blob: b9ed8b25c108c69d68889b5b6eadac9878eb7bd9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/string.h>
8#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040010#include <linux/fdtable.h>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/tty.h>
14#include <linux/namei.h>
15#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080016#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070017#include <linux/securebits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080020#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090021#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040030#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050031#include <linux/ima.h>
Eric Paris2dfc1ca2009-12-17 20:30:52 -050032#include <linux/dnotify.h>
Al Viro3f6d0782013-02-24 13:49:08 -050033#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Eric Parise81e3f42009-12-04 15:47:36 -050035#include "internal.h"
36
NeilBrown4a301312006-01-08 01:02:39 -080037int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
38 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070039{
Amerigo Wang939a9422009-08-20 19:29:03 -070040 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 struct iattr newattrs;
42
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
44 if (length < 0)
45 return -EINVAL;
46
47 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080048 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080049 if (filp) {
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
52 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Linus Torvalds7b82dc02007-05-08 20:10:00 -070054 /* Remove suid/sgid on truncate too */
Amerigo Wang939a9422009-08-20 19:29:03 -070055 ret = should_remove_suid(dentry);
56 if (ret)
57 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070058
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080059 mutex_lock(&dentry->d_inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -040060 /* Note any delegations or leases have already been broken: */
61 ret = notify_change(dentry, &newattrs, NULL);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080062 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070063 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064}
65
David Howellsa02de962012-12-20 21:52:36 +000066long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070067{
Al Viro2d8f3032008-07-22 09:59:21 -040068 struct inode *inode;
David Howellsa02de962012-12-20 21:52:36 +000069 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
David Howellsa02de962012-12-20 21:52:36 +000071 inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
73 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 if (S_ISDIR(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000075 return -EISDIR;
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 if (!S_ISREG(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000077 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
David Howellsa02de962012-12-20 21:52:36 +000079 error = mnt_want_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 if (error)
David Howellsa02de962012-12-20 21:52:36 +000081 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Al Viro256984a2008-07-22 08:09:30 -040083 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080084 if (error)
85 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
87 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020088 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080089 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 error = get_write_access(inode);
92 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -080093 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
david m. richter97003822007-07-31 00:39:12 -070095 /*
96 * Make sure that there are no leases. get_write_access() protects
97 * against the truncate racing with a lease-granting setlease().
98 */
Al Viro8737c932009-12-24 06:47:55 -050099 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700100 if (error)
101 goto put_write_and_out;
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900104 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000105 error = security_path_truncate(path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500106 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000107 error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108
david m. richter97003822007-07-31 00:39:12 -0700109put_write_and_out:
110 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800111mnt_drop_write_and_out:
David Howellsa02de962012-12-20 21:52:36 +0000112 mnt_drop_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113out:
114 return error;
115}
David Howellsa02de962012-12-20 21:52:36 +0000116EXPORT_SYMBOL_GPL(vfs_truncate);
117
118static long do_sys_truncate(const char __user *pathname, loff_t length)
119{
Jeff Layton48f75302012-12-11 12:10:11 -0500120 unsigned int lookup_flags = LOOKUP_FOLLOW;
David Howellsa02de962012-12-20 21:52:36 +0000121 struct path path;
122 int error;
123
124 if (length < 0) /* sorry, but loff_t says... */
125 return -EINVAL;
126
Jeff Layton48f75302012-12-11 12:10:11 -0500127retry:
128 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
David Howellsa02de962012-12-20 21:52:36 +0000129 if (!error) {
130 error = vfs_truncate(&path, length);
131 path_put(&path);
132 }
Jeff Layton48f75302012-12-11 12:10:11 -0500133 if (retry_estale(error, lookup_flags)) {
134 lookup_flags |= LOOKUP_REVAL;
135 goto retry;
136 }
David Howellsa02de962012-12-20 21:52:36 +0000137 return error;
138}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200140SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200142 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143}
144
Al Viro3f6d0782013-02-24 13:49:08 -0500145#ifdef CONFIG_COMPAT
146COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
147{
148 return do_sys_truncate(path, length);
149}
150#endif
151
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800152static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153{
