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" |
06ae43f3 | 20 | #include "internal.h" |
1da177e4 LT |
21 | |
22 | #include <asm/uaccess.h> | |
23 | #include <asm/unistd.h> | |
24 | ||
4b6f5d20 | 25 | const struct file_operations generic_ro_fops = { |
1da177e4 | 26 | .llseek = generic_file_llseek, |
543ade1f BP |
27 | .read = do_sync_read, |
28 | .aio_read = generic_file_aio_read, | |
1da177e4 | 29 | .mmap = generic_file_readonly_mmap, |
534f2aaa | 30 | .splice_read = generic_file_splice_read, |
1da177e4 LT |
31 | }; |
32 | ||
33 | EXPORT_SYMBOL(generic_ro_fops); | |
34 | ||
cccb5a1e | 35 | static inline int unsigned_offsets(struct file *file) |
4a3956c7 | 36 | { |
cccb5a1e | 37 | return file->f_mode & FMODE_UNSIGNED_OFFSET; |
4a3956c7 KH |
38 | } |
39 | ||
ef3d0fd2 AK |
40 | static loff_t lseek_execute(struct file *file, struct inode *inode, |
41 | loff_t offset, loff_t maxsize) | |
42 | { | |
43 | if (offset < 0 && !unsigned_offsets(file)) | |
44 | return -EINVAL; | |
45 | if (offset > maxsize) | |
46 | return -EINVAL; | |
47 | ||
48 | if (offset != file->f_pos) { | |
49 | file->f_pos = offset; | |
50 | file->f_version = 0; | |
51 | } | |
52 | return offset; | |
53 | } | |
54 | ||
3a8cff4f | 55 | /** |
5760495a | 56 | * generic_file_llseek_size - generic llseek implementation for regular files |
3a8cff4f CH |
57 | * @file: file structure to seek on |
58 | * @offset: file offset to seek to | |
965c8e59 | 59 | * @whence: type of seek |
e8b96eb5 ES |
60 | * @size: max size of this file in file system |
61 | * @eof: offset used for SEEK_END position | |
3a8cff4f | 62 | * |
5760495a | 63 | * This is a variant of generic_file_llseek that allows passing in a custom |
e8b96eb5 | 64 | * maximum file size and a custom EOF position, for e.g. hashed directories |
ef3d0fd2 AK |
65 | * |
66 | * Synchronization: | |
5760495a | 67 | * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms) |
ef3d0fd2 AK |
68 | * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes. |
69 | * read/writes behave like SEEK_SET against seeks. | |
3a8cff4f | 70 | */ |
9465efc9 | 71 | loff_t |
965c8e59 | 72 | generic_file_llseek_size(struct file *file, loff_t offset, int whence, |
e8b96eb5 | 73 | loff_t maxsize, loff_t eof) |
1da177e4 | 74 | { |
1da177e4 LT |
75 | struct inode *inode = file->f_mapping->host; |
76 | ||
965c8e59 | 77 | switch (whence) { |
3a8cff4f | 78 | case SEEK_END: |
e8b96eb5 | 79 | offset += eof; |
3a8cff4f CH |
80 | break; |
81 | case SEEK_CUR: | |
5b6f1eb9 AK |
82 | /* |
83 | * Here we special-case the lseek(fd, 0, SEEK_CUR) | |
84 | * position-querying operation. Avoid rewriting the "same" | |
85 | * f_pos value back to the file because a concurrent read(), | |
86 | * write() or lseek() might have altered it | |
87 | */ | |
88 | if (offset == 0) | |
89 | return file->f_pos; | |
ef3d0fd2 AK |
90 | /* |
91 | * f_lock protects against read/modify/write race with other | |
92 | * SEEK_CURs. Note that parallel writes and reads behave | |
93 | * like SEEK_SET. | |
94 | */ | |
95 | spin_lock(&file->f_lock); | |
96 | offset = lseek_execute(file, inode, file->f_pos + offset, | |
5760495a | 97 | maxsize); |
ef3d0fd2 AK |
98 | spin_unlock(&file->f_lock); |
99 | return offset; | |
982d8165 JB |
100 | case SEEK_DATA: |
101 | /* | |
102 | * In the generic case the entire file is data, so as long as | |
103 | * offset isn't at the end of the file then the offset is data. | |
104 | */ | |
e8b96eb5 | 105 | if (offset >= eof) |
982d8165 JB |
106 | return -ENXIO; |
107 | break; | |
108 | case SEEK_HOLE: | |
109 | /* | |
110 | * There is a virtual hole at the end of the file, so as long as | |
111 | * offset isn't i_size or larger, return i_size. | |
112 | */ | |
e8b96eb5 | 113 | if (offset >= eof) |
982d8165 | 114 | return -ENXIO; |
e8b96eb5 | 115 | offset = eof; |
982d8165 | 116 | break; |
1da177e4 | 117 | } |
3a8cff4f | 118 | |
5760495a AK |
119 | return lseek_execute(file, inode, offset, maxsize); |
120 | } | |
121 | EXPORT_SYMBOL(generic_file_llseek_size); | |
122 | ||
123 | /** | |
124 | * generic_file_llseek - generic llseek implementation for regular files | |
125 | * @file: file structure to seek on | |
126 | * @offset: file offset to seek to | |
965c8e59 | 127 | * @whence: type of seek |
