blob: 2ff88766123750a3ae2cee33ddeea13b9323067a [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
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -070037int do_truncate2(struct vfsmount *mnt, struct dentry *dentry, loff_t length,
38 unsigned int time_attrs, 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
Al Viro59551022016-01-22 15:40:57 -050061 inode_lock(dentry->d_inode);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -040062 /* Note any delegations or leases have already been broken: */
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -070063 ret = notify_change2(mnt, dentry, &newattrs, NULL);
Al Viro59551022016-01-22 15:40:57 -050064 inode_unlock(dentry->d_inode);
Amerigo Wang939a9422009-08-20 19:29:03 -070065 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066}
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -070067int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
68 struct file *filp)
69{
70 return do_truncate2(NULL, dentry, length, time_attrs, filp);
71}
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Al Viro7df818b2016-03-25 14:24:09 -040073long vfs_truncate(const struct path *path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070074{
Al Viro2d8f3032008-07-22 09:59:21 -040075 struct inode *inode;
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -070076 struct vfsmount *mnt;
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +020077 struct dentry *upperdentry;
David Howellsa02de962012-12-20 21:52:36 +000078 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079
David Howellsa02de962012-12-20 21:52:36 +000080 inode = path->dentry->d_inode;
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -070081 mnt = path->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
83 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 if (S_ISDIR(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000085 return -EISDIR;
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 if (!S_ISREG(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000087 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
David Howellsa02de962012-12-20 21:52:36 +000089 error = mnt_want_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 if (error)
David Howellsa02de962012-12-20 21:52:36 +000091 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -070093 error = inode_permission2(mnt, inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080094 if (error)
95 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020098 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080099 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +0200101 /*
102 * If this is an overlayfs then do as if opening the file so we get
103 * write access on the upper inode, not on the overlay inode. For
104 * non-overlay filesystems d_real() is an identity function.
105 */
106 upperdentry = d_real(path->dentry, NULL, O_WRONLY);
107 error = PTR_ERR(upperdentry);
108 if (IS_ERR(upperdentry))
109 goto mnt_drop_write_and_out;
110
111 error = get_write_access(upperdentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800113 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
david m. richter97003822007-07-31 00:39:12 -0700115 /*
116 * Make sure that there are no leases. get_write_access() protects
117 * against the truncate racing with a lease-granting setlease().
118 */
Al Viro8737c932009-12-24 06:47:55 -0500119 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700120 if (error)
121 goto put_write_and_out;
122
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900124 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000125 error = security_path_truncate(path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500126 if (!error)
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700127 error = do_truncate2(mnt, path->dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
david m. richter97003822007-07-31 00:39:12 -0700129put_write_and_out:
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +0200130 put_write_access(upperdentry->d_inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800131mnt_drop_write_and_out:
David Howellsa02de962012-12-20 21:52:36 +0000132 mnt_drop_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133out:
134 return error;
135}
David Howellsa02de962012-12-20 21:52:36 +0000136EXPORT_SYMBOL_GPL(vfs_truncate);
137
138static long do_sys_truncate(const char __user *pathname, loff_t length)
139{
Jeff Layton48f75302012-12-11 12:10:11 -0500140 unsigned int lookup_flags = LOOKUP_FOLLOW;
David Howellsa02de962012-12-20 21:52:36 +0000141 struct path path;
142 int error;
143
144 if (length < 0) /* sorry, but loff_t says... */
145 return -EINVAL;
146
Jeff Layton48f75302012-12-11 12:10:11 -0500147retry:
148 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
David Howellsa02de962012-12-20 21:52:36 +0000149 if (!error) {
150 error = vfs_truncate(&path, length);
151 path_put(&path);
152 }
Jeff Layton48f75302012-12-11 12:10:11 -0500153 if (retry_estale(error, lookup_flags)) {
154 lookup_flags |= LOOKUP_REVAL;
155 goto retry;
156 }
David Howellsa02de962012-12-20 21:52:36 +0000157 return error;
158}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200160SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200162 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163}
164
Al Viro3f6d0782013-02-24 13:49:08 -0500165#ifdef CONFIG_COMPAT
166COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
167{
168 return do_sys_truncate(path, length);
169}
170#endif
171
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800172static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173{
Al Virobf2965d2012-08-26 20:13:36 -0400174 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 struct dentry *dentry;
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700176 struct vfsmount *mnt;
Al Viro2903ff02012-08-28 12:52:22 -0400177 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 int error;
179
180 error = -EINVAL;
181 if (length < 0)
182 goto out;
183 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400184 f = fdget(fd);
185 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 goto out;
187
188 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400189 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 small = 0;
191
Al Viro2903ff02012-08-28 12:52:22 -0400192 dentry = f.file->f_path.dentry;
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700193 mnt = f.file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 inode = dentry->d_inode;
195 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400196 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 goto out_putf;
198
199 error = -EINVAL;
200 /* Cannot ftruncate over 2^31 bytes without large file support */
201 if (small && length > MAX_NON_LFS)
202 goto out_putf;
203
204 error = -EPERM;
205 if (IS_APPEND(inode))
206 goto out_putf;
207
Jan Kara14da9202012-06-12 16:20:37 +0200208 sb_start_write(inode->i_sb);
