4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/utime.h>
10 #include <linux/file.h>
11 #include <linux/smp_lock.h>
12 #include <linux/quotaops.h>
13 #include <linux/dnotify.h>
14 #include <linux/module.h>
15 #include <linux/slab.h>
16 #include <linux/tty.h>
17 #include <linux/namei.h>
18 #include <linux/backing-dev.h>
19 #include <linux/security.h>
20 #include <linux/mount.h>
21 #include <linux/vfs.h>
22 #include <asm/uaccess.h>
24 #include <linux/pagemap.h>
25 #include <linux/syscalls.h>
26 #include <linux/vs_limit.h>
27 #include <linux/vs_dlimit.h>
28 #include <linux/vserver/xid.h>
30 #include <asm/unistd.h>
32 int vfs_statfs(struct super_block *sb, struct kstatfs *buf)
38 if (sb->s_op->statfs) {
39 memset(buf, 0, sizeof(*buf));
40 retval = security_sb_statfs(sb);
43 retval = sb->s_op->statfs(sb, buf);
44 if (retval == 0 && buf->f_frsize == 0)
45 buf->f_frsize = buf->f_bsize;
47 if (!vx_check(0, VX_ADMIN|VX_WATCH))
48 vx_vsi_statfs(sb, buf);
53 EXPORT_SYMBOL(vfs_statfs);
55 static int vfs_statfs_native(struct super_block *sb, struct statfs *buf)
60 retval = vfs_statfs(sb, &st);
64 if (sizeof(*buf) == sizeof(st))
65 memcpy(buf, &st, sizeof(st));
67 if (sizeof buf->f_blocks == 4) {
68 if ((st.f_blocks | st.f_bfree | st.f_bavail) &
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 super_block *sb, struct statfs64 *buf)
103 retval = vfs_statfs(sb, &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.dentry->d_inode->i_sb, &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.dentry->d_inode->i_sb, &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_dentry->d_inode->i_sb, &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_dentry->d_inode->i_sb, &tmp);
193 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
200 int do_truncate(struct dentry *dentry, loff_t length)
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 | ATTR_CTIME;
212 down(&dentry->d_inode->i_sem);
213 err = notify_change(dentry, &newattrs);
214 up(&dentry->d_inode->i_sem);
218 static inline long do_sys_truncate(const char __user * path, loff_t length)
221 struct inode * inode;
225 if (length < 0) /* sorry, but loff_t says... */
228 error = user_path_walk(path, &nd);
231 inode = nd.dentry->d_inode;
233 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
235 if (S_ISDIR(inode->i_mode))
239 if (!S_ISREG(inode->i_mode))
242 error = permission(inode,MAY_WRITE,&nd);
247 if (IS_RDONLY(inode))
251 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
255 * Make sure that there are no leases.
257 error = break_lease(inode, FMODE_WRITE);
261 error = get_write_access(inode);
265 error = locks_verify_truncate(inode, NULL, length);
268 error = do_truncate(nd.dentry, length);
270 put_write_access(inode);
278 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
280 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
281 return do_sys_truncate(path, (long)length);
284 static inline long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
286 struct inode * inode;
287 struct dentry *dentry;
299 /* explicitly opened as large or we are on 64-bit box */
300 if (file->f_flags & O_LARGEFILE)
303 dentry = file->f_dentry;
304 inode = dentry->d_inode;
306 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
310 /* Cannot ftruncate over 2^31 bytes without large file support */
311 if (small && length > MAX_NON_LFS)
315 if (IS_APPEND(inode))
318 error = locks_verify_truncate(inode, file, length);
320 error = do_truncate(dentry, length);
327 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
329 return do_sys_ftruncate(fd, length, 1);
332 /* LFS versions of truncate are only needed on 32 bit machines */
333 #if BITS_PER_LONG == 32
334 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
336 return do_sys_truncate(path, length);
339 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
341 return do_sys_ftruncate(fd, length, 0);
345 #ifdef __ARCH_WANT_SYS_UTIME
348 * sys_utime() can be implemented in user-level using sys_utimes().
