[PATCH] spanned_pages is not updated at a case of memory hot-add
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / read_write.c
blob5bc0e9234f9d7c5dd21b1e244127fece7bfa905c
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>
17 #include <linux/pagemap.h>
19 #include <asm/uaccess.h>
20 #include <asm/unistd.h>
22 const struct file_operations generic_ro_fops = {
23 .llseek = generic_file_llseek,
24 .read = generic_file_read,
25 .mmap = generic_file_readonly_mmap,
26 .sendfile = generic_file_sendfile,
29 EXPORT_SYMBOL(generic_ro_fops);
31 loff_t generic_file_llseek(struct file *file, loff_t offset, int origin)
33 long long retval;
34 struct inode *inode = file->f_mapping->host;
36 mutex_lock(&inode->i_mutex);
37 switch (origin) {
38 case 2:
39 offset += inode->i_size;
40 break;
41 case 1:
42 offset += file->f_pos;
44 retval = -EINVAL;
45 if (offset>=0 && offset<=inode->i_sb->s_maxbytes) {
46 if (offset != file->f_pos) {
47 file->f_pos = offset;
48 file->f_version = 0;
50 retval = offset;
52 mutex_unlock(&inode->i_mutex);
53 return retval;
56 EXPORT_SYMBOL(generic_file_llseek);
58 loff_t remote_llseek(struct file *file, loff_t offset, int origin)
60 long long retval;
62 lock_kernel();
63 switch (origin) {
64 case 2:
65 offset += i_size_read(file->f_dentry->d_inode);
66 break;
67 case 1:
68 offset += file->f_pos;
70 retval = -EINVAL;
71 if (offset>=0 && offset<=file->f_dentry->d_inode->i_sb->s_maxbytes) {
72 if (offset != file->f_pos) {
73 file->f_pos = offset;
74 file->f_version = 0;
76 retval = offset;
78 unlock_kernel();
79 return retval;
81 EXPORT_SYMBOL(remote_llseek);
83 loff_t no_llseek(struct file *file, loff_t offset, int origin)
85 return -ESPIPE;
87 EXPORT_SYMBOL(no_llseek);
89 loff_t default_llseek(struct file *file, loff_t offset, int origin)
91 long long retval;
93 lock_kernel();
94 switch (origin) {
95 case 2:
96 offset += i_size_read(file->f_dentry->d_inode);
97 break;
98 case 1:
99 offset += file->f_pos;
101 retval = -EINVAL;
102 if (offset >= 0) {
103 if (offset != file->f_pos) {
104 file->f_pos = offset;
105 file->f_version = 0;
107 retval = offset;
109 unlock_kernel();
110 return retval;
112 EXPORT_SYMBOL(default_llseek);
114 loff_t vfs_llseek(struct file *file, loff_t offset, int origin)
116 loff_t (*fn)(struct file *, loff_t, int);
118 fn = no_llseek;
119 if (file->f_mode & FMODE_LSEEK) {
120 fn = default_llseek;
121 if (file->f_op && file->f_op->llseek)
122 fn = file->f_op->llseek;
124 return fn(file, offset, origin);
126 EXPORT_SYMBOL(vfs_llseek);
128 asmlinkage off_t sys_lseek(unsigned int fd, off_t offset, unsigned int origin)
130 off_t retval;
131 struct file * file;
132 int fput_needed;
134 retval = -EBADF;
135 file = fget_light(fd, &fput_needed);
136 if (!file)
137 goto bad;
139 retval = -EINVAL;
140 if (origin <= 2) {
141 loff_t res = vfs_llseek(file, offset, origin);
142 retval = res;
143 if (res != (loff_t)retval)
144 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
146 fput_light(file, fput_needed);
147 bad:
148 return retval;
151 #ifdef __ARCH_WANT_SYS_LLSEEK
152 asmlinkage long sys_llseek(unsigned int fd, unsigned long offset_high,
153 unsigned long offset_low, loff_t __user * result,
154 unsigned int origin)
156 int retval;
157 struct file * file;
158 loff_t offset;
159 int fput_needed;
161 retval = -EBADF;
162 file = fget_light(fd, &fput_needed);
163 if (!file)
164 goto bad;
166 retval = -EINVAL;
167 if (origin > 2)
168 goto out_putf;
170 offset = vfs_llseek(file, ((loff_t) offset_high << 32) | offset_low,
171 origin);
173 retval = (int)offset;
174 if (offset >= 0) {
175 retval = -EFAULT;
176 if (!copy_to_user(result, &offset, sizeof(offset)))
177 retval = 0;
179 out_putf:
180 fput_light(file, fput_needed);
181 bad:
182 return retval;
184 #endif
187 * rw_verify_area doesn't like huge counts. We limit
188 * them to something that fits in "int" so that others
189 * won't have to do range checks all the time.