5760495a AK |
128 | * |
129 | * This is a generic implemenation of ->llseek useable for all normal local | |
130 | * filesystems. It just updates the file offset to the value specified by | |
965c8e59 | 131 | * @offset and @whence under i_mutex. |
5760495a | 132 | */ |
965c8e59 | 133 | loff_t generic_file_llseek(struct file *file, loff_t offset, int whence) |
5760495a AK |
134 | { |
135 | struct inode *inode = file->f_mapping->host; | |
136 | ||
965c8e59 | 137 | return generic_file_llseek_size(file, offset, whence, |
e8b96eb5 ES |
138 | inode->i_sb->s_maxbytes, |
139 | i_size_read(inode)); | |
1da177e4 | 140 | } |
9465efc9 | 141 | EXPORT_SYMBOL(generic_file_llseek); |
1da177e4 | 142 | |
ae6afc3f B |
143 | /** |
144 | * noop_llseek - No Operation Performed llseek implementation | |
145 | * @file: file structure to seek on | |
146 | * @offset: file offset to seek to | |
965c8e59 | 147 | * @whence: type of seek |
ae6afc3f B |
148 | * |
149 | * This is an implementation of ->llseek useable for the rare special case when | |
150 | * userspace expects the seek to succeed but the (device) file is actually not | |
151 | * able to perform the seek. In this case you use noop_llseek() instead of | |
152 | * falling back to the default implementation of ->llseek. | |
153 | */ | |
965c8e59 | 154 | loff_t noop_llseek(struct file *file, loff_t offset, int whence) |
ae6afc3f B |
155 | { |
156 | return file->f_pos; | |
157 | } | |
158 | EXPORT_SYMBOL(noop_llseek); | |
159 | ||
965c8e59 | 160 | loff_t no_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
161 | { |
162 | return -ESPIPE; | |
163 | } | |
164 | EXPORT_SYMBOL(no_llseek); | |
165 | ||
965c8e59 | 166 | loff_t default_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 167 | { |
496ad9aa | 168 | struct inode *inode = file_inode(file); |
16abef0e | 169 | loff_t retval; |
1da177e4 | 170 | |
982d8165 | 171 | mutex_lock(&inode->i_mutex); |
965c8e59 | 172 | switch (whence) { |
7b8e8924 | 173 | case SEEK_END: |
982d8165 | 174 | offset += i_size_read(inode); |
1da177e4 | 175 | break; |
7b8e8924 | 176 | case SEEK_CUR: |
5b6f1eb9 AK |
177 | if (offset == 0) { |
178 | retval = file->f_pos; | |
179 | goto out; | |
180 | } | |
1da177e4 | 181 | offset += file->f_pos; |
982d8165 JB |
182 | break; |
183 | case SEEK_DATA: | |
184 | /* | |
185 | * In the generic case the entire file is data, so as | |
186 | * long as offset isn't at the end of the file then the | |
187 | * offset is data. | |
188 | */ | |
bacb2d81 DC |
189 | if (offset >= inode->i_size) { |
190 | retval = -ENXIO; | |
191 | goto out; | |
192 | } | |
982d8165 JB |
193 | break; |
194 | case SEEK_HOLE: | |
195 | /* | |
196 | * There is a virtual hole at the end of the file, so | |
197 | * as long as offset isn't i_size or larger, return | |
198 | * i_size. | |
199 | */ | |
bacb2d81 DC |
200 | if (offset >= inode->i_size) { |
201 | retval = -ENXIO; | |
202 | goto out; | |
203 | } | |
982d8165 JB |
204 | offset = inode->i_size; |
205 | break; | |
1da177e4 LT |
206 | } |
207 | retval = -EINVAL; | |
cccb5a1e | 208 | if (offset >= 0 || unsigned_offsets(file)) { |
1da177e4 LT |
209 | if (offset != file->f_pos) { |
210 | file->f_pos = offset; | |
211 | file->f_version = 0; | |
212 | } | |
213 | retval = offset; | |
214 | } | |
5b6f1eb9 | 215 | out: |
982d8165 | 216 | mutex_unlock(&inode->i_mutex); |
1da177e4 LT |
217 | return retval; |
218 | } | |
219 | EXPORT_SYMBOL(default_llseek); | |
220 | ||
965c8e59 | 221 | loff_t vfs_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 LT |
222 | { |
223 | loff_t (*fn)(struct file *, loff_t, int); | |
224 | ||
225 | fn = no_llseek; | |
226 | if (file->f_mode & FMODE_LSEEK) { | |
1da177e4 LT |
227 | if (file->f_op && file->f_op->llseek) |
228 | fn = file->f_op->llseek; | |
229 | } | |
965c8e59 | 230 | return fn(file, offset, whence); |
1da177e4 LT |
231 | } |
232 | EXPORT_SYMBOL(vfs_llseek); | |
233 | ||
965c8e59 | 234 | SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence) |
1da177e4 LT |
235 | { |
236 | off_t retval; | |
2903ff01 AV |
237 | struct fd f = fdget(fd); |
238 | if (!f.file) | |
239 | return -EBADF; | |
1da177e4 LT |
240 | |
241 | retval = -EINVAL; | |
965c8e59 AM |
242 | if (whence <= SEEK_MAX) { |
243 | loff_t res = vfs_llseek(f.file, offset, whence); | |
