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/vs_base.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;
211 down(&dentry->d_inode->i_sem);
212 down_write(&dentry->d_inode->i_alloc_sem);
213 err = notify_change(dentry, &newattrs);
214 up_write(&dentry->d_inode->i_alloc_sem);
215 up(&dentry->d_inode->i_sem);
219 static inline long do_sys_truncate(const char __user * path, loff_t length)
222 struct inode * inode;
226 if (length < 0) /* sorry, but loff_t says... */
229 error = user_path_walk(path, &nd);
232 inode = nd.dentry->d_inode;
234 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
236 if (S_ISDIR(inode->i_mode))
240 if (!S_ISREG(inode->i_mode))
243 error = permission(inode,MAY_WRITE,&nd);
248 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
252 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
256 * Make sure that there are no leases.
258 error = break_lease(inode, FMODE_WRITE);
262 error = get_write_access(inode);
266 error = locks_verify_truncate(inode, NULL, length);
269 error = do_truncate(nd.dentry, length);
271 put_write_access(inode);
279 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
281 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
282 return do_sys_truncate(path, (long)length);
285 static inline long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
287 struct inode * inode;
288 struct dentry *dentry;
300 /* explicitly opened as large or we are on 64-bit box */
301 if (file->f_flags & O_LARGEFILE)
304 dentry = file->f_dentry;
305 inode = dentry->d_inode;
307 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
311 /* Cannot ftruncate over 2^31 bytes without large file support */
312 if (small && length > MAX_NON_LFS)
316 if (IS_APPEND(inode))
319 error = locks_verify_truncate(inode, file, length);
321 error = do_truncate(dentry, length);
328 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
330 return do_sys_ftruncate(fd, length, 1);
333 /* LFS versions of truncate are only needed on 32 bit machines */
334 #if BITS_PER_LONG == 32
335 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
337 return do_sys_truncate(path, length);
340 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
342 return do_sys_ftruncate(fd, length, 0);
346 #ifdef __ARCH_WANT_SYS_UTIME
349 * sys_utime() can be implemented in user-level using sys_utimes().
350 * Is this for backwards compatibility? If so, why not move it
351 * into the appropriate arch directory (for those architectures that
355 /* If times==NULL, set access and modification to current time,
356 * must be owner or have write permission.
357 * Else, update from *times, must be owner or super user.
359 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
363 struct inode * inode;
364 struct iattr newattrs;
366 error = user_path_walk(filename, &nd);
369 inode = nd.dentry->d_inode;
372 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
375 /* Don't worry, the checks are done in inode_change_ok() */
376 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
379 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
382 error = get_user(newattrs.ia_atime.tv_sec, ×->actime);
383 newattrs.ia_atime.tv_nsec = 0;
385 error = get_user(newattrs.ia_mtime.tv_sec, ×->modtime);
386 newattrs.ia_mtime.tv_nsec = 0;
390 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
393 if (IS_IMMUTABLE(inode))
396 if (current->fsuid != inode->i_uid &&
397 (error = permission(inode,MAY_WRITE,&nd)) != 0)
401 error = notify_change(nd.dentry, &newattrs);
411 /* If times==NULL, set access and modification to current time,
412 * must be owner or have write permission.
413 * Else, update from *times, must be owner or super user.
415 long do_utimes(char __user * filename, struct timeval * times)
419 struct inode * inode;
420 struct iattr newattrs;
422 error = user_path_walk(filename, &nd);
426 inode = nd.dentry->d_inode;
429 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
432 /* Don't worry, the checks are done in inode_change_ok() */
433 newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
436 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
439 newattrs.ia_atime.tv_sec = times[0].tv_sec;
440 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
441 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
442 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
443 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
446 if (IS_IMMUTABLE(inode))
449 if (current->fsuid != inode->i_uid &&
450 (error = permission(inode,MAY_WRITE,&nd)) != 0)
454 error = notify_change(nd.dentry, &newattrs);
462 asmlinkage long sys_utimes(char __user * filename, struct timeval __user * utimes)
464 struct timeval times[2];
466 if (utimes && copy_from_user(×, utimes, sizeof(times)))
468 return do_utimes(filename, utimes ? times : NULL);
473 * access() needs to use the real uid/gid, not the effective uid/gid.
474 * We do this by temporarily clearing all FS-related capabilities and
475 * switching the fsuid/fsgid around to the real ones.
477 asmlinkage long sys_access(const char __user * filename, int mode)
480 int old_fsuid, old_fsgid;
481 kernel_cap_t old_cap;
484 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
487 old_fsuid = current->fsuid;
488 old_fsgid = current->fsgid;
489 old_cap = current->cap_effective;
491 current->fsuid = current->uid;
492 current->fsgid = current->gid;
495 * Clear the capabilities if we switch to a non-root user
497 * FIXME: There is a race here against sys_capset. The
498 * capabilities can change yet we will restore the old
499 * value below. We should hold task_capabilities_lock,
500 * but we cannot because user_path_walk can sleep.
