xref: /openbmc/linux/fs/read_write.c (revision 1da177e4)
1 /*
2  *  linux/fs/read_write.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  */
6 
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/dnotify.h>
14 #include <linux/security.h>
15 #include <linux/module.h>
16 #include <linux/syscalls.h>
17 
18 #include <asm/uaccess.h>
19 #include <asm/unistd.h>
20 
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,
26 };
27 
28 EXPORT_SYMBOL(generic_ro_fops);
29 
30 loff_t generic_file_llseek(struct file *file, loff_t offset, int origin)
31 {
32 	long long retval;
33 	struct inode *inode = file->f_mapping->host;
34 
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;
42 	}
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;
48 		}
49 		retval = offset;
50 	}
51 	up(&inode->i_sem);
52 	return retval;
53 }
54 
55 EXPORT_SYMBOL(generic_file_llseek);
56 
57 loff_t remote_llseek(struct file *file, loff_t offset, int origin)
58 {
59 	long long retval;
60 
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;
68 	}
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;
74 		}
75 		retval = offset;
76 	}
77 	unlock_kernel();
78 	return retval;
79 }
80 EXPORT_SYMBOL(remote_llseek);
81 
82 loff_t no_llseek(struct file *file, loff_t offset, int origin)
83 {
84 	return -ESPIPE;
85 }
86 EXPORT_SYMBOL(no_llseek);
87 
88 loff_t default_llseek(struct file *file, loff_t offset, int origin)
89 {
90 	long long retval;
91 
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;
99 	}
100 	retval = -EINVAL;
101 	if (offset >= 0) {
102 		if (offset != file->f_pos) {
103 			file->f_pos = offset;
104 			file->f_version = 0;
105 		}
106 		retval = offset;
107 	}
108 	unlock_kernel();
109 	return retval;
110 }
111 EXPORT_SYMBOL(default_llseek);
112 
113 loff_t vfs_llseek(struct file *file, loff_t offset, int origin)
114 {
115 	loff_t (*fn)(struct file *, loff_t, int);
116 
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;
122 	}
123 	return fn(file, offset, origin);
124 }
125 EXPORT_SYMBOL(vfs_llseek);
126 
127 asmlinkage off_t sys_lseek(unsigned int fd, off_t offset, unsigned int origin)
128 {
129 	off_t retval;
130 	struct file * file;
131 	int fput_needed;
132 
133 	retval = -EBADF;
134 	file = fget_light(fd, &fput_needed);
135 	if (!file)
136 		goto bad;
137 
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 */
144 	}
145 	fput_light(file, fput_needed);
146 bad:
147 	return retval;
148 }
149 
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)
154 {
155 	int retval;
156 	struct file * file;
157 	loff_t offset;
158 	int fput_needed;
159 
160 	retval = -EBADF;
161 	file = fget_light(fd, &fput_needed);
162 	if (!file)
163 		goto bad;
164 
165 	retval = -EINVAL;
166 	if (origin > 2)
167 		goto out_putf;
168 
169 	offset = vfs_llseek(file, ((loff_t) offset_high << 32) | offset_low,
170 			origin);
171 
172 	retval = (int)offset;
173 	if (offset >= 0) {
174 		retval = -EFAULT;
175 		if (!copy_to_user(result, &offset, sizeof(offset)))
176 			retval = 0;
177 	}
178 out_putf:
179 	fput_light(file, fput_needed);
180 bad:
181 	return retval;
182 }
183 #endif
184 
185 
186 int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
187 {
188 	struct inode *inode;
189 	loff_t pos;
190 
191 	if (unlikely(count > file->f_maxcount))
192 		goto Einval;
193 	pos = *ppos;
194 	if (unlikely((pos < 0) || (loff_t) (pos + count) < 0))
195 		goto Einval;
196 
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;
