blob: bc132e167d2dd2e361fd8fb0354f987f2fa296dd [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
Jan Kara14da9202012-06-12 16:20:37 +0200167 sb_start_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 error = locks_verify_truncate(inode, file, length);
169 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900170 error = security_path_truncate(&file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900171 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700172 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Jan Kara14da9202012-06-12 16:20:37 +0200173 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174out_putf:
175 fput(file);
176out:
177 return error;
178}
179
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100180SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700182 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700183 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700184 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700185 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186}
187
188/* LFS versions of truncate are only needed on 32 bit machines */
189#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100190SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
192 return do_sys_truncate(path, length);
193}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100194#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
195asmlinkage long SyS_truncate64(long path, loff_t length)
196{
197 return SYSC_truncate64((const char __user *) path, length);
198}
199SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
200#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100202SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700204 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700205 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700206 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700207 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100209#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
210asmlinkage long SyS_ftruncate64(long fd, loff_t length)
211{
212 return SYSC_ftruncate64((unsigned int) fd, length);
213}
214SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100216#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400218
219int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400220{
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400221 struct inode *inode = file->f_path.dentry->d_inode;
222 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400223
224 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400225 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400226
227 /* Return error if mode is not supported */
Josef Bacik79124f182010-11-17 20:46:15 -0500228 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
229 return -EOPNOTSUPP;
230
231 /* Punch hole must have keep size set */
232 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
233 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400234 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400235
Amit Arora97ac7352007-07-17 21:42:44 -0400236 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400237 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100238
239 /* It's not possible punch hole on append only file */
240 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
241 return -EPERM;
242
243 if (IS_IMMUTABLE(inode))
244 return -EPERM;
245
Amit Arora97ac7352007-07-17 21:42:44 -0400246 /*
247 * Revalidate the write permissions, in case security policy has
248 * changed since the files were opened.
249 */
250 ret = security_file_permission(file, MAY_WRITE);
251 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400252 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400253
Amit Arora97ac7352007-07-17 21:42:44 -0400254 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400255 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400256
Amit Arora97ac7352007-07-17 21:42:44 -0400257 /*
258 * Let individual file system decide if it supports preallocation
259 * for directories or not.
260 */
261 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400262 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400263
Amit Arora97ac7352007-07-17 21:42:44 -0400264 /* Check for wrap through zero too */
265 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400266 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400267
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100268 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400269 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400270
Jan Kara14da9202012-06-12 16:20:37 +0200271 sb_start_write(inode->i_sb);
272 ret = file->f_op->fallocate(file, mode, offset, len);
273 sb_end_write(inode->i_sb);
274 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400275}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400276
277SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
278{
279 struct file *file;
280 int error = -EBADF;
281
282 file = fget(fd);
283 if (file) {
284 error = do_fallocate(file, mode, offset, len);
285 fput(file);
286 }
287
288 return error;
289}
290
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100291#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
292asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
293{
294 return SYSC_fallocate((int)fd, (int)mode, offset, len);
295}
296SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
297#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400298
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299/*
300 * access() needs to use the real uid/gid, not the effective uid/gid.
301 * We do this by temporarily clearing all FS-related capabilities and
302 * switching the fsuid/fsgid around to the real ones.
303 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100304SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305{
David Howellsd84f4f92008-11-14 10:39:23 +1100306 const struct cred *old_cred;
307 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400308 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400309 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 int res;
311
312 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
313 return -EINVAL;
314
David Howellsd84f4f92008-11-14 10:39:23 +1100315 override_cred = prepare_creds();
316 if (!override_cred)
317 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
David Howellsd84f4f92008-11-14 10:39:23 +1100319 override_cred->fsuid = override_cred->uid;
320 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700322 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100323 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800324 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
325 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100326 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700327 else
David Howellsd84f4f92008-11-14 10:39:23 +1100328 override_cred->cap_effective =
329 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700330 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
David Howellsd84f4f92008-11-14 10:39:23 +1100332 old_cred = override_creds(override_cred);
333
Al Viro2d8f3032008-07-22 09:59:21 -0400334 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700335 if (res)
336 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Al Viro2d8f3032008-07-22 09:59:21 -0400338 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400339
340 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400341 /*
342 * MAY_EXEC on regular files is denied if the fs is mounted
343 * with the "noexec" flag.
