read_write.c 26.7 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10 11
/*
 *  linux/fs/read_write.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 */

#include <linux/slab.h> 
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/file.h>
#include <linux/uio.h>
12
#include <linux/aio.h>
Robert Love's avatar
Robert Love committed
13
#include <linux/fsnotify.h>
Linus Torvalds's avatar
Linus Torvalds committed
14
#include <linux/security.h>
15
#include <linux/export.h>
Linus Torvalds's avatar
Linus Torvalds committed
16
#include <linux/syscalls.h>
17
#include <linux/pagemap.h>
18
#include <linux/splice.h>
19
#include <linux/compat.h>
20
#include "internal.h"
Linus Torvalds's avatar
Linus Torvalds committed
21 22 23 24

#include <asm/uaccess.h>
#include <asm/unistd.h>

Al Viro's avatar
Al Viro committed
25 26 27 28
typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
typedef ssize_t (*iov_fn_t)(struct kiocb *, const struct iovec *,
		unsigned long, loff_t);

29
const struct file_operations generic_ro_fops = {
Linus Torvalds's avatar
Linus Torvalds committed
30
	.llseek		= generic_file_llseek,
31 32
	.read		= do_sync_read,
	.aio_read	= generic_file_aio_read,
Linus Torvalds's avatar
Linus Torvalds committed
33
	.mmap		= generic_file_readonly_mmap,
34
	.splice_read	= generic_file_splice_read,
Linus Torvalds's avatar
Linus Torvalds committed
35 36 37 38
};

EXPORT_SYMBOL(generic_ro_fops);

39
static inline int unsigned_offsets(struct file *file)
40
{
41
	return file->f_mode & FMODE_UNSIGNED_OFFSET;
42 43
}

44 45 46 47 48 49 50 51 52 53 54 55 56 57 58
static loff_t lseek_execute(struct file *file, struct inode *inode,
		loff_t offset, loff_t maxsize)
{
	if (offset < 0 && !unsigned_offsets(file))
		return -EINVAL;
	if (offset > maxsize)
		return -EINVAL;

	if (offset != file->f_pos) {
		file->f_pos = offset;
		file->f_version = 0;
	}
	return offset;
}

59
/**
60
 * generic_file_llseek_size - generic llseek implementation for regular files
61 62
 * @file:	file structure to seek on
 * @offset:	file offset to seek to
63
 * @whence:	type of seek
64 65
 * @size:	max size of this file in file system
 * @eof:	offset used for SEEK_END position
66
 *
67
 * This is a variant of generic_file_llseek that allows passing in a custom
68
 * maximum file size and a custom EOF position, for e.g. hashed directories
69 70
 *
 * Synchronization:
71
 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
72 73
 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
 * read/writes behave like SEEK_SET against seeks.
74
 */
75
loff_t
76
generic_file_llseek_size(struct file *file, loff_t offset, int whence,
77
		loff_t maxsize, loff_t eof)
Linus Torvalds's avatar
Linus Torvalds committed
78 79 80
{
	struct inode *inode = file->f_mapping->host;

81
	switch (whence) {
82
	case SEEK_END:
83
		offset += eof;
84 85
		break;
	case SEEK_CUR:
86 87 88 89 90 91 92 93
		/*
		 * Here we special-case the lseek(fd, 0, SEEK_CUR)
		 * position-querying operation.  Avoid rewriting the "same"
		 * f_pos value back to the file because a concurrent read(),
		 * write() or lseek() might have altered it
		 */
		if (offset == 0)
			return file->f_pos;
94 95 96 97 98 99 100
		/*
		 * f_lock protects against read/modify/write race with other
		 * SEEK_CURs. Note that parallel writes and reads behave
		 * like SEEK_SET.
		 */
		spin_lock(&file->f_lock);
		offset = lseek_execute(file, inode, file->f_pos + offset,
101
				       maxsize);
102 103
		spin_unlock(&file->f_lock);
		return offset;
104 105 106 107 108
	case SEEK_DATA:
		/*
		 * In the generic case the entire file is data, so as long as
		 * offset isn't at the end of the file then the offset is data.
		 */
109
		if (offset >= eof)
110 111 112 113 114 115 116
			return -ENXIO;
		break;
	case SEEK_HOLE:
		/*
		 * There is a virtual hole at the end of the file, so as long as
		 * offset isn't i_size or larger, return i_size.
		 */
117
		if (offset >= eof)
118
			return -ENXIO;
119
		offset = eof;
120
		break;
Linus Torvalds's avatar
Linus Torvalds committed
121
	}
122

123 124 125 126 127 128 129 130
	return lseek_execute(file, inode, offset, maxsize);
}
EXPORT_SYMBOL(generic_file_llseek_size);

