blob: e17cc79bd88a38a94e748cc09683940bb407ebb0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/string.h>
8#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040010#include <linux/fdtable.h>
Robert Love0eeca282005-07-12 17:06:03 -040011#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/tty.h>
14#include <linux/namei.h>
15#include <linux/backing-dev.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080016#include <linux/capability.h>
Andrew G. Morgan086f7312008-07-04 09:59:58 -070017#include <linux/securebits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/mount.h>
Ulrich Drepper5590ff02006-01-18 17:43:53 -080020#include <linux/fcntl.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090021#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <asm/uaccess.h>
23#include <linux/fs.h>
Yoav Zachef3daed2005-06-23 00:09:58 -070024#include <linux/personality.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/pagemap.h>
26#include <linux/syscalls.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070027#include <linux/rcupdate.h>
Amy Griffis73241cc2005-11-03 16:00:25 +000028#include <linux/audit.h>
Amit Arora97ac7352007-07-17 21:42:44 -040029#include <linux/falloc.h>
Al Viro5ad4e532009-03-29 19:50:06 -040030#include <linux/fs_struct.h>
Al Virob65a9cf2009-12-16 06:27:40 -050031#include <linux/ima.h>
Eric Paris2dfc1ca2009-12-17 20:30:52 -050032#include <linux/dnotify.h>
Al Viro3f6d0782013-02-24 13:49:08 -050033#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Eric Parise81e3f42009-12-04 15:47:36 -050035#include "internal.h"
36
NeilBrown4a301312006-01-08 01:02:39 -080037int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
38 struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070039{
Amerigo Wang939a9422009-08-20 19:29:03 -070040 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 struct iattr newattrs;
42
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
44 if (length < 0)
45 return -EINVAL;
46
47 newattrs.ia_size = length;
NeilBrown4a301312006-01-08 01:02:39 -080048 newattrs.ia_valid = ATTR_SIZE | time_attrs;
Miklos Szeredicc4e69d2005-11-07 00:59:49 -080049 if (filp) {
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
52 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Jan Kara45f147a2015-05-21 16:05:55 +020054 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
56 if (ret < 0)
57 return ret;
Amerigo Wang939a9422009-08-20 19:29:03 -070058 if (ret)
59 newattrs.ia_valid |= ret | ATTR_FORCE;
Linus Torvalds7b82dc02007-05-08 20:10:00 -070060
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: */
63 ret = notify_change(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}
67
Al Viro7df818b2016-03-25 14:24:09 -040068long vfs_truncate(const struct path *path, loff_t length)
Linus Torvalds1da177e2005-04-16 15:20:36 -070069{
Al Viro2d8f3032008-07-22 09:59:21 -040070 struct inode *inode;
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +020071 struct dentry *upperdentry;
David Howellsa02de962012-12-20 21:52:36 +000072 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
David Howellsa02de962012-12-20 21:52:36 +000074 inode = path->dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
76 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 if (S_ISDIR(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000078 return -EISDIR;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 if (!S_ISREG(inode->i_mode))
David Howellsa02de962012-12-20 21:52:36 +000080 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
David Howellsa02de962012-12-20 21:52:36 +000082 error = mnt_want_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 if (error)
David Howellsa02de962012-12-20 21:52:36 +000084 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
Al Viro256984a2008-07-22 08:09:30 -040086 error = inode_permission(inode, MAY_WRITE);
Dave Hansen9ac9b842008-02-15 14:37:52 -080087 if (error)
88 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
90 error = -EPERM;
Miklos Szeredic82e42d2008-06-24 16:50:12 +020091 if (IS_APPEND(inode))
Dave Hansen9ac9b842008-02-15 14:37:52 -080092 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +020094 /*
95 * If this is an overlayfs then do as if opening the file so we get
96 * write access on the upper inode, not on the overlay inode. For
97 * non-overlay filesystems d_real() is an identity function.
98 */
99 upperdentry = d_real(path->dentry, NULL, O_WRONLY);
100 error = PTR_ERR(upperdentry);
101 if (IS_ERR(upperdentry))
102 goto mnt_drop_write_and_out;
103
104 error = get_write_access(upperdentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 if (error)
Dave Hansen9ac9b842008-02-15 14:37:52 -0800106 goto mnt_drop_write_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107
david m. richter97003822007-07-31 00:39:12 -0700108 /*
109 * Make sure that there are no leases. get_write_access() protects
110 * against the truncate racing with a lease-granting setlease().
111 */
Al Viro8737c932009-12-24 06:47:55 -0500112 error = break_lease(inode, O_WRONLY);
david m. richter97003822007-07-31 00:39:12 -0700113 if (error)
114 goto put_write_and_out;
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 error = locks_verify_truncate(inode, NULL, length);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900117 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000118 error = security_path_truncate(path);
Christoph Hellwig907f4552010-03-03 09:05:06 -0500119 if (!error)
David Howellsa02de962012-12-20 21:52:36 +0000120 error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
david m. richter97003822007-07-31 00:39:12 -0700122put_write_and_out:
Miklos Szeredi4d0c5ba2016-09-16 12:44:21 +0200123 put_write_access(upperdentry->d_inode);
Dave Hansen9ac9b842008-02-15 14:37:52 -0800124mnt_drop_write_and_out:
David Howellsa02de962012-12-20 21:52:36 +0000125 mnt_drop_write(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126out:
127 return error;
128}
David Howellsa02de962012-12-20 21:52:36 +0000129EXPORT_SYMBOL_GPL(vfs_truncate);
130
131static long do_sys_truncate(const char __user *pathname, loff_t length)
132{
Jeff Layton48f75302012-12-11 12:10:11 -0500133 unsigned int lookup_flags = LOOKUP_FOLLOW;
David Howellsa02de962012-12-20 21:52:36 +0000134 struct path path;
135 int error;
136
137 if (length < 0) /* sorry, but loff_t says... */
138 return -EINVAL;
139
Jeff Layton48f75302012-12-11 12:10:11 -0500140retry:
141 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
David Howellsa02de962012-12-20 21:52:36 +0000142 if (!error) {
143 error = vfs_truncate(&path, length);
144 path_put(&path);
145 }
Jeff Layton48f75302012-12-11 12:10:11 -0500146 if (retry_estale(error, lookup_flags)) {
147 lookup_flags |= LOOKUP_REVAL;
148 goto retry;
149 }
David Howellsa02de962012-12-20 21:52:36 +0000150 return error;
151}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200153SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154{
Heiko Carstens4fd8da82009-09-23 17:49:55 +0200155 return do_sys_truncate(path, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156}
157
Al Viro3f6d0782013-02-24 13:49:08 -0500158#ifdef CONFIG_COMPAT
159COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
160{
161 return do_sys_truncate(path, length);
162}
163#endif
164
Matt Mackallb01ec0e2006-01-08 01:05:20 -0800165static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166{
Al Virobf2965d2012-08-26 20:13:36 -0400167 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 struct dentry *dentry;
