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>
25 * generic_fillattr - Fill in the basic attributes from the inode struct
26 * @inode: Inode to use as the source
27 * @stat: Where to fill in the attributes
29 * Fill in the basic attributes in the kstat structure from data that's to be
30 * found on the VFS inode structure. This is the default if no getattr inode
31 * operation is supplied.
33 void generic_fillattr(struct inode *inode, struct kstat *stat)
35 stat->dev = inode->i_sb->s_dev;
36 stat->ino = inode->i_ino;
37 stat->mode = inode->i_mode;
38 stat->nlink = inode->i_nlink;
39 stat->uid = inode->i_uid;
40 stat->gid = inode->i_gid;
41 stat->rdev = inode->i_rdev;
42 stat->size = i_size_read(inode);
43 stat->atime = inode->i_atime;
44 stat->mtime = inode->i_mtime;
45 stat->ctime = inode->i_ctime;
46 stat->blksize = i_blocksize(inode);
47 stat->blocks = inode->i_blocks;
49 if (IS_NOATIME(inode))
50 stat->result_mask &= ~STATX_ATIME;
51 if (IS_AUTOMOUNT(inode))
52 stat->attributes |= STATX_ATTR_AUTOMOUNT;
54 EXPORT_SYMBOL(generic_fillattr);
57 * vfs_getattr_nosec - getattr without security checks
58 * @path: file to get attributes from
59 * @stat: structure to return attributes in
60 * @request_mask: STATX_xxx flags indicating what the caller wants
61 * @query_flags: Query mode (KSTAT_QUERY_FLAGS)
63 * Get attributes without calling security_inode_getattr.
65 * Currently the only caller other than vfs_getattr is internal to the
66 * filehandle lookup code, which uses only the inode number and returns no
67 * attributes to any user. Any other code probably wants vfs_getattr.
69 int vfs_getattr_nosec(const struct path *path, struct kstat *stat,
70 u32 request_mask, unsigned int query_flags)
72 struct inode *inode = d_backing_inode(path->dentry);
74 memset(stat, 0, sizeof(*stat));
75 stat->result_mask |= STATX_BASIC_STATS;
76 request_mask &= STATX_ALL;
77 query_flags &= KSTAT_QUERY_FLAGS;
78 if (inode->i_op->getattr)
79 return inode->i_op->getattr(path, stat, request_mask,
82 generic_fillattr(inode, stat);
85 EXPORT_SYMBOL(vfs_getattr_nosec);
88 * vfs_getattr - Get the enhanced basic attributes of a file
89 * @path: The file of interest
90 * @stat: Where to return the statistics
91 * @request_mask: STATX_xxx flags indicating what the caller wants
92 * @query_flags: Query mode (KSTAT_QUERY_FLAGS)
94 * Ask the filesystem for a file's attributes. The caller must indicate in
95 * request_mask and query_flags to indicate what they want.
97 * If the file is remote, the filesystem can be forced to update the attributes
98 * from the backing store by passing AT_STATX_FORCE_SYNC in query_flags or can
99 * suppress the update by passing AT_STATX_DONT_SYNC.
101 * Bits must have been set in request_mask to indicate which attributes the
102 * caller wants retrieving. Any such attribute not requested may be returned
103 * anyway, but the value may be approximate, and, if remote, may not have been
104 * synchronised with the server.
106 * 0 will be returned on success, and a -ve error code if unsuccessful.
108 int vfs_getattr(const struct path *path, struct kstat *stat,
109 u32 request_mask, unsigned int query_flags)
113 retval = security_inode_getattr(path);
116 return vfs_getattr_nosec(path, stat, request_mask, query_flags);
118 EXPORT_SYMBOL(vfs_getattr);
121 * vfs_statx_fd - Get the enhanced basic attributes by file descriptor
122 * @fd: The file descriptor referring to the file of interest
123 * @stat: The result structure to fill in.
124 * @request_mask: STATX_xxx flags indicating what the caller wants
125 * @query_flags: Query mode (KSTAT_QUERY_FLAGS)
127 * This function is a wrapper around vfs_getattr(). The main difference is
128 * that it uses a file descriptor to determine the file location.
130 * 0 will be returned on success, and a -ve error code if unsuccessful.
