blob: d36c42ff019d425e43dd2bdf21ac66842919ee54 [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
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500225int vfs_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 */
Namjae Jeon00f5e612014-02-24 10:58:15 +1100234 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
Lukas Czerner409332b2014-03-13 19:07:42 +1100235 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
236 return -EOPNOTSUPP;
237
238 /* Punch hole and zero range are mutually exclusive */
239 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
240 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
Josef Bacik79124f182010-11-17 20:46:15 -0500241 return -EOPNOTSUPP;
242
243 /* Punch hole must have keep size set */
244 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
245 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400246 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400247
Namjae Jeon00f5e612014-02-24 10:58:15 +1100248 /* Collapse range should only be used exclusively. */
249 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
250 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
251 return -EINVAL;
252
Amit Arora97ac7352007-07-17 21:42:44 -0400253 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400254 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100255
Namjae Jeon00f5e612014-02-24 10:58:15 +1100256 /*
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400257 * We can only allow pure fallocate on append only files
Namjae Jeon00f5e612014-02-24 10:58:15 +1100258 */
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400259 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100260 return -EPERM;
261
262 if (IS_IMMUTABLE(inode))
263 return -EPERM;
264
Amit Arora97ac7352007-07-17 21:42:44 -0400265 /*
Eric Biggers6d2b6172014-06-24 23:45:08 -0500266 * We cannot allow any fallocate operation on an active swapfile
Lukas Czerner0790b312014-04-12 10:05:37 -0400267 */
268 if (IS_SWAPFILE(inode))
Eric Biggers6d2b6172014-06-24 23:45:08 -0500269 return -ETXTBSY;
Lukas Czerner0790b312014-04-12 10:05:37 -0400270
271 /*
Amit Arora97ac7352007-07-17 21:42:44 -0400272 * Revalidate the write permissions, in case security policy has
273 * changed since the files were opened.
274 */
275 ret = security_file_permission(file, MAY_WRITE);
276 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400277 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400278
Amit Arora97ac7352007-07-17 21:42:44 -0400279 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400280 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400281
Amit Arora97ac7352007-07-17 21:42:44 -0400282 /*
283 * Let individual file system decide if it supports preallocation
284 * for directories or not.
285 */
286 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400287 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400288
Amit Arora97ac7352007-07-17 21:42:44 -0400289 /* Check for wrap through zero too */
290 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400291 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400292
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100293 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400294 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400295
Jan Kara14da9202012-06-12 16:20:37 +0200296 sb_start_write(inode->i_sb);
297 ret = file->f_op->fallocate(file, mode, offset, len);
Heinrich Schuchardt820c12d2014-12-12 16:58:34 -0800298
299 /*
300 * Create inotify and fanotify events.
301 *
302 * To keep the logic simple always create events if fallocate succeeds.
303 * This implies that events are even created if the file size remains
304 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
305 */
306 if (ret == 0)
307 fsnotify_modify(file);
308
Jan Kara14da9202012-06-12 16:20:37 +0200309 sb_end_write(inode->i_sb);
310 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400311}
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500312EXPORT_SYMBOL_GPL(vfs_fallocate);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400313
Al Viro4a0fd5b2013-01-21 15:16:58 -0500314SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400315{
Al Viro2903ff02012-08-28 12:52:22 -0400316 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400317 int error = -EBADF;
318
Al Viro2903ff02012-08-28 12:52:22 -0400319 if (f.file) {
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500320 error = vfs_fallocate(f.file, mode, offset, len);
Al Viro2903ff02012-08-28 12:52:22 -0400321 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400322 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400323 return error;
324}
325
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326/*
327 * access() needs to use the real uid/gid, not the effective uid/gid.
328 * We do this by temporarily clearing all FS-related capabilities and
329 * switching the fsuid/fsgid around to the real ones.
