Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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> | |
1da177e4 | 9 | #include <linux/file.h> |
9f3acc31 | 10 | #include <linux/fdtable.h> |
0eeca283 | 11 | #include <linux/fsnotify.h> |
1da177e4 | 12 | #include <linux/module.h> |
1da177e4 LT |
13 | #include <linux/tty.h> |
14 | #include <linux/namei.h> | |
15 | #include <linux/backing-dev.h> | |
16f7e0fe | 16 | #include <linux/capability.h> |
086f7316 | 17 | #include <linux/securebits.h> |
1da177e4 LT |
18 | #include <linux/security.h> |
19 | #include <linux/mount.h> | |
5590ff0d | 20 | #include <linux/fcntl.h> |
5a0e3ad6 | 21 | #include <linux/slab.h> |
1da177e4 LT |
22 | #include <asm/uaccess.h> |
23 | #include <linux/fs.h> | |
ef3daeda | 24 | #include <linux/personality.h> |
1da177e4 LT |
25 | #include <linux/pagemap.h> |
26 | #include <linux/syscalls.h> | |
ab2af1f5 | 27 | #include <linux/rcupdate.h> |
73241ccc | 28 | #include <linux/audit.h> |
97ac7350 | 29 | #include <linux/falloc.h> |
5ad4e53b | 30 | #include <linux/fs_struct.h> |
b65a9cfc | 31 | #include <linux/ima.h> |
2dfc1cae | 32 | #include <linux/dnotify.h> |
1da177e4 | 33 | |
e81e3f4d EP |
34 | #include "internal.h" |
35 | ||
4a30131e N |
36 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
37 | struct file *filp) | |
1da177e4 | 38 | { |
939a9421 | 39 | int ret; |
1da177e4 LT |
40 | struct iattr newattrs; |
41 | ||
42 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
43 | if (length < 0) | |
44 | return -EINVAL; | |
45 | ||
46 | newattrs.ia_size = length; | |
4a30131e | 47 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
48 | if (filp) { |
49 | newattrs.ia_file = filp; | |
50 | newattrs.ia_valid |= ATTR_FILE; | |
51 | } | |
1da177e4 | 52 | |
7b82dc0e | 53 | /* Remove suid/sgid on truncate too */ |
939a9421 AW |
54 | ret = should_remove_suid(dentry); |
55 | if (ret) | |
56 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 57 | |
1b1dcc1b | 58 | mutex_lock(&dentry->d_inode->i_mutex); |
939a9421 | 59 | ret = notify_change(dentry, &newattrs); |
1b1dcc1b | 60 | mutex_unlock(&dentry->d_inode->i_mutex); |
939a9421 | 61 | return ret; |
1da177e4 LT |
62 | } |
63 | ||
2d8f3038 | 64 | static long do_sys_truncate(const char __user *pathname, loff_t length) |
1da177e4 | 65 | { |
2d8f3038 AV |
66 | struct path path; |
67 | struct inode *inode; | |
1da177e4 LT |
68 | int error; |
69 | ||
70 | error = -EINVAL; | |
71 | if (length < 0) /* sorry, but loff_t says... */ | |
72 | goto out; | |
73 | ||
2d8f3038 | 74 | error = user_path(pathname, &path); |
1da177e4 LT |
75 | if (error) |
76 | goto out; | |
2d8f3038 | 77 | inode = path.dentry->d_inode; |
1da177e4 LT |
78 | |
79 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
80 | error = -EISDIR; | |
81 | if (S_ISDIR(inode->i_mode)) | |
82 | goto dput_and_out; | |
83 | ||
84 | error = -EINVAL; | |
85 | if (!S_ISREG(inode->i_mode)) | |
86 | goto dput_and_out; | |
87 | ||
2d8f3038 | 88 | error = mnt_want_write(path.mnt); |
1da177e4 LT |
89 | if (error) |
90 | goto dput_and_out; | |
91 | ||
256984a8 | 92 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
93 | if (error) |
94 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
95 | |
96 | error = -EPERM; | |
c82e42da | 97 | if (IS_APPEND(inode)) |
9ac9b847 | 98 | goto mnt_drop_write_and_out; |
1da177e4 | 99 | |
9700382c | 100 | error = get_write_access(inode); |
1da177e4 | 101 | if (error) |
9ac9b847 | 102 | goto mnt_drop_write_and_out; |
1da177e4 | 103 | |
9700382c | 104 | /* |
105 | * Make sure that there are no leases. get_write_access() protects | |
106 | * against the truncate racing with a lease-granting setlease(). | |
107 | */ | |
8737c930 | 108 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 109 | if (error) |
9700382c | 110 | goto put_write_and_out; |
1da177e4 LT |
111 | |
112 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 113 | if (!error) |
ea0d3ab2 | 114 | error = security_path_truncate(&path); |
907f4554 | 115 | if (!error) |
2d8f3038 | 116 | error = do_truncate(path.dentry, length, 0, NULL); |
1da177e4 | 117 | |
9700382c | 118 | put_write_and_out: |
119 | put_write_access(inode); | |
9ac9b847 | 120 | mnt_drop_write_and_out: |
2d8f3038 | 121 | mnt_drop_write(path.mnt); |
1da177e4 | 122 | dput_and_out: |
2d8f3038 | 123 | path_put(&path); |
1da177e4 LT |
124 | out: |
125 | return error; | |
126 | } | |
127 | ||
4fd8da8d | 128 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 129 | { |
4fd8da8d | 130 | return do_sys_truncate(path, length); |
1da177e4 LT |
131 | } |
132 | ||
b01ec0ef | 133 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 LT |
134 | { |
135 | struct inode * inode; | |
136 | struct dentry *dentry; | |
137 | struct file * file; | |
138 | int error; | |
139 | ||
140 | error = -EINVAL; | |
141 | if (length < 0) | |
142 | goto out; | |
143 | error = -EBADF; | |
144 | file = fget(fd); | |