191 #define MAX_RW_COUNT (INT_MAX & PAGE_CACHE_MASK)
193 int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
195 struct inode *inode;
196 loff_t pos;
198 if (unlikely((ssize_t) count < 0))
199 goto Einval;
200 pos = *ppos;
201 if (unlikely((pos < 0) || (loff_t) (pos + count) < 0))
202 goto Einval;
204 inode = file->f_dentry->d_inode;
205 if (unlikely(inode->i_flock && MANDATORY_LOCK(inode))) {
206 int retval = locks_mandatory_area(
207 read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
208 inode, file, pos, count);
209 if (retval < 0)
210 return retval;
212 return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
214 Einval:
215 return -EINVAL;
218 static void wait_on_retry_sync_kiocb(struct kiocb *iocb)
220 set_current_state(TASK_UNINTERRUPTIBLE);
221 if (!kiocbIsKicked(iocb))
222 schedule();
223 else
224 kiocbClearKicked(iocb);
225 __set_current_state(TASK_RUNNING);
228 ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
230 struct kiocb kiocb;
231 ssize_t ret;
233 init_sync_kiocb(&kiocb, filp);
234 kiocb.ki_pos = *ppos;
235 while (-EIOCBRETRY ==
236 (ret = filp->f_op->aio_read(&kiocb, buf, len, kiocb.ki_pos)))
237 wait_on_retry_sync_kiocb(&kiocb);
239 if (-EIOCBQUEUED == ret)
240 ret = wait_on_sync_kiocb(&kiocb);
241 *ppos = kiocb.ki_pos;
242 return ret;
245 EXPORT_SYMBOL(do_sync_read);
247 ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
249 ssize_t ret;
251 if (!(file->f_mode & FMODE_READ))
252 return -EBADF;
253 if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
254 return -EINVAL;
255 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
256 return -EFAULT;
258 ret = rw_verify_area(READ, file, pos, count);
259 if (ret >= 0) {
260 count = ret;
261 ret = security_file_permission (file, MAY_READ);
262 if (!ret) {
263 if (file->f_op->read)
264 ret = file->f_op->read(file, buf, count, pos);
265 else
266 ret = do_sync_read(file, buf, count, pos);
267 if (ret > 0) {
268 fsnotify_access(file->f_dentry);
269 current->rchar += ret;
271 current->syscr++;
275 return ret;
278 EXPORT_SYMBOL(vfs_read);
280 ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
282 struct kiocb kiocb;
283 ssize_t ret;
285 init_sync_kiocb(&kiocb, filp);
286 kiocb.ki_pos = *ppos;
287 while (-EIOCBRETRY ==
288 (ret = filp->f_op->aio_write(&kiocb, buf, len, kiocb.ki_pos)))
289 wait_on_retry_sync_kiocb(&kiocb);
291 if (-EIOCBQUEUED == ret)
292 ret = wait_on_sync_kiocb(&kiocb);
293 *ppos = kiocb.ki_pos;
294 return ret;
297 EXPORT_SYMBOL(do_sync_write);
299 ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
301 ssize_t ret;
303 if (!(file->f_mode & FMODE_WRITE))
304 return -EBADF;
305 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
306 return -EINVAL;
307 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
308 return -EFAULT;
310 ret = rw_verify_area(WRITE, file, pos, count);
311 if (ret >= 0) {
312 count = ret;
313 ret = security_file_permission (file, MAY_WRITE);
314 if (!ret) {
315 if (file->f_op->write)
316 ret = file->f_op->write(file, buf, count, pos);
317 else
318 ret = do_sync_write(file, buf, count, pos);
319 if (ret > 0) {
