xref: /linux/fs/stat.c (revision fd7d598270724cc787982ea48bbe17ad383a8b7f)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/stat.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  */
7 
8 #include <linux/blkdev.h>
9 #include <linux/export.h>
10 #include <linux/mm.h>
11 #include <linux/errno.h>
12 #include <linux/file.h>
13 #include <linux/highuid.h>
14 #include <linux/fs.h>
15 #include <linux/namei.h>
16 #include <linux/security.h>
17 #include <linux/cred.h>
18 #include <linux/syscalls.h>
19 #include <linux/pagemap.h>
20 #include <linux/compat.h>
21 #include <linux/iversion.h>
22 
23 #include <linux/uaccess.h>
24 #include <asm/unistd.h>
25 
26 #include "internal.h"
27 #include "mount.h"
28 
29 /**
30  * generic_fillattr - Fill in the basic attributes from the inode struct
31  * @idmap:		idmap of the mount the inode was found from
32  * @request_mask:	statx request_mask
33  * @inode:		Inode to use as the source
34  * @stat:		Where to fill in the attributes
35  *
36  * Fill in the basic attributes in the kstat structure from data that's to be
37  * found on the VFS inode structure.  This is the default if no getattr inode
38  * operation is supplied.
39  *
40  * If the inode has been found through an idmapped mount the idmap of
41  * the vfsmount must be passed through @idmap. This function will then
42  * take care to map the inode according to @idmap before filling in the
43  * uid and gid filds. On non-idmapped mounts or if permission checking is to be
44  * performed on the raw inode simply passs @nop_mnt_idmap.
45  */
46 void generic_fillattr(struct mnt_idmap *idmap, u32 request_mask,
47 		      struct inode *inode, struct kstat *stat)
48 {
49 	vfsuid_t vfsuid = i_uid_into_vfsuid(idmap, inode);
50 	vfsgid_t vfsgid = i_gid_into_vfsgid(idmap, inode);
51 
52 	stat->dev = inode->i_sb->s_dev;
53 	stat->ino = inode->i_ino;
54 	stat->mode = inode->i_mode;
55 	stat->nlink = inode->i_nlink;
56 	stat->uid = vfsuid_into_kuid(vfsuid);
57 	stat->gid = vfsgid_into_kgid(vfsgid);
58 	stat->rdev = inode->i_rdev;
59 	stat->size = i_size_read(inode);
60 	stat->atime = inode->i_atime;
61 	stat->mtime = inode->i_mtime;
62 	stat->ctime = inode_get_ctime(inode);
63 	stat->blksize = i_blocksize(inode);
64 	stat->blocks = inode->i_blocks;
65 
66 	if ((request_mask & STATX_CHANGE_COOKIE) && IS_I_VERSION(inode)) {
67 		stat->result_mask |= STATX_CHANGE_COOKIE;
68 		stat->change_cookie = inode_query_iversion(inode);
69 	}
70 
71 }
72 EXPORT_SYMBOL(generic_fillattr);
73 
74 /**
75  * generic_fill_statx_attr - Fill in the statx attributes from the inode flags
76  * @inode:	Inode to use as the source
77  * @stat:	Where to fill in the attribute flags
78  *
79  * Fill in the STATX_ATTR_* flags in the kstat structure for properties of the
80  * inode that are published on i_flags and enforced by the VFS.
81  */
82 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat)
83 {
84 	if (inode->i_flags & S_IMMUTABLE)
85 		stat->attributes |= STATX_ATTR_IMMUTABLE;
86 	if (inode->i_flags & S_APPEND)
87 		stat->attributes |= STATX_ATTR_APPEND;
88 	stat->attributes_mask |= KSTAT_ATTR_VFS_FLAGS;
89 }
90 EXPORT_SYMBOL(generic_fill_statx_attr);
91 
92 /**
93  * vfs_getattr_nosec - getattr without security checks
94  * @path: file to get attributes from
95  * @stat: structure to return attributes in
96  * @request_mask: STATX_xxx flags indicating what the caller wants
97  * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
98  *
99  * Get attributes without calling security_inode_getattr.
100  *
101  * Currently the only caller other than vfs_getattr is internal to the
102  * filehandle lookup code, which uses only the inode number and returns no
103  * attributes to any user.  Any other code probably wants vfs_getattr.
