Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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> | |
0eeca283 | 12 | #include <linux/fsnotify.h> |
1da177e4 | 13 | #include <linux/security.h> |
630d9c47 | 14 | #include <linux/export.h> |
1da177e4 | 15 | #include <linux/syscalls.h> |
e28cc715 | 16 | #include <linux/pagemap.h> |
d6b29d7c | 17 | #include <linux/splice.h> |
561c6731 | 18 | #include <linux/compat.h> |
ee0b3e67 | 19 | #include "read_write.h" |
1da177e4 LT |
20 | |
21 | #include <asm/uaccess.h> | |
22 | #include <asm/unistd.h> | |
23 | ||
4b6f5d20 | 24 | const struct file_operations generic_ro_fops = { |
1da177e4 | 25 | .llseek = generic_file_llseek, |
543ade1f BP |
26 | .read = do_sync_read, |
27 | .aio_read = generic_file_aio_read, | |
1da177e4 | 28 | .mmap = generic_file_readonly_mmap, |
534f2aaa | 29 | .splice_read = generic_file_splice_read, |
1da177e4 LT |
30 | }; |
31 | ||
32 | EXPORT_SYMBOL(generic_ro_fops); | |
33 | ||
cccb5a1e | 34 | static inline int unsigned_offsets(struct file *file) |
4a3956c7 | 35 | { |
cccb5a1e | 36 | return file->f_mode & FMODE_UNSIGNED_OFFSET; |
4a3956c7 KH |
37 | } |
38 | ||
ef3d0fd2 AK |
39 | static loff_t lseek_execute(struct file *file, struct inode *inode, |
40 | loff_t offset, loff_t maxsize) | |
41 | { | |
42 | if (offset < 0 && !unsigned_offsets(file)) | |
43 | return -EINVAL; | |
44 | if (offset > maxsize) | |
45 | return -EINVAL; | |
46 | ||
47 | if (offset != file->f_pos) { | |
48 | file->f_pos = offset; | |
49 | file->f_version = 0; | |
50 | } | |
51 | return offset; | |
52 | } | |
53 | ||
3a8cff4f | 54 | /** |
5760495a | 55 | * generic_file_llseek_size - generic llseek implementation for regular files |
3a8cff4f CH |
56 | * @file: file structure to seek on |
57 | * @offset: file offset to seek to | |
965c8e59 | 58 | * @whence: type of seek |
e8b96eb5 ES |
59 | * @size: max size of this file in file system |
60 | * @eof: offset used for SEEK_END position | |
3a8cff4f | 61 | * |
5760495a | 62 | * This is a variant of generic_file_llseek that allows passing in a custom |
e8b96eb5 | 63 | * maximum file size and a custom EOF position, for e.g. hashed directories |
ef3d0fd2 AK |
64 | * |
65 | * Synchronization: | |
5760495a | 66 | * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms) |
ef3d0fd2 AK |
67 | * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes. |
68 | * read/writes behave like SEEK_SET against seeks. | |
3a8cff4f | 69 | */ |
9465efc9 | 70 | loff_t |
965c8e59 | 71 | generic_file_llseek_size(struct file *file, loff_t offset, int whence, |
e8b96eb5 | 72 | loff_t maxsize, loff_t eof) |
1da177e4 | 73 | { |
1da177e4 LT |
74 | struct inode *inode = file->f_mapping->host; |
75 | ||
965c8e59 | 76 | switch (whence) { |
3a8cff4f | 77 | case SEEK_END: |
e8b96eb5 | 78 | offset += eof; |
3a8cff4f CH |
79 | break; |
80 | case SEEK_CUR: | |
5b6f1eb9 AK |
81 | /* |
82 | * Here we special-case the lseek(fd, 0, SEEK_CUR) | |
83 | * position-querying operation. Avoid rewriting the "same" | |
84 | * f_pos value back to the file because a concurrent read(), | |
85 | * write() or lseek() might have altered it | |
86 | */ | |
87 | if (offset == 0) | |
88 | return file->f_pos; | |
ef3d0fd2 AK |
89 | /* |
90 | * f_lock protects against read/modify/write race with other | |
91 | * SEEK_CURs. Note that parallel writes and reads behave | |
92 | * like SEEK_SET. | |
93 | */ | |
94 | spin_lock(&file->f_lock); | |
95 | offset = lseek_execute(file, inode, file->f_pos + offset, | |
5760495a | 96 | maxsize); |
ef3d0fd2 AK |
97 | spin_unlock(&file->f_lock); |
98 | return offset; | |
982d8165 JB |
99 | case SEEK_DATA: |
100 | /* | |
101 | * In the generic case the entire file is data, so as long as | |
102 | * offset isn't at the end of the file then the offset is data. | |
103 | */ | |
e8b96eb5 | 104 | if (offset >= eof) |
982d8165 JB |
105 | return -ENXIO; |
106 | break; | |
107 | case SEEK_HOLE: | |
108 | /* | |
109 | * There is a virtual hole at the end of the file, so as long as | |
110 | * offset isn't i_size or larger, return i_size. | |
111 | */ | |
e8b96eb5 | 112 | if (offset >= eof) |
982d8165 | 113 | return -ENXIO; |
e8b96eb5 | 114 | offset = eof; |
982d8165 | 115 | break; |
1da177e4 | 116 | } |
3a8cff4f | 117 | |
