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>
17 #include <linux/pagemap.h>
18 #include <linux/splice.h>
19 #include "read_write.h"
21 #include <asm/uaccess.h>
22 #include <asm/unistd.h>
24 const struct file_operations generic_ro_fops = {
25 .llseek = generic_file_llseek,
27 .aio_read = generic_file_aio_read,
28 .mmap = generic_file_readonly_mmap,
29 .splice_read = generic_file_splice_read,
32 EXPORT_SYMBOL(generic_ro_fops);
34 loff_t generic_file_llseek(struct file *file, loff_t offset, int origin)
37 struct inode *inode = file->f_mapping->host;
39 mutex_lock(&inode->i_mutex);
42 offset += inode->i_size;
45 offset += file->f_pos;
48 if (offset>=0 && offset<=inode->i_sb->s_maxbytes) {
49 if (offset != file->f_pos) {
55 mutex_unlock(&inode->i_mutex);
59 EXPORT_SYMBOL(generic_file_llseek);
61 loff_t remote_llseek(struct file *file, loff_t offset, int origin)
68 offset += i_size_read(file->f_path.dentry->d_inode);
71 offset += file->f_pos;
74 if (offset>=0 && offset<=file->f_path.dentry->d_inode->i_sb->s_maxbytes) {
75 if (offset != file->f_pos) {
84 EXPORT_SYMBOL(remote_llseek);
86 loff_t no_llseek(struct file *file, loff_t offset, int origin)
90 EXPORT_SYMBOL(no_llseek);
92 loff_t default_llseek(struct file *file, loff_t offset, int origin)
99 offset += i_size_read(file->f_path.dentry->d_inode);
102 offset += file->f_pos;
106 if (offset != file->f_pos) {
107 file->f_pos = offset;
115 EXPORT_SYMBOL(default_llseek);
117 loff_t vfs_llseek(struct file *file, loff_t offset, int origin)
119 loff_t (*fn)(struct file *, loff_t, int);
122 if (file->f_mode & FMODE_LSEEK) {
124 if (file->f_op && file->f_op->llseek)
125 fn = file->f_op->llseek;
127 return fn(file, offset, origin);
129 EXPORT_SYMBOL(vfs_llseek);
131 asmlinkage off_t sys_lseek(unsigned int fd, off_t offset, unsigned int origin)
138 file = fget_light(fd, &fput_needed);
143 if (origin <= SEEK_MAX) {
144 loff_t res = vfs_llseek(file, offset, origin);
146 if (res != (loff_t)retval)
147 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
149 fput_light(file, fput_needed);
154 #ifdef __ARCH_WANT_SYS_LLSEEK
155 asmlinkage long sys_llseek(unsigned int fd, unsigned long offset_high,
156 unsigned long offset_low, loff_t __user * result,
165 file = fget_light(fd, &fput_needed);
170 if (origin > SEEK_MAX)
173 offset = vfs_llseek(file, ((loff_t) offset_high << 32) | offset_low,
176 retval = (int)offset;
179 if (!copy_to_user(result, &offset, sizeof(offset)))
183 fput_light(file, fput_needed);
190 * rw_verify_area doesn't like huge counts. We limit
191 * them to something that fits in "int" so that others
192 * won't have to do range checks all the time.
194 #define MAX_RW_COUNT (INT_MAX & PAGE_CACHE_MASK)
196 int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
201 inode = file->f_path.dentry->d_inode;
202 if (unlikely((ssize_t) count < 0))
205 if (unlikely((pos < 0) || (loff_t) (pos + count) < 0))
208 if (unlikely(inode->i_flock && MANDATORY_LOCK(inode))) {
209 int retval = locks_mandatory_area(
210 read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
211 inode, file, pos, count);
215 return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
221 static void wait_on_retry_sync_kiocb(struct kiocb *iocb)
223 set_current_state(TASK_UNINTERRUPTIBLE);
224 if (!kiocbIsKicked(iocb))
227 kiocbClearKicked(iocb);
228 __set_current_state(TASK_RUNNING);
231 ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
233 struct iovec iov = { .iov_base = buf, .iov_len = len };
237 init_sync_kiocb(&kiocb, filp);
238 kiocb.ki_pos = *ppos;
242 ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
243 if (ret != -EIOCBRETRY)
245 wait_on_retry_sync_kiocb(&kiocb);
248 if (-EIOCBQUEUED == ret)
249 ret = wait_on_sync_kiocb(&kiocb);
250 *ppos = kiocb.ki_pos;
254 EXPORT_SYMBOL(do_sync_read);
256 ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
260 if (!(file->f_mode & FMODE_READ))
262 if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
264 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
267 ret = rw_verify_area(READ, file, pos, count);