132 int vfs_statx_fd(unsigned int fd, struct kstat *stat,
133 u32 request_mask, unsigned int query_flags)
138 if (query_flags & ~KSTAT_QUERY_FLAGS)
143 error = vfs_getattr(&f.file->f_path, stat,
144 request_mask, query_flags);
149 EXPORT_SYMBOL(vfs_statx_fd);
152 * vfs_statx - Get basic and extra attributes by filename
153 * @dfd: A file descriptor representing the base dir for a relative filename
154 * @filename: The name of the file of interest
155 * @flags: Flags to control the query
156 * @stat: The result structure to fill in.
157 * @request_mask: STATX_xxx flags indicating what the caller wants
159 * This function is a wrapper around vfs_getattr(). The main difference is
160 * that it uses a filename and base directory to determine the file location.
161 * Additionally, the use of AT_SYMLINK_NOFOLLOW in flags will prevent a symlink
162 * at the given name from being referenced.
164 * 0 will be returned on success, and a -ve error code if unsuccessful.
166 int vfs_statx(int dfd, const char __user *filename, int flags,
167 struct kstat *stat, u32 request_mask)
171 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT;
173 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT |
174 AT_EMPTY_PATH | KSTAT_QUERY_FLAGS)) != 0)
177 if (flags & AT_SYMLINK_NOFOLLOW)
178 lookup_flags &= ~LOOKUP_FOLLOW;
179 if (flags & AT_NO_AUTOMOUNT)
180 lookup_flags &= ~LOOKUP_AUTOMOUNT;
181 if (flags & AT_EMPTY_PATH)
182 lookup_flags |= LOOKUP_EMPTY;
185 error = user_path_at(dfd, filename, lookup_flags, &path);
189 error = vfs_getattr(&path, stat, request_mask, flags);
191 if (retry_estale(error, lookup_flags)) {
192 lookup_flags |= LOOKUP_REVAL;
198 EXPORT_SYMBOL(vfs_statx);
201 #ifdef __ARCH_WANT_OLD_STAT
204 * For backward compatibility? Maybe this should be moved
205 * into arch/i386 instead?
207 static int cp_old_stat(struct kstat *stat, struct __old_kernel_stat __user * statbuf)
209 static int warncount = 5;
210 struct __old_kernel_stat tmp;
214 printk(KERN_WARNING "VFS: Warning: %s using old stat() call. Recompile your binary.\n",
216 } else if (warncount < 0) {
217 /* it's laughable, but... */
221 memset(&tmp, 0, sizeof(struct __old_kernel_stat));
222 tmp.st_dev = old_encode_dev(stat->dev);
223 tmp.st_ino = stat->ino;
224 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
226 tmp.st_mode = stat->mode;
227 tmp.st_nlink = stat->nlink;
228 if (tmp.st_nlink != stat->nlink)
230 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
231 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
232 tmp.st_rdev = old_encode_dev(stat->rdev);
233 #if BITS_PER_LONG == 32
234 if (stat->size > MAX_NON_LFS)
237 tmp.st_size = stat->size;
238 tmp.st_atime = stat->atime.tv_sec;
239 tmp.st_mtime = stat->mtime.tv_sec;
240 tmp.st_ctime = stat->ctime.tv_sec;
241 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
244 SYSCALL_DEFINE2(stat, const char __user *, filename,
245 struct __old_kernel_stat __user *, statbuf)
250 error = vfs_stat(filename, &stat);
254 return cp_old_stat(&stat, statbuf);
257 SYSCALL_DEFINE2(lstat, const char __user *, filename,
258 struct __old_kernel_stat __user *, statbuf)
263 error = vfs_lstat(filename, &stat);
267 return cp_old_stat(&stat, statbuf);
270 SYSCALL_DEFINE2(fstat, unsigned int, fd, struct __old_kernel_stat __user *, statbuf)
273 int error = vfs_fstat(fd, &stat);
276 error = cp_old_stat(&stat, statbuf);
