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 0xffffffff00000000ULL)
71 * f_files and f_ffree may be -1; it's okay to stuff
74 if (st.f_files != -1 &&
75 (st.f_files & 0xffffffff00000000ULL))
77 if (st.f_ffree != -1 &&
78 (st.f_ffree & 0xffffffff00000000ULL))
82 buf->f_type = st.f_type;
83 buf->f_bsize = st.f_bsize;
84 buf->f_blocks = st.f_blocks;
85 buf->f_bfree = st.f_bfree;
86 buf->f_bavail = st.f_bavail;
87 buf->f_files = st.f_files;
88 buf->f_ffree = st.f_ffree;
89 buf->f_fsid = st.f_fsid;
90 buf->f_namelen = st.f_namelen;
91 buf->f_frsize = st.f_frsize;
92 memset(buf->f_spare, 0, sizeof(buf->f_spare));
97 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
102 retval = vfs_statfs(dentry, &st);
106 if (sizeof(*buf) == sizeof(st))
107 memcpy(buf, &st, sizeof(st));
109 buf->f_type = st.f_type;
110 buf->f_bsize = st.f_bsize;
111 buf->f_blocks = st.f_blocks;
112 buf->f_bfree = st.f_bfree;
113 buf->f_bavail = st.f_bavail;
114 buf->f_files = st.f_files;
115 buf->f_ffree = st.f_ffree;
116 buf->f_fsid = st.f_fsid;
117 buf->f_namelen = st.f_namelen;
118 buf->f_frsize = st.f_frsize;
119 memset(buf->f_spare, 0, sizeof(buf->f_spare));
124 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
129 error = user_path_walk(path, &nd);
132 error = vfs_statfs_native(nd.path.dentry, &tmp);
133 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
141 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
146 if (sz != sizeof(*buf))
148 error = user_path_walk(path, &nd);
151 error = vfs_statfs64(nd.path.dentry, &tmp);
152 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
160 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
170 error = vfs_statfs_native(file->f_path.dentry, &tmp);
171 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
178 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
184 if (sz != sizeof(*buf))
191 error = vfs_statfs64(file->f_path.dentry, &tmp);
192 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
199 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
203 struct iattr newattrs;
205 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
209 newattrs.ia_size = length;
210 newattrs.ia_valid = ATTR_SIZE | time_attrs;
212 newattrs.ia_file = filp;
213 newattrs.ia_valid |= ATTR_FILE;
216 /* Remove suid/sgid on truncate too */
217 newattrs.ia_valid |= should_remove_suid(dentry);
219 mutex_lock(&dentry->d_inode->i_mutex);
220 err = notify_change(dentry, &newattrs);
221 mutex_unlock(&dentry->d_inode->i_mutex);
225 static long do_sys_truncate(const char __user * path, loff_t length)
228 struct inode * inode;
232 if (length < 0) /* sorry, but loff_t says... */
235 error = user_path_walk(path, &nd);
238 inode = nd.path.dentry->d_inode;
240 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
242 if (S_ISDIR(inode->i_mode))
246 if (!S_ISREG(inode->i_mode))
249 error = mnt_want_write(nd.path.mnt);
253 error = vfs_permission(&nd, MAY_WRITE);
255 goto mnt_drop_write_and_out;
258 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
259 goto mnt_drop_write_and_out;
261 error = get_write_access(inode);
263 goto mnt_drop_write_and_out;
266 * Make sure that there are no leases. get_write_access() protects
267 * against the truncate racing with a lease-granting setlease().
