usr/src/uts/common/inet/ip/icmp.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */
/* Copyright (c) 1990 Mentat Inc. */

#include <sys/types.h>
#include <sys/stream.h>
#include <sys/stropts.h>
#include <sys/strlog.h>
#include <sys/strsun.h>
#define	_SUN_TPI_VERSION 2
#include <sys/tihdr.h>
#include <sys/timod.h>
#include <sys/ddi.h>
#include <sys/sunddi.h>
#include <sys/strsubr.h>
#include <sys/cmn_err.h>
#include <sys/debug.h>
#include <sys/kmem.h>
#include <sys/policy.h>
#include <sys/priv.h>
#include <sys/zone.h>
#include <sys/time.h>

#include <sys/sockio.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/isa_defs.h>
#include <sys/suntpi.h>
#include <sys/xti_inet.h>
#include <sys/netstack.h>

#include <net/route.h>
#include <net/if.h>

#include <netinet/in.h>
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <inet/common.h>
#include <inet/ip.h>
#include <inet/ip6.h>
#include <inet/proto_set.h>
#include <inet/nd.h>
#include <inet/optcom.h>
#include <inet/snmpcom.h>
#include <inet/kstatcom.h>
#include <inet/rawip_impl.h>

#include <netinet/ip_mroute.h>
#include <inet/tcp.h>
#include <net/pfkeyv2.h>
#include <inet/ipsec_info.h>
#include <inet/ipclassifier.h>

#include <sys/tsol/label.h>
#include <sys/tsol/tnet.h>

#include <inet/ip_ire.h>
#include <inet/ip_if.h>

#include <inet/ip_impl.h>
#include <sys/disp.h>

/*
 * Synchronization notes:
 *
 * RAWIP is MT and uses the usual kernel synchronization primitives. There is
 * locks, which is icmp_rwlock. We also use conn_lock when updating things
 * which affect the IP classifier lookup.
 * The lock order is icmp_rwlock -> conn_lock.
 *
 * The icmp_rwlock:
 * This protects most of the other fields in the icmp_t. The exact list of
 * fields which are protected by each of the above locks is documented in
 * the icmp_t structure definition.
 *
 * Plumbing notes:
 * ICMP is always a device driver. For compatibility with mibopen() code
 * it is possible to I_PUSH "icmp", but that results in pushing a passthrough
 * dummy module.
 */

static void	icmp_addr_req(queue_t *q, mblk_t *mp);
static void	icmp_tpi_bind(queue_t *q, mblk_t *mp);
static int	icmp_bind_proto(conn_t *connp);
static int	icmp_build_hdrs(icmp_t *icmp);
static void	icmp_capability_req(queue_t *q, mblk_t *mp);
static int	icmp_close(queue_t *q, int flags);
static void	icmp_tpi_connect(queue_t *q, mblk_t *mp);
static void	icmp_tpi_disconnect(queue_t *q, mblk_t *mp);
static void	icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
		    int sys_error);
static void	icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
		    t_scalar_t t_error, int sys_error);
static void	icmp_icmp_error(conn_t *connp, mblk_t *mp);
static void	icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp);
static void	icmp_info_req(queue_t *q, mblk_t *mp);
static void	icmp_input(void *, mblk_t *, void *);
static conn_t 	*icmp_open(int family, cred_t *credp, int *err, int flags);
static int	icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
		    cred_t *credp);
static int	icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
		    cred_t *credp);
static int	icmp_unitdata_opt_process(queue_t *q, mblk_t *mp,
		    int *errorp, void *thisdg_attrs);
static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
int		icmp_opt_set(conn_t *connp, uint_t optset_context,
		    int level, int name, uint_t inlen,
		    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
		    void *thisdg_attrs, cred_t *cr);
int		icmp_opt_get(conn_t *connp, int level, int name,
		    uchar_t *ptr);
static int	icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
static boolean_t icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt);
static int	icmp_param_set(queue_t *q, mblk_t *mp, char *value,
		    caddr_t cp, cred_t *cr);
static int	icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
		    uchar_t *ptr, int len);
static int	icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
		    cred_t *cr);
static void	icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err);
static void	icmp_tpi_unbind(queue_t *q, mblk_t *mp);
static int	icmp_update_label(icmp_t *icmp, mblk_t *mp, ipaddr_t dst);
static void	icmp_wput(queue_t *q, mblk_t *mp);
static void	icmp_wput_fallback(queue_t *q, mblk_t *mp);
static int	raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp,
		    sin6_t *sin6, ip6_pkt_t *ipp);
static int	raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp,
		    ipaddr_t v4dst, ip4_pkt_t *pktinfop);
static void	icmp_wput_other(queue_t *q, mblk_t *mp);
static void	icmp_wput_iocdata(queue_t *q, mblk_t *mp);
static void	icmp_wput_restricted(queue_t *q, mblk_t *mp);
static void	icmp_ulp_recv(conn_t *, mblk_t *);

static void	*rawip_stack_init(netstackid_t stackid, netstack_t *ns);
static void	rawip_stack_fini(netstackid_t stackid, void *arg);

static void	*rawip_kstat_init(netstackid_t stackid);
static void	rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp);
static int	rawip_kstat_update(kstat_t *kp, int rw);
static void	rawip_stack_shutdown(netstackid_t stackid, void *arg);
static int	rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa,
		    uint_t *salenp);
static int	rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa,
		    uint_t *salenp);

int		rawip_getsockname(sock_lower_handle_t, struct sockaddr *,
		    socklen_t *, cred_t *);
int		rawip_getpeername(sock_lower_handle_t, struct sockaddr *,
		    socklen_t *, cred_t *);

static struct module_info icmp_mod_info =  {
	5707, "icmp", 1, INFPSZ, 512, 128
};

/*
 * Entry points for ICMP as a device.
 * We have separate open functions for the /dev/icmp and /dev/icmp6 devices.
 */
static struct qinit icmprinitv4 = {
	NULL, NULL, icmp_openv4, icmp_close, NULL, &icmp_mod_info
};

static struct qinit icmprinitv6 = {
	NULL, NULL, icmp_openv6, icmp_close, NULL, &icmp_mod_info
};

static struct qinit icmpwinit = {
	(pfi_t)icmp_wput, NULL, NULL, NULL, NULL, &icmp_mod_info
};

/* ICMP entry point during fallback */
static struct qinit icmp_fallback_sock_winit = {
	(pfi_t)icmp_wput_fallback, NULL, NULL, NULL, NULL, &icmp_mod_info
};

/* For AF_INET aka /dev/icmp */
struct streamtab icmpinfov4 = {
	&icmprinitv4, &icmpwinit
};

/* For AF_INET6 aka /dev/icmp6 */
struct streamtab icmpinfov6 = {
	&icmprinitv6, &icmpwinit
};

static sin_t	sin_null;	/* Zero address for quick clears */
static sin6_t	sin6_null;	/* Zero address for quick clears */

/* Default structure copied into T_INFO_ACK messages */
static struct T_info_ack icmp_g_t_info_ack = {
	T_INFO_ACK,
	IP_MAXPACKET,	 /* TSDU_size.  icmp allows maximum size messages. */
	T_INVALID,	/* ETSDU_size.  icmp does not support expedited data. */
	T_INVALID,	/* CDATA_size. icmp does not support connect data. */
	T_INVALID,	/* DDATA_size. icmp does not support disconnect data. */
	0,		/* ADDR_size - filled in later. */
	0,		/* OPT_size - not initialized here */
	IP_MAXPACKET,	/* TIDU_size.  icmp allows maximum size messages. */
	T_CLTS,		/* SERV_type.  icmp supports connection-less. */
	TS_UNBND,	/* CURRENT_state.  This is set from icmp_state. */
	(XPG4_1|SENDZERO) /* PROVIDER_flag */
};

/*
 * Table of ND variables supported by icmp.  These are loaded into is_nd
 * when the stack instance is created.
 * All of these are alterable, within the min/max values given, at run time.
 */
static icmpparam_t	icmp_param_arr[] = {
	/* min	max	value	name */
	{ 0,	128,	32,	"icmp_wroff_extra" },
	{ 1,	255,	255,	"icmp_ipv4_ttl" },
	{ 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS,	"icmp_ipv6_hoplimit"},
	{ 0,	1,	1,	"icmp_bsd_compat" },
	{ 4096,	65536,	8192,	"icmp_xmit_hiwat"},
	{ 0,	65536,	1024,	"icmp_xmit_lowat"},
	{ 4096,	65536,	8192,	"icmp_recv_hiwat"},
	{ 65536, 1024*1024*1024, 256*1024,	"icmp_max_buf"},
};
#define	is_wroff_extra			is_param_arr[0].icmp_param_value
#define	is_ipv4_ttl			is_param_arr[1].icmp_param_value
#define	is_ipv6_hoplimit		is_param_arr[2].icmp_param_value
#define	is_bsd_compat			is_param_arr[3].icmp_param_value
#define	is_xmit_hiwat			is_param_arr[4].icmp_param_value
#define	is_xmit_lowat			is_param_arr[5].icmp_param_value
#define	is_recv_hiwat			is_param_arr[6].icmp_param_value
#define	is_max_buf			is_param_arr[7].icmp_param_value

static int rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len);
static int rawip_do_connect(conn_t *connp, const struct sockaddr *sa,
    socklen_t len, cred_t *cr);
static void rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error);

/*
 * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
 * passed to icmp_wput.
 * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP
 * protocol type placed in the message following the address. A T_BIND_ACK
 * message is returned by ip_bind_v4/v6.
 */
static void
icmp_tpi_bind(queue_t *q, mblk_t *mp)
{
	int	error;
	struct sockaddr *sa;
	struct T_bind_req *tbr;
	socklen_t	len;
	sin_t	*sin;
	sin6_t	*sin6;
	icmp_t		*icmp;
	conn_t	*connp = Q_TO_CONN(q);
	mblk_t *mp1;
	cred_t *cr;

