blob: 377eb25b6abfd84a11dadb716cb03cffe6b32f99 [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>
Al Viro5ad4e532009-03-29 19:50:06 -040032#include <linux/fs_struct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
David Howells726c3342006-06-23 02:02:58 -070034int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070035{
36 int retval = -ENODEV;
37
David Howells726c3342006-06-23 02:02:58 -070038 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 retval = -ENOSYS;
David Howells726c3342006-06-23 02:02:58 -070040 if (dentry->d_sb->s_op->statfs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 memset(buf, 0, sizeof(*buf));
David Howells726c3342006-06-23 02:02:58 -070042 retval = security_sb_statfs(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 if (retval)
44 return retval;
David Howells726c3342006-06-23 02:02:58 -070045 retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 if (retval == 0 && buf->f_frsize == 0)
47 buf->f_frsize = buf->f_bsize;
48 }
49 }
50 return retval;
51}
52
53EXPORT_SYMBOL(vfs_statfs);
54
David Howells726c3342006-06-23 02:02:58 -070055static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070056{
57 struct kstatfs st;
58 int retval;
59
David Howells726c3342006-06-23 02:02:58 -070060 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 if (retval)
62 return retval;
63
64 if (sizeof(*buf) == sizeof(st))
65 memcpy(buf, &st, sizeof(st));
66 else {
67 if (sizeof buf->f_blocks == 4) {
Jon Tollefsonf4a67cc2008-07-23 21:27:55 -070068 if ((st.f_blocks | st.f_bfree | st.f_bavail |
69 st.f_bsize | st.f_frsize) &
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 0xffffffff00000000ULL)
71 return -EOVERFLOW;
72 /*
73 * f_files and f_ffree may be -1; it's okay to stuff
74 * that into 32 bits
75 */
76 if (st.f_files != -1 &&
77 (st.f_files & 0xffffffff00000000ULL))
78 return -EOVERFLOW;
79 if (st.f_ffree != -1 &&
80 (st.f_ffree & 0xffffffff00000000ULL))
81 return -EOVERFLOW;
82 }
83
84 buf->f_type = st.f_type;
85 buf->f_bsize = st.f_bsize;
86 buf->f_blocks = st.f_blocks;
87 buf->f_bfree = st.f_bfree;
88 buf->f_bavail = st.f_bavail;
89 buf->f_files = st.f_files;
90 buf->f_ffree = st.f_ffree;
91 buf->f_fsid = st.f_fsid;
92 buf->f_namelen = st.f_namelen;
93 buf->f_frsize = st.f_frsize;
94 memset(buf->f_spare, 0, sizeof(buf->f_spare));
95 }
96 return 0;
97}
98
David Howells726c3342006-06-23 02:02:58 -070099static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100{
101 struct kstatfs st;
102 int retval;
103
David Howells726c3342006-06-23 02:02:58 -0700104 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 if (retval)
106 return retval;
107
108 if (sizeof(*buf) == sizeof(st))
109 memcpy(buf, &st, sizeof(st));
110 else {
111 buf->f_type = st.f_type;
112 buf->f_bsize = st.f_bsize;
113 buf->f_blocks = st.f_blocks;
114 buf->f_bfree = st.f_bfree;
115 buf->f_bavail = st.f_bavail;
116 buf->f_files = st.f_files;
117 buf->f_ffree = st.f_ffree;
118 buf->f_fsid = st.f_fsid;
119 buf->f_namelen = st.f_namelen;
120 buf->f_frsize = st.f_frsize;
121 memset(buf->f_spare, 0, sizeof(buf->f_spare));
122 }
123 return 0;
124}
125
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100126SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127{
Al Viro2d8f3032008-07-22 09:59:21 -0400128 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 int error;
130
Al Viro2d8f3032008-07-22 09:59:21 -0400131 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 if (!error) {
133 struct statfs tmp;
Al Viro2d8f3032008-07-22 09:59:21 -0400134 error = vfs_statfs_native(path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
136 error = -EFAULT;
Al Viro2d8f3032008-07-22 09:59:21 -0400137 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138 }
139 return error;
140}
141
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100142SYSCALL_DEFINE3(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
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100160SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
162 struct file * file;
163 struct statfs tmp;
164 int error;
165
166 error = -EBADF;
167 file = fget(fd);
168 if (!file)
169 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800170 error = vfs_statfs_native(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
172 error = -EFAULT;
173 fput(file);
174out:
175 return error;
176}
177
Heiko Carstens257ac262009-01-14 14:14:13 +0100178SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179{
180 struct file * file;
181 struct statfs64 tmp;
182 int error;
183
184 if (sz != sizeof(*buf))
185 return -EINVAL;
186
187 error = -EBADF;
188 file = fget(fd);
189 if (!file)
190 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800191 error = vfs_statfs64(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
193 error = -EFAULT;
194 fput(file);
195out:
196 return error;
197}
198
NeilBrown4a301312006-01-08 01:02:39 -0800199int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
200 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
202 int err;
203 struct iattr newattrs;
204
205 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
206 if (length < 0)
207 return -EINVAL;
208
209 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -0800210 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -0800211 if (filp) {
212 newattrs.ia_file = filp;
213 newattrs.ia_valid |= ATTR_FILE;
214 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Linus Torvalds7b82dc02007-05-08 20:10:00 -0700216 /* Remove suid/sgid on truncate too */
217 newattrs.ia_valid |= should_remove_suid(dentry);
218
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800219 mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800221 mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return err;
223}
224
Al Viro2d8f3032008-07-22 09:59:21 -0400225static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226{
Al Viro2d8f3032008-07-22 09:59:21 -0400227 struct path path;
