blob: 7931f76acc2beede0298747df1d505b30cd225b4 [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>
Al Viro3f6d0782013-02-24 13:49:08 -050033#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Eric Parise81e3f42009-12-04 15:47:36 -050035#include "internal.h"
36
NeilBrown4a301312006-01-08 01:02:39 -080037int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
38 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070039{
Amerigo Wang939a9422009-08-20 19:29:03 -070040 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 struct iattr newattrs;
42
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
44 if (length < 0)
45 return -EINVAL;
46
47 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080048 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080049 if (filp) {
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
52 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Linus Torvalds7b82dc02007-05-08 20:10:00 -070054 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -070055 ret = should_remove_suid(dentry);
56 if (ret)
57 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070058
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080059 mutex_lock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070060 ret = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080061 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070062 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063}
64
David Howellsa02de962012-12-20 21:52:36 +000065long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070066{
Al Viro2d8f3032008-07-22 09:59:21 -040067 struct inode *inode;
David Howellsa02de962012-12-20 21:52:36 +000068 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
David Howellsa02de962012-12-20 21:52:36 +000070 inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
72 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 if (S_ISDIR(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000074 return -EISDIR;
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (!S_ISREG(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000076 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
David Howellsa02de962012-12-20 21:52:36 +000078 error = mnt_want_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 if (error)
David Howellsa02de962012-12-20 21:52:36 +000080 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
Al Viro256984a2008-07-22 08:09:30 -040082 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080083 if (error)
84 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
86 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020087 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080088 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 error = get_write_access(inode);
91 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -080092 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
david m. richter97003822007-07-31 00:39:12 -070094 /*
95 * Make sure that there are no leases. get_write_access() protects
96 * against the truncate racing with a lease-granting setlease().
97 */
Al Viro8737c932009-12-24 06:47:55 -050098 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -070099 if (error)
100 goto put_write_and_out;
101
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900103 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000104 error = security_path_truncate(path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500105 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000106 error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107
david m. richter97003822007-07-31 00:39:12 -0700108put_write_and_out:
109 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800110mnt_drop_write_and_out:
David Howellsa02de962012-12-20 21:52:36 +0000111 mnt_drop_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112out:
113 return error;
114}
David Howellsa02de962012-12-20 21:52:36 +0000115EXPORT_SYMBOL_GPL(vfs_truncate);
116
117static long do_sys_truncate(const char __user *pathname, loff_t length)
118{
Jeff Layton48f75302012-12-11 12:10:11 -0500119 unsigned int lookup_flags = LOOKUP_FOLLOW;
David Howellsa02de962012-12-20 21:52:36 +0000120 struct path path;
121 int error;
122
123 if (length < 0) /* sorry, but loff_t says... */
124 return -EINVAL;
125
Jeff Layton48f75302012-12-11 12:10:11 -0500126retry:
127 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
David Howellsa02de962012-12-20 21:52:36 +0000128 if (!error) {
129 error = vfs_truncate(&path, length);
130 path_put(&path);
131 }
Jeff Layton48f75302012-12-11 12:10:11 -0500132 if (retry_estale(error, lookup_flags)) {
133 lookup_flags |= LOOKUP_REVAL;
134 goto retry;
135 }
David Howellsa02de962012-12-20 21:52:36 +0000136 return error;
137}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200139SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200141 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142}
143
Al Viro3f6d0782013-02-24 13:49:08 -0500144#ifdef CONFIG_COMPAT
145COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
146{
147 return do_sys_truncate(path, length);
148}
149#endif
150
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800151static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152{
Al Virobf2965d2012-08-26 20:13:36 -0400153 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -0400155 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 int error;
157
158 error = -EINVAL;
159 if (length < 0)
160 goto out;
161 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400162 f = fdget(fd);
163 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 goto out;
165
166 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400167 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 small = 0;
169
Al Viro2903ff02012-08-28 12:52:22 -0400170 dentry = f.file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 inode = dentry->d_inode;
172 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400173 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 goto out_putf;
175
176 error = -EINVAL;
177 /* Cannot ftruncate over 2^31 bytes without large file support */
178 if (small && length > MAX_NON_LFS)
179 goto out_putf;
180
181 error = -EPERM;
182 if (IS_APPEND(inode))
183 goto out_putf;
184
Jan Kara14da9202012-06-12 16:20:37 +0200185 sb_start_write(inode->i_sb);
Al Viro2903ff02012-08-28 12:52:22 -0400186 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400188 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900189 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400190 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200191 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400193 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194out:
195 return error;
196}
197
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100198SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199{
Al Viro2cf09662013-01-21 15:25:54 -0500200 return do_sys_ftruncate(fd, length, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201}
202
Al Viro3f6d0782013-02-24 13:49:08 -0500203#ifdef CONFIG_COMPAT
204COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
205{
206 return do_sys_ftruncate(fd, length, 1);
207}
208#endif
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210/* LFS versions of truncate are only needed on 32 bit machines */
211#if BITS_PER_LONG == 32
Al Viro4a0fd5b2013-01-21 15:16:58 -0500212SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
214 return do_sys_truncate(path, length);
215}
216
Al Viro4a0fd5b2013-01-21 15:16:58 -0500217SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218{
Al Viro2cf09662013-01-21 15:25:54 -0500219 return do_sys_ftruncate(fd, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100221#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400223
224int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400225{
Al Viro496ad9a2013-01-23 17:07:38 -0500226 struct inode *inode = file_inode(file);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400227 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400228
229 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400230 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400231
232 /* Return error if mode is not supported */
Josef Bacik79124f182010-11-17 20:46:15 -0500233 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
234 return -EOPNOTSUPP;
235
236 /* Punch hole must have keep size set */
237 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
238 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400239 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400240
Amit Arora97ac7352007-07-17 21:42:44 -0400241 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400242 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100243
244 /* It's not possible punch hole on append only file */
245 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
246 return -EPERM;
247
248 if (IS_IMMUTABLE(inode))
249 return -EPERM;
250
Amit Arora97ac7352007-07-17 21:42:44 -0400251 /*
252 * Revalidate the write permissions, in case security policy has
253 * changed since the files were opened.
254 */
255 ret = security_file_permission(file, MAY_WRITE);
256 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400257 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400258
Amit Arora97ac7352007-07-17 21:42:44 -0400259 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400260 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400261
Amit Arora97ac7352007-07-17 21:42:44 -0400262 /*
263 * Let individual file system decide if it supports preallocation
264 * for directories or not.
265 */
266 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400267 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400268
Amit Arora97ac7352007-07-17 21:42:44 -0400269 /* Check for wrap through zero too */
270 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400271 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400272
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100273 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400274 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400275
Jan Kara14da9202012-06-12 16:20:37 +0200276 sb_start_write(inode->i_sb);
277 ret = file->f_op->fallocate(file, mode, offset, len);
278 sb_end_write(inode->i_sb);
279 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400280}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400281
Al Viro4a0fd5b2013-01-21 15:16:58 -0500282SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400283{
Al Viro2903ff02012-08-28 12:52:22 -0400284 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400285 int error = -EBADF;
286
Al Viro2903ff02012-08-28 12:52:22 -0400287 if (f.file) {
288 error = do_fallocate(f.file, mode, offset, len);
289 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400290 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400291 return error;
292}
293
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;
Jeff Layton87fa5592012-12-11 12:10:11 -0500306 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
308 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
309 return -EINVAL;
310
David Howellsd84f4f92008-11-14 10:39:23 +1100311 override_cred = prepare_creds();
312 if (!override_cred)
313 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
David Howellsd84f4f92008-11-14 10:39:23 +1100315 override_cred->fsuid = override_cred->uid;
316 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700318 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100319 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800320 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
321 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100322 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700323 else
David Howellsd84f4f92008-11-14 10:39:23 +1100324 override_cred->cap_effective =
325 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700326 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327
David Howellsd84f4f92008-11-14 10:39:23 +1100328 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500329retry:
330 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700331 if (res)
332 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
Al Viro2d8f3032008-07-22 09:59:21 -0400334 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400335
336 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400337 /*
338 * MAY_EXEC on regular files is denied if the fs is mounted
339 * with the "noexec" flag.
340 */
341 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400342 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400343 goto out_path_release;
344 }
345
Al Viro256984a2008-07-22 08:09:30 -0400346 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700347 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400348 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700349 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800350 /*
351 * This is a rare case where using __mnt_is_readonly()
352 * is OK without a mnt_want/drop_write() pair. Since
353 * no actual write to the fs is performed here, we do
354 * not need to telegraph to that to anyone.
