blob: 304c1c7814cbacfec0695e8714c5557ea8ee0e9b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/string.h>
8#include <linux/mm.h>
9#include <linux/utime.h>
10#include <linux/file.h>
11#include <linux/smp_lock.h>
12#include <linux/quotaops.h>
Robert Love0eeca282005-07-12 17:06:03 -040013#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/module.h>
15#include <linux/slab.h>
16#include <linux/tty.h>
17#include <linux/namei.h>
18#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080019#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/security.h>
21#include <linux/mount.h>
22#include <linux/vfs.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080023#include <linux/fcntl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <asm/uaccess.h>
25#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070026#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/pagemap.h>
28#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070029#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000030#include <linux/audit.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#include <asm/unistd.h>
33
David Howells726c3342006-06-23 02:02:58 -070034int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070035{
36 int retval = -ENODEV;
37
David Howells726c3342006-06-23 02:02:58 -070038 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 retval = -ENOSYS;
David Howells726c3342006-06-23 02:02:58 -070040 if (dentry->d_sb->s_op->statfs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 memset(buf, 0, sizeof(*buf));
David Howells726c3342006-06-23 02:02:58 -070042 retval = security_sb_statfs(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 if (retval)
44 return retval;
David Howells726c3342006-06-23 02:02:58 -070045 retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 if (retval == 0 && buf->f_frsize == 0)
47 buf->f_frsize = buf->f_bsize;
48 }
49 }
50 return retval;
51}
52
53EXPORT_SYMBOL(vfs_statfs);
54
David Howells726c3342006-06-23 02:02:58 -070055static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070056{
57 struct kstatfs st;
58 int retval;
59
David Howells726c3342006-06-23 02:02:58 -070060 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 if (retval)
62 return retval;
63
64 if (sizeof(*buf) == sizeof(st))
65 memcpy(buf, &st, sizeof(st));
66 else {
67 if (sizeof buf->f_blocks == 4) {
68 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
69 0xffffffff00000000ULL)
70 return -EOVERFLOW;
71 /*
72 * f_files and f_ffree may be -1; it's okay to stuff
73 * that into 32 bits
74 */
75 if (st.f_files != -1 &&
76 (st.f_files & 0xffffffff00000000ULL))
77 return -EOVERFLOW;
78 if (st.f_ffree != -1 &&
79 (st.f_ffree & 0xffffffff00000000ULL))
80 return -EOVERFLOW;
81 }
82
83 buf->f_type = st.f_type;
84 buf->f_bsize = st.f_bsize;
85 buf->f_blocks = st.f_blocks;
86 buf->f_bfree = st.f_bfree;
87 buf->f_bavail = st.f_bavail;
88 buf->f_files = st.f_files;
89 buf->f_ffree = st.f_ffree;
90 buf->f_fsid = st.f_fsid;
91 buf->f_namelen = st.f_namelen;
92 buf->f_frsize = st.f_frsize;
93 memset(buf->f_spare, 0, sizeof(buf->f_spare));
94 }
95 return 0;
96}
97
David Howells726c3342006-06-23 02:02:58 -070098static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070099{
100 struct kstatfs st;
101 int retval;
102
David Howells726c3342006-06-23 02:02:58 -0700103 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 if (retval)
105 return retval;
106
107 if (sizeof(*buf) == sizeof(st))
108 memcpy(buf, &st, sizeof(st));
109 else {
110 buf->f_type = st.f_type;
111 buf->f_bsize = st.f_bsize;
112 buf->f_blocks = st.f_blocks;
113 buf->f_bfree = st.f_bfree;
114 buf->f_bavail = st.f_bavail;
115 buf->f_files = st.f_files;
116 buf->f_ffree = st.f_ffree;
117 buf->f_fsid = st.f_fsid;
118 buf->f_namelen = st.f_namelen;
119 buf->f_frsize = st.f_frsize;
120 memset(buf->f_spare, 0, sizeof(buf->f_spare));
121 }
122 return 0;
123}
124
125asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
126{
127 struct nameidata nd;
128 int error;
129
130 error = user_path_walk(path, &nd);
131 if (!error) {
132 struct statfs tmp;
David Howells726c3342006-06-23 02:02:58 -0700133 error = vfs_statfs_native(nd.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
135 error = -EFAULT;
136 path_release(&nd);
137 }
138 return error;
139}
140
141
142asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
143{
144 struct nameidata nd;
145 long error;
146
147 if (sz != sizeof(*buf))
148 return -EINVAL;
149 error = user_path_walk(path, &nd);
150 if (!error) {
151 struct statfs64 tmp;
David Howells726c3342006-06-23 02:02:58 -0700152 error = vfs_statfs64(nd.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
154 error = -EFAULT;
155 path_release(&nd);
156 }
157 return error;
158}
159
160
161asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
162{
163 struct file * file;
164 struct statfs tmp;
165 int error;
166
167 error = -EBADF;
168 file = fget(fd);
169 if (!file)
170 goto out;
David Howells726c3342006-06-23 02:02:58 -0700171 error = vfs_statfs_native(file->f_dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
173 error = -EFAULT;
174 fput(file);