/**
 * generic_file_llseek - generic llseek implementation for regular files
 * @file:	file structure to seek on
 * @offset:	file offset to seek to
131
 * @whence:	type of seek
132 133 134
 *
 * This is a generic implemenation of ->llseek useable for all normal local
 * filesystems.  It just updates the file offset to the value specified by
135
 * @offset and @whence.
136
 */
137
loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
138 139 140
{
	struct inode *inode = file->f_mapping->host;

141
	return generic_file_llseek_size(file, offset, whence,
142 143
					inode->i_sb->s_maxbytes,
					i_size_read(inode));
Linus Torvalds's avatar
Linus Torvalds committed
144
}
145
EXPORT_SYMBOL(generic_file_llseek);
Linus Torvalds's avatar
Linus Torvalds committed
146

147 148 149 150
/**
 * noop_llseek - No Operation Performed llseek implementation
 * @file:	file structure to seek on
 * @offset:	file offset to seek to
151
 * @whence:	type of seek
152 153 154 155 156 157
 *
 * This is an implementation of ->llseek useable for the rare special case when
 * userspace expects the seek to succeed but the (device) file is actually not
 * able to perform the seek. In this case you use noop_llseek() instead of
 * falling back to the default implementation of ->llseek.
 */
158
loff_t noop_llseek(struct file *file, loff_t offset, int whence)
159 160 161 162 163
{
	return file->f_pos;
}
EXPORT_SYMBOL(noop_llseek);

164
loff_t no_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds's avatar
Linus Torvalds committed
165 166 167 168 169
{
	return -ESPIPE;
}
EXPORT_SYMBOL(no_llseek);

170
loff_t default_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds's avatar
Linus Torvalds committed
171
{
Al Viro's avatar
Al Viro committed
172
	struct inode *inode = file_inode(file);
173
	loff_t retval;
Linus Torvalds's avatar
Linus Torvalds committed
174

175
	mutex_lock(&inode->i_mutex);
176
	switch (whence) {
177
		case SEEK_END:
178
			offset += i_size_read(inode);
Linus Torvalds's avatar
Linus Torvalds committed
179
			break;
180
		case SEEK_CUR:
181 182 183 184
			if (offset == 0) {
				retval = file->f_pos;
				goto out;
			}
Linus Torvalds's avatar
Linus Torvalds committed
185
			offset += file->f_pos;
186 187 188 189 190 191 192
			break;
		case SEEK_DATA:
			/*
			 * In the generic case the entire file is data, so as
			 * long as offset isn't at the end of the file then the
			 * offset is data.
			 */
193 194 195 196
			if (offset >= inode->i_size) {
				retval = -ENXIO;
				goto out;
			}
197 198 199 200 201 202 203
			break;
		case SEEK_HOLE:
			/*
			 * There is a virtual hole at the end of the file, so
			 * as long as offset isn't i_size or larger, return
			 * i_size.
			 */
204 205 206 207
			if (offset >= inode->i_size) {
				retval = -ENXIO;
				goto out;
			}
208 209
			offset = inode->i_size;
			break;
Linus Torvalds's avatar
Linus Torvalds committed
210 211
	}
	retval = -EINVAL;
212
	if (offset >= 0 || unsigned_offsets(file)) {
Linus Torvalds's avatar
Linus Torvalds committed
213 214 215 216 217 218
		if (offset != file->f_pos) {
			file->f_pos = offset;
			file->f_version = 0;
		}
		retval = offset;
	}
219
out:
220
	mutex_unlock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
221 222 223 224
	return retval;
}
EXPORT_SYMBOL(default_llseek);

225
loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds's avatar
Linus Torvalds committed
226 227 228 229 230 231 232 233
{
	loff_t (*fn)(struct file *, loff_t, int);

	fn = no_llseek;
	if (file->f_mode & FMODE_LSEEK) {
		if (file->f_op && file->f_op->llseek)
			fn = file->f_op->llseek;
	}
234
	return fn(file, offset, whence);
Linus Torvalds's avatar
Linus Torvalds committed
235 236 237
}
EXPORT_SYMBOL(vfs_llseek);

238
SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
Linus Torvalds's avatar
Linus Torvalds committed
239 240
{
	off_t retval;
241 242 243
	struct fd f = fdget(fd);
	if (!f.file)
		return -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
244 245

	retval = -EINVAL;
246 247
	if (whence <= SEEK_MAX) {
		loff_t res = vfs_llseek(f.file, offset, whence);
Linus Torvalds's avatar
Linus Torvalds committed
248 249 250 251
		retval = res;
		if (res != (loff_t)retval)
			retval = -EOVERFLOW;	/* LFS: should only happen on 32 bit platforms */
	}
252
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
253 254 255
	return retval;
}

256 257 258 259 260 261 262
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
{
	return sys_lseek(fd, offset, whence);
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
263
#ifdef __ARCH_WANT_SYS_LLSEEK
264 265
SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
		unsigned long, offset_low, loff_t __user *, result,
266
		unsigned int, whence)
Linus Torvalds's avatar
Linus Torvalds committed
267 268
{
	int retval;
269
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
270 271
	loff_t offset;

