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 * @mnt_userns: user namespace of the mount the inode was found from
30 * @inode: Inode to use as the source
31 * @stat: Where to fill in the attributes
33 * Fill in the basic attributes in the kstat structure from data that's to be
34 * found on the VFS inode structure. This is the default if no getattr inode
35 * operation is supplied.
37 * If the inode has been found through an idmapped mount the user namespace of
38 * the vfsmount must be passed through @mnt_userns. This function will then
39 * take care to map the inode according to @mnt_userns before filling in the
40 * uid and gid filds. On non-idmapped mounts or if permission checking is to be
41 * performed on the raw inode simply passs init_user_ns.
43 void generic_fillattr(struct user_namespace *mnt_userns, struct inode *inode,
46 stat->dev = inode->i_sb->s_dev;
47 stat->ino = inode->i_ino;
48 stat->mode = inode->i_mode;
49 stat->nlink = inode->i_nlink;
50 stat->uid = i_uid_into_mnt(mnt_userns, inode);
51 stat->gid = i_gid_into_mnt(mnt_userns, inode);
52 stat->rdev = inode->i_rdev;
53 stat->size = i_size_read(inode);
54 stat->atime = inode->i_atime;
55 stat->mtime = inode->i_mtime;
56 stat->ctime = inode->i_ctime;
57 stat->blksize = i_blocksize(inode);
58 stat->blocks = inode->i_blocks;
60 EXPORT_SYMBOL(generic_fillattr);
63 * vfs_getattr_nosec - getattr without security checks
64 * @path: file to get attributes from
65 * @stat: structure to return attributes in
66 * @request_mask: STATX_xxx flags indicating what the caller wants
67 * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
69 * Get attributes without calling security_inode_getattr.
71 * Currently the only caller other than vfs_getattr is internal to the
72 * filehandle lookup code, which uses only the inode number and returns no
73 * attributes to any user. Any other code probably wants vfs_getattr.
75 int vfs_getattr_nosec(const struct path *path, struct kstat *stat,
76 u32 request_mask, unsigned int query_flags)
78 struct user_namespace *mnt_userns;
79 struct inode *inode = d_backing_inode(path->dentry);
81 memset(stat, 0, sizeof(*stat));
82 stat->result_mask |= STATX_BASIC_STATS;
83 query_flags &= AT_STATX_SYNC_TYPE;
85 /* allow the fs to override these if it really wants to */
86 /* SB_NOATIME means filesystem supplies dummy atime value */
87 if (inode->i_sb->s_flags & SB_NOATIME)
88 stat->result_mask &= ~STATX_ATIME;
91 * Note: If you add another clause to set an attribute flag, please
92 * update attributes_mask below.
94 if (IS_AUTOMOUNT(inode))
95 stat->attributes |= STATX_ATTR_AUTOMOUNT;
98 stat->attributes |= STATX_ATTR_DAX;
100 stat->attributes_mask |= (STATX_ATTR_AUTOMOUNT |
103 mnt_userns = mnt_user_ns(path->mnt);
104 if (inode->i_op->getattr)
105 return inode->i_op->getattr(mnt_userns, path, stat,
106 request_mask, query_flags);
108 generic_fillattr(mnt_userns, inode, stat);
111 EXPORT_SYMBOL(vfs_getattr_nosec);
114 * vfs_getattr - Get the enhanced basic attributes of a file
115 * @path: The file of interest
116 * @stat: Where to return the statistics
117 * @request_mask: STATX_xxx flags indicating what the caller wants
118 * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
120 * Ask the filesystem for a file's attributes. The caller must indicate in
121 * request_mask and query_flags to indicate what they want.
123 * If the file is remote, the filesystem can be forced to update the attributes
124 * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
125 * suppress the update by passing AT_STATX_DONT_SYNC.
127 * Bits must have been set in request_mask to indicate which attributes the
128 * caller wants retrieving. Any such attribute not requested may be returned
129 * anyway, but the value may be approximate, and, if remote, may not have been
130 * synchronised with the server.
132 * 0 will be returned on success, and a -ve error code if unsuccessful.
134 int vfs_getattr(const struct path *path, struct kstat *stat,
135 u32 request_mask, unsigned int query_flags)
139 retval = security_inode_getattr(path);
142 return vfs_getattr_nosec(path, stat, request_mask, query_flags);
144 EXPORT_SYMBOL(vfs_getattr);
147 * vfs_fstat - Get the basic attributes by file descriptor
148 * @fd: The file descriptor referring to the file of interest
149 * @stat: The result structure to fill in.
151 * This function is a wrapper around vfs_getattr(). The main difference is
152 * that it uses a file descriptor to determine the file location.
154 * 0 will be returned on success, and a -ve error code if unsuccessful.
