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> |
3f6d078d | 33 | #include <linux/compat.h> |
1da177e4 | 34 | |
e81e3f4d EP |
35 | #include "internal.h" |
36 | ||
4a30131e N |
37 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
38 | struct file *filp) | |
1da177e4 | 39 | { |
939a9421 | 40 | int ret; |
1da177e4 LT |
41 | 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; | |
4a30131e | 48 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
49 | if (filp) { |
50 | newattrs.ia_file = filp; | |
51 | newattrs.ia_valid |= ATTR_FILE; | |
52 | } | |
1da177e4 | 53 | |
7b82dc0e | 54 | /* Remove suid/sgid on truncate too */ |
939a9421 AW |
55 | ret = should_remove_suid(dentry); |
56 | if (ret) | |
57 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 58 | |
1b1dcc1b | 59 | mutex_lock(&dentry->d_inode->i_mutex); |
27ac0ffe BF |
60 | /* Note any delegations or leases have already been broken: */ |
61 | ret = notify_change(dentry, &newattrs, NULL); | |
1b1dcc1b | 62 | mutex_unlock(&dentry->d_inode->i_mutex); |
939a9421 | 63 | return ret; |
1da177e4 LT |
64 | } |
65 | ||
a02de960 | 66 | long vfs_truncate(struct path *path, loff_t length) |
1da177e4 | 67 | { |
2d8f3038 | 68 | struct inode *inode; |
a02de960 | 69 | long error; |
1da177e4 | 70 | |
a02de960 | 71 | inode = path->dentry->d_inode; |
1da177e4 LT |
72 | |
73 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 74 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 75 | return -EISDIR; |
1da177e4 | 76 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 77 | return -EINVAL; |
1da177e4 | 78 | |
a02de960 | 79 | error = mnt_want_write(path->mnt); |
1da177e4 | 80 | if (error) |
a02de960 | 81 | goto out; |
1da177e4 | 82 | |
256984a8 | 83 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
84 | if (error) |
85 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
86 | |
87 | error = -EPERM; | |
c82e42da | 88 | if (IS_APPEND(inode)) |
9ac9b847 | 89 | goto mnt_drop_write_and_out; |
1da177e4 | 90 | |
9700382c | 91 | error = get_write_access(inode); |
1da177e4 | 92 | if (error) |
9ac9b847 | 93 | goto mnt_drop_write_and_out; |
1da177e4 | 94 | |
9700382c | 95 | /* |
96 | * Make sure that there are no leases. get_write_access() protects | |
97 | * against the truncate racing with a lease-granting setlease(). | |
98 | */ | |
8737c930 | 99 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 100 | if (error) |
9700382c | 101 | goto put_write_and_out; |
1da177e4 LT |
102 | |
103 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 104 | if (!error) |
a02de960 | 105 | error = security_path_truncate(path); |
907f4554 | 106 | if (!error) |
a02de960 | 107 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 108 | |
9700382c | 109 | put_write_and_out: |
110 | put_write_access(inode); | |
9ac9b847 | 111 | mnt_drop_write_and_out: |
a02de960 | 112 | mnt_drop_write(path->mnt); |
1da177e4 LT |
113 | out: |
114 | return error; | |
115 | } | |
a02de960 DH |
116 | EXPORT_SYMBOL_GPL(vfs_truncate); |
117 | ||
118 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
119 | { | |
48f7530d | 120 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
121 | struct path path; |
122 | int error; | |
123 | ||
124 | if (length < 0) /* sorry, but loff_t says... */ | |
125 | return -EINVAL; | |
126 | ||
48f7530d JL |
127 | retry: |
128 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
129 | if (!error) { |
130 | error = vfs_truncate(&path, length); | |
131 | path_put(&path); | |
132 | } | |
48f7530d JL |
133 | if (retry_estale(error, lookup_flags)) { |
134 | lookup_flags |= LOOKUP_REVAL; | |
135 | goto retry; | |
136 | } | |
a02de960 DH |
137 | return error; |
138 | } | |
1da177e4 | 139 | |
4fd8da8d | 140 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 141 | { |
4fd8da8d | 142 | return do_sys_truncate(path, length); |
1da177e4 LT |
143 | } |
144 | ||
3f6d078d AV |
145 | #ifdef CONFIG_COMPAT |
146 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
147 | { | |
148 | return do_sys_truncate(path, length); | |
149 | } | |
150 | #endif | |
151 | ||
b01ec0ef | 152 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 153 | { |
bf2965d5 | 154 | struct inode *inode; |
1da177e4 | 155 | struct dentry *dentry; |
2903ff01 | 156 | struct fd f; |
1da177e4 LT |
157 | int error; |
158 | ||
159 | error = -EINVAL; | |
160 | if (length < 0) | |
161 | goto out; | |
162 | error = -EBADF; | |
2903ff01 AV |
163 | f = fdget(fd); |
164 | if (!f.file) | |
1da177e4 LT |
165 | goto out; |
166 | ||
167 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 168 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
169 | small = 0; |
170 | ||
2903ff01 | 171 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
172 | inode = dentry->d_inode; |
173 | error = -EINVAL; | |
2903ff01 | 174 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
175 | goto out_putf; |
176 | ||
177 | error = -EINVAL; | |
