blob: b6f1e96a7c0b331b3e5a5d9bb4c014c5c9edc54b [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
Jan Kara45f147a2015-05-21 16:05:55 +020054 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
56 if (ret < 0)
57 return ret;
Amerigo Wang939a9422009-08-20 19:29:03 -070058 if (ret)
59 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070060
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080061 mutex_lock(&dentry->d_inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -040062 /* Note any delegations or leases have already been broken: */
63 ret = notify_change(dentry, &newattrs, NULL);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -080064 mutex_unlock(&dentry->d_inode->i_mutex);
Amerigo Wang939a9422009-08-20 19:29:03 -070065 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066}
67
David Howellsa02de962012-12-20 21:52:36 +000068long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070069{
Al Viro2d8f3032008-07-22 09:59:21 -040070 struct inode *inode;
David Howellsa02de962012-12-20 21:52:36 +000071 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
David Howellsa02de962012-12-20 21:52:36 +000073 inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
75 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 if (S_ISDIR(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000077 return -EISDIR;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 if (!S_ISREG(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000079 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070080
David Howellsa02de962012-12-20 21:52:36 +000081 error = mnt_want_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 if (error)
David Howellsa02de962012-12-20 21:52:36 +000083 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Al Viro256984a2008-07-22 08:09:30 -040085 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080086 if (error)
87 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
89 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020090 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080091 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 error = get_write_access(inode);
94 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -080095 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
david m. richter97003822007-07-31 00:39:12 -070097 /*
98 * Make sure that there are no leases. get_write_access() protects
99 * against the truncate racing with a lease-granting setlease().
100 */
Al Viro8737c932009-12-24 06:47:55 -0500101 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700102 if (error)
103 goto put_write_and_out;
104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900106 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000107 error = security_path_truncate(path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500108 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000109 error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
david m. richter97003822007-07-31 00:39:12 -0700111put_write_and_out:
112 put_write_access(inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800113mnt_drop_write_and_out:
David Howellsa02de962012-12-20 21:52:36 +0000114 mnt_drop_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115out:
116 return error;
117}
David Howellsa02de962012-12-20 21:52:36 +0000118EXPORT_SYMBOL_GPL(vfs_truncate);
119
120static long do_sys_truncate(const char __user *pathname, loff_t length)
121{
Jeff Layton48f75302012-12-11 12:10:11 -0500122 unsigned int lookup_flags = LOOKUP_FOLLOW;
David Howellsa02de962012-12-20 21:52:36 +0000123 struct path path;
124 int error;
125
126 if (length < 0) /* sorry, but loff_t says... */
127 return -EINVAL;
128
Jeff Layton48f75302012-12-11 12:10:11 -0500129retry:
130 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
David Howellsa02de962012-12-20 21:52:36 +0000131 if (!error) {
132 error = vfs_truncate(&path, length);
133 path_put(&path);
134 }
Jeff Layton48f75302012-12-11 12:10:11 -0500135 if (retry_estale(error, lookup_flags)) {
136 lookup_flags |= LOOKUP_REVAL;
137 goto retry;
138 }
David Howellsa02de962012-12-20 21:52:36 +0000139 return error;
140}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200142SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200144 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145}
146
Al Viro3f6d0782013-02-24 13:49:08 -0500147#ifdef CONFIG_COMPAT
148COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
149{
150 return do_sys_truncate(path, length);
151}
152#endif
153
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800154static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155{
Al Virobf2965d2012-08-26 20:13:36 -0400156 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -0400158 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 int error;
160
161 error = -EINVAL;
162 if (length < 0)
163 goto out;
164 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400165 f = fdget(fd);
166 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 goto out;
168
169 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400170 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 small = 0;
172
Al Viro2903ff02012-08-28 12:52:22 -0400173 dentry = f.file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 inode = dentry->d_inode;
175 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400176 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 goto out_putf;
178
179 error = -EINVAL;
180 /* Cannot ftruncate over 2^31 bytes without large file support */
181 if (small && length > MAX_NON_LFS)
182 goto out_putf;
183
184 error = -EPERM;
185 if (IS_APPEND(inode))
186 goto out_putf;
187
Jan Kara14da9202012-06-12 16:20:37 +0200188 sb_start_write(inode->i_sb);
Al Viro2903ff02012-08-28 12:52:22 -0400189 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400191 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900192 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400193 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200194 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400196 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197out:
198 return error;
199}
200