	/*
	 * All Solaris components should pass a db_credp
	 * for this TPI message, hence we ASSERT.
	 * But in case there is some other M_PROTO that looks
	 * like a TPI message sent by some other kernel
	 * component, we check and return an error.
	 */
	cr = msg_getcred(mp, NULL);
	ASSERT(cr != NULL);
	if (cr == NULL) {
		icmp_err_ack(q, mp, TSYSERR, EINVAL);
		return;
	}

	icmp = connp->conn_icmp;
	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
		    "icmp_bind: bad req, len %u",
		    (uint_t)(mp->b_wptr - mp->b_rptr));
		icmp_err_ack(q, mp, TPROTO, 0);
		return;
	}

	if (icmp->icmp_state != TS_UNBND) {
		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
		    "icmp_bind: bad state, %d", icmp->icmp_state);
		icmp_err_ack(q, mp, TOUTSTATE, 0);
		return;
	}

	/*
	 * Reallocate the message to make sure we have enough room for an
	 * address and the protocol type.
	 */
	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
	if (!mp1) {
		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
		return;
	}
	mp = mp1;

	/* Reset the message type in preparation for shipping it back. */
	DB_TYPE(mp) = M_PCPROTO;
	tbr = (struct T_bind_req *)mp->b_rptr;
	len = tbr->ADDR_length;
	switch (len) {
	case 0:	/* request for a generic port */
		tbr->ADDR_offset = sizeof (struct T_bind_req);
		if (icmp->icmp_family == AF_INET) {
			tbr->ADDR_length = sizeof (sin_t);
			sin = (sin_t *)&tbr[1];
			*sin = sin_null;
			sin->sin_family = AF_INET;
			mp->b_wptr = (uchar_t *)&sin[1];
			sa = (struct sockaddr *)sin;
			len = sizeof (sin_t);
		} else {
			ASSERT(icmp->icmp_family == AF_INET6);
			tbr->ADDR_length = sizeof (sin6_t);
			sin6 = (sin6_t *)&tbr[1];
			*sin6 = sin6_null;
			sin6->sin6_family = AF_INET6;
			mp->b_wptr = (uchar_t *)&sin6[1];
			sa = (struct sockaddr *)sin6;
			len = sizeof (sin6_t);
		}
		break;

	case sizeof (sin_t):	/* Complete IPv4 address */
		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
		    sizeof (sin_t));
		break;

	case sizeof (sin6_t):	/* Complete IPv6 address */
		sa = (struct sockaddr *)mi_offset_param(mp,
		    tbr->ADDR_offset, sizeof (sin6_t));
		break;

	default:
		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
		    "icmp_bind: bad ADDR_length %d", tbr->ADDR_length);
		icmp_err_ack(q, mp, TBADADDR, 0);
		return;
	}

	error = rawip_do_bind(connp, sa, len);
done:
	ASSERT(mp->b_cont == NULL);
	if (error != 0) {
		if (error > 0) {
			icmp_err_ack(q, mp, TSYSERR, error);
		} else {
			icmp_err_ack(q, mp, -error, 0);
		}
	} else {
		tbr->PRIM_type = T_BIND_ACK;
		qreply(q, mp);
	}
}

static int
rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len)
{
	sin_t		*sin;
	sin6_t		*sin6;
	icmp_t		*icmp;
	int		error = 0;
	mblk_t		*ire_mp;


	icmp = connp->conn_icmp;

	if (sa == NULL || !OK_32PTR((char *)sa)) {
		return (EINVAL);
	}

	/*
	 * The state must be TS_UNBND. TPI mandates that users must send
	 * TPI primitives only 1 at a time and wait for the response before
	 * sending the next primitive.
	 */
	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
	if (icmp->icmp_state != TS_UNBND || icmp->icmp_pending_op != -1) {
		error = -TOUTSTATE;
		goto done;
	}

	ASSERT(len != 0);
	switch (len) {
	case sizeof (sin_t):    /* Complete IPv4 address */
		sin = (sin_t *)sa;
		if (sin->sin_family != AF_INET ||
		    icmp->icmp_family != AF_INET) {
			/* TSYSERR, EAFNOSUPPORT */
			error = EAFNOSUPPORT;
			goto done;
		}
		break;
	case sizeof (sin6_t): /* Complete IPv6 address */
		sin6 = (sin6_t *)sa;
		if (sin6->sin6_family != AF_INET6 ||
		    icmp->icmp_family != AF_INET6) {
			/* TSYSERR, EAFNOSUPPORT */
			error = EAFNOSUPPORT;
			goto done;
		}
		/* No support for mapped addresses on raw sockets */
		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
			/* TSYSERR, EADDRNOTAVAIL */
			error = EADDRNOTAVAIL;
			goto done;
		}
		break;

	default:
		/* TBADADDR */
		error = EADDRNOTAVAIL;
		goto done;
	}

	icmp->icmp_pending_op = T_BIND_REQ;
	icmp->icmp_state = TS_IDLE;

	/*
	 * Copy the source address into our icmp structure.  This address
	 * may still be zero; if so, ip will fill in the correct address
	 * each time an outbound packet is passed to it.
	 * If we are binding to a broadcast or multicast address then
	 * rawip_post_ip_bind_connect will clear the source address.
	 */

	if (icmp->icmp_family == AF_INET) {
		ASSERT(sin != NULL);
		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr,
		    &icmp->icmp_v6src);
		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
		    icmp->icmp_ip_snd_options_len;
		icmp->icmp_bound_v6src = icmp->icmp_v6src;
	} else {
		int error;

		ASSERT(sin6 != NULL);
		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
		icmp->icmp_v6src = sin6->sin6_addr;
		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
		icmp->icmp_bound_v6src = icmp->icmp_v6src;

		/* Rebuild the header template */
		error = icmp_build_hdrs(icmp);
		if (error != 0) {
			icmp->icmp_pending_op = -1;
			/*
			 * TSYSERR
			 */
			goto done;
		}
	}

	ire_mp = NULL;
	if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) {
		/*
		 * request an IRE if src not 0 (INADDR_ANY)
		 */
		ire_mp = allocb(sizeof (ire_t), BPRI_HI);
		if (ire_mp == NULL) {
			icmp->icmp_pending_op = -1;
			error = ENOMEM;
			goto done;
		}
		DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;
	}
done:
	rw_exit(&icmp->icmp_rwlock);
	if (error != 0)
		return (error);

	if (icmp->icmp_family == AF_INET6) {
		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
		    &sin6->sin6_addr, sin6->sin6_port, B_TRUE);
	} else {
		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
		    sin->sin_addr.s_addr, sin->sin_port, B_TRUE);
	}
	rawip_post_ip_bind_connect(icmp, ire_mp, error);
	return (error);
}

static void
rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error)
{
	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
	if (icmp->icmp_state == TS_UNBND) {
		/*
		 * not yet bound - bind sent by icmp_bind_proto.
		 */
		rw_exit(&icmp->icmp_rwlock);
		return;
	}
	ASSERT(icmp->icmp_pending_op != -1);
	icmp->icmp_pending_op = -1;

	if (error != 0) {
		if (icmp->icmp_state == TS_DATA_XFER) {
			/* Connect failed */
			/* Revert back to the bound source */
			icmp->icmp_v6src = icmp->icmp_bound_v6src;
			icmp->icmp_state = TS_IDLE;
			if (icmp->icmp_family == AF_INET6)
				(void) icmp_build_hdrs(icmp);
		} else {
			V6_SET_ZERO(icmp->icmp_v6src);
			V6_SET_ZERO(icmp->icmp_bound_v6src);
			icmp->icmp_state = TS_UNBND;
			if (icmp->icmp_family == AF_INET6)
				(void) icmp_build_hdrs(icmp);
		}
	} else {
		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
			ire_t *ire;

			ire = (ire_t *)ire_mp->b_rptr;
			/*
			 * If a broadcast/multicast address was bound set
			 * the source address to 0.
			 * This ensures no datagrams with broadcast address
			 * as source address are emitted (which would violate
			 * RFC1122 - Hosts requirements)
			 * Note: we get IRE_BROADCAST for IPv6
			 * to "mark" a multicast local address.
			 */


			if (ire->ire_type == IRE_BROADCAST &&
			    icmp->icmp_state != TS_DATA_XFER) {
				/*
				 * This was just a local bind to a
				 * MC/broadcast addr
				 */
				V6_SET_ZERO(icmp->icmp_v6src);
				if (icmp->icmp_family == AF_INET6)
					(void) icmp_build_hdrs(icmp);
			}
		}

	}
	rw_exit(&icmp->icmp_rwlock);
	if (ire_mp != NULL)
		freeb(ire_mp);
}

/*
 * Send message to IP to just bind to the protocol.
 */
static int
icmp_bind_proto(conn_t *connp)
{
	icmp_t	*icmp;
	int	error;

	icmp = connp->conn_icmp;

	if (icmp->icmp_family == AF_INET6)
		error = ip_proto_bind_laddr_v6(connp, NULL, icmp->icmp_proto,
		    &sin6_null.sin6_addr, 0, B_TRUE);
	else
		error = ip_proto_bind_laddr_v4(connp, NULL, icmp->icmp_proto,
		    sin_null.sin_addr.s_addr, 0, B_TRUE);

	rawip_post_ip_bind_connect(icmp, NULL, error);
	return (error);
}

static void
icmp_tpi_connect(queue_t *q, mblk_t *mp)
{
	conn_t	*connp = Q_TO_CONN(q);
	struct T_conn_req	*tcr;
	icmp_t	*icmp;
	struct sockaddr *sa;
	socklen_t len;
	int error;
	cred_t *cr;

	/*
	 * All Solaris components should pass a db_credp
	 * for this TPI message, hence we ASSERT.
	 * But in case there is some other M_PROTO that looks
	 * like a TPI message sent by some other kernel
	 * component, we check and return an error.
	 */
	cr = msg_getcred(mp, NULL);
	ASSERT(cr != NULL);
	if (cr == NULL) {
		icmp_err_ack(q, mp, TSYSERR, EINVAL);
		return;
	}

	icmp = connp->conn_icmp;
	tcr = (struct T_conn_req *)mp->b_rptr;
	/* Sanity checks */
	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
		icmp_err_ack(q, mp, TPROTO, 0);
		return;
	}

	if (tcr->OPT_length != 0) {
		icmp_err_ack(q, mp, TBADOPT, 0);
		return;
	}

	len = tcr->DEST_length;

	switch (len) {
	default:
		icmp_err_ack(q, mp, TBADADDR, 0);
		return;
	case sizeof (sin_t):
		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
		    sizeof (sin_t));
		break;
	case sizeof (sin6_t):
		sa = (struct sockaddr *)mi_offset_param(mp,
		    tcr->DEST_offset, sizeof (sin6_t));
		break;
	}