349 * Is this for backwards compatibility? If so, why not move it
350 * into the appropriate arch directory (for those architectures that
354 /* If times==NULL, set access and modification to current time,
355 * must be owner or have write permission.
356 * Else, update from *times, must be owner or super user.
358 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
362 struct inode * inode;
363 struct iattr newattrs;
365 error = user_path_walk(filename, &nd);
368 inode = nd.dentry->d_inode;
371 if (IS_RDONLY(inode))
374 /* Don't worry, the checks are done in inode_change_ok() */
375 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
378 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
381 error = get_user(newattrs.ia_atime.tv_sec, ×->actime);
382 newattrs.ia_atime.tv_nsec = 0;
384 error = get_user(newattrs.ia_mtime.tv_sec, ×->modtime);
385 newattrs.ia_mtime.tv_nsec = 0;
389 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
392 if (IS_IMMUTABLE(inode))
395 if (current->fsuid != inode->i_uid &&
396 (error = permission(inode,MAY_WRITE,&nd)) != 0)
400 error = notify_change(nd.dentry, &newattrs);
410 /* If times==NULL, set access and modification to current time,
411 * must be owner or have write permission.
412 * Else, update from *times, must be owner or super user.
414 long do_utimes(char __user * filename, struct timeval * times)
418 struct inode * inode;
419 struct iattr newattrs;
421 error = user_path_walk(filename, &nd);
425 inode = nd.dentry->d_inode;
428 if (IS_RDONLY(inode))
431 /* Don't worry, the checks are done in inode_change_ok() */
432 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
435 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
438 newattrs.ia_atime.tv_sec = times[0].tv_sec;
439 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
440 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
441 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
442 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
445 if (IS_IMMUTABLE(inode))
448 if (current->fsuid != inode->i_uid &&
449 (error = permission(inode,MAY_WRITE,&nd)) != 0)
453 error = notify_change(nd.dentry, &newattrs);
461 asmlinkage long sys_utimes(char __user * filename, struct timeval __user * utimes)
463 struct timeval times[2];
465 if (utimes && copy_from_user(×, utimes, sizeof(times)))
467 return do_utimes(filename, utimes ? times : NULL);
472 * access() needs to use the real uid/gid, not the effective uid/gid.
473 * We do this by temporarily clearing all FS-related capabilities and
474 * switching the fsuid/fsgid around to the real ones.
476 asmlinkage long sys_access(const char __user * filename, int mode)
479 int old_fsuid, old_fsgid;
480 kernel_cap_t old_cap;
483 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
486 old_fsuid = current->fsuid;
487 old_fsgid = current->fsgid;
488 old_cap = current->cap_effective;
490 current->fsuid = current->uid;
491 current->fsgid = current->gid;
494 * Clear the capabilities if we switch to a non-root user
496 * FIXME: There is a race here against sys_capset. The
497 * capabilities can change yet we will restore the old
498 * value below. We should hold task_capabilities_lock,
499 * but we cannot because user_path_walk can sleep.