178 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
179 | if (small && length > MAX_NON_LFS) | |
180 | goto out_putf; | |
181 | ||
182 | error = -EPERM; | |
183 | if (IS_APPEND(inode)) | |
184 | goto out_putf; | |
185 | ||
14da9200 | 186 | sb_start_write(inode->i_sb); |
2903ff01 | 187 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 188 | if (!error) |
2903ff01 | 189 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 190 | if (!error) |
2903ff01 | 191 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 192 | sb_end_write(inode->i_sb); |
1da177e4 | 193 | out_putf: |
2903ff01 | 194 | fdput(f); |
1da177e4 LT |
195 | out: |
196 | return error; | |
197 | } | |
198 | ||
bdc480e3 | 199 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 200 | { |
2cf09666 | 201 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
202 | } |
203 | ||
3f6d078d AV |
204 | #ifdef CONFIG_COMPAT |
205 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
206 | { | |
207 | return do_sys_ftruncate(fd, length, 1); | |
208 | } | |
209 | #endif | |
210 | ||
1da177e4 LT |
211 | /* LFS versions of truncate are only needed on 32 bit machines */ |
212 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 213 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
214 | { |
215 | return do_sys_truncate(path, length); | |
216 | } | |
217 | ||
4a0fd5bf | 218 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 219 | { |
2cf09666 | 220 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 221 | } |
6673e0c3 | 222 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 223 | |
3e63cbb1 AJ |
224 | |
225 | int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
97ac7350 | 226 | { |
496ad9aa | 227 | struct inode *inode = file_inode(file); |
3e63cbb1 | 228 | long ret; |
97ac7350 AA |
229 | |
230 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 231 | return -EINVAL; |
97ac7350 AA |
232 | |
233 | /* Return error if mode is not supported */ | |
00f5e619 | 234 | if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | |
409332b6 LC |
235 | FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE)) |
236 | return -EOPNOTSUPP; | |
237 | ||
238 | /* Punch hole and zero range are mutually exclusive */ | |
239 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
240 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
241 | return -EOPNOTSUPP; |
242 | ||
243 | /* Punch hole must have keep size set */ | |
244 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
245 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 246 | return -EOPNOTSUPP; |
97ac7350 | 247 | |
00f5e619 NJ |
248 | /* Collapse range should only be used exclusively. */ |
249 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
250 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
251 | return -EINVAL; | |
252 | ||
97ac7350 | 253 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 254 | return -EBADF; |
1ca551c6 | 255 | |
00f5e619 | 256 | /* |
8fc61d92 | 257 | * We can only allow pure fallocate on append only files |
00f5e619 | 258 | */ |
8fc61d92 | 259 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
260 | return -EPERM; |
261 | ||
262 | if (IS_IMMUTABLE(inode)) | |
263 | return -EPERM; | |
264 | ||
97ac7350 AA |
265 | /* |
266 | * Revalidate the write permissions, in case security policy has | |
267 | * changed since the files were opened. | |
268 | */ | |
269 | ret = security_file_permission(file, MAY_WRITE); | |
270 | if (ret) | |
3e63cbb1 | 271 | return ret; |
97ac7350 | 272 | |
97ac7350 | 273 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 274 | return -ESPIPE; |
97ac7350 | 275 | |
97ac7350 AA |
276 | /* |
277 | * Let individual file system decide if it supports preallocation | |
278 | * for directories or not. | |
279 | */ | |
280 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 281 | return -ENODEV; |
97ac7350 | 282 | |
97ac7350 AA |
283 | /* Check for wrap through zero too */ |
284 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 285 | return -EFBIG; |
97ac7350 | 286 | |
00f5e619 NJ |
287 | /* |
288 | * There is no need to overlap collapse range with EOF, in which case | |
289 | * it is effectively a truncate operation | |
290 | */ | |
291 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
292 | (offset + len >= i_size_read(inode))) | |
293 | return -EINVAL; | |
294 | ||
2fe17c10 | 295 | if (!file->f_op->fallocate) |
3e63cbb1 | 296 | return -EOPNOTSUPP; |
97ac7350 | 297 | |
14da9200 JK |
298 | sb_start_write(inode->i_sb); |
299 | ret = file->f_op->fallocate(file, mode, offset, len); | |
300 | sb_end_write(inode->i_sb); | |
301 | return ret; | |
3e63cbb1 AJ |
302 | } |
303 | ||
4a0fd5bf | 304 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 305 | { |
2903ff01 | 306 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
307 | int error = -EBADF; |
308 | ||
2903ff01 AV |
309 | if (f.file) { |
