xref: /freebsd-9.3-release/sys/fs/coda/coda_vnops.c

/*-
 *             Coda: an Experimental Distributed File System
 *                              Release 3.1
 *
 *           Copyright (c) 1987-1998 Carnegie Mellon University
 *                          All Rights Reserved
 *
 * Permission  to  use, copy, modify and distribute this software and its
 * documentation is hereby granted,  provided  that  both  the  copyright
 * notice  and  this  permission  notice  appear  in  all  copies  of the
 * software, derivative works or  modified  versions,  and  any  portions
 * thereof, and that both notices appear in supporting documentation, and
 * that credit is given to Carnegie Mellon University  in  all  documents
 * and publicity pertaining to direct or indirect use of this code or its
 * derivatives.
 *
 * CODA IS AN EXPERIMENTAL SOFTWARE SYSTEM AND IS  KNOWN  TO  HAVE  BUGS,
 * SOME  OF  WHICH MAY HAVE SERIOUS CONSEQUENCES.  CARNEGIE MELLON ALLOWS
 * FREE USE OF THIS SOFTWARE IN ITS "AS IS" CONDITION.   CARNEGIE  MELLON
 * DISCLAIMS  ANY  LIABILITY  OF  ANY  KIND  FOR  ANY  DAMAGES WHATSOEVER
 * RESULTING DIRECTLY OR INDIRECTLY FROM THE USE OF THIS SOFTWARE  OR  OF
 * ANY DERIVATIVE WORK.
 *
 * Carnegie  Mellon  encourages  users  of  this  software  to return any
 * improvements or extensions that  they  make,  and  to  grant  Carnegie
 * Mellon the rights to redistribute these changes without encumbrance.
 *
 *  	@(#) src/sys/coda/coda_vnops.c,v 1.1.1.1 1998/08/29 21:14:52 rvb Exp $
 */
/*
 * Mach Operating System
 * Copyright (c) 1990 Carnegie-Mellon University
 * Copyright (c) 1989 Carnegie-Mellon University
 * All rights reserved.  The CMU software License Agreement specifies
 * the terms and conditions for use and redistribution.
 */

/*
 * This code was written for the Coda filesystem at Carnegie Mellon
 * University.  Contributers include David Steere, James Kistler, and
 * M. Satyanarayanan.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD: head/sys/fs/coda/coda_vnops.c 176307 2008-02-15 11:58:11Z rwatson $");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/acct.h>
#include <sys/errno.h>
#include <sys/fcntl.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/file.h>		/* Must come after sys/malloc.h */
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/uio.h>
#include <sys/unistd.h>

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

#include <fs/coda/coda.h>
#include <fs/coda/cnode.h>
#include <fs/coda/coda_vnops.h>
#include <fs/coda/coda_venus.h>
#include <fs/coda/coda_opstats.h>
#include <fs/coda/coda_subr.h>
#include <fs/coda/coda_pioctl.h>

/*
 * These flags select various performance enhancements.
 */
static int coda_attr_cache = 1;		/* Set to cache attributes. */
static int coda_symlink_cache = 1;	/* Set to cache symbolic links. */
static int coda_access_cache = 1;	/* Set to cache some access checks. */

/*
 * Structure to keep track of vfs calls.
 */
static struct coda_op_stats coda_vnodeopstats[CODA_VNODEOPS_SIZE];

#define	MARK_ENTRY(op)		(coda_vnodeopstats[op].entries++)
#define	MARK_INT_SAT(op)	(coda_vnodeopstats[op].sat_intrn++)
#define	MARK_INT_FAIL(op)	(coda_vnodeopstats[op].unsat_intrn++)
#define	MARK_INT_GEN(op)	(coda_vnodeopstats[op].gen_intrn++)

/*
 * What we are delaying for in printf.
 */
int coda_printf_delay = 0;	/* In microseconds */
int coda_vnop_print_entry = 0;
static int coda_lockdebug = 0;

/*
 * Some FreeBSD details:
 *
 * codadev_modevent is called at boot time or module load time.
 */
#define	ENTRY do {							\
	if (coda_vnop_print_entry)					\
		myprintf(("Entered %s\n", __func__));			\
} while (0)

/*
 * Definition of the vnode operation vector.
 */
struct vop_vector coda_vnodeops = {
	.vop_default = &default_vnodeops,
	.vop_cachedlookup = coda_lookup,	/* uncached lookup */
	.vop_lookup = vfs_cache_lookup,		/* namecache lookup */
	.vop_create = coda_create,		/* create */
	.vop_open = coda_open,			/* open */
	.vop_close = coda_close,		/* close */
	.vop_access = coda_access,		/* access */
	.vop_getattr = coda_getattr,		/* getattr */
	.vop_setattr = coda_setattr,		/* setattr */
	.vop_read = coda_read,			/* read */
	.vop_write = coda_write,		/* write */
	.vop_ioctl = coda_ioctl,		/* ioctl */
	.vop_fsync = coda_fsync,		/* fsync */
	.vop_remove = coda_remove,		/* remove */
	.vop_link = coda_link,			/* link */
	.vop_rename = coda_rename,		/* rename */
	.vop_mkdir = coda_mkdir,		/* mkdir */
	.vop_rmdir = coda_rmdir,		/* rmdir */
	.vop_symlink = coda_symlink,		/* symlink */
	.vop_readdir = coda_readdir,		/* readdir */
	.vop_readlink = coda_readlink,		/* readlink */
	.vop_inactive = coda_inactive,		/* inactive */
	.vop_reclaim = coda_reclaim,		/* reclaim */
	.vop_lock1 = coda_lock,			/* lock */
	.vop_unlock = coda_unlock,		/* unlock */
	.vop_bmap = VOP_EOPNOTSUPP,		/* bmap */
	.vop_print = VOP_NULL,			/* print */
	.vop_islocked = coda_islocked,		/* islocked */
	.vop_pathconf = coda_pathconf,		/* pathconf */
	.vop_poll = vop_stdpoll,
	.vop_getpages = vop_stdgetpages,	/* pager intf.*/
	.vop_putpages = vop_stdputpages,	/* pager intf.*/
	.vop_getwritemount = vop_stdgetwritemount,
#if 0
	/* missing */
	.vop_cachedlookup = ufs_lookup,
	.vop_whiteout =	ufs_whiteout,
#endif

};

static void	coda_print_vattr(struct vattr *attr);

int
coda_vnodeopstats_init(void)
{
	int i;

	for(i=0; i<CODA_VNODEOPS_SIZE; i++) {
		coda_vnodeopstats[i].opcode = i;
		coda_vnodeopstats[i].entries = 0;
		coda_vnodeopstats[i].sat_intrn = 0;
		coda_vnodeopstats[i].unsat_intrn = 0;
		coda_vnodeopstats[i].gen_intrn = 0;
	}
	return (0);
}

/*
 * coda_open calls Venus which returns an open file descriptor the cache file
 * holding the data.  We get the vnode while we are still in the context of
 * the venus process in coda_psdev.c.  This vnode is then passed back to the
 * caller and opened.
 */
int
coda_open(struct vop_open_args *ap)
{

