]>
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 | ||
ccae47c6 SJR |
37 | #define CREATE_TRACE_POINTS |
38 | #include <trace/events/fs.h> | |
39 | ||
4a30131e N |
40 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
41 | struct file *filp) | |
1da177e4 | 42 | { |
939a9421 | 43 | int ret; |
1da177e4 LT |
44 | struct iattr newattrs; |
45 | ||
46 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
47 | if (length < 0) | |
48 | return -EINVAL; | |
49 | ||
50 | newattrs.ia_size = length; | |
4a30131e | 51 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
52 | if (filp) { |
53 | newattrs.ia_file = filp; | |
54 | newattrs.ia_valid |= ATTR_FILE; | |
55 | } | |
1da177e4 | 56 | |
45f147a1 JK |
57 | /* Remove suid, sgid, and file capabilities on truncate too */ |
58 | ret = dentry_needs_remove_privs(dentry); | |
59 | if (ret < 0) | |
60 | return ret; | |
939a9421 AW |
61 | if (ret) |
62 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 63 | |
1b1dcc1b | 64 | mutex_lock(&dentry->d_inode->i_mutex); |
27ac0ffe BF |
65 | /* Note any delegations or leases have already been broken: */ |
66 | ret = notify_change(dentry, &newattrs, NULL); | |
1b1dcc1b | 67 | mutex_unlock(&dentry->d_inode->i_mutex); |
939a9421 | 68 | return ret; |
1da177e4 | 69 | } |
dee5220c | 70 | EXPORT_SYMBOL(do_truncate); |
1da177e4 | 71 | |
a02de960 | 72 | long vfs_truncate(struct path *path, loff_t length) |
1da177e4 | 73 | { |
2d8f3038 | 74 | struct inode *inode; |
a02de960 | 75 | long error; |
1da177e4 | 76 | |
a02de960 | 77 | inode = path->dentry->d_inode; |
1da177e4 LT |
78 | |
79 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 80 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 81 | return -EISDIR; |
1da177e4 | 82 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 83 | return -EINVAL; |
1da177e4 | 84 | |
a02de960 | 85 | error = mnt_want_write(path->mnt); |
1da177e4 | 86 | if (error) |
a02de960 | 87 | goto out; |
1da177e4 | 88 | |
256984a8 | 89 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
90 | if (error) |
91 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
92 | |
93 | error = -EPERM; | |
c82e42da | 94 | if (IS_APPEND(inode)) |
9ac9b847 | 95 | goto mnt_drop_write_and_out; |
1da177e4 | 96 | |
9700382c | 97 | error = get_write_access(inode); |
1da177e4 | 98 | if (error) |
9ac9b847 | 99 | goto mnt_drop_write_and_out; |
1da177e4 | 100 | |
9700382c | 101 | /* |
102 | * Make sure that there are no leases. get_write_access() protects | |
103 | * against the truncate racing with a lease-granting setlease(). | |
104 | */ | |
8737c930 | 105 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 106 | if (error) |
9700382c | 107 | goto put_write_and_out; |
1da177e4 LT |
108 | |
109 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 110 | if (!error) |
a02de960 | 111 | error = security_path_truncate(path); |
907f4554 | 112 | if (!error) |
a02de960 | 113 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 114 | |
9700382c | 115 | put_write_and_out: |
116 | put_write_access(inode); | |
9ac9b847 | 117 | mnt_drop_write_and_out: |
a02de960 | 118 | mnt_drop_write(path->mnt); |
1da177e4 LT |
119 | out: |
120 | return error; | |
121 | } | |
a02de960 DH |
122 | EXPORT_SYMBOL_GPL(vfs_truncate); |
123 | ||
124 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
125 | { | |
48f7530d | 126 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
127 | struct path path; |
128 | int error; | |
129 | ||
130 | if (length < 0) /* sorry, but loff_t says... */ | |
131 | return -EINVAL; | |
132 | ||
48f7530d JL |
133 | retry: |
134 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
135 | if (!error) { |
136 | error = vfs_truncate(&path, length); | |
137 | path_put(&path); | |
138 | } | |
48f7530d JL |
139 | if (retry_estale(error, lookup_flags)) { |
140 | lookup_flags |= LOOKUP_REVAL; | |
141 | goto retry; | |
142 | } | |
a02de960 DH |
143 | return error; |
144 | } | |
1da177e4 | 145 | |
4fd8da8d | 146 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 147 | { |
4fd8da8d | 148 | return do_sys_truncate(path, length); |
1da177e4 LT |
149 | } |
150 | ||
3f6d078d AV |
151 | #ifdef CONFIG_COMPAT |
152 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
153 | { | |
154 | return do_sys_truncate(path, length); | |
155 | } | |
156 | #endif | |
157 | ||
b01ec0ef | 158 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 159 | { |
bf2965d5 | 160 | struct inode *inode; |
1da177e4 | 161 | struct dentry *dentry; |
2903ff01 | 162 | struct fd f; |
1da177e4 LT |
163 | int error; |
164 | ||
165 | error = -EINVAL; | |
166 | if (length < 0) | |
167 | goto out; | |
168 | error = -EBADF; | |
2903ff01 AV |
169 | f = fdget(fd); |
170 | if (!f.file) | |
1da177e4 LT |
171 | goto out; |
172 | ||
173 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 174 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
175 | small = 0; |
176 | ||
2903ff01 | 177 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
178 | inode = dentry->d_inode; |