310 | error = do_fallocate(f.file, mode, offset, len); | |
311 | fdput(f); | |
3e63cbb1 | 312 | } |
3e63cbb1 | 313 | return error; |
97ac7350 | 314 | } |
3e63cbb1 | 315 | |
1da177e4 LT |
316 | /* |
317 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
318 | * We do this by temporarily clearing all FS-related capabilities and | |
319 | * switching the fsuid/fsgid around to the real ones. | |
320 | */ | |
6559eed8 | 321 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 322 | { |
d84f4f99 DH |
323 | const struct cred *old_cred; |
324 | struct cred *override_cred; | |
2d8f3038 | 325 | struct path path; |
256984a8 | 326 | struct inode *inode; |
1da177e4 | 327 | int res; |
87fa5595 | 328 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
329 | |
330 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
331 | return -EINVAL; | |
332 | ||
d84f4f99 DH |
333 | override_cred = prepare_creds(); |
334 | if (!override_cred) | |
335 | return -ENOMEM; | |
1da177e4 | 336 | |
d84f4f99 DH |
337 | override_cred->fsuid = override_cred->uid; |
338 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 339 | |
086f7316 | 340 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 341 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
342 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
343 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 344 | cap_clear(override_cred->cap_effective); |
086f7316 | 345 | else |
d84f4f99 DH |
346 | override_cred->cap_effective = |
347 | override_cred->cap_permitted; | |
086f7316 | 348 | } |
1da177e4 | 349 | |
d84f4f99 | 350 | old_cred = override_creds(override_cred); |
87fa5595 JL |
351 | retry: |
352 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
353 | if (res) |
354 | goto out; | |
355 | ||
2d8f3038 | 356 | inode = path.dentry->d_inode; |
256984a8 AV |
357 | |
358 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
359 | /* |
360 | * MAY_EXEC on regular files is denied if the fs is mounted | |
361 | * with the "noexec" flag. | |
362 | */ | |
363 | res = -EACCES; | |
2d8f3038 | 364 | if (path.mnt->mnt_flags & MNT_NOEXEC) |
30524472 AV |
365 | goto out_path_release; |
366 | } | |
367 | ||
256984a8 | 368 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 369 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 370 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 371 | goto out_path_release; |
2f676cbc DH |
372 | /* |
373 | * This is a rare case where using __mnt_is_readonly() | |
374 | * is OK without a mnt_want/drop_write() pair. Since | |
375 | * no actual write to the fs is performed here, we do | |
376 | * not need to telegraph to that to anyone. | |
377 | * | |
378 | * By doing this, we accept that this access is | |
379 | * inherently racy and know that the fs may change | |
380 | * state before we even see this result. | |
381 | */ | |
2d8f3038 | 382 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 383 | res = -EROFS; |
1da177e4 | 384 | |
6902d925 | 385 | out_path_release: |
2d8f3038 | 386 | path_put(&path); |
87fa5595 JL |
387 | if (retry_estale(res, lookup_flags)) { |
388 | lookup_flags |= LOOKUP_REVAL; | |
389 | goto retry; | |
390 | } | |
6902d925 | 391 | out: |
d84f4f99 DH |
392 | revert_creds(old_cred); |
393 | put_cred(override_cred); | |
1da177e4 LT |
394 | return res; |
395 | } | |
396 | ||
ca013e94 | 397 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
398 | { |
399 | return sys_faccessat(AT_FDCWD, filename, mode); | |
400 | } | |
401 | ||
3cdad428 | 402 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 403 | { |
2d8f3038 | 404 | struct path path; |
1da177e4 | 405 | int error; |
0291c0a5 JL |
406 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
407 | retry: | |
408 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
409 | if (error) |
410 | goto out; | |
411 | ||
9cfcac81 | 412 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
413 | if (error) |
414 | goto dput_and_out; | |
415 | ||
2d8f3038 | 416 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
417 | |
418 | dput_and_out: | |
2d8f3038 | 419 | path_put(&path); |
0291c0a5 JL |
420 | if (retry_estale(error, lookup_flags)) { |
421 | lookup_flags |= LOOKUP_REVAL; | |
422 | goto retry; | |
423 | } | |
1da177e4 LT |
424 | out: |
425 | return error; | |
426 | } | |
427 | ||
3cdad428 | 428 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 429 | { |
2903ff01 | 430 | struct fd f = fdget_raw(fd); |
1da177e4 | 431 | struct inode *inode; |
2903ff01 | 432 | int error = -EBADF; |
1da177e4 LT |
433 | |
434 | error = -EBADF; | |
2903ff01 | 435 | if (!f.file) |
1da177e4 LT |
436 | goto out; |
437 | ||
496ad9aa | 438 | inode = file_inode(f.file); |
1da177e4 LT |
439 | |
440 | error = -ENOTDIR; | |
441 | if (!S_ISDIR(inode->i_mode)) | |
442 | goto out_putf; | |
443 | ||
9cfcac81 | 444 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 445 | if (!error) |
