blob: 74e5cd9f718e8b43eb17e8dad2f28074591fbaaf [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>
20#include <linux/vfs.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080021#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090022#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <asm/uaccess.h>
24#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070025#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/pagemap.h>
27#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070028#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000029#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040030#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040031#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050032#include <linux/ima.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Eric Parise81e3f42009-12-04 15:47:36 -050034#include "internal.h"
35
David Howells726c3342006-06-23 02:02:58 -070036int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070037{
38 int retval = -ENODEV;
39
David Howells726c3342006-06-23 02:02:58 -070040 if (dentry) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 retval = -ENOSYS;
David Howells726c3342006-06-23 02:02:58 -070042 if (dentry->d_sb->s_op->statfs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 memset(buf, 0, sizeof(*buf));
David Howells726c3342006-06-23 02:02:58 -070044 retval = security_sb_statfs(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 if (retval)
46 return retval;
David Howells726c3342006-06-23 02:02:58 -070047 retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 if (retval == 0 && buf->f_frsize == 0)
49 buf->f_frsize = buf->f_bsize;
50 }
51 }
52 return retval;
53}
54
55EXPORT_SYMBOL(vfs_statfs);
56
David Howells726c3342006-06-23 02:02:58 -070057static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -070058{
59 struct kstatfs st;
60 int retval;
61
David Howells726c3342006-06-23 02:02:58 -070062 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 if (retval)
64 return retval;
65
66 if (sizeof(*buf) == sizeof(st))
67 memcpy(buf, &st, sizeof(st));
68 else {
69 if (sizeof buf->f_blocks == 4) {
Jon Tollefsonf4a67cc2008-07-23 21:27:55 -070070 if ((st.f_blocks | st.f_bfree | st.f_bavail |
71 st.f_bsize | st.f_frsize) &
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 0xffffffff00000000ULL)
73 return -EOVERFLOW;
74 /*
75 * f_files and f_ffree may be -1; it's okay to stuff
76 * that into 32 bits
77 */
78 if (st.f_files != -1 &&
79 (st.f_files & 0xffffffff00000000ULL))
80 return -EOVERFLOW;
81 if (st.f_ffree != -1 &&
82 (st.f_ffree & 0xffffffff00000000ULL))
83 return -EOVERFLOW;
84 }
85
86 buf->f_type = st.f_type;
87 buf->f_bsize = st.f_bsize;
88 buf->f_blocks = st.f_blocks;
89 buf->f_bfree = st.f_bfree;
90 buf->f_bavail = st.f_bavail;
91 buf->f_files = st.f_files;
92 buf->f_ffree = st.f_ffree;
93 buf->f_fsid = st.f_fsid;
94 buf->f_namelen = st.f_namelen;
95 buf->f_frsize = st.f_frsize;
96 memset(buf->f_spare, 0, sizeof(buf->f_spare));
97 }
98 return 0;
99}
100
David Howells726c3342006-06-23 02:02:58 -0700101static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102{
103 struct kstatfs st;
104 int retval;
105
David Howells726c3342006-06-23 02:02:58 -0700106 retval = vfs_statfs(dentry, &st);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 if (retval)
108 return retval;
109
110 if (sizeof(*buf) == sizeof(st))
111 memcpy(buf, &st, sizeof(st));
112 else {
113 buf->f_type = st.f_type;
114 buf->f_bsize = st.f_bsize;
115 buf->f_blocks = st.f_blocks;
116 buf->f_bfree = st.f_bfree;
117 buf->f_bavail = st.f_bavail;
118 buf->f_files = st.f_files;
119 buf->f_ffree = st.f_ffree;
120 buf->f_fsid = st.f_fsid;
121 buf->f_namelen = st.f_namelen;
122 buf->f_frsize = st.f_frsize;
123 memset(buf->f_spare, 0, sizeof(buf->f_spare));
124 }
125 return 0;
126}
127
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100128SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
Al Viro2d8f3032008-07-22 09:59:21 -0400130 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 int error;
132
Al Viro2d8f3032008-07-22 09:59:21 -0400133 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 if (!error) {
135 struct statfs tmp;
Al Viro2d8f3032008-07-22 09:59:21 -0400136 error = vfs_statfs_native(path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
138 error = -EFAULT;
Al Viro2d8f3032008-07-22 09:59:21 -0400139 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 }
141 return error;
142}
143
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100144SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145{
Al Viro2d8f3032008-07-22 09:59:21 -0400146 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 long error;
148
149 if (sz != sizeof(*buf))
150 return -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -0400151 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 if (!error) {
153 struct statfs64 tmp;
Al Viro2d8f3032008-07-22 09:59:21 -0400154 error = vfs_statfs64(path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
156 error = -EFAULT;
Al Viro2d8f3032008-07-22 09:59:21 -0400157 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 }
159 return error;
160}
161
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100162SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163{
164 struct file * file;
165 struct statfs tmp;
166 int error;
167
168 error = -EBADF;
169 file = fget(fd);
170 if (!file)
171 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800172 error = vfs_statfs_native(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
174 error = -EFAULT;
175 fput(file);
176out:
177 return error;
178}
179
Heiko Carstens257ac262009-01-14 14:14:13 +0100180SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181{
182 struct file * file;
183 struct statfs64 tmp;
184 int error;
185
186 if (sz != sizeof(*buf))
187 return -EINVAL;
188
189 error = -EBADF;
190 file = fget(fd);
191 if (!file)
192 goto out;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800193 error = vfs_statfs64(file->f_path.dentry, &tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
195 error = -EFAULT;
196 fput(file);
197out:
198 return error;
199}
200
NeilBrown4a301312006-01-08 01:02:39 -0800201int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
202 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Amerigo Wang939a9422009-08-20 19:29:03 -0700204 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 struct iattr newattrs;
206
207 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
208 if (length < 0)
209 return -EINVAL;
210
211 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -0800212 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -0800213 if (filp) {
214 newattrs.ia_file = filp;
215 newattrs.ia_valid |= ATTR_FILE;
216 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Linus Torvalds7b82dc02007-05-08 20:10:00 -0700218 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -0700219 ret = should_remove_suid(dentry);
220 if (ret)
221 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -0700222
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800223 mutex_lock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -0700224 ret = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800225 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -0700226 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227}
228
Al Viro2d8f3032008-07-22 09:59:21 -0400229static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
Al Viro2d8f3032008-07-22 09:59:21 -0400231 struct path path;
232 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 int error;
234
235 error = -EINVAL;
236 if (length < 0) /* sorry, but loff_t says... */
237 goto out;
238
Al Viro2d8f3032008-07-22 09:59:21 -0400239 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 if (error)
241 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400242 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
244 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
245 error = -EISDIR;
246 if (S_ISDIR(inode->i_mode))
247 goto dput_and_out;
248
249 error = -EINVAL;
250 if (!S_ISREG(inode->i_mode))
251 goto dput_and_out;
252
Al Viro2d8f3032008-07-22 09:59:21 -0400253 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 if (error)
255 goto dput_and_out;
256
Al Viro256984a2008-07-22 08:09:30 -0400257 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800258 if (error)
259 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
261 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +0200262 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -0800263 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 error = get_write_access(inode);
266 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800267 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
david m. richter97003822007-07-31 00:39:12 -0700269 /*
270 * Make sure that there are no leases. get_write_access() protects
271 * against the truncate racing with a lease-granting setlease().
272 */
Al Viro8737c932009-12-24 06:47:55 -0500273 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700274 if (error)
275 goto put_write_and_out;
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900278 if (!error)
279 error = security_path_truncate(&path, length, 0);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500280 if (!error)
Al Viro2d8f3032008-07-22 09:59:21 -0400281 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282
david m. richter97003822007-07-31 00:39:12 -0700283put_write_and_out:
284 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800285mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400286 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400288 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289out:
290 return error;
291}
292
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200293SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200295 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296}
297
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800298static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299{
300 struct inode * inode;
301 struct dentry *dentry;
302 struct file * file;
303 int error;
304
305 error = -EINVAL;
306 if (length < 0)
307 goto out;
308 error = -EBADF;
309 file = fget(fd);
310 if (!file)
311 goto out;
312
313 /* explicitly opened as large or we are on 64-bit box */
314 if (file->f_flags & O_LARGEFILE)
315 small = 0;
316
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800317 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 inode = dentry->d_inode;
319 error = -EINVAL;
320 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
321 goto out_putf;
322
323 error = -EINVAL;
324 /* Cannot ftruncate over 2^31 bytes without large file support */
325 if (small && length > MAX_NON_LFS)
326 goto out_putf;
327
328 error = -EPERM;
329 if (IS_APPEND(inode))
330 goto out_putf;
331
332 error = locks_verify_truncate(inode, file, length);
333 if (!error)
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900334 error = security_path_truncate(&file->f_path, length,
335 ATTR_MTIME|ATTR_CTIME);
336 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700337 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338out_putf:
339 fput(file);
340out:
341 return error;
342}
343
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100344SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700346 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700347 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700348 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700349 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350}
351
352/* LFS versions of truncate are only needed on 32 bit machines */
353#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100354SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355{
356 return do_sys_truncate(path, length);
357}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100358#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
359asmlinkage long SyS_truncate64(long path, loff_t length)
360{
361 return SYSC_truncate64((const char __user *) path, length);
362}
363SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