	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
	if (error != 0) {
		icmp_err_ack(q, mp, TSYSERR, error);
		return;
	}

	error = rawip_do_connect(connp, sa, len, cr);
	if (error != 0) {
		if (error < 0) {
			icmp_err_ack(q, mp, -error, 0);
		} else {
			icmp_err_ack(q, mp, 0, error);
		}
	} else {
		mblk_t *mp1;

		/*
		 * We have to send a connection confirmation to
		 * keep TLI happy.
		 */
		if (icmp->icmp_family == AF_INET) {
			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
			    sizeof (sin_t), NULL, 0);
		} else {
			ASSERT(icmp->icmp_family == AF_INET6);
			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
			    sizeof (sin6_t), NULL, 0);
		}
		if (mp1 == NULL) {
			rw_exit(&icmp->icmp_rwlock);
			icmp_err_ack(q, mp, TSYSERR, ENOMEM);
			return;
		}

		/*
		 * Send ok_ack for T_CONN_REQ
		 */
		mp = mi_tpi_ok_ack_alloc(mp);
		if (mp == NULL) {
			/* Unable to reuse the T_CONN_REQ for the ack. */
			freemsg(mp1);
			icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
			return;
		}
		putnext(connp->conn_rq, mp);
		putnext(connp->conn_rq, mp1);
	}
}

static int
rawip_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len,
    cred_t *cr)
{
	icmp_t	*icmp;
	sin_t	*sin;
	sin6_t	*sin6;
	mblk_t  *ire_mp;
	int	error;
	ipaddr_t	v4dst;
	in6_addr_t	v6dst;

	icmp = connp->conn_icmp;

	if (sa == NULL || !OK_32PTR((char *)sa)) {
		return (EINVAL);
	}

	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
	if (ire_mp == NULL)
		return (ENOMEM);
	DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;


	ASSERT(sa != NULL && len != 0);

	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
		rw_exit(&icmp->icmp_rwlock);
		freeb(ire_mp);
		return (-TOUTSTATE);
	}

	switch (len) {
	case sizeof (sin_t):
		sin = (sin_t *)sa;

		ASSERT(icmp->icmp_family == AF_INET);
		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);

		v4dst = sin->sin_addr.s_addr;
		/*
		 * Interpret a zero destination to mean loopback.
		 * Update the T_CONN_REQ (sin/sin6) since it is used to
		 * generate the T_CONN_CON.
		 */
		if (v4dst == INADDR_ANY) {
			v4dst = htonl(INADDR_LOOPBACK);
		}

		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
		    icmp->icmp_ip_snd_options_len;
		icmp->icmp_v6dst.sin6_addr = v6dst;
		icmp->icmp_v6dst.sin6_family = AF_INET6;
		icmp->icmp_v6dst.sin6_flowinfo = 0;
		icmp->icmp_v6dst.sin6_port = 0;

		/*
		 * If the destination address is multicast and
		 * an outgoing multicast interface has been set,
		 * use the address of that interface as our
		 * source address if no source address has been set.
		 */
		if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY &&
		    CLASSD(v4dst) &&
		    icmp->icmp_multicast_if_addr != INADDR_ANY) {
			IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr,
			    &icmp->icmp_v6src);
		}
		break;
	case sizeof (sin6_t):
		sin6 = (sin6_t *)sa;

		/* No support for mapped addresses on raw sockets */
		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
			rw_exit(&icmp->icmp_rwlock);
			freeb(ire_mp);
			return (EADDRNOTAVAIL);
		}

		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
		ASSERT(icmp->icmp_family == AF_INET6);

		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;

		icmp->icmp_v6dst = *sin6;
		icmp->icmp_v6dst.sin6_port = 0;

		/*
		 * Interpret a zero destination to mean loopback.
		 * Update the T_CONN_REQ (sin/sin6) since it is used to
		 * generate the T_CONN_CON.
		 */
		if (IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6dst.sin6_addr)) {
			icmp->icmp_v6dst.sin6_addr = ipv6_loopback;
		}
		/*
		 * If the destination address is multicast and
		 * an outgoing multicast interface has been set,
		 * then the ip bind logic will pick the correct source
		 * address (i.e. matching the outgoing multicast interface).
		 */
		break;
	}

	icmp->icmp_pending_op = T_CONN_REQ;

	if (icmp->icmp_state == TS_DATA_XFER) {
		/* Already connected - clear out state */
		icmp->icmp_v6src = icmp->icmp_bound_v6src;
		icmp->icmp_state = TS_IDLE;
	}

	icmp->icmp_state = TS_DATA_XFER;
	rw_exit(&icmp->icmp_rwlock);

	if (icmp->icmp_family == AF_INET6) {
		error = ip_proto_bind_connected_v6(connp, &ire_mp,
		    icmp->icmp_proto, &icmp->icmp_v6src, 0,
		    &icmp->icmp_v6dst.sin6_addr,
		    NULL, sin6->sin6_port, B_TRUE, B_TRUE, cr);
	} else {
		error = ip_proto_bind_connected_v4(connp, &ire_mp,
		    icmp->icmp_proto, &V4_PART_OF_V6(icmp->icmp_v6src), 0,
		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr), sin->sin_port,
		    B_TRUE, B_TRUE, cr);
	}
	rawip_post_ip_bind_connect(icmp, ire_mp, error);
	return (error);
}

static void
icmp_close_free(conn_t *connp)
{
	icmp_t *icmp = connp->conn_icmp;

	/* If there are any options associated with the stream, free them. */
	if (icmp->icmp_ip_snd_options != NULL) {
		mi_free((char *)icmp->icmp_ip_snd_options);
		icmp->icmp_ip_snd_options = NULL;
		icmp->icmp_ip_snd_options_len = 0;
	}

	if (icmp->icmp_filter != NULL) {
		kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t));
		icmp->icmp_filter = NULL;
	}

	/* Free memory associated with sticky options */
	if (icmp->icmp_sticky_hdrs_len != 0) {
		kmem_free(icmp->icmp_sticky_hdrs,
		    icmp->icmp_sticky_hdrs_len);
		icmp->icmp_sticky_hdrs = NULL;
		icmp->icmp_sticky_hdrs_len = 0;
	}
	ip6_pkt_free(&icmp->icmp_sticky_ipp);

	/*
	 * Clear any fields which the kmem_cache constructor clears.
	 * Only icmp_connp needs to be preserved.
	 * TBD: We should make this more efficient to avoid clearing
	 * everything.
	 */
	ASSERT(icmp->icmp_connp == connp);
	bzero(icmp, sizeof (icmp_t));
	icmp->icmp_connp = connp;
}

static int
rawip_do_close(conn_t *connp)
{
	ASSERT(connp != NULL && IPCL_IS_RAWIP(connp));

	ip_quiesce_conn(connp);

	if (!IPCL_IS_NONSTR(connp)) {
		qprocsoff(connp->conn_rq);
	}

	ASSERT(connp->conn_icmp->icmp_fallback_queue_head == NULL &&
	    connp->conn_icmp->icmp_fallback_queue_tail == NULL);
	icmp_close_free(connp);

	/*
	 * Now we are truly single threaded on this stream, and can
	 * delete the things hanging off the connp, and finally the connp.
	 * We removed this connp from the fanout list, it cannot be
	 * accessed thru the fanouts, and we already waited for the
	 * conn_ref to drop to 0. We are already in close, so
	 * there cannot be any other thread from the top. qprocsoff
	 * has completed, and service has completed or won't run in
	 * future.
	 */
	ASSERT(connp->conn_ref == 1);

	if (!IPCL_IS_NONSTR(connp)) {
		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
	} else {
		ip_free_helper_stream(connp);
	}

	connp->conn_ref--;
	ipcl_conn_destroy(connp);

	return (0);
}

static int
icmp_close(queue_t *q, int flags)
{
	conn_t  *connp;

	if (flags & SO_FALLBACK) {
		/*
		 * stream is being closed while in fallback
		 * simply free the resources that were allocated
		 */
		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
		qprocsoff(q);
		goto done;
	}

	connp = Q_TO_CONN(q);
	(void) rawip_do_close(connp);
done:
	q->q_ptr = WR(q)->q_ptr = NULL;
	return (0);
}

/*
 * This routine handles each T_DISCON_REQ message passed to icmp
 * as an indicating that ICMP is no longer connected. This results
 * in sending a T_BIND_REQ to IP to restore the binding to just
 * the local address.
 *
 * The disconnect completes in rawip_post_ip_bind_connect.
 */
static int
icmp_do_disconnect(conn_t *connp)
{
	icmp_t	*icmp;
	mblk_t	*ire_mp;
	int error;

	icmp = connp->conn_icmp;
	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
	if (icmp->icmp_state != TS_DATA_XFER || icmp->icmp_pending_op != -1) {
		rw_exit(&icmp->icmp_rwlock);
		return (-TOUTSTATE);
	}
	icmp->icmp_pending_op = T_DISCON_REQ;
	icmp->icmp_v6src = icmp->icmp_bound_v6src;
	icmp->icmp_state = TS_IDLE;


	if (icmp->icmp_family == AF_INET6) {
		/* Rebuild the header template */
		error = icmp_build_hdrs(icmp);
		if (error != 0) {
			icmp->icmp_pending_op = -1;
			rw_exit(&icmp->icmp_rwlock);
			return (error);
		}
	}

	rw_exit(&icmp->icmp_rwlock);
	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
	if (ire_mp == NULL) {
		return (ENOMEM);
	}

	if (icmp->icmp_family == AF_INET6) {
		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
		    &icmp->icmp_bound_v6src, 0, B_TRUE);
	} else {

		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
		    V4_PART_OF_V6(icmp->icmp_bound_v6src), 0, B_TRUE);
	}

	rawip_post_ip_bind_connect(icmp, ire_mp, error);

	return (error);
}

static void
icmp_tpi_disconnect(queue_t *q, mblk_t *mp)
{
	conn_t	*connp = Q_TO_CONN(q);
	int	error;

