blob: 31c6efa431839e41f4b39b314cb640dd89b68ecc [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>
Kent Overstreeta27bb332013-05-07 16:19:08 -070012#include <linux/aio.h>
Robert Love0eeca282005-07-12 17:06:03 -040013#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/security.h>
Paul Gortmaker630d9c42011-11-16 23:57:37 -050015#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/syscalls.h>
Linus Torvaldse28cc712006-01-04 16:20:40 -080017#include <linux/pagemap.h>
Jens Axboed6b29d72007-06-04 09:59:47 +020018#include <linux/splice.h>
Al Viro561c6732013-02-24 10:52:26 -050019#include <linux/compat.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
Al Viroc0bd14af2013-05-04 15:00:54 -040025typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
26typedef ssize_t (*iov_fn_t)(struct kiocb *, const struct iovec *,
27 unsigned long, loff_t);
28
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080029const struct file_operations generic_ro_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070030 .llseek = generic_file_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -070031 .read = do_sync_read,
32 .aio_read = generic_file_aio_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070033 .mmap = generic_file_readonly_mmap,
Jens Axboe534f2aa2007-06-01 14:52:37 +020034 .splice_read = generic_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070035};
36
37EXPORT_SYMBOL(generic_ro_fops);
38
Al Virocccb5a12010-12-17 07:44:05 -050039static inline int unsigned_offsets(struct file *file)
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070040{
Al Virocccb5a12010-12-17 07:44:05 -050041 return file->f_mode & FMODE_UNSIGNED_OFFSET;
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070042}
43
Jie Liu46a1c2c2013-06-25 12:02:13 +080044/**
45 * vfs_setpos - update the file offset for lseek
46 * @file: file structure in question
47 * @offset: file offset to seek to
48 * @maxsize: maximum file size
49 *
50 * This is a low-level filesystem helper for updating the file offset to
51 * the value specified by @offset if the given offset is valid and it is
52 * not equal to the current file offset.
53 *
54 * Return the specified offset on success and -EINVAL on invalid offset.
55 */
56loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
Andi Kleenef3d0fd2011-09-15 16:06:48 -070057{
58 if (offset < 0 && !unsigned_offsets(file))
59 return -EINVAL;
60 if (offset > maxsize)
61 return -EINVAL;
62
63 if (offset != file->f_pos) {
64 file->f_pos = offset;
65 file->f_version = 0;
66 }
67 return offset;
68}
Jie Liu46a1c2c2013-06-25 12:02:13 +080069EXPORT_SYMBOL(vfs_setpos);
Andi Kleenef3d0fd2011-09-15 16:06:48 -070070
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020071/**
Andi Kleen57604952011-09-15 16:06:50 -070072 * generic_file_llseek_size - generic llseek implementation for regular files
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020073 * @file: file structure to seek on
74 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -080075 * @whence: type of seek
Eric Sandeene8b96eb2012-04-30 13:11:29 -050076 * @size: max size of this file in file system
77 * @eof: offset used for SEEK_END position
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020078 *
Andi Kleen57604952011-09-15 16:06:50 -070079 * This is a variant of generic_file_llseek that allows passing in a custom
Eric Sandeene8b96eb2012-04-30 13:11:29 -050080 * maximum file size and a custom EOF position, for e.g. hashed directories
Andi Kleenef3d0fd2011-09-15 16:06:48 -070081 *
82 * Synchronization:
Andi Kleen57604952011-09-15 16:06:50 -070083 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
Andi Kleenef3d0fd2011-09-15 16:06:48 -070084 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
85 * read/writes behave like SEEK_SET against seeks.
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020086 */
Andi Kleen9465efc2008-06-27 11:05:24 +020087loff_t
Andrew Morton965c8e52012-12-17 15:59:39 -080088generic_file_llseek_size(struct file *file, loff_t offset, int whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -050089 loff_t maxsize, loff_t eof)
Linus Torvalds1da177e2005-04-16 15:20:36 -070090{
Andrew Morton965c8e52012-12-17 15:59:39 -080091 switch (whence) {
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020092 case SEEK_END:
Eric Sandeene8b96eb2012-04-30 13:11:29 -050093 offset += eof;
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020094 break;
95 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -080096 /*
97 * Here we special-case the lseek(fd, 0, SEEK_CUR)
98 * position-querying operation. Avoid rewriting the "same"
99 * f_pos value back to the file because a concurrent read(),
100 * write() or lseek() might have altered it
101 */
102 if (offset == 0)
103 return file->f_pos;
Andi Kleenef3d0fd2011-09-15 16:06:48 -0700104 /*
105 * f_lock protects against read/modify/write race with other
106 * SEEK_CURs. Note that parallel writes and reads behave
107 * like SEEK_SET.
108 */
109 spin_lock(&file->f_lock);
Jie Liu46a1c2c2013-06-25 12:02:13 +0800110 offset = vfs_setpos(file, file->f_pos + offset, maxsize);
Andi Kleenef3d0fd2011-09-15 16:06:48 -0700111 spin_unlock(&file->f_lock);
112 return offset;
Josef Bacik982d8162011-07-18 13:21:35 -0400113 case SEEK_DATA:
114 /*
115 * In the generic case the entire file is data, so as long as
116 * offset isn't at the end of the file then the offset is data.
117 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500118 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400119 return -ENXIO;
120 break;
121 case SEEK_HOLE:
122 /*
123 * There is a virtual hole at the end of the file, so as long as
124 * offset isn't i_size or larger, return i_size.
