1 /*
2 * linux/fs/open.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7 #include <linux/string.h>
8 #include <linux/mm.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/fsnotify.h>
12 #include <linux/module.h>
13 #include <linux/tty.h>
14 #include <linux/namei.h>
15 #include <linux/backing-dev.h>
16 #include <linux/capability.h>
17 #include <linux/securebits.h>
18 #include <linux/security.h>
19 #include <linux/mount.h>
20 #include <linux/fcntl.h>
21 #include <linux/slab.h>
22 #include <asm/uaccess.h>
23 #include <linux/fs.h>
24 #include <linux/personality.h>
25 #include <linux/pagemap.h>
26 #include <linux/syscalls.h>
27 #include <linux/rcupdate.h>
28 #include <linux/audit.h>
29 #include <linux/falloc.h>
30 #include <linux/fs_struct.h>
31 #include <linux/ima.h>
32 #include <linux/dnotify.h>
33 #include <linux/compat.h>
34
35 #include "internal.h"
36
do_truncate2(struct vfsmount * mnt,struct dentry * dentry,loff_t length,unsigned int time_attrs,struct file * filp)37 int do_truncate2(struct vfsmount *mnt, struct dentry *dentry, loff_t length,
38 unsigned int time_attrs, struct file *filp)
39 {
40 int ret;
41 struct iattr newattrs;
42
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
44 if (length < 0)
45 return -EINVAL;
46
47 newattrs.ia_size = length;
48 newattrs.ia_valid = ATTR_SIZE | time_attrs;
49 if (filp) {
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
52 }
53
54 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
56 if (ret < 0)
57 return ret;
58 if (ret)
59 newattrs.ia_valid |= ret | ATTR_FORCE;
60
61 inode_lock(dentry->d_inode);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change2(mnt, dentry, &newattrs, NULL);
64 inode_unlock(dentry->d_inode);
65 return ret;
66 }
do_truncate(struct dentry * dentry,loff_t length,unsigned int time_attrs,struct file * filp)67 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
68 struct file *filp)
69 {
70 return do_truncate2(NULL, dentry, length, time_attrs, filp);
71 }
72
vfs_truncate(const struct path * path,loff_t length)73 long vfs_truncate(const struct path *path, loff_t length)
74 {
75 struct inode *inode;
76 struct vfsmount *mnt;
77 struct dentry *upperdentry;
78 long error;
79
80 inode = path->dentry->d_inode;
81 mnt = path->mnt;
82
83 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
84 if (S_ISDIR(inode->i_mode))
85 return -EISDIR;
86 if (!S_ISREG(inode->i_mode))
87 return -EINVAL;
88
89 error = mnt_want_write(path->mnt);
90 if (error)
91 goto out;
92
93 error = inode_permission2(mnt, inode, MAY_WRITE);
94 if (error)
95 goto mnt_drop_write_and_out;
96
97 error = -EPERM;
98 if (IS_APPEND(inode))
99 goto mnt_drop_write_and_out;
100
101 /*
102 * If this is an overlayfs then do as if opening the file so we get
103 * write access on the upper inode, not on the overlay inode. For
104 * non-overlay filesystems d_real() is an identity function.
105 */
106 upperdentry = d_real(path->dentry, NULL, O_WRONLY);
107 error = PTR_ERR(upperdentry);
108 if (IS_ERR(upperdentry))
109 goto mnt_drop_write_and_out;
110
111 error = get_write_access(upperdentry->d_inode);
112 if (error)
113 goto mnt_drop_write_and_out;
114
115 /*
116 * Make sure that there are no leases. get_write_access() protects
117 * against the truncate racing with a lease-granting setlease().