281 #endif /* __ARCH_WANT_OLD_STAT */
283 #if BITS_PER_LONG == 32
284 # define choose_32_64(a,b) a
286 # define choose_32_64(a,b) b
289 #ifndef INIT_STRUCT_STAT_PADDING
290 # define INIT_STRUCT_STAT_PADDING(st) memset(&st, 0, sizeof(st))
293 static int cp_new_stat(struct kstat *stat, struct stat __user *statbuf)
297 if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
299 if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
301 #if BITS_PER_LONG == 32
302 if (stat->size > MAX_NON_LFS)
306 INIT_STRUCT_STAT_PADDING(tmp);
307 tmp.st_dev = new_encode_dev(stat->dev);
308 tmp.st_ino = stat->ino;
309 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
311 tmp.st_mode = stat->mode;
312 tmp.st_nlink = stat->nlink;
313 if (tmp.st_nlink != stat->nlink)
315 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
316 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
317 tmp.st_rdev = new_encode_dev(stat->rdev);
318 tmp.st_size = stat->size;
319 tmp.st_atime = stat->atime.tv_sec;
320 tmp.st_mtime = stat->mtime.tv_sec;
321 tmp.st_ctime = stat->ctime.tv_sec;
322 #ifdef STAT_HAVE_NSEC
323 tmp.st_atime_nsec = stat->atime.tv_nsec;
324 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
325 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
327 tmp.st_blocks = stat->blocks;
328 tmp.st_blksize = stat->blksize;
329 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
332 SYSCALL_DEFINE2(newstat, const char __user *, filename,
333 struct stat __user *, statbuf)
336 int error = vfs_stat(filename, &stat);
340 return cp_new_stat(&stat, statbuf);
343 SYSCALL_DEFINE2(newlstat, const char __user *, filename,
344 struct stat __user *, statbuf)
349 error = vfs_lstat(filename, &stat);
353 return cp_new_stat(&stat, statbuf);
356 #if !defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_SYS_NEWFSTATAT)
357 SYSCALL_DEFINE4(newfstatat, int, dfd, const char __user *, filename,
358 struct stat __user *, statbuf, int, flag)
363 error = vfs_fstatat(dfd, filename, &stat, flag);
366 return cp_new_stat(&stat, statbuf);
370 SYSCALL_DEFINE2(newfstat, unsigned int, fd, struct stat __user *, statbuf)
373 int error = vfs_fstat(fd, &stat);
376 error = cp_new_stat(&stat, statbuf);
381 SYSCALL_DEFINE4(readlinkat, int, dfd, const char __user *, pathname,
382 char __user *, buf, int, bufsiz)
387 unsigned int lookup_flags = LOOKUP_EMPTY;
393 error = user_path_at_empty(dfd, pathname, lookup_flags, &path, &empty);
395 struct inode *inode = d_backing_inode(path.dentry);
397 error = empty ? -ENOENT : -EINVAL;
399 * AFS mountpoints allow readlink(2) but are not symlinks
401 if (d_is_symlink(path.dentry) || inode->i_op->readlink) {
402 error = security_inode_readlink(path.dentry);
405 error = vfs_readlink(path.dentry, buf, bufsiz);
409 if (retry_estale(error, lookup_flags)) {
410 lookup_flags |= LOOKUP_REVAL;
417 SYSCALL_DEFINE3(readlink, const char __user *, path, char __user *, buf,
420 return sys_readlinkat(AT_FDCWD, path, buf, bufsiz);
424 /* ---------- LFS-64 ----------- */
425 #if defined(__ARCH_WANT_STAT64) || defined(__ARCH_WANT_COMPAT_STAT64)
427 #ifndef INIT_STRUCT_STAT64_PADDING
428 # define INIT_STRUCT_STAT64_PADDING(st) memset(&st, 0, sizeof(st))
431 static long cp_new_stat64(struct kstat *stat, struct stat64 __user *statbuf)
435 INIT_STRUCT_STAT64_PADDING(tmp);
437 /* mips has weird padding, so we don't get 64 bits there */
438 tmp.st_dev = new_encode_dev(stat->dev);