201 
202 Einval:
203 	return -EINVAL;
204 }
205 
206 ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
207 {
208 	struct kiocb kiocb;
209 	ssize_t ret;
210 
211 	init_sync_kiocb(&kiocb, filp);
212 	kiocb.ki_pos = *ppos;
213 	ret = filp->f_op->aio_read(&kiocb, buf, len, kiocb.ki_pos);
214 	if (-EIOCBQUEUED == ret)
215 		ret = wait_on_sync_kiocb(&kiocb);
216 	*ppos = kiocb.ki_pos;
217 	return ret;
218 }
219 
220 EXPORT_SYMBOL(do_sync_read);
221 
222 ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
223 {
224 	ssize_t ret;
225 
226 	if (!(file->f_mode & FMODE_READ))
227 		return -EBADF;
228 	if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
229 		return -EINVAL;
230 	if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
231 		return -EFAULT;
232 
233 	ret = rw_verify_area(READ, file, pos, count);
234 	if (!ret) {
235 		ret = security_file_permission (file, MAY_READ);
236 		if (!ret) {
237 			if (file->f_op->read)
238 				ret = file->f_op->read(file, buf, count, pos);
239 			else
240 				ret = do_sync_read(file, buf, count, pos);
241 			if (ret > 0) {
242 				dnotify_parent(file->f_dentry, DN_ACCESS);
243 				current->rchar += ret;
244 			}
245 			current->syscr++;
246 		}
247 	}
248 
249 	return ret;
250 }
251 
252 EXPORT_SYMBOL(vfs_read);
253 
254 ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
255 {
256 	struct kiocb kiocb;
257 	ssize_t ret;
258 
259 	init_sync_kiocb(&kiocb, filp);
260 	kiocb.ki_pos = *ppos;
261 	ret = filp->f_op->aio_write(&kiocb, buf, len, kiocb.ki_pos);
262 	if (-EIOCBQUEUED == ret)
263 		ret = wait_on_sync_kiocb(&kiocb);
264 	*ppos = kiocb.ki_pos;
265 	return ret;
266 }
267 
268 EXPORT_SYMBOL(do_sync_write);
269 
270 ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
271 {
272 	ssize_t ret;
273 
274 	if (!(file->f_mode & FMODE_WRITE))
275 		return -EBADF;
276 	if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
277 		return -EINVAL;
278 	if (unlikely(!access_ok(VERIFY_READ, buf, count)))
279 		return -EFAULT;
280 
281 	ret = rw_verify_area(WRITE, file, pos, count);
282 	if (!ret) {
283 		ret = security_file_permission (file, MAY_WRITE);
284 		if (!ret) {
285 			if (file->f_op->write)
286 				ret = file->f_op->write(file, buf, count, pos);
287 			else
288 				ret = do_sync_write(file, buf, count, pos);
289 			if (ret > 0) {
290 				dnotify_parent(file->f_dentry, DN_MODIFY);
291 				current->wchar += ret;
292 			}
293 			current->syscw++;
294 		}
295 	}
296 
297 	return ret;
298 }
299 
300 EXPORT_SYMBOL(vfs_write);
301 
302 static inline loff_t file_pos_read(struct file *file)
303 {
304 	return file->f_pos;
305 }
306 
307 static inline void file_pos_write(struct file *file, loff_t pos)
308 {
309 	file->f_pos = pos;
310 }
311 
312 asmlinkage ssize_t sys_read(unsigned int fd, char __user * buf, size_t count)
313 {
314 	struct file *file;
315 	ssize_t ret = -EBADF;
316 	int fput_needed;
317 
318 	file = fget_light(fd, &fput_needed);
319 	if (file) {
320 		loff_t pos = file_pos_read(file);
321 		ret = vfs_read(file, buf, count, &pos);
322 		file_pos_write(file, pos);
323 		fput_light(file, fput_needed);
324 	}
325 
326 	return ret;
327 }
328 EXPORT_SYMBOL_GPL(sys_read);
329 
330 asmlinkage ssize_t sys_write(unsigned int fd, const char __user * buf, size_t count)
331 {
332 	struct file *file;
333 	ssize_t ret = -EBADF;
334 	int fput_needed;
335 
336 	file = fget_light(fd, &fput_needed);
337 	if (file) {
338 		loff_t pos = file_pos_read(file);
