blob: 1540632d8387fe98a51d0193201346acb18ae70e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/string.h>
8#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040010#include <linux/fdtable.h>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/tty.h>
14#include <linux/namei.h>
15#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080016#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070017#include <linux/securebits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080020#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090021#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040030#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050031#include <linux/ima.h>
Eric Paris2dfc1ca2009-12-17 20:30:52 -050032#include <linux/dnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Eric Parise81e3f42009-12-04 15:47:36 -050034#include "internal.h"
35
NeilBrown4a301312006-01-08 01:02:39 -080036int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
37 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070038{
Amerigo Wang939a9422009-08-20 19:29:03 -070039 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 struct iattr newattrs;
41
42 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
43 if (length < 0)
44 return -EINVAL;
45
46 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080047 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080048 if (filp) {
49 newattrs.ia_file = filp;
50 newattrs.ia_valid |= ATTR_FILE;
51 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Linus Torvalds7b82dc02007-05-08 20:10:00 -070053 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -070054 ret = should_remove_suid(dentry);
55 if (ret)
56 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070057
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080058 mutex_lock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070059 ret = notify_change(dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080060 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070061 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062}
63
Al Viro2d8f3032008-07-22 09:59:21 -040064static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
Al Viro2d8f3032008-07-22 09:59:21 -040066 struct path path;
67 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 int error;
69
70 error = -EINVAL;
71 if (length < 0) /* sorry, but loff_t says... */
72 goto out;
73
Al Viro2d8f3032008-07-22 09:59:21 -040074 error = user_path(pathname, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (error)
76 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -040077 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
80 error = -EISDIR;
81 if (S_ISDIR(inode->i_mode))
82 goto dput_and_out;
83
84 error = -EINVAL;
85 if (!S_ISREG(inode->i_mode))
86 goto dput_and_out;
87
Al Viro2d8f3032008-07-22 09:59:21 -040088 error = mnt_want_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 if (error)
90 goto dput_and_out;
91
Al Viro256984a2008-07-22 08:09:30 -040092 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080093 if (error)
94 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020097 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080098 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 error = get_write_access(inode);
101 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800102 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
david m. richter97003822007-07-31 00:39:12 -0700104 /*
105 * Make sure that there are no leases. get_write_access() protects
106 * against the truncate racing with a lease-granting setlease().
107 */
Al Viro8737c932009-12-24 06:47:55 -0500108 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700109 if (error)
110 goto put_write_and_out;
111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900113 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900114 error = security_path_truncate(&path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500115 if (!error)
Al Viro2d8f3032008-07-22 09:59:21 -0400116 error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
david m. richter97003822007-07-31 00:39:12 -0700118put_write_and_out:
119 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800120mnt_drop_write_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400121 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400123 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124out:
125 return error;
126}
127
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200128SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200130 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131}
132
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800133static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134{
135 struct inode * inode;
136 struct dentry *dentry;
137 struct file * file;
138 int error;
139
140 error = -EINVAL;
141 if (length < 0)
142 goto out;
143 error = -EBADF;
144 file = fget(fd);
145 if (!file)
146 goto out;
147
148 /* explicitly opened as large or we are on 64-bit box */
149 if (file->f_flags & O_LARGEFILE)
150 small = 0;
151
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800152 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 inode = dentry->d_inode;
154 error = -EINVAL;
155 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
156 goto out_putf;
157
158 error = -EINVAL;
159 /* Cannot ftruncate over 2^31 bytes without large file support */
160 if (small && length > MAX_NON_LFS)
161 goto out_putf;
162
163 error = -EPERM;
164 if (IS_APPEND(inode))
165 goto out_putf;
166
167 error = locks_verify_truncate(inode, file, length);
168 if (!error)
Tetsuo Handaea0d3ab2010-06-02 13:24:43 +0900169 error = security_path_truncate(&file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900170 if (!error)
Peter Staubach6e656be2006-06-25 05:48:36 -0700171 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172out_putf:
173 fput(file);
174out:
175 return error;
176}
177
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100178SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700180 long ret = do_sys_ftruncate(fd, length, 1);
Linus Torvalds385910f2006-04-18 13:22:59 -0700181 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700182 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700183 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184}
185
186/* LFS versions of truncate are only needed on 32 bit machines */
187#if BITS_PER_LONG == 32
