blob: e6ddc8dceb96fc48a8fbe379799d0efed52d2b4a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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>
Robert Love0eeca282005-07-12 17:06:03 -040012#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/security.h>
Paul Gortmaker630d9c42011-11-16 23:57:37 -050014#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/syscalls.h>
Linus Torvaldse28cc712006-01-04 16:20:40 -080016#include <linux/pagemap.h>
Jens Axboed6b29d72007-06-04 09:59:47 +020017#include <linux/splice.h>
Al Viro561c6732013-02-24 10:52:26 -050018#include <linux/compat.h>
Badari Pulavartyee0b3e62006-09-30 23:28:47 -070019#include "read_write.h"
Al Viro06ae43f2013-03-20 13:19:30 -040020#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
22#include <asm/uaccess.h>
23#include <asm/unistd.h>
24
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080025const struct file_operations generic_ro_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 .llseek = generic_file_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -070027 .read = do_sync_read,
28 .aio_read = generic_file_aio_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070029 .mmap = generic_file_readonly_mmap,
Jens Axboe534f2aa2007-06-01 14:52:37 +020030 .splice_read = generic_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070031};
32
33EXPORT_SYMBOL(generic_ro_fops);
34
Al Virocccb5a12010-12-17 07:44:05 -050035static inline int unsigned_offsets(struct file *file)
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070036{
Al Virocccb5a12010-12-17 07:44:05 -050037 return file->f_mode & FMODE_UNSIGNED_OFFSET;
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070038}
39
Andi Kleenef3d0fd2011-09-15 16:06:48 -070040static loff_t lseek_execute(struct file *file, struct inode *inode,
41 loff_t offset, loff_t maxsize)
42{
43 if (offset < 0 && !unsigned_offsets(file))
44 return -EINVAL;
45 if (offset > maxsize)
46 return -EINVAL;
47
48 if (offset != file->f_pos) {
49 file->f_pos = offset;
50 file->f_version = 0;
51 }
52 return offset;
53}
54
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020055/**
Andi Kleen57604952011-09-15 16:06:50 -070056 * generic_file_llseek_size - generic llseek implementation for regular files
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020057 * @file: file structure to seek on
58 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -080059 * @whence: type of seek
Eric Sandeene8b96eb2012-04-30 13:11:29 -050060 * @size: max size of this file in file system
61 * @eof: offset used for SEEK_END position
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020062 *
Andi Kleen57604952011-09-15 16:06:50 -070063 * This is a variant of generic_file_llseek that allows passing in a custom
Eric Sandeene8b96eb2012-04-30 13:11:29 -050064 * maximum file size and a custom EOF position, for e.g. hashed directories
Andi Kleenef3d0fd2011-09-15 16:06:48 -070065 *
66 * Synchronization:
Andi Kleen57604952011-09-15 16:06:50 -070067 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
Andi Kleenef3d0fd2011-09-15 16:06:48 -070068 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
69 * read/writes behave like SEEK_SET against seeks.
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020070 */
Andi Kleen9465efc2008-06-27 11:05:24 +020071loff_t
Andrew Morton965c8e52012-12-17 15:59:39 -080072generic_file_llseek_size(struct file *file, loff_t offset, int whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -050073 loff_t maxsize, loff_t eof)
Linus Torvalds1da177e2005-04-16 15:20:36 -070074{
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 struct inode *inode = file->f_mapping->host;
76
Andrew Morton965c8e52012-12-17 15:59:39 -080077 switch (whence) {
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020078 case SEEK_END:
Eric Sandeene8b96eb2012-04-30 13:11:29 -050079 offset += eof;
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020080 break;
81 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -080082 /*
83 * Here we special-case the lseek(fd, 0, SEEK_CUR)
84 * position-querying operation. Avoid rewriting the "same"
85 * f_pos value back to the file because a concurrent read(),
86 * write() or lseek() might have altered it
87 */
88 if (offset == 0)
89 return file->f_pos;
Andi Kleenef3d0fd2011-09-15 16:06:48 -070090 /*
91 * f_lock protects against read/modify/write race with other
92 * SEEK_CURs. Note that parallel writes and reads behave
93 * like SEEK_SET.