104  */
105 int vfs_getattr_nosec(const struct path *path, struct kstat *stat,
106 		      u32 request_mask, unsigned int query_flags)
107 {
108 	struct mnt_idmap *idmap;
109 	struct inode *inode = d_backing_inode(path->dentry);
110 
111 	memset(stat, 0, sizeof(*stat));
112 	stat->result_mask |= STATX_BASIC_STATS;
113 	query_flags &= AT_STATX_SYNC_TYPE;
114 
115 	/* allow the fs to override these if it really wants to */
116 	/* SB_NOATIME means filesystem supplies dummy atime value */
117 	if (inode->i_sb->s_flags & SB_NOATIME)
118 		stat->result_mask &= ~STATX_ATIME;
119 
120 	/*
121 	 * Note: If you add another clause to set an attribute flag, please
122 	 * update attributes_mask below.
123 	 */
124 	if (IS_AUTOMOUNT(inode))
125 		stat->attributes |= STATX_ATTR_AUTOMOUNT;
126 
127 	if (IS_DAX(inode))
128 		stat->attributes |= STATX_ATTR_DAX;
129 
130 	stat->attributes_mask |= (STATX_ATTR_AUTOMOUNT |
131 				  STATX_ATTR_DAX);
132 
133 	idmap = mnt_idmap(path->mnt);
134 	if (inode->i_op->getattr)
135 		return inode->i_op->getattr(idmap, path, stat,
136 					    request_mask, query_flags);
137 
138 	generic_fillattr(idmap, request_mask, inode, stat);
139 	return 0;
140 }
141 EXPORT_SYMBOL(vfs_getattr_nosec);
142 
143 /*
144  * vfs_getattr - Get the enhanced basic attributes of a file
145  * @path: The file of interest
146  * @stat: Where to return the statistics
147  * @request_mask: STATX_xxx flags indicating what the caller wants
148  * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
149  *
150  * Ask the filesystem for a file's attributes.  The caller must indicate in
151  * request_mask and query_flags to indicate what they want.
152  *
153  * If the file is remote, the filesystem can be forced to update the attributes
154  * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
155  * suppress the update by passing AT_STATX_DONT_SYNC.
156  *
157  * Bits must have been set in request_mask to indicate which attributes the
158  * caller wants retrieving.  Any such attribute not requested may be returned
159  * anyway, but the value may be approximate, and, if remote, may not have been
160  * synchronised with the server.
161  *
162  * 0 will be returned on success, and a -ve error code if unsuccessful.
163  */
164 int vfs_getattr(const struct path *path, struct kstat *stat,
165 		u32 request_mask, unsigned int query_flags)
166 {
167 	int retval;
168 
169 	retval = security_inode_getattr(path);
170 	if (retval)
171 		return retval;
172 	return vfs_getattr_nosec(path, stat, request_mask, query_flags);
173 }
174 EXPORT_SYMBOL(vfs_getattr);
175 
176 /**
177  * vfs_fstat - Get the basic attributes by file descriptor
178  * @fd: The file descriptor referring to the file of interest
179  * @stat: The result structure to fill in.
180  *
181  * This function is a wrapper around vfs_getattr().  The main difference is
182  * that it uses a file descriptor to determine the file location.
183  *
184  * 0 will be returned on success, and a -ve error code if unsuccessful.
185  */
186 int vfs_fstat(int fd, struct kstat *stat)
187 {
188 	struct fd f;
189 	int error;
190 
191 	f = fdget_raw(fd);
192 	if (!f.file)
193 		return -EBADF;
194 	error = vfs_getattr(&f.file->f_path, stat, STATX_BASIC_STATS, 0);
195 	fdput(f);
196 	return error;
197 }
198 
199 int getname_statx_lookup_flags(int flags)
200 {
201 	int lookup_flags = 0;
202 
203 	if (!(flags & AT_SYMLINK_NOFOLLOW))
204 		lookup_flags |= LOOKUP_FOLLOW;
205 	if (!(flags & AT_NO_AUTOMOUNT))
206 		lookup_flags |= LOOKUP_AUTOMOUNT;
207 	if (flags & AT_EMPTY_PATH)
208 		lookup_flags |= LOOKUP_EMPTY;
209 
210 	return lookup_flags;
211 }
212 
213 /**
214  * vfs_statx - Get basic and extra attributes by filename
215  * @dfd: A file descriptor representing the base dir for a relative filename
216  * @filename: The name of the file of interest
217  * @flags: Flags to control the query
218  * @stat: The result structure to fill in.
