blob: 933b53a375b4cd872c1e9e8e314ea57a13837980 [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>
Zach Brown29732932015-11-10 16:53:30 -050019#include <linux/mount.h>
Wouter van Kesteren2feb55f2016-02-16 22:20:59 +010020#include <linux/fs.h>
Al Viro06ae43f2013-03-20 13:19:30 -040021#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
23#include <asm/uaccess.h>
24#include <asm/unistd.h>
25
Al Viroc0bd14af2013-05-04 15:00:54 -040026typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
Al Viro293bc982014-02-11 18:37:41 -050027typedef ssize_t (*iter_fn_t)(struct kiocb *, struct iov_iter *);
Al Viroc0bd14af2013-05-04 15:00:54 -040028
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,
Al Viroaad4f8b2014-04-02 14:33:16 -040031 .read_iter = generic_file_read_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -070032 .mmap = generic_file_readonly_mmap,
Jens Axboe534f2aa2007-06-01 14:52:37 +020033 .splice_read = generic_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070034};
35
36EXPORT_SYMBOL(generic_ro_fops);
37
Al Virocccb5a12010-12-17 07:44:05 -050038static inline int unsigned_offsets(struct file *file)
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070039{
Al Virocccb5a12010-12-17 07:44:05 -050040 return file->f_mode & FMODE_UNSIGNED_OFFSET;
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -070041}
42
Jie Liu46a1c2c2013-06-25 12:02:13 +080043/**
44 * vfs_setpos - update the file offset for lseek
45 * @file: file structure in question
46 * @offset: file offset to seek to
47 * @maxsize: maximum file size
48 *
49 * This is a low-level filesystem helper for updating the file offset to
50 * the value specified by @offset if the given offset is valid and it is
51 * not equal to the current file offset.
52 *
53 * Return the specified offset on success and -EINVAL on invalid offset.
54 */
55loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
Andi Kleenef3d0fd2011-09-15 16:06:48 -070056{
57 if (offset < 0 && !unsigned_offsets(file))
58 return -EINVAL;
59 if (offset > maxsize)
60 return -EINVAL;
61
62 if (offset != file->f_pos) {
63 file->f_pos = offset;
64 file->f_version = 0;
65 }
66 return offset;
67}
Jie Liu46a1c2c2013-06-25 12:02:13 +080068EXPORT_SYMBOL(vfs_setpos);
Andi Kleenef3d0fd2011-09-15 16:06:48 -070069
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020070/**
Andi Kleen57604952011-09-15 16:06:50 -070071 * generic_file_llseek_size - generic llseek implementation for regular files
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020072 * @file: file structure to seek on
73 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -080074 * @whence: type of seek
Eric Sandeene8b96eb2012-04-30 13:11:29 -050075 * @size: max size of this file in file system
76 * @eof: offset used for SEEK_END position
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020077 *
Andi Kleen57604952011-09-15 16:06:50 -070078 * This is a variant of generic_file_llseek that allows passing in a custom
Eric Sandeene8b96eb2012-04-30 13:11:29 -050079 * maximum file size and a custom EOF position, for e.g. hashed directories
Andi Kleenef3d0fd2011-09-15 16:06:48 -070080 *
81 * Synchronization:
Andi Kleen57604952011-09-15 16:06:50 -070082 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
Andi Kleenef3d0fd2011-09-15 16:06:48 -070083 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
84 * read/writes behave like SEEK_SET against seeks.
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020085 */
Andi Kleen9465efc2008-06-27 11:05:24 +020086loff_t
Andrew Morton965c8e52012-12-17 15:59:39 -080087generic_file_llseek_size(struct file *file, loff_t offset, int whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -050088 loff_t maxsize, loff_t eof)
Linus Torvalds1da177e2005-04-16 15:20:36 -070089{
Andrew Morton965c8e52012-12-17 15:59:39 -080090 switch (whence) {
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020091 case SEEK_END:
Eric Sandeene8b96eb2012-04-30 13:11:29 -050092 offset += eof;
Christoph Hellwig3a8cff42008-08-11 15:37:17 +020093 break;
94 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -080095 /*
96 * Here we special-case the lseek(fd, 0, SEEK_CUR)
97 * position-querying operation. Avoid rewriting the "same"
98 * f_pos value back to the file because a concurrent read(),
99 * write() or lseek() might have altered it
100 */
101 if (offset == 0)
102 return file->f_pos;
Andi Kleenef3d0fd2011-09-15 16:06:48 -0700103 /*
104 * f_lock protects against read/modify/write race with other
105 * SEEK_CURs. Note that parallel writes and reads behave
106 * like SEEK_SET.
107 */
108 spin_lock(&file->f_lock);
Jie Liu46a1c2c2013-06-25 12:02:13 +0800109 offset = vfs_setpos(file, file->f_pos + offset, maxsize);
Andi Kleenef3d0fd2011-09-15 16:06:48 -0700110 spin_unlock(&file->f_lock);
111 return offset;
Josef Bacik982d8162011-07-18 13:21:35 -0400112 case SEEK_DATA:
113 /*
114 * In the generic case the entire file is data, so as long as
115 * offset isn't at the end of the file then the offset is data.
116 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500117 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400118 return -ENXIO;
119 break;
120 case SEEK_HOLE:
121 /*
122 * There is a virtual hole at the end of the file, so as long as
123 * offset isn't i_size or larger, return i_size.
124 */
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500125 if (offset >= eof)
Josef Bacik982d8162011-07-18 13:21:35 -0400126 return -ENXIO;
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500127 offset = eof;
Josef Bacik982d8162011-07-18 13:21:35 -0400128 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 }
Christoph Hellwig3a8cff42008-08-11 15:37:17 +0200130
Jie Liu46a1c2c2013-06-25 12:02:13 +0800131 return vfs_setpos(file, offset, maxsize);
Andi Kleen57604952011-09-15 16:06:50 -0700132}
133EXPORT_SYMBOL(generic_file_llseek_size);
134
135/**
136 * generic_file_llseek - generic llseek implementation for regular files
137 * @file: file structure to seek on
138 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800139 * @whence: type of seek
Andi Kleen57604952011-09-15 16:06:50 -0700140 *
141 * This is a generic implemenation of ->llseek useable for all normal local
142 * filesystems. It just updates the file offset to the value specified by
Ming Lei546ae2d2013-04-29 15:06:07 -0700143 * @offset and @whence.
Andi Kleen57604952011-09-15 16:06:50 -0700144 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800145loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
Andi Kleen57604952011-09-15 16:06:50 -0700146{
147 struct inode *inode = file->f_mapping->host;
148
Andrew Morton965c8e52012-12-17 15:59:39 -0800149 return generic_file_llseek_size(file, offset, whence,
Eric Sandeene8b96eb2012-04-30 13:11:29 -0500150 inode->i_sb->s_maxbytes,
151 i_size_read(inode));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152}
Andi Kleen9465efc2008-06-27 11:05:24 +0200153EXPORT_SYMBOL(generic_file_llseek);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154
jan Blunckae6afc32010-05-26 14:44:48 -0700155/**
Al Viro1bf9d142013-06-16 20:27:42 +0400156 * fixed_size_llseek - llseek implementation for fixed-sized devices
157 * @file: file structure to seek on
158 * @offset: file offset to seek to
159 * @whence: type of seek
160 * @size: size of the file
161 *
162 */
163loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size)
164{
165 switch (whence) {
166 case SEEK_SET: case SEEK_CUR: case SEEK_END:
167 return generic_file_llseek_size(file, offset, whence,
168 size, size);
169 default:
170 return -EINVAL;
171 }
172}
173EXPORT_SYMBOL(fixed_size_llseek);
174
175/**
Al Virob25472f2015-12-05 22:04:48 -0500176 * no_seek_end_llseek - llseek implementation for fixed-sized devices
177 * @file: file structure to seek on
178 * @offset: file offset to seek to
179 * @whence: type of seek
180 *
181 */
182loff_t no_seek_end_llseek(struct file *file, loff_t offset, int whence)
183{
184 switch (whence) {
185 case SEEK_SET: case SEEK_CUR:
186 return generic_file_llseek_size(file, offset, whence,
Wouter van Kesteren2feb55f2016-02-16 22:20:59 +0100187 OFFSET_MAX, 0);
Al Virob25472f2015-12-05 22:04:48 -0500188 default:
189 return -EINVAL;
190 }
191}
192EXPORT_SYMBOL(no_seek_end_llseek);
193
194/**
195 * no_seek_end_llseek_size - llseek implementation for fixed-sized devices
196 * @file: file structure to seek on
197 * @offset: file offset to seek to
198 * @whence: type of seek
199 * @size: maximal offset allowed
200 *
201 */
202loff_t no_seek_end_llseek_size(struct file *file, loff_t offset, int whence, loff_t size)
203{
204 switch (whence) {
205 case SEEK_SET: case SEEK_CUR:
206 return generic_file_llseek_size(file, offset, whence,
207 size, 0);
208 default:
209 return -EINVAL;
210 }
211}
212EXPORT_SYMBOL(no_seek_end_llseek_size);
213
214/**
jan Blunckae6afc32010-05-26 14:44:48 -0700215 * noop_llseek - No Operation Performed llseek implementation
216 * @file: file structure to seek on
217 * @offset: file offset to seek to
Andrew Morton965c8e52012-12-17 15:59:39 -0800218 * @whence: type of seek
jan Blunckae6afc32010-05-26 14:44:48 -0700219 *
220 * This is an implementation of ->llseek useable for the rare special case when
221 * userspace expects the seek to succeed but the (device) file is actually not
222 * able to perform the seek. In this case you use noop_llseek() instead of
223 * falling back to the default implementation of ->llseek.