125 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500126 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400127 return -ENXIO;
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500128 offset = eof;
Josef Bacik982d8162011-07-18 13:21:35 -0400129 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 }
Christoph Hellwig3a8cff42008-08-11 15:37:17 +0200131
Jie Liu46a1c2c2013-06-25 12:02:13 +0800132 return vfs_setpos(file, offset, maxsize);
Andi Kleen57604952011-09-15 16:06:50 -0700133}
134EXPORT_SYMBOL(generic_file_llseek_size);
135
136/**
137 * generic_file_llseek - generic llseek implementation for regular files
138 * @file: file structure to seek on
139 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800140 * @whence: type of seek
Andi Kleen57604952011-09-15 16:06:50 -0700141 *
142 * This is a generic implemenation of ->llseek useable for all normal local
143 * filesystems. It just updates the file offset to the value specified by
Ming Lei546ae2d2013-04-29 15:06:07 -0700144 * @offset and @whence.
Andi Kleen57604952011-09-15 16:06:50 -0700145 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800146loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
Andi Kleen57604952011-09-15 16:06:50 -0700147{
148 struct inode *inode = file->f_mapping->host;
149
Andrew Morton965c8e52012-12-17 15:59:39 -0800150 return generic_file_llseek_size(file, offset, whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500151 inode->i_sb->s_maxbytes,
152 i_size_read(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153}
Andi Kleen9465efc2008-06-27 11:05:24 +0200154EXPORT_SYMBOL(generic_file_llseek);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
jan Blunckae6afc32010-05-26 14:44:48 -0700156/**
Al Viro1bf9d142013-06-16 20:27:42 +0400157 * fixed_size_llseek - llseek implementation for fixed-sized devices
158 * @file: file structure to seek on
159 * @offset: file offset to seek to
160 * @whence: type of seek
161 * @size: size of the file
162 *
163 */
164loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size)
165{
166 switch (whence) {
167 case SEEK_SET: case SEEK_CUR: case SEEK_END:
168 return generic_file_llseek_size(file, offset, whence,
169 size, size);
170 default:
171 return -EINVAL;
172 }
173}
174EXPORT_SYMBOL(fixed_size_llseek);
175
176/**
jan Blunckae6afc32010-05-26 14:44:48 -0700177 * noop_llseek - No Operation Performed llseek implementation
178 * @file: file structure to seek on
179 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800180 * @whence: type of seek
jan Blunckae6afc32010-05-26 14:44:48 -0700181 *
182 * This is an implementation of ->llseek useable for the rare special case when
183 * userspace expects the seek to succeed but the (device) file is actually not
184 * able to perform the seek. In this case you use noop_llseek() instead of
185 * falling back to the default implementation of ->llseek.
186 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800187loff_t noop_llseek(struct file *file, loff_t offset, int whence)
jan Blunckae6afc32010-05-26 14:44:48 -0700188{
189 return file->f_pos;
190}
191EXPORT_SYMBOL(noop_llseek);
192
Andrew Morton965c8e52012-12-17 15:59:39 -0800193loff_t no_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194{
195 return -ESPIPE;
196}
197EXPORT_SYMBOL(no_llseek);
198
Andrew Morton965c8e52012-12-17 15:59:39 -0800199loff_t default_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200{
Al Viro496ad9a2013-01-23 17:07:38 -0500201 struct inode *inode = file_inode(file);
David Sterba16abef02008-04-22 15:09:22 +0200202 loff_t retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
Josef Bacik982d8162011-07-18 13:21:35 -0400204 mutex_lock(&inode->i_mutex);
Andrew Morton965c8e52012-12-17 15:59:39 -0800205 switch (whence) {
Chris Snook7b8e8922007-05-08 00:24:13 -0700206 case SEEK_END:
Josef Bacik982d8162011-07-18 13:21:35 -0400207 offset += i_size_read(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 break;
Chris Snook7b8e8922007-05-08 00:24:13 -0700209 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800210 if (offset == 0) {
211 retval = file->f_pos;
212 goto out;
213 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 offset += file->f_pos;
Josef Bacik982d8162011-07-18 13:21:35 -0400215 break;
216 case SEEK_DATA:
217 /*
218 * In the generic case the entire file is data, so as
219 * long as offset isn't at the end of the file then the
220 * offset is data.
221 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300222 if (offset >= inode->i_size) {
223 retval = -ENXIO;
224 goto out;
225 }
Josef Bacik982d8162011-07-18 13:21:35 -0400226 break;
227 case SEEK_HOLE:
228 /*
229 * There is a virtual hole at the end of the file, so
230 * as long as offset isn't i_size or larger, return
231 * i_size.
