blob: 52647be277a2c6bbe7e6ba5a9e0603d07ad33c71 [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>
Al Viro9f3acc32008-04-24 07:44:08 -040010#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/quotaops.h>
Robert Love0eeca282005-07-12 17:06:03 -040012#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/module.h>
14#include <linux/slab.h>
15#include <linux/tty.h>
16#include <linux/namei.h>
17#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080018#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070019#include <linux/securebits.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>
Amit Arora97ac7352007-07-17 21:42:44 -040031#include <linux/falloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
David Howells726c3342006-06-23 02:02:58 -070033int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070034{
35 int retval = -ENODEV;
36
David Howells726c3342006-06-23 02:02:58 -070037 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070038 retval = -ENOSYS;
David Howells726c3342006-06-23 02:02:58 -070039 if (dentry->d_sb->s_op->statfs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 memset(buf, 0, sizeof(*buf));
David Howells726c3342006-06-23 02:02:58 -070041 retval = security_sb_statfs(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 if (retval)
43 return retval;
David Howells726c3342006-06-23 02:02:58 -070044 retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 if (retval == 0 && buf->f_frsize == 0)
46 buf->f_frsize = buf->f_bsize;
47 }
48 }
49 return retval;
50}
51
52EXPORT_SYMBOL(vfs_statfs);
53
David Howells726c3342006-06-23 02:02:58 -070054static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070055{
56 struct kstatfs st;
57 int retval;
58
David Howells726c3342006-06-23 02:02:58 -070059 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 if (retval)
61 return retval;
62
63 if (sizeof(*buf) == sizeof(st))
64 memcpy(buf, &st, sizeof(st));
65 else {
66 if (sizeof buf->f_blocks == 4) {
Jon Tollefsonf4a67cc2008-07-23 21:27:55 -070067 if ((st.f_blocks | st.f_bfree | st.f_bavail |
68 st.f_bsize | st.f_frsize) &
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 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
Al Viro2d8f3032008-07-22 09:59:21 -0400125asmlinkage long sys_statfs(const char __user *pathname, struct statfs __user * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126{
Al Viro2d8f3032008-07-22 09:59:21 -0400127 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 int error;
129
Al Viro2d8f3032008-07-22 09:59:21 -0400130 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 if (!error) {
132 struct statfs tmp;
Al Viro2d8f3032008-07-22 09:59:21 -0400133 error = vfs_statfs_native(path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
135 error = -EFAULT;
Al Viro2d8f3032008-07-22 09:59:21 -0400136 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 }
138 return error;
139}
140
141
Al Viro2d8f3032008-07-22 09:59:21 -0400142asmlinkage long sys_statfs64(const char __user *pathname, size_t sz, struct statfs64 __user *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
Al Viro2d8f3032008-07-22 09:59:21 -0400144 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 long error;
146
147 if (sz != sizeof(*buf))
148 return -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -0400149 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 if (!error) {
151 struct statfs64 tmp;
Al Viro2d8f3032008-07-22 09:59:21 -0400152 error = vfs_statfs64(path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
154 error = -EFAULT;
Al Viro2d8f3032008-07-22 09:59:21 -0400155 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 }
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;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800171 error = vfs_statfs_native(file->f_path.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;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800192 error = vfs_statfs64(file->f_path.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
Linus Torvalds7b82dc02007-05-08 20:10:00 -0700217 /* Remove suid/sgid on truncate too */
218 newattrs.ia_valid |= should_remove_suid(dentry);
219
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800220 mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800222 mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 return err;
224}
225
Al Viro2d8f3032008-07-22 09:59:21 -0400226static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
Al Viro2d8f3032008-07-22 09:59:21 -0400228 struct path path;
229 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 int error;
231
232 error = -EINVAL;
233 if (length < 0) /* sorry, but loff_t says... */
234 goto out;
235
Al Viro2d8f3032008-07-22 09:59:21 -0400236 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 if (error)
238 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400239 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
242 error = -EISDIR;
243 if (S_ISDIR(inode->i_mode))
244 goto dput_and_out;
245
246 error = -EINVAL;
247 if (!S_ISREG(inode->i_mode))
248 goto dput_and_out;
249
Al Viro2d8f3032008-07-22 09:59:21 -0400250 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 if (error)
252 goto dput_and_out;
253
Al Viro256984a2008-07-22 08:09:30 -0400254 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800255 if (error)
256 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +0200259 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -0800260 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 error = get_write_access(inode);
263 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800264 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
david m. richter97003822007-07-31 00:39:12 -0700266 /*
267 * Make sure that there are no leases. get_write_access() protects
268 * against the truncate racing with a lease-granting setlease().
