]>
Commit | Line | Data |
---|---|---|
1 | // SPDX-License-Identifier: GPL-2.0-only | |
2 | /* | |
3 | * linux/fs/open.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | #include <linux/string.h> | |
9 | #include <linux/mm.h> | |
10 | #include <linux/file.h> | |
11 | #include <linux/fdtable.h> | |
12 | #include <linux/fsnotify.h> | |
13 | #include <linux/module.h> | |
14 | #include <linux/tty.h> | |
15 | #include <linux/namei.h> | |
16 | #include <linux/backing-dev.h> | |
17 | #include <linux/capability.h> | |
18 | #include <linux/securebits.h> | |
19 | #include <linux/security.h> | |
20 | #include <linux/mount.h> | |
21 | #include <linux/fcntl.h> | |
22 | #include <linux/slab.h> | |
23 | #include <linux/uaccess.h> | |
24 | #include <linux/fs.h> | |
25 | #include <linux/personality.h> | |
26 | #include <linux/pagemap.h> | |
27 | #include <linux/syscalls.h> | |
28 | #include <linux/rcupdate.h> | |
29 | #include <linux/audit.h> | |
30 | #include <linux/falloc.h> | |
31 | #include <linux/fs_struct.h> | |
32 | #include <linux/ima.h> | |
33 | #include <linux/dnotify.h> | |
34 | #include <linux/compat.h> | |
35 | ||
36 | #include "internal.h" | |
37 | ||
38 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, | |
39 | struct file *filp) | |
40 | { | |
41 | int ret; | |
42 | struct iattr newattrs; | |
43 | ||
44 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
45 | if (length < 0) | |
46 | return -EINVAL; | |
47 | ||
48 | newattrs.ia_size = length; | |
49 | newattrs.ia_valid = ATTR_SIZE | time_attrs; | |
50 | if (filp) { | |
51 | newattrs.ia_file = filp; | |
52 | newattrs.ia_valid |= ATTR_FILE; | |
53 | } | |
54 | ||
55 | /* Remove suid, sgid, and file capabilities on truncate too */ | |
56 | ret = dentry_needs_remove_privs(dentry); | |
57 | if (ret < 0) | |
58 | return ret; | |
59 | if (ret) | |
60 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
61 | ||
62 | inode_lock(dentry->d_inode); | |
63 | /* Note any delegations or leases have already been broken: */ | |
64 | ret = notify_change(dentry, &newattrs, NULL); | |
65 | inode_unlock(dentry->d_inode); | |
66 | return ret; | |
67 | } | |
68 | ||
69 | long vfs_truncate(const struct path *path, loff_t length) | |
70 | { | |
71 | struct inode *inode; | |
72 | long error; | |
73 | ||
74 | inode = path->dentry->d_inode; | |
75 | ||
76 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
77 | if (S_ISDIR(inode->i_mode)) | |
78 | return -EISDIR; | |
79 | if (!S_ISREG(inode->i_mode)) | |
80 | return -EINVAL; | |
81 | ||
82 | error = mnt_want_write(path->mnt); | |
83 | if (error) | |
84 | goto out; | |
85 | ||
86 | error = inode_permission(inode, MAY_WRITE); | |
87 | if (error) | |
88 | goto mnt_drop_write_and_out; | |
89 | ||
90 | error = -EPERM; | |
91 | if (IS_APPEND(inode)) | |
92 | goto mnt_drop_write_and_out; | |
93 | ||
94 | error = get_write_access(inode); | |
95 | if (error) | |
96 | goto mnt_drop_write_and_out; | |
97 | ||
98 | /* | |
99 | * Make sure that there are no leases. get_write_access() protects | |
100 | * against the truncate racing with a lease-granting setlease(). | |
101 | */ | |
102 | error = break_lease(inode, O_WRONLY); | |
103 | if (error) | |
104 | goto put_write_and_out; | |
105 | ||
106 | error = locks_verify_truncate(inode, NULL, length); | |
107 | if (!error) | |
108 | error = security_path_truncate(path); | |
109 | if (!error) | |
110 | error = do_truncate(path->dentry, length, 0, NULL); | |
111 | ||
112 | put_write_and_out: | |
113 | put_write_access(inode); | |
114 | mnt_drop_write_and_out: | |
115 | mnt_drop_write(path->mnt); | |
116 | out: | |
117 | return error; | |
118 | } | |
119 | EXPORT_SYMBOL_GPL(vfs_truncate); | |
120 | ||
121 | long do_sys_truncate(const char __user *pathname, loff_t length) | |
122 | { | |
123 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
124 | struct path path; | |
125 | int error; | |
126 | ||
127 | if (length < 0) /* sorry, but loff_t says... */ | |
128 | return -EINVAL; | |
129 | ||
130 | retry: | |
131 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
132 | if (!error) { | |
133 | error = vfs_truncate(&path, length); | |
134 | path_put(&path); | |
135 | } | |
136 | if (retry_estale(error, lookup_flags)) { | |
137 | lookup_flags |= LOOKUP_REVAL; | |
138 | goto retry; | |
139 | } | |
140 | return error; | |
141 | } | |
142 | ||
143 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) | |
144 | { | |
145 | return do_sys_truncate(path, length); | |
146 | } | |
147 | ||
148 | #ifdef CONFIG_COMPAT | |
149 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
150 | { | |
151 | return do_sys_truncate(path, length); | |
152 | } | |
153 | #endif | |
154 | ||
155 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) | |
156 | { | |
157 | struct inode *inode; | |
158 | struct dentry *dentry; | |
159 | struct fd f; | |
160 | int error; | |
161 | ||
162 | error = -EINVAL; | |
163 | if (length < 0) | |
164 | goto out; | |
165 | error = -EBADF; | |
166 | f = fdget(fd); | |
167 | if (!f.file) | |
168 | goto out; | |
169 | ||
170 | /* explicitly opened as large or we are on 64-bit box */ | |
171 | if (f.file->f_flags & O_LARGEFILE) | |
172 | small = 0; | |
173 | ||
174 | dentry = f.file->f_path.dentry; | |
175 | inode = dentry->d_inode; | |
176 | error = -EINVAL; | |
177 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) | |
178 | goto out_putf; | |
179 | ||
180 | error = -EINVAL; | |
181 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
