]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/nfs/nfs_syscalls.c
xnu-2422.1.72.tar.gz
[apple/xnu.git] / bsd / nfs / nfs_syscalls.c
index 910b82c0a751740dd5622522b7a45ff126d911d6..ceeb803ff5a5db0091b32cdfc6770ac19e42ce8f 100644 (file)
@@ -1,16 +1,19 @@
 /*
 /*
- * Copyright (c) 2000-2003 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2011 Apple Inc.  All rights reserved.
  *
  *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
- * compliance with the License. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
 /*
  */
 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
 /*
  *     @(#)nfs_syscalls.c      8.5 (Berkeley) 3/30/95
  * FreeBSD-Id: nfs_syscalls.c,v 1.32 1997/11/07 08:53:25 phk Exp $
  */
  *     @(#)nfs_syscalls.c      8.5 (Berkeley) 3/30/95
  * FreeBSD-Id: nfs_syscalls.c,v 1.32 1997/11/07 08:53:25 phk Exp $
  */
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ */
 
 #include <sys/param.h>
 #include <sys/systm.h>
 
 #include <sys/param.h>
 #include <sys/systm.h>
-/* XXX CSM 11/25/97 FreeBSD's generated syscall prototypes */
-#ifdef notyet
-#include <sys/sysproto.h>
-#endif
 #include <sys/kernel.h>
 #include <sys/kernel.h>
-#include <sys/file.h>
+#include <sys/file_internal.h>
 #include <sys/filedesc.h>
 #include <sys/stat.h>
 #include <sys/filedesc.h>
 #include <sys/stat.h>
-#include <sys/vnode.h>
-#include <sys/mount.h>
-#include <sys/proc.h>
+#include <sys/vnode_internal.h>
+#include <sys/mount_internal.h>
+#include <sys/proc_internal.h> /* for fdflags */
+#include <sys/kauth.h>
 #include <sys/sysctl.h>
 #include <sys/ubc.h>
 #include <sys/uio.h>
 #include <sys/malloc.h>
 #include <sys/sysctl.h>
 #include <sys/ubc.h>
 #include <sys/uio.h>
 #include <sys/malloc.h>
-#include <sys/mbuf.h>
+#include <sys/kpi_mbuf.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/domain.h>
 #include <sys/protosw.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/domain.h>
 #include <sys/protosw.h>
-#include <sys/namei.h>
 #include <sys/fcntl.h>
 #include <sys/lockf.h>
 #include <sys/syslog.h>
 #include <sys/user.h>
 #include <sys/fcntl.h>
 #include <sys/lockf.h>
 #include <sys/syslog.h>
 #include <sys/user.h>
-#include <machine/spl.h>
+#include <sys/sysproto.h>
+#include <sys/kpi_socket.h>
+#include <sys/fsevents.h>
+#include <libkern/OSAtomic.h>
+#include <kern/thread_call.h>
+#include <kern/task.h>
+
+#include <security/audit/audit.h>
 
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 
 #include <netinet/in.h>
 #include <netinet/tcp.h>
-#if 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/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/nfs_gss.h>
 #include <nfs/nfsmount.h>
 #include <nfs/nfsnode.h>
 #include <nfs/nfsmount.h>
 #include <nfs/nfsnode.h>
-#include <nfs/nqnfs.h>
-#include <nfs/nfsrtt.h>
 #include <nfs/nfs_lock.h>
 #include <nfs/nfs_lock.h>
+#if CONFIG_MACF
+#include <security/mac_framework.h>
+#endif
+
+kern_return_t  thread_terminate(thread_t); /* XXX */
+
+#if NFSSERVER
+
+extern int (*nfsrv_procs[NFS_NPROCS])(struct nfsrv_descript *nd,
+                                           struct nfsrv_sock *slp,
+                                           vfs_context_t ctx,
+                                           mbuf_t *mrepp);
+extern int nfsrv_wg_delay;
+extern int nfsrv_wg_delay_v3;
+
+static int nfsrv_require_resv_port = 0;
+static int nfsrv_deadsock_timer_on = 0;
 
 
-/* Global defs. */
-extern int (*nfsrv3_procs[NFS_NPROCS]) __P((struct nfsrv_descript *nd,
-                                           struct nfssvc_sock *slp,
-                                           struct proc *procp,
-                                           struct mbuf **mreqp));
-extern int nfs_numasync;
-extern int nfs_ioddelwri;
-extern time_t nqnfsstarttime;
-extern int nqsrv_writeslack;
-extern int nfsrtton;
-extern struct nfsstats nfsstats;
-extern int nfsrvw_procrastinate;
-extern int nfsrvw_procrastinate_v3;
-struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock;
-static int nuidhash_max = NFS_MAXUIDHASH;
-
-static void    nfsrv_zapsock __P((struct nfssvc_sock *slp));
-static int     nfssvc_iod __P((struct proc *));
-
-#define        TRUE    1
-#define        FALSE   0
-
-static int nfs_asyncdaemon[NFS_MAXASYNCDAEMON];
-
-#ifndef NFS_NOSERVER
-int nfsd_waiting = 0;
-static struct nfsdrt nfsdrt;
-static int nfs_numnfsd = 0;
-static int notstarted = 1;
-static int modify_flag = 0;
-static void    nfsd_rt __P((int sotype, struct nfsrv_descript *nd,
-                            int cacherep));
-static int     nfssvc_addsock __P((struct file *, struct mbuf *,
-                                   struct proc *));
-static int     nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *));
-
-static int nfs_privport = 0;
-/* XXX CSM 11/25/97 Upgrade sysctl.h someday */
-#ifdef notyet
-SYSCTL_INT(_vfs_nfs, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW, &nfs_privport, 0, "");
-SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay, CTLFLAG_RW, &nfsrvw_procrastinate, 0, "");
-SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay_v3, CTLFLAG_RW, &nfsrvw_procrastinate_v3, 0, "");
+int    nfssvc_export(user_addr_t argp);
+int    nfssvc_nfsd(void);
+int    nfssvc_addsock(socket_t, mbuf_t);
+void   nfsrv_zapsock(struct nfsrv_sock *);
+void   nfsrv_slpderef(struct nfsrv_sock *);
+void   nfsrv_slpfree(struct nfsrv_sock *);
+
+#endif /* NFSSERVER */
+
+/*
+ * sysctl stuff
+ */
+SYSCTL_DECL(_vfs_generic);
+SYSCTL_NODE(_vfs_generic, OID_AUTO, nfs, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "nfs hinge");
+
+#if NFSCLIENT
+SYSCTL_NODE(_vfs_generic_nfs, OID_AUTO, client, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "nfs client hinge");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, initialdowndelay, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_tprintf_initial_delay, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, nextdowndelay, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_tprintf_delay, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, iosize, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_iosize, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, access_cache_timeout, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_access_cache_timeout, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, allow_async, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_allow_async, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, statfs_rate_limit, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_statfs_rate_limit, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, nfsiod_thread_max, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsiod_thread_max, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, nfsiod_thread_count, CTLFLAG_RD | CTLFLAG_LOCKED, &nfsiod_thread_count, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, lockd_mounts, CTLFLAG_RD | CTLFLAG_LOCKED, &nfs_lockd_mounts, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, max_async_writes, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_max_async_writes, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, single_des, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_single_des, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, access_delete, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_access_delete, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, access_dotzfs, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_access_dotzfs, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, access_for_getattr, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_access_for_getattr, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, idmap_ctrl, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_idmap_ctrl, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, callback_port, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_callback_port, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, is_mobile, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_is_mobile, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_client, OID_AUTO, squishy_flags, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_squishy_flags, 0, "");
+SYSCTL_UINT(_vfs_generic_nfs_client, OID_AUTO, debug_ctl, CTLFLAG_RW | CTLFLAG_LOCKED, &nfs_debug_ctl, 0, "");
+
+
+#endif /* NFSCLIENT */
+
+#if NFSSERVER
+SYSCTL_NODE(_vfs_generic_nfs, OID_AUTO, server, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "nfs server hinge");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, wg_delay, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_wg_delay, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, wg_delay_v3, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_wg_delay_v3, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, require_resv_port, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_require_resv_port, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, async, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_async, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, export_hash_size, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_export_hash_size, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, reqcache_size, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_reqcache_size, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, request_queue_length, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_sock_max_rec_queue_length, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, user_stats, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_user_stat_enabled, 0, "");
+SYSCTL_UINT(_vfs_generic_nfs_server, OID_AUTO, gss_context_ttl, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_gss_context_ttl, 0, "");
+#if CONFIG_FSE
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, fsevents, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_fsevents_enabled, 0, "");
 #endif
 #endif
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, nfsd_thread_max, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsd_thread_max, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, nfsd_thread_count, CTLFLAG_RD | CTLFLAG_LOCKED, &nfsd_thread_count, 0, "");
+#ifdef NFS_UC_Q_DEBUG
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, use_upcall_svc, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_uc_use_proxy, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, upcall_queue_limit, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_uc_queue_limit, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, upcall_queue_max_seen, CTLFLAG_RW | CTLFLAG_LOCKED, &nfsrv_uc_queue_max_seen, 0, "");
+SYSCTL_INT(_vfs_generic_nfs_server, OID_AUTO, upcall_queue_count, CTLFLAG_RD | CTLFLAG_LOCKED, (int *)&nfsrv_uc_queue_count, 0, "");
+#endif
+#endif /* NFSSERVER */
+
+
+#if NFSCLIENT
+
+int
+nfsclnt(proc_t p, struct nfsclnt_args *uap, __unused int *retval)
+{
+       struct lockd_ans la;
+       int error;
+
+       switch (uap->flag) {
+       case NFSCLNT_LOCKDANS:
+               error = copyin(uap->argp, &la, sizeof(la));
+               if (!error)
+                       error = nfslockdans(p, &la);
+               break;
+       case NFSCLNT_LOCKDNOTIFY:
+               error = nfslockdnotify(p, uap->argp);
+               break;
+       default:
+               error = EINVAL;
+       }
+       return (error);
+}
+
+/*
+ * Asynchronous I/O threads for client NFS.
+ * They do read-ahead and write-behind operations on the block I/O cache.
+ *
+ * The pool of up to nfsiod_thread_max threads is launched on demand and exit
+ * when unused for a while.  There are as many nfsiod structs as there are
+ * nfsiod threads; however there's no strict tie between a thread and a struct.
+ * Each thread puts an nfsiod on the free list and sleeps on it.  When it wakes
+ * up, it removes the next struct nfsiod from the queue and services it.  Then
+ * it will put the struct at the head of free list and sleep on it.
+ * Async requests will pull the next struct nfsiod from the head of the free list,
+ * put it on the work queue, and wake whatever thread is waiting on that struct.
+ */
+
+/*
+ * nfsiod thread exit routine
+ *
+ * Must be called with nfsiod_mutex held so that the
+ * decision to terminate is atomic with the termination.
+ */
+void
+nfsiod_terminate(struct nfsiod *niod)
+{
+       nfsiod_thread_count--;
+       lck_mtx_unlock(nfsiod_mutex);
+       if (niod)
+               FREE(niod, M_TEMP);
+       else
+               printf("nfsiod: terminating without niod\n");
+       thread_terminate(current_thread());
+       /*NOTREACHED*/
+}
+
+/* nfsiod thread startup routine */
+void
+nfsiod_thread(void)
+{
+       struct nfsiod *niod;
+       int error;
+
+       MALLOC(niod, struct nfsiod *, sizeof(struct nfsiod), M_TEMP, M_WAITOK);
+       if (!niod) {
+               lck_mtx_lock(nfsiod_mutex);
+               nfsiod_thread_count--;
+               wakeup(current_thread());
+               lck_mtx_unlock(nfsiod_mutex);
+               thread_terminate(current_thread());
+               /*NOTREACHED*/
+       }
+       bzero(niod, sizeof(*niod));
+       lck_mtx_lock(nfsiod_mutex);
+       TAILQ_INSERT_HEAD(&nfsiodfree, niod, niod_link);
+       wakeup(current_thread());
+       error = msleep0(niod, nfsiod_mutex, PWAIT | PDROP, "nfsiod", NFS_ASYNCTHREADMAXIDLE*hz, nfsiod_continue);
+       /* shouldn't return... so we have an error */
+       /* remove an old nfsiod struct and terminate */
+       lck_mtx_lock(nfsiod_mutex);
+       if ((niod = TAILQ_LAST(&nfsiodfree, nfsiodlist)))
+               TAILQ_REMOVE(&nfsiodfree, niod, niod_link);
+       nfsiod_terminate(niod);
+       /*NOTREACHED*/
+}
+
+/*
+ * Start up another nfsiod thread.
+ * (unless we're already maxed out and there are nfsiods running)
+ */
+int
+nfsiod_start(void)
+{
+       thread_t thd = THREAD_NULL;
+
+       lck_mtx_lock(nfsiod_mutex);
+       if ((nfsiod_thread_count >= NFSIOD_MAX) && (nfsiod_thread_count > 0)) {
+               lck_mtx_unlock(nfsiod_mutex);
+               return (EBUSY);
+       }
+       nfsiod_thread_count++;
+       if (kernel_thread_start((thread_continue_t)nfsiod_thread, NULL, &thd) != KERN_SUCCESS) {
+               lck_mtx_unlock(nfsiod_mutex);
+               return (EBUSY);
+       }
+       /* wait for the thread to complete startup */
+       msleep(thd, nfsiod_mutex, PWAIT | PDROP, "nfsiodw", NULL);
+       thread_deallocate(thd);
+       return (0);
+}
+
+/*
+ * Continuation for Asynchronous I/O threads for NFS client.
+ *
+ * Grab an nfsiod struct to work on, do some work, then drop it
+ */
+int
+nfsiod_continue(int error)
+{
+       struct nfsiod *niod;
+       struct nfsmount *nmp;
+       struct nfsreq *req, *treq;
+       struct nfs_reqqhead iodq;
+       int morework;
+
+       lck_mtx_lock(nfsiod_mutex);
+       niod = TAILQ_FIRST(&nfsiodwork);
+       if (!niod) {
+               /* there's no work queued up */
+               /* remove an old nfsiod struct and terminate */
+               if ((niod = TAILQ_LAST(&nfsiodfree, nfsiodlist)))
+                       TAILQ_REMOVE(&nfsiodfree, niod, niod_link);
+               nfsiod_terminate(niod);
+               /*NOTREACHED*/
+       }
+       TAILQ_REMOVE(&nfsiodwork, niod, niod_link);
+
+worktodo:
+       while ((nmp = niod->niod_nmp)) {
+               /* 
+                * Service this mount's async I/O queue.
+                *
+                * In order to ensure some level of fairness between mounts,
+                * we grab all the work up front before processing it so any
+                * new work that arrives will be serviced on a subsequent
+                * iteration - and we have a chance to see if other work needs
+                * to be done (e.g. the delayed write queue needs to be pushed
+                * or other mounts are waiting for an nfsiod).
+                */
+               /* grab the current contents of the queue */
+               TAILQ_INIT(&iodq);
+               TAILQ_CONCAT(&iodq, &nmp->nm_iodq, r_achain);
+               lck_mtx_unlock(nfsiod_mutex);
+
+               /* process the queue */
+               TAILQ_FOREACH_SAFE(req, &iodq, r_achain, treq) {
+                       TAILQ_REMOVE(&iodq, req, r_achain);
+                       req->r_achain.tqe_next = NFSREQNOLIST;
+                       req->r_callback.rcb_func(req);
+               }
+
+               /* now check if there's more/other work to be done */
+               lck_mtx_lock(nfsiod_mutex);
+               morework = !TAILQ_EMPTY(&nmp->nm_iodq);
+               if (!morework || !TAILQ_EMPTY(&nfsiodmounts)) {
+                       /* we're going to stop working on this mount */
+                       if (morework) /* mount still needs more work so queue it up */
+                               TAILQ_INSERT_TAIL(&nfsiodmounts, nmp, nm_iodlink);
+                       nmp->nm_niod = NULL;
+                       niod->niod_nmp = NULL;
+               }
+       }
+
+       /* loop if there's still a mount to work on */
+       if (!niod->niod_nmp && !TAILQ_EMPTY(&nfsiodmounts)) {
+               niod->niod_nmp = TAILQ_FIRST(&nfsiodmounts);
+               TAILQ_REMOVE(&nfsiodmounts, niod->niod_nmp, nm_iodlink);
+       }
+       if (niod->niod_nmp)
+               goto worktodo;
+
+       /* queue ourselves back up - if there aren't too many threads running */
+       if (nfsiod_thread_count <= NFSIOD_MAX) {
+               TAILQ_INSERT_HEAD(&nfsiodfree, niod, niod_link);
+               error = msleep0(niod, nfsiod_mutex, PWAIT | PDROP, "nfsiod", NFS_ASYNCTHREADMAXIDLE*hz, nfsiod_continue);
+               /* shouldn't return... so we have an error */
+               /* remove an old nfsiod struct and terminate */
+               lck_mtx_lock(nfsiod_mutex);
+               if ((niod = TAILQ_LAST(&nfsiodfree, nfsiodlist)))
+                       TAILQ_REMOVE(&nfsiodfree, niod, niod_link);
+       }
+       nfsiod_terminate(niod);
+       /*NOTREACHED*/
+       return (0);
+}
+
+#endif /* NFSCLIENT */
+
+
+#if NFSSERVER
 