Al Virobf2965d2012-08-26 20:13:36 -0400154 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -0400156 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 int error;
158
159 error = -EINVAL;
160 if (length < 0)
161 goto out;
162 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400163 f = fdget(fd);
164 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 goto out;
166
167 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400168 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 small = 0;
170
Al Viro2903ff02012-08-28 12:52:22 -0400171 dentry = f.file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 inode = dentry->d_inode;
173 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400174 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 goto out_putf;
176
177 error = -EINVAL;
178 /* Cannot ftruncate over 2^31 bytes without large file support */
179 if (small && length > MAX_NON_LFS)
180 goto out_putf;
181
182 error = -EPERM;
183 if (IS_APPEND(inode))
184 goto out_putf;
185
Jan Kara14da9202012-06-12 16:20:37 +0200186 sb_start_write(inode->i_sb);
Al Viro2903ff02012-08-28 12:52:22 -0400187 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400189 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900190 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400191 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200192 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400194 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195out:
196 return error;
197}
198
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100199SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200{
Al Viro2cf09662013-01-21 15:25:54 -0500201 return do_sys_ftruncate(fd, length, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202}
203
Al Viro3f6d0782013-02-24 13:49:08 -0500204#ifdef CONFIG_COMPAT
205COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
206{
207 return do_sys_ftruncate(fd, length, 1);
208}
209#endif
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211/* LFS versions of truncate are only needed on 32 bit machines */
212#if BITS_PER_LONG == 32
Al Viro4a0fd5b2013-01-21 15:16:58 -0500213SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 return do_sys_truncate(path, length);
216}
217
Al Viro4a0fd5b2013-01-21 15:16:58 -0500218SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219{
Al Viro2cf09662013-01-21 15:25:54 -0500220 return do_sys_ftruncate(fd, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100222#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400224
225int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400226{
Al Viro496ad9a2013-01-23 17:07:38 -0500227 struct inode *inode = file_inode(file);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400228 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400229
230 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400231 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400232
233 /* Return error if mode is not supported */
Josef Bacik79124f182010-11-17 20:46:15 -0500234 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
235 return -EOPNOTSUPP;
236
237 /* Punch hole must have keep size set */
238 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
239 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400240 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400241
Amit Arora97ac7352007-07-17 21:42:44 -0400242 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400243 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100244
245 /* It's not possible punch hole on append only file */
246 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
247 return -EPERM;
248
249 if (IS_IMMUTABLE(inode))
250 return -EPERM;
251
Amit Arora97ac7352007-07-17 21:42:44 -0400252 /*
253 * Revalidate the write permissions, in case security policy has
254 * changed since the files were opened.
255 */
256 ret = security_file_permission(file, MAY_WRITE);
257 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400258 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400259
Amit Arora97ac7352007-07-17 21:42:44 -0400260 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400261 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400262
Amit Arora97ac7352007-07-17 21:42:44 -0400263 /*
264 * Let individual file system decide if it supports preallocation
265 * for directories or not.
266 */
267 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400268 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400269
Amit Arora97ac7352007-07-17 21:42:44 -0400270 /* Check for wrap through zero too */
271 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400272 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400273
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100274 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400275 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400276
Jan Kara14da9202012-06-12 16:20:37 +0200277 sb_start_write(inode->i_sb);
278 ret = file->f_op->fallocate(file, mode, offset, len);
279 sb_end_write(inode->i_sb);
280 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400281}
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400282
Al Viro4a0fd5b2013-01-21 15:16:58 -0500283SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400284{
Al Viro2903ff02012-08-28 12:52:22 -0400285 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400286 int error = -EBADF;
287
Al Viro2903ff02012-08-28 12:52:22 -0400288 if (f.file) {
289 error = do_fallocate(f.file, mode, offset, len);
290 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400291 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400292 return error;
293}
294
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295/*
296 * access() needs to use the real uid/gid, not the effective uid/gid.
297 * We do this by temporarily clearing all FS-related capabilities and
298 * switching the fsuid/fsgid around to the real ones.
