blob: 1e914b397e129f311875eb15d0c7679ace93aa3e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/string.h>
8#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040010#include <linux/fdtable.h>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/tty.h>
14#include <linux/namei.h>
15#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080016#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070017#include <linux/securebits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080020#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090021#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040030#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050031#include <linux/ima.h>
Eric Paris2dfc1ca2009-12-17 20:30:52 -050032#include <linux/dnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Eric Parise81e3f42009-12-04 15:47:36 -050034#include "internal.h"
35
NeilBrown4a301312006-01-08 01:02:39 -080036int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
37 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070038{
Amerigo Wang939a9422009-08-20 19:29:03 -070039 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 struct iattr newattrs;
41
42 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
43 if (length < 0)
44 return -EINVAL;
45
46 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080047 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080048 if (filp) {
49 newattrs.ia_file = filp;
50 newattrs.ia_valid |= ATTR_FILE;
51 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Linus Torvalds7b82dc02007-05-08 20:10:00 -070053 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -070054 ret = should_remove_suid(dentry);
55 if (ret)
56 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070057
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080058 mutex_lock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070059 ret = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080060 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070061 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062}
63
Al Viro2d8f3032008-07-22 09:59:21 -040064static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
Al Viro2d8f3032008-07-22 09:59:21 -040066 struct path path;
67 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 int error;
69
70 error = -EINVAL;
71 if (length < 0) /* sorry, but loff_t says... */
72 goto out;
73
Al Viro2d8f3032008-07-22 09:59:21 -040074 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (error)
76 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -040077 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
80 error = -EISDIR;
81 if (S_ISDIR(inode->i_mode))
82 goto dput_and_out;
83
84 error = -EINVAL;
85 if (!S_ISREG(inode->i_mode))
86 goto dput_and_out;
87
Al Viro2d8f3032008-07-22 09:59:21 -040088 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 if (error)
90 goto dput_and_out;
91
Al Viro256984a2008-07-22 08:09:30 -040092 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080093 if (error)
94 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020097 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080098 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 error = get_write_access(inode);
101 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800102 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
david m. richter97003822007-07-31 00:39:12 -0700104 /*
105 * Make sure that there are no leases. get_write_access() protects
106 * against the truncate racing with a lease-granting setlease().
107 */
Al Viro8737c932009-12-24 06:47:55 -0500108 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700109 if (error)
110 goto put_write_and_out;
111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900113 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900114 error = security_path_truncate(&path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500115 if (!error)
Al Viro2d8f3032008-07-22 09:59:21 -0400116 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
david m. richter97003822007-07-31 00:39:12 -0700118put_write_and_out:
119 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800120mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400121 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400123 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124out:
125 return error;
126}
127
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200128SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200130 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131}
132
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800133static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134{
135 struct inode * inode;
136 struct dentry *dentry;
137 struct file * file;
138 int error;
139
140 error = -EINVAL;
141 if (length < 0)
142 goto out;
143 error = -EBADF;
144 file = fget(fd);
145 if (!file)
146 goto out;
147
148 /* explicitly opened as large or we are on 64-bit box */
149 if (file->f_flags & O_LARGEFILE)
150 small = 0;
151
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800152 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 inode = dentry->d_inode;
154 error = -EINVAL;
155 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
156 goto out_putf;
157
158 error = -EINVAL;
159 /* Cannot ftruncate over 2^31 bytes without large file support */
160 if (small && length > MAX_NON_LFS)
161 goto out_putf;
162
163 error = -EPERM;
164 if (IS_APPEND(inode))
165 goto out_putf;
166
167 error = locks_verify_truncate(inode, file, length);
168 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900169 error = security_path_truncate(&file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900170 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700171 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172out_putf:
173 fput(file);
174out:
175 return error;
176}
177
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100178SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700180 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700181 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700182 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700183 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184}
185
186/* LFS versions of truncate are only needed on 32 bit machines */
187#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100188SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
190 return do_sys_truncate(path, length);
191}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100192#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
193asmlinkage long SyS_truncate64(long path, loff_t length)
194{
195 return SYSC_truncate64((const char __user *) path, length);
196}
197SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
198#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100200SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700202 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700203 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700204 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700205 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100207#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
208asmlinkage long SyS_ftruncate64(long fd, loff_t length)
209{
210 return SYSC_ftruncate64((unsigned int) fd, length);
211}
212SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100214#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400216
217int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400218{
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400219 struct inode *inode = file->f_path.dentry->d_inode;
220 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400221
222 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400223 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400224
225 /* Return error if mode is not supported */
Josef Bacik79124f182010-11-17 20:46:15 -0500226 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
227 return -EOPNOTSUPP;
228
229 /* Punch hole must have keep size set */
230 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
231 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400232 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400233
Amit Arora97ac7352007-07-17 21:42:44 -0400234 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400235 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100236
237 /* It's not possible punch hole on append only file */
238 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
239 return -EPERM;
240
241 if (IS_IMMUTABLE(inode))
242 return -EPERM;
243
Amit Arora97ac7352007-07-17 21:42:44 -0400244 /*
245 * Revalidate the write permissions, in case security policy has
246 * changed since the files were opened.
