blob: 22c41b543f2dfc00ce74665e4fdf24db1d8321ac [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 */
Josef Bacik79124f12010-11-17 20:46:15 -0500226 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
227 return -EOPNOTSUPP;
228
229 /* Punch hole must have keep size set */
230 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
231 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400232 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400233
Amit Arora97ac7352007-07-17 21:42:44 -0400234 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400235 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100236
237 /* It's not possible punch hole on append only file */
238 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
239 return -EPERM;
240
241 if (IS_IMMUTABLE(inode))
242 return -EPERM;
243
Amit Arora97ac7352007-07-17 21:42:44 -0400244 /*
245 * Revalidate the write permissions, in case security policy has
246 * changed since the files were opened.
247 */
248 ret = security_file_permission(file, MAY_WRITE);
249 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400250 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400251
Amit Arora97ac7352007-07-17 21:42:44 -0400252 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400253 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400254
Amit Arora97ac7352007-07-17 21:42:44 -0400255 /*
256 * Let individual file system decide if it supports preallocation
257 * for directories or not.
258 */
259 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400260 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400261
Amit Arora97ac7352007-07-17 21:42:44 -0400262 /* Check for wrap through zero too */
263 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400264 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400265
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100266 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400267 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400268
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100269 return file->f_op->fallocate(file, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400270}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400271
272SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
273{
274 struct file *file;
275 int error = -EBADF;
276
277 file = fget(fd);
278 if (file) {
279 error = do_fallocate(file, mode, offset, len);
280 fput(file);
281 }
282
283 return error;
284}
285
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100286#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
287asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
288{
289 return SYSC_fallocate((int)fd, (int)mode, offset, len);
290}
291SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
292#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294/*
295 * access() needs to use the real uid/gid, not the effective uid/gid.
296 * We do this by temporarily clearing all FS-related capabilities and
297 * switching the fsuid/fsgid around to the real ones.
298 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100299SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300{
David Howellsd84f4f92008-11-14 10:39:23 +1100301 const struct cred *old_cred;
302 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400303 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400304 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 int res;
306
307 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
308 return -EINVAL;
309
David Howellsd84f4f92008-11-14 10:39:23 +1100310 override_cred = prepare_creds();
311 if (!override_cred)
312 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
David Howellsd84f4f92008-11-14 10:39:23 +1100314 override_cred->fsuid = override_cred->uid;
315 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700317 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100318 /* Clear the capabilities if we switch to a non-root user */
David Howellsd84f4f92008-11-14 10:39:23 +1100319 if (override_cred->uid)
320 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700321 else
David Howellsd84f4f92008-11-14 10:39:23 +1100322 override_cred->cap_effective =
323 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700324 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
David Howellsd84f4f92008-11-14 10:39:23 +1100326 old_cred = override_creds(override_cred);
327
Al Viro2d8f3032008-07-22 09:59:21 -0400328 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700329 if (res)
330 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Al Viro2d8f3032008-07-22 09:59:21 -0400332 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400333
334 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400335 /*
336 * MAY_EXEC on regular files is denied if the fs is mounted
337 * with the "noexec" flag.
338 */
339 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400340 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400341 goto out_path_release;
342 }
343
Al Viro256984a2008-07-22 08:09:30 -0400344 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700345 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400346 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700347 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800348 /*
349 * This is a rare case where using __mnt_is_readonly()
350 * is OK without a mnt_want/drop_write() pair. Since
351 * no actual write to the fs is performed here, we do
352 * not need to telegraph to that to anyone.
353 *
354 * By doing this, we accept that this access is
355 * inherently racy and know that the fs may change
356 * state before we even see this result.