145 | if (!file) | |
146 | goto out; | |
147 | ||
148 | /* explicitly opened as large or we are on 64-bit box */ | |
149 | if (file->f_flags & O_LARGEFILE) | |
150 | small = 0; | |
151 | ||
0f7fc9e4 | 152 | dentry = file->f_path.dentry; |
1da177e4 LT |
153 | inode = dentry->d_inode; |
154 | error = -EINVAL; | |
155 | if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE)) | |
156 | goto out_putf; | |
157 | ||
158 | error = -EINVAL; | |
159 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
160 | if (small && length > MAX_NON_LFS) | |
161 | goto out_putf; | |
162 | ||
163 | error = -EPERM; | |
164 | if (IS_APPEND(inode)) | |
165 | goto out_putf; | |
166 | ||
167 | error = locks_verify_truncate(inode, file, length); | |
be6d3e56 | 168 | if (!error) |
ea0d3ab2 | 169 | error = security_path_truncate(&file->f_path); |
1da177e4 | 170 | if (!error) |
6e656be8 | 171 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file); |
1da177e4 LT |
172 | out_putf: |
173 | fput(file); | |
174 | out: | |
175 | return error; | |
176 | } | |
177 | ||
bdc480e3 | 178 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 179 | { |
0a489cb3 | 180 | long ret = do_sys_ftruncate(fd, length, 1); |
385910f2 | 181 | /* avoid REGPARM breakage on x86: */ |
54a01510 | 182 | asmlinkage_protect(2, ret, fd, length); |
0a489cb3 | 183 | return ret; |
1da177e4 LT |
184 | } |
185 | ||
186 | /* LFS versions of truncate are only needed on 32 bit machines */ | |
187 | #if BITS_PER_LONG == 32 | |
6673e0c3 | 188 | SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length) |
1da177e4 LT |
189 | { |
190 | return do_sys_truncate(path, length); | |
191 | } | |
6673e0c3 HC |
192 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
193 | asmlinkage long SyS_truncate64(long path, loff_t length) | |
194 | { | |
195 | return SYSC_truncate64((const char __user *) path, length); | |
196 | } | |
197 | SYSCALL_ALIAS(sys_truncate64, SyS_truncate64); | |
198 | #endif | |
1da177e4 | 199 | |
6673e0c3 | 200 | SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length) |
1da177e4 | 201 | { |
0a489cb3 | 202 | long ret = do_sys_ftruncate(fd, length, 0); |
385910f2 | 203 | /* avoid REGPARM breakage on x86: */ |
54a01510 | 204 | asmlinkage_protect(2, ret, fd, length); |
0a489cb3 | 205 | return ret; |
1da177e4 | 206 | } |
6673e0c3 HC |
207 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
208 | asmlinkage long SyS_ftruncate64(long fd, loff_t length) | |
209 | { | |
210 | return SYSC_ftruncate64((unsigned int) fd, length); | |
211 | } | |
212 | SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64); | |
1da177e4 | 213 | #endif |
6673e0c3 | 214 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 215 | |
3e63cbb1 AJ |
216 | |
217 | int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
97ac7350 | 218 | { |
3e63cbb1 AJ |
219 | struct inode *inode = file->f_path.dentry->d_inode; |
220 | long ret; | |
97ac7350 AA |
221 | |
222 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 223 | return -EINVAL; |
97ac7350 AA |
224 | |
225 | /* Return error if mode is not supported */ | |
79124f18 JB |
226 | if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE)) |
227 | return -EOPNOTSUPP; | |
228 | ||
229 | /* Punch hole must have keep size set */ | |
230 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
231 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 232 | return -EOPNOTSUPP; |
97ac7350 | 233 | |
97ac7350 | 234 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 235 | return -EBADF; |
1ca551c6 MS |
236 | |
237 | /* It's not possible punch hole on append only file */ | |
238 | if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode)) | |
239 | return -EPERM; | |
240 | ||
241 | if (IS_IMMUTABLE(inode)) | |
242 | return -EPERM; | |
243 | ||
97ac7350 AA |
244 | /* |
245 | * Revalidate the write permissions, in case security policy has | |
246 | * changed since the files were opened. | |
247 | */ | |
248 | ret = security_file_permission(file, MAY_WRITE); | |
249 | if (ret) | |
3e63cbb1 | 250 | return ret; |
97ac7350 | 251 | |
97ac7350 | 252 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 253 | return -ESPIPE; |
97ac7350 | 254 | |
97ac7350 AA |
255 | /* |
256 | * Let individual file system decide if it supports preallocation | |
257 | * for directories or not. | |
258 | */ | |
259 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 260 | return -ENODEV; |
97ac7350 | 261 | |
97ac7350 AA |
262 | /* Check for wrap through zero too */ |
263 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 264 | return -EFBIG; |
97ac7350 | 265 | |
2fe17c10 | 266 | if (!file->f_op->fallocate) |
3e63cbb1 | 267 | return -EOPNOTSUPP; |
97ac7350 | 268 | |
2fe17c10 | 269 | return file->f_op->fallocate(file, mode, offset, len); |
3e63cbb1 AJ |
270 | } |
271 | ||
272 | SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len) | |
273 | { | |
