/*-
 * Copyright (c) 1982, 1986, 1988, 1990, 1993
 *	The Regents of the University of California.
 * Copyright (c) 2004 The FreeBSD Foundation
 * Copyright (c) 2004-2008 Robert N. M. Watson
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)uipc_socket.c	8.3 (Berkeley) 4/15/94
 */

/*
 * Comments on the socket life cycle:
 *
 * soalloc() sets of socket layer state for a socket, called only by
 * socreate() and sonewconn().  Socket layer private.
 *
 * sodealloc() tears down socket layer state for a socket, called only by
 * sofree() and sonewconn().  Socket layer private.
 *
 * pru_attach() associates protocol layer state with an allocated socket;
 * called only once, may fail, aborting socket allocation.  This is called
 * from socreate() and sonewconn().  Socket layer private.
 *
 * pru_detach() disassociates protocol layer state from an attached socket,
 * and will be called exactly once for sockets in which pru_attach() has
 * been successfully called.  If pru_attach() returned an error,
 * pru_detach() will not be called.  Socket layer private.
 *
 * pru_abort() and pru_close() notify the protocol layer that the last
 * consumer of a socket is starting to tear down the socket, and that the
 * protocol should terminate the connection.  Historically, pru_abort() also
 * detached protocol state from the socket state, but this is no longer the
 * case.
 *
 * socreate() creates a socket and attaches protocol state.  This is a public
 * interface that may be used by socket layer consumers to create new
 * sockets.
 *
 * sonewconn() creates a socket and attaches protocol state.  This is a
 * public interface  that may be used by protocols to create new sockets when
 * a new connection is received and will be available for accept() on a
 * listen socket.
 *
 * soclose() destroys a socket after possibly waiting for it to disconnect.
 * This is a public interface that socket consumers should use to close and
 * release a socket when done with it.
 *
 * soabort() destroys a socket without waiting for it to disconnect (used
 * only for incoming connections that are already partially or fully
 * connected).  This is used internally by the socket layer when clearing
 * listen socket queues (due to overflow or close on the listen socket), but
 * is also a public interface protocols may use to abort connections in
 * their incomplete listen queues should they no longer be required.  Sockets
 * placed in completed connection listen queues should not be aborted for
 * reasons described in the comment above the soclose() implementation.  This
 * is not a general purpose close routine, and except in the specific
 * circumstances described here, should not be used.
 *
 * sofree() will free a socket and its protocol state if all references on
 * the socket have been released, and is the public interface to attempt to
 * free a socket when a reference is removed.  This is a socket layer private
 * interface.
 *
 * NOTE: In addition to socreate() and soclose(), which provide a single
 * socket reference to the consumer to be managed as required, there are two
 * calls to explicitly manage socket references, soref(), and sorele().
 * Currently, these are generally required only when transitioning a socket
 * from a listen queue to a file descriptor, in order to prevent garbage
 * collection of the socket at an untimely moment.  For a number of reasons,
 * these interfaces are not preferred, and should be avoided.
 * 
 * NOTE: With regard to VNETs the general rule is that callers do not set
 * curvnet. Exceptions to this rule include soabort(), sodisconnect(),
 * sofree() (and with that sorele(), sotryfree()), as well as sonewconn()
 * and sorflush(), which are usually called from a pre-set VNET context.
 * sopoll() currently does not need a VNET context to be set.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD: release/9.1.0/sys/kern/uipc_socket.c 233353 2012-03-23 11:26:54Z kib $");

#include "opt_inet.h"
#include "opt_inet6.h"
#include "opt_zero.h"
#include "opt_compat.h"
#include "opt_passiveinet.h"
#include "opt_promiscinet.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/fcntl.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/mac.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/mutex.h>
#include <sys/domain.h>
#include <sys/file.h>			/* for struct knote */
#include <sys/kernel.h>
#include <sys/event.h>
#include <sys/eventhandler.h>
#include <sys/poll.h>
#include <sys/proc.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/resourcevar.h>
#include <net/route.h>
#include <sys/signalvar.h>
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/uio.h>
#include <sys/jail.h>

#include <net/vnet.h>

#ifdef MAC
#include <security/mac/mac_framework.h>
#endif /* MAC */

#ifdef PROMISCUOUS_INET
#include <netinet/in_promisc.h>
#endif

#include <vm/uma.h>

#ifdef COMPAT_FREEBSD32
#include <sys/mount.h>
#include <sys/sysent.h>
#include <compat/freebsd32/freebsd32.h>
#endif

static int	soreceive_rcvoob(struct socket *so, struct uio *uio,
		    int flags);

static void	filt_sordetach(struct knote *kn);
static int	filt_soread(struct knote *kn, long hint);
static void	filt_sowdetach(struct knote *kn);
static int	filt_sowrite(struct knote *kn, long hint);
static int	filt_solisten(struct knote *kn, long hint);

static struct filterops solisten_filtops = {
	.f_isfd = 1,
	.f_detach = filt_sordetach,
	.f_event = filt_solisten,
};
static struct filterops soread_filtops = {
	.f_isfd = 1,
	.f_detach = filt_sordetach,
	.f_event = filt_soread,
};
static struct filterops sowrite_filtops = {
	.f_isfd = 1,
	.f_detach = filt_sowdetach,
	.f_event = filt_sowrite,
};

uma_zone_t socket_zone;
so_gen_t	so_gencnt;	/* generation count for sockets */

int	maxsockets;

MALLOC_DEFINE(M_SONAME, "soname", "socket name");
MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");

#define	VNET_SO_ASSERT(so)						\
	VNET_ASSERT(curvnet != NULL,					\
	    ("%s:%d curvnet is NULL, so=%p", __func__, __LINE__, (so)));

static int somaxconn = SOMAXCONN;
static int sysctl_somaxconn(SYSCTL_HANDLER_ARGS);
/* XXX: we dont have SYSCTL_USHORT */
SYSCTL_PROC(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLTYPE_UINT | CTLFLAG_RW,
    0, sizeof(int), sysctl_somaxconn, "I", "Maximum pending socket connection "
    "queue size");
static int numopensockets;
SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD,
    &numopensockets, 0, "Number of open sockets");
#ifdef ZERO_COPY_SOCKETS
/* These aren't static because they're used in other files. */
int so_zero_copy_send = 1;
int so_zero_copy_receive = 1;
SYSCTL_NODE(_kern_ipc, OID_AUTO, zero_copy, CTLFLAG_RD, 0,
    "Zero copy controls");
SYSCTL_INT(_kern_ipc_zero_copy, OID_AUTO, receive, CTLFLAG_RW,
    &so_zero_copy_receive, 0, "Enable zero copy receive");
SYSCTL_INT(_kern_ipc_zero_copy, OID_AUTO, send, CTLFLAG_RW,
    &so_zero_copy_send, 0, "Enable zero copy send");
#endif /* ZERO_COPY_SOCKETS */

/*
 * accept_mtx locks down per-socket fields relating to accept queues.  See
 * socketvar.h for an annotation of the protected fields of struct socket.
 */
struct mtx accept_mtx;
MTX_SYSINIT(accept_mtx, &accept_mtx, "accept", MTX_DEF);