269 error = break_lease(inode, FMODE_WRITE);
271 goto put_write_and_out;
273 error = locks_verify_truncate(inode, NULL, length);
276 error = do_truncate(nd.path.dentry, length, 0, NULL);
280 put_write_access(inode);
281 mnt_drop_write_and_out:
282 mnt_drop_write(nd.path.mnt);
289 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
291 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
292 return do_sys_truncate(path, (long)length);
295 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
297 struct inode * inode;
298 struct dentry *dentry;
310 /* explicitly opened as large or we are on 64-bit box */
311 if (file->f_flags & O_LARGEFILE)
314 dentry = file->f_path.dentry;
315 inode = dentry->d_inode;
317 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
321 /* Cannot ftruncate over 2^31 bytes without large file support */
322 if (small && length > MAX_NON_LFS)
326 if (IS_APPEND(inode))
329 error = locks_verify_truncate(inode, file, length);
331 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
338 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
340 long ret = do_sys_ftruncate(fd, length, 1);
341 /* avoid REGPARM breakage on x86: */
342 asmlinkage_protect(2, ret, fd, length);
346 /* LFS versions of truncate are only needed on 32 bit machines */
347 #if BITS_PER_LONG == 32
348 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
350 return do_sys_truncate(path, length);
353 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
355 long ret = do_sys_ftruncate(fd, length, 0);
356 /* avoid REGPARM breakage on x86: */
357 asmlinkage_protect(2, ret, fd, length);
362 asmlinkage long sys_fallocate(int fd, int mode, loff_t offset, loff_t len)
368 if (offset < 0 || len <= 0)
371 /* Return error if mode is not supported */
373 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
380 if (!(file->f_mode & FMODE_WRITE))
383 * Revalidate the write permissions, in case security policy has
384 * changed since the files were opened.
386 ret = security_file_permission(file, MAY_WRITE);
390 inode = file->f_path.dentry->d_inode;
393 if (S_ISFIFO(inode->i_mode))
398 * Let individual file system decide if it supports preallocation
399 * for directories or not.
401 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
405 /* Check for wrap through zero too */
406 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
409 if (inode->i_op && inode->i_op->fallocate)
410 ret = inode->i_op->fallocate(inode, mode, offset, len);
421 * access() needs to use the real uid/gid, not the effective uid/gid.
422 * We do this by temporarily clearing all FS-related capabilities and
423 * switching the fsuid/fsgid around to the real ones.
425 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
428 int old_fsuid, old_fsgid;
429 kernel_cap_t uninitialized_var(old_cap); /* !SECURE_NO_SETUID_FIXUP */
432 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
435 old_fsuid = current->fsuid;
436 old_fsgid = current->fsgid;
438 current->fsuid = current->uid;
439 current->fsgid = current->gid;
441 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
443 * Clear the capabilities if we switch to a non-root user
445 #ifndef CONFIG_SECURITY_FILE_CAPABILITIES
447 * FIXME: There is a race here against sys_capset. The
448 * capabilities can change yet we will restore the old
449 * value below. We should hold task_capabilities_lock,
450 * but we cannot because user_path_walk can sleep.
452 #endif /* ndef CONFIG_SECURITY_FILE_CAPABILITIES */
454 old_cap = cap_set_effective(__cap_empty_set);
456 old_cap = cap_set_effective(current->cap_permitted);
459 res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
463 res = vfs_permission(&nd, mode);
464 /* SuS v2 requires we report a read only fs too */
465 if(res || !(mode & S_IWOTH) ||
466 special_file(nd.path.dentry->d_inode->i_mode))
467 goto out_path_release;
469 * This is a rare case where using __mnt_is_readonly()
470 * is OK without a mnt_want/drop_write() pair. Since
471 * no actual write to the fs is performed here, we do
472 * not need to telegraph to that to anyone.
474 * By doing this, we accept that this access is
475 * inherently racy and know that the fs may change
476 * state before we even see this result.