5760495a AK |
118 | return lseek_execute(file, inode, offset, maxsize); |
119 | } | |
120 | EXPORT_SYMBOL(generic_file_llseek_size); | |
121 | ||
122 | /** | |
123 | * generic_file_llseek - generic llseek implementation for regular files | |
124 | * @file: file structure to seek on | |
125 | * @offset: file offset to seek to | |
965c8e59 | 126 | * @whence: type of seek |
5760495a AK |
127 | * |
128 | * This is a generic implemenation of ->llseek useable for all normal local | |
129 | * filesystems. It just updates the file offset to the value specified by | |
965c8e59 | 130 | * @offset and @whence under i_mutex. |
5760495a | 131 | */ |
965c8e59 | 132 | loff_t generic_file_llseek(struct file *file, loff_t offset, int whence) |
5760495a AK |
133 | { |
134 | struct inode *inode = file->f_mapping->host; | |
135 | ||
965c8e59 | 136 | return generic_file_llseek_size(file, offset, whence, |
e8b96eb5 ES |
137 | inode->i_sb->s_maxbytes, |
138 | i_size_read(inode)); | |
1da177e4 | 139 | } |
9465efc9 | 140 | EXPORT_SYMBOL(generic_file_llseek); |
1da177e4 | 141 | |
ae6afc3f B |
142 | /** |
143 | * noop_llseek - No Operation Performed llseek implementation | |
144 | * @file: file structure to seek on | |
145 | * @offset: file offset to seek to | |
965c8e59 | 146 | * @whence: type of seek |
ae6afc3f B |
147 | * |
148 | * This is an implementation of ->llseek useable for the rare special case when | |
149 | * userspace expects the seek to succeed but the (device) file is actually not | |
150 | * able to perform the seek. In this case you use noop_llseek() instead of | |
151 | * falling back to the default implementation of ->llseek. | |
152 | */ | |
965c8e59 | 153 | loff_t noop_llseek(struct file *file, loff_t offset, int whence) |
ae6afc3f B |
154 | { |
155 | return file->f_pos; | |
156 | } | |
157 | EXPORT_SYMBOL(noop_llseek); | |
158 | ||
965c8e59 | 159 | loff_t no_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
160 | { |
161 | return -ESPIPE; | |
162 | } | |
163 | EXPORT_SYMBOL(no_llseek); | |
164 | ||
965c8e59 | 165 | loff_t default_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 166 | { |
496ad9aa | 167 | struct inode *inode = file_inode(file); |
16abef0e | 168 | loff_t retval; |
1da177e4 | 169 | |
982d8165 | 170 | mutex_lock(&inode->i_mutex); |
965c8e59 | 171 | switch (whence) { |
7b8e8924 | 172 | case SEEK_END: |
982d8165 | 173 | offset += i_size_read(inode); |
1da177e4 | 174 | break; |
7b8e8924 | 175 | case SEEK_CUR: |
5b6f1eb9 AK |
176 | if (offset == 0) { |
177 | retval = file->f_pos; | |
178 | goto out; | |
179 | } | |
1da177e4 | 180 | offset += file->f_pos; |
982d8165 JB |
181 | break; |
182 | case SEEK_DATA: | |
183 | /* | |
184 | * In the generic case the entire file is data, so as | |
185 | * long as offset isn't at the end of the file then the | |
186 | * offset is data. | |
187 | */ | |
bacb2d81 DC |
188 | if (offset >= inode->i_size) { |
189 | retval = -ENXIO; | |
190 | goto out; | |
191 | } | |
982d8165 JB |
192 | break; |
193 | case SEEK_HOLE: | |
194 | /* | |
195 | * There is a virtual hole at the end of the file, so | |
196 | * as long as offset isn't i_size or larger, return | |
197 | * i_size. | |
198 | */ | |
bacb2d81 DC |
199 | if (offset >= inode->i_size) { |
200 | retval = -ENXIO; | |
201 | goto out; | |
202 | } | |
982d8165 JB |
203 | offset = inode->i_size; |
204 | break; | |
1da177e4 LT |
205 | } |
206 | retval = -EINVAL; | |
cccb5a1e | 207 | if (offset >= 0 || unsigned_offsets(file)) { |
1da177e4 LT |
208 | if (offset != file->f_pos) { |
209 | file->f_pos = offset; | |
210 | file->f_version = 0; | |
211 | } | |
212 | retval = offset; | |
213 | } | |
5b6f1eb9 | 214 | out: |
982d8165 | 215 | mutex_unlock(&inode->i_mutex); |
1da177e4 LT |
216 | return retval; |
217 | } | |
218 | EXPORT_SYMBOL(default_llseek); | |
219 | ||
965c8e59 | 220 | loff_t vfs_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
221 | { |
222 | loff_t (*fn)(struct file *, loff_t, int); | |
223 | ||
224 | fn = no_llseek; | |
225 | if (file->f_mode & FMODE_LSEEK) { | |
1da177e4 LT |
226 | if (file->f_op && file->f_op->llseek) |
227 | fn = file->f_op->llseek; | |
228 | } | |
965c8e59 | 229 | return fn(file, offset, whence); |
1da177e4 LT |
230 | } |
231 | EXPORT_SYMBOL(vfs_llseek); | |