 /*
  * NFS server system calls
 
 /*
  * NFS server system calls
@@ -160,44 +407,98 @@ SYSCTL_INT(_vfs_nfs, OID_AUTO, gatherdelay_v3, CTLFLAG_RW, &nfsrvw_procrastinate
 /*
  * Get file handle system call
  */
 /*
  * Get file handle system call
  */
-#ifndef _SYS_SYSPROTO_H_
-struct getfh_args {
-       char    *fname;
-       fhandle_t *fhp;
-};
-#endif
 int
 int
-getfh(p, uap)
-       struct proc *p;
-       register struct getfh_args *uap;
+getfh(proc_t p, struct getfh_args *uap, __unused int *retval)
 {
 {
-       register struct vnode *vp;
-       fhandle_t fh;
-       int error;
+       vnode_t vp;
+       struct nfs_filehandle nfh;
+       int error, fhlen, fidlen;
        struct nameidata nd;
        struct nameidata nd;
+       char path[MAXPATHLEN], *ptr;
+       size_t pathlen;
+       struct nfs_exportfs *nxfs;
+       struct nfs_export *nx;
 
        /*
         * Must be super user
         */
 
        /*
         * Must be super user
         */
-       error = suser(p->p_ucred, &p->p_acflag);
-       if(error)
+       error = proc_suser(p);
+       if (error)
                return (error);
                return (error);
-       NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1, UIO_USERSPACE,                    uap->fname, p);
+
+       error = copyinstr(uap->fname, path, MAXPATHLEN, &pathlen);
+       if (!error)
+               error = copyin(uap->fhp, &fhlen, sizeof(fhlen));
+       if (error)
+               return (error);
+       /* limit fh size to length specified (or v3 size by default) */
+       if ((fhlen != NFSV2_MAX_FH_SIZE) && (fhlen != NFSV3_MAX_FH_SIZE))
+               fhlen = NFSV3_MAX_FH_SIZE;
+       fidlen = fhlen - sizeof(struct nfs_exphandle);
+
+       if (!nfsrv_is_initialized())
+               return (EINVAL);
+
+       NDINIT(&nd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1, 
+                       UIO_SYSSPACE, CAST_USER_ADDR_T(path), vfs_context_current());
        error = namei(&nd);
        if (error)
                return (error);
        error = namei(&nd);
        if (error)
                return (error);
+       nameidone(&nd);
+
        vp = nd.ni_vp;
        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);
+
+       // find exportfs that matches f_mntonname
+       lck_rw_lock_shared(&nfsrv_export_rwlock);
+       ptr = vnode_mount(vp)->mnt_vfsstat.f_mntonname;
+       LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
+               if (!strncmp(nxfs->nxfs_path, ptr, MAXPATHLEN))
+                       break;
+       }
+       if (!nxfs || strncmp(nxfs->nxfs_path, path, strlen(nxfs->nxfs_path))) {
+               error = EINVAL;
+               goto out;
+       }
+       // find export that best matches remainder of path
+       ptr = path + strlen(nxfs->nxfs_path);
+       while (*ptr && (*ptr == '/'))
+               ptr++;
+       LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
+               int len = strlen(nx->nx_path);
+               if (len == 0)  // we've hit the export entry for the root directory
+                       break;
+               if (!strncmp(nx->nx_path, ptr, len))
+                       break;
+       }
+       if (!nx) {
+               error = EINVAL;
+               goto out;
+       }
+
+       bzero(&nfh, sizeof(nfh));
+       nfh.nfh_xh.nxh_version = htonl(NFS_FH_VERSION);
+       nfh.nfh_xh.nxh_fsid = htonl(nxfs->nxfs_id);
+       nfh.nfh_xh.nxh_expid = htonl(nx->nx_id);
+       nfh.nfh_xh.nxh_flags = 0;
+       nfh.nfh_xh.nxh_reserved = 0;
+       nfh.nfh_len = fidlen;
+       error = VFS_VPTOFH(vp, (int*)&nfh.nfh_len, &nfh.nfh_fid[0], NULL);
+       if (nfh.nfh_len > (uint32_t)fidlen)
+               error = EOVERFLOW;
+       nfh.nfh_xh.nxh_fidlen = nfh.nfh_len;
+       nfh.nfh_len += sizeof(nfh.nfh_xh);
+       nfh.nfh_fhp = (u_char*)&nfh.nfh_xh;
+
+out:
+       lck_rw_done(&nfsrv_export_rwlock);
+       vnode_put(vp);
        if (error)
                return (error);
        if (error)
                return (error);
-       error = copyout((caddr_t)&fh, (caddr_t)uap->fhp, sizeof (fh));
+       error = copyout((caddr_t)&nfh, uap->fhp, sizeof(fhandle_t));
        return (error);
 }
 
        return (error);
 }
 
-#endif /* NFS_NOSERVER */
+extern const struct fileops vnops;
 
 /*
  * syscall for the rpc.lockd to use to translate a NFS file handle into
 
 /*
  * syscall for the rpc.lockd to use to translate a NFS file handle into
@@ -206,138 +507,111 @@ getfh(p, uap)
  * warning: do not remove the suser() call or this becomes one giant
  * security hole.
  */
  * warning: do not remove the suser() call or this becomes one giant
  * security hole.
  */
-#ifndef _SYS_SYSPROTO_H_
-struct fhopen_args {
-       const struct fhandle *u_fhp;
-       int flags;
-};
-#endif
 int
 int
-fhopen(p, uap, retval)
-       struct proc *p;
-       register struct fhopen_args *uap;
-       register_t *retval;
+fhopen( proc_t p,
+       struct fhopen_args *uap,
+       int32_t *retval)
 {
 {
-       struct mount *mp;
-       struct vnode *vp;
-       struct fhandle fhp;
-       struct vattr vat;
-       struct vattr *vap = &vat;
+       vnode_t vp;
+       struct nfs_filehandle nfh;
+       struct nfs_export *nx;
+       struct nfs_export_options *nxo;
        struct flock lf;
        struct flock lf;
-       struct file *fp;
-       register struct filedesc *fdp = p->p_fd;
-       int fmode, mode, error, type;
-       struct file *nfp;
+       struct fileproc *fp, *nfp;
+       int fmode, error, type;
        int indx;
        int indx;
-        struct ucred *credanon;
-        int exflags;
-       struct ucred *cred = p->p_ucred;
-       int didhold = 0;
-       extern struct fileops vnops;
+       vfs_context_t ctx = vfs_context_current();
+       kauth_action_t action;
 