355 *
356 * By doing this, we accept that this access is
357 * inherently racy and know that the fs may change
358 * state before we even see this result.
359 */
Al Viro2d8f3032008-07-22 09:59:21 -0400360 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700361 res = -EROFS;
362
363out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400364 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500365 if (retry_estale(res, lookup_flags)) {
366 lookup_flags |= LOOKUP_REVAL;
367 goto retry;
368 }
Dave Hansen6902d922006-09-30 23:29:01 -0700369out:
David Howellsd84f4f92008-11-14 10:39:23 +1100370 revert_creds(old_cred);
371 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 return res;
373}
374
Heiko Carstensca013e92009-01-14 14:14:19 +0100375SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800376{
377 return sys_faccessat(AT_FDCWD, filename, mode);
378}
379
Heiko Carstens3cdad422009-01-14 14:14:22 +0100380SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381{
Al Viro2d8f3032008-07-22 09:59:21 -0400382 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500384 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
385retry:
386 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 if (error)
388 goto out;
389
Eric Paris9cfcac82010-07-23 11:43:51 -0400390 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 if (error)
392 goto dput_and_out;
393
Al Viro2d8f3032008-07-22 09:59:21 -0400394 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
396dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400397 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500398 if (retry_estale(error, lookup_flags)) {
399 lookup_flags |= LOOKUP_REVAL;
400 goto retry;
401 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402out:
403 return error;
404}
405
Heiko Carstens3cdad422009-01-14 14:14:22 +0100406SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407{
Al Viro2903ff02012-08-28 12:52:22 -0400408 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400410 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
412 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400413 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 goto out;
415
Al Viro496ad9a2013-01-23 17:07:38 -0500416 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
418 error = -ENOTDIR;
419 if (!S_ISDIR(inode->i_mode))
420 goto out_putf;
421
Eric Paris9cfcac82010-07-23 11:43:51 -0400422 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400424 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400426 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427out:
428 return error;
429}
430
Heiko Carstens3480b252009-01-14 14:14:16 +0100431SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
Al Viro2d8f3032008-07-22 09:59:21 -0400433 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500435 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
436retry:
437 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 if (error)
439 goto out;
440
Eric Paris9cfcac82010-07-23 11:43:51 -0400441 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 if (error)
443 goto dput_and_out;
444
445 error = -EPERM;
Eric W. Biedermana85fb272012-07-31 01:14:12 -0700446 if (!nsown_capable(CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900448 error = security_path_chroot(&path);
449 if (error)
450 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
Al Viro2d8f3032008-07-22 09:59:21 -0400452 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 error = 0;
454dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400455 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500456 if (retry_estale(error, lookup_flags)) {
457 lookup_flags |= LOOKUP_REVAL;
458 goto retry;
459 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460out:
461 return error;
462}
463
Al Viroe57712e2011-07-26 04:15:54 -0400464static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465{
Al Viroe57712e2011-07-26 04:15:54 -0400466 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400468 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
Al Viroe57712e2011-07-26 04:15:54 -0400470 error = mnt_want_write(path->mnt);
471 if (error)
472 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900473 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500474 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400475 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900476 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
478 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400479 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900480out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800481 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400482 mnt_drop_write(path->mnt);
483 return error;
484}
485
Al Viro49f0a072011-07-26 04:22:01 -0400486SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400487{
488 struct file * file;
489 int err = -EBADF;
490
491 file = fget(fd);
492 if (file) {
Jeff Laytonbfcec702012-10-10 15:25:23 -0400493 audit_inode(NULL, file->f_path.dentry, 0);
Al Viroe57712e2011-07-26 04:15:54 -0400494 err = chmod_common(&file->f_path, mode);
495 fput(file);
496 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 return err;
498}
499
Al Viro49f0a072011-07-26 04:22:01 -0400500SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501{
Al Viro2d8f3032008-07-22 09:59:21 -0400502 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 int error;