1da177e4 LT |
244 | retval = res; |
245 | if (res != (loff_t)retval) | |
246 | retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */ | |
247 | } | |
2903ff01 | 248 | fdput(f); |
1da177e4 LT |
249 | return retval; |
250 | } | |
251 | ||
561c6731 AV |
252 | #ifdef CONFIG_COMPAT |
253 | COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence) | |
254 | { | |
255 | return sys_lseek(fd, offset, whence); | |
256 | } | |
257 | #endif | |
258 | ||
1da177e4 | 259 | #ifdef __ARCH_WANT_SYS_LLSEEK |
003d7ab4 HC |
260 | SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high, |
261 | unsigned long, offset_low, loff_t __user *, result, | |
965c8e59 | 262 | unsigned int, whence) |
1da177e4 LT |
263 | { |
264 | int retval; | |
2903ff01 | 265 | struct fd f = fdget(fd); |
1da177e4 | 266 | loff_t offset; |
1da177e4 | 267 | |
2903ff01 AV |
268 | if (!f.file) |
269 | return -EBADF; | |
1da177e4 LT |
270 | |
271 | retval = -EINVAL; | |
965c8e59 | 272 | if (whence > SEEK_MAX) |
1da177e4 LT |
273 | goto out_putf; |
274 | ||
2903ff01 | 275 | offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low, |
965c8e59 | 276 | whence); |
1da177e4 LT |
277 | |
278 | retval = (int)offset; | |
279 | if (offset >= 0) { | |
280 | retval = -EFAULT; | |
281 | if (!copy_to_user(result, &offset, sizeof(offset))) | |
282 | retval = 0; | |
283 | } | |
284 | out_putf: | |
2903ff01 | 285 | fdput(f); |
1da177e4 LT |
286 | return retval; |
287 | } | |
288 | #endif | |
289 | ||
e28cc715 LT |
290 | /* |
291 | * rw_verify_area doesn't like huge counts. We limit | |
292 | * them to something that fits in "int" so that others | |
293 | * won't have to do range checks all the time. | |
294 | */ | |
1da177e4 LT |
295 | int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count) |
296 | { | |
297 | struct inode *inode; | |
298 | loff_t pos; | |
c43e259c | 299 | int retval = -EINVAL; |
1da177e4 | 300 | |
496ad9aa | 301 | inode = file_inode(file); |
e28cc715 | 302 | if (unlikely((ssize_t) count < 0)) |
c43e259c | 303 | return retval; |
1da177e4 | 304 | pos = *ppos; |
cccb5a1e AV |
305 | if (unlikely(pos < 0)) { |
306 | if (!unsigned_offsets(file)) | |
307 | return retval; | |
308 | if (count >= -pos) /* both values are in 0..LLONG_MAX */ | |
309 | return -EOVERFLOW; | |
310 | } else if (unlikely((loff_t) (pos + count) < 0)) { | |
311 | if (!unsigned_offsets(file)) | |
4a3956c7 KH |
312 | return retval; |
313 | } | |
1da177e4 | 314 | |
a16877ca | 315 | if (unlikely(inode->i_flock && mandatory_lock(inode))) { |
c43e259c | 316 | retval = locks_mandatory_area( |
e28cc715 LT |
317 | read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE, |
318 | inode, file, pos, count); | |
319 | if (retval < 0) | |
320 | return retval; | |
321 | } | |
c43e259c JM |
322 | retval = security_file_permission(file, |
323 | read_write == READ ? MAY_READ : MAY_WRITE); | |
324 | if (retval) | |
325 | return retval; | |
e28cc715 | 326 | return count > MAX_RW_COUNT ? MAX_RW_COUNT : count; |
1da177e4 LT |
327 | } |
328 | ||
63e68809 BL |
329 | static void wait_on_retry_sync_kiocb(struct kiocb *iocb) |
330 | { | |
331 | set_current_state(TASK_UNINTERRUPTIBLE); | |
332 | if (!kiocbIsKicked(iocb)) | |
333 | schedule(); | |
334 | else | |
335 | kiocbClearKicked(iocb); | |
336 | __set_current_state(TASK_RUNNING); | |
337 | } | |
338 | ||
1da177e4 LT |
339 | ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) |
340 | { | |
027445c3 | 341 | struct iovec iov = { .iov_base = buf, .iov_len = len }; |
1da177e4 LT |
342 | struct kiocb kiocb; |
343 | ssize_t ret; | |
344 | ||
345 | init_sync_kiocb(&kiocb, filp); | |
346 | kiocb.ki_pos = *ppos; | |
027445c3 | 347 | kiocb.ki_left = len; |
61964eba | 348 | kiocb.ki_nbytes = len; |
027445c3 BP |
349 | |
350 | for (;;) { | |
351 | ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos); | |
352 | if (ret != -EIOCBRETRY) | |
353 | break; | |
63e68809 | 354 | wait_on_retry_sync_kiocb(&kiocb); |
027445c3 | 355 | } |
63e68809 | 356 | |
1da177e4 LT |
357 | if (-EIOCBQUEUED == ret) |
358 | ret = wait_on_sync_kiocb(&kiocb); | |
359 | *ppos = kiocb.ki_pos; | |
360 | return ret; | |
361 | } | |
362 | ||
363 | EXPORT_SYMBOL(do_sync_read); | |
364 | ||
365 | ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos) | |