269 */
270 error = break_lease(inode, FMODE_WRITE);
271 if (error)
272 goto put_write_and_out;
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 error = locks_verify_truncate(inode, NULL, length);
275 if (!error) {
276 DQUOT_INIT(inode);
Al Viro2d8f3032008-07-22 09:59:21 -0400277 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
david m. richter97003822007-07-31 00:39:12 -0700280put_write_and_out:
281 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800282mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400283 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400285 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286out:
287 return error;
288}
289
290asmlinkage long sys_truncate(const char __user * path, unsigned long length)
291{
292 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
293 return do_sys_truncate(path, (long)length);
294}
295
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800296static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
298 struct inode * inode;
299 struct dentry *dentry;
300 struct file * file;
301 int error;
302
303 error = -EINVAL;
304 if (length < 0)
305 goto out;
306 error = -EBADF;
307 file = fget(fd);
308 if (!file)
309 goto out;
310
311 /* explicitly opened as large or we are on 64-bit box */
312 if (file->f_flags & O_LARGEFILE)
313 small = 0;
314
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800315 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 inode = dentry->d_inode;
317 error = -EINVAL;
318 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
319 goto out_putf;
320
321 error = -EINVAL;
322 /* Cannot ftruncate over 2^31 bytes without large file support */
323 if (small && length > MAX_NON_LFS)
324 goto out_putf;
325
326 error = -EPERM;
327 if (IS_APPEND(inode))
328 goto out_putf;
329
330 error = locks_verify_truncate(inode, file, length);
331 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700332 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333out_putf:
334 fput(file);
335out:
336 return error;
337}
338
339asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
340{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700341 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700342 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700343 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700344 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345}
346
347/* LFS versions of truncate are only needed on 32 bit machines */
348#if BITS_PER_LONG == 32
349asmlinkage long sys_truncate64(const char __user * path, loff_t length)
350{
351 return do_sys_truncate(path, length);
352}
353
354asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
355{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700356 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700357 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700358 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700359 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360}
361#endif
362
Amit Arora97ac7352007-07-17 21:42:44 -0400363asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
364{
365 struct file *file;
366 struct inode *inode;
367 long ret = -EINVAL;
368
369 if (offset < 0 || len <= 0)
370 goto out;
371
372 /* Return error if mode is not supported */
373 ret = -EOPNOTSUPP;
374 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
375 goto out;
376
377 ret = -EBADF;
378 file = fget(fd);
379 if (!file)
380 goto out;
381 if (!(file->f_mode & FMODE_WRITE))
382 goto out_fput;
383 /*
384 * Revalidate the write permissions, in case security policy has
385 * changed since the files were opened.
386 */
387 ret = security_file_permission(file, MAY_WRITE);
388 if (ret)
389 goto out_fput;
390
391 inode = file->f_path.dentry->d_inode;
392
393 ret = -ESPIPE;
394 if (S_ISFIFO(inode->i_mode))
395 goto out_fput;
396
397 ret = -ENODEV;
398 /*
399 * Let individual file system decide if it supports preallocation
400 * for directories or not.
401 */
402 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
403 goto out_fput;
404
405 ret = -EFBIG;
406 /* Check for wrap through zero too */
407 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
408 goto out_fput;
409
410 if (inode->i_op && inode->i_op->fallocate)
411 ret = inode->i_op->fallocate(inode, mode, offset, len);
412 else
Ulrich Drepper0d786d42007-07-23 18:43:46 -0700413 ret = -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400414
415out_fput:
416 fput(file);
417out:
418 return ret;
419}
420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421/*
422 * access() needs to use the real uid/gid, not the effective uid/gid.
