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 (KSTAT_QUERY_FLAGS)
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 &= KSTAT_QUERY_FLAGS;
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;
80 if (IS_AUTOMOUNT(inode))
81 stat->attributes |= STATX_ATTR_AUTOMOUNT;
84 stat->attributes |= STATX_ATTR_DAX;
86 if (inode->i_op->getattr)
87 return inode->i_op->getattr(path, stat, request_mask,
90 generic_fillattr(inode, stat);
93 EXPORT_SYMBOL(vfs_getattr_nosec);
96 * vfs_getattr - Get the enhanced basic attributes of a file
97 * @path: The file of interest
98 * @stat: Where to return the statistics
99 * @request_mask: STATX_xxx flags indicating what the caller wants
100 * @query_flags: Query mode (KSTAT_QUERY_FLAGS)
102 * Ask the filesystem for a file's attributes. The caller must indicate in
103 * request_mask and query_flags to indicate what they want.
105 * If the file is remote, the filesystem can be forced to update the attributes
106 * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
107 * suppress the update by passing AT_STATX_DONT_SYNC.
109 * Bits must have been set in request_mask to indicate which attributes the
110 * caller wants retrieving. Any such attribute not requested may be returned
111 * anyway, but the value may be approximate, and, if remote, may not have been
112 * synchronised with the server.
114 * 0 will be returned on success, and a -ve error code if unsuccessful.
116 int vfs_getattr(const struct path *path, struct kstat *stat,
117 u32 request_mask, unsigned int query_flags)
121 retval = security_inode_getattr(path);
124 return vfs_getattr_nosec(path, stat, request_mask, query_flags);
126 EXPORT_SYMBOL(vfs_getattr);
129 * vfs_statx_fd - Get the enhanced basic attributes by file descriptor
130 * @fd: The file descriptor referring to the file of interest
131 * @stat: The result structure to fill in.
132 * @request_mask: STATX_xxx flags indicating what the caller wants
133 * @query_flags: Query mode (KSTAT_QUERY_FLAGS)
135 * This function is a wrapper around vfs_getattr(). The main difference is
136 * that it uses a file descriptor to determine the file location.
138 * 0 will be returned on success, and a -ve error code if unsuccessful.
140 int vfs_statx_fd(unsigned int fd, struct kstat *stat,
141 u32 request_mask, unsigned int query_flags)
146 if (query_flags & ~KSTAT_QUERY_FLAGS)
151 error = vfs_getattr(&f.file->f_path, stat,
152 request_mask, query_flags);
157 EXPORT_SYMBOL(vfs_statx_fd);
159 static inline unsigned vfs_stat_set_lookup_flags(unsigned *lookup_flags,
162 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT |
163 AT_EMPTY_PATH | KSTAT_QUERY_FLAGS)) != 0)
166 *lookup_flags = LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT;
167 if (flags & AT_SYMLINK_NOFOLLOW)
168 *lookup_flags &= ~LOOKUP_FOLLOW;
169 if (flags & AT_NO_AUTOMOUNT)
170 *lookup_flags &= ~LOOKUP_AUTOMOUNT;
171 if (flags & AT_EMPTY_PATH)
172 *lookup_flags |= LOOKUP_EMPTY;
178 * vfs_statx - Get basic and extra attributes by filename
179 * @dfd: A file descriptor representing the base dir for a relative filename
180 * @filename: The name of the file of interest
181 * @flags: Flags to control the query
182 * @stat: The result structure to fill in.
183 * @request_mask: STATX_xxx flags indicating what the caller wants
185 * This function is a wrapper around vfs_getattr(). The main difference is
186 * that it uses a filename and base directory to determine the file location.
187 * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
188 * at the given name from being referenced.
190 * 0 will be returned on success, and a -ve error code if unsuccessful.
