21 files changed, 17956 insertions, 0 deletions

diff --git a/sys/nfs/nfs.h b/sys/nfs/nfs.h
new file mode 100644
index 000000000000..89c60944f62e
--- /dev/null
+++ b/sys/nfs/nfs.h
@@ -0,0 +1,566 @@
+/*
+ * Copyright (c) 1989, 1993, 1995
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs.h	8.4 (Berkeley) 5/1/95
+ */
+
+#ifndef _NFS_NFS_H_
+#define _NFS_NFS_H_
+
+/*
+ * Tunable constants for nfs
+ */
+
+#define	NFS_MAXIOVEC	34
+#define NFS_TICKINTVL	5		/* Desired time for a tick (msec) */
+#define NFS_HZ		(hz / nfs_ticks) /* Ticks/sec */
+#define	NFS_TIMEO	(1 * NFS_HZ)	/* Default timeout = 1 second */
+#define	NFS_MINTIMEO	(1 * NFS_HZ)	/* Min timeout to use */
+#define	NFS_MAXTIMEO	(60 * NFS_HZ)	/* Max timeout to backoff to */
+#define	NFS_MINIDEMTIMEO (5 * NFS_HZ)	/* Min timeout for non-idempotent ops*/
+#define	NFS_MAXREXMIT	100		/* Stop counting after this many */
+#define	NFS_MAXWINDOW	1024		/* Max number of outstanding requests */
+#define	NFS_RETRANS	10		/* Num of retrans for soft mounts */
+#define	NFS_MAXGRPS	16		/* Max. size of groups list */
+#ifndef NFS_MINATTRTIMO
+#define	NFS_MINATTRTIMO 5		/* Attribute cache timeout in sec */
+#endif
+#ifndef NFS_MAXATTRTIMO
+#define	NFS_MAXATTRTIMO 60
+#endif
+#define	NFS_WSIZE	8192		/* Def. write data size <= 8192 */
+#define	NFS_RSIZE	8192		/* Def. read data size <= 8192 */
+#define NFS_READDIRSIZE	8192		/* Def. readdir size */
+#define	NFS_DEFRAHEAD	1		/* Def. read ahead # blocks */
+#define	NFS_MAXRAHEAD	4		/* Max. read ahead # blocks */
+#define	NFS_MAXUIDHASH	64		/* Max. # of hashed uid entries/mp */
+#define	NFS_MAXASYNCDAEMON 	20	/* Max. number async_daemons runable */
+#define NFS_MAXGATHERDELAY	100	/* Max. write gather delay (msec) */
+#ifndef NFS_GATHERDELAY
+#define NFS_GATHERDELAY		10	/* Default write gather delay (msec) */
+#endif
+#define	NFS_DIRBLKSIZ	4096		/* Must be a multiple of DIRBLKSIZ */
+
+/*
+ * Oddballs
+ */
+#define	NMOD(a)		((a) % nfs_asyncdaemons)
+#define NFS_CMPFH(n, f, s) \
+	((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s)))
+#define NFS_ISV3(v)	(VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3)
+#define NFS_SRVMAXDATA(n) \
+		(((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \
+		 NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA)
+
+/*
+ * XXX
+ * The B_INVAFTERWRITE flag should be set to whatever is required by the
+ * buffer cache code to say "Invalidate the block after it is written back".
+ */
+#define	B_INVAFTERWRITE	B_INVAL
+
+/*
+ * The IO_METASYNC flag should be implemented for local file systems.
+ * (Until then, it is nothin at all.)
+ */
+#ifndef IO_METASYNC
+#define IO_METASYNC	0
+#endif
+
+/*
+ * Set the attribute timeout based on how recently the file has been modified.
+ */
+#define	NFS_ATTRTIMEO(np) \
+	((((np)->n_flag & NMODIFIED) || \
+	 (time.tv_sec - (np)->n_mtime) / 10 < NFS_MINATTRTIMO) ? NFS_MINATTRTIMO : \
+	 ((time.tv_sec - (np)->n_mtime) / 10 > NFS_MAXATTRTIMO ? NFS_MAXATTRTIMO : \
+	  (time.tv_sec - (np)->n_mtime) / 10))
+
+/*
+ * Expected allocation sizes for major data structures. If the actual size
+ * of the structure exceeds these sizes, then malloc() will be allocating
+ * almost twice the memory required. This is used in nfs_init() to warn
+ * the sysadmin that the size of a structure should be reduced.
+ * (These sizes are always a power of 2. If the kernel malloc() changes
+ *  to one that does not allocate space in powers of 2 size, then this all
+ *  becomes bunk!)
+ */
+#define NFS_NODEALLOC	256
+#define NFS_MNTALLOC	512
+#define NFS_SVCALLOC	256
+#define NFS_UIDALLOC	128
+
+/*
+ * Arguments to mount NFS
+ */
+#define NFS_ARGSVERSION	3		/* change when nfs_args changes */
+struct nfs_args {
+	int		version;	/* args structure version number */
+	struct sockaddr	*addr;		/* file server address */
+	int		addrlen;	/* length of address */
+	int		sotype;		/* Socket type */
+	int		proto;		/* and Protocol */
+	u_char		*fh;		/* File handle to be mounted */
+	int		fhsize;		/* Size, in bytes, of fh */
+	int		flags;		/* flags */
+	int		wsize;		/* write size in bytes */
+	int		rsize;		/* read size in bytes */
+	int		readdirsize;	/* readdir size in bytes */
+	int		timeo;		/* initial timeout in .1 secs */
+	int		retrans;	/* times to retry send */
+	int		maxgrouplist;	/* Max. size of group list */
+	int		readahead;	/* # of blocks to readahead */
+	int		leaseterm;	/* Term (sec) of lease */
+	int		deadthresh;	/* Retrans threshold */
+	char		*hostname;	/* server's name */
+};
+
+/*
+ * NFS mount option flags
+ */
+#define	NFSMNT_SOFT		0x00000001  /* soft mount (hard is default) */
+#define	NFSMNT_WSIZE		0x00000002  /* set write size */
+#define	NFSMNT_RSIZE		0x00000004  /* set read size */
+#define	NFSMNT_TIMEO		0x00000008  /* set initial timeout */
+#define	NFSMNT_RETRANS		0x00000010  /* set number of request retries */
+#define	NFSMNT_MAXGRPS		0x00000020  /* set maximum grouplist size */
+#define	NFSMNT_INT		0x00000040  /* allow interrupts on hard mount */
+#define	NFSMNT_NOCONN		0x00000080  /* Don't Connect the socket */
+#define	NFSMNT_NQNFS		0x00000100  /* Use Nqnfs protocol */
+#define	NFSMNT_NFSV3		0x00000200  /* Use NFS Version 3 protocol */
+#define	NFSMNT_KERB		0x00000400  /* Use Kerberos authentication */
+#define	NFSMNT_DUMBTIMR		0x00000800  /* Don't estimate rtt dynamically */
+#define	NFSMNT_LEASETERM	0x00001000  /* set lease term (nqnfs) */
+#define	NFSMNT_READAHEAD	0x00002000  /* set read ahead */
+#define	NFSMNT_DEADTHRESH	0x00004000  /* set dead server retry thresh */
+#define	NFSMNT_RESVPORT		0x00008000  /* Allocate a reserved port */
+#define	NFSMNT_RDIRPLUS		0x00010000  /* Use Readdirplus for V3 */
+#define	NFSMNT_READDIRSIZE	0x00020000  /* Set readdir size */
+#define	NFSMNT_INTERNAL		0xfffc0000  /* Bits set internally */
+#define NFSMNT_HASWRITEVERF	0x00040000  /* Has write verifier for V3 */
+#define NFSMNT_GOTPATHCONF	0x00080000  /* Got the V3 pathconf info */
+#define NFSMNT_GOTFSINFO	0x00100000  /* Got the V3 fsinfo */
+#define	NFSMNT_MNTD		0x00200000  /* Mnt server for mnt point */
+#define	NFSMNT_DISMINPROG	0x00400000  /* Dismount in progress */
+#define	NFSMNT_DISMNT		0x00800000  /* Dismounted */
+#define	NFSMNT_SNDLOCK		0x01000000  /* Send socket lock */
+#define	NFSMNT_WANTSND		0x02000000  /* Want above */
+#define	NFSMNT_RCVLOCK		0x04000000  /* Rcv socket lock */
+#define	NFSMNT_WANTRCV		0x08000000  /* Want above */
+#define	NFSMNT_WAITAUTH		0x10000000  /* Wait for authentication */
+#define	NFSMNT_HASAUTH		0x20000000  /* Has authenticator */
+#define	NFSMNT_WANTAUTH		0x40000000  /* Wants an authenticator */
+#define	NFSMNT_AUTHERR		0x80000000  /* Authentication error */
+
+/*
+ * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs
+ * should ever try and use it.
+ */
+struct nfsd_args {
+	int	sock;		/* Socket to serve */
+	caddr_t	name;		/* Client addr for connection based sockets */
+	int	namelen;	/* Length of name */
+};
+
+struct nfsd_srvargs {
+	struct nfsd	*nsd_nfsd;	/* Pointer to in kernel nfsd struct */
+	uid_t		nsd_uid;	/* Effective uid mapped to cred */
+	u_long		nsd_haddr;	/* Ip address of client */
+	struct ucred	nsd_cr;		/* Cred. uid maps to */
+	int		nsd_authlen;	/* Length of auth string (ret) */
+	u_char		*nsd_authstr;	/* Auth string (ret) */
+	int		nsd_verflen;	/* and the verfier */
+	u_char		*nsd_verfstr;
+	struct timeval	nsd_timestamp;	/* timestamp from verifier */
+	u_long		nsd_ttl;	/* credential ttl (sec) */
+	NFSKERBKEY_T	nsd_key;	/* Session key */
+};
+
+struct nfsd_cargs {
+	char		*ncd_dirp;	/* Mount dir path */
+	uid_t		ncd_authuid;	/* Effective uid */
+	int		ncd_authtype;	/* Type of authenticator */
+	int		ncd_authlen;	/* Length of authenticator string */
+	u_char		*ncd_authstr;	/* Authenticator string */
+	int		ncd_verflen;	/* and the verifier */
+	u_char		*ncd_verfstr;
+	NFSKERBKEY_T	ncd_key;	/* Session key */
+};
+
+/*
+ * Stats structure
+ */
+struct nfsstats {
+	int	attrcache_hits;
+	int	attrcache_misses;
+	int	lookupcache_hits;
+	int	lookupcache_misses;
+	int	direofcache_hits;
+	int	direofcache_misses;
+	int	biocache_reads;
+	int	read_bios;
+	int	read_physios;
+	int	biocache_writes;
+	int	write_bios;
+	int	write_physios;
+	int	biocache_readlinks;
+	int	readlink_bios;
+	int	biocache_readdirs;
+	int	readdir_bios;
+	int	rpccnt[NFS_NPROCS];
+	int	rpcretries;
+	int	srvrpccnt[NFS_NPROCS];
+	int	srvrpc_errs;
+	int	srv_errs;
+	int	rpcrequests;
+	int	rpctimeouts;
+	int	rpcunexpected;
+	int	rpcinvalid;
+	int	srvcache_inproghits;
+	int	srvcache_idemdonehits;
+	int	srvcache_nonidemdonehits;
+	int	srvcache_misses;
+	int	srvnqnfs_leases;
+	int	srvnqnfs_maxleases;
+	int	srvnqnfs_getleases;
+	int	srvvop_writes;
+};
+
+/*
+ * Flags for nfssvc() system call.
+ */
+#define	NFSSVC_BIOD	0x002
+#define	NFSSVC_NFSD	0x004
+#define	NFSSVC_ADDSOCK	0x008
+#define	NFSSVC_AUTHIN	0x010
+#define	NFSSVC_GOTAUTH	0x040
+#define	NFSSVC_AUTHINFAIL 0x080
+#define	NFSSVC_MNTD	0x100
+
+/*
+ * fs.nfs sysctl(3) identifiers
+ */
+#define NFS_NFSSTATS	1		/* struct: struct nfsstats */
+
+#define FS_NFS_NAMES { \
+		       { 0, 0 }, \
+		       { "nfsstats", CTLTYPE_STRUCT }, \
+}
+
+/*
+ * The set of signals the interrupt an I/O in progress for NFSMNT_INT mounts.
+ * What should be in this set is open to debate, but I believe that since
+ * I/O system calls on ufs are never interrupted by signals the set should
+ * be minimal. My reasoning is that many current programs that use signals
+ * such as SIGALRM will not expect file I/O system calls to be interrupted
+ * by them and break.
+ */
+#ifdef KERNEL
+
+struct uio; struct buf; struct vattr; struct nameidata;	/* XXX */
+
+#define	NFSINT_SIGMASK	(sigmask(SIGINT)|sigmask(SIGTERM)|sigmask(SIGKILL)| \
+			 sigmask(SIGHUP)|sigmask(SIGQUIT))
+
+/*
+ * Socket errors ignored for connectionless sockets??
+ * For now, ignore them all
+ */
+#define	NFSIGNORE_SOERROR(s, e) \
+		((e) != EINTR && (e) != ERESTART && (e) != EWOULDBLOCK && \
+		((s) & PR_CONNREQUIRED) == 0)
+
+/*
+ * Nfs outstanding request list element
+ */
+struct nfsreq {
+	TAILQ_ENTRY(nfsreq) r_chain;
+	struct mbuf	*r_mreq;
+	struct mbuf	*r_mrep;
+	struct mbuf	*r_md;
+	caddr_t		r_dpos;
+	struct nfsmount *r_nmp;
+	struct vnode	*r_vp;
+	u_long		r_xid;
+	int		r_flags;	/* flags on request, see below */
+	int		r_retry;	/* max retransmission count */
+	int		r_rexmit;	/* current retrans count */
+	int		r_timer;	/* tick counter on reply */
+	int		r_procnum;	/* NFS procedure number */
+	int		r_rtt;		/* RTT for rpc */
+	struct proc	*r_procp;	/* Proc that did I/O system call */
+};
+
+/*
+ * Queue head for nfsreq's
+ */
+TAILQ_HEAD(, nfsreq) nfs_reqq;
+
+/* Flag values for r_flags */
+#define R_TIMING	0x01		/* timing request (in mntp) */
+#define R_SENT		0x02		/* request has been sent */
+#define	R_SOFTTERM	0x04		/* soft mnt, too many retries */
+#define	R_INTR		0x08		/* intr mnt, signal pending */
+#define	R_SOCKERR	0x10		/* Fatal error on socket */
+#define	R_TPRINTFMSG	0x20		/* Did a tprintf msg. */
+#define	R_MUSTRESEND	0x40		/* Must resend request */
+#define	R_GETONEREP	0x80		/* Probe for one reply only */
+
+/*
+ * A list of nfssvc_sock structures is maintained with all the sockets
+ * that require service by the nfsd.
+ * The nfsuid structs hang off of the nfssvc_sock structs in both lru
+ * and uid hash lists.
+ */
+#ifndef NFS_UIDHASHSIZ
+#define	NFS_UIDHASHSIZ	29	/* Tune the size of nfssvc_sock with this */
+#endif
+#define	NUIDHASH(sock, uid) \
+	(&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ])
+#ifndef NFS_WDELAYHASHSIZ
+#define	NFS_WDELAYHASHSIZ 16	/* and with this */
+#endif
+#define	NWDELAYHASH(sock, f) \
+	(&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ])
+#ifndef NFS_MUIDHASHSIZ
+#define NFS_MUIDHASHSIZ	67	/* Tune the size of nfsmount with this */
+#endif
+#define	NMUIDHASH(nmp, uid) \
+	(&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ])
+#define	NFSNOHASH(fhsum) \
+	(&nfsnodehashtbl[(fhsum) & nfsnodehash])
+
+/*
+ * Network address hash list element
+ */
+union nethostaddr {
+	u_long had_inetaddr;
+	struct mbuf *had_nam;
+};
+
+struct nfsuid {
+	TAILQ_ENTRY(nfsuid) nu_lru;	/* LRU chain */
+	LIST_ENTRY(nfsuid) nu_hash;	/* Hash list */
+	int		nu_flag;	/* Flags */
+	union nethostaddr nu_haddr;	/* Host addr. for dgram sockets */
+	struct ucred	nu_cr;		/* Cred uid mapped to */
+	int		nu_expire;	/* Expiry time (sec) */
+	struct timeval	nu_timestamp;	/* Kerb. timestamp */
+	u_long		nu_nickname;	/* Nickname on server */
+	NFSKERBKEY_T	nu_key;		/* and session key */
+};
+
+#define	nu_inetaddr	nu_haddr.had_inetaddr
+#define	nu_nam		nu_haddr.had_nam
+/* Bits for nu_flag */
+#define	NU_INETADDR	0x1
+#define NU_NAM		0x2
+#define NU_NETFAM(u)	(((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO)
+
+struct nfssvc_sock {
+	TAILQ_ENTRY(nfssvc_sock) ns_chain;	/* List of all nfssvc_sock's */
+	TAILQ_HEAD(, nfsuid) ns_uidlruhead;
+	struct file	*ns_fp;
+	struct socket	*ns_so;
+	struct mbuf	*ns_nam;
+	struct mbuf	*ns_raw;
+	struct mbuf	*ns_rawend;
+	struct mbuf	*ns_rec;
+	struct mbuf	*ns_recend;
+	struct mbuf	*ns_frag;
+	int		ns_flag;
+	int		ns_solock;
+	int		ns_cc;
+	int		ns_reclen;
+	int		ns_numuids;
+	u_long		ns_sref;
+	LIST_HEAD(, nfsrv_descript) ns_tq;	/* Write gather lists */
+	LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ];
+	LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ];
+};
+
+/* Bits for "ns_flag" */
+#define	SLP_VALID	0x01
+#define	SLP_DOREC	0x02
+#define	SLP_NEEDQ	0x04
+#define	SLP_DISCONN	0x08
+#define	SLP_GETSTREAM	0x10
+#define	SLP_LASTFRAG	0x20
+#define SLP_ALLFLAGS	0xff
+
+TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead;
+int nfssvc_sockhead_flag;
+#define	SLP_INIT	0x01
+#define	SLP_WANTINIT	0x02
+
+/*
+ * One of these structures is allocated for each nfsd.
+ */
+struct nfsd {
+	TAILQ_ENTRY(nfsd) nfsd_chain;	/* List of all nfsd's */
+	int		nfsd_flag;	/* NFSD_ flags */
+	struct nfssvc_sock *nfsd_slp;	/* Current socket */
+	int		nfsd_authlen;	/* Authenticator len */
+	u_char		nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */
+	int		nfsd_verflen;	/* and the Verifier */
+	u_char		nfsd_verfstr[RPCVERF_MAXSIZ];
+	struct proc	*nfsd_procp;	/* Proc ptr */
+	struct nfsrv_descript *nfsd_nd;	/* Associated nfsrv_descript */
+};
+
+/* Bits for "nfsd_flag" */
+#define	NFSD_WAITING	0x01
+#define	NFSD_REQINPROG	0x02
+#define	NFSD_NEEDAUTH	0x04
+#define	NFSD_AUTHFAIL	0x08
+
+/*
+ * This structure is used by the server for describing each request.
+ * Some fields are used only when write request gathering is performed.
+ */
+struct nfsrv_descript {
+	u_quad_t		nd_time;	/* Write deadline (usec) */
+	off_t			nd_off;		/* Start byte offset */
+	off_t			nd_eoff;	/* and end byte offset */
+	LIST_ENTRY(nfsrv_descript) nd_hash;	/* Hash list */
+	LIST_ENTRY(nfsrv_descript) nd_tq;		/* and timer list */
+	LIST_HEAD(,nfsrv_descript) nd_coalesce;	/* coalesced writes */
+	struct mbuf		*nd_mrep;	/* Request mbuf list */
+	struct mbuf		*nd_md;		/* Current dissect mbuf */
+	struct mbuf		*nd_mreq;	/* Reply mbuf list */
+	struct mbuf		*nd_nam;	/* and socket addr */
+	struct mbuf		*nd_nam2;	/* return socket addr */
+	caddr_t			nd_dpos;	/* Current dissect pos */
+	int			nd_procnum;	/* RPC # */
+	int			nd_stable;	/* storage type */
+	int			nd_flag;	/* nd_flag */
+	int			nd_len;		/* Length of this write */
+	int			nd_repstat;	/* Reply status */
+	u_long			nd_retxid;	/* Reply xid */
+	u_long			nd_duration;	/* Lease duration */
+	struct timeval		nd_starttime;	/* Time RPC initiated */
+	fhandle_t		nd_fh;		/* File handle */
+	struct ucred		nd_cr;		/* Credentials */
+};
+
+/* Bits for "nd_flag" */
+#define	ND_READ		LEASE_READ
+#define ND_WRITE	LEASE_WRITE
+#define ND_CHECK	0x04
+#define ND_LEASE	(ND_READ | ND_WRITE | ND_CHECK)
+#define ND_NFSV3	0x08
+#define ND_NQNFS	0x10
+#define ND_KERBNICK	0x20
+#define ND_KERBFULL	0x40
+#define ND_KERBAUTH	(ND_KERBNICK | ND_KERBFULL)
+
+TAILQ_HEAD(, nfsd) nfsd_head;
+int nfsd_head_flag;
+#define	NFSD_CHECKSLP	0x01
+
+/*
+ * These macros compare nfsrv_descript structures.
+ */
+#define NFSW_CONTIG(o, n) \
+		((o)->nd_eoff >= (n)->nd_off && \
+		 !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH))
+
+#define NFSW_SAMECRED(o, n) \
+	(((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \
+ 	 !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \
+		sizeof (struct ucred)))
+
+int	nfs_reply __P((struct nfsreq *));
+int	nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int));
+int	nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *));
+int	nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *));
+int	nfs_sndlock __P((int *,struct nfsreq *));
+int	nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *));
+int	nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int));
+int	nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *));
+int	nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *));
+int	nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *));
+int	nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *));
+int	nfs_asyncio __P((struct buf *,struct ucred *));
+int	nfs_doio __P((struct buf *,struct ucred *,struct proc *));
+int	nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *));
+int	nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *));
+int	nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *));
+int	nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *));
+void	nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *));
+void	nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **));
+void	nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **));
+int	nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int));
+int	nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *));
+int	netaddr_match __P((int,union nethostaddr *,struct mbuf *));
+int	nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *));
+int	nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *));
+int	nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int));
+void	nfsm_adj __P((struct mbuf *,int,int));
+int	nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *));
+void	nfsrv_initcache __P((void));
+int	nfs_rcvlock __P((struct nfsreq *));
+int	nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T));
+int	nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int));
+int	nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *));
+int	nfs_msg __P((struct proc *,char *,char *));
+int	nfs_adv __P((struct mbuf **,caddr_t *,int,int));
+int	nfsrv_getstream __P((struct nfssvc_sock *,int));
+void	nfs_nhinit __P((void));
+void	nfs_timer __P((void*));
+u_long nfs_hash __P((nfsfh_t *,int));
+int	nfssvc_iod __P((struct proc *));
+int	nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *));
+int	nfssvc_addsock __P((struct file *,struct mbuf *));
+int	nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **));
+int	nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **));
+void	nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *));
+int	mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **));
+int	nfs_connect __P((struct nfsmount *,struct nfsreq *));
+int	nfs_getattrcache __P((struct vnode *,struct vattr *));
+int	nfsm_strtmbuf __P((struct mbuf **,char **,char *,long));
+int	nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *));
+int	nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *));
+void	nfsrv_init __P((int));
+void	nfs_clearcommit __P((struct mount *));
+int	nfsrv_errmap __P((struct nfsrv_descript *, int));
+void	nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *));
+void	nfsrvw_sort __P((gid_t [],int));
+void	nfsrv_setcred __P((struct ucred *,struct ucred *));
+int	nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int));
+int	nfs_writebp __P((struct buf *,int));
+#endif	/* KERNEL */
+
+#endif
diff --git a/sys/nfs/nfs_bio.c b/sys/nfs/nfs_bio.c
new file mode 100644
index 000000000000..2cce1fdd3a26
--- /dev/null
+++ b/sys/nfs/nfs_bio.c
@@ -0,0 +1,903 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_bio.c	8.9 (Berkeley) 3/30/95
+ */
+
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/resourcevar.h>
+#include <sys/signalvar.h>
+#include <sys/proc.h>
+#include <sys/buf.h>
+#include <sys/vnode.h>
+#include <sys/trace.h>
+#include <sys/mount.h>
+#include <sys/kernel.h>
+
+#include <vm/vm.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsmount.h>
+#include <nfs/nqnfs.h>
+#include <nfs/nfsnode.h>
+
+struct buf *nfs_getcacheblk();
+extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON];
+extern int nfs_numasync;
+extern struct nfsstats nfsstats;
+
+/*
+ * Vnode op for read using bio
+ * Any similarity to readip() is purely coincidental
+ */
+int
+nfs_bioread(vp, uio, ioflag, cred)
+	register struct vnode *vp;
+	register struct uio *uio;
+	int ioflag;
+	struct ucred *cred;
+{
+	register struct nfsnode *np = VTONFS(vp);
+	register int biosize, diff, i;
+	struct buf *bp = 0, *rabp;
+	struct vattr vattr;
+	struct proc *p;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	daddr_t lbn, bn, bn2, rabn;
+	caddr_t baddr;
+	int got_buf = 0, nra, error = 0, n = 0, on = 0, not_readin;
+	nfsquad_t tquad;
+
+#ifdef DIAGNOSTIC
+	if (uio->uio_rw != UIO_READ)
+		panic("nfs_read mode");
+#endif
+	if (uio->uio_resid == 0)
+		return (0);
+	if (uio->uio_offset < 0)
+		return (EINVAL);
+	p = uio->uio_procp;
+	if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3)
+		(void)nfs_fsinfo(nmp, vp, cred, p);
+	biosize = nmp->nm_rsize;
+	/*
+	 * For nfs, cache consistency can only be maintained approximately.
+	 * Although RFC1094 does not specify the criteria, the following is
+	 * believed to be compatible with the reference port.
+	 * For nqnfs, full cache consistency is maintained within the loop.
+	 * For nfs:
+	 * If the file's modify time on the server has changed since the
+	 * last read rpc or you have written to the file,
+	 * you may have lost data cache consistency with the
+	 * server, so flush all of the file's data out of the cache.
+	 * Then force a getattr rpc to ensure that you have up to date
+	 * attributes.
+	 * NB: This implies that cache data can be read when up to
+	 * NFS_ATTRTIMEO seconds out of date. If you find that you need current
+	 * attributes this could be forced by setting n_attrstamp to 0 before
+	 * the VOP_GETATTR() call.
+	 */
+	if ((nmp->nm_flag & NFSMNT_NQNFS) == 0 && vp->v_type != VLNK) {
+		if (np->n_flag & NMODIFIED) {
+			if (vp->v_type != VREG) {
+				if (vp->v_type != VDIR)
+					panic("nfs: bioread, not dir");
+				nfs_invaldir(vp);
+				error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+				if (error)
+					return (error);
+			}
+			np->n_attrstamp = 0;
+			error = VOP_GETATTR(vp, &vattr, cred, p);
+			if (error)
+				return (error);
+			np->n_mtime = vattr.va_mtime.ts_sec;
+		} else {
+			error = VOP_GETATTR(vp, &vattr, cred, p);
+			if (error)
+				return (error);
+			if (np->n_mtime != vattr.va_mtime.ts_sec) {
+				if (vp->v_type == VDIR)
+					nfs_invaldir(vp);
+				error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+				if (error)
+					return (error);
+				np->n_mtime = vattr.va_mtime.ts_sec;
+			}
+		}
+	}
+	do {
+
+	    /*
+	     * Get a valid lease. If cached data is stale, flush it.
+	     */
+	    if (nmp->nm_flag & NFSMNT_NQNFS) {
+		if (NQNFS_CKINVALID(vp, np, ND_READ)) {
+		    do {
+			error = nqnfs_getlease(vp, ND_READ, cred, p);
+		    } while (error == NQNFS_EXPIRED);
+		    if (error)
+			return (error);
+		    if (np->n_lrev != np->n_brev ||
+			(np->n_flag & NQNFSNONCACHE) ||
+			((np->n_flag & NMODIFIED) && vp->v_type == VDIR)) {
+			if (vp->v_type == VDIR)
+			    nfs_invaldir(vp);
+			error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+			if (error)
+			    return (error);
+			np->n_brev = np->n_lrev;
+		    }
+		} else if (vp->v_type == VDIR && (np->n_flag & NMODIFIED)) {
+		    nfs_invaldir(vp);
+		    error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+		    if (error)
+			return (error);
+		}
+	    }
+	    if (np->n_flag & NQNFSNONCACHE) {
+		switch (vp->v_type) {
+		case VREG:
+			return (nfs_readrpc(vp, uio, cred));
+		case VLNK:
+			return (nfs_readlinkrpc(vp, uio, cred));
+		case VDIR:
+			break;
+		default:
+			printf(" NQNFSNONCACHE: type %x unexpected\n",	
+				vp->v_type);
+		};
+	    }
+	    baddr = (caddr_t)0;
+	    switch (vp->v_type) {
+	    case VREG:
+		nfsstats.biocache_reads++;
+		lbn = uio->uio_offset / biosize;
+		on = uio->uio_offset & (biosize - 1);
+		bn = lbn * (biosize / DEV_BSIZE);
+		not_readin = 1;
+
+		/*
+		 * Start the read ahead(s), as required.
+		 */
+		if (nfs_numasync > 0 && nmp->nm_readahead > 0) {
+		    for (nra = 0; nra < nmp->nm_readahead &&
+			(lbn + 1 + nra) * biosize < np->n_size; nra++) {
+			rabn = (lbn + 1 + nra) * (biosize / DEV_BSIZE);
+			if (!incore(vp, rabn)) {
+			    rabp = nfs_getcacheblk(vp, rabn, biosize, p);
+			    if (!rabp)
+				return (EINTR);
+			    if ((rabp->b_flags & (B_DELWRI | B_DONE)) == 0) {
+				rabp->b_flags |= (B_READ | B_ASYNC);
+				if (nfs_asyncio(rabp, cred)) {
+				    rabp->b_flags |= B_INVAL;
+				    brelse(rabp);
+				}
+			    } else
+				brelse(rabp);
+			}
+		    }
+		}
+
+		/*
+		 * If the block is in the cache and has the required data
+		 * in a valid region, just copy it out.
+		 * Otherwise, get the block and write back/read in,
+		 * as required.
+		 */
+		if ((bp = incore(vp, bn)) &&
+		    (bp->b_flags & (B_BUSY | B_WRITEINPROG)) ==
+		    (B_BUSY | B_WRITEINPROG))
+			got_buf = 0;
+		else {
+again:
+			bp = nfs_getcacheblk(vp, bn, biosize, p);
+			if (!bp)
+				return (EINTR);
+			got_buf = 1;
+			if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
+				bp->b_flags |= B_READ;
+				not_readin = 0;
+				error = nfs_doio(bp, cred, p);
+				if (error) {
+				    brelse(bp);
+				    return (error);
+				}
+			}
+		}
+		n = min((unsigned)(biosize - on), uio->uio_resid);
+		diff = np->n_size - uio->uio_offset;
+		if (diff < n)
+			n = diff;
+		if (not_readin && n > 0) {
+			if (on < bp->b_validoff || (on + n) > bp->b_validend) {
+				if (!got_buf) {
+				    bp = nfs_getcacheblk(vp, bn, biosize, p);
+				    if (!bp)
+					return (EINTR);
+				    got_buf = 1;
+				}
+				bp->b_flags |= B_INVAFTERWRITE;
+				if (bp->b_dirtyend > 0) {
+				    if ((bp->b_flags & B_DELWRI) == 0)
+					panic("nfsbioread");
+				    if (VOP_BWRITE(bp) == EINTR)
+					return (EINTR);
+				} else
+				    brelse(bp);
+				goto again;
+			}
+		}
+		vp->v_lastr = lbn;
+		diff = (on >= bp->b_validend) ? 0 : (bp->b_validend - on);
+		if (diff < n)
+			n = diff;
+		break;
+	    case VLNK:
+		nfsstats.biocache_readlinks++;
+		bp = nfs_getcacheblk(vp, (daddr_t)0, NFS_MAXPATHLEN, p);
+		if (!bp)
+			return (EINTR);
+		if ((bp->b_flags & B_DONE) == 0) {
+			bp->b_flags |= B_READ;
+			error = nfs_doio(bp, cred, p);
+			if (error) {
+				brelse(bp);
+				return (error);
+			}
+		}
+		n = min(uio->uio_resid, NFS_MAXPATHLEN - bp->b_resid);
+		got_buf = 1;
+		on = 0;
+		break;
+	    case VDIR:
+		nfsstats.biocache_readdirs++;
+		lbn = uio->uio_offset / NFS_DIRBLKSIZ;
+		on = uio->uio_offset & (NFS_DIRBLKSIZ - 1);
+		bp = nfs_getcacheblk(vp, lbn, NFS_DIRBLKSIZ, p);
+		if (!bp)
+		    return (EINTR);
+		if ((bp->b_flags & B_DONE) == 0) {
+		    bp->b_flags |= B_READ;
+		    error = nfs_doio(bp, cred, p);
+		    if (error) {
+			brelse(bp);
+			while (error == NFSERR_BAD_COOKIE) {
+			    nfs_invaldir(vp);
+			    error = nfs_vinvalbuf(vp, 0, cred, p, 1);
+			    /*
+			     * Yuck! The directory has been modified on the
+			     * server. The only way to get the block is by
+			     * reading from the beginning to get all the
+			     * offset cookies.
+			     */
+			    for (i = 0; i <= lbn && !error; i++) {
+				bp = nfs_getcacheblk(vp, i, NFS_DIRBLKSIZ, p);
+				if (!bp)
+				    return (EINTR);
+				if ((bp->b_flags & B_DONE) == 0) {
+				    bp->b_flags |= B_READ;
+				    error = nfs_doio(bp, cred, p);
+				    if (error)
+					brelse(bp);
+				}
+			    }
+			}
+			if (error)
+			    return (error);
+		    }
+		}
+
+		/*
+		 * If not eof and read aheads are enabled, start one.
+		 * (You need the current block first, so that you have the
+		 *  directory offset cookie of the next block.)
+		 */
+		if (nfs_numasync > 0 && nmp->nm_readahead > 0 &&
+		    (np->n_direofoffset == 0 ||
+		    (lbn + 1) * NFS_DIRBLKSIZ < np->n_direofoffset) &&
+		    !(np->n_flag & NQNFSNONCACHE) &&
+		    !incore(vp, lbn + 1)) {
+			rabp = nfs_getcacheblk(vp, lbn + 1, NFS_DIRBLKSIZ, p);
+			if (rabp) {
+			    if ((rabp->b_flags & (B_DONE | B_DELWRI)) == 0) {
+				rabp->b_flags |= (B_READ | B_ASYNC);
+				if (nfs_asyncio(rabp, cred)) {
+				    rabp->b_flags |= B_INVAL;
+				    brelse(rabp);
+				}
+			    } else
+				brelse(rabp);
+			}
+		}
+		n = min(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid - on);
+		got_buf = 1;
+		break;
+	    default:
+		printf(" nfsbioread: type %x unexpected\n",vp->v_type);
+		break;
+	    };
+
+	    if (n > 0) {
+		if (!baddr)
+			baddr = bp->b_data;
+		error = uiomove(baddr + on, (int)n, uio);
+	    }
+	    switch (vp->v_type) {
+	    case VREG:
+		break;
+	    case VLNK:
+		n = 0;
+		break;
+	    case VDIR:
+		if (np->n_flag & NQNFSNONCACHE)
+			bp->b_flags |= B_INVAL;
+		break;
+	    default:
+		printf(" nfsbioread: type %x unexpected\n",vp->v_type);
+	    }
+	    if (got_buf)
+		brelse(bp);
+	} while (error == 0 && uio->uio_resid > 0 && n > 0);
+	return (error);
+}
+
+/*
+ * Vnode op for write using bio
+ */
+int
+nfs_write(ap)
+	struct vop_write_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register int biosize;
+	register struct uio *uio = ap->a_uio;
+	struct proc *p = uio->uio_procp;
+	register struct vnode *vp = ap->a_vp;
+	struct nfsnode *np = VTONFS(vp);
+	register struct ucred *cred = ap->a_cred;
+	int ioflag = ap->a_ioflag;
+	struct buf *bp;
+	struct vattr vattr;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	daddr_t lbn, bn;
+	int n, on, error = 0, iomode, must_commit;
+
+#ifdef DIAGNOSTIC
+	if (uio->uio_rw != UIO_WRITE)
+		panic("nfs_write mode");
+	if (uio->uio_segflg == UIO_USERSPACE && uio->uio_procp != curproc)
+		panic("nfs_write proc");
+#endif
+	if (vp->v_type != VREG)
+		return (EIO);
+	if (np->n_flag & NWRITEERR) {
+		np->n_flag &= ~NWRITEERR;
+		return (np->n_error);
+	}
+	if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3)
+		(void)nfs_fsinfo(nmp, vp, cred, p);
+	if (ioflag & (IO_APPEND | IO_SYNC)) {
+		if (np->n_flag & NMODIFIED) {
+			np->n_attrstamp = 0;
+			error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+			if (error)
+				return (error);
+		}
+		if (ioflag & IO_APPEND) {
+			np->n_attrstamp = 0;
+			error = VOP_GETATTR(vp, &vattr, cred, p);
+			if (error)
+				return (error);
+			uio->uio_offset = np->n_size;
+		}
+	}
+	if (uio->uio_offset < 0)
+		return (EINVAL);
+	if (uio->uio_resid == 0)
+		return (0);
+	/*
+	 * Maybe this should be above the vnode op call, but so long as
+	 * file servers have no limits, i don't think it matters
+	 */
+	if (p && uio->uio_offset + uio->uio_resid >
+	      p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
+		psignal(p, SIGXFSZ);
+		return (EFBIG);
+	}
+	/*
+	 * I use nm_rsize, not nm_wsize so that all buffer cache blocks
+	 * will be the same size within a filesystem. nfs_writerpc will
+	 * still use nm_wsize when sizing the rpc's.
+	 */
+	biosize = nmp->nm_rsize;
+	do {
+
+		/*
+		 * XXX make sure we aren't cached in the VM page cache
+		 */
+		(void)vnode_pager_uncache(vp);
+
+		/*
+		 * Check for a valid write lease.
+		 */
+		if ((nmp->nm_flag & NFSMNT_NQNFS) &&
+		    NQNFS_CKINVALID(vp, np, ND_WRITE)) {
+			do {
+				error = nqnfs_getlease(vp, ND_WRITE, cred, p);
+			} while (error == NQNFS_EXPIRED);
+			if (error)
+				return (error);
+			if (np->n_lrev != np->n_brev ||
+			    (np->n_flag & NQNFSNONCACHE)) {
+				error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+				if (error)
+					return (error);
+				np->n_brev = np->n_lrev;
+			}
+		}
+		if ((np->n_flag & NQNFSNONCACHE) && uio->uio_iovcnt == 1) {
+		    iomode = NFSV3WRITE_FILESYNC;
+		    error = nfs_writerpc(vp, uio, cred, &iomode, &must_commit);
+		    if (must_commit)
+			nfs_clearcommit(vp->v_mount);
+		    return (error);
+		}
+		nfsstats.biocache_writes++;
+		lbn = uio->uio_offset / biosize;
+		on = uio->uio_offset & (biosize-1);
+		n = min((unsigned)(biosize - on), uio->uio_resid);
+		bn = lbn * (biosize / DEV_BSIZE);
+again:
+		bp = nfs_getcacheblk(vp, bn, biosize, p);
+		if (!bp)
+			return (EINTR);
+		if (bp->b_wcred == NOCRED) {
+			crhold(cred);
+			bp->b_wcred = cred;
+		}
+		np->n_flag |= NMODIFIED;
+		if (uio->uio_offset + n > np->n_size) {
+			np->n_size = uio->uio_offset + n;
+			vnode_pager_setsize(vp, (u_long)np->n_size);
+		}
+
+		/*
+		 * If the new write will leave a contiguous dirty
+		 * area, just update the b_dirtyoff and b_dirtyend,
+		 * otherwise force a write rpc of the old dirty area.
+		 */
+		if (bp->b_dirtyend > 0 &&
+		    (on > bp->b_dirtyend || (on + n) < bp->b_dirtyoff)) {
+			bp->b_proc = p;
+			if (VOP_BWRITE(bp) == EINTR)
+				return (EINTR);
+			goto again;
+		}
+
+		/*
+		 * Check for valid write lease and get one as required.
+		 * In case getblk() and/or bwrite() delayed us.
+		 */
+		if ((nmp->nm_flag & NFSMNT_NQNFS) &&
+		    NQNFS_CKINVALID(vp, np, ND_WRITE)) {
+			do {
+				error = nqnfs_getlease(vp, ND_WRITE, cred, p);
+			} while (error == NQNFS_EXPIRED);
+			if (error) {
+				brelse(bp);
+				return (error);
+			}
+			if (np->n_lrev != np->n_brev ||
+			    (np->n_flag & NQNFSNONCACHE)) {
+				brelse(bp);
+				error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+				if (error)
+					return (error);
+				np->n_brev = np->n_lrev;
+				goto again;
+			}
+		}
+		error = uiomove((char *)bp->b_data + on, n, uio);
+		if (error) {
+			bp->b_flags |= B_ERROR;
+			brelse(bp);
+			return (error);
+		}
+		if (bp->b_dirtyend > 0) {
+			bp->b_dirtyoff = min(on, bp->b_dirtyoff);
+			bp->b_dirtyend = max((on + n), bp->b_dirtyend);
+		} else {
+			bp->b_dirtyoff = on;
+			bp->b_dirtyend = on + n;
+		}
+		if (bp->b_validend == 0 || bp->b_validend < bp->b_dirtyoff ||
+		    bp->b_validoff > bp->b_dirtyend) {
+			bp->b_validoff = bp->b_dirtyoff;
+			bp->b_validend = bp->b_dirtyend;
+		} else {
+			bp->b_validoff = min(bp->b_validoff, bp->b_dirtyoff);
+			bp->b_validend = max(bp->b_validend, bp->b_dirtyend);
+		}
+		/*
+		 * If the lease is non-cachable or IO_SYNC do bwrite().
+		 */
+		if ((np->n_flag & NQNFSNONCACHE) || (ioflag & IO_SYNC)) {
+			bp->b_proc = p;
+			error = VOP_BWRITE(bp);
+			if (error)
+				return (error);
+			if (np->n_flag & NQNFSNONCACHE) {
+				error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
+				if (error)
+					return (error);
+			}
+		} else if ((n + on) == biosize &&
+			(nmp->nm_flag & NFSMNT_NQNFS) == 0) {
+			bp->b_proc = (struct proc *)0;
+			bp->b_flags |= B_ASYNC;
+			(void)nfs_writebp(bp, 0);
+		} else
+			bdwrite(bp);
+	} while (uio->uio_resid > 0 && n > 0);
+	return (0);
+}
+
+/*
+ * Get an nfs cache block.
+ * Allocate a new one if the block isn't currently in the cache
+ * and return the block marked busy. If the calling process is
+ * interrupted by a signal for an interruptible mount point, return
+ * NULL.
+ */
+struct buf *
+nfs_getcacheblk(vp, bn, size, p)
+	struct vnode *vp;
+	daddr_t bn;
+	int size;
+	struct proc *p;
+{
+	register struct buf *bp;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+
+	if (nmp->nm_flag & NFSMNT_INT) {
+		bp = getblk(vp, bn, size, PCATCH, 0);
+		while (bp == (struct buf *)0) {
+			if (nfs_sigintr(nmp, (struct nfsreq *)0, p))
+				return ((struct buf *)0);
+			bp = getblk(vp, bn, size, 0, 2 * hz);
+		}
+	} else
+		bp = getblk(vp, bn, size, 0, 0);
+	return (bp);
+}
+
+/*
+ * Flush and invalidate all dirty buffers. If another process is already
+ * doing the flush, just wait for completion.
+ */
+int
+nfs_vinvalbuf(vp, flags, cred, p, intrflg)
+	struct vnode *vp;
+	int flags;
+	struct ucred *cred;
+	struct proc *p;
+	int intrflg;
+{
+	register struct nfsnode *np = VTONFS(vp);
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	int error = 0, slpflag, slptimeo;
+
+	if ((nmp->nm_flag & NFSMNT_INT) == 0)
+		intrflg = 0;
+	if (intrflg) {
+		slpflag = PCATCH;
+		slptimeo = 2 * hz;
+	} else {
+		slpflag = 0;
+		slptimeo = 0;
+	}
+	/*
+	 * First wait for any other process doing a flush to complete.
+	 */
+	while (np->n_flag & NFLUSHINPROG) {
+		np->n_flag |= NFLUSHWANT;
+		error = tsleep((caddr_t)&np->n_flag, PRIBIO + 2, "nfsvinval",
+			slptimeo);
+		if (error && intrflg && nfs_sigintr(nmp, (struct nfsreq *)0, p))
+			return (EINTR);
+	}
+
+	/*
+	 * Now, flush as required.
+	 */
+	np->n_flag |= NFLUSHINPROG;
+	error = vinvalbuf(vp, flags, cred, p, slpflag, 0);
+	while (error) {
+		if (intrflg && nfs_sigintr(nmp, (struct nfsreq *)0, p)) {
+			np->n_flag &= ~NFLUSHINPROG;
+			if (np->n_flag & NFLUSHWANT) {
+				np->n_flag &= ~NFLUSHWANT;
+				wakeup((caddr_t)&np->n_flag);
+			}
+			return (EINTR);
+		}
+		error = vinvalbuf(vp, flags, cred, p, 0, slptimeo);
+	}
+	np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
+	if (np->n_flag & NFLUSHWANT) {
+		np->n_flag &= ~NFLUSHWANT;
+		wakeup((caddr_t)&np->n_flag);
+	}
+	return (0);
+}
+
+/*
+ * Initiate asynchronous I/O. Return an error if no nfsiods are available.
+ * This is mainly to avoid queueing async I/O requests when the nfsiods
+ * are all hung on a dead server.
+ */
+int
+nfs_asyncio(bp, cred)
+	register struct buf *bp;
+	struct ucred *cred;
+{
+	register int i;
+
+	if (nfs_numasync == 0)
+		return (EIO);
+	for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
+	    if (nfs_iodwant[i]) {
+		if (bp->b_flags & B_READ) {
+			if (bp->b_rcred == NOCRED && cred != NOCRED) {
+				crhold(cred);
+				bp->b_rcred = cred;
+			}
+		} else {
+			bp->b_flags |= B_WRITEINPROG;
+			if (bp->b_wcred == NOCRED && cred != NOCRED) {
+				crhold(cred);
+				bp->b_wcred = cred;
+			}
+		}
+	
+		TAILQ_INSERT_TAIL(&nfs_bufq, bp, b_freelist);
+		nfs_iodwant[i] = (struct proc *)0;
+		wakeup((caddr_t)&nfs_iodwant[i]);
+		return (0);
+	    }
+
+	/*
+	 * If it is a read or a write already marked B_WRITEINPROG or B_NOCACHE
+	 * return EIO so the process will call nfs_doio() and do it
+	 * synchronously.
+	 */
+	if (bp->b_flags & (B_READ | B_WRITEINPROG | B_NOCACHE))
+		return (EIO);
+
+	/*
+	 * Just turn the async write into a delayed write, instead of
+	 * doing in synchronously. Hopefully, at least one of the nfsiods
+	 * is currently doing a write for this file and will pick up the
+	 * delayed writes before going back to sleep.
+	 */
+	bp->b_flags |= B_DELWRI;
+	reassignbuf(bp, bp->b_vp);
+	biodone(bp);
+	return (0);
+}
+
+/*
+ * Do an I/O operation to/from a cache block. This may be called
+ * synchronously or from an nfsiod.
+ */
+int
+nfs_doio(bp, cr, p)
+	register struct buf *bp;
+	struct ucred *cr;
+	struct proc *p;
+{
+	register struct uio *uiop;
+	register struct vnode *vp;
+	struct nfsnode *np;
+	struct nfsmount *nmp;
+	int error = 0, diff, len, iomode, must_commit = 0;
+	struct uio uio;
+	struct iovec io;
+	nfsquad_t tquad;
+
+	vp = bp->b_vp;
+	np = VTONFS(vp);
+	nmp = VFSTONFS(vp->v_mount);
+	uiop = &uio;
+	uiop->uio_iov = &io;
+	uiop->uio_iovcnt = 1;
+	uiop->uio_segflg = UIO_SYSSPACE;
+	uiop->uio_procp = p;
+
+	/*
+	 * Historically, paging was done with physio, but no more.
+	 */
+	if (bp->b_flags & B_PHYS) {
+	    /*
+	     * ...though reading /dev/drum still gets us here.
+	     */
+	    io.iov_len = uiop->uio_resid = bp->b_bcount;
+	    /* mapping was done by vmapbuf() */
+	    io.iov_base = bp->b_data;
+	    uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
+	    if (bp->b_flags & B_READ) {
+		uiop->uio_rw = UIO_READ;
+		nfsstats.read_physios++;
+		error = nfs_readrpc(vp, uiop, cr);
+	    } else
+		panic("physio write");
+	    if (error) {
+		bp->b_flags |= B_ERROR;
+		bp->b_error = error;
+	    }
+	} else if (bp->b_flags & B_READ) {
+	    io.iov_len = uiop->uio_resid = bp->b_bcount;
+	    io.iov_base = bp->b_data;
+	    uiop->uio_rw = UIO_READ;
+	    switch (vp->v_type) {
+	    case VREG:
+		uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
+		nfsstats.read_bios++;
+		error = nfs_readrpc(vp, uiop, cr);
+		if (!error) {
+		    bp->b_validoff = 0;
+		    if (uiop->uio_resid) {
+			/*
+			 * If len > 0, there is a hole in the file and
+			 * no writes after the hole have been pushed to
+			 * the server yet.
+			 * Just zero fill the rest of the valid area.
+			 */
+			diff = bp->b_bcount - uiop->uio_resid;
+			len = np->n_size - (((u_quad_t)bp->b_blkno) * DEV_BSIZE
+				+ diff);
+			if (len > 0) {
+			    len = min(len, uiop->uio_resid);
+			    bzero((char *)bp->b_data + diff, len);
+			    bp->b_validend = diff + len;
+			} else
+			    bp->b_validend = diff;
+		    } else
+			bp->b_validend = bp->b_bcount;
+		}
+		if (p && (vp->v_flag & VTEXT) &&
+			(((nmp->nm_flag & NFSMNT_NQNFS) &&
+			  NQNFS_CKINVALID(vp, np, ND_READ) &&
+			  np->n_lrev != np->n_brev) ||
+			 (!(nmp->nm_flag & NFSMNT_NQNFS) &&
+			  np->n_mtime != np->n_vattr.va_mtime.ts_sec))) {
+			uprintf("Process killed due to text file modification\n");
+			psignal(p, SIGKILL);
+			p->p_flag |= P_NOSWAP;
+		}
+		break;
+	    case VLNK:
+		uiop->uio_offset = (off_t)0;
+		nfsstats.readlink_bios++;
+		error = nfs_readlinkrpc(vp, uiop, cr);
+		break;
+	    case VDIR:
+		nfsstats.readdir_bios++;
+		uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ;
+		if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
+			error = nfs_readdirplusrpc(vp, uiop, cr);
+			if (error == NFSERR_NOTSUPP)
+				nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
+		}
+		if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
+			error = nfs_readdirrpc(vp, uiop, cr);
+		break;
+	    default:
+		printf("nfs_doio:  type %x unexpected\n",vp->v_type);
+		break;
+	    };
+	    if (error) {
+		bp->b_flags |= B_ERROR;
+		bp->b_error = error;
+	    }
+	} else {
+	    io.iov_len = uiop->uio_resid = bp->b_dirtyend
+		- bp->b_dirtyoff;
+	    uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE
+		+ bp->b_dirtyoff;
+	    io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
+	    uiop->uio_rw = UIO_WRITE;
+	    nfsstats.write_bios++;
+	    if ((bp->b_flags & (B_ASYNC | B_NEEDCOMMIT | B_NOCACHE)) == B_ASYNC)
+		iomode = NFSV3WRITE_UNSTABLE;
+	    else
+		iomode = NFSV3WRITE_FILESYNC;
+	    bp->b_flags |= B_WRITEINPROG;
+	    error = nfs_writerpc(vp, uiop, cr, &iomode, &must_commit);
+	    if (!error && iomode == NFSV3WRITE_UNSTABLE)
+		bp->b_flags |= B_NEEDCOMMIT;
+	    else
+		bp->b_flags &= ~B_NEEDCOMMIT;
+	    bp->b_flags &= ~B_WRITEINPROG;
+
+	    /*
+	     * For an interrupted write, the buffer is still valid and the
+	     * write hasn't been pushed to the server yet, so we can't set
+	     * B_ERROR and report the interruption by setting B_EINTR. For
+	     * the B_ASYNC case, B_EINTR is not relevant, so the rpc attempt
+	     * is essentially a noop.
+	     * For the case of a V3 write rpc not being committed to stable
+	     * storage, the block is still dirty and requires either a commit
+	     * rpc or another write rpc with iomode == NFSV3WRITE_FILESYNC
+	     * before the block is reused. This is indicated by setting the
+	     * B_DELWRI and B_NEEDCOMMIT flags.
+	     */
+	    if (error == EINTR || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
+		bp->b_flags |= B_DELWRI;
+
+		/*
+		 * Since for the B_ASYNC case, nfs_bwrite() has reassigned the
+		 * buffer to the clean list, we have to reassign it back to the
+		 * dirty one. Ugh.
+		 */
+		if (bp->b_flags & B_ASYNC)
+		    reassignbuf(bp, vp);
+		else
+		    bp->b_flags |= B_EINTR;
+	    } else {
+		if (error) {
+		    bp->b_flags |= B_ERROR;
+		    bp->b_error = np->n_error = error;
+		    np->n_flag |= NWRITEERR;
+		}
+		bp->b_dirtyoff = bp->b_dirtyend = 0;
+	    }
+	}
+	bp->b_resid = uiop->uio_resid;
+	if (must_commit)
+		nfs_clearcommit(vp->v_mount);
+	biodone(bp);
+	return (error);
+}
diff --git a/sys/nfs/nfs_node.c b/sys/nfs/nfs_node.c
new file mode 100644
index 000000000000..ffd4b08f2b4c
--- /dev/null
+++ b/sys/nfs/nfs_node.c
@@ -0,0 +1,255 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_node.c	8.6 (Berkeley) 5/22/95
+ */
+
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/proc.h>
+#include <sys/mount.h>
+#include <sys/namei.h>
+#include <sys/vnode.h>
+#include <sys/kernel.h>
+#include <sys/malloc.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfsmount.h>
+#include <nfs/nqnfs.h>
+
+LIST_HEAD(nfsnodehashhead, nfsnode) *nfsnodehashtbl;
+u_long nfsnodehash;
+
+#define TRUE	1
+#define	FALSE	0
+
+/*
+ * Initialize hash links for nfsnodes
+ * and build nfsnode free list.
+ */
+void
+nfs_nhinit()
+{
+
+#ifndef lint
+	if ((sizeof(struct nfsnode) - 1) & sizeof(struct nfsnode))
+		printf("nfs_nhinit: bad size %d\n", sizeof(struct nfsnode));
+#endif /* not lint */
+	nfsnodehashtbl = hashinit(desiredvnodes, M_NFSNODE, &nfsnodehash);
+}
+
+/*
+ * Compute an entry in the NFS hash table structure
+ */
+u_long
+nfs_hash(fhp, fhsize)
+	register nfsfh_t *fhp;
+	int fhsize;
+{
+	register u_char *fhpp;
+	register u_long fhsum;
+	register int i;
+
+	fhpp = &fhp->fh_bytes[0];
+	fhsum = 0;
+	for (i = 0; i < fhsize; i++)
+		fhsum += *fhpp++;
+	return (fhsum);
+}
+
+/*
+ * Look up a vnode/nfsnode by file handle.
+ * Callers must check for mount points!!
+ * In all cases, a pointer to a
+ * nfsnode structure is returned.
+ */
+int
+nfs_nget(mntp, fhp, fhsize, npp)
+	struct mount *mntp;
+	register nfsfh_t *fhp;
+	int fhsize;
+	struct nfsnode **npp;
+{
+	struct proc *p = curproc;	/* XXX */
+	struct nfsnode *np;
+	struct nfsnodehashhead *nhpp;
+	register struct vnode *vp;
+	extern int (**nfsv2_vnodeop_p)();
+	struct vnode *nvp;
+	int error;
+
+	nhpp = NFSNOHASH(nfs_hash(fhp, fhsize));
+loop:
+	for (np = nhpp->lh_first; np != 0; np = np->n_hash.le_next) {
+		if (mntp != NFSTOV(np)->v_mount || np->n_fhsize != fhsize ||
+		    bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize))
+			continue;
+		vp = NFSTOV(np);
+		if (vget(vp, LK_EXCLUSIVE, p))
+			goto loop;
+		*npp = np;
+		return(0);
+	}
+	error = getnewvnode(VT_NFS, mntp, nfsv2_vnodeop_p, &nvp);
+	if (error) {
+		*npp = 0;
+		return (error);
+	}
+	vp = nvp;
+	MALLOC(np, struct nfsnode *, sizeof *np, M_NFSNODE, M_WAITOK);
+	bzero((caddr_t)np, sizeof *np);
+	vp->v_data = np;
+	np->n_vnode = vp;
+	/*
+	 * Insert the nfsnode in the hash queue for its new file handle
+	 */
+	LIST_INSERT_HEAD(nhpp, np, n_hash);
+	if (fhsize > NFS_SMALLFH) {
+		MALLOC(np->n_fhp, nfsfh_t *, fhsize, M_NFSBIGFH, M_WAITOK);
+	} else
+		np->n_fhp = &np->n_fh;
+	bcopy((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize);
+	np->n_fhsize = fhsize;
+	*npp = np;
+	return (0);
+}
+
+int
+nfs_inactive(ap)
+	struct vop_inactive_args /* {
+		struct vnode *a_vp;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct nfsnode *np;
+	register struct sillyrename *sp;
+	struct proc *p = curproc;	/* XXX */
+	extern int prtactive;
+
+	np = VTONFS(ap->a_vp);
+	if (prtactive && ap->a_vp->v_usecount != 0)
+		vprint("nfs_inactive: pushing active", ap->a_vp);
+	if (ap->a_vp->v_type != VDIR)
+		sp = np->n_sillyrename;
+	else
+		sp = (struct sillyrename *)0;
+	np->n_sillyrename = (struct sillyrename *)0;
+	if (sp) {
+		/*
+		 * Remove the silly file that was rename'd earlier
+		 */
+		(void) nfs_vinvalbuf(ap->a_vp, 0, sp->s_cred, p, 1);
+		nfs_removeit(sp);
+		crfree(sp->s_cred);
+		vrele(sp->s_dvp);
+		FREE((caddr_t)sp, M_NFSREQ);
+	}
+	np->n_flag &= (NMODIFIED | NFLUSHINPROG | NFLUSHWANT | NQNFSEVICTED |
+		NQNFSNONCACHE | NQNFSWRITE);
+	VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
+	return (0);
+}
+
+/*
+ * Reclaim an nfsnode so that it can be used for other purposes.
+ */
+int
+nfs_reclaim(ap)
+	struct vop_reclaim_args /* {
+		struct vnode *a_vp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	register struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	register struct nfsdmap *dp, *dp2;
+	extern int prtactive;
+
+	if (prtactive && vp->v_usecount != 0)
+		vprint("nfs_reclaim: pushing active", vp);
+
+	LIST_REMOVE(np, n_hash);
+
+	/*
+	 * For nqnfs, take it off the timer queue as required.
+	 */
+	if ((nmp->nm_flag & NFSMNT_NQNFS) && np->n_timer.cqe_next != 0) {
+		CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer);
+	}
+
+	/*
+	 * Free up any directory cookie structures and
+	 * large file handle structures that might be associated with
+	 * this nfs node.
+	 */
+	if (vp->v_type == VDIR) {
+		dp = np->n_cookies.lh_first;
+		while (dp) {
+			dp2 = dp;
+			dp = dp->ndm_list.le_next;
+			FREE((caddr_t)dp2, M_NFSDIROFF);
+		}
+	}
+	if (np->n_fhsize > NFS_SMALLFH) {
+		FREE((caddr_t)np->n_fhp, M_NFSBIGFH);
+	}
+
+	cache_purge(vp);
+	FREE(vp->v_data, M_NFSNODE);
+	vp->v_data = (void *)0;
+	return (0);
+}
+
+/*
+ * Nfs abort op, called after namei() when a CREATE/DELETE isn't actually
+ * done. Currently nothing to do.
+ */
+/* ARGSUSED */
+int
+nfs_abortop(ap)
+	struct vop_abortop_args /* {
+		struct vnode *a_dvp;
+		struct componentname *a_cnp;
+	} */ *ap;
+{
+
+	if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF)
+		FREE(ap->a_cnp->cn_pnbuf, M_NAMEI);
+	return (0);
+}
diff --git a/sys/nfs/nfs_nqlease.c b/sys/nfs/nfs_nqlease.c
new file mode 100644
index 000000000000..5c348b3d8413
--- /dev/null
+++ b/sys/nfs/nfs_nqlease.c
@@ -0,0 +1,1244 @@
+/*
+ * Copyright (c) 1992, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_nqlease.c	8.9 (Berkeley) 5/20/95
+ */
+
+
+/*
+ * References:
+ *	Cary G. Gray and David R. Cheriton, "Leases: An Efficient Fault-Tolerant
+ *		Mechanism for Distributed File Cache Consistency",
+ *		In Proc. of the Twelfth ACM Symposium on Operating Systems
+ *		Principals, pg. 202-210, Litchfield Park, AZ, Dec. 1989.
+ *	Michael N. Nelson, Brent B. Welch and John K. Ousterhout, "Caching
+ *		in the Sprite Network File System", ACM TOCS 6(1),
+ *		pages 134-154, February 1988.
+ *	V. Srinivasan and Jeffrey C. Mogul, "Spritely NFS: Implementation and
+ *		Performance of Cache-Consistency Protocols", Digital
+ *		Equipment Corporation WRL Research Report 89/5, May 1989.
+ */
+#include <sys/param.h>
+#include <sys/vnode.h>
+#include <sys/mount.h>
+#include <sys/kernel.h>
+#include <sys/proc.h>
+#include <sys/systm.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+#include <sys/file.h>
+#include <sys/buf.h>
+#include <sys/stat.h>
+#include <sys/protosw.h>
+
+#include <netinet/in.h>
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nqnfs.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfsmount.h>
+
+time_t nqnfsstarttime = (time_t)0;
+int nqsrv_clockskew = NQ_CLOCKSKEW;
+int nqsrv_writeslack = NQ_WRITESLACK;
+int nqsrv_maxlease = NQ_MAXLEASE;
+int nqsrv_maxnumlease = NQ_MAXNUMLEASE;
+void nqsrv_instimeq(), nqsrv_send_eviction(), nfs_sndunlock();
+void nqsrv_unlocklease(), nqsrv_waitfor_expiry(), nfsrv_slpderef();
+void nqsrv_addhost(), nqsrv_locklease(), nqnfs_serverd();
+void nqnfs_clientlease();
+struct mbuf *nfsm_rpchead();
+
+/*
+ * Signifies which rpcs can have piggybacked lease requests
+ */
+int nqnfs_piggy[NFS_NPROCS] = {
+	0,
+	0,
+	ND_WRITE,
+	ND_READ,
+	0,
+	ND_READ,
+	ND_READ,
+	ND_WRITE,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+	ND_READ,
+	ND_READ,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+	0,
+};
+
+extern nfstype nfsv2_type[9];
+extern nfstype nfsv3_type[9];
+extern struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock;
+extern int nfsd_waiting;
+extern struct nfsstats nfsstats;
+extern int nfs_mount_type;
+
+#define TRUE	1
+#define	FALSE	0
+
+/*
+ * Get or check for a lease for "vp", based on ND_CHECK flag.
+ * The rules are as follows:
+ * - if a current non-caching lease, reply non-caching
+ * - if a current lease for same host only, extend lease
+ * - if a read cachable lease and a read lease request
+ *	add host to list any reply cachable
+ * - else { set non-cachable for read-write sharing }
+ *	send eviction notice messages to all other hosts that have lease
+ *	wait for lease termination { either by receiving vacated messages
+ *					from all the other hosts or expiry
+ *					via. timeout }
+ *	modify lease to non-cachable
+ * - else if no current lease, issue new one
+ * - reply
+ * - return boolean TRUE iff nam should be m_freem()'d
+ * NB: Since nqnfs_serverd() is called from a timer, any potential tsleep()
+ *     in here must be framed by nqsrv_locklease() and nqsrv_unlocklease().
+ *     nqsrv_locklease() is coded such that at least one of LC_LOCKED and
+ *     LC_WANTED is set whenever a process is tsleeping in it. The exception
+ *     is when a new lease is being allocated, since it is not in the timer
+ *     queue yet. (Ditto for the splsoftclock() and splx(s) calls)
+ */
+int
+nqsrv_getlease(vp, duration, flags, slp, procp, nam, cachablep, frev, cred)
+	struct vnode *vp;
+	u_long *duration;
+	int flags;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf *nam;
+	int *cachablep;
+	u_quad_t *frev;
+	struct ucred *cred;
+{
+	register struct nqlease *lp;
+	register struct nqfhhashhead *lpp = 0;
+	register struct nqhost *lph = 0;
+	struct nqlease *tlp;
+	struct nqm **lphp;
+	struct vattr vattr;
+	fhandle_t fh;
+	int i, ok, error, s;
+
+	if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK)
+		return (0);
+	if (*duration > nqsrv_maxlease)
+		*duration = nqsrv_maxlease;
+	error = VOP_GETATTR(vp, &vattr, cred, procp);
+	if (error)
+		return (error);
+	*frev = vattr.va_filerev;
+	s = splsoftclock();
+	tlp = vp->v_lease;
+	if ((flags & ND_CHECK) == 0)
+		nfsstats.srvnqnfs_getleases++;
+	if (tlp == (struct nqlease *)0) {
+
+		/*
+		 * Find the lease by searching the hash list.
+		 */
+		fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+		error = VFS_VPTOFH(vp, &fh.fh_fid);
+		if (error) {
+			splx(s);
+			return (error);
+		}
+		lpp = NQFHHASH(fh.fh_fid.fid_data);
+		for (lp = lpp->lh_first; lp != 0; lp = lp->lc_hash.le_next)
+			if (fh.fh_fsid.val[0] == lp->lc_fsid.val[0] &&
+			    fh.fh_fsid.val[1] == lp->lc_fsid.val[1] &&
+			    !bcmp(fh.fh_fid.fid_data, lp->lc_fiddata,
+				  fh.fh_fid.fid_len - sizeof (long))) {
+				/* Found it */
+				lp->lc_vp = vp;
+				vp->v_lease = lp;
+				tlp = lp;
+				break;
+			}
+	} else
+		lp = tlp;
+	if (lp) {
+		if ((lp->lc_flag & LC_NONCACHABLE) ||
+		    (lp->lc_morehosts == (struct nqm *)0 &&
+		     nqsrv_cmpnam(slp, nam, &lp->lc_host)))
+			goto doreply;
+		if ((flags & ND_READ) && (lp->lc_flag & LC_WRITE) == 0) {
+			if (flags & ND_CHECK)
+				goto doreply;
+			if (nqsrv_cmpnam(slp, nam, &lp->lc_host))
+				goto doreply;
+			i = 0;
+			if (lp->lc_morehosts) {
+				lph = lp->lc_morehosts->lpm_hosts;
+				lphp = &lp->lc_morehosts->lpm_next;
+				ok = 1;
+			} else {
+				lphp = &lp->lc_morehosts;
+				ok = 0;
+			}
+			while (ok && (lph->lph_flag & LC_VALID)) {
+				if (nqsrv_cmpnam(slp, nam, lph))
+					goto doreply;
+				if (++i == LC_MOREHOSTSIZ) {
+					i = 0;
+					if (*lphp) {
+						lph = (*lphp)->lpm_hosts;
+						lphp = &((*lphp)->lpm_next);
+					} else
+						ok = 0;
+				} else
+					lph++;
+			}
+			nqsrv_locklease(lp);
+			if (!ok) {
+				*lphp = (struct nqm *)
+					malloc(sizeof (struct nqm),
+						M_NQMHOST, M_WAITOK);
+				bzero((caddr_t)*lphp, sizeof (struct nqm));
+				lph = (*lphp)->lpm_hosts;
+			}
+			nqsrv_addhost(lph, slp, nam);
+			nqsrv_unlocklease(lp);
+		} else {
+			lp->lc_flag |= LC_NONCACHABLE;
+			nqsrv_locklease(lp);
+			nqsrv_send_eviction(vp, lp, slp, nam, cred);
+			nqsrv_waitfor_expiry(lp);
+			nqsrv_unlocklease(lp);
+		}
+doreply:
+		/*
+		 * Update the lease and return
+		 */
+		if ((flags & ND_CHECK) == 0)
+			nqsrv_instimeq(lp, *duration);
+		if (lp->lc_flag & LC_NONCACHABLE)
+			*cachablep = 0;
+		else {
+			*cachablep = 1;
+			if (flags & ND_WRITE)
+				lp->lc_flag |= LC_WRITTEN;
+		}
+		splx(s);
+		return (0);
+	}
+	splx(s);
+	if (flags & ND_CHECK)
+		return (0);
+
+	/*
+	 * Allocate new lease
+	 * The value of nqsrv_maxnumlease should be set generously, so that
+	 * the following "printf" happens infrequently.