224 */
Andrew Morton965c8e52012-12-17 15:59:39 -0800225loff_t noop_llseek(struct file *file, loff_t offset, int whence)
jan Blunckae6afc32010-05-26 14:44:48 -0700226{
227 return file->f_pos;
228}
229EXPORT_SYMBOL(noop_llseek);
230
Andrew Morton965c8e52012-12-17 15:59:39 -0800231loff_t no_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232{
233 return -ESPIPE;
234}
235EXPORT_SYMBOL(no_llseek);
236
Andrew Morton965c8e52012-12-17 15:59:39 -0800237loff_t default_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238{
Al Viro496ad9a2013-01-23 17:07:38 -0500239 struct inode *inode = file_inode(file);
David Sterba16abef02008-04-22 15:09:22 +0200240 loff_t retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
Al Viro59551022016-01-22 15:40:57 -0500242 inode_lock(inode);
Andrew Morton965c8e52012-12-17 15:59:39 -0800243 switch (whence) {
Chris Snook7b8e8922007-05-08 00:24:13 -0700244 case SEEK_END:
Josef Bacik982d8162011-07-18 13:21:35 -0400245 offset += i_size_read(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 break;
Chris Snook7b8e8922007-05-08 00:24:13 -0700247 case SEEK_CUR:
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800248 if (offset == 0) {
249 retval = file->f_pos;
250 goto out;
251 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 offset += file->f_pos;
Josef Bacik982d8162011-07-18 13:21:35 -0400253 break;
254 case SEEK_DATA:
255 /*
256 * In the generic case the entire file is data, so as
257 * long as offset isn't at the end of the file then the
258 * offset is data.
259 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300260 if (offset >= inode->i_size) {
261 retval = -ENXIO;
262 goto out;
263 }
Josef Bacik982d8162011-07-18 13:21:35 -0400264 break;
265 case SEEK_HOLE:
266 /*
267 * There is a virtual hole at the end of the file, so
268 * as long as offset isn't i_size or larger, return
269 * i_size.
270 */
Dan Carpenterbacb2d82011-07-26 17:25:20 +0300271 if (offset >= inode->i_size) {
272 retval = -ENXIO;
273 goto out;
274 }
Josef Bacik982d8162011-07-18 13:21:35 -0400275 offset = inode->i_size;
276 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 }
278 retval = -EINVAL;
Al Virocccb5a12010-12-17 07:44:05 -0500279 if (offset >= 0 || unsigned_offsets(file)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 if (offset != file->f_pos) {
281 file->f_pos = offset;
282 file->f_version = 0;
283 }
284 retval = offset;
285 }
Alain Knaff5b6f1eb2008-11-10 17:08:08 -0800286out:
Al Viro59551022016-01-22 15:40:57 -0500287 inode_unlock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return retval;
289}
290EXPORT_SYMBOL(default_llseek);
291
Andrew Morton965c8e52012-12-17 15:59:39 -0800292loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293{
294 loff_t (*fn)(struct file *, loff_t, int);
295
296 fn = no_llseek;
297 if (file->f_mode & FMODE_LSEEK) {
Al Viro72c2d532013-09-22 16:27:52 -0400298 if (file->f_op->llseek)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 fn = file->f_op->llseek;
300 }
Andrew Morton965c8e52012-12-17 15:59:39 -0800301 return fn(file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302}
303EXPORT_SYMBOL(vfs_llseek);
304
Andrew Morton965c8e52012-12-17 15:59:39 -0800305SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306{
307 off_t retval;
Linus Torvalds9c225f22014-03-03 09:36:58 -0800308 struct fd f = fdget_pos(fd);
Al Viro2903ff02012-08-28 12:52:22 -0400309 if (!f.file)
310 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
312 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800313 if (whence <= SEEK_MAX) {
314 loff_t res = vfs_llseek(f.file, offset, whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 retval = res;
316 if (res != (loff_t)retval)
317 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
318 }
Linus Torvalds9c225f22014-03-03 09:36:58 -0800319 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 return retval;
321}
322
Al Viro561c6732013-02-24 10:52:26 -0500323#ifdef CONFIG_COMPAT
324COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
325{
326 return sys_lseek(fd, offset, whence);
327}
328#endif
329
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330#ifdef __ARCH_WANT_SYS_LLSEEK
Heiko Carstens003d7ab2009-01-14 14:14:21 +0100331SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
332 unsigned long, offset_low, loff_t __user *, result,
Andrew Morton965c8e52012-12-17 15:59:39 -0800333 unsigned int, whence)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334{
335 int retval;
Eric Biggersd7a15f82014-03-16 14:24:08 -0500336 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 loff_t offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
Al Viro2903ff02012-08-28 12:52:22 -0400339 if (!f.file)
340 return -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
342 retval = -EINVAL;
Andrew Morton965c8e52012-12-17 15:59:39 -0800343 if (whence > SEEK_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 goto out_putf;
345
Al Viro2903ff02012-08-28 12:52:22 -0400346 offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
Andrew Morton965c8e52012-12-17 15:59:39 -0800347 whence);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 retval = (int)offset;
350 if (offset >= 0) {
351 retval = -EFAULT;
352 if (!copy_to_user(result, &offset, sizeof(offset)))
353 retval = 0;
354 }
355out_putf:
Eric Biggersd7a15f82014-03-16 14:24:08 -0500356 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 return retval;
358}
359#endif
360
Christoph Hellwigdbe4e192015-01-25 21:11:59 +0100361ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos)
362{
363 struct kiocb kiocb;
364 ssize_t ret;
365
366 if (!file->f_op->read_iter)
367 return -EINVAL;
368
369 init_sync_kiocb(&kiocb, file);
370 kiocb.ki_pos = *ppos;
Christoph Hellwigdbe4e192015-01-25 21:11:59 +0100371
372 iter->type |= READ;
373 ret = file->f_op->read_iter(&kiocb, iter);
Christoph Hellwig599bd192015-02-11 19:59:44 +0100374 BUG_ON(ret == -EIOCBQUEUED);
Christoph Hellwigdbe4e192015-01-25 21:11:59 +0100375 if (ret > 0)
376 *ppos = kiocb.ki_pos;
377 return ret;
378}
379EXPORT_SYMBOL(vfs_iter_read);
380
381ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos)
382{
383 struct kiocb kiocb;
384 ssize_t ret;
385
386 if (!file->f_op->write_iter)
387 return -EINVAL;
388
389 init_sync_kiocb(&kiocb, file);
390 kiocb.ki_pos = *ppos;
Christoph Hellwigdbe4e192015-01-25 21:11:59 +0100391
392 iter->type |= WRITE;
393 ret = file->f_op->write_iter(&kiocb, iter);
Christoph Hellwig599bd192015-02-11 19:59:44 +0100394 BUG_ON(ret == -EIOCBQUEUED);
Christoph Hellwigdbe4e192015-01-25 21:11:59 +0100395 if (ret > 0)
396 *ppos = kiocb.ki_pos;
397 return ret;
398}
399EXPORT_SYMBOL(vfs_iter_write);
400
Al Viro68d70d02013-06-19 15:26:04 +0400401int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402{
403 struct inode *inode;
404 loff_t pos;
James Morrisc43e2592008-01-12 22:05:48 +1100405 int retval = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
Al Viro496ad9a2013-01-23 17:07:38 -0500407 inode = file_inode(file);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800408 if (unlikely((ssize_t) count < 0))
James Morrisc43e2592008-01-12 22:05:48 +1100409 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 pos = *ppos;
Al Virocccb5a12010-12-17 07:44:05 -0500411 if (unlikely(pos < 0)) {
412 if (!unsigned_offsets(file))
413 return retval;
414 if (count >= -pos) /* both values are in 0..LLONG_MAX */
415 return -EOVERFLOW;
416 } else if (unlikely((loff_t) (pos + count) < 0)) {
417 if (!unsigned_offsets(file))
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700418 return retval;
419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Jeff Laytonbd61e0a2015-01-16 15:05:55 -0500421 if (unlikely(inode->i_flctx && mandatory_lock(inode))) {
Christoph Hellwigacc15572015-12-03 12:59:49 +0100422 retval = locks_mandatory_area(inode, file, pos, pos + count - 1,
423 read_write == READ ? F_RDLCK : F_WRLCK);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800424 if (retval < 0)
425 return retval;
426 }
Al Virobc613842016-03-31 21:48:20 -0400427 return security_file_permission(file,
James Morrisc43e2592008-01-12 22:05:48 +1100428 read_write == READ ? MAY_READ : MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429}
430
Al Viro5d5d5682015-04-03 15:41:18 -0400431static ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
Al Viro293bc982014-02-11 18:37:41 -0500432{
433 struct iovec iov = { .iov_base = buf, .iov_len = len };
434 struct kiocb kiocb;
435 struct iov_iter iter;
436 ssize_t ret;
437
438 init_sync_kiocb(&kiocb, filp);
439 kiocb.ki_pos = *ppos;
Al Viro293bc982014-02-11 18:37:41 -0500440 iov_iter_init(&iter, READ, &iov, 1, len);
441
442 ret = filp->f_op->read_iter(&kiocb, &iter);
Christoph Hellwig599bd192015-02-11 19:59:44 +0100443 BUG_ON(ret == -EIOCBQUEUED);
Al Viro293bc982014-02-11 18:37:41 -0500444 *ppos = kiocb.ki_pos;
