[PATCH] Kprobes: Use RCU for (un)register synchronization - base changes
[linux-2.6/linux-2.6-openrd.git] / fs / read_write.c
bloba091ee4f430df6cc83317b59c4a2fc363c005080
1 /*
2 * linux/fs/read_write.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
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)
32 long long retval;
33 struct inode *inode = file->f_mapping->host;
35 down(&inode->i_sem);
36 switch (origin) {
37 case 2:
38 offset += inode->i_size;
39 break;
40 case 1:
41 offset += file->f_pos;
43 retval = -EINVAL;
44 if (offset>=0 && offset<=inode->i_sb->s_maxbytes) {
45 if (offset != file->f_pos) {
46 file->f_pos = offset;
47 file->f_version = 0;
49 retval = offset;
51 up(&inode->i_sem);
52 return retval;
55 EXPORT_SYMBOL(generic_file_llseek);
57 loff_t remote_llseek(struct file *file, loff_t offset, int origin)
59 long long retval;
61 lock_kernel();
62 switch (origin) {
63 case 2:
64 offset += i_size_read(file->f_dentry->d_inode);
65 break;
66 case 1:
67 offset += file->f_pos;
69 retval = -EINVAL;
70 if (offset>=0 && offset<=file->f_dentry->d_inode->i_sb->s_maxbytes) {
71 if (offset != file->f_pos) {
72 file->f_pos = offset;
73 file->f_version = 0;
75 retval = offset;
77 unlock_kernel();
78 return retval;
80 EXPORT_SYMBOL(remote_llseek);
82 loff_t no_llseek(struct file *file, loff_t offset, int origin)
84 return -ESPIPE;
86 EXPORT_SYMBOL(no_llseek);
88 loff_t default_llseek(struct file *file, loff_t offset, int origin)
90 long long retval;
92 lock_kernel();
93 switch (origin) {
94 case 2:
95 offset += i_size_read(file->f_dentry->d_inode);
96 break;
97 case 1:
98 offset += file->f_pos;
100 retval = -EINVAL;
101 if (offset >= 0) {
102 if (offset != file->f_pos) {
103 file->f_pos = offset;
104 file->f_version = 0;
106 retval = offset;
108 unlock_kernel();
109 return retval;
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);
117 fn = no_llseek;
118 if (file->f_mode & FMODE_LSEEK) {
119 fn = default_llseek;
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)
129 off_t retval;
130 struct file * file;
131 int fput_needed;
133 retval = -EBADF;
134 file = fget_light(fd, &fput_needed);
135 if (!file)
136 goto bad;
138 retval = -EINVAL;
139 if (origin <= 2) {
140 loff_t res = vfs_llseek(file, offset, origin);
141 retval = res;
142 if (res != (loff_t)retval)
143 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
145 fput_light(file, fput_needed);
146 bad:
147 return retval;
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,
153 unsigned int origin)
155 int retval;
156 struct file * file;
157 loff_t offset;
158 int fput_needed;
160 retval = -EBADF;
161 file = fget_light(fd, &fput_needed);
162 if (!file)
163 goto bad;
165 retval = -EINVAL;
166 if (origin > 2)
167 goto out_putf;
169 offset = vfs_llseek(file, ((loff_t) offset_high << 32) | offset_low,
170 origin);
172 retval = (int)offset;
173 if (offset >= 0) {
174 retval = -EFAULT;
175 if (!copy_to_user(result, &offset, sizeof(offset)))
176 retval = 0;
178 out_putf:
179 fput_light(file, fput_needed);
180 bad:
181 return retval;
183 #endif
186 int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
188 struct inode *inode;
189 loff_t pos;
191 if (unlikely(count > INT_MAX))
192 goto Einval;
193 pos = *ppos;
194 if (unlikely((pos < 0) || (loff_t) (pos + count) < 0))
195 goto Einval;
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);
200 return 0;
202 Einval:
203 return -EINVAL;
206 static void wait_on_retry_sync_kiocb(struct kiocb *iocb)
208 set_current_state(TASK_UNINTERRUPTIBLE);
209 if (!kiocbIsKicked(iocb))
210 schedule();
211 else
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)
218 struct kiocb kiocb;
219 ssize_t ret;
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;
230 return ret;
233 EXPORT_SYMBOL(do_sync_read);
235 ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
237 ssize_t ret;
239 if (!(file->f_mode & FMODE_READ))
240 return -EBADF;
241 if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
242 return -EINVAL;
243 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
244 return -EFAULT;
246 ret = rw_verify_area(READ, file, pos, count);
247 if (!ret) {
248 ret = security_file_permission (file, MAY_READ);
249 if (!ret) {
250 if (file->f_op->read)
251 ret = file->f_op->read(file, buf, count, pos);
252 else
253 ret = do_sync_read(file, buf, count, pos);
254 if (ret > 0) {
255 fsnotify_access(file->f_dentry);
256 current->rchar += ret;
258 current->syscr++;