344 */
345 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400346 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400347 goto out_path_release;
348 }
349
Al Viro256984a2008-07-22 08:09:30 -0400350 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700351 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400352 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700353 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800354 /*
355 * This is a rare case where using __mnt_is_readonly()
356 * is OK without a mnt_want/drop_write() pair. Since
357 * no actual write to the fs is performed here, we do
358 * not need to telegraph to that to anyone.
359 *
360 * By doing this, we accept that this access is
361 * inherently racy and know that the fs may change
362 * state before we even see this result.
363 */
Al Viro2d8f3032008-07-22 09:59:21 -0400364 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700365 res = -EROFS;
366
367out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400368 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700369out:
David Howellsd84f4f92008-11-14 10:39:23 +1100370 revert_creds(old_cred);
371 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 return res;
373}
374
Heiko Carstensca013e92009-01-14 14:14:19 +0100375SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800376{
377 return sys_faccessat(AT_FDCWD, filename, mode);
378}
379
Heiko Carstens3cdad422009-01-14 14:14:22 +0100380SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381{
Al Viro2d8f3032008-07-22 09:59:21 -0400382 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 int error;
384
Al Viro2d8f3032008-07-22 09:59:21 -0400385 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 if (error)
387 goto out;
388
Eric Paris9cfcac82010-07-23 11:43:51 -0400389 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 if (error)
391 goto dput_and_out;
392
Al Viro2d8f3032008-07-22 09:59:21 -0400393 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
395dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400396 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397out:
398 return error;
399}
400
Heiko Carstens3cdad422009-01-14 14:14:22 +0100401SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402{
403 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 struct inode *inode;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700405 int error, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
407 error = -EBADF;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700408 file = fget_raw_light(fd, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 if (!file)
410 goto out;
411
Jan Blunckac748a02008-02-14 19:34:39 -0800412 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413
414 error = -ENOTDIR;
415 if (!S_ISDIR(inode->i_mode))
416 goto out_putf;
417
Eric Paris9cfcac82010-07-23 11:43:51 -0400418 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800420 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421out_putf:
Linus Torvalds332a2e12012-07-07 10:17:00 -0700422 fput_light(file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423out:
424 return error;
425}
426
Heiko Carstens3480b252009-01-14 14:14:16 +0100427SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428{
Al Viro2d8f3032008-07-22 09:59:21 -0400429 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 int error;
431
Al Viro2d8f3032008-07-22 09:59:21 -0400432 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 if (error)
434 goto out;
435
Eric Paris9cfcac82010-07-23 11:43:51 -0400436 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 if (error)
438 goto dput_and_out;
439
440 error = -EPERM;
441 if (!capable(CAP_SYS_CHROOT))
442 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900443 error = security_path_chroot(&path);
444 if (error)
445 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
Al Viro2d8f3032008-07-22 09:59:21 -0400447 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 error = 0;
449dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400450 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451out:
452 return error;
453}
454
Al Viroe57712e2011-07-26 04:15:54 -0400455static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456{
Al Viroe57712e2011-07-26 04:15:54 -0400457 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400459 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
Al Viroe57712e2011-07-26 04:15:54 -0400461 error = mnt_want_write(path->mnt);
462 if (error)
463 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900464 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500465 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400466 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900467 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
469 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400470 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900471out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800472 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400473 mnt_drop_write(path->mnt);
474 return error;
475}
476
Al Viro49f0a072011-07-26 04:22:01 -0400477SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400478{
479 struct file * file;
480 int err = -EBADF;
481
482 file = fget(fd);
483 if (file) {
484 audit_inode(NULL, file->f_path.dentry);
485 err = chmod_common(&file->f_path, mode);
486 fput(file);
487 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 return err;
489}
490
Al Viro49f0a072011-07-26 04:22:01 -0400491SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
Al Viro2d8f3032008-07-22 09:59:21 -0400493 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Al Viro2d8f3032008-07-22 09:59:21 -0400496 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400497 if (!error) {
498 error = chmod_common(&path, mode);
499 path_put(&path);
500 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 return error;
502}
503
Al Viro49f0a072011-07-26 04:22:01 -0400504SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800505{
506 return sys_fchmodat(AT_FDCWD, filename, mode);
507}
508
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900509static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900511 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 int error;
513 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800514 kuid_t uid;
515 kgid_t gid;
516
517 uid = make_kuid(current_user_ns(), user);
518 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 newattrs.ia_valid = ATTR_CTIME;
521 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800522 if (!uid_valid(uid))
523 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800525 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 }
527 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800528 if (!gid_valid(gid))