232 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300233 if (offset >= inode->i_size) {
234 retval = -ENXIO;
235 goto out;
236 }
Josef Bacik982d8162011-07-18 13:21:35 -0400237 offset = inode->i_size;
238 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 }
240 retval = -EINVAL;
Al Virocccb5a12010-12-17 07:44:05 -0500241 if (offset >= 0 || unsigned_offsets(file)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 if (offset != file->f_pos) {
243 file->f_pos = offset;
244 file->f_version = 0;
245 }
246 retval = offset;
247 }
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800248out:
Josef Bacik982d8162011-07-18 13:21:35 -0400249 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 return retval;
251}
252EXPORT_SYMBOL(default_llseek);
253
Andrew Morton965c8e52012-12-17 15:59:39 -0800254loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255{
256 loff_t (*fn)(struct file *, loff_t, int);
257
258 fn = no_llseek;
259 if (file->f_mode & FMODE_LSEEK) {
Al Viro72c2d532013-09-22 16:27:52 -0400260 if (file->f_op->llseek)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 fn = file->f_op->llseek;
262 }
Andrew Morton965c8e52012-12-17 15:59:39 -0800263 return fn(file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264}
265EXPORT_SYMBOL(vfs_llseek);
266
Linus Torvalds9c225f22014-03-03 09:36:58 -0800267static inline struct fd fdget_pos(int fd)
268{
Al Virobd2a31d2014-03-04 14:54:22 -0500269 return __to_fd(__fdget_pos(fd));
Linus Torvalds9c225f22014-03-03 09:36:58 -0800270}
271
272static inline void fdput_pos(struct fd f)
273{
274 if (f.flags & FDPUT_POS_UNLOCK)
275 mutex_unlock(&f.file->f_pos_lock);
276 fdput(f);
277}
278
Andrew Morton965c8e52012-12-17 15:59:39 -0800279SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280{
281 off_t retval;
Linus Torvalds9c225f22014-03-03 09:36:58 -0800282 struct fd f = fdget_pos(fd);
Al Viro2903ff02012-08-28 12:52:22 -0400283 if (!f.file)
284 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
286 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800287 if (whence <= SEEK_MAX) {
288 loff_t res = vfs_llseek(f.file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 retval = res;
290 if (res != (loff_t)retval)
291 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
292 }
Linus Torvalds9c225f22014-03-03 09:36:58 -0800293 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 return retval;
295}
296
Al Viro561c6732013-02-24 10:52:26 -0500297#ifdef CONFIG_COMPAT
298COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
299{
300 return sys_lseek(fd, offset, whence);
301}
302#endif
303
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304#ifdef __ARCH_WANT_SYS_LLSEEK
Heiko Carstens003d7ab2009-01-14 14:14:21 +0100305SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
306 unsigned long, offset_low, loff_t __user *, result,
Andrew Morton965c8e52012-12-17 15:59:39 -0800307 unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308{
309 int retval;
Eric Biggersd7a15f82014-03-16 14:24:08 -0500310 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 loff_t offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Al Viro2903ff02012-08-28 12:52:22 -0400313 if (!f.file)
314 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800317 if (whence > SEEK_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 goto out_putf;
319
Al Viro2903ff02012-08-28 12:52:22 -0400320 offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
Andrew Morton965c8e52012-12-17 15:59:39 -0800321 whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
323 retval = (int)offset;
324 if (offset >= 0) {
325 retval = -EFAULT;
326 if (!copy_to_user(result, &offset, sizeof(offset)))
327 retval = 0;
328 }
329out_putf:
Eric Biggersd7a15f82014-03-16 14:24:08 -0500330 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 return retval;
332}
333#endif
334
Linus Torvaldse28cc712006-01-04 16:20:40 -0800335/*
336 * rw_verify_area doesn't like huge counts. We limit
337 * them to something that fits in "int" so that others
338 * won't have to do range checks all the time.
339 */
Al Viro68d70d02013-06-19 15:26:04 +0400340int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341{
342 struct inode *inode;
343 loff_t pos;
James Morrisc43e2592008-01-12 22:05:48 +1100344 int retval = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
Al Viro496ad9a2013-01-23 17:07:38 -0500346 inode = file_inode(file);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800347 if (unlikely((ssize_t) count < 0))
James Morrisc43e2592008-01-12 22:05:48 +1100348 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 pos = *ppos;
Al Virocccb5a12010-12-17 07:44:05 -0500350 if (unlikely(pos < 0)) {
351 if (!unsigned_offsets(file))
352 return retval;
353 if (count >= -pos) /* both values are in 0..LLONG_MAX */
354 return -EOVERFLOW;
355 } else if (unlikely((loff_t) (pos + count) < 0)) {
356 if (!unsigned_offsets(file))
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700357 return retval;
358 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
Pavel Emelyanova16877c2007-10-01 14:41:11 -0700360 if (unlikely(inode->i_flock && mandatory_lock(inode))) {
James Morrisc43e2592008-01-12 22:05:48 +1100361 retval = locks_mandatory_area(
Linus Torvaldse28cc712006-01-04 16:20:40 -0800362 read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
363 inode, file, pos, count);
364 if (retval < 0)
365 return retval;
366 }
James Morrisc43e2592008-01-12 22:05:48 +1100367 retval = security_file_permission(file,
368 read_write == READ ? MAY_READ : MAY_WRITE);
369 if (retval)
370 return retval;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800371 return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
374ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
375{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700376 struct iovec iov = { .iov_base = buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 struct kiocb kiocb;
378 ssize_t ret;
379
380 init_sync_kiocb(&kiocb, filp);
381 kiocb.ki_pos = *ppos;
David Howells61964eb2010-03-24 17:09:19 +0000382 kiocb.ki_nbytes = len;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700383
Zach Brown41003a72013-05-07 16:18:25 -0700384 ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 if (-EIOCBQUEUED == ret)
386 ret = wait_on_sync_kiocb(&kiocb);
387 *ppos = kiocb.ki_pos;
388 return ret;
389}
390
391EXPORT_SYMBOL(do_sync_read);
392
393ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
394{
395 ssize_t ret;
396
397 if (!(file->f_mode & FMODE_READ))
398 return -EBADF;
Al Viro72c2d532013-09-22 16:27:52 -0400399 if (!file->f_op->read && !file->f_op->aio_read)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 return -EINVAL;
401 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
402 return -EFAULT;
403
404 ret = rw_verify_area(READ, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800405 if (ret >= 0) {
406 count = ret;
James Morrisc43e2592008-01-12 22:05:48 +1100407 if (file->f_op->read)
408 ret = file->f_op->read(file, buf, count, pos);
409 else
410 ret = do_sync_read(file, buf, count, pos);
411 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500412 fsnotify_access(file);
James Morrisc43e2592008-01-12 22:05:48 +1100413 add_rchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 }
James Morrisc43e2592008-01-12 22:05:48 +1100415 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 }
417
418 return ret;
419}
420
421EXPORT_SYMBOL(vfs_read);
422
423ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
424{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700425 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 struct kiocb kiocb;
427 ssize_t ret;
428
429 init_sync_kiocb(&kiocb, filp);
430 kiocb.ki_pos = *ppos;