	/*
	 * FreeBSD can pass the O_EXCL flag in mode, even though the check
	 * has already happened.  Venus defensively assumes that if open is
	 * passed the EXCL, it must be a bug.  We strip the flag here.
	 */
	/* true args */
	struct vnode **vpp = &(ap->a_vp);
	struct cnode *cp = VTOC(*vpp);
	int flag = ap->a_mode & (~O_EXCL);
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
	int error;
	struct vnode *vp;

	MARK_ENTRY(CODA_OPEN_STATS);

	/*
	 * Check for open of control file.
	 */
	if (IS_CTL_VP(*vpp)) {
		/* XXX */
		/* if (WRITEABLE(flag)) */
		if (flag & (FWRITE | O_TRUNC | O_CREAT | O_EXCL)) {
			MARK_INT_FAIL(CODA_OPEN_STATS);
			return (EACCES);
		}
		MARK_INT_SAT(CODA_OPEN_STATS);
		return (0);
	}
	error = venus_open(vtomi((*vpp)), &cp->c_fid, flag, cred,
	    td->td_proc, &vp);
	if (error)
		return (error);
	CODADEBUG(CODA_OPEN, myprintf(("open: vp %p result %d\n", vp,
	    error)););

	/*
	 * Save the vnode pointer for the cache file.
	 */
	if (cp->c_ovp == NULL) {
		cp->c_ovp = vp;
	} else {
		if (cp->c_ovp != vp)
			panic("coda_open: cp->c_ovp != ITOV(ip)");
	}
	cp->c_ocount++;

	/*
	 * Flush the attribute cached if writing the file.
	 */
	if (flag & FWRITE) {
		cp->c_owrite++;
		cp->c_flags &= ~C_VATTR;
	}

	/*
	 * Open the cache file.
	 */
	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
	error = VOP_OPEN(vp, flag, cred, td, NULL);
	if (error) {
		VOP_UNLOCK(vp, 0);
    		printf("coda_open: VOP_OPEN on container failed %d\n", error);
		return (error);
	}
	(*vpp)->v_object = vp->v_object;
	VOP_UNLOCK(vp, 0);
	return (0);
}

/*
 * Close the cache file used for I/O and notify Venus.
 */
int
coda_close(struct vop_close_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	int flag = ap->a_fflag;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
	int error;

	MARK_ENTRY(CODA_CLOSE_STATS);

	/*
	 * Check for close of control file.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_SAT(CODA_CLOSE_STATS);
		return (0);
	}
	if (cp->c_ovp) {
		vn_lock(cp->c_ovp, LK_EXCLUSIVE | LK_RETRY);
		/* Do errors matter here? */
		VOP_CLOSE(cp->c_ovp, flag, cred, td);
		vput(cp->c_ovp);
	}
#ifdef CODA_VERBOSE
	else
		printf("coda_close: NO container vp %p/cp %p\n", vp, cp);
#endif
	if (--cp->c_ocount == 0)
		cp->c_ovp = NULL;

	/*
	 * File was opened for write.
	 */
	if (flag & FWRITE)
		--cp->c_owrite;
	if (!IS_UNMOUNTING(cp))
		error = venus_close(vtomi(vp), &cp->c_fid, flag, cred,
		    td->td_proc);
	else
		error = ENODEV;
	CODADEBUG(CODA_CLOSE, myprintf(("close: result %d\n",error)););
	return (error);
}

int
coda_read(struct vop_read_args *ap)
{

	ENTRY;
	return (coda_rdwr(ap->a_vp, ap->a_uio, UIO_READ, ap->a_ioflag,
	    ap->a_cred, ap->a_uio->uio_td));
}

int
coda_write(struct vop_write_args *ap)
{

	ENTRY;
	return (coda_rdwr(ap->a_vp, ap->a_uio, UIO_WRITE, ap->a_ioflag,
	    ap->a_cred, ap->a_uio->uio_td));
}

int
coda_rdwr(struct vnode *vp, struct uio *uiop, enum uio_rw rw, int ioflag,
    struct ucred *cred, struct thread *td)
{
	/* upcall decl */
	/* NOTE: container file operation!!! */
	/* locals */
	struct cnode *cp = VTOC(vp);
	struct vnode *cfvp = cp->c_ovp;
	int opened_internally = 0;
	int error = 0;

	MARK_ENTRY(CODA_RDWR_STATS);
	CODADEBUG(CODA_RDWR, myprintf(("coda_rdwr(%d, %p, %d, %lld, %d)\n",
	    rw, (void *)uiop->uio_iov->iov_base, uiop->uio_resid,
	    (long long)uiop->uio_offset, uiop->uio_segflg)););

	/*
	 * Check for rdwr of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_RDWR_STATS);
		return (EINVAL);
	}

	/*
	 * If file is not already open this must be a page {read,write}
	 * request and we should open it internally.
	 */
	if (cfvp == NULL) {
		opened_internally = 1;
		MARK_INT_GEN(CODA_OPEN_STATS);
		error = VOP_OPEN(vp, (rw == UIO_READ ? FREAD : FWRITE), cred,
		    td, NULL);
#ifdef CODA_VERBOSE
		printf("coda_rdwr: Internally Opening %p\n", vp);
#endif
		if (error) {
			printf("coda_rdwr: VOP_OPEN on container failed "
			    "%d\n", error);
			return (error);
		}
		cfvp = cp->c_ovp;
	}

	/*
	 * Have UFS handle the call.
	 */
	CODADEBUG(CODA_RDWR, myprintf(("indirect rdwr: fid = %s, refcnt = "
	    "%d\n", coda_f2s(&cp->c_fid), CTOV(cp)->v_usecount)););
	vn_lock(cfvp, LK_EXCLUSIVE | LK_RETRY);
	if (rw == UIO_READ) {
		error = VOP_READ(cfvp, uiop, ioflag, cred);
	} else {
		error = VOP_WRITE(cfvp, uiop, ioflag, cred);
		/*
		 * ufs_write updates the vnode_pager_setsize for the
		 * vnode/object.
		 *
		 * XXX: Since we now share vm objects between layers, this is
		 * probably unnecessary.
		 */
		{
			struct vattr attr;
			if (VOP_GETATTR(cfvp, &attr, cred, td) == 0)
				vnode_pager_setsize(vp, attr.va_size);
		}
	}
	VOP_UNLOCK(cfvp, 0);
	if (error)
		MARK_INT_FAIL(CODA_RDWR_STATS);
	else
		MARK_INT_SAT(CODA_RDWR_STATS);

	/*
	 * Do an internal close if necessary.
	 */
	if (opened_internally) {
		MARK_INT_GEN(CODA_CLOSE_STATS);
		(void)VOP_CLOSE(vp, (rw == UIO_READ ? FREAD : FWRITE), cred,
		    td);
	}