478 if (__mnt_is_readonly(nd.path.mnt))
484 current->fsuid = old_fsuid;
485 current->fsgid = old_fsgid;
487 if (!issecure(SECURE_NO_SETUID_FIXUP))
488 cap_set_effective(old_cap);
493 asmlinkage long sys_access(const char __user *filename, int mode)
495 return sys_faccessat(AT_FDCWD, filename, mode);
498 asmlinkage long sys_chdir(const char __user * filename)
503 error = __user_walk(filename,
504 LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
508 error = vfs_permission(&nd, MAY_EXEC);
512 set_fs_pwd(current->fs, &nd.path);
520 asmlinkage long sys_fchdir(unsigned int fd)
531 inode = file->f_path.dentry->d_inode;
534 if (!S_ISDIR(inode->i_mode))
537 error = file_permission(file, MAY_EXEC);
539 set_fs_pwd(current->fs, &file->f_path);
546 asmlinkage long sys_chroot(const char __user * filename)
551 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
555 error = vfs_permission(&nd, MAY_EXEC);
560 if (!capable(CAP_SYS_CHROOT))
563 set_fs_root(current->fs, &nd.path);
572 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
574 struct inode * inode;
575 struct dentry * dentry;
578 struct iattr newattrs;
584 dentry = file->f_path.dentry;
585 inode = dentry->d_inode;
587 audit_inode(NULL, dentry);
589 err = mnt_want_write(file->f_path.mnt);
593 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
595 mutex_lock(&inode->i_mutex);
596 if (mode == (mode_t) -1)
597 mode = inode->i_mode;
598 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
599 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
600 err = notify_change(dentry, &newattrs);
601 mutex_unlock(&inode->i_mutex);
604 mnt_drop_write(file->f_path.mnt);
611 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
615 struct inode * inode;
617 struct iattr newattrs;
619 error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
622 inode = nd.path.dentry->d_inode;
624 error = mnt_want_write(nd.path.mnt);
629 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
632 mutex_lock(&inode->i_mutex);
633 if (mode == (mode_t) -1)
634 mode = inode->i_mode;
635 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
636 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
637 error = notify_change(nd.path.dentry, &newattrs);
638 mutex_unlock(&inode->i_mutex);
641 mnt_drop_write(nd.path.mnt);
648 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
650 return sys_fchmodat(AT_FDCWD, filename, mode);
653 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
655 struct inode * inode;
657 struct iattr newattrs;
660 if (!(inode = dentry->d_inode)) {
661 printk(KERN_ERR "chown_common: NULL inode\n");
665 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
667 newattrs.ia_valid = ATTR_CTIME;
668 if (user != (uid_t) -1) {
669 newattrs.ia_valid |= ATTR_UID;
670 newattrs.ia_uid = user;
672 if (group != (gid_t) -1) {
673 newattrs.ia_valid |= ATTR_GID;
674 newattrs.ia_gid = group;
676 if (!S_ISDIR(inode->i_mode))
678 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
679 mutex_lock(&inode->i_mutex);
680 error = notify_change(dentry, &newattrs);
681 mutex_unlock(&inode->i_mutex);
686 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
691 error = user_path_walk(filename, &nd);
694 error = mnt_want_write(nd.path.mnt);
697 error = chown_common(nd.path.dentry, user, group);
698 mnt_drop_write(nd.path.mnt);
705 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
706 gid_t group, int flag)
712 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
715 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
716 error = __user_walk_fd(dfd, filename, follow, &nd);
719 error = mnt_want_write(nd.path.mnt);
722 error = chown_common(nd.path.dentry, user, group);
723 mnt_drop_write(nd.path.mnt);
730 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
735 error = user_path_walk_link(filename, &nd);
738 error = mnt_want_write(nd.path.mnt);
741 error = chown_common(nd.path.dentry, user, group);
742 mnt_drop_write(nd.path.mnt);
750 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
754 struct dentry * dentry;
760 error = mnt_want_write(file->f_path.mnt);
763 dentry = file->f_path.dentry;
764 audit_inode(NULL, dentry);
765 error = chown_common(dentry, user, group);
766 mnt_drop_write(file->f_path.mnt);
774 * You have to be very careful that these write
775 * counts get cleaned up in error cases and
776 * upon __fput(). This should probably never
777 * be called outside of __dentry_open().
779 static inline int __get_file_write_access(struct inode *inode,
780 struct vfsmount *mnt)
783 error = get_write_access(inode);
787 * Do not take mount writer counts on
788 * special files since no writes to
789 * the mount itself will occur.
791 if (!special_file(inode->i_mode)) {
793 * Balanced in __fput()
795 error = mnt_want_write(mnt);
797 put_write_access(inode);
802 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
803 int flags, struct file *f,
804 int (*open)(struct inode *, struct file *))
810 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
811 FMODE_PREAD | FMODE_PWRITE;
812 inode = dentry->d_inode;
813 if (f->f_mode & FMODE_WRITE) {
814 error = __get_file_write_access(inode, mnt);
817 if (!special_file(inode->i_mode))
821 f->f_mapping = inode->i_mapping;
822 f->f_path.dentry = dentry;
825 f->f_op = fops_get(inode->i_fop);
826 file_move(f, &inode->i_sb->s_files);
828 error = security_dentry_open(f);
832 if (!open && f->f_op)
833 open = f->f_op->open;
835 error = open(inode, f);
840 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
842 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
844 /* NB: we're sure to have correct a_ops only after f_op->open */
845 if (f->f_flags & O_DIRECT) {
846 if (!f->f_mapping->a_ops ||
847 ((!f->f_mapping->a_ops->direct_IO) &&
848 (!f->f_mapping->a_ops->get_xip_mem))) {
850 f = ERR_PTR(-EINVAL);
858 if (f->f_mode & FMODE_WRITE) {
859 put_write_access(inode);
860 if (!special_file(inode->i_mode)) {
862 * We don't consider this a real
863 * mnt_want/drop_write() pair
864 * because it all happenend right
865 * here, so just reset the state.
