blob: f83ca80cc59a1d72c8b933d7562efc853cfe3f8c [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 Bacik79124f182010-11-17 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
Heiko Carstensa26eab22009-01-14 14:14:17 +0100449SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450{
451 struct inode * inode;
452 struct dentry * dentry;
453 struct file * file;
454 int err = -EBADF;
455 struct iattr newattrs;
456
457 file = fget(fd);
458 if (!file)
459 goto out;
460
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800461 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 inode = dentry->d_inode;
463
Al Viro5a190ae2007-06-07 12:19:32 -0400464 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000465
npiggin@suse.de96029c42009-04-26 20:25:55 +1000466 err = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800467 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 goto out_putf;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900469 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900470 err = security_path_chmod(dentry, file->f_vfsmnt, mode);
471 if (err)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900472 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 if (mode == (mode_t) -1)
474 mode = inode->i_mode;
475 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
476 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
477 err = notify_change(dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900478out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800479 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800480 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481out_putf:
482 fput(file);
483out:
484 return err;
485}
486
Heiko Carstens6559eed82009-01-14 14:14:32 +0100487SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488{
Al Viro2d8f3032008-07-22 09:59:21 -0400489 struct path path;
490 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 int error;
492 struct iattr newattrs;
493
Al Viro2d8f3032008-07-22 09:59:21 -0400494 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 if (error)
496 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400497 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Al Viro2d8f3032008-07-22 09:59:21 -0400499 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800500 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 goto dput_and_out;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900502 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900503 error = security_path_chmod(path.dentry, path.mnt, mode);
504 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900505 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 if (mode == (mode_t) -1)
507 mode = inode->i_mode;
508 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
509 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400510 error = notify_change(path.dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900511out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800512 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400513 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400515 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516out:
517 return error;
518}
519
Heiko Carstensa26eab22009-01-14 14:14:17 +0100520SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800521{
522 return sys_fchmodat(AT_FDCWD, filename, mode);
523}
524
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900525static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900527 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 int error;
529 struct iattr newattrs;
530
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 newattrs.ia_valid = ATTR_CTIME;
532 if (user != (uid_t) -1) {
533 newattrs.ia_valid |= ATTR_UID;
534 newattrs.ia_uid = user;
535 }
536 if (group != (gid_t) -1) {
537 newattrs.ia_valid |= ATTR_GID;
538 newattrs.ia_gid = group;
539 }
540 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700541 newattrs.ia_valid |=
542 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800543 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900544 error = security_path_chown(path, user, group);
545 if (!error)
546 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800547 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200548
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 return error;
550}
551
Heiko Carstensca013e92009-01-14 14:14:19 +0100552SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
Al Viro2d8f3032008-07-22 09:59:21 -0400554 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 int error;
556
Al Viro2d8f3032008-07-22 09:59:21 -0400557 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700558 if (error)
559 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400560 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800561 if (error)
562 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900563 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400564 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800565out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400566 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700567out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 return error;
569}
570
Heiko Carstens6559eed82009-01-14 14:14:32 +0100571SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
572 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800573{
Al Viro2d8f3032008-07-22 09:59:21 -0400574 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800575 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400576 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800577
Al Viro65cfc672011-03-13 15:56:26 -0400578 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800579 goto out;
580
Al Viro65cfc672011-03-13 15:56:26 -0400581 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
582 if (flag & AT_EMPTY_PATH)
583 lookup_flags |= LOOKUP_EMPTY;
584 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700585 if (error)
586 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400587 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800588 if (error)
589 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900590 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400591 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800592out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400593 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800594out:
595 return error;
596}
597
Heiko Carstensca013e92009-01-14 14:14:19 +0100598SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
Al Viro2d8f3032008-07-22 09:59:21 -0400600 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 int error;
602
Al Viro2d8f3032008-07-22 09:59:21 -0400603 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700604 if (error)
605 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400606 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800607 if (error)
608 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900609 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400610 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800611out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400612 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700613out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 return error;
615}
616
Heiko Carstensca013e92009-01-14 14:14:19 +0100617SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618{
619 struct file * file;
620 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700621 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700624 if (!file)
625 goto out;
626
npiggin@suse.de96029c42009-04-26 20:25:55 +1000627 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800628 if (error)
629 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800630 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400631 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900632 error = chown_common(&file->f_path, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800633 mnt_drop_write(file->f_path.mnt);
634out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700635 fput(file);
636out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 return error;
638}
639
Dave Hansen4a3fd212008-02-15 14:37:48 -0800640/*
641 * You have to be very careful that these write
642 * counts get cleaned up in error cases and
643 * upon __fput(). This should probably never
644 * be called outside of __dentry_open().
