blob: bb98d2fe809f09fa382a9d6224231206e1161962 [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
125asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
126{
127 struct nameidata nd;
128 int error;
129
130 error = user_path_walk(path, &nd);
131 if (!error) {
132 struct statfs tmp;
Jan Blunck4ac91372008-02-14 19:34:32 -0800133 error = vfs_statfs_native(nd.path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
135 error = -EFAULT;
Jan Blunck1d957f92008-02-14 19:34:35 -0800136 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 }
138 return error;
139}
140
141
142asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
143{
144 struct nameidata nd;
145 long error;
146
147 if (sz != sizeof(*buf))
148 return -EINVAL;
149 error = user_path_walk(path, &nd);
150 if (!error) {
151 struct statfs64 tmp;
Jan Blunck4ac91372008-02-14 19:34:32 -0800152 error = vfs_statfs64(nd.path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
154 error = -EFAULT;
Jan Blunck1d957f92008-02-14 19:34:35 -0800155 path_put(&nd.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
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800226static long do_sys_truncate(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
228 struct nameidata nd;
229 struct inode * inode;
230 int error;
231
232 error = -EINVAL;
233 if (length < 0) /* sorry, but loff_t says... */
234 goto out;
235
236 error = user_path_walk(path, &nd);
237 if (error)
238 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800239 inode = nd.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
Dave Hansen9ac9b842008-02-15 14:37:52 -0800250 error = mnt_want_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 if (error)
252 goto dput_and_out;
253
Dave Hansen9ac9b842008-02-15 14:37:52 -0800254 error = vfs_permission(&nd, MAY_WRITE);
255 if (error)
256 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 error = -EPERM;
259 if (IS_IMMUTABLE(inode) || 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);
Jan Blunck4ac91372008-02-14 19:34:32 -0800277 error = do_truncate(nd.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:
283 mnt_drop_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284dput_and_out:
Jan Blunck1d957f92008-02-14 19:34:35 -0800285 path_put(&nd.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{
428 struct nameidata nd;
429 int old_fsuid, old_fsgid;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700430 kernel_cap_t uninitialized_var(old_cap); /* !SECURE_NO_SETUID_FIXUP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 int res;
432
433 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
434 return -EINVAL;
435
436 old_fsuid = current->fsuid;
437 old_fsgid = current->fsgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 current->fsuid = current->uid;
440 current->fsgid = current->gid;
441
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700442 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
443 /*
444 * Clear the capabilities if we switch to a non-root user
445 */
446#ifndef CONFIG_SECURITY_FILE_CAPABILITIES
447 /*
448 * FIXME: There is a race here against sys_capset. The
449 * capabilities can change yet we will restore the old
450 * value below. We should hold task_capabilities_lock,
451 * but we cannot because user_path_walk can sleep.
452 */
453#endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
454 if (current->uid)
455 old_cap = cap_set_effective(__cap_empty_set);
456 else
457 old_cap = cap_set_effective(current->cap_permitted);
458 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800460 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700461 if (res)
462 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
Dave Hansen6902d922006-09-30 23:29:01 -0700464 res = vfs_permission(&nd, mode);
465 /* SuS v2 requires we report a read only fs too */
466 if(res || !(mode & S_IWOTH) ||
Jan Blunck4ac91372008-02-14 19:34:32 -0800467 special_file(nd.path.dentry->d_inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700468 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800469 /*
470 * This is a rare case where using __mnt_is_readonly()
471 * is OK without a mnt_want/drop_write() pair. Since
472 * no actual write to the fs is performed here, we do
473 * not need to telegraph to that to anyone.
474 *
475 * By doing this, we accept that this access is
476 * inherently racy and know that the fs may change
477 * state before we even see this result.
