blob: d74e1983e8dc478145dc8da369c0479bb947d8c2 [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>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/tty.h>
14#include <linux/namei.h>
15#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080016#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070017#include <linux/securebits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080020#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090021#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040030#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050031#include <linux/ima.h>
Eric Paris2dfc1ca2009-12-17 20:30:52 -050032#include <linux/dnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Eric Parise81e3f42009-12-04 15:47:36 -050034#include "internal.h"
35
NeilBrown4a301312006-01-08 01:02:39 -080036int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
37 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070038{
Amerigo Wang939a9422009-08-20 19:29:03 -070039 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 struct iattr newattrs;
41
42 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
43 if (length < 0)
44 return -EINVAL;
45
46 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080047 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080048 if (filp) {
49 newattrs.ia_file = filp;
50 newattrs.ia_valid |= ATTR_FILE;
51 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Linus Torvalds7b82dc02007-05-08 20:10:00 -070053 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -070054 ret = should_remove_suid(dentry);
55 if (ret)
56 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070057
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080058 mutex_lock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070059 ret = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080060 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070061 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062}
63
Al Viro2d8f3032008-07-22 09:59:21 -040064static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
Al Viro2d8f3032008-07-22 09:59:21 -040066 struct path path;
67 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 int error;
69
70 error = -EINVAL;
71 if (length < 0) /* sorry, but loff_t says... */
72 goto out;
73
Al Viro2d8f3032008-07-22 09:59:21 -040074 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (error)
76 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -040077 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
80 error = -EISDIR;
81 if (S_ISDIR(inode->i_mode))
82 goto dput_and_out;
83
84 error = -EINVAL;
85 if (!S_ISREG(inode->i_mode))
86 goto dput_and_out;
87
Al Viro2d8f3032008-07-22 09:59:21 -040088 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 if (error)
90 goto dput_and_out;
91
Al Viro256984a2008-07-22 08:09:30 -040092 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080093 if (error)
94 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020097 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080098 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 error = get_write_access(inode);
101 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800102 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
david m. richter97003822007-07-31 00:39:12 -0700104 /*
105 * Make sure that there are no leases. get_write_access() protects
106 * against the truncate racing with a lease-granting setlease().
107 */
Al Viro8737c932009-12-24 06:47:55 -0500108 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700109 if (error)
110 goto put_write_and_out;
111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900113 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900114 error = security_path_truncate(&path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500115 if (!error)
Al Viro2d8f3032008-07-22 09:59:21 -0400116 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
david m. richter97003822007-07-31 00:39:12 -0700118put_write_and_out:
119 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800120mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400121 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400123 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124out:
125 return error;
126}
127
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200128SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200130 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131}
132
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800133static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134{
135 struct inode * inode;
136 struct dentry *dentry;
137 struct file * file;
138 int error;
139
140 error = -EINVAL;
141 if (length < 0)
142 goto out;
143 error = -EBADF;
144 file = fget(fd);
145 if (!file)
146 goto out;
147
148 /* explicitly opened as large or we are on 64-bit box */
149 if (file->f_flags & O_LARGEFILE)
150 small = 0;
151
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800152 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 inode = dentry->d_inode;
154 error = -EINVAL;
155 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
156 goto out_putf;
157
158 error = -EINVAL;
159 /* Cannot ftruncate over 2^31 bytes without large file support */
160 if (small && length > MAX_NON_LFS)
161 goto out_putf;
162
163 error = -EPERM;
164 if (IS_APPEND(inode))
165 goto out_putf;
166
167 error = locks_verify_truncate(inode, file, length);
168 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900169 error = security_path_truncate(&file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900170 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700171 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172out_putf:
173 fput(file);
174out:
175 return error;
176}
177
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100178SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700180 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700181 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700182 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700183 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184}
185
186/* LFS versions of truncate are only needed on 32 bit machines */
187#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100188SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
190 return do_sys_truncate(path, length);
191}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100192#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
193asmlinkage long SyS_truncate64(long path, loff_t length)
194{
195 return SYSC_truncate64((const char __user *) path, length);
196}
197SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
198#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100200SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700202 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700203 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700204 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700205 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100207#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
208asmlinkage long SyS_ftruncate64(long fd, loff_t length)
209{
210 return SYSC_ftruncate64((unsigned int) fd, length);
211}
212SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100214#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400216
217int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400218{
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400219 struct inode *inode = file->f_path.dentry->d_inode;
220 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400221
222 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400223 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400224
225 /* Return error if mode is not supported */
Amit Arora97ac7352007-07-17 21:42:44 -0400226 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400227 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400228
Amit Arora97ac7352007-07-17 21:42:44 -0400229 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400230 return -EBADF;
Amit Arora97ac7352007-07-17 21:42:44 -0400231 /*
232 * Revalidate the write permissions, in case security policy has
233 * changed since the files were opened.
