blob: 279aacf256001b356e42da929dc8149c054d5029 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/quotaops.h>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
13#include <linux/slab.h>
14#include <linux/tty.h>
15#include <linux/namei.h>
16#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080017#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
20#include <linux/vfs.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080021#include <linux/fcntl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
David Howells726c3342006-06-23 02:02:58 -070031int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070032{
33 int retval = -ENODEV;
34
David Howells726c3342006-06-23 02:02:58 -070035 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 retval = -ENOSYS;
David Howells726c3342006-06-23 02:02:58 -070037 if (dentry->d_sb->s_op->statfs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070038 memset(buf, 0, sizeof(*buf));
David Howells726c3342006-06-23 02:02:58 -070039 retval = security_sb_statfs(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 if (retval)
41 return retval;
David Howells726c3342006-06-23 02:02:58 -070042 retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 if (retval == 0 && buf->f_frsize == 0)
44 buf->f_frsize = buf->f_bsize;
45 }
46 }
47 return retval;
48}
49
50EXPORT_SYMBOL(vfs_statfs);
51
David Howells726c3342006-06-23 02:02:58 -070052static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070053{
54 struct kstatfs st;
55 int retval;
56
David Howells726c3342006-06-23 02:02:58 -070057 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 if (retval)
59 return retval;
60
61 if (sizeof(*buf) == sizeof(st))
62 memcpy(buf, &st, sizeof(st));
63 else {
64 if (sizeof buf->f_blocks == 4) {
65 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
66 0xffffffff00000000ULL)
67 return -EOVERFLOW;
68 /*
69 * f_files and f_ffree may be -1; it's okay to stuff
70 * that into 32 bits
71 */
72 if (st.f_files != -1 &&
73 (st.f_files & 0xffffffff00000000ULL))
74 return -EOVERFLOW;
75 if (st.f_ffree != -1 &&
76 (st.f_ffree & 0xffffffff00000000ULL))
77 return -EOVERFLOW;
78 }
79
80 buf->f_type = st.f_type;
81 buf->f_bsize = st.f_bsize;
82 buf->f_blocks = st.f_blocks;
83 buf->f_bfree = st.f_bfree;
84 buf->f_bavail = st.f_bavail;
85 buf->f_files = st.f_files;
86 buf->f_ffree = st.f_ffree;
87 buf->f_fsid = st.f_fsid;
88 buf->f_namelen = st.f_namelen;
89 buf->f_frsize = st.f_frsize;
90 memset(buf->f_spare, 0, sizeof(buf->f_spare));
91 }
92 return 0;
93}
94
David Howells726c3342006-06-23 02:02:58 -070095static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070096{
97 struct kstatfs st;
98 int retval;
99
David Howells726c3342006-06-23 02:02:58 -0700100 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 if (retval)
102 return retval;
103
104 if (sizeof(*buf) == sizeof(st))
105 memcpy(buf, &st, sizeof(st));
106 else {
107 buf->f_type = st.f_type;
108 buf->f_bsize = st.f_bsize;
109 buf->f_blocks = st.f_blocks;
110 buf->f_bfree = st.f_bfree;
111 buf->f_bavail = st.f_bavail;
112 buf->f_files = st.f_files;
113 buf->f_ffree = st.f_ffree;
114 buf->f_fsid = st.f_fsid;
115 buf->f_namelen = st.f_namelen;
116 buf->f_frsize = st.f_frsize;
117 memset(buf->f_spare, 0, sizeof(buf->f_spare));
118 }
119 return 0;
120}
121
122asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
123{
124 struct nameidata nd;
125 int error;
126
127 error = user_path_walk(path, &nd);
128 if (!error) {
129 struct statfs tmp;
Jan Blunck4ac91372008-02-14 19:34:32 -0800130 error = vfs_statfs_native(nd.path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
132 error = -EFAULT;
133 path_release(&nd);
134 }
135 return error;
136}
137
138
139asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
140{
141 struct nameidata nd;
142 long error;
143
144 if (sz != sizeof(*buf))
145 return -EINVAL;
146 error = user_path_walk(path, &nd);
147 if (!error) {
148 struct statfs64 tmp;
Jan Blunck4ac91372008-02-14 19:34:32 -0800149 error = vfs_statfs64(nd.path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
151 error = -EFAULT;
152 path_release(&nd);
153 }
154 return error;
155}
156
157
158asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
159{
160 struct file * file;
161 struct statfs tmp;
162 int error;
163
164 error = -EBADF;
165 file = fget(fd);
166 if (!file)
167 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800168 error = vfs_statfs_native(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
170 error = -EFAULT;
171 fput(file);
172out:
173 return error;
174}
175
176asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
177{
178 struct file * file;
179 struct statfs64 tmp;
180 int error;
181
182 if (sz != sizeof(*buf))
183 return -EINVAL;
184
185 error = -EBADF;
186 file = fget(fd);
187 if (!file)
188 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800189 error = vfs_statfs64(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