423 * We do this by temporarily clearing all FS-related capabilities and
424 * switching the fsuid/fsgid around to the real ones.
425 */
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800426asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
Al Viro2d8f3032008-07-22 09:59:21 -0400428 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400429 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 int old_fsuid, old_fsgid;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700431 kernel_cap_t uninitialized_var(old_cap); /* !SECURE_NO_SETUID_FIXUP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 int res;
433
434 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
435 return -EINVAL;
436
437 old_fsuid = current->fsuid;
438 old_fsgid = current->fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440 current->fsuid = current->uid;
441 current->fsgid = current->gid;
442
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700443 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
444 /*
445 * Clear the capabilities if we switch to a non-root user
446 */
447#ifndef CONFIG_SECURITY_FILE_CAPABILITIES
448 /*
449 * FIXME: There is a race here against sys_capset. The
450 * capabilities can change yet we will restore the old
451 * value below. We should hold task_capabilities_lock,
Al Viro2d8f3032008-07-22 09:59:21 -0400452 * but we cannot because user_path_at can sleep.
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700453 */
454#endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
455 if (current->uid)
456 old_cap = cap_set_effective(__cap_empty_set);
457 else
458 old_cap = cap_set_effective(current->cap_permitted);
459 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
Al Viro2d8f3032008-07-22 09:59:21 -0400461 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700462 if (res)
463 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
Al Viro2d8f3032008-07-22 09:59:21 -0400465 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400466
467 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400468 /*
469 * MAY_EXEC on regular files is denied if the fs is mounted
470 * with the "noexec" flag.
471 */
472 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400473 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400474 goto out_path_release;
475 }
476
Al Viro256984a2008-07-22 08:09:30 -0400477 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700478 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400479 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700480 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800481 /*
482 * This is a rare case where using __mnt_is_readonly()
483 * is OK without a mnt_want/drop_write() pair. Since
484 * no actual write to the fs is performed here, we do
485 * not need to telegraph to that to anyone.
486 *
487 * By doing this, we accept that this access is
488 * inherently racy and know that the fs may change
489 * state before we even see this result.
490 */
Al Viro2d8f3032008-07-22 09:59:21 -0400491 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700492 res = -EROFS;
493
494out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400495 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700496out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 current->fsuid = old_fsuid;
498 current->fsgid = old_fsgid;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700499
500 if (!issecure(SECURE_NO_SETUID_FIXUP))
501 cap_set_effective(old_cap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
503 return res;
504}
505
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800506asmlinkage long sys_access(const char __user *filename, int mode)
507{
508 return sys_faccessat(AT_FDCWD, filename, mode);
509}
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511asmlinkage long sys_chdir(const char __user * filename)
512{
Al Viro2d8f3032008-07-22 09:59:21 -0400513 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 int error;
515
Al Viro2d8f3032008-07-22 09:59:21 -0400516 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 if (error)
518 goto out;
519
Al Viro2d8f3032008-07-22 09:59:21 -0400520 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 if (error)
522 goto dput_and_out;
523
Al Viro2d8f3032008-07-22 09:59:21 -0400524 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
526dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400527 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528out:
529 return error;
530}
531
532asmlinkage long sys_fchdir(unsigned int fd)
533{
534 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 int error;
537
538 error = -EBADF;
539 file = fget(fd);
540 if (!file)
541 goto out;
542
Jan Blunckac748a02008-02-14 19:34:39 -0800543 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
545 error = -ENOTDIR;
546 if (!S_ISDIR(inode->i_mode))
547 goto out_putf;
548
Al Viro256984a2008-07-22 08:09:30 -0400549 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800551 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552out_putf:
553 fput(file);
554out:
555 return error;
556}
557
558asmlinkage long sys_chroot(const char __user * filename)
559{
Al Viro2d8f3032008-07-22 09:59:21 -0400560 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 int error;
562
Al Viro2d8f3032008-07-22 09:59:21 -0400563 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 if (error)
565 goto out;
566
Al Viro2d8f3032008-07-22 09:59:21 -0400567 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 if (error)
569 goto dput_and_out;
570
571 error = -EPERM;
572 if (!capable(CAP_SYS_CHROOT))
573 goto dput_and_out;
574
Al Viro2d8f3032008-07-22 09:59:21 -0400575 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 error = 0;
577dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400578 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579out:
580 return error;
581}
582
583asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
584{
585 struct inode * inode;
586 struct dentry * dentry;
587 struct file * file;
588 int err = -EBADF;
589 struct iattr newattrs;
590
591 file = fget(fd);
592 if (!file)
593 goto out;
594
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800595 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 inode = dentry->d_inode;
597
Al Viro5a190ae2007-06-07 12:19:32 -0400598 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000599
Dave Hansen2af482a2008-02-15 14:37:50 -0800600 err = mnt_want_write(file->f_path.mnt);
601 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 goto out_putf;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800603 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 if (mode == (mode_t) -1)