94 */
95 spin_lock(&file->f_lock);
96 offset = lseek_execute(file, inode, file->f_pos + offset,
Andi Kleen57604952011-09-15 16:06:50 -070097 maxsize);
Andi Kleenef3d0fd2011-09-15 16:06:48 -070098 spin_unlock(&file->f_lock);
99 return offset;
Josef Bacik982d8162011-07-18 13:21:35 -0400100 case SEEK_DATA:
101 /*
102 * In the generic case the entire file is data, so as long as
103 * offset isn't at the end of the file then the offset is data.
104 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500105 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400106 return -ENXIO;
107 break;
108 case SEEK_HOLE:
109 /*
110 * There is a virtual hole at the end of the file, so as long as
111 * offset isn't i_size or larger, return i_size.
112 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500113 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400114 return -ENXIO;
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500115 offset = eof;
Josef Bacik982d8162011-07-18 13:21:35 -0400116 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 }
Christoph Hellwig3a8cff42008-08-11 15:37:17 +0200118
Andi Kleen57604952011-09-15 16:06:50 -0700119 return lseek_execute(file, inode, offset, maxsize);
120}
121EXPORT_SYMBOL(generic_file_llseek_size);
122
123/**
124 * generic_file_llseek - generic llseek implementation for regular files
125 * @file: file structure to seek on
126 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800127 * @whence: type of seek
Andi Kleen57604952011-09-15 16:06:50 -0700128 *
129 * This is a generic implemenation of ->llseek useable for all normal local
130 * filesystems. It just updates the file offset to the value specified by
Andrew Morton965c8e52012-12-17 15:59:39 -0800131 * @offset and @whence under i_mutex.
Andi Kleen57604952011-09-15 16:06:50 -0700132 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800133loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
Andi Kleen57604952011-09-15 16:06:50 -0700134{
135 struct inode *inode = file->f_mapping->host;
136
Andrew Morton965c8e52012-12-17 15:59:39 -0800137 return generic_file_llseek_size(file, offset, whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500138 inode->i_sb->s_maxbytes,
139 i_size_read(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140}
Andi Kleen9465efc2008-06-27 11:05:24 +0200141EXPORT_SYMBOL(generic_file_llseek);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142
jan Blunckae6afc32010-05-26 14:44:48 -0700143/**
144 * noop_llseek - No Operation Performed llseek implementation
145 * @file: file structure to seek on
146 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800147 * @whence: type of seek
jan Blunckae6afc32010-05-26 14:44:48 -0700148 *
149 * This is an implementation of ->llseek useable for the rare special case when
150 * userspace expects the seek to succeed but the (device) file is actually not
151 * able to perform the seek. In this case you use noop_llseek() instead of
152 * falling back to the default implementation of ->llseek.
153 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800154loff_t noop_llseek(struct file *file, loff_t offset, int whence)
jan Blunckae6afc32010-05-26 14:44:48 -0700155{
156 return file->f_pos;
157}
158EXPORT_SYMBOL(noop_llseek);
159
Andrew Morton965c8e52012-12-17 15:59:39 -0800160loff_t no_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
162 return -ESPIPE;
163}
164EXPORT_SYMBOL(no_llseek);
165
Andrew Morton965c8e52012-12-17 15:59:39 -0800166loff_t default_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167{
Al Viro496ad9a2013-01-23 17:07:38 -0500168 struct inode *inode = file_inode(file);
David Sterba16abef02008-04-22 15:09:22 +0200169 loff_t retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
Josef Bacik982d8162011-07-18 13:21:35 -0400171 mutex_lock(&inode->i_mutex);
Andrew Morton965c8e52012-12-17 15:59:39 -0800172 switch (whence) {
Chris Snook7b8e8922007-05-08 00:24:13 -0700173 case SEEK_END:
Josef Bacik982d8162011-07-18 13:21:35 -0400174 offset += i_size_read(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 break;
Chris Snook7b8e8922007-05-08 00:24:13 -0700176 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800177 if (offset == 0) {
178 retval = file->f_pos;
179 goto out;
180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 offset += file->f_pos;
Josef Bacik982d8162011-07-18 13:21:35 -0400182 break;
183 case SEEK_DATA:
184 /*
185 * In the generic case the entire file is data, so as
186 * long as offset isn't at the end of the file then the
187 * offset is data.