445 return ret;
446}
447
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200448ssize_t __vfs_read(struct file *file, char __user *buf, size_t count,
449 loff_t *pos)
450{
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200451 if (file->f_op->read)
Al Viro3d04c8a2015-04-03 15:09:18 -0400452 return file->f_op->read(file, buf, count, pos);
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200453 else if (file->f_op->read_iter)
Al Viro3d04c8a2015-04-03 15:09:18 -0400454 return new_sync_read(file, buf, count, pos);
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200455 else
Al Viro3d04c8a2015-04-03 15:09:18 -0400456 return -EINVAL;
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200457}
Al Viro3d04c8a2015-04-03 15:09:18 -0400458EXPORT_SYMBOL(__vfs_read);
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
461{
462 ssize_t ret;
463
464 if (!(file->f_mode & FMODE_READ))
465 return -EBADF;
Al Viro7f7f25e2014-02-11 17:49:24 -0500466 if (!(file->f_mode & FMODE_CAN_READ))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 return -EINVAL;
468 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
469 return -EFAULT;
470
471 ret = rw_verify_area(READ, file, pos, count);
Al Virobc613842016-03-31 21:48:20 -0400472 if (!ret) {
473 if (count > MAX_RW_COUNT)
474 count = MAX_RW_COUNT;
Dmitry Kasatkin6fb50322014-11-05 17:01:17 +0200475 ret = __vfs_read(file, buf, count, pos);
James Morrisc43e2592008-01-12 22:05:48 +1100476 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500477 fsnotify_access(file);
James Morrisc43e2592008-01-12 22:05:48 +1100478 add_rchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 }
James Morrisc43e2592008-01-12 22:05:48 +1100480 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 }
482
483 return ret;
484}
485
486EXPORT_SYMBOL(vfs_read);
487
Al Viro5d5d5682015-04-03 15:41:18 -0400488static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
Al Viro293bc982014-02-11 18:37:41 -0500489{
490 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
491 struct kiocb kiocb;
492 struct iov_iter iter;
493 ssize_t ret;
494
495 init_sync_kiocb(&kiocb, filp);
496 kiocb.ki_pos = *ppos;
Al Viro293bc982014-02-11 18:37:41 -0500497 iov_iter_init(&iter, WRITE, &iov, 1, len);
498
499 ret = filp->f_op->write_iter(&kiocb, &iter);
Christoph Hellwig599bd192015-02-11 19:59:44 +0100500 BUG_ON(ret == -EIOCBQUEUED);
Al Virof765b132015-04-06 20:50:38 -0400501 if (ret > 0)
502 *ppos = kiocb.ki_pos;
Al Viro293bc982014-02-11 18:37:41 -0500503 return ret;
504}
505
Al Viro493c84c2015-04-03 15:06:43 -0400506ssize_t __vfs_write(struct file *file, const char __user *p, size_t count,
507 loff_t *pos)
508{
509 if (file->f_op->write)
510 return file->f_op->write(file, p, count, pos);
Al Viro493c84c2015-04-03 15:06:43 -0400511 else if (file->f_op->write_iter)
512 return new_sync_write(file, p, count, pos);
513 else
514 return -EINVAL;
515}
516EXPORT_SYMBOL(__vfs_write);
517
Al Viro06ae43f2013-03-20 13:19:30 -0400518ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
519{
520 mm_segment_t old_fs;
521 const char __user *p;
522 ssize_t ret;
523
Al Viro7f7f25e2014-02-11 17:49:24 -0500524 if (!(file->f_mode & FMODE_CAN_WRITE))
Al Viro3e84f482013-03-27 15:20:30 +0000525 return -EINVAL;
526
Al Viro06ae43f2013-03-20 13:19:30 -0400527 old_fs = get_fs();
528 set_fs(get_ds());
529 p = (__force const char __user *)buf;
530 if (count > MAX_RW_COUNT)
531 count = MAX_RW_COUNT;
Al Viro493c84c2015-04-03 15:06:43 -0400532 ret = __vfs_write(file, p, count, pos);
Al Viro06ae43f2013-03-20 13:19:30 -0400533 set_fs(old_fs);
534 if (ret > 0) {
535 fsnotify_modify(file);
536 add_wchar(current, ret);
537 }
538 inc_syscw(current);
539 return ret;
540}
541
Al Viro2ec3a122014-08-19 11:48:09 -0400542EXPORT_SYMBOL(__kernel_write);
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
545{
546 ssize_t ret;
547
548 if (!(file->f_mode & FMODE_WRITE))
549 return -EBADF;
Al Viro7f7f25e2014-02-11 17:49:24 -0500550 if (!(file->f_mode & FMODE_CAN_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 return -EINVAL;
552 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
553 return -EFAULT;
554
555 ret = rw_verify_area(WRITE, file, pos, count);
Al Virobc613842016-03-31 21:48:20 -0400556 if (!ret) {
557 if (count > MAX_RW_COUNT)
558 count = MAX_RW_COUNT;
Al Viro03d95eb2013-03-20 13:04:20 -0400559 file_start_write(file);
Al Viro493c84c2015-04-03 15:06:43 -0400560 ret = __vfs_write(file, buf, count, pos);
James Morrisc43e2592008-01-12 22:05:48 +1100561 if (ret > 0) {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500562 fsnotify_modify(file);
James Morrisc43e2592008-01-12 22:05:48 +1100563 add_wchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 }
James Morrisc43e2592008-01-12 22:05:48 +1100565 inc_syscw(current);
Al Viro03d95eb2013-03-20 13:04:20 -0400566 file_end_write(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 }
568
569 return ret;
570}
571
572EXPORT_SYMBOL(vfs_write);
573
574static inline loff_t file_pos_read(struct file *file)
575{
576 return file->f_pos;
577}
578
579static inline void file_pos_write(struct file *file, loff_t pos)
580{
581 file->f_pos = pos;
582}
583
Heiko Carstens3cdad422009-01-14 14:14:22 +0100584SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800586 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
Al Viro2903ff02012-08-28 12:52:22 -0400589 if (f.file) {
590 loff_t pos = file_pos_read(f.file);
591 ret = vfs_read(f.file, buf, count, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400592 if (ret >= 0)
593 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800594 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 return ret;
597}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
Heiko Carstens3cdad422009-01-14 14:14:22 +0100599SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
600 size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800602 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
Al Viro2903ff02012-08-28 12:52:22 -0400605 if (f.file) {
606 loff_t pos = file_pos_read(f.file);
607 ret = vfs_write(f.file, buf, count, &pos);
Al Viro5faf1532013-06-15 05:49:36 +0400608 if (ret >= 0)
609 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800610 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 }
612
613 return ret;
614}
615
Al Viro4a0fd5b2013-01-21 15:16:58 -0500616SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
617 size_t, count, loff_t, pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618{
Al Viro2903ff02012-08-28 12:52:22 -0400619 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622 if (pos < 0)
623 return -EINVAL;
624
Al Viro2903ff02012-08-28 12:52:22 -0400625 f = fdget(fd);
626 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400628 if (f.file->f_mode & FMODE_PREAD)
629 ret = vfs_read(f.file, buf, count, &pos);
630 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 }
632
633 return ret;
634}
635
Al Viro4a0fd5b2013-01-21 15:16:58 -0500636SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
637 size_t, count, loff_t, pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
Al Viro2903ff02012-08-28 12:52:22 -0400639 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
642 if (pos < 0)
643 return -EINVAL;
644
Al Viro2903ff02012-08-28 12:52:22 -0400645 f = fdget(fd);
646 if (f.file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400648 if (f.file->f_mode & FMODE_PWRITE)
649 ret = vfs_write(f.file, buf, count, &pos);
650 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 }
652
653 return ret;
654}
655
656/*
657 * Reduce an iovec's length in-place. Return the resulting number of segments
658 */
659unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
660{
661 unsigned long seg = 0;
662 size_t len = 0;
663
664 while (seg < nr_segs) {
665 seg++;
666 if (len + iov->iov_len >= to) {
667 iov->iov_len = to - len;
668 break;
669 }
670 len += iov->iov_len;
671 iov++;
672 }
673 return seg;
674}
Eric Sandeen19295522008-01-28 23:58:27 -0500675EXPORT_SYMBOL(iov_shorten);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Al Viroac15ac02015-03-20 20:10:21 -0400677static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter,
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100678 loff_t *ppos, iter_fn_t fn, int flags)
Al Viro293bc982014-02-11 18:37:41 -0500679{
680 struct kiocb kiocb;
Al Viro293bc982014-02-11 18:37:41 -0500681 ssize_t ret;
682
Christoph Hellwige864f392016-04-07 08:52:03 -0700683 if (flags & ~(RWF_HIPRI | RWF_DSYNC | RWF_SYNC))
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100684 return -EOPNOTSUPP;
685
Al Viro293bc982014-02-11 18:37:41 -0500686 init_sync_kiocb(&kiocb, filp);
Christoph Hellwig97be7eb2016-03-03 16:04:01 +0100687 if (flags & RWF_HIPRI)
688 kiocb.ki_flags |= IOCB_HIPRI;
Christoph Hellwige864f392016-04-07 08:52:03 -0700689 if (flags & RWF_DSYNC)