+	 */
+	if (nfsstats.srvnqnfs_leases > nqsrv_maxnumlease) {
+		printf("Nqnfs server, too many leases\n");
+		do {
+			(void) tsleep((caddr_t)&lbolt, PSOCK,
+					"nqsrvnuml", 0);
+		} while (nfsstats.srvnqnfs_leases > nqsrv_maxnumlease);
+	}
+	MALLOC(lp, struct nqlease *, sizeof (struct nqlease), M_NQLEASE, M_WAITOK);
+	bzero((caddr_t)lp, sizeof (struct nqlease));
+	if (flags & ND_WRITE)
+		lp->lc_flag |= (LC_WRITE | LC_WRITTEN);
+	nqsrv_addhost(&lp->lc_host, slp, nam);
+	lp->lc_vp = vp;
+	lp->lc_fsid = fh.fh_fsid;
+	bcopy(fh.fh_fid.fid_data, lp->lc_fiddata,
+		fh.fh_fid.fid_len - sizeof (long));
+	if(!lpp)
+		panic("nfs_nqlease.c: Phoney lpp");
+	LIST_INSERT_HEAD(lpp, lp, lc_hash);
+	vp->v_lease = lp;
+	s = splsoftclock();
+	nqsrv_instimeq(lp, *duration);
+	splx(s);
+	*cachablep = 1;
+	if (++nfsstats.srvnqnfs_leases > nfsstats.srvnqnfs_maxleases)
+		nfsstats.srvnqnfs_maxleases = nfsstats.srvnqnfs_leases;
+	return (0);
+}
+
+/*
+ * Local lease check for server syscalls.
+ * Just set up args and let nqsrv_getlease() do the rest.
+ */
+int
+nqnfs_vop_lease_check(ap)
+	struct vop_lease_args /* {
+		struct vnode *a_vp;
+		struct proc *a_p;
+		struct ucred *a_cred;
+		int a_flag;
+	} */ *ap;
+{
+	u_long duration = 0;
+	int cache;
+	u_quad_t frev;
+
+	(void) nqsrv_getlease(ap->a_vp, &duration, ND_CHECK | ap->a_flag,
+	    NQLOCALSLP, ap->a_p, (struct mbuf *)0, &cache, &frev, ap->a_cred);
+	return (0);
+}
+
+/*
+ * Add a host to an nqhost structure for a lease.
+ */
+void
+nqsrv_addhost(lph, slp, nam)
+	register struct nqhost *lph;
+	struct nfssvc_sock *slp;
+	struct mbuf *nam;
+{
+	register struct sockaddr_in *saddr;
+
+	if (slp == NQLOCALSLP)
+		lph->lph_flag |= (LC_VALID | LC_LOCAL);
+	else if (slp == nfs_udpsock) {
+		saddr = mtod(nam, struct sockaddr_in *);
+		lph->lph_flag |= (LC_VALID | LC_UDP);
+		lph->lph_inetaddr = saddr->sin_addr.s_addr;
+		lph->lph_port = saddr->sin_port;
+	} else if (slp == nfs_cltpsock) {
+		lph->lph_nam = m_copym(nam, 0, M_COPYALL, M_WAIT);
+		lph->lph_flag |= (LC_VALID | LC_CLTP);
+	} else {
+		lph->lph_flag |= (LC_VALID | LC_SREF);
+		lph->lph_slp = slp;
+		slp->ns_sref++;
+	}
+}
+
+/*
+ * Update the lease expiry time and position it in the timer queue correctly.
+ */
+void
+nqsrv_instimeq(lp, duration)
+	register struct nqlease *lp;
+	u_long duration;
+{
+	register struct nqlease *tlp;
+	time_t newexpiry;
+
+	newexpiry = time.tv_sec + duration + nqsrv_clockskew;
+	if (lp->lc_expiry == newexpiry)
+		return;
+	if (lp->lc_timer.cqe_next != 0) {
+		CIRCLEQ_REMOVE(&nqtimerhead, lp, lc_timer);
+	}
+	lp->lc_expiry = newexpiry;
+
+	/*
+	 * Find where in the queue it should be.
+	 */
+	tlp = nqtimerhead.cqh_last;
+	while (tlp != (void *)&nqtimerhead && tlp->lc_expiry > newexpiry)
+		tlp = tlp->lc_timer.cqe_prev;
+#ifdef HASNVRAM
+	if (tlp == nqtimerhead.cqh_last)
+		NQSTORENOVRAM(newexpiry);
+#endif /* HASNVRAM */
+	if (tlp == (void *)&nqtimerhead) {
+		CIRCLEQ_INSERT_HEAD(&nqtimerhead, lp, lc_timer);
+	} else {
+		CIRCLEQ_INSERT_AFTER(&nqtimerhead, tlp, lp, lc_timer);
+	}
+}
+
+/*
+ * Compare the requesting host address with the lph entry in the lease.
+ * Return true iff it is the same.
+ * This is somewhat messy due to the union in the nqhost structure.
+ * The local host is indicated by the special value of NQLOCALSLP for slp.
+ */
+int
+nqsrv_cmpnam(slp, nam, lph)
+	register struct nfssvc_sock *slp;
+	struct mbuf *nam;
+	register struct nqhost *lph;
+{
+	register struct sockaddr_in *saddr;
+	struct mbuf *addr;
+	union nethostaddr lhaddr;
+	int ret;
+
+	if (slp == NQLOCALSLP) {
+		if (lph->lph_flag & LC_LOCAL)
+			return (1);
+		else
+			return (0);
+	}
+	if (slp == nfs_udpsock || slp == nfs_cltpsock)
+		addr = nam;
+	else
+		addr = slp->ns_nam;
+	if (lph->lph_flag & LC_UDP)
+		ret = netaddr_match(AF_INET, &lph->lph_haddr, addr);
+	else if (lph->lph_flag & LC_CLTP)
+		ret = netaddr_match(AF_ISO, &lph->lph_claddr, addr);
+	else {
+		if ((lph->lph_slp->ns_flag & SLP_VALID) == 0)
+			return (0);
+		saddr = mtod(lph->lph_slp->ns_nam, struct sockaddr_in *);
+		if (saddr->sin_family == AF_INET)
+			lhaddr.had_inetaddr = saddr->sin_addr.s_addr;
+		else
+			lhaddr.had_nam = lph->lph_slp->ns_nam;
+		ret = netaddr_match(saddr->sin_family, &lhaddr, addr);
+	}
+	return (ret);
+}
+
+/*
+ * Send out eviction notice messages to all other hosts for the lease.
+ */
+void
+nqsrv_send_eviction(vp, lp, slp, nam, cred)
+	struct vnode *vp;
+	register struct nqlease *lp;
+	struct nfssvc_sock *slp;
+	struct mbuf *nam;
+	struct ucred *cred;
+{
+	register struct nqhost *lph = &lp->lc_host;
+	register struct mbuf *m;
+	register int siz;
+	struct nqm *lphnext = lp->lc_morehosts;
+	struct mbuf *mreq, *mb, *mb2, *nam2, *mheadend;
+	struct socket *so;
+	struct sockaddr_in *saddr;
+	fhandle_t *fhp;
+	caddr_t bpos, cp;
+	u_long xid;
+	int len = 1, ok = 1, i = 0;
+	int sotype, *solockp;
+
+	while (ok && (lph->lph_flag & LC_VALID)) {
+		if (nqsrv_cmpnam(slp, nam, lph))
+			lph->lph_flag |= LC_VACATED;
+		else if ((lph->lph_flag & (LC_LOCAL | LC_VACATED)) == 0) {
+			if (lph->lph_flag & LC_UDP) {
+				MGET(nam2, M_WAIT, MT_SONAME);
+				saddr = mtod(nam2, struct sockaddr_in *);
+				nam2->m_len = saddr->sin_len =
+					sizeof (struct sockaddr_in);
+				saddr->sin_family = AF_INET;
+				saddr->sin_addr.s_addr = lph->lph_inetaddr;
+				saddr->sin_port = lph->lph_port;
+				so = nfs_udpsock->ns_so;
+			} else if (lph->lph_flag & LC_CLTP) {
+				nam2 = lph->lph_nam;
+				so = nfs_cltpsock->ns_so;
+			} else if (lph->lph_slp->ns_flag & SLP_VALID) {
+				nam2 = (struct mbuf *)0;
+				so = lph->lph_slp->ns_so;
+			} else
+				goto nextone;
+			sotype = so->so_type;
+			if (so->so_proto->pr_flags & PR_CONNREQUIRED)
+				solockp = &lph->lph_slp->ns_solock;
+			else
+				solockp = (int *)0;
+			nfsm_reqhead((struct vnode *)0, NQNFSPROC_EVICTED,
+				NFSX_V3FH);
+			nfsm_build(cp, caddr_t, NFSX_V3FH);
+			bzero(cp, NFSX_V3FH);
+			fhp = (fhandle_t *)cp;
+			fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+			VFS_VPTOFH(vp, &fhp->fh_fid);
+			m = mreq;
+			siz = 0;
+			while (m) {
+				siz += m->m_len;
+				m = m->m_next;
+			}
+			if (siz <= 0 || siz > NFS_MAXPACKET) {
+				printf("mbuf siz=%d\n",siz);
+				panic("Bad nfs svc reply");
+			}
+			m = nfsm_rpchead(cred, (NFSMNT_NFSV3 | NFSMNT_NQNFS),
+				NQNFSPROC_EVICTED,
+				RPCAUTH_UNIX, 5 * NFSX_UNSIGNED, (char *)0,
+				mreq, siz, &mheadend, &xid);
+			/*
+			 * For stream protocols, prepend a Sun RPC
+			 * Record Mark.
+			 */
+			if (sotype == SOCK_STREAM) {
+				M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
+				*mtod(m, u_long *) = htonl(0x80000000 |
+					(m->m_pkthdr.len - NFSX_UNSIGNED));
+			}
+			if (((lph->lph_flag & (LC_UDP | LC_CLTP)) == 0 &&
+			    (lph->lph_slp->ns_flag & SLP_VALID) == 0) ||
+			    (solockp && (*solockp & NFSMNT_SNDLOCK)))
+				m_freem(m);
+			else {
+				if (solockp)
+					*solockp |= NFSMNT_SNDLOCK;
+				(void) nfs_send(so, nam2, m,
+						(struct nfsreq *)0);
+				if (solockp)
+					nfs_sndunlock(solockp);
+			}
+			if (lph->lph_flag & LC_UDP)
+				MFREE(nam2, m);
+		}
+nextone:
+		if (++i == len) {
+			if (lphnext) {
+				i = 0;
+				len = LC_MOREHOSTSIZ;
+				lph = lphnext->lpm_hosts;
+				lphnext = lphnext->lpm_next;
+			} else
+				ok = 0;
+		} else
+			lph++;
+	}
+}
+
+/*
+ * Wait for the lease to expire.
+ * This will occur when all clients have sent "vacated" messages to
+ * this server OR when it expires do to timeout.
+ */
+void
+nqsrv_waitfor_expiry(lp)
+	register struct nqlease *lp;
+{
+	register struct nqhost *lph;
+	register int i;
+	struct nqm *lphnext;
+	int len, ok;
+
+tryagain:
+	if (time.tv_sec > lp->lc_expiry)
+		return;
+	lph = &lp->lc_host;
+	lphnext = lp->lc_morehosts;
+	len = 1;
+	i = 0;
+	ok = 1;
+	while (ok && (lph->lph_flag & LC_VALID)) {
+		if ((lph->lph_flag & (LC_LOCAL | LC_VACATED)) == 0) {
+			lp->lc_flag |= LC_EXPIREDWANTED;
+			(void) tsleep((caddr_t)&lp->lc_flag, PSOCK,
+					"nqexp", 0);
+			goto tryagain;
+		}
+		if (++i == len) {
+			if (lphnext) {
+				i = 0;
+				len = LC_MOREHOSTSIZ;
+				lph = lphnext->lpm_hosts;
+				lphnext = lphnext->lpm_next;
+			} else
+				ok = 0;
+		} else
+			lph++;
+	}
+}
+
+/*
+ * Nqnfs server timer that maintains the server lease queue.
+ * Scan the lease queue for expired entries:
+ * - when one is found, wakeup anyone waiting for it
+ *   else dequeue and free
+ */
+void
+nqnfs_serverd()
+{
+	register struct nqlease *lp;
+	register struct nqhost *lph;
+	struct nqlease *nextlp;
+	struct nqm *lphnext, *olphnext;
+	struct mbuf *n;
+	int i, len, ok;
+
+	for (lp = nqtimerhead.cqh_first; lp != (void *)&nqtimerhead;
+	    lp = nextlp) {
+		if (lp->lc_expiry >= time.tv_sec)
+			break;
+		nextlp = lp->lc_timer.cqe_next;
+		if (lp->lc_flag & LC_EXPIREDWANTED) {
+			lp->lc_flag &= ~LC_EXPIREDWANTED;
+			wakeup((caddr_t)&lp->lc_flag);
+		} else if ((lp->lc_flag & (LC_LOCKED | LC_WANTED)) == 0) {
+		    /*
+		     * Make a best effort at keeping a write caching lease long
+		     * enough by not deleting it until it has been explicitly
+		     * vacated or there have been no writes in the previous
+		     * write_slack seconds since expiry and the nfsds are not
+		     * all busy. The assumption is that if the nfsds are not
+		     * all busy now (no queue of nfs requests), then the client
+		     * would have been able to do at least one write to the
+		     * file during the last write_slack seconds if it was still
+		     * trying to push writes to the server.
+		     */
+		    if ((lp->lc_flag & (LC_WRITE | LC_VACATED)) == LC_WRITE &&
+			((lp->lc_flag & LC_WRITTEN) || nfsd_waiting == 0)) {
+			lp->lc_flag &= ~LC_WRITTEN;
+			nqsrv_instimeq(lp, nqsrv_writeslack);
+		    } else {
+			CIRCLEQ_REMOVE(&nqtimerhead, lp, lc_timer);
+			LIST_REMOVE(lp, lc_hash);
+			/*
+			 * This soft reference may no longer be valid, but
+			 * no harm done. The worst case is if the vnode was
+			 * recycled and has another valid lease reference,
+			 * which is dereferenced prematurely.
+			 */
+			lp->lc_vp->v_lease = (struct nqlease *)0;
+			lph = &lp->lc_host;
+			lphnext = lp->lc_morehosts;
+			olphnext = (struct nqm *)0;
+			len = 1;
+			i = 0;
+			ok = 1;
+			while (ok && (lph->lph_flag & LC_VALID)) {
+				if (lph->lph_flag & LC_CLTP)
+					MFREE(lph->lph_nam, n);
+				if (lph->lph_flag & LC_SREF)
+					nfsrv_slpderef(lph->lph_slp);
+				if (++i == len) {
+					if (olphnext) {
+						free((caddr_t)olphnext, M_NQMHOST);
+						olphnext = (struct nqm *)0;
+					}
+					if (lphnext) {
+						olphnext = lphnext;
+						i = 0;
+						len = LC_MOREHOSTSIZ;
+						lph = lphnext->lpm_hosts;
+						lphnext = lphnext->lpm_next;
+					} else
+						ok = 0;
+				} else
+					lph++;
+			}
+			FREE((caddr_t)lp, M_NQLEASE);
+			if (olphnext)
+				free((caddr_t)olphnext, M_NQMHOST);
+			nfsstats.srvnqnfs_leases--;
+		    }
+		}
+	}
+}
+
+/*
+ * Called from nfssvc_nfsd() for a getlease rpc request.
+ * Do the from/to xdr translation and call nqsrv_getlease() to
+ * do the real work.
+ */
+int
+nqnfsrv_getlease(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nfs_fattr *fp;
+	struct vattr va;
+	register struct vattr *vap = &va;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	u_quad_t frev;
+	caddr_t bpos;
+	int error = 0;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	int flags, rdonly, cache;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+	flags = fxdr_unsigned(int, *tl++);
+	nfsd->nd_duration = fxdr_unsigned(int, *tl);
+	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
+		(nfsd->nd_flag & ND_KERBAUTH));
+	if (error)
+		nfsm_reply(0);
+	if (rdonly && flags == ND_WRITE) {
+		vput(vp);
+		error = EROFS;
+		nfsm_reply(0);
+	}
+	(void) nqsrv_getlease(vp, &nfsd->nd_duration, flags, slp, procp,
+		nam, &cache, &frev, cred);
+	error = VOP_GETATTR(vp, vap, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
+	nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+	*tl++ = txdr_unsigned(cache);
+	*tl++ = txdr_unsigned(nfsd->nd_duration);
+	txdr_hyper(&frev, tl);
+	nfsm_build(fp, struct nfs_fattr *, NFSX_V3FATTR);
+	nfsm_srvfillattr(vap, fp);
+	nfsm_srvdone;
+}
+
+/*
+ * Called from nfssvc_nfsd() when a "vacated" message is received from a
+ * client. Find the entry and expire it.
+ */
+int
+nqnfsrv_vacated(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nqlease *lp;
+	register struct nqhost *lph;
+	struct nqlease *tlp = (struct nqlease *)0;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	struct nqm *lphnext;
+	struct mbuf *mreq, *mb;
+	int error = 0, i, len, ok, gotit = 0, cache = 0;
+	char *cp2, *bpos;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	m_freem(mrep);
+	/*
+	 * Find the lease by searching the hash list.
+	 */
+	for (lp = NQFHHASH(fhp->fh_fid.fid_data)->lh_first; lp != 0;
+	    lp = lp->lc_hash.le_next)
+		if (fhp->fh_fsid.val[0] == lp->lc_fsid.val[0] &&
+		    fhp->fh_fsid.val[1] == lp->lc_fsid.val[1] &&
+		    !bcmp(fhp->fh_fid.fid_data, lp->lc_fiddata,
+			  MAXFIDSZ)) {
+			/* Found it */
+			tlp = lp;
+			break;
+		}
+	if (tlp) {
+		lp = tlp;
+		len = 1;
+		i = 0;
+		lph = &lp->lc_host;
+		lphnext = lp->lc_morehosts;
+		ok = 1;
+		while (ok && (lph->lph_flag & LC_VALID)) {
+			if (nqsrv_cmpnam(slp, nam, lph)) {
+				lph->lph_flag |= LC_VACATED;
+				gotit++;
+				break;
+			}
+			if (++i == len) {
+				if (lphnext) {
+					len = LC_MOREHOSTSIZ;
+					i = 0;
+					lph = lphnext->lpm_hosts;
+					lphnext = lphnext->lpm_next;
+				} else
+					ok = 0;
+			} else
+				lph++;
+		}
+		if ((lp->lc_flag & LC_EXPIREDWANTED) && gotit) {
+			lp->lc_flag &= ~LC_EXPIREDWANTED;
+			wakeup((caddr_t)&lp->lc_flag);
+		}
+nfsmout:
+		return (EPERM);
+	}
+	return (EPERM);
+}
+
+/*
+ * Client get lease rpc function.
+ */
+int
+nqnfs_getlease(vp, rwflag, cred, p)
+	register struct vnode *vp;
+	int rwflag;
+	struct ucred *cred;
+	struct proc *p;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	register struct nfsnode *np;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	caddr_t bpos, dpos, cp2;
+	time_t reqtime;
+	int error = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int cachable;
+	u_quad_t frev;
+	
+	nfsstats.rpccnt[NQNFSPROC_GETLEASE]++;
+	mb = mreq = nfsm_reqh(vp, NQNFSPROC_GETLEASE, NFSX_V3FH+2*NFSX_UNSIGNED,
+		 &bpos);
+	nfsm_fhtom(vp, 1);
+	nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+	*tl++ = txdr_unsigned(rwflag);
+	*tl = txdr_unsigned(nmp->nm_leaseterm);
+	reqtime = time.tv_sec;
+	nfsm_request(vp, NQNFSPROC_GETLEASE, p, cred);
+	np = VTONFS(vp);
+	nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED);
+	cachable = fxdr_unsigned(int, *tl++);
+	reqtime += fxdr_unsigned(int, *tl++);
+	if (reqtime > time.tv_sec) {
+		fxdr_hyper(tl, &frev);
+		nqnfs_clientlease(nmp, np, rwflag, cachable, reqtime, frev);
+		nfsm_loadattr(vp, (struct vattr *)0);
+	} else
+		error = NQNFS_EXPIRED;
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * Client vacated message function.
+ */
+int
+nqnfs_vacated(vp, cred)
+	register struct vnode *vp;
+	struct ucred *cred;
+{
+	register caddr_t cp;
+	register struct mbuf *m;
+	register int i;
+	register u_long *tl;
+	register long t1, t2;
+	caddr_t bpos;
+	u_long xid;
+	int error = 0;
+	struct mbuf *mreq, *mb, *mb2, *mheadend;
+	struct nfsmount *nmp;
+	struct nfsreq myrep;
+	
+	nmp = VFSTONFS(vp->v_mount);
+	nfsstats.rpccnt[NQNFSPROC_VACATED]++;
+	nfsm_reqhead(vp, NQNFSPROC_VACATED, NFSX_V3FH);
+	nfsm_fhtom(vp, 1);
+	m = mreq;
+	i = 0;
+	while (m) {
+		i += m->m_len;
+		m = m->m_next;
+	}
+	m = nfsm_rpchead(cred, nmp->nm_flag, NQNFSPROC_VACATED,
+		RPCAUTH_UNIX, 5 * NFSX_UNSIGNED, (char *)0,
+		mreq, i, &mheadend, &xid);
+	if (nmp->nm_sotype == SOCK_STREAM) {
+		M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
+		*mtod(m, u_long *) = htonl(0x80000000 | (m->m_pkthdr.len -
+			NFSX_UNSIGNED));
+	}
+	myrep.r_flags = 0;
+	myrep.r_nmp = nmp;
+	if (nmp->nm_soflags & PR_CONNREQUIRED)
+		(void) nfs_sndlock(&nmp->nm_flag, (struct nfsreq *)0);
+	(void) nfs_send(nmp->nm_so, nmp->nm_nam, m, &myrep);
+	if (nmp->nm_soflags & PR_CONNREQUIRED)
+		nfs_sndunlock(&nmp->nm_flag);
+nfsmout:
+	return (error);
+}
+
+/*
+ * Called for client side callbacks
+ */
+int
+nqnfs_callback(nmp, mrep, md, dpos)
+	struct nfsmount *nmp;
+	struct mbuf *mrep, *md;
+	caddr_t dpos;
+{
+	register struct vnode *vp;
+	register u_long *tl;
+	register long t1;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct nfsnode *np;
+	struct nfsd tnfsd;
+	struct nfssvc_sock *slp;
+	struct nfsrv_descript ndesc;
+	register struct nfsrv_descript *nfsd = &ndesc;
+	struct mbuf **mrq = (struct mbuf **)0, *mb, *mreq;
+	int error = 0, cache = 0;
+	char *cp2, *bpos;
+	u_quad_t frev;
+
+#ifndef nolint
+	slp = NULL;
+#endif
+	nfsd->nd_mrep = mrep;
+	nfsd->nd_md = md;
+	nfsd->nd_dpos = dpos;
+	error = nfs_getreq(nfsd, &tnfsd, FALSE);
+	if (error)
+		return (error);
+	md = nfsd->nd_md;
+	dpos = nfsd->nd_dpos;
+	if (nfsd->nd_procnum != NQNFSPROC_EVICTED) {
+		m_freem(mrep);
+		return (EPERM);
+	}
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	m_freem(mrep);
+	error = nfs_nget(nmp->nm_mountp, (nfsfh_t *)fhp, NFSX_V3FH, &np);
+	if (error)
+		return (error);
+	vp = NFSTOV(np);
+	if (np->n_timer.cqe_next != 0) {
+		np->n_expiry = 0;
+		np->n_flag |= NQNFSEVICTED;
+		if (nmp->nm_timerhead.cqh_first != np) {
+			CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer);
+			CIRCLEQ_INSERT_HEAD(&nmp->nm_timerhead, np, n_timer);
+		}
+	}
+	vrele(vp);
+	nfsm_srvdone;
+}
+
+/*
+ * Nqnfs client helper daemon. Runs once a second to expire leases.
+ * It also get authorization strings for "kerb" mounts.
+ * It must start at the beginning of the list again after any potential
+ * "sleep" since nfs_reclaim() called from vclean() can pull a node off
+ * the list asynchronously.
+ */
+int
+nqnfs_clientd(nmp, cred, ncd, flag, argp, p)
+	register struct nfsmount *nmp;
+	struct ucred *cred;
+	struct nfsd_cargs *ncd;
+	int flag;
+	caddr_t argp;
+	struct proc *p;
+{
+	register struct nfsnode *np;
+	struct vnode *vp;
+	struct nfsreq myrep;
+	struct nfsuid *nuidp, *nnuidp;
+	int error = 0, vpid;
+
+	/*
+	 * First initialize some variables
+	 */
+
+	/*
+	 * If an authorization string is being passed in, get it.
+	 */
+	if ((flag & NFSSVC_GOTAUTH) &&
+	    (nmp->nm_flag & (NFSMNT_WAITAUTH | NFSMNT_DISMNT)) == 0) {
+	    if (nmp->nm_flag & NFSMNT_HASAUTH)
+		panic("cld kerb");
+	    if ((flag & NFSSVC_AUTHINFAIL) == 0) {
+		if (ncd->ncd_authlen <= nmp->nm_authlen &&
+		    ncd->ncd_verflen <= nmp->nm_verflen &&
+		    !copyin(ncd->ncd_authstr,nmp->nm_authstr,ncd->ncd_authlen)&&
+		    !copyin(ncd->ncd_verfstr,nmp->nm_verfstr,ncd->ncd_verflen)){
+		    nmp->nm_authtype = ncd->ncd_authtype;
+		    nmp->nm_authlen = ncd->ncd_authlen;
+		    nmp->nm_verflen = ncd->ncd_verflen;
+#ifdef NFSKERB
+		    nmp->nm_key = ncd->ncd_key;
+#endif
+		} else
+		    nmp->nm_flag |= NFSMNT_AUTHERR;
+	    } else
+		nmp->nm_flag |= NFSMNT_AUTHERR;
+	    nmp->nm_flag |= NFSMNT_HASAUTH;
+	    wakeup((caddr_t)&nmp->nm_authlen);
+	} else
+	    nmp->nm_flag |= NFSMNT_WAITAUTH;
+
+	/*
+	 * Loop every second updating queue until there is a termination sig.
+	 */
+	while ((nmp->nm_flag & NFSMNT_DISMNT) == 0) {
+	    if (nmp->nm_flag & NFSMNT_NQNFS) {
+		/*
+		 * If there are no outstanding requests (and therefore no
+		 * processes in nfs_reply) and there is data in the receive
+		 * queue, poke for callbacks.
+		 */
+		if (nfs_reqq.tqh_first == 0 && nmp->nm_so &&
+		    nmp->nm_so->so_rcv.sb_cc > 0) {
+		    myrep.r_flags = R_GETONEREP;
+		    myrep.r_nmp = nmp;
+		    myrep.r_mrep = (struct mbuf *)0;
+		    myrep.r_procp = (struct proc *)0;
+		    (void) nfs_reply(&myrep);
+		}
+
+		/*
+		 * Loop through the leases, updating as required.
+		 */
+		np = nmp->nm_timerhead.cqh_first;
+		while (np != (void *)&nmp->nm_timerhead &&
+		       (nmp->nm_flag & NFSMNT_DISMINPROG) == 0) {
+			vp = NFSTOV(np);
+			vpid = vp->v_id;
+			if (np->n_expiry < time.tv_sec) {
+			   if (vget(vp, LK_EXCLUSIVE, p) == 0) {
+			     nmp->nm_inprog = vp;
+			     if (vpid == vp->v_id) {
+				CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer);
+				np->n_timer.cqe_next = 0;
+				if ((np->n_flag & (NMODIFIED | NQNFSEVICTED))
+				    && vp->v_type == VREG) {
+					if (np->n_flag & NQNFSEVICTED) {
+						(void) nfs_vinvalbuf(vp,
+						       V_SAVE, cred, p, 0);
+						np->n_flag &= ~NQNFSEVICTED;
+						(void) nqnfs_vacated(vp, cred);
+					} else {
+						(void) VOP_FSYNC(vp, cred,
+						    MNT_WAIT, p);
+						np->n_flag &= ~NMODIFIED;
+					}
+				}
+			      }
+			      vrele(vp);
+			      nmp->nm_inprog = NULLVP;
+			    }
+			} else if ((np->n_expiry - NQ_RENEWAL) < time.tv_sec) {
+			    if ((np->n_flag & (NQNFSWRITE | NQNFSNONCACHE))
+				 == NQNFSWRITE && vp->v_dirtyblkhd.lh_first &&
+				 vget(vp, LK_EXCLUSIVE, p) == 0) {
+				 nmp->nm_inprog = vp;
+				 if (vpid == vp->v_id &&
+				     nqnfs_getlease(vp, ND_WRITE, cred, p)==0)
+					np->n_brev = np->n_lrev;
+				 vrele(vp);
+				 nmp->nm_inprog = NULLVP;
+			    }
+			} else
+				break;
+			if (np == nmp->nm_timerhead.cqh_first)
+				break;
+			np = nmp->nm_timerhead.cqh_first;
+		}
+	    }
+
+	    /*
+	     * Get an authorization string, if required.
+	     */
+	    if ((nmp->nm_flag & (NFSMNT_WAITAUTH | NFSMNT_DISMNT | NFSMNT_HASAUTH)) == 0) {
+		ncd->ncd_authuid = nmp->nm_authuid;
+		if (copyout((caddr_t)ncd, argp, sizeof (struct nfsd_cargs)))
+			nmp->nm_flag |= NFSMNT_WAITAUTH;
+		else
+			return (ENEEDAUTH);
+	    }
+
+	    /*
+	     * Wait a bit (no pun) and do it again.
+	     */
+	    if ((nmp->nm_flag & NFSMNT_DISMNT) == 0 &&
+		(nmp->nm_flag & (NFSMNT_WAITAUTH | NFSMNT_HASAUTH))) {
+		    error = tsleep((caddr_t)&nmp->nm_authstr, PSOCK | PCATCH,
+			"nqnfstimr", hz / 3);
+		    if (error == EINTR || error == ERESTART)
+			(void) dounmount(nmp->nm_mountp, 0, p);
+	    }
+	}
+
+	/*
+	 * Finally, we can free up the mount structure.
+	 */
+	for (nuidp = nmp->nm_uidlruhead.tqh_first; nuidp != 0; nuidp = nnuidp) {
+		nnuidp = nuidp->nu_lru.tqe_next;
+		LIST_REMOVE(nuidp, nu_hash);
+		TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
+		free((caddr_t)nuidp, M_NFSUID);
+	}
+	free((caddr_t)nmp, M_NFSMNT);
+	if (error == EWOULDBLOCK)
+		error = 0;
+	return (error);
+}
+
+/*
+ * Adjust all timer queue expiry times when the time of day clock is changed.
+ * Called from the settimeofday() syscall.
+ */
+void
+nqnfs_lease_updatetime(deltat)
+	register int deltat;
+{
+	struct proc *p = curproc;	/* XXX */
+	struct nqlease *lp;
+	struct nfsnode *np;
+	struct mount *mp, *nxtmp;
+	struct nfsmount *nmp;
+	int s;
+
+	if (nqnfsstarttime != 0)
+		nqnfsstarttime += deltat;
+	s = splsoftclock();
+	for (lp = nqtimerhead.cqh_first; lp != (void *)&nqtimerhead;
+	    lp = lp->lc_timer.cqe_next)
+		lp->lc_expiry += deltat;
+	splx(s);
+
+	/*
+	 * Search the mount list for all nqnfs mounts and do their timer
+	 * queues.
+	 */
+	simple_lock(&mountlist_slock);
+	for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nxtmp) {
+		if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) {
+			nxtmp = mp->mnt_list.cqe_next;
+			continue;
+		}
+		if (mp->mnt_stat.f_type == nfs_mount_type) {
+			nmp = VFSTONFS(mp);
+			if (nmp->nm_flag & NFSMNT_NQNFS) {
+				for (np = nmp->nm_timerhead.cqh_first;
+				    np != (void *)&nmp->nm_timerhead;
+				    np = np->n_timer.cqe_next) {
+					np->n_expiry += deltat;
+				}
+			}
+		}
+		simple_lock(&mountlist_slock);
+		nxtmp = mp->mnt_list.cqe_next;
+		vfs_unbusy(mp, p);
+	}
+	simple_unlock(&mountlist_slock);
+}
+
+/*
+ * Lock a server lease.
+ */
+void
+nqsrv_locklease(lp)
+	struct nqlease *lp;
+{
+
+	while (lp->lc_flag & LC_LOCKED) {
+		lp->lc_flag |= LC_WANTED;
+		(void) tsleep((caddr_t)lp, PSOCK, "nqlc", 0);
+	}
+	lp->lc_flag |= LC_LOCKED;
+	lp->lc_flag &= ~LC_WANTED;
+}
+
+/*
+ * Unlock a server lease.
+ */
+void
+nqsrv_unlocklease(lp)
+	struct nqlease *lp;
+{
+
+	lp->lc_flag &= ~LC_LOCKED;
+	if (lp->lc_flag & LC_WANTED)
+		wakeup((caddr_t)lp);
+}
+
+/*
+ * Update a client lease.
+ */
+void
+nqnfs_clientlease(nmp, np, rwflag, cachable, expiry, frev)
+	register struct nfsmount *nmp;
+	register struct nfsnode *np;
+	int rwflag, cachable;
+	time_t expiry;
+	u_quad_t frev;
+{
+	register struct nfsnode *tp;
+
+	if (np->n_timer.cqe_next != 0) {
+		CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer);
+		if (rwflag == ND_WRITE)
+			np->n_flag |= NQNFSWRITE;
+	} else if (rwflag == ND_READ)
+		np->n_flag &= ~NQNFSWRITE;
+	else
+		np->n_flag |= NQNFSWRITE;
+	if (cachable)
+		np->n_flag &= ~NQNFSNONCACHE;
+	else
+		np->n_flag |= NQNFSNONCACHE;
+	np->n_expiry = expiry;
+	np->n_lrev = frev;
+	tp = nmp->nm_timerhead.cqh_last;
+	while (tp != (void *)&nmp->nm_timerhead && tp->n_expiry > np->n_expiry)
+		tp = tp->n_timer.cqe_prev;
+	if (tp == (void *)&nmp->nm_timerhead) {
+		CIRCLEQ_INSERT_HEAD(&nmp->nm_timerhead, np, n_timer);
+	} else {
+		CIRCLEQ_INSERT_AFTER(&nmp->nm_timerhead, tp, np, n_timer);
+	}
+}
diff --git a/sys/nfs/nfs_serv.c b/sys/nfs/nfs_serv.c
new file mode 100644
index 000000000000..3e6ef3d86b05
--- /dev/null
+++ b/sys/nfs/nfs_serv.c
@@ -0,0 +1,3337 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_serv.c	8.7 (Berkeley) 5/14/95
+ */
+
+/*
+ * nfs version 2 and 3 server calls to vnode ops
+ * - these routines generally have 3 phases
+ *   1 - break down and validate rpc request in mbuf list
+ *   2 - do the vnode ops for the request
+ *       (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c)
+ *   3 - build the rpc reply in an mbuf list
+ *   nb:
+ *	- do not mix the phases, since the nfsm_?? macros can return failures
+ *	  on a bad rpc or similar and do not do any vrele() or vput()'s
+ *
+ *      - the nfsm_reply() macro generates an nfs rpc reply with the nfs
+ *	error number iff error != 0 whereas
+ *	returning an error from the server function implies a fatal error
+ *	such as a badly constructed rpc request that should be dropped without
+ *	a reply.
+ *	For Version 3, nfsm_reply() does not return for the error case, since
+ *	most version 3 rpcs return more than the status for error cases.
+ */
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/proc.h>
+#include <sys/file.h>
+#include <sys/namei.h>
+#include <sys/vnode.h>
+#include <sys/mount.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+#include <sys/mbuf.h>
+#include <sys/dirent.h>
+#include <sys/stat.h>
+#include <sys/kernel.h>
+#include <ufs/ufs/dir.h>
+
+#include <vm/vm.h>
+
+#include <nfs/nfsproto.h>
+#include <nfs/rpcv2.h>
+#include <nfs/nfs.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nqnfs.h>
+
+/* Global vars */
+extern u_long nfs_xdrneg1;
+extern u_long nfs_false, nfs_true;
+extern enum vtype nv3tov_type[8];
+extern struct nfsstats nfsstats;
+nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON,
+		      NFCHR, NFNON };
+nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK,
+		      NFFIFO, NFNON };
+int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
+
+/*
+ * nfs v3 access service
+ */
+int
+nfsrv3_access(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, getret;
+	char *cp2;
+	struct mbuf *mb, *mreq, *mb2;
+	struct vattr vattr, *vap = &vattr;
+	u_long testmode, nfsmode;
+	u_quad_t frev;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(1, (struct vattr *)0);
+		return (0);
+	}
+	nfsmode = fxdr_unsigned(u_long, *tl);
+	if ((nfsmode & NFSV3ACCESS_READ) &&
+		nfsrv_access(vp, VREAD, cred, rdonly, procp))
+		nfsmode &= ~NFSV3ACCESS_READ;
+	if (vp->v_type == VDIR)
+		testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
+			NFSV3ACCESS_DELETE);
+	else
+		testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
+	if ((nfsmode & testmode) &&
+		nfsrv_access(vp, VWRITE, cred, rdonly, procp))
+		nfsmode &= ~testmode;
+	if (vp->v_type == VDIR)
+		testmode = NFSV3ACCESS_LOOKUP;
+	else
+		testmode = NFSV3ACCESS_EXECUTE;
+	if ((nfsmode & testmode) &&
+		nfsrv_access(vp, VEXEC, cred, rdonly, procp))
+		nfsmode &= ~testmode;
+	getret = VOP_GETATTR(vp, vap, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
+	nfsm_srvpostop_attr(getret, vap);
+	nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+	*tl = txdr_unsigned(nfsmode);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs getattr service
+ */
+int
+nfsrv_getattr(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nfs_fattr *fp;
+	struct vattr va;
+	register struct vattr *vap = &va;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(0);
+		return (0);
+	}
+	nqsrv_getl(vp, ND_READ);
+	error = VOP_GETATTR(vp, vap, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
+	if (error)
+		return (0);
+	nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
+	nfsm_srvfillattr(vap, fp);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs setattr service
+ */
+int
+nfsrv_setattr(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct vattr va, preat;
+	register struct vattr *vap = &va;
+	register struct nfsv2_sattr *sp;
+	register struct nfs_fattr *fp;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	u_quad_t frev;
+	struct timespec guard;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	VATTR_NULL(vap);
+	if (v3) {
+		nfsm_srvsattr(vap);
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		gcheck = fxdr_unsigned(int, *tl);
+		if (gcheck) {
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+			fxdr_nfsv3time(tl, &guard);
+		}
+	} else {
+		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		/*
+		 * Nah nah nah nah na nah
+		 * There is a bug in the Sun client that puts 0xffff in the mode
+		 * field of sattr when it should put in 0xffffffff. The u_short
+		 * doesn't sign extend.
+		 * --> check the low order 2 bytes for 0xffff
+		 */
+		if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
+			vap->va_mode = nfstov_mode(sp->sa_mode);
+		if (sp->sa_uid != nfs_xdrneg1)
+			vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
+		if (sp->sa_gid != nfs_xdrneg1)
+			vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
+		if (sp->sa_size != nfs_xdrneg1)
+			vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
+		if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
+#ifdef notyet
+			fxdr_nfsv2time(&sp->sa_atime, &vap->va_atime);
+#else
+			vap->va_atime.ts_sec =
+				fxdr_unsigned(long, sp->sa_atime.nfsv2_sec);
+			vap->va_atime.ts_nsec = 0;
+#endif
+		}
+		if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
+			fxdr_nfsv2time(&sp->sa_mtime, &vap->va_mtime);
+
+	}
+
+	/*
+	 * Now that we have all the fields, lets do it.
+	 */
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
+		return (0);
+	}
+	nqsrv_getl(vp, ND_WRITE);
+	if (v3) {
+		error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
+		if (!error && gcheck &&
+			(preat.va_ctime.ts_sec != guard.ts_sec ||
+			 preat.va_ctime.ts_nsec != guard.ts_nsec))
+			error = NFSERR_NOT_SYNC;
+		if (error) {
+			vput(vp);
+			nfsm_reply(NFSX_WCCDATA(v3));
+			nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
+			return (0);
+		}
+	}
+
+	/*
+	 * If the size is being changed write acces is required, otherwise
+	 * just check for a read only file system.
+	 */
+	if (vap->va_size == ((u_quad_t)((quad_t) -1))) {
+		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
+			error = EROFS;
+			goto out;
+		}
+	} else {
+		if (vp->v_type == VDIR) {
+			error = EISDIR;
+			goto out;
+		} else if (error = nfsrv_access(vp, VWRITE, cred, rdonly,
+			procp))
+			goto out;
+	}
+	error = VOP_SETATTR(vp, vap, cred, procp);
+	postat_ret = VOP_GETATTR(vp, vap, cred, procp);
+	if (!error)
+		error = postat_ret;
+out:
+	vput(vp);
+	nfsm_reply(NFSX_WCCORFATTR(v3));
+	if (v3) {
+		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
+		return (0);
+	} else {
+		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+		nfsm_srvfillattr(vap, fp);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs lookup rpc
+ */
+int
+nfsrv_lookup(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nfs_fattr *fp;
+	struct nameidata nd;
+	struct vnode *vp, *dirp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register caddr_t cp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, dirattr_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vattr va, dirattr, *vap = &va;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = LOOKUP;
+	nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
+				procp);
+		vrele(dirp);
+	}
+	if (error) {
+		nfsm_reply(NFSX_POSTOPATTR(v3));
+		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
+		return (0);
+	}
+	nqsrv_getl(nd.ni_startdir, ND_READ);
+	vrele(nd.ni_startdir);
+	FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	vp = nd.ni_vp;
+	bzero((caddr_t)fhp, sizeof(nfh));
+	fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+	error = VFS_VPTOFH(vp, &fhp->fh_fid);
+	if (!error)
+		error = VOP_GETATTR(vp, vap, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
+	if (error) {
+		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
+		return (0);
+	}
+	nfsm_srvfhtom(fhp, v3);
+	if (v3) {
+		nfsm_srvpostop_attr(0, vap);
+		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
+	} else {
+		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+		nfsm_srvfillattr(vap, fp);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs readlink service
+ */
+int
+nfsrv_readlink(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
+	register struct iovec *ivp = iv;
+	register struct mbuf *mp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, i, tlen, len, getret;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mb2, *mp2, *mp3, *mreq;
+	struct vnode *vp;
+	struct vattr attr;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct uio io, *uiop = &io;
+	u_quad_t frev;
+
+#ifndef nolint
+	mp2 = mp3 = (struct mbuf *)0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	len = 0;
+	i = 0;
+	while (len < NFS_MAXPATHLEN) {
+		MGET(mp, M_WAIT, MT_DATA);
+		MCLGET(mp, M_WAIT);
+		mp->m_len = NFSMSIZ(mp);
+		if (len == 0)
+			mp3 = mp2 = mp;
+		else {
+			mp2->m_next = mp;
+			mp2 = mp;
+		}
+		if ((len+mp->m_len) > NFS_MAXPATHLEN) {
+			mp->m_len = NFS_MAXPATHLEN-len;
+			len = NFS_MAXPATHLEN;
+		} else
+			len += mp->m_len;
+		ivp->iov_base = mtod(mp, caddr_t);
+		ivp->iov_len = mp->m_len;
+		i++;
+		ivp++;
+	}
+	uiop->uio_iov = iv;
+	uiop->uio_iovcnt = i;
+	uiop->uio_offset = 0;
+	uiop->uio_resid = len;
+	uiop->uio_rw = UIO_READ;
+	uiop->uio_segflg = UIO_SYSSPACE;
+	uiop->uio_procp = (struct proc *)0;
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		m_freem(mp3);
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(1, (struct vattr *)0);
+		return (0);
+	}
+	if (vp->v_type != VLNK) {
+		if (v3)
+			error = EINVAL;
+		else
+			error = ENXIO;
+		goto out;
+	}
+	nqsrv_getl(vp, ND_READ);
+	error = VOP_READLINK(vp, uiop, cred);
+out:
+	getret = VOP_GETATTR(vp, &attr, cred, procp);
+	vput(vp);
+	if (error)
+		m_freem(mp3);
+	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
+	if (v3) {
+		nfsm_srvpostop_attr(getret, &attr);
+		if (error)
+			return (0);
+	}
+	if (uiop->uio_resid > 0) {
+		len -= uiop->uio_resid;
+		tlen = nfsm_rndup(len);
+		nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
+	}
+	nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+	*tl = txdr_unsigned(len);
+	mb->m_next = mp3;
+	nfsm_srvdone;
+}
+
+/*
+ * nfs read service
+ */
+int
+nfsrv_read(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct iovec *iv;
+	struct iovec *iv2;
+	register struct mbuf *m;
+	register struct nfs_fattr *fp;
+	register u_long *tl;
+	register long t1;
+	register int i;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
+	int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct mbuf *m2;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct uio io, *uiop = &io;
+	struct vattr va, *vap = &va;
+	off_t off;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (v3) {
+		nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+		fxdr_hyper(tl, &off);
+	} else {
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		off = (off_t)fxdr_unsigned(u_long, *tl);
+	}
+	nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(1, (struct vattr *)0);
+		return (0);
+	}
+	if (vp->v_type != VREG) {
+		if (v3)
+			error = EINVAL;
+		else
+			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
+	}
+	if (!error) {
+	    nqsrv_getl(vp, ND_READ);
+	    if (error = nfsrv_access(vp, VREAD, cred, rdonly, procp))
+		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp);
+	}
+	getret = VOP_GETATTR(vp, vap, cred, procp);
+	if (!error)
+		error = getret;
+	if (error) {
+		vput(vp);
+		nfsm_reply(NFSX_POSTOPATTR(v3));
+		nfsm_srvpostop_attr(getret, vap);
+		return (0);
+	}
+	if (off >= vap->va_size)
+		cnt = 0;
+	else if ((off + reqlen) > vap->va_size)
+		cnt = nfsm_rndup(vap->va_size - off);
+	else
+		cnt = reqlen;
+	nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
+	if (v3) {
+		nfsm_build(tl, u_long *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
+		*tl++ = nfs_true;
+		fp = (struct nfs_fattr *)tl;
+		tl += (NFSX_V3FATTR / sizeof (u_long));
+	} else {
+		nfsm_build(tl, u_long *, NFSX_V2FATTR + NFSX_UNSIGNED);
+		fp = (struct nfs_fattr *)tl;
+		tl += (NFSX_V2FATTR / sizeof (u_long));
+	}
+	len = left = cnt;
+	if (cnt > 0) {
+		/*
+		 * Generate the mbuf list with the uio_iov ref. to it.
+		 */
+		i = 0;
+		m = m2 = mb;
+		while (left > 0) {
+			siz = min(M_TRAILINGSPACE(m), left);
+			if (siz > 0) {
+				left -= siz;
+				i++;
+			}
+			if (left > 0) {
+				MGET(m, M_WAIT, MT_DATA);
+				MCLGET(m, M_WAIT);
+				m->m_len = 0;
+				m2->m_next = m;
+				m2 = m;
+			}
+		}
+		MALLOC(iv, struct iovec *, i * sizeof (struct iovec),
+		       M_TEMP, M_WAITOK);
+		uiop->uio_iov = iv2 = iv;
+		m = mb;
+		left = cnt;
+		i = 0;
+		while (left > 0) {
+			if (m == NULL)
+				panic("nfsrv_read iov");
+			siz = min(M_TRAILINGSPACE(m), left);
+			if (siz > 0) {
+				iv->iov_base = mtod(m, caddr_t) + m->m_len;
+				iv->iov_len = siz;
+				m->m_len += siz;
+				left -= siz;
+				iv++;
+				i++;
+			}
+			m = m->m_next;
+		}
+		uiop->uio_iovcnt = i;
+		uiop->uio_offset = off;
+		uiop->uio_resid = cnt;
+		uiop->uio_rw = UIO_READ;
+		uiop->uio_segflg = UIO_SYSSPACE;
+		error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
+		off = uiop->uio_offset;
+		FREE((caddr_t)iv2, M_TEMP);
+		if (error || (getret = VOP_GETATTR(vp, vap, cred, procp))) {
+			if (!error)
+				error = getret;
+			m_freem(mreq);
+			vput(vp);
+			nfsm_reply(NFSX_POSTOPATTR(v3));
+			nfsm_srvpostop_attr(getret, vap);
+			return (0);
+		}
+	} else
+		uiop->uio_resid = 0;
+	vput(vp);
+	nfsm_srvfillattr(vap, fp);
+	len -= uiop->uio_resid;
+	tlen = nfsm_rndup(len);
+	if (cnt != tlen || tlen != len)
+		nfsm_adj(mb, cnt - tlen, tlen - len);
+	if (v3) {
+		*tl++ = txdr_unsigned(len);
+		if (len < reqlen)
+			*tl++ = nfs_true;
+		else
+			*tl++ = nfs_false;
+	}
+	*tl = txdr_unsigned(len);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs write service
+ */
+int
+nfsrv_write(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct iovec *ivp;
+	register int i, cnt;
+	register struct mbuf *mp;
+	register struct nfs_fattr *fp;
+	struct iovec *iv;
+	struct vattr va, forat;
+	register struct vattr *vap = &va;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, siz, len, xfer, forat_ret = 1;
+	int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
+	int stable = NFSV3WRITE_FILESYNC;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct uio io, *uiop = &io;
+	off_t off;
+	u_quad_t frev;
+
+	if (mrep == NULL) {
+		*mrq = NULL;
+		return (0);
+	}
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (v3) {
+		nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED);
+		fxdr_hyper(tl, &off);
+		tl += 3;
+		stable = fxdr_unsigned(int, *tl++);
+	} else {
+		nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED);
+		off = (off_t)fxdr_unsigned(u_long, *++tl);
+		tl += 2;
+	}
+	retlen = len = fxdr_unsigned(long, *tl);
+	cnt = i = 0;
+
+	/*
+	 * For NFS Version 2, it is not obvious what a write of zero length
+	 * should do, but I might as well be consistent with Version 3,
+	 * which is to return ok so long as there are no permission problems.
+	 */
+	if (len > 0) {
+	    zeroing = 1;
+	    mp = mrep;
+	    while (mp) {
+		if (mp == md) {
+			zeroing = 0;
+			adjust = dpos - mtod(mp, caddr_t);
+			mp->m_len -= adjust;
+			if (mp->m_len > 0 && adjust > 0)
+				NFSMADV(mp, adjust);
+		}
+		if (zeroing)
+			mp->m_len = 0;
+		else if (mp->m_len > 0) {
+			i += mp->m_len;
+			if (i > len) {
+				mp->m_len -= (i - len);
+				zeroing	= 1;
+			}
+			if (mp->m_len > 0)
+				cnt++;
+		}
+		mp = mp->m_next;
+	    }
+	}
+	if (len > NFS_MAXDATA || len < 0 || i < len) {
+		error = EIO;
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
+		return (0);
+	}
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
+		return (0);
+	}
+	if (v3)
+		forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
+	if (vp->v_type != VREG) {
+		if (v3)
+			error = EINVAL;
+		else
+			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
+	}
+	if (!error) {
+		nqsrv_getl(vp, ND_WRITE);
+		error = nfsrv_access(vp, VWRITE, cred, rdonly, procp);
+	}
+	if (error) {
+		vput(vp);
+		nfsm_reply(NFSX_WCCDATA(v3));
+		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
+		return (0);
+	}
+
+	if (len > 0) {
+	    MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP,
+		M_WAITOK);
+	    uiop->uio_iov = iv = ivp;
+	    uiop->uio_iovcnt = cnt;
+	    mp = mrep;
+	    while (mp) {
+		if (mp->m_len > 0) {
+			ivp->iov_base = mtod(mp, caddr_t);
+			ivp->iov_len = mp->m_len;
+			ivp++;
+		}
+		mp = mp->m_next;
+	    }
+
+	    /*
+	     * XXX
+	     * The IO_METASYNC flag indicates that all metadata (and not just
+	     * enough to ensure data integrity) mus be written to stable storage
+	     * synchronously.
+	     * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
+	     */
+	    if (stable == NFSV3WRITE_UNSTABLE)
+		ioflags = IO_NODELOCKED;
+	    else if (stable == NFSV3WRITE_DATASYNC)
+		ioflags = (IO_SYNC | IO_NODELOCKED);
+	    else
+		ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
+	    uiop->uio_resid = len;
+	    uiop->uio_rw = UIO_WRITE;
+	    uiop->uio_segflg = UIO_SYSSPACE;
+	    uiop->uio_procp = (struct proc *)0;
+	    uiop->uio_offset = off;
+	    error = VOP_WRITE(vp, uiop, ioflags, cred);
+	    nfsstats.srvvop_writes++;
+	    FREE((caddr_t)iv, M_TEMP);
+	}
+	aftat_ret = VOP_GETATTR(vp, vap, cred, procp);
+	vput(vp);
+	if (!error)
+		error = aftat_ret;
+	nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
+		2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
+	if (v3) {
+		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
+		if (error)
+			return (0);
+		nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+		*tl++ = txdr_unsigned(retlen);
+		if (stable == NFSV3WRITE_UNSTABLE)
+			*tl++ = txdr_unsigned(stable);
+		else
+			*tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
+		/*
+		 * Actually, there is no need to txdr these fields,
+		 * but it may make the values more human readable,
+		 * for debugging purposes.
+		 */
+		*tl++ = txdr_unsigned(boottime.tv_sec);
+		*tl = txdr_unsigned(boottime.tv_usec);
+	} else {
+		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+		nfsm_srvfillattr(vap, fp);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * NFS write service with write gathering support. Called when
+ * nfsrvw_procrastinate > 0.
+ * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
+ * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
+ * Jan. 1994.
+ */
+int
+nfsrv_writegather(ndp, slp, procp, mrq)
+	struct nfsrv_descript **ndp;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	register struct iovec *ivp;
+	register struct mbuf *mp;
+	register struct nfsrv_descript *wp, *nfsd, *owp, *swp;
+	register struct nfs_fattr *fp;
+	register int i;
+	struct iovec *iov;
+	struct nfsrvw_delayhash *wpp;
+	struct ucred *cred;
+	struct vattr va, forat;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos, dpos;
+	int error = 0, rdonly, cache, len, forat_ret = 1;
+	int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq, *mrep, *md;
+	struct vnode *vp;
+	struct uio io, *uiop = &io;
+	off_t off;
+	u_quad_t frev, cur_usec;
+
+#ifndef nolint
+	i = 0;
+	len = 0;
+#endif
+	*mrq = NULL;
+	if (*ndp) {
+	    nfsd = *ndp;
+	    *ndp = NULL;
+	    mrep = nfsd->nd_mrep;
+	    md = nfsd->nd_md;
+	    dpos = nfsd->nd_dpos;
+	    cred = &nfsd->nd_cr;
+	    v3 = (nfsd->nd_flag & ND_NFSV3);
+	    LIST_INIT(&nfsd->nd_coalesce);
+	    nfsd->nd_mreq = NULL;
+	    nfsd->nd_stable = NFSV3WRITE_FILESYNC;
+	    cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
+	    nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
+    
+	    /*
+	     * Now, get the write header..
+	     */
+	    nfsm_srvmtofh(&nfsd->nd_fh);
+	    if (v3) {
+		nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED);
+		fxdr_hyper(tl, &nfsd->nd_off);
+		tl += 3;
+		nfsd->nd_stable = fxdr_unsigned(int, *tl++);
+	    } else {
+		nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED);
+		nfsd->nd_off = (off_t)fxdr_unsigned(u_long, *++tl);
+		tl += 2;
+	    }
+	    len = fxdr_unsigned(long, *tl);
+	    nfsd->nd_len = len;
+	    nfsd->nd_eoff = nfsd->nd_off + len;
+    
+	    /*
+	     * Trim the header out of the mbuf list and trim off any trailing
+	     * junk so that the mbuf list has only the write data.
+	     */
+	    zeroing = 1;
+	    i = 0;
+	    mp = mrep;
+	    while (mp) {
+		if (mp == md) {
+		    zeroing = 0;
+		    adjust = dpos - mtod(mp, caddr_t);
+		    mp->m_len -= adjust;
+		    if (mp->m_len > 0 && adjust > 0)
+			NFSMADV(mp, adjust);
+		}
+		if (zeroing)
+		    mp->m_len = 0;
+		else {
+		    i += mp->m_len;
+		    if (i > len) {
+			mp->m_len -= (i - len);
+			zeroing = 1;
+		    }
+		}
+		mp = mp->m_next;
+	    }
+	    if (len > NFS_MAXDATA || len < 0  || i < len) {
+nfsmout:
+		m_freem(mrep);
+		error = EIO;
+		nfsm_writereply(2 * NFSX_UNSIGNED, v3);
+		if (v3)
+		    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
+		nfsd->nd_mreq = mreq;
+		nfsd->nd_mrep = NULL;
+		nfsd->nd_time = 0;
+	    }
+    
+	    /*
+	     * Add this entry to the hash and time queues.
+	     */
+	    s = splsoftclock();
+	    owp = NULL;
+	    wp = slp->ns_tq.lh_first;
+	    while (wp && wp->nd_time < nfsd->nd_time) {
+		owp = wp;
+		wp = wp->nd_tq.le_next;
+	    }
+	    if (owp) {
+		LIST_INSERT_AFTER(owp, nfsd, nd_tq);
+	    } else {
+		LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
+	    }
+	    if (nfsd->nd_mrep) {
+		wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
+		owp = NULL;
+		wp = wpp->lh_first;
+		while (wp &&
+		    bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
+		    owp = wp;
+		    wp = wp->nd_hash.le_next;
+		}
+		while (wp && wp->nd_off < nfsd->nd_off &&
+		    !bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
+		    owp = wp;
+		    wp = wp->nd_hash.le_next;
+		}
+		if (owp) {
+		    LIST_INSERT_AFTER(owp, nfsd, nd_hash);
+
+		    /*
+		     * Search the hash list for overlapping entries and
+		     * coalesce.
+		     */
+		    for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
+			wp = nfsd->nd_hash.le_next;
+			if (NFSW_SAMECRED(owp, nfsd))
+			    nfsrvw_coalesce(owp, nfsd);
+		    }
+		} else {
+		    LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
+		}
+	    }
+	    splx(s);
+	}
+    
+	/*
+	 * Now, do VOP_WRITE()s for any one(s) that need to be done now
+	 * and generate the associated reply mbuf list(s).
+	 */
+loop1:
+	cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
+	s = splsoftclock();
+	for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
+		owp = nfsd->nd_tq.le_next;
+		if (nfsd->nd_time > cur_usec)
+		    break;
+		if (nfsd->nd_mreq)
+		    continue;
+		LIST_REMOVE(nfsd, nd_tq);
+		LIST_REMOVE(nfsd, nd_hash);
+		splx(s);
+		mrep = nfsd->nd_mrep;
+		nfsd->nd_mrep = NULL;
+		cred = &nfsd->nd_cr;
+		v3 = (nfsd->nd_flag & ND_NFSV3);
+		forat_ret = aftat_ret = 1;
+		error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, 
+		    nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH));
+		if (!error) {
+		    if (v3)
+			forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
+		    if (vp->v_type != VREG) {
+			if (v3)
+			    error = EINVAL;
+			else
+			    error = (vp->v_type == VDIR) ? EISDIR : EACCES;
+		    }
+		} else
+		    vp = NULL;
+		if (!error) {
+		    nqsrv_getl(vp, ND_WRITE);
+		    error = nfsrv_access(vp, VWRITE, cred, rdonly, procp);
+		}
+    
+		if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
+		    ioflags = IO_NODELOCKED;
+		else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
+		    ioflags = (IO_SYNC | IO_NODELOCKED);
+		else
+		    ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
+		uiop->uio_rw = UIO_WRITE;
+		uiop->uio_segflg = UIO_SYSSPACE;
+		uiop->uio_procp = (struct proc *)0;
+		uiop->uio_offset = nfsd->nd_off;
+		uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
+		if (uiop->uio_resid > 0) {
+		    mp = mrep;
+		    i = 0;
+		    while (mp) {
+			if (mp->m_len > 0)
+			    i++;
+			mp = mp->m_next;
+		    }
+		    uiop->uio_iovcnt = i;
+		    MALLOC(iov, struct iovec *, i * sizeof (struct iovec), 
+			M_TEMP, M_WAITOK);
+		    uiop->uio_iov = ivp = iov;
+		    mp = mrep;
+		    while (mp) {
+			if (mp->m_len > 0) {
+			    ivp->iov_base = mtod(mp, caddr_t);
+			    ivp->iov_len = mp->m_len;
+			    ivp++;
+			}
+			mp = mp->m_next;
+		    }
+		    if (!error) {
+			error = VOP_WRITE(vp, uiop, ioflags, cred);
+			nfsstats.srvvop_writes++;
+		    }
+		    FREE((caddr_t)iov, M_TEMP);
+		}
+		m_freem(mrep);
+		if (vp) {
+		    aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
+		    vput(vp);
+		}
+
+		/*
+		 * Loop around generating replies for all write rpcs that have
+		 * now been completed.
+		 */
+		swp = nfsd;
+		do {
+		    if (error) {
+			nfsm_writereply(NFSX_WCCDATA(v3), v3);
+			if (v3) {
+			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
+			}
+		    } else {
+			nfsm_writereply(NFSX_PREOPATTR(v3) +
+			    NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
+			    NFSX_WRITEVERF(v3), v3);
+			if (v3) {
+			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
+			    nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+			    *tl++ = txdr_unsigned(nfsd->nd_len);
+			    *tl++ = txdr_unsigned(swp->nd_stable);
+			    /*
+			     * Actually, there is no need to txdr these fields,
+			     * but it may make the values more human readable,
+			     * for debugging purposes.
+			     */
+			    *tl++ = txdr_unsigned(boottime.tv_sec);
+			    *tl = txdr_unsigned(boottime.tv_usec);
+			} else {
+			    nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+			    nfsm_srvfillattr(&va, fp);
+			}
+		    }
+		    nfsd->nd_mreq = mreq;
+		    if (nfsd->nd_mrep)
+			panic("nfsrv_write: nd_mrep not free");
+
+		    /*
+		     * Done. Put it at the head of the timer queue so that
+		     * the final phase can return the reply.
+		     */
+		    s = splsoftclock();
+		    if (nfsd != swp) {
+			nfsd->nd_time = 0;
+			LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
+		    }
+		    nfsd = swp->nd_coalesce.lh_first;
+		    if (nfsd) {
+			LIST_REMOVE(nfsd, nd_tq);
+		    }
+		    splx(s);
+		} while (nfsd);
+		s = splsoftclock();
+		swp->nd_time = 0;
+		LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
+		splx(s);
+		goto loop1;
+	}
+	splx(s);
+
+	/*
+	 * Search for a reply to return.
+	 */
+	s = splsoftclock();
+	for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
+		if (nfsd->nd_mreq) {
+		    LIST_REMOVE(nfsd, nd_tq);
+		    *mrq = nfsd->nd_mreq;
+		    *ndp = nfsd;
+		    break;
+		}
+	splx(s);
+	return (0);
+}
+
+/*
+ * Coalesce the write request nfsd into owp. To do this we must:
+ * - remove nfsd from the queues
+ * - merge nfsd->nd_mrep into owp->nd_mrep
+ * - update the nd_eoff and nd_stable for owp
+ * - put nfsd on owp's nd_coalesce list
+ * NB: Must be called at splsoftclock().
+ */
+void
+nfsrvw_coalesce(owp, nfsd)
+        register struct nfsrv_descript *owp;
+        register struct nfsrv_descript *nfsd;
+{
+        register int overlap;
+        register struct mbuf *mp;
+
+        LIST_REMOVE(nfsd, nd_hash);
+        LIST_REMOVE(nfsd, nd_tq);
+        if (owp->nd_eoff < nfsd->nd_eoff) {
+            overlap = owp->nd_eoff - nfsd->nd_off;
+            if (overlap < 0)
+                panic("nfsrv_coalesce: bad off");
+            if (overlap > 0)
+                m_adj(nfsd->nd_mrep, overlap);
+            mp = owp->nd_mrep;
+            while (mp->m_next)
+                mp = mp->m_next;
+            mp->m_next = nfsd->nd_mrep;
+            owp->nd_eoff = nfsd->nd_eoff;
+        } else
+            m_freem(nfsd->nd_mrep);
+        nfsd->nd_mrep = NULL;
+        if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
+            owp->nd_stable = NFSV3WRITE_FILESYNC;
+        else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
+            owp->nd_stable == NFSV3WRITE_UNSTABLE)
+            owp->nd_stable = NFSV3WRITE_DATASYNC;
+        LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
+}
+
+/*
+ * Sort the group list in increasing numerical order.
+ * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
+ *  that used to be here.)
+ */
+void
+nfsrvw_sort(list, num)
+        register gid_t *list;
+        register int num;
+{
+	register int i, j;
+	gid_t v;
+
+	/* Insertion sort. */
+	for (i = 1; i < num; i++) {
+		v = list[i];
+		/* find correct slot for value v, moving others up */
+		for (j = i; --j >= 0 && v < list[j];)
+			list[j + 1] = list[j];
+		list[j + 1] = v;
+	}
+}
+
+/*
+ * copy credentials making sure that the result can be compared with bcmp().
+ */
+void
+nfsrv_setcred(incred, outcred)
+	register struct ucred *incred, *outcred;
+{
+	register int i;
+
+	bzero((caddr_t)outcred, sizeof (struct ucred));
+	outcred->cr_ref = 1;
+	outcred->cr_uid = incred->cr_uid;
+	outcred->cr_ngroups = incred->cr_ngroups;
+	for (i = 0; i < incred->cr_ngroups; i++)
+		outcred->cr_groups[i] = incred->cr_groups[i];
+	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
+}
+
+/*
+ * nfs create service
+ * now does a truncate to 0 length via. setattr if it already exists
+ */
+int
+nfsrv_create(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nfs_fattr *fp;
+	struct vattr va, dirfor, diraft;
+	register struct vattr *vap = &va;
+	register struct nfsv2_sattr *sp;
+	register u_long *tl;
+	struct nameidata nd;
+	register caddr_t cp;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdev, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp, *dirp = (struct vnode *)0;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev, tempsize;
+	u_char cverf[NFSX_V3CREATEVERF];
+
+#ifndef nolint
+	rdev = 0;
+#endif
+	nd.ni_cnd.cn_nameiop = 0;
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = CREATE;
+	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else {
+			vrele(dirp);
+			dirp = (struct vnode *)0;
+		}
+	}
+	if (error) {
+		nfsm_reply(NFSX_WCCDATA(v3));
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		if (dirp)
+			vrele(dirp);
+		return (0);
+	}
+	VATTR_NULL(vap);
+	if (v3) {
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		how = fxdr_unsigned(int, *tl);
+		switch (how) {
+		case NFSV3CREATE_GUARDED:
+			if (nd.ni_vp) {
+				error = EEXIST;
+				break;
+			}
+		case NFSV3CREATE_UNCHECKED:
+			nfsm_srvsattr(vap);
+			break;
+		case NFSV3CREATE_EXCLUSIVE:
+			nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
+			bcopy(cp, cverf, NFSX_V3CREATEVERF);
+			exclusive_flag = 1;
+			if (nd.ni_vp == NULL)
+				vap->va_mode = 0;
+			break;
+		};
+		vap->va_type = VREG;
+	} else {
+		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		vap->va_type = IFTOVT(fxdr_unsigned(u_long, sp->sa_mode));
+		if (vap->va_type == VNON)
+			vap->va_type = VREG;
+		vap->va_mode = nfstov_mode(sp->sa_mode);
+		switch (vap->va_type) {
+		case VREG:
+			tsize = fxdr_unsigned(long, sp->sa_size);
+			if (tsize != -1)
+				vap->va_size = (u_quad_t)tsize;
+			break;
+		case VCHR:
+		case VBLK:
+		case VFIFO:
+			rdev = fxdr_unsigned(long, sp->sa_size);
+			break;
+		};
+	}
+
+	/*
+	 * Iff doesn't exist, create it
+	 * otherwise just truncate to 0 length
+	 *   should I set the mode too ??
+	 */
+	if (nd.ni_vp == NULL) {
+		if (vap->va_type == VREG || vap->va_type == VSOCK) {
+			vrele(nd.ni_startdir);
+			nqsrv_getl(nd.ni_dvp, ND_WRITE);
+			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap);
+			if (!error) {
+				FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+				if (exclusive_flag) {
+					exclusive_flag = 0;
+					VATTR_NULL(vap);
+					bcopy(cverf, (caddr_t)&vap->va_atime,
+						NFSX_V3CREATEVERF);
+					error = VOP_SETATTR(nd.ni_vp, vap, cred,
+						procp);
+				}
+			}
+		} else if (vap->va_type == VCHR || vap->va_type == VBLK ||
+			vap->va_type == VFIFO) {
+			if (vap->va_type == VCHR && rdev == 0xffffffff)
+				vap->va_type = VFIFO;
+			if (error = suser(cred, (u_short *)0)) {
+				vrele(nd.ni_startdir);
+				free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+				vput(nd.ni_dvp);
+				nfsm_reply(0);
+				return (error);
+			} else
+				vap->va_rdev = (dev_t)rdev;
+			nqsrv_getl(nd.ni_dvp, ND_WRITE);
+			if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) {
+				vrele(nd.ni_startdir);
+				nfsm_reply(0);
+			}
+			nd.ni_cnd.cn_nameiop = LOOKUP;
+			nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
+			nd.ni_cnd.cn_proc = procp;
+			nd.ni_cnd.cn_cred = cred;
+			if (error = lookup(&nd)) {
+				free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+				nfsm_reply(0);
+			}
+			FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+			if (nd.ni_cnd.cn_flags & ISSYMLINK) {
+				vrele(nd.ni_dvp);
+				vput(nd.ni_vp);
+				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+				error = EINVAL;
+				nfsm_reply(0);
+			}
+		} else {
+			vrele(nd.ni_startdir);
+			free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+			vput(nd.ni_dvp);
+			error = ENXIO;
+		}
+		vp = nd.ni_vp;
+	} else {
+		vrele(nd.ni_startdir);
+		free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+		vp = nd.ni_vp;
+		if (nd.ni_dvp == vp)
+			vrele(nd.ni_dvp);
+		else
+			vput(nd.ni_dvp);
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		if (vap->va_size != -1) {
+			error = nfsrv_access(vp, VWRITE, cred,
+			    (nd.ni_cnd.cn_flags & RDONLY), procp);
+			if (!error) {
+				nqsrv_getl(vp, ND_WRITE);
+				tempsize = vap->va_size;
+				VATTR_NULL(vap);
+				vap->va_size = tempsize;
+				error = VOP_SETATTR(vp, vap, cred,
+					 procp);
+			}
+			if (error)
+				vput(vp);
+		}
+	}
+	if (!error) {
+		bzero((caddr_t)fhp, sizeof(nfh));
+		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+		error = VFS_VPTOFH(vp, &fhp->fh_fid);
+		if (!error)
+			error = VOP_GETATTR(vp, vap, cred, procp);
+		vput(vp);
+	}
+	if (v3) {
+		if (exclusive_flag && !error &&
+			bcmp(cverf, (caddr_t)&vap->va_atime, NFSX_V3CREATEVERF))
+			error = EEXIST;
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
+	if (v3) {
+		if (!error) {
+			nfsm_srvpostop_fh(fhp);
+			nfsm_srvpostop_attr(0, vap);
+		}
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+	} else {
+		nfsm_srvfhtom(fhp, v3);
+		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+		nfsm_srvfillattr(vap, fp);
+	}
+	return (0);
+nfsmout:
+	if (dirp)
+		vrele(dirp);
+	if (nd.ni_cnd.cn_nameiop) {
+		vrele(nd.ni_startdir);
+		free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	}
+	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+	if (nd.ni_dvp == nd.ni_vp)
+		vrele(nd.ni_dvp);
+	else
+		vput(nd.ni_dvp);
+	if (nd.ni_vp)
+		vput(nd.ni_vp);
+	return (error);
+}
+
+/*
+ * nfs v3 mknod service
+ */
+int
+nfsrv_mknod(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct nfs_fattr *fp;
+	struct vattr va, dirfor, diraft;
+	register struct vattr *vap = &va;
+	register u_long *tl;
+	struct nameidata nd;
+	register caddr_t cp;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
+	u_long major, minor;
+	enum vtype vtyp;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp, *dirp = (struct vnode *)0;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+	nd.ni_cnd.cn_nameiop = 0;
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = CREATE;
+	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp)
+		dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
+	if (error) {
+		nfsm_reply(NFSX_WCCDATA(1));
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		if (dirp)
+			vrele(dirp);
+		return (0);
+	}
+	nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+	vtyp = nfsv3tov_type(*tl);
+	if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
+		vrele(nd.ni_startdir);
+		free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
+		error = NFSERR_BADTYPE;
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		vput(nd.ni_dvp);
+		goto out;
+	}
+	VATTR_NULL(vap);
+	nfsm_srvsattr(vap);
+	if (vtyp == VCHR || vtyp == VBLK) {
+		nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+		major = fxdr_unsigned(u_long, *tl++);
+		minor = fxdr_unsigned(u_long, *tl);
+		vap->va_rdev = makedev(major, minor);
+	}
+
+	/*
+	 * Iff doesn't exist, create it.