274 | struct file *file; | |
275 | int error = -EBADF; | |
276 | ||
277 | file = fget(fd); | |
278 | if (file) { | |
279 | error = do_fallocate(file, mode, offset, len); | |
280 | fput(file); | |
281 | } | |
282 | ||
283 | return error; | |
97ac7350 | 284 | } |
3e63cbb1 | 285 | |
6673e0c3 HC |
286 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
287 | asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len) | |
288 | { | |
289 | return SYSC_fallocate((int)fd, (int)mode, offset, len); | |
290 | } | |
291 | SYSCALL_ALIAS(sys_fallocate, SyS_fallocate); | |
292 | #endif | |
97ac7350 | 293 | |
1da177e4 LT |
294 | /* |
295 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
296 | * We do this by temporarily clearing all FS-related capabilities and | |
297 | * switching the fsuid/fsgid around to the real ones. | |
298 | */ | |
6559eed8 | 299 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 300 | { |
d84f4f99 DH |
301 | const struct cred *old_cred; |
302 | struct cred *override_cred; | |
2d8f3038 | 303 | struct path path; |
256984a8 | 304 | struct inode *inode; |
1da177e4 LT |
305 | int res; |
306 | ||
307 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
308 | return -EINVAL; | |
309 | ||
d84f4f99 DH |
310 | override_cred = prepare_creds(); |
311 | if (!override_cred) | |
312 | return -ENOMEM; | |
1da177e4 | 313 | |
d84f4f99 DH |
314 | override_cred->fsuid = override_cred->uid; |
315 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 316 | |
086f7316 | 317 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 318 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
319 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
320 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 321 | cap_clear(override_cred->cap_effective); |
086f7316 | 322 | else |
d84f4f99 DH |
323 | override_cred->cap_effective = |
324 | override_cred->cap_permitted; | |
086f7316 | 325 | } |
1da177e4 | 326 | |
d84f4f99 DH |
327 | old_cred = override_creds(override_cred); |
328 | ||
2d8f3038 | 329 | res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); |
6902d925 DH |
330 | if (res) |
331 | goto out; | |
332 | ||
2d8f3038 | 333 | inode = path.dentry->d_inode; |
256984a8 AV |
334 | |
335 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
336 | /* |
337 | * MAY_EXEC on regular files is denied if the fs is mounted | |
338 | * with the "noexec" flag. | |
339 | */ | |
340 | res = -EACCES; | |
2d8f3038 | 341 | if (path.mnt->mnt_flags & MNT_NOEXEC) |
30524472 AV |
342 | goto out_path_release; |
343 | } | |
344 | ||
256984a8 | 345 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 346 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 347 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 348 | goto out_path_release; |
2f676cbc DH |
349 | /* |
350 | * This is a rare case where using __mnt_is_readonly() | |
351 | * is OK without a mnt_want/drop_write() pair. Since | |
352 | * no actual write to the fs is performed here, we do | |
353 | * not need to telegraph to that to anyone. | |
354 | * | |
355 | * By doing this, we accept that this access is | |
356 | * inherently racy and know that the fs may change | |
357 | * state before we even see this result. | |
358 | */ | |
2d8f3038 | 359 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 360 | res = -EROFS; |
1da177e4 | 361 | |
6902d925 | 362 | out_path_release: |
2d8f3038 | 363 | path_put(&path); |
6902d925 | 364 | out: |
d84f4f99 DH |
365 | revert_creds(old_cred); |
366 | put_cred(override_cred); | |
1da177e4 LT |
367 | return res; |
368 | } | |
369 | ||
ca013e94 | 370 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
371 | { |
372 | return sys_faccessat(AT_FDCWD, filename, mode); | |
373 | } | |
374 | ||
3cdad428 | 375 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 376 | { |
2d8f3038 | 377 | struct path path; |
1da177e4 LT |
378 | int error; |
379 | ||
2d8f3038 | 380 | error = user_path_dir(filename, &path); |
1da177e4 LT |
381 | if (error) |
382 | goto out; | |
383 | ||
9cfcac81 | 384 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
385 | if (error) |
386 | goto dput_and_out; | |
387 | ||
2d8f3038 | 388 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
389 | |
390 | dput_and_out: | |
2d8f3038 | 391 | path_put(&path); |
1da177e4 LT |
392 | out: |
393 | return error; | |
394 | } | |
395 | ||
3cdad428 | 396 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 LT |
397 | { |
398 | struct file *file; | |
1da177e4 | 399 | struct inode *inode; |
332a2e12 | 400 | int error, fput_needed; |
1da177e4 LT |
401 | |
402 | error = -EBADF; | |
332a2e12 | 403 | file = fget_raw_light(fd, &fput_needed); |
1da177e4 LT |
404 | if (!file) |
405 | goto out; | |
406 | ||
ac748a09 | 407 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
408 | |
409 | error = -ENOTDIR; | |
410 | if (!S_ISDIR(inode->i_mode)) | |
411 | goto out_putf; | |
412 | ||