188 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300189 if (offset >= inode->i_size) {
190 retval = -ENXIO;
191 goto out;
192 }
Josef Bacik982d8162011-07-18 13:21:35 -0400193 break;
194 case SEEK_HOLE:
195 /*
196 * There is a virtual hole at the end of the file, so
197 * as long as offset isn't i_size or larger, return
198 * i_size.
199 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300200 if (offset >= inode->i_size) {
201 retval = -ENXIO;
202 goto out;
203 }
Josef Bacik982d8162011-07-18 13:21:35 -0400204 offset = inode->i_size;
205 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 }
207 retval = -EINVAL;
Al Virocccb5a12010-12-17 07:44:05 -0500208 if (offset >= 0 || unsigned_offsets(file)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 if (offset != file->f_pos) {
210 file->f_pos = offset;
211 file->f_version = 0;
212 }
213 retval = offset;
214 }
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800215out:
Josef Bacik982d8162011-07-18 13:21:35 -0400216 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 return retval;
218}
219EXPORT_SYMBOL(default_llseek);
220
Andrew Morton965c8e52012-12-17 15:59:39 -0800221loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
223 loff_t (*fn)(struct file *, loff_t, int);
224
225 fn = no_llseek;
226 if (file->f_mode & FMODE_LSEEK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 if (file->f_op && file->f_op->llseek)
228 fn = file->f_op->llseek;
229 }
Andrew Morton965c8e52012-12-17 15:59:39 -0800230 return fn(file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231}
232EXPORT_SYMBOL(vfs_llseek);
233
Andrew Morton965c8e52012-12-17 15:59:39 -0800234SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235{
236 off_t retval;
Al Viro2903ff02012-08-28 12:52:22 -0400237 struct fd f = fdget(fd);
238 if (!f.file)
239 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800242 if (whence <= SEEK_MAX) {
243 loff_t res = vfs_llseek(f.file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 retval = res;
245 if (res != (loff_t)retval)
246 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
247 }
Al Viro2903ff02012-08-28 12:52:22 -0400248 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 return retval;
250}
251
Al Viro561c6732013-02-24 10:52:26 -0500252#ifdef CONFIG_COMPAT
253COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
254{
255 return sys_lseek(fd, offset, whence);
256}
257#endif
258
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259#ifdef __ARCH_WANT_SYS_LLSEEK
Heiko Carstens003d7ab2009-01-14 14:14:21 +0100260SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
261 unsigned long, offset_low, loff_t __user *, result,
Andrew Morton965c8e52012-12-17 15:59:39 -0800262 unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263{
264 int retval;
Al Viro2903ff02012-08-28 12:52:22 -0400265 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 loff_t offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Al Viro2903ff02012-08-28 12:52:22 -0400268 if (!f.file)
269 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
271 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800272 if (whence > SEEK_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 goto out_putf;
274
Al Viro2903ff02012-08-28 12:52:22 -0400275 offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
Andrew Morton965c8e52012-12-17 15:59:39 -0800276 whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
278 retval = (int)offset;
279 if (offset >= 0) {
280 retval = -EFAULT;
281 if (!copy_to_user(result, &offset, sizeof(offset)))
282 retval = 0;
283 }
284out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400285 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 return retval;
287}
288#endif
289
Linus Torvaldse28cc712006-01-04 16:20:40 -0800290/*
291 * rw_verify_area doesn't like huge counts. We limit
292 * them to something that fits in "int" so that others
293 * won't have to do range checks all the time.
