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
do_truncate(struct dentry * dentry,loff_t length,unsigned int time_attrs,struct file * filp)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
vfs_truncate(const struct path * path,loff_t length)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
do_sys_truncate(const char __user * pathname,loff_t length)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
SYSCALL_DEFINE2(truncate,const char __user *,path,long,length)143 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
144 {
145 return do_sys_truncate(path, length);
146 }
147
148 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate,const char __user *,path,compat_off_t,length)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
do_sys_ftruncate(unsigned int fd,loff_t length,int small)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
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_NS_GPL(vfs_fallocate, ANDROID_GKI_VFS_EXPORT_ONLY);
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 */
do_faccessat(int dfd,const char __user * filename,int mode)348 long do_faccessat(int dfd, const char __user *filename, int mode)
349 {
350 const struct cred *old_cred;
351 struct cred *override_cred;
352 struct path path;
353 struct inode *inode;
354 int res;
355 unsigned int lookup_flags = LOOKUP_FOLLOW;
356
357 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
358 return -EINVAL;
359
360 override_cred = prepare_creds();
361 if (!override_cred)
362 return -ENOMEM;
363
364 override_cred->fsuid = override_cred->uid;
365 override_cred->fsgid = override_cred->gid;
366
367 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
368 /* Clear the capabilities if we switch to a non-root user */
369 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
370 if (!uid_eq(override_cred->uid, root_uid))
371 cap_clear(override_cred->cap_effective);
372 else
373 override_cred->cap_effective =
374 override_cred->cap_permitted;
375 }
376
377 /*
378 * The new set of credentials can *only* be used in
379 * task-synchronous circumstances, and does not need
380 * RCU freeing, unless somebody then takes a separate
381 * reference to it.
382 *
383 * NOTE! This is _only_ true because this credential
384 * is used purely for override_creds() that installs
385 * it as the subjective cred. Other threads will be
386 * accessing ->real_cred, not the subjective cred.
387 *
388 * If somebody _does_ make a copy of this (using the
389 * 'get_current_cred()' function), that will clear the
390 * non_rcu field, because now that other user may be
391 * expecting RCU freeing. But normal thread-synchronous
392 * cred accesses will keep things non-RCY.
393 */
394 override_cred->non_rcu = 1;
395
396 old_cred = override_creds(override_cred);
397 retry:
398 res = user_path_at(dfd, filename, lookup_flags, &path);
399 if (res)
400 goto out;
401
402 inode = d_backing_inode(path.dentry);
403
404 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
405 /*
406 * MAY_EXEC on regular files is denied if the fs is mounted
407 * with the "noexec" flag.
408 */
409 res = -EACCES;
410 if (path_noexec(&path))
411 goto out_path_release;
412 }
413
414 res = inode_permission(inode, mode | MAY_ACCESS);
415 /* SuS v2 requires we report a read only fs too */
416 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
417 goto out_path_release;
418 /*
419 * This is a rare case where using __mnt_is_readonly()
420 * is OK without a mnt_want/drop_write() pair. Since
421 * no actual write to the fs is performed here, we do
422 * not need to telegraph to that to anyone.
423 *
424 * By doing this, we accept that this access is
425 * inherently racy and know that the fs may change
426 * state before we even see this result.