690 kiocb.ki_flags |= IOCB_DSYNC;
691 if (flags & RWF_SYNC)
692 kiocb.ki_flags |= (IOCB_DSYNC | IOCB_SYNC);
Al Viro293bc982014-02-11 18:37:41 -0500693 kiocb.ki_pos = *ppos;
Al Viro293bc982014-02-11 18:37:41 -0500694
Al Viroac15ac02015-03-20 20:10:21 -0400695 ret = fn(&kiocb, iter);
Christoph Hellwig599bd192015-02-11 19:59:44 +0100696 BUG_ON(ret == -EIOCBQUEUED);
Al Viro293bc982014-02-11 18:37:41 -0500697 *ppos = kiocb.ki_pos;
698 return ret;
699}
700
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700701/* Do it by hand, with file-ops */
Al Viroac15ac02015-03-20 20:10:21 -0400702static ssize_t do_loop_readv_writev(struct file *filp, struct iov_iter *iter,
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100703 loff_t *ppos, io_fn_t fn, int flags)
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700704{
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700705 ssize_t ret = 0;
706
Christoph Hellwig97be7eb2016-03-03 16:04:01 +0100707 if (flags & ~RWF_HIPRI)
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100708 return -EOPNOTSUPP;
709
Al Viroac15ac02015-03-20 20:10:21 -0400710 while (iov_iter_count(iter)) {
711 struct iovec iovec = iov_iter_iovec(iter);
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700712 ssize_t nr;
713
Al Viroac15ac02015-03-20 20:10:21 -0400714 nr = fn(filp, iovec.iov_base, iovec.iov_len, ppos);
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700715
716 if (nr < 0) {
717 if (!ret)
718 ret = nr;
719 break;
720 }
721 ret += nr;
Al Viroac15ac02015-03-20 20:10:21 -0400722 if (nr != iovec.iov_len)
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700723 break;
Al Viroac15ac02015-03-20 20:10:21 -0400724 iov_iter_advance(iter, nr);
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700725 }
726
727 return ret;
728}
729
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730/* A write operation does a read from user space and vice versa */
731#define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
732
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700733ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
734 unsigned long nr_segs, unsigned long fast_segs,
735 struct iovec *fast_pointer,
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700736 struct iovec **ret_pointer)
Linus Torvalds435f49a2010-10-29 10:36:49 -0700737{
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700738 unsigned long seg;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700739 ssize_t ret;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700740 struct iovec *iov = fast_pointer;
741
Linus Torvalds435f49a2010-10-29 10:36:49 -0700742 /*
743 * SuS says "The readv() function *may* fail if the iovcnt argument
744 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
745 * traditionally returned zero for zero segments, so...
746 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700747 if (nr_segs == 0) {
748 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700749 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700750 }
751
Linus Torvalds435f49a2010-10-29 10:36:49 -0700752 /*
753 * First get the "struct iovec" from user memory and
754 * verify all the pointers
755 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700756 if (nr_segs > UIO_MAXIOV) {
757 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700758 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700759 }
760 if (nr_segs > fast_segs) {
Linus Torvalds435f49a2010-10-29 10:36:49 -0700761 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700762 if (iov == NULL) {
763 ret = -ENOMEM;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700764 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700765 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700766 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700767 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
768 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700769 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700770 }
771
Linus Torvalds435f49a2010-10-29 10:36:49 -0700772 /*
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700773 * According to the Single Unix Specification we should return EINVAL
774 * if an element length is < 0 when cast to ssize_t or if the
775 * total length would overflow the ssize_t return value of the
776 * system call.
Linus Torvalds435f49a2010-10-29 10:36:49 -0700777 *
778 * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
779 * overflow case.
780 */
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700781 ret = 0;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700782 for (seg = 0; seg < nr_segs; seg++) {
783 void __user *buf = iov[seg].iov_base;
784 ssize_t len = (ssize_t)iov[seg].iov_len;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700785
786 /* see if we we're about to use an invalid len or if
787 * it's about to overflow ssize_t */
Linus Torvalds435f49a2010-10-29 10:36:49 -0700788 if (len < 0) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700789 ret = -EINVAL;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700790 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700791 }
Christopher Yeohac34ebb2012-05-31 16:26:42 -0700792 if (type >= 0
Christopher Yeohfcf63402011-10-31 17:06:39 -0700793 && unlikely(!access_ok(vrfy_dir(type), buf, len))) {
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700794 ret = -EFAULT;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700795 goto out;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700796 }
Linus Torvalds435f49a2010-10-29 10:36:49 -0700797 if (len > MAX_RW_COUNT - ret) {
798 len = MAX_RW_COUNT - ret;
799 iov[seg].iov_len = len;
800 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700801 ret += len;
Linus Torvalds435f49a2010-10-29 10:36:49 -0700802 }
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700803out:
804 *ret_pointer = iov;
805 return ret;
806}
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808static ssize_t do_readv_writev(int type, struct file *file,
809 const struct iovec __user * uvector,
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100810 unsigned long nr_segs, loff_t *pos,
811 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 size_t tot_len;
814 struct iovec iovstack[UIO_FASTIOV];
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700815 struct iovec *iov = iovstack;
Al Viroac15ac02015-03-20 20:10:21 -0400816 struct iov_iter iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 io_fn_t fn;
Al Viro293bc982014-02-11 18:37:41 -0500819 iter_fn_t iter_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Al Viro0504c072015-03-21 19:40:11 -0400821 ret = import_iovec(type, uvector, nr_segs,
822 ARRAY_SIZE(iovstack), &iov, &iter);
823 if (ret < 0)
824 return ret;
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700825
Al Viro0504c072015-03-21 19:40:11 -0400826 tot_len = iov_iter_count(&iter);
827 if (!tot_len)
828 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 ret = rw_verify_area(type, file, pos, tot_len);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800830 if (ret < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 goto out;
832
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 if (type == READ) {
834 fn = file->f_op->read;
Al Viro293bc982014-02-11 18:37:41 -0500835 iter_fn = file->f_op->read_iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 } else {
837 fn = (io_fn_t)file->f_op->write;
Al Viro293bc982014-02-11 18:37:41 -0500838 iter_fn = file->f_op->write_iter;
Al Viro03d95eb2013-03-20 13:04:20 -0400839 file_start_write(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 }
841
Al Viro293bc982014-02-11 18:37:41 -0500842 if (iter_fn)
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100843 ret = do_iter_readv_writev(file, &iter, pos, iter_fn, flags);
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700844 else
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100845 ret = do_loop_readv_writev(file, &iter, pos, fn, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
Al Viro03d95eb2013-03-20 13:04:20 -0400847 if (type != READ)
848 file_end_write(file);
849
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850out:
Al Viro0504c072015-03-21 19:40:11 -0400851 kfree(iov);
Robert Love0eeca282005-07-12 17:06:03 -0400852 if ((ret + (type == READ)) > 0) {
853 if (type == READ)
Eric Paris2a12a9d2009-12-17 21:24:21 -0500854 fsnotify_access(file);
Robert Love0eeca282005-07-12 17:06:03 -0400855 else
Eric Paris2a12a9d2009-12-17 21:24:21 -0500856 fsnotify_modify(file);
Robert Love0eeca282005-07-12 17:06:03 -0400857 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859}
860
861ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100862 unsigned long vlen, loff_t *pos, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863{