232 | ||
965c8e59 | 233 | SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence) |
1da177e4 LT |
234 | { |
235 | off_t retval; | |
2903ff01 AV |
236 | struct fd f = fdget(fd); |
237 | if (!f.file) | |
238 | return -EBADF; | |
1da177e4 LT |
239 | |
240 | retval = -EINVAL; | |
965c8e59 AM |
241 | if (whence <= SEEK_MAX) { |
242 | loff_t res = vfs_llseek(f.file, offset, whence); | |
1da177e4 LT |
243 | retval = res; |
244 | if (res != (loff_t)retval) | |
245 | retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */ | |
246 | } | |
2903ff01 | 247 | fdput(f); |
1da177e4 LT |
248 | return retval; |
249 | } | |
250 | ||
561c6731 AV |
251 | #ifdef CONFIG_COMPAT |
252 | COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence) | |
253 | { | |
254 | return sys_lseek(fd, offset, whence); | |
255 | } | |
256 | #endif | |
257 | ||
1da177e4 | 258 | #ifdef __ARCH_WANT_SYS_LLSEEK |
003d7ab4 HC |
259 | SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high, |
260 | unsigned long, offset_low, loff_t __user *, result, | |
965c8e59 | 261 | unsigned int, whence) |
1da177e4 LT |
262 | { |
263 | int retval; | |
2903ff01 | 264 | struct fd f = fdget(fd); |
1da177e4 | 265 | loff_t offset; |
1da177e4 | 266 | |
2903ff01 AV |
267 | if (!f.file) |
268 | return -EBADF; | |
1da177e4 LT |
269 | |
270 | retval = -EINVAL; | |
965c8e59 | 271 | if (whence > SEEK_MAX) |
1da177e4 LT |
272 | goto out_putf; |
273 | ||
2903ff01 | 274 | offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low, |
965c8e59 | 275 | whence); |
1da177e4 LT |
276 | |
277 | retval = (int)offset; | |
278 | if (offset >= 0) { | |
279 | retval = -EFAULT; | |
280 | if (!copy_to_user(result, &offset, sizeof(offset))) | |
281 | retval = 0; | |
282 | } | |
283 | out_putf: | |
2903ff01 | 284 | fdput(f); |
1da177e4 LT |
285 | return retval; |
286 | } | |
287 | #endif | |
288 | ||
e28cc715 LT |
289 | /* |
290 | * rw_verify_area doesn't like huge counts. We limit | |
291 | * them to something that fits in "int" so that others | |
292 | * won't have to do range checks all the time. | |
293 | */ | |
1da177e4 LT |
294 | int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count) |
295 | { | |
296 | struct inode *inode; | |
297 | loff_t pos; | |
c43e259c | 298 | int retval = -EINVAL; |
1da177e4 | 299 | |
496ad9aa | 300 | inode = file_inode(file); |
e28cc715 | 301 | if (unlikely((ssize_t) count < 0)) |
c43e259c | 302 | return retval; |
1da177e4 | 303 | pos = *ppos; |
cccb5a1e AV |
304 | if (unlikely(pos < 0)) { |
305 | if (!unsigned_offsets(file)) | |
306 | return retval; | |
307 | if (count >= -pos) /* both values are in 0..LLONG_MAX */ | |
308 | return -EOVERFLOW; | |
309 | } else if (unlikely((loff_t) (pos + count) < 0)) { | |
310 | if (!unsigned_offsets(file)) | |
4a3956c7 KH |
311 | return retval; |
312 | } | |
1da177e4 | 313 | |
a16877ca | 314 | if (unlikely(inode->i_flock && mandatory_lock(inode))) { |
c43e259c | 315 | retval = locks_mandatory_area( |
e28cc715 LT |
316 | read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE, |
317 | inode, file, pos, count); | |
318 | if (retval < 0) | |
319 | return retval; | |
320 | } | |
c43e259c JM |
321 | retval = security_file_permission(file, |
322 | read_write == READ ? MAY_READ : MAY_WRITE); | |
323 | if (retval) | |
324 | return retval; | |
e28cc715 | 325 | return count > MAX_RW_COUNT ? MAX_RW_COUNT : count; |
1da177e4 LT |
326 | } |
327 | ||
63e68809 BL |
328 | static void wait_on_retry_sync_kiocb(struct kiocb *iocb) |
329 | { | |
330 | set_current_state(TASK_UNINTERRUPTIBLE); | |
331 | if (!kiocbIsKicked(iocb)) | |
332 | schedule(); | |
333 | else | |
334 | kiocbClearKicked(iocb); | |
335 | __set_current_state(TASK_RUNNING); | |
336 | } | |
337 | ||
1da177e4 LT |
338 | ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) |
339 | { | |
027445c3 | 340 | struct iovec iov = { .iov_base = buf, .iov_len = len }; |
1da177e4 LT |
341 | struct kiocb kiocb; |
342 | ssize_t ret; | |
343 | ||
344 | init_sync_kiocb(&kiocb, filp); | |
345 | kiocb.ki_pos = *ppos; | |
027445c3 | 346 | kiocb.ki_left = len; |
61964eba | 347 | kiocb.ki_nbytes = len; |
027445c3 BP |
348 | |
349 | for (;;) { | |
350 | ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos); | |
351 | if (ret != -EIOCBRETRY) | |