9cfcac81 | 413 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 414 | if (!error) |
ac748a09 | 415 | set_fs_pwd(current->fs, &file->f_path); |
1da177e4 | 416 | out_putf: |
332a2e12 | 417 | fput_light(file, fput_needed); |
1da177e4 LT |
418 | out: |
419 | return error; | |
420 | } | |
421 | ||
3480b257 | 422 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 423 | { |
2d8f3038 | 424 | struct path path; |
1da177e4 LT |
425 | int error; |
426 | ||
2d8f3038 | 427 | error = user_path_dir(filename, &path); |
1da177e4 LT |
428 | if (error) |
429 | goto out; | |
430 | ||
9cfcac81 | 431 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
432 | if (error) |
433 | goto dput_and_out; | |
434 | ||
435 | error = -EPERM; | |
436 | if (!capable(CAP_SYS_CHROOT)) | |
437 | goto dput_and_out; | |
8b8efb44 TH |
438 | error = security_path_chroot(&path); |
439 | if (error) | |
440 | goto dput_and_out; | |
1da177e4 | 441 | |
2d8f3038 | 442 | set_fs_root(current->fs, &path); |
1da177e4 LT |
443 | error = 0; |
444 | dput_and_out: | |
2d8f3038 | 445 | path_put(&path); |
1da177e4 LT |
446 | out: |
447 | return error; | |
448 | } | |
449 | ||
e57712eb | 450 | static int chmod_common(struct path *path, umode_t mode) |
1da177e4 | 451 | { |
e57712eb | 452 | struct inode *inode = path->dentry->d_inode; |
1da177e4 | 453 | struct iattr newattrs; |
e57712eb | 454 | int error; |
1da177e4 | 455 | |
e57712eb AV |
456 | error = mnt_want_write(path->mnt); |
457 | if (error) | |
458 | return error; | |
fe542cf5 | 459 | mutex_lock(&inode->i_mutex); |
cdcf116d | 460 | error = security_path_chmod(path, mode); |
e57712eb | 461 | if (error) |
fe542cf5 | 462 | goto out_unlock; |
1da177e4 LT |
463 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
464 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
e57712eb | 465 | error = notify_change(path->dentry, &newattrs); |
fe542cf5 | 466 | out_unlock: |
1b1dcc1b | 467 | mutex_unlock(&inode->i_mutex); |
e57712eb AV |
468 | mnt_drop_write(path->mnt); |
469 | return error; | |
470 | } | |
471 | ||
49f0a076 | 472 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb AV |
473 | { |
474 | struct file * file; | |
475 | int err = -EBADF; | |
476 | ||
477 | file = fget(fd); | |
478 | if (file) { | |
479 | audit_inode(NULL, file->f_path.dentry); | |
480 | err = chmod_common(&file->f_path, mode); | |
481 | fput(file); | |
482 | } | |
1da177e4 LT |
483 | return err; |
484 | } | |
485 | ||
49f0a076 | 486 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 487 | { |
2d8f3038 | 488 | struct path path; |
1da177e4 | 489 | int error; |
1da177e4 | 490 | |
2d8f3038 | 491 | error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); |
e57712eb AV |
492 | if (!error) { |
493 | error = chmod_common(&path, mode); | |
494 | path_put(&path); | |
495 | } | |
1da177e4 LT |
496 | return error; |
497 | } | |
498 | ||
49f0a076 | 499 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
500 | { |
501 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
502 | } | |
503 | ||
fe542cf5 | 504 | static int chown_common(struct path *path, uid_t user, gid_t group) |
1da177e4 | 505 | { |
fe542cf5 | 506 | struct inode *inode = path->dentry->d_inode; |
1da177e4 LT |
507 | int error; |
508 | struct iattr newattrs; | |
52137abe EB |
509 | kuid_t uid; |
510 | kgid_t gid; | |
511 | ||
512 | uid = make_kuid(current_user_ns(), user); | |
513 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 514 | |
1da177e4 LT |
515 | newattrs.ia_valid = ATTR_CTIME; |
516 | if (user != (uid_t) -1) { | |
52137abe EB |
517 | if (!uid_valid(uid)) |
518 | return -EINVAL; | |
1da177e4 | 519 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 520 | newattrs.ia_uid = uid; |
1da177e4 LT |
521 | } |
522 | if (group != (gid_t) -1) { | |
52137abe EB |
523 | if (!gid_valid(gid)) |
524 | return -EINVAL; | |
1da177e4 | 525 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 526 | newattrs.ia_gid = gid; |
1da177e4 LT |
527 | } |
528 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
529 | newattrs.ia_valid |= |
530 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
1b1dcc1b | 531 | mutex_lock(&inode->i_mutex); |
fe542cf5 TH |
532 | error = security_path_chown(path, user, group); |
533 | if (!error) | |
534 | error = notify_change(path->dentry, &newattrs); | |
1b1dcc1b | 535 | mutex_unlock(&inode->i_mutex); |
beb29e05 | 536 | |
1da177e4 LT |
537 | return error; |
538 | } | |
539 | ||
6559eed8 HC |
540 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
541 | gid_t, group, int, flag) | |
5590ff0d | 542 | { |
2d8f3038 | 543 | struct path path; |
5590ff0d | 544 | int error = -EINVAL; |
65cfc672 | 545 | int lookup_flags; |
5590ff0d | 546 | |
65cfc672 | 547 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
548 | goto out; |
549 | ||
65cfc672 AV |
550 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
551 | if (flag & AT_EMPTY_PATH) | |