192 int vfs_statx(int dfd, const char __user *filename, int flags,
193 struct kstat *stat, u32 request_mask)
197 unsigned lookup_flags;
199 if (vfs_stat_set_lookup_flags(&lookup_flags, flags))
202 error = user_path_at(dfd, filename, lookup_flags, &path);
206 error = vfs_getattr(&path, stat, request_mask, flags);
207 stat->mnt_id = real_mount(path.mnt)->mnt_id;
208 stat->result_mask |= STATX_MNT_ID;
209 if (path.mnt->mnt_root == path.dentry)
210 stat->attributes |= STATX_ATTR_MOUNT_ROOT;
211 stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
213 if (retry_estale(error, lookup_flags)) {
214 lookup_flags |= LOOKUP_REVAL;
220 EXPORT_SYMBOL(vfs_statx);
223 #ifdef __ARCH_WANT_OLD_STAT
226 * For backward compatibility? Maybe this should be moved
227 * into arch/i386 instead?
229 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
231 static int warncount = 5;
232 struct __old_kernel_stat tmp;
236 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
238 } else if (warncount < 0) {
239 /* it's laughable, but... */
243 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
244 tmp.st_dev = old_encode_dev(stat->dev);
245 tmp.st_ino = stat->ino;
246 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
248 tmp.st_mode = stat->mode;
249 tmp.st_nlink = stat->nlink;
250 if (tmp.st_nlink != stat->nlink)
252 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
253 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
254 tmp.st_rdev = old_encode_dev(stat->rdev);
255 #if BITS_PER_LONG == 32
256 if (stat->size > MAX_NON_LFS)
259 tmp.st_size = stat->size;
260 tmp.st_atime = stat->atime.tv_sec;
261 tmp.st_mtime = stat->mtime.tv_sec;
262 tmp.st_ctime = stat->ctime.tv_sec;
263 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
266 SYSCALL_DEFINE2(stat, const char __user *, filename,
267 struct __old_kernel_stat __user *, statbuf)
272 error = vfs_stat(filename, &stat);
276 return cp_old_stat(&stat, statbuf);
279 SYSCALL_DEFINE2(lstat, const char __user *, filename,
280 struct __old_kernel_stat __user *, statbuf)
285 error = vfs_lstat(filename, &stat);
289 return cp_old_stat(&stat, statbuf);
292 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
295 int error = vfs_fstat(fd, &stat);
298 error = cp_old_stat(&stat, statbuf);
303 #endif /* __ARCH_WANT_OLD_STAT */
305 #ifdef __ARCH_WANT_NEW_STAT
307 #if BITS_PER_LONG == 32
308 # define choose_32_64(a,b) a
310 # define choose_32_64(a,b) b
313 #define valid_dev(x) choose_32_64(old_valid_dev(x),true)
314 #define encode_dev(x) choose_32_64(old_encode_dev,new_encode_dev)(x)
316 #ifndef INIT_STRUCT_STAT_PADDING
317 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
320 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
324 if (!valid_dev(stat->dev) || !valid_dev(stat->rdev))
326 #if BITS_PER_LONG == 32
327 if (stat->size > MAX_NON_LFS)
331 INIT_STRUCT_STAT_PADDING(tmp);
332 tmp.st_dev = encode_dev(stat->dev);
333 tmp.st_ino = stat->ino;
334 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
336 tmp.st_mode = stat->mode;
337 tmp.st_nlink = stat->nlink;
338 if (tmp.st_nlink != stat->nlink)
340 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
341 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
342 tmp.st_rdev = encode_dev(stat->rdev);
343 tmp.st_size = stat->size;
344 tmp.st_atime = stat->atime.tv_sec;
345 tmp.st_mtime = stat->mtime.tv_sec;
346 tmp.st_ctime = stat->ctime.tv_sec;
347 #ifdef STAT_HAVE_NSEC
348 tmp.st_atime_nsec = stat->atime.tv_nsec;
349 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
350 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
352 tmp.st_blocks = stat->blocks;
353 tmp.st_blksize = stat->blksize;
354 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
357 SYSCALL_DEFINE2(newstat, const char __user *, filename,
358 struct stat __user *, statbuf)
361 int error = vfs_stat(filename, &stat);
365 return cp_new_stat(&stat, statbuf);
368 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
369 struct stat __user *, statbuf)
374 error = vfs_lstat(filename, &stat);
378 return cp_new_stat(&stat, statbuf);
381 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
382 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
383 struct stat __user *, statbuf, int, flag)
388 error = vfs_fstatat(dfd, filename, &stat, flag);
391 return cp_new_stat(&stat, statbuf);
395 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
398 int error = vfs_fstat(fd, &stat);
401 error = cp_new_stat(&stat, statbuf);
407 static int do_readlinkat(int dfd, const char __user *pathname,
408 char __user *buf, int bufsiz)