605 mode = inode->i_mode;
606 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
607 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
608 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800609 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800610 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611out_putf:
612 fput(file);
613out:
614 return err;
615}
616
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800617asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
618 mode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619{
Al Viro2d8f3032008-07-22 09:59:21 -0400620 struct path path;
621 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 int error;
623 struct iattr newattrs;
624
Al Viro2d8f3032008-07-22 09:59:21 -0400625 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 if (error)
627 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400628 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Al Viro2d8f3032008-07-22 09:59:21 -0400630 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800631 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 goto dput_and_out;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800633 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 if (mode == (mode_t) -1)
635 mode = inode->i_mode;
636 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
637 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400638 error = notify_change(path.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800639 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400640 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400642 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643out:
644 return error;
645}
646
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800647asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
648{
649 return sys_fchmodat(AT_FDCWD, filename, mode);
650}
651
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
653{
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200654 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 int error;
656 struct iattr newattrs;
657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 newattrs.ia_valid = ATTR_CTIME;
659 if (user != (uid_t) -1) {
660 newattrs.ia_valid |= ATTR_UID;
661 newattrs.ia_uid = user;
662 }
663 if (group != (gid_t) -1) {
664 newattrs.ia_valid |= ATTR_GID;
665 newattrs.ia_gid = group;
666 }
667 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700668 newattrs.ia_valid |=
669 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800670 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 error = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800672 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200673
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 return error;
675}
676
677asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
678{
Al Viro2d8f3032008-07-22 09:59:21 -0400679 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 int error;
681
Al Viro2d8f3032008-07-22 09:59:21 -0400682 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700683 if (error)
684 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400685 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800686 if (error)
687 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400688 error = chown_common(path.dentry, user, group);
689 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800690out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400691 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700692out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 return error;
694}
695
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800696asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
697 gid_t group, int flag)
698{
Al Viro2d8f3032008-07-22 09:59:21 -0400699 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800700 int error = -EINVAL;
701 int follow;
702
703 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
704 goto out;
705
706 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
Al Viro2d8f3032008-07-22 09:59:21 -0400707 error = user_path_at(dfd, filename, follow, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700708 if (error)
709 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400710 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800711 if (error)
712 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400713 error = chown_common(path.dentry, user, group);
714 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800715out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400716 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800717out:
718 return error;
719}
720
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
722{
Al Viro2d8f3032008-07-22 09:59:21 -0400723 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 int error;
725
Al Viro2d8f3032008-07-22 09:59:21 -0400726 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700727 if (error)
728 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400729 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800730 if (error)
731 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400732 error = chown_common(path.dentry, user, group);
733 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800734out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400735 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700736out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 return error;
738}
739
740
741asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
742{
743 struct file * file;
744 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700745 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
747 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700748 if (!file)
749 goto out;
750
Dave Hansen2af482a2008-02-15 14:37:50 -0800751 error = mnt_want_write(file->f_path.mnt);
752 if (error)
753 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800754 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400755 audit_inode(NULL, dentry);
Dave Hansen6902d922006-09-30 23:29:01 -0700756 error = chown_common(dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800757 mnt_drop_write(file->f_path.mnt);
758out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700759 fput(file);
760out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 return error;
762}
763
Dave Hansen4a3fd212008-02-15 14:37:48 -0800764/*
765 * You have to be very careful that these write
766 * counts get cleaned up in error cases and
767 * upon __fput(). This should probably never
768 * be called outside of __dentry_open().