864 if (!(file->f_mode & FMODE_READ))
865 return -EBADF;
Al Viro7f7f25e2014-02-11 17:49:24 -0500866 if (!(file->f_mode & FMODE_CAN_READ))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 return -EINVAL;
868
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100869 return do_readv_writev(READ, file, vec, vlen, pos, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
872EXPORT_SYMBOL(vfs_readv);
873
874ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100875 unsigned long vlen, loff_t *pos, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876{
877 if (!(file->f_mode & FMODE_WRITE))
878 return -EBADF;
Al Viro7f7f25e2014-02-11 17:49:24 -0500879 if (!(file->f_mode & FMODE_CAN_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 return -EINVAL;
881
Christoph Hellwig793b80e2016-03-03 16:03:58 +0100882 return do_readv_writev(WRITE, file, vec, vlen, pos, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883}
884
885EXPORT_SYMBOL(vfs_writev);
886
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100887static ssize_t do_readv(unsigned long fd, const struct iovec __user *vec,
888 unsigned long vlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800890 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Al Viro2903ff02012-08-28 12:52:22 -0400893 if (f.file) {
894 loff_t pos = file_pos_read(f.file);
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100895 ret = vfs_readv(f.file, vec, vlen, &pos, flags);
Al Viro5faf1532013-06-15 05:49:36 +0400896 if (ret >= 0)
897 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800898 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 }
900
901 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800902 add_rchar(current, ret);
903 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 return ret;
905}
906
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100907static ssize_t do_writev(unsigned long fd, const struct iovec __user *vec,
908 unsigned long vlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909{
Linus Torvalds9c225f22014-03-03 09:36:58 -0800910 struct fd f = fdget_pos(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 ssize_t ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
Al Viro2903ff02012-08-28 12:52:22 -0400913 if (f.file) {
914 loff_t pos = file_pos_read(f.file);
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100915 ret = vfs_writev(f.file, vec, vlen, &pos, flags);
Al Viro5faf1532013-06-15 05:49:36 +0400916 if (ret >= 0)
917 file_pos_write(f.file, pos);
Linus Torvalds9c225f22014-03-03 09:36:58 -0800918 fdput_pos(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 }
920
921 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800922 add_wchar(current, ret);
923 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 return ret;
925}
926
Linus Torvalds601cc112009-04-03 08:03:22 -0700927static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700928{
Linus Torvalds601cc112009-04-03 08:03:22 -0700929#define HALF_LONG_BITS (BITS_PER_LONG / 2)
930 return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
931}
932
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100933static ssize_t do_preadv(unsigned long fd, const struct iovec __user *vec,
934 unsigned long vlen, loff_t pos, int flags)
Linus Torvalds601cc112009-04-03 08:03:22 -0700935{
Al Viro2903ff02012-08-28 12:52:22 -0400936 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700937 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700938
939 if (pos < 0)
940 return -EINVAL;
941
Al Viro2903ff02012-08-28 12:52:22 -0400942 f = fdget(fd);
943 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700944 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400945 if (f.file->f_mode & FMODE_PREAD)
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100946 ret = vfs_readv(f.file, vec, vlen, &pos, flags);
Al Viro2903ff02012-08-28 12:52:22 -0400947 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700948 }
949
950 if (ret > 0)
951 add_rchar(current, ret);
952 inc_syscr(current);
953 return ret;
954}
955
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100956static ssize_t do_pwritev(unsigned long fd, const struct iovec __user *vec,
957 unsigned long vlen, loff_t pos, int flags)
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700958{
Al Viro2903ff02012-08-28 12:52:22 -0400959 struct fd f;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700960 ssize_t ret = -EBADF;
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700961
962 if (pos < 0)
963 return -EINVAL;
964
Al Viro2903ff02012-08-28 12:52:22 -0400965 f = fdget(fd);
966 if (f.file) {
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700967 ret = -ESPIPE;
Al Viro2903ff02012-08-28 12:52:22 -0400968 if (f.file->f_mode & FMODE_PWRITE)
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100969 ret = vfs_writev(f.file, vec, vlen, &pos, flags);
Al Viro2903ff02012-08-28 12:52:22 -0400970 fdput(f);
Gerd Hoffmannf3554f42009-04-02 16:59:23 -0700971 }
972
973 if (ret > 0)
974 add_wchar(current, ret);
975 inc_syscw(current);
976 return ret;
977}
978
Milosz Tanskif17d8b32016-03-03 16:03:59 +0100979SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
980 unsigned long, vlen)
981{
982 return do_readv(fd, vec, vlen, 0);
983}
984
985SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
986 unsigned long, vlen)
987{
988 return do_writev(fd, vec, vlen, 0);
989}
990
991SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
992 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
993{
994 loff_t pos = pos_from_hilo(pos_h, pos_l);
995
996 return do_preadv(fd, vec, vlen, pos, 0);
997}
998
999SYSCALL_DEFINE6(preadv2, unsigned long, fd, const struct iovec __user *, vec,
1000 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
1001 int, flags)
1002{
1003 loff_t pos = pos_from_hilo(pos_h, pos_l);
1004
1005 if (pos == -1)
1006 return do_readv(fd, vec, vlen, flags);
1007
1008 return do_preadv(fd, vec, vlen, pos, flags);
1009}
1010
1011SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
1012 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
1013{
1014 loff_t pos = pos_from_hilo(pos_h, pos_l);
1015
1016 return do_pwritev(fd, vec, vlen, pos, 0);
1017}
1018
1019SYSCALL_DEFINE6(pwritev2, unsigned long, fd, const struct iovec __user *, vec,
1020 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
1021 int, flags)
1022{
1023 loff_t pos = pos_from_hilo(pos_h, pos_l);
1024
1025 if (pos == -1)
1026 return do_writev(fd, vec, vlen, flags);
1027
1028 return do_pwritev(fd, vec, vlen, pos, flags);
1029}
1030
Al Viro72ec3512013-03-20 10:42:10 -04001031#ifdef CONFIG_COMPAT
1032
1033static ssize_t compat_do_readv_writev(int type, struct file *file,
1034 const struct compat_iovec __user *uvector,
Christoph Hellwig793b80e2016-03-03 16:03:58 +01001035 unsigned long nr_segs, loff_t *pos,
1036 int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001037{
1038 compat_ssize_t tot_len;
1039 struct iovec iovstack[UIO_FASTIOV];
1040 struct iovec *iov = iovstack;
Al Viroac15ac02015-03-20 20:10:21 -04001041 struct iov_iter iter;
Al Viro72ec3512013-03-20 10:42:10 -04001042 ssize_t ret;
1043 io_fn_t fn;
Al Viro293bc982014-02-11 18:37:41 -05001044 iter_fn_t iter_fn;
Al Viro72ec3512013-03-20 10:42:10 -04001045
Al Viro0504c072015-03-21 19:40:11 -04001046 ret = compat_import_iovec(type, uvector, nr_segs,
1047 UIO_FASTIOV, &iov, &iter);
1048 if (ret < 0)
1049 return ret;
Al Viro72ec3512013-03-20 10:42:10 -04001050
Al Viro0504c072015-03-21 19:40:11 -04001051 tot_len = iov_iter_count(&iter);
1052 if (!tot_len)
1053 goto out;
Al Viro72ec3512013-03-20 10:42:10 -04001054 ret = rw_verify_area(type, file, pos, tot_len);
1055 if (ret < 0)
1056 goto out;
1057
Al Viro72ec3512013-03-20 10:42:10 -04001058 if (type == READ) {
1059 fn = file->f_op->read;
Al Viro293bc982014-02-11 18:37:41 -05001060 iter_fn = file->f_op->read_iter;
Al Viro72ec3512013-03-20 10:42:10 -04001061 } else {
1062 fn = (io_fn_t)file->f_op->write;
Al Viro293bc982014-02-11 18:37:41 -05001063 iter_fn = file->f_op->write_iter;
Al Viro03d95eb2013-03-20 13:04:20 -04001064 file_start_write(file);
Al Viro72ec3512013-03-20 10:42:10 -04001065 }
1066
Al Viro293bc982014-02-11 18:37:41 -05001067 if (iter_fn)
Christoph Hellwig793b80e2016-03-03 16:03:58 +01001068 ret = do_iter_readv_writev(file, &iter, pos, iter_fn, flags);
Al Viro03d95eb2013-03-20 13:04:20 -04001069 else
Christoph Hellwig793b80e2016-03-03 16:03:58 +01001070 ret = do_loop_readv_writev(file, &iter, pos, fn, flags);
Al Viro72ec3512013-03-20 10:42:10 -04001071
Al Viro03d95eb2013-03-20 13:04:20 -04001072 if (type != READ)
1073 file_end_write(file);
1074
Al Viro72ec3512013-03-20 10:42:10 -04001075out:
Al Viro0504c072015-03-21 19:40:11 -04001076 kfree(iov);
Al Viro72ec3512013-03-20 10:42:10 -04001077 if ((ret + (type == READ)) > 0) {
1078 if (type == READ)
