blob: e52389e1f05b4c010b49ef9a6e4d8cb11bfac504 [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 Bacik79124f12010-11-17 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;
Amit Arora97ac7352007-07-17 21:42:44 -0400236 /*
237 * Revalidate the write permissions, in case security policy has
238 * changed since the files were opened.
239 */
240 ret = security_file_permission(file, MAY_WRITE);
241 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400242 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400243
Amit Arora97ac7352007-07-17 21:42:44 -0400244 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400245 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400246
Amit Arora97ac7352007-07-17 21:42:44 -0400247 /*
248 * Let individual file system decide if it supports preallocation
249 * for directories or not.
250 */
251 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400252 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400253
Amit Arora97ac7352007-07-17 21:42:44 -0400254 /* Check for wrap through zero too */
255 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400256 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400257
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100258 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400259 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400260
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100261 return file->f_op->fallocate(file, mode, offset, len);
Amit Arora97ac7352007-07-17 21:42:44 -0400262}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400263
264SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
265{
266 struct file *file;
267 int error = -EBADF;
268
269 file = fget(fd);
270 if (file) {
271 error = do_fallocate(file, mode, offset, len);
272 fput(file);
273 }
274
275 return error;
276}
277
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100278#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
279asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
280{
281 return SYSC_fallocate((int)fd, (int)mode, offset, len);
282}
283SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
284#endif
Amit Arora97ac7352007-07-17 21:42:44 -0400285
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286/*
287 * access() needs to use the real uid/gid, not the effective uid/gid.
288 * We do this by temporarily clearing all FS-related capabilities and
289 * switching the fsuid/fsgid around to the real ones.
290 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100291SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
David Howellsd84f4f92008-11-14 10:39:23 +1100293 const struct cred *old_cred;
294 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400295 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400296 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 int res;
298
299 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
300 return -EINVAL;
301
David Howellsd84f4f92008-11-14 10:39:23 +1100302 override_cred = prepare_creds();
303 if (!override_cred)
304 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
David Howellsd84f4f92008-11-14 10:39:23 +1100306 override_cred->fsuid = override_cred->uid;
307 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700309 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100310 /* Clear the capabilities if we switch to a non-root user */
David Howellsd84f4f92008-11-14 10:39:23 +1100311 if (override_cred->uid)
312 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700313 else
David Howellsd84f4f92008-11-14 10:39:23 +1100314 override_cred->cap_effective =
315 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700316 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
David Howellsd84f4f92008-11-14 10:39:23 +1100318 old_cred = override_creds(override_cred);
319
Al Viro2d8f3032008-07-22 09:59:21 -0400320 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700321 if (res)
322 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Al Viro2d8f3032008-07-22 09:59:21 -0400324 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400325
326 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400327 /*
328 * MAY_EXEC on regular files is denied if the fs is mounted
329 * with the "noexec" flag.
330 */
331 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400332 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400333 goto out_path_release;
334 }
335
Al Viro256984a2008-07-22 08:09:30 -0400336 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700337 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400338 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700339 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800340 /*
341 * This is a rare case where using __mnt_is_readonly()
342 * is OK without a mnt_want/drop_write() pair. Since
343 * no actual write to the fs is performed here, we do
344 * not need to telegraph to that to anyone.
345 *
346 * By doing this, we accept that this access is
347 * inherently racy and know that the fs may change
348 * state before we even see this result.