179 | error = -EINVAL; | |
2903ff01 | 180 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
181 | goto out_putf; |
182 | ||
183 | error = -EINVAL; | |
184 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
185 | if (small && length > MAX_NON_LFS) | |
186 | goto out_putf; | |
187 | ||
188 | error = -EPERM; | |
189 | if (IS_APPEND(inode)) | |
190 | goto out_putf; | |
191 | ||
14da9200 | 192 | sb_start_write(inode->i_sb); |
2903ff01 | 193 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 194 | if (!error) |
2903ff01 | 195 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 196 | if (!error) |
2903ff01 | 197 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 198 | sb_end_write(inode->i_sb); |
1da177e4 | 199 | out_putf: |
2903ff01 | 200 | fdput(f); |
1da177e4 LT |
201 | out: |
202 | return error; | |
203 | } | |
204 | ||
bdc480e3 | 205 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 206 | { |
2cf09666 | 207 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
208 | } |
209 | ||
3f6d078d AV |
210 | #ifdef CONFIG_COMPAT |
211 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
212 | { | |
213 | return do_sys_ftruncate(fd, length, 1); | |
214 | } | |
215 | #endif | |
216 | ||
1da177e4 LT |
217 | /* LFS versions of truncate are only needed on 32 bit machines */ |
218 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 219 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
220 | { |
221 | return do_sys_truncate(path, length); | |
222 | } | |
223 | ||
4a0fd5bf | 224 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 225 | { |
2cf09666 | 226 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 227 | } |
6673e0c3 | 228 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 229 | |
3e63cbb1 | 230 | |
72c72bdf | 231 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 232 | { |
496ad9aa | 233 | struct inode *inode = file_inode(file); |
3e63cbb1 | 234 | long ret; |
97ac7350 AA |
235 | |
236 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 237 | return -EINVAL; |
97ac7350 AA |
238 | |
239 | /* Return error if mode is not supported */ | |
dd46c787 | 240 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
241 | return -EOPNOTSUPP; |
242 | ||
243 | /* Punch hole and zero range are mutually exclusive */ | |
244 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
245 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
246 | return -EOPNOTSUPP; |
247 | ||
248 | /* Punch hole must have keep size set */ | |
249 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
250 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 251 | return -EOPNOTSUPP; |
97ac7350 | 252 | |
00f5e619 NJ |
253 | /* Collapse range should only be used exclusively. */ |
254 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
255 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
256 | return -EINVAL; | |
257 | ||
dd46c787 NJ |
258 | /* Insert range should only be used exclusively. */ |
259 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
260 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
261 | return -EINVAL; | |
262 | ||
97ac7350 | 263 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 264 | return -EBADF; |
1ca551c6 | 265 | |
00f5e619 | 266 | /* |
8fc61d92 | 267 | * We can only allow pure fallocate on append only files |
00f5e619 | 268 | */ |
8fc61d92 | 269 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
270 | return -EPERM; |
271 | ||
272 | if (IS_IMMUTABLE(inode)) | |
273 | return -EPERM; | |
274 | ||
0790b31b | 275 | /* |
6d2b6170 | 276 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
277 | */ |
278 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 279 | return -ETXTBSY; |
0790b31b | 280 | |
97ac7350 AA |
281 | /* |
282 | * Revalidate the write permissions, in case security policy has | |
283 | * changed since the files were opened. | |
284 | */ | |
285 | ret = security_file_permission(file, MAY_WRITE); | |
286 | if (ret) | |
3e63cbb1 | 287 | return ret; |
97ac7350 | 288 | |
97ac7350 | 289 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 290 | return -ESPIPE; |
97ac7350 | 291 | |
97ac7350 AA |
292 | /* |
293 | * Let individual file system decide if it supports preallocation | |
294 | * for directories or not. | |
295 | */ | |
296 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 297 | return -ENODEV; |
97ac7350 | 298 | |
97ac7350 AA |
299 | /* Check for wrap through zero too */ |
300 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 301 | return -EFBIG; |
97ac7350 | 302 | |
2fe17c10 | 303 | if (!file->f_op->fallocate) |
3e63cbb1 | 304 | return -EOPNOTSUPP; |
97ac7350 | 305 | |
14da9200 JK |
306 | sb_start_write(inode->i_sb); |
307 | ret = file->f_op->fallocate(file, mode, offset, len); | |
820c12d5 HS |
308 | |
309 | /* | |
310 | * Create inotify and fanotify events. | |
311 | * | |
312 | * To keep the logic simple always create events if fallocate succeeds. | |
313 | * This implies that events are even created if the file size remains | |