270 ret = security_file_permission (file, MAY_READ);
272 if (file->f_op->read)
273 ret = file->f_op->read(file, buf, count, pos);
275 ret = do_sync_read(file, buf, count, pos);
277 fsnotify_access(file->f_path.dentry);
278 add_rchar(current, ret);
287 EXPORT_SYMBOL(vfs_read);
289 ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
291 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
295 init_sync_kiocb(&kiocb, filp);
296 kiocb.ki_pos = *ppos;
300 ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
301 if (ret != -EIOCBRETRY)
303 wait_on_retry_sync_kiocb(&kiocb);
306 if (-EIOCBQUEUED == ret)
307 ret = wait_on_sync_kiocb(&kiocb);
308 *ppos = kiocb.ki_pos;
312 EXPORT_SYMBOL(do_sync_write);
314 ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
318 if (!(file->f_mode & FMODE_WRITE))
320 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
322 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
325 ret = rw_verify_area(WRITE, file, pos, count);
328 ret = security_file_permission (file, MAY_WRITE);
330 if (file->f_op->write)
331 ret = file->f_op->write(file, buf, count, pos);
333 ret = do_sync_write(file, buf, count, pos);
335 fsnotify_modify(file->f_path.dentry);
336 add_wchar(current, ret);
345 EXPORT_SYMBOL(vfs_write);
347 static inline loff_t file_pos_read(struct file *file)
352 static inline void file_pos_write(struct file *file, loff_t pos)
357 asmlinkage ssize_t sys_read(unsigned int fd, char __user * buf, size_t count)
360 ssize_t ret = -EBADF;
363 file = fget_light(fd, &fput_needed);
365 loff_t pos = file_pos_read(file);
366 ret = vfs_read(file, buf, count, &pos);
367 file_pos_write(file, pos);
368 fput_light(file, fput_needed);
373 EXPORT_SYMBOL_GPL(sys_read);
375 asmlinkage ssize_t sys_write(unsigned int fd, const char __user * buf, size_t count)
378 ssize_t ret = -EBADF;
381 file = fget_light(fd, &fput_needed);
383 loff_t pos = file_pos_read(file);
384 ret = vfs_write(file, buf, count, &pos);
385 file_pos_write(file, pos);
386 fput_light(file, fput_needed);
392 asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf,
393 size_t count, loff_t pos)
396 ssize_t ret = -EBADF;
402 file = fget_light(fd, &fput_needed);
405 if (file->f_mode & FMODE_PREAD)
406 ret = vfs_read(file, buf, count, &pos);
407 fput_light(file, fput_needed);
413 asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf,
414 size_t count, loff_t pos)
417 ssize_t ret = -EBADF;
423 file = fget_light(fd, &fput_needed);
426 if (file->f_mode & FMODE_PWRITE)
427 ret = vfs_write(file, buf, count, &pos);
428 fput_light(file, fput_needed);
435 * Reduce an iovec's length in-place. Return the resulting number of segments
437 unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
439 unsigned long seg = 0;
442 while (seg < nr_segs) {
444 if (len + iov->iov_len >= to) {
445 iov->iov_len = to - len;
454 ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
455 unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
460 init_sync_kiocb(&kiocb, filp);
461 kiocb.ki_pos = *ppos;
463 kiocb.ki_nbytes = len;
466 ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
467 if (ret != -EIOCBRETRY)
469 wait_on_retry_sync_kiocb(&kiocb);
472 if (ret == -EIOCBQUEUED)
473 ret = wait_on_sync_kiocb(&kiocb);
474 *ppos = kiocb.ki_pos;
478 /* Do it by hand, with file-ops */
479 ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
480 unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
482 struct iovec *vector = iov;
485 while (nr_segs > 0) {
490 base = vector->iov_base;
491 len = vector->iov_len;
495 nr = fn(filp, base, len, ppos);
510 /* A write operation does a read from user space and vice versa */
511 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
513 ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
514 unsigned long nr_segs, unsigned long fast_segs,
515 struct iovec *fast_pointer,
516 struct iovec **ret_pointer)
520 struct iovec *iov = fast_pointer;
523 * SuS says "The readv() function *may* fail if the iovcnt argument
524 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
525 * traditionally returned zero for zero segments, so...
