4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/quotaops.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/tty.h>
16 #include <linux/namei.h>
17 #include <linux/backing-dev.h>
18 #include <linux/capability.h>
19 #include <linux/securebits.h>
20 #include <linux/security.h>
21 #include <linux/mount.h>
22 #include <linux/vfs.h>
23 #include <linux/fcntl.h>
24 #include <asm/uaccess.h>
26 #include <linux/personality.h>
27 #include <linux/pagemap.h>
28 #include <linux/syscalls.h>
29 #include <linux/rcupdate.h>
30 #include <linux/audit.h>
31 #include <linux/falloc.h>
33 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
39 if (dentry->d_sb->s_op->statfs) {
40 memset(buf, 0, sizeof(*buf));
41 retval = security_sb_statfs(dentry);
44 retval = dentry->d_sb->s_op->statfs(dentry, buf);
45 if (retval == 0 && buf->f_frsize == 0)
46 buf->f_frsize = buf->f_bsize;
52 EXPORT_SYMBOL(vfs_statfs);
54 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
59 retval = vfs_statfs(dentry, &st);
63 if (sizeof(*buf) == sizeof(st))
64 memcpy(buf, &st, sizeof(st));
66 if (sizeof buf->f_blocks == 4) {
67 if ((st.f_blocks | st.f_bfree | st.f_bavail |
68 st.f_bsize | st.f_frsize) &
69 0xffffffff00000000ULL)
72 * f_files and f_ffree may be -1; it's okay to stuff
75 if (st.f_files != -1 &&
76 (st.f_files & 0xffffffff00000000ULL))
78 if (st.f_ffree != -1 &&
79 (st.f_ffree & 0xffffffff00000000ULL))
83 buf->f_type = st.f_type;
84 buf->f_bsize = st.f_bsize;
85 buf->f_blocks = st.f_blocks;
86 buf->f_bfree = st.f_bfree;
87 buf->f_bavail = st.f_bavail;
88 buf->f_files = st.f_files;
89 buf->f_ffree = st.f_ffree;
90 buf->f_fsid = st.f_fsid;
91 buf->f_namelen = st.f_namelen;
92 buf->f_frsize = st.f_frsize;
93 memset(buf->f_spare, 0, sizeof(buf->f_spare));
98 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
103 retval = vfs_statfs(dentry, &st);
107 if (sizeof(*buf) == sizeof(st))
108 memcpy(buf, &st, sizeof(st));
110 buf->f_type = st.f_type;
111 buf->f_bsize = st.f_bsize;
112 buf->f_blocks = st.f_blocks;
113 buf->f_bfree = st.f_bfree;
114 buf->f_bavail = st.f_bavail;
115 buf->f_files = st.f_files;
116 buf->f_ffree = st.f_ffree;
117 buf->f_fsid = st.f_fsid;
118 buf->f_namelen = st.f_namelen;
119 buf->f_frsize = st.f_frsize;
120 memset(buf->f_spare, 0, sizeof(buf->f_spare));
125 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
130 error = user_path_walk(path, &nd);
133 error = vfs_statfs_native(nd.path.dentry, &tmp);
134 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
142 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
147 if (sz != sizeof(*buf))
149 error = user_path_walk(path, &nd);
152 error = vfs_statfs64(nd.path.dentry, &tmp);
153 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
161 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
171 error = vfs_statfs_native(file->f_path.dentry, &tmp);
172 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
179 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
185 if (sz != sizeof(*buf))
192 error = vfs_statfs64(file->f_path.dentry, &tmp);
193 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
200 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
204 struct iattr newattrs;
206 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
210 newattrs.ia_size = length;
211 newattrs.ia_valid = ATTR_SIZE | time_attrs;
213 newattrs.ia_file = filp;
214 newattrs.ia_valid |= ATTR_FILE;
217 /* Remove suid/sgid on truncate too */
218 newattrs.ia_valid |= should_remove_suid(dentry);
220 mutex_lock(&dentry->d_inode->i_mutex);
221 err = notify_change(dentry, &newattrs);
222 mutex_unlock(&dentry->d_inode->i_mutex);
226 static long do_sys_truncate(const char __user * path, loff_t length)
229 struct inode * inode;
233 if (length < 0) /* sorry, but loff_t says... */
236 error = user_path_walk(path, &nd);
239 inode = nd.path.dentry->d_inode;
241 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
243 if (S_ISDIR(inode->i_mode))
247 if (!S_ISREG(inode->i_mode))
250 error = mnt_want_write(nd.path.mnt);
254 error = vfs_permission(&nd, MAY_WRITE);
256 goto mnt_drop_write_and_out;
259 if (IS_APPEND(inode))
260 goto mnt_drop_write_and_out;
262 error = get_write_access(inode);
264 goto mnt_drop_write_and_out;
267 * Make sure that there are no leases. get_write_access() protects
268 * against the truncate racing with a lease-granting setlease().