228 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 int error;
230
231 error = -EINVAL;
232 if (length < 0) /* sorry, but loff_t says... */
233 goto out;
234
Al Viro2d8f3032008-07-22 09:59:21 -0400235 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 if (error)
237 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400238 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
241 error = -EISDIR;
242 if (S_ISDIR(inode->i_mode))
243 goto dput_and_out;
244
245 error = -EINVAL;
246 if (!S_ISREG(inode->i_mode))
247 goto dput_and_out;
248
Al Viro2d8f3032008-07-22 09:59:21 -0400249 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 if (error)
251 goto dput_and_out;
252
Al Viro256984a2008-07-22 08:09:30 -0400253 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800254 if (error)
255 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
257 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +0200258 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -0800259 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 error = get_write_access(inode);
262 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800263 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
david m. richter97003822007-07-31 00:39:12 -0700265 /*
266 * Make sure that there are no leases. get_write_access() protects
267 * against the truncate racing with a lease-granting setlease().
268 */
269 error = break_lease(inode, FMODE_WRITE);
270 if (error)
271 goto put_write_and_out;
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900274 if (!error)
275 error = security_path_truncate(&path, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 if (!error) {
Jan Kara9e3509e2009-01-26 16:45:12 +0100277 vfs_dq_init(inode);
Al Viro2d8f3032008-07-22 09:59:21 -0400278 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
david m. richter97003822007-07-31 00:39:12 -0700281put_write_and_out:
282 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800283mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400284 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400286 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287out:
288 return error;
289}
290
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100291SYSCALL_DEFINE2(truncate, const char __user *, path, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
293 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
294 return do_sys_truncate(path, (long)length);
295}
296
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800297static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298{
299 struct inode * inode;
300 struct dentry *dentry;
301 struct file * file;
302 int error;
303
304 error = -EINVAL;
305 if (length < 0)
306 goto out;
307 error = -EBADF;
308 file = fget(fd);
309 if (!file)
310 goto out;
311
312 /* explicitly opened as large or we are on 64-bit box */
313 if (file->f_flags & O_LARGEFILE)
314 small = 0;
315
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800316 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 inode = dentry->d_inode;
318 error = -EINVAL;
319 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
320 goto out_putf;
321
322 error = -EINVAL;
323 /* Cannot ftruncate over 2^31 bytes without large file support */
324 if (small && length > MAX_NON_LFS)
325 goto out_putf;
326
327 error = -EPERM;
328 if (IS_APPEND(inode))
329 goto out_putf;
330
331 error = locks_verify_truncate(inode, file, length);
332 if (!error)
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900333 error = security_path_truncate(&file->f_path, length,
334 ATTR_MTIME|ATTR_CTIME);
335 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700336 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337out_putf:
338 fput(file);
339out:
340 return error;
341}
342
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100343SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700345 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700346 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700347 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700348 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349}
350
351/* LFS versions of truncate are only needed on 32 bit machines */
352#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100353SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
355 return do_sys_truncate(path, length);
356}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100357#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
358asmlinkage long SyS_truncate64(long path, loff_t length)
359{
360 return SYSC_truncate64((const char __user *) path, length);
361}
362SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
363#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100365SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700367 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700368 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700369 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700370 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100372#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
373asmlinkage long SyS_ftruncate64(long fd, loff_t length)
374{
375 return SYSC_ftruncate64((unsigned int) fd, length);
376}
377SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100379#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100381SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400382{
383 struct file *file;
384 struct inode *inode;
385 long ret = -EINVAL;
386
387 if (offset < 0 || len <= 0)
388 goto out;
389
390 /* Return error if mode is not supported */
391 ret = -EOPNOTSUPP;
392 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
393 goto out;
394
395 ret = -EBADF;
396 file = fget(fd);
397 if (!file)
398 goto out;
399 if (!(file->f_mode & FMODE_WRITE))
400 goto out_fput;
401 /*
402 * Revalidate the write permissions, in case security policy has
403 * changed since the files were opened.