Jeff Layton14ff6902012-12-11 12:10:13 -0500504 unsigned int lookup_flags = LOOKUP_FOLLOW;
505retry:
506 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400507 if (!error) {
508 error = chmod_common(&path, mode);
509 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500510 if (retry_estale(error, lookup_flags)) {
511 lookup_flags |= LOOKUP_REVAL;
512 goto retry;
513 }
Al Viroe57712e2011-07-26 04:15:54 -0400514 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 return error;
516}
517
Al Viro49f0a072011-07-26 04:22:01 -0400518SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800519{
520 return sys_fchmodat(AT_FDCWD, filename, mode);
521}
522
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900523static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900525 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 int error;
527 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800528 kuid_t uid;
529 kgid_t gid;
530
531 uid = make_kuid(current_user_ns(), user);
532 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 newattrs.ia_valid = ATTR_CTIME;
535 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800536 if (!uid_valid(uid))
537 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800539 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 }
541 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800542 if (!gid_valid(gid))
543 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800545 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 }
547 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700548 newattrs.ia_valid |=
549 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800550 mutex_lock(&inode->i_mutex);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600551 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900552 if (!error)
553 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800554 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return error;
557}
558
Heiko Carstens6559eed82009-01-14 14:14:32 +0100559SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
560 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800561{
Al Viro2d8f3032008-07-22 09:59:21 -0400562 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800563 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400564 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800565
Al Viro65cfc672011-03-13 15:56:26 -0400566 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800567 goto out;
568
Al Viro65cfc672011-03-13 15:56:26 -0400569 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
570 if (flag & AT_EMPTY_PATH)
571 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500572retry:
Al Viro65cfc672011-03-13 15:56:26 -0400573 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700574 if (error)
575 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400576 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800577 if (error)
578 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900579 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400580 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800581out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400582 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500583 if (retry_estale(error, lookup_flags)) {
584 lookup_flags |= LOOKUP_REVAL;
585 goto retry;
586 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800587out:
588 return error;
589}
590
David Howells55e4def2012-06-25 12:55:09 +0100591SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
592{
593 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
594}
595
Heiko Carstensca013e92009-01-14 14:14:19 +0100596SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597{
David Howells55e4def2012-06-25 12:55:09 +0100598 return sys_fchownat(AT_FDCWD, filename, user, group,
599 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600}
601
Heiko Carstensca013e92009-01-14 14:14:19 +0100602SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
Al Viro2903ff02012-08-28 12:52:22 -0400604 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 int error = -EBADF;
606
Al Viro2903ff02012-08-28 12:52:22 -0400607 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700608 goto out;
609
Al Viro2903ff02012-08-28 12:52:22 -0400610 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800611 if (error)
612 goto out_fput;
Jeff Laytonbfcec702012-10-10 15:25:23 -0400613 audit_inode(NULL, f.file->f_path.dentry, 0);
Al Viro2903ff02012-08-28 12:52:22 -0400614 error = chown_common(&f.file->f_path, user, group);
615 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800616out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400617 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700618out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 return error;
620}
621
Dave Hansen4a3fd212008-02-15 14:37:48 -0800622/*
623 * You have to be very careful that these write
624 * counts get cleaned up in error cases and
625 * upon __fput(). This should probably never
626 * be called outside of __dentry_open().
627 */
628static inline int __get_file_write_access(struct inode *inode,
629 struct vfsmount *mnt)
630{
631 int error;
632 error = get_write_access(inode);
633 if (error)
634 return error;
635 /*
636 * Do not take mount writer counts on
637 * special files since no writes to
638 * the mount itself will occur.