364#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100366SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700368 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700369 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700370 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700371 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100373#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
374asmlinkage long SyS_ftruncate64(long fd, loff_t length)
375{
376 return SYSC_ftruncate64((unsigned int) fd, length);
377}
378SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100380#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400382
383int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400384{
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400385 struct inode *inode = file->f_path.dentry->d_inode;
386 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400387
388 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400389 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400390
391 /* Return error if mode is not supported */
Amit Arora97ac7352007-07-17 21:42:44 -0400392 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400393 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400394
Amit Arora97ac7352007-07-17 21:42:44 -0400395 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400396 return -EBADF;
Amit Arora97ac7352007-07-17 21:42:44 -0400397 /*
398 * Revalidate the write permissions, in case security policy has
399 * changed since the files were opened.
400 */
401 ret = security_file_permission(file, MAY_WRITE);
402 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400403 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400404
Amit Arora97ac7352007-07-17 21:42:44 -0400405 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400406 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400407
Amit Arora97ac7352007-07-17 21:42:44 -0400408 /*
409 * Let individual file system decide if it supports preallocation
410 * for directories or not.
411 */
412 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400413 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400414
Amit Arora97ac7352007-07-17 21:42:44 -0400415 /* Check for wrap through zero too */
416 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400417 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400418
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400419 if (!inode->i_op->fallocate)
420 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400421
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400422 return inode->i_op->fallocate(inode, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400423}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400424
425SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
426{
427 struct file *file;
428 int error = -EBADF;
429
430 file = fget(fd);
431 if (file) {
432 error = do_fallocate(file, mode, offset, len);
433 fput(file);
434 }
435
436 return error;
437}
438
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100439#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
440asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
441{
442 return SYSC_fallocate((int)fd, (int)mode, offset, len);
443}
444SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
445#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447/*
448 * access() needs to use the real uid/gid, not the effective uid/gid.
449 * We do this by temporarily clearing all FS-related capabilities and
450 * switching the fsuid/fsgid around to the real ones.
451 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100452SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
David Howellsd84f4f92008-11-14 10:39:23 +1100454 const struct cred *old_cred;
455 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400456 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400457 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 int res;
459
460 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
461 return -EINVAL;
462
David Howellsd84f4f92008-11-14 10:39:23 +1100463 override_cred = prepare_creds();
464 if (!override_cred)
465 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
David Howellsd84f4f92008-11-14 10:39:23 +1100467 override_cred->fsuid = override_cred->uid;
468 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700470 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100471 /* Clear the capabilities if we switch to a non-root user */
David Howellsd84f4f92008-11-14 10:39:23 +1100472 if (override_cred->uid)
473 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700474 else
David Howellsd84f4f92008-11-14 10:39:23 +1100475 override_cred->cap_effective =
476 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700477 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
David Howellsd84f4f92008-11-14 10:39:23 +1100479 old_cred = override_creds(override_cred);
480
Al Viro2d8f3032008-07-22 09:59:21 -0400481 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700482 if (res)
483 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
Al Viro2d8f3032008-07-22 09:59:21 -0400485 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400486
487 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400488 /*
489 * MAY_EXEC on regular files is denied if the fs is mounted
490 * with the "noexec" flag.
491 */
492 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400493 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400494 goto out_path_release;
495 }
496
Al Viro256984a2008-07-22 08:09:30 -0400497 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700498 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400499 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700500 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800501 /*
502 * This is a rare case where using __mnt_is_readonly()
503 * is OK without a mnt_want/drop_write() pair. Since
504 * no actual write to the fs is performed here, we do
505 * not need to telegraph to that to anyone.
506 *
507 * By doing this, we accept that this access is
508 * inherently racy and know that the fs may change
509 * state before we even see this result.
