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_fstat - Get the basic attributes by file descriptor
130 * @fd: The file descriptor referring to the file of interest
131 * @stat: The result structure to fill in.
133 * This function is a wrapper around vfs_getattr(). The main difference is
134 * that it uses a file descriptor to determine the file location.
136 * 0 will be returned on success, and a -ve error code if unsuccessful.
138 int vfs_fstat(int fd, struct kstat *stat)
146 error = vfs_getattr(&f.file->f_path, stat, STATX_BASIC_STATS, 0);
151 static inline unsigned vfs_stat_set_lookup_flags(unsigned *lookup_flags,
154 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT |
155 AT_EMPTY_PATH | KSTAT_QUERY_FLAGS)) != 0)
158 *lookup_flags = LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT;
159 if (flags & AT_SYMLINK_NOFOLLOW)
160 *lookup_flags &= ~LOOKUP_FOLLOW;
161 if (flags & AT_NO_AUTOMOUNT)
162 *lookup_flags &= ~LOOKUP_AUTOMOUNT;
163 if (flags & AT_EMPTY_PATH)
164 *lookup_flags |= LOOKUP_EMPTY;
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 int vfs_statx(int dfd, const char __user *filename, int flags,
185 struct kstat *stat, u32 request_mask)
189 unsigned lookup_flags;
191 if (vfs_stat_set_lookup_flags(&lookup_flags, flags))
194 error = user_path_at(dfd, filename, lookup_flags, &path);
198 error = vfs_getattr(&path, stat, request_mask, flags);
199 stat->mnt_id = real_mount(path.mnt)->mnt_id;
200 stat->result_mask |= STATX_MNT_ID;
201 if (path.mnt->mnt_root == path.dentry)
202 stat->attributes |= STATX_ATTR_MOUNT_ROOT;
203 stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
205 if (retry_estale(error, lookup_flags)) {
206 lookup_flags |= LOOKUP_REVAL;
212 EXPORT_SYMBOL(vfs_statx);
215 #ifdef __ARCH_WANT_OLD_STAT
218 * For backward compatibility? Maybe this should be moved
219 * into arch/i386 instead?
221 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
223 static int warncount = 5;
224 struct __old_kernel_stat tmp;
228 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
230 } else if (warncount < 0) {
231 /* it's laughable, but... */
235 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
236 tmp.st_dev = old_encode_dev(stat->dev);
237 tmp.st_ino = stat->ino;
238 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
240 tmp.st_mode = stat->mode;
241 tmp.st_nlink = stat->nlink;
242 if (tmp.st_nlink != stat->nlink)
244 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
245 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
246 tmp.st_rdev = old_encode_dev(stat->rdev);
247 #if BITS_PER_LONG == 32
248 if (stat->size > MAX_NON_LFS)
251 tmp.st_size = stat->size;
252 tmp.st_atime = stat->atime.tv_sec;
253 tmp.st_mtime = stat->mtime.tv_sec;
254 tmp.st_ctime = stat->ctime.tv_sec;
255 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
258 SYSCALL_DEFINE2(stat, const char __user *, filename,
259 struct __old_kernel_stat __user *, statbuf)
264 error = vfs_stat(filename, &stat);
268 return cp_old_stat(&stat, statbuf);
271 SYSCALL_DEFINE2(lstat, const char __user *, filename,
272 struct __old_kernel_stat __user *, statbuf)
277 error = vfs_lstat(filename, &stat);
281 return cp_old_stat(&stat, statbuf);
284 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
287 int error = vfs_fstat(fd, &stat);
290 error = cp_old_stat(&stat, statbuf);
295 #endif /* __ARCH_WANT_OLD_STAT */
297 #ifdef __ARCH_WANT_NEW_STAT
299 #if BITS_PER_LONG == 32
300 # define choose_32_64(a,b) a
302 # define choose_32_64(a,b) b
305 #define valid_dev(x) choose_32_64(old_valid_dev(x),true)
306 #define encode_dev(x) choose_32_64(old_encode_dev,new_encode_dev)(x)
308 #ifndef INIT_STRUCT_STAT_PADDING
309 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
312 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
316 if (!valid_dev(stat->dev) || !valid_dev(stat->rdev))
318 #if BITS_PER_LONG == 32
319 if (stat->size > MAX_NON_LFS)
323 INIT_STRUCT_STAT_PADDING(tmp);
324 tmp.st_dev = encode_dev(stat->dev);
325 tmp.st_ino = stat->ino;
326 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
328 tmp.st_mode = stat->mode;
329 tmp.st_nlink = stat->nlink;
330 if (tmp.st_nlink != stat->nlink)
332 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
333 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
334 tmp.st_rdev = encode_dev(stat->rdev);
335 tmp.st_size = stat->size;
336 tmp.st_atime = stat->atime.tv_sec;
337 tmp.st_mtime = stat->mtime.tv_sec;
338 tmp.st_ctime = stat->ctime.tv_sec;
339 #ifdef STAT_HAVE_NSEC
340 tmp.st_atime_nsec = stat->atime.tv_nsec;
341 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
342 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
344 tmp.st_blocks = stat->blocks;
345 tmp.st_blksize = stat->blksize;
346 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
349 SYSCALL_DEFINE2(newstat, const char __user *, filename,
350 struct stat __user *, statbuf)
353 int error = vfs_stat(filename, &stat);
357 return cp_new_stat(&stat, statbuf);
360 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
361 struct stat __user *, statbuf)
366 error = vfs_lstat(filename, &stat);