272 273
	if (!f.file)
		return -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
274 275

	retval = -EINVAL;
276
	if (whence > SEEK_MAX)
Linus Torvalds's avatar
Linus Torvalds committed
277 278
		goto out_putf;

279
	offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
280
			whence);
Linus Torvalds's avatar
Linus Torvalds committed
281 282 283 284 285 286 287 288

	retval = (int)offset;
	if (offset >= 0) {
		retval = -EFAULT;
		if (!copy_to_user(result, &offset, sizeof(offset)))
			retval = 0;
	}
out_putf:
289
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
290 291 292 293
	return retval;
}
#endif

294 295 296 297 298
/*
 * rw_verify_area doesn't like huge counts. We limit
 * them to something that fits in "int" so that others
 * won't have to do range checks all the time.
 */
Linus Torvalds's avatar
Linus Torvalds committed
299 300 301 302
int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
{
	struct inode *inode;
	loff_t pos;
303
	int retval = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
304

Al Viro's avatar
Al Viro committed
305
	inode = file_inode(file);
306
	if (unlikely((ssize_t) count < 0))
307
		return retval;
Linus Torvalds's avatar
Linus Torvalds committed
308
	pos = *ppos;
309 310 311 312 313 314 315
	if (unlikely(pos < 0)) {
		if (!unsigned_offsets(file))
			return retval;
		if (count >= -pos) /* both values are in 0..LLONG_MAX */
			return -EOVERFLOW;
	} else if (unlikely((loff_t) (pos + count) < 0)) {
		if (!unsigned_offsets(file))
316 317
			return retval;
	}
Linus Torvalds's avatar
Linus Torvalds committed
318

319
	if (unlikely(inode->i_flock && mandatory_lock(inode))) {
320
		retval = locks_mandatory_area(
321 322 323 324 325
			read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
			inode, file, pos, count);
		if (retval < 0)
			return retval;
	}
326 327 328 329
	retval = security_file_permission(file,
				read_write == READ ? MAY_READ : MAY_WRITE);
	if (retval)
		return retval;
330
	return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
Linus Torvalds's avatar
Linus Torvalds committed
331 332 333 334
}

ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
{
335
	struct iovec iov = { .iov_base = buf, .iov_len = len };
Linus Torvalds's avatar
Linus Torvalds committed
336 337 338 339 340
	struct kiocb kiocb;
	ssize_t ret;

	init_sync_kiocb(&kiocb, filp);
	kiocb.ki_pos = *ppos;
341
	kiocb.ki_left = len;
342
	kiocb.ki_nbytes = len;
343

Zach Brown's avatar
Zach Brown committed
344
	ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
Linus Torvalds's avatar
Linus Torvalds committed
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364
	if (-EIOCBQUEUED == ret)
		ret = wait_on_sync_kiocb(&kiocb);
	*ppos = kiocb.ki_pos;
	return ret;
}

EXPORT_SYMBOL(do_sync_read);

ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
{
	ssize_t ret;

	if (!(file->f_mode & FMODE_READ))
		return -EBADF;
	if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
		return -EINVAL;
	if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
		return -EFAULT;

	ret = rw_verify_area(READ, file, pos, count);
365 366
	if (ret >= 0) {
		count = ret;
367 368 369 370 371
		if (file->f_op->read)
			ret = file->f_op->read(file, buf, count, pos);
		else
			ret = do_sync_read(file, buf, count, pos);
		if (ret > 0) {
372
			fsnotify_access(file);
373
			add_rchar(current, ret);
Linus Torvalds's avatar
Linus Torvalds committed
374
		}
375
		inc_syscr(current);
Linus Torvalds's avatar
Linus Torvalds committed
376 377 378 379 380 381 382 383 384
	}

	return ret;
}

EXPORT_SYMBOL(vfs_read);

ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
{
385
	struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
Linus Torvalds's avatar
Linus Torvalds committed
386 387 388 389 390
	struct kiocb kiocb;
	ssize_t ret;

	init_sync_kiocb(&kiocb, filp);
	kiocb.ki_pos = *ppos;
391
	kiocb.ki_left = len;
392
	kiocb.ki_nbytes = len;
393

Zach Brown's avatar
Zach Brown committed
394
	ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
Linus Torvalds's avatar
Linus Torvalds committed
395 396 397 398 399 400 401 402
	if (-EIOCBQUEUED == ret)
		ret = wait_on_sync_kiocb(&kiocb);
	*ppos = kiocb.ki_pos;
	return ret;
}

EXPORT_SYMBOL(do_sync_write);

403 404 405 406 407 408
ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
{
	mm_segment_t old_fs;
	const char __user *p;
	ssize_t ret;

409 410 411
	if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
		return -EINVAL;