299 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100300SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301{
David Howellsd84f4f92008-11-14 10:39:23 +1100302 const struct cred *old_cred;
303 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400304 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400305 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 int res;
Jeff Layton87fa5592012-12-11 12:10:11 -0500307 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
310 return -EINVAL;
311
David Howellsd84f4f92008-11-14 10:39:23 +1100312 override_cred = prepare_creds();
313 if (!override_cred)
314 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
David Howellsd84f4f92008-11-14 10:39:23 +1100316 override_cred->fsuid = override_cred->uid;
317 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700319 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100320 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800321 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
322 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100323 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700324 else
David Howellsd84f4f92008-11-14 10:39:23 +1100325 override_cred->cap_effective =
326 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700327 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
David Howellsd84f4f92008-11-14 10:39:23 +1100329 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500330retry:
331 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700332 if (res)
333 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
Al Viro2d8f3032008-07-22 09:59:21 -0400335 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400336
337 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400338 /*
339 * MAY_EXEC on regular files is denied if the fs is mounted
340 * with the "noexec" flag.
341 */
342 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400343 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400344 goto out_path_release;
345 }
346
Al Viro256984a2008-07-22 08:09:30 -0400347 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700348 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400349 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700350 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800351 /*
352 * This is a rare case where using __mnt_is_readonly()
353 * is OK without a mnt_want/drop_write() pair. Since
354 * no actual write to the fs is performed here, we do
355 * not need to telegraph to that to anyone.
356 *
357 * By doing this, we accept that this access is
358 * inherently racy and know that the fs may change
359 * state before we even see this result.
360 */
Al Viro2d8f3032008-07-22 09:59:21 -0400361 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700362 res = -EROFS;
363
364out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400365 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500366 if (retry_estale(res, lookup_flags)) {
367 lookup_flags |= LOOKUP_REVAL;
368 goto retry;
369 }
Dave Hansen6902d922006-09-30 23:29:01 -0700370out:
David Howellsd84f4f92008-11-14 10:39:23 +1100371 revert_creds(old_cred);
372 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 return res;
374}
375
Heiko Carstensca013e92009-01-14 14:14:19 +0100376SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800377{
378 return sys_faccessat(AT_FDCWD, filename, mode);
379}
380
Heiko Carstens3cdad422009-01-14 14:14:22 +0100381SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382{
Al Viro2d8f3032008-07-22 09:59:21 -0400383 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500385 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
386retry:
387 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 if (error)
389 goto out;
390
Eric Paris9cfcac82010-07-23 11:43:51 -0400391 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 if (error)
393 goto dput_and_out;
394
Al Viro2d8f3032008-07-22 09:59:21 -0400395 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
397dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400398 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500399 if (retry_estale(error, lookup_flags)) {
400 lookup_flags |= LOOKUP_REVAL;
401 goto retry;
402 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403out:
404 return error;
405}
406
Heiko Carstens3cdad422009-01-14 14:14:22 +0100407SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408{
Al Viro2903ff02012-08-28 12:52:22 -0400409 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400411 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
413 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400414 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 goto out;
416
Al Viro496ad9a2013-01-23 17:07:38 -0500417 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418
419 error = -ENOTDIR;
420 if (!S_ISDIR(inode->i_mode))
421 goto out_putf;
422
Eric Paris9cfcac82010-07-23 11:43:51 -0400423 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400425 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400427 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428out:
429 return error;
430}
431
Heiko Carstens3480b252009-01-14 14:14:16 +0100432SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
Al Viro2d8f3032008-07-22 09:59:21 -0400434 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500436 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
437retry:
438 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 if (error)
440 goto out;
441
Eric Paris9cfcac82010-07-23 11:43:51 -0400442 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 if (error)
444 goto dput_and_out;
445
446 error = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700447 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900449 error = security_path_chroot(&path);
450 if (error)
451 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Al Viro2d8f3032008-07-22 09:59:21 -0400453 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 error = 0;
455dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400456 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500457 if (retry_estale(error, lookup_flags)) {
458 lookup_flags |= LOOKUP_REVAL;
459 goto retry;
460 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461out:
462 return error;
463}
464
Al Viroe57712e2011-07-26 04:15:54 -0400465static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
Al Viroe57712e2011-07-26 04:15:54 -0400467 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400468 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400470 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Al Viroe57712e2011-07-26 04:15:54 -0400472 error = mnt_want_write(path->mnt);
473 if (error)
474 return error;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400475retry_deleg:
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900476 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500477 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400478 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900479 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
481 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400482 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900483out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800484 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400485 if (delegated_inode) {
486 error = break_deleg_wait(&delegated_inode);
487 if (!error)
488 goto retry_deleg;
489 }
Al Viroe57712e2011-07-26 04:15:54 -0400490 mnt_drop_write(path->mnt);
491 return error;
492}
493
Al Viro49f0a072011-07-26 04:22:01 -0400494SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400495{
Al Viro173c8402013-08-30 12:48:53 -0400496 struct fd f = fdget(fd);
Al Viroe57712e2011-07-26 04:15:54 -0400497 int err = -EBADF;
498
Al Viro173c8402013-08-30 12:48:53 -0400499 if (f.file) {
500 audit_inode(NULL, f.file->f_path.dentry, 0);
501 err = chmod_common(&f.file->f_path, mode);
502 fdput(f);
Al Viroe57712e2011-07-26 04:15:54 -0400503 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 return err;
505}
506
Al Viro49f0a072011-07-26 04:22:01 -0400507SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508{
Al Viro2d8f3032008-07-22 09:59:21 -0400509 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 int error;
Jeff Layton14ff6902012-12-11 12:10:13 -0500511 unsigned int lookup_flags = LOOKUP_FOLLOW;
512retry:
513 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400514 if (!error) {
515 error = chmod_common(&path, mode);
516 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500517 if (retry_estale(error, lookup_flags)) {
518 lookup_flags |= LOOKUP_REVAL;
519 goto retry;
520 }
Al Viroe57712e2011-07-26 04:15:54 -0400521 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 return error;
523}
524
Al Viro49f0a072011-07-26 04:22:01 -0400525SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800526{
527 return sys_fchmodat(AT_FDCWD, filename, mode);
528}
529
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900530static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900532 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400533 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 int error;
535 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800536 kuid_t uid;
537 kgid_t gid;
538
539 uid = make_kuid(current_user_ns(), user);
540 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 newattrs.ia_valid = ATTR_CTIME;
543 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800544 if (!uid_valid(uid))
545 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800547 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 }
549 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800550 if (!gid_valid(gid))
551 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800553 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 }
555 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700556 newattrs.ia_valid |=
557 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400558retry_deleg:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800559 mutex_lock(&inode->i_mutex);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600560 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900561 if (!error)
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400562 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800563 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400564 if (delegated_inode) {
565 error = break_deleg_wait(&delegated_inode);
566 if (!error)
567 goto retry_deleg;
568 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 return error;
570}
571
Heiko Carstens6559eed82009-01-14 14:14:32 +0100572SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
573 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800574{
Al Viro2d8f3032008-07-22 09:59:21 -0400575 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800576 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400577 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800578
Al Viro65cfc672011-03-13 15:56:26 -0400579 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800580 goto out;
581
Al Viro65cfc672011-03-13 15:56:26 -0400582 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
583 if (flag & AT_EMPTY_PATH)
584 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500585retry:
Al Viro65cfc672011-03-13 15:56:26 -0400586 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700587 if (error)
588 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400589 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800590 if (error)
591 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900592 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400593 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800594out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400595 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500596 if (retry_estale(error, lookup_flags)) {
597 lookup_flags |= LOOKUP_REVAL;
598 goto retry;
599 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800600out:
601 return error;
602}
603
David Howells55e4def2012-06-25 12:55:09 +0100604SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
605{
606 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
607}
608
Heiko Carstensca013e92009-01-14 14:14:19 +0100609SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610{
David Howells55e4def2012-06-25 12:55:09 +0100611 return sys_fchownat(AT_FDCWD, filename, user, group,
612 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613}
614
Heiko Carstensca013e92009-01-14 14:14:19 +0100615SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616{
Al Viro2903ff02012-08-28 12:52:22 -0400617 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 int error = -EBADF;
619
Al Viro2903ff02012-08-28 12:52:22 -0400620 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700621 goto out;
622
Al Viro2903ff02012-08-28 12:52:22 -0400623 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800624 if (error)
625 goto out_fput;
Jeff Laytonbfcec702012-10-10 15:25:23 -0400626 audit_inode(NULL, f.file->f_path.dentry, 0);
Al Viro2903ff02012-08-28 12:52:22 -0400627 error = chown_common(&f.file->f_path, user, group);
628 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800629out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400630 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700631out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 return error;
633}
634
Dave Hansen4a3fd212008-02-15 14:37:48 -0800635/*
636 * You have to be very careful that these write
637 * counts get cleaned up in error cases and
638 * upon __fput(). This should probably never
639 * be called outside of __dentry_open().