/*
 * so_global_mtx protects so_gencnt, numopensockets, and the per-socket
 * so_gencnt field.
 */
static struct mtx so_global_mtx;
MTX_SYSINIT(so_global_mtx, &so_global_mtx, "so_glabel", MTX_DEF);

/*
 * General IPC sysctl name space, used by sockets and a variety of other IPC
 * types.
 */
SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");

/*
 * Sysctl to get and set the maximum global sockets limit.  Notify protocols
 * of the change so that they can update their dependent limits as required.
 */
static int
sysctl_maxsockets(SYSCTL_HANDLER_ARGS)
{
	int error, newmaxsockets;

	newmaxsockets = maxsockets;
	error = sysctl_handle_int(oidp, &newmaxsockets, 0, req);
	if (error == 0 && req->newptr) {
		if (newmaxsockets > maxsockets) {
			maxsockets = newmaxsockets;
			if (maxsockets > ((maxfiles / 4) * 3)) {
				maxfiles = (maxsockets * 5) / 4;
				maxfilesperproc = (maxfiles * 9) / 10;
			}
			EVENTHANDLER_INVOKE(maxsockets_change);
		} else
			error = EINVAL;
	}
	return (error);
}

SYSCTL_PROC(_kern_ipc, OID_AUTO, maxsockets, CTLTYPE_INT|CTLFLAG_RW,
    &maxsockets, 0, sysctl_maxsockets, "IU",
    "Maximum number of sockets avaliable");

/*
 * Initialise maxsockets.  This SYSINIT must be run after
 * tunable_mbinit().
 */
static void
init_maxsockets(void *ignored)
{

	TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
	maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
}
SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);

/*
 * Socket operation routines.  These routines are called by the routines in
 * sys_socket.c or from a system process, and implement the semantics of
 * socket operations by switching out to the protocol specific routines.
 */

/*
 * Get a socket structure from our zone, and initialize it.  Note that it
 * would probably be better to allocate socket and PCB at the same time, but
 * I'm not convinced that all the protocols can be easily modified to do
 * this.
 *
 * soalloc() returns a socket with a ref count of 0.
 */
static struct socket *
soalloc(struct vnet *vnet)
{
	struct socket *so;

	so = uma_zalloc(socket_zone, M_NOWAIT | M_ZERO);
	if (so == NULL)
		return (NULL);
#ifdef MAC
	if (mac_socket_init(so, M_NOWAIT) != 0) {
		uma_zfree(socket_zone, so);
		return (NULL);
	}
#endif
#ifdef PROMISCUOUS_INET
	if (in_promisc_socket_init(so, M_NOWAIT) != 0) {
#ifdef MAC
		mac_socket_destroy(so);
#endif
		uma_zfree(socket_zone, so);
		return (NULL);
	}
#endif /* PROMISCUOUS_INET */
	SOCKBUF_LOCK_INIT(&so->so_snd, "so_snd");
	SOCKBUF_LOCK_INIT(&so->so_rcv, "so_rcv");
	sx_init(&so->so_snd.sb_sx, "so_snd_sx");
	sx_init(&so->so_rcv.sb_sx, "so_rcv_sx");
	TAILQ_INIT(&so->so_aiojobq);
	mtx_lock(&so_global_mtx);
	so->so_gencnt = ++so_gencnt;
	++numopensockets;
#ifdef VIMAGE
	VNET_ASSERT(vnet != NULL, ("%s:%d vnet is NULL, so=%p",
	    __func__, __LINE__, so));
	vnet->vnet_sockcnt++;
	so->so_vnet = vnet;
#endif
	mtx_unlock(&so_global_mtx);
	return (so);
}

/*
 * Free the storage associated with a socket at the socket layer, tear down
 * locks, labels, etc.  All protocol state is assumed already to have been
 * torn down (and possibly never set up) by the caller.
 */
static void
sodealloc(struct socket *so)
{

	KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count));
	KASSERT(so->so_pcb == NULL, ("sodealloc(): so_pcb != NULL"));

	mtx_lock(&so_global_mtx);
	so->so_gencnt = ++so_gencnt;
	--numopensockets;	/* Could be below, but faster here. */
#ifdef VIMAGE
	VNET_ASSERT(so->so_vnet != NULL, ("%s:%d so_vnet is NULL, so=%p",
	    __func__, __LINE__, so));
	so->so_vnet->vnet_sockcnt--;
#endif
	mtx_unlock(&so_global_mtx);
	if (so->so_rcv.sb_hiwat)
		(void)chgsbsize(so->so_cred->cr_uidinfo,
		    &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
	if (so->so_snd.sb_hiwat)
		(void)chgsbsize(so->so_cred->cr_uidinfo,
		    &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
#ifdef INET
	/* remove acccept filter if one is present. */
	if (so->so_accf != NULL)
		do_setopt_accept_filter(so, NULL);
#endif
#ifdef PROMISCUOUS_INET
	in_promisc_socket_destroy(so);
#endif
#ifdef MAC
	mac_socket_destroy(so);
#endif
	crfree(so->so_cred);
	sx_destroy(&so->so_snd.sb_sx);
	sx_destroy(&so->so_rcv.sb_sx);
	SOCKBUF_LOCK_DESTROY(&so->so_snd);
	SOCKBUF_LOCK_DESTROY(&so->so_rcv);
	uma_zfree(socket_zone, so);
}

/*
 * socreate returns a socket with a ref count of 1.  The socket should be
 * closed with soclose().
 */
int
socreate(int dom, struct socket **aso, int type, int proto,
    struct ucred *cred, struct thread *td)
{
	struct protosw *prp;
	struct socket *so;
	int error;

	if (proto)
		prp = pffindproto(dom, proto, type);
	else
		prp = pffindtype(dom, type);

	if (prp == NULL || prp->pr_usrreqs->pru_attach == NULL ||
	    prp->pr_usrreqs->pru_attach == pru_attach_notsupp)
		return (EPROTONOSUPPORT);

	if (prison_check_af(cred, prp->pr_domain->dom_family) != 0)
		return (EPROTONOSUPPORT);

	if (prp->pr_type != type)
		return (EPROTOTYPE);
	so = soalloc(CRED_TO_VNET(cred));
	if (so == NULL)
		return (ENOBUFS);

	TAILQ_INIT(&so->so_incomp);
	TAILQ_INIT(&so->so_comp);
	so->so_type = type;
	so->so_cred = crhold(cred);
	if ((prp->pr_domain->dom_family == PF_INET) ||
	    (prp->pr_domain->dom_family == PF_INET6) ||
	    (prp->pr_domain->dom_family == PF_ROUTE))
		so->so_fibnum = td->td_proc->p_fibnum;
	else
		so->so_fibnum = 0;
	so->so_proto = prp;
#ifdef MAC
	mac_socket_create(cred, so);
#endif
	knlist_init_mtx(&so->so_rcv.sb_sel.si_note, SOCKBUF_MTX(&so->so_rcv));
	knlist_init_mtx(&so->so_snd.sb_sel.si_note, SOCKBUF_MTX(&so->so_snd));
	so->so_count = 1;
	/*
	 * Auto-sizing of socket buffers is managed by the protocols and
	 * the appropriate flags must be set in the pru_attach function.
	 */
	CURVNET_SET(so->so_vnet);
	error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
	CURVNET_RESTORE();
	if (error) {
		KASSERT(so->so_count == 1, ("socreate: so_count %d",
		    so->so_count));
		so->so_count = 0;
		sodealloc(so);
		return (error);
	}
	*aso = so;
	return (0);
}