320 fsnotify_modify(file->f_dentry);
321 current->wchar += ret;
323 current->syscw++;
327 return ret;
330 EXPORT_SYMBOL(vfs_write);
332 static inline loff_t file_pos_read(struct file *file)
334 return file->f_pos;
337 static inline void file_pos_write(struct file *file, loff_t pos)
339 file->f_pos = pos;
342 asmlinkage ssize_t sys_read(unsigned int fd, char __user * buf, size_t count)
344 struct file *file;
345 ssize_t ret = -EBADF;
346 int fput_needed;
348 file = fget_light(fd, &fput_needed);
349 if (file) {
350 loff_t pos = file_pos_read(file);
351 ret = vfs_read(file, buf, count, &pos);
352 file_pos_write(file, pos);
353 fput_light(file, fput_needed);
356 return ret;
358 EXPORT_SYMBOL_GPL(sys_read);
360 asmlinkage ssize_t sys_write(unsigned int fd, const char __user * buf, size_t count)
362 struct file *file;
363 ssize_t ret = -EBADF;
364 int fput_needed;
366 file = fget_light(fd, &fput_needed);
367 if (file) {
368 loff_t pos = file_pos_read(file);
369 ret = vfs_write(file, buf, count, &pos);
370 file_pos_write(file, pos);
371 fput_light(file, fput_needed);
374 return ret;
377 asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf,
378 size_t count, loff_t pos)
380 struct file *file;
381 ssize_t ret = -EBADF;
382 int fput_needed;
384 if (pos < 0)
385 return -EINVAL;
387 file = fget_light(fd, &fput_needed);
388 if (file) {
389 ret = -ESPIPE;
390 if (file->f_mode & FMODE_PREAD)
391 ret = vfs_read(file, buf, count, &pos);
392 fput_light(file, fput_needed);
395 return ret;
398 asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf,
399 size_t count, loff_t pos)
401 struct file *file;
402 ssize_t ret = -EBADF;
403 int fput_needed;
405 if (pos < 0)
406 return -EINVAL;
408 file = fget_light(fd, &fput_needed);
409 if (file) {
410 ret = -ESPIPE;
411 if (file->f_mode & FMODE_PWRITE)
412 ret = vfs_write(file, buf, count, &pos);
413 fput_light(file, fput_needed);
416 return ret;
420 * Reduce an iovec's length in-place. Return the resulting number of segments
422 unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
424 unsigned long seg = 0;
425 size_t len = 0;
427 while (seg < nr_segs) {
428 seg++;
429 if (len + iov->iov_len >= to) {
430 iov->iov_len = to - len;
431 break;
433 len += iov->iov_len;
434 iov++;
436 return seg;
439 EXPORT_SYMBOL(iov_shorten);
441 /* A write operation does a read from user space and vice versa */
442 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
444 static ssize_t do_readv_writev(int type, struct file *file,
445 const struct iovec __user * uvector,
446 unsigned long nr_segs, loff_t *pos)
448 typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
449 typedef ssize_t (*iov_fn_t)(struct file *, const struct iovec *, unsigned long, loff_t *);
451 size_t tot_len;
452 struct iovec iovstack[UIO_FASTIOV];
453 struct iovec *iov=iovstack, *vector;
454 ssize_t ret;
455 int seg;
456 io_fn_t fn;
457 iov_fn_t fnv;
460 * SuS says "The readv() function *may* fail if the iovcnt argument
461 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
462 * traditionally returned zero for zero segments, so...