478 */
479 if (__mnt_is_readonly(nd.path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700480 res = -EROFS;
481
482out_path_release:
Jan Blunck1d957f92008-02-14 19:34:35 -0800483 path_put(&nd.path);
Dave Hansen6902d922006-09-30 23:29:01 -0700484out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 current->fsuid = old_fsuid;
486 current->fsgid = old_fsgid;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700487
488 if (!issecure(SECURE_NO_SETUID_FIXUP))
489 cap_set_effective(old_cap);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 return res;
492}
493
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800494asmlinkage long sys_access(const char __user *filename, int mode)
495{
496 return sys_faccessat(AT_FDCWD, filename, mode);
497}
498
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499asmlinkage long sys_chdir(const char __user * filename)
500{
501 struct nameidata nd;
502 int error;
503
Miklos Szeredi650a8982006-09-29 01:59:35 -0700504 error = __user_walk(filename,
505 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 if (error)
507 goto out;
508
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800509 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 if (error)
511 goto dput_and_out;
512
Jan Blunckac748a02008-02-14 19:34:39 -0800513 set_fs_pwd(current->fs, &nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
515dput_and_out:
Jan Blunck1d957f92008-02-14 19:34:35 -0800516 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517out:
518 return error;
519}
520
521asmlinkage long sys_fchdir(unsigned int fd)
522{
523 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 int error;
526
527 error = -EBADF;
528 file = fget(fd);
529 if (!file)
530 goto out;
531
Jan Blunckac748a02008-02-14 19:34:39 -0800532 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534 error = -ENOTDIR;
535 if (!S_ISDIR(inode->i_mode))
536 goto out_putf;
537
Christoph Hellwig8c744fb2005-11-08 21:35:04 -0800538 error = file_permission(file, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800540 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541out_putf:
542 fput(file);
543out:
544 return error;
545}
546
547asmlinkage long sys_chroot(const char __user * filename)
548{
549 struct nameidata nd;
550 int error;
551
552 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
553 if (error)
554 goto out;
555
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800556 error = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 if (error)
558 goto dput_and_out;
559
560 error = -EPERM;
561 if (!capable(CAP_SYS_CHROOT))
562 goto dput_and_out;
563
Jan Blunckac748a02008-02-14 19:34:39 -0800564 set_fs_root(current->fs, &nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 set_fs_altroot();
566 error = 0;
567dput_and_out:
Jan Blunck1d957f92008-02-14 19:34:35 -0800568 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569out:
570 return error;
571}
572
573asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
574{
575 struct inode * inode;
576 struct dentry * dentry;
577 struct file * file;
578 int err = -EBADF;
579 struct iattr newattrs;
580
581 file = fget(fd);
582 if (!file)
583 goto out;
584
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800585 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 inode = dentry->d_inode;
587
Al Viro5a190ae2007-06-07 12:19:32 -0400588 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000589
Dave Hansen2af482a2008-02-15 14:37:50 -0800590 err = mnt_want_write(file->f_path.mnt);
591 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 goto out_putf;
593 err = -EPERM;
594 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
Dave Hansen2af482a2008-02-15 14:37:50 -0800595 goto out_drop_write;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800596 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 if (mode == (mode_t) -1)
598 mode = inode->i_mode;
599 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
600 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
601 err = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800602 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Dave Hansen2af482a2008-02-15 14:37:50 -0800604out_drop_write:
605 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606out_putf:
607 fput(file);
608out:
609 return err;
610}
611
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800612asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
613 mode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614{
615 struct nameidata nd;
616 struct inode * inode;
617 int error;
618 struct iattr newattrs;
619
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800620 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 if (error)
622 goto out;
Jan Blunck4ac91372008-02-14 19:34:32 -0800623 inode = nd.path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624
Dave Hansen2af482a2008-02-15 14:37:50 -0800625 error = mnt_want_write(nd.path.mnt);
626 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 goto dput_and_out;
628
629 error = -EPERM;
630 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
Dave Hansen2af482a2008-02-15 14:37:50 -0800631 goto out_drop_write;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
