blob: 44da0feeca2c94c980c2723bda63b1fcf74caedc [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{
Al Virobf2965d2012-08-26 20:13:36 -0400135 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -0400137 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138 int error;
139
140 error = -EINVAL;
141 if (length < 0)
142 goto out;
143 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400144 f = fdget(fd);
145 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 goto out;
147
148 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400149 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 small = 0;
151
Al Viro2903ff02012-08-28 12:52:22 -0400152 dentry = f.file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 inode = dentry->d_inode;
154 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400155 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 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);
Al Viro2903ff02012-08-28 12:52:22 -0400168 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400170 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900171 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400172 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200173 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400175 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176out:
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{
Al Viro2903ff02012-08-28 12:52:22 -0400279 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400280 int error = -EBADF;
281
Al Viro2903ff02012-08-28 12:52:22 -0400282 if (f.file) {
283 error = do_fallocate(f.file, mode, offset, len);
284 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400285 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400286 return error;
287}
288
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100289#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
290asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
291{
292 return SYSC_fallocate((int)fd, (int)mode, offset, len);
293}
294SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
295#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400296
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297/*
298 * access() needs to use the real uid/gid, not the effective uid/gid.
299 * We do this by temporarily clearing all FS-related capabilities and
300 * switching the fsuid/fsgid around to the real ones.
301 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100302SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303{
David Howellsd84f4f92008-11-14 10:39:23 +1100304 const struct cred *old_cred;
305 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400306 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400307 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 int res;
309
310 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
311 return -EINVAL;
312
David Howellsd84f4f92008-11-14 10:39:23 +1100313 override_cred = prepare_creds();
314 if (!override_cred)
315 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
David Howellsd84f4f92008-11-14 10:39:23 +1100317 override_cred->fsuid = override_cred->uid;
318 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700320 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100321 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800322 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
323 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100324 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700325 else
David Howellsd84f4f92008-11-14 10:39:23 +1100326 override_cred->cap_effective =
327 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700328 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
David Howellsd84f4f92008-11-14 10:39:23 +1100330 old_cred = override_creds(override_cred);
331
Al Viro2d8f3032008-07-22 09:59:21 -0400332 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700333 if (res)
334 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
Al Viro2d8f3032008-07-22 09:59:21 -0400336 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400337
338 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400339 /*
340 * MAY_EXEC on regular files is denied if the fs is mounted
341 * with the "noexec" flag.
342 */
343 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400344 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400345 goto out_path_release;
346 }
347
Al Viro256984a2008-07-22 08:09:30 -0400348 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700349 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400350 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700351 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800352 /*
353 * This is a rare case where using __mnt_is_readonly()
354 * is OK without a mnt_want/drop_write() pair. Since
355 * no actual write to the fs is performed here, we do
356 * not need to telegraph to that to anyone.
357 *
358 * By doing this, we accept that this access is
359 * inherently racy and know that the fs may change
360 * state before we even see this result.
361 */
Al Viro2d8f3032008-07-22 09:59:21 -0400362 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700363 res = -EROFS;
364
365out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400366 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700367out:
David Howellsd84f4f92008-11-14 10:39:23 +1100368 revert_creds(old_cred);
369 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 return res;
371}
372
Heiko Carstensca013e92009-01-14 14:14:19 +0100373SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800374{
375 return sys_faccessat(AT_FDCWD, filename, mode);
376}
377
Heiko Carstens3cdad422009-01-14 14:14:22 +0100378SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
Al Viro2d8f3032008-07-22 09:59:21 -0400380 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 int error;
382
Al Viro2d8f3032008-07-22 09:59:21 -0400383 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 if (error)
385 goto out;
386
Eric Paris9cfcac82010-07-23 11:43:51 -0400387 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 if (error)
389 goto dput_and_out;
390
Al Viro2d8f3032008-07-22 09:59:21 -0400391 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400394 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395out:
396 return error;
397}
398
Heiko Carstens3cdad422009-01-14 14:14:22 +0100399SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400{
Al Viro2903ff02012-08-28 12:52:22 -0400401 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400403 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400406 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 goto out;
408
Al Viro2903ff02012-08-28 12:52:22 -0400409 inode = f.file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
411 error = -ENOTDIR;
412 if (!S_ISDIR(inode->i_mode))
413 goto out_putf;
414
Eric Paris9cfcac82010-07-23 11:43:51 -0400415 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400417 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400419 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420out:
421 return error;
422}
423
Heiko Carstens3480b252009-01-14 14:14:16 +0100424SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425{