404 */
405 ret = security_file_permission(file, MAY_WRITE);
406 if (ret)
407 goto out_fput;
408
409 inode = file->f_path.dentry->d_inode;
410
411 ret = -ESPIPE;
412 if (S_ISFIFO(inode->i_mode))
413 goto out_fput;
414
415 ret = -ENODEV;
416 /*
417 * Let individual file system decide if it supports preallocation
418 * for directories or not.
419 */
420 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
421 goto out_fput;
422
423 ret = -EFBIG;
424 /* Check for wrap through zero too */
425 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
426 goto out_fput;
427
Al Viroacfa4382008-12-04 10:06:33 -0500428 if (inode->i_op->fallocate)
Amit Arora97ac7352007-07-17 21:42:44 -0400429 ret = inode->i_op->fallocate(inode, mode, offset, len);
430 else
Ulrich Drepper0d786d42007-07-23 18:43:46 -0700431 ret = -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400432
433out_fput:
434 fput(file);
435out:
436 return ret;
437}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100438#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
439asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
440{
441 return SYSC_fallocate((int)fd, (int)mode, offset, len);
442}
443SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
444#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400445
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446/*
447 * access() needs to use the real uid/gid, not the effective uid/gid.
448 * We do this by temporarily clearing all FS-related capabilities and
449 * switching the fsuid/fsgid around to the real ones.
450 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100451SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452{
David Howellsd84f4f92008-11-14 10:39:23 +1100453 const struct cred *old_cred;
454 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400455 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400456 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 int res;
458
459 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
460 return -EINVAL;
461
David Howellsd84f4f92008-11-14 10:39:23 +1100462 override_cred = prepare_creds();
463 if (!override_cred)
464 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
David Howellsd84f4f92008-11-14 10:39:23 +1100466 override_cred->fsuid = override_cred->uid;
467 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700469 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100470 /* Clear the capabilities if we switch to a non-root user */
David Howellsd84f4f92008-11-14 10:39:23 +1100471 if (override_cred->uid)
472 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700473 else
David Howellsd84f4f92008-11-14 10:39:23 +1100474 override_cred->cap_effective =
475 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700476 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
David Howellsd84f4f92008-11-14 10:39:23 +1100478 old_cred = override_creds(override_cred);
479
Al Viro2d8f3032008-07-22 09:59:21 -0400480 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700481 if (res)
482 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
Al Viro2d8f3032008-07-22 09:59:21 -0400484 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400485
486 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400487 /*
488 * MAY_EXEC on regular files is denied if the fs is mounted
489 * with the "noexec" flag.
490 */
491 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400492 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400493 goto out_path_release;
494 }
495
Al Viro256984a2008-07-22 08:09:30 -0400496 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700497 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400498 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700499 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800500 /*
501 * This is a rare case where using __mnt_is_readonly()
502 * is OK without a mnt_want/drop_write() pair. Since
503 * no actual write to the fs is performed here, we do
504 * not need to telegraph to that to anyone.
505 *
506 * By doing this, we accept that this access is
507 * inherently racy and know that the fs may change
508 * state before we even see this result.
