Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/read_write.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
b12fb7f4 | 8 | #include <linux/slab.h> |
1da177e4 | 9 | #include <linux/stat.h> |
b12fb7f4 | 10 | #include <linux/sched/xacct.h> |
1da177e4 LT |
11 | #include <linux/fcntl.h> |
12 | #include <linux/file.h> | |
13 | #include <linux/uio.h> | |
0eeca283 | 14 | #include <linux/fsnotify.h> |
1da177e4 | 15 | #include <linux/security.h> |
630d9c47 | 16 | #include <linux/export.h> |
1da177e4 | 17 | #include <linux/syscalls.h> |
e28cc715 | 18 | #include <linux/pagemap.h> |
d6b29d7c | 19 | #include <linux/splice.h> |
561c6731 | 20 | #include <linux/compat.h> |
29732938 | 21 | #include <linux/mount.h> |
2feb55f8 | 22 | #include <linux/fs.h> |
06ae43f3 | 23 | #include "internal.h" |
1da177e4 | 24 | |
7c0f6ba6 | 25 | #include <linux/uaccess.h> |
1da177e4 LT |
26 | #include <asm/unistd.h> |
27 | ||
4b6f5d20 | 28 | const struct file_operations generic_ro_fops = { |
1da177e4 | 29 | .llseek = generic_file_llseek, |
aad4f8bb | 30 | .read_iter = generic_file_read_iter, |
1da177e4 | 31 | .mmap = generic_file_readonly_mmap, |
534f2aaa | 32 | .splice_read = generic_file_splice_read, |
1da177e4 LT |
33 | }; |
34 | ||
35 | EXPORT_SYMBOL(generic_ro_fops); | |
36 | ||
ddef7ed2 | 37 | static inline bool unsigned_offsets(struct file *file) |
4a3956c7 | 38 | { |
cccb5a1e | 39 | return file->f_mode & FMODE_UNSIGNED_OFFSET; |
4a3956c7 KH |
40 | } |
41 | ||
46a1c2c7 JL |
42 | /** |
43 | * vfs_setpos - update the file offset for lseek | |
44 | * @file: file structure in question | |
45 | * @offset: file offset to seek to | |
46 | * @maxsize: maximum file size | |
47 | * | |
48 | * This is a low-level filesystem helper for updating the file offset to | |
49 | * the value specified by @offset if the given offset is valid and it is | |
50 | * not equal to the current file offset. | |
51 | * | |
52 | * Return the specified offset on success and -EINVAL on invalid offset. | |
53 | */ | |
54 | loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize) | |
ef3d0fd2 AK |
55 | { |
56 | if (offset < 0 && !unsigned_offsets(file)) | |
57 | return -EINVAL; | |
58 | if (offset > maxsize) | |
59 | return -EINVAL; | |
60 | ||
61 | if (offset != file->f_pos) { | |
62 | file->f_pos = offset; | |
63 | file->f_version = 0; | |
64 | } | |
65 | return offset; | |
66 | } | |
46a1c2c7 | 67 | EXPORT_SYMBOL(vfs_setpos); |
ef3d0fd2 | 68 | |
3a8cff4f | 69 | /** |
5760495a | 70 | * generic_file_llseek_size - generic llseek implementation for regular files |
3a8cff4f CH |
71 | * @file: file structure to seek on |
72 | * @offset: file offset to seek to | |
965c8e59 | 73 | * @whence: type of seek |
e8b96eb5 ES |
74 | * @size: max size of this file in file system |
75 | * @eof: offset used for SEEK_END position | |
3a8cff4f | 76 | * |
5760495a | 77 | * This is a variant of generic_file_llseek that allows passing in a custom |
e8b96eb5 | 78 | * maximum file size and a custom EOF position, for e.g. hashed directories |
ef3d0fd2 AK |
79 | * |
80 | * Synchronization: | |
5760495a | 81 | * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms) |
ef3d0fd2 AK |
82 | * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes. |
83 | * read/writes behave like SEEK_SET against seeks. | |
3a8cff4f | 84 | */ |
9465efc9 | 85 | loff_t |
965c8e59 | 86 | generic_file_llseek_size(struct file *file, loff_t offset, int whence, |
e8b96eb5 | 87 | loff_t maxsize, loff_t eof) |
1da177e4 | 88 | { |
965c8e59 | 89 | switch (whence) { |
3a8cff4f | 90 | case SEEK_END: |
e8b96eb5 | 91 | offset += eof; |
3a8cff4f CH |
92 | break; |
93 | case SEEK_CUR: | |
5b6f1eb9 AK |
94 | /* |
95 | * Here we special-case the lseek(fd, 0, SEEK_CUR) | |
96 | * position-querying operation. Avoid rewriting the "same" | |
97 | * f_pos value back to the file because a concurrent read(), | |
98 | * write() or lseek() might have altered it | |
99 | */ | |
100 | if (offset == 0) | |
101 | return file->f_pos; | |
ef3d0fd2 AK |
102 | /* |
103 | * f_lock protects against read/modify/write race with other | |
104 | * SEEK_CURs. Note that parallel writes and reads behave | |
105 | * like SEEK_SET. | |
106 | */ | |
107 | spin_lock(&file->f_lock); | |
46a1c2c7 | 108 | offset = vfs_setpos(file, file->f_pos + offset, maxsize); |
ef3d0fd2 AK |
109 | spin_unlock(&file->f_lock); |
110 | return offset; | |
982d8165 JB |
111 | case SEEK_DATA: |
112 | /* | |
113 | * In the generic case the entire file is data, so as long as | |
114 | * offset isn't at the end of the file then the offset is data. | |
115 | */ | |
fc46820b | 116 | if ((unsigned long long)offset >= eof) |
982d8165 JB |
117 | return -ENXIO; |
118 | break; | |
119 | case SEEK_HOLE: | |
120 | /* | |
121 | * There is a virtual hole at the end of the file, so as long as | |
122 | * offset isn't i_size or larger, return i_size. | |
123 | */ | |
fc46820b | 124 | if ((unsigned long long)offset >= eof) |
982d8165 | 125 | return -ENXIO; |
e8b96eb5 | 126 | offset = eof; |
982d8165 | 127 | break; |
1da177e4 | 128 | } |
3a8cff4f | 129 | |
46a1c2c7 | 130 | return vfs_setpos(file, offset, maxsize); |
5760495a AK |
131 | } |
132 | EXPORT_SYMBOL(generic_file_llseek_size); | |
133 | ||
134 | /** | |
135 | * generic_file_llseek - generic llseek implementation for regular files | |
136 | * @file: file structure to seek on | |
137 | * @offset: file offset to seek to | |
965c8e59 | 138 | * @whence: type of seek |
5760495a AK |
139 | * |
140 | * This is a generic implemenation of ->llseek useable for all normal local | |
141 | * filesystems. It just updates the file offset to the value specified by | |
546ae2d2 | 142 | * @offset and @whence. |
5760495a | 143 | */ |
965c8e59 | 144 | loff_t generic_file_llseek(struct file *file, loff_t offset, int whence) |
5760495a AK |
145 | { |
146 | struct inode *inode = file->f_mapping->host; | |
147 | ||
965c8e59 | 148 | return generic_file_llseek_size(file, offset, whence, |
e8b96eb5 ES |
149 | inode->i_sb->s_maxbytes, |
150 | i_size_read(inode)); | |
1da177e4 | 151 | } |
9465efc9 | 152 | EXPORT_SYMBOL(generic_file_llseek); |
1da177e4 | 153 | |
1bf9d14d AV |
154 | /** |
155 | * fixed_size_llseek - llseek implementation for fixed-sized devices | |
156 | * @file: file structure to seek on | |
157 | * @offset: file offset to seek to | |
158 | * @whence: type of seek | |
159 | * @size: size of the file | |
160 | * | |
161 | */ | |
162 | loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size) | |
163 | { | |
164 | switch (whence) { | |
165 | case SEEK_SET: case SEEK_CUR: case SEEK_END: | |
166 | return generic_file_llseek_size(file, offset, whence, | |
167 | size, size); | |
168 | default: | |
169 | return -EINVAL; | |
170 | } | |
171 | } | |
172 | EXPORT_SYMBOL(fixed_size_llseek); | |
173 | ||
b25472f9 AV |
174 | /** |
175 | * no_seek_end_llseek - llseek implementation for fixed-sized devices | |
176 | * @file: file structure to seek on | |
177 | * @offset: file offset to seek to | |
178 | * @whence: type of seek | |
179 | * | |
180 | */ | |
181 | loff_t no_seek_end_llseek(struct file *file, loff_t offset, int whence) | |
182 | { | |
183 | switch (whence) { | |
184 | case SEEK_SET: case SEEK_CUR: | |
185 | return generic_file_llseek_size(file, offset, whence, | |
2feb55f8 | 186 | OFFSET_MAX, 0); |
b25472f9 AV |
187 | default: |
188 | return -EINVAL; | |
189 | } | |
190 | } | |
191 | EXPORT_SYMBOL(no_seek_end_llseek); | |
192 | ||
193 | /** | |
194 | * no_seek_end_llseek_size - llseek implementation for fixed-sized devices | |
195 | * @file: file structure to seek on | |
196 | * @offset: file offset to seek to | |
197 | * @whence: type of seek | |
198 | * @size: maximal offset allowed | |
199 | * | |
200 | */ | |
201 | loff_t no_seek_end_llseek_size(struct file *file, loff_t offset, int whence, loff_t size) | |
202 | { | |
203 | switch (whence) { | |
204 | case SEEK_SET: case SEEK_CUR: | |
205 | return generic_file_llseek_size(file, offset, whence, | |
206 | size, 0); | |
207 | default: | |
208 | return -EINVAL; | |
209 | } | |
210 | } | |
211 | EXPORT_SYMBOL(no_seek_end_llseek_size); | |
212 | ||
ae6afc3f B |
213 | /** |
214 | * noop_llseek - No Operation Performed llseek implementation | |
215 | * @file: file structure to seek on | |
216 | * @offset: file offset to seek to | |
965c8e59 | 217 | * @whence: type of seek |
ae6afc3f B |
218 | * |
219 | * This is an implementation of ->llseek useable for the rare special case when | |
220 | * userspace expects the seek to succeed but the (device) file is actually not | |
221 | * able to perform the seek. In this case you use noop_llseek() instead of | |
222 | * falling back to the default implementation of ->llseek. | |
223 | */ | |
965c8e59 | 224 | loff_t noop_llseek(struct file *file, loff_t offset, int whence) |
ae6afc3f B |
225 | { |
226 | return file->f_pos; | |
227 | } | |
228 | EXPORT_SYMBOL(noop_llseek); | |
229 | ||
965c8e59 | 230 | loff_t no_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
231 | { |
232 | return -ESPIPE; | |
233 | } | |
234 | EXPORT_SYMBOL(no_llseek); | |
235 | ||
965c8e59 | 236 | loff_t default_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 237 | { |
496ad9aa | 238 | struct inode *inode = file_inode(file); |
16abef0e | 239 | loff_t retval; |
1da177e4 | 240 | |
5955102c | 241 | inode_lock(inode); |
965c8e59 | 242 | switch (whence) { |
7b8e8924 | 243 | case SEEK_END: |
982d8165 | 244 | offset += i_size_read(inode); |
1da177e4 | 245 | break; |
7b8e8924 | 246 | case SEEK_CUR: |
5b6f1eb9 AK |
247 | if (offset == 0) { |
248 | retval = file->f_pos; | |
249 | goto out; | |
250 | } | |
1da177e4 | 251 | offset += file->f_pos; |
982d8165 JB |
252 | break; |
253 | case SEEK_DATA: | |
254 | /* | |
255 | * In the generic case the entire file is data, so as | |
256 | * long as offset isn't at the end of the file then the | |
257 | * offset is data. | |
258 | */ | |
bacb2d81 DC |
259 | if (offset >= inode->i_size) { |
260 | retval = -ENXIO; | |
261 | goto out; | |
262 | } | |
982d8165 JB |
263 | break; |
264 | case SEEK_HOLE: | |
265 | /* | |
266 | * There is a virtual hole at the end of the file, so | |
267 | * as long as offset isn't i_size or larger, return | |
268 | * i_size. | |
269 | */ | |
bacb2d81 DC |
270 | if (offset >= inode->i_size) { |
271 | retval = -ENXIO; | |
272 | goto out; | |
273 | } | |
982d8165 JB |
274 | offset = inode->i_size; |