118 */
119 error = break_lease(inode, O_WRONLY);
120 if (error)
121 goto put_write_and_out;
122
123 error = locks_verify_truncate(inode, NULL, length);
124 if (!error)
125 error = security_path_truncate(path);
126 if (!error)
127 error = do_truncate2(mnt, path->dentry, length, 0, NULL);
128
129 put_write_and_out:
130 put_write_access(upperdentry->d_inode);
131 mnt_drop_write_and_out:
132 mnt_drop_write(path->mnt);
133 out:
134 return error;
135 }
136 EXPORT_SYMBOL_GPL(vfs_truncate);
137
do_sys_truncate(const char __user * pathname,loff_t length)138 static long do_sys_truncate(const char __user *pathname, loff_t length)
139 {
140 unsigned int lookup_flags = LOOKUP_FOLLOW;
141 struct path path;
142 int error;
143
144 if (length < 0) /* sorry, but loff_t says... */
145 return -EINVAL;
146
147 retry:
148 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
149 if (!error) {
150 error = vfs_truncate(&path, length);
151 path_put(&path);
152 }
153 if (retry_estale(error, lookup_flags)) {
154 lookup_flags |= LOOKUP_REVAL;
155 goto retry;
156 }
157 return error;
158 }
159
SYSCALL_DEFINE2(truncate,const char __user *,path,long,length)160 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
161 {
162 return do_sys_truncate(path, length);
163 }
164
165 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate,const char __user *,path,compat_off_t,length)166 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
167 {
168 return do_sys_truncate(path, length);
169 }
170 #endif
171
do_sys_ftruncate(unsigned int fd,loff_t length,int small)172 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
173 {
174 struct inode *inode;
175 struct dentry *dentry;
176 struct vfsmount *mnt;
177 struct fd f;
178 int error;
179
180 error = -EINVAL;
181 if (length < 0)
182 goto out;
183 error = -EBADF;
184 f = fdget(fd);
185 if (!f.file)
186 goto out;
187
188 /* explicitly opened as large or we are on 64-bit box */
189 if (f.file->f_flags & O_LARGEFILE)
190 small = 0;
191
192 dentry = f.file->f_path.dentry;
193 mnt = f.file->f_path.mnt;
194 inode = dentry->d_inode;
195 error = -EINVAL;
196 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
197 goto out_putf;
198
199 error = -EINVAL;
200 /* Cannot ftruncate over 2^31 bytes without large file support */
201 if (small && length > MAX_NON_LFS)
202 goto out_putf;
203
204 error = -EPERM;
205 if (IS_APPEND(inode))
206 goto out_putf;
207
208 sb_start_write(inode->i_sb);
209 error = locks_verify_truncate(inode, f.file, length);
210 if (!error)
211 error = security_path_truncate(&f.file->f_path);
212 if (!error)
213 error = do_truncate2(mnt, dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
214 sb_end_write(inode->i_sb);
215 out_putf:
216 fdput(f);
217 out:
218 return error;
219 }
220
SYSCALL_DEFINE2(ftruncate,unsigned int,fd,unsigned long,length)221 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
222 {
223 return do_sys_ftruncate(fd, length, 1);
224 }
225
226 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(ftruncate,unsigned int,fd,compat_ulong_t,length)227 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
228 {
229 return do_sys_ftruncate(fd, length, 1);
230 }
231 #endif
232
233 /* LFS versions of truncate are only needed on 32 bit machines */
234 #if BITS_PER_LONG == 32
SYSCALL_DEFINE2(truncate64,const char __user *,path,loff_t,length)235 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
236 {
237 return do_sys_truncate(path, length);
238 }
239
SYSCALL_DEFINE2(ftruncate64,unsigned int,fd,loff_t,length)240 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
241 {
242 return do_sys_ftruncate(fd, length, 0);
243 }
244 #endif /* BITS_PER_LONG == 32 */
245
246
vfs_fallocate(struct file * file,int mode,loff_t offset,loff_t len)247 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
248 {
249 struct inode *inode = file_inode(file);
250 long ret;
251
252 if (offset < 0 || len <= 0)
253 return -EINVAL;
254
255 /* Return error if mode is not supported */
256 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
257 return -EOPNOTSUPP;
258
259 /* Punch hole and zero range are mutually exclusive */
260 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
261 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
262 return -EOPNOTSUPP;
263
264 /* Punch hole must have keep size set */
265 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
266 !(mode & FALLOC_FL_KEEP_SIZE))
267 return -EOPNOTSUPP;
268
269 /* Collapse range should only be used exclusively. */
270 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
271 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
272 return -EINVAL;
273
274 /* Insert range should only be used exclusively. */
275 if ((mode & FALLOC_FL_INSERT_RANGE) &&
276 (mode & ~FALLOC_FL_INSERT_RANGE))
277 return -EINVAL;
278
279 /* Unshare range should only be used with allocate mode. */
280 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
281 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
282 return -EINVAL;
283
284 if (!(file->f_mode & FMODE_WRITE))
285 return -EBADF;
286
287 /*
288 * We can only allow pure fallocate on append only files
289 */
290 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
291 return -EPERM;
292
293 if (IS_IMMUTABLE(inode))
294 return -EPERM;
295
296 /*
297 * We cannot allow any fallocate operation on an active swapfile
298 */
299 if (IS_SWAPFILE(inode))
300 return -ETXTBSY;
301
302 /*
303 * Revalidate the write permissions, in case security policy has
304 * changed since the files were opened.
305 */
306 ret = security_file_permission(file, MAY_WRITE);
307 if (ret)
308 return ret;
309
310 if (S_ISFIFO(inode->i_mode))
311 return -ESPIPE;
312
313 /*
314 * Let individual file system decide if it supports preallocation
315 * for directories or not.