2903ff01 | 446 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 447 | out_putf: |
2903ff01 | 448 | fdput(f); |
1da177e4 LT |
449 | out: |
450 | return error; | |
451 | } | |
452 | ||
3480b257 | 453 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 454 | { |
2d8f3038 | 455 | struct path path; |
1da177e4 | 456 | int error; |
2771261e JL |
457 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
458 | retry: | |
459 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
460 | if (error) |
461 | goto out; | |
462 | ||
9cfcac81 | 463 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
464 | if (error) |
465 | goto dput_and_out; | |
466 | ||
467 | error = -EPERM; | |
c7b96acf | 468 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 469 | goto dput_and_out; |
8b8efb44 TH |
470 | error = security_path_chroot(&path); |
471 | if (error) | |
472 | goto dput_and_out; | |
1da177e4 | 473 | |
2d8f3038 | 474 | set_fs_root(current->fs, &path); |
1da177e4 LT |
475 | error = 0; |
476 | dput_and_out: | |
2d8f3038 | 477 | path_put(&path); |
2771261e JL |
478 | if (retry_estale(error, lookup_flags)) { |
479 | lookup_flags |= LOOKUP_REVAL; | |
480 | goto retry; | |
481 | } | |
1da177e4 LT |
482 | out: |
483 | return error; | |
484 | } | |
485 | ||
e57712eb | 486 | static int chmod_common(struct path *path, umode_t mode) |
1da177e4 | 487 | { |
e57712eb | 488 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 489 | struct inode *delegated_inode = NULL; |
1da177e4 | 490 | struct iattr newattrs; |
e57712eb | 491 | int error; |
1da177e4 | 492 | |
e57712eb AV |
493 | error = mnt_want_write(path->mnt); |
494 | if (error) | |
495 | return error; | |
27ac0ffe | 496 | retry_deleg: |
fe542cf5 | 497 | mutex_lock(&inode->i_mutex); |
cdcf116d | 498 | error = security_path_chmod(path, mode); |
e57712eb | 499 | if (error) |
fe542cf5 | 500 | goto out_unlock; |
1da177e4 LT |
501 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
502 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 503 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 504 | out_unlock: |
1b1dcc1b | 505 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
506 | if (delegated_inode) { |
507 | error = break_deleg_wait(&delegated_inode); | |
508 | if (!error) | |
509 | goto retry_deleg; | |
510 | } | |
e57712eb AV |
511 | mnt_drop_write(path->mnt); |
512 | return error; | |
513 | } | |
514 | ||
49f0a076 | 515 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 516 | { |
173c8401 | 517 | struct fd f = fdget(fd); |
e57712eb AV |
518 | int err = -EBADF; |
519 | ||
173c8401 AV |
520 | if (f.file) { |
521 | audit_inode(NULL, f.file->f_path.dentry, 0); | |
522 | err = chmod_common(&f.file->f_path, mode); | |
523 | fdput(f); | |
e57712eb | 524 | } |
1da177e4 LT |
525 | return err; |
526 | } | |
527 | ||
49f0a076 | 528 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 529 | { |
2d8f3038 | 530 | struct path path; |
1da177e4 | 531 | int error; |
14ff690c JL |
532 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
533 | retry: | |
534 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
535 | if (!error) { |
536 | error = chmod_common(&path, mode); | |
537 | path_put(&path); | |
14ff690c JL |
538 | if (retry_estale(error, lookup_flags)) { |
539 | lookup_flags |= LOOKUP_REVAL; | |
540 | goto retry; | |
541 | } | |
e57712eb | 542 | } |
1da177e4 LT |
543 | return error; |
544 | } | |
545 | ||
49f0a076 | 546 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
547 | { |
548 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
549 | } | |
550 | ||
fe542cf5 | 551 | static int chown_common(struct path *path, uid_t user, gid_t group) |
1da177e4 | 552 | { |
fe542cf5 | 553 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 554 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
555 | int error; |
556 | struct iattr newattrs; | |
52137abe EB |
557 | kuid_t uid; |
558 | kgid_t gid; | |
559 | ||
560 | uid = make_kuid(current_user_ns(), user); | |
561 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 562 | |
1da177e4 LT |
563 | newattrs.ia_valid = ATTR_CTIME; |
564 | if (user != (uid_t) -1) { | |
52137abe EB |
565 | if (!uid_valid(uid)) |
566 | return -EINVAL; | |
1da177e4 | 567 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 568 | newattrs.ia_uid = uid; |
1da177e4 LT |
569 | } |
570 | if (group != (gid_t) -1) { | |
52137abe EB |
571 | if (!gid_valid(gid)) |
572 | return -EINVAL; | |
1da177e4 | 573 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 574 | newattrs.ia_gid = gid; |
1da177e4 LT |
575 | } |
576 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
577 | newattrs.ia_valid |= |
578 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
27ac0ffe | 579 | retry_deleg: |
1b1dcc1b | 580 | mutex_lock(&inode->i_mutex); |
d2b31ca6 | 581 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 582 | if (!error) |
27ac0ffe | 583 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