275 | break; | |
1da177e4 LT |
276 | } |
277 | retval = -EINVAL; | |
cccb5a1e | 278 | if (offset >= 0 || unsigned_offsets(file)) { |
1da177e4 LT |
279 | if (offset != file->f_pos) { |
280 | file->f_pos = offset; | |
281 | file->f_version = 0; | |
282 | } | |
283 | retval = offset; | |
284 | } | |
5b6f1eb9 | 285 | out: |
5955102c | 286 | inode_unlock(inode); |
1da177e4 LT |
287 | return retval; |
288 | } | |
289 | EXPORT_SYMBOL(default_llseek); | |
290 | ||
965c8e59 | 291 | loff_t vfs_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
292 | { |
293 | loff_t (*fn)(struct file *, loff_t, int); | |
294 | ||
295 | fn = no_llseek; | |
296 | if (file->f_mode & FMODE_LSEEK) { | |
72c2d531 | 297 | if (file->f_op->llseek) |
1da177e4 LT |
298 | fn = file->f_op->llseek; |
299 | } | |
965c8e59 | 300 | return fn(file, offset, whence); |
1da177e4 LT |
301 | } |
302 | EXPORT_SYMBOL(vfs_llseek); | |
303 | ||
bef17329 | 304 | static off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence) |
1da177e4 LT |
305 | { |
306 | off_t retval; | |
9c225f26 | 307 | struct fd f = fdget_pos(fd); |
2903ff01 AV |
308 | if (!f.file) |
309 | return -EBADF; | |
1da177e4 LT |
310 | |
311 | retval = -EINVAL; | |
965c8e59 AM |
312 | if (whence <= SEEK_MAX) { |
313 | loff_t res = vfs_llseek(f.file, offset, whence); | |
1da177e4 LT |
314 | retval = res; |
315 | if (res != (loff_t)retval) | |
316 | retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */ | |
317 | } | |
9c225f26 | 318 | fdput_pos(f); |
1da177e4 LT |
319 | return retval; |
320 | } | |
321 | ||
76847e43 DB |
322 | SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence) |
323 | { | |
324 | return ksys_lseek(fd, offset, whence); | |
325 | } | |
326 | ||
561c6731 AV |
327 | #ifdef CONFIG_COMPAT |
328 | COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence) | |
329 | { | |
76847e43 | 330 | return ksys_lseek(fd, offset, whence); |
561c6731 AV |
331 | } |
332 | #endif | |
333 | ||
9e62ccec MS |
334 | #if !defined(CONFIG_64BIT) || defined(CONFIG_COMPAT) || \ |
335 | defined(__ARCH_WANT_SYS_LLSEEK) | |
003d7ab4 HC |
336 | SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high, |
337 | unsigned long, offset_low, loff_t __user *, result, | |
965c8e59 | 338 | unsigned int, whence) |
1da177e4 LT |
339 | { |
340 | int retval; | |
d7a15f8d | 341 | struct fd f = fdget_pos(fd); |
1da177e4 | 342 | loff_t offset; |
1da177e4 | 343 | |
2903ff01 AV |
344 | if (!f.file) |
345 | return -EBADF; | |
1da177e4 LT |
346 | |
347 | retval = -EINVAL; | |
965c8e59 | 348 | if (whence > SEEK_MAX) |
1da177e4 LT |
349 | goto out_putf; |
350 | ||
2903ff01 | 351 | offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low, |
965c8e59 | 352 | whence); |
1da177e4 LT |
353 | |
354 | retval = (int)offset; | |
355 | if (offset >= 0) { | |
356 | retval = -EFAULT; | |
357 | if (!copy_to_user(result, &offset, sizeof(offset))) | |
358 | retval = 0; | |
359 | } | |
360 | out_putf: | |
d7a15f8d | 361 | fdput_pos(f); |
1da177e4 LT |
362 | return retval; |
363 | } | |
364 | #endif | |
365 | ||
68d70d03 | 366 | int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count) |
1da177e4 LT |
367 | { |
368 | struct inode *inode; | |
c43e259c | 369 | int retval = -EINVAL; |
1da177e4 | 370 | |
496ad9aa | 371 | inode = file_inode(file); |
e28cc715 | 372 | if (unlikely((ssize_t) count < 0)) |
c43e259c | 373 | return retval; |
1da177e4 | 374 | |
438ab720 KS |
375 | /* |
376 | * ranged mandatory locking does not apply to streams - it makes sense | |
377 | * only for files where position has a meaning. | |
378 | */ | |
379 | if (ppos) { | |
380 | loff_t pos = *ppos; | |
381 | ||
382 | if (unlikely(pos < 0)) { | |
383 | if (!unsigned_offsets(file)) | |
384 | return retval; | |
385 | if (count >= -pos) /* both values are in 0..LLONG_MAX */ | |
386 | return -EOVERFLOW; | |
387 | } else if (unlikely((loff_t) (pos + count) < 0)) { | |
388 | if (!unsigned_offsets(file)) | |
389 | return retval; | |
390 | } | |
391 | ||
392 | if (unlikely(inode->i_flctx && mandatory_lock(inode))) { | |
393 | retval = locks_mandatory_area(inode, file, pos, pos + count - 1, | |
394 | read_write == READ ? F_RDLCK : F_WRLCK); | |
395 | if (retval < 0) | |
396 | return retval; | |
397 | } | |
e28cc715 | 398 | } |
438ab720 | 399 | |
bc61384d | 400 | return security_file_permission(file, |
c43e259c | 401 | read_write == READ ? MAY_READ : MAY_WRITE); |
1da177e4 LT |
402 | } |
403 | ||
5d5d5689 | 404 | static ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) |
293bc982 AV |
405 | { |
406 | struct iovec iov = { .iov_base = buf, .iov_len = len }; | |
407 | struct kiocb kiocb; | |
408 | struct iov_iter iter; | |
409 | ssize_t ret; | |
410 | ||
411 | init_sync_kiocb(&kiocb, filp); | |
438ab720 | 412 | kiocb.ki_pos = (ppos ? *ppos : 0); |
293bc982 AV |
413 | iov_iter_init(&iter, READ, &iov, 1, len); |
414 | ||
bb7462b6 | 415 | ret = call_read_iter(filp, &kiocb, &iter); |
599bd19b | 416 | BUG_ON(ret == -EIOCBQUEUED); |
438ab720 KS |
417 | if (ppos) |
418 | *ppos = kiocb.ki_pos; | |
293bc982 AV |
419 | return ret; |
420 | } | |
421 | ||
61a707c5 CH |
422 | ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos) |
423 | { | |
424 | mm_segment_t old_fs = get_fs(); | |
425 | ssize_t ret; | |
426 | ||
427 | if (WARN_ON_ONCE(!(file->f_mode & FMODE_READ))) | |
428 | return -EINVAL; | |
429 | if (!(file->f_mode & FMODE_CAN_READ)) | |
430 | return -EINVAL; | |
431 | ||
432 | if (count > MAX_RW_COUNT) | |
433 | count = MAX_RW_COUNT; | |
434 | set_fs(KERNEL_DS); | |
775802c0 CH |
435 | if (file->f_op->read) |
436 | ret = file->f_op->read(file, (void __user *)buf, count, pos); | |
437 | else if (file->f_op->read_iter) | |
438 | ret = new_sync_read(file, (void __user *)buf, count, pos); | |
439 | else | |
440 | ret = -EINVAL; | |
61a707c5 CH |
441 | set_fs(old_fs); |
442 | if (ret > 0) { | |
443 | fsnotify_access(file); | |
444 | add_rchar(current, ret); | |
445 | } | |
446 | inc_syscr(current); | |
447 | return ret; | |
448 | } | |
449 | ||
bdd1d2d3 | 450 | ssize_t kernel_read(struct file *file, void *buf, size_t count, loff_t *pos) |
c41fbad0 | 451 | { |
6209dd91 | 452 | ssize_t ret; |
c41fbad0 | 453 | |
6209dd91 CH |
454 | ret = rw_verify_area(READ, file, pos, count); |
455 | if (ret) | |
456 | return ret; | |
457 | return __kernel_read(file, buf, count, pos); | |
c41fbad0 CH |
458 | } |
459 | EXPORT_SYMBOL(kernel_read); | |
6fb5032e | 460 | |
1da177e4 LT |
461 | ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos) |
462 | { | |
463 | ssize_t ret; | |
464 | ||
465 | if (!(file->f_mode & FMODE_READ)) | |
466 | return -EBADF; | |
7f7f25e8 | 467 | if (!(file->f_mode & FMODE_CAN_READ)) |
1da177e4 | 468 | return -EINVAL; |
96d4f267 | 469 | if (unlikely(!access_ok(buf, count))) |
1da177e4 LT |
470 | return -EFAULT; |
471 | ||
472 | ret = rw_verify_area(READ, file, pos, count); | |
775802c0 CH |
473 | if (ret) |
474 | return ret; | |
475 | if (count > MAX_RW_COUNT) | |
476 | count = MAX_RW_COUNT; | |
1da177e4 | 477 | |
775802c0 CH |
478 | if (file->f_op->read) |
479 | ret = file->f_op->read(file, buf, count, pos); | |
480 | else if (file->f_op->read_iter) | |
481 | ret = new_sync_read(file, buf, count, pos); | |
482 | else | |
483 | ret = -EINVAL; | |
484 | if (ret > 0) { | |
485 | fsnotify_access(file); | |
486 | add_rchar(current, ret); | |
487 | } | |
488 | inc_syscr(current); | |
1da177e4 LT |
489 | return ret; |
490 | } | |
491 | ||
5d5d5689 | 492 | static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) |
293bc982 AV |
493 | { |
494 | struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len }; | |
495 | struct kiocb kiocb; | |
496 | struct iov_iter iter; | |
497 | ssize_t ret; | |
498 | ||
499 | init_sync_kiocb(&kiocb, filp); | |
438ab720 | 500 | kiocb.ki_pos = (ppos ? *ppos : 0); |
293bc982 AV |
501 | iov_iter_init(&iter, WRITE, &iov, 1, len); |
502 | ||
bb7462b6 | 503 | ret = call_write_iter(filp, &kiocb, &iter); |
599bd19b | 504 | BUG_ON(ret == -EIOCBQUEUED); |
438ab720 | 505 | if (ret > 0 && ppos) |
f765b134 | 506 | *ppos = kiocb.ki_pos; |
293bc982 AV |
507 | return ret; |
508 | } | |
509 | ||
81238b2c | 510 | /* caller is responsible for file_start_write/file_end_write */ |
73e18f7c | 511 | ssize_t __kernel_write(struct file *file, const void *buf, size_t count, loff_t *pos) |
06ae43f3 AV |
512 | { |
513 | mm_segment_t old_fs; | |
514 | const char __user *p; | |
515 | ssize_t ret; | |
516 | ||
a01ac27b CH |
517 | if (WARN_ON_ONCE(!(file->f_mode & FMODE_WRITE))) |
518 | return -EBADF; | |
7f7f25e8 | 519 | if (!(file->f_mode & FMODE_CAN_WRITE)) |
3e84f48e AV |
520 | return -EINVAL; |
521 | ||
06ae43f3 | 522 | old_fs = get_fs(); |
736706be | 523 | set_fs(KERNEL_DS); |
06ae43f3 AV |
524 | p = (__force const char __user *)buf; |
525 | if (count > MAX_RW_COUNT) | |
526 | count = MAX_RW_COUNT; | |
53ad8626 CH |
527 | if (file->f_op->write) |
528 | ret = file->f_op->write(file, p, count, pos); | |
529 | else if (file->f_op->write_iter) | |
530 | ret = new_sync_write(file, p, count, pos); | |
531 | else | |
532 | ret = -EINVAL; | |
06ae43f3 AV |
533 | 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 | } | |
90fb7027 LT |
541 | /* |
542 | * This "EXPORT_SYMBOL_GPL()" is more of a "EXPORT_SYMBOL_DONTUSE()", | |
543 | * but autofs is one of the few internal kernel users that actually | |
544 | * wants this _and_ can be built as a module. So we need to export | |
545 | * this symbol for autofs, even though it really isn't appropriate | |
546 | * for any other kernel modules. | |
547 | */ | |
548 | EXPORT_SYMBOL_GPL(__kernel_write); | |
2ec3a12a | 549 | |
e13ec939 CH |
550 | ssize_t kernel_write(struct file *file, const void *buf, size_t count, |
551 | loff_t *pos) | |
ac452aca | 552 | { |
81238b2c | 553 | ssize_t ret; |
ac452aca | 554 | |
81238b2c CH |
555 | ret = rw_verify_area(WRITE, file, pos, count); |
556 | if (ret) | |
557 | return ret; | |
ac452aca | 558 | |
81238b2c CH |
559 | file_start_write(file); |
560 | ret = __kernel_write(file, buf, count, pos); | |
561 | file_end_write(file); | |
562 | return ret; | |
ac452aca CH |
563 | } |
564 | EXPORT_SYMBOL(kernel_write); | |
565 | ||
1da177e4 LT |
566 | ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos) |
567 | { | |
568 | ssize_t ret; | |
569 | ||
570 | if (!(file->f_mode & FMODE_WRITE)) | |
571 | return -EBADF; | |