464 ret = 0;
465 if (nr_segs == 0)
466 goto out;
469 * First get the "struct iovec" from user memory and
470 * verify all the pointers
472 ret = -EINVAL;
473 if (nr_segs > UIO_MAXIOV)
474 goto out;
475 if (!file->f_op)
476 goto out;
477 if (nr_segs > UIO_FASTIOV) {
478 ret = -ENOMEM;
479 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
480 if (!iov)
481 goto out;
483 ret = -EFAULT;
484 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector)))
485 goto out;
488 * Single unix specification:
489 * We should -EINVAL if an element length is not >= 0 and fitting an
490 * ssize_t. The total length is fitting an ssize_t
492 * Be careful here because iov_len is a size_t not an ssize_t
494 tot_len = 0;
495 ret = -EINVAL;
496 for (seg = 0; seg < nr_segs; seg++) {
497 void __user *buf = iov[seg].iov_base;
498 ssize_t len = (ssize_t)iov[seg].iov_len;
500 if (len < 0) /* size_t not fitting an ssize_t .. */
501 goto out;
502 if (unlikely(!access_ok(vrfy_dir(type), buf, len)))
503 goto Efault;
504 tot_len += len;
505 if ((ssize_t)tot_len < 0) /* maths overflow on the ssize_t */
506 goto out;
508 if (tot_len == 0) {
509 ret = 0;
510 goto out;
513 ret = rw_verify_area(type, file, pos, tot_len);
514 if (ret < 0)
515 goto out;
516 ret = security_file_permission(file, type == READ ? MAY_READ : MAY_WRITE);
517 if (ret)
518 goto out;
520 fnv = NULL;
521 if (type == READ) {
522 fn = file->f_op->read;
523 fnv = file->f_op->readv;
524 } else {
525 fn = (io_fn_t)file->f_op->write;
526 fnv = file->f_op->writev;
528 if (fnv) {
529 ret = fnv(file, iov, nr_segs, pos);
530 goto out;
533 /* Do it by hand, with file-ops */
534 ret = 0;
535 vector = iov;
536 while (nr_segs > 0) {
537 void __user * base;
538 size_t len;
539 ssize_t nr;
541 base = vector->iov_base;
542 len = vector->iov_len;
543 vector++;
544 nr_segs--;
546 nr = fn(file, base, len, pos);
548 if (nr < 0) {
549 if (!ret) ret = nr;
550 break;
552 ret += nr;
553 if (nr != len)
554 break;
556 out:
557 if (iov != iovstack)
558 kfree(iov);
559 if ((ret + (type == READ)) > 0) {
560 if (type == READ)
561 fsnotify_access(file->f_dentry);
562 else
563 fsnotify_modify(file->f_dentry);
565 return ret;
566 Efault:
567 ret = -EFAULT;
568 goto out;
571 ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
572 unsigned long vlen, loff_t *pos)
574 if (!(file->f_mode & FMODE_READ))
575 return -EBADF;
576 if (!file->f_op || (!file->f_op->readv && !file->f_op->read))
577 return -EINVAL;
579 return do_readv_writev(READ, file, vec, vlen, pos);
582 EXPORT_SYMBOL(vfs_readv);
584 ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
585 unsigned long vlen, loff_t *pos)
587 if (!(file->f_mode & FMODE_WRITE))
588 return -EBADF;
589 if (!file->f_op || (!file->f_op->writev && !file->f_op->write))
590 return -EINVAL;
592 return do_readv_writev(WRITE, file, vec, vlen, pos);
595 EXPORT_SYMBOL(vfs_writev);
597 asmlinkage ssize_t
598 sys_readv(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
600 struct file *file;
601 ssize_t ret = -EBADF;
602 int fput_needed;
604 file = fget_light(fd, &fput_needed);
605 if (file) {
606 loff_t pos = file_pos_read(file);
607 ret = vfs_readv(file, vec, vlen, &pos);
608 file_pos_write(file, pos);
609 fput_light(file, fput_needed);
612 if (ret > 0)
613 current->rchar += ret;
614 current->syscr++;
615 return ret;
618 asmlinkage ssize_t
619 sys_writev(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
621 struct file *file;
622 ssize_t ret = -EBADF;
623 int fput_needed;
625 file = fget_light(fd, &fput_needed);
626 if (file) {
627 loff_t pos = file_pos_read(file);
628 ret = vfs_writev(file, vec, vlen, &pos);
629 file_pos_write(file, pos);
630 fput_light(file, fput_needed);
633 if (ret > 0)
634 current->wchar += ret;
635 current->syscw++;
636 return ret;
639 static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
640 size_t count, loff_t max)
642 struct file * in_file, * out_file;
643 struct inode * in_inode, * out_inode;
644 loff_t pos;
645 ssize_t retval;
646 int fput_needed_in, fput_needed_out;
649 * Get input file, and verify that it is ok..