David Howells61964eb2010-03-24 17:09:19 +0000431 kiocb.ki_nbytes = len;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700432
Zach Brown41003a72013-05-07 16:18:25 -0700433 ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (-EIOCBQUEUED == ret)
435 ret = wait_on_sync_kiocb(&kiocb);
436 *ppos = kiocb.ki_pos;
437 return ret;
438}
439
440EXPORT_SYMBOL(do_sync_write);
441
Al Viro06ae43f2013-03-20 13:19:30 -0400442ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
443{
444 mm_segment_t old_fs;
445 const char __user *p;
446 ssize_t ret;
447
Al Viro72c2d532013-09-22 16:27:52 -0400448 if (!file->f_op->write && !file->f_op->aio_write)
Al Viro3e84f482013-03-27 15:20:30 +0000449 return -EINVAL;
450
Al Viro06ae43f2013-03-20 13:19:30 -0400451 old_fs = get_fs();
452 set_fs(get_ds());
453 p = (__force const char __user *)buf;
454 if (count > MAX_RW_COUNT)
455 count = MAX_RW_COUNT;
456 if (file->f_op->write)
457 ret = file->f_op->write(file, p, count, pos);
458 else
459 ret = do_sync_write(file, p, count, pos);
460 set_fs(old_fs);
461 if (ret > 0) {
462 fsnotify_modify(file);
463 add_wchar(current, ret);
464 }
465 inc_syscw(current);
466 return ret;
467}
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
470{
471 ssize_t ret;
472
473 if (!(file->f_mode & FMODE_WRITE))
474 return -EBADF;
Al Viro72c2d532013-09-22 16:27:52 -0400475 if (!file->f_op->write && !file->f_op->aio_write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 return -EINVAL;
477 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
478 return -EFAULT;
479
480 ret = rw_verify_area(WRITE, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800481 if (ret >= 0) {
482 count = ret;
Al Viro03d95eb2013-03-20 13:04:20 -0400483 file_start_write(file);
James Morrisc43e2592008-01-12 22:05:48 +1100484 if (file->f_op->write)
485 ret = file->f_op->write(file, buf, count, pos);
486 else
487 ret = do_sync_write(file, buf, count, pos);
488 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500489 fsnotify_modify(file);
James Morrisc43e2592008-01-12 22:05:48 +1100490 add_wchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 }
James Morrisc43e2592008-01-12 22:05:48 +1100492 inc_syscw(current);
Al Viro03d95eb2013-03-20 13:04:20 -0400493 file_end_write(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 }
495
496 return ret;
497}
498
499EXPORT_SYMBOL(vfs_write);
500
501static inline loff_t file_pos_read(struct file *file)
502{
503 return file->f_pos;
504}
505
506static inline void file_pos_write(struct file *file, loff_t pos)
507{
508 file->f_pos = pos;
509}
510
Heiko Carstens3cdad422009-01-14 14:14:22 +0100511SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800513 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Al Viro2903ff02012-08-28 12:52:22 -0400516 if (f.file) {
517 loff_t pos = file_pos_read(f.file);
518 ret = vfs_read(f.file, buf, count, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400519 if (ret >= 0)
520 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800521 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 return ret;
524}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
Heiko Carstens3cdad422009-01-14 14:14:22 +0100526SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
527 size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800529 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Al Viro2903ff02012-08-28 12:52:22 -0400532 if (f.file) {
533 loff_t pos = file_pos_read(f.file);
534 ret = vfs_write(f.file, buf, count, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400535 if (ret >= 0)
536 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800537 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 }
539
540 return ret;
541}
542
Al Viro4a0fd5b2013-01-21 15:16:58 -0500543SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
544 size_t, count, loff_t, pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
Al Viro2903ff02012-08-28 12:52:22 -0400546 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548
549 if (pos < 0)
550 return -EINVAL;
551
Al Viro2903ff02012-08-28 12:52:22 -0400552 f = fdget(fd);
553 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400555 if (f.file->f_mode & FMODE_PREAD)
556 ret = vfs_read(f.file, buf, count, &pos);
557 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
559
560 return ret;
561}
562
Al Viro4a0fd5b2013-01-21 15:16:58 -0500563SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
564 size_t, count, loff_t, pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565{
Al Viro2903ff02012-08-28 12:52:22 -0400566 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
569 if (pos < 0)
570 return -EINVAL;
571
Al Viro2903ff02012-08-28 12:52:22 -0400572 f = fdget(fd);
573 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400575 if (f.file->f_mode & FMODE_PWRITE)
576 ret = vfs_write(f.file, buf, count, &pos);
577 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 }
579
580 return ret;
581}
582
583/*
584 * Reduce an iovec's length in-place. Return the resulting number of segments
585 */
586unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
587{
588 unsigned long seg = 0;
589 size_t len = 0;
590
591 while (seg < nr_segs) {
592 seg++;
593 if (len + iov->iov_len >= to) {
594 iov->iov_len = to - len;
595 break;
596 }
597 len += iov->iov_len;
598 iov++;
599 }
600 return seg;
601}
Eric Sandeen19295522008-01-28 23:58:27 -0500602EXPORT_SYMBOL(iov_shorten);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Al Viro72ec3512013-03-20 10:42:10 -0400604static ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700605 unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
606{
607 struct kiocb kiocb;
608 ssize_t ret;
609
610 init_sync_kiocb(&kiocb, filp);
611 kiocb.ki_pos = *ppos;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700612 kiocb.ki_nbytes = len;
613
Zach Brown41003a72013-05-07 16:18:25 -0700614 ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700615 if (ret == -EIOCBQUEUED)
616 ret = wait_on_sync_kiocb(&kiocb);
617 *ppos = kiocb.ki_pos;
618 return ret;
619}
620
621/* Do it by hand, with file-ops */
Al Viro72ec3512013-03-20 10:42:10 -0400622static ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700623 unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
624{
625 struct iovec *vector = iov;
626 ssize_t ret = 0;
627
628 while (nr_segs > 0) {
629 void __user *base;
630 size_t len;
631 ssize_t nr;
632
633 base = vector->iov_base;
634 len = vector->iov_len;
635 vector++;
636 nr_segs--;
637
638 nr = fn(filp, base, len, ppos);
639
640 if (nr < 0) {
641 if (!ret)
642 ret = nr;
643 break;
644 }
645 ret += nr;
646 if (nr != len)
647 break;
648 }
649
650 return ret;
651}
652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653/* A write operation does a read from user space and vice versa */
654#define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
655
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700656ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
657 unsigned long nr_segs, unsigned long fast_segs,
658 struct iovec *fast_pointer,
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700659 struct iovec **ret_pointer)
Linus Torvalds435f49a2010-10-29 10:36:49 -0700660{
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700661 unsigned long seg;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700662 ssize_t ret;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700663 struct iovec *iov = fast_pointer;
664
Linus Torvalds435f49a2010-10-29 10:36:49 -0700665 /*
666 * SuS says "The readv() function *may* fail if the iovcnt argument
667 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
668 * traditionally returned zero for zero segments, so...