156 int vfs_fstat(int fd, struct kstat *stat)
164 error = vfs_getattr(&f.file->f_path, stat, STATX_BASIC_STATS, 0);
170 * vfs_statx - Get basic and extra attributes by filename
171 * @dfd: A file descriptor representing the base dir for a relative filename
172 * @filename: The name of the file of interest
173 * @flags: Flags to control the query
174 * @stat: The result structure to fill in.
175 * @request_mask: STATX_xxx flags indicating what the caller wants
177 * This function is a wrapper around vfs_getattr(). The main difference is
178 * that it uses a filename and base directory to determine the file location.
179 * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
180 * at the given name from being referenced.
182 * 0 will be returned on success, and a -ve error code if unsuccessful.
184 static int vfs_statx(int dfd, const char __user *filename, int flags,
185 struct kstat *stat, u32 request_mask)
188 unsigned lookup_flags = 0;
191 if (flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT | AT_EMPTY_PATH |
195 if (!(flags & AT_SYMLINK_NOFOLLOW))
196 lookup_flags |= LOOKUP_FOLLOW;
197 if (!(flags & AT_NO_AUTOMOUNT))
198 lookup_flags |= LOOKUP_AUTOMOUNT;
199 if (flags & AT_EMPTY_PATH)
200 lookup_flags |= LOOKUP_EMPTY;
203 error = user_path_at(dfd, filename, lookup_flags, &path);
207 error = vfs_getattr(&path, stat, request_mask, flags);
208 stat->mnt_id = real_mount(path.mnt)->mnt_id;
209 stat->result_mask |= STATX_MNT_ID;
210 if (path.mnt->mnt_root == path.dentry)
211 stat->attributes |= STATX_ATTR_MOUNT_ROOT;
212 stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
214 if (retry_estale(error, lookup_flags)) {
215 lookup_flags |= LOOKUP_REVAL;
222 int vfs_fstatat(int dfd, const char __user *filename,
223 struct kstat *stat, int flags)
225 return vfs_statx(dfd, filename, flags | AT_NO_AUTOMOUNT,
226 stat, STATX_BASIC_STATS);
229 #ifdef __ARCH_WANT_OLD_STAT
232 * For backward compatibility? Maybe this should be moved
233 * into arch/i386 instead?
235 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
237 static int warncount = 5;
238 struct __old_kernel_stat tmp;
242 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
244 } else if (warncount < 0) {
245 /* it's laughable, but... */
249 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
250 tmp.st_dev = old_encode_dev(stat->dev);
251 tmp.st_ino = stat->ino;
252 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
254 tmp.st_mode = stat->mode;
255 tmp.st_nlink = stat->nlink;
256 if (tmp.st_nlink != stat->nlink)
258 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
259 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
260 tmp.st_rdev = old_encode_dev(stat->rdev);
261 #if BITS_PER_LONG == 32
262 if (stat->size > MAX_NON_LFS)
265 tmp.st_size = stat->size;
266 tmp.st_atime = stat->atime.tv_sec;
267 tmp.st_mtime = stat->mtime.tv_sec;
268 tmp.st_ctime = stat->ctime.tv_sec;
269 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
272 SYSCALL_DEFINE2(stat, const char __user *, filename,
273 struct __old_kernel_stat __user *, statbuf)
278 error = vfs_stat(filename, &stat);
282 return cp_old_stat(&stat, statbuf);
285 SYSCALL_DEFINE2(lstat, const char __user *, filename,
286 struct __old_kernel_stat __user *, statbuf)
291 error = vfs_lstat(filename, &stat);
295 return cp_old_stat(&stat, statbuf);
298 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
301 int error = vfs_fstat(fd, &stat);
304 error = cp_old_stat(&stat, statbuf);
309 #endif /* __ARCH_WANT_OLD_STAT */
311 #ifdef __ARCH_WANT_NEW_STAT
313 #if BITS_PER_LONG == 32
314 # define choose_32_64(a,b) a
316 # define choose_32_64(a,b) b
319 #define valid_dev(x) choose_32_64(old_valid_dev(x),true)
320 #define encode_dev(x) choose_32_64(old_encode_dev,new_encode_dev)(x)
322 #ifndef INIT_STRUCT_STAT_PADDING
323 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
326 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
330 if (!valid_dev(stat->dev) || !valid_dev(stat->rdev))
332 #if BITS_PER_LONG == 32
333 if (stat->size > MAX_NON_LFS)
337 INIT_STRUCT_STAT_PADDING(tmp);
338 tmp.st_dev = encode_dev(stat->dev);
339 tmp.st_ino = stat->ino;
340 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
342 tmp.st_mode = stat->mode;
343 tmp.st_nlink = stat->nlink;
344 if (tmp.st_nlink != stat->nlink)
346 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
347 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
348 tmp.st_rdev = encode_dev(stat->rdev);
349 tmp.st_size = stat->size;
350 tmp.st_atime = stat->atime.tv_sec;
351 tmp.st_mtime = stat->mtime.tv_sec;
352 tmp.st_ctime = stat->ctime.tv_sec;
353 #ifdef STAT_HAVE_NSEC
354 tmp.st_atime_nsec = stat->atime.tv_nsec;
355 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
356 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
358 tmp.st_blocks = stat->blocks;
359 tmp.st_blksize = stat->blksize;
360 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
363 SYSCALL_DEFINE2(newstat, const char __user *, filename,
364 struct stat __user *, statbuf)
367 int error = vfs_stat(filename, &stat);
371 return cp_new_stat(&stat, statbuf);
374 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