412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
	old_fs = get_fs();
	set_fs(get_ds());
	p = (__force const char __user *)buf;
	if (count > MAX_RW_COUNT)
		count =  MAX_RW_COUNT;
	if (file->f_op->write)
		ret = file->f_op->write(file, p, count, pos);
	else
		ret = do_sync_write(file, p, count, pos);
	set_fs(old_fs);
	if (ret > 0) {
		fsnotify_modify(file);
		add_wchar(current, ret);
	}
	inc_syscw(current);
	return ret;
}

Linus Torvalds's avatar
Linus Torvalds committed
430 431 432 433 434 435 436 437 438 439 440 441
ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
{
	ssize_t ret;

	if (!(file->f_mode & FMODE_WRITE))
		return -EBADF;
	if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
		return -EINVAL;
	if (unlikely(!access_ok(VERIFY_READ, buf, count)))
		return -EFAULT;

	ret = rw_verify_area(WRITE, file, pos, count);
442 443
	if (ret >= 0) {
		count = ret;
444
		file_start_write(file);
445 446 447 448 449
		if (file->f_op->write)
			ret = file->f_op->write(file, buf, count, pos);
		else
			ret = do_sync_write(file, buf, count, pos);
		if (ret > 0) {
450
			fsnotify_modify(file);
451
			add_wchar(current, ret);
Linus Torvalds's avatar
Linus Torvalds committed
452
		}
453
		inc_syscw(current);
454
		file_end_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
	}

	return ret;
}

EXPORT_SYMBOL(vfs_write);

static inline loff_t file_pos_read(struct file *file)
{
	return file->f_pos;
}

static inline void file_pos_write(struct file *file, loff_t pos)
{
	file->f_pos = pos;
}

472
SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
Linus Torvalds's avatar
Linus Torvalds committed
473
{
474
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
475 476
	ssize_t ret = -EBADF;

477 478 479 480 481
	if (f.file) {
		loff_t pos = file_pos_read(f.file);
		ret = vfs_read(f.file, buf, count, &pos);
		file_pos_write(f.file, pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
482 483 484 485
	}
	return ret;
}

486 487
SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
		size_t, count)
Linus Torvalds's avatar
Linus Torvalds committed
488
{
489
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
490 491
	ssize_t ret = -EBADF;

492 493 494 495 496
	if (f.file) {
		loff_t pos = file_pos_read(f.file);
		ret = vfs_write(f.file, buf, count, &pos);
		file_pos_write(f.file, pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
497 498 499 500 501
	}

	return ret;
}

502 503
SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
			size_t, count, loff_t, pos)
Linus Torvalds's avatar
Linus Torvalds committed
504
{
505
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
506 507 508 509 510
	ssize_t ret = -EBADF;

	if (pos < 0)
		return -EINVAL;

511 512
	f = fdget(fd);
	if (f.file) {
Linus Torvalds's avatar
Linus Torvalds committed
513
		ret = -ESPIPE;
514 515 516
		if (f.file->f_mode & FMODE_PREAD)
			ret = vfs_read(f.file, buf, count, &pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
517 518 519 520 521
	}

	return ret;
}

522 523
SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
			 size_t, count, loff_t, pos)
Linus Torvalds's avatar
Linus Torvalds committed
524
{
525
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
526 527 528 529 530
	ssize_t ret = -EBADF;

	if (pos < 0)
		return -EINVAL;

531 532
	f = fdget(fd);
	if (f.file) {
Linus Torvalds's avatar
Linus Torvalds committed
533
		ret = -ESPIPE;
534 535 536
		if (f.file->f_mode & FMODE_PWRITE)  
			ret = vfs_write(f.file, buf, count, &pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560
	}

	return ret;
}

/*
 * Reduce an iovec's length in-place.  Return the resulting number of segments
 */
unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
{
	unsigned long seg = 0;
	size_t len = 0;

	while (seg < nr_segs) {
		seg++;
		if (len + iov->iov_len >= to) {
			iov->iov_len = to - len;
			break;
		}
		len += iov->iov_len;
		iov++;
	}
	return seg;
}
561
EXPORT_SYMBOL(iov_shorten);
Linus Torvalds's avatar
Linus Torvalds committed
562

563
static ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
564 565 566 567 568 569 570 571 572 573
		unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
{
	struct kiocb kiocb;
	ssize_t ret;

	init_sync_kiocb(&kiocb, filp);
	kiocb.ki_pos = *ppos;
	kiocb.ki_left = len;
	kiocb.ki_nbytes = len;

Zach Brown's avatar
Zach Brown committed
574
	ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
575 576 577 578 579 580 581
	if (ret == -EIOCBQUEUED)
		ret = wait_on_sync_kiocb(&kiocb);
	*ppos = kiocb.ki_pos;
	return ret;
}

