1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/export.h>
10 #include <linux/errno.h>
11 #include <linux/file.h>
12 #include <linux/highuid.h>
14 #include <linux/namei.h>
15 #include <linux/security.h>
16 #include <linux/cred.h>
17 #include <linux/syscalls.h>
18 #include <linux/pagemap.h>
19 #include <linux/compat.h>
21 #include <linux/uaccess.h>
22 #include <asm/unistd.h>
28 * generic_fillattr - Fill in the basic attributes from the inode struct
29 * @inode: Inode to use as the source
30 * @stat: Where to fill in the attributes
32 * Fill in the basic attributes in the kstat structure from data that's to be
33 * found on the VFS inode structure. This is the default if no getattr inode
34 * operation is supplied.
36 void generic_fillattr(struct inode *inode, struct kstat *stat)
38 stat->dev = inode->i_sb->s_dev;
39 stat->ino = inode->i_ino;
40 stat->mode = inode->i_mode;
41 stat->nlink = inode->i_nlink;
42 stat->uid = inode->i_uid;
43 stat->gid = inode->i_gid;
44 stat->rdev = inode->i_rdev;
45 stat->size = i_size_read(inode);
46 stat->atime = inode->i_atime;
47 stat->mtime = inode->i_mtime;
48 stat->ctime = inode->i_ctime;
49 stat->blksize = i_blocksize(inode);
50 stat->blocks = inode->i_blocks;
52 EXPORT_SYMBOL(generic_fillattr);
55 * vfs_getattr_nosec - getattr without security checks
56 * @path: file to get attributes from
57 * @stat: structure to return attributes in
58 * @request_mask: STATX_xxx flags indicating what the caller wants
59 * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
61 * Get attributes without calling security_inode_getattr.
63 * Currently the only caller other than vfs_getattr is internal to the
64 * filehandle lookup code, which uses only the inode number and returns no
65 * attributes to any user. Any other code probably wants vfs_getattr.
67 int vfs_getattr_nosec(const struct path *path, struct kstat *stat,
68 u32 request_mask, unsigned int query_flags)
70 struct inode *inode = d_backing_inode(path->dentry);
72 memset(stat, 0, sizeof(*stat));
73 stat->result_mask |= STATX_BASIC_STATS;
74 query_flags &= AT_STATX_SYNC_TYPE;
76 /* allow the fs to override these if it really wants to */
77 /* SB_NOATIME means filesystem supplies dummy atime value */
78 if (inode->i_sb->s_flags & SB_NOATIME)
79 stat->result_mask &= ~STATX_ATIME;
82 * Note: If you add another clause to set an attribute flag, please
83 * update attributes_mask below.
85 if (IS_AUTOMOUNT(inode))
86 stat->attributes |= STATX_ATTR_AUTOMOUNT;
89 stat->attributes |= STATX_ATTR_DAX;
91 stat->attributes_mask |= (STATX_ATTR_AUTOMOUNT |
94 if (inode->i_op->getattr)
95 return inode->i_op->getattr(path, stat, request_mask,
98 generic_fillattr(inode, stat);
101 EXPORT_SYMBOL(vfs_getattr_nosec);
104 * vfs_getattr - Get the enhanced basic attributes of a file
105 * @path: The file of interest
106 * @stat: Where to return the statistics
107 * @request_mask: STATX_xxx flags indicating what the caller wants
108 * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
110 * Ask the filesystem for a file's attributes. The caller must indicate in
111 * request_mask and query_flags to indicate what they want.
113 * If the file is remote, the filesystem can be forced to update the attributes
114 * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
115 * suppress the update by passing AT_STATX_DONT_SYNC.
117 * Bits must have been set in request_mask to indicate which attributes the
118 * caller wants retrieving. Any such attribute not requested may be returned
119 * anyway, but the value may be approximate, and, if remote, may not have been
120 * synchronised with the server.
122 * 0 will be returned on success, and a -ve error code if unsuccessful.
124 int vfs_getattr(const struct path *path, struct kstat *stat,
125 u32 request_mask, unsigned int query_flags)
129 retval = security_inode_getattr(path);
132 return vfs_getattr_nosec(path, stat, request_mask, query_flags);
134 EXPORT_SYMBOL(vfs_getattr);
137 * vfs_fstat - Get the basic attributes by file descriptor
138 * @fd: The file descriptor referring to the file of interest
139 * @stat: The result structure to fill in.
141 * This function is a wrapper around vfs_getattr(). The main difference is
142 * that it uses a file descriptor to determine the file location.
144 * 0 will be returned on success, and a -ve error code if unsuccessful.