552 | lookup_flags |= LOOKUP_EMPTY; | |
553 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
554 | if (error) |
555 | goto out; | |
2d8f3038 | 556 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
557 | if (error) |
558 | goto out_release; | |
fe542cf5 | 559 | error = chown_common(&path, user, group); |
2d8f3038 | 560 | mnt_drop_write(path.mnt); |
2af482a7 | 561 | out_release: |
2d8f3038 | 562 | path_put(&path); |
5590ff0d UD |
563 | out: |
564 | return error; | |
565 | } | |
566 | ||
55e4def0 | 567 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 568 | { |
55e4def0 DH |
569 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
570 | } | |
1da177e4 | 571 | |
55e4def0 DH |
572 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
573 | { | |
574 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
575 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
576 | } |
577 | ||
ca013e94 | 578 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 LT |
579 | { |
580 | struct file * file; | |
581 | int error = -EBADF; | |
6902d925 | 582 | struct dentry * dentry; |
1da177e4 LT |
583 | |
584 | file = fget(fd); | |
6902d925 DH |
585 | if (!file) |
586 | goto out; | |
587 | ||
96029c4e | 588 | error = mnt_want_write_file(file); |
2af482a7 DH |
589 | if (error) |
590 | goto out_fput; | |
0f7fc9e4 | 591 | dentry = file->f_path.dentry; |
5a190ae6 | 592 | audit_inode(NULL, dentry); |
fe542cf5 | 593 | error = chown_common(&file->f_path, user, group); |
2a79f17e | 594 | mnt_drop_write_file(file); |
2af482a7 | 595 | out_fput: |
6902d925 DH |
596 | fput(file); |
597 | out: | |
1da177e4 LT |
598 | return error; |
599 | } | |
600 | ||
4a3fd211 DH |
601 | /* |
602 | * You have to be very careful that these write | |
603 | * counts get cleaned up in error cases and | |
604 | * upon __fput(). This should probably never | |
605 | * be called outside of __dentry_open(). | |
606 | */ | |
607 | static inline int __get_file_write_access(struct inode *inode, | |
608 | struct vfsmount *mnt) | |
609 | { | |
610 | int error; | |
611 | error = get_write_access(inode); | |
612 | if (error) | |
613 | return error; | |
614 | /* | |
615 | * Do not take mount writer counts on | |
616 | * special files since no writes to | |
617 | * the mount itself will occur. | |
618 | */ | |
619 | if (!special_file(inode->i_mode)) { | |
620 | /* | |
621 | * Balanced in __fput() | |
622 | */ | |
623 | error = mnt_want_write(mnt); | |
624 | if (error) | |
625 | put_write_access(inode); | |
626 | } | |
627 | return error; | |
628 | } | |
629 | ||
90ad1a8e MS |
630 | int open_check_o_direct(struct file *f) |
631 | { | |
632 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
633 | if (f->f_flags & O_DIRECT) { | |
634 | if (!f->f_mapping->a_ops || | |
635 | ((!f->f_mapping->a_ops->direct_IO) && | |
636 | (!f->f_mapping->a_ops->get_xip_mem))) { | |
637 | return -EINVAL; | |
638 | } | |
639 | } | |
640 | return 0; | |
641 | } | |
642 | ||
02e5180d | 643 | static int do_dentry_open(struct file *f, |
96b7e579 AV |
644 | int (*open)(struct inode *, struct file *), |
645 | const struct cred *cred) | |
1da177e4 | 646 | { |
1abf0c71 | 647 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
648 | struct inode *inode; |
649 | int error; | |
650 | ||
5300990c | 651 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 652 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 AV |
653 | |
654 | if (unlikely(f->f_flags & O_PATH)) | |
655 | f->f_mode = FMODE_PATH; | |
656 | ||
02e5180d | 657 | inode = f->f_path.dentry->d_inode; |
1da177e4 | 658 | if (f->f_mode & FMODE_WRITE) { |
02e5180d | 659 | error = __get_file_write_access(inode, f->f_path.mnt); |
1da177e4 LT |
660 | if (error) |
661 | goto cleanup_file; | |
ad775f5a DH |
662 | if (!special_file(inode->i_mode)) |
663 | file_take_write(f); | |
1da177e4 LT |
664 | } |
665 | ||
666 | f->f_mapping = inode->i_mapping; | |
1da177e4 | 667 | f->f_pos = 0; |
ee2ffa0d | 668 | file_sb_list_add(f, inode->i_sb); |
1da177e4 | 669 | |
1abf0c71 AV |
670 | if (unlikely(f->f_mode & FMODE_PATH)) { |
671 | f->f_op = &empty_fops; | |
96b7e579 | 672 | return 0; |
1abf0c71 AV |
673 | } |
674 | ||
675 | f->f_op = fops_get(inode->i_fop); | |
676 | ||
83d49856 | 677 | error = security_file_open(f, cred); |
788e7dd4 YN |
678 | if (error) |
679 | goto cleanup_all; | |
680 | ||
f3c7691e BF |
681 | error = break_lease(inode, f->f_flags); |
682 | if (error) | |
683 | goto cleanup_all; | |
684 | ||
834f2a4a TM |
685 | if (!open && f->f_op) |
686 | open = f->f_op->open; | |
687 | if (open) { | |
688 | error = open(inode, f); | |
1da177e4 LT |
689 | if (error) |
690 | goto cleanup_all; | |
691 | } | |
890275b5 MZ |
692 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
693 | i_readcount_inc(inode); | |
834f2a4a | 694 | |
1da177e4 LT |
695 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
696 | ||