769 */
770static inline int __get_file_write_access(struct inode *inode,
771 struct vfsmount *mnt)
772{
773 int error;
774 error = get_write_access(inode);
775 if (error)
776 return error;
777 /*
778 * Do not take mount writer counts on
779 * special files since no writes to
780 * the mount itself will occur.
781 */
782 if (!special_file(inode->i_mode)) {
783 /*
784 * Balanced in __fput()
785 */
786 error = mnt_want_write(mnt);
787 if (error)
788 put_write_access(inode);
789 }
790 return error;
791}
792
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700793static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Trond Myklebust834f2a42005-10-18 14:20:16 -0700794 int flags, struct file *f,
795 int (*open)(struct inode *, struct file *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 struct inode *inode;
798 int error;
799
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 f->f_flags = flags;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700801 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
802 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 inode = dentry->d_inode;
804 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800805 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 if (error)
807 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800808 if (!special_file(inode->i_mode))
809 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 }
811
812 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800813 f->f_path.dentry = dentry;
814 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 f->f_pos = 0;
816 f->f_op = fops_get(inode->i_fop);
817 file_move(f, &inode->i_sb->s_files);
818
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900819 error = security_dentry_open(f);
820 if (error)
821 goto cleanup_all;
822
Trond Myklebust834f2a42005-10-18 14:20:16 -0700823 if (!open && f->f_op)
824 open = f->f_op->open;
825 if (open) {
826 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 if (error)
828 goto cleanup_all;
829 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700830
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
832
833 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
834
835 /* NB: we're sure to have correct a_ops only after f_op->open */
836 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700837 if (!f->f_mapping->a_ops ||
838 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700839 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 fput(f);
841 f = ERR_PTR(-EINVAL);
842 }
843 }
844
845 return f;
846
847cleanup_all:
848 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800849 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800851 if (!special_file(inode->i_mode)) {
852 /*
853 * We don't consider this a real
854 * mnt_want/drop_write() pair
855 * because it all happenend right
856 * here, so just reset the state.
857 */
858 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800859 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800860 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800861 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 file_kill(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800863 f->f_path.dentry = NULL;
864 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865cleanup_file:
866 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 dput(dentry);
868 mntput(mnt);
869 return ERR_PTR(error);
870}
871
Trond Myklebust834f2a42005-10-18 14:20:16 -0700872/**
873 * lookup_instantiate_filp - instantiates the open intent filp
874 * @nd: pointer to nameidata
875 * @dentry: pointer to dentry
876 * @open: open callback
877 *
878 * Helper for filesystems that want to use lookup open intents and pass back
879 * a fully instantiated struct file to the caller.
880 * This function is meant to be called from within a filesystem's
881 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800882 * Beware of calling it for non-regular files! Those ->open methods might block
883 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
884 * leading to a deadlock, as nobody can open that fifo anymore, because
885 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700886 * Note that in case of error, nd->intent.open.file is destroyed, but the
887 * path information remains valid.
888 * If the open callback is set to NULL, then the standard f_op->open()
889 * filesystem callback is substituted.
890 */
891struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
892 int (*open)(struct inode *, struct file *))
893{
894 if (IS_ERR(nd->intent.open.file))
895 goto out;
896 if (IS_ERR(dentry))
897 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800898 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700899 nd->intent.open.flags - 1,
900 nd->intent.open.file,
901 open);
902out:
903 return nd->intent.open.file;
904out_err:
905 release_open_intent(nd);
906 nd->intent.open.file = (struct file *)dentry;
907 goto out;
908}
909EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
910
911/**
912 * nameidata_to_filp - convert a nameidata to an open filp.
