blk-mq: don't count completed flush data request as inflight in case of quiesce
[linux-block.git] / fs / read_write.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4
LT
2/*
3 * linux/fs/read_write.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */
7
b12fb7f4 8#include <linux/slab.h>
1da177e4 9#include <linux/stat.h>
b12fb7f4 10#include <linux/sched/xacct.h>
1da177e4
LT
11#include <linux/fcntl.h>
12#include <linux/file.h>
13#include <linux/uio.h>
0eeca283 14#include <linux/fsnotify.h>
1da177e4 15#include <linux/security.h>
630d9c47 16#include <linux/export.h>
1da177e4 17#include <linux/syscalls.h>
e28cc715 18#include <linux/pagemap.h>
d6b29d7c 19#include <linux/splice.h>
561c6731 20#include <linux/compat.h>
29732938 21#include <linux/mount.h>
2feb55f8 22#include <linux/fs.h>
06ae43f3 23#include "internal.h"
1da177e4 24
7c0f6ba6 25#include <linux/uaccess.h>
1da177e4
LT
26#include <asm/unistd.h>
27
4b6f5d20 28const struct file_operations generic_ro_fops = {
1da177e4 29 .llseek = generic_file_llseek,
aad4f8bb 30 .read_iter = generic_file_read_iter,
1da177e4 31 .mmap = generic_file_readonly_mmap,
2cb1e089 32 .splice_read = filemap_splice_read,
1da177e4
LT
33};
34
35EXPORT_SYMBOL(generic_ro_fops);
36
ddef7ed2 37static inline bool unsigned_offsets(struct file *file)
4a3956c7 38{
cccb5a1e 39 return file->f_mode & FMODE_UNSIGNED_OFFSET;
4a3956c7
KH
40}
41
46a1c2c7
JL
42/**
43 * vfs_setpos - update the file offset for lseek
44 * @file: file structure in question
45 * @offset: file offset to seek to
46 * @maxsize: maximum file size
47 *
48 * This is a low-level filesystem helper for updating the file offset to
49 * the value specified by @offset if the given offset is valid and it is
50 * not equal to the current file offset.
51 *
52 * Return the specified offset on success and -EINVAL on invalid offset.
53 */
54loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
ef3d0fd2
AK
55{
56 if (offset < 0 && !unsigned_offsets(file))
57 return -EINVAL;
58 if (offset > maxsize)
59 return -EINVAL;
60
61 if (offset != file->f_pos) {
62 file->f_pos = offset;
63 file->f_version = 0;
64 }
65 return offset;
66}
46a1c2c7 67EXPORT_SYMBOL(vfs_setpos);
ef3d0fd2 68
3a8cff4f 69/**
5760495a 70 * generic_file_llseek_size - generic llseek implementation for regular files
3a8cff4f
CH
71 * @file: file structure to seek on
72 * @offset: file offset to seek to
965c8e59 73 * @whence: type of seek
89cbd4c0 74 * @maxsize: max size of this file in file system
e8b96eb5 75 * @eof: offset used for SEEK_END position
3a8cff4f 76 *
5760495a 77 * This is a variant of generic_file_llseek that allows passing in a custom
e8b96eb5 78 * maximum file size and a custom EOF position, for e.g. hashed directories
ef3d0fd2
AK
79 *
80 * Synchronization:
5760495a 81 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
ef3d0fd2
AK
82 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
83 * read/writes behave like SEEK_SET against seeks.
3a8cff4f 84 */
9465efc9 85loff_t
965c8e59 86generic_file_llseek_size(struct file *file, loff_t offset, int whence,
e8b96eb5 87 loff_t maxsize, loff_t eof)
1da177e4 88{
965c8e59 89 switch (whence) {
3a8cff4f 90 case SEEK_END:
e8b96eb5 91 offset += eof;
3a8cff4f
CH
92 break;
93 case SEEK_CUR:
5b6f1eb9
AK
94 /*
95 * Here we special-case the lseek(fd, 0, SEEK_CUR)
96 * position-querying operation. Avoid rewriting the "same"
97 * f_pos value back to the file because a concurrent read(),
98 * write() or lseek() might have altered it
99 */
100 if (offset == 0)
101 return file->f_pos;
ef3d0fd2
AK
102 /*
103 * f_lock protects against read/modify/write race with other
104 * SEEK_CURs. Note that parallel writes and reads behave
105 * like SEEK_SET.
106 */
107 spin_lock(&file->f_lock);
46a1c2c7 108 offset = vfs_setpos(file, file->f_pos + offset, maxsize);
ef3d0fd2
AK
109 spin_unlock(&file->f_lock);
110 return offset;
982d8165
JB
111 case SEEK_DATA:
112 /*
113 * In the generic case the entire file is data, so as long as
114 * offset isn't at the end of the file then the offset is data.
115 */
fc46820b 116 if ((unsigned long long)offset >= eof)
982d8165
JB
117 return -ENXIO;
118 break;
119 case SEEK_HOLE:
120 /*
121 * There is a virtual hole at the end of the file, so as long as
122 * offset isn't i_size or larger, return i_size.
123 */
fc46820b 124 if ((unsigned long long)offset >= eof)
982d8165 125 return -ENXIO;
e8b96eb5 126 offset = eof;
982d8165 127 break;
1da177e4 128 }
3a8cff4f 129
46a1c2c7 130 return vfs_setpos(file, offset, maxsize);
5760495a
AK
131}
132EXPORT_SYMBOL(generic_file_llseek_size);
133
134/**
135 * generic_file_llseek - generic llseek implementation for regular files
136 * @file: file structure to seek on
137 * @offset: file offset to seek to
965c8e59 138 * @whence: type of seek
5760495a
AK
139 *
140 * This is a generic implemenation of ->llseek useable for all normal local
141 * filesystems. It just updates the file offset to the value specified by
546ae2d2 142 * @offset and @whence.
5760495a 143 */
965c8e59 144loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
5760495a
AK
145{
146 struct inode *inode = file->f_mapping->host;
147
965c8e59 148 return generic_file_llseek_size(file, offset, whence,
e8b96eb5
ES
149 inode->i_sb->s_maxbytes,
150 i_size_read(inode));
1da177e4 151}
9465efc9 152EXPORT_SYMBOL(generic_file_llseek);
1da177e4 153
1bf9d14d
AV
154/**
155 * fixed_size_llseek - llseek implementation for fixed-sized devices
156 * @file: file structure to seek on
157 * @offset: file offset to seek to
158 * @whence: type of seek
159 * @size: size of the file
160 *
161 */
162loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size)
163{
164 switch (whence) {
165 case SEEK_SET: case SEEK_CUR: case SEEK_END:
166 return generic_file_llseek_size(file, offset, whence,
167 size, size);
168 default:
169 return -EINVAL;
170 }
171}
172EXPORT_SYMBOL(fixed_size_llseek);
173
b25472f9
AV
174/**
175 * no_seek_end_llseek - llseek implementation for fixed-sized devices
176 * @file: file structure to seek on
177 * @offset: file offset to seek to
178 * @whence: type of seek
179 *
180 */
181loff_t no_seek_end_llseek(struct file *file, loff_t offset, int whence)
182{
183 switch (whence) {
184 case SEEK_SET: case SEEK_CUR:
185 return generic_file_llseek_size(file, offset, whence,
2feb55f8 186 OFFSET_MAX, 0);
b25472f9
AV
187 default:
188 return -EINVAL;
189 }
190}
191EXPORT_SYMBOL(no_seek_end_llseek);
192
193/**
194 * no_seek_end_llseek_size - llseek implementation for fixed-sized devices
195 * @file: file structure to seek on
196 * @offset: file offset to seek to
197 * @whence: type of seek
198 * @size: maximal offset allowed
199 *
200 */
201loff_t no_seek_end_llseek_size(struct file *file, loff_t offset, int whence, loff_t size)
202{
203 switch (whence) {
204 case SEEK_SET: case SEEK_CUR:
205 return generic_file_llseek_size(file, offset, whence,
206 size, 0);
207 default:
208 return -EINVAL;
209 }
210}
211EXPORT_SYMBOL(no_seek_end_llseek_size);
212
ae6afc3f
B
213/**
214 * noop_llseek - No Operation Performed llseek implementation
215 * @file: file structure to seek on
216 * @offset: file offset to seek to
965c8e59 217 * @whence: type of seek
ae6afc3f
B
218 *
219 * This is an implementation of ->llseek useable for the rare special case when
220 * userspace expects the seek to succeed but the (device) file is actually not
221 * able to perform the seek. In this case you use noop_llseek() instead of
222 * falling back to the default implementation of ->llseek.