510 */
Al Viro2d8f3032008-07-22 09:59:21 -0400511 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700512 res = -EROFS;
513
514out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400515 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700516out:
David Howellsd84f4f92008-11-14 10:39:23 +1100517 revert_creds(old_cred);
518 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 return res;
520}
521
Heiko Carstensca013e92009-01-14 14:14:19 +0100522SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800523{
524 return sys_faccessat(AT_FDCWD, filename, mode);
525}
526
Heiko Carstens3cdad422009-01-14 14:14:22 +0100527SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528{
Al Viro2d8f3032008-07-22 09:59:21 -0400529 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 int error;
531
Al Viro2d8f3032008-07-22 09:59:21 -0400532 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 if (error)
534 goto out;
535
Al Viro2d8f3032008-07-22 09:59:21 -0400536 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 if (error)
538 goto dput_and_out;
539
Al Viro2d8f3032008-07-22 09:59:21 -0400540 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
542dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400543 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544out:
545 return error;
546}
547
Heiko Carstens3cdad422009-01-14 14:14:22 +0100548SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549{
550 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 int error;
553
554 error = -EBADF;
555 file = fget(fd);
556 if (!file)
557 goto out;
558
Jan Blunckac748a02008-02-14 19:34:39 -0800559 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560
561 error = -ENOTDIR;
562 if (!S_ISDIR(inode->i_mode))
563 goto out_putf;
564
Al Viro256984a2008-07-22 08:09:30 -0400565 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800567 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568out_putf:
569 fput(file);
570out:
571 return error;
572}
573
Heiko Carstens3480b252009-01-14 14:14:16 +0100574SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575{
Al Viro2d8f3032008-07-22 09:59:21 -0400576 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 int error;
578
Al Viro2d8f3032008-07-22 09:59:21 -0400579 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 if (error)
581 goto out;
582
Al Viro2d8f3032008-07-22 09:59:21 -0400583 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 if (error)
585 goto dput_and_out;
586
587 error = -EPERM;
588 if (!capable(CAP_SYS_CHROOT))
589 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900590 error = security_path_chroot(&path);
591 if (error)
592 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
Al Viro2d8f3032008-07-22 09:59:21 -0400594 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 error = 0;
596dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400597 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598out:
599 return error;
600}
601
Heiko Carstensa26eab22009-01-14 14:14:17 +0100602SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 struct inode * inode;
605 struct dentry * dentry;
606 struct file * file;
607 int err = -EBADF;
608 struct iattr newattrs;
609
610 file = fget(fd);
611 if (!file)
612 goto out;
613
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800614 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 inode = dentry->d_inode;
616
Al Viro5a190ae2007-06-07 12:19:32 -0400617 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000618
npiggin@suse.de96029c42009-04-26 20:25:55 +1000619 err = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800620 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 goto out_putf;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900622 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900623 err = security_path_chmod(dentry, file->f_vfsmnt, mode);
624 if (err)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900625 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 if (mode == (mode_t) -1)
627 mode = inode->i_mode;
628 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
629 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
630 err = notify_change(dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900631out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800632 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800633 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634out_putf:
635 fput(file);
636out:
637 return err;
638}
639
Heiko Carstens6559eed82009-01-14 14:14:32 +0100640SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641{
Al Viro2d8f3032008-07-22 09:59:21 -0400642 struct path path;
643 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 int error;
645 struct iattr newattrs;
646
Al Viro2d8f3032008-07-22 09:59:21 -0400647 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 if (error)
649 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400650 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
Al Viro2d8f3032008-07-22 09:59:21 -0400652 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800653 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 goto dput_and_out;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900655 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900656 error = security_path_chmod(path.dentry, path.mnt, mode);
657 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900658 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 if (mode == (mode_t) -1)
660 mode = inode->i_mode;
661 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
662 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400663 error = notify_change(path.dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900664out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800665 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400666 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400668 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669out:
670 return error;
671}
672
Heiko Carstensa26eab22009-01-14 14:14:17 +0100673SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800674{
675 return sys_fchmodat(AT_FDCWD, filename, mode);
676}
677
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900678static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900680 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 int error;
682 struct iattr newattrs;
683
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 newattrs.ia_valid = ATTR_CTIME;
685 if (user != (uid_t) -1) {
686 newattrs.ia_valid |= ATTR_UID;
687 newattrs.ia_uid = user;
688 }
689 if (group != (gid_t) -1) {
690 newattrs.ia_valid |= ATTR_GID;
691 newattrs.ia_gid = group;
692 }
693 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700694 newattrs.ia_valid |=
695 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800696 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900697 error = security_path_chown(path, user, group);
698 if (!error)