Al Viro2903ff02012-08-28 12:52:22 -0400169 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 int error;
171
172 error = -EINVAL;
173 if (length < 0)
174 goto out;
175 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400176 f = fdget(fd);
177 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 goto out;
179
180 /* explicitly opened as large or we are on 64-bit box */
Al Viro2903ff02012-08-28 12:52:22 -0400181 if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 small = 0;
183
Al Viro2903ff02012-08-28 12:52:22 -0400184 dentry = f.file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 inode = dentry->d_inode;
186 error = -EINVAL;
Al Viro2903ff02012-08-28 12:52:22 -0400187 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 goto out_putf;
189
190 error = -EINVAL;
191 /* Cannot ftruncate over 2^31 bytes without large file support */
192 if (small && length > MAX_NON_LFS)
193 goto out_putf;
194
195 error = -EPERM;
196 if (IS_APPEND(inode))
197 goto out_putf;
198
Jan Kara14da9202012-06-12 16:20:37 +0200199 sb_start_write(inode->i_sb);
Al Viro2903ff02012-08-28 12:52:22 -0400200 error = locks_verify_truncate(inode, f.file, length);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400202 error = security_path_truncate(&f.file->f_path);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900203 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400204 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
Jan Kara14da9202012-06-12 16:20:37 +0200205 sb_end_write(inode->i_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400207 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208out:
209 return error;
210}
211
Heiko Carstensbdc480e2009-01-14 14:14:12 +0100212SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213{
Al Viro2cf09662013-01-21 15:25:54 -0500214 return do_sys_ftruncate(fd, length, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215}
216
Al Viro3f6d0782013-02-24 13:49:08 -0500217#ifdef CONFIG_COMPAT
218COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
219{
220 return do_sys_ftruncate(fd, length, 1);
221}
222#endif
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224/* LFS versions of truncate are only needed on 32 bit machines */
225#if BITS_PER_LONG == 32
Al Viro4a0fd5b2013-01-21 15:16:58 -0500226SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
228 return do_sys_truncate(path, length);
229}
230
Al Viro4a0fd5b2013-01-21 15:16:58 -0500231SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232{
Al Viro2cf09662013-01-21 15:25:54 -0500233 return do_sys_ftruncate(fd, length, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234}
Heiko Carstens6673e0c2009-01-14 14:14:02 +0100235#endif /* BITS_PER_LONG == 32 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400237
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500238int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
Amit Arora97ac7352007-07-17 21:42:44 -0400239{
Al Viro496ad9a2013-01-23 17:07:38 -0500240 struct inode *inode = file_inode(file);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400241 long ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400242
243 if (offset < 0 || len <= 0)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400244 return -EINVAL;
Amit Arora97ac7352007-07-17 21:42:44 -0400245
246 /* Return error if mode is not supported */
Namjae Jeondd46c782015-03-25 15:07:05 +1100247 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
Lukas Czerner409332b2014-03-13 19:07:42 +1100248 return -EOPNOTSUPP;
249
250 /* Punch hole and zero range are mutually exclusive */
251 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
252 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
Josef Bacik79124f182010-11-17 20:46:15 -0500253 return -EOPNOTSUPP;
254
255 /* Punch hole must have keep size set */
256 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
257 !(mode & FALLOC_FL_KEEP_SIZE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400258 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400259
Namjae Jeon00f5e612014-02-24 10:58:15 +1100260 /* Collapse range should only be used exclusively. */
261 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
262 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
263 return -EINVAL;
264
Namjae Jeondd46c782015-03-25 15:07:05 +1100265 /* Insert range should only be used exclusively. */
266 if ((mode & FALLOC_FL_INSERT_RANGE) &&
267 (mode & ~FALLOC_FL_INSERT_RANGE))
268 return -EINVAL;
269
Darrick J. Wong71be6b42016-10-03 09:11:14 -0700270 /* Unshare range should only be used with allocate mode. */
271 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
272 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
273 return -EINVAL;
274
Amit Arora97ac7352007-07-17 21:42:44 -0400275 if (!(file->f_mode & FMODE_WRITE))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400276 return -EBADF;
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100277
Namjae Jeon00f5e612014-02-24 10:58:15 +1100278 /*
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400279 * We can only allow pure fallocate on append only files
Namjae Jeon00f5e612014-02-24 10:58:15 +1100280 */
Lukas Czerner8fc61d92014-04-12 09:51:34 -0400281 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
Marco Stornelli1ca551c2011-03-05 11:10:19 +0100282 return -EPERM;
283
284 if (IS_IMMUTABLE(inode))
285 return -EPERM;
286
Amit Arora97ac7352007-07-17 21:42:44 -0400287 /*
Eric Biggers6d2b6172014-06-24 23:45:08 -0500288 * We cannot allow any fallocate operation on an active swapfile
Lukas Czerner0790b312014-04-12 10:05:37 -0400289 */
290 if (IS_SWAPFILE(inode))
Eric Biggers6d2b6172014-06-24 23:45:08 -0500291 return -ETXTBSY;
Lukas Czerner0790b312014-04-12 10:05:37 -0400292
293 /*
Amit Arora97ac7352007-07-17 21:42:44 -0400294 * Revalidate the write permissions, in case security policy has
295 * changed since the files were opened.
296 */
297 ret = security_file_permission(file, MAY_WRITE);
298 if (ret)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400299 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400300
Amit Arora97ac7352007-07-17 21:42:44 -0400301 if (S_ISFIFO(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400302 return -ESPIPE;
Amit Arora97ac7352007-07-17 21:42:44 -0400303
Amit Arora97ac7352007-07-17 21:42:44 -0400304 /*
305 * Let individual file system decide if it supports preallocation
306 * for directories or not.
307 */
Darrick J. Wong25f4c412016-10-11 13:51:11 -0700308 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) &&
309 !S_ISBLK(inode->i_mode))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400310 return -ENODEV;
Amit Arora97ac7352007-07-17 21:42:44 -0400311
Amit Arora97ac7352007-07-17 21:42:44 -0400312 /* Check for wrap through zero too */
313 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400314 return -EFBIG;
Amit Arora97ac7352007-07-17 21:42:44 -0400315
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100316 if (!file->f_op->fallocate)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400317 return -EOPNOTSUPP;
Amit Arora97ac7352007-07-17 21:42:44 -0400318
Jan Kara14da9202012-06-12 16:20:37 +0200319 sb_start_write(inode->i_sb);
320 ret = file->f_op->fallocate(file, mode, offset, len);
Heinrich Schuchardt820c12d2014-12-12 16:58:34 -0800321
322 /*
323 * Create inotify and fanotify events.
324 *
325 * To keep the logic simple always create events if fallocate succeeds.
