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;
89 if (IS_AUTOMOUNT(inode))
90 stat->attributes |= STATX_ATTR_AUTOMOUNT;
93 stat->attributes |= STATX_ATTR_DAX;
95 mnt_userns = mnt_user_ns(path->mnt);
96 if (inode->i_op->getattr)
97 return inode->i_op->getattr(mnt_userns, path, stat,
98 request_mask, query_flags);
100 generic_fillattr(mnt_userns, inode, stat);
103 EXPORT_SYMBOL(vfs_getattr_nosec);
106 * vfs_getattr - Get the enhanced basic attributes of a file
107 * @path: The file of interest
108 * @stat: Where to return the statistics
109 * @request_mask: STATX_xxx flags indicating what the caller wants
110 * @query_flags: Query mode (AT_STATX_SYNC_TYPE)
112 * Ask the filesystem for a file's attributes. The caller must indicate in
113 * request_mask and query_flags to indicate what they want.
115 * If the file is remote, the filesystem can be forced to update the attributes
116 * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
117 * suppress the update by passing AT_STATX_DONT_SYNC.
119 * Bits must have been set in request_mask to indicate which attributes the
120 * caller wants retrieving. Any such attribute not requested may be returned
121 * anyway, but the value may be approximate, and, if remote, may not have been
122 * synchronised with the server.
124 * 0 will be returned on success, and a -ve error code if unsuccessful.
126 int vfs_getattr(const struct path *path, struct kstat *stat,
127 u32 request_mask, unsigned int query_flags)
131 retval = security_inode_getattr(path);
134 return vfs_getattr_nosec(path, stat, request_mask, query_flags);
136 EXPORT_SYMBOL(vfs_getattr);
139 * vfs_fstat - Get the basic attributes by file descriptor
140 * @fd: The file descriptor referring to the file of interest
141 * @stat: The result structure to fill in.
143 * This function is a wrapper around vfs_getattr(). The main difference is
144 * that it uses a file descriptor to determine the file location.
146 * 0 will be returned on success, and a -ve error code if unsuccessful.
148 int vfs_fstat(int fd, struct kstat *stat)
156 error = vfs_getattr(&f.file->f_path, stat, STATX_BASIC_STATS, 0);
162 * vfs_statx - Get basic and extra attributes by filename
163 * @dfd: A file descriptor representing the base dir for a relative filename
164 * @filename: The name of the file of interest
165 * @flags: Flags to control the query
166 * @stat: The result structure to fill in.
167 * @request_mask: STATX_xxx flags indicating what the caller wants
169 * This function is a wrapper around vfs_getattr(). The main difference is
170 * that it uses a filename and base directory to determine the file location.
171 * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
172 * at the given name from being referenced.
174 * 0 will be returned on success, and a -ve error code if unsuccessful.
176 static int vfs_statx(int dfd, const char __user *filename, int flags,
177 struct kstat *stat, u32 request_mask)
180 unsigned lookup_flags = 0;
183 if (flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT | AT_EMPTY_PATH |
187 if (!(flags & AT_SYMLINK_NOFOLLOW))
188 lookup_flags |= LOOKUP_FOLLOW;
189 if (!(flags & AT_NO_AUTOMOUNT))
190 lookup_flags |= LOOKUP_AUTOMOUNT;
191 if (flags & AT_EMPTY_PATH)
192 lookup_flags |= LOOKUP_EMPTY;
195 error = user_path_at(dfd, filename, lookup_flags, &path);
199 error = vfs_getattr(&path, stat, request_mask, flags);
200 stat->mnt_id = real_mount(path.mnt)->mnt_id;
201 stat->result_mask |= STATX_MNT_ID;
202 if (path.mnt->mnt_root == path.dentry)
203 stat->attributes |= STATX_ATTR_MOUNT_ROOT;
204 stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
206 if (retry_estale(error, lookup_flags)) {
207 lookup_flags |= LOOKUP_REVAL;
214 int vfs_fstatat(int dfd, const char __user *filename,
215 struct kstat *stat, int flags)
217 return vfs_statx(dfd, filename, flags | AT_NO_AUTOMOUNT,
218 stat, STATX_BASIC_STATS);
221 #ifdef __ARCH_WANT_OLD_STAT
224 * For backward compatibility? Maybe this should be moved
225 * into arch/i386 instead?