375 struct stat __user *, statbuf)
380 error = vfs_lstat(filename, &stat);
384 return cp_new_stat(&stat, statbuf);
387 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
388 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
389 struct stat __user *, statbuf, int, flag)
394 error = vfs_fstatat(dfd, filename, &stat, flag);
397 return cp_new_stat(&stat, statbuf);
401 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
404 int error = vfs_fstat(fd, &stat);
407 error = cp_new_stat(&stat, statbuf);
413 static int do_readlinkat(int dfd, const char __user *pathname,
414 char __user *buf, int bufsiz)
419 unsigned int lookup_flags = LOOKUP_EMPTY;
425 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
427 struct inode *inode = d_backing_inode(path.dentry);
429 error = empty ? -ENOENT : -EINVAL;
431 * AFS mountpoints allow readlink(2) but are not symlinks
433 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
434 error = security_inode_readlink(path.dentry);
437 error = vfs_readlink(path.dentry, buf, bufsiz);
441 if (retry_estale(error, lookup_flags)) {
442 lookup_flags |= LOOKUP_REVAL;
449 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
450 char __user *, buf, int, bufsiz)
452 return do_readlinkat(dfd, pathname, buf, bufsiz);
455 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
458 return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
462 /* ---------- LFS-64 ----------- */
463 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
465 #ifndef INIT_STRUCT_STAT64_PADDING
466 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
469 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
473 INIT_STRUCT_STAT64_PADDING(tmp);
475 /* mips has weird padding, so we don't get 64 bits there */
476 tmp.st_dev = new_encode_dev(stat->dev);
477 tmp.st_rdev = new_encode_dev(stat->rdev);
479 tmp.st_dev = huge_encode_dev(stat->dev);
480 tmp.st_rdev = huge_encode_dev(stat->rdev);
482 tmp.st_ino = stat->ino;
483 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
485 #ifdef STAT64_HAS_BROKEN_ST_INO
486 tmp.__st_ino = stat->ino;
488 tmp.st_mode = stat->mode;
489 tmp.st_nlink = stat->nlink;
490 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
491 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
492 tmp.st_atime = stat->atime.tv_sec;
493 tmp.st_atime_nsec = stat->atime.tv_nsec;
494 tmp.st_mtime = stat->mtime.tv_sec;
495 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
496 tmp.st_ctime = stat->ctime.tv_sec;
497 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
498 tmp.st_size = stat->size;
499 tmp.st_blocks = stat->blocks;
500 tmp.st_blksize = stat->blksize;
501 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
504 SYSCALL_DEFINE2(stat64, const char __user *, filename,
505 struct stat64 __user *, statbuf)
508 int error = vfs_stat(filename, &stat);
511 error = cp_new_stat64(&stat, statbuf);
516 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
517 struct stat64 __user *, statbuf)
520 int error = vfs_lstat(filename, &stat);
523 error = cp_new_stat64(&stat, statbuf);
528 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
531 int error = vfs_fstat(fd, &stat);
534 error = cp_new_stat64(&stat, statbuf);
539 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
540 struct stat64 __user *, statbuf, int, flag)
545 error = vfs_fstatat(dfd, filename, &stat, flag);
548 return cp_new_stat64(&stat, statbuf);
550 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
552 static noinline_for_stack int
553 cp_statx(const struct kstat *stat, struct statx __user *buffer)
557 memset(&tmp, 0, sizeof(tmp));
559 tmp.stx_mask = stat->result_mask;
560 tmp.stx_blksize = stat->blksize;
561 tmp.stx_attributes = stat->attributes;
562 tmp.stx_nlink = stat->nlink;
563 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
564 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
565 tmp.stx_mode = stat->mode;
566 tmp.stx_ino = stat->ino;
567 tmp.stx_size = stat->size;
568 tmp.stx_blocks = stat->blocks;
569 tmp.stx_attributes_mask = stat->attributes_mask;
570 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
571 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
572 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
573 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
574 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
575 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
576 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
577 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
578 tmp.stx_rdev_major = MAJOR(stat->rdev);
579 tmp.stx_rdev_minor = MINOR(stat->rdev);
580 tmp.stx_dev_major = MAJOR(stat->dev);
581 tmp.stx_dev_minor = MINOR(stat->dev);
582 tmp.stx_mnt_id = stat->mnt_id;
584 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
587 int do_statx(int dfd, const char __user *filename, unsigned flags,
588 unsigned int mask, struct statx __user *buffer)
593 if (mask & STATX__RESERVED)
595 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
598 error = vfs_statx(dfd, filename, flags, &stat, mask);
602 return cp_statx(&stat, buffer);
606 * sys_statx - System call to get enhanced stats
607 * @dfd: Base directory to pathwalk from *or* fd to stat.