697 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
698 | ||
96b7e579 | 699 | return 0; |
1da177e4 LT |
700 | |
701 | cleanup_all: | |
702 | fops_put(f->f_op); | |
c3c4f694 | 703 | file_sb_list_del(f); |
4a3fd211 | 704 | if (f->f_mode & FMODE_WRITE) { |
1da177e4 | 705 | put_write_access(inode); |
ad775f5a DH |
706 | if (!special_file(inode->i_mode)) { |
707 | /* | |
708 | * We don't consider this a real | |
709 | * mnt_want/drop_write() pair | |
710 | * because it all happenend right | |
711 | * here, so just reset the state. | |
712 | */ | |
713 | file_reset_write(f); | |
02e5180d | 714 | mnt_drop_write(f->f_path.mnt); |
ad775f5a | 715 | } |
4a3fd211 | 716 | } |
1da177e4 | 717 | cleanup_file: |
02e5180d AV |
718 | path_put(&f->f_path); |
719 | f->f_path.mnt = NULL; | |
720 | f->f_path.dentry = NULL; | |
96b7e579 | 721 | return error; |
1da177e4 LT |
722 | } |
723 | ||
d18e9008 MS |
724 | /** |
725 | * finish_open - finish opening a file | |
726 | * @od: opaque open data | |
727 | * @dentry: pointer to dentry | |
728 | * @open: open callback | |
729 | * | |
730 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
731 | * | |
732 | * If the open callback is set to NULL, then the standard f_op->open() | |
733 | * filesystem callback is substituted. | |
734 | */ | |
30d90494 AV |
735 | int finish_open(struct file *file, struct dentry *dentry, |
736 | int (*open)(struct inode *, struct file *), | |
737 | int *opened) | |
d18e9008 | 738 | { |
96b7e579 | 739 | int error; |
3d8a00d2 | 740 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 741 | |
30d90494 | 742 | mntget(file->f_path.mnt); |
02e5180d | 743 | file->f_path.dentry = dget(dentry); |
d18e9008 | 744 | |
02e5180d | 745 | error = do_dentry_open(file, open, current_cred()); |
96b7e579 | 746 | if (!error) |
47237687 | 747 | *opened |= FILE_OPENED; |
d18e9008 | 748 | |
96b7e579 | 749 | return error; |
d18e9008 MS |
750 | } |
751 | EXPORT_SYMBOL(finish_open); | |
752 | ||
753 | /** | |
754 | * finish_no_open - finish ->atomic_open() without opening the file | |
755 | * | |
756 | * @od: opaque open data | |
757 | * @dentry: dentry or NULL (as returned from ->lookup()) | |
758 | * | |
759 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
760 | * The filesystem's atomic_open() method shall return NULL after calling this. | |
761 | */ | |
e45198a6 | 762 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 763 | { |
30d90494 | 764 | file->f_path.dentry = dentry; |
e45198a6 | 765 | return 1; |
d18e9008 MS |
766 | } |
767 | EXPORT_SYMBOL(finish_no_open); | |
768 | ||
765927b2 | 769 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 770 | const struct cred *cred) |
a1a5b3d9 PS |
771 | { |
772 | int error; | |
773 | struct file *f; | |
774 | ||
e0e81739 DH |
775 | validate_creds(cred); |
776 | ||
c212f9aa | 777 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 778 | BUG_ON(!path->mnt); |
322ee5b3 | 779 | |
a1a5b3d9 PS |
780 | error = -ENFILE; |
781 | f = get_empty_filp(); | |
765927b2 | 782 | if (f == NULL) |
a1a5b3d9 PS |
783 | return ERR_PTR(error); |
784 | ||
482928d5 | 785 | f->f_flags = flags; |
765927b2 AV |
786 | f->f_path = *path; |
787 | path_get(&f->f_path); | |
02e5180d | 788 | error = do_dentry_open(f, NULL, cred); |
2a027e7a AV |
789 | if (!error) { |
790 | error = open_check_o_direct(f); | |
791 | if (error) { | |
792 | fput(f); | |
793 | f = ERR_PTR(error); | |
794 | } | |
795 | } else { | |
796 | put_filp(f); | |
797 | f = ERR_PTR(error); | |
798 | } | |
799 | return f; | |
a1a5b3d9 | 800 | } |
1da177e4 LT |
801 | EXPORT_SYMBOL(dentry_open); |
802 | ||
b01ec0ef | 803 | static void __put_unused_fd(struct files_struct *files, unsigned int fd) |
1da177e4 | 804 | { |
badf1662 | 805 | struct fdtable *fdt = files_fdtable(files); |
1dce27c5 | 806 | __clear_open_fd(fd, fdt); |
0c9e63fd ED |
807 | if (fd < files->next_fd) |
808 | files->next_fd = fd; | |
1da177e4 LT |
809 | } |
810 | ||
fc9b52cd | 811 | void put_unused_fd(unsigned int fd) |
1da177e4 LT |
812 | { |
813 | struct files_struct *files = current->files; | |
814 | spin_lock(&files->file_lock); | |
815 | __put_unused_fd(files, fd); | |
816 | spin_unlock(&files->file_lock); | |
817 | } | |
818 | ||
819 | EXPORT_SYMBOL(put_unused_fd); | |
820 | ||
821 | /* | |
5590ff0d | 822 | * Install a file pointer in the fd array. |
1da177e4 LT |
823 | * |
824 | * The VFS is full of places where we drop the files lock between | |
825 | * setting the open_fds bitmap and installing the file in the file | |
826 | * array. At any such point, we are vulnerable to a dup2() race | |
827 | * installing a file in the array before us. We need to detect this and | |
828 | * fput() the struct file we are about to overwrite in this case. | |
829 | * | |
830 | * It should never happen - if we allow dup2() do it, _really_ bad things | |
831 | * will follow. | |
832 | */ | |
833 | ||