191 error = -EFAULT;
192 fput(file);
193out:
194 return error;
195}
196
NeilBrown4a301312006-01-08 01:02:39 -0800197int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
198 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199{
200 int err;
201 struct iattr newattrs;
202
203 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
204 if (length < 0)
205 return -EINVAL;
206
207 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -0800208 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -0800209 if (filp) {
210 newattrs.ia_file = filp;
211 newattrs.ia_valid |= ATTR_FILE;
212 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
Linus Torvalds7b82dc02007-05-08 20:10:00 -0700214 /* Remove suid/sgid on truncate too */
215 newattrs.ia_valid |= should_remove_suid(dentry);
216
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;
Jan Blunck4ac91372008-02-14 19:34:32 -0800236 inode = nd.path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237
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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 error = get_write_access(inode);
260 if (error)
261 goto dput_and_out;
262
david m. richter97003822007-07-31 00:39:12 -0700263 /*
264 * Make sure that there are no leases. get_write_access() protects
265 * against the truncate racing with a lease-granting setlease().
266 */
267 error = break_lease(inode, FMODE_WRITE);
268 if (error)
269 goto put_write_and_out;
270
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 error = locks_verify_truncate(inode, NULL, length);
272 if (!error) {
273 DQUOT_INIT(inode);
Jan Blunck4ac91372008-02-14 19:34:32 -0800274 error = do_truncate(nd.path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
david m. richter97003822007-07-31 00:39:12 -0700277put_write_and_out:
278 put_write_access(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279dput_and_out:
280 path_release(&nd);
281out:
282 return error;
283}
284
285asmlinkage long sys_truncate(const char __user * path, unsigned long length)
286{
287 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
288 return do_sys_truncate(path, (long)length);
289}
290
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800291static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
293 struct inode * inode;
294 struct dentry *dentry;
295 struct file * file;
296 int error;
297
298 error = -EINVAL;
299 if (length < 0)
300 goto out;
301 error = -EBADF;
302 file = fget(fd);
303 if (!file)
304 goto out;
305
306 /* explicitly opened as large or we are on 64-bit box */
307 if (file->f_flags & O_LARGEFILE)
308 small = 0;
309
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800310 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 inode = dentry->d_inode;
312 error = -EINVAL;
313 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
314 goto out_putf;
315
316 error = -EINVAL;
317 /* Cannot ftruncate over 2^31 bytes without large file support */
318 if (small && length > MAX_NON_LFS)
319 goto out_putf;
320
321 error = -EPERM;
322 if (IS_APPEND(inode))
323 goto out_putf;
324
325 error = locks_verify_truncate(inode, file, length);
326 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700327 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328out_putf:
329 fput(file);
330out:
331 return error;
332}
333
334asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
335{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700336 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700337 /* avoid REGPARM breakage on x86: */
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700338 prevent_tail_call(ret);
339 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340}
341
342/* LFS versions of truncate are only needed on 32 bit machines */
343#if BITS_PER_LONG == 32
344asmlinkage long sys_truncate64(const char __user * path, loff_t length)
345{
346 return do_sys_truncate(path, length);
347}
348
349asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
350{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700351 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700352 /* avoid REGPARM breakage on x86: */
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700353 prevent_tail_call(ret);
354 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355}
356#endif
357
Amit Arora97ac7352007-07-17 21:42:44 -0400358asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
359{
360 struct file *file;
361 struct inode *inode;
362 long ret = -EINVAL;
363
364 if (offset < 0 || len <= 0)
365 goto out;
366
367 /* Return error if mode is not supported */
368 ret = -EOPNOTSUPP;
369 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
370 goto out;
371
372 ret = -EBADF;
373 file = fget(fd);
374 if (!file)
375 goto out;
376 if (!(file->f_mode & FMODE_WRITE))
377 goto out_fput;
378 /*
379 * Revalidate the write permissions, in case security policy has
380 * changed since the files were opened.