529 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800531 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 }
533 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700534 newattrs.ia_valid |=
535 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800536 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900537 error = security_path_chown(path, user, group);
538 if (!error)
539 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800540 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200541
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 return error;
543}
544
Heiko Carstens6559eed82009-01-14 14:14:32 +0100545SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
546 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800547{
Al Viro2d8f3032008-07-22 09:59:21 -0400548 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800549 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400550 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800551
Al Viro65cfc672011-03-13 15:56:26 -0400552 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800553 goto out;
554
Al Viro65cfc672011-03-13 15:56:26 -0400555 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
556 if (flag & AT_EMPTY_PATH)
557 lookup_flags |= LOOKUP_EMPTY;
558 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700559 if (error)
560 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400561 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800562 if (error)
563 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900564 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400565 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800566out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400567 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800568out:
569 return error;
570}
571
David Howells55e4def2012-06-25 12:55:09 +0100572SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
573{
574 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
575}
576
Heiko Carstensca013e92009-01-14 14:14:19 +0100577SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578{
David Howells55e4def2012-06-25 12:55:09 +0100579 return sys_fchownat(AT_FDCWD, filename, user, group,
580 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581}
582
Heiko Carstensca013e92009-01-14 14:14:19 +0100583SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584{
585 struct file * file;
586 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700587 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
589 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700590 if (!file)
591 goto out;
592
npiggin@suse.de96029c42009-04-26 20:25:55 +1000593 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800594 if (error)
595 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800596 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400597 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900598 error = chown_common(&file->f_path, user, group);
Al Viro2a79f172011-12-09 08:06:57 -0500599 mnt_drop_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800600out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700601 fput(file);
602out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 return error;
604}
605
Dave Hansen4a3fd212008-02-15 14:37:48 -0800606/*
607 * You have to be very careful that these write
608 * counts get cleaned up in error cases and
609 * upon __fput(). This should probably never
610 * be called outside of __dentry_open().
611 */
612static inline int __get_file_write_access(struct inode *inode,
613 struct vfsmount *mnt)
614{
615 int error;
616 error = get_write_access(inode);
617 if (error)
618 return error;
619 /*
620 * Do not take mount writer counts on
621 * special files since no writes to
622 * the mount itself will occur.
623 */
624 if (!special_file(inode->i_mode)) {
625 /*
626 * Balanced in __fput()
627 */
Jan Karaeb04c282012-06-12 16:20:35 +0200628 error = __mnt_want_write(mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800629 if (error)
630 put_write_access(inode);
631 }
632 return error;
633}
634
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200635int open_check_o_direct(struct file *f)
636{
637 /* NB: we're sure to have correct a_ops only after f_op->open */
638 if (f->f_flags & O_DIRECT) {
639 if (!f->f_mapping->a_ops ||
640 ((!f->f_mapping->a_ops->direct_IO) &&
641 (!f->f_mapping->a_ops->get_xip_mem))) {
642 return -EINVAL;
643 }
644 }
645 return 0;
646}
647
Al Viro02e51802012-06-10 14:32:45 -0400648static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400649 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
Al Virob5bcdda2012-07-20 23:28:46 +0400662 path_get(&f->f_path);
Al Viro02e51802012-06-10 14:32:45 -0400663 inode = f->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 if (f->f_mode & FMODE_WRITE) {
Al Viro02e51802012-06-10 14:32:45 -0400665 error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 if (error)
667 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800668 if (!special_file(inode->i_mode))
669 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 }
671
672 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 f->f_pos = 0;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000674 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Al Viro1abf0c72011-03-13 03:51:11 -0400676 if (unlikely(f->f_mode & FMODE_PATH)) {
677 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400678 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400679 }
680
681 f->f_op = fops_get(inode->i_fop);
682
Eric Paris83d49852012-04-04 13:45:40 -0400683 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900684 if (error)
685 goto cleanup_all;
686
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400687 error = break_lease(inode, f->f_flags);
688 if (error)
689 goto cleanup_all;
690
Trond Myklebust834f2a42005-10-18 14:20:16 -0700691 if (!open && f->f_op)
692 open = f->f_op->open;
693 if (open) {
694 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 if (error)
696 goto cleanup_all;
697 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400698 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
699 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700700
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
702
703 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
704
Al Viro96b7e572012-06-10 14:22:04 -0400705 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
707cleanup_all:
708 fops_put(f->f_op);
Al Viroc3c4f692012-06-23 22:49:45 +0400709 file_sb_list_del(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800710 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800712 if (!special_file(inode->i_mode)) {
713 /*
714 * We don't consider this a real
715 * mnt_want/drop_write() pair
716 * because it all happenend right
717 * here, so just reset the state.