357 */
Al Viro2d8f3032008-07-22 09:59:21 -0400358 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700359 res = -EROFS;
360
361out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400362 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700363out:
David Howellsd84f4f92008-11-14 10:39:23 +1100364 revert_creds(old_cred);
365 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 return res;
367}
368
Heiko Carstensca013e92009-01-14 14:14:19 +0100369SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800370{
371 return sys_faccessat(AT_FDCWD, filename, mode);
372}
373
Heiko Carstens3cdad422009-01-14 14:14:22 +0100374SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375{
Al Viro2d8f3032008-07-22 09:59:21 -0400376 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 int error;
378
Al Viro2d8f3032008-07-22 09:59:21 -0400379 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 if (error)
381 goto out;
382
Eric Paris9cfcac82010-07-23 11:43:51 -0400383 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 if (error)
385 goto dput_and_out;
386
Al Viro2d8f3032008-07-22 09:59:21 -0400387 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400390 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391out:
392 return error;
393}
394
Heiko Carstens3cdad422009-01-14 14:14:22 +0100395SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396{
397 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 int error;
400
401 error = -EBADF;
402 file = fget(fd);
403 if (!file)
404 goto out;
405
Jan Blunckac748a02008-02-14 19:34:39 -0800406 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408 error = -ENOTDIR;
409 if (!S_ISDIR(inode->i_mode))
410 goto out_putf;
411
Eric Paris9cfcac82010-07-23 11:43:51 -0400412 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800414 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415out_putf:
416 fput(file);
417out:
418 return error;
419}
420
Heiko Carstens3480b252009-01-14 14:14:16 +0100421SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422{
Al Viro2d8f3032008-07-22 09:59:21 -0400423 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 int error;
425
Al Viro2d8f3032008-07-22 09:59:21 -0400426 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 if (error)
428 goto out;
429
Eric Paris9cfcac82010-07-23 11:43:51 -0400430 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 if (error)
432 goto dput_and_out;
433
434 error = -EPERM;
435 if (!capable(CAP_SYS_CHROOT))
436 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900437 error = security_path_chroot(&path);
438 if (error)
439 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Al Viro2d8f3032008-07-22 09:59:21 -0400441 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 error = 0;
443dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400444 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445out:
446 return error;
447}
448
Al Viroe57712e2011-07-26 04:15:54 -0400449static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450{
Al Viroe57712e2011-07-26 04:15:54 -0400451 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400453 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
Al Viroe57712e2011-07-26 04:15:54 -0400455 error = mnt_want_write(path->mnt);
456 if (error)
457 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900458 mutex_lock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400459 error = security_path_chmod(path->dentry, path->mnt, mode);
460 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900461 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
463 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400464 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900465out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800466 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400467 mnt_drop_write(path->mnt);
468 return error;
469}
470
471SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
472{
473 struct file * file;
474 int err = -EBADF;
475
476 file = fget(fd);
477 if (file) {
478 audit_inode(NULL, file->f_path.dentry);
479 err = chmod_common(&file->f_path, mode);
480 fput(file);
481 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 return err;
483}
484
Heiko Carstens6559eed82009-01-14 14:14:32 +0100485SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486{
Al Viro2d8f3032008-07-22 09:59:21 -0400487 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
Al Viro2d8f3032008-07-22 09:59:21 -0400490 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400491 if (!error) {
492 error = chmod_common(&path, mode);
493 path_put(&path);
494 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 return error;
496}
497
Heiko Carstensa26eab22009-01-14 14:14:17 +0100498SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800499{
500 return sys_fchmodat(AT_FDCWD, filename, mode);
501}
502
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900503static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900505 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 int error;
507 struct iattr newattrs;
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 newattrs.ia_valid = ATTR_CTIME;
510 if (user != (uid_t) -1) {
511 newattrs.ia_valid |= ATTR_UID;
512 newattrs.ia_uid = user;
513 }
514 if (group != (gid_t) -1) {
515 newattrs.ia_valid |= ATTR_GID;
516 newattrs.ia_gid = group;
517 }
518 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700519 newattrs.ia_valid |=
520 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800521 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900522 error = security_path_chown(path, user, group);
523 if (!error)
524 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800525 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 return error;
528}
529
Heiko Carstensca013e92009-01-14 14:14:19 +0100530SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
Al Viro2d8f3032008-07-22 09:59:21 -0400532 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 int error;
534
Al Viro2d8f3032008-07-22 09:59:21 -0400535 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700536 if (error)
537 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400538 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800539 if (error)
540 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900541 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400542 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800543out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400544 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700545out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 return error;
547}
548