349 */
Al Viro2d8f3032008-07-22 09:59:21 -0400350 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700351 res = -EROFS;
352
353out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400354 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700355out:
David Howellsd84f4f92008-11-14 10:39:23 +1100356 revert_creds(old_cred);
357 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 return res;
359}
360
Heiko Carstensca013e92009-01-14 14:14:19 +0100361SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800362{
363 return sys_faccessat(AT_FDCWD, filename, mode);
364}
365
Heiko Carstens3cdad422009-01-14 14:14:22 +0100366SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
Al Viro2d8f3032008-07-22 09:59:21 -0400368 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 int error;
370
Al Viro2d8f3032008-07-22 09:59:21 -0400371 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 if (error)
373 goto out;
374
Eric Paris9cfcac82010-07-23 11:43:51 -0400375 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 if (error)
377 goto dput_and_out;
378
Al Viro2d8f3032008-07-22 09:59:21 -0400379 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
381dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400382 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383out:
384 return error;
385}
386
Heiko Carstens3cdad422009-01-14 14:14:22 +0100387SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388{
389 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 int error;
392
393 error = -EBADF;
394 file = fget(fd);
395 if (!file)
396 goto out;
397
Jan Blunckac748a02008-02-14 19:34:39 -0800398 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
400 error = -ENOTDIR;
401 if (!S_ISDIR(inode->i_mode))
402 goto out_putf;
403
Eric Paris9cfcac82010-07-23 11:43:51 -0400404 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 if (!error)
Jan Blunckac748a02008-02-14 19:34:39 -0800406 set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407out_putf:
408 fput(file);
409out:
410 return error;
411}
412
Heiko Carstens3480b252009-01-14 14:14:16 +0100413SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414{
Al Viro2d8f3032008-07-22 09:59:21 -0400415 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 int error;
417
Al Viro2d8f3032008-07-22 09:59:21 -0400418 error = user_path_dir(filename, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 if (error)
420 goto out;
421
Eric Paris9cfcac82010-07-23 11:43:51 -0400422 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 if (error)
424 goto dput_and_out;
425
426 error = -EPERM;
427 if (!capable(CAP_SYS_CHROOT))
428 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900429 error = security_path_chroot(&path);
430 if (error)
431 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
Al Viro2d8f3032008-07-22 09:59:21 -0400433 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 error = 0;
435dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400436 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437out:
438 return error;
439}
440
Heiko Carstensa26eab22009-01-14 14:14:17 +0100441SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442{
443 struct inode * inode;
444 struct dentry * dentry;
445 struct file * file;
446 int err = -EBADF;
447 struct iattr newattrs;
448
449 file = fget(fd);
450 if (!file)
451 goto out;
452
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800453 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 inode = dentry->d_inode;
455
Al Viro5a190ae2007-06-07 12:19:32 -0400456 audit_inode(NULL, dentry);
Amy Griffis73241cc2005-11-03 16:00:25 +0000457
npiggin@suse.de96029c42009-04-26 20:25:55 +1000458 err = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800459 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 goto out_putf;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900461 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900462 err = security_path_chmod(dentry, file->f_vfsmnt, mode);
463 if (err)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900464 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (mode == (mode_t) -1)
466 mode = inode->i_mode;
467 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
468 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
469 err = notify_change(dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900470out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800471 mutex_unlock(&inode->i_mutex);
Dave Hansen2af482a2008-02-15 14:37:50 -0800472 mnt_drop_write(file->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473out_putf:
474 fput(file);
475out:
476 return err;
477}
478
Heiko Carstens6559eed82009-01-14 14:14:32 +0100479SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480{
Al Viro2d8f3032008-07-22 09:59:21 -0400481 struct path path;
482 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 int error;
484 struct iattr newattrs;
485
Al Viro2d8f3032008-07-22 09:59:21 -0400486 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 if (error)
488 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400489 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Al Viro2d8f3032008-07-22 09:59:21 -0400491 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800492 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 goto dput_and_out;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900494 mutex_lock(&inode->i_mutex);
Tetsuo Handa89eda062009-10-04 21:49:47 +0900495 error = security_path_chmod(path.dentry, path.mnt, mode);
496 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900497 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 if (mode == (mode_t) -1)
499 mode = inode->i_mode;
500 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
501 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Al Viro2d8f3032008-07-22 09:59:21 -0400502 error = notify_change(path.dentry, &newattrs);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900503out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800504 mutex_unlock(&inode->i_mutex);