352 | break; | |
63e68809 | 353 | wait_on_retry_sync_kiocb(&kiocb); |
027445c3 | 354 | } |
63e68809 | 355 | |
1da177e4 LT |
356 | if (-EIOCBQUEUED == ret) |
357 | ret = wait_on_sync_kiocb(&kiocb); | |
358 | *ppos = kiocb.ki_pos; | |
359 | return ret; | |
360 | } | |
361 | ||
362 | EXPORT_SYMBOL(do_sync_read); | |
363 | ||
364 | ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos) | |
365 | { | |
366 | ssize_t ret; | |
367 | ||
368 | if (!(file->f_mode & FMODE_READ)) | |
369 | return -EBADF; | |
370 | if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read)) | |
371 | return -EINVAL; | |
372 | if (unlikely(!access_ok(VERIFY_WRITE, buf, count))) | |
373 | return -EFAULT; | |
374 | ||
375 | ret = rw_verify_area(READ, file, pos, count); | |
e28cc715 LT |
376 | if (ret >= 0) { |
377 | count = ret; | |
c43e259c JM |
378 | if (file->f_op->read) |
379 | ret = file->f_op->read(file, buf, count, pos); | |
380 | else | |
381 | ret = do_sync_read(file, buf, count, pos); | |
382 | if (ret > 0) { | |
2a12a9d7 | 383 | fsnotify_access(file); |
c43e259c | 384 | add_rchar(current, ret); |
1da177e4 | 385 | } |
c43e259c | 386 | inc_syscr(current); |
1da177e4 LT |
387 | } |
388 | ||
389 | return ret; | |
390 | } | |
391 | ||
392 | EXPORT_SYMBOL(vfs_read); | |
393 | ||
394 | ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) | |
395 | { | |
027445c3 | 396 | struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len }; |
1da177e4 LT |
397 | struct kiocb kiocb; |
398 | ssize_t ret; | |
399 | ||
400 | init_sync_kiocb(&kiocb, filp); | |
401 | kiocb.ki_pos = *ppos; | |
027445c3 | 402 | kiocb.ki_left = len; |
61964eba | 403 | kiocb.ki_nbytes = len; |
027445c3 BP |
404 | |
405 | for (;;) { | |
406 | ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos); | |
407 | if (ret != -EIOCBRETRY) | |
408 | break; | |
63e68809 | 409 | wait_on_retry_sync_kiocb(&kiocb); |
027445c3 | 410 | } |
63e68809 | 411 | |
1da177e4 LT |
412 | if (-EIOCBQUEUED == ret) |
413 | ret = wait_on_sync_kiocb(&kiocb); | |
414 | *ppos = kiocb.ki_pos; | |
415 | return ret; | |
416 | } | |
417 | ||
418 | EXPORT_SYMBOL(do_sync_write); | |
419 | ||
420 | ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos) | |
421 | { | |
422 | ssize_t ret; | |
423 | ||
424 | if (!(file->f_mode & FMODE_WRITE)) | |
425 | return -EBADF; | |
426 | if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write)) | |
427 | return -EINVAL; | |
428 | if (unlikely(!access_ok(VERIFY_READ, buf, count))) | |
429 | return -EFAULT; | |
430 | ||
431 | ret = rw_verify_area(WRITE, file, pos, count); | |
e28cc715 LT |
432 | if (ret >= 0) { |
433 | count = ret; | |
c43e259c JM |
434 | if (file->f_op->write) |
435 | ret = file->f_op->write(file, buf, count, pos); | |
436 | else | |
437 | ret = do_sync_write(file, buf, count, pos); | |
438 | if (ret > 0) { | |
2a12a9d7 | 439 | fsnotify_modify(file); |
c43e259c | 440 | add_wchar(current, ret); |
1da177e4 | 441 | } |
c43e259c | 442 | inc_syscw(current); |
1da177e4 LT |
443 | } |
444 | ||
445 | return ret; | |
446 | } | |
447 | ||
448 | EXPORT_SYMBOL(vfs_write); | |
449 | ||
450 | static inline loff_t file_pos_read(struct file *file) | |
451 | { | |
452 | return file->f_pos; | |
453 | } | |
454 | ||
455 | static inline void file_pos_write(struct file *file, loff_t pos) | |
456 | { | |
457 | file->f_pos = pos; | |
458 | } | |
459 | ||
3cdad428 | 460 | SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count) |
1da177e4 | 461 | { |
2903ff01 | 462 | struct fd f = fdget(fd); |
1da177e4 | 463 | ssize_t ret = -EBADF; |
1da177e4 | 464 | |
2903ff01 AV |
465 | if (f.file) { |
466 | loff_t pos = file_pos_read(f.file); | |
467 | ret = vfs_read(f.file, buf, count, &pos); | |
468 | file_pos_write(f.file, pos); | |
469 | fdput(f); | |
1da177e4 | 470 | } |
1da177e4 LT |
471 | return ret; |
472 | } | |
1da177e4 | 473 | |
3cdad428 HC |
474 | SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf, |
475 | size_t, count) | |
1da177e4 | 476 | { |
2903ff01 | 477 | struct fd f = fdget(fd); |
1da177e4 | 478 | ssize_t ret = -EBADF; |
1da177e4 | 479 | |
2903ff01 AV |
480 | if (f.file) { |
481 | loff_t pos = file_pos_read(f.file); | |
482 | ret = vfs_write(f.file, buf, count, &pos); | |
483 | file_pos_write(f.file, pos); | |