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100188SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
190 return do_sys_truncate(path, length);
191}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100192#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
193asmlinkage long SyS_truncate64(long path, loff_t length)
194{
195 return SYSC_truncate64((const char __user *) path, length);
196}
197SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
198#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100200SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700202 long ret = do_sys_ftruncate(fd, length, 0);
Linus Torvalds385910f2006-04-18 13:22:59 -0700203 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700204 asmlinkage_protect(2, ret, fd, length);
Linus Torvalds0a489cb2006-04-18 13:02:48 -0700205 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100207#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
208asmlinkage long SyS_ftruncate64(long fd, loff_t length)
209{
210 return SYSC_ftruncate64((unsigned int) fd, length);
211}
212SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213#endif
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100214#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400216
217int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400218{
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400219 struct inode *inode = file->f_path.dentry->d_inode;
220 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400221
222 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400223 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400224
225 /* Return error if mode is not supported */
Josef Bacik79124f182010-11-17 20:46:15 -0500226 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
227 return -EOPNOTSUPP;
228
229 /* Punch hole must have keep size set */
230 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
231 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400232 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400233
Amit Arora97ac7352007-07-17 21:42:44 -0400234 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400235 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100236
237 /* It's not possible punch hole on append only file */
238 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
239 return -EPERM;
240
241 if (IS_IMMUTABLE(inode))
242 return -EPERM;
243
Amit Arora97ac7352007-07-17 21:42:44 -0400244 /*
245 * Revalidate the write permissions, in case security policy has
246 * changed since the files were opened.
247 */
248 ret = security_file_permission(file, MAY_WRITE);
249 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400250 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400251
Amit Arora97ac7352007-07-17 21:42:44 -0400252 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400253 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400254
Amit Arora97ac7352007-07-17 21:42:44 -0400255 /*
256 * Let individual file system decide if it supports preallocation
257 * for directories or not.
258 */
259 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400260 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400261
Amit Arora97ac7352007-07-17 21:42:44 -0400262 /* Check for wrap through zero too */
263 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400264 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400265
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100266 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400267 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400268
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100269 return file->f_op->fallocate(file, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400270}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400271
272SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
273{
274 struct file *file;
275 int error = -EBADF;
276
277 file = fget(fd);
278 if (file) {
279 error = do_fallocate(file, mode, offset, len);
280 fput(file);
281 }
282
283 return error;
284}
285
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100286#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
287asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
288{
289 return SYSC_fallocate((int)fd, (int)mode, offset, len);
290}
291SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
292#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294/*
295 * access() needs to use the real uid/gid, not the effective uid/gid.
296 * We do this by temporarily clearing all FS-related capabilities and
297 * switching the fsuid/fsgid around to the real ones.
298 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100299SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300{
David Howellsd84f4f92008-11-14 10:39:23 +1100301 const struct cred *old_cred;
302 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400303 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400304 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 int res;
306
307 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
308 return -EINVAL;
309
David Howellsd84f4f92008-11-14 10:39:23 +1100310 override_cred = prepare_creds();
311 if (!override_cred)
312 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
David Howellsd84f4f92008-11-14 10:39:23 +1100314 override_cred->fsuid = override_cred->uid;
315 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700317 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100318 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800319 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
320 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100321 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700322 else
David Howellsd84f4f92008-11-14 10:39:23 +1100323 override_cred->cap_effective =
324 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700325 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326
David Howellsd84f4f92008-11-14 10:39:23 +1100327 old_cred = override_creds(override_cred);
328
Al Viro2d8f3032008-07-22 09:59:21 -0400329 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700330 if (res)
331 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Al Viro2d8f3032008-07-22 09:59:21 -0400333 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400334
335 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400336 /*
337 * MAY_EXEC on regular files is denied if the fs is mounted
338 * with the "noexec" flag.