294 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
296{
297 struct inode *inode;
298 loff_t pos;
James Morrisc43e2592008-01-12 22:05:48 +1100299 int retval = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
Al Viro496ad9a2013-01-23 17:07:38 -0500301 inode = file_inode(file);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800302 if (unlikely((ssize_t) count < 0))
James Morrisc43e2592008-01-12 22:05:48 +1100303 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 pos = *ppos;
Al Virocccb5a12010-12-17 07:44:05 -0500305 if (unlikely(pos < 0)) {
306 if (!unsigned_offsets(file))
307 return retval;
308 if (count >= -pos) /* both values are in 0..LLONG_MAX */
309 return -EOVERFLOW;
310 } else if (unlikely((loff_t) (pos + count) < 0)) {
311 if (!unsigned_offsets(file))
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700312 return retval;
313 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
Pavel Emelyanova16877c2007-10-01 14:41:11 -0700315 if (unlikely(inode->i_flock && mandatory_lock(inode))) {
James Morrisc43e2592008-01-12 22:05:48 +1100316 retval = locks_mandatory_area(
Linus Torvaldse28cc712006-01-04 16:20:40 -0800317 read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
318 inode, file, pos, count);
319 if (retval < 0)
320 return retval;
321 }
James Morrisc43e2592008-01-12 22:05:48 +1100322 retval = security_file_permission(file,
323 read_write == READ ? MAY_READ : MAY_WRITE);
324 if (retval)
325 return retval;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800326 return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327}
328
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700329static void wait_on_retry_sync_kiocb(struct kiocb *iocb)
330{
331 set_current_state(TASK_UNINTERRUPTIBLE);
332 if (!kiocbIsKicked(iocb))
333 schedule();
334 else
335 kiocbClearKicked(iocb);
336 __set_current_state(TASK_RUNNING);
337}
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
340{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700341 struct iovec iov = { .iov_base = buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 struct kiocb kiocb;
343 ssize_t ret;
344
345 init_sync_kiocb(&kiocb, filp);
346 kiocb.ki_pos = *ppos;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700347 kiocb.ki_left = len;
David Howells61964eb2010-03-24 17:09:19 +0000348 kiocb.ki_nbytes = len;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700349
350 for (;;) {
351 ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
352 if (ret != -EIOCBRETRY)
353 break;
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700354 wait_on_retry_sync_kiocb(&kiocb);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700355 }
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700356
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 if (-EIOCBQUEUED == ret)
358 ret = wait_on_sync_kiocb(&kiocb);
359 *ppos = kiocb.ki_pos;
360 return ret;
361}
362
363EXPORT_SYMBOL(do_sync_read);
364
365ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
366{
367 ssize_t ret;
368
369 if (!(file->f_mode & FMODE_READ))
370 return -EBADF;
371 if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
372 return -EINVAL;
373 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
374 return -EFAULT;
375
376 ret = rw_verify_area(READ, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800377 if (ret >= 0) {
378 count = ret;
James Morrisc43e2592008-01-12 22:05:48 +1100379 if (file->f_op->read)
380 ret = file->f_op->read(file, buf, count, pos);
381 else
382 ret = do_sync_read(file, buf, count, pos);
383 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500384 fsnotify_access(file);
James Morrisc43e2592008-01-12 22:05:48 +1100385 add_rchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 }
James Morrisc43e2592008-01-12 22:05:48 +1100387 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 }
389
390 return ret;
391}
392
393EXPORT_SYMBOL(vfs_read);
394
395ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
396{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700397 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 struct kiocb kiocb;
399 ssize_t ret;
400
401 init_sync_kiocb(&kiocb, filp);
402 kiocb.ki_pos = *ppos;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700403 kiocb.ki_left = len;
David Howells61964eb2010-03-24 17:09:19 +0000404 kiocb.ki_nbytes = len;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700405
406 for (;;) {
407 ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
408 if (ret != -EIOCBRETRY)
409 break;
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700410 wait_on_retry_sync_kiocb(&kiocb);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700411 }
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700412
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 if (-EIOCBQUEUED == ret)
414 ret = wait_on_sync_kiocb(&kiocb);
415 *ppos = kiocb.ki_pos;
416 return ret;
417}
418
419EXPORT_SYMBOL(do_sync_write);
420
Al Viro06ae43f2013-03-20 13:19:30 -0400421ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
422{
423 mm_segment_t old_fs;
424 const char __user *p;
425 ssize_t ret;
426
Al Viro3e84f482013-03-27 15:20:30 +0000427 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
428 return -EINVAL;
429
Al Viro06ae43f2013-03-20 13:19:30 -0400430 old_fs = get_fs();
431 set_fs(get_ds());
432 p = (__force const char __user *)buf;
433 if (count > MAX_RW_COUNT)
434 count = MAX_RW_COUNT;
435 if (file->f_op->write)
436 ret = file->f_op->write(file, p, count, pos);
437 else
438 ret = do_sync_write(file, p, count, pos);
439 set_fs(old_fs);
440 if (ret > 0) {
441 fsnotify_modify(file);