314 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
315 | */ | |
316 | if (ret == 0) | |
317 | fsnotify_modify(file); | |
318 | ||
14da9200 JK |
319 | sb_end_write(inode->i_sb); |
320 | return ret; | |
3e63cbb1 | 321 | } |
72c72bdf | 322 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 323 | |
4a0fd5bf | 324 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 325 | { |
2903ff01 | 326 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
327 | int error = -EBADF; |
328 | ||
2903ff01 | 329 | if (f.file) { |
72c72bdf | 330 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 331 | fdput(f); |
3e63cbb1 | 332 | } |
3e63cbb1 | 333 | return error; |
97ac7350 | 334 | } |
3e63cbb1 | 335 | |
1da177e4 LT |
336 | /* |
337 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
338 | * We do this by temporarily clearing all FS-related capabilities and | |
339 | * switching the fsuid/fsgid around to the real ones. | |
340 | */ | |
6559eed8 | 341 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 342 | { |
d84f4f99 DH |
343 | const struct cred *old_cred; |
344 | struct cred *override_cred; | |
2d8f3038 | 345 | struct path path; |
256984a8 | 346 | struct inode *inode; |
1da177e4 | 347 | int res; |
87fa5595 | 348 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
349 | |
350 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
351 | return -EINVAL; | |
352 | ||
d84f4f99 DH |
353 | override_cred = prepare_creds(); |
354 | if (!override_cred) | |
355 | return -ENOMEM; | |
1da177e4 | 356 | |
d84f4f99 DH |
357 | override_cred->fsuid = override_cred->uid; |
358 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 359 | |
086f7316 | 360 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 361 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
362 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
363 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 364 | cap_clear(override_cred->cap_effective); |
086f7316 | 365 | else |
d84f4f99 DH |
366 | override_cred->cap_effective = |
367 | override_cred->cap_permitted; | |
086f7316 | 368 | } |
1da177e4 | 369 | |
d84f4f99 | 370 | old_cred = override_creds(override_cred); |
87fa5595 JL |
371 | retry: |
372 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
373 | if (res) |
374 | goto out; | |
375 | ||
63afdfc7 | 376 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
377 | |
378 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
379 | /* |
380 | * MAY_EXEC on regular files is denied if the fs is mounted | |
381 | * with the "noexec" flag. | |
382 | */ | |
383 | res = -EACCES; | |
90f8572b | 384 | if (path_noexec(&path)) |
30524472 AV |
385 | goto out_path_release; |
386 | } | |
387 | ||
256984a8 | 388 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 389 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 390 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 391 | goto out_path_release; |
2f676cbc DH |
392 | /* |
393 | * This is a rare case where using __mnt_is_readonly() | |
394 | * is OK without a mnt_want/drop_write() pair. Since | |
395 | * no actual write to the fs is performed here, we do | |
396 | * not need to telegraph to that to anyone. | |
397 | * | |
398 | * By doing this, we accept that this access is | |
399 | * inherently racy and know that the fs may change | |
400 | * state before we even see this result. | |
401 | */ | |
2d8f3038 | 402 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 403 | res = -EROFS; |
1da177e4 | 404 | |
6902d925 | 405 | out_path_release: |
2d8f3038 | 406 | path_put(&path); |
87fa5595 JL |
407 | if (retry_estale(res, lookup_flags)) { |
408 | lookup_flags |= LOOKUP_REVAL; | |
409 | goto retry; | |
410 | } | |
6902d925 | 411 | out: |
d84f4f99 DH |
412 | revert_creds(old_cred); |
413 | put_cred(override_cred); | |
1da177e4 LT |
414 | return res; |
415 | } | |
416 | ||
ca013e94 | 417 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
418 | { |
419 | return sys_faccessat(AT_FDCWD, filename, mode); | |
420 | } | |
421 | ||
3cdad428 | 422 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 423 | { |
2d8f3038 | 424 | struct path path; |
1da177e4 | 425 | int error; |
0291c0a5 JL |
426 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
427 | retry: | |
428 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
429 | if (error) |
430 | goto out; | |
431 | ||
9cfcac81 | 432 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
433 | if (error) |
434 | goto dput_and_out; | |
435 | ||
2d8f3038 | 436 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
437 | |
438 | dput_and_out: | |
2d8f3038 | 439 | path_put(&path); |
0291c0a5 JL |
440 | if (retry_estale(error, lookup_flags)) { |
441 | lookup_flags |= LOOKUP_REVAL; | |
442 | goto retry; | |
443 | } | |
1da177e4 LT |
444 | out: |
445 | return error; | |
446 | } | |
447 | ||
3cdad428 | 448 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 449 | { |
2903ff01 | 450 | struct fd f = fdget_raw(fd); |
1da177e4 | 451 | struct inode *inode; |
2903ff01 | 452 | int error = -EBADF; |
1da177e4 LT |
