• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 #include <linux/mnt_idmapping.h>
36 
37 #include "internal.h"
38 #include <trace/hooks/syscall_check.h>
39 
do_truncate(struct user_namespace * mnt_userns,struct dentry * dentry,loff_t length,unsigned int time_attrs,struct file * filp)40 int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry,
41 		loff_t length, unsigned int time_attrs, struct file *filp)
42 {
43 	int ret;
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;
51 	newattrs.ia_valid = ATTR_SIZE | time_attrs;
52 	if (filp) {
53 		newattrs.ia_file = filp;
54 		newattrs.ia_valid |= ATTR_FILE;
55 	}
56 
57 	/* Remove suid, sgid, and file capabilities on truncate too */
58 	ret = dentry_needs_remove_privs(mnt_userns, dentry);
59 	if (ret < 0)
60 		return ret;
61 	if (ret)
62 		newattrs.ia_valid |= ret | ATTR_FORCE;
63 
64 	inode_lock(dentry->d_inode);
65 	/* Note any delegations or leases have already been broken: */
66 	ret = notify_change(mnt_userns, dentry, &newattrs, NULL);
67 	inode_unlock(dentry->d_inode);
68 	return ret;
69 }
70 
vfs_truncate(const struct path * path,loff_t length)71 long vfs_truncate(const struct path *path, loff_t length)
72 {
73 	struct user_namespace *mnt_userns;
74 	struct inode *inode;
75 	long error;
76 
77 	inode = path->dentry->d_inode;
78 
79 	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
80 	if (S_ISDIR(inode->i_mode))
81 		return -EISDIR;
82 	if (!S_ISREG(inode->i_mode))
83 		return -EINVAL;
84 
85 	error = mnt_want_write(path->mnt);
86 	if (error)
87 		goto out;
88 
89 	mnt_userns = mnt_user_ns(path->mnt);
90 	error = inode_permission(mnt_userns, inode, MAY_WRITE);
91 	if (error)
92 		goto mnt_drop_write_and_out;
93 
94 	error = -EPERM;
95 	if (IS_APPEND(inode))
96 		goto mnt_drop_write_and_out;
97 
98 	error = get_write_access(inode);
99 	if (error)
100 		goto mnt_drop_write_and_out;
101 
102 	/*
103 	 * Make sure that there are no leases.  get_write_access() protects
104 	 * against the truncate racing with a lease-granting setlease().
105 	 */
106 	error = break_lease(inode, O_WRONLY);
107 	if (error)
108 		goto put_write_and_out;
109 
110 	error = security_path_truncate(path);
111 	if (!error)
112 		error = do_truncate(mnt_userns, path->dentry, length, 0, NULL);
113 
114 put_write_and_out:
115 	put_write_access(inode);
116 mnt_drop_write_and_out:
117 	mnt_drop_write(path->mnt);
118 out:
119 	return error;
120 }
121 EXPORT_SYMBOL_GPL(vfs_truncate);
122 
do_sys_truncate(const char __user * pathname,loff_t length)123 long do_sys_truncate(const char __user *pathname, loff_t length)
124 {
125 	unsigned int lookup_flags = LOOKUP_FOLLOW;
126 	struct path path;
127 	int error;
128 
129 	if (length < 0)	/* sorry, but loff_t says... */
130 		return -EINVAL;
131 
132 retry:
133 	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
134 	if (!error) {
135 		error = vfs_truncate(&path, length);
136 		path_put(&path);
137 	}
138 	if (retry_estale(error, lookup_flags)) {
139 		lookup_flags |= LOOKUP_REVAL;
140 		goto retry;
141 	}
142 	return error;
143 }
144 
SYSCALL_DEFINE2(truncate,const char __user *,path,long,length)145 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
146 {
147 	return do_sys_truncate(path, length);
148 }
149 
150 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate,const char __user *,path,compat_off_t,length)151 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
152 {
153 	return do_sys_truncate(path, length);
154 }
155 #endif
156 
do_sys_ftruncate(unsigned int fd,loff_t length,int small)157 long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
158 {
159 	struct inode *inode;
160 	struct dentry *dentry;
161 	struct fd f;
162 	int error;
163 
164 	error = -EINVAL;
165 	if (length < 0)
166 		goto out;
167 	error = -EBADF;
168 	f = fdget(fd);
169 	if (!f.file)
170 		goto out;
171 
172 	/* explicitly opened as large or we are on 64-bit box */
173 	if (f.file->f_flags & O_LARGEFILE)
174 		small = 0;
175 
176 	dentry = f.file->f_path.dentry;
177 	inode = dentry->d_inode;
178 	error = -EINVAL;
179 	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
180 		goto out_putf;
181 
182 	error = -EINVAL;
183 	/* Cannot ftruncate over 2^31 bytes without large file support */
184 	if (small && length > MAX_NON_LFS)
185 		goto out_putf;
186 
187 	error = -EPERM;
188 	/* Check IS_APPEND on real upper inode */
189 	if (IS_APPEND(file_inode(f.file)))
190 		goto out_putf;
191 	sb_start_write(inode->i_sb);
192 	error = security_path_truncate(&f.file->f_path);
193 	if (!error)
194 		error = do_truncate(file_mnt_user_ns(f.file), dentry, length,
195 				    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 
SYSCALL_DEFINE2(ftruncate,unsigned int,fd,unsigned long,length)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
COMPAT_SYSCALL_DEFINE2(ftruncate,unsigned int,fd,compat_ulong_t,length)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
SYSCALL_DEFINE2(truncate64,const char __user *,path,loff_t,length)217 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
218 {
219 	return do_sys_truncate(path, length);
220 }
221 
SYSCALL_DEFINE2(ftruncate64,unsigned int,fd,loff_t,length)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 
vfs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)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 
ksys_fallocate(int fd,int mode,loff_t offset,loff_t len)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 
SYSCALL_DEFINE4(fallocate,int,fd,int,mode,loff_t,offset,loff_t,len)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  */
access_override_creds(void)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 
do_faccessat(int dfd,const char __user * filename,int mode,int flags)397 static long do_faccessat(int dfd, const char __user *filename, int mode, int flags)
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 = NULL;
404 
405 	if (mode & ~S_IRWXO)	/* where's F_OK, X_OK, W_OK, R_OK? */
406 		return -EINVAL;
407 
408 	if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH))
409 		return -EINVAL;
410 
411 	if (flags & AT_SYMLINK_NOFOLLOW)
412 		lookup_flags &= ~LOOKUP_FOLLOW;
413 	if (flags & AT_EMPTY_PATH)
414 		lookup_flags |= LOOKUP_EMPTY;
415 
416 	if (!(flags & AT_EACCESS)) {
417 		old_cred = access_override_creds();
418 		if (!old_cred)
419 			return -ENOMEM;
420 	}
421 
422 retry:
423 	res = user_path_at(dfd, filename, lookup_flags, &path);
424 	if (res)
425 		goto out;
426 
427 	inode = d_backing_inode(path.dentry);
428 
429 	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
430 		/*
431 		 * MAY_EXEC on regular files is denied if the fs is mounted
432 		 * with the "noexec" flag.