503 cap_clear(current->cap_effective);
505 current->cap_effective = current->cap_permitted;
507 res = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
509 res = permission(nd.dentry->d_inode, mode, &nd);
510 /* SuS v2 requires we report a read only fs too */
511 if(!res && (mode & S_IWOTH)
512 && (IS_RDONLY(nd.dentry->d_inode) || MNT_IS_RDONLY(nd.mnt))
513 && !special_file(nd.dentry->d_inode->i_mode))
518 current->fsuid = old_fsuid;
519 current->fsgid = old_fsgid;
520 current->cap_effective = old_cap;
525 asmlinkage long sys_chdir(const char __user * filename)
530 error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
534 error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
538 set_fs_pwd(current->fs, nd.mnt, nd.dentry);
546 EXPORT_SYMBOL_GPL(sys_chdir);
548 asmlinkage long sys_fchdir(unsigned int fd)
551 struct dentry *dentry;
553 struct vfsmount *mnt;
561 dentry = file->f_dentry;
562 mnt = file->f_vfsmnt;
563 inode = dentry->d_inode;
566 if (!S_ISDIR(inode->i_mode))
569 error = permission(inode, MAY_EXEC, NULL);
571 set_fs_pwd(current->fs, mnt, dentry);
578 asmlinkage long sys_chroot(const char __user * filename)
583 error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
587 error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
592 if (!capable(CAP_SYS_CHROOT))
595 set_fs_root(current->fs, nd.mnt, nd.dentry);
604 EXPORT_SYMBOL_GPL(sys_chroot);
606 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
608 struct inode * inode;
609 struct dentry * dentry;
612 struct iattr newattrs;
618 dentry = file->f_dentry;
619 inode = dentry->d_inode;
622 if (IS_RDONLY(inode) || (file && MNT_IS_RDONLY(file->f_vfsmnt)))
625 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
628 if (mode == (mode_t) -1)
629 mode = inode->i_mode;
630 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
631 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
632 err = notify_change(dentry, &newattrs);
641 asmlinkage long sys_chmod(const char __user * filename, mode_t mode)
644 struct inode * inode;
646 struct iattr newattrs;
648 error = user_path_walk(filename, &nd);
651 inode = nd.dentry->d_inode;
654 if (IS_RDONLY(inode) || MNT_IS_RDONLY(nd.mnt))
658 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
662 if (mode == (mode_t) -1)
663 mode = inode->i_mode;
664 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
665 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
666 error = notify_change(nd.dentry, &newattrs);
675 static int chown_common(struct dentry *dentry, struct vfsmount *mnt,
676 uid_t user, gid_t group)
678 struct inode * inode;
680 struct iattr newattrs;
683 if (!(inode = dentry->d_inode)) {
684 printk(KERN_ERR "chown_common: NULL inode\n");
688 if (IS_RDONLY(inode) || MNT_IS_RDONLY(mnt))
691 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
694 newattrs.ia_valid = ATTR_CTIME;
695 if (user != (uid_t) -1) {
696 newattrs.ia_valid |= ATTR_UID;
697 newattrs.ia_uid = vx_map_uid(user);
699 if (group != (gid_t) -1) {
700 newattrs.ia_valid |= ATTR_GID;
701 newattrs.ia_gid = vx_map_gid(group);
703 if (!S_ISDIR(inode->i_mode))
704 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
706 error = notify_change(dentry, &newattrs);
712 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
717 error = user_path_walk(filename, &nd);
719 error = chown_common(nd.dentry, nd.mnt, user, group);
725 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
730 error = user_path_walk_link(filename, &nd);
732 error = chown_common(nd.dentry, nd.mnt, user, group);
739 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
746 error = chown_common(file->f_dentry, file->f_vfsmnt, user, group);
753 * Note that while the flag value (low two bits) for sys_open means:
759 * 00 - no permissions needed
760 * 01 - read-permission
761 * 10 - write-permission
763 * for the internal routines (ie open_namei()/follow_link() etc). 00 is
766 struct file *filp_open(const char * filename, int flags, int mode)
768 int namei_flags, error;
772 if ((namei_flags+1) & O_ACCMODE)
774 if (namei_flags & O_TRUNC)
777 error = open_namei(filename, namei_flags, mode, &nd);
779 return dentry_open(nd.dentry, nd.mnt, flags);
781 return ERR_PTR(error);
784 EXPORT_SYMBOL(filp_open);
786 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
793 f = get_empty_filp();
797 f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
798 inode = dentry->d_inode;
799 if (f->f_mode & FMODE_WRITE) {
800 error = get_write_access(inode);
805 f->f_mapping = inode->i_mapping;
806 f->f_dentry = dentry;
809 f->f_op = fops_get(inode->i_fop);
810 file_move(f, &inode->i_sb->s_files);
812 if (f->f_op && f->f_op->open) {
813 error = f->f_op->open(inode,f);
817 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
819 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
821 /* NB: we're sure to have correct a_ops only after f_op->open */
822 if (f->f_flags & O_DIRECT) {
823 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) {
825 f = ERR_PTR(-EINVAL);
833 if (f->f_mode & FMODE_WRITE)
834 put_write_access(inode);
843 return ERR_PTR(error);
846 EXPORT_SYMBOL(dentry_open);
849 * Find an empty file descriptor entry, and mark it busy.