699 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800700 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 return error;
703}
704
Heiko Carstensca013e92009-01-14 14:14:19 +0100705SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706{
Al Viro2d8f3032008-07-22 09:59:21 -0400707 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 int error;
709
Al Viro2d8f3032008-07-22 09:59:21 -0400710 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700711 if (error)
712 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400713 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800714 if (error)
715 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900716 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400717 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800718out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400719 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700720out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 return error;
722}
723
Heiko Carstens6559eed82009-01-14 14:14:32 +0100724SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
725 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800726{
Al Viro2d8f3032008-07-22 09:59:21 -0400727 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800728 int error = -EINVAL;
729 int follow;
730
731 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
732 goto out;
733
734 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
Al Viro2d8f3032008-07-22 09:59:21 -0400735 error = user_path_at(dfd, filename, follow, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700736 if (error)
737 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400738 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800739 if (error)
740 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900741 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400742 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800743out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400744 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800745out:
746 return error;
747}
748
Heiko Carstensca013e92009-01-14 14:14:19 +0100749SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750{
Al Viro2d8f3032008-07-22 09:59:21 -0400751 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 int error;
753
Al Viro2d8f3032008-07-22 09:59:21 -0400754 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700755 if (error)
756 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400757 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800758 if (error)
759 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900760 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400761 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800762out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400763 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700764out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 return error;
766}
767
Heiko Carstensca013e92009-01-14 14:14:19 +0100768SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769{
770 struct file * file;
771 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700772 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700775 if (!file)
776 goto out;
777
npiggin@suse.de96029c42009-04-26 20:25:55 +1000778 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800779 if (error)
780 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800781 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400782 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900783 error = chown_common(&file->f_path, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800784 mnt_drop_write(file->f_path.mnt);
785out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700786 fput(file);
787out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 return error;
789}
790
Dave Hansen4a3fd212008-02-15 14:37:48 -0800791/*
792 * You have to be very careful that these write
793 * counts get cleaned up in error cases and
794 * upon __fput(). This should probably never
795 * be called outside of __dentry_open().
796 */
797static inline int __get_file_write_access(struct inode *inode,
798 struct vfsmount *mnt)
799{
800 int error;
801 error = get_write_access(inode);
802 if (error)
803 return error;
804 /*
805 * Do not take mount writer counts on
806 * special files since no writes to
807 * the mount itself will occur.
808 */
809 if (!special_file(inode->i_mode)) {
810 /*
811 * Balanced in __fput()
812 */
813 error = mnt_want_write(mnt);
814 if (error)
815 put_write_access(inode);
816 }
817 return error;
818}
819
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700820static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Al Viro482928d2009-12-19 10:10:39 -0500821 struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100822 int (*open)(struct inode *, struct file *),
823 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 struct inode *inode;
826 int error;
827
Al Viro53009902009-12-19 10:15:07 -0500828 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700829 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 inode = dentry->d_inode;
831 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800832 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 if (error)
834 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800835 if (!special_file(inode->i_mode))
836 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 }
838
839 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800840 f->f_path.dentry = dentry;
841 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 f->f_pos = 0;
843 f->f_op = fops_get(inode->i_fop);
844 file_move(f, &inode->i_sb->s_files);
845
David Howells745ca242008-11-14 10:39:22 +1100846 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900847 if (error)
848 goto cleanup_all;
849
Trond Myklebust834f2a42005-10-18 14:20:16 -0700850 if (!open && f->f_op)
851 open = f->f_op->open;
852 if (open) {
853 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 if (error)
855 goto cleanup_all;
856 }
Al Virob65a9cf2009-12-16 06:27:40 -0500857 ima_counts_get(f);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700858
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
860
861 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
862
863 /* NB: we're sure to have correct a_ops only after f_op->open */
864 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700865 if (!f->f_mapping->a_ops ||
866 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700867 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 fput(f);
869 f = ERR_PTR(-EINVAL);
870 }
871 }
872
873 return f;
874
875cleanup_all:
876 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800877 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800879 if (!special_file(inode->i_mode)) {
880 /*
881 * We don't consider this a real
882 * mnt_want/drop_write() pair
883 * because it all happenend right
884 * here, so just reset the state.