433 		 */
434 		res = -EACCES;
435 		if (path_noexec(&path))
436 			goto out_path_release;
437 	}
438 
439 	res = inode_permission(mnt_user_ns(path.mnt), inode, mode | MAY_ACCESS);
440 	/* SuS v2 requires we report a read only fs too */
441 	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
442 		goto out_path_release;
443 	/*
444 	 * This is a rare case where using __mnt_is_readonly()
445 	 * is OK without a mnt_want/drop_write() pair.  Since
446 	 * no actual write to the fs is performed here, we do
447 	 * not need to telegraph to that to anyone.
448 	 *
449 	 * By doing this, we accept that this access is
450 	 * inherently racy and know that the fs may change
451 	 * state before we even see this result.
452 	 */
453 	if (__mnt_is_readonly(path.mnt))
454 		res = -EROFS;
455 
456 out_path_release:
457 	path_put(&path);
458 	if (retry_estale(res, lookup_flags)) {
459 		lookup_flags |= LOOKUP_REVAL;
460 		goto retry;
461 	}
462 out:
463 	if (old_cred)
464 		revert_creds(old_cred);
465 
466 	return res;
467 }
468 
SYSCALL_DEFINE3(faccessat,int,dfd,const char __user *,filename,int,mode)469 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
470 {
471 	return do_faccessat(dfd, filename, mode, 0);
472 }
473 
SYSCALL_DEFINE4(faccessat2,int,dfd,const char __user *,filename,int,mode,int,flags)474 SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode,
475 		int, flags)
476 {
477 	return do_faccessat(dfd, filename, mode, flags);
478 }
479 
SYSCALL_DEFINE2(access,const char __user *,filename,int,mode)480 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
481 {
482 	return do_faccessat(AT_FDCWD, filename, mode, 0);
483 }
484 
SYSCALL_DEFINE1(chdir,const char __user *,filename)485 SYSCALL_DEFINE1(chdir, const char __user *, filename)
486 {
487 	struct path path;
488 	int error;
489 	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
490 retry:
491 	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
492 	if (error)
493 		goto out;
494 
495 	error = path_permission(&path, MAY_EXEC | MAY_CHDIR);
496 	if (error)
497 		goto dput_and_out;
498 
499 	set_fs_pwd(current->fs, &path);
500 
501 dput_and_out:
502 	path_put(&path);
503 	if (retry_estale(error, lookup_flags)) {
504 		lookup_flags |= LOOKUP_REVAL;
505 		goto retry;
506 	}
507 out:
508 	return error;
509 }
510 
SYSCALL_DEFINE1(fchdir,unsigned int,fd)511 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
512 {
513 	struct fd f = fdget_raw(fd);
514 	int error;
515 
516 	error = -EBADF;
517 	if (!f.file)
518 		goto out;
519 
520 	error = -ENOTDIR;
521 	if (!d_can_lookup(f.file->f_path.dentry))
522 		goto out_putf;
523 
524 	error = file_permission(f.file, MAY_EXEC | MAY_CHDIR);
525 	if (!error)
526 		set_fs_pwd(current->fs, &f.file->f_path);
527 out_putf:
528 	fdput(f);
529 out:
530 	return error;
531 }
532 
SYSCALL_DEFINE1(chroot,const char __user *,filename)533 SYSCALL_DEFINE1(chroot, const char __user *, filename)
534 {
535 	struct path path;
536 	int error;
537 	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
538 retry:
539 	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
540 	if (error)
541 		goto out;
542 
543 	error = path_permission(&path, MAY_EXEC | MAY_CHDIR);
544 	if (error)
545 		goto dput_and_out;
546 
547 	error = -EPERM;
548 	if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
549 		goto dput_and_out;
550 	error = security_path_chroot(&path);
551 	if (error)
552 		goto dput_and_out;
553 
554 	set_fs_root(current->fs, &path);
555 	error = 0;
556 dput_and_out:
557 	path_put(&path);
558 	if (retry_estale(error, lookup_flags)) {
559 		lookup_flags |= LOOKUP_REVAL;
560 		goto retry;
561 	}
562 out:
563 	return error;
564 }
565 
chmod_common(const struct path * path,umode_t mode)566 int chmod_common(const struct path *path, umode_t mode)
567 {
568 	struct inode *inode = path->dentry->d_inode;
569 	struct inode *delegated_inode = NULL;
570 	struct iattr newattrs;
571 	int error;
572 
573 	error = mnt_want_write(path->mnt);
574 	if (error)
575 		return error;
576 retry_deleg:
577 	inode_lock(inode);
578 	error = security_path_chmod(path, mode);
579 	if (error)
580 		goto out_unlock;
581 	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
582 	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
583 	error = notify_change(mnt_user_ns(path->mnt), path->dentry,
584 			      &newattrs, &delegated_inode);
585 out_unlock:
586 	inode_unlock(inode);
587 	if (delegated_inode) {