+	 */
+	if (nd.ni_vp) {
+		vrele(nd.ni_startdir);
+		free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
+		error = EEXIST;
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		vput(nd.ni_dvp);
+		goto out;
+	}
+	vap->va_type = vtyp;
+	if (vtyp == VSOCK) {
+		vrele(nd.ni_startdir);
+		nqsrv_getl(nd.ni_dvp, ND_WRITE);
+		error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap);
+		if (!error)
+			FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	} else {
+		if (error = suser(cred, (u_short *)0)) {
+			vrele(nd.ni_startdir);
+			free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
+			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+			vput(nd.ni_dvp);
+			goto out;
+		}
+		nqsrv_getl(nd.ni_dvp, ND_WRITE);
+		if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) {
+			vrele(nd.ni_startdir);
+			goto out;
+		}
+		nd.ni_cnd.cn_nameiop = LOOKUP;
+		nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
+		nd.ni_cnd.cn_proc = procp;
+		nd.ni_cnd.cn_cred = procp->p_ucred;
+		error = lookup(&nd);
+		FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+		if (error)
+			goto out;
+		if (nd.ni_cnd.cn_flags & ISSYMLINK) {
+			vrele(nd.ni_dvp);
+			vput(nd.ni_vp);
+			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+			error = EINVAL;
+		}
+	}
+out:
+	vp = nd.ni_vp;
+	if (!error) {
+		bzero((caddr_t)fhp, sizeof(nfh));
+		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+		error = VFS_VPTOFH(vp, &fhp->fh_fid);
+		if (!error)
+			error = VOP_GETATTR(vp, vap, cred, procp);
+		vput(vp);
+	}
+	diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+	vrele(dirp);
+	nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
+	if (!error) {
+		nfsm_srvpostop_fh(fhp);
+		nfsm_srvpostop_attr(0, vap);
+	}
+	nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+	return (0);
+nfsmout:
+	if (dirp)
+		vrele(dirp);
+	if (nd.ni_cnd.cn_nameiop) {
+		vrele(nd.ni_startdir);
+		free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	}
+	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+	if (nd.ni_dvp == nd.ni_vp)
+		vrele(nd.ni_dvp);
+	else
+		vput(nd.ni_dvp);
+	if (nd.ni_vp)
+		vput(nd.ni_vp);
+	return (error);
+}
+
+/*
+ * nfs remove service
+ */
+int
+nfsrv_remove(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct nameidata nd;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mreq, *mb2;
+	struct vnode *vp, *dirp;
+	struct vattr dirfor, diraft;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+#ifndef nolint
+	vp = (struct vnode *)0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = DELETE;
+	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else
+			vrele(dirp);
+	}
+	if (!error) {
+		vp = nd.ni_vp;
+		if (vp->v_type == VDIR &&
+			(error = suser(cred, (u_short *)0)))
+			goto out;
+		/*
+		 * The root of a mounted filesystem cannot be deleted.
+		 */
+		if (vp->v_flag & VROOT) {
+			error = EBUSY;
+			goto out;
+		}
+		if (vp->v_flag & VTEXT)
+			(void) vnode_pager_uncache(vp);
+out:
+		if (!error) {
+			nqsrv_getl(nd.ni_dvp, ND_WRITE);
+			nqsrv_getl(vp, ND_WRITE);
+			error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
+		} else {
+			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+			if (nd.ni_dvp == vp)
+				vrele(nd.ni_dvp);
+			else
+				vput(nd.ni_dvp);
+			vput(vp);
+		}
+	}
+	if (dirp && v3) {
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	nfsm_reply(NFSX_WCCDATA(v3));
+	if (v3) {
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		return (0);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs rename service
+ */
+int
+nfsrv_rename(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1;
+	int tdirfor_ret = 1, tdiraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mreq, *mb2;
+	struct nameidata fromnd, tond;
+	struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
+	struct vnode *tdirp = (struct vnode *)0;
+	struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
+	nfsfh_t fnfh, tnfh;
+	fhandle_t *ffhp, *tfhp;
+	u_quad_t frev;
+	uid_t saved_uid;
+
+#ifndef nolint
+	fvp = (struct vnode *)0;
+#endif
+	ffhp = &fnfh.fh_generic;
+	tfhp = &tnfh.fh_generic;
+	fromnd.ni_cnd.cn_nameiop = 0;
+	tond.ni_cnd.cn_nameiop = 0;
+	nfsm_srvmtofh(ffhp);
+	nfsm_srvnamesiz(len);
+	/*
+	 * Remember our original uid so that we can reset cr_uid before
+	 * the second nfs_namei() call, in case it is remapped.
+	 */
+	saved_uid = cred->cr_uid;
+	fromnd.ni_cnd.cn_cred = cred;
+	fromnd.ni_cnd.cn_nameiop = DELETE;
+	fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
+	error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
+		&dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (fdirp) {
+		if (v3)
+			fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
+				procp);
+		else {
+			vrele(fdirp);
+			fdirp = (struct vnode *)0;
+		}
+	}
+	if (error) {
+		nfsm_reply(2 * NFSX_WCCDATA(v3));
+		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
+		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
+		if (fdirp)
+			vrele(fdirp);
+		return (0);
+	}
+	fvp = fromnd.ni_vp;
+	nfsm_srvmtofh(tfhp);
+	nfsm_strsiz(len2, NFS_MAXNAMLEN);
+	cred->cr_uid = saved_uid;
+	tond.ni_cnd.cn_cred = cred;
+	tond.ni_cnd.cn_nameiop = RENAME;
+	tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
+	error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
+		&dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (tdirp) {
+		if (v3)
+			tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
+				procp);
+		else {
+			vrele(tdirp);
+			tdirp = (struct vnode *)0;
+		}
+	}
+	if (error) {
+		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
+		vrele(fromnd.ni_dvp);
+		vrele(fvp);
+		goto out1;
+	}
+	tdvp = tond.ni_dvp;
+	tvp = tond.ni_vp;
+	if (tvp != NULL) {
+		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
+			if (v3)
+				error = EEXIST;
+			else
+				error = EISDIR;
+			goto out;
+		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
+			if (v3)
+				error = EEXIST;
+			else
+				error = ENOTDIR;
+			goto out;
+		}
+		if (tvp->v_type == VDIR && tvp->v_mountedhere) {
+			if (v3)
+				error = EXDEV;
+			else
+				error = ENOTEMPTY;
+			goto out;
+		}
+	}
+	if (fvp->v_type == VDIR && fvp->v_mountedhere) {
+		if (v3)
+			error = EXDEV;
+		else
+			error = ENOTEMPTY;
+		goto out;
+	}
+	if (fvp->v_mount != tdvp->v_mount) {
+		if (v3)
+			error = EXDEV;
+		else
+			error = ENOTEMPTY;
+		goto out;
+	}
+	if (fvp == tdvp)
+		if (v3)
+			error = EINVAL;
+		else
+			error = ENOTEMPTY;
+	/*
+	 * If source is the same as the destination (that is the
+	 * same vnode with the same name in the same directory),
+	 * then there is nothing to do.
+	 */
+	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
+	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
+	    !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
+	      fromnd.ni_cnd.cn_namelen))
+		error = -1;
+out:
+	if (!error) {
+		nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
+		nqsrv_getl(tdvp, ND_WRITE);
+		if (tvp)
+			nqsrv_getl(tvp, ND_WRITE);
+		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
+				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
+	} else {
+		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
+		if (tdvp == tvp)
+			vrele(tdvp);
+		else
+			vput(tdvp);
+		if (tvp)
+			vput(tvp);
+		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
+		vrele(fromnd.ni_dvp);
+		vrele(fvp);
+		if (error == -1)
+			error = 0;
+	}
+	vrele(tond.ni_startdir);
+	FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
+out1:
+	if (fdirp) {
+		fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
+		vrele(fdirp);
+	}
+	if (tdirp) {
+		tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
+		vrele(tdirp);
+	}
+	vrele(fromnd.ni_startdir);
+	FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
+	nfsm_reply(2 * NFSX_WCCDATA(v3));
+	if (v3) {
+		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
+		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
+	}
+	return (0);
+
+nfsmout:
+	if (fdirp)
+		vrele(fdirp);
+	if (tdirp)
+		vrele(tdirp);
+	if (tond.ni_cnd.cn_nameiop) {
+		vrele(tond.ni_startdir);
+		FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
+	}
+	if (fromnd.ni_cnd.cn_nameiop) {
+		vrele(fromnd.ni_startdir);
+		FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
+		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
+		vrele(fromnd.ni_dvp);
+		vrele(fvp);
+	}
+	return (error);
+}
+
+/*
+ * nfs link service
+ */
+int
+nfsrv_link(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct nameidata nd;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
+	int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mreq, *mb2;
+	struct vnode *vp, *xp, *dirp = (struct vnode *)0;
+	struct vattr dirfor, diraft, at;
+	nfsfh_t nfh, dnfh;
+	fhandle_t *fhp, *dfhp;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	dfhp = &dnfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvmtofh(dfhp);
+	nfsm_srvnamesiz(len);
+	if (error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
+		nfsm_srvpostop_attr(getret, &at);
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		return (0);
+	}
+	if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)))
+		goto out1;
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = CREATE;
+	nd.ni_cnd.cn_flags = LOCKPARENT;
+	error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else {
+			vrele(dirp);
+			dirp = (struct vnode *)0;
+		}
+	}
+	if (error)
+		goto out1;
+	xp = nd.ni_vp;
+	if (xp != NULL) {
+		error = EEXIST;
+		goto out;
+	}
+	xp = nd.ni_dvp;
+	if (vp->v_mount != xp->v_mount)
+		error = EXDEV;
+out:
+	if (!error) {
+		nqsrv_getl(vp, ND_WRITE);
+		nqsrv_getl(xp, ND_WRITE);
+		error = VOP_LINK(vp, nd.ni_dvp, &nd.ni_cnd);
+	} else {
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		if (nd.ni_dvp == nd.ni_vp)
+			vrele(nd.ni_dvp);
+		else
+			vput(nd.ni_dvp);
+		if (nd.ni_vp)
+			vrele(nd.ni_vp);
+	}
+out1:
+	if (v3)
+		getret = VOP_GETATTR(vp, &at, cred, procp);
+	if (dirp) {
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	vrele(vp);
+	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
+	if (v3) {
+		nfsm_srvpostop_attr(getret, &at);
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		return (0);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs symbolic link service
+ */
+int
+nfsrv_symlink(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct vattr va, dirfor, diraft;
+	struct nameidata nd;
+	register struct vattr *vap = &va;
+	register u_long *tl;
+	register long t1;
+	struct nfsv2_sattr *sp;
+	char *bpos, *cp, *pathcp = (char *)0, *cp2;
+	struct uio io;
+	struct iovec iv;
+	int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	struct mbuf *mb, *mreq, *mb2;
+	struct vnode *dirp = (struct vnode *)0;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+	nd.ni_cnd.cn_nameiop = 0;
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = CREATE;
+	nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else {
+			vrele(dirp);
+			dirp = (struct vnode *)0;
+		}
+	}
+	if (error)
+		goto out;
+	VATTR_NULL(vap);
+	if (v3)
+		nfsm_srvsattr(vap);
+	nfsm_strsiz(len2, NFS_MAXPATHLEN);
+	MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK);
+	iv.iov_base = pathcp;
+	iv.iov_len = len2;
+	io.uio_resid = len2;
+	io.uio_offset = 0;
+	io.uio_iov = &iv;
+	io.uio_iovcnt = 1;
+	io.uio_segflg = UIO_SYSSPACE;
+	io.uio_rw = UIO_READ;
+	io.uio_procp = (struct proc *)0;
+	nfsm_mtouio(&io, len2);
+	if (!v3) {
+		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		vap->va_mode = fxdr_unsigned(u_short, sp->sa_mode);
+	}
+	*(pathcp + len2) = '\0';
+	if (nd.ni_vp) {
+		vrele(nd.ni_startdir);
+		free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		if (nd.ni_dvp == nd.ni_vp)
+			vrele(nd.ni_dvp);
+		else
+			vput(nd.ni_dvp);
+		vrele(nd.ni_vp);
+		error = EEXIST;
+		goto out;
+	}
+	nqsrv_getl(nd.ni_dvp, ND_WRITE);
+	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap, pathcp);
+	if (error)
+		vrele(nd.ni_startdir);
+	else {
+	    if (v3) {
+		nd.ni_cnd.cn_nameiop = LOOKUP;
+		nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW);
+		nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF);
+		nd.ni_cnd.cn_proc = procp;
+		nd.ni_cnd.cn_cred = cred;
+		error = lookup(&nd);
+		if (!error) {
+			bzero((caddr_t)fhp, sizeof(nfh));
+			fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
+			error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
+			if (!error)
+				error = VOP_GETATTR(nd.ni_vp, vap, cred,
+					procp);
+			vput(nd.ni_vp);
+		}
+	    } else
+		vrele(nd.ni_startdir);
+	    FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	}
+out:
+	if (pathcp)
+		FREE(pathcp, M_TEMP);
+	if (dirp) {
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
+	if (v3) {
+		if (!error) {
+			nfsm_srvpostop_fh(fhp);
+			nfsm_srvpostop_attr(0, vap);
+		}
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+	}
+	return (0);
+nfsmout:
+	if (nd.ni_cnd.cn_nameiop) {
+		vrele(nd.ni_startdir);
+		free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
+	}
+	if (dirp)
+		vrele(dirp);
+	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+	if (nd.ni_dvp == nd.ni_vp)
+		vrele(nd.ni_dvp);
+	else
+		vput(nd.ni_dvp);
+	if (nd.ni_vp)
+		vrele(nd.ni_vp);
+	if (pathcp)
+		FREE(pathcp, M_TEMP);
+	return (error);
+}
+
+/*
+ * nfs mkdir service
+ */
+int
+nfsrv_mkdir(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct vattr va, dirfor, diraft;
+	register struct vattr *vap = &va;
+	register struct nfs_fattr *fp;
+	struct nameidata nd;
+	register caddr_t cp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp, *dirp = (struct vnode *)0;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = CREATE;
+	nd.ni_cnd.cn_flags = LOCKPARENT;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else {
+			vrele(dirp);
+			dirp = (struct vnode *)0;
+		}
+	}
+	if (error) {
+		nfsm_reply(NFSX_WCCDATA(v3));
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		if (dirp)
+			vrele(dirp);
+		return (0);
+	}
+	VATTR_NULL(vap);
+	if (v3) {
+		nfsm_srvsattr(vap);
+	} else {
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		vap->va_mode = nfstov_mode(*tl++);
+	}
+	vap->va_type = VDIR;
+	vp = nd.ni_vp;
+	if (vp != NULL) {
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		if (nd.ni_dvp == vp)
+			vrele(nd.ni_dvp);
+		else
+			vput(nd.ni_dvp);
+		vrele(vp);
+		error = EEXIST;
+		goto out;
+	}
+	nqsrv_getl(nd.ni_dvp, ND_WRITE);
+	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap);
+	if (!error) {
+		vp = nd.ni_vp;
+		bzero((caddr_t)fhp, sizeof(nfh));
+		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+		error = VFS_VPTOFH(vp, &fhp->fh_fid);
+		if (!error)
+			error = VOP_GETATTR(vp, vap, cred, procp);
+		vput(vp);
+	}
+out:
+	if (dirp) {
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
+	if (v3) {
+		if (!error) {
+			nfsm_srvpostop_fh(fhp);
+			nfsm_srvpostop_attr(0, vap);
+		}
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+	} else {
+		nfsm_srvfhtom(fhp, v3);
+		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
+		nfsm_srvfillattr(vap, fp);
+	}
+	return (0);
+nfsmout:
+	if (dirp)
+		vrele(dirp);
+	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+	if (nd.ni_dvp == nd.ni_vp)
+		vrele(nd.ni_dvp);
+	else
+		vput(nd.ni_dvp);
+	if (nd.ni_vp)
+		vrele(nd.ni_vp);
+	return (error);
+}
+
+/*
+ * nfs rmdir service
+ */
+int
+nfsrv_rmdir(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mreq, *mb2;
+	struct vnode *vp, *dirp = (struct vnode *)0;
+	struct vattr dirfor, diraft;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct nameidata nd;
+	u_quad_t frev;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_srvnamesiz(len);
+	nd.ni_cnd.cn_cred = cred;
+	nd.ni_cnd.cn_nameiop = DELETE;
+	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
+	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
+		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH));
+	if (dirp) {
+		if (v3)
+			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
+				procp);
+		else {
+			vrele(dirp);
+			dirp = (struct vnode *)0;
+		}
+	}
+	if (error) {
+		nfsm_reply(NFSX_WCCDATA(v3));
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		if (dirp)
+			vrele(dirp);
+		return (0);
+	}
+	vp = nd.ni_vp;
+	if (vp->v_type != VDIR) {
+		error = ENOTDIR;
+		goto out;
+	}
+	/*
+	 * No rmdir "." please.
+	 */
+	if (nd.ni_dvp == vp) {
+		error = EINVAL;
+		goto out;
+	}
+	/*
+	 * The root of a mounted filesystem cannot be deleted.
+	 */
+	if (vp->v_flag & VROOT)
+		error = EBUSY;
+out:
+	if (!error) {
+		nqsrv_getl(nd.ni_dvp, ND_WRITE);
+		nqsrv_getl(vp, ND_WRITE);
+		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
+	} else {
+		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
+		if (nd.ni_dvp == nd.ni_vp)
+			vrele(nd.ni_dvp);
+		else
+			vput(nd.ni_dvp);
+		vput(vp);
+	}
+	if (dirp) {
+		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
+		vrele(dirp);
+	}
+	nfsm_reply(NFSX_WCCDATA(v3));
+	if (v3) {
+		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
+		return (0);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs readdir service
+ * - mallocs what it thinks is enough to read
+ *	count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
+ * - calls VOP_READDIR()
+ * - loops around building the reply
+ *	if the output generated exceeds count break out of loop
+ *	The nfsm_clget macro is used here so that the reply will be packed
+ *	tightly in mbuf clusters.
+ * - it only knows that it has encountered eof when the VOP_READDIR()
+ *	reads nothing
+ * - as such one readdir rpc will return eof false although you are there
+ *	and then the next will return eof
+ * - it trims out records with d_fileno == 0
+ *	this doesn't matter for Unix clients, but they might confuse clients
+ *	for other os'.
+ * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
+ *	than requested, but this may not apply to all filesystems. For
+ *	example, client NFS does not { although it is never remote mounted
+ *	anyhow }
+ *     The alternate call nfsrv_readdirplus() does lookups as well.
+ * PS: The NFS protocol spec. does not clarify what the "count" byte
+ *	argument is a count of.. just name strings and file id's or the
+ *	entire reply rpc or ...
+ *	I tried just file name and id sizes and it confused the Sun client,
+ *	so I am using the full rpc size now. The "paranoia.." comment refers
+ *	to including the status longwords that are not a part of the dir.
+ *	"entry" structures, but are in the rpc.
+ */
+struct flrep {
+	nfsuint64	fl_off;
+	u_long		fl_postopok;
+	u_long		fl_fattr[NFSX_V3FATTR / sizeof (u_long)];
+	u_long		fl_fhok;
+	u_long		fl_fhsize;
+	u_long		fl_nfh[NFSX_V3FH / sizeof (u_long)];
+};
+
+int
+nfsrv_readdir(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register char *bp, *be;
+	register struct mbuf *mp;
+	register struct dirent *dp;
+	register caddr_t cp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	struct mbuf *mb, *mb2, *mreq, *mp2;
+	char *cpos, *cend, *cp2, *rbuf;
+	struct vnode *vp;
+	struct vattr at;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct uio io;
+	struct iovec iv;
+	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
+	int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	u_quad_t frev, off, toff, verf;
+	u_long *cookies = NULL, *cookiep;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (v3) {
+		nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED);
+		fxdr_hyper(tl, &toff);
+		tl += 2;
+		fxdr_hyper(tl, &verf);
+		tl += 2;
+	} else {
+		nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+		toff = fxdr_unsigned(u_quad_t, *tl++);
+	}
+	off = toff;
+	cnt = fxdr_unsigned(int, *tl);
+	siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
+	xfer = NFS_SRVMAXDATA(nfsd);
+	if (siz > xfer)
+		siz = xfer;
+	fullsiz = siz;
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	nqsrv_getl(vp, ND_READ);
+	if (v3) {
+		error = getret = VOP_GETATTR(vp, &at, cred, procp);
+		if (!error && toff && verf != at.va_filerev)
+			error = NFSERR_BAD_COOKIE;
+	}
+	if (!error)
+		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp);
+	if (error) {
+		vput(vp);
+		nfsm_reply(NFSX_POSTOPATTR(v3));
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	VOP_UNLOCK(vp, 0, procp);
+	MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
+again:
+	iv.iov_base = rbuf;
+	iv.iov_len = fullsiz;
+	io.uio_iov = &iv;
+	io.uio_iovcnt = 1;
+	io.uio_offset = (off_t)off;
+	io.uio_resid = fullsiz;
+	io.uio_segflg = UIO_SYSSPACE;
+	io.uio_rw = UIO_READ;
+	io.uio_procp = (struct proc *)0;
+	eofflag = 0;
+	if (cookies) {
+		free((caddr_t)cookies, M_TEMP);
+		cookies = NULL;
+	}
+	error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies);
+	off = (off_t)io.uio_offset;
+	if (!cookies && !error)
+		error = NFSERR_PERM;
+	if (v3) {
+		getret = VOP_GETATTR(vp, &at, cred, procp);
+		if (!error)
+			error = getret;
+	}
+	if (error) {
+		vrele(vp);
+		free((caddr_t)rbuf, M_TEMP);
+		if (cookies)
+			free((caddr_t)cookies, M_TEMP);
+		nfsm_reply(NFSX_POSTOPATTR(v3));
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	if (io.uio_resid) {
+		siz -= io.uio_resid;
+
+		/*
+		 * If nothing read, return eof
+		 * rpc reply
+		 */
+		if (siz == 0) {
+			vrele(vp);
+			nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
+				2 * NFSX_UNSIGNED);
+			if (v3) {
+				nfsm_srvpostop_attr(getret, &at);
+				nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+				txdr_hyper(&at.va_filerev, tl);
+				tl += 2;
+			} else
+				nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = nfs_false;
+			*tl = nfs_true;
+			FREE((caddr_t)rbuf, M_TEMP);
+			FREE((caddr_t)cookies, M_TEMP);
+			return (0);
+		}
+	}
+
+	/*
+	 * Check for degenerate cases of nothing useful read.
+	 * If so go try again
+	 */
+	cpos = rbuf;
+	cend = rbuf + siz;
+	dp = (struct dirent *)cpos;
+	cookiep = cookies;
+	while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) {
+		cpos += dp->d_reclen;
+		dp = (struct dirent *)cpos;
+		cookiep++;
+		ncookies--;
+	}
+	if (cpos >= cend || ncookies == 0) {
+		toff = off;
+		siz = fullsiz;
+		goto again;
+	}
+
+	len = 3 * NFSX_UNSIGNED;	/* paranoia, probably can be 0 */
+	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
+	if (v3) {
+		nfsm_srvpostop_attr(getret, &at);
+		nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+		txdr_hyper(&at.va_filerev, tl);
+	}
+	mp = mp2 = mb;
+	bp = bpos;
+	be = bp + M_TRAILINGSPACE(mp);
+
+	/* Loop through the records and build reply */
+	while (cpos < cend && ncookies > 0) {
+		if (dp->d_fileno != 0) {
+			nlen = dp->d_namlen;
+			rem = nfsm_rndup(nlen)-nlen;
+			len += (4 * NFSX_UNSIGNED + nlen + rem);
+			if (v3)
+				len += 2 * NFSX_UNSIGNED;
+			if (len > cnt) {
+				eofflag = 0;
+				break;
+			}
+			/*
+			 * Build the directory record xdr from
+			 * the dirent entry.
+			 */
+			nfsm_clget;
+			*tl = nfs_true;
+			bp += NFSX_UNSIGNED;
+			if (v3) {
+				nfsm_clget;
+				*tl = 0;
+				bp += NFSX_UNSIGNED;
+			}
+			nfsm_clget;
+			*tl = txdr_unsigned(dp->d_fileno);
+			bp += NFSX_UNSIGNED;
+			nfsm_clget;
+			*tl = txdr_unsigned(nlen);
+			bp += NFSX_UNSIGNED;
+	
+			/* And loop around copying the name */
+			xfer = nlen;
+			cp = dp->d_name;
+			while (xfer > 0) {
+				nfsm_clget;
+				if ((bp+xfer) > be)
+					tsiz = be-bp;
+				else
+					tsiz = xfer;
+				bcopy(cp, bp, tsiz);
+				bp += tsiz;
+				xfer -= tsiz;
+				if (xfer > 0)
+					cp += tsiz;
+			}
+			/* And null pad to a long boundary */
+			for (i = 0; i < rem; i++)
+				*bp++ = '\0';
+			nfsm_clget;
+	
+			/* Finish off the record */
+			if (v3) {
+				*tl = 0;
+				bp += NFSX_UNSIGNED;
+				nfsm_clget;
+			}
+			*tl = txdr_unsigned(*cookiep);
+			bp += NFSX_UNSIGNED;
+		}
+		cpos += dp->d_reclen;
+		dp = (struct dirent *)cpos;
+		cookiep++;
+		ncookies--;
+	}
+	vrele(vp);
+	nfsm_clget;
+	*tl = nfs_false;
+	bp += NFSX_UNSIGNED;
+	nfsm_clget;
+	if (eofflag)
+		*tl = nfs_true;
+	else
+		*tl = nfs_false;
+	bp += NFSX_UNSIGNED;
+	if (mp != mb) {
+		if (bp < be)
+			mp->m_len = bp - mtod(mp, caddr_t);
+	} else
+		mp->m_len += bp - bpos;
+	FREE((caddr_t)rbuf, M_TEMP);
+	FREE((caddr_t)cookies, M_TEMP);
+	nfsm_srvdone;
+}
+
+int
+nfsrv_readdirplus(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register char *bp, *be;
+	register struct mbuf *mp;
+	register struct dirent *dp;
+	register caddr_t cp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	struct mbuf *mb, *mb2, *mreq, *mp2;
+	char *cpos, *cend, *cp2, *rbuf;
+	struct vnode *vp, *nvp;
+	struct flrep fl;
+	nfsfh_t nfh;
+	fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
+	struct uio io;
+	struct iovec iv;
+	struct vattr va, at, *vap = &va;
+	struct nfs_fattr *fp;
+	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
+	int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
+	u_quad_t frev, off, toff, verf;
+	u_long *cookies = NULL, *cookiep;
+
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED);
+	fxdr_hyper(tl, &toff);
+	tl += 2;
+	fxdr_hyper(tl, &verf);
+	tl += 2;
+	siz = fxdr_unsigned(int, *tl++);
+	cnt = fxdr_unsigned(int, *tl);
+	off = toff;
+	siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
+	xfer = NFS_SRVMAXDATA(nfsd);
+	if (siz > xfer)
+		siz = xfer;
+	fullsiz = siz;
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	error = getret = VOP_GETATTR(vp, &at, cred, procp);
+	if (!error && toff && verf != at.va_filerev)
+		error = NFSERR_BAD_COOKIE;
+	if (!error) {
+		nqsrv_getl(vp, ND_READ);
+		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp);
+	}
+	if (error) {
+		vput(vp);
+		nfsm_reply(NFSX_V3POSTOPATTR);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	VOP_UNLOCK(vp, 0, procp);
+	MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
+again:
+	iv.iov_base = rbuf;
+	iv.iov_len = fullsiz;
+	io.uio_iov = &iv;
+	io.uio_iovcnt = 1;
+	io.uio_offset = (off_t)off;
+	io.uio_resid = fullsiz;
+	io.uio_segflg = UIO_SYSSPACE;
+	io.uio_rw = UIO_READ;
+	io.uio_procp = (struct proc *)0;
+	eofflag = 0;
+	if (cookies) {
+		free((caddr_t)cookies, M_TEMP);
+		cookies = NULL;
+	}
+	error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies);
+	off = (u_quad_t)io.uio_offset;
+	getret = VOP_GETATTR(vp, &at, cred, procp);
+	if (!cookies && !error)
+		error = NFSERR_PERM;
+	if (!error)
+		error = getret;
+	if (error) {
+		vrele(vp);
+		if (cookies)
+			free((caddr_t)cookies, M_TEMP);
+		free((caddr_t)rbuf, M_TEMP);
+		nfsm_reply(NFSX_V3POSTOPATTR);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	if (io.uio_resid) {
+		siz -= io.uio_resid;
+
+		/*
+		 * If nothing read, return eof
+		 * rpc reply
+		 */
+		if (siz == 0) {
+			vrele(vp);
+			nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
+				2 * NFSX_UNSIGNED);
+			nfsm_srvpostop_attr(getret, &at);
+			nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+			txdr_hyper(&at.va_filerev, tl);
+			tl += 2;
+			*tl++ = nfs_false;
+			*tl = nfs_true;
+			FREE((caddr_t)cookies, M_TEMP);
+			FREE((caddr_t)rbuf, M_TEMP);
+			return (0);
+		}
+	}
+
+	/*
+	 * Check for degenerate cases of nothing useful read.
+	 * If so go try again
+	 */
+	cpos = rbuf;
+	cend = rbuf + siz;
+	dp = (struct dirent *)cpos;
+	cookiep = cookies;
+	while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) {
+		cpos += dp->d_reclen;
+		dp = (struct dirent *)cpos;
+		cookiep++;
+		ncookies--;
+	}
+	if (cpos >= cend || ncookies == 0) {
+		toff = off;
+		siz = fullsiz;
+		goto again;
+	}
+
+	dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
+	nfsm_reply(cnt);
+	nfsm_srvpostop_attr(getret, &at);
+	nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+	txdr_hyper(&at.va_filerev, tl);
+	mp = mp2 = mb;
+	bp = bpos;
+	be = bp + M_TRAILINGSPACE(mp);
+
+	/* Loop through the records and build reply */
+	while (cpos < cend && ncookies > 0) {
+		if (dp->d_fileno != 0) {
+			nlen = dp->d_namlen;
+			rem = nfsm_rndup(nlen)-nlen;
+	
+			/*
+			 * For readdir_and_lookup get the vnode using
+			 * the file number.
+			 */
+			if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
+				goto invalid;
+			bzero((caddr_t)nfhp, NFSX_V3FH);
+			nfhp->fh_fsid =
+				nvp->v_mount->mnt_stat.f_fsid;
+			if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
+				vput(nvp);
+				goto invalid;
+			}
+			if (VOP_GETATTR(nvp, vap, cred, procp)) {
+				vput(nvp);
+				goto invalid;
+			}
+			vput(nvp);
+
+			/*
+			 * If either the dircount or maxcount will be
+			 * exceeded, get out now. Both of these lengths
+			 * are calculated conservatively, including all
+			 * XDR overheads.
+			 */
+			len += (7 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
+				NFSX_V3POSTOPATTR);
+			dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
+			if (len > cnt || dirlen > fullsiz) {
+				eofflag = 0;
+				break;
+			}
+
+			/*
+			 * Build the directory record xdr from
+			 * the dirent entry.
+			 */
+			fp = (struct nfs_fattr *)&fl.fl_fattr;
+			nfsm_srvfillattr(vap, fp);
+			fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
+			fl.fl_fhok = nfs_true;
+			fl.fl_postopok = nfs_true;
+			fl.fl_off.nfsuquad[0] = 0;
+			fl.fl_off.nfsuquad[1] = txdr_unsigned(*cookiep);
+
+			nfsm_clget;
+			*tl = nfs_true;
+			bp += NFSX_UNSIGNED;
+			nfsm_clget;
+			*tl = 0;
+			bp += NFSX_UNSIGNED;
+			nfsm_clget;
+			*tl = txdr_unsigned(dp->d_fileno);
+			bp += NFSX_UNSIGNED;
+			nfsm_clget;
+			*tl = txdr_unsigned(nlen);
+			bp += NFSX_UNSIGNED;
+	
+			/* And loop around copying the name */
+			xfer = nlen;
+			cp = dp->d_name;
+			while (xfer > 0) {
+				nfsm_clget;
+				if ((bp + xfer) > be)
+					tsiz = be - bp;
+				else
+					tsiz = xfer;
+				bcopy(cp, bp, tsiz);
+				bp += tsiz;
+				xfer -= tsiz;
+				if (xfer > 0)
+					cp += tsiz;
+			}
+			/* And null pad to a long boundary */
+			for (i = 0; i < rem; i++)
+				*bp++ = '\0';
+	
+			/*
+			 * Now copy the flrep structure out.
+			 */
+			xfer = sizeof (struct flrep);
+			cp = (caddr_t)&fl;
+			while (xfer > 0) {
+				nfsm_clget;
+				if ((bp + xfer) > be)
+					tsiz = be - bp;
+				else
+					tsiz = xfer;
+				bcopy(cp, bp, tsiz);
+				bp += tsiz;
+				xfer -= tsiz;
+				if (xfer > 0)
+					cp += tsiz;
+			}
+		}
+invalid:
+		cpos += dp->d_reclen;
+		dp = (struct dirent *)cpos;
+		cookiep++;
+		ncookies--;
+	}
+	vrele(vp);
+	nfsm_clget;
+	*tl = nfs_false;
+	bp += NFSX_UNSIGNED;
+	nfsm_clget;
+	if (eofflag)
+		*tl = nfs_true;
+	else
+		*tl = nfs_false;
+	bp += NFSX_UNSIGNED;
+	if (mp != mb) {
+		if (bp < be)
+			mp->m_len = bp - mtod(mp, caddr_t);
+	} else
+		mp->m_len += bp - bpos;
+	FREE((caddr_t)cookies, M_TEMP);
+	FREE((caddr_t)rbuf, M_TEMP);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs commit service
+ */
+int
+nfsrv_commit(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	struct vattr bfor, aft;
+	struct vnode *vp;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	u_quad_t frev, off;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+
+	/*
+	 * XXX At this time VOP_FSYNC() does not accept offset and byte
+	 * count parameters, so these arguments are useless (someday maybe).
+	 */
+	fxdr_hyper(tl, &off);
+	tl += 2;
+	cnt = fxdr_unsigned(int, *tl);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(2 * NFSX_UNSIGNED);
+		nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
+		return (0);
+	}
+	for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
+	error = VOP_FSYNC(vp, cred, MNT_WAIT, procp);
+	aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
+	nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
+	if (!error) {
+		nfsm_build(tl, u_long *, NFSX_V3WRITEVERF);
+		*tl++ = txdr_unsigned(boottime.tv_sec);
+		*tl = txdr_unsigned(boottime.tv_usec);
+	} else
+		return (0);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs statfs service
+ */
+int
+nfsrv_statfs(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register struct statfs *sf;
+	register struct nfs_statfs *sfp;
+	register u_long *tl;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, getret = 1;
+	int v3 = (nfsd->nd_flag & ND_NFSV3);
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp;
+	struct vattr at;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	struct statfs statfs;
+	u_quad_t frev, tval;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	sf = &statfs;
+	error = VFS_STATFS(vp->v_mount, sf, procp);
+	getret = VOP_GETATTR(vp, &at, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
+	if (v3)
+		nfsm_srvpostop_attr(getret, &at);
+	if (error)
+		return (0);
+	nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
+	if (v3) {
+		tval = (u_quad_t)sf->f_blocks;
+		tval *= (u_quad_t)sf->f_bsize;
+		txdr_hyper(&tval, &sfp->sf_tbytes);
+		tval = (u_quad_t)sf->f_bfree;
+		tval *= (u_quad_t)sf->f_bsize;
+		txdr_hyper(&tval, &sfp->sf_fbytes);
+		tval = (u_quad_t)sf->f_bavail;
+		tval *= (u_quad_t)sf->f_bsize;
+		txdr_hyper(&tval, &sfp->sf_abytes);
+		sfp->sf_tfiles.nfsuquad[0] = 0;
+		sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files);
+		sfp->sf_ffiles.nfsuquad[0] = 0;
+		sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
+		sfp->sf_afiles.nfsuquad[0] = 0;
+		sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
+		sfp->sf_invarsec = 0;
+	} else {
+		sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
+		sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
+		sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
+		sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
+		sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
+	}
+	nfsm_srvdone;
+}
+
+/*
+ * nfs fsinfo service
+ */
+int
+nfsrv_fsinfo(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register u_long *tl;
+	register struct nfsv3_fsinfo *sip;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, getret = 1, pref;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp;
+	struct vattr at;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	getret = VOP_GETATTR(vp, &at, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
+	nfsm_srvpostop_attr(getret, &at);
+	nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
+
+	/*
+	 * XXX
+	 * There should be file system VFS OP(s) to get this information.
+	 * For now, assume ufs.
+	 */
+	if (slp->ns_so->so_type == SOCK_DGRAM)
+		pref = NFS_MAXDGRAMDATA;
+	else
+		pref = NFS_MAXDATA;
+	sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
+	sip->fs_rtpref = txdr_unsigned(pref);
+	sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
+	sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
+	sip->fs_wtpref = txdr_unsigned(pref);
+	sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
+	sip->fs_dtpref = txdr_unsigned(pref);
+	sip->fs_maxfilesize.nfsuquad[0] = 0xffffffff;
+	sip->fs_maxfilesize.nfsuquad[1] = 0xffffffff;
+	sip->fs_timedelta.nfsv3_sec = 0;
+	sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
+	sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
+		NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
+		NFSV3FSINFO_CANSETTIME);
+	nfsm_srvdone;
+}
+
+/*
+ * nfs pathconf service
+ */
+int
+nfsrv_pathconf(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	register u_long *tl;
+	register struct nfsv3_pathconf *pc;
+	register long t1;
+	caddr_t bpos;
+	int error = 0, rdonly, cache, getret = 1, linkmax, namemax;
+	int chownres, notrunc;
+	char *cp2;
+	struct mbuf *mb, *mb2, *mreq;
+	struct vnode *vp;
+	struct vattr at;
+	nfsfh_t nfh;
+	fhandle_t *fhp;
+	u_quad_t frev;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	fhp = &nfh.fh_generic;
+	nfsm_srvmtofh(fhp);
+	if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
+		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) {
+		nfsm_reply(NFSX_UNSIGNED);
+		nfsm_srvpostop_attr(getret, &at);
+		return (0);
+	}
+	error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
+	if (!error)
+		error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
+	if (!error)
+		error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
+	if (!error)
+		error = VOP_PATHCONF(vp, _PC_NO_TRUNC, &notrunc);
+	getret = VOP_GETATTR(vp, &at, cred, procp);
+	vput(vp);
+	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
+	nfsm_srvpostop_attr(getret, &at);
+	if (error)
+		return (0);
+	nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
+
+	pc->pc_linkmax = txdr_unsigned(linkmax);
+	pc->pc_namemax = txdr_unsigned(namemax);
+	pc->pc_notrunc = txdr_unsigned(notrunc);
+	pc->pc_chownrestricted = txdr_unsigned(chownres);
+
+	/*
+	 * These should probably be supported by VOP_PATHCONF(), but
+	 * until msdosfs is exportable (why would you want to?), the
+	 * Unix defaults should be ok.
+	 */
+	pc->pc_caseinsensitive = nfs_false;
+	pc->pc_casepreserving = nfs_true;
+	nfsm_srvdone;
+}
+
+/*
+ * Null operation, used by clients to ping server
+ */
+/* ARGSUSED */
+int
+nfsrv_null(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	caddr_t bpos;
+	int error = NFSERR_RETVOID, cache;
+	struct mbuf *mb, *mreq;
+	u_quad_t frev;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	nfsm_reply(0);
+	return (0);
+}
+
+/*
+ * No operation, used for obsolete procedures
+ */
+/* ARGSUSED */
+int
+nfsrv_noop(nfsd, slp, procp, mrq)
+	struct nfsrv_descript *nfsd;
+	struct nfssvc_sock *slp;
+	struct proc *procp;
+	struct mbuf **mrq;
+{
+	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
+	struct mbuf *nam = nfsd->nd_nam;
+	caddr_t dpos = nfsd->nd_dpos;
+	struct ucred *cred = &nfsd->nd_cr;
+	caddr_t bpos;
+	int error, cache;
+	struct mbuf *mb, *mreq;
+	u_quad_t frev;
+
+#ifndef nolint
+	cache = 0;
+#endif
+	if (nfsd->nd_repstat)
+		error = nfsd->nd_repstat;
+	else
+		error = EPROCUNAVAIL;
+	nfsm_reply(0);
+	return (0);
+}
+
+/*
+ * Perform access checking for vnodes obtained from file handles that would
+ * refer to files already opened by a Unix client. You cannot just use
+ * vn_writechk() and VOP_ACCESS() for two reasons.
+ * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
+ * 2 - The owner is to be given access irrespective of mode bits so that
+ *     processes that chmod after opening a file don't break. I don't like
+ *     this because it opens a security hole, but since the nfs server opens
+ *     a security hole the size of a barn door anyhow, what the heck.
+ */
+int
+nfsrv_access(vp, flags, cred, rdonly, p)
+	register struct vnode *vp;
+	int flags;
+	register struct ucred *cred;
+	int rdonly;
+	struct proc *p;
+{
+	struct vattr vattr;
+	int error;
+	if (flags & VWRITE) {
+		/* Just vn_writechk() changed to check rdonly */
+		/*
+		 * Disallow write attempts on read-only file systems;
+		 * unless the file is a socket or a block or character
+		 * device resident on the file system.
+		 */
+		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
+			switch (vp->v_type) {
+			case VREG: case VDIR: case VLNK:
+				return (EROFS);
+			}
+		}
+		/*
+		 * If there's shared text associated with
+		 * the inode, try to free it up once.  If
+		 * we fail, we can't allow writing.
+		 */
+		if ((vp->v_flag & VTEXT) && !vnode_pager_uncache(vp))
+			return (ETXTBSY);
+	}
+	if (error = VOP_GETATTR(vp, &vattr, cred, p))
+		return (error);
+	if ((error = VOP_ACCESS(vp, flags, cred, p)) &&
+	    cred->cr_uid != vattr.va_uid)
+		return (error);
+	return (0);
+}
diff --git a/sys/nfs/nfs_socket.c b/sys/nfs/nfs_socket.c
new file mode 100644
index 000000000000..4326da07d534
--- /dev/null
+++ b/sys/nfs/nfs_socket.c
@@ -0,0 +1,2182 @@
+/*
+ * Copyright (c) 1989, 1991, 1993, 1995
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_socket.c	8.5 (Berkeley) 3/30/95
+ */
+
+/*
+ * Socket operations for use by nfs
+ */
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/proc.h>
+#include <sys/mount.h>
+#include <sys/kernel.h>
+#include <sys/mbuf.h>
+#include <sys/vnode.h>
+#include <sys/domain.h>
+#include <sys/protosw.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+#include <sys/syslog.h>
+#include <sys/tprintf.h>
+
+#include <netinet/in.h>
+#include <netinet/tcp.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nfsmount.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfsrtt.h>
+#include <nfs/nqnfs.h>
+
+#define	TRUE	1
+#define	FALSE	0
+
+/*
+ * Estimate rto for an nfs rpc sent via. an unreliable datagram.
+ * Use the mean and mean deviation of rtt for the appropriate type of rpc
+ * for the frequent rpcs and a default for the others.
+ * The justification for doing "other" this way is that these rpcs
+ * happen so infrequently that timer est. would probably be stale.
+ * Also, since many of these rpcs are
+ * non-idempotent, a conservative timeout is desired.
+ * getattr, lookup - A+2D
+ * read, write     - A+4D
+ * other           - nm_timeo
+ */
+#define	NFS_RTO(n, t) \
+	((t) == 0 ? (n)->nm_timeo : \
+	 ((t) < 3 ? \
+	  (((((n)->nm_srtt[t-1] + 3) >> 2) + (n)->nm_sdrtt[t-1] + 1) >> 1) : \
+	  ((((n)->nm_srtt[t-1] + 7) >> 3) + (n)->nm_sdrtt[t-1] + 1)))
+#define	NFS_SRTT(r)	(r)->r_nmp->nm_srtt[proct[(r)->r_procnum] - 1]
+#define	NFS_SDRTT(r)	(r)->r_nmp->nm_sdrtt[proct[(r)->r_procnum] - 1]
+/*
+ * External data, mostly RPC constants in XDR form
+ */
+extern u_long rpc_reply, rpc_msgdenied, rpc_mismatch, rpc_vers, rpc_auth_unix,
+	rpc_msgaccepted, rpc_call, rpc_autherr,
+	rpc_auth_kerb;
+extern u_long nfs_prog, nqnfs_prog;
+extern time_t nqnfsstarttime;
+extern struct nfsstats nfsstats;
+extern int nfsv3_procid[NFS_NPROCS];
+extern int nfs_ticks;
+
+/*
+ * Defines which timer to use for the procnum.
+ * 0 - default
+ * 1 - getattr
+ * 2 - lookup
+ * 3 - read
+ * 4 - write
+ */
+static int proct[NFS_NPROCS] = {
+	0, 1, 0, 2, 1, 3, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 3, 3, 0, 0, 0, 0, 0,
+	0, 0, 0,
+};
+
+/*
+ * There is a congestion window for outstanding rpcs maintained per mount
+ * point. The cwnd size is adjusted in roughly the way that:
+ * Van Jacobson, Congestion avoidance and Control, In "Proceedings of
+ * SIGCOMM '88". ACM, August 1988.
+ * describes for TCP. The cwnd size is chopped in half on a retransmit timeout
+ * and incremented by 1/cwnd when each rpc reply is received and a full cwnd
+ * of rpcs is in progress.
+ * (The sent count and cwnd are scaled for integer arith.)
+ * Variants of "slow start" were tried and were found to be too much of a
+ * performance hit (ave. rtt 3 times larger),
+ * I suspect due to the large rtt that nfs rpcs have.
+ */
+#define	NFS_CWNDSCALE	256
+#define	NFS_MAXCWND	(NFS_CWNDSCALE * 32)
+static int nfs_backoff[8] = { 2, 4, 8, 16, 32, 64, 128, 256, };
+int	nfs_sbwait();
+void	nfs_disconnect(), nfs_realign(), nfsrv_wakenfsd(), nfs_sndunlock();
+void	nfs_rcvunlock(), nqnfs_serverd(), nqnfs_clientlease();
+struct mbuf *nfsm_rpchead();
+int nfsrtton = 0;
+struct nfsrtt nfsrtt;
+
+int	nfsrv_null(),
+	nfsrv_getattr(),
+	nfsrv_setattr(),
+	nfsrv_lookup(),
+	nfsrv3_access(),
+	nfsrv_readlink(),
+	nfsrv_read(),
+	nfsrv_write(),
+	nfsrv_create(),
+	nfsrv_mknod(),
+	nfsrv_remove(),
+	nfsrv_rename(),
+	nfsrv_link(),
+	nfsrv_symlink(),
+	nfsrv_mkdir(),
+	nfsrv_rmdir(),
+	nfsrv_readdir(),
+	nfsrv_readdirplus(),
+	nfsrv_statfs(),
+	nfsrv_fsinfo(),
+	nfsrv_pathconf(),
+	nfsrv_commit(),
+	nfsrv_noop(),
+	nqnfsrv_getlease(),
+	nqnfsrv_vacated();
+
+int (*nfsrv3_procs[NFS_NPROCS])() = {
+	nfsrv_null,
+	nfsrv_getattr,
+	nfsrv_setattr,
+	nfsrv_lookup,
+	nfsrv3_access,
+	nfsrv_readlink,
+	nfsrv_read,
+	nfsrv_write,
+	nfsrv_create,
+	nfsrv_mkdir,
+	nfsrv_symlink,
+	nfsrv_mknod,
+	nfsrv_remove,
+	nfsrv_rmdir,
+	nfsrv_rename,
+	nfsrv_link,
+	nfsrv_readdir,
+	nfsrv_readdirplus,
+	nfsrv_statfs,
+	nfsrv_fsinfo,
+	nfsrv_pathconf,
+	nfsrv_commit,
+	nqnfsrv_getlease,
+	nqnfsrv_vacated,
+	nfsrv_noop,
+	nfsrv_noop
+};
+
+/*
+ * Initialize sockets and congestion for a new NFS connection.
+ * We do not free the sockaddr if error.
+ */
+int
+nfs_connect(nmp, rep)
+	register struct nfsmount *nmp;
+	struct nfsreq *rep;
+{
+	register struct socket *so;
+	int s, error, rcvreserve, sndreserve;
+	struct sockaddr *saddr;
+	struct sockaddr_in *sin;
+	struct mbuf *m;
+	u_short tport;
+
+	nmp->nm_so = (struct socket *)0;
+	saddr = mtod(nmp->nm_nam, struct sockaddr *);
+	error = socreate(saddr->sa_family, &nmp->nm_so, nmp->nm_sotype, 
+		nmp->nm_soproto);
+	if (error)
+		goto bad;
+	so = nmp->nm_so;
+	nmp->nm_soflags = so->so_proto->pr_flags;
+
+	/*
+	 * Some servers require that the client port be a reserved port number.
+	 */
+	if (saddr->sa_family == AF_INET && (nmp->nm_flag & NFSMNT_RESVPORT)) {
+		MGET(m, M_WAIT, MT_SONAME);
+		sin = mtod(m, struct sockaddr_in *);
+		sin->sin_len = m->m_len = sizeof (struct sockaddr_in);
+		sin->sin_family = AF_INET;
+		sin->sin_addr.s_addr = INADDR_ANY;
+		tport = IPPORT_RESERVED - 1;
+		sin->sin_port = htons(tport);
+		while ((error = sobind(so, m)) == EADDRINUSE &&
+		       --tport > IPPORT_RESERVED / 2)
+			sin->sin_port = htons(tport);
+		m_freem(m);
+		if (error)
+			goto bad;
+	}
+
+	/*
+	 * Protocols that do not require connections may be optionally left
+	 * unconnected for servers that reply from a port other than NFS_PORT.
+	 */
+	if (nmp->nm_flag & NFSMNT_NOCONN) {
+		if (nmp->nm_soflags & PR_CONNREQUIRED) {
+			error = ENOTCONN;
+			goto bad;
+		}
+	} else {
+		error = soconnect(so, nmp->nm_nam);
+		if (error)
+			goto bad;
+
+		/*
+		 * Wait for the connection to complete. Cribbed from the
+		 * connect system call but with the wait timing out so
+		 * that interruptible mounts don't hang here for a long time.
+		 */
+		s = splnet();
+		while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
+			(void) tsleep((caddr_t)&so->so_timeo, PSOCK,
+				"nfscon", 2 * hz);
+			if ((so->so_state & SS_ISCONNECTING) &&
+			    so->so_error == 0 && rep &&
+			    (error = nfs_sigintr(nmp, rep, rep->r_procp))) {
+				so->so_state &= ~SS_ISCONNECTING;
+				splx(s);
+				goto bad;
+			}
+		}
+		if (so->so_error) {
+			error = so->so_error;
+			so->so_error = 0;
+			splx(s);
+			goto bad;
+		}
+		splx(s);
+	}
+	if (nmp->nm_flag & (NFSMNT_SOFT | NFSMNT_INT)) {
+		so->so_rcv.sb_timeo = (5 * hz);
+		so->so_snd.sb_timeo = (5 * hz);
+	} else {
+		so->so_rcv.sb_timeo = 0;
+		so->so_snd.sb_timeo = 0;
+	}
+	if (nmp->nm_sotype == SOCK_DGRAM) {
+		sndreserve = nmp->nm_wsize + NFS_MAXPKTHDR;
+		rcvreserve = nmp->nm_rsize + NFS_MAXPKTHDR;
+	} else if (nmp->nm_sotype == SOCK_SEQPACKET) {
+		sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR) * 2;
+		rcvreserve = (nmp->nm_rsize + NFS_MAXPKTHDR) * 2;
+	} else {
+		if (nmp->nm_sotype != SOCK_STREAM)
+			panic("nfscon sotype");
+		if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
+			MGET(m, M_WAIT, MT_SOOPTS);
+			*mtod(m, int *) = 1;
+			m->m_len = sizeof(int);
+			sosetopt(so, SOL_SOCKET, SO_KEEPALIVE, m);
+		}
+		if (so->so_proto->pr_protocol == IPPROTO_TCP) {
+			MGET(m, M_WAIT, MT_SOOPTS);
+			*mtod(m, int *) = 1;
+			m->m_len = sizeof(int);
+			sosetopt(so, IPPROTO_TCP, TCP_NODELAY, m);
+		}
+		sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR + sizeof (u_long))
+				* 2;
+		rcvreserve = (nmp->nm_rsize + NFS_MAXPKTHDR + sizeof (u_long))
+				* 2;
+	}
+	error = soreserve(so, sndreserve, rcvreserve);
+	if (error)
+		goto bad;
+	so->so_rcv.sb_flags |= SB_NOINTR;
+	so->so_snd.sb_flags |= SB_NOINTR;
+
+	/* Initialize other non-zero congestion variables */
+	nmp->nm_srtt[0] = nmp->nm_srtt[1] = nmp->nm_srtt[2] = nmp->nm_srtt[3] =
+		nmp->nm_srtt[4] = (NFS_TIMEO << 3);
+	nmp->nm_sdrtt[0] = nmp->nm_sdrtt[1] = nmp->nm_sdrtt[2] =
+		nmp->nm_sdrtt[3] = nmp->nm_sdrtt[4] = 0;
+	nmp->nm_cwnd = NFS_MAXCWND / 2;	    /* Initial send window */
+	nmp->nm_sent = 0;
+	nmp->nm_timeouts = 0;
+	return (0);
+
+bad:
+	nfs_disconnect(nmp);
+	return (error);
+}
+
+/*
+ * Reconnect routine:
+ * Called when a connection is broken on a reliable protocol.
+ * - clean up the old socket
+ * - nfs_connect() again
+ * - set R_MUSTRESEND for all outstanding requests on mount point
+ * If this fails the mount point is DEAD!
+ * nb: Must be called with the nfs_sndlock() set on the mount point.
+ */
+int
+nfs_reconnect(rep)
+	register struct nfsreq *rep;
+{
+	register struct nfsreq *rp;
+	register struct nfsmount *nmp = rep->r_nmp;
+	int error;
+
+	nfs_disconnect(nmp);
+	while ((error = nfs_connect(nmp, rep))) {
+		if (error == EINTR || error == ERESTART)
+			return (EINTR);
+		(void) tsleep((caddr_t)&lbolt, PSOCK, "nfscon", 0);
+	}
+
+	/*
+	 * Loop through outstanding request list and fix up all requests
+	 * on old socket.
+	 */
+	for (rp = nfs_reqq.tqh_first; rp != 0; rp = rp->r_chain.tqe_next) {
+		if (rp->r_nmp == nmp)
+			rp->r_flags |= R_MUSTRESEND;
+	}
+	return (0);
+}
+
+/*
+ * NFS disconnect. Clean up and unlink.
+ */
+void
+nfs_disconnect(nmp)
+	register struct nfsmount *nmp;
+{
+	register struct socket *so;
+
+	if (nmp->nm_so) {
+		so = nmp->nm_so;
+		nmp->nm_so = (struct socket *)0;
+		soshutdown(so, 2);
+		soclose(so);
+	}
+}
+
+/*
+ * This is the nfs send routine. For connection based socket types, it
+ * must be called with an nfs_sndlock() on the socket.
+ * "rep == NULL" indicates that it has been called from a server.
+ * For the client side:
+ * - return EINTR if the RPC is terminated, 0 otherwise
+ * - set R_MUSTRESEND if the send fails for any reason
+ * - do any cleanup required by recoverable socket errors (???)
+ * For the server side:
+ * - return EINTR or ERESTART if interrupted by a signal
+ * - return EPIPE if a connection is lost for connection based sockets (TCP...)
+ * - do any cleanup required by recoverable socket errors (???)
+ */
+int
+nfs_send(so, nam, top, rep)
+	register struct socket *so;
+	struct mbuf *nam;
+	register struct mbuf *top;
+	struct nfsreq *rep;
+{
+	struct mbuf *sendnam;
+	int error, soflags, flags;
+
+	if (rep) {
+		if (rep->r_flags & R_SOFTTERM) {
+			m_freem(top);
+			return (EINTR);
+		}
+		if ((so = rep->r_nmp->nm_so) == NULL) {
+			rep->r_flags |= R_MUSTRESEND;
+			m_freem(top);
+			return (0);
+		}
+		rep->r_flags &= ~R_MUSTRESEND;
+		soflags = rep->r_nmp->nm_soflags;
+	} else
+		soflags = so->so_proto->pr_flags;
+	if ((soflags & PR_CONNREQUIRED) || (so->so_state & SS_ISCONNECTED))
+		sendnam = (struct mbuf *)0;
+	else
+		sendnam = nam;
+	if (so->so_type == SOCK_SEQPACKET)
+		flags = MSG_EOR;
+	else
+		flags = 0;
+
+	error = sosend(so, sendnam, (struct uio *)0, top,
+		(struct mbuf *)0, flags);
+	if (error) {
+		if (rep) {
+			log(LOG_INFO, "nfs send error %d for server %s\n",error,
+			    rep->r_nmp->nm_mountp->mnt_stat.f_mntfromname);
+			/*
+			 * Deal with errors for the client side.
+			 */
+			if (rep->r_flags & R_SOFTTERM)
+				error = EINTR;
+			else
+				rep->r_flags |= R_MUSTRESEND;
+		} else
+			log(LOG_INFO, "nfsd send error %d\n", error);
+
+		/*
+		 * Handle any recoverable (soft) socket errors here. (???)
+		 */
+		if (error != EINTR && error != ERESTART &&
+			error != EWOULDBLOCK && error != EPIPE)
+			error = 0;
+	}
+	return (error);
+}
+
+/*
+ * Receive a Sun RPC Request/Reply. For SOCK_DGRAM, the work is all
+ * done by soreceive(), but for SOCK_STREAM we must deal with the Record
+ * Mark and consolidate the data into a new mbuf list.
+ * nb: Sometimes TCP passes the data up to soreceive() in long lists of
+ *     small mbufs.
+ * For SOCK_STREAM we must be very careful to read an entire record once
+ * we have read any of it, even if the system call has been interrupted.
+ */
+int
+nfs_receive(rep, aname, mp)
+	register struct nfsreq *rep;
+	struct mbuf **aname;
+	struct mbuf **mp;
+{
+	register struct socket *so;
+	struct uio auio;
+	struct iovec aio;
+	register struct mbuf *m;
+	struct mbuf *control;
+	u_long len;
+	struct mbuf **getnam;
+	int error, sotype, rcvflg;
+	struct proc *p = curproc;	/* XXX */
+
+	/*
+	 * Set up arguments for soreceive()
+	 */
+	*mp = (struct mbuf *)0;
+	*aname = (struct mbuf *)0;
+	sotype = rep->r_nmp->nm_sotype;
+
+	/*
+	 * For reliable protocols, lock against other senders/receivers
+	 * in case a reconnect is necessary.
+	 * For SOCK_STREAM, first get the Record Mark to find out how much
+	 * more there is to get.
+	 * We must lock the socket against other receivers
+	 * until we have an entire rpc request/reply.
+	 */
+	if (sotype != SOCK_DGRAM) {
+		error = nfs_sndlock(&rep->r_nmp->nm_flag, rep);
+		if (error)
+			return (error);
+tryagain:
+		/*
+		 * Check for fatal errors and resending request.
+		 */
+		/*
+		 * Ugh: If a reconnect attempt just happened, nm_so
+		 * would have changed. NULL indicates a failed
+		 * attempt that has essentially shut down this
+		 * mount point.
+		 */
+		if (rep->r_mrep || (rep->r_flags & R_SOFTTERM)) {
+			nfs_sndunlock(&rep->r_nmp->nm_flag);
+			return (EINTR);
+		}
+		so = rep->r_nmp->nm_so;
+		if (!so) {
+			error = nfs_reconnect(rep); 
+			if (error) {
+				nfs_sndunlock(&rep->r_nmp->nm_flag);
+				return (error);
+			}
+			goto tryagain;
+		}
+		while (rep->r_flags & R_MUSTRESEND) {
+			m = m_copym(rep->r_mreq, 0, M_COPYALL, M_WAIT);
+			nfsstats.rpcretries++;
+			error = nfs_send(so, rep->r_nmp->nm_nam, m, rep);
+			if (error) {
+				if (error == EINTR || error == ERESTART ||
+				    (error = nfs_reconnect(rep))) {
+					nfs_sndunlock(&rep->r_nmp->nm_flag);
+					return (error);
+				}
+				goto tryagain;
+			}
+		}
+		nfs_sndunlock(&rep->r_nmp->nm_flag);
+		if (sotype == SOCK_STREAM) {
+			aio.iov_base = (caddr_t) &len;
+			aio.iov_len = sizeof(u_long);
+			auio.uio_iov = &aio;
+			auio.uio_iovcnt = 1;
+			auio.uio_segflg = UIO_SYSSPACE;
+			auio.uio_rw = UIO_READ;
+			auio.uio_offset = 0;
+			auio.uio_resid = sizeof(u_long);
+			auio.uio_procp = p;
+			do {
+			   rcvflg = MSG_WAITALL;
+			   error = soreceive(so, (struct mbuf **)0, &auio,
+				(struct mbuf **)0, (struct mbuf **)0, &rcvflg);
+			   if (error == EWOULDBLOCK && rep) {
+				if (rep->r_flags & R_SOFTTERM)
+					return (EINTR);
+			   }
+			} while (error == EWOULDBLOCK);
+			if (!error && auio.uio_resid > 0) {
+			    log(LOG_INFO,
+				 "short receive (%d/%d) from nfs server %s\n",
+				 sizeof(u_long) - auio.uio_resid,
+				 sizeof(u_long),
+				 rep->r_nmp->nm_mountp->mnt_stat.f_mntfromname);
+			    error = EPIPE;
+			}
+			if (error)
+				goto errout;
+			len = ntohl(len) & ~0x80000000;
+			/*
+			 * This is SERIOUS! We are out of sync with the sender
+			 * and forcing a disconnect/reconnect is all I can do.
+			 */
+			if (len > NFS_MAXPACKET) {
+			    log(LOG_ERR, "%s (%d) from nfs server %s\n",
+				"impossible packet length",
+				len,
+				rep->r_nmp->nm_mountp->mnt_stat.f_mntfromname);
+			    error = EFBIG;
+			    goto errout;
+			}
+			auio.uio_resid = len;
+			do {
+			    rcvflg = MSG_WAITALL;
+			    error =  soreceive(so, (struct mbuf **)0,
+				&auio, mp, (struct mbuf **)0, &rcvflg);
+			} while (error == EWOULDBLOCK || error == EINTR ||
+				 error == ERESTART);
+			if (!error && auio.uio_resid > 0) {
+			    log(LOG_INFO,
+				"short receive (%d/%d) from nfs server %s\n",
+				len - auio.uio_resid, len,
+				rep->r_nmp->nm_mountp->mnt_stat.f_mntfromname);
+			    error = EPIPE;
+			}
+		} else {
+			/*
+			 * NB: Since uio_resid is big, MSG_WAITALL is ignored
+			 * and soreceive() will return when it has either a
+			 * control msg or a data msg.
+			 * We have no use for control msg., but must grab them
+			 * and then throw them away so we know what is going
+			 * on.
+			 */
+			auio.uio_resid = len = 100000000; /* Anything Big */
+			auio.uio_procp = p;
+			do {
+			    rcvflg = 0;
+			    error =  soreceive(so, (struct mbuf **)0,
+				&auio, mp, &control, &rcvflg);
+			    if (control)
+				m_freem(control);
+			    if (error == EWOULDBLOCK && rep) {
+				if (rep->r_flags & R_SOFTTERM)
+					return (EINTR);
+			    }
+			} while (error == EWOULDBLOCK ||
+				 (!error && *mp == NULL && control));
+			if ((rcvflg & MSG_EOR) == 0)
+				printf("Egad!!\n");
+			if (!error && *mp == NULL)
+				error = EPIPE;
+			len -= auio.uio_resid;
+		}
+errout:
+		if (error && error != EINTR && error != ERESTART) {
+			m_freem(*mp);
+			*mp = (struct mbuf *)0;
+			if (error != EPIPE)
+				log(LOG_INFO,
+				    "receive error %d from nfs server %s\n",
+				    error,
+				 rep->r_nmp->nm_mountp->mnt_stat.f_mntfromname);
+			error = nfs_sndlock(&rep->r_nmp->nm_flag, rep);
+			if (!error)
+				error = nfs_reconnect(rep);
+			if (!error)
+				goto tryagain;
+		}
+	} else {
+		if ((so = rep->r_nmp->nm_so) == NULL)
+			return (EACCES);
+		if (so->so_state & SS_ISCONNECTED)
+			getnam = (struct mbuf **)0;
+		else
+			getnam = aname;
+		auio.uio_resid = len = 1000000;
+		auio.uio_procp = p;
+		do {
+			rcvflg = 0;
+			error =  soreceive(so, getnam, &auio, mp,
+				(struct mbuf **)0, &rcvflg);
+			if (error == EWOULDBLOCK &&
+			    (rep->r_flags & R_SOFTTERM))
+				return (EINTR);
+		} while (error == EWOULDBLOCK);
+		len -= auio.uio_resid;
+	}
+	if (error) {
+		m_freem(*mp);
+		*mp = (struct mbuf *)0;
+	}
+	/*
+	 * Search for any mbufs that are not a multiple of 4 bytes long
+	 * or with m_data not longword aligned.
+	 * These could cause pointer alignment problems, so copy them to
+	 * well aligned mbufs.
+	 */
+	nfs_realign(*mp, 5 * NFSX_UNSIGNED);
+	return (error);
+}
+
+/*
+ * Implement receipt of reply on a socket.
+ * We must search through the list of received datagrams matching them
+ * with outstanding requests using the xid, until ours is found.
+ */
+/* ARGSUSED */
+int
+nfs_reply(myrep)
+	struct nfsreq *myrep;
+{
+	register struct nfsreq *rep;
+	register struct nfsmount *nmp = myrep->r_nmp;
+	register long t1;
+	struct mbuf *mrep, *nam, *md;
+	u_long rxid, *tl;
+	caddr_t dpos, cp2;
+	int error;
+
+	/*
+	 * Loop around until we get our own reply
+	 */
+	for (;;) {
+		/*
+		 * Lock against other receivers so that I don't get stuck in
+		 * sbwait() after someone else has received my reply for me.
+		 * Also necessary for connection based protocols to avoid
+		 * race conditions during a reconnect.
+		 */
+		error = nfs_rcvlock(myrep);
+		if (error)
+			return (error);
+		/* Already received, bye bye */
+		if (myrep->r_mrep != NULL) {
+			nfs_rcvunlock(&nmp->nm_flag);
+			return (0);
+		}
+		/*
+		 * Get the next Rpc reply off the socket
+		 */
+		error = nfs_receive(myrep, &nam, &mrep);
+		nfs_rcvunlock(&nmp->nm_flag);
+		if (error) {
+
+			/*
+			 * Ignore routing errors on connectionless protocols??