381 */
382 ret = security_file_permission(file, MAY_WRITE);
383 if (ret)
384 goto out_fput;
385
386 inode = file->f_path.dentry->d_inode;
387
388 ret = -ESPIPE;
389 if (S_ISFIFO(inode->i_mode))
390 goto out_fput;
391
392 ret = -ENODEV;
393 /*
394 * Let individual file system decide if it supports preallocation
395 * for directories or not.
396 */
397 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
398 goto out_fput;
399
400 ret = -EFBIG;
401 /* Check for wrap through zero too */
402 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
403 goto out_fput;
404
405 if (inode->i_op && inode->i_op->fallocate)
406 ret = inode->i_op->fallocate(inode, mode, offset, len);
407 else
Ulrich Drepper0d786d42007-07-23 18:43:46 -0700408 ret = -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400409
410out_fput:
411 fput(file);
412out:
413 return ret;
414}
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416/*
417 * access() needs to use the real uid/gid, not the effective uid/gid.
418 * We do this by temporarily clearing all FS-related capabilities and
419 * switching the fsuid/fsgid around to the real ones.
420 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800421asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422{
423 struct nameidata nd;
424 int old_fsuid, old_fsgid;
425 kernel_cap_t old_cap;
426 int res;
427
428 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
429 return -EINVAL;
430
431 old_fsuid = current->fsuid;
432 old_fsgid = current->fsgid;
433 old_cap = current->cap_effective;
434
435 current->fsuid = current->uid;
436 current->fsgid = current->gid;
437
438 /*
439 * Clear the capabilities if we switch to a non-root user
440 *
441 * FIXME: There is a race here against sys_capset. The
442 * capabilities can change yet we will restore the old
443 * value below. We should hold task_capabilities_lock,
444 * but we cannot because user_path_walk can sleep.
445 */
446 if (current->uid)
447 cap_clear(current->cap_effective);
448 else
449 current->cap_effective = current->cap_permitted;
450
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800451 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700452 if (res)
453 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
Dave Hansen6902d922006-09-30 23:29:01 -0700455 res = vfs_permission(&nd, mode);
456 /* SuS v2 requires we report a read only fs too */
457 if(res || !(mode & S_IWOTH) ||
Jan Blunck4ac91372008-02-14 19:34:32 -0800458 special_file(nd.path.dentry->d_inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700459 goto out_path_release;
460
Jan Blunck4ac91372008-02-14 19:34:32 -0800461 if(IS_RDONLY(nd.path.dentry->d_inode))
Dave Hansen6902d922006-09-30 23:29:01 -0700462 res = -EROFS;
463
464out_path_release:
465 path_release(&nd);
466out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 current->fsuid = old_fsuid;
468 current->fsgid = old_fsgid;
469 current->cap_effective = old_cap;
470
471 return res;
472}
473
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800474asmlinkage long sys_access(const char __user *filename, int mode)
475{
476 return sys_faccessat(AT_FDCWD, filename, mode);
477}
478
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479asmlinkage long sys_chdir(const char __user * filename)
480{
481 struct nameidata nd;
482 int error;
483
Miklos Szeredi650a8982006-09-29 01:59:35 -0700484 error = __user_walk(filename,
485 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 if (error)
487 goto out;
488
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800489 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 if (error)
491 goto dput_and_out;
492
Jan Blunck4ac91372008-02-14 19:34:32 -0800493 set_fs_pwd(current->fs, nd.path.mnt, nd.path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494
495dput_and_out:
496 path_release(&nd);
497out:
498 return error;
499}
500
501asmlinkage long sys_fchdir(unsigned int fd)
502{
503 struct file *file;
504 struct dentry *dentry;
505 struct inode *inode;
506 struct vfsmount *mnt;
507 int error;
508
509 error = -EBADF;
510 file = fget(fd);
511 if (!file)
512 goto out;
513
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800514 dentry = file->f_path.dentry;
515 mnt = file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 inode = dentry->d_inode;
517
518 error = -ENOTDIR;
519 if (!S_ISDIR(inode->i_mode))
520 goto out_putf;
521
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800522 error = file_permission(file, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 if (!error)
524 set_fs_pwd(current->fs, mnt, dentry);
525out_putf:
526 fput(file);
527out:
528 return error;
529}
530
531asmlinkage long sys_chroot(const char __user * filename)
532{
533 struct nameidata nd;
534 int error;
535
536 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
537 if (error)
538 goto out;
539
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800540 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 if (error)
542 goto dput_and_out;
543
544 error = -EPERM;
545 if (!capable(CAP_SYS_CHROOT))
546 goto dput_and_out;
547
Jan Blunck4ac91372008-02-14 19:34:32 -0800548 set_fs_root(current->fs, nd.path.mnt, nd.path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 set_fs_altroot();
550 error = 0;
551dput_and_out:
552 path_release(&nd);
553out:
554 return error;
555}
556
557asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