339 */
340 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400341 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400342 goto out_path_release;
343 }
344
Al Viro256984a2008-07-22 08:09:30 -0400345 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700346 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400347 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700348 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800349 /*
350 * This is a rare case where using __mnt_is_readonly()
351 * is OK without a mnt_want/drop_write() pair. Since
352 * no actual write to the fs is performed here, we do
353 * not need to telegraph to that to anyone.
354 *
355 * By doing this, we accept that this access is
356 * inherently racy and know that the fs may change
357 * state before we even see this result.
358 */
Al Viro2d8f3032008-07-22 09:59:21 -0400359 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700360 res = -EROFS;
361
362out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400363 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700364out:
David Howellsd84f4f92008-11-14 10:39:23 +1100365 revert_creds(old_cred);
366 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 return res;
368}
369
Heiko Carstensca013e92009-01-14 14:14:19 +0100370SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800371{
372 return sys_faccessat(AT_FDCWD, filename, mode);
373}
374
Heiko Carstens3cdad422009-01-14 14:14:22 +0100375SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376{
Al Viro2d8f3032008-07-22 09:59:21 -0400377 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 int error;
379
Al Viro2d8f3032008-07-22 09:59:21 -0400380 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 if (error)
382 goto out;
383
Eric Paris9cfcac82010-07-23 11:43:51 -0400384 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 if (error)
386 goto dput_and_out;
387
Al Viro2d8f3032008-07-22 09:59:21 -0400388 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
390dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400391 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392out:
393 return error;
394}
395
Heiko Carstens3cdad422009-01-14 14:14:22 +0100396SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397{
398 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 struct inode *inode;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700400 int error, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
402 error = -EBADF;
Linus Torvalds332a2e12012-07-07 10:17:00 -0700403 file = fget_raw_light(fd, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 if (!file)
405 goto out;
406
Jan Blunckac748a02008-02-14 19:34:39 -0800407 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
409 error = -ENOTDIR;
410 if (!S_ISDIR(inode->i_mode))
411 goto out_putf;
412
Eric Paris9cfcac82010-07-23 11:43:51 -0400413 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800415 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416out_putf:
Linus Torvalds332a2e12012-07-07 10:17:00 -0700417 fput_light(file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418out:
419 return error;
420}
421
Heiko Carstens3480b252009-01-14 14:14:16 +0100422SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423{
Al Viro2d8f3032008-07-22 09:59:21 -0400424 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 int error;
426
Al Viro2d8f3032008-07-22 09:59:21 -0400427 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 if (error)
429 goto out;
430
Eric Paris9cfcac82010-07-23 11:43:51 -0400431 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (error)
433 goto dput_and_out;
434
435 error = -EPERM;
436 if (!capable(CAP_SYS_CHROOT))
437 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900438 error = security_path_chroot(&path);
439 if (error)
440 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Al Viro2d8f3032008-07-22 09:59:21 -0400442 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 error = 0;
444dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400445 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446out:
447 return error;
448}
449
Al Viroe57712e2011-07-26 04:15:54 -0400450static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
Al Viroe57712e2011-07-26 04:15:54 -0400452 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400454 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
Al Viroe57712e2011-07-26 04:15:54 -0400456 error = mnt_want_write(path->mnt);
457 if (error)
458 return error;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900459 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500460 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400461 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900462 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
464 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viroe57712e2011-07-26 04:15:54 -0400465 error = notify_change(path->dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900466out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800467 mutex_unlock(&inode->i_mutex);
Al Viroe57712e2011-07-26 04:15:54 -0400468 mnt_drop_write(path->mnt);
469 return error;
470}
471
Al Viro49f0a072011-07-26 04:22:01 -0400472SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400473{
474 struct file * file;
475 int err = -EBADF;
476
477 file = fget(fd);
478 if (file) {
479 audit_inode(NULL, file->f_path.dentry);
480 err = chmod_common(&file->f_path, mode);
481 fput(file);
482 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 return err;
484}
485