7f7f25e8 | 572 | if (!(file->f_mode & FMODE_CAN_WRITE)) |
1da177e4 | 573 | return -EINVAL; |
96d4f267 | 574 | if (unlikely(!access_ok(buf, count))) |
1da177e4 LT |
575 | return -EFAULT; |
576 | ||
577 | ret = rw_verify_area(WRITE, file, pos, count); | |
53ad8626 CH |
578 | if (ret) |
579 | return ret; | |
580 | if (count > MAX_RW_COUNT) | |
581 | count = MAX_RW_COUNT; | |
582 | file_start_write(file); | |
583 | if (file->f_op->write) | |
584 | ret = file->f_op->write(file, buf, count, pos); | |
585 | else if (file->f_op->write_iter) | |
586 | ret = new_sync_write(file, buf, count, pos); | |
587 | else | |
588 | ret = -EINVAL; | |
589 | if (ret > 0) { | |
590 | fsnotify_modify(file); | |
591 | add_wchar(current, ret); | |
1da177e4 | 592 | } |
53ad8626 CH |
593 | inc_syscw(current); |
594 | file_end_write(file); | |
1da177e4 LT |
595 | return ret; |
596 | } | |
597 | ||
438ab720 KS |
598 | /* file_ppos returns &file->f_pos or NULL if file is stream */ |
599 | static inline loff_t *file_ppos(struct file *file) | |
1da177e4 | 600 | { |
438ab720 | 601 | return file->f_mode & FMODE_STREAM ? NULL : &file->f_pos; |
1da177e4 LT |
602 | } |
603 | ||
3ce4a7bf | 604 | ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count) |
1da177e4 | 605 | { |
9c225f26 | 606 | struct fd f = fdget_pos(fd); |
1da177e4 | 607 | ssize_t ret = -EBADF; |
1da177e4 | 608 | |
2903ff01 | 609 | if (f.file) { |
438ab720 KS |
610 | loff_t pos, *ppos = file_ppos(f.file); |
611 | if (ppos) { | |
612 | pos = *ppos; | |
613 | ppos = &pos; | |
614 | } | |
615 | ret = vfs_read(f.file, buf, count, ppos); | |
616 | if (ret >= 0 && ppos) | |
617 | f.file->f_pos = pos; | |
9c225f26 | 618 | fdput_pos(f); |
1da177e4 | 619 | } |
1da177e4 LT |
620 | return ret; |
621 | } | |
1da177e4 | 622 | |
3ce4a7bf DB |
623 | SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count) |
624 | { | |
625 | return ksys_read(fd, buf, count); | |
626 | } | |
627 | ||
e7a3e8b2 | 628 | ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count) |
1da177e4 | 629 | { |
9c225f26 | 630 | struct fd f = fdget_pos(fd); |
1da177e4 | 631 | ssize_t ret = -EBADF; |
1da177e4 | 632 | |
2903ff01 | 633 | if (f.file) { |
438ab720 KS |
634 | loff_t pos, *ppos = file_ppos(f.file); |
635 | if (ppos) { | |
636 | pos = *ppos; | |
637 | ppos = &pos; | |
638 | } | |
639 | ret = vfs_write(f.file, buf, count, ppos); | |
640 | if (ret >= 0 && ppos) | |
641 | f.file->f_pos = pos; | |
9c225f26 | 642 | fdput_pos(f); |
1da177e4 LT |
643 | } |
644 | ||
645 | return ret; | |
646 | } | |
647 | ||
e7a3e8b2 DB |
648 | SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf, |
649 | size_t, count) | |
650 | { | |
651 | return ksys_write(fd, buf, count); | |
652 | } | |
653 | ||
36028d5d DB |
654 | ssize_t ksys_pread64(unsigned int fd, char __user *buf, size_t count, |
655 | loff_t pos) | |
1da177e4 | 656 | { |
2903ff01 | 657 | struct fd f; |
1da177e4 | 658 | ssize_t ret = -EBADF; |
1da177e4 LT |
659 | |
660 | if (pos < 0) | |
661 | return -EINVAL; | |
662 | ||
2903ff01 AV |
663 | f = fdget(fd); |
664 | if (f.file) { | |
1da177e4 | 665 | ret = -ESPIPE; |
2903ff01 AV |
666 | if (f.file->f_mode & FMODE_PREAD) |
667 | ret = vfs_read(f.file, buf, count, &pos); | |
668 | fdput(f); | |
1da177e4 LT |
669 | } |
670 | ||
671 | return ret; | |
672 | } | |
673 | ||
36028d5d DB |
674 | SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf, |
675 | size_t, count, loff_t, pos) | |
676 | { | |
677 | return ksys_pread64(fd, buf, count, pos); | |
678 | } | |
679 | ||
680 | ssize_t ksys_pwrite64(unsigned int fd, const char __user *buf, | |
681 | size_t count, loff_t pos) | |
1da177e4 | 682 | { |
2903ff01 | 683 | struct fd f; |
1da177e4 | 684 | ssize_t ret = -EBADF; |
1da177e4 LT |
685 | |
686 | if (pos < 0) | |
687 | return -EINVAL; | |
688 | ||
2903ff01 AV |
689 | f = fdget(fd); |
690 | if (f.file) { | |
1da177e4 | 691 | ret = -ESPIPE; |
2903ff01 AV |
692 | if (f.file->f_mode & FMODE_PWRITE) |
693 | ret = vfs_write(f.file, buf, count, &pos); | |
694 | fdput(f); | |
1da177e4 LT |
695 | } |
696 | ||
697 | return ret; | |
698 | } | |
699 | ||
36028d5d DB |
700 | SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf, |
701 | size_t, count, loff_t, pos) | |
702 | { | |
703 | return ksys_pwrite64(fd, buf, count, pos); | |
704 | } | |
705 | ||
ac15ac06 | 706 | static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter, |
ddef7ed2 | 707 | loff_t *ppos, int type, rwf_t flags) |
293bc982 AV |
708 | { |
709 | struct kiocb kiocb; | |
293bc982 AV |
710 | ssize_t ret; |
711 | ||
712 | init_sync_kiocb(&kiocb, filp); | |
fdd2f5b7 GR |
713 | ret = kiocb_set_rw_flags(&kiocb, flags); |
714 | if (ret) | |
715 | return ret; | |
438ab720 | 716 | kiocb.ki_pos = (ppos ? *ppos : 0); |
293bc982 | 717 | |
0f78d06a | 718 | if (type == READ) |
bb7462b6 | 719 | ret = call_read_iter(filp, &kiocb, iter); |
0f78d06a | 720 | else |
bb7462b6 | 721 | ret = call_write_iter(filp, &kiocb, iter); |
599bd19b | 722 | BUG_ON(ret == -EIOCBQUEUED); |
438ab720 KS |
723 | if (ppos) |
724 | *ppos = kiocb.ki_pos; | |
293bc982 AV |
725 | return ret; |
726 | } | |
727 | ||
ee0b3e67 | 728 | /* Do it by hand, with file-ops */ |
ac15ac06 | 729 | static ssize_t do_loop_readv_writev(struct file *filp, struct iov_iter *iter, |
ddef7ed2 | 730 | loff_t *ppos, int type, rwf_t flags) |
ee0b3e67 | 731 | { |
ee0b3e67 BP |
732 | ssize_t ret = 0; |
733 | ||
97be7ebe | 734 | if (flags & ~RWF_HIPRI) |
793b80ef CH |
735 | return -EOPNOTSUPP; |
736 | ||
ac15ac06 AV |
737 | while (iov_iter_count(iter)) { |
738 | struct iovec iovec = iov_iter_iovec(iter); | |
ee0b3e67 BP |
739 | ssize_t nr; |
740 | ||
0f78d06a MS |
741 | if (type == READ) { |
742 | nr = filp->f_op->read(filp, iovec.iov_base, | |
743 | iovec.iov_len, ppos); | |
744 | } else { | |
745 | nr = filp->f_op->write(filp, iovec.iov_base, | |
746 | iovec.iov_len, ppos); | |
747 | } | |
ee0b3e67 BP |
748 | |
749 | if (nr < 0) { | |
750 | if (!ret) | |
751 | ret = nr; | |
752 | break; | |
753 | } | |
754 | ret += nr; | |
ac15ac06 | 755 | if (nr != iovec.iov_len) |
ee0b3e67 | 756 | break; |
ac15ac06 | 757 | iov_iter_advance(iter, nr); |
ee0b3e67 BP |
758 | } |
759 | ||
760 | return ret; | |
761 | } | |
762 | ||
ffecee4f VN |
763 | /** |
764 | * rw_copy_check_uvector() - Copy an array of &struct iovec from userspace | |
765 | * into the kernel and check that it is valid. | |
766 | * | |
767 | * @type: One of %CHECK_IOVEC_ONLY, %READ, or %WRITE. | |
768 | * @uvector: Pointer to the userspace array. | |
769 | * @nr_segs: Number of elements in userspace array. | |
770 | * @fast_segs: Number of elements in @fast_pointer. | |
771 | * @fast_pointer: Pointer to (usually small on-stack) kernel array. | |
772 | * @ret_pointer: (output parameter) Pointer to a variable that will point to | |
773 | * either @fast_pointer, a newly allocated kernel array, or NULL, | |
774 | * depending on which array was used. | |
775 | * | |
776 | * This function copies an array of &struct iovec of @nr_segs from | |
777 | * userspace into the kernel and checks that each element is valid (e.g. | |
778 | * it does not point to a kernel address or cause overflow by being too | |
779 | * large, etc.). | |
780 | * | |
781 | * As an optimization, the caller may provide a pointer to a small | |
782 | * on-stack array in @fast_pointer, typically %UIO_FASTIOV elements long | |
783 | * (the size of this array, or 0 if unused, should be given in @fast_segs). | |
784 | * | |
785 | * @ret_pointer will always point to the array that was used, so the | |
786 | * caller must take care not to call kfree() on it e.g. in case the | |
787 | * @fast_pointer array was used and it was allocated on the stack. | |
788 | * | |
789 | * Return: The total number of bytes covered by the iovec array on success | |
790 | * or a negative error code on error. | |
791 | */ | |
eed4e51f BP |
792 | ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector, |
793 | unsigned long nr_segs, unsigned long fast_segs, | |
794 | struct iovec *fast_pointer, | |
ac34ebb3 | 795 | struct iovec **ret_pointer) |
435f49a5 | 796 | { |
eed4e51f | 797 | unsigned long seg; |
435f49a5 | 798 | ssize_t ret; |
eed4e51f BP |
799 | struct iovec *iov = fast_pointer; |
800 | ||
435f49a5 LT |
801 | /* |
802 | * SuS says "The readv() function *may* fail if the iovcnt argument | |
803 | * was less than or equal to 0, or greater than {IOV_MAX}. Linux has | |
804 | * traditionally returned zero for zero segments, so... | |
805 | */ | |
eed4e51f BP |
806 | if (nr_segs == 0) { |
807 | ret = 0; | |
435f49a5 | 808 | goto out; |
eed4e51f BP |
809 | } |
810 | ||
435f49a5 LT |
811 | /* |
812 | * First get the "struct iovec" from user memory and | |
813 | * verify all the pointers | |
814 | */ | |
eed4e51f BP |
815 | if (nr_segs > UIO_MAXIOV) { |
816 | ret = -EINVAL; | |
435f49a5 | 817 | goto out; |
eed4e51f BP |
818 | } |
819 | if (nr_segs > fast_segs) { | |
6da2ec56 | 820 | iov = kmalloc_array(nr_segs, sizeof(struct iovec), GFP_KERNEL); |
eed4e51f BP |
821 | if (iov == NULL) { |
822 | ret = -ENOMEM; | |
435f49a5 | 823 | goto out; |
eed4e51f | 824 | } |
435f49a5 | 825 | } |
eed4e51f BP |
826 | if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) { |
827 | ret = -EFAULT; | |
435f49a5 | 828 | goto out; |
eed4e51f BP |
829 | } |
830 | ||
435f49a5 | 831 | /* |
eed4e51f BP |
832 | * According to the Single Unix Specification we should return EINVAL |
833 | * if an element length is < 0 when cast to ssize_t or if the | |
834 | * total length would overflow the ssize_t return value of the | |
835 | * system call. | |
435f49a5 LT |
836 | * |
837 | * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the | |
838 | * overflow case. | |
839 | */ | |
eed4e51f | 840 | ret = 0; |
435f49a5 LT |
841 | for (seg = 0; seg < nr_segs; seg++) { |
842 | void __user *buf = iov[seg].iov_base; | |
843 | ssize_t len = (ssize_t)iov[seg].iov_len; | |
eed4e51f BP |
844 | |
845 | /* see if we we're about to use an invalid len or if | |
846 | * it's about to overflow ssize_t */ | |
435f49a5 | 847 | if (len < 0) { |
eed4e51f | 848 | ret = -EINVAL; |
435f49a5 | 849 | goto out; |
eed4e51f | 850 | } |
ac34ebb3 | 851 | if (type >= 0 |
96d4f267 | 852 | && unlikely(!access_ok(buf, len))) { |
eed4e51f | 853 | ret = -EFAULT; |
435f49a5 LT |
854 | goto out; |
855 | } | |
856 | if (len > MAX_RW_COUNT - ret) { | |
857 | len = MAX_RW_COUNT - ret; | |
858 | iov[seg].iov_len = len; | |