502 cap_clear(current->cap_effective);
504 current->cap_effective = current->cap_permitted;
506 res = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
508 res = permission(nd.dentry->d_inode, mode, &nd);
509 /* SuS v2 requires we report a read only fs too */
510 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
511 && !special_file(nd.dentry->d_inode->i_mode))
516 current->fsuid = old_fsuid;
517 current->fsgid = old_fsgid;
518 current->cap_effective = old_cap;
523 asmlinkage long sys_chdir(const char __user * filename)
528 error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
532 error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
536 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
544 asmlinkage long sys_fchdir(unsigned int fd)
547 struct dentry *dentry;
549 struct vfsmount *mnt;
557 dentry = file->f_dentry;
558 mnt = file->f_vfsmnt;
559 inode = dentry->d_inode;
562 if (!S_ISDIR(inode->i_mode))
565 error = permission(inode, MAY_EXEC, NULL);
567 set_fs_pwd(current->fs, mnt, dentry);
574 asmlinkage long sys_chroot(const char __user * filename)
579 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
583 error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
588 if (!capable(CAP_SYS_CHROOT))
591 set_fs_root(current->fs, nd.mnt, nd.dentry);
600 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
602 struct inode * inode;
603 struct dentry * dentry;
606 struct iattr newattrs;
612 dentry = file->f_dentry;
613 inode = dentry->d_inode;
616 if (IS_RDONLY(inode))
619 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
622 if (mode == (mode_t) -1)
623 mode = inode->i_mode;
624 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
625 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
626 err = notify_change(dentry, &newattrs);
635 asmlinkage long sys_chmod(const char __user * filename, mode_t mode)
638 struct inode * inode;
640 struct iattr newattrs;
642 error = user_path_walk(filename, &nd);
645 inode = nd.dentry->d_inode;
648 if (IS_RDONLY(inode))
652 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
656 if (mode == (mode_t) -1)
657 mode = inode->i_mode;
658 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
659 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
660 error = notify_change(nd.dentry, &newattrs);
669 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
671 struct inode * inode;
673 struct iattr newattrs;
676 if (!(inode = dentry->d_inode)) {
677 printk(KERN_ERR "chown_common: NULL inode\n");
681 if (IS_RDONLY(inode))
684 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
686 newattrs.ia_valid = ATTR_CTIME;
687 if (user != (uid_t) -1) {
688 newattrs.ia_valid |= ATTR_UID;
689 newattrs.ia_uid = vx_map_uid(user);
691 if (group != (gid_t) -1) {
692 newattrs.ia_valid |= ATTR_GID;
693 newattrs.ia_gid = vx_map_gid(group);
695 if (!S_ISDIR(inode->i_mode))
696 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
698 error = notify_change(dentry, &newattrs);
704 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
709 error = user_path_walk(filename, &nd);
711 error = chown_common(nd.dentry, user, group);
717 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
722 error = user_path_walk_link(filename, &nd);
724 error = chown_common(nd.dentry, user, group);
731 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
738 error = chown_common(file->f_dentry, user, group);
745 * Note that while the flag value (low two bits) for sys_open means:
751 * 00 - no permissions needed
752 * 01 - read-permission
753 * 10 - write-permission
755 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
758 struct file *filp_open(const char * filename, int flags, int mode)
760 int namei_flags, error;
764 if ((namei_flags+1) & O_ACCMODE)
766 if (namei_flags & O_TRUNC)
769 error = open_namei(filename, namei_flags, mode, &nd);
771 return dentry_open(nd.dentry, nd.mnt, flags);
773 return ERR_PTR(error);
776 EXPORT_SYMBOL(filp_open);
778 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
785 f = get_empty_filp();
789 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
790 inode = dentry->d_inode;
791 if (f->f_mode & FMODE_WRITE) {
792 error = get_write_access(inode);
797 f->f_mapping = inode->i_mapping;
798 f->f_dentry = dentry;
801 f->f_op = fops_get(inode->i_fop);
802 file_move(f, &inode->i_sb->s_files);
804 if (f->f_op && f->f_op->open) {
805 error = f->f_op->open(inode,f);
809 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
811 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
813 /* NB: we're sure to have correct a_ops only after f_op->open */
814 if (f->f_flags & O_DIRECT) {
815 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) {
817 f = ERR_PTR(-EINVAL);
825 if (f->f_mode & FMODE_WRITE)
826 put_write_access(inode);
835 return ERR_PTR(error);
838 EXPORT_SYMBOL(dentry_open);
841 * Find an empty file descriptor entry, and mark it busy.
843 int get_unused_fd(void)
845 struct files_struct * files = current->files;
849 spin_lock(&files->file_lock);
852 fd = find_next_zero_bit(files->open_fds->fds_bits,
857 * N.B. For clone tasks sharing a files structure, this test
858 * will limit the total number of files that can be opened.
860 if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
863 /* Do we need to expand the fd array or fd set? */
864 error = expand_files(files, fd);
870 * If we needed to expand the fs array we
871 * might have blocked - try again.