223 */
965c8e59 224loff_t noop_llseek(struct file *file, loff_t offset, int whence)
ae6afc3f
B
225{
226 return file->f_pos;
227}
228EXPORT_SYMBOL(noop_llseek);
229
965c8e59 230loff_t default_llseek(struct file *file, loff_t offset, int whence)
1da177e4 231{
496ad9aa 232 struct inode *inode = file_inode(file);
16abef0e 233 loff_t retval;
1da177e4 234
5955102c 235 inode_lock(inode);
965c8e59 236 switch (whence) {
7b8e8924 237 case SEEK_END:
982d8165 238 offset += i_size_read(inode);
1da177e4 239 break;
7b8e8924 240 case SEEK_CUR:
5b6f1eb9
AK
241 if (offset == 0) {
242 retval = file->f_pos;
243 goto out;
244 }
1da177e4 245 offset += file->f_pos;
982d8165
JB
246 break;
247 case SEEK_DATA:
248 /*
249 * In the generic case the entire file is data, so as
250 * long as offset isn't at the end of the file then the
251 * offset is data.
252 */
bacb2d81
DC
253 if (offset >= inode->i_size) {
254 retval = -ENXIO;
255 goto out;
256 }
982d8165
JB
257 break;
258 case SEEK_HOLE:
259 /*
260 * There is a virtual hole at the end of the file, so
261 * as long as offset isn't i_size or larger, return
262 * i_size.
263 */
bacb2d81
DC
264 if (offset >= inode->i_size) {
265 retval = -ENXIO;
266 goto out;
267 }
982d8165
JB
268 offset = inode->i_size;
269 break;
1da177e4
LT
270 }
271 retval = -EINVAL;
cccb5a1e 272 if (offset >= 0 || unsigned_offsets(file)) {
1da177e4
LT
273 if (offset != file->f_pos) {
274 file->f_pos = offset;
275 file->f_version = 0;
276 }
277 retval = offset;
278 }
5b6f1eb9 279out:
5955102c 280 inode_unlock(inode);
1da177e4
LT
281 return retval;
282}
283EXPORT_SYMBOL(default_llseek);
284
965c8e59 285loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
1da177e4 286{
4e3299ea
JD
287 if (!(file->f_mode & FMODE_LSEEK))
288 return -ESPIPE;
289 return file->f_op->llseek(file, offset, whence);
1da177e4
LT
290}
291EXPORT_SYMBOL(vfs_llseek);
292
bef17329 293static off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence)
1da177e4
LT
294{
295 off_t retval;
9c225f26 296 struct fd f = fdget_pos(fd);
2903ff01
AV
297 if (!f.file)
298 return -EBADF;
1da177e4
LT
299
300 retval = -EINVAL;
965c8e59
AM
301 if (whence <= SEEK_MAX) {
302 loff_t res = vfs_llseek(f.file, offset, whence);
1da177e4
LT
303 retval = res;
304 if (res != (loff_t)retval)
305 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
306 }
9c225f26 307 fdput_pos(f);
1da177e4
LT
308 return retval;
309}
310
76847e43
DB
311SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
312{
313 return ksys_lseek(fd, offset, whence);
314}
315
561c6731
AV
316#ifdef CONFIG_COMPAT
317COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
318{
76847e43 319 return ksys_lseek(fd, offset, whence);
561c6731
AV
320}
321#endif
322
9e62ccec
MS
323#if !defined(CONFIG_64BIT) || defined(CONFIG_COMPAT) || \
324 defined(__ARCH_WANT_SYS_LLSEEK)
003d7ab4
HC
325SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
326 unsigned long, offset_low, loff_t __user *, result,
965c8e59 327 unsigned int, whence)
1da177e4
LT
328{
329 int retval;
d7a15f8d 330 struct fd f = fdget_pos(fd);
1da177e4 331 loff_t offset;
1da177e4 332
2903ff01
AV
333 if (!f.file)
334 return -EBADF;
1da177e4
LT
335
336 retval = -EINVAL;
965c8e59 337 if (whence > SEEK_MAX)
1da177e4
LT
338 goto out_putf;
339
2903ff01 340 offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
965c8e59 341 whence);
1da177e4
LT
342
343 retval = (int)offset;
344 if (offset >= 0) {
345 retval = -EFAULT;
346 if (!copy_to_user(result, &offset, sizeof(offset)))
347 retval = 0;
348 }
349out_putf:
d7a15f8d 350 fdput_pos(f);
1da177e4
LT
351 return retval;
352}
353#endif
354
68d70d03 355int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count)
1da177e4 356{
e28cc715 357 if (unlikely((ssize_t) count < 0))
2949e842 358 return -EINVAL;
1da177e4 359
438ab720
KS
360 if (ppos) {
361 loff_t pos = *ppos;
362
363 if (unlikely(pos < 0)) {
364 if (!unsigned_offsets(file))
2949e842 365 return -EINVAL;
438ab720
KS
366 if (count >= -pos) /* both values are in 0..LLONG_MAX */
367 return -EOVERFLOW;
368 } else if (unlikely((loff_t) (pos + count) < 0)) {
369 if (!unsigned_offsets(file))
2949e842 370 return -EINVAL;
438ab720 371 }
e28cc715 372 }
438ab720 373
bc61384d 374 return security_file_permission(file,
c43e259c 375 read_write == READ ? MAY_READ : MAY_WRITE);
1da177e4 376}
87112933 377EXPORT_SYMBOL(rw_verify_area);
1da177e4 378
5d5d5689 379static ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
293bc982 380{
293bc982
AV
381 struct kiocb kiocb;
382 struct iov_iter iter;
383 ssize_t ret;
384
385 init_sync_kiocb(&kiocb, filp);
438ab720 386 kiocb.ki_pos = (ppos ? *ppos : 0);
de4eda9d 387 iov_iter_ubuf(&iter, ITER_DEST, buf, len);
293bc982 388
bb7462b6 389 ret = call_read_iter(filp, &kiocb, &iter);
599bd19b 390 BUG_ON(ret == -EIOCBQUEUED);
438ab720
KS
391 if (ppos)
392 *ppos = kiocb.ki_pos;
293bc982
AV
393 return ret;
394}
395
4d03e3cc
CH
396static int warn_unsupported(struct file *file, const char *op)
397{
398 pr_warn_ratelimited(
399 "kernel %s not supported for file %pD4 (pid: %d comm: %.20s)\n",
400 op, file, current->pid, current->comm);
401 return -EINVAL;
402}
403
61a707c5
CH
404ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
405{
4d03e3cc
CH
406 struct kvec iov = {
407 .iov_base = buf,
408 .iov_len = min_t(size_t, count, MAX_RW_COUNT),
409 };
410 struct kiocb kiocb;
411 struct iov_iter iter;
61a707c5
CH
412 ssize_t ret;
413
414 if (WARN_ON_ONCE(!(file->f_mode & FMODE_READ)))
415 return -EINVAL;
416 if (!(file->f_mode & FMODE_CAN_READ))
417 return -EINVAL;
4d03e3cc
CH
418 /*
419 * Also fail if ->read_iter and ->read are both wired up as that
420 * implies very convoluted semantics.