339 		ret = vfs_write(file, buf, count, &pos);
340 		file_pos_write(file, pos);
341 		fput_light(file, fput_needed);
342 	}
343 
344 	return ret;
345 }
346 
347 asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf,
348 			     size_t count, loff_t pos)
349 {
350 	struct file *file;
351 	ssize_t ret = -EBADF;
352 	int fput_needed;
353 
354 	if (pos < 0)
355 		return -EINVAL;
356 
357 	file = fget_light(fd, &fput_needed);
358 	if (file) {
359 		ret = -ESPIPE;
360 		if (file->f_mode & FMODE_PREAD)
361 			ret = vfs_read(file, buf, count, &pos);
362 		fput_light(file, fput_needed);
363 	}
364 
365 	return ret;
366 }
367 
368 asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf,
369 			      size_t count, loff_t pos)
370 {
371 	struct file *file;
372 	ssize_t ret = -EBADF;
373 	int fput_needed;
374 
375 	if (pos < 0)
376 		return -EINVAL;
377 
378 	file = fget_light(fd, &fput_needed);
379 	if (file) {
380 		ret = -ESPIPE;
381 		if (file->f_mode & FMODE_PWRITE)
382 			ret = vfs_write(file, buf, count, &pos);
383 		fput_light(file, fput_needed);
384 	}
385 
386 	return ret;
387 }
388 
389 /*
390  * Reduce an iovec's length in-place.  Return the resulting number of segments
391  */
392 unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
393 {
394 	unsigned long seg = 0;
395 	size_t len = 0;
396 
397 	while (seg < nr_segs) {
398 		seg++;
399 		if (len + iov->iov_len >= to) {
400 			iov->iov_len = to - len;
401 			break;
402 		}
403 		len += iov->iov_len;
404 		iov++;
405 	}
406 	return seg;
407 }
408 
409 EXPORT_SYMBOL(iov_shorten);
410 
411 /* A write operation does a read from user space and vice versa */
412 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
413 
414 static ssize_t do_readv_writev(int type, struct file *file,
415 			       const struct iovec __user * uvector,
416 			       unsigned long nr_segs, loff_t *pos)
417 {
418 	typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
419 	typedef ssize_t (*iov_fn_t)(struct file *, const struct iovec *, unsigned long, loff_t *);
420 
421 	size_t tot_len;
422 	struct iovec iovstack[UIO_FASTIOV];
423 	struct iovec *iov=iovstack, *vector;
424 	ssize_t ret;
425 	int seg;
426 	io_fn_t fn;
427 	iov_fn_t fnv;
428 
429 	/*
430 	 * SuS says "The readv() function *may* fail if the iovcnt argument
431 	 * was less than or equal to 0, or greater than {IOV_MAX}.  Linux has
432 	 * traditionally returned zero for zero segments, so...
433 	 */
434 	ret = 0;
435 	if (nr_segs == 0)
436 		goto out;
437 
438 	/*
439 	 * First get the "struct iovec" from user memory and
440 	 * verify all the pointers
441 	 */
442 	ret = -EINVAL;
443 	if ((nr_segs > UIO_MAXIOV) || (nr_segs <= 0))
444 		goto out;
445 	if (!file->f_op)
446 		goto out;
447 	if (nr_segs > UIO_FASTIOV) {
448 		ret = -ENOMEM;
449 		iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
450 		if (!iov)
451 			goto out;
452 	}
453 	ret = -EFAULT;
454 	if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector)))
455 		goto out;
456 
457 	/*
458 	 * Single unix specification:
459 	 * We should -EINVAL if an element length is not >= 0 and fitting an
460 	 * ssize_t.  The total length is fitting an ssize_t
461 	 *
462 	 * Be careful here because iov_len is a size_t not an ssize_t
463 	 */
464 	tot_len = 0;
465 	ret = -EINVAL;
466 	for (seg = 0; seg < nr_segs; seg++) {
467 		void __user *buf = iov[seg].iov_base;
468 		ssize_t len = (ssize_t)iov[seg].iov_len;
469 