484 | fdput(f); | |
1da177e4 LT |
485 | } |
486 | ||
487 | return ret; | |
488 | } | |
489 | ||
6673e0c3 HC |
490 | SYSCALL_DEFINE(pread64)(unsigned int fd, char __user *buf, |
491 | size_t count, loff_t pos) | |
1da177e4 | 492 | { |
2903ff01 | 493 | struct fd f; |
1da177e4 | 494 | ssize_t ret = -EBADF; |
1da177e4 LT |
495 | |
496 | if (pos < 0) | |
497 | return -EINVAL; | |
498 | ||
2903ff01 AV |
499 | f = fdget(fd); |
500 | if (f.file) { | |
1da177e4 | 501 | ret = -ESPIPE; |
2903ff01 AV |
502 | if (f.file->f_mode & FMODE_PREAD) |
503 | ret = vfs_read(f.file, buf, count, &pos); | |
504 | fdput(f); | |
1da177e4 LT |
505 | } |
506 | ||
507 | return ret; | |
508 | } | |
6673e0c3 HC |
509 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
510 | asmlinkage long SyS_pread64(long fd, long buf, long count, loff_t pos) | |
511 | { | |
512 | return SYSC_pread64((unsigned int) fd, (char __user *) buf, | |
513 | (size_t) count, pos); | |
514 | } | |
515 | SYSCALL_ALIAS(sys_pread64, SyS_pread64); | |
516 | #endif | |
1da177e4 | 517 | |
6673e0c3 HC |
518 | SYSCALL_DEFINE(pwrite64)(unsigned int fd, const char __user *buf, |
519 | size_t count, loff_t pos) | |
1da177e4 | 520 | { |
2903ff01 | 521 | struct fd f; |
1da177e4 | 522 | ssize_t ret = -EBADF; |
1da177e4 LT |
523 | |
524 | if (pos < 0) | |
525 | return -EINVAL; | |
526 | ||
2903ff01 AV |
527 | f = fdget(fd); |
528 | if (f.file) { | |
1da177e4 | 529 | ret = -ESPIPE; |
2903ff01 AV |
530 | if (f.file->f_mode & FMODE_PWRITE) |
531 | ret = vfs_write(f.file, buf, count, &pos); | |
532 | fdput(f); | |
1da177e4 LT |
533 | } |
534 | ||
535 | return ret; | |
536 | } | |
6673e0c3 HC |
537 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
538 | asmlinkage long SyS_pwrite64(long fd, long buf, long count, loff_t pos) | |
539 | { | |
540 | return SYSC_pwrite64((unsigned int) fd, (const char __user *) buf, | |
541 | (size_t) count, pos); | |
542 | } | |
543 | SYSCALL_ALIAS(sys_pwrite64, SyS_pwrite64); | |
544 | #endif | |
1da177e4 LT |
545 | |
546 | /* | |
547 | * Reduce an iovec's length in-place. Return the resulting number of segments | |
548 | */ | |
549 | unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to) | |
550 | { | |
551 | unsigned long seg = 0; | |
552 | size_t len = 0; | |
553 | ||
554 | while (seg < nr_segs) { | |
555 | seg++; | |
556 | if (len + iov->iov_len >= to) { | |
557 | iov->iov_len = to - len; | |
558 | break; | |
559 | } | |
560 | len += iov->iov_len; | |
561 | iov++; | |
562 | } | |
563 | return seg; | |
564 | } | |
19295529 | 565 | EXPORT_SYMBOL(iov_shorten); |
1da177e4 | 566 | |
ee0b3e67 BP |
567 | ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov, |
568 | unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn) | |
569 | { | |
570 | struct kiocb kiocb; | |
571 | ssize_t ret; | |
572 | ||
573 | init_sync_kiocb(&kiocb, filp); | |
574 | kiocb.ki_pos = *ppos; | |
575 | kiocb.ki_left = len; | |
576 | kiocb.ki_nbytes = len; | |
577 | ||
578 | for (;;) { | |
579 | ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos); | |
580 | if (ret != -EIOCBRETRY) | |
581 | break; | |
582 | wait_on_retry_sync_kiocb(&kiocb); | |
583 | } | |
584 | ||
585 | if (ret == -EIOCBQUEUED) | |
586 | ret = wait_on_sync_kiocb(&kiocb); | |
587 | *ppos = kiocb.ki_pos; | |
588 | return ret; | |
589 | } | |
590 | ||
591 | /* Do it by hand, with file-ops */ | |
592 | ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov, | |
593 | unsigned long nr_segs, loff_t *ppos, io_fn_t fn) | |
594 | { | |
595 | struct iovec *vector = iov; | |
596 | ssize_t ret = 0; | |
597 | ||
598 | while (nr_segs > 0) { | |
599 | void __user *base; | |
600 | size_t len; | |
601 | ssize_t nr; | |
602 | ||
603 | base = vector->iov_base; | |
604 | len = vector->iov_len; | |
605 | vector++; | |
606 | nr_segs--; | |
607 | ||
608 | nr = fn(filp, base, len, ppos); | |
609 | ||
610 | if (nr < 0) { | |
611 | if (!ret) | |
612 | ret = nr; | |
613 | break; | |
614 | } | |
615 | ret += nr; | |
616 | if (nr != len) | |
617 | break; | |
618 | } | |
619 | ||
620 | return ret; | |
621 | } | |
622 | ||
1da177e4 LT |
623 | /* A write operation does a read from user space and vice versa */ |
624 | #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ) | |
625 | ||
eed4e51f BP |