eed4e51f | 859 | } |
eed4e51f | 860 | ret += len; |
435f49a5 | 861 | } |
eed4e51f BP |
862 | out: |
863 | *ret_pointer = iov; | |
864 | return ret; | |
865 | } | |
866 | ||
f5029855 AV |
867 | #ifdef CONFIG_COMPAT |
868 | ssize_t compat_rw_copy_check_uvector(int type, | |
869 | const struct compat_iovec __user *uvector, unsigned long nr_segs, | |
870 | unsigned long fast_segs, struct iovec *fast_pointer, | |
871 | struct iovec **ret_pointer) | |
872 | { | |
873 | compat_ssize_t tot_len; | |
874 | struct iovec *iov = *ret_pointer = fast_pointer; | |
875 | ssize_t ret = 0; | |
876 | int seg; | |
877 | ||
878 | /* | |
879 | * SuS says "The readv() function *may* fail if the iovcnt argument | |
880 | * was less than or equal to 0, or greater than {IOV_MAX}. Linux has | |
881 | * traditionally returned zero for zero segments, so... | |
882 | */ | |
883 | if (nr_segs == 0) | |
884 | goto out; | |
885 | ||
886 | ret = -EINVAL; | |
887 | if (nr_segs > UIO_MAXIOV) | |
888 | goto out; | |
889 | if (nr_segs > fast_segs) { | |
890 | ret = -ENOMEM; | |
6da2ec56 | 891 | iov = kmalloc_array(nr_segs, sizeof(struct iovec), GFP_KERNEL); |
f5029855 AV |
892 | if (iov == NULL) |
893 | goto out; | |
894 | } | |
895 | *ret_pointer = iov; | |
896 | ||
897 | ret = -EFAULT; | |
96d4f267 | 898 | if (!access_ok(uvector, nr_segs*sizeof(*uvector))) |
f5029855 AV |
899 | goto out; |
900 | ||
901 | /* | |
902 | * Single unix specification: | |
903 | * We should -EINVAL if an element length is not >= 0 and fitting an | |
904 | * ssize_t. | |
905 | * | |
906 | * In Linux, the total length is limited to MAX_RW_COUNT, there is | |
907 | * no overflow possibility. | |
908 | */ | |
909 | tot_len = 0; | |
910 | ret = -EINVAL; | |
911 | for (seg = 0; seg < nr_segs; seg++) { | |
912 | compat_uptr_t buf; | |
913 | compat_ssize_t len; | |
914 | ||
915 | if (__get_user(len, &uvector->iov_len) || | |
916 | __get_user(buf, &uvector->iov_base)) { | |
917 | ret = -EFAULT; | |
918 | goto out; | |
919 | } | |
920 | if (len < 0) /* size_t not fitting in compat_ssize_t .. */ | |
921 | goto out; | |
922 | if (type >= 0 && | |
96d4f267 | 923 | !access_ok(compat_ptr(buf), len)) { |
f5029855 AV |
924 | ret = -EFAULT; |
925 | goto out; | |
926 | } | |
927 | if (len > MAX_RW_COUNT - tot_len) | |
928 | len = MAX_RW_COUNT - tot_len; | |
929 | tot_len += len; | |
930 | iov->iov_base = compat_ptr(buf); | |
931 | iov->iov_len = (compat_size_t) len; | |
932 | uvector++; | |
933 | iov++; | |
934 | } | |
935 | ret = tot_len; | |
936 | ||
937 | out: | |
938 | return ret; | |
939 | } | |
940 | #endif | |
941 | ||
19c73586 | 942 | static ssize_t do_iter_read(struct file *file, struct iov_iter *iter, |
ddef7ed2 | 943 | loff_t *pos, rwf_t flags) |
1da177e4 | 944 | { |
1da177e4 | 945 | size_t tot_len; |
7687a7a4 | 946 | ssize_t ret = 0; |
1da177e4 | 947 | |
edab5fe3 CH |
948 | if (!(file->f_mode & FMODE_READ)) |
949 | return -EBADF; | |
950 | if (!(file->f_mode & FMODE_CAN_READ)) | |
951 | return -EINVAL; | |
952 | ||
7687a7a4 | 953 | tot_len = iov_iter_count(iter); |
0504c074 AV |
954 | if (!tot_len) |
955 | goto out; | |
19c73586 | 956 | ret = rw_verify_area(READ, file, pos, tot_len); |
e28cc715 | 957 | if (ret < 0) |
19c73586 | 958 | return ret; |
1da177e4 | 959 | |
19c73586 CH |
960 | if (file->f_op->read_iter) |
961 | ret = do_iter_readv_writev(file, iter, pos, READ, flags); | |
ee0b3e67 | 962 | else |
19c73586 | 963 | ret = do_loop_readv_writev(file, iter, pos, READ, flags); |
1da177e4 | 964 | out: |
19c73586 CH |
965 | if (ret >= 0) |
966 | fsnotify_access(file); | |
1da177e4 | 967 | return ret; |
1da177e4 LT |
968 | } |
969 | ||
5dcdc43e JX |
970 | ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, |
971 | struct iov_iter *iter) | |
972 | { | |
973 | size_t tot_len; | |
974 | ssize_t ret = 0; | |
975 | ||
976 | if (!file->f_op->read_iter) | |
977 | return -EINVAL; | |
978 | if (!(file->f_mode & FMODE_READ)) | |
979 | return -EBADF; | |
980 | if (!(file->f_mode & FMODE_CAN_READ)) | |
981 | return -EINVAL; | |
982 | ||
983 | tot_len = iov_iter_count(iter); | |
984 | if (!tot_len) | |
985 | goto out; | |
986 | ret = rw_verify_area(READ, file, &iocb->ki_pos, tot_len); | |
987 | if (ret < 0) | |
988 | return ret; | |
989 | ||
990 | ret = call_read_iter(file, iocb, iter); | |
991 | out: | |
992 | if (ret >= 0) | |
993 | fsnotify_access(file); | |
994 | return ret; | |
995 | } | |
996 | EXPORT_SYMBOL(vfs_iocb_iter_read); | |
997 | ||
18e9710e | 998 | ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, |
ddef7ed2 | 999 | rwf_t flags) |
7687a7a4 | 1000 | { |
18e9710e CH |
1001 | if (!file->f_op->read_iter) |
1002 | return -EINVAL; | |
1003 | return do_iter_read(file, iter, ppos, flags); | |
1004 | } | |
1005 | EXPORT_SYMBOL(vfs_iter_read); | |
7687a7a4 | 1006 | |
19c73586 | 1007 | static ssize_t do_iter_write(struct file *file, struct iov_iter *iter, |
ddef7ed2 | 1008 | loff_t *pos, rwf_t flags) |
19c73586 CH |
1009 | { |
1010 | size_t tot_len; | |
1011 | ssize_t ret = 0; | |
03d95eb2 | 1012 | |
edab5fe3 CH |
1013 | if (!(file->f_mode & FMODE_WRITE)) |
1014 | return -EBADF; | |
1015 | if (!(file->f_mode & FMODE_CAN_WRITE)) | |
1016 | return -EINVAL; | |
1017 | ||
19c73586 CH |
1018 | tot_len = iov_iter_count(iter); |
1019 | if (!tot_len) | |
1020 | return 0; | |
1021 | ret = rw_verify_area(WRITE, file, pos, tot_len); | |
7687a7a4 MS |
1022 | if (ret < 0) |
1023 | return ret; | |
1024 | ||
19c73586 CH |
1025 | if (file->f_op->write_iter) |
1026 | ret = do_iter_readv_writev(file, iter, pos, WRITE, flags); | |
1027 | else | |
1028 | ret = do_loop_readv_writev(file, iter, pos, WRITE, flags); | |
19c73586 CH |
1029 | if (ret > 0) |
1030 | fsnotify_modify(file); | |
7687a7a4 MS |
1031 | return ret; |
1032 | } | |
1033 | ||
5dcdc43e JX |
1034 | ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, |
1035 | struct iov_iter *iter) | |
1036 | { | |
1037 | size_t tot_len; | |
1038 | ssize_t ret = 0; | |
1039 | ||
1040 | if (!file->f_op->write_iter) | |
1041 | return -EINVAL; | |
1042 | if (!(file->f_mode & FMODE_WRITE)) | |
1043 | return -EBADF; | |
1044 | if (!(file->f_mode & FMODE_CAN_WRITE)) | |
1045 | return -EINVAL; | |
1046 | ||
1047 | tot_len = iov_iter_count(iter); | |
1048 | if (!tot_len) | |
1049 | return 0; | |
1050 | ret = rw_verify_area(WRITE, file, &iocb->ki_pos, tot_len); | |
1051 | if (ret < 0) | |
1052 | return ret; | |
1053 | ||
1054 | ret = call_write_iter(file, iocb, iter); | |
1055 | if (ret > 0) | |
1056 | fsnotify_modify(file); | |
1057 | ||
1058 | return ret; | |
1059 | } | |
1060 | EXPORT_SYMBOL(vfs_iocb_iter_write); | |
1061 | ||
abbb6589 | 1062 | ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, |
ddef7ed2 | 1063 | rwf_t flags) |
abbb6589 CH |
1064 | { |
1065 | if (!file->f_op->write_iter) | |
1066 | return -EINVAL; | |
1067 | return do_iter_write(file, iter, ppos, flags); | |
1068 | } | |
1069 | EXPORT_SYMBOL(vfs_iter_write); | |
1070 | ||
1da177e4 | 1071 | ssize_t vfs_readv(struct file *file, const struct iovec __user *vec, |
ddef7ed2 | 1072 | unsigned long vlen, loff_t *pos, rwf_t flags) |
1da177e4 | 1073 | { |
7687a7a4 MS |
1074 | struct iovec iovstack[UIO_FASTIOV]; |
1075 | struct iovec *iov = iovstack; | |
1076 | struct iov_iter iter; | |
1077 | ssize_t ret; | |
1da177e4 | 1078 | |
251b42a1 | 1079 | ret = import_iovec(READ, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter); |
edab5fe3 CH |
1080 | if (ret >= 0) { |
1081 | ret = do_iter_read(file, &iter, pos, flags); | |
1082 | kfree(iov); | |
1083 | } | |
1da177e4 | 1084 | |
251b42a1 CH |
1085 | return ret; |
1086 | } | |
1da177e4 | 1087 | |
9725d4ce | 1088 | static ssize_t vfs_writev(struct file *file, const struct iovec __user *vec, |
ddef7ed2 | 1089 | unsigned long vlen, loff_t *pos, rwf_t flags) |
1da177e4 | 1090 | { |
251b42a1 CH |
1091 | struct iovec iovstack[UIO_FASTIOV]; |
1092 | struct iovec *iov = iovstack; | |
1093 | struct iov_iter iter; | |
1094 | ssize_t ret; | |
1da177e4 | 1095 | |
251b42a1 | 1096 | ret = import_iovec(WRITE, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter); |
edab5fe3 | 1097 | if (ret >= 0) { |
62473a2d | 1098 | file_start_write(file); |
edab5fe3 | 1099 | ret = do_iter_write(file, &iter, pos, flags); |
62473a2d | 1100 | file_end_write(file); |
edab5fe3 CH |
1101 | kfree(iov); |
1102 | } | |
251b42a1 | 1103 | return ret; |
1da177e4 | 1104 | } |
1da177e4 | 1105 | |
f17d8b35 | 1106 | static ssize_t do_readv(unsigned long fd, const struct iovec __user *vec, |
ddef7ed2 | 1107 | unsigned long vlen, rwf_t flags) |
1da177e4 | 1108 | { |
9c225f26 | 1109 | struct fd f = fdget_pos(fd); |
1da177e4 | 1110 | ssize_t ret = -EBADF; |
1da177e4 | 1111 | |
2903ff01 | 1112 | if (f.file) { |
438ab720 KS |
1113 | loff_t pos, *ppos = file_ppos(f.file); |
1114 | if (ppos) { | |
1115 | pos = *ppos; | |
1116 | ppos = &pos; | |
1117 | } | |
1118 | ret = vfs_readv(f.file, vec, vlen, ppos, flags); | |
1119 | if (ret >= 0 && ppos) | |
1120 | f.file->f_pos = pos; | |
9c225f26 | 1121 | fdput_pos(f); |
1da177e4 LT |
1122 | } |
1123 | ||
1124 | if (ret > 0) | |
4b98d11b AD |
1125 | add_rchar(current, ret); |
1126 | inc_syscr(current); | |
1da177e4 LT |
1127 | return ret; |
1128 | } | |
1129 | ||
f17d8b35 | 1130 | static ssize_t do_writev(unsigned long fd, const struct iovec __user *vec, |
ddef7ed2 | 1131 | unsigned long vlen, rwf_t flags) |
1da177e4 | 1132 | { |
9c225f26 | 1133 | struct fd f = fdget_pos(fd); |
1da177e4 | 1134 | ssize_t ret = -EBADF; |
1da177e4 | 1135 | |
2903ff01 | 1136 | if (f.file) { |
438ab720 KS |
1137 | loff_t pos, *ppos = file_ppos(f.file); |
1138 | if (ppos) { | |
1139 | pos = *ppos; | |
1140 | ppos = &pos; | |
1141 | } | |
1142 | ret = vfs_writev(f.file, vec, vlen, ppos, flags); | |
1143 | if (ret >= 0 && ppos) | |
1144 | f.file->f_pos = pos; | |
9c225f26 | 1145 | fdput_pos(f); |
1da177e4 LT |
1146 | } |
1147 | ||
1148 | if (ret > 0) | |
4b98d11b AD |
1149 | add_wchar(current, ret); |
1150 | inc_syscw(current); | |
1da177e4 LT |
1151 | return ret; |
1152 | } | |
1153 | ||
601cc11d LT |
1154 | static inline loff_t pos_from_hilo(unsigned long high, unsigned long low) |
1155 | { | |
1156 | #define HALF_LONG_BITS (BITS_PER_LONG / 2) | |
1157 | return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low; | |
1158 | } | |
1159 | ||
f17d8b35 | 1160 | static ssize_t do_preadv(unsigned long fd, const struct iovec __user *vec, |