Al Viro2903ff02012-08-28 12:52:22 -0400209 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400211 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900212 if (!error)
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700213 error = do_truncate2(mnt, dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200214 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400216 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217out:
218 return error;
219}
220
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100221SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
Al Viro2cf09662013-01-21 15:25:54 -0500223 return do_sys_ftruncate(fd, length, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224}
225
Al Viro3f6d0782013-02-24 13:49:08 -0500226#ifdef CONFIG_COMPAT
227COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
228{
229 return do_sys_ftruncate(fd, length, 1);
230}
231#endif
232
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233/* LFS versions of truncate are only needed on 32 bit machines */
234#if BITS_PER_LONG == 32
Al Viro4a0fd5b2013-01-21 15:16:58 -0500235SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236{
237 return do_sys_truncate(path, length);
238}
239
Al Viro4a0fd5b2013-01-21 15:16:58 -0500240SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
Al Viro2cf09662013-01-21 15:25:54 -0500242 return do_sys_ftruncate(fd, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100244#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400246
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500247int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400248{
Al Viro496ad9a2013-01-23 17:07:38 -0500249 struct inode *inode = file_inode(file);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400250 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400251
252 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400253 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400254
255 /* Return error if mode is not supported */
Namjae Jeondd46c782015-03-25 15:07:05 +1100256 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
Lukas Czerner409332b2014-03-13 19:07:42 +1100257 return -EOPNOTSUPP;
258
259 /* Punch hole and zero range are mutually exclusive */
260 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
261 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
Josef Bacik79124f182010-11-17 20:46:15 -0500262 return -EOPNOTSUPP;
263
264 /* Punch hole must have keep size set */
265 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
266 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400267 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400268
Namjae Jeon00f5e612014-02-24 10:58:15 +1100269 /* Collapse range should only be used exclusively. */
270 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
271 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
272 return -EINVAL;
273
Namjae Jeondd46c782015-03-25 15:07:05 +1100274 /* Insert range should only be used exclusively. */
275 if ((mode & FALLOC_FL_INSERT_RANGE) &&
276 (mode & ~FALLOC_FL_INSERT_RANGE))
277 return -EINVAL;
278
Darrick J. Wong71be6b42016-10-03 09:11:14 -0700279 /* Unshare range should only be used with allocate mode. */
280 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
281 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
282 return -EINVAL;
283
Amit Arora97ac7352007-07-17 21:42:44 -0400284 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400285 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100286
Namjae Jeon00f5e612014-02-24 10:58:15 +1100287 /*
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400288 * We can only allow pure fallocate on append only files
Namjae Jeon00f5e612014-02-24 10:58:15 +1100289 */
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400290 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100291 return -EPERM;
292
293 if (IS_IMMUTABLE(inode))
294 return -EPERM;
295
Amit Arora97ac7352007-07-17 21:42:44 -0400296 /*
Eric Biggers6d2b6172014-06-24 23:45:08 -0500297 * We cannot allow any fallocate operation on an active swapfile
Lukas Czerner0790b312014-04-12 10:05:37 -0400298 */
299 if (IS_SWAPFILE(inode))
Eric Biggers6d2b6172014-06-24 23:45:08 -0500300 return -ETXTBSY;
Lukas Czerner0790b312014-04-12 10:05:37 -0400301
302 /*
Amit Arora97ac7352007-07-17 21:42:44 -0400303 * Revalidate the write permissions, in case security policy has
304 * changed since the files were opened.
305 */
306 ret = security_file_permission(file, MAY_WRITE);
307 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400308 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400309
Amit Arora97ac7352007-07-17 21:42:44 -0400310 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400311 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400312
Amit Arora97ac7352007-07-17 21:42:44 -0400313 /*
314 * Let individual file system decide if it supports preallocation
315 * for directories or not.
316 */
Darrick J. Wong25f4c412016-10-11 13:51:11 -0700317 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) &&
318 !S_ISBLK(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400319 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400320
Amit Arora97ac7352007-07-17 21:42:44 -0400321 /* Check for wrap through zero too */
322 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400323 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400324
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100325 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400326 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400327
Jan Kara14da9202012-06-12 16:20:37 +0200328 sb_start_write(inode->i_sb);
329 ret = file->f_op->fallocate(file, mode, offset, len);
Heinrich Schuchardt820c12d2014-12-12 16:58:34 -0800330
331 /*
332 * Create inotify and fanotify events.
333 *
334 * To keep the logic simple always create events if fallocate succeeds.
335 * This implies that events are even created if the file size remains
336 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
337 */
338 if (ret == 0)
339 fsnotify_modify(file);
340
Jan Kara14da9202012-06-12 16:20:37 +0200341 sb_end_write(inode->i_sb);
342 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400343}
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500344EXPORT_SYMBOL_GPL(vfs_fallocate);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400345
Al Viro4a0fd5b2013-01-21 15:16:58 -0500346SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400347{
Al Viro2903ff02012-08-28 12:52:22 -0400348 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400349 int error = -EBADF;
350
Al Viro2903ff02012-08-28 12:52:22 -0400351 if (f.file) {
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500352 error = vfs_fallocate(f.file, mode, offset, len);
Al Viro2903ff02012-08-28 12:52:22 -0400353 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400354 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400355 return error;
356}
357
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358/*
359 * access() needs to use the real uid/gid, not the effective uid/gid.