427 */
428 if (__mnt_is_readonly(path.mnt))
429 res = -EROFS;
430
431 out_path_release:
432 path_put(&path);
433 if (retry_estale(res, lookup_flags)) {
434 lookup_flags |= LOOKUP_REVAL;
435 goto retry;
436 }
437 out:
438 revert_creds(old_cred);
439 put_cred(override_cred);
440 return res;
441 }
442
SYSCALL_DEFINE3(faccessat,int,dfd,const char __user *,filename,int,mode)443 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
444 {
445 return do_faccessat(dfd, filename, mode);
446 }
447
SYSCALL_DEFINE2(access,const char __user *,filename,int,mode)448 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
449 {
450 return do_faccessat(AT_FDCWD, filename, mode);
451 }
452
ksys_chdir(const char __user * filename)453 int ksys_chdir(const char __user *filename)
454 {
455 struct path path;
456 int error;
457 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
458 retry:
459 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
460 if (error)
461 goto out;
462
463 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
464 if (error)
465 goto dput_and_out;
466
467 set_fs_pwd(current->fs, &path);
468
469 dput_and_out:
470 path_put(&path);
471 if (retry_estale(error, lookup_flags)) {
472 lookup_flags |= LOOKUP_REVAL;
473 goto retry;
474 }
475 out:
476 return error;
477 }
478
SYSCALL_DEFINE1(chdir,const char __user *,filename)479 SYSCALL_DEFINE1(chdir, const char __user *, filename)
480 {
481 return ksys_chdir(filename);
482 }
483
SYSCALL_DEFINE1(fchdir,unsigned int,fd)484 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
485 {
486 struct fd f = fdget_raw(fd);
487 int error;
488
489 error = -EBADF;
490 if (!f.file)
491 goto out;
492
493 error = -ENOTDIR;
494 if (!d_can_lookup(f.file->f_path.dentry))
495 goto out_putf;
496
497 error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR);
498 if (!error)
499 set_fs_pwd(current->fs, &f.file->f_path);
500 out_putf:
501 fdput(f);
502 out:
503 return error;
504 }
505
ksys_chroot(const char __user * filename)506 int ksys_chroot(const char __user *filename)
507 {
508 struct path path;
509 int error;
510 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
511 retry:
512 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
513 if (error)
514 goto out;
515
516 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
517 if (error)
518 goto dput_and_out;
519
520 error = -EPERM;
521 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
522 goto dput_and_out;
523 error = security_path_chroot(&path);
524 if (error)
525 goto dput_and_out;
526
527 set_fs_root(current->fs, &path);
528 error = 0;
529 dput_and_out:
530 path_put(&path);
531 if (retry_estale(error, lookup_flags)) {
532 lookup_flags |= LOOKUP_REVAL;
533 goto retry;
534 }
535 out:
536 return error;
537 }
538
SYSCALL_DEFINE1(chroot,const char __user *,filename)539 SYSCALL_DEFINE1(chroot, const char __user *, filename)
540 {
541 return ksys_chroot(filename);
542 }
543
chmod_common(const struct path * path,umode_t mode)544 static int chmod_common(const struct path *path, umode_t mode)
545 {
546 struct inode *inode = path->dentry->d_inode;
547 struct inode *delegated_inode = NULL;
548 struct iattr newattrs;
549 int error;
550
551 error = mnt_want_write(path->mnt);
552 if (error)
553 return error;
554 retry_deleg:
555 inode_lock(inode);
556 error = security_path_chmod(path, mode);
557 if (error)
558 goto out_unlock;
559 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
560 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
561 error = notify_change(path->dentry, &newattrs, &delegated_inode);
562 out_unlock:
563 inode_unlock(inode);
564 if (delegated_inode) {
565 error = break_deleg_wait(&delegated_inode);
566 if (!error)
567 goto retry_deleg;
568 }
569 mnt_drop_write(path->mnt);
570 return error;
571 }
572
vfs_fchmod(struct file * file,umode_t mode)573 int vfs_fchmod(struct file *file, umode_t mode)
574 {
575 audit_file(file);
576 return chmod_common(&file->f_path, mode);
577 }
578
ksys_fchmod(unsigned int fd,umode_t mode)579 int ksys_fchmod(unsigned int fd, umode_t mode)
580 {
581 struct fd f = fdget(fd);
582 int err = -EBADF;
583
584 if (f.file) {
585 err = vfs_fchmod(f.file, mode);
586 fdput(f);
587 }
588 return err;
589 }
590
SYSCALL_DEFINE2(fchmod,unsigned int,fd,umode_t,mode)591 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
592 {
593 return ksys_fchmod(fd, mode);
594 }
595
do_fchmodat(int dfd,const char __user * filename,umode_t mode)596 int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