421 */
422 if (unlikely(!file->f_op->read_iter || file->f_op->read))
423 return warn_unsupported(file, "read");
61a707c5 424
4d03e3cc 425 init_sync_kiocb(&kiocb, file);
7b84b665 426 kiocb.ki_pos = pos ? *pos : 0;
de4eda9d 427 iov_iter_kvec(&iter, ITER_DEST, &iov, 1, iov.iov_len);
4d03e3cc 428 ret = file->f_op->read_iter(&kiocb, &iter);
61a707c5 429 if (ret > 0) {
7b84b665
MWO
430 if (pos)
431 *pos = kiocb.ki_pos;
61a707c5
CH
432 fsnotify_access(file);
433 add_rchar(current, ret);
434 }
435 inc_syscr(current);
436 return ret;
437}
438
bdd1d2d3 439ssize_t kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
c41fbad0 440{
6209dd91 441 ssize_t ret;
c41fbad0 442
6209dd91
CH
443 ret = rw_verify_area(READ, file, pos, count);
444 if (ret)
445 return ret;
446 return __kernel_read(file, buf, count, pos);
c41fbad0
CH
447}
448EXPORT_SYMBOL(kernel_read);
6fb5032e 449
1da177e4
LT
450ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
451{
452 ssize_t ret;
453
454 if (!(file->f_mode & FMODE_READ))
455 return -EBADF;
7f7f25e8 456 if (!(file->f_mode & FMODE_CAN_READ))
1da177e4 457 return -EINVAL;
96d4f267 458 if (unlikely(!access_ok(buf, count)))
1da177e4
LT
459 return -EFAULT;
460
461 ret = rw_verify_area(READ, file, pos, count);
775802c0
CH
462 if (ret)
463 return ret;
464 if (count > MAX_RW_COUNT)
465 count = MAX_RW_COUNT;
1da177e4 466
775802c0
CH
467 if (file->f_op->read)
468 ret = file->f_op->read(file, buf, count, pos);
469 else if (file->f_op->read_iter)
470 ret = new_sync_read(file, buf, count, pos);
471 else
472 ret = -EINVAL;
473 if (ret > 0) {
474 fsnotify_access(file);
475 add_rchar(current, ret);
476 }
477 inc_syscr(current);
1da177e4
LT
478 return ret;
479}
480
5d5d5689 481static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
293bc982 482{
293bc982
AV
483 struct kiocb kiocb;
484 struct iov_iter iter;
485 ssize_t ret;
486
487 init_sync_kiocb(&kiocb, filp);
438ab720 488 kiocb.ki_pos = (ppos ? *ppos : 0);
de4eda9d 489 iov_iter_ubuf(&iter, ITER_SOURCE, (void __user *)buf, len);
293bc982 490
bb7462b6 491 ret = call_write_iter(filp, &kiocb, &iter);
599bd19b 492 BUG_ON(ret == -EIOCBQUEUED);
438ab720 493 if (ret > 0 && ppos)
f765b134 494 *ppos = kiocb.ki_pos;
293bc982
AV
495 return ret;
496}
497
81238b2c 498/* caller is responsible for file_start_write/file_end_write */
06bbaa6d 499ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *pos)
06ae43f3 500{
4d03e3cc 501 struct kiocb kiocb;
06ae43f3
AV
502 ssize_t ret;
503
a01ac27b
CH
504 if (WARN_ON_ONCE(!(file->f_mode & FMODE_WRITE)))
505 return -EBADF;
7f7f25e8 506 if (!(file->f_mode & FMODE_CAN_WRITE))
3e84f48e 507 return -EINVAL;
4d03e3cc
CH
508 /*
509 * Also fail if ->write_iter and ->write are both wired up as that
510 * implies very convoluted semantics.
511 */
512 if (unlikely(!file->f_op->write_iter || file->f_op->write))
513 return warn_unsupported(file, "write");
3e84f48e 514
4d03e3cc 515 init_sync_kiocb(&kiocb, file);
4c207ef4 516 kiocb.ki_pos = pos ? *pos : 0;
06bbaa6d 517 ret = file->f_op->write_iter(&kiocb, from);
06ae43f3 518 if (ret > 0) {
4c207ef4
MWO
519 if (pos)
520 *pos = kiocb.ki_pos;
06ae43f3
AV
521 fsnotify_modify(file);
522 add_wchar(current, ret);
523 }
524 inc_syscw(current);
525 return ret;
526}
06bbaa6d
AV
527
528/* caller is responsible for file_start_write/file_end_write */
529ssize_t __kernel_write(struct file *file, const void *buf, size_t count, loff_t *pos)
530{
531 struct kvec iov = {
532 .iov_base = (void *)buf,
533 .iov_len = min_t(size_t, count, MAX_RW_COUNT),
534 };
535 struct iov_iter iter;
de4eda9d 536 iov_iter_kvec(&iter, ITER_SOURCE, &iov, 1, iov.iov_len);
06bbaa6d
AV
537 return __kernel_write_iter(file, &iter, pos);
538}
90fb7027
LT
539/*
540 * This "EXPORT_SYMBOL_GPL()" is more of a "EXPORT_SYMBOL_DONTUSE()",
541 * but autofs is one of the few internal kernel users that actually
542 * wants this _and_ can be built as a module. So we need to export
543 * this symbol for autofs, even though it really isn't appropriate
544 * for any other kernel modules.
545 */
546EXPORT_SYMBOL_GPL(__kernel_write);
2ec3a12a 547
e13ec939
CH
548ssize_t kernel_write(struct file *file, const void *buf, size_t count,
549 loff_t *pos)
ac452aca 550{
81238b2c 551 ssize_t ret;
ac452aca 552
81238b2c
CH
553 ret = rw_verify_area(WRITE, file, pos, count);
554 if (ret)
555 return ret;
ac452aca 556
81238b2c
CH
557 file_start_write(file);
558 ret = __kernel_write(file, buf, count, pos);
559 file_end_write(file);
560 return ret;
ac452aca
CH
561}
562EXPORT_SYMBOL(kernel_write);
563
1da177e4
LT
564ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
565{
566 ssize_t ret;
567
568 if (!(file->f_mode & FMODE_WRITE))
569 return -EBADF;
7f7f25e8 570 if (!(file->f_mode & FMODE_CAN_WRITE))
1da177e4 571 return -EINVAL;
96d4f267 572 if (unlikely(!access_ok(buf, count)))
1da177e4
LT
573 return -EFAULT;
574
575 ret = rw_verify_area(WRITE, file, pos, count);
53ad8626
CH
576 if (ret)
577 return ret;
578 if (count > MAX_RW_COUNT)
579 count = MAX_RW_COUNT;
580 file_start_write(file);
581 if (file->f_op->write)
582 ret = file->f_op->write(file, buf, count, pos);
583 else if (file->f_op->write_iter)
584 ret = new_sync_write(file, buf, count, pos);
585 else
586 ret = -EINVAL;
587 if (ret > 0) {
588 fsnotify_modify(file);
589 add_wchar(current, ret);
1da177e4 590 }
53ad8626
CH
591 inc_syscw(current);
592 file_end_write(file);
1da177e4
LT
593 return ret;
594}
595
438ab720
KS
596/* file_ppos returns &file->f_pos or NULL if file is stream */
597static inline loff_t *file_ppos(struct file *file)
1da177e4 598{
438ab720 599 return file->f_mode & FMODE_STREAM ? NULL : &file->f_pos;
1da177e4
LT
600}
601
3ce4a7bf 602ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count)
1da177e4 603{
9c225f26 604 struct fd f = fdget_pos(fd);
1da177e4 605 ssize_t ret = -EBADF;
1da177e4 606
2903ff01 607 if (f.file) {
438ab720
KS
608 loff_t pos, *ppos = file_ppos(f.file);
609 if (ppos) {
610 pos = *ppos;
611 ppos = &pos;
612 }
613 ret = vfs_read(f.file, buf, count, ppos);
614 if (ret >= 0 && ppos)
615 f.file->f_pos = pos;
9c225f26 616 fdput_pos(f);
1da177e4 617 }
1da177e4
LT
618 return ret;
619}
1da177e4 620
3ce4a7bf
DB
621SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
622{
623 return ksys_read(fd, buf, count);
624}
625
e7a3e8b2 626ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count)
1da177e4 627{
9c225f26 628 struct fd f = fdget_pos(fd);
1da177e4 629 ssize_t ret = -EBADF;
1da177e4 630
2903ff01 631 if (f.file) {
438ab720
KS