175out:
176 return error;
177}
178
179asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
180{
181 struct file * file;
182 struct statfs64 tmp;
183 int error;
184
185 if (sz != sizeof(*buf))
186 return -EINVAL;
187
188 error = -EBADF;
189 file = fget(fd);
190 if (!file)
191 goto out;
David Howells726c3342006-06-23 02:02:58 -0700192 error = vfs_statfs64(file->f_dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
194 error = -EFAULT;
195 fput(file);
196out:
197 return error;
198}
199
NeilBrown4a301312006-01-08 01:02:39 -0800200int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
201 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
203 int err;
204 struct iattr newattrs;
205
206 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
207 if (length < 0)
208 return -EINVAL;
209
210 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -0800211 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -0800212 if (filp) {
213 newattrs.ia_file = filp;
214 newattrs.ia_valid |= ATTR_FILE;
215 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800217 mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800219 mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
221}
222
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800223static long do_sys_truncate(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
225 struct nameidata nd;
226 struct inode * inode;
227 int error;
228
229 error = -EINVAL;
230 if (length < 0) /* sorry, but loff_t says... */
231 goto out;
232
233 error = user_path_walk(path, &nd);
234 if (error)
235 goto out;
236 inode = nd.dentry->d_inode;
237
238 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
239 error = -EISDIR;
240 if (S_ISDIR(inode->i_mode))
241 goto dput_and_out;
242
243 error = -EINVAL;
244 if (!S_ISREG(inode->i_mode))
245 goto dput_and_out;
246
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800247 error = vfs_permission(&nd, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 if (error)
249 goto dput_and_out;
250
251 error = -EROFS;
252 if (IS_RDONLY(inode))
253 goto dput_and_out;
254
255 error = -EPERM;
256 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
257 goto dput_and_out;
258
259 /*
260 * Make sure that there are no leases.
261 */
262 error = break_lease(inode, FMODE_WRITE);
263 if (error)
264 goto dput_and_out;
265
266 error = get_write_access(inode);
267 if (error)
268 goto dput_and_out;
269
270 error = locks_verify_truncate(inode, NULL, length);
271 if (!error) {
272 DQUOT_INIT(inode);
NeilBrown4a301312006-01-08 01:02:39 -0800273 error = do_truncate(nd.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 }
275 put_write_access(inode);
276
277dput_and_out:
278 path_release(&nd);
279out:
280 return error;
281}
282
283asmlinkage long sys_truncate(const char __user * path, unsigned long length)
284{
285 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
286 return do_sys_truncate(path, (long)length);
287}
288
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800289static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290{
291 struct inode * inode;
292 struct dentry *dentry;
293 struct file * file;
294 int error;
295
296 error = -EINVAL;
297 if (length < 0)
298 goto out;
299 error = -EBADF;
300 file = fget(fd);
301 if (!file)
302 goto out;
303
304 /* explicitly opened as large or we are on 64-bit box */
305 if (file->f_flags & O_LARGEFILE)
306 small = 0;
307
308 dentry = file->f_dentry;
309 inode = dentry->d_inode;
310 error = -EINVAL;
311 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
312 goto out_putf;
313
314 error = -EINVAL;
315 /* Cannot ftruncate over 2^31 bytes without large file support */
316 if (small && length > MAX_NON_LFS)
317 goto out_putf;
318
319 error = -EPERM;
320 if (IS_APPEND(inode))
321 goto out_putf;
322
323 error = locks_verify_truncate(inode, file, length);
324 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700325 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326out_putf:
327 fput(file);
328out:
329 return error;
330}
331
332asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
333{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700334 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700335 /* avoid REGPARM breakage on x86: */
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700336 prevent_tail_call(ret);
337 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338}
339
340/* LFS versions of truncate are only needed on 32 bit machines */
341#if BITS_PER_LONG == 32
342asmlinkage long sys_truncate64(const char __user * path, loff_t length)
343{
344 return do_sys_truncate(path, length);
345}
346
347asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
348{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700349 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700350 /* avoid REGPARM breakage on x86: */
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700351 prevent_tail_call(ret);
352 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353}
354#endif
355
356#ifdef __ARCH_WANT_SYS_UTIME
357
358/*
359 * sys_utime() can be implemented in user-level using sys_utimes().