597 {
598 struct path path;
599 int error;
600 unsigned int lookup_flags = LOOKUP_FOLLOW;
601 retry:
602 error = user_path_at(dfd, filename, lookup_flags, &path);
603 if (!error) {
604 error = chmod_common(&path, mode);
605 path_put(&path);
606 if (retry_estale(error, lookup_flags)) {
607 lookup_flags |= LOOKUP_REVAL;
608 goto retry;
609 }
610 }
611 return error;
612 }
613
SYSCALL_DEFINE3(fchmodat,int,dfd,const char __user *,filename,umode_t,mode)614 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
615 umode_t, mode)
616 {
617 return do_fchmodat(dfd, filename, mode);
618 }
619
SYSCALL_DEFINE2(chmod,const char __user *,filename,umode_t,mode)620 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
621 {
622 return do_fchmodat(AT_FDCWD, filename, mode);
623 }
624
chown_common(const struct path * path,uid_t user,gid_t group)625 static int chown_common(const struct path *path, uid_t user, gid_t group)
626 {
627 struct inode *inode = path->dentry->d_inode;
628 struct inode *delegated_inode = NULL;
629 int error;
630 struct iattr newattrs;
631 kuid_t uid;
632 kgid_t gid;
633
634 uid = make_kuid(current_user_ns(), user);
635 gid = make_kgid(current_user_ns(), group);
636
637 retry_deleg:
638 newattrs.ia_valid = ATTR_CTIME;
639 if (user != (uid_t) -1) {
640 if (!uid_valid(uid))
641 return -EINVAL;
642 newattrs.ia_valid |= ATTR_UID;
643 newattrs.ia_uid = uid;
644 }
645 if (group != (gid_t) -1) {
646 if (!gid_valid(gid))
647 return -EINVAL;
648 newattrs.ia_valid |= ATTR_GID;
649 newattrs.ia_gid = gid;
650 }
651 if (!S_ISDIR(inode->i_mode))
652 newattrs.ia_valid |=
653 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
654 inode_lock(inode);
655 error = security_path_chown(path, uid, gid);
656 if (!error)
657 error = notify_change(path->dentry, &newattrs, &delegated_inode);
658 inode_unlock(inode);
659 if (delegated_inode) {
660 error = break_deleg_wait(&delegated_inode);
661 if (!error)
662 goto retry_deleg;
663 }
664 return error;
665 }
666
do_fchownat(int dfd,const char __user * filename,uid_t user,gid_t group,int flag)667 int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
668 int flag)
669 {
670 struct path path;
671 int error = -EINVAL;
672 int lookup_flags;
673
674 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
675 goto out;
676
677 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
678 if (flag & AT_EMPTY_PATH)
679 lookup_flags |= LOOKUP_EMPTY;
680 retry:
681 error = user_path_at(dfd, filename, lookup_flags, &path);
682 if (error)
683 goto out;
684 error = mnt_want_write(path.mnt);
685 if (error)
686 goto out_release;
687 error = chown_common(&path, user, group);
688 mnt_drop_write(path.mnt);
689 out_release:
690 path_put(&path);
691 if (retry_estale(error, lookup_flags)) {
692 lookup_flags |= LOOKUP_REVAL;
693 goto retry;
694 }
695 out:
696 return error;
697 }
698
SYSCALL_DEFINE5(fchownat,int,dfd,const char __user *,filename,uid_t,user,gid_t,group,int,flag)699 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
700 gid_t, group, int, flag)
701 {
702 return do_fchownat(dfd, filename, user, group, flag);
703 }
704
SYSCALL_DEFINE3(chown,const char __user *,filename,uid_t,user,gid_t,group)705 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
706 {
707 return do_fchownat(AT_FDCWD, filename, user, group, 0);
708 }
709
SYSCALL_DEFINE3(lchown,const char __user *,filename,uid_t,user,gid_t,group)710 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
711 {
712 return do_fchownat(AT_FDCWD, filename, user, group,
713 AT_SYMLINK_NOFOLLOW);
714 }
715
vfs_fchown(struct file * file,uid_t user,gid_t group)716 int vfs_fchown(struct file *file, uid_t user, gid_t group)
717 {
718 int error;
719
720 error = mnt_want_write_file(file);
721 if (error)
722 return error;
723 audit_file(file);
724 error = chown_common(&file->f_path, user, group);
725 mnt_drop_write_file(file);
726 return error;
727 }
728
ksys_fchown(unsigned int fd,uid_t user,gid_t group)729 int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
730 {
731 struct fd f = fdget(fd);
732 int error = -EBADF;
733
734 if (f.file) {
735 error = vfs_fchown(f.file, user, group);
736 fdput(f);
737 }
738 return error;
739 }
740
SYSCALL_DEFINE3(fchown,unsigned int,fd,uid_t,user,gid_t,group)741 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
742 {
743 return ksys_fchown(fd, user, group);
744 }
745