360 * We do this by temporarily clearing all FS-related capabilities and
361 * switching the fsuid/fsgid around to the real ones.
362 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100363SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364{
David Howellsd84f4f92008-11-14 10:39:23 +1100365 const struct cred *old_cred;
366 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400367 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400368 struct inode *inode;
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700369 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 int res;
Jeff Layton87fa5592012-12-11 12:10:11 -0500371 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
374 return -EINVAL;
375
David Howellsd84f4f92008-11-14 10:39:23 +1100376 override_cred = prepare_creds();
377 if (!override_cred)
378 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379
David Howellsd84f4f92008-11-14 10:39:23 +1100380 override_cred->fsuid = override_cred->uid;
381 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700383 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100384 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800385 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
386 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100387 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700388 else
David Howellsd84f4f92008-11-14 10:39:23 +1100389 override_cred->cap_effective =
390 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700391 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
Linus Torvalds50810012019-07-11 09:54:40 -0700393 /*
394 * The new set of credentials can *only* be used in
395 * task-synchronous circumstances, and does not need
396 * RCU freeing, unless somebody then takes a separate
397 * reference to it.
398 *
399 * NOTE! This is _only_ true because this credential
400 * is used purely for override_creds() that installs
401 * it as the subjective cred. Other threads will be
402 * accessing ->real_cred, not the subjective cred.
403 *
404 * If somebody _does_ make a copy of this (using the
405 * 'get_current_cred()' function), that will clear the
406 * non_rcu field, because now that other user may be
407 * expecting RCU freeing. But normal thread-synchronous
408 * cred accesses will keep things non-RCY.
409 */
410 override_cred->non_rcu = 1;
411
David Howellsd84f4f92008-11-14 10:39:23 +1100412 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500413retry:
414 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700415 if (res)
416 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
David Howells63afdfc2015-05-06 15:59:00 +0100418 inode = d_backing_inode(path.dentry);
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700419 mnt = path.mnt;
Al Viro256984a2008-07-22 08:09:30 -0400420
421 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400422 /*
423 * MAY_EXEC on regular files is denied if the fs is mounted
424 * with the "noexec" flag.
425 */
426 res = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500427 if (path_noexec(&path))
Al Viro30524472008-07-22 00:02:33 -0400428 goto out_path_release;
429 }
430
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700431 res = inode_permission2(mnt, inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700432 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400433 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700434 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800435 /*
436 * This is a rare case where using __mnt_is_readonly()
437 * is OK without a mnt_want/drop_write() pair. Since
438 * no actual write to the fs is performed here, we do
439 * not need to telegraph to that to anyone.
440 *
441 * By doing this, we accept that this access is
442 * inherently racy and know that the fs may change
443 * state before we even see this result.
444 */
Al Viro2d8f3032008-07-22 09:59:21 -0400445 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700446 res = -EROFS;
447
448out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400449 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500450 if (retry_estale(res, lookup_flags)) {
451 lookup_flags |= LOOKUP_REVAL;
452 goto retry;
453 }
Dave Hansen6902d922006-09-30 23:29:01 -0700454out:
David Howellsd84f4f92008-11-14 10:39:23 +1100455 revert_creds(old_cred);
456 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 return res;
458}
459
Heiko Carstensca013e92009-01-14 14:14:19 +0100460SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800461{
462 return sys_faccessat(AT_FDCWD, filename, mode);
463}
464
Heiko Carstens3cdad422009-01-14 14:14:22 +0100465SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
Al Viro2d8f3032008-07-22 09:59:21 -0400467 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500469 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
470retry:
471 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 if (error)
473 goto out;
474
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700475 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 if (error)
477 goto dput_and_out;
478
Al Viro2d8f3032008-07-22 09:59:21 -0400479 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
481dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400482 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500483 if (retry_estale(error, lookup_flags)) {
484 lookup_flags |= LOOKUP_REVAL;
485 goto retry;
486 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487out:
488 return error;
489}
490
Heiko Carstens3cdad422009-01-14 14:14:22 +0100491SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
Al Viro2903ff02012-08-28 12:52:22 -0400493 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 struct inode *inode;
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700495 struct vfsmount *mnt;
Al Viro2903ff02012-08-28 12:52:22 -0400496 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
498 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400499 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 goto out;
501
Al Viro496ad9a2013-01-23 17:07:38 -0500502 inode = file_inode(f.file);
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700503 mnt = f.file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
505 error = -ENOTDIR;
506 if (!S_ISDIR(inode->i_mode))
507 goto out_putf;
508
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700509 error = inode_permission2(mnt, inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400511 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400513 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514out:
515 return error;
516}
517
Heiko Carstens3480b252009-01-14 14:14:16 +0100518SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
Al Viro2d8f3032008-07-22 09:59:21 -0400520 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500522 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
523retry:
524 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (error)
526 goto out;
527
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700528 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 if (error)
530 goto dput_and_out;
531
532 error = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700533 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900535 error = security_path_chroot(&path);
536 if (error)
537 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Al Viro2d8f3032008-07-22 09:59:21 -0400539 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 error = 0;
541dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400542 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500543 if (retry_estale(error, lookup_flags)) {
544 lookup_flags |= LOOKUP_REVAL;
545 goto retry;
546 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547out:
548 return error;
549}
550
Al Virobe01f9f2016-03-25 14:56:23 -0400551static int chmod_common(const struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
Al Viroe57712e2011-07-26 04:15:54 -0400553 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400554 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400556 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557
Al Viroe57712e2011-07-26 04:15:54 -0400558 error = mnt_want_write(path->mnt);
559 if (error)
560 return error;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400561retry_deleg:
Al Viro59551022016-01-22 15:40:57 -0500562 inode_lock(inode);
Al Virocdcf1162011-12-08 10:51:53 -0500563 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400564 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900565 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
567 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700568 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900569out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500570 inode_unlock(inode);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400571 if (delegated_inode) {
572 error = break_deleg_wait(&delegated_inode);
573 if (!error)
574 goto retry_deleg;
575 }
Al Viroe57712e2011-07-26 04:15:54 -0400576 mnt_drop_write(path->mnt);
577 return error;
578}
579
Al Viro49f0a072011-07-26 04:22:01 -0400580SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400581{
Al Viro173c8402013-08-30 12:48:53 -0400582 struct fd f = fdget(fd);
Al Viroe57712e2011-07-26 04:15:54 -0400583 int err = -EBADF;
584
Al Viro173c8402013-08-30 12:48:53 -0400585 if (f.file) {
Al Viro9f45f5b2014-10-31 17:44:57 -0400586 audit_file(f.file);
Al Viro173c8402013-08-30 12:48:53 -0400587 err = chmod_common(&f.file->f_path, mode);
588 fdput(f);
Al Viroe57712e2011-07-26 04:15:54 -0400589 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 return err;
591}
592
Al Viro49f0a072011-07-26 04:22:01 -0400593SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594{
Al Viro2d8f3032008-07-22 09:59:21 -0400595 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 int error;
Jeff Layton14ff6902012-12-11 12:10:13 -0500597 unsigned int lookup_flags = LOOKUP_FOLLOW;
598retry:
599 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400600 if (!error) {
601 error = chmod_common(&path, mode);
602 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500603 if (retry_estale(error, lookup_flags)) {
604 lookup_flags |= LOOKUP_REVAL;
605 goto retry;
606 }
Al Viroe57712e2011-07-26 04:15:54 -0400607 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 return error;
609}
610
Al Viro49f0a072011-07-26 04:22:01 -0400611SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800612{
613 return sys_fchmodat(AT_FDCWD, filename, mode);
614}
615
Al Viro7fd25da2016-03-25 14:44:41 -0400616static int chown_common(const struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900618 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400619 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 int error;
621 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800622 kuid_t uid;
623 kgid_t gid;
624
625 uid = make_kuid(current_user_ns(), user);
626 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Andrew Elblec1b89402015-02-23 08:51:24 -0500628retry_deleg:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 newattrs.ia_valid = ATTR_CTIME;
630 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800631 if (!uid_valid(uid))
632 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800634 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 }
636 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800637 if (!gid_valid(gid))
638 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800640 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 }
642 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700643 newattrs.ia_valid |=
644 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Al Viro59551022016-01-22 15:40:57 -0500645 inode_lock(inode);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600646 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900647 if (!error)
Daniel Rosenberg2757e9b2016-10-26 16:33:11 -0700648 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
Al Viro59551022016-01-22 15:40:57 -0500649 inode_unlock(inode);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400650 if (delegated_inode) {
651 error = break_deleg_wait(&delegated_inode);
652 if (!error)
653 goto retry_deleg;
654 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 return error;
656}
657
Heiko Carstens6559eed82009-01-14 14:14:32 +0100658SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
659 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800660{
Al Viro2d8f3032008-07-22 09:59:21 -0400661 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800662 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400663 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800664
Al Viro65cfc672011-03-13 15:56:26 -0400665 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800666 goto out;
667
Al Viro65cfc672011-03-13 15:56:26 -0400668 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
669 if (flag & AT_EMPTY_PATH)
670 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500671retry:
Al Viro65cfc672011-03-13 15:56:26 -0400672 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700673 if (error)
674 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400675 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800676 if (error)
677 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900678 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400679 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800680out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400681 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500682 if (retry_estale(error, lookup_flags)) {
683 lookup_flags |= LOOKUP_REVAL;
684 goto retry;
685 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800686out:
687 return error;
688}
689