639 */
640 if (!special_file(inode->i_mode)) {
641 /*
642 * Balanced in __fput()
643 */
Jan Karaeb04c282012-06-12 16:20:35 +0200644 error = __mnt_want_write(mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800645 if (error)
646 put_write_access(inode);
647 }
648 return error;
649}
650
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200651int open_check_o_direct(struct file *f)
652{
653 /* NB: we're sure to have correct a_ops only after f_op->open */
654 if (f->f_flags & O_DIRECT) {
655 if (!f->f_mapping->a_ops ||
656 ((!f->f_mapping->a_ops->direct_IO) &&
657 (!f->f_mapping->a_ops->get_xip_mem))) {
658 return -EINVAL;
659 }
660 }
661 return 0;
662}
663
Al Viro02e51802012-06-10 14:32:45 -0400664static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400665 int (*open)(struct inode *, struct file *),
666 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667{
Al Viro1abf0c72011-03-13 03:51:11 -0400668 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 struct inode *inode;
670 int error;
671
Al Viro53009902009-12-19 10:15:07 -0500672 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700673 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400674
675 if (unlikely(f->f_flags & O_PATH))
676 f->f_mode = FMODE_PATH;
677
Al Virob5bcdda2012-07-20 23:28:46 +0400678 path_get(&f->f_path);
Al Virodd379782013-03-01 19:48:30 -0500679 inode = f->f_inode = f->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (f->f_mode & FMODE_WRITE) {
Al Viro02e51802012-06-10 14:32:45 -0400681 error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 if (error)
683 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800684 if (!special_file(inode->i_mode))
685 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 }
687
688 f->f_mapping = inode->i_mapping;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000689 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Al Viro1abf0c72011-03-13 03:51:11 -0400691 if (unlikely(f->f_mode & FMODE_PATH)) {
692 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400693 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400694 }
695
696 f->f_op = fops_get(inode->i_fop);
697
Eric Paris83d49852012-04-04 13:45:40 -0400698 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900699 if (error)
700 goto cleanup_all;
701
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400702 error = break_lease(inode, f->f_flags);
703 if (error)
704 goto cleanup_all;
705
Trond Myklebust834f2a42005-10-18 14:20:16 -0700706 if (!open && f->f_op)
707 open = f->f_op->open;
708 if (open) {
709 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 if (error)
711 goto cleanup_all;
712 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400713 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
714 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
717
718 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
719
Al Viro96b7e572012-06-10 14:22:04 -0400720 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
722cleanup_all:
723 fops_put(f->f_op);
Al Viroc3c4f692012-06-23 22:49:45 +0400724 file_sb_list_del(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800725 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800727 if (!special_file(inode->i_mode)) {
728 /*
729 * We don't consider this a real
730 * mnt_want/drop_write() pair
731 * because it all happenend right
732 * here, so just reset the state.
733 */
734 file_reset_write(f);
Al Virofe7c8052012-08-04 08:39:23 +0400735 __mnt_drop_write(f->f_path.mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800736 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800737 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400739 path_put(&f->f_path);
740 f->f_path.mnt = NULL;
741 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500742 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400743 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744}
745
Trond Myklebust834f2a42005-10-18 14:20:16 -0700746/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200747 * finish_open - finish opening a file
748 * @od: opaque open data
749 * @dentry: pointer to dentry
750 * @open: open callback
751 *
752 * This can be used to finish opening a file passed to i_op->atomic_open().
753 *
754 * If the open callback is set to NULL, then the standard f_op->open()
755 * filesystem callback is substituted.
756 */
Al Viro30d90492012-06-22 12:40:19 +0400757int finish_open(struct file *file, struct dentry *dentry,
758 int (*open)(struct inode *, struct file *),
759 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200760{
Al Viro96b7e572012-06-10 14:22:04 -0400761 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400762 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200763
Al Virob5bcdda2012-07-20 23:28:46 +0400764 file->f_path.dentry = dentry;
Al Viro02e51802012-06-10 14:32:45 -0400765 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400766 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400767 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200768
Al Viro96b7e572012-06-10 14:22:04 -0400769 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200770}
771EXPORT_SYMBOL(finish_open);
772
773/**
774 * finish_no_open - finish ->atomic_open() without opening the file
775 *
776 * @od: opaque open data
777 * @dentry: dentry or NULL (as returned from ->lookup())
778 *
779 * This can be used to set the result of a successful lookup in ->atomic_open().
780 * The filesystem's atomic_open() method shall return NULL after calling this.