146 int vfs_fstat(int fd, struct kstat *stat)
154 error = vfs_getattr(&f.file->f_path, stat, STATX_BASIC_STATS, 0);
160 * vfs_statx - Get basic and extra attributes by filename
161 * @dfd: A file descriptor representing the base dir for a relative filename
162 * @filename: The name of the file of interest
163 * @flags: Flags to control the query
164 * @stat: The result structure to fill in.
165 * @request_mask: STATX_xxx flags indicating what the caller wants
167 * This function is a wrapper around vfs_getattr(). The main difference is
168 * that it uses a filename and base directory to determine the file location.
169 * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
170 * at the given name from being referenced.
172 * 0 will be returned on success, and a -ve error code if unsuccessful.
174 static int vfs_statx(int dfd, const char __user *filename, int flags,
175 struct kstat *stat, u32 request_mask)
178 unsigned lookup_flags = 0;
181 if (flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT | AT_EMPTY_PATH |
185 if (!(flags & AT_SYMLINK_NOFOLLOW))
186 lookup_flags |= LOOKUP_FOLLOW;
187 if (!(flags & AT_NO_AUTOMOUNT))
188 lookup_flags |= LOOKUP_AUTOMOUNT;
189 if (flags & AT_EMPTY_PATH)
190 lookup_flags |= LOOKUP_EMPTY;
193 error = user_path_at(dfd, filename, lookup_flags, &path);
197 error = vfs_getattr(&path, stat, request_mask, flags);
198 stat->mnt_id = real_mount(path.mnt)->mnt_id;
199 stat->result_mask |= STATX_MNT_ID;
200 if (path.mnt->mnt_root == path.dentry)
201 stat->attributes |= STATX_ATTR_MOUNT_ROOT;
202 stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
204 if (retry_estale(error, lookup_flags)) {
205 lookup_flags |= LOOKUP_REVAL;
212 int vfs_fstatat(int dfd, const char __user *filename,
213 struct kstat *stat, int flags)
215 return vfs_statx(dfd, filename, flags | AT_NO_AUTOMOUNT,
216 stat, STATX_BASIC_STATS);
219 #ifdef __ARCH_WANT_OLD_STAT
222 * For backward compatibility? Maybe this should be moved
223 * into arch/i386 instead?
225 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
227 static int warncount = 5;
228 struct __old_kernel_stat tmp;
232 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
234 } else if (warncount < 0) {
235 /* it's laughable, but... */
239 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
240 tmp.st_dev = old_encode_dev(stat->dev);
241 tmp.st_ino = stat->ino;
242 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
244 tmp.st_mode = stat->mode;
245 tmp.st_nlink = stat->nlink;
246 if (tmp.st_nlink != stat->nlink)
248 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
249 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
250 tmp.st_rdev = old_encode_dev(stat->rdev);
251 #if BITS_PER_LONG == 32
252 if (stat->size > MAX_NON_LFS)
255 tmp.st_size = stat->size;
256 tmp.st_atime = stat->atime.tv_sec;
257 tmp.st_mtime = stat->mtime.tv_sec;
258 tmp.st_ctime = stat->ctime.tv_sec;
259 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
262 SYSCALL_DEFINE2(stat, const char __user *, filename,
263 struct __old_kernel_stat __user *, statbuf)
268 error = vfs_stat(filename, &stat);
272 return cp_old_stat(&stat, statbuf);
275 SYSCALL_DEFINE2(lstat, const char __user *, filename,
276 struct __old_kernel_stat __user *, statbuf)
281 error = vfs_lstat(filename, &stat);
285 return cp_old_stat(&stat, statbuf);
288 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
291 int error = vfs_fstat(fd, &stat);
294 error = cp_old_stat(&stat, statbuf);
299 #endif /* __ARCH_WANT_OLD_STAT */
301 #ifdef __ARCH_WANT_NEW_STAT
303 #if BITS_PER_LONG == 32
304 # define choose_32_64(a,b) a
306 # define choose_32_64(a,b) b
309 #ifndef INIT_STRUCT_STAT_PADDING
310 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
313 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
317 if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
319 if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
321 #if BITS_PER_LONG == 32
322 if (stat->size > MAX_NON_LFS)
326 INIT_STRUCT_STAT_PADDING(tmp);
327 tmp.st_dev = new_encode_dev(stat->dev);
328 tmp.st_ino = stat->ino;
329 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
331 tmp.st_mode = stat->mode;
332 tmp.st_nlink = stat->nlink;
333 if (tmp.st_nlink != stat->nlink)
335 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
336 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
337 tmp.st_rdev = new_encode_dev(stat->rdev);
338 tmp.st_size = stat->size;
339 tmp.st_atime = stat->atime.tv_sec;
340 tmp.st_mtime = stat->mtime.tv_sec;
341 tmp.st_ctime = stat->ctime.tv_sec;
342 #ifdef STAT_HAVE_NSEC
343 tmp.st_atime_nsec = stat->atime.tv_nsec;
344 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
345 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
347 tmp.st_blocks = stat->blocks;
348 tmp.st_blksize = stat->blksize;
349 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
352 SYSCALL_DEFINE2(newstat, const char __user *, filename,
353 struct stat __user *, statbuf)
356 int error = vfs_stat(filename, &stat);
360 return cp_new_stat(&stat, statbuf);
363 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
364 struct stat __user *, statbuf)