640 */
641static inline int __get_file_write_access(struct inode *inode,
642 struct vfsmount *mnt)
643{
644 int error;
645 error = get_write_access(inode);
646 if (error)
647 return error;
648 /*
649 * Do not take mount writer counts on
650 * special files since no writes to
651 * the mount itself will occur.
652 */
653 if (!special_file(inode->i_mode)) {
654 /*
655 * Balanced in __fput()
656 */
Jan Karaeb04c282012-06-12 16:20:35 +0200657 error = __mnt_want_write(mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800658 if (error)
659 put_write_access(inode);
660 }
661 return error;
662}
663
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200664int open_check_o_direct(struct file *f)
665{
666 /* NB: we're sure to have correct a_ops only after f_op->open */
667 if (f->f_flags & O_DIRECT) {
668 if (!f->f_mapping->a_ops ||
669 ((!f->f_mapping->a_ops->direct_IO) &&
670 (!f->f_mapping->a_ops->get_xip_mem))) {
671 return -EINVAL;
672 }
673 }
674 return 0;
675}
676
Al Viro02e51802012-06-10 14:32:45 -0400677static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400678 int (*open)(struct inode *, struct file *),
679 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680{
Al Viro1abf0c72011-03-13 03:51:11 -0400681 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 struct inode *inode;
683 int error;
684
Al Viro53009902009-12-19 10:15:07 -0500685 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700686 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400687
688 if (unlikely(f->f_flags & O_PATH))
689 f->f_mode = FMODE_PATH;
690
Al Virob5bcdda2012-07-20 23:28:46 +0400691 path_get(&f->f_path);
Al Virodd379782013-03-01 19:48:30 -0500692 inode = f->f_inode = f->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 if (f->f_mode & FMODE_WRITE) {
Al Viro02e51802012-06-10 14:32:45 -0400694 error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 if (error)
696 goto cleanup_file;
Dave Hansenad775f52008-02-15 14:38:01 -0800697 if (!special_file(inode->i_mode))
698 file_take_write(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 }
700
701 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Al Viro1abf0c72011-03-13 03:51:11 -0400703 if (unlikely(f->f_mode & FMODE_PATH)) {
704 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400705 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400706 }
707
Linus Torvalds9c225f22014-03-03 09:36:58 -0800708 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
709 if (S_ISREG(inode->i_mode))
710 f->f_mode |= FMODE_ATOMIC_POS;
711
Al Viro1abf0c72011-03-13 03:51:11 -0400712 f->f_op = fops_get(inode->i_fop);
Al Viro72c2d532013-09-22 16:27:52 -0400713 if (unlikely(WARN_ON(!f->f_op))) {
714 error = -ENODEV;
715 goto cleanup_all;
716 }
Al Viro1abf0c72011-03-13 03:51:11 -0400717
Eric Paris83d49852012-04-04 13:45:40 -0400718 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900719 if (error)
720 goto cleanup_all;
721
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400722 error = break_lease(inode, f->f_flags);
723 if (error)
724 goto cleanup_all;
725
Al Viro72c2d532013-09-22 16:27:52 -0400726 if (!open)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700727 open = f->f_op->open;
728 if (open) {
729 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 if (error)
731 goto cleanup_all;
732 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400733 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
734 i_readcount_inc(inode);
Trond Myklebust834f2a42005-10-18 14:20:16 -0700735
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
737
738 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
739
Al Viro96b7e572012-06-10 14:22:04 -0400740 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742cleanup_all:
743 fops_put(f->f_op);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800744 if (f->f_mode & FMODE_WRITE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 put_write_access(inode);
Dave Hansenad775f52008-02-15 14:38:01 -0800746 if (!special_file(inode->i_mode)) {
747 /*
748 * We don't consider this a real
749 * mnt_want/drop_write() pair
750 * because it all happenend right
751 * here, so just reset the state.