	/*
	 * Invalidate cached attributes if writing.
	 */
	if (rw == UIO_WRITE)
		cp->c_flags &= ~C_VATTR;
	return (error);
}

int
coda_ioctl(struct vop_ioctl_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	int com = ap->a_command;
	caddr_t data = ap->a_data;
	int flag = ap->a_fflag;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
	int error;
	struct vnode *tvp;
	struct nameidata ndp;
	struct PioctlData *iap = (struct PioctlData *)data;

	MARK_ENTRY(CODA_IOCTL_STATS);
	CODADEBUG(CODA_IOCTL, myprintf(("in coda_ioctl on %s\n", iap->path)););

	/*
	 * Don't check for operation on a dying object, for ctlvp it
	 * shouldn't matter.
	 *
	 * Must be control object to succeed.
	 */
	if (!IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_IOCTL_STATS);
		CODADEBUG(CODA_IOCTL, myprintf(("coda_ioctl error: vp != "
		    "ctlvp")););
		return (EOPNOTSUPP);
	}

	/*
	 * Look up the pathname.
	 *
	 * Should we use the name cache here? It would get it from lookupname
	 * sooner or later anyway, right?
	 */
	NDINIT(&ndp, LOOKUP, (iap->follow ? FOLLOW : NOFOLLOW),
	    UIO_USERSPACE, iap->path, td);
	error = namei(&ndp);
	tvp = ndp.ni_vp;
	if (error) {
		MARK_INT_FAIL(CODA_IOCTL_STATS);
		CODADEBUG(CODA_IOCTL, myprintf(("coda_ioctl error: lookup "
		    "returns %d\n", error)););
		return (error);
	}

	/*
	 * Make sure this is a coda style cnode, but it may be a different
	 * vfsp.
	 */
	if (tvp->v_op != &coda_vnodeops) {
		vrele(tvp);
		NDFREE(&ndp, NDF_ONLY_PNBUF);
		MARK_INT_FAIL(CODA_IOCTL_STATS);
		CODADEBUG(CODA_IOCTL,
		myprintf(("coda_ioctl error: %s not a coda object\n",
		    iap->path)););
		return (EINVAL);
	}
	if (iap->vi.in_size > VC_MAXDATASIZE) {
		NDFREE(&ndp, 0);
		return (EINVAL);
	}
	error = venus_ioctl(vtomi(tvp), &((VTOC(tvp))->c_fid), com, flag,
	    data, cred, td->td_proc);
	if (error)
		MARK_INT_FAIL(CODA_IOCTL_STATS);
	else
		CODADEBUG(CODA_IOCTL, myprintf(("Ioctl returns %d \n",
		    error)););
	vrele(tvp);
	NDFREE(&ndp, NDF_ONLY_PNBUF);
	return (error);
}

/*
 * To reduce the cost of a user-level venus;we cache attributes in the
 * kernel.  Each cnode has storage allocated for an attribute.  If c_vattr is
 * valid, return a reference to it.  Otherwise, get the attributes from venus
 * and store them in the cnode.  There is some question if this method is a
 * security leak.  But I think that in order to make this call, the user must
 * have done a lookup and opened the file, and therefore should already have
 * access.
 */
int
coda_getattr(struct vop_getattr_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct vattr *vap = ap->a_vap;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
    	struct vnode *convp;
	int error, size;

	MARK_ENTRY(CODA_GETATTR_STATS);
	if (IS_UNMOUNTING(cp))
		return (ENODEV);

	/*
	 * Check for getattr of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_GETATTR_STATS);
		return (ENOENT);
	}

	/*
	 * Check to see if the attributes have already been cached.
	 */
	if (VALID_VATTR(cp)) {
		CODADEBUG(CODA_GETATTR, myprintf(("attr cache hit: %s\n",
		    coda_f2s(&cp->c_fid))););
		CODADEBUG(CODA_GETATTR, if (!(codadebug & ~CODA_GETATTR))
		    coda_print_vattr(&cp->c_vattr););
		*vap = cp->c_vattr;
		MARK_INT_SAT(CODA_GETATTR_STATS);
		return (0);
	}
    	error = venus_getattr(vtomi(vp), &cp->c_fid, cred, td->td_proc, vap);
	if (!error) {
		CODADEBUG(CODA_GETATTR, myprintf(("getattr miss %s: result "
		    "%d\n", coda_f2s(&cp->c_fid), error)););
		CODADEBUG(CODA_GETATTR, if (!(codadebug & ~CODA_GETATTR))
		    coda_print_vattr(vap););

		/*
		 * XXX: Since we now share vm objects between layers, this is
		 * probably unnecessary.
		 */
		size = vap->va_size;
    		convp = cp->c_ovp;
		if (convp != NULL)
			vnode_pager_setsize(convp, size);

		/*
		 * If not open for write, store attributes in cnode.
		 */
		if ((cp->c_owrite == 0) && (coda_attr_cache)) {
			cp->c_vattr = *vap;
			cp->c_flags |= C_VATTR;
		}
	}
	return (error);
}

int
coda_setattr(struct vop_setattr_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct vattr *vap = ap->a_vap;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
    	struct vnode *convp;
	int error, size;

	MARK_ENTRY(CODA_SETATTR_STATS);

	/*
	 * Check for setattr of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_SETATTR_STATS);
		return (ENOENT);
	}
	if (codadebug & CODADBGMSK(CODA_SETATTR))
		coda_print_vattr(vap);
	error = venus_setattr(vtomi(vp), &cp->c_fid, vap, cred, td->td_proc);
	if (!error)
		cp->c_flags &= ~(C_VATTR | C_ACCCACHE);

	/*
	 * XXX: Since we now share vm objects between layers, this is
	 * probably unnecessary.
	 *
	 * XXX: Shouldn't we only be doing this "set" if C_VATTR remains
	 * valid after venus_setattr()?
	 */
	size = vap->va_size;
    	convp = cp->c_ovp;
	if (size != VNOVAL && convp != NULL)
		vnode_pager_setsize(convp, size);
	CODADEBUG(CODA_SETATTR,	myprintf(("setattr %d\n", error)););
	return (error);
}

int
coda_access(struct vop_access_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	int mode = ap->a_mode;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_td;
	/* locals */
	int error;

	MARK_ENTRY(CODA_ACCESS_STATS);

	/*
	 * Check for access of control object.  Only read access is allowed
	 * on it.
	 */
	if (IS_CTL_VP(vp)) {
		/*
		 * Bogus hack - all will be marked as successes.
		 */
		MARK_INT_SAT(CODA_ACCESS_STATS);
		return (((mode & VREAD) && !(mode & (VWRITE | VEXEC)))
		    ? 0 : EACCES);
	}