442 add_wchar(current, ret);
443 }
444 inc_syscw(current);
445 return ret;
446}
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
449{
450 ssize_t ret;
451
452 if (!(file->f_mode & FMODE_WRITE))
453 return -EBADF;
454 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
455 return -EINVAL;
456 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
457 return -EFAULT;
458
459 ret = rw_verify_area(WRITE, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800460 if (ret >= 0) {
461 count = ret;
James Morrisc43e2592008-01-12 22:05:48 +1100462 if (file->f_op->write)
463 ret = file->f_op->write(file, buf, count, pos);
464 else
465 ret = do_sync_write(file, buf, count, pos);
466 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500467 fsnotify_modify(file);
James Morrisc43e2592008-01-12 22:05:48 +1100468 add_wchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 }
James Morrisc43e2592008-01-12 22:05:48 +1100470 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 }
472
473 return ret;
474}
475
476EXPORT_SYMBOL(vfs_write);
477
478static inline loff_t file_pos_read(struct file *file)
479{
480 return file->f_pos;
481}
482
483static inline void file_pos_write(struct file *file, loff_t pos)
484{
485 file->f_pos = pos;
486}
487
Heiko Carstens3cdad422009-01-14 14:14:22 +0100488SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
Al Viro2903ff02012-08-28 12:52:22 -0400490 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
Al Viro2903ff02012-08-28 12:52:22 -0400493 if (f.file) {
494 loff_t pos = file_pos_read(f.file);
495 ret = vfs_read(f.file, buf, count, &pos);
496 file_pos_write(f.file, pos);
497 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 return ret;
500}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Heiko Carstens3cdad422009-01-14 14:14:22 +0100502SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
503 size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
Al Viro2903ff02012-08-28 12:52:22 -0400505 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Al Viro2903ff02012-08-28 12:52:22 -0400508 if (f.file) {
509 loff_t pos = file_pos_read(f.file);
510 ret = vfs_write(f.file, buf, count, &pos);
511 file_pos_write(f.file, pos);
512 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 }
514
515 return ret;
516}
517
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100518SYSCALL_DEFINE(pread64)(unsigned int fd, char __user *buf,
519 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520{
Al Viro2903ff02012-08-28 12:52:22 -0400521 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
524 if (pos < 0)
525 return -EINVAL;
526
Al Viro2903ff02012-08-28 12:52:22 -0400527 f = fdget(fd);
528 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400530 if (f.file->f_mode & FMODE_PREAD)
531 ret = vfs_read(f.file, buf, count, &pos);
532 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 }
534
535 return ret;
536}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100537#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
538asmlinkage long SyS_pread64(long fd, long buf, long count, loff_t pos)
539{
540 return SYSC_pread64((unsigned int) fd, (char __user *) buf,
541 (size_t) count, pos);
542}
543SYSCALL_ALIAS(sys_pread64, SyS_pread64);
544#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100546SYSCALL_DEFINE(pwrite64)(unsigned int fd, const char __user *buf,
547 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548{
Al Viro2903ff02012-08-28 12:52:22 -0400549 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
552 if (pos < 0)
553 return -EINVAL;
554
Al Viro2903ff02012-08-28 12:52:22 -0400555 f = fdget(fd);
556 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400558 if (f.file->f_mode & FMODE_PWRITE)
559 ret = vfs_write(f.file, buf, count, &pos);
560 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 }
562
563 return ret;
564}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100565#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
566asmlinkage long SyS_pwrite64(long fd, long buf, long count, loff_t pos)
567{
568 return SYSC_pwrite64((unsigned int) fd, (const char __user *) buf,
569 (size_t) count, pos);
570}
571SYSCALL_ALIAS(sys_pwrite64, SyS_pwrite64);
572#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
574/*
575 * Reduce an iovec's length in-place. Return the resulting number of segments
576 */
577unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
578{
579 unsigned long seg = 0;
580 size_t len = 0;
581
582 while (seg < nr_segs) {
583 seg++;
584 if (len + iov->iov_len >= to) {
585 iov->iov_len = to - len;
586 break;
587 }
588 len += iov->iov_len;
589 iov++;
590 }
591 return seg;
592}
Eric Sandeen19295522008-01-28 23:58:27 -0500593EXPORT_SYMBOL(iov_shorten);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700595ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
596 unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
597{
598 struct kiocb kiocb;
599 ssize_t ret;
600
601 init_sync_kiocb(&kiocb, filp);
602 kiocb.ki_pos = *ppos;
603 kiocb.ki_left = len;
604 kiocb.ki_nbytes = len;
605
606 for (;;) {
607 ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
608 if (ret != -EIOCBRETRY)
609 break;
610 wait_on_retry_sync_kiocb(&kiocb);
611 }
612
613 if (ret == -EIOCBQUEUED)
614 ret = wait_on_sync_kiocb(&kiocb);
615 *ppos = kiocb.ki_pos;
616 return ret;
617}
618
619/* Do it by hand, with file-ops */
620ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
621 unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
622{
623 struct iovec *vector = iov;
624 ssize_t ret = 0;
625
626 while (nr_segs > 0) {
627 void __user *base;
628 size_t len;
629 ssize_t nr;