do_dentry_open(struct file * f,struct inode * inode,int (* open)(struct inode *,struct file *))746 static int do_dentry_open(struct file *f,
747 struct inode *inode,
748 int (*open)(struct inode *, struct file *))
749 {
750 static const struct file_operations empty_fops = {};
751 int error;
752
753 path_get(&f->f_path);
754 f->f_inode = inode;
755 f->f_mapping = inode->i_mapping;
756
757 /* Ensure that we skip any errors that predate opening of the file */
758 f->f_wb_err = filemap_sample_wb_err(f->f_mapping);
759
760 if (unlikely(f->f_flags & O_PATH)) {
761 f->f_mode = FMODE_PATH | FMODE_OPENED;
762 f->f_op = &empty_fops;
763 return 0;
764 }
765
766 /* Any file opened for execve()/uselib() has to be a regular file. */
767 if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) {
768 error = -EACCES;
769 goto cleanup_file;
770 }
771
772 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
773 error = get_write_access(inode);
774 if (unlikely(error))
775 goto cleanup_file;
776 error = __mnt_want_write(f->f_path.mnt);
777 if (unlikely(error)) {
778 put_write_access(inode);
779 goto cleanup_file;
780 }
781 f->f_mode |= FMODE_WRITER;
782 }
783
784 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
785 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
786 f->f_mode |= FMODE_ATOMIC_POS;
787
788 f->f_op = fops_get(inode->i_fop);
789 if (WARN_ON(!f->f_op)) {
790 error = -ENODEV;
791 goto cleanup_all;
792 }
793
794 error = security_file_open(f);
795 if (error)
796 goto cleanup_all;
797
798 error = break_lease(locks_inode(f), f->f_flags);
799 if (error)
800 goto cleanup_all;
801
802 /* normally all 3 are set; ->open() can clear them if needed */
803 f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
804 if (!open)
805 open = f->f_op->open;
806 if (open) {
807 error = open(inode, f);
808 if (error)
809 goto cleanup_all;
810 }
811 f->f_mode |= FMODE_OPENED;
812 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
813 i_readcount_inc(inode);
814 if ((f->f_mode & FMODE_READ) &&
815 likely(f->f_op->read || f->f_op->read_iter))
816 f->f_mode |= FMODE_CAN_READ;
817 if ((f->f_mode & FMODE_WRITE) &&
818 likely(f->f_op->write || f->f_op->write_iter))
819 f->f_mode |= FMODE_CAN_WRITE;
820
821 f->f_write_hint = WRITE_LIFE_NOT_SET;
822 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
823
824 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
825
826 /* NB: we're sure to have correct a_ops only after f_op->open */
827 if (f->f_flags & O_DIRECT) {
828 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
829 return -EINVAL;
830 }
831
832 /*
833 * XXX: Huge page cache doesn't support writing yet. Drop all page
834 * cache for this file before processing writes.
835 */
836 if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping))
837 truncate_pagecache(inode, 0);
838
839 return 0;
840
841 cleanup_all:
842 if (WARN_ON_ONCE(error > 0))
843 error = -EINVAL;
844 fops_put(f->f_op);
845 if (f->f_mode & FMODE_WRITER) {
846 put_write_access(inode);
847 __mnt_drop_write(f->f_path.mnt);
848 }
849 cleanup_file:
850 path_put(&f->f_path);
851 f->f_path.mnt = NULL;
852 f->f_path.dentry = NULL;
853 f->f_inode = NULL;
854 return error;
855 }
856
857 /**
858 * finish_open - finish opening a file
859 * @file: file pointer
860 * @dentry: pointer to dentry
861 * @open: open callback
862 * @opened: state of open
863 *
864 * This can be used to finish opening a file passed to i_op->atomic_open().
865 *
866 * If the open callback is set to NULL, then the standard f_op->open()
867 * filesystem callback is substituted.
868 *
869 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
870 * the return value of d_splice_alias(), then the caller needs to perform dput()
871 * on it after finish_open().
872 *
873 * Returns zero on success or -errno if the open failed.
874 */
finish_open(struct file * file,struct dentry * dentry,int (* open)(struct inode *,struct file *))875 int finish_open(struct file *file, struct dentry *dentry,
876 int (*open)(struct inode *, struct file *))
877 {
878 BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */
879
880 file->f_path.dentry = dentry;
881 return do_dentry_open(file, d_backing_inode(dentry), open);
882 }
883 EXPORT_SYMBOL(finish_open);
884
885 /**
886 * finish_no_open - finish ->atomic_open() without opening the file
887 *
888 * @file: file pointer
889 * @dentry: dentry or NULL (as returned from ->lookup())
890 *
891 * This can be used to set the result of a successful lookup in ->atomic_open().