1079 fsnotify_access(file);
1080 else
1081 fsnotify_modify(file);
1082 }
1083 return ret;
1084}
1085
1086static size_t compat_readv(struct file *file,
1087 const struct compat_iovec __user *vec,
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001088 unsigned long vlen, loff_t *pos, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001089{
1090 ssize_t ret = -EBADF;
1091
1092 if (!(file->f_mode & FMODE_READ))
1093 goto out;
1094
1095 ret = -EINVAL;
Al Viro7f7f25e2014-02-11 17:49:24 -05001096 if (!(file->f_mode & FMODE_CAN_READ))
Al Viro72ec3512013-03-20 10:42:10 -04001097 goto out;
1098
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001099 ret = compat_do_readv_writev(READ, file, vec, vlen, pos, flags);
Al Viro72ec3512013-03-20 10:42:10 -04001100
1101out:
1102 if (ret > 0)
1103 add_rchar(current, ret);
1104 inc_syscr(current);
1105 return ret;
1106}
1107
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001108static size_t do_compat_readv(compat_ulong_t fd,
1109 const struct compat_iovec __user *vec,
1110 compat_ulong_t vlen, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001111{
Linus Torvalds9c225f22014-03-03 09:36:58 -08001112 struct fd f = fdget_pos(fd);
Al Viro72ec3512013-03-20 10:42:10 -04001113 ssize_t ret;
1114 loff_t pos;
1115
1116 if (!f.file)
1117 return -EBADF;
1118 pos = f.file->f_pos;
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001119 ret = compat_readv(f.file, vec, vlen, &pos, flags);
Al Viro5faf1532013-06-15 05:49:36 +04001120 if (ret >= 0)
1121 f.file->f_pos = pos;
Linus Torvalds9c225f22014-03-03 09:36:58 -08001122 fdput_pos(f);
Al Viro72ec3512013-03-20 10:42:10 -04001123 return ret;
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001124
Al Viro72ec3512013-03-20 10:42:10 -04001125}
1126
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001127COMPAT_SYSCALL_DEFINE3(readv, compat_ulong_t, fd,
1128 const struct compat_iovec __user *,vec,
1129 compat_ulong_t, vlen)
1130{
1131 return do_compat_readv(fd, vec, vlen, 0);
1132}
1133
1134static long do_compat_preadv64(unsigned long fd,
Heiko Carstens378a10f2014-03-05 10:43:51 +01001135 const struct compat_iovec __user *vec,
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001136 unsigned long vlen, loff_t pos, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001137{
1138 struct fd f;
1139 ssize_t ret;
1140
1141 if (pos < 0)
1142 return -EINVAL;
1143 f = fdget(fd);
1144 if (!f.file)
1145 return -EBADF;
1146 ret = -ESPIPE;
1147 if (f.file->f_mode & FMODE_PREAD)
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001148 ret = compat_readv(f.file, vec, vlen, &pos, flags);
Al Viro72ec3512013-03-20 10:42:10 -04001149 fdput(f);
1150 return ret;
1151}
1152
Heiko Carstens378a10f2014-03-05 10:43:51 +01001153#ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1154COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
1155 const struct compat_iovec __user *,vec,
1156 unsigned long, vlen, loff_t, pos)
1157{
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001158 return do_compat_preadv64(fd, vec, vlen, pos, 0);
Heiko Carstens378a10f2014-03-05 10:43:51 +01001159}
1160#endif
1161
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001162COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -04001163 const struct compat_iovec __user *,vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001164 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
Al Viro72ec3512013-03-20 10:42:10 -04001165{
1166 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
Heiko Carstens378a10f2014-03-05 10:43:51 +01001167
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001168 return do_compat_preadv64(fd, vec, vlen, pos, 0);
1169}
1170
1171COMPAT_SYSCALL_DEFINE6(preadv2, compat_ulong_t, fd,
1172 const struct compat_iovec __user *,vec,
1173 compat_ulong_t, vlen, u32, pos_low, u32, pos_high,
1174 int, flags)
1175{
1176 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1177
1178 if (pos == -1)
1179 return do_compat_readv(fd, vec, vlen, flags);
1180
1181 return do_compat_preadv64(fd, vec, vlen, pos, flags);
Al Viro72ec3512013-03-20 10:42:10 -04001182}
1183
1184static size_t compat_writev(struct file *file,
1185 const struct compat_iovec __user *vec,
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001186 unsigned long vlen, loff_t *pos, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001187{
1188 ssize_t ret = -EBADF;
1189
1190 if (!(file->f_mode & FMODE_WRITE))
1191 goto out;
1192
1193 ret = -EINVAL;
Al Viro7f7f25e2014-02-11 17:49:24 -05001194 if (!(file->f_mode & FMODE_CAN_WRITE))
Al Viro72ec3512013-03-20 10:42:10 -04001195 goto out;
1196
Christoph Hellwig793b80e2016-03-03 16:03:58 +01001197 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos, 0);
Al Viro72ec3512013-03-20 10:42:10 -04001198
1199out:
1200 if (ret > 0)
1201 add_wchar(current, ret);
1202 inc_syscw(current);
1203 return ret;
1204}
1205
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001206static size_t do_compat_writev(compat_ulong_t fd,
1207 const struct compat_iovec __user* vec,
1208 compat_ulong_t vlen, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001209{
Linus Torvalds9c225f22014-03-03 09:36:58 -08001210 struct fd f = fdget_pos(fd);
Al Viro72ec3512013-03-20 10:42:10 -04001211 ssize_t ret;
1212 loff_t pos;
1213
1214 if (!f.file)
1215 return -EBADF;
1216 pos = f.file->f_pos;
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001217 ret = compat_writev(f.file, vec, vlen, &pos, flags);
Al Viro5faf1532013-06-15 05:49:36 +04001218 if (ret >= 0)
1219 f.file->f_pos = pos;
Linus Torvalds9c225f22014-03-03 09:36:58 -08001220 fdput_pos(f);
Al Viro72ec3512013-03-20 10:42:10 -04001221 return ret;
1222}
1223
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001224COMPAT_SYSCALL_DEFINE3(writev, compat_ulong_t, fd,
1225 const struct compat_iovec __user *, vec,
1226 compat_ulong_t, vlen)
1227{
1228 return do_compat_writev(fd, vec, vlen, 0);
1229}
1230
1231static long do_compat_pwritev64(unsigned long fd,
Heiko Carstens378a10f2014-03-05 10:43:51 +01001232 const struct compat_iovec __user *vec,
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001233 unsigned long vlen, loff_t pos, int flags)
Al Viro72ec3512013-03-20 10:42:10 -04001234{
1235 struct fd f;
1236 ssize_t ret;
1237
1238 if (pos < 0)
1239 return -EINVAL;
1240 f = fdget(fd);
1241 if (!f.file)
1242 return -EBADF;
1243 ret = -ESPIPE;
1244 if (f.file->f_mode & FMODE_PWRITE)
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001245 ret = compat_writev(f.file, vec, vlen, &pos, flags);
Al Viro72ec3512013-03-20 10:42:10 -04001246 fdput(f);
1247 return ret;
1248}
1249
Heiko Carstens378a10f2014-03-05 10:43:51 +01001250#ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1251COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
1252 const struct compat_iovec __user *,vec,
1253 unsigned long, vlen, loff_t, pos)
1254{
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001255 return do_compat_pwritev64(fd, vec, vlen, pos, 0);
Heiko Carstens378a10f2014-03-05 10:43:51 +01001256}
1257#endif
1258
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001259COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd,
Al Viro72ec3512013-03-20 10:42:10 -04001260 const struct compat_iovec __user *,vec,
Heiko Carstensdfd948e2014-01-29 14:05:44 -08001261 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
Al Viro72ec3512013-03-20 10:42:10 -04001262{
1263 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
Heiko Carstens378a10f2014-03-05 10:43:51 +01001264
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001265 return do_compat_pwritev64(fd, vec, vlen, pos, 0);
Al Viro72ec3512013-03-20 10:42:10 -04001266}
Milosz Tanskif17d8b32016-03-03 16:03:59 +01001267
1268COMPAT_SYSCALL_DEFINE6(pwritev2, compat_ulong_t, fd,
1269 const struct compat_iovec __user *,vec,
1270 compat_ulong_t, vlen, u32, pos_low, u32, pos_high, int, flags)
1271{
1272 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1273
1274 if (pos == -1)
1275 return do_compat_writev(fd, vec, vlen, flags);
1276
1277 return do_compat_pwritev64(fd, vec, vlen, pos, flags);
1278}
1279
Al Viro72ec3512013-03-20 10:42:10 -04001280#endif
1281
Al Viro19f4fc32013-02-24 02:17:03 -05001282static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
1283 size_t count, loff_t max)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284{
Al Viro2903ff02012-08-28 12:52:22 -04001285 struct fd in, out;
1286 struct inode *in_inode, *out_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 loff_t pos;
Al Viro7995bd22013-06-20 18:58:36 +04001288 loff_t out_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 ssize_t retval;
Al Viro2903ff02012-08-28 12:52:22 -04001290 int fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291
1292 /*
1293 * Get input file, and verify that it is ok..