360 * Is this for backwards compatibility? If so, why not move it
361 * into the appropriate arch directory (for those architectures that
362 * need it).
363 */
364
365/* If times==NULL, set access and modification to current time,
366 * must be owner or have write permission.
367 * Else, update from *times, must be owner or super user.
368 */
369asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
370{
371 int error;
372 struct nameidata nd;
373 struct inode * inode;
374 struct iattr newattrs;
375
376 error = user_path_walk(filename, &nd);
377 if (error)
378 goto out;
379 inode = nd.dentry->d_inode;
380
381 error = -EROFS;
382 if (IS_RDONLY(inode))
383 goto dput_and_out;
384
385 /* Don't worry, the checks are done in inode_change_ok() */
386 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
387 if (times) {
388 error = -EPERM;
389 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
390 goto dput_and_out;
391
392 error = get_user(newattrs.ia_atime.tv_sec, &times->actime);
393 newattrs.ia_atime.tv_nsec = 0;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800394 if (!error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 error = get_user(newattrs.ia_mtime.tv_sec, &times->modtime);
396 newattrs.ia_mtime.tv_nsec = 0;
397 if (error)
398 goto dput_and_out;
399
400 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
401 } else {
402 error = -EACCES;
403 if (IS_IMMUTABLE(inode))
404 goto dput_and_out;
405
406 if (current->fsuid != inode->i_uid &&
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800407 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 goto dput_and_out;
409 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800410 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 error = notify_change(nd.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800412 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413dput_and_out:
414 path_release(&nd);
415out:
416 return error;
417}
418
419#endif
420
421/* If times==NULL, set access and modification to current time,
422 * must be owner or have write permission.
423 * Else, update from *times, must be owner or super user.
424 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800425long do_utimes(int dfd, char __user *filename, struct timeval *times)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426{
427 int error;
428 struct nameidata nd;
429 struct inode * inode;
430 struct iattr newattrs;
431
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800432 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434 if (error)
435 goto out;
436 inode = nd.dentry->d_inode;
437
438 error = -EROFS;
439 if (IS_RDONLY(inode))
440 goto dput_and_out;
441
442 /* Don't worry, the checks are done in inode_change_ok() */
443 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
444 if (times) {
445 error = -EPERM;
446 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
447 goto dput_and_out;
448
449 newattrs.ia_atime.tv_sec = times[0].tv_sec;
450 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
451 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
452 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
453 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
454 } else {
455 error = -EACCES;
456 if (IS_IMMUTABLE(inode))
457 goto dput_and_out;
458
459 if (current->fsuid != inode->i_uid &&
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800460 (error = vfs_permission(&nd, MAY_WRITE)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 goto dput_and_out;
462 }
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800463 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 error = notify_change(nd.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800465 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466dput_and_out:
467 path_release(&nd);
468out:
469 return error;
470}
471
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800472asmlinkage long sys_futimesat(int dfd, char __user *filename, struct timeval __user *utimes)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473{
474 struct timeval times[2];
475
476 if (utimes && copy_from_user(&times, utimes, sizeof(times)))
477 return -EFAULT;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800478 return do_utimes(dfd, filename, utimes ? times : NULL);
479}
480
481asmlinkage long sys_utimes(char __user *filename, struct timeval __user *utimes)
482{
483 return sys_futimesat(AT_FDCWD, filename, utimes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484}
485
486
487/*
488 * access() needs to use the real uid/gid, not the effective uid/gid.