1b1dcc1b | 584 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
585 | if (delegated_inode) { |
586 | error = break_deleg_wait(&delegated_inode); | |
587 | if (!error) | |
588 | goto retry_deleg; | |
589 | } | |
1da177e4 LT |
590 | return error; |
591 | } | |
592 | ||
6559eed8 HC |
593 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
594 | gid_t, group, int, flag) | |
5590ff0d | 595 | { |
2d8f3038 | 596 | struct path path; |
5590ff0d | 597 | int error = -EINVAL; |
65cfc672 | 598 | int lookup_flags; |
5590ff0d | 599 | |
65cfc672 | 600 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
601 | goto out; |
602 | ||
65cfc672 AV |
603 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
604 | if (flag & AT_EMPTY_PATH) | |
605 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 606 | retry: |
65cfc672 | 607 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
608 | if (error) |
609 | goto out; | |
2d8f3038 | 610 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
611 | if (error) |
612 | goto out_release; | |
fe542cf5 | 613 | error = chown_common(&path, user, group); |
2d8f3038 | 614 | mnt_drop_write(path.mnt); |
2af482a7 | 615 | out_release: |
2d8f3038 | 616 | path_put(&path); |
99a5df37 JL |
617 | if (retry_estale(error, lookup_flags)) { |
618 | lookup_flags |= LOOKUP_REVAL; | |
619 | goto retry; | |
620 | } | |
5590ff0d UD |
621 | out: |
622 | return error; | |
623 | } | |
624 | ||
55e4def0 | 625 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 626 | { |
55e4def0 DH |
627 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
628 | } | |
1da177e4 | 629 | |
55e4def0 DH |
630 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
631 | { | |
632 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
633 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
634 | } |
635 | ||
ca013e94 | 636 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 637 | { |
2903ff01 | 638 | struct fd f = fdget(fd); |
1da177e4 LT |
639 | int error = -EBADF; |
640 | ||
2903ff01 | 641 | if (!f.file) |
6902d925 DH |
642 | goto out; |
643 | ||
2903ff01 | 644 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
645 | if (error) |
646 | goto out_fput; | |
bfcec708 | 647 | audit_inode(NULL, f.file->f_path.dentry, 0); |
2903ff01 AV |
648 | error = chown_common(&f.file->f_path, user, group); |
649 | mnt_drop_write_file(f.file); | |
2af482a7 | 650 | out_fput: |
2903ff01 | 651 | fdput(f); |
6902d925 | 652 | out: |
1da177e4 LT |
653 | return error; |
654 | } | |
655 | ||
4a3fd211 DH |
656 | /* |
657 | * You have to be very careful that these write | |
658 | * counts get cleaned up in error cases and | |
659 | * upon __fput(). This should probably never | |
660 | * be called outside of __dentry_open(). | |
661 | */ | |
662 | static inline int __get_file_write_access(struct inode *inode, | |
663 | struct vfsmount *mnt) | |
664 | { | |
665 | int error; | |
666 | error = get_write_access(inode); | |
667 | if (error) | |
668 | return error; | |
669 | /* | |
670 | * Do not take mount writer counts on | |
671 | * special files since no writes to | |
672 | * the mount itself will occur. | |
673 | */ | |
674 | if (!special_file(inode->i_mode)) { | |
675 | /* | |
676 | * Balanced in __fput() | |
677 | */ | |
eb04c282 | 678 | error = __mnt_want_write(mnt); |
4a3fd211 DH |
679 | if (error) |
680 | put_write_access(inode); | |
681 | } | |
682 | return error; | |
683 | } | |
684 | ||
90ad1a8e MS |
685 | int open_check_o_direct(struct file *f) |
686 | { | |
687 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
688 | if (f->f_flags & O_DIRECT) { | |
689 | if (!f->f_mapping->a_ops || | |
690 | ((!f->f_mapping->a_ops->direct_IO) && | |
691 | (!f->f_mapping->a_ops->get_xip_mem))) { | |
692 | return -EINVAL; | |
693 | } | |
694 | } | |
695 | return 0; | |
696 | } | |
697 | ||
02e5180d | 698 | static int do_dentry_open(struct file *f, |
96b7e579 AV |
699 | int (*open)(struct inode *, struct file *), |
700 | const struct cred *cred) | |
1da177e4 | 701 | { |
1abf0c71 | 702 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
703 | struct inode *inode; |
704 | int error; | |
705 | ||
5300990c | 706 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 707 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 AV |
708 | |
709 | if (unlikely(f->f_flags & O_PATH)) | |
710 | f->f_mode = FMODE_PATH; | |
711 | ||
b5bcdda3 | 712 | path_get(&f->f_path); |
dd37978c | 713 | inode = f->f_inode = f->f_path.dentry->d_inode; |
1da177e4 | 714 | if (f->f_mode & FMODE_WRITE) { |
02e5180d | 715 | error = __get_file_write_access(inode, f->f_path.mnt); |
1da177e4 LT |
716 | if (error) |
717 | goto cleanup_file; | |
ad775f5a DH |
718 | if (!special_file(inode->i_mode)) |
719 | file_take_write(f); | |
1da177e4 LT |
720 | } |
721 | ||
722 | f->f_mapping = inode->i_mapping; | |
1da177e4 | 723 | |
1abf0c71 AV |
724 | if (unlikely(f->f_mode & FMODE_PATH)) { |