247 */
248 ret = security_file_permission(file, MAY_WRITE);
249 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400250 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400251
Amit Arora97ac7352007-07-17 21:42:44 -0400252 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400253 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400254
Amit Arora97ac7352007-07-17 21:42:44 -0400255 /*
256 * Let individual file system decide if it supports preallocation
257 * for directories or not.
258 */
259 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400260 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400261
Amit Arora97ac7352007-07-17 21:42:44 -0400262 /* Check for wrap through zero too */
263 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400264 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400265
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100266 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400267 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400268
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100269 return file->f_op->fallocate(file, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400270}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400271
272SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
273{
274 struct file *file;
275 int error = -EBADF;
276
277 file = fget(fd);
278 if (file) {
279 error = do_fallocate(file, mode, offset, len);
280 fput(file);
281 }
282
283 return error;
284}
285
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100286#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
287asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
288{
289 return SYSC_fallocate((int)fd, (int)mode, offset, len);
290}
291SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
292#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294/*
295 * access() needs to use the real uid/gid, not the effective uid/gid.
296 * We do this by temporarily clearing all FS-related capabilities and
297 * switching the fsuid/fsgid around to the real ones.
298 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100299SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300{
David Howellsd84f4f92008-11-14 10:39:23 +1100301 const struct cred *old_cred;
302 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400303 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400304 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 int res;
306
307 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
308 return -EINVAL;
309
David Howellsd84f4f92008-11-14 10:39:23 +1100310 override_cred = prepare_creds();
311 if (!override_cred)
312 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
David Howellsd84f4f92008-11-14 10:39:23 +1100314 override_cred->fsuid = override_cred->uid;
315 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700317 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100318 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800319 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
320 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100321 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700322 else
David Howellsd84f4f92008-11-14 10:39:23 +1100323 override_cred->cap_effective =
324 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700325 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326
David Howellsd84f4f92008-11-14 10:39:23 +1100327 old_cred = override_creds(override_cred);
328
Al Viro2d8f3032008-07-22 09:59:21 -0400329 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700330 if (res)
331 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Al Viro2d8f3032008-07-22 09:59:21 -0400333 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400334
335 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400336 /*
337 * MAY_EXEC on regular files is denied if the fs is mounted
338 * with the "noexec" flag.
339 */
340 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400341 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400342 goto out_path_release;
343 }
344
Al Viro256984a2008-07-22 08:09:30 -0400345 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700346 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400347 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700348 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800349 /*
350 * This is a rare case where using __mnt_is_readonly()
351 * is OK without a mnt_want/drop_write() pair. Since
352 * no actual write to the fs is performed here, we do
353 * not need to telegraph to that to anyone.
354 *
355 * By doing this, we accept that this access is
356 * inherently racy and know that the fs may change
357 * state before we even see this result.