509 */
Al Viro2d8f3032008-07-22 09:59:21 -0400510 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700511 res = -EROFS;
512
513out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400514 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700515out:
David Howellsd84f4f92008-11-14 10:39:23 +1100516 revert_creds(old_cred);
517 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 return res;
519}
520
Heiko Carstensca013e92009-01-14 14:14:19 +0100521SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800522{
523 return sys_faccessat(AT_FDCWD, filename, mode);
524}
525
Heiko Carstens3cdad422009-01-14 14:14:22 +0100526SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527{
Al Viro2d8f3032008-07-22 09:59:21 -0400528 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 int error;
530
Al Viro2d8f3032008-07-22 09:59:21 -0400531 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 if (error)
533 goto out;
534
Al Viro2d8f3032008-07-22 09:59:21 -0400535 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 if (error)
537 goto dput_and_out;
538
Al Viro2d8f3032008-07-22 09:59:21 -0400539 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
541dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400542 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543out:
544 return error;
545}
546
Heiko Carstens3cdad422009-01-14 14:14:22 +0100547SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548{
549 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 int error;
552
553 error = -EBADF;
554 file = fget(fd);
555 if (!file)
556 goto out;
557
Jan Blunckac748a02008-02-14 19:34:39 -0800558 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
560 error = -ENOTDIR;
561 if (!S_ISDIR(inode->i_mode))
562 goto out_putf;
563
Al Viro256984a2008-07-22 08:09:30 -0400564 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800566 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567out_putf:
568 fput(file);
569out:
570 return error;
571}
572
Heiko Carstens3480b252009-01-14 14:14:16 +0100573SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574{
Al Viro2d8f3032008-07-22 09:59:21 -0400575 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 int error;
577
Al Viro2d8f3032008-07-22 09:59:21 -0400578 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 if (error)
580 goto out;
581
Al Viro2d8f3032008-07-22 09:59:21 -0400582 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 if (error)
584 goto dput_and_out;
585
586 error = -EPERM;
587 if (!capable(CAP_SYS_CHROOT))
588 goto dput_and_out;
589
Al Viro2d8f3032008-07-22 09:59:21 -0400590 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 error = 0;
592dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400593 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594out:
595 return error;
596}
597
Heiko Carstensa26eab22009-01-14 14:14:17 +0100598SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
600 struct inode * inode;
601 struct dentry * dentry;
602 struct file * file;
603 int err = -EBADF;
604 struct iattr newattrs;
605
606 file = fget(fd);
607 if (!file)
608 goto out;
609
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800610 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 inode = dentry->d_inode;
612
Al Viro5a190ae2007-06-07 12:19:32 -0400613 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000614
Dave Hansen2af482a2008-02-15 14:37:50 -0800615 err = mnt_want_write(file->f_path.mnt);
616 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 goto out_putf;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800618 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 if (mode == (mode_t) -1)
620 mode = inode->i_mode;
621 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
622 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
623 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800624 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800625 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626out_putf:
627 fput(file);
628out:
629 return err;
630}
631
Heiko Carstens6559eed82009-01-14 14:14:32 +0100632SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
Al Viro2d8f3032008-07-22 09:59:21 -0400634 struct path path;
635 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 int error;
637 struct iattr newattrs;
638
Al Viro2d8f3032008-07-22 09:59:21 -0400639 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 if (error)
641 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400642 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
Al Viro2d8f3032008-07-22 09:59:21 -0400644 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800645 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 goto dput_and_out;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800647 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 if (mode == (mode_t) -1)
649 mode = inode->i_mode;
650 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
651 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400652 error = notify_change(path.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800653 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400654 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400656 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657out:
658 return error;
659}
660
Heiko Carstensa26eab22009-01-14 14:14:17 +0100661SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800662{
663 return sys_fchmodat(AT_FDCWD, filename, mode);
664}
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
667{
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200668 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 int error;
670 struct iattr newattrs;
671
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 newattrs.ia_valid = ATTR_CTIME;
673 if (user != (uid_t) -1) {
674 newattrs.ia_valid |= ATTR_UID;
675 newattrs.ia_uid = user;
676 }
677 if (group != (gid_t) -1) {
678 newattrs.ia_valid |= ATTR_GID;
679 newattrs.ia_gid = group;
680 }
681 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700682 newattrs.ia_valid |=
683 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800684 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 error = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800686 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200687
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 return error;
689}
690
Heiko Carstensca013e92009-01-14 14:14:19 +0100691SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692{
Al Viro2d8f3032008-07-22 09:59:21 -0400693 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 int error;