326 * This implies that events are even created if the file size remains
327 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
328 */
329 if (ret == 0)
330 fsnotify_modify(file);
331
Jan Kara14da9202012-06-12 16:20:37 +0200332 sb_end_write(inode->i_sb);
333 return ret;
Amit Arora97ac7352007-07-17 21:42:44 -0400334}
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500335EXPORT_SYMBOL_GPL(vfs_fallocate);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400336
Al Viro4a0fd5b2013-01-21 15:16:58 -0500337SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400338{
Al Viro2903ff02012-08-28 12:52:22 -0400339 struct fd f = fdget(fd);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400340 int error = -EBADF;
341
Al Viro2903ff02012-08-28 12:52:22 -0400342 if (f.file) {
Anna Schumaker72c72bd2014-11-07 14:44:25 -0500343 error = vfs_fallocate(f.file, mode, offset, len);
Al Viro2903ff02012-08-28 12:52:22 -0400344 fdput(f);
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400345 }
Ankit Jain3e63cbb2009-06-19 14:28:07 -0400346 return error;
347}
348
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349/*
350 * access() needs to use the real uid/gid, not the effective uid/gid.
351 * We do this by temporarily clearing all FS-related capabilities and
352 * switching the fsuid/fsgid around to the real ones.
353 */
Heiko Carstens6559eed82009-01-14 14:14:32 +0100354SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355{
David Howellsd84f4f92008-11-14 10:39:23 +1100356 const struct cred *old_cred;
357 struct cred *override_cred;
Al Viro2d8f3032008-07-22 09:59:21 -0400358 struct path path;
Al Viro256984a2008-07-22 08:09:30 -0400359 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 int res;
Jeff Layton87fa5592012-12-11 12:10:11 -0500361 unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
363 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
364 return -EINVAL;
365
David Howellsd84f4f92008-11-14 10:39:23 +1100366 override_cred = prepare_creds();
367 if (!override_cred)
368 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
David Howellsd84f4f92008-11-14 10:39:23 +1100370 override_cred->fsuid = override_cred->uid;
371 override_cred->fsgid = override_cred->gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700373 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells1cdcbec2008-11-14 10:39:14 +1100374 /* Clear the capabilities if we switch to a non-root user */
Eric W. Biederman18815a12012-02-07 16:45:47 -0800375 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
376 if (!uid_eq(override_cred->uid, root_uid))
David Howellsd84f4f92008-11-14 10:39:23 +1100377 cap_clear(override_cred->cap_effective);
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700378 else
David Howellsd84f4f92008-11-14 10:39:23 +1100379 override_cred->cap_effective =
380 override_cred->cap_permitted;
Andrew G. Morgan086f7312008-07-04 09:59:58 -0700381 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
Linus Torvalds50810012019-07-11 09:54:40 -0700383 /*
384 * The new set of credentials can *only* be used in
385 * task-synchronous circumstances, and does not need
386 * RCU freeing, unless somebody then takes a separate
387 * reference to it.
388 *
389 * NOTE! This is _only_ true because this credential
390 * is used purely for override_creds() that installs
391 * it as the subjective cred. Other threads will be
392 * accessing ->real_cred, not the subjective cred.
393 *
394 * If somebody _does_ make a copy of this (using the
395 * 'get_current_cred()' function), that will clear the
396 * non_rcu field, because now that other user may be
397 * expecting RCU freeing. But normal thread-synchronous
398 * cred accesses will keep things non-RCY.
399 */
400 override_cred->non_rcu = 1;
401
David Howellsd84f4f92008-11-14 10:39:23 +1100402 old_cred = override_creds(override_cred);
Jeff Layton87fa5592012-12-11 12:10:11 -0500403retry:
404 res = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700405 if (res)
406 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
David Howells63afdfc2015-05-06 15:59:00 +0100408 inode = d_backing_inode(path.dentry);
Al Viro256984a2008-07-22 08:09:30 -0400409
410 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
Al Viro30524472008-07-22 00:02:33 -0400411 /*
412 * MAY_EXEC on regular files is denied if the fs is mounted
413 * with the "noexec" flag.
414 */
415 res = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500416 if (path_noexec(&path))
Al Viro30524472008-07-22 00:02:33 -0400417 goto out_path_release;
418 }
419
Al Viro256984a2008-07-22 08:09:30 -0400420 res = inode_permission(inode, mode | MAY_ACCESS);
Dave Hansen6902d922006-09-30 23:29:01 -0700421 /* SuS v2 requires we report a read only fs too */
Al Viro256984a2008-07-22 08:09:30 -0400422 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
Dave Hansen6902d922006-09-30 23:29:01 -0700423 goto out_path_release;
Dave Hansen2f676cb2008-02-15 14:37:55 -0800424 /*
425 * This is a rare case where using __mnt_is_readonly()
426 * is OK without a mnt_want/drop_write() pair. Since
427 * no actual write to the fs is performed here, we do
428 * not need to telegraph to that to anyone.
429 *
430 * By doing this, we accept that this access is
431 * inherently racy and know that the fs may change
432 * state before we even see this result.