558{
559 struct inode * inode;
560 struct dentry * dentry;
561 struct file * file;
562 int err = -EBADF;
563 struct iattr newattrs;
564
565 file = fget(fd);
566 if (!file)
567 goto out;
568
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800569 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 inode = dentry->d_inode;
571
Al Viro5a190ae2007-06-07 12:19:32 -0400572 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 err = -EROFS;
575 if (IS_RDONLY(inode))
576 goto out_putf;
577 err = -EPERM;
578 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
579 goto out_putf;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800580 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 if (mode == (mode_t) -1)
582 mode = inode->i_mode;
583 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
584 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
585 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800586 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
588out_putf:
589 fput(file);
590out:
591 return err;
592}
593
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800594asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
595 mode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596{
597 struct nameidata nd;
598 struct inode * inode;
599 int error;
600 struct iattr newattrs;
601
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800602 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 if (error)
604 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800605 inode = nd.path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
607 error = -EROFS;
608 if (IS_RDONLY(inode))
609 goto dput_and_out;
610
611 error = -EPERM;
612 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
613 goto dput_and_out;
614
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800615 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 if (mode == (mode_t) -1)
617 mode = inode->i_mode;
618 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
619 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Jan Blunck4ac91372008-02-14 19:34:32 -0800620 error = notify_change(nd.path.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800621 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623dput_and_out:
624 path_release(&nd);
625out:
626 return error;
627}
628
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800629asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
630{
631 return sys_fchmodat(AT_FDCWD, filename, mode);
632}
633
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
635{
636 struct inode * inode;
637 int error;
638 struct iattr newattrs;
639
640 error = -ENOENT;
641 if (!(inode = dentry->d_inode)) {
642 printk(KERN_ERR "chown_common: NULL inode\n");
643 goto out;
644 }
645 error = -EROFS;
646 if (IS_RDONLY(inode))
647 goto out;
648 error = -EPERM;
649 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
650 goto out;
651 newattrs.ia_valid = ATTR_CTIME;
652 if (user != (uid_t) -1) {
653 newattrs.ia_valid |= ATTR_UID;
654 newattrs.ia_uid = user;
655 }
656 if (group != (gid_t) -1) {
657 newattrs.ia_valid |= ATTR_GID;
658 newattrs.ia_gid = group;
659 }
660 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700661 newattrs.ia_valid |=
662 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800663 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 error = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800665 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666out:
667 return error;
668}
669
670asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
671{
672 struct nameidata nd;
673 int error;
674
675 error = user_path_walk(filename, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700676 if (error)
677 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800678 error = chown_common(nd.path.dentry, user, group);
Dave Hansen6902d922006-09-30 23:29:01 -0700679 path_release(&nd);
680out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 return error;
682}
683
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800684asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
685 gid_t group, int flag)
686{
687 struct nameidata nd;
688 int error = -EINVAL;
689 int follow;
690
691 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
692 goto out;
693
694 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
695 error = __user_walk_fd(dfd, filename, follow, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700696 if (error)
697 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800698 error = chown_common(nd.path.dentry, user, group);
Dave Hansen6902d922006-09-30 23:29:01 -0700699 path_release(&nd);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800700out:
701 return error;
702}
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
705{
706 struct nameidata nd;
707 int error;
708
709 error = user_path_walk_link(filename, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700710 if (error)
711 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800712 error = chown_common(nd.path.dentry, user, group);
Dave Hansen6902d922006-09-30 23:29:01 -0700713 path_release(&nd);
714out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 return error;
716}
717
718
719asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
720{
721 struct file * file;
722 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700723 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
