[dragonfly.git] / sys / kern / vfs_syscalls.c

/*
 * Copyright (c) 1989, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)vfs_syscalls.c	8.13 (Berkeley) 4/15/94
 * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $
 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.115 2007/05/06 19:23:31 dillon Exp $
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/sysent.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mountctl.h>
#include <sys/sysproto.h>
#include <sys/filedesc.h>
#include <sys/kernel.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/linker.h>
#include <sys/stat.h>
#include <sys/unistd.h>
#include <sys/vnode.h>
#include <sys/proc.h>
#include <sys/namei.h>
#include <sys/nlookup.h>
#include <sys/dirent.h>
#include <sys/extattr.h>
#include <sys/spinlock.h>
#include <sys/kern_syscall.h>
#include <sys/objcache.h>
#include <sys/sysctl.h>
#include <sys/file2.h>
#include <sys/spinlock2.h>

#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>

#include <machine/limits.h>
#include <machine/stdarg.h>

#include <vfs/union/union.h>

static void mount_warning(struct mount *mp, const char *ctl, ...);
static int mount_path(struct proc *p, struct mount *mp, char **rb, char **fb);
static int checkvp_chdir (struct vnode *vn, struct thread *td);
static void checkdirs (struct nchandle *old_nch, struct nchandle *new_nch);
static int chroot_refuse_vdir_fds (struct filedesc *fdp);
static int chroot_visible_mnt(struct mount *mp, struct proc *p);
static int getutimes (const struct timeval *, struct timespec *);
static int setfown (struct vnode *, uid_t, gid_t);
static int setfmode (struct vnode *, int);
static int setfflags (struct vnode *, int);
static int setutimes (struct vnode *, const struct timespec *, int);
static int	usermount = 0;	/* if 1, non-root can mount fs. */

int (*union_dircheckp) (struct thread *, struct vnode **, struct file *);

SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");

/*
 * Virtual File System System Calls
 */

/*
 * Mount a file system.
 */
/*
 * mount_args(char *type, char *path, int flags, caddr_t data)
 */
/* ARGSUSED */
int
sys_mount(struct mount_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct nchandle nch;
	struct mount *mp;
	struct vfsconf *vfsp;
	int error, flag = 0, flag2 = 0;
	int hasmount;
	struct vattr va;
	struct nlookupdata nd;
	char fstypename[MFSNAMELEN];
	struct ucred *cred = p->p_ucred;

	KKASSERT(p);
	if (cred->cr_prison != NULL)
		return (EPERM);
	if (usermount == 0 && (error = suser(td)))
		return (error);
	/*
	 * Do not allow NFS export by non-root users.
	 */
	if (uap->flags & MNT_EXPORTED) {
		error = suser(td);
		if (error)
			return (error);
	}
	/*
	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users
	 */
	if (suser(td)) 
		uap->flags |= MNT_NOSUID | MNT_NODEV;

	/*
	 * Lookup the requested path and extract the nch and vnode.
	 */
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0) {
		if ((error = nlookup(&nd)) == 0) {
			if (nd.nl_nch.ncp->nc_vp == NULL)
				error = ENOENT;
		}
	}
	if (error) {
		nlookup_done(&nd);
		return (error);
	}

	/*
	 * Extract the locked+refd ncp and cleanup the nd structure
	 */
	nch = nd.nl_nch;
	cache_zero(&nd.nl_nch);
	nlookup_done(&nd);

	if ((nch.ncp->nc_flag & NCF_ISMOUNTPT) && cache_findmount(&nch))
		hasmount = 1;
	else
		hasmount = 0;


	/*
	 * now we have the locked ref'd nch and unreferenced vnode.
	 */
	vp = nch.ncp->nc_vp;
	if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
		cache_put(&nch);
		return (error);
	}
	cache_unlock(&nch);

	/*
	 * Now we have an unlocked ref'd nch and a locked ref'd vp
	 */
	if (uap->flags & MNT_UPDATE) {
		if ((vp->v_flag & VROOT) == 0) {
			cache_drop(&nch);
			vput(vp);
			return (EINVAL);
		}
		mp = vp->v_mount;
		flag = mp->mnt_flag;
		flag2 = mp->mnt_kern_flag;
		/*
		 * We only allow the filesystem to be reloaded if it
		 * is currently mounted read-only.
		 */
		if ((uap->flags & MNT_RELOAD) &&
		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
			cache_drop(&nch);
			vput(vp);
			return (EOPNOTSUPP);	/* Needs translation */
		}
		/*
		 * Only root, or the user that did the original mount is
		 * permitted to update it.
		 */
		if (mp->mnt_stat.f_owner != cred->cr_uid &&
		    (error = suser(td))) {
			cache_drop(&nch);
			vput(vp);
			return (error);
		}
		if (vfs_busy(mp, LK_NOWAIT)) {
			cache_drop(&nch);
			vput(vp);
			return (EBUSY);
		}
		if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
			cache_drop(&nch);
			vfs_unbusy(mp);
			vput(vp);
			return (EBUSY);
		}
		vp->v_flag |= VMOUNT;
		mp->mnt_flag |=
		    uap->flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
		vn_unlock(vp);
		goto update;
	}
	/*
	 * If the user is not root, ensure that they own the directory
	 * onto which we are attempting to mount.
	 */
	if ((error = VOP_GETATTR(vp, &va)) ||
	    (va.va_uid != cred->cr_uid && (error = suser(td)))) {
		cache_drop(&nch);
		vput(vp);
		return (error);
	}
	if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
		cache_drop(&nch);
		vput(vp);
		return (error);
	}
	if (vp->v_type != VDIR) {
		cache_drop(&nch);
		vput(vp);
		return (ENOTDIR);
	}
	if ((error = copyinstr(uap->type, fstypename, MFSNAMELEN, NULL)) != 0) {
		cache_drop(&nch);
		vput(vp);
		return (error);
	}
	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
		if (!strcmp(vfsp->vfc_name, fstypename))
			break;
	}
	if (vfsp == NULL) {
		linker_file_t lf;

		/* Only load modules for root (very important!) */
		if ((error = suser(td)) != 0) {
			cache_drop(&nch);
			vput(vp);
			return error;
		}
		error = linker_load_file(fstypename, &lf);
		if (error || lf == NULL) {
			cache_drop(&nch);
			vput(vp);
			if (lf == NULL)
				error = ENODEV;
			return error;
		}
		lf->userrefs++;
		/* lookup again, see if the VFS was loaded */
		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
			if (!strcmp(vfsp->vfc_name, fstypename))
				break;
		}
		if (vfsp == NULL) {
			lf->userrefs--;
			linker_file_unload(lf);
			cache_drop(&nch);
			vput(vp);
			return (ENODEV);
		}
	}
	if ((vp->v_flag & VMOUNT) != 0 || hasmount) {
		cache_drop(&nch);
		vput(vp);
		return (EBUSY);
	}
	vp->v_flag |= VMOUNT;

	/*
	 * Allocate and initialize the filesystem.
	 */
	mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
	TAILQ_INIT(&mp->mnt_nvnodelist);
	TAILQ_INIT(&mp->mnt_reservedvnlist);
	TAILQ_INIT(&mp->mnt_jlist);
	mp->mnt_nvnodelistsize = 0;
	lockinit(&mp->mnt_lock, "vfslock", 0, 0);
	vfs_busy(mp, LK_NOWAIT);
	mp->mnt_op = vfsp->vfc_vfsops;
	mp->mnt_vfc = vfsp;
	vfsp->vfc_refcount++;
	mp->mnt_stat.f_type = vfsp->vfc_typenum;
	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
	mp->mnt_stat.f_owner = cred->cr_uid;
	mp->mnt_iosize_max = DFLTPHYS;
	vn_unlock(vp);
update:
	/*
	 * Set the mount level flags.
	 */
	if (uap->flags & MNT_RDONLY)
		mp->mnt_flag |= MNT_RDONLY;
	else if (mp->mnt_flag & MNT_RDONLY)
		mp->mnt_kern_flag |= MNTK_WANTRDWR;
	mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME |
	    MNT_NOSYMFOLLOW | MNT_IGNORE |
	    MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
	mp->mnt_flag |= uap->flags & (MNT_NOSUID | MNT_NOEXEC |
	    MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE |
	    MNT_NOSYMFOLLOW | MNT_IGNORE |
	    MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR);
	/*
	 * Mount the filesystem.
	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
	 * get. 
	 */
	error = VFS_MOUNT(mp, uap->path, uap->data, cred);
	if (mp->mnt_flag & MNT_UPDATE) {
		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
			mp->mnt_flag &= ~MNT_RDONLY;
		mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE);
		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
		if (error) {
			mp->mnt_flag = flag;
			mp->mnt_kern_flag = flag2;
		}
		vfs_unbusy(mp);
		vp->v_flag &= ~VMOUNT;
		vrele(vp);
		cache_drop(&nch);
		return (error);
	}
	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
	/*
	 * Put the new filesystem on the mount list after root.  The mount
	 * point gets its own mnt_ncmountpt (unless the VFS already set one
	 * up) which represents the root of the mount.  The lookup code
	 * detects the mount point going forward and checks the root of
	 * the mount going backwards.
	 *
	 * It is not necessary to invalidate or purge the vnode underneath
	 * because elements under the mount will be given their own glue
	 * namecache record.
	 */
	if (!error) {
		if (mp->mnt_ncmountpt.ncp == NULL) {
			/* 
			 * allocate, then unlock, but leave the ref intact 
			 */
			cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
			cache_unlock(&mp->mnt_ncmountpt);
		}
		mp->mnt_ncmounton = nch;		/* inherits ref */
		nch.ncp->nc_flag |= NCF_ISMOUNTPT;