227 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
229 static int warncount = 5;
230 struct __old_kernel_stat tmp;
234 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
236 } else if (warncount < 0) {
237 /* it's laughable, but... */
241 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
242 tmp.st_dev = old_encode_dev(stat->dev);
243 tmp.st_ino = stat->ino;
244 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
246 tmp.st_mode = stat->mode;
247 tmp.st_nlink = stat->nlink;
248 if (tmp.st_nlink != stat->nlink)
250 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
251 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
252 tmp.st_rdev = old_encode_dev(stat->rdev);
253 #if BITS_PER_LONG == 32
254 if (stat->size > MAX_NON_LFS)
257 tmp.st_size = stat->size;
258 tmp.st_atime = stat->atime.tv_sec;
259 tmp.st_mtime = stat->mtime.tv_sec;
260 tmp.st_ctime = stat->ctime.tv_sec;
261 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
264 SYSCALL_DEFINE2(stat, const char __user *, filename,
265 struct __old_kernel_stat __user *, statbuf)
270 error = vfs_stat(filename, &stat);
274 return cp_old_stat(&stat, statbuf);
277 SYSCALL_DEFINE2(lstat, const char __user *, filename,
278 struct __old_kernel_stat __user *, statbuf)
283 error = vfs_lstat(filename, &stat);
287 return cp_old_stat(&stat, statbuf);
290 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
293 int error = vfs_fstat(fd, &stat);
296 error = cp_old_stat(&stat, statbuf);
301 #endif /* __ARCH_WANT_OLD_STAT */
303 #ifdef __ARCH_WANT_NEW_STAT
305 #if BITS_PER_LONG == 32
306 # define choose_32_64(a,b) a
308 # define choose_32_64(a,b) b
311 #define valid_dev(x) choose_32_64(old_valid_dev(x),true)
312 #define encode_dev(x) choose_32_64(old_encode_dev,new_encode_dev)(x)
314 #ifndef INIT_STRUCT_STAT_PADDING
315 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
318 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
322 if (!valid_dev(stat->dev) || !valid_dev(stat->rdev))
324 #if BITS_PER_LONG == 32
325 if (stat->size > MAX_NON_LFS)
329 INIT_STRUCT_STAT_PADDING(tmp);
330 tmp.st_dev = encode_dev(stat->dev);
331 tmp.st_ino = stat->ino;
332 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
334 tmp.st_mode = stat->mode;
335 tmp.st_nlink = stat->nlink;
336 if (tmp.st_nlink != stat->nlink)
338 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
339 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
340 tmp.st_rdev = encode_dev(stat->rdev);
341 tmp.st_size = stat->size;
342 tmp.st_atime = stat->atime.tv_sec;
343 tmp.st_mtime = stat->mtime.tv_sec;
344 tmp.st_ctime = stat->ctime.tv_sec;
345 #ifdef STAT_HAVE_NSEC
346 tmp.st_atime_nsec = stat->atime.tv_nsec;
347 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
348 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
350 tmp.st_blocks = stat->blocks;
351 tmp.st_blksize = stat->blksize;
352 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
355 SYSCALL_DEFINE2(newstat, const char __user *, filename,
356 struct stat __user *, statbuf)
359 int error = vfs_stat(filename, &stat);
363 return cp_new_stat(&stat, statbuf);
366 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
367 struct stat __user *, statbuf)
372 error = vfs_lstat(filename, &stat);
376 return cp_new_stat(&stat, statbuf);
379 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
380 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
381 struct stat __user *, statbuf, int, flag)
386 error = vfs_fstatat(dfd, filename, &stat, flag);
389 return cp_new_stat(&stat, statbuf);
393 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
396 int error = vfs_fstat(fd, &stat);
399 error = cp_new_stat(&stat, statbuf);
405 static int do_readlinkat(int dfd, const char __user *pathname,
406 char __user *buf, int bufsiz)
411 unsigned int lookup_flags = LOOKUP_EMPTY;
417 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
419 struct inode *inode = d_backing_inode(path.dentry);
421 error = empty ? -ENOENT : -EINVAL;
423 * AFS mountpoints allow readlink(2) but are not symlinks
425 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
426 error = security_inode_readlink(path.dentry);
429 error = vfs_readlink(path.dentry, buf, bufsiz);