645 */
646static inline int __get_file_write_access(struct inode *inode,
647 struct vfsmount *mnt)
648{
649 int error;
650 error = get_write_access(inode);
651 if (error)
652 return error;
653 /*
654 * Do not take mount writer counts on
655 * special files since no writes to
656 * the mount itself will occur.
657 */
658 if (!special_file(inode->i_mode)) {
659 /*
660 * Balanced in __fput()
661 */
662 error = mnt_want_write(mnt);
663 if (error)
664 put_write_access(inode);
665 }
666 return error;
667}
668
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700669static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Al Viro482928d2009-12-19 10:10:39 -0500670 struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100671 int (*open)(struct inode *, struct file *),
672 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673{
Al Viro1abf0c72011-03-13 03:51:11 -0400674 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 struct inode *inode;
676 int error;
677
Al Viro53009902009-12-19 10:15:07 -0500678 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700679 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400680
681 if (unlikely(f->f_flags & O_PATH))
682 f->f_mode = FMODE_PATH;
683
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 inode = dentry->d_inode;
685 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800686 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 if (error)
688 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800689 if (!special_file(inode->i_mode))
690 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 }
692
693 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800694 f->f_path.dentry = dentry;
695 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 f->f_pos = 0;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000697 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Al Viro1abf0c72011-03-13 03:51:11 -0400699 if (unlikely(f->f_mode & FMODE_PATH)) {
700 f->f_op = &empty_fops;
701 return f;
702 }
703
704 f->f_op = fops_get(inode->i_fop);
705
David Howells745ca242008-11-14 10:39:22 +1100706 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900707 if (error)
708 goto cleanup_all;
709
Trond Myklebust834f2a42005-10-18 14:20:16 -0700710 if (!open && f->f_op)
711 open = f->f_op->open;
712 if (open) {
713 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 if (error)
715 goto cleanup_all;
716 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400717 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
718 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
721
722 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
723
724 /* NB: we're sure to have correct a_ops only after f_op->open */
725 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700726 if (!f->f_mapping->a_ops ||
727 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700728 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 fput(f);
730 f = ERR_PTR(-EINVAL);
731 }
732 }
733
734 return f;
735
736cleanup_all:
737 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800738 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800740 if (!special_file(inode->i_mode)) {
741 /*
742 * We don't consider this a real
743 * mnt_want/drop_write() pair
744 * because it all happenend right
745 * here, so just reset the state.
746 */
747 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800748 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800749 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800750 }
Nick Pigginee2ffa02010-08-18 04:37:35 +1000751 file_sb_list_del(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800752 f->f_path.dentry = NULL;
753 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754cleanup_file:
755 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 dput(dentry);
757 mntput(mnt);
758 return ERR_PTR(error);
759}
760
Trond Myklebust834f2a42005-10-18 14:20:16 -0700761/**
762 * lookup_instantiate_filp - instantiates the open intent filp
763 * @nd: pointer to nameidata
764 * @dentry: pointer to dentry
765 * @open: open callback
766 *
767 * Helper for filesystems that want to use lookup open intents and pass back
768 * a fully instantiated struct file to the caller.