		/* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
		vp->v_flag &= ~VMOUNT;
		mountlist_insert(mp, MNTINS_LAST);
		checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
		vn_unlock(vp);
		error = vfs_allocate_syncvnode(mp);
		vfs_unbusy(mp);
		error = VFS_START(mp, 0);
		vrele(vp);
	} else {
		vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
		vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
		vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
		vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
		vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
		vp->v_flag &= ~VMOUNT;
		mp->mnt_vfc->vfc_refcount--;
		vfs_unbusy(mp);
		kfree(mp, M_MOUNT);
		cache_drop(&nch);
		vput(vp);
	}
	return (error);
}

/*
 * Scan all active processes to see if any of them have a current
 * or root directory onto which the new filesystem has just been
 * mounted. If so, replace them with the new mount point.
 *
 * The passed ncp is ref'd and locked (from the mount code) and
 * must be associated with the vnode representing the root of the
 * mount point.
 */
struct checkdirs_info {
	struct nchandle old_nch;
	struct nchandle new_nch;
	struct vnode *old_vp;
	struct vnode *new_vp;
};

static int checkdirs_callback(struct proc *p, void *data);

static void
checkdirs(struct nchandle *old_nch, struct nchandle *new_nch)
{
	struct checkdirs_info info;
	struct vnode *olddp;
	struct vnode *newdp;
	struct mount *mp;

	/*
	 * If the old mount point's vnode has a usecount of 1, it is not
	 * being held as a descriptor anywhere.
	 */
	olddp = old_nch->ncp->nc_vp;
	if (olddp == NULL || olddp->v_sysref.refcnt == 1)
		return;

	/*
	 * Force the root vnode of the new mount point to be resolved
	 * so we can update any matching processes.
	 */
	mp = new_nch->mount;
	if (VFS_ROOT(mp, &newdp))
		panic("mount: lost mount");
	cache_setunresolved(new_nch);
	cache_setvp(new_nch, newdp);

	/*
	 * Special handling of the root node
	 */
	if (rootvnode == olddp) {
		vref(newdp);
		vfs_cache_setroot(newdp, cache_hold(new_nch));
	}

	/*
	 * Pass newdp separately so the callback does not have to access
	 * it via new_nch->ncp->nc_vp.
	 */
	info.old_nch = *old_nch;
	info.new_nch = *new_nch;
	info.new_vp = newdp;
	allproc_scan(checkdirs_callback, &info);
	vput(newdp);
}

/*
 * NOTE: callback is not MP safe because the scanned process's filedesc
 * structure can be ripped out from under us, amoung other things.
 */
static int
checkdirs_callback(struct proc *p, void *data)
{
	struct checkdirs_info *info = data;
	struct filedesc *fdp;
	struct nchandle ncdrop1;
	struct nchandle ncdrop2;
	struct vnode *vprele1;
	struct vnode *vprele2;

	if ((fdp = p->p_fd) != NULL) {
		cache_zero(&ncdrop1);
		cache_zero(&ncdrop2);
		vprele1 = NULL;
		vprele2 = NULL;

		/*
		 * MPUNSAFE - XXX fdp can be pulled out from under a
		 * foreign process.
		 *
		 * A shared filedesc is ok, we don't have to copy it
		 * because we are making this change globally.
		 */
		spin_lock_wr(&fdp->fd_spin);
		if (fdp->fd_ncdir.mount == info->old_nch.mount &&
		    fdp->fd_ncdir.ncp == info->old_nch.ncp) {
			vprele1 = fdp->fd_cdir;
			vref(info->new_vp);
			fdp->fd_cdir = info->new_vp;
			ncdrop1 = fdp->fd_ncdir;
			cache_copy(&info->new_nch, &fdp->fd_ncdir);
		}
		if (fdp->fd_nrdir.mount == info->old_nch.mount &&
		    fdp->fd_nrdir.ncp == info->old_nch.ncp) {
			vprele2 = fdp->fd_rdir;
			vref(info->new_vp);
			fdp->fd_rdir = info->new_vp;
			ncdrop2 = fdp->fd_nrdir;
			cache_copy(&info->new_nch, &fdp->fd_nrdir);
		}
		spin_unlock_wr(&fdp->fd_spin);
		if (ncdrop1.ncp)
			cache_drop(&ncdrop1);
		if (ncdrop2.ncp)
			cache_drop(&ncdrop2);
		if (vprele1)
			vrele(vprele1);
		if (vprele2)
			vrele(vprele2);
	}
	return(0);
}

/*
 * Unmount a file system.
 *
 * Note: unmount takes a path to the vnode mounted on as argument,
 * not special file (as before).
 */
/*
 * umount_args(char *path, int flags)
 */
/* ARGSUSED */
int
sys_unmount(struct unmount_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct mount *mp = NULL;
	int error;
	struct nlookupdata nd;

	KKASSERT(p);
	if (p->p_ucred->cr_prison != NULL)
		return (EPERM);
	if (usermount == 0 && (error = suser(td)))
		return (error);

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = nlookup(&nd);
	if (error)
		goto out;

	mp = nd.nl_nch.mount;

	/*
	 * Only root, or the user that did the original mount is
	 * permitted to unmount this filesystem.
	 */
	if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) &&
	    (error = suser(td)))
		goto out;

	/*
	 * Don't allow unmounting the root file system.
	 */
	if (mp->mnt_flag & MNT_ROOTFS) {
		error = EINVAL;
		goto out;
	}

	/*
	 * Must be the root of the filesystem
	 */
	if (nd.nl_nch.ncp != mp->mnt_ncmountpt.ncp) {
		error = EINVAL;
		goto out;
	}

out:
	nlookup_done(&nd);
	if (error)
		return (error);
	return (dounmount(mp, uap->flags));
}

/*
 * Do the actual file system unmount.
 */
static int
dounmount_interlock(struct mount *mp)
{
	if (mp->mnt_kern_flag & MNTK_UNMOUNT)
		return (EBUSY);
	mp->mnt_kern_flag |= MNTK_UNMOUNT;
	return(0);
}

int
dounmount(struct mount *mp, int flags)
{
	struct namecache *ncp;
	struct nchandle nch;
	int error;
	int async_flag;
	int lflags;
	int freeok = 1;

	/*
	 * Exclusive access for unmounting purposes
	 */
	if ((error = mountlist_interlock(dounmount_interlock, mp)) != 0)
		return (error);

	/*
	 * Allow filesystems to detect that a forced unmount is in progress.
	 */
	if (flags & MNT_FORCE)
		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
	lflags = LK_EXCLUSIVE | ((flags & MNT_FORCE) ? 0 : LK_NOWAIT);
	error = lockmgr(&mp->mnt_lock, lflags);
	if (error) {
		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
		if (mp->mnt_kern_flag & MNTK_MWAIT)
			wakeup(mp);
		return (error);
	}

	if (mp->mnt_flag & MNT_EXPUBLIC)
		vfs_setpublicfs(NULL, NULL, NULL);

	vfs_msync(mp, MNT_WAIT);
	async_flag = mp->mnt_flag & MNT_ASYNC;
	mp->mnt_flag &=~ MNT_ASYNC;

	/*
	 * If this filesystem isn't aliasing other filesystems,
	 * try to invalidate any remaining namecache entries and
	 * check the count afterwords.
	 */
	if ((mp->mnt_kern_flag & MNTK_NCALIASED) == 0) {
		cache_lock(&mp->mnt_ncmountpt);
		cache_inval(&mp->mnt_ncmountpt, CINV_DESTROY|CINV_CHILDREN);
		cache_unlock(&mp->mnt_ncmountpt);

		if ((ncp = mp->mnt_ncmountpt.ncp) != NULL &&
		    (ncp->nc_refs != 1 || TAILQ_FIRST(&ncp->nc_list))) {

			if ((flags & MNT_FORCE) == 0) {
				error = EBUSY;
				mount_warning(mp, "Cannot unmount: "
						  "%d namecache "
						  "references still "
						  "present",
						  ncp->nc_refs - 1);
			} else {
				mount_warning(mp, "Forced unmount: "
						  "%d namecache "
						  "references still "
						  "present",
						  ncp->nc_refs - 1);
				freeok = 0;
			}
		}
	}

	/*
	 * nchandle records ref the mount structure.  Expect a count of 1
	 * (our mount->mnt_ncmountpt).
	 */
	if (mp->mnt_refs != 1) {
		if ((flags & MNT_FORCE) == 0) {
			mount_warning(mp, "Cannot unmount: "
					  "%d process references still "
					  "present", mp->mnt_refs);
			error = EBUSY;
		} else {
			mount_warning(mp, "Forced unmount: "
					  "%d process references still "
					  "present", mp->mnt_refs);
			freeok = 0;
		}
	}

	if (error == 0) {
		if (mp->mnt_syncer != NULL)
			vrele(mp->mnt_syncer);
		if (((mp->mnt_flag & MNT_RDONLY) ||
		     (error = VFS_SYNC(mp, MNT_WAIT)) == 0) ||
		    (flags & MNT_FORCE)) {
			error = VFS_UNMOUNT(mp, flags);
		}
	}
	if (error) {
		if (mp->mnt_syncer == NULL)
			vfs_allocate_syncvnode(mp);
		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
		mp->mnt_flag |= async_flag;
		lockmgr(&mp->mnt_lock, LK_RELEASE);
		if (mp->mnt_kern_flag & MNTK_MWAIT)
			wakeup(mp);
		return (error);
	}
	/*
	 * Clean up any journals still associated with the mount after
	 * filesystem activity has ceased.
	 */
	journal_remove_all_journals(mp, 
	    ((flags & MNT_FORCE) ? MC_JOURNAL_STOP_IMM : 0));

	mountlist_remove(mp);