ddef7ed2 | 1161 | unsigned long vlen, loff_t pos, rwf_t flags) |
f3554f4b | 1162 | { |
2903ff01 | 1163 | struct fd f; |
f3554f4b | 1164 | ssize_t ret = -EBADF; |
f3554f4b GH |
1165 | |
1166 | if (pos < 0) | |
1167 | return -EINVAL; | |
1168 | ||
2903ff01 AV |
1169 | f = fdget(fd); |
1170 | if (f.file) { | |
f3554f4b | 1171 | ret = -ESPIPE; |
2903ff01 | 1172 | if (f.file->f_mode & FMODE_PREAD) |
f17d8b35 | 1173 | ret = vfs_readv(f.file, vec, vlen, &pos, flags); |
2903ff01 | 1174 | fdput(f); |
f3554f4b GH |
1175 | } |
1176 | ||
1177 | if (ret > 0) | |
1178 | add_rchar(current, ret); | |
1179 | inc_syscr(current); | |
1180 | return ret; | |
1181 | } | |
1182 | ||
f17d8b35 | 1183 | static ssize_t do_pwritev(unsigned long fd, const struct iovec __user *vec, |
ddef7ed2 | 1184 | unsigned long vlen, loff_t pos, rwf_t flags) |
f3554f4b | 1185 | { |
2903ff01 | 1186 | struct fd f; |
f3554f4b | 1187 | ssize_t ret = -EBADF; |
f3554f4b GH |
1188 | |
1189 | if (pos < 0) | |
1190 | return -EINVAL; | |
1191 | ||
2903ff01 AV |
1192 | f = fdget(fd); |
1193 | if (f.file) { | |
f3554f4b | 1194 | ret = -ESPIPE; |
2903ff01 | 1195 | if (f.file->f_mode & FMODE_PWRITE) |
f17d8b35 | 1196 | ret = vfs_writev(f.file, vec, vlen, &pos, flags); |
2903ff01 | 1197 | fdput(f); |
f3554f4b GH |
1198 | } |
1199 | ||
1200 | if (ret > 0) | |
1201 | add_wchar(current, ret); | |
1202 | inc_syscw(current); | |
1203 | return ret; | |
1204 | } | |
1205 | ||
f17d8b35 MT |
1206 | SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec, |
1207 | unsigned long, vlen) | |
1208 | { | |
1209 | return do_readv(fd, vec, vlen, 0); | |
1210 | } | |
1211 | ||
1212 | SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec, | |
1213 | unsigned long, vlen) | |
1214 | { | |
1215 | return do_writev(fd, vec, vlen, 0); | |
1216 | } | |
1217 | ||
1218 | SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec, | |
1219 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) | |
1220 | { | |
1221 | loff_t pos = pos_from_hilo(pos_h, pos_l); | |
1222 | ||
1223 | return do_preadv(fd, vec, vlen, pos, 0); | |
1224 | } | |
1225 | ||
1226 | SYSCALL_DEFINE6(preadv2, unsigned long, fd, const struct iovec __user *, vec, | |
1227 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h, | |
ddef7ed2 | 1228 | rwf_t, flags) |
f17d8b35 MT |
1229 | { |
1230 | loff_t pos = pos_from_hilo(pos_h, pos_l); | |
1231 | ||
1232 | if (pos == -1) | |
1233 | return do_readv(fd, vec, vlen, flags); | |
1234 | ||
1235 | return do_preadv(fd, vec, vlen, pos, flags); | |
1236 | } | |
1237 | ||
1238 | SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec, | |
1239 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) | |
1240 | { | |
1241 | loff_t pos = pos_from_hilo(pos_h, pos_l); | |
1242 | ||
1243 | return do_pwritev(fd, vec, vlen, pos, 0); | |
1244 | } | |
1245 | ||
1246 | SYSCALL_DEFINE6(pwritev2, unsigned long, fd, const struct iovec __user *, vec, | |
1247 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h, | |
ddef7ed2 | 1248 | rwf_t, flags) |
f17d8b35 MT |
1249 | { |
1250 | loff_t pos = pos_from_hilo(pos_h, pos_l); | |
1251 | ||
1252 | if (pos == -1) | |
1253 | return do_writev(fd, vec, vlen, flags); | |
1254 | ||
1255 | return do_pwritev(fd, vec, vlen, pos, flags); | |
1256 | } | |
1257 | ||
72ec3516 | 1258 | #ifdef CONFIG_COMPAT |
72ec3516 AV |
1259 | static size_t compat_readv(struct file *file, |
1260 | const struct compat_iovec __user *vec, | |
ddef7ed2 | 1261 | unsigned long vlen, loff_t *pos, rwf_t flags) |
72ec3516 | 1262 | { |
72ec3516 AV |
1263 | struct iovec iovstack[UIO_FASTIOV]; |
1264 | struct iovec *iov = iovstack; | |
ac15ac06 | 1265 | struct iov_iter iter; |
72ec3516 | 1266 | ssize_t ret; |
72ec3516 | 1267 | |
26c87fb7 | 1268 | ret = compat_import_iovec(READ, vec, vlen, UIO_FASTIOV, &iov, &iter); |
edab5fe3 CH |
1269 | if (ret >= 0) { |
1270 | ret = do_iter_read(file, &iter, pos, flags); | |
1271 | kfree(iov); | |
1272 | } | |
72ec3516 AV |
1273 | if (ret > 0) |
1274 | add_rchar(current, ret); | |
1275 | inc_syscr(current); | |
1276 | return ret; | |
1277 | } | |
1278 | ||
f17d8b35 MT |
1279 | static size_t do_compat_readv(compat_ulong_t fd, |
1280 | const struct compat_iovec __user *vec, | |
ddef7ed2 | 1281 | compat_ulong_t vlen, rwf_t flags) |
72ec3516 | 1282 | { |
9c225f26 | 1283 | struct fd f = fdget_pos(fd); |
72ec3516 AV |
1284 | ssize_t ret; |
1285 | loff_t pos; | |
1286 | ||
1287 | if (!f.file) | |
1288 | return -EBADF; | |
1289 | pos = f.file->f_pos; | |
f17d8b35 | 1290 | ret = compat_readv(f.file, vec, vlen, &pos, flags); |
5faf153e AV |
1291 | if (ret >= 0) |
1292 | f.file->f_pos = pos; | |
9c225f26 | 1293 | fdput_pos(f); |
72ec3516 | 1294 | return ret; |
f17d8b35 | 1295 | |
72ec3516 AV |
1296 | } |
1297 | ||
f17d8b35 MT |
1298 | COMPAT_SYSCALL_DEFINE3(readv, compat_ulong_t, fd, |
1299 | const struct compat_iovec __user *,vec, | |
1300 | compat_ulong_t, vlen) | |
1301 | { | |
1302 | return do_compat_readv(fd, vec, vlen, 0); | |
1303 | } | |
1304 | ||
1305 | static long do_compat_preadv64(unsigned long fd, | |
378a10f3 | 1306 | const struct compat_iovec __user *vec, |
ddef7ed2 | 1307 | unsigned long vlen, loff_t pos, rwf_t flags) |
72ec3516 AV |
1308 | { |
1309 | struct fd f; | |
1310 | ssize_t ret; | |
1311 | ||
1312 | if (pos < 0) | |
1313 | return -EINVAL; | |
1314 | f = fdget(fd); | |
1315 | if (!f.file) | |
1316 | return -EBADF; | |
1317 | ret = -ESPIPE; | |
1318 | if (f.file->f_mode & FMODE_PREAD) | |
f17d8b35 | 1319 | ret = compat_readv(f.file, vec, vlen, &pos, flags); |
72ec3516 AV |
1320 | fdput(f); |
1321 | return ret; | |
1322 | } | |
1323 | ||
378a10f3 HC |
1324 | #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64 |
1325 | COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd, | |
1326 | const struct compat_iovec __user *,vec, | |
1327 | unsigned long, vlen, loff_t, pos) | |
1328 | { | |
f17d8b35 | 1329 | return do_compat_preadv64(fd, vec, vlen, pos, 0); |
378a10f3 HC |
1330 | } |
1331 | #endif | |
1332 | ||
dfd948e3 | 1333 | COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd, |
72ec3516 | 1334 | const struct compat_iovec __user *,vec, |
dfd948e3 | 1335 | compat_ulong_t, vlen, u32, pos_low, u32, pos_high) |
72ec3516 AV |
1336 | { |
1337 | loff_t pos = ((loff_t)pos_high << 32) | pos_low; | |
378a10f3 | 1338 | |
f17d8b35 MT |
1339 | return do_compat_preadv64(fd, vec, vlen, pos, 0); |
1340 | } | |
1341 | ||
3ebfd81f L |
1342 | #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2 |
1343 | COMPAT_SYSCALL_DEFINE5(preadv64v2, unsigned long, fd, | |
1344 | const struct compat_iovec __user *,vec, | |
ddef7ed2 | 1345 | unsigned long, vlen, loff_t, pos, rwf_t, flags) |
3ebfd81f | 1346 | { |
cc4b1242 AJ |
1347 | if (pos == -1) |
1348 | return do_compat_readv(fd, vec, vlen, flags); | |
1349 | ||
3ebfd81f L |
1350 | return do_compat_preadv64(fd, vec, vlen, pos, flags); |
1351 | } | |
1352 | #endif | |
1353 | ||
f17d8b35 MT |
1354 | COMPAT_SYSCALL_DEFINE6(preadv2, compat_ulong_t, fd, |
1355 | const struct compat_iovec __user *,vec, | |
1356 | compat_ulong_t, vlen, u32, pos_low, u32, pos_high, | |
ddef7ed2 | 1357 | rwf_t, flags) |
f17d8b35 MT |
1358 | { |
1359 | loff_t pos = ((loff_t)pos_high << 32) | pos_low; | |
1360 | ||
1361 | if (pos == -1) | |
1362 | return do_compat_readv(fd, vec, vlen, flags); | |
1363 | ||
1364 | return do_compat_preadv64(fd, vec, vlen, pos, flags); | |
72ec3516 AV |
1365 | } |
1366 | ||
1367 | static size_t compat_writev(struct file *file, | |
1368 | const struct compat_iovec __user *vec, | |
ddef7ed2 | 1369 | unsigned long vlen, loff_t *pos, rwf_t flags) |
72ec3516 | 1370 | { |
26c87fb7 CH |
1371 | struct iovec iovstack[UIO_FASTIOV]; |
1372 | struct iovec *iov = iovstack; | |
1373 | struct iov_iter iter; | |
edab5fe3 | 1374 | ssize_t ret; |
72ec3516 | 1375 | |
26c87fb7 | 1376 | ret = compat_import_iovec(WRITE, vec, vlen, UIO_FASTIOV, &iov, &iter); |
edab5fe3 | 1377 | if (ret >= 0) { |
62473a2d | 1378 | file_start_write(file); |
edab5fe3 | 1379 | ret = do_iter_write(file, &iter, pos, flags); |
62473a2d | 1380 | file_end_write(file); |
edab5fe3 CH |
1381 | kfree(iov); |
1382 | } | |
72ec3516 AV |
1383 | if (ret > 0) |
1384 | add_wchar(current, ret); | |
1385 | inc_syscw(current); | |
1386 | return ret; | |
1387 | } | |
1388 | ||
f17d8b35 MT |
1389 | static size_t do_compat_writev(compat_ulong_t fd, |
1390 | const struct compat_iovec __user* vec, | |
ddef7ed2 | 1391 | compat_ulong_t vlen, rwf_t flags) |
72ec3516 | 1392 | { |
9c225f26 | 1393 | struct fd f = fdget_pos(fd); |
72ec3516 AV |
1394 | ssize_t ret; |
1395 | loff_t pos; | |
1396 | ||
1397 | if (!f.file) | |
1398 | return -EBADF; | |
1399 | pos = f.file->f_pos; | |
f17d8b35 | 1400 | ret = compat_writev(f.file, vec, vlen, &pos, flags); |
5faf153e AV |
1401 | if (ret >= 0) |
1402 | f.file->f_pos = pos; | |
9c225f26 | 1403 | fdput_pos(f); |
72ec3516 AV |
1404 | return ret; |
1405 | } | |
1406 | ||
f17d8b35 MT |
1407 | COMPAT_SYSCALL_DEFINE3(writev, compat_ulong_t, fd, |
1408 | const struct compat_iovec __user *, vec, | |
1409 | compat_ulong_t, vlen) | |
1410 | { | |
1411 | return do_compat_writev(fd, vec, vlen, 0); | |
1412 | } | |
1413 | ||
1414 | static long do_compat_pwritev64(unsigned long fd, | |
378a10f3 | 1415 | const struct compat_iovec __user *vec, |
ddef7ed2 | 1416 | unsigned long vlen, loff_t pos, rwf_t flags) |
72ec3516 AV |
1417 | { |
1418 | struct fd f; | |
1419 | ssize_t ret; | |
1420 | ||
1421 | if (pos < 0) | |
1422 | return -EINVAL; | |
1423 | f = fdget(fd); | |
1424 | if (!f.file) | |
1425 | return -EBADF; | |
1426 | ret = -ESPIPE; | |
1427 | if (f.file->f_mode & FMODE_PWRITE) | |
f17d8b35 | 1428 | ret = compat_writev(f.file, vec, vlen, &pos, flags); |
72ec3516 AV |
1429 | fdput(f); |
1430 | return ret; | |
1431 | } | |
1432 | ||
378a10f3 HC |
1433 | #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64 |
1434 | COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd, | |
1435 | const struct compat_iovec __user *,vec, | |
1436 | unsigned long, vlen, loff_t, pos) | |
1437 | { | |
f17d8b35 | 1438 | return do_compat_pwritev64(fd, vec, vlen, pos, 0); |
378a10f3 HC |
1439 | } |
1440 | #endif | |
1441 | ||
dfd948e3 | 1442 | COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd, |
72ec3516 | 1443 | const struct compat_iovec __user *,vec, |
dfd948e3 | 1444 | compat_ulong_t, vlen, u32, pos_low, u32, pos_high) |