182 | if (small && length > MAX_NON_LFS) | |
183 | goto out_putf; | |
184 | ||
185 | error = -EPERM; | |
186 | /* Check IS_APPEND on real upper inode */ | |
187 | if (IS_APPEND(file_inode(f.file))) | |
188 | goto out_putf; | |
189 | ||
190 | sb_start_write(inode->i_sb); | |
191 | error = locks_verify_truncate(inode, f.file, length); | |
192 | if (!error) | |
193 | error = security_path_truncate(&f.file->f_path); | |
194 | if (!error) | |
195 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); | |
196 | sb_end_write(inode->i_sb); | |
197 | out_putf: | |
198 | fdput(f); | |
199 | out: | |
200 | return error; | |
201 | } | |
202 | ||
203 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) | |
204 | { | |
205 | return do_sys_ftruncate(fd, length, 1); | |
206 | } | |
207 | ||
208 | #ifdef CONFIG_COMPAT | |
209 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
210 | { | |
211 | return do_sys_ftruncate(fd, length, 1); | |
212 | } | |
213 | #endif | |
214 | ||
215 | /* LFS versions of truncate are only needed on 32 bit machines */ | |
216 | #if BITS_PER_LONG == 32 | |
217 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) | |
218 | { | |
219 | return do_sys_truncate(path, length); | |
220 | } | |
221 | ||
222 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) | |
223 | { | |
224 | return do_sys_ftruncate(fd, length, 0); | |
225 | } | |
226 | #endif /* BITS_PER_LONG == 32 */ | |
227 | ||
228 | ||
229 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
230 | { | |
231 | struct inode *inode = file_inode(file); | |
232 | long ret; | |
233 | ||
234 | if (offset < 0 || len <= 0) | |
235 | return -EINVAL; | |
236 | ||
237 | /* Return error if mode is not supported */ | |
238 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) | |
239 | return -EOPNOTSUPP; | |
240 | ||
241 | /* Punch hole and zero range are mutually exclusive */ | |
242 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
243 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
244 | return -EOPNOTSUPP; | |
245 | ||
246 | /* Punch hole must have keep size set */ | |
247 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
248 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
249 | return -EOPNOTSUPP; | |
250 | ||
251 | /* Collapse range should only be used exclusively. */ | |
252 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
253 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
254 | return -EINVAL; | |
255 | ||
256 | /* Insert range should only be used exclusively. */ | |
257 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
258 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
259 | return -EINVAL; | |
260 | ||
261 | /* Unshare range should only be used with allocate mode. */ | |
262 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
263 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
264 | return -EINVAL; | |
265 | ||
266 | if (!(file->f_mode & FMODE_WRITE)) | |
267 | return -EBADF; | |
268 | ||
269 | /* | |
270 | * We can only allow pure fallocate on append only files | |
271 | */ | |
272 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) | |
273 | return -EPERM; | |
274 | ||
275 | if (IS_IMMUTABLE(inode)) | |
276 | return -EPERM; | |
277 | ||
278 | /* | |
279 | * We cannot allow any fallocate operation on an active swapfile | |
280 | */ | |
281 | if (IS_SWAPFILE(inode)) | |
282 | return -ETXTBSY; | |
283 | ||
284 | /* | |
285 | * Revalidate the write permissions, in case security policy has | |
286 | * changed since the files were opened. | |
287 | */ | |
288 | ret = security_file_permission(file, MAY_WRITE); | |
289 | if (ret) | |
290 | return ret; | |
291 | ||
292 | if (S_ISFIFO(inode->i_mode)) | |
293 | return -ESPIPE; | |
294 | ||
295 | if (S_ISDIR(inode->i_mode)) | |
296 | return -EISDIR; | |
297 | ||
298 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
299 | return -ENODEV; | |
300 | ||
301 | /* Check for wrap through zero too */ | |
302 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
303 | return -EFBIG; | |
304 | ||
305 | if (!file->f_op->fallocate) | |
306 | return -EOPNOTSUPP; | |
307 | ||
308 | file_start_write(file); | |
309 | ret = file->f_op->fallocate(file, mode, offset, len); | |
310 | ||
311 | /* | |
312 | * Create inotify and fanotify events. | |
313 | * | |
314 | * To keep the logic simple always create events if fallocate succeeds. | |
315 | * This implies that events are even created if the file size remains | |
316 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
317 | */ | |
318 | if (ret == 0) | |
319 | fsnotify_modify(file); | |
320 | ||
321 | file_end_write(file); | |
322 | return ret; | |
323 | } | |
324 | EXPORT_SYMBOL_GPL(vfs_fallocate); | |
325 | ||
326 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) | |
327 | { | |
328 | struct fd f = fdget(fd); | |
329 | int error = -EBADF; | |
330 | ||
331 | if (f.file) { | |
332 | error = vfs_fallocate(f.file, mode, offset, len); | |
333 | fdput(f); | |
334 | } | |
335 | return error; | |
336 | } | |
337 | ||
338 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) | |
339 | { | |
340 | return ksys_fallocate(fd, mode, offset, len); | |
341 | } | |
342 | ||
343 | /* | |
344 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
345 | * We do this by temporarily clearing all FS-related capabilities and | |
346 | * switching the fsuid/fsgid around to the real ones. | |
347 | */ | |
348 | static const struct cred *access_override_creds(void) | |
349 | { | |