433 */
Al Viro2d8f3032008-07-22 09:59:21 -0400434 if (__mnt_is_readonly(path.mnt))
Dave Hansen6902d922006-09-30 23:29:01 -0700435 res = -EROFS;
436
437out_path_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400438 path_put(&path);
Jeff Layton87fa5592012-12-11 12:10:11 -0500439 if (retry_estale(res, lookup_flags)) {
440 lookup_flags |= LOOKUP_REVAL;
441 goto retry;
442 }
Dave Hansen6902d922006-09-30 23:29:01 -0700443out:
David Howellsd84f4f92008-11-14 10:39:23 +1100444 revert_creds(old_cred);
445 put_cred(override_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 return res;
447}
448
Heiko Carstensca013e92009-01-14 14:14:19 +0100449SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800450{
451 return sys_faccessat(AT_FDCWD, filename, mode);
452}
453
Heiko Carstens3cdad422009-01-14 14:14:22 +0100454SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455{
Al Viro2d8f3032008-07-22 09:59:21 -0400456 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 int error;
Jeff Layton0291c0a2012-12-11 12:10:12 -0500458 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
459retry:
460 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 if (error)
462 goto out;
463
Eric Paris9cfcac82010-07-23 11:43:51 -0400464 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (error)
466 goto dput_and_out;
467
Al Viro2d8f3032008-07-22 09:59:21 -0400468 set_fs_pwd(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400471 path_put(&path);
Jeff Layton0291c0a2012-12-11 12:10:12 -0500472 if (retry_estale(error, lookup_flags)) {
473 lookup_flags |= LOOKUP_REVAL;
474 goto retry;
475 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476out:
477 return error;
478}
479
Heiko Carstens3cdad422009-01-14 14:14:22 +0100480SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481{
Al Viro2903ff02012-08-28 12:52:22 -0400482 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 struct inode *inode;
Al Viro2903ff02012-08-28 12:52:22 -0400484 int error = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
486 error = -EBADF;
Al Viro2903ff02012-08-28 12:52:22 -0400487 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 goto out;
489
Al Viro496ad9a2013-01-23 17:07:38 -0500490 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 error = -ENOTDIR;
493 if (!S_ISDIR(inode->i_mode))
494 goto out_putf;
495
Eric Paris9cfcac82010-07-23 11:43:51 -0400496 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 if (!error)
Al Viro2903ff02012-08-28 12:52:22 -0400498 set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499out_putf:
Al Viro2903ff02012-08-28 12:52:22 -0400500 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501out:
502 return error;
503}
504
Heiko Carstens3480b252009-01-14 14:14:16 +0100505SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
Al Viro2d8f3032008-07-22 09:59:21 -0400507 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 int error;
Jeff Layton27712612012-12-20 17:08:32 -0500509 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
510retry:
511 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 if (error)
513 goto out;
514
Eric Paris9cfcac82010-07-23 11:43:51 -0400515 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 if (error)
517 goto dput_and_out;
518
519 error = -EPERM;
Eric W. Biedermanc7b96ac2013-03-20 12:49:49 -0700520 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 goto dput_and_out;
Tetsuo Handa8b8efb42009-10-04 21:49:48 +0900522 error = security_path_chroot(&path);
523 if (error)
524 goto dput_and_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
Al Viro2d8f3032008-07-22 09:59:21 -0400526 set_fs_root(current->fs, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 error = 0;
528dput_and_out:
Al Viro2d8f3032008-07-22 09:59:21 -0400529 path_put(&path);
Jeff Layton27712612012-12-20 17:08:32 -0500530 if (retry_estale(error, lookup_flags)) {
531 lookup_flags |= LOOKUP_REVAL;
532 goto retry;
533 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534out:
535 return error;
536}
537
Al Virobe01f9f2016-03-25 14:56:23 -0400538static int chmod_common(const struct path *path, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539{
Al Viroe57712e2011-07-26 04:15:54 -0400540 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400541 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 struct iattr newattrs;
Al Viroe57712e2011-07-26 04:15:54 -0400543 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Al Viroe57712e2011-07-26 04:15:54 -0400545 error = mnt_want_write(path->mnt);
546 if (error)
547 return error;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400548retry_deleg:
Al Viro59551022016-01-22 15:40:57 -0500549 inode_lock(inode);
Al Virocdcf1162011-12-08 10:51:53 -0500550 error = security_path_chmod(path, mode);
Al Viroe57712e2011-07-26 04:15:54 -0400551 if (error)
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900552 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
554 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400555 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900556out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500557 inode_unlock(inode);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400558 if (delegated_inode) {
559 error = break_deleg_wait(&delegated_inode);
560 if (!error)
561 goto retry_deleg;
562 }
Al Viroe57712e2011-07-26 04:15:54 -0400563 mnt_drop_write(path->mnt);
564 return error;
565}
566
Al Viro49f0a072011-07-26 04:22:01 -0400567SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
Al Viroe57712e2011-07-26 04:15:54 -0400568{
Al Viro173c8402013-08-30 12:48:53 -0400569 struct fd f = fdget(fd);
Al Viroe57712e2011-07-26 04:15:54 -0400570 int err = -EBADF;
571
Al Viro173c8402013-08-30 12:48:53 -0400572 if (f.file) {
Al Viro9f45f5b2014-10-31 17:44:57 -0400573 audit_file(f.file);
Al Viro173c8402013-08-30 12:48:53 -0400574 err = chmod_common(&f.file->f_path, mode);
575 fdput(f);
Al Viroe57712e2011-07-26 04:15:54 -0400576 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 return err;
578}
579
Al Viro49f0a072011-07-26 04:22:01 -0400580SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581{
Al Viro2d8f3032008-07-22 09:59:21 -0400582 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 int error;
Jeff Layton14ff6902012-12-11 12:10:13 -0500584 unsigned int lookup_flags = LOOKUP_FOLLOW;
585retry:
586 error = user_path_at(dfd, filename, lookup_flags, &path);
Al Viroe57712e2011-07-26 04:15:54 -0400587 if (!error) {
588 error = chmod_common(&path, mode);
589 path_put(&path);
Jeff Layton14ff6902012-12-11 12:10:13 -0500590 if (retry_estale(error, lookup_flags)) {
591 lookup_flags |= LOOKUP_REVAL;
592 goto retry;
593 }
Al Viroe57712e2011-07-26 04:15:54 -0400594 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 return error;
596}
597
Al Viro49f0a072011-07-26 04:22:01 -0400598SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800599{
600 return sys_fchmodat(AT_FDCWD, filename, mode);
601}
602
Al Viro7fd25da2016-03-25 14:44:41 -0400603static int chown_common(const struct path *path, uid_t user, gid_t group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604{
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900605 struct inode *inode = path->dentry->d_inode;
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400606 struct inode *delegated_inode = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 int error;
608 struct iattr newattrs;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800609 kuid_t uid;
610 kgid_t gid;
611
612 uid = make_kuid(current_user_ns(), user);
613 gid = make_kgid(current_user_ns(), group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614
Andrew Elblec1b89402015-02-23 08:51:24 -0500615retry_deleg:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 newattrs.ia_valid = ATTR_CTIME;