781 */
Al Viroe45198a2012-06-10 06:48:09 -0400782int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200783{
Al Viro30d90492012-06-22 12:40:19 +0400784 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400785 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200786}
787EXPORT_SYMBOL(finish_no_open);
788
Al Viro765927b2012-06-26 21:58:53 +0400789struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100790 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700791{
792 int error;
793 struct file *f;
794
David Howellse0e81732009-09-02 09:13:40 +0100795 validate_creds(cred);
796
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900797 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400798 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800799
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700800 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500801 if (!IS_ERR(f)) {
802 f->f_flags = flags;
803 f->f_path = *path;
804 error = do_dentry_open(f, NULL, cred);
805 if (!error) {
806 /* from now on we need fput() to dispose of f */
807 error = open_check_o_direct(f);
808 if (error) {
809 fput(f);
810 f = ERR_PTR(error);
811 }
812 } else {
813 put_filp(f);
Al Viro2a027e72012-06-10 14:24:38 -0400814 f = ERR_PTR(error);
815 }
Al Viro2a027e72012-06-10 14:24:38 -0400816 }
817 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700818}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819EXPORT_SYMBOL(dentry_open);
820
Al Viroa218d0f2011-11-21 14:59:34 -0500821static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500822{
823 int lookup_flags = 0;
824 int acc_mode;
825
Andy Lutomirskie305f482013-08-01 21:07:52 -0700826 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200827 op->mode = (mode & S_IALLUGO) | S_IFREG;
828 else
829 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500830
831 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400832 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500833
834 /*
835 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
836 * check for O_DSYNC if the need any syncing at all we enforce it's
837 * always set instead of having to deal with possibly weird behaviour
838 * for malicious applications setting only __O_SYNC.
839 */
840 if (flags & __O_SYNC)
841 flags |= O_DSYNC;
842
Al Virobb458c62013-07-13 13:26:37 +0400843 if (flags & __O_TMPFILE) {
844 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400845 return -EINVAL;
846 acc_mode = MAY_OPEN | ACC_MODE(flags);
Al Viroba57ea62013-07-20 03:11:32 +0400847 if (!(acc_mode & MAY_WRITE))
848 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400849 } else if (flags & O_PATH) {
850 /*
851 * If we have O_PATH in the open flag. Then we
852 * cannot have anything other than the below set of flags
853 */
Al Viro1abf0c72011-03-13 03:51:11 -0400854 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
855 acc_mode = 0;
856 } else {
857 acc_mode = MAY_OPEN | ACC_MODE(flags);
858 }
Al Viro47c805d2011-02-23 17:44:09 -0500859
Al Viro1abf0c72011-03-13 03:51:11 -0400860 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500861
862 /* O_TRUNC implies we need access checks for write permissions */
863 if (flags & O_TRUNC)
864 acc_mode |= MAY_WRITE;
865
866 /* Allow the LSM permission hook to distinguish append
867 access from general write access. */
868 if (flags & O_APPEND)
869 acc_mode |= MAY_APPEND;
870
871 op->acc_mode = acc_mode;
872
Al Viro1abf0c72011-03-13 03:51:11 -0400873 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
874
Al Viro47c805d2011-02-23 17:44:09 -0500875 if (flags & O_CREAT) {
876 op->intent |= LOOKUP_CREATE;
877 if (flags & O_EXCL)
878 op->intent |= LOOKUP_EXCL;
879 }
880
881 if (flags & O_DIRECTORY)
882 lookup_flags |= LOOKUP_DIRECTORY;
883 if (!(flags & O_NOFOLLOW))
884 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +0400885 op->lookup_flags = lookup_flags;
886 return 0;
Al Viro47c805d2011-02-23 17:44:09 -0500887}
888
889/**
Jeff Layton669abf42012-10-10 16:43:10 -0400890 * file_open_name - open file and return file pointer
891 *
892 * @name: struct filename containing path to open
893 * @flags: open flags as per the open(2) second argument
894 * @mode: mode for the new file if O_CREAT is set, else ignored
895 *
896 * This is the helper to open a file from kernelspace if you really
897 * have to. But in generally you should not do this, so please move
898 * along, nothing to see here..