	/*
	 * We maintain a one-entry LRU positive access cache with each cnode.
	 * In principle we could also track negative results, and for more
	 * than one uid, but we don't yet.  Venus is responsible for
	 * invalidating this cache as required.
	 */
	if (coda_access_cache && VALID_ACCCACHE(cp) &&
	    (cred->cr_uid == cp->c_cached_uid) &&
	    (mode & cp->c_cached_mode) == mode) {
		MARK_INT_SAT(CODA_ACCESS_STATS);
		return (0);
	}
	error = venus_access(vtomi(vp), &cp->c_fid, mode, cred, td->td_proc);
	if (error == 0 && coda_access_cache) {
		/*-
		 * When we have a new successful request, we consider three
		 * cases:
		 *
		 * - No initialized access cache, in which case cache the
		 *   result.
		 * - Cached result for a different user, in which case we
		 *   replace the entry.
		 * - Cached result for the same user, in which case we add
		 *   any newly granted rights to the cached mode.
		 *
		 * XXXRW: If we ever move to something more interesting than
		 * uid-based token lookup, we'll need to change this.
		 */
		cp->c_flags |= C_ACCCACHE;
		if (cp->c_cached_uid != cred->cr_uid) {
			cp->c_cached_mode = mode;
			cp->c_cached_uid = cred->cr_uid;
		} else
			cp->c_cached_mode |= mode;
	}
	return (error);
}

int
coda_readlink(struct vop_readlink_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct uio *uiop = ap->a_uio;
	struct ucred *cred = ap->a_cred;
	struct thread *td = ap->a_uio->uio_td;
	/* locals */
	int error;
	char *str;
	int len;

	MARK_ENTRY(CODA_READLINK_STATS);

	/*
	 * Check for readlink of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_READLINK_STATS);
		return (ENOENT);
	}
	if ((coda_symlink_cache) && (VALID_SYMLINK(cp))) {
		/*
		 * Symlink was cached.
		 */
		uiop->uio_rw = UIO_READ;
		error = uiomove(cp->c_symlink, (int)cp->c_symlen, uiop);
		if (error)
			MARK_INT_FAIL(CODA_READLINK_STATS);
		else
			MARK_INT_SAT(CODA_READLINK_STATS);
		return (error);
	}
	error = venus_readlink(vtomi(vp), &cp->c_fid, cred, td != NULL ?
	    td->td_proc : NULL, &str, &len);
	if (!error) {
		uiop->uio_rw = UIO_READ;
		error = uiomove(str, len, uiop);
		if (coda_symlink_cache) {
			cp->c_symlink = str;
			cp->c_symlen = len;
			cp->c_flags |= C_SYMLINK;
		} else
			CODA_FREE(str, len);
	}
	CODADEBUG(CODA_READLINK, myprintf(("in readlink result %d\n",
	    error)););
	return (error);
}

int
coda_fsync(struct vop_fsync_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct thread *td = ap->a_td;
	/* locals */
	struct vnode *convp = cp->c_ovp;
	int error;

	MARK_ENTRY(CODA_FSYNC_STATS);

	/*
	 * Check for fsync on an unmounting object.
	 *
	 * XXX: Is this comment true on FreeBSD?  It seems likely, since
	 * unmounting is fairly non-atomic.
	 *
	 * The NetBSD kernel, in it's infinite wisdom, can try to fsync after
	 * an unmount has been initiated.  This is a Bad Thing, which we have
	 * to avoid.  Not a legitimate failure for stats.
	 */
	if (IS_UNMOUNTING(cp))
		return (ENODEV);

	/*
	 * Check for fsync of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_SAT(CODA_FSYNC_STATS);
		return (0);
	}
	if (convp != NULL) {
		vn_lock(convp, LK_EXCLUSIVE | LK_RETRY);
		VOP_FSYNC(convp, MNT_WAIT, td);
		VOP_UNLOCK(convp, 0);
	}

	/*
	 * We see fsyncs with usecount == 1 then usecount == 0.  For now we
	 * ignore them.
	 */
#if 0
	VI_LOCK(vp);
	if (!vp->v_usecount) {
		printf("coda_fsync on vnode %p with %d usecount.  "
		    "c_flags = %x (%x)\n", vp, vp->v_usecount, cp->c_flags,
		    cp->c_flags&C_PURGING);
	}
	VI_UNLOCK(vp);
#endif

	/*
	 * We can expect fsync on any vnode at all if venus is purging it.
	 * Venus can't very well answer the fsync request, now can it?
	 * Hopefully, it won't have to, because hopefully, venus preserves
	 * the (possibly untrue) invariant that it never purges an open
	 * vnode.  Hopefully.
	 */
	if (cp->c_flags & C_PURGING)
		return (0);

	/* XXX: needs research */
	return (0);
	error = venus_fsync(vtomi(vp), &cp->c_fid, td->td_proc);
	CODADEBUG(CODA_FSYNC, myprintf(("in fsync result %d\n", error)););
	return (error);
}

int
coda_inactive(struct vop_inactive_args *ap)
{
	/*
	 * XXX - at the moment, inactive doesn't look at cred, and doesn't
	 * have a proc pointer.  Oops.
	 */
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct ucred *cred __attribute__((unused)) = NULL;
	struct thread *td __attribute__((unused)) = curthread;
	/* upcall decl */
	/* locals */

	/*
	 * We don't need to send inactive to venus - DCS.
	 */
	MARK_ENTRY(CODA_INACTIVE_STATS);
	CODADEBUG(CODA_INACTIVE, myprintf(("in inactive, %s, vfsp %p\n",
	    coda_f2s(&cp->c_fid), vp->v_mount)););
	vp->v_object = NULL;

	/*
	 * If an array has been allocated to hold the symlink, deallocate it.
	 */
	if ((coda_symlink_cache) && (VALID_SYMLINK(cp))) {
		if (cp->c_symlink == NULL)
			panic("coda_inactive: null symlink pointer in cnode");
		CODA_FREE(cp->c_symlink, cp->c_symlen);
		cp->c_flags &= ~C_SYMLINK;
		cp->c_symlen = 0;
	}

	/*
	 * Remove it from the table so it can't be found.
	 */
	coda_unsave(cp);
	if ((struct coda_mntinfo *)(vp->v_mount->mnt_data) == NULL) {
		myprintf(("Help! vfsp->vfs_data was NULL, but vnode %p "
		    "wasn't dying\n", vp));
		panic("badness in coda_inactive\n");
	}
	if (IS_UNMOUNTING(cp)) {
#ifdef	DEBUG
		printf("coda_inactive: IS_UNMOUNTING use %d: vp %p, cp %p\n",
		    vrefcnt(vp), vp, cp);
		if (cp->c_ovp != NULL)
			printf("coda_inactive: cp->ovp != NULL use %d: vp "
			    "%p, cp %p\n", vrefcnt(vp), vp, cp);
#endif
	} else
		vgone(vp);
	MARK_INT_SAT(CODA_INACTIVE_STATS);
	return (0);
}

/*
 * Remote filesystem operations having to do with directory manipulation.
 */

/*
 * In FreeBSD, lookup returns the vnode locked.
 */
int
coda_lookup(struct vop_cachedlookup_args *ap)
{
	/* true args */
	struct vnode *dvp = ap->a_dvp;
	struct cnode *dcp = VTOC(dvp);
	struct vnode **vpp = ap->a_vpp;
	/*
	 * It looks as though ap->a_cnp->ni_cnd->cn_nameptr holds the rest of
	 * the string to xlate, and that we must try to get at least
	 * ap->a_cnp->ni_cnd->cn_namelen of those characters to macth.  I
	 * could be wrong.
	 */
	struct componentname  *cnp = ap->a_cnp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* locals */
	struct cnode *cp;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
	CodaFid VFid;
	int vtype;
	int error = 0;