626 | ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector, |
627 | unsigned long nr_segs, unsigned long fast_segs, | |
628 | struct iovec *fast_pointer, | |
ac34ebb3 | 629 | struct iovec **ret_pointer) |
435f49a5 | 630 | { |
eed4e51f | 631 | unsigned long seg; |
435f49a5 | 632 | ssize_t ret; |
eed4e51f BP |
633 | struct iovec *iov = fast_pointer; |
634 | ||
435f49a5 LT |
635 | /* |
636 | * SuS says "The readv() function *may* fail if the iovcnt argument | |
637 | * was less than or equal to 0, or greater than {IOV_MAX}. Linux has | |
638 | * traditionally returned zero for zero segments, so... | |
639 | */ | |
eed4e51f BP |
640 | if (nr_segs == 0) { |
641 | ret = 0; | |
435f49a5 | 642 | goto out; |
eed4e51f BP |
643 | } |
644 | ||
435f49a5 LT |
645 | /* |
646 | * First get the "struct iovec" from user memory and | |
647 | * verify all the pointers | |
648 | */ | |
eed4e51f BP |
649 | if (nr_segs > UIO_MAXIOV) { |
650 | ret = -EINVAL; | |
435f49a5 | 651 | goto out; |
eed4e51f BP |
652 | } |
653 | if (nr_segs > fast_segs) { | |
435f49a5 | 654 | iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL); |
eed4e51f BP |
655 | if (iov == NULL) { |
656 | ret = -ENOMEM; | |
435f49a5 | 657 | goto out; |
eed4e51f | 658 | } |
435f49a5 | 659 | } |
eed4e51f BP |
660 | if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) { |
661 | ret = -EFAULT; | |
435f49a5 | 662 | goto out; |
eed4e51f BP |
663 | } |
664 | ||
435f49a5 | 665 | /* |
eed4e51f BP |
666 | * According to the Single Unix Specification we should return EINVAL |
667 | * if an element length is < 0 when cast to ssize_t or if the | |
668 | * total length would overflow the ssize_t return value of the | |
669 | * system call. | |
435f49a5 LT |
670 | * |
671 | * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the | |
672 | * overflow case. | |
673 | */ | |
eed4e51f | 674 | ret = 0; |
435f49a5 LT |
675 | for (seg = 0; seg < nr_segs; seg++) { |
676 | void __user *buf = iov[seg].iov_base; | |
677 | ssize_t len = (ssize_t)iov[seg].iov_len; | |
eed4e51f BP |
678 | |
679 | /* see if we we're about to use an invalid len or if | |
680 | * it's about to overflow ssize_t */ | |
435f49a5 | 681 | if (len < 0) { |
eed4e51f | 682 | ret = -EINVAL; |
435f49a5 | 683 | goto out; |
eed4e51f | 684 | } |
ac34ebb3 | 685 | if (type >= 0 |
fcf63409 | 686 | && unlikely(!access_ok(vrfy_dir(type), buf, len))) { |
eed4e51f | 687 | ret = -EFAULT; |
435f49a5 LT |
688 | goto out; |
689 | } | |
690 | if (len > MAX_RW_COUNT - ret) { | |
691 | len = MAX_RW_COUNT - ret; | |
692 | iov[seg].iov_len = len; | |
eed4e51f | 693 | } |
eed4e51f | 694 | ret += len; |
435f49a5 | 695 | } |
eed4e51f BP |
696 | out: |
697 | *ret_pointer = iov; | |
698 | return ret; | |
699 | } | |
700 | ||
1da177e4 LT |
701 | static ssize_t do_readv_writev(int type, struct file *file, |
702 | const struct iovec __user * uvector, | |
703 | unsigned long nr_segs, loff_t *pos) | |
704 | { | |
1da177e4 LT |
705 | size_t tot_len; |
706 | struct iovec iovstack[UIO_FASTIOV]; | |
ee0b3e67 | 707 | struct iovec *iov = iovstack; |
1da177e4 | 708 | ssize_t ret; |
1da177e4 LT |
709 | io_fn_t fn; |
710 | iov_fn_t fnv; | |
711 | ||
eed4e51f BP |
712 | if (!file->f_op) { |
713 | ret = -EINVAL; | |
1da177e4 | 714 | goto out; |
1da177e4 | 715 | } |
1da177e4 | 716 | |
eed4e51f | 717 | ret = rw_copy_check_uvector(type, uvector, nr_segs, |
ac34ebb3 | 718 | ARRAY_SIZE(iovstack), iovstack, &iov); |
eed4e51f | 719 | if (ret <= 0) |
1da177e4 | 720 | goto out; |
1da177e4 | 721 | |
eed4e51f | 722 | tot_len = ret; |
1da177e4 | 723 | ret = rw_verify_area(type, file, pos, tot_len); |
e28cc715 | 724 | if (ret < 0) |
411b67b4 | 725 | goto out; |
1da177e4 LT |
726 | |
727 | fnv = NULL; | |
728 | if (type == READ) { | |
729 | fn = file->f_op->read; | |
ee0b3e67 | 730 | fnv = file->f_op->aio_read; |
1da177e4 LT |
731 | } else { |
732 | fn = (io_fn_t)file->f_op->write; | |
ee0b3e67 | 733 | fnv = file->f_op->aio_write; |
1da177e4 LT |
734 | } |
735 | ||
ee0b3e67 BP |
736 | if (fnv) |
737 | ret = do_sync_readv_writev(file, iov, nr_segs, tot_len, | |
738 | pos, fnv); | |
739 | else | |
740 | ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn); | |
1da177e4 | 741 | |
1da177e4 LT |
742 | out: |
743 | if (iov != iovstack) | |