588 		error = break_deleg_wait(&delegated_inode);
589 		if (!error)
590 			goto retry_deleg;
591 	}
592 	mnt_drop_write(path->mnt);
593 	return error;
594 }
595 
vfs_fchmod(struct file * file,umode_t mode)596 int vfs_fchmod(struct file *file, umode_t mode)
597 {
598 	audit_file(file);
599 	return chmod_common(&file->f_path, mode);
600 }
601 
SYSCALL_DEFINE2(fchmod,unsigned int,fd,umode_t,mode)602 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
603 {
604 	struct fd f = fdget(fd);
605 	int err = -EBADF;
606 
607 	if (f.file) {
608 		err = vfs_fchmod(f.file, mode);
609 		fdput(f);
610 	}
611 	return err;
612 }
613 
do_fchmodat(int dfd,const char __user * filename,umode_t mode)614 static int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
615 {
616 	struct path path;
617 	int error;
618 	unsigned int lookup_flags = LOOKUP_FOLLOW;
619 retry:
620 	error = user_path_at(dfd, filename, lookup_flags, &path);
621 	if (!error) {
622 		error = chmod_common(&path, mode);
623 		path_put(&path);
624 		if (retry_estale(error, lookup_flags)) {
625 			lookup_flags |= LOOKUP_REVAL;
626 			goto retry;
627 		}
628 	}
629 	return error;
630 }
631 
SYSCALL_DEFINE3(fchmodat,int,dfd,const char __user *,filename,umode_t,mode)632 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
633 		umode_t, mode)
634 {
635 	return do_fchmodat(dfd, filename, mode);
636 }
637 
SYSCALL_DEFINE2(chmod,const char __user *,filename,umode_t,mode)638 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
639 {
640 	return do_fchmodat(AT_FDCWD, filename, mode);
641 }
642 
chown_common(const struct path * path,uid_t user,gid_t group)643 int chown_common(const struct path *path, uid_t user, gid_t group)
644 {
645 	struct user_namespace *mnt_userns, *fs_userns;
646 	struct inode *inode = path->dentry->d_inode;
647 	struct inode *delegated_inode = NULL;
648 	int error;
649 	struct iattr newattrs;
650 	kuid_t uid;
651 	kgid_t gid;
652 
653 	uid = make_kuid(current_user_ns(), user);
654 	gid = make_kgid(current_user_ns(), group);
655 
656 	mnt_userns = mnt_user_ns(path->mnt);
657 	fs_userns = i_user_ns(inode);
658 	uid = mapped_kuid_user(mnt_userns, fs_userns, uid);
659 	gid = mapped_kgid_user(mnt_userns, fs_userns, gid);
660 
661 retry_deleg:
662 	newattrs.ia_valid =  ATTR_CTIME;
663 	if (user != (uid_t) -1) {
664 		if (!uid_valid(uid))
665 			return -EINVAL;
666 		newattrs.ia_valid |= ATTR_UID;
667 		newattrs.ia_uid = uid;
668 	}
669 	if (group != (gid_t) -1) {
670 		if (!gid_valid(gid))
671 			return -EINVAL;
672 		newattrs.ia_valid |= ATTR_GID;
673 		newattrs.ia_gid = gid;
674 	}
675 	inode_lock(inode);
676 	if (!S_ISDIR(inode->i_mode))
677 		newattrs.ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV |
678 				     setattr_should_drop_sgid(mnt_userns, inode);
679 	error = security_path_chown(path, uid, gid);
680 	if (!error)
681 		error = notify_change(mnt_userns, path->dentry, &newattrs,
682 				      &delegated_inode);
683 	inode_unlock(inode);
684 	if (delegated_inode) {
685 		error = break_deleg_wait(&delegated_inode);
686 		if (!error)
687 			goto retry_deleg;
688 	}
689 	return error;
690 }
691 
do_fchownat(int dfd,const char __user * filename,uid_t user,gid_t group,int flag)692 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
693 		int flag)
694 {
695 	struct path path;
696 	int error = -EINVAL;
697 	int lookup_flags;
698 
699 	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
700 		goto out;
701 
702 	lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
703 	if (flag & AT_EMPTY_PATH)
704 		lookup_flags |= LOOKUP_EMPTY;
705 retry:
706 	error = user_path_at(dfd, filename, lookup_flags, &path);
707 	if (error)
708 		goto out;
709 	error = mnt_want_write(path.mnt);
710 	if (error)
711 		goto out_release;
712 	error = chown_common(&path, user, group);
713 	mnt_drop_write(path.mnt);
714 out_release:
715 	path_put(&path);
716 	if (retry_estale(error, lookup_flags)) {
717 		lookup_flags |= LOOKUP_REVAL;
718 		goto retry;
719 	}
720 out:
721 	return error;
722 }
723 
SYSCALL_DEFINE5(fchownat,int,dfd,const char __user *,filename,uid_t,user,gid_t,group,int,flag)724 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
725 		gid_t, group, int, flag)
726 {
727 	return do_fchownat(dfd, filename, user, group, flag);
728 }
729 
SYSCALL_DEFINE3(chown,const char __user *,filename,uid_t,user,gid_t,group)730 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
731 {
732 	return do_fchownat(AT_FDCWD, filename, user, group, 0);