Al Viro2d8f3032008-07-22 09:59:21 -0400505 mnt_drop_write(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400507 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508out:
509 return error;
510}
511
Heiko Carstensa26eab22009-01-14 14:14:17 +0100512SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800513{
514 return sys_fchmodat(AT_FDCWD, filename, mode);
515}
516
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900517static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900519 struct inode *inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 int error;
521 struct iattr newattrs;
522
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 newattrs.ia_valid = ATTR_CTIME;
524 if (user != (uid_t) -1) {
525 newattrs.ia_valid |= ATTR_UID;
526 newattrs.ia_uid = user;
527 }
528 if (group != (gid_t) -1) {
529 newattrs.ia_valid |= ATTR_GID;
530 newattrs.ia_gid = group;
531 }
532 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700533 newattrs.ia_valid |=
534 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800535 mutex_lock(&inode->i_mutex);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900536 error = security_path_chown(path, user, group);
537 if (!error)
538 error = notify_change(path->dentry, &newattrs);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800539 mutex_unlock(&inode->i_mutex);
Miklos Szeredibeb29e02008-07-01 15:01:29 +0200540
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 return error;
542}
543
Heiko Carstensca013e92009-01-14 14:14:19 +0100544SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
Al Viro2d8f3032008-07-22 09:59:21 -0400546 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 int error;
548
Al Viro2d8f3032008-07-22 09:59:21 -0400549 error = user_path(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700550 if (error)
551 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400552 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800553 if (error)
554 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900555 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400556 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800557out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400558 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700559out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 return error;
561}
562
Heiko Carstens6559eed82009-01-14 14:14:32 +0100563SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
564 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800565{
Al Viro2d8f3032008-07-22 09:59:21 -0400566 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800567 int error = -EINVAL;
568 int follow;
569
570 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
571 goto out;
572
573 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
Al Viro2d8f3032008-07-22 09:59:21 -0400574 error = user_path_at(dfd, filename, follow, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700575 if (error)
576 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400577 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800578 if (error)
579 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900580 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400581 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800582out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400583 path_put(&path);
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800584out:
585 return error;
586}
587
Heiko Carstensca013e92009-01-14 14:14:19 +0100588SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589{
Al Viro2d8f3032008-07-22 09:59:21 -0400590 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 int error;
592
Al Viro2d8f3032008-07-22 09:59:21 -0400593 error = user_lpath(filename, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700594 if (error)
595 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400596 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800597 if (error)
598 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900599 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400600 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800601out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400602 path_put(&path);
Dave Hansen6902d922006-09-30 23:29:01 -0700603out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 return error;
605}
606
Heiko Carstensca013e92009-01-14 14:14:19 +0100607SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608{
609 struct file * file;
610 int error = -EBADF;
Dave Hansen6902d922006-09-30 23:29:01 -0700611 struct dentry * dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 file = fget(fd);
Dave Hansen6902d922006-09-30 23:29:01 -0700614 if (!file)
615 goto out;
616
npiggin@suse.de96029c42009-04-26 20:25:55 +1000617 error = mnt_want_write_file(file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800618 if (error)
619 goto out_fput;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800620 dentry = file->f_path.dentry;
Al Viro5a190ae2007-06-07 12:19:32 -0400621 audit_inode(NULL, dentry);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900622 error = chown_common(&file->f_path, user, group);
Dave Hansen2af482a2008-02-15 14:37:50 -0800623 mnt_drop_write(file->f_path.mnt);
624out_fput:
Dave Hansen6902d922006-09-30 23:29:01 -0700625 fput(file);
626out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 return error;
628}
629
Dave Hansen4a3fd212008-02-15 14:37:48 -0800630/*
631 * You have to be very careful that these write
632 * counts get cleaned up in error cases and
633 * upon __fput(). This should probably never
634 * be called outside of __dentry_open().
635 */
636static inline int __get_file_write_access(struct inode *inode,
637 struct vfsmount *mnt)
638{
639 int error;
640 error = get_write_access(inode);
641 if (error)
642 return error;
643 /*
644 * Do not take mount writer counts on
645 * special files since no writes to
646 * the mount itself will occur.