744 | kfree(iov); | |
0eeca283 RL |
745 | if ((ret + (type == READ)) > 0) { |
746 | if (type == READ) | |
2a12a9d7 | 747 | fsnotify_access(file); |
0eeca283 | 748 | else |
2a12a9d7 | 749 | fsnotify_modify(file); |
0eeca283 | 750 | } |
1da177e4 | 751 | return ret; |
1da177e4 LT |
752 | } |
753 | ||
754 | ssize_t vfs_readv(struct file *file, const struct iovec __user *vec, | |
755 | unsigned long vlen, loff_t *pos) | |
756 | { | |
757 | if (!(file->f_mode & FMODE_READ)) | |
758 | return -EBADF; | |
ee0b3e67 | 759 | if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read)) |
1da177e4 LT |
760 | return -EINVAL; |
761 | ||
762 | return do_readv_writev(READ, file, vec, vlen, pos); | |
763 | } | |
764 | ||
765 | EXPORT_SYMBOL(vfs_readv); | |
766 | ||
767 | ssize_t vfs_writev(struct file *file, const struct iovec __user *vec, | |
768 | unsigned long vlen, loff_t *pos) | |
769 | { | |
770 | if (!(file->f_mode & FMODE_WRITE)) | |
771 | return -EBADF; | |
ee0b3e67 | 772 | if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write)) |
1da177e4 LT |
773 | return -EINVAL; |
774 | ||
775 | return do_readv_writev(WRITE, file, vec, vlen, pos); | |
776 | } | |
777 | ||
778 | EXPORT_SYMBOL(vfs_writev); | |
779 | ||
3cdad428 HC |
780 | SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec, |
781 | unsigned long, vlen) | |
1da177e4 | 782 | { |
2903ff01 | 783 | struct fd f = fdget(fd); |
1da177e4 | 784 | ssize_t ret = -EBADF; |
1da177e4 | 785 | |
2903ff01 AV |
786 | if (f.file) { |
787 | loff_t pos = file_pos_read(f.file); | |
788 | ret = vfs_readv(f.file, vec, vlen, &pos); | |
789 | file_pos_write(f.file, pos); | |
790 | fdput(f); | |
1da177e4 LT |
791 | } |
792 | ||
793 | if (ret > 0) | |
4b98d11b AD |
794 | add_rchar(current, ret); |
795 | inc_syscr(current); | |
1da177e4 LT |
796 | return ret; |
797 | } | |
798 | ||
3cdad428 HC |
799 | SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec, |
800 | unsigned long, vlen) | |
1da177e4 | 801 | { |
2903ff01 | 802 | struct fd f = fdget(fd); |
1da177e4 | 803 | ssize_t ret = -EBADF; |
1da177e4 | 804 | |
2903ff01 AV |
805 | if (f.file) { |
806 | loff_t pos = file_pos_read(f.file); | |
807 | ret = vfs_writev(f.file, vec, vlen, &pos); | |
808 | file_pos_write(f.file, pos); | |
809 | fdput(f); | |
1da177e4 LT |
810 | } |
811 | ||
812 | if (ret > 0) | |
4b98d11b AD |
813 | add_wchar(current, ret); |
814 | inc_syscw(current); | |
1da177e4 LT |
815 | return ret; |
816 | } | |
817 | ||
601cc11d LT |
818 | static inline loff_t pos_from_hilo(unsigned long high, unsigned long low) |
819 | { | |
820 | #define HALF_LONG_BITS (BITS_PER_LONG / 2) | |
821 | return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low; | |
822 | } | |
823 | ||
f3554f4b | 824 | SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec, |
601cc11d | 825 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) |
f3554f4b | 826 | { |
601cc11d | 827 | loff_t pos = pos_from_hilo(pos_h, pos_l); |
2903ff01 | 828 | struct fd f; |
f3554f4b | 829 | ssize_t ret = -EBADF; |
f3554f4b GH |
830 | |
831 | if (pos < 0) | |
832 | return -EINVAL; | |
833 | ||
2903ff01 AV |
834 | f = fdget(fd); |
835 | if (f.file) { | |
f3554f4b | 836 | ret = -ESPIPE; |
2903ff01 AV |
837 | if (f.file->f_mode & FMODE_PREAD) |
838 | ret = vfs_readv(f.file, vec, vlen, &pos); | |
839 | fdput(f); | |
f3554f4b GH |
840 | } |
841 | ||
842 | if (ret > 0) | |
843 | add_rchar(current, ret); | |
844 | inc_syscr(current); | |
845 | return ret; | |
846 | } | |
847 | ||
848 | SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec, | |
601cc11d | 849 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) |
f3554f4b | 850 | { |
601cc11d | 851 | loff_t pos = pos_from_hilo(pos_h, pos_l); |
2903ff01 | 852 | struct fd f; |
f3554f4b | 853 | ssize_t ret = -EBADF; |
f3554f4b GH |
854 | |
855 | if (pos < 0) | |
856 | return -EINVAL; | |
857 | ||
2903ff01 AV |
858 | f = fdget(fd); |
859 | if (f.file) { | |
f3554f4b | 860 | ret = -ESPIPE; |
2903ff01 AV |
861 | if (f.file->f_mode & FMODE_PWRITE) |
862 | ret = vfs_writev(f.file, vec, vlen, &pos); | |
863 | fdput(f); | |
f3554f4b GH |
864 | } |
865 | ||
866 | if (ret > 0) | |
867 | add_wchar(current, ret); | |
868 | inc_syscw(current); | |
869 | return ret; | |
870 | } | |
871 | ||