	/*
	 * Remove any installed vnode ops here so the individual VFSs don't
	 * have to.
	 */
	vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
	vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
	vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
	vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
	vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);

	if (mp->mnt_ncmountpt.ncp != NULL) {
		nch = mp->mnt_ncmountpt;
		cache_zero(&mp->mnt_ncmountpt);
		cache_clrmountpt(&nch);
		cache_drop(&nch);
	}
	if (mp->mnt_ncmounton.ncp != NULL) {
		nch = mp->mnt_ncmounton;
		cache_zero(&mp->mnt_ncmounton);
		cache_clrmountpt(&nch);
		cache_drop(&nch);
	}

	mp->mnt_vfc->vfc_refcount--;
	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
		panic("unmount: dangling vnode");
	lockmgr(&mp->mnt_lock, LK_RELEASE);
	if (mp->mnt_kern_flag & MNTK_MWAIT)
		wakeup(mp);
	if (freeok)
		kfree(mp, M_MOUNT);
	return (0);
}

static
void
mount_warning(struct mount *mp, const char *ctl, ...)
{
	char *ptr;
	char *buf;
	__va_list va;

	__va_start(va, ctl);
	if (cache_fullpath(NULL, &mp->mnt_ncmounton, &ptr, &buf) == 0) {
		kprintf("unmount(%s): ", ptr);
		kvprintf(ctl, va);
		kprintf("\n");
		kfree(buf, M_TEMP);
	} else {
		kprintf("unmount(%p): ", mp);
		kvprintf(ctl, va);
		kprintf("\n");
	}
	__va_end(va);
}

/*
 * Shim cache_fullpath() to handle the case where a process is chrooted into
 * a subdirectory of a mount.  In this case if the root mount matches the
 * process root directory's mount we have to specify the process's root
 * directory instead of the mount point, because the mount point might
 * be above the root directory.
 */
static
int
mount_path(struct proc *p, struct mount *mp, char **rb, char **fb)
{
	struct nchandle *nch;

	if (p && p->p_fd->fd_nrdir.mount == mp)
		nch = &p->p_fd->fd_nrdir;
	else
		nch = &mp->mnt_ncmountpt;
	return(cache_fullpath(p, nch, rb, fb));
}

/*
 * Sync each mounted filesystem.
 */

#ifdef DEBUG
static int syncprt = 0;
SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
#endif /* DEBUG */

static int sync_callback(struct mount *mp, void *data);

/* ARGSUSED */
int
sys_sync(struct sync_args *uap)
{
	mountlist_scan(sync_callback, NULL, MNTSCAN_FORWARD);
#ifdef DEBUG
	/*
	 * print out buffer pool stat information on each sync() call.
	 */
	if (syncprt)
		vfs_bufstats();
#endif /* DEBUG */
	return (0);
}

static
int
sync_callback(struct mount *mp, void *data __unused)
{
	int asyncflag;

	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
		asyncflag = mp->mnt_flag & MNT_ASYNC;
		mp->mnt_flag &= ~MNT_ASYNC;
		vfs_msync(mp, MNT_NOWAIT);
		VFS_SYNC(mp, MNT_NOWAIT);
		mp->mnt_flag |= asyncflag;
	}
	return(0);
}

/* XXX PRISON: could be per prison flag */
static int prison_quotas;
#if 0
SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
#endif

/*
 *  quotactl_args(char *path, int fcmd, int uid, caddr_t arg)
 *
 * Change filesystem quotas.
 */
/* ARGSUSED */
int
sys_quotactl(struct quotactl_args *uap)
{
	struct nlookupdata nd;
	struct thread *td;
	struct proc *p;
	struct mount *mp;
	int error;

	td = curthread;
	p = td->td_proc;
	if (p->p_ucred->cr_prison && !prison_quotas)
		return (EPERM);

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = nlookup(&nd);
	if (error == 0) {
		mp = nd.nl_nch.mount;
		error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
				    uap->arg, nd.nl_cred);
	}
	nlookup_done(&nd);
	return (error);
}

/*
 * mountctl(char *path, int op, int fd, const void *ctl, int ctllen,
 *		void *buf, int buflen)
 *
 * This function operates on a mount point and executes the specified
 * operation using the specified control data, and possibly returns data.
 *
 * The actual number of bytes stored in the result buffer is returned, 0
 * if none, otherwise an error is returned.
 */
/* ARGSUSED */
int
sys_mountctl(struct mountctl_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	void *ctl = NULL;
	void *buf = NULL;
	char *path = NULL;
	int error;

	/*
	 * Sanity and permissions checks.  We must be root.
	 */
	KKASSERT(p);
	if (p->p_ucred->cr_prison != NULL)
		return (EPERM);
	if ((error = suser(td)) != 0)
		return (error);

	/*
	 * Argument length checks
	 */
	if (uap->ctllen < 0 || uap->ctllen > 1024)
		return (EINVAL);
	if (uap->buflen < 0 || uap->buflen > 16 * 1024)
		return (EINVAL);
	if (uap->path == NULL)
		return (EINVAL);

	/*
	 * Allocate the necessary buffers and copyin data
	 */
	path = objcache_get(namei_oc, M_WAITOK);
	error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
	if (error)
		goto done;

	if (uap->ctllen) {
		ctl = kmalloc(uap->ctllen + 1, M_TEMP, M_WAITOK|M_ZERO);
		error = copyin(uap->ctl, ctl, uap->ctllen);
		if (error)
			goto done;
	}
	if (uap->buflen)
		buf = kmalloc(uap->buflen + 1, M_TEMP, M_WAITOK|M_ZERO);

	/*
	 * Validate the descriptor
	 */
	fp = holdfp(p->p_fd, uap->fd, -1);
	if (fp == NULL) {
		error = EBADF;
		goto done;
	}

	/*
	 * Execute the internal kernel function and clean up.
	 */
	error = kern_mountctl(path, uap->op, fp, ctl, uap->ctllen, buf, uap->buflen, &uap->sysmsg_result);
	if (fp)
		fdrop(fp);
	if (error == 0 && uap->sysmsg_result > 0)
		error = copyout(buf, uap->buf, uap->sysmsg_result);
done:
	if (path)
		objcache_put(namei_oc, path);
	if (ctl)
		kfree(ctl, M_TEMP);
	if (buf)
		kfree(buf, M_TEMP);
	return (error);
}

/*
 * Execute a mount control operation by resolving the path to a mount point
 * and calling vop_mountctl().  
 */
int
kern_mountctl(const char *path, int op, struct file *fp, 
		const void *ctl, int ctllen, 
		void *buf, int buflen, int *res)
{
	struct vnode *vp;
	struct mount *mp;
	struct nlookupdata nd;
	int error;

	*res = 0;
	vp = NULL;
	error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
	if (error == 0)
		error = nlookup(&nd);
	if (error == 0)
		error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
	nlookup_done(&nd);
	if (error)
		return (error);

	mp = vp->v_mount;

	/*
	 * Must be the root of the filesystem
	 */
	if ((vp->v_flag & VROOT) == 0) {
		vput(vp);
		return (EINVAL);
	}
	error = vop_mountctl(mp->mnt_vn_use_ops, op, fp, ctl, ctllen, 
				buf, buflen, res);
	vput(vp);
	return (error);
}

int
kern_statfs(struct nlookupdata *nd, struct statfs *buf)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct mount *mp;
	struct statfs *sp;
	char *fullpath, *freepath;
	int error;

	if ((error = nlookup(nd)) != 0)
		return (error);
	mp = nd->nl_nch.mount;
	sp = &mp->mnt_stat;
	if ((error = VFS_STATFS(mp, sp, nd->nl_cred)) != 0)
		return (error);

	error = mount_path(p, mp, &fullpath, &freepath);
	if (error)
		return(error);
	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
	strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
	kfree(freepath, M_TEMP);

	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
	bcopy(sp, buf, sizeof(*buf));
	/* Only root should have access to the fsid's. */
	if (suser(td))
		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
	return (0);
}

/*
 * statfs_args(char *path, struct statfs *buf)
 *
 * Get filesystem statistics.
 */
int
sys_statfs(struct statfs_args *uap)
{
	struct nlookupdata nd;
	struct statfs buf;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_statfs(&nd, &buf);
	nlookup_done(&nd);
	if (error == 0)
		error = copyout(&buf, uap->buf, sizeof(*uap->buf));
	return (error);
}

int
kern_fstatfs(int fd, struct statfs *buf)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	struct mount *mp;
	struct statfs *sp;
	char *fullpath, *freepath;
	int error;

	KKASSERT(p);
	if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
		return (error);
	mp = ((struct vnode *)fp->f_data)->v_mount;
	if (mp == NULL) {
		error = EBADF;
		goto done;
	}
	if (fp->f_cred == NULL) {
		error = EINVAL;
		goto done;
	}
	sp = &mp->mnt_stat;
	if ((error = VFS_STATFS(mp, sp, fp->f_cred)) != 0)
		goto done;

	if ((error = mount_path(p, mp, &fullpath, &freepath)) != 0)
		goto done;
	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
	strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
	kfree(freepath, M_TEMP);

	sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
	bcopy(sp, buf, sizeof(*buf));