#ifdef REGRESSION
static int regression_sonewconn_earlytest = 1;
SYSCTL_INT(_regression, OID_AUTO, sonewconn_earlytest, CTLFLAG_RW,
    &regression_sonewconn_earlytest, 0, "Perform early sonewconn limit test");
#endif

/*
 * When an attempt at a new connection is noted on a socket which accepts
 * connections, sonewconn is called.  If the connection is possible (subject
 * to space constraints, etc.) then we allocate a new structure, properly
 * linked into the data structure of the original socket, and return this.
 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
 *
 * Note: the ref count on the socket is 0 on return.
 */
struct socket *
sonewconn(struct socket *head, int connstatus)
{
	struct socket *so;
	int over;

	ACCEPT_LOCK();
	over = (head->so_qlen > 3 * head->so_qlimit / 2);
	ACCEPT_UNLOCK();
#ifdef REGRESSION
	if (regression_sonewconn_earlytest && over)
#else
	if (over)
#endif
		return (NULL);
	VNET_ASSERT(head->so_vnet != NULL, ("%s:%d so_vnet is NULL, head=%p",
	    __func__, __LINE__, head));
	so = soalloc(head->so_vnet);
	if (so == NULL)
		return (NULL);
	if ((head->so_options & SO_ACCEPTFILTER) != 0)
		connstatus = 0;
	so->so_head = head;
	so->so_type = head->so_type;
	so->so_options = head->so_options &~ SO_ACCEPTCONN;
	so->so_linger = head->so_linger;
	so->so_state = head->so_state | SS_NOFDREF;
	so->so_fibnum = head->so_fibnum;
	so->so_proto = head->so_proto;
	so->so_cred = crhold(head->so_cred);
#ifdef MAC
	mac_socket_newconn(head, so);
#endif
#ifdef PROMISCUOUS_INET
	in_promisc_socket_newconn(head, so);
#endif
	knlist_init_mtx(&so->so_rcv.sb_sel.si_note, SOCKBUF_MTX(&so->so_rcv));
	knlist_init_mtx(&so->so_snd.sb_sel.si_note, SOCKBUF_MTX(&so->so_snd));
	VNET_SO_ASSERT(head);
	if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
	    (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
		sodealloc(so);
		return (NULL);
	}
	so->so_rcv.sb_lowat = head->so_rcv.sb_lowat;
	so->so_snd.sb_lowat = head->so_snd.sb_lowat;
	so->so_rcv.sb_timeo = head->so_rcv.sb_timeo;
	so->so_snd.sb_timeo = head->so_snd.sb_timeo;
	so->so_rcv.sb_flags |= head->so_rcv.sb_flags & SB_AUTOSIZE;
	so->so_snd.sb_flags |= head->so_snd.sb_flags & SB_AUTOSIZE;
	so->so_state |= connstatus;
	ACCEPT_LOCK();
	if (connstatus) {
		TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
		so->so_qstate |= SQ_COMP;
		head->so_qlen++;
	} else {
		/*
		 * Keep removing sockets from the head until there's room for
		 * us to insert on the tail.  In pre-locking revisions, this
		 * was a simple if(), but as we could be racing with other
		 * threads and soabort() requires dropping locks, we must
		 * loop waiting for the condition to be true.
		 */
		while (head->so_incqlen > head->so_qlimit) {
			struct socket *sp;
			sp = TAILQ_FIRST(&head->so_incomp);
			TAILQ_REMOVE(&head->so_incomp, sp, so_list);
			head->so_incqlen--;
			sp->so_qstate &= ~SQ_INCOMP;
			sp->so_head = NULL;
			ACCEPT_UNLOCK();
			soabort(sp);
			ACCEPT_LOCK();
		}
		TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
		so->so_qstate |= SQ_INCOMP;
		head->so_incqlen++;
	}
	ACCEPT_UNLOCK();
	if (connstatus) {
		sorwakeup(head);
		wakeup_one(&head->so_timeo);
	}
	return (so);
}

#ifdef PASSIVE_INET
/*
 * When a new connection is completed on a listening socket that is
 * configured for passive reassembly, sonewconn_passive_client is called to
 * create a socket representing the client side of the reassembled
 * connection.  We allocate a new structure, inherit configuration from the
 * listening socket, and return this.
 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTING.
 *
 * Note: the ref count on the socket is 1 on return.
 */
struct socket *
sonewconn_passive_client(struct socket *head, int connstatus)
{
	struct socket *so;

	VNET_ASSERT(head->so_vnet != NULL, ("%s:%d so_vnet is NULL, head=%p",
	    __func__, __LINE__, head));
	so = soalloc(head->so_vnet);
	if (so == NULL)
		return (NULL);
	so->so_head = NULL; /* just inheriting from head, not otherwise associating */
	so->so_type = head->so_type;
	so->so_options = head->so_options &~ SO_ACCEPTCONN;
	so->so_linger = head->so_linger;
	so->so_state = head->so_state | SS_NOFDREF;
	so->so_fibnum = head->so_fibnum;
	so->so_proto = head->so_proto;
	so->so_cred = crhold(head->so_cred);
#ifdef MAC
	mac_socket_newconn(head, so);
#endif
#ifdef PROMISCUOUS_INET
	in_promisc_socket_newconn(head, so);
#endif
	knlist_init_mtx(&so->so_rcv.sb_sel.si_note, SOCKBUF_MTX(&so->so_rcv));
	knlist_init_mtx(&so->so_snd.sb_sel.si_note, SOCKBUF_MTX(&so->so_snd));
	VNET_SO_ASSERT(head);
	if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
	    (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
		sodealloc(so);
		return (NULL);
	}
	so->so_rcv.sb_lowat = head->so_rcv.sb_lowat;
	so->so_snd.sb_lowat = head->so_snd.sb_lowat;
	so->so_rcv.sb_timeo = head->so_rcv.sb_timeo;
	so->so_snd.sb_timeo = head->so_snd.sb_timeo;
	so->so_rcv.sb_flags |= head->so_rcv.sb_flags & SB_AUTOSIZE;
	so->so_snd.sb_flags |= head->so_snd.sb_flags & SB_AUTOSIZE;
	so->so_state |= connstatus;

	return (so);
}
#endif /* PASSIVE_INET */

int
sobind(struct socket *so, struct sockaddr *nam, struct thread *td)
{
	int error;

	CURVNET_SET(so->so_vnet);
	error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
	CURVNET_RESTORE();
	return error;
}