Al Viro49f0a072011-07-26 04:22:01 -0400486SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Al Viro2d8f3032008-07-22 09:59:21 -0400488 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Al Viro2d8f3032008-07-22 09:59:21 -0400491 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400492 if (!error) {
493 error = chmod_common(&path, mode);
494 path_put(&path);
495 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 return error;
497}
498
Al Viro49f0a072011-07-26 04:22:01 -0400499SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800500{
501 return sys_fchmodat(AT_FDCWD, filename, mode);
502}
503
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900504static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900506 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 int error;
508 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800509 kuid_t uid;
510 kgid_t gid;
511
512 uid = make_kuid(current_user_ns(), user);
513 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 newattrs.ia_valid = ATTR_CTIME;
516 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800517 if (!uid_valid(uid))
518 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800520 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 }
522 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800523 if (!gid_valid(gid))
524 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800526 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 }
528 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700529 newattrs.ia_valid |=
530 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800531 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900532 error = security_path_chown(path, user, group);
533 if (!error)
534 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800535 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 return error;
538}
539
Heiko Carstensca013e92009-01-14 14:14:19 +0100540SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541{
Al Viro2d8f3032008-07-22 09:59:21 -0400542 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 int error;
544
Al Viro2d8f3032008-07-22 09:59:21 -0400545 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700546 if (error)
547 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400548 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800549 if (error)
550 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900551 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400552 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800553out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400554 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700555out:
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;
572 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700573 if (error)
574 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400575 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800576 if (error)
577 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900578 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400579 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800580out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400581 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800582out:
583 return error;
584}
585
Heiko Carstensca013e92009-01-14 14:14:19 +0100586SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587{
Al Viro2d8f3032008-07-22 09:59:21 -0400588 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 int error;
590
Al Viro2d8f3032008-07-22 09:59:21 -0400591 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700592 if (error)
593 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400594 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800595 if (error)
596 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900597 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400598 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800599out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400600 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700601out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 return error;
603}
604
Heiko Carstensca013e92009-01-14 14:14:19 +0100605SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606{
607 struct file * file;
608 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700609 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610
611 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700612 if (!file)
613 goto out;
614
npiggin@suse.de96029c42009-04-26 20:25:55 +1000615 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800616 if (error)
617 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800618 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400619 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900620 error = chown_common(&file->f_path, user, group);
Al Viro2a79f172011-12-09 08:06:57 -0500621 mnt_drop_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800622out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700623 fput(file);
624out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 return error;
626}
627
Dave Hansen4a3fd212008-02-15 14:37:48 -0800628/*
629 * You have to be very careful that these write
630 * counts get cleaned up in error cases and
631 * upon __fput(). This should probably never
632 * be called outside of __dentry_open().
633 */
634static inline int __get_file_write_access(struct inode *inode,
635 struct vfsmount *mnt)
636{
637 int error;
638 error = get_write_access(inode);
639 if (error)
640 return error;
641 /*
642 * Do not take mount writer counts on
643 * special files since no writes to
644 * the mount itself will occur.