	/* Only root should have access to the fsid's. */
	if (suser(td))
		buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0;
	error = 0;
done:
	fdrop(fp);
	return (error);
}

/*
 * fstatfs_args(int fd, struct statfs *buf)
 *
 * Get filesystem statistics.
 */
int
sys_fstatfs(struct fstatfs_args *uap)
{
	struct statfs buf;
	int error;

	error = kern_fstatfs(uap->fd, &buf);

	if (error == 0)
		error = copyout(&buf, uap->buf, sizeof(*uap->buf));
	return (error);
}

/*
 * getfsstat_args(struct statfs *buf, long bufsize, int flags)
 *
 * Get statistics on all filesystems.
 */

struct getfsstat_info {
	struct statfs *sfsp;
	long count;
	long maxcount;
	int error;
	int flags;
	struct proc *p;
};

static int getfsstat_callback(struct mount *, void *);

/* ARGSUSED */
int
sys_getfsstat(struct getfsstat_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct getfsstat_info info;

	bzero(&info, sizeof(info));

	info.maxcount = uap->bufsize / sizeof(struct statfs);
	info.sfsp = uap->buf;
	info.count = 0;
	info.flags = uap->flags;
	info.p = p;

	mountlist_scan(getfsstat_callback, &info, MNTSCAN_FORWARD);
	if (info.sfsp && info.count > info.maxcount)
		uap->sysmsg_result = info.maxcount;
	else
		uap->sysmsg_result = info.count;
	return (info.error);
}

static int
getfsstat_callback(struct mount *mp, void *data)
{
	struct getfsstat_info *info = data;
	struct statfs *sp;
	char *freepath;
	char *fullpath;
	int error;

	if (info->sfsp && info->count < info->maxcount) {
		if (info->p && !chroot_visible_mnt(mp, info->p))
			return(0);
		sp = &mp->mnt_stat;

		/*
		 * If MNT_NOWAIT or MNT_LAZY is specified, do not
		 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
		 * overrides MNT_WAIT.
		 */
		if (((info->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
		    (info->flags & MNT_WAIT)) &&
		    (error = VFS_STATFS(mp, sp, info->p->p_ucred))) {
			return(0);
		}
		sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;

		error = mount_path(info->p, mp, &fullpath, &freepath);
		if (error) {
			info->error = error;
			return(-1);
		}
		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
		strlcpy(sp->f_mntonname, fullpath, sizeof(sp->f_mntonname));
		kfree(freepath, M_TEMP);

		error = copyout(sp, info->sfsp, sizeof(*sp));
		if (error) {
			info->error = error;
			return (-1);
		}
		++info->sfsp;
	}
	info->count++;
	return(0);
}

/*
 * fchdir_args(int fd)
 *
 * Change current working directory to a given file descriptor.
 */
/* ARGSUSED */
int
sys_fchdir(struct fchdir_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct filedesc *fdp = p->p_fd;
	struct vnode *vp, *ovp;
	struct mount *mp;
	struct file *fp;
	struct nchandle nch, onch, tnch;
	int error;

	if ((error = holdvnode(fdp, uap->fd, &fp)) != 0)
		return (error);
	vp = (struct vnode *)fp->f_data;
	vref(vp);
	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
	if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL)
		error = ENOTDIR;
	else
		error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
	if (error) {
		vput(vp);
		fdrop(fp);
		return (error);
	}
	cache_copy(&fp->f_nchandle, &nch);

	/*
	 * If the ncp has become a mount point, traverse through
	 * the mount point.
	 */

	while (!error && (nch.ncp->nc_flag & NCF_ISMOUNTPT) &&
	       (mp = cache_findmount(&nch)) != NULL
	) {
		error = nlookup_mp(mp, &tnch);
		if (error == 0) {
			cache_unlock(&tnch);	/* leave ref intact */
			vput(vp);
			vp = tnch.ncp->nc_vp;
			error = vget(vp, LK_SHARED);
			KKASSERT(error == 0);
			cache_drop(&nch);
			nch = tnch;
		}
	}
	if (error == 0) {
		ovp = fdp->fd_cdir;
		onch = fdp->fd_ncdir;
		vn_unlock(vp);		/* leave ref intact */
		fdp->fd_cdir = vp;
		fdp->fd_ncdir = nch;
		cache_drop(&onch);
		vrele(ovp);
	} else {
		cache_drop(&nch);
		vput(vp);
	}
	fdrop(fp);
	return (error);
}

int
kern_chdir(struct nlookupdata *nd)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct filedesc *fdp = p->p_fd;
	struct vnode *vp, *ovp;
	struct nchandle onch;
	int error;

	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
		return (ENOENT);
	if ((error = vget(vp, LK_SHARED)) != 0)
		return (error);

	error = checkvp_chdir(vp, td);
	vn_unlock(vp);
	if (error == 0) {
		ovp = fdp->fd_cdir;
		onch = fdp->fd_ncdir;
		cache_unlock(&nd->nl_nch);	/* leave reference intact */
		fdp->fd_ncdir = nd->nl_nch;
		fdp->fd_cdir = vp;
		cache_drop(&onch);
		vrele(ovp);
		cache_zero(&nd->nl_nch);
	} else {
		vrele(vp);
	}
	return (error);
}

/*
 * chdir_args(char *path)
 *
 * Change current working directory (``.'').
 */
int
sys_chdir(struct chdir_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_chdir(&nd);
	nlookup_done(&nd);
	return (error);
}

/*
 * Helper function for raised chroot(2) security function:  Refuse if
 * any filedescriptors are open directories.
 */
static int
chroot_refuse_vdir_fds(struct filedesc *fdp)
{
	struct vnode *vp;
	struct file *fp;
	int error;
	int fd;

	for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
		if ((error = holdvnode(fdp, fd, &fp)) != 0)
			continue;
		vp = (struct vnode *)fp->f_data;
		if (vp->v_type != VDIR) {
			fdrop(fp);
			continue;
		}
		fdrop(fp);
		return(EPERM);
	}
	return (0);
}

/*
 * This sysctl determines if we will allow a process to chroot(2) if it
 * has a directory open:
 *	0: disallowed for all processes.
 *	1: allowed for processes that were not already chroot(2)'ed.
 *	2: allowed for all processes.
 */

static int chroot_allow_open_directories = 1;

SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
     &chroot_allow_open_directories, 0, "");

/*
 * chroot to the specified namecache entry.  We obtain the vp from the
 * namecache data.  The passed ncp must be locked and referenced and will
 * remain locked and referenced on return.
 */
int
kern_chroot(struct nchandle *nch)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct filedesc *fdp = p->p_fd;
	struct vnode *vp;
	int error;

	/*
	 * Only root can chroot
	 */
	if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)
		return (error);

	/*
	 * Disallow open directory descriptors (fchdir() breakouts).
	 */
	if (chroot_allow_open_directories == 0 ||
	   (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
		if ((error = chroot_refuse_vdir_fds(fdp)) != 0)
			return (error);
	}
	if ((vp = nch->ncp->nc_vp) == NULL)
		return (ENOENT);

	if ((error = vget(vp, LK_SHARED)) != 0)
		return (error);

	/*
	 * Check the validity of vp as a directory to change to and 
	 * associate it with rdir/jdir.
	 */
	error = checkvp_chdir(vp, td);
	vn_unlock(vp);			/* leave reference intact */
	if (error == 0) {
		vrele(fdp->fd_rdir);
		fdp->fd_rdir = vp;	/* reference inherited by fd_rdir */
		cache_drop(&fdp->fd_nrdir);
		cache_copy(nch, &fdp->fd_nrdir);
		if (fdp->fd_jdir == NULL) {
			fdp->fd_jdir = vp;
			vref(fdp->fd_jdir);
			cache_copy(nch, &fdp->fd_njdir);
		}
	} else {
		vrele(vp);
	}
	return (error);
}

/*
 * chroot_args(char *path)
 *
 * Change notion of root (``/'') directory.
 */
/* ARGSUSED */
int
sys_chroot(struct chroot_args *uap)
{
	struct thread *td = curthread;
	struct nlookupdata nd;
	int error;

	KKASSERT(td->td_proc);
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error) {
		nlookup_done(&nd);
		return(error);
	}
	error = nlookup(&nd);
	if (error == 0)
		error = kern_chroot(&nd.nl_nch);
	nlookup_done(&nd);
	return(error);
}

/*
 * Common routine for chroot and chdir.  Given a locked, referenced vnode,
 * determine whether it is legal to chdir to the vnode.  The vnode's state
 * is not changed by this call.
 */
int
checkvp_chdir(struct vnode *vp, struct thread *td)
{
	int error;

	if (vp->v_type != VDIR)
		error = ENOTDIR;
	else
		error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred);
	return (error);
}

int
kern_open(struct nlookupdata *nd, int oflags, int mode, int *res)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct lwp *lp = td->td_lwp;
	struct filedesc *fdp = p->p_fd;
	int cmode, flags;
	struct file *nfp;
	struct file *fp;
	struct vnode *vp;
	int type, indx, error;
	struct flock lf;

	if ((oflags & O_ACCMODE) == O_ACCMODE)
		return (EINVAL);
	flags = FFLAGS(oflags);
	error = falloc(p, &nfp, NULL);
	if (error)
		return (error);
	fp = nfp;
	cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;

	/*
	 * XXX p_dupfd is a real mess.  It allows a device to return a
	 * file descriptor to be duplicated rather then doing the open
	 * itself.
	 */
	lp->lwp_dupfd = -1;