733 }
734 
SYSCALL_DEFINE3(lchown,const char __user *,filename,uid_t,user,gid_t,group)735 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
736 {
737 	return do_fchownat(AT_FDCWD, filename, user, group,
738 			   AT_SYMLINK_NOFOLLOW);
739 }
740 
vfs_fchown(struct file * file,uid_t user,gid_t group)741 int vfs_fchown(struct file *file, uid_t user, gid_t group)
742 {
743 	int error;
744 
745 	error = mnt_want_write_file(file);
746 	if (error)
747 		return error;
748 	audit_file(file);
749 	error = chown_common(&file->f_path, user, group);
750 	mnt_drop_write_file(file);
751 	return error;
752 }
753 
ksys_fchown(unsigned int fd,uid_t user,gid_t group)754 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
755 {
756 	struct fd f = fdget(fd);
757 	int error = -EBADF;
758 
759 	if (f.file) {
760 		error = vfs_fchown(f.file, user, group);
761 		fdput(f);
762 	}
763 	return error;
764 }
765 
SYSCALL_DEFINE3(fchown,unsigned int,fd,uid_t,user,gid_t,group)766 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
767 {
768 	return ksys_fchown(fd, user, group);
769 }
770 
do_dentry_open(struct file * f,struct inode * inode,int (* open)(struct inode *,struct file *))771 static int do_dentry_open(struct file *f,
772 			  struct inode *inode,
773 			  int (*open)(struct inode *, struct file *))
774 {
775 	static const struct file_operations empty_fops = {};
776 	int error;
777 
778 	path_get(&f->f_path);
779 	f->f_inode = inode;
780 	f->f_mapping = inode->i_mapping;
781 	f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
782 	f->f_sb_err = file_sample_sb_err(f);
783 
784 	if (unlikely(f->f_flags & O_PATH)) {
785 		f->f_mode = FMODE_PATH | FMODE_OPENED;
786 		f->f_op = &empty_fops;
787 		return 0;
788 	}
789 
790 	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) {
791 		i_readcount_inc(inode);
792 	} else if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
793 		error = get_write_access(inode);
794 		if (unlikely(error))
795 			goto cleanup_file;
796 		error = __mnt_want_write(f->f_path.mnt);
797 		if (unlikely(error)) {
798 			put_write_access(inode);
799 			goto cleanup_file;
800 		}
801 		f->f_mode |= FMODE_WRITER;
802 	}
803 
804 	/* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
805 	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
806 		f->f_mode |= FMODE_ATOMIC_POS;
807 
808 	f->f_op = fops_get(inode->i_fop);
809 	if (WARN_ON(!f->f_op)) {
810 		error = -ENODEV;
811 		goto cleanup_all;
812 	}
813 	trace_android_vh_check_file_open(f);
814 
815 	error = security_file_open(f);
816 	if (error)
817 		goto cleanup_all;
818 
819 	error = break_lease(locks_inode(f), f->f_flags);
820 	if (error)
821 		goto cleanup_all;
822 
823 	/* normally all 3 are set; ->open() can clear them if needed */
824 	f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
825 	if (!open)
826 		open = f->f_op->open;
827 	if (open) {
828 		error = open(inode, f);
829 		if (error)
830 			goto cleanup_all;
831 	}
832 	f->f_mode |= FMODE_OPENED;
833 	if ((f->f_mode & FMODE_READ) &&
834 	     likely(f->f_op->read || f->f_op->read_iter))
835 		f->f_mode |= FMODE_CAN_READ;
836 	if ((f->f_mode & FMODE_WRITE) &&
837 	     likely(f->f_op->write || f->f_op->write_iter))
838 		f->f_mode |= FMODE_CAN_WRITE;
839 
840 	f->f_write_hint = WRITE_LIFE_NOT_SET;
841 	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
842 
843 	file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
844 
845 	/* NB: we're sure to have correct a_ops only after f_op->open */
846 	if (f->f_flags & O_DIRECT) {
847 		if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
848 			return -EINVAL;
849 	}
850 
851 	/*
852 	 * XXX: Huge page cache doesn't support writing yet. Drop all page
853 	 * cache for this file before processing writes.
854 	 */
855 	if (f->f_mode & FMODE_WRITE) {
856 		/*
857 		 * Paired with smp_mb() in collapse_file() to ensure nr_thps
858 		 * is up to date and the update to i_writecount by
859 		 * get_write_access() is visible. Ensures subsequent insertion
860 		 * of THPs into the page cache will fail.
861 		 */
862 		smp_mb();
863 		if (filemap_nr_thps(inode->i_mapping)) {
864 			struct address_space *mapping = inode->i_mapping;
865 
866 			filemap_invalidate_lock(inode->i_mapping);
867 			/*
868 			 * unmap_mapping_range just need to be called once
869 			 * here, because the private pages is not need to be
870 			 * unmapped mapping (e.g. data segment of dynamic
871 			 * shared libraries here).