366 | { | |
367 | ssize_t ret; | |
368 | ||
369 | if (!(file->f_mode & FMODE_READ)) | |
370 | return -EBADF; | |
371 | if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read)) | |
372 | return -EINVAL; | |
373 | if (unlikely(!access_ok(VERIFY_WRITE, buf, count))) | |
374 | return -EFAULT; | |
375 | ||
376 | ret = rw_verify_area(READ, file, pos, count); | |
e28cc715 LT |
377 | if (ret >= 0) { |
378 | count = ret; | |
c43e259c JM |
379 | if (file->f_op->read) |
380 | ret = file->f_op->read(file, buf, count, pos); | |
381 | else | |
382 | ret = do_sync_read(file, buf, count, pos); | |
383 | if (ret > 0) { | |
2a12a9d7 | 384 | fsnotify_access(file); |
c43e259c | 385 | add_rchar(current, ret); |
1da177e4 | 386 | } |
c43e259c | 387 | inc_syscr(current); |
1da177e4 LT |
388 | } |
389 | ||
390 | return ret; | |
391 | } | |
392 | ||
393 | EXPORT_SYMBOL(vfs_read); | |
394 | ||
395 | ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) | |
396 | { | |
027445c3 | 397 | struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len }; |
1da177e4 LT |
398 | struct kiocb kiocb; |
399 | ssize_t ret; | |
400 | ||
401 | init_sync_kiocb(&kiocb, filp); | |
402 | kiocb.ki_pos = *ppos; | |
027445c3 | 403 | kiocb.ki_left = len; |
61964eba | 404 | kiocb.ki_nbytes = len; |
027445c3 BP |
405 | |
406 | for (;;) { | |
407 | ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos); | |
408 | if (ret != -EIOCBRETRY) | |
409 | break; | |
63e68809 | 410 | wait_on_retry_sync_kiocb(&kiocb); |
027445c3 | 411 | } |
63e68809 | 412 | |
1da177e4 LT |
413 | if (-EIOCBQUEUED == ret) |
414 | ret = wait_on_sync_kiocb(&kiocb); | |
415 | *ppos = kiocb.ki_pos; | |
416 | return ret; | |
417 | } | |
418 | ||
419 | EXPORT_SYMBOL(do_sync_write); | |
420 | ||
06ae43f3 AV |
421 | ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos) |
422 | { | |
423 | mm_segment_t old_fs; | |
424 | const char __user *p; | |
425 | ssize_t ret; | |
426 | ||
3e84f48e AV |
427 | if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write)) |
428 | return -EINVAL; | |
429 | ||
06ae43f3 AV |
430 | old_fs = get_fs(); |
431 | set_fs(get_ds()); | |
432 | p = (__force const char __user *)buf; | |
433 | if (count > MAX_RW_COUNT) | |
434 | count = MAX_RW_COUNT; | |
435 | if (file->f_op->write) | |
436 | ret = file->f_op->write(file, p, count, pos); | |
437 | else | |
438 | ret = do_sync_write(file, p, count, pos); | |
439 | set_fs(old_fs); | |
440 | if (ret > 0) { | |
441 | fsnotify_modify(file); | |
442 | add_wchar(current, ret); | |
443 | } | |
444 | inc_syscw(current); | |
445 | return ret; | |
446 | } | |
447 | ||
1da177e4 LT |
448 | ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos) |
449 | { | |
450 | ssize_t ret; | |
451 | ||
452 | if (!(file->f_mode & FMODE_WRITE)) | |
453 | return -EBADF; | |
454 | if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write)) | |
455 | return -EINVAL; | |
456 | if (unlikely(!access_ok(VERIFY_READ, buf, count))) | |
457 | return -EFAULT; | |
458 | ||
459 | ret = rw_verify_area(WRITE, file, pos, count); | |
e28cc715 LT |
460 | if (ret >= 0) { |
461 | count = ret; | |
c43e259c JM |
462 | if (file->f_op->write) |
463 | ret = file->f_op->write(file, buf, count, pos); | |
464 | else | |
465 | ret = do_sync_write(file, buf, count, pos); | |
466 | if (ret > 0) { | |
2a12a9d7 | 467 | fsnotify_modify(file); |
c43e259c | 468 | add_wchar(current, ret); |
1da177e4 | 469 | } |
c43e259c | 470 | inc_syscw(current); |
1da177e4 LT |
471 | } |
472 | ||
473 | return ret; | |
474 | } | |
475 | ||
476 | EXPORT_SYMBOL(vfs_write); | |
477 | ||
478 | static inline loff_t file_pos_read(struct file *file) | |
479 | { | |
480 | return file->f_pos; | |
481 | } | |
482 | ||
483 | static inline void file_pos_write(struct file *file, loff_t pos) | |
484 | { | |
485 | file->f_pos = pos; | |
486 | } | |
487 | ||
3cdad428 | 488 | SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count) |
1da177e4 | 489 | { |
2903ff01 | 490 | struct fd f = fdget(fd); |
1da177e4 | 491 | ssize_t ret = -EBADF; |
1da177e4 | 492 | |
2903ff01 AV |
493 | if (f.file) { |
494 | loff_t pos = file_pos_read(f.file); | |
495 | ret = vfs_read(f.file, buf, count, &pos); | |
496 | file_pos_write(f.file, pos); | |
497 | fdput(f); | |
1da177e4 | 498 | } |