/*
 * solisten() transitions a socket from a non-listening state to a listening
 * state, but can also be used to update the listen queue depth on an
 * existing listen socket.  The protocol will call back into the sockets
 * layer using solisten_proto_check() and solisten_proto() to check and set
 * socket-layer listen state.  Call backs are used so that the protocol can
 * acquire both protocol and socket layer locks in whatever order is required
 * by the protocol.
 *
 * Protocol implementors are advised to hold the socket lock across the
 * socket-layer test and set to avoid races at the socket layer.
 */
int
solisten(struct socket *so, int backlog, struct thread *td)
{
	int error;

	CURVNET_SET(so->so_vnet);
	error = (*so->so_proto->pr_usrreqs->pru_listen)(so, backlog, td);
	CURVNET_RESTORE();
	return error;
}

int
solisten_proto_check(struct socket *so)
{

	SOCK_LOCK_ASSERT(so);

	if (so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
	    SS_ISDISCONNECTING))
		return (EINVAL);
	return (0);
}

void
solisten_proto(struct socket *so, int backlog)
{

	SOCK_LOCK_ASSERT(so);

	if (backlog < 0 || backlog > somaxconn)
		backlog = somaxconn;
	so->so_qlimit = backlog;
	so->so_options |= SO_ACCEPTCONN;
}

/*
 * Evaluate the reference count and named references on a socket; if no
 * references remain, free it.  This should be called whenever a reference is
 * released, such as in sorele(), but also when named reference flags are
 * cleared in socket or protocol code.
 *
 * sofree() will free the socket if:
 *
 * - There are no outstanding file descriptor references or related consumers
 *   (so_count == 0).
 *
 * - The socket has been closed by user space, if ever open (SS_NOFDREF).
 *
 * - The protocol does not have an outstanding strong reference on the socket
 *   (SS_PROTOREF).
 *
 * - The socket is not in a completed connection queue, so a process has been
 *   notified that it is present.  If it is removed, the user process may
 *   block in accept() despite select() saying the socket was ready.
 */
void
sofree(struct socket *so)
{
	struct protosw *pr = so->so_proto;
	struct socket *head;

	ACCEPT_LOCK_ASSERT();
	SOCK_LOCK_ASSERT(so);

	if ((so->so_state & SS_NOFDREF) == 0 || so->so_count != 0 ||
	    (so->so_state & SS_PROTOREF) || (so->so_qstate & SQ_COMP)) {
		SOCK_UNLOCK(so);
		ACCEPT_UNLOCK();
		return;
	}

	head = so->so_head;
	if (head != NULL) {
		KASSERT((so->so_qstate & SQ_COMP) != 0 ||
		    (so->so_qstate & SQ_INCOMP) != 0,
		    ("sofree: so_head != NULL, but neither SQ_COMP nor "
		    "SQ_INCOMP"));
		KASSERT((so->so_qstate & SQ_COMP) == 0 ||
		    (so->so_qstate & SQ_INCOMP) == 0,
		    ("sofree: so->so_qstate is SQ_COMP and also SQ_INCOMP"));
		TAILQ_REMOVE(&head->so_incomp, so, so_list);
		head->so_incqlen--;
		so->so_qstate &= ~SQ_INCOMP;
		so->so_head = NULL;
	}
	KASSERT((so->so_qstate & SQ_COMP) == 0 &&
	    (so->so_qstate & SQ_INCOMP) == 0,
	    ("sofree: so_head == NULL, but still SQ_COMP(%d) or SQ_INCOMP(%d)",
	    so->so_qstate & SQ_COMP, so->so_qstate & SQ_INCOMP));
	if (so->so_options & SO_ACCEPTCONN) {
		KASSERT((TAILQ_EMPTY(&so->so_comp)), ("sofree: so_comp populated"));
		KASSERT((TAILQ_EMPTY(&so->so_incomp)), ("sofree: so_comp populated"));
	}
	SOCK_UNLOCK(so);
	ACCEPT_UNLOCK();

	VNET_SO_ASSERT(so);
	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose != NULL)
		(*pr->pr_domain->dom_dispose)(so->so_rcv.sb_mb);
	if (pr->pr_usrreqs->pru_detach != NULL)
		(*pr->pr_usrreqs->pru_detach)(so);

	/*
	 * From this point on, we assume that no other references to this
	 * socket exist anywhere else in the stack.  Therefore, no locks need
	 * to be acquired or held.
	 *
	 * We used to do a lot of socket buffer and socket locking here, as
	 * well as invoke sorflush() and perform wakeups.  The direct call to
	 * dom_dispose() and sbrelease_internal() are an inlining of what was
	 * necessary from sorflush().
	 *
	 * Notice that the socket buffer and kqueue state are torn down
	 * before calling pru_detach.  This means that protocols shold not
	 * assume they can perform socket wakeups, etc, in their detach code.
	 */
	sbdestroy(&so->so_snd, so);
	sbdestroy(&so->so_rcv, so);
	seldrain(&so->so_snd.sb_sel);
	seldrain(&so->so_rcv.sb_sel);
	knlist_destroy(&so->so_rcv.sb_sel.si_note);
	knlist_destroy(&so->so_snd.sb_sel.si_note);
	sodealloc(so);
}

/*
 * Close a socket on last file table reference removal.  Initiate disconnect
 * if connected.  Free socket when disconnect complete.
 *
 * This function will sorele() the socket.  Note that soclose() may be called
 * prior to the ref count reaching zero.  The actual socket structure will
 * not be freed until the ref count reaches zero.
 */
int
soclose(struct socket *so)
{
	int error = 0;

	KASSERT(!(so->so_state & SS_NOFDREF), ("soclose: SS_NOFDREF on enter"));

	CURVNET_SET(so->so_vnet);
	funsetown(&so->so_sigio);
	if (so->so_state & SS_ISCONNECTED) {
		if ((so->so_state & SS_ISDISCONNECTING) == 0) {
			error = sodisconnect(so);
			if (error) {
				if (error == ENOTCONN)
					error = 0;
				goto drop;
			}
		}
		if (so->so_options & SO_LINGER) {
			if ((so->so_state & SS_ISDISCONNECTING) &&
			    (so->so_state & SS_NBIO))
				goto drop;
			while (so->so_state & SS_ISCONNECTED) {
				error = tsleep(&so->so_timeo,
				    PSOCK | PCATCH, "soclos", so->so_linger * hz);
				if (error)
					break;
			}
		}
	}

drop:
	if (so->so_proto->pr_usrreqs->pru_close != NULL)
		(*so->so_proto->pr_usrreqs->pru_close)(so);
	if (so->so_options & SO_ACCEPTCONN) {
		struct socket *sp;
		ACCEPT_LOCK();
		while ((sp = TAILQ_FIRST(&so->so_incomp)) != NULL) {
			TAILQ_REMOVE(&so->so_incomp, sp, so_list);
			so->so_incqlen--;
			sp->so_qstate &= ~SQ_INCOMP;
			sp->so_head = NULL;
			ACCEPT_UNLOCK();
			soabort(sp);
			ACCEPT_LOCK();
		}
		while ((sp = TAILQ_FIRST(&so->so_comp)) != NULL) {
			TAILQ_REMOVE(&so->so_comp, sp, so_list);
			so->so_qlen--;
			sp->so_qstate &= ~SQ_COMP;
			sp->so_head = NULL;
			ACCEPT_UNLOCK();
			soabort(sp);
			ACCEPT_LOCK();
		}
		ACCEPT_UNLOCK();
	}
	ACCEPT_LOCK();
	SOCK_LOCK(so);
	KASSERT((so->so_state & SS_NOFDREF) == 0, ("soclose: NOFDREF"));
	so->so_state |= SS_NOFDREF;
	sorele(so);
	CURVNET_RESTORE();
	return (error);
}