262 return 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)
269 struct kiocb kiocb;
270 ssize_t ret;
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;
281 return ret;
284 EXPORT_SYMBOL(do_sync_write);
286 ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
288 ssize_t ret;
290 if (!(file->f_mode & FMODE_WRITE))
291 return -EBADF;
292 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
293 return -EINVAL;
294 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
295 return -EFAULT;
297 ret = rw_verify_area(WRITE, file, pos, count);
298 if (!ret) {
299 ret = security_file_permission (file, MAY_WRITE);
300 if (!ret) {
301 if (file->f_op->write)
302 ret = file->f_op->write(file, buf, count, pos);
303 else
304 ret = do_sync_write(file, buf, count, pos);
305 if (ret > 0) {
306 fsnotify_modify(file->f_dentry);
307 current->wchar += ret;
309 current->syscw++;
313 return ret;
316 EXPORT_SYMBOL(vfs_write);
318 static inline loff_t file_pos_read(struct file *file)
320 return file->f_pos;
323 static inline void file_pos_write(struct file *file, loff_t pos)
325 file->f_pos = pos;
328 asmlinkage ssize_t sys_read(unsigned int fd, char __user * buf, size_t count)
330 struct file *file;
331 ssize_t ret = -EBADF;
332 int fput_needed;
334 file = fget_light(fd, &fput_needed);
335 if (file) {
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);
342 return ret;
344 EXPORT_SYMBOL_GPL(sys_read);
346 asmlinkage ssize_t sys_write(unsigned int fd, const char __user * buf, size_t count)
348 struct file *file;
349 ssize_t ret = -EBADF;
350 int fput_needed;
352 file = fget_light(fd, &fput_needed);
353 if (file) {
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);
360 return ret;
363 asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf,
364 size_t count, loff_t pos)
366 struct file *file;
367 ssize_t ret = -EBADF;
368 int fput_needed;
370 if (pos < 0)
371 return -EINVAL;
373 file = fget_light(fd, &fput_needed);
374 if (file) {
375 ret = -ESPIPE;
376 if (file->f_mode & FMODE_PREAD)
377 ret = vfs_read(file, buf, count, &pos);
378 fput_light(file, fput_needed);
381 return ret;
384 asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf,
385 size_t count, loff_t pos)
387 struct file *file;
388 ssize_t ret = -EBADF;
389 int fput_needed;
391 if (pos < 0)
392 return -EINVAL;
394 file = fget_light(fd, &fput_needed);
395 if (file) {
396 ret = -ESPIPE;
397 if (file->f_mode & FMODE_PWRITE)
398 ret = vfs_write(file, buf, count, &pos);
399 fput_light(file, fput_needed);
402 return ret;
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;
411 size_t len = 0;
413 while (seg < nr_segs) {
414 seg++;
415 if (len + iov->iov_len >= to) {
416 iov->iov_len = to - len;
417 break;
419 len += iov->iov_len;
420 iov++;
422 return seg;
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 *);
437 size_t tot_len;
438 struct iovec iovstack[UIO_FASTIOV];
439 struct iovec *iov=iovstack, *vector;
440 ssize_t ret;
441 int seg;
442 io_fn_t fn;
443 iov_fn_t fnv;
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...
450 ret = 0;
451 if (nr_segs == 0)
452 goto out;
455 * First get the "struct iovec" from user memory and
456 * verify all the pointers
458 ret = -EINVAL;
459 if ((nr_segs > UIO_MAXIOV) || (nr_segs <= 0))
460 goto out;
461 if (!file->f_op)
462 goto out;
463 if (nr_segs > UIO_FASTIOV) {
464 ret = -ENOMEM;
465 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
466 if (!iov)
467 goto out;
469 ret = -EFAULT;
470 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector)))
471 goto out;
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
480 tot_len = 0;
481 ret = -EINVAL;
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 .. */
487 goto out;
488 if (unlikely(!access_ok(vrfy_dir(type), buf, len)))
489 goto Efault;
490 tot_len += len;
491 if ((ssize_t)tot_len < 0) /* maths overflow on the ssize_t */
492 goto out;
494 if (tot_len == 0) {
495 ret = 0;
496 goto out;
499 ret = rw_verify_area(type, file, pos, tot_len);
500 if (ret)
501 goto out;
502 ret = security_file_permission(file, type == READ ? MAY_READ : MAY_WRITE);
503 if (ret)
504 goto out;
506 fnv = NULL;
507 if (type == READ) {
508 fn = file->f_op->read;
509 fnv = file->f_op->readv;
510 } else {
511 fn = (io_fn_t)file->f_op->write;
512 fnv = file->f_op->writev;
514 if (fnv) {
515 ret = fnv(file, iov, nr_segs, pos);
516 goto out;
519 /* Do it by hand, with file-ops */
520 ret = 0;
521 vector = iov;
522 while (nr_segs > 0) {
523 void __user * base;