630
631 base = vector->iov_base;
632 len = vector->iov_len;
633 vector++;
634 nr_segs--;
635
636 nr = fn(filp, base, len, ppos);
637
638 if (nr < 0) {
639 if (!ret)
640 ret = nr;
641 break;
642 }
643 ret += nr;
644 if (nr != len)
645 break;
646 }
647
648 return ret;
649}
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651/* A write operation does a read from user space and vice versa */
652#define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
653
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700654ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
655 unsigned long nr_segs, unsigned long fast_segs,
656 struct iovec *fast_pointer,
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700657 struct iovec **ret_pointer)
Linus Torvalds435f49a2010-10-29 10:36:49 -0700658{
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700659 unsigned long seg;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700660 ssize_t ret;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700661 struct iovec *iov = fast_pointer;
662
Linus Torvalds435f49a2010-10-29 10:36:49 -0700663 /*
664 * SuS says "The readv() function *may* fail if the iovcnt argument
665 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
666 * traditionally returned zero for zero segments, so...
667 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700668 if (nr_segs == 0) {
669 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700670 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700671 }
672
Linus Torvalds435f49a2010-10-29 10:36:49 -0700673 /*
674 * First get the "struct iovec" from user memory and
675 * verify all the pointers
676 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700677 if (nr_segs > UIO_MAXIOV) {
678 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700679 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700680 }
681 if (nr_segs > fast_segs) {
Linus Torvalds435f49a2010-10-29 10:36:49 -0700682 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700683 if (iov == NULL) {
684 ret = -ENOMEM;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700685 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700686 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700687 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700688 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
689 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700690 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700691 }
692
Linus Torvalds435f49a2010-10-29 10:36:49 -0700693 /*
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700694 * According to the Single Unix Specification we should return EINVAL
695 * if an element length is < 0 when cast to ssize_t or if the
696 * total length would overflow the ssize_t return value of the
697 * system call.
Linus Torvalds435f49a2010-10-29 10:36:49 -0700698 *
699 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
700 * overflow case.
701 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700702 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700703 for (seg = 0; seg < nr_segs; seg++) {
704 void __user *buf = iov[seg].iov_base;
705 ssize_t len = (ssize_t)iov[seg].iov_len;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700706
707 /* see if we we're about to use an invalid len or if
708 * it's about to overflow ssize_t */
Linus Torvalds435f49a2010-10-29 10:36:49 -0700709 if (len < 0) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700710 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700711 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700712 }
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700713 if (type >= 0
Christopher Yeohfcf63402011-10-31 17:06:39 -0700714 && unlikely(!access_ok(vrfy_dir(type), buf, len))) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700715 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700716 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700717 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700718 if (len > MAX_RW_COUNT - ret) {
719 len = MAX_RW_COUNT - ret;
720 iov[seg].iov_len = len;
721 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700722 ret += len;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700723 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700724out:
725 *ret_pointer = iov;
726 return ret;
727}
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729static ssize_t do_readv_writev(int type, struct file *file,
730 const struct iovec __user * uvector,
731 unsigned long nr_segs, loff_t *pos)
732{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 size_t tot_len;
734 struct iovec iovstack[UIO_FASTIOV];
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700735 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 io_fn_t fn;
738 iov_fn_t fnv;
739
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700740 if (!file->f_op) {
741 ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 goto out;
743 }
744
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700745 ret = rw_copy_check_uvector(type, uvector, nr_segs,
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700746 ARRAY_SIZE(iovstack), iovstack, &iov);
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700747 if (ret <= 0)
748 goto out;
749
750 tot_len = ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 ret = rw_verify_area(type, file, pos, tot_len);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800752 if (ret < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 goto out;
754
755 fnv = NULL;
756 if (type == READ) {