/*
 * soabort() is used to abruptly tear down a connection, such as when a
 * resource limit is reached (listen queue depth exceeded), or if a listen
 * socket is closed while there are sockets waiting to be accepted.
 *
 * This interface is tricky, because it is called on an unreferenced socket,
 * and must be called only by a thread that has actually removed the socket
 * from the listen queue it was on, or races with other threads are risked.
 *
 * This interface will call into the protocol code, so must not be called
 * with any socket locks held.  Protocols do call it while holding their own
 * recursible protocol mutexes, but this is something that should be subject
 * to review in the future.
 */
void
soabort(struct socket *so)
{

	/*
	 * In as much as is possible, assert that no references to this
	 * socket are held.  This is not quite the same as asserting that the
	 * current thread is responsible for arranging for no references, but
	 * is as close as we can get for now.
	 */
	KASSERT(so->so_count == 0, ("soabort: so_count"));
	KASSERT((so->so_state & SS_PROTOREF) == 0, ("soabort: SS_PROTOREF"));
	KASSERT(so->so_state & SS_NOFDREF, ("soabort: !SS_NOFDREF"));
	KASSERT((so->so_state & SQ_COMP) == 0, ("soabort: SQ_COMP"));
	KASSERT((so->so_state & SQ_INCOMP) == 0, ("soabort: SQ_INCOMP"));
	VNET_SO_ASSERT(so);

	if (so->so_proto->pr_usrreqs->pru_abort != NULL)
		(*so->so_proto->pr_usrreqs->pru_abort)(so);
	ACCEPT_LOCK();
	SOCK_LOCK(so);
	sofree(so);
}

int
soaccept(struct socket *so, struct sockaddr **nam)
{
	int error;

	SOCK_LOCK(so);
	KASSERT((so->so_state & SS_NOFDREF) != 0, ("soaccept: !NOFDREF"));
	so->so_state &= ~SS_NOFDREF;
	SOCK_UNLOCK(so);

	CURVNET_SET(so->so_vnet);
	error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
	CURVNET_RESTORE();
	return (error);
}

int
soconnect(struct socket *so, struct sockaddr *nam, struct thread *td)
{
	int error;

	if (so->so_options & SO_ACCEPTCONN)
		return (EOPNOTSUPP);

	CURVNET_SET(so->so_vnet);
	/*
	 * If protocol is connection-based, can only connect once.
	 * Otherwise, if connected, try to disconnect first.  This allows
	 * user to disconnect by connecting to, e.g., a null address.
	 */
	if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
	    (error = sodisconnect(so)))) {
		error = EISCONN;
	} else {
		/*
		 * Prevent accumulated error from previous connection from
		 * biting us.
		 */
		so->so_error = 0;
		error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
	}
	CURVNET_RESTORE();

	return (error);
}

int
soconnect2(struct socket *so1, struct socket *so2)
{
	int error;

	CURVNET_SET(so1->so_vnet);
	error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
	CURVNET_RESTORE();
	return (error);
}

int
sodisconnect(struct socket *so)
{
	int error;

	if ((so->so_state & SS_ISCONNECTED) == 0)
		return (ENOTCONN);
	if (so->so_state & SS_ISDISCONNECTING)
		return (EALREADY);
	VNET_SO_ASSERT(so);
	error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
	return (error);
}

#ifdef ZERO_COPY_SOCKETS
struct so_zerocopy_stats{
	int size_ok;
	int align_ok;
	int found_ifp;
};
struct so_zerocopy_stats so_zerocp_stats = {0,0,0};
#include <netinet/in.h>
#include <net/route.h>
#include <netinet/in_pcb.h>
#include <vm/vm.h>
#include <vm/vm_page.h>
#include <vm/vm_object.h>

/*
 * sosend_copyin() is only used if zero copy sockets are enabled.  Otherwise
 * sosend_dgram() and sosend_generic() use m_uiotombuf().
 * 
 * sosend_copyin() accepts a uio and prepares an mbuf chain holding part or
 * all of the data referenced by the uio.  If desired, it uses zero-copy.
 * *space will be updated to reflect data copied in.
 *
 * NB: If atomic I/O is requested, the caller must already have checked that
 * space can hold resid bytes.
 *
 * NB: In the event of an error, the caller may need to free the partial
 * chain pointed to by *mpp.  The contents of both *uio and *space may be
 * modified even in the case of an error.
 */
static int
sosend_copyin(struct uio *uio, struct mbuf **retmp, int atomic, long *space,
    int flags)
{
	struct mbuf *m, **mp, *top;
	long len;
	ssize_t resid;
	int error;
#ifdef ZERO_COPY_SOCKETS
	int cow_send;
#endif

	*retmp = top = NULL;
	mp = &top;
	len = 0;
	resid = uio->uio_resid;
	error = 0;
	do {
#ifdef ZERO_COPY_SOCKETS
		cow_send = 0;
#endif /* ZERO_COPY_SOCKETS */
		if (resid >= MINCLSIZE) {
#ifdef ZERO_COPY_SOCKETS
			if (top == NULL) {
				m = m_gethdr(M_WAITOK, MT_DATA);
				m->m_pkthdr.len = 0;
				m->m_pkthdr.rcvif = NULL;
			} else
				m = m_get(M_WAITOK, MT_DATA);
			if (so_zero_copy_send &&
			    resid>=PAGE_SIZE &&
			    *space>=PAGE_SIZE &&
			    uio->uio_iov->iov_len>=PAGE_SIZE) {
				so_zerocp_stats.size_ok++;
				so_zerocp_stats.align_ok++;
				cow_send = socow_setup(m, uio);
				len = cow_send;
			}
			if (!cow_send) {
				m_clget(m, M_WAITOK);
				len = min(min(MCLBYTES, resid), *space);
			}
#else /* ZERO_COPY_SOCKETS */
			if (top == NULL) {
				m = m_getcl(M_WAIT, MT_DATA, M_PKTHDR);
				m->m_pkthdr.len = 0;
				m->m_pkthdr.rcvif = NULL;
			} else
				m = m_getcl(M_WAIT, MT_DATA, 0);
			len = min(min(MCLBYTES, resid), *space);
#endif /* ZERO_COPY_SOCKETS */
		} else {
			if (top == NULL) {
				m = m_gethdr(M_WAIT, MT_DATA);
				m->m_pkthdr.len = 0;
				m->m_pkthdr.rcvif = NULL;

				len = min(min(MHLEN, resid), *space);
				/*
				 * For datagram protocols, leave room
				 * for protocol headers in first mbuf.
				 */
				if (atomic && m && len < MHLEN)
					MH_ALIGN(m, len);
			} else {
				m = m_get(M_WAIT, MT_DATA);
				len = min(min(MLEN, resid), *space);
			}
		}
		if (m == NULL) {
			error = ENOBUFS;
			goto out;
		}