1294 */
1295 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001296 in = fdget(in_fd);
1297 if (!in.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 goto out;
Al Viro2903ff02012-08-28 12:52:22 -04001299 if (!(in.file->f_mode & FMODE_READ))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 goto fput_in;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 retval = -ESPIPE;
Al Viro7995bd22013-06-20 18:58:36 +04001302 if (!ppos) {
1303 pos = in.file->f_pos;
1304 } else {
1305 pos = *ppos;
Al Viro2903ff02012-08-28 12:52:22 -04001306 if (!(in.file->f_mode & FMODE_PREAD))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 goto fput_in;
Al Viro7995bd22013-06-20 18:58:36 +04001308 }
1309 retval = rw_verify_area(READ, in.file, &pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -08001310 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 goto fput_in;
Al Virobc613842016-03-31 21:48:20 -04001312 if (count > MAX_RW_COUNT)
1313 count = MAX_RW_COUNT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 /*
1316 * Get output file, and verify that it is ok..
1317 */
1318 retval = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -04001319 out = fdget(out_fd);
1320 if (!out.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 goto fput_in;
Al Viro2903ff02012-08-28 12:52:22 -04001322 if (!(out.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 goto fput_out;
1324 retval = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -05001325 in_inode = file_inode(in.file);
1326 out_inode = file_inode(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001327 out_pos = out.file->f_pos;
1328 retval = rw_verify_area(WRITE, out.file, &out_pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -08001329 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 goto fput_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 if (!max)
1333 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 if (unlikely(pos + count > max)) {
1336 retval = -EOVERFLOW;
1337 if (pos >= max)
1338 goto fput_out;
1339 count = max - pos;
1340 }
1341
Jens Axboed96e6e72007-06-11 12:18:52 +02001342 fl = 0;
Jens Axboe534f2aa2007-06-01 14:52:37 +02001343#if 0
Jens Axboed96e6e72007-06-11 12:18:52 +02001344 /*
1345 * We need to debate whether we can enable this or not. The
1346 * man page documents EAGAIN return for the output at least,
1347 * and the application is arguably buggy if it doesn't expect
1348 * EAGAIN on a non-blocking file descriptor.
1349 */
Al Viro2903ff02012-08-28 12:52:22 -04001350 if (in.file->f_flags & O_NONBLOCK)
Jens Axboed96e6e72007-06-11 12:18:52 +02001351 fl = SPLICE_F_NONBLOCK;
Jens Axboe534f2aa2007-06-01 14:52:37 +02001352#endif
Al Viro50cd2c52013-05-23 20:10:34 -04001353 file_start_write(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001354 retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl);
Al Viro50cd2c52013-05-23 20:10:34 -04001355 file_end_write(out.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
1357 if (retval > 0) {
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08001358 add_rchar(current, retval);
1359 add_wchar(current, retval);
Scott Wolchoka68c2f12012-12-20 15:05:52 -08001360 fsnotify_access(in.file);
1361 fsnotify_modify(out.file);
Al Viro7995bd22013-06-20 18:58:36 +04001362 out.file->f_pos = out_pos;
1363 if (ppos)
1364 *ppos = pos;
1365 else
1366 in.file->f_pos = pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08001369 inc_syscr(current);
1370 inc_syscw(current);
Al Viro7995bd22013-06-20 18:58:36 +04001371 if (pos > max)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 retval = -EOVERFLOW;
1373
1374fput_out:
Al Viro2903ff02012-08-28 12:52:22 -04001375 fdput(out);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376fput_in:
Al Viro2903ff02012-08-28 12:52:22 -04001377 fdput(in);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378out:
1379 return retval;
1380}
1381
Heiko Carstens002c8972009-01-14 14:14:18 +01001382SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383{
1384 loff_t pos;
1385 off_t off;
1386 ssize_t ret;
1387
1388 if (offset) {
1389 if (unlikely(get_user(off, offset)))
1390 return -EFAULT;
1391 pos = off;
1392 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1393 if (unlikely(put_user(pos, offset)))
1394 return -EFAULT;
1395 return ret;
1396 }
1397
1398 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1399}
1400
Heiko Carstens002c8972009-01-14 14:14:18 +01001401SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402{
1403 loff_t pos;
1404 ssize_t ret;
1405
1406 if (offset) {
1407 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1408 return -EFAULT;
1409 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1410 if (unlikely(put_user(pos, offset)))
1411 return -EFAULT;
1412 return ret;
1413 }
1414
1415 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1416}
Al Viro19f4fc32013-02-24 02:17:03 -05001417
1418#ifdef CONFIG_COMPAT
1419COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
1420 compat_off_t __user *, offset, compat_size_t, count)
1421{
1422 loff_t pos;
1423 off_t off;
1424 ssize_t ret;
1425
1426 if (offset) {
1427 if (unlikely(get_user(off, offset)))
1428 return -EFAULT;
1429 pos = off;
1430 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1431 if (unlikely(put_user(pos, offset)))
1432 return -EFAULT;
1433 return ret;
1434 }
1435
1436 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1437}
1438
1439COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
1440 compat_loff_t __user *, offset, compat_size_t, count)
1441{
1442 loff_t pos;
1443 ssize_t ret;
1444
1445 if (offset) {
1446 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1447 return -EFAULT;
1448 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1449 if (unlikely(put_user(pos, offset)))
1450 return -EFAULT;
1451 return ret;
1452 }
1453
1454 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1455}
1456#endif
Zach Brown29732932015-11-10 16:53:30 -05001457
1458/*
1459 * copy_file_range() differs from regular file read and write in that it
1460 * specifically allows return partial success. When it does so is up to
1461 * the copy_file_range method.