645 */
646 if (!special_file(inode->i_mode)) {
647 /*
648 * Balanced in __fput()
649 */
650 error = mnt_want_write(mnt);
651 if (error)
652 put_write_access(inode);
653 }
654 return error;
655}
656
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200657int open_check_o_direct(struct file *f)
658{
659 /* NB: we're sure to have correct a_ops only after f_op->open */
660 if (f->f_flags & O_DIRECT) {
661 if (!f->f_mapping->a_ops ||
662 ((!f->f_mapping->a_ops->direct_IO) &&
663 (!f->f_mapping->a_ops->get_xip_mem))) {
664 return -EINVAL;
665 }
666 }
667 return 0;
668}
669
670static struct file *do_dentry_open(struct dentry *dentry, struct vfsmount *mnt,
671 struct file *f,
672 int (*open)(struct inode *, struct file *),
673 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674{
Al Viro1abf0c72011-03-13 03:51:11 -0400675 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 struct inode *inode;
677 int error;
678
Al Viro53009902009-12-19 10:15:07 -0500679 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700680 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400681
682 if (unlikely(f->f_flags & O_PATH))
683 f->f_mode = FMODE_PATH;
684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 inode = dentry->d_inode;
686 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800687 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 if (error)
689 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800690 if (!special_file(inode->i_mode))
691 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 }
693
694 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800695 f->f_path.dentry = dentry;
696 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 f->f_pos = 0;
Nick Pigginee2ffa02010-08-18 04:37:35 +1000698 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Al Viro1abf0c72011-03-13 03:51:11 -0400700 if (unlikely(f->f_mode & FMODE_PATH)) {
701 f->f_op = &empty_fops;
702 return f;
703 }
704
705 f->f_op = fops_get(inode->i_fop);
706
Eric Paris83d49852012-04-04 13:45:40 -0400707 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900708 if (error)
709 goto cleanup_all;
710
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400711 error = break_lease(inode, f->f_flags);
712 if (error)
713 goto cleanup_all;
714
Trond Myklebust834f2a42005-10-18 14:20:16 -0700715 if (!open && f->f_op)
716 open = f->f_op->open;
717 if (open) {
718 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 if (error)
720 goto cleanup_all;
721 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400722 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
723 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700724
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
726
727 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 return f;
730
731cleanup_all:
732 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800733 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800735 if (!special_file(inode->i_mode)) {
736 /*
737 * We don't consider this a real
738 * mnt_want/drop_write() pair
739 * because it all happenend right
740 * here, so just reset the state.
741 */
742 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800743 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800744 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800745 }
Nick Pigginee2ffa02010-08-18 04:37:35 +1000746 file_sb_list_del(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800747 f->f_path.dentry = NULL;
748 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749cleanup_file:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 dput(dentry);
751 mntput(mnt);
752 return ERR_PTR(error);
753}
754
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200755static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
756 struct file *f,
757 int (*open)(struct inode *, struct file *),
758 const struct cred *cred)
759{
760 struct file *res = do_dentry_open(dentry, mnt, f, open, cred);
761 if (!IS_ERR(res)) {
762 int error = open_check_o_direct(f);
763 if (error) {
764 fput(res);
765 res = ERR_PTR(error);
766 }
Miklos Szeredi78f71ef2012-05-21 17:30:16 +0200767 } else {
768 put_filp(f);
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200769 }
770 return res;
771}
772
Trond Myklebust834f2a42005-10-18 14:20:16 -0700773/**
774 * lookup_instantiate_filp - instantiates the open intent filp
775 * @nd: pointer to nameidata
776 * @dentry: pointer to dentry
777 * @open: open callback
778 *
779 * Helper for filesystems that want to use lookup open intents and pass back
780 * a fully instantiated struct file to the caller.
781 * This function is meant to be called from within a filesystem's
782 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800783 * Beware of calling it for non-regular files! Those ->open methods might block
784 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
785 * leading to a deadlock, as nobody can open that fifo anymore, because
786 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700787 * Note that in case of error, nd->intent.open.file is destroyed, but the
788 * path information remains valid.
789 * If the open callback is set to NULL, then the standard f_op->open()
790 * filesystem callback is substituted.
791 */
792struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
793 int (*open)(struct inode *, struct file *))
794{
David Howells745ca242008-11-14 10:39:22 +1100795 const struct cred *cred = current_cred();
796
Trond Myklebust834f2a42005-10-18 14:20:16 -0700797 if (IS_ERR(nd->intent.open.file))
798 goto out;
799 if (IS_ERR(dentry))
800 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800801 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700802 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100803 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700804out:
805 return nd->intent.open.file;
806out_err:
807 release_open_intent(nd);
Konstantin Khlebnikov5a9a4362011-07-17 15:35:12 +0400808 nd->intent.open.file = ERR_CAST(dentry);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700809 goto out;
810}
811EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
812
813/**
814 * nameidata_to_filp - convert a nameidata to an open filp.