Al Viro2d8f3032008-07-22 09:59:21 -0400426 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 int error;
428
Al Viro2d8f3032008-07-22 09:59:21 -0400429 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 if (error)
431 goto out;
432
Eric Paris9cfcac82010-07-23 11:43:51 -0400433 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (error)
435 goto dput_and_out;
436
437 error = -EPERM;
438 if (!capable(CAP_SYS_CHROOT))
439 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900440 error = security_path_chroot(&path);
441 if (error)
442 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
Al Viro2d8f3032008-07-22 09:59:21 -0400444 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 error = 0;
446dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400447 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448out:
449 return error;
450}
451
Al Viroe57712e2011-07-26 04:15:54 -0400452static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
Al Viroe57712e2011-07-26 04:15:54 -0400454 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400456 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457
Al Viroe57712e2011-07-26 04:15:54 -0400458 error = mnt_want_write(path->mnt);
459 if (error)
460 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900461 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500462 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400463 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900464 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
466 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400467 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900468out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800469 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400470 mnt_drop_write(path->mnt);
471 return error;
472}
473
Al Viro49f0a072011-07-26 04:22:01 -0400474SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400475{
476 struct file * file;
477 int err = -EBADF;
478
479 file = fget(fd);
480 if (file) {
481 audit_inode(NULL, file->f_path.dentry);
482 err = chmod_common(&file->f_path, mode);
483 fput(file);
484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 return err;
486}
487
Al Viro49f0a072011-07-26 04:22:01 -0400488SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
Al Viro2d8f3032008-07-22 09:59:21 -0400490 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
Al Viro2d8f3032008-07-22 09:59:21 -0400493 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400494 if (!error) {
495 error = chmod_common(&path, mode);
496 path_put(&path);
497 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 return error;
499}
500
Al Viro49f0a072011-07-26 04:22:01 -0400501SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800502{
503 return sys_fchmodat(AT_FDCWD, filename, mode);
504}
505
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900506static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900508 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 int error;
510 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800511 kuid_t uid;
512 kgid_t gid;
513
514 uid = make_kuid(current_user_ns(), user);
515 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 newattrs.ia_valid = ATTR_CTIME;
518 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800519 if (!uid_valid(uid))
520 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800522 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 }
524 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800525 if (!gid_valid(gid))
526 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800528 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 }
530 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700531 newattrs.ia_valid |=
532 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800533 mutex_lock(&inode->i_mutex);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600534 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900535 if (!error)
536 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800537 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200538
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 return error;
540}
541
Heiko Carstens6559eed82009-01-14 14:14:32 +0100542SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
543 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800544{
Al Viro2d8f3032008-07-22 09:59:21 -0400545 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800546 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400547 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800548
Al Viro65cfc672011-03-13 15:56:26 -0400549 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800550 goto out;
551
Al Viro65cfc672011-03-13 15:56:26 -0400552 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
553 if (flag & AT_EMPTY_PATH)
554 lookup_flags |= LOOKUP_EMPTY;
555 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700556 if (error)
557 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400558 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800559 if (error)
560 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900561 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400562 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800563out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400564 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800565out:
566 return error;
567}
568
David Howells55e4def2012-06-25 12:55:09 +0100569SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
570{
571 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
572}
573
Heiko Carstensca013e92009-01-14 14:14:19 +0100574SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575{
David Howells55e4def2012-06-25 12:55:09 +0100576 return sys_fchownat(AT_FDCWD, filename, user, group,
577 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578}
579
Heiko Carstensca013e92009-01-14 14:14:19 +0100580SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581{
Al Viro2903ff02012-08-28 12:52:22 -0400582 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 int error = -EBADF;
584
Al Viro2903ff02012-08-28 12:52:22 -0400585 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700586 goto out;
587
Al Viro2903ff02012-08-28 12:52:22 -0400588 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800589 if (error)
590 goto out_fput;
Al Viro2903ff02012-08-28 12:52:22 -0400591 audit_inode(NULL, f.file->f_path.dentry);
592 error = chown_common(&f.file->f_path, user, group);
593 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800594out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400595 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700596out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 return error;
598}
599
Dave Hansen4a3fd212008-02-15 14:37:48 -0800600/*
601 * You have to be very careful that these write
602 * counts get cleaned up in error cases and
603 * upon __fput(). This should probably never
604 * be called outside of __dentry_open().
605 */
606static inline int __get_file_write_access(struct inode *inode,
607 struct vfsmount *mnt)
608{
609 int error;
610 error = get_write_access(inode);
611 if (error)
612 return error;
613 /*
614 * Do not take mount writer counts on
615 * special files since no writes to
616 * the mount itself will occur.