872 f->f_path.dentry = NULL;
873 f->f_path.mnt = NULL;
878 return ERR_PTR(error);
882 * lookup_instantiate_filp - instantiates the open intent filp
883 * @nd: pointer to nameidata
884 * @dentry: pointer to dentry
885 * @open: open callback
887 * Helper for filesystems that want to use lookup open intents and pass back
888 * a fully instantiated struct file to the caller.
889 * This function is meant to be called from within a filesystem's
891 * Beware of calling it for non-regular files! Those ->open methods might block
892 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
893 * leading to a deadlock, as nobody can open that fifo anymore, because
894 * another process to open fifo will block on locked parent when doing lookup).
895 * Note that in case of error, nd->intent.open.file is destroyed, but the
896 * path information remains valid.
897 * If the open callback is set to NULL, then the standard f_op->open()
898 * filesystem callback is substituted.
900 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
901 int (*open)(struct inode *, struct file *))
903 if (IS_ERR(nd->intent.open.file))
907 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
908 nd->intent.open.flags - 1,
909 nd->intent.open.file,
912 return nd->intent.open.file;
914 release_open_intent(nd);
915 nd->intent.open.file = (struct file *)dentry;
918 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
921 * nameidata_to_filp - convert a nameidata to an open filp.
922 * @nd: pointer to nameidata
925 * Note that this function destroys the original nameidata
927 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
931 /* Pick up the filp from the open intent */
932 filp = nd->intent.open.file;
933 /* Has the filesystem initialised the file for us? */
934 if (filp->f_path.dentry == NULL)
935 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
943 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
946 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
952 * We must always pass in a valid mount pointer. Historically
953 * callers got away with not passing it, but we must enforce this at
954 * the earliest possible point now to avoid strange problems deep in the
958 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
960 return ERR_PTR(-EINVAL);
964 f = get_empty_filp();
968 return ERR_PTR(error);
971 return __dentry_open(dentry, mnt, flags, f, NULL);
973 EXPORT_SYMBOL(dentry_open);
976 * Find an empty file descriptor entry, and mark it busy.
978 int get_unused_fd_flags(int flags)
980 struct files_struct * files = current->files;
985 spin_lock(&files->file_lock);
988 fdt = files_fdtable(files);
989 fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
993 * N.B. For clone tasks sharing a files structure, this test
994 * will limit the total number of files that can be opened.
996 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
999 /* Do we need to expand the fd array or fd set? */
1000 error = expand_files(files, fd);
1006 * If we needed to expand the fs array we
1007 * might have blocked - try again.
1013 FD_SET(fd, fdt->open_fds);
1014 if (flags & O_CLOEXEC)
1015 FD_SET(fd, fdt->close_on_exec);
1017 FD_CLR(fd, fdt->close_on_exec);
1018 files->next_fd = fd + 1;
1021 if (fdt->fd[fd] != NULL) {
1022 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1029 spin_unlock(&files->file_lock);
1033 int get_unused_fd(void)
1035 return get_unused_fd_flags(0);
1038 EXPORT_SYMBOL(get_unused_fd);
1040 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1042 struct fdtable *fdt = files_fdtable(files);
1043 __FD_CLR(fd, fdt->open_fds);
1044 if (fd < files->next_fd)
1045 files->next_fd = fd;
1048 void put_unused_fd(unsigned int fd)
1050 struct files_struct *files = current->files;
1051 spin_lock(&files->file_lock);
1052 __put_unused_fd(files, fd);
1053 spin_unlock(&files->file_lock);
1056 EXPORT_SYMBOL(put_unused_fd);
1059 * Install a file pointer in the fd array.