		*space -= len;
#ifdef ZERO_COPY_SOCKETS
		if (cow_send)
			error = 0;
		else
#endif /* ZERO_COPY_SOCKETS */
		error = uiomove(mtod(m, void *), (int)len, uio);
		resid = uio->uio_resid;
		m->m_len = len;
		*mp = m;
		top->m_pkthdr.len += len;
		if (error)
			goto out;
		mp = &m->m_next;
		if (resid <= 0) {
			if (flags & MSG_EOR)
				top->m_flags |= M_EOR;
			break;
		}
	} while (*space > 0 && atomic);
out:
	*retmp = top;
	return (error);
}
#endif /*ZERO_COPY_SOCKETS*/

#define	SBLOCKWAIT(f)	(((f) & MSG_DONTWAIT) ? 0 : SBL_WAIT)

int
sosend_dgram(struct socket *so, struct sockaddr *addr, struct uio *uio,
    struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
{
	long space;
	ssize_t resid;
	int clen = 0, error, dontroute;
#ifdef ZERO_COPY_SOCKETS
	int atomic = sosendallatonce(so) || top;
#endif

	KASSERT(so->so_type == SOCK_DGRAM, ("sodgram_send: !SOCK_DGRAM"));
	KASSERT(so->so_proto->pr_flags & PR_ATOMIC,
	    ("sodgram_send: !PR_ATOMIC"));

	if (uio != NULL)
		resid = uio->uio_resid;
	else
		resid = top->m_pkthdr.len;
	/*
	 * In theory resid should be unsigned.  However, space must be
	 * signed, as it might be less than 0 if we over-committed, and we
	 * must use a signed comparison of space and resid.  On the other
	 * hand, a negative resid causes us to loop sending 0-length
	 * segments to the protocol.
	 */
	if (resid < 0) {
		error = EINVAL;
		goto out;
	}

	dontroute =
	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0;
	if (td != NULL)
		td->td_ru.ru_msgsnd++;
	if (control != NULL)
		clen = control->m_len;

	SOCKBUF_LOCK(&so->so_snd);
	if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
		SOCKBUF_UNLOCK(&so->so_snd);
		error = EPIPE;
		goto out;
	}
	if (so->so_error) {
		error = so->so_error;
		so->so_error = 0;
		SOCKBUF_UNLOCK(&so->so_snd);
		goto out;
	}
	if ((so->so_state & SS_ISCONNECTED) == 0) {
		/*
		 * `sendto' and `sendmsg' is allowed on a connection-based
		 * socket if it supports implied connect.  Return ENOTCONN if
		 * not connected and no address is supplied.
		 */
		if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
		    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
			if ((so->so_state & SS_ISCONFIRMING) == 0 &&
			    !(resid == 0 && clen != 0)) {
				SOCKBUF_UNLOCK(&so->so_snd);
				error = ENOTCONN;
				goto out;
			}
		} else if (addr == NULL) {
			if (so->so_proto->pr_flags & PR_CONNREQUIRED)
				error = ENOTCONN;
			else
				error = EDESTADDRREQ;
			SOCKBUF_UNLOCK(&so->so_snd);
			goto out;
		}
	}

	/*
	 * Do we need MSG_OOB support in SOCK_DGRAM?  Signs here may be a
	 * problem and need fixing.
	 */
	space = sbspace(&so->so_snd);
	if (flags & MSG_OOB)
		space += 1024;
	space -= clen;
	SOCKBUF_UNLOCK(&so->so_snd);
	if (resid > space) {
		error = EMSGSIZE;
		goto out;
	}
	if (uio == NULL) {
		resid = 0;
		if (flags & MSG_EOR)
			top->m_flags |= M_EOR;
	} else {
#ifdef ZERO_COPY_SOCKETS
		error = sosend_copyin(uio, &top, atomic, &space, flags);
		if (error)
			goto out;
#else
		/*
		 * Copy the data from userland into a mbuf chain.
		 * If no data is to be copied in, a single empty mbuf
		 * is returned.
		 */
		top = m_uiotombuf(uio, M_WAITOK, space, max_hdr,
		    (M_PKTHDR | ((flags & MSG_EOR) ? M_EOR : 0)));
		if (top == NULL) {
			error = EFAULT;	/* only possible error */
			goto out;
		}
		space -= resid - uio->uio_resid;
#endif
		resid = uio->uio_resid;
	}
	KASSERT(resid == 0, ("sosend_dgram: resid != 0"));
	/*
	 * XXXRW: Frobbing SO_DONTROUTE here is even worse without sblock
	 * than with.
	 */
	if (dontroute) {
		SOCK_LOCK(so);
		so->so_options |= SO_DONTROUTE;
		SOCK_UNLOCK(so);
	}
	/*
	 * XXX all the SBS_CANTSENDMORE checks previously done could be out
	 * of date.  We could have recieved a reset packet in an interrupt or
	 * maybe we slept while doing page faults in uiomove() etc.  We could
	 * probably recheck again inside the locking protection here, but
	 * there are probably other places that this also happens.  We must
	 * rethink this.
	 */
	VNET_SO_ASSERT(so);
	error = (*so->so_proto->pr_usrreqs->pru_send)(so,
	    (flags & MSG_OOB) ? PRUS_OOB :
	/*
	 * If the user set MSG_EOF, the protocol understands this flag and
	 * nothing left to send then use PRU_SEND_EOF instead of PRU_SEND.
	 */
	    ((flags & MSG_EOF) &&
	     (so->so_proto->pr_flags & PR_IMPLOPCL) &&
	     (resid <= 0)) ?
		PRUS_EOF :
		/* If there is more to send set PRUS_MORETOCOME */
		(resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
		top, addr, control, td);
	if (dontroute) {
		SOCK_LOCK(so);
		so->so_options &= ~SO_DONTROUTE;
		SOCK_UNLOCK(so);
	}
	clen = 0;
	control = NULL;
	top = NULL;
out:
	if (top != NULL)
		m_freem(top);
	if (control != NULL)
		m_freem(control);
	return (error);
}