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;
Jan Blunck4ac91372008-02-14 19:34:32 -0800638 error = notify_change(nd.path.dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800639 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
Dave Hansen2af482a2008-02-15 14:37:50 -0800641out_drop_write:
642 mnt_drop_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643dput_and_out:
Jan Blunck1d957f92008-02-14 19:34:35 -0800644 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645out:
646 return error;
647}
648
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800649asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
650{
651 return sys_fchmodat(AT_FDCWD, filename, mode);
652}
653
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
655{
656 struct inode * inode;
657 int error;
658 struct iattr newattrs;
659
660 error = -ENOENT;
661 if (!(inode = dentry->d_inode)) {
662 printk(KERN_ERR "chown_common: NULL inode\n");
663 goto out;
664 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 error = -EPERM;
666 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
667 goto out;
668 newattrs.ia_valid = ATTR_CTIME;
669 if (user != (uid_t) -1) {
670 newattrs.ia_valid |= ATTR_UID;
671 newattrs.ia_uid = user;
672 }
673 if (group != (gid_t) -1) {
674 newattrs.ia_valid |= ATTR_GID;
675 newattrs.ia_gid = group;
676 }
677 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700678 newattrs.ia_valid |=
679 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800680 mutex_lock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 error = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800682 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683out:
684 return error;
685}
686
687asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
688{
689 struct nameidata nd;
690 int error;
691
692 error = user_path_walk(filename, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700693 if (error)
694 goto out;
Dave Hansen2af482a2008-02-15 14:37:50 -0800695 error = mnt_want_write(nd.path.mnt);
696 if (error)
697 goto out_release;
Jan Blunck4ac91372008-02-14 19:34:32 -0800698 error = chown_common(nd.path.dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800699 mnt_drop_write(nd.path.mnt);
700out_release:
Jan Blunck1d957f92008-02-14 19:34:35 -0800701 path_put(&nd.path);
Dave Hansen6902d922006-09-30 23:29:01 -0700702out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return error;
704}
705
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800706asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
707 gid_t group, int flag)
708{
709 struct nameidata nd;
710 int error = -EINVAL;
711 int follow;
712
713 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
714 goto out;
715
716 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
717 error = __user_walk_fd(dfd, filename, follow, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700718 if (error)
719 goto out;
Dave Hansen2af482a2008-02-15 14:37:50 -0800720 error = mnt_want_write(nd.path.mnt);
721 if (error)
722 goto out_release;
Jan Blunck4ac91372008-02-14 19:34:32 -0800723 error = chown_common(nd.path.dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800724 mnt_drop_write(nd.path.mnt);
725out_release:
Jan Blunck1d957f92008-02-14 19:34:35 -0800726 path_put(&nd.path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800727out:
728 return error;
729}
730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
732{
733 struct nameidata nd;
734 int error;
735
736 error = user_path_walk_link(filename, &nd);
Dave Hansen6902d922006-09-30 23:29:01 -0700737 if (error)
738 goto out;
Dave Hansen2af482a2008-02-15 14:37:50 -0800739 error = mnt_want_write(nd.path.mnt);
740 if (error)
741 goto out_release;
Jan Blunck4ac91372008-02-14 19:34:32 -0800742 error = chown_common(nd.path.dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800743 mnt_drop_write(nd.path.mnt);
744out_release:
Jan Blunck1d957f92008-02-14 19:34:35 -0800745 path_put(&nd.path);
Dave Hansen6902d922006-09-30 23:29:01 -0700746out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 return error;
748}
749
750
751asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
752{
753 struct file * file;
754 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700755 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756
757 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700758 if (!file)
759 goto out;
760
Dave Hansen2af482a2008-02-15 14:37:50 -0800761 error = mnt_want_write(file->f_path.mnt);
762 if (error)
763 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800764 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400765 audit_inode(NULL, dentry);
Dave Hansen6902d922006-09-30 23:29:01 -0700766 error = chown_common(dentry, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800767 mnt_drop_write(file->f_path.mnt);
768out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700769 fput(file);
770out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 return error;
772}
773
Dave Hansen4a3fd212008-02-15 14:37:48 -0800774/*
775 * You have to be very careful that these write
776 * counts get cleaned up in error cases and
777 * upon __fput(). This should probably never
778 * be called outside of __dentry_open().