632 loff_t pos, *ppos = file_ppos(f.file);
633 if (ppos) {
634 pos = *ppos;
635 ppos = &pos;
636 }
637 ret = vfs_write(f.file, buf, count, ppos);
638 if (ret >= 0 && ppos)
639 f.file->f_pos = pos;
9c225f26 640 fdput_pos(f);
1da177e4
LT
641 }
642
643 return ret;
644}
645
e7a3e8b2
DB
646SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
647 size_t, count)
648{
649 return ksys_write(fd, buf, count);
650}
651
36028d5d
DB
652ssize_t ksys_pread64(unsigned int fd, char __user *buf, size_t count,
653 loff_t pos)
1da177e4 654{
2903ff01 655 struct fd f;
1da177e4 656 ssize_t ret = -EBADF;
1da177e4
LT
657
658 if (pos < 0)
659 return -EINVAL;
660
2903ff01
AV
661 f = fdget(fd);
662 if (f.file) {
1da177e4 663 ret = -ESPIPE;
2903ff01
AV
664 if (f.file->f_mode & FMODE_PREAD)
665 ret = vfs_read(f.file, buf, count, &pos);
666 fdput(f);
1da177e4
LT
667 }
668
669 return ret;
670}
671
36028d5d
DB
672SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
673 size_t, count, loff_t, pos)
674{
675 return ksys_pread64(fd, buf, count, pos);
676}
677
59c10c52
GR
678#if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PREAD64)
679COMPAT_SYSCALL_DEFINE5(pread64, unsigned int, fd, char __user *, buf,
680 size_t, count, compat_arg_u64_dual(pos))
681{
682 return ksys_pread64(fd, buf, count, compat_arg_u64_glue(pos));
683}
684#endif
685
36028d5d
DB
686ssize_t ksys_pwrite64(unsigned int fd, const char __user *buf,
687 size_t count, loff_t pos)
1da177e4 688{
2903ff01 689 struct fd f;
1da177e4 690 ssize_t ret = -EBADF;
1da177e4
LT
691
692 if (pos < 0)
693 return -EINVAL;
694
2903ff01
AV
695 f = fdget(fd);
696 if (f.file) {
1da177e4 697 ret = -ESPIPE;
2903ff01
AV
698 if (f.file->f_mode & FMODE_PWRITE)
699 ret = vfs_write(f.file, buf, count, &pos);
700 fdput(f);
1da177e4
LT
701 }
702
703 return ret;
704}
705
36028d5d
DB
706SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
707 size_t, count, loff_t, pos)
708{
709 return ksys_pwrite64(fd, buf, count, pos);
710}
711
59c10c52
GR
712#if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PWRITE64)
713COMPAT_SYSCALL_DEFINE5(pwrite64, unsigned int, fd, const char __user *, buf,
714 size_t, count, compat_arg_u64_dual(pos))
715{
716 return ksys_pwrite64(fd, buf, count, compat_arg_u64_glue(pos));
717}
718#endif
719
ac15ac06 720static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter,
ddef7ed2 721 loff_t *ppos, int type, rwf_t flags)
293bc982
AV
722{
723 struct kiocb kiocb;
293bc982
AV
724 ssize_t ret;
725
726 init_sync_kiocb(&kiocb, filp);
fdd2f5b7
GR
727 ret = kiocb_set_rw_flags(&kiocb, flags);
728 if (ret)
729 return ret;
438ab720 730 kiocb.ki_pos = (ppos ? *ppos : 0);
293bc982 731
0f78d06a 732 if (type == READ)
bb7462b6 733 ret = call_read_iter(filp, &kiocb, iter);
0f78d06a 734 else
bb7462b6 735 ret = call_write_iter(filp, &kiocb, iter);
599bd19b 736 BUG_ON(ret == -EIOCBQUEUED);
438ab720
KS
737 if (ppos)
738 *ppos = kiocb.ki_pos;
293bc982
AV
739 return ret;
740}
741
ee0b3e67 742/* Do it by hand, with file-ops */
ac15ac06 743static ssize_t do_loop_readv_writev(struct file *filp, struct iov_iter *iter,
ddef7ed2 744 loff_t *ppos, int type, rwf_t flags)
ee0b3e67 745{
ee0b3e67
BP
746 ssize_t ret = 0;
747
97be7ebe 748 if (flags & ~RWF_HIPRI)
793b80ef
CH
749 return -EOPNOTSUPP;
750
ac15ac06 751 while (iov_iter_count(iter)) {
ee0b3e67
BP
752 ssize_t nr;
753
0f78d06a 754 if (type == READ) {
95e49cf8
JA
755 nr = filp->f_op->read(filp, iter_iov_addr(iter),
756 iter_iov_len(iter), ppos);
0f78d06a 757 } else {
95e49cf8
JA
758 nr = filp->f_op->write(filp, iter_iov_addr(iter),
759 iter_iov_len(iter), ppos);
0f78d06a 760 }
ee0b3e67
BP
761
762 if (nr < 0) {
763 if (!ret)
764 ret = nr;
765 break;
766 }
767 ret += nr;
95e49cf8 768 if (nr != iter_iov_len(iter))
ee0b3e67 769 break;
ac15ac06 770 iov_iter_advance(iter, nr);
ee0b3e67
BP
771 }
772
773 return ret;
774}
775
19c73586 776static ssize_t do_iter_read(struct file *file, struct iov_iter *iter,
ddef7ed2 777 loff_t *pos, rwf_t flags)
1da177e4 778{
1da177e4 779 size_t tot_len;
7687a7a4 780 ssize_t ret = 0;
1da177e4 781
edab5fe3
CH
782 if (!(file->f_mode & FMODE_READ))
783 return -EBADF;
784 if (!(file->f_mode & FMODE_CAN_READ))
785 return -EINVAL;
786
7687a7a4 787 tot_len = iov_iter_count(iter);
0504c074
AV
788 if (!tot_len)
789 goto out;
19c73586 790 ret = rw_verify_area(READ, file, pos, tot_len);
e28cc715 791 if (ret < 0)
19c73586 792 return ret;
1da177e4 793
19c73586
CH
794 if (file->f_op->read_iter)
795 ret = do_iter_readv_writev(file, iter, pos, READ, flags);
ee0b3e67 796 else
19c73586 797 ret = do_loop_readv_writev(file, iter, pos, READ, flags);
1da177e4 798out:
19c73586
CH
799 if (ret >= 0)
800 fsnotify_access(file);
1da177e4 801 return ret;
1da177e4
LT
802}
803
5dcdc43e
JX
804ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
805 struct iov_iter *iter)
806{
807 size_t tot_len;
808 ssize_t ret = 0;
809
810 if (!file->f_op->read_iter)
811 return -EINVAL;
812 if (!(file->f_mode & FMODE_READ))
813 return -EBADF;
814 if (!(file->f_mode & FMODE_CAN_READ))
815 return -EINVAL;
816
817 tot_len = iov_iter_count(iter);
818 if (!tot_len)
819 goto out;
820 ret = rw_verify_area(READ, file, &iocb->ki_pos, tot_len);
821 if (ret < 0)
822 return ret;
823
824 ret = call_read_iter(file, iocb, iter);
825out:
826 if (ret >= 0)
827 fsnotify_access(file);
828 return ret;
829}
830EXPORT_SYMBOL(vfs_iocb_iter_read);
831
18e9710e 832ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
ddef7ed2 833 rwf_t flags)
7687a7a4 834{
18e9710e
CH
835 if (!file->f_op->read_iter)
836 return -EINVAL;
837 return do_iter_read(file, iter, ppos, flags);
838}
839EXPORT_SYMBOL(vfs_iter_read);
7687a7a4 840
19c73586 841static ssize_t do_iter_write(struct file *file, struct iov_iter *iter,
ddef7ed2 842 loff_t *pos, rwf_t flags)
19c73586
CH
843{
844 size_t tot_len;
845 ssize_t ret = 0;
03d95eb2 846
edab5fe3
CH
847 if (!(file->f_mode & FMODE_WRITE))
848 return -EBADF;
849 if (!(file->f_mode & FMODE_CAN_WRITE))
850 return -EINVAL;
851
19c73586
CH
852 tot_len = iov_iter_count(iter);
853 if (!tot_len)
854 return 0;
855 ret = rw_verify_area(WRITE, file, pos, tot_len);
7687a7a4
MS
856 if (ret < 0)
857 return ret;
858
19c73586
CH
859 if (file->f_op->write_iter)
860 ret = do_iter_readv_writev(file, iter, pos, WRITE, flags);
861 else
862 ret = do_loop_readv_writev(file, iter, pos, WRITE, flags);
19c73586
CH
863 if (ret > 0)
864 fsnotify_modify(file);
7687a7a4
MS
865 return ret;
866}
867
5dcdc43e
JX
868ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