/*
 * Send on a socket.  If send must go all at once and message is larger than
 * send buffering, then hard error.  Lock against other senders.  If must go
 * all at once and not enough room now, then inform user that this would
 * block and do nothing.  Otherwise, if nonblocking, send as much as
 * possible.  The data to be sent is described by "uio" if nonzero, otherwise
 * by the mbuf chain "top" (which must be null if uio is not).  Data provided
 * in mbuf chain must be small enough to send all at once.
 *
 * Returns nonzero on error, timeout or signal; callers must check for short
 * counts if EINTR/ERESTART are returned.  Data and control buffers are freed
 * on return.
 */
int
sosend_generic(struct socket *so, struct sockaddr *addr, struct uio *uio,
    struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
{
	long space;
	ssize_t resid;
	int clen = 0, error, dontroute;
	int atomic = sosendallatonce(so) || top;

	if (uio != NULL)
		resid = uio->uio_resid;
	else
		resid = top->m_pkthdr.len;
	/*
	 * In theory resid should be unsigned.  However, space must be
	 * signed, as it might be less than 0 if we over-committed, and we
	 * must use a signed comparison of space and resid.  On the other
	 * hand, a negative resid causes us to loop sending 0-length
	 * segments to the protocol.
	 *
	 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
	 * type sockets since that's an error.
	 */
	if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
		error = EINVAL;
		goto out;
	}

	dontroute =
	    (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
	    (so->so_proto->pr_flags & PR_ATOMIC);
	if (td != NULL)
		td->td_ru.ru_msgsnd++;
	if (control != NULL)
		clen = control->m_len;

	error = sblock(&so->so_snd, SBLOCKWAIT(flags));
	if (error)
		goto out;

restart:
	do {
		SOCKBUF_LOCK(&so->so_snd);
		if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
			SOCKBUF_UNLOCK(&so->so_snd);
			error = EPIPE;
			goto release;
		}
		if (so->so_error) {
			error = so->so_error;
			so->so_error = 0;
			SOCKBUF_UNLOCK(&so->so_snd);
			goto release;
		}
		if ((so->so_state & SS_ISCONNECTED) == 0) {
			/*
			 * `sendto' and `sendmsg' is allowed on a connection-
			 * based socket if it supports implied connect.
			 * Return ENOTCONN if not connected and no address is
			 * supplied.
			 */
			if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
			    (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
				if ((so->so_state & SS_ISCONFIRMING) == 0 &&
				    !(resid == 0 && clen != 0)) {
					SOCKBUF_UNLOCK(&so->so_snd);
					error = ENOTCONN;
					goto release;
				}
			} else if (addr == NULL) {
				SOCKBUF_UNLOCK(&so->so_snd);
				if (so->so_proto->pr_flags & PR_CONNREQUIRED)
					error = ENOTCONN;
				else
					error = EDESTADDRREQ;
				goto release;
			}
		}
		space = sbspace(&so->so_snd);
		if (flags & MSG_OOB)
			space += 1024;
		if ((atomic && resid > so->so_snd.sb_hiwat) ||
		    clen > so->so_snd.sb_hiwat) {
			SOCKBUF_UNLOCK(&so->so_snd);
			error = EMSGSIZE;
			goto release;
		}
		if (space < resid + clen &&
		    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
			if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) {
				if (so->so_upcallprep.soup_send) {
					so->so_upcallprep.soup_send(so,
						so->so_upcallprep.soup_send_arg,
						resid);
				}
				SOCKBUF_UNLOCK(&so->so_snd);
				error = EWOULDBLOCK;
				goto release;
			}
			error = sbwait(&so->so_snd);
			SOCKBUF_UNLOCK(&so->so_snd);
			if (error)
				goto release;
			goto restart;
		}
		SOCKBUF_UNLOCK(&so->so_snd);
		space -= clen;
		do {
			if (uio == NULL) {
				resid = 0;
				if (flags & MSG_EOR)
					top->m_flags |= M_EOR;
			} else {
#ifdef ZERO_COPY_SOCKETS
				error = sosend_copyin(uio, &top, atomic,
				    &space, flags);
				if (error != 0)
					goto release;
#else
				/*
				 * Copy the data from userland into a mbuf
				 * chain.  If no data is to be copied in,
				 * a single empty mbuf is returned.
				 */
				top = m_uiotombuf(uio, M_WAITOK, space,
				    (atomic ? max_hdr : 0),
				    (atomic ? M_PKTHDR : 0) |
				    ((flags & MSG_EOR) ? M_EOR : 0));
				if (top == NULL) {
					error = EFAULT; /* only possible error */
					goto release;
				}
				space -= resid - uio->uio_resid;
#endif
				resid = uio->uio_resid;
			}
			if (dontroute) {
				SOCK_LOCK(so);
				so->so_options |= SO_DONTROUTE;
				SOCK_UNLOCK(so);
			}
			/*
			 * XXX all the SBS_CANTSENDMORE checks previously
			 * done could be out of date.  We could have recieved
			 * a reset packet in an interrupt or maybe we slept
			 * while doing page faults in uiomove() etc.  We
			 * could probably recheck again inside the locking
			 * protection here, but there are probably other
			 * places that this also happens.  We must rethink
			 * this.
			 */
			VNET_SO_ASSERT(so);
			error = (*so->so_proto->pr_usrreqs->pru_send)(so,
			    (flags & MSG_OOB) ? PRUS_OOB :
			/*
			 * If the user set MSG_EOF, the protocol understands
			 * this flag and nothing left to send then use
			 * PRU_SEND_EOF instead of PRU_SEND.
			 */
			    ((flags & MSG_EOF) &&
			     (so->so_proto->pr_flags & PR_IMPLOPCL) &&
			     (resid <= 0)) ?
				PRUS_EOF :
			/* If there is more to send set PRUS_MORETOCOME. */
			    (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
			    top, addr, control, td);
			if (dontroute) {
				SOCK_LOCK(so);
				so->so_options &= ~SO_DONTROUTE;
				SOCK_UNLOCK(so);
			}
			clen = 0;
			control = NULL;
			top = NULL;
			if (error)
				goto release;
		} while (resid && space > 0);
	} while (resid);

release:
	sbunlock(&so->so_snd);
out:
	if (top != NULL)
		m_freem(top);
	if (control != NULL)
		m_freem(control);
	return (error);
}

int
sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
    struct mbuf *top, struct mbuf *control, int flags, struct thread *td)
{
	int error;

	CURVNET_SET(so->so_vnet);
	error = so->so_proto->pr_usrreqs->pru_sosend(so, addr, uio, top,
	    control, flags, td);
	CURVNET_RESTORE();
	return (error);
}

/*
 * The part of soreceive() that implements reading non-inline out-of-band
 * data from a socket.  For more complete comments, see soreceive(), from
 * which this code originated.
 *
 * Note that soreceive_rcvoob(), unlike the remainder of soreceive(), is
 * unable to return an mbuf chain to the caller.
 */
static int
soreceive_rcvoob(struct socket *so, struct uio *uio, int flags)
{
	struct protosw *pr = so->so_proto;
	struct mbuf *m;
	int error;

	KASSERT(flags & MSG_OOB, ("soreceive_rcvoob: (flags & MSG_OOB) == 0"));
	VNET_SO_ASSERT(so);

	m = m_get(M_WAIT, MT_DATA);
	error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
	if (error)
		goto bad;
	do {
#ifdef ZERO_COPY_SOCKETS
		if (so_zero_copy_receive) {
			int disposable;

			if ((m->m_flags & M_EXT)
			 && (m->m_ext.ext_type == EXT_DISPOSABLE))
				disposable = 1;
			else
				disposable = 0;

			error = uiomoveco(mtod(m, void *),
					  min(uio->uio_resid, m->m_len),
					  uio, disposable);
		} else
#endif /* ZERO_COPY_SOCKETS */
		error = uiomove(mtod(m, void *),
		    (int) min(uio->uio_resid, m->m_len), uio);
		m = m_free(m);
	} while (uio->uio_resid && error == 0 && m);
bad:
	if (m != NULL)
		m_freem(m);
	return (error);
}