234 */
235 ret = security_file_permission(file, MAY_WRITE);
236 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400237 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400238
Amit Arora97ac7352007-07-17 21:42:44 -0400239 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400240 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400241
Amit Arora97ac7352007-07-17 21:42:44 -0400242 /*
243 * Let individual file system decide if it supports preallocation
244 * for directories or not.
245 */
246 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400247 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400248
Amit Arora97ac7352007-07-17 21:42:44 -0400249 /* Check for wrap through zero too */
250 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400251 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400252
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400253 if (!inode->i_op->fallocate)
254 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400255
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400256 return inode->i_op->fallocate(inode, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400257}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400258
259SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
260{
261 struct file *file;
262 int error = -EBADF;
263
264 file = fget(fd);
265 if (file) {
266 error = do_fallocate(file, mode, offset, len);
267 fput(file);
268 }
269
270 return error;
271}
272
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100273#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
274asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
275{
276 return SYSC_fallocate((int)fd, (int)mode, offset, len);
277}
278SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
279#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281/*
282 * access() needs to use the real uid/gid, not the effective uid/gid.
283 * We do this by temporarily clearing all FS-related capabilities and
284 * switching the fsuid/fsgid around to the real ones.
285 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100286SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287{
David Howellsd84f4f92008-11-14 10:39:23 +1100288 const struct cred *old_cred;
289 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400290 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400291 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 int res;
293
294 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
295 return -EINVAL;
296
David Howellsd84f4f92008-11-14 10:39:23 +1100297 override_cred = prepare_creds();
298 if (!override_cred)
299 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
David Howellsd84f4f92008-11-14 10:39:23 +1100301 override_cred->fsuid = override_cred->uid;
302 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700304 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100305 /* Clear the capabilities if we switch to a non-root user */
David Howellsd84f4f92008-11-14 10:39:23 +1100306 if (override_cred->uid)
307 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700308 else
David Howellsd84f4f92008-11-14 10:39:23 +1100309 override_cred->cap_effective =
310 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700311 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
David Howellsd84f4f92008-11-14 10:39:23 +1100313 old_cred = override_creds(override_cred);
314
Al Viro2d8f3032008-07-22 09:59:21 -0400315 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700316 if (res)
317 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
Al Viro2d8f3032008-07-22 09:59:21 -0400319 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400320
321 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400322 /*
323 * MAY_EXEC on regular files is denied if the fs is mounted
324 * with the "noexec" flag.
325 */
326 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400327 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400328 goto out_path_release;
329 }
330
Al Viro256984a2008-07-22 08:09:30 -0400331 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700332 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400333 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700334 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800335 /*
336 * This is a rare case where using __mnt_is_readonly()
337 * is OK without a mnt_want/drop_write() pair. Since
338 * no actual write to the fs is performed here, we do
339 * not need to telegraph to that to anyone.
340 *
341 * By doing this, we accept that this access is
342 * inherently racy and know that the fs may change
343 * state before we even see this result.