489 * We do this by temporarily clearing all FS-related capabilities and
490 * switching the fsuid/fsgid around to the real ones.
491 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800492asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 struct nameidata nd;
495 int old_fsuid, old_fsgid;
496 kernel_cap_t old_cap;
497 int res;
498
499 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
500 return -EINVAL;
501
502 old_fsuid = current->fsuid;
503 old_fsgid = current->fsgid;
504 old_cap = current->cap_effective;
505
506 current->fsuid = current->uid;
507 current->fsgid = current->gid;
508
509 /*
510 * Clear the capabilities if we switch to a non-root user
511 *
512 * FIXME: There is a race here against sys_capset. The
513 * capabilities can change yet we will restore the old
514 * value below. We should hold task_capabilities_lock,
515 * but we cannot because user_path_walk can sleep.
516 */
517 if (current->uid)
518 cap_clear(current->cap_effective);
519 else
520 current->cap_effective = current->cap_permitted;
521
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800522 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 if (!res) {
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800524 res = vfs_permission(&nd, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 /* SuS v2 requires we report a read only fs too */
526 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
527 && !special_file(nd.dentry->d_inode->i_mode))
528 res = -EROFS;
529 path_release(&nd);
530 }
531
532 current->fsuid = old_fsuid;
533 current->fsgid = old_fsgid;
534 current->cap_effective = old_cap;
535
536 return res;
537}
538
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800539asmlinkage long sys_access(const char __user *filename, int mode)
540{
541 return sys_faccessat(AT_FDCWD, filename, mode);
542}
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544asmlinkage long sys_chdir(const char __user * filename)
545{
546 struct nameidata nd;
547 int error;
548
Miklos Szeredi650a8982006-09-29 01:59:35 -0700549 error = __user_walk(filename,
550 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 if (error)
552 goto out;
553
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800554 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 if (error)
556 goto dput_and_out;
557
558 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
559
560dput_and_out:
561 path_release(&nd);
562out:
563 return error;
564}
565
566asmlinkage long sys_fchdir(unsigned int fd)
567{
568 struct file *file;
569 struct dentry *dentry;
570 struct inode *inode;
571 struct vfsmount *mnt;
572 int error;
573
574 error = -EBADF;
575 file = fget(fd);
576 if (!file)
577 goto out;
578
579 dentry = file->f_dentry;
580 mnt = file->f_vfsmnt;
581 inode = dentry->d_inode;
582
583 error = -ENOTDIR;
584 if (!S_ISDIR(inode->i_mode))
585 goto out_putf;
586
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800587 error = file_permission(file, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 if (!error)
589 set_fs_pwd(current->fs, mnt, dentry);
590out_putf:
591 fput(file);
592out:
593 return error;
594}
595
596asmlinkage long sys_chroot(const char __user * filename)
597{
598 struct nameidata nd;
599 int error;
600
601 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
602 if (error)
603 goto out;
604
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800605 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 if (error)
607 goto dput_and_out;
608
609 error = -EPERM;
610 if (!capable(CAP_SYS_CHROOT))
611 goto dput_and_out;
612
613 set_fs_root(current->fs, nd.mnt, nd.dentry);
614 set_fs_altroot();
615 error = 0;
616dput_and_out:
617 path_release(&nd);
618out:
619 return error;
620}
621
622asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
623{
624 struct inode * inode;
625 struct dentry * dentry;
626 struct file * file;
627 int err = -EBADF;
628 struct iattr newattrs;
629
630 file = fget(fd);
631 if (!file)
632 goto out;
633
634 dentry = file->f_dentry;
635 inode = dentry->d_inode;
636
Amy Griffis9c937dc2006-06-08 23:19:31 -0400637 audit_inode(NULL, inode);
Amy Griffis73241cc2005-11-03 16:00:25 +0000638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 err = -EROFS;
640 if (IS_RDONLY(inode))
641 goto out_putf;
642 err = -EPERM;
643 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
644 goto out_putf;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800645 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 if (mode == (mode_t) -1)
647 mode = inode->i_mode;
648 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
649 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