David Howells55e4def2012-06-25 12:55:09 +0100690SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
691{
692 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
693}
694
Heiko Carstensca013e92009-01-14 14:14:19 +0100695SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696{
David Howells55e4def2012-06-25 12:55:09 +0100697 return sys_fchownat(AT_FDCWD, filename, user, group,
698 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699}
700
Heiko Carstensca013e92009-01-14 14:14:19 +0100701SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702{
Al Viro2903ff02012-08-28 12:52:22 -0400703 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 int error = -EBADF;
705
Al Viro2903ff02012-08-28 12:52:22 -0400706 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700707 goto out;
708
Al Viro2903ff02012-08-28 12:52:22 -0400709 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800710 if (error)
711 goto out_fput;
Al Viro9f45f5b2014-10-31 17:44:57 -0400712 audit_file(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400713 error = chown_common(&f.file->f_path, user, group);
714 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800715out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400716 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700717out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 return error;
719}
720
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200721int open_check_o_direct(struct file *f)
722{
723 /* NB: we're sure to have correct a_ops only after f_op->open */
724 if (f->f_flags & O_DIRECT) {
Matthew Wilcoxe748dcd2015-02-16 15:59:12 -0800725 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200726 return -EINVAL;
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200727 }
728 return 0;
729}
730
Al Viro02e51802012-06-10 14:32:45 -0400731static int do_dentry_open(struct file *f,
David Howells4bacc9c2015-06-18 14:32:31 +0100732 struct inode *inode,
Al Viro96b7e572012-06-10 14:22:04 -0400733 int (*open)(struct inode *, struct file *),
734 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735{
Al Viro1abf0c72011-03-13 03:51:11 -0400736 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 int error;
738
Al Viro53009902009-12-19 10:15:07 -0500739 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700740 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400741
Al Virob5bcdda2012-07-20 23:28:46 +0400742 path_get(&f->f_path);
David Howells4bacc9c2015-06-18 14:32:31 +0100743 f->f_inode = inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
Al Viro3f4d5a02014-03-14 09:43:29 -0400746 if (unlikely(f->f_flags & O_PATH)) {
747 f->f_mode = FMODE_PATH;
Al Viro1abf0c72011-03-13 03:51:11 -0400748 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400749 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400750 }
751
Tetsuo Handa7d7637d2019-03-28 20:43:30 -0700752 /* Any file opened for execve()/uselib() has to be a regular file. */
753 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
754 error = -EACCES;
755 goto cleanup_file;
756 }
757
Al Virodd209082014-03-14 10:56:20 -0400758 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400759 error = get_write_access(inode);
Al Viro3f4d5a02014-03-14 09:43:29 -0400760 if (unlikely(error))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 goto cleanup_file;
Al Viro0ccb2862014-03-14 10:40:46 -0400762 error = __mnt_want_write(f->f_path.mnt);
Al Viro3f4d5a02014-03-14 09:43:29 -0400763 if (unlikely(error)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400764 put_write_access(inode);
765 goto cleanup_file;
766 }
Al Viro83f936c2014-03-14 12:02:47 -0400767 f->f_mode |= FMODE_WRITER;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 }
769
Linus Torvalds9c225f22014-03-03 09:36:58 -0800770 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
Al Viro63b6df12016-04-20 17:08:21 -0400771 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
Linus Torvalds9c225f22014-03-03 09:36:58 -0800772 f->f_mode |= FMODE_ATOMIC_POS;
773
Al Viro1abf0c72011-03-13 03:51:11 -0400774 f->f_op = fops_get(inode->i_fop);
Al Viro72c2d532013-09-22 16:27:52 -0400775 if (unlikely(WARN_ON(!f->f_op))) {
776 error = -ENODEV;
777 goto cleanup_all;
778 }
Al Viro1abf0c72011-03-13 03:51:11 -0400779
Eric Paris83d49852012-04-04 13:45:40 -0400780 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900781 if (error)
782 goto cleanup_all;
783
Miklos Szeredic568d682016-09-16 12:44:20 +0200784 error = break_lease(locks_inode(f), f->f_flags);
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400785 if (error)
786 goto cleanup_all;
787
Al Viro72c2d532013-09-22 16:27:52 -0400788 if (!open)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700789 open = f->f_op->open;
790 if (open) {
791 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 if (error)
793 goto cleanup_all;
794 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400795 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
796 i_readcount_inc(inode);
Al Viro293bc982014-02-11 18:37:41 -0500797 if ((f->f_mode & FMODE_READ) &&
Al Viro84363182015-04-04 01:14:53 -0400798 likely(f->f_op->read || f->f_op->read_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500799 f->f_mode |= FMODE_CAN_READ;
Al Viro293bc982014-02-11 18:37:41 -0500800 if ((f->f_mode & FMODE_WRITE) &&
Al Viro84363182015-04-04 01:14:53 -0400801 likely(f->f_op->write || f->f_op->write_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500802 f->f_mode |= FMODE_CAN_WRITE;
Trond Myklebust834f2a42005-10-18 14:20:16 -0700803
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
805
806 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
807
Al Viro96b7e572012-06-10 14:22:04 -0400808 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810cleanup_all:
811 fops_put(f->f_op);
Al Viro83f936c2014-03-14 12:02:47 -0400812 if (f->f_mode & FMODE_WRITER) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 put_write_access(inode);
Al Viro83f936c2014-03-14 12:02:47 -0400814 __mnt_drop_write(f->f_path.mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800815 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400817 path_put(&f->f_path);
818 f->f_path.mnt = NULL;
819 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500820 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400821 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822}
823
Trond Myklebust834f2a42005-10-18 14:20:16 -0700824/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200825 * finish_open - finish opening a file
Miklos Szeredi0854d452013-09-16 14:51:55 +0200826 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200827 * @dentry: pointer to dentry
828 * @open: open callback
Miklos Szeredi0854d452013-09-16 14:51:55 +0200829 * @opened: state of open
Miklos Szeredid18e9002012-06-05 15:10:17 +0200830 *
831 * This can be used to finish opening a file passed to i_op->atomic_open().