330 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100331SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332{
David Howellsd84f4f92008-11-14 10:39:23 +1100333 const struct cred *old_cred;
334 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400335 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400336 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 int res;
Jeff Layton87fa5592012-12-11 12:10:11 -0500338 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
340 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
341 return -EINVAL;
342
David Howellsd84f4f92008-11-14 10:39:23 +1100343 override_cred = prepare_creds();
344 if (!override_cred)
345 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346
David Howellsd84f4f92008-11-14 10:39:23 +1100347 override_cred->fsuid = override_cred->uid;
348 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700350 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100351 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800352 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
353 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100354 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700355 else
David Howellsd84f4f92008-11-14 10:39:23 +1100356 override_cred->cap_effective =
357 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700358 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
David Howellsd84f4f92008-11-14 10:39:23 +1100360 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500361retry:
362 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700363 if (res)
364 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Al Viro2d8f3032008-07-22 09:59:21 -0400366 inode = path.dentry->d_inode;
Al Viro256984a2008-07-22 08:09:30 -0400367
368 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400369 /*
370 * MAY_EXEC on regular files is denied if the fs is mounted
371 * with the "noexec" flag.
372 */
373 res = -EACCES;
Al Viro2d8f3032008-07-22 09:59:21 -0400374 if (path.mnt->mnt_flags & MNT_NOEXEC)
Al Viro30524472008-07-22 00:02:33 -0400375 goto out_path_release;
376 }
377
Al Viro256984a2008-07-22 08:09:30 -0400378 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700379 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400380 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700381 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800382 /*
383 * This is a rare case where using __mnt_is_readonly()
384 * is OK without a mnt_want/drop_write() pair. Since
385 * no actual write to the fs is performed here, we do
386 * not need to telegraph to that to anyone.
387 *
388 * By doing this, we accept that this access is
389 * inherently racy and know that the fs may change
390 * state before we even see this result.
391 */
Al Viro2d8f3032008-07-22 09:59:21 -0400392 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700393 res = -EROFS;
394
395out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400396 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500397 if (retry_estale(res, lookup_flags)) {
398 lookup_flags |= LOOKUP_REVAL;
399 goto retry;
400 }
Dave Hansen6902d922006-09-30 23:29:01 -0700401out:
David Howellsd84f4f92008-11-14 10:39:23 +1100402 revert_creds(old_cred);
403 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 return res;
405}
406
Heiko Carstensca013e92009-01-14 14:14:19 +0100407SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800408{
409 return sys_faccessat(AT_FDCWD, filename, mode);
410}
411
Heiko Carstens3cdad422009-01-14 14:14:22 +0100412SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413{
Al Viro2d8f3032008-07-22 09:59:21 -0400414 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500416 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
417retry:
418 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 if (error)
420 goto out;
421
Eric Paris9cfcac82010-07-23 11:43:51 -0400422 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 if (error)
424 goto dput_and_out;
425
Al Viro2d8f3032008-07-22 09:59:21 -0400426 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400429 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500430 if (retry_estale(error, lookup_flags)) {
431 lookup_flags |= LOOKUP_REVAL;
432 goto retry;
433 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434out:
435 return error;
436}
437
Heiko Carstens3cdad422009-01-14 14:14:22 +0100438SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439{
Al Viro2903ff02012-08-28 12:52:22 -0400440 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400442 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
444 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400445 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 goto out;
447
Al Viro496ad9a2013-01-23 17:07:38 -0500448 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450 error = -ENOTDIR;
451 if (!S_ISDIR(inode->i_mode))
452 goto out_putf;
453
Eric Paris9cfcac82010-07-23 11:43:51 -0400454 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400456 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400458 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459out:
460 return error;
461}
462
Heiko Carstens3480b252009-01-14 14:14:16 +0100463SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464{
Al Viro2d8f3032008-07-22 09:59:21 -0400465 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500467 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
468retry:
469 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 if (error)
471 goto out;
472
Eric Paris9cfcac82010-07-23 11:43:51 -0400473 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 if (error)
475 goto dput_and_out;
476
477 error = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700478 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900480 error = security_path_chroot(&path);
481 if (error)
482 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
Al Viro2d8f3032008-07-22 09:59:21 -0400484 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 error = 0;
486dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400487 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500488 if (retry_estale(error, lookup_flags)) {