316 */
317 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) &&
318 !S_ISBLK(inode->i_mode))
319 return -ENODEV;
320
321 /* Check for wrap through zero too */
322 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
323 return -EFBIG;
324
325 if (!file->f_op->fallocate)
326 return -EOPNOTSUPP;
327
328 sb_start_write(inode->i_sb);
329 ret = file->f_op->fallocate(file, mode, offset, len);
330
331 /*
332 * Create inotify and fanotify events.
333 *
334 * To keep the logic simple always create events if fallocate succeeds.
335 * This implies that events are even created if the file size remains
336 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
337 */
338 if (ret == 0)
339 fsnotify_modify(file);
340
341 sb_end_write(inode->i_sb);
342 return ret;
343 }
344 EXPORT_SYMBOL_GPL(vfs_fallocate);
345
SYSCALL_DEFINE4(fallocate,int,fd,int,mode,loff_t,offset,loff_t,len)346 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
347 {
348 struct fd f = fdget(fd);
349 int error = -EBADF;
350
351 if (f.file) {
352 error = vfs_fallocate(f.file, mode, offset, len);
353 fdput(f);
354 }
355 return error;
356 }
357
358 /*
359 * access() needs to use the real uid/gid, not the effective uid/gid.
360 * We do this by temporarily clearing all FS-related capabilities and
361 * switching the fsuid/fsgid around to the real ones.
362 */
SYSCALL_DEFINE3(faccessat,int,dfd,const char __user *,filename,int,mode)363 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
364 {
365 const struct cred *old_cred;
366 struct cred *override_cred;
367 struct path path;
368 struct inode *inode;
369 struct vfsmount *mnt;
370 int res;
371 unsigned int lookup_flags = LOOKUP_FOLLOW;
372
373 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
374 return -EINVAL;
375
376 override_cred = prepare_creds();
377 if (!override_cred)
378 return -ENOMEM;
379
380 override_cred->fsuid = override_cred->uid;
381 override_cred->fsgid = override_cred->gid;
382
383 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
384 /* Clear the capabilities if we switch to a non-root user */
385 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
386 if (!uid_eq(override_cred->uid, root_uid))
387 cap_clear(override_cred->cap_effective);
388 else
389 override_cred->cap_effective =
390 override_cred->cap_permitted;
391 }
392
393 old_cred = override_creds(override_cred);
394 retry:
395 res = user_path_at(dfd, filename, lookup_flags, &path);
396 if (res)
397 goto out;
398
399 inode = d_backing_inode(path.dentry);
400 mnt = path.mnt;
401
402 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
403 /*
404 * MAY_EXEC on regular files is denied if the fs is mounted
405 * with the "noexec" flag.
406 */
407 res = -EACCES;
408 if (path_noexec(&path))
409 goto out_path_release;
410 }
411
412 res = inode_permission2(mnt, inode, mode | MAY_ACCESS);
413 /* SuS v2 requires we report a read only fs too */
414 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
415 goto out_path_release;
416 /*
417 * This is a rare case where using __mnt_is_readonly()
418 * is OK without a mnt_want/drop_write() pair. Since
419 * no actual write to the fs is performed here, we do
420 * not need to telegraph to that to anyone.
421 *
422 * By doing this, we accept that this access is
423 * inherently racy and know that the fs may change
424 * state before we even see this result.