725 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700726 if (!file)
727 goto out;
728
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800729 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400730 audit_inode(NULL, dentry);
Dave Hansen6902d922006-09-30 23:29:01 -0700731 error = chown_common(dentry, user, group);
732 fput(file);
733out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 return error;
735}
736
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700737static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Trond Myklebust834f2a42005-10-18 14:20:16 -0700738 int flags, struct file *f,
739 int (*open)(struct inode *, struct file *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 struct inode *inode;
742 int error;
743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 f->f_flags = flags;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700745 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
746 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 inode = dentry->d_inode;
748 if (f->f_mode & FMODE_WRITE) {
749 error = get_write_access(inode);
750 if (error)
751 goto cleanup_file;
752 }
753
754 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800755 f->f_path.dentry = dentry;
756 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 f->f_pos = 0;
758 f->f_op = fops_get(inode->i_fop);
759 file_move(f, &inode->i_sb->s_files);
760
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900761 error = security_dentry_open(f);
762 if (error)
763 goto cleanup_all;
764
Trond Myklebust834f2a42005-10-18 14:20:16 -0700765 if (!open && f->f_op)
766 open = f->f_op->open;
767 if (open) {
768 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 if (error)
770 goto cleanup_all;
771 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700772
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
774
775 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
776
777 /* NB: we're sure to have correct a_ops only after f_op->open */
778 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700779 if (!f->f_mapping->a_ops ||
780 ((!f->f_mapping->a_ops->direct_IO) &&
781 (!f->f_mapping->a_ops->get_xip_page))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 fput(f);
783 f = ERR_PTR(-EINVAL);
784 }
785 }
786
787 return f;
788
789cleanup_all:
790 fops_put(f->f_op);
791 if (f->f_mode & FMODE_WRITE)
792 put_write_access(inode);
793 file_kill(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800794 f->f_path.dentry = NULL;
795 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796cleanup_file:
797 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 dput(dentry);
799 mntput(mnt);
800 return ERR_PTR(error);
801}
802
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700803/*
804 * Note that while the flag value (low two bits) for sys_open means:
805 * 00 - read-only
806 * 01 - write-only
807 * 10 - read-write
808 * 11 - special
809 * it is changed into
810 * 00 - no permissions needed
811 * 01 - read-permission
812 * 10 - write-permission
813 * 11 - read-write
814 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
815 * used by symlinks.
816 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800817static struct file *do_filp_open(int dfd, const char *filename, int flags,
818 int mode)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700819{
820 int namei_flags, error;
821 struct nameidata nd;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700822
823 namei_flags = flags;
824 if ((namei_flags+1) & O_ACCMODE)
825 namei_flags++;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700826
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800827 error = open_namei(dfd, filename, namei_flags, mode, &nd);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700828 if (!error)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700829 return nameidata_to_filp(&nd, flags);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700830
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700831 return ERR_PTR(error);
832}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800833
834struct file *filp_open(const char *filename, int flags, int mode)
835{
836 return do_filp_open(AT_FDCWD, filename, flags, mode);
837}
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700838EXPORT_SYMBOL(filp_open);
839
Trond Myklebust834f2a42005-10-18 14:20:16 -0700840/**
841 * lookup_instantiate_filp - instantiates the open intent filp
842 * @nd: pointer to nameidata
843 * @dentry: pointer to dentry
844 * @open: open callback
845 *
846 * Helper for filesystems that want to use lookup open intents and pass back
847 * a fully instantiated struct file to the caller.
848 * This function is meant to be called from within a filesystem's
849 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800850 * Beware of calling it for non-regular files! Those ->open methods might block
851 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
852 * leading to a deadlock, as nobody can open that fifo anymore, because
853 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700854 * Note that in case of error, nd->intent.open.file is destroyed, but the
855 * path information remains valid.
856 * If the open callback is set to NULL, then the standard f_op->open()
857 * filesystem callback is substituted.