1da177e4 LT |
499 | return ret; |
500 | } | |
1da177e4 | 501 | |
3cdad428 HC |
502 | SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf, |
503 | size_t, count) | |
1da177e4 | 504 | { |
2903ff01 | 505 | struct fd f = fdget(fd); |
1da177e4 | 506 | ssize_t ret = -EBADF; |
1da177e4 | 507 | |
2903ff01 AV |
508 | if (f.file) { |
509 | loff_t pos = file_pos_read(f.file); | |
510 | ret = vfs_write(f.file, buf, count, &pos); | |
511 | file_pos_write(f.file, pos); | |
512 | fdput(f); | |
1da177e4 LT |
513 | } |
514 | ||
515 | return ret; | |
516 | } | |
517 | ||
6673e0c3 HC |
518 | SYSCALL_DEFINE(pread64)(unsigned int fd, 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_PREAD) |
531 | ret = vfs_read(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_pread64(long fd, long buf, long count, loff_t pos) | |
539 | { | |
540 | return SYSC_pread64((unsigned int) fd, (char __user *) buf, | |
541 | (size_t) count, pos); | |
542 | } | |
543 | SYSCALL_ALIAS(sys_pread64, SyS_pread64); | |
544 | #endif | |
1da177e4 | 545 | |
6673e0c3 HC |
546 | SYSCALL_DEFINE(pwrite64)(unsigned int fd, const char __user *buf, |
547 | size_t count, loff_t pos) | |
1da177e4 | 548 | { |
2903ff01 | 549 | struct fd f; |
1da177e4 | 550 | ssize_t ret = -EBADF; |
1da177e4 LT |
551 | |
552 | if (pos < 0) | |
553 | return -EINVAL; | |
554 | ||
2903ff01 AV |
555 | f = fdget(fd); |
556 | if (f.file) { | |
1da177e4 | 557 | ret = -ESPIPE; |
2903ff01 AV |
558 | if (f.file->f_mode & FMODE_PWRITE) |
559 | ret = vfs_write(f.file, buf, count, &pos); | |
560 | fdput(f); | |
1da177e4 LT |
561 | } |
562 | ||
563 | return ret; | |
564 | } | |
6673e0c3 HC |
565 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
566 | asmlinkage long SyS_pwrite64(long fd, long buf, long count, loff_t pos) | |
567 | { | |
568 | return SYSC_pwrite64((unsigned int) fd, (const char __user *) buf, | |
569 | (size_t) count, pos); | |
570 | } | |
571 | SYSCALL_ALIAS(sys_pwrite64, SyS_pwrite64); | |
572 | #endif | |
1da177e4 LT |
573 | |
574 | /* | |
575 | * Reduce an iovec's length in-place. Return the resulting number of segments | |
576 | */ | |
577 | unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to) | |
578 | { | |
579 | unsigned long seg = 0; | |
580 | size_t len = 0; | |
581 | ||
582 | while (seg < nr_segs) { | |
583 | seg++; | |
584 | if (len + iov->iov_len >= to) { | |
585 | iov->iov_len = to - len; | |
586 | break; | |
587 | } | |
588 | len += iov->iov_len; | |
589 | iov++; | |
590 | } | |
591 | return seg; | |
592 | } | |
19295529 | 593 | EXPORT_SYMBOL(iov_shorten); |
1da177e4 | 594 | |
ee0b3e67 BP |
595 | ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov, |
596 | unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn) | |
597 | { | |
598 | struct kiocb kiocb; | |
599 | ssize_t ret; | |
600 | ||
601 | init_sync_kiocb(&kiocb, filp); | |
602 | kiocb.ki_pos = *ppos; | |
603 | kiocb.ki_left = len; | |
604 | kiocb.ki_nbytes = len; | |
605 | ||
606 | for (;;) { | |
607 | ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos); | |
608 | if (ret != -EIOCBRETRY) | |
609 | break; | |
610 | wait_on_retry_sync_kiocb(&kiocb); | |
611 | } | |
612 | ||
613 | if (ret == -EIOCBQUEUED) | |
614 | ret = wait_on_sync_kiocb(&kiocb); | |
615 | *ppos = kiocb.ki_pos; | |
616 | return ret; | |
617 | } | |
618 | ||
619 | /* Do it by hand, with file-ops */ | |
620 | ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov, | |
621 | unsigned long nr_segs, loff_t *ppos, io_fn_t fn) | |
622 | { | |
623 | struct iovec *vector = iov; | |
624 | ssize_t ret = 0; | |
625 | ||
626 | while (nr_segs > 0) { | |
627 | void __user *base; | |
628 | size_t len; | |
629 | ssize_t nr; | |
630 | ||
631 | base = vector->iov_base; | |
632 | len = vector->iov_len; | |
633 | vector++; | |
634 | nr_segs--; | |
635 | ||
636 | nr = fn(filp, base, len, ppos); | |
637 | ||
638 | if (nr < 0) { | |
639 | if (!ret) | |
640 | ret = nr; | |
641 | break; | |
642 | } | |
643 | ret += nr; | |
644 | if (nr != len) | |
645 | break; | |
646 | } | |
647 | ||
648 | return ret; | |
649 | } | |
650 | ||
1da177e4 LT |
651 | /* A write operation does a read from user space and vice versa */ |
652 | #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ) | |
653 | ||
eed4e51f BP |