	MARK_ENTRY(CODA_LOOKUP_STATS);
	CODADEBUG(CODA_LOOKUP, myprintf(("lookup: %s in %s\n", nm,
	    coda_f2s(&dcp->c_fid))););

	/*
	 * Check for lookup of control object.
	 */
	if (IS_CTL_NAME(dvp, nm, len)) {
		*vpp = coda_ctlvp;
		vref(*vpp);
		MARK_INT_SAT(CODA_LOOKUP_STATS);
		goto exit;
	}
	if (len+1 > CODA_MAXNAMLEN) {
		MARK_INT_FAIL(CODA_LOOKUP_STATS);
		CODADEBUG(CODA_LOOKUP, myprintf(("name too long: lookup, "
		    "%s (%s)\n", coda_f2s(&dcp->c_fid), nm)););
		*vpp = NULL;
		error = EINVAL;
		goto exit;
	}

	error = venus_lookup(vtomi(dvp), &dcp->c_fid, nm, len, cred,
	    td->td_proc, &VFid, &vtype);
	if (error) {
		MARK_INT_FAIL(CODA_LOOKUP_STATS);
		CODADEBUG(CODA_LOOKUP, myprintf(("lookup error on %s "
		    "(%s)%d\n", coda_f2s(&dcp->c_fid), nm, error)););
		*vpp = NULL;
	} else {
		MARK_INT_SAT(CODA_LOOKUP_STATS);
		CODADEBUG(CODA_LOOKUP, myprintf(("lookup: %s type %o "
		    "result %d\n", coda_f2s(&VFid), vtype, error)););
		cp = make_coda_node(&VFid, dvp->v_mount, vtype);
    		*vpp = CTOV(cp);

    		/*
		 * Enter the new vnode in the namecache only if the top bit
		 * isn't set.
		 *
		 * And don't enter a new vnode for an invalid one!
		 */
		if (!(vtype & CODA_NOCACHE) && (cnp->cn_flags & MAKEENTRY))
			cache_enter(dvp, *vpp, cnp);
	}
exit:
	/*
	 * If we are creating, and this was the last name to be looked up,
	 * and the error was ENOENT, then there really shouldn't be an error
	 * and we can make the leaf NULL and return success.  Since this is
	 * supposed to work under Mach as well as FreeBSD, we're leaving this
	 * fn wrapped.  We also must tell lookup/namei that we need to save
	 * the last component of the name.  (Create will have to free the
	 * name buffer later...lucky us...).
	 */
	if (((cnp->cn_nameiop == CREATE) || (cnp->cn_nameiop == RENAME))
	    && (cnp->cn_flags & ISLASTCN) && (error == ENOENT)) {
		error = EJUSTRETURN;
		cnp->cn_flags |= SAVENAME;
		*ap->a_vpp = NULL;
	}

	/*
	 * If we are removing, and we are at the last element, and we found
	 * it, then we need to keep the name around so that the removal will
	 * go ahead as planned.  Unfortunately, this will probably also lock
	 * the to-be-removed vnode, which may or may not be a good idea.
	 * I'll have to look at the bits of coda_remove to make sure.  We'll
	 * only save the name if we did in fact find the name, otherwise
	 * coda_remove won't have a chance to free the pathname.
	 */
	if ((cnp->cn_nameiop == DELETE) && (cnp->cn_flags & ISLASTCN)
	    && !error)
		cnp->cn_flags |= SAVENAME;

	/*
	 * If the lookup went well, we need to (potentially?) unlock the
	 * parent, and lock the child.  We are only responsible for checking
	 * to see if the parent is supposed to be unlocked before we return.
	 * We must always lock the child (provided there is one, and (the
	 * parent isn't locked or it isn't the same as the parent.)  Simple,
	 * huh?  We can never leave the parent locked unless we are ISLASTCN.
	 */
	if (!error || (error == EJUSTRETURN)) {
		if (cnp->cn_flags & ISDOTDOT) {
			VOP_UNLOCK(dvp, 0);
			/*
			 * The parent is unlocked.  As long as there is a
			 * child, lock it without bothering to check anything
			 * else.
			 */
			if (*ap->a_vpp)
				vn_lock(*ap->a_vpp, LK_EXCLUSIVE | LK_RETRY);
			vn_lock(dvp, LK_RETRY|LK_EXCLUSIVE);
		} else {
			/*
			 * The parent is locked, and may be the same as the
			 * child.  If different, go ahead and lock it.
			 */
			if (*ap->a_vpp && (*ap->a_vpp != dvp))
				vn_lock(*ap->a_vpp, LK_EXCLUSIVE | LK_RETRY);
		}
	} else {
		/*
		 * If the lookup failed, we need to ensure that the leaf is
		 * NULL.
		 *
		 * Don't change any locking?
		 */
		*ap->a_vpp = NULL;
	}
	return (error);
}

/*ARGSUSED*/
int
coda_create(struct vop_create_args *ap)
{
	/* true args */
	struct vnode *dvp = ap->a_dvp;
	struct cnode *dcp = VTOC(dvp);
	struct vattr *va = ap->a_vap;
	int exclusive = 1;
	int mode = ap->a_vap->va_mode;
	struct vnode **vpp = ap->a_vpp;
	struct componentname  *cnp = ap->a_cnp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* locals */
	int error;
	struct cnode *cp;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
	CodaFid VFid;
	struct vattr attr;

	MARK_ENTRY(CODA_CREATE_STATS);

	/*
	 * All creates are exclusive XXX.
	 *
	 * I'm assuming the 'mode' argument is the file mode bits XXX.
	 *
	 * Check for create of control object.
	 */
	if (IS_CTL_NAME(dvp, nm, len)) {
		*vpp = (struct vnode *)0;
		MARK_INT_FAIL(CODA_CREATE_STATS);
		return (EACCES);
	}
	error = venus_create(vtomi(dvp), &dcp->c_fid, nm, len, exclusive,
	    mode, va, cred, td->td_proc, &VFid, &attr);
	if (!error) {
		/*
		 * If this is an exclusive create, panic if the file already
		 * exists.
		 *
		 * Venus should have detected the file and reported EEXIST.
		 */
		if ((exclusive == 1) && (coda_find(&VFid) != NULL))
	  	  	panic("cnode existed for newly created file!");
		cp = make_coda_node(&VFid, dvp->v_mount, attr.va_type);
		*vpp = CTOV(cp);

		/*
		 * Update va to reflect the new attributes.
		 */
		(*va) = attr;

		/*
		 * Update the attribute cache and mark it as valid.
		 */
		if (coda_attr_cache) {
			VTOC(*vpp)->c_vattr = attr;
			VTOC(*vpp)->c_flags |= C_VATTR;
		}