270 error = break_lease(inode, FMODE_WRITE);
272 goto put_write_and_out;
274 error = locks_verify_truncate(inode, NULL, length);
277 error = do_truncate(nd.path.dentry, length, 0, NULL);
281 put_write_access(inode);
282 mnt_drop_write_and_out:
283 mnt_drop_write(nd.path.mnt);
290 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
292 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
293 return do_sys_truncate(path, (long)length);
296 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
298 struct inode * inode;
299 struct dentry *dentry;
311 /* explicitly opened as large or we are on 64-bit box */
312 if (file->f_flags & O_LARGEFILE)
315 dentry = file->f_path.dentry;
316 inode = dentry->d_inode;
318 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
322 /* Cannot ftruncate over 2^31 bytes without large file support */
323 if (small && length > MAX_NON_LFS)
327 if (IS_APPEND(inode))
330 error = locks_verify_truncate(inode, file, length);
332 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
339 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
341 long ret = do_sys_ftruncate(fd, length, 1);
342 /* avoid REGPARM breakage on x86: */
343 asmlinkage_protect(2, ret, fd, length);
347 /* LFS versions of truncate are only needed on 32 bit machines */
348 #if BITS_PER_LONG == 32
349 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
351 return do_sys_truncate(path, length);
354 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
356 long ret = do_sys_ftruncate(fd, length, 0);
357 /* avoid REGPARM breakage on x86: */
358 asmlinkage_protect(2, ret, fd, length);
363 asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
369 if (offset < 0 || len <= 0)
372 /* Return error if mode is not supported */
374 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
381 if (!(file->f_mode & FMODE_WRITE))
384 * Revalidate the write permissions, in case security policy has
385 * changed since the files were opened.
387 ret = security_file_permission(file, MAY_WRITE);
391 inode = file->f_path.dentry->d_inode;
394 if (S_ISFIFO(inode->i_mode))
399 * Let individual file system decide if it supports preallocation
400 * for directories or not.
402 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
406 /* Check for wrap through zero too */
407 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
410 if (inode->i_op && inode->i_op->fallocate)
411 ret = inode->i_op->fallocate(inode, mode, offset, len);
422 * access() needs to use the real uid/gid, not the effective uid/gid.
423 * We do this by temporarily clearing all FS-related capabilities and
424 * switching the fsuid/fsgid around to the real ones.
426 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
429 int old_fsuid, old_fsgid;
430 kernel_cap_t uninitialized_var(old_cap); /* !SECURE_NO_SETUID_FIXUP */
433 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
436 old_fsuid = current->fsuid;
437 old_fsgid = current->fsgid;
439 current->fsuid = current->uid;
440 current->fsgid = current->gid;
442 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
444 * Clear the capabilities if we switch to a non-root user
446 #ifndef CONFIG_SECURITY_FILE_CAPABILITIES
448 * FIXME: There is a race here against sys_capset. The
449 * capabilities can change yet we will restore the old
450 * value below. We should hold task_capabilities_lock,
451 * but we cannot because user_path_walk can sleep.
453 #endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
455 old_cap = cap_set_effective(__cap_empty_set);
457 old_cap = cap_set_effective(current->cap_permitted);
460 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
464 if ((mode & MAY_EXEC) && S_ISREG(nd.path.dentry->d_inode->i_mode)) {
466 * MAY_EXEC on regular files is denied if the fs is mounted
467 * with the "noexec" flag.
470 if (nd.path.mnt->mnt_flags & MNT_NOEXEC)
471 goto out_path_release;
474 res = vfs_permission(&nd, mode | MAY_ACCESS);
475 /* SuS v2 requires we report a read only fs too */
476 if(res || !(mode & S_IWOTH) ||
477 special_file(nd.path.dentry->d_inode->i_mode))
478 goto out_path_release;
480 * This is a rare case where using __mnt_is_readonly()
481 * is OK without a mnt_want/drop_write() pair. Since
482 * no actual write to the fs is performed here, we do
483 * not need to telegraph to that to anyone.
485 * By doing this, we accept that this access is
486 * inherently racy and know that the fs may change
487 * state before we even see this result.