654 | ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector, |
655 | unsigned long nr_segs, unsigned long fast_segs, | |
656 | struct iovec *fast_pointer, | |
ac34ebb3 | 657 | struct iovec **ret_pointer) |
435f49a5 | 658 | { |
eed4e51f | 659 | unsigned long seg; |
435f49a5 | 660 | ssize_t ret; |
eed4e51f BP |
661 | struct iovec *iov = fast_pointer; |
662 | ||
435f49a5 LT |
663 | /* |
664 | * SuS says "The readv() function *may* fail if the iovcnt argument | |
665 | * was less than or equal to 0, or greater than {IOV_MAX}. Linux has | |
666 | * traditionally returned zero for zero segments, so... | |
667 | */ | |
eed4e51f BP |
668 | if (nr_segs == 0) { |
669 | ret = 0; | |
435f49a5 | 670 | goto out; |
eed4e51f BP |
671 | } |
672 | ||
435f49a5 LT |
673 | /* |
674 | * First get the "struct iovec" from user memory and | |
675 | * verify all the pointers | |
676 | */ | |
eed4e51f BP |
677 | if (nr_segs > UIO_MAXIOV) { |
678 | ret = -EINVAL; | |
435f49a5 | 679 | goto out; |
eed4e51f BP |
680 | } |
681 | if (nr_segs > fast_segs) { | |
435f49a5 | 682 | iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL); |
eed4e51f BP |
683 | if (iov == NULL) { |
684 | ret = -ENOMEM; | |
435f49a5 | 685 | goto out; |
eed4e51f | 686 | } |
435f49a5 | 687 | } |
eed4e51f BP |
688 | if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) { |
689 | ret = -EFAULT; | |
435f49a5 | 690 | goto out; |
eed4e51f BP |
691 | } |
692 | ||
435f49a5 | 693 | /* |
eed4e51f BP |
694 | * According to the Single Unix Specification we should return EINVAL |
695 | * if an element length is < 0 when cast to ssize_t or if the | |
696 | * total length would overflow the ssize_t return value of the | |
697 | * system call. | |
435f49a5 LT |
698 | * |
699 | * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the | |
700 | * overflow case. | |
701 | */ | |
eed4e51f | 702 | ret = 0; |
435f49a5 LT |
703 | for (seg = 0; seg < nr_segs; seg++) { |
704 | void __user *buf = iov[seg].iov_base; | |
705 | ssize_t len = (ssize_t)iov[seg].iov_len; | |
eed4e51f BP |
706 | |
707 | /* see if we we're about to use an invalid len or if | |
708 | * it's about to overflow ssize_t */ | |
435f49a5 | 709 | if (len < 0) { |
eed4e51f | 710 | ret = -EINVAL; |
435f49a5 | 711 | goto out; |
eed4e51f | 712 | } |
ac34ebb3 | 713 | if (type >= 0 |
fcf63409 | 714 | && unlikely(!access_ok(vrfy_dir(type), buf, len))) { |
eed4e51f | 715 | ret = -EFAULT; |
435f49a5 LT |
716 | goto out; |
717 | } | |
718 | if (len > MAX_RW_COUNT - ret) { | |
719 | len = MAX_RW_COUNT - ret; | |
720 | iov[seg].iov_len = len; | |
eed4e51f | 721 | } |
eed4e51f | 722 | ret += len; |
435f49a5 | 723 | } |
eed4e51f BP |
724 | out: |
725 | *ret_pointer = iov; | |
726 | return ret; | |
727 | } | |
728 | ||
1da177e4 LT |
729 | static ssize_t do_readv_writev(int type, struct file *file, |
730 | const struct iovec __user * uvector, | |
731 | unsigned long nr_segs, loff_t *pos) | |
732 | { | |
1da177e4 LT |
733 | size_t tot_len; |
734 | struct iovec iovstack[UIO_FASTIOV]; | |
ee0b3e67 | 735 | struct iovec *iov = iovstack; |
1da177e4 | 736 | ssize_t ret; |
1da177e4 LT |
737 | io_fn_t fn; |
738 | iov_fn_t fnv; | |
739 | ||
eed4e51f BP |
740 | if (!file->f_op) { |
741 | ret = -EINVAL; | |
1da177e4 | 742 | goto out; |
1da177e4 | 743 | } |
1da177e4 | 744 | |
eed4e51f | 745 | ret = rw_copy_check_uvector(type, uvector, nr_segs, |
ac34ebb3 | 746 | ARRAY_SIZE(iovstack), iovstack, &iov); |
eed4e51f | 747 | if (ret <= 0) |
1da177e4 | 748 | goto out; |
1da177e4 | 749 | |
eed4e51f | 750 | tot_len = ret; |
1da177e4 | 751 | ret = rw_verify_area(type, file, pos, tot_len); |
e28cc715 | 752 | if (ret < 0) |
411b67b4 | 753 | goto out; |
1da177e4 LT |
754 | |
755 | fnv = NULL; | |
756 | if (type == READ) { | |
757 | fn = file->f_op->read; | |
ee0b3e67 | 758 | fnv = file->f_op->aio_read; |
1da177e4 LT |
759 | } else { |
760 | fn = (io_fn_t)file->f_op->write; | |
ee0b3e67 | 761 | fnv = file->f_op->aio_write; |
1da177e4 LT |
762 | } |
763 | ||
ee0b3e67 BP |
764 | if (fnv) |
765 | ret = do_sync_readv_writev(file, iov, nr_segs, tot_len, | |
766 | pos, fnv); | |
767 | else | |
768 | ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn); | |
1da177e4 | 769 | |
1da177e4 LT |
770 | out: |
771 | if (iov != iovstack) | |
772 | kfree(iov); | |