647 */
648 if (!special_file(inode->i_mode)) {
649 /*
650 * Balanced in __fput()
651 */
652 error = mnt_want_write(mnt);
653 if (error)
654 put_write_access(inode);
655 }
656 return error;
657}
658
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700659static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
Al Viro482928d2009-12-19 10:10:39 -0500660 struct file *f,
David Howells745ca242008-11-14 10:39:22 +1100661 int (*open)(struct inode *, struct file *),
662 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 struct inode *inode;
665 int error;
666
Al Viro53009902009-12-19 10:15:07 -0500667 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700668 FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 inode = dentry->d_inode;
670 if (f->f_mode & FMODE_WRITE) {
Dave Hansen4a3fd212008-02-15 14:37:48 -0800671 error = __get_file_write_access(inode, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 if (error)
673 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800674 if (!special_file(inode->i_mode))
675 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 }
677
678 f->f_mapping = inode->i_mapping;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800679 f->f_path.dentry = dentry;
680 f->f_path.mnt = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 f->f_pos = 0;
682 f->f_op = fops_get(inode->i_fop);
Nick Pigginee2ffa02010-08-18 04:37:35 +1000683 file_sb_list_add(f, inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
David Howells745ca242008-11-14 10:39:22 +1100685 error = security_dentry_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900686 if (error)
687 goto cleanup_all;
688
Trond Myklebust834f2a42005-10-18 14:20:16 -0700689 if (!open && f->f_op)
690 open = f->f_op->open;
691 if (open) {
692 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 if (error)
694 goto cleanup_all;
695 }
Al Virob65a9cf2009-12-16 06:27:40 -0500696 ima_counts_get(f);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
699
700 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
701
702 /* NB: we're sure to have correct a_ops only after f_op->open */
703 if (f->f_flags & O_DIRECT) {
Carsten Otteceffc072005-06-23 22:05:25 -0700704 if (!f->f_mapping->a_ops ||
705 ((!f->f_mapping->a_ops->direct_IO) &&
Nick Piggin70688e42008-04-28 02:13:02 -0700706 (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 fput(f);
708 f = ERR_PTR(-EINVAL);
709 }
710 }
711
712 return f;
713
714cleanup_all:
715 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800716 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800718 if (!special_file(inode->i_mode)) {
719 /*
720 * We don't consider this a real
721 * mnt_want/drop_write() pair
722 * because it all happenend right
723 * here, so just reset the state.
724 */
725 file_reset_write(f);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800726 mnt_drop_write(mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800727 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800728 }
Nick Pigginee2ffa02010-08-18 04:37:35 +1000729 file_sb_list_del(f);
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800730 f->f_path.dentry = NULL;
731 f->f_path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732cleanup_file:
733 put_filp(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 dput(dentry);
735 mntput(mnt);
736 return ERR_PTR(error);
737}
738
Trond Myklebust834f2a42005-10-18 14:20:16 -0700739/**
740 * lookup_instantiate_filp - instantiates the open intent filp
741 * @nd: pointer to nameidata
742 * @dentry: pointer to dentry
743 * @open: open callback
744 *
745 * Helper for filesystems that want to use lookup open intents and pass back
746 * a fully instantiated struct file to the caller.
747 * This function is meant to be called from within a filesystem's
748 * lookup method.
Oleg Drokin9a56c212006-03-25 03:07:02 -0800749 * Beware of calling it for non-regular files! Those ->open methods might block
750 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
751 * leading to a deadlock, as nobody can open that fifo anymore, because
752 * another process to open fifo will block on locked parent when doing lookup).
Trond Myklebust834f2a42005-10-18 14:20:16 -0700753 * Note that in case of error, nd->intent.open.file is destroyed, but the
754 * path information remains valid.
755 * If the open callback is set to NULL, then the standard f_op->open()
756 * filesystem callback is substituted.
757 */
758struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
759 int (*open)(struct inode *, struct file *))
760{
David Howells745ca242008-11-14 10:39:22 +1100761 const struct cred *cred = current_cred();
762
Trond Myklebust834f2a42005-10-18 14:20:16 -0700763 if (IS_ERR(nd->intent.open.file))
764 goto out;
765 if (IS_ERR(dentry))
766 goto out_err;
Jan Blunck4ac91372008-02-14 19:34:32 -0800767 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
Trond Myklebust834f2a42005-10-18 14:20:16 -0700768 nd->intent.open.file,
David Howells745ca242008-11-14 10:39:22 +1100769 open, cred);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700770out:
771 return nd->intent.open.file;
772out_err:
773 release_open_intent(nd);
774 nd->intent.open.file = (struct file *)dentry;
775 goto out;
776}
777EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
778
779/**
780 * nameidata_to_filp - convert a nameidata to an open filp.