869 struct iov_iter *iter)
870{
871 size_t tot_len;
872 ssize_t ret = 0;
873
874 if (!file->f_op->write_iter)
875 return -EINVAL;
876 if (!(file->f_mode & FMODE_WRITE))
877 return -EBADF;
878 if (!(file->f_mode & FMODE_CAN_WRITE))
879 return -EINVAL;
880
881 tot_len = iov_iter_count(iter);
882 if (!tot_len)
883 return 0;
884 ret = rw_verify_area(WRITE, file, &iocb->ki_pos, tot_len);
885 if (ret < 0)
886 return ret;
887
888 ret = call_write_iter(file, iocb, iter);
889 if (ret > 0)
890 fsnotify_modify(file);
891
892 return ret;
893}
894EXPORT_SYMBOL(vfs_iocb_iter_write);
895
abbb6589 896ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
ddef7ed2 897 rwf_t flags)
abbb6589
CH
898{
899 if (!file->f_op->write_iter)
900 return -EINVAL;
901 return do_iter_write(file, iter, ppos, flags);
902}
903EXPORT_SYMBOL(vfs_iter_write);
904
36e2c742 905static ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
ddef7ed2 906 unsigned long vlen, loff_t *pos, rwf_t flags)
1da177e4 907{
7687a7a4
MS
908 struct iovec iovstack[UIO_FASTIOV];
909 struct iovec *iov = iovstack;
910 struct iov_iter iter;
911 ssize_t ret;
1da177e4 912
de4eda9d 913 ret = import_iovec(ITER_DEST, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter);
edab5fe3
CH
914 if (ret >= 0) {
915 ret = do_iter_read(file, &iter, pos, flags);
916 kfree(iov);
917 }
1da177e4 918
251b42a1
CH
919 return ret;
920}
1da177e4 921
9725d4ce 922static ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
ddef7ed2 923 unsigned long vlen, loff_t *pos, rwf_t flags)
1da177e4 924{
251b42a1
CH
925 struct iovec iovstack[UIO_FASTIOV];
926 struct iovec *iov = iovstack;
927 struct iov_iter iter;
928 ssize_t ret;
1da177e4 929
de4eda9d 930 ret = import_iovec(ITER_SOURCE, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter);
edab5fe3 931 if (ret >= 0) {
62473a2d 932 file_start_write(file);
edab5fe3 933 ret = do_iter_write(file, &iter, pos, flags);
62473a2d 934 file_end_write(file);
edab5fe3
CH
935 kfree(iov);
936 }
251b42a1 937 return ret;
1da177e4 938}
1da177e4 939
f17d8b35 940static ssize_t do_readv(unsigned long fd, const struct iovec __user *vec,
ddef7ed2 941 unsigned long vlen, rwf_t flags)
1da177e4 942{
9c225f26 943 struct fd f = fdget_pos(fd);
1da177e4 944 ssize_t ret = -EBADF;
1da177e4 945
2903ff01 946 if (f.file) {
438ab720
KS
947 loff_t pos, *ppos = file_ppos(f.file);
948 if (ppos) {
949 pos = *ppos;
950 ppos = &pos;
951 }
952 ret = vfs_readv(f.file, vec, vlen, ppos, flags);
953 if (ret >= 0 && ppos)
954 f.file->f_pos = pos;
9c225f26 955 fdput_pos(f);
1da177e4
LT
956 }
957
958 if (ret > 0)
4b98d11b
AD
959 add_rchar(current, ret);
960 inc_syscr(current);
1da177e4
LT
961 return ret;
962}
963
f17d8b35 964static ssize_t do_writev(unsigned long fd, const struct iovec __user *vec,
ddef7ed2 965 unsigned long vlen, rwf_t flags)
1da177e4 966{
9c225f26 967 struct fd f = fdget_pos(fd);
1da177e4 968 ssize_t ret = -EBADF;
1da177e4 969
2903ff01 970 if (f.file) {
438ab720
KS
971 loff_t pos, *ppos = file_ppos(f.file);
972 if (ppos) {
973 pos = *ppos;
974 ppos = &pos;
975 }
976 ret = vfs_writev(f.file, vec, vlen, ppos, flags);
977 if (ret >= 0 && ppos)
978 f.file->f_pos = pos;
9c225f26 979 fdput_pos(f);
1da177e4
LT
980 }
981
982 if (ret > 0)
4b98d11b
AD
983 add_wchar(current, ret);
984 inc_syscw(current);
1da177e4
LT
985 return ret;
986}
987
601cc11d
LT
988static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
989{
990#define HALF_LONG_BITS (BITS_PER_LONG / 2)
991 return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
992}
993
f17d8b35 994static ssize_t do_preadv(unsigned long fd, const struct iovec __user *vec,
ddef7ed2 995 unsigned long vlen, loff_t pos, rwf_t flags)
f3554f4b 996{
2903ff01 997 struct fd f;
f3554f4b 998 ssize_t ret = -EBADF;
f3554f4b
GH
999
1000 if (pos < 0)
1001 return -EINVAL;
1002
2903ff01
AV
1003 f = fdget(fd);
1004 if (f.file) {
f3554f4b 1005 ret = -ESPIPE;
2903ff01 1006 if (f.file->f_mode & FMODE_PREAD)
f17d8b35 1007 ret = vfs_readv(f.file, vec, vlen, &pos, flags);
2903ff01 1008 fdput(f);
f3554f4b
GH
1009 }
1010
1011 if (ret > 0)
1012 add_rchar(current, ret);
1013 inc_syscr(current);
1014 return ret;
1015}
1016
f17d8b35 1017static ssize_t do_pwritev(unsigned long fd, const struct iovec __user *vec,
ddef7ed2 1018 unsigned long vlen, loff_t pos, rwf_t flags)
f3554f4b 1019{
2903ff01 1020 struct fd f;
f3554f4b 1021 ssize_t ret = -EBADF;
f3554f4b
GH
1022
1023 if (pos < 0)
1024 return -EINVAL;
1025
2903ff01
AV
1026 f = fdget(fd);
1027 if (f.file) {
f3554f4b 1028 ret = -ESPIPE;
2903ff01 1029 if (f.file->f_mode & FMODE_PWRITE)
f17d8b35 1030 ret = vfs_writev(f.file, vec, vlen, &pos, flags);
2903ff01 1031 fdput(f);
f3554f4b
GH
1032 }
1033
1034 if (ret > 0)
1035 add_wchar(current, ret);
1036 inc_syscw(current);
1037 return ret;
1038}
1039
f17d8b35
MT
1040SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
1041 unsigned long, vlen)
1042{
1043 return do_readv(fd, vec, vlen, 0);
1044}
1045
1046SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
1047 unsigned long, vlen)
1048{
1049 return do_writev(fd, vec, vlen, 0);
1050}
1051
1052SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
1053 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
1054{
1055 loff_t pos = pos_from_hilo(pos_h, pos_l);
1056
1057 return do_preadv(fd, vec, vlen, pos, 0);
1058}
1059
1060SYSCALL_DEFINE6(preadv2, unsigned long, fd, const struct iovec __user *, vec,
1061 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
ddef7ed2 1062 rwf_t, flags)
f17d8b35
MT
1063{
1064 loff_t pos = pos_from_hilo(pos_h, pos_l);
1065
1066 if (pos == -1)
1067 return do_readv(fd, vec, vlen, flags);
1068
1069 return do_preadv(fd, vec, vlen, pos, flags);
1070}
1071
1072SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
1073 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
1074{
1075 loff_t pos = pos_from_hilo(pos_h, pos_l);
1076
1077 return do_pwritev(fd, vec, vlen, pos, 0);
1078}
1079
1080SYSCALL_DEFINE6(pwritev2, unsigned long, fd, const struct iovec __user *, vec,
1081 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
ddef7ed2 1082 rwf_t, flags)
f17d8b35
MT
1083{
1084 loff_t pos = pos_from_hilo(pos_h, pos_l);
1085
1086 if (pos == -1)
1087 return do_writev(fd, vec, vlen, flags);
1088
1089 return do_pwritev(fd, vec, vlen, pos, flags);
1090}
1091
3523a9d4
CH
1092/*
1093 * Various compat syscalls. Note that they all pretend to take a native
1094 * iovec - import_iovec will properly treat those as compat_iovecs based on
1095 * in_compat_syscall().