72ec3516 AV |
1445 | { |
1446 | loff_t pos = ((loff_t)pos_high << 32) | pos_low; | |
378a10f3 | 1447 | |
f17d8b35 | 1448 | return do_compat_pwritev64(fd, vec, vlen, pos, 0); |
72ec3516 | 1449 | } |
f17d8b35 | 1450 | |
3ebfd81f L |
1451 | #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2 |
1452 | COMPAT_SYSCALL_DEFINE5(pwritev64v2, unsigned long, fd, | |
1453 | const struct compat_iovec __user *,vec, | |
ddef7ed2 | 1454 | unsigned long, vlen, loff_t, pos, rwf_t, flags) |
3ebfd81f | 1455 | { |
cc4b1242 AJ |
1456 | if (pos == -1) |
1457 | return do_compat_writev(fd, vec, vlen, flags); | |
1458 | ||
3ebfd81f L |
1459 | return do_compat_pwritev64(fd, vec, vlen, pos, flags); |
1460 | } | |
1461 | #endif | |
1462 | ||
f17d8b35 MT |
1463 | COMPAT_SYSCALL_DEFINE6(pwritev2, compat_ulong_t, fd, |
1464 | const struct compat_iovec __user *,vec, | |
ddef7ed2 | 1465 | compat_ulong_t, vlen, u32, pos_low, u32, pos_high, rwf_t, flags) |
f17d8b35 MT |
1466 | { |
1467 | loff_t pos = ((loff_t)pos_high << 32) | pos_low; | |
1468 | ||
1469 | if (pos == -1) | |
1470 | return do_compat_writev(fd, vec, vlen, flags); | |
1471 | ||
1472 | return do_compat_pwritev64(fd, vec, vlen, pos, flags); | |
72ec3516 | 1473 | } |
f17d8b35 | 1474 | |
72ec3516 AV |
1475 | #endif |
1476 | ||
19f4fc3a AV |
1477 | static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos, |
1478 | size_t count, loff_t max) | |
1da177e4 | 1479 | { |
2903ff01 AV |
1480 | struct fd in, out; |
1481 | struct inode *in_inode, *out_inode; | |
1da177e4 | 1482 | loff_t pos; |
7995bd28 | 1483 | loff_t out_pos; |
1da177e4 | 1484 | ssize_t retval; |
2903ff01 | 1485 | int fl; |
1da177e4 LT |
1486 | |
1487 | /* | |
1488 | * Get input file, and verify that it is ok.. | |
1489 | */ | |
1490 | retval = -EBADF; | |
2903ff01 AV |
1491 | in = fdget(in_fd); |
1492 | if (!in.file) | |
1da177e4 | 1493 | goto out; |
2903ff01 | 1494 | if (!(in.file->f_mode & FMODE_READ)) |
1da177e4 | 1495 | goto fput_in; |
1da177e4 | 1496 | retval = -ESPIPE; |
7995bd28 AV |
1497 | if (!ppos) { |
1498 | pos = in.file->f_pos; | |
1499 | } else { | |
1500 | pos = *ppos; | |
2903ff01 | 1501 | if (!(in.file->f_mode & FMODE_PREAD)) |
1da177e4 | 1502 | goto fput_in; |
7995bd28 AV |
1503 | } |
1504 | retval = rw_verify_area(READ, in.file, &pos, count); | |
e28cc715 | 1505 | if (retval < 0) |
1da177e4 | 1506 | goto fput_in; |
bc61384d AV |
1507 | if (count > MAX_RW_COUNT) |
1508 | count = MAX_RW_COUNT; | |
1da177e4 | 1509 | |
1da177e4 LT |
1510 | /* |
1511 | * Get output file, and verify that it is ok.. | |
1512 | */ | |
1513 | retval = -EBADF; | |
2903ff01 AV |
1514 | out = fdget(out_fd); |
1515 | if (!out.file) | |
1da177e4 | 1516 | goto fput_in; |
2903ff01 | 1517 | if (!(out.file->f_mode & FMODE_WRITE)) |
1da177e4 | 1518 | goto fput_out; |
496ad9aa AV |
1519 | in_inode = file_inode(in.file); |
1520 | out_inode = file_inode(out.file); | |
7995bd28 AV |
1521 | out_pos = out.file->f_pos; |
1522 | retval = rw_verify_area(WRITE, out.file, &out_pos, count); | |
e28cc715 | 1523 | if (retval < 0) |
1da177e4 LT |
1524 | goto fput_out; |
1525 | ||
1da177e4 LT |
1526 | if (!max) |
1527 | max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes); | |
1528 | ||
1da177e4 LT |
1529 | if (unlikely(pos + count > max)) { |
1530 | retval = -EOVERFLOW; | |
1531 | if (pos >= max) | |
1532 | goto fput_out; | |
1533 | count = max - pos; | |
1534 | } | |
1535 | ||
d96e6e71 | 1536 | fl = 0; |
534f2aaa | 1537 | #if 0 |
d96e6e71 JA |
1538 | /* |
1539 | * We need to debate whether we can enable this or not. The | |
1540 | * man page documents EAGAIN return for the output at least, | |
1541 | * and the application is arguably buggy if it doesn't expect | |
1542 | * EAGAIN on a non-blocking file descriptor. | |
1543 | */ | |
2903ff01 | 1544 | if (in.file->f_flags & O_NONBLOCK) |
d96e6e71 | 1545 | fl = SPLICE_F_NONBLOCK; |
534f2aaa | 1546 | #endif |
50cd2c57 | 1547 | file_start_write(out.file); |
7995bd28 | 1548 | retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl); |
50cd2c57 | 1549 | file_end_write(out.file); |
1da177e4 LT |
1550 | |
1551 | if (retval > 0) { | |
4b98d11b AD |
1552 | add_rchar(current, retval); |
1553 | add_wchar(current, retval); | |
a68c2f12 SW |
1554 | fsnotify_access(in.file); |
1555 | fsnotify_modify(out.file); | |
7995bd28 AV |
1556 | out.file->f_pos = out_pos; |
1557 | if (ppos) | |
1558 | *ppos = pos; | |
1559 | else | |
1560 | in.file->f_pos = pos; | |
1da177e4 | 1561 | } |
1da177e4 | 1562 | |
4b98d11b AD |
1563 | inc_syscr(current); |
1564 | inc_syscw(current); | |
7995bd28 | 1565 | if (pos > max) |
1da177e4 LT |
1566 | retval = -EOVERFLOW; |
1567 | ||
1568 | fput_out: | |
2903ff01 | 1569 | fdput(out); |
1da177e4 | 1570 | fput_in: |
2903ff01 | 1571 | fdput(in); |
1da177e4 LT |
1572 | out: |
1573 | return retval; | |
1574 | } | |
1575 | ||
002c8976 | 1576 | SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count) |
1da177e4 LT |
1577 | { |
1578 | loff_t pos; | |
1579 | off_t off; | |
1580 | ssize_t ret; | |
1581 | ||
1582 | if (offset) { | |
1583 | if (unlikely(get_user(off, offset))) | |
1584 | return -EFAULT; | |
1585 | pos = off; | |
1586 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS); | |
1587 | if (unlikely(put_user(pos, offset))) | |
1588 | return -EFAULT; | |
1589 | return ret; | |
1590 | } | |
1591 | ||
1592 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1593 | } | |
1594 | ||
002c8976 | 1595 | SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count) |
1da177e4 LT |
1596 | { |
1597 | loff_t pos; | |
1598 | ssize_t ret; | |
1599 | ||
1600 | if (offset) { | |
1601 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t)))) | |
1602 | return -EFAULT; | |
1603 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0); | |
1604 | if (unlikely(put_user(pos, offset))) | |
1605 | return -EFAULT; | |
1606 | return ret; | |
1607 | } | |
1608 | ||
1609 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1610 | } | |
19f4fc3a AV |
1611 | |
1612 | #ifdef CONFIG_COMPAT | |
1613 | COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, | |
1614 | compat_off_t __user *, offset, compat_size_t, count) | |
1615 | { | |
1616 | loff_t pos; | |
1617 | off_t off; | |
1618 | ssize_t ret; | |
1619 | ||
1620 | if (offset) { | |
1621 | if (unlikely(get_user(off, offset))) | |
1622 | return -EFAULT; | |
1623 | pos = off; | |
1624 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS); | |
1625 | if (unlikely(put_user(pos, offset))) | |
1626 | return -EFAULT; | |
1627 | return ret; | |
1628 | } | |
1629 | ||
1630 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1631 | } | |
1632 | ||
1633 | COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, | |
1634 | compat_loff_t __user *, offset, compat_size_t, count) | |
1635 | { | |
1636 | loff_t pos; | |
1637 | ssize_t ret; | |
1638 | ||
1639 | if (offset) { | |
1640 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t)))) | |
1641 | return -EFAULT; | |
1642 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0); | |
1643 | if (unlikely(put_user(pos, offset))) | |
1644 | return -EFAULT; | |
1645 | return ret; | |
1646 | } | |
1647 | ||
1648 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1649 | } | |
1650 | #endif | |
29732938 | 1651 | |
f16acc9d DC |
1652 | /** |
1653 | * generic_copy_file_range - copy data between two files | |
1654 | * @file_in: file structure to read from | |
1655 | * @pos_in: file offset to read from | |
1656 | * @file_out: file structure to write data to | |
1657 | * @pos_out: file offset to write data to | |
1658 | * @len: amount of data to copy | |
1659 | * @flags: copy flags | |
1660 | * | |
1661 | * This is a generic filesystem helper to copy data from one file to another. | |
1662 | * It has no constraints on the source or destination file owners - the files | |
1663 | * can belong to different superblocks and different filesystem types. Short | |
1664 | * copies are allowed. | |
1665 | * | |
1666 | * This should be called from the @file_out filesystem, as per the | |
1667 | * ->copy_file_range() method. | |
1668 | * | |
1669 | * Returns the number of bytes copied or a negative error indicating the | |
1670 | * failure. | |
1671 | */ | |
1672 | ||
1673 | ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in, | |
1674 | struct file *file_out, loff_t pos_out, | |
1675 | size_t len, unsigned int flags) | |
1676 | { | |
1677 | return do_splice_direct(file_in, &pos_in, file_out, &pos_out, | |
1678 | len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0); | |
1679 | } | |
1680 | EXPORT_SYMBOL(generic_copy_file_range); | |
1681 | ||
64bf5ff5 DC |
1682 | static ssize_t do_copy_file_range(struct file *file_in, loff_t pos_in, |
1683 | struct file *file_out, loff_t pos_out, | |
1684 | size_t len, unsigned int flags) | |
1685 | { | |
5dae222a AG |
1686 | /* |
1687 | * Although we now allow filesystems to handle cross sb copy, passing | |
1688 | * a file of the wrong filesystem type to filesystem driver can result | |
1689 | * in an attempt to dereference the wrong type of ->private_data, so | |
1690 | * avoid doing that until we really have a good reason. NFS defines | |
1691 | * several different file_system_type structures, but they all end up | |
1692 | * using the same ->copy_file_range() function pointer. | |
1693 | */ | |
1694 | if (file_out->f_op->copy_file_range && | |
1695 | file_out->f_op->copy_file_range == file_in->f_op->copy_file_range) | |
64bf5ff5 DC |
1696 | return file_out->f_op->copy_file_range(file_in, pos_in, |
1697 | file_out, pos_out, | |
1698 | len, flags); | |
1699 | ||
1700 | return generic_copy_file_range(file_in, pos_in, file_out, pos_out, len, | |
1701 | flags); | |
1702 | } | |
1703 | ||
29732938 ZB |
1704 | /* |
1705 | * copy_file_range() differs from regular file read and write in that it | |
1706 | * specifically allows return partial success. When it does so is up to | |
1707 | * the copy_file_range method. | |
1708 | */ | |
1709 | ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in, | |
1710 | struct file *file_out, loff_t pos_out, | |
1711 | size_t len, unsigned int flags) | |
1712 | { | |
29732938 ZB |
1713 | ssize_t ret; |
1714 | ||
1715 | if (flags != 0) | |
1716 | return -EINVAL; | |
1717 | ||
96e6e8f4 AG |
1718 | ret = generic_copy_file_checks(file_in, pos_in, file_out, pos_out, &len, |
1719 | flags); | |
a3171351 AG |