752 */
753 file_reset_write(f);
Al Virofe7c8052012-08-04 08:39:23 +0400754 __mnt_drop_write(f->f_path.mnt);
Dave Hansenad775f52008-02-15 14:38:01 -0800755 }
Dave Hansen4a3fd212008-02-15 14:37:48 -0800756 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400758 path_put(&f->f_path);
759 f->f_path.mnt = NULL;
760 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500761 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400762 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763}
764
Trond Myklebust834f2a42005-10-18 14:20:16 -0700765/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200766 * finish_open - finish opening a file
Miklos Szeredi0854d452013-09-16 14:51:55 +0200767 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200768 * @dentry: pointer to dentry
769 * @open: open callback
Miklos Szeredi0854d452013-09-16 14:51:55 +0200770 * @opened: state of open
Miklos Szeredid18e9002012-06-05 15:10:17 +0200771 *
772 * This can be used to finish opening a file passed to i_op->atomic_open().
773 *
774 * If the open callback is set to NULL, then the standard f_op->open()
775 * filesystem callback is substituted.
Miklos Szeredi0854d452013-09-16 14:51:55 +0200776 *
777 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
778 * the return value of d_splice_alias(), then the caller needs to perform dput()
779 * on it after finish_open().
780 *
781 * On successful return @file is a fully instantiated open file. After this, if
782 * an error occurs in ->atomic_open(), it needs to clean up with fput().
783 *
784 * Returns zero on success or -errno if the open failed.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200785 */
Al Viro30d90492012-06-22 12:40:19 +0400786int finish_open(struct file *file, struct dentry *dentry,
787 int (*open)(struct inode *, struct file *),
788 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200789{
Al Viro96b7e572012-06-10 14:22:04 -0400790 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400791 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200792
Al Virob5bcdda2012-07-20 23:28:46 +0400793 file->f_path.dentry = dentry;
Al Viro02e51802012-06-10 14:32:45 -0400794 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400795 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400796 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200797
Al Viro96b7e572012-06-10 14:22:04 -0400798 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200799}
800EXPORT_SYMBOL(finish_open);
801
802/**
803 * finish_no_open - finish ->atomic_open() without opening the file
804 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200805 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200806 * @dentry: dentry or NULL (as returned from ->lookup())
807 *
808 * This can be used to set the result of a successful lookup in ->atomic_open().
Miklos Szeredi0854d452013-09-16 14:51:55 +0200809 *
810 * NB: unlike finish_open() this function does consume the dentry reference and
811 * the caller need not dput() it.
812 *
813 * Returns "1" which must be the return value of ->atomic_open() after having
814 * called this function.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200815 */
Al Viroe45198a2012-06-10 06:48:09 -0400816int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200817{
Al Viro30d90492012-06-22 12:40:19 +0400818 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400819 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200820}
821EXPORT_SYMBOL(finish_no_open);
822
Al Viro765927b2012-06-26 21:58:53 +0400823struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100824 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700825{
826 int error;
827 struct file *f;
828
David Howellse0e81732009-09-02 09:13:40 +0100829 validate_creds(cred);
830
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900831 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400832 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800833
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700834 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500835 if (!IS_ERR(f)) {
836 f->f_flags = flags;
837 f->f_path = *path;
838 error = do_dentry_open(f, NULL, cred);
839 if (!error) {
840 /* from now on we need fput() to dispose of f */
841 error = open_check_o_direct(f);
842 if (error) {
843 fput(f);
844 f = ERR_PTR(error);
845 }
846 } else {
847 put_filp(f);
Al Viro2a027e72012-06-10 14:24:38 -0400848 f = ERR_PTR(error);
849 }
Al Viro2a027e72012-06-10 14:24:38 -0400850 }
851 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700852}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853EXPORT_SYMBOL(dentry_open);
854
Al Viroa218d0f2011-11-21 14:59:34 -0500855static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500856{
857 int lookup_flags = 0;
858 int acc_mode;
859
Andy Lutomirskie305f482013-08-01 21:07:52 -0700860 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200861 op->mode = (mode & S_IALLUGO) | S_IFREG;
862 else
863 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500864
865 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400866 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500867
868 /*
869 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
870 * check for O_DSYNC if the need any syncing at all we enforce it's
871 * always set instead of having to deal with possibly weird behaviour
872 * for malicious applications setting only __O_SYNC.