617 if (user != (uid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800618 if (!uid_valid(uid))
619 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 newattrs.ia_valid |= ATTR_UID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800621 newattrs.ia_uid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 }
623 if (group != (gid_t) -1) {
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800624 if (!gid_valid(gid))
625 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 newattrs.ia_valid |= ATTR_GID;
Eric W. Biederman52137ab2012-03-03 19:52:01 -0800627 newattrs.ia_gid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 }
629 if (!S_ISDIR(inode->i_mode))
Serge E. Hallynb5376772007-10-16 23:31:36 -0700630 newattrs.ia_valid |=
631 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Al Viro59551022016-01-22 15:40:57 -0500632 inode_lock(inode);
Eric W. Biedermand2b31ca2012-06-01 16:14:19 -0600633 error = security_path_chown(path, uid, gid);
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900634 if (!error)
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400635 error = notify_change(path->dentry, &newattrs, &delegated_inode);
Al Viro59551022016-01-22 15:40:57 -0500636 inode_unlock(inode);
J. Bruce Fields27ac0ff2011-09-20 17:19:26 -0400637 if (delegated_inode) {
638 error = break_deleg_wait(&delegated_inode);
639 if (!error)
640 goto retry_deleg;
641 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 return error;
643}
644
Heiko Carstens6559eed82009-01-14 14:14:32 +0100645SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
646 gid_t, group, int, flag)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800647{
Al Viro2d8f3032008-07-22 09:59:21 -0400648 struct path path;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800649 int error = -EINVAL;
Al Viro65cfc672011-03-13 15:56:26 -0400650 int lookup_flags;
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800651
Al Viro65cfc672011-03-13 15:56:26 -0400652 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800653 goto out;
654
Al Viro65cfc672011-03-13 15:56:26 -0400655 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
656 if (flag & AT_EMPTY_PATH)
657 lookup_flags |= LOOKUP_EMPTY;
Jeff Layton99a5df32012-12-11 12:10:13 -0500658retry:
Al Viro65cfc672011-03-13 15:56:26 -0400659 error = user_path_at(dfd, filename, lookup_flags, &path);
Dave Hansen6902d922006-09-30 23:29:01 -0700660 if (error)
661 goto out;
Al Viro2d8f3032008-07-22 09:59:21 -0400662 error = mnt_want_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800663 if (error)
664 goto out_release;
Tetsuo Handafe542cf2009-11-22 11:49:55 +0900665 error = chown_common(&path, user, group);
Al Viro2d8f3032008-07-22 09:59:21 -0400666 mnt_drop_write(path.mnt);
Dave Hansen2af482a2008-02-15 14:37:50 -0800667out_release:
Al Viro2d8f3032008-07-22 09:59:21 -0400668 path_put(&path);
Jeff Layton99a5df32012-12-11 12:10:13 -0500669 if (retry_estale(error, lookup_flags)) {
670 lookup_flags |= LOOKUP_REVAL;
671 goto retry;
672 }
Ulrich Drepper5590ff02006-01-18 17:43:53 -0800673out:
674 return error;
675}
676
David Howells55e4def2012-06-25 12:55:09 +0100677SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
678{
679 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
680}
681
Heiko Carstensca013e92009-01-14 14:14:19 +0100682SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683{
David Howells55e4def2012-06-25 12:55:09 +0100684 return sys_fchownat(AT_FDCWD, filename, user, group,
685 AT_SYMLINK_NOFOLLOW);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686}
687
Heiko Carstensca013e92009-01-14 14:14:19 +0100688SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689{
Al Viro2903ff02012-08-28 12:52:22 -0400690 struct fd f = fdget(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 int error = -EBADF;
692
Al Viro2903ff02012-08-28 12:52:22 -0400693 if (!f.file)
Dave Hansen6902d922006-09-30 23:29:01 -0700694 goto out;
695
Al Viro2903ff02012-08-28 12:52:22 -0400696 error = mnt_want_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800697 if (error)
698 goto out_fput;
Al Viro9f45f5b2014-10-31 17:44:57 -0400699 audit_file(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400700 error = chown_common(&f.file->f_path, user, group);
701 mnt_drop_write_file(f.file);
Dave Hansen2af482a2008-02-15 14:37:50 -0800702out_fput:
Al Viro2903ff02012-08-28 12:52:22 -0400703 fdput(f);
Dave Hansen6902d922006-09-30 23:29:01 -0700704out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 return error;
706}
707
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200708int open_check_o_direct(struct file *f)
709{
710 /* NB: we're sure to have correct a_ops only after f_op->open */
711 if (f->f_flags & O_DIRECT) {
Matthew Wilcoxe748dcd2015-02-16 15:59:12 -0800712 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200713 return -EINVAL;
Miklos Szeredi90ad1a82012-05-21 17:30:15 +0200714 }
715 return 0;
716}
717
Al Viro02e51802012-06-10 14:32:45 -0400718static int do_dentry_open(struct file *f,
David Howells4bacc9c2015-06-18 14:32:31 +0100719 struct inode *inode,
Al Viro96b7e572012-06-10 14:22:04 -0400720 int (*open)(struct inode *, struct file *),
721 const struct cred *cred)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722{
Al Viro1abf0c72011-03-13 03:51:11 -0400723 static const struct file_operations empty_fops = {};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 int error;
725
Al Viro53009902009-12-19 10:15:07 -0500726 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700727 FMODE_PREAD | FMODE_PWRITE;
Al Viro1abf0c72011-03-13 03:51:11 -0400728
Al Virob5bcdda2012-07-20 23:28:46 +0400729 path_get(&f->f_path);
David Howells4bacc9c2015-06-18 14:32:31 +0100730 f->f_inode = inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 f->f_mapping = inode->i_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732
Al Viro3f4d5a02014-03-14 09:43:29 -0400733 if (unlikely(f->f_flags & O_PATH)) {
734 f->f_mode = FMODE_PATH;
Al Viro1abf0c72011-03-13 03:51:11 -0400735 f->f_op = &empty_fops;
Al Viro96b7e572012-06-10 14:22:04 -0400736 return 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400737 }
738
Tetsuo Handa7d7637d2019-03-28 20:43:30 -0700739 /* Any file opened for execve()/uselib() has to be a regular file. */
740 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
741 error = -EACCES;
742 goto cleanup_file;
743 }
744
Al Virodd209082014-03-14 10:56:20 -0400745 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400746 error = get_write_access(inode);
Al Viro3f4d5a02014-03-14 09:43:29 -0400747 if (unlikely(error))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto cleanup_file;
Al Viro0ccb2862014-03-14 10:40:46 -0400749 error = __mnt_want_write(f->f_path.mnt);
Al Viro3f4d5a02014-03-14 09:43:29 -0400750 if (unlikely(error)) {
Al Viro0ccb2862014-03-14 10:40:46 -0400751 put_write_access(inode);
752 goto cleanup_file;
753 }
Al Viro83f936c2014-03-14 12:02:47 -0400754 f->f_mode |= FMODE_WRITER;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 }
756
Linus Torvalds9c225f22014-03-03 09:36:58 -0800757 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
Al Viro63b6df12016-04-20 17:08:21 -0400758 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
Linus Torvalds9c225f22014-03-03 09:36:58 -0800759 f->f_mode |= FMODE_ATOMIC_POS;
760
Al Viro1abf0c72011-03-13 03:51:11 -0400761 f->f_op = fops_get(inode->i_fop);
Al Viro72c2d532013-09-22 16:27:52 -0400762 if (unlikely(WARN_ON(!f->f_op))) {
763 error = -ENODEV;
764 goto cleanup_all;
765 }
Al Viro1abf0c72011-03-13 03:51:11 -0400766
Eric Paris83d49852012-04-04 13:45:40 -0400767 error = security_file_open(f, cred);
Yuichi Nakamura788e7dd2007-09-14 09:27:07 +0900768 if (error)
769 goto cleanup_all;
770
Miklos Szeredic568d682016-09-16 12:44:20 +0200771 error = break_lease(locks_inode(f), f->f_flags);