	/*
	 * Allocate the largest primitive we need to send back
	 * T_error_ack is > than T_ok_ack
	 */
	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
	if (mp == NULL) {
		/* Unable to reuse the T_DISCON_REQ for the ack. */
		icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
		return;
	}

	error = icmp_do_disconnect(connp);

	if (error != 0) {
		if (error > 0) {
			icmp_err_ack(q, mp, 0, error);
		} else {
			icmp_err_ack(q, mp, -error, 0);
		}
	} else {
		mp = mi_tpi_ok_ack_alloc(mp);
		ASSERT(mp != NULL);
		qreply(q, mp);
	}

}

static int
icmp_disconnect(conn_t *connp)
{
	int	error;
	icmp_t	*icmp = connp->conn_icmp;

	icmp->icmp_dgram_errind = B_FALSE;

	error = icmp_do_disconnect(connp);

	if (error < 0)
		error = proto_tlitosyserr(-error);
	return (error);
}

/* This routine creates a T_ERROR_ACK message and passes it upstream. */
static void
icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
{
	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
		qreply(q, mp);
}

/* Shorthand to generate and send TPI error acks to our client */
static void
icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
    t_scalar_t t_error, int sys_error)
{
	struct T_error_ack	*teackp;

	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
		teackp = (struct T_error_ack *)mp->b_rptr;
		teackp->ERROR_prim = primitive;
		teackp->TLI_error = t_error;
		teackp->UNIX_error = sys_error;
		qreply(q, mp);
	}
}

/*
 * icmp_icmp_error is called by icmp_input to process ICMP
 * messages passed up by IP.
 * Generates the appropriate permanent (non-transient) errors.
 * Assumes that IP has pulled up everything up to and including
 * the ICMP header.
 */
static void
icmp_icmp_error(conn_t *connp, mblk_t *mp)
{
	icmph_t *icmph;
	ipha_t	*ipha;
	int	iph_hdr_length;
	sin_t	sin;
	mblk_t	*mp1;
	int	error = 0;
	icmp_t	*icmp = connp->conn_icmp;

	ipha = (ipha_t *)mp->b_rptr;

	ASSERT(OK_32PTR(mp->b_rptr));

	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
		icmp_icmp_error_ipv6(connp, mp);
		return;
	}

	/*
	 * icmp does not support v4 mapped addresses
	 * so we can never be here for a V6 socket
	 * i.e. icmp_family == AF_INET6
	 */
	ASSERT((IPH_HDR_VERSION(ipha) == IPV4_VERSION) &&
	    (icmp->icmp_family == AF_INET));

	ASSERT(icmp->icmp_family == AF_INET);

	/* Skip past the outer IP and ICMP headers */
	iph_hdr_length = IPH_HDR_LENGTH(ipha);
	icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]);
	ipha = (ipha_t *)&icmph[1];
	iph_hdr_length = IPH_HDR_LENGTH(ipha);

	switch (icmph->icmph_type) {
	case ICMP_DEST_UNREACHABLE:
		switch (icmph->icmph_code) {
		case ICMP_FRAGMENTATION_NEEDED:
			/*
			 * IP has already adjusted the path MTU.
			 */
			break;
		case ICMP_PORT_UNREACHABLE:
		case ICMP_PROTOCOL_UNREACHABLE:
			error = ECONNREFUSED;
			break;
		default:
			/* Transient errors */
			break;
		}
		break;
	default:
		/* Transient errors */
		break;
	}
	if (error == 0) {
		freemsg(mp);
		return;
	}

	/*
	 * Deliver T_UDERROR_IND when the application has asked for it.
	 * The socket layer enables this automatically when connected.
	 */
	if (!icmp->icmp_dgram_errind) {
		freemsg(mp);
		return;
	}

	sin = sin_null;
	sin.sin_family = AF_INET;
	sin.sin_addr.s_addr = ipha->ipha_dst;

	if (IPCL_IS_NONSTR(connp)) {
		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
		if (icmp->icmp_state == TS_DATA_XFER) {
			if (sin.sin_addr.s_addr ==
			    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr)) {
				rw_exit(&icmp->icmp_rwlock);
				(*connp->conn_upcalls->su_set_error)
				    (connp->conn_upper_handle, error);
				goto done;
			}
		} else {
			icmp->icmp_delayed_error = error;
			*((sin_t *)&icmp->icmp_delayed_addr) = sin;
		}
		rw_exit(&icmp->icmp_rwlock);
	} else {
		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL,
		    0, error);
		if (mp1 != NULL)
			putnext(connp->conn_rq, mp1);
	}
done:
	ASSERT(!RW_ISWRITER(&icmp->icmp_rwlock));
	freemsg(mp);
}

/*
 * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6
 * for IPv6 packets.
 * Send permanent (non-transient) errors upstream.
 * Assumes that IP has pulled up all the extension headers as well
 * as the ICMPv6 header.
 */
static void
icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
{
	icmp6_t		*icmp6;
	ip6_t		*ip6h, *outer_ip6h;
	uint16_t	iph_hdr_length;
	uint8_t		*nexthdrp;
	sin6_t		sin6;
	mblk_t		*mp1;
	int		error = 0;
	icmp_t		*icmp = connp->conn_icmp;

	outer_ip6h = (ip6_t *)mp->b_rptr;
	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
	else
		iph_hdr_length = IPV6_HDR_LEN;

	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
	ip6h = (ip6_t *)&icmp6[1];
	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
		freemsg(mp);
		return;
	}

	switch (icmp6->icmp6_type) {
	case ICMP6_DST_UNREACH:
		switch (icmp6->icmp6_code) {
		case ICMP6_DST_UNREACH_NOPORT:
			error = ECONNREFUSED;
			break;
		case ICMP6_DST_UNREACH_ADMIN:
		case ICMP6_DST_UNREACH_NOROUTE:
		case ICMP6_DST_UNREACH_BEYONDSCOPE:
		case ICMP6_DST_UNREACH_ADDR:
			/* Transient errors */
			break;
		default:
			break;
		}
		break;
	case ICMP6_PACKET_TOO_BIG: {
		struct T_unitdata_ind	*tudi;
		struct T_opthdr		*toh;
		size_t			udi_size;
		mblk_t			*newmp;
		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
		    sizeof (struct ip6_mtuinfo);
		sin6_t			*sin6;
		struct ip6_mtuinfo	*mtuinfo;

		/*
		 * If the application has requested to receive path mtu
		 * information, send up an empty message containing an
		 * IPV6_PATHMTU ancillary data item.
		 */
		if (!icmp->icmp_ipv6_recvpathmtu)
			break;

		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
		    opt_length;
		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
			BUMP_MIB(&icmp->icmp_is->is_rawip_mib, rawipInErrors);
			break;
		}

		/*
		 * newmp->b_cont is left to NULL on purpose.  This is an
		 * empty message containing only ancillary data.
		 */
		newmp->b_datap->db_type = M_PROTO;
		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
		newmp->b_wptr = (uchar_t *)tudi + udi_size;
		tudi->PRIM_type = T_UNITDATA_IND;
		tudi->SRC_length = sizeof (sin6_t);
		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
		tudi->OPT_length = opt_length;

		sin6 = (sin6_t *)&tudi[1];
		bzero(sin6, sizeof (sin6_t));
		sin6->sin6_family = AF_INET6;
		sin6->sin6_addr = icmp->icmp_v6dst.sin6_addr;

		toh = (struct T_opthdr *)&sin6[1];
		toh->level = IPPROTO_IPV6;
		toh->name = IPV6_PATHMTU;
		toh->len = opt_length;
		toh->status = 0;

		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
		/*
		 * We've consumed everything we need from the original
		 * message.  Free it, then send our empty message.
		 */
		freemsg(mp);
		icmp_ulp_recv(connp, newmp);

		return;
	}
	case ICMP6_TIME_EXCEEDED:
		/* Transient errors */
		break;
	case ICMP6_PARAM_PROB:
		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
		    (uchar_t *)nexthdrp) {
			error = ECONNREFUSED;
			break;
		}
		break;
	}
	if (error == 0) {
		freemsg(mp);
		return;
	}

	/*
	 * Deliver T_UDERROR_IND when the application has asked for it.
	 * The socket layer enables this automatically when connected.
	 */
	if (!icmp->icmp_dgram_errind) {
		freemsg(mp);
		return;
	}

	sin6 = sin6_null;
	sin6.sin6_family = AF_INET6;
	sin6.sin6_addr = ip6h->ip6_dst;
	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;

	if (IPCL_IS_NONSTR(connp)) {
		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
		if (icmp->icmp_state == TS_DATA_XFER) {
			if (IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
			    &icmp->icmp_v6dst.sin6_addr)) {
				rw_exit(&icmp->icmp_rwlock);
				(*connp->conn_upcalls->su_set_error)
				    (connp->conn_upper_handle, error);
				goto done;
			}
		} else {
			icmp->icmp_delayed_error = error;
			*((sin6_t *)&icmp->icmp_delayed_addr) = sin6;
		}
		rw_exit(&icmp->icmp_rwlock);
	} else {
		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
		    NULL, 0, error);
		if (mp1 != NULL)
			putnext(connp->conn_rq, mp1);
	}
done:
	ASSERT(!RW_ISWRITER(&icmp->icmp_rwlock));
	freemsg(mp);
}

/*
 * This routine responds to T_ADDR_REQ messages.  It is called by icmp_wput.
 * The local address is filled in if endpoint is bound. The remote address
 * is filled in if remote address has been precified ("connected endpoint")
 * (The concept of connected CLTS sockets is alien to published TPI
 *  but we support it anyway).
 */
static void
icmp_addr_req(queue_t *q, mblk_t *mp)
{
	icmp_t	*icmp = Q_TO_ICMP(q);
	mblk_t	*ackmp;
	struct T_addr_ack *taa;