769 * This function is meant to be called from within a filesystem's
770 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800771 * Beware of calling it for non-regular files! Those ->open methods might block
772 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
773 * leading to a deadlock, as nobody can open that fifo anymore, because
774 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700775 * Note that in case of error, nd->intent.open.file is destroyed, but the
776 * path information remains valid.
777 * If the open callback is set to NULL, then the standard f_op->open()
778 * filesystem callback is substituted.
779 */
780struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
781 int (*open)(struct inode *, struct file *))
782{
David Howells745ca242008-11-14 10:39:22 +1100783 const struct cred *cred = current_cred();
784
Trond Myklebust834f2a42005-10-18 14:20:16 -0700785 if (IS_ERR(nd->intent.open.file))
786 goto out;
787 if (IS_ERR(dentry))
788 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800789 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700790 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100791 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700792out:
793 return nd->intent.open.file;
794out_err:
795 release_open_intent(nd);
796 nd->intent.open.file = (struct file *)dentry;
797 goto out;
798}
799EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
800
801/**
802 * nameidata_to_filp - convert a nameidata to an open filp.
803 * @nd: pointer to nameidata
804 * @flags: open flags
805 *
806 * Note that this function destroys the original nameidata
807 */
Al Viro482928d2009-12-19 10:10:39 -0500808struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700809{
David Howells745ca242008-11-14 10:39:22 +1100810 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700811 struct file *filp;
812
813 /* Pick up the filp from the open intent */
814 filp = nd->intent.open.file;
Linus Torvalds2dab5972011-02-11 15:53:38 -0800815 nd->intent.open.file = NULL;
816
Trond Myklebust834f2a42005-10-18 14:20:16 -0700817 /* Has the filesystem initialised the file for us? */
Al Virod893f1b2010-10-29 03:30:42 -0400818 if (filp->f_path.dentry == NULL) {
819 path_get(&nd->path);
Al Viro482928d2009-12-19 10:10:39 -0500820 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
David Howells745ca242008-11-14 10:39:22 +1100821 NULL, cred);
Al Virod893f1b2010-10-29 03:30:42 -0400822 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700823 return filp;
824}
825
Peter Staubach6fdcc212005-11-07 00:59:42 -0800826/*
827 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
828 * error.
829 */
David Howells745ca242008-11-14 10:39:22 +1100830struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
831 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700832{
833 int error;
834 struct file *f;
835
David Howellse0e81732009-09-02 09:13:40 +0100836 validate_creds(cred);
837
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800838 /*
839 * We must always pass in a valid mount pointer. Historically
840 * callers got away with not passing it, but we must enforce this at
841 * the earliest possible point now to avoid strange problems deep in the
842 * filesystem stack.
843 */
844 if (!mnt) {
845 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
846 dump_stack();
847 return ERR_PTR(-EINVAL);
848 }
849
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700850 error = -ENFILE;
851 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800852 if (f == NULL) {
853 dput(dentry);
854 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700855 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800856 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700857
Al Viro482928d2009-12-19 10:10:39 -0500858 f->f_flags = flags;
859 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700860}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861EXPORT_SYMBOL(dentry_open);
862
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800863static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700865 struct fdtable *fdt = files_fdtable(files);
866 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800867 if (fd < files->next_fd)
868 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869}
870
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800871void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
873 struct files_struct *files = current->files;
874 spin_lock(&files->file_lock);
875 __put_unused_fd(files, fd);
876 spin_unlock(&files->file_lock);
877}
878
879EXPORT_SYMBOL(put_unused_fd);
880
881/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800882 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 *
884 * The VFS is full of places where we drop the files lock between
885 * setting the open_fds bitmap and installing the file in the file
886 * array. At any such point, we are vulnerable to a dup2() race
887 * installing a file in the array before us. We need to detect this and
888 * fput() the struct file we are about to overwrite in this case.
889 *
890 * It should never happen - if we allow dup2() do it, _really_ bad things
891 * will follow.