650 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800651 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653out_putf:
654 fput(file);
655out:
656 return err;
657}
658
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800659asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
660 mode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661{
662 struct nameidata nd;
663 struct inode * inode;
664 int error;
665 struct iattr newattrs;
666
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800667 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 if (error)
669 goto out;
670 inode = nd.dentry->d_inode;
671
672 error = -EROFS;
673 if (IS_RDONLY(inode))
674 goto dput_and_out;
675
676 error = -EPERM;
677 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
678 goto dput_and_out;
679
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800680 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 if (mode == (mode_t) -1)
682 mode = inode->i_mode;
683 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
684 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
685 error = notify_change(nd.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800686 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687
688dput_and_out:
689 path_release(&nd);
690out:
691 return error;
692}
693
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800694asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
695{
696 return sys_fchmodat(AT_FDCWD, filename, mode);
697}
698
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
700{
701 struct inode * inode;
702 int error;
703 struct iattr newattrs;
704
705 error = -ENOENT;
706 if (!(inode = dentry->d_inode)) {
707 printk(KERN_ERR "chown_common: NULL inode\n");
708 goto out;
709 }
710 error = -EROFS;
711 if (IS_RDONLY(inode))
712 goto out;
713 error = -EPERM;
714 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
715 goto out;
716 newattrs.ia_valid = ATTR_CTIME;
717 if (user != (uid_t) -1) {
718 newattrs.ia_valid |= ATTR_UID;
719 newattrs.ia_uid = user;
720 }
721 if (group != (gid_t) -1) {
722 newattrs.ia_valid |= ATTR_GID;
723 newattrs.ia_gid = group;
724 }
725 if (!S_ISDIR(inode->i_mode))
726 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800727 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 error = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800729 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730out:
731 return error;
732}
733
734asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
735{
736 struct nameidata nd;
737 int error;
738
739 error = user_path_walk(filename, &nd);
740 if (!error) {
741 error = chown_common(nd.dentry, user, group);
742 path_release(&nd);
743 }
744 return error;
745}
746
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800747asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
748 gid_t group, int flag)
749{
750 struct nameidata nd;
751 int error = -EINVAL;
752 int follow;
753
754 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
755 goto out;
756
757 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
758 error = __user_walk_fd(dfd, filename, follow, &nd);
759 if (!error) {
760 error = chown_common(nd.dentry, user, group);
761 path_release(&nd);
762 }
763out:
764 return error;
765}
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
768{
769 struct nameidata nd;
770 int error;
771
772 error = user_path_walk_link(filename, &nd);
773 if (!error) {
774 error = chown_common(nd.dentry, user, group);
775 path_release(&nd);
776 }
777 return error;
778}
779
780
781asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
782{
783 struct file * file;
784 int error = -EBADF;
785
786 file = fget(fd);
787 if (file) {
Amy Griffis73241cc2005-11-03 16:00:25 +0000788 struct dentry * dentry;
789 dentry = file->f_dentry;
Amy Griffis9c937dc2006-06-08 23:19:31 -0400790 audit_inode(NULL, dentry->d_inode);
Amy Griffis73241cc2005-11-03 16:00:25 +0000791 error = chown_common(dentry, user, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 fput(file);
793 }
794 return error;
795}
796
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700797static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Trond Myklebust834f2a42005-10-18 14:20:16 -0700798 int flags, struct file *f,
799 int (*open)(struct inode *, struct file *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 struct inode *inode;
802 int error;
803
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 f->f_flags = flags;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700805 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
806 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 inode = dentry->d_inode;