8f9c0119 CM |
872 | ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos, size_t count, |
873 | loff_t max) | |
1da177e4 | 874 | { |
2903ff01 AV |
875 | struct fd in, out; |
876 | struct inode *in_inode, *out_inode; | |
1da177e4 LT |
877 | loff_t pos; |
878 | ssize_t retval; | |
2903ff01 | 879 | int fl; |
1da177e4 LT |
880 | |
881 | /* | |
882 | * Get input file, and verify that it is ok.. | |
883 | */ | |
884 | retval = -EBADF; | |
2903ff01 AV |
885 | in = fdget(in_fd); |
886 | if (!in.file) | |
1da177e4 | 887 | goto out; |
2903ff01 | 888 | if (!(in.file->f_mode & FMODE_READ)) |
1da177e4 | 889 | goto fput_in; |
1da177e4 LT |
890 | retval = -ESPIPE; |
891 | if (!ppos) | |
2903ff01 | 892 | ppos = &in.file->f_pos; |
1da177e4 | 893 | else |
2903ff01 | 894 | if (!(in.file->f_mode & FMODE_PREAD)) |
1da177e4 | 895 | goto fput_in; |
2903ff01 | 896 | retval = rw_verify_area(READ, in.file, ppos, count); |
e28cc715 | 897 | if (retval < 0) |
1da177e4 | 898 | goto fput_in; |
e28cc715 | 899 | count = retval; |
1da177e4 | 900 | |
1da177e4 LT |
901 | /* |
902 | * Get output file, and verify that it is ok.. | |
903 | */ | |
904 | retval = -EBADF; | |
2903ff01 AV |
905 | out = fdget(out_fd); |
906 | if (!out.file) | |
1da177e4 | 907 | goto fput_in; |
2903ff01 | 908 | if (!(out.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
909 | goto fput_out; |
910 | retval = -EINVAL; | |
496ad9aa AV |
911 | in_inode = file_inode(in.file); |
912 | out_inode = file_inode(out.file); | |
2903ff01 | 913 | retval = rw_verify_area(WRITE, out.file, &out.file->f_pos, count); |
e28cc715 | 914 | if (retval < 0) |
1da177e4 | 915 | goto fput_out; |
e28cc715 | 916 | count = retval; |
1da177e4 | 917 | |
1da177e4 LT |
918 | if (!max) |
919 | max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes); | |
920 | ||
921 | pos = *ppos; | |
1da177e4 LT |
922 | if (unlikely(pos + count > max)) { |
923 | retval = -EOVERFLOW; | |
924 | if (pos >= max) | |
925 | goto fput_out; | |
926 | count = max - pos; | |
927 | } | |
928 | ||
d96e6e71 | 929 | fl = 0; |
534f2aaa | 930 | #if 0 |
d96e6e71 JA |
931 | /* |
932 | * We need to debate whether we can enable this or not. The | |
933 | * man page documents EAGAIN return for the output at least, | |
934 | * and the application is arguably buggy if it doesn't expect | |
935 | * EAGAIN on a non-blocking file descriptor. | |
936 | */ | |
2903ff01 | 937 | if (in.file->f_flags & O_NONBLOCK) |
d96e6e71 | 938 | fl = SPLICE_F_NONBLOCK; |
534f2aaa | 939 | #endif |
2903ff01 | 940 | retval = do_splice_direct(in.file, ppos, out.file, count, fl); |
1da177e4 LT |
941 | |
942 | if (retval > 0) { | |
4b98d11b AD |
943 | add_rchar(current, retval); |
944 | add_wchar(current, retval); | |
a68c2f12 SW |
945 | fsnotify_access(in.file); |
946 | fsnotify_modify(out.file); | |
1da177e4 | 947 | } |
1da177e4 | 948 | |
4b98d11b AD |
949 | inc_syscr(current); |
950 | inc_syscw(current); | |
1da177e4 LT |
951 | if (*ppos > max) |
952 | retval = -EOVERFLOW; | |
953 | ||
954 | fput_out: | |
2903ff01 | 955 | fdput(out); |
1da177e4 | 956 | fput_in: |
2903ff01 | 957 | fdput(in); |
1da177e4 LT |
958 | out: |
959 | return retval; | |
960 | } | |
961 | ||
002c8976 | 962 | SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count) |
1da177e4 LT |
963 | { |
964 | loff_t pos; | |
965 | off_t off; | |
966 | ssize_t ret; | |
967 | ||
968 | if (offset) { | |
969 | if (unlikely(get_user(off, offset))) | |
970 | return -EFAULT; | |
971 | pos = off; | |
972 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS); | |
973 | if (unlikely(put_user(pos, offset))) | |
974 | return -EFAULT; | |
975 | return ret; | |
976 | } | |
977 | ||
978 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
979 | } | |
980 | ||
002c8976 | 981 | SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count) |
1da177e4 LT |
982 | { |
983 | loff_t pos; | |
984 | ssize_t ret; | |
985 | ||
986 | if (offset) { | |
987 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t)))) | |
988 | return -EFAULT; | |
989 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0); | |
990 | if (unlikely(put_user(pos, offset))) | |
991 | return -EFAULT; | |
992 | return ret; | |
993 | } | |
994 | ||
995 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
996 | } |