858 */
859struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
860 int (*open)(struct inode *, struct file *))
861{
862 if (IS_ERR(nd->intent.open.file))
863 goto out;
864 if (IS_ERR(dentry))
865 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800866 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700867 nd->intent.open.flags - 1,
868 nd->intent.open.file,
869 open);
870out:
871 return nd->intent.open.file;
872out_err:
873 release_open_intent(nd);
874 nd->intent.open.file = (struct file *)dentry;
875 goto out;
876}
877EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
878
879/**
880 * nameidata_to_filp - convert a nameidata to an open filp.
881 * @nd: pointer to nameidata
882 * @flags: open flags
883 *
884 * Note that this function destroys the original nameidata
885 */
886struct file *nameidata_to_filp(struct nameidata *nd, int flags)
887{
888 struct file *filp;
889
890 /* Pick up the filp from the open intent */
891 filp = nd->intent.open.file;
892 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800893 if (filp->f_path.dentry == NULL)
Jan Blunck4ac91372008-02-14 19:34:32 -0800894 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
895 NULL);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700896 else
897 path_release(nd);
898 return filp;
899}
900
Peter Staubach6fdcc212005-11-07 00:59:42 -0800901/*
902 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
903 * error.
904 */
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700905struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
906{
907 int error;
908 struct file *f;
909
910 error = -ENFILE;
911 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800912 if (f == NULL) {
913 dput(dentry);
914 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700915 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800916 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700917
Trond Myklebust834f2a42005-10-18 14:20:16 -0700918 return __dentry_open(dentry, mnt, flags, f, NULL);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700919}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920EXPORT_SYMBOL(dentry_open);
921
922/*
923 * Find an empty file descriptor entry, and mark it busy.
924 */
Ulrich Drepper4a195422007-07-15 23:40:34 -0700925int get_unused_fd_flags(int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926{
927 struct files_struct * files = current->files;
928 int fd, error;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700929 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
931 error = -EMFILE;
932 spin_lock(&files->file_lock);
933
934repeat:
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700935 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800936 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800937 files->next_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939 /*
940 * N.B. For clone tasks sharing a files structure, this test
941 * will limit the total number of files that can be opened.
942 */
943 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
944 goto out;
945
946 /* Do we need to expand the fd array or fd set? */
947 error = expand_files(files, fd);
948 if (error < 0)
949 goto out;
950
951 if (error) {
952 /*
953 * If we needed to expand the fs array we
954 * might have blocked - try again.
955 */
956 error = -EMFILE;
957 goto repeat;
958 }
959
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700960 FD_SET(fd, fdt->open_fds);
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700961 if (flags & O_CLOEXEC)
962 FD_SET(fd, fdt->close_on_exec);
963 else
964 FD_CLR(fd, fdt->close_on_exec);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800965 files->next_fd = fd + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966#if 1
967 /* Sanity check */
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700968 if (fdt->fd[fd] != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700970 fdt->fd[fd] = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 }
972#endif
973 error = fd;
974
975out:
976 spin_unlock(&files->file_lock);
977 return error;
978}
979
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700980int get_unused_fd(void)
981{
982 return get_unused_fd_flags(0);
983}
984
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985EXPORT_SYMBOL(get_unused_fd);
986
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800987static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700989 struct fdtable *fdt = files_fdtable(files);
990 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800991 if (fd < files->next_fd)
992 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993}
994
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800995void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996{
997 struct files_struct *files = current->files;
998 spin_lock(&files->file_lock);
999 __put_unused_fd(files, fd);
1000 spin_unlock(&files->file_lock);
1001}
1002
1003EXPORT_SYMBOL(put_unused_fd);
1004
1005/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001006 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 *
1008 * The VFS is full of places where we drop the files lock between
1009 * setting the open_fds bitmap and installing the file in the file
1010 * array. At any such point, we are vulnerable to a dup2() race
1011 * installing a file in the array before us. We need to detect this and
1012 * fput() the struct file we are about to overwrite in this case.
1013 *
1014 * It should never happen - if we allow dup2() do it, _really_ bad things
1015 * will follow.