	/* Make it large enough for worst case */
	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
	    2 * sizeof (sin6_t), 1);
	if (ackmp == NULL) {
		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
		return;
	}
	taa = (struct T_addr_ack *)ackmp->b_rptr;

	bzero(taa, sizeof (struct T_addr_ack));
	ackmp->b_wptr = (uchar_t *)&taa[1];

	taa->PRIM_type = T_ADDR_ACK;
	ackmp->b_datap->db_type = M_PCPROTO;
	rw_enter(&icmp->icmp_rwlock, RW_READER);
	/*
	 * Note: Following code assumes 32 bit alignment of basic
	 * data structures like sin_t and struct T_addr_ack.
	 */
	if (icmp->icmp_state != TS_UNBND) {
		/*
		 * Fill in local address
		 */
		taa->LOCADDR_offset = sizeof (*taa);
		if (icmp->icmp_family == AF_INET) {
			sin_t	*sin;

			taa->LOCADDR_length = sizeof (sin_t);
			sin = (sin_t *)&taa[1];
			/* Fill zeroes and then intialize non-zero fields */
			*sin = sin_null;
			sin->sin_family = AF_INET;
			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src,
				    sin->sin_addr.s_addr);
			} else {
				/*
				 * INADDR_ANY
				 * icmp_v6src is not set, we might be bound to
				 * broadcast/multicast. Use icmp_bound_v6src as
				 * local address instead (that could
				 * also still be INADDR_ANY)
				 */
				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src,
				    sin->sin_addr.s_addr);
			}
			ackmp->b_wptr = (uchar_t *)&sin[1];
		} else {
			sin6_t	*sin6;

			ASSERT(icmp->icmp_family == AF_INET6);
			taa->LOCADDR_length = sizeof (sin6_t);
			sin6 = (sin6_t *)&taa[1];
			/* Fill zeroes and then intialize non-zero fields */
			*sin6 = sin6_null;
			sin6->sin6_family = AF_INET6;
			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
				sin6->sin6_addr = icmp->icmp_v6src;
			} else {
				/*
				 * UNSPECIFIED
				 * icmp_v6src is not set, we might be bound to
				 * broadcast/multicast. Use icmp_bound_v6src as
				 * local address instead (that could
				 * also still be UNSPECIFIED)
				 */
				sin6->sin6_addr = icmp->icmp_bound_v6src;
			}
			ackmp->b_wptr = (uchar_t *)&sin6[1];
		}
	}
	rw_exit(&icmp->icmp_rwlock);
	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
	qreply(q, ackmp);
}

static void
icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp)
{
	*tap = icmp_g_t_info_ack;

	if (icmp->icmp_family == AF_INET6)
		tap->ADDR_size = sizeof (sin6_t);
	else
		tap->ADDR_size = sizeof (sin_t);
	tap->CURRENT_state = icmp->icmp_state;
	tap->OPT_size = icmp_max_optsize;
}

static void
icmp_do_capability_ack(icmp_t *icmp, struct T_capability_ack *tcap,
    t_uscalar_t cap_bits1)
{
	tcap->CAP_bits1 = 0;

	if (cap_bits1 & TC1_INFO) {
		icmp_copy_info(&tcap->INFO_ack, icmp);
		tcap->CAP_bits1 |= TC1_INFO;
	}
}

/*
 * This routine responds to T_CAPABILITY_REQ messages.  It is called by
 * icmp_wput.  Much of the T_CAPABILITY_ACK information is copied from
 * icmp_g_t_info_ack.  The current state of the stream is copied from
 * icmp_state.
 */
static void
icmp_capability_req(queue_t *q, mblk_t *mp)
{
	icmp_t			*icmp = Q_TO_ICMP(q);
	t_uscalar_t		cap_bits1;
	struct T_capability_ack	*tcap;

	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;

	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
	    mp->b_datap->db_type, T_CAPABILITY_ACK);
	if (!mp)
		return;

	tcap = (struct T_capability_ack *)mp->b_rptr;

	icmp_do_capability_ack(icmp, tcap, cap_bits1);

	qreply(q, mp);
}

/*
 * This routine responds to T_INFO_REQ messages.  It is called by icmp_wput.
 * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack.
 * The current state of the stream is copied from icmp_state.
 */
static void
icmp_info_req(queue_t *q, mblk_t *mp)
{
	icmp_t	*icmp = Q_TO_ICMP(q);

	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
	    T_INFO_ACK);
	if (!mp)
		return;
	icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp);
	qreply(q, mp);
}

/* For /dev/icmp aka AF_INET open */
static int
icmp_tpi_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
    int family)
{
	conn_t *connp;
	dev_t	conn_dev;
	icmp_stack_t *is;
	int	error;

	conn_dev = NULL;

	/* If the stream is already open, return immediately. */
	if (q->q_ptr != NULL)
		return (0);

	if (sflag == MODOPEN)
		return (EINVAL);

	/*
	 * Since ICMP is not used so heavily, allocating from the small
	 * arena should be sufficient.
	 */
	if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0) {
		return (EBUSY);
	}

	if (flag & SO_FALLBACK) {
		/*
		 * Non streams socket needs a stream to fallback to
		 */
		RD(q)->q_ptr = (void *)conn_dev;
		WR(q)->q_qinfo = &icmp_fallback_sock_winit;
		WR(q)->q_ptr = (void *)ip_minor_arena_sa;
		qprocson(q);
		return (0);
	}

	connp = icmp_open(family, credp, &error, KM_SLEEP);
	if (connp == NULL) {
		ASSERT(error != NULL);
		inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
		return (error);
	}

	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
	connp->conn_dev = conn_dev;
	connp->conn_minor_arena = ip_minor_arena_sa;

	is = connp->conn_icmp->icmp_is;

	/*
	 * Initialize the icmp_t structure for this stream.
	 */
	q->q_ptr = connp;
	WR(q)->q_ptr = connp;
	connp->conn_rq = q;
	connp->conn_wq = WR(q);

	if (connp->conn_icmp->icmp_family == AF_INET6) {
		/* Build initial header template for transmit */
		rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
		if ((error = icmp_build_hdrs(connp->conn_icmp)) != 0) {
			rw_exit(&connp->conn_icmp->icmp_rwlock);
			inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
			ipcl_conn_destroy(connp);
			return (error);
		}
		rw_exit(&connp->conn_icmp->icmp_rwlock);
	}


	q->q_hiwat = is->is_recv_hiwat;
	WR(q)->q_hiwat = is->is_xmit_hiwat;
	WR(q)->q_lowat = is->is_xmit_lowat;

	qprocson(q);

	/* Set the Stream head write offset. */
	(void) proto_set_tx_wroff(q, connp,
	    connp->conn_icmp->icmp_max_hdr_len + is->is_wroff_extra);
	(void) proto_set_rx_hiwat(connp->conn_rq, connp, q->q_hiwat);

	mutex_enter(&connp->conn_lock);
	connp->conn_state_flags &= ~CONN_INCIPIENT;
	mutex_exit(&connp->conn_lock);

	return (0);
}

/* For /dev/icmp4 aka AF_INET open */
static int
icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
{
	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET));
}

/* For /dev/icmp6 aka AF_INET6 open */
static int
icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
{
	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET6));
}

/*
 * This is the open routine for icmp.  It allocates a icmp_t structure for
 * the stream and, on the first open of the module, creates an ND table.
 */
/* ARGSUSED */
static conn_t *
icmp_open(int family, cred_t *credp, int *err, int flags)
{
	icmp_t	*icmp;
	conn_t *connp;
	zoneid_t zoneid;
	netstack_t *ns;
	icmp_stack_t *is;
	boolean_t isv6 = B_FALSE;

	*err = secpolicy_net_icmpaccess(credp);
	if (*err != 0)
		return (NULL);

	if (family == AF_INET6)
		isv6 = B_TRUE;
	ns = netstack_find_by_cred(credp);
	ASSERT(ns != NULL);
	is = ns->netstack_icmp;
	ASSERT(is != NULL);

	/*
	 * For exclusive stacks we set the zoneid to zero
	 * to make ICMP operate as if in the global zone.
	 */
	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
		zoneid = GLOBAL_ZONEID;
	else
		zoneid = crgetzoneid(credp);

	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);

	connp = ipcl_conn_create(IPCL_RAWIPCONN, flags, ns);
	icmp = connp->conn_icmp;
	icmp->icmp_v6dst = sin6_null;

	/*
	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
	 * done by netstack_find_by_cred()
	 */
	netstack_rele(ns);

	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
	ASSERT(connp->conn_ulp == IPPROTO_ICMP);
	ASSERT(connp->conn_icmp == icmp);
	ASSERT(icmp->icmp_connp == connp);

	/* Set the initial state of the stream and the privilege status. */
	icmp->icmp_state = TS_UNBND;
	if (isv6) {
		icmp->icmp_ipversion = IPV6_VERSION;
		icmp->icmp_family = AF_INET6;
		connp->conn_ulp = IPPROTO_ICMPV6;
		/* May be changed by a SO_PROTOTYPE socket option. */
		icmp->icmp_proto = IPPROTO_ICMPV6;
		icmp->icmp_checksum_off = 2;	/* Offset for icmp6_cksum */
		icmp->icmp_max_hdr_len = IPV6_HDR_LEN;
		icmp->icmp_ttl = (uint8_t)is->is_ipv6_hoplimit;
		connp->conn_af_isv6 = B_TRUE;
		connp->conn_flags |= IPCL_ISV6;
	} else {
		icmp->icmp_ipversion = IPV4_VERSION;
		icmp->icmp_family = AF_INET;
		/* May be changed by a SO_PROTOTYPE socket option. */
		icmp->icmp_proto = IPPROTO_ICMP;
		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH;
		icmp->icmp_ttl = (uint8_t)is->is_ipv4_ttl;
		connp->conn_af_isv6 = B_FALSE;
		connp->conn_flags &= ~IPCL_ISV6;
	}
	icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
	icmp->icmp_pending_op = -1;
	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
	connp->conn_zoneid = zoneid;

	/*
	 * If the caller has the process-wide flag set, then default to MAC
	 * exempt mode.  This allows read-down to unlabeled hosts.
	 */
	if (getpflags(NET_MAC_AWARE, credp) != 0)
		connp->conn_mac_exempt = B_TRUE;

	connp->conn_ulp_labeled = is_system_labeled();

	icmp->icmp_is = is;

	connp->conn_recv = icmp_input;
	crhold(credp);
	connp->conn_cred = credp;

	rw_exit(&icmp->icmp_rwlock);

	connp->conn_flow_cntrld = B_FALSE;
	return (connp);
}

/*
 * Which ICMP options OK to set through T_UNITDATA_REQ...
 */
/* ARGSUSED */
static boolean_t
icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
{
	return (B_TRUE);
}