219  * @request_mask: STATX_xxx flags indicating what the caller wants
220  *
221  * This function is a wrapper around vfs_getattr().  The main difference is
222  * that it uses a filename and base directory to determine the file location.
223  * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
224  * at the given name from being referenced.
225  *
226  * 0 will be returned on success, and a -ve error code if unsuccessful.
227  */
228 static int vfs_statx(int dfd, struct filename *filename, int flags,
229 	      struct kstat *stat, u32 request_mask)
230 {
231 	struct path path;
232 	unsigned int lookup_flags = getname_statx_lookup_flags(flags);
233 	int error;
234 
235 	if (flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT | AT_EMPTY_PATH |
236 		      AT_STATX_SYNC_TYPE))
237 		return -EINVAL;
238 
239 retry:
240 	error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
241 	if (error)
242 		goto out;
243 
244 	error = vfs_getattr(&path, stat, request_mask, flags);
245 
246 	stat->mnt_id = real_mount(path.mnt)->mnt_id;
247 	stat->result_mask |= STATX_MNT_ID;
248 
249 	if (path.mnt->mnt_root == path.dentry)
250 		stat->attributes |= STATX_ATTR_MOUNT_ROOT;
251 	stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
252 
253 	/* Handle STATX_DIOALIGN for block devices. */
254 	if (request_mask & STATX_DIOALIGN) {
255 		struct inode *inode = d_backing_inode(path.dentry);
256 
257 		if (S_ISBLK(inode->i_mode))
258 			bdev_statx_dioalign(inode, stat);
259 	}
260 
261 	path_put(&path);
262 	if (retry_estale(error, lookup_flags)) {
263 		lookup_flags |= LOOKUP_REVAL;
264 		goto retry;
265 	}
266 out:
267 	return error;
268 }
269 
270 int vfs_fstatat(int dfd, const char __user *filename,
271 			      struct kstat *stat, int flags)
272 {
273 	int ret;
274 	int statx_flags = flags | AT_NO_AUTOMOUNT;
275 	struct filename *name;
276 
277 	/*
278 	 * Work around glibc turning fstat() into fstatat(AT_EMPTY_PATH)
279 	 *
280 	 * If AT_EMPTY_PATH is set, we expect the common case to be that
281 	 * empty path, and avoid doing all the extra pathname work.
282 	 */
283 	if (dfd >= 0 && flags == AT_EMPTY_PATH) {
284 		char c;
285 
286 		ret = get_user(c, filename);
287 		if (unlikely(ret))
288 			return ret;
289 
290 		if (likely(!c))
291 			return vfs_fstat(dfd, stat);
292 	}
293 
294 	name = getname_flags(filename, getname_statx_lookup_flags(statx_flags), NULL);
295 	ret = vfs_statx(dfd, name, statx_flags, stat, STATX_BASIC_STATS);
296 	putname(name);
297 
298 	return ret;
299 }
300 
301 #ifdef __ARCH_WANT_OLD_STAT
302 
303 /*
304  * For backward compatibility?  Maybe this should be moved
305  * into arch/i386 instead?