781 * @nd: pointer to nameidata
782 * @flags: open flags
783 *
784 * Note that this function destroys the original nameidata
785 */
Al Viro482928d2009-12-19 10:10:39 -0500786struct file *nameidata_to_filp(struct nameidata *nd)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700787{
David Howells745ca242008-11-14 10:39:22 +1100788 const struct cred *cred = current_cred();
Trond Myklebust834f2a42005-10-18 14:20:16 -0700789 struct file *filp;
790
791 /* Pick up the filp from the open intent */
792 filp = nd->intent.open.file;
793 /* Has the filesystem initialised the file for us? */
Al Virod893f1b2010-10-29 03:30:42 -0400794 if (filp->f_path.dentry == NULL) {
795 path_get(&nd->path);
Al Viro482928d2009-12-19 10:10:39 -0500796 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
David Howells745ca242008-11-14 10:39:22 +1100797 NULL, cred);
Al Virod893f1b2010-10-29 03:30:42 -0400798 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700799 return filp;
800}
801
Peter Staubach6fdcc212005-11-07 00:59:42 -0800802/*
803 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
804 * error.
805 */
David Howells745ca242008-11-14 10:39:22 +1100806struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
807 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700808{
809 int error;
810 struct file *f;
811
David Howellse0e81732009-09-02 09:13:40 +0100812 validate_creds(cred);
813
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800814 /*
815 * We must always pass in a valid mount pointer. Historically
816 * callers got away with not passing it, but we must enforce this at
817 * the earliest possible point now to avoid strange problems deep in the
818 * filesystem stack.
819 */
820 if (!mnt) {
821 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
822 dump_stack();
823 return ERR_PTR(-EINVAL);
824 }
825
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700826 error = -ENFILE;
827 f = get_empty_filp();
Peter Staubach6fdcc212005-11-07 00:59:42 -0800828 if (f == NULL) {
829 dput(dentry);
830 mntput(mnt);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700831 return ERR_PTR(error);
Peter Staubach6fdcc212005-11-07 00:59:42 -0800832 }
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700833
Al Viro482928d2009-12-19 10:10:39 -0500834 f->f_flags = flags;
835 return __dentry_open(dentry, mnt, f, NULL, cred);
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700836}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837EXPORT_SYMBOL(dentry_open);
838
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800839static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840{
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700841 struct fdtable *fdt = files_fdtable(files);
842 __FD_CLR(fd, fdt->open_fds);
Eric Dumazet0c9e63f2006-03-23 03:00:12 -0800843 if (fd < files->next_fd)
844 files->next_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845}
846
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800847void put_unused_fd(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 struct files_struct *files = current->files;
850 spin_lock(&files->file_lock);
851 __put_unused_fd(files, fd);
852 spin_unlock(&files->file_lock);
853}
854
855EXPORT_SYMBOL(put_unused_fd);
856
857/*
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800858 * Install a file pointer in the fd array.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 *
860 * The VFS is full of places where we drop the files lock between
861 * setting the open_fds bitmap and installing the file in the file
862 * array. At any such point, we are vulnerable to a dup2() race
863 * installing a file in the array before us. We need to detect this and
864 * fput() the struct file we are about to overwrite in this case.
865 *
866 * It should never happen - if we allow dup2() do it, _really_ bad things
867 * will follow.
868 */
869
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -0800870void fd_install(unsigned int fd, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871{
872 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700873 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700875 fdt = files_fdtable(files);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700876 BUG_ON(fdt->fd[fd] != NULL);
877 rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 spin_unlock(&files->file_lock);
879}
880
881EXPORT_SYMBOL(fd_install);
882
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800883long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884{
Miklos Szeredie922efc2005-09-06 15:18:25 -0700885 char *tmp = getname(filename);
886 int fd = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 if (!IS_ERR(tmp)) {
Ulrich Drepperf23513e2007-07-15 23:40:32 -0700889 fd = get_unused_fd_flags(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 if (fd >= 0) {
Al Viro6e8341a2009-04-06 11:16:22 -0400891 struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700892 if (IS_ERR(f)) {
893 put_unused_fd(fd);
894 fd = PTR_ERR(f);
895 } else {
Eric Paris2a12a9d2009-12-17 21:24:21 -0500896 fsnotify_open(f);
Telemaque Ndizihiwefed2fc12005-06-23 00:10:33 -0700897 fd_install(fd, f);
898 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 putname(tmp);
901 }
902 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903}
Miklos Szeredie922efc2005-09-06 15:18:25 -0700904
Heiko Carstensca013e92009-01-14 14:14:19 +0100905SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -0700906{
Linus Torvalds385910f2006-04-18 13:22:59 -0700907 long ret;
908
Miklos Szeredie922efc2005-09-06 15:18:25 -0700909 if (force_o_largefile())
910 flags |= O_LARGEFILE;
911
Linus Torvalds385910f2006-04-18 13:22:59 -0700912 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
913 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700914 asmlinkage_protect(3, ret, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700915 return ret;
Miklos Szeredie922efc2005-09-06 15:18:25 -0700916}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
Heiko Carstens6559eed82009-01-14 14:14:32 +0100918SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
919 int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800920{
Linus Torvalds385910f2006-04-18 13:22:59 -0700921 long ret;
922
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800923 if (force_o_largefile())
924 flags |= O_LARGEFILE;
925
Linus Torvalds385910f2006-04-18 13:22:59 -0700926 ret = do_sys_open(dfd, filename, flags, mode);
927 /* avoid REGPARM breakage on x86: */
Roland McGrath54a01512008-04-10 15:37:38 -0700928 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
Linus Torvalds385910f2006-04-18 13:22:59 -0700929 return ret;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800930}
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800931
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932#ifndef __alpha__
933
934/*
935 * For backward compatibility? Maybe this should be moved
936 * into arch/i386 instead?