+			 */
+			if (NFSIGNORE_SOERROR(nmp->nm_soflags, error)) {
+				nmp->nm_so->so_error = 0;
+				if (myrep->r_flags & R_GETONEREP)
+					return (0);
+				continue;
+			}
+			return (error);
+		}
+		if (nam)
+			m_freem(nam);
+	
+		/*
+		 * Get the xid and check that it is an rpc reply
+		 */
+		md = mrep;
+		dpos = mtod(md, caddr_t);
+		nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED);
+		rxid = *tl++;
+		if (*tl != rpc_reply) {
+			if (nmp->nm_flag & NFSMNT_NQNFS) {
+				if (nqnfs_callback(nmp, mrep, md, dpos))
+					nfsstats.rpcinvalid++;
+			} else {
+				nfsstats.rpcinvalid++;
+				m_freem(mrep);
+			}
+nfsmout:
+			if (myrep->r_flags & R_GETONEREP)
+				return (0);
+			continue;
+		}
+
+		/*
+		 * Loop through the request list to match up the reply
+		 * Iff no match, just drop the datagram
+		 */
+		for (rep = nfs_reqq.tqh_first; rep != 0;
+		    rep = rep->r_chain.tqe_next) {
+			if (rep->r_mrep == NULL && rxid == rep->r_xid) {
+				/* Found it.. */
+				rep->r_mrep = mrep;
+				rep->r_md = md;
+				rep->r_dpos = dpos;
+				if (nfsrtton) {
+					struct rttl *rt;
+
+					rt = &nfsrtt.rttl[nfsrtt.pos];
+					rt->proc = rep->r_procnum;
+					rt->rto = NFS_RTO(nmp, proct[rep->r_procnum]);
+					rt->sent = nmp->nm_sent;
+					rt->cwnd = nmp->nm_cwnd;
+					rt->srtt = nmp->nm_srtt[proct[rep->r_procnum] - 1];
+					rt->sdrtt = nmp->nm_sdrtt[proct[rep->r_procnum] - 1];
+					rt->fsid = nmp->nm_mountp->mnt_stat.f_fsid;
+					rt->tstamp = time;
+					if (rep->r_flags & R_TIMING)
+						rt->rtt = rep->r_rtt;
+					else
+						rt->rtt = 1000000;
+					nfsrtt.pos = (nfsrtt.pos + 1) % NFSRTTLOGSIZ;
+				}
+				/*
+				 * Update congestion window.
+				 * Do the additive increase of
+				 * one rpc/rtt.
+				 */
+				if (nmp->nm_cwnd <= nmp->nm_sent) {
+					nmp->nm_cwnd +=
+					   (NFS_CWNDSCALE * NFS_CWNDSCALE +
+					   (nmp->nm_cwnd >> 1)) / nmp->nm_cwnd;
+					if (nmp->nm_cwnd > NFS_MAXCWND)
+						nmp->nm_cwnd = NFS_MAXCWND;
+				}
+				rep->r_flags &= ~R_SENT;
+				nmp->nm_sent -= NFS_CWNDSCALE;
+				/*
+				 * Update rtt using a gain of 0.125 on the mean
+				 * and a gain of 0.25 on the deviation.
+				 */
+				if (rep->r_flags & R_TIMING) {
+					/*
+					 * Since the timer resolution of
+					 * NFS_HZ is so course, it can often
+					 * result in r_rtt == 0. Since
+					 * r_rtt == N means that the actual
+					 * rtt is between N+dt and N+2-dt ticks,
+					 * add 1.
+					 */
+					t1 = rep->r_rtt + 1;
+					t1 -= (NFS_SRTT(rep) >> 3);
+					NFS_SRTT(rep) += t1;
+					if (t1 < 0)
+						t1 = -t1;
+					t1 -= (NFS_SDRTT(rep) >> 2);
+					NFS_SDRTT(rep) += t1;
+				}
+				nmp->nm_timeouts = 0;
+				break;
+			}
+		}
+		/*
+		 * If not matched to a request, drop it.
+		 * If it's mine, get out.
+		 */
+		if (rep == 0) {
+			nfsstats.rpcunexpected++;
+			m_freem(mrep);
+		} else if (rep == myrep) {
+			if (rep->r_mrep == NULL)
+				panic("nfsreply nil");
+			return (0);
+		}
+		if (myrep->r_flags & R_GETONEREP)
+			return (0);
+	}
+}
+
+/*
+ * nfs_request - goes something like this
+ *	- fill in request struct
+ *	- links it into list
+ *	- calls nfs_send() for first transmit
+ *	- calls nfs_receive() to get reply
+ *	- break down rpc header and return with nfs reply pointed to
+ *	  by mrep or error
+ * nb: always frees up mreq mbuf list
+ */
+int
+nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp)
+	struct vnode *vp;
+	struct mbuf *mrest;
+	int procnum;
+	struct proc *procp;
+	struct ucred *cred;
+	struct mbuf **mrp;
+	struct mbuf **mdp;
+	caddr_t *dposp;
+{
+	register struct mbuf *m, *mrep;
+	register struct nfsreq *rep;
+	register u_long *tl;
+	register int i;
+	struct nfsmount *nmp;
+	struct mbuf *md, *mheadend;
+	struct nfsnode *np;
+	char nickv[RPCX_NICKVERF];
+	time_t reqtime, waituntil;
+	caddr_t dpos, cp2;
+	int t1, nqlflag, cachable, s, error = 0, mrest_len, auth_len, auth_type;
+	int trylater_delay = NQ_TRYLATERDEL, trylater_cnt = 0, failed_auth = 0;
+	int verf_len, verf_type;
+	u_long xid;
+	u_quad_t frev;
+	char *auth_str, *verf_str;
+	NFSKERBKEY_T key;		/* save session key */
+
+	nmp = VFSTONFS(vp->v_mount);
+	MALLOC(rep, struct nfsreq *, sizeof(struct nfsreq), M_NFSREQ, M_WAITOK);
+	rep->r_nmp = nmp;
+	rep->r_vp = vp;
+	rep->r_procp = procp;
+	rep->r_procnum = procnum;
+	i = 0;
+	m = mrest;
+	while (m) {
+		i += m->m_len;
+		m = m->m_next;
+	}
+	mrest_len = i;
+
+	/*
+	 * Get the RPC header with authorization.
+	 */
+kerbauth:
+	verf_str = auth_str = (char *)0;
+	if (nmp->nm_flag & NFSMNT_KERB) {
+		verf_str = nickv;
+		verf_len = sizeof (nickv);
+		auth_type = RPCAUTH_KERB4;
+		bzero((caddr_t)key, sizeof (key));
+		if (failed_auth || nfs_getnickauth(nmp, cred, &auth_str,
+			&auth_len, verf_str, verf_len)) {
+			error = nfs_getauth(nmp, rep, cred, &auth_str,
+				&auth_len, verf_str, &verf_len, key);
+			if (error) {
+				free((caddr_t)rep, M_NFSREQ);
+				m_freem(mrest);
+				return (error);
+			}
+		}
+	} else {
+		auth_type = RPCAUTH_UNIX;
+		if (cred->cr_ngroups < 1)
+			panic("nfsreq nogrps");
+		auth_len = ((((cred->cr_ngroups - 1) > nmp->nm_numgrps) ?
+			nmp->nm_numgrps : (cred->cr_ngroups - 1)) << 2) +
+			5 * NFSX_UNSIGNED;
+	}
+	m = nfsm_rpchead(cred, nmp->nm_flag, procnum, auth_type, auth_len,
+	     auth_str, verf_len, verf_str, mrest, mrest_len, &mheadend, &xid);
+	if (auth_str)
+		free(auth_str, M_TEMP);
+
+	/*
+	 * For stream protocols, insert a Sun RPC Record Mark.
+	 */
+	if (nmp->nm_sotype == SOCK_STREAM) {
+		M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
+		*mtod(m, u_long *) = htonl(0x80000000 |
+			 (m->m_pkthdr.len - NFSX_UNSIGNED));
+	}
+	rep->r_mreq = m;
+	rep->r_xid = xid;
+tryagain:
+	if (nmp->nm_flag & NFSMNT_SOFT)
+		rep->r_retry = nmp->nm_retry;
+	else
+		rep->r_retry = NFS_MAXREXMIT + 1;	/* past clip limit */
+	rep->r_rtt = rep->r_rexmit = 0;
+	if (proct[procnum] > 0)
+		rep->r_flags = R_TIMING;
+	else
+		rep->r_flags = 0;
+	rep->r_mrep = NULL;
+
+	/*
+	 * Do the client side RPC.
+	 */
+	nfsstats.rpcrequests++;
+	/*
+	 * Chain request into list of outstanding requests. Be sure
+	 * to put it LAST so timer finds oldest requests first.
+	 */
+	s = splsoftclock();
+	TAILQ_INSERT_TAIL(&nfs_reqq, rep, r_chain);
+
+	/* Get send time for nqnfs */
+	reqtime = time.tv_sec;
+
+	/*
+	 * If backing off another request or avoiding congestion, don't
+	 * send this one now but let timer do it. If not timing a request,
+	 * do it now.
+	 */
+	if (nmp->nm_so && (nmp->nm_sotype != SOCK_DGRAM ||
+		(nmp->nm_flag & NFSMNT_DUMBTIMR) ||
+		nmp->nm_sent < nmp->nm_cwnd)) {
+		splx(s);
+		if (nmp->nm_soflags & PR_CONNREQUIRED)
+			error = nfs_sndlock(&nmp->nm_flag, rep);
+		if (!error) {
+			m = m_copym(m, 0, M_COPYALL, M_WAIT);
+			error = nfs_send(nmp->nm_so, nmp->nm_nam, m, rep);
+			if (nmp->nm_soflags & PR_CONNREQUIRED)
+				nfs_sndunlock(&nmp->nm_flag);
+		}
+		if (!error && (rep->r_flags & R_MUSTRESEND) == 0) {
+			nmp->nm_sent += NFS_CWNDSCALE;
+			rep->r_flags |= R_SENT;
+		}
+	} else {
+		splx(s);
+		rep->r_rtt = -1;
+	}
+
+	/*
+	 * Wait for the reply from our send or the timer's.
+	 */
+	if (!error || error == EPIPE)
+		error = nfs_reply(rep);
+
+	/*
+	 * RPC done, unlink the request.
+	 */
+	s = splsoftclock();
+	TAILQ_REMOVE(&nfs_reqq, rep, r_chain);
+	splx(s);
+
+	/*
+	 * Decrement the outstanding request count.
+	 */
+	if (rep->r_flags & R_SENT) {
+		rep->r_flags &= ~R_SENT;	/* paranoia */
+		nmp->nm_sent -= NFS_CWNDSCALE;
+	}
+
+	/*
+	 * If there was a successful reply and a tprintf msg.
+	 * tprintf a response.
+	 */
+	if (!error && (rep->r_flags & R_TPRINTFMSG))
+		nfs_msg(rep->r_procp, nmp->nm_mountp->mnt_stat.f_mntfromname,
+		    "is alive again");
+	mrep = rep->r_mrep;
+	md = rep->r_md;
+	dpos = rep->r_dpos;
+	if (error) {
+		m_freem(rep->r_mreq);
+		free((caddr_t)rep, M_NFSREQ);
+		return (error);
+	}
+
+	/*
+	 * break down the rpc header and check if ok
+	 */
+	nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+	if (*tl++ == rpc_msgdenied) {
+		if (*tl == rpc_mismatch)
+			error = EOPNOTSUPP;
+		else if ((nmp->nm_flag & NFSMNT_KERB) && *tl++ == rpc_autherr) {
+			if (!failed_auth) {
+				failed_auth++;
+				mheadend->m_next = (struct mbuf *)0;
+				m_freem(mrep);
+				m_freem(rep->r_mreq);
+				goto kerbauth;
+			} else
+				error = EAUTH;
+		} else
+			error = EACCES;
+		m_freem(mrep);
+		m_freem(rep->r_mreq);
+		free((caddr_t)rep, M_NFSREQ);
+		return (error);
+	}
+
+	/*
+	 * Grab any Kerberos verifier, otherwise just throw it away.
+	 */
+	verf_type = fxdr_unsigned(int, *tl++);
+	i = fxdr_unsigned(int, *tl);
+	if ((nmp->nm_flag & NFSMNT_KERB) && verf_type == RPCAUTH_KERB4) {
+		error = nfs_savenickauth(nmp, cred, i, key, &md, &dpos, mrep);
+		if (error)
+			goto nfsmout;
+	} else if (i > 0)
+		nfsm_adv(nfsm_rndup(i));
+	nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+	/* 0 == ok */
+	if (*tl == 0) {
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		if (*tl != 0) {
+			error = fxdr_unsigned(int, *tl);
+			if ((nmp->nm_flag & NFSMNT_NFSV3) &&
+				error == NFSERR_TRYLATER) {
+				m_freem(mrep);
+				error = 0;
+				waituntil = time.tv_sec + trylater_delay;
+				while (time.tv_sec < waituntil)
+					(void) tsleep((caddr_t)&lbolt,
+						PSOCK, "nqnfstry", 0);
+				trylater_delay *= nfs_backoff[trylater_cnt];
+				if (trylater_cnt < 7)
+					trylater_cnt++;
+				goto tryagain;
+			}
+
+			/*
+			 * If the File Handle was stale, invalidate the
+			 * lookup cache, just in case.
+			 */
+			if (error == ESTALE)
+				cache_purge(vp);
+			if (nmp->nm_flag & NFSMNT_NFSV3) {
+				*mrp = mrep;
+				*mdp = md;
+				*dposp = dpos;
+				error |= NFSERR_RETERR;
+			} else
+				m_freem(mrep);
+			m_freem(rep->r_mreq);
+			free((caddr_t)rep, M_NFSREQ);
+			return (error);
+		}
+
+		/*
+		 * For nqnfs, get any lease in reply
+		 */
+		if (nmp->nm_flag & NFSMNT_NQNFS) {
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			if (*tl) {
+				np = VTONFS(vp);
+				nqlflag = fxdr_unsigned(int, *tl);
+				nfsm_dissect(tl, u_long *, 4*NFSX_UNSIGNED);
+				cachable = fxdr_unsigned(int, *tl++);
+				reqtime += fxdr_unsigned(int, *tl++);
+				if (reqtime > time.tv_sec) {
+				    fxdr_hyper(tl, &frev);
+				    nqnfs_clientlease(nmp, np, nqlflag,
+					cachable, reqtime, frev);
+				}
+			}
+		}
+		*mrp = mrep;
+		*mdp = md;
+		*dposp = dpos;
+		m_freem(rep->r_mreq);
+		FREE((caddr_t)rep, M_NFSREQ);
+		return (0);
+	}
+	m_freem(mrep);
+	error = EPROTONOSUPPORT;
+nfsmout:
+	m_freem(rep->r_mreq);
+	free((caddr_t)rep, M_NFSREQ);
+	return (error);
+}
+
+/*
+ * Generate the rpc reply header
+ * siz arg. is used to decide if adding a cluster is worthwhile
+ */
+int
+nfs_rephead(siz, nd, slp, err, cache, frev, mrq, mbp, bposp)
+	int siz;
+	struct nfsrv_descript *nd;
+	struct nfssvc_sock *slp;
+	int err;
+	int cache;
+	u_quad_t *frev;
+	struct mbuf **mrq;
+	struct mbuf **mbp;
+	caddr_t *bposp;
+{
+	register u_long *tl;
+	register struct mbuf *mreq;
+	caddr_t bpos;
+	struct mbuf *mb, *mb2;
+
+	MGETHDR(mreq, M_WAIT, MT_DATA);
+	mb = mreq;
+	/*
+	 * If this is a big reply, use a cluster else
+	 * try and leave leading space for the lower level headers.
+	 */
+	siz += RPC_REPLYSIZ;
+	if (siz >= MINCLSIZE) {
+		MCLGET(mreq, M_WAIT);
+	} else
+		mreq->m_data += max_hdr;
+	tl = mtod(mreq, u_long *);
+	mreq->m_len = 6 * NFSX_UNSIGNED;
+	bpos = ((caddr_t)tl) + mreq->m_len;
+	*tl++ = txdr_unsigned(nd->nd_retxid);
+	*tl++ = rpc_reply;
+	if (err == ERPCMISMATCH || (err & NFSERR_AUTHERR)) {
+		*tl++ = rpc_msgdenied;
+		if (err & NFSERR_AUTHERR) {
+			*tl++ = rpc_autherr;
+			*tl = txdr_unsigned(err & ~NFSERR_AUTHERR);
+			mreq->m_len -= NFSX_UNSIGNED;
+			bpos -= NFSX_UNSIGNED;
+		} else {
+			*tl++ = rpc_mismatch;
+			*tl++ = txdr_unsigned(RPC_VER2);
+			*tl = txdr_unsigned(RPC_VER2);
+		}
+	} else {
+		*tl++ = rpc_msgaccepted;
+
+		/*
+		 * For Kerberos authentication, we must send the nickname
+		 * verifier back, otherwise just RPCAUTH_NULL.
+		 */
+		if (nd->nd_flag & ND_KERBFULL) {
+		    register struct nfsuid *nuidp;
+		    struct timeval ktvin, ktvout;
+		    NFSKERBKEYSCHED_T keys;	/* stores key schedule */
+
+		    for (nuidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first;
+			nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
+			if (nuidp->nu_cr.cr_uid == nd->nd_cr.cr_uid &&
+			    (!nd->nd_nam2 || netaddr_match(NU_NETFAM(nuidp),
+			     &nuidp->nu_haddr, nd->nd_nam2)))
+			    break;
+		    }
+		    if (nuidp) {
+			ktvin.tv_sec =
+			    txdr_unsigned(nuidp->nu_timestamp.tv_sec - 1);
+			ktvin.tv_usec =
+			    txdr_unsigned(nuidp->nu_timestamp.tv_usec);
+
+			/*
+			 * Encrypt the timestamp in ecb mode using the
+			 * session key.
+			 */
+#ifdef NFSKERB
+			XXX
+#endif
+
+			*tl++ = rpc_auth_kerb;
+			*tl++ = txdr_unsigned(3 * NFSX_UNSIGNED);
+			*tl = ktvout.tv_sec;
+			nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED);
+			*tl++ = ktvout.tv_usec;
+			*tl++ = txdr_unsigned(nuidp->nu_cr.cr_uid);
+		    } else {
+			*tl++ = 0;
+			*tl++ = 0;
+		    }
+		} else {
+			*tl++ = 0;
+			*tl++ = 0;
+		}
+		switch (err) {
+		case EPROGUNAVAIL:
+			*tl = txdr_unsigned(RPC_PROGUNAVAIL);
+			break;
+		case EPROGMISMATCH:
+			*tl = txdr_unsigned(RPC_PROGMISMATCH);
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			if (nd->nd_flag & ND_NQNFS) {
+				*tl++ = txdr_unsigned(3);
+				*tl = txdr_unsigned(3);
+			} else {
+				*tl++ = txdr_unsigned(2);
+				*tl = txdr_unsigned(3);
+			}
+			break;
+		case EPROCUNAVAIL:
+			*tl = txdr_unsigned(RPC_PROCUNAVAIL);
+			break;
+		case EBADRPC:
+			*tl = txdr_unsigned(RPC_GARBAGE);
+			break;
+		default:
+			*tl = 0;
+			if (err != NFSERR_RETVOID) {
+				nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+				if (err)
+				    *tl = txdr_unsigned(nfsrv_errmap(nd, err));
+				else
+				    *tl = 0;
+			}
+			break;
+		};
+	}
+
+	/*
+	 * For nqnfs, piggyback lease as requested.
+	 */
+	if ((nd->nd_flag & ND_NQNFS) && err == 0) {
+		if (nd->nd_flag & ND_LEASE) {
+			nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED);
+			*tl++ = txdr_unsigned(nd->nd_flag & ND_LEASE);
+			*tl++ = txdr_unsigned(cache);
+			*tl++ = txdr_unsigned(nd->nd_duration);
+			txdr_hyper(frev, tl);
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = 0;
+		}
+	}
+	*mrq = mreq;
+	*mbp = mb;
+	*bposp = bpos;
+	if (err != 0 && err != NFSERR_RETVOID)
+		nfsstats.srvrpc_errs++;
+	return (0);
+}
+
+/*
+ * Nfs timer routine
+ * Scan the nfsreq list and retranmit any requests that have timed out
+ * To avoid retransmission attempts on STREAM sockets (in the future) make
+ * sure to set the r_retry field to 0 (implies nm_retry == 0).
+ */
+void
+nfs_timer(arg)
+	void *arg;	/* never used */
+{
+	register struct nfsreq *rep;
+	register struct mbuf *m;
+	register struct socket *so;
+	register struct nfsmount *nmp;
+	register int timeo;
+	register struct nfssvc_sock *slp;
+	static long lasttime = 0;
+	int s, error;
+	u_quad_t cur_usec;
+
+	s = splnet();
+	for (rep = nfs_reqq.tqh_first; rep != 0; rep = rep->r_chain.tqe_next) {
+		nmp = rep->r_nmp;
+		if (rep->r_mrep || (rep->r_flags & R_SOFTTERM))
+			continue;
+		if (nfs_sigintr(nmp, rep, rep->r_procp)) {
+			rep->r_flags |= R_SOFTTERM;
+			continue;
+		}
+		if (rep->r_rtt >= 0) {
+			rep->r_rtt++;
+			if (nmp->nm_flag & NFSMNT_DUMBTIMR)
+				timeo = nmp->nm_timeo;
+			else
+				timeo = NFS_RTO(nmp, proct[rep->r_procnum]);
+			if (nmp->nm_timeouts > 0)
+				timeo *= nfs_backoff[nmp->nm_timeouts - 1];
+			if (rep->r_rtt <= timeo)
+				continue;
+			if (nmp->nm_timeouts < 8)
+				nmp->nm_timeouts++;
+		}
+		/*
+		 * Check for server not responding
+		 */
+		if ((rep->r_flags & R_TPRINTFMSG) == 0 &&
+		     rep->r_rexmit > nmp->nm_deadthresh) {
+			nfs_msg(rep->r_procp,
+			    nmp->nm_mountp->mnt_stat.f_mntfromname,
+			    "not responding");
+			rep->r_flags |= R_TPRINTFMSG;
+		}
+		if (rep->r_rexmit >= rep->r_retry) {	/* too many */
+			nfsstats.rpctimeouts++;
+			rep->r_flags |= R_SOFTTERM;
+			continue;
+		}
+		if (nmp->nm_sotype != SOCK_DGRAM) {
+			if (++rep->r_rexmit > NFS_MAXREXMIT)
+				rep->r_rexmit = NFS_MAXREXMIT;
+			continue;
+		}
+		if ((so = nmp->nm_so) == NULL)
+			continue;
+
+		/*
+		 * If there is enough space and the window allows..
+		 *	Resend it
+		 * Set r_rtt to -1 in case we fail to send it now.
+		 */
+		rep->r_rtt = -1;
+		if (sbspace(&so->so_snd) >= rep->r_mreq->m_pkthdr.len &&
+		   ((nmp->nm_flag & NFSMNT_DUMBTIMR) ||
+		    (rep->r_flags & R_SENT) ||
+		    nmp->nm_sent < nmp->nm_cwnd) &&
+		   (m = m_copym(rep->r_mreq, 0, M_COPYALL, M_DONTWAIT))){
+			if ((nmp->nm_flag & NFSMNT_NOCONN) == 0)
+			    error = (*so->so_proto->pr_usrreq)(so, PRU_SEND, m,
+			    (struct mbuf *)0, (struct mbuf *)0);
+			else
+			    error = (*so->so_proto->pr_usrreq)(so, PRU_SEND, m,
+			    nmp->nm_nam, (struct mbuf *)0);
+			if (error) {
+				if (NFSIGNORE_SOERROR(nmp->nm_soflags, error))
+					so->so_error = 0;
+			} else {
+				/*
+				 * Iff first send, start timing
+				 * else turn timing off, backoff timer
+				 * and divide congestion window by 2.
+				 */
+				if (rep->r_flags & R_SENT) {
+					rep->r_flags &= ~R_TIMING;
+					if (++rep->r_rexmit > NFS_MAXREXMIT)
+						rep->r_rexmit = NFS_MAXREXMIT;
+					nmp->nm_cwnd >>= 1;
+					if (nmp->nm_cwnd < NFS_CWNDSCALE)
+						nmp->nm_cwnd = NFS_CWNDSCALE;
+					nfsstats.rpcretries++;
+				} else {
+					rep->r_flags |= R_SENT;
+					nmp->nm_sent += NFS_CWNDSCALE;
+				}
+				rep->r_rtt = 0;
+			}
+		}
+	}
+
+	/*
+	 * Call the nqnfs server timer once a second to handle leases.
+	 */
+	if (lasttime != time.tv_sec) {
+		lasttime = time.tv_sec;
+		nqnfs_serverd();
+	}
+
+	/*
+	 * Scan the write gathering queues for writes that need to be
+	 * completed now.
+	 */
+	cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
+	for (slp = nfssvc_sockhead.tqh_first; slp != 0;
+	    slp = slp->ns_chain.tqe_next) {
+	    if (slp->ns_tq.lh_first && slp->ns_tq.lh_first->nd_time<=cur_usec)
+		nfsrv_wakenfsd(slp);
+	}
+	splx(s);
+	timeout(nfs_timer, (void *)0, nfs_ticks);
+}
+
+/*
+ * Test for a termination condition pending on the process.
+ * This is used for NFSMNT_INT mounts.
+ */
+int
+nfs_sigintr(nmp, rep, p)
+	struct nfsmount *nmp;
+	struct nfsreq *rep;
+	register struct proc *p;
+{
+
+	if (rep && (rep->r_flags & R_SOFTTERM))
+		return (EINTR);
+	if (!(nmp->nm_flag & NFSMNT_INT))
+		return (0);
+	if (p && p->p_siglist &&
+	    (((p->p_siglist & ~p->p_sigmask) & ~p->p_sigignore) &
+	    NFSINT_SIGMASK))
+		return (EINTR);
+	return (0);
+}
+
+/*
+ * Lock a socket against others.
+ * Necessary for STREAM sockets to ensure you get an entire rpc request/reply
+ * and also to avoid race conditions between the processes with nfs requests
+ * in progress when a reconnect is necessary.
+ */
+int
+nfs_sndlock(flagp, rep)
+	register int *flagp;
+	struct nfsreq *rep;
+{
+	struct proc *p;
+	int slpflag = 0, slptimeo = 0;
+
+	if (rep) {
+		p = rep->r_procp;
+		if (rep->r_nmp->nm_flag & NFSMNT_INT)
+			slpflag = PCATCH;
+	} else
+		p = (struct proc *)0;
+	while (*flagp & NFSMNT_SNDLOCK) {
+		if (nfs_sigintr(rep->r_nmp, rep, p))
+			return (EINTR);
+		*flagp |= NFSMNT_WANTSND;
+		(void) tsleep((caddr_t)flagp, slpflag | (PZERO - 1), "nfsndlck",
+			slptimeo);
+		if (slpflag == PCATCH) {
+			slpflag = 0;
+			slptimeo = 2 * hz;
+		}
+	}
+	*flagp |= NFSMNT_SNDLOCK;
+	return (0);
+}
+
+/*
+ * Unlock the stream socket for others.
+ */
+void
+nfs_sndunlock(flagp)
+	register int *flagp;
+{
+
+	if ((*flagp & NFSMNT_SNDLOCK) == 0)
+		panic("nfs sndunlock");
+	*flagp &= ~NFSMNT_SNDLOCK;
+	if (*flagp & NFSMNT_WANTSND) {
+		*flagp &= ~NFSMNT_WANTSND;
+		wakeup((caddr_t)flagp);
+	}
+}
+
+int
+nfs_rcvlock(rep)
+	register struct nfsreq *rep;
+{
+	register int *flagp = &rep->r_nmp->nm_flag;
+	int slpflag, slptimeo = 0;
+
+	if (*flagp & NFSMNT_INT)
+		slpflag = PCATCH;
+	else
+		slpflag = 0;
+	while (*flagp & NFSMNT_RCVLOCK) {
+		if (nfs_sigintr(rep->r_nmp, rep, rep->r_procp))
+			return (EINTR);
+		*flagp |= NFSMNT_WANTRCV;
+		(void) tsleep((caddr_t)flagp, slpflag | (PZERO - 1), "nfsrcvlk",
+			slptimeo);
+		if (slpflag == PCATCH) {
+			slpflag = 0;
+			slptimeo = 2 * hz;
+		}
+	}
+	*flagp |= NFSMNT_RCVLOCK;
+	return (0);
+}
+
+/*
+ * Unlock the stream socket for others.
+ */
+void
+nfs_rcvunlock(flagp)
+	register int *flagp;
+{
+
+	if ((*flagp & NFSMNT_RCVLOCK) == 0)
+		panic("nfs rcvunlock");
+	*flagp &= ~NFSMNT_RCVLOCK;
+	if (*flagp & NFSMNT_WANTRCV) {
+		*flagp &= ~NFSMNT_WANTRCV;
+		wakeup((caddr_t)flagp);
+	}
+}
+
+/*
+ * Check for badly aligned mbuf data areas and
+ * realign data in an mbuf list by copying the data areas up, as required.
+ */
+void
+nfs_realign(m, hsiz)
+	register struct mbuf *m;
+	int hsiz;
+{
+	register struct mbuf *m2;
+	register int siz, mlen, olen;
+	register caddr_t tcp, fcp;
+	struct mbuf *mnew;
+
+	while (m) {
+	    /*
+	     * This never happens for UDP, rarely happens for TCP
+	     * but frequently happens for iso transport.
+	     */
+	    if ((m->m_len & 0x3) || (mtod(m, int) & 0x3)) {
+		olen = m->m_len;
+		fcp = mtod(m, caddr_t);
+		if ((int)fcp & 0x3) {
+			m->m_flags &= ~M_PKTHDR;
+			if (m->m_flags & M_EXT)
+				m->m_data = m->m_ext.ext_buf +
+					((m->m_ext.ext_size - olen) & ~0x3);
+			else
+				m->m_data = m->m_dat;
+		}
+		m->m_len = 0;
+		tcp = mtod(m, caddr_t);
+		mnew = m;
+		m2 = m->m_next;
+	
+		/*
+		 * If possible, only put the first invariant part
+		 * of the RPC header in the first mbuf.
+		 */
+		mlen = M_TRAILINGSPACE(m);
+		if (olen <= hsiz && mlen > hsiz)
+			mlen = hsiz;
+	
+		/*
+		 * Loop through the mbuf list consolidating data.
+		 */
+		while (m) {
+			while (olen > 0) {
+				if (mlen == 0) {
+					m2->m_flags &= ~M_PKTHDR;
+					if (m2->m_flags & M_EXT)
+						m2->m_data = m2->m_ext.ext_buf;
+					else
+						m2->m_data = m2->m_dat;
+					m2->m_len = 0;
+					mlen = M_TRAILINGSPACE(m2);
+					tcp = mtod(m2, caddr_t);
+					mnew = m2;
+					m2 = m2->m_next;
+				}
+				siz = min(mlen, olen);
+				if (tcp != fcp)
+					bcopy(fcp, tcp, siz);
+				mnew->m_len += siz;
+				mlen -= siz;
+				olen -= siz;
+				tcp += siz;
+				fcp += siz;
+			}
+			m = m->m_next;
+			if (m) {
+				olen = m->m_len;
+				fcp = mtod(m, caddr_t);
+			}
+		}
+	
+		/*
+		 * Finally, set m_len == 0 for any trailing mbufs that have
+		 * been copied out of.
+		 */
+		while (m2) {
+			m2->m_len = 0;
+			m2 = m2->m_next;
+		}
+		return;
+	    }
+	    m = m->m_next;
+	}
+}
+
+/*
+ * Socket upcall routine for the nfsd sockets.
+ * The caddr_t arg is a pointer to the "struct nfssvc_sock".
+ * Essentially do as much as possible non-blocking, else punt and it will
+ * be called with M_WAIT from an nfsd.
+ */
+void
+nfsrv_rcv(so, arg, waitflag)
+	struct socket *so;
+	caddr_t arg;
+	int waitflag;
+{
+	register struct nfssvc_sock *slp = (struct nfssvc_sock *)arg;
+	register struct mbuf *m;
+	struct mbuf *mp, *nam;
+	struct uio auio;
+	int flags, error;
+
+	if ((slp->ns_flag & SLP_VALID) == 0)
+		return;
+#ifdef notdef
+	/*
+	 * Define this to test for nfsds handling this under heavy load.
+	 */
+	if (waitflag == M_DONTWAIT) {
+		slp->ns_flag |= SLP_NEEDQ; goto dorecs;
+	}
+#endif
+	auio.uio_procp = NULL;
+	if (so->so_type == SOCK_STREAM) {
+		/*
+		 * If there are already records on the queue, defer soreceive()
+		 * to an nfsd so that there is feedback to the TCP layer that
+		 * the nfs servers are heavily loaded.
+		 */
+		if (slp->ns_rec && waitflag == M_DONTWAIT) {
+			slp->ns_flag |= SLP_NEEDQ;
+			goto dorecs;
+		}
+
+		/*
+		 * Do soreceive().
+		 */
+		auio.uio_resid = 1000000000;
+		flags = MSG_DONTWAIT;
+		error = soreceive(so, &nam, &auio, &mp, (struct mbuf **)0, &flags);
+		if (error || mp == (struct mbuf *)0) {
+			if (error == EWOULDBLOCK)
+				slp->ns_flag |= SLP_NEEDQ;
+			else
+				slp->ns_flag |= SLP_DISCONN;
+			goto dorecs;
+		}
+		m = mp;
+		if (slp->ns_rawend) {
+			slp->ns_rawend->m_next = m;
+			slp->ns_cc += 1000000000 - auio.uio_resid;
+		} else {
+			slp->ns_raw = m;
+			slp->ns_cc = 1000000000 - auio.uio_resid;
+		}
+		while (m->m_next)
+			m = m->m_next;
+		slp->ns_rawend = m;
+
+		/*
+		 * Now try and parse record(s) out of the raw stream data.
+		 */
+		error = nfsrv_getstream(slp, waitflag);
+		if (error) {
+			if (error == EPERM)
+				slp->ns_flag |= SLP_DISCONN;
+			else
+				slp->ns_flag |= SLP_NEEDQ;
+		}
+	} else {
+		do {
+			auio.uio_resid = 1000000000;
+			flags = MSG_DONTWAIT;
+			error = soreceive(so, &nam, &auio, &mp,
+						(struct mbuf **)0, &flags);
+			if (mp) {
+				nfs_realign(mp, 10 * NFSX_UNSIGNED);
+				if (nam) {
+					m = nam;
+					m->m_next = mp;
+				} else
+					m = mp;
+				if (slp->ns_recend)
+					slp->ns_recend->m_nextpkt = m;
+				else
+					slp->ns_rec = m;
+				slp->ns_recend = m;
+				m->m_nextpkt = (struct mbuf *)0;
+			}
+			if (error) {
+				if ((so->so_proto->pr_flags & PR_CONNREQUIRED)
+					&& error != EWOULDBLOCK) {
+					slp->ns_flag |= SLP_DISCONN;
+					goto dorecs;
+				}
+			}
+		} while (mp);
+	}
+
+	/*
+	 * Now try and process the request records, non-blocking.
+	 */
+dorecs:
+	if (waitflag == M_DONTWAIT &&
+		(slp->ns_rec || (slp->ns_flag & (SLP_NEEDQ | SLP_DISCONN))))
+		nfsrv_wakenfsd(slp);
+}
+
+/*
+ * Try and extract an RPC request from the mbuf data list received on a
+ * stream socket. The "waitflag" argument indicates whether or not it
+ * can sleep.
+ */
+int
+nfsrv_getstream(slp, waitflag)
+	register struct nfssvc_sock *slp;
+	int waitflag;
+{
+	register struct mbuf *m, **mpp;
+	register char *cp1, *cp2;
+	register int len;
+	struct mbuf *om, *m2, *recm = 0;
+	u_long recmark;
+
+	if (slp->ns_flag & SLP_GETSTREAM)
+		panic("nfs getstream");
+	slp->ns_flag |= SLP_GETSTREAM;
+	for (;;) {
+	    if (slp->ns_reclen == 0) {
+		if (slp->ns_cc < NFSX_UNSIGNED) {
+			slp->ns_flag &= ~SLP_GETSTREAM;
+			return (0);
+		}
+		m = slp->ns_raw;
+		if (m->m_len >= NFSX_UNSIGNED) {
+			bcopy(mtod(m, caddr_t), (caddr_t)&recmark, NFSX_UNSIGNED);
+			m->m_data += NFSX_UNSIGNED;
+			m->m_len -= NFSX_UNSIGNED;
+		} else {
+			cp1 = (caddr_t)&recmark;
+			cp2 = mtod(m, caddr_t);
+			while (cp1 < ((caddr_t)&recmark) + NFSX_UNSIGNED) {
+				while (m->m_len == 0) {
+					m = m->m_next;
+					cp2 = mtod(m, caddr_t);
+				}
+				*cp1++ = *cp2++;
+				m->m_data++;
+				m->m_len--;
+			}
+		}
+		slp->ns_cc -= NFSX_UNSIGNED;
+		recmark = ntohl(recmark);
+		slp->ns_reclen = recmark & ~0x80000000;
+		if (recmark & 0x80000000)
+			slp->ns_flag |= SLP_LASTFRAG;
+		else
+			slp->ns_flag &= ~SLP_LASTFRAG;
+		if (slp->ns_reclen < NFS_MINPACKET || slp->ns_reclen > NFS_MAXPACKET) {
+			slp->ns_flag &= ~SLP_GETSTREAM;
+			return (EPERM);
+		}
+	    }
+
+	    /*
+	     * Now get the record part.
+	     */
+	    if (slp->ns_cc == slp->ns_reclen) {
+		recm = slp->ns_raw;
+		slp->ns_raw = slp->ns_rawend = (struct mbuf *)0;
+		slp->ns_cc = slp->ns_reclen = 0;
+	    } else if (slp->ns_cc > slp->ns_reclen) {
+		len = 0;
+		m = slp->ns_raw;
+		om = (struct mbuf *)0;
+		while (len < slp->ns_reclen) {
+			if ((len + m->m_len) > slp->ns_reclen) {
+				m2 = m_copym(m, 0, slp->ns_reclen - len,
+					waitflag);
+				if (m2) {
+					if (om) {
+						om->m_next = m2;
+						recm = slp->ns_raw;
+					} else
+						recm = m2;
+					m->m_data += slp->ns_reclen - len;
+					m->m_len -= slp->ns_reclen - len;
+					len = slp->ns_reclen;
+				} else {
+					slp->ns_flag &= ~SLP_GETSTREAM;
+					return (EWOULDBLOCK);
+				}
+			} else if ((len + m->m_len) == slp->ns_reclen) {
+				om = m;
+				len += m->m_len;
+				m = m->m_next;
+				recm = slp->ns_raw;
+				om->m_next = (struct mbuf *)0;
+			} else {
+				om = m;
+				len += m->m_len;
+				m = m->m_next;
+			}
+		}
+		slp->ns_raw = m;
+		slp->ns_cc -= len;
+		slp->ns_reclen = 0;
+	    } else {
+		slp->ns_flag &= ~SLP_GETSTREAM;
+		return (0);
+	    }
+
+	    /*
+	     * Accumulate the fragments into a record.
+	     */
+	    mpp = &slp->ns_frag;
+	    while (*mpp)
+		mpp = &((*mpp)->m_next);
+	    *mpp = recm;
+	    if (slp->ns_flag & SLP_LASTFRAG) {
+		nfs_realign(slp->ns_frag, 10 * NFSX_UNSIGNED);
+		if (slp->ns_recend)
+		    slp->ns_recend->m_nextpkt = slp->ns_frag;
+		else
+		    slp->ns_rec = slp->ns_frag;
+		slp->ns_recend = slp->ns_frag;
+		slp->ns_frag = (struct mbuf *)0;
+	    }
+	}
+}
+
+/*
+ * Parse an RPC header.
+ */
+int
+nfsrv_dorec(slp, nfsd, ndp)
+	register struct nfssvc_sock *slp;
+	struct nfsd *nfsd;
+	struct nfsrv_descript **ndp;
+{
+	register struct mbuf *m, *nam;
+	register struct nfsrv_descript *nd;
+	int error;
+
+	*ndp = NULL;
+	if ((slp->ns_flag & SLP_VALID) == 0 ||
+	    (m = slp->ns_rec) == (struct mbuf *)0)
+		return (ENOBUFS);
+	slp->ns_rec = m->m_nextpkt;
+	if (slp->ns_rec)
+		m->m_nextpkt = (struct mbuf *)0;
+	else
+		slp->ns_recend = (struct mbuf *)0;
+	if (m->m_type == MT_SONAME) {
+		nam = m;
+		m = m->m_next;
+		nam->m_next = NULL;
+	} else
+		nam = NULL;
+	MALLOC(nd, struct nfsrv_descript *, sizeof (struct nfsrv_descript),
+		M_NFSRVDESC, M_WAITOK);
+	nd->nd_md = nd->nd_mrep = m;
+	nd->nd_nam2 = nam;
+	nd->nd_dpos = mtod(m, caddr_t);
+	error = nfs_getreq(nd, nfsd, TRUE);
+	if (error) {
+		m_freem(nam);
+		free((caddr_t)nd, M_NFSRVDESC);
+		return (error);
+	}
+	*ndp = nd;
+	nfsd->nfsd_nd = nd;
+	return (0);
+}
+
+/*
+ * Parse an RPC request
+ * - verify it
+ * - fill in the cred struct.
+ */
+int
+nfs_getreq(nd, nfsd, has_header)
+	register struct nfsrv_descript *nd;
+	struct nfsd *nfsd;
+	int has_header;
+{
+	register int len, i;
+	register u_long *tl;
+	register long t1;
+	struct uio uio;
+	struct iovec iov;
+	caddr_t dpos, cp2, cp;
+	u_long nfsvers, auth_type;
+	uid_t nickuid;
+	int error = 0, nqnfs = 0, ticklen;
+	struct mbuf *mrep, *md;
+	register struct nfsuid *nuidp;
+	struct timeval tvin, tvout;
+	NFSKERBKEYSCHED_T keys;	/* stores key schedule */
+
+	mrep = nd->nd_mrep;
+	md = nd->nd_md;
+	dpos = nd->nd_dpos;
+	if (has_header) {
+		nfsm_dissect(tl, u_long *, 10 * NFSX_UNSIGNED);
+		nd->nd_retxid = fxdr_unsigned(u_long, *tl++);
+		if (*tl++ != rpc_call) {
+			m_freem(mrep);
+			return (EBADRPC);
+		}
+	} else
+		nfsm_dissect(tl, u_long *, 8 * NFSX_UNSIGNED);
+	nd->nd_repstat = 0;
+	nd->nd_flag = 0;
+	if (*tl++ != rpc_vers) {
+		nd->nd_repstat = ERPCMISMATCH;
+		nd->nd_procnum = NFSPROC_NOOP;
+		return (0);
+	}
+	if (*tl != nfs_prog) {
+		if (*tl == nqnfs_prog)
+			nqnfs++;
+		else {
+			nd->nd_repstat = EPROGUNAVAIL;
+			nd->nd_procnum = NFSPROC_NOOP;
+			return (0);
+		}
+	}
+	tl++;
+	nfsvers = fxdr_unsigned(u_long, *tl++);
+	if (((nfsvers < NFS_VER2 || nfsvers > NFS_VER3) && !nqnfs) ||
+		(nfsvers != NQNFS_VER3 && nqnfs)) {
+		nd->nd_repstat = EPROGMISMATCH;
+		nd->nd_procnum = NFSPROC_NOOP;
+		return (0);
+	}
+	if (nqnfs)
+		nd->nd_flag = (ND_NFSV3 | ND_NQNFS);
+	else if (nfsvers == NFS_VER3)
+		nd->nd_flag = ND_NFSV3;
+	nd->nd_procnum = fxdr_unsigned(u_long, *tl++);
+	if (nd->nd_procnum == NFSPROC_NULL)
+		return (0);
+	if (nd->nd_procnum >= NFS_NPROCS ||
+		(!nqnfs && nd->nd_procnum >= NQNFSPROC_GETLEASE) ||
+		(!nd->nd_flag && nd->nd_procnum > NFSV2PROC_STATFS)) {
+		nd->nd_repstat = EPROCUNAVAIL;
+		nd->nd_procnum = NFSPROC_NOOP;
+		return (0);
+	}
+	if ((nd->nd_flag & ND_NFSV3) == 0)
+		nd->nd_procnum = nfsv3_procid[nd->nd_procnum];
+	auth_type = *tl++;
+	len = fxdr_unsigned(int, *tl++);
+	if (len < 0 || len > RPCAUTH_MAXSIZ) {
+		m_freem(mrep);
+		return (EBADRPC);
+	}
+
+	nd->nd_flag &= ~ND_KERBAUTH;
+	/*
+	 * Handle auth_unix or auth_kerb.
+	 */
+	if (auth_type == rpc_auth_unix) {
+		len = fxdr_unsigned(int, *++tl);
+		if (len < 0 || len > NFS_MAXNAMLEN) {
+			m_freem(mrep);
+			return (EBADRPC);
+		}
+		nfsm_adv(nfsm_rndup(len));
+		nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+		bzero((caddr_t)&nd->nd_cr, sizeof (struct ucred));
+		nd->nd_cr.cr_ref = 1;
+		nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++);
+		nd->nd_cr.cr_gid = fxdr_unsigned(gid_t, *tl++);
+		len = fxdr_unsigned(int, *tl);
+		if (len < 0 || len > RPCAUTH_UNIXGIDS) {
+			m_freem(mrep);
+			return (EBADRPC);
+		}
+		nfsm_dissect(tl, u_long *, (len + 2) * NFSX_UNSIGNED);
+		for (i = 1; i <= len; i++)
+		    if (i < NGROUPS)
+			nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++);
+		    else
+			tl++;
+		nd->nd_cr.cr_ngroups = (len >= NGROUPS) ? NGROUPS : (len + 1);
+		if (nd->nd_cr.cr_ngroups > 1)
+		    nfsrvw_sort(nd->nd_cr.cr_groups, nd->nd_cr.cr_ngroups);
+		len = fxdr_unsigned(int, *++tl);
+		if (len < 0 || len > RPCAUTH_MAXSIZ) {
+			m_freem(mrep);
+			return (EBADRPC);
+		}
+		if (len > 0)
+			nfsm_adv(nfsm_rndup(len));
+	} else if (auth_type == rpc_auth_kerb) {
+		switch (fxdr_unsigned(int, *tl++)) {
+		case RPCAKN_FULLNAME:
+			ticklen = fxdr_unsigned(int, *tl);
+			*((u_long *)nfsd->nfsd_authstr) = *tl;
+			uio.uio_resid = nfsm_rndup(ticklen) + NFSX_UNSIGNED;
+			nfsd->nfsd_authlen = uio.uio_resid + NFSX_UNSIGNED;
+			if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) {
+				m_freem(mrep);
+				return (EBADRPC);
+			}
+			uio.uio_offset = 0;
+			uio.uio_iov = &iov;
+			uio.uio_iovcnt = 1;
+			uio.uio_segflg = UIO_SYSSPACE;
+			iov.iov_base = (caddr_t)&nfsd->nfsd_authstr[4];
+			iov.iov_len = RPCAUTH_MAXSIZ - 4;
+			nfsm_mtouio(&uio, uio.uio_resid);
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+			if (*tl++ != rpc_auth_kerb ||
+				fxdr_unsigned(int, *tl) != 4 * NFSX_UNSIGNED) {
+				printf("Bad kerb verifier\n");
+				nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+			nfsm_dissect(cp, caddr_t, 4 * NFSX_UNSIGNED);
+			tl = (u_long *)cp;
+			if (fxdr_unsigned(int, *tl) != RPCAKN_FULLNAME) {
+				printf("Not fullname kerb verifier\n");
+				nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+			cp += NFSX_UNSIGNED;
+			bcopy(cp, nfsd->nfsd_verfstr, 3 * NFSX_UNSIGNED);
+			nfsd->nfsd_verflen = 3 * NFSX_UNSIGNED;
+			nd->nd_flag |= ND_KERBFULL;
+			nfsd->nfsd_flag |= NFSD_NEEDAUTH;
+			break;
+		case RPCAKN_NICKNAME:
+			if (len != 2 * NFSX_UNSIGNED) {
+				printf("Kerb nickname short\n");
+				nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADCRED);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+			nickuid = fxdr_unsigned(uid_t, *tl);
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+			if (*tl++ != rpc_auth_kerb ||
+				fxdr_unsigned(int, *tl) != 3 * NFSX_UNSIGNED) {
+				printf("Kerb nick verifier bad\n");
+				nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+			nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+			tvin.tv_sec = *tl++;
+			tvin.tv_usec = *tl;
+
+			for (nuidp = NUIDHASH(nfsd->nfsd_slp,nickuid)->lh_first;
+			    nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
+				if (nuidp->nu_cr.cr_uid == nickuid &&
+				    (!nd->nd_nam2 ||
+				     netaddr_match(NU_NETFAM(nuidp),
+				      &nuidp->nu_haddr, nd->nd_nam2)))
+					break;
+			}
+			if (!nuidp) {
+				nd->nd_repstat =
+					(NFSERR_AUTHERR|AUTH_REJECTCRED);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+
+			/*
+			 * Now, decrypt the timestamp using the session key
+			 * and validate it.
+			 */
+#ifdef NFSKERB
+			XXX
+#endif
+
+			tvout.tv_sec = fxdr_unsigned(long, tvout.tv_sec);
+			tvout.tv_usec = fxdr_unsigned(long, tvout.tv_usec);
+			if (nuidp->nu_expire < time.tv_sec ||
+			    nuidp->nu_timestamp.tv_sec > tvout.tv_sec ||
+			    (nuidp->nu_timestamp.tv_sec == tvout.tv_sec &&
+			     nuidp->nu_timestamp.tv_usec > tvout.tv_usec)) {
+				nuidp->nu_expire = 0;
+				nd->nd_repstat =
+				    (NFSERR_AUTHERR|AUTH_REJECTVERF);
+				nd->nd_procnum = NFSPROC_NOOP;
+				return (0);
+			}
+			nfsrv_setcred(&nuidp->nu_cr, &nd->nd_cr);
+			nd->nd_flag |= ND_KERBNICK;
+		};
+	} else {
+		nd->nd_repstat = (NFSERR_AUTHERR | AUTH_REJECTCRED);
+		nd->nd_procnum = NFSPROC_NOOP;
+		return (0);
+	}
+
+	/*
+	 * For nqnfs, get piggybacked lease request.
+	 */
+	if (nqnfs && nd->nd_procnum != NQNFSPROC_EVICTED) {
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		nd->nd_flag |= fxdr_unsigned(int, *tl);
+		if (nd->nd_flag & ND_LEASE) {
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			nd->nd_duration = fxdr_unsigned(int, *tl);
+		} else
+			nd->nd_duration = NQ_MINLEASE;
+	} else
+		nd->nd_duration = NQ_MINLEASE;
+	nd->nd_md = md;
+	nd->nd_dpos = dpos;
+	return (0);
+nfsmout:
+	return (error);
+}
+
+/*
+ * Search for a sleeping nfsd and wake it up.
+ * SIDE EFFECT: If none found, set NFSD_CHECKSLP flag, so that one of the
+ * running nfsds will go look for the work in the nfssvc_sock list.
+ */
+void
+nfsrv_wakenfsd(slp)
+	struct nfssvc_sock *slp;
+{
+	register struct nfsd *nd;
+
+	if ((slp->ns_flag & SLP_VALID) == 0)
+		return;
+	for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nfsd_chain.tqe_next) {
+		if (nd->nfsd_flag & NFSD_WAITING) {
+			nd->nfsd_flag &= ~NFSD_WAITING;
+			if (nd->nfsd_slp)
+				panic("nfsd wakeup");
+			slp->ns_sref++;
+			nd->nfsd_slp = slp;
+			wakeup((caddr_t)nd);
+			return;
+		}
+	}
+	slp->ns_flag |= SLP_DOREC;
+	nfsd_head_flag |= NFSD_CHECKSLP;
+}
+
+int
+nfs_msg(p, server, msg)
+	struct proc *p;
+	char *server, *msg;
+{
+	tpr_t tpr;
+
+	if (p)
+		tpr = tprintf_open(p);
+	else
+		tpr = NULL;
+	tprintf(tpr, "nfs server %s: %s\n", server, msg);
+	tprintf_close(tpr);
+	return (0);
+}
diff --git a/sys/nfs/nfs_srvcache.c b/sys/nfs/nfs_srvcache.c
new file mode 100644
index 000000000000..954987bf4c55
--- /dev/null
+++ b/sys/nfs/nfs_srvcache.c
@@ -0,0 +1,336 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_srvcache.c	8.3 (Berkeley) 3/30/95
+ */
+
+/*
+ * Reference: Chet Juszczak, "Improving the Performance and Correctness
+ *		of an NFS Server", in Proc. Winter 1989 USENIX Conference,
+ *		pages 53-63. San Diego, February 1989.
+ */
+#include <sys/param.h>
+#include <sys/vnode.h>
+#include <sys/mount.h>
+#include <sys/kernel.h>
+#include <sys/systm.h>
+#include <sys/proc.h>
+#include <sys/mbuf.h>
+#include <sys/malloc.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+
+#include <netinet/in.h>
+#ifdef ISO
+#include <netiso/iso.h>
+#endif
+#include <nfs/nfsm_subs.h>
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsrvcache.h>
+#include <nfs/nqnfs.h>
+
+extern struct nfsstats nfsstats;
+extern int nfsv2_procid[NFS_NPROCS];
+long numnfsrvcache, desirednfsrvcache = NFSRVCACHESIZ;
+
+#define	NFSRCHASH(xid) \
+	(&nfsrvhashtbl[((xid) + ((xid) >> 24)) & nfsrvhash])
+LIST_HEAD(nfsrvhash, nfsrvcache) *nfsrvhashtbl;
+TAILQ_HEAD(nfsrvlru, nfsrvcache) nfsrvlruhead;
+u_long nfsrvhash;
+
+#define TRUE	1
+#define	FALSE	0
+
+#define	NETFAMILY(rp) \
+		(((rp)->rc_flag & RC_INETADDR) ? AF_INET : AF_ISO)
+
+/*
+ * Static array that defines which nfs rpc's are nonidempotent
+ */
+int nonidempotent[NFS_NPROCS] = {
+	FALSE,
+	FALSE,
+	TRUE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+};
+
+/* True iff the rpc reply is an nfs status ONLY! */
+static int nfsv2_repstat[NFS_NPROCS] = {
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	FALSE,
+	TRUE,
+	TRUE,
+	TRUE,
+	TRUE,
+	FALSE,
+	TRUE,
+	FALSE,
+	FALSE,
+};
+
+/*
+ * Initialize the server request cache list
+ */
+void
+nfsrv_initcache()
+{
+
+	nfsrvhashtbl = hashinit(desirednfsrvcache, M_NFSD, &nfsrvhash);
+	TAILQ_INIT(&nfsrvlruhead);
+}
+
+/*
+ * Look for the request in the cache
+ * If found then
+ *    return action and optionally reply
+ * else
+ *    insert it in the cache
+ *
+ * The rules are as follows:
+ * - if in progress, return DROP request
+ * - if completed within DELAY of the current time, return DROP it
+ * - if completed a longer time ago return REPLY if the reply was cached or
+ *   return DOIT
+ * Update/add new request at end of lru list
+ */
+int
+nfsrv_getcache(nd, slp, repp)
+	register struct nfsrv_descript *nd;
+	struct nfssvc_sock *slp;
+	struct mbuf **repp;
+{
+	register struct nfsrvcache *rp;
+	struct mbuf *mb;
+	struct sockaddr_in *saddr;
+	caddr_t bpos;
+	int ret;
+
+	/*
+	 * Don't cache recent requests for reliable transport protocols.
+	 * (Maybe we should for the case of a reconnect, but..)
+	 */
+	if (!nd->nd_nam2)
+		return (RC_DOIT);
+loop:
+	for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0;
+	    rp = rp->rc_hash.le_next) {
+	    if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc &&
+		netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) {
+			if ((rp->rc_flag & RC_LOCKED) != 0) {
+				rp->rc_flag |= RC_WANTED;
+				(void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0);
+				goto loop;
+			}
+			rp->rc_flag |= RC_LOCKED;
+			/* If not at end of LRU chain, move it there */
+			if (rp->rc_lru.tqe_next) {
+				TAILQ_REMOVE(&nfsrvlruhead, rp, rc_lru);
+				TAILQ_INSERT_TAIL(&nfsrvlruhead, rp, rc_lru);
+			}
+			if (rp->rc_state == RC_UNUSED)
+				panic("nfsrv cache");
+			if (rp->rc_state == RC_INPROG) {
+				nfsstats.srvcache_inproghits++;
+				ret = RC_DROPIT;
+			} else if (rp->rc_flag & RC_REPSTATUS) {
+				nfsstats.srvcache_nonidemdonehits++;
+				nfs_rephead(0, nd, slp, rp->rc_status,
+				   0, (u_quad_t *)0, repp, &mb, &bpos);
+				ret = RC_REPLY;
+			} else if (rp->rc_flag & RC_REPMBUF) {
+				nfsstats.srvcache_nonidemdonehits++;
+				*repp = m_copym(rp->rc_reply, 0, M_COPYALL,
+						M_WAIT);
+				ret = RC_REPLY;
+			} else {
+				nfsstats.srvcache_idemdonehits++;
+				rp->rc_state = RC_INPROG;
+				ret = RC_DOIT;
+			}
+			rp->rc_flag &= ~RC_LOCKED;
+			if (rp->rc_flag & RC_WANTED) {
+				rp->rc_flag &= ~RC_WANTED;
+				wakeup((caddr_t)rp);
+			}
+			return (ret);
+		}
+	}
+	nfsstats.srvcache_misses++;
+	if (numnfsrvcache < desirednfsrvcache) {
+		rp = (struct nfsrvcache *)malloc((u_long)sizeof *rp,
+		    M_NFSD, M_WAITOK);
+		bzero((char *)rp, sizeof *rp);
+		numnfsrvcache++;
+		rp->rc_flag = RC_LOCKED;
+	} else {
+		rp = nfsrvlruhead.tqh_first;
+		while ((rp->rc_flag & RC_LOCKED) != 0) {
+			rp->rc_flag |= RC_WANTED;
+			(void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0);
+			rp = nfsrvlruhead.tqh_first;
+		}
+		rp->rc_flag |= RC_LOCKED;
+		LIST_REMOVE(rp, rc_hash);
+		TAILQ_REMOVE(&nfsrvlruhead, rp, rc_lru);
+		if (rp->rc_flag & RC_REPMBUF)
+			m_freem(rp->rc_reply);
+		if (rp->rc_flag & RC_NAM)
+			MFREE(rp->rc_nam, mb);
+		rp->rc_flag &= (RC_LOCKED | RC_WANTED);
+	}
+	TAILQ_INSERT_TAIL(&nfsrvlruhead, rp, rc_lru);
+	rp->rc_state = RC_INPROG;
+	rp->rc_xid = nd->nd_retxid;
+	saddr = mtod(nd->nd_nam, struct sockaddr_in *);
+	switch (saddr->sin_family) {
+	case AF_INET:
+		rp->rc_flag |= RC_INETADDR;
+		rp->rc_inetaddr = saddr->sin_addr.s_addr;
+		break;
+	case AF_ISO:
+	default:
+		rp->rc_flag |= RC_NAM;
+		rp->rc_nam = m_copym(nd->nd_nam, 0, M_COPYALL, M_WAIT);
+		break;
+	};
+	rp->rc_proc = nd->nd_procnum;
+	LIST_INSERT_HEAD(NFSRCHASH(nd->nd_retxid), rp, rc_hash);
+	rp->rc_flag &= ~RC_LOCKED;
+	if (rp->rc_flag & RC_WANTED) {
+		rp->rc_flag &= ~RC_WANTED;
+		wakeup((caddr_t)rp);
+	}
+	return (RC_DOIT);
+}
+
+/*
+ * Update a request cache entry after the rpc has been done
+ */
+void
+nfsrv_updatecache(nd, repvalid, repmbuf)
+	register struct nfsrv_descript *nd;
+	int repvalid;
+	struct mbuf *repmbuf;
+{
+	register struct nfsrvcache *rp;
+
+	if (!nd->nd_nam2)
+		return;
+loop:
+	for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0;
+	    rp = rp->rc_hash.le_next) {
+	    if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc &&
+		netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) {
+			if ((rp->rc_flag & RC_LOCKED) != 0) {
+				rp->rc_flag |= RC_WANTED;
+				(void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0);
+				goto loop;
+			}
+			rp->rc_flag |= RC_LOCKED;
+			rp->rc_state = RC_DONE;
+			/*
+			 * If we have a valid reply update status and save
+			 * the reply for non-idempotent rpc's.
+			 */
+			if (repvalid && nonidempotent[nd->nd_procnum]) {
+				if ((nd->nd_flag & ND_NFSV3) == 0 &&
+				  nfsv2_repstat[nfsv2_procid[nd->nd_procnum]]) {
+					rp->rc_status = nd->nd_repstat;
+					rp->rc_flag |= RC_REPSTATUS;
+				} else {
+					rp->rc_reply = m_copym(repmbuf,
+						0, M_COPYALL, M_WAIT);
+					rp->rc_flag |= RC_REPMBUF;
+				}
+			}
+			rp->rc_flag &= ~RC_LOCKED;
+			if (rp->rc_flag & RC_WANTED) {
+				rp->rc_flag &= ~RC_WANTED;
+				wakeup((caddr_t)rp);
+			}
+			return;
+		}
+	}
+}
+
+/*
+ * Clean out the cache. Called when the last nfsd terminates.
+ */
+void
+nfsrv_cleancache()
+{
+	register struct nfsrvcache *rp, *nextrp;
+
+	for (rp = nfsrvlruhead.tqh_first; rp != 0; rp = nextrp) {
+		nextrp = rp->rc_lru.tqe_next;
+		LIST_REMOVE(rp, rc_hash);
+		TAILQ_REMOVE(&nfsrvlruhead, rp, rc_lru);
+		free(rp, M_NFSD);
+	}
+	numnfsrvcache = 0;
+}
diff --git a/sys/nfs/nfs_subs.c b/sys/nfs/nfs_subs.c
new file mode 100644
index 000000000000..5211c5cdf09b
--- /dev/null
+++ b/sys/nfs/nfs_subs.c
@@ -0,0 +1,1851 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_subs.c	8.8 (Berkeley) 5/22/95
+ */
+
+
+/*
+ * These functions support the macros and help fiddle mbuf chains for
+ * the nfs op functions. They do things like create the rpc header and
+ * copy data between mbuf chains and uio lists.
+ */
+#include <sys/param.h>
+#include <sys/proc.h>
+#include <sys/systm.h>
+#include <sys/kernel.h>
+#include <sys/mount.h>
+#include <sys/vnode.h>
+#include <sys/namei.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/stat.h>
+#include <sys/malloc.h>
+#ifdef VFS_LKM
+#include <sys/sysent.h>
+#include <sys/syscall.h>
+#endif
+
+#include <vm/vm.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfs.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nfsmount.h>
+#include <nfs/nqnfs.h>
+#include <nfs/nfsrtt.h>
+
+#include <miscfs/specfs/specdev.h>
+
+#include <netinet/in.h>
+#ifdef ISO
+#include <netiso/iso.h>
+#endif
+
+/*
+ * Data items converted to xdr at startup, since they are constant
+ * This is kinda hokey, but may save a little time doing byte swaps
+ */
+u_long nfs_xdrneg1;
+u_long rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
+	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
+	rpc_auth_kerb;
+u_long nfs_prog, nqnfs_prog, nfs_true, nfs_false;
+
+/* And other global data */
+static u_long nfs_xid = 0;
+enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
+enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
+int nfs_mount_type;
+int nfs_ticks;
+
+/*
+ * Mapping of old NFS Version 2 RPC numbers to generic numbers.
+ */
+int nfsv3_procid[NFS_NPROCS] = {
+	NFSPROC_NULL,
+	NFSPROC_GETATTR,
+	NFSPROC_SETATTR,
+	NFSPROC_NOOP,
+	NFSPROC_LOOKUP,
+	NFSPROC_READLINK,
+	NFSPROC_READ,
+	NFSPROC_NOOP,
+	NFSPROC_WRITE,
+	NFSPROC_CREATE,
+	NFSPROC_REMOVE,
+	NFSPROC_RENAME,
+	NFSPROC_LINK,
+	NFSPROC_SYMLINK,
+	NFSPROC_MKDIR,
+	NFSPROC_RMDIR,
+	NFSPROC_READDIR,
+	NFSPROC_FSSTAT,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP,
+	NFSPROC_NOOP
+};
+
+/*
+ * and the reverse mapping from generic to Version 2 procedure numbers
+ */
+int nfsv2_procid[NFS_NPROCS] = {
+	NFSV2PROC_NULL,
+	NFSV2PROC_GETATTR,
+	NFSV2PROC_SETATTR,
+	NFSV2PROC_LOOKUP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_READLINK,
+	NFSV2PROC_READ,
+	NFSV2PROC_WRITE,
+	NFSV2PROC_CREATE,
+	NFSV2PROC_MKDIR,
+	NFSV2PROC_SYMLINK,
+	NFSV2PROC_CREATE,
+	NFSV2PROC_REMOVE,
+	NFSV2PROC_RMDIR,
+	NFSV2PROC_RENAME,
+	NFSV2PROC_LINK,
+	NFSV2PROC_READDIR,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_STATFS,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+	NFSV2PROC_NOOP,
+};
+
+/*
+ * Maps errno values to nfs error numbers.
+ * Use NFSERR_IO as the catch all for ones not specifically defined in
+ * RFC 1094.
+ */
+static u_char nfsrv_v2errmap[ELAST] = {
+  NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
+  NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
+  NFSERR_IO,
+};
+
+/*
+ * Maps errno values to nfs error numbers.
+ * Although it is not obvious whether or not NFS clients really care if
+ * a returned error value is in the specified list for the procedure, the
+ * safest thing to do is filter them appropriately. For Version 2, the
+ * X/Open XNFS document is the only specification that defines error values
+ * for each RPC (The RFC simply lists all possible error values for all RPCs),
+ * so I have decided to not do this for Version 2.
+ * The first entry is the default error return and the rest are the valid
+ * errors for that RPC in increasing numeric order.
+ */
+static short nfsv3err_null[] = {
+	0,
+	0,
+};
+
+static short nfsv3err_getattr[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_setattr[] = {
+	NFSERR_IO,
+	NFSERR_PERM,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_INVAL,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOT_SYNC,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_lookup[] = {
+	NFSERR_IO,
+	NFSERR_NOENT,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_NOTDIR,
+	NFSERR_NAMETOL,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_access[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_readlink[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_INVAL,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_read[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_NXIO,
+	NFSERR_ACCES,
+	NFSERR_INVAL,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_write[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_INVAL,
+	NFSERR_FBIG,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_create[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_NOTDIR,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_mkdir[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_NOTDIR,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_symlink[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_NOTDIR,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_mknod[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_NOTDIR,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	NFSERR_BADTYPE,
+	0,
+};
+
+static short nfsv3err_remove[] = {
+	NFSERR_IO,
+	NFSERR_NOENT,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_NOTDIR,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_rmdir[] = {
+	NFSERR_IO,
+	NFSERR_NOENT,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_NOTDIR,
+	NFSERR_INVAL,
+	NFSERR_ROFS,
+	NFSERR_NAMETOL,
+	NFSERR_NOTEMPTY,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_rename[] = {
+	NFSERR_IO,
+	NFSERR_NOENT,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_XDEV,
+	NFSERR_NOTDIR,
+	NFSERR_ISDIR,
+	NFSERR_INVAL,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_MLINK,
+	NFSERR_NAMETOL,
+	NFSERR_NOTEMPTY,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_link[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_EXIST,
+	NFSERR_XDEV,
+	NFSERR_NOTDIR,
+	NFSERR_INVAL,
+	NFSERR_NOSPC,
+	NFSERR_ROFS,
+	NFSERR_MLINK,
+	NFSERR_NAMETOL,
+	NFSERR_DQUOT,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_NOTSUPP,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_readdir[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_NOTDIR,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_BAD_COOKIE,
+	NFSERR_TOOSMALL,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_readdirplus[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_ACCES,
+	NFSERR_NOTDIR,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_BAD_COOKIE,
+	NFSERR_NOTSUPP,
+	NFSERR_TOOSMALL,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_fsstat[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_fsinfo[] = {
+	NFSERR_STALE,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_pathconf[] = {
+	NFSERR_STALE,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short nfsv3err_commit[] = {
+	NFSERR_IO,
+	NFSERR_IO,
+	NFSERR_STALE,
+	NFSERR_BADHANDLE,
+	NFSERR_SERVERFAULT,
+	0,
+};
+
+static short *nfsrv_v3errmap[] = {
+	nfsv3err_null,
+	nfsv3err_getattr,
+	nfsv3err_setattr,
+	nfsv3err_lookup,
+	nfsv3err_access,
+	nfsv3err_readlink,
+	nfsv3err_read,
+	nfsv3err_write,
+	nfsv3err_create,
+	nfsv3err_mkdir,
+	nfsv3err_symlink,
+	nfsv3err_mknod,
+	nfsv3err_remove,
+	nfsv3err_rmdir,
+	nfsv3err_rename,
+	nfsv3err_link,
+	nfsv3err_readdir,
+	nfsv3err_readdirplus,
+	nfsv3err_fsstat,
+	nfsv3err_fsinfo,
+	nfsv3err_pathconf,
+	nfsv3err_commit,
+};
+
+extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON];
+extern struct nfsrtt nfsrtt;
+extern time_t nqnfsstarttime;
+extern int nqsrv_clockskew;
+extern int nqsrv_writeslack;
+extern int nqsrv_maxlease;
+extern struct nfsstats nfsstats;
+extern int nqnfs_piggy[NFS_NPROCS];
+extern nfstype nfsv2_type[9];
+extern nfstype nfsv3_type[9];
+extern struct nfsnodehashhead *nfsnodehashtbl;
+extern u_long nfsnodehash;
+
+#ifdef VFS_LKM
+struct getfh_args;
+extern int getfh(struct proc *, struct getfh_args *, int *);
+struct nfssvc_args;
+extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
+#endif
+
+LIST_HEAD(nfsnodehashhead, nfsnode);
+
+/*
+ * Create the header for an rpc request packet
+ * The hsiz is the size of the rest of the nfs request header.
+ * (just used to decide if a cluster is a good idea)
+ */
+struct mbuf *
+nfsm_reqh(vp, procid, hsiz, bposp)
+	struct vnode *vp;
+	u_long procid;
+	int hsiz;
+	caddr_t *bposp;
+{
+	register struct mbuf *mb;
+	register u_long *tl;
+	register caddr_t bpos;
+	struct mbuf *mb2;
+	struct nfsmount *nmp;
+	int nqflag;
+
+	MGET(mb, M_WAIT, MT_DATA);
+	if (hsiz >= MINCLSIZE)
+		MCLGET(mb, M_WAIT);
+	mb->m_len = 0;
+	bpos = mtod(mb, caddr_t);
+	
+	/*
+	 * For NQNFS, add lease request.