832 *
833 * If the open callback is set to NULL, then the standard f_op->open()
834 * filesystem callback is substituted.
Miklos Szeredi0854d452013-09-16 14:51:55 +0200835 *
836 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
837 * the return value of d_splice_alias(), then the caller needs to perform dput()
838 * on it after finish_open().
839 *
840 * On successful return @file is a fully instantiated open file. After this, if
841 * an error occurs in ->atomic_open(), it needs to clean up with fput().
842 *
843 * Returns zero on success or -errno if the open failed.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200844 */
Al Viro30d90492012-06-22 12:40:19 +0400845int finish_open(struct file *file, struct dentry *dentry,
846 int (*open)(struct inode *, struct file *),
847 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200848{
Al Viro96b7e572012-06-10 14:22:04 -0400849 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400850 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200851
Al Virob5bcdda2012-07-20 23:28:46 +0400852 file->f_path.dentry = dentry;
David Howells4bacc9c2015-06-18 14:32:31 +0100853 error = do_dentry_open(file, d_backing_inode(dentry), open,
854 current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400855 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400856 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200857
Al Viro96b7e572012-06-10 14:22:04 -0400858 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200859}
860EXPORT_SYMBOL(finish_open);
861
862/**
863 * finish_no_open - finish ->atomic_open() without opening the file
864 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200865 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200866 * @dentry: dentry or NULL (as returned from ->lookup())
867 *
868 * This can be used to set the result of a successful lookup in ->atomic_open().
Miklos Szeredi0854d452013-09-16 14:51:55 +0200869 *
870 * NB: unlike finish_open() this function does consume the dentry reference and
871 * the caller need not dput() it.
872 *
873 * Returns "1" which must be the return value of ->atomic_open() after having
874 * called this function.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200875 */
Al Viroe45198a2012-06-10 06:48:09 -0400876int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200877{
Al Viro30d90492012-06-22 12:40:19 +0400878 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400879 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200880}
881EXPORT_SYMBOL(finish_no_open);
882
Miklos Szeredi9bf39ab2015-06-19 10:29:13 +0200883char *file_path(struct file *filp, char *buf, int buflen)
884{
885 return d_path(&filp->f_path, buf, buflen);
886}
887EXPORT_SYMBOL(file_path);
888
David Howells4bacc9c2015-06-18 14:32:31 +0100889/**
890 * vfs_open - open the file at the given path
891 * @path: path to open
892 * @file: newly allocated file with f_flag initialized
893 * @cred: credentials to use
894 */
895int vfs_open(const struct path *path, struct file *file,
896 const struct cred *cred)
897{
Miklos Szeredi2d902672016-06-30 08:53:27 +0200898 struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags);
Miklos Szeredi54d5ca82016-05-11 01:16:37 +0200899
Miklos Szeredi2d902672016-06-30 08:53:27 +0200900 if (IS_ERR(dentry))
901 return PTR_ERR(dentry);
David Howells4bacc9c2015-06-18 14:32:31 +0100902
903 file->f_path = *path;
Miklos Szeredi2d902672016-06-30 08:53:27 +0200904 return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
David Howells4bacc9c2015-06-18 14:32:31 +0100905}
906
Al Viro765927b2012-06-26 21:58:53 +0400907struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100908 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700909{
910 int error;
911 struct file *f;
912
David Howellse0e81732009-09-02 09:13:40 +0100913 validate_creds(cred);
914
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900915 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400916 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800917
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700918 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500919 if (!IS_ERR(f)) {
920 f->f_flags = flags;
Miklos Szeredi4aa7c632014-10-24 00:14:35 +0200921 error = vfs_open(path, f, cred);
Al Viro1afc99b2013-02-14 20:41:04 -0500922 if (!error) {
923 /* from now on we need fput() to dispose of f */
924 error = open_check_o_direct(f);
925 if (error) {
926 fput(f);
927 f = ERR_PTR(error);
928 }
929 } else {
930 put_filp(f);
Al Viro2a027e72012-06-10 14:24:38 -0400931 f = ERR_PTR(error);
932 }
Al Viro2a027e72012-06-10 14:24:38 -0400933 }
934 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700935}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936EXPORT_SYMBOL(dentry_open);
937
Al Viroa218d0f2011-11-21 14:59:34 -0500938static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500939{
940 int lookup_flags = 0;
Al Viro62fb4a12015-12-26 22:33:24 -0500941 int acc_mode = ACC_MODE(flags);
Al Viro47c805d2011-02-23 17:44:09 -0500942
Christoph Hellwig26ff0652017-04-27 09:42:25 +0200943 /*
944 * Clear out all open flags we don't know about so that we don't report
945 * them in fcntl(F_GETFD) or similar interfaces.