433 if (retry_estale(error, lookup_flags)) {
434 lookup_flags |= LOOKUP_REVAL;
441 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
442 char __user *, buf, int, bufsiz)
444 return do_readlinkat(dfd, pathname, buf, bufsiz);
447 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
450 return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
454 /* ---------- LFS-64 ----------- */
455 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
457 #ifndef INIT_STRUCT_STAT64_PADDING
458 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
461 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
465 INIT_STRUCT_STAT64_PADDING(tmp);
467 /* mips has weird padding, so we don't get 64 bits there */
468 tmp.st_dev = new_encode_dev(stat->dev);
469 tmp.st_rdev = new_encode_dev(stat->rdev);
471 tmp.st_dev = huge_encode_dev(stat->dev);
472 tmp.st_rdev = huge_encode_dev(stat->rdev);
474 tmp.st_ino = stat->ino;
475 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
477 #ifdef STAT64_HAS_BROKEN_ST_INO
478 tmp.__st_ino = stat->ino;
480 tmp.st_mode = stat->mode;
481 tmp.st_nlink = stat->nlink;
482 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
483 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
484 tmp.st_atime = stat->atime.tv_sec;
485 tmp.st_atime_nsec = stat->atime.tv_nsec;
486 tmp.st_mtime = stat->mtime.tv_sec;
487 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
488 tmp.st_ctime = stat->ctime.tv_sec;
489 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
490 tmp.st_size = stat->size;
491 tmp.st_blocks = stat->blocks;
492 tmp.st_blksize = stat->blksize;
493 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
496 SYSCALL_DEFINE2(stat64, const char __user *, filename,
497 struct stat64 __user *, statbuf)
500 int error = vfs_stat(filename, &stat);
503 error = cp_new_stat64(&stat, statbuf);
508 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
509 struct stat64 __user *, statbuf)
512 int error = vfs_lstat(filename, &stat);
515 error = cp_new_stat64(&stat, statbuf);
520 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
523 int error = vfs_fstat(fd, &stat);
526 error = cp_new_stat64(&stat, statbuf);
531 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
532 struct stat64 __user *, statbuf, int, flag)
537 error = vfs_fstatat(dfd, filename, &stat, flag);
540 return cp_new_stat64(&stat, statbuf);
542 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
544 static noinline_for_stack int
545 cp_statx(const struct kstat *stat, struct statx __user *buffer)
549 memset(&tmp, 0, sizeof(tmp));
551 tmp.stx_mask = stat->result_mask;
552 tmp.stx_blksize = stat->blksize;
553 tmp.stx_attributes = stat->attributes;
554 tmp.stx_nlink = stat->nlink;
555 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
556 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
557 tmp.stx_mode = stat->mode;
558 tmp.stx_ino = stat->ino;
559 tmp.stx_size = stat->size;
560 tmp.stx_blocks = stat->blocks;
561 tmp.stx_attributes_mask = stat->attributes_mask;
562 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
563 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
564 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
565 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
566 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
567 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
568 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
569 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
570 tmp.stx_rdev_major = MAJOR(stat->rdev);
571 tmp.stx_rdev_minor = MINOR(stat->rdev);
572 tmp.stx_dev_major = MAJOR(stat->dev);
573 tmp.stx_dev_minor = MINOR(stat->dev);
574 tmp.stx_mnt_id = stat->mnt_id;
576 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
579 int do_statx(int dfd, const char __user *filename, unsigned flags,
580 unsigned int mask, struct statx __user *buffer)
585 if (mask & STATX__RESERVED)
587 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
590 error = vfs_statx(dfd, filename, flags, &stat, mask);
594 return cp_statx(&stat, buffer);
598 * sys_statx - System call to get enhanced stats
599 * @dfd: Base directory to pathwalk from *or* fd to stat.