1720 | if (unlikely(ret)) |
1721 | return ret; | |
11cbfb10 | 1722 | |
29732938 | 1723 | ret = rw_verify_area(READ, file_in, &pos_in, len); |
bc61384d AV |
1724 | if (unlikely(ret)) |
1725 | return ret; | |
1726 | ||
1727 | ret = rw_verify_area(WRITE, file_out, &pos_out, len); | |
1728 | if (unlikely(ret)) | |
29732938 ZB |
1729 | return ret; |
1730 | ||
29732938 ZB |
1731 | if (len == 0) |
1732 | return 0; | |
1733 | ||
bfe219d3 | 1734 | file_start_write(file_out); |
29732938 | 1735 | |
a76b5b04 CH |
1736 | /* |
1737 | * Try cloning first, this is supported by more file systems, and | |
1738 | * more efficient if both clone and copy are supported (e.g. NFS). | |
1739 | */ | |
5dae222a AG |
1740 | if (file_in->f_op->remap_file_range && |
1741 | file_inode(file_in)->i_sb == file_inode(file_out)->i_sb) { | |
42ec3d4c DW |
1742 | loff_t cloned; |
1743 | ||
1744 | cloned = file_in->f_op->remap_file_range(file_in, pos_in, | |
1745 | file_out, pos_out, | |
eca3654e DW |
1746 | min_t(loff_t, MAX_RW_COUNT, len), |
1747 | REMAP_FILE_CAN_SHORTEN); | |
42ec3d4c DW |
1748 | if (cloned > 0) { |
1749 | ret = cloned; | |
a76b5b04 CH |
1750 | goto done; |
1751 | } | |
1752 | } | |
1753 | ||
64bf5ff5 DC |
1754 | ret = do_copy_file_range(file_in, pos_in, file_out, pos_out, len, |
1755 | flags); | |
1756 | WARN_ON_ONCE(ret == -EOPNOTSUPP); | |
a76b5b04 | 1757 | done: |
29732938 ZB |
1758 | if (ret > 0) { |
1759 | fsnotify_access(file_in); | |
1760 | add_rchar(current, ret); | |
1761 | fsnotify_modify(file_out); | |
1762 | add_wchar(current, ret); | |
1763 | } | |
a76b5b04 | 1764 | |
29732938 ZB |
1765 | inc_syscr(current); |
1766 | inc_syscw(current); | |
1767 | ||
bfe219d3 | 1768 | file_end_write(file_out); |
29732938 ZB |
1769 | |
1770 | return ret; | |
1771 | } | |
1772 | EXPORT_SYMBOL(vfs_copy_file_range); | |
1773 | ||
1774 | SYSCALL_DEFINE6(copy_file_range, int, fd_in, loff_t __user *, off_in, | |
1775 | int, fd_out, loff_t __user *, off_out, | |
1776 | size_t, len, unsigned int, flags) | |
1777 | { | |
1778 | loff_t pos_in; | |
1779 | loff_t pos_out; | |
1780 | struct fd f_in; | |
1781 | struct fd f_out; | |
1782 | ssize_t ret = -EBADF; | |
1783 | ||
1784 | f_in = fdget(fd_in); | |
1785 | if (!f_in.file) | |
1786 | goto out2; | |
1787 | ||
1788 | f_out = fdget(fd_out); | |
1789 | if (!f_out.file) | |
1790 | goto out1; | |
1791 | ||
1792 | ret = -EFAULT; | |
1793 | if (off_in) { | |
1794 | if (copy_from_user(&pos_in, off_in, sizeof(loff_t))) | |
1795 | goto out; | |
1796 | } else { | |
1797 | pos_in = f_in.file->f_pos; | |
1798 | } | |
1799 | ||
1800 | if (off_out) { | |
1801 | if (copy_from_user(&pos_out, off_out, sizeof(loff_t))) | |
1802 | goto out; | |
1803 | } else { | |
1804 | pos_out = f_out.file->f_pos; | |
1805 | } | |
1806 | ||
1807 | ret = vfs_copy_file_range(f_in.file, pos_in, f_out.file, pos_out, len, | |
1808 | flags); | |
1809 | if (ret > 0) { | |
1810 | pos_in += ret; | |
1811 | pos_out += ret; | |
1812 | ||
1813 | if (off_in) { | |
1814 | if (copy_to_user(off_in, &pos_in, sizeof(loff_t))) | |
1815 | ret = -EFAULT; | |
1816 | } else { | |
1817 | f_in.file->f_pos = pos_in; | |
1818 | } | |
1819 | ||
1820 | if (off_out) { | |
1821 | if (copy_to_user(off_out, &pos_out, sizeof(loff_t))) | |
1822 | ret = -EFAULT; | |
1823 | } else { | |
1824 | f_out.file->f_pos = pos_out; | |
1825 | } | |
1826 | } | |
1827 | ||
1828 | out: | |
1829 | fdput(f_out); | |
1830 | out1: | |
1831 | fdput(f_in); | |
1832 | out2: | |
1833 | return ret; | |
1834 | } | |
04b38d60 | 1835 | |
42ec3d4c DW |
1836 | static int remap_verify_area(struct file *file, loff_t pos, loff_t len, |
1837 | bool write) | |
04b38d60 CH |
1838 | { |
1839 | struct inode *inode = file_inode(file); | |
1840 | ||
42ec3d4c | 1841 | if (unlikely(pos < 0 || len < 0)) |
04b38d60 CH |
1842 | return -EINVAL; |
1843 | ||
1844 | if (unlikely((loff_t) (pos + len) < 0)) | |
1845 | return -EINVAL; | |
1846 | ||
1847 | if (unlikely(inode->i_flctx && mandatory_lock(inode))) { | |
1848 | loff_t end = len ? pos + len - 1 : OFFSET_MAX; | |
1849 | int retval; | |
1850 | ||
1851 | retval = locks_mandatory_area(inode, file, pos, end, | |
1852 | write ? F_WRLCK : F_RDLCK); | |
1853 | if (retval < 0) | |
1854 | return retval; | |
1855 | } | |
1856 | ||
1857 | return security_file_permission(file, write ? MAY_WRITE : MAY_READ); | |
1858 | } | |
07d19dc9 DW |
1859 | /* |
1860 | * Ensure that we don't remap a partial EOF block in the middle of something | |
1861 | * else. Assume that the offsets have already been checked for block | |
1862 | * alignment. | |
1863 | * | |
a5e6ea18 FM |
1864 | * For clone we only link a partial EOF block above or at the destination file's |
1865 | * EOF. For deduplication we accept a partial EOF block only if it ends at the | |
1866 | * destination file's EOF (can not link it into the middle of a file). | |
eca3654e DW |
1867 | * |
1868 | * Shorten the request if possible. | |
07d19dc9 DW |
1869 | */ |
1870 | static int generic_remap_check_len(struct inode *inode_in, | |
1871 | struct inode *inode_out, | |
1872 | loff_t pos_out, | |
42ec3d4c | 1873 | loff_t *len, |
a91ae49b | 1874 | unsigned int remap_flags) |
07d19dc9 DW |
1875 | { |
1876 | u64 blkmask = i_blocksize(inode_in) - 1; | |
eca3654e | 1877 | loff_t new_len = *len; |
07d19dc9 DW |
1878 | |
1879 | if ((*len & blkmask) == 0) | |
1880 | return 0; | |
1881 | ||
a5e6ea18 | 1882 | if (pos_out + *len < i_size_read(inode_out)) |
eca3654e | 1883 | new_len &= ~blkmask; |
07d19dc9 | 1884 | |
eca3654e DW |
1885 | if (new_len == *len) |
1886 | return 0; | |
1887 | ||
1888 | if (remap_flags & REMAP_FILE_CAN_SHORTEN) { | |
1889 | *len = new_len; | |
1890 | return 0; | |
1891 | } | |
1892 | ||
1893 | return (remap_flags & REMAP_FILE_DEDUP) ? -EBADE : -EINVAL; | |
07d19dc9 | 1894 | } |
04b38d60 | 1895 | |
edc58dd0 | 1896 | /* Read a page's worth of file data into the page cache. */ |
c32e5f39 DW |
1897 | static struct page *vfs_dedupe_get_page(struct inode *inode, loff_t offset) |
1898 | { | |
1899 | struct page *page; | |
1900 | ||
1901 | page = read_mapping_page(inode->i_mapping, offset >> PAGE_SHIFT, NULL); | |
1902 | if (IS_ERR(page)) | |
1903 | return page; | |
1904 | if (!PageUptodate(page)) { | |
1905 | put_page(page); | |
1906 | return ERR_PTR(-EIO); | |
1907 | } | |
c32e5f39 DW |
1908 | return page; |
1909 | } | |
1910 | ||
edc58dd0 DW |
1911 | /* |
1912 | * Lock two pages, ensuring that we lock in offset order if the pages are from | |
1913 | * the same file. | |
1914 | */ | |
1915 | static void vfs_lock_two_pages(struct page *page1, struct page *page2) | |
1916 | { | |
1917 | /* Always lock in order of increasing index. */ | |
1918 | if (page1->index > page2->index) | |
1919 | swap(page1, page2); | |
1920 | ||
1921 | lock_page(page1); | |
1922 | if (page1 != page2) | |
1923 | lock_page(page2); | |
1924 | } | |
1925 | ||
1926 | /* Unlock two pages, being careful not to unlock the same page twice. */ | |
1927 | static void vfs_unlock_two_pages(struct page *page1, struct page *page2) | |
1928 | { | |
1929 | unlock_page(page1); | |
1930 | if (page1 != page2) | |
1931 | unlock_page(page2); | |
1932 | } | |
1933 | ||
c32e5f39 DW |
1934 | /* |
1935 | * Compare extents of two files to see if they are the same. | |
1936 | * Caller must have locked both inodes to prevent write races. | |
1937 | */ | |
1938 | static int vfs_dedupe_file_range_compare(struct inode *src, loff_t srcoff, | |
1939 | struct inode *dest, loff_t destoff, | |
1940 | loff_t len, bool *is_same) | |
1941 | { | |
1942 | loff_t src_poff; | |
1943 | loff_t dest_poff; | |
1944 | void *src_addr; | |
1945 | void *dest_addr; | |
1946 | struct page *src_page; | |
1947 | struct page *dest_page; | |
1948 | loff_t cmp_len; | |
1949 | bool same; | |
1950 | int error; | |
1951 | ||
1952 | error = -EINVAL; | |
1953 | same = true; | |
1954 | while (len) { | |
1955 | src_poff = srcoff & (PAGE_SIZE - 1); | |
1956 | dest_poff = destoff & (PAGE_SIZE - 1); | |
1957 | cmp_len = min(PAGE_SIZE - src_poff, | |
1958 | PAGE_SIZE - dest_poff); | |
1959 | cmp_len = min(cmp_len, len); | |
1960 | if (cmp_len <= 0) | |
1961 | goto out_error; | |
1962 | ||
1963 | src_page = vfs_dedupe_get_page(src, srcoff); | |
1964 | if (IS_ERR(src_page)) { | |
1965 | error = PTR_ERR(src_page); | |
1966 | goto out_error; | |
1967 | } | |
1968 | dest_page = vfs_dedupe_get_page(dest, destoff); | |
1969 | if (IS_ERR(dest_page)) { | |
1970 | error = PTR_ERR(dest_page); | |
c32e5f39 DW |
1971 | put_page(src_page); |
1972 | goto out_error; | |
1973 | } | |
edc58dd0 DW |
1974 | |
1975 | vfs_lock_two_pages(src_page, dest_page); | |
1976 | ||
1977 | /* | |
1978 | * Now that we've locked both pages, make sure they're still | |
1979 | * mapped to the file data we're interested in. If not, | |
1980 | * someone is invalidating pages on us and we lose. | |
1981 | */ | |
1982 | if (!PageUptodate(src_page) || !PageUptodate(dest_page) || | |
1983 | src_page->mapping != src->i_mapping || | |
1984 | dest_page->mapping != dest->i_mapping) { | |
1985 | same = false; | |
1986 | goto unlock; | |
1987 | } | |
1988 | ||
c32e5f39 DW |
1989 | src_addr = kmap_atomic(src_page); |
1990 | dest_addr = kmap_atomic(dest_page); | |
1991 | ||
1992 | flush_dcache_page(src_page); | |
1993 | flush_dcache_page(dest_page); | |
1994 | ||
1995 | if (memcmp(src_addr + src_poff, dest_addr + dest_poff, cmp_len)) | |
1996 | same = false; | |
1997 | ||
1998 | kunmap_atomic(dest_addr); | |
1999 | kunmap_atomic(src_addr); | |
edc58dd0 DW |
2000 | unlock: |
2001 | vfs_unlock_two_pages(src_page, dest_page); | |
c32e5f39 DW |
2002 | put_page(dest_page); |
2003 | put_page(src_page); | |
2004 | ||
2005 | if (!same) | |
2006 | break; | |
2007 | ||
2008 | srcoff += cmp_len; | |
2009 | destoff += cmp_len; | |
2010 | len -= cmp_len; | |
2011 | } | |
2012 | ||
2013 | *is_same = same; | |
2014 | return 0; | |
2015 | ||
2016 | out_error: | |
2017 | return error; | |
2018 | } | |
04b38d60 | 2019 | |
876bec6f DW |
2020 | /* |
2021 | * Check that the two inodes are eligible for cloning, the ranges make | |
2022 | * sense, and then flush all dirty data. Caller must ensure that the | |
2023 | * inodes have been locked against any other modifications. | |
22725ce4 | 2024 | * |
8c5c836b DW |
2025 | * If there's an error, then the usual negative error code is returned. |
2026 | * Otherwise returns 0 with *len set to the request length. | |