489 if (__mnt_is_readonly(nd.path.mnt))
495 current->fsuid = old_fsuid;
496 current->fsgid = old_fsgid;
498 if (!issecure(SECURE_NO_SETUID_FIXUP))
499 cap_set_effective(old_cap);
504 asmlinkage long sys_access(const char __user *filename, int mode)
506 return sys_faccessat(AT_FDCWD, filename, mode);
509 asmlinkage long sys_chdir(const char __user * filename)
514 error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
518 error = vfs_permission(&nd, MAY_EXEC | MAY_ACCESS);
522 set_fs_pwd(current->fs, &nd.path);
530 asmlinkage long sys_fchdir(unsigned int fd)
541 inode = file->f_path.dentry->d_inode;
544 if (!S_ISDIR(inode->i_mode))
547 error = file_permission(file, MAY_EXEC | MAY_ACCESS);
549 set_fs_pwd(current->fs, &file->f_path);
556 asmlinkage long sys_chroot(const char __user * filename)
561 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &nd);
565 error = vfs_permission(&nd, MAY_EXEC | MAY_ACCESS);
570 if (!capable(CAP_SYS_CHROOT))
573 set_fs_root(current->fs, &nd.path);
581 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
583 struct inode * inode;
584 struct dentry * dentry;
587 struct iattr newattrs;
593 dentry = file->f_path.dentry;
594 inode = dentry->d_inode;
596 audit_inode(NULL, dentry);
598 err = mnt_want_write(file->f_path.mnt);
601 mutex_lock(&inode->i_mutex);
602 if (mode == (mode_t) -1)
603 mode = inode->i_mode;
604 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
605 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
606 err = notify_change(dentry, &newattrs);
607 mutex_unlock(&inode->i_mutex);
608 mnt_drop_write(file->f_path.mnt);
615 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
619 struct inode * inode;
621 struct iattr newattrs;
623 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
626 inode = nd.path.dentry->d_inode;
628 error = mnt_want_write(nd.path.mnt);
631 mutex_lock(&inode->i_mutex);
632 if (mode == (mode_t) -1)
633 mode = inode->i_mode;
634 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
635 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
636 error = notify_change(nd.path.dentry, &newattrs);
637 mutex_unlock(&inode->i_mutex);
638 mnt_drop_write(nd.path.mnt);
645 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
647 return sys_fchmodat(AT_FDCWD, filename, mode);
650 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
652 struct inode *inode = dentry->d_inode;
654 struct iattr newattrs;
656 newattrs.ia_valid = ATTR_CTIME;
657 if (user != (uid_t) -1) {
658 newattrs.ia_valid |= ATTR_UID;
659 newattrs.ia_uid = user;
661 if (group != (gid_t) -1) {
662 newattrs.ia_valid |= ATTR_GID;
663 newattrs.ia_gid = group;
665 if (!S_ISDIR(inode->i_mode))
667 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
668 mutex_lock(&inode->i_mutex);
669 error = notify_change(dentry, &newattrs);
670 mutex_unlock(&inode->i_mutex);
675 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
680 error = user_path_walk(filename, &nd);
683 error = mnt_want_write(nd.path.mnt);
686 error = chown_common(nd.path.dentry, user, group);
687 mnt_drop_write(nd.path.mnt);
694 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
695 gid_t group, int flag)
701 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
704 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
705 error = __user_walk_fd(dfd, filename, follow, &nd);
708 error = mnt_want_write(nd.path.mnt);
711 error = chown_common(nd.path.dentry, user, group);
712 mnt_drop_write(nd.path.mnt);
719 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
724 error = user_path_walk_link(filename, &nd);
727 error = mnt_want_write(nd.path.mnt);
730 error = chown_common(nd.path.dentry, user, group);
731 mnt_drop_write(nd.path.mnt);
739 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
743 struct dentry * dentry;
749 error = mnt_want_write(file->f_path.mnt);
752 dentry = file->f_path.dentry;
753 audit_inode(NULL, dentry);
754 error = chown_common(dentry, user, group);
755 mnt_drop_write(file->f_path.mnt);
763 * You have to be very careful that these write
764 * counts get cleaned up in error cases and
765 * upon __fput(). This should probably never
766 * be called outside of __dentry_open().
768 static inline int __get_file_write_access(struct inode *inode,
769 struct vfsmount *mnt)
772 error = get_write_access(inode);
776 * Do not take mount writer counts on
777 * special files since no writes to
778 * the mount itself will occur.