725 | f->f_op = &empty_fops; | |
96b7e579 | 726 | return 0; |
1abf0c71 AV |
727 | } |
728 | ||
9c225f26 LT |
729 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
730 | if (S_ISREG(inode->i_mode)) | |
731 | f->f_mode |= FMODE_ATOMIC_POS; | |
732 | ||
1abf0c71 | 733 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
734 | if (unlikely(WARN_ON(!f->f_op))) { |
735 | error = -ENODEV; | |
736 | goto cleanup_all; | |
737 | } | |
1abf0c71 | 738 | |
83d49856 | 739 | error = security_file_open(f, cred); |
788e7dd4 YN |
740 | if (error) |
741 | goto cleanup_all; | |
742 | ||
f3c7691e BF |
743 | error = break_lease(inode, f->f_flags); |
744 | if (error) | |
745 | goto cleanup_all; | |
746 | ||
72c2d531 | 747 | if (!open) |
834f2a4a TM |
748 | open = f->f_op->open; |
749 | if (open) { | |
750 | error = open(inode, f); | |
1da177e4 LT |
751 | if (error) |
752 | goto cleanup_all; | |
753 | } | |
890275b5 MZ |
754 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
755 | i_readcount_inc(inode); | |
834f2a4a | 756 | |
1da177e4 LT |
757 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
758 | ||
759 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
760 | ||
96b7e579 | 761 | return 0; |
1da177e4 LT |
762 | |
763 | cleanup_all: | |
764 | fops_put(f->f_op); | |
4a3fd211 | 765 | if (f->f_mode & FMODE_WRITE) { |
1da177e4 | 766 | put_write_access(inode); |
ad775f5a DH |
767 | if (!special_file(inode->i_mode)) { |
768 | /* | |
769 | * We don't consider this a real | |
770 | * mnt_want/drop_write() pair | |
771 | * because it all happenend right | |
772 | * here, so just reset the state. | |
773 | */ | |
774 | file_reset_write(f); | |
fe7c8051 | 775 | __mnt_drop_write(f->f_path.mnt); |
ad775f5a | 776 | } |
4a3fd211 | 777 | } |
1da177e4 | 778 | cleanup_file: |
02e5180d AV |
779 | path_put(&f->f_path); |
780 | f->f_path.mnt = NULL; | |
781 | f->f_path.dentry = NULL; | |
dd37978c | 782 | f->f_inode = NULL; |
96b7e579 | 783 | return error; |
1da177e4 LT |
784 | } |
785 | ||
d18e9008 MS |
786 | /** |
787 | * finish_open - finish opening a file | |
0854d450 | 788 | * @file: file pointer |
d18e9008 MS |
789 | * @dentry: pointer to dentry |
790 | * @open: open callback | |
0854d450 | 791 | * @opened: state of open |
d18e9008 MS |
792 | * |
793 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
794 | * | |
795 | * If the open callback is set to NULL, then the standard f_op->open() | |
796 | * filesystem callback is substituted. | |
0854d450 MS |
797 | * |
798 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
799 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
800 | * on it after finish_open(). | |
801 | * | |
802 | * On successful return @file is a fully instantiated open file. After this, if | |
803 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
804 | * | |
805 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 806 | */ |
30d90494 AV |
807 | int finish_open(struct file *file, struct dentry *dentry, |
808 | int (*open)(struct inode *, struct file *), | |
809 | int *opened) | |
d18e9008 | 810 | { |
96b7e579 | 811 | int error; |
3d8a00d2 | 812 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 813 | |
b5bcdda3 | 814 | file->f_path.dentry = dentry; |
02e5180d | 815 | error = do_dentry_open(file, open, current_cred()); |
96b7e579 | 816 | if (!error) |
47237687 | 817 | *opened |= FILE_OPENED; |
d18e9008 | 818 | |
96b7e579 | 819 | return error; |
d18e9008 MS |
820 | } |
821 | EXPORT_SYMBOL(finish_open); | |
822 | ||
823 | /** | |
824 | * finish_no_open - finish ->atomic_open() without opening the file | |
825 | * | |
0854d450 | 826 | * @file: file pointer |
d18e9008 MS |
827 | * @dentry: dentry or NULL (as returned from ->lookup()) |
828 | * | |
829 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
830 | * |
831 | * NB: unlike finish_open() this function does consume the dentry reference and | |
832 | * the caller need not dput() it. | |
833 | * | |
834 | * Returns "1" which must be the return value of ->atomic_open() after having | |
835 | * called this function. | |
d18e9008 | 836 | */ |
e45198a6 | 837 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 838 | { |
30d90494 | 839 | file->f_path.dentry = dentry; |
e45198a6 | 840 | return 1; |
d18e9008 MS |
841 | } |
842 | EXPORT_SYMBOL(finish_no_open); | |
843 | ||
765927b2 | 844 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 845 | const struct cred *cred) |
a1a5b3d9 PS |
846 | { |
847 | int error; | |
848 | struct file *f; | |
849 | ||
e0e81739 DH |
850 | validate_creds(cred); |
851 | ||
c212f9aa | 852 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 853 | BUG_ON(!path->mnt); |
322ee5b3 | 854 | |
a1a5b3d9 | 855 | f = get_empty_filp(); |
1afc99be AV |
856 | if (!IS_ERR(f)) { |
857 | f->f_flags = flags; | |
858 | f->f_path = *path; | |
859 | error = do_dentry_open(f, NULL, cred); | |
860 | if (!error) { | |
