd
, vec
, vlen
, 0);
1081 SYSCALL_DEFINE5(preadv
, unsigned long, fd
, const struct iovec __user
*, vec
,
1082 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1084 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1086 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1089 SYSCALL_DEFINE6(preadv2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1090 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1093 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1096 return do_readv(fd
, vec
, vlen
, flags
);
1098 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1101 SYSCALL_DEFINE5(pwritev
, unsigned long, fd
, const struct iovec __user
*, vec
,
1102 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
)
1104 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1106 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1109 SYSCALL_DEFINE6(pwritev2
, unsigned long, fd
, const struct iovec __user
*, vec
,
1110 unsigned long, vlen
, unsigned long, pos_l
, unsigned long, pos_h
,
1113 loff_t pos
= pos_from_hilo(pos_h
, pos_l
);
1116 return do_writev(fd
, vec
, vlen
, flags
);
1118 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1122 * Various compat syscalls. Note that they all pretend to take a native
1123 * iovec - import_iovec will properly treat those as compat_iovecs based on
1124 * in_compat_syscall().
1126 #ifdef CONFIG_COMPAT
1127 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1128 COMPAT_SYSCALL_DEFINE4(preadv64
, unsigned long, fd
,
1129 const struct iovec __user
*, vec
,
1130 unsigned long, vlen
, loff_t
, pos
)
1132 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1136 COMPAT_SYSCALL_DEFINE5(preadv
, compat_ulong_t
, fd
,
1137 const struct iovec __user
*, vec
,
1138 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1140 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1142 return do_preadv(fd
, vec
, vlen
, pos
, 0);
1145 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1146 COMPAT_SYSCALL_DEFINE5(preadv64v2
, unsigned long, fd
,
1147 const struct iovec __user
*, vec
,
1148 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1151 return do_readv(fd
, vec
, vlen
, flags
);
1152 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1156 COMPAT_SYSCALL_DEFINE6(preadv2
, compat_ulong_t
, fd
,
1157 const struct iovec __user
*, vec
,
1158 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
,
1161 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1164 return do_readv(fd
, vec
, vlen
, flags
);
1165 return do_preadv(fd
, vec
, vlen
, pos
, flags
);
1168 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1169 COMPAT_SYSCALL_DEFINE4(pwritev64
, unsigned long, fd
,
1170 const struct iovec __user
*, vec
,
1171 unsigned long, vlen
, loff_t
, pos
)
1173 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1177 COMPAT_SYSCALL_DEFINE5(pwritev
, compat_ulong_t
, fd
,
1178 const struct iovec __user
*,vec
,
1179 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
)
1181 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1183 return do_pwritev(fd
, vec
, vlen
, pos
, 0);
1186 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1187 COMPAT_SYSCALL_DEFINE5(pwritev64v2
, unsigned long, fd
,
1188 const struct iovec __user
*, vec
,
1189 unsigned long, vlen
, loff_t
, pos
, rwf_t
, flags
)
1192 return do_writev(fd
, vec
, vlen
, flags
);
1193 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1197 COMPAT_SYSCALL_DEFINE6(pwritev2
, compat_ulong_t
, fd
,
1198 const struct iovec __user
*,vec
,
1199 compat_ulong_t
, vlen
, u32
, pos_low
, u32
, pos_high
, rwf_t
, flags
)
1201 loff_t pos
= ((loff_t
)pos_high
<< 32) | pos_low
;
1204 return do_writev(fd
, vec
, vlen
, flags
);
1205 return do_pwritev(fd
, vec
, vlen
, pos
, flags
);
1207 #endif /* CONFIG_COMPAT */
1209 static ssize_t
do_sendfile(int out_fd
, int in_fd
, loff_t
*ppos
,
1210 size_t count
, loff_t max
)
1213 struct inode
*in_inode
, *out_inode
;
1214 struct pipe_inode_info
*opipe
;
1221 * Get input file, and verify that it is ok..
1227 if (!(in
.file
->f_mode
& FMODE_READ
))
1231 pos
= in
.file
->f_pos
;
1234 if (!(in
.file
->f_mode
& FMODE_PREAD
))
1237 retval
= rw_verify_area(READ
, in
.file
, &pos
, count
);
1240 if (count
> MAX_RW_COUNT
)
1241 count
= MAX_RW_COUNT
;
1244 * Get output file, and verify that it is ok..