1016 */
1017
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001018void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019{
1020 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001021 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001023 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001024 BUG_ON(fdt->fd[fd] != NULL);
1025 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 spin_unlock(&files->file_lock);
1027}
1028
1029EXPORT_SYMBOL(fd_install);
1030
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001031long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001033 char *tmp = getname(filename);
1034 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001037 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 if (fd >= 0) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001039 struct file *f = do_filp_open(dfd, tmp, flags, mode);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001040 if (IS_ERR(f)) {
1041 put_unused_fd(fd);
1042 fd = PTR_ERR(f);
1043 } else {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001044 fsnotify_open(f->f_path.dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001045 fd_install(fd, f);
1046 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 putname(tmp);
1049 }
1050 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001052
1053asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1054{
Linus Torvalds385910f2006-04-18 13:22:59 -07001055 long ret;
1056
Miklos Szeredie922efc2005-09-06 15:18:25 -07001057 if (force_o_largefile())
1058 flags |= O_LARGEFILE;
1059
Linus Torvalds385910f2006-04-18 13:22:59 -07001060 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1061 /* avoid REGPARM breakage on x86: */
1062 prevent_tail_call(ret);
1063 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001064}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001066asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1067 int mode)
1068{
Linus Torvalds385910f2006-04-18 13:22:59 -07001069 long ret;
1070
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001071 if (force_o_largefile())
1072 flags |= O_LARGEFILE;
1073
Linus Torvalds385910f2006-04-18 13:22:59 -07001074 ret = do_sys_open(dfd, filename, flags, mode);
1075 /* avoid REGPARM breakage on x86: */
1076 prevent_tail_call(ret);
1077 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001078}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001079
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080#ifndef __alpha__
1081
1082/*
1083 * For backward compatibility? Maybe this should be moved
1084 * into arch/i386 instead?
1085 */
1086asmlinkage long sys_creat(const char __user * pathname, int mode)
1087{
1088 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1089}
1090
1091#endif
1092
1093/*
1094 * "id" is the POSIX thread ID. We use the
1095 * files pointer for this..
1096 */
1097int filp_close(struct file *filp, fl_owner_t id)
1098{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001099 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100
1101 if (!file_count(filp)) {
1102 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001103 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 }
1105
Christoph Lameter45778ca2005-06-23 00:10:17 -07001106 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001107 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 dnotify_flush(filp, id);
1110 locks_remove_posix(filp, id);
1111 fput(filp);
1112 return retval;
1113}
1114
1115EXPORT_SYMBOL(filp_close);
1116
1117/*
1118 * Careful here! We test whether the file pointer is NULL before
1119 * releasing the fd. This ensures that one clone task can't release
1120 * an fd while another clone is opening it.
1121 */
1122asmlinkage long sys_close(unsigned int fd)
1123{
1124 struct file * filp;
1125 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001126 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001127 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
1129 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001130 fdt = files_fdtable(files);
1131 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001133 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 if (!filp)
1135 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001136 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001137 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 __put_unused_fd(files, fd);
1139 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001140 retval = filp_close(filp, files);
1141
1142 /* can't restart close syscall because file table entry was cleared */
1143 if (unlikely(retval == -ERESTARTSYS ||
1144 retval == -ERESTARTNOINTR ||
1145 retval == -ERESTARTNOHAND ||
1146 retval == -ERESTART_RESTARTBLOCK))
1147 retval = -EINTR;
1148
1149 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
1151out_unlock:
1152 spin_unlock(&files->file_lock);
1153 return -EBADF;
1154}
1155
1156EXPORT_SYMBOL(sys_close);
1157
1158/*
1159 * This routine simulates a hangup on the tty, to arrange that users
1160 * are given clean terminals at login time.
1161 */
1162asmlinkage long sys_vhangup(void)
1163{
1164 if (capable(CAP_SYS_TTY_CONFIG)) {
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001165 /* XXX: this needs locking */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 tty_vhangup(current->signal->tty);
1167 return 0;
1168 }
1169 return -EPERM;
1170}
1171
1172/*
1173 * Called when an inode is about to be open.
1174 * We use this to disallow opening large files on 32bit systems if
1175 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1176 * on this flag in sys_open.
1177 */
1178int generic_file_open(struct inode * inode, struct file * filp)
1179{
1180 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001181 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 return 0;
1183}
1184
1185EXPORT_SYMBOL(generic_file_open);
1186
1187/*
1188 * This is used by subsystems that don't want seekable
1189 * file descriptors
1190 */
1191int nonseekable_open(struct inode *inode, struct file *filp)
1192{
1193 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1194 return 0;
1195}
1196
1197EXPORT_SYMBOL(nonseekable_open);