369 error = vfs_lstat(filename, &stat);
373 return cp_new_stat(&stat, statbuf);
376 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
377 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
378 struct stat __user *, statbuf, int, flag)
383 error = vfs_fstatat(dfd, filename, &stat, flag);
386 return cp_new_stat(&stat, statbuf);
390 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
393 int error = vfs_fstat(fd, &stat);
396 error = cp_new_stat(&stat, statbuf);
402 static int do_readlinkat(int dfd, const char __user *pathname,
403 char __user *buf, int bufsiz)
408 unsigned int lookup_flags = LOOKUP_EMPTY;
414 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
416 struct inode *inode = d_backing_inode(path.dentry);
418 error = empty ? -ENOENT : -EINVAL;
420 * AFS mountpoints allow readlink(2) but are not symlinks
422 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
423 error = security_inode_readlink(path.dentry);
426 error = vfs_readlink(path.dentry, buf, bufsiz);
430 if (retry_estale(error, lookup_flags)) {
431 lookup_flags |= LOOKUP_REVAL;
438 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
439 char __user *, buf, int, bufsiz)
441 return do_readlinkat(dfd, pathname, buf, bufsiz);
444 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
447 return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
451 /* ---------- LFS-64 ----------- */
452 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
454 #ifndef INIT_STRUCT_STAT64_PADDING
455 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
458 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
462 INIT_STRUCT_STAT64_PADDING(tmp);
464 /* mips has weird padding, so we don't get 64 bits there */
465 tmp.st_dev = new_encode_dev(stat->dev);
466 tmp.st_rdev = new_encode_dev(stat->rdev);
468 tmp.st_dev = huge_encode_dev(stat->dev);
469 tmp.st_rdev = huge_encode_dev(stat->rdev);
471 tmp.st_ino = stat->ino;
472 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
474 #ifdef STAT64_HAS_BROKEN_ST_INO
475 tmp.__st_ino = stat->ino;
477 tmp.st_mode = stat->mode;
478 tmp.st_nlink = stat->nlink;
479 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
480 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
481 tmp.st_atime = stat->atime.tv_sec;
482 tmp.st_atime_nsec = stat->atime.tv_nsec;
483 tmp.st_mtime = stat->mtime.tv_sec;
484 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
485 tmp.st_ctime = stat->ctime.tv_sec;
486 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
487 tmp.st_size = stat->size;
488 tmp.st_blocks = stat->blocks;
489 tmp.st_blksize = stat->blksize;
490 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
493 SYSCALL_DEFINE2(stat64, const char __user *, filename,
494 struct stat64 __user *, statbuf)
497 int error = vfs_stat(filename, &stat);
500 error = cp_new_stat64(&stat, statbuf);
505 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
506 struct stat64 __user *, statbuf)
509 int error = vfs_lstat(filename, &stat);
512 error = cp_new_stat64(&stat, statbuf);
517 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
520 int error = vfs_fstat(fd, &stat);
523 error = cp_new_stat64(&stat, statbuf);
528 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
529 struct stat64 __user *, statbuf, int, flag)
534 error = vfs_fstatat(dfd, filename, &stat, flag);
537 return cp_new_stat64(&stat, statbuf);
539 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
541 static noinline_for_stack int
542 cp_statx(const struct kstat *stat, struct statx __user *buffer)
546 memset(&tmp, 0, sizeof(tmp));
548 tmp.stx_mask = stat->result_mask;
549 tmp.stx_blksize = stat->blksize;
550 tmp.stx_attributes = stat->attributes;
551 tmp.stx_nlink = stat->nlink;
552 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
553 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
554 tmp.stx_mode = stat->mode;
555 tmp.stx_ino = stat->ino;
556 tmp.stx_size = stat->size;
557 tmp.stx_blocks = stat->blocks;
558 tmp.stx_attributes_mask = stat->attributes_mask;
559 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
560 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
561 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
562 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
563 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
564 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
565 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
566 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
567 tmp.stx_rdev_major = MAJOR(stat->rdev);
568 tmp.stx_rdev_minor = MINOR(stat->rdev);
569 tmp.stx_dev_major = MAJOR(stat->dev);
570 tmp.stx_dev_minor = MINOR(stat->dev);
571 tmp.stx_mnt_id = stat->mnt_id;
573 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
576 int do_statx(int dfd, const char __user *filename, unsigned flags,
577 unsigned int mask, struct statx __user *buffer)
582 if (mask & STATX__RESERVED)
584 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
587 error = vfs_statx(dfd, filename, flags, &stat, mask);
591 return cp_statx(&stat, buffer);
595 * sys_statx - System call to get enhanced stats
596 * @dfd: Base directory to pathwalk from *or* fd to stat.