453 | |
454 | error = -EBADF; | |
2903ff01 | 455 | if (!f.file) |
1da177e4 LT |
456 | goto out; |
457 | ||
496ad9aa | 458 | inode = file_inode(f.file); |
1da177e4 LT |
459 | |
460 | error = -ENOTDIR; | |
461 | if (!S_ISDIR(inode->i_mode)) | |
462 | goto out_putf; | |
463 | ||
9cfcac81 | 464 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 465 | if (!error) |
2903ff01 | 466 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 467 | out_putf: |
2903ff01 | 468 | fdput(f); |
1da177e4 LT |
469 | out: |
470 | return error; | |
471 | } | |
472 | ||
3480b257 | 473 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 474 | { |
2d8f3038 | 475 | struct path path; |
1da177e4 | 476 | int error; |
2771261e JL |
477 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
478 | retry: | |
479 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
480 | if (error) |
481 | goto out; | |
482 | ||
9cfcac81 | 483 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
484 | if (error) |
485 | goto dput_and_out; | |
486 | ||
487 | error = -EPERM; | |
c7b96acf | 488 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 489 | goto dput_and_out; |
8b8efb44 TH |
490 | error = security_path_chroot(&path); |
491 | if (error) | |
492 | goto dput_and_out; | |
1da177e4 | 493 | |
2d8f3038 | 494 | set_fs_root(current->fs, &path); |
1da177e4 LT |
495 | error = 0; |
496 | dput_and_out: | |
2d8f3038 | 497 | path_put(&path); |
2771261e JL |
498 | if (retry_estale(error, lookup_flags)) { |
499 | lookup_flags |= LOOKUP_REVAL; | |
500 | goto retry; | |
501 | } | |
1da177e4 LT |
502 | out: |
503 | return error; | |
504 | } | |
505 | ||
e57712eb | 506 | static int chmod_common(struct path *path, umode_t mode) |
1da177e4 | 507 | { |
e57712eb | 508 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 509 | struct inode *delegated_inode = NULL; |
1da177e4 | 510 | struct iattr newattrs; |
e57712eb | 511 | int error; |
1da177e4 | 512 | |
e57712eb AV |
513 | error = mnt_want_write(path->mnt); |
514 | if (error) | |
515 | return error; | |
27ac0ffe | 516 | retry_deleg: |
fe542cf5 | 517 | mutex_lock(&inode->i_mutex); |
cdcf116d | 518 | error = security_path_chmod(path, mode); |
e57712eb | 519 | if (error) |
fe542cf5 | 520 | goto out_unlock; |
1da177e4 LT |
521 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
522 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 523 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 524 | out_unlock: |
1b1dcc1b | 525 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
526 | if (delegated_inode) { |
527 | error = break_deleg_wait(&delegated_inode); | |
528 | if (!error) | |
529 | goto retry_deleg; | |
530 | } | |
e57712eb AV |
531 | mnt_drop_write(path->mnt); |
532 | return error; | |
533 | } | |
534 | ||
49f0a076 | 535 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 536 | { |
173c8401 | 537 | struct fd f = fdget(fd); |
e57712eb AV |
538 | int err = -EBADF; |
539 | ||
173c8401 | 540 | if (f.file) { |
9f45f5bf | 541 | audit_file(f.file); |
173c8401 AV |
542 | err = chmod_common(&f.file->f_path, mode); |
543 | fdput(f); | |
e57712eb | 544 | } |
1da177e4 LT |
545 | return err; |
546 | } | |
547 | ||
49f0a076 | 548 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 549 | { |
2d8f3038 | 550 | struct path path; |
1da177e4 | 551 | int error; |
14ff690c JL |
552 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
553 | retry: | |
554 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
555 | if (!error) { |
556 | error = chmod_common(&path, mode); | |
557 | path_put(&path); | |
14ff690c JL |
558 | if (retry_estale(error, lookup_flags)) { |
559 | lookup_flags |= LOOKUP_REVAL; | |
560 | goto retry; | |
561 | } | |
e57712eb | 562 | } |
1da177e4 LT |
563 | return error; |
564 | } | |
565 | ||
49f0a076 | 566 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
567 | { |
568 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
569 | } | |
570 | ||
fe542cf5 | 571 | static int chown_common(struct path *path, uid_t user, gid_t group) |
1da177e4 | 572 | { |
fe542cf5 | 573 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 574 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
575 | int error; |
576 | struct iattr newattrs; | |
52137abe EB |
577 | kuid_t uid; |
578 | kgid_t gid; | |
579 | ||
580 | uid = make_kuid(current_user_ns(), user); | |
581 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 582 | |
c1b8940b | 583 | retry_deleg: |
1da177e4 LT |
584 | newattrs.ia_valid = ATTR_CTIME; |
585 | if (user != (uid_t) -1) { | |
52137abe EB |
586 | if (!uid_valid(uid)) |
587 | return -EINVAL; | |
1da177e4 | 588 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 589 | newattrs.ia_uid = uid; |
1da177e4 LT |
590 | } |
591 | if (group != (gid_t) -1) { | |
52137abe EB |
592 | if (!gid_valid(gid)) |
593 | return -EINVAL; | |
1da177e4 | 594 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 595 | newattrs.ia_gid = gid; |
1da177e4 LT |
596 | } |
597 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