/* Do it by hand, with file-ops */
582
static ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
		unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
{
	struct iovec *vector = iov;
	ssize_t ret = 0;

	while (nr_segs > 0) {
		void __user *base;
		size_t len;
		ssize_t nr;

		base = vector->iov_base;
		len = vector->iov_len;
		vector++;
		nr_segs--;

		nr = fn(filp, base, len, ppos);

		if (nr < 0) {
			if (!ret)
				ret = nr;
			break;
		}
		ret += nr;
		if (nr != len)
			break;
	}

	return ret;
}

Linus Torvalds's avatar
Linus Torvalds committed
613 614 615
/* A write operation does a read from user space and vice versa */
#define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)

616 617 618
ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
			      unsigned long nr_segs, unsigned long fast_segs,
			      struct iovec *fast_pointer,
619
			      struct iovec **ret_pointer)
620
{
621
	unsigned long seg;
622
	ssize_t ret;
623 624
	struct iovec *iov = fast_pointer;

625 626 627 628 629
	/*
	 * SuS says "The readv() function *may* fail if the iovcnt argument
	 * was less than or equal to 0, or greater than {IOV_MAX}.  Linux has
	 * traditionally returned zero for zero segments, so...
	 */
630 631
	if (nr_segs == 0) {
		ret = 0;
632
		goto out;
633 634
	}

635 636 637 638
	/*
	 * First get the "struct iovec" from user memory and
	 * verify all the pointers
	 */
639 640
	if (nr_segs > UIO_MAXIOV) {
		ret = -EINVAL;
641
		goto out;
642 643
	}
	if (nr_segs > fast_segs) {
644
		iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
645 646
		if (iov == NULL) {
			ret = -ENOMEM;
647
			goto out;
648
		}
649
	}
650 651
	if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
		ret = -EFAULT;
652
		goto out;
653 654
	}

655
	/*
656 657 658 659
	 * According to the Single Unix Specification we should return EINVAL
	 * if an element length is < 0 when cast to ssize_t or if the
	 * total length would overflow the ssize_t return value of the
	 * system call.
660 661 662 663
	 *
	 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
	 * overflow case.
	 */
664
	ret = 0;
665 666 667
	for (seg = 0; seg < nr_segs; seg++) {
		void __user *buf = iov[seg].iov_base;
		ssize_t len = (ssize_t)iov[seg].iov_len;
668 669 670

		/* see if we we're about to use an invalid len or if
		 * it's about to overflow ssize_t */
671
		if (len < 0) {
672
			ret = -EINVAL;
673
			goto out;
674
		}
675
		if (type >= 0
Christopher Yeoh's avatar
Christopher Yeoh committed
676
		    && unlikely(!access_ok(vrfy_dir(type), buf, len))) {
677
			ret = -EFAULT;
678 679 680 681 682
			goto out;
		}
		if (len > MAX_RW_COUNT - ret) {
			len = MAX_RW_COUNT - ret;
			iov[seg].iov_len = len;
683 684
		}
		ret += len;
685
	}
686 687 688 689 690
out:
	*ret_pointer = iov;
	return ret;
}

Linus Torvalds's avatar
Linus Torvalds committed
691 692 693 694 695 696
static ssize_t do_readv_writev(int type, struct file *file,
			       const struct iovec __user * uvector,
			       unsigned long nr_segs, loff_t *pos)
{
	size_t tot_len;
	struct iovec iovstack[UIO_FASTIOV];
697
	struct iovec *iov = iovstack;
Linus Torvalds's avatar
Linus Torvalds committed
698 699 700 701
	ssize_t ret;
	io_fn_t fn;
	iov_fn_t fnv;

702 703
	if (!file->f_op) {
		ret = -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
704 705 706
		goto out;
	}

707
	ret = rw_copy_check_uvector(type, uvector, nr_segs,
708
				    ARRAY_SIZE(iovstack), iovstack, &iov);
709
	if (ret <= 0)
Linus Torvalds's avatar
Linus Torvalds committed
710 711
		goto out;

712
	tot_len = ret;
Linus Torvalds's avatar
Linus Torvalds committed
713
	ret = rw_verify_area(type, file, pos, tot_len);
714
	if (ret < 0)
715
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
716 717 718 719

	fnv = NULL;
	if (type == READ) {
		fn = file->f_op->read;
720
		fnv = file->f_op->aio_read;
Linus Torvalds's avatar
Linus Torvalds committed
721 722
	} else {
		fn = (io_fn_t)file->f_op->write;
723
		fnv = file->f_op->aio_write;
724
		file_start_write(file);
Linus Torvalds's avatar
Linus Torvalds committed
725 726
	}

727 728 729 730 731
	if (fnv)
		ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
						pos, fnv);
	else
		ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
Linus Torvalds's avatar
Linus Torvalds committed
732

733 734 735
	if (type != READ)
		file_end_write(file);