306  */
307 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
308 {
309 	static int warncount = 5;
310 	struct __old_kernel_stat tmp;
311 
312 	if (warncount > 0) {
313 		warncount--;
314 		printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
315 			current->comm);
316 	} else if (warncount < 0) {
317 		/* it's laughable, but... */
318 		warncount = 0;
319 	}
320 
321 	memset(&tmp, 0, sizeof(struct __old_kernel_stat));
322 	tmp.st_dev = old_encode_dev(stat->dev);
323 	tmp.st_ino = stat->ino;
324 	if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
325 		return -EOVERFLOW;
326 	tmp.st_mode = stat->mode;
327 	tmp.st_nlink = stat->nlink;
328 	if (tmp.st_nlink != stat->nlink)
329 		return -EOVERFLOW;
330 	SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
331 	SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
332 	tmp.st_rdev = old_encode_dev(stat->rdev);
333 #if BITS_PER_LONG == 32
334 	if (stat->size > MAX_NON_LFS)
335 		return -EOVERFLOW;
336 #endif
337 	tmp.st_size = stat->size;
338 	tmp.st_atime = stat->atime.tv_sec;
339 	tmp.st_mtime = stat->mtime.tv_sec;
340 	tmp.st_ctime = stat->ctime.tv_sec;
341 	return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
342 }
343 
344 SYSCALL_DEFINE2(stat, const char __user *, filename,
345 		struct __old_kernel_stat __user *, statbuf)
346 {
347 	struct kstat stat;
348 	int error;
349 
350 	error = vfs_stat(filename, &stat);
351 	if (error)
352 		return error;
353 
354 	return cp_old_stat(&stat, statbuf);
355 }
356 
357 SYSCALL_DEFINE2(lstat, const char __user *, filename,
358 		struct __old_kernel_stat __user *, statbuf)
359 {
360 	struct kstat stat;
361 	int error;
362 
363 	error = vfs_lstat(filename, &stat);
364 	if (error)
365 		return error;
366 
367 	return cp_old_stat(&stat, statbuf);
368 }
369 
370 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
371 {
372 	struct kstat stat;
373 	int error = vfs_fstat(fd, &stat);
374 
375 	if (!error)
376 		error = cp_old_stat(&stat, statbuf);
377 
378 	return error;
379 }
380 
381 #endif /* __ARCH_WANT_OLD_STAT */
382 
383 #ifdef __ARCH_WANT_NEW_STAT
384 
385 #ifndef INIT_STRUCT_STAT_PADDING
386 #  define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
387 #endif
388 
389 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
390 {
391 	struct stat tmp;
392 
393 	if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
394 		return -EOVERFLOW;
395 	if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
396 		return -EOVERFLOW;
397 #if BITS_PER_LONG == 32
398 	if (stat->size > MAX_NON_LFS)
399 		return -EOVERFLOW;
400 #endif
401 
402 	INIT_STRUCT_STAT_PADDING(tmp);
403 	tmp.st_dev = new_encode_dev(stat->dev);
404 	tmp.st_ino = stat->ino;
405 	if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
406 		return -EOVERFLOW;
407 	tmp.st_mode = stat->mode;
408 	tmp.st_nlink = stat->nlink;
409 	if (tmp.st_nlink != stat->nlink)
410 		return -EOVERFLOW;
411 	SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
412 	SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
413 	tmp.st_rdev = new_encode_dev(stat->rdev);
414 	tmp.st_size = stat->size;
415 	tmp.st_atime = stat->atime.tv_sec;
416 	tmp.st_mtime = stat->mtime.tv_sec;
417 	tmp.st_ctime = stat->ctime.tv_sec;
418 #ifdef STAT_HAVE_NSEC
419 	tmp.st_atime_nsec = stat->atime.tv_nsec;
420 	tmp.st_mtime_nsec = stat->mtime.tv_nsec;
421 	tmp.st_ctime_nsec = stat->ctime.tv_nsec;
422 #endif
423 	tmp.st_blocks = stat->blocks;
424 	tmp.st_blksize = stat->blksize;
425 	return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
426 }
427 
428 SYSCALL_DEFINE2(newstat, const char __user *, filename,
429 		struct stat __user *, statbuf)
430 {
431 	struct kstat stat;
432 	int error = vfs_stat(filename, &stat);
433 
434 	if (error)
435 		return error;
436 	return cp_new_stat(&stat, statbuf);
437 }
438 
439 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
440 		struct stat __user *, statbuf)
441 {
442 	struct kstat stat;
443 	int error;
444 
445 	error = vfs_lstat(filename, &stat);