617 */
618 if (!special_file(inode->i_mode)) {
619 /*
620 * Balanced in __fput()
621 */
Jan Karaeb04c282012-06-12 16:20:35 +0200622 error = __mnt_want_write(mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800623 if (error)
624 put_write_access(inode);
625 }
626 return error;
627}
628
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200629int open_check_o_direct(struct file *f)
630{
631 /* NB: we're sure to have correct a_ops only after f_op->open */
632 if (f->f_flags & O_DIRECT) {
633 if (!f->f_mapping->a_ops ||
634 ((!f->f_mapping->a_ops->direct_IO) &&
635 (!f->f_mapping->a_ops->get_xip_mem))) {
636 return -EINVAL;
637 }
638 }
639 return 0;
640}
641
Al Viro02e51802012-06-10 14:32:45 -0400642static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400643 int (*open)(struct inode *, struct file *),
644 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645{
Al Viro1abf0c72011-03-13 03:51:11 -0400646 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 struct inode *inode;
648 int error;
649
Al Viro53009902009-12-19 10:15:07 -0500650 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700651 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400652
653 if (unlikely(f->f_flags & O_PATH))
654 f->f_mode = FMODE_PATH;
655
Al Virob5bcdda2012-07-20 23:28:46 +0400656 path_get(&f->f_path);
Al Viro02e51802012-06-10 14:32:45 -0400657 inode = f->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 if (f->f_mode & FMODE_WRITE) {
Al Viro02e51802012-06-10 14:32:45 -0400659 error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 if (error)
661 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800662 if (!special_file(inode->i_mode))
663 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 }
665
666 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 f->f_pos = 0;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000668 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
Al Viro1abf0c72011-03-13 03:51:11 -0400670 if (unlikely(f->f_mode & FMODE_PATH)) {
671 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400672 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400673 }
674
675 f->f_op = fops_get(inode->i_fop);
676
Eric Paris83d49852012-04-04 13:45:40 -0400677 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900678 if (error)
679 goto cleanup_all;
680
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400681 error = break_lease(inode, f->f_flags);
682 if (error)
683 goto cleanup_all;
684
Trond Myklebust834f2a42005-10-18 14:20:16 -0700685 if (!open && f->f_op)
686 open = f->f_op->open;
687 if (open) {
688 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 if (error)
690 goto cleanup_all;
691 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400692 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
693 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700694
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
696
697 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
698
Al Viro96b7e572012-06-10 14:22:04 -0400699 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
701cleanup_all:
702 fops_put(f->f_op);
Al Viroc3c4f692012-06-23 22:49:45 +0400703 file_sb_list_del(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800704 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800706 if (!special_file(inode->i_mode)) {
707 /*
708 * We don't consider this a real
709 * mnt_want/drop_write() pair
710 * because it all happenend right
711 * here, so just reset the state.
712 */
713 file_reset_write(f);
Al Virofe7c8052012-08-04 08:39:23 +0400714 __mnt_drop_write(f->f_path.mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800715 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800716 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400718 path_put(&f->f_path);
719 f->f_path.mnt = NULL;
720 f->f_path.dentry = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400721 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722}
723
Trond Myklebust834f2a42005-10-18 14:20:16 -0700724/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200725 * finish_open - finish opening a file
726 * @od: opaque open data
727 * @dentry: pointer to dentry
728 * @open: open callback
729 *
730 * This can be used to finish opening a file passed to i_op->atomic_open().
731 *
732 * If the open callback is set to NULL, then the standard f_op->open()
733 * filesystem callback is substituted.
734 */
Al Viro30d90492012-06-22 12:40:19 +0400735int finish_open(struct file *file, struct dentry *dentry,
736 int (*open)(struct inode *, struct file *),
737 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200738{
Al Viro96b7e572012-06-10 14:22:04 -0400739 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400740 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200741
Al Virob5bcdda2012-07-20 23:28:46 +0400742 file->f_path.dentry = dentry;
Al Viro02e51802012-06-10 14:32:45 -0400743 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400744 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400745 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200746
Al Viro96b7e572012-06-10 14:22:04 -0400747 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200748}
749EXPORT_SYMBOL(finish_open);
750
751/**
752 * finish_no_open - finish ->atomic_open() without opening the file
753 *
754 * @od: opaque open data
755 * @dentry: dentry or NULL (as returned from ->lookup())
756 *
757 * This can be used to set the result of a successful lookup in ->atomic_open().