Heiko Carstens6559eed82009-01-14 14:14:32 +0100549SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
550 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800551{
Al Viro2d8f3032008-07-22 09:59:21 -0400552 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800553 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400554 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800555
Al Viro65cfc672011-03-13 15:56:26 -0400556 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800557 goto out;
558
Al Viro65cfc672011-03-13 15:56:26 -0400559 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
560 if (flag & AT_EMPTY_PATH)
561 lookup_flags |= LOOKUP_EMPTY;
562 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700563 if (error)
564 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400565 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800566 if (error)
567 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900568 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400569 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800570out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400571 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800572out:
573 return error;
574}
575
Heiko Carstensca013e92009-01-14 14:14:19 +0100576SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577{
Al Viro2d8f3032008-07-22 09:59:21 -0400578 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 int error;
580
Al Viro2d8f3032008-07-22 09:59:21 -0400581 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700582 if (error)
583 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400584 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800585 if (error)
586 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900587 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400588 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800589out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400590 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700591out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 return error;
593}
594
Heiko Carstensca013e92009-01-14 14:14:19 +0100595SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596{
597 struct file * file;
598 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700599 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700602 if (!file)
603 goto out;
604
npiggin@suse.de96029c42009-04-26 20:25:55 +1000605 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800606 if (error)
607 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800608 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400609 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900610 error = chown_common(&file->f_path, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800611 mnt_drop_write(file->f_path.mnt);
612out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700613 fput(file);
614out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 return error;
616}
617
Dave Hansen4a3fd212008-02-15 14:37:48 -0800618/*
619 * You have to be very careful that these write
620 * counts get cleaned up in error cases and
621 * upon __fput(). This should probably never
622 * be called outside of __dentry_open().
623 */
624static inline int __get_file_write_access(struct inode *inode,
625 struct vfsmount *mnt)
626{
627 int error;
628 error = get_write_access(inode);
629 if (error)
630 return error;
631 /*
632 * Do not take mount writer counts on
633 * special files since no writes to
634 * the mount itself will occur.
635 */
636 if (!special_file(inode->i_mode)) {
637 /*
638 * Balanced in __fput()
639 */
640 error = mnt_want_write(mnt);
641 if (error)
642 put_write_access(inode);
643 }
644 return error;
645}
646
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700647static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Al Viro482928d2009-12-19 10:10:39 -0500648 struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100649 int (*open)(struct inode *, struct file *),
650 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651{
Al Viro1abf0c72011-03-13 03:51:11 -0400652 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 struct inode *inode;
654 int error;
655
Al Viro53009902009-12-19 10:15:07 -0500656 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700657 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400658
659 if (unlikely(f->f_flags & O_PATH))
660 f->f_mode = FMODE_PATH;
661
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 inode = dentry->d_inode;
663 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800664 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 if (error)
666 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800667 if (!special_file(inode->i_mode))
668 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 }
670
671 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800672 f->f_path.dentry = dentry;
673 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 f->f_pos = 0;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000675 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Al Viro1abf0c72011-03-13 03:51:11 -0400677 if (unlikely(f->f_mode & FMODE_PATH)) {
678 f->f_op = &empty_fops;
679 return f;
680 }
681
682 f->f_op = fops_get(inode->i_fop);
683
David Howells745ca242008-11-14 10:39:22 +1100684 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900685 if (error)
686 goto cleanup_all;
687
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400688 error = break_lease(inode, f->f_flags);
689 if (error)
690 goto cleanup_all;
691
Trond Myklebust834f2a42005-10-18 14:20:16 -0700692 if (!open && f->f_op)
693 open = f->f_op->open;
694 if (open) {
695 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 if (error)
697 goto cleanup_all;
698 }
Mimi Zohar890275b2010-11-02 10:13:07 -0400699 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
700 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
703
704 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
705
706 /* NB: we're sure to have correct a_ops only after f_op->open */
707 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700708 if (!f->f_mapping->a_ops ||
709 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700710 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 fput(f);
712 f = ERR_PTR(-EINVAL);
713 }
714 }
715
716 return f;
717
718cleanup_all:
719 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800720 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800722 if (!special_file(inode->i_mode)) {
723 /*
724 * We don't consider this a real
725 * mnt_want/drop_write() pair
726 * because it all happenend right
727 * here, so just reset the state.