718 */
719 file_reset_write(f);
Al Virofe7c8052012-08-04 08:39:23 +0400720 __mnt_drop_write(f->f_path.mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800721 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800722 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400724 path_put(&f->f_path);
725 f->f_path.mnt = NULL;
726 f->f_path.dentry = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400727 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728}
729
Trond Myklebust834f2a42005-10-18 14:20:16 -0700730/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200731 * finish_open - finish opening a file
732 * @od: opaque open data
733 * @dentry: pointer to dentry
734 * @open: open callback
735 *
736 * This can be used to finish opening a file passed to i_op->atomic_open().
737 *
738 * If the open callback is set to NULL, then the standard f_op->open()
739 * filesystem callback is substituted.
740 */
Al Viro30d90492012-06-22 12:40:19 +0400741int finish_open(struct file *file, struct dentry *dentry,
742 int (*open)(struct inode *, struct file *),
743 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200744{
Al Viro96b7e572012-06-10 14:22:04 -0400745 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400746 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200747
Al Virob5bcdda2012-07-20 23:28:46 +0400748 file->f_path.dentry = dentry;
Al Viro02e51802012-06-10 14:32:45 -0400749 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400750 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400751 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200752
Al Viro96b7e572012-06-10 14:22:04 -0400753 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200754}
755EXPORT_SYMBOL(finish_open);
756
757/**
758 * finish_no_open - finish ->atomic_open() without opening the file
759 *
760 * @od: opaque open data
761 * @dentry: dentry or NULL (as returned from ->lookup())
762 *
763 * This can be used to set the result of a successful lookup in ->atomic_open().
764 * The filesystem's atomic_open() method shall return NULL after calling this.
765 */
Al Viroe45198a2012-06-10 06:48:09 -0400766int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200767{
Al Viro30d90492012-06-22 12:40:19 +0400768 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400769 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200770}
771EXPORT_SYMBOL(finish_no_open);
772
Al Viro765927b2012-06-26 21:58:53 +0400773struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100774 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700775{
776 int error;
777 struct file *f;
778
David Howellse0e81732009-09-02 09:13:40 +0100779 validate_creds(cred);
780
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900781 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400782 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800783
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700784 error = -ENFILE;
785 f = get_empty_filp();
Al Viro765927b2012-06-26 21:58:53 +0400786 if (f == NULL)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700787 return ERR_PTR(error);
788
Al Viro482928d2009-12-19 10:10:39 -0500789 f->f_flags = flags;
Al Viro765927b2012-06-26 21:58:53 +0400790 f->f_path = *path;
Al Viro02e51802012-06-10 14:32:45 -0400791 error = do_dentry_open(f, NULL, cred);
Al Viro2a027e72012-06-10 14:24:38 -0400792 if (!error) {
793 error = open_check_o_direct(f);
794 if (error) {
795 fput(f);
796 f = ERR_PTR(error);
797 }
798 } else {
799 put_filp(f);
800 f = ERR_PTR(error);
801 }
802 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700803}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804EXPORT_SYMBOL(dentry_open);
805
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800806static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700808 struct fdtable *fdt = files_fdtable(files);
David Howells1dce27c2012-02-16 17:49:42 +0000809 __clear_open_fd(fd, fdt);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800810 if (fd < files->next_fd)
811 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800814void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
816 struct files_struct *files = current->files;
817 spin_lock(&files->file_lock);
818 __put_unused_fd(files, fd);
819 spin_unlock(&files->file_lock);
820}
821
822EXPORT_SYMBOL(put_unused_fd);
823
824/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800825 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 *
827 * The VFS is full of places where we drop the files lock between
828 * setting the open_fds bitmap and installing the file in the file
829 * array. At any such point, we are vulnerable to a dup2() race
830 * installing a file in the array before us. We need to detect this and
831 * fput() the struct file we are about to overwrite in this case.