872 			 */
873 			unmap_mapping_range(mapping, 0, 0, 0);
874 			truncate_inode_pages(mapping, 0);
875 			filemap_invalidate_unlock(inode->i_mapping);
876 		}
877 	}
878 
879 	return 0;
880 
881 cleanup_all:
882 	if (WARN_ON_ONCE(error > 0))
883 		error = -EINVAL;
884 	fops_put(f->f_op);
885 	put_file_access(f);
886 cleanup_file:
887 	path_put(&f->f_path);
888 	f->f_path.mnt = NULL;
889 	f->f_path.dentry = NULL;
890 	f->f_inode = NULL;
891 	return error;
892 }
893 
894 /**
895  * finish_open - finish opening a file
896  * @file: file pointer
897  * @dentry: pointer to dentry
898  * @open: open callback
899  * @opened: state of open
900  *
901  * This can be used to finish opening a file passed to i_op->atomic_open().
902  *
903  * If the open callback is set to NULL, then the standard f_op->open()
904  * filesystem callback is substituted.
905  *
906  * NB: the dentry reference is _not_ consumed.  If, for example, the dentry is
907  * the return value of d_splice_alias(), then the caller needs to perform dput()
908  * on it after finish_open().
909  *
910  * Returns zero on success or -errno if the open failed.
911  */
finish_open(struct file * file,struct dentry * dentry,int (* open)(struct inode *,struct file *))912 int finish_open(struct file *file, struct dentry *dentry,
913 		int (*open)(struct inode *, struct file *))
914 {
915 	BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
916 
917 	file->f_path.dentry = dentry;
918 	return do_dentry_open(file, d_backing_inode(dentry), open);
919 }
920 EXPORT_SYMBOL(finish_open);
921 
922 /**
923  * finish_no_open - finish ->atomic_open() without opening the file
924  *
925  * @file: file pointer
926  * @dentry: dentry or NULL (as returned from ->lookup())
927  *
928  * This can be used to set the result of a successful lookup in ->atomic_open().
929  *
930  * NB: unlike finish_open() this function does consume the dentry reference and
931  * the caller need not dput() it.
932  *
933  * Returns "0" which must be the return value of ->atomic_open() after having
934  * called this function.
935  */
finish_no_open(struct file * file,struct dentry * dentry)936 int finish_no_open(struct file *file, struct dentry *dentry)
937 {
938 	file->f_path.dentry = dentry;
939 	return 0;
940 }
941 EXPORT_SYMBOL(finish_no_open);
942 
file_path(struct file * filp,char * buf,int buflen)943 char *file_path(struct file *filp, char *buf, int buflen)
944 {
945 	return d_path(&filp->f_path, buf, buflen);
946 }
947 EXPORT_SYMBOL(file_path);
948 
949 /**
950  * vfs_open - open the file at the given path
951  * @path: path to open
952  * @file: newly allocated file with f_flag initialized
953  * @cred: credentials to use
954  */
vfs_open(const struct path * path,struct file * file)955 int vfs_open(const struct path *path, struct file *file)
956 {
957 	file->f_path = *path;
958 	return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
959 }
960 
dentry_open(const struct path * path,int flags,const struct cred * cred)961 struct file *dentry_open(const struct path *path, int flags,
962 			 const struct cred *cred)
963 {
964 	int error;
965 	struct file *f;
966 
967 	validate_creds(cred);
968 
969 	/* We must always pass in a valid mount pointer. */
970 	BUG_ON(!path->mnt);
971 
972 	f = alloc_empty_file(flags, cred);
973 	if (!IS_ERR(f)) {
974 		error = vfs_open(path, f);
975 		if (error) {
976 			fput(f);
977 			f = ERR_PTR(error);
978 		}
979 	}
980 	return f;
981 }
982 EXPORT_SYMBOL_NS(dentry_open, ANDROID_GKI_VFS_EXPORT_ONLY);
983 
open_with_fake_path(const struct path * path,int flags,struct inode * inode,const struct cred * cred)984 struct file *open_with_fake_path(const struct path *path, int flags,
985 				struct inode *inode, const struct cred *cred)
986 {
987 	struct file *f = alloc_empty_file_noaccount(flags, cred);
988 	if (!IS_ERR(f)) {
989 		int error;
990 
991 		f->f_path = *path;
992 		error = do_dentry_open(f, inode, NULL);
993 		if (error) {
994 			fput(f);
995 			f = ERR_PTR(error);
996 		}
997 	}
998 	return f;
999 }
1000 EXPORT_SYMBOL(open_with_fake_path);
1001 
1002 #define WILL_CREATE(flags)	(flags & (O_CREAT | __O_TMPFILE))
1003 #define O_PATH_FLAGS		(O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC)
1004 
build_open_how(int flags,umode_t mode)1005 inline struct open_how build_open_how(int flags, umode_t mode)
1006 {
1007 	struct open_how how = {
1008 		.flags = flags & VALID_OPEN_FLAGS,
1009 		.mode = mode & S_IALLUGO,
1010 	};
1011 
1012 	/* O_PATH beats everything else. */
1013 	if (how.flags & O_PATH)
1014 		how.flags &= O_PATH_FLAGS;
1015 	/* Modes should only be set for create-like flags. */
1016 	if (!WILL_CREATE(how.flags))
1017 		how.mode = 0;
1018 	return how;
1019 }
1020 
build_open_flags(const struct open_how * how,struct open_flags * op)1021 inline int build_open_flags(const struct open_how *how, struct open_flags *op)
1022 {
1023 	u64 flags = how->flags;
1024 	u64 strip = FMODE_NONOTIFY | O_CLOEXEC;
1025 	int lookup_flags = 0;
1026 	int acc_mode = ACC_MODE(flags);
1027 
1028 	BUILD_BUG_ON_MSG(upper_32_bits(VALID_OPEN_FLAGS),
1029 			 "struct open_flags doesn't yet handle flags > 32 bits");
1030 
1031 	/*
1032 	 * Strip flags that either shouldn't be set by userspace like
1033 	 * FMODE_NONOTIFY or that aren't relevant in determining struct
1034 	 * open_flags like O_CLOEXEC.