413 unsigned int lookup_flags = LOOKUP_EMPTY;
419 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
421 struct inode *inode = d_backing_inode(path.dentry);
423 error = empty ? -ENOENT : -EINVAL;
425 * AFS mountpoints allow readlink(2) but are not symlinks
427 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
428 error = security_inode_readlink(path.dentry);
431 error = vfs_readlink(path.dentry, buf, bufsiz);
435 if (retry_estale(error, lookup_flags)) {
436 lookup_flags |= LOOKUP_REVAL;
443 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
444 char __user *, buf, int, bufsiz)
446 return do_readlinkat(dfd, pathname, buf, bufsiz);
449 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
452 return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
456 /* ---------- LFS-64 ----------- */
457 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
459 #ifndef INIT_STRUCT_STAT64_PADDING
460 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
463 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
467 INIT_STRUCT_STAT64_PADDING(tmp);
469 /* mips has weird padding, so we don't get 64 bits there */
470 tmp.st_dev = new_encode_dev(stat->dev);
471 tmp.st_rdev = new_encode_dev(stat->rdev);
473 tmp.st_dev = huge_encode_dev(stat->dev);
474 tmp.st_rdev = huge_encode_dev(stat->rdev);
476 tmp.st_ino = stat->ino;
477 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
479 #ifdef STAT64_HAS_BROKEN_ST_INO
480 tmp.__st_ino = stat->ino;
482 tmp.st_mode = stat->mode;
483 tmp.st_nlink = stat->nlink;
484 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
485 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
486 tmp.st_atime = stat->atime.tv_sec;
487 tmp.st_atime_nsec = stat->atime.tv_nsec;
488 tmp.st_mtime = stat->mtime.tv_sec;
489 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
490 tmp.st_ctime = stat->ctime.tv_sec;
491 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
492 tmp.st_size = stat->size;
493 tmp.st_blocks = stat->blocks;
494 tmp.st_blksize = stat->blksize;
495 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
498 SYSCALL_DEFINE2(stat64, const char __user *, filename,
499 struct stat64 __user *, statbuf)
502 int error = vfs_stat(filename, &stat);
505 error = cp_new_stat64(&stat, statbuf);
510 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
511 struct stat64 __user *, statbuf)
514 int error = vfs_lstat(filename, &stat);
517 error = cp_new_stat64(&stat, statbuf);
522 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
525 int error = vfs_fstat(fd, &stat);
528 error = cp_new_stat64(&stat, statbuf);
533 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
534 struct stat64 __user *, statbuf, int, flag)
539 error = vfs_fstatat(dfd, filename, &stat, flag);
542 return cp_new_stat64(&stat, statbuf);
544 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
546 static noinline_for_stack int
547 cp_statx(const struct kstat *stat, struct statx __user *buffer)
551 memset(&tmp, 0, sizeof(tmp));
553 tmp.stx_mask = stat->result_mask;
554 tmp.stx_blksize = stat->blksize;
555 tmp.stx_attributes = stat->attributes;
556 tmp.stx_nlink = stat->nlink;
557 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
558 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
559 tmp.stx_mode = stat->mode;
560 tmp.stx_ino = stat->ino;
561 tmp.stx_size = stat->size;
562 tmp.stx_blocks = stat->blocks;
563 tmp.stx_attributes_mask = stat->attributes_mask;
564 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
565 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
566 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
567 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
568 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
569 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
570 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
571 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
572 tmp.stx_rdev_major = MAJOR(stat->rdev);
573 tmp.stx_rdev_minor = MINOR(stat->rdev);
574 tmp.stx_dev_major = MAJOR(stat->dev);
575 tmp.stx_dev_minor = MINOR(stat->dev);
576 tmp.stx_mnt_id = stat->mnt_id;
578 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
581 int do_statx(int dfd, const char __user *filename, unsigned flags,
582 unsigned int mask, struct statx __user *buffer)
587 if (mask & STATX__RESERVED)
589 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
592 error = vfs_statx(dfd, filename, flags, &stat, mask);
596 return cp_statx(&stat, buffer);
600 * sys_statx - System call to get enhanced stats
601 * @dfd: Base directory to pathwalk from *or* fd to stat.