446 	if (error)
447 		return error;
448 
449 	return cp_new_stat(&stat, statbuf);
450 }
451 
452 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
453 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
454 		struct stat __user *, statbuf, int, flag)
455 {
456 	struct kstat stat;
457 	int error;
458 
459 	error = vfs_fstatat(dfd, filename, &stat, flag);
460 	if (error)
461 		return error;
462 	return cp_new_stat(&stat, statbuf);
463 }
464 #endif
465 
466 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
467 {
468 	struct kstat stat;
469 	int error = vfs_fstat(fd, &stat);
470 
471 	if (!error)
472 		error = cp_new_stat(&stat, statbuf);
473 
474 	return error;
475 }
476 #endif
477 
478 static int do_readlinkat(int dfd, const char __user *pathname,
479 			 char __user *buf, int bufsiz)
480 {
481 	struct path path;
482 	int error;
483 	int empty = 0;
484 	unsigned int lookup_flags = LOOKUP_EMPTY;
485 
486 	if (bufsiz <= 0)
487 		return -EINVAL;
488 
489 retry:
490 	error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
491 	if (!error) {
492 		struct inode *inode = d_backing_inode(path.dentry);
493 
494 		error = empty ? -ENOENT : -EINVAL;
495 		/*
496 		 * AFS mountpoints allow readlink(2) but are not symlinks
497 		 */
498 		if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
499 			error = security_inode_readlink(path.dentry);
500 			if (!error) {
501 				touch_atime(&path);
502 				error = vfs_readlink(path.dentry, buf, bufsiz);
503 			}
504 		}
505 		path_put(&path);
506 		if (retry_estale(error, lookup_flags)) {
507 			lookup_flags |= LOOKUP_REVAL;
508 			goto retry;
509 		}
510 	}
511 	return error;
512 }
513 
514 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
515 		char __user *, buf, int, bufsiz)
516 {
517 	return do_readlinkat(dfd, pathname, buf, bufsiz);
518 }
519 
520 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
521 		int, bufsiz)
522 {
523 	return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
524 }
525 
526 
527 /* ---------- LFS-64 ----------- */
528 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
529 
530 #ifndef INIT_STRUCT_STAT64_PADDING
531 #  define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
532 #endif
533 
534 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
535 {
536 	struct stat64 tmp;
537 
538 	INIT_STRUCT_STAT64_PADDING(tmp);
539 #ifdef CONFIG_MIPS
540 	/* mips has weird padding, so we don't get 64 bits there */
541 	tmp.st_dev = new_encode_dev(stat->dev);
542 	tmp.st_rdev = new_encode_dev(stat->rdev);
543 #else
544 	tmp.st_dev = huge_encode_dev(stat->dev);
545 	tmp.st_rdev = huge_encode_dev(stat->rdev);
546 #endif
547 	tmp.st_ino = stat->ino;
548 	if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
549 		return -EOVERFLOW;
550 #ifdef STAT64_HAS_BROKEN_ST_INO
551 	tmp.__st_ino = stat->ino;
552 #endif
553 	tmp.st_mode = stat->mode;
554 	tmp.st_nlink = stat->nlink;
555 	tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
556 	tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
557 	tmp.st_atime = stat->atime.tv_sec;
558 	tmp.st_atime_nsec = stat->atime.tv_nsec;
559 	tmp.st_mtime = stat->mtime.tv_sec;
560 	tmp.st_mtime_nsec = stat->mtime.tv_nsec;
561 	tmp.st_ctime = stat->ctime.tv_sec;
562 	tmp.st_ctime_nsec = stat->ctime.tv_nsec;
563 	tmp.st_size = stat->size;
564 	tmp.st_blocks = stat->blocks;
565 	tmp.st_blksize = stat->blksize;
566 	return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
567 }
568 
569 SYSCALL_DEFINE2(stat64, const char __user *, filename,
570 		struct stat64 __user *, statbuf)
571 {
572 	struct kstat stat;
573 	int error = vfs_stat(filename, &stat);
574 
575 	if (!error)
576 		error = cp_new_stat64(&stat, statbuf);
577 
578 	return error;
579 }
580 
581 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
582 		struct stat64 __user *, statbuf)
583 {
584 	struct kstat stat;
585 	int error = vfs_lstat(filename, &stat);
586 
587 	if (!error)
588 		error = cp_new_stat64(&stat, statbuf);