1247 out
= fdget(out_fd
);
1250 if (!(out
.file
->f_mode
& FMODE_WRITE
))
1252 in_inode
= file_inode(in
.file
);
1253 out_inode
= file_inode(out
.file
);
1254 out_pos
= out
.file
->f_pos
;
1257 max
= min(in_inode
->i_sb
->s_maxbytes
, out_inode
->i_sb
->s_maxbytes
);
1259 if (unlikely(pos
+ count
> max
)) {
1260 retval
= -EOVERFLOW
;
1269 * We need to debate whether we can enable this or not. The
1270 * man page documents EAGAIN return for the output at least,
1271 * and the application is arguably buggy if it doesn't expect
1272 * EAGAIN on a non-blocking file descriptor.
1274 if (in
.file
->f_flags
& O_NONBLOCK
)
1275 fl
= SPLICE_F_NONBLOCK
;
1277 opipe
= get_pipe_info(out
.file
, true);
1279 retval
= rw_verify_area(WRITE
, out
.file
, &out_pos
, count
);
1282 file_start_write(out
.file
);
1283 retval
= do_splice_direct(in
.file
, &pos
, out
.file
, &out_pos
,
1285 file_end_write(out
.file
);
1287 if (out
.file
->f_flags
& O_NONBLOCK
)
1288 fl
|= SPLICE_F_NONBLOCK
;
1290 retval
= splice_file_to_pipe(in
.file
, opipe
, &pos
, count
, fl
);
1294 add_rchar(current
, retval
);
1295 add_wchar(current
, retval
);
1296 fsnotify_access(in
.file
);
1297 fsnotify_modify(out
.file
);
1298 out
.file
->f_pos
= out_pos
;
1302 in
.file
->f_pos
= pos
;
1308 retval
= -EOVERFLOW
;
1318 SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
, off_t __user
*, offset
, size_t, count
)
1325 if (unlikely(get_user(off
, offset
)))
1328 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1329 if (unlikely(put_user(pos
, offset
)))
1334 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1337 SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
, loff_t __user
*, offset
, size_t, count
)
1343 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1345 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1346 if (unlikely(put_user(pos
, offset
)))
1351 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1354 #ifdef CONFIG_COMPAT
1355 COMPAT_SYSCALL_DEFINE4(sendfile
, int, out_fd
, int, in_fd
,
1356 compat_off_t __user
*, offset
, compat_size_t
, count
)
1363 if (unlikely(get_user(off
, offset
)))
1366 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, MAX_NON_LFS
);
1367 if (unlikely(put_user(pos
, offset
)))
1372 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1375 COMPAT_SYSCALL_DEFINE4(sendfile64
, int, out_fd
, int, in_fd
,
1376 compat_loff_t __user
*, offset
, compat_size_t
, count
)
1382 if (unlikely(copy_from_user(&pos
, offset
, sizeof(loff_t
))))
1384 ret
= do_sendfile(out_fd
, in_fd
, &pos
, count
, 0);
1385 if (unlikely(put_user(pos
, offset
)))
1390 return do_sendfile(out_fd
, in_fd
, NULL
, count
, 0);
1395 * generic_copy_file_range - copy data between two files
1396 * @file_in: file structure to read from
1397 * @pos_in: file offset to read from
1398 * @file_out: file structure to write data to
1399 * @pos_out: file offset to write data to
1400 * @len: amount of data to copy
1401 * @flags: copy flags
1403 * This is a generic filesystem helper to copy data from one file to another.
1404 * It has no constraints on the source or destination file owners - the files
1405 * can belong to different superblocks and different filesystem types. Short
1406 * copies are allowed.
1408 * This should be called from the @file_out filesystem, as per the
1409 * ->copy_file_range() method.
1411 * Returns the number of bytes copied or a negative error indicating the
1415 ssize_t
generic_copy_file_range(struct file
*file_in
, loff_t pos_in
,
1416 struct file
*file_out
, loff_t pos_out
,
1417 size_t len
, unsigned int flags
)
1419 lockdep_assert(sb_write_started(file_inode(file_out
)->i_sb
));
1421 return do_splice_direct(file_in
, &pos_in
, file_out
, &pos_out
,
1422 len
> MAX_RW_COUNT
? MAX_RW_COUNT
: len
, 0);
1424 EXPORT_SYMBOL(generic_copy_file_range
);
1427 * Performs necessary checks before doing a file copy
1429 * Can adjust amount of bytes to copy via @req_count argument.
1430 * Returns appropriate error code that caller should return or
1431 * zero in case the copy should be allowed.