489 lookup_flags |= LOOKUP_REVAL;
490 goto retry;
491 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492out:
493 return error;
494}
495
Al Viroe57712e2011-07-26 04:15:54 -0400496static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
Al Viroe57712e2011-07-26 04:15:54 -0400498 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400499 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400501 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
Al Viroe57712e2011-07-26 04:15:54 -0400503 error = mnt_want_write(path->mnt);
504 if (error)
505 return error;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400506retry_deleg:
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900507 mutex_lock(&inode->i_mutex);
Al Virocdcf1162011-12-08 10:51:53 -0500508 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400509 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900510 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
512 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400513 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900514out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800515 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400516 if (delegated_inode) {
517 error = break_deleg_wait(&delegated_inode);
518 if (!error)
519 goto retry_deleg;
520 }
Al Viroe57712e2011-07-26 04:15:54 -0400521 mnt_drop_write(path->mnt);
522 return error;
523}
524
Al Viro49f0a072011-07-26 04:22:01 -0400525SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400526{
Al Viro173c8402013-08-30 12:48:53 -0400527 struct fd f = fdget(fd);
Al Viroe57712e2011-07-26 04:15:54 -0400528 int err = -EBADF;
529
Al Viro173c8402013-08-30 12:48:53 -0400530 if (f.file) {
Al Viro9f45f5b2014-10-31 17:44:57 -0400531 audit_file(f.file);
Al Viro173c8402013-08-30 12:48:53 -0400532 err = chmod_common(&f.file->f_path, mode);
533 fdput(f);
Al Viroe57712e2011-07-26 04:15:54 -0400534 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 return err;
536}
537
Al Viro49f0a072011-07-26 04:22:01 -0400538SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539{
Al Viro2d8f3032008-07-22 09:59:21 -0400540 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 int error;
Jeff Layton14ff6902012-12-11 12:10:13 -0500542 unsigned int lookup_flags = LOOKUP_FOLLOW;
543retry:
544 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400545 if (!error) {
546 error = chmod_common(&path, mode);
547 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500548 if (retry_estale(error, lookup_flags)) {
549 lookup_flags |= LOOKUP_REVAL;
550 goto retry;
551 }
Al Viroe57712e2011-07-26 04:15:54 -0400552 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 return error;
554}
555
Al Viro49f0a072011-07-26 04:22:01 -0400556SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800557{
558 return sys_fchmodat(AT_FDCWD, filename, mode);
559}
560
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900561static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900563 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400564 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 int error;
566 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800567 kuid_t uid;
568 kgid_t gid;
569
570 uid = make_kuid(current_user_ns(), user);
571 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 newattrs.ia_valid = ATTR_CTIME;
574 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800575 if (!uid_valid(uid))
576 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800578 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 }
580 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800581 if (!gid_valid(gid))
582 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800584 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 }
586 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700587 newattrs.ia_valid |=
588 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400589retry_deleg:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800590 mutex_lock(&inode->i_mutex);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600591 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900592 if (!error)
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400593 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800594 mutex_unlock(&inode->i_mutex);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400595 if (delegated_inode) {
596 error = break_deleg_wait(&delegated_inode);
597 if (!error)
598 goto retry_deleg;
599 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 return error;
601}
602
Heiko Carstens6559eed82009-01-14 14:14:32 +0100603SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
604 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800605{
Al Viro2d8f3032008-07-22 09:59:21 -0400606 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800607 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400608 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800609
Al Viro65cfc672011-03-13 15:56:26 -0400610 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800611 goto out;
612
Al Viro65cfc672011-03-13 15:56:26 -0400613 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
614 if (flag & AT_EMPTY_PATH)
615 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500616retry:
Al Viro65cfc672011-03-13 15:56:26 -0400617 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700618 if (error)
619 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400620 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800621 if (error)
622 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900623 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400624 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800625out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400626 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500627 if (retry_estale(error, lookup_flags)) {
628 lookup_flags |= LOOKUP_REVAL;
629 goto retry;
630 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800631out:
632 return error;
633}
634
David Howells55e4def2012-06-25 12:55:09 +0100635SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
636{
637 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
638}
639