425 */
426 if (__mnt_is_readonly(path.mnt))
427 res = -EROFS;
428
429 out_path_release:
430 path_put(&path);
431 if (retry_estale(res, lookup_flags)) {
432 lookup_flags |= LOOKUP_REVAL;
433 goto retry;
434 }
435 out:
436 revert_creds(old_cred);
437 put_cred(override_cred);
438 return res;
439 }
440
SYSCALL_DEFINE2(access,const char __user *,filename,int,mode)441 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
442 {
443 return sys_faccessat(AT_FDCWD, filename, mode);
444 }
445
SYSCALL_DEFINE1(chdir,const char __user *,filename)446 SYSCALL_DEFINE1(chdir, const char __user *, filename)
447 {
448 struct path path;
449 int error;
450 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
451 retry:
452 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
453 if (error)
454 goto out;
455
456 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
457 if (error)
458 goto dput_and_out;
459
460 set_fs_pwd(current->fs, &path);
461
462 dput_and_out:
463 path_put(&path);
464 if (retry_estale(error, lookup_flags)) {
465 lookup_flags |= LOOKUP_REVAL;
466 goto retry;
467 }
468 out:
469 return error;
470 }
471
SYSCALL_DEFINE1(fchdir,unsigned int,fd)472 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
473 {
474 struct fd f = fdget_raw(fd);
475 struct inode *inode;
476 struct vfsmount *mnt;
477 int error = -EBADF;
478
479 error = -EBADF;
480 if (!f.file)
481 goto out;
482
483 inode = file_inode(f.file);
484 mnt = f.file->f_path.mnt;
485
486 error = -ENOTDIR;
487 if (!S_ISDIR(inode->i_mode))
488 goto out_putf;
489
490 error = inode_permission2(mnt, inode, MAY_EXEC | MAY_CHDIR);
491 if (!error)
492 set_fs_pwd(current->fs, &f.file->f_path);
493 out_putf:
494 fdput(f);
495 out:
496 return error;
497 }
498
SYSCALL_DEFINE1(chroot,const char __user *,filename)499 SYSCALL_DEFINE1(chroot, const char __user *, filename)
500 {
501 struct path path;
502 int error;
503 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
504 retry:
505 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
506 if (error)
507 goto out;
508
509 error = inode_permission2(path.mnt, path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
510 if (error)
511 goto dput_and_out;
512
513 error = -EPERM;
514 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
515 goto dput_and_out;
516 error = security_path_chroot(&path);
517 if (error)
518 goto dput_and_out;
519
520 set_fs_root(current->fs, &path);
521 error = 0;
522 dput_and_out:
523 path_put(&path);
524 if (retry_estale(error, lookup_flags)) {
525 lookup_flags |= LOOKUP_REVAL;
526 goto retry;
527 }
528 out:
529 return error;
530 }
531
chmod_common(const struct path * path,umode_t mode)532 static int chmod_common(const struct path *path, umode_t mode)
533 {
534 struct inode *inode = path->dentry->d_inode;
535 struct inode *delegated_inode = NULL;
536 struct iattr newattrs;
537 int error;
538
539 error = mnt_want_write(path->mnt);
540 if (error)
541 return error;
542 retry_deleg:
543 inode_lock(inode);
544 error = security_path_chmod(path, mode);
545 if (error)
546 goto out_unlock;
547 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
548 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
549 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
550 out_unlock:
551 inode_unlock(inode);
552 if (delegated_inode) {
553 error = break_deleg_wait(&delegated_inode);
554 if (!error)
555 goto retry_deleg;
556 }
557 mnt_drop_write(path->mnt);
558 return error;
559 }
560
SYSCALL_DEFINE2(fchmod,unsigned int,fd,umode_t,mode)561 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
562 {
563 struct fd f = fdget(fd);
564 int err = -EBADF;
565
566 if (f.file) {
567 audit_file(f.file);
568 err = chmod_common(&f.file->f_path, mode);
569 fdput(f);
570 }
571 return err;
572 }
573
SYSCALL_DEFINE3(fchmodat,int,dfd,const char __user *,filename,umode_t,mode)574 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
575 {
576 struct path path;
577 int error;
578 unsigned int lookup_flags = LOOKUP_FOLLOW;
579 retry:
580 error = user_path_at(dfd, filename, lookup_flags, &path);
581 if (!error) {
582 error = chmod_common(&path, mode);
583 path_put(&path);
584 if (retry_estale(error, lookup_flags)) {
585 lookup_flags |= LOOKUP_REVAL;
586 goto retry;
587 }
588 }
589 return error;
590 }
591
SYSCALL_DEFINE2(chmod,const char __user *,filename,umode_t,mode)592 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
593 {
594 return sys_fchmodat(AT_FDCWD, filename, mode);
595 }
596
chown_common(const struct path * path,uid_t user,gid_t group)597 static int chown_common(const struct path *path, uid_t user, gid_t group)
598 {
599 struct inode *inode = path->dentry->d_inode;
600 struct inode *delegated_inode = NULL;
601 int error;
602 struct iattr newattrs;
603 kuid_t uid;
604 kgid_t gid;
605
606 uid = make_kuid(current_user_ns(), user);
607 gid = make_kgid(current_user_ns(), group);
608
609 retry_deleg:
610 newattrs.ia_valid = ATTR_CTIME;
611 if (user != (uid_t) -1) {
612 if (!uid_valid(uid))
613 return -EINVAL;
614 newattrs.ia_valid |= ATTR_UID;
615 newattrs.ia_uid = uid;
616 }
617 if (group != (gid_t) -1) {
618 if (!gid_valid(gid))
619 return -EINVAL;
620 newattrs.ia_valid |= ATTR_GID;