+	 */
+	if (vp) {
+		nmp = VFSTONFS(vp->v_mount);
+		if (nmp->nm_flag & NFSMNT_NQNFS) {
+			nqflag = NQNFS_NEEDLEASE(vp, procid);
+			if (nqflag) {
+				nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED);
+				*tl++ = txdr_unsigned(nqflag);
+				*tl = txdr_unsigned(nmp->nm_leaseterm);
+			} else {
+				nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+				*tl = 0;
+			}
+		}
+	}
+	/* Finally, return values */
+	*bposp = bpos;
+	return (mb);
+}
+
+/*
+ * Build the RPC header and fill in the authorization info.
+ * The authorization string argument is only used when the credentials
+ * come from outside of the kernel.
+ * Returns the head of the mbuf list.
+ */
+struct mbuf *
+nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
+	verf_str, mrest, mrest_len, mbp, xidp)
+	register struct ucred *cr;
+	int nmflag;
+	int procid;
+	int auth_type;
+	int auth_len;
+	char *auth_str;
+	int verf_len;
+	char *verf_str;
+	struct mbuf *mrest;
+	int mrest_len;
+	struct mbuf **mbp;
+	u_long *xidp;
+{
+	register struct mbuf *mb;
+	register u_long *tl;
+	register caddr_t bpos;
+	register int i;
+	struct mbuf *mreq, *mb2;
+	int siz, grpsiz, authsiz;
+
+	authsiz = nfsm_rndup(auth_len);
+	MGETHDR(mb, M_WAIT, MT_DATA);
+	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
+		MCLGET(mb, M_WAIT);
+	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
+		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
+	} else {
+		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
+	}
+	mb->m_len = 0;
+	mreq = mb;
+	bpos = mtod(mb, caddr_t);
+
+	/*
+	 * First the RPC header.
+	 */
+	nfsm_build(tl, u_long *, 8 * NFSX_UNSIGNED);
+	if (++nfs_xid == 0)
+		nfs_xid++;
+	*tl++ = *xidp = txdr_unsigned(nfs_xid);
+	*tl++ = rpc_call;
+	*tl++ = rpc_vers;
+	if (nmflag & NFSMNT_NQNFS) {
+		*tl++ = txdr_unsigned(NQNFS_PROG);
+		*tl++ = txdr_unsigned(NQNFS_VER3);
+	} else {
+		*tl++ = txdr_unsigned(NFS_PROG);
+		if (nmflag & NFSMNT_NFSV3)
+			*tl++ = txdr_unsigned(NFS_VER3);
+		else
+			*tl++ = txdr_unsigned(NFS_VER2);
+	}
+	if (nmflag & NFSMNT_NFSV3)
+		*tl++ = txdr_unsigned(procid);
+	else
+		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
+
+	/*
+	 * And then the authorization cred.
+	 */
+	*tl++ = txdr_unsigned(auth_type);
+	*tl = txdr_unsigned(authsiz);
+	switch (auth_type) {
+	case RPCAUTH_UNIX:
+		nfsm_build(tl, u_long *, auth_len);
+		*tl++ = 0;		/* stamp ?? */
+		*tl++ = 0;		/* NULL hostname */
+		*tl++ = txdr_unsigned(cr->cr_uid);
+		*tl++ = txdr_unsigned(cr->cr_groups[0]);
+		grpsiz = (auth_len >> 2) - 5;
+		*tl++ = txdr_unsigned(grpsiz);
+		for (i = 1; i <= grpsiz; i++)
+			*tl++ = txdr_unsigned(cr->cr_groups[i]);
+		break;
+	case RPCAUTH_KERB4:
+		siz = auth_len;
+		while (siz > 0) {
+			if (M_TRAILINGSPACE(mb) == 0) {
+				MGET(mb2, M_WAIT, MT_DATA);
+				if (siz >= MINCLSIZE)
+					MCLGET(mb2, M_WAIT);
+				mb->m_next = mb2;
+				mb = mb2;
+				mb->m_len = 0;
+				bpos = mtod(mb, caddr_t);
+			}
+			i = min(siz, M_TRAILINGSPACE(mb));
+			bcopy(auth_str, bpos, i);
+			mb->m_len += i;
+			auth_str += i;
+			bpos += i;
+			siz -= i;
+		}
+		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
+			for (i = 0; i < siz; i++)
+				*bpos++ = '\0';
+			mb->m_len += siz;
+		}
+		break;
+	};
+
+	/*
+	 * And the verifier...
+	 */
+	nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+	if (verf_str) {
+		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
+		*tl = txdr_unsigned(verf_len);
+		siz = verf_len;
+		while (siz > 0) {
+			if (M_TRAILINGSPACE(mb) == 0) {
+				MGET(mb2, M_WAIT, MT_DATA);
+				if (siz >= MINCLSIZE)
+					MCLGET(mb2, M_WAIT);
+				mb->m_next = mb2;
+				mb = mb2;
+				mb->m_len = 0;
+				bpos = mtod(mb, caddr_t);
+			}
+			i = min(siz, M_TRAILINGSPACE(mb));
+			bcopy(verf_str, bpos, i);
+			mb->m_len += i;
+			verf_str += i;
+			bpos += i;
+			siz -= i;
+		}
+		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
+			for (i = 0; i < siz; i++)
+				*bpos++ = '\0';
+			mb->m_len += siz;
+		}
+	} else {
+		*tl++ = txdr_unsigned(RPCAUTH_NULL);
+		*tl = 0;
+	}
+	mb->m_next = mrest;
+	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
+	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
+	*mbp = mb;
+	return (mreq);
+}
+
+/*
+ * copies mbuf chain to the uio scatter/gather list
+ */
+int
+nfsm_mbuftouio(mrep, uiop, siz, dpos)
+	struct mbuf **mrep;
+	register struct uio *uiop;
+	int siz;
+	caddr_t *dpos;
+{
+	register char *mbufcp, *uiocp;
+	register int xfer, left, len;
+	register struct mbuf *mp;
+	long uiosiz, rem;
+	int error = 0;
+
+	mp = *mrep;
+	mbufcp = *dpos;
+	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
+	rem = nfsm_rndup(siz)-siz;
+	while (siz > 0) {
+		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
+			return (EFBIG);
+		left = uiop->uio_iov->iov_len;
+		uiocp = uiop->uio_iov->iov_base;
+		if (left > siz)
+			left = siz;
+		uiosiz = left;
+		while (left > 0) {
+			while (len == 0) {
+				mp = mp->m_next;
+				if (mp == NULL)
+					return (EBADRPC);
+				mbufcp = mtod(mp, caddr_t);
+				len = mp->m_len;
+			}
+			xfer = (left > len) ? len : left;
+#ifdef notdef
+			/* Not Yet.. */
+			if (uiop->uio_iov->iov_op != NULL)
+				(*(uiop->uio_iov->iov_op))
+				(mbufcp, uiocp, xfer);
+			else
+#endif
+			if (uiop->uio_segflg == UIO_SYSSPACE)
+				bcopy(mbufcp, uiocp, xfer);
+			else
+				copyout(mbufcp, uiocp, xfer);
+			left -= xfer;
+			len -= xfer;
+			mbufcp += xfer;
+			uiocp += xfer;
+			uiop->uio_offset += xfer;
+			uiop->uio_resid -= xfer;
+		}
+		if (uiop->uio_iov->iov_len <= siz) {
+			uiop->uio_iovcnt--;
+			uiop->uio_iov++;
+		} else {
+			uiop->uio_iov->iov_base += uiosiz;
+			uiop->uio_iov->iov_len -= uiosiz;
+		}
+		siz -= uiosiz;
+	}
+	*dpos = mbufcp;
+	*mrep = mp;
+	if (rem > 0) {
+		if (len < rem)
+			error = nfs_adv(mrep, dpos, rem, len);
+		else
+			*dpos += rem;
+	}
+	return (error);
+}
+
+/*
+ * copies a uio scatter/gather list to an mbuf chain...
+ */
+int
+nfsm_uiotombuf(uiop, mq, siz, bpos)
+	register struct uio *uiop;
+	struct mbuf **mq;
+	int siz;
+	caddr_t *bpos;
+{
+	register char *uiocp;
+	register struct mbuf *mp, *mp2;
+	register int xfer, left, mlen;
+	int uiosiz, clflg, rem;
+	char *cp;
+
+	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
+		clflg = 1;
+	else
+		clflg = 0;
+	rem = nfsm_rndup(siz)-siz;
+	mp = mp2 = *mq;
+	while (siz > 0) {
+		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
+			return (EINVAL);
+		left = uiop->uio_iov->iov_len;
+		uiocp = uiop->uio_iov->iov_base;
+		if (left > siz)
+			left = siz;
+		uiosiz = left;
+		while (left > 0) {
+			mlen = M_TRAILINGSPACE(mp);
+			if (mlen == 0) {
+				MGET(mp, M_WAIT, MT_DATA);
+				if (clflg)
+					MCLGET(mp, M_WAIT);
+				mp->m_len = 0;
+				mp2->m_next = mp;
+				mp2 = mp;
+				mlen = M_TRAILINGSPACE(mp);
+			}
+			xfer = (left > mlen) ? mlen : left;
+#ifdef notdef
+			/* Not Yet.. */
+			if (uiop->uio_iov->iov_op != NULL)
+				(*(uiop->uio_iov->iov_op))
+				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
+			else
+#endif
+			if (uiop->uio_segflg == UIO_SYSSPACE)
+				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
+			else
+				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
+			mp->m_len += xfer;
+			left -= xfer;
+			uiocp += xfer;
+			uiop->uio_offset += xfer;
+			uiop->uio_resid -= xfer;
+		}
+		if (uiop->uio_iov->iov_len <= siz) {
+			uiop->uio_iovcnt--;
+			uiop->uio_iov++;
+		} else {
+			uiop->uio_iov->iov_base += uiosiz;
+			uiop->uio_iov->iov_len -= uiosiz;
+		}
+		siz -= uiosiz;
+	}
+	if (rem > 0) {
+		if (rem > M_TRAILINGSPACE(mp)) {
+			MGET(mp, M_WAIT, MT_DATA);
+			mp->m_len = 0;
+			mp2->m_next = mp;
+		}
+		cp = mtod(mp, caddr_t)+mp->m_len;
+		for (left = 0; left < rem; left++)
+			*cp++ = '\0';
+		mp->m_len += rem;
+		*bpos = cp;
+	} else
+		*bpos = mtod(mp, caddr_t)+mp->m_len;
+	*mq = mp;
+	return (0);
+}
+
+/*
+ * Help break down an mbuf chain by setting the first siz bytes contiguous
+ * pointed to by returned val.
+ * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
+ * cases. (The macros use the vars. dpos and dpos2)
+ */
+int
+nfsm_disct(mdp, dposp, siz, left, cp2)
+	struct mbuf **mdp;
+	caddr_t *dposp;
+	int siz;
+	int left;
+	caddr_t *cp2;
+{
+	register struct mbuf *mp, *mp2;
+	register int siz2, xfer;
+	register caddr_t p;
+
+	mp = *mdp;
+	while (left == 0) {
+		*mdp = mp = mp->m_next;
+		if (mp == NULL)
+			return (EBADRPC);
+		left = mp->m_len;
+		*dposp = mtod(mp, caddr_t);
+	}
+	if (left >= siz) {
+		*cp2 = *dposp;
+		*dposp += siz;
+	} else if (mp->m_next == NULL) {
+		return (EBADRPC);
+	} else if (siz > MHLEN) {
+		panic("nfs S too big");
+	} else {
+		MGET(mp2, M_WAIT, MT_DATA);
+		mp2->m_next = mp->m_next;
+		mp->m_next = mp2;
+		mp->m_len -= left;
+		mp = mp2;
+		*cp2 = p = mtod(mp, caddr_t);
+		bcopy(*dposp, p, left);		/* Copy what was left */
+		siz2 = siz-left;
+		p += left;
+		mp2 = mp->m_next;
+		/* Loop around copying up the siz2 bytes */
+		while (siz2 > 0) {
+			if (mp2 == NULL)
+				return (EBADRPC);
+			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
+			if (xfer > 0) {
+				bcopy(mtod(mp2, caddr_t), p, xfer);
+				NFSMADV(mp2, xfer);
+				mp2->m_len -= xfer;
+				p += xfer;
+				siz2 -= xfer;
+			}
+			if (siz2 > 0)
+				mp2 = mp2->m_next;
+		}
+		mp->m_len = siz;
+		*mdp = mp2;
+		*dposp = mtod(mp2, caddr_t);
+	}
+	return (0);
+}
+
+/*
+ * Advance the position in the mbuf chain.
+ */
+int
+nfs_adv(mdp, dposp, offs, left)
+	struct mbuf **mdp;
+	caddr_t *dposp;
+	int offs;
+	int left;
+{
+	register struct mbuf *m;
+	register int s;
+
+	m = *mdp;
+	s = left;
+	while (s < offs) {
+		offs -= s;
+		m = m->m_next;
+		if (m == NULL)
+			return (EBADRPC);
+		s = m->m_len;
+	}
+	*mdp = m;
+	*dposp = mtod(m, caddr_t)+offs;
+	return (0);
+}
+
+/*
+ * Copy a string into mbufs for the hard cases...
+ */
+int
+nfsm_strtmbuf(mb, bpos, cp, siz)
+	struct mbuf **mb;
+	char **bpos;
+	char *cp;
+	long siz;
+{
+	register struct mbuf *m1 = 0, *m2;
+	long left, xfer, len, tlen;
+	u_long *tl;
+	int putsize;
+
+	putsize = 1;
+	m2 = *mb;
+	left = M_TRAILINGSPACE(m2);
+	if (left > 0) {
+		tl = ((u_long *)(*bpos));
+		*tl++ = txdr_unsigned(siz);
+		putsize = 0;
+		left -= NFSX_UNSIGNED;
+		m2->m_len += NFSX_UNSIGNED;
+		if (left > 0) {
+			bcopy(cp, (caddr_t) tl, left);
+			siz -= left;
+			cp += left;
+			m2->m_len += left;
+			left = 0;
+		}
+	}
+	/* Loop around adding mbufs */
+	while (siz > 0) {
+		MGET(m1, M_WAIT, MT_DATA);
+		if (siz > MLEN)
+			MCLGET(m1, M_WAIT);
+		m1->m_len = NFSMSIZ(m1);
+		m2->m_next = m1;
+		m2 = m1;
+		tl = mtod(m1, u_long *);
+		tlen = 0;
+		if (putsize) {
+			*tl++ = txdr_unsigned(siz);
+			m1->m_len -= NFSX_UNSIGNED;
+			tlen = NFSX_UNSIGNED;
+			putsize = 0;
+		}
+		if (siz < m1->m_len) {
+			len = nfsm_rndup(siz);
+			xfer = siz;
+			if (xfer < len)
+				*(tl+(xfer>>2)) = 0;
+		} else {
+			xfer = len = m1->m_len;
+		}
+		bcopy(cp, (caddr_t) tl, xfer);
+		m1->m_len = len+tlen;
+		siz -= xfer;
+		cp += xfer;
+	}
+	*mb = m1;
+	*bpos = mtod(m1, caddr_t)+m1->m_len;
+	return (0);
+}
+
+/*
+ * Called once to initialize data structures...
+ */
+int
+nfs_init(vfsp)
+	struct vfsconf *vfsp;
+{
+	register int i;
+
+	/*
+	 * Check to see if major data structures haven't bloated.
+	 */
+	if (sizeof (struct nfsnode) > NFS_NODEALLOC) {
+		printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC);
+		printf("Try reducing NFS_SMALLFH\n");
+	}
+	if (sizeof (struct nfsmount) > NFS_MNTALLOC) {
+		printf("struct nfsmount bloated (> %dbytes)\n", NFS_MNTALLOC);
+		printf("Try reducing NFS_MUIDHASHSIZ\n");
+	}
+	if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) {
+		printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC);
+		printf("Try reducing NFS_UIDHASHSIZ\n");
+	}
+	if (sizeof (struct nfsuid) > NFS_UIDALLOC) {
+		printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC);
+		printf("Try unionizing the nu_nickname and nu_flag fields\n");
+	}
+	nfs_mount_type = vfsp->vfc_typenum;
+	nfsrtt.pos = 0;
+	rpc_vers = txdr_unsigned(RPC_VER2);
+	rpc_call = txdr_unsigned(RPC_CALL);
+	rpc_reply = txdr_unsigned(RPC_REPLY);
+	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
+	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
+	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
+	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
+	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
+	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
+	nfs_prog = txdr_unsigned(NFS_PROG);
+	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
+	nfs_true = txdr_unsigned(TRUE);
+	nfs_false = txdr_unsigned(FALSE);
+	nfs_xdrneg1 = txdr_unsigned(-1);
+	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
+	if (nfs_ticks < 1)
+		nfs_ticks = 1;
+	/* Ensure async daemons disabled */
+	for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
+		nfs_iodwant[i] = (struct proc *)0;
+	TAILQ_INIT(&nfs_bufq);
+	nfs_nhinit();			/* Init the nfsnode table */
+	nfsrv_init(0);			/* Init server data structures */
+	nfsrv_initcache();		/* Init the server request cache */
+
+	/*
+	 * Initialize the nqnfs server stuff.
+	 */
+	if (nqnfsstarttime == 0) {
+		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
+			+ nqsrv_clockskew + nqsrv_writeslack;
+		NQLOADNOVRAM(nqnfsstarttime);
+		CIRCLEQ_INIT(&nqtimerhead);
+		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
+	}
+
+	/*
+	 * Initialize reply list and start timer
+	 */
+	TAILQ_INIT(&nfs_reqq);
+	nfs_timer(0);
+	return (0);
+}
+
+/*
+ * Attribute cache routines.
+ * nfs_loadattrcache() - loads or updates the cache contents from attributes
+ *	that are on the mbuf list
+ * nfs_getattrcache() - returns valid attributes if found in cache, returns
+ *	error otherwise
+ */
+
+/*
+ * Load the attribute cache (that lives in the nfsnode entry) with
+ * the values on the mbuf list and
+ * Iff vap not NULL
+ *    copy the attributes to *vaper
+ */
+int
+nfs_loadattrcache(vpp, mdp, dposp, vaper)
+	struct vnode **vpp;
+	struct mbuf **mdp;
+	caddr_t *dposp;
+	struct vattr *vaper;
+{
+	register struct vnode *vp = *vpp;
+	register struct vattr *vap;
+	register struct nfs_fattr *fp;
+	extern int (**spec_nfsv2nodeop_p)();
+	register struct nfsnode *np;
+	register struct nfsnodehashhead *nhpp;
+	register long t1;
+	caddr_t cp2;
+	int error = 0, rdev;
+	struct mbuf *md;
+	enum vtype vtyp;
+	u_short vmode;
+	struct timespec mtime;
+	struct vnode *nvp;
+	quad_t tval;
+	int v3 = NFS_ISV3(vp);
+
+	md = *mdp;
+	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
+	if (error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2))
+		return (error);
+	fp = (struct nfs_fattr *)cp2;
+	if (v3) {
+		vtyp = nfsv3tov_type(fp->fa_type);
+		vmode = fxdr_unsigned(u_short, fp->fa_mode);
+		rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1),
+			fxdr_unsigned(u_char, fp->fa3_rdev.specdata2));
+		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
+	} else {
+		vtyp = nfsv2tov_type(fp->fa_type);
+		vmode = fxdr_unsigned(u_short, fp->fa_mode);
+		if (vtyp == VNON || vtyp == VREG)
+			vtyp = IFTOVT(vmode);
+		rdev = fxdr_unsigned(long, fp->fa2_rdev);
+		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
+
+		/*
+		 * Really ugly NFSv2 kludge.
+		 */
+		if (vtyp == VCHR && rdev == 0xffffffff)
+			vtyp = VFIFO;
+	}
+
+	/*
+	 * If v_type == VNON it is a new node, so fill in the v_type,
+	 * n_mtime fields. Check to see if it represents a special 
+	 * device, and if so, check for a possible alias. Once the
+	 * correct vnode has been obtained, fill in the rest of the
+	 * information.
+	 */
+	np = VTONFS(vp);
+	if (vp->v_type == VNON) {
+		vp->v_type = vtyp;
+		if (vp->v_type == VFIFO) {
+			extern int (**fifo_nfsv2nodeop_p)();
+			vp->v_op = fifo_nfsv2nodeop_p;
+		}
+		if (vp->v_type == VCHR || vp->v_type == VBLK) {
+			vp->v_op = spec_nfsv2nodeop_p;
+			nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
+			if (nvp) {
+				/*
+				 * Discard unneeded vnode, but save its nfsnode.
+				 * Since the nfsnode does not have a lock, its
+				 * vnode lock has to be carried over.
+				 */
+				nvp->v_vnlock = vp->v_vnlock;
+				vp->v_vnlock = NULL;
+				nvp->v_data = vp->v_data;
+				vp->v_data = NULL;
+				vp->v_op = spec_vnodeop_p;
+				vrele(vp);
+				vgone(vp);
+				/*
+				 * Reinitialize aliased node.
+				 */
+				np->n_vnode = nvp;
+				*vpp = vp = nvp;
+			}
+		}
+		np->n_mtime = mtime.ts_sec;
+	}
+	vap = &np->n_vattr;
+	vap->va_type = vtyp;
+	vap->va_mode = (vmode & 07777);
+	vap->va_rdev = (dev_t)rdev;
+	vap->va_mtime = mtime;
+	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
+	if (v3) {
+		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
+		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
+		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
+		fxdr_hyper(&fp->fa3_size, &vap->va_size);
+		vap->va_blocksize = NFS_FABLKSIZE;
+		fxdr_hyper(&fp->fa3_used, &vap->va_bytes);
+		vap->va_fileid = fxdr_unsigned(int, fp->fa3_fileid.nfsuquad[1]);
+		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
+		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
+		vap->va_flags = 0;
+		vap->va_filerev = 0;
+	} else {
+		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
+		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
+		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
+		vap->va_size = fxdr_unsigned(u_long, fp->fa2_size);
+		vap->va_blocksize = fxdr_unsigned(long, fp->fa2_blocksize);
+		vap->va_bytes = fxdr_unsigned(long, fp->fa2_blocks) * NFS_FABLKSIZE;
+		vap->va_fileid = fxdr_unsigned(long, fp->fa2_fileid);
+		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
+		vap->va_flags = 0;
+		vap->va_ctime.ts_sec = fxdr_unsigned(long, fp->fa2_ctime.nfsv2_sec);
+		vap->va_ctime.ts_nsec = 0;
+		vap->va_gen = fxdr_unsigned(u_long, fp->fa2_ctime.nfsv2_usec);
+		vap->va_filerev = 0;
+	}
+	if (vap->va_size != np->n_size) {
+		if (vap->va_type == VREG) {
+			if (np->n_flag & NMODIFIED) {
+				if (vap->va_size < np->n_size)
+					vap->va_size = np->n_size;
+				else
+					np->n_size = vap->va_size;
+			} else
+				np->n_size = vap->va_size;
+			vnode_pager_setsize(vp, (u_long)np->n_size);
+		} else
+			np->n_size = vap->va_size;
+	}
+	np->n_attrstamp = time.tv_sec;
+	if (vaper != NULL) {
+		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
+		if (np->n_flag & NCHG) {
+			if (np->n_flag & NACC)
+				vaper->va_atime = np->n_atim;
+			if (np->n_flag & NUPD)
+				vaper->va_mtime = np->n_mtim;
+		}
+	}
+	return (0);
+}
+
+/*
+ * Check the time stamp
+ * If the cache is valid, copy contents to *vap and return 0
+ * otherwise return an error
+ */
+int
+nfs_getattrcache(vp, vaper)
+	register struct vnode *vp;
+	struct vattr *vaper;
+{
+	register struct nfsnode *np = VTONFS(vp);
+	register struct vattr *vap;
+
+	if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
+		nfsstats.attrcache_misses++;
+		return (ENOENT);
+	}
+	nfsstats.attrcache_hits++;
+	vap = &np->n_vattr;
+	if (vap->va_size != np->n_size) {
+		if (vap->va_type == VREG) {
+			if (np->n_flag & NMODIFIED) {
+				if (vap->va_size < np->n_size)
+					vap->va_size = np->n_size;
+				else
+					np->n_size = vap->va_size;
+			} else
+				np->n_size = vap->va_size;
+			vnode_pager_setsize(vp, (u_long)np->n_size);
+		} else
+			np->n_size = vap->va_size;
+	}
+	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
+	if (np->n_flag & NCHG) {
+		if (np->n_flag & NACC)
+			vaper->va_atime = np->n_atim;
+		if (np->n_flag & NUPD)
+			vaper->va_mtime = np->n_mtim;
+	}
+	return (0);
+}
+
+/*
+ * Set up nameidata for a lookup() call and do it
+ */
+int
+nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag)
+	register struct nameidata *ndp;
+	fhandle_t *fhp;
+	int len;
+	struct nfssvc_sock *slp;
+	struct mbuf *nam;
+	struct mbuf **mdp;
+	caddr_t *dposp;
+	struct vnode **retdirp;
+	struct proc *p;
+	int kerbflag;
+{
+	register int i, rem;
+	register struct mbuf *md;
+	register char *fromcp, *tocp;
+	struct vnode *dp;
+	int error, rdonly;
+	struct componentname *cnp = &ndp->ni_cnd;
+
+	*retdirp = (struct vnode *)0;
+	MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK);
+	/*
+	 * Copy the name from the mbuf list to ndp->ni_pnbuf
+	 * and set the various ndp fields appropriately.
+	 */
+	fromcp = *dposp;
+	tocp = cnp->cn_pnbuf;
+	md = *mdp;
+	rem = mtod(md, caddr_t) + md->m_len - fromcp;
+	cnp->cn_hash = 0;
+	for (i = 0; i < len; i++) {
+		while (rem == 0) {
+			md = md->m_next;
+			if (md == NULL) {
+				error = EBADRPC;
+				goto out;
+			}
+			fromcp = mtod(md, caddr_t);
+			rem = md->m_len;
+		}
+		if (*fromcp == '\0' || *fromcp == '/') {
+			error = EACCES;
+			goto out;
+		}
+		cnp->cn_hash += (unsigned char)*fromcp;
+		*tocp++ = *fromcp++;
+		rem--;
+	}
+	*tocp = '\0';
+	*mdp = md;
+	*dposp = fromcp;
+	len = nfsm_rndup(len)-len;
+	if (len > 0) {
+		if (rem >= len)
+			*dposp += len;
+		else if (error = nfs_adv(mdp, dposp, len, rem))
+			goto out;
+	}
+	ndp->ni_pathlen = tocp - cnp->cn_pnbuf;
+	cnp->cn_nameptr = cnp->cn_pnbuf;
+	/*
+	 * Extract and set starting directory.
+	 */
+	if (error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
+	    nam, &rdonly, kerbflag))
+		goto out;
+	if (dp->v_type != VDIR) {
+		vrele(dp);
+		error = ENOTDIR;
+		goto out;
+	}
+	VREF(dp);
+	*retdirp = dp;
+	ndp->ni_startdir = dp;
+	if (rdonly)
+		cnp->cn_flags |= (NOCROSSMOUNT | RDONLY);
+	else
+		cnp->cn_flags |= NOCROSSMOUNT;
+	/*
+	 * And call lookup() to do the real work
+	 */
+	cnp->cn_proc = p;
+	if (error = lookup(ndp))
+		goto out;
+	/*
+	 * Check for encountering a symbolic link
+	 */
+	if (cnp->cn_flags & ISSYMLINK) {
+		if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
+			vput(ndp->ni_dvp);
+		else
+			vrele(ndp->ni_dvp);
+		vput(ndp->ni_vp);
+		ndp->ni_vp = NULL;
+		error = EINVAL;
+		goto out;
+	}
+	/*
+	 * Check for saved name request
+	 */
+	if (cnp->cn_flags & (SAVENAME | SAVESTART)) {
+		cnp->cn_flags |= HASBUF;
+		return (0);
+	}
+out:
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	return (error);
+}
+
+/*
+ * A fiddled version of m_adj() that ensures null fill to a long
+ * boundary and only trims off the back end
+ */
+void
+nfsm_adj(mp, len, nul)
+	struct mbuf *mp;
+	register int len;
+	int nul;
+{
+	register struct mbuf *m;
+	register int count, i;
+	register char *cp;
+
+	/*
+	 * Trim from tail.  Scan the mbuf chain,
+	 * calculating its length and finding the last mbuf.
+	 * If the adjustment only affects this mbuf, then just
+	 * adjust and return.  Otherwise, rescan and truncate
+	 * after the remaining size.
+	 */
+	count = 0;
+	m = mp;
+	for (;;) {
+		count += m->m_len;
+		if (m->m_next == (struct mbuf *)0)
+			break;
+		m = m->m_next;
+	}
+	if (m->m_len > len) {
+		m->m_len -= len;
+		if (nul > 0) {
+			cp = mtod(m, caddr_t)+m->m_len-nul;
+			for (i = 0; i < nul; i++)
+				*cp++ = '\0';
+		}
+		return;
+	}
+	count -= len;
+	if (count < 0)
+		count = 0;
+	/*
+	 * Correct length for chain is "count".
+	 * Find the mbuf with last data, adjust its length,
+	 * and toss data from remaining mbufs on chain.
+	 */
+	for (m = mp; m; m = m->m_next) {
+		if (m->m_len >= count) {
+			m->m_len = count;
+			if (nul > 0) {
+				cp = mtod(m, caddr_t)+m->m_len-nul;
+				for (i = 0; i < nul; i++)
+					*cp++ = '\0';
+			}
+			break;
+		}
+		count -= m->m_len;
+	}
+	for (m = m->m_next;m;m = m->m_next)
+		m->m_len = 0;
+}
+
+/*
+ * Make these functions instead of macros, so that the kernel text size
+ * doesn't get too big...
+ */
+void
+nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
+	struct nfsrv_descript *nfsd;
+	int before_ret;
+	register struct vattr *before_vap;
+	int after_ret;
+	struct vattr *after_vap;
+	struct mbuf **mbp;
+	char **bposp;
+{
+	register struct mbuf *mb = *mbp, *mb2;
+	register char *bpos = *bposp;
+	register u_long *tl;
+
+	if (before_ret) {
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+		*tl = nfs_false;
+	} else {
+		nfsm_build(tl, u_long *, 7 * NFSX_UNSIGNED);
+		*tl++ = nfs_true;
+		txdr_hyper(&(before_vap->va_size), tl);
+		tl += 2;
+		txdr_nfsv3time(&(before_vap->va_mtime), tl);
+		tl += 2;
+		txdr_nfsv3time(&(before_vap->va_ctime), tl);
+	}
+	*bposp = bpos;
+	*mbp = mb;
+	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
+}
+
+void
+nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
+	struct nfsrv_descript *nfsd;
+	int after_ret;
+	struct vattr *after_vap;
+	struct mbuf **mbp;
+	char **bposp;
+{
+	register struct mbuf *mb = *mbp, *mb2;
+	register char *bpos = *bposp;
+	register u_long *tl;
+	register struct nfs_fattr *fp;
+
+	if (after_ret) {
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+		*tl = nfs_false;
+	} else {
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FATTR);
+		*tl++ = nfs_true;
+		fp = (struct nfs_fattr *)tl;
+		nfsm_srvfattr(nfsd, after_vap, fp);
+	}
+	*mbp = mb;
+	*bposp = bpos;
+}
+
+void
+nfsm_srvfattr(nfsd, vap, fp)
+	register struct nfsrv_descript *nfsd;
+	register struct vattr *vap;
+	register struct nfs_fattr *fp;
+{
+
+	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
+	fp->fa_uid = txdr_unsigned(vap->va_uid);
+	fp->fa_gid = txdr_unsigned(vap->va_gid);
+	if (nfsd->nd_flag & ND_NFSV3) {
+		fp->fa_type = vtonfsv3_type(vap->va_type);
+		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
+		txdr_hyper(&vap->va_size, &fp->fa3_size);
+		txdr_hyper(&vap->va_bytes, &fp->fa3_used);
+		fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
+		fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
+		fp->fa3_fsid.nfsuquad[0] = 0;
+		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
+		fp->fa3_fileid.nfsuquad[0] = 0;
+		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
+		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
+		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
+		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
+	} else {
+		fp->fa_type = vtonfsv2_type(vap->va_type);
+		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
+		fp->fa2_size = txdr_unsigned(vap->va_size);
+		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
+		if (vap->va_type == VFIFO)
+			fp->fa2_rdev = 0xffffffff;
+		else
+			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
+		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
+		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
+		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
+		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
+		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
+		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
+	}
+}
+
+/*
+ * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
+ * 	- look up fsid in mount list (if not found ret error)
+ *	- get vp and export rights by calling VFS_FHTOVP()
+ *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
+ *	- if not lockflag unlock it with VOP_UNLOCK()
+ */
+int
+nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag)
+	fhandle_t *fhp;
+	int lockflag;
+	struct vnode **vpp;
+	struct ucred *cred;
+	struct nfssvc_sock *slp;
+	struct mbuf *nam;
+	int *rdonlyp;
+	int kerbflag;
+{
+	struct proc *p = curproc;	/* XXX */
+	register struct mount *mp;
+	register struct nfsuid *uidp;
+	register int i;
+	struct ucred *credanon;
+	int error, exflags;
+
+	*vpp = (struct vnode *)0;
+	mp = vfs_getvfs(&fhp->fh_fsid);
+	if (!mp)
+		return (ESTALE);
+	error = VFS_FHTOVP(mp, &fhp->fh_fid, nam, vpp, &exflags, &credanon);
+	if (error)
+		return (error);
+	/*
+	 * Check/setup credentials.
+	 */
+	if (exflags & MNT_EXKERB) {
+		if (!kerbflag) {
+			vput(*vpp);
+			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
+		}
+	} else if (kerbflag) {
+		vput(*vpp);
+		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
+	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
+		cred->cr_uid = credanon->cr_uid;
+		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
+			cred->cr_groups[i] = credanon->cr_groups[i];
+		cred->cr_ngroups = i;
+	}
+	if (exflags & MNT_EXRDONLY)
+		*rdonlyp = 1;
+	else
+		*rdonlyp = 0;
+	if (!lockflag)
+		VOP_UNLOCK(*vpp, 0, p);
+	return (0);
+}
+
+/*
+ * This function compares two net addresses by family and returns TRUE
+ * if they are the same host.
+ * If there is any doubt, return FALSE.
+ * The AF_INET family is handled as a special case so that address mbufs
+ * don't need to be saved to store "struct in_addr", which is only 4 bytes.
+ */
+int
+netaddr_match(family, haddr, nam)
+	int family;
+	union nethostaddr *haddr;
+	struct mbuf *nam;
+{
+	register struct sockaddr_in *inetaddr;
+
+	switch (family) {
+	case AF_INET:
+		inetaddr = mtod(nam, struct sockaddr_in *);
+		if (inetaddr->sin_family == AF_INET &&
+		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
+			return (1);
+		break;
+#ifdef ISO
+	case AF_ISO:
+	    {
+		register struct sockaddr_iso *isoaddr1, *isoaddr2;
+
+		isoaddr1 = mtod(nam, struct sockaddr_iso *);
+		isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
+		if (isoaddr1->siso_family == AF_ISO &&
+		    isoaddr1->siso_nlen > 0 &&
+		    isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
+		    SAME_ISOADDR(isoaddr1, isoaddr2))
+			return (1);
+		break;
+	    }
+#endif	/* ISO */
+	default:
+		break;
+	};
+	return (0);
+}
+
+static nfsuint64 nfs_nullcookie = { 0, 0 };
+/*
+ * This function finds the directory cookie that corresponds to the
+ * logical byte offset given.
+ */
+nfsuint64 *
+nfs_getcookie(np, off, add)
+	register struct nfsnode *np;
+	off_t off;
+	int add;
+{
+	register struct nfsdmap *dp, *dp2;
+	register int pos;
+
+	pos = off / NFS_DIRBLKSIZ;
+	if (pos == 0) {
+#ifdef DIAGNOSTIC
+		if (add)
+			panic("nfs getcookie add at 0");
+#endif
+		return (&nfs_nullcookie);
+	}
+	pos--;
+	dp = np->n_cookies.lh_first;
+	if (!dp) {
+		if (add) {
+			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
+				M_NFSDIROFF, M_WAITOK);
+			dp->ndm_eocookie = 0;
+			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
+		} else
+			return ((nfsuint64 *)0);
+	}
+	while (pos >= NFSNUMCOOKIES) {
+		pos -= NFSNUMCOOKIES;
+		if (dp->ndm_list.le_next) {
+			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
+				pos >= dp->ndm_eocookie)
+				return ((nfsuint64 *)0);
+			dp = dp->ndm_list.le_next;
+		} else if (add) {
+			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
+				M_NFSDIROFF, M_WAITOK);
+			dp2->ndm_eocookie = 0;
+			LIST_INSERT_AFTER(dp, dp2, ndm_list);
+			dp = dp2;
+		} else
+			return ((nfsuint64 *)0);
+	}
+	if (pos >= dp->ndm_eocookie) {
+		if (add)
+			dp->ndm_eocookie = pos + 1;
+		else
+			return ((nfsuint64 *)0);
+	}
+	return (&dp->ndm_cookies[pos]);
+}
+
+/*
+ * Invalidate cached directory information, except for the actual directory
+ * blocks (which are invalidated separately).
+ * Done mainly to avoid the use of stale offset cookies.
+ */
+void
+nfs_invaldir(vp)
+	register struct vnode *vp;
+{
+	register struct nfsnode *np = VTONFS(vp);
+
+#ifdef DIAGNOSTIC
+	if (vp->v_type != VDIR)
+		panic("nfs: invaldir not dir");
+#endif
+	np->n_direofoffset = 0;
+	np->n_cookieverf.nfsuquad[0] = 0;
+	np->n_cookieverf.nfsuquad[1] = 0;
+	if (np->n_cookies.lh_first)
+		np->n_cookies.lh_first->ndm_eocookie = 0;
+}
+
+/*
+ * The write verifier has changed (probably due to a server reboot), so all
+ * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
+ * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
+ * flag. Once done the new write verifier can be set for the mount point.
+ */
+void
+nfs_clearcommit(mp)
+	struct mount *mp;
+{
+	register struct vnode *vp, *nvp;
+	register struct buf *bp, *nbp;
+	int s;
+
+	s = splbio();
+loop:
+	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
+		if (vp->v_mount != mp)	/* Paranoia */
+			goto loop;
+		nvp = vp->v_mntvnodes.le_next;
+		for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
+			nbp = bp->b_vnbufs.le_next;
+			if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT))
+				== (B_DELWRI | B_NEEDCOMMIT))
+				bp->b_flags &= ~B_NEEDCOMMIT;
+		}
+	}
+	splx(s);
+}
+
+/*
+ * Map errnos to NFS error numbers. For Version 3 also filter out error
+ * numbers not specified for the associated procedure.
+ */
+int
+nfsrv_errmap(nd, err)
+	struct nfsrv_descript *nd;
+	register int err;
+{
+	register short *defaulterrp, *errp;
+
+	if (nd->nd_flag & ND_NFSV3) {
+	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
+		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
+		while (*++errp) {
+			if (*errp == err)
+				return (err);
+			else if (*errp > err)
+				break;
+		}
+		return ((int)*defaulterrp);
+	    } else
+		return (err & 0xffff);
+	}
+	if (err <= ELAST)
+		return ((int)nfsrv_v2errmap[err - 1]);
+	return (NFSERR_IO);
+}
diff --git a/sys/nfs/nfs_syscalls.c b/sys/nfs/nfs_syscalls.c
new file mode 100644
index 000000000000..aac38e4f667f
--- /dev/null
+++ b/sys/nfs/nfs_syscalls.c
@@ -0,0 +1,1125 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_syscalls.c	8.5 (Berkeley) 3/30/95
+ */
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/kernel.h>
+#include <sys/file.h>
+#include <sys/stat.h>
+#include <sys/vnode.h>
+#include <sys/mount.h>
+#include <sys/proc.h>
+#include <sys/uio.h>
+#include <sys/malloc.h>
+#include <sys/buf.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+#include <sys/domain.h>
+#include <sys/protosw.h>
+#include <sys/namei.h>
+#include <sys/syslog.h>
+
+#include <netinet/in.h>
+#include <netinet/tcp.h>
+#ifdef ISO
+#include <netiso/iso.h>
+#endif
+#include <nfs/xdr_subs.h>
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nfsrvcache.h>
+#include <nfs/nfsmount.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nqnfs.h>
+#include <nfs/nfsrtt.h>
+
+void	nfsrv_zapsock	__P((struct nfssvc_sock *));
+
+/* Global defs. */
+extern int (*nfsrv3_procs[NFS_NPROCS])();
+extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON];
+extern int nfs_numasync;
+extern time_t nqnfsstarttime;
+extern int nqsrv_writeslack;
+extern int nfsrtton;
+extern struct nfsstats nfsstats;
+extern int nfsrvw_procrastinate;
+struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock;
+int nuidhash_max = NFS_MAXUIDHASH;
+static int nfs_numnfsd = 0;
+int nfsd_waiting = 0;
+static int notstarted = 1;
+static int modify_flag = 0;
+static struct nfsdrt nfsdrt;
+void nfsrv_cleancache(), nfsrv_rcv(), nfsrv_wakenfsd(), nfs_sndunlock();
+static void nfsd_rt();
+void nfsrv_slpderef();
+
+#define	TRUE	1
+#define	FALSE	0
+
+static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON];
+/*
+ * NFS server system calls
+ * getfh() lives here too, but maybe should move to kern/vfs_syscalls.c
+ */
+
+/*
+ * Get file handle system call
+ */
+struct getfh_args {
+	char	*fname;
+	fhandle_t *fhp;
+};
+int
+getfh(p, uap, retval)
+	struct proc *p;
+	register struct getfh_args *uap;
+	int *retval;
+{
+	register struct vnode *vp;
+	fhandle_t fh;
+	int error;
+	struct nameidata nd;
+
+	/*
+	 * Must be super user
+	 */
+	error = suser(p->p_ucred, &p->p_acflag);
+	if(error)
+		return (error);
+	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, p);
+	error = namei(&nd);
+	if (error)
+		return (error);
+	vp = nd.ni_vp;
+	bzero((caddr_t)&fh, sizeof(fh));
+	fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
+	error = VFS_VPTOFH(vp, &fh.fh_fid);
+	vput(vp);
+	if (error)
+		return (error);
+	error = copyout((caddr_t)&fh, (caddr_t)uap->fhp, sizeof (fh));
+	return (error);
+}
+
+/*
+ * Nfs server psuedo system call for the nfsd's
+ * Based on the flag value it either:
+ * - adds a socket to the selection list
+ * - remains in the kernel as an nfsd
+ * - remains in the kernel as an nfsiod
+ */
+struct nfssvc_args {
+	int flag;
+	caddr_t argp;
+};
+int
+nfssvc(p, uap, retval)
+	struct proc *p;
+	register struct nfssvc_args *uap;
+	int *retval;
+{
+	struct nameidata nd;
+	struct file *fp;
+	struct mbuf *nam;
+	struct nfsd_args nfsdarg;
+	struct nfsd_srvargs nfsd_srvargs, *nsd = &nfsd_srvargs;
+	struct nfsd_cargs ncd;
+	struct nfsd *nfsd;
+	struct nfssvc_sock *slp;
+	struct nfsuid *nuidp;
+	struct nfsmount *nmp;
+	int error;
+
+	/*
+	 * Must be super user
+	 */
+	error = suser(p->p_ucred, &p->p_acflag);
+	if(error)
+		return (error);
+	while (nfssvc_sockhead_flag & SLP_INIT) {
+		 nfssvc_sockhead_flag |= SLP_WANTINIT;
+		(void) tsleep((caddr_t)&nfssvc_sockhead, PSOCK, "nfsd init", 0);
+	}
+	if (uap->flag & NFSSVC_BIOD)
+		error = nfssvc_iod(p);
+	else if (uap->flag & NFSSVC_MNTD) {
+		error = copyin(uap->argp, (caddr_t)&ncd, sizeof (ncd));
+		if (error)
+			return (error);
+		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
+			ncd.ncd_dirp, p);
+		error = namei(&nd);
+		if (error)
+			return (error);
+		if ((nd.ni_vp->v_flag & VROOT) == 0)
+			error = EINVAL;
+		nmp = VFSTONFS(nd.ni_vp->v_mount);
+		vput(nd.ni_vp);
+		if (error)
+			return (error);
+		if ((nmp->nm_flag & NFSMNT_MNTD) &&
+			(uap->flag & NFSSVC_GOTAUTH) == 0)
+			return (0);
+		nmp->nm_flag |= NFSMNT_MNTD;
+		error = nqnfs_clientd(nmp, p->p_ucred, &ncd, uap->flag,
+			uap->argp, p);
+	} else if (uap->flag & NFSSVC_ADDSOCK) {
+		error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
+		if (error)
+			return (error);
+		error = getsock(p->p_fd, nfsdarg.sock, &fp);
+		if (error)
+			return (error);
+		/*
+		 * Get the client address for connected sockets.
+		 */
+		if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
+			nam = (struct mbuf *)0;
+		else {
+			error = sockargs(&nam, nfsdarg.name, nfsdarg.namelen,
+				MT_SONAME);
+			if (error)
+				return (error);
+		}
+		error = nfssvc_addsock(fp, nam);
+	} else {
+		error = copyin(uap->argp, (caddr_t)nsd, sizeof (*nsd));
+		if (error)
+			return (error);
+		if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) &&
+			(nfsd->nfsd_slp->ns_flag & SLP_VALID)) {
+			slp = nfsd->nfsd_slp;
+
+			/*
+			 * First check to see if another nfsd has already
+			 * added this credential.
+			 */
+			for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first;
+			    nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
+				if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid &&
+				    (!nfsd->nfsd_nd->nd_nam2 ||
+				     netaddr_match(NU_NETFAM(nuidp),
+				     &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2)))
+					break;
+			}
+			if (nuidp) {
+			    nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr);
+			    nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
+			} else {
+			    /*
+			     * Nope, so we will.
+			     */
+			    if (slp->ns_numuids < nuidhash_max) {
+				slp->ns_numuids++;
+				nuidp = (struct nfsuid *)
+				   malloc(sizeof (struct nfsuid), M_NFSUID,
+					M_WAITOK);
+			    } else
+				nuidp = (struct nfsuid *)0;
+			    if ((slp->ns_flag & SLP_VALID) == 0) {
+				if (nuidp)
+				    free((caddr_t)nuidp, M_NFSUID);
+			    } else {
+				if (nuidp == (struct nfsuid *)0) {
+				    nuidp = slp->ns_uidlruhead.tqh_first;
+				    LIST_REMOVE(nuidp, nu_hash);
+				    TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp,
+					nu_lru);
+				    if (nuidp->nu_flag & NU_NAM)
+					m_freem(nuidp->nu_nam);
+			        }
+				nuidp->nu_flag = 0;
+				nuidp->nu_cr = nsd->nsd_cr;
+				if (nuidp->nu_cr.cr_ngroups > NGROUPS)
+				    nuidp->nu_cr.cr_ngroups = NGROUPS;
+				nuidp->nu_cr.cr_ref = 1;
+				nuidp->nu_timestamp = nsd->nsd_timestamp;
+				nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl;
+				/*
+				 * and save the session key in nu_key.
+				 */
+				bcopy(nsd->nsd_key, nuidp->nu_key,
+				    sizeof (nsd->nsd_key));
+				if (nfsd->nfsd_nd->nd_nam2) {
+				    struct sockaddr_in *saddr;
+
+				    saddr = mtod(nfsd->nfsd_nd->nd_nam2,
+					 struct sockaddr_in *);
+				    switch (saddr->sin_family) {
+				    case AF_INET:
+					nuidp->nu_flag |= NU_INETADDR;
+					nuidp->nu_inetaddr =
+					     saddr->sin_addr.s_addr;
+					break;
+				    case AF_ISO:
+				    default:
+					nuidp->nu_flag |= NU_NAM;
+					nuidp->nu_nam = m_copym(
+					    nfsd->nfsd_nd->nd_nam2, 0,
+					     M_COPYALL, M_WAIT);
+					break;
+				    };
+				}
+				TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp,
+					nu_lru);
+				LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid),
+					nuidp, nu_hash);
+				nfsrv_setcred(&nuidp->nu_cr,
+				    &nfsd->nfsd_nd->nd_cr);
+				nfsd->nfsd_nd->nd_flag |= ND_KERBFULL;
+			    }
+			}
+		}
+		if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd))
+			nfsd->nfsd_flag |= NFSD_AUTHFAIL;
+		error = nfssvc_nfsd(nsd, uap->argp, p);
+	}
+	if (error == EINTR || error == ERESTART)
+		error = 0;
+	return (error);
+}
+
+/*
+ * Adds a socket to the list for servicing by nfsds.
+ */
+int
+nfssvc_addsock(fp, mynam)
+	struct file *fp;
+	struct mbuf *mynam;
+{
+	register struct mbuf *m;
+	register int siz;
+	register struct nfssvc_sock *slp;
+	register struct socket *so;
+	struct nfssvc_sock *tslp;
+	int error, s;
+
+	so = (struct socket *)fp->f_data;
+	tslp = (struct nfssvc_sock *)0;
+	/*
+	 * Add it to the list, as required.
+	 */
+	if (so->so_proto->pr_protocol == IPPROTO_UDP) {
+		tslp = nfs_udpsock;
+		if (tslp->ns_flag & SLP_VALID) {
+			m_freem(mynam);
+			return (EPERM);
+		}
+#ifdef ISO
+	} else if (so->so_proto->pr_protocol == ISOPROTO_CLTP) {
+		tslp = nfs_cltpsock;
+		if (tslp->ns_flag & SLP_VALID) {
+			m_freem(mynam);
+			return (EPERM);
+		}
+#endif /* ISO */
+	}
+	if (so->so_type == SOCK_STREAM)
+		siz = NFS_MAXPACKET + sizeof (u_long);
+	else
+		siz = NFS_MAXPACKET;
+	error = soreserve(so, siz, siz); 
+	if (error) {
+		m_freem(mynam);
+		return (error);
+	}
+
+	/*
+	 * Set protocol specific options { for now TCP only } and
+	 * reserve some space. For datagram sockets, this can get called
+	 * repeatedly for the same socket, but that isn't harmful.
+	 */
+	if (so->so_type == SOCK_STREAM) {
+		MGET(m, M_WAIT, MT_SOOPTS);
+		*mtod(m, int *) = 1;
+		m->m_len = sizeof(int);
+		sosetopt(so, SOL_SOCKET, SO_KEEPALIVE, m);
+	}
+	if (so->so_proto->pr_domain->dom_family == AF_INET &&
+	    so->so_proto->pr_protocol == IPPROTO_TCP) {
+		MGET(m, M_WAIT, MT_SOOPTS);
+		*mtod(m, int *) = 1;
+		m->m_len = sizeof(int);
+		sosetopt(so, IPPROTO_TCP, TCP_NODELAY, m);
+	}
+	so->so_rcv.sb_flags &= ~SB_NOINTR;
+	so->so_rcv.sb_timeo = 0;
+	so->so_snd.sb_flags &= ~SB_NOINTR;
+	so->so_snd.sb_timeo = 0;
+	if (tslp)
+		slp = tslp;
+	else {
+		slp = (struct nfssvc_sock *)
+			malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
+		bzero((caddr_t)slp, sizeof (struct nfssvc_sock));
+		TAILQ_INIT(&slp->ns_uidlruhead);
+		TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
+	}
+	slp->ns_so = so;
+	slp->ns_nam = mynam;
+	fp->f_count++;
+	slp->ns_fp = fp;
+	s = splnet();
+	so->so_upcallarg = (caddr_t)slp;
+	so->so_upcall = nfsrv_rcv;
+	slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
+	nfsrv_wakenfsd(slp);
+	splx(s);
+	return (0);
+}
+
+/*
+ * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
+ * until it is killed by a signal.
+ */
+int
+nfssvc_nfsd(nsd, argp, p)
+	struct nfsd_srvargs *nsd;
+	caddr_t argp;
+	struct proc *p;
+{
+	register struct mbuf *m;
+	register int siz;
+	register struct nfssvc_sock *slp;
+	register struct socket *so;
+	register int *solockp;
+	struct nfsd *nfsd = nsd->nsd_nfsd;
+	struct nfsrv_descript *nd = NULL;
+	struct mbuf *mreq;
+	struct nfsuid *uidp;
+	int error = 0, cacherep, s, sotype, writes_todo;
+	u_quad_t cur_usec;
+
+#ifndef nolint
+	cacherep = RC_DOIT;
+	writes_todo = 0;
+#endif
+	s = splnet();
+	if (nfsd == (struct nfsd *)0) {
+		nsd->nsd_nfsd = nfsd = (struct nfsd *)
+			malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK);
+		bzero((caddr_t)nfsd, sizeof (struct nfsd));
+		nfsd->nfsd_procp = p;
+		TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
+		nfs_numnfsd++;
+	}
+	/*
+	 * Loop getting rpc requests until SIGKILL.
+	 */
+	for (;;) {
+		if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
+			while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
+			    (nfsd_head_flag & NFSD_CHECKSLP) == 0) {
+				nfsd->nfsd_flag |= NFSD_WAITING;
+				nfsd_waiting++;
+				error = tsleep((caddr_t)nfsd, PSOCK | PCATCH,
+				    "nfsd", 0);
+				nfsd_waiting--;
+				if (error)
+					goto done;
+			}
+			if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 &&
+			    (nfsd_head_flag & NFSD_CHECKSLP) != 0) {
+				for (slp = nfssvc_sockhead.tqh_first; slp != 0;
+				    slp = slp->ns_chain.tqe_next) {
+				    if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
+					== (SLP_VALID | SLP_DOREC)) {
+					    slp->ns_flag &= ~SLP_DOREC;
+					    slp->ns_sref++;
+					    nfsd->nfsd_slp = slp;
+					    break;
+				    }
+				}
+				if (slp == 0)
+					nfsd_head_flag &= ~NFSD_CHECKSLP;
+			}
+			if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
+				continue;
+			if (slp->ns_flag & SLP_VALID) {
+				if (slp->ns_flag & SLP_DISCONN)
+					nfsrv_zapsock(slp);
+				else if (slp->ns_flag & SLP_NEEDQ) {
+					slp->ns_flag &= ~SLP_NEEDQ;
+					(void) nfs_sndlock(&slp->ns_solock,
+						(struct nfsreq *)0);
+					nfsrv_rcv(slp->ns_so, (caddr_t)slp,
+						M_WAIT);
+					nfs_sndunlock(&slp->ns_solock);
+				}
+				error = nfsrv_dorec(slp, nfsd, &nd);
+				cur_usec = (u_quad_t)time.tv_sec * 1000000 +
+					(u_quad_t)time.tv_usec;
+				if (error && slp->ns_tq.lh_first &&
+				    slp->ns_tq.lh_first->nd_time <= cur_usec) {
+					error = 0;
+					cacherep = RC_DOIT;
+					writes_todo = 1;
+				} else
+					writes_todo = 0;
+				nfsd->nfsd_flag |= NFSD_REQINPROG;
+			}
+		} else {
+			error = 0;
+			slp = nfsd->nfsd_slp;
+		}
+		if (error || (slp->ns_flag & SLP_VALID) == 0) {
+			if (nd) {
+				free((caddr_t)nd, M_NFSRVDESC);
+				nd = NULL;
+			}
+			nfsd->nfsd_slp = (struct nfssvc_sock *)0;
+			nfsd->nfsd_flag &= ~NFSD_REQINPROG;
+			nfsrv_slpderef(slp);
+			continue;
+		}
+		splx(s);
+		so = slp->ns_so;
+		sotype = so->so_type;
+		if (so->so_proto->pr_flags & PR_CONNREQUIRED)
+			solockp = &slp->ns_solock;
+		else
+			solockp = (int *)0;
+		if (nd) {
+		    nd->nd_starttime = time;
+		    if (nd->nd_nam2)
+			nd->nd_nam = nd->nd_nam2;
+		    else
+			nd->nd_nam = slp->ns_nam;
+
+		    /*
+		     * Check to see if authorization is needed.
+		     */
+		    if (nfsd->nfsd_flag & NFSD_NEEDAUTH) {
+			nfsd->nfsd_flag &= ~NFSD_NEEDAUTH;
+			nsd->nsd_haddr = mtod(nd->nd_nam,
+			    struct sockaddr_in *)->sin_addr.s_addr;
+			nsd->nsd_authlen = nfsd->nfsd_authlen;
+			nsd->nsd_verflen = nfsd->nfsd_verflen;
+			if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr,
+				nfsd->nfsd_authlen) &&
+			    !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr,
+				nfsd->nfsd_verflen) &&
+			    !copyout((caddr_t)nsd, argp, sizeof (*nsd)))
+			    return (ENEEDAUTH);
+			cacherep = RC_DROPIT;
+		    } else
+			cacherep = nfsrv_getcache(nd, slp, &mreq);
+
+		    /*
+		     * Check for just starting up for NQNFS and send
+		     * fake "try again later" replies to the NQNFS clients.
+		     */
+		    if (notstarted && nqnfsstarttime <= time.tv_sec) {
+			if (modify_flag) {
+				nqnfsstarttime = time.tv_sec + nqsrv_writeslack;
+				modify_flag = 0;
+			} else
+				notstarted = 0;
+		    }
+		    if (notstarted) {
+			if ((nd->nd_flag & ND_NQNFS) == 0)
+				cacherep = RC_DROPIT;
+			else if (nd->nd_procnum != NFSPROC_WRITE) {
+				nd->nd_procnum = NFSPROC_NOOP;
+				nd->nd_repstat = NQNFS_TRYLATER;
+				cacherep = RC_DOIT;
+			} else
+				modify_flag = 1;
+		    } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) {
+			nfsd->nfsd_flag &= ~NFSD_AUTHFAIL;
+			nd->nd_procnum = NFSPROC_NOOP;
+			nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
+			cacherep = RC_DOIT;
+		    }
+		}
+
+		/*
+		 * Loop to get all the write rpc relies that have been
+		 * gathered together.
+		 */
+		do {
+		    switch (cacherep) {
+		    case RC_DOIT:
+			if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
+			    nfsrvw_procrastinate > 0 && !notstarted))
+			    error = nfsrv_writegather(&nd, slp,
+				nfsd->nfsd_procp, &mreq);
+			else
+			    error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
+				slp, nfsd->nfsd_procp, &mreq);
+			if (mreq == NULL)
+				break;
+			if (error) {
+				if (nd->nd_procnum != NQNFSPROC_VACATED)
+					nfsstats.srv_errs++;
+				nfsrv_updatecache(nd, FALSE, mreq);
+				if (nd->nd_nam2)
+					m_freem(nd->nd_nam2);
+				break;
+			}
+			nfsstats.srvrpccnt[nd->nd_procnum]++;
+			nfsrv_updatecache(nd, TRUE, mreq);
+			nd->nd_mrep = (struct mbuf *)0;
+		    case RC_REPLY:
+			m = mreq;
+			siz = 0;
+			while (m) {
+				siz += m->m_len;
+				m = m->m_next;
+			}
+			if (siz <= 0 || siz > NFS_MAXPACKET) {
+				printf("mbuf siz=%d\n",siz);
+				panic("Bad nfs svc reply");
+			}
+			m = mreq;
+			m->m_pkthdr.len = siz;
+			m->m_pkthdr.rcvif = (struct ifnet *)0;
+			/*
+			 * For stream protocols, prepend a Sun RPC
+			 * Record Mark.
+			 */
+			if (sotype == SOCK_STREAM) {
+				M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
+				*mtod(m, u_long *) = htonl(0x80000000 | siz);
+			}
+			if (solockp)
+				(void) nfs_sndlock(solockp, (struct nfsreq *)0);
+			if (slp->ns_flag & SLP_VALID)
+			    error = nfs_send(so, nd->nd_nam2, m, NULL);
+			else {
+			    error = EPIPE;
+			    m_freem(m);
+			}
+			if (nfsrtton)
+				nfsd_rt(sotype, nd, cacherep);
+			if (nd->nd_nam2)
+				MFREE(nd->nd_nam2, m);
+			if (nd->nd_mrep)
+				m_freem(nd->nd_mrep);
+			if (error == EPIPE)
+				nfsrv_zapsock(slp);
+			if (solockp)
+				nfs_sndunlock(solockp);
+			if (error == EINTR || error == ERESTART) {
+				free((caddr_t)nd, M_NFSRVDESC);
+				nfsrv_slpderef(slp);
+				s = splnet();
+				goto done;
+			}
+			break;
+		    case RC_DROPIT:
+			if (nfsrtton)
+				nfsd_rt(sotype, nd, cacherep);
+			m_freem(nd->nd_mrep);
+			m_freem(nd->nd_nam2);
+			break;
+		    };
+		    if (nd) {
+			FREE((caddr_t)nd, M_NFSRVDESC);
+			nd = NULL;
+		    }
+
+		    /*
+		     * Check to see if there are outstanding writes that
+		     * need to be serviced.
+		     */
+		    cur_usec = (u_quad_t)time.tv_sec * 1000000 +
+			(u_quad_t)time.tv_usec;
+		    s = splsoftclock();
+		    if (slp->ns_tq.lh_first &&
+			slp->ns_tq.lh_first->nd_time <= cur_usec) {
+			cacherep = RC_DOIT;
+			writes_todo = 1;
+		    } else
+			writes_todo = 0;
+		    splx(s);
+		} while (writes_todo);
+		s = splnet();
+		if (nfsrv_dorec(slp, nfsd, &nd)) {
+			nfsd->nfsd_flag &= ~NFSD_REQINPROG;
+			nfsd->nfsd_slp = NULL;
+			nfsrv_slpderef(slp);
+		}
+	}
+done:
+	TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
+	splx(s);
+	free((caddr_t)nfsd, M_NFSD);
+	nsd->nsd_nfsd = (struct nfsd *)0;
+	if (--nfs_numnfsd == 0)
+		nfsrv_init(TRUE);	/* Reinitialize everything */
+	return (error);
+}
+
+/*
+ * Asynchronous I/O daemons for client nfs.
+ * They do read-ahead and write-behind operations on the block I/O cache.
+ * Never returns unless it fails or gets killed.
+ */
+int
+nfssvc_iod(p)
+	struct proc *p;
+{
+	register struct buf *bp, *nbp;
+	register int i, myiod;
+	struct vnode *vp;
+	int error = 0, s;
+
+	/*
+	 * Assign my position or return error if too many already running
+	 */
+	myiod = -1;
+	for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
+		if (nfs_asyncdaemon[i] == 0) {
+			nfs_asyncdaemon[i]++;
+			myiod = i;
+			break;
+		}
+	if (myiod == -1)
+		return (EBUSY);
+	nfs_numasync++;
+	/*
+	 * Just loop around doin our stuff until SIGKILL
+	 */
+	for (;;) {
+	    while (nfs_bufq.tqh_first == NULL && error == 0) {
+		nfs_iodwant[myiod] = p;
+		error = tsleep((caddr_t)&nfs_iodwant[myiod],
+			PWAIT | PCATCH, "nfsidl", 0);
+	    }
+	    while ((bp = nfs_bufq.tqh_first) != NULL) {
+		/* Take one off the front of the list */
+		TAILQ_REMOVE(&nfs_bufq, bp, b_freelist);
+		if (bp->b_flags & B_READ)
+		    (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0);
+		else do {
+		    /*
+		     * Look for a delayed write for the same vnode, so I can do 
+		     * it now. We must grab it before calling nfs_doio() to
+		     * avoid any risk of the vnode getting vclean()'d while
+		     * we are doing the write rpc.
+		     */
+		    vp = bp->b_vp;
+		    s = splbio();
+		    for (nbp = vp->v_dirtyblkhd.lh_first; nbp;
+			nbp = nbp->b_vnbufs.le_next) {
+			if ((nbp->b_flags &
+			    (B_BUSY|B_DELWRI|B_NEEDCOMMIT|B_NOCACHE))!=B_DELWRI)
+			    continue;
+			bremfree(nbp);
+			nbp->b_flags |= (B_BUSY|B_ASYNC);
+			break;
+		    }
+		    splx(s);
+		    /*
+		     * For the delayed write, do the first part of nfs_bwrite()
+		     * up to, but not including nfs_strategy().
+		     */
+		    if (nbp) {
+			nbp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI);
+			reassignbuf(nbp, nbp->b_vp);
+			nbp->b_vp->v_numoutput++;
+		    }
+		    (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0);
+		} while (bp = nbp);
+	    }
+	    if (error) {
+		nfs_asyncdaemon[myiod] = 0;
+		nfs_numasync--;
+		return (error);
+	    }
+	}
+}
+
+/*
+ * Shut down a socket associated with an nfssvc_sock structure.
+ * Should be called with the send lock set, if required.
+ * The trick here is to increment the sref at the start, so that the nfsds
+ * will stop using it and clear ns_flag at the end so that it will not be
+ * reassigned during cleanup.
+ */
+void
+nfsrv_zapsock(slp)
+	register struct nfssvc_sock *slp;
+{
+	register struct nfsuid *nuidp, *nnuidp;
+	register struct nfsrv_descript *nwp, *nnwp;
+	struct socket *so;
+	struct file *fp;
+	struct mbuf *m;
+	int s;
+
+	slp->ns_flag &= ~SLP_ALLFLAGS;
+	fp = slp->ns_fp;
+	if (fp) {
+		slp->ns_fp = (struct file *)0;
+		so = slp->ns_so;
+		so->so_upcall = NULL;
+		soshutdown(so, 2);
+		closef(fp, (struct proc *)0);
+		if (slp->ns_nam)
+			MFREE(slp->ns_nam, m);
+		m_freem(slp->ns_raw);
+		m_freem(slp->ns_rec);
+		for (nuidp = slp->ns_uidlruhead.tqh_first; nuidp != 0;
+		    nuidp = nnuidp) {
+			nnuidp = nuidp->nu_lru.tqe_next;
+			LIST_REMOVE(nuidp, nu_hash);
+			TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru);
+			if (nuidp->nu_flag & NU_NAM)
+				m_freem(nuidp->nu_nam);
+			free((caddr_t)nuidp, M_NFSUID);
+		}
+		s = splsoftclock();
+		for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
+			nnwp = nwp->nd_tq.le_next;
+			LIST_REMOVE(nwp, nd_tq);
+			free((caddr_t)nwp, M_NFSRVDESC);
+		}
+		LIST_INIT(&slp->ns_tq);
+		splx(s);
+	}
+}
+
+/*
+ * Get an authorization string for the uid by having the mount_nfs sitting
+ * on this mount point porpous out of the kernel and do it.
+ */
+int
+nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key)
+	register struct nfsmount *nmp;
+	struct nfsreq *rep;
+	struct ucred *cred;
+	char **auth_str;
+	int *auth_len;
+	char *verf_str;
+	int *verf_len;
+	NFSKERBKEY_T key;		/* return session key */
+{
+	int error = 0;
+
+	while ((nmp->nm_flag & NFSMNT_WAITAUTH) == 0) {
+		nmp->nm_flag |= NFSMNT_WANTAUTH;
+		(void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK,
+			"nfsauth1", 2 * hz);
+		error = nfs_sigintr(nmp, rep, rep->r_procp);
+		if (error) {
+			nmp->nm_flag &= ~NFSMNT_WANTAUTH;
+			return (error);
+		}
+	}
+	nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH);
+	nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
+	nmp->nm_authlen = RPCAUTH_MAXSIZ;
+	nmp->nm_verfstr = verf_str;
+	nmp->nm_verflen = *verf_len;
+	nmp->nm_authuid = cred->cr_uid;
+	wakeup((caddr_t)&nmp->nm_authstr);
+
+	/*
+	 * And wait for mount_nfs to do its stuff.
+	 */
+	while ((nmp->nm_flag & NFSMNT_HASAUTH) == 0 && error == 0) {
+		(void) tsleep((caddr_t)&nmp->nm_authlen, PSOCK,
+			"nfsauth2", 2 * hz);
+		error = nfs_sigintr(nmp, rep, rep->r_procp);
+	}
+	if (nmp->nm_flag & NFSMNT_AUTHERR) {
+		nmp->nm_flag &= ~NFSMNT_AUTHERR;
+		error = EAUTH;
+	}
+	if (error)
+		free((caddr_t)*auth_str, M_TEMP);
+	else {
+		*auth_len = nmp->nm_authlen;
+		*verf_len = nmp->nm_verflen;
+		bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key));
+	}
+	nmp->nm_flag &= ~NFSMNT_HASAUTH;
+	nmp->nm_flag |= NFSMNT_WAITAUTH;
+	if (nmp->nm_flag & NFSMNT_WANTAUTH) {
+		nmp->nm_flag &= ~NFSMNT_WANTAUTH;
+		wakeup((caddr_t)&nmp->nm_authtype);
+	}
+	return (error);
+}
+
+/*
+ * Get a nickname authenticator and verifier.
+ */
+int
+nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len)
+	struct nfsmount *nmp;
+	struct ucred *cred;
+	char **auth_str;
+	int *auth_len;
+	char *verf_str;
+	int verf_len;
+{
+	register struct nfsuid *nuidp;
+	register u_long *nickp, *verfp;
+	struct timeval ktvin, ktvout;
+	NFSKERBKEYSCHED_T keys;	/* stores key schedule */
+
+#ifdef DIAGNOSTIC
+	if (verf_len < (4 * NFSX_UNSIGNED))
+		panic("nfs_getnickauth verf too small");
+#endif
+	for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first;
+	    nuidp != 0; nuidp = nuidp->nu_hash.le_next) {
+		if (nuidp->nu_cr.cr_uid == cred->cr_uid)
+			break;
+	}
+	if (!nuidp || nuidp->nu_expire < time.tv_sec)
+		return (EACCES);
+
+	/*
+	 * Move to the end of the lru list (end of lru == most recently used).
+	 */
+	TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
+	TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
+
+	nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK);
+	*nickp++ = txdr_unsigned(RPCAKN_NICKNAME);
+	*nickp = txdr_unsigned(nuidp->nu_nickname);
+	*auth_str = (char *)nickp;
+	*auth_len = 2 * NFSX_UNSIGNED;
+
+	/*
+	 * Now we must encrypt the verifier and package it up.