669 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700670 if (nr_segs == 0) {
671 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700672 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700673 }
674
Linus Torvalds435f49a2010-10-29 10:36:49 -0700675 /*
676 * First get the "struct iovec" from user memory and
677 * verify all the pointers
678 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700679 if (nr_segs > UIO_MAXIOV) {
680 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700681 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700682 }
683 if (nr_segs > fast_segs) {
Linus Torvalds435f49a2010-10-29 10:36:49 -0700684 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700685 if (iov == NULL) {
686 ret = -ENOMEM;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700687 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700688 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700689 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700690 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
691 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700692 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700693 }
694
Linus Torvalds435f49a2010-10-29 10:36:49 -0700695 /*
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700696 * According to the Single Unix Specification we should return EINVAL
697 * if an element length is < 0 when cast to ssize_t or if the
698 * total length would overflow the ssize_t return value of the
699 * system call.
Linus Torvalds435f49a2010-10-29 10:36:49 -0700700 *
701 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
702 * overflow case.
703 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700704 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700705 for (seg = 0; seg < nr_segs; seg++) {
706 void __user *buf = iov[seg].iov_base;
707 ssize_t len = (ssize_t)iov[seg].iov_len;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700708
709 /* see if we we're about to use an invalid len or if
710 * it's about to overflow ssize_t */
Linus Torvalds435f49a2010-10-29 10:36:49 -0700711 if (len < 0) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700712 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700713 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700714 }
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700715 if (type >= 0
Christopher Yeohfcf63402011-10-31 17:06:39 -0700716 && unlikely(!access_ok(vrfy_dir(type), buf, len))) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700717 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700718 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700719 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700720 if (len > MAX_RW_COUNT - ret) {
721 len = MAX_RW_COUNT - ret;
722 iov[seg].iov_len = len;
723 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700724 ret += len;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700725 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700726out:
727 *ret_pointer = iov;
728 return ret;
729}
730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731static ssize_t do_readv_writev(int type, struct file *file,
732 const struct iovec __user * uvector,
733 unsigned long nr_segs, loff_t *pos)
734{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 size_t tot_len;
736 struct iovec iovstack[UIO_FASTIOV];
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700737 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 io_fn_t fn;
740 iov_fn_t fnv;
741
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700742 ret = rw_copy_check_uvector(type, uvector, nr_segs,
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700743 ARRAY_SIZE(iovstack), iovstack, &iov);
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700744 if (ret <= 0)
745 goto out;
746
747 tot_len = ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 ret = rw_verify_area(type, file, pos, tot_len);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800749 if (ret < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 goto out;
751
752 fnv = NULL;
753 if (type == READ) {
754 fn = file->f_op->read;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700755 fnv = file->f_op->aio_read;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 } else {
757 fn = (io_fn_t)file->f_op->write;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700758 fnv = file->f_op->aio_write;
Al Viro03d95eb2013-03-20 13:04:20 -0400759 file_start_write(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 }
761
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700762 if (fnv)
763 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
764 pos, fnv);
765 else
766 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
Al Viro03d95eb2013-03-20 13:04:20 -0400768 if (type != READ)
769 file_end_write(file);
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771out:
772 if (iov != iovstack)
773 kfree(iov);
Robert Love0eeca282005-07-12 17:06:03 -0400774 if ((ret + (type == READ)) > 0) {
775 if (type == READ)
Eric Paris2a12a9d2009-12-17 21:24:21 -0500776 fsnotify_access(file);
Robert Love0eeca282005-07-12 17:06:03 -0400777 else
Eric Paris2a12a9d2009-12-17 21:24:21 -0500778 fsnotify_modify(file);
Robert Love0eeca282005-07-12 17:06:03 -0400779 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
783ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
784 unsigned long vlen, loff_t *pos)
785{
786 if (!(file->f_mode & FMODE_READ))
787 return -EBADF;
Al Viro72c2d532013-09-22 16:27:52 -0400788 if (!file->f_op->aio_read && !file->f_op->read)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 return -EINVAL;
790
791 return do_readv_writev(READ, file, vec, vlen, pos);
792}
793
794EXPORT_SYMBOL(vfs_readv);
795
796ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