1096 */
72ec3516 1097#ifdef CONFIG_COMPAT
378a10f3
HC
1098#ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1099COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
3523a9d4 1100 const struct iovec __user *, vec,
378a10f3
HC
1101 unsigned long, vlen, loff_t, pos)
1102{
3523a9d4 1103 return do_preadv(fd, vec, vlen, pos, 0);
378a10f3
HC
1104}
1105#endif
1106
dfd948e3 1107COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd,
3523a9d4 1108 const struct iovec __user *, vec,
dfd948e3 1109 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
72ec3516
AV
1110{
1111 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
378a10f3 1112
3523a9d4 1113 return do_preadv(fd, vec, vlen, pos, 0);
f17d8b35
MT
1114}
1115
3ebfd81f
L
1116#ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1117COMPAT_SYSCALL_DEFINE5(preadv64v2, unsigned long, fd,
3523a9d4 1118 const struct iovec __user *, vec,
ddef7ed2 1119 unsigned long, vlen, loff_t, pos, rwf_t, flags)
3ebfd81f 1120{
cc4b1242 1121 if (pos == -1)
3523a9d4
CH
1122 return do_readv(fd, vec, vlen, flags);
1123 return do_preadv(fd, vec, vlen, pos, flags);
3ebfd81f
L
1124}
1125#endif
1126
f17d8b35 1127COMPAT_SYSCALL_DEFINE6(preadv2, compat_ulong_t, fd,
3523a9d4 1128 const struct iovec __user *, vec,
f17d8b35 1129 compat_ulong_t, vlen, u32, pos_low, u32, pos_high,
ddef7ed2 1130 rwf_t, flags)
f17d8b35
MT
1131{
1132 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1133
1134 if (pos == -1)
3523a9d4
CH
1135 return do_readv(fd, vec, vlen, flags);
1136 return do_preadv(fd, vec, vlen, pos, flags);
72ec3516
AV
1137}
1138
378a10f3
HC
1139#ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1140COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
3523a9d4 1141 const struct iovec __user *, vec,
378a10f3
HC
1142 unsigned long, vlen, loff_t, pos)
1143{
3523a9d4 1144 return do_pwritev(fd, vec, vlen, pos, 0);
378a10f3
HC
1145}
1146#endif
1147
dfd948e3 1148COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd,
3523a9d4 1149 const struct iovec __user *,vec,
dfd948e3 1150 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
72ec3516
AV
1151{
1152 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
378a10f3 1153
3523a9d4 1154 return do_pwritev(fd, vec, vlen, pos, 0);
72ec3516 1155}
f17d8b35 1156
3ebfd81f
L
1157#ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1158COMPAT_SYSCALL_DEFINE5(pwritev64v2, unsigned long, fd,
3523a9d4 1159 const struct iovec __user *, vec,
ddef7ed2 1160 unsigned long, vlen, loff_t, pos, rwf_t, flags)
3ebfd81f 1161{
cc4b1242 1162 if (pos == -1)
3523a9d4
CH
1163 return do_writev(fd, vec, vlen, flags);
1164 return do_pwritev(fd, vec, vlen, pos, flags);
3ebfd81f
L
1165}
1166#endif
1167
f17d8b35 1168COMPAT_SYSCALL_DEFINE6(pwritev2, compat_ulong_t, fd,
3523a9d4 1169 const struct iovec __user *,vec,
ddef7ed2 1170 compat_ulong_t, vlen, u32, pos_low, u32, pos_high, rwf_t, flags)
f17d8b35
MT
1171{
1172 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1173
1174 if (pos == -1)
3523a9d4
CH
1175 return do_writev(fd, vec, vlen, flags);
1176 return do_pwritev(fd, vec, vlen, pos, flags);
72ec3516 1177}
3523a9d4 1178#endif /* CONFIG_COMPAT */
72ec3516 1179
19f4fc3a
AV
1180static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
1181 size_t count, loff_t max)
1da177e4 1182{
2903ff01
AV
1183 struct fd in, out;
1184 struct inode *in_inode, *out_inode;
b964bf53 1185 struct pipe_inode_info *opipe;
1da177e4 1186 loff_t pos;
7995bd28 1187 loff_t out_pos;
1da177e4 1188 ssize_t retval;
2903ff01 1189 int fl;
1da177e4
LT
1190
1191 /*
1192 * Get input file, and verify that it is ok..
1193 */
1194 retval = -EBADF;
2903ff01
AV
1195 in = fdget(in_fd);
1196 if (!in.file)
1da177e4 1197 goto out;
2903ff01 1198 if (!(in.file->f_mode & FMODE_READ))
1da177e4 1199 goto fput_in;
1da177e4 1200 retval = -ESPIPE;
7995bd28
AV
1201 if (!ppos) {
1202 pos = in.file->f_pos;
1203 } else {
1204 pos = *ppos;
2903ff01 1205 if (!(in.file->f_mode & FMODE_PREAD))
1da177e4 1206 goto fput_in;
7995bd28
AV
1207 }
1208 retval = rw_verify_area(READ, in.file, &pos, count);
e28cc715 1209 if (retval < 0)
1da177e4 1210 goto fput_in;
bc61384d
AV
1211 if (count > MAX_RW_COUNT)
1212 count = MAX_RW_COUNT;
1da177e4 1213
1da177e4
LT
1214 /*
1215 * Get output file, and verify that it is ok..
1216 */
1217 retval = -EBADF;
2903ff01
AV
1218 out = fdget(out_fd);
1219 if (!out.file)
1da177e4 1220 goto fput_in;
2903ff01 1221 if (!(out.file->f_mode & FMODE_WRITE))
1da177e4 1222 goto fput_out;
496ad9aa
AV
1223 in_inode = file_inode(in.file);
1224 out_inode = file_inode(out.file);
7995bd28 1225 out_pos = out.file->f_pos;
1da177e4 1226
1da177e4
LT
1227 if (!max)
1228 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
1229
1da177e4
LT
1230 if (unlikely(pos + count > max)) {
1231 retval = -EOVERFLOW;
1232 if (pos >= max)
1233 goto fput_out;
1234 count = max - pos;
1235 }
1236
d96e6e71 1237 fl = 0;
534f2aaa 1238#if 0
d96e6e71
JA
1239 /*
1240 * We need to debate whether we can enable this or not. The
1241 * man page documents EAGAIN return for the output at least,
1242 * and the application is arguably buggy if it doesn't expect
1243 * EAGAIN on a non-blocking file descriptor.