		/*
		 * Invalidate the parent's attr cache, the modification time
		 * has changed.
		 */
		VTOC(dvp)->c_flags &= ~C_VATTR;
		cache_enter(dvp, *vpp, cnp);
		CODADEBUG(CODA_CREATE, myprintf(("create: %s, result %d\n",
		    coda_f2s(&VFid), error)););
	} else {
		*vpp = (struct vnode *)0;
		CODADEBUG(CODA_CREATE, myprintf(("create error %d\n",
		    error)););
	}
	if (!error) {
		if (cnp->cn_flags & MAKEENTRY)
			cache_enter(dvp, *vpp, cnp);
		if (cnp->cn_flags & LOCKLEAF)
			vn_lock(*ap->a_vpp, LK_EXCLUSIVE | LK_RETRY);
	} else if (error == ENOENT) {
		/*
		 * XXXRW: We only enter a negative entry if ENOENT is
		 * returned, not other errors.  But will Venus invalidate dvp
		 * properly in all cases when new files appear via the
		 * network rather than a local operation?
		 */
		if (cnp->cn_flags & MAKEENTRY)
			cache_enter(dvp, NULL, cnp);
	}
	return (error);
}

int
coda_remove(struct vop_remove_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct vnode *dvp = ap->a_dvp;
	struct cnode *cp = VTOC(dvp);
	struct componentname  *cnp = ap->a_cnp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* locals */
	int error;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
#if 0
	struct cnode *tp;
#endif

	MARK_ENTRY(CODA_REMOVE_STATS);
	CODADEBUG(CODA_REMOVE, myprintf(("remove: %s in %s\n", nm,
	    coda_f2s(&cp->c_fid))););

	/*
	 * Check for remove of control object.
	 */
	if (IS_CTL_NAME(dvp, nm, len)) {
		MARK_INT_FAIL(CODA_REMOVE_STATS);
		return (ENOENT);
	}

	/*
	 * Invalidate the parent's attr cache, the modification time has
	 * changed.  We don't yet know if the last reference to the file is
	 * being removed, but we do know the reference count on the child has
	 * changed, so invalidate its attr cache also.
	 */
	VTOC(dvp)->c_flags &= ~C_VATTR;
	VTOC(vp)->c_flags &= ~(C_VATTR | C_ACCCACHE);
	error = venus_remove(vtomi(dvp), &cp->c_fid, nm, len, cred,
	    td->td_proc);
	cache_purge(vp);
	CODADEBUG(CODA_REMOVE, myprintf(("in remove result %d\n",error)););
	return (error);
}

int
coda_link(struct vop_link_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct vnode *tdvp = ap->a_tdvp;
	struct cnode *tdcp = VTOC(tdvp);
	struct componentname *cnp = ap->a_cnp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* locals */
	int error;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;

	MARK_ENTRY(CODA_LINK_STATS);

	if (codadebug & CODADBGMSK(CODA_LINK)) {
		myprintf(("nb_link:   vp fid: %s\n", coda_f2s(&cp->c_fid)));
		myprintf(("nb_link: tdvp fid: %s)\n",
		    coda_f2s(&tdcp->c_fid)));
	}
	if (codadebug & CODADBGMSK(CODA_LINK)) {
		myprintf(("link:   vp fid: %s\n", coda_f2s(&cp->c_fid)));
		myprintf(("link: tdvp fid: %s\n", coda_f2s(&tdcp->c_fid)));
	}

	/*
	 * Check for link to/from control object.
	 */
	if (IS_CTL_NAME(tdvp, nm, len) || IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_LINK_STATS);
		return (EACCES);
	}
	error = venus_link(vtomi(vp), &cp->c_fid, &tdcp->c_fid, nm, len,
	    cred, td->td_proc);

	/*
	 * Invalidate the parent's attr cache, the modification time has
	 * changed.
	 */
	VTOC(tdvp)->c_flags &= ~C_VATTR;
	VTOC(vp)->c_flags &= ~C_VATTR;
	CODADEBUG(CODA_LINK, myprintf(("in link result %d\n",error)););
	return (error);
}

int
coda_rename(struct vop_rename_args *ap)
{
	/* true args */
	struct vnode *fvp = ap->a_fvp;
	struct vnode *tvp = ap->a_tvp;
	struct vnode *odvp = ap->a_fdvp;
	struct cnode *odcp = VTOC(odvp);
	struct componentname  *fcnp = ap->a_fcnp;
	struct vnode *ndvp = ap->a_tdvp;
	struct cnode *ndcp = VTOC(ndvp);
	struct componentname  *tcnp = ap->a_tcnp;
	struct ucred *cred = fcnp->cn_cred;
	struct thread *td = fcnp->cn_thread;
	/* true args */
	int error;
	const char *fnm = fcnp->cn_nameptr;
	int flen = fcnp->cn_namelen;
	const char *tnm = tcnp->cn_nameptr;
	int tlen = tcnp->cn_namelen;

	MARK_ENTRY(CODA_RENAME_STATS);

	/*
	 * Check for rename involving control object.
	 */
	if (IS_CTL_NAME(odvp, fnm, flen) || IS_CTL_NAME(ndvp, tnm, tlen)) {
		MARK_INT_FAIL(CODA_RENAME_STATS);
		return (EACCES);
	}

	/*
	 * Remove the entries for both source and target directories, which
	 * should catch references to the children.  Perhaps we could purge
	 * less?
	 */
	cache_purge(odvp);
	cache_purge(ndvp);

	/*
	 * Invalidate parent directories as modification times have changed.
	 * Invalidate access cache on renamed file as rights may have
	 * changed.
	 */
	VTOC(odvp)->c_flags &= ~C_VATTR;
	VTOC(ndvp)->c_flags &= ~C_VATTR;
	VTOC(fvp)->c_flags &= ~C_ACCCACHE;
	if (flen+1 > CODA_MAXNAMLEN) {
		MARK_INT_FAIL(CODA_RENAME_STATS);
		error = EINVAL;
		goto exit;
	}
	if (tlen+1 > CODA_MAXNAMLEN) {
		MARK_INT_FAIL(CODA_RENAME_STATS);
		error = EINVAL;
		goto exit;
	}
	error = venus_rename(vtomi(odvp), &odcp->c_fid, &ndcp->c_fid, fnm,
	    flen, tnm, tlen, cred, td->td_proc);
exit:
	CODADEBUG(CODA_RENAME, myprintf(("in rename result %d\n",error)););

	/*
	 * Update namecache to reflect that the names of various objects may
	 * have changed (or gone away entirely).
	 */
	cache_purge(fvp);
	cache_purge(tvp);

	/*
	 * Release parents first, then children.
	 */
	vrele(odvp);
	if (tvp) {
		if (tvp == ndvp)
			vrele(ndvp);
		else
			vput(ndvp);
		vput(tvp);
	} else
		vput(ndvp);
	vrele(fvp);
	return (error);
}

int
coda_mkdir(struct vop_mkdir_args *ap)
{
	/* true args */
	struct vnode *dvp = ap->a_dvp;
	struct cnode *dcp = VTOC(dvp);
	struct componentname  *cnp = ap->a_cnp;
	struct vattr *va = ap->a_vap;
	struct vnode **vpp = ap->a_vpp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* locals */
	int error;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
	struct cnode *cp;
	CodaFid VFid;
	struct vattr ova;