892 */
893
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800894void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
896 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700897 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700899 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700900 BUG_ON(fdt->fd[fd] != NULL);
901 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 spin_unlock(&files->file_lock);
903}
904
905EXPORT_SYMBOL(fd_install);
906
Al Viro47c805d2011-02-23 17:44:09 -0500907static inline int build_open_flags(int flags, int mode, struct open_flags *op)
908{
909 int lookup_flags = 0;
910 int acc_mode;
911
912 if (!(flags & O_CREAT))
913 mode = 0;
914 op->mode = mode;
915
916 /* Must never be set by userspace */
917 flags &= ~FMODE_NONOTIFY;
918
919 /*
920 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
921 * check for O_DSYNC if the need any syncing at all we enforce it's
922 * always set instead of having to deal with possibly weird behaviour
923 * for malicious applications setting only __O_SYNC.
924 */
925 if (flags & __O_SYNC)
926 flags |= O_DSYNC;
927
Al Viro1abf0c72011-03-13 03:51:11 -0400928 /*
929 * If we have O_PATH in the open flag. Then we
930 * cannot have anything other than the below set of flags
931 */
932 if (flags & O_PATH) {
933 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
934 acc_mode = 0;
935 } else {
936 acc_mode = MAY_OPEN | ACC_MODE(flags);
937 }
Al Viro47c805d2011-02-23 17:44:09 -0500938
Al Viro1abf0c72011-03-13 03:51:11 -0400939 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500940
941 /* O_TRUNC implies we need access checks for write permissions */
942 if (flags & O_TRUNC)
943 acc_mode |= MAY_WRITE;
944
945 /* Allow the LSM permission hook to distinguish append
946 access from general write access. */
947 if (flags & O_APPEND)
948 acc_mode |= MAY_APPEND;
949
950 op->acc_mode = acc_mode;
951
Al Viro1abf0c72011-03-13 03:51:11 -0400952 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
953
Al Viro47c805d2011-02-23 17:44:09 -0500954 if (flags & O_CREAT) {
955 op->intent |= LOOKUP_CREATE;
956 if (flags & O_EXCL)
957 op->intent |= LOOKUP_EXCL;
958 }
959
960 if (flags & O_DIRECTORY)
961 lookup_flags |= LOOKUP_DIRECTORY;
962 if (!(flags & O_NOFOLLOW))
963 lookup_flags |= LOOKUP_FOLLOW;
964 return lookup_flags;
965}
966
967/**
968 * filp_open - open file and return file pointer
969 *
970 * @filename: path to open
971 * @flags: open flags as per the open(2) second argument
972 * @mode: mode for the new file if O_CREAT is set, else ignored
973 *
974 * This is the helper to open a file from kernelspace if you really
975 * have to. But in generally you should not do this, so please move
976 * along, nothing to see here..
977 */
978struct file *filp_open(const char *filename, int flags, int mode)
979{
980 struct open_flags op;
981 int lookup = build_open_flags(flags, mode, &op);
982 return do_filp_open(AT_FDCWD, filename, &op, lookup);
983}
984EXPORT_SYMBOL(filp_open);
985
Al Viro73d049a2011-03-11 12:08:24 -0500986struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
987 const char *filename, int flags)
988{
989 struct open_flags op;
990 int lookup = build_open_flags(flags, 0, &op);
991 if (flags & O_CREAT)
992 return ERR_PTR(-EINVAL);
993 if (!filename && (flags & O_DIRECTORY))
994 if (!dentry->d_inode->i_op->lookup)
995 return ERR_PTR(-ENOTDIR);
996 return do_file_open_root(dentry, mnt, filename, &op, lookup);
997}
998EXPORT_SYMBOL(file_open_root);
999
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001000long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
Al Viro47c805d2011-02-23 17:44:09 -05001002 struct open_flags op;
1003 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001004 char *tmp = getname(filename);
1005 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001008 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -05001010 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001011 if (IS_ERR(f)) {
1012 put_unused_fd(fd);
1013 fd = PTR_ERR(f);
1014 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -05001015 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001016 fd_install(fd, f);
1017 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 putname(tmp);
1020 }
1021 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001023
Heiko Carstensca013e92009-01-14 14:14:19 +01001024SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001025{
Linus Torvalds385910f2006-04-18 13:22:59 -07001026 long ret;
1027
Miklos Szeredie922efc2005-09-06 15:18:25 -07001028 if (force_o_largefile())
1029 flags |= O_LARGEFILE;
1030
Linus Torvalds385910f2006-04-18 13:22:59 -07001031 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1032 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001033 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001034 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001035}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
Heiko Carstens6559eed82009-01-14 14:14:32 +01001037SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1038 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001039{
Linus Torvalds385910f2006-04-18 13:22:59 -07001040 long ret;
1041
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001042 if (force_o_largefile())
1043 flags |= O_LARGEFILE;
1044
Linus Torvalds385910f2006-04-18 13:22:59 -07001045 ret = do_sys_open(dfd, filename, flags, mode);
1046 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001047 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001048 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001049}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001050
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051#ifndef __alpha__
1052
1053/*
1054 * For backward compatibility? Maybe this should be moved
1055 * into arch/i386 instead?
