blob: 1c177f29e1b72be0aa63f38496ed81727d01f653 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/read_write.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/slab.h>
8#include <linux/stat.h>
9#include <linux/fcntl.h>
10#include <linux/file.h>
11#include <linux/uio.h>
12#include <linux/smp_lock.h>
Robert Love0eeca282005-07-12 17:06:03 -040013#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/security.h>
15#include <linux/module.h>
16#include <linux/syscalls.h>
Linus Torvaldse28cc712006-01-04 16:20:40 -080017#include <linux/pagemap.h>
Jens Axboed6b29d72007-06-04 09:59:47 +020018#include <linux/splice.h>
Badari Pulavartyee0b3e62006-09-30 23:28:47 -070019#include "read_write.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070020
21#include <asm/uaccess.h>
22#include <asm/unistd.h>
23
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080024const struct file_operations generic_ro_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070025 .llseek = generic_file_llseek,
Badari Pulavarty543ade12006-09-30 23:28:48 -070026 .read = do_sync_read,
27 .aio_read = generic_file_aio_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070028 .mmap = generic_file_readonly_mmap,
Jens Axboe534f2aa2007-06-01 14:52:37 +020029 .splice_read = generic_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070030};
31
32EXPORT_SYMBOL(generic_ro_fops);
33
34loff_t generic_file_llseek(struct file *file, loff_t offset, int origin)
35{
36 long long retval;
37 struct inode *inode = file->f_mapping->host;
38
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080039 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 switch (origin) {
Chris Snook7b8e8922007-05-08 00:24:13 -070041 case SEEK_END:
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 offset += inode->i_size;
43 break;
Chris Snook7b8e8922007-05-08 00:24:13 -070044 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 offset += file->f_pos;
46 }
47 retval = -EINVAL;
48 if (offset>=0 && offset<=inode->i_sb->s_maxbytes) {
49 if (offset != file->f_pos) {
50 file->f_pos = offset;
51 file->f_version = 0;
52 }
53 retval = offset;
54 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080055 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070056 return retval;
57}
58
59EXPORT_SYMBOL(generic_file_llseek);
60
61loff_t remote_llseek(struct file *file, loff_t offset, int origin)
62{
63 long long retval;
64
65 lock_kernel();
66 switch (origin) {
Chris Snook7b8e8922007-05-08 00:24:13 -070067 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080068 offset += i_size_read(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 break;
Chris Snook7b8e8922007-05-08 00:24:13 -070070 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 offset += file->f_pos;
72 }
73 retval = -EINVAL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080074 if (offset>=0 && offset<=file->f_path.dentry->d_inode->i_sb->s_maxbytes) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (offset != file->f_pos) {
76 file->f_pos = offset;
77 file->f_version = 0;
78 }
79 retval = offset;
80 }
81 unlock_kernel();
82 return retval;
83}
84EXPORT_SYMBOL(remote_llseek);
85
86loff_t no_llseek(struct file *file, loff_t offset, int origin)
87{
88 return -ESPIPE;
89}
90EXPORT_SYMBOL(no_llseek);
91
92loff_t default_llseek(struct file *file, loff_t offset, int origin)
93{
94 long long retval;
95
96 lock_kernel();
97 switch (origin) {
Chris Snook7b8e8922007-05-08 00:24:13 -070098 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -080099 offset += i_size_read(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 break;
Chris Snook7b8e8922007-05-08 00:24:13 -0700101 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 offset += file->f_pos;
103 }
104 retval = -EINVAL;
105 if (offset >= 0) {
106 if (offset != file->f_pos) {
107 file->f_pos = offset;
108 file->f_version = 0;
109 }
110 retval = offset;
111 }
112 unlock_kernel();
113 return retval;
114}
115EXPORT_SYMBOL(default_llseek);
116
117loff_t vfs_llseek(struct file *file, loff_t offset, int origin)
118{
119 loff_t (*fn)(struct file *, loff_t, int);
120
121 fn = no_llseek;
122 if (file->f_mode & FMODE_LSEEK) {