1462 */
1463ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
1464 struct file *file_out, loff_t pos_out,
1465 size_t len, unsigned int flags)
1466{
1467 struct inode *inode_in = file_inode(file_in);
1468 struct inode *inode_out = file_inode(file_out);
1469 ssize_t ret;
1470
1471 if (flags != 0)
1472 return -EINVAL;
1473
Zach Brown29732932015-11-10 16:53:30 -05001474 ret = rw_verify_area(READ, file_in, &pos_in, len);
Al Virobc613842016-03-31 21:48:20 -04001475 if (unlikely(ret))
1476 return ret;
1477
1478 ret = rw_verify_area(WRITE, file_out, &pos_out, len);
1479 if (unlikely(ret))
Zach Brown29732932015-11-10 16:53:30 -05001480 return ret;
1481
1482 if (!(file_in->f_mode & FMODE_READ) ||
1483 !(file_out->f_mode & FMODE_WRITE) ||
Anna Schumakereac70052015-11-10 16:53:33 -05001484 (file_out->f_flags & O_APPEND))
Zach Brown29732932015-11-10 16:53:30 -05001485 return -EBADF;
1486
1487 /* this could be relaxed once a method supports cross-fs copies */
1488 if (inode_in->i_sb != inode_out->i_sb)
1489 return -EXDEV;
1490
1491 if (len == 0)
1492 return 0;
1493
1494 ret = mnt_want_write_file(file_out);
1495 if (ret)
1496 return ret;
1497
Anna Schumakereac70052015-11-10 16:53:33 -05001498 ret = -EOPNOTSUPP;
1499 if (file_out->f_op->copy_file_range)
1500 ret = file_out->f_op->copy_file_range(file_in, pos_in, file_out,
1501 pos_out, len, flags);
1502 if (ret == -EOPNOTSUPP)
1503 ret = do_splice_direct(file_in, &pos_in, file_out, &pos_out,
1504 len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
1505
Zach Brown29732932015-11-10 16:53:30 -05001506 if (ret > 0) {
1507 fsnotify_access(file_in);
1508 add_rchar(current, ret);
1509 fsnotify_modify(file_out);
1510 add_wchar(current, ret);
1511 }
1512 inc_syscr(current);
1513 inc_syscw(current);
1514
1515 mnt_drop_write_file(file_out);
1516
1517 return ret;
1518}
1519EXPORT_SYMBOL(vfs_copy_file_range);
1520
1521SYSCALL_DEFINE6(copy_file_range, int, fd_in, loff_t __user *, off_in,
1522 int, fd_out, loff_t __user *, off_out,
1523 size_t, len, unsigned int, flags)
1524{
1525 loff_t pos_in;
1526 loff_t pos_out;
1527 struct fd f_in;
1528 struct fd f_out;
1529 ssize_t ret = -EBADF;
1530
1531 f_in = fdget(fd_in);
1532 if (!f_in.file)
1533 goto out2;
1534
1535 f_out = fdget(fd_out);
1536 if (!f_out.file)
1537 goto out1;
1538
1539 ret = -EFAULT;
1540 if (off_in) {
1541 if (copy_from_user(&pos_in, off_in, sizeof(loff_t)))
1542 goto out;
1543 } else {
1544 pos_in = f_in.file->f_pos;
1545 }
1546
1547 if (off_out) {
1548 if (copy_from_user(&pos_out, off_out, sizeof(loff_t)))
1549 goto out;
1550 } else {
1551 pos_out = f_out.file->f_pos;
1552 }
1553
1554 ret = vfs_copy_file_range(f_in.file, pos_in, f_out.file, pos_out, len,
1555 flags);
1556 if (ret > 0) {
1557 pos_in += ret;
1558 pos_out += ret;
1559
1560 if (off_in) {
1561 if (copy_to_user(off_in, &pos_in, sizeof(loff_t)))
1562 ret = -EFAULT;
1563 } else {
1564 f_in.file->f_pos = pos_in;
1565 }
1566
1567 if (off_out) {
1568 if (copy_to_user(off_out, &pos_out, sizeof(loff_t)))
1569 ret = -EFAULT;
1570 } else {
1571 f_out.file->f_pos = pos_out;
1572 }
1573 }
1574
1575out:
1576 fdput(f_out);
1577out1:
1578 fdput(f_in);
1579out2:
1580 return ret;
1581}
Christoph Hellwig04b38d62015-12-03 12:59:50 +01001582
1583static int clone_verify_area(struct file *file, loff_t pos, u64 len, bool write)
1584{
1585 struct inode *inode = file_inode(file);
1586
1587 if (unlikely(pos < 0))
1588 return -EINVAL;
1589
1590 if (unlikely((loff_t) (pos + len) < 0))
1591 return -EINVAL;
1592
1593 if (unlikely(inode->i_flctx && mandatory_lock(inode))) {
1594 loff_t end = len ? pos + len - 1 : OFFSET_MAX;
1595 int retval;
1596
1597 retval = locks_mandatory_area(inode, file, pos, end,
1598 write ? F_WRLCK : F_RDLCK);
1599 if (retval < 0)
1600 return retval;
1601 }
1602
1603 return security_file_permission(file, write ? MAY_WRITE : MAY_READ);
1604}
1605
1606int vfs_clone_file_range(struct file *file_in, loff_t pos_in,
1607 struct file *file_out, loff_t pos_out, u64 len)
1608{
1609 struct inode *inode_in = file_inode(file_in);
1610 struct inode *inode_out = file_inode(file_out);
1611 int ret;
1612
1613 if (inode_in->i_sb != inode_out->i_sb ||
1614 file_in->f_path.mnt != file_out->f_path.mnt)
1615 return -EXDEV;
1616
1617 if (S_ISDIR(inode_in->i_mode) || S_ISDIR(inode_out->i_mode))
1618 return -EISDIR;
1619 if (!S_ISREG(inode_in->i_mode) || !S_ISREG(inode_out->i_mode))
Darrick J. Wongd79bdd52015-12-19 00:55:52 -08001620 return -EINVAL;
Christoph Hellwig04b38d62015-12-03 12:59:50 +01001621
1622 if (!(file_in->f_mode & FMODE_READ) ||
1623 !(file_out->f_mode & FMODE_WRITE) ||
Christoph Hellwig0fcbf992016-02-26 18:53:12 +01001624 (file_out->f_flags & O_APPEND))
Christoph Hellwig04b38d62015-12-03 12:59:50 +01001625 return -EBADF;
1626
Christoph Hellwig0fcbf992016-02-26 18:53:12 +01001627 if (!file_in->f_op->clone_file_range)
1628 return -EOPNOTSUPP;
1629
Christoph Hellwig04b38d62015-12-03 12:59:50 +01001630 ret = clone_verify_area(file_in, pos_in, len, false);
1631 if (ret)
1632 return ret;
1633
1634 ret = clone_verify_area(file_out, pos_out, len, true);
1635 if (ret)
1636 return ret;
1637
1638 if (pos_in + len > i_size_read(inode_in))
1639 return -EINVAL;
1640
1641 ret = mnt_want_write_file(file_out);
1642 if (ret)
1643 return ret;
1644
1645 ret = file_in->f_op->clone_file_range(file_in, pos_in,
1646 file_out, pos_out, len);
1647 if (!ret) {
1648 fsnotify_access(file_in);
1649 fsnotify_modify(file_out);
1650 }
1651
1652 mnt_drop_write_file(file_out);
1653 return ret;
1654}
1655EXPORT_SYMBOL(vfs_clone_file_range);
Darrick J. Wong54dbc152015-12-19 00:55:59 -08001656
1657int vfs_dedupe_file_range(struct file *file, struct file_dedupe_range *same)
1658{
1659 struct file_dedupe_range_info *info;
1660 struct inode *src = file_inode(file);
1661 u64 off;
1662 u64 len;
1663 int i;
1664 int ret;
1665 bool is_admin = capable(CAP_SYS_ADMIN);
1666 u16 count = same->dest_count;
1667 struct file *dst_file;
1668 loff_t dst_off;
1669 ssize_t deduped;
1670
1671 if (!(file->f_mode & FMODE_READ))
1672 return -EINVAL;
1673
1674 if (same->reserved1 || same->reserved2)
1675 return -EINVAL;
1676
1677 off = same->src_offset;
1678 len = same->src_length;
1679
1680 ret = -EISDIR;
1681 if (S_ISDIR(src->i_mode))
1682 goto out;
1683
1684 ret = -EINVAL;
1685 if (!S_ISREG(src->i_mode))
1686 goto out;
1687
1688 ret = clone_verify_area(file, off, len, false);
1689 if (ret < 0)
1690 goto out;
1691 ret = 0;
1692
1693 /* pre-format output fields to sane values */
1694 for (i = 0; i < count; i++) {
1695 same->info[i].bytes_deduped = 0ULL;
1696 same->info[i].status = FILE_DEDUPE_RANGE_SAME;
1697 }
1698
1699 for (i = 0, info = same->info; i < count; i++, info++) {
1700 struct inode *dst;
1701 struct fd dst_fd = fdget(info->dest_fd);
1702
1703 dst_file = dst_fd.file;
1704 if (!dst_file) {
1705 info->status = -EBADF;
1706 goto next_loop;
1707 }
1708 dst = file_inode(dst_file);
1709
1710 ret = mnt_want_write_file(dst_file);
1711 if (ret) {
1712 info->status = ret;
1713 goto next_loop;
1714 }
1715
1716 dst_off = info->dest_offset;
1717 ret = clone_verify_area(dst_file, dst_off, len, true);
1718 if (ret < 0) {
1719 info->status = ret;
1720 goto next_file;
1721 }
1722 ret = 0;
1723
1724 if (info->reserved) {
1725 info->status = -EINVAL;
1726 } else if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
1727 info->status = -EINVAL;
1728 } else if (file->f_path.mnt != dst_file->f_path.mnt) {
1729 info->status = -EXDEV;
1730 } else if (S_ISDIR(dst->i_mode)) {
1731 info->status = -EISDIR;
1732 } else if (dst_file->f_op->dedupe_file_range == NULL) {
1733 info->status = -EINVAL;
1734 } else {
1735 deduped = dst_file->f_op->dedupe_file_range(file, off,
1736 len, dst_file,
1737 info->dest_offset);
1738 if (deduped == -EBADE)
1739 info->status = FILE_DEDUPE_RANGE_DIFFERS;
1740 else if (deduped < 0)
1741 info->status = deduped;
1742 else
1743 info->bytes_deduped += deduped;
1744 }
1745
1746next_file:
1747 mnt_drop_write_file(dst_file);
1748next_loop:
1749 fdput(dst_fd);
Darrick J. Wonge62e5602016-01-22 16:58:28 -08001750
1751 if (fatal_signal_pending(current))
1752 goto out;
Darrick J. Wong54dbc152015-12-19 00:55:59 -08001753 }
1754
1755out:
1756 return ret;
1757}
1758EXPORT_SYMBOL(vfs_dedupe_file_range);