621 newattrs.ia_gid = gid;
622 }
623 if (!S_ISDIR(inode->i_mode))
624 newattrs.ia_valid |=
625 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
626 inode_lock(inode);
627 error = security_path_chown(path, uid, gid);
628 if (!error)
629 error = notify_change2(path->mnt, path->dentry, &newattrs, &delegated_inode);
630 inode_unlock(inode);
631 if (delegated_inode) {
632 error = break_deleg_wait(&delegated_inode);
633 if (!error)
634 goto retry_deleg;
635 }
636 return error;
637 }
638
SYSCALL_DEFINE5(fchownat,int,dfd,const char __user *,filename,uid_t,user,gid_t,group,int,flag)639 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
640 gid_t, group, int, flag)
641 {
642 struct path path;
643 int error = -EINVAL;
644 int lookup_flags;
645
646 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
647 goto out;
648
649 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
650 if (flag & AT_EMPTY_PATH)
651 lookup_flags |= LOOKUP_EMPTY;
652 retry:
653 error = user_path_at(dfd, filename, lookup_flags, &path);
654 if (error)
655 goto out;
656 error = mnt_want_write(path.mnt);
657 if (error)
658 goto out_release;
659 error = chown_common(&path, user, group);
660 mnt_drop_write(path.mnt);
661 out_release:
662 path_put(&path);
663 if (retry_estale(error, lookup_flags)) {
664 lookup_flags |= LOOKUP_REVAL;
665 goto retry;
666 }
667 out:
668 return error;
669 }
670
SYSCALL_DEFINE3(chown,const char __user *,filename,uid_t,user,gid_t,group)671 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
672 {
673 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
674 }
675
SYSCALL_DEFINE3(lchown,const char __user *,filename,uid_t,user,gid_t,group)676 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
677 {
678 return sys_fchownat(AT_FDCWD, filename, user, group,
679 AT_SYMLINK_NOFOLLOW);
680 }
681
SYSCALL_DEFINE3(fchown,unsigned int,fd,uid_t,user,gid_t,group)682 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
683 {
684 struct fd f = fdget(fd);
685 int error = -EBADF;
686
687 if (!f.file)
688 goto out;
689
690 error = mnt_want_write_file(f.file);
691 if (error)
692 goto out_fput;
693 audit_file(f.file);
694 error = chown_common(&f.file->f_path, user, group);
695 mnt_drop_write_file(f.file);
696 out_fput:
697 fdput(f);
698 out:
699 return error;
700 }
701
open_check_o_direct(struct file * f)702 int open_check_o_direct(struct file *f)
703 {
704 /* NB: we're sure to have correct a_ops only after f_op->open */
705 if (f->f_flags & O_DIRECT) {
706 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
707 return -EINVAL;
708 }
709 return 0;
710 }
711
do_dentry_open(struct file * f,struct inode * inode,int (* open)(struct inode *,struct file *),const struct cred * cred)712 static int do_dentry_open(struct file *f,
713 struct inode *inode,
714 int (*open)(struct inode *, struct file *),
715 const struct cred *cred)
716 {
717 static const struct file_operations empty_fops = {};
718 int error;
719
720 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
721 FMODE_PREAD | FMODE_PWRITE;
722
723 path_get(&f->f_path);
724 f->f_inode = inode;
725 f->f_mapping = inode->i_mapping;
726
727 if (unlikely(f->f_flags & O_PATH)) {
728 f->f_mode = FMODE_PATH;
729 f->f_op = &empty_fops;
730 return 0;
731 }
732
733 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
734 error = get_write_access(inode);
735 if (unlikely(error))
736 goto cleanup_file;
737 error = __mnt_want_write(f->f_path.mnt);
738 if (unlikely(error)) {
739 put_write_access(inode);
740 goto cleanup_file;
741 }
742 f->f_mode |= FMODE_WRITER;
743 }
744
745 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
746 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
747 f->f_mode |= FMODE_ATOMIC_POS;
748
749 f->f_op = fops_get(inode->i_fop);
750 if (unlikely(WARN_ON(!f->f_op))) {
751 error = -ENODEV;
752 goto cleanup_all;
753 }
754
755 error = security_file_open(f, cred);
756 if (error)
757 goto cleanup_all;
758
759 error = break_lease(locks_inode(f), f->f_flags);
760 if (error)
761 goto cleanup_all;
762
763 if (!open)
764 open = f->f_op->open;
765 if (open) {
766 error = open(inode, f);
767 if (error)
768 goto cleanup_all;
769 }
770 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
771 i_readcount_inc(inode);
772 if ((f->f_mode & FMODE_READ) &&
773 likely(f->f_op->read || f->f_op->read_iter))
774 f->f_mode |= FMODE_CAN_READ;
775 if ((f->f_mode & FMODE_WRITE) &&
776 likely(f->f_op->write || f->f_op->write_iter))
777 f->f_mode |= FMODE_CAN_WRITE;
778
779 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
780
781 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
782
783 return 0;
784
785 cleanup_all:
786 fops_put(f->f_op);
787 if (f->f_mode & FMODE_WRITER) {
788 put_write_access(inode);
789 __mnt_drop_write(f->f_path.mnt);
790 }
791 cleanup_file:
792 path_put(&f->f_path);
793 f->f_path.mnt = NULL;
794 f->f_path.dentry = NULL;
795 f->f_inode = NULL;
796 return error;
797 }
798
799 /**
800 * finish_open - finish opening a file
801 * @file: file pointer
802 * @dentry: pointer to dentry
803 * @open: open callback
804 * @opened: state of open
805 *
806 * This can be used to finish opening a file passed to i_op->atomic_open().