123 fn = default_llseek;
124 if (file->f_op && file->f_op->llseek)
125 fn = file->f_op->llseek;
126 }
127 return fn(file, offset, origin);
128}
129EXPORT_SYMBOL(vfs_llseek);
130
131asmlinkage off_t sys_lseek(unsigned int fd, off_t offset, unsigned int origin)
132{
133 off_t retval;
134 struct file * file;
135 int fput_needed;
136
137 retval = -EBADF;
138 file = fget_light(fd, &fput_needed);
139 if (!file)
140 goto bad;
141
142 retval = -EINVAL;
Chris Snook1ae70752007-05-08 00:24:15 -0700143 if (origin <= SEEK_MAX) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 loff_t res = vfs_llseek(file, offset, origin);
145 retval = res;
146 if (res != (loff_t)retval)
147 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
148 }
149 fput_light(file, fput_needed);
150bad:
151 return retval;
152}
153
154#ifdef __ARCH_WANT_SYS_LLSEEK
155asmlinkage long sys_llseek(unsigned int fd, unsigned long offset_high,
156 unsigned long offset_low, loff_t __user * result,
157 unsigned int origin)
158{
159 int retval;
160 struct file * file;
161 loff_t offset;
162 int fput_needed;
163
164 retval = -EBADF;
165 file = fget_light(fd, &fput_needed);
166 if (!file)
167 goto bad;
168
169 retval = -EINVAL;
Chris Snook1ae70752007-05-08 00:24:15 -0700170 if (origin > SEEK_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 goto out_putf;
172
173 offset = vfs_llseek(file, ((loff_t) offset_high << 32) | offset_low,
174 origin);
175
176 retval = (int)offset;
177 if (offset >= 0) {
178 retval = -EFAULT;
179 if (!copy_to_user(result, &offset, sizeof(offset)))
180 retval = 0;
181 }
182out_putf:
183 fput_light(file, fput_needed);
184bad:
185 return retval;
186}
187#endif
188
Linus Torvaldse28cc712006-01-04 16:20:40 -0800189/*
190 * rw_verify_area doesn't like huge counts. We limit
191 * them to something that fits in "int" so that others
192 * won't have to do range checks all the time.
193 */
194#define MAX_RW_COUNT (INT_MAX & PAGE_CACHE_MASK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
196int rw_verify_area(int read_write, struct file *file, loff_t *ppos, size_t count)
197{
198 struct inode *inode;
199 loff_t pos;
James Morrisc43e2592008-01-12 22:05:48 +1100200 int retval = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
Eric Dumazet163da952007-02-12 00:52:24 -0800202 inode = file->f_path.dentry->d_inode;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800203 if (unlikely((ssize_t) count < 0))
James Morrisc43e2592008-01-12 22:05:48 +1100204 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 pos = *ppos;
206 if (unlikely((pos < 0) || (loff_t) (pos + count) < 0))
James Morrisc43e2592008-01-12 22:05:48 +1100207 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
Pavel Emelyanova16877c2007-10-01 14:41:11 -0700209 if (unlikely(inode->i_flock && mandatory_lock(inode))) {
James Morrisc43e2592008-01-12 22:05:48 +1100210 retval = locks_mandatory_area(
Linus Torvaldse28cc712006-01-04 16:20:40 -0800211 read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
212 inode, file, pos, count);
213 if (retval < 0)
214 return retval;
215 }
James Morrisc43e2592008-01-12 22:05:48 +1100216 retval = security_file_permission(file,
217 read_write == READ ? MAY_READ : MAY_WRITE);
218 if (retval)
219 return retval;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800220 return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221}
222
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700223static void wait_on_retry_sync_kiocb(struct kiocb *iocb)
224{
225 set_current_state(TASK_UNINTERRUPTIBLE);
226 if (!kiocbIsKicked(iocb))
227 schedule();
228 else
229 kiocbClearKicked(iocb);
230 __set_current_state(TASK_RUNNING);
231}
232
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
234{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700235 struct iovec iov = { .iov_base = buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 struct kiocb kiocb;
237 ssize_t ret;
238
239 init_sync_kiocb(&kiocb, filp);
240 kiocb.ki_pos = *ppos;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700241 kiocb.ki_left = len;
242
243 for (;;) {
244 ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