597 * @filename: File to stat or "" with AT_EMPTY_PATH
598 * @flags: AT_* flags to control pathwalk.
599 * @mask: Parts of statx struct actually required.
600 * @buffer: Result buffer.
602 * Note that fstat() can be emulated by setting dfd to the fd of interest,
603 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
605 SYSCALL_DEFINE5(statx,
606 int, dfd, const char __user *, filename, unsigned, flags,
608 struct statx __user *, buffer)
610 return do_statx(dfd, filename, flags, mask, buffer);
614 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
616 struct compat_stat tmp;
618 if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
620 if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
623 memset(&tmp, 0, sizeof(tmp));
624 tmp.st_dev = new_encode_dev(stat->dev);
625 tmp.st_ino = stat->ino;
626 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
628 tmp.st_mode = stat->mode;
629 tmp.st_nlink = stat->nlink;
630 if (tmp.st_nlink != stat->nlink)
632 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
633 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
634 tmp.st_rdev = new_encode_dev(stat->rdev);
635 if ((u64) stat->size > MAX_NON_LFS)
637 tmp.st_size = stat->size;
638 tmp.st_atime = stat->atime.tv_sec;
639 tmp.st_atime_nsec = stat->atime.tv_nsec;
640 tmp.st_mtime = stat->mtime.tv_sec;
641 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
642 tmp.st_ctime = stat->ctime.tv_sec;
643 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
644 tmp.st_blocks = stat->blocks;
645 tmp.st_blksize = stat->blksize;
646 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
649 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
650 struct compat_stat __user *, statbuf)
655 error = vfs_stat(filename, &stat);
658 return cp_compat_stat(&stat, statbuf);
661 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
662 struct compat_stat __user *, statbuf)
667 error = vfs_lstat(filename, &stat);
670 return cp_compat_stat(&stat, statbuf);
673 #ifndef __ARCH_WANT_STAT64
674 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
675 const char __user *, filename,
676 struct compat_stat __user *, statbuf, int, flag)
681 error = vfs_fstatat(dfd, filename, &stat, flag);
684 return cp_compat_stat(&stat, statbuf);
688 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
689 struct compat_stat __user *, statbuf)
692 int error = vfs_fstat(fd, &stat);
695 error = cp_compat_stat(&stat, statbuf);
700 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
701 void __inode_add_bytes(struct inode *inode, loff_t bytes)
703 inode->i_blocks += bytes >> 9;
705 inode->i_bytes += bytes;
706 if (inode->i_bytes >= 512) {
708 inode->i_bytes -= 512;
711 EXPORT_SYMBOL(__inode_add_bytes);
713 void inode_add_bytes(struct inode *inode, loff_t bytes)
715 spin_lock(&inode->i_lock);
716 __inode_add_bytes(inode, bytes);
717 spin_unlock(&inode->i_lock);
720 EXPORT_SYMBOL(inode_add_bytes);
722 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
724 inode->i_blocks -= bytes >> 9;
726 if (inode->i_bytes < bytes) {
728 inode->i_bytes += 512;
730 inode->i_bytes -= bytes;
733 EXPORT_SYMBOL(__inode_sub_bytes);
735 void inode_sub_bytes(struct inode *inode, loff_t bytes)
737 spin_lock(&inode->i_lock);
738 __inode_sub_bytes(inode, bytes);
739 spin_unlock(&inode->i_lock);
742 EXPORT_SYMBOL(inode_sub_bytes);
744 loff_t inode_get_bytes(struct inode *inode)
748 spin_lock(&inode->i_lock);
749 ret = __inode_get_bytes(inode);
750 spin_unlock(&inode->i_lock);
754 EXPORT_SYMBOL(inode_get_bytes);
756 void inode_set_bytes(struct inode *inode, loff_t bytes)
758 /* Caller is here responsible for sufficient locking
759 * (ie. inode->i_lock) */
760 inode->i_blocks = bytes >> 9;
761 inode->i_bytes = bytes & 511;
764 EXPORT_SYMBOL(inode_set_bytes);