757 fn = file->f_op->read;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700758 fnv = file->f_op->aio_read;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 } else {
760 fn = (io_fn_t)file->f_op->write;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700761 fnv = file->f_op->aio_write;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 }
763
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700764 if (fnv)
765 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
766 pos, fnv);
767 else
768 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770out:
771 if (iov != iovstack)
772 kfree(iov);
Robert Love0eeca282005-07-12 17:06:03 -0400773 if ((ret + (type == READ)) > 0) {
774 if (type == READ)
Eric Paris2a12a9d2009-12-17 21:24:21 -0500775 fsnotify_access(file);
Robert Love0eeca282005-07-12 17:06:03 -0400776 else
Eric Paris2a12a9d2009-12-17 21:24:21 -0500777 fsnotify_modify(file);
Robert Love0eeca282005-07-12 17:06:03 -0400778 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780}
781
782ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
783 unsigned long vlen, loff_t *pos)
784{
785 if (!(file->f_mode & FMODE_READ))
786 return -EBADF;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700787 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 return -EINVAL;
789
790 return do_readv_writev(READ, file, vec, vlen, pos);
791}
792
793EXPORT_SYMBOL(vfs_readv);
794
795ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
796 unsigned long vlen, loff_t *pos)
797{
798 if (!(file->f_mode & FMODE_WRITE))
799 return -EBADF;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700800 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 return -EINVAL;
802
803 return do_readv_writev(WRITE, file, vec, vlen, pos);
804}
805
806EXPORT_SYMBOL(vfs_writev);
807
Heiko Carstens3cdad422009-01-14 14:14:22 +0100808SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
809 unsigned long, vlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Al Viro2903ff02012-08-28 12:52:22 -0400811 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Al Viro2903ff02012-08-28 12:52:22 -0400814 if (f.file) {
815 loff_t pos = file_pos_read(f.file);
816 ret = vfs_readv(f.file, vec, vlen, &pos);
817 file_pos_write(f.file, pos);
818 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 }
820
821 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800822 add_rchar(current, ret);
823 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 return ret;
825}
826
Heiko Carstens3cdad422009-01-14 14:14:22 +0100827SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
828 unsigned long, vlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829{
Al Viro2903ff02012-08-28 12:52:22 -0400830 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Al Viro2903ff02012-08-28 12:52:22 -0400833 if (f.file) {
834 loff_t pos = file_pos_read(f.file);
835 ret = vfs_writev(f.file, vec, vlen, &pos);
836 file_pos_write(f.file, pos);
837 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 }
839
840 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800841 add_wchar(current, ret);
842 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 return ret;
844}
845
Linus Torvalds601cc112009-04-03 08:03:22 -0700846static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700847{
Linus Torvalds601cc112009-04-03 08:03:22 -0700848#define HALF_LONG_BITS (BITS_PER_LONG / 2)
849 return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
850}
851
852SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
853 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
854{
855 loff_t pos = pos_from_hilo(pos_h, pos_l);
Al Viro2903ff02012-08-28 12:52:22 -0400856 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700857 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700858
859 if (pos < 0)
860 return -EINVAL;
861
Al Viro2903ff02012-08-28 12:52:22 -0400862 f = fdget(fd);
863 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700864 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400865 if (f.file->f_mode & FMODE_PREAD)
866 ret = vfs_readv(f.file, vec, vlen, &pos);
867 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700868 }
869
870 if (ret > 0)
871 add_rchar(current, ret);
872 inc_syscr(current);
873 return ret;
874}
875
876SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
Linus Torvalds601cc112009-04-03 08:03:22 -0700877 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700878{
Linus Torvalds601cc112009-04-03 08:03:22 -0700879 loff_t pos = pos_from_hilo(pos_h, pos_l);
Al Viro2903ff02012-08-28 12:52:22 -0400880 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700881 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700882
883 if (pos < 0)
884 return -EINVAL;
885
Al Viro2903ff02012-08-28 12:52:22 -0400886 f = fdget(fd);
887 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700888 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400889 if (f.file->f_mode & FMODE_PWRITE)
890 ret = vfs_writev(f.file, vec, vlen, &pos);
891 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700892 }
893
894 if (ret > 0)
895 add_wchar(current, ret);
896 inc_syscw(current);
897 return ret;
898}
899
Catalin Marinas8f9c0112012-09-19 12:01:52 +0100900ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos, size_t count,
901 loff_t max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902{
Al Viro2903ff02012-08-28 12:52:22 -0400903 struct fd in, out;
904 struct inode *in_inode, *out_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 loff_t pos;
906 ssize_t retval;
Al Viro2903ff02012-08-28 12:52:22 -0400907 int fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908
909 /*
910 * Get input file, and verify that it is ok..