245 if (ret != -EIOCBRETRY)
246 break;
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700247 wait_on_retry_sync_kiocb(&kiocb);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700248 }
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 if (-EIOCBQUEUED == ret)
251 ret = wait_on_sync_kiocb(&kiocb);
252 *ppos = kiocb.ki_pos;
253 return ret;
254}
255
256EXPORT_SYMBOL(do_sync_read);
257
258ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
259{
260 ssize_t ret;
261
262 if (!(file->f_mode & FMODE_READ))
263 return -EBADF;
264 if (!file->f_op || (!file->f_op->read && !file->f_op->aio_read))
265 return -EINVAL;
266 if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
267 return -EFAULT;
268
269 ret = rw_verify_area(READ, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800270 if (ret >= 0) {
271 count = ret;
James Morrisc43e2592008-01-12 22:05:48 +1100272 if (file->f_op->read)
273 ret = file->f_op->read(file, buf, count, pos);
274 else
275 ret = do_sync_read(file, buf, count, pos);
276 if (ret > 0) {
277 fsnotify_access(file->f_path.dentry);
278 add_rchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 }
James Morrisc43e2592008-01-12 22:05:48 +1100280 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 }
282
283 return ret;
284}
285
286EXPORT_SYMBOL(vfs_read);
287
288ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
289{
Badari Pulavarty027445c2006-09-30 23:28:46 -0700290 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 struct kiocb kiocb;
292 ssize_t ret;
293
294 init_sync_kiocb(&kiocb, filp);
295 kiocb.ki_pos = *ppos;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700296 kiocb.ki_left = len;
297
298 for (;;) {
299 ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
300 if (ret != -EIOCBRETRY)
301 break;
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700302 wait_on_retry_sync_kiocb(&kiocb);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700303 }
Benjamin LaHaise63e68802005-06-23 00:10:27 -0700304
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 if (-EIOCBQUEUED == ret)
306 ret = wait_on_sync_kiocb(&kiocb);
307 *ppos = kiocb.ki_pos;
308 return ret;
309}
310
311EXPORT_SYMBOL(do_sync_write);
312
313ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
314{
315 ssize_t ret;
316
317 if (!(file->f_mode & FMODE_WRITE))
318 return -EBADF;
319 if (!file->f_op || (!file->f_op->write && !file->f_op->aio_write))
320 return -EINVAL;
321 if (unlikely(!access_ok(VERIFY_READ, buf, count)))
322 return -EFAULT;
323
324 ret = rw_verify_area(WRITE, file, pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800325 if (ret >= 0) {
326 count = ret;
James Morrisc43e2592008-01-12 22:05:48 +1100327 if (file->f_op->write)
328 ret = file->f_op->write(file, buf, count, pos);
329 else
330 ret = do_sync_write(file, buf, count, pos);
331 if (ret > 0) {
332 fsnotify_modify(file->f_path.dentry);
333 add_wchar(current, ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 }
James Morrisc43e2592008-01-12 22:05:48 +1100335 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 }
337
338 return ret;
339}
340
341EXPORT_SYMBOL(vfs_write);
342
343static inline loff_t file_pos_read(struct file *file)
344{
345 return file->f_pos;
346}
347
348static inline void file_pos_write(struct file *file, loff_t pos)
349{
350 file->f_pos = pos;
351}
352
353asmlinkage ssize_t sys_read(unsigned int fd, char __user * buf, size_t count)
354{
355 struct file *file;
356 ssize_t ret = -EBADF;
357 int fput_needed;
358
359 file = fget_light(fd, &fput_needed);
360 if (file) {
361 loff_t pos = file_pos_read(file);
362 ret = vfs_read(file, buf, count, &pos);
363 file_pos_write(file, pos);
364 fput_light(file, fput_needed);
365 }
366
367 return ret;
368}
Arjan van de Vencb51f972007-11-14 17:00:10 -0800369EXPORT_UNUSED_SYMBOL_GPL(sys_read); /* to be deleted for 2.6.25 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
371asmlinkage ssize_t sys_write(unsigned int fd, const char __user * buf, size_t count)
372{
373 struct file *file;
374 ssize_t ret = -EBADF;
375 int fput_needed;
376
377 file = fget_light(fd, &fput_needed);
378 if (file) {