892 *
893 * NB: unlike finish_open() this function does consume the dentry reference and
894 * the caller need not dput() it.
895 *
896 * Returns "0" which must be the return value of ->atomic_open() after having
897 * called this function.
898 */
finish_no_open(struct file * file,struct dentry * dentry)899 int finish_no_open(struct file *file, struct dentry *dentry)
900 {
901 file->f_path.dentry = dentry;
902 return 0;
903 }
904 EXPORT_SYMBOL(finish_no_open);
905
file_path(struct file * filp,char * buf,int buflen)906 char *file_path(struct file *filp, char *buf, int buflen)
907 {
908 return d_path(&filp->f_path, buf, buflen);
909 }
910 EXPORT_SYMBOL(file_path);
911
912 /**
913 * vfs_open - open the file at the given path
914 * @path: path to open
915 * @file: newly allocated file with f_flag initialized
916 * @cred: credentials to use
917 */
vfs_open(const struct path * path,struct file * file)918 int vfs_open(const struct path *path, struct file *file)
919 {
920 file->f_path = *path;
921 return do_dentry_open(file, d_backing_inode(path->dentry), NULL);
922 }
923
dentry_open(const struct path * path,int flags,const struct cred * cred)924 struct file *dentry_open(const struct path *path, int flags,
925 const struct cred *cred)
926 {
927 int error;
928 struct file *f;
929
930 validate_creds(cred);
931
932 /* We must always pass in a valid mount pointer. */
933 BUG_ON(!path->mnt);
934
935 f = alloc_empty_file(flags, cred);
936 if (!IS_ERR(f)) {
937 error = vfs_open(path, f);
938 if (error) {
939 fput(f);
940 f = ERR_PTR(error);
941 }
942 }
943 return f;
944 }
945 EXPORT_SYMBOL_NS(dentry_open, ANDROID_GKI_VFS_EXPORT_ONLY);
946
open_with_fake_path(const struct path * path,int flags,struct inode * inode,const struct cred * cred)947 struct file *open_with_fake_path(const struct path *path, int flags,
948 struct inode *inode, const struct cred *cred)
949 {
950 struct file *f = alloc_empty_file_noaccount(flags, cred);
951 if (!IS_ERR(f)) {
952 int error;
953
954 f->f_path = *path;
955 error = do_dentry_open(f, inode, NULL);
956 if (error) {
957 fput(f);
958 f = ERR_PTR(error);
959 }
960 }
961 return f;
962 }
963 EXPORT_SYMBOL(open_with_fake_path);
964
build_open_flags(int flags,umode_t mode,struct open_flags * op)965 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
966 {
967 int lookup_flags = 0;
968 int acc_mode = ACC_MODE(flags);
969
970 /*
971 * Clear out all open flags we don't know about so that we don't report
972 * them in fcntl(F_GETFD) or similar interfaces.
973 */
974 flags &= VALID_OPEN_FLAGS;
975
976 if (flags & (O_CREAT | __O_TMPFILE))
977 op->mode = (mode & S_IALLUGO) | S_IFREG;
978 else
979 op->mode = 0;
980
981 /* Must never be set by userspace */
982 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
983
984 /*
985 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
986 * check for O_DSYNC if the need any syncing at all we enforce it's
987 * always set instead of having to deal with possibly weird behaviour
988 * for malicious applications setting only __O_SYNC.
989 */
990 if (flags & __O_SYNC)
991 flags |= O_DSYNC;
992
993 if (flags & __O_TMPFILE) {
994 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
995 return -EINVAL;
996 if (!(acc_mode & MAY_WRITE))
997 return -EINVAL;
998 } else if (flags & O_PATH) {
999 /*
1000 * If we have O_PATH in the open flag. Then we
1001 * cannot have anything other than the below set of flags
1002 */
1003 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
1004 acc_mode = 0;
1005 }
1006
1007 op->open_flag = flags;
1008
1009 /* O_TRUNC implies we need access checks for write permissions */
1010 if (flags & O_TRUNC)
1011 acc_mode |= MAY_WRITE;
1012
1013 /* Allow the LSM permission hook to distinguish append
1014 access from general write access. */
1015 if (flags & O_APPEND)
1016 acc_mode |= MAY_APPEND;
1017
1018 op->acc_mode = acc_mode;
1019
1020 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
1021
1022 if (flags & O_CREAT) {
1023 op->intent |= LOOKUP_CREATE;
1024 if (flags & O_EXCL)
1025 op->intent |= LOOKUP_EXCL;
1026 }
1027
1028 if (flags & O_DIRECTORY)
1029 lookup_flags |= LOOKUP_DIRECTORY;
1030 if (!(flags & O_NOFOLLOW))
1031 lookup_flags |= LOOKUP_FOLLOW;
1032 op->lookup_flags = lookup_flags;
1033 return 0;
1034 }
1035
1036 /**
1037 * file_open_name - open file and return file pointer
1038 *
1039 * @name: struct filename containing path to open
1040 * @flags: open flags as per the open(2) second argument
1041 * @mode: mode for the new file if O_CREAT is set, else ignored
1042 *
1043 * This is the helper to open a file from kernelspace if you really
1044 * have to. But in generally you should not do this, so please move
1045 * along, nothing to see here..