695
Al Viro2d8f3032008-07-22 09:59:21 -0400696 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700697 if (error)
698 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400699 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800700 if (error)
701 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400702 error = chown_common(path.dentry, user, group);
703 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800704out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400705 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700706out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 return error;
708}
709
Heiko Carstens6559eed82009-01-14 14:14:32 +0100710SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
711 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800712{
Al Viro2d8f3032008-07-22 09:59:21 -0400713 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800714 int error = -EINVAL;
715 int follow;
716
717 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
718 goto out;
719
720 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
Al Viro2d8f3032008-07-22 09:59:21 -0400721 error = user_path_at(dfd, filename, follow, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700722 if (error)
723 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400724 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800725 if (error)
726 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400727 error = chown_common(path.dentry, user, group);
728 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800729out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400730 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800731out:
732 return error;
733}
734
Heiko Carstensca013e92009-01-14 14:14:19 +0100735SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
Al Viro2d8f3032008-07-22 09:59:21 -0400737 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 int error;
739
Al Viro2d8f3032008-07-22 09:59:21 -0400740 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700741 if (error)
742 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400743 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800744 if (error)
745 goto out_release;
Al Viro2d8f3032008-07-22 09:59:21 -0400746 error = chown_common(path.dentry, user, group);
747 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800748out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400749 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700750out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 return error;
752}
753
Heiko Carstensca013e92009-01-14 14:14:19 +0100754SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 struct file * file;
757 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700758 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700761 if (!file)
762 goto out;
763
Dave Hansen2af482a2008-02-15 14:37:50 -0800764 error = mnt_want_write(file->f_path.mnt);
765 if (error)
766 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800767 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400768 audit_inode(NULL, dentry);
Dave Hansen6902d922006-09-30 23:29:01 -0700769 error = chown_common(dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800770 mnt_drop_write(file->f_path.mnt);
771out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700772 fput(file);
773out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 return error;
775}
776
Dave Hansen4a3fd212008-02-15 14:37:48 -0800777/*
778 * You have to be very careful that these write
779 * counts get cleaned up in error cases and
780 * upon __fput(). This should probably never
781 * be called outside of __dentry_open().
782 */
783static inline int __get_file_write_access(struct inode *inode,
784 struct vfsmount *mnt)
785{
786 int error;
787 error = get_write_access(inode);
788 if (error)
789 return error;
790 /*
791 * Do not take mount writer counts on
792 * special files since no writes to
793 * the mount itself will occur.
794 */
795 if (!special_file(inode->i_mode)) {
796 /*
797 * Balanced in __fput()
798 */
799 error = mnt_want_write(mnt);
800 if (error)
801 put_write_access(inode);
802 }
803 return error;
804}
805
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700806static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Trond Myklebust834f2a42005-10-18 14:20:16 -0700807 int flags, struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100808 int (*open)(struct inode *, struct file *),
809 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 struct inode *inode;
812 int error;
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 f->f_flags = flags;
Al Viroaeb5d722008-09-02 15:28:45 -0400815 f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700816 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 inode = dentry->d_inode;
818 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800819 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 if (error)
821 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800822 if (!special_file(inode->i_mode))
823 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 }
825
826 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800827 f->f_path.dentry = dentry;
828 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 f->f_pos = 0;
830 f->f_op = fops_get(inode->i_fop);
831 file_move(f, &inode->i_sb->s_files);
832
David Howells745ca242008-11-14 10:39:22 +1100833 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900834 if (error)
835 goto cleanup_all;
836
Trond Myklebust834f2a42005-10-18 14:20:16 -0700837 if (!open && f->f_op)
838 open = f->f_op->open;
839 if (open) {
840 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 if (error)
842 goto cleanup_all;
843 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
846
847 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
848
849 /* NB: we're sure to have correct a_ops only after f_op->open */
850 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700851 if (!f->f_mapping->a_ops ||
852 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700853 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 fput(f);
855 f = ERR_PTR(-EINVAL);
856 }
857 }
858
859 return f;
860
861cleanup_all:
862 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800863 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800865 if (!special_file(inode->i_mode)) {
866 /*
867 * We don't consider this a real
868 * mnt_want/drop_write() pair
869 * because it all happenend right
870 * here, so just reset the state.