1244 */
2903ff01 1245 if (in.file->f_flags & O_NONBLOCK)
d96e6e71 1246 fl = SPLICE_F_NONBLOCK;
534f2aaa 1247#endif
b964bf53
AV
1248 opipe = get_pipe_info(out.file, true);
1249 if (!opipe) {
1250 retval = rw_verify_area(WRITE, out.file, &out_pos, count);
1251 if (retval < 0)
1252 goto fput_out;
1253 file_start_write(out.file);
1254 retval = do_splice_direct(in.file, &pos, out.file, &out_pos,
1255 count, fl);
1256 file_end_write(out.file);
1257 } else {
bdeb77bc
AV
1258 if (out.file->f_flags & O_NONBLOCK)
1259 fl |= SPLICE_F_NONBLOCK;
1260
b964bf53
AV
1261 retval = splice_file_to_pipe(in.file, opipe, &pos, count, fl);
1262 }
1da177e4
LT
1263
1264 if (retval > 0) {
4b98d11b
AD
1265 add_rchar(current, retval);
1266 add_wchar(current, retval);
a68c2f12
SW
1267 fsnotify_access(in.file);
1268 fsnotify_modify(out.file);
7995bd28
AV
1269 out.file->f_pos = out_pos;
1270 if (ppos)
1271 *ppos = pos;
1272 else
1273 in.file->f_pos = pos;
1da177e4 1274 }
1da177e4 1275
4b98d11b
AD
1276 inc_syscr(current);
1277 inc_syscw(current);
7995bd28 1278 if (pos > max)
1da177e4
LT
1279 retval = -EOVERFLOW;
1280
1281fput_out:
2903ff01 1282 fdput(out);
1da177e4 1283fput_in:
2903ff01 1284 fdput(in);
1da177e4
LT
1285out:
1286 return retval;
1287}
1288
002c8976 1289SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
1da177e4
LT
1290{
1291 loff_t pos;
1292 off_t off;
1293 ssize_t ret;
1294
1295 if (offset) {
1296 if (unlikely(get_user(off, offset)))
1297 return -EFAULT;
1298 pos = off;
1299 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1300 if (unlikely(put_user(pos, offset)))
1301 return -EFAULT;
1302 return ret;
1303 }
1304
1305 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1306}
1307
002c8976 1308SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
1da177e4
LT
1309{
1310 loff_t pos;
1311 ssize_t ret;
1312
1313 if (offset) {
1314 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1315 return -EFAULT;
1316 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1317 if (unlikely(put_user(pos, offset)))
1318 return -EFAULT;
1319 return ret;
1320 }
1321
1322 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1323}
19f4fc3a
AV
1324
1325#ifdef CONFIG_COMPAT
1326COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
1327 compat_off_t __user *, offset, compat_size_t, count)
1328{
1329 loff_t pos;
1330 off_t off;
1331 ssize_t ret;
1332
1333 if (offset) {
1334 if (unlikely(get_user(off, offset)))
1335 return -EFAULT;
1336 pos = off;
1337 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1338 if (unlikely(put_user(pos, offset)))
1339 return -EFAULT;
1340 return ret;
1341 }
1342
1343 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1344}
1345
1346COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
1347 compat_loff_t __user *, offset, compat_size_t, count)
1348{
1349 loff_t pos;
1350 ssize_t ret;
1351
1352 if (offset) {
1353 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1354 return -EFAULT;
1355 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1356 if (unlikely(put_user(pos, offset)))
1357 return -EFAULT;
1358 return ret;
1359 }
1360
1361 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1362}
1363#endif
29732938 1364
f16acc9d
DC
1365/**
1366 * generic_copy_file_range - copy data between two files
1367 * @file_in: file structure to read from
1368 * @pos_in: file offset to read from
1369 * @file_out: file structure to write data to
1370 * @pos_out: file offset to write data to
1371 * @len: amount of data to copy
1372 * @flags: copy flags
1373 *
1374 * This is a generic filesystem helper to copy data from one file to another.
1375 * It has no constraints on the source or destination file owners - the files
1376 * can belong to different superblocks and different filesystem types. Short
1377 * copies are allowed.
1378 *
1379 * This should be called from the @file_out filesystem, as per the
1380 * ->copy_file_range() method.
1381 *
1382 * Returns the number of bytes copied or a negative error indicating the
1383 * failure.
1384 */
1385
1386ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
1387 struct file *file_out, loff_t pos_out,
1388 size_t len, unsigned int flags)
1389{
10bc8e4a
AG
1390 lockdep_assert(sb_write_started(file_inode(file_out)->i_sb));
1391
f16acc9d
DC
1392 return do_splice_direct(file_in, &pos_in, file_out, &pos_out,
1393 len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
1394}
1395EXPORT_SYMBOL(generic_copy_file_range);
1396
407e9c63
DW
1397/*
1398 * Performs necessary checks before doing a file copy
1399 *
1400 * Can adjust amount of bytes to copy via @req_count argument.
1401 * Returns appropriate error code that caller should return or
1402 * zero in case the copy should be allowed.
1403 */
1404static int generic_copy_file_checks(struct file *file_in, loff_t pos_in,
1405 struct file *file_out, loff_t pos_out,
1406 size_t *req_count, unsigned int flags)
1407{
1408 struct inode *inode_in = file_inode(file_in);
1409 struct inode *inode_out = file_inode(file_out);
1410 uint64_t count = *req_count;
1411 loff_t size_in;
1412 int ret;
1413
1414 ret = generic_file_rw_checks(file_in, file_out);
1415 if (ret)
1416 return ret;
1417
868f9f2f
AG
1418 /*
1419 * We allow some filesystems to handle cross sb copy, but passing
1420 * a file of the wrong filesystem type to filesystem driver can result
1421 * in an attempt to dereference the wrong type of ->private_data, so
1422 * avoid doing that until we really have a good reason.
1423 *
1424 * nfs and cifs define several different file_system_type structures
1425 * and several different sets of file_operations, but they all end up
1426 * using the same ->copy_file_range() function pointer.
1427 */
10bc8e4a
AG
1428 if (flags & COPY_FILE_SPLICE) {
1429 /* cross sb splice is allowed */
1430 } else if (file_out->f_op->copy_file_range) {
868f9f2f
AG
1431 if (file_in->f_op->copy_file_range !=
1432 file_out->f_op->copy_file_range)
1433 return -EXDEV;
1434 } else if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb) {
1435 return -EXDEV;
1436 }
1437
407e9c63
DW
1438 /* Don't touch certain kinds of inodes */
1439 if (IS_IMMUTABLE(inode_out))
1440 return -EPERM;
1441
1442 if (IS_SWAPFILE(inode_in) || IS_SWAPFILE(inode_out))
1443 return -ETXTBSY;
1444
1445 /* Ensure offsets don't wrap. */
1446 if (pos_in + count < pos_in || pos_out + count < pos_out)
1447 return -EOVERFLOW;
1448
1449 /* Shorten the copy to EOF */
1450 size_in = i_size_read(inode_in);
1451 if (pos_in >= size_in)
1452 count = 0;
1453 else
1454 count = min(count, size_in - (uint64_t)pos_in);
1455
1456 ret = generic_write_check_limits(file_out, pos_out, &count);
1457 if (ret)
1458 return ret;
1459
1460 /* Don't allow overlapped copying within the same file. */
1461 if (inode_in == inode_out &&
1462 pos_out + count > pos_in &&
1463 pos_out < pos_in + count)
1464 return -EINVAL;
1465
1466 *req_count = count;
1467 return 0;
1468}
1469
29732938
ZB
1470/*
1471 * copy_file_range() differs from regular file read and write in that it
1472 * specifically allows return partial success. When it does so is up to
1473 * the copy_file_range method.
1474 */
1475ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
1476 struct file *file_out, loff_t pos_out,
1477 size_t len, unsigned int flags)
1478{
29732938 1479 ssize_t ret;
10bc8e4a 1480 bool splice = flags & COPY_FILE_SPLICE;
29732938 1481
10bc8e4a 1482 if (flags & ~COPY_FILE_SPLICE)
29732938
ZB
1483 return -EINVAL;
1484
96e6e8f4
AG
1485 ret = generic_copy_file_checks(file_in, pos_in, file_out, pos_out, &len,
1486 flags);
a3171351
AG
1487 if (unlikely(ret))
1488 return ret;
11cbfb10 1489
29732938 1490 ret = rw_verify_area(READ, file_in, &pos_in, len);
bc61384d
AV
1491 if (unlikely(ret))
1492 return ret;
1493
1494 ret = rw_verify_area(WRITE, file_out, &pos_out, len);
1495 if (unlikely(ret))
29732938
ZB
1496 return ret;
1497
29732938
ZB
1498 if (len == 0)
1499 return 0;
1500
bfe219d3 1501 file_start_write(file_out);
29732938 1502
a76b5b04 1503 /*
868f9f2f
AG
1504 * Cloning is supported by more file systems, so we implement copy on
1505 * same sb using clone, but for filesystems where both clone and copy
1506 * are supported (e.g. nfs,cifs), we only call the copy method.
a76b5b04 1507 */
10bc8e4a 1508 if (!splice && file_out->f_op->copy_file_range) {
868f9f2f
AG
1509 ret = file_out->f_op->copy_file_range(file_in, pos_in,
1510 file_out, pos_out,
1511 len, flags);
1512 goto done;
1513 }
1514
10bc8e4a 1515 if (!splice && file_in->f_op->remap_file_range &&
5dae222a 1516 file_inode(file_in)->i_sb == file_inode(file_out)->i_sb) {
868f9f2f 1517 ret = file_in->f_op->remap_file_range(file_in, pos_in,
42ec3d4c 1518 file_out, pos_out,
eca3654e
DW
1519 min_t(loff_t, MAX_RW_COUNT, len),
1520 REMAP_FILE_CAN_SHORTEN);
868f9f2f 1521 if (ret > 0)
a76b5b04 1522 goto done;
a76b5b04
CH
1523 }
1524
868f9f2f
AG
1525 /*
1526 * We can get here for same sb copy of filesystems that do not implement
1527 * ->copy_file_range() in case filesystem does not support clone or in
1528 * case filesystem supports clone but rejected the clone request (e.g.