832 *
833 * It should never happen - if we allow dup2() do it, _really_ bad things
834 * will follow.
835 */
836
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800837void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
839 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700840 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700842 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700843 BUG_ON(fdt->fd[fd] != NULL);
844 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 spin_unlock(&files->file_lock);
846}
847
848EXPORT_SYMBOL(fd_install);
849
Al Viroa218d0f2011-11-21 14:59:34 -0500850static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500851{
852 int lookup_flags = 0;
853 int acc_mode;
854
855 if (!(flags & O_CREAT))
856 mode = 0;
857 op->mode = mode;
858
859 /* Must never be set by userspace */
860 flags &= ~FMODE_NONOTIFY;
861
862 /*
863 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
864 * check for O_DSYNC if the need any syncing at all we enforce it's
865 * always set instead of having to deal with possibly weird behaviour
866 * for malicious applications setting only __O_SYNC.
867 */
868 if (flags & __O_SYNC)
869 flags |= O_DSYNC;
870
Al Viro1abf0c72011-03-13 03:51:11 -0400871 /*
872 * If we have O_PATH in the open flag. Then we
873 * cannot have anything other than the below set of flags
874 */
875 if (flags & O_PATH) {
876 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
877 acc_mode = 0;
878 } else {
879 acc_mode = MAY_OPEN | ACC_MODE(flags);
880 }
Al Viro47c805d2011-02-23 17:44:09 -0500881
Al Viro1abf0c72011-03-13 03:51:11 -0400882 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500883
884 /* O_TRUNC implies we need access checks for write permissions */
885 if (flags & O_TRUNC)
886 acc_mode |= MAY_WRITE;
887
888 /* Allow the LSM permission hook to distinguish append
889 access from general write access. */
890 if (flags & O_APPEND)
891 acc_mode |= MAY_APPEND;
892
893 op->acc_mode = acc_mode;
894
Al Viro1abf0c72011-03-13 03:51:11 -0400895 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
896
Al Viro47c805d2011-02-23 17:44:09 -0500897 if (flags & O_CREAT) {
898 op->intent |= LOOKUP_CREATE;
899 if (flags & O_EXCL)
900 op->intent |= LOOKUP_EXCL;
901 }
902
903 if (flags & O_DIRECTORY)
904 lookup_flags |= LOOKUP_DIRECTORY;
905 if (!(flags & O_NOFOLLOW))
906 lookup_flags |= LOOKUP_FOLLOW;
907 return lookup_flags;
908}
909
910/**
911 * filp_open - open file and return file pointer
912 *
913 * @filename: path to open
914 * @flags: open flags as per the open(2) second argument
915 * @mode: mode for the new file if O_CREAT is set, else ignored
916 *
917 * This is the helper to open a file from kernelspace if you really
918 * have to. But in generally you should not do this, so please move
919 * along, nothing to see here..