608 * @filename: File to stat or "" with AT_EMPTY_PATH
609 * @flags: AT_* flags to control pathwalk.
610 * @mask: Parts of statx struct actually required.
611 * @buffer: Result buffer.
613 * Note that fstat() can be emulated by setting dfd to the fd of interest,
614 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
616 SYSCALL_DEFINE5(statx,
617 int, dfd, const char __user *, filename, unsigned, flags,
619 struct statx __user *, buffer)
621 return do_statx(dfd, filename, flags, mask, buffer);
625 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
627 struct compat_stat tmp;
629 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
632 memset(&tmp, 0, sizeof(tmp));
633 tmp.st_dev = old_encode_dev(stat->dev);
634 tmp.st_ino = stat->ino;
635 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
637 tmp.st_mode = stat->mode;
638 tmp.st_nlink = stat->nlink;
639 if (tmp.st_nlink != stat->nlink)
641 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
642 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
643 tmp.st_rdev = old_encode_dev(stat->rdev);
644 if ((u64) stat->size > MAX_NON_LFS)
646 tmp.st_size = stat->size;
647 tmp.st_atime = stat->atime.tv_sec;
648 tmp.st_atime_nsec = stat->atime.tv_nsec;
649 tmp.st_mtime = stat->mtime.tv_sec;
650 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
651 tmp.st_ctime = stat->ctime.tv_sec;
652 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
653 tmp.st_blocks = stat->blocks;
654 tmp.st_blksize = stat->blksize;
655 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
658 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
659 struct compat_stat __user *, statbuf)
664 error = vfs_stat(filename, &stat);
667 return cp_compat_stat(&stat, statbuf);
670 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
671 struct compat_stat __user *, statbuf)
676 error = vfs_lstat(filename, &stat);
679 return cp_compat_stat(&stat, statbuf);
682 #ifndef __ARCH_WANT_STAT64
683 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
684 const char __user *, filename,
685 struct compat_stat __user *, statbuf, int, flag)
690 error = vfs_fstatat(dfd, filename, &stat, flag);
693 return cp_compat_stat(&stat, statbuf);
697 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
698 struct compat_stat __user *, statbuf)
701 int error = vfs_fstat(fd, &stat);
704 error = cp_compat_stat(&stat, statbuf);
709 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
710 void __inode_add_bytes(struct inode *inode, loff_t bytes)
712 inode->i_blocks += bytes >> 9;
714 inode->i_bytes += bytes;
715 if (inode->i_bytes >= 512) {
717 inode->i_bytes -= 512;
720 EXPORT_SYMBOL(__inode_add_bytes);
722 void inode_add_bytes(struct inode *inode, loff_t bytes)
724 spin_lock(&inode->i_lock);
725 __inode_add_bytes(inode, bytes);
726 spin_unlock(&inode->i_lock);
729 EXPORT_SYMBOL(inode_add_bytes);
731 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
733 inode->i_blocks -= bytes >> 9;
735 if (inode->i_bytes < bytes) {
737 inode->i_bytes += 512;
739 inode->i_bytes -= bytes;
742 EXPORT_SYMBOL(__inode_sub_bytes);
744 void inode_sub_bytes(struct inode *inode, loff_t bytes)
746 spin_lock(&inode->i_lock);
747 __inode_sub_bytes(inode, bytes);
748 spin_unlock(&inode->i_lock);
751 EXPORT_SYMBOL(inode_sub_bytes);
753 loff_t inode_get_bytes(struct inode *inode)
757 spin_lock(&inode->i_lock);
758 ret = __inode_get_bytes(inode);
759 spin_unlock(&inode->i_lock);
763 EXPORT_SYMBOL(inode_get_bytes);
765 void inode_set_bytes(struct inode *inode, loff_t bytes)
767 /* Caller is here responsible for sufficient locking
768 * (ie. inode->i_lock) */
769 inode->i_blocks = bytes >> 9;
770 inode->i_bytes = bytes & 511;
773 EXPORT_SYMBOL(inode_set_bytes);