1433 static int generic_copy_file_checks(struct file
*file_in
, loff_t pos_in
,
1434 struct file
*file_out
, loff_t pos_out
,
1435 size_t *req_count
, unsigned int flags
)
1437 struct inode
*inode_in
= file_inode(file_in
);
1438 struct inode
*inode_out
= file_inode(file_out
);
1439 uint64_t count
= *req_count
;
1443 ret
= generic_file_rw_checks(file_in
, file_out
);
1448 * We allow some filesystems to handle cross sb copy, but passing
1449 * a file of the wrong filesystem type to filesystem driver can result
1450 * in an attempt to dereference the wrong type of ->private_data, so
1451 * avoid doing that until we really have a good reason.
1453 * nfs and cifs define several different file_system_type structures
1454 * and several different sets of file_operations, but they all end up
1455 * using the same ->copy_file_range() function pointer.
1457 if (flags
& COPY_FILE_SPLICE
) {
1458 /* cross sb splice is allowed */
1459 } else if (file_out
->f_op
->copy_file_range
) {
1460 if (file_in
->f_op
->copy_file_range
!=
1461 file_out
->f_op
->copy_file_range
)
1463 } else if (file_inode(file_in
)->i_sb
!= file_inode(file_out
)->i_sb
) {
1467 /* Don't touch certain kinds of inodes */
1468 if (IS_IMMUTABLE(inode_out
))
1471 if (IS_SWAPFILE(inode_in
) || IS_SWAPFILE(inode_out
))
1474 /* Ensure offsets don't wrap. */
1475 if (pos_in
+ count
< pos_in
|| pos_out
+ count
< pos_out
)
1478 /* Shorten the copy to EOF */
1479 size_in
= i_size_read(inode_in
);
1480 if (pos_in
>= size_in
)
1483 count
= min(count
, size_in
- (uint64_t)pos_in
);
1485 ret
= generic_write_check_limits(file_out
, pos_out
, &count
);
1489 /* Don't allow overlapped copying within the same file. */
1490 if (inode_in
== inode_out
&&
1491 pos_out
+ count
> pos_in
&&
1492 pos_out
< pos_in
+ count
)
1500 * copy_file_range() differs from regular file read and write in that it
1501 * specifically allows return partial success. When it does so is up to
1502 * the copy_file_range method.
1504 ssize_t
vfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
1505 struct file
*file_out
, loff_t pos_out
,
1506 size_t len
, unsigned int flags
)
1509 bool splice
= flags
& COPY_FILE_SPLICE
;
1511 if (flags
& ~COPY_FILE_SPLICE
)
1514 ret
= generic_copy_file_checks(file_in
, pos_in
, file_out
, pos_out
, &len
,
1519 ret
= rw_verify_area(READ
, file_in
, &pos_in
, len
);
1523 ret
= rw_verify_area(WRITE
, file_out
, &pos_out
, len
);
1530 file_start_write(file_out
);
1533 * Cloning is supported by more file systems, so we implement copy on
1534 * same sb using clone, but for filesystems where both clone and copy
1535 * are supported (e.g. nfs,cifs), we only call the copy method.
1537 if (!splice
&& file_out
->f_op
->copy_file_range
) {
1538 ret
= file_out
->f_op
->copy_file_range(file_in
, pos_in
,
1544 if (!splice
&& file_in
->f_op
->remap_file_range
&&
1545 file_inode(file_in
)->i_sb
== file_inode(file_out
)->i_sb
) {
1546 ret
= file_in
->f_op
->remap_file_range(file_in
, pos_in
,
1548 min_t(loff_t
, MAX_RW_COUNT
, len
),
1549 REMAP_FILE_CAN_SHORTEN
);
1555 * We can get here for same sb copy of filesystems that do not implement
1556 * ->copy_file_range() in case filesystem does not support clone or in
1557 * case filesystem supports clone but rejected the clone request (e.g.
1558 * because it was not block aligned).
1560 * In both cases, fall back to kernel copy so we are able to maintain a
1561 * consistent story about which filesystems support copy_file_range()
1562 * and which filesystems do not, that will allow userspace tools to
1563 * make consistent desicions w.r.t using copy_file_range().
1565 * We also get here if caller (e.g. nfsd) requested COPY_FILE_SPLICE.