920 */
Al Viroa218d0f2011-11-21 14:59:34 -0500921struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500922{
923 struct open_flags op;
924 int lookup = build_open_flags(flags, mode, &op);
925 return do_filp_open(AT_FDCWD, filename, &op, lookup);
926}
927EXPORT_SYMBOL(filp_open);
928
Al Viro73d049a2011-03-11 12:08:24 -0500929struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
930 const char *filename, int flags)
931{
932 struct open_flags op;
933 int lookup = build_open_flags(flags, 0, &op);
934 if (flags & O_CREAT)
935 return ERR_PTR(-EINVAL);
936 if (!filename && (flags & O_DIRECTORY))
937 if (!dentry->d_inode->i_op->lookup)
938 return ERR_PTR(-ENOTDIR);
939 return do_file_open_root(dentry, mnt, filename, &op, lookup);
940}
941EXPORT_SYMBOL(file_open_root);
942
Al Viroa218d0f2011-11-21 14:59:34 -0500943long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944{
Al Viro47c805d2011-02-23 17:44:09 -0500945 struct open_flags op;
946 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -0700947 char *tmp = getname(filename);
948 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700951 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -0500953 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700954 if (IS_ERR(f)) {
955 put_unused_fd(fd);
956 fd = PTR_ERR(f);
957 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500958 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700959 fd_install(fd, f);
960 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 putname(tmp);
963 }
964 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700966
Al Viroa218d0f2011-11-21 14:59:34 -0500967SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700968{
Linus Torvalds385910f2006-04-18 13:22:59 -0700969 long ret;
970
Miklos Szeredie922efc2005-09-06 15:18:25 -0700971 if (force_o_largefile())
972 flags |= O_LARGEFILE;
973
Linus Torvalds385910f2006-04-18 13:22:59 -0700974 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
975 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700976 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700977 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -0700978}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
Heiko Carstens6559eed82009-01-14 14:14:32 +0100980SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -0500981 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800982{
Linus Torvalds385910f2006-04-18 13:22:59 -0700983 long ret;
984
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800985 if (force_o_largefile())
986 flags |= O_LARGEFILE;
987
Linus Torvalds385910f2006-04-18 13:22:59 -0700988 ret = do_sys_open(dfd, filename, flags, mode);
989 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700990 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700991 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800992}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800993
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994#ifndef __alpha__
995
996/*
997 * For backward compatibility? Maybe this should be moved
998 * into arch/i386 instead?
999 */
Al Viroa218d0f2011-11-21 14:59:34 -05001000SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
1002 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1003}
1004
1005#endif
1006
1007/*
1008 * "id" is the POSIX thread ID. We use the
1009 * files pointer for this..
1010 */
1011int filp_close(struct file *filp, fl_owner_t id)
1012{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001013 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 if (!file_count(filp)) {
1016 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001017 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 }
1019
Christoph Lameter45778ca2005-06-23 00:10:17 -07001020 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001021 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
Al Viro1abf0c72011-03-13 03:51:11 -04001023 if (likely(!(filp->f_mode & FMODE_PATH))) {
1024 dnotify_flush(filp, id);
1025 locks_remove_posix(filp, id);
1026 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 fput(filp);
1028 return retval;
1029}
1030
1031EXPORT_SYMBOL(filp_close);
1032
1033/*
1034 * Careful here! We test whether the file pointer is NULL before
1035 * releasing the fd. This ensures that one clone task can't release
1036 * an fd while another clone is opening it.
1037 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001038SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039{
1040 struct file * filp;
1041 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001042 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001043 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
1045 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001046 fdt = files_fdtable(files);
1047 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001049 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 if (!filp)
1051 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001052 rcu_assign_pointer(fdt->fd[fd], NULL);
David Howells1dce27c2012-02-16 17:49:42 +00001053 __clear_close_on_exec(fd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 __put_unused_fd(files, fd);
1055 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001056 retval = filp_close(filp, files);
1057
1058 /* can't restart close syscall because file table entry was cleared */
1059 if (unlikely(retval == -ERESTARTSYS ||
1060 retval == -ERESTARTNOINTR ||
1061 retval == -ERESTARTNOHAND ||
1062 retval == -ERESTART_RESTARTBLOCK))
1063 retval = -EINTR;
1064
1065 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067out_unlock:
1068 spin_unlock(&files->file_lock);
1069 return -EBADF;
1070}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071EXPORT_SYMBOL(sys_close);
1072
1073/*
1074 * This routine simulates a hangup on the tty, to arrange that users
1075 * are given clean terminals at login time.
1076 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001077SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078{
1079 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001080 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 return 0;
1082 }
1083 return -EPERM;
1084}
1085
1086/*
1087 * Called when an inode is about to be open.
1088 * We use this to disallow opening large files on 32bit systems if
1089 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1090 * on this flag in sys_open.
1091 */
1092int generic_file_open(struct inode * inode, struct file * filp)
1093{
1094 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001095 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 return 0;
1097}
1098
1099EXPORT_SYMBOL(generic_file_open);
1100
1101/*
1102 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001103 * file descriptors. The function is not supposed to ever fail, the only
1104 * reason it returns an 'int' and not 'void' is so that it can be plugged
1105 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 */
1107int nonseekable_open(struct inode *inode, struct file *filp)
1108{
1109 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1110 return 0;
1111}
1112
1113EXPORT_SYMBOL(nonseekable_open);