808 if (f->f_mode & FMODE_WRITE) {
809 error = get_write_access(inode);
810 if (error)
811 goto cleanup_file;
812 }
813
814 f->f_mapping = inode->i_mapping;
815 f->f_dentry = dentry;
816 f->f_vfsmnt = mnt;
817 f->f_pos = 0;
818 f->f_op = fops_get(inode->i_fop);
819 file_move(f, &inode->i_sb->s_files);
820
Trond Myklebust834f2a42005-10-18 14:20:16 -0700821 if (!open && f->f_op)
822 open = f->f_op->open;
823 if (open) {
824 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 if (error)
826 goto cleanup_all;
827 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700828
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
830
831 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
832
833 /* NB: we're sure to have correct a_ops only after f_op->open */
834 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700835 if (!f->f_mapping->a_ops ||
836 ((!f->f_mapping->a_ops->direct_IO) &&
837 (!f->f_mapping->a_ops->get_xip_page))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 fput(f);
839 f = ERR_PTR(-EINVAL);
840 }
841 }
842
843 return f;
844
845cleanup_all:
846 fops_put(f->f_op);
847 if (f->f_mode & FMODE_WRITE)
848 put_write_access(inode);
849 file_kill(f);
850 f->f_dentry = NULL;
851 f->f_vfsmnt = NULL;
852cleanup_file:
853 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 dput(dentry);
855 mntput(mnt);
856 return ERR_PTR(error);
857}
858
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700859/*
860 * Note that while the flag value (low two bits) for sys_open means:
861 * 00 - read-only
862 * 01 - write-only
863 * 10 - read-write
864 * 11 - special
865 * it is changed into
866 * 00 - no permissions needed
867 * 01 - read-permission
868 * 10 - write-permission
869 * 11 - read-write
870 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
871 * used by symlinks.
872 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800873static struct file *do_filp_open(int dfd, const char *filename, int flags,
874 int mode)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700875{
876 int namei_flags, error;
877 struct nameidata nd;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700878
879 namei_flags = flags;
880 if ((namei_flags+1) & O_ACCMODE)
881 namei_flags++;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700882
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800883 error = open_namei(dfd, filename, namei_flags, mode, &nd);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700884 if (!error)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700885 return nameidata_to_filp(&nd, flags);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700886
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700887 return ERR_PTR(error);
888}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800889
890struct file *filp_open(const char *filename, int flags, int mode)
891{
892 return do_filp_open(AT_FDCWD, filename, flags, mode);
893}
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700894EXPORT_SYMBOL(filp_open);
895
Trond Myklebust834f2a42005-10-18 14:20:16 -0700896/**
897 * lookup_instantiate_filp - instantiates the open intent filp
898 * @nd: pointer to nameidata
899 * @dentry: pointer to dentry
900 * @open: open callback
901 *
902 * Helper for filesystems that want to use lookup open intents and pass back
903 * a fully instantiated struct file to the caller.
904 * This function is meant to be called from within a filesystem's
905 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800906 * Beware of calling it for non-regular files! Those ->open methods might block
907 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
908 * leading to a deadlock, as nobody can open that fifo anymore, because
909 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700910 * Note that in case of error, nd->intent.open.file is destroyed, but the
911 * path information remains valid.
912 * If the open callback is set to NULL, then the standard f_op->open()
913 * filesystem callback is substituted.
914 */
915struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
916 int (*open)(struct inode *, struct file *))
917{
918 if (IS_ERR(nd->intent.open.file))
919 goto out;
920 if (IS_ERR(dentry))
921 goto out_err;
922 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
923 nd->intent.open.flags - 1,
924 nd->intent.open.file,
925 open);
926out:
927 return nd->intent.open.file;
928out_err:
929 release_open_intent(nd);
930 nd->intent.open.file = (struct file *)dentry;
931 goto out;
932}
933EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
934
935/**
936 * nameidata_to_filp - convert a nameidata to an open filp.