/*
 * This routine gets default values of certain options whose default
 * values are maintained by protcol specific code
 */
/* ARGSUSED */
int
icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr)
{
	icmp_t *icmp = Q_TO_ICMP(q);
	icmp_stack_t *is = icmp->icmp_is;
	int *i1 = (int *)ptr;

	switch (level) {
	case IPPROTO_IP:
		switch (name) {
		case IP_MULTICAST_TTL:
			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
			return (sizeof (uchar_t));
		case IP_MULTICAST_LOOP:
			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
			return (sizeof (uchar_t));
		}
		break;
	case IPPROTO_IPV6:
		switch (name) {
		case IPV6_MULTICAST_HOPS:
			*i1 = IP_DEFAULT_MULTICAST_TTL;
			return (sizeof (int));
		case IPV6_MULTICAST_LOOP:
			*i1 = IP_DEFAULT_MULTICAST_LOOP;
			return (sizeof (int));
		case IPV6_UNICAST_HOPS:
			*i1 = is->is_ipv6_hoplimit;
			return (sizeof (int));
		}
		break;
	case IPPROTO_ICMPV6:
		switch (name) {
		case ICMP6_FILTER:
			/* Make it look like "pass all" */
			ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
			return (sizeof (icmp6_filter_t));
		}
		break;
	}
	return (-1);
}

/*
 * This routine retrieves the current status of socket options.
 * It returns the size of the option retrieved.
 */
int
icmp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
{
	icmp_t		*icmp = connp->conn_icmp;
	icmp_stack_t	*is = icmp->icmp_is;
	int		*i1 = (int *)ptr;
	ip6_pkt_t	*ipp = &icmp->icmp_sticky_ipp;
	int		ret = 0;

	ASSERT(RW_READ_HELD(&icmp->icmp_rwlock));
	switch (level) {
	case SOL_SOCKET:
		switch (name) {
		case SO_DEBUG:
			*i1 = icmp->icmp_debug;
			break;
		case SO_TYPE:
			*i1 = SOCK_RAW;
			break;
		case SO_PROTOTYPE:
			*i1 = icmp->icmp_proto;
			break;
		case SO_REUSEADDR:
			*i1 = icmp->icmp_reuseaddr;
			break;

		/*
		 * The following three items are available here,
		 * but are only meaningful to IP.
		 */
		case SO_DONTROUTE:
			*i1 = icmp->icmp_dontroute;
			break;
		case SO_USELOOPBACK:
			*i1 = icmp->icmp_useloopback;
			break;
		case SO_BROADCAST:
			*i1 = icmp->icmp_broadcast;
			break;

		case SO_SNDBUF:
			ASSERT(icmp->icmp_xmit_hiwat <= INT_MAX);
			*i1 = icmp->icmp_xmit_hiwat;
			break;
		case SO_RCVBUF:
			ASSERT(icmp->icmp_recv_hiwat <= INT_MAX);
			*i1 = icmp->icmp_recv_hiwat;
			break;
		case SO_DGRAM_ERRIND:
			*i1 = icmp->icmp_dgram_errind;
			break;
		case SO_TIMESTAMP:
			*i1 = icmp->icmp_timestamp;
			break;
		case SO_MAC_EXEMPT:
			*i1 = connp->conn_mac_exempt;
			break;
		case SO_DOMAIN:
			*i1 = icmp->icmp_family;
			break;

		/*
		 * Following four not meaningful for icmp
		 * Action is same as "default" to which we fallthrough
		 * so we keep them in comments.
		 * case SO_LINGER:
		 * case SO_KEEPALIVE:
		 * case SO_OOBINLINE:
		 * case SO_ALLZONES:
		 */
		default:
			ret = -1;
			goto done;
		}
		break;
	case IPPROTO_IP:
		/*
		 * Only allow IPv4 option processing on IPv4 sockets.
		 */
		if (icmp->icmp_family != AF_INET) {
			ret = -1;
			goto done;
		}

		switch (name) {
		case IP_OPTIONS:
		case T_IP_OPTIONS:
			/* Options are passed up with each packet */
			ret = 0;
			goto done;
		case IP_HDRINCL:
			*i1 = (int)icmp->icmp_hdrincl;
			break;
		case IP_TOS:
		case T_IP_TOS:
			*i1 = (int)icmp->icmp_type_of_service;
			break;
		case IP_TTL:
			*i1 = (int)icmp->icmp_ttl;
			break;
		case IP_MULTICAST_IF:
			/* 0 address if not set */
			*(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr;
			ret = sizeof (ipaddr_t);
			goto done;
		case IP_MULTICAST_TTL:
			*(uchar_t *)ptr = icmp->icmp_multicast_ttl;
			ret = sizeof (uchar_t);
			goto done;
		case IP_MULTICAST_LOOP:
			*ptr = connp->conn_multicast_loop;
			ret = sizeof (uint8_t);
			goto done;
		case IP_BOUND_IF:
			/* Zero if not set */
			*i1 = icmp->icmp_bound_if;
			break;	/* goto sizeof (int) option return */
		case IP_UNSPEC_SRC:
			*ptr = icmp->icmp_unspec_source;
			break;	/* goto sizeof (int) option return */
		case IP_RECVIF:
			*ptr = icmp->icmp_recvif;
			break;	/* goto sizeof (int) option return */
		case IP_BROADCAST_TTL:
			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
			return (sizeof (uchar_t));
		case IP_RECVPKTINFO:
			/*
			 * This also handles IP_PKTINFO.
			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
			 * Differentiation is based on the size of the argument
			 * passed in.
			 * This option is handled in IP which will return an
			 * error for IP_PKTINFO as it's not supported as a
			 * sticky option.
			 */
			ret = -EINVAL;
			goto done;
		/*
		 * Cannot "get" the value of following options
		 * at this level. Action is same as "default" to
		 * which we fallthrough so we keep them in comments.
		 *
		 * case IP_ADD_MEMBERSHIP:
		 * case IP_DROP_MEMBERSHIP:
		 * case IP_BLOCK_SOURCE:
		 * case IP_UNBLOCK_SOURCE:
		 * case IP_ADD_SOURCE_MEMBERSHIP:
		 * case IP_DROP_SOURCE_MEMBERSHIP:
		 * case MCAST_JOIN_GROUP:
		 * case MCAST_LEAVE_GROUP:
		 * case MCAST_BLOCK_SOURCE:
		 * case MCAST_UNBLOCK_SOURCE:
		 * case MCAST_JOIN_SOURCE_GROUP:
		 * case MCAST_LEAVE_SOURCE_GROUP:
		 * case MRT_INIT:
		 * case MRT_DONE:
		 * case MRT_ADD_VIF:
		 * case MRT_DEL_VIF:
		 * case MRT_ADD_MFC:
		 * case MRT_DEL_MFC:
		 * case MRT_VERSION:
		 * case MRT_ASSERT:
		 * case IP_SEC_OPT:
		 * case IP_NEXTHOP:
		 */
		default:
			ret = -1;
			goto done;
		}
		break;
	case IPPROTO_IPV6:
		/*
		 * Only allow IPv6 option processing on native IPv6 sockets.
		 */
		if (icmp->icmp_family != AF_INET6) {
			ret = -1;
			goto done;
		}
		switch (name) {
		case IPV6_UNICAST_HOPS:
			*i1 = (unsigned int)icmp->icmp_ttl;
			break;
		case IPV6_MULTICAST_IF:
			/* 0 index if not set */
			*i1 = icmp->icmp_multicast_if_index;
			break;
		case IPV6_MULTICAST_HOPS:
			*i1 = icmp->icmp_multicast_ttl;
			break;
		case IPV6_MULTICAST_LOOP:
			*i1 = connp->conn_multicast_loop;
			break;
		case IPV6_BOUND_IF:
			/* Zero if not set */
			*i1 = icmp->icmp_bound_if;
			break;
		case IPV6_UNSPEC_SRC:
			*i1 = icmp->icmp_unspec_source;
			break;
		case IPV6_CHECKSUM:
			/*
			 * Return offset or -1 if no checksum offset.
			 * Does not apply to IPPROTO_ICMPV6
			 */
			if (icmp->icmp_proto == IPPROTO_ICMPV6) {
				ret = -1;
				goto done;
			}

			if (icmp->icmp_raw_checksum) {
				*i1 = icmp->icmp_checksum_off;
			} else {
				*i1 = -1;
			}
			break;
		case IPV6_JOIN_GROUP:
		case IPV6_LEAVE_GROUP:
		case MCAST_JOIN_GROUP:
		case MCAST_LEAVE_GROUP:
		case MCAST_BLOCK_SOURCE:
		case MCAST_UNBLOCK_SOURCE:
		case MCAST_JOIN_SOURCE_GROUP:
		case MCAST_LEAVE_SOURCE_GROUP:
			/* cannot "get" the value for these */
			ret = -1;
			goto done;
		case IPV6_RECVPKTINFO:
			*i1 = icmp->icmp_ip_recvpktinfo;
			break;
		case IPV6_RECVTCLASS:
			*i1 = icmp->icmp_ipv6_recvtclass;
			break;
		case IPV6_RECVPATHMTU:
			*i1 = icmp->icmp_ipv6_recvpathmtu;
			break;
		case IPV6_V6ONLY:
			*i1 = 1;
			break;
		case IPV6_RECVHOPLIMIT:
			*i1 = icmp->icmp_ipv6_recvhoplimit;
			break;
		case IPV6_RECVHOPOPTS:
			*i1 = icmp->icmp_ipv6_recvhopopts;
			break;
		case IPV6_RECVDSTOPTS:
			*i1 = icmp->icmp_ipv6_recvdstopts;
			break;
		case _OLD_IPV6_RECVDSTOPTS:
			*i1 = icmp->icmp_old_ipv6_recvdstopts;
			break;
		case IPV6_RECVRTHDRDSTOPTS:
			*i1 = icmp->icmp_ipv6_recvrtdstopts;
			break;
		case IPV6_RECVRTHDR:
			*i1 = icmp->icmp_ipv6_recvrthdr;
			break;
		case IPV6_PKTINFO: {
			/* XXX assumes that caller has room for max size! */
			struct in6_pktinfo *pkti;

			pkti = (struct in6_pktinfo *)ptr;
			if (ipp->ipp_fields & IPPF_IFINDEX)
				pkti->ipi6_ifindex = ipp->ipp_ifindex;
			else
				pkti->ipi6_ifindex = 0;
			if (ipp->ipp_fields & IPPF_ADDR)
				pkti->ipi6_addr = ipp->ipp_addr;
			else
				pkti->ipi6_addr = ipv6_all_zeros;
			ret = sizeof (struct in6_pktinfo);
			goto done;
		}
		case IPV6_NEXTHOP: {
			sin6_t *sin6 = (sin6_t *)ptr;