876bec6f | 2027 | */ |
a83ab01a DW |
2028 | int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, |
2029 | struct file *file_out, loff_t pos_out, | |
42ec3d4c | 2030 | loff_t *len, unsigned int remap_flags) |
876bec6f | 2031 | { |
1383a7ed DW |
2032 | struct inode *inode_in = file_inode(file_in); |
2033 | struct inode *inode_out = file_inode(file_out); | |
876bec6f DW |
2034 | bool same_inode = (inode_in == inode_out); |
2035 | int ret; | |
2036 | ||
2037 | /* Don't touch certain kinds of inodes */ | |
2038 | if (IS_IMMUTABLE(inode_out)) | |
2039 | return -EPERM; | |
2040 | ||
2041 | if (IS_SWAPFILE(inode_in) || IS_SWAPFILE(inode_out)) | |
2042 | return -ETXTBSY; | |
2043 | ||
2044 | /* Don't reflink dirs, pipes, sockets... */ | |
2045 | if (S_ISDIR(inode_in->i_mode) || S_ISDIR(inode_out->i_mode)) | |
2046 | return -EISDIR; | |
2047 | if (!S_ISREG(inode_in->i_mode) || !S_ISREG(inode_out->i_mode)) | |
2048 | return -EINVAL; | |
2049 | ||
876bec6f DW |
2050 | /* Zero length dedupe exits immediately; reflink goes to EOF. */ |
2051 | if (*len == 0) { | |
1383a7ed DW |
2052 | loff_t isize = i_size_read(inode_in); |
2053 | ||
a91ae49b | 2054 | if ((remap_flags & REMAP_FILE_DEDUP) || pos_in == isize) |
876bec6f | 2055 | return 0; |
22725ce4 DW |
2056 | if (pos_in > isize) |
2057 | return -EINVAL; | |
876bec6f | 2058 | *len = isize - pos_in; |
2c5773f1 DW |
2059 | if (*len == 0) |
2060 | return 0; | |
876bec6f DW |
2061 | } |
2062 | ||
1383a7ed DW |
2063 | /* Check that we don't violate system file offset limits. */ |
2064 | ret = generic_remap_checks(file_in, pos_in, file_out, pos_out, len, | |
3d28193e | 2065 | remap_flags); |
1383a7ed DW |
2066 | if (ret) |
2067 | return ret; | |
876bec6f DW |
2068 | |
2069 | /* Wait for the completion of any pending IOs on both files */ | |
2070 | inode_dio_wait(inode_in); | |
2071 | if (!same_inode) | |
2072 | inode_dio_wait(inode_out); | |
2073 | ||
2074 | ret = filemap_write_and_wait_range(inode_in->i_mapping, | |
2075 | pos_in, pos_in + *len - 1); | |
2076 | if (ret) | |
2077 | return ret; | |
2078 | ||
2079 | ret = filemap_write_and_wait_range(inode_out->i_mapping, | |
2080 | pos_out, pos_out + *len - 1); | |
2081 | if (ret) | |
2082 | return ret; | |
2083 | ||
2084 | /* | |
2085 | * Check that the extents are the same. | |
2086 | */ | |
a91ae49b | 2087 | if (remap_flags & REMAP_FILE_DEDUP) { |
876bec6f DW |
2088 | bool is_same = false; |
2089 | ||
2090 | ret = vfs_dedupe_file_range_compare(inode_in, pos_in, | |
2091 | inode_out, pos_out, *len, &is_same); | |
2092 | if (ret) | |
2093 | return ret; | |
2094 | if (!is_same) | |
2095 | return -EBADE; | |
2096 | } | |
2097 | ||
07d19dc9 | 2098 | ret = generic_remap_check_len(inode_in, inode_out, pos_out, len, |
a91ae49b | 2099 | remap_flags); |
07d19dc9 DW |
2100 | if (ret) |
2101 | return ret; | |
2102 | ||
8dde90bc | 2103 | /* If can't alter the file contents, we're done. */ |
e38f7f53 AG |
2104 | if (!(remap_flags & REMAP_FILE_DEDUP)) |
2105 | ret = file_modified(file_out); | |
8dde90bc | 2106 | |
e38f7f53 | 2107 | return ret; |
876bec6f | 2108 | } |
a83ab01a | 2109 | EXPORT_SYMBOL(generic_remap_file_range_prep); |
876bec6f | 2110 | |
42ec3d4c | 2111 | loff_t do_clone_file_range(struct file *file_in, loff_t pos_in, |
452ce659 DW |
2112 | struct file *file_out, loff_t pos_out, |
2113 | loff_t len, unsigned int remap_flags) | |
04b38d60 | 2114 | { |
42ec3d4c | 2115 | loff_t ret; |
04b38d60 | 2116 | |
6744557b | 2117 | WARN_ON_ONCE(remap_flags & REMAP_FILE_DEDUP); |
04b38d60 | 2118 | |
913b86e9 AG |
2119 | /* |
2120 | * FICLONE/FICLONERANGE ioctls enforce that src and dest files are on | |
2121 | * the same mount. Practically, they only need to be on the same file | |
2122 | * system. | |
2123 | */ | |
a3171351 | 2124 | if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb) |
04b38d60 CH |
2125 | return -EXDEV; |
2126 | ||
a3171351 AG |
2127 | ret = generic_file_rw_checks(file_in, file_out); |
2128 | if (ret < 0) | |
2129 | return ret; | |
04b38d60 | 2130 | |
2e5dfc99 | 2131 | if (!file_in->f_op->remap_file_range) |
0fcbf996 CH |
2132 | return -EOPNOTSUPP; |
2133 | ||
6095028b | 2134 | ret = remap_verify_area(file_in, pos_in, len, false); |
04b38d60 CH |
2135 | if (ret) |
2136 | return ret; | |
2137 | ||
6095028b | 2138 | ret = remap_verify_area(file_out, pos_out, len, true); |
04b38d60 CH |
2139 | if (ret) |
2140 | return ret; | |
2141 | ||
2e5dfc99 | 2142 | ret = file_in->f_op->remap_file_range(file_in, pos_in, |
452ce659 | 2143 | file_out, pos_out, len, remap_flags); |
42ec3d4c DW |
2144 | if (ret < 0) |
2145 | return ret; | |
04b38d60 | 2146 | |
42ec3d4c DW |
2147 | fsnotify_access(file_in); |
2148 | fsnotify_modify(file_out); | |
04b38d60 CH |
2149 | return ret; |
2150 | } | |
a725356b AG |
2151 | EXPORT_SYMBOL(do_clone_file_range); |
2152 | ||
42ec3d4c | 2153 | loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, |
452ce659 DW |
2154 | struct file *file_out, loff_t pos_out, |
2155 | loff_t len, unsigned int remap_flags) | |
a725356b | 2156 | { |
42ec3d4c | 2157 | loff_t ret; |
a725356b AG |
2158 | |
2159 | file_start_write(file_out); | |
452ce659 DW |
2160 | ret = do_clone_file_range(file_in, pos_in, file_out, pos_out, len, |
2161 | remap_flags); | |
a725356b AG |
2162 | file_end_write(file_out); |
2163 | ||
2164 | return ret; | |
2165 | } | |
04b38d60 | 2166 | EXPORT_SYMBOL(vfs_clone_file_range); |
54dbc151 | 2167 | |
5de4480a MF |
2168 | /* Check whether we are allowed to dedupe the destination file */ |
2169 | static bool allow_file_dedupe(struct file *file) | |
2170 | { | |
2171 | if (capable(CAP_SYS_ADMIN)) | |
2172 | return true; | |
2173 | if (file->f_mode & FMODE_WRITE) | |
2174 | return true; | |
2175 | if (uid_eq(current_fsuid(), file_inode(file)->i_uid)) | |
2176 | return true; | |
2177 | if (!inode_permission(file_inode(file), MAY_WRITE)) | |
2178 | return true; | |
2179 | return false; | |
2180 | } | |
2181 | ||
42ec3d4c DW |
2182 | loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, |
2183 | struct file *dst_file, loff_t dst_pos, | |
df365836 | 2184 | loff_t len, unsigned int remap_flags) |
1b4f42a1 | 2185 | { |
42ec3d4c | 2186 | loff_t ret; |
1b4f42a1 | 2187 | |
eca3654e DW |
2188 | WARN_ON_ONCE(remap_flags & ~(REMAP_FILE_DEDUP | |
2189 | REMAP_FILE_CAN_SHORTEN)); | |
1b4f42a1 MS |
2190 | |
2191 | ret = mnt_want_write_file(dst_file); | |
2192 | if (ret) | |
2193 | return ret; | |
2194 | ||
6095028b | 2195 | ret = remap_verify_area(dst_file, dst_pos, len, true); |
1b4f42a1 MS |
2196 | if (ret < 0) |
2197 | goto out_drop_write; | |
2198 | ||
85c95f20 | 2199 | ret = -EPERM; |
5de4480a | 2200 | if (!allow_file_dedupe(dst_file)) |
1b4f42a1 MS |
2201 | goto out_drop_write; |
2202 | ||
2203 | ret = -EXDEV; | |
2204 | if (src_file->f_path.mnt != dst_file->f_path.mnt) | |
2205 | goto out_drop_write; | |
2206 | ||
2207 | ret = -EISDIR; | |
2208 | if (S_ISDIR(file_inode(dst_file)->i_mode)) | |
2209 | goto out_drop_write; | |
2210 | ||
2211 | ret = -EINVAL; | |
2e5dfc99 | 2212 | if (!dst_file->f_op->remap_file_range) |
1b4f42a1 MS |
2213 | goto out_drop_write; |
2214 | ||
9aae2050 DW |
2215 | if (len == 0) { |
2216 | ret = 0; | |
2217 | goto out_drop_write; | |
2218 | } | |
2219 | ||
2e5dfc99 | 2220 | ret = dst_file->f_op->remap_file_range(src_file, src_pos, dst_file, |
df365836 | 2221 | dst_pos, len, remap_flags | REMAP_FILE_DEDUP); |
1b4f42a1 MS |
2222 | out_drop_write: |
2223 | mnt_drop_write_file(dst_file); | |
2224 | ||
2225 | return ret; | |
2226 | } | |
f1825366 | 2227 | EXPORT_SYMBOL(vfs_dedupe_file_range_one); |
1b4f42a1 | 2228 | |
54dbc151 DW |
2229 | int vfs_dedupe_file_range(struct file *file, struct file_dedupe_range *same) |
2230 | { | |
2231 | struct file_dedupe_range_info *info; | |
2232 | struct inode *src = file_inode(file); | |
2233 | u64 off; | |
2234 | u64 len; | |
2235 | int i; | |
2236 | int ret; | |
54dbc151 | 2237 | u16 count = same->dest_count; |
42ec3d4c | 2238 | loff_t deduped; |
54dbc151 DW |
2239 | |
2240 | if (!(file->f_mode & FMODE_READ)) | |
2241 | return -EINVAL; | |
2242 | ||
2243 | if (same->reserved1 || same->reserved2) | |
2244 | return -EINVAL; | |
2245 | ||
2246 | off = same->src_offset; | |
2247 | len = same->src_length; | |
2248 | ||
54dbc151 | 2249 | if (S_ISDIR(src->i_mode)) |
494633fa | 2250 | return -EISDIR; |
54dbc151 | 2251 | |
54dbc151 | 2252 | if (!S_ISREG(src->i_mode)) |
494633fa DC |
2253 | return -EINVAL; |
2254 | ||
2255 | if (!file->f_op->remap_file_range) | |
2256 | return -EOPNOTSUPP; | |
54dbc151 | 2257 | |
6095028b | 2258 | ret = remap_verify_area(file, off, len, false); |
54dbc151 | 2259 | if (ret < 0) |
494633fa | 2260 | return ret; |
54dbc151 DW |
2261 | ret = 0; |
2262 | ||
22725ce4 DW |
2263 | if (off + len > i_size_read(src)) |
2264 | return -EINVAL; | |
2265 | ||
92b66d2c MS |
2266 | /* Arbitrary 1G limit on a single dedupe request, can be raised. */ |
2267 | len = min_t(u64, len, 1 << 30); | |
2268 | ||
54dbc151 DW |
2269 | /* pre-format output fields to sane values */ |
2270 | for (i = 0; i < count; i++) { | |
2271 | same->info[i].bytes_deduped = 0ULL; | |
2272 | same->info[i].status = FILE_DEDUPE_RANGE_SAME; | |
2273 | } | |
2274 | ||
2275 | for (i = 0, info = same->info; i < count; i++, info++) { | |
54dbc151 | 2276 | struct fd dst_fd = fdget(info->dest_fd); |
1b4f42a1 | 2277 | struct file *dst_file = dst_fd.file; |
54dbc151 | 2278 | |
54dbc151 DW |
2279 | if (!dst_file) { |
2280 | info->status = -EBADF; | |
2281 | goto next_loop; | |
2282 | } | |
54dbc151 DW |
2283 | |
2284 | if (info->reserved) { | |
2285 | info->status = -EINVAL; | |
1b4f42a1 | 2286 | goto next_fdput; |
54dbc151 DW |
2287 | } |
2288 | ||
1b4f42a1 | 2289 | deduped = vfs_dedupe_file_range_one(file, off, dst_file, |
df365836 | 2290 | info->dest_offset, len, |
eca3654e | 2291 | REMAP_FILE_CAN_SHORTEN); |
1b4f42a1 MS |
2292 | if (deduped == -EBADE) |
2293 | info->status = FILE_DEDUPE_RANGE_DIFFERS; | |
2294 | else if (deduped < 0) | |
2295 | info->status = deduped; | |
2296 | else | |
2297 | info->bytes_deduped = len; | |
2298 | ||
22762711 | 2299 | next_fdput: |
54dbc151 | 2300 | fdput(dst_fd); |
22762711 | 2301 | next_loop: |
e62e560f | 2302 | if (fatal_signal_pending(current)) |
494633fa | 2303 | break; |
54dbc151 | 2304 | } |
54dbc151 DW |
2305 | return ret; |
2306 | } | |
2307 | EXPORT_SYMBOL(vfs_dedupe_file_range); |