Linus Torvalds's avatar
Linus Torvalds committed
736 737 738
out:
	if (iov != iovstack)
		kfree(iov);
Robert Love's avatar
Robert Love committed
739 740
	if ((ret + (type == READ)) > 0) {
		if (type == READ)
741
			fsnotify_access(file);
Robert Love's avatar
Robert Love committed
742
		else
743
			fsnotify_modify(file);
Robert Love's avatar
Robert Love committed
744
	}
Linus Torvalds's avatar
Linus Torvalds committed
745 746 747 748 749 750 751 752
	return ret;
}

ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
		  unsigned long vlen, loff_t *pos)
{
	if (!(file->f_mode & FMODE_READ))
		return -EBADF;
753
	if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
Linus Torvalds's avatar
Linus Torvalds committed
754 755 756 757 758 759 760 761 762 763 764 765
		return -EINVAL;

	return do_readv_writev(READ, file, vec, vlen, pos);
}

EXPORT_SYMBOL(vfs_readv);

ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
		   unsigned long vlen, loff_t *pos)
{
	if (!(file->f_mode & FMODE_WRITE))
		return -EBADF;
766
	if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
Linus Torvalds's avatar
Linus Torvalds committed
767 768 769 770 771 772 773
		return -EINVAL;

	return do_readv_writev(WRITE, file, vec, vlen, pos);
}

EXPORT_SYMBOL(vfs_writev);

774 775
SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
		unsigned long, vlen)
Linus Torvalds's avatar
Linus Torvalds committed
776
{
777
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
778 779
	ssize_t ret = -EBADF;

780 781 782 783 784
	if (f.file) {
		loff_t pos = file_pos_read(f.file);
		ret = vfs_readv(f.file, vec, vlen, &pos);
		file_pos_write(f.file, pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
785 786 787
	}

	if (ret > 0)
788 789
		add_rchar(current, ret);
	inc_syscr(current);
Linus Torvalds's avatar
Linus Torvalds committed
790 791 792
	return ret;
}

793 794
SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
		unsigned long, vlen)
Linus Torvalds's avatar
Linus Torvalds committed
795
{
796
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
797 798
	ssize_t ret = -EBADF;

799 800 801 802 803
	if (f.file) {
		loff_t pos = file_pos_read(f.file);
		ret = vfs_writev(f.file, vec, vlen, &pos);
		file_pos_write(f.file, pos);
		fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
804 805 806
	}

	if (ret > 0)
807 808
		add_wchar(current, ret);
	inc_syscw(current);
Linus Torvalds's avatar
Linus Torvalds committed
809 810 811
	return ret;
}

812 813 814 815 816 817
static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
{
#define HALF_LONG_BITS (BITS_PER_LONG / 2)
	return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
}

818
SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
819
		unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
820
{
821
	loff_t pos = pos_from_hilo(pos_h, pos_l);
822
	struct fd f;
823 824 825 826 827
	ssize_t ret = -EBADF;

	if (pos < 0)
		return -EINVAL;

828 829
	f = fdget(fd);
	if (f.file) {
830
		ret = -ESPIPE;
831 832 833
		if (f.file->f_mode & FMODE_PREAD)
			ret = vfs_readv(f.file, vec, vlen, &pos);
		fdput(f);
834 835 836 837 838 839 840 841 842
	}

	if (ret > 0)
		add_rchar(current, ret);
	inc_syscr(current);
	return ret;
}

SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
843
		unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
844
{
845
	loff_t pos = pos_from_hilo(pos_h, pos_l);
846
	struct fd f;
847 848 849 850 851
	ssize_t ret = -EBADF;

	if (pos < 0)
		return -EINVAL;

852 853
	f = fdget(fd);
	if (f.file) {
854
		ret = -ESPIPE;
855 856 857
		if (f.file->f_mode & FMODE_PWRITE)
			ret = vfs_writev(f.file, vec, vlen, &pos);
		fdput(f);
858 859 860 861 862 863 864 865
	}

	if (ret > 0)
		add_wchar(current, ret);
	inc_syscw(current);
	return ret;
}

866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903
#ifdef CONFIG_COMPAT

static ssize_t compat_do_readv_writev(int type, struct file *file,
			       const struct compat_iovec __user *uvector,
			       unsigned long nr_segs, loff_t *pos)
{
	compat_ssize_t tot_len;
	struct iovec iovstack[UIO_FASTIOV];
	struct iovec *iov = iovstack;
	ssize_t ret;
	io_fn_t fn;
	iov_fn_t fnv;

	ret = -EINVAL;
	if (!file->f_op)
		goto out;

	ret = -EFAULT;
	if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
		goto out;

	ret = compat_rw_copy_check_uvector(type, uvector, nr_segs,
					       UIO_FASTIOV, iovstack, &iov);
	if (ret <= 0)
		goto out;

	tot_len = ret;
	ret = rw_verify_area(type, file, pos, tot_len);
	if (ret < 0)
		goto out;

	fnv = NULL;
	if (type == READ) {
		fn = file->f_op->read;
		fnv = file->f_op->aio_read;
	} else {
		fn = (io_fn_t)file->f_op->write;
		fnv = file->f_op->aio_write;
904
		file_start_write(file);
905 906
	}