Heiko Carstensca013e92009-01-14 14:14:19 +0100640SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641{
David Howells55e4def2012-06-25 12:55:09 +0100642 return sys_fchownat(AT_FDCWD, filename, user, group,
643 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644}
645
Heiko Carstensca013e92009-01-14 14:14:19 +0100646SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647{
Al Viro2903ff02012-08-28 12:52:22 -0400648 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 int error = -EBADF;
650
Al Viro2903ff02012-08-28 12:52:22 -0400651 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700652 goto out;
653
Al Viro2903ff02012-08-28 12:52:22 -0400654 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800655 if (error)
656 goto out_fput;
Al Viro9f45f5b2014-10-31 17:44:57 -0400657 audit_file(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400658 error = chown_common(&f.file->f_path, user, group);
659 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800660out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400661 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700662out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 return error;
664}
665
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200666int open_check_o_direct(struct file *f)
667{
668 /* NB: we're sure to have correct a_ops only after f_op->open */
669 if (f->f_flags & O_DIRECT) {
670 if (!f->f_mapping->a_ops ||
671 ((!f->f_mapping->a_ops->direct_IO) &&
672 (!f->f_mapping->a_ops->get_xip_mem))) {
673 return -EINVAL;
674 }
675 }
676 return 0;
677}
678
Al Viro02e51802012-06-10 14:32:45 -0400679static int do_dentry_open(struct file *f,
Al Viro96b7e572012-06-10 14:22:04 -0400680 int (*open)(struct inode *, struct file *),
681 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682{
Al Viro1abf0c72011-03-13 03:51:11 -0400683 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 struct inode *inode;
685 int error;
686
Al Viro53009902009-12-19 10:15:07 -0500687 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700688 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400689
Al Virob5bcdda2012-07-20 23:28:46 +0400690 path_get(&f->f_path);
Al Virodd379782013-03-01 19:48:30 -0500691 inode = f->f_inode = f->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
Al Viro3f4d5a02014-03-14 09:43:29 -0400694 if (unlikely(f->f_flags & O_PATH)) {
695 f->f_mode = FMODE_PATH;
Al Viro1abf0c72011-03-13 03:51:11 -0400696 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400697 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400698 }
699
Al Virodd209082014-03-14 10:56:20 -0400700 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400701 error = get_write_access(inode);
Al Viro3f4d5a02014-03-14 09:43:29 -0400702 if (unlikely(error))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 goto cleanup_file;
Al Viro0ccb2862014-03-14 10:40:46 -0400704 error = __mnt_want_write(f->f_path.mnt);
Al Viro3f4d5a02014-03-14 09:43:29 -0400705 if (unlikely(error)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400706 put_write_access(inode);
707 goto cleanup_file;
708 }
Al Viro83f936c2014-03-14 12:02:47 -0400709 f->f_mode |= FMODE_WRITER;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 }
711
Linus Torvalds9c225f22014-03-03 09:36:58 -0800712 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
713 if (S_ISREG(inode->i_mode))
714 f->f_mode |= FMODE_ATOMIC_POS;
715
Al Viro1abf0c72011-03-13 03:51:11 -0400716 f->f_op = fops_get(inode->i_fop);
Al Viro72c2d532013-09-22 16:27:52 -0400717 if (unlikely(WARN_ON(!f->f_op))) {
718 error = -ENODEV;
719 goto cleanup_all;
720 }
Al Viro1abf0c72011-03-13 03:51:11 -0400721
Eric Paris83d49852012-04-04 13:45:40 -0400722 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900723 if (error)
724 goto cleanup_all;
725
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400726 error = break_lease(inode, f->f_flags);
727 if (error)
728 goto cleanup_all;
729
Al Viro72c2d532013-09-22 16:27:52 -0400730 if (!open)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700731 open = f->f_op->open;
732 if (open) {
733 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 if (error)
735 goto cleanup_all;
736 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400737 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
738 i_readcount_inc(inode);
Al Viro293bc982014-02-11 18:37:41 -0500739 if ((f->f_mode & FMODE_READ) &&
740 likely(f->f_op->read || f->f_op->aio_read || f->f_op->read_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500741 f->f_mode |= FMODE_CAN_READ;
Al Viro293bc982014-02-11 18:37:41 -0500742 if ((f->f_mode & FMODE_WRITE) &&
743 likely(f->f_op->write || f->f_op->aio_write || f->f_op->write_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500744 f->f_mode |= FMODE_CAN_WRITE;
Trond Myklebust834f2a42005-10-18 14:20:16 -0700745
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
747
748 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
749
Al Viro96b7e572012-06-10 14:22:04 -0400750 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
752cleanup_all:
753 fops_put(f->f_op);
Al Viro83f936c2014-03-14 12:02:47 -0400754 if (f->f_mode & FMODE_WRITER) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 put_write_access(inode);
Al Viro83f936c2014-03-14 12:02:47 -0400756 __mnt_drop_write(f->f_path.mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800757 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400759 path_put(&f->f_path);
760 f->f_path.mnt = NULL;
761 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500762 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400763 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764}
765
Trond Myklebust834f2a42005-10-18 14:20:16 -0700766/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200767 * finish_open - finish opening a file
Miklos Szeredi0854d452013-09-16 14:51:55 +0200768 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200769 * @dentry: pointer to dentry
770 * @open: open callback
Miklos Szeredi0854d452013-09-16 14:51:55 +0200771 * @opened: state of open
Miklos Szeredid18e9002012-06-05 15:10:17 +0200772 *
773 * This can be used to finish opening a file passed to i_op->atomic_open().