        /*
         * Must be super user
         */
 
        /*
         * Must be super user
         */
-       error = suser(cred, &p->p_acflag);
-       if (error)
+       error = suser(vfs_context_ucred(ctx), 0);
+       if (error) {
                return (error);
                return (error);
+       }
+
+       if (!nfsrv_is_initialized()) {
+               return (EINVAL);
+       }
 
        fmode = FFLAGS(uap->flags);
        /* why not allow a non-read/write open for our lockd? */
        if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
                return (EINVAL);
 
        fmode = FFLAGS(uap->flags);
        /* why not allow a non-read/write open for our lockd? */
        if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
                return (EINVAL);
-       error = copyin((void*)uap->u_fhp, &fhp, sizeof(fhp));
+
+       error = copyin(uap->u_fhp, &nfh.nfh_len, sizeof(nfh.nfh_len));
        if (error)
                return (error);
        if (error)
                return (error);
-       /* find the mount point */
-       mp = vfs_getvfs(&fhp.fh_fsid);
-       if (mp == NULL)
-               return (ESTALE);      
-       /* now give me my vnode, it gets returned to me locked */
-/* XXX CSM need to split VFS_CHECKEXP out of VFS_FHTOVP? */
-       error = VFS_FHTOVP(mp, &fhp.fh_fid, NULL, &vp, &exflags, &credanon);
+       if ((nfh.nfh_len < (int)sizeof(struct nfs_exphandle)) ||
+           (nfh.nfh_len > (int)NFSV3_MAX_FH_SIZE))
+               return (EINVAL);
+       error = copyin(uap->u_fhp, &nfh, sizeof(nfh.nfh_len) + nfh.nfh_len);
        if (error)
                return (error);
        if (error)
                return (error);
+       nfh.nfh_fhp = (u_char*)&nfh.nfh_xh;
+
+       lck_rw_lock_shared(&nfsrv_export_rwlock);
+       /* now give me my vnode, it gets returned to me with a reference */
+       error = nfsrv_fhtovp(&nfh, NULL, &vp, &nx, &nxo);
+       lck_rw_done(&nfsrv_export_rwlock);
+       if (error) {
+               if (error == NFSERR_TRYLATER)
+                       error = EAGAIN; // XXX EBUSY? Or just leave as TRYLATER?
+               return (error);
+       }
+
        /*
        /*
-        * from now on we have to make sure not
-        * to forget about the vnode
-        * any error that causes an abort must vput(vp)
-        * just set error = err and 'goto bad;'.
+        * From now on we have to make sure not
+        * to forget about the vnode.
+        * Any error that causes an abort must vnode_put(vp).
+        * Just set error = err and 'goto bad;'.
         */
 
        /*
         * from vn_open  
         */      
         */
 
        /*
         * from vn_open  
         */      
-       if (vp->v_type == VSOCK) {
+       if (vnode_vtype(vp) == VSOCK) {
                error = EOPNOTSUPP;
                goto bad;      
        }
 
                error = EOPNOTSUPP;
                goto bad;      
        }
 
-       if (UBCINFOEXISTS(vp) && ((didhold = ubc_hold(vp)) == 0)) {
-               error = ENOENT;
+       /* disallow write operations on directories */
+       if (vnode_isdir(vp) && (fmode & (FWRITE | O_TRUNC))) {
+               error = EISDIR;
                goto bad;
        }
 
                goto bad;
        }
 
-       if (fmode & FREAD && fmode & (FWRITE | O_TRUNC)) {
-               int err = 0;
-               if (vp->v_type == VDIR)
-                       err = EISDIR;
-               else
-                       err = vn_writechk(vp);
-               if (err && !(error = VOP_ACCESS(vp, VREAD, cred, p)))
-                       error = err;
-               if (error || (error = VOP_ACCESS(vp, VREAD|VWRITE, cred, p)))
-                       goto bad;
-       } else if (fmode & FREAD) {
-               if ((error = VOP_ACCESS(vp, VREAD, cred, p)))
-                       goto bad;
-       } else if (fmode & (FWRITE | O_TRUNC)) {
-               if (vp->v_type == VDIR) {
-                       error = EISDIR;
-                       goto bad;
-               }
-               if ((error = vn_writechk(vp)) ||
-                   (error = VOP_ACCESS(vp, VWRITE, cred, p)))
-                       goto bad;
-       }
-       if (fmode & O_TRUNC) {
-               VOP_UNLOCK(vp, 0, p);                     /* XXX */
-               VOP_LEASE(vp, p, cred, LEASE_WRITE);
-               vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);       /* XXX */
-               VATTR_NULL(vap);
-               vap->va_size = 0;
-               error = VOP_SETATTR(vp, vap, cred, p);
-               if (error)
-                       goto bad;
-       }
+       /* compute action to be authorized */
+       action = 0;
+       if (fmode & FREAD)
+               action |= KAUTH_VNODE_READ_DATA;
+       if (fmode & (FWRITE | O_TRUNC))
+               action |= KAUTH_VNODE_WRITE_DATA;
+       if ((error = vnode_authorize(vp, NULL, action, ctx)) != 0)
+               goto bad;
 
 
-       error = VOP_OPEN(vp, fmode, cred, p);
-       if (error)
+       if ((error = VNOP_OPEN(vp, fmode, ctx)))
+               goto bad;
+       if ((error = vnode_ref_ext(vp, fmode, 0)))
                goto bad;
 
                goto bad;
 
-       if (fmode & FWRITE)
-               if (++vp->v_writecount <= 0)
-                       panic("fhopen: v_writecount");
        /*
         * end of vn_open code
         */
 
        /*
         * end of vn_open code
         */
 
-       if ((error = falloc(p, &nfp, &indx)) != 0) {
-               if (fmode & FWRITE)
-                       vp->v_writecount--;
+       // starting here... error paths should call vn_close/vnode_put
+       if ((error = falloc(p, &nfp, &indx, ctx)) != 0) {
+               vn_close(vp, fmode & FMASK, ctx);
                goto bad;
        }
        fp = nfp;
 
                goto bad;
        }
        fp = nfp;
 
-       /*
-        * Hold an extra reference to avoid having fp ripped out       
-        * from under us while we block in the lock op
-        */      
-       fref(fp);
-       nfp->f_data = (caddr_t)vp;
-       nfp->f_flag = fmode & FMASK;
-       nfp->f_ops = &vnops;     
-       nfp->f_type = DTYPE_VNODE;
+       fp->f_fglob->fg_flag = fmode & FMASK;
+       fp->f_fglob->fg_ops = &vnops;
+       fp->f_fglob->fg_data = (caddr_t)vp;
+
+       // XXX do we really need to support this with fhopen()?
        if (fmode & (O_EXLOCK | O_SHLOCK)) {
                lf.l_whence = SEEK_SET;
                lf.l_start = 0;
        if (fmode & (O_EXLOCK | O_SHLOCK)) {
                lf.l_whence = SEEK_SET;
                lf.l_start = 0;
@@ -349,460 +623,422 @@ fhopen(p, uap, retval)
                type = F_FLOCK;
                if ((fmode & FNONBLOCK) == 0)
                        type |= F_WAIT;
                type = F_FLOCK;
                if ((fmode & FNONBLOCK) == 0)
                        type |= F_WAIT;
-               VOP_UNLOCK(vp, 0, p);
-               if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
-                           type)) != 0) {
-                       (void) vn_close(vp, fp->f_flag, fp->f_cred, p);
-                       ffree(fp);
-                       fdrelse(p, indx);
-                       /*
-                        * release our private reference
-                        */
-                       frele(fp);
+               if ((error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, type, ctx, NULL))) {
+                       struct vfs_context context = *vfs_context_current();
+                       /* Modify local copy (to not damage thread copy) */
+                       context.vc_ucred = fp->f_fglob->fg_cred;
 
 
+                       vn_close(vp, fp->f_fglob->fg_flag, &context);
+                       fp_free(p, indx, fp);
                        return (error);
                }
                        return (error);
                }
-               vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
-               fp->f_flag |= FHASLOCK;
+               fp->f_fglob->fg_flag |= FHASLOCK;
        }
 
        }
 
-       VOP_UNLOCK(vp, 0, p);
-       *fdflags(p, indx) &= ~UF_RESERVED;
-       frele(fp);
+       vnode_put(vp);
+
+       proc_fdlock(p);
+       procfdtbl_releasefd(p, indx, NULL);
+       fp_drop(p, indx, fp, 1);
+       proc_fdunlock(p);
+
        *retval = indx;
        return (0);
 
 bad:
        *retval = indx;
        return (0);
 
 bad:
-       VOP_UNLOCK(vp, 0, p);
-       if (didhold)
-               ubc_rele(vp);
-       vrele(vp);
+       vnode_put(vp);
        return (error);
 }
 
 /*
        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
+ * NFS server pseudo system call
  */
  */
-#ifndef _SYS_SYSPROTO_H_
-struct nfssvc_args {
-       int flag;
-       caddr_t argp;
-};
-#endif
 int
 int
-nfssvc(p, uap)
-       struct proc *p;
-       register struct nfssvc_args *uap;
+nfssvc(proc_t p, struct nfssvc_args *uap, __unused int *retval)
 {
 {
-#ifndef NFS_NOSERVER
-       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;
-       struct timeval now;
-#endif /* NFS_NOSERVER */
+       mbuf_t nam;
+       struct user_nfsd_args user_nfsdarg;
+       socket_t so;
        int error;
 
        int error;
 
+       AUDIT_ARG(cmd, uap->flag);
+
        /*
        /*
-        * Must be super user
+        * Must be super user for most operations (export ops checked later).
         */
         */
-       error = suser(p->p_ucred, &p->p_acflag);
-       if(error)
+       if ((uap->flag != NFSSVC_EXPORT) && ((error = proc_suser(p))))
                return (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);
-#ifdef NFS_NOSERVER
-       else
-               error = ENXIO;
-#else /* !NFS_NOSERVER */
-       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 CONFIG_MACF
+       error = mac_system_check_nfsd(kauth_cred_get());
+       if (error)
+               return (error);
+#endif
+
+       /* make sure NFS server data structures have been initialized */
+       nfsrv_init();
 
 
-               if ((nmp->nm_state & NFSSTA_MNTD) &&
-                       (uap->flag & NFSSVC_GOTAUTH) == 0)
-                       return (0);
-               nmp->nm_state |= NFSSTA_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 (uap->flag & NFSSVC_ADDSOCK) {
+               if (IS_64BIT_PROCESS(p)) {
+                       error = copyin(uap->argp, (caddr_t)&user_nfsdarg, sizeof(user_nfsdarg));
+               } else {
+                       struct nfsd_args    tmp_args;
+                       error = copyin(uap->argp, (caddr_t)&tmp_args, sizeof(tmp_args));
+                       if (error == 0) {
+                               user_nfsdarg.sock = tmp_args.sock;
+                               user_nfsdarg.name = CAST_USER_ADDR_T(tmp_args.name);
+                               user_nfsdarg.namelen = tmp_args.namelen;
+                       }
+               }
                if (error)
                        return (error);
                if (error)
                        return (error);
-               error = getsock(p->p_fd, nfsdarg.sock, &fp);
+               /* get the socket */
+               error = file_socket(user_nfsdarg.sock, &so);
                if (error)
                        return (error);
                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)
+               /* Get the client address for connected sockets. */
+               if (user_nfsdarg.name == USER_ADDR_NULL || user_nfsdarg.namelen == 0) {
+                       nam = NULL;
+               } else {
+                       error = sockargs(&nam, user_nfsdarg.name, user_nfsdarg.namelen, MBUF_TYPE_SONAME);
+                       if (error) {
+                               /* drop the iocount file_socket() grabbed on the file descriptor */
+                               file_drop(user_nfsdarg.sock);
                                return (error);
                                return (error);
-               }
-               error = nfssvc_addsock(fp, nam, p);
-       } 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_ZONE(sizeof (struct nfsuid),
-                                                       M_NFSUID, M_WAITOK);
-                           } else
-                               nuidp = (struct nfsuid *)0;
-                           if ((slp->ns_flag & SLP_VALID) == 0) {
-                               if (nuidp)
-                                   FREE_ZONE((caddr_t)nuidp,
-                                       sizeof (struct nfsuid), 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;
-                               microtime(&now);
-                               nuidp->nu_expire = now.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);
+               /*
+                * nfssvc_addsock() will grab a retain count on the socket
+                * to keep the socket from being closed when nfsd closes its
+                * file descriptor for it.
+                */
+               error = nfssvc_addsock(so, nam);
+               /* drop the iocount file_socket() grabbed on the file descriptor */
+               file_drop(user_nfsdarg.sock);
+       } else if (uap->flag & NFSSVC_NFSD) {
+               error = nfssvc_nfsd();
+       } else if (uap->flag & NFSSVC_EXPORT) {
+               error = nfssvc_export(uap->argp);
+       } else {
+               error = EINVAL;
        }
        }
-#endif /* NFS_NOSERVER */
        if (error == EINTR || error == ERESTART)
                error = 0;
        return (error);
 }
 
        if (error == EINTR || error == ERESTART)
                error = 0;
        return (error);
 }
 