758 * The filesystem's atomic_open() method shall return NULL after calling this.
759 */
Al Viroe45198a2012-06-10 06:48:09 -0400760int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200761{
Al Viro30d90492012-06-22 12:40:19 +0400762 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400763 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200764}
765EXPORT_SYMBOL(finish_no_open);
766
Al Viro765927b2012-06-26 21:58:53 +0400767struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100768 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700769{
770 int error;
771 struct file *f;
772
David Howellse0e81732009-09-02 09:13:40 +0100773 validate_creds(cred);
774
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900775 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400776 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800777
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700778 error = -ENFILE;
779 f = get_empty_filp();
Al Viro765927b2012-06-26 21:58:53 +0400780 if (f == NULL)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700781 return ERR_PTR(error);
782
Al Viro482928d2009-12-19 10:10:39 -0500783 f->f_flags = flags;
Al Viro765927b2012-06-26 21:58:53 +0400784 f->f_path = *path;
Al Viro02e51802012-06-10 14:32:45 -0400785 error = do_dentry_open(f, NULL, cred);
Al Viro2a027e72012-06-10 14:24:38 -0400786 if (!error) {
787 error = open_check_o_direct(f);
788 if (error) {
789 fput(f);
790 f = ERR_PTR(error);
791 }
792 } else {
793 put_filp(f);
794 f = ERR_PTR(error);
795 }
796 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700797}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798EXPORT_SYMBOL(dentry_open);
799
Al Viroa218d0f2011-11-21 14:59:34 -0500800static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500801{
802 int lookup_flags = 0;
803 int acc_mode;
804
Miklos Szeredie68726f2012-08-15 13:01:24 +0200805 if (flags & O_CREAT)
806 op->mode = (mode & S_IALLUGO) | S_IFREG;
807 else
808 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500809
810 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400811 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500812
813 /*
814 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
815 * check for O_DSYNC if the need any syncing at all we enforce it's
816 * always set instead of having to deal with possibly weird behaviour
817 * for malicious applications setting only __O_SYNC.
818 */
819 if (flags & __O_SYNC)
820 flags |= O_DSYNC;
821
Al Viro1abf0c72011-03-13 03:51:11 -0400822 /*
823 * If we have O_PATH in the open flag. Then we
824 * cannot have anything other than the below set of flags
825 */
826 if (flags & O_PATH) {
827 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
828 acc_mode = 0;
829 } else {
830 acc_mode = MAY_OPEN | ACC_MODE(flags);
831 }
Al Viro47c805d2011-02-23 17:44:09 -0500832
Al Viro1abf0c72011-03-13 03:51:11 -0400833 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500834
835 /* O_TRUNC implies we need access checks for write permissions */
836 if (flags & O_TRUNC)
837 acc_mode |= MAY_WRITE;
838
839 /* Allow the LSM permission hook to distinguish append
840 access from general write access. */
841 if (flags & O_APPEND)
842 acc_mode |= MAY_APPEND;
843
844 op->acc_mode = acc_mode;
845
Al Viro1abf0c72011-03-13 03:51:11 -0400846 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
847
Al Viro47c805d2011-02-23 17:44:09 -0500848 if (flags & O_CREAT) {
849 op->intent |= LOOKUP_CREATE;
850 if (flags & O_EXCL)
851 op->intent |= LOOKUP_EXCL;
852 }
853
854 if (flags & O_DIRECTORY)
855 lookup_flags |= LOOKUP_DIRECTORY;
856 if (!(flags & O_NOFOLLOW))
857 lookup_flags |= LOOKUP_FOLLOW;
858 return lookup_flags;
859}
860
861/**
862 * filp_open - open file and return file pointer
863 *
864 * @filename: path to open
865 * @flags: open flags as per the open(2) second argument
866 * @mode: mode for the new file if O_CREAT is set, else ignored
867 *
868 * This is the helper to open a file from kernelspace if you really
869 * have to. But in generally you should not do this, so please move
870 * along, nothing to see here..