350 | const struct cred *old_cred; | |
351 | struct cred *override_cred; | |
352 | ||
353 | override_cred = prepare_creds(); | |
354 | if (!override_cred) | |
355 | return NULL; | |
356 | ||
357 | override_cred->fsuid = override_cred->uid; | |
358 | override_cred->fsgid = override_cred->gid; | |
359 | ||
360 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
361 | /* Clear the capabilities if we switch to a non-root user */ | |
362 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); | |
363 | if (!uid_eq(override_cred->uid, root_uid)) | |
364 | cap_clear(override_cred->cap_effective); | |
365 | else | |
366 | override_cred->cap_effective = | |
367 | override_cred->cap_permitted; | |
368 | } | |
369 | ||
370 | /* | |
371 | * The new set of credentials can *only* be used in | |
372 | * task-synchronous circumstances, and does not need | |
373 | * RCU freeing, unless somebody then takes a separate | |
374 | * reference to it. | |
375 | * | |
376 | * NOTE! This is _only_ true because this credential | |
377 | * is used purely for override_creds() that installs | |
378 | * it as the subjective cred. Other threads will be | |
379 | * accessing ->real_cred, not the subjective cred. | |
380 | * | |
381 | * If somebody _does_ make a copy of this (using the | |
382 | * 'get_current_cred()' function), that will clear the | |
383 | * non_rcu field, because now that other user may be | |
384 | * expecting RCU freeing. But normal thread-synchronous | |
385 | * cred accesses will keep things non-RCY. | |
386 | */ | |
387 | override_cred->non_rcu = 1; | |
388 | ||
389 | old_cred = override_creds(override_cred); | |
390 | ||
391 | /* override_cred() gets its own ref */ | |
392 | put_cred(override_cred); | |
393 | ||
394 | return old_cred; | |
395 | } | |
396 | ||
397 | long do_faccessat(int dfd, const char __user *filename, int mode) | |
398 | { | |
399 | struct path path; | |
400 | struct inode *inode; | |
401 | int res; | |
402 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
403 | const struct cred *old_cred; | |
404 | ||
405 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
406 | return -EINVAL; | |
407 | ||
408 | old_cred = access_override_creds(); | |
409 | if (!old_cred) | |
410 | return -ENOMEM; | |
411 | ||
412 | retry: | |
413 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
414 | if (res) | |
415 | goto out; | |
416 | ||
417 | inode = d_backing_inode(path.dentry); | |
418 | ||
419 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
420 | /* | |
421 | * MAY_EXEC on regular files is denied if the fs is mounted | |
422 | * with the "noexec" flag. | |
423 | */ | |
424 | res = -EACCES; | |
425 | if (path_noexec(&path)) | |
426 | goto out_path_release; | |
427 | } | |
428 | ||
429 | res = inode_permission(inode, mode | MAY_ACCESS); | |
430 | /* SuS v2 requires we report a read only fs too */ | |
431 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) | |
432 | goto out_path_release; | |
433 | /* | |
434 | * This is a rare case where using __mnt_is_readonly() | |
435 | * is OK without a mnt_want/drop_write() pair. Since | |
436 | * no actual write to the fs is performed here, we do | |
437 | * not need to telegraph to that to anyone. | |
438 | * | |
439 | * By doing this, we accept that this access is | |
440 | * inherently racy and know that the fs may change | |
441 | * state before we even see this result. | |
442 | */ | |
443 | if (__mnt_is_readonly(path.mnt)) | |
444 | res = -EROFS; | |
445 | ||
446 | out_path_release: | |
447 | path_put(&path); | |
448 | if (retry_estale(res, lookup_flags)) { | |
449 | lookup_flags |= LOOKUP_REVAL; | |
450 | goto retry; | |
451 | } | |
452 | out: | |
453 | revert_creds(old_cred); | |
454 | return res; | |
455 | } | |
456 | ||
457 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) | |
458 | { | |
459 | return do_faccessat(dfd, filename, mode); | |
460 | } | |
461 | ||
462 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) | |
463 | { | |
464 | return do_faccessat(AT_FDCWD, filename, mode); | |
465 | } | |
466 | ||
467 | int ksys_chdir(const char __user *filename) | |
468 | { | |
469 | struct path path; | |
470 | int error; | |
471 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
472 | retry: | |
473 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
474 | if (error) | |
475 | goto out; | |
476 | ||
477 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); | |
478 | if (error) | |
479 | goto dput_and_out; | |
480 | ||
481 | set_fs_pwd(current->fs, &path); | |
482 | ||
483 | dput_and_out: | |
484 | path_put(&path); | |
485 | if (retry_estale(error, lookup_flags)) { | |
486 | lookup_flags |= LOOKUP_REVAL; | |
487 | goto retry; | |
488 | } | |
489 | out: | |
490 | return error; | |
491 | } | |
492 | ||
493 | SYSCALL_DEFINE1(chdir, const char __user *, filename) | |
494 | { | |
495 | return ksys_chdir(filename); | |
496 | } | |
497 | ||
498 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) | |
499 | { | |
500 | struct fd f = fdget_raw(fd); | |
501 | int error; | |
502 | ||
503 | error = -EBADF; | |
504 | if (!f.file) | |
505 | goto out; | |
506 | ||
507 | error = -ENOTDIR; | |
508 | if (!d_can_lookup(f.file->f_path.dentry)) | |
509 | goto out_putf; | |
510 | ||
511 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); | |
512 | if (!error) | |
513 | set_fs_pwd(current->fs, &f.file->f_path); | |
514 | out_putf: | |
515 | fdput(f); | |
516 | out: | |
517 | return error; | |
518 | } | |
519 | ||