598 | newattrs.ia_valid |= |
599 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
1b1dcc1b | 600 | mutex_lock(&inode->i_mutex); |
d2b31ca6 | 601 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 602 | if (!error) |
27ac0ffe | 603 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
1b1dcc1b | 604 | mutex_unlock(&inode->i_mutex); |
27ac0ffe BF |
605 | if (delegated_inode) { |
606 | error = break_deleg_wait(&delegated_inode); | |
607 | if (!error) | |
608 | goto retry_deleg; | |
609 | } | |
1da177e4 LT |
610 | return error; |
611 | } | |
612 | ||
6559eed8 HC |
613 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
614 | gid_t, group, int, flag) | |
5590ff0d | 615 | { |
2d8f3038 | 616 | struct path path; |
5590ff0d | 617 | int error = -EINVAL; |
65cfc672 | 618 | int lookup_flags; |
5590ff0d | 619 | |
65cfc672 | 620 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
621 | goto out; |
622 | ||
65cfc672 AV |
623 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
624 | if (flag & AT_EMPTY_PATH) | |
625 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 626 | retry: |
65cfc672 | 627 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
628 | if (error) |
629 | goto out; | |
2d8f3038 | 630 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
631 | if (error) |
632 | goto out_release; | |
fe542cf5 | 633 | error = chown_common(&path, user, group); |
2d8f3038 | 634 | mnt_drop_write(path.mnt); |
2af482a7 | 635 | out_release: |
2d8f3038 | 636 | path_put(&path); |
99a5df37 JL |
637 | if (retry_estale(error, lookup_flags)) { |
638 | lookup_flags |= LOOKUP_REVAL; | |
639 | goto retry; | |
640 | } | |
5590ff0d UD |
641 | out: |
642 | return error; | |
643 | } | |
644 | ||
55e4def0 | 645 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 646 | { |
55e4def0 DH |
647 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
648 | } | |
1da177e4 | 649 | |
55e4def0 DH |
650 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
651 | { | |
652 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
653 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
654 | } |
655 | ||
ca013e94 | 656 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 657 | { |
2903ff01 | 658 | struct fd f = fdget(fd); |
1da177e4 LT |
659 | int error = -EBADF; |
660 | ||
2903ff01 | 661 | if (!f.file) |
6902d925 DH |
662 | goto out; |
663 | ||
2903ff01 | 664 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
665 | if (error) |
666 | goto out_fput; | |
9f45f5bf | 667 | audit_file(f.file); |
2903ff01 AV |
668 | error = chown_common(&f.file->f_path, user, group); |
669 | mnt_drop_write_file(f.file); | |
2af482a7 | 670 | out_fput: |
2903ff01 | 671 | fdput(f); |
6902d925 | 672 | out: |
1da177e4 LT |
673 | return error; |
674 | } | |
675 | ||
90ad1a8e MS |
676 | int open_check_o_direct(struct file *f) |
677 | { | |
678 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
679 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 680 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 681 | return -EINVAL; |
90ad1a8e MS |
682 | } |
683 | return 0; | |
684 | } | |
dee5220c | 685 | EXPORT_SYMBOL(open_check_o_direct); |
90ad1a8e | 686 | |
02e5180d | 687 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 688 | struct inode *inode, |
96b7e579 AV |
689 | int (*open)(struct inode *, struct file *), |
690 | const struct cred *cred) | |
1da177e4 | 691 | { |
1abf0c71 | 692 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
693 | int error; |
694 | ||
5300990c | 695 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 696 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 697 | |
b5bcdda3 | 698 | path_get(&f->f_path); |
4bacc9c9 | 699 | f->f_inode = inode; |
1da177e4 | 700 | f->f_mapping = inode->i_mapping; |
1da177e4 | 701 | |
3f4d5a00 AV |
702 | if (unlikely(f->f_flags & O_PATH)) { |
703 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 704 | f->f_op = &empty_fops; |
96b7e579 | 705 | return 0; |
1abf0c71 AV |
706 | } |
707 | ||
dd20908a | 708 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 709 | error = get_write_access(inode); |
3f4d5a00 | 710 | if (unlikely(error)) |
1da177e4 | 711 | goto cleanup_file; |
0ccb2863 | 712 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 713 | if (unlikely(error)) { |
0ccb2863 AV |
714 | put_write_access(inode); |
715 | goto cleanup_file; | |
716 | } | |
83f936c7 | 717 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
718 | } |
719 | ||
9c225f26 LT |
720 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
721 | if (S_ISREG(inode->i_mode)) | |
722 | f->f_mode |= FMODE_ATOMIC_POS; | |
723 | ||
1abf0c71 | 724 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
725 | if (unlikely(WARN_ON(!f->f_op))) { |
726 | error = -ENODEV; | |
727 | goto cleanup_all; | |
728 | } | |
1abf0c71 | 729 | |
83d49856 | 730 | error = security_file_open(f, cred); |
788e7dd4 YN |
731 | if (error) |
732 | goto cleanup_all; | |
733 | ||
f3c7691e BF |
734 | error = break_lease(inode, f->f_flags); |
735 | if (error) | |