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100201SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
Al Viro2cf09662013-01-21 15:25:54 -0500203 return do_sys_ftruncate(fd, length, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204}
205
Al Viro3f6d0782013-02-24 13:49:08 -0500206#ifdef CONFIG_COMPAT
207COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
208{
209 return do_sys_ftruncate(fd, length, 1);
210}
211#endif
212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213/* LFS versions of truncate are only needed on 32 bit machines */
214#if BITS_PER_LONG == 32
Al Viro4a0fd5b2013-01-21 15:16:58 -0500215SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216{
217 return do_sys_truncate(path, length);
218}
219
Al Viro4a0fd5b2013-01-21 15:16:58 -0500220SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221{
Al Viro2cf09662013-01-21 15:25:54 -0500222 return do_sys_ftruncate(fd, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100224#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400226
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500227int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400228{
Al Viro496ad9a2013-01-23 17:07:38 -0500229 struct inode *inode = file_inode(file);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400230 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400231
232 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400233 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400234
235 /* Return error if mode is not supported */
Namjae Jeondd46c782015-03-25 15:07:05 +1100236 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
Lukas Czerner409332b2014-03-13 19:07:42 +1100237 return -EOPNOTSUPP;
238
239 /* Punch hole and zero range are mutually exclusive */
240 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
241 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
Josef Bacik79124f182010-11-17 20:46:15 -0500242 return -EOPNOTSUPP;
243
244 /* Punch hole must have keep size set */
245 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
246 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400247 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400248
Namjae Jeon00f5e612014-02-24 10:58:15 +1100249 /* Collapse range should only be used exclusively. */
250 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
251 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
252 return -EINVAL;
253
Namjae Jeondd46c782015-03-25 15:07:05 +1100254 /* Insert range should only be used exclusively. */
255 if ((mode & FALLOC_FL_INSERT_RANGE) &&
256 (mode & ~FALLOC_FL_INSERT_RANGE))
257 return -EINVAL;
258
Amit Arora97ac7352007-07-17 21:42:44 -0400259 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400260 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100261
Namjae Jeon00f5e612014-02-24 10:58:15 +1100262 /*
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400263 * We can only allow pure fallocate on append only files
Namjae Jeon00f5e612014-02-24 10:58:15 +1100264 */
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400265 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100266 return -EPERM;
267
268 if (IS_IMMUTABLE(inode))
269 return -EPERM;
270
Amit Arora97ac7352007-07-17 21:42:44 -0400271 /*
Eric Biggers6d2b6172014-06-24 23:45:08 -0500272 * We cannot allow any fallocate operation on an active swapfile
Lukas Czerner0790b312014-04-12 10:05:37 -0400273 */
274 if (IS_SWAPFILE(inode))
Eric Biggers6d2b6172014-06-24 23:45:08 -0500275 return -ETXTBSY;
Lukas Czerner0790b312014-04-12 10:05:37 -0400276
277 /*
Amit Arora97ac7352007-07-17 21:42:44 -0400278 * Revalidate the write permissions, in case security policy has
279 * changed since the files were opened.
280 */
281 ret = security_file_permission(file, MAY_WRITE);
282 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400283 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400284
Amit Arora97ac7352007-07-17 21:42:44 -0400285 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400286 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400287
Amit Arora97ac7352007-07-17 21:42:44 -0400288 /*
289 * Let individual file system decide if it supports preallocation
290 * for directories or not.
291 */
292 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400293 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400294
Amit Arora97ac7352007-07-17 21:42:44 -0400295 /* Check for wrap through zero too */
296 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400297 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400298
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100299 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400300 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400301
Jan Kara14da9202012-06-12 16:20:37 +0200302 sb_start_write(inode->i_sb);
303 ret = file->f_op->fallocate(file, mode, offset, len);
Heinrich Schuchardt820c12d2014-12-12 16:58:34 -0800304
305 /*
306 * Create inotify and fanotify events.
307 *
308 * To keep the logic simple always create events if fallocate succeeds.
309 * This implies that events are even created if the file size remains
310 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
311 */
312 if (ret == 0)
313 fsnotify_modify(file);
314
Jan Kara14da9202012-06-12 16:20:37 +0200315 sb_end_write(inode->i_sb);
316 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400317}
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500318EXPORT_SYMBOL_GPL(vfs_fallocate);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400319
Al Viro4a0fd5b2013-01-21 15:16:58 -0500320SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400321{
Al Viro2903ff02012-08-28 12:52:22 -0400322 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400323 int error = -EBADF;
324
Al Viro2903ff02012-08-28 12:52:22 -0400325 if (f.file) {
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500326 error = vfs_fallocate(f.file, mode, offset, len);
Al Viro2903ff02012-08-28 12:52:22 -0400327 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400328 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400329 return error;
330}
331
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332/*
333 * access() needs to use the real uid/gid, not the effective uid/gid.