439 tmp.st_rdev = new_encode_dev(stat->rdev);
441 tmp.st_dev = huge_encode_dev(stat->dev);
442 tmp.st_rdev = huge_encode_dev(stat->rdev);
444 tmp.st_ino = stat->ino;
445 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
447 #ifdef STAT64_HAS_BROKEN_ST_INO
448 tmp.__st_ino = stat->ino;
450 tmp.st_mode = stat->mode;
451 tmp.st_nlink = stat->nlink;
452 tmp.st_uid = from_kuid_munged(current_user_ns(), stat->uid);
453 tmp.st_gid = from_kgid_munged(current_user_ns(), stat->gid);
454 tmp.st_atime = stat->atime.tv_sec;
455 tmp.st_atime_nsec = stat->atime.tv_nsec;
456 tmp.st_mtime = stat->mtime.tv_sec;
457 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
458 tmp.st_ctime = stat->ctime.tv_sec;
459 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
460 tmp.st_size = stat->size;
461 tmp.st_blocks = stat->blocks;
462 tmp.st_blksize = stat->blksize;
463 return copy_to_user(statbuf,&tmp,sizeof(tmp)) ? -EFAULT : 0;
466 SYSCALL_DEFINE2(stat64, const char __user *, filename,
467 struct stat64 __user *, statbuf)
470 int error = vfs_stat(filename, &stat);
473 error = cp_new_stat64(&stat, statbuf);
478 SYSCALL_DEFINE2(lstat64, const char __user *, filename,
479 struct stat64 __user *, statbuf)
482 int error = vfs_lstat(filename, &stat);
485 error = cp_new_stat64(&stat, statbuf);
490 SYSCALL_DEFINE2(fstat64, unsigned long, fd, struct stat64 __user *, statbuf)
493 int error = vfs_fstat(fd, &stat);
496 error = cp_new_stat64(&stat, statbuf);
501 SYSCALL_DEFINE4(fstatat64, int, dfd, const char __user *, filename,
502 struct stat64 __user *, statbuf, int, flag)
507 error = vfs_fstatat(dfd, filename, &stat, flag);
510 return cp_new_stat64(&stat, statbuf);
512 #endif /* __ARCH_WANT_STAT64 || __ARCH_WANT_COMPAT_STAT64 */
514 static noinline_for_stack int
515 cp_statx(const struct kstat *stat, struct statx __user *buffer)
519 memset(&tmp, 0, sizeof(tmp));
521 tmp.stx_mask = stat->result_mask;
522 tmp.stx_blksize = stat->blksize;
523 tmp.stx_attributes = stat->attributes;
524 tmp.stx_nlink = stat->nlink;
525 tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
526 tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
527 tmp.stx_mode = stat->mode;
528 tmp.stx_ino = stat->ino;
529 tmp.stx_size = stat->size;
530 tmp.stx_blocks = stat->blocks;
531 tmp.stx_attributes_mask = stat->attributes_mask;
532 tmp.stx_atime.tv_sec = stat->atime.tv_sec;
533 tmp.stx_atime.tv_nsec = stat->atime.tv_nsec;
534 tmp.stx_btime.tv_sec = stat->btime.tv_sec;
535 tmp.stx_btime.tv_nsec = stat->btime.tv_nsec;
536 tmp.stx_ctime.tv_sec = stat->ctime.tv_sec;
537 tmp.stx_ctime.tv_nsec = stat->ctime.tv_nsec;
538 tmp.stx_mtime.tv_sec = stat->mtime.tv_sec;
539 tmp.stx_mtime.tv_nsec = stat->mtime.tv_nsec;
540 tmp.stx_rdev_major = MAJOR(stat->rdev);
541 tmp.stx_rdev_minor = MINOR(stat->rdev);
542 tmp.stx_dev_major = MAJOR(stat->dev);
543 tmp.stx_dev_minor = MINOR(stat->dev);
545 return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
549 * sys_statx - System call to get enhanced stats
550 * @dfd: Base directory to pathwalk from *or* fd to stat.
551 * @filename: File to stat or "" with AT_EMPTY_PATH
552 * @flags: AT_* flags to control pathwalk.
553 * @mask: Parts of statx struct actually required.
554 * @buffer: Result buffer.
556 * Note that fstat() can be emulated by setting dfd to the fd of interest,
557 * supplying "" as the filename and setting AT_EMPTY_PATH in the flags.