911 */
912 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400913 in = fdget(in_fd);
914 if (!in.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 goto out;
Al Viro2903ff02012-08-28 12:52:22 -0400916 if (!(in.file->f_mode & FMODE_READ))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 goto fput_in;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 retval = -ESPIPE;
919 if (!ppos)
Al Viro2903ff02012-08-28 12:52:22 -0400920 ppos = &in.file->f_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 else
Al Viro2903ff02012-08-28 12:52:22 -0400922 if (!(in.file->f_mode & FMODE_PREAD))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 goto fput_in;
Al Viro2903ff02012-08-28 12:52:22 -0400924 retval = rw_verify_area(READ, in.file, ppos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800925 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 goto fput_in;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800927 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 /*
930 * Get output file, and verify that it is ok..
931 */
932 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400933 out = fdget(out_fd);
934 if (!out.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 goto fput_in;
Al Viro2903ff02012-08-28 12:52:22 -0400936 if (!(out.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 goto fput_out;
938 retval = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -0500939 in_inode = file_inode(in.file);
940 out_inode = file_inode(out.file);
Al Viro2903ff02012-08-28 12:52:22 -0400941 retval = rw_verify_area(WRITE, out.file, &out.file->f_pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800942 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 goto fput_out;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800944 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 if (!max)
947 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
948
949 pos = *ppos;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 if (unlikely(pos + count > max)) {
951 retval = -EOVERFLOW;
952 if (pos >= max)
953 goto fput_out;
954 count = max - pos;
955 }
956
Jens Axboed96e6e72007-06-11 12:18:52 +0200957 fl = 0;
Jens Axboe534f2aa2007-06-01 14:52:37 +0200958#if 0
Jens Axboed96e6e72007-06-11 12:18:52 +0200959 /*
960 * We need to debate whether we can enable this or not. The
961 * man page documents EAGAIN return for the output at least,
962 * and the application is arguably buggy if it doesn't expect
963 * EAGAIN on a non-blocking file descriptor.
964 */
Al Viro2903ff02012-08-28 12:52:22 -0400965 if (in.file->f_flags & O_NONBLOCK)
Jens Axboed96e6e72007-06-11 12:18:52 +0200966 fl = SPLICE_F_NONBLOCK;
Jens Axboe534f2aa2007-06-01 14:52:37 +0200967#endif
Al Viro2903ff02012-08-28 12:52:22 -0400968 retval = do_splice_direct(in.file, ppos, out.file, count, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
970 if (retval > 0) {
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800971 add_rchar(current, retval);
972 add_wchar(current, retval);
Scott Wolchoka68c2f12012-12-20 15:05:52 -0800973 fsnotify_access(in.file);
974 fsnotify_modify(out.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800977 inc_syscr(current);
978 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 if (*ppos > max)
980 retval = -EOVERFLOW;
981
982fput_out:
Al Viro2903ff02012-08-28 12:52:22 -0400983 fdput(out);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984fput_in:
Al Viro2903ff02012-08-28 12:52:22 -0400985 fdput(in);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986out:
987 return retval;
988}
989
Heiko Carstens002c8972009-01-14 14:14:18 +0100990SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991{
992 loff_t pos;
993 off_t off;
994 ssize_t ret;
995
996 if (offset) {
997 if (unlikely(get_user(off, offset)))
998 return -EFAULT;
999 pos = off;
1000 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1001 if (unlikely(put_user(pos, offset)))
1002 return -EFAULT;
1003 return ret;
1004 }
1005
1006 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1007}
1008
Heiko Carstens002c8972009-01-14 14:14:18 +01001009SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010{
1011 loff_t pos;
1012 ssize_t ret;
1013
1014 if (offset) {
1015 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1016 return -EFAULT;
1017 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1018 if (unlikely(put_user(pos, offset)))
1019 return -EFAULT;
1020 return ret;
1021 }
1022
1023 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1024}