780 if (!special_file(inode->i_mode)) {
782 * Balanced in __fput()
784 error = mnt_want_write(mnt);
786 put_write_access(inode);
791 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
792 int flags, struct file *f,
793 int (*open)(struct inode *, struct file *))
799 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
800 FMODE_PREAD | FMODE_PWRITE;
801 inode = dentry->d_inode;
802 if (f->f_mode & FMODE_WRITE) {
803 error = __get_file_write_access(inode, mnt);
806 if (!special_file(inode->i_mode))
810 f->f_mapping = inode->i_mapping;
811 f->f_path.dentry = dentry;
814 f->f_op = fops_get(inode->i_fop);
815 file_move(f, &inode->i_sb->s_files);
817 error = security_dentry_open(f);
821 if (!open && f->f_op)
822 open = f->f_op->open;
824 error = open(inode, f);
829 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
831 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
833 /* NB: we're sure to have correct a_ops only after f_op->open */
834 if (f->f_flags & O_DIRECT) {
835 if (!f->f_mapping->a_ops ||
836 ((!f->f_mapping->a_ops->direct_IO) &&
837 (!f->f_mapping->a_ops->get_xip_mem))) {
839 f = ERR_PTR(-EINVAL);
847 if (f->f_mode & FMODE_WRITE) {
848 put_write_access(inode);
849 if (!special_file(inode->i_mode)) {
851 * We don't consider this a real
852 * mnt_want/drop_write() pair
853 * because it all happenend right
854 * here, so just reset the state.
861 f->f_path.dentry = NULL;
862 f->f_path.mnt = NULL;
867 return ERR_PTR(error);
871 * lookup_instantiate_filp - instantiates the open intent filp
872 * @nd: pointer to nameidata
873 * @dentry: pointer to dentry
874 * @open: open callback
876 * Helper for filesystems that want to use lookup open intents and pass back
877 * a fully instantiated struct file to the caller.
878 * This function is meant to be called from within a filesystem's
880 * Beware of calling it for non-regular files! Those ->open methods might block
881 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
882 * leading to a deadlock, as nobody can open that fifo anymore, because
883 * another process to open fifo will block on locked parent when doing lookup).
884 * Note that in case of error, nd->intent.open.file is destroyed, but the
885 * path information remains valid.
886 * If the open callback is set to NULL, then the standard f_op->open()
887 * filesystem callback is substituted.
889 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
890 int (*open)(struct inode *, struct file *))
892 if (IS_ERR(nd->intent.open.file))
896 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
897 nd->intent.open.flags - 1,
898 nd->intent.open.file,
901 return nd->intent.open.file;
903 release_open_intent(nd);
904 nd->intent.open.file = (struct file *)dentry;
907 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
910 * nameidata_to_filp - convert a nameidata to an open filp.
911 * @nd: pointer to nameidata
914 * Note that this function destroys the original nameidata
916 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
920 /* Pick up the filp from the open intent */
921 filp = nd->intent.open.file;
922 /* Has the filesystem initialised the file for us? */
923 if (filp->f_path.dentry == NULL)
924 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
932 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
935 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
941 * We must always pass in a valid mount pointer. Historically
942 * callers got away with not passing it, but we must enforce this at
943 * the earliest possible point now to avoid strange problems deep in the
947 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
949 return ERR_PTR(-EINVAL);
953 f = get_empty_filp();
957 return ERR_PTR(error);
960 return __dentry_open(dentry, mnt, flags, f, NULL);
962 EXPORT_SYMBOL(dentry_open);
965 * Find an empty file descriptor entry, and mark it busy.
967 int get_unused_fd_flags(int flags)
969 struct files_struct * files = current->files;
974 spin_lock(&files->file_lock);
977 fdt = files_fdtable(files);
978 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
982 * N.B. For clone tasks sharing a files structure, this test
983 * will limit the total number of files that can be opened.
985 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
988 /* Do we need to expand the fd array or fd set? */
989 error = expand_files(files, fd);
995 * If we needed to expand the fs array we
996 * might have blocked - try again.
1002 FD_SET(fd, fdt->open_fds);
1003 if (flags & O_CLOEXEC)
1004 FD_SET(fd, fdt->close_on_exec);
1006 FD_CLR(fd, fdt->close_on_exec);
1007 files->next_fd = fd + 1;
1010 if (fdt->fd[fd] != NULL) {
1011 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1018 spin_unlock(&files->file_lock);
1022 int get_unused_fd(void)
1024 return get_unused_fd_flags(0);
1027 EXPORT_SYMBOL(get_unused_fd);
1029 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1031 struct fdtable *fdt = files_fdtable(files);
1032 __FD_CLR(fd, fdt->open_fds);
1033 if (fd < files->next_fd)
1034 files->next_fd = fd;
1037 void put_unused_fd(unsigned int fd)
1039 struct files_struct *files = current->files;
1040 spin_lock(&files->file_lock);
1041 __put_unused_fd(files, fd);
1042 spin_unlock(&files->file_lock);
1045 EXPORT_SYMBOL(put_unused_fd);
1048 * Install a file pointer in the fd array.