774 *
775 * If the open callback is set to NULL, then the standard f_op->open()
776 * filesystem callback is substituted.
Miklos Szeredi0854d452013-09-16 14:51:55 +0200777 *
778 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
779 * the return value of d_splice_alias(), then the caller needs to perform dput()
780 * on it after finish_open().
781 *
782 * On successful return @file is a fully instantiated open file. After this, if
783 * an error occurs in ->atomic_open(), it needs to clean up with fput().
784 *
785 * Returns zero on success or -errno if the open failed.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200786 */
Al Viro30d90492012-06-22 12:40:19 +0400787int finish_open(struct file *file, struct dentry *dentry,
788 int (*open)(struct inode *, struct file *),
789 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200790{
Al Viro96b7e572012-06-10 14:22:04 -0400791 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400792 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200793
Al Virob5bcdda2012-07-20 23:28:46 +0400794 file->f_path.dentry = dentry;
Al Viro02e51802012-06-10 14:32:45 -0400795 error = do_dentry_open(file, open, current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400796 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400797 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200798
Al Viro96b7e572012-06-10 14:22:04 -0400799 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200800}
801EXPORT_SYMBOL(finish_open);
802
803/**
804 * finish_no_open - finish ->atomic_open() without opening the file
805 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200806 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200807 * @dentry: dentry or NULL (as returned from ->lookup())
808 *
809 * This can be used to set the result of a successful lookup in ->atomic_open().
Miklos Szeredi0854d452013-09-16 14:51:55 +0200810 *
811 * NB: unlike finish_open() this function does consume the dentry reference and
812 * the caller need not dput() it.
813 *
814 * Returns "1" which must be the return value of ->atomic_open() after having
815 * called this function.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200816 */
Al Viroe45198a2012-06-10 06:48:09 -0400817int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200818{
Al Viro30d90492012-06-22 12:40:19 +0400819 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400820 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200821}
822EXPORT_SYMBOL(finish_no_open);
823
Al Viro765927b2012-06-26 21:58:53 +0400824struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100825 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700826{
827 int error;
828 struct file *f;
829
David Howellse0e81732009-09-02 09:13:40 +0100830 validate_creds(cred);
831
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900832 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400833 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800834
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700835 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500836 if (!IS_ERR(f)) {
837 f->f_flags = flags;
Miklos Szeredi4aa7c632014-10-24 00:14:35 +0200838 error = vfs_open(path, f, cred);
Al Viro1afc99b2013-02-14 20:41:04 -0500839 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
Miklos Szeredi4aa7c632014-10-24 00:14:35 +0200855/**
856 * vfs_open - open the file at the given path
857 * @path: path to open
858 * @filp: newly allocated file with f_flag initialized
859 * @cred: credentials to use
860 */
861int vfs_open(const struct path *path, struct file *filp,
862 const struct cred *cred)
863{
864 struct inode *inode = path->dentry->d_inode;
865
866 if (inode->i_op->dentry_open)
867 return inode->i_op->dentry_open(path->dentry, filp, cred);
868 else {
869 filp->f_path = *path;
870 return do_dentry_open(filp, NULL, cred);
871 }
872}
873EXPORT_SYMBOL(vfs_open);
874
Al Viroa218d0f2011-11-21 14:59:34 -0500875static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500876{
877 int lookup_flags = 0;
878 int acc_mode;
879
Andy Lutomirskie305f482013-08-01 21:07:52 -0700880 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200881 op->mode = (mode & S_IALLUGO) | S_IFREG;
882 else
883 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500884
885 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400886 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500887
888 /*
889 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
890 * check for O_DSYNC if the need any syncing at all we enforce it's
891 * always set instead of having to deal with possibly weird behaviour
892 * for malicious applications setting only __O_SYNC.