J. Bruce Fieldsf3c7691e2011-09-21 10:58:13 -0400772 if (error)
773 goto cleanup_all;
774
Al Viro72c2d532013-09-22 16:27:52 -0400775 if (!open)
Trond Myklebust834f2a42005-10-18 14:20:16 -0700776 open = f->f_op->open;
777 if (open) {
778 error = open(inode, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 if (error)
780 goto cleanup_all;
781 }
Mimi Zohar890275b52010-11-02 10:13:07 -0400782 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
783 i_readcount_inc(inode);
Al Viro293bc982014-02-11 18:37:41 -0500784 if ((f->f_mode & FMODE_READ) &&
Al Viro84363182015-04-04 01:14:53 -0400785 likely(f->f_op->read || f->f_op->read_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500786 f->f_mode |= FMODE_CAN_READ;
Al Viro293bc982014-02-11 18:37:41 -0500787 if ((f->f_mode & FMODE_WRITE) &&
Al Viro84363182015-04-04 01:14:53 -0400788 likely(f->f_op->write || f->f_op->write_iter))
Al Viro7f7f25e2014-02-11 17:49:24 -0500789 f->f_mode |= FMODE_CAN_WRITE;
Trond Myklebust834f2a42005-10-18 14:20:16 -0700790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
792
793 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
794
Al Viro96b7e572012-06-10 14:22:04 -0400795 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
797cleanup_all:
798 fops_put(f->f_op);
Al Viro83f936c2014-03-14 12:02:47 -0400799 if (f->f_mode & FMODE_WRITER) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 put_write_access(inode);
Al Viro83f936c2014-03-14 12:02:47 -0400801 __mnt_drop_write(f->f_path.mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800802 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803cleanup_file:
Al Viro02e51802012-06-10 14:32:45 -0400804 path_put(&f->f_path);
805 f->f_path.mnt = NULL;
806 f->f_path.dentry = NULL;
Al Virodd379782013-03-01 19:48:30 -0500807 f->f_inode = NULL;
Al Viro96b7e572012-06-10 14:22:04 -0400808 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809}
810
Trond Myklebust834f2a42005-10-18 14:20:16 -0700811/**
Miklos Szeredid18e9002012-06-05 15:10:17 +0200812 * finish_open - finish opening a file
Miklos Szeredi0854d452013-09-16 14:51:55 +0200813 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200814 * @dentry: pointer to dentry
815 * @open: open callback
Miklos Szeredi0854d452013-09-16 14:51:55 +0200816 * @opened: state of open
Miklos Szeredid18e9002012-06-05 15:10:17 +0200817 *
818 * This can be used to finish opening a file passed to i_op->atomic_open().
819 *
820 * If the open callback is set to NULL, then the standard f_op->open()
821 * filesystem callback is substituted.
Miklos Szeredi0854d452013-09-16 14:51:55 +0200822 *
823 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
824 * the return value of d_splice_alias(), then the caller needs to perform dput()
825 * on it after finish_open().
826 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200827 * Returns zero on success or -errno if the open failed.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200828 */
Al Viro30d90492012-06-22 12:40:19 +0400829int finish_open(struct file *file, struct dentry *dentry,
830 int (*open)(struct inode *, struct file *),
831 int *opened)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200832{
Al Viro96b7e572012-06-10 14:22:04 -0400833 int error;
Al Viro3d8a00d2012-06-10 05:04:43 -0400834 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredid18e9002012-06-05 15:10:17 +0200835
Al Virob5bcdda2012-07-20 23:28:46 +0400836 file->f_path.dentry = dentry;
David Howells4bacc9c2015-06-18 14:32:31 +0100837 error = do_dentry_open(file, d_backing_inode(dentry), open,
838 current_cred());
Al Viro96b7e572012-06-10 14:22:04 -0400839 if (!error)
Al Viro47237682012-06-10 05:01:45 -0400840 *opened |= FILE_OPENED;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200841
Al Viro96b7e572012-06-10 14:22:04 -0400842 return error;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200843}
844EXPORT_SYMBOL(finish_open);
845
846/**
847 * finish_no_open - finish ->atomic_open() without opening the file
848 *
Miklos Szeredi0854d452013-09-16 14:51:55 +0200849 * @file: file pointer
Miklos Szeredid18e9002012-06-05 15:10:17 +0200850 * @dentry: dentry or NULL (as returned from ->lookup())
851 *
852 * This can be used to set the result of a successful lookup in ->atomic_open().
Miklos Szeredi0854d452013-09-16 14:51:55 +0200853 *
854 * NB: unlike finish_open() this function does consume the dentry reference and
855 * the caller need not dput() it.
856 *
857 * Returns "1" which must be the return value of ->atomic_open() after having
858 * called this function.
Miklos Szeredid18e9002012-06-05 15:10:17 +0200859 */
Al Viroe45198a2012-06-10 06:48:09 -0400860int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredid18e9002012-06-05 15:10:17 +0200861{
Al Viro30d90492012-06-22 12:40:19 +0400862 file->f_path.dentry = dentry;
Al Viroe45198a2012-06-10 06:48:09 -0400863 return 1;
Miklos Szeredid18e9002012-06-05 15:10:17 +0200864}
865EXPORT_SYMBOL(finish_no_open);
866
Miklos Szeredi9bf39ab2015-06-19 10:29:13 +0200867char *file_path(struct file *filp, char *buf, int buflen)
868{
869 return d_path(&filp->f_path, buf, buflen);
870}
871EXPORT_SYMBOL(file_path);
872
David Howells4bacc9c2015-06-18 14:32:31 +0100873/**
874 * vfs_open - open the file at the given path
875 * @path: path to open
876 * @file: newly allocated file with f_flag initialized
877 * @cred: credentials to use
878 */
879int vfs_open(const struct path *path, struct file *file,
880 const struct cred *cred)
881{
Miklos Szeredi2d902672016-06-30 08:53:27 +0200882 struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags);
Miklos Szeredi54d5ca82016-05-11 01:16:37 +0200883
Miklos Szeredi2d902672016-06-30 08:53:27 +0200884 if (IS_ERR(dentry))
885 return PTR_ERR(dentry);
David Howells4bacc9c2015-06-18 14:32:31 +0100886
887 file->f_path = *path;
Miklos Szeredi2d902672016-06-30 08:53:27 +0200888 return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
David Howells4bacc9c2015-06-18 14:32:31 +0100889}
890
Al Viro765927b2012-06-26 21:58:53 +0400891struct file *dentry_open(const struct path *path, int flags,
David Howells745ca242008-11-14 10:39:22 +1100892 const struct cred *cred)
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700893{
894 int error;
895 struct file *f;
896
David Howellse0e81732009-09-02 09:13:40 +0100897 validate_creds(cred);
898
Tetsuo Handac212f9a2011-01-19 21:08:41 +0900899 /* We must always pass in a valid mount pointer. */
Al Viro765927b2012-06-26 21:58:53 +0400900 BUG_ON(!path->mnt);
Christoph Hellwig322ee5b2008-02-15 14:37:24 -0800901
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700902 f = get_empty_filp();
Al Viro1afc99b2013-02-14 20:41:04 -0500903 if (!IS_ERR(f)) {
904 f->f_flags = flags;
Miklos Szeredi4aa7c632014-10-24 00:14:35 +0200905 error = vfs_open(path, f, cred);
Al Viro1afc99b2013-02-14 20:41:04 -0500906 if (!error) {
907 /* from now on we need fput() to dispose of f */
908 error = open_check_o_direct(f);
909 if (error) {
910 fput(f);
911 f = ERR_PTR(error);
912 }
913 } else {
914 put_filp(f);
Al Viro2a027e72012-06-10 14:24:38 -0400915 f = ERR_PTR(error);
916 }
Al Viro2a027e72012-06-10 14:24:38 -0400917 }
918 return f;
Peter Staubacha1a5b3d2005-09-13 01:25:12 -0700919}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920EXPORT_SYMBOL(dentry_open);
921
Al Viroa218d0f2011-11-21 14:59:34 -0500922static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
Al Viro47c805d2011-02-23 17:44:09 -0500923{
924 int lookup_flags = 0;
Al Viro62fb4a12015-12-26 22:33:24 -0500925 int acc_mode = ACC_MODE(flags);
Al Viro47c805d2011-02-23 17:44:09 -0500926
Christoph Hellwig26ff0652017-04-27 09:42:25 +0200927 /*
928 * Clear out all open flags we don't know about so that we don't report
929 * them in fcntl(F_GETFD) or similar interfaces.