779 */
780static inline int __get_file_write_access(struct inode *inode,
781 struct vfsmount *mnt)
782{
783 int error;
784 error = get_write_access(inode);
785 if (error)
786 return error;
787 /*
788 * Do not take mount writer counts on
789 * special files since no writes to
790 * the mount itself will occur.
791 */
792 if (!special_file(inode->i_mode)) {
793 /*
794 * Balanced in __fput()
795 */
796 error = mnt_want_write(mnt);
797 if (error)
798 put_write_access(inode);
799 }
800 return error;
801}
802
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700803static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Trond Myklebust834f2a42005-10-18 14:20:16 -0700804 int flags, struct file *f,
805 int (*open)(struct inode *, struct file *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 struct inode *inode;
808 int error;
809
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 f->f_flags = flags;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700811 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
812 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 inode = dentry->d_inode;
814 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800815 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 if (error)
817 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800818 if (!special_file(inode->i_mode))
819 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 }
821
822 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800823 f->f_path.dentry = dentry;
824 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 f->f_pos = 0;
826 f->f_op = fops_get(inode->i_fop);
827 file_move(f, &inode->i_sb->s_files);
828
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900829 error = security_dentry_open(f);
830 if (error)
831 goto cleanup_all;
832
Trond Myklebust834f2a42005-10-18 14:20:16 -0700833 if (!open && f->f_op)
834 open = f->f_op->open;
835 if (open) {
836 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 if (error)
838 goto cleanup_all;
839 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700840
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
842
843 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
844
845 /* NB: we're sure to have correct a_ops only after f_op->open */
846 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700847 if (!f->f_mapping->a_ops ||
848 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700849 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 fput(f);
851 f = ERR_PTR(-EINVAL);
852 }
853 }
854
855 return f;
856
857cleanup_all:
858 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800859 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800861 if (!special_file(inode->i_mode)) {
862 /*
863 * We don't consider this a real
864 * mnt_want/drop_write() pair
865 * because it all happenend right
866 * here, so just reset the state.
867 */
868 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800869 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800870 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800871 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 file_kill(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800873 f->f_path.dentry = NULL;
874 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875cleanup_file:
876 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 dput(dentry);
878 mntput(mnt);
879 return ERR_PTR(error);
880}
881
Trond Myklebust834f2a42005-10-18 14:20:16 -0700882/**
883 * lookup_instantiate_filp - instantiates the open intent filp
884 * @nd: pointer to nameidata
885 * @dentry: pointer to dentry
886 * @open: open callback
887 *
888 * Helper for filesystems that want to use lookup open intents and pass back
889 * a fully instantiated struct file to the caller.
890 * This function is meant to be called from within a filesystem's
891 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800892 * Beware of calling it for non-regular files! Those ->open methods might block
893 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
894 * leading to a deadlock, as nobody can open that fifo anymore, because
895 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700896 * Note that in case of error, nd->intent.open.file is destroyed, but the
897 * path information remains valid.
898 * If the open callback is set to NULL, then the standard f_op->open()
899 * filesystem callback is substituted.
900 */
901struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
902 int (*open)(struct inode *, struct file *))
903{
904 if (IS_ERR(nd->intent.open.file))
905 goto out;
906 if (IS_ERR(dentry))
907 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800908 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700909 nd->intent.open.flags - 1,
910 nd->intent.open.file,
911 open);
912out:
913 return nd->intent.open.file;
914out_err:
915 release_open_intent(nd);
916 nd->intent.open.file = (struct file *)dentry;
917 goto out;
918}
919EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
920
921/**
922 * nameidata_to_filp - convert a nameidata to an open filp.
923 * @nd: pointer to nameidata
924 * @flags: open flags
925 *
926 * Note that this function destroys the original nameidata
927 */
928struct file *nameidata_to_filp(struct nameidata *nd, int flags)
929{
930 struct file *filp;
931
932 /* Pick up the filp from the open intent */
933 filp = nd->intent.open.file;
934 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800935 if (filp->f_path.dentry == NULL)
Jan Blunck4ac91372008-02-14 19:34:32 -0800936 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
937 NULL);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700938 else
Jan Blunck1d957f92008-02-14 19:34:35 -0800939 path_put(&nd->path);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700940 return filp;
941}
942
Peter Staubach6fdcc212005-11-07 00:59:42 -0800943/*
944 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
945 * error.