937 * @nd: pointer to nameidata
938 * @flags: open flags
939 *
940 * Note that this function destroys the original nameidata
941 */
942struct file *nameidata_to_filp(struct nameidata *nd, int flags)
943{
944 struct file *filp;
945
946 /* Pick up the filp from the open intent */
947 filp = nd->intent.open.file;
948 /* Has the filesystem initialised the file for us? */
949 if (filp->f_dentry == NULL)
950 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
951 else
952 path_release(nd);
953 return filp;
954}
955
Peter Staubach6fdcc212005-11-07 00:59:42 -0800956/*
957 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
958 * error.
959 */
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700960struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
961{
962 int error;
963 struct file *f;
964
965 error = -ENFILE;
966 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800967 if (f == NULL) {
968 dput(dentry);
969 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700970 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800971 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700972
Trond Myklebust834f2a42005-10-18 14:20:16 -0700973 return __dentry_open(dentry, mnt, flags, f, NULL);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700974}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975EXPORT_SYMBOL(dentry_open);
976
977/*
978 * Find an empty file descriptor entry, and mark it busy.
979 */
980int get_unused_fd(void)
981{
982 struct files_struct * files = current->files;
983 int fd, error;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700984 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
986 error = -EMFILE;
987 spin_lock(&files->file_lock);
988
989repeat:
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700990 fdt = files_fdtable(files);
991 fd = find_next_zero_bit(fdt->open_fds->fds_bits,
992 fdt->max_fdset,
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800993 files->next_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
995 /*
996 * N.B. For clone tasks sharing a files structure, this test
997 * will limit the total number of files that can be opened.
998 */
999 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
1000 goto out;
1001
1002 /* Do we need to expand the fd array or fd set? */
1003 error = expand_files(files, fd);
1004 if (error < 0)
1005 goto out;
1006
1007 if (error) {
1008 /*
1009 * If we needed to expand the fs array we
1010 * might have blocked - try again.
1011 */
1012 error = -EMFILE;
1013 goto repeat;
1014 }
1015
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001016 FD_SET(fd, fdt->open_fds);
1017 FD_CLR(fd, fdt->close_on_exec);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001018 files->next_fd = fd + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019#if 1
1020 /* Sanity check */
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001021 if (fdt->fd[fd] != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001023 fdt->fd[fd] = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 }
1025#endif
1026 error = fd;
1027
1028out:
1029 spin_unlock(&files->file_lock);
1030 return error;
1031}
1032
1033EXPORT_SYMBOL(get_unused_fd);
1034
Matt Mackallb01ec0e2006-01-08 01:05:20 -08001035static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036{
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001037 struct fdtable *fdt = files_fdtable(files);
1038 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001039 if (fd < files->next_fd)
1040 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041}
1042
1043void fastcall put_unused_fd(unsigned int fd)
1044{
1045 struct files_struct *files = current->files;
1046 spin_lock(&files->file_lock);
1047 __put_unused_fd(files, fd);
1048 spin_unlock(&files->file_lock);
1049}
1050
1051EXPORT_SYMBOL(put_unused_fd);
1052
1053/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001054 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 *
1056 * The VFS is full of places where we drop the files lock between
1057 * setting the open_fds bitmap and installing the file in the file
1058 * array. At any such point, we are vulnerable to a dup2() race
1059 * installing a file in the array before us. We need to detect this and
1060 * fput() the struct file we are about to overwrite in this case.
1061 *
1062 * It should never happen - if we allow dup2() do it, _really_ bad things
1063 * will follow.