379 loff_t pos = file_pos_read(file);
380 ret = vfs_write(file, buf, count, &pos);
381 file_pos_write(file, pos);
382 fput_light(file, fput_needed);
383 }
384
385 return ret;
386}
387
388asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf,
389 size_t count, loff_t pos)
390{
391 struct file *file;
392 ssize_t ret = -EBADF;
393 int fput_needed;
394
395 if (pos < 0)
396 return -EINVAL;
397
398 file = fget_light(fd, &fput_needed);
399 if (file) {
400 ret = -ESPIPE;
401 if (file->f_mode & FMODE_PREAD)
402 ret = vfs_read(file, buf, count, &pos);
403 fput_light(file, fput_needed);
404 }
405
406 return ret;
407}
408
409asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf,
410 size_t count, loff_t pos)
411{
412 struct file *file;
413 ssize_t ret = -EBADF;
414 int fput_needed;
415
416 if (pos < 0)
417 return -EINVAL;
418
419 file = fget_light(fd, &fput_needed);
420 if (file) {
421 ret = -ESPIPE;
422 if (file->f_mode & FMODE_PWRITE)
423 ret = vfs_write(file, buf, count, &pos);
424 fput_light(file, fput_needed);
425 }
426
427 return ret;
428}
429
430/*
431 * Reduce an iovec's length in-place. Return the resulting number of segments
432 */
433unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
434{
435 unsigned long seg = 0;
436 size_t len = 0;
437
438 while (seg < nr_segs) {
439 seg++;
440 if (len + iov->iov_len >= to) {
441 iov->iov_len = to - len;
442 break;
443 }
444 len += iov->iov_len;
445 iov++;
446 }
447 return seg;
448}
Eric Sandeen19295522008-01-28 23:58:27 -0500449EXPORT_SYMBOL(iov_shorten);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700451ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
452 unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
453{
454 struct kiocb kiocb;
455 ssize_t ret;
456
457 init_sync_kiocb(&kiocb, filp);
458 kiocb.ki_pos = *ppos;
459 kiocb.ki_left = len;
460 kiocb.ki_nbytes = len;
461
462 for (;;) {
463 ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
464 if (ret != -EIOCBRETRY)
465 break;
466 wait_on_retry_sync_kiocb(&kiocb);
467 }
468
469 if (ret == -EIOCBQUEUED)
470 ret = wait_on_sync_kiocb(&kiocb);
471 *ppos = kiocb.ki_pos;
472 return ret;
473}
474
475/* Do it by hand, with file-ops */
476ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
477 unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
478{
479 struct iovec *vector = iov;
480 ssize_t ret = 0;
481
482 while (nr_segs > 0) {
483 void __user *base;
484 size_t len;
485 ssize_t nr;
486
487 base = vector->iov_base;
488 len = vector->iov_len;
489 vector++;
490 nr_segs--;
491
492 nr = fn(filp, base, len, ppos);
493
494 if (nr < 0) {
495 if (!ret)
496 ret = nr;
497 break;
498 }
499 ret += nr;
500 if (nr != len)
501 break;
502 }
503
504 return ret;
505}
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507/* A write operation does a read from user space and vice versa */
508#define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
509
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700510ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
511 unsigned long nr_segs, unsigned long fast_segs,
512 struct iovec *fast_pointer,
513 struct iovec **ret_pointer)
514 {
515 unsigned long seg;
516 ssize_t ret;
517 struct iovec *iov = fast_pointer;
518
519 /*
520 * SuS says "The readv() function *may* fail if the iovcnt argument
521 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
522 * traditionally returned zero for zero segments, so...
523 */
524 if (nr_segs == 0) {
525 ret = 0;
526 goto out;
527 }
528
529 /*
530 * First get the "struct iovec" from user memory and
531 * verify all the pointers
532 */
533 if (nr_segs > UIO_MAXIOV) {
534 ret = -EINVAL;
535 goto out;
536 }
537 if (nr_segs > fast_segs) {
538 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
539 if (iov == NULL) {
540 ret = -ENOMEM;
541 goto out;
542 }
543 }
544 if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
545 ret = -EFAULT;
546 goto out;
547 }
548
549 /*
550 * According to the Single Unix Specification we should return EINVAL
551 * if an element length is < 0 when cast to ssize_t or if the
552 * total length would overflow the ssize_t return value of the
553 * system call.