728 */
729 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800730 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800731 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800732 }
Nick Pigginee2ffa02010-08-18 04:37:35 +1000733 file_sb_list_del(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800734 f->f_path.dentry = NULL;
735 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736cleanup_file:
737 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 dput(dentry);
739 mntput(mnt);
740 return ERR_PTR(error);
741}
742
Trond Myklebust834f2a42005-10-18 14:20:16 -0700743/**
744 * lookup_instantiate_filp - instantiates the open intent filp
745 * @nd: pointer to nameidata
746 * @dentry: pointer to dentry
747 * @open: open callback
748 *
749 * Helper for filesystems that want to use lookup open intents and pass back
750 * a fully instantiated struct file to the caller.
751 * This function is meant to be called from within a filesystem's
752 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800753 * Beware of calling it for non-regular files! Those ->open methods might block
754 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
755 * leading to a deadlock, as nobody can open that fifo anymore, because
756 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700757 * Note that in case of error, nd->intent.open.file is destroyed, but the
758 * path information remains valid.
759 * If the open callback is set to NULL, then the standard f_op->open()
760 * filesystem callback is substituted.
761 */
762struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
763 int (*open)(struct inode *, struct file *))
764{
David Howells745ca242008-11-14 10:39:22 +1100765 const struct cred *cred = current_cred();
766
Trond Myklebust834f2a42005-10-18 14:20:16 -0700767 if (IS_ERR(nd->intent.open.file))
768 goto out;
769 if (IS_ERR(dentry))
770 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800771 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700772 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100773 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700774out:
775 return nd->intent.open.file;
776out_err:
777 release_open_intent(nd);
Konstantin Khlebnikov5a9a4362011-07-17 15:35:12 +0400778 nd->intent.open.file = ERR_CAST(dentry);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700779 goto out;
780}
781EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
782
783/**
784 * nameidata_to_filp - convert a nameidata to an open filp.
785 * @nd: pointer to nameidata
786 * @flags: open flags
787 *
788 * Note that this function destroys the original nameidata
789 */
Al Viro482928d2009-12-19 10:10:39 -0500790struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700791{
David Howells745ca242008-11-14 10:39:22 +1100792 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700793 struct file *filp;
794
795 /* Pick up the filp from the open intent */
796 filp = nd->intent.open.file;
Linus Torvalds2dab5972011-02-11 15:53:38 -0800797 nd->intent.open.file = NULL;
798
Trond Myklebust834f2a42005-10-18 14:20:16 -0700799 /* Has the filesystem initialised the file for us? */
Al Virod893f1b2010-10-29 03:30:42 -0400800 if (filp->f_path.dentry == NULL) {
801 path_get(&nd->path);
Al Viro482928d2009-12-19 10:10:39 -0500802 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
David Howells745ca242008-11-14 10:39:22 +1100803 NULL, cred);
Al Virod893f1b2010-10-29 03:30:42 -0400804 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700805 return filp;
806}
807
Peter Staubach6fdcc212005-11-07 00:59:42 -0800808/*
809 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
810 * error.
811 */
David Howells745ca242008-11-14 10:39:22 +1100812struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
813 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700814{
815 int error;
816 struct file *f;
817
David Howellse0e81732009-09-02 09:13:40 +0100818 validate_creds(cred);
819
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900820 /* We must always pass in a valid mount pointer. */
821 BUG_ON(!mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800822
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700823 error = -ENFILE;
824 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800825 if (f == NULL) {
826 dput(dentry);
827 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700828 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800829 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700830
Al Viro482928d2009-12-19 10:10:39 -0500831 f->f_flags = flags;
832 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700833}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834EXPORT_SYMBOL(dentry_open);
835
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800836static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700838 struct fdtable *fdt = files_fdtable(files);
839 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800840 if (fd < files->next_fd)
841 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842}
843
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800844void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 struct files_struct *files = current->files;
847 spin_lock(&files->file_lock);
848 __put_unused_fd(files, fd);
849 spin_unlock(&files->file_lock);
850}
851
852EXPORT_SYMBOL(put_unused_fd);
853
854/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800855 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 *
857 * The VFS is full of places where we drop the files lock between
858 * setting the open_fds bitmap and installing the file in the file
859 * array. At any such point, we are vulnerable to a dup2() race
860 * installing a file in the array before us. We need to detect this and
861 * fput() the struct file we are about to overwrite in this case.