370 return cp_new_stat(&stat, statbuf);
373 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
374 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
375 struct stat __user *, statbuf, int, flag)
380 error = vfs_fstatat(dfd, filename, &stat, flag);
383 return cp_new_stat(&stat, statbuf);
387 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
390 int error = vfs_fstat(fd, &stat);
393 error = cp_new_stat(&stat, statbuf);
399 static int do_readlinkat(int dfd, const char __user *pathname,
400 char __user *buf, int bufsiz)
405 unsigned int lookup_flags = LOOKUP_EMPTY;
411 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
413 struct inode *inode = d_backing_inode(path.dentry);
415 error = empty ? -ENOENT : -EINVAL;
417 * AFS mountpoints allow readlink(2) but are not symlinks
419 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
420 error = security_inode_readlink(path.dentry);
423 error = vfs_readlink(path.dentry, buf, bufsiz);
427 if (retry_estale(error, lookup_flags)) {
428 lookup_flags |= LOOKUP_REVAL;
435 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
436 char __user *, buf, int, bufsiz)
438 return do_readlinkat(dfd, pathname, buf, bufsiz);
441 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
444 return do_readlinkat(AT_FDCWD, path, buf, bufsiz);
448 /* ---------- LFS-64 ----------- */
449 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
451 #ifndef INIT_STRUCT_STAT64_PADDING
452 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
455 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
459 INIT_STRUCT_STAT64_PADDING(tmp);
461 /* mips has weird padding, so we don't get 64 bits there */
462 tmp.st_dev = new_encode_dev(stat->dev);
463 tmp.st_rdev = new_encode_dev(stat->rdev);
465 tmp.st_dev = huge_encode_dev(stat->dev);
466 tmp.st_rdev = huge_encode_dev(stat->rdev);
468 tmp.st_ino = stat->ino;
469 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
471 #ifdef STAT64_HAS_BROKEN_ST_INO
472 tmp.__st_ino = stat->ino;
474 tmp.st_mode = stat->mode;
475 tmp.st_nlink = stat->nlink;
476 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
477 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
478 tmp.st_atime = stat->atime.tv_sec;
479 tmp.st_atime_nsec = stat->atime.tv_nsec;
480 tmp.st_mtime = stat->mtime.tv_sec;
481 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
482 tmp.st_ctime = stat->ctime.tv_sec;
483 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
484 tmp.st_size = stat->size;
485 tmp.st_blocks = stat->blocks;
486 tmp.st_blksize = stat->blksize;
487 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
490 SYSCALL_DEFINE2(stat64, const char __user *, filename,
491 struct stat64 __user *, statbuf)
494 int error = vfs_stat(filename, &stat);
497 error = cp_new_stat64(&stat, statbuf);
502 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
503 struct stat64 __user *, statbuf)
506 int error = vfs_lstat(filename, &stat);
509 error = cp_new_stat64(&stat, statbuf);
514 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
517 int error = vfs_fstat(fd, &stat);
520 error = cp_new_stat64(&stat, statbuf);
525 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
526 struct stat64 __user *, statbuf, int, flag)
531 error = vfs_fstatat(dfd, filename, &stat, flag);
534 return cp_new_stat64(&stat, statbuf);
536 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
538 static noinline_for_stack int
539 cp_statx(const struct kstat *stat, struct statx __user *buffer)
543 memset(&tmp, 0, sizeof(tmp));
545 tmp.stx_mask = stat->result_mask;
546 tmp.stx_blksize = stat->blksize;
547 tmp.stx_attributes = stat->attributes;
548 tmp.stx_nlink = stat->nlink;
549 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
550 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
551 tmp.stx_mode = stat->mode;
552 tmp.stx_ino = stat->ino;
553 tmp.stx_size = stat->size;
554 tmp.stx_blocks = stat->blocks;
555 tmp.stx_attributes_mask = stat->attributes_mask;
556 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
557 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
558 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
559 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
560 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
561 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
562 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
563 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
564 tmp.stx_rdev_major = MAJOR(stat->rdev);
565 tmp.stx_rdev_minor = MINOR(stat->rdev);
566 tmp.stx_dev_major = MAJOR(stat->dev);
567 tmp.stx_dev_minor = MINOR(stat->dev);
568 tmp.stx_mnt_id = stat->mnt_id;
570 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
573 int do_statx(int dfd, const char __user *filename, unsigned flags,
574 unsigned int mask, struct statx __user *buffer)
579 if (mask & STATX__RESERVED)
581 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
584 error = vfs_statx(dfd, filename, flags, &stat, mask);
588 return cp_statx(&stat, buffer);
592 * sys_statx - System call to get enhanced stats
593 * @dfd: Base directory to pathwalk from *or* fd to stat.