815 * @nd: pointer to nameidata
816 * @flags: open flags
817 *
818 * Note that this function destroys the original nameidata
819 */
Al Viro482928d2009-12-19 10:10:39 -0500820struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700821{
David Howells745ca242008-11-14 10:39:22 +1100822 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700823 struct file *filp;
824
825 /* Pick up the filp from the open intent */
826 filp = nd->intent.open.file;
Linus Torvalds2dab5972011-02-11 15:53:38 -0800827
Trond Myklebust834f2a42005-10-18 14:20:16 -0700828 /* Has the filesystem initialised the file for us? */
Miklos Szeredi50ee93a2012-05-21 17:30:18 +0200829 if (filp->f_path.dentry != NULL) {
830 nd->intent.open.file = NULL;
831 } else {
Miklos Szeredi91daee92012-05-21 17:30:17 +0200832 struct file *res;
833
Al Virod893f1b2010-10-29 03:30:42 -0400834 path_get(&nd->path);
Miklos Szeredi91daee92012-05-21 17:30:17 +0200835 res = do_dentry_open(nd->path.dentry, nd->path.mnt,
836 filp, NULL, cred);
837 if (!IS_ERR(res)) {
838 int error;
839
Miklos Szeredi50ee93a2012-05-21 17:30:18 +0200840 nd->intent.open.file = NULL;
Miklos Szeredi91daee92012-05-21 17:30:17 +0200841 BUG_ON(res != filp);
842
843 error = open_check_o_direct(filp);
844 if (error) {
845 fput(filp);
846 filp = ERR_PTR(error);
847 }
848 } else {
Miklos Szeredi50ee93a2012-05-21 17:30:18 +0200849 /* Allow nd->intent.open.file to be recycled */
Miklos Szeredi91daee92012-05-21 17:30:17 +0200850 filp = res;
851 }
Al Virod893f1b2010-10-29 03:30:42 -0400852 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700853 return filp;
854}
855
Peter Staubach6fdcc212005-11-07 00:59:42 -0800856/*
857 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
858 * error.
859 */
David Howells745ca242008-11-14 10:39:22 +1100860struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
861 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700862{
863 int error;
864 struct file *f;
865
David Howellse0e81732009-09-02 09:13:40 +0100866 validate_creds(cred);
867
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900868 /* We must always pass in a valid mount pointer. */
869 BUG_ON(!mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800870
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700871 error = -ENFILE;
872 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800873 if (f == NULL) {
874 dput(dentry);
875 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700876 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800877 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700878
Al Viro482928d2009-12-19 10:10:39 -0500879 f->f_flags = flags;
880 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700881}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882EXPORT_SYMBOL(dentry_open);
883
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800884static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700886 struct fdtable *fdt = files_fdtable(files);
David Howells1dce27c2012-02-16 17:49:42 +0000887 __clear_open_fd(fd, fdt);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800888 if (fd < files->next_fd)
889 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890}
891
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800892void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
894 struct files_struct *files = current->files;
895 spin_lock(&files->file_lock);
896 __put_unused_fd(files, fd);
897 spin_unlock(&files->file_lock);
898}
899
900EXPORT_SYMBOL(put_unused_fd);
901
902/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800903 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 *
905 * The VFS is full of places where we drop the files lock between
906 * setting the open_fds bitmap and installing the file in the file
907 * array. At any such point, we are vulnerable to a dup2() race
908 * installing a file in the array before us. We need to detect this and
909 * fput() the struct file we are about to overwrite in this case.
910 *
911 * It should never happen - if we allow dup2() do it, _really_ bad things
912 * will follow.
913 */
914
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800915void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916{
917 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700918 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700920 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700921 BUG_ON(fdt->fd[fd] != NULL);
922 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 spin_unlock(&files->file_lock);
924}
925
926EXPORT_SYMBOL(fd_install);
927
Al Viroa218d0f2011-11-21 14:59:34 -0500928static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500929{
930 int lookup_flags = 0;
931 int acc_mode;
932
933 if (!(flags & O_CREAT))
934 mode = 0;
935 op->mode = mode;
936
937 /* Must never be set by userspace */
938 flags &= ~FMODE_NONOTIFY;
939
940 /*
941 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
942 * check for O_DSYNC if the need any syncing at all we enforce it's
943 * always set instead of having to deal with possibly weird behaviour
944 * for malicious applications setting only __O_SYNC.