1050 * The VFS is full of places where we drop the files lock between
1051 * setting the open_fds bitmap and installing the file in the file
1052 * array. At any such point, we are vulnerable to a dup2() race
1053 * installing a file in the array before us. We need to detect this and
1054 * fput() the struct file we are about to overwrite in this case.
1056 * It should never happen - if we allow dup2() do it, _really_ bad things
1060 void fd_install(unsigned int fd, struct file *file)
1062 struct files_struct *files = current->files;
1063 struct fdtable *fdt;
1064 spin_lock(&files->file_lock);
1065 fdt = files_fdtable(files);
1066 BUG_ON(fdt->fd[fd] != NULL);
1067 rcu_assign_pointer(fdt->fd[fd], file);
1068 spin_unlock(&files->file_lock);
1071 EXPORT_SYMBOL(fd_install);
1073 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1075 char *tmp = getname(filename);
1076 int fd = PTR_ERR(tmp);
1079 fd = get_unused_fd_flags(flags);
1081 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1086 fsnotify_open(f->f_path.dentry);
1095 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1099 if (force_o_largefile())
1100 flags |= O_LARGEFILE;
1102 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1103 /* avoid REGPARM breakage on x86: */
1104 asmlinkage_protect(3, ret, filename, flags, mode);
1108 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1113 if (force_o_largefile())
1114 flags |= O_LARGEFILE;
1116 ret = do_sys_open(dfd, filename, flags, mode);
1117 /* avoid REGPARM breakage on x86: */
1118 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1125 * For backward compatibility? Maybe this should be moved
1126 * into arch/i386 instead?
1128 asmlinkage long sys_creat(const char __user * pathname, int mode)
1130 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1136 * "id" is the POSIX thread ID. We use the
1137 * files pointer for this..
1139 int filp_close(struct file *filp, fl_owner_t id)
1143 if (!file_count(filp)) {
1144 printk(KERN_ERR "VFS: Close: file count is 0\n");
1148 if (filp->f_op && filp->f_op->flush)
1149 retval = filp->f_op->flush(filp, id);
1151 dnotify_flush(filp, id);
1152 locks_remove_posix(filp, id);
1157 EXPORT_SYMBOL(filp_close);
1160 * Careful here! We test whether the file pointer is NULL before
1161 * releasing the fd. This ensures that one clone task can't release
1162 * an fd while another clone is opening it.
1164 asmlinkage long sys_close(unsigned int fd)
1167 struct files_struct *files = current->files;
1168 struct fdtable *fdt;
1171 spin_lock(&files->file_lock);
1172 fdt = files_fdtable(files);
1173 if (fd >= fdt->max_fds)
1178 rcu_assign_pointer(fdt->fd[fd], NULL);
1179 FD_CLR(fd, fdt->close_on_exec);
1180 __put_unused_fd(files, fd);
1181 spin_unlock(&files->file_lock);
1182 retval = filp_close(filp, files);
1184 /* can't restart close syscall because file table entry was cleared */
1185 if (unlikely(retval == -ERESTARTSYS ||
1186 retval == -ERESTARTNOINTR ||
1187 retval == -ERESTARTNOHAND ||
1188 retval == -ERESTART_RESTARTBLOCK))
1194 spin_unlock(&files->file_lock);
1198 EXPORT_SYMBOL(sys_close);
1201 * This routine simulates a hangup on the tty, to arrange that users
1202 * are given clean terminals at login time.
1204 asmlinkage long sys_vhangup(void)
1206 if (capable(CAP_SYS_TTY_CONFIG)) {
1207 /* XXX: this needs locking */
1208 tty_vhangup(current->signal->tty);
1215 * Called when an inode is about to be open.
1216 * We use this to disallow opening large files on 32bit systems if
1217 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1218 * on this flag in sys_open.
1220 int generic_file_open(struct inode * inode, struct file * filp)
1222 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1227 EXPORT_SYMBOL(generic_file_open);
1230 * This is used by subsystems that don't want seekable
1233 int nonseekable_open(struct inode *inode, struct file *filp)
1235 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1239 EXPORT_SYMBOL(nonseekable_open);