470 		if (unlikely(!access_ok(vrfy_dir(type), buf, len)))
471 			goto Efault;
472 		if (len < 0)	/* size_t not fitting an ssize_t .. */
473 			goto out;
474 		tot_len += len;
475 		if ((ssize_t)tot_len < 0) /* maths overflow on the ssize_t */
476 			goto out;
477 	}
478 	if (tot_len == 0) {
479 		ret = 0;
480 		goto out;
481 	}
482 
483 	ret = rw_verify_area(type, file, pos, tot_len);
484 	if (ret)
485 		goto out;
486 
487 	fnv = NULL;
488 	if (type == READ) {
489 		fn = file->f_op->read;
490 		fnv = file->f_op->readv;
491 	} else {
492 		fn = (io_fn_t)file->f_op->write;
493 		fnv = file->f_op->writev;
494 	}
495 	if (fnv) {
496 		ret = fnv(file, iov, nr_segs, pos);
497 		goto out;
498 	}
499 
500 	/* Do it by hand, with file-ops */
501 	ret = 0;
502 	vector = iov;
503 	while (nr_segs > 0) {
504 		void __user * base;
505 		size_t len;
506 		ssize_t nr;
507 
508 		base = vector->iov_base;
509 		len = vector->iov_len;
510 		vector++;
511 		nr_segs--;
512 
513 		nr = fn(file, base, len, pos);
514 
515 		if (nr < 0) {
516 			if (!ret) ret = nr;
517 			break;
518 		}
519 		ret += nr;
520 		if (nr != len)
521 			break;
522 	}
523 out:
524 	if (iov != iovstack)
525 		kfree(iov);
526 	if ((ret + (type == READ)) > 0)
527 		dnotify_parent(file->f_dentry,
528 				(type == READ) ? DN_ACCESS : DN_MODIFY);
529 	return ret;
530 Efault:
531 	ret = -EFAULT;
532 	goto out;
533 }
534 
535 ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
536 		  unsigned long vlen, loff_t *pos)
537 {
538 	if (!(file->f_mode & FMODE_READ))
539 		return -EBADF;
540 	if (!file->f_op || (!file->f_op->readv && !file->f_op->read))
541 		return -EINVAL;
542 
543 	return do_readv_writev(READ, file, vec, vlen, pos);
544 }
545 
546 EXPORT_SYMBOL(vfs_readv);
547 
548 ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
549 		   unsigned long vlen, loff_t *pos)
550 {
551 	if (!(file->f_mode & FMODE_WRITE))
552 		return -EBADF;
553 	if (!file->f_op || (!file->f_op->writev && !file->f_op->write))
554 		return -EINVAL;
555 
556 	return do_readv_writev(WRITE, file, vec, vlen, pos);
557 }
558 
559 EXPORT_SYMBOL(vfs_writev);
560 
561 asmlinkage ssize_t
562 sys_readv(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
563 {
564 	struct file *file;
565 	ssize_t ret = -EBADF;
566 	int fput_needed;
567 
568 	file = fget_light(fd, &fput_needed);
569 	if (file) {
570 		loff_t pos = file_pos_read(file);
571 		ret = vfs_readv(file, vec, vlen, &pos);
572 		file_pos_write(file, pos);
573 		fput_light(file, fput_needed);
574 	}
575 
576 	if (ret > 0)
577 		current->rchar += ret;
578 	current->syscr++;
579 	return ret;
580 }
581 
582 asmlinkage ssize_t
583 sys_writev(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
584 {
585 	struct file *file;
586 	ssize_t ret = -EBADF;
587 	int fput_needed;
588 
589 	file = fget_light(fd, &fput_needed);
590 	if (file) {
591 		loff_t pos = file_pos_read(file);
592 		ret = vfs_writev(file, vec, vlen, &pos);
593 		file_pos_write(file, pos);
594 		fput_light(file, fput_needed);
595 	}
596 
597 	if (ret > 0)
598 		current->wchar += ret;
599 	current->syscw++;
600 	return ret;
601 }
602 
603 static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
604 			   size_t count, loff_t max)
605 {
606 	struct file * in_file, * out_file;
607 	struct inode * in_inode, * out_inode;
608 	loff_t pos;
609 	ssize_t retval;
610 	int fput_needed_in, fput_needed_out;
611 
612 	/*
613 	 * Get input file, and verify that it is ok..