893 */
894 if (flags & __O_SYNC)
895 flags |= O_DSYNC;
896
Al Virobb458c62013-07-13 13:26:37 +0400897 if (flags & __O_TMPFILE) {
898 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400899 return -EINVAL;
900 acc_mode = MAY_OPEN | ACC_MODE(flags);
Al Viroba57ea62013-07-20 03:11:32 +0400901 if (!(acc_mode & MAY_WRITE))
902 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400903 } else if (flags & O_PATH) {
904 /*
905 * If we have O_PATH in the open flag. Then we
906 * cannot have anything other than the below set of flags
907 */
Al Viro1abf0c72011-03-13 03:51:11 -0400908 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
909 acc_mode = 0;
910 } else {
911 acc_mode = MAY_OPEN | ACC_MODE(flags);
912 }
Al Viro47c805d2011-02-23 17:44:09 -0500913
Al Viro1abf0c72011-03-13 03:51:11 -0400914 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500915
916 /* O_TRUNC implies we need access checks for write permissions */
917 if (flags & O_TRUNC)
918 acc_mode |= MAY_WRITE;
919
920 /* Allow the LSM permission hook to distinguish append
921 access from general write access. */
922 if (flags & O_APPEND)
923 acc_mode |= MAY_APPEND;
924
925 op->acc_mode = acc_mode;
926
Al Viro1abf0c72011-03-13 03:51:11 -0400927 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
928
Al Viro47c805d2011-02-23 17:44:09 -0500929 if (flags & O_CREAT) {
930 op->intent |= LOOKUP_CREATE;
931 if (flags & O_EXCL)
932 op->intent |= LOOKUP_EXCL;
933 }
934
935 if (flags & O_DIRECTORY)
936 lookup_flags |= LOOKUP_DIRECTORY;
937 if (!(flags & O_NOFOLLOW))
938 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +0400939 op->lookup_flags = lookup_flags;
940 return 0;
Al Viro47c805d2011-02-23 17:44:09 -0500941}
942
943/**
Jeff Layton669abf42012-10-10 16:43:10 -0400944 * file_open_name - open file and return file pointer
945 *
946 * @name: struct filename containing path to open
947 * @flags: open flags as per the open(2) second argument
948 * @mode: mode for the new file if O_CREAT is set, else ignored
949 *
950 * This is the helper to open a file from kernelspace if you really
951 * have to. But in generally you should not do this, so please move
952 * along, nothing to see here..
953 */
954struct file *file_open_name(struct filename *name, int flags, umode_t mode)
955{
956 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400957 int err = build_open_flags(flags, mode, &op);
958 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -0400959}
960
961/**
Al Viro47c805d2011-02-23 17:44:09 -0500962 * filp_open - open file and return file pointer
963 *
964 * @filename: path to open
965 * @flags: open flags as per the open(2) second argument
966 * @mode: mode for the new file if O_CREAT is set, else ignored
967 *
968 * This is the helper to open a file from kernelspace if you really
969 * have to. But in generally you should not do this, so please move
970 * along, nothing to see here..