1046 */
file_open_name(struct filename * name,int flags,umode_t mode)1047 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
1048 {
1049 struct open_flags op;
1050 int err = build_open_flags(flags, mode, &op);
1051 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1052 }
1053
1054 /**
1055 * filp_open - open file and return file pointer
1056 *
1057 * @filename: path to open
1058 * @flags: open flags as per the open(2) second argument
1059 * @mode: mode for the new file if O_CREAT is set, else ignored
1060 *
1061 * This is the helper to open a file from kernelspace if you really
1062 * have to. But in generally you should not do this, so please move
1063 * along, nothing to see here..
1064 */
filp_open(const char * filename,int flags,umode_t mode)1065 struct file *filp_open(const char *filename, int flags, umode_t mode)
1066 {
1067 struct filename *name = getname_kernel(filename);
1068 struct file *file = ERR_CAST(name);
1069
1070 if (!IS_ERR(name)) {
1071 file = file_open_name(name, flags, mode);
1072 putname(name);
1073 }
1074 return file;
1075 }
1076 EXPORT_SYMBOL(filp_open);
1077
file_open_root(struct dentry * dentry,struct vfsmount * mnt,const char * filename,int flags,umode_t mode)1078 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1079 const char *filename, int flags, umode_t mode)
1080 {
1081 struct open_flags op;
1082 int err = build_open_flags(flags, mode, &op);
1083 if (err)
1084 return ERR_PTR(err);
1085 return do_file_open_root(dentry, mnt, filename, &op);
1086 }
1087 EXPORT_SYMBOL(file_open_root);
1088
do_sys_open(int dfd,const char __user * filename,int flags,umode_t mode)1089 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1090 {
1091 struct open_flags op;
1092 int fd = build_open_flags(flags, mode, &op);
1093 struct filename *tmp;
1094
1095 if (fd)
1096 return fd;
1097
1098 tmp = getname(filename);
1099 if (IS_ERR(tmp))
1100 return PTR_ERR(tmp);
1101
1102 fd = get_unused_fd_flags(flags);
1103 if (fd >= 0) {
1104 struct file *f = do_filp_open(dfd, tmp, &op);
1105 if (IS_ERR(f)) {
1106 put_unused_fd(fd);
1107 fd = PTR_ERR(f);
1108 } else {
1109 fsnotify_open(f);
1110 fd_install(fd, f);
1111 }
1112 }
1113 putname(tmp);
1114 return fd;
1115 }
1116
SYSCALL_DEFINE3(open,const char __user *,filename,int,flags,umode_t,mode)1117 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1118 {
1119 if (force_o_largefile())
1120 flags |= O_LARGEFILE;
1121
1122 return do_sys_open(AT_FDCWD, filename, flags, mode);
1123 }
1124
SYSCALL_DEFINE4(openat,int,dfd,const char __user *,filename,int,flags,umode_t,mode)1125 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1126 umode_t, mode)
1127 {
1128 if (force_o_largefile())
1129 flags |= O_LARGEFILE;
1130
1131 return do_sys_open(dfd, filename, flags, mode);
1132 }
1133
1134 #ifdef CONFIG_COMPAT
1135 /*
1136 * Exactly like sys_open(), except that it doesn't set the
1137 * O_LARGEFILE flag.
1138 */
COMPAT_SYSCALL_DEFINE3(open,const char __user *,filename,int,flags,umode_t,mode)1139 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1140 {
1141 return do_sys_open(AT_FDCWD, filename, flags, mode);
1142 }
1143
1144 /*
1145 * Exactly like sys_openat(), except that it doesn't set the
1146 * O_LARGEFILE flag.