344 */
Al Viro2d8f3032008-07-22 09:59:21 -0400345 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700346 res = -EROFS;
347
348out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400349 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700350out:
David Howellsd84f4f92008-11-14 10:39:23 +1100351 revert_creds(old_cred);
352 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 return res;
354}
355
Heiko Carstensca013e92009-01-14 14:14:19 +0100356SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800357{
358 return sys_faccessat(AT_FDCWD, filename, mode);
359}
360
Heiko Carstens3cdad422009-01-14 14:14:22 +0100361SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362{
Al Viro2d8f3032008-07-22 09:59:21 -0400363 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 int error;
365
Al Viro2d8f3032008-07-22 09:59:21 -0400366 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 if (error)
368 goto out;
369
Eric Paris9cfcac82010-07-23 11:43:51 -0400370 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 if (error)
372 goto dput_and_out;
373
Al Viro2d8f3032008-07-22 09:59:21 -0400374 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
376dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400377 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378out:
379 return error;
380}
381
Heiko Carstens3cdad422009-01-14 14:14:22 +0100382SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383{
384 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 int error;
387
388 error = -EBADF;
389 file = fget(fd);
390 if (!file)
391 goto out;
392
Jan Blunckac748a02008-02-14 19:34:39 -0800393 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
395 error = -ENOTDIR;
396 if (!S_ISDIR(inode->i_mode))
397 goto out_putf;
398
Eric Paris9cfcac82010-07-23 11:43:51 -0400399 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800401 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402out_putf:
403 fput(file);
404out:
405 return error;
406}
407
Heiko Carstens3480b252009-01-14 14:14:16 +0100408SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409{
Al Viro2d8f3032008-07-22 09:59:21 -0400410 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 int error;
412
Al Viro2d8f3032008-07-22 09:59:21 -0400413 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 if (error)
415 goto out;
416
Eric Paris9cfcac82010-07-23 11:43:51 -0400417 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 if (error)
419 goto dput_and_out;
420
421 error = -EPERM;
422 if (!capable(CAP_SYS_CHROOT))
423 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900424 error = security_path_chroot(&path);
425 if (error)
426 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
Al Viro2d8f3032008-07-22 09:59:21 -0400428 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 error = 0;
430dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400431 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432out:
433 return error;
434}
435
Heiko Carstensa26eab22009-01-14 14:14:17 +0100436SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437{
438 struct inode * inode;
439 struct dentry * dentry;
440 struct file * file;
441 int err = -EBADF;
442 struct iattr newattrs;
443
444 file = fget(fd);
445 if (!file)
446 goto out;
447
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800448 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 inode = dentry->d_inode;
450
Al Viro5a190ae2007-06-07 12:19:32 -0400451 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000452
npiggin@suse.de96029c42009-04-26 20:25:55 +1000453 err = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800454 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 goto out_putf;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900456 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900457 err = security_path_chmod(dentry, file->f_vfsmnt, mode);
458 if (err)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900459 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 if (mode == (mode_t) -1)
461 mode = inode->i_mode;
462 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
463 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
464 err = notify_change(dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900465out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800466 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800467 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468out_putf:
469 fput(file);
470out:
471 return err;
472}
473
Heiko Carstens6559eed82009-01-14 14:14:32 +0100474SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475{
Al Viro2d8f3032008-07-22 09:59:21 -0400476 struct path path;
477 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 int error;
479 struct iattr newattrs;
480
Al Viro2d8f3032008-07-22 09:59:21 -0400481 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (error)
483 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400484 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Al Viro2d8f3032008-07-22 09:59:21 -0400486 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800487 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 goto dput_and_out;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900489 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900490 error = security_path_chmod(path.dentry, path.mnt, mode);
491 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900492 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 if (mode == (mode_t) -1)
494 mode = inode->i_mode;
495 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
496 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400497 error = notify_change(path.dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900498out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800499 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400500 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400502 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503out:
504 return error;
505}
506
Heiko Carstensa26eab22009-01-14 14:14:17 +0100507SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800508{
509 return sys_fchmodat(AT_FDCWD, filename, mode);
510}
511
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900512static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900514 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 int error;
516 struct iattr newattrs;
517
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 newattrs.ia_valid = ATTR_CTIME;
519 if (user != (uid_t) -1) {
520 newattrs.ia_valid |= ATTR_UID;
521 newattrs.ia_uid = user;
522 }
523 if (group != (gid_t) -1) {
524 newattrs.ia_valid |= ATTR_GID;
525 newattrs.ia_gid = group;
526 }
527 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700528 newattrs.ia_valid |=
529 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800530 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900531 error = security_path_chown(path, user, group);
532 if (!error)