946 */
947 flags &= VALID_OPEN_FLAGS;
948
Andy Lutomirskie305f482013-08-01 21:07:52 -0700949 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200950 op->mode = (mode & S_IALLUGO) | S_IFREG;
951 else
952 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500953
954 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400955 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500956
957 /*
958 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
959 * check for O_DSYNC if the need any syncing at all we enforce it's
960 * always set instead of having to deal with possibly weird behaviour
961 * for malicious applications setting only __O_SYNC.
962 */
963 if (flags & __O_SYNC)
964 flags |= O_DSYNC;
965
Al Virobb458c62013-07-13 13:26:37 +0400966 if (flags & __O_TMPFILE) {
967 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400968 return -EINVAL;
Al Viroba57ea62013-07-20 03:11:32 +0400969 if (!(acc_mode & MAY_WRITE))
970 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400971 } else if (flags & O_PATH) {
972 /*
973 * If we have O_PATH in the open flag. Then we
974 * cannot have anything other than the below set of flags
975 */
Al Viro1abf0c72011-03-13 03:51:11 -0400976 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
977 acc_mode = 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400978 }
Al Viro47c805d2011-02-23 17:44:09 -0500979
Al Viro1abf0c72011-03-13 03:51:11 -0400980 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500981
982 /* O_TRUNC implies we need access checks for write permissions */
983 if (flags & O_TRUNC)
984 acc_mode |= MAY_WRITE;
985
986 /* Allow the LSM permission hook to distinguish append
987 access from general write access. */
988 if (flags & O_APPEND)
989 acc_mode |= MAY_APPEND;
990
991 op->acc_mode = acc_mode;
992
Al Viro1abf0c72011-03-13 03:51:11 -0400993 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
994
Al Viro47c805d2011-02-23 17:44:09 -0500995 if (flags & O_CREAT) {
996 op->intent |= LOOKUP_CREATE;
997 if (flags & O_EXCL)
998 op->intent |= LOOKUP_EXCL;
999 }
1000
1001 if (flags & O_DIRECTORY)
1002 lookup_flags |= LOOKUP_DIRECTORY;
1003 if (!(flags & O_NOFOLLOW))
1004 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +04001005 op->lookup_flags = lookup_flags;
1006 return 0;
Al Viro47c805d2011-02-23 17:44:09 -05001007}
1008
1009/**
Jeff Layton669abf42012-10-10 16:43:10 -04001010 * file_open_name - open file and return file pointer
1011 *
1012 * @name: struct filename containing path to open
1013 * @flags: open flags as per the open(2) second argument
1014 * @mode: mode for the new file if O_CREAT is set, else ignored
1015 *
1016 * This is the helper to open a file from kernelspace if you really
1017 * have to. But in generally you should not do this, so please move
1018 * along, nothing to see here..
1019 */
1020struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1021{
1022 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001023 int err = build_open_flags(flags, mode, &op);
1024 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -04001025}
1026
1027/**
Al Viro47c805d2011-02-23 17:44:09 -05001028 * filp_open - open file and return file pointer
1029 *
1030 * @filename: path to open
1031 * @flags: open flags as per the open(2) second argument
1032 * @mode: mode for the new file if O_CREAT is set, else ignored
1033 *
1034 * This is the helper to open a file from kernelspace if you really
1035 * have to. But in generally you should not do this, so please move
1036 * along, nothing to see here..