600 * @filename: File to stat or "" with AT_EMPTY_PATH
601 * @flags: AT_* flags to control pathwalk.
602 * @mask: Parts of statx struct actually required.
603 * @buffer: Result buffer.
605 * Note that fstat() can be emulated by setting dfd to the fd of interest,
606 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
608 SYSCALL_DEFINE5(statx,
609 int, dfd, const char __user *, filename, unsigned, flags,
611 struct statx __user *, buffer)
613 return do_statx(dfd, filename, flags, mask, buffer);
617 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
619 struct compat_stat tmp;
621 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
624 memset(&tmp, 0, sizeof(tmp));
625 tmp.st_dev = old_encode_dev(stat->dev);
626 tmp.st_ino = stat->ino;
627 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
629 tmp.st_mode = stat->mode;
630 tmp.st_nlink = stat->nlink;
631 if (tmp.st_nlink != stat->nlink)
633 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
634 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
635 tmp.st_rdev = old_encode_dev(stat->rdev);
636 if ((u64) stat->size > MAX_NON_LFS)
638 tmp.st_size = stat->size;
639 tmp.st_atime = stat->atime.tv_sec;
640 tmp.st_atime_nsec = stat->atime.tv_nsec;
641 tmp.st_mtime = stat->mtime.tv_sec;
642 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
643 tmp.st_ctime = stat->ctime.tv_sec;
644 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
645 tmp.st_blocks = stat->blocks;
646 tmp.st_blksize = stat->blksize;
647 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
650 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
651 struct compat_stat __user *, statbuf)
656 error = vfs_stat(filename, &stat);
659 return cp_compat_stat(&stat, statbuf);
662 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
663 struct compat_stat __user *, statbuf)
668 error = vfs_lstat(filename, &stat);
671 return cp_compat_stat(&stat, statbuf);
674 #ifndef __ARCH_WANT_STAT64
675 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
676 const char __user *, filename,
677 struct compat_stat __user *, statbuf, int, flag)
682 error = vfs_fstatat(dfd, filename, &stat, flag);
685 return cp_compat_stat(&stat, statbuf);
689 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
690 struct compat_stat __user *, statbuf)
693 int error = vfs_fstat(fd, &stat);
696 error = cp_compat_stat(&stat, statbuf);
701 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
702 void __inode_add_bytes(struct inode *inode, loff_t bytes)
704 inode->i_blocks += bytes >> 9;
706 inode->i_bytes += bytes;
707 if (inode->i_bytes >= 512) {
709 inode->i_bytes -= 512;
712 EXPORT_SYMBOL(__inode_add_bytes);
714 void inode_add_bytes(struct inode *inode, loff_t bytes)
716 spin_lock(&inode->i_lock);
717 __inode_add_bytes(inode, bytes);
718 spin_unlock(&inode->i_lock);
721 EXPORT_SYMBOL(inode_add_bytes);
723 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
725 inode->i_blocks -= bytes >> 9;
727 if (inode->i_bytes < bytes) {
729 inode->i_bytes += 512;
731 inode->i_bytes -= bytes;
734 EXPORT_SYMBOL(__inode_sub_bytes);
736 void inode_sub_bytes(struct inode *inode, loff_t bytes)
738 spin_lock(&inode->i_lock);
739 __inode_sub_bytes(inode, bytes);
740 spin_unlock(&inode->i_lock);
743 EXPORT_SYMBOL(inode_sub_bytes);
745 loff_t inode_get_bytes(struct inode *inode)
749 spin_lock(&inode->i_lock);
750 ret = __inode_get_bytes(inode);
751 spin_unlock(&inode->i_lock);
755 EXPORT_SYMBOL(inode_get_bytes);
757 void inode_set_bytes(struct inode *inode, loff_t bytes)
759 /* Caller is here responsible for sufficient locking
760 * (ie. inode->i_lock) */
761 inode->i_blocks = bytes >> 9;
762 inode->i_bytes = bytes & 511;
765 EXPORT_SYMBOL(inode_set_bytes);