334 * We do this by temporarily clearing all FS-related capabilities and
335 * switching the fsuid/fsgid around to the real ones.
336 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100337SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338{
David Howellsd84f4f92008-11-14 10:39:23 +1100339 const struct cred *old_cred;
340 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400341 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400342 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 int res;
Jeff Layton87fa5592012-12-11 12:10:11 -0500344 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
346 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
347 return -EINVAL;
348
David Howellsd84f4f92008-11-14 10:39:23 +1100349 override_cred = prepare_creds();
350 if (!override_cred)
351 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352
David Howellsd84f4f92008-11-14 10:39:23 +1100353 override_cred->fsuid = override_cred->uid;
354 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700356 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100357 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800358 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
359 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100360 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700361 else
David Howellsd84f4f92008-11-14 10:39:23 +1100362 override_cred->cap_effective =
363 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700364 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
David Howellsd84f4f92008-11-14 10:39:23 +1100366 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500367retry:
368 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700369 if (res)
370 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
David Howells63afdfc2015-05-06 15:59:00 +0100372 inode = d_backing_inode(path.dentry);
Al Viro256984a2008-07-22 08:09:30 -0400373
374 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400375 /*
376 * MAY_EXEC on regular files is denied if the fs is mounted
377 * with the "noexec" flag.
378 */
379 res = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500380 if (path_noexec(&path))
Al Viro30524472008-07-22 00:02:33 -0400381 goto out_path_release;
382 }
383
Al Viro256984a2008-07-22 08:09:30 -0400384 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700385 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400386 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700387 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800388 /*
389 * This is a rare case where using __mnt_is_readonly()
390 * is OK without a mnt_want/drop_write() pair. Since
391 * no actual write to the fs is performed here, we do
392 * not need to telegraph to that to anyone.
393 *
394 * By doing this, we accept that this access is
395 * inherently racy and know that the fs may change
396 * state before we even see this result.
397 */
Al Viro2d8f3032008-07-22 09:59:21 -0400398 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700399 res = -EROFS;
400
401out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400402 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500403 if (retry_estale(res, lookup_flags)) {
404 lookup_flags |= LOOKUP_REVAL;
405 goto retry;
406 }
Dave Hansen6902d922006-09-30 23:29:01 -0700407out:
David Howellsd84f4f92008-11-14 10:39:23 +1100408 revert_creds(old_cred);
409 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 return res;
411}
412
Heiko Carstensca013e92009-01-14 14:14:19 +0100413SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800414{
415 return sys_faccessat(AT_FDCWD, filename, mode);
416}
417
Heiko Carstens3cdad422009-01-14 14:14:22 +0100418SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419{
Al Viro2d8f3032008-07-22 09:59:21 -0400420 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500422 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
423retry:
424 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 if (error)
426 goto out;
427
Eric Paris9cfcac82010-07-23 11:43:51 -0400428 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 if (error)
430 goto dput_and_out;
431
Al Viro2d8f3032008-07-22 09:59:21 -0400432 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400435 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500436 if (retry_estale(error, lookup_flags)) {
437 lookup_flags |= LOOKUP_REVAL;
438 goto retry;
439 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440out:
441 return error;
442}
443
Heiko Carstens3cdad422009-01-14 14:14:22 +0100444SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445{
Al Viro2903ff02012-08-28 12:52:22 -0400446 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400448 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400451 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 goto out;
453
Al Viro496ad9a2013-01-23 17:07:38 -0500454 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
456 error = -ENOTDIR;
457 if (!S_ISDIR(inode->i_mode))
458 goto out_putf;
459
Eric Paris9cfcac82010-07-23 11:43:51 -0400460 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400462 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400464 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465out:
466 return error;
467}
468
Heiko Carstens3480b252009-01-14 14:14:16 +0100469SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
Al Viro2d8f3032008-07-22 09:59:21 -0400471 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500473 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
474retry:
475 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 if (error)
477 goto out;
478
Eric Paris9cfcac82010-07-23 11:43:51 -0400479 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 if (error)
481 goto dput_and_out;
482
483 error = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700484 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900486 error = security_path_chroot(&path);
487 if (error)
488 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
Al Viro2d8f3032008-07-22 09:59:21 -0400490 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 error = 0;
492dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400493 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500494 if (retry_estale(error, lookup_flags)) {