520 | int ksys_chroot(const char __user *filename) | |
521 | { | |
522 | struct path path; | |
523 | int error; | |
524 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
525 | retry: | |
526 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
527 | if (error) | |
528 | goto out; | |
529 | ||
530 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); | |
531 | if (error) | |
532 | goto dput_and_out; | |
533 | ||
534 | error = -EPERM; | |
535 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) | |
536 | goto dput_and_out; | |
537 | error = security_path_chroot(&path); | |
538 | if (error) | |
539 | goto dput_and_out; | |
540 | ||
541 | set_fs_root(current->fs, &path); | |
542 | error = 0; | |
543 | dput_and_out: | |
544 | path_put(&path); | |
545 | if (retry_estale(error, lookup_flags)) { | |
546 | lookup_flags |= LOOKUP_REVAL; | |
547 | goto retry; | |
548 | } | |
549 | out: | |
550 | return error; | |
551 | } | |
552 | ||
553 | SYSCALL_DEFINE1(chroot, const char __user *, filename) | |
554 | { | |
555 | return ksys_chroot(filename); | |
556 | } | |
557 | ||
558 | static int chmod_common(const struct path *path, umode_t mode) | |
559 | { | |
560 | struct inode *inode = path->dentry->d_inode; | |
561 | struct inode *delegated_inode = NULL; | |
562 | struct iattr newattrs; | |
563 | int error; | |
564 | ||
565 | error = mnt_want_write(path->mnt); | |
566 | if (error) | |
567 | return error; | |
568 | retry_deleg: | |
569 | inode_lock(inode); | |
570 | error = security_path_chmod(path, mode); | |
571 | if (error) | |
572 | goto out_unlock; | |
573 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); | |
574 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
575 | error = notify_change(path->dentry, &newattrs, &delegated_inode); | |
576 | out_unlock: | |
577 | inode_unlock(inode); | |
578 | if (delegated_inode) { | |
579 | error = break_deleg_wait(&delegated_inode); | |
580 | if (!error) | |
581 | goto retry_deleg; | |
582 | } | |
583 | mnt_drop_write(path->mnt); | |
584 | return error; | |
585 | } | |
586 | ||
587 | int ksys_fchmod(unsigned int fd, umode_t mode) | |
588 | { | |
589 | struct fd f = fdget(fd); | |
590 | int err = -EBADF; | |
591 | ||
592 | if (f.file) { | |
593 | audit_file(f.file); | |
594 | err = chmod_common(&f.file->f_path, mode); | |
595 | fdput(f); | |
596 | } | |
597 | return err; | |
598 | } | |
599 | ||
600 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) | |
601 | { | |
602 | return ksys_fchmod(fd, mode); | |
603 | } | |
604 | ||
605 | int do_fchmodat(int dfd, const char __user *filename, umode_t mode) | |
606 | { | |
607 | struct path path; | |
608 | int error; | |
609 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
610 | retry: | |
611 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
612 | if (!error) { | |
613 | error = chmod_common(&path, mode); | |
614 | path_put(&path); | |
615 | if (retry_estale(error, lookup_flags)) { | |
616 | lookup_flags |= LOOKUP_REVAL; | |
617 | goto retry; | |
618 | } | |
619 | } | |
620 | return error; | |
621 | } | |
622 | ||
623 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, | |
624 | umode_t, mode) | |
625 | { | |
626 | return do_fchmodat(dfd, filename, mode); | |
627 | } | |
628 | ||
629 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) | |
630 | { | |
631 | return do_fchmodat(AT_FDCWD, filename, mode); | |
632 | } | |
633 | ||
634 | static int chown_common(const struct path *path, uid_t user, gid_t group) | |
635 | { | |
636 | struct inode *inode = path->dentry->d_inode; | |
637 | struct inode *delegated_inode = NULL; | |
638 | int error; | |
639 | struct iattr newattrs; | |
640 | kuid_t uid; | |
641 | kgid_t gid; | |
642 | ||
643 | uid = make_kuid(current_user_ns(), user); | |
644 | gid = make_kgid(current_user_ns(), group); | |
645 | ||
646 | retry_deleg: | |
647 | newattrs.ia_valid = ATTR_CTIME; | |
648 | if (user != (uid_t) -1) { | |
649 | if (!uid_valid(uid)) | |
650 | return -EINVAL; | |
651 | newattrs.ia_valid |= ATTR_UID; | |
652 | newattrs.ia_uid = uid; | |
653 | } | |
654 | if (group != (gid_t) -1) { | |
655 | if (!gid_valid(gid)) | |
656 | return -EINVAL; | |
657 | newattrs.ia_valid |= ATTR_GID; | |
658 | newattrs.ia_gid = gid; | |
659 | } | |
660 | if (!S_ISDIR(inode->i_mode)) | |
661 | newattrs.ia_valid |= | |
662 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
663 | inode_lock(inode); | |
664 | error = security_path_chown(path, uid, gid); | |
665 | if (!error) | |
666 | error = notify_change(path->dentry, &newattrs, &delegated_inode); | |
667 | inode_unlock(inode); | |
668 | if (delegated_inode) { | |
669 | error = break_deleg_wait(&delegated_inode); | |
670 | if (!error) | |
671 | goto retry_deleg; | |
672 | } | |
673 | return error; | |
674 | } | |
675 | ||
676 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, | |
677 | int flag) | |
678 | { | |
679 | struct path path; | |
680 | int error = -EINVAL; | |
681 | int lookup_flags; | |
682 | ||
683 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) | |
684 | goto out; | |
685 | ||
686 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; | |
687 | if (flag & AT_EMPTY_PATH) | |
688 | lookup_flags |= LOOKUP_EMPTY; | |
689 | retry: | |
690 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
691 | if (error) | |
692 | goto out; | |
693 | error = mnt_want_write(path.mnt); | |
694 | if (error) | |
695 | goto out_release; | |
696 | error = chown_common(&path, user, group); | |
697 | mnt_drop_write(path.mnt); | |