885 */
886 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800887 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800888 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800889 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 file_kill(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800891 f->f_path.dentry = NULL;
892 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893cleanup_file:
894 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 dput(dentry);
896 mntput(mnt);
897 return ERR_PTR(error);
898}
899
Trond Myklebust834f2a42005-10-18 14:20:16 -0700900/**
901 * lookup_instantiate_filp - instantiates the open intent filp
902 * @nd: pointer to nameidata
903 * @dentry: pointer to dentry
904 * @open: open callback
905 *
906 * Helper for filesystems that want to use lookup open intents and pass back
907 * a fully instantiated struct file to the caller.
908 * This function is meant to be called from within a filesystem's
909 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800910 * Beware of calling it for non-regular files! Those ->open methods might block
911 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
912 * leading to a deadlock, as nobody can open that fifo anymore, because
913 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700914 * Note that in case of error, nd->intent.open.file is destroyed, but the
915 * path information remains valid.
916 * If the open callback is set to NULL, then the standard f_op->open()
917 * filesystem callback is substituted.
918 */
919struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
920 int (*open)(struct inode *, struct file *))
921{
David Howells745ca242008-11-14 10:39:22 +1100922 const struct cred *cred = current_cred();
923
Trond Myklebust834f2a42005-10-18 14:20:16 -0700924 if (IS_ERR(nd->intent.open.file))
925 goto out;
926 if (IS_ERR(dentry))
927 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800928 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700929 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100930 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700931out:
932 return nd->intent.open.file;
933out_err:
934 release_open_intent(nd);
935 nd->intent.open.file = (struct file *)dentry;
936 goto out;
937}
938EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
939
940/**
941 * nameidata_to_filp - convert a nameidata to an open filp.
942 * @nd: pointer to nameidata
943 * @flags: open flags
944 *
945 * Note that this function destroys the original nameidata
946 */
Al Viro482928d2009-12-19 10:10:39 -0500947struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700948{
David Howells745ca242008-11-14 10:39:22 +1100949 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700950 struct file *filp;
951
952 /* Pick up the filp from the open intent */
953 filp = nd->intent.open.file;
954 /* Has the filesystem initialised the file for us? */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800955 if (filp->f_path.dentry == NULL)
Al Viro482928d2009-12-19 10:10:39 -0500956 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
David Howells745ca242008-11-14 10:39:22 +1100957 NULL, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700958 else
Jan Blunck1d957f92008-02-14 19:34:35 -0800959 path_put(&nd->path);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700960 return filp;
961}
962
Peter Staubach6fdcc212005-11-07 00:59:42 -0800963/*
964 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
965 * error.
966 */
David Howells745ca242008-11-14 10:39:22 +1100967struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
968 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700969{
970 int error;
971 struct file *f;
972
David Howellse0e81732009-09-02 09:13:40 +0100973 validate_creds(cred);
974
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800975 /*
976 * We must always pass in a valid mount pointer. Historically
977 * callers got away with not passing it, but we must enforce this at
978 * the earliest possible point now to avoid strange problems deep in the
979 * filesystem stack.
980 */
981 if (!mnt) {
982 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
983 dump_stack();
984 return ERR_PTR(-EINVAL);
985 }
986
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700987 error = -ENFILE;
988 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800989 if (f == NULL) {
990 dput(dentry);
991 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700992 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800993 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700994
Al Viro482928d2009-12-19 10:10:39 -0500995 f->f_flags = flags;
996 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700997}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998EXPORT_SYMBOL(dentry_open);
999
Matt Mackallb01ec0e2006-01-08 01:05:20 -08001000static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001002 struct fdtable *fdt = files_fdtable(files);
1003 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -08001004 if (fd < files->next_fd)
1005 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006}
1007
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001008void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009{
1010 struct files_struct *files = current->files;
1011 spin_lock(&files->file_lock);
1012 __put_unused_fd(files, fd);
1013 spin_unlock(&files->file_lock);
1014}
1015
1016EXPORT_SYMBOL(put_unused_fd);
1017
1018/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001019 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 *
1021 * The VFS is full of places where we drop the files lock between
1022 * setting the open_fds bitmap and installing the file in the file
1023 * array. At any such point, we are vulnerable to a dup2() race
1024 * installing a file in the array before us. We need to detect this and
1025 * fput() the struct file we are about to overwrite in this case.
1026 *
1027 * It should never happen - if we allow dup2() do it, _really_ bad things
1028 * will follow.