495 lookup_flags |= LOOKUP_REVAL;
496 goto retry;
497 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498out:
499 return error;
500}
501
Al Viroe57712e2011-07-26 04:15:54 -0400502static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503{
Al Viroe57712e2011-07-26 04:15:54 -0400504 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400505 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400507 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
Al Viroe57712e2011-07-26 04:15:54 -0400509 error = mnt_want_write(path->mnt);
510 if (error)
511 return error;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400512retry_deleg:
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900513 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500514 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400515 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900516 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
518 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400519 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900520out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800521 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400522 if (delegated_inode) {
523 error = break_deleg_wait(&delegated_inode);
524 if (!error)
525 goto retry_deleg;
526 }
Al Viroe57712e2011-07-26 04:15:54 -0400527 mnt_drop_write(path->mnt);
528 return error;
529}
530
Al Viro49f0a072011-07-26 04:22:01 -0400531SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400532{
Al Viro173c8402013-08-30 12:48:53 -0400533 struct fd f = fdget(fd);
Al Viroe57712e2011-07-26 04:15:54 -0400534 int err = -EBADF;
535
Al Viro173c8402013-08-30 12:48:53 -0400536 if (f.file) {
Al Viro9f45f5b2014-10-31 17:44:57 -0400537 audit_file(f.file);
Al Viro173c8402013-08-30 12:48:53 -0400538 err = chmod_common(&f.file->f_path, mode);
539 fdput(f);
Al Viroe57712e2011-07-26 04:15:54 -0400540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 return err;
542}
543
Al Viro49f0a072011-07-26 04:22:01 -0400544SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
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;
Jeff Layton14ff6902012-12-11 12:10:13 -0500548 unsigned int lookup_flags = LOOKUP_FOLLOW;
549retry:
550 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400551 if (!error) {
552 error = chmod_common(&path, mode);
553 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500554 if (retry_estale(error, lookup_flags)) {
555 lookup_flags |= LOOKUP_REVAL;
556 goto retry;
557 }
Al Viroe57712e2011-07-26 04:15:54 -0400558 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 return error;
560}
561
Al Viro49f0a072011-07-26 04:22:01 -0400562SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800563{
564 return sys_fchmodat(AT_FDCWD, filename, mode);
565}
566
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900567static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900569 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400570 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 int error;
572 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800573 kuid_t uid;
574 kgid_t gid;
575
576 uid = make_kuid(current_user_ns(), user);
577 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
Andrew Elblec1b89402015-02-23 08:51:24 -0500579retry_deleg:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 newattrs.ia_valid = ATTR_CTIME;
581 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800582 if (!uid_valid(uid))
583 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800585 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 }
587 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800588 if (!gid_valid(gid))
589 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800591 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 }
593 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700594 newattrs.ia_valid |=
595 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800596 mutex_lock(&inode->i_mutex);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600597 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900598 if (!error)
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400599 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800600 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400601 if (delegated_inode) {
602 error = break_deleg_wait(&delegated_inode);
603 if (!error)
604 goto retry_deleg;
605 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 return error;
607}
608
Heiko Carstens6559eed82009-01-14 14:14:32 +0100609SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
610 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800611{
Al Viro2d8f3032008-07-22 09:59:21 -0400612 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800613 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400614 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800615
Al Viro65cfc672011-03-13 15:56:26 -0400616 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800617 goto out;
618
Al Viro65cfc672011-03-13 15:56:26 -0400619 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
620 if (flag & AT_EMPTY_PATH)
621 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500622retry:
Al Viro65cfc672011-03-13 15:56:26 -0400623 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700624 if (error)
625 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400626 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800627 if (error)
628 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900629 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400630 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800631out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400632 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500633 if (retry_estale(error, lookup_flags)) {
634 lookup_flags |= LOOKUP_REVAL;
635 goto retry;
636 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800637out:
638 return error;
639}
640
David Howells55e4def2012-06-25 12:55:09 +0100641SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
642{
643 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
644}
645
Heiko Carstensca013e92009-01-14 14:14:19 +0100646SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647{