1147 */
COMPAT_SYSCALL_DEFINE4(openat,int,dfd,const char __user *,filename,int,flags,umode_t,mode)1148 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1149 {
1150 return do_sys_open(dfd, filename, flags, mode);
1151 }
1152 #endif
1153
1154 #ifndef __alpha__
1155
1156 /*
1157 * For backward compatibility? Maybe this should be moved
1158 * into arch/i386 instead?
1159 */
SYSCALL_DEFINE2(creat,const char __user *,pathname,umode_t,mode)1160 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1161 {
1162 return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1163 }
1164
1165 #endif
1166
1167 /*
1168 * "id" is the POSIX thread ID. We use the
1169 * files pointer for this..
1170 */
filp_close(struct file * filp,fl_owner_t id)1171 int filp_close(struct file *filp, fl_owner_t id)
1172 {
1173 int retval = 0;
1174
1175 if (!file_count(filp)) {
1176 printk(KERN_ERR "VFS: Close: file count is 0\n");
1177 return 0;
1178 }
1179
1180 if (filp->f_op->flush)
1181 retval = filp->f_op->flush(filp, id);
1182
1183 if (likely(!(filp->f_mode & FMODE_PATH))) {
1184 dnotify_flush(filp, id);
1185 locks_remove_posix(filp, id);
1186 }
1187 fput(filp);
1188 return retval;
1189 }
1190
1191 EXPORT_SYMBOL(filp_close);
1192
1193 /*
1194 * Careful here! We test whether the file pointer is NULL before
1195 * releasing the fd. This ensures that one clone task can't release
1196 * an fd while another clone is opening it.
1197 */
SYSCALL_DEFINE1(close,unsigned int,fd)1198 SYSCALL_DEFINE1(close, unsigned int, fd)
1199 {
1200 int retval = __close_fd(current->files, fd);
1201
1202 /* can't restart close syscall because file table entry was cleared */
1203 if (unlikely(retval == -ERESTARTSYS ||
1204 retval == -ERESTARTNOINTR ||
1205 retval == -ERESTARTNOHAND ||
1206 retval == -ERESTART_RESTARTBLOCK))
1207 retval = -EINTR;
1208
1209 return retval;
1210 }
1211
1212 /*
1213 * This routine simulates a hangup on the tty, to arrange that users
1214 * are given clean terminals at login time.
1215 */
SYSCALL_DEFINE0(vhangup)1216 SYSCALL_DEFINE0(vhangup)
1217 {
1218 if (capable(CAP_SYS_TTY_CONFIG)) {
1219 tty_vhangup_self();
1220 return 0;
1221 }
1222 return -EPERM;
1223 }
1224
1225 /*
1226 * Called when an inode is about to be open.
1227 * We use this to disallow opening large files on 32bit systems if
1228 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1229 * on this flag in sys_open.
1230 */
generic_file_open(struct inode * inode,struct file * filp)1231 int generic_file_open(struct inode * inode, struct file * filp)
1232 {
1233 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1234 return -EOVERFLOW;
1235 return 0;
1236 }
1237
1238 EXPORT_SYMBOL(generic_file_open);
1239
1240 /*
1241 * This is used by subsystems that don't want seekable
1242 * file descriptors. The function is not supposed to ever fail, the only
1243 * reason it returns an 'int' and not 'void' is so that it can be plugged
1244 * directly into file_operations structure.
1245 */
nonseekable_open(struct inode * inode,struct file * filp)1246 int nonseekable_open(struct inode *inode, struct file *filp)
1247 {
1248 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1249 return 0;
1250 }
1251
1252 EXPORT_SYMBOL(nonseekable_open);
1253
1254 /*
1255 * stream_open is used by subsystems that want stream-like file descriptors.
1256 * Such file descriptors are not seekable and don't have notion of position
1257 * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL).
1258 * Contrary to file descriptors of other regular files, .read() and .write()
1259 * can run simultaneously.
1260 *
1261 * stream_open never fails and is marked to return int so that it could be
1262 * directly used as file_operations.open .
1263 */
stream_open(struct inode * inode,struct file * filp)1264 int stream_open(struct inode *inode, struct file *filp)
1265 {
1266 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS);
1267 filp->f_mode |= FMODE_STREAM;
1268 return 0;
1269 }
1270
1271 EXPORT_SYMBOL(stream_open);
1272