358 */
Al Viro2d8f3032008-07-22 09:59:21 -0400359 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700360 res = -EROFS;
361
362out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400363 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700364out:
David Howellsd84f4f92008-11-14 10:39:23 +1100365 revert_creds(old_cred);
366 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 return res;
368}
369
Heiko Carstensca013e92009-01-14 14:14:19 +0100370SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800371{
372 return sys_faccessat(AT_FDCWD, filename, mode);
373}
374
Heiko Carstens3cdad422009-01-14 14:14:22 +0100375SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376{
Al Viro2d8f3032008-07-22 09:59:21 -0400377 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 int error;
379
Al Viro2d8f3032008-07-22 09:59:21 -0400380 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 if (error)
382 goto out;
383
Eric Paris9cfcac82010-07-23 11:43:51 -0400384 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 if (error)
386 goto dput_and_out;
387
Al Viro2d8f3032008-07-22 09:59:21 -0400388 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
390dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400391 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392out:
393 return error;
394}
395
Heiko Carstens3cdad422009-01-14 14:14:22 +0100396SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397{
398 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 struct inode *inode;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700400 int error, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
402 error = -EBADF;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700403 file = fget_raw_light(fd, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 if (!file)
405 goto out;
406
Jan Blunckac748a02008-02-14 19:34:39 -0800407 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
409 error = -ENOTDIR;
410 if (!S_ISDIR(inode->i_mode))
411 goto out_putf;
412
Eric Paris9cfcac82010-07-23 11:43:51 -0400413 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800415 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416out_putf:
Linus Torvalds332a2e12012-07-07 10:17:00 -0700417 fput_light(file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418out:
419 return error;
420}
421
Heiko Carstens3480b252009-01-14 14:14:16 +0100422SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423{
Al Viro2d8f3032008-07-22 09:59:21 -0400424 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 int error;
426
Al Viro2d8f3032008-07-22 09:59:21 -0400427 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 if (error)
429 goto out;
430
Eric Paris9cfcac82010-07-23 11:43:51 -0400431 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (error)
433 goto dput_and_out;
434
435 error = -EPERM;
436 if (!capable(CAP_SYS_CHROOT))
437 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900438 error = security_path_chroot(&path);
439 if (error)
440 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Al Viro2d8f3032008-07-22 09:59:21 -0400442 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 error = 0;
444dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400445 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446out:
447 return error;
448}
449
Al Viroe57712e2011-07-26 04:15:54 -0400450static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
Al Viroe57712e2011-07-26 04:15:54 -0400452 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400454 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
Al Viroe57712e2011-07-26 04:15:54 -0400456 error = mnt_want_write(path->mnt);
457 if (error)
458 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900459 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500460 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400461 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900462 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
464 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400465 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900466out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800467 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400468 mnt_drop_write(path->mnt);
469 return error;
470}
471
Al Viro49f0a072011-07-26 04:22:01 -0400472SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400473{
474 struct file * file;
475 int err = -EBADF;
476
477 file = fget(fd);
478 if (file) {
479 audit_inode(NULL, file->f_path.dentry);
480 err = chmod_common(&file->f_path, mode);
481 fput(file);
482 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 return err;
484}
485
Al Viro49f0a072011-07-26 04:22:01 -0400486SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Al Viro2d8f3032008-07-22 09:59:21 -0400488 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Al Viro2d8f3032008-07-22 09:59:21 -0400491 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400492 if (!error) {
493 error = chmod_common(&path, mode);
494 path_put(&path);
495 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 return error;
497}
498
Al Viro49f0a072011-07-26 04:22:01 -0400499SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800500{
501 return sys_fchmodat(AT_FDCWD, filename, mode);
502}
503
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900504static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900506 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 int error;
508 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800509 kuid_t uid;
510 kgid_t gid;
511
512 uid = make_kuid(current_user_ns(), user);
513 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 newattrs.ia_valid = ATTR_CTIME;
516 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800517 if (!uid_valid(uid))
518 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800520 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 }
522 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800523 if (!gid_valid(gid))
524 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800526 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 }
528 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700529 newattrs.ia_valid |=
530 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800531 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900532 error = security_path_chown(path, user, group);
533 if (!error)
534 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800535 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 return error;
538}
539
Heiko Carstens6559eed82009-01-14 14:14:32 +0100540SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
541 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800542{
Al Viro2d8f3032008-07-22 09:59:21 -0400543 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800544 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400545 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800546
Al Viro65cfc672011-03-13 15:56:26 -0400547 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800548 goto out;
549
Al Viro65cfc672011-03-13 15:56:26 -0400550 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
551 if (flag & AT_EMPTY_PATH)
552 lookup_flags |= LOOKUP_EMPTY;
553 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700554 if (error)
555 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400556 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800557 if (error)
558 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900559 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400560 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800561out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400562 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800563out:
564 return error;
565}
566
David Howells55e4def2012-06-25 12:55:09 +0100567SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
568{
569 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
570}
571
Heiko Carstensca013e92009-01-14 14:14:19 +0100572SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573{
David Howells55e4def2012-06-25 12:55:09 +0100574 return sys_fchownat(AT_FDCWD, filename, user, group,
575 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
577
Heiko Carstensca013e92009-01-14 14:14:19 +0100578SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579{
580 struct file * file;
581 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700582 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700585 if (!file)
586 goto out;
587
npiggin@suse.de96029c42009-04-26 20:25:55 +1000588 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800589 if (error)
590 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800591 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400592 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900593 error = chown_common(&file->f_path, user, group);
Al Viro2a79f172011-12-09 08:06:57 -0500594 mnt_drop_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800595out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700596 fput(file);
597out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 return error;
599}
600
Dave Hansen4a3fd212008-02-15 14:37:48 -0800601/*
602 * You have to be very careful that these write
603 * counts get cleaned up in error cases and
604 * upon __fput(). This should probably never
605 * be called outside of __dentry_open().