945 */
946 if (flags & __O_SYNC)
947 flags |= O_DSYNC;
948
Al Viro1abf0c72011-03-13 03:51:11 -0400949 /*
950 * If we have O_PATH in the open flag. Then we
951 * cannot have anything other than the below set of flags
952 */
953 if (flags & O_PATH) {
954 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
955 acc_mode = 0;
956 } else {
957 acc_mode = MAY_OPEN | ACC_MODE(flags);
958 }
Al Viro47c805d2011-02-23 17:44:09 -0500959
Al Viro1abf0c72011-03-13 03:51:11 -0400960 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500961
962 /* O_TRUNC implies we need access checks for write permissions */
963 if (flags & O_TRUNC)
964 acc_mode |= MAY_WRITE;
965
966 /* Allow the LSM permission hook to distinguish append
967 access from general write access. */
968 if (flags & O_APPEND)
969 acc_mode |= MAY_APPEND;
970
971 op->acc_mode = acc_mode;
972
Al Viro1abf0c72011-03-13 03:51:11 -0400973 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
974
Al Viro47c805d2011-02-23 17:44:09 -0500975 if (flags & O_CREAT) {
976 op->intent |= LOOKUP_CREATE;
977 if (flags & O_EXCL)
978 op->intent |= LOOKUP_EXCL;
979 }
980
981 if (flags & O_DIRECTORY)
982 lookup_flags |= LOOKUP_DIRECTORY;
983 if (!(flags & O_NOFOLLOW))
984 lookup_flags |= LOOKUP_FOLLOW;
985 return lookup_flags;
986}
987
988/**
989 * filp_open - open file and return file pointer
990 *
991 * @filename: path to open
992 * @flags: open flags as per the open(2) second argument
993 * @mode: mode for the new file if O_CREAT is set, else ignored
994 *
995 * This is the helper to open a file from kernelspace if you really
996 * have to. But in generally you should not do this, so please move
997 * along, nothing to see here..
998 */
Al Viroa218d0f2011-11-21 14:59:34 -0500999struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -05001000{
1001 struct open_flags op;
1002 int lookup = build_open_flags(flags, mode, &op);
1003 return do_filp_open(AT_FDCWD, filename, &op, lookup);
1004}
1005EXPORT_SYMBOL(filp_open);
1006
Al Viro73d049a2011-03-11 12:08:24 -05001007struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1008 const char *filename, int flags)
1009{
1010 struct open_flags op;
1011 int lookup = build_open_flags(flags, 0, &op);
1012 if (flags & O_CREAT)
1013 return ERR_PTR(-EINVAL);
1014 if (!filename && (flags & O_DIRECTORY))
1015 if (!dentry->d_inode->i_op->lookup)
1016 return ERR_PTR(-ENOTDIR);
1017 return do_file_open_root(dentry, mnt, filename, &op, lookup);
1018}
1019EXPORT_SYMBOL(file_open_root);
1020
Al Viroa218d0f2011-11-21 14:59:34 -05001021long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022{
Al Viro47c805d2011-02-23 17:44:09 -05001023 struct open_flags op;
1024 int lookup = build_open_flags(flags, mode, &op);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001025 char *tmp = getname(filename);
1026 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -07001029 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 if (fd >= 0) {
Al Viro47c805d2011-02-23 17:44:09 -05001031 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001032 if (IS_ERR(f)) {
1033 put_unused_fd(fd);
1034 fd = PTR_ERR(f);
1035 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -05001036 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -07001037 fd_install(fd, f);
1038 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 putname(tmp);
1041 }
1042 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001044
Al Viroa218d0f2011-11-21 14:59:34 -05001045SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001046{
Linus Torvalds385910f2006-04-18 13:22:59 -07001047 long ret;
1048
Miklos Szeredie922efc2005-09-06 15:18:25 -07001049 if (force_o_largefile())
1050 flags |= O_LARGEFILE;
1051
Linus Torvalds385910f2006-04-18 13:22:59 -07001052 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1053 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001054 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001055 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -07001056}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057
Heiko Carstens6559eed82009-01-14 14:14:32 +01001058SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001059 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001060{
Linus Torvalds385910f2006-04-18 13:22:59 -07001061 long ret;
1062
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001063 if (force_o_largefile())
1064 flags |= O_LARGEFILE;
1065
Linus Torvalds385910f2006-04-18 13:22:59 -07001066 ret = do_sys_open(dfd, filename, flags, mode);
1067 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -07001068 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -07001069 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001070}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001071
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072#ifndef __alpha__
1073
1074/*
1075 * For backward compatibility? Maybe this should be moved
1076 * into arch/i386 instead?