861 | /* from now on we need fput() to dispose of f */ | |
862 | error = open_check_o_direct(f); | |
863 | if (error) { | |
864 | fput(f); | |
865 | f = ERR_PTR(error); | |
866 | } | |
867 | } else { | |
868 | put_filp(f); | |
2a027e7a AV |
869 | f = ERR_PTR(error); |
870 | } | |
2a027e7a AV |
871 | } |
872 | return f; | |
a1a5b3d9 | 873 | } |
1da177e4 LT |
874 | EXPORT_SYMBOL(dentry_open); |
875 | ||
a218d0fd | 876 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
877 | { |
878 | int lookup_flags = 0; | |
879 | int acc_mode; | |
880 | ||
e305f48b | 881 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
882 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
883 | else | |
884 | op->mode = 0; | |
47c805dc AV |
885 | |
886 | /* Must never be set by userspace */ | |
c6f3d811 | 887 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
888 | |
889 | /* | |
890 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
891 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
892 | * always set instead of having to deal with possibly weird behaviour | |
893 | * for malicious applications setting only __O_SYNC. | |
894 | */ | |
895 | if (flags & __O_SYNC) | |
896 | flags |= O_DSYNC; | |
897 | ||
bb458c64 AV |
898 | if (flags & __O_TMPFILE) { |
899 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d AV |
900 | return -EINVAL; |
901 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
ba57ea64 AV |
902 | if (!(acc_mode & MAY_WRITE)) |
903 | return -EINVAL; | |
60545d0d AV |
904 | } else if (flags & O_PATH) { |
905 | /* | |
906 | * If we have O_PATH in the open flag. Then we | |
907 | * cannot have anything other than the below set of flags | |
908 | */ | |
1abf0c71 AV |
909 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
910 | acc_mode = 0; | |
911 | } else { | |
912 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
913 | } | |
47c805dc | 914 | |
1abf0c71 | 915 | op->open_flag = flags; |
47c805dc AV |
916 | |
917 | /* O_TRUNC implies we need access checks for write permissions */ | |
918 | if (flags & O_TRUNC) | |
919 | acc_mode |= MAY_WRITE; | |
920 | ||
921 | /* Allow the LSM permission hook to distinguish append | |
922 | access from general write access. */ | |
923 | if (flags & O_APPEND) | |
924 | acc_mode |= MAY_APPEND; | |
925 | ||
926 | op->acc_mode = acc_mode; | |
927 | ||
1abf0c71 AV |
928 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
929 | ||
47c805dc AV |
930 | if (flags & O_CREAT) { |
931 | op->intent |= LOOKUP_CREATE; | |
932 | if (flags & O_EXCL) | |
933 | op->intent |= LOOKUP_EXCL; | |
934 | } | |
935 | ||
936 | if (flags & O_DIRECTORY) | |
937 | lookup_flags |= LOOKUP_DIRECTORY; | |
938 | if (!(flags & O_NOFOLLOW)) | |
939 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
940 | op->lookup_flags = lookup_flags; |
941 | return 0; | |
47c805dc AV |
942 | } |
943 | ||
669abf4e JL |
944 | /** |
945 | * file_open_name - open file and return file pointer | |
946 | * | |
947 | * @name: struct filename containing path to open | |
948 | * @flags: open flags as per the open(2) second argument | |
949 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
950 | * | |
951 | * This is the helper to open a file from kernelspace if you really | |
952 | * have to. But in generally you should not do this, so please move | |
953 | * along, nothing to see here.. | |
954 | */ | |
955 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
956 | { | |
957 | struct open_flags op; | |
f9652e10 AV |
958 | int err = build_open_flags(flags, mode, &op); |
959 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
960 | } |
961 | ||
47c805dc AV |
962 | /** |
963 | * filp_open - open file and return file pointer | |
964 | * | |
965 | * @filename: path to open | |
966 | * @flags: open flags as per the open(2) second argument | |
967 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
968 | * | |
969 | * This is the helper to open a file from kernelspace if you really | |
970 | * have to. But in generally you should not do this, so please move | |
971 | * along, nothing to see here.. | |
972 | */ | |
a218d0fd | 973 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 974 | { |
669abf4e JL |
975 | struct filename name = {.name = filename}; |
976 | return file_open_name(&name, flags, mode); | |
47c805dc AV |
977 | } |
978 | EXPORT_SYMBOL(filp_open); | |
979 | ||
73d049a4 AV |
980 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
981 | const char *filename, int flags) | |
982 | { | |
983 | struct open_flags op; | |
f9652e10 AV |
984 | int err = build_open_flags(flags, 0, &op); |
985 | if (err) | |
986 | return ERR_PTR(err); | |
73d049a4 AV |
987 | if (flags & O_CREAT) |
988 | return ERR_PTR(-EINVAL); | |
989 | if (!filename && (flags & O_DIRECTORY)) | |
990 | if (!dentry->d_inode->i_op->lookup) | |
991 | return ERR_PTR(-ENOTDIR); | |
f9652e10 | 992 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
993 | } |