606 */
607static inline int __get_file_write_access(struct inode *inode,
608 struct vfsmount *mnt)
609{
610 int error;
611 error = get_write_access(inode);
612 if (error)
613 return error;
614 /*
615 * Do not take mount writer counts on
616 * special files since no writes to
617 * the mount itself will occur.
618 */
619 if (!special_file(inode->i_mode)) {
620 /*
621 * Balanced in __fput()
622 */
623 error = mnt_want_write(mnt);
624 if (error)
625 put_write_access(inode);
626 }
627 return error;
628}
629
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200630int open_check_o_direct(struct file *f)
631{
632 /* NB: we're sure to have correct a_ops only after f_op->open */
633 if (f->f_flags & O_DIRECT) {
634 if (!f->f_mapping->a_ops ||
635 ((!f->f_mapping->a_ops->direct_IO) &&
636 (!f->f_mapping->a_ops->get_xip_mem))) {
637 return -EINVAL;
638 }
639 }
640 return 0;
641}
642
Al Viro02e51802012-06-10 14:32:45 -0400643static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400644 int (*open)(struct inode *, struct file *),
645 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646{
Al Viro1abf0c72011-03-13 03:51:11 -0400647 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 struct inode *inode;
649 int error;
650
Al Viro53009902009-12-19 10:15:07 -0500651 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700652 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400653
654 if (unlikely(f->f_flags & O_PATH))
655 f->f_mode = FMODE_PATH;
656
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 Viro02e51802012-06-10 14:32:45 -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 Viro30d90492012-06-22 12:40:19 +0400742 mntget(file->f_path.mnt);
Al Viro02e51802012-06-10 14:32:45 -0400743 file->f_path.dentry = dget(dentry);
Miklos Szeredid18e9002012-06-05 15:10:17 +0200744
Al Viro02e51802012-06-10 14:32:45 -0400745 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400746 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400747 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200748
Al Viro96b7e572012-06-10 14:22:04 -0400749 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200750}
751EXPORT_SYMBOL(finish_open);
752
753/**
754 * finish_no_open - finish ->atomic_open() without opening the file
755 *
756 * @od: opaque open data
757 * @dentry: dentry or NULL (as returned from ->lookup())
758 *
759 * This can be used to set the result of a successful lookup in ->atomic_open().
760 * The filesystem's atomic_open() method shall return NULL after calling this.
761 */
Al Viroe45198a2012-06-10 06:48:09 -0400762int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200763{
Al Viro30d90492012-06-22 12:40:19 +0400764 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400765 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200766}
767EXPORT_SYMBOL(finish_no_open);
768
Al Viro765927b2012-06-26 21:58:53 +0400769struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100770 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700771{
772 int error;
773 struct file *f;
774
David Howellse0e81732009-09-02 09:13:40 +0100775 validate_creds(cred);
776
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900777 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400778 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800779
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700780 error = -ENFILE;
781 f = get_empty_filp();
Al Viro765927b2012-06-26 21:58:53 +0400782 if (f == NULL)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700783 return ERR_PTR(error);
784
Al Viro482928d2009-12-19 10:10:39 -0500785 f->f_flags = flags;
Al Viro765927b2012-06-26 21:58:53 +0400786 f->f_path = *path;
787 path_get(&f->f_path);
Al Viro02e51802012-06-10 14:32:45 -0400788 error = do_dentry_open(f, NULL, cred);
Al Viro2a027e72012-06-10 14:24:38 -0400789 if (!error) {
790 error = open_check_o_direct(f);
791 if (error) {
792 fput(f);
793 f = ERR_PTR(error);
794 }
795 } else {
796 put_filp(f);
797 f = ERR_PTR(error);
798 }
799 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700800}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801EXPORT_SYMBOL(dentry_open);
802
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800803static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700805 struct fdtable *fdt = files_fdtable(files);
David Howells1dce27c2012-02-16 17:49:42 +0000806 __clear_open_fd(fd, fdt);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800807 if (fd < files->next_fd)
808 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809}
810
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800811void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
813 struct files_struct *files = current->files;
814 spin_lock(&files->file_lock);
815 __put_unused_fd(files, fd);
816 spin_unlock(&files->file_lock);
817}
818
819EXPORT_SYMBOL(put_unused_fd);
820
821/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800822 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 *
824 * The VFS is full of places where we drop the files lock between
825 * setting the open_fds bitmap and installing the file in the file
826 * array. At any such point, we are vulnerable to a dup2() race
827 * installing a file in the array before us. We need to detect this and
828 * fput() the struct file we are about to overwrite in this case.