	/*
	 * Call vn_open() to do the lookup and assign the vnode to the 
	 * file pointer.  vn_open() does not change the ref count on fp
	 * and the vnode, on success, will be inherited by the file pointer
	 * and unlocked.
	 */
	nd->nl_flags |= NLC_LOCKVP;
	error = vn_open(nd, fp, flags, cmode);
	nlookup_done(nd);
	if (error) {
		/*
		 * handle special fdopen() case.  bleh.  dupfdopen() is
		 * responsible for dropping the old contents of ofiles[indx]
		 * if it succeeds.
		 *
		 * Note that fsetfd() will add a ref to fp which represents
		 * the fd_files[] assignment.  We must still drop our
		 * reference.
		 */
		if ((error == ENODEV || error == ENXIO) && lp->lwp_dupfd >= 0) {
			if (fdalloc(p, 0, &indx) == 0) {
				error = dupfdopen(p, indx, lp->lwp_dupfd, flags, error);
				if (error == 0) {
					*res = indx;
					fdrop(fp);	/* our ref */
					return (0);
				}
				fsetfd(p, NULL, indx);
			}
		}
		fdrop(fp);	/* our ref */
		if (error == ERESTART)
			error = EINTR;
		return (error);
	}

	/*
	 * ref the vnode for ourselves so it can't be ripped out from under
	 * is.  XXX need an ND flag to request that the vnode be returned
	 * anyway.
	 *
	 * Reserve a file descriptor but do not assign it until the open
	 * succeeds.
	 */
	vp = (struct vnode *)fp->f_data;
	vref(vp);
	if ((error = fdalloc(p, 0, &indx)) != 0) {
		fdrop(fp);
		vrele(vp);
		return (error);
	}

	/*
	 * If no error occurs the vp will have been assigned to the file
	 * pointer.
	 */
	lp->lwp_dupfd = 0;

	if (flags & (O_EXLOCK | O_SHLOCK)) {
		lf.l_whence = SEEK_SET;
		lf.l_start = 0;
		lf.l_len = 0;
		if (flags & O_EXLOCK)
			lf.l_type = F_WRLCK;
		else
			lf.l_type = F_RDLCK;
		if (flags & FNONBLOCK)
			type = 0;
		else
			type = F_WAIT;

		if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) {
			/*
			 * lock request failed.  Clean up the reserved
			 * descriptor.
			 */
			vrele(vp);
			fsetfd(p, NULL, indx);
			fdrop(fp);
			return (error);
		}
		fp->f_flag |= FHASLOCK;
	}
#if 0
	/*
	 * Assert that all regular file vnodes were created with a object.
	 */
	KASSERT(vp->v_type != VREG || vp->v_object != NULL,
		("open: regular file has no backing object after vn_open"));
#endif

	vrele(vp);

	/*
	 * release our private reference, leaving the one associated with the
	 * descriptor table intact.
	 */
	fsetfd(p, fp, indx);
	fdrop(fp);
	*res = indx;
	return (0);
}

/*
 * open_args(char *path, int flags, int mode)
 *
 * Check permissions, allocate an open file structure,
 * and call the device open routine if any.
 */
int
sys_open(struct open_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0) {
		error = kern_open(&nd, uap->flags,
				    uap->mode, &uap->sysmsg_result);
	}
	nlookup_done(&nd);
	return (error);
}

int
kern_mknod(struct nlookupdata *nd, int mode, int dev)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct vattr vattr;
	int error;
	int whiteout = 0;

	KKASSERT(p);

	switch (mode & S_IFMT) {
	case S_IFCHR:
	case S_IFBLK:
		error = suser(td);
		break;
	default:
		error = suser_cred(p->p_ucred, PRISON_ROOT);
		break;
	}
	if (error)
		return (error);

	bwillwrite();
	nd->nl_flags |= NLC_CREATE;
	if ((error = nlookup(nd)) != 0)
		return (error);
	if (nd->nl_nch.ncp->nc_vp)
		return (EEXIST);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);

	VATTR_NULL(&vattr);
	vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
	vattr.va_rdev = dev;
	whiteout = 0;

	switch (mode & S_IFMT) {
	case S_IFMT:	/* used by badsect to flag bad sectors */
		vattr.va_type = VBAD;
		break;
	case S_IFCHR:
		vattr.va_type = VCHR;
		break;
	case S_IFBLK:
		vattr.va_type = VBLK;
		break;
	case S_IFWHT:
		whiteout = 1;
		break;
	default:
		error = EINVAL;
		break;
	}
	if (error == 0) {
		if (whiteout) {
			error = VOP_NWHITEOUT(&nd->nl_nch, nd->nl_cred, NAMEI_CREATE);
		} else {
			vp = NULL;
			error = VOP_NMKNOD(&nd->nl_nch, &vp, nd->nl_cred, &vattr);
			if (error == 0)
				vput(vp);
		}
	}
	return (error);
}

/*
 * mknod_args(char *path, int mode, int dev)
 *
 * Create a special file.
 */
int
sys_mknod(struct mknod_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_mknod(&nd, uap->mode, uap->dev);
	nlookup_done(&nd);
	return (error);
}

int
kern_mkfifo(struct nlookupdata *nd, int mode)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vattr vattr;
	struct vnode *vp;
	int error;

	bwillwrite();

	nd->nl_flags |= NLC_CREATE;
	if ((error = nlookup(nd)) != 0)
		return (error);
	if (nd->nl_nch.ncp->nc_vp)
		return (EEXIST);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);

	VATTR_NULL(&vattr);
	vattr.va_type = VFIFO;
	vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask;
	vp = NULL;
	error = VOP_NMKNOD(&nd->nl_nch, &vp, nd->nl_cred, &vattr);
	if (error == 0)
		vput(vp);
	return (error);
}

/*
 * mkfifo_args(char *path, int mode)
 *
 * Create a named pipe.
 */
int
sys_mkfifo(struct mkfifo_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_mkfifo(&nd, uap->mode);
	nlookup_done(&nd);
	return (error);
}

static int hardlink_check_uid = 0;
SYSCTL_INT(_kern, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
    &hardlink_check_uid, 0, 
    "Unprivileged processes cannot create hard links to files owned by other "
    "users");
static int hardlink_check_gid = 0;
SYSCTL_INT(_kern, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
    &hardlink_check_gid, 0,
    "Unprivileged processes cannot create hard links to files owned by other "
    "groups");

static int
can_hardlink(struct vnode *vp, struct thread *td, struct ucred *cred)
{
	struct vattr va;
	int error;

	/*
	 * Shortcut if disabled
	 */
	if (hardlink_check_uid == 0 && hardlink_check_gid == 0)
		return (0);

	/*
	 * root cred can always hardlink
	 */
	if (suser_cred(cred, PRISON_ROOT) == 0)
		return (0);

	/*
	 * Otherwise only if the originating file is owned by the
	 * same user or group.  Note that any group is allowed if
	 * the file is owned by the caller.
	 */
	error = VOP_GETATTR(vp, &va);
	if (error != 0)
		return (error);
	
	if (hardlink_check_uid) {
		if (cred->cr_uid != va.va_uid)
			return (EPERM);
	}
	
	if (hardlink_check_gid) {
		if (cred->cr_uid != va.va_uid && !groupmember(va.va_gid, cred))
			return (EPERM);
	}

	return (0);
}

int
kern_link(struct nlookupdata *nd, struct nlookupdata *linknd)
{
	struct thread *td = curthread;
	struct vnode *vp;
	int error;

	/*
	 * Lookup the source and obtained a locked vnode.
	 *
	 * XXX relookup on vget failure / race ?
	 */
	bwillwrite();
	if ((error = nlookup(nd)) != 0)
		return (error);
	vp = nd->nl_nch.ncp->nc_vp;
	KKASSERT(vp != NULL);
	if (vp->v_type == VDIR)
		return (EPERM);		/* POSIX */
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);
	if ((error = vget(vp, LK_EXCLUSIVE)) != 0)
		return (error);

	/*
	 * Unlock the source so we can lookup the target without deadlocking
	 * (XXX vp is locked already, possible other deadlock?).  The target
	 * must not exist.
	 */
	KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
	nd->nl_flags &= ~NLC_NCPISLOCKED;
	cache_unlock(&nd->nl_nch);

	linknd->nl_flags |= NLC_CREATE;
	if ((error = nlookup(linknd)) != 0) {
		vput(vp);
		return (error);
	}
	if (linknd->nl_nch.ncp->nc_vp) {
		vput(vp);
		return (EEXIST);
	}

	/*
	 * Finally run the new API VOP.
	 */
	error = can_hardlink(vp, td, td->td_proc->p_ucred);
	if (error == 0)
		error = VOP_NLINK(&linknd->nl_nch, vp, linknd->nl_cred);
	vput(vp);
	return (error);
}

/*
 * link_args(char *path, char *link)
 *
 * Make a hard file link.
 */
int
sys_link(struct link_args *uap)
{
	struct nlookupdata nd, linknd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0) {
		error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0);
		if (error == 0)
			error = kern_link(&nd, &linknd);
		nlookup_done(&linknd);
	}
	nlookup_done(&nd);
	return (error);
}

int
kern_symlink(struct nlookupdata *nd, char *path, int mode)
{
	struct vattr vattr;
	struct vnode *vp;
	int error;

	bwillwrite();
	nd->nl_flags |= NLC_CREATE;
	if ((error = nlookup(nd)) != 0)
		return (error);
	if (nd->nl_nch.ncp->nc_vp)
		return (EEXIST);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);
	VATTR_NULL(&vattr);
	vattr.va_mode = mode;
	error = VOP_NSYMLINK(&nd->nl_nch, &vp, nd->nl_cred, &vattr, path);
	if (error == 0)
		vput(vp);
	return (error);
}