David Howells55e4def2012-06-25 12:55:09 +0100648 return sys_fchownat(AT_FDCWD, filename, user, group,
649 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650}
651
Heiko Carstensca013e92009-01-14 14:14:19 +0100652SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653{
Al Viro2903ff02012-08-28 12:52:22 -0400654 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 int error = -EBADF;
656
Al Viro2903ff02012-08-28 12:52:22 -0400657 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700658 goto out;
659
Al Viro2903ff02012-08-28 12:52:22 -0400660 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800661 if (error)
662 goto out_fput;
Al Viro9f45f5b2014-10-31 17:44:57 -0400663 audit_file(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400664 error = chown_common(&f.file->f_path, user, group);
665 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800666out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400667 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700668out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 return error;
670}
671
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200672int open_check_o_direct(struct file *f)
673{
674 /* NB: we're sure to have correct a_ops only after f_op->open */
675 if (f->f_flags & O_DIRECT) {
Matthew Wilcoxe748dcd2015-02-16 15:59:12 -0800676 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200677 return -EINVAL;
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200678 }
679 return 0;
680}
681
Al Viro02e51802012-06-10 14:32:45 -0400682static int do_dentry_open(struct file *f,
David Howells4bacc9c2015-06-18 14:32:31 +0100683 struct inode *inode,
Al Viro96b7e572012-06-10 14:22:04 -0400684 int (*open)(struct inode *, struct file *),
685 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686{
Al Viro1abf0c72011-03-13 03:51:11 -0400687 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 int error;
689
Al Viro53009902009-12-19 10:15:07 -0500690 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700691 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400692
Al Virob5bcdda2012-07-20 23:28:46 +0400693 path_get(&f->f_path);
David Howells4bacc9c2015-06-18 14:32:31 +0100694 f->f_inode = inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
Al Viro3f4d5a02014-03-14 09:43:29 -0400697 if (unlikely(f->f_flags & O_PATH)) {
698 f->f_mode = FMODE_PATH;
Al Viro1abf0c72011-03-13 03:51:11 -0400699 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400700 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400701 }
702
Al Virodd209082014-03-14 10:56:20 -0400703 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400704 error = get_write_access(inode);
Al Viro3f4d5a02014-03-14 09:43:29 -0400705 if (unlikely(error))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 goto cleanup_file;
Al Viro0ccb2862014-03-14 10:40:46 -0400707 error = __mnt_want_write(f->f_path.mnt);
Al Viro3f4d5a02014-03-14 09:43:29 -0400708 if (unlikely(error)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400709 put_write_access(inode);
710 goto cleanup_file;
711 }
Al Viro83f936c2014-03-14 12:02:47 -0400712 f->f_mode |= FMODE_WRITER;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 }
714
Linus Torvalds9c225f22014-03-03 09:36:58 -0800715 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
716 if (S_ISREG(inode->i_mode))
717 f->f_mode |= FMODE_ATOMIC_POS;
718
Al Viro1abf0c72011-03-13 03:51:11 -0400719 f->f_op = fops_get(inode->i_fop);
Al Viro72c2d532013-09-22 16:27:52 -0400720 if (unlikely(WARN_ON(!f->f_op))) {
721 error = -ENODEV;
722 goto cleanup_all;
723 }
Al Viro1abf0c72011-03-13 03:51:11 -0400724
Eric Paris83d49852012-04-04 13:45:40 -0400725 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900726 if (error)
727 goto cleanup_all;
728
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400729 error = break_lease(inode, f->f_flags);
730 if (error)
731 goto cleanup_all;
732
Al Viro72c2d532013-09-22 16:27:52 -0400733 if (!open)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700734 open = f->f_op->open;
735 if (open) {
736 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 if (error)
738 goto cleanup_all;
739 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400740 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
741 i_readcount_inc(inode);
Al Viro293bc982014-02-11 18:37:41 -0500742 if ((f->f_mode & FMODE_READ) &&
Al Viro84363182015-04-04 01:14:53 -0400743 likely(f->f_op->read || f->f_op->read_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500744 f->f_mode |= FMODE_CAN_READ;
Al Viro293bc982014-02-11 18:37:41 -0500745 if ((f->f_mode & FMODE_WRITE) &&
Al Viro84363182015-04-04 01:14:53 -0400746 likely(f->f_op->write || f->f_op->write_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500747 f->f_mode |= FMODE_CAN_WRITE;
Trond Myklebust834f2a42005-10-18 14:20:16 -0700748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
750
751 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
752
Al Viro96b7e572012-06-10 14:22:04 -0400753 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
755cleanup_all:
756 fops_put(f->f_op);
Al Viro83f936c2014-03-14 12:02:47 -0400757 if (f->f_mode & FMODE_WRITER) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 put_write_access(inode);
Al Viro83f936c2014-03-14 12:02:47 -0400759 __mnt_drop_write(f->f_path.mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800760 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400762 path_put(&f->f_path);
763 f->f_path.mnt = NULL;
764 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500765 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400766 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767}
768
Trond Myklebust834f2a42005-10-18 14:20:16 -0700769/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200770 * finish_open - finish opening a file
Miklos Szeredi0854d452013-09-16 14:51:55 +0200771 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200772 * @dentry: pointer to dentry
773 * @open: open callback
Miklos Szeredi0854d452013-09-16 14:51:55 +0200774 * @opened: state of open
Miklos Szeredid18e9002012-06-05 15:10:17 +0200775 *
776 * This can be used to finish opening a file passed to i_op->atomic_open().