877 FD_SET(fd, files->open_fds);
878 FD_CLR(fd, files->close_on_exec);
879 files->next_fd = fd + 1;
883 if (files->fd[fd] != NULL) {
884 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
885 files->fd[fd] = NULL;
891 spin_unlock(&files->file_lock);
895 EXPORT_SYMBOL(get_unused_fd);
897 static inline void __put_unused_fd(struct files_struct *files, unsigned int fd)
899 __FD_CLR(fd, files->open_fds);
900 if (fd < files->next_fd)
905 void fastcall put_unused_fd(unsigned int fd)
907 struct files_struct *files = current->files;
908 spin_lock(&files->file_lock);
909 __put_unused_fd(files, fd);
910 spin_unlock(&files->file_lock);
913 EXPORT_SYMBOL(put_unused_fd);
916 * Install a file pointer in the fd array.
918 * The VFS is full of places where we drop the files lock between
919 * setting the open_fds bitmap and installing the file in the file
920 * array. At any such point, we are vulnerable to a dup2() race
921 * installing a file in the array before us. We need to detect this and
922 * fput() the struct file we are about to overwrite in this case.
924 * It should never happen - if we allow dup2() do it, _really_ bad things
928 void fastcall fd_install(unsigned int fd, struct file * file)
930 struct files_struct *files = current->files;
931 spin_lock(&files->file_lock);
932 if (unlikely(files->fd[fd] != NULL))
934 files->fd[fd] = file;
935 spin_unlock(&files->file_lock);
938 EXPORT_SYMBOL(fd_install);
940 asmlinkage long sys_open(const char __user * filename, int flags, int mode)
945 #if BITS_PER_LONG != 32
946 flags |= O_LARGEFILE;
948 tmp = getname(filename);
951 fd = get_unused_fd();
953 struct file *f = filp_open(tmp, flags, mode);
969 EXPORT_SYMBOL_GPL(sys_open);
974 * For backward compatibility? Maybe this should be moved
975 * into arch/i386 instead?
977 asmlinkage long sys_creat(const char __user * pathname, int mode)
979 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
985 * "id" is the POSIX thread ID. We use the
986 * files pointer for this..
988 int filp_close(struct file *filp, fl_owner_t id)
992 /* Report and clear outstanding errors */
993 retval = filp->f_error;
997 if (!file_count(filp)) {
998 printk(KERN_ERR "VFS: Close: file count is 0\n");
1002 if (filp->f_op && filp->f_op->flush) {
1003 int err = filp->f_op->flush(filp);
1008 dnotify_flush(filp, id);
1009 locks_remove_posix(filp, id);
1014 EXPORT_SYMBOL(filp_close);
1017 * Careful here! We test whether the file pointer is NULL before
1018 * releasing the fd. This ensures that one clone task can't release
1019 * an fd while another clone is opening it.
1021 asmlinkage long sys_close(unsigned int fd)
1024 struct files_struct *files = current->files;
1026 spin_lock(&files->file_lock);
1027 if (fd >= files->max_fds)
1029 filp = files->fd[fd];
1032 files->fd[fd] = NULL;
1033 FD_CLR(fd, files->close_on_exec);
1034 __put_unused_fd(files, fd);
1035 spin_unlock(&files->file_lock);
1036 return filp_close(filp, files);
1039 spin_unlock(&files->file_lock);
1043 EXPORT_SYMBOL(sys_close);
1046 * This routine simulates a hangup on the tty, to arrange that users
1047 * are given clean terminals at login time.
1049 asmlinkage long sys_vhangup(void)
1051 if (capable(CAP_SYS_TTY_CONFIG)) {
1052 tty_vhangup(current->signal->tty);
1059 * Called when an inode is about to be open.
1060 * We use this to disallow opening large files on 32bit systems if
1061 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1062 * on this flag in sys_open.
1064 int generic_file_open(struct inode * inode, struct file * filp)
1066 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1071 EXPORT_SYMBOL(generic_file_open);
1074 * This is used by subsystems that don't want seekable
1077 int nonseekable_open(struct inode *inode, struct file *filp)
1079 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1083 EXPORT_SYMBOL(nonseekable_open);