698 | out_release: | |
699 | path_put(&path); | |
700 | if (retry_estale(error, lookup_flags)) { | |
701 | lookup_flags |= LOOKUP_REVAL; | |
702 | goto retry; | |
703 | } | |
704 | out: | |
705 | return error; | |
706 | } | |
707 | ||
708 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, | |
709 | gid_t, group, int, flag) | |
710 | { | |
711 | return do_fchownat(dfd, filename, user, group, flag); | |
712 | } | |
713 | ||
714 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) | |
715 | { | |
716 | return do_fchownat(AT_FDCWD, filename, user, group, 0); | |
717 | } | |
718 | ||
719 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) | |
720 | { | |
721 | return do_fchownat(AT_FDCWD, filename, user, group, | |
722 | AT_SYMLINK_NOFOLLOW); | |
723 | } | |
724 | ||
725 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) | |
726 | { | |
727 | struct fd f = fdget(fd); | |
728 | int error = -EBADF; | |
729 | ||
730 | if (!f.file) | |
731 | goto out; | |
732 | ||
733 | error = mnt_want_write_file(f.file); | |
734 | if (error) | |
735 | goto out_fput; | |
736 | audit_file(f.file); | |
737 | error = chown_common(&f.file->f_path, user, group); | |
738 | mnt_drop_write_file(f.file); | |
739 | out_fput: | |
740 | fdput(f); | |
741 | out: | |
742 | return error; | |
743 | } | |
744 | ||
745 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) | |
746 | { | |
747 | return ksys_fchown(fd, user, group); | |
748 | } | |
749 | ||
750 | static int do_dentry_open(struct file *f, | |
751 | struct inode *inode, | |
752 | int (*open)(struct inode *, struct file *)) | |
753 | { | |
754 | static const struct file_operations empty_fops = {}; | |
755 | int error; | |
756 | ||
757 | path_get(&f->f_path); | |
758 | f->f_inode = inode; | |
759 | f->f_mapping = inode->i_mapping; | |
760 | ||
761 | /* Ensure that we skip any errors that predate opening of the file */ | |
762 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); | |
763 | ||
764 | if (unlikely(f->f_flags & O_PATH)) { | |
765 | f->f_mode = FMODE_PATH | FMODE_OPENED; | |
766 | f->f_op = &empty_fops; | |
767 | return 0; | |
768 | } | |
769 | ||
770 | /* Any file opened for execve()/uselib() has to be a regular file. */ | |
771 | if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) { | |
772 | error = -EACCES; | |
773 | goto cleanup_file; | |
774 | } | |
775 | ||
776 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { | |
777 | error = get_write_access(inode); | |
778 | if (unlikely(error)) | |
779 | goto cleanup_file; | |
780 | error = __mnt_want_write(f->f_path.mnt); | |
781 | if (unlikely(error)) { | |
782 | put_write_access(inode); | |
783 | goto cleanup_file; | |
784 | } | |
785 | f->f_mode |= FMODE_WRITER; | |
786 | } | |
787 | ||
788 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ | |
789 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
790 | f->f_mode |= FMODE_ATOMIC_POS; | |
791 | ||
792 | f->f_op = fops_get(inode->i_fop); | |
793 | if (WARN_ON(!f->f_op)) { | |
794 | error = -ENODEV; | |
795 | goto cleanup_all; | |
796 | } | |
797 | ||
798 | error = security_file_open(f); | |
799 | if (error) | |
800 | goto cleanup_all; | |
801 | ||
802 | error = break_lease(locks_inode(f), f->f_flags); | |
803 | if (error) | |
804 | goto cleanup_all; | |
805 | ||
806 | /* normally all 3 are set; ->open() can clear them if needed */ | |
807 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
808 | if (!open) | |
809 | open = f->f_op->open; | |
810 | if (open) { | |
811 | error = open(inode, f); | |
812 | if (error) | |
813 | goto cleanup_all; | |
814 | } | |
815 | f->f_mode |= FMODE_OPENED; | |
816 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) | |
817 | i_readcount_inc(inode); | |
818 | if ((f->f_mode & FMODE_READ) && | |
819 | likely(f->f_op->read || f->f_op->read_iter)) | |
820 | f->f_mode |= FMODE_CAN_READ; | |
821 | if ((f->f_mode & FMODE_WRITE) && | |
822 | likely(f->f_op->write || f->f_op->write_iter)) | |
823 | f->f_mode |= FMODE_CAN_WRITE; | |
824 | ||
825 | f->f_write_hint = WRITE_LIFE_NOT_SET; | |
826 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); | |
827 | ||
828 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
829 | ||
830 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
831 | if (f->f_flags & O_DIRECT) { | |
832 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) | |
833 | return -EINVAL; | |
834 | } | |
835 | ||
836 | /* | |
837 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
838 | * cache for this file before processing writes. | |
839 | */ | |
840 | if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping)) | |
841 | truncate_pagecache(inode, 0); | |
842 | ||
843 | return 0; | |
844 | ||
845 | cleanup_all: | |
846 | if (WARN_ON_ONCE(error > 0)) | |
847 | error = -EINVAL; | |
848 | fops_put(f->f_op); | |
849 | if (f->f_mode & FMODE_WRITER) { | |
850 | put_write_access(inode); | |
851 | __mnt_drop_write(f->f_path.mnt); | |
852 | } | |
853 | cleanup_file: | |
854 | path_put(&f->f_path); | |
855 | f->f_path.mnt = NULL; | |
856 | f->f_path.dentry = NULL; | |
857 | f->f_inode = NULL; | |
858 | return error; | |
859 | } | |
860 | ||
861 | /** | |
862 | * finish_open - finish opening a file | |
863 | * @file: file pointer | |
864 | * @dentry: pointer to dentry | |
865 | * @open: open callback | |
866 | * @opened: state of open | |
867 | * | |
868 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
869 | * | |
870 | * If the open callback is set to NULL, then the standard f_op->open() | |