1567 ret
= generic_copy_file_range(file_in
, pos_in
, file_out
, pos_out
, len
,
1572 fsnotify_access(file_in
);
1573 add_rchar(current
, ret
);
1574 fsnotify_modify(file_out
);
1575 add_wchar(current
, ret
);
1581 file_end_write(file_out
);
1585 EXPORT_SYMBOL(vfs_copy_file_range
);
1587 SYSCALL_DEFINE6(copy_file_range
, int, fd_in
, loff_t __user
*, off_in
,
1588 int, fd_out
, loff_t __user
*, off_out
,
1589 size_t, len
, unsigned int, flags
)
1595 ssize_t ret
= -EBADF
;
1597 f_in
= fdget(fd_in
);
1601 f_out
= fdget(fd_out
);
1607 if (copy_from_user(&pos_in
, off_in
, sizeof(loff_t
)))
1610 pos_in
= f_in
.file
->f_pos
;
1614 if (copy_from_user(&pos_out
, off_out
, sizeof(loff_t
)))
1617 pos_out
= f_out
.file
->f_pos
;
1624 ret
= vfs_copy_file_range(f_in
.file
, pos_in
, f_out
.file
, pos_out
, len
,
1631 if (copy_to_user(off_in
, &pos_in
, sizeof(loff_t
)))
1634 f_in
.file
->f_pos
= pos_in
;
1638 if (copy_to_user(off_out
, &pos_out
, sizeof(loff_t
)))
1641 f_out
.file
->f_pos
= pos_out
;
1654 * Don't operate on ranges the page cache doesn't support, and don't exceed the
1655 * LFS limits. If pos is under the limit it becomes a short access. If it
1656 * exceeds the limit we return -EFBIG.
1658 int generic_write_check_limits(struct file
*file
, loff_t pos
, loff_t
*count
)
1660 struct inode
*inode
= file
->f_mapping
->host
;
1661 loff_t max_size
= inode
->i_sb
->s_maxbytes
;
1662 loff_t limit
= rlimit(RLIMIT_FSIZE
);
1664 if (limit
!= RLIM_INFINITY
) {
1666 send_sig(SIGXFSZ
, current
, 0);
1669 *count
= min(*count
, limit
- pos
);
1672 if (!(file
->f_flags
& O_LARGEFILE
))
1673 max_size
= MAX_NON_LFS
;
1675 if (unlikely(pos
>= max_size
))
1678 *count
= min(*count
, max_size
- pos
);
1683 /* Like generic_write_checks(), but takes size of write instead of iter. */
1684 int generic_write_checks_count(struct kiocb
*iocb
, loff_t
*count
)
1686 struct file
*file
= iocb
->ki_filp
;
1687 struct inode
*inode
= file
->f_mapping
->host
;
1689 if (IS_SWAPFILE(inode
))
1695 if (iocb
->ki_flags
& IOCB_APPEND
)
1696 iocb
->ki_pos
= i_size_read(inode
);
1698 if ((iocb
->ki_flags
& IOCB_NOWAIT
) &&
1699 !((iocb
->ki_flags
& IOCB_DIRECT
) ||
1700 (file
->f_mode
& FMODE_BUF_WASYNC
)))
1703 return generic_write_check_limits(iocb
->ki_filp
, iocb
->ki_pos
, count
);
1705 EXPORT_SYMBOL(generic_write_checks_count
);
1708 * Performs necessary checks before doing a write
1710 * Can adjust writing position or amount of bytes to write.
1711 * Returns appropriate error code that caller should return or
1712 * zero in case that write should be allowed.
1714 ssize_t
generic_write_checks(struct kiocb
*iocb
, struct iov_iter
*from
)
1716 loff_t count
= iov_iter_count(from
);
1719 ret
= generic_write_checks_count(iocb
, &count
);
1723 iov_iter_truncate(from
, count
);
1724 return iov_iter_count(from
);
1726 EXPORT_SYMBOL(generic_write_checks
);
1729 * Performs common checks before doing a file copy/clone
1730 * from @file_in to @file_out.
1732 int generic_file_rw_checks(struct file
*file_in
, struct file
*file_out
)
1734 struct inode
*inode_in
= file_inode(file_in
);
1735 struct inode
*inode_out
= file_inode(file_out
);
1737 /* Don't copy dirs, pipes, sockets... */
1738 if (S_ISDIR(inode_in
->i_mode
) || S_ISDIR(inode_out
->i_mode
))
1740 if (!S_ISREG(inode_in
->i_mode
) || !S_ISREG(inode_out
->i_mode
))
1743 if (!(file_in
->f_mode
& FMODE_READ
) ||
1744 !(file_out
->f_mode
& FMODE_WRITE
) ||
1745 (file_out
->f_flags
& O_APPEND
))