559 SYSCALL_DEFINE5(statx,
560 int, dfd, const char __user *, filename, unsigned, flags,
562 struct statx __user *, buffer)
567 if (mask & STATX__RESERVED)
569 if ((flags & AT_STATX_SYNC_TYPE) == AT_STATX_SYNC_TYPE)
572 error = vfs_statx(dfd, filename, flags, &stat, mask);
576 return cp_statx(&stat, buffer);
580 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
582 struct compat_stat tmp;
584 if (sizeof(tmp.st_dev) < 4 && !old_valid_dev(stat->dev))
586 if (sizeof(tmp.st_rdev) < 4 && !old_valid_dev(stat->rdev))
589 memset(&tmp, 0, sizeof(tmp));
590 tmp.st_dev = new_encode_dev(stat->dev);
591 tmp.st_ino = stat->ino;
592 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
594 tmp.st_mode = stat->mode;
595 tmp.st_nlink = stat->nlink;
596 if (tmp.st_nlink != stat->nlink)
598 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
599 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
600 tmp.st_rdev = new_encode_dev(stat->rdev);
601 if ((u64) stat->size > MAX_NON_LFS)
603 tmp.st_size = stat->size;
604 tmp.st_atime = stat->atime.tv_sec;
605 tmp.st_atime_nsec = stat->atime.tv_nsec;
606 tmp.st_mtime = stat->mtime.tv_sec;
607 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
608 tmp.st_ctime = stat->ctime.tv_sec;
609 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
610 tmp.st_blocks = stat->blocks;
611 tmp.st_blksize = stat->blksize;
612 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
615 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
616 struct compat_stat __user *, statbuf)
621 error = vfs_stat(filename, &stat);
624 return cp_compat_stat(&stat, statbuf);
627 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
628 struct compat_stat __user *, statbuf)
633 error = vfs_lstat(filename, &stat);
636 return cp_compat_stat(&stat, statbuf);
639 #ifndef __ARCH_WANT_STAT64
640 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
641 const char __user *, filename,
642 struct compat_stat __user *, statbuf, int, flag)
647 error = vfs_fstatat(dfd, filename, &stat, flag);
650 return cp_compat_stat(&stat, statbuf);
654 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
655 struct compat_stat __user *, statbuf)
658 int error = vfs_fstat(fd, &stat);
661 error = cp_compat_stat(&stat, statbuf);
666 /* Caller is here responsible for sufficient locking (ie. inode->i_lock) */
667 void __inode_add_bytes(struct inode *inode, loff_t bytes)
669 inode->i_blocks += bytes >> 9;
671 inode->i_bytes += bytes;
672 if (inode->i_bytes >= 512) {
674 inode->i_bytes -= 512;
677 EXPORT_SYMBOL(__inode_add_bytes);
679 void inode_add_bytes(struct inode *inode, loff_t bytes)
681 spin_lock(&inode->i_lock);
682 __inode_add_bytes(inode, bytes);
683 spin_unlock(&inode->i_lock);
686 EXPORT_SYMBOL(inode_add_bytes);
688 void __inode_sub_bytes(struct inode *inode, loff_t bytes)
690 inode->i_blocks -= bytes >> 9;
692 if (inode->i_bytes < bytes) {
694 inode->i_bytes += 512;
696 inode->i_bytes -= bytes;
699 EXPORT_SYMBOL(__inode_sub_bytes);
701 void inode_sub_bytes(struct inode *inode, loff_t bytes)
703 spin_lock(&inode->i_lock);
704 __inode_sub_bytes(inode, bytes);
705 spin_unlock(&inode->i_lock);
708 EXPORT_SYMBOL(inode_sub_bytes);
710 loff_t inode_get_bytes(struct inode *inode)
714 spin_lock(&inode->i_lock);
715 ret = __inode_get_bytes(inode);
716 spin_unlock(&inode->i_lock);
720 EXPORT_SYMBOL(inode_get_bytes);
722 void inode_set_bytes(struct inode *inode, loff_t bytes)
724 /* Caller is here responsible for sufficient locking
725 * (ie. inode->i_lock) */
726 inode->i_blocks = bytes >> 9;
727 inode->i_bytes = bytes & 511;
730 EXPORT_SYMBOL(inode_set_bytes);