871 | * filesystem callback is substituted. | |
872 | * | |
873 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
874 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
875 | * on it after finish_open(). | |
876 | * | |
877 | * Returns zero on success or -errno if the open failed. | |
878 | */ | |
879 | int finish_open(struct file *file, struct dentry *dentry, | |
880 | int (*open)(struct inode *, struct file *)) | |
881 | { | |
882 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ | |
883 | ||
884 | file->f_path.dentry = dentry; | |
885 | return do_dentry_open(file, d_backing_inode(dentry), open); | |
886 | } | |
887 | EXPORT_SYMBOL(finish_open); | |
888 | ||
889 | /** | |
890 | * finish_no_open - finish ->atomic_open() without opening the file | |
891 | * | |
892 | * @file: file pointer | |
893 | * @dentry: dentry or NULL (as returned from ->lookup()) | |
894 | * | |
895 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
896 | * | |
897 | * NB: unlike finish_open() this function does consume the dentry reference and | |
898 | * the caller need not dput() it. | |
899 | * | |
900 | * Returns "0" which must be the return value of ->atomic_open() after having | |
901 | * called this function. | |
902 | */ | |
903 | int finish_no_open(struct file *file, struct dentry *dentry) | |
904 | { | |
905 | file->f_path.dentry = dentry; | |
906 | return 0; | |
907 | } | |
908 | EXPORT_SYMBOL(finish_no_open); | |
909 | ||
910 | char *file_path(struct file *filp, char *buf, int buflen) | |
911 | { | |
912 | return d_path(&filp->f_path, buf, buflen); | |
913 | } | |
914 | EXPORT_SYMBOL(file_path); | |
915 | ||
916 | /** | |
917 | * vfs_open - open the file at the given path | |
918 | * @path: path to open | |
919 | * @file: newly allocated file with f_flag initialized | |
920 | * @cred: credentials to use | |
921 | */ | |
922 | int vfs_open(const struct path *path, struct file *file) | |
923 | { | |
924 | file->f_path = *path; | |
925 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); | |
926 | } | |
927 | ||
928 | struct file *dentry_open(const struct path *path, int flags, | |
929 | const struct cred *cred) | |
930 | { | |
931 | int error; | |
932 | struct file *f; | |
933 | ||
934 | validate_creds(cred); | |
935 | ||
936 | /* We must always pass in a valid mount pointer. */ | |
937 | BUG_ON(!path->mnt); | |
938 | ||
939 | f = alloc_empty_file(flags, cred); | |
940 | if (!IS_ERR(f)) { | |
941 | error = vfs_open(path, f); | |
942 | if (error) { | |
943 | fput(f); | |
944 | f = ERR_PTR(error); | |
945 | } | |
946 | } | |
947 | return f; | |
948 | } | |
949 | EXPORT_SYMBOL(dentry_open); | |
950 | ||
951 | struct file *open_with_fake_path(const struct path *path, int flags, | |
952 | struct inode *inode, const struct cred *cred) | |
953 | { | |
954 | struct file *f = alloc_empty_file_noaccount(flags, cred); | |
955 | if (!IS_ERR(f)) { | |
956 | int error; | |
957 | ||
958 | f->f_path = *path; | |
959 | error = do_dentry_open(f, inode, NULL); | |
960 | if (error) { | |
961 | fput(f); | |
962 | f = ERR_PTR(error); | |
963 | } | |
964 | } | |
965 | return f; | |
966 | } | |
967 | EXPORT_SYMBOL(open_with_fake_path); | |
968 | ||
969 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) | |
970 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
971 | ||
972 | inline struct open_how build_open_how(int flags, umode_t mode) | |
973 | { | |
974 | struct open_how how = { | |
975 | .flags = flags & VALID_OPEN_FLAGS, | |
976 | .mode = mode & S_IALLUGO, | |
977 | }; | |
978 | ||
979 | /* O_PATH beats everything else. */ | |
980 | if (how.flags & O_PATH) | |
981 | how.flags &= O_PATH_FLAGS; | |
982 | /* Modes should only be set for create-like flags. */ | |
983 | if (!WILL_CREATE(how.flags)) | |
984 | how.mode = 0; | |
985 | return how; | |
986 | } | |
987 | ||
988 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) | |
989 | { | |
990 | int flags = how->flags; | |
991 | int lookup_flags = 0; | |
992 | int acc_mode = ACC_MODE(flags); | |
993 | ||
994 | /* Must never be set by userspace */ | |
995 | flags &= ~(FMODE_NONOTIFY | O_CLOEXEC); | |
996 | ||
997 | /* | |
998 | * Older syscalls implicitly clear all of the invalid flags or argument | |
999 | * values before calling build_open_flags(), but openat2(2) checks all | |
1000 | * of its arguments. | |
1001 | */ | |
1002 | if (flags & ~VALID_OPEN_FLAGS) | |
1003 | return -EINVAL; | |
1004 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1005 | return -EINVAL; | |
1006 | ||
1007 | /* Deal with the mode. */ | |
1008 | if (WILL_CREATE(flags)) { | |
1009 | if (how->mode & ~S_IALLUGO) | |
1010 | return -EINVAL; | |
1011 | op->mode = how->mode | S_IFREG; | |
1012 | } else { | |
1013 | if (how->mode != 0) | |
1014 | return -EINVAL; | |
1015 | op->mode = 0; | |
1016 | } | |
1017 | ||
1018 | /* | |
1019 | * In order to ensure programs get explicit errors when trying to use | |
1020 | * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it | |
1021 | * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we | |
1022 | * have to require userspace to explicitly set it. | |
1023 | */ | |
1024 | if (flags & __O_TMPFILE) { | |
1025 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
1026 | return -EINVAL; | |
1027 | if (!(acc_mode & MAY_WRITE)) | |
1028 | return -EINVAL; | |
1029 | } | |
1030 | if (flags & O_PATH) { | |
1031 | /* O_PATH only permits certain other flags to be set. */ | |