-#ifndef NFS_NOSERVER
 /*
  * Adds a socket to the list for servicing by nfsds.
  */
 /*
  * Adds a socket to the list for servicing by nfsds.
  */
-static int
-nfssvc_addsock(fp, mynam, p)
-       struct file *fp;
-       struct mbuf *mynam;
-       struct proc *p;
+int
+nfssvc_addsock(socket_t so, mbuf_t 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.
-        */
-       thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-       if (so->so_proto->pr_protocol == IPPROTO_UDP) {
-               tslp = nfs_udpsock;
-               if (tslp->ns_flag & SLP_VALID) {
-                       m_freem(mynam);
-                       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-                       return (EPERM);
-               }
-#if ISO
-       } else if (so->so_proto->pr_protocol == ISOPROTO_CLTP) {
-               tslp = nfs_cltpsock;
-               if (tslp->ns_flag & SLP_VALID) {
-                       m_freem(mynam);
-                       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-                       return (EPERM);
-               }
-#endif /* ISO */
+       struct nfsrv_sock *slp;
+       int error = 0, sodomain, sotype, soprotocol, on = 1;
+       int first;
+       struct timeval timeo;
+
+       /* make sure mbuf constants are set up */
+       if (!nfs_mbuf_mhlen)
+               nfs_mbuf_init();
+
+       sock_gettype(so, &sodomain, &sotype, &soprotocol);
+
+       /* There should be only one UDP socket for each of IPv4 and IPv6 */
+       if ((sodomain == AF_INET) && (soprotocol == IPPROTO_UDP) && nfsrv_udpsock) {
+               mbuf_freem(mynam);
+               return (EEXIST);
        }
        }
-       /* reserve buffer space for 2 maximally-sized packets */
-       siz = NFS_MAXPACKET;
-       if (so->so_type == SOCK_STREAM)
-               siz += sizeof (u_long);
-       siz *= 2;
-       if (siz > NFS_MAXSOCKBUF)
-               siz = NFS_MAXSOCKBUF;
-       error = soreserve(so, siz, siz);
-       if (error) {
-               m_freem(mynam);
-               thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-               return (error);
+       if ((sodomain == AF_INET6) && (soprotocol == IPPROTO_UDP) && nfsrv_udp6sock) {
+               mbuf_freem(mynam);
+               return (EEXIST);
        }
 
        }
 
+       /* Set protocol options and reserve some space (for UDP). */
+       if (sotype == SOCK_STREAM)
+               sock_setsockopt(so, SOL_SOCKET, SO_KEEPALIVE, &on, sizeof(on));
+       if ((sodomain == AF_INET) && (soprotocol == IPPROTO_TCP))
+               sock_setsockopt(so, IPPROTO_TCP, TCP_NODELAY, &on, sizeof(on));
+       if (sotype == SOCK_DGRAM) { /* set socket buffer sizes for UDP */
+               int reserve = NFS_UDPSOCKBUF;
+               error |= sock_setsockopt(so, SOL_SOCKET, SO_SNDBUF, &reserve, sizeof(reserve));
+               error |= sock_setsockopt(so, SOL_SOCKET, SO_RCVBUF, &reserve, sizeof(reserve));
+               if (error) {
+                       log(LOG_INFO, "nfssvc_addsock: UDP socket buffer setting error(s) %d\n", error);
+                       error = 0;
+               }
+       }
+       sock_nointerrupt(so, 0);
+
        /*
        /*
-        * 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.
+        * Set socket send/receive timeouts.
+        * Receive timeout shouldn't matter, but setting the send timeout
+        * will make sure that an unresponsive client can't hang the server.
         */
         */
-       if (so->so_type == SOCK_STREAM) {
-               struct sockopt sopt;
-               int val;
-
-               bzero(&sopt, sizeof sopt);
-               sopt.sopt_dir = SOPT_SET;
-               sopt.sopt_level = SOL_SOCKET;
-               sopt.sopt_name = SO_KEEPALIVE;
-               sopt.sopt_val = &val;
-               sopt.sopt_valsize = sizeof val;
-               val = 1;
-               sosetopt(so, &sopt);
-       }
-       if (so->so_proto->pr_domain->dom_family == AF_INET &&
-           so->so_proto->pr_protocol == IPPROTO_TCP) {
-               struct sockopt sopt;
-               int val;
-
-               bzero(&sopt, sizeof sopt);
-               sopt.sopt_dir = SOPT_SET;
-               sopt.sopt_level = IPPROTO_TCP;
-               sopt.sopt_name = TCP_NODELAY;
-               sopt.sopt_val = &val;
-               sopt.sopt_valsize = sizeof val;
-               val = 1;
-               sosetopt(so, &sopt);
+       timeo.tv_usec = 0;
+       timeo.tv_sec = 1;
+       error |= sock_setsockopt(so, SOL_SOCKET, SO_RCVTIMEO, &timeo, sizeof(timeo));
+       timeo.tv_sec = 30;
+       error |= sock_setsockopt(so, SOL_SOCKET, SO_SNDTIMEO, &timeo, sizeof(timeo));
+       if (error) {
+               log(LOG_INFO, "nfssvc_addsock: socket timeout setting error(s) %d\n", error);
+               error = 0;
        }
 
        }
 
-       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;
-       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-       if (tslp)
-               slp = tslp;
-       else {
-               MALLOC(slp, struct nfssvc_sock *, 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);
+       MALLOC(slp, struct nfsrv_sock *, sizeof(struct nfsrv_sock), M_NFSSVC, M_WAITOK);
+       if (!slp) {
+               mbuf_freem(mynam);
+               return (ENOMEM);
+       }
+       bzero((caddr_t)slp, sizeof (struct nfsrv_sock));
+       lck_rw_init(&slp->ns_rwlock, nfsrv_slp_rwlock_group, LCK_ATTR_NULL);
+       lck_mtx_init(&slp->ns_wgmutex, nfsrv_slp_mutex_group, LCK_ATTR_NULL);
+
+       lck_mtx_lock(nfsd_mutex);
+
+       if (soprotocol == IPPROTO_UDP) {
+               if (sodomain == AF_INET) {
+                       /* There should be only one UDP/IPv4 socket */
+                       if (nfsrv_udpsock) {
+                               lck_mtx_unlock(nfsd_mutex);
+                               nfsrv_slpfree(slp);
+                               mbuf_freem(mynam);
+                               return (EEXIST);
+                       }
+                       nfsrv_udpsock = slp;
+               }
+               if (sodomain == AF_INET6) {
+                       /* There should be only one UDP/IPv6 socket */
+                       if (nfsrv_udp6sock) {
+                               lck_mtx_unlock(nfsd_mutex);
+                               nfsrv_slpfree(slp);
+                               mbuf_freem(mynam);
+                               return (EEXIST);
+                       }
+                       nfsrv_udp6sock = slp;
+               }
        }
        }
+
+       /* add the socket to the list */
+       first = TAILQ_EMPTY(&nfsrv_socklist);
+       TAILQ_INSERT_TAIL(&nfsrv_socklist, slp, ns_chain);
+
+       sock_retain(so); /* grab a retain count on the socket */
        slp->ns_so = so;
        slp->ns_so = so;
+       slp->ns_sotype = sotype;
        slp->ns_nam = mynam;
        slp->ns_nam = mynam;
-       slp->ns_fp = fp;
-       (void)fref(fp);
-       thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-       s = splnet();
-       so->so_upcallarg = (caddr_t)slp;
-       so->so_upcall = nfsrv_rcv;
-       so->so_rcv.sb_flags |= SB_UPCALL; /* required for freebsd merge */
-       slp->ns_nflag = SLPN_NEEDQ;
-       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-       slp->ns_flag = SLP_VALID;
+
+       /* set up the socket up-call */
+       nfsrv_uc_addsock(slp, first);
+
+       /* mark that the socket is not in the nfsrv_sockwg list */
+       slp->ns_wgq.tqe_next = SLPNOLIST;
+       
+       slp->ns_flag = SLP_VALID | SLP_NEEDQ;
+
        nfsrv_wakenfsd(slp);
        nfsrv_wakenfsd(slp);
-       splx(s);
+       lck_mtx_unlock(nfsd_mutex);
+
        return (0);
 }
 
 /*
        return (0);
 }
 
 /*
- * Called by nfssvc() for nfsds. Just loops around servicing rpc requests
- * until it is killed by a signal.
+ * nfssvc_nfsd()
+ *
+ * nfsd theory of operation:
+ *
+ * The first nfsd thread stays in user mode accepting new TCP connections
+ * which are then added via the "addsock" call.  The rest of the nfsd threads
+ * simply call into the kernel and remain there in a loop handling NFS
+ * requests until killed by a signal.
+ * 
+ * There's a list of nfsd threads (nfsd_head).
+ * There's an nfsd queue that contains only those nfsds that are
+ *   waiting for work to do (nfsd_queue).
+ *
+ * There's a list of all NFS sockets (nfsrv_socklist) and two queues for
+ *   managing the work on the sockets:
+ *   nfsrv_sockwait - sockets w/new data waiting to be worked on
+ *   nfsrv_sockwork - sockets being worked on which may have more work to do
+ *   nfsrv_sockwg -- sockets which have pending write gather data
+ * When a socket receives data, if it is not currently queued, it
+ *   will be placed at the end of the "wait" queue.
+ * Whenever a socket needs servicing we make sure it is queued and
+ *   wake up a waiting nfsd (if there is one).
+ *
+ * nfsds will service at most 8 requests from the same socket before
+ *   defecting to work on another socket.
+ * nfsds will defect immediately if there are any sockets in the "wait" queue
+ * nfsds looking for a socket to work on check the "wait" queue first and
+ *   then check the "work" queue.
+ * When an nfsd starts working on a socket, it removes it from the head of
+ *   the queue it's currently on and moves it to the end of the "work" queue.
+ * When nfsds are checking the queues for work, any sockets found not to 
+ *   have any work are simply dropped from the queue.
+ *
  */
  */
-static int
-nfssvc_nfsd(nsd, argp, p)
-       struct nfsd_srvargs *nsd;
-       caddr_t argp;
-       struct proc *p;
+int
+nfssvc_nfsd(void)
 {
 {
-       register struct mbuf *m;
-       register int siz;
-       register struct nfssvc_sock *slp;
-       register struct socket *so;
-       struct nfsd *nfsd = nsd->nsd_nfsd;
+       mbuf_t m, mrep;
+       struct nfsrv_sock *slp;
+       struct nfsd *nfsd;
        struct nfsrv_descript *nd = NULL;
        struct nfsrv_descript *nd = NULL;
-       struct mbuf *mreq;
-       int error = 0, cacherep, s, sotype, writes_todo;
-       int procrastinate;
+       int error = 0, cacherep, writes_todo;
+       int siz, procrastinate, opcnt = 0;
        u_quad_t cur_usec;
        struct timeval now;
        u_quad_t cur_usec;
        struct timeval now;
+       struct vfs_context context;
+       struct timespec to;
 