1037 */
Al Viroa218d0f2011-11-21 14:59:34 -05001038struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -05001039{
Paul Moore51689102015-01-22 00:00:03 -05001040 struct filename *name = getname_kernel(filename);
1041 struct file *file = ERR_CAST(name);
1042
1043 if (!IS_ERR(name)) {
1044 file = file_open_name(name, flags, mode);
1045 putname(name);
1046 }
1047 return file;
Al Viro47c805d2011-02-23 17:44:09 -05001048}
1049EXPORT_SYMBOL(filp_open);
1050
Al Viro73d049a2011-03-11 12:08:24 -05001051struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
Jann Horn378c6522016-03-22 14:25:36 -07001052 const char *filename, int flags, umode_t mode)
Al Viro73d049a2011-03-11 12:08:24 -05001053{
1054 struct open_flags op;
Jann Horn378c6522016-03-22 14:25:36 -07001055 int err = build_open_flags(flags, mode, &op);
Al Virof9652e12013-06-11 08:23:01 +04001056 if (err)
1057 return ERR_PTR(err);
Al Virof9652e12013-06-11 08:23:01 +04001058 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -05001059}
1060EXPORT_SYMBOL(file_open_root);
1061
James Bottomley9a08c352016-02-17 16:49:38 -08001062struct file *filp_clone_open(struct file *oldfile)
1063{
1064 struct file *file;
1065 int retval;
1066
1067 file = get_empty_filp();
1068 if (IS_ERR(file))
1069 return file;
1070
1071 file->f_flags = oldfile->f_flags;
1072 retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred);
1073 if (retval) {
1074 put_filp(file);
1075 return ERR_PTR(retval);
1076 }
1077
1078 return file;
1079}
1080EXPORT_SYMBOL(filp_clone_open);
1081
Al Viroa218d0f2011-11-21 14:59:34 -05001082long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083{
Al Viro47c805d2011-02-23 17:44:09 -05001084 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001085 int fd = build_open_flags(flags, mode, &op);
1086 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087
Al Virof9652e12013-06-11 08:23:01 +04001088 if (fd)
1089 return fd;
1090
1091 tmp = getname(filename);
1092 if (IS_ERR(tmp))
1093 return PTR_ERR(tmp);
1094
1095 fd = get_unused_fd_flags(flags);
1096 if (fd >= 0) {
1097 struct file *f = do_filp_open(dfd, tmp, &op);
1098 if (IS_ERR(f)) {
1099 put_unused_fd(fd);
1100 fd = PTR_ERR(f);
1101 } else {
1102 fsnotify_open(f);
1103 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 }
Al Virof9652e12013-06-11 08:23:01 +04001106 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001109
Al Viroa218d0f2011-11-21 14:59:34 -05001110SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001111{
1112 if (force_o_largefile())
1113 flags |= O_LARGEFILE;
1114
Al Viro2cf09662013-01-21 15:25:54 -05001115 return do_sys_open(AT_FDCWD, filename, flags, mode);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001116}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117
Heiko Carstens6559eed82009-01-14 14:14:32 +01001118SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001119 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001120{
1121 if (force_o_largefile())
1122 flags |= O_LARGEFILE;
1123
Al Viro2cf09662013-01-21 15:25:54 -05001124 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001125}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127#ifndef __alpha__
1128
1129/*
1130 * For backward compatibility? Maybe this should be moved
1131 * into arch/i386 instead?
1132 */
Al Viroa218d0f2011-11-21 14:59:34 -05001133SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134{
1135 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1136}
1137
1138#endif
1139
1140/*
1141 * "id" is the POSIX thread ID. We use the
1142 * files pointer for this..
1143 */
1144int filp_close(struct file *filp, fl_owner_t id)
1145{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001146 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148 if (!file_count(filp)) {
1149 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001150 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 }
1152
Al Viro72c2d532013-09-22 16:27:52 -04001153 if (filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001154 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Al Viro1abf0c72011-03-13 03:51:11 -04001156 if (likely(!(filp->f_mode & FMODE_PATH))) {
1157 dnotify_flush(filp, id);
1158 locks_remove_posix(filp, id);
1159 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 fput(filp);
1161 return retval;
1162}
1163
1164EXPORT_SYMBOL(filp_close);
1165
1166/*
1167 * Careful here! We test whether the file pointer is NULL before
1168 * releasing the fd. This ensures that one clone task can't release
1169 * an fd while another clone is opening it.
1170 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001171SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
Al Viro483ce1d2012-08-19 12:04:24 -04001173 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001174
1175 /* can't restart close syscall because file table entry was cleared */
1176 if (unlikely(retval == -ERESTARTSYS ||
1177 retval == -ERESTARTNOINTR ||
1178 retval == -ERESTARTNOHAND ||
1179 retval == -ERESTART_RESTARTBLOCK))
1180 retval = -EINTR;
1181
1182 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184EXPORT_SYMBOL(sys_close);
1185
1186/*
1187 * This routine simulates a hangup on the tty, to arrange that users
1188 * are given clean terminals at login time.
1189 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001190SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191{
1192 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001193 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 return 0;
1195 }
1196 return -EPERM;
1197}
1198
1199/*
1200 * Called when an inode is about to be open.
1201 * We use this to disallow opening large files on 32bit systems if
1202 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1203 * on this flag in sys_open.
1204 */
1205int generic_file_open(struct inode * inode, struct file * filp)
1206{
1207 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001208 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 return 0;
1210}
1211
1212EXPORT_SYMBOL(generic_file_open);
1213
1214/*
1215 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001216 * file descriptors. The function is not supposed to ever fail, the only
1217 * reason it returns an 'int' and not 'void' is so that it can be plugged
1218 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 */
1220int nonseekable_open(struct inode *inode, struct file *filp)
1221{
1222 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1223 return 0;
1224}
1225
1226EXPORT_SYMBOL(nonseekable_open);
Kirill Smelkov9c829b62019-03-26 22:20:43 +00001227
1228/*
1229 * stream_open is used by subsystems that want stream-like file descriptors.
1230 * Such file descriptors are not seekable and don't have notion of position
1231 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1232 * files, .read() and .write() can run simultaneously.
1233 *
1234 * stream_open never fails and is marked to return int so that it could be
1235 * directly used as file_operations.open .
1236 */
1237int stream_open(struct inode *inode, struct file *filp)
1238{
1239 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1240 filp->f_mode |= FMODE_STREAM;
1241 return 0;
1242}
1243
1244EXPORT_SYMBOL(stream_open);