1032 | if (flags & ~O_PATH_FLAGS) | |
1033 | return -EINVAL; | |
1034 | acc_mode = 0; | |
1035 | } | |
1036 | ||
1037 | /* | |
1038 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1039 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1040 | * always set instead of having to deal with possibly weird behaviour | |
1041 | * for malicious applications setting only __O_SYNC. | |
1042 | */ | |
1043 | if (flags & __O_SYNC) | |
1044 | flags |= O_DSYNC; | |
1045 | ||
1046 | op->open_flag = flags; | |
1047 | ||
1048 | /* O_TRUNC implies we need access checks for write permissions */ | |
1049 | if (flags & O_TRUNC) | |
1050 | acc_mode |= MAY_WRITE; | |
1051 | ||
1052 | /* Allow the LSM permission hook to distinguish append | |
1053 | access from general write access. */ | |
1054 | if (flags & O_APPEND) | |
1055 | acc_mode |= MAY_APPEND; | |
1056 | ||
1057 | op->acc_mode = acc_mode; | |
1058 | ||
1059 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; | |
1060 | ||
1061 | if (flags & O_CREAT) { | |
1062 | op->intent |= LOOKUP_CREATE; | |
1063 | if (flags & O_EXCL) { | |
1064 | op->intent |= LOOKUP_EXCL; | |
1065 | flags |= O_NOFOLLOW; | |
1066 | } | |
1067 | } | |
1068 | ||
1069 | if (flags & O_DIRECTORY) | |
1070 | lookup_flags |= LOOKUP_DIRECTORY; | |
1071 | if (!(flags & O_NOFOLLOW)) | |
1072 | lookup_flags |= LOOKUP_FOLLOW; | |
1073 | ||
1074 | if (how->resolve & RESOLVE_NO_XDEV) | |
1075 | lookup_flags |= LOOKUP_NO_XDEV; | |
1076 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1077 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1078 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1079 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1080 | if (how->resolve & RESOLVE_BENEATH) | |
1081 | lookup_flags |= LOOKUP_BENEATH; | |
1082 | if (how->resolve & RESOLVE_IN_ROOT) | |
1083 | lookup_flags |= LOOKUP_IN_ROOT; | |
1084 | ||
1085 | op->lookup_flags = lookup_flags; | |
1086 | return 0; | |
1087 | } | |
1088 | ||
1089 | /** | |
1090 | * file_open_name - open file and return file pointer | |
1091 | * | |
1092 | * @name: struct filename containing path to open | |
1093 | * @flags: open flags as per the open(2) second argument | |
1094 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1095 | * | |
1096 | * This is the helper to open a file from kernelspace if you really | |
1097 | * have to. But in generally you should not do this, so please move | |
1098 | * along, nothing to see here.. | |
1099 | */ | |
1100 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1101 | { | |
1102 | struct open_flags op; | |
1103 | struct open_how how = build_open_how(flags, mode); | |
1104 | int err = build_open_flags(&how, &op); | |
1105 | if (err) | |
1106 | return ERR_PTR(err); | |
1107 | return do_filp_open(AT_FDCWD, name, &op); | |
1108 | } | |
1109 | ||
1110 | /** | |
1111 | * filp_open - open file and return file pointer | |
1112 | * | |
1113 | * @filename: path to open | |
1114 | * @flags: open flags as per the open(2) second argument | |
1115 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1116 | * | |
1117 | * This is the helper to open a file from kernelspace if you really | |
1118 | * have to. But in generally you should not do this, so please move | |
1119 | * along, nothing to see here.. | |
1120 | */ | |
1121 | struct file *filp_open(const char *filename, int flags, umode_t mode) | |
1122 | { | |
1123 | struct filename *name = getname_kernel(filename); | |
1124 | struct file *file = ERR_CAST(name); | |
1125 | ||
1126 | if (!IS_ERR(name)) { | |
1127 | file = file_open_name(name, flags, mode); | |
1128 | putname(name); | |
1129 | } | |
1130 | return file; | |
1131 | } | |
1132 | EXPORT_SYMBOL(filp_open); | |
1133 | ||
1134 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, | |
1135 | const char *filename, int flags, umode_t mode) | |
1136 | { | |
1137 | struct open_flags op; | |
1138 | struct open_how how = build_open_how(flags, mode); | |
1139 | int err = build_open_flags(&how, &op); | |
1140 | if (err) | |
1141 | return ERR_PTR(err); | |
1142 | return do_file_open_root(dentry, mnt, filename, &op); | |
1143 | } | |
1144 | EXPORT_SYMBOL(file_open_root); | |
1145 | ||
1146 | static long do_sys_openat2(int dfd, const char __user *filename, | |
1147 | struct open_how *how) | |
1148 | { | |
1149 | struct open_flags op; | |
1150 | int fd = build_open_flags(how, &op); | |
1151 | struct filename *tmp; | |
1152 | ||
1153 | if (fd) | |
1154 | return fd; | |
1155 | ||
1156 | tmp = getname(filename); | |
1157 | if (IS_ERR(tmp)) | |
1158 | return PTR_ERR(tmp); | |
1159 | ||
1160 | fd = get_unused_fd_flags(how->flags); | |
1161 | if (fd >= 0) { | |
1162 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1163 | if (IS_ERR(f)) { | |
1164 | put_unused_fd(fd); | |
1165 | fd = PTR_ERR(f); | |
1166 | } else { | |
1167 | fsnotify_open(f); | |
1168 | fd_install(fd, f); | |
1169 | } | |
1170 | } | |
1171 | putname(tmp); | |
1172 | return fd; | |
1173 | } | |
1174 | ||
1175 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) | |
1176 | { | |
1177 | struct open_how how = build_open_how(flags, mode); | |
1178 | return do_sys_openat2(dfd, filename, &how); | |
1179 | } | |
1180 | ||
1181 | ||
1182 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1183 | { | |
1184 | return ksys_open(filename, flags, mode); | |
1185 | } | |
1186 | ||
1187 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, | |
1188 | umode_t, mode) | |
1189 | { | |
1190 | if (force_o_largefile()) | |