907
	if (fnv)
908 909
		ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
						pos, fnv);
910
	else
911 912
		ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);

913 914 915
	if (type != READ)
		file_end_write(file);

916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
out:
	if (iov != iovstack)
		kfree(iov);
	if ((ret + (type == READ)) > 0) {
		if (type == READ)
			fsnotify_access(file);
		else
			fsnotify_modify(file);
	}
	return ret;
}

static size_t compat_readv(struct file *file,
			   const struct compat_iovec __user *vec,
			   unsigned long vlen, loff_t *pos)
{
	ssize_t ret = -EBADF;

	if (!(file->f_mode & FMODE_READ))
		goto out;

	ret = -EINVAL;
	if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
		goto out;

	ret = compat_do_readv_writev(READ, file, vec, vlen, pos);

out:
	if (ret > 0)
		add_rchar(current, ret);
	inc_syscr(current);
	return ret;
}

950
COMPAT_SYSCALL_DEFINE3(readv, compat_ulong_t, fd,
951
		const struct compat_iovec __user *,vec,
952
		compat_ulong_t, vlen)
953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985
{
	struct fd f = fdget(fd);
	ssize_t ret;
	loff_t pos;

	if (!f.file)
		return -EBADF;
	pos = f.file->f_pos;
	ret = compat_readv(f.file, vec, vlen, &pos);
	f.file->f_pos = pos;
	fdput(f);
	return ret;
}

COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
		const struct compat_iovec __user *,vec,
		unsigned long, vlen, loff_t, pos)
{
	struct fd f;
	ssize_t ret;

	if (pos < 0)
		return -EINVAL;
	f = fdget(fd);
	if (!f.file)
		return -EBADF;
	ret = -ESPIPE;
	if (f.file->f_mode & FMODE_PREAD)
		ret = compat_readv(f.file, vec, vlen, &pos);
	fdput(f);
	return ret;
}

986
COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd,
987
		const struct compat_iovec __user *,vec,
988
		compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
{
	loff_t pos = ((loff_t)pos_high << 32) | pos_low;
	return compat_sys_preadv64(fd, vec, vlen, pos);
}

static size_t compat_writev(struct file *file,
			    const struct compat_iovec __user *vec,
			    unsigned long vlen, loff_t *pos)
{
	ssize_t ret = -EBADF;

	if (!(file->f_mode & FMODE_WRITE))
		goto out;

	ret = -EINVAL;
	if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
		goto out;

	ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);

out:
	if (ret > 0)
		add_wchar(current, ret);
	inc_syscw(current);
	return ret;
}

1016
COMPAT_SYSCALL_DEFINE3(writev, compat_ulong_t, fd,
1017
		const struct compat_iovec __user *, vec,
1018
		compat_ulong_t, vlen)
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
{
	struct fd f = fdget(fd);
	ssize_t ret;
	loff_t pos;

	if (!f.file)
		return -EBADF;
	pos = f.file->f_pos;
	ret = compat_writev(f.file, vec, vlen, &pos);
	f.file->f_pos = pos;
	fdput(f);
	return ret;
}

COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
		const struct compat_iovec __user *,vec,
		unsigned long, vlen, loff_t, pos)
{
	struct fd f;
	ssize_t ret;

	if (pos < 0)
		return -EINVAL;
	f = fdget(fd);
	if (!f.file)
		return -EBADF;
	ret = -ESPIPE;
	if (f.file->f_mode & FMODE_PWRITE)
		ret = compat_writev(f.file, vec, vlen, &pos);
	fdput(f);
	return ret;
}

1052
COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd,
1053
		const struct compat_iovec __user *,vec,
1054
		compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
1055 1056 1057 1058 1059 1060
{
	loff_t pos = ((loff_t)pos_high << 32) | pos_low;
	return compat_sys_pwritev64(fd, vec, vlen, pos);
}
#endif

1061 1062
static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
		  	   size_t count, loff_t max)
Linus Torvalds's avatar
Linus Torvalds committed
1063
{
1064 1065
	struct fd in, out;
	struct inode *in_inode, *out_inode;
Linus Torvalds's avatar
Linus Torvalds committed
1066
	loff_t pos;
1067
	loff_t out_pos;
Linus Torvalds's avatar
Linus Torvalds committed
1068
	ssize_t retval;
1069
	int fl;
Linus Torvalds's avatar
Linus Torvalds committed
1070 1071 1072 1073 1074