+	 */
+	verfp = (u_long *)verf_str;
+	*verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
+	if (time.tv_sec > nuidp->nu_timestamp.tv_sec ||
+	    (time.tv_sec == nuidp->nu_timestamp.tv_sec &&
+	     time.tv_usec > nuidp->nu_timestamp.tv_usec))
+		nuidp->nu_timestamp = time;
+	else
+		nuidp->nu_timestamp.tv_usec++;
+	ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec);
+	ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec);
+
+	/*
+	 * Now encrypt the timestamp verifier in ecb mode using the session
+	 * key.
+	 */
+#ifdef NFSKERB
+	XXX
+#endif
+
+	*verfp++ = ktvout.tv_sec;
+	*verfp++ = ktvout.tv_usec;
+	*verfp = 0;
+	return (0);
+}
+
+/*
+ * Save the current nickname in a hash list entry on the mount point.
+ */
+int
+nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep)
+	register struct nfsmount *nmp;
+	struct ucred *cred;
+	int len;
+	NFSKERBKEY_T key;
+	struct mbuf **mdp;
+	char **dposp;
+	struct mbuf *mrep;
+{
+	register struct nfsuid *nuidp;
+	register u_long *tl;
+	register long t1;
+	struct mbuf *md = *mdp;
+	struct timeval ktvin, ktvout;
+	u_long nick;
+	NFSKERBKEYSCHED_T keys;
+	char *dpos = *dposp, *cp2;
+	int deltasec, error = 0;
+
+	if (len == (3 * NFSX_UNSIGNED)) {
+		nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+		ktvin.tv_sec = *tl++;
+		ktvin.tv_usec = *tl++;
+		nick = fxdr_unsigned(u_long, *tl);
+
+		/*
+		 * Decrypt the timestamp in ecb mode.
+		 */
+#ifdef NFSKERB
+		XXX
+#endif
+		ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
+		ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
+		deltasec = time.tv_sec - ktvout.tv_sec;
+		if (deltasec < 0)
+			deltasec = -deltasec;
+		/*
+		 * If ok, add it to the hash list for the mount point.
+		 */
+		if (deltasec <= NFS_KERBCLOCKSKEW) {
+			if (nmp->nm_numuids < nuidhash_max) {
+				nmp->nm_numuids++;
+				nuidp = (struct nfsuid *)
+				   malloc(sizeof (struct nfsuid), M_NFSUID,
+					M_WAITOK);
+			} else {
+				nuidp = nmp->nm_uidlruhead.tqh_first;
+				LIST_REMOVE(nuidp, nu_hash);
+				TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp,
+					nu_lru);
+			}
+			nuidp->nu_flag = 0;
+			nuidp->nu_cr.cr_uid = cred->cr_uid;
+			nuidp->nu_expire = time.tv_sec + NFS_KERBTTL;
+			nuidp->nu_timestamp = ktvout;
+			nuidp->nu_nickname = nick;
+			bcopy(key, nuidp->nu_key, sizeof (key));
+			TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp,
+				nu_lru);
+			LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid),
+				nuidp, nu_hash);
+		}
+	} else
+		nfsm_adv(nfsm_rndup(len));
+nfsmout:
+	*mdp = md;
+	*dposp = dpos;
+	return (error);
+}
+
+/*
+ * Derefence a server socket structure. If it has no more references and
+ * is no longer valid, you can throw it away.
+ */
+void
+nfsrv_slpderef(slp)
+	register struct nfssvc_sock *slp;
+{
+	if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
+		TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
+		free((caddr_t)slp, M_NFSSVC);
+	}
+}
+
+/*
+ * Initialize the data structures for the server.
+ * Handshake with any new nfsds starting up to avoid any chance of
+ * corruption.
+ */
+void
+nfsrv_init(terminating)
+	int terminating;
+{
+	register struct nfssvc_sock *slp, *nslp;
+
+	if (nfssvc_sockhead_flag & SLP_INIT)
+		panic("nfsd init");
+	nfssvc_sockhead_flag |= SLP_INIT;
+	if (terminating) {
+		for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = nslp) {
+			nslp = slp->ns_chain.tqe_next;
+			if (slp->ns_flag & SLP_VALID)
+				nfsrv_zapsock(slp);
+			TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
+			free((caddr_t)slp, M_NFSSVC);
+		}
+		nfsrv_cleancache();	/* And clear out server cache */
+	}
+
+	TAILQ_INIT(&nfssvc_sockhead);
+	nfssvc_sockhead_flag &= ~SLP_INIT;
+	if (nfssvc_sockhead_flag & SLP_WANTINIT) {
+		nfssvc_sockhead_flag &= ~SLP_WANTINIT;
+		wakeup((caddr_t)&nfssvc_sockhead);
+	}
+
+	TAILQ_INIT(&nfsd_head);
+	nfsd_head_flag &= ~NFSD_CHECKSLP;
+
+	nfs_udpsock = (struct nfssvc_sock *)
+	    malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
+	bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock));
+	TAILQ_INIT(&nfs_udpsock->ns_uidlruhead);
+	TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
+
+	nfs_cltpsock = (struct nfssvc_sock *)
+	    malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK);
+	bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock));
+	TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead);
+	TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
+}
+
+/*
+ * Add entries to the server monitor log.
+ */
+static void
+nfsd_rt(sotype, nd, cacherep)
+	int sotype;
+	register struct nfsrv_descript *nd;
+	int cacherep;
+{
+	register struct drt *rt;
+
+	rt = &nfsdrt.drt[nfsdrt.pos];
+	if (cacherep == RC_DOIT)
+		rt->flag = 0;
+	else if (cacherep == RC_REPLY)
+		rt->flag = DRT_CACHEREPLY;
+	else
+		rt->flag = DRT_CACHEDROP;
+	if (sotype == SOCK_STREAM)
+		rt->flag |= DRT_TCP;
+	if (nd->nd_flag & ND_NQNFS)
+		rt->flag |= DRT_NQNFS;
+	else if (nd->nd_flag & ND_NFSV3)
+		rt->flag |= DRT_NFSV3;
+	rt->proc = nd->nd_procnum;
+	if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET)
+	    rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr;
+	else
+	    rt->ipadr = INADDR_ANY;
+	rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) +
+		(time.tv_usec - nd->nd_starttime.tv_usec);
+	rt->tstamp = time;
+	nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
+}
diff --git a/sys/nfs/nfs_vfsops.c b/sys/nfs/nfs_vfsops.c
new file mode 100644
index 000000000000..54e2a71f871f
--- /dev/null
+++ b/sys/nfs/nfs_vfsops.c
@@ -0,0 +1,931 @@
+/*
+ * Copyright (c) 1989, 1993, 1995
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_vfsops.c	8.12 (Berkeley) 5/20/95
+ */
+
+#include <sys/param.h>
+#include <sys/conf.h>
+#include <sys/ioctl.h>
+#include <sys/signal.h>
+#include <sys/proc.h>
+#include <sys/namei.h>
+#include <sys/vnode.h>
+#include <sys/kernel.h>
+#include <sys/mount.h>
+#include <sys/buf.h>
+#include <sys/mbuf.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+#include <sys/systm.h>
+
+#include <net/if.h>
+#include <net/route.h>
+#include <netinet/in.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsmount.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nfsdiskless.h>
+#include <nfs/nqnfs.h>
+
+struct nfsstats nfsstats;
+static int nfs_sysctl(int *, u_int, void *, size_t *, void *, size_t,
+		      struct proc *);
+extern int nfs_ticks;
+
+/*
+ * nfs vfs operations.
+ */
+struct vfsops nfs_vfsops = {
+	nfs_mount,
+	nfs_start,
+	nfs_unmount,
+	nfs_root,
+	nfs_quotactl,
+	nfs_statfs,
+	nfs_sync,
+	nfs_vget,
+	nfs_fhtovp,
+	nfs_vptofh,
+	nfs_init,
+	nfs_sysctl
+};
+
+/*
+ * This structure must be filled in by a primary bootstrap or bootstrap
+ * server for a diskless/dataless machine. It is initialized below just
+ * to ensure that it is allocated to initialized data (.data not .bss).
+ */
+struct nfs_diskless nfs_diskless = { 0 };
+int nfs_diskless_valid = 0;
+
+void nfs_disconnect __P((struct nfsmount *));
+void nfsargs_ntoh __P((struct nfs_args *));
+int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, 
+	struct proc *));
+static int nfs_mountdiskless __P((char *, char *, int, struct sockaddr_in *,
+	struct nfs_args *, struct proc *, struct vnode **, struct mount **));
+
+/*
+ * nfs statfs call
+ */
+int
+nfs_statfs(mp, sbp, p)
+	struct mount *mp;
+	register struct statfs *sbp;
+	struct proc *p;
+{
+	register struct vnode *vp;
+	register struct nfs_statfs *sfp;
+	register caddr_t cp;
+	register u_long *tl;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	struct nfsmount *nmp = VFSTONFS(mp);
+	int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct ucred *cred;
+	struct nfsnode *np;
+	u_quad_t tquad;
+
+#ifndef nolint
+	sfp = (struct nfs_statfs *)0;
+#endif
+	error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
+	if (error)
+		return (error);
+	vp = NFSTOV(np);
+	cred = crget();
+	cred->cr_ngroups = 1;
+	if (v3 && (nmp->nm_flag & NFSMNT_GOTFSINFO) == 0)
+		(void)nfs_fsinfo(nmp, vp, cred, p);
+	nfsstats.rpccnt[NFSPROC_FSSTAT]++;
+	nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
+	nfsm_fhtom(vp, v3);
+	nfsm_request(vp, NFSPROC_FSSTAT, p, cred);
+	if (v3)
+		nfsm_postop_attr(vp, retattr);
+	if (!error)
+		nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
+	sbp->f_iosize = min(nmp->nm_rsize, nmp->nm_wsize);
+	if (v3) {
+		sbp->f_bsize = NFS_FABLKSIZE;
+		fxdr_hyper(&sfp->sf_tbytes, &tquad);
+		sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
+		fxdr_hyper(&sfp->sf_fbytes, &tquad);
+		sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
+		fxdr_hyper(&sfp->sf_abytes, &tquad);
+		sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
+		sbp->f_files = (fxdr_unsigned(long, sfp->sf_tfiles.nfsuquad[1])
+			& 0x7fffffff);
+		sbp->f_ffree = (fxdr_unsigned(long, sfp->sf_ffiles.nfsuquad[1])
+			& 0x7fffffff);
+	} else {
+		sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize);
+		sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks);
+		sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree);
+		sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail);
+		sbp->f_files = 0;
+		sbp->f_ffree = 0;
+	}
+	if (sbp != &mp->mnt_stat) {
+		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
+		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
+	}
+	nfsm_reqdone;
+	vrele(vp);
+	crfree(cred);
+	return (error);
+}
+
+/*
+ * nfs version 3 fsinfo rpc call
+ */
+int
+nfs_fsinfo(nmp, vp, cred, p)
+	register struct nfsmount *nmp;
+	register struct vnode *vp;
+	struct ucred *cred;
+	struct proc *p;
+{
+	register struct nfsv3_fsinfo *fsp;
+	register caddr_t cp;
+	register long t1, t2;
+	register u_long *tl, pref, max;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, retattr;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+
+	nfsstats.rpccnt[NFSPROC_FSINFO]++;
+	nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
+	nfsm_fhtom(vp, 1);
+	nfsm_request(vp, NFSPROC_FSINFO, p, cred);
+	nfsm_postop_attr(vp, retattr);
+	if (!error) {
+		nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
+		pref = fxdr_unsigned(u_long, fsp->fs_wtpref);
+		if (pref < nmp->nm_wsize)
+			nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
+				~(NFS_FABLKSIZE - 1);
+		max = fxdr_unsigned(u_long, fsp->fs_wtmax);
+		if (max < nmp->nm_wsize) {
+			nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
+			if (nmp->nm_wsize == 0)
+				nmp->nm_wsize = max;
+		}
+		pref = fxdr_unsigned(u_long, fsp->fs_rtpref);
+		if (pref < nmp->nm_rsize)
+			nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
+				~(NFS_FABLKSIZE - 1);
+		max = fxdr_unsigned(u_long, fsp->fs_rtmax);
+		if (max < nmp->nm_rsize) {
+			nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
+			if (nmp->nm_rsize == 0)
+				nmp->nm_rsize = max;
+		}
+		pref = fxdr_unsigned(u_long, fsp->fs_dtpref);
+		if (pref < nmp->nm_readdirsize)
+			nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
+				~(NFS_DIRBLKSIZ - 1);
+		if (max < nmp->nm_readdirsize) {
+			nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
+			if (nmp->nm_readdirsize == 0)
+				nmp->nm_readdirsize = max;
+		}
+		nmp->nm_flag |= NFSMNT_GOTFSINFO;
+	}
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * Mount a remote root fs via. nfs. This depends on the info in the
+ * nfs_diskless structure that has been filled in properly by some primary
+ * bootstrap.
+ * It goes something like this:
+ * - do enough of "ifconfig" by calling ifioctl() so that the system
+ *   can talk to the server
+ * - If nfs_diskless.mygateway is filled in, use that address as
+ *   a default gateway.
+ * - hand craft the swap nfs vnode hanging off a fake mount point
+ *	if swdevt[0].sw_dev == NODEV
+ * - build the rootfs mount point and call mountnfs() to do the rest.
+ */
+int
+nfs_mountroot()
+{
+	struct mount *mp, *swap_mp;
+	struct nfs_diskless *nd = &nfs_diskless;
+	struct socket *so;
+	struct vnode *vp;
+	struct proc *p = curproc;		/* XXX */
+	int error, i;
+	u_long l;
+	char buf[128];
+
+	/*
+	 * XXX time must be non-zero when we init the interface or else
+	 * the arp code will wedge...
+	 */
+	if (time.tv_sec == 0)
+		time.tv_sec = 1;
+
+	/* 
+	 * XXX splnet, so networks will receive...
+	 */
+	splnet();
+
+#ifdef notyet
+	/* Set up swap credentials. */
+	proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
+	proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
+	if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
+		NGROUPS)
+		proc0.p_ucred->cr_ngroups = NGROUPS;
+	for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
+	    proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
+#endif
+
+	/*
+	 * Do enough of ifconfig(8) so that the critical net interface can
+	 * talk to the server.
+	 */
+	error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0);
+	if (error) {
+		printf("nfs_mountroot: socreate(%04x): %d",
+			nd->myif.ifra_addr.sa_family, error);
+		return (error);
+	}
+
+	/*
+	 * We might not have been told the right interface, so we pass
+	 * over the first ten interfaces of the same kind, until we get
+	 * one of them configured.
+	 */
+
+	for (i = strlen(nd->myif.ifra_name) - 1;
+		nd->myif.ifra_name[i] >= '0' && 
+		nd->myif.ifra_name[i] <= '9';
+		nd->myif.ifra_name[i] ++) {
+		error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, p);
+		if(!error) 
+			break;
+	}
+	if (error) {
+		printf("nfs_mountroot: SIOCAIFADDR: %d", error);
+		return (error);
+	}
+	soclose(so);
+
+	/*
+	 * If the gateway field is filled in, set it as the default route.
+	 */
+	if (nd->mygateway.sin_len != 0) {
+		struct sockaddr_in mask, sin;
+
+		bzero((caddr_t)&mask, sizeof(mask));
+		sin = mask;
+		sin.sin_family = AF_INET;
+		sin.sin_len = sizeof(sin);
+		error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
+		    (struct sockaddr *)&nd->mygateway,
+		    (struct sockaddr *)&mask,
+		    RTF_UP | RTF_GATEWAY, (struct rtentry **)0);
+		if (error) {
+			printf("nfs_mountroot: RTM_ADD: %d", error);
+			return (error);
+		}
+	}
+
+	swap_mp = NULL;
+	if (nd->swap_nblks) {
+		/*
+		 * Create a fake mount point just for the swap vnode so that the
+		 * swap file can be on a different server from the rootfs.
+		 */
+		nd->swap_args.fh = nd->swap_fh;
+		/*
+		 * If using nfsv3_diskless, replace NFSX_V2FH with
+		 * nd->swap_fhsize.
+		 */
+		nd->swap_args.fhsize = NFSX_V2FH;
+		l = ntohl(nd->swap_saddr.sin_addr.s_addr);
+		sprintf(buf,"%ld.%ld.%ld.%ld:%s",
+			(l >> 24) & 0xff, (l >> 16) & 0xff,
+			(l >>  8) & 0xff, (l >>  0) & 0xff,nd->swap_hostnam);
+		printf("NFS SWAP: %s\n",buf);
+		if (error = nfs_mountdiskless(buf, "/swap", 0,
+		    &nd->swap_saddr, &nd->swap_args, p, &vp, &swap_mp))
+			return (error);
+		vfs_unbusy(swap_mp, p);
+
+		for (i=0;swdevt[i].sw_dev != NODEV;i++) ;
+
+		/*
+		 * Since the swap file is not the root dir of a file system,
+		 * hack it to a regular file.
+		 */
+		vp->v_type = VREG;
+		vp->v_flag = 0;
+		swapdev_vp = vp;
+		VREF(vp);
+		swdevt[i].sw_vp = vp;
+		swdevt[i].sw_nblks = nd->swap_nblks*2;
+
+		if (!swdevt[i].sw_nblks) {
+			swdevt[i].sw_nblks = 2048;
+			printf("defaulting to %d kbyte.\n",
+				swdevt[i].sw_nblks/2);
+		} else
+			printf("using %d kbyte.\n",swdevt[i].sw_nblks/2);
+	}
+
+	/*
+	 * Create the rootfs mount point.
+	 */
+	nd->root_args.fh = nd->root_fh;
+	/*
+	 * If using nfsv3_diskless, replace NFSX_V2FH with nd->root_fhsize.
+	 */
+	nd->root_args.fhsize = NFSX_V2FH;
+	l = ntohl(nd->swap_saddr.sin_addr.s_addr);
+	sprintf(buf,"%ld.%ld.%ld.%ld:%s",
+		(l >> 24) & 0xff, (l >> 16) & 0xff,
+		(l >>  8) & 0xff, (l >>  0) & 0xff,nd->root_hostnam);
+	printf("NFS ROOT: %s\n",buf);
+	if (error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
+	    &nd->root_saddr, &nd->root_args, p, &vp, &mp)) {
+		if (swap_mp) {
+			mp->mnt_vfc->vfc_refcount--;
+			free(swap_mp, M_MOUNT);
+		}
+		return (error);
+	}
+
+	simple_lock(&mountlist_slock);
+	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
+	simple_unlock(&mountlist_slock);
+	rootvp = vp;
+	vfs_unbusy(mp, p);
+
+	/*
+	 * This is not really an nfs issue, but it is much easier to
+	 * set hostname here and then let the "/etc/rc.xxx" files
+	 * mount the right /var based upon its preset value.
+	 */
+	bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
+	hostname[MAXHOSTNAMELEN - 1] = '\0';
+	for (i = 0; i < MAXHOSTNAMELEN; i++)
+		if (hostname[i] == '\0')
+			break;
+	hostnamelen = i;
+	inittodr(ntohl(nd->root_time));
+	return (0);
+}
+
+/*
+ * Internal version of mount system call for diskless setup.
+ */
+static int
+nfs_mountdiskless(path, which, mountflag, sin, args, p, vpp, mpp)
+	char *path;
+	char *which;
+	int mountflag;
+	struct sockaddr_in *sin;
+	struct nfs_args *args;
+	struct proc *p;
+	struct vnode **vpp;
+	struct mount **mpp;
+{
+	struct mount *mp;
+	struct mbuf *m;
+	int error;
+
+	if (error = vfs_rootmountalloc("nfs", path, &mp)) {
+		printf("nfs_mountroot: NFS not configured");
+		return (error);
+	}
+	mp->mnt_flag = mountflag;
+	MGET(m, MT_SONAME, M_WAITOK);
+	bcopy((caddr_t)sin, mtod(m, caddr_t), sin->sin_len);
+	m->m_len = sin->sin_len;
+	if (error = mountnfs(args, mp, m, which, path, vpp)) {
+		printf("nfs_mountroot: mount %s on %s: %d", path, which, error);
+		mp->mnt_vfc->vfc_refcount--;
+		vfs_unbusy(mp, p);
+		free(mp, M_MOUNT);
+		return (error);
+	}
+	(void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
+	*mpp = mp;
+	return (0);
+}
+
+/*
+ * VFS Operations.
+ *
+ * mount system call
+ * It seems a bit dumb to copyinstr() the host and path here and then
+ * bcopy() them in mountnfs(), but I wanted to detect errors before
+ * doing the sockargs() call because sockargs() allocates an mbuf and
+ * an error after that means that I have to release the mbuf.
+ */
+/* ARGSUSED */
+int
+nfs_mount(mp, path, data, ndp, p)
+	struct mount *mp;
+	char *path;
+	caddr_t data;
+	struct nameidata *ndp;
+	struct proc *p;
+{
+	int error;
+	struct nfs_args args;
+	struct mbuf *nam;
+	struct vnode *vp;
+	char pth[MNAMELEN], hst[MNAMELEN];
+	u_int len;
+	u_char nfh[NFSX_V3FHMAX];
+
+	error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
+	if (error)
+		return (error);
+	if (args.version != NFS_ARGSVERSION)
+		return (EPROGMISMATCH);
+	error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
+	if (error)
+		return (error);
+	error = copyinstr(path, pth, MNAMELEN-1, &len);
+	if (error)
+		return (error);
+	bzero(&pth[len], MNAMELEN - len);
+	error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
+	if (error)
+		return (error);
+	bzero(&hst[len], MNAMELEN - len);
+	/* sockargs() call must be after above copyin() calls */
+	error = sockargs(&nam, (caddr_t)args.addr, args.addrlen, MT_SONAME);
+	if (error)
+		return (error);
+	args.fh = nfh;
+	error = mountnfs(&args, mp, nam, pth, hst, &vp);
+	return (error);
+}
+
+/*
+ * Common code for mount and mountroot
+ */
+int
+mountnfs(argp, mp, nam, pth, hst, vpp)
+	register struct nfs_args *argp;
+	register struct mount *mp;
+	struct mbuf *nam;
+	char *pth, *hst;
+	struct vnode **vpp;
+{
+	register struct nfsmount *nmp;
+	struct nfsnode *np;
+	int error, maxio;
+
+	if (mp->mnt_flag & MNT_UPDATE) {
+		nmp = VFSTONFS(mp);
+		/* update paths, file handles, etc, here	XXX */
+		m_freem(nam);
+		return (0);
+	} else {
+		MALLOC(nmp, struct nfsmount *, sizeof (struct nfsmount),
+		    M_NFSMNT, M_WAITOK);
+		bzero((caddr_t)nmp, sizeof (struct nfsmount));
+		TAILQ_INIT(&nmp->nm_uidlruhead);
+		mp->mnt_data = (qaddr_t)nmp;
+	}
+	vfs_getnewfsid(mp);
+	nmp->nm_mountp = mp;
+	nmp->nm_flag = argp->flags;
+	if (nmp->nm_flag & NFSMNT_NQNFS)
+		/*
+		 * We have to set mnt_maxsymlink to a non-zero value so
+		 * that COMPAT_43 routines will know that we are setting
+		 * the d_type field in directories (and can zero it for
+		 * unsuspecting binaries).
+		 */
+		mp->mnt_maxsymlinklen = 1;
+	nmp->nm_timeo = NFS_TIMEO;
+	nmp->nm_retry = NFS_RETRANS;
+	nmp->nm_wsize = NFS_WSIZE;
+	nmp->nm_rsize = NFS_RSIZE;
+	nmp->nm_readdirsize = NFS_READDIRSIZE;
+	nmp->nm_numgrps = NFS_MAXGRPS;
+	nmp->nm_readahead = NFS_DEFRAHEAD;
+	nmp->nm_leaseterm = NQ_DEFLEASE;
+	nmp->nm_deadthresh = NQ_DEADTHRESH;
+	CIRCLEQ_INIT(&nmp->nm_timerhead);
+	nmp->nm_inprog = NULLVP;
+	nmp->nm_fhsize = argp->fhsize;
+	bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
+	bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
+	bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
+	nmp->nm_nam = nam;
+
+	if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
+		nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
+		if (nmp->nm_timeo < NFS_MINTIMEO)
+			nmp->nm_timeo = NFS_MINTIMEO;
+		else if (nmp->nm_timeo > NFS_MAXTIMEO)
+			nmp->nm_timeo = NFS_MAXTIMEO;
+	}
+
+	if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
+		nmp->nm_retry = argp->retrans;
+		if (nmp->nm_retry > NFS_MAXREXMIT)
+			nmp->nm_retry = NFS_MAXREXMIT;
+	}
+
+	if (argp->flags & NFSMNT_NFSV3) {
+		if (argp->sotype == SOCK_DGRAM)
+			maxio = NFS_MAXDGRAMDATA;
+		else
+			maxio = NFS_MAXDATA;
+	} else
+		maxio = NFS_V2MAXDATA;
+
+	if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
+		nmp->nm_wsize = argp->wsize;
+		/* Round down to multiple of blocksize */
+		nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
+		if (nmp->nm_wsize <= 0)
+			nmp->nm_wsize = NFS_FABLKSIZE;
+	}
+	if (nmp->nm_wsize > maxio)
+		nmp->nm_wsize = maxio;
+	if (nmp->nm_wsize > MAXBSIZE)
+		nmp->nm_wsize = MAXBSIZE;
+
+	if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
+		nmp->nm_rsize = argp->rsize;
+		/* Round down to multiple of blocksize */
+		nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
+		if (nmp->nm_rsize <= 0)
+			nmp->nm_rsize = NFS_FABLKSIZE;
+	}
+	if (nmp->nm_rsize > maxio)
+		nmp->nm_rsize = maxio;
+	if (nmp->nm_rsize > MAXBSIZE)
+		nmp->nm_rsize = MAXBSIZE;
+
+	if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
+		nmp->nm_readdirsize = argp->readdirsize;
+		/* Round down to multiple of blocksize */
+		nmp->nm_readdirsize &= ~(NFS_DIRBLKSIZ - 1);
+		if (nmp->nm_readdirsize < NFS_DIRBLKSIZ)
+			nmp->nm_readdirsize = NFS_DIRBLKSIZ;
+	}
+	if (nmp->nm_readdirsize > maxio)
+		nmp->nm_readdirsize = maxio;
+
+	if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0 &&
+		argp->maxgrouplist <= NFS_MAXGRPS)
+		nmp->nm_numgrps = argp->maxgrouplist;
+	if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0 &&
+		argp->readahead <= NFS_MAXRAHEAD)
+		nmp->nm_readahead = argp->readahead;
+	if ((argp->flags & NFSMNT_LEASETERM) && argp->leaseterm >= 2 &&
+		argp->leaseterm <= NQ_MAXLEASE)
+		nmp->nm_leaseterm = argp->leaseterm;
+	if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1 &&
+		argp->deadthresh <= NQ_NEVERDEAD)
+		nmp->nm_deadthresh = argp->deadthresh;
+	/* Set up the sockets and per-host congestion */
+	nmp->nm_sotype = argp->sotype;
+	nmp->nm_soproto = argp->proto;
+
+	/*
+	 * For Connection based sockets (TCP,...) defer the connect until
+	 * the first request, in case the server is not responding.
+	 */
+	if (nmp->nm_sotype == SOCK_DGRAM &&
+		(error = nfs_connect(nmp, (struct nfsreq *)0)))
+		goto bad;
+
+	/*
+	 * This is silly, but it has to be set so that vinifod() works.
+	 * We do not want to do an nfs_statfs() here since we can get
+	 * stuck on a dead server and we are holding a lock on the mount
+	 * point.
+	 */
+	mp->mnt_stat.f_iosize = NFS_MAXDGRAMDATA;
+	/*
+	 * A reference count is needed on the nfsnode representing the
+	 * remote root.  If this object is not persistent, then backward
+	 * traversals of the mount point (i.e. "..") will not work if
+	 * the nfsnode gets flushed out of the cache. Ufs does not have
+	 * this problem, because one can identify root inodes by their
+	 * number == ROOTINO (2).
+	 */
+	error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
+	if (error)
+		goto bad;
+	*vpp = NFSTOV(np);
+
+	return (0);
+bad:
+	nfs_disconnect(nmp);
+	free((caddr_t)nmp, M_NFSMNT);
+	m_freem(nam);
+	return (error);
+}
+
+/*
+ * unmount system call
+ */
+int
+nfs_unmount(mp, mntflags, p)
+	struct mount *mp;
+	int mntflags;
+	struct proc *p;
+{
+	register struct nfsmount *nmp;
+	struct nfsnode *np;
+	struct vnode *vp;
+	int error, flags = 0;
+
+	if (mntflags & MNT_FORCE)
+		flags |= FORCECLOSE;
+	nmp = VFSTONFS(mp);
+	/*
+	 * Goes something like this..
+	 * - Check for activity on the root vnode (other than ourselves).
+	 * - Call vflush() to clear out vnodes for this file system,
+	 *   except for the root vnode.
+	 * - Decrement reference on the vnode representing remote root.
+	 * - Close the socket
+	 * - Free up the data structures
+	 */
+	/*
+	 * We need to decrement the ref. count on the nfsnode representing
+	 * the remote root.  See comment in mountnfs().  The VFS unmount()
+	 * has done vput on this vnode, otherwise we would get deadlock!
+	 */
+	error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
+	if (error)
+		return(error);
+	vp = NFSTOV(np);
+	if (vp->v_usecount > 2) {
+		vput(vp);
+		return (EBUSY);
+	}
+
+	/*
+	 * Must handshake with nqnfs_clientd() if it is active.
+	 */
+	nmp->nm_flag |= NFSMNT_DISMINPROG;
+	while (nmp->nm_inprog != NULLVP)
+		(void) tsleep((caddr_t)&lbolt, PSOCK, "nfsdism", 0);
+	error = vflush(mp, vp, flags);
+	if (error) {
+		vput(vp);
+		nmp->nm_flag &= ~NFSMNT_DISMINPROG;
+		return (error);
+	}
+
+	/*
+	 * We are now committed to the unmount.
+	 * For NQNFS, let the server daemon free the nfsmount structure.
+	 */
+	if (nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB))
+		nmp->nm_flag |= NFSMNT_DISMNT;
+
+	/*
+	 * There are two reference counts to get rid of here.
+	 */
+	vrele(vp);
+	vrele(vp);
+	vgone(vp);
+	nfs_disconnect(nmp);
+	m_freem(nmp->nm_nam);
+
+	if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB)) == 0)
+		free((caddr_t)nmp, M_NFSMNT);
+	return (0);
+}
+
+/*
+ * Return root of a filesystem
+ */
+int
+nfs_root(mp, vpp)
+	struct mount *mp;
+	struct vnode **vpp;
+{
+	register struct vnode *vp;
+	struct nfsmount *nmp;
+	struct nfsnode *np;
+	int error;
+
+	nmp = VFSTONFS(mp);
+	error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
+	if (error)
+		return (error);
+	vp = NFSTOV(np);
+	vp->v_type = VDIR;
+	vp->v_flag = VROOT;
+	*vpp = vp;
+	return (0);
+}
+
+extern int syncprt;
+
+/*
+ * Flush out the buffer cache
+ */
+/* ARGSUSED */
+int
+nfs_sync(mp, waitfor, cred, p)
+	struct mount *mp;
+	int waitfor;
+	struct ucred *cred;
+	struct proc *p;
+{
+	register struct vnode *vp;
+	int error, allerror = 0;
+
+	/*
+	 * Force stale buffer cache information to be flushed.
+	 */
+loop:
+	for (vp = mp->mnt_vnodelist.lh_first;
+	     vp != NULL;
+	     vp = vp->v_mntvnodes.le_next) {
+		/*
+		 * If the vnode that we are about to sync is no longer
+		 * associated with this mount point, start over.
+		 */
+		if (vp->v_mount != mp)
+			goto loop;
+		if (VOP_ISLOCKED(vp) || vp->v_dirtyblkhd.lh_first == NULL)
+			continue;
+		if (vget(vp, LK_EXCLUSIVE, p))
+			goto loop;
+		error = VOP_FSYNC(vp, cred, waitfor, p);
+		if (error)
+			allerror = error;
+		vput(vp);
+	}
+	return (allerror);
+}
+
+/*
+ * NFS flat namespace lookup.
+ * Currently unsupported.
+ */
+/* ARGSUSED */
+int
+nfs_vget(mp, ino, vpp)
+	struct mount *mp;
+	ino_t ino;
+	struct vnode **vpp;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * At this point, this should never happen
+ */
+/* ARGSUSED */
+int
+nfs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
+	register struct mount *mp;
+	struct fid *fhp;
+	struct mbuf *nam;
+	struct vnode **vpp;
+	int *exflagsp;
+	struct ucred **credanonp;
+{
+
+	return (EINVAL);
+}
+
+/*
+ * Vnode pointer to File handle, should never happen either
+ */
+/* ARGSUSED */
+int
+nfs_vptofh(vp, fhp)
+	struct vnode *vp;
+	struct fid *fhp;
+{
+
+	return (EINVAL);
+}
+
+/*
+ * Vfs start routine, a no-op.
+ */
+/* ARGSUSED */
+int
+nfs_start(mp, flags, p)
+	struct mount *mp;
+	int flags;
+	struct proc *p;
+{
+
+	return (0);
+}
+
+/*
+ * Do operations associated with quotas, not supported
+ */
+/* ARGSUSED */
+int
+nfs_quotactl(mp, cmd, uid, arg, p)
+	struct mount *mp;
+	int cmd;
+	uid_t uid;
+	caddr_t arg;
+	struct proc *p;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * Do that sysctl thang...
+ */
+static int
+nfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
+	   size_t newlen, struct proc *p)
+{
+	int rv;
+
+	/*
+	 * All names at this level are terminal.
+	 */
+	if(namelen > 1)
+		return ENOTDIR;	/* overloaded */
+
+	switch(name[0]) {
+	case NFS_NFSSTATS:
+		if(!oldp) {
+			*oldlenp = sizeof nfsstats;
+			return 0;
+		}
+
+		if(*oldlenp < sizeof nfsstats) {
+			*oldlenp = sizeof nfsstats;
+			return ENOMEM;
+		}
+
+		rv = copyout(&nfsstats, oldp, sizeof nfsstats);
+		if(rv) return rv;
+
+		if(newp && newlen != sizeof nfsstats)
+			return EINVAL;
+
+		if(newp) {
+			return copyin(newp, &nfsstats, sizeof nfsstats);
+		}
+		return 0;
+
+	default:
+		return EOPNOTSUPP;
+	}
+}
+
diff --git a/sys/nfs/nfs_vnops.c b/sys/nfs/nfs_vnops.c
new file mode 100644
index 000000000000..c72f573b9ea5
--- /dev/null
+++ b/sys/nfs/nfs_vnops.c
@@ -0,0 +1,3250 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfs_vnops.c	8.16 (Berkeley) 5/27/95
+ */
+
+
+/*
+ * vnode op calls for Sun NFS version 2 and 3
+ */
+
+#include <sys/param.h>
+#include <sys/kernel.h>
+#include <sys/systm.h>
+#include <sys/resourcevar.h>
+#include <sys/proc.h>
+#include <sys/mount.h>
+#include <sys/buf.h>
+#include <sys/malloc.h>
+#include <sys/mbuf.h>
+#include <sys/conf.h>
+#include <sys/namei.h>
+#include <sys/vnode.h>
+#include <sys/dirent.h>
+#include <sys/fcntl.h>
+#include <ufs/ufs/dir.h>
+
+#include <vm/vm.h>
+
+#include <miscfs/specfs/specdev.h>
+#include <miscfs/fifofs/fifo.h>
+
+#include <nfs/rpcv2.h>
+#include <nfs/nfsproto.h>
+#include <nfs/nfs.h>
+#include <nfs/nfsnode.h>
+#include <nfs/nfsmount.h>
+#include <nfs/xdr_subs.h>
+#include <nfs/nfsm_subs.h>
+#include <nfs/nqnfs.h>
+
+#include <net/if.h>
+#include <netinet/in.h>
+#include <netinet/in_var.h>
+
+/* Defs */
+#define	TRUE	1
+#define	FALSE	0
+
+/*
+ * Global vfs data structures for nfs
+ */
+int (**nfsv2_vnodeop_p)();
+struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = {
+	{ &vop_default_desc, vn_default_error },
+	{ &vop_lookup_desc, nfs_lookup },	/* lookup */
+	{ &vop_create_desc, nfs_create },	/* create */
+	{ &vop_mknod_desc, nfs_mknod },		/* mknod */
+	{ &vop_open_desc, nfs_open },		/* open */
+	{ &vop_close_desc, nfs_close },		/* close */
+	{ &vop_access_desc, nfs_access },	/* access */
+	{ &vop_getattr_desc, nfs_getattr },	/* getattr */
+	{ &vop_setattr_desc, nfs_setattr },	/* setattr */
+	{ &vop_read_desc, nfs_read },		/* read */
+	{ &vop_write_desc, nfs_write },		/* write */
+	{ &vop_lease_desc, nfs_lease_check },	/* lease */
+	{ &vop_ioctl_desc, nfs_ioctl },		/* ioctl */
+	{ &vop_select_desc, nfs_select },	/* select */
+	{ &vop_revoke_desc, nfs_revoke },	/* revoke */
+	{ &vop_mmap_desc, nfs_mmap },		/* mmap */
+	{ &vop_fsync_desc, nfs_fsync },		/* fsync */
+	{ &vop_seek_desc, nfs_seek },		/* seek */
+	{ &vop_remove_desc, nfs_remove },	/* remove */
+	{ &vop_link_desc, nfs_link },		/* link */
+	{ &vop_rename_desc, nfs_rename },	/* rename */
+	{ &vop_mkdir_desc, nfs_mkdir },		/* mkdir */
+	{ &vop_rmdir_desc, nfs_rmdir },		/* rmdir */
+	{ &vop_symlink_desc, nfs_symlink },	/* symlink */
+	{ &vop_readdir_desc, nfs_readdir },	/* readdir */
+	{ &vop_readlink_desc, nfs_readlink },	/* readlink */
+	{ &vop_abortop_desc, nfs_abortop },	/* abortop */
+	{ &vop_inactive_desc, nfs_inactive },	/* inactive */
+	{ &vop_reclaim_desc, nfs_reclaim },	/* reclaim */
+	{ &vop_lock_desc, nfs_lock },		/* lock */
+	{ &vop_unlock_desc, nfs_unlock },	/* unlock */
+	{ &vop_bmap_desc, nfs_bmap },		/* bmap */
+	{ &vop_strategy_desc, nfs_strategy },	/* strategy */
+	{ &vop_print_desc, nfs_print },		/* print */
+	{ &vop_islocked_desc, nfs_islocked },	/* islocked */
+	{ &vop_pathconf_desc, nfs_pathconf },	/* pathconf */
+	{ &vop_advlock_desc, nfs_advlock },	/* advlock */
+	{ &vop_blkatoff_desc, nfs_blkatoff },	/* blkatoff */
+	{ &vop_valloc_desc, nfs_valloc },	/* valloc */
+	{ &vop_reallocblks_desc, nfs_reallocblks },	/* reallocblks */
+	{ &vop_vfree_desc, nfs_vfree },		/* vfree */
+	{ &vop_truncate_desc, nfs_truncate },	/* truncate */
+	{ &vop_update_desc, nfs_update },	/* update */
+	{ &vop_bwrite_desc, nfs_bwrite },
+	{ (struct vnodeop_desc*)NULL, (int(*)())NULL }
+};
+struct vnodeopv_desc nfsv2_vnodeop_opv_desc =
+	{ &nfsv2_vnodeop_p, nfsv2_vnodeop_entries };
+
+/*
+ * Special device vnode ops
+ */
+int (**spec_nfsv2nodeop_p)();
+struct vnodeopv_entry_desc spec_nfsv2nodeop_entries[] = {
+	{ &vop_default_desc, vn_default_error },
+	{ &vop_lookup_desc, spec_lookup },	/* lookup */
+	{ &vop_create_desc, spec_create },	/* create */
+	{ &vop_mknod_desc, spec_mknod },	/* mknod */
+	{ &vop_open_desc, spec_open },		/* open */
+	{ &vop_close_desc, nfsspec_close },	/* close */
+	{ &vop_access_desc, nfsspec_access },	/* access */
+	{ &vop_getattr_desc, nfs_getattr },	/* getattr */
+	{ &vop_setattr_desc, nfs_setattr },	/* setattr */
+	{ &vop_read_desc, nfsspec_read },	/* read */
+	{ &vop_write_desc, nfsspec_write },	/* write */
+	{ &vop_lease_desc, spec_lease_check },	/* lease */
+	{ &vop_ioctl_desc, spec_ioctl },	/* ioctl */
+	{ &vop_select_desc, spec_select },	/* select */
+	{ &vop_revoke_desc, spec_revoke },	/* revoke */
+	{ &vop_mmap_desc, spec_mmap },		/* mmap */
+	{ &vop_fsync_desc, nfs_fsync },		/* fsync */
+	{ &vop_seek_desc, spec_seek },		/* seek */
+	{ &vop_remove_desc, spec_remove },	/* remove */
+	{ &vop_link_desc, spec_link },		/* link */
+	{ &vop_rename_desc, spec_rename },	/* rename */
+	{ &vop_mkdir_desc, spec_mkdir },	/* mkdir */
+	{ &vop_rmdir_desc, spec_rmdir },	/* rmdir */
+	{ &vop_symlink_desc, spec_symlink },	/* symlink */
+	{ &vop_readdir_desc, spec_readdir },	/* readdir */
+	{ &vop_readlink_desc, spec_readlink },	/* readlink */
+	{ &vop_abortop_desc, spec_abortop },	/* abortop */
+	{ &vop_inactive_desc, nfs_inactive },	/* inactive */
+	{ &vop_reclaim_desc, nfs_reclaim },	/* reclaim */
+	{ &vop_lock_desc, nfs_lock },		/* lock */
+	{ &vop_unlock_desc, nfs_unlock },	/* unlock */
+	{ &vop_bmap_desc, spec_bmap },		/* bmap */
+	{ &vop_strategy_desc, spec_strategy },	/* strategy */
+	{ &vop_print_desc, nfs_print },		/* print */
+	{ &vop_islocked_desc, nfs_islocked },	/* islocked */
+	{ &vop_pathconf_desc, spec_pathconf },	/* pathconf */
+	{ &vop_advlock_desc, spec_advlock },	/* advlock */
+	{ &vop_blkatoff_desc, spec_blkatoff },	/* blkatoff */
+	{ &vop_valloc_desc, spec_valloc },	/* valloc */
+	{ &vop_reallocblks_desc, spec_reallocblks },	/* reallocblks */
+	{ &vop_vfree_desc, spec_vfree },	/* vfree */
+	{ &vop_truncate_desc, spec_truncate },	/* truncate */
+	{ &vop_update_desc, nfs_update },	/* update */
+	{ &vop_bwrite_desc, vn_bwrite },
+	{ (struct vnodeop_desc*)NULL, (int(*)())NULL }
+};
+struct vnodeopv_desc spec_nfsv2nodeop_opv_desc =
+	{ &spec_nfsv2nodeop_p, spec_nfsv2nodeop_entries };
+
+int (**fifo_nfsv2nodeop_p)();
+struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = {
+	{ &vop_default_desc, vn_default_error },
+	{ &vop_lookup_desc, fifo_lookup },	/* lookup */
+	{ &vop_create_desc, fifo_create },	/* create */
+	{ &vop_mknod_desc, fifo_mknod },	/* mknod */
+	{ &vop_open_desc, fifo_open },		/* open */
+	{ &vop_close_desc, nfsfifo_close },	/* close */
+	{ &vop_access_desc, nfsspec_access },	/* access */
+	{ &vop_getattr_desc, nfs_getattr },	/* getattr */
+	{ &vop_setattr_desc, nfs_setattr },	/* setattr */
+	{ &vop_read_desc, nfsfifo_read },	/* read */
+	{ &vop_write_desc, nfsfifo_write },	/* write */
+	{ &vop_lease_desc, fifo_lease_check },	/* lease */
+	{ &vop_ioctl_desc, fifo_ioctl },	/* ioctl */
+	{ &vop_select_desc, fifo_select },	/* select */
+	{ &vop_revoke_desc, fifo_revoke },	/* revoke */
+	{ &vop_mmap_desc, fifo_mmap },		/* mmap */
+	{ &vop_fsync_desc, nfs_fsync },		/* fsync */
+	{ &vop_seek_desc, fifo_seek },		/* seek */
+	{ &vop_remove_desc, fifo_remove },	/* remove */
+	{ &vop_link_desc, fifo_link },		/* link */
+	{ &vop_rename_desc, fifo_rename },	/* rename */
+	{ &vop_mkdir_desc, fifo_mkdir },	/* mkdir */
+	{ &vop_rmdir_desc, fifo_rmdir },	/* rmdir */
+	{ &vop_symlink_desc, fifo_symlink },	/* symlink */
+	{ &vop_readdir_desc, fifo_readdir },	/* readdir */
+	{ &vop_readlink_desc, fifo_readlink },	/* readlink */
+	{ &vop_abortop_desc, fifo_abortop },	/* abortop */
+	{ &vop_inactive_desc, nfs_inactive },	/* inactive */
+	{ &vop_reclaim_desc, nfs_reclaim },	/* reclaim */
+	{ &vop_lock_desc, nfs_lock },		/* lock */
+	{ &vop_unlock_desc, nfs_unlock },	/* unlock */
+	{ &vop_bmap_desc, fifo_bmap },		/* bmap */
+	{ &vop_strategy_desc, fifo_badop },	/* strategy */
+	{ &vop_print_desc, nfs_print },		/* print */
+	{ &vop_islocked_desc, nfs_islocked },	/* islocked */
+	{ &vop_pathconf_desc, fifo_pathconf },	/* pathconf */
+	{ &vop_advlock_desc, fifo_advlock },	/* advlock */
+	{ &vop_blkatoff_desc, fifo_blkatoff },	/* blkatoff */
+	{ &vop_valloc_desc, fifo_valloc },	/* valloc */
+	{ &vop_reallocblks_desc, fifo_reallocblks },	/* reallocblks */
+	{ &vop_vfree_desc, fifo_vfree },	/* vfree */
+	{ &vop_truncate_desc, fifo_truncate },	/* truncate */
+	{ &vop_update_desc, nfs_update },	/* update */
+	{ &vop_bwrite_desc, vn_bwrite },
+	{ (struct vnodeop_desc*)NULL, (int(*)())NULL }
+};
+struct vnodeopv_desc fifo_nfsv2nodeop_opv_desc =
+	{ &fifo_nfsv2nodeop_p, fifo_nfsv2nodeop_entries };
+
+void nqnfs_clientlease();
+int nfs_commit();
+
+/*
+ * Global variables
+ */
+extern u_long nfs_true, nfs_false;
+extern struct nfsstats nfsstats;
+extern nfstype nfsv3_type[9];
+struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON];
+int nfs_numasync = 0;
+#define	DIRHDSIZ	(sizeof (struct dirent) - (MAXNAMLEN + 1))
+
+/*
+ * nfs null call from vfs.
+ */
+int
+nfs_null(vp, cred, procp)
+	struct vnode *vp;
+	struct ucred *cred;
+	struct proc *procp;
+{
+	caddr_t bpos, dpos;
+	int error = 0;
+	struct mbuf *mreq, *mrep, *md, *mb;
+	
+	nfsm_reqhead(vp, NFSPROC_NULL, 0);
+	nfsm_request(vp, NFSPROC_NULL, procp, cred);
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * nfs access vnode op.
+ * For nfs version 2, just return ok. File accesses may fail later.
+ * For nfs version 3, use the access rpc to check accessibility. If file modes
+ * are changed on the server, accesses might still fail later.
+ */
+int
+nfs_access(ap)
+	struct vop_access_args /* {
+		struct vnode *a_vp;
+		int  a_mode;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register u_long *tl;
+	register caddr_t cp;
+	register int t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, attrflag;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	u_long mode, rmode;
+	int v3 = NFS_ISV3(vp);
+
+	/*
+	 * Disallow write attempts on filesystems mounted read-only;
+	 * unless the file is a socket, fifo, or a block or character
+	 * device resident on the filesystem.
+	 */
+	if ((ap->a_mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
+		switch (vp->v_type) {
+		case VREG: case VDIR: case VLNK:
+			return (EROFS);
+		}
+	}
+	/*
+	 * For nfs v3, do an access rpc, otherwise you are stuck emulating
+	 * ufs_access() locally using the vattr. This may not be correct,
+	 * since the server may apply other access criteria such as
+	 * client uid-->server uid mapping that we do not know about, but
+	 * this is better than just returning anything that is lying about
+	 * in the cache.
+	 */
+	if (v3) {
+		nfsstats.rpccnt[NFSPROC_ACCESS]++;
+		nfsm_reqhead(vp, NFSPROC_ACCESS, NFSX_FH(v3) + NFSX_UNSIGNED);
+		nfsm_fhtom(vp, v3);
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+		if (ap->a_mode & VREAD)
+			mode = NFSV3ACCESS_READ;
+		else
+			mode = 0;
+		if (vp->v_type == VDIR) {
+			if (ap->a_mode & VWRITE)
+				mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
+					 NFSV3ACCESS_DELETE);
+			if (ap->a_mode & VEXEC)
+				mode |= NFSV3ACCESS_LOOKUP;
+		} else {
+			if (ap->a_mode & VWRITE)
+				mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
+			if (ap->a_mode & VEXEC)
+				mode |= NFSV3ACCESS_EXECUTE;
+		}
+		*tl = txdr_unsigned(mode);
+		nfsm_request(vp, NFSPROC_ACCESS, ap->a_p, ap->a_cred);
+		nfsm_postop_attr(vp, attrflag);
+		if (!error) {
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			rmode = fxdr_unsigned(u_long, *tl);
+			/*
+			 * The NFS V3 spec does not clarify whether or not
+			 * the returned access bits can be a superset of
+			 * the ones requested, so...
+			 */
+			if ((rmode & mode) != mode)
+				error = EACCES;
+		}
+		nfsm_reqdone;
+		return (error);
+	} else
+		return (nfsspec_access(ap));
+}
+
+/*
+ * nfs open vnode op
+ * Check to see if the type is ok
+ * and that deletion is not in progress.
+ * For paged in text files, you will need to flush the page cache
+ * if consistency is lost.
+ */
+/* ARGSUSED */
+int
+nfs_open(ap)
+	struct vop_open_args /* {
+		struct vnode *a_vp;
+		int  a_mode;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	struct nfsnode *np = VTONFS(vp);
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	struct vattr vattr;
+	int error;
+
+	if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK)
+{ printf("open eacces vtyp=%d\n",vp->v_type);
+		return (EACCES);
+}
+	/*
+	 * Get a valid lease. If cached data is stale, flush it.
+	 */
+	if (nmp->nm_flag & NFSMNT_NQNFS) {
+		if (NQNFS_CKINVALID(vp, np, ND_READ)) {
+		    do {
+			error = nqnfs_getlease(vp, ND_READ, ap->a_cred,
+			    ap->a_p);
+		    } while (error == NQNFS_EXPIRED);
+		    if (error)
+			return (error);
+		    if (np->n_lrev != np->n_brev ||
+			(np->n_flag & NQNFSNONCACHE)) {
+			if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
+				ap->a_p, 1)) == EINTR)
+				return (error);
+			(void) vnode_pager_uncache(vp);
+			np->n_brev = np->n_lrev;
+		    }
+		}
+	} else {
+		if (np->n_flag & NMODIFIED) {
+			if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
+				ap->a_p, 1)) == EINTR)
+				return (error);
+			(void) vnode_pager_uncache(vp);
+			np->n_attrstamp = 0;
+			if (vp->v_type == VDIR)
+				np->n_direofoffset = 0;
+			error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p);
+			if (error)
+				return (error);
+			np->n_mtime = vattr.va_mtime.ts_sec;
+		} else {
+			error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p);
+			if (error)
+				return (error);
+			if (np->n_mtime != vattr.va_mtime.ts_sec) {
+				if (vp->v_type == VDIR)
+					np->n_direofoffset = 0;
+				if ((error = nfs_vinvalbuf(vp, V_SAVE,
+					ap->a_cred, ap->a_p, 1)) == EINTR)
+					return (error);
+				(void) vnode_pager_uncache(vp);
+				np->n_mtime = vattr.va_mtime.ts_sec;
+			}
+		}
+	}
+	if ((nmp->nm_flag & NFSMNT_NQNFS) == 0)
+		np->n_attrstamp = 0; /* For Open/Close consistency */
+	return (0);
+}
+
+/*
+ * nfs close vnode op
+ * What an NFS client should do upon close after writing is a debatable issue.
+ * Most NFS clients push delayed writes to the server upon close, basically for
+ * two reasons:
+ * 1 - So that any write errors may be reported back to the client process
+ *     doing the close system call. By far the two most likely errors are
+ *     NFSERR_NOSPC and NFSERR_DQUOT to indicate space allocation failure.
+ * 2 - To put a worst case upper bound on cache inconsistency between
+ *     multiple clients for the file.
+ * There is also a consistency problem for Version 2 of the protocol w.r.t.
+ * not being able to tell if other clients are writing a file concurrently,
+ * since there is no way of knowing if the changed modify time in the reply
+ * is only due to the write for this client.
+ * (NFS Version 3 provides weak cache consistency data in the reply that
+ *  should be sufficient to detect and handle this case.)
+ *
+ * The current code does the following:
+ * for NFS Version 2 - play it safe and flush/invalidate all dirty buffers
+ * for NFS Version 3 - flush dirty buffers to the server but don't invalidate
+ *                     or commit them (this satisfies 1 and 2 except for the
+ *                     case where the server crashes after this close but
+ *                     before the commit RPC, which is felt to be "good
+ *                     enough". Changing the last argument to nfs_flush() to
+ *                     a 1 would force a commit operation, if it is felt a
+ *                     commit is necessary now.
+ * for NQNFS         - do nothing now, since 2 is dealt with via leases and
+ *                     1 should be dealt with via an fsync() system call for
+ *                     cases where write errors are important.
+ */
+/* ARGSUSED */
+int
+nfs_close(ap)
+	struct vop_close_args /* {
+		struct vnodeop_desc *a_desc;
+		struct vnode *a_vp;
+		int  a_fflag;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	int error = 0;
+
+	if (vp->v_type == VREG) {
+	    if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) == 0 &&
+		(np->n_flag & NMODIFIED)) {
+		if (NFS_ISV3(vp))
+		    error = nfs_flush(vp, ap->a_cred, MNT_WAIT, ap->a_p, 0);
+		else
+		    error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1);
+		np->n_attrstamp = 0;
+	    }
+	    if (np->n_flag & NWRITEERR) {
+		np->n_flag &= ~NWRITEERR;
+		error = np->n_error;
+	    }
+	}
+	return (error);
+}
+
+/*
+ * nfs getattr call from vfs.
+ */
+int
+nfs_getattr(ap)
+	struct vop_getattr_args /* {
+		struct vnode *a_vp;
+		struct vattr *a_vap;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	register caddr_t cp;
+	register u_long *tl;
+	register int t1, t2;
+	caddr_t bpos, dpos;
+	int error = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(vp);
+	
+	/*
+	 * Update local times for special files.
+	 */
+	if (np->n_flag & (NACC | NUPD))
+		np->n_flag |= NCHG;
+	/*
+	 * First look in the cache.
+	 */
+	if (nfs_getattrcache(vp, ap->a_vap) == 0)
+		return (0);
+	nfsstats.rpccnt[NFSPROC_GETATTR]++;
+	nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH(v3));
+	nfsm_fhtom(vp, v3);
+	nfsm_request(vp, NFSPROC_GETATTR, ap->a_p, ap->a_cred);
+	if (!error)
+		nfsm_loadattr(vp, ap->a_vap);
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * nfs setattr call.
+ */
+int
+nfs_setattr(ap)
+	struct vop_setattr_args /* {
+		struct vnodeop_desc *a_desc;
+		struct vnode *a_vp;
+		struct vattr *a_vap;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	register struct vattr *vap = ap->a_vap;
+	int error = 0;
+	u_quad_t tsize;
+
+#ifndef nolint
+	tsize = (u_quad_t)0;
+#endif
+	/*
+	 * Disallow write attempts if the filesystem is mounted read-only.
+	 */
+  	if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL ||
+	    vap->va_gid != (gid_t)VNOVAL || vap->va_atime.ts_sec != VNOVAL ||
+	    vap->va_mtime.ts_sec != VNOVAL || vap->va_mode != (mode_t)VNOVAL) &&
+	    (vp->v_mount->mnt_flag & MNT_RDONLY))
+		return (EROFS);
+	if (vap->va_size != VNOVAL) {
+ 		switch (vp->v_type) {
+ 		case VDIR:
+ 			return (EISDIR);
+ 		case VCHR:
+ 		case VBLK:
+ 		case VSOCK:
+ 		case VFIFO:
+			if (vap->va_mtime.ts_sec == VNOVAL &&
+			    vap->va_atime.ts_sec == VNOVAL &&
+			    vap->va_mode == (u_short)VNOVAL &&
+			    vap->va_uid == (uid_t)VNOVAL &&
+			    vap->va_gid == (gid_t)VNOVAL)
+				return (0);
+ 			vap->va_size = VNOVAL;
+ 			break;
+ 		default:
+			/*
+			 * Disallow write attempts if the filesystem is
+			 * mounted read-only.
+			 */
+			if (vp->v_mount->mnt_flag & MNT_RDONLY)
+				return (EROFS);
+ 			if (np->n_flag & NMODIFIED) {
+ 			    if (vap->va_size == 0)
+ 				error = nfs_vinvalbuf(vp, 0,
+ 					ap->a_cred, ap->a_p, 1);
+ 			    else
+ 				error = nfs_vinvalbuf(vp, V_SAVE,
+ 					ap->a_cred, ap->a_p, 1);
+ 			    if (error)
+ 				return (error);
+ 			}
+ 			tsize = np->n_size;
+ 			np->n_size = np->n_vattr.va_size = vap->va_size;
+ 			vnode_pager_setsize(vp, (u_long)np->n_size);
+  		};
+  	} else if ((vap->va_mtime.ts_sec != VNOVAL ||
+		vap->va_atime.ts_sec != VNOVAL) && (np->n_flag & NMODIFIED) &&
+		vp->v_type == VREG &&
+  		(error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
+		 ap->a_p, 1)) == EINTR)
+		return (error);
+	error = nfs_setattrrpc(vp, vap, ap->a_cred, ap->a_p);
+	if (error) {
+		np->n_size = np->n_vattr.va_size = tsize;
+		vnode_pager_setsize(vp, (u_long)np->n_size);
+	}
+	return (error);
+}
+
+/*
+ * Do an nfs setattr rpc.
+ */
+int
+nfs_setattrrpc(vp, vap, cred, procp)
+	register struct vnode *vp;
+	register struct vattr *vap;
+	struct ucred *cred;
+	struct proc *procp;
+{
+	register struct nfsv2_sattr *sp;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	u_long *tl;
+	int error = 0, wccflag = NFSV3_WCCRATTR;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	u_quad_t frev;
+	int v3 = NFS_ISV3(vp);
+
+	nfsstats.rpccnt[NFSPROC_SETATTR]++;
+	nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH(v3) + NFSX_SATTR(v3));
+	nfsm_fhtom(vp, v3);
+	if (v3) {
+		if (vap->va_mode != (u_short)VNOVAL) {
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = nfs_true;
+			*tl = txdr_unsigned(vap->va_mode);
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = nfs_false;
+		}
+		if (vap->va_uid != (uid_t)VNOVAL) {
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = nfs_true;
+			*tl = txdr_unsigned(vap->va_uid);
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = nfs_false;
+		}
+		if (vap->va_gid != (gid_t)VNOVAL) {
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = nfs_true;
+			*tl = txdr_unsigned(vap->va_gid);
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = nfs_false;
+		}
+		if (vap->va_size != VNOVAL) {
+			nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED);
+			*tl++ = nfs_true;
+			txdr_hyper(&vap->va_size, tl);
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = nfs_false;
+		}
+		if (vap->va_atime.ts_sec != VNOVAL) {
+			if (vap->va_atime.ts_sec != time.tv_sec) {
+				nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED);
+				*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
+				txdr_nfsv3time(&vap->va_atime, tl);
+			} else {
+				nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+				*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
+			}
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
+		}
+		if (vap->va_mtime.ts_sec != VNOVAL) {
+			if (vap->va_mtime.ts_sec != time.tv_sec) {
+				nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED);
+				*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
+				txdr_nfsv3time(&vap->va_atime, tl);
+			} else {
+				nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+				*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
+			}
+		} else {
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+			*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
+		}
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+		*tl = nfs_false;
+	} else {
+		nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		if (vap->va_mode == (u_short)VNOVAL)
+			sp->sa_mode = VNOVAL;
+		else
+			sp->sa_mode = vtonfsv2_mode(vp->v_type, vap->va_mode);
+		if (vap->va_uid == (uid_t)VNOVAL)
+			sp->sa_uid = VNOVAL;
+		else
+			sp->sa_uid = txdr_unsigned(vap->va_uid);
+		if (vap->va_gid == (gid_t)VNOVAL)
+			sp->sa_gid = VNOVAL;
+		else
+			sp->sa_gid = txdr_unsigned(vap->va_gid);
+		sp->sa_size = txdr_unsigned(vap->va_size);
+		txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
+		txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
+	}
+	nfsm_request(vp, NFSPROC_SETATTR, procp, cred);
+	if (v3) {
+		nfsm_wcc_data(vp, wccflag);
+	} else
+		nfsm_loadattr(vp, (struct vattr *)0);
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * nfs lookup call, one step at a time...
+ * First look in cache
+ * If not found, unlock the directory nfsnode and do the rpc
+ */
+int
+nfs_lookup(ap)
+	struct vop_lookup_args /* {
+		struct vnodeop_desc *a_desc;
+		struct vnode *a_dvp;
+		struct vnode **a_vpp;
+		struct componentname *a_cnp;
+	} */ *ap;
+{
+	register struct componentname *cnp = ap->a_cnp;
+	register struct vnode *dvp = ap->a_dvp;
+	register struct vnode **vpp = ap->a_vpp;
+	register int flags = cnp->cn_flags;
+	register struct proc *p = cnp->cn_proc;
+	register struct vnode *newvp;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	struct nfsmount *nmp;
+	caddr_t bpos, dpos, cp2;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	long len;
+	nfsfh_t *fhp;
+	struct nfsnode *np;
+	int lockparent, wantparent, error = 0, attrflag, fhsize;
+	int v3 = NFS_ISV3(dvp);
+
+	if ((flags & ISLASTCN) && (dvp->v_mount->mnt_flag & MNT_RDONLY) &&
+	    (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
+		return (EROFS);
+	*vpp = NULLVP;
+	if (dvp->v_type != VDIR)
+		return (ENOTDIR);
+	lockparent = flags & LOCKPARENT;
+	wantparent = flags & (LOCKPARENT|WANTPARENT);
+	nmp = VFSTONFS(dvp->v_mount);
+	np = VTONFS(dvp);
+	if ((error = cache_lookup(dvp, vpp, cnp)) && error != ENOENT) {
+		struct vattr vattr;
+		int vpid;
+
+		newvp = *vpp;
+		vpid = newvp->v_id;
+		/*
+		 * See the comment starting `Step through' in ufs/ufs_lookup.c
+		 * for an explanation of the locking protocol
+		 */
+		if (dvp == newvp) {
+			VREF(newvp);
+			error = 0;
+		} else
+			error = vget(newvp, LK_EXCLUSIVE, p);
+		if (!error) {
+			if (vpid == newvp->v_id) {
+			   if (!VOP_GETATTR(newvp, &vattr, cnp->cn_cred, cnp->cn_proc)
+			    && vattr.va_ctime.ts_sec == VTONFS(newvp)->n_ctime) {
+				nfsstats.lookupcache_hits++;
+				if (cnp->cn_nameiop != LOOKUP &&
+				    (flags & ISLASTCN))
+					cnp->cn_flags |= SAVENAME;
+				return (0);
+			   }
+			   cache_purge(newvp);
+			}
+			vrele(newvp);
+		}
+		*vpp = NULLVP;
+	}
+	error = 0;
+	newvp = NULLVP;
+	nfsstats.lookupcache_misses++;
+	nfsstats.rpccnt[NFSPROC_LOOKUP]++;
+	len = cnp->cn_namelen;
+	nfsm_reqhead(dvp, NFSPROC_LOOKUP,
+		NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN);
+	nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred);
+	if (error) {
+		nfsm_postop_attr(dvp, attrflag);
+		m_freem(mrep);
+		goto nfsmout;
+	}
+	nfsm_getfh(fhp, fhsize, v3);
+
+	/*
+	 * Handle RENAME case...
+	 */
+	if (cnp->cn_nameiop == RENAME && wantparent && (flags & ISLASTCN)) {
+		if (NFS_CMPFH(np, fhp, fhsize)) {
+			m_freem(mrep);
+			return (EISDIR);
+		}
+		if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) {
+			m_freem(mrep);
+			return (error);
+		}
+		newvp = NFSTOV(np);
+		if (v3) {
+			nfsm_postop_attr(newvp, attrflag);
+			nfsm_postop_attr(dvp, attrflag);
+		} else
+			nfsm_loadattr(newvp, (struct vattr *)0);
+		*vpp = newvp;
+		m_freem(mrep);
+		cnp->cn_flags |= SAVENAME;
+		return (0);
+	}
+
+	if (NFS_CMPFH(np, fhp, fhsize)) {
+		VREF(dvp);
+		newvp = dvp;
+	} else {
+		if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) {
+			m_freem(mrep);
+			return (error);
+		}
+		newvp = NFSTOV(np);
+	}
+	if (v3) {
+		nfsm_postop_attr(newvp, attrflag);
+		nfsm_postop_attr(dvp, attrflag);
+	} else
+		nfsm_loadattr(newvp, (struct vattr *)0);
+	if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN))
+		cnp->cn_flags |= SAVENAME;
+	if ((cnp->cn_flags & MAKEENTRY) &&
+	    (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN))) {
+		np->n_ctime = np->n_vattr.va_ctime.ts_sec;
+		cache_enter(dvp, newvp, cnp);
+	}
+	*vpp = newvp;
+	nfsm_reqdone;
+	if (error) {
+		if (newvp != NULLVP)
+			vrele(newvp);
+		if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) &&
+		    (flags & ISLASTCN) && error == ENOENT) {
+			if (dvp->v_mount->mnt_flag & MNT_RDONLY)
+				error = EROFS;
+			else
+				error = EJUSTRETURN;
+		}
+		if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN))
+			cnp->cn_flags |= SAVENAME;
+	}
+	return (error);
+}
+
+/*
+ * nfs read call.
+ * Just call nfs_bioread() to do the work.
+ */
+int
+nfs_read(ap)
+	struct vop_read_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+
+	if (vp->v_type != VREG)
+		return (EPERM);
+	return (nfs_bioread(vp, ap->a_uio, ap->a_ioflag, ap->a_cred));
+}
+
+/*
+ * nfs readlink call
+ */
+int
+nfs_readlink(ap)
+	struct vop_readlink_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+
+	if (vp->v_type != VLNK)
+		return (EPERM);
+	return (nfs_bioread(vp, ap->a_uio, 0, ap->a_cred));
+}
+
+/*
+ * Do a readlink rpc.
+ * Called by nfs_doio() from below the buffer cache.
+ */
+int
+nfs_readlinkrpc(vp, uiop, cred)
+	register struct vnode *vp;
+	struct uio *uiop;
+	struct ucred *cred;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, len, attrflag;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(vp);
+
+	nfsstats.rpccnt[NFSPROC_READLINK]++;
+	nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH(v3));
+	nfsm_fhtom(vp, v3);
+	nfsm_request(vp, NFSPROC_READLINK, uiop->uio_procp, cred);
+	if (v3)
+		nfsm_postop_attr(vp, attrflag);
+	if (!error) {
+		nfsm_strsiz(len, NFS_MAXPATHLEN);
+		nfsm_mtouio(uiop, len);
+	}
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * nfs read rpc call
+ * Ditto above
+ */
+int
+nfs_readrpc(vp, uiop, cred)
+	register struct vnode *vp;
+	struct uio *uiop;
+	struct ucred *cred;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct nfsmount *nmp;
+	int error = 0, len, retlen, tsiz, eof, attrflag;
+	int v3 = NFS_ISV3(vp);
+
+#ifndef nolint
+	eof = 0;
+#endif
+	nmp = VFSTONFS(vp->v_mount);
+	tsiz = uiop->uio_resid;
+	if (uiop->uio_offset + tsiz > 0xffffffff && !v3)
+		return (EFBIG);
+	while (tsiz > 0) {
+		nfsstats.rpccnt[NFSPROC_READ]++;
+		len = (tsiz > nmp->nm_rsize) ? nmp->nm_rsize : tsiz;
+		nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH(v3) + NFSX_UNSIGNED * 3);
+		nfsm_fhtom(vp, v3);
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED * 3);
+		if (v3) {
+			txdr_hyper(&uiop->uio_offset, tl);
+			*(tl + 2) = txdr_unsigned(len);
+		} else {
+			*tl++ = txdr_unsigned(uiop->uio_offset);
+			*tl++ = txdr_unsigned(len);
+			*tl = 0;
+		}
+		nfsm_request(vp, NFSPROC_READ, uiop->uio_procp, cred);
+		if (v3) {
+			nfsm_postop_attr(vp, attrflag);
+			if (error) {
+				m_freem(mrep);
+				goto nfsmout;
+			}
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+			eof = fxdr_unsigned(int, *(tl + 1));
+		} else
+			nfsm_loadattr(vp, (struct vattr *)0);
+		nfsm_strsiz(retlen, nmp->nm_rsize);
+		nfsm_mtouio(uiop, retlen);
+		m_freem(mrep);
+		tsiz -= retlen;
+		if (v3) {
+			if (eof || retlen == 0)
+				tsiz = 0;
+		} else if (retlen < len)
+			tsiz = 0;
+	}
+nfsmout:
+	return (error);
+}
+
+/*
+ * nfs write call
+ */
+int
+nfs_writerpc(vp, uiop, cred, iomode, must_commit)
+	register struct vnode *vp;
+	register struct uio *uiop;
+	struct ucred *cred;
+	int *iomode, *must_commit;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register int t1, t2, backup;
+	caddr_t bpos, dpos, cp2;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	struct nfsnode *np = VTONFS(vp);
+	u_quad_t frev;
+	int error = 0, len, tsiz, wccflag = NFSV3_WCCRATTR, rlen, commit;
+	int v3 = NFS_ISV3(vp), committed = NFSV3WRITE_FILESYNC;
+
+#ifndef DIAGNOSTIC
+	if (uiop->uio_iovcnt != 1)
+		panic("nfs: writerpc iovcnt > 1");
+#endif
+	*must_commit = 0;
+	tsiz = uiop->uio_resid;
+	if (uiop->uio_offset + tsiz > 0xffffffff && !v3)
+		return (EFBIG);
+	while (tsiz > 0) {
+		nfsstats.rpccnt[NFSPROC_WRITE]++;
+		len = (tsiz > nmp->nm_wsize) ? nmp->nm_wsize : tsiz;
+		nfsm_reqhead(vp, NFSPROC_WRITE,
+			NFSX_FH(v3) + 5 * NFSX_UNSIGNED + nfsm_rndup(len));
+		nfsm_fhtom(vp, v3);
+		if (v3) {
+			nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED);
+			txdr_hyper(&uiop->uio_offset, tl);
+			tl += 2;
+			*tl++ = txdr_unsigned(len);
+			*tl++ = txdr_unsigned(*iomode);
+		} else {
+			nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED);
+			*++tl = txdr_unsigned(uiop->uio_offset);
+			tl += 2;
+		}
+		*tl = txdr_unsigned(len);
+		nfsm_uiotom(uiop, len);
+		nfsm_request(vp, NFSPROC_WRITE, uiop->uio_procp, cred);
+		if (v3) {
+			wccflag = NFSV3_WCCCHK;
+			nfsm_wcc_data(vp, wccflag);
+			if (!error) {
+				nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED +
+					NFSX_V3WRITEVERF);
+				rlen = fxdr_unsigned(int, *tl++);
+				if (rlen == 0) {
+					error = NFSERR_IO;
+					break;
+				} else if (rlen < len) {
+					backup = len - rlen;
+					uiop->uio_iov->iov_base -= backup;
+					uiop->uio_iov->iov_len += backup;
+					uiop->uio_offset -= backup;
+					uiop->uio_resid += backup;
+					len = rlen;
+				}
+				commit = fxdr_unsigned(int, *tl++);
+
+				/*
+				 * Return the lowest committment level
+				 * obtained by any of the RPCs.