807 *
808 * If the open callback is set to NULL, then the standard f_op->open()
809 * filesystem callback is substituted.
810 *
811 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
812 * the return value of d_splice_alias(), then the caller needs to perform dput()
813 * on it after finish_open().
814 *
815 * On successful return @file is a fully instantiated open file. After this, if
816 * an error occurs in ->atomic_open(), it needs to clean up with fput().
817 *
818 * Returns zero on success or -errno if the open failed.
819 */
finish_open(struct file * file,struct dentry * dentry,int (* open)(struct inode *,struct file *),int * opened)820 int finish_open(struct file *file, struct dentry *dentry,
821 int (*open)(struct inode *, struct file *),
822 int *opened)
823 {
824 int error;
825 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
826
827 file->f_path.dentry = dentry;
828 error = do_dentry_open(file, d_backing_inode(dentry), open,
829 current_cred());
830 if (!error)
831 *opened |= FILE_OPENED;
832
833 return error;
834 }
835 EXPORT_SYMBOL(finish_open);
836
837 /**
838 * finish_no_open - finish ->atomic_open() without opening the file
839 *
840 * @file: file pointer
841 * @dentry: dentry or NULL (as returned from ->lookup())
842 *
843 * This can be used to set the result of a successful lookup in ->atomic_open().
844 *
845 * NB: unlike finish_open() this function does consume the dentry reference and
846 * the caller need not dput() it.
847 *
848 * Returns "1" which must be the return value of ->atomic_open() after having
849 * called this function.
850 */
finish_no_open(struct file * file,struct dentry * dentry)851 int finish_no_open(struct file *file, struct dentry *dentry)
852 {
853 file->f_path.dentry = dentry;
854 return 1;
855 }
856 EXPORT_SYMBOL(finish_no_open);
857
file_path(struct file * filp,char * buf,int buflen)858 char *file_path(struct file *filp, char *buf, int buflen)
859 {
860 return d_path(&filp->f_path, buf, buflen);
861 }
862 EXPORT_SYMBOL(file_path);
863
864 /**
865 * vfs_open - open the file at the given path
866 * @path: path to open
867 * @file: newly allocated file with f_flag initialized
868 * @cred: credentials to use
869 */
vfs_open(const struct path * path,struct file * file,const struct cred * cred)870 int vfs_open(const struct path *path, struct file *file,
871 const struct cred *cred)
872 {
873 struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags);
874
875 if (IS_ERR(dentry))
876 return PTR_ERR(dentry);
877
878 file->f_path = *path;
879 return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
880 }
881
dentry_open(const struct path * path,int flags,const struct cred * cred)882 struct file *dentry_open(const struct path *path, int flags,
883 const struct cred *cred)
884 {
885 int error;
886 struct file *f;
887
888 validate_creds(cred);
889
890 /* We must always pass in a valid mount pointer. */
891 BUG_ON(!path->mnt);
892
893 f = get_empty_filp();
894 if (!IS_ERR(f)) {
895 f->f_flags = flags;
896 error = vfs_open(path, f, cred);
897 if (!error) {
898 /* from now on we need fput() to dispose of f */
899 error = open_check_o_direct(f);
900 if (error) {
901 fput(f);
902 f = ERR_PTR(error);
903 }
904 } else {
905 put_filp(f);
906 f = ERR_PTR(error);
907 }
908 }
909 return f;
910 }
911 EXPORT_SYMBOL(dentry_open);
912
build_open_flags(int flags,umode_t mode,struct open_flags * op)913 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
914 {
915 int lookup_flags = 0;
916 int acc_mode = ACC_MODE(flags);
917
918 /*
919 * Clear out all open flags we don't know about so that we don't report
920 * them in fcntl(F_GETFD) or similar interfaces.