	/*
	 * Get input file, and verify that it is ok..
	 */
	retval = -EBADF;
1075 1076
	in = fdget(in_fd);
	if (!in.file)
Linus Torvalds's avatar
Linus Torvalds committed
1077
		goto out;
1078
	if (!(in.file->f_mode & FMODE_READ))
Linus Torvalds's avatar
Linus Torvalds committed
1079 1080
		goto fput_in;
	retval = -ESPIPE;
1081 1082 1083 1084
	if (!ppos) {
		pos = in.file->f_pos;
	} else {
		pos = *ppos;
1085
		if (!(in.file->f_mode & FMODE_PREAD))
Linus Torvalds's avatar
Linus Torvalds committed
1086
			goto fput_in;
1087 1088
	}
	retval = rw_verify_area(READ, in.file, &pos, count);
1089
	if (retval < 0)
Linus Torvalds's avatar
Linus Torvalds committed
1090
		goto fput_in;
1091
	count = retval;
Linus Torvalds's avatar
Linus Torvalds committed
1092 1093 1094 1095 1096

	/*
	 * Get output file, and verify that it is ok..
	 */
	retval = -EBADF;
1097 1098
	out = fdget(out_fd);
	if (!out.file)
Linus Torvalds's avatar
Linus Torvalds committed
1099
		goto fput_in;
1100
	if (!(out.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
1101 1102
		goto fput_out;
	retval = -EINVAL;
Al Viro's avatar
Al Viro committed
1103 1104
	in_inode = file_inode(in.file);
	out_inode = file_inode(out.file);
1105 1106
	out_pos = out.file->f_pos;
	retval = rw_verify_area(WRITE, out.file, &out_pos, count);
1107
	if (retval < 0)
Linus Torvalds's avatar
Linus Torvalds committed
1108
		goto fput_out;
1109
	count = retval;
Linus Torvalds's avatar
Linus Torvalds committed
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120

	if (!max)
		max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);

	if (unlikely(pos + count > max)) {
		retval = -EOVERFLOW;
		if (pos >= max)
			goto fput_out;
		count = max - pos;
	}

1121
	fl = 0;
1122
#if 0
1123 1124 1125 1126 1127 1128
	/*
	 * We need to debate whether we can enable this or not. The
	 * man page documents EAGAIN return for the output at least,
	 * and the application is arguably buggy if it doesn't expect
	 * EAGAIN on a non-blocking file descriptor.
	 */
1129
	if (in.file->f_flags & O_NONBLOCK)
1130
		fl = SPLICE_F_NONBLOCK;
1131
#endif
1132
	retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl);
Linus Torvalds's avatar
Linus Torvalds committed
1133 1134

	if (retval > 0) {
1135 1136
		add_rchar(current, retval);
		add_wchar(current, retval);
1137 1138
		fsnotify_access(in.file);
		fsnotify_modify(out.file);
1139 1140 1141 1142 1143
		out.file->f_pos = out_pos;
		if (ppos)
			*ppos = pos;
		else
			in.file->f_pos = pos;
Linus Torvalds's avatar
Linus Torvalds committed
1144 1145
	}

1146 1147
	inc_syscr(current);
	inc_syscw(current);
1148
	if (pos > max)
Linus Torvalds's avatar
Linus Torvalds committed
1149 1150 1151
		retval = -EOVERFLOW;

fput_out:
1152
	fdput(out);
Linus Torvalds's avatar
Linus Torvalds committed
1153
fput_in:
1154
	fdput(in);
Linus Torvalds's avatar
Linus Torvalds committed
1155 1156 1157 1158
out:
	return retval;
}

1159
SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
Linus Torvalds's avatar
Linus Torvalds committed
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
{
	loff_t pos;
	off_t off;
	ssize_t ret;

	if (offset) {
		if (unlikely(get_user(off, offset)))
			return -EFAULT;
		pos = off;
		ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
		if (unlikely(put_user(pos, offset)))
			return -EFAULT;
		return ret;
	}

	return do_sendfile(out_fd, in_fd, NULL, count, 0);
}

1178
SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
Linus Torvalds's avatar
Linus Torvalds committed
1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
{
	loff_t pos;
	ssize_t ret;

	if (offset) {
		if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
			return -EFAULT;
		ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
		if (unlikely(put_user(pos, offset)))
			return -EFAULT;
		return ret;
	}

	return do_sendfile(out_fd, in_fd, NULL, count, 0);
}
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233

#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
		compat_off_t __user *, offset, compat_size_t, count)
{
	loff_t pos;
	off_t off;
	ssize_t ret;

	if (offset) {
		if (unlikely(get_user(off, offset)))
			return -EFAULT;
		pos = off;
		ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
		if (unlikely(put_user(pos, offset)))
			return -EFAULT;
		return ret;
	}

	return do_sendfile(out_fd, in_fd, NULL, count, 0);
}

COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
		compat_loff_t __user *, offset, compat_size_t, count)
{
	loff_t pos;
	ssize_t ret;

	if (offset) {
		if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
			return -EFAULT;
		ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
		if (unlikely(put_user(pos, offset)))
			return -EFAULT;
		return ret;
	}

	return do_sendfile(out_fd, in_fd, NULL, count, 0);
}
#endif