	MARK_ENTRY(CODA_MKDIR_STATS);

	/*
	 * Check for mkdir of target object.
	 */
	if (IS_CTL_NAME(dvp, nm, len)) {
		*vpp = (struct vnode *)0;
		MARK_INT_FAIL(CODA_MKDIR_STATS);
		return (EACCES);
	}
	if (len+1 > CODA_MAXNAMLEN) {
		*vpp = (struct vnode *)0;
		MARK_INT_FAIL(CODA_MKDIR_STATS);
		return (EACCES);
	}
	error = venus_mkdir(vtomi(dvp), &dcp->c_fid, nm, len, va, cred,
	    td->td_proc, &VFid, &ova);
	if (!error) {
		if (coda_find(&VFid) != NULL)
			panic("cnode existed for newly created directory!");
		cp =  make_coda_node(&VFid, dvp->v_mount, va->va_type);
		*vpp = CTOV(cp);

		/*
		 * Enter the new vnode in the Name Cache.
		 */
		if (cnp->cn_flags & MAKEENTRY)
			cache_enter(dvp, *vpp, cnp);

		/*
		 * Update the attr cache and mark as valid.
		 */
		if (coda_attr_cache) {
			VTOC(*vpp)->c_vattr = ova;
			VTOC(*vpp)->c_flags |= C_VATTR;
		}

		/*
		 * Invalidate the parent's attr cache, the modification time
		 * has changed.
		 */
		VTOC(dvp)->c_flags &= ~C_VATTR;
		vn_lock(*vpp, LK_EXCLUSIVE | LK_RETRY);
		CODADEBUG( CODA_MKDIR, myprintf(("mkdir: %s result %d\n",
		    coda_f2s(&VFid), error)););
	} else {
		*vpp = NULL;
		CODADEBUG(CODA_MKDIR, myprintf(("mkdir error %d\n",error)););
	}
	return (error);
}

int
coda_rmdir(struct vop_rmdir_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct vnode *dvp = ap->a_dvp;
	struct cnode *dcp = VTOC(dvp);
	struct componentname  *cnp = ap->a_cnp;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	/* true args */
	int error;
	const char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
#if 0
	struct cnode *cp;
#endif

	MARK_ENTRY(CODA_RMDIR_STATS);

	/*
	 * Check for rmdir of control object.
	 */
	if (IS_CTL_NAME(dvp, nm, len)) {
		MARK_INT_FAIL(CODA_RMDIR_STATS);
		return (ENOENT);
	}

	/*
	 * Possibly somewhat conservative purging, perhaps we just need to
	 * purge vp?
	 */
	cache_purge(dvp);
	cache_purge(vp);

	/*
	 * Invalidate the parent's attr cache, the modification time has
	 * changed.
	 */
	dcp->c_flags &= ~C_VATTR;
	error = venus_rmdir(vtomi(dvp), &dcp->c_fid, nm, len, cred,
	    td->td_proc);
	CODADEBUG(CODA_RMDIR, myprintf(("in rmdir result %d\n", error)););
	return (error);
}

int
coda_symlink(struct vop_symlink_args *ap)
{
	/* true args */
	struct vnode *tdvp = ap->a_dvp;
	struct cnode *tdcp = VTOC(tdvp);
	struct componentname *cnp = ap->a_cnp;
	struct vattr *tva = ap->a_vap;
	char *path = ap->a_target;
	struct ucred *cred = cnp->cn_cred;
	struct thread *td = cnp->cn_thread;
	struct vnode **vpp = ap->a_vpp;
	/* locals */
	int error;

	/*-
	 * XXX I'm assuming the following things about coda_symlink's
	 * arguments:
	 *       t(foo) is the new name/parent/etc being created.
	 *       lname is the contents of the new symlink.
	 */
	char *nm = cnp->cn_nameptr;
	int len = cnp->cn_namelen;
	int plen = strlen(path);

	/*
	 * Here's the strategy for the moment: perform the symlink, then do a
	 * lookup to grab the resulting vnode.  I know this requires two
	 * communications with Venus for a new sybolic link, but that's the
	 * way the ball bounces.  I don't yet want to change the way the Mach
	 * symlink works.  When Mach support is deprecated, we should change
	 * symlink so that the common case returns the resultant vnode in a
	 * vpp argument.
	 */
	MARK_ENTRY(CODA_SYMLINK_STATS);

	/*
	 * Check for symlink of control object.
	 */
	if (IS_CTL_NAME(tdvp, nm, len)) {
		MARK_INT_FAIL(CODA_SYMLINK_STATS);
		return (EACCES);
	}
	if (plen+1 > CODA_MAXPATHLEN) {
		MARK_INT_FAIL(CODA_SYMLINK_STATS);
		return (EINVAL);
	}
	if (len+1 > CODA_MAXNAMLEN) {
		MARK_INT_FAIL(CODA_SYMLINK_STATS);
		error = EINVAL;
		goto exit;
	}
	error = venus_symlink(vtomi(tdvp), &tdcp->c_fid, path, plen, nm, len,
	    tva, cred, td->td_proc);

	/*
	 * Invalidate the parent's attr cache, the modification time has
	 * changed.
	 */
	tdcp->c_flags &= ~C_VATTR;
	if (error == 0)
		error = VOP_LOOKUP(tdvp, vpp, cnp);
exit:
	CODADEBUG(CODA_SYMLINK, myprintf(("in symlink result %d\n",error)););
	return (error);
}

/*
 * Read directory entries.
 *
 * XXX: This forwards the operator straight to the cache vnode using
 * VOP_READDIR(), rather than calling venus_readdir().  Why?
 */
int
coda_readdir(struct vop_readdir_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	struct uio *uiop = ap->a_uio;
	struct ucred *cred = ap->a_cred;
	int *eofflag = ap->a_eofflag;
	u_long **cookies = ap->a_cookies;
	int *ncookies = ap->a_ncookies;
	struct thread *td = ap->a_uio->uio_td;
	/* upcall decl */
	/* locals */
	int error = 0;
	int opened_internally = 0;

	MARK_ENTRY(CODA_READDIR_STATS);
	CODADEBUG(CODA_READDIR, myprintf(("coda_readdir(%p, %d, %lld, %d)\n",
	    (void *)uiop->uio_iov->iov_base, uiop->uio_resid,
	    (long long)uiop->uio_offset, uiop->uio_segflg)););

	/*
	 * Check for readdir of control object.
	 */
	if (IS_CTL_VP(vp)) {
		MARK_INT_FAIL(CODA_READDIR_STATS);
		return (ENOENT);
	}

	/*
	 * If directory is not already open do an "internal open" on it.
	 *
	 * XXX: Why would this happen?  For files, there's memory mapping,
	 * execution, and other kernel access paths such as ktrace.  For
	 * directories, it is less clear.
	 */
	if (cp->c_ovp == NULL) {
		opened_internally = 1;
		MARK_INT_GEN(CODA_OPEN_STATS);
		error = VOP_OPEN(vp, FREAD, cred, td, NULL);
		printf("coda_readdir: Internally Opening %p\n", vp);
		if (error) {
			printf("coda_readdir: VOP_OPEN on container failed "
			   "%d\n", error);
			return (error);
		}
	}