/*
 * symlink(char *path, char *link)
 *
 * Make a symbolic link.
 */
int
sys_symlink(struct symlink_args *uap)
{
	struct thread *td = curthread;
	struct nlookupdata nd;
	char *path;
	int error;
	int mode;

	path = objcache_get(namei_oc, M_WAITOK);
	error = copyinstr(uap->path, path, MAXPATHLEN, NULL);
	if (error == 0) {
		error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0);
		if (error == 0) {
			mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask;
			error = kern_symlink(&nd, path, mode);
		}
		nlookup_done(&nd);
	}
	objcache_put(namei_oc, path);
	return (error);
}

/*
 * undelete_args(char *path)
 *
 * Delete a whiteout from the filesystem.
 */
/* ARGSUSED */
int
sys_undelete(struct undelete_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	bwillwrite();
	nd.nl_flags |= NLC_DELETE;
	if (error == 0)
		error = nlookup(&nd);
	if (error == 0)
		error = ncp_writechk(&nd.nl_nch);
	if (error == 0)
		error = VOP_NWHITEOUT(&nd.nl_nch, nd.nl_cred, NAMEI_DELETE);
	nlookup_done(&nd);
	return (error);
}

int
kern_unlink(struct nlookupdata *nd)
{
	int error;

	bwillwrite();
	nd->nl_flags |= NLC_DELETE;
	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);
	error = VOP_NREMOVE(&nd->nl_nch, nd->nl_cred);
	return (error);
}

/*
 * unlink_args(char *path)
 *
 * Delete a name from the filesystem.
 */
int
sys_unlink(struct unlink_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_unlink(&nd);
	nlookup_done(&nd);
	return (error);
}

int
kern_lseek(int fd, off_t offset, int whence, off_t *res)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	struct vattr vattr;
	int error;

	fp = holdfp(p->p_fd, fd, -1);
	if (fp == NULL)
		return (EBADF);
	if (fp->f_type != DTYPE_VNODE) {
		error = ESPIPE;
		goto done;
	}

	switch (whence) {
	case L_INCR:
		fp->f_offset += offset;
		error = 0;
		break;
	case L_XTND:
		error = VOP_GETATTR((struct vnode *)fp->f_data, &vattr);
		if (error == 0)
			fp->f_offset = offset + vattr.va_size;
		break;
	case L_SET:
		fp->f_offset = offset;
		error = 0;
		break;
	default:
		error = EINVAL;
		break;
	}
	*res = fp->f_offset;
done:
	fdrop(fp);
	return (error);
}

/*
 * lseek_args(int fd, int pad, off_t offset, int whence)
 *
 * Reposition read/write file offset.
 */
int
sys_lseek(struct lseek_args *uap)
{
	int error;

	error = kern_lseek(uap->fd, uap->offset, uap->whence,
	    &uap->sysmsg_offset);

	return (error);
}

int
kern_access(struct nlookupdata *nd, int aflags)
{
	struct vnode *vp;
	int error, flags;

	if ((error = nlookup(nd)) != 0)
		return (error);
retry:
	error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
	if (error)
		return (error);

	/* Flags == 0 means only check for existence. */
	if (aflags) {
		flags = 0;
		if (aflags & R_OK)
			flags |= VREAD;
		if (aflags & W_OK)
			flags |= VWRITE;
		if (aflags & X_OK)
			flags |= VEXEC;
		if ((flags & VWRITE) == 0 || 
		    (error = vn_writechk(vp, &nd->nl_nch)) == 0)
			error = VOP_ACCESS(vp, flags, nd->nl_cred);

		/*
		 * If the file handle is stale we have to re-resolve the
		 * entry.  This is a hack at the moment.
		 */
		if (error == ESTALE) {
			cache_setunresolved(&nd->nl_nch);
			error = cache_resolve(&nd->nl_nch, nd->nl_cred);
			if (error == 0) {
				vput(vp);
				vp = NULL;
				goto retry;
			}
		}
	}
	vput(vp);
	return (error);
}

/*
 * access_args(char *path, int flags)
 *
 * Check access permissions.
 */
int
sys_access(struct access_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_access(&nd, uap->flags);
	nlookup_done(&nd);
	return (error);
}

int
kern_stat(struct nlookupdata *nd, struct stat *st)
{
	int error;
	struct vnode *vp;
	thread_t td;

	if ((error = nlookup(nd)) != 0)
		return (error);
again:
	if ((vp = nd->nl_nch.ncp->nc_vp) == NULL)
		return (ENOENT);

	td = curthread;
	if ((error = vget(vp, LK_SHARED)) != 0)
		return (error);
	error = vn_stat(vp, st, nd->nl_cred);

	/*
	 * If the file handle is stale we have to re-resolve the entry.  This
	 * is a hack at the moment.
	 */
	if (error == ESTALE) {
		cache_setunresolved(&nd->nl_nch);
		error = cache_resolve(&nd->nl_nch, nd->nl_cred);
		if (error == 0) {
			vput(vp);
			goto again;
		}
	}
	vput(vp);
	return (error);
}

/*
 * stat_args(char *path, struct stat *ub)
 *
 * Get file status; this version follows links.
 */
int
sys_stat(struct stat_args *uap)
{
	struct nlookupdata nd;
	struct stat st;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0) {
		error = kern_stat(&nd, &st);
		if (error == 0)
			error = copyout(&st, uap->ub, sizeof(*uap->ub));
	}
	nlookup_done(&nd);
	return (error);
}

/*
 * lstat_args(char *path, struct stat *ub)
 *
 * Get file status; this version does not follow links.
 */
int
sys_lstat(struct lstat_args *uap)
{
	struct nlookupdata nd;
	struct stat st;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0) {
		error = kern_stat(&nd, &st);
		if (error == 0)
			error = copyout(&st, uap->ub, sizeof(*uap->ub));
	}
	nlookup_done(&nd);
	return (error);
}

/*
 * pathconf_Args(char *path, int name)
 *
 * Get configurable pathname variables.
 */
/* ARGSUSED */
int
sys_pathconf(struct pathconf_args *uap)
{
	struct nlookupdata nd;
	struct vnode *vp;
	int error;

	vp = NULL;
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = nlookup(&nd);
	if (error == 0)
		error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &vp);
	nlookup_done(&nd);
	if (error == 0) {
		error = VOP_PATHCONF(vp, uap->name, uap->sysmsg_fds);
		vput(vp);
	}
	return (error);
}

/*
 * XXX: daver
 * kern_readlink isn't properly split yet.  There is a copyin burried
 * in VOP_READLINK().
 */
int
kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct iovec aiov;
	struct uio auio;
	int error;

	if ((error = nlookup(nd)) != 0)
		return (error);
	error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &vp);
	if (error)
		return (error);
	if (vp->v_type != VLNK) {
		error = EINVAL;
	} else {
		aiov.iov_base = buf;
		aiov.iov_len = count;
		auio.uio_iov = &aiov;
		auio.uio_iovcnt = 1;
		auio.uio_offset = 0;
		auio.uio_rw = UIO_READ;
		auio.uio_segflg = UIO_USERSPACE;
		auio.uio_td = td;
		auio.uio_resid = count;
		error = VOP_READLINK(vp, &auio, p->p_ucred);
	}
	vput(vp);
	*res = count - auio.uio_resid;
	return (error);
}

/*
 * readlink_args(char *path, char *buf, int count)
 *
 * Return target name of a symbolic link.
 */
int
sys_readlink(struct readlink_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0) {
		error = kern_readlink(&nd, uap->buf, uap->count,
					&uap->sysmsg_result);
	}
	nlookup_done(&nd);
	return (error);
}

static int
setfflags(struct vnode *vp, int flags)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	int error;
	struct vattr vattr;

	/*
	 * Prevent non-root users from setting flags on devices.  When
	 * a device is reused, users can retain ownership of the device
	 * if they are allowed to set flags and programs assume that
	 * chown can't fail when done as root.
	 */
	if ((vp->v_type == VCHR || vp->v_type == VBLK) && 
	    ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0))
		return (error);

	/*
	 * note: vget is required for any operation that might mod the vnode
	 * so VINACTIVE is properly cleared.
	 */
	if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_flags = flags;
		error = VOP_SETATTR(vp, &vattr, p->p_ucred);
		vput(vp);
	}
	return (error);
}

/*
 * chflags(char *path, int flags)
 *
 * Change flags of a file given a path name.
 */
/* ARGSUSED */
int
sys_chflags(struct chflags_args *uap)
{
	struct nlookupdata nd;
	struct vnode *vp;
	int error;

	vp = NULL;
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	/* XXX Add NLC flag indicating modifying operation? */
	if (error == 0)
		error = nlookup(&nd);
	if (error == 0)
		error = ncp_writechk(&nd.nl_nch);
	if (error == 0)
		error = cache_vref(&nd.nl_nch, nd.nl_cred, &vp);
	nlookup_done(&nd);
	if (error == 0) {
		error = setfflags(vp, uap->flags);
		vrele(vp);
	}
	return (error);
}

/*
 * fchflags_args(int fd, int flags)
 *
 * Change flags of a file given a file descriptor.
 */
/* ARGSUSED */
int
sys_fchflags(struct fchflags_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	int error;

	if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
		return (error);
	if (fp->f_nchandle.ncp)
		error = ncp_writechk(&fp->f_nchandle);
	if (error == 0)
		error = setfflags((struct vnode *) fp->f_data, uap->flags);
	fdrop(fp);
	return (error);
}

static int
setfmode(struct vnode *vp, int mode)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	int error;
	struct vattr vattr;