533 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800534 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 return error;
537}
538
Heiko Carstensca013e92009-01-14 14:14:19 +0100539SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
Al Viro2d8f3032008-07-22 09:59:21 -0400541 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 int error;
543
Al Viro2d8f3032008-07-22 09:59:21 -0400544 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700545 if (error)
546 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400547 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800548 if (error)
549 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900550 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400551 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800552out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400553 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700554out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return error;
556}
557
Heiko Carstens6559eed82009-01-14 14:14:32 +0100558SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
559 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800560{
Al Viro2d8f3032008-07-22 09:59:21 -0400561 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800562 int error = -EINVAL;
563 int follow;
564
565 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
566 goto out;
567
568 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
Al Viro2d8f3032008-07-22 09:59:21 -0400569 error = user_path_at(dfd, filename, follow, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700570 if (error)
571 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400572 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800573 if (error)
574 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900575 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400576 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800577out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400578 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800579out:
580 return error;
581}
582
Heiko Carstensca013e92009-01-14 14:14:19 +0100583SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584{
Al Viro2d8f3032008-07-22 09:59:21 -0400585 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 int error;
587
Al Viro2d8f3032008-07-22 09:59:21 -0400588 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700589 if (error)
590 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400591 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800592 if (error)
593 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900594 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400595 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800596out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400597 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700598out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 return error;
600}
601
Heiko Carstensca013e92009-01-14 14:14:19 +0100602SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 struct file * file;
605 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700606 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
608 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700609 if (!file)
610 goto out;
611
npiggin@suse.de96029c42009-04-26 20:25:55 +1000612 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800613 if (error)
614 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800615 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400616 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900617 error = chown_common(&file->f_path, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800618 mnt_drop_write(file->f_path.mnt);
619out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700620 fput(file);
621out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 return error;
623}
624
Dave Hansen4a3fd212008-02-15 14:37:48 -0800625/*
626 * You have to be very careful that these write
627 * counts get cleaned up in error cases and
628 * upon __fput(). This should probably never
629 * be called outside of __dentry_open().
630 */
631static inline int __get_file_write_access(struct inode *inode,
632 struct vfsmount *mnt)
633{
634 int error;
635 error = get_write_access(inode);
636 if (error)
637 return error;
638 /*
639 * Do not take mount writer counts on
640 * special files since no writes to
641 * the mount itself will occur.
642 */
643 if (!special_file(inode->i_mode)) {
644 /*
645 * Balanced in __fput()
646 */
647 error = mnt_want_write(mnt);
648 if (error)
649 put_write_access(inode);
650 }
651 return error;
652}
653
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700654static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Al Viro482928d2009-12-19 10:10:39 -0500655 struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100656 int (*open)(struct inode *, struct file *),
657 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 struct inode *inode;
660 int error;
661
Al Viro53009902009-12-19 10:15:07 -0500662 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700663 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 inode = dentry->d_inode;
665 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800666 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 if (error)
668 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800669 if (!special_file(inode->i_mode))
670 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 }
672
673 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800674 f->f_path.dentry = dentry;
675 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 f->f_pos = 0;
677 f->f_op = fops_get(inode->i_fop);
Nick Pigginee2ffa02010-08-18 04:37:35 +1000678 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
David Howells745ca242008-11-14 10:39:22 +1100680 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900681 if (error)
682 goto cleanup_all;
683
Trond Myklebust834f2a42005-10-18 14:20:16 -0700684 if (!open && f->f_op)
685 open = f->f_op->open;
686 if (open) {
687 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 if (error)
689 goto cleanup_all;
690 }
Al Virob65a9cf2009-12-16 06:27:40 -0500691 ima_counts_get(f);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700692
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
694
695 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
696
697 /* NB: we're sure to have correct a_ops only after f_op->open */
698 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700699 if (!f->f_mapping->a_ops ||
700 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700701 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 fput(f);
703 f = ERR_PTR(-EINVAL);
704 }
705 }
706
707 return f;
708
709cleanup_all:
710 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800711 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800713 if (!special_file(inode->i_mode)) {
714 /*
715 * We don't consider this a real
716 * mnt_want/drop_write() pair
717 * because it all happenend right
718 * here, so just reset the state.