1077 */
Al Viroa218d0f2011-11-21 14:59:34 -05001078SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079{
1080 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1081}
1082
1083#endif
1084
1085/*
1086 * "id" is the POSIX thread ID. We use the
1087 * files pointer for this..
1088 */
1089int filp_close(struct file *filp, fl_owner_t id)
1090{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001091 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092
1093 if (!file_count(filp)) {
1094 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001095 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 }
1097
Christoph Lameter45778ca2005-06-23 00:10:17 -07001098 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001099 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100
Al Viro1abf0c72011-03-13 03:51:11 -04001101 if (likely(!(filp->f_mode & FMODE_PATH))) {
1102 dnotify_flush(filp, id);
1103 locks_remove_posix(filp, id);
1104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 fput(filp);
1106 return retval;
1107}
1108
1109EXPORT_SYMBOL(filp_close);
1110
1111/*
1112 * Careful here! We test whether the file pointer is NULL before
1113 * releasing the fd. This ensures that one clone task can't release
1114 * an fd while another clone is opening it.
1115 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001116SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117{
1118 struct file * filp;
1119 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001120 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -07001121 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122
1123 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001124 fdt = files_fdtable(files);
1125 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001127 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 if (!filp)
1129 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -07001130 rcu_assign_pointer(fdt->fd[fd], NULL);
David Howells1dce27c2012-02-16 17:49:42 +00001131 __clear_close_on_exec(fd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 __put_unused_fd(files, fd);
1133 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -07001134 retval = filp_close(filp, files);
1135
1136 /* can't restart close syscall because file table entry was cleared */
1137 if (unlikely(retval == -ERESTARTSYS ||
1138 retval == -ERESTARTNOINTR ||
1139 retval == -ERESTARTNOHAND ||
1140 retval == -ERESTART_RESTARTBLOCK))
1141 retval = -EINTR;
1142
1143 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
1145out_unlock:
1146 spin_unlock(&files->file_lock);
1147 return -EBADF;
1148}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149EXPORT_SYMBOL(sys_close);
1150
1151/*
1152 * This routine simulates a hangup on the tty, to arrange that users
1153 * are given clean terminals at login time.
1154 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001155SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156{
1157 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001158 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 return 0;
1160 }
1161 return -EPERM;
1162}
1163
1164/*
1165 * Called when an inode is about to be open.
1166 * We use this to disallow opening large files on 32bit systems if
1167 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1168 * on this flag in sys_open.
1169 */
1170int generic_file_open(struct inode * inode, struct file * filp)
1171{
1172 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001173 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 return 0;
1175}
1176
1177EXPORT_SYMBOL(generic_file_open);
1178
1179/*
1180 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001181 * file descriptors. The function is not supposed to ever fail, the only
1182 * reason it returns an 'int' and not 'void' is so that it can be plugged
1183 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 */
1185int nonseekable_open(struct inode *inode, struct file *filp)
1186{
1187 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1188 return 0;
1189}
1190
1191EXPORT_SYMBOL(nonseekable_open);