0eeca283 RL |
773 | if ((ret + (type == READ)) > 0) { |
774 | if (type == READ) | |
2a12a9d7 | 775 | fsnotify_access(file); |
0eeca283 | 776 | else |
2a12a9d7 | 777 | fsnotify_modify(file); |
0eeca283 | 778 | } |
1da177e4 | 779 | return ret; |
1da177e4 LT |
780 | } |
781 | ||
782 | ssize_t vfs_readv(struct file *file, const struct iovec __user *vec, | |
783 | unsigned long vlen, loff_t *pos) | |
784 | { | |
785 | if (!(file->f_mode & FMODE_READ)) | |
786 | return -EBADF; | |
ee0b3e67 | 787 | if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read)) |
1da177e4 LT |
788 | return -EINVAL; |
789 | ||
790 | return do_readv_writev(READ, file, vec, vlen, pos); | |
791 | } | |
792 | ||
793 | EXPORT_SYMBOL(vfs_readv); | |
794 | ||
795 | ssize_t vfs_writev(struct file *file, const struct iovec __user *vec, | |
796 | unsigned long vlen, loff_t *pos) | |
797 | { | |
798 | if (!(file->f_mode & FMODE_WRITE)) | |
799 | return -EBADF; | |
ee0b3e67 | 800 | if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write)) |
1da177e4 LT |
801 | return -EINVAL; |
802 | ||
803 | return do_readv_writev(WRITE, file, vec, vlen, pos); | |
804 | } | |
805 | ||
806 | EXPORT_SYMBOL(vfs_writev); | |
807 | ||
3cdad428 HC |
808 | SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec, |
809 | unsigned long, vlen) | |
1da177e4 | 810 | { |
2903ff01 | 811 | struct fd f = fdget(fd); |
1da177e4 | 812 | ssize_t ret = -EBADF; |
1da177e4 | 813 | |
2903ff01 AV |
814 | if (f.file) { |
815 | loff_t pos = file_pos_read(f.file); | |
816 | ret = vfs_readv(f.file, vec, vlen, &pos); | |
817 | file_pos_write(f.file, pos); | |
818 | fdput(f); | |
1da177e4 LT |
819 | } |
820 | ||
821 | if (ret > 0) | |
4b98d11b AD |
822 | add_rchar(current, ret); |
823 | inc_syscr(current); | |
1da177e4 LT |
824 | return ret; |
825 | } | |
826 | ||
3cdad428 HC |
827 | SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec, |
828 | unsigned long, vlen) | |
1da177e4 | 829 | { |
2903ff01 | 830 | struct fd f = fdget(fd); |
1da177e4 | 831 | ssize_t ret = -EBADF; |
1da177e4 | 832 | |
2903ff01 AV |
833 | if (f.file) { |
834 | loff_t pos = file_pos_read(f.file); | |
835 | ret = vfs_writev(f.file, vec, vlen, &pos); | |
836 | file_pos_write(f.file, pos); | |
837 | fdput(f); | |
1da177e4 LT |
838 | } |
839 | ||
840 | if (ret > 0) | |
4b98d11b AD |
841 | add_wchar(current, ret); |
842 | inc_syscw(current); | |
1da177e4 LT |
843 | return ret; |
844 | } | |
845 | ||
601cc11d LT |
846 | static inline loff_t pos_from_hilo(unsigned long high, unsigned long low) |
847 | { | |
848 | #define HALF_LONG_BITS (BITS_PER_LONG / 2) | |
849 | return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low; | |
850 | } | |
851 | ||
f3554f4b | 852 | SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec, |
601cc11d | 853 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) |
f3554f4b | 854 | { |
601cc11d | 855 | loff_t pos = pos_from_hilo(pos_h, pos_l); |
2903ff01 | 856 | struct fd f; |
f3554f4b | 857 | ssize_t ret = -EBADF; |
f3554f4b GH |
858 | |
859 | if (pos < 0) | |
860 | return -EINVAL; | |
861 | ||
2903ff01 AV |
862 | f = fdget(fd); |
863 | if (f.file) { | |
f3554f4b | 864 | ret = -ESPIPE; |
2903ff01 AV |
865 | if (f.file->f_mode & FMODE_PREAD) |
866 | ret = vfs_readv(f.file, vec, vlen, &pos); | |
867 | fdput(f); | |
f3554f4b GH |
868 | } |
869 | ||
870 | if (ret > 0) | |
871 | add_rchar(current, ret); | |
872 | inc_syscr(current); | |
873 | return ret; | |
874 | } | |
875 | ||
876 | SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec, | |
601cc11d | 877 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h) |
f3554f4b | 878 | { |
601cc11d | 879 | loff_t pos = pos_from_hilo(pos_h, pos_l); |
2903ff01 | 880 | struct fd f; |
f3554f4b | 881 | ssize_t ret = -EBADF; |
f3554f4b GH |
882 | |
883 | if (pos < 0) | |
884 | return -EINVAL; | |
885 | ||
2903ff01 AV |
886 | f = fdget(fd); |
887 | if (f.file) { | |
f3554f4b | 888 | ret = -ESPIPE; |
2903ff01 AV |
889 | if (f.file->f_mode & FMODE_PWRITE) |
890 | ret = vfs_writev(f.file, vec, vlen, &pos); | |
891 | fdput(f); | |
f3554f4b GH |
892 | } |
893 | ||
894 | if (ret > 0) | |
895 | add_wchar(current, ret); | |
896 | inc_syscw(current); | |
897 | return ret; | |
898 | } | |
899 | ||
8f9c0119 CM |