913 * @nd: pointer to nameidata
914 * @flags: open flags
915 *
916 * Note that this function destroys the original nameidata
917 */
918struct file *nameidata_to_filp(struct nameidata *nd, int flags)
919{
920 struct file *filp;
921
922 /* Pick up the filp from the open intent */
923 filp = nd->intent.open.file;
924 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800925 if (filp->f_path.dentry == NULL)
Jan Blunck4ac91372008-02-14 19:34:32 -0800926 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
927 NULL);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700928 else
Jan Blunck1d957f92008-02-14 19:34:35 -0800929 path_put(&nd->path);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700930 return filp;
931}
932
Peter Staubach6fdcc212005-11-07 00:59:42 -0800933/*
934 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
935 * error.
936 */
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700937struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
938{
939 int error;
940 struct file *f;
941
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800942 /*
943 * We must always pass in a valid mount pointer. Historically
944 * callers got away with not passing it, but we must enforce this at
945 * the earliest possible point now to avoid strange problems deep in the
946 * filesystem stack.
947 */
948 if (!mnt) {
949 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
950 dump_stack();
951 return ERR_PTR(-EINVAL);
952 }
953
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700954 error = -ENFILE;
955 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800956 if (f == NULL) {
957 dput(dentry);
958 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700959 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800960 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700961
Trond Myklebust834f2a42005-10-18 14:20:16 -0700962 return __dentry_open(dentry, mnt, flags, f, NULL);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700963}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964EXPORT_SYMBOL(dentry_open);
965
966/*
967 * Find an empty file descriptor entry, and mark it busy.
968 */
Ulrich Drepper4a195422007-07-15 23:40:34 -0700969int get_unused_fd_flags(int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
971 struct files_struct * files = current->files;
972 int fd, error;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700973 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 spin_lock(&files->file_lock);
976
977repeat:
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700978 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800979 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800980 files->next_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 /* Do we need to expand the fd array or fd set? */
983 error = expand_files(files, fd);
984 if (error < 0)
985 goto out;
986
987 if (error) {
988 /*
989 * If we needed to expand the fs array we
990 * might have blocked - try again.
991 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 goto repeat;
993 }
994
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700995 FD_SET(fd, fdt->open_fds);
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700996 if (flags & O_CLOEXEC)
997 FD_SET(fd, fdt->close_on_exec);
998 else
999 FD_CLR(fd, fdt->close_on_exec);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001000 files->next_fd = fd + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001#if 1
1002 /* Sanity check */
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001003 if (fdt->fd[fd] != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001005 fdt->fd[fd] = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 }
1007#endif
1008 error = fd;
1009
1010out:
1011 spin_unlock(&files->file_lock);
1012 return error;
1013}
1014
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001015int get_unused_fd(void)
1016{
1017 return get_unused_fd_flags(0);
1018}
1019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020EXPORT_SYMBOL(get_unused_fd);
1021
Matt Mackallb01ec0e2006-01-08 01:05:20 -08001022static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001024 struct fdtable *fdt = files_fdtable(files);
1025 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001026 if (fd < files->next_fd)
1027 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028}
1029
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001030void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031{
1032 struct files_struct *files = current->files;
1033 spin_lock(&files->file_lock);
1034 __put_unused_fd(files, fd);
1035 spin_unlock(&files->file_lock);
1036}
1037
1038EXPORT_SYMBOL(put_unused_fd);
1039
1040/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001041 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 *
1043 * The VFS is full of places where we drop the files lock between
1044 * setting the open_fds bitmap and installing the file in the file
1045 * array. At any such point, we are vulnerable to a dup2() race
1046 * installing a file in the array before us. We need to detect this and
1047 * fput() the struct file we are about to overwrite in this case.
1048 *
1049 * It should never happen - if we allow dup2() do it, _really_ bad things
1050 * will follow.