1061 * The VFS is full of places where we drop the files lock between
1062 * setting the open_fds bitmap and installing the file in the file
1063 * array. At any such point, we are vulnerable to a dup2() race
1064 * installing a file in the array before us. We need to detect this and
1065 * fput() the struct file we are about to overwrite in this case.
1067 * It should never happen - if we allow dup2() do it, _really_ bad things
1071 void fd_install(unsigned int fd, struct file *file)
1073 struct files_struct *files = current->files;
1074 struct fdtable *fdt;
1075 spin_lock(&files->file_lock);
1076 fdt = files_fdtable(files);
1077 BUG_ON(fdt->fd[fd] != NULL);
1078 rcu_assign_pointer(fdt->fd[fd], file);
1079 spin_unlock(&files->file_lock);
1082 EXPORT_SYMBOL(fd_install);
1084 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1086 char *tmp = getname(filename);
1087 int fd = PTR_ERR(tmp);
1090 fd = get_unused_fd_flags(flags);
1092 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1097 fsnotify_open(f->f_path.dentry);
1106 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1110 if (force_o_largefile())
1111 flags |= O_LARGEFILE;
1113 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1114 /* avoid REGPARM breakage on x86: */
1115 asmlinkage_protect(3, ret, filename, flags, mode);
1119 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1124 if (force_o_largefile())
1125 flags |= O_LARGEFILE;
1127 ret = do_sys_open(dfd, filename, flags, mode);
1128 /* avoid REGPARM breakage on x86: */
1129 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1136 * For backward compatibility? Maybe this should be moved
1137 * into arch/i386 instead?
1139 asmlinkage long sys_creat(const char __user * pathname, int mode)
1141 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1147 * "id" is the POSIX thread ID. We use the
1148 * files pointer for this..
1150 int filp_close(struct file *filp, fl_owner_t id)
1154 if (!file_count(filp)) {
1155 printk(KERN_ERR "VFS: Close: file count is 0\n");
1159 if (filp->f_op && filp->f_op->flush)
1160 retval = filp->f_op->flush(filp, id);
1162 dnotify_flush(filp, id);
1163 locks_remove_posix(filp, id);
1168 EXPORT_SYMBOL(filp_close);
1171 * Careful here! We test whether the file pointer is NULL before
1172 * releasing the fd. This ensures that one clone task can't release
1173 * an fd while another clone is opening it.
1175 asmlinkage long sys_close(unsigned int fd)
1178 struct files_struct *files = current->files;
1179 struct fdtable *fdt;
1182 spin_lock(&files->file_lock);
1183 fdt = files_fdtable(files);
1184 if (fd >= fdt->max_fds)
1189 rcu_assign_pointer(fdt->fd[fd], NULL);
1190 FD_CLR(fd, fdt->close_on_exec);
1191 __put_unused_fd(files, fd);
1192 spin_unlock(&files->file_lock);
1193 retval = filp_close(filp, files);
1195 /* can't restart close syscall because file table entry was cleared */
1196 if (unlikely(retval == -ERESTARTSYS ||
1197 retval == -ERESTARTNOINTR ||
1198 retval == -ERESTARTNOHAND ||
1199 retval == -ERESTART_RESTARTBLOCK))
1205 spin_unlock(&files->file_lock);
1209 EXPORT_SYMBOL(sys_close);
1212 * This routine simulates a hangup on the tty, to arrange that users
1213 * are given clean terminals at login time.
1215 asmlinkage long sys_vhangup(void)
1217 if (capable(CAP_SYS_TTY_CONFIG)) {
1218 /* XXX: this needs locking */
1219 tty_vhangup(current->signal->tty);
1226 * Called when an inode is about to be open.
1227 * We use this to disallow opening large files on 32bit systems if
1228 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1229 * on this flag in sys_open.
1231 int generic_file_open(struct inode * inode, struct file * filp)
1233 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1238 EXPORT_SYMBOL(generic_file_open);
1241 * This is used by subsystems that don't want seekable
1244 int nonseekable_open(struct inode *inode, struct file *filp)
1246 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1250 EXPORT_SYMBOL(nonseekable_open);