1064 */
1065
1066void fastcall fd_install(unsigned int fd, struct file * file)
1067{
1068 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001069 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001071 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001072 BUG_ON(fdt->fd[fd] != NULL);
1073 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 spin_unlock(&files->file_lock);
1075}
1076
1077EXPORT_SYMBOL(fd_install);
1078
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001079long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001081 char *tmp = getname(filename);
1082 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 if (!IS_ERR(tmp)) {
1085 fd = get_unused_fd();
1086 if (fd >= 0) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001087 struct file *f = do_filp_open(dfd, tmp, flags, mode);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001088 if (IS_ERR(f)) {
1089 put_unused_fd(fd);
1090 fd = PTR_ERR(f);
1091 } else {
Robert Love0eeca282005-07-12 17:06:03 -04001092 fsnotify_open(f->f_dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001093 fd_install(fd, f);
1094 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 putname(tmp);
1097 }
1098 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001100
1101asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1102{
Linus Torvalds385910f2006-04-18 13:22:59 -07001103 long ret;
1104
Miklos Szeredie922efc2005-09-06 15:18:25 -07001105 if (force_o_largefile())
1106 flags |= O_LARGEFILE;
1107
Linus Torvalds385910f2006-04-18 13:22:59 -07001108 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1109 /* avoid REGPARM breakage on x86: */
1110 prevent_tail_call(ret);
1111 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001112}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113EXPORT_SYMBOL_GPL(sys_open);
1114
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001115asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1116 int mode)
1117{
Linus Torvalds385910f2006-04-18 13:22:59 -07001118 long ret;
1119
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001120 if (force_o_largefile())
1121 flags |= O_LARGEFILE;
1122
Linus Torvalds385910f2006-04-18 13:22:59 -07001123 ret = do_sys_open(dfd, filename, flags, mode);
1124 /* avoid REGPARM breakage on x86: */
1125 prevent_tail_call(ret);
1126 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001127}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129#ifndef __alpha__
1130
1131/*
1132 * For backward compatibility? Maybe this should be moved
1133 * into arch/i386 instead?
1134 */
1135asmlinkage long sys_creat(const char __user * pathname, int mode)
1136{
1137 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1138}
1139
1140#endif
1141
1142/*
1143 * "id" is the POSIX thread ID. We use the
1144 * files pointer for this..
1145 */
1146int filp_close(struct file *filp, fl_owner_t id)
1147{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001148 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
1150 if (!file_count(filp)) {
1151 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001152 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 }
1154
Christoph Lameter45778ca2005-06-23 00:10:17 -07001155 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001156 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157
1158 dnotify_flush(filp, id);
1159 locks_remove_posix(filp, id);
1160 fput(filp);
1161 return retval;
1162}
1163
1164EXPORT_SYMBOL(filp_close);
1165
1166/*
1167 * Careful here! We test whether the file pointer is NULL before
1168 * releasing the fd. This ensures that one clone task can't release
1169 * an fd while another clone is opening it.
1170 */
1171asmlinkage long sys_close(unsigned int fd)
1172{
1173 struct file * filp;
1174 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001175 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001176 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
1178 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001179 fdt = files_fdtable(files);
1180 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001182 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 if (!filp)
1184 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001185 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001186 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 __put_unused_fd(files, fd);
1188 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001189 retval = filp_close(filp, files);
1190
1191 /* can't restart close syscall because file table entry was cleared */
1192 if (unlikely(retval == -ERESTARTSYS ||
1193 retval == -ERESTARTNOINTR ||
1194 retval == -ERESTARTNOHAND ||
1195 retval == -ERESTART_RESTARTBLOCK))
1196 retval = -EINTR;
1197
1198 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
1200out_unlock:
1201 spin_unlock(&files->file_lock);
1202 return -EBADF;
1203}
1204
1205EXPORT_SYMBOL(sys_close);
1206
1207/*
1208 * This routine simulates a hangup on the tty, to arrange that users
1209 * are given clean terminals at login time.
1210 */
1211asmlinkage long sys_vhangup(void)
1212{
1213 if (capable(CAP_SYS_TTY_CONFIG)) {
1214 tty_vhangup(current->signal->tty);
1215 return 0;
1216 }
1217 return -EPERM;
1218}
1219
1220/*
1221 * Called when an inode is about to be open.
1222 * We use this to disallow opening large files on 32bit systems if
1223 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1224 * on this flag in sys_open.
1225 */
1226int generic_file_open(struct inode * inode, struct file * filp)
1227{
1228 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1229 return -EFBIG;
1230 return 0;
1231}
1232
1233EXPORT_SYMBOL(generic_file_open);
1234
1235/*
1236 * This is used by subsystems that don't want seekable
1237 * file descriptors
1238 */
1239int nonseekable_open(struct inode *inode, struct file *filp)
1240{
1241 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1242 return 0;
1243}
1244
1245EXPORT_SYMBOL(nonseekable_open);