1029 */
1030
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -08001031void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032{
1033 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001034 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001036 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001037 BUG_ON(fdt->fd[fd] != NULL);
1038 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 spin_unlock(&files->file_lock);
1040}
1041
1042EXPORT_SYMBOL(fd_install);
1043
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001044long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045{
Miklos Szeredie922efc2005-09-06 15:18:25 -07001046 char *tmp = getname(filename);
1047 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001050 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 if (fd >= 0) {
Al Viro6e8341a2009-04-06 11:16:22 -04001052 struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001053 if (IS_ERR(f)) {
1054 put_unused_fd(fd);
1055 fd = PTR_ERR(f);
1056 } else {
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001057 fsnotify_open(f->f_path.dentry);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001058 fd_install(fd, f);
1059 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 putname(tmp);
1062 }
1063 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001065
Heiko Carstensca013e92009-01-14 14:14:19 +01001066SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001067{
Linus Torvalds385910f2006-04-18 13:22:59 -07001068 long ret;
1069
Miklos Szeredie922efc2005-09-06 15:18:25 -07001070 if (force_o_largefile())
1071 flags |= O_LARGEFILE;
1072
Linus Torvalds385910f2006-04-18 13:22:59 -07001073 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1074 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001075 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001076 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001077}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
Heiko Carstens6559eed82009-01-14 14:14:32 +01001079SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1080 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001081{
Linus Torvalds385910f2006-04-18 13:22:59 -07001082 long ret;
1083
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001084 if (force_o_largefile())
1085 flags |= O_LARGEFILE;
1086
Linus Torvalds385910f2006-04-18 13:22:59 -07001087 ret = do_sys_open(dfd, filename, flags, mode);
1088 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001089 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001090 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001091}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001092
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093#ifndef __alpha__
1094
1095/*
1096 * For backward compatibility? Maybe this should be moved
1097 * into arch/i386 instead?
1098 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001099SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100{
1101 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1102}
1103
1104#endif
1105
1106/*
1107 * "id" is the POSIX thread ID. We use the
1108 * files pointer for this..
1109 */
1110int filp_close(struct file *filp, fl_owner_t id)
1111{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001112 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
1114 if (!file_count(filp)) {
1115 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001116 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 }
1118
Christoph Lameter45778ca2005-06-23 00:10:17 -07001119 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001120 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
1122 dnotify_flush(filp, id);
1123 locks_remove_posix(filp, id);
1124 fput(filp);
1125 return retval;
1126}
1127
1128EXPORT_SYMBOL(filp_close);
1129
1130/*
1131 * Careful here! We test whether the file pointer is NULL before
1132 * releasing the fd. This ensures that one clone task can't release
1133 * an fd while another clone is opening it.
1134 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001135SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136{
1137 struct file * filp;
1138 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001139 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001140 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141
1142 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001143 fdt = files_fdtable(files);
1144 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001146 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 if (!filp)
1148 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001149 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001150 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 __put_unused_fd(files, fd);
1152 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001153 retval = filp_close(filp, files);
1154
1155 /* can't restart close syscall because file table entry was cleared */
1156 if (unlikely(retval == -ERESTARTSYS ||
1157 retval == -ERESTARTNOINTR ||
1158 retval == -ERESTARTNOHAND ||
1159 retval == -ERESTART_RESTARTBLOCK))
1160 retval = -EINTR;
1161
1162 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164out_unlock:
1165 spin_unlock(&files->file_lock);
1166 return -EBADF;
1167}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168EXPORT_SYMBOL(sys_close);
1169
1170/*
1171 * This routine simulates a hangup on the tty, to arrange that users
1172 * are given clean terminals at login time.
1173 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001174SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175{
1176 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001177 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 return 0;
1179 }
1180 return -EPERM;
1181}
1182
1183/*
1184 * Called when an inode is about to be open.
1185 * We use this to disallow opening large files on 32bit systems if
1186 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1187 * on this flag in sys_open.
1188 */
1189int generic_file_open(struct inode * inode, struct file * filp)
1190{
1191 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001192 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 return 0;
1194}
1195
1196EXPORT_SYMBOL(generic_file_open);
1197
1198/*
1199 * This is used by subsystems that don't want seekable
1200 * file descriptors
1201 */
1202int nonseekable_open(struct inode *inode, struct file *filp)
1203{
1204 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1205 return 0;
1206}
1207
1208EXPORT_SYMBOL(nonseekable_open);