871 */
872 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800873 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800874 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800875 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 file_kill(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800877 f->f_path.dentry = NULL;
878 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879cleanup_file:
880 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 dput(dentry);
882 mntput(mnt);
883 return ERR_PTR(error);
884}
885
Trond Myklebust834f2a42005-10-18 14:20:16 -0700886/**
887 * lookup_instantiate_filp - instantiates the open intent filp
888 * @nd: pointer to nameidata
889 * @dentry: pointer to dentry
890 * @open: open callback
891 *
892 * Helper for filesystems that want to use lookup open intents and pass back
893 * a fully instantiated struct file to the caller.
894 * This function is meant to be called from within a filesystem's
895 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800896 * Beware of calling it for non-regular files! Those ->open methods might block
897 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
898 * leading to a deadlock, as nobody can open that fifo anymore, because
899 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700900 * Note that in case of error, nd->intent.open.file is destroyed, but the
901 * path information remains valid.
902 * If the open callback is set to NULL, then the standard f_op->open()
903 * filesystem callback is substituted.
904 */
905struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
906 int (*open)(struct inode *, struct file *))
907{
David Howells745ca242008-11-14 10:39:22 +1100908 const struct cred *cred = current_cred();
909
Trond Myklebust834f2a42005-10-18 14:20:16 -0700910 if (IS_ERR(nd->intent.open.file))
911 goto out;
912 if (IS_ERR(dentry))
913 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800914 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700915 nd->intent.open.flags - 1,
916 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100917 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700918out:
919 return nd->intent.open.file;
920out_err:
921 release_open_intent(nd);
922 nd->intent.open.file = (struct file *)dentry;
923 goto out;
924}
925EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
926
927/**
928 * nameidata_to_filp - convert a nameidata to an open filp.
929 * @nd: pointer to nameidata
930 * @flags: open flags
931 *
932 * Note that this function destroys the original nameidata
933 */
934struct file *nameidata_to_filp(struct nameidata *nd, int flags)
935{
David Howells745ca242008-11-14 10:39:22 +1100936 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700937 struct file *filp;
938
939 /* Pick up the filp from the open intent */
940 filp = nd->intent.open.file;
941 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800942 if (filp->f_path.dentry == NULL)
Jan Blunck4ac91372008-02-14 19:34:32 -0800943 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
David Howells745ca242008-11-14 10:39:22 +1100944 NULL, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700945 else
Jan Blunck1d957f92008-02-14 19:34:35 -0800946 path_put(&nd->path);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700947 return filp;
948}
949
Peter Staubach6fdcc212005-11-07 00:59:42 -0800950/*
951 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
952 * error.
953 */
David Howells745ca242008-11-14 10:39:22 +1100954struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
955 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700956{
957 int error;
958 struct file *f;
959
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800960 /*
961 * We must always pass in a valid mount pointer. Historically
962 * callers got away with not passing it, but we must enforce this at
963 * the earliest possible point now to avoid strange problems deep in the
964 * filesystem stack.
965 */
966 if (!mnt) {
967 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
968 dump_stack();
969 return ERR_PTR(-EINVAL);
970 }
971
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700972 error = -ENFILE;
973 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800974 if (f == NULL) {
975 dput(dentry);
976 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700977 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800978 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700979
David Howells745ca242008-11-14 10:39:22 +1100980 return __dentry_open(dentry, mnt, flags, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700981}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982EXPORT_SYMBOL(dentry_open);
983
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800984static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700986 struct fdtable *fdt = files_fdtable(files);
987 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800988 if (fd < files->next_fd)
989 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990}
991
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800992void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993{
994 struct files_struct *files = current->files;
995 spin_lock(&files->file_lock);
996 __put_unused_fd(files, fd);
997 spin_unlock(&files->file_lock);
998}
999
1000EXPORT_SYMBOL(put_unused_fd);
1001
1002/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001003 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 *
1005 * The VFS is full of places where we drop the files lock between
1006 * setting the open_fds bitmap and installing the file in the file
1007 * array. At any such point, we are vulnerable to a dup2() race
1008 * installing a file in the array before us. We need to detect this and
1009 * fput() the struct file we are about to overwrite in this case.
1010 *
1011 * It should never happen - if we allow dup2() do it, _really_ bad things
1012 * will follow.