fc9b52cd | 834 | void fd_install(unsigned int fd, struct file *file) |
1da177e4 LT |
835 | { |
836 | struct files_struct *files = current->files; | |
badf1662 | 837 | struct fdtable *fdt; |
1da177e4 | 838 | spin_lock(&files->file_lock); |
badf1662 | 839 | fdt = files_fdtable(files); |
ab2af1f5 DS |
840 | BUG_ON(fdt->fd[fd] != NULL); |
841 | rcu_assign_pointer(fdt->fd[fd], file); | |
1da177e4 LT |
842 | spin_unlock(&files->file_lock); |
843 | } | |
844 | ||
845 | EXPORT_SYMBOL(fd_install); | |
846 | ||
a218d0fd | 847 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
848 | { |
849 | int lookup_flags = 0; | |
850 | int acc_mode; | |
851 | ||
852 | if (!(flags & O_CREAT)) | |
853 | mode = 0; | |
854 | op->mode = mode; | |
855 | ||
856 | /* Must never be set by userspace */ | |
857 | flags &= ~FMODE_NONOTIFY; | |
858 | ||
859 | /* | |
860 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
861 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
862 | * always set instead of having to deal with possibly weird behaviour | |
863 | * for malicious applications setting only __O_SYNC. | |
864 | */ | |
865 | if (flags & __O_SYNC) | |
866 | flags |= O_DSYNC; | |
867 | ||
1abf0c71 AV |
868 | /* |
869 | * If we have O_PATH in the open flag. Then we | |
870 | * cannot have anything other than the below set of flags | |
871 | */ | |
872 | if (flags & O_PATH) { | |
873 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; | |
874 | acc_mode = 0; | |
875 | } else { | |
876 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
877 | } | |
47c805dc | 878 | |
1abf0c71 | 879 | op->open_flag = flags; |
47c805dc AV |
880 | |
881 | /* O_TRUNC implies we need access checks for write permissions */ | |
882 | if (flags & O_TRUNC) | |
883 | acc_mode |= MAY_WRITE; | |
884 | ||
885 | /* Allow the LSM permission hook to distinguish append | |
886 | access from general write access. */ | |
887 | if (flags & O_APPEND) | |
888 | acc_mode |= MAY_APPEND; | |
889 | ||
890 | op->acc_mode = acc_mode; | |
891 | ||
1abf0c71 AV |
892 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
893 | ||
47c805dc AV |
894 | if (flags & O_CREAT) { |
895 | op->intent |= LOOKUP_CREATE; | |
896 | if (flags & O_EXCL) | |
897 | op->intent |= LOOKUP_EXCL; | |
898 | } | |
899 | ||
900 | if (flags & O_DIRECTORY) | |
901 | lookup_flags |= LOOKUP_DIRECTORY; | |
902 | if (!(flags & O_NOFOLLOW)) | |
903 | lookup_flags |= LOOKUP_FOLLOW; | |
904 | return lookup_flags; | |
905 | } | |
906 | ||
907 | /** | |
908 | * filp_open - open file and return file pointer | |
909 | * | |
910 | * @filename: path to open | |
911 | * @flags: open flags as per the open(2) second argument | |
912 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
913 | * | |
914 | * This is the helper to open a file from kernelspace if you really | |
915 | * have to. But in generally you should not do this, so please move | |
916 | * along, nothing to see here.. | |
917 | */ | |
a218d0fd | 918 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc AV |
919 | { |
920 | struct open_flags op; | |
921 | int lookup = build_open_flags(flags, mode, &op); | |
922 | return do_filp_open(AT_FDCWD, filename, &op, lookup); | |
923 | } | |
924 | EXPORT_SYMBOL(filp_open); | |
925 | ||
73d049a4 AV |
926 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
927 | const char *filename, int flags) | |
928 | { | |
929 | struct open_flags op; | |
930 | int lookup = build_open_flags(flags, 0, &op); | |
931 | if (flags & O_CREAT) | |
932 | return ERR_PTR(-EINVAL); | |
933 | if (!filename && (flags & O_DIRECTORY)) | |
934 | if (!dentry->d_inode->i_op->lookup) | |
935 | return ERR_PTR(-ENOTDIR); | |
936 | return do_file_open_root(dentry, mnt, filename, &op, lookup); | |
937 | } | |
938 | EXPORT_SYMBOL(file_open_root); | |
939 | ||
a218d0fd | 940 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 941 | { |
47c805dc AV |
942 | struct open_flags op; |
943 | int lookup = build_open_flags(flags, mode, &op); | |
e922efc3 MS |
944 | char *tmp = getname(filename); |
945 | int fd = PTR_ERR(tmp); | |
1da177e4 | 946 | |
1da177e4 | 947 | if (!IS_ERR(tmp)) { |
f23513e8 | 948 | fd = get_unused_fd_flags(flags); |
1da177e4 | 949 | if (fd >= 0) { |
47c805dc | 950 | struct file *f = do_filp_open(dfd, tmp, &op, lookup); |
fed2fc18 TN |
951 | if (IS_ERR(f)) { |
952 | put_unused_fd(fd); | |
953 | fd = PTR_ERR(f); | |
954 | } else { | |
2a12a9d7 | 955 | fsnotify_open(f); |
fed2fc18 TN |
956 | fd_install(fd, f); |
957 | } | |
1da177e4 | 958 | } |
1da177e4 LT |
959 | putname(tmp); |
960 | } | |
961 | return fd; | |
1da177e4 | 962 | } |
e922efc3 | 963 | |
a218d0fd | 964 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 | 965 | { |
385910f2 LT |
966 | long ret; |
967 | ||
e922efc3 MS |
968 | if (force_o_largefile()) |
969 | flags |= O_LARGEFILE; | |
970 | ||
385910f2 LT |
971 | ret = do_sys_open(AT_FDCWD, filename, flags, mode); |