 #ifndef nolint
        cacherep = RC_DOIT;
        writes_todo = 0;
 #endif
 
 #ifndef nolint
        cacherep = RC_DOIT;
        writes_todo = 0;
 #endif
-       s = splnet();
-       if (nfsd == (struct nfsd *)0) {
-               MALLOC(nfsd, struct nfsd *, sizeof(struct nfsd), M_NFSD, M_WAITOK);
-               nsd->nsd_nfsd = nfsd;
-               bzero((caddr_t)nfsd, sizeof (struct nfsd));
-               nfsd->nfsd_procp = p;
-               TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
-               nfs_numnfsd++;
-       }
+
+       MALLOC(nfsd, struct nfsd *, sizeof(struct nfsd), M_NFSD, M_WAITOK);
+       if (!nfsd)
+               return (ENOMEM);
+       bzero(nfsd, sizeof(struct nfsd));
+       lck_mtx_lock(nfsd_mutex);
+       if (nfsd_thread_count++ == 0)
+               nfsrv_initcache();              /* Init the server request cache */
+       
+       TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
+       lck_mtx_unlock(nfsd_mutex);
+
+       context.vc_thread = current_thread();
+
+       /* Set time out so that nfsd threads can wake up a see if they are still needed. */
+       to.tv_sec = 5;
+       to.tv_nsec = 0;
+
        /*
         * Loop getting rpc requests until SIGKILL.
         */
        for (;;) {
        /*
         * 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) {
+               if (nfsd_thread_max <= 0) {
+                       /* NFS server shutting down, get out ASAP */
+                       error = EINTR;
+                       slp = nfsd->nfsd_slp;
+               } else if (nfsd->nfsd_flag & NFSD_REQINPROG) {
+                       /* already have some work to do */
+                       error = 0;
+                       slp = nfsd->nfsd_slp;
+               } else {
+                       /* need to find work to do */
+                       error = 0;
+                       lck_mtx_lock(nfsd_mutex);
+                       while (!nfsd->nfsd_slp && TAILQ_EMPTY(&nfsrv_sockwait) && TAILQ_EMPTY(&nfsrv_sockwork)) {
+                               if (nfsd_thread_count > nfsd_thread_max) {
+                                       /*
+                                        * If we have no socket and there are more
+                                        * nfsd threads than configured, let's exit.
+                                        */
+                                       error = 0;
+                                       goto done;
+                               }
                                nfsd->nfsd_flag |= NFSD_WAITING;
                                nfsd->nfsd_flag |= NFSD_WAITING;
-                               nfsd_waiting++;
-                               error = tsleep((caddr_t)nfsd, PSOCK | PCATCH,
-                                   "nfsd", 0);
-                               nfsd_waiting--;
-                               if (error)
+                               TAILQ_INSERT_HEAD(&nfsd_queue, nfsd, nfsd_queue);
+                               error = msleep(nfsd, nfsd_mutex, PSOCK | PCATCH, "nfsd", &to);
+                               if (error) {
+                                       if (nfsd->nfsd_flag & NFSD_WAITING) {
+                                               TAILQ_REMOVE(&nfsd_queue, nfsd, nfsd_queue);
+                                               nfsd->nfsd_flag &= ~NFSD_WAITING;
+                                       }
+                                       if (error == EWOULDBLOCK)
+                                               continue;
                                        goto done;
                                        goto done;
+                               }
+                       }
+                       slp = nfsd->nfsd_slp;
+                       if (!slp && !TAILQ_EMPTY(&nfsrv_sockwait)) {
+                               /* look for a socket to work on in the wait queue */
+                               while ((slp = TAILQ_FIRST(&nfsrv_sockwait))) {
+                                       lck_rw_lock_exclusive(&slp->ns_rwlock);
+                                       /* remove from the head of the queue */
+                                       TAILQ_REMOVE(&nfsrv_sockwait, slp, ns_svcq);
+                                       slp->ns_flag &= ~SLP_WAITQ;
+                                       if ((slp->ns_flag & SLP_VALID) && (slp->ns_flag & SLP_WORKTODO))
+                                               break;
+                                       /* nothing to do, so skip this socket */
+                                       lck_rw_done(&slp->ns_rwlock);
+                               }
                        }
                        }
-                       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 && !TAILQ_EMPTY(&nfsrv_sockwork)) {
+                               /* look for a socket to work on in the work queue */
+                               while ((slp = TAILQ_FIRST(&nfsrv_sockwork))) {
+                                       lck_rw_lock_exclusive(&slp->ns_rwlock);
+                                       /* remove from the head of the queue */
+                                       TAILQ_REMOVE(&nfsrv_sockwork, slp, ns_svcq);
+                                       slp->ns_flag &= ~SLP_WORKQ;
+                                       if ((slp->ns_flag & SLP_VALID) && (slp->ns_flag & SLP_WORKTODO))
+                                               break;
+                                       /* nothing to do, so skip this socket */
+                                       lck_rw_done(&slp->ns_rwlock);
                                }
                                }
-                               if (slp == 0)
-                                       nfsd_head_flag &= ~NFSD_CHECKSLP;
                        }
                        }
-                       if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0)
+                       if (!nfsd->nfsd_slp && slp) {
+                               /* we found a socket to work on, grab a reference */
+                               slp->ns_sref++;
+                               nfsd->nfsd_slp = slp;
+                               opcnt = 0;
+                               /* and put it at the back of the work queue */
+                               TAILQ_INSERT_TAIL(&nfsrv_sockwork, slp, ns_svcq);
+                               slp->ns_flag |= SLP_WORKQ;
+                               lck_rw_done(&slp->ns_rwlock);
+                       }
+                       lck_mtx_unlock(nfsd_mutex);
+                       if (!slp)
                                continue;
                                continue;
+                       lck_rw_lock_exclusive(&slp->ns_rwlock);
                        if (slp->ns_flag & SLP_VALID) {
                        if (slp->ns_flag & SLP_VALID) {
-                               nfs_slplock(slp, 1);
-                               if (slp->ns_nflag & SLPN_DISCONN) {
+                               if ((slp->ns_flag & (SLP_NEEDQ|SLP_DISCONN)) == SLP_NEEDQ) {
+                                       slp->ns_flag &= ~SLP_NEEDQ;
+                                       nfsrv_rcv_locked(slp->ns_so, slp, MBUF_WAITOK);
+                               }
+                               if (slp->ns_flag & SLP_DISCONN)
                                        nfsrv_zapsock(slp);
                                        nfsrv_zapsock(slp);
-                                       thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-                               } else if (slp->ns_nflag & SLPN_NEEDQ) {
-                                       thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-                                       slp->ns_nflag &= ~SLPN_NEEDQ;
-                                       nfsrv_rcv(slp->ns_so, (caddr_t)slp,
-                                               M_WAIT);
-                               } else
-                                       thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
                                error = nfsrv_dorec(slp, nfsd, &nd);
                                error = nfsrv_dorec(slp, nfsd, &nd);
-                               thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-                               nfs_slpunlock(slp);
-                               microuptime(&now);
-                               cur_usec = (u_quad_t)now.tv_sec * 1000000 +
-                                       (u_quad_t)now.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;
+                               if (error == EINVAL) {  // RPCSEC_GSS drop
+                                       if (slp->ns_sotype == SOCK_STREAM)
+                                               nfsrv_zapsock(slp); // drop connection
+                               }
+                               writes_todo = 0;
+                               if (error && (slp->ns_wgtime || (slp->ns_flag & SLP_DOWRITES))) {
+                                       microuptime(&now);
+                                       cur_usec = (u_quad_t)now.tv_sec * 1000000 +
+                                               (u_quad_t)now.tv_usec;
+                                       if (slp->ns_wgtime <= cur_usec) {
+                                               error = 0;
+                                               cacherep = RC_DOIT;
+                                               writes_todo = 1;
+                                       }
+                                       slp->ns_flag &= ~SLP_DOWRITES;
+                               }
                                nfsd->nfsd_flag |= NFSD_REQINPROG;
                        }
                                nfsd->nfsd_flag |= NFSD_REQINPROG;
                        }
-               } else {
-                       error = 0;
-                       slp = nfsd->nfsd_slp;
+                       lck_rw_done(&slp->ns_rwlock);
                }
                }
-               if (error || (slp->ns_flag & SLP_VALID) == 0) {
+               if (error || (slp && !(slp->ns_flag & SLP_VALID))) {
                        if (nd) {
                        if (nd) {
+                               nfsm_chain_cleanup(&nd->nd_nmreq);
                                if (nd->nd_nam2)
                                if (nd->nd_nam2)
-                                       m_freem(nd->nd_nam2);
-                               FREE_ZONE((caddr_t)nd,
-                                               sizeof *nd, M_NFSRVDESC);
+                                       mbuf_freem(nd->nd_nam2);
+                               if (IS_VALID_CRED(nd->nd_cr))
+                                       kauth_cred_unref(&nd->nd_cr);
+                               if (nd->nd_gss_context)
+                                       nfs_gss_svc_ctx_deref(nd->nd_gss_context);
+                               FREE_ZONE(nd, sizeof(*nd), M_NFSRVDESC);
                                nd = NULL;
                        }
                                nd = NULL;
                        }
-                       nfsd->nfsd_slp = (struct nfssvc_sock *)0;
+                       nfsd->nfsd_slp = NULL;
                        nfsd->nfsd_flag &= ~NFSD_REQINPROG;
                        nfsd->nfsd_flag &= ~NFSD_REQINPROG;
-                       nfsrv_slpderef(slp);
+                       if (slp)
+                               nfsrv_slpderef(slp);
+                       if (nfsd_thread_max <= 0)
+                               break;
                        continue;
                }
                        continue;
                }
-               splx(s);
-               so = slp->ns_so;
-               sotype = so->so_type;
                if (nd) {
                    microuptime(&nd->nd_starttime);
                    if (nd->nd_nam2)
                if (nd) {
                    microuptime(&nd->nd_starttime);
                    if (nd->nd_nam2)
@@ -810,171 +1046,159 @@ nfssvc_nfsd(nsd, argp, p)
                    else
                        nd->nd_nam = slp->ns_nam;
 
                    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);
+                   cacherep = nfsrv_getcache(nd, slp, &mrep);
 
 
-                   /*
-                    * Check for just starting up for NQNFS and send
-                    * fake "try again later" replies to the NQNFS clients.
-                    */
-                   microtime(&now);
-                   if (notstarted && nqnfsstarttime <= now.tv_sec) {
-                       if (modify_flag) {
-                               nqnfsstarttime = now.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;
-                   } else if (nfs_privport) {
-                       /* Check if source port is privileged */
-                       u_short port;
-                       struct sockaddr *nam = mtod(nd->nd_nam, struct sockaddr*);
-                       struct sockaddr_in *sin;
-
-                       sin = (struct sockaddr_in *)nam;
-                       port = ntohs(sin->sin_port);
-                       if (port >= IPPORT_RESERVED && 
-                           nd->nd_procnum != NFSPROC_NULL) {
+                   if (nfsrv_require_resv_port) {
+                       /* Check if source port is a reserved port */
+                       in_port_t port = 0;
+                       struct sockaddr *saddr = mbuf_data(nd->nd_nam);
+
+                       if (saddr->sa_family == AF_INET)
+                               port = ntohs(((struct sockaddr_in*)saddr)->sin_port);
+                       else if (saddr->sa_family == AF_INET6)
+                               port = ntohs(((struct sockaddr_in6*)saddr)->sin6_port);
+                       if ((port >= IPPORT_RESERVED) && (nd->nd_procnum != NFSPROC_NULL)) {
                            nd->nd_procnum = NFSPROC_NOOP;
                            nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
                            cacherep = RC_DOIT;
                            nd->nd_procnum = NFSPROC_NOOP;
                            nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
                            cacherep = RC_DOIT;
-                           printf("NFS request from unprivileged port (%s:%d)\n",
-                                  (char *)(inet_ntoa(sin->sin_addr)), port);
                        }
                    }
 