946 */
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700947struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
948{
949 int error;
950 struct file *f;
951
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800952 /*
953 * We must always pass in a valid mount pointer. Historically
954 * callers got away with not passing it, but we must enforce this at
955 * the earliest possible point now to avoid strange problems deep in the
956 * filesystem stack.
957 */
958 if (!mnt) {
959 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
960 dump_stack();
961 return ERR_PTR(-EINVAL);
962 }
963
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700964 error = -ENFILE;
965 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800966 if (f == NULL) {
967 dput(dentry);
968 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700969 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800970 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700971
Trond Myklebust834f2a42005-10-18 14:20:16 -0700972 return __dentry_open(dentry, mnt, flags, f, NULL);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700973}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974EXPORT_SYMBOL(dentry_open);
975
976/*
977 * Find an empty file descriptor entry, and mark it busy.
978 */
Ulrich Drepper4a195422007-07-15 23:40:34 -0700979int get_unused_fd_flags(int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980{
981 struct files_struct * files = current->files;
982 int fd, error;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700983 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984
985 error = -EMFILE;
986 spin_lock(&files->file_lock);
987
988repeat:
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700989 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800990 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800991 files->next_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992
993 /*
994 * N.B. For clone tasks sharing a files structure, this test
995 * will limit the total number of files that can be opened.
996 */
997 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
998 goto out;
999
1000 /* Do we need to expand the fd array or fd set? */
1001 error = expand_files(files, fd);
1002 if (error < 0)
1003 goto out;
1004
1005 if (error) {
1006 /*
1007 * If we needed to expand the fs array we
1008 * might have blocked - try again.
1009 */
1010 error = -EMFILE;
1011 goto repeat;
1012 }
1013
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001014 FD_SET(fd, fdt->open_fds);
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001015 if (flags & O_CLOEXEC)
1016 FD_SET(fd, fdt->close_on_exec);
1017 else
1018 FD_CLR(fd, fdt->close_on_exec);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001019 files->next_fd = fd + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020#if 1
1021 /* Sanity check */
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001022 if (fdt->fd[fd] != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001024 fdt->fd[fd] = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 }
1026#endif
1027 error = fd;
1028
1029out:
1030 spin_unlock(&files->file_lock);
1031 return error;
1032}
1033
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001034int get_unused_fd(void)
1035{
1036 return get_unused_fd_flags(0);
1037}
1038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039EXPORT_SYMBOL(get_unused_fd);
1040
Matt Mackallb01ec0e2006-01-08 01:05:20 -08001041static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042{
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001043 struct fdtable *fdt = files_fdtable(files);
1044 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001045 if (fd < files->next_fd)
1046 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047}
1048
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001049void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050{
1051 struct files_struct *files = current->files;
1052 spin_lock(&files->file_lock);
1053 __put_unused_fd(files, fd);
1054 spin_unlock(&files->file_lock);
1055}
1056
1057EXPORT_SYMBOL(put_unused_fd);
1058
1059/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001060 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 *
1062 * The VFS is full of places where we drop the files lock between
1063 * setting the open_fds bitmap and installing the file in the file
1064 * array. At any such point, we are vulnerable to a dup2() race
1065 * installing a file in the array before us. We need to detect this and
1066 * fput() the struct file we are about to overwrite in this case.
1067 *
1068 * It should never happen - if we allow dup2() do it, _really_ bad things
1069 * will follow.