589 
590 	return error;
591 }
592 
593 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
594 {
595 	struct kstat stat;
596 	int error = vfs_fstat(fd, &stat);
597 
598 	if (!error)
599 		error = cp_new_stat64(&stat, statbuf);
600 
601 	return error;
602 }
603 
604 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
605 		struct stat64 __user *, statbuf, int, flag)
606 {
607 	struct kstat stat;
608 	int error;
609 
610 	error = vfs_fstatat(dfd, filename, &stat, flag);
611 	if (error)
612 		return error;
613 	return cp_new_stat64(&stat, statbuf);
614 }
615 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
616 
617 static noinline_for_stack int
618 cp_statx(const struct kstat *stat, struct statx __user *buffer)
619 {
620 	struct statx tmp;
621 
622 	memset(&tmp, 0, sizeof(tmp));
623 
624 	/* STATX_CHANGE_COOKIE is kernel-only for now */
625 	tmp.stx_mask = stat->result_mask & ~STATX_CHANGE_COOKIE;
626 	tmp.stx_blksize = stat->blksize;
627 	/* STATX_ATTR_CHANGE_MONOTONIC is kernel-only for now */
628 	tmp.stx_attributes = stat->attributes & ~STATX_ATTR_CHANGE_MONOTONIC;
629 	tmp.stx_nlink = stat->nlink;
630 	tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
631 	tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
632 	tmp.stx_mode = stat->mode;
633 	tmp.stx_ino = stat->ino;
634 	tmp.stx_size = stat->size;
635 	tmp.stx_blocks = stat->blocks;
636 	tmp.stx_attributes_mask = stat->attributes_mask;
637 	tmp.stx_atime.tv_sec = stat->atime.tv_sec;
638 	tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
639 	tmp.stx_btime.tv_sec = stat->btime.tv_sec;
640 	tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
641 	tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
642 	tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
643 	tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
644 	tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
645 	tmp.stx_rdev_major = MAJOR(stat->rdev);
646 	tmp.stx_rdev_minor = MINOR(stat->rdev);
647 	tmp.stx_dev_major = MAJOR(stat->dev);
648 	tmp.stx_dev_minor = MINOR(stat->dev);
649 	tmp.stx_mnt_id = stat->mnt_id;
650 	tmp.stx_dio_mem_align = stat->dio_mem_align;
651 	tmp.stx_dio_offset_align = stat->dio_offset_align;
652 
653 	return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
654 }
655 
656 int do_statx(int dfd, struct filename *filename, unsigned int flags,
657 	     unsigned int mask, struct statx __user *buffer)
658 {
659 	struct kstat stat;
660 	int error;
661 
662 	if (mask & STATX__RESERVED)
663 		return -EINVAL;
664 	if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
665 		return -EINVAL;
666 
667 	/* STATX_CHANGE_COOKIE is kernel-only for now. Ignore requests
668 	 * from userland.
669 	 */
670 	mask &= ~STATX_CHANGE_COOKIE;
671 
672 	error = vfs_statx(dfd, filename, flags, &stat, mask);
673 	if (error)
674 		return error;
675 
676 	return cp_statx(&stat, buffer);
677 }
678 
679 /**
680  * sys_statx - System call to get enhanced stats
681  * @dfd: Base directory to pathwalk from *or* fd to stat.
682  * @filename: File to stat or "" with AT_EMPTY_PATH
683  * @flags: AT_* flags to control pathwalk.
684  * @mask: Parts of statx struct actually required.
685  * @buffer: Result buffer.
686  *
687  * Note that fstat() can be emulated by setting dfd to the fd of interest,
688  * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
689  */
690 SYSCALL_DEFINE5(statx,
691 		int, dfd, const char __user *, filename, unsigned, flags,
692 		unsigned int, mask,
693 		struct statx __user *, buffer)
694 {
695 	int ret;
696 	struct filename *name;
697 
698 	name = getname_flags(filename, getname_statx_lookup_flags(flags), NULL);
699 	ret = do_statx(dfd, name, flags, mask, buffer);
700 	putname(name);
701 
702 	return ret;
703 }
704 
705 #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_STAT)
706 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
707 {
708 	struct compat_stat tmp;
709 
710 	if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