	/*
	 * note: vget is required for any operation that might mod the vnode
	 * so VINACTIVE is properly cleared.
	 */
	if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_mode = mode & ALLPERMS;
		error = VOP_SETATTR(vp, &vattr, p->p_ucred);
		vput(vp);
	}
	return error;
}

int
kern_chmod(struct nlookupdata *nd, int mode)
{
	struct vnode *vp;
	int error;

	/* XXX Add NLC flag indicating modifying operation? */
	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
		return (error);
	if ((error = ncp_writechk(&nd->nl_nch)) == 0)
		error = setfmode(vp, mode);
	vrele(vp);
	return (error);
}

/*
 * chmod_args(char *path, int mode)
 *
 * Change mode of a file given path name.
 */
/* ARGSUSED */
int
sys_chmod(struct chmod_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_chmod(&nd, uap->mode);
	nlookup_done(&nd);
	return (error);
}

/*
 * lchmod_args(char *path, int mode)
 *
 * Change mode of a file given path name (don't follow links.)
 */
/* ARGSUSED */
int
sys_lchmod(struct lchmod_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_chmod(&nd, uap->mode);
	nlookup_done(&nd);
	return (error);
}

/*
 * fchmod_args(int fd, int mode)
 *
 * Change mode of a file given a file descriptor.
 */
/* ARGSUSED */
int
sys_fchmod(struct fchmod_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	int error;

	if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
		return (error);
	if (fp->f_nchandle.ncp)
		error = ncp_writechk(&fp->f_nchandle);
	if (error == 0)
		error = setfmode((struct vnode *)fp->f_data, uap->mode);
	fdrop(fp);
	return (error);
}

static int
setfown(struct vnode *vp, uid_t uid, gid_t gid)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	int error;
	struct vattr vattr;

	/*
	 * note: vget is required for any operation that might mod the vnode
	 * so VINACTIVE is properly cleared.
	 */
	if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_uid = uid;
		vattr.va_gid = gid;
		error = VOP_SETATTR(vp, &vattr, p->p_ucred);
		vput(vp);
	}
	return error;
}

int
kern_chown(struct nlookupdata *nd, int uid, int gid)
{
	struct vnode *vp;
	int error;

	/* XXX Add NLC flag indicating modifying operation? */
	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
		return (error);
	if ((error = ncp_writechk(&nd->nl_nch)) == 0)
		error = setfown(vp, uid, gid);
	vrele(vp);
	return (error);
}

/*
 * chown(char *path, int uid, int gid)
 *
 * Set ownership given a path name.
 */
int
sys_chown(struct chown_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_chown(&nd, uap->uid, uap->gid);
	nlookup_done(&nd);
	return (error);
}

/*
 * lchown_args(char *path, int uid, int gid)
 *
 * Set ownership given a path name, do not cross symlinks.
 */
int
sys_lchown(struct lchown_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_chown(&nd, uap->uid, uap->gid);
	nlookup_done(&nd);
	return (error);
}

/*
 * fchown_args(int fd, int uid, int gid)
 *
 * Set ownership given a file descriptor.
 */
/* ARGSUSED */
int
sys_fchown(struct fchown_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct file *fp;
	int error;

	if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
		return (error);
	if (fp->f_nchandle.ncp)
		error = ncp_writechk(&fp->f_nchandle);
	if (error == 0)
		error = setfown((struct vnode *)fp->f_data, uap->uid, uap->gid);
	fdrop(fp);
	return (error);
}

static int
getutimes(const struct timeval *tvp, struct timespec *tsp)
{
	struct timeval tv[2];

	if (tvp == NULL) {
		microtime(&tv[0]);
		TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
		tsp[1] = tsp[0];
	} else {
		TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
		TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
	}
	return 0;
}

static int
setutimes(struct vnode *vp, const struct timespec *ts, int nullflag)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	int error;
	struct vattr vattr;

	/*
	 * note: vget is required for any operation that might mod the vnode
	 * so VINACTIVE is properly cleared.
	 */
	if ((error = vget(vp, LK_EXCLUSIVE)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_atime = ts[0];
		vattr.va_mtime = ts[1];
		if (nullflag)
			vattr.va_vaflags |= VA_UTIMES_NULL;
		error = VOP_SETATTR(vp, &vattr, p->p_ucred);
		vput(vp);
	}
	return error;
}

int
kern_utimes(struct nlookupdata *nd, struct timeval *tptr)
{
	struct timespec ts[2];
	struct vnode *vp;
	int error;

	if ((error = getutimes(tptr, ts)) != 0)
		return (error);
	/* XXX Add NLC flag indicating modifying operation? */
	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);
	if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
		return (error);
	error = setutimes(vp, ts, tptr == NULL);
	vrele(vp);
	return (error);
}

/*
 * utimes_args(char *path, struct timeval *tptr)
 *
 * Set the access and modification times of a file.
 */
int
sys_utimes(struct utimes_args *uap)
{
	struct timeval tv[2];
	struct nlookupdata nd;
	int error;

	if (uap->tptr) {
 		error = copyin(uap->tptr, tv, sizeof(tv));
		if (error)
			return (error);
	}
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_utimes(&nd, uap->tptr ? tv : NULL);
	nlookup_done(&nd);
	return (error);
}

/*
 * lutimes_args(char *path, struct timeval *tptr)
 *
 * Set the access and modification times of a file.
 */
int
sys_lutimes(struct lutimes_args *uap)
{
	struct timeval tv[2];
	struct nlookupdata nd;
	int error;

	if (uap->tptr) {
		error = copyin(uap->tptr, tv, sizeof(tv));
		if (error)
			return (error);
	}
	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_utimes(&nd, uap->tptr ? tv : NULL);
	nlookup_done(&nd);
	return (error);
}

int
kern_futimes(int fd, struct timeval *tptr)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct timespec ts[2];
	struct file *fp;
	int error;

	error = getutimes(tptr, ts);
	if (error)
		return (error);
	if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
		return (error);
	if (fp->f_nchandle.ncp)
		error = ncp_writechk(&fp->f_nchandle);
	if (error == 0)
		error =  setutimes((struct vnode *)fp->f_data, ts, tptr == NULL);
	fdrop(fp);
	return (error);
}

/*
 * futimes_args(int fd, struct timeval *tptr)
 *
 * Set the access and modification times of a file.
 */
int
sys_futimes(struct futimes_args *uap)
{
	struct timeval tv[2];
	int error;

	if (uap->tptr) {
		error = copyin(uap->tptr, tv, sizeof(tv));
		if (error)
			return (error);
	}

	error = kern_futimes(uap->fd, uap->tptr ? tv : NULL);

	return (error);
}

int
kern_truncate(struct nlookupdata *nd, off_t length)
{
	struct vnode *vp;
	struct vattr vattr;
	int error;

	if (length < 0)
		return(EINVAL);
	/* XXX Add NLC flag indicating modifying operation? */
	if ((error = nlookup(nd)) != 0)
		return (error);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);
	if ((error = cache_vref(&nd->nl_nch, nd->nl_cred, &vp)) != 0)
		return (error);
	if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY)) != 0) {
		vrele(vp);
		return (error);
	}
	if (vp->v_type == VDIR) {
		error = EISDIR;
	} else if ((error = vn_writechk(vp, &nd->nl_nch)) == 0 &&
	    (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_size = length;
		error = VOP_SETATTR(vp, &vattr, nd->nl_cred);
	}
	vput(vp);
	return (error);
}

/*
 * truncate(char *path, int pad, off_t length)
 *
 * Truncate a file given its path name.
 */
int
sys_truncate(struct truncate_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
	if (error == 0)
		error = kern_truncate(&nd, uap->length);
	nlookup_done(&nd);
	return error;
}

int
kern_ftruncate(int fd, off_t length)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vattr vattr;
	struct vnode *vp;
	struct file *fp;
	int error;

	if (length < 0)
		return(EINVAL);
	if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
		return (error);
	if (fp->f_nchandle.ncp) {
		error = ncp_writechk(&fp->f_nchandle);
		if (error)
			goto done;
	}
	if ((fp->f_flag & FWRITE) == 0) {
		error = EINVAL;
		goto done;
	}
	vp = (struct vnode *)fp->f_data;
	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
	if (vp->v_type == VDIR) {
		error = EISDIR;
	} else if ((error = vn_writechk(vp, NULL)) == 0) {
		VATTR_NULL(&vattr);
		vattr.va_size = length;
		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
	}
	vn_unlock(vp);
done:
	fdrop(fp);
	return (error);
}

/*
 * ftruncate_args(int fd, int pad, off_t length)
 *
 * Truncate a file given a file descriptor.
 */
int
sys_ftruncate(struct ftruncate_args *uap)
{
	int error;

	error = kern_ftruncate(uap->fd, uap->length);

	return (error);
}

/*
 * fsync(int fd)
 *
 * Sync an open file.
 */
/* ARGSUSED */
int
sys_fsync(struct fsync_args *uap)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct file *fp;
	vm_object_t obj;
	int error;

	if ((error = holdvnode(p->p_fd, uap->fd, &fp)) != 0)
		return (error);
	vp = (struct vnode *)fp->f_data;
	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
	if ((obj = vp->v_object) != NULL)
		vm_object_page_clean(obj, 0, 0, 0);
	if ((error = VOP_FSYNC(vp, MNT_WAIT)) == 0 &&
	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) &&
	    bioops.io_fsync) {
		error = (*bioops.io_fsync)(vp);
	}
	vn_unlock(vp);
	fdrop(fp);
	return (error);
}

int
kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond)
{
	struct nchandle fnchd;
	struct nchandle tnchd;
	struct namecache *ncp;
	struct mount *mp;
	int error;

	bwillwrite();
	if ((error = nlookup(fromnd)) != 0)
		return (error);
	if ((fnchd.ncp = fromnd->nl_nch.ncp->nc_parent) == NULL)
		return (ENOENT);
	fnchd.mount = fromnd->nl_nch.mount;
	cache_hold(&fnchd);