1529 * because it was not block aligned).
1530 *
1531 * In both cases, fall back to kernel copy so we are able to maintain a
1532 * consistent story about which filesystems support copy_file_range()
1533 * and which filesystems do not, that will allow userspace tools to
1534 * make consistent desicions w.r.t using copy_file_range().
10bc8e4a
AG
1535 *
1536 * We also get here if caller (e.g. nfsd) requested COPY_FILE_SPLICE.
868f9f2f
AG
1537 */
1538 ret = generic_copy_file_range(file_in, pos_in, file_out, pos_out, len,
1539 flags);
1540
a76b5b04 1541done:
29732938
ZB
1542 if (ret > 0) {
1543 fsnotify_access(file_in);
1544 add_rchar(current, ret);
1545 fsnotify_modify(file_out);
1546 add_wchar(current, ret);
1547 }
a76b5b04 1548
29732938
ZB
1549 inc_syscr(current);
1550 inc_syscw(current);
1551
bfe219d3 1552 file_end_write(file_out);
29732938
ZB
1553
1554 return ret;
1555}
1556EXPORT_SYMBOL(vfs_copy_file_range);
1557
1558SYSCALL_DEFINE6(copy_file_range, int, fd_in, loff_t __user *, off_in,
1559 int, fd_out, loff_t __user *, off_out,
1560 size_t, len, unsigned int, flags)
1561{
1562 loff_t pos_in;
1563 loff_t pos_out;
1564 struct fd f_in;
1565 struct fd f_out;
1566 ssize_t ret = -EBADF;
1567
1568 f_in = fdget(fd_in);
1569 if (!f_in.file)
1570 goto out2;
1571
1572 f_out = fdget(fd_out);
1573 if (!f_out.file)
1574 goto out1;
1575
1576 ret = -EFAULT;
1577 if (off_in) {
1578 if (copy_from_user(&pos_in, off_in, sizeof(loff_t)))
1579 goto out;
1580 } else {
1581 pos_in = f_in.file->f_pos;
1582 }
1583
1584 if (off_out) {
1585 if (copy_from_user(&pos_out, off_out, sizeof(loff_t)))
1586 goto out;
1587 } else {
1588 pos_out = f_out.file->f_pos;
1589 }
1590
10bc8e4a
AG
1591 ret = -EINVAL;
1592 if (flags != 0)
1593 goto out;
1594
29732938
ZB
1595 ret = vfs_copy_file_range(f_in.file, pos_in, f_out.file, pos_out, len,
1596 flags);
1597 if (ret > 0) {
1598 pos_in += ret;
1599 pos_out += ret;
1600
1601 if (off_in) {
1602 if (copy_to_user(off_in, &pos_in, sizeof(loff_t)))
1603 ret = -EFAULT;
1604 } else {
1605 f_in.file->f_pos = pos_in;
1606 }
1607
1608 if (off_out) {
1609 if (copy_to_user(off_out, &pos_out, sizeof(loff_t)))
1610 ret = -EFAULT;
1611 } else {
1612 f_out.file->f_pos = pos_out;
1613 }
1614 }
1615
1616out:
1617 fdput(f_out);
1618out1:
1619 fdput(f_in);
1620out2:
1621 return ret;
1622}
04b38d60 1623
edc58dd0 1624/*
407e9c63
DW
1625 * Don't operate on ranges the page cache doesn't support, and don't exceed the
1626 * LFS limits. If pos is under the limit it becomes a short access. If it
1627 * exceeds the limit we return -EFBIG.
edc58dd0 1628 */
407e9c63 1629int generic_write_check_limits(struct file *file, loff_t pos, loff_t *count)
edc58dd0 1630{
407e9c63
DW
1631 struct inode *inode = file->f_mapping->host;
1632 loff_t max_size = inode->i_sb->s_maxbytes;
1633 loff_t limit = rlimit(RLIMIT_FSIZE);
edc58dd0 1634
407e9c63
DW
1635 if (limit != RLIM_INFINITY) {
1636 if (pos >= limit) {
1637 send_sig(SIGXFSZ, current, 0);
1638 return -EFBIG;
edc58dd0 1639 }
407e9c63
DW
1640 *count = min(*count, limit - pos);
1641 }
edc58dd0 1642
407e9c63
DW
1643 if (!(file->f_flags & O_LARGEFILE))
1644 max_size = MAX_NON_LFS;
c32e5f39 1645
407e9c63
DW
1646 if (unlikely(pos >= max_size))
1647 return -EFBIG;
c32e5f39 1648
407e9c63 1649 *count = min(*count, max_size - pos);
c32e5f39 1650
c32e5f39 1651 return 0;
c32e5f39 1652}
04b38d60 1653
f6f7a25a
OS
1654/* Like generic_write_checks(), but takes size of write instead of iter. */
1655int generic_write_checks_count(struct kiocb *iocb, loff_t *count)
876bec6f 1656{
407e9c63
DW
1657 struct file *file = iocb->ki_filp;
1658 struct inode *inode = file->f_mapping->host;
876bec6f 1659
407e9c63 1660 if (IS_SWAPFILE(inode))
876bec6f
DW
1661 return -ETXTBSY;
1662
f6f7a25a 1663 if (!*count)
407e9c63 1664 return 0;
5de4480a 1665
407e9c63
DW
1666 if (iocb->ki_flags & IOCB_APPEND)
1667 iocb->ki_pos = i_size_read(inode);
1b4f42a1 1668
80175539
SR
1669 if ((iocb->ki_flags & IOCB_NOWAIT) &&
1670 !((iocb->ki_flags & IOCB_DIRECT) ||
1671 (file->f_mode & FMODE_BUF_WASYNC)))
407e9c63 1672 return -EINVAL;
1b4f42a1 1673
f6f7a25a
OS
1674 return generic_write_check_limits(iocb->ki_filp, iocb->ki_pos, count);
1675}
1676EXPORT_SYMBOL(generic_write_checks_count);
1677
1678/*
1679 * Performs necessary checks before doing a write
1680 *
1681 * Can adjust writing position or amount of bytes to write.
1682 * Returns appropriate error code that caller should return or
1683 * zero in case that write should be allowed.
1684 */
1685ssize_t generic_write_checks(struct kiocb *iocb, struct iov_iter *from)
1686{
1687 loff_t count = iov_iter_count(from);
1688 int ret;
1689
1690 ret = generic_write_checks_count(iocb, &count);
1b4f42a1
MS
1691 if (ret)
1692 return ret;
1693
407e9c63
DW
1694 iov_iter_truncate(from, count);
1695 return iov_iter_count(from);
1b4f42a1 1696}
407e9c63 1697EXPORT_SYMBOL(generic_write_checks);
1b4f42a1 1698
407e9c63
DW
1699/*
1700 * Performs common checks before doing a file copy/clone
1701 * from @file_in to @file_out.
1702 */
1703int generic_file_rw_checks(struct file *file_in, struct file *file_out)
54dbc151 1704{
407e9c63
DW
1705 struct inode *inode_in = file_inode(file_in);
1706 struct inode *inode_out = file_inode(file_out);
54dbc151 1707
407e9c63
DW
1708 /* Don't copy dirs, pipes, sockets... */
1709 if (S_ISDIR(inode_in->i_mode) || S_ISDIR(inode_out->i_mode))
494633fa 1710 return -EISDIR;
407e9c63 1711 if (!S_ISREG(inode_in->i_mode) || !S_ISREG(inode_out->i_mode))
22725ce4
DW
1712 return -EINVAL;
1713
407e9c63
DW
1714 if (!(file_in->f_mode & FMODE_READ) ||
1715 !(file_out->f_mode & FMODE_WRITE) ||
1716 (file_out->f_flags & O_APPEND))
1717 return -EBADF;
1b4f42a1 1718
407e9c63 1719 return 0;
54dbc151 1720}