1035 	 */
1036 	flags &= ~strip;
1037 
1038 	/*
1039 	 * Older syscalls implicitly clear all of the invalid flags or argument
1040 	 * values before calling build_open_flags(), but openat2(2) checks all
1041 	 * of its arguments.
1042 	 */
1043 	if (flags & ~VALID_OPEN_FLAGS)
1044 		return -EINVAL;
1045 	if (how->resolve & ~VALID_RESOLVE_FLAGS)
1046 		return -EINVAL;
1047 
1048 	/* Scoping flags are mutually exclusive. */
1049 	if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT))
1050 		return -EINVAL;
1051 
1052 	/* Deal with the mode. */
1053 	if (WILL_CREATE(flags)) {
1054 		if (how->mode & ~S_IALLUGO)
1055 			return -EINVAL;
1056 		op->mode = how->mode | S_IFREG;
1057 	} else {
1058 		if (how->mode != 0)
1059 			return -EINVAL;
1060 		op->mode = 0;
1061 	}
1062 
1063 	/*
1064 	 * In order to ensure programs get explicit errors when trying to use
1065 	 * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it
1066 	 * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we
1067 	 * have to require userspace to explicitly set it.
1068 	 */
1069 	if (flags & __O_TMPFILE) {
1070 		if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
1071 			return -EINVAL;
1072 		if (!(acc_mode & MAY_WRITE))
1073 			return -EINVAL;
1074 	}
1075 	if (flags & O_PATH) {
1076 		/* O_PATH only permits certain other flags to be set. */
1077 		if (flags & ~O_PATH_FLAGS)
1078 			return -EINVAL;
1079 		acc_mode = 0;
1080 	}
1081 
1082 	/*
1083 	 * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
1084 	 * check for O_DSYNC if the need any syncing at all we enforce it's
1085 	 * always set instead of having to deal with possibly weird behaviour
1086 	 * for malicious applications setting only __O_SYNC.
1087 	 */
1088 	if (flags & __O_SYNC)
1089 		flags |= O_DSYNC;
1090 
1091 	op->open_flag = flags;
1092 
1093 	/* O_TRUNC implies we need access checks for write permissions */
1094 	if (flags & O_TRUNC)
1095 		acc_mode |= MAY_WRITE;
1096 
1097 	/* Allow the LSM permission hook to distinguish append
1098 	   access from general write access. */
1099 	if (flags & O_APPEND)
1100 		acc_mode |= MAY_APPEND;
1101 
1102 	op->acc_mode = acc_mode;
1103 
1104 	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1105 
1106 	if (flags & O_CREAT) {
1107 		op->intent |= LOOKUP_CREATE;
1108 		if (flags & O_EXCL) {
1109 			op->intent |= LOOKUP_EXCL;
1110 			flags |= O_NOFOLLOW;
1111 		}
1112 	}
1113 
1114 	if (flags & O_DIRECTORY)
1115 		lookup_flags |= LOOKUP_DIRECTORY;
1116 	if (!(flags & O_NOFOLLOW))
1117 		lookup_flags |= LOOKUP_FOLLOW;
1118 
1119 	if (how->resolve & RESOLVE_NO_XDEV)
1120 		lookup_flags |= LOOKUP_NO_XDEV;
1121 	if (how->resolve & RESOLVE_NO_MAGICLINKS)
1122 		lookup_flags |= LOOKUP_NO_MAGICLINKS;
1123 	if (how->resolve & RESOLVE_NO_SYMLINKS)
1124 		lookup_flags |= LOOKUP_NO_SYMLINKS;
1125 	if (how->resolve & RESOLVE_BENEATH)
1126 		lookup_flags |= LOOKUP_BENEATH;
1127 	if (how->resolve & RESOLVE_IN_ROOT)
1128 		lookup_flags |= LOOKUP_IN_ROOT;
1129 	if (how->resolve & RESOLVE_CACHED) {
1130 		/* Don't bother even trying for create/truncate/tmpfile open */
1131 		if (flags & (O_TRUNC | O_CREAT | __O_TMPFILE))
1132 			return -EAGAIN;
1133 		lookup_flags |= LOOKUP_CACHED;
1134 	}
1135 
1136 	op->lookup_flags = lookup_flags;
1137 	return 0;
1138 }
1139 
1140 /**
1141  * file_open_name - open file and return file pointer
1142  *
1143  * @name:	struct filename containing path to open
1144  * @flags:	open flags as per the open(2) second argument
1145  * @mode:	mode for the new file if O_CREAT is set, else ignored
1146  *
1147  * This is the helper to open a file from kernelspace if you really
1148  * have to.  But in generally you should not do this, so please move
1149  * along, nothing to see here..