                }
 
                /*
                        }
                    }
 
                }
 
                /*
-                * Loop to get all the write rpc relies that have been
+                * Loop to get all the write RPC replies that have been
                 * gathered together.
                 */
                do {
                    switch (cacherep) {
                    case RC_DOIT:
                 * gathered together.
                 */
                do {
                    switch (cacherep) {
                    case RC_DOIT:
-                       if (nd && (nd->nd_flag & ND_NFSV3))
-                           procrastinate = nfsrvw_procrastinate_v3;
+                       if (nd && (nd->nd_vers == NFS_VER3))
+                           procrastinate = nfsrv_wg_delay_v3;
                        else
                        else
-                           procrastinate = nfsrvw_procrastinate;
-                       if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE &&
-                           procrastinate > 0 && !notstarted))
-                           error = nfsrv_writegather(&nd, slp,
-                               nfsd->nfsd_procp, &mreq);
+                           procrastinate = nfsrv_wg_delay;
+                       lck_rw_lock_shared(&nfsrv_export_rwlock);
+                       context.vc_ucred = NULL;
+                       if (writes_todo || ((nd->nd_procnum == NFSPROC_WRITE) && (procrastinate > 0)))
+                               error = nfsrv_writegather(&nd, slp, &context, &mrep);
                        else
                        else
-                           error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
-                               slp, nfsd->nfsd_procp, &mreq);
-                       if (mreq == NULL)
+                               error = (*(nfsrv_procs[nd->nd_procnum]))(nd, slp, &context, &mrep);
+                       lck_rw_done(&nfsrv_export_rwlock);
+                       if (mrep == NULL) {
+                               /*
+                                * If this is a stream socket and we are not going
+                                * to send a reply we better close the connection
+                                * so the client doesn't hang.
+                                */
+                               if (error && slp->ns_sotype == SOCK_STREAM) {
+                                       lck_rw_lock_exclusive(&slp->ns_rwlock);
+                                       nfsrv_zapsock(slp);
+                                       lck_rw_done(&slp->ns_rwlock);
+                                       printf("NFS server: NULL reply from proc = %d error = %d\n",
+                                               nd->nd_procnum, error);
+                               }
                                break;
                                break;
+
+                       }
                        if (error) {
                        if (error) {
-                               if (nd->nd_procnum != NQNFSPROC_VACATED)
-                                       nfsstats.srv_errs++;
-                               nfsrv_updatecache(nd, FALSE, mreq);
+                               OSAddAtomic64(1, &nfsstats.srv_errs);
+                               nfsrv_updatecache(nd, FALSE, mrep);
                                if (nd->nd_nam2) {
                                if (nd->nd_nam2) {
-                                       m_freem(nd->nd_nam2);
+                                       mbuf_freem(nd->nd_nam2);
                                        nd->nd_nam2 = NULL;
                                }
                                break;
                        }
                                        nd->nd_nam2 = NULL;
                                }
                                break;
                        }
-                       nfsstats.srvrpccnt[nd->nd_procnum]++;
-                       nfsrv_updatecache(nd, TRUE, mreq);
-                       nd->nd_mrep = (struct mbuf *)0;
+                       OSAddAtomic64(1, &nfsstats.srvrpccnt[nd->nd_procnum]);
+                       nfsrv_updatecache(nd, TRUE, mrep);
+                       /* FALLTHRU */
+
                    case RC_REPLY:
                    case RC_REPLY:
-                       m = mreq;
+                       if (nd->nd_gss_mb != NULL) {    // It's RPCSEC_GSS
+                               /*
+                                * Need to checksum or encrypt the reply
+                                */
+                               error = nfs_gss_svc_protect_reply(nd, mrep);
+                               if (error) {
+                                       mbuf_freem(mrep);
+                                       break;
+                               }
+                       }
+
+                       /*
+                        * Get the total size of the reply
+                        */
+                       m = mrep;
                        siz = 0;
                        while (m) {
                        siz = 0;
                        while (m) {
-                               siz += m->m_len;
-                               m = m->m_next;
+                               siz += mbuf_len(m);
+                               m = mbuf_next(m);
                        }
                        if (siz <= 0 || siz > NFS_MAXPACKET) {
                                printf("mbuf siz=%d\n",siz);
                                panic("Bad nfs svc reply");
                        }
                        }
                        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;
+                       m = mrep;
+                       mbuf_pkthdr_setlen(m, siz);
+                       error = mbuf_pkthdr_setrcvif(m, NULL);
+                       if (error)
+                               panic("nfsd setrcvif failed: %d", error);
                        /*
                         * For stream protocols, prepend a Sun RPC
                         * Record Mark.
                         */
                        /*
                         * 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 (slp->ns_sotype == SOCK_STREAM) {
+                               error = mbuf_prepend(&m, NFSX_UNSIGNED, MBUF_WAITOK);
+                               if (!error)
+                                       *(u_int32_t*)mbuf_data(m) = htonl(0x80000000 | siz);
                        }
                        }
-                       if (so->so_proto->pr_flags & PR_CONNREQUIRED)
-                               (void) nfs_slplock(slp, 1);
-                       if (slp->ns_flag & SLP_VALID)
-                           error = nfs_send(so, nd->nd_nam2, m, NULL);
-                       else {
-                           error = EPIPE;
-                           m_freem(m);
+                       if (!error) {
+                               if (slp->ns_flag & SLP_VALID) {
+                                   error = nfsrv_send(slp, nd->nd_nam2, m);
+                               } else {
+                                   error = EPIPE;
+                                   mbuf_freem(m);
+                               }
+                       } else {
+                               mbuf_freem(m);
                        }
                        }
-                       mreq = NULL;
-                       if (nfsrtton)
-                               nfsd_rt(sotype, nd, cacherep);
+                       mrep = NULL;
                        if (nd->nd_nam2) {
                        if (nd->nd_nam2) {
-                               MFREE(nd->nd_nam2, m);
+                               mbuf_freem(nd->nd_nam2);
                                nd->nd_nam2 = NULL;
                        }
                                nd->nd_nam2 = NULL;
                        }
-                       if (nd->nd_mrep) {
-                               m_freem(nd->nd_mrep);
-                               nd->nd_mrep = NULL;
-                       }
-                       if (error == EPIPE)
+                       if (error == EPIPE) {
+                               lck_rw_lock_exclusive(&slp->ns_rwlock);
                                nfsrv_zapsock(slp);
                                nfsrv_zapsock(slp);
-                       if (so->so_proto->pr_flags & PR_CONNREQUIRED)
-                               nfs_slpunlock(slp);
+                               lck_rw_done(&slp->ns_rwlock);
+                       }
                        if (error == EINTR || error == ERESTART) {
                        if (error == EINTR || error == ERESTART) {
-                               FREE_ZONE((caddr_t)nd,
-                                               sizeof *nd, M_NFSRVDESC);
+                               nfsm_chain_cleanup(&nd->nd_nmreq);
+                               if (IS_VALID_CRED(nd->nd_cr))
+                                       kauth_cred_unref(&nd->nd_cr);
+                               if (nd->nd_gss_context)
+                                       nfs_gss_svc_ctx_deref(nd->nd_gss_context);
+                               FREE_ZONE(nd, sizeof(*nd), M_NFSRVDESC);
                                nfsrv_slpderef(slp);
                                nfsrv_slpderef(slp);
-                               s = splnet();
+                               lck_mtx_lock(nfsd_mutex);
                                goto done;
                        }
                        break;
                    case RC_DROPIT:
                                goto done;
                        }
                        break;
                    case RC_DROPIT:
-                       if (nfsrtton)
-                               nfsd_rt(sotype, nd, cacherep);
-                       m_freem(nd->nd_mrep);
-                       m_freem(nd->nd_nam2);
-                       nd->nd_mrep = nd->nd_nam2 = NULL;
+                       mbuf_freem(nd->nd_nam2);
+                       nd->nd_nam2 = NULL;
                        break;
                    };
                        break;
                    };
+                   opcnt++;
                    if (nd) {
                    if (nd) {
-                       if (nd->nd_mrep)
-                               m_freem(nd->nd_mrep);
+                       nfsm_chain_cleanup(&nd->nd_nmreq);
                        if (nd->nd_nam2)
                        if (nd->nd_nam2)
-                               m_freem(nd->nd_nam2);
-                       FREE_ZONE((caddr_t)nd, sizeof *nd, M_NFSRVDESC);
+                               mbuf_freem(nd->nd_nam2);
+                       if (IS_VALID_CRED(nd->nd_cr))
+                               kauth_cred_unref(&nd->nd_cr);
+                       if (nd->nd_gss_context)
+                               nfs_gss_svc_ctx_deref(nd->nd_gss_context);
+                       FREE_ZONE(nd, sizeof(*nd), M_NFSRVDESC);
                        nd = NULL;
                    }
 
                        nd = NULL;
                    }
 
@@ -982,611 +1206,340 @@ nfssvc_nfsd(nsd, argp, p)
                     * Check to see if there are outstanding writes that
                     * need to be serviced.
                     */
                     * Check to see if there are outstanding writes that
                     * need to be serviced.
                     */
-                   microuptime(&now);
-                   cur_usec = (u_quad_t)now.tv_sec * 1000000 +
-                       (u_quad_t)now.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);
+                   writes_todo = 0;
+                   if (slp->ns_wgtime) {
+                       microuptime(&now);
+                       cur_usec = (u_quad_t)now.tv_sec * 1000000 +
+                               (u_quad_t)now.tv_usec;
+                       if (slp->ns_wgtime <= cur_usec) {
+                           cacherep = RC_DOIT;
+                           writes_todo = 1;
+                       }
+                   }
                } while (writes_todo);
                } while (writes_todo);
-               s = splnet();
-               thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-               if (nfsrv_dorec(slp, nfsd, &nd)) {
-                       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
+
+               nd = NULL;
+               if (TAILQ_EMPTY(&nfsrv_sockwait) && (opcnt < 8)) {
+                       lck_rw_lock_exclusive(&slp->ns_rwlock);
+                       error = nfsrv_dorec(slp, nfsd, &nd);
+                       if (error == EINVAL) {  // RPCSEC_GSS drop
+                               if (slp->ns_sotype == SOCK_STREAM)
+                                       nfsrv_zapsock(slp); // drop connection
+                       }
+                       lck_rw_done(&slp->ns_rwlock);
+               }
+               if (!nd) {
+                       /* drop our reference on the socket */
                        nfsd->nfsd_flag &= ~NFSD_REQINPROG;
                        nfsd->nfsd_slp = NULL;
                        nfsrv_slpderef(slp);
                        nfsd->nfsd_flag &= ~NFSD_REQINPROG;
                        nfsd->nfsd_slp = NULL;
                        nfsrv_slpderef(slp);
-               } else
-                       thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
+               }
        }
        }
+       lck_mtx_lock(nfsd_mutex);
 done:
        TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
 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 */
+       FREE(nfsd, M_NFSD);
+       if (--nfsd_thread_count == 0)
+               nfsrv_cleanup();
+       lck_mtx_unlock(nfsd_mutex);
        return (error);
 }
        return (error);
 }
-#endif /* NFS_NOSERVER */
-
-int nfs_defect = 0;
-/* XXX CSM 11/25/97 Upgrade sysctl.h someday */
-#ifdef notyet
-SYSCTL_INT(_vfs_nfs, OID_AUTO, defect, CTLFLAG_RW, &nfs_defect, 0, "");
-#endif
 