994 | EXPORT_SYMBOL(file_open_root); | |
995 | ||
a218d0fd | 996 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 997 | { |
47c805dc | 998 | struct open_flags op; |
f9652e10 AV |
999 | int fd = build_open_flags(flags, mode, &op); |
1000 | struct filename *tmp; | |
1001 | ||
1002 | if (fd) | |
1003 | return fd; | |
1004 | ||
1005 | tmp = getname(filename); | |
1006 | if (IS_ERR(tmp)) | |
1007 | return PTR_ERR(tmp); | |
1008 | ||
1009 | fd = get_unused_fd_flags(flags); | |
1010 | if (fd >= 0) { | |
1011 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1012 | if (IS_ERR(f)) { | |
1013 | put_unused_fd(fd); | |
1014 | fd = PTR_ERR(f); | |
1015 | } else { | |
1016 | fsnotify_open(f); | |
1017 | fd_install(fd, f); | |
1da177e4 | 1018 | } |
1da177e4 | 1019 | } |
f9652e10 | 1020 | putname(tmp); |
1da177e4 | 1021 | return fd; |
1da177e4 | 1022 | } |
e922efc3 | 1023 | |
a218d0fd | 1024 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1025 | { |
1026 | if (force_o_largefile()) | |
1027 | flags |= O_LARGEFILE; | |
1028 | ||
2cf09666 | 1029 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1030 | } |
1da177e4 | 1031 | |
6559eed8 | 1032 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1033 | umode_t, mode) |
5590ff0d UD |
1034 | { |
1035 | if (force_o_largefile()) | |
1036 | flags |= O_LARGEFILE; | |
1037 | ||
2cf09666 | 1038 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1039 | } |
5590ff0d | 1040 | |
1da177e4 LT |
1041 | #ifndef __alpha__ |
1042 | ||
1043 | /* | |
1044 | * For backward compatibility? Maybe this should be moved | |
1045 | * into arch/i386 instead? | |
1046 | */ | |
a218d0fd | 1047 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1048 | { |
1049 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1050 | } | |
1051 | ||
1052 | #endif | |
1053 | ||
1054 | /* | |
1055 | * "id" is the POSIX thread ID. We use the | |
1056 | * files pointer for this.. | |
1057 | */ | |
1058 | int filp_close(struct file *filp, fl_owner_t id) | |
1059 | { | |
45778ca8 | 1060 | int retval = 0; |
1da177e4 LT |
1061 | |
1062 | if (!file_count(filp)) { | |
1063 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1064 | return 0; |
1da177e4 LT |
1065 | } |
1066 | ||
72c2d531 | 1067 | if (filp->f_op->flush) |
75e1fcc0 | 1068 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1069 | |
1abf0c71 AV |
1070 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1071 | dnotify_flush(filp, id); | |
1072 | locks_remove_posix(filp, id); | |
1073 | } | |
1da177e4 LT |
1074 | fput(filp); |
1075 | return retval; | |
1076 | } | |
1077 | ||
1078 | EXPORT_SYMBOL(filp_close); | |
1079 | ||
1080 | /* | |
1081 | * Careful here! We test whether the file pointer is NULL before | |
1082 | * releasing the fd. This ensures that one clone task can't release | |
1083 | * an fd while another clone is opening it. | |
1084 | */ | |
ca013e94 | 1085 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1086 | { |
483ce1d4 | 1087 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1088 | |
1089 | /* can't restart close syscall because file table entry was cleared */ | |
1090 | if (unlikely(retval == -ERESTARTSYS || | |
1091 | retval == -ERESTARTNOINTR || | |
1092 | retval == -ERESTARTNOHAND || | |
1093 | retval == -ERESTART_RESTARTBLOCK)) | |
1094 | retval = -EINTR; | |
1095 | ||
1096 | return retval; | |
1da177e4 | 1097 | } |
1da177e4 LT |
1098 | EXPORT_SYMBOL(sys_close); |
1099 | ||
1100 | /* | |
1101 | * This routine simulates a hangup on the tty, to arrange that users | |
1102 | * are given clean terminals at login time. | |
1103 | */ | |
ca013e94 | 1104 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1105 | { |
1106 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1107 | tty_vhangup_self(); |
1da177e4 LT |
1108 | return 0; |
1109 | } | |
1110 | return -EPERM; | |
1111 | } | |
1112 | ||
1113 | /* | |
1114 | * Called when an inode is about to be open. | |
1115 | * We use this to disallow opening large files on 32bit systems if | |
1116 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1117 | * on this flag in sys_open. | |
1118 | */ | |
1119 | int generic_file_open(struct inode * inode, struct file * filp) | |
1120 | { | |
1121 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1122 | return -EOVERFLOW; |
1da177e4 LT |
1123 | return 0; |
1124 | } | |
1125 | ||
1126 | EXPORT_SYMBOL(generic_file_open); | |
1127 | ||
1128 | /* | |
1129 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1130 | * file descriptors. The function is not supposed to ever fail, the only |
1131 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1132 | * directly into file_operations structure. | |
1da177e4 LT |
1133 | */ |
1134 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1135 | { | |
1136 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1137 | return 0; | |
1138 | } | |
1139 | ||
1140 | EXPORT_SYMBOL(nonseekable_open); |