			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
				return (0);
			*sin6 = sin6_null;
			sin6->sin6_family = AF_INET6;
			sin6->sin6_addr = ipp->ipp_nexthop;
			ret = (sizeof (sin6_t));
			goto done;
		}
		case IPV6_HOPOPTS:
			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
				return (0);
			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
				return (0);
			bcopy((char *)ipp->ipp_hopopts +
			    icmp->icmp_label_len_v6, ptr,
			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
			if (icmp->icmp_label_len_v6 > 0) {
				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
				ptr[1] = (ipp->ipp_hopoptslen -
				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
			}
			ret = (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
			goto done;
		case IPV6_RTHDRDSTOPTS:
			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
				return (0);
			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
			ret = ipp->ipp_rtdstoptslen;
			goto done;
		case IPV6_RTHDR:
			if (!(ipp->ipp_fields & IPPF_RTHDR))
				return (0);
			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
			ret = ipp->ipp_rthdrlen;
			goto done;
		case IPV6_DSTOPTS:
			if (!(ipp->ipp_fields & IPPF_DSTOPTS)) {
				ret = 0;
				goto done;
			}
			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
			ret = ipp->ipp_dstoptslen;
			goto done;
		case IPV6_PATHMTU:
			if (!(ipp->ipp_fields & IPPF_PATHMTU)) {
				ret = 0;
			} else {
				ret = ip_fill_mtuinfo(
				    &icmp->icmp_v6dst.sin6_addr, 0,
				    (struct ip6_mtuinfo *)ptr,
				    is->is_netstack);
			}
			goto done;
		case IPV6_TCLASS:
			if (ipp->ipp_fields & IPPF_TCLASS)
				*i1 = ipp->ipp_tclass;
			else
				*i1 = IPV6_FLOW_TCLASS(
				    IPV6_DEFAULT_VERS_AND_FLOW);
			break;
		default:
			ret = -1;
			goto done;
		}
		break;
	case IPPROTO_ICMPV6:
		/*
		 * Only allow IPv6 option processing on native IPv6 sockets.
		 */
		if (icmp->icmp_family != AF_INET6) {
			ret = -1;
		}

		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
			ret = -1;
		}

		switch (name) {
		case ICMP6_FILTER:
			if (icmp->icmp_filter == NULL) {
				/* Make it look like "pass all" */
				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
			} else {
				(void) bcopy(icmp->icmp_filter, ptr,
				    sizeof (icmp6_filter_t));
			}
			ret = sizeof (icmp6_filter_t);
			goto done;
		default:
			ret = -1;
			goto done;
		}
	default:
		ret = -1;
		goto done;
	}
	ret = sizeof (int);
done:
	return (ret);
}

/*
 * This routine retrieves the current status of socket options.
 * It returns the size of the option retrieved.
 */
int
icmp_tpi_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
{
	conn_t  *connp = Q_TO_CONN(q);
	icmp_t	*icmp = connp->conn_icmp;
	int 	err;

	rw_enter(&icmp->icmp_rwlock, RW_READER);
	err = icmp_opt_get(connp, level, name, ptr);
	rw_exit(&icmp->icmp_rwlock);
	return (err);
}

int
icmp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
    void *thisdg_attrs, boolean_t checkonly)
{

	int	*i1 = (int *)invalp;
	boolean_t onoff = (*i1 == 0) ? 0 : 1;
	icmp_t *icmp = connp->conn_icmp;
	icmp_stack_t *is = icmp->icmp_is;
	int	error;

	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
	/*
	 * For fixed length options, no sanity check
	 * of passed in length is done. It is assumed *_optcom_req()
	 * routines do the right thing.
	 */
	switch (level) {
	case SOL_SOCKET:
		switch (name) {
		case SO_DEBUG:
			if (!checkonly)
				icmp->icmp_debug = onoff;
			break;
		case SO_PROTOTYPE:
			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
			    secpolicy_net_rawaccess(cr) != 0) {
				*outlenp = 0;
				return (EACCES);
			}
			/* Can't use IPPROTO_RAW with IPv6 */
			if ((*i1 & 0xFF) == IPPROTO_RAW &&
			    icmp->icmp_family == AF_INET6) {
				*outlenp = 0;
				return (EPROTONOSUPPORT);
			}
			if (checkonly) {
				/* T_CHECK case */
				*(int *)outvalp = (*i1 & 0xFF);
				break;
			}
			icmp->icmp_proto = *i1 & 0xFF;
			if ((icmp->icmp_proto == IPPROTO_RAW ||
			    icmp->icmp_proto == IPPROTO_IGMP) &&
			    icmp->icmp_family == AF_INET)
				icmp->icmp_hdrincl = 1;
			else
				icmp->icmp_hdrincl = 0;

			if (icmp->icmp_family == AF_INET6 &&
			    icmp->icmp_proto == IPPROTO_ICMPV6) {
				/* Set offset for icmp6_cksum */
				icmp->icmp_raw_checksum = 0;
				icmp->icmp_checksum_off = 2;
			}
			if (icmp->icmp_proto == IPPROTO_UDP ||
			    icmp->icmp_proto == IPPROTO_TCP ||
			    icmp->icmp_proto == IPPROTO_SCTP) {
				icmp->icmp_no_tp_cksum = 1;
				icmp->icmp_sticky_ipp.ipp_fields |=
				    IPPF_NO_CKSUM;
			} else {
				icmp->icmp_no_tp_cksum = 0;
				icmp->icmp_sticky_ipp.ipp_fields &=
				    ~IPPF_NO_CKSUM;
			}

			if (icmp->icmp_filter != NULL &&
			    icmp->icmp_proto != IPPROTO_ICMPV6) {
				kmem_free(icmp->icmp_filter,
				    sizeof (icmp6_filter_t));
				icmp->icmp_filter = NULL;
			}

			/* Rebuild the header template */
			error = icmp_build_hdrs(icmp);
			if (error != 0) {
				*outlenp = 0;
				return (error);
			}

			/*
			 * For SCTP, we don't use icmp_bind_proto() for
			 * raw socket binding.  Note that we do not need
			 * to set *outlenp.
			 * FIXME: how does SCTP work?
			 */
			if (icmp->icmp_proto == IPPROTO_SCTP)
				return (0);

			*outlenp = sizeof (int);
			*(int *)outvalp = *i1 & 0xFF;

			/* Drop lock across the bind operation */
			rw_exit(&icmp->icmp_rwlock);
			(void) icmp_bind_proto(connp);
			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
			return (0);
		case SO_REUSEADDR:
			if (!checkonly) {
				icmp->icmp_reuseaddr = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;

		/*
		 * The following three items are available here,
		 * but are only meaningful to IP.
		 */
		case SO_DONTROUTE:
			if (!checkonly) {
				icmp->icmp_dontroute = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case SO_USELOOPBACK:
			if (!checkonly) {
				icmp->icmp_useloopback = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case SO_BROADCAST:
			if (!checkonly) {
				icmp->icmp_broadcast = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;

		case SO_SNDBUF:
			if (*i1 > is->is_max_buf) {
				*outlenp = 0;
				return (ENOBUFS);
			}
			if (!checkonly) {
				if (!IPCL_IS_NONSTR(connp)) {
					connp->conn_wq->q_hiwat = *i1;
				}
				icmp->icmp_xmit_hiwat = *i1;
			}
			break;
		case SO_RCVBUF:
			if (*i1 > is->is_max_buf) {
				*outlenp = 0;
				return (ENOBUFS);
			}
			if (!checkonly) {
				icmp->icmp_recv_hiwat = *i1;
				rw_exit(&icmp->icmp_rwlock);
				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
				    *i1);
				rw_enter(&icmp->icmp_rwlock, RW_WRITER);
			}
			break;
		case SO_DGRAM_ERRIND:
			if (!checkonly)
				icmp->icmp_dgram_errind = onoff;
			break;
		case SO_ALLZONES:
			/*
			 * "soft" error (negative)
			 * option not handled at this level
			 * Note: Do not modify *outlenp
			 */
			return (-EINVAL);
		case SO_TIMESTAMP:
			if (!checkonly) {
				icmp->icmp_timestamp = onoff;
			}
			break;
		case SO_MAC_EXEMPT:
			/*
			 * "soft" error (negative)
			 * option not handled at this level
			 * Note: Do not modify *outlenp
			 */
			return (-EINVAL);
		case SO_RCVTIMEO:
		case SO_SNDTIMEO:
			/*
			 * Pass these two options in order for third part
			 * protocol usage. Here just return directly.
			 */
			return (0);
		/*
		 * Following three not meaningful for icmp
		 * Action is same as "default" so we keep them
		 * in comments.
		 * case SO_LINGER:
		 * case SO_KEEPALIVE:
		 * case SO_OOBINLINE:
		 */
		default:
			*outlenp = 0;
			return (EINVAL);
		}
		break;
	case IPPROTO_IP:
		/*
		 * Only allow IPv4 option processing on IPv4 sockets.
		 */
		if (icmp->icmp_family != AF_INET) {
			*outlenp = 0;
			return (ENOPROTOOPT);
		}
		switch (name) {
		case IP_OPTIONS:
		case T_IP_OPTIONS:
			/* Save options for use by IP. */
			if ((inlen & 0x3) ||
			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
				*outlenp = 0;
				return (EINVAL);
			}
			if (checkonly)
				break;