/*
 * Following replacement or removal of the first mbuf on the first mbuf chain
 * of a socket buffer, push necessary state changes back into the socket
 * buffer so that other consumers see the values consistently.  'nextrecord'
 * is the callers locally stored value of the original value of
 * sb->sb_mb->m_nextpkt which must be restored when the lead mbuf changes.
 * NOTE: 'nextrecord' may be NULL.
 */
static __inline void
sockbuf_pushsync(struct sockbuf *sb, struct mbuf *nextrecord)
{

	SOCKBUF_LOCK_ASSERT(sb);
	/*
	 * First, update for the new value of nextrecord.  If necessary, make
	 * it the first record.
	 */
	if (sb->sb_mb != NULL)
		sb->sb_mb->m_nextpkt = nextrecord;
	else
		sb->sb_mb = nextrecord;

        /*
         * Now update any dependent socket buffer fields to reflect the new
         * state.  This is an expanded inline of SB_EMPTY_FIXUP(), with the
	 * addition of a second clause that takes care of the case where
	 * sb_mb has been updated, but remains the last record.
         */
        if (sb->sb_mb == NULL) {
                sb->sb_mbtail = NULL;
                sb->sb_lastrecord = NULL;
        } else if (sb->sb_mb->m_nextpkt == NULL)
                sb->sb_lastrecord = sb->sb_mb;
}


/*
 * Implement receive operations on a socket.  We depend on the way that
 * records are added to the sockbuf by sbappend.  In particular, each record
 * (mbufs linked through m_next) must begin with an address if the protocol
 * so specifies, followed by an optional mbuf or mbufs containing ancillary
 * data, and then zero or more mbufs of data.  In order to allow parallelism
 * between network receive and copying to user space, as well as avoid
 * sleeping with a mutex held, we release the socket buffer mutex during the
 * user space copy.  Although the sockbuf is locked, new data may still be
 * appended, and thus we must maintain consistency of the sockbuf during that
 * time.
 *
 * The caller may receive the data as a single mbuf chain by supplying an
 * mbuf **mp0 for use in returning the chain.  The uio is then used only for
 * the count in uio_resid.
 */
int
soreceive_generic(struct socket *so, struct sockaddr **psa, struct uio *uio,
    struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
{
	struct mbuf *m, **mp;
	int flags, error, offset;
	ssize_t len;
	struct protosw *pr = so->so_proto;
	struct mbuf *nextrecord;
	int moff, type = 0, last_m_flags, hole_break = 0;
	ssize_t orig_resid = uio->uio_resid;

	mp = mp0;
	if (psa != NULL)
		*psa = NULL;
	if (controlp != NULL)
		*controlp = NULL;
	if (flagsp != NULL) {
		hole_break = *flagsp & MSG_HOLE_BREAK;
		*flagsp &= ~MSG_HOLE_BREAK;
		flags = *flagsp &~ MSG_EOR;
	} else
		flags = 0;
	if (flags & MSG_OOB)
		return (soreceive_rcvoob(so, uio, flags));
	if (mp != NULL)
		*mp = NULL;
	if ((pr->pr_flags & PR_WANTRCVD) && (so->so_state & SS_ISCONFIRMING)
	    && uio->uio_resid) {
		VNET_SO_ASSERT(so);
		(*pr->pr_usrreqs->pru_rcvd)(so, 0);
	}

	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
	if (error)
		return (error);

restart:
	SOCKBUF_LOCK(&so->so_rcv);
	m = so->so_rcv.sb_mb;
	/*
	 * If we have less data than requested, block awaiting more (subject
	 * to any timeout) if:
	 *   1. the current count is less than the low water mark, or
	 *   2. MSG_WAITALL is set, and it is possible to do the entire
	 *	receive operation at once if we block (resid <= hiwat).
	 *   3. MSG_DONTWAIT is not set
	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
	 * we have to do the receive in sections, and thus risk returning a
	 * short count if a timeout or signal occurs after we start.
	 */
	if (m == NULL || (((flags & MSG_DONTWAIT) == 0 &&
	    so->so_rcv.sb_cc < uio->uio_resid) &&
	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
	    m->m_nextpkt == NULL && (pr->pr_flags & PR_ATOMIC) == 0)) {
		KASSERT(m != NULL || !so->so_rcv.sb_cc,
		    ("receive: m == %p so->so_rcv.sb_cc == %u",
		    m, so->so_rcv.sb_cc));
		if (so->so_error) {
			if (m != NULL)
				goto dontblock;
			error = so->so_error;
			if ((flags & MSG_PEEK) == 0)
				so->so_error = 0;
			SOCKBUF_UNLOCK(&so->so_rcv);
			goto release;
		}
		SOCKBUF_LOCK_ASSERT(&so->so_rcv);
		if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
			if (m == NULL) {
				SOCKBUF_UNLOCK(&so->so_rcv);
				goto release;
			} else
				goto dontblock;
		}
		for (; m != NULL; m = m->m_next)
			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
				m = so->so_rcv.sb_mb;
				goto dontblock;
			}
		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
			SOCKBUF_UNLOCK(&so->so_rcv);
			error = ENOTCONN;
			goto release;
		}
		if (uio->uio_resid == 0) {
			SOCKBUF_UNLOCK(&so->so_rcv);
			goto release;
		}
		if ((so->so_state & SS_NBIO) ||
		    (flags & (MSG_DONTWAIT|MSG_NBIO))) {
			if (so->so_upcallprep.soup_receive != NULL) {
				so->so_upcallprep.soup_receive(so,
					so->so_upcallprep.soup_receive_arg,
					orig_resid - uio->uio_resid, uio->uio_resid);
			}
			SOCKBUF_UNLOCK(&so->so_rcv);
			error = EWOULDBLOCK;
			goto release;
		}
		SBLASTRECORDCHK(&so->so_rcv);
		SBLASTMBUFCHK(&so->so_rcv);
		error = sbwait(&so->so_rcv);
		SOCKBUF_UNLOCK(&so->so_rcv);
		if (error)
			goto release;
		goto restart;
	}
dontblock:
	/*
	 * From this point onward, we maintain 'nextrecord' as a cache of the
	 * pointer to the next record in the socket buffer.  We must keep the
	 * various socket buffer pointers and local stack versions of the
	 * pointers in sync, pushing out modifications before dropping the
	 * socket buffer mutex, and re-reading them when picking it up.
	 *
	 * Otherwise, we will race with the network stack appending new data
	 * or records onto the socket buffer by using inconsistent/stale
	 * versions of the field, possibly resulting in socket buffer
	 * corruption.
	 *
	 * By holding the high-level sblock(), we prevent simultaneous
	 * readers from pulling off the front of the socket buffer.
	 */
	SOCKBUF_LOCK_ASSERT(&so->so_rcv);
	if (uio->u…