1051 */
1052
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001053void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054{
1055 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001056 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001058 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001059 BUG_ON(fdt->fd[fd] != NULL);
1060 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 spin_unlock(&files->file_lock);
1062}
1063
1064EXPORT_SYMBOL(fd_install);
1065
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001066long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001068 char *tmp = getname(filename);
1069 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001072 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 if (fd >= 0) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001074 struct file *f = do_filp_open(dfd, tmp, flags, mode);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001075 if (IS_ERR(f)) {
1076 put_unused_fd(fd);
1077 fd = PTR_ERR(f);
1078 } else {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001079 fsnotify_open(f->f_path.dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001080 fd_install(fd, f);
1081 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 putname(tmp);
1084 }
1085 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001087
1088asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1089{
Linus Torvalds385910f2006-04-18 13:22:59 -07001090 long ret;
1091
Miklos Szeredie922efc2005-09-06 15:18:25 -07001092 if (force_o_largefile())
1093 flags |= O_LARGEFILE;
1094
Linus Torvalds385910f2006-04-18 13:22:59 -07001095 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1096 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001097 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001098 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001099}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001101asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1102 int mode)
1103{
Linus Torvalds385910f2006-04-18 13:22:59 -07001104 long ret;
1105
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001106 if (force_o_largefile())
1107 flags |= O_LARGEFILE;
1108
Linus Torvalds385910f2006-04-18 13:22:59 -07001109 ret = do_sys_open(dfd, filename, flags, mode);
1110 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001111 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001112 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001113}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001114
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115#ifndef __alpha__
1116
1117/*
1118 * For backward compatibility? Maybe this should be moved
1119 * into arch/i386 instead?
1120 */
1121asmlinkage long sys_creat(const char __user * pathname, int mode)
1122{
1123 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1124}
1125
1126#endif
1127
1128/*
1129 * "id" is the POSIX thread ID. We use the
1130 * files pointer for this..
1131 */
1132int filp_close(struct file *filp, fl_owner_t id)
1133{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001134 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 if (!file_count(filp)) {
1137 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001138 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 }
1140
Christoph Lameter45778ca2005-06-23 00:10:17 -07001141 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001142 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
1144 dnotify_flush(filp, id);
1145 locks_remove_posix(filp, id);
1146 fput(filp);
1147 return retval;
1148}
1149
1150EXPORT_SYMBOL(filp_close);
1151
1152/*
1153 * Careful here! We test whether the file pointer is NULL before
1154 * releasing the fd. This ensures that one clone task can't release
1155 * an fd while another clone is opening it.
1156 */
1157asmlinkage long sys_close(unsigned int fd)
1158{
1159 struct file * filp;
1160 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001161 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001162 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001165 fdt = files_fdtable(files);
1166 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001168 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 if (!filp)
1170 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001171 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001172 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 __put_unused_fd(files, fd);
1174 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001175 retval = filp_close(filp, files);
1176
1177 /* can't restart close syscall because file table entry was cleared */
1178 if (unlikely(retval == -ERESTARTSYS ||
1179 retval == -ERESTARTNOINTR ||
1180 retval == -ERESTARTNOHAND ||
1181 retval == -ERESTART_RESTARTBLOCK))
1182 retval = -EINTR;
1183
1184 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
1186out_unlock:
1187 spin_unlock(&files->file_lock);
1188 return -EBADF;
1189}
1190
1191EXPORT_SYMBOL(sys_close);
1192
1193/*
1194 * This routine simulates a hangup on the tty, to arrange that users
1195 * are given clean terminals at login time.
1196 */
1197asmlinkage long sys_vhangup(void)
1198{
1199 if (capable(CAP_SYS_TTY_CONFIG)) {
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001200 /* XXX: this needs locking */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 tty_vhangup(current->signal->tty);
1202 return 0;
1203 }
1204 return -EPERM;
1205}
1206
1207/*
1208 * Called when an inode is about to be open.
1209 * We use this to disallow opening large files on 32bit systems if
1210 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1211 * on this flag in sys_open.
1212 */
1213int generic_file_open(struct inode * inode, struct file * filp)
1214{
1215 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001216 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 return 0;
1218}
1219
1220EXPORT_SYMBOL(generic_file_open);
1221
1222/*
1223 * This is used by subsystems that don't want seekable
1224 * file descriptors
1225 */
1226int nonseekable_open(struct inode *inode, struct file *filp)
1227{
1228 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1229 return 0;
1230}
1231
1232EXPORT_SYMBOL(nonseekable_open);