921 */
922 flags &= VALID_OPEN_FLAGS;
923
924 if (flags & (O_CREAT | __O_TMPFILE))
925 op->mode = (mode & S_IALLUGO) | S_IFREG;
926 else
927 op->mode = 0;
928
929 /* Must never be set by userspace */
930 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
931
932 /*
933 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
934 * check for O_DSYNC if the need any syncing at all we enforce it's
935 * always set instead of having to deal with possibly weird behaviour
936 * for malicious applications setting only __O_SYNC.
937 */
938 if (flags & __O_SYNC)
939 flags |= O_DSYNC;
940
941 if (flags & __O_TMPFILE) {
942 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
943 return -EINVAL;
944 if (!(acc_mode & MAY_WRITE))
945 return -EINVAL;
946 } else if (flags & O_PATH) {
947 /*
948 * If we have O_PATH in the open flag. Then we
949 * cannot have anything other than the below set of flags
950 */
951 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
952 acc_mode = 0;
953 }
954
955 op->open_flag = flags;
956
957 /* O_TRUNC implies we need access checks for write permissions */
958 if (flags & O_TRUNC)
959 acc_mode |= MAY_WRITE;
960
961 /* Allow the LSM permission hook to distinguish append
962 access from general write access. */
963 if (flags & O_APPEND)
964 acc_mode |= MAY_APPEND;
965
966 op->acc_mode = acc_mode;
967
968 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
969
970 if (flags & O_CREAT) {
971 op->intent |= LOOKUP_CREATE;
972 if (flags & O_EXCL)
973 op->intent |= LOOKUP_EXCL;
974 }
975
976 if (flags & O_DIRECTORY)
977 lookup_flags |= LOOKUP_DIRECTORY;
978 if (!(flags & O_NOFOLLOW))
979 lookup_flags |= LOOKUP_FOLLOW;
980 op->lookup_flags = lookup_flags;
981 return 0;
982 }
983
984 /**
985 * file_open_name - open file and return file pointer
986 *
987 * @name: struct filename containing path to open
988 * @flags: open flags as per the open(2) second argument
989 * @mode: mode for the new file if O_CREAT is set, else ignored
990 *
991 * This is the helper to open a file from kernelspace if you really
992 * have to. But in generally you should not do this, so please move
993 * along, nothing to see here..
994 */
file_open_name(struct filename * name,int flags,umode_t mode)995 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
996 {
997 struct open_flags op;
998 int err = build_open_flags(flags, mode, &op);
999 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
1000 }
1001
1002 /**
1003 * filp_open - open file and return file pointer
1004 *
1005 * @filename: path to open
1006 * @flags: open flags as per the open(2) second argument
1007 * @mode: mode for the new file if O_CREAT is set, else ignored
1008 *
1009 * This is the helper to open a file from kernelspace if you really
1010 * have to. But in generally you should not do this, so please move
1011 * along, nothing to see here..
1012 */
filp_open(const char * filename,int flags,umode_t mode)1013 struct file *filp_open(const char *filename, int flags, umode_t mode)
1014 {
1015 struct filename *name = getname_kernel(filename);
1016 struct file *file = ERR_CAST(name);
1017
1018 if (!IS_ERR(name)) {
1019 file = file_open_name(name, flags, mode);
1020 putname(name);
1021 }
1022 return file;
1023 }
1024 EXPORT_SYMBOL(filp_open);
1025
file_open_root(struct dentry * dentry,struct vfsmount * mnt,const char * filename,int flags,umode_t mode)1026 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1027 const char *filename, int flags, umode_t mode)
1028 {
1029 struct open_flags op;
1030 int err = build_open_flags(flags, mode, &op);
1031 if (err)
1032 return ERR_PTR(err);
1033 return do_file_open_root(dentry, mnt, filename, &op);
1034 }
1035 EXPORT_SYMBOL(file_open_root);
1036
filp_clone_open(struct file * oldfile)1037 struct file *filp_clone_open(struct file *oldfile)
1038 {
1039 struct file *file;
1040 int retval;
1041
1042 file = get_empty_filp();
1043 if (IS_ERR(file))
1044 return file;
1045
1046 file->f_flags = oldfile->f_flags;
1047 retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred);
1048 if (retval) {
1049 put_filp(file);
1050 return ERR_PTR(retval);
1051 }
1052
1053 return file;
1054 }
1055 EXPORT_SYMBOL(filp_clone_open);
1056
do_sys_open(int dfd,const char __user * filename,int flags,umode_t mode)1057 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1058 {
1059 struct open_flags op;
1060 int fd = build_open_flags(flags, mode, &op);
1061 struct filename *tmp;
1062
1063 if (fd)
1064 return fd;
1065
1066 tmp = getname(filename);
1067 if (IS_ERR(tmp))
1068 return PTR_ERR(tmp);
1069
1070 fd = get_unused_fd_flags(flags);
1071 if (fd >= 0) {
1072 struct file *f = do_filp_open(dfd, tmp, &op);
1073 if (IS_ERR(f)) {
1074 put_unused_fd(fd);
1075 fd = PTR_ERR(f);
1076 } else {
1077 fsnotify_open(f);
1078 fd_install(fd, f);
1079 }
1080 }
1081 putname(tmp);
1082 return fd;
1083 }
1084
SYSCALL_DEFINE3(open,const char __user *,filename,int,flags,umode_t,mode)1085 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1086 {
1087 if (force_o_largefile())
1088 flags |= O_LARGEFILE;
1089
1090 return do_sys_open(AT_FDCWD, filename, flags, mode);
1091 }
1092
SYSCALL_DEFINE4(openat,int,dfd,const char __user *,filename,int,flags,umode_t,mode)1093 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1094 umode_t, mode)
1095 {
1096 if (force_o_largefile())
1097 flags |= O_LARGEFILE;
1098
1099 return do_sys_open(dfd, filename, flags, mode);
1100 }
1101
1102 #ifndef __alpha__
1103
1104 /*
1105 * For backward compatibility? Maybe this should be moved
1106 * into arch/i386 instead?