	/*
	 * unlock the source nch so we can lookup the target nch without
	 * deadlocking.  The target may or may not exist so we do not check
	 * for a target vp like kern_mkdir() and other creation functions do.
	 *
	 * The source and target directories are ref'd and rechecked after
	 * everything is relocked to determine if the source or target file
	 * has been renamed.
	 */
	KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED);
	fromnd->nl_flags &= ~NLC_NCPISLOCKED;
	cache_unlock(&fromnd->nl_nch);

	tond->nl_flags |= NLC_CREATE;
	if ((error = nlookup(tond)) != 0) {
		cache_drop(&fnchd);
		return (error);
	}
	if ((tnchd.ncp = tond->nl_nch.ncp->nc_parent) == NULL) {
		cache_drop(&fnchd);
		return (ENOENT);
	}
	tnchd.mount = tond->nl_nch.mount;
	cache_hold(&tnchd);

	/*
	 * If the source and target are the same there is nothing to do
	 */
	if (fromnd->nl_nch.ncp == tond->nl_nch.ncp) {
		cache_drop(&fnchd);
		cache_drop(&tnchd);
		return (0);
	}

	/*
	 * Mount points cannot be renamed or overwritten
	 */
	if ((fromnd->nl_nch.ncp->nc_flag | tond->nl_nch.ncp->nc_flag) &
	    NCF_ISMOUNTPT
	) {
		cache_drop(&fnchd);
		cache_drop(&tnchd);
		return (EINVAL);
	}

	/*
	 * relock the source ncp.  NOTE AFTER RELOCKING: the source ncp
	 * may have become invalid while it was unlocked, nc_vp and nc_mount
	 * could be NULL.
	 */
	if (cache_lock_nonblock(&fromnd->nl_nch) == 0) {
		cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
	} else if (fromnd->nl_nch.ncp > tond->nl_nch.ncp) {
		cache_lock(&fromnd->nl_nch);
		cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
	} else {
		cache_unlock(&tond->nl_nch);
		cache_lock(&fromnd->nl_nch);
		cache_resolve(&fromnd->nl_nch, fromnd->nl_cred);
		cache_lock(&tond->nl_nch);
		cache_resolve(&tond->nl_nch, tond->nl_cred);
	}
	fromnd->nl_flags |= NLC_NCPISLOCKED;

	/*
	 * make sure the parent directories linkages are the same
	 */
	if (fnchd.ncp != fromnd->nl_nch.ncp->nc_parent ||
	    tnchd.ncp != tond->nl_nch.ncp->nc_parent) {
		cache_drop(&fnchd);
		cache_drop(&tnchd);
		return (ENOENT);
	}

	/*
	 * Both the source and target must be within the same filesystem and
	 * in the same filesystem as their parent directories within the
	 * namecache topology.
	 *
	 * NOTE: fromnd's nc_mount or nc_vp could be NULL.
	 */
	mp = fnchd.mount;
	if (mp != tnchd.mount || mp != fromnd->nl_nch.mount ||
	    mp != tond->nl_nch.mount) {
		cache_drop(&fnchd);
		cache_drop(&tnchd);
		return (EXDEV);
	}

	/*
	 * Make sure the mount point is writable
	 */
	if ((error = ncp_writechk(&tond->nl_nch)) != 0) {
		cache_drop(&fnchd);
		cache_drop(&tnchd);
		return (error);
	}

	/*
	 * If the target exists and either the source or target is a directory,
	 * then both must be directories.
	 *
	 * Due to relocking of the source, fromnd->nl_nch.ncp->nc_vp might h
	 * have become NULL.
	 */
	if (tond->nl_nch.ncp->nc_vp) {
		if (fromnd->nl_nch.ncp->nc_vp == NULL) {
			error = ENOENT;
		} else if (fromnd->nl_nch.ncp->nc_vp->v_type == VDIR) {
			if (tond->nl_nch.ncp->nc_vp->v_type != VDIR)
				error = ENOTDIR;
		} else if (tond->nl_nch.ncp->nc_vp->v_type == VDIR) {
			error = EISDIR;
		}
	}

	/*
	 * You cannot rename a source into itself or a subdirectory of itself.
	 * We check this by travsersing the target directory upwards looking
	 * for a match against the source.
	 */
	if (error == 0) {
		for (ncp = tnchd.ncp; ncp; ncp = ncp->nc_parent) {
			if (fromnd->nl_nch.ncp == ncp) {
				error = EINVAL;
				break;
			}
		}
	}

	cache_drop(&fnchd);
	cache_drop(&tnchd);

	/*
	 * Even though the namespaces are different, they may still represent
	 * hardlinks to the same file.  The filesystem might have a hard time
	 * with this so we issue a NREMOVE of the source instead of a NRENAME
	 * when we detect the situation.
	 */
	if (error == 0) {
		if (fromnd->nl_nch.ncp->nc_vp == tond->nl_nch.ncp->nc_vp) {
			error = VOP_NREMOVE(&fromnd->nl_nch, fromnd->nl_cred);
		} else {
			error = VOP_NRENAME(&fromnd->nl_nch, &tond->nl_nch, 
					    tond->nl_cred);
		}
	}
	return (error);
}

/*
 * rename_args(char *from, char *to)
 *
 * Rename files.  Source and destination must either both be directories,
 * or both not be directories.  If target is a directory, it must be empty.
 */
int
sys_rename(struct rename_args *uap)
{
	struct nlookupdata fromnd, tond;
	int error;

	error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0);
	if (error == 0) {
		error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0);
		if (error == 0)
			error = kern_rename(&fromnd, &tond);
		nlookup_done(&tond);
	}
	nlookup_done(&fromnd);
	return (error);
}

int
kern_mkdir(struct nlookupdata *nd, int mode)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct vattr vattr;
	int error;

	bwillwrite();
	nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE;
	if ((error = nlookup(nd)) != 0)
		return (error);

	if (nd->nl_nch.ncp->nc_vp)
		return (EEXIST);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);

	VATTR_NULL(&vattr);
	vattr.va_type = VDIR;
	vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask;

	vp = NULL;
	error = VOP_NMKDIR(&nd->nl_nch, &vp, p->p_ucred, &vattr);
	if (error == 0)
		vput(vp);
	return (error);
}

/*
 * mkdir_args(char *path, int mode)
 *
 * Make a directory file.
 */
/* ARGSUSED */
int
sys_mkdir(struct mkdir_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_mkdir(&nd, uap->mode);
	nlookup_done(&nd);
	return (error);
}

int
kern_rmdir(struct nlookupdata *nd)
{
	int error;

	bwillwrite();
	nd->nl_flags |= NLC_DELETE;
	if ((error = nlookup(nd)) != 0)
		return (error);

	/*
	 * Do not allow directories representing mount points to be
	 * deleted, even if empty.  Check write perms on mount point
	 * in case the vnode is aliased (aka nullfs).
	 */
	if (nd->nl_nch.ncp->nc_flag & (NCF_ISMOUNTPT))
		return (EINVAL);
	if ((error = ncp_writechk(&nd->nl_nch)) != 0)
		return (error);

	error = VOP_NRMDIR(&nd->nl_nch, nd->nl_cred);
	return (error);
}

/*
 * rmdir_args(char *path)
 *
 * Remove a directory file.
 */
/* ARGSUSED */
int
sys_rmdir(struct rmdir_args *uap)
{
	struct nlookupdata nd;
	int error;

	error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
	if (error == 0)
		error = kern_rmdir(&nd);
	nlookup_done(&nd);
	return (error);
}

int
kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res,
    enum uio_seg direction)
{
	struct thread *td = curthread;
	struct proc *p = td->td_proc;
	struct vnode *vp;
	struct file *fp;
	struct uio auio;
	struct iovec aiov;
	long loff;
	int error, eofflag;

	if ((error = holdvnode(p->p_fd, fd, &fp)) != 0)
		return (error);
	if ((fp->f_flag & FREAD) == 0) {
		error = EBADF;
		goto done;
	}
	vp = (struct vnode *)fp->f_data;
unionread:
	if (vp->v_type != VDIR) {
		error = EINVAL;
		goto done;
	}
	aiov.iov_base = buf;
	aiov.iov_len = count;
	auio.uio_iov = &aiov;
	auio.uio_iovcnt = 1;
	auio.uio_rw = UIO_READ;
	auio.uio_segflg = direction;
	auio.uio_td = td;
	auio.uio_resid = count;
	loff = auio.uio_offset = fp->f_offset;
	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL);
	fp->f_offset = auio.uio_offset;
	if (error)
		goto done;
	if (count == auio.uio_resid) {
		if (union_dircheckp) {
			error = union_dircheckp(td, &vp, fp);
			if (error == -1)
				goto unionread;
			if (error)
				goto done;
		}
#if 0
		if ((vp->v_flag & VROOT) &&
		    (vp->v_mount->mnt_flag & MNT_UNION)) {
			struct vnode *tvp = vp;
			vp = vp->v_mount->mnt_vnodecovered;
			vref(vp);
			fp->f_data = vp;