+				 */
+				if (committed == NFSV3WRITE_FILESYNC)
+					committed = commit;
+				else if (committed == NFSV3WRITE_DATASYNC &&
+					commit == NFSV3WRITE_UNSTABLE)
+					committed = commit;
+				if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0) {
+				    bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
+					NFSX_V3WRITEVERF);
+				    nmp->nm_flag |= NFSMNT_HASWRITEVERF;
+				} else if (bcmp((caddr_t)tl,
+				    (caddr_t)nmp->nm_verf, NFSX_V3WRITEVERF)) {
+				    *must_commit = 1;
+				    bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
+					NFSX_V3WRITEVERF);
+				}
+			}
+		} else
+		    nfsm_loadattr(vp, (struct vattr *)0);
+		if (wccflag)
+		    VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.ts_sec;
+		m_freem(mrep);
+		tsiz -= len;
+	}
+nfsmout:
+	*iomode = committed;
+	if (error)
+		uiop->uio_resid = tsiz;
+	return (error);
+}
+
+/*
+ * nfs mknod rpc
+ * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
+ * mode set to specify the file type and the size field for rdev.
+ */
+int
+nfs_mknodrpc(dvp, vpp, cnp, vap)
+	register struct vnode *dvp;
+	register struct vnode **vpp;
+	register struct componentname *cnp;
+	register struct vattr *vap;
+{
+	register struct nfsv2_sattr *sp;
+	register struct nfsv3_sattr *sp3;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	struct vnode *newvp = (struct vnode *)0;
+	struct nfsnode *np;
+	struct vattr vattr;
+	char *cp2;
+	caddr_t bpos, dpos;
+	int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	u_long rdev;
+	int v3 = NFS_ISV3(dvp);
+
+	if (vap->va_type == VCHR || vap->va_type == VBLK)
+		rdev = txdr_unsigned(vap->va_rdev);
+	else if (vap->va_type == VFIFO || vap->va_type == VSOCK)
+		rdev = 0xffffffff;
+	else {
+		VOP_ABORTOP(dvp, cnp);
+		vput(dvp);
+		return (EOPNOTSUPP);
+	}
+	if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) {
+		VOP_ABORTOP(dvp, cnp);
+		vput(dvp);
+		return (error);
+	}
+	nfsstats.rpccnt[NFSPROC_MKNOD]++;
+	nfsm_reqhead(dvp, NFSPROC_MKNOD, NFSX_FH(v3) + 4 * NFSX_UNSIGNED +
+		+ nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
+	if (v3) {
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3SRVSATTR);
+		*tl++ = vtonfsv3_type(vap->va_type);
+		sp3 = (struct nfsv3_sattr *)tl;
+		nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid);
+		if (vap->va_type == VCHR || vap->va_type == VBLK) {
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = txdr_unsigned(major(vap->va_rdev));
+			*tl = txdr_unsigned(minor(vap->va_rdev));
+		}
+	} else {
+		nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
+		sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid);
+		sp->sa_gid = txdr_unsigned(vattr.va_gid);
+		sp->sa_size = rdev;
+		txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
+		txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
+	}
+	nfsm_request(dvp, NFSPROC_MKNOD, cnp->cn_proc, cnp->cn_cred);
+	if (!error) {
+		nfsm_mtofh(dvp, newvp, v3, gotvp);
+		if (!gotvp) {
+			if (newvp) {
+				vrele(newvp);
+				newvp = (struct vnode *)0;
+			}
+			error = nfs_lookitup(dvp, cnp->cn_nameptr,
+			    cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np);
+			if (!error)
+				newvp = NFSTOV(np);
+		}
+	}
+	if (v3)
+		nfsm_wcc_data(dvp, wccflag);
+	nfsm_reqdone;
+	if (error) {
+		if (newvp)
+			vrele(newvp);
+	} else {
+		if (cnp->cn_flags & MAKEENTRY)
+			cache_enter(dvp, newvp, cnp);
+		*vpp = newvp;
+	}
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	vrele(dvp);
+	return (error);
+}
+
+/*
+ * nfs mknod vop
+ * just call nfs_mknodrpc() to do the work.
+ */
+/* ARGSUSED */
+int
+nfs_mknod(ap)
+	struct vop_mknod_args /* {
+		struct vnode *a_dvp;
+		struct vnode **a_vpp;
+		struct componentname *a_cnp;
+		struct vattr *a_vap;
+	} */ *ap;
+{
+	struct vnode *newvp;
+	int error;
+
+	error = nfs_mknodrpc(ap->a_dvp, &newvp, ap->a_cnp, ap->a_vap);
+	if (!error)
+		vrele(newvp);
+	return (error);
+}
+
+static u_long create_verf;
+/*
+ * nfs file create call
+ */
+int
+nfs_create(ap)
+	struct vop_create_args /* {
+		struct vnode *a_dvp;
+		struct vnode **a_vpp;
+		struct componentname *a_cnp;
+		struct vattr *a_vap;
+	} */ *ap;
+{
+	register struct vnode *dvp = ap->a_dvp;
+	register struct vattr *vap = ap->a_vap;
+	register struct componentname *cnp = ap->a_cnp;
+	register struct nfsv2_sattr *sp;
+	register struct nfsv3_sattr *sp3;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	struct nfsnode *np = (struct nfsnode *)0;
+	struct vnode *newvp = (struct vnode *)0;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0, fmode = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct vattr vattr;
+	int v3 = NFS_ISV3(dvp);
+
+	/*
+	 * Oops, not for me..
+	 */
+	if (vap->va_type == VSOCK)
+		return (nfs_mknodrpc(dvp, ap->a_vpp, cnp, vap));
+
+	if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) {
+		VOP_ABORTOP(dvp, cnp);
+		vput(dvp);
+		return (error);
+	}
+	if (vap->va_vaflags & VA_EXCLUSIVE)
+		fmode |= O_EXCL;
+again:
+	nfsstats.rpccnt[NFSPROC_CREATE]++;
+	nfsm_reqhead(dvp, NFSPROC_CREATE, NFSX_FH(v3) + 2 * NFSX_UNSIGNED +
+		nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
+	if (v3) {
+		nfsm_build(tl, u_long *, NFSX_UNSIGNED);
+		if (fmode & O_EXCL) {
+		    *tl = txdr_unsigned(NFSV3CREATE_EXCLUSIVE);
+		    nfsm_build(tl, u_long *, NFSX_V3CREATEVERF);
+		    if (in_ifaddr)
+			*tl++ = IA_SIN(in_ifaddr)->sin_addr.s_addr;
+		    else
+			*tl++ = create_verf;
+		    *tl = ++create_verf;
+		} else {
+		    *tl = txdr_unsigned(NFSV3CREATE_UNCHECKED);
+		    nfsm_build(tl, u_long *, NFSX_V3SRVSATTR);
+		    sp3 = (struct nfsv3_sattr *)tl;
+		    nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid);
+		}
+	} else {
+		nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
+		sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid);
+		sp->sa_gid = txdr_unsigned(vattr.va_gid);
+		sp->sa_size = 0;
+		txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
+		txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
+	}
+	nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred);
+	if (!error) {
+		nfsm_mtofh(dvp, newvp, v3, gotvp);
+		if (!gotvp) {
+			if (newvp) {
+				vrele(newvp);
+				newvp = (struct vnode *)0;
+			}
+			error = nfs_lookitup(dvp, cnp->cn_nameptr,
+			    cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np);
+			if (!error)
+				newvp = NFSTOV(np);
+		}
+	}
+	if (v3)
+		nfsm_wcc_data(dvp, wccflag);
+	nfsm_reqdone;
+	if (error) {
+		if (v3 && (fmode & O_EXCL) && error == NFSERR_NOTSUPP) {
+			fmode &= ~O_EXCL;
+			goto again;
+		}
+		if (newvp)
+			vrele(newvp);
+	} else if (v3 && (fmode & O_EXCL))
+		error = nfs_setattrrpc(newvp, vap, cnp->cn_cred, cnp->cn_proc);
+	if (!error) {
+		if (cnp->cn_flags & MAKEENTRY)
+			cache_enter(dvp, newvp, cnp);
+		*ap->a_vpp = newvp;
+	}
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	vrele(dvp);
+	return (error);
+}
+
+/*
+ * nfs file remove call
+ * To try and make nfs semantics closer to ufs semantics, a file that has
+ * other processes using the vnode is renamed instead of removed and then
+ * removed later on the last close.
+ * - If v_usecount > 1
+ *	  If a rename is not already in the works
+ *	     call nfs_sillyrename() to set it up
+ *     else
+ *	  do the remove rpc
+ */
+int
+nfs_remove(ap)
+	struct vop_remove_args /* {
+		struct vnodeop_desc *a_desc;
+		struct vnode * a_dvp;
+		struct vnode * a_vp;
+		struct componentname * a_cnp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct vnode *dvp = ap->a_dvp;
+	register struct componentname *cnp = ap->a_cnp;
+	register struct nfsnode *np = VTONFS(vp);
+	register u_long *tl;
+	register caddr_t cp;
+	register long t2;
+	caddr_t bpos, dpos;
+	int error = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct vattr vattr;
+	int v3 = NFS_ISV3(dvp);
+
+#ifndef DIAGNOSTIC
+	if ((cnp->cn_flags & HASBUF) == 0)
+		panic("nfs_remove: no name");
+	if (vp->v_usecount < 1)
+		panic("nfs_remove: bad v_usecount");
+#endif
+	if (vp->v_usecount == 1 || (np->n_sillyrename &&
+	    VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) == 0 &&
+	    vattr.va_nlink > 1)) {
+		/*
+		 * Purge the name cache so that the chance of a lookup for
+		 * the name succeeding while the remove is in progress is
+		 * minimized. Without node locking it can still happen, such
+		 * that an I/O op returns ESTALE, but since you get this if
+		 * another host removes the file..
+		 */
+		cache_purge(vp);
+		/*
+		 * throw away biocache buffers, mainly to avoid
+		 * unnecessary delayed writes later.
+		 */
+		error = nfs_vinvalbuf(vp, 0, cnp->cn_cred, cnp->cn_proc, 1);
+		/* Do the rpc */
+		if (error != EINTR)
+			error = nfs_removerpc(dvp, cnp->cn_nameptr,
+				cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc);
+		/*
+		 * Kludge City: If the first reply to the remove rpc is lost..
+		 *   the reply to the retransmitted request will be ENOENT
+		 *   since the file was in fact removed
+		 *   Therefore, we cheat and return success.
+		 */
+		if (error == ENOENT)
+			error = 0;
+	} else if (!np->n_sillyrename)
+		error = nfs_sillyrename(dvp, vp, cnp);
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	np->n_attrstamp = 0;
+	vrele(dvp);
+	vrele(vp);
+	return (error);
+}
+
+/*
+ * nfs file remove rpc called from nfs_inactive
+ */
+int
+nfs_removeit(sp)
+	register struct sillyrename *sp;
+{
+
+	return (nfs_removerpc(sp->s_dvp, sp->s_name, sp->s_namlen, sp->s_cred,
+		(struct proc *)0));
+}
+
+/*
+ * Nfs remove rpc, called from nfs_remove() and nfs_removeit().
+ */
+int
+nfs_removerpc(dvp, name, namelen, cred, proc)
+	register struct vnode *dvp;
+	char *name;
+	int namelen;
+	struct ucred *cred;
+	struct proc *proc;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, wccflag = NFSV3_WCCRATTR;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(dvp);
+
+	nfsstats.rpccnt[NFSPROC_REMOVE]++;
+	nfsm_reqhead(dvp, NFSPROC_REMOVE,
+		NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(namelen));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(name, namelen, NFS_MAXNAMLEN);
+	nfsm_request(dvp, NFSPROC_REMOVE, proc, cred);
+	if (v3)
+		nfsm_wcc_data(dvp, wccflag);
+	nfsm_reqdone;
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	return (error);
+}
+
+/*
+ * nfs file rename call
+ */
+int
+nfs_rename(ap)
+	struct vop_rename_args  /* {
+		struct vnode *a_fdvp;
+		struct vnode *a_fvp;
+		struct componentname *a_fcnp;
+		struct vnode *a_tdvp;
+		struct vnode *a_tvp;
+		struct componentname *a_tcnp;
+	} */ *ap;
+{
+	register struct vnode *fvp = ap->a_fvp;
+	register struct vnode *tvp = ap->a_tvp;
+	register struct vnode *fdvp = ap->a_fdvp;
+	register struct vnode *tdvp = ap->a_tdvp;
+	register struct componentname *tcnp = ap->a_tcnp;
+	register struct componentname *fcnp = ap->a_fcnp;
+	int error;
+
+#ifndef DIAGNOSTIC
+	if ((tcnp->cn_flags & HASBUF) == 0 ||
+	    (fcnp->cn_flags & HASBUF) == 0)
+		panic("nfs_rename: no name");
+#endif
+	/* Check for cross-device rename */
+	if ((fvp->v_mount != tdvp->v_mount) ||
+	    (tvp && (fvp->v_mount != tvp->v_mount))) {
+		error = EXDEV;
+		goto out;
+	}
+
+	/*
+	 * If the tvp exists and is in use, sillyrename it before doing the
+	 * rename of the new file over it.
+	 */
+	if (tvp && tvp->v_usecount > 1 && !VTONFS(tvp)->n_sillyrename &&
+		!nfs_sillyrename(tdvp, tvp, tcnp)) {
+		vrele(tvp);
+		tvp = NULL;
+	}
+
+	error = nfs_renamerpc(fdvp, fcnp->cn_nameptr, fcnp->cn_namelen,
+		tdvp, tcnp->cn_nameptr, tcnp->cn_namelen, tcnp->cn_cred,
+		tcnp->cn_proc);
+
+	if (fvp->v_type == VDIR) {
+		if (tvp != NULL && tvp->v_type == VDIR)
+			cache_purge(tdvp);
+		cache_purge(fdvp);
+	}
+out:
+	if (tdvp == tvp)
+		vrele(tdvp);
+	else
+		vput(tdvp);
+	if (tvp)
+		vput(tvp);
+	vrele(fdvp);
+	vrele(fvp);
+	/*
+	 * Kludge: Map ENOENT => 0 assuming that it is a reply to a retry.
+	 */
+	if (error == ENOENT)
+		error = 0;
+	return (error);
+}
+
+/*
+ * nfs file rename rpc called from nfs_remove() above
+ */
+int
+nfs_renameit(sdvp, scnp, sp)
+	struct vnode *sdvp;
+	struct componentname *scnp;
+	register struct sillyrename *sp;
+{
+	return (nfs_renamerpc(sdvp, scnp->cn_nameptr, scnp->cn_namelen,
+		sdvp, sp->s_name, sp->s_namlen, scnp->cn_cred, scnp->cn_proc));
+}
+
+/*
+ * Do an nfs rename rpc. Called from nfs_rename() and nfs_renameit().
+ */
+int
+nfs_renamerpc(fdvp, fnameptr, fnamelen, tdvp, tnameptr, tnamelen, cred, proc)
+	register struct vnode *fdvp;
+	char *fnameptr;
+	int fnamelen;
+	register struct vnode *tdvp;
+	char *tnameptr;
+	int tnamelen;
+	struct ucred *cred;
+	struct proc *proc;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, fwccflag = NFSV3_WCCRATTR, twccflag = NFSV3_WCCRATTR;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(fdvp);
+
+	nfsstats.rpccnt[NFSPROC_RENAME]++;
+	nfsm_reqhead(fdvp, NFSPROC_RENAME,
+		(NFSX_FH(v3) + NFSX_UNSIGNED)*2 + nfsm_rndup(fnamelen) +
+		nfsm_rndup(tnamelen));
+	nfsm_fhtom(fdvp, v3);
+	nfsm_strtom(fnameptr, fnamelen, NFS_MAXNAMLEN);
+	nfsm_fhtom(tdvp, v3);
+	nfsm_strtom(tnameptr, tnamelen, NFS_MAXNAMLEN);
+	nfsm_request(fdvp, NFSPROC_RENAME, proc, cred);
+	if (v3) {
+		nfsm_wcc_data(fdvp, fwccflag);
+		nfsm_wcc_data(tdvp, twccflag);
+	}
+	nfsm_reqdone;
+	VTONFS(fdvp)->n_flag |= NMODIFIED;
+	VTONFS(tdvp)->n_flag |= NMODIFIED;
+	if (!fwccflag)
+		VTONFS(fdvp)->n_attrstamp = 0;
+	if (!twccflag)
+		VTONFS(tdvp)->n_attrstamp = 0;
+	return (error);
+}
+
+/*
+ * nfs hard link create call
+ */
+int
+nfs_link(ap)
+	struct vop_link_args /* {
+		struct vnode *a_vp;
+		struct vnode *a_tdvp;
+		struct componentname *a_cnp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct vnode *tdvp = ap->a_tdvp;
+	register struct componentname *cnp = ap->a_cnp;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, wccflag = NFSV3_WCCRATTR, attrflag = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(vp);
+
+	if (vp->v_mount != tdvp->v_mount) {
+		/*VOP_ABORTOP(vp, cnp);*/
+		if (tdvp == vp)
+			vrele(tdvp);
+		else
+			vput(tdvp);
+		return (EXDEV);
+	}
+
+	/*
+	 * Push all writes to the server, so that the attribute cache
+	 * doesn't get "out of sync" with the server.
+	 * XXX There should be a better way!
+	 */
+	VOP_FSYNC(vp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc);
+
+	nfsstats.rpccnt[NFSPROC_LINK]++;
+	nfsm_reqhead(vp, NFSPROC_LINK,
+		NFSX_FH(v3)*2 + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen));
+	nfsm_fhtom(vp, v3);
+	nfsm_fhtom(tdvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
+	nfsm_request(vp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred);
+	if (v3) {
+		nfsm_postop_attr(vp, attrflag);
+		nfsm_wcc_data(tdvp, wccflag);
+	}
+	nfsm_reqdone;
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VTONFS(tdvp)->n_flag |= NMODIFIED;
+	if (!attrflag)
+		VTONFS(vp)->n_attrstamp = 0;
+	if (!wccflag)
+		VTONFS(tdvp)->n_attrstamp = 0;
+	vrele(tdvp);
+	/*
+	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
+	 */
+	if (error == EEXIST)
+		error = 0;
+	return (error);
+}
+
+/*
+ * nfs symbolic link create call
+ */
+int
+nfs_symlink(ap)
+	struct vop_symlink_args /* {
+		struct vnode *a_dvp;
+		struct vnode **a_vpp;
+		struct componentname *a_cnp;
+		struct vattr *a_vap;
+		char *a_target;
+	} */ *ap;
+{
+	register struct vnode *dvp = ap->a_dvp;
+	register struct vattr *vap = ap->a_vap;
+	register struct componentname *cnp = ap->a_cnp;
+	register struct nfsv2_sattr *sp;
+	register struct nfsv3_sattr *sp3;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int slen, error = 0, wccflag = NFSV3_WCCRATTR, gotvp;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct vnode *newvp = (struct vnode *)0;
+	int v3 = NFS_ISV3(dvp);
+
+	nfsstats.rpccnt[NFSPROC_SYMLINK]++;
+	slen = strlen(ap->a_target);
+	nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH(v3) + 2*NFSX_UNSIGNED +
+	    nfsm_rndup(cnp->cn_namelen) + nfsm_rndup(slen) + NFSX_SATTR(v3));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
+	if (v3) {
+		nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR);
+		nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid,
+			cnp->cn_cred->cr_gid);
+	}
+	nfsm_strtom(ap->a_target, slen, NFS_MAXPATHLEN);
+	if (!v3) {
+		nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		sp->sa_mode = vtonfsv2_mode(VLNK, vap->va_mode);
+		sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid);
+		sp->sa_gid = txdr_unsigned(cnp->cn_cred->cr_gid);
+		sp->sa_size = -1;
+		txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
+		txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
+	}
+	nfsm_request(dvp, NFSPROC_SYMLINK, cnp->cn_proc, cnp->cn_cred);
+	if (v3) {
+		if (!error)
+			nfsm_mtofh(dvp, newvp, v3, gotvp);
+		nfsm_wcc_data(dvp, wccflag);
+	}
+	nfsm_reqdone;
+	if (newvp)
+		vrele(newvp);
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	vrele(dvp);
+	/*
+	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
+	 */
+	if (error == EEXIST)
+		error = 0;
+	return (error);
+}
+
+/*
+ * nfs make dir call
+ */
+int
+nfs_mkdir(ap)
+	struct vop_mkdir_args /* {
+		struct vnode *a_dvp;
+		struct vnode **a_vpp;
+		struct componentname *a_cnp;
+		struct vattr *a_vap;
+	} */ *ap;
+{
+	register struct vnode *dvp = ap->a_dvp;
+	register struct vattr *vap = ap->a_vap;
+	register struct componentname *cnp = ap->a_cnp;
+	register struct nfsv2_sattr *sp;
+	register struct nfsv3_sattr *sp3;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	register int len;
+	struct nfsnode *np = (struct nfsnode *)0;
+	struct vnode *newvp = (struct vnode *)0;
+	caddr_t bpos, dpos, cp2;
+	nfsfh_t *fhp;
+	int error = 0, wccflag = NFSV3_WCCRATTR, attrflag;
+	int fhsize, gotvp = 0;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	struct vattr vattr;
+	int v3 = NFS_ISV3(dvp);
+
+	if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) {
+		VOP_ABORTOP(dvp, cnp);
+		vput(dvp);
+		return (error);
+	}
+	len = cnp->cn_namelen;
+	nfsstats.rpccnt[NFSPROC_MKDIR]++;
+	nfsm_reqhead(dvp, NFSPROC_MKDIR,
+	  NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len) + NFSX_SATTR(v3));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN);
+	if (v3) {
+		nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR);
+		nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid);
+	} else {
+		nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
+		sp->sa_mode = vtonfsv2_mode(VDIR, vap->va_mode);
+		sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid);
+		sp->sa_gid = txdr_unsigned(vattr.va_gid);
+		sp->sa_size = -1;
+		txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
+		txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
+	}
+	nfsm_request(dvp, NFSPROC_MKDIR, cnp->cn_proc, cnp->cn_cred);
+	if (!error)
+		nfsm_mtofh(dvp, newvp, v3, gotvp);
+	if (v3)
+		nfsm_wcc_data(dvp, wccflag);
+	nfsm_reqdone;
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	/*
+	 * Kludge: Map EEXIST => 0 assuming that you have a reply to a retry
+	 * if we can succeed in looking up the directory.
+	 */
+	if (error == EEXIST || (!error && !gotvp)) {
+		if (newvp) {
+			vrele(newvp);
+			newvp = (struct vnode *)0;
+		}
+		error = nfs_lookitup(dvp, cnp->cn_nameptr, len, cnp->cn_cred,
+			cnp->cn_proc, &np);
+		if (!error) {
+			newvp = NFSTOV(np);
+			if (newvp->v_type != VDIR)
+				error = EEXIST;
+		}
+	}
+	if (error) {
+		if (newvp)
+			vrele(newvp);
+	} else
+		*ap->a_vpp = newvp;
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	vrele(dvp);
+	return (error);
+}
+
+/*
+ * nfs remove directory call
+ */
+int
+nfs_rmdir(ap)
+	struct vop_rmdir_args /* {
+		struct vnode *a_dvp;
+		struct vnode *a_vp;
+		struct componentname *a_cnp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct vnode *dvp = ap->a_dvp;
+	register struct componentname *cnp = ap->a_cnp;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	caddr_t bpos, dpos, cp2;
+	int error = 0, wccflag = NFSV3_WCCRATTR;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	int v3 = NFS_ISV3(dvp);
+
+	if (dvp == vp) {
+		vrele(dvp);
+		vrele(dvp);
+		FREE(cnp->cn_pnbuf, M_NAMEI);
+		return (EINVAL);
+	}
+	nfsstats.rpccnt[NFSPROC_RMDIR]++;
+	nfsm_reqhead(dvp, NFSPROC_RMDIR,
+		NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
+	nfsm_request(dvp, NFSPROC_RMDIR, cnp->cn_proc, cnp->cn_cred);
+	if (v3)
+		nfsm_wcc_data(dvp, wccflag);
+	nfsm_reqdone;
+	FREE(cnp->cn_pnbuf, M_NAMEI);
+	VTONFS(dvp)->n_flag |= NMODIFIED;
+	if (!wccflag)
+		VTONFS(dvp)->n_attrstamp = 0;
+	cache_purge(dvp);
+	cache_purge(vp);
+	vrele(vp);
+	vrele(dvp);
+	/*
+	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
+	 */
+	if (error == ENOENT)
+		error = 0;
+	return (error);
+}
+
+/*
+ * nfs readdir call
+ */
+int
+nfs_readdir(ap)
+	struct vop_readdir_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	register struct uio *uio = ap->a_uio;
+	int tresid, error;
+	struct vattr vattr;
+
+	if (vp->v_type != VDIR)
+		return (EPERM);
+	/*
+	 * First, check for hit on the EOF offset cache
+	 */
+	if (np->n_direofoffset > 0 && uio->uio_offset >= np->n_direofoffset &&
+	    (np->n_flag & NMODIFIED) == 0) {
+		if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) {
+			if (NQNFS_CKCACHABLE(vp, ND_READ)) {
+				nfsstats.direofcache_hits++;
+				return (0);
+			}
+		} else if (VOP_GETATTR(vp, &vattr, ap->a_cred, uio->uio_procp) == 0 &&
+			np->n_mtime == vattr.va_mtime.ts_sec) {
+			nfsstats.direofcache_hits++;
+			return (0);
+		}
+	}
+
+	/*
+	 * Call nfs_bioread() to do the real work.
+	 */
+	tresid = uio->uio_resid;
+	error = nfs_bioread(vp, uio, 0, ap->a_cred);
+
+	if (!error && uio->uio_resid == tresid)
+		nfsstats.direofcache_misses++;
+	return (error);
+}
+
+/*
+ * Readdir rpc call.
+ * Called from below the buffer cache by nfs_doio().
+ */
+int
+nfs_readdirrpc(vp, uiop, cred)
+	struct vnode *vp;
+	register struct uio *uiop;
+	struct ucred *cred;
+{
+	register int len, left;
+	register struct dirent *dp;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	register nfsuint64 *cookiep;
+	caddr_t bpos, dpos, cp2;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	nfsuint64 cookie;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	struct nfsnode *dnp = VTONFS(vp);
+	nfsfh_t *fhp;
+	u_quad_t frev, fileno;
+	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1, i;
+	int cachable, attrflag, fhsize;
+	int v3 = NFS_ISV3(vp);
+
+#ifndef nolint
+	dp = (struct dirent *)0;
+#endif
+#ifndef DIAGNOSTIC
+	if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (NFS_DIRBLKSIZ - 1)) ||
+		(uiop->uio_resid & (NFS_DIRBLKSIZ - 1)))
+		panic("nfs readdirrpc bad uio");
+#endif
+
+	/*
+	 * If there is no cookie, assume end of directory.
+	 */
+	cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0);
+	if (cookiep)
+		cookie = *cookiep;
+	else
+		return (0);
+	/*
+	 * Loop around doing readdir rpc's of size nm_readdirsize
+	 * truncated to a multiple of DIRBLKSIZ.
+	 * The stopping criteria is EOF or buffer full.
+	 */
+	while (more_dirs && bigenough) {
+		nfsstats.rpccnt[NFSPROC_READDIR]++;
+		nfsm_reqhead(vp, NFSPROC_READDIR, NFSX_FH(v3) +
+			NFSX_READDIR(v3));
+		nfsm_fhtom(vp, v3);
+		if (v3) {
+			nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED);
+			*tl++ = cookie.nfsuquad[0];
+			*tl++ = cookie.nfsuquad[1];
+			*tl++ = dnp->n_cookieverf.nfsuquad[0];
+			*tl++ = dnp->n_cookieverf.nfsuquad[1];
+		} else {
+			nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED);
+			*tl++ = cookie.nfsuquad[0];
+		}
+		*tl = txdr_unsigned(nmp->nm_readdirsize);
+		nfsm_request(vp, NFSPROC_READDIR, uiop->uio_procp, cred);
+		if (v3) {
+			nfsm_postop_attr(vp, attrflag);
+			if (!error) {
+				nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+				dnp->n_cookieverf.nfsuquad[0] = *tl++;
+				dnp->n_cookieverf.nfsuquad[1] = *tl;
+			} else {
+				m_freem(mrep);
+				goto nfsmout;
+			}
+		}
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+		more_dirs = fxdr_unsigned(int, *tl);
+	
+		/* loop thru the dir entries, doctoring them to 4bsd form */
+		while (more_dirs && bigenough) {
+			if (v3) {
+				nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+				fxdr_hyper(tl, &fileno);
+				len = fxdr_unsigned(int, *(tl + 2));
+			} else {
+				nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+				fileno = fxdr_unsigned(u_quad_t, *tl++);
+				len = fxdr_unsigned(int, *tl);
+			}
+			if (len <= 0 || len > NFS_MAXNAMLEN) {
+				error = EBADRPC;
+				m_freem(mrep);
+				goto nfsmout;
+			}
+			tlen = nfsm_rndup(len);
+			if (tlen == len)
+				tlen += 4;	/* To ensure null termination */
+			left = DIRBLKSIZ - blksiz;
+			if ((tlen + DIRHDSIZ) > left) {
+				dp->d_reclen += left;
+				uiop->uio_iov->iov_base += left;
+				uiop->uio_iov->iov_len -= left;
+				uiop->uio_offset += left;
+				uiop->uio_resid -= left;
+				blksiz = 0;
+			}
+			if ((tlen + DIRHDSIZ) > uiop->uio_resid)
+				bigenough = 0;
+			if (bigenough) {
+				dp = (struct dirent *)uiop->uio_iov->iov_base;
+				dp->d_fileno = (int)fileno;
+				dp->d_namlen = len;
+				dp->d_reclen = tlen + DIRHDSIZ;
+				dp->d_type = DT_UNKNOWN;
+				blksiz += dp->d_reclen;
+				if (blksiz == DIRBLKSIZ)
+					blksiz = 0;
+				uiop->uio_offset += DIRHDSIZ;
+				uiop->uio_resid -= DIRHDSIZ;
+				uiop->uio_iov->iov_base += DIRHDSIZ;
+				uiop->uio_iov->iov_len -= DIRHDSIZ;
+				nfsm_mtouio(uiop, len);
+				cp = uiop->uio_iov->iov_base;
+				tlen -= len;
+				*cp = '\0';	/* null terminate */
+				uiop->uio_iov->iov_base += tlen;
+				uiop->uio_iov->iov_len -= tlen;
+				uiop->uio_offset += tlen;
+				uiop->uio_resid -= tlen;
+			} else
+				nfsm_adv(nfsm_rndup(len));
+			if (v3) {
+				nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+			} else {
+				nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED);
+			}
+			if (bigenough) {
+				cookie.nfsuquad[0] = *tl++;
+				if (v3)
+					cookie.nfsuquad[1] = *tl++;
+			} else if (v3)
+				tl += 2;
+			else
+				tl++;
+			more_dirs = fxdr_unsigned(int, *tl);
+		}
+		/*
+		 * If at end of rpc data, get the eof boolean
+		 */
+		if (!more_dirs) {
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			more_dirs = (fxdr_unsigned(int, *tl) == 0);
+		}
+		m_freem(mrep);
+	}
+	/*
+	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
+	 * by increasing d_reclen for the last record.
+	 */
+	if (blksiz > 0) {
+		left = DIRBLKSIZ - blksiz;
+		dp->d_reclen += left;
+		uiop->uio_iov->iov_base += left;
+		uiop->uio_iov->iov_len -= left;
+		uiop->uio_offset += left;
+		uiop->uio_resid -= left;
+	}
+
+	/*
+	 * We are now either at the end of the directory or have filled the
+	 * block.
+	 */
+	if (bigenough)
+		dnp->n_direofoffset = uiop->uio_offset;
+	else {
+		if (uiop->uio_resid > 0)
+			printf("EEK! readdirrpc resid > 0\n");
+		cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1);
+		*cookiep = cookie;
+	}
+nfsmout:
+	return (error);
+}
+
+/*
+ * NFS V3 readdir plus RPC. Used in place of nfs_readdirrpc().
+ */
+int
+nfs_readdirplusrpc(vp, uiop, cred)
+	struct vnode *vp;
+	register struct uio *uiop;
+	struct ucred *cred;
+{
+	register int len, left;
+	register struct dirent *dp;
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	register struct vnode *newvp;
+	register nfsuint64 *cookiep;
+	caddr_t bpos, dpos, cp2, dpossav1, dpossav2;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2, *mdsav1, *mdsav2;
+	struct nameidata nami, *ndp = &nami;
+	struct componentname *cnp = &ndp->ni_cnd;
+	nfsuint64 cookie;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	struct nfsnode *dnp = VTONFS(vp), *np;
+	nfsfh_t *fhp;
+	u_quad_t frev, fileno;
+	int error = 0, tlen, more_dirs = 1, blksiz = 0, doit, bigenough = 1, i;
+	int cachable, attrflag, fhsize;
+
+#ifndef nolint
+	dp = (struct dirent *)0;
+#endif
+#ifndef DIAGNOSTIC
+	if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (DIRBLKSIZ - 1)) ||
+		(uiop->uio_resid & (DIRBLKSIZ - 1)))
+		panic("nfs readdirplusrpc bad uio");
+#endif
+	ndp->ni_dvp = vp;
+	newvp = NULLVP;
+
+	/*
+	 * If there is no cookie, assume end of directory.
+	 */
+	cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0);
+	if (cookiep)
+		cookie = *cookiep;
+	else
+		return (0);
+	/*
+	 * Loop around doing readdir rpc's of size nm_readdirsize
+	 * truncated to a multiple of DIRBLKSIZ.
+	 * The stopping criteria is EOF or buffer full.
+	 */
+	while (more_dirs && bigenough) {
+		nfsstats.rpccnt[NFSPROC_READDIRPLUS]++;
+		nfsm_reqhead(vp, NFSPROC_READDIRPLUS,
+			NFSX_FH(1) + 6 * NFSX_UNSIGNED);
+		nfsm_fhtom(vp, 1);
+ 		nfsm_build(tl, u_long *, 6 * NFSX_UNSIGNED);
+		*tl++ = cookie.nfsuquad[0];
+		*tl++ = cookie.nfsuquad[1];
+		*tl++ = dnp->n_cookieverf.nfsuquad[0];
+		*tl++ = dnp->n_cookieverf.nfsuquad[1];
+		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
+		*tl = txdr_unsigned(nmp->nm_rsize);
+		nfsm_request(vp, NFSPROC_READDIRPLUS, uiop->uio_procp, cred);
+		nfsm_postop_attr(vp, attrflag);
+		if (error) {
+			m_freem(mrep);
+			goto nfsmout;
+		}
+		nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+		dnp->n_cookieverf.nfsuquad[0] = *tl++;
+		dnp->n_cookieverf.nfsuquad[1] = *tl++;
+		more_dirs = fxdr_unsigned(int, *tl);
+	
+		/* loop thru the dir entries, doctoring them to 4bsd form */
+		while (more_dirs && bigenough) {
+			nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+			fxdr_hyper(tl, &fileno);
+			len = fxdr_unsigned(int, *(tl + 2));
+			if (len <= 0 || len > NFS_MAXNAMLEN) {
+				error = EBADRPC;
+				m_freem(mrep);
+				goto nfsmout;
+			}
+			tlen = nfsm_rndup(len);
+			if (tlen == len)
+				tlen += 4;	/* To ensure null termination*/
+			left = DIRBLKSIZ - blksiz;
+			if ((tlen + DIRHDSIZ) > left) {
+				dp->d_reclen += left;
+				uiop->uio_iov->iov_base += left;
+				uiop->uio_iov->iov_len -= left;
+				uiop->uio_offset += left;
+				uiop->uio_resid -= left;
+				blksiz = 0;
+			}
+			if ((tlen + DIRHDSIZ) > uiop->uio_resid)
+				bigenough = 0;
+			if (bigenough) {
+				dp = (struct dirent *)uiop->uio_iov->iov_base;
+				dp->d_fileno = (int)fileno;
+				dp->d_namlen = len;
+				dp->d_reclen = tlen + DIRHDSIZ;
+				dp->d_type = DT_UNKNOWN;
+				blksiz += dp->d_reclen;
+				if (blksiz == DIRBLKSIZ)
+					blksiz = 0;
+				uiop->uio_offset += DIRHDSIZ;
+				uiop->uio_resid -= DIRHDSIZ;
+				uiop->uio_iov->iov_base += DIRHDSIZ;
+				uiop->uio_iov->iov_len -= DIRHDSIZ;
+				cnp->cn_nameptr = uiop->uio_iov->iov_base;
+				cnp->cn_namelen = len;
+				nfsm_mtouio(uiop, len);
+				cp = uiop->uio_iov->iov_base;
+				tlen -= len;
+				*cp = '\0';
+				uiop->uio_iov->iov_base += tlen;
+				uiop->uio_iov->iov_len -= tlen;
+				uiop->uio_offset += tlen;
+				uiop->uio_resid -= tlen;
+			} else
+				nfsm_adv(nfsm_rndup(len));
+			nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED);
+			if (bigenough) {
+				cookie.nfsuquad[0] = *tl++;
+				cookie.nfsuquad[1] = *tl++;
+			} else
+				tl += 2;
+
+			/*
+			 * Since the attributes are before the file handle
+			 * (sigh), we must skip over the attributes and then
+			 * come back and get them.
+			 */
+			attrflag = fxdr_unsigned(int, *tl);
+			if (attrflag) {
+			    dpossav1 = dpos;
+			    mdsav1 = md;
+			    nfsm_adv(NFSX_V3FATTR);
+			    nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			    doit = fxdr_unsigned(int, *tl);
+			    if (doit) {
+				nfsm_getfh(fhp, fhsize, 1);
+				if (NFS_CMPFH(dnp, fhp, fhsize)) {
+				    VREF(vp);
+				    newvp = vp;
+				    np = dnp;
+				} else {
+				    if (error = nfs_nget(vp->v_mount, fhp,
+					fhsize, &np))
+					doit = 0;
+				    else
+					newvp = NFSTOV(np);
+				}
+			    }
+			    if (doit) {
+				dpossav2 = dpos;
+				dpos = dpossav1;
+				mdsav2 = md;
+				md = mdsav1;
+				nfsm_loadattr(newvp, (struct vattr *)0);
+				dpos = dpossav2;
+				md = mdsav2;
+				dp->d_type =
+				    IFTODT(VTTOIF(np->n_vattr.va_type));
+				ndp->ni_vp = newvp;
+				cnp->cn_hash = 0;
+				for (cp = cnp->cn_nameptr, i = 1; i <= len;
+				    i++, cp++)
+				    cnp->cn_hash += (unsigned char)*cp * i;
+				if (cnp->cn_namelen <= NCHNAMLEN)
+				    cache_enter(ndp->ni_dvp, ndp->ni_vp, cnp);
+			    }
+			} else {
+			    /* Just skip over the file handle */
+			    nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			    i = fxdr_unsigned(int, *tl);
+			    nfsm_adv(nfsm_rndup(i));
+			}
+			if (newvp != NULLVP) {
+			    vrele(newvp);
+			    newvp = NULLVP;
+			}
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			more_dirs = fxdr_unsigned(int, *tl);
+		}
+		/*
+		 * If at end of rpc data, get the eof boolean
+		 */
+		if (!more_dirs) {
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED);
+			more_dirs = (fxdr_unsigned(int, *tl) == 0);
+		}
+		m_freem(mrep);
+	}
+	/*
+	 * Fill last record, iff any, out to a multiple of NFS_DIRBLKSIZ
+	 * by increasing d_reclen for the last record.
+	 */
+	if (blksiz > 0) {
+		left = DIRBLKSIZ - blksiz;
+		dp->d_reclen += left;
+		uiop->uio_iov->iov_base += left;
+		uiop->uio_iov->iov_len -= left;
+		uiop->uio_offset += left;
+		uiop->uio_resid -= left;
+	}
+
+	/*
+	 * We are now either at the end of the directory or have filled the
+	 * block.
+	 */
+	if (bigenough)
+		dnp->n_direofoffset = uiop->uio_offset;
+	else {
+		if (uiop->uio_resid > 0)
+			printf("EEK! readdirplusrpc resid > 0\n");
+		cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1);
+		*cookiep = cookie;
+	}
+nfsmout:
+	if (newvp != NULLVP)
+		vrele(newvp);
+	return (error);
+}
+static char hextoasc[] = "0123456789abcdef";
+
+/*
+ * Silly rename. To make the NFS filesystem that is stateless look a little
+ * more like the "ufs" a remove of an active vnode is translated to a rename
+ * to a funny looking filename that is removed by nfs_inactive on the
+ * nfsnode. There is the potential for another process on a different client
+ * to create the same funny name between the nfs_lookitup() fails and the
+ * nfs_rename() completes, but...
+ */
+int
+nfs_sillyrename(dvp, vp, cnp)
+	struct vnode *dvp, *vp;
+	struct componentname *cnp;
+{
+	register struct sillyrename *sp;
+	struct nfsnode *np;
+	int error;
+	short pid;
+
+	cache_purge(dvp);
+	np = VTONFS(vp);
+#ifndef DIAGNOSTIC
+	if (vp->v_type == VDIR)
+		panic("nfs: sillyrename dir");
+#endif
+	MALLOC(sp, struct sillyrename *, sizeof (struct sillyrename),
+		M_NFSREQ, M_WAITOK);
+	sp->s_cred = crdup(cnp->cn_cred);
+	sp->s_dvp = dvp;
+	VREF(dvp);
+
+	/* Fudge together a funny name */
+	pid = cnp->cn_proc->p_pid;
+	bcopy(".nfsAxxxx4.4", sp->s_name, 13);
+	sp->s_namlen = 12;
+	sp->s_name[8] = hextoasc[pid & 0xf];
+	sp->s_name[7] = hextoasc[(pid >> 4) & 0xf];
+	sp->s_name[6] = hextoasc[(pid >> 8) & 0xf];
+	sp->s_name[5] = hextoasc[(pid >> 12) & 0xf];
+
+	/* Try lookitups until we get one that isn't there */
+	while (nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred,
+		cnp->cn_proc, (struct nfsnode **)0) == 0) {
+		sp->s_name[4]++;
+		if (sp->s_name[4] > 'z') {
+			error = EINVAL;
+			goto bad;
+		}
+	}
+	if (error = nfs_renameit(dvp, cnp, sp))
+		goto bad;
+	error = nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred,
+		cnp->cn_proc, &np);
+	np->n_sillyrename = sp;
+	return (0);
+bad:
+	vrele(sp->s_dvp);
+	crfree(sp->s_cred);
+	free((caddr_t)sp, M_NFSREQ);
+	return (error);
+}
+
+/*
+ * Look up a file name and optionally either update the file handle or
+ * allocate an nfsnode, depending on the value of npp.
+ * npp == NULL	--> just do the lookup
+ * *npp == NULL --> allocate a new nfsnode and make sure attributes are
+ *			handled too
+ * *npp != NULL --> update the file handle in the vnode
+ */
+int
+nfs_lookitup(dvp, name, len, cred, procp, npp)
+	register struct vnode *dvp;
+	char *name;
+	int len;
+	struct ucred *cred;
+	struct proc *procp;
+	struct nfsnode **npp;
+{
+	register u_long *tl;
+	register caddr_t cp;
+	register long t1, t2;
+	struct vnode *newvp = (struct vnode *)0;
+	struct nfsnode *np, *dnp = VTONFS(dvp);
+	caddr_t bpos, dpos, cp2;
+	int error = 0, fhlen, attrflag;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	nfsfh_t *nfhp;
+	int v3 = NFS_ISV3(dvp);
+
+	nfsstats.rpccnt[NFSPROC_LOOKUP]++;
+	nfsm_reqhead(dvp, NFSPROC_LOOKUP,
+		NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len));
+	nfsm_fhtom(dvp, v3);
+	nfsm_strtom(name, len, NFS_MAXNAMLEN);
+	nfsm_request(dvp, NFSPROC_LOOKUP, procp, cred);
+	if (npp && !error) {
+		nfsm_getfh(nfhp, fhlen, v3);
+		if (*npp) {
+		    np = *npp;
+		    if (np->n_fhsize > NFS_SMALLFH && fhlen <= NFS_SMALLFH) {
+			free((caddr_t)np->n_fhp, M_NFSBIGFH);
+			np->n_fhp = &np->n_fh;
+		    } else if (np->n_fhsize <= NFS_SMALLFH && fhlen>NFS_SMALLFH)
+			np->n_fhp =(nfsfh_t *)malloc(fhlen,M_NFSBIGFH,M_WAITOK);
+		    bcopy((caddr_t)nfhp, (caddr_t)np->n_fhp, fhlen);
+		    np->n_fhsize = fhlen;
+		    newvp = NFSTOV(np);
+		} else if (NFS_CMPFH(dnp, nfhp, fhlen)) {
+		    VREF(dvp);
+		    newvp = dvp;
+		} else {
+		    error = nfs_nget(dvp->v_mount, nfhp, fhlen, &np);
+		    if (error) {
+			m_freem(mrep);
+			return (error);
+		    }
+		    newvp = NFSTOV(np);
+		}
+		if (v3) {
+			nfsm_postop_attr(newvp, attrflag);
+			if (!attrflag && *npp == NULL) {
+				m_freem(mrep);
+				vrele(newvp);
+				return (ENOENT);
+			}
+		} else
+			nfsm_loadattr(newvp, (struct vattr *)0);
+	}
+	nfsm_reqdone;
+	if (npp && *npp == NULL) {
+		if (error) {
+			if (newvp)
+				vrele(newvp);
+		} else
+			*npp = np;
+	}
+	return (error);
+}
+
+/*
+ * Nfs Version 3 commit rpc
+ */
+int
+nfs_commit(vp, offset, cnt, cred, procp)
+	register struct vnode *vp;
+	u_quad_t offset;
+	int cnt;
+	struct ucred *cred;
+	struct proc *procp;
+{
+	register caddr_t cp;
+	register u_long *tl;
+	register int t1, t2;
+	register struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	caddr_t bpos, dpos, cp2;
+	int error = 0, wccflag = NFSV3_WCCRATTR;
+	struct mbuf *mreq, *mrep, *md, *mb, *mb2;
+	
+	if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0)
+		return (0);
+	nfsstats.rpccnt[NFSPROC_COMMIT]++;
+	nfsm_reqhead(vp, NFSPROC_COMMIT, NFSX_FH(1));
+	nfsm_fhtom(vp, 1);
+	nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED);
+	txdr_hyper(&offset, tl);
+	tl += 2;
+	*tl = txdr_unsigned(cnt);
+	nfsm_request(vp, NFSPROC_COMMIT, procp, cred);
+	nfsm_wcc_data(vp, wccflag);
+	if (!error) {
+		nfsm_dissect(tl, u_long *, NFSX_V3WRITEVERF);
+		if (bcmp((caddr_t)nmp->nm_verf, (caddr_t)tl,
+			NFSX_V3WRITEVERF)) {
+			bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
+				NFSX_V3WRITEVERF);
+			error = NFSERR_STALEWRITEVERF;
+		}
+	}
+	nfsm_reqdone;
+	return (error);
+}
+
+/*
+ * Kludge City..
+ * - make nfs_bmap() essentially a no-op that does no translation
+ * - do nfs_strategy() by doing I/O with nfs_readrpc/nfs_writerpc
+ *   (Maybe I could use the process's page mapping, but I was concerned that
+ *    Kernel Write might not be enabled and also figured copyout() would do
+ *    a lot more work than bcopy() and also it currently happens in the
+ *    context of the swapper process (2).
+ */
+int
+nfs_bmap(ap)
+	struct vop_bmap_args /* {
+		struct vnode *a_vp;
+		daddr_t  a_bn;
+		struct vnode **a_vpp;
+		daddr_t *a_bnp;
+		int *a_runp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+
+	if (ap->a_vpp != NULL)
+		*ap->a_vpp = vp;
+	if (ap->a_bnp != NULL)
+		*ap->a_bnp = ap->a_bn * btodb(vp->v_mount->mnt_stat.f_iosize);
+	return (0);
+}
+
+/*
+ * Strategy routine.
+ * For async requests when nfsiod(s) are running, queue the request by
+ * calling nfs_asyncio(), otherwise just all nfs_doio() to do the
+ * request.
+ */
+int
+nfs_strategy(ap)
+	struct vop_strategy_args *ap;
+{
+	register struct buf *bp = ap->a_bp;
+	struct ucred *cr;
+	struct proc *p;
+	int error = 0;
+
+	if (bp->b_flags & B_PHYS)
+		panic("nfs physio");
+	if (bp->b_flags & B_ASYNC)
+		p = (struct proc *)0;
+	else
+		p = curproc;	/* XXX */
+	if (bp->b_flags & B_READ)
+		cr = bp->b_rcred;
+	else
+		cr = bp->b_wcred;
+	/*
+	 * If the op is asynchronous and an i/o daemon is waiting
+	 * queue the request, wake it up and wait for completion
+	 * otherwise just do it ourselves.
+	 */
+	if ((bp->b_flags & B_ASYNC) == 0 ||
+		nfs_asyncio(bp, NOCRED))
+		error = nfs_doio(bp, cr, p);
+	return (error);
+}
+
+/*
+ * Mmap a file
+ *
+ * NB Currently unsupported.
+ */
+/* ARGSUSED */
+int
+nfs_mmap(ap)
+	struct vop_mmap_args /* {
+		struct vnode *a_vp;
+		int  a_fflags;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+
+	return (EINVAL);
+}
+
+/*
+ * fsync vnode op. Just call nfs_flush() with commit == 1.
+ */
+/* ARGSUSED */
+int
+nfs_fsync(ap)
+	struct vop_fsync_args /* {
+		struct vnodeop_desc *a_desc;
+		struct vnode * a_vp;
+		struct ucred * a_cred;
+		int  a_waitfor;
+		struct proc * a_p;
+	} */ *ap;
+{
+
+	return (nfs_flush(ap->a_vp, ap->a_cred, ap->a_waitfor, ap->a_p, 1));
+}
+
+/*
+ * Flush all the blocks associated with a vnode.
+ * 	Walk through the buffer pool and push any dirty pages
+ *	associated with the vnode.
+ */
+int
+nfs_flush(vp, cred, waitfor, p, commit)
+	register struct vnode *vp;
+	struct ucred *cred;
+	int waitfor;
+	struct proc *p;
+	int commit;
+{
+	register struct nfsnode *np = VTONFS(vp);
+	register struct buf *bp;
+	register int i;
+	struct buf *nbp;
+	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
+	int s, error = 0, slptimeo = 0, slpflag = 0, retv, bvecpos;
+	int passone = 1;
+	u_quad_t off = (u_quad_t)-1, endoff = 0, toff;
+#ifndef NFS_COMMITBVECSIZ
+#define NFS_COMMITBVECSIZ	20
+#endif
+	struct buf *bvec[NFS_COMMITBVECSIZ];
+
+	if (nmp->nm_flag & NFSMNT_INT)
+		slpflag = PCATCH;
+	if (!commit)
+		passone = 0;
+	/*
+	 * A b_flags == (B_DELWRI | B_NEEDCOMMIT) block has been written to the
+	 * server, but nas not been committed to stable storage on the server
+	 * yet. On the first pass, the byte range is worked out and the commit
+	 * rpc is done. On the second pass, nfs_writebp() is called to do the
+	 * job.
+	 */
+again:
+	bvecpos = 0;
+	if (NFS_ISV3(vp) && commit) {
+		s = splbio();
+		for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
+			nbp = bp->b_vnbufs.le_next;
+			if (bvecpos >= NFS_COMMITBVECSIZ)
+				break;
+			if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT))
+				!= (B_DELWRI | B_NEEDCOMMIT))
+				continue;
+			bremfree(bp);
+			bp->b_flags |= (B_BUSY | B_WRITEINPROG);
+			/*
+			 * A list of these buffers is kept so that the
+			 * second loop knows which buffers have actually
+			 * been committed. This is necessary, since there
+			 * may be a race between the commit rpc and new
+			 * uncommitted writes on the file.
+			 */
+			bvec[bvecpos++] = bp;
+			toff = ((u_quad_t)bp->b_blkno) * DEV_BSIZE +
+				bp->b_dirtyoff;
+			if (toff < off)
+				off = toff;
+			toff += (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff);
+			if (toff > endoff)
+				endoff = toff;
+		}
+		splx(s);
+	}
+	if (bvecpos > 0) {
+		/*
+		 * Commit data on the server, as required.
+		 */
+		retv = nfs_commit(vp, off, (int)(endoff - off), cred, p);
+		if (retv == NFSERR_STALEWRITEVERF)
+			nfs_clearcommit(vp->v_mount);
+		/*
+		 * Now, either mark the blocks I/O done or mark the
+		 * blocks dirty, depending on whether the commit
+		 * succeeded.
+		 */
+		for (i = 0; i < bvecpos; i++) {
+			bp = bvec[i];
+			bp->b_flags &= ~(B_NEEDCOMMIT | B_WRITEINPROG);
+			if (retv) {
+			    brelse(bp);
+			} else {
+			    vp->v_numoutput++;
+			    bp->b_flags |= B_ASYNC;
+			    bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI);
+			    bp->b_dirtyoff = bp->b_dirtyend = 0;
+			    reassignbuf(bp, vp);
+			    biodone(bp);
+			}
+		}
+	}
+
+	/*
+	 * Start/do any write(s) that are required.
+	 */
+loop:
+	s = splbio();
+	for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
+		nbp = bp->b_vnbufs.le_next;
+		if (bp->b_flags & B_BUSY) {
+			if (waitfor != MNT_WAIT || passone)
+				continue;
+			bp->b_flags |= B_WANTED;
+			error = tsleep((caddr_t)bp, slpflag | (PRIBIO + 1),
+				"nfsfsync", slptimeo);
+			splx(s);
+			if (error) {
+			    if (nfs_sigintr(nmp, (struct nfsreq *)0, p))
+				return (EINTR);
+			    if (slpflag == PCATCH) {
+				slpflag = 0;
+				slptimeo = 2 * hz;
+			    }
+			}
+			goto loop;
+		}
+		if ((bp->b_flags & B_DELWRI) == 0)
+			panic("nfs_fsync: not dirty");
+		if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT))
+			continue;
+		bremfree(bp);
+		if (passone || !commit)
+		    bp->b_flags |= (B_BUSY|B_ASYNC);
+		else
+		    bp->b_flags |= (B_BUSY|B_ASYNC|B_WRITEINPROG|B_NEEDCOMMIT);
+		splx(s);
+		VOP_BWRITE(bp);
+		goto loop;
+	}
+	splx(s);
+	if (passone) {
+		passone = 0;
+		goto again;
+	}
+	if (waitfor == MNT_WAIT) {
+		while (vp->v_numoutput) {
+			vp->v_flag |= VBWAIT;
+			error = tsleep((caddr_t)&vp->v_numoutput,
+				slpflag | (PRIBIO + 1), "nfsfsync", slptimeo);
+			if (error) {
+			    if (nfs_sigintr(nmp, (struct nfsreq *)0, p))
+				return (EINTR);
+			    if (slpflag == PCATCH) {
+				slpflag = 0;
+				slptimeo = 2 * hz;
+			    }
+			}
+		}
+		if (vp->v_dirtyblkhd.lh_first && commit) {
+#ifndef DIAGNOSTIC
+			vprint("nfs_fsync: dirty", vp);
+#endif
+			goto loop;
+		}
+	}
+	if (np->n_flag & NWRITEERR) {
+		error = np->n_error;
+		np->n_flag &= ~NWRITEERR;
+	}
+	return (error);
+}
+
+/*
+ * Return POSIX pathconf information applicable to nfs.
+ *
+ * The NFS V2 protocol doesn't support this, so just return EINVAL
+ * for V2.
+ */
+/* ARGSUSED */
+int
+nfs_pathconf(ap)
+	struct vop_pathconf_args /* {
+		struct vnode *a_vp;
+		int a_name;
+		int *a_retval;
+	} */ *ap;
+{
+
+	return (EINVAL);
+}
+
+/*
+ * NFS advisory byte-level locks.
+ * Currently unsupported.
+ */
+int
+nfs_advlock(ap)
+	struct vop_advlock_args /* {
+		struct vnode *a_vp;
+		caddr_t  a_id;
+		int  a_op;
+		struct flock *a_fl;
+		int  a_flags;
+	} */ *ap;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * Print out the contents of an nfsnode.
+ */
+int
+nfs_print(ap)
+	struct vop_print_args /* {
+		struct vnode *a_vp;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+
+	printf("tag VT_NFS, fileid %ld fsid 0x%lx",
+		np->n_vattr.va_fileid, np->n_vattr.va_fsid);
+	if (vp->v_type == VFIFO)
+		fifo_printinfo(vp);
+	printf("\n");
+	return (0);
+}
+
+/*
+ * NFS directory offset lookup.
+ * Currently unsupported.
+ */
+int
+nfs_blkatoff(ap)
+	struct vop_blkatoff_args /* {
+		struct vnode *a_vp;
+		off_t a_offset;
+		char **a_res;
+		struct buf **a_bpp;
+	} */ *ap;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * NFS flat namespace allocation.
+ * Currently unsupported.
+ */
+int
+nfs_valloc(ap)
+	struct vop_valloc_args /* {
+		struct vnode *a_pvp;
+		int a_mode;
+		struct ucred *a_cred;
+		struct vnode **a_vpp;
+	} */ *ap;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * NFS flat namespace free.
+ * Currently unsupported.
+ */
+int
+nfs_vfree(ap)
+	struct vop_vfree_args /* {
+		struct vnode *a_pvp;
+		ino_t a_ino;
+		int a_mode;
+	} */ *ap;
+{
+
+	return (EOPNOTSUPP);
+}
+
+/*
+ * NFS file truncation.
+ */
+int
+nfs_truncate(ap)
+	struct vop_truncate_args /* {
+		struct vnode *a_vp;
+		off_t a_length;
+		int a_flags;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+
+	/* Use nfs_setattr */
+	printf("nfs_truncate: need to implement!!");
+	return (EOPNOTSUPP);
+}
+
+/*
+ * NFS update.
+ */
+int
+nfs_update(ap)
+	struct vop_update_args /* {
+		struct vnode *a_vp;
+		struct timeval *a_ta;
+		struct timeval *a_tm;
+		int a_waitfor;
+	} */ *ap;
+{
+
+	/* Use nfs_setattr */
+	printf("nfs_update: need to implement!!");
+	return (EOPNOTSUPP);
+}
+
+/*
+ * Just call nfs_writebp() with the force argument set to 1.
+ */
+int
+nfs_bwrite(ap)
+	struct vop_bwrite_args /* {
+		struct vnode *a_bp;
+	} */ *ap;
+{
+
+	return (nfs_writebp(ap->a_bp, 1));
+}
+
+/*
+ * This is a clone of vn_bwrite(), except that B_WRITEINPROG isn't set unless
+ * the force flag is one and it also handles the B_NEEDCOMMIT flag.
+ */
+int
+nfs_writebp(bp, force)
+	register struct buf *bp;
+	int force;
+{
+	register int oldflags = bp->b_flags, retv = 1;
+	register struct proc *p = curproc;	/* XXX */
+	off_t off;
+
+	if(!(bp->b_flags & B_BUSY))
+		panic("bwrite: buffer is not busy???");
+
+	bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI);
+
+	if (oldflags & B_ASYNC) {
+		if (oldflags & B_DELWRI) {
+			reassignbuf(bp, bp->b_vp);
+		} else if (p) {
+			++p->p_stats->p_ru.ru_oublock;
+		}
+	}
+	bp->b_vp->v_numoutput++;
+
+	/*
+	 * If B_NEEDCOMMIT is set, a commit rpc may do the trick. If not
+	 * an actual write will have to be scheduled via. VOP_STRATEGY().
+	 * If B_WRITEINPROG is already set, then push it with a write anyhow.
+	 */
+	if (oldflags & (B_NEEDCOMMIT | B_WRITEINPROG) == B_NEEDCOMMIT) {
+		off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
+		bp->b_flags |= B_WRITEINPROG;
+		retv = nfs_commit(bp->b_vp, off, bp->b_dirtyend-bp->b_dirtyoff,
+			bp->b_wcred, bp->b_proc);
+		bp->b_flags &= ~B_WRITEINPROG;
+		if (!retv) {
+			bp->b_dirtyoff = bp->b_dirtyend = 0;
+			bp->b_flags &= ~B_NEEDCOMMIT;
+			biodone(bp);
+		} else if (retv == NFSERR_STALEWRITEVERF)
+			nfs_clearcommit(bp->b_vp->v_mount);
+	}
+	if (retv) {
+		if (force)
+			bp->b_flags |= B_WRITEINPROG;
+		VOP_STRATEGY(bp);
+	}
+
+	if( (oldflags & B_ASYNC) == 0) {
+		int rtval = biowait(bp);
+		if (oldflags & B_DELWRI) {
+			reassignbuf(bp, bp->b_vp);
+		} else if (p) {
+			++p->p_stats->p_ru.ru_oublock;
+		}
+		brelse(bp);
+		return (rtval);
+	} 
+
+	return (0);
+}
+
+/*
+ * nfs special file access vnode op.
+ * Essentially just get vattr and then imitate iaccess() since the device is
+ * local to the client.
+ */
+int
+nfsspec_access(ap)
+	struct vop_access_args /* {
+		struct vnode *a_vp;
+		int  a_mode;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vattr *vap;
+	register gid_t *gp;
+	register struct ucred *cred = ap->a_cred;
+	struct vnode *vp = ap->a_vp;
+	mode_t mode = ap->a_mode;
+	struct vattr vattr;
+	register int i;
+	int error;
+
+	/*
+	 * Disallow write attempts on filesystems mounted read-only;
+	 * unless the file is a socket, fifo, or a block or character
+	 * device resident on the filesystem.
+	 */
+	if ((mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
+		switch (vp->v_type) {
+		case VREG: case VDIR: case VLNK:
+			return (EROFS);
+		}
+	}
+	/*
+	 * If you're the super-user,
+	 * you always get access.
+	 */
+	if (cred->cr_uid == 0)
+		return (0);
+	vap = &vattr;
+	error = VOP_GETATTR(vp, vap, cred, ap->a_p);
+	if (error)
+		return (error);
+	/*
+	 * Access check is based on only one of owner, group, public.
+	 * If not owner, then check group. If not a member of the
+	 * group, then check public access.
+	 */
+	if (cred->cr_uid != vap->va_uid) {
+		mode >>= 3;
+		gp = cred->cr_groups;
+		for (i = 0; i < cred->cr_ngroups; i++, gp++)
+			if (vap->va_gid == *gp)
+				goto found;
+		mode >>= 3;
+found:
+		;
+	}
+	error = (vap->va_mode & mode) == mode ? 0 : EACCES;
+	return (error);
+}
+
+/*
+ * Read wrapper for special devices.
+ */
+int
+nfsspec_read(ap)
+	struct vop_read_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register struct nfsnode *np = VTONFS(ap->a_vp);
+
+	/*
+	 * Set access flag.
+	 */
+	np->n_flag |= NACC;
+	np->n_atim.ts_sec = time.tv_sec;
+	np->n_atim.ts_nsec = time.tv_usec * 1000;
+	return (VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap));
+}
+
+/*
+ * Write wrapper for special devices.
+ */
+int
+nfsspec_write(ap)
+	struct vop_write_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	register struct nfsnode *np = VTONFS(ap->a_vp);
+
+	/*
+	 * Set update flag.
+	 */
+	np->n_flag |= NUPD;
+	np->n_mtim.ts_sec = time.tv_sec;
+	np->n_mtim.ts_nsec = time.tv_usec * 1000;
+	return (VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap));
+}
+
+/*
+ * Close wrapper for special devices.
+ *
+ * Update the times on the nfsnode then do device close.
+ */
+int
+nfsspec_close(ap)
+	struct vop_close_args /* {
+		struct vnode *a_vp;
+		int  a_fflag;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	struct vattr vattr;
+
+	if (np->n_flag & (NACC | NUPD)) {
+		np->n_flag |= NCHG;
+		if (vp->v_usecount == 1 &&
+		    (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
+			VATTR_NULL(&vattr);
+			if (np->n_flag & NACC)
+				vattr.va_atime = np->n_atim;
+			if (np->n_flag & NUPD)
+				vattr.va_mtime = np->n_mtim;
+			(void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p);
+		}
+	}
+	return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
+}
+
+/*
+ * Read wrapper for fifos.
+ */
+int
+nfsfifo_read(ap)
+	struct vop_read_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	extern int (**fifo_vnodeop_p)();
+	register struct nfsnode *np = VTONFS(ap->a_vp);
+
+	/*
+	 * Set access flag.
+	 */
+	np->n_flag |= NACC;
+	np->n_atim.ts_sec = time.tv_sec;
+	np->n_atim.ts_nsec = time.tv_usec * 1000;
+	return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap));
+}
+
+/*
+ * Write wrapper for fifos.
+ */
+int
+nfsfifo_write(ap)
+	struct vop_write_args /* {
+		struct vnode *a_vp;
+		struct uio *a_uio;
+		int  a_ioflag;
+		struct ucred *a_cred;
+	} */ *ap;
+{
+	extern int (**fifo_vnodeop_p)();
+	register struct nfsnode *np = VTONFS(ap->a_vp);
+
+	/*
+	 * Set update flag.
+	 */
+	np->n_flag |= NUPD;
+	np->n_mtim.ts_sec = time.tv_sec;
+	np->n_mtim.ts_nsec = time.tv_usec * 1000;
+	return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap));
+}
+
+/*
+ * Close wrapper for fifos.
+ *
+ * Update the times on the nfsnode then do fifo close.
+ */
+int
+nfsfifo_close(ap)
+	struct vop_close_args /* {
+		struct vnode *a_vp;
+		int  a_fflag;
+		struct ucred *a_cred;
+		struct proc *a_p;
+	} */ *ap;
+{
+	register struct vnode *vp = ap->a_vp;
+	register struct nfsnode *np = VTONFS(vp);
+	struct vattr vattr;
+	extern int (**fifo_vnodeop_p)();
+
+	if (np->n_flag & (NACC | NUPD)) {
+		if (np->n_flag & NACC) {
+			np->n_atim.ts_sec = time.tv_sec;
+			np->n_atim.ts_nsec = time.tv_usec * 1000;
+		}
+		if (np->n_flag & NUPD) {
+			np->n_mtim.ts_sec = time.tv_sec;
+			np->n_mtim.ts_nsec = time.tv_usec * 1000;
+		}
+		np->n_flag |= NCHG;
+		if (vp->v_usecount == 1 &&
+		    (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
+			VATTR_NULL(&vattr);
+			if (np->n_flag & NACC)
+				vattr.va_atime = np->n_atim;
+			if (np->n_flag & NUPD)
+				vattr.va_mtime = np->n_mtim;
+			(void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p);
+		}
+	}
+	return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
+}
diff --git a/sys/nfs/nfsdiskless.h b/sys/nfs/nfsdiskless.h
new file mode 100644
index 000000000000..9f864a47d751
--- /dev/null
+++ b/sys/nfs/nfsdiskless.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 1991, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsdiskless.h	8.2 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NFSDISKLESS_H_
+#define _NFS_NFSDISKLESS_H_
+
+/*
+ * Structure that must be initialized for a diskless nfs client.
+ * This structure is used by nfs_mountroot() to set up the root and swap
+ * vnodes plus do a partial ifconfig(8) and route(8) so that the critical net
+ * interface can communicate with the server.
+ * The primary bootstrap is expected to fill in the appropriate fields before
+ * starting the kernel. Whether or not the swap area is nfs mounted is
+ * determined by the value in swdevt[0]. (equal to NODEV --> swap over nfs)
+ * Currently only works for AF_INET protocols.
+ * NB: All fields are stored in net byte order to avoid hassles with
+ * client/server byte ordering differences.