736 | goto cleanup_all; | |
737 | ||
72c2d531 | 738 | if (!open) |
834f2a4a TM |
739 | open = f->f_op->open; |
740 | if (open) { | |
741 | error = open(inode, f); | |
1da177e4 LT |
742 | if (error) |
743 | goto cleanup_all; | |
744 | } | |
890275b5 MZ |
745 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
746 | i_readcount_inc(inode); | |
293bc982 | 747 | if ((f->f_mode & FMODE_READ) && |
84363182 | 748 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 749 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 750 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 751 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 752 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 753 | |
1da177e4 LT |
754 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
755 | ||
756 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
757 | ||
96b7e579 | 758 | return 0; |
1da177e4 LT |
759 | |
760 | cleanup_all: | |
761 | fops_put(f->f_op); | |
83f936c7 | 762 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 763 | put_write_access(inode); |
83f936c7 | 764 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 765 | } |
1da177e4 | 766 | cleanup_file: |
02e5180d AV |
767 | path_put(&f->f_path); |
768 | f->f_path.mnt = NULL; | |
769 | f->f_path.dentry = NULL; | |
dd37978c | 770 | f->f_inode = NULL; |
96b7e579 | 771 | return error; |
1da177e4 LT |
772 | } |
773 | ||
d18e9008 MS |
774 | /** |
775 | * finish_open - finish opening a file | |
0854d450 | 776 | * @file: file pointer |
d18e9008 MS |
777 | * @dentry: pointer to dentry |
778 | * @open: open callback | |
0854d450 | 779 | * @opened: state of open |
d18e9008 MS |
780 | * |
781 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
782 | * | |
783 | * If the open callback is set to NULL, then the standard f_op->open() | |
784 | * filesystem callback is substituted. | |
0854d450 MS |
785 | * |
786 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
787 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
788 | * on it after finish_open(). | |
789 | * | |
790 | * On successful return @file is a fully instantiated open file. After this, if | |
791 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
792 | * | |
793 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 794 | */ |
30d90494 AV |
795 | int finish_open(struct file *file, struct dentry *dentry, |
796 | int (*open)(struct inode *, struct file *), | |
797 | int *opened) | |
d18e9008 | 798 | { |
96b7e579 | 799 | int error; |
3d8a00d2 | 800 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 801 | |
b5bcdda3 | 802 | file->f_path.dentry = dentry; |
4bacc9c9 DH |
803 | error = do_dentry_open(file, d_backing_inode(dentry), open, |
804 | current_cred()); | |
96b7e579 | 805 | if (!error) |
47237687 | 806 | *opened |= FILE_OPENED; |
d18e9008 | 807 | |
96b7e579 | 808 | return error; |
d18e9008 MS |
809 | } |
810 | EXPORT_SYMBOL(finish_open); | |
811 | ||
812 | /** | |
813 | * finish_no_open - finish ->atomic_open() without opening the file | |
814 | * | |
0854d450 | 815 | * @file: file pointer |
d18e9008 MS |
816 | * @dentry: dentry or NULL (as returned from ->lookup()) |
817 | * | |
818 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
819 | * |
820 | * NB: unlike finish_open() this function does consume the dentry reference and | |
821 | * the caller need not dput() it. | |
822 | * | |
823 | * Returns "1" which must be the return value of ->atomic_open() after having | |
824 | * called this function. | |
d18e9008 | 825 | */ |
e45198a6 | 826 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 827 | { |
30d90494 | 828 | file->f_path.dentry = dentry; |
e45198a6 | 829 | return 1; |
d18e9008 MS |
830 | } |
831 | EXPORT_SYMBOL(finish_no_open); | |
832 | ||
9bf39ab2 MS |
833 | char *file_path(struct file *filp, char *buf, int buflen) |
834 | { | |
835 | return d_path(&filp->f_path, buf, buflen); | |
836 | } | |
837 | EXPORT_SYMBOL(file_path); | |
838 | ||
4bacc9c9 DH |
839 | /** |
840 | * vfs_open - open the file at the given path | |
841 | * @path: path to open | |
842 | * @file: newly allocated file with f_flag initialized | |
843 | * @cred: credentials to use | |
844 | */ | |
845 | int vfs_open(const struct path *path, struct file *file, | |
846 | const struct cred *cred) | |
847 | { | |
234c6ee4 | 848 | struct inode *inode = vfs_select_inode(path->dentry, file->f_flags); |
4bacc9c9 | 849 | |
234c6ee4 MS |
850 | if (IS_ERR(inode)) |
851 | return PTR_ERR(inode); | |
4bacc9c9 | 852 | |
234c6ee4 | 853 | file->f_path = *path; |
4bacc9c9 DH |
854 | return do_dentry_open(file, inode, NULL, cred); |
855 | } | |
856 | ||
765927b2 | 857 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 858 | const struct cred *cred) |
a1a5b3d9 PS |
859 | { |
860 | int error; | |
861 | struct file *f; | |
862 | ||
e0e81739 DH |
863 | validate_creds(cred); |
864 | ||
c212f9aa | 865 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 866 | BUG_ON(!path->mnt); |
322ee5b3 | 867 | |
a1a5b3d9 | 868 | f = get_empty_filp(); |