 
-#ifndef _SYS_SYSPROTO_H_
-struct nfsclnt_args {
-       int flag;
-       caddr_t argp;
-};
-#endif
 int
 int
-nfsclnt(struct proc *p, struct nfsclnt_args *uap)
+nfssvc_export(user_addr_t argp)
 {
 {
-       struct lockd_ans la;
-       int error;
-
-       if (uap->flag == NFSCLNT_LOCKDWAIT) {
-               return (nfslockdwait(p));
-       }
-       if (uap->flag == NFSCLNT_LOCKDANS) {
-               error = copyin(uap->argp, &la, sizeof(la));
-               return (error != 0 ? error : nfslockdans(p, &la));
-       }
-       if (uap->flag == NFSCLNT_LOCKDFD)
-               return (nfslockdfd(p, (int)uap->argp));
-       return EINVAL;
-}
-
+       int error = 0, is_64bit;
+       struct user_nfs_export_args unxa;
+       vfs_context_t ctx = vfs_context_current();
 
 
-static int nfssvc_iod_continue(int);
-
-/*
- * 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.
- */
-static int
-nfssvc_iod(p)
-       struct proc *p;
-{
-       register int i, myiod;
-       struct nfsmount *nmp;
-       int error = 0;
-       struct uthread *ut;
+       is_64bit = IS_64BIT_PROCESS(vfs_context_proc(ctx));
 
 
-       /*
-        * 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;
+       /* copy in pointers to path and export args */
+       if (is_64bit) {
+               error = copyin(argp, (caddr_t)&unxa, sizeof(unxa));
+       } else {
+               struct nfs_export_args tnxa;
+               error = copyin(argp, (caddr_t)&tnxa, sizeof(tnxa));
+               if (error == 0) {
+                       /* munge into LP64 version of nfs_export_args structure */
+                       unxa.nxa_fsid = tnxa.nxa_fsid;
+                       unxa.nxa_expid = tnxa.nxa_expid;
+                       unxa.nxa_fspath = CAST_USER_ADDR_T(tnxa.nxa_fspath);
+                       unxa.nxa_exppath = CAST_USER_ADDR_T(tnxa.nxa_exppath);
+                       unxa.nxa_flags = tnxa.nxa_flags;
+                       unxa.nxa_netcount = tnxa.nxa_netcount;
+                       unxa.nxa_nets = CAST_USER_ADDR_T(tnxa.nxa_nets);
                }
                }
-       if (myiod == -1)
-               return (EBUSY);
-       nfs_numasync++;
-
-       /* stuff myiod into uthread to get off local stack for continuation */
-
-       ut = (struct uthread *)get_bsdthread_info(current_act());
-       ut->uu_state.uu_nfs_myiod = myiod;  /* squirrel away for continuation */
-
-       nfssvc_iod_continue(0);
-       /* NOTREACHED */
-
-}
-
-/*
- * Continuation for Asynchronous I/O daemons for client nfs.
- */
-static int
-nfssvc_iod_continue(error)
-{
-       register struct nfsbuf *bp;
-       register int i, myiod;
-       struct nfsmount *nmp;
-       struct uthread *ut;
-       struct proc *p;
-
-       /*
-        * real myiod is stored in uthread, recover it
-        */
-       ut = (struct uthread *)get_bsdthread_info(current_act());
-       myiod = ut->uu_state.uu_nfs_myiod;
-       p = current_proc();
+       }
+       if (error)
+               return (error);
 
 
-       /*
-        * Just loop around doin our stuff until SIGKILL
-        *  - actually we don't loop with continuations...
-        */
-       for (;;) {
-           while (((nmp = nfs_iodmount[myiod]) == NULL
-                   || nmp->nm_bufq.tqh_first == NULL)
-                  && error == 0 && nfs_ioddelwri == 0) {
-               if (nmp)
-                   nmp->nm_bufqiods--;
-               nfs_iodwant[myiod] = p;
-               nfs_iodmount[myiod] = NULL;
-               error = tsleep0((caddr_t)&nfs_iodwant[myiod],
-                       PWAIT | PCATCH, "nfsidl", 0, nfssvc_iod_continue);
-               /* NOTREACHED */
-           }
-           if (error) {
-               nfs_asyncdaemon[myiod] = 0;
-               if (nmp) nmp->nm_bufqiods--;
-               nfs_iodwant[myiod] = NULL;
-               nfs_iodmount[myiod] = NULL;
-               nfs_numasync--;
-               if (error == EINTR || error == ERESTART)
-                 error = 0;
-               unix_syscall_return(error);
-           }
-           if (nmp != NULL) {
-               while ((bp = nmp->nm_bufq.tqh_first) != NULL) {
-                   /* Take one off the front of the list */
-                   TAILQ_REMOVE(&nmp->nm_bufq, bp, nb_free);
-                   bp->nb_free.tqe_next = NFSNOLIST;
-                   nmp->nm_bufqlen--;
-                   if (nmp->nm_bufqwant && nmp->nm_bufqlen < 2 * nfs_numasync) {
-                       nmp->nm_bufqwant = FALSE;
-                       wakeup(&nmp->nm_bufq);
-                   }
-                   if (ISSET(bp->nb_flags, NB_READ))
-                       (void) nfs_doio(bp, bp->nb_rcred, (struct proc *)0);
-                   else
-                       (void) nfs_doio(bp, bp->nb_wcred, (struct proc *)0);
+       error = nfsrv_export(&unxa, ctx);
 
 
-                   /*
-                    * If there are more than one iod on this mount, then defect
-                    * so that the iods can be shared out fairly between the mounts
-                    */
-                   if (nfs_defect && nmp->nm_bufqiods > 1) {
-                       NFS_DPF(ASYNCIO,
-                               ("nfssvc_iod: iod %d defecting from mount %p\n",
-                                myiod, nmp));
-                       nfs_iodmount[myiod] = NULL;
-                       nmp->nm_bufqiods--;
-                       break;
-                   }
-               }
-           }
-           if (nfs_ioddelwri) {
-               i = 0;
-               nfs_ioddelwri = 0;
-               while (i < 8 && (bp = TAILQ_FIRST(&nfsbufdelwri)) != NULL) {
-                       struct nfsnode *np = VTONFS(bp->nb_vp);
-                       nfs_buf_remfree(bp);
-                       if (ISSET(bp->nb_flags, NB_NEEDCOMMIT)) {
-                               /* put buffer at end of delwri list */
-                               TAILQ_INSERT_TAIL(&nfsbufdelwri, bp, nb_free);
-                               nfsbufdelwricnt++;
-                               nfs_flushcommits(np->n_vnode, (struct proc *)0);
-                       } else {
-                               SET(bp->nb_flags, (NB_BUSY | NB_ASYNC | NB_IOD));
-                               nfs_buf_write(bp);
-                       }
-                       i++;
-               }
-           }
-       }
+       return (error);
 }
 
 /*
 }
 
 /*
- * Shut down a socket associated with an nfssvc_sock structure.
+ * Shut down a socket associated with an nfsrv_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.
  */
  * 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.
  */
-static void
-nfsrv_zapsock(slp)
-       register struct nfssvc_sock *slp;
+void
+nfsrv_zapsock(struct nfsrv_sock *slp)
 {
 {
-       register struct nfsuid *nuidp, *nnuidp;
-       register struct nfsrv_descript *nwp, *nnwp;
-       struct socket *so;
-       struct file *fp;
-       struct mbuf *m;
-       int s;
+       socket_t so;
 
 
+       if ((slp->ns_flag & SLP_VALID) == 0)
+               return;
        slp->ns_flag &= ~SLP_ALLFLAGS;
        slp->ns_flag &= ~SLP_ALLFLAGS;
-       slp->ns_nflag &= ~SLP_ALLFLAGS;
-       fp = slp->ns_fp;
-       if (fp) {
-               slp->ns_fp = (struct file *)0;
-               so = slp->ns_so;
-               thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL);
-               so->so_upcall = NULL;
-               so->so_rcv.sb_flags &= ~SB_UPCALL;
-               soshutdown(so, 2);
-               if (slp->ns_nam)
-                       MFREE(slp->ns_nam, m);
-               m_freem(slp->ns_raw);
-               m_freem(slp->ns_rec);
-               slp->ns_nam = slp->ns_raw = slp->ns_rec = NULL;
-               thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL);
-               closef(fp, (struct proc *)0);
-               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_ZONE((caddr_t)nuidp,
-                                       sizeof (struct nfsuid), 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_ZONE((caddr_t)nwp, sizeof *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_state & NFSSTA_WAITAUTH) == 0) {
-               nmp->nm_state |= NFSSTA_WANTAUTH;
-               (void) tsleep((caddr_t)&nmp->nm_authtype, PSOCK,
-                       "nfsauth1", 2 * hz);
-               error = nfs_sigintr(nmp, rep, rep->r_procp);
-               if (error) {
-                       nmp->nm_state &= ~NFSSTA_WANTAUTH;
-                       return (error);
-               }
-       }
-       nmp->nm_state &= ~(NFSSTA_WAITAUTH | NFSSTA_WANTAUTH);
-       MALLOC(*auth_str, char *, RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK);
-       nmp->nm_authstr = *auth_str;
-       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_state & NFSSTA_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_state & NFSSTA_AUTHERR) {
-               nmp->nm_state &= ~NFSSTA_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_state &= ~NFSSTA_HASAUTH;
-       nmp->nm_state |= NFSSTA_WAITAUTH;
-       if (nmp->nm_state & NFSSTA_WANTAUTH) {
-               nmp->nm_state &= ~NFSSTA_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, now;
 
 
-#if 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;
-       }
-       microtime(&now);
-       if (!nuidp || nuidp->nu_expire < now.tv_sec)
-               return (EACCES);
+       so = slp->ns_so;
+       if (so == NULL)
+               return;
 
        /*
 
        /*
-        * Move to the end of the lru list (end of lru == most recently used).
+        * Attempt to deter future up-calls, but leave the
+        * up-call info in place to avoid a race with the
+        * networking code.
         */
         */
-       TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru);
-       TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru);
+       socket_lock(so, 1);
+       so->so_rcv.sb_flags &= ~SB_UPCALL;
+       socket_unlock(so, 1);
 
 
-       MALLOC(nickp, u_long *, 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;
+       sock_shutdown(so, SHUT_RDWR);
 
        /*
 
        /*
-        * Now we must encrypt the verifier and package it up.
+        * Remove from the up-call queue
         */
         */
-       verfp = (u_long *)verf_str;
-       *verfp++ = txdr_unsigned(RPCAKN_NICKNAME);
-       microtime(&now);
-       if (now.tv_sec > nuidp->nu_timestamp.tv_sec ||
-           (now.tv_sec == nuidp->nu_timestamp.tv_sec &&
-            now.tv_usec > nuidp->nu_timestamp.tv_usec))
-               nuidp->nu_timestamp = now;
-       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.
-        */
-#if NFSKERB
-       XXX
-#endif
-
-       *verfp++ = ktvout.tv_sec;
-       *verfp++ = ktvout.tv_usec;
-       *verfp = 0;
-       return (0);
+       nfsrv_uc_dequeue(slp);
 }
 
 /*
 }
 
 /*
- * Save the current nickname in a hash list entry on the mount point.
+ * cleanup and release a server socket structure.
  */
  */
-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;
+void
+nfsrv_slpfree(struct nfsrv_sock *slp)
 {
 {
-       register struct nfsuid *nuidp;
-       register u_long *tl;
-       register long t1;
-       struct mbuf *md = *mdp;
-       struct timeval ktvin, ktvout, now;
-       u_long nick;
-       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);
+       struct nfsrv_descript *nwp, *nnwp;
 