602 * @filename: File to stat or "" with AT_EMPTY_PATH
603 * @flags: AT_* flags to control pathwalk.
604 * @mask: Parts of statx struct actually required.
605 * @buffer: Result buffer.
607 * Note that fstat() can be emulated by setting dfd to the fd of interest,
608 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
610 SYSCALL_DEFINE5(statx,
611 int, dfd, const char __user *, filename, unsigned, flags,
613 struct statx __user *, buffer)
615 return do_statx(dfd, filename, flags, mask, buffer);
619 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
621 struct compat_stat tmp;
623 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
626 memset(&tmp, 0, sizeof(tmp));
627 tmp.st_dev = old_encode_dev(stat->dev);
628 tmp.st_ino = stat->ino;
629 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
631 tmp.st_mode = stat->mode;
632 tmp.st_nlink = stat->nlink;
633 if (tmp.st_nlink != stat->nlink)
635 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
636 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
637 tmp.st_rdev = old_encode_dev(stat->rdev);
638 if ((u64) stat->size > MAX_NON_LFS)
640 tmp.st_size = stat->size;
641 tmp.st_atime = stat->atime.tv_sec;
642 tmp.st_atime_nsec = stat->atime.tv_nsec;
643 tmp.st_mtime = stat->mtime.tv_sec;
644 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
645 tmp.st_ctime = stat->ctime.tv_sec;
646 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
647 tmp.st_blocks = stat->blocks;
648 tmp.st_blksize = stat->blksize;
649 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
652 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
653 struct compat_stat __user *, statbuf)
658 error = vfs_stat(filename, &stat);
661 return cp_compat_stat(&stat, statbuf);
664 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
665 struct compat_stat __user *, statbuf)
670 error = vfs_lstat(filename, &stat);
673 return cp_compat_stat(&stat, statbuf);
676 #ifndef __ARCH_WANT_STAT64
677 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
678 const char __user *, filename,
679 struct compat_stat __user *, statbuf, int, flag)
684 error = vfs_fstatat(dfd, filename, &stat, flag);
687 return cp_compat_stat(&stat, statbuf);
691 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
692 struct compat_stat __user *, statbuf)
695 int error = vfs_fstat(fd, &stat);
698 error = cp_compat_stat(&stat, statbuf);
703 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
704 void __inode_add_bytes(struct inode *inode, loff_t bytes)
706 inode->i_blocks += bytes >> 9;
708 inode->i_bytes += bytes;
709 if (inode->i_bytes >= 512) {
711 inode->i_bytes -= 512;
714 EXPORT_SYMBOL(__inode_add_bytes);
716 void inode_add_bytes(struct inode *inode, loff_t bytes)
718 spin_lock(&inode->i_lock);
719 __inode_add_bytes(inode, bytes);
720 spin_unlock(&inode->i_lock);
723 EXPORT_SYMBOL(inode_add_bytes);
725 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
727 inode->i_blocks -= bytes >> 9;
729 if (inode->i_bytes < bytes) {
731 inode->i_bytes += 512;
733 inode->i_bytes -= bytes;
736 EXPORT_SYMBOL(__inode_sub_bytes);
738 void inode_sub_bytes(struct inode *inode, loff_t bytes)
740 spin_lock(&inode->i_lock);
741 __inode_sub_bytes(inode, bytes);
742 spin_unlock(&inode->i_lock);
745 EXPORT_SYMBOL(inode_sub_bytes);
747 loff_t inode_get_bytes(struct inode *inode)
751 spin_lock(&inode->i_lock);
752 ret = __inode_get_bytes(inode);
753 spin_unlock(&inode->i_lock);
757 EXPORT_SYMBOL(inode_get_bytes);
759 void inode_set_bytes(struct inode *inode, loff_t bytes)
761 /* Caller is here responsible for sufficient locking
762 * (ie. inode->i_lock) */
763 inode->i_blocks = bytes >> 9;
764 inode->i_bytes = bytes & 511;
767 EXPORT_SYMBOL(inode_set_bytes);