873 */
874 if (flags & __O_SYNC)
875 flags |= O_DSYNC;
876
Al Virobb458c62013-07-13 13:26:37 +0400877 if (flags & __O_TMPFILE) {
878 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400879 return -EINVAL;
880 acc_mode = MAY_OPEN | ACC_MODE(flags);
Al Viroba57ea62013-07-20 03:11:32 +0400881 if (!(acc_mode & MAY_WRITE))
882 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400883 } else if (flags & O_PATH) {
884 /*
885 * If we have O_PATH in the open flag. Then we
886 * cannot have anything other than the below set of flags
887 */
Al Viro1abf0c72011-03-13 03:51:11 -0400888 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
889 acc_mode = 0;
890 } else {
891 acc_mode = MAY_OPEN | ACC_MODE(flags);
892 }
Al Viro47c805d2011-02-23 17:44:09 -0500893
Al Viro1abf0c72011-03-13 03:51:11 -0400894 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500895
896 /* O_TRUNC implies we need access checks for write permissions */
897 if (flags & O_TRUNC)
898 acc_mode |= MAY_WRITE;
899
900 /* Allow the LSM permission hook to distinguish append
901 access from general write access. */
902 if (flags & O_APPEND)
903 acc_mode |= MAY_APPEND;
904
905 op->acc_mode = acc_mode;
906
Al Viro1abf0c72011-03-13 03:51:11 -0400907 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
908
Al Viro47c805d2011-02-23 17:44:09 -0500909 if (flags & O_CREAT) {
910 op->intent |= LOOKUP_CREATE;
911 if (flags & O_EXCL)
912 op->intent |= LOOKUP_EXCL;
913 }
914
915 if (flags & O_DIRECTORY)
916 lookup_flags |= LOOKUP_DIRECTORY;
917 if (!(flags & O_NOFOLLOW))
918 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +0400919 op->lookup_flags = lookup_flags;
920 return 0;
Al Viro47c805d2011-02-23 17:44:09 -0500921}
922
923/**
Jeff Layton669abf42012-10-10 16:43:10 -0400924 * file_open_name - open file and return file pointer
925 *
926 * @name: struct filename containing path to open
927 * @flags: open flags as per the open(2) second argument
928 * @mode: mode for the new file if O_CREAT is set, else ignored
929 *
930 * This is the helper to open a file from kernelspace if you really
931 * have to. But in generally you should not do this, so please move
932 * along, nothing to see here..
933 */
934struct file *file_open_name(struct filename *name, int flags, umode_t mode)
935{
936 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400937 int err = build_open_flags(flags, mode, &op);
938 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -0400939}
940
941/**
Al Viro47c805d2011-02-23 17:44:09 -0500942 * filp_open - open file and return file pointer
943 *
944 * @filename: path to open
945 * @flags: open flags as per the open(2) second argument
946 * @mode: mode for the new file if O_CREAT is set, else ignored
947 *
948 * This is the helper to open a file from kernelspace if you really
949 * have to. But in generally you should not do this, so please move
950 * along, nothing to see here..