829 *
830 * It should never happen - if we allow dup2() do it, _really_ bad things
831 * will follow.
832 */
833
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800834void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
836 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700837 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700839 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700840 BUG_ON(fdt->fd[fd] != NULL);
841 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 spin_unlock(&files->file_lock);
843}
844
845EXPORT_SYMBOL(fd_install);
846
Al Viroa218d0f2011-11-21 14:59:34 -0500847static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500848{
849 int lookup_flags = 0;
850 int acc_mode;
851
852 if (!(flags & O_CREAT))
853 mode = 0;
854 op->mode = mode;
855
856 /* Must never be set by userspace */
857 flags &= ~FMODE_NONOTIFY;
858
859 /*
860 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
861 * check for O_DSYNC if the need any syncing at all we enforce it's
862 * always set instead of having to deal with possibly weird behaviour
863 * for malicious applications setting only __O_SYNC.
864 */
865 if (flags & __O_SYNC)
866 flags |= O_DSYNC;
867
Al Viro1abf0c72011-03-13 03:51:11 -0400868 /*
869 * If we have O_PATH in the open flag. Then we
870 * cannot have anything other than the below set of flags
871 */
872 if (flags & O_PATH) {
873 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
874 acc_mode = 0;
875 } else {
876 acc_mode = MAY_OPEN | ACC_MODE(flags);
877 }
Al Viro47c805d2011-02-23 17:44:09 -0500878
Al Viro1abf0c72011-03-13 03:51:11 -0400879 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500880
881 /* O_TRUNC implies we need access checks for write permissions */
882 if (flags & O_TRUNC)
883 acc_mode |= MAY_WRITE;
884
885 /* Allow the LSM permission hook to distinguish append
886 access from general write access. */
887 if (flags & O_APPEND)
888 acc_mode |= MAY_APPEND;
889
890 op->acc_mode = acc_mode;
891
Al Viro1abf0c72011-03-13 03:51:11 -0400892 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
893
Al Viro47c805d2011-02-23 17:44:09 -0500894 if (flags & O_CREAT) {
895 op->intent |= LOOKUP_CREATE;
896 if (flags & O_EXCL)
897 op->intent |= LOOKUP_EXCL;
898 }
899
900 if (flags & O_DIRECTORY)
901 lookup_flags |= LOOKUP_DIRECTORY;
902 if (!(flags & O_NOFOLLOW))
903 lookup_flags |= LOOKUP_FOLLOW;
904 return lookup_flags;
905}
906
907/**
908 * filp_open - open file and return file pointer
909 *
910 * @filename: path to open
911 * @flags: open flags as per the open(2) second argument
912 * @mode: mode for the new file if O_CREAT is set, else ignored
913 *
914 * This is the helper to open a file from kernelspace if you really
915 * have to. But in generally you should not do this, so please move
916 * along, nothing to see here..