797 unsigned long vlen, loff_t *pos)
798{
799 if (!(file->f_mode & FMODE_WRITE))
800 return -EBADF;
Al Viro72c2d532013-09-22 16:27:52 -0400801 if (!file->f_op->aio_write && !file->f_op->write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return -EINVAL;
803
804 return do_readv_writev(WRITE, file, vec, vlen, pos);
805}
806
807EXPORT_SYMBOL(vfs_writev);
808
Heiko Carstens3cdad422009-01-14 14:14:22 +0100809SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
810 unsigned long, vlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800812 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
Al Viro2903ff02012-08-28 12:52:22 -0400815 if (f.file) {
816 loff_t pos = file_pos_read(f.file);
817 ret = vfs_readv(f.file, vec, vlen, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400818 if (ret >= 0)
819 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800820 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 }
822
823 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800824 add_rchar(current, ret);
825 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 return ret;
827}
828
Heiko Carstens3cdad422009-01-14 14:14:22 +0100829SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
830 unsigned long, vlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800832 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Al Viro2903ff02012-08-28 12:52:22 -0400835 if (f.file) {
836 loff_t pos = file_pos_read(f.file);
837 ret = vfs_writev(f.file, vec, vlen, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400838 if (ret >= 0)
839 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800840 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 }
842
843 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800844 add_wchar(current, ret);
845 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 return ret;
847}
848
Linus Torvalds601cc112009-04-03 08:03:22 -0700849static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700850{
Linus Torvalds601cc112009-04-03 08:03:22 -0700851#define HALF_LONG_BITS (BITS_PER_LONG / 2)
852 return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
853}
854
855SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
856 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
857{
858 loff_t pos = pos_from_hilo(pos_h, pos_l);
Al Viro2903ff02012-08-28 12:52:22 -0400859 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700860 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700861
862 if (pos < 0)
863 return -EINVAL;
864
Al Viro2903ff02012-08-28 12:52:22 -0400865 f = fdget(fd);
866 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700867 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400868 if (f.file->f_mode & FMODE_PREAD)
869 ret = vfs_readv(f.file, vec, vlen, &pos);
870 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700871 }
872
873 if (ret > 0)
874 add_rchar(current, ret);
875 inc_syscr(current);
876 return ret;
877}
878
879SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
Linus Torvalds601cc112009-04-03 08:03:22 -0700880 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700881{
Linus Torvalds601cc112009-04-03 08:03:22 -0700882 loff_t pos = pos_from_hilo(pos_h, pos_l);
Al Viro2903ff02012-08-28 12:52:22 -0400883 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700884 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700885
886 if (pos < 0)
887 return -EINVAL;
888
Al Viro2903ff02012-08-28 12:52:22 -0400889 f = fdget(fd);
890 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700891 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400892 if (f.file->f_mode & FMODE_PWRITE)
893 ret = vfs_writev(f.file, vec, vlen, &pos);
894 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700895 }
896
897 if (ret > 0)
898 add_wchar(current, ret);
899 inc_syscw(current);
900 return ret;
901}
902
Al Viro72ec3512013-03-20 10:42:10 -0400903#ifdef CONFIG_COMPAT
904
905static ssize_t compat_do_readv_writev(int type, struct file *file,
906 const struct compat_iovec __user *uvector,
907 unsigned long nr_segs, loff_t *pos)
908{
909 compat_ssize_t tot_len;
910 struct iovec iovstack[UIO_FASTIOV];
911 struct iovec *iov = iovstack;
912 ssize_t ret;
913 io_fn_t fn;
914 iov_fn_t fnv;
915
Al Viro72ec3512013-03-20 10:42:10 -0400916 ret = compat_rw_copy_check_uvector(type, uvector, nr_segs,
917 UIO_FASTIOV, iovstack, &iov);
918 if (ret <= 0)
919 goto out;
920
921 tot_len = ret;
922 ret = rw_verify_area(type, file, pos, tot_len);
923 if (ret < 0)
924 goto out;
925
926 fnv = NULL;
927 if (type == READ) {
928 fn = file->f_op->read;
929 fnv = file->f_op->aio_read;
930 } else {
931 fn = (io_fn_t)file->f_op->write;
932 fnv = file->f_op->aio_write;
Al Viro03d95eb2013-03-20 13:04:20 -0400933 file_start_write(file);
Al Viro72ec3512013-03-20 10:42:10 -0400934 }
935
Al Viro03d95eb2013-03-20 13:04:20 -0400936 if (fnv)
Al Viro72ec3512013-03-20 10:42:10 -0400937 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
938 pos, fnv);
Al Viro03d95eb2013-03-20 13:04:20 -0400939 else
Al Viro72ec3512013-03-20 10:42:10 -0400940 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
941
Al Viro03d95eb2013-03-20 13:04:20 -0400942 if (type != READ)
943 file_end_write(file);
944
Al Viro72ec3512013-03-20 10:42:10 -0400945out:
946 if (iov != iovstack)
947 kfree(iov);
948 if ((ret + (type == READ)) > 0) {
949 if (type == READ)
950 fsnotify_access(file);
951 else
952 fsnotify_modify(file);
953 }
954 return ret;
955}
956
957static size_t compat_readv(struct file *file,
958 const struct compat_iovec __user *vec,
959 unsigned long vlen, loff_t *pos)
960{
961 ssize_t ret = -EBADF;
962
963 if (!(file->f_mode & FMODE_READ))
964 goto out;
965
966 ret = -EINVAL;
Al Viro72c2d532013-09-22 16:27:52 -0400967 if (!file->f_op->aio_read && !file->f_op->read)
Al Viro72ec3512013-03-20 10:42:10 -0400968 goto out;