1150  */
file_open_name(struct filename * name,int flags,umode_t mode)1151 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1152 {
1153 	struct open_flags op;
1154 	struct open_how how = build_open_how(flags, mode);
1155 	int err = build_open_flags(&how, &op);
1156 	if (err)
1157 		return ERR_PTR(err);
1158 	return do_filp_open(AT_FDCWD, name, &op);
1159 }
1160 
1161 /**
1162  * filp_open - open file and return file pointer
1163  *
1164  * @filename:	path to open
1165  * @flags:	open flags as per the open(2) second argument
1166  * @mode:	mode for the new file if O_CREAT is set, else ignored
1167  *
1168  * This is the helper to open a file from kernelspace if you really
1169  * have to.  But in generally you should not do this, so please move
1170  * along, nothing to see here..
1171  */
filp_open(const char * filename,int flags,umode_t mode)1172 struct file *filp_open(const char *filename, int flags, umode_t mode)
1173 {
1174 	struct filename *name = getname_kernel(filename);
1175 	struct file *file = ERR_CAST(name);
1176 
1177 	if (!IS_ERR(name)) {
1178 		file = file_open_name(name, flags, mode);
1179 		putname(name);
1180 	}
1181 	return file;
1182 }
1183 EXPORT_SYMBOL_NS(filp_open, ANDROID_GKI_VFS_EXPORT_ONLY);
1184 
1185 
1186 /* ANDROID: Allow drivers to open only block files from kernel mode */
filp_open_block(const char * filename,int flags,umode_t mode)1187 struct file *filp_open_block(const char *filename, int flags, umode_t mode)
1188 {
1189 	struct file *file;
1190 
1191 	file = filp_open(filename, flags, mode);
1192 	if (IS_ERR(file))
1193 		goto err_out;
1194 
1195 	/* Drivers should only be allowed to open block devices */
1196 	if (!S_ISBLK(file->f_mapping->host->i_mode)) {
1197 		filp_close(file, NULL);
1198 		file = ERR_PTR(-ENOTBLK);
1199 	}
1200 
1201 err_out:
1202 	return file;
1203 }
1204 EXPORT_SYMBOL_GPL(filp_open_block);
1205 
file_open_root(const struct path * root,const char * filename,int flags,umode_t mode)1206 struct file *file_open_root(const struct path *root,
1207 			    const char *filename, int flags, umode_t mode)
1208 {
1209 	struct open_flags op;
1210 	struct open_how how = build_open_how(flags, mode);
1211 	int err = build_open_flags(&how, &op);
1212 	if (err)
1213 		return ERR_PTR(err);
1214 	return do_file_open_root(root, filename, &op);
1215 }
1216 EXPORT_SYMBOL(file_open_root);
1217 
do_sys_openat2(int dfd,const char __user * filename,struct open_how * how)1218 static long do_sys_openat2(int dfd, const char __user *filename,
1219 			   struct open_how *how)
1220 {
1221 	struct open_flags op;
1222 	int fd = build_open_flags(how, &op);
1223 	struct filename *tmp;
1224 
1225 	if (fd)
1226 		return fd;
1227 
1228 	tmp = getname(filename);
1229 	if (IS_ERR(tmp))
1230 		return PTR_ERR(tmp);
1231 
1232 	fd = get_unused_fd_flags(how->flags);
1233 	if (fd >= 0) {
1234 		struct file *f = do_filp_open(dfd, tmp, &op);
1235 		if (IS_ERR(f)) {
1236 			put_unused_fd(fd);
1237 			fd = PTR_ERR(f);
1238 		} else {
1239 			fsnotify_open(f);
1240 			fd_install(fd, f);
1241 		}
1242 	}
1243 	putname(tmp);
1244 	return fd;
1245 }
1246 
do_sys_open(int dfd,const char __user * filename,int flags,umode_t mode)1247 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1248 {
1249 	struct open_how how = build_open_how(flags, mode);
1250 	return do_sys_openat2(dfd, filename, &how);
1251 }
1252 
1253 
SYSCALL_DEFINE3(open,const char __user *,filename,int,flags,umode_t,mode)1254 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1255 {
1256 	if (force_o_largefile())
1257 		flags |= O_LARGEFILE;
1258 	return do_sys_open(AT_FDCWD, filename, flags, mode);
1259 }
1260 
SYSCALL_DEFINE4(openat,int,dfd,const char __user *,filename,int,flags,umode_t,mode)1261 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1262 		umode_t, mode)
1263 {
1264 	if (force_o_largefile())
1265 		flags |= O_LARGEFILE;
1266 	return do_sys_open(dfd, filename, flags, mode);
1267 }
1268 
SYSCALL_DEFINE4(openat2,int,dfd,const char __user *,filename,struct open_how __user *,how,size_t,usize)1269 SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename,
1270 		struct open_how __user *, how, size_t, usize)
1271 {
1272 	int err;
1273 	struct open_how tmp;
1274 
1275 	BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0);
1276 	BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST);
1277 
1278 	if (unlikely(usize < OPEN_HOW_SIZE_VER0))
1279 		return -EINVAL;
1280 
1281 	err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize);
1282 	if (err)
1283 		return err;
1284 
1285 	/* O_LARGEFILE is only allowed for non-O_PATH. */
1286 	if (!(tmp.flags & O_PATH) && force_o_largefile())
1287 		tmp.flags |= O_LARGEFILE;
1288 
1289 	return do_sys_openat2(dfd, filename, &tmp);
1290 }
1291 
1292 #ifdef CONFIG_COMPAT
1293 /*
1294  * Exactly like sys_open(), except that it doesn't set the
1295  * O_LARGEFILE flag.
1296  */
COMPAT_SYSCALL_DEFINE3(open,const char __user *,filename,int,flags,umode_t,mode)1297 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1298 {
1299 	return do_sys_open(AT_FDCWD, filename, flags, mode);
1300 }
1301 
1302 /*
1303  * Exactly like sys_openat(), except that it doesn't set the
1304  * O_LARGEFILE flag.