			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
			    &icmp->icmp_ip_snd_options_len,
			    icmp->icmp_label_len, invalp, inlen)) {
				*outlenp = 0;
				return (ENOMEM);
			}

			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
			    icmp->icmp_ip_snd_options_len;
			rw_exit(&icmp->icmp_rwlock);
			(void) proto_set_tx_wroff(connp->conn_rq == NULL ? NULL:
			    RD(connp->conn_rq), connp,
			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
			break;
		case IP_HDRINCL:
			if (!checkonly)
				icmp->icmp_hdrincl = onoff;
			break;
		case IP_TOS:
		case T_IP_TOS:
			if (!checkonly) {
				icmp->icmp_type_of_service = (uint8_t)*i1;
			}
			break;
		case IP_TTL:
			if (!checkonly) {
				icmp->icmp_ttl = (uint8_t)*i1;
			}
			break;
		case IP_MULTICAST_IF:
			/*
			 * TODO should check OPTMGMT reply and undo this if
			 * there is an error.
			 */
			if (!checkonly) {
				icmp->icmp_multicast_if_addr = *i1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IP_MULTICAST_TTL:
			if (!checkonly)
				icmp->icmp_multicast_ttl = *invalp;
			break;
		case IP_MULTICAST_LOOP:
			if (!checkonly) {
				connp->conn_multicast_loop =
				    (*invalp == 0) ? 0 : 1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IP_BOUND_IF:
			if (!checkonly) {
				icmp->icmp_bound_if = *i1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IP_UNSPEC_SRC:
			if (!checkonly) {
				icmp->icmp_unspec_source = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IP_BROADCAST_TTL:
			if (!checkonly)
				connp->conn_broadcast_ttl = *invalp;
			break;
		case IP_RECVIF:
			if (!checkonly) {
				icmp->icmp_recvif = onoff;
			}
			/*
			 * pass to ip
			 */
			return (-EINVAL);
		case IP_PKTINFO: {
			/*
			 * This also handles IP_RECVPKTINFO.
			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
			 * Differentiation is based on the size of the argument
			 * passed in.
			 */
			struct in_pktinfo *pktinfop;
			ip4_pkt_t *attr_pktinfop;

			if (checkonly)
				break;

			if (inlen == sizeof (int)) {
				/*
				 * This is IP_RECVPKTINFO option.
				 * Keep a local copy of wether this option is
				 * set or not and pass it down to IP for
				 * processing.
				 */
				icmp->icmp_ip_recvpktinfo = onoff;
				return (-EINVAL);
			}


			if (inlen != sizeof (struct in_pktinfo)) {
				return (EINVAL);
			}

			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
			    == NULL) {
				/*
				 * sticky option is not supported
				 */
				return (EINVAL);
			}

			pktinfop = (struct in_pktinfo *)invalp;

			/*
			 * Atleast one of the values should be specified
			 */
			if (pktinfop->ipi_ifindex == 0 &&
			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
				return (EINVAL);
			}

			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
		}
			break;
		case IP_ADD_MEMBERSHIP:
		case IP_DROP_MEMBERSHIP:
		case IP_BLOCK_SOURCE:
		case IP_UNBLOCK_SOURCE:
		case IP_ADD_SOURCE_MEMBERSHIP:
		case IP_DROP_SOURCE_MEMBERSHIP:
		case MCAST_JOIN_GROUP:
		case MCAST_LEAVE_GROUP:
		case MCAST_BLOCK_SOURCE:
		case MCAST_UNBLOCK_SOURCE:
		case MCAST_JOIN_SOURCE_GROUP:
		case MCAST_LEAVE_SOURCE_GROUP:
		case MRT_INIT:
		case MRT_DONE:
		case MRT_ADD_VIF:
		case MRT_DEL_VIF:
		case MRT_ADD_MFC:
		case MRT_DEL_MFC:
		case MRT_VERSION:
		case MRT_ASSERT:
		case IP_SEC_OPT:
		case IP_NEXTHOP:
			/*
			 * "soft" error (negative)
			 * option not handled at this level
			 * Note: Do not modify *outlenp
			 */
			return (-EINVAL);
		default:
			*outlenp = 0;
			return (EINVAL);
		}
		break;
	case IPPROTO_IPV6: {
		ip6_pkt_t		*ipp;
		boolean_t		sticky;

		if (icmp->icmp_family != AF_INET6) {
			*outlenp = 0;
			return (ENOPROTOOPT);
		}
		/*
		 * Deal with both sticky options and ancillary data
		 */
		if (thisdg_attrs == NULL) {
			/* sticky options, or none */
			ipp = &icmp->icmp_sticky_ipp;
			sticky = B_TRUE;
		} else {
			/* ancillary data */
			ipp = (ip6_pkt_t *)thisdg_attrs;
			sticky = B_FALSE;
		}

		switch (name) {
		case IPV6_MULTICAST_IF:
			if (!checkonly) {
				icmp->icmp_multicast_if_index = *i1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_UNICAST_HOPS:
			/* -1 means use default */
			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
				*outlenp = 0;
				return (EINVAL);
			}
			if (!checkonly) {
				if (*i1 == -1) {
					icmp->icmp_ttl = ipp->ipp_unicast_hops =
					    is->is_ipv6_hoplimit;
					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
					/* Pass modified value to IP. */
					*i1 = ipp->ipp_hoplimit;
				} else {
					icmp->icmp_ttl = ipp->ipp_unicast_hops =
					    (uint8_t)*i1;
					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
				}
				/* Rebuild the header template */
				error = icmp_build_hdrs(icmp);
				if (error != 0) {
					*outlenp = 0;
					return (error);
				}
			}
			break;
		case IPV6_MULTICAST_HOPS:
			/* -1 means use default */
			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
				*outlenp = 0;
				return (EINVAL);
			}
			if (!checkonly) {
				if (*i1 == -1) {
					icmp->icmp_multicast_ttl =
					    ipp->ipp_multicast_hops =
					    IP_DEFAULT_MULTICAST_TTL;
					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
					/* Pass modified value to IP. */
					*i1 = icmp->icmp_multicast_ttl;
				} else {
					icmp->icmp_multicast_ttl =
					    ipp->ipp_multicast_hops =
					    (uint8_t)*i1;
					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
				}
			}
			break;
		case IPV6_MULTICAST_LOOP:
			if (*i1 != 0 && *i1 != 1) {
				*outlenp = 0;
				return (EINVAL);
			}
			if (!checkonly) {
				connp->conn_multicast_loop = *i1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_CHECKSUM:
			/*
			 * Integer offset into the user data of where the
			 * checksum is located.
			 * Offset of -1 disables option.
			 * Does not apply to IPPROTO_ICMPV6.
			 */
			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
				*outlenp = 0;
				return (EINVAL);
			}
			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
				/* Negative or not 16 bit aligned offset */
				*outlenp = 0;
				return (EINVAL);
			}
			if (checkonly)
				break;

			if (*i1 == -1) {
				icmp->icmp_raw_checksum = 0;
				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
			} else {
				icmp->icmp_raw_checksum = 1;
				icmp->icmp_checksum_off = *i1;
				ipp->ipp_fields |= IPPF_RAW_CKSUM;
			}
			/* Rebuild the header template */
			error = icmp_build_hdrs(icmp);
			if (error != 0) {
				*outlenp = 0;
				return (error);
			}
			break;
		case IPV6_JOIN_GROUP:
		case IPV6_LEAVE_GROUP:
		case MCAST_JOIN_GROUP:
		case MCAST_LEAVE_GROUP:
		case MCAST_BLOCK_SOURCE:
		case MCAST_UNBLOCK_SOURCE:
		case MCAST_JOIN_SOURCE_GROUP:
		case MCAST_LEAVE_SOURCE_GROUP:
			/*
			 * "soft" error (negative)
			 * option not handled at this level
			 * Note: Do not modify *outlenp
			 */
			return (-EINVAL);
		case IPV6_BOUND_IF:
			if (!checkonly) {
				icmp->icmp_bound_if = *i1;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_UNSPEC_SRC:
			if (!checkonly) {
				icmp->icmp_unspec_source = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVTCLASS:
			if (!checkonly) {
				icmp->icmp_ipv6_recvtclass = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		/*
		 * Set boolean switches for ancillary data delivery
		 */
		case IPV6_RECVPKTINFO:
			if (!checkonly) {
				icmp->icmp_ip_recvpktinfo = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVPATHMTU:
			if (!checkonly) {
				icmp->icmp_ipv6_recvpathmtu = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVHOPLIMIT:
			if (!checkonly) {
				icmp->icmp_ipv6_recvhoplimit = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVHOPOPTS:
			if (!checkonly) {
				icmp->icmp_ipv6_recvhopopts = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVDSTOPTS:
			if (!checkonly) {
				icmp->icmp_ipv6_recvdstopts = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case _OLD_IPV6_RECVDSTOPTS:
			if (!checkonly)
				icmp->icmp_old_ipv6_recvdstopts = onoff;
			break;
		case IPV6_RECVRTHDRDSTOPTS:
			if (!checkonly) {
				icmp->icmp_ipv6_recvrtdstopts = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		case IPV6_RECVRTHDR:
			if (!checkonly) {
				icmp->icmp_ipv6_recvrthdr = onoff;
				PASS_OPT_TO_IP(connp);
			}
			break;
		/*
		 * Set sticky options or ancillary data.
		 * If sticky options, (re)build any extension headers
		 * that might be needed as a result.
		 */
		case IPV6_PKTINFO:
			/*
			 * The source address and ifindex are verified
			 * in ip_opt_set(). For ancillary data the
			 * source address is checked in ip_wput_v6.
			 */
			if (inlen != 0 && inlen !=
			    sizeof (struct in6_pktinfo)) {
				return (EINVAL);
			}
			if (checkonly)
				break;

			if (inlen == 0) {
				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
				ipp->ipp_sticky_ignored |=
				    (IPPF_IFINDEX|IPPF_ADDR);
			} else {
				struct in6_pktinfo *pkti;

				pkti = (struct in6_pktinfo *)invalp;
				ipp->ipp_ifindex = pkti->ipi6_ifindex;
				ipp->ipp_addr = pkti->ipi6_addr;
				if (ipp->ipp_ifindex != 0)