554 */
555 ret = 0;
556 for (seg = 0; seg < nr_segs; seg++) {
557 void __user *buf = iov[seg].iov_base;
558 ssize_t len = (ssize_t)iov[seg].iov_len;
559
560 /* see if we we're about to use an invalid len or if
561 * it's about to overflow ssize_t */
562 if (len < 0 || (ret + len < ret)) {
563 ret = -EINVAL;
564 goto out;
565 }
566 if (unlikely(!access_ok(vrfy_dir(type), buf, len))) {
567 ret = -EFAULT;
568 goto out;
569 }
570
571 ret += len;
572 }
573out:
574 *ret_pointer = iov;
575 return ret;
576}
577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578static ssize_t do_readv_writev(int type, struct file *file,
579 const struct iovec __user * uvector,
580 unsigned long nr_segs, loff_t *pos)
581{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 size_t tot_len;
583 struct iovec iovstack[UIO_FASTIOV];
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700584 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 io_fn_t fn;
587 iov_fn_t fnv;
588
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700589 if (!file->f_op) {
590 ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 goto out;
592 }
593
Badari Pulavartyeed4e512006-09-30 23:28:49 -0700594 ret = rw_copy_check_uvector(type, uvector, nr_segs,
595 ARRAY_SIZE(iovstack), iovstack, &iov);
596 if (ret <= 0)
597 goto out;
598
599 tot_len = ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 ret = rw_verify_area(type, file, pos, tot_len);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800601 if (ret < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 goto out;
603
604 fnv = NULL;
605 if (type == READ) {
606 fn = file->f_op->read;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700607 fnv = file->f_op->aio_read;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 } else {
609 fn = (io_fn_t)file->f_op->write;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700610 fnv = file->f_op->aio_write;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 }
612
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700613 if (fnv)
614 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
615 pos, fnv);
616 else
617 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619out:
620 if (iov != iovstack)
621 kfree(iov);
Robert Love0eeca282005-07-12 17:06:03 -0400622 if ((ret + (type == READ)) > 0) {
623 if (type == READ)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800624 fsnotify_access(file->f_path.dentry);
Robert Love0eeca282005-07-12 17:06:03 -0400625 else
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800626 fsnotify_modify(file->f_path.dentry);
Robert Love0eeca282005-07-12 17:06:03 -0400627 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629}
630
631ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
632 unsigned long vlen, loff_t *pos)
633{
634 if (!(file->f_mode & FMODE_READ))
635 return -EBADF;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700636 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 return -EINVAL;
638
639 return do_readv_writev(READ, file, vec, vlen, pos);
640}
641
642EXPORT_SYMBOL(vfs_readv);
643
644ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
645 unsigned long vlen, loff_t *pos)
646{
647 if (!(file->f_mode & FMODE_WRITE))
648 return -EBADF;
Badari Pulavartyee0b3e62006-09-30 23:28:47 -0700649 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 return -EINVAL;
651
652 return do_readv_writev(WRITE, file, vec, vlen, pos);
653}
654
655EXPORT_SYMBOL(vfs_writev);
656
657asmlinkage ssize_t
658sys_readv(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
659{
660 struct file *file;
661 ssize_t ret = -EBADF;
662 int fput_needed;
663
664 file = fget_light(fd, &fput_needed);
665 if (file) {
666 loff_t pos = file_pos_read(file);
667 ret = vfs_readv(file, vec, vlen, &pos);
668 file_pos_write(file, pos);
669 fput_light(file, fput_needed);
670 }
671
672 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800673 add_rchar(current, ret);
674 inc_syscr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 return ret;
676}
677
678asmlinkage ssize_t
679sys_writev(unsigned long fd, const struct iovec __user *vec, unsigned long vlen)
680{
681 struct file *file;
682 ssize_t ret = -EBADF;
683 int fput_needed;
684
685 file = fget_light(fd, &fput_needed);
686 if (file) {
687 loff_t pos = file_pos_read(file);
688 ret = vfs_writev(file, vec, vlen, &pos);
689 file_pos_write(file, pos);
690 fput_light(file, fput_needed);
691 }
692
693 if (ret > 0)
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800694 add_wchar(current, ret);
695 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 return ret;
697}
698
699static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
700 size_t count, loff_t max)
701{
702 struct file * in_file, * out_file;
703 struct inode * in_inode, * out_inode;
704 loff_t pos;
705 ssize_t retval;
Jens Axboe534f2aa2007-06-01 14:52:37 +0200706 int fput_needed_in, fput_needed_out, fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
708 /*
709 * Get input file, and verify that it is ok..