930 */
931 flags &= VALID_OPEN_FLAGS;
932
Andy Lutomirskie305f482013-08-01 21:07:52 -0700933 if (flags & (O_CREAT | __O_TMPFILE))
Miklos Szeredie68726f2012-08-15 13:01:24 +0200934 op->mode = (mode & S_IALLUGO) | S_IFREG;
935 else
936 op->mode = 0;
Al Viro47c805d2011-02-23 17:44:09 -0500937
938 /* Must never be set by userspace */
Al Viroc6f3d812012-08-26 11:01:04 -0400939 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
Al Viro47c805d2011-02-23 17:44:09 -0500940
941 /*
942 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
943 * check for O_DSYNC if the need any syncing at all we enforce it's
944 * always set instead of having to deal with possibly weird behaviour
945 * for malicious applications setting only __O_SYNC.
946 */
947 if (flags & __O_SYNC)
948 flags |= O_DSYNC;
949
Al Virobb458c62013-07-13 13:26:37 +0400950 if (flags & __O_TMPFILE) {
951 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
Al Viro60545d02013-06-07 01:20:27 -0400952 return -EINVAL;
Al Viroba57ea62013-07-20 03:11:32 +0400953 if (!(acc_mode & MAY_WRITE))
954 return -EINVAL;
Al Viro60545d02013-06-07 01:20:27 -0400955 } else if (flags & O_PATH) {
956 /*
957 * If we have O_PATH in the open flag. Then we
958 * cannot have anything other than the below set of flags
959 */
Al Viro1abf0c72011-03-13 03:51:11 -0400960 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
961 acc_mode = 0;
Al Viro1abf0c72011-03-13 03:51:11 -0400962 }
Al Viro47c805d2011-02-23 17:44:09 -0500963
Al Viro1abf0c72011-03-13 03:51:11 -0400964 op->open_flag = flags;
Al Viro47c805d2011-02-23 17:44:09 -0500965
966 /* O_TRUNC implies we need access checks for write permissions */
967 if (flags & O_TRUNC)
968 acc_mode |= MAY_WRITE;
969
970 /* Allow the LSM permission hook to distinguish append
971 access from general write access. */
972 if (flags & O_APPEND)
973 acc_mode |= MAY_APPEND;
974
975 op->acc_mode = acc_mode;
976
Al Viro1abf0c72011-03-13 03:51:11 -0400977 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
978
Al Viro47c805d2011-02-23 17:44:09 -0500979 if (flags & O_CREAT) {
980 op->intent |= LOOKUP_CREATE;
981 if (flags & O_EXCL)
982 op->intent |= LOOKUP_EXCL;
983 }
984
985 if (flags & O_DIRECTORY)
986 lookup_flags |= LOOKUP_DIRECTORY;
987 if (!(flags & O_NOFOLLOW))
988 lookup_flags |= LOOKUP_FOLLOW;
Al Virof9652e12013-06-11 08:23:01 +0400989 op->lookup_flags = lookup_flags;
990 return 0;
Al Viro47c805d2011-02-23 17:44:09 -0500991}
992
993/**
Jeff Layton669abf42012-10-10 16:43:10 -0400994 * file_open_name - open file and return file pointer
995 *
996 * @name: struct filename containing path to open
997 * @flags: open flags as per the open(2) second argument
998 * @mode: mode for the new file if O_CREAT is set, else ignored
999 *
1000 * This is the helper to open a file from kernelspace if you really
1001 * have to. But in generally you should not do this, so please move
1002 * along, nothing to see here..
1003 */
1004struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1005{
1006 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001007 int err = build_open_flags(flags, mode, &op);
1008 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
Jeff Layton669abf42012-10-10 16:43:10 -04001009}
1010
1011/**
Al Viro47c805d2011-02-23 17:44:09 -05001012 * filp_open - open file and return file pointer
1013 *
1014 * @filename: path to open
1015 * @flags: open flags as per the open(2) second argument
1016 * @mode: mode for the new file if O_CREAT is set, else ignored
1017 *
1018 * This is the helper to open a file from kernelspace if you really
1019 * have to. But in generally you should not do this, so please move
1020 * along, nothing to see here..