917 */
Al Viroa218d0f2011-11-21 14:59:34 -0500918struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500919{
920 struct open_flags op;
921 int lookup = build_open_flags(flags, mode, &op);
922 return do_filp_open(AT_FDCWD, filename, &op, lookup);
923}
924EXPORT_SYMBOL(filp_open);
925
Al Viro73d049a2011-03-11 12:08:24 -0500926struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
927 const char *filename, int flags)
928{
929 struct open_flags op;
930 int lookup = build_open_flags(flags, 0, &op);
931 if (flags & O_CREAT)
932 return ERR_PTR(-EINVAL);
933 if (!filename && (flags & O_DIRECTORY))
934 if (!dentry->d_inode->i_op->lookup)
935 return ERR_PTR(-ENOTDIR);
936 return do_file_open_root(dentry, mnt, filename, &op, lookup);
937}
938EXPORT_SYMBOL(file_open_root);
939
Al Viroa218d0f2011-11-21 14:59:34 -0500940long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941{
Al Viro47c805d2011-02-23 17:44:09 -0500942 struct open_flags op;
943 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -0700944 char *tmp = getname(filename);
945 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700948 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -0500950 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700951 if (IS_ERR(f)) {
952 put_unused_fd(fd);
953 fd = PTR_ERR(f);
954 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500955 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700956 fd_install(fd, f);
957 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 putname(tmp);
960 }
961 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700963
Al Viroa218d0f2011-11-21 14:59:34 -0500964SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700965{
Linus Torvalds385910f2006-04-18 13:22:59 -0700966 long ret;
967
Miklos Szeredie922efc2005-09-06 15:18:25 -0700968 if (force_o_largefile())
969 flags |= O_LARGEFILE;
970
Linus Torvalds385910f2006-04-18 13:22:59 -0700971 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
972 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700973 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700974 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -0700975}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
Heiko Carstens6559eed82009-01-14 14:14:32 +0100977SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -0500978 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800979{
Linus Torvalds385910f2006-04-18 13:22:59 -0700980 long ret;
981
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800982 if (force_o_largefile())
983 flags |= O_LARGEFILE;
984
Linus Torvalds385910f2006-04-18 13:22:59 -0700985 ret = do_sys_open(dfd, filename, flags, mode);
986 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700987 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700988 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800989}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991#ifndef __alpha__
992
993/*
994 * For backward compatibility? Maybe this should be moved
995 * into arch/i386 instead?
996 */
Al Viroa218d0f2011-11-21 14:59:34 -0500997SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998{
999 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1000}
1001
1002#endif
1003
1004/*
1005 * "id" is the POSIX thread ID. We use the
1006 * files pointer for this..
1007 */
1008int filp_close(struct file *filp, fl_owner_t id)
1009{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001010 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
1012 if (!file_count(filp)) {
1013 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001014 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 }
1016
Christoph Lameter45778ca2005-06-23 00:10:17 -07001017 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001018 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
Al Viro1abf0c72011-03-13 03:51:11 -04001020 if (likely(!(filp->f_mode & FMODE_PATH))) {
1021 dnotify_flush(filp, id);
1022 locks_remove_posix(filp, id);
1023 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 fput(filp);
1025 return retval;
1026}
1027
1028EXPORT_SYMBOL(filp_close);
1029
1030/*
1031 * Careful here! We test whether the file pointer is NULL before
1032 * releasing the fd. This ensures that one clone task can't release
1033 * an fd while another clone is opening it.
1034 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001035SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036{
1037 struct file * filp;
1038 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001039 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001040 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041
1042 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001043 fdt = files_fdtable(files);
1044 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001046 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 if (!filp)
1048 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001049 rcu_assign_pointer(fdt->fd[fd], NULL);
David Howells1dce27c2012-02-16 17:49:42 +00001050 __clear_close_on_exec(fd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 __put_unused_fd(files, fd);
1052 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001053 retval = filp_close(filp, files);
1054
1055 /* can't restart close syscall because file table entry was cleared */
1056 if (unlikely(retval == -ERESTARTSYS ||
1057 retval == -ERESTARTNOINTR ||
1058 retval == -ERESTARTNOHAND ||
1059 retval == -ERESTART_RESTARTBLOCK))
1060 retval = -EINTR;
1061
1062 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063
1064out_unlock:
1065 spin_unlock(&files->file_lock);
1066 return -EBADF;
1067}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068EXPORT_SYMBOL(sys_close);
1069
1070/*
1071 * This routine simulates a hangup on the tty, to arrange that users
1072 * are given clean terminals at login time.
1073 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001074SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075{
1076 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001077 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 return 0;
1079 }
1080 return -EPERM;
1081}
1082
1083/*
1084 * Called when an inode is about to be open.
1085 * We use this to disallow opening large files on 32bit systems if
1086 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1087 * on this flag in sys_open.
1088 */
1089int generic_file_open(struct inode * inode, struct file * filp)
1090{
1091 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001092 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 return 0;
1094}
1095
1096EXPORT_SYMBOL(generic_file_open);
1097
1098/*
1099 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001100 * file descriptors. The function is not supposed to ever fail, the only
1101 * reason it returns an 'int' and not 'void' is so that it can be plugged
1102 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 */
1104int nonseekable_open(struct inode *inode, struct file *filp)
1105{
1106 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1107 return 0;
1108}
1109
1110EXPORT_SYMBOL(nonseekable_open);