614 	 */
615 	retval = -EBADF;
616 	in_file = fget_light(in_fd, &fput_needed_in);
617 	if (!in_file)
618 		goto out;
619 	if (!(in_file->f_mode & FMODE_READ))
620 		goto fput_in;
621 	retval = -EINVAL;
622 	in_inode = in_file->f_dentry->d_inode;
623 	if (!in_inode)
624 		goto fput_in;
625 	if (!in_file->f_op || !in_file->f_op->sendfile)
626 		goto fput_in;
627 	retval = -ESPIPE;
628 	if (!ppos)
629 		ppos = &in_file->f_pos;
630 	else
631 		if (!(in_file->f_mode & FMODE_PREAD))
632 			goto fput_in;
633 	retval = rw_verify_area(READ, in_file, ppos, count);
634 	if (retval)
635 		goto fput_in;
636 
637 	retval = security_file_permission (in_file, MAY_READ);
638 	if (retval)
639 		goto fput_in;
640 
641 	/*
642 	 * Get output file, and verify that it is ok..
643 	 */
644 	retval = -EBADF;
645 	out_file = fget_light(out_fd, &fput_needed_out);
646 	if (!out_file)
647 		goto fput_in;
648 	if (!(out_file->f_mode & FMODE_WRITE))
649 		goto fput_out;
650 	retval = -EINVAL;
651 	if (!out_file->f_op || !out_file->f_op->sendpage)
652 		goto fput_out;
653 	out_inode = out_file->f_dentry->d_inode;
654 	retval = rw_verify_area(WRITE, out_file, &out_file->f_pos, count);
655 	if (retval)
656 		goto fput_out;
657 
658 	retval = security_file_permission (out_file, MAY_WRITE);
659 	if (retval)
660 		goto fput_out;
661 
662 	if (!max)
663 		max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
664 
665 	pos = *ppos;
666 	retval = -EINVAL;
667 	if (unlikely(pos < 0))
668 		goto fput_out;
669 	if (unlikely(pos + count > max)) {
670 		retval = -EOVERFLOW;
671 		if (pos >= max)
672 			goto fput_out;
673 		count = max - pos;
674 	}
675 
676 	retval = in_file->f_op->sendfile(in_file, ppos, count, file_send_actor, out_file);
677 
678 	if (retval > 0) {
679 		current->rchar += retval;
680 		current->wchar += retval;
681 	}
682 	current->syscr++;
683 	current->syscw++;
684 
685 	if (*ppos > max)
686 		retval = -EOVERFLOW;
687 
688 fput_out:
689 	fput_light(out_file, fput_needed_out);
690 fput_in:
691 	fput_light(in_file, fput_needed_in);
692 out:
693 	return retval;
694 }
695 
696 asmlinkage ssize_t sys_sendfile(int out_fd, int in_fd, off_t __user *offset, size_t count)
697 {
698 	loff_t pos;
699 	off_t off;
700 	ssize_t ret;
701 
702 	if (offset) {
703 		if (unlikely(get_user(off, offset)))
704 			return -EFAULT;
705 		pos = off;
706 		ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
707 		if (unlikely(put_user(pos, offset)))
708 			return -EFAULT;
709 		return ret;
710 	}
711 
712 	return do_sendfile(out_fd, in_fd, NULL, count, 0);
713 }
714 
715 asmlinkage ssize_t sys_sendfile64(int out_fd, int in_fd, loff_t __user *offset, size_t count)
716 {
717 	loff_t pos;
718 	ssize_t ret;
719 
720 	if (offset) {
721 		if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
722 			return -EFAULT;
723 		ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
724 		if (unlikely(put_user(pos, offset)))
725 			return -EFAULT;
726 		return ret;
727 	}
728 
729 	return do_sendfile(out_fd, in_fd, NULL, count, 0);
730 }
731