1107 */
SYSCALL_DEFINE2(creat,const char __user *,pathname,umode_t,mode)1108 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1109 {
1110 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1111 }
1112
1113 #endif
1114
1115 /*
1116 * "id" is the POSIX thread ID. We use the
1117 * files pointer for this..
1118 */
filp_close(struct file * filp,fl_owner_t id)1119 int filp_close(struct file *filp, fl_owner_t id)
1120 {
1121 int retval = 0;
1122
1123 if (!file_count(filp)) {
1124 printk(KERN_ERR "VFS: Close: file count is 0\n");
1125 return 0;
1126 }
1127
1128 if (filp->f_op->flush)
1129 retval = filp->f_op->flush(filp, id);
1130
1131 if (likely(!(filp->f_mode & FMODE_PATH))) {
1132 dnotify_flush(filp, id);
1133 locks_remove_posix(filp, id);
1134 }
1135 fput(filp);
1136 return retval;
1137 }
1138
1139 EXPORT_SYMBOL(filp_close);
1140
1141 /*
1142 * Careful here! We test whether the file pointer is NULL before
1143 * releasing the fd. This ensures that one clone task can't release
1144 * an fd while another clone is opening it.
1145 */
SYSCALL_DEFINE1(close,unsigned int,fd)1146 SYSCALL_DEFINE1(close, unsigned int, fd)
1147 {
1148 int retval = __close_fd(current->files, fd);
1149
1150 /* can't restart close syscall because file table entry was cleared */
1151 if (unlikely(retval == -ERESTARTSYS ||
1152 retval == -ERESTARTNOINTR ||
1153 retval == -ERESTARTNOHAND ||
1154 retval == -ERESTART_RESTARTBLOCK))
1155 retval = -EINTR;
1156
1157 return retval;
1158 }
1159 EXPORT_SYMBOL(sys_close);
1160
1161 /*
1162 * This routine simulates a hangup on the tty, to arrange that users
1163 * are given clean terminals at login time.
1164 */
SYSCALL_DEFINE0(vhangup)1165 SYSCALL_DEFINE0(vhangup)
1166 {
1167 if (capable(CAP_SYS_TTY_CONFIG)) {
1168 tty_vhangup_self();
1169 return 0;
1170 }
1171 return -EPERM;
1172 }
1173
1174 /*
1175 * Called when an inode is about to be open.
1176 * We use this to disallow opening large files on 32bit systems if
1177 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1178 * on this flag in sys_open.
1179 */
generic_file_open(struct inode * inode,struct file * filp)1180 int generic_file_open(struct inode * inode, struct file * filp)
1181 {
1182 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1183 return -EOVERFLOW;
1184 return 0;
1185 }
1186
1187 EXPORT_SYMBOL(generic_file_open);
1188
1189 /*
1190 * This is used by subsystems that don't want seekable
1191 * file descriptors. The function is not supposed to ever fail, the only
1192 * reason it returns an 'int' and not 'void' is so that it can be plugged
1193 * directly into file_operations structure.
1194 */
nonseekable_open(struct inode * inode,struct file * filp)1195 int nonseekable_open(struct inode *inode, struct file *filp)
1196 {
1197 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1198 return 0;
1199 }
1200
1201 EXPORT_SYMBOL(nonseekable_open);
1202