777 *
778 * If the open callback is set to NULL, then the standard f_op->open()
779 * filesystem callback is substituted.
Miklos Szeredi0854d452013-09-16 14:51:55 +0200780 *
781 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
782 * the return value of d_splice_alias(), then the caller needs to perform dput()
783 * on it after finish_open().
784 *
785 * On successful return @file is a fully instantiated open file. After this, if
786 * an error occurs in ->atomic_open(), it needs to clean up with fput().
787 *
788 * Returns zero on success or -errno if the open failed.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200789 */
Al Viro30d90492012-06-22 12:40:19 +0400790int finish_open(struct file *file, struct dentry *dentry,
791 int (*open)(struct inode *, struct file *),
792 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200793{
Al Viro96b7e572012-06-10 14:22:04 -0400794 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400795 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200796
Al Virob5bcdda2012-07-20 23:28:46 +0400797 file->f_path.dentry = dentry;
David Howells4bacc9c2015-06-18 14:32:31 +0100798 error = do_dentry_open(file, d_backing_inode(dentry), open,
799 current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400800 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400801 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200802
Al Viro96b7e572012-06-10 14:22:04 -0400803 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200804}
805EXPORT_SYMBOL(finish_open);
806
807/**
808 * finish_no_open - finish ->atomic_open() without opening the file
809 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200810 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200811 * @dentry: dentry or NULL (as returned from ->lookup())
812 *
813 * This can be used to set the result of a successful lookup in ->atomic_open().
Miklos Szeredi0854d452013-09-16 14:51:55 +0200814 *
815 * NB: unlike finish_open() this function does consume the dentry reference and
816 * the caller need not dput() it.
817 *
818 * Returns "1" which must be the return value of ->atomic_open() after having
819 * called this function.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200820 */
Al Viroe45198a2012-06-10 06:48:09 -0400821int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200822{
Al Viro30d90492012-06-22 12:40:19 +0400823 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400824 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200825}
826EXPORT_SYMBOL(finish_no_open);
827
Miklos Szeredi9bf39ab2015-06-19 10:29:13 +0200828char *file_path(struct file *filp, char *buf, int buflen)
829{
830 return d_path(&filp->f_path, buf, buflen);
831}
832EXPORT_SYMBOL(file_path);
833
David Howells4bacc9c2015-06-18 14:32:31 +0100834/**
835 * vfs_open - open the file at the given path
836 * @path: path to open
837 * @file: newly allocated file with f_flag initialized
838 * @cred: credentials to use
839 */
840int vfs_open(const struct path *path, struct file *file,
841 const struct cred *cred)
842{
843 struct dentry *dentry = path->dentry;
844 struct inode *inode = dentry->d_inode;
845
846 file->f_path = *path;
847 if (dentry->d_flags & DCACHE_OP_SELECT_INODE) {
848 inode = dentry->d_op->d_select_inode(dentry, file->f_flags);
849 if (IS_ERR(inode))
850 return PTR_ERR(inode);
851 }
852
853 return do_dentry_open(file, inode, NULL, cred);
854}
855
Al Viro765927b2012-06-26 21:58:53 +0400856struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100857 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700858{
859 int error;
860 struct file *f;
861
David Howellse0e81732009-09-02 09:13:40 +0100862 validate_creds(cred);
863
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900864 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400865 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800866
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700867 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500868 if (!IS_ERR(f)) {
869 f->f_flags = flags;
Miklos Szeredi4aa7c632014-10-24 00:14:35 +0200870 error = vfs_open(path, f, cred);
Al Viro1afc99b2013-02-14 20:41:04 -0500871 if (!error) {
872 /* from now on we need fput() to dispose of f */
873 error = open_check_o_direct(f);
874 if (error) {
875 fput(f);
876 f = ERR_PTR(error);
877 }
878 } else {
879 put_filp(f);
Al Viro2a027e72012-06-10 14:24:38 -0400880 f = ERR_PTR(error);
881 }
Al Viro2a027e72012-06-10 14:24:38 -0400882 }
883 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700884}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885EXPORT_SYMBOL(dentry_open);
886
Al Viroa218d0f2011-11-21 14:59:34 -0500887static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500888{
889 int lookup_flags = 0;
890 int acc_mode;
891
Andy Lutomirskie305f482013-08-01 21:07:52 -0700892 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200893 op->mode = (mode & S_IALLUGO) | S_IFREG;
894 else
895 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500896
897 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400898 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500899
900 /*
901 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
902 * check for O_DSYNC if the need any syncing at all we enforce it's
903 * always set instead of having to deal with possibly weird behaviour
904 * for malicious applications setting only __O_SYNC.