719 */
720 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800721 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800722 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800723 }
Nick Pigginee2ffa02010-08-18 04:37:35 +1000724 file_sb_list_del(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800725 f->f_path.dentry = NULL;
726 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727cleanup_file:
728 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 dput(dentry);
730 mntput(mnt);
731 return ERR_PTR(error);
732}
733
Trond Myklebust834f2a42005-10-18 14:20:16 -0700734/**
735 * lookup_instantiate_filp - instantiates the open intent filp
736 * @nd: pointer to nameidata
737 * @dentry: pointer to dentry
738 * @open: open callback
739 *
740 * Helper for filesystems that want to use lookup open intents and pass back
741 * a fully instantiated struct file to the caller.
742 * This function is meant to be called from within a filesystem's
743 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800744 * Beware of calling it for non-regular files! Those ->open methods might block
745 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
746 * leading to a deadlock, as nobody can open that fifo anymore, because
747 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700748 * Note that in case of error, nd->intent.open.file is destroyed, but the
749 * path information remains valid.
750 * If the open callback is set to NULL, then the standard f_op->open()
751 * filesystem callback is substituted.
752 */
753struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
754 int (*open)(struct inode *, struct file *))
755{
David Howells745ca242008-11-14 10:39:22 +1100756 const struct cred *cred = current_cred();
757
Trond Myklebust834f2a42005-10-18 14:20:16 -0700758 if (IS_ERR(nd->intent.open.file))
759 goto out;
760 if (IS_ERR(dentry))
761 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800762 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700763 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100764 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700765out:
766 return nd->intent.open.file;
767out_err:
768 release_open_intent(nd);
769 nd->intent.open.file = (struct file *)dentry;
770 goto out;
771}
772EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
773
774/**
775 * nameidata_to_filp - convert a nameidata to an open filp.
776 * @nd: pointer to nameidata
777 * @flags: open flags
778 *
779 * Note that this function destroys the original nameidata
780 */
Al Viro482928d2009-12-19 10:10:39 -0500781struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700782{
David Howells745ca242008-11-14 10:39:22 +1100783 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700784 struct file *filp;
785
786 /* Pick up the filp from the open intent */
787 filp = nd->intent.open.file;
788 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800789 if (filp->f_path.dentry == NULL)
Al Viro482928d2009-12-19 10:10:39 -0500790 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
David Howells745ca242008-11-14 10:39:22 +1100791 NULL, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700792 else
Jan Blunck1d957f92008-02-14 19:34:35 -0800793 path_put(&nd->path);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700794 return filp;
795}
796
Peter Staubach6fdcc212005-11-07 00:59:42 -0800797/*
798 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
799 * error.
800 */
David Howells745ca242008-11-14 10:39:22 +1100801struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
802 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700803{
804 int error;
805 struct file *f;
806
David Howellse0e81732009-09-02 09:13:40 +0100807 validate_creds(cred);
808
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800809 /*
810 * We must always pass in a valid mount pointer. Historically
811 * callers got away with not passing it, but we must enforce this at
812 * the earliest possible point now to avoid strange problems deep in the
813 * filesystem stack.
814 */
815 if (!mnt) {
816 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
817 dump_stack();
818 return ERR_PTR(-EINVAL);
819 }
820
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700821 error = -ENFILE;
822 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800823 if (f == NULL) {
824 dput(dentry);
825 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700826 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800827 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700828
Al Viro482928d2009-12-19 10:10:39 -0500829 f->f_flags = flags;
830 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700831}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832EXPORT_SYMBOL(dentry_open);
833
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800834static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700836 struct fdtable *fdt = files_fdtable(files);
837 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800838 if (fd < files->next_fd)
839 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840}
841
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800842void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843{
844 struct files_struct *files = current->files;
845 spin_lock(&files->file_lock);
846 __put_unused_fd(files, fd);
847 spin_unlock(&files->file_lock);
848}
849
850EXPORT_SYMBOL(put_unused_fd);
851
852/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800853 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 *
855 * The VFS is full of places where we drop the files lock between
856 * setting the open_fds bitmap and installing the file in the file
857 * array. At any such point, we are vulnerable to a dup2() race
858 * installing a file in the array before us. We need to detect this and
859 * fput() the struct file we are about to overwrite in this case.
860 *
861 * It should never happen - if we allow dup2() do it, _really_ bad things
862 * will follow.