 
-               /*
-                * Decrypt the timestamp in ecb mode.
-                */
-#if NFSKERB
-               XXX
-#endif
-               ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec);
-               ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec);
-               microtime(&now);
-               deltasec = now.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++;
-                               MALLOC_ZONE(nuidp, struct nfsuid *,
-                                               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 = now.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);
-}
+       if (slp->ns_so) {
+               sock_release(slp->ns_so);
+               slp->ns_so = NULL;
+       }
+       if (slp->ns_nam)
+               mbuf_free(slp->ns_nam);
+       if (slp->ns_raw)
+               mbuf_freem(slp->ns_raw);
+       if (slp->ns_rec)
+               mbuf_freem(slp->ns_rec);
+       if (slp->ns_frag)
+               mbuf_freem(slp->ns_frag);
+       slp->ns_nam = slp->ns_raw = slp->ns_rec = slp->ns_frag = NULL;
+       slp->ns_reccnt = 0;
+
+       if (slp->ns_ua)
+               FREE(slp->ns_ua, M_NFSSVC);
+
+       for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) {
+               nnwp = nwp->nd_tq.le_next;
+               LIST_REMOVE(nwp, nd_tq);
+               nfsm_chain_cleanup(&nwp->nd_nmreq);
+               if (nwp->nd_mrep)
+                       mbuf_freem(nwp->nd_mrep);
+               if (nwp->nd_nam2)
+                       mbuf_freem(nwp->nd_nam2);
+               if (IS_VALID_CRED(nwp->nd_cr))
+                       kauth_cred_unref(&nwp->nd_cr);
+               if (nwp->nd_gss_context)
+                       nfs_gss_svc_ctx_deref(nwp->nd_gss_context);
+               FREE_ZONE(nwp, sizeof(*nwp), M_NFSRVDESC);
+       }
+       LIST_INIT(&slp->ns_tq);
 
 
-#ifndef NFS_NOSERVER
+       lck_rw_destroy(&slp->ns_rwlock, nfsrv_slp_rwlock_group);
+       lck_mtx_destroy(&slp->ns_wgmutex, nfsrv_slp_mutex_group);
+       FREE(slp, M_NFSSVC);
+}
 
 /*
  * Derefence a server socket structure. If it has no more references and
  * is no longer valid, you can throw it away.
  */
 void
 
 /*
  * 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;
+nfsrv_slpderef(struct nfsrv_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);
+       struct timeval now;
+
+       lck_mtx_lock(nfsd_mutex);
+       lck_rw_lock_exclusive(&slp->ns_rwlock);
+       slp->ns_sref--;
+
+       if (slp->ns_sref || (slp->ns_flag & SLP_VALID)) {
+               if ((slp->ns_flag & SLP_QUEUED) && !(slp->ns_flag & SLP_WORKTODO)) {
+                       /* remove socket from queue since there's no work */
+                       if (slp->ns_flag & SLP_WAITQ)
+                               TAILQ_REMOVE(&nfsrv_sockwait, slp, ns_svcq);
+                       else
+                               TAILQ_REMOVE(&nfsrv_sockwork, slp, ns_svcq);
+                       slp->ns_flag &= ~SLP_QUEUED;
+               }
+               lck_rw_done(&slp->ns_rwlock);
+               lck_mtx_unlock(nfsd_mutex);
+               return;
        }
        }
-}
 
 
-/*
- * Lock a socket against others.
- */
-int
-nfs_slplock(slp, wait)
-       register struct nfssvc_sock *slp;
-       int wait;
-{
-       int *statep = &slp->ns_solock;
+       /* This socket is no longer valid, so we'll get rid of it */
+
+       if (slp->ns_flag & SLP_QUEUED) {
+               if (slp->ns_flag & SLP_WAITQ)
+                       TAILQ_REMOVE(&nfsrv_sockwait, slp, ns_svcq);
+               else
+                       TAILQ_REMOVE(&nfsrv_sockwork, slp, ns_svcq);
+               slp->ns_flag &= ~SLP_QUEUED;
+       }
 
 
-       if (!wait && (*statep & NFSSTA_SNDLOCK))
-               return(0);      /* already locked, fail */
-       while (*statep & NFSSTA_SNDLOCK) {
-               *statep |= NFSSTA_WANTSND;
-               (void) tsleep((caddr_t)statep, PZERO - 1, "nfsslplck", 0);
+       /*
+        * Queue the socket up for deletion
+        * and start the timer to delete it
+        * after it has been in limbo for
+        * a while.
+        */
+       microuptime(&now);
+       slp->ns_timestamp = now.tv_sec;
+       TAILQ_REMOVE(&nfsrv_socklist, slp, ns_chain);
+       TAILQ_INSERT_TAIL(&nfsrv_deadsocklist, slp, ns_chain);
+       if (!nfsrv_deadsock_timer_on) {
+               nfsrv_deadsock_timer_on = 1;
+               nfs_interval_timer_start(nfsrv_deadsock_timer_call,
+                               NFSRV_DEADSOCKDELAY * 1000);
        }
        }
-       *statep |= NFSSTA_SNDLOCK;
-       return (1);
+
+       lck_rw_done(&slp->ns_rwlock);
+       /* now remove from the write gather socket list */ 
+       if (slp->ns_wgq.tqe_next != SLPNOLIST) {
+               TAILQ_REMOVE(&nfsrv_sockwg, slp, ns_wgq);
+               slp->ns_wgq.tqe_next = SLPNOLIST;
+       }
+       lck_mtx_unlock(nfsd_mutex);
 }
 
 /*
 }
 
 /*
- * Unlock the stream socket for others.
+ * Check periodically for dead sockets pending delete.
+ * If a socket has been dead for more than NFSRV_DEADSOCKDELAY
+ * seconds then we assume it's safe to free.
  */
 void
  */
 void
-nfs_slpunlock(slp)
-       struct nfssvc_sock *slp;
+nfsrv_deadsock_timer(__unused void *param0, __unused void *param1)
 {
 {
-       int *statep = &slp->ns_solock;
-
-       if ((*statep & NFSSTA_SNDLOCK) == 0)
-               panic("nfs slpunlock");
-       *statep &= ~NFSSTA_SNDLOCK;
-       if (*statep & NFSSTA_WANTSND) {
-               *statep &= ~NFSSTA_WANTSND;
-               wakeup((caddr_t)statep);
+       struct nfsrv_sock *slp;
+       struct timeval now;
+       time_t time_to_wait;
+
+       microuptime(&now);
+       lck_mtx_lock(nfsd_mutex);
+
+       while ((slp = TAILQ_FIRST(&nfsrv_deadsocklist))) {
+               if ((slp->ns_timestamp + NFSRV_DEADSOCKDELAY) > now.tv_sec)
+                       break;
+               TAILQ_REMOVE(&nfsrv_deadsocklist, slp, ns_chain);
+               nfsrv_slpfree(slp);
        }
        }
+       if (TAILQ_EMPTY(&nfsrv_deadsocklist)) {
+               nfsrv_deadsock_timer_on = 0;
+               lck_mtx_unlock(nfsd_mutex);
+               return;
+       }
+       time_to_wait = (slp->ns_timestamp + NFSRV_DEADSOCKDELAY) - now.tv_sec;
+       if (time_to_wait < 1)
+               time_to_wait = 1;
+
+       lck_mtx_unlock(nfsd_mutex);
+
+       nfs_interval_timer_start(nfsrv_deadsock_timer_call,
+               time_to_wait * 1000);
 }
 
 /*
 }
 
 /*
- * Initialize the data structures for the server.
- * Handshake with any new nfsds starting up to avoid any chance of
- * corruption.
+ * Clean up the data structures for the server.
  */
 void
  */
 void
-nfsrv_init(terminating)
-       int terminating;
+nfsrv_cleanup(void)
 {
 {
-       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);
+       struct nfsrv_sock *slp, *nslp;
+       struct timeval now;
+#if CONFIG_FSE
+       struct nfsrv_fmod *fp, *nfp;
+       int i;
+#endif
+
+       microuptime(&now);
+       for (slp = TAILQ_FIRST(&nfsrv_socklist); slp != 0; slp = nslp) {
+               nslp = TAILQ_NEXT(slp, ns_chain);
+               if (slp->ns_flag & SLP_VALID) {
+                       lck_rw_lock_exclusive(&slp->ns_rwlock);
+                       nfsrv_zapsock(slp);
+                       lck_rw_done(&slp->ns_rwlock);
+               }
+               if (slp->ns_flag & SLP_QUEUED) {
+                       if (slp->ns_flag & SLP_WAITQ)
+                               TAILQ_REMOVE(&nfsrv_sockwait, slp, ns_svcq);
+                       else
+                               TAILQ_REMOVE(&nfsrv_sockwork, slp, ns_svcq);
+                       slp->ns_flag &= ~SLP_QUEUED;
+               }
+               if (slp->ns_wgq.tqe_next != SLPNOLIST) {
+                       TAILQ_REMOVE(&nfsrv_sockwg, slp, ns_wgq);
+                       slp->ns_wgq.tqe_next = SLPNOLIST;
+               }
+               /* queue the socket up for deletion */
+               slp->ns_timestamp = now.tv_sec;
+               TAILQ_REMOVE(&nfsrv_socklist, slp, ns_chain);
+               TAILQ_INSERT_TAIL(&nfsrv_deadsocklist, slp, ns_chain);
+               if (!nfsrv_deadsock_timer_on) {
+                       nfsrv_deadsock_timer_on = 1;
+                       nfs_interval_timer_start(nfsrv_deadsock_timer_call,
+                               NFSRV_DEADSOCKDELAY * 1000);
                }
                }
-               nfsrv_cleancache();     /* And clear out server cache */
-/* XXX CSM 12/4/97 Revisit when enabling WebNFS */
-#ifdef notyet
-       } else
-               nfs_pub.np_valid = 0;
-#else
        }
        }
-#endif
 
 
-       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);
+#if CONFIG_FSE
+       /*
+        * Flush pending file write fsevents
+        */
+       lck_mtx_lock(nfsrv_fmod_mutex);
+       for (i = 0; i < NFSRVFMODHASHSZ; i++) {
+               for (fp = LIST_FIRST(&nfsrv_fmod_hashtbl[i]); fp; fp = nfp) {
+                       /*
+                        * Fire off the content modified fsevent for each
+                        * entry, remove it from the list, and free it.
+                        */
+                       if (nfsrv_fsevents_enabled) {
+                               fp->fm_context.vc_thread = current_thread();
+                               add_fsevent(FSE_CONTENT_MODIFIED, &fp->fm_context,
+                                               FSE_ARG_VNODE, fp->fm_vp,
+                                               FSE_ARG_DONE);
+                       }
+                       vnode_put(fp->fm_vp);
+                       kauth_cred_unref(&fp->fm_context.vc_ucred);
+                       nfp = LIST_NEXT(fp, fm_link);
+                       LIST_REMOVE(fp, fm_link);
+                       FREE(fp, M_TEMP);
+               }
        }
        }
+       nfsrv_fmod_pending = 0;
+       lck_mtx_unlock(nfsrv_fmod_mutex);
+#endif
 
 
-       TAILQ_INIT(&nfsd_head);
-       nfsd_head_flag &= ~NFSD_CHECKSLP;
+       nfsrv_uc_cleanup();     /* Stop nfs socket up-call threads */
+       
+       nfs_gss_svc_cleanup();  /* Remove any RPCSEC_GSS contexts */
 
 
-       MALLOC(nfs_udpsock, struct nfssvc_sock *, 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);
+       nfsrv_cleancache();     /* And clear out server cache */
 
 
-       MALLOC(nfs_cltpsock, struct nfssvc_sock *, 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);
+       nfsrv_udpsock = NULL;
+       nfsrv_udp6sock = NULL;
 }
 
 }
 
-/*
- * 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;
-       struct timeval now;
-
-       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;
-       microuptime(&now);
-       rt->resptime = ((now.tv_sec - nd->nd_starttime.tv_sec) * 1000000) +
-               (now.tv_usec - nd->nd_starttime.tv_usec);
-       microtime(&rt->tstamp); // XXX unused
-       nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ;
-}
 #endif /* NFS_NOSERVER */
 #endif /* NFS_NOSERVER */