1056 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001057SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058{
1059 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1060}
1061
1062#endif
1063
1064/*
1065 * "id" is the POSIX thread ID. We use the
1066 * files pointer for this..
1067 */
1068int filp_close(struct file *filp, fl_owner_t id)
1069{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001070 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
1072 if (!file_count(filp)) {
1073 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001074 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 }
1076
Christoph Lameter45778ca2005-06-23 00:10:17 -07001077 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001078 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079
Al Viro1abf0c72011-03-13 03:51:11 -04001080 if (likely(!(filp->f_mode & FMODE_PATH))) {
1081 dnotify_flush(filp, id);
1082 locks_remove_posix(filp, id);
1083 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 fput(filp);
1085 return retval;
1086}
1087
1088EXPORT_SYMBOL(filp_close);
1089
1090/*
1091 * Careful here! We test whether the file pointer is NULL before
1092 * releasing the fd. This ensures that one clone task can't release
1093 * an fd while another clone is opening it.
1094 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001095SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096{
1097 struct file * filp;
1098 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001099 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001100 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101
1102 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001103 fdt = files_fdtable(files);
1104 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001106 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 if (!filp)
1108 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001109 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001110 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 __put_unused_fd(files, fd);
1112 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001113 retval = filp_close(filp, files);
1114
1115 /* can't restart close syscall because file table entry was cleared */
1116 if (unlikely(retval == -ERESTARTSYS ||
1117 retval == -ERESTARTNOINTR ||
1118 retval == -ERESTARTNOHAND ||
1119 retval == -ERESTART_RESTARTBLOCK))
1120 retval = -EINTR;
1121
1122 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
1124out_unlock:
1125 spin_unlock(&files->file_lock);
1126 return -EBADF;
1127}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128EXPORT_SYMBOL(sys_close);
1129
1130/*
1131 * This routine simulates a hangup on the tty, to arrange that users
1132 * are given clean terminals at login time.
1133 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001134SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135{
1136 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001137 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 return 0;
1139 }
1140 return -EPERM;
1141}
1142
1143/*
1144 * Called when an inode is about to be open.
1145 * We use this to disallow opening large files on 32bit systems if
1146 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1147 * on this flag in sys_open.
1148 */
1149int generic_file_open(struct inode * inode, struct file * filp)
1150{
1151 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001152 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 return 0;
1154}
1155
1156EXPORT_SYMBOL(generic_file_open);
1157
1158/*
1159 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001160 * file descriptors. The function is not supposed to ever fail, the only
1161 * reason it returns an 'int' and not 'void' is so that it can be plugged
1162 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 */
1164int nonseekable_open(struct inode *inode, struct file *filp)
1165{
1166 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1167 return 0;
1168}
1169
1170EXPORT_SYMBOL(nonseekable_open);