871 */
Al Viroa218d0f2011-11-21 14:59:34 -0500872struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500873{
874 struct open_flags op;
875 int lookup = build_open_flags(flags, mode, &op);
876 return do_filp_open(AT_FDCWD, filename, &op, lookup);
877}
878EXPORT_SYMBOL(filp_open);
879
Al Viro73d049a2011-03-11 12:08:24 -0500880struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
881 const char *filename, int flags)
882{
883 struct open_flags op;
884 int lookup = build_open_flags(flags, 0, &op);
885 if (flags & O_CREAT)
886 return ERR_PTR(-EINVAL);
887 if (!filename && (flags & O_DIRECTORY))
888 if (!dentry->d_inode->i_op->lookup)
889 return ERR_PTR(-ENOTDIR);
890 return do_file_open_root(dentry, mnt, filename, &op, lookup);
891}
892EXPORT_SYMBOL(file_open_root);
893
Al Viroa218d0f2011-11-21 14:59:34 -0500894long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
Al Viro47c805d2011-02-23 17:44:09 -0500896 struct open_flags op;
897 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -0700898 char *tmp = getname(filename);
899 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700902 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -0500904 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700905 if (IS_ERR(f)) {
906 put_unused_fd(fd);
907 fd = PTR_ERR(f);
908 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500909 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700910 fd_install(fd, f);
911 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 putname(tmp);
914 }
915 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700917
Al Viroa218d0f2011-11-21 14:59:34 -0500918SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700919{
Linus Torvalds385910f2006-04-18 13:22:59 -0700920 long ret;
921
Miklos Szeredie922efc2005-09-06 15:18:25 -0700922 if (force_o_largefile())
923 flags |= O_LARGEFILE;
924
Linus Torvalds385910f2006-04-18 13:22:59 -0700925 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
926 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700927 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700928 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -0700929}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
Heiko Carstens6559eed82009-01-14 14:14:32 +0100931SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -0500932 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800933{
Linus Torvalds385910f2006-04-18 13:22:59 -0700934 long ret;
935
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800936 if (force_o_largefile())
937 flags |= O_LARGEFILE;
938
Linus Torvalds385910f2006-04-18 13:22:59 -0700939 ret = do_sys_open(dfd, filename, flags, mode);
940 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700941 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700942 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800943}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800944
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945#ifndef __alpha__
946
947/*
948 * For backward compatibility? Maybe this should be moved
949 * into arch/i386 instead?
950 */
Al Viroa218d0f2011-11-21 14:59:34 -0500951SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952{
953 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
954}
955
956#endif
957
958/*
959 * "id" is the POSIX thread ID. We use the
960 * files pointer for this..
961 */
962int filp_close(struct file *filp, fl_owner_t id)
963{
Christoph Lameter45778ca2005-06-23 00:10:17 -0700964 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
966 if (!file_count(filp)) {
967 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -0700968 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 }
970
Christoph Lameter45778ca2005-06-23 00:10:17 -0700971 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700972 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
Al Viro1abf0c72011-03-13 03:51:11 -0400974 if (likely(!(filp->f_mode & FMODE_PATH))) {
975 dnotify_flush(filp, id);
976 locks_remove_posix(filp, id);
977 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 fput(filp);
979 return retval;
980}
981
982EXPORT_SYMBOL(filp_close);
983
984/*
985 * Careful here! We test whether the file pointer is NULL before
986 * releasing the fd. This ensures that one clone task can't release
987 * an fd while another clone is opening it.
988 */
Heiko Carstensca013e92009-01-14 14:14:19 +0100989SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
Al Viro483ce1d2012-08-19 12:04:24 -0400991 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -0700992
993 /* can't restart close syscall because file table entry was cleared */
994 if (unlikely(retval == -ERESTARTSYS ||
995 retval == -ERESTARTNOINTR ||
996 retval == -ERESTARTNOHAND ||
997 retval == -ERESTART_RESTARTBLOCK))
998 retval = -EINTR;
999
1000 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002EXPORT_SYMBOL(sys_close);
1003
1004/*
1005 * This routine simulates a hangup on the tty, to arrange that users
1006 * are given clean terminals at login time.
1007 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001008SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009{
1010 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001011 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 return 0;
1013 }
1014 return -EPERM;
1015}
1016
1017/*
1018 * Called when an inode is about to be open.
1019 * We use this to disallow opening large files on 32bit systems if
1020 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1021 * on this flag in sys_open.
1022 */
1023int generic_file_open(struct inode * inode, struct file * filp)
1024{
1025 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001026 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 return 0;
1028}
1029
1030EXPORT_SYMBOL(generic_file_open);
1031
1032/*
1033 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001034 * file descriptors. The function is not supposed to ever fail, the only
1035 * reason it returns an 'int' and not 'void' is so that it can be plugged
1036 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 */
1038int nonseekable_open(struct inode *inode, struct file *filp)
1039{
1040 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1041 return 0;
1042}
1043
1044EXPORT_SYMBOL(nonseekable_open);