533 * First get the "struct iovec" from user memory and
534 * verify all the pointers
536 if (nr_segs > UIO_MAXIOV) {
540 if (nr_segs > fast_segs) {
541 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
547 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
553 * According to the Single Unix Specification we should return EINVAL
554 * if an element length is < 0 when cast to ssize_t or if the
555 * total length would overflow the ssize_t return value of the
559 for (seg = 0; seg < nr_segs; seg++) {
560 void __user *buf = iov[seg].iov_base;
561 ssize_t len = (ssize_t)iov[seg].iov_len;
563 /* see if we we're about to use an invalid len or if
564 * it's about to overflow ssize_t */
565 if (len < 0 || (ret + len < ret)) {
569 if (unlikely(!access_ok(vrfy_dir(type), buf, len))) {
581 static ssize_t do_readv_writev(int type, struct file *file,
582 const struct iovec __user * uvector,
583 unsigned long nr_segs, loff_t *pos)
586 struct iovec iovstack[UIO_FASTIOV];
587 struct iovec *iov = iovstack;
597 ret = rw_copy_check_uvector(type, uvector, nr_segs,
598 ARRAY_SIZE(iovstack), iovstack, &iov);
603 ret = rw_verify_area(type, file, pos, tot_len);
606 ret = security_file_permission(file, type == READ ? MAY_READ : MAY_WRITE);
612 fn = file->f_op->read;
613 fnv = file->f_op->aio_read;
615 fn = (io_fn_t)file->f_op->write;
616 fnv = file->f_op->aio_write;
620 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
623 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
628 if ((ret + (type == READ)) > 0) {
630 fsnotify_access(file->f_path.dentry);
632 fsnotify_modify(file->f_path.dentry);
637 ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
638 unsigned long vlen, loff_t *pos)
640 if (!(file->f_mode & FMODE_READ))
642 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
645 return do_readv_writev(READ, file, vec, vlen, pos);
648 EXPORT_SYMBOL(vfs_readv);
650 ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
651 unsigned long vlen, loff_t *pos)
653 if (!(file->f_mode & FMODE_WRITE))
655 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
658 return do_readv_writev(WRITE, file, vec, vlen, pos);
661 EXPORT_SYMBOL(vfs_writev);
664 sys_readv(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
667 ssize_t ret = -EBADF;
670 file = fget_light(fd, &fput_needed);
672 loff_t pos = file_pos_read(file);
673 ret = vfs_readv(file, vec, vlen, &pos);
674 file_pos_write(file, pos);
675 fput_light(file, fput_needed);
679 add_rchar(current, ret);
685 sys_writev(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
688 ssize_t ret = -EBADF;
691 file = fget_light(fd, &fput_needed);
693 loff_t pos = file_pos_read(file);
694 ret = vfs_writev(file, vec, vlen, &pos);
695 file_pos_write(file, pos);
696 fput_light(file, fput_needed);
700 add_wchar(current, ret);
705 static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
706 size_t count, loff_t max)
708 struct file * in_file, * out_file;
709 struct inode * in_inode, * out_inode;
712 int fput_needed_in, fput_needed_out, fl;
715 * Get input file, and verify that it is ok..
718 in_file = fget_light(in_fd, &fput_needed_in);
721 if (!(in_file->f_mode & FMODE_READ))
724 in_inode = in_file->f_path.dentry->d_inode;
727 if (!in_file->f_op || !in_file->f_op->splice_read)
731 ppos = &in_file->f_pos;
733 if (!(in_file->f_mode & FMODE_PREAD))
735 retval = rw_verify_area(READ, in_file, ppos, count);
740 retval = security_file_permission (in_file, MAY_READ);
745 * Get output file, and verify that it is ok..
748 out_file = fget_light(out_fd, &fput_needed_out);
751 if (!(out_file->f_mode & FMODE_WRITE))
754 if (!out_file->f_op || !out_file->f_op->sendpage)
756 out_inode = out_file->f_path.dentry->d_inode;
757 retval = rw_verify_area(WRITE, out_file, &out_file->f_pos, count);
762 retval = security_file_permission (out_file, MAY_WRITE);
767 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
771 if (unlikely(pos < 0))
773 if (unlikely(pos + count > max)) {
783 * We need to debate whether we can enable this or not. The
784 * man page documents EAGAIN return for the output at least,
785 * and the application is arguably buggy if it doesn't expect
786 * EAGAIN on a non-blocking file descriptor.
788 if (in_file->f_flags & O_NONBLOCK)
789 fl = SPLICE_F_NONBLOCK;
791 retval = do_splice_direct(in_file, ppos, out_file, count, fl);
794 add_rchar(current, retval);
795 add_wchar(current, retval);
804 fput_light(out_file, fput_needed_out);
806 fput_light(in_file, fput_needed_in);
811 asmlinkage ssize_t sys_sendfile(int out_fd, int in_fd, off_t __user *offset, size_t count)
818 if (unlikely(get_user(off, offset)))
821 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
822 if (unlikely(put_user(pos, offset)))
827 return do_sendfile(out_fd, in_fd, NULL, count, 0);
830 asmlinkage ssize_t sys_sendfile64(int out_fd, int in_fd, loff_t __user *offset, size_t count)
836 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
838 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
839 if (unlikely(put_user(pos, offset)))
844 return do_sendfile(out_fd, in_fd, NULL, count, 0);