651 retval = -EBADF;
652 in_file = fget_light(in_fd, &fput_needed_in);
653 if (!in_file)
654 goto out;
655 if (!(in_file->f_mode & FMODE_READ))
656 goto fput_in;
657 retval = -EINVAL;
658 in_inode = in_file->f_dentry->d_inode;
659 if (!in_inode)
660 goto fput_in;
661 if (!in_file->f_op || !in_file->f_op->sendfile)
662 goto fput_in;
663 retval = -ESPIPE;
664 if (!ppos)
665 ppos = &in_file->f_pos;
666 else
667 if (!(in_file->f_mode & FMODE_PREAD))
668 goto fput_in;
669 retval = rw_verify_area(READ, in_file, ppos, count);
670 if (retval < 0)
671 goto fput_in;
672 count = retval;
674 retval = security_file_permission (in_file, MAY_READ);
675 if (retval)
676 goto fput_in;
679 * Get output file, and verify that it is ok..
681 retval = -EBADF;
682 out_file = fget_light(out_fd, &fput_needed_out);
683 if (!out_file)
684 goto fput_in;
685 if (!(out_file->f_mode & FMODE_WRITE))
686 goto fput_out;
687 retval = -EINVAL;
688 if (!out_file->f_op || !out_file->f_op->sendpage)
689 goto fput_out;
690 out_inode = out_file->f_dentry->d_inode;
691 retval = rw_verify_area(WRITE, out_file, &out_file->f_pos, count);
692 if (retval < 0)
693 goto fput_out;
694 count = retval;
696 retval = security_file_permission (out_file, MAY_WRITE);
697 if (retval)
698 goto fput_out;
700 if (!max)
701 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
703 pos = *ppos;
704 retval = -EINVAL;
705 if (unlikely(pos < 0))
706 goto fput_out;
707 if (unlikely(pos + count > max)) {
708 retval = -EOVERFLOW;
709 if (pos >= max)
710 goto fput_out;
711 count = max - pos;
714 retval = in_file->f_op->sendfile(in_file, ppos, count, file_send_actor, out_file);
716 if (retval > 0) {
717 current->rchar += retval;
718 current->wchar += retval;
720 current->syscr++;
721 current->syscw++;
723 if (*ppos > max)
724 retval = -EOVERFLOW;
726 fput_out:
727 fput_light(out_file, fput_needed_out);
728 fput_in:
729 fput_light(in_file, fput_needed_in);
730 out:
731 return retval;
734 asmlinkage ssize_t sys_sendfile(int out_fd, int in_fd, off_t __user *offset, size_t count)
736 loff_t pos;
737 off_t off;
738 ssize_t ret;
740 if (offset) {
741 if (unlikely(get_user(off, offset)))
742 return -EFAULT;
743 pos = off;
744 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
745 if (unlikely(put_user(pos, offset)))
746 return -EFAULT;
747 return ret;
750 return do_sendfile(out_fd, in_fd, NULL, count, 0);
753 asmlinkage ssize_t sys_sendfile64(int out_fd, int in_fd, loff_t __user *offset, size_t count)
755 loff_t pos;
756 ssize_t ret;
758 if (offset) {
759 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
760 return -EFAULT;
761 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
762 if (unlikely(put_user(pos, offset)))
763 return -EFAULT;
764 return ret;
767 return do_sendfile(out_fd, in_fd, NULL, count, 0);