862 *
863 * It should never happen - if we allow dup2() do it, _really_ bad things
864 * will follow.
865 */
866
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800867void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
869 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700870 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700872 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700873 BUG_ON(fdt->fd[fd] != NULL);
874 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 spin_unlock(&files->file_lock);
876}
877
878EXPORT_SYMBOL(fd_install);
879
Al Viro47c805d2011-02-23 17:44:09 -0500880static inline int build_open_flags(int flags, int mode, struct open_flags *op)
881{
882 int lookup_flags = 0;
883 int acc_mode;
884
885 if (!(flags & O_CREAT))
886 mode = 0;
887 op->mode = mode;
888
889 /* Must never be set by userspace */
890 flags &= ~FMODE_NONOTIFY;
891
892 /*
893 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
894 * check for O_DSYNC if the need any syncing at all we enforce it's
895 * always set instead of having to deal with possibly weird behaviour
896 * for malicious applications setting only __O_SYNC.
897 */
898 if (flags & __O_SYNC)
899 flags |= O_DSYNC;
900
Al Viro1abf0c72011-03-13 03:51:11 -0400901 /*
902 * If we have O_PATH in the open flag. Then we
903 * cannot have anything other than the below set of flags
904 */
905 if (flags & O_PATH) {
906 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
907 acc_mode = 0;
908 } else {
909 acc_mode = MAY_OPEN | ACC_MODE(flags);
910 }
Al Viro47c805d2011-02-23 17:44:09 -0500911
Al Viro1abf0c72011-03-13 03:51:11 -0400912 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500913
914 /* O_TRUNC implies we need access checks for write permissions */
915 if (flags & O_TRUNC)
916 acc_mode |= MAY_WRITE;
917
918 /* Allow the LSM permission hook to distinguish append
919 access from general write access. */
920 if (flags & O_APPEND)
921 acc_mode |= MAY_APPEND;
922
923 op->acc_mode = acc_mode;
924
Al Viro1abf0c72011-03-13 03:51:11 -0400925 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
926
Al Viro47c805d2011-02-23 17:44:09 -0500927 if (flags & O_CREAT) {
928 op->intent |= LOOKUP_CREATE;
929 if (flags & O_EXCL)
930 op->intent |= LOOKUP_EXCL;
931 }
932
933 if (flags & O_DIRECTORY)
934 lookup_flags |= LOOKUP_DIRECTORY;
935 if (!(flags & O_NOFOLLOW))
936 lookup_flags |= LOOKUP_FOLLOW;
937 return lookup_flags;
938}
939
940/**
941 * filp_open - open file and return file pointer
942 *
943 * @filename: path to open
944 * @flags: open flags as per the open(2) second argument
945 * @mode: mode for the new file if O_CREAT is set, else ignored
946 *
947 * This is the helper to open a file from kernelspace if you really
948 * have to. But in generally you should not do this, so please move
949 * along, nothing to see here..