900 | ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos, size_t count, |
901 | loff_t max) | |
1da177e4 | 902 | { |
2903ff01 AV |
903 | struct fd in, out; |
904 | struct inode *in_inode, *out_inode; | |
1da177e4 LT |
905 | loff_t pos; |
906 | ssize_t retval; | |
2903ff01 | 907 | int fl; |
1da177e4 LT |
908 | |
909 | /* | |
910 | * Get input file, and verify that it is ok.. | |
911 | */ | |
912 | retval = -EBADF; | |
2903ff01 AV |
913 | in = fdget(in_fd); |
914 | if (!in.file) | |
1da177e4 | 915 | goto out; |
2903ff01 | 916 | if (!(in.file->f_mode & FMODE_READ)) |
1da177e4 | 917 | goto fput_in; |
1da177e4 LT |
918 | retval = -ESPIPE; |
919 | if (!ppos) | |
2903ff01 | 920 | ppos = &in.file->f_pos; |
1da177e4 | 921 | else |
2903ff01 | 922 | if (!(in.file->f_mode & FMODE_PREAD)) |
1da177e4 | 923 | goto fput_in; |
2903ff01 | 924 | retval = rw_verify_area(READ, in.file, ppos, count); |
e28cc715 | 925 | if (retval < 0) |
1da177e4 | 926 | goto fput_in; |
e28cc715 | 927 | count = retval; |
1da177e4 | 928 | |
1da177e4 LT |
929 | /* |
930 | * Get output file, and verify that it is ok.. | |
931 | */ | |
932 | retval = -EBADF; | |
2903ff01 AV |
933 | out = fdget(out_fd); |
934 | if (!out.file) | |
1da177e4 | 935 | goto fput_in; |
2903ff01 | 936 | if (!(out.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
937 | goto fput_out; |
938 | retval = -EINVAL; | |
496ad9aa AV |
939 | in_inode = file_inode(in.file); |
940 | out_inode = file_inode(out.file); | |
2903ff01 | 941 | retval = rw_verify_area(WRITE, out.file, &out.file->f_pos, count); |
e28cc715 | 942 | if (retval < 0) |
1da177e4 | 943 | goto fput_out; |
e28cc715 | 944 | count = retval; |
1da177e4 | 945 | |
1da177e4 LT |
946 | if (!max) |
947 | max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes); | |
948 | ||
949 | pos = *ppos; | |
1da177e4 LT |
950 | if (unlikely(pos + count > max)) { |
951 | retval = -EOVERFLOW; | |
952 | if (pos >= max) | |
953 | goto fput_out; | |
954 | count = max - pos; | |
955 | } | |
956 | ||
d96e6e71 | 957 | fl = 0; |
534f2aaa | 958 | #if 0 |
d96e6e71 JA |
959 | /* |
960 | * We need to debate whether we can enable this or not. The | |
961 | * man page documents EAGAIN return for the output at least, | |
962 | * and the application is arguably buggy if it doesn't expect | |
963 | * EAGAIN on a non-blocking file descriptor. | |
964 | */ | |
2903ff01 | 965 | if (in.file->f_flags & O_NONBLOCK) |
d96e6e71 | 966 | fl = SPLICE_F_NONBLOCK; |
534f2aaa | 967 | #endif |
2903ff01 | 968 | retval = do_splice_direct(in.file, ppos, out.file, count, fl); |
1da177e4 LT |
969 | |
970 | if (retval > 0) { | |
4b98d11b AD |
971 | add_rchar(current, retval); |
972 | add_wchar(current, retval); | |
a68c2f12 SW |
973 | fsnotify_access(in.file); |
974 | fsnotify_modify(out.file); | |
1da177e4 | 975 | } |
1da177e4 | 976 | |
4b98d11b AD |
977 | inc_syscr(current); |
978 | inc_syscw(current); | |
1da177e4 LT |
979 | if (*ppos > max) |
980 | retval = -EOVERFLOW; | |
981 | ||
982 | fput_out: | |
2903ff01 | 983 | fdput(out); |
1da177e4 | 984 | fput_in: |
2903ff01 | 985 | fdput(in); |
1da177e4 LT |
986 | out: |
987 | return retval; | |
988 | } | |
989 | ||
002c8976 | 990 | SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count) |
1da177e4 LT |
991 | { |
992 | loff_t pos; | |
993 | off_t off; | |
994 | ssize_t ret; | |
995 | ||
996 | if (offset) { | |
997 | if (unlikely(get_user(off, offset))) | |
998 | return -EFAULT; | |
999 | pos = off; | |
1000 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS); | |
1001 | if (unlikely(put_user(pos, offset))) | |
1002 | return -EFAULT; | |
1003 | return ret; | |
1004 | } | |
1005 | ||
1006 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1007 | } | |
1008 | ||
002c8976 | 1009 | SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count) |
1da177e4 LT |
1010 | { |
1011 | loff_t pos; | |
1012 | ssize_t ret; | |
1013 | ||
1014 | if (offset) { | |
1015 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t)))) | |
1016 | return -EFAULT; | |
1017 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0); | |
1018 | if (unlikely(put_user(pos, offset))) | |
1019 | return -EFAULT; | |
1020 | return ret; | |
1021 | } | |
1022 | ||
1023 | return do_sendfile(out_fd, in_fd, NULL, count, 0); | |
1024 | } |