1305  */
COMPAT_SYSCALL_DEFINE4(openat,int,dfd,const char __user *,filename,int,flags,umode_t,mode)1306 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1307 {
1308 	return do_sys_open(dfd, filename, flags, mode);
1309 }
1310 #endif
1311 
1312 #ifndef __alpha__
1313 
1314 /*
1315  * For backward compatibility?  Maybe this should be moved
1316  * into arch/i386 instead?
1317  */
SYSCALL_DEFINE2(creat,const char __user *,pathname,umode_t,mode)1318 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1319 {
1320 	int flags = O_CREAT | O_WRONLY | O_TRUNC;
1321 
1322 	if (force_o_largefile())
1323 		flags |= O_LARGEFILE;
1324 	return do_sys_open(AT_FDCWD, pathname, flags, mode);
1325 }
1326 #endif
1327 
1328 /*
1329  * "id" is the POSIX thread ID. We use the
1330  * files pointer for this..
1331  */
filp_close(struct file * filp,fl_owner_t id)1332 int filp_close(struct file *filp, fl_owner_t id)
1333 {
1334 	int retval = 0;
1335 
1336 	if (!file_count(filp)) {
1337 		printk(KERN_ERR "VFS: Close: file count is 0\n");
1338 		return 0;
1339 	}
1340 
1341 	if (filp->f_op->flush)
1342 		retval = filp->f_op->flush(filp, id);
1343 
1344 	if (likely(!(filp->f_mode & FMODE_PATH))) {
1345 		dnotify_flush(filp, id);
1346 		locks_remove_posix(filp, id);
1347 	}
1348 	fput(filp);
1349 	return retval;
1350 }
1351 
1352 EXPORT_SYMBOL(filp_close);
1353 
1354 /*
1355  * Careful here! We test whether the file pointer is NULL before
1356  * releasing the fd. This ensures that one clone task can't release
1357  * an fd while another clone is opening it.
1358  */
SYSCALL_DEFINE1(close,unsigned int,fd)1359 SYSCALL_DEFINE1(close, unsigned int, fd)
1360 {
1361 	int retval = close_fd(fd);
1362 
1363 	/* can't restart close syscall because file table entry was cleared */
1364 	if (unlikely(retval == -ERESTARTSYS ||
1365 		     retval == -ERESTARTNOINTR ||
1366 		     retval == -ERESTARTNOHAND ||
1367 		     retval == -ERESTART_RESTARTBLOCK))
1368 		retval = -EINTR;
1369 
1370 	return retval;
1371 }
1372 
1373 /**
1374  * close_range() - Close all file descriptors in a given range.
1375  *
1376  * @fd:     starting file descriptor to close
1377  * @max_fd: last file descriptor to close
1378  * @flags:  reserved for future extensions
1379  *
1380  * This closes a range of file descriptors. All file descriptors
1381  * from @fd up to and including @max_fd are closed.
1382  * Currently, errors to close a given file descriptor are ignored.
1383  */
SYSCALL_DEFINE3(close_range,unsigned int,fd,unsigned int,max_fd,unsigned int,flags)1384 SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd,
1385 		unsigned int, flags)
1386 {
1387 	return __close_range(fd, max_fd, flags);
1388 }
1389 
1390 /*
1391  * This routine simulates a hangup on the tty, to arrange that users
1392  * are given clean terminals at login time.
1393  */
SYSCALL_DEFINE0(vhangup)1394 SYSCALL_DEFINE0(vhangup)
1395 {
1396 	if (capable(CAP_SYS_TTY_CONFIG)) {
1397 		tty_vhangup_self();
1398 		return 0;
1399 	}
1400 	return -EPERM;
1401 }
1402 
1403 /*
1404  * Called when an inode is about to be open.
1405  * We use this to disallow opening large files on 32bit systems if
1406  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1407  * on this flag in sys_open.
1408  */
generic_file_open(struct inode * inode,struct file * filp)1409 int generic_file_open(struct inode * inode, struct file * filp)
1410 {
1411 	if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1412 		return -EOVERFLOW;
1413 	return 0;
1414 }
1415 
1416 EXPORT_SYMBOL_NS(generic_file_open, ANDROID_GKI_VFS_EXPORT_ONLY);
1417 
1418 /*
1419  * This is used by subsystems that don't want seekable
1420  * file descriptors. The function is not supposed to ever fail, the only
1421  * reason it returns an 'int' and not 'void' is so that it can be plugged
1422  * directly into file_operations structure.
1423  */
nonseekable_open(struct inode * inode,struct file * filp)1424 int nonseekable_open(struct inode *inode, struct file *filp)
1425 {
1426 	filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1427 	return 0;
1428 }
1429 
1430 EXPORT_SYMBOL(nonseekable_open);
1431 
1432 /*
1433  * stream_open is used by subsystems that want stream-like file descriptors.
1434  * Such file descriptors are not seekable and don't have notion of position
1435  * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL).
1436  * Contrary to file descriptors of other regular files, .read() and .write()
1437  * can run simultaneously.
1438  *
1439  * stream_open never fails and is marked to return int so that it could be
1440  * directly used as file_operations.open .
1441  */
stream_open(struct inode * inode,struct file * filp)1442 int stream_open(struct inode *inode, struct file *filp)
1443 {
1444 	filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1445 	filp->f_mode |= FMODE_STREAM;
1446 	return 0;
1447 }
1448 
1449 EXPORT_SYMBOL(stream_open);
1450