950 */
951struct file *filp_open(const char *filename, int flags, int mode)
952{
953 struct open_flags op;
954 int lookup = build_open_flags(flags, mode, &op);
955 return do_filp_open(AT_FDCWD, filename, &op, lookup);
956}
957EXPORT_SYMBOL(filp_open);
958
Al Viro73d049a2011-03-11 12:08:24 -0500959struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
960 const char *filename, int flags)
961{
962 struct open_flags op;
963 int lookup = build_open_flags(flags, 0, &op);
964 if (flags & O_CREAT)
965 return ERR_PTR(-EINVAL);
966 if (!filename && (flags & O_DIRECTORY))
967 if (!dentry->d_inode->i_op->lookup)
968 return ERR_PTR(-ENOTDIR);
969 return do_file_open_root(dentry, mnt, filename, &op, lookup);
970}
971EXPORT_SYMBOL(file_open_root);
972
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800973long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974{
Al Viro47c805d2011-02-23 17:44:09 -0500975 struct open_flags op;
976 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -0700977 char *tmp = getname(filename);
978 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700981 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -0500983 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700984 if (IS_ERR(f)) {
985 put_unused_fd(fd);
986 fd = PTR_ERR(f);
987 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500988 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700989 fd_install(fd, f);
990 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 putname(tmp);
993 }
994 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700996
Heiko Carstensca013e92009-01-14 14:14:19 +0100997SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700998{
Linus Torvalds385910f2006-04-18 13:22:59 -0700999 long ret;
1000
Miklos Szeredie922efc2005-09-06 15:18:25 -07001001 if (force_o_largefile())
1002 flags |= O_LARGEFILE;
1003
Linus Torvalds385910f2006-04-18 13:22:59 -07001004 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1005 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001006 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001007 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001008}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
Heiko Carstens6559eed82009-01-14 14:14:32 +01001010SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1011 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001012{
Linus Torvalds385910f2006-04-18 13:22:59 -07001013 long ret;
1014
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001015 if (force_o_largefile())
1016 flags |= O_LARGEFILE;
1017
Linus Torvalds385910f2006-04-18 13:22:59 -07001018 ret = do_sys_open(dfd, filename, flags, mode);
1019 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001020 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001021 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001022}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001023
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024#ifndef __alpha__
1025
1026/*
1027 * For backward compatibility? Maybe this should be moved
1028 * into arch/i386 instead?
1029 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001030SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031{
1032 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1033}
1034
1035#endif
1036
1037/*
1038 * "id" is the POSIX thread ID. We use the
1039 * files pointer for this..
1040 */
1041int filp_close(struct file *filp, fl_owner_t id)
1042{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001043 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
1045 if (!file_count(filp)) {
1046 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001047 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 }
1049
Christoph Lameter45778ca2005-06-23 00:10:17 -07001050 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001051 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
Al Viro1abf0c72011-03-13 03:51:11 -04001053 if (likely(!(filp->f_mode & FMODE_PATH))) {
1054 dnotify_flush(filp, id);
1055 locks_remove_posix(filp, id);
1056 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 fput(filp);
1058 return retval;
1059}
1060
1061EXPORT_SYMBOL(filp_close);
1062
1063/*
1064 * Careful here! We test whether the file pointer is NULL before
1065 * releasing the fd. This ensures that one clone task can't release
1066 * an fd while another clone is opening it.
1067 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001068SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069{
1070 struct file * filp;
1071 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001072 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001073 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074
1075 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001076 fdt = files_fdtable(files);
1077 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001079 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 if (!filp)
1081 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001082 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001083 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 __put_unused_fd(files, fd);
1085 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001086 retval = filp_close(filp, files);
1087
1088 /* can't restart close syscall because file table entry was cleared */
1089 if (unlikely(retval == -ERESTARTSYS ||
1090 retval == -ERESTARTNOINTR ||
1091 retval == -ERESTARTNOHAND ||
1092 retval == -ERESTART_RESTARTBLOCK))
1093 retval = -EINTR;
1094
1095 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
1097out_unlock:
1098 spin_unlock(&files->file_lock);
1099 return -EBADF;
1100}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101EXPORT_SYMBOL(sys_close);
1102
1103/*
1104 * This routine simulates a hangup on the tty, to arrange that users
1105 * are given clean terminals at login time.
1106 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001107SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108{
1109 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001110 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 return 0;
1112 }
1113 return -EPERM;
1114}
1115
1116/*
1117 * Called when an inode is about to be open.
1118 * We use this to disallow opening large files on 32bit systems if
1119 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1120 * on this flag in sys_open.
1121 */
1122int generic_file_open(struct inode * inode, struct file * filp)
1123{
1124 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001125 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 return 0;
1127}
1128
1129EXPORT_SYMBOL(generic_file_open);
1130
1131/*
1132 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001133 * file descriptors. The function is not supposed to ever fail, the only
1134 * reason it returns an 'int' and not 'void' is so that it can be plugged
1135 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 */
1137int nonseekable_open(struct inode *inode, struct file *filp)
1138{
1139 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1140 return 0;
1141}
1142
1143EXPORT_SYMBOL(nonseekable_open);