951 */
Al Viroa218d0f2011-11-21 14:59:34 -0500952struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500953{
Jeff Layton669abf42012-10-10 16:43:10 -0400954 struct filename name = {.name = filename};
955 return file_open_name(&name, flags, mode);
Al Viro47c805d2011-02-23 17:44:09 -0500956}
957EXPORT_SYMBOL(filp_open);
958
Al Viro73d049a2011-03-11 12:08:24 -0500959struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
960 const char *filename, int flags)
961{
962 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400963 int err = build_open_flags(flags, 0, &op);
964 if (err)
965 return ERR_PTR(err);
Al Viro73d049a2011-03-11 12:08:24 -0500966 if (flags & O_CREAT)
967 return ERR_PTR(-EINVAL);
968 if (!filename && (flags & O_DIRECTORY))
969 if (!dentry->d_inode->i_op->lookup)
970 return ERR_PTR(-ENOTDIR);
Al Virof9652e12013-06-11 08:23:01 +0400971 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -0500972}
973EXPORT_SYMBOL(file_open_root);
974
Al Viroa218d0f2011-11-21 14:59:34 -0500975long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976{
Al Viro47c805d2011-02-23 17:44:09 -0500977 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400978 int fd = build_open_flags(flags, mode, &op);
979 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Al Virof9652e12013-06-11 08:23:01 +0400981 if (fd)
982 return fd;
983
984 tmp = getname(filename);
985 if (IS_ERR(tmp))
986 return PTR_ERR(tmp);
987
988 fd = get_unused_fd_flags(flags);
989 if (fd >= 0) {
990 struct file *f = do_filp_open(dfd, tmp, &op);
991 if (IS_ERR(f)) {
992 put_unused_fd(fd);
993 fd = PTR_ERR(f);
994 } else {
995 fsnotify_open(f);
996 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 }
Al Virof9652e12013-06-11 08:23:01 +0400999 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001002
Al Viroa218d0f2011-11-21 14:59:34 -05001003SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001004{
1005 if (force_o_largefile())
1006 flags |= O_LARGEFILE;
1007
Al Viro2cf09662013-01-21 15:25:54 -05001008 return do_sys_open(AT_FDCWD, filename, flags, mode);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001009}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Heiko Carstens6559eed82009-01-14 14:14:32 +01001011SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001012 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001013{
1014 if (force_o_largefile())
1015 flags |= O_LARGEFILE;
1016
Al Viro2cf09662013-01-21 15:25:54 -05001017 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001018}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020#ifndef __alpha__
1021
1022/*
1023 * For backward compatibility? Maybe this should be moved
1024 * into arch/i386 instead?
1025 */
Al Viroa218d0f2011-11-21 14:59:34 -05001026SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027{
1028 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1029}
1030
1031#endif
1032
1033/*
1034 * "id" is the POSIX thread ID. We use the
1035 * files pointer for this..
1036 */
1037int filp_close(struct file *filp, fl_owner_t id)
1038{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001039 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
1041 if (!file_count(filp)) {
1042 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001043 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 }
1045
Al Viro72c2d532013-09-22 16:27:52 -04001046 if (filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001047 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Al Viro1abf0c72011-03-13 03:51:11 -04001049 if (likely(!(filp->f_mode & FMODE_PATH))) {
1050 dnotify_flush(filp, id);
1051 locks_remove_posix(filp, id);
1052 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 fput(filp);
1054 return retval;
1055}
1056
1057EXPORT_SYMBOL(filp_close);
1058
1059/*
1060 * Careful here! We test whether the file pointer is NULL before
1061 * releasing the fd. This ensures that one clone task can't release
1062 * an fd while another clone is opening it.
1063 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001064SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065{
Al Viro483ce1d2012-08-19 12:04:24 -04001066 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001067
1068 /* can't restart close syscall because file table entry was cleared */
1069 if (unlikely(retval == -ERESTARTSYS ||
1070 retval == -ERESTARTNOINTR ||
1071 retval == -ERESTARTNOHAND ||
1072 retval == -ERESTART_RESTARTBLOCK))
1073 retval = -EINTR;
1074
1075 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077EXPORT_SYMBOL(sys_close);
1078
1079/*
1080 * This routine simulates a hangup on the tty, to arrange that users
1081 * are given clean terminals at login time.
1082 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001083SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
1085 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001086 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 return 0;
1088 }
1089 return -EPERM;
1090}
1091
1092/*
1093 * Called when an inode is about to be open.
1094 * We use this to disallow opening large files on 32bit systems if
1095 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1096 * on this flag in sys_open.
1097 */
1098int generic_file_open(struct inode * inode, struct file * filp)
1099{
1100 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001101 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 return 0;
1103}
1104
1105EXPORT_SYMBOL(generic_file_open);
1106
1107/*
1108 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001109 * file descriptors. The function is not supposed to ever fail, the only
1110 * reason it returns an 'int' and not 'void' is so that it can be plugged
1111 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 */
1113int nonseekable_open(struct inode *inode, struct file *filp)
1114{
1115 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1116 return 0;
1117}
1118
1119EXPORT_SYMBOL(nonseekable_open);