1070 */
1071
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001072void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073{
1074 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001075 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001077 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001078 BUG_ON(fdt->fd[fd] != NULL);
1079 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 spin_unlock(&files->file_lock);
1081}
1082
1083EXPORT_SYMBOL(fd_install);
1084
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001085long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001087 char *tmp = getname(filename);
1088 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001091 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 if (fd >= 0) {
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001093 struct file *f = do_filp_open(dfd, tmp, flags, mode);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001094 if (IS_ERR(f)) {
1095 put_unused_fd(fd);
1096 fd = PTR_ERR(f);
1097 } else {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001098 fsnotify_open(f->f_path.dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001099 fd_install(fd, f);
1100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 putname(tmp);
1103 }
1104 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001106
1107asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1108{
Linus Torvalds385910f2006-04-18 13:22:59 -07001109 long ret;
1110
Miklos Szeredie922efc2005-09-06 15:18:25 -07001111 if (force_o_largefile())
1112 flags |= O_LARGEFILE;
1113
Linus Torvalds385910f2006-04-18 13:22:59 -07001114 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1115 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001116 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001117 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001118}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001120asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1121 int mode)
1122{
Linus Torvalds385910f2006-04-18 13:22:59 -07001123 long ret;
1124
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001125 if (force_o_largefile())
1126 flags |= O_LARGEFILE;
1127
Linus Torvalds385910f2006-04-18 13:22:59 -07001128 ret = do_sys_open(dfd, filename, flags, mode);
1129 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001130 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001131 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001132}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001133
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134#ifndef __alpha__
1135
1136/*
1137 * For backward compatibility? Maybe this should be moved
1138 * into arch/i386 instead?
1139 */
1140asmlinkage long sys_creat(const char __user * pathname, int mode)
1141{
1142 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1143}
1144
1145#endif
1146
1147/*
1148 * "id" is the POSIX thread ID. We use the
1149 * files pointer for this..
1150 */
1151int filp_close(struct file *filp, fl_owner_t id)
1152{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001153 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
1155 if (!file_count(filp)) {
1156 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001157 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 }
1159
Christoph Lameter45778ca2005-06-23 00:10:17 -07001160 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001161 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163 dnotify_flush(filp, id);
1164 locks_remove_posix(filp, id);
1165 fput(filp);
1166 return retval;
1167}
1168
1169EXPORT_SYMBOL(filp_close);
1170
1171/*
1172 * Careful here! We test whether the file pointer is NULL before
1173 * releasing the fd. This ensures that one clone task can't release
1174 * an fd while another clone is opening it.
1175 */
1176asmlinkage long sys_close(unsigned int fd)
1177{
1178 struct file * filp;
1179 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001180 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001181 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
1183 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001184 fdt = files_fdtable(files);
1185 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001187 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 if (!filp)
1189 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001190 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001191 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 __put_unused_fd(files, fd);
1193 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001194 retval = filp_close(filp, files);
1195
1196 /* can't restart close syscall because file table entry was cleared */
1197 if (unlikely(retval == -ERESTARTSYS ||
1198 retval == -ERESTARTNOINTR ||
1199 retval == -ERESTARTNOHAND ||
1200 retval == -ERESTART_RESTARTBLOCK))
1201 retval = -EINTR;
1202
1203 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
1205out_unlock:
1206 spin_unlock(&files->file_lock);
1207 return -EBADF;
1208}
1209
1210EXPORT_SYMBOL(sys_close);
1211
1212/*
1213 * This routine simulates a hangup on the tty, to arrange that users
1214 * are given clean terminals at login time.
1215 */
1216asmlinkage long sys_vhangup(void)
1217{
1218 if (capable(CAP_SYS_TTY_CONFIG)) {
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001219 /* XXX: this needs locking */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 tty_vhangup(current->signal->tty);
1221 return 0;
1222 }
1223 return -EPERM;
1224}
1225
1226/*
1227 * Called when an inode is about to be open.
1228 * We use this to disallow opening large files on 32bit systems if
1229 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1230 * on this flag in sys_open.
1231 */
1232int generic_file_open(struct inode * inode, struct file * filp)
1233{
1234 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001235 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 return 0;
1237}
1238
1239EXPORT_SYMBOL(generic_file_open);
1240
1241/*
1242 * This is used by subsystems that don't want seekable
1243 * file descriptors
1244 */
1245int nonseekable_open(struct inode *inode, struct file *filp)
1246{
1247 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1248 return 0;
1249}
1250
1251EXPORT_SYMBOL(nonseekable_open);