1013 */
1014
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001015void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016{
1017 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001018 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001020 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001021 BUG_ON(fdt->fd[fd] != NULL);
1022 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 spin_unlock(&files->file_lock);
1024}
1025
1026EXPORT_SYMBOL(fd_install);
1027
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001028long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001030 char *tmp = getname(filename);
1031 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001034 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 if (fd >= 0) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001036 struct file *f = do_filp_open(dfd, tmp, flags, mode);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001037 if (IS_ERR(f)) {
1038 put_unused_fd(fd);
1039 fd = PTR_ERR(f);
1040 } else {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001041 fsnotify_open(f->f_path.dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001042 fd_install(fd, f);
1043 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 putname(tmp);
1046 }
1047 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001049
Heiko Carstensca013e92009-01-14 14:14:19 +01001050SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001051{
Linus Torvalds385910f2006-04-18 13:22:59 -07001052 long ret;
1053
Miklos Szeredie922efc2005-09-06 15:18:25 -07001054 if (force_o_largefile())
1055 flags |= O_LARGEFILE;
1056
Linus Torvalds385910f2006-04-18 13:22:59 -07001057 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1058 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001059 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001060 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001061}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Heiko Carstens6559eed82009-01-14 14:14:32 +01001063SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1064 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001065{
Linus Torvalds385910f2006-04-18 13:22:59 -07001066 long ret;
1067
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001068 if (force_o_largefile())
1069 flags |= O_LARGEFILE;
1070
Linus Torvalds385910f2006-04-18 13:22:59 -07001071 ret = do_sys_open(dfd, filename, flags, mode);
1072 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001073 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001074 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001075}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001076
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077#ifndef __alpha__
1078
1079/*
1080 * For backward compatibility? Maybe this should be moved
1081 * into arch/i386 instead?
1082 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001083SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
1085 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1086}
1087
1088#endif
1089
1090/*
1091 * "id" is the POSIX thread ID. We use the
1092 * files pointer for this..
1093 */
1094int filp_close(struct file *filp, fl_owner_t id)
1095{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001096 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 if (!file_count(filp)) {
1099 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001100 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 }
1102
Christoph Lameter45778ca2005-06-23 00:10:17 -07001103 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001104 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
1106 dnotify_flush(filp, id);
1107 locks_remove_posix(filp, id);
1108 fput(filp);
1109 return retval;
1110}
1111
1112EXPORT_SYMBOL(filp_close);
1113
1114/*
1115 * Careful here! We test whether the file pointer is NULL before
1116 * releasing the fd. This ensures that one clone task can't release
1117 * an fd while another clone is opening it.
1118 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001119SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120{
1121 struct file * filp;
1122 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001123 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001124 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125
1126 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001127 fdt = files_fdtable(files);
1128 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001130 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (!filp)
1132 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001133 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001134 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 __put_unused_fd(files, fd);
1136 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001137 retval = filp_close(filp, files);
1138
1139 /* can't restart close syscall because file table entry was cleared */
1140 if (unlikely(retval == -ERESTARTSYS ||
1141 retval == -ERESTARTNOINTR ||
1142 retval == -ERESTARTNOHAND ||
1143 retval == -ERESTART_RESTARTBLOCK))
1144 retval = -EINTR;
1145
1146 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148out_unlock:
1149 spin_unlock(&files->file_lock);
1150 return -EBADF;
1151}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152EXPORT_SYMBOL(sys_close);
1153
1154/*
1155 * This routine simulates a hangup on the tty, to arrange that users
1156 * are given clean terminals at login time.
1157 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001158SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159{
1160 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001161 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 return 0;
1163 }
1164 return -EPERM;
1165}
1166
1167/*
1168 * Called when an inode is about to be open.
1169 * We use this to disallow opening large files on 32bit systems if
1170 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1171 * on this flag in sys_open.
1172 */
1173int generic_file_open(struct inode * inode, struct file * filp)
1174{
1175 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001176 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 return 0;
1178}
1179
1180EXPORT_SYMBOL(generic_file_open);
1181
1182/*
1183 * This is used by subsystems that don't want seekable
1184 * file descriptors
1185 */
1186int nonseekable_open(struct inode *inode, struct file *filp)
1187{
1188 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1189 return 0;
1190}
1191
1192EXPORT_SYMBOL(nonseekable_open);