1afc99be AV |
869 | if (!IS_ERR(f)) { |
870 | f->f_flags = flags; | |
4aa7c634 | 871 | error = vfs_open(path, f, cred); |
1afc99be AV |
872 | if (!error) { |
873 | /* from now on we need fput() to dispose of f */ | |
874 | error = open_check_o_direct(f); | |
875 | if (error) { | |
876 | fput(f); | |
877 | f = ERR_PTR(error); | |
878 | } | |
879 | } else { | |
880 | put_filp(f); | |
2a027e7a AV |
881 | f = ERR_PTR(error); |
882 | } | |
2a027e7a AV |
883 | } |
884 | return f; | |
a1a5b3d9 | 885 | } |
1da177e4 LT |
886 | EXPORT_SYMBOL(dentry_open); |
887 | ||
a218d0fd | 888 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
889 | { |
890 | int lookup_flags = 0; | |
891 | int acc_mode; | |
892 | ||
e305f48b | 893 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
894 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
895 | else | |
896 | op->mode = 0; | |
47c805dc AV |
897 | |
898 | /* Must never be set by userspace */ | |
c6f3d811 | 899 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
900 | |
901 | /* | |
902 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
903 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
904 | * always set instead of having to deal with possibly weird behaviour | |
905 | * for malicious applications setting only __O_SYNC. | |
906 | */ | |
907 | if (flags & __O_SYNC) | |
908 | flags |= O_DSYNC; | |
909 | ||
bb458c64 AV |
910 | if (flags & __O_TMPFILE) { |
911 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d AV |
912 | return -EINVAL; |
913 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
ba57ea64 AV |
914 | if (!(acc_mode & MAY_WRITE)) |
915 | return -EINVAL; | |
60545d0d AV |
916 | } else if (flags & O_PATH) { |
917 | /* | |
918 | * If we have O_PATH in the open flag. Then we | |
919 | * cannot have anything other than the below set of flags | |
920 | */ | |
1abf0c71 AV |
921 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
922 | acc_mode = 0; | |
923 | } else { | |
924 | acc_mode = MAY_OPEN | ACC_MODE(flags); | |
925 | } | |
47c805dc | 926 | |
1abf0c71 | 927 | op->open_flag = flags; |
47c805dc AV |
928 | |
929 | /* O_TRUNC implies we need access checks for write permissions */ | |
930 | if (flags & O_TRUNC) | |
931 | acc_mode |= MAY_WRITE; | |
932 | ||
933 | /* Allow the LSM permission hook to distinguish append | |
934 | access from general write access. */ | |
935 | if (flags & O_APPEND) | |
936 | acc_mode |= MAY_APPEND; | |
937 | ||
938 | op->acc_mode = acc_mode; | |
939 | ||
1abf0c71 AV |
940 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
941 | ||
47c805dc AV |
942 | if (flags & O_CREAT) { |
943 | op->intent |= LOOKUP_CREATE; | |
944 | if (flags & O_EXCL) | |
945 | op->intent |= LOOKUP_EXCL; | |
946 | } | |
947 | ||
948 | if (flags & O_DIRECTORY) | |
949 | lookup_flags |= LOOKUP_DIRECTORY; | |
950 | if (!(flags & O_NOFOLLOW)) | |
951 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
952 | op->lookup_flags = lookup_flags; |
953 | return 0; | |
47c805dc AV |
954 | } |
955 | ||
669abf4e JL |
956 | /** |
957 | * file_open_name - open file and return file pointer | |
958 | * | |
959 | * @name: struct filename containing path to open | |
960 | * @flags: open flags as per the open(2) second argument | |
961 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
962 | * | |
963 | * This is the helper to open a file from kernelspace if you really | |
964 | * have to. But in generally you should not do this, so please move | |
965 | * along, nothing to see here.. | |
966 | */ | |
967 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
968 | { | |
969 | struct open_flags op; | |
f9652e10 AV |
970 | int err = build_open_flags(flags, mode, &op); |
971 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
972 | } |
973 | ||
47c805dc AV |
974 | /** |
975 | * filp_open - open file and return file pointer | |
976 | * | |
977 | * @filename: path to open | |
978 | * @flags: open flags as per the open(2) second argument | |
979 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
980 | * | |
981 | * This is the helper to open a file from kernelspace if you really | |
982 | * have to. But in generally you should not do this, so please move | |
983 | * along, nothing to see here.. | |
984 | */ | |
a218d0fd | 985 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 986 | { |
51689104 PM |
987 | struct filename *name = getname_kernel(filename); |
988 | struct file *file = ERR_CAST(name); | |
989 | ||
990 | if (!IS_ERR(name)) { | |
991 | file = file_open_name(name, flags, mode); | |
992 | putname(name); | |
993 | } | |
994 | return file; | |
47c805dc AV |
995 | } |
996 | EXPORT_SYMBOL(filp_open); | |
997 | ||
73d049a4 | 998 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
975abf7f | 999 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1000 | { |
1001 | struct open_flags op; | |
975abf7f | 1002 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1003 | if (err) |
1004 | return ERR_PTR(err); | |
f9652e10 | 1005 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1006 | } |