1191 | flags |= O_LARGEFILE; | |
1192 | return do_sys_open(dfd, filename, flags, mode); | |
1193 | } | |
1194 | ||
1195 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, | |
1196 | struct open_how __user *, how, size_t, usize) | |
1197 | { | |
1198 | int err; | |
1199 | struct open_how tmp; | |
1200 | ||
1201 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1202 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1203 | ||
1204 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1205 | return -EINVAL; | |
1206 | ||
1207 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1208 | if (err) | |
1209 | return err; | |
1210 | ||
1211 | /* O_LARGEFILE is only allowed for non-O_PATH. */ | |
1212 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1213 | tmp.flags |= O_LARGEFILE; | |
1214 | ||
1215 | return do_sys_openat2(dfd, filename, &tmp); | |
1216 | } | |
1217 | ||
1218 | #ifdef CONFIG_COMPAT | |
1219 | /* | |
1220 | * Exactly like sys_open(), except that it doesn't set the | |
1221 | * O_LARGEFILE flag. | |
1222 | */ | |
1223 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1224 | { | |
1225 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1226 | } | |
1227 | ||
1228 | /* | |
1229 | * Exactly like sys_openat(), except that it doesn't set the | |
1230 | * O_LARGEFILE flag. | |
1231 | */ | |
1232 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1233 | { | |
1234 | return do_sys_open(dfd, filename, flags, mode); | |
1235 | } | |
1236 | #endif | |
1237 | ||
1238 | #ifndef __alpha__ | |
1239 | ||
1240 | /* | |
1241 | * For backward compatibility? Maybe this should be moved | |
1242 | * into arch/i386 instead? | |
1243 | */ | |
1244 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) | |
1245 | { | |
1246 | return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1247 | } | |
1248 | ||
1249 | #endif | |
1250 | ||
1251 | /* | |
1252 | * "id" is the POSIX thread ID. We use the | |
1253 | * files pointer for this.. | |
1254 | */ | |
1255 | int filp_close(struct file *filp, fl_owner_t id) | |
1256 | { | |
1257 | int retval = 0; | |
1258 | ||
1259 | if (!file_count(filp)) { | |
1260 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
1261 | return 0; | |
1262 | } | |
1263 | ||
1264 | if (filp->f_op->flush) | |
1265 | retval = filp->f_op->flush(filp, id); | |
1266 | ||
1267 | if (likely(!(filp->f_mode & FMODE_PATH))) { | |
1268 | dnotify_flush(filp, id); | |
1269 | locks_remove_posix(filp, id); | |
1270 | } | |
1271 | fput(filp); | |
1272 | return retval; | |
1273 | } | |
1274 | ||
1275 | EXPORT_SYMBOL(filp_close); | |
1276 | ||
1277 | /* | |
1278 | * Careful here! We test whether the file pointer is NULL before | |
1279 | * releasing the fd. This ensures that one clone task can't release | |
1280 | * an fd while another clone is opening it. | |
1281 | */ | |
1282 | SYSCALL_DEFINE1(close, unsigned int, fd) | |
1283 | { | |
1284 | int retval = __close_fd(current->files, fd); | |
1285 | ||
1286 | /* can't restart close syscall because file table entry was cleared */ | |
1287 | if (unlikely(retval == -ERESTARTSYS || | |
1288 | retval == -ERESTARTNOINTR || | |
1289 | retval == -ERESTARTNOHAND || | |
1290 | retval == -ERESTART_RESTARTBLOCK)) | |
1291 | retval = -EINTR; | |
1292 | ||
1293 | return retval; | |
1294 | } | |
1295 | ||
1296 | /* | |
1297 | * This routine simulates a hangup on the tty, to arrange that users | |
1298 | * are given clean terminals at login time. | |
1299 | */ | |
1300 | SYSCALL_DEFINE0(vhangup) | |
1301 | { | |
1302 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
1303 | tty_vhangup_self(); | |
1304 | return 0; | |
1305 | } | |
1306 | return -EPERM; | |
1307 | } | |
1308 | ||
1309 | /* | |
1310 | * Called when an inode is about to be open. | |
1311 | * We use this to disallow opening large files on 32bit systems if | |
1312 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1313 | * on this flag in sys_open. | |
1314 | */ | |
1315 | int generic_file_open(struct inode * inode, struct file * filp) | |
1316 | { | |
1317 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
1318 | return -EOVERFLOW; | |
1319 | return 0; | |
1320 | } | |
1321 | ||
1322 | EXPORT_SYMBOL(generic_file_open); | |
1323 | ||
1324 | /* | |
1325 | * This is used by subsystems that don't want seekable | |
1326 | * file descriptors. The function is not supposed to ever fail, the only | |
1327 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1328 | * directly into file_operations structure. | |
1329 | */ | |
1330 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1331 | { | |
1332 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1333 | return 0; | |
1334 | } | |
1335 | ||
1336 | EXPORT_SYMBOL(nonseekable_open); | |
1337 | ||
1338 | /* | |
1339 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1340 | * Such file descriptors are not seekable and don't have notion of position | |
1341 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). | |
1342 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1343 | * can run simultaneously. | |
1344 | * | |
1345 | * stream_open never fails and is marked to return int so that it could be | |
1346 | * directly used as file_operations.open . | |
1347 | */ | |
1348 | int stream_open(struct inode *inode, struct file *filp) | |
1349 | { | |
1350 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); | |
1351 | filp->f_mode |= FMODE_STREAM; | |
1352 | return 0; | |
1353 | } | |
1354 | ||
1355 | EXPORT_SYMBOL(stream_open); |