1021 */
Al Viroa218d0f2011-11-21 14:59:34 -05001022struct file *filp_open(const char *filename, int flags, umode_t mode)
Al Viro47c805d2011-02-23 17:44:09 -05001023{
Paul Moore51689102015-01-22 00:00:03 -05001024 struct filename *name = getname_kernel(filename);
1025 struct file *file = ERR_CAST(name);
1026
1027 if (!IS_ERR(name)) {
1028 file = file_open_name(name, flags, mode);
1029 putname(name);
1030 }
1031 return file;
Al Viro47c805d2011-02-23 17:44:09 -05001032}
1033EXPORT_SYMBOL(filp_open);
1034
Al Viro73d049a2011-03-11 12:08:24 -05001035struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
Jann Horn378c6522016-03-22 14:25:36 -07001036 const char *filename, int flags, umode_t mode)
Al Viro73d049a2011-03-11 12:08:24 -05001037{
1038 struct open_flags op;
Jann Horn378c6522016-03-22 14:25:36 -07001039 int err = build_open_flags(flags, mode, &op);
Al Virof9652e12013-06-11 08:23:01 +04001040 if (err)
1041 return ERR_PTR(err);
Al Virof9652e12013-06-11 08:23:01 +04001042 return do_file_open_root(dentry, mnt, filename, &op);
Al Viro73d049a2011-03-11 12:08:24 -05001043}
1044EXPORT_SYMBOL(file_open_root);
1045
James Bottomley9a08c352016-02-17 16:49:38 -08001046struct file *filp_clone_open(struct file *oldfile)
1047{
1048 struct file *file;
1049 int retval;
1050
1051 file = get_empty_filp();
1052 if (IS_ERR(file))
1053 return file;
1054
1055 file->f_flags = oldfile->f_flags;
1056 retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred);
1057 if (retval) {
1058 put_filp(file);
1059 return ERR_PTR(retval);
1060 }
1061
1062 return file;
1063}
1064EXPORT_SYMBOL(filp_clone_open);
1065
Al Viroa218d0f2011-11-21 14:59:34 -05001066long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067{
Al Viro47c805d2011-02-23 17:44:09 -05001068 struct open_flags op;
Al Virof9652e12013-06-11 08:23:01 +04001069 int fd = build_open_flags(flags, mode, &op);
1070 struct filename *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Al Virof9652e12013-06-11 08:23:01 +04001072 if (fd)
1073 return fd;
1074
1075 tmp = getname(filename);
1076 if (IS_ERR(tmp))
1077 return PTR_ERR(tmp);
1078
1079 fd = get_unused_fd_flags(flags);
1080 if (fd >= 0) {
1081 struct file *f = do_filp_open(dfd, tmp, &op);
1082 if (IS_ERR(f)) {
1083 put_unused_fd(fd);
1084 fd = PTR_ERR(f);
1085 } else {
1086 fsnotify_open(f);
1087 fd_install(fd, f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 }
Al Virof9652e12013-06-11 08:23:01 +04001090 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 return fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092}
Miklos Szeredie922efc2005-09-06 15:18:25 -07001093
Al Viroa218d0f2011-11-21 14:59:34 -05001094SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
Miklos Szeredie922efc2005-09-06 15:18:25 -07001095{
1096 if (force_o_largefile())
1097 flags |= O_LARGEFILE;
1098
Al Viro2cf09662013-01-21 15:25:54 -05001099 return do_sys_open(AT_FDCWD, filename, flags, mode);
Miklos Szeredie922efc2005-09-06 15:18:25 -07001100}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101
Heiko Carstens6559eed82009-01-14 14:14:32 +01001102SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
Al Viroa218d0f2011-11-21 14:59:34 -05001103 umode_t, mode)
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001104{
1105 if (force_o_largefile())
1106 flags |= O_LARGEFILE;
1107
Al Viro2cf09662013-01-21 15:25:54 -05001108 return do_sys_open(dfd, filename, flags, mode);
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001109}
Ulrich Drepper5590ff02006-01-18 17:43:53 -08001110
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111#ifndef __alpha__
1112
1113/*
1114 * For backward compatibility? Maybe this should be moved
1115 * into arch/i386 instead?
1116 */
Al Viroa218d0f2011-11-21 14:59:34 -05001117SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118{
1119 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1120}
1121
1122#endif
1123
1124/*
1125 * "id" is the POSIX thread ID. We use the
1126 * files pointer for this..
1127 */
1128int filp_close(struct file *filp, fl_owner_t id)
1129{
Christoph Lameter45778ca2005-06-23 00:10:17 -07001130 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131
1132 if (!file_count(filp)) {
1133 printk(KERN_ERR "VFS: Close: file count is 0\n");
Christoph Lameter45778ca2005-06-23 00:10:17 -07001134 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 }
1136
Al Viro72c2d532013-09-22 16:27:52 -04001137 if (filp->f_op->flush)
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001138 retval = filp->f_op->flush(filp, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Al Viro1abf0c72011-03-13 03:51:11 -04001140 if (likely(!(filp->f_mode & FMODE_PATH))) {
1141 dnotify_flush(filp, id);
1142 locks_remove_posix(filp, id);
1143 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 fput(filp);
1145 return retval;
1146}
1147
1148EXPORT_SYMBOL(filp_close);
1149
1150/*
1151 * Careful here! We test whether the file pointer is NULL before
1152 * releasing the fd. This ensures that one clone task can't release
1153 * an fd while another clone is opening it.
1154 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001155SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156{
Al Viro483ce1d2012-08-19 12:04:24 -04001157 int retval = __close_fd(current->files, fd);
Ernie Petridesee731f42006-09-29 02:00:13 -07001158
1159 /* can't restart close syscall because file table entry was cleared */
1160 if (unlikely(retval == -ERESTARTSYS ||
1161 retval == -ERESTARTNOINTR ||
1162 retval == -ERESTARTNOHAND ||
1163 retval == -ERESTART_RESTARTBLOCK))
1164 retval = -EINTR;
1165
1166 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168EXPORT_SYMBOL(sys_close);
1169
1170/*
1171 * This routine simulates a hangup on the tty, to arrange that users
1172 * are given clean terminals at login time.
1173 */
Heiko Carstensca013e92009-01-14 14:14:19 +01001174SYSCALL_DEFINE0(vhangup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175{
1176 if (capable(CAP_SYS_TTY_CONFIG)) {
Alan Cox2cb59982008-10-13 10:40:30 +01001177 tty_vhangup_self();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 return 0;
1179 }
1180 return -EPERM;
1181}
1182
1183/*
1184 * Called when an inode is about to be open.
1185 * We use this to disallow opening large files on 32bit systems if
1186 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1187 * on this flag in sys_open.
1188 */
1189int generic_file_open(struct inode * inode, struct file * filp)
1190{
1191 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
Alan Coxa9c62a12007-10-16 23:30:22 -07001192 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 return 0;
1194}
1195
1196EXPORT_SYMBOL(generic_file_open);
1197
1198/*
1199 * This is used by subsystems that don't want seekable
Dmitry Torokhov06b1e102010-08-10 18:01:33 -07001200 * file descriptors. The function is not supposed to ever fail, the only
1201 * reason it returns an 'int' and not 'void' is so that it can be plugged
1202 * directly into file_operations structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 */
1204int nonseekable_open(struct inode *inode, struct file *filp)
1205{
1206 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1207 return 0;
1208}
1209
1210EXPORT_SYMBOL(nonseekable_open);
Kirill Smelkov9c829b62019-03-26 22:20:43 +00001211
1212/*
1213 * stream_open is used by subsystems that want stream-like file descriptors.
1214 * Such file descriptors are not seekable and don't have notion of position
1215 * (file.f_pos is always 0). Contrary to file descriptors of other regular
1216 * files, .read() and .write() can run simultaneously.
1217 *
1218 * stream_open never fails and is marked to return int so that it could be
1219 * directly used as file_operations.open .
1220 */
1221int stream_open(struct inode *inode, struct file *filp)
1222{
1223 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1224 filp->f_mode |= FMODE_STREAM;
1225 return 0;
1226}
1227
1228EXPORT_SYMBOL(stream_open);