899 */
900struct file *file_open_name(struct filename *name, int flags, umode_t mode)
901{
902 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400903 int err = build_open_flags(flags, mode, &op);
904 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -0400905}
906
907/**
Al Viro47c805d2011-02-23 17:44:09 -0500908 * 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{
Jeff Layton669abf42012-10-10 16:43:10 -0400920 struct filename name = {.name = filename};
921 return file_open_name(&name, flags, mode);
Al Viro47c805d2011-02-23 17:44:09 -0500922}
923EXPORT_SYMBOL(filp_open);
924
Al Viro73d049a2011-03-11 12:08:24 -0500925struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
926 const char *filename, int flags)
927{
928 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400929 int err = build_open_flags(flags, 0, &op);
930 if (err)
931 return ERR_PTR(err);
Al Viro73d049a2011-03-11 12:08:24 -0500932 if (flags & O_CREAT)
933 return ERR_PTR(-EINVAL);
934 if (!filename && (flags & O_DIRECTORY))
935 if (!dentry->d_inode->i_op->lookup)
936 return ERR_PTR(-ENOTDIR);
Al Virof9652e12013-06-11 08:23:01 +0400937 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -0500938}
939EXPORT_SYMBOL(file_open_root);
940
Al Viroa218d0f2011-11-21 14:59:34 -0500941long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942{
Al Viro47c805d2011-02-23 17:44:09 -0500943 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400944 int fd = build_open_flags(flags, mode, &op);
945 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
Al Virof9652e12013-06-11 08:23:01 +0400947 if (fd)
948 return fd;
949
950 tmp = getname(filename);
951 if (IS_ERR(tmp))
952 return PTR_ERR(tmp);
953
954 fd = get_unused_fd_flags(flags);
955 if (fd >= 0) {
956 struct file *f = do_filp_open(dfd, tmp, &op);
957 if (IS_ERR(f)) {
958 put_unused_fd(fd);
959 fd = PTR_ERR(f);
960 } else {
961 fsnotify_open(f);
962 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 }
Al Virof9652e12013-06-11 08:23:01 +0400965 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700968
Al Viroa218d0f2011-11-21 14:59:34 -0500969SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700970{
971 if (force_o_largefile())
972 flags |= O_LARGEFILE;
973
Al Viro2cf09662013-01-21 15:25:54 -0500974 return do_sys_open(AT_FDCWD, filename, flags, mode);
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{
980 if (force_o_largefile())
981 flags |= O_LARGEFILE;
982
Al Viro2cf09662013-01-21 15:25:54 -0500983 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800984}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986#ifndef __alpha__
987
988/*
989 * For backward compatibility? Maybe this should be moved
990 * into arch/i386 instead?
991 */
Al Viroa218d0f2011-11-21 14:59:34 -0500992SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993{
994 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
995}
996
997#endif
998
999/*
1000 * "id" is the POSIX thread ID. We use the
1001 * files pointer for this..
1002 */
1003int filp_close(struct file *filp, fl_owner_t id)
1004{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001005 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
1007 if (!file_count(filp)) {
1008 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001009 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 }
1011
Christoph Lameter45778ca2005-06-23 00:10:17 -07001012 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001013 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Al Viro1abf0c72011-03-13 03:51:11 -04001015 if (likely(!(filp->f_mode & FMODE_PATH))) {
1016 dnotify_flush(filp, id);
1017 locks_remove_posix(filp, id);
1018 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 fput(filp);
1020 return retval;
1021}
1022
1023EXPORT_SYMBOL(filp_close);
1024
1025/*
1026 * Careful here! We test whether the file pointer is NULL before
1027 * releasing the fd. This ensures that one clone task can't release
1028 * an fd while another clone is opening it.
1029 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001030SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031{
Al Viro483ce1d2012-08-19 12:04:24 -04001032 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001033
1034 /* can't restart close syscall because file table entry was cleared */
1035 if (unlikely(retval == -ERESTARTSYS ||
1036 retval == -ERESTARTNOINTR ||
1037 retval == -ERESTARTNOHAND ||
1038 retval == -ERESTART_RESTARTBLOCK))
1039 retval = -EINTR;
1040
1041 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043EXPORT_SYMBOL(sys_close);
1044
1045/*
1046 * This routine simulates a hangup on the tty, to arrange that users
1047 * are given clean terminals at login time.
1048 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001049SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050{
1051 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001052 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 return 0;
1054 }
1055 return -EPERM;
1056}
1057
1058/*
1059 * Called when an inode is about to be open.
1060 * We use this to disallow opening large files on 32bit systems if
1061 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1062 * on this flag in sys_open.
1063 */
1064int generic_file_open(struct inode * inode, struct file * filp)
1065{
1066 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001067 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 return 0;
1069}
1070
1071EXPORT_SYMBOL(generic_file_open);
1072
1073/*
1074 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001075 * file descriptors. The function is not supposed to ever fail, the only
1076 * reason it returns an 'int' and not 'void' is so that it can be plugged
1077 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 */
1079int nonseekable_open(struct inode *inode, struct file *filp)
1080{
1081 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1082 return 0;
1083}
1084
1085EXPORT_SYMBOL(nonseekable_open);