711 		return -EOVERFLOW;
712 	if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
713 		return -EOVERFLOW;
714 
715 	memset(&tmp, 0, sizeof(tmp));
716 	tmp.st_dev = new_encode_dev(stat->dev);
717 	tmp.st_ino = stat->ino;
718 	if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
719 		return -EOVERFLOW;
720 	tmp.st_mode = stat->mode;
721 	tmp.st_nlink = stat->nlink;
722 	if (tmp.st_nlink != stat->nlink)
723 		return -EOVERFLOW;
724 	SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
725 	SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
726 	tmp.st_rdev = new_encode_dev(stat->rdev);
727 	if ((u64) stat->size > MAX_NON_LFS)
728 		return -EOVERFLOW;
729 	tmp.st_size = stat->size;
730 	tmp.st_atime = stat->atime.tv_sec;
731 	tmp.st_atime_nsec = stat->atime.tv_nsec;
732 	tmp.st_mtime = stat->mtime.tv_sec;
733 	tmp.st_mtime_nsec = stat->mtime.tv_nsec;
734 	tmp.st_ctime = stat->ctime.tv_sec;
735 	tmp.st_ctime_nsec = stat->ctime.tv_nsec;
736 	tmp.st_blocks = stat->blocks;
737 	tmp.st_blksize = stat->blksize;
738 	return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
739 }
740 
741 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
742 		       struct compat_stat __user *, statbuf)
743 {
744 	struct kstat stat;
745 	int error;
746 
747 	error = vfs_stat(filename, &stat);
748 	if (error)
749 		return error;
750 	return cp_compat_stat(&stat, statbuf);
751 }
752 
753 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
754 		       struct compat_stat __user *, statbuf)
755 {
756 	struct kstat stat;
757 	int error;
758 
759 	error = vfs_lstat(filename, &stat);
760 	if (error)
761 		return error;
762 	return cp_compat_stat(&stat, statbuf);
763 }
764 
765 #ifndef __ARCH_WANT_STAT64
766 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
767 		       const char __user *, filename,
768 		       struct compat_stat __user *, statbuf, int, flag)
769 {
770 	struct kstat stat;
771 	int error;
772 
773 	error = vfs_fstatat(dfd, filename, &stat, flag);
774 	if (error)
775 		return error;
776 	return cp_compat_stat(&stat, statbuf);
777 }
778 #endif
779 
780 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
781 		       struct compat_stat __user *, statbuf)
782 {
783 	struct kstat stat;
784 	int error = vfs_fstat(fd, &stat);
785 
786 	if (!error)
787 		error = cp_compat_stat(&stat, statbuf);
788 	return error;
789 }
790 #endif
791 
792 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
793 void __inode_add_bytes(struct inode *inode, loff_t bytes)
794 {
795 	inode->i_blocks += bytes >> 9;
796 	bytes &= 511;
797 	inode->i_bytes += bytes;
798 	if (inode->i_bytes >= 512) {
799 		inode->i_blocks++;
800 		inode->i_bytes -= 512;
801 	}
802 }
803 EXPORT_SYMBOL(__inode_add_bytes);
804 
805 void inode_add_bytes(struct inode *inode, loff_t bytes)
806 {
807 	spin_lock(&inode->i_lock);
808 	__inode_add_bytes(inode, bytes);
809 	spin_unlock(&inode->i_lock);
810 }
811 
812 EXPORT_SYMBOL(inode_add_bytes);
813 
814 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
815 {
816 	inode->i_blocks -= bytes >> 9;
817 	bytes &= 511;
818 	if (inode->i_bytes < bytes) {
819 		inode->i_blocks--;
820 		inode->i_bytes += 512;
821 	}
822 	inode->i_bytes -= bytes;
823 }
824 
825 EXPORT_SYMBOL(__inode_sub_bytes);
826 
827 void inode_sub_bytes(struct inode *inode, loff_t bytes)
828 {
829 	spin_lock(&inode->i_lock);
830 	__inode_sub_bytes(inode, bytes);
831 	spin_unlock(&inode->i_lock);
832 }
833 
834 EXPORT_SYMBOL(inode_sub_bytes);
835 
836 loff_t inode_get_bytes(struct inode *inode)
837 {
838 	loff_t ret;
839 
840 	spin_lock(&inode->i_lock);
841 	ret = __inode_get_bytes(inode);
842 	spin_unlock(&inode->i_lock);
843 	return ret;
844 }
845 
846 EXPORT_SYMBOL(inode_get_bytes);
847 
848 void inode_set_bytes(struct inode *inode, loff_t bytes)
849 {
850 	/* Caller is here responsible for sufficient locking
851 	 * (ie. inode->i_lock) */
852 	inode->i_blocks = bytes >> 9;
853 	inode->i_bytes = bytes & 511;
854 }
855 
856 EXPORT_SYMBOL(inode_set_bytes);
857