710 */
711 retval = -EBADF;
712 in_file = fget_light(in_fd, &fput_needed_in);
713 if (!in_file)
714 goto out;
715 if (!(in_file->f_mode & FMODE_READ))
716 goto fput_in;
717 retval = -EINVAL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800718 in_inode = in_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 if (!in_inode)
720 goto fput_in;
Jens Axboed96e6e72007-06-11 12:18:52 +0200721 if (!in_file->f_op || !in_file->f_op->splice_read)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 goto fput_in;
723 retval = -ESPIPE;
724 if (!ppos)
725 ppos = &in_file->f_pos;
726 else
727 if (!(in_file->f_mode & FMODE_PREAD))
728 goto fput_in;
729 retval = rw_verify_area(READ, in_file, ppos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800730 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 goto fput_in;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800732 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 /*
735 * Get output file, and verify that it is ok..
736 */
737 retval = -EBADF;
738 out_file = fget_light(out_fd, &fput_needed_out);
739 if (!out_file)
740 goto fput_in;
741 if (!(out_file->f_mode & FMODE_WRITE))
742 goto fput_out;
743 retval = -EINVAL;
744 if (!out_file->f_op || !out_file->f_op->sendpage)
745 goto fput_out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800746 out_inode = out_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 retval = rw_verify_area(WRITE, out_file, &out_file->f_pos, count);
Linus Torvaldse28cc712006-01-04 16:20:40 -0800748 if (retval < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 goto fput_out;
Linus Torvaldse28cc712006-01-04 16:20:40 -0800750 count = retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 if (!max)
753 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
754
755 pos = *ppos;
756 retval = -EINVAL;
757 if (unlikely(pos < 0))
758 goto fput_out;
759 if (unlikely(pos + count > max)) {
760 retval = -EOVERFLOW;
761 if (pos >= max)
762 goto fput_out;
763 count = max - pos;
764 }
765
Jens Axboed96e6e72007-06-11 12:18:52 +0200766 fl = 0;
Jens Axboe534f2aa2007-06-01 14:52:37 +0200767#if 0
Jens Axboed96e6e72007-06-11 12:18:52 +0200768 /*
769 * We need to debate whether we can enable this or not. The
770 * man page documents EAGAIN return for the output at least,
771 * and the application is arguably buggy if it doesn't expect
772 * EAGAIN on a non-blocking file descriptor.
773 */
774 if (in_file->f_flags & O_NONBLOCK)
775 fl = SPLICE_F_NONBLOCK;
Jens Axboe534f2aa2007-06-01 14:52:37 +0200776#endif
Jens Axboed96e6e72007-06-11 12:18:52 +0200777 retval = do_splice_direct(in_file, ppos, out_file, count, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 if (retval > 0) {
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800780 add_rchar(current, retval);
781 add_wchar(current, retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
Alexey Dobriyan4b98d112007-02-10 01:46:45 -0800784 inc_syscr(current);
785 inc_syscw(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 if (*ppos > max)
787 retval = -EOVERFLOW;
788
789fput_out:
790 fput_light(out_file, fput_needed_out);
791fput_in:
792 fput_light(in_file, fput_needed_in);
793out:
794 return retval;
795}
796
797asmlinkage ssize_t sys_sendfile(int out_fd, int in_fd, off_t __user *offset, size_t count)
798{
799 loff_t pos;
800 off_t off;
801 ssize_t ret;
802
803 if (offset) {
804 if (unlikely(get_user(off, offset)))
805 return -EFAULT;
806 pos = off;
807 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
808 if (unlikely(put_user(pos, offset)))
809 return -EFAULT;
810 return ret;
811 }
812
813 return do_sendfile(out_fd, in_fd, NULL, count, 0);
814}
815
816asmlinkage ssize_t sys_sendfile64(int out_fd, int in_fd, loff_t __user *offset, size_t count)
817{
818 loff_t pos;
819 ssize_t ret;
820
821 if (offset) {
822 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
823 return -EFAULT;
824 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
825 if (unlikely(put_user(pos, offset)))
826 return -EFAULT;
827 return ret;
828 }
829
830 return do_sendfile(out_fd, in_fd, NULL, count, 0);
831}