905 */
906 if (flags & __O_SYNC)
907 flags |= O_DSYNC;
908
Al Virobb458c62013-07-13 13:26:37 +0400909 if (flags & __O_TMPFILE) {
910 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400911 return -EINVAL;
912 acc_mode = MAY_OPEN | ACC_MODE(flags);
Al Viroba57ea62013-07-20 03:11:32 +0400913 if (!(acc_mode & MAY_WRITE))
914 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400915 } else if (flags & O_PATH) {
916 /*
917 * If we have O_PATH in the open flag. Then we
918 * cannot have anything other than the below set of flags
919 */
Al Viro1abf0c72011-03-13 03:51:11 -0400920 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
921 acc_mode = 0;
922 } else {
923 acc_mode = MAY_OPEN | ACC_MODE(flags);
924 }
Al Viro47c805d2011-02-23 17:44:09 -0500925
Al Viro1abf0c72011-03-13 03:51:11 -0400926 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500927
928 /* O_TRUNC implies we need access checks for write permissions */
929 if (flags & O_TRUNC)
930 acc_mode |= MAY_WRITE;
931
932 /* Allow the LSM permission hook to distinguish append
933 access from general write access. */
934 if (flags & O_APPEND)
935 acc_mode |= MAY_APPEND;
936
937 op->acc_mode = acc_mode;
938
Al Viro1abf0c72011-03-13 03:51:11 -0400939 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
940
Al Viro47c805d2011-02-23 17:44:09 -0500941 if (flags & O_CREAT) {
942 op->intent |= LOOKUP_CREATE;
943 if (flags & O_EXCL)
944 op->intent |= LOOKUP_EXCL;
945 }
946
947 if (flags & O_DIRECTORY)
948 lookup_flags |= LOOKUP_DIRECTORY;
949 if (!(flags & O_NOFOLLOW))
950 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +0400951 op->lookup_flags = lookup_flags;
952 return 0;
Al Viro47c805d2011-02-23 17:44:09 -0500953}
954
955/**
Jeff Layton669abf42012-10-10 16:43:10 -0400956 * file_open_name - open file and return file pointer
957 *
958 * @name: struct filename containing path to open
959 * @flags: open flags as per the open(2) second argument
960 * @mode: mode for the new file if O_CREAT is set, else ignored
961 *
962 * This is the helper to open a file from kernelspace if you really
963 * have to. But in generally you should not do this, so please move
964 * along, nothing to see here..
965 */
966struct file *file_open_name(struct filename *name, int flags, umode_t mode)
967{
968 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400969 int err = build_open_flags(flags, mode, &op);
970 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -0400971}
972
973/**
Al Viro47c805d2011-02-23 17:44:09 -0500974 * filp_open - open file and return file pointer
975 *
976 * @filename: path to open
977 * @flags: open flags as per the open(2) second argument
978 * @mode: mode for the new file if O_CREAT is set, else ignored
979 *
980 * This is the helper to open a file from kernelspace if you really
981 * have to. But in generally you should not do this, so please move
982 * along, nothing to see here..