594 * @filename: File to stat or "" with AT_EMPTY_PATH
595 * @flags: AT_* flags to control pathwalk.
596 * @mask: Parts of statx struct actually required.
597 * @buffer: Result buffer.
599 * Note that fstat() can be emulated by setting dfd to the fd of interest,
600 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
602 SYSCALL_DEFINE5(statx,
603 int, dfd, const char __user *, filename, unsigned, flags,
605 struct statx __user *, buffer)
607 return do_statx(dfd, filename, flags, mask, buffer);
611 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
613 struct compat_stat tmp;
615 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
618 memset(&tmp, 0, sizeof(tmp));
619 tmp.st_dev = old_encode_dev(stat->dev);
620 tmp.st_ino = stat->ino;
621 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
623 tmp.st_mode = stat->mode;
624 tmp.st_nlink = stat->nlink;
625 if (tmp.st_nlink != stat->nlink)
627 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
628 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
629 tmp.st_rdev = old_encode_dev(stat->rdev);
630 if ((u64) stat->size > MAX_NON_LFS)
632 tmp.st_size = stat->size;
633 tmp.st_atime = stat->atime.tv_sec;
634 tmp.st_atime_nsec = stat->atime.tv_nsec;
635 tmp.st_mtime = stat->mtime.tv_sec;
636 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
637 tmp.st_ctime = stat->ctime.tv_sec;
638 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
639 tmp.st_blocks = stat->blocks;
640 tmp.st_blksize = stat->blksize;
641 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
644 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
645 struct compat_stat __user *, statbuf)
650 error = vfs_stat(filename, &stat);
653 return cp_compat_stat(&stat, statbuf);
656 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
657 struct compat_stat __user *, statbuf)
662 error = vfs_lstat(filename, &stat);
665 return cp_compat_stat(&stat, statbuf);
668 #ifndef __ARCH_WANT_STAT64
669 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
670 const char __user *, filename,
671 struct compat_stat __user *, statbuf, int, flag)
676 error = vfs_fstatat(dfd, filename, &stat, flag);
679 return cp_compat_stat(&stat, statbuf);
683 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
684 struct compat_stat __user *, statbuf)
687 int error = vfs_fstat(fd, &stat);
690 error = cp_compat_stat(&stat, statbuf);
695 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
696 void __inode_add_bytes(struct inode *inode, loff_t bytes)
698 inode->i_blocks += bytes >> 9;
700 inode->i_bytes += bytes;
701 if (inode->i_bytes >= 512) {
703 inode->i_bytes -= 512;
706 EXPORT_SYMBOL(__inode_add_bytes);
708 void inode_add_bytes(struct inode *inode, loff_t bytes)
710 spin_lock(&inode->i_lock);
711 __inode_add_bytes(inode, bytes);
712 spin_unlock(&inode->i_lock);
715 EXPORT_SYMBOL(inode_add_bytes);
717 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
719 inode->i_blocks -= bytes >> 9;
721 if (inode->i_bytes < bytes) {
723 inode->i_bytes += 512;
725 inode->i_bytes -= bytes;
728 EXPORT_SYMBOL(__inode_sub_bytes);
730 void inode_sub_bytes(struct inode *inode, loff_t bytes)
732 spin_lock(&inode->i_lock);
733 __inode_sub_bytes(inode, bytes);
734 spin_unlock(&inode->i_lock);
737 EXPORT_SYMBOL(inode_sub_bytes);
739 loff_t inode_get_bytes(struct inode *inode)
743 spin_lock(&inode->i_lock);
744 ret = __inode_get_bytes(inode);
745 spin_unlock(&inode->i_lock);
749 EXPORT_SYMBOL(inode_get_bytes);
751 void inode_set_bytes(struct inode *inode, loff_t bytes)
753 /* Caller is here responsible for sufficient locking
754 * (ie. inode->i_lock) */
755 inode->i_blocks = bytes >> 9;
756 inode->i_bytes = bytes & 511;
759 EXPORT_SYMBOL(inode_set_bytes);