863 */
864
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800865void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866{
867 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700868 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700870 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700871 BUG_ON(fdt->fd[fd] != NULL);
872 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 spin_unlock(&files->file_lock);
874}
875
876EXPORT_SYMBOL(fd_install);
877
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800878long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879{
Miklos Szeredie922efc2005-09-06 15:18:25 -0700880 char *tmp = getname(filename);
881 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700884 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 if (fd >= 0) {
Al Viro6e8341a2009-04-06 11:16:22 -0400886 struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700887 if (IS_ERR(f)) {
888 put_unused_fd(fd);
889 fd = PTR_ERR(f);
890 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500891 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700892 fd_install(fd, f);
893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 putname(tmp);
896 }
897 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700899
Heiko Carstensca013e92009-01-14 14:14:19 +0100900SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700901{
Linus Torvalds385910f2006-04-18 13:22:59 -0700902 long ret;
903
Miklos Szeredie922efc2005-09-06 15:18:25 -0700904 if (force_o_largefile())
905 flags |= O_LARGEFILE;
906
Linus Torvalds385910f2006-04-18 13:22:59 -0700907 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
908 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700909 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700910 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -0700911}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
Heiko Carstens6559eed82009-01-14 14:14:32 +0100913SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
914 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800915{
Linus Torvalds385910f2006-04-18 13:22:59 -0700916 long ret;
917
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800918 if (force_o_largefile())
919 flags |= O_LARGEFILE;
920
Linus Torvalds385910f2006-04-18 13:22:59 -0700921 ret = do_sys_open(dfd, filename, flags, mode);
922 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700923 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700924 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800925}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800926
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927#ifndef __alpha__
928
929/*
930 * For backward compatibility? Maybe this should be moved
931 * into arch/i386 instead?
932 */
Heiko Carstens002c8972009-01-14 14:14:18 +0100933SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934{
935 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
936}
937
938#endif
939
940/*
941 * "id" is the POSIX thread ID. We use the
942 * files pointer for this..
943 */
944int filp_close(struct file *filp, fl_owner_t id)
945{
Christoph Lameter45778ca2005-06-23 00:10:17 -0700946 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947
948 if (!file_count(filp)) {
949 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -0700950 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 }
952
Christoph Lameter45778ca2005-06-23 00:10:17 -0700953 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700954 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
956 dnotify_flush(filp, id);
957 locks_remove_posix(filp, id);
958 fput(filp);
959 return retval;
960}
961
962EXPORT_SYMBOL(filp_close);
963
964/*
965 * Careful here! We test whether the file pointer is NULL before
966 * releasing the fd. This ensures that one clone task can't release
967 * an fd while another clone is opening it.
968 */
Heiko Carstensca013e92009-01-14 14:14:19 +0100969SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
971 struct file * filp;
972 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700973 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -0700974 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
976 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700977 fdt = files_fdtable(files);
978 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700980 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 if (!filp)
982 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700983 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700984 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 __put_unused_fd(files, fd);
986 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -0700987 retval = filp_close(filp, files);
988
989 /* can't restart close syscall because file table entry was cleared */
990 if (unlikely(retval == -ERESTARTSYS ||
991 retval == -ERESTARTNOINTR ||
992 retval == -ERESTARTNOHAND ||
993 retval == -ERESTART_RESTARTBLOCK))
994 retval = -EINTR;
995
996 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
998out_unlock:
999 spin_unlock(&files->file_lock);
1000 return -EBADF;
1001}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002EXPORT_SYMBOL(sys_close);
1003
1004/*
1005 * This routine simulates a hangup on the tty, to arrange that users
1006 * are given clean terminals at login time.
1007 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001008SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009{
1010 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001011 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 return 0;
1013 }
1014 return -EPERM;
1015}
1016
1017/*
1018 * Called when an inode is about to be open.
1019 * We use this to disallow opening large files on 32bit systems if
1020 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1021 * on this flag in sys_open.
1022 */
1023int generic_file_open(struct inode * inode, struct file * filp)
1024{
1025 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001026 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 return 0;
1028}
1029
1030EXPORT_SYMBOL(generic_file_open);
1031
1032/*
1033 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001034 * file descriptors. The function is not supposed to ever fail, the only
1035 * reason it returns an 'int' and not 'void' is so that it can be plugged
1036 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 */
1038int nonseekable_open(struct inode *inode, struct file *filp)
1039{
1040 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1041 return 0;
1042}
1043
1044EXPORT_SYMBOL(nonseekable_open);