1007 | EXPORT_SYMBOL(file_open_root); | |
1008 | ||
a218d0fd | 1009 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1010 | { |
47c805dc | 1011 | struct open_flags op; |
f9652e10 AV |
1012 | int fd = build_open_flags(flags, mode, &op); |
1013 | struct filename *tmp; | |
1014 | ||
1015 | if (fd) | |
1016 | return fd; | |
1017 | ||
1018 | tmp = getname(filename); | |
1019 | if (IS_ERR(tmp)) | |
1020 | return PTR_ERR(tmp); | |
1021 | ||
1022 | fd = get_unused_fd_flags(flags); | |
1023 | if (fd >= 0) { | |
1024 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1025 | if (IS_ERR(f)) { | |
1026 | put_unused_fd(fd); | |
1027 | fd = PTR_ERR(f); | |
1028 | } else { | |
1029 | fsnotify_open(f); | |
1030 | fd_install(fd, f); | |
ccae47c6 | 1031 | trace_do_sys_open(tmp->name, flags, mode); |
1da177e4 | 1032 | } |
1da177e4 | 1033 | } |
f9652e10 | 1034 | putname(tmp); |
1da177e4 | 1035 | return fd; |
1da177e4 | 1036 | } |
e922efc3 | 1037 | |
a218d0fd | 1038 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1039 | { |
1040 | if (force_o_largefile()) | |
1041 | flags |= O_LARGEFILE; | |
1042 | ||
2cf09666 | 1043 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1044 | } |
1da177e4 | 1045 | |
6559eed8 | 1046 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1047 | umode_t, mode) |
5590ff0d UD |
1048 | { |
1049 | if (force_o_largefile()) | |
1050 | flags |= O_LARGEFILE; | |
1051 | ||
2cf09666 | 1052 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1053 | } |
5590ff0d | 1054 | |
1da177e4 LT |
1055 | #ifndef __alpha__ |
1056 | ||
1057 | /* | |
1058 | * For backward compatibility? Maybe this should be moved | |
1059 | * into arch/i386 instead? | |
1060 | */ | |
a218d0fd | 1061 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1062 | { |
1063 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1064 | } | |
1065 | ||
1066 | #endif | |
1067 | ||
1068 | /* | |
1069 | * "id" is the POSIX thread ID. We use the | |
1070 | * files pointer for this.. | |
1071 | */ | |
1072 | int filp_close(struct file *filp, fl_owner_t id) | |
1073 | { | |
45778ca8 | 1074 | int retval = 0; |
1da177e4 LT |
1075 | |
1076 | if (!file_count(filp)) { | |
1077 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1078 | return 0; |
1da177e4 LT |
1079 | } |
1080 | ||
72c2d531 | 1081 | if (filp->f_op->flush) |
75e1fcc0 | 1082 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1083 | |
1abf0c71 AV |
1084 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1085 | dnotify_flush(filp, id); | |
1086 | locks_remove_posix(filp, id); | |
1087 | } | |
1da177e4 LT |
1088 | fput(filp); |
1089 | return retval; | |
1090 | } | |
1091 | ||
1092 | EXPORT_SYMBOL(filp_close); | |
1093 | ||
1094 | /* | |
1095 | * Careful here! We test whether the file pointer is NULL before | |
1096 | * releasing the fd. This ensures that one clone task can't release | |
1097 | * an fd while another clone is opening it. | |
1098 | */ | |
ca013e94 | 1099 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1100 | { |
483ce1d4 | 1101 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1102 | |
1103 | /* can't restart close syscall because file table entry was cleared */ | |
1104 | if (unlikely(retval == -ERESTARTSYS || | |
1105 | retval == -ERESTARTNOINTR || | |
1106 | retval == -ERESTARTNOHAND || | |
1107 | retval == -ERESTART_RESTARTBLOCK)) | |
1108 | retval = -EINTR; | |
1109 | ||
1110 | return retval; | |
1da177e4 | 1111 | } |
1da177e4 LT |
1112 | EXPORT_SYMBOL(sys_close); |
1113 | ||
1114 | /* | |
1115 | * This routine simulates a hangup on the tty, to arrange that users | |
1116 | * are given clean terminals at login time. | |
1117 | */ | |
ca013e94 | 1118 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1119 | { |
1120 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1121 | tty_vhangup_self(); |
1da177e4 LT |
1122 | return 0; |
1123 | } | |
1124 | return -EPERM; | |
1125 | } | |
1126 | ||
1127 | /* | |
1128 | * Called when an inode is about to be open. | |
1129 | * We use this to disallow opening large files on 32bit systems if | |
1130 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1131 | * on this flag in sys_open. | |
1132 | */ | |
1133 | int generic_file_open(struct inode * inode, struct file * filp) | |
1134 | { | |
1135 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1136 | return -EOVERFLOW; |
1da177e4 LT |
1137 | return 0; |
1138 | } | |
1139 | ||
1140 | EXPORT_SYMBOL(generic_file_open); | |
1141 | ||
1142 | /* | |
1143 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1144 | * file descriptors. The function is not supposed to ever fail, the only |
1145 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1146 | * directly into file_operations structure. | |
1da177e4 LT |
1147 | */ |
1148 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1149 | { | |
1150 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1151 | return 0; | |
1152 | } | |
1153 | ||
1154 | EXPORT_SYMBOL(nonseekable_open); |