	/*
	 * Have UFS handle the call.
	 */
	CODADEBUG(CODA_READDIR, myprintf(("indirect readdir: fid = %s, "
	    "refcnt = %d\n", coda_f2s(&cp->c_fid), vp->v_usecount)););
	vn_lock(cp->c_ovp, LK_EXCLUSIVE | LK_RETRY);
	error = VOP_READDIR(cp->c_ovp, uiop, cred, eofflag, ncookies,
	    cookies);
	VOP_UNLOCK(cp->c_ovp, 0);
	if (error)
		MARK_INT_FAIL(CODA_READDIR_STATS);
	else
		MARK_INT_SAT(CODA_READDIR_STATS);

	/*
	 * Do an "internal close" if necessary.
	 */
	if (opened_internally) {
		MARK_INT_GEN(CODA_CLOSE_STATS);
		(void)VOP_CLOSE(vp, FREAD, cred, td);
	}
	return (error);
}

int
coda_reclaim(struct vop_reclaim_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	/* upcall decl */
	/* locals */

	/*
	 * Forced unmount/flush will let vnodes with non-zero use be
	 * destroyed!
	 */
	ENTRY;

	if (IS_UNMOUNTING(cp)) {
#ifdef	DEBUG
		if (VTOC(vp)->c_ovp) {
			if (IS_UNMOUNTING(cp))
				printf("coda_reclaim: c_ovp not void: vp "
				    "%p, cp %p\n", vp, cp);
		}
#endif
	} else {
		if (prtactive && vp->v_usecount != 0)
			vprint("coda_reclaim: pushing active", vp);
	}
	cache_purge(vp);
	coda_free(VTOC(vp));
	vp->v_data = NULL;
	vp->v_object = NULL;
	return (0);
}

int
coda_lock(struct vop_lock1_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	/* upcall decl */
	/* locals */

	ENTRY;
	if ((ap->a_flags & LK_INTERLOCK) == 0) {
		VI_LOCK(vp);
		ap->a_flags |= LK_INTERLOCK;
	}
	if (coda_lockdebug)
		myprintf(("Attempting lock on %s\n", coda_f2s(&cp->c_fid)));
	return (vop_stdlock(ap));
}

int
coda_unlock(struct vop_unlock_args *ap)
{
	/* true args */
	struct vnode *vp = ap->a_vp;
	struct cnode *cp = VTOC(vp);
	/* upcall decl */
	/* locals */

	ENTRY;
	if (coda_lockdebug)
		myprintf(("Attempting unlock on %s\n",
		    coda_f2s(&cp->c_fid)));
	return (vop_stdunlock(ap));
}

int
coda_islocked(struct vop_islocked_args *ap)
{
	/* true args */

	ENTRY;
	return (vop_stdislocked(ap));
}

static void
coda_print_vattr(struct vattr *attr)
{
	char *typestr;

	switch (attr->va_type) {
	case VNON:
		typestr = "VNON";
		break;

	case VREG:
		typestr = "VREG";
		break;

	case VDIR:
		typestr = "VDIR";
		break;

	case VBLK:
		typestr = "VBLK";
		break;

	case VCHR:
		typestr = "VCHR";
		break;

	case VLNK:
		typestr = "VLNK";
		break;

	case VSOCK:
		typestr = "VSCK";
		break;

	case VFIFO:
		typestr = "VFFO";
		break;

	case VBAD:
		typestr = "VBAD";
		break;

	default:
		typestr = "????";
		break;
	}
	myprintf(("attr: type %s mode %d uid %d gid %d fsid %d rdev %d\n",
	    typestr, (int)attr->va_mode, (int)attr->va_uid,
	    (int)attr->va_gid, (int)attr->va_fsid, (int)attr->va_rdev));
	myprintf(("      fileid %d nlink %d size %d blocksize %d bytes %d\n",
	    (int)attr->va_fileid, (int)attr->va_nlink, (int)attr->va_size,
	    (int)attr->va_blocksize,(int)attr->va_bytes));
	myprintf(("      gen %ld flags %ld vaflags %d\n", attr->va_gen,
	    attr->va_flags, attr->va_vaflags));
	myprintf(("      atime sec %d nsec %d\n", (int)attr->va_atime.tv_sec,
	    (int)attr->va_atime.tv_nsec));
	myprintf(("      mtime sec %d nsec %d\n", (int)attr->va_mtime.tv_sec,
	    (int)attr->va_mtime.tv_nsec));
	myprintf(("      ctime sec %d nsec %d\n", (int)attr->va_ctime.tv_sec,
	    (int)attr->va_ctime.tv_nsec));
}

/*
 * How to print a ucred.
 */
void
coda_print_cred(struct ucred *cred)
{
	int i;

	myprintf(("ref %d\tuid %d\n",cred->cr_ref,cred->cr_uid));
	for (i=0; i < cred->cr_ngroups; i++)
		myprintf(("\tgroup %d: (%d)\n",i,cred->cr_groups[i]));
	myprintf(("\n"));
}

/*
 * Return a vnode for the given fid.  If no cnode exists for this fid create
 * one and put it in a table hashed by coda_f2i().  If the cnode for this fid
 * is already in the table return it (ref count is incremented by coda_find.
 * The cnode will be flushed from the table when coda_inactive calls
 * coda_unsave.
 */
struct cnode *
make_coda_node(CodaFid *fid, struct mount *vfsp, short type)
{
	struct cnode *cp;
	struct vnode *vp;
	int err;

	/*
	 * XXXRW: This really needs a moderate amount of reworking.  We need
	 * to properly tolerate failures of getnewvnode() and insmntque(),
	 * and callers need to be able to accept an error back from
	 * make_coda_node.  There may also be more general issues in how we
	 * handle forced unmount.  Finally, if/when Coda loses its dependency
	 * on Giant, the ordering of this needs rethinking.
	 */
	cp = coda_find(fid);
	if (cp != NULL) {
		vref(CTOV(cp));
		return (cp);
	}
	cp = coda_alloc();
	cp->c_fid = *fid;
	err = getnewvnode("coda", vfsp, &coda_vnodeops, &vp);
	if (err)
		panic("coda: getnewvnode returned error %d\n", err);
	vp->v_data = cp;
	vp->v_type = type;
	cp->c_vnode = vp;
	coda_save(cp);
	err = insmntque(vp, vfsp);
	if (err != 0)
		printf("coda: insmntque failed: error %d", err);
	return (cp);
}

int
coda_pathconf(struct vop_pathconf_args *ap)
{

	switch (ap->a_name) {
	case _PC_NAME_MAX:
		*ap->a_retval = CODA_MAXNAMLEN;
		return (0);

	case _PC_PATH_MAX:
		*ap->a_retval = CODA_MAXPATHLEN;
		return (0);

	default:
		return (vop_stdpathconf(ap));
	}
}