969
970 ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
971
972out:
973 if (ret > 0)
974 add_rchar(current, ret);
975 inc_syscr(current);
976 return ret;
977}
978
Heiko Carstensdfd948e2014-01-29 14:05:44 -0800979COMPAT_SYSCALL_DEFINE3(readv, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -0400980 const struct compat_iovec __user *,vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -0800981 compat_ulong_t, vlen)
Al Viro72ec3512013-03-20 10:42:10 -0400982{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800983 struct fd f = fdget_pos(fd);
Al Viro72ec3512013-03-20 10:42:10 -0400984 ssize_t ret;
985 loff_t pos;
986
987 if (!f.file)
988 return -EBADF;
989 pos = f.file->f_pos;
990 ret = compat_readv(f.file, vec, vlen, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400991 if (ret >= 0)
992 f.file->f_pos = pos;
Linus Torvalds9c225f22014-03-03 09:36:58 -0800993 fdput_pos(f);
Al Viro72ec3512013-03-20 10:42:10 -0400994 return ret;
995}
996
Heiko Carstens378a10f2014-03-05 10:43:51 +0100997static long __compat_sys_preadv64(unsigned long fd,
998 const struct compat_iovec __user *vec,
999 unsigned long vlen, loff_t pos)
Al Viro72ec3512013-03-20 10:42:10 -04001000{
1001 struct fd f;
1002 ssize_t ret;
1003
1004 if (pos < 0)
1005 return -EINVAL;
1006 f = fdget(fd);
1007 if (!f.file)
1008 return -EBADF;
1009 ret = -ESPIPE;
1010 if (f.file->f_mode & FMODE_PREAD)
1011 ret = compat_readv(f.file, vec, vlen, &pos);
1012 fdput(f);
1013 return ret;
1014}
1015
Heiko Carstens378a10f2014-03-05 10:43:51 +01001016#ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1017COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
1018 const struct compat_iovec __user *,vec,
1019 unsigned long, vlen, loff_t, pos)
1020{
1021 return __compat_sys_preadv64(fd, vec, vlen, pos);
1022}
1023#endif
1024
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001025COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -04001026 const struct compat_iovec __user *,vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001027 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
Al Viro72ec3512013-03-20 10:42:10 -04001028{
1029 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
Heiko Carstens378a10f2014-03-05 10:43:51 +01001030
1031 return __compat_sys_preadv64(fd, vec, vlen, pos);
Al Viro72ec3512013-03-20 10:42:10 -04001032}
1033
1034static size_t compat_writev(struct file *file,
1035 const struct compat_iovec __user *vec,
1036 unsigned long vlen, loff_t *pos)
1037{
1038 ssize_t ret = -EBADF;
1039
1040 if (!(file->f_mode & FMODE_WRITE))
1041 goto out;
1042
1043 ret = -EINVAL;
Al Viro72c2d532013-09-22 16:27:52 -04001044 if (!file->f_op->aio_write && !file->f_op->write)
Al Viro72ec3512013-03-20 10:42:10 -04001045 goto out;
1046
1047 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
1048
1049out:
1050 if (ret > 0)
1051 add_wchar(current, ret);
1052 inc_syscw(current);
1053 return ret;
1054}
1055
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001056COMPAT_SYSCALL_DEFINE3(writev, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -04001057 const struct compat_iovec __user *, vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001058 compat_ulong_t, vlen)
Al Viro72ec3512013-03-20 10:42:10 -04001059{
Linus Torvalds9c225f22014-03-03 09:36:58 -08001060 struct fd f = fdget_pos(fd);
Al Viro72ec3512013-03-20 10:42:10 -04001061 ssize_t ret;
1062 loff_t pos;
1063
1064 if (!f.file)
1065 return -EBADF;
1066 pos = f.file->f_pos;
1067 ret = compat_writev(f.file, vec, vlen, &pos);
Al Viro5faf1532013-06-15 05:49:36 +04001068 if (ret >= 0)
1069 f.file->f_pos = pos;
Linus Torvalds9c225f22014-03-03 09:36:58 -08001070 fdput_pos(f);
Al Viro72ec3512013-03-20 10:42:10 -04001071 return ret;
1072}
1073
Heiko Carstens378a10f2014-03-05 10:43:51 +01001074static long __compat_sys_pwritev64(unsigned long fd,
1075 const struct compat_iovec __user *vec,
1076 unsigned long vlen, loff_t pos)
Al Viro72ec3512013-03-20 10:42:10 -04001077{
1078 struct fd f;
1079 ssize_t ret;
1080
1081 if (pos < 0)
1082 return -EINVAL;
1083 f = fdget(fd);
1084 if (!f.file)
1085 return -EBADF;
1086 ret = -ESPIPE;
1087 if (f.file->f_mode & FMODE_PWRITE)
1088 ret = compat_writev(f.file, vec, vlen, &pos);
1089 fdput(f);
1090 return ret;
1091}
1092
Heiko Carstens378a10f2014-03-05 10:43:51 +01001093#ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1094COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
1095 const struct compat_iovec __user *,vec,
1096 unsigned long, vlen, loff_t, pos)
1097{
1098 return __compat_sys_pwritev64(fd, vec, vlen, pos);
1099}
1100#endif
1101
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001102COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -04001103 const struct compat_iovec __user *,vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001104 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
Al Viro72ec3512013-03-20 10:42:10 -04001105{
1106 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
Heiko Carstens378a10f2014-03-05 10:43:51 +01001107
1108 return __compat_sys_pwritev64(fd, vec, vlen, pos);
Al Viro72ec3512013-03-20 10:42:10 -04001109}
1110#endif
1111
Al Viro19f4fc32013-02-24 02:17:03 -05001112static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
1113 size_t count, loff_t max)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114{
Al Viro2903ff02012-08-28 12:52:22 -04001115 struct fd in, out;
1116 struct inode *in_inode, *out_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 loff_t pos;
Al Viro7995bd22013-06-20 18:58:36 +04001118 loff_t out_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 ssize_t retval;
Al Viro2903ff02012-08-28 12:52:22 -04001120 int fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
1122 /*
1123 * Get input file, and verify that it is ok..