524 size_t len;
525 ssize_t nr;
527 base = vector->iov_base;
528 len = vector->iov_len;
529 vector++;
530 nr_segs--;
532 nr = fn(file, base, len, pos);
534 if (nr < 0) {
535 if (!ret) ret = nr;
536 break;
538 ret += nr;
539 if (nr != len)
540 break;
542 out:
543 if (iov != iovstack)
544 kfree(iov);
545 if ((ret + (type == READ)) > 0) {
546 if (type == READ)
547 fsnotify_access(file->f_dentry);
548 else
549 fsnotify_modify(file->f_dentry);
551 return ret;
552 Efault:
553 ret = -EFAULT;
554 goto out;
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))
561 return -EBADF;
562 if (!file->f_op || (!file->f_op->readv && !file->f_op->read))
563 return -EINVAL;
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))
574 return -EBADF;
575 if (!file->f_op || (!file->f_op->writev && !file->f_op->write))
576 return -EINVAL;
578 return do_readv_writev(WRITE, file, vec, vlen, pos);
581 EXPORT_SYMBOL(vfs_writev);
583 asmlinkage ssize_t
584 sys_readv(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
586 struct file *file;
587 ssize_t ret = -EBADF;
588 int fput_needed;
590 file = fget_light(fd, &fput_needed);
591 if (file) {
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);
598 if (ret > 0)
599 current->rchar += ret;
600 current->syscr++;
601 return ret;
604 asmlinkage ssize_t
605 sys_writev(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
607 struct file *file;
608 ssize_t ret = -EBADF;
609 int fput_needed;
611 file = fget_light(fd, &fput_needed);
612 if (file) {
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);
619 if (ret > 0)
620 current->wchar += ret;
621 current->syscw++;
622 return 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;
630 loff_t pos;
631 ssize_t retval;
632 int fput_needed_in, fput_needed_out;
635 * Get input file, and verify that it is ok..
637 retval = -EBADF;
638 in_file = fget_light(in_fd, &fput_needed_in);
639 if (!in_file)
640 goto out;
641 if (!(in_file->f_mode & FMODE_READ))
642 goto fput_in;
643 retval = -EINVAL;
644 in_inode = in_file->f_dentry->d_inode;
645 if (!in_inode)
646 goto fput_in;
647 if (!in_file->f_op || !in_file->f_op->sendfile)
648 goto fput_in;
649 retval = -ESPIPE;
650 if (!ppos)
651 ppos = &in_file->f_pos;
652 else
653 if (!(in_file->f_mode & FMODE_PREAD))
654 goto fput_in;
655 retval = rw_verify_area(READ, in_file, ppos, count);
656 if (retval)
657 goto fput_in;
659 retval = security_file_permission (in_file, MAY_READ);
660 if (retval)
661 goto fput_in;
664 * Get output file, and verify that it is ok..
666 retval = -EBADF;
667 out_file = fget_light(out_fd, &fput_needed_out);
668 if (!out_file)
669 goto fput_in;
670 if (!(out_file->f_mode & FMODE_WRITE))
671 goto fput_out;
672 retval = -EINVAL;
673 if (!out_file->f_op || !out_file->f_op->sendpage)
674 goto fput_out;
675 out_inode = out_file->f_dentry->d_inode;
676 retval = rw_verify_area(WRITE, out_file, &out_file->f_pos, count);
677 if (retval)
678 goto fput_out;
680 retval = security_file_permission (out_file, MAY_WRITE);
681 if (retval)
682 goto fput_out;
684 if (!max)
685 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
687 pos = *ppos;
688 retval = -EINVAL;
689 if (unlikely(pos < 0))
690 goto fput_out;
691 if (unlikely(pos + count > max)) {
692 retval = -EOVERFLOW;
693 if (pos >= max)
694 goto fput_out;
695 count = max - pos;
698 retval = in_file->f_op->sendfile(in_file, ppos, count, file_send_actor, out_file);
700 if (retval > 0) {
701 current->rchar += retval;
702 current->wchar += retval;
704 current->syscr++;
705 current->syscw++;
707 if (*ppos > max)
708 retval = -EOVERFLOW;
710 fput_out:
711 fput_light(out_file, fput_needed_out);
712 fput_in:
713 fput_light(in_file, fput_needed_in);
714 out:
715 return retval;
718 asmlinkage ssize_t sys_sendfile(int out_fd, int in_fd, off_t __user *offset, size_t count)
720 loff_t pos;
721 off_t off;
722 ssize_t ret;
724 if (offset) {
725 if (unlikely(get_user(off, offset)))
726 return -EFAULT;
727 pos = off;
728 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
729 if (unlikely(put_user(pos, offset)))
730 return -EFAULT;
731 return ret;
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)
739 loff_t pos;
740 ssize_t ret;
742 if (offset) {
743 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
744 return -EFAULT;
745 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
746 if (unlikely(put_user(pos, offset)))
747 return -EFAULT;
748 return ret;
751 return do_sendfile(out_fd, in_fd, NULL, count, 0);