937 */
Heiko Carstens002c8972009-01-14 14:14:18 +0100938SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939{
940 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
941}
942
943#endif
944
945/*
946 * "id" is the POSIX thread ID. We use the
947 * files pointer for this..
948 */
949int filp_close(struct file *filp, fl_owner_t id)
950{
Christoph Lameter45778ca2005-06-23 00:10:17 -0700951 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952
953 if (!file_count(filp)) {
954 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -0700955 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 }
957
Christoph Lameter45778ca2005-06-23 00:10:17 -0700958 if (filp->f_op && filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700959 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960
961 dnotify_flush(filp, id);
962 locks_remove_posix(filp, id);
963 fput(filp);
964 return retval;
965}
966
967EXPORT_SYMBOL(filp_close);
968
969/*
970 * Careful here! We test whether the file pointer is NULL before
971 * releasing the fd. This ensures that one clone task can't release
972 * an fd while another clone is opening it.
973 */
Heiko Carstensca013e92009-01-14 14:14:19 +0100974SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975{
976 struct file * filp;
977 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700978 struct fdtable *fdt;
Ernie Petridesee731f42006-09-29 02:00:13 -0700979 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
981 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700982 fdt = files_fdtable(files);
983 if (fd >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 goto out_unlock;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700985 filp = fdt->fd[fd];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 if (!filp)
987 goto out_unlock;
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -0700988 rcu_assign_pointer(fdt->fd[fd], NULL);
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700989 FD_CLR(fd, fdt->close_on_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 __put_unused_fd(files, fd);
991 spin_unlock(&files->file_lock);
Ernie Petridesee731f42006-09-29 02:00:13 -0700992 retval = filp_close(filp, files);
993
994 /* can't restart close syscall because file table entry was cleared */
995 if (unlikely(retval == -ERESTARTSYS ||
996 retval == -ERESTARTNOINTR ||
997 retval == -ERESTARTNOHAND ||
998 retval == -ERESTART_RESTARTBLOCK))
999 retval = -EINTR;
1000
1001 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
1003out_unlock:
1004 spin_unlock(&files->file_lock);
1005 return -EBADF;
1006}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007EXPORT_SYMBOL(sys_close);
1008
1009/*
1010 * This routine simulates a hangup on the tty, to arrange that users
1011 * are given clean terminals at login time.
1012 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001013SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014{
1015 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001016 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 return 0;
1018 }
1019 return -EPERM;
1020}
1021
1022/*
1023 * Called when an inode is about to be open.
1024 * We use this to disallow opening large files on 32bit systems if
1025 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1026 * on this flag in sys_open.
1027 */
1028int generic_file_open(struct inode * inode, struct file * filp)
1029{
1030 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001031 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 return 0;
1033}
1034
1035EXPORT_SYMBOL(generic_file_open);
1036
1037/*
1038 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001039 * file descriptors. The function is not supposed to ever fail, the only
1040 * reason it returns an 'int' and not 'void' is so that it can be plugged
1041 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 */
1043int nonseekable_open(struct inode *inode, struct file *filp)
1044{
1045 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1046 return 0;
1047}
1048
1049EXPORT_SYMBOL(nonseekable_open);