1124 */
1125 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001126 in = fdget(in_fd);
1127 if (!in.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 goto out;
Al Viro2903ff02012-08-28 12:52:22 -04001129 if (!(in.file->f_mode & FMODE_READ))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 goto fput_in;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 retval = -ESPIPE;
Al Viro7995bd22013-06-20 18:58:36 +04001132 if (!ppos) {
1133 pos = in.file->f_pos;
1134 } else {
1135 pos = *ppos;
Al Viro2903ff02012-08-28 12:52:22 -04001136 if (!(in.file->f_mode & FMODE_PREAD))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 goto fput_in;
Al Viro7995bd22013-06-20 18:58:36 +04001138 }
1139 retval = rw_verify_area(READ, in.file, &pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -08001140 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 goto fput_in;
Linus Torvaldse28cc712006-01-04 16:20:40 -08001142 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 /*
1145 * Get output file, and verify that it is ok..
1146 */
1147 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001148 out = fdget(out_fd);
1149 if (!out.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 goto fput_in;
Al Viro2903ff02012-08-28 12:52:22 -04001151 if (!(out.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 goto fput_out;
1153 retval = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -05001154 in_inode = file_inode(in.file);
1155 out_inode = file_inode(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001156 out_pos = out.file->f_pos;
1157 retval = rw_verify_area(WRITE, out.file, &out_pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -08001158 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 goto fput_out;
Linus Torvaldse28cc712006-01-04 16:20:40 -08001160 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 if (!max)
1163 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
1164
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 if (unlikely(pos + count > max)) {
1166 retval = -EOVERFLOW;
1167 if (pos >= max)
1168 goto fput_out;
1169 count = max - pos;
1170 }
1171
Jens Axboed96e6e72007-06-11 12:18:52 +02001172 fl = 0;
Jens Axboe534f2aa2007-06-01 14:52:37 +02001173#if 0
Jens Axboed96e6e72007-06-11 12:18:52 +02001174 /*
1175 * We need to debate whether we can enable this or not. The
1176 * man page documents EAGAIN return for the output at least,
1177 * and the application is arguably buggy if it doesn't expect
1178 * EAGAIN on a non-blocking file descriptor.
1179 */
Al Viro2903ff02012-08-28 12:52:22 -04001180 if (in.file->f_flags & O_NONBLOCK)
Jens Axboed96e6e72007-06-11 12:18:52 +02001181 fl = SPLICE_F_NONBLOCK;
Jens Axboe534f2aa2007-06-01 14:52:37 +02001182#endif
Al Viro50cd2c52013-05-23 20:10:34 -04001183 file_start_write(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001184 retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl);
Al Viro50cd2c52013-05-23 20:10:34 -04001185 file_end_write(out.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 if (retval > 0) {
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08001188 add_rchar(current, retval);
1189 add_wchar(current, retval);
Scott Wolchoka68c2f12012-12-20 15:05:52 -08001190 fsnotify_access(in.file);
1191 fsnotify_modify(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001192 out.file->f_pos = out_pos;
1193 if (ppos)
1194 *ppos = pos;
1195 else
1196 in.file->f_pos = pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08001199 inc_syscr(current);
1200 inc_syscw(current);
Al Viro7995bd22013-06-20 18:58:36 +04001201 if (pos > max)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 retval = -EOVERFLOW;
1203
1204fput_out:
Al Viro2903ff02012-08-28 12:52:22 -04001205 fdput(out);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206fput_in:
Al Viro2903ff02012-08-28 12:52:22 -04001207 fdput(in);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208out:
1209 return retval;
1210}
1211
Heiko Carstens002c8972009-01-14 14:14:18 +01001212SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213{
1214 loff_t pos;
1215 off_t off;
1216 ssize_t ret;
1217
1218 if (offset) {
1219 if (unlikely(get_user(off, offset)))
1220 return -EFAULT;
1221 pos = off;
1222 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1223 if (unlikely(put_user(pos, offset)))
1224 return -EFAULT;
1225 return ret;
1226 }
1227
1228 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1229}
1230
Heiko Carstens002c8972009-01-14 14:14:18 +01001231SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232{
1233 loff_t pos;
1234 ssize_t ret;
1235
1236 if (offset) {
1237 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1238 return -EFAULT;
1239 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1240 if (unlikely(put_user(pos, offset)))
1241 return -EFAULT;
1242 return ret;
1243 }
1244
1245 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1246}
Al Viro19f4fc32013-02-24 02:17:03 -05001247
1248#ifdef CONFIG_COMPAT
1249COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
1250 compat_off_t __user *, offset, compat_size_t, count)
1251{
1252 loff_t pos;
1253 off_t off;
1254 ssize_t ret;
1255
1256 if (offset) {
1257 if (unlikely(get_user(off, offset)))
1258 return -EFAULT;
1259 pos = off;
1260 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1261 if (unlikely(put_user(pos, offset)))
1262 return -EFAULT;
1263 return ret;
1264 }
1265
1266 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1267}
1268
1269COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
1270 compat_loff_t __user *, offset, compat_size_t, count)
1271{
1272 loff_t pos;
1273 ssize_t ret;
1274
1275 if (offset) {
1276 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1277 return -EFAULT;
1278 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1279 if (unlikely(put_user(pos, offset)))
1280 return -EFAULT;
1281 return ret;
1282 }
1283
1284 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1285}
1286#endif