971 */
Al Viroa218d0f2011-11-21 14:59:34 -0500972struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -0500973{
Paul Moore51689102015-01-22 00:00:03 -0500974 struct filename *name = getname_kernel(filename);
975 struct file *file = ERR_CAST(name);
976
977 if (!IS_ERR(name)) {
978 file = file_open_name(name, flags, mode);
979 putname(name);
980 }
981 return file;
Al Viro47c805d2011-02-23 17:44:09 -0500982}
983EXPORT_SYMBOL(filp_open);
984
Al Viro73d049a2011-03-11 12:08:24 -0500985struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
986 const char *filename, int flags)
987{
988 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +0400989 int err = build_open_flags(flags, 0, &op);
990 if (err)
991 return ERR_PTR(err);
Al Viro73d049a2011-03-11 12:08:24 -0500992 if (flags & O_CREAT)
993 return ERR_PTR(-EINVAL);
994 if (!filename && (flags & O_DIRECTORY))
995 if (!dentry->d_inode->i_op->lookup)
996 return ERR_PTR(-ENOTDIR);
Al Virof9652e12013-06-11 08:23:01 +0400997 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -0500998}
999EXPORT_SYMBOL(file_open_root);
1000
Al Viroa218d0f2011-11-21 14:59:34 -05001001long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002{
Al Viro47c805d2011-02-23 17:44:09 -05001003 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001004 int fd = build_open_flags(flags, mode, &op);
1005 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Al Virof9652e12013-06-11 08:23:01 +04001007 if (fd)
1008 return fd;
1009
1010 tmp = getname(filename);
1011 if (IS_ERR(tmp))
1012 return PTR_ERR(tmp);
1013
1014 fd = get_unused_fd_flags(flags);
1015 if (fd >= 0) {
1016 struct file *f = do_filp_open(dfd, tmp, &op);
1017 if (IS_ERR(f)) {
1018 put_unused_fd(fd);
1019 fd = PTR_ERR(f);
1020 } else {
1021 fsnotify_open(f);
1022 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 }
Al Virof9652e12013-06-11 08:23:01 +04001025 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001028
Al Viroa218d0f2011-11-21 14:59:34 -05001029SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001030{
1031 if (force_o_largefile())
1032 flags |= O_LARGEFILE;
1033
Al Viro2cf09662013-01-21 15:25:54 -05001034 return do_sys_open(AT_FDCWD, filename, flags, mode);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001035}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
Heiko Carstens6559eed82009-01-14 14:14:32 +01001037SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001038 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001039{
1040 if (force_o_largefile())
1041 flags |= O_LARGEFILE;
1042
Al Viro2cf09662013-01-21 15:25:54 -05001043 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001044}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046#ifndef __alpha__
1047
1048/*
1049 * For backward compatibility? Maybe this should be moved
1050 * into arch/i386 instead?
1051 */
Al Viroa218d0f2011-11-21 14:59:34 -05001052SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
1054 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1055}
1056
1057#endif
1058
1059/*
1060 * "id" is the POSIX thread ID. We use the
1061 * files pointer for this..
1062 */
1063int filp_close(struct file *filp, fl_owner_t id)
1064{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001065 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067 if (!file_count(filp)) {
1068 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001069 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 }
1071
Al Viro72c2d532013-09-22 16:27:52 -04001072 if (filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001073 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074
Al Viro1abf0c72011-03-13 03:51:11 -04001075 if (likely(!(filp->f_mode & FMODE_PATH))) {
1076 dnotify_flush(filp, id);
1077 locks_remove_posix(filp, id);
1078 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 fput(filp);
1080 return retval;
1081}
1082
1083EXPORT_SYMBOL(filp_close);
1084
1085/*
1086 * Careful here! We test whether the file pointer is NULL before
1087 * releasing the fd. This ensures that one clone task can't release
1088 * an fd while another clone is opening it.
1089 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001090SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091{
Al Viro483ce1d2012-08-19 12:04:24 -04001092 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001093
1094 /* can't restart close syscall because file table entry was cleared */
1095 if (unlikely(retval == -ERESTARTSYS ||
1096 retval == -ERESTARTNOINTR ||
1097 retval == -ERESTARTNOHAND ||
1098 retval == -ERESTART_RESTARTBLOCK))
1099 retval = -EINTR;
1100
1101 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103EXPORT_SYMBOL(sys_close);
1104
1105/*
1106 * This routine simulates a hangup on the tty, to arrange that users
1107 * are given clean terminals at login time.
1108 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001109SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110{
1111 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001112 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 return 0;
1114 }
1115 return -EPERM;
1116}
1117
1118/*
1119 * Called when an inode is about to be open.
1120 * We use this to disallow opening large files on 32bit systems if
1121 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1122 * on this flag in sys_open.
1123 */
1124int generic_file_open(struct inode * inode, struct file * filp)
1125{
1126 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001127 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 return 0;
1129}
1130
1131EXPORT_SYMBOL(generic_file_open);
1132
1133/*
1134 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001135 * file descriptors. The function is not supposed to ever fail, the only
1136 * reason it returns an 'int' and not 'void' is so that it can be plugged
1137 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 */
1139int nonseekable_open(struct inode *inode, struct file *filp)
1140{
1141 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1142 return 0;
1143}
1144
1145EXPORT_SYMBOL(nonseekable_open);