851 int get_unused_fd(void)
853 struct files_struct * files = current->files;
857 spin_lock(&files->file_lock);
860 fd = find_next_zero_bit(files->open_fds->fds_bits,
865 * N.B. For clone tasks sharing a files structure, this test
866 * will limit the total number of files that can be opened.
868 if (fd >= current->rlim[RLIMIT_NOFILE].rlim_cur)
871 /* Do we need to expand the fdset array? */
872 if (fd >= files->max_fdset) {
873 error = expand_fdset(files, fd);
882 * Check whether we need to expand the fd array.
884 if (fd >= files->max_fds) {
885 error = expand_fd_array(files, fd);
893 FD_SET(fd, files->open_fds);
894 FD_CLR(fd, files->close_on_exec);
895 files->next_fd = fd + 1;
896 // vx_openfd_inc(fd);
899 if (files->fd[fd] != NULL) {
900 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
901 files->fd[fd] = NULL;
907 spin_unlock(&files->file_lock);
911 EXPORT_SYMBOL(get_unused_fd);
913 static inline void __put_unused_fd(struct files_struct *files, unsigned int fd)
915 __FD_CLR(fd, files->open_fds);
916 if (fd < files->next_fd)
918 // vx_openfd_dec(fd);
921 void fastcall put_unused_fd(unsigned int fd)
923 struct files_struct *files = current->files;
924 spin_lock(&files->file_lock);
925 __put_unused_fd(files, fd);
926 spin_unlock(&files->file_lock);
929 EXPORT_SYMBOL(put_unused_fd);
932 * Install a file pointer in the fd array.
934 * The VFS is full of places where we drop the files lock between
935 * setting the open_fds bitmap and installing the file in the file
936 * array. At any such point, we are vulnerable to a dup2() race
937 * installing a file in the array before us. We need to detect this and
938 * fput() the struct file we are about to overwrite in this case.
940 * It should never happen - if we allow dup2() do it, _really_ bad things
944 void fastcall fd_install(unsigned int fd, struct file * file)
946 struct files_struct *files = current->files;
947 spin_lock(&files->file_lock);
948 if (unlikely(files->fd[fd] != NULL))
950 files->fd[fd] = file;
951 spin_unlock(&files->file_lock);
954 EXPORT_SYMBOL(fd_install);
956 asmlinkage long sys_open(const char __user * filename, int flags, int mode)
961 #if BITS_PER_LONG != 32
962 flags |= O_LARGEFILE;
964 tmp = getname(filename);
967 fd = get_unused_fd();
969 struct file *f = filp_open(tmp, flags, mode);
989 * For backward compatibility? Maybe this should be moved
990 * into arch/i386 instead?
992 asmlinkage long sys_creat(const char __user * pathname, int mode)
994 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1000 * "id" is the POSIX thread ID. We use the
1001 * files pointer for this..
1003 int filp_close(struct file *filp, fl_owner_t id)
1007 /* Report and clear outstanding errors */
1008 retval = filp->f_error;
1012 if (!file_count(filp)) {
1013 printk(KERN_ERR "VFS: Close: file count is 0\n");
1017 if (filp->f_op && filp->f_op->flush) {
1018 int err = filp->f_op->flush(filp);
1023 dnotify_flush(filp, id);
1024 locks_remove_posix(filp, id);
1029 EXPORT_SYMBOL(filp_close);
1032 * Careful here! We test whether the file pointer is NULL before
1033 * releasing the fd. This ensures that one clone task can't release
1034 * an fd while another clone is opening it.
1036 asmlinkage long sys_close(unsigned int fd)
1039 struct files_struct *files = current->files;
1041 spin_lock(&files->file_lock);
1042 if (fd >= files->max_fds)
1044 filp = files->fd[fd];
1047 files->fd[fd] = NULL;
1048 FD_CLR(fd, files->close_on_exec);
1049 __put_unused_fd(files, fd);
1050 spin_unlock(&files->file_lock);
1051 return filp_close(filp, files);
1054 spin_unlock(&files->file_lock);
1058 EXPORT_SYMBOL(sys_close);
1061 * This routine simulates a hangup on the tty, to arrange that users
1062 * are given clean terminals at login time.
1064 asmlinkage long sys_vhangup(void)
1066 if (capable(CAP_SYS_TTY_CONFIG)) {
1067 tty_vhangup(current->signal->tty);
1074 * Called when an inode is about to be open.
1075 * We use this to disallow opening large files on 32bit systems if
1076 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1077 * on this flag in sys_open.
1079 int generic_file_open(struct inode * inode, struct file * filp)
1081 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1086 EXPORT_SYMBOL(generic_file_open);
1089 * This is used by subsystems that don't want seekable
1092 int nonseekable_open(struct inode *inode, struct file *filp)
1094 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1098 EXPORT_SYMBOL(nonseekable_open);