972 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 973 | asmlinkage_protect(3, ret, filename, flags, mode); |
385910f2 | 974 | return ret; |
e922efc3 | 975 | } |
1da177e4 | 976 | |
6559eed8 | 977 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 978 | umode_t, mode) |
5590ff0d | 979 | { |
385910f2 LT |
980 | long ret; |
981 | ||
5590ff0d UD |
982 | if (force_o_largefile()) |
983 | flags |= O_LARGEFILE; | |
984 | ||
385910f2 LT |
985 | ret = do_sys_open(dfd, filename, flags, mode); |
986 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 987 | asmlinkage_protect(4, ret, dfd, filename, flags, mode); |
385910f2 | 988 | return ret; |
5590ff0d | 989 | } |
5590ff0d | 990 | |
1da177e4 LT |
991 | #ifndef __alpha__ |
992 | ||
993 | /* | |
994 | * For backward compatibility? Maybe this should be moved | |
995 | * into arch/i386 instead? | |
996 | */ | |
a218d0fd | 997 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
998 | { |
999 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1000 | } | |
1001 | ||
1002 | #endif | |
1003 | ||
1004 | /* | |
1005 | * "id" is the POSIX thread ID. We use the | |
1006 | * files pointer for this.. | |
1007 | */ | |
1008 | int filp_close(struct file *filp, fl_owner_t id) | |
1009 | { | |
45778ca8 | 1010 | int retval = 0; |
1da177e4 LT |
1011 | |
1012 | if (!file_count(filp)) { | |
1013 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1014 | return 0; |
1da177e4 LT |
1015 | } |
1016 | ||
45778ca8 | 1017 | if (filp->f_op && filp->f_op->flush) |
75e1fcc0 | 1018 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1019 | |
1abf0c71 AV |
1020 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1021 | dnotify_flush(filp, id); | |
1022 | locks_remove_posix(filp, id); | |
1023 | } | |
1da177e4 LT |
1024 | fput(filp); |
1025 | return retval; | |
1026 | } | |
1027 | ||
1028 | EXPORT_SYMBOL(filp_close); | |
1029 | ||
1030 | /* | |
1031 | * Careful here! We test whether the file pointer is NULL before | |
1032 | * releasing the fd. This ensures that one clone task can't release | |
1033 | * an fd while another clone is opening it. | |
1034 | */ | |
ca013e94 | 1035 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 LT |
1036 | { |
1037 | struct file * filp; | |
1038 | struct files_struct *files = current->files; | |
badf1662 | 1039 | struct fdtable *fdt; |
ee731f4f | 1040 | int retval; |
1da177e4 LT |
1041 | |
1042 | spin_lock(&files->file_lock); | |
badf1662 DS |
1043 | fdt = files_fdtable(files); |
1044 | if (fd >= fdt->max_fds) | |
1da177e4 | 1045 | goto out_unlock; |
badf1662 | 1046 | filp = fdt->fd[fd]; |
1da177e4 LT |
1047 | if (!filp) |
1048 | goto out_unlock; | |
ab2af1f5 | 1049 | rcu_assign_pointer(fdt->fd[fd], NULL); |
1dce27c5 | 1050 | __clear_close_on_exec(fd, fdt); |
1da177e4 LT |
1051 | __put_unused_fd(files, fd); |
1052 | spin_unlock(&files->file_lock); | |
ee731f4f EP |
1053 | retval = filp_close(filp, files); |
1054 | ||
1055 | /* can't restart close syscall because file table entry was cleared */ | |
1056 | if (unlikely(retval == -ERESTARTSYS || | |
1057 | retval == -ERESTARTNOINTR || | |
1058 | retval == -ERESTARTNOHAND || | |
1059 | retval == -ERESTART_RESTARTBLOCK)) | |
1060 | retval = -EINTR; | |
1061 | ||
1062 | return retval; | |
1da177e4 LT |
1063 | |
1064 | out_unlock: | |
1065 | spin_unlock(&files->file_lock); | |
1066 | return -EBADF; | |
1067 | } | |
1da177e4 LT |
1068 | EXPORT_SYMBOL(sys_close); |
1069 | ||
1070 | /* | |
1071 | * This routine simulates a hangup on the tty, to arrange that users | |
1072 | * are given clean terminals at login time. | |
1073 | */ | |
ca013e94 | 1074 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1075 | { |
1076 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1077 | tty_vhangup_self(); |
1da177e4 LT |
1078 | return 0; |
1079 | } | |
1080 | return -EPERM; | |
1081 | } | |
1082 | ||
1083 | /* | |
1084 | * Called when an inode is about to be open. | |
1085 | * We use this to disallow opening large files on 32bit systems if | |
1086 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1087 | * on this flag in sys_open. | |
1088 | */ | |
1089 | int generic_file_open(struct inode * inode, struct file * filp) | |
1090 | { | |
1091 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1092 | return -EOVERFLOW; |
1da177e4 LT |
1093 | return 0; |
1094 | } | |
1095 | ||
1096 | EXPORT_SYMBOL(generic_file_open); | |
1097 | ||
1098 | /* | |
1099 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1100 | * file descriptors. The function is not supposed to ever fail, the only |
1101 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1102 | * directly into file_operations structure. | |
1da177e4 LT |
1103 | */ |
1104 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1105 | { | |
1106 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1107 | return 0; | |
1108 | } | |
1109 | ||
1110 | EXPORT_SYMBOL(nonseekable_open); |