983 */
Al Viroa218d0f2011-11-21 14:59:34 -0500984struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500985{
Paul Moore51689102015-01-22 00:00:03 -0500986 struct filename *name = getname_kernel(filename);
987 struct file *file = ERR_CAST(name);
988
989 if (!IS_ERR(name)) {
990 file = file_open_name(name, flags, mode);
991 putname(name);
992 }
993 return file;
Al Viro47c805d2011-02-23 17:44:09 -0500994}
995EXPORT_SYMBOL(filp_open);
996
Al Viro73d049a2011-03-11 12:08:24 -0500997struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
998 const char *filename, int flags)
999{
1000 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001001 int err = build_open_flags(flags, 0, &op);
1002 if (err)
1003 return ERR_PTR(err);
Al Viro73d049a2011-03-11 12:08:24 -05001004 if (flags & O_CREAT)
1005 return ERR_PTR(-EINVAL);
Al Virof9652e12013-06-11 08:23:01 +04001006 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -05001007}
1008EXPORT_SYMBOL(file_open_root);
1009
Al Viroa218d0f2011-11-21 14:59:34 -05001010long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011{
Al Viro47c805d2011-02-23 17:44:09 -05001012 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001013 int fd = build_open_flags(flags, mode, &op);
1014 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
Al Virof9652e12013-06-11 08:23:01 +04001016 if (fd)
1017 return fd;
1018
1019 tmp = getname(filename);
1020 if (IS_ERR(tmp))
1021 return PTR_ERR(tmp);
1022
1023 fd = get_unused_fd_flags(flags);
1024 if (fd >= 0) {
1025 struct file *f = do_filp_open(dfd, tmp, &op);
1026 if (IS_ERR(f)) {
1027 put_unused_fd(fd);
1028 fd = PTR_ERR(f);
1029 } else {
1030 fsnotify_open(f);
1031 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 }
Al Virof9652e12013-06-11 08:23:01 +04001034 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001037
Al Viroa218d0f2011-11-21 14:59:34 -05001038SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001039{
1040 if (force_o_largefile())
1041 flags |= O_LARGEFILE;
1042
Al Viro2cf09662013-01-21 15:25:54 -05001043 return do_sys_open(AT_FDCWD, filename, flags, mode);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001044}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045
Heiko Carstens6559eed82009-01-14 14:14:32 +01001046SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001047 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001048{
1049 if (force_o_largefile())
1050 flags |= O_LARGEFILE;
1051
Al Viro2cf09662013-01-21 15:25:54 -05001052 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001053}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001054
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055#ifndef __alpha__
1056
1057/*
1058 * For backward compatibility? Maybe this should be moved
1059 * into arch/i386 instead?
1060 */
Al Viroa218d0f2011-11-21 14:59:34 -05001061SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062{
1063 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1064}
1065
1066#endif
1067
1068/*
1069 * "id" is the POSIX thread ID. We use the
1070 * files pointer for this..
1071 */
1072int filp_close(struct file *filp, fl_owner_t id)
1073{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001074 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
1076 if (!file_count(filp)) {
1077 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001078 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 }
1080
Al Viro72c2d532013-09-22 16:27:52 -04001081 if (filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001082 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Al Viro1abf0c72011-03-13 03:51:11 -04001084 if (likely(!(filp->f_mode & FMODE_PATH))) {
1085 dnotify_flush(filp, id);
1086 locks_remove_posix(filp, id);
1087 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 fput(filp);
1089 return retval;
1090}
1091
1092EXPORT_SYMBOL(filp_close);
1093
1094/*
1095 * Careful here! We test whether the file pointer is NULL before
1096 * releasing the fd. This ensures that one clone task can't release
1097 * an fd while another clone is opening it.
1098 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001099SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100{
Al Viro483ce1d2012-08-19 12:04:24 -04001101 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001102
1103 /* can't restart close syscall because file table entry was cleared */
1104 if (unlikely(retval == -ERESTARTSYS ||
1105 retval == -ERESTARTNOINTR ||
1106 retval == -ERESTARTNOHAND ||
1107 retval == -ERESTART_RESTARTBLOCK))
1108 retval = -EINTR;
1109
1110 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112EXPORT_SYMBOL(sys_close);
1113
1114/*
1115 * This routine simulates a hangup on the tty, to arrange that users
1116 * are given clean terminals at login time.
1117 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001118SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119{
1120 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001121 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 return 0;
1123 }
1124 return -EPERM;
1125}
1126
1127/*
1128 * Called when an inode is about to be open.
1129 * We use this to disallow opening large files on 32bit systems if
1130 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1131 * on this flag in sys_open.
1132 */
1133int generic_file_open(struct inode * inode, struct file * filp)
1134{
1135 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001136 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 return 0;
1138}
1139
1140EXPORT_SYMBOL(generic_file_open);
1141
1142/*
1143 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001144 * file descriptors. The function is not supposed to ever fail, the only
1145 * reason it returns an 'int' and not 'void' is so that it can be plugged
1146 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 */
1148int nonseekable_open(struct inode *inode, struct file *filp)
1149{
1150 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1151 return 0;
1152}
1153
1154EXPORT_SYMBOL(nonseekable_open);