+ */
+
+/*
+ * I have defined a new structure that can handle an NFS Version 3 file handle
+ * but the kernel still expects the old Version 2 one to be provided. The
+ * changes required in nfs_vfsops.c for using the new are documented there in
+ * comments. (I felt that breaking network booting code by changing this
+ * structure would not be prudent at this time, since almost all servers are
+ * still Version 2 anyhow.)
+ */
+struct nfsv3_diskless {
+	struct ifaliasreq myif;			/* Default interface */
+	struct sockaddr_in mygateway;		/* Default gateway */
+	struct nfs_args	swap_args;		/* Mount args for swap file */
+	int		swap_fhsize;		/* Size of file handle */
+	u_char		swap_fh[NFSX_V3FHMAX];	/* Swap file's file handle */
+	struct sockaddr_in swap_saddr;		/* Address of swap server */
+	char		swap_hostnam[MNAMELEN];	/* Host name for mount pt */
+	int		swap_nblks;		/* Size of server swap file */
+	struct ucred	swap_ucred;		/* Swap credentials */
+	struct nfs_args	root_args;		/* Mount args for root fs */
+	int		root_fhsize;		/* Size of root file handle */
+	u_char		root_fh[NFSX_V3FHMAX];	/* File handle of root dir */
+	struct sockaddr_in root_saddr;		/* Address of root server */
+	char		root_hostnam[MNAMELEN];	/* Host name for mount pt */
+	long		root_time;		/* Timestamp of root fs */
+	char		my_hostnam[MAXHOSTNAMELEN]; /* Client host name */
+};
+
+struct nfs_diskless {
+	struct ifaliasreq myif;			/* Default interface */
+	struct sockaddr_in mygateway;		/* Default gateway */
+	struct nfs_args	swap_args;		/* Mount args for swap file */
+	u_char		swap_fh[NFSX_V2FH];	/* Swap file's file handle */
+	struct sockaddr_in swap_saddr;		/* Address of swap server */
+	char		swap_hostnam[MNAMELEN];	/* Host name for mount pt */
+	int		swap_nblks;		/* Size of server swap file */
+	struct ucred	swap_ucred;		/* Swap credentials */
+	struct nfs_args	root_args;		/* Mount args for root fs */
+	u_char		root_fh[NFSX_V2FH];	/* File handle of root dir */
+	struct sockaddr_in root_saddr;		/* Address of root server */
+	char		root_hostnam[MNAMELEN];	/* Host name for mount pt */
+	long		root_time;		/* Timestamp of root fs */
+	char		my_hostnam[MAXHOSTNAMELEN]; /* Client host name */
+};
+
+#endif
diff --git a/sys/nfs/nfsm_subs.h b/sys/nfs/nfsm_subs.h
new file mode 100644
index 000000000000..95c16462e4e8
--- /dev/null
+++ b/sys/nfs/nfsm_subs.h
@@ -0,0 +1,439 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsm_subs.h	8.2 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NFSM_SUBS_H_
+#define _NFS_NFSM_SUBS_H_
+
+
+/*
+ * These macros do strange and peculiar things to mbuf chains for
+ * the assistance of the nfs code. To attempt to use them for any
+ * other purpose will be dangerous. (they make weird assumptions)
+ */
+
+/*
+ * First define what the actual subs. return
+ */
+extern struct mbuf *nfsm_reqh();
+
+#define	M_HASCL(m)	((m)->m_flags & M_EXT)
+#define	NFSMINOFF(m) \
+		if (M_HASCL(m)) \
+			(m)->m_data = (m)->m_ext.ext_buf; \
+		else if ((m)->m_flags & M_PKTHDR) \
+			(m)->m_data = (m)->m_pktdat; \
+		else \
+			(m)->m_data = (m)->m_dat
+#define	NFSMADV(m, s)	(m)->m_data += (s)
+#define	NFSMSIZ(m)	((M_HASCL(m))?MCLBYTES: \
+				(((m)->m_flags & M_PKTHDR)?MHLEN:MLEN))
+
+/*
+ * Now for the macros that do the simple stuff and call the functions
+ * for the hard stuff.
+ * These macros use several vars. declared in nfsm_reqhead and these
+ * vars. must not be used elsewhere unless you are careful not to corrupt
+ * them. The vars. starting with pN and tN (N=1,2,3,..) are temporaries
+ * that may be used so long as the value is not expected to retained
+ * after a macro.
+ * I know, this is kind of dorkey, but it makes the actual op functions
+ * fairly clean and deals with the mess caused by the xdr discriminating
+ * unions.
+ */
+
+#define	nfsm_build(a,c,s) \
+		{ if ((s) > M_TRAILINGSPACE(mb)) { \
+			MGET(mb2, M_WAIT, MT_DATA); \
+			if ((s) > MLEN) \
+				panic("build > MLEN"); \
+			mb->m_next = mb2; \
+			mb = mb2; \
+			mb->m_len = 0; \
+			bpos = mtod(mb, caddr_t); \
+		} \
+		(a) = (c)(bpos); \
+		mb->m_len += (s); \
+		bpos += (s); }
+
+#define	nfsm_dissect(a, c, s) \
+		{ t1 = mtod(md, caddr_t)+md->m_len-dpos; \
+		if (t1 >= (s)) { \
+			(a) = (c)(dpos); \
+			dpos += (s); \
+		} else if (t1 = nfsm_disct(&md, &dpos, (s), t1, &cp2)) { \
+			error = t1; \
+			m_freem(mrep); \
+			goto nfsmout; \
+		} else { \
+			(a) = (c)cp2; \
+		} }
+
+#define nfsm_fhtom(v, v3) \
+	      { if (v3) { \
+			t2 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; \
+			if (t2 <= M_TRAILINGSPACE(mb)) { \
+				nfsm_build(tl, u_long *, t2); \
+				*tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); \
+				*(tl + ((t2>>2) - 2)) = 0; \
+				bcopy((caddr_t)VTONFS(v)->n_fhp,(caddr_t)tl, \
+					VTONFS(v)->n_fhsize); \
+			} else if (t2 = nfsm_strtmbuf(&mb, &bpos, \
+				(caddr_t)VTONFS(v)->n_fhp, VTONFS(v)->n_fhsize)) { \
+				error = t2; \
+				m_freem(mreq); \
+				goto nfsmout; \
+			} \
+		} else { \
+			nfsm_build(cp, caddr_t, NFSX_V2FH); \
+			bcopy((caddr_t)VTONFS(v)->n_fhp, cp, NFSX_V2FH); \
+		} }
+
+#define nfsm_srvfhtom(f, v3) \
+		{ if (v3) { \
+			nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FH); \
+			*tl++ = txdr_unsigned(NFSX_V3FH); \
+			bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \
+		} else { \
+			nfsm_build(cp, caddr_t, NFSX_V2FH); \
+			bcopy((caddr_t)(f), cp, NFSX_V2FH); \
+		} }
+
+#define nfsm_srvpostop_fh(f) \
+		{ nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED + NFSX_V3FH); \
+		*tl++ = nfs_true; \
+		*tl++ = txdr_unsigned(NFSX_V3FH); \
+		bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \
+		}
+
+#define nfsm_mtofh(d, v, v3, f) \
+		{ struct nfsnode *ttnp; nfsfh_t *ttfhp; int ttfhsize; \
+		if (v3) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			(f) = fxdr_unsigned(int, *tl); \
+		} else \
+			(f) = 1; \
+		if (f) { \
+			nfsm_getfh(ttfhp, ttfhsize, (v3)); \
+			if (t1 = nfs_nget((d)->v_mount, ttfhp, ttfhsize, \
+				&ttnp)) { \
+				error = t1; \
+				m_freem(mrep); \
+				goto nfsmout; \
+			} \
+			(v) = NFSTOV(ttnp); \
+		} \
+		if (v3) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			if (f) \
+				(f) = fxdr_unsigned(int, *tl); \
+			else if (fxdr_unsigned(int, *tl)) \
+				nfsm_adv(NFSX_V3FATTR); \
+		} \
+		if (f) \
+			nfsm_loadattr((v), (struct vattr *)0); \
+		}
+
+#define nfsm_getfh(f, s, v3) \
+		{ if (v3) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			if (((s) = fxdr_unsigned(int, *tl)) <= 0 || \
+				(s) > NFSX_V3FHMAX) { \
+				m_freem(mrep); \
+				error = EBADRPC; \
+				goto nfsmout; \
+			} \
+		} else \
+			(s) = NFSX_V2FH; \
+		nfsm_dissect((f), nfsfh_t *, nfsm_rndup(s)); }
+
+#define	nfsm_loadattr(v, a) \
+		{ struct vnode *ttvp = (v); \
+		if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, (a))) { \
+			error = t1; \
+			m_freem(mrep); \
+			goto nfsmout; \
+		} \
+		(v) = ttvp; }
+
+#define	nfsm_postop_attr(v, f) \
+		{ struct vnode *ttvp = (v); \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if ((f) = fxdr_unsigned(int, *tl)) { \
+			if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, \
+				(struct vattr *)0)) { \
+				error = t1; \
+				(f) = 0; \
+				m_freem(mrep); \
+				goto nfsmout; \
+			} \
+			(v) = ttvp; \
+		} }
+
+/* Used as (f) for nfsm_wcc_data() */
+#define NFSV3_WCCRATTR	0
+#define NFSV3_WCCCHK	1
+
+#define	nfsm_wcc_data(v, f) \
+		{ int ttattrf, ttretf = 0; \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if (*tl == nfs_true) { \
+			nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); \
+			if (f) \
+				ttretf = (VTONFS(v)->n_mtime == \
+					fxdr_unsigned(u_long, *(tl + 2))); \
+		} \
+		nfsm_postop_attr((v), ttattrf); \
+		if (f) { \
+			(f) = ttretf; \
+		} else { \
+			(f) = ttattrf; \
+		} }
+
+#define nfsm_v3sattr(s, a, u, g) \
+		{ (s)->sa_modetrue = nfs_true; \
+		(s)->sa_mode = vtonfsv3_mode((a)->va_mode); \
+		(s)->sa_uidtrue = nfs_true; \
+		(s)->sa_uid = txdr_unsigned(u); \
+		(s)->sa_gidtrue = nfs_true; \
+		(s)->sa_gid = txdr_unsigned(g); \
+		(s)->sa_sizefalse = nfs_false; \
+		(s)->sa_atimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \
+		txdr_nfsv3time(&(a)->va_atime, &(s)->sa_atime); \
+		(s)->sa_mtimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \
+		txdr_nfsv3time(&(a)->va_mtime, &(s)->sa_mtime); \
+		}
+
+#define	nfsm_strsiz(s,m) \
+		{ nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \
+		if (((s) = fxdr_unsigned(long,*tl)) > (m)) { \
+			m_freem(mrep); \
+			error = EBADRPC; \
+			goto nfsmout; \
+		} }
+
+#define	nfsm_srvstrsiz(s,m) \
+		{ nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \
+		if (((s) = fxdr_unsigned(long,*tl)) > (m) || (s) <= 0) { \
+			error = EBADRPC; \
+			nfsm_reply(0); \
+		} }
+
+#define	nfsm_srvnamesiz(s) \
+		{ nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \
+		if (((s) = fxdr_unsigned(long,*tl)) > NFS_MAXNAMLEN) \
+			error = NFSERR_NAMETOL; \
+		if ((s) <= 0) \
+			error = EBADRPC; \
+		if (error) \
+			nfsm_reply(0); \
+		}
+
+#define nfsm_mtouio(p,s) \
+		if ((s) > 0 && \
+		   (t1 = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \
+			error = t1; \
+			m_freem(mrep); \
+			goto nfsmout; \
+		}
+
+#define nfsm_uiotom(p,s) \
+		if (t1 = nfsm_uiotombuf((p),&mb,(s),&bpos)) { \
+			error = t1; \
+			m_freem(mreq); \
+			goto nfsmout; \
+		}
+
+#define	nfsm_reqhead(v,a,s) \
+		mb = mreq = nfsm_reqh((v),(a),(s),&bpos)
+
+#define nfsm_reqdone	m_freem(mrep); \
+		nfsmout: 
+
+#define nfsm_rndup(a)	(((a)+3)&(~0x3))
+
+#define	nfsm_request(v, t, p, c)	\
+		if (error = nfs_request((v), mreq, (t), (p), \
+		   (c), &mrep, &md, &dpos)) { \
+			if (error & NFSERR_RETERR) \
+				error &= ~NFSERR_RETERR; \
+			else \
+				goto nfsmout; \
+		}
+
+#define	nfsm_strtom(a,s,m) \
+		if ((s) > (m)) { \
+			m_freem(mreq); \
+			error = ENAMETOOLONG; \
+			goto nfsmout; \
+		} \
+		t2 = nfsm_rndup(s)+NFSX_UNSIGNED; \
+		if (t2 <= M_TRAILINGSPACE(mb)) { \
+			nfsm_build(tl,u_long *,t2); \
+			*tl++ = txdr_unsigned(s); \
+			*(tl+((t2>>2)-2)) = 0; \
+			bcopy((caddr_t)(a), (caddr_t)tl, (s)); \
+		} else if (t2 = nfsm_strtmbuf(&mb, &bpos, (a), (s))) { \
+			error = t2; \
+			m_freem(mreq); \
+			goto nfsmout; \
+		}
+
+#define	nfsm_srvdone \
+		nfsmout: \
+		return(error)
+
+#define	nfsm_reply(s) \
+		{ \
+		nfsd->nd_repstat = error; \
+		if (error && !(nfsd->nd_flag & ND_NFSV3)) \
+		   (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \
+			mrq, &mb, &bpos); \
+		else \
+		   (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \
+			mrq, &mb, &bpos); \
+		m_freem(mrep); \
+		mreq = *mrq; \
+		if (error && (!(nfsd->nd_flag & ND_NFSV3) || \
+			error == EBADRPC)) \
+			return(0); \
+		}
+
+#define	nfsm_writereply(s, v3) \
+		{ \
+		nfsd->nd_repstat = error; \
+		if (error && !(v3)) \
+		   (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \
+			&mreq, &mb, &bpos); \
+		else \
+		   (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \
+			&mreq, &mb, &bpos); \
+		}
+
+#define	nfsm_adv(s) \
+		{ t1 = mtod(md, caddr_t)+md->m_len-dpos; \
+		if (t1 >= (s)) { \
+			dpos += (s); \
+		} else if (t1 = nfs_adv(&md, &dpos, (s), t1)) { \
+			error = t1; \
+			m_freem(mrep); \
+			goto nfsmout; \
+		} }
+
+#define nfsm_srvmtofh(f) \
+		{ if (nfsd->nd_flag & ND_NFSV3) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			if (fxdr_unsigned(int, *tl) != NFSX_V3FH) { \
+				error = EBADRPC; \
+				nfsm_reply(0); \
+			} \
+		} \
+		nfsm_dissect(tl, u_long *, NFSX_V3FH); \
+		bcopy((caddr_t)tl, (caddr_t)(f), NFSX_V3FH); \
+		if ((nfsd->nd_flag & ND_NFSV3) == 0) \
+			nfsm_adv(NFSX_V2FH - NFSX_V3FH); \
+		}
+
+#define	nfsm_clget \
+		if (bp >= be) { \
+			if (mp == mb) \
+				mp->m_len += bp-bpos; \
+			MGET(mp, M_WAIT, MT_DATA); \
+			MCLGET(mp, M_WAIT); \
+			mp->m_len = NFSMSIZ(mp); \
+			mp2->m_next = mp; \
+			mp2 = mp; \
+			bp = mtod(mp, caddr_t); \
+			be = bp+mp->m_len; \
+		} \
+		tl = (u_long *)bp
+
+#define	nfsm_srvfillattr(a, f) \
+		nfsm_srvfattr(nfsd, (a), (f))
+
+#define nfsm_srvwcc_data(br, b, ar, a) \
+		nfsm_srvwcc(nfsd, (br), (b), (ar), (a), &mb, &bpos)
+
+#define nfsm_srvpostop_attr(r, a) \
+		nfsm_srvpostopattr(nfsd, (r), (a), &mb, &bpos)
+
+#define nfsm_srvsattr(a) \
+		{ nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if (*tl == nfs_true) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			(a)->va_mode = nfstov_mode(*tl); \
+		} \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if (*tl == nfs_true) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			(a)->va_uid = fxdr_unsigned(uid_t, *tl); \
+		} \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if (*tl == nfs_true) { \
+			nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+			(a)->va_gid = fxdr_unsigned(gid_t, *tl); \
+		} \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		if (*tl == nfs_true) { \
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \
+			fxdr_hyper(tl, &(a)->va_size); \
+		} \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		switch (fxdr_unsigned(int, *tl)) { \
+		case NFSV3SATTRTIME_TOCLIENT: \
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \
+			fxdr_nfsv3time(tl, &(a)->va_atime); \
+			break; \
+		case NFSV3SATTRTIME_TOSERVER: \
+			(a)->va_atime.ts_sec = time.tv_sec; \
+			(a)->va_atime.ts_nsec = time.tv_usec * 1000; \
+			break; \
+		}; \
+		nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \
+		switch (fxdr_unsigned(int, *tl)) { \
+		case NFSV3SATTRTIME_TOCLIENT: \
+			nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \
+			fxdr_nfsv3time(tl, &(a)->va_mtime); \
+			break; \
+		case NFSV3SATTRTIME_TOSERVER: \
+			(a)->va_mtime.ts_sec = time.tv_sec; \
+			(a)->va_mtime.ts_nsec = time.tv_usec * 1000; \
+			break; \
+		}; }
+
+#endif
diff --git a/sys/nfs/nfsmount.h b/sys/nfs/nfsmount.h
new file mode 100644
index 000000000000..78fe04a33376
--- /dev/null
+++ b/sys/nfs/nfsmount.h
@@ -0,0 +1,141 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsmount.h	8.3 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NFSMOUNT_H_
+#define _NFS_NFSMOUNT_H_
+
+/*
+ * Mount structure.
+ * One allocated on every NFS mount.
+ * Holds NFS specific information for mount.
+ */
+struct	nfsmount {
+	int	nm_flag;		/* Flags for soft/hard... */
+	struct	mount *nm_mountp;	/* Vfs structure for this filesystem */
+	int	nm_numgrps;		/* Max. size of groupslist */
+	u_char	nm_fh[NFSX_V3FHMAX];	/* File handle of root dir */
+	int	nm_fhsize;		/* Size of root file handle */
+	struct	socket *nm_so;		/* Rpc socket */
+	int	nm_sotype;		/* Type of socket */
+	int	nm_soproto;		/* and protocol */
+	int	nm_soflags;		/* pr_flags for socket protocol */
+	struct	mbuf *nm_nam;		/* Addr of server */
+	int	nm_timeo;		/* Init timer for NFSMNT_DUMBTIMR */
+	int	nm_retry;		/* Max retries */
+	int	nm_srtt[4];		/* Timers for rpcs */
+	int	nm_sdrtt[4];
+	int	nm_sent;		/* Request send count */
+	int	nm_cwnd;		/* Request send window */
+	int	nm_timeouts;		/* Request timeouts */
+	int	nm_deadthresh;		/* Threshold of timeouts-->dead server*/
+	int	nm_rsize;		/* Max size of read rpc */
+	int	nm_wsize;		/* Max size of write rpc */
+	int	nm_readdirsize;		/* Size of a readdir rpc */
+	int	nm_readahead;		/* Num. of blocks to readahead */
+	int	nm_leaseterm;		/* Term (sec) for NQNFS lease */
+	CIRCLEQ_HEAD(, nfsnode) nm_timerhead; /* Head of lease timer queue */
+	struct vnode *nm_inprog;	/* Vnode in prog by nqnfs_clientd() */
+	uid_t	nm_authuid;		/* Uid for authenticator */
+	int	nm_authtype;		/* Authenticator type */
+	int	nm_authlen;		/* and length */
+	char	*nm_authstr;		/* Authenticator string */
+	char	*nm_verfstr;		/* and the verifier */
+	int	nm_verflen;
+	u_char	nm_verf[NFSX_V3WRITEVERF]; /* V3 write verifier */
+	NFSKERBKEY_T nm_key;		/* and the session key */
+	int	nm_numuids;		/* Number of nfsuid mappings */
+	TAILQ_HEAD(, nfsuid) nm_uidlruhead; /* Lists of nfsuid mappings */
+	LIST_HEAD(, nfsuid) nm_uidhashtbl[NFS_MUIDHASHSIZ];
+};
+
+#ifdef KERNEL
+/*
+ * Convert mount ptr to nfsmount ptr.
+ */
+#define VFSTONFS(mp)	((struct nfsmount *)((mp)->mnt_data))
+#endif /* KERNEL */
+
+/*
+ * Prototypes for NFS mount operations
+ */
+int	nfs_mount __P((
+		struct mount *mp,
+		char *path,
+		caddr_t data,
+		struct nameidata *ndp,
+		struct proc *p));
+int	nfs_start __P((
+		struct mount *mp,
+		int flags,
+		struct proc *p));
+int	nfs_unmount __P((
+		struct mount *mp,
+		int mntflags,
+		struct proc *p));
+int	nfs_root __P((
+		struct mount *mp,
+		struct vnode **vpp));
+int	nfs_quotactl __P((
+		struct mount *mp,
+		int cmds,
+		uid_t uid,
+		caddr_t arg,
+		struct proc *p));
+int	nfs_statfs __P((
+		struct mount *mp,
+		struct statfs *sbp,
+		struct proc *p));
+int	nfs_sync __P((
+		struct mount *mp,
+		int waitfor,
+		struct ucred *cred,
+		struct proc *p));
+int	nfs_fhtovp __P((
+		struct mount *mp,
+		struct fid *fhp,
+		struct mbuf *nam,
+		struct vnode **vpp,
+		int *exflagsp,
+		struct ucred **credanonp));
+int	nfs_vptofh __P((
+		struct vnode *vp,
+		struct fid *fhp));
+int	nfs_init __P(());
+
+#endif
diff --git a/sys/nfs/nfsnode.h b/sys/nfs/nfsnode.h
new file mode 100644
index 000000000000..06107d5fdfe4
--- /dev/null
+++ b/sys/nfs/nfsnode.h
@@ -0,0 +1,220 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsnode.h	8.9 (Berkeley) 5/14/95
+ */
+
+
+#ifndef _NFS_NFSNODE_H_
+#define _NFS_NFSNODE_H_
+
+#ifndef _NFS_NFS_H_
+#include <nfs/nfs.h>
+#endif
+
+/*
+ * Silly rename structure that hangs off the nfsnode until the name
+ * can be removed by nfs_inactive()
+ */
+struct sillyrename {
+	struct	ucred *s_cred;
+	struct	vnode *s_dvp;
+	long	s_namlen;
+	char	s_name[20];
+};
+
+/*
+ * This structure is used to save the logical directory offset to
+ * NFS cookie mappings.
+ * The mappings are stored in a list headed
+ * by n_cookies, as required.
+ * There is one mapping for each NFS_DIRBLKSIZ bytes of directory information
+ * stored in increasing logical offset byte order.
+ */
+#define NFSNUMCOOKIES		31
+
+struct nfsdmap {
+	LIST_ENTRY(nfsdmap)	ndm_list;
+	int			ndm_eocookie;
+	nfsuint64		ndm_cookies[NFSNUMCOOKIES];
+};
+
+/*
+ * The nfsnode is the nfs equivalent to ufs's inode. Any similarity
+ * is purely coincidental.
+ * There is a unique nfsnode allocated for each active file,
+ * each current directory, each mounted-on file, text file, and the root.
+ * An nfsnode is 'named' by its file handle. (nget/nfs_node.c)
+ * If this structure exceeds 256 bytes (it is currently 256 using 4.4BSD-Lite
+ * type definitions), file handles of > 32 bytes should probably be split out
+ * into a separate MALLOC()'d data structure. (Reduce the size of nfsfh_t by
+ * changing the definition in sys/mount.h of NFS_SMALLFH.)
+ * NB: Hopefully the current order of the fields is such that everything will
+ *     be well aligned and, therefore, tightly packed.
+ */
+struct nfsnode {
+	LIST_ENTRY(nfsnode)	n_hash;		/* Hash chain */
+	CIRCLEQ_ENTRY(nfsnode)	n_timer;	/* Nqnfs timer chain */
+	u_quad_t		n_size;		/* Current size of file */
+	u_quad_t		n_brev;		/* Modify rev when cached */
+	u_quad_t		n_lrev;		/* Modify rev for lease */
+	struct vattr		n_vattr;	/* Vnode attribute cache */
+	time_t			n_attrstamp;	/* Attr. cache timestamp */
+	time_t			n_mtime;	/* Prev modify time. */
+	time_t			n_ctime;	/* Prev create time. */
+	time_t			n_expiry;	/* Lease expiry time */
+	nfsfh_t			*n_fhp;		/* NFS File Handle */
+	struct vnode		*n_vnode;	/* associated vnode */
+	struct lockf		*n_lockf;	/* Locking record of file */
+	int			n_error;	/* Save write error value */
+	union {
+		struct timespec	nf_atim;	/* Special file times */
+		nfsuint64	nd_cookieverf;	/* Cookie verifier (dir only) */
+	} n_un1;
+	union {
+		struct timespec	nf_mtim;
+		off_t		nd_direof;	/* Dir. EOF offset cache */
+	} n_un2;
+	union {
+		struct sillyrename *nf_silly;	/* Ptr to silly rename struct */
+		LIST_HEAD(, nfsdmap) nd_cook;	/* cookies */
+	} n_un3;
+	short			n_fhsize;	/* size in bytes, of fh */
+	short			n_flag;		/* Flag for locking.. */
+	nfsfh_t			n_fh;		/* Small File Handle */
+};
+
+#define n_atim		n_un1.nf_atim
+#define n_mtim		n_un2.nf_mtim
+#define n_sillyrename	n_un3.nf_silly
+#define n_cookieverf	n_un1.nd_cookieverf
+#define n_direofoffset	n_un2.nd_direof
+#define n_cookies	n_un3.nd_cook
+
+/*
+ * Flags for n_flag
+ */
+#define	NFLUSHWANT	0x0001	/* Want wakeup from a flush in prog. */
+#define	NFLUSHINPROG	0x0002	/* Avoid multiple calls to vinvalbuf() */
+#define	NMODIFIED	0x0004	/* Might have a modified buffer in bio */
+#define	NWRITEERR	0x0008	/* Flag write errors so close will know */
+#define	NQNFSNONCACHE	0x0020	/* Non-cachable lease */
+#define	NQNFSWRITE	0x0040	/* Write lease */
+#define	NQNFSEVICTED	0x0080	/* Has been evicted */
+#define	NACC		0x0100	/* Special file accessed */
+#define	NUPD		0x0200	/* Special file updated */
+#define	NCHG		0x0400	/* Special file times changed */
+
+/*
+ * Convert between nfsnode pointers and vnode pointers
+ */
+#define VTONFS(vp)	((struct nfsnode *)(vp)->v_data)
+#define NFSTOV(np)	((struct vnode *)(np)->n_vnode)
+
+/*
+ * Queue head for nfsiod's
+ */
+TAILQ_HEAD(, buf) nfs_bufq;
+
+#ifdef KERNEL
+/*
+ * Prototypes for NFS vnode operations
+ */
+int	nfs_lookup __P((struct vop_lookup_args *));
+int	nfs_create __P((struct vop_create_args *));
+int	nfs_mknod __P((struct vop_mknod_args *));
+int	nfs_open __P((struct vop_open_args *));
+int	nfs_close __P((struct vop_close_args *));
+int	nfsspec_close __P((struct vop_close_args *));
+int	nfsfifo_close __P((struct vop_close_args *));
+int	nfs_access __P((struct vop_access_args *));
+int	nfsspec_access __P((struct vop_access_args *));
+int	nfs_getattr __P((struct vop_getattr_args *));
+int	nfs_setattr __P((struct vop_setattr_args *));
+int	nfs_read __P((struct vop_read_args *));
+int	nfs_write __P((struct vop_write_args *));
+#define	nfs_lease_check ((int (*) __P((struct  vop_lease_args *)))nullop)
+#define nqnfs_vop_lease_check	lease_check
+int	nfsspec_read __P((struct vop_read_args *));
+int	nfsspec_write __P((struct vop_write_args *));
+int	nfsfifo_read __P((struct vop_read_args *));
+int	nfsfifo_write __P((struct vop_write_args *));
+#define nfs_ioctl ((int (*) __P((struct  vop_ioctl_args *)))enoioctl)
+#define nfs_select ((int (*) __P((struct  vop_select_args *)))seltrue)
+#define nfs_revoke vop_revoke
+int	nfs_mmap __P((struct vop_mmap_args *));
+int	nfs_fsync __P((struct vop_fsync_args *));
+#define nfs_seek ((int (*) __P((struct  vop_seek_args *)))nullop)
+int	nfs_remove __P((struct vop_remove_args *));
+int	nfs_link __P((struct vop_link_args *));
+int	nfs_rename __P((struct vop_rename_args *));
+int	nfs_mkdir __P((struct vop_mkdir_args *));
+int	nfs_rmdir __P((struct vop_rmdir_args *));
+int	nfs_symlink __P((struct vop_symlink_args *));
+int	nfs_readdir __P((struct vop_readdir_args *));
+int	nfs_readlink __P((struct vop_readlink_args *));
+int	nfs_abortop __P((struct vop_abortop_args *));
+int	nfs_inactive __P((struct vop_inactive_args *));
+int	nfs_reclaim __P((struct vop_reclaim_args *));
+#define nfs_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock)
+#define nfs_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock)
+#define nfs_islocked ((int (*) __P((struct vop_islocked_args *)))vop_noislocked)
+int	nfs_bmap __P((struct vop_bmap_args *));
+int	nfs_strategy __P((struct vop_strategy_args *));
+int	nfs_print __P((struct vop_print_args *));
+int	nfs_pathconf __P((struct vop_pathconf_args *));
+int	nfs_advlock __P((struct vop_advlock_args *));
+int	nfs_blkatoff __P((struct vop_blkatoff_args *));
+int	nfs_bwrite __P((struct vop_bwrite_args *));
+int	nfs_vget __P((struct mount *, ino_t, struct vnode **));
+int	nfs_valloc __P((struct vop_valloc_args *));
+#define nfs_reallocblks \
+	((int (*) __P((struct  vop_reallocblks_args *)))eopnotsupp)
+int	nfs_vfree __P((struct vop_vfree_args *));
+int	nfs_truncate __P((struct vop_truncate_args *));
+int	nfs_update __P((struct vop_update_args *));
+
+/* other stuff */
+int	nfs_removeit __P((struct sillyrename *));
+int	nfs_nget __P((struct mount *,nfsfh_t *,int,struct nfsnode **));
+int	nfs_lookitup __P((struct vnode *,char *,int,struct ucred *,struct proc *,struct nfsnode **));
+int	nfs_sillyrename __P((struct vnode *,struct vnode *,struct componentname *));
+nfsuint64 *nfs_getcookie __P((struct nfsnode *, off_t, int));
+void nfs_invaldir __P((struct vnode *));
+#define nqnfs_lease_updatetime	lease_updatetime
+
+#endif /* KERNEL */
+
+#endif
diff --git a/sys/nfs/nfsproto.h b/sys/nfs/nfsproto.h
new file mode 100644
index 000000000000..7dd909010ad4
--- /dev/null
+++ b/sys/nfs/nfsproto.h
@@ -0,0 +1,441 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsproto.h	8.2 (Berkeley) 3/30/95
+ */
+
+#ifndef _NFS_NFSPROTO_H_
+#define _NFS_NFSPROTO_H_
+
+
+/*
+ * nfs definitions as per the Version 2 and 3 specs
+ */
+
+/*
+ * Constants as defined in the Sun NFS Version 2 and 3 specs.
+ * "NFS: Network File System Protocol Specification" RFC1094
+ * and in the "NFS: Network File System Version 3 Protocol
+ * Specification"
+ */
+
+#define NFS_PORT	2049
+#define	NFS_PROG	100003
+#define NFS_VER2	2
+#define	NFS_VER3	3
+#define NFS_V2MAXDATA	8192
+#define	NFS_MAXDGRAMDATA 16384
+#define	NFS_MAXDATA	32768
+#define	NFS_MAXPATHLEN	1024
+#define	NFS_MAXNAMLEN	255
+#define	NFS_MAXPKTHDR	404
+#define NFS_MAXPACKET	(NFS_MAXPKTHDR + NFS_MAXDATA)
+#define	NFS_MINPACKET	20
+#define	NFS_FABLKSIZE	512	/* Size in bytes of a block wrt fa_blocks */
+
+/* Stat numbers for rpc returns (version 2 and 3) */
+#define	NFS_OK			0
+#define	NFSERR_PERM		1
+#define	NFSERR_NOENT		2
+#define	NFSERR_IO		5
+#define	NFSERR_NXIO		6
+#define	NFSERR_ACCES		13
+#define	NFSERR_EXIST		17
+#define	NFSERR_XDEV		18	/* Version 3 only */
+#define	NFSERR_NODEV		19
+#define	NFSERR_NOTDIR		20
+#define	NFSERR_ISDIR		21
+#define	NFSERR_INVAL		22	/* Version 3 only */
+#define	NFSERR_FBIG		27
+#define	NFSERR_NOSPC		28
+#define	NFSERR_ROFS		30
+#define	NFSERR_MLINK		31	/* Version 3 only */
+#define	NFSERR_NAMETOL		63
+#define	NFSERR_NOTEMPTY		66
+#define	NFSERR_DQUOT		69
+#define	NFSERR_STALE		70
+#define	NFSERR_REMOTE		71	/* Version 3 only */
+#define	NFSERR_WFLUSH		99	/* Version 2 only */
+#define	NFSERR_BADHANDLE	10001	/* The rest Version 3 only */
+#define	NFSERR_NOT_SYNC		10002
+#define	NFSERR_BAD_COOKIE	10003
+#define	NFSERR_NOTSUPP		10004
+#define	NFSERR_TOOSMALL		10005
+#define	NFSERR_SERVERFAULT	10006
+#define	NFSERR_BADTYPE		10007
+#define	NFSERR_JUKEBOX		10008
+#define NFSERR_TRYLATER		NFSERR_JUKEBOX
+#define	NFSERR_STALEWRITEVERF	30001	/* Fake return for nfs_commit() */
+
+#define NFSERR_RETVOID		0x20000000 /* Return void, not error */
+#define NFSERR_AUTHERR		0x40000000 /* Mark an authentication error */
+#define NFSERR_RETERR		0x80000000 /* Mark an error return for V3 */
+
+/* Sizes in bytes of various nfs rpc components */
+#define	NFSX_UNSIGNED	4
+
+/* specific to NFS Version 2 */
+#define	NFSX_V2FH	32
+#define	NFSX_V2FATTR	68
+#define	NFSX_V2SATTR	32
+#define	NFSX_V2COOKIE	4
+#define NFSX_V2STATFS	20
+
+/* specific to NFS Version 3 */
+#define NFSX_V3FH		(sizeof (fhandle_t)) /* size this server uses */
+#define	NFSX_V3FHMAX		64	/* max. allowed by protocol */
+#define NFSX_V3FATTR		84
+#define NFSX_V3SATTR		60	/* max. all fields filled in */
+#define NFSX_V3SRVSATTR		(sizeof (struct nfsv3_sattr))
+#define NFSX_V3POSTOPATTR	(NFSX_V3FATTR + NFSX_UNSIGNED)
+#define NFSX_V3WCCDATA		(NFSX_V3POSTOPATTR + 8 * NFSX_UNSIGNED)
+#define NFSX_V3COOKIEVERF 	8
+#define NFSX_V3WRITEVERF 	8
+#define NFSX_V3CREATEVERF	8
+#define NFSX_V3STATFS		52
+#define NFSX_V3FSINFO		48
+#define NFSX_V3PATHCONF		24
+
+/* variants for both versions */
+#define NFSX_FH(v3)		((v3) ? (NFSX_V3FHMAX + NFSX_UNSIGNED) : \
+					NFSX_V2FH)
+#define NFSX_SRVFH(v3)		((v3) ? NFSX_V3FH : NFSX_V2FH)
+#define	NFSX_FATTR(v3)		((v3) ? NFSX_V3FATTR : NFSX_V2FATTR)
+#define NFSX_PREOPATTR(v3)	((v3) ? (7 * NFSX_UNSIGNED) : 0)
+#define NFSX_POSTOPATTR(v3)	((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : 0)
+#define NFSX_POSTOPORFATTR(v3)	((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : \
+					NFSX_V2FATTR)
+#define NFSX_WCCDATA(v3)	((v3) ? NFSX_V3WCCDATA : 0)
+#define NFSX_WCCORFATTR(v3)	((v3) ? NFSX_V3WCCDATA : NFSX_V2FATTR)
+#define	NFSX_SATTR(v3)		((v3) ? NFSX_V3SATTR : NFSX_V2SATTR)
+#define	NFSX_COOKIEVERF(v3)	((v3) ? NFSX_V3COOKIEVERF : 0)
+#define	NFSX_WRITEVERF(v3)	((v3) ? NFSX_V3WRITEVERF : 0)
+#define NFSX_READDIR(v3)	((v3) ? (5 * NFSX_UNSIGNED) : \
+					(2 * NFSX_UNSIGNED))
+#define	NFSX_STATFS(v3)		((v3) ? NFSX_V3STATFS : NFSX_V2STATFS)
+
+/* nfs rpc procedure numbers (before version mapping) */
+#define	NFSPROC_NULL		0
+#define	NFSPROC_GETATTR		1
+#define	NFSPROC_SETATTR		2
+#define	NFSPROC_LOOKUP		3
+#define	NFSPROC_ACCESS		4
+#define	NFSPROC_READLINK	5
+#define	NFSPROC_READ		6
+#define	NFSPROC_WRITE		7
+#define	NFSPROC_CREATE		8
+#define	NFSPROC_MKDIR		9
+#define	NFSPROC_SYMLINK		10
+#define	NFSPROC_MKNOD		11
+#define	NFSPROC_REMOVE		12
+#define	NFSPROC_RMDIR		13
+#define	NFSPROC_RENAME		14
+#define	NFSPROC_LINK		15
+#define	NFSPROC_READDIR		16
+#define	NFSPROC_READDIRPLUS	17
+#define	NFSPROC_FSSTAT		18
+#define	NFSPROC_FSINFO		19
+#define	NFSPROC_PATHCONF	20
+#define	NFSPROC_COMMIT		21
+
+/* And leasing (nqnfs) procedure numbers (must be last) */
+#define	NQNFSPROC_GETLEASE	22
+#define	NQNFSPROC_VACATED	23
+#define	NQNFSPROC_EVICTED	24
+
+#define NFSPROC_NOOP		25
+#define	NFS_NPROCS		26
+
+/* Actual Version 2 procedure numbers */
+#define	NFSV2PROC_NULL		0
+#define	NFSV2PROC_GETATTR	1
+#define	NFSV2PROC_SETATTR	2
+#define	NFSV2PROC_NOOP		3
+#define	NFSV2PROC_ROOT		NFSV2PROC_NOOP	/* Obsolete */
+#define	NFSV2PROC_LOOKUP	4
+#define	NFSV2PROC_READLINK	5
+#define	NFSV2PROC_READ		6
+#define	NFSV2PROC_WRITECACHE	NFSV2PROC_NOOP	/* Obsolete */
+#define	NFSV2PROC_WRITE		8
+#define	NFSV2PROC_CREATE	9
+#define	NFSV2PROC_REMOVE	10
+#define	NFSV2PROC_RENAME	11
+#define	NFSV2PROC_LINK		12
+#define	NFSV2PROC_SYMLINK	13
+#define	NFSV2PROC_MKDIR		14
+#define	NFSV2PROC_RMDIR		15
+#define	NFSV2PROC_READDIR	16
+#define	NFSV2PROC_STATFS	17
+
+/*
+ * Constants used by the Version 3 protocol for various RPCs
+ */
+#define NFSV3SATTRTIME_DONTCHANGE	0
+#define NFSV3SATTRTIME_TOSERVER		1
+#define NFSV3SATTRTIME_TOCLIENT		2
+
+#define NFSV3ACCESS_READ		0x01
+#define NFSV3ACCESS_LOOKUP		0x02
+#define NFSV3ACCESS_MODIFY		0x04
+#define NFSV3ACCESS_EXTEND		0x08
+#define NFSV3ACCESS_DELETE		0x10
+#define NFSV3ACCESS_EXECUTE		0x20
+
+#define NFSV3WRITE_UNSTABLE		0
+#define NFSV3WRITE_DATASYNC		1
+#define NFSV3WRITE_FILESYNC		2
+
+#define NFSV3CREATE_UNCHECKED		0
+#define NFSV3CREATE_GUARDED		1
+#define NFSV3CREATE_EXCLUSIVE		2
+
+#define NFSV3FSINFO_LINK		0x01
+#define NFSV3FSINFO_SYMLINK		0x02
+#define NFSV3FSINFO_HOMOGENEOUS		0x08
+#define NFSV3FSINFO_CANSETTIME		0x10
+
+/* Conversion macros */
+#define	vtonfsv2_mode(t,m) \
+		txdr_unsigned(((t) == VFIFO) ? MAKEIMODE(VCHR, (m)) : \
+				MAKEIMODE((t), (m)))
+#define vtonfsv3_mode(m)	txdr_unsigned((m) & 07777)
+#define	nfstov_mode(a)		(fxdr_unsigned(u_short, (a))&07777)
+#define	vtonfsv2_type(a)	txdr_unsigned(nfsv2_type[((long)(a))])
+#define	vtonfsv3_type(a)	txdr_unsigned(nfsv3_type[((long)(a))])
+#define	nfsv2tov_type(a)	nv2tov_type[fxdr_unsigned(u_long,(a))&0x7]
+#define	nfsv3tov_type(a)	nv3tov_type[fxdr_unsigned(u_long,(a))&0x7]
+
+/* File types */
+typedef enum { NFNON=0, NFREG=1, NFDIR=2, NFBLK=3, NFCHR=4, NFLNK=5,
+	NFSOCK=6, NFFIFO=7 } nfstype;
+
+/* Structs for common parts of the rpc's */
+/*
+ * File Handle (32 bytes for version 2), variable up to 64 for version 3.
+ * File Handles of up to NFS_SMALLFH in size are stored directly in the
+ * nfs node, whereas larger ones are malloc'd. (This never happens when
+ * NFS_SMALLFH is set to 64.)
+ * NFS_SMALLFH should be in the range of 32 to 64 and be divisible by 4.
+ */
+#ifndef NFS_SMALLFH
+#define NFS_SMALLFH	64
+#endif
+union nfsfh {
+	fhandle_t	fh_generic;
+	u_char		fh_bytes[NFS_SMALLFH];
+};
+typedef union nfsfh nfsfh_t;
+
+struct nfsv2_time {
+	u_long	nfsv2_sec;
+	u_long	nfsv2_usec;
+};
+typedef struct nfsv2_time	nfstime2;
+
+struct nfsv3_time {
+	u_long	nfsv3_sec;
+	u_long	nfsv3_nsec;
+};
+typedef struct nfsv3_time	nfstime3;
+
+/*
+ * Quads are defined as arrays of 2 longs to ensure dense packing for the
+ * protocol and to facilitate xdr conversion.
+ */
+struct nfs_uquad {
+	u_long	nfsuquad[2];
+};
+typedef	struct nfs_uquad	nfsuint64;
+
+/*
+ * Used to convert between two u_longs and a u_quad_t.
+ */
+union nfs_quadconvert {
+	u_long		lval[2];
+	u_quad_t	qval;
+};
+typedef union nfs_quadconvert	nfsquad_t;
+
+/*
+ * NFS Version 3 special file number.
+ */
+struct nfsv3_spec {
+	u_long	specdata1;
+	u_long	specdata2;
+};
+typedef	struct nfsv3_spec	nfsv3spec;
+
+/*
+ * File attributes and setable attributes. These structures cover both
+ * NFS version 2 and the version 3 protocol. Note that the union is only
+ * used so that one pointer can refer to both variants. These structures
+ * go out on the wire and must be densely packed, so no quad data types
+ * are used. (all fields are longs or u_longs or structures of same)
+ * NB: You can't do sizeof(struct nfs_fattr), you must use the
+ *     NFSX_FATTR(v3) macro.
+ */
+struct nfs_fattr {
+	u_long	fa_type;
+	u_long	fa_mode;
+	u_long	fa_nlink;
+	u_long	fa_uid;
+	u_long	fa_gid;
+	union {
+		struct {
+			u_long		nfsv2fa_size;
+			u_long		nfsv2fa_blocksize;
+			u_long		nfsv2fa_rdev;
+			u_long		nfsv2fa_blocks;
+			u_long		nfsv2fa_fsid;
+			u_long		nfsv2fa_fileid;
+			nfstime2	nfsv2fa_atime;
+			nfstime2	nfsv2fa_mtime;
+			nfstime2	nfsv2fa_ctime;
+		} fa_nfsv2;
+		struct {
+			nfsuint64	nfsv3fa_size;
+			nfsuint64	nfsv3fa_used;
+			nfsv3spec	nfsv3fa_rdev;
+			nfsuint64	nfsv3fa_fsid;
+			nfsuint64	nfsv3fa_fileid;
+			nfstime3	nfsv3fa_atime;
+			nfstime3	nfsv3fa_mtime;
+			nfstime3	nfsv3fa_ctime;
+		} fa_nfsv3;
+	} fa_un;
+};
+
+/* and some ugly defines for accessing union components */
+#define	fa2_size		fa_un.fa_nfsv2.nfsv2fa_size
+#define	fa2_blocksize		fa_un.fa_nfsv2.nfsv2fa_blocksize
+#define	fa2_rdev		fa_un.fa_nfsv2.nfsv2fa_rdev
+#define	fa2_blocks		fa_un.fa_nfsv2.nfsv2fa_blocks
+#define	fa2_fsid		fa_un.fa_nfsv2.nfsv2fa_fsid
+#define	fa2_fileid		fa_un.fa_nfsv2.nfsv2fa_fileid
+#define	fa2_atime		fa_un.fa_nfsv2.nfsv2fa_atime
+#define	fa2_mtime		fa_un.fa_nfsv2.nfsv2fa_mtime
+#define	fa2_ctime		fa_un.fa_nfsv2.nfsv2fa_ctime
+#define	fa3_size		fa_un.fa_nfsv3.nfsv3fa_size
+#define	fa3_used		fa_un.fa_nfsv3.nfsv3fa_used
+#define	fa3_rdev		fa_un.fa_nfsv3.nfsv3fa_rdev
+#define	fa3_fsid		fa_un.fa_nfsv3.nfsv3fa_fsid
+#define	fa3_fileid		fa_un.fa_nfsv3.nfsv3fa_fileid
+#define	fa3_atime		fa_un.fa_nfsv3.nfsv3fa_atime
+#define	fa3_mtime		fa_un.fa_nfsv3.nfsv3fa_mtime
+#define	fa3_ctime		fa_un.fa_nfsv3.nfsv3fa_ctime
+
+struct nfsv2_sattr {
+	u_long		sa_mode;
+	u_long		sa_uid;
+	u_long		sa_gid;
+	u_long		sa_size;
+	nfstime2	sa_atime;
+	nfstime2	sa_mtime;
+};
+
+/*
+ * NFS Version 3 sattr structure for the new node creation case.
+ */
+struct nfsv3_sattr {
+	u_long		sa_modetrue;
+	u_long		sa_mode;
+	u_long		sa_uidtrue;
+	u_long		sa_uid;
+	u_long		sa_gidtrue;
+	u_long		sa_gid;
+	u_long		sa_sizefalse;
+	u_long		sa_atimetype;
+	nfstime3	sa_atime;
+	u_long		sa_mtimetype;
+	nfstime3	sa_mtime;
+};
+
+struct nfs_statfs {
+	union {
+		struct {
+			u_long		nfsv2sf_tsize;
+			u_long		nfsv2sf_bsize;
+			u_long		nfsv2sf_blocks;
+			u_long		nfsv2sf_bfree;
+			u_long		nfsv2sf_bavail;
+		} sf_nfsv2;
+		struct {
+			nfsuint64	nfsv3sf_tbytes;
+			nfsuint64	nfsv3sf_fbytes;
+			nfsuint64	nfsv3sf_abytes;
+			nfsuint64	nfsv3sf_tfiles;
+			nfsuint64	nfsv3sf_ffiles;
+			nfsuint64	nfsv3sf_afiles;
+			u_long		nfsv3sf_invarsec;
+		} sf_nfsv3;
+	} sf_un;
+};
+
+#define sf_tsize	sf_un.sf_nfsv2.nfsv2sf_tsize
+#define sf_bsize	sf_un.sf_nfsv2.nfsv2sf_bsize
+#define sf_blocks	sf_un.sf_nfsv2.nfsv2sf_blocks
+#define sf_bfree	sf_un.sf_nfsv2.nfsv2sf_bfree
+#define sf_bavail	sf_un.sf_nfsv2.nfsv2sf_bavail
+#define sf_tbytes	sf_un.sf_nfsv3.nfsv3sf_tbytes
+#define sf_fbytes	sf_un.sf_nfsv3.nfsv3sf_fbytes
+#define sf_abytes	sf_un.sf_nfsv3.nfsv3sf_abytes
+#define sf_tfiles	sf_un.sf_nfsv3.nfsv3sf_tfiles
+#define sf_ffiles	sf_un.sf_nfsv3.nfsv3sf_ffiles
+#define sf_afiles	sf_un.sf_nfsv3.nfsv3sf_afiles
+#define sf_invarsec	sf_un.sf_nfsv3.nfsv3sf_invarsec
+
+struct nfsv3_fsinfo {
+	u_long		fs_rtmax;
+	u_long		fs_rtpref;
+	u_long		fs_rtmult;
+	u_long		fs_wtmax;
+	u_long		fs_wtpref;
+	u_long		fs_wtmult;
+	u_long		fs_dtpref;
+	nfsuint64	fs_maxfilesize;
+	nfstime3	fs_timedelta;
+	u_long		fs_properties;
+};
+
+struct nfsv3_pathconf {
+	u_long		pc_linkmax;
+	u_long		pc_namemax;
+	u_long		pc_notrunc;
+	u_long		pc_chownrestricted;
+	u_long		pc_caseinsensitive;
+	u_long		pc_casepreserving;
+};
+
+#endif
diff --git a/sys/nfs/nfsrtt.h b/sys/nfs/nfsrtt.h
new file mode 100644
index 000000000000..da15ece23337
--- /dev/null
+++ b/sys/nfs/nfsrtt.h
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 1992, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsrtt.h	8.2 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NFSRTT_H_
+#define _NFS_NFSRTT_H_
+
+/*
+ * Definitions for performance monitor.
+ * The client and server logging are turned on by setting the global
+ * constant "nfsrtton" to 1.
+ */
+#define	NFSRTTLOGSIZ	128
+
+/*
+ * Circular log of client side rpc activity. Each log entry is for one
+ * rpc filled in upon completion. (ie. in order of completion)
+ * The "pos" is the table index for the "next" entry, therefore the
+ * list goes from nfsrtt.rttl[pos] --> nfsrtt.rttl[pos - 1] in
+ * chronological order of completion.
+ */
+struct nfsrtt {
+	int pos;			/* Position in array for next entry */
+	struct rttl {
+		int	proc;		/* NFS procedure number */
+		int	rtt;		/* Measured round trip time */
+		int	rto;		/* Round Trip Timeout */
+		int	sent;		/* # rpcs in progress */
+		int	cwnd;		/* Send window */
+		int	srtt;		/* Ave Round Trip Time */
+		int	sdrtt;		/* Ave mean deviation of RTT */
+		fsid_t	fsid;		/* Fsid for mount point */
+		struct timeval tstamp;	/* Timestamp of log entry */
+	} rttl[NFSRTTLOGSIZ];
+};
+
+/*
+ * And definitions for server side performance monitor.
+ * The log organization is the same as above except it is filled in at the
+ * time the server sends the rpc reply.
+ */
+
+/*
+ * Bits for the flags field.
+ */
+#define	DRT_NQNFS	0x01	/* Rpc used Nqnfs protocol */
+#define	DRT_TCP		0x02	/* Client used TCP transport */
+#define	DRT_CACHEREPLY	0x04	/* Reply was from recent request cache */
+#define	DRT_CACHEDROP	0x08	/* Rpc request dropped, due to recent reply */
+#define DRT_NFSV3	0x10	/* Rpc used NFS Version 3 */
+
+/*
+ * Server log structure
+ * NB: ipadr == INADDR_ANY indicates a client using a non IP protocol.
+ *	(ISO perhaps?)
+ */
+struct nfsdrt {
+	int pos;			/* Position of next log entry */
+	struct drt {
+		int	flag;		/* Bits as defined above */
+		int	proc;		/* NFS procedure number */
+		u_long	ipadr;		/* IP address of client */
+		int	resptime;	/* Response time (usec) */
+		struct timeval tstamp;	/* Timestamp of log entry */
+	} drt[NFSRTTLOGSIZ];
+};
+
+#endif
diff --git a/sys/nfs/nfsrvcache.h b/sys/nfs/nfsrvcache.h
new file mode 100644
index 000000000000..15f0c1bcf21a
--- /dev/null
+++ b/sys/nfs/nfsrvcache.h
@@ -0,0 +1,88 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nfsrvcache.h	8.3 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NFSRVCACHE_H_
+#define _NFS_NFSRVCACHE_H_
+
+/*
+ * Definitions for the server recent request cache
+ */
+
+#define	NFSRVCACHESIZ	64
+
+struct nfsrvcache {
+	TAILQ_ENTRY(nfsrvcache) rc_lru;		/* LRU chain */
+	LIST_ENTRY(nfsrvcache) rc_hash;		/* Hash chain */
+	u_long	rc_xid;				/* rpc id number */
+	union {
+		struct mbuf *ru_repmb;		/* Reply mbuf list OR */
+		int ru_repstat;			/* Reply status */
+	} rc_un;
+	union nethostaddr rc_haddr;		/* Host address */
+	short	rc_proc;			/* rpc proc number */
+	u_char	rc_state;		/* Current state of request */
+	u_char	rc_flag;		/* Flag bits */
+};
+
+#define	rc_reply	rc_un.ru_repmb
+#define	rc_status	rc_un.ru_repstat
+#define	rc_inetaddr	rc_haddr.had_inetaddr
+#define	rc_nam		rc_haddr.had_nam
+
+/* Cache entry states */
+#define	RC_UNUSED	0
+#define	RC_INPROG	1
+#define	RC_DONE		2
+
+/* Return values */
+#define	RC_DROPIT	0
+#define	RC_REPLY	1
+#define	RC_DOIT		2
+#define	RC_CHECKIT	3
+
+/* Flag bits */
+#define	RC_LOCKED	0x01
+#define	RC_WANTED	0x02
+#define	RC_REPSTATUS	0x04
+#define	RC_REPMBUF	0x08
+#define	RC_NQNFS	0x10
+#define	RC_INETADDR	0x20
+#define	RC_NAM		0x40
+
+#endif
diff --git a/sys/nfs/nqnfs.h b/sys/nfs/nqnfs.h
new file mode 100644
index 000000000000..254d5c4c9215
--- /dev/null
+++ b/sys/nfs/nqnfs.h
@@ -0,0 +1,213 @@
+/*
+ * Copyright (c) 1992, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)nqnfs.h	8.3 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_NQNFS_H_
+#define _NFS_NQNFS_H_
+
+/*
+ * Definitions for NQNFS (Not Quite NFS) cache consistency protocol.
+ */
+
+/* Tunable constants */
+#define	NQ_CLOCKSKEW	3	/* Clock skew factor (sec) */
+#define	NQ_WRITESLACK	5	/* Delay for write cache flushing */
+#define	NQ_MAXLEASE	60	/* Max lease duration (sec) */
+#define	NQ_MINLEASE	5	/* Min lease duration (sec) */
+#define	NQ_DEFLEASE	30	/* Default lease duration (sec) */
+#define	NQ_RENEWAL	3	/* Time before expiry (sec) to renew */
+#define	NQ_TRYLATERDEL	15	/* Initial try later delay (sec) */
+#define	NQ_MAXNUMLEASE	2048	/* Upper bound on number of server leases */
+#define	NQ_DEADTHRESH	NQ_NEVERDEAD	/* Default nm_deadthresh */
+#define	NQ_NEVERDEAD	9	/* Greater than max. nm_timeouts */
+#define	NQLCHSZ		256	/* Server hash table size */
+
+#define	NQNFS_PROG	300105	/* As assigned by Sun */
+#define	NQNFS_VER3	3
+#define	NQNFS_EVICTSIZ	156	/* Size of eviction request in bytes */
+
+/*
+ * Definitions used for saving the "last lease expires" time in Non-volatile
+ * RAM on the server. The default definitions below assume that NOVRAM is not
+ * available.
+ */
+#ifdef HASNVRAM
+#  undef HASNVRAM
+#endif
+#define	NQSTORENOVRAM(t)
+#define	NQLOADNOVRAM(t)
+
+/*
+ * Defn and structs used on the server to maintain state for current leases.
+ * The list of host(s) that hold the lease are kept as nqhost structures.
+ * The first one lives in nqlease and any others are held in a linked
+ * list of nqm structures hanging off of nqlease.
+ *
+ * Each nqlease structure is chained into two lists. The first is a list
+ * ordered by increasing expiry time for nqsrv_timer() and the second is a chain
+ * hashed on lc_fh.
+ */
+#define	LC_MOREHOSTSIZ	10
+
+struct nqhost {
+	union {
+		struct {
+			u_short udp_flag;
+			u_short	udp_port;
+			union nethostaddr udp_haddr;
+		} un_udp;
+		struct {
+			u_short connless_flag;
+			u_short connless_spare;
+			union nethostaddr connless_haddr;
+		} un_connless;
+		struct {
+			u_short conn_flag;
+			u_short conn_spare;
+			struct nfssvc_sock *conn_slp;
+		} un_conn;
+	} lph_un;
+};
+#define	lph_flag	lph_un.un_udp.udp_flag
+#define	lph_port	lph_un.un_udp.udp_port
+#define	lph_haddr	lph_un.un_udp.udp_haddr
+#define	lph_inetaddr	lph_un.un_udp.udp_haddr.had_inetaddr
+#define	lph_claddr	lph_un.un_connless.connless_haddr
+#define	lph_nam		lph_un.un_connless.connless_haddr.had_nam
+#define	lph_slp		lph_un.un_conn.conn_slp
+
+struct nqlease {
+	LIST_ENTRY(nqlease) lc_hash;	/* Fhandle hash list */
+	CIRCLEQ_ENTRY(nqlease) lc_timer; /* Timer queue list */
+	time_t		lc_expiry;	/* Expiry time (sec) */
+	struct nqhost	lc_host;	/* Host that got lease */
+	struct nqm	*lc_morehosts;	/* Other hosts that share read lease */
+	fsid_t		lc_fsid;	/* Fhandle */
+	char		lc_fiddata[MAXFIDSZ];
+	struct vnode	*lc_vp;		/* Soft reference to associated vnode */
+};
+#define	lc_flag		lc_host.lph_un.un_udp.udp_flag
+
+/* lc_flag bits */
+#define	LC_VALID	0x0001	/* Host address valid */
+#define	LC_WRITE	0x0002	/* Write cache */
+#define	LC_NONCACHABLE	0x0004	/* Non-cachable lease */
+#define	LC_LOCKED	0x0008	/* Locked */
+#define	LC_WANTED	0x0010	/* Lock wanted */
+#define	LC_EXPIREDWANTED 0x0020	/* Want lease when expired */
+#define	LC_UDP		0x0040	/* Host address for udp socket */
+#define	LC_CLTP		0x0080	/* Host address for other connectionless */
+#define	LC_LOCAL	0x0100	/* Host is server */
+#define	LC_VACATED	0x0200	/* Host has vacated lease */
+#define	LC_WRITTEN	0x0400	/* Recently wrote to the leased file */
+#define	LC_SREF		0x0800	/* Holds a nfssvc_sock reference */
+
+struct nqm {
+	struct nqm	*lpm_next;
+	struct nqhost	lpm_hosts[LC_MOREHOSTSIZ];
+};
+
+/*
+ * Special value for slp for local server calls.
+ */
+#define	NQLOCALSLP	((struct nfssvc_sock *) -1)
+
+/*
+ * Server side macros.
+ */
+#define	nqsrv_getl(v, l) \
+		(void) nqsrv_getlease((v), &nfsd->nd_duration, \
+		 ((nfsd->nd_flag & ND_LEASE) ? (nfsd->nd_flag & ND_LEASE) : \
+		 ((l) | ND_CHECK)), \
+		 slp, procp, nfsd->nd_nam, &cache, &frev, cred)
+
+/*
+ * Client side macros that check for a valid lease.
+ */
+#define	NQNFS_CKINVALID(v, n, f) \
+ ((time.tv_sec > (n)->n_expiry && \
+ VFSTONFS((v)->v_mount)->nm_timeouts < VFSTONFS((v)->v_mount)->nm_deadthresh) \
+  || ((f) == ND_WRITE && ((n)->n_flag & NQNFSWRITE) == 0))
+
+#define	NQNFS_CKCACHABLE(v, f) \
+ ((time.tv_sec <= VTONFS(v)->n_expiry || \
+  VFSTONFS((v)->v_mount)->nm_timeouts >= VFSTONFS((v)->v_mount)->nm_deadthresh) \
+   && (VTONFS(v)->n_flag & NQNFSNONCACHE) == 0 && \
+   ((f) == ND_READ || (VTONFS(v)->n_flag & NQNFSWRITE)))
+
+#define	NQNFS_NEEDLEASE(v, p) \
+		(time.tv_sec > VTONFS(v)->n_expiry ? \
+		 ((VTONFS(v)->n_flag & NQNFSEVICTED) ? 0 : nqnfs_piggy[p]) : \
+		 (((time.tv_sec + NQ_RENEWAL) > VTONFS(v)->n_expiry && \
+		   nqnfs_piggy[p]) ? \
+		   ((VTONFS(v)->n_flag & NQNFSWRITE) ? \
+		    ND_WRITE : nqnfs_piggy[p]) : 0))
+
+/*
+ * List head for timer queue.
+ */
+CIRCLEQ_HEAD(, nqlease) nqtimerhead;
+
+/*
+ * List head for the file handle hash table.
+ */
+#define	NQFHHASH(f) \
+	(&nqfhhashtbl[(*((u_long *)(f))) & nqfhhash])
+LIST_HEAD(nqfhhashhead, nqlease) *nqfhhashtbl;
+u_long nqfhhash;
+
+/*
+ * Nqnfs return status numbers.
+ */
+#define	NQNFS_EXPIRED	500
+#define	NQNFS_TRYLATER	501
+
+#ifdef KERNEL
+void	nqnfs_lease_updatetime __P((int));
+int	nqsrv_cmpnam __P((struct nfssvc_sock *,struct mbuf *,struct nqhost *));
+int	nqsrv_getlease __P((struct vnode *, u_long *, int,
+		struct nfssvc_sock *, struct proc *, struct mbuf *, int *,
+		u_quad_t *, struct ucred *));
+int	nqnfs_getlease __P((struct vnode *, int, struct ucred *,struct proc *));
+int	nqnfs_callback __P((struct nfsmount *, struct mbuf *, struct mbuf *,
+		caddr_t));
+int	nqnfs_clientd __P((struct nfsmount *, struct ucred *,
+		struct nfsd_cargs *, int, caddr_t, struct proc *));
+#endif
+
+#endif
diff --git a/sys/nfs/rpcv2.h b/sys/nfs/rpcv2.h
new file mode 100644
index 000000000000..bb6cd3daf76a
--- /dev/null
+++ b/sys/nfs/rpcv2.h
@@ -0,0 +1,141 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)rpcv2.h	8.2 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_RPCV2_H_
+#define _NFS_RPCV2_H_
+
+/*
+ * Definitions for Sun RPC Version 2, from
+ * "RPC: Remote Procedure Call Protocol Specification" RFC1057
+ */
+
+/* Version # */
+#define	RPC_VER2	2
+
+/* Authentication */
+#define	RPCAUTH_NULL	0
+#define	RPCAUTH_UNIX	1
+#define	RPCAUTH_SHORT	2
+#define RPCAUTH_KERB4	4
+#define	RPCAUTH_NQNFS	300000
+#define	RPCAUTH_MAXSIZ	400
+#define	RPCVERF_MAXSIZ	12	/* For Kerb, can actually be 400 */
+#define	RPCAUTH_UNIXGIDS 16
+
+/*
+ * Constants associated with authentication flavours.
+ */
+#define RPCAKN_FULLNAME	0
+#define RPCAKN_NICKNAME	1
+
+/* Rpc Constants */
+#define	RPC_CALL	0
+#define	RPC_REPLY	1
+#define	RPC_MSGACCEPTED	0
+#define	RPC_MSGDENIED	1
+#define	RPC_PROGUNAVAIL	1
+#define	RPC_PROGMISMATCH	2
+#define	RPC_PROCUNAVAIL	3
+#define	RPC_GARBAGE	4		/* I like this one */
+#define	RPC_MISMATCH	0
+#define	RPC_AUTHERR	1
+
+/* Authentication failures */
+#define	AUTH_BADCRED	1
+#define	AUTH_REJECTCRED	2
+#define	AUTH_BADVERF	3
+#define	AUTH_REJECTVERF	4
+#define	AUTH_TOOWEAK	5		/* Give em wheaties */
+
+/* Sizes of rpc header parts */
+#define	RPC_SIZ		24
+#define	RPC_REPLYSIZ	28
+
+/* RPC Prog definitions */
+#define	RPCPROG_MNT	100005
+#define	RPCMNT_VER1	1
+#define	RPCMNT_MOUNT	1
+#define	RPCMNT_DUMP	2
+#define	RPCMNT_UMOUNT	3
+#define	RPCMNT_UMNTALL	4
+#define	RPCMNT_EXPORT	5
+#define	RPCMNT_NAMELEN	255
+#define	RPCMNT_PATHLEN	1024
+#define	RPCPROG_NFS	100003
+
+/*
+ * Structures used for RPCAUTH_KERB4.
+ */
+struct nfsrpc_fullverf {
+	u_long		t1;
+	u_long		t2;
+	u_long		w2;
+};
+
+struct nfsrpc_fullblock {
+	u_long		t1;
+	u_long		t2;
+	u_long		w1;
+	u_long		w2;
+};
+
+struct nfsrpc_nickverf {
+	u_long			kind;
+	struct nfsrpc_fullverf	verf;
+};
+
+/*
+ * and their sizes in bytes.. If sizeof (struct nfsrpc_xx) != these
+ * constants, well then things will break in mount_nfs and nfsd.
+ */
+#define RPCX_FULLVERF	12
+#define RPCX_FULLBLOCK	16
+#define RPCX_NICKVERF	16
+
+#ifdef NFSKERB
+XXX
+#else
+typedef u_char			NFSKERBKEY_T[2];
+typedef u_char			NFSKERBKEYSCHED_T[2];
+#endif
+#define NFS_KERBSRV	"rcmd"		/* Kerberos Service for NFS */
+#define NFS_KERBTTL	(30 * 60)	/* Credential ttl (sec) */
+#define NFS_KERBCLOCKSKEW (5 * 60)	/* Clock skew (sec) */
+#define NFS_KERBW1(t)	(*((u_long *)(&((t).dat[((t).length + 3) & ~0x3]))))
+#endif
diff --git a/sys/nfs/xdr_subs.h b/sys/nfs/xdr_subs.h
new file mode 100644
index 000000000000..2c433a63aedc
--- /dev/null
+++ b/sys/nfs/xdr_subs.h
@@ -0,0 +1,90 @@
+/*
+ * Copyright (c) 1989, 1993
+ *	The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Rick Macklem at The University of Guelph.
+ *
+ * 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.
+ *
+ *	@(#)xdr_subs.h	8.3 (Berkeley) 3/30/95
+ */
+
+
+#ifndef _NFS_XDR_SUBS_H_
+#define _NFS_XDR_SUBS_H_
+
+/*
+ * Macros used for conversion to/from xdr representation by nfs...
+ * These use the MACHINE DEPENDENT routines ntohl, htonl
+ * As defined by "XDR: External Data Representation Standard" RFC1014
+ *
+ * To simplify the implementation, we use ntohl/htonl even on big-endian
+ * machines, and count on them being `#define'd away.  Some of these
+ * might be slightly more efficient as quad_t copies on a big-endian,
+ * but we cannot count on their alignment anyway.
+ */
+
+#define	fxdr_unsigned(t, v)	((t)ntohl((long)(v)))
+#define	txdr_unsigned(v)	(htonl((long)(v)))
+
+#define	fxdr_nfsv2time(f, t) { \
+	(t)->ts_sec = ntohl(((struct nfsv2_time *)(f))->nfsv2_sec); \
+	if (((struct nfsv2_time *)(f))->nfsv2_usec != 0xffffffff) \
+		(t)->ts_nsec = 1000 * ntohl(((struct nfsv2_time *)(f))->nfsv2_usec); \
+	else \
+		(t)->ts_nsec = 0; \
+}
+#define	txdr_nfsv2time(f, t) { \
+	((struct nfsv2_time *)(t))->nfsv2_sec = htonl((f)->ts_sec); \
+	if ((f)->ts_nsec != -1) \
+		((struct nfsv2_time *)(t))->nfsv2_usec = htonl((f)->ts_nsec / 1000); \
+	else \
+		((struct nfsv2_time *)(t))->nfsv2_usec = 0xffffffff; \
+}
+
+#define	fxdr_nfsv3time(f, t) { \
+	(t)->ts_sec = ntohl(((struct nfsv3_time *)(f))->nfsv3_sec); \
+	(t)->ts_nsec = ntohl(((struct nfsv3_time *)(f))->nfsv3_nsec); \
+}
+#define	txdr_nfsv3time(f, t) { \
+	((struct nfsv3_time *)(t))->nfsv3_sec = htonl((f)->ts_sec); \
+	((struct nfsv3_time *)(t))->nfsv3_nsec = htonl((f)->ts_nsec); \
+}
+
+#define	fxdr_hyper(f, t) { \
+	((long *)(t))[_QUAD_HIGHWORD] = ntohl(((long *)(f))[0]); \
+	((long *)(t))[_QUAD_LOWWORD] = ntohl(((long *)(f))[1]); \
+}
+#define	txdr_hyper(f, t) { \
+	((long *)(t))[0] = htonl(((long *)(f))[_QUAD_HIGHWORD]); \
+	((long *)(t))[1] = htonl(((long *)(f))[_QUAD_LOWWORD]); \
+}
+
+#endif