xref: /titanic_52/usr/src/uts/common/inet/ip/icmp.c (revision e11c3f44f531fdff80941ce57c065d2ae861cefc)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/stropts.h>
30 #include <sys/strlog.h>
31 #include <sys/strsun.h>
32 #define	_SUN_TPI_VERSION 2
33 #include <sys/tihdr.h>
34 #include <sys/timod.h>
35 #include <sys/ddi.h>
36 #include <sys/sunddi.h>
37 #include <sys/strsubr.h>
38 #include <sys/cmn_err.h>
39 #include <sys/debug.h>
40 #include <sys/kmem.h>
41 #include <sys/policy.h>
42 #include <sys/priv.h>
43 #include <sys/zone.h>
44 #include <sys/time.h>
45 
46 #include <sys/sockio.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/isa_defs.h>
50 #include <sys/suntpi.h>
51 #include <sys/xti_inet.h>
52 #include <sys/netstack.h>
53 
54 #include <net/route.h>
55 #include <net/if.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/ip6.h>
59 #include <netinet/icmp6.h>
60 #include <inet/common.h>
61 #include <inet/ip.h>
62 #include <inet/ip6.h>
63 #include <inet/proto_set.h>
64 #include <inet/nd.h>
65 #include <inet/optcom.h>
66 #include <inet/snmpcom.h>
67 #include <inet/kstatcom.h>
68 #include <inet/rawip_impl.h>
69 
70 #include <netinet/ip_mroute.h>
71 #include <inet/tcp.h>
72 #include <net/pfkeyv2.h>
73 #include <inet/ipsec_info.h>
74 #include <inet/ipclassifier.h>
75 
76 #include <sys/tsol/label.h>
77 #include <sys/tsol/tnet.h>
78 
79 #include <inet/ip_ire.h>
80 #include <inet/ip_if.h>
81 
82 #include <inet/ip_impl.h>
83 #include <sys/disp.h>
84 
85 /*
86  * Synchronization notes:
87  *
88  * RAWIP is MT and uses the usual kernel synchronization primitives. There is
89  * locks, which is icmp_rwlock. We also use conn_lock when updating things
90  * which affect the IP classifier lookup.
91  * The lock order is icmp_rwlock -> conn_lock.
92  *
93  * The icmp_rwlock:
94  * This protects most of the other fields in the icmp_t. The exact list of
95  * fields which are protected by each of the above locks is documented in
96  * the icmp_t structure definition.
97  *
98  * Plumbing notes:
99  * ICMP is always a device driver. For compatibility with mibopen() code
100  * it is possible to I_PUSH "icmp", but that results in pushing a passthrough
101  * dummy module.
102  */
103 
104 static void	icmp_addr_req(queue_t *q, mblk_t *mp);
105 static void	icmp_tpi_bind(queue_t *q, mblk_t *mp);
106 static int	icmp_bind_proto(conn_t *connp);
107 static int	icmp_build_hdrs(icmp_t *icmp);
108 static void	icmp_capability_req(queue_t *q, mblk_t *mp);
109 static int	icmp_close(queue_t *q, int flags);
110 static void	icmp_tpi_connect(queue_t *q, mblk_t *mp);
111 static void	icmp_tpi_disconnect(queue_t *q, mblk_t *mp);
112 static void	icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
113 		    int sys_error);
114 static void	icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
115 		    t_scalar_t t_error, int sys_error);
116 static void	icmp_icmp_error(conn_t *connp, mblk_t *mp);
117 static void	icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp);
118 static void	icmp_info_req(queue_t *q, mblk_t *mp);
119 static void	icmp_input(void *, mblk_t *, void *);
120 static conn_t 	*icmp_open(int family, cred_t *credp, int *err, int flags);
121 static int	icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
122 		    cred_t *credp);
123 static int	icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
124 		    cred_t *credp);
125 static int	icmp_unitdata_opt_process(queue_t *q, mblk_t *mp,
126 		    int *errorp, void *thisdg_attrs);
127 static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
128 int		icmp_opt_set(conn_t *connp, uint_t optset_context,
129 		    int level, int name, uint_t inlen,
130 		    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
131 		    void *thisdg_attrs, cred_t *cr);
132 int		icmp_opt_get(conn_t *connp, int level, int name,
133 		    uchar_t *ptr);
134 static int	icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
135 static boolean_t icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt);
136 static int	icmp_param_set(queue_t *q, mblk_t *mp, char *value,
137 		    caddr_t cp, cred_t *cr);
138 static int	icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
139 		    uchar_t *ptr, int len);
140 static int	icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
141 		    cred_t *cr);
142 static void	icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err);
143 static void	icmp_tpi_unbind(queue_t *q, mblk_t *mp);
144 static void	icmp_wput(queue_t *q, mblk_t *mp);
145 static void	icmp_wput_fallback(queue_t *q, mblk_t *mp);
146 static int	raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp,
147 		    sin6_t *sin6, ip6_pkt_t *ipp);
148 static int	raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp,
149 		    ipaddr_t v4dst, ip4_pkt_t *pktinfop);
150 static void	icmp_wput_other(queue_t *q, mblk_t *mp);
151 static void	icmp_wput_iocdata(queue_t *q, mblk_t *mp);
152 static void	icmp_wput_restricted(queue_t *q, mblk_t *mp);
153 
154 static void	*rawip_stack_init(netstackid_t stackid, netstack_t *ns);
155 static void	rawip_stack_fini(netstackid_t stackid, void *arg);
156 
157 static void	*rawip_kstat_init(netstackid_t stackid);
158 static void	rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp);
159 static int	rawip_kstat_update(kstat_t *kp, int rw);
160 static void	rawip_stack_shutdown(netstackid_t stackid, void *arg);
161 static int	rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa,
162 		    uint_t *salenp);
163 static int	rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa,
164 		    uint_t *salenp);
165 
166 int		rawip_getsockname(sock_lower_handle_t, struct sockaddr *,
167 		    socklen_t *, cred_t *);
168 int		rawip_getpeername(sock_lower_handle_t, struct sockaddr *,
169 		    socklen_t *, cred_t *);
170 
171 static struct module_info icmp_mod_info =  {
172 	5707, "icmp", 1, INFPSZ, 512, 128
173 };
174 
175 /*
176  * Entry points for ICMP as a device.
177  * We have separate open functions for the /dev/icmp and /dev/icmp6 devices.
178  */
179 static struct qinit icmprinitv4 = {
180 	NULL, NULL, icmp_openv4, icmp_close, NULL, &icmp_mod_info
181 };
182 
183 static struct qinit icmprinitv6 = {
184 	NULL, NULL, icmp_openv6, icmp_close, NULL, &icmp_mod_info
185 };
186 
187 static struct qinit icmpwinit = {
188 	(pfi_t)icmp_wput, NULL, NULL, NULL, NULL, &icmp_mod_info
189 };
190 
191 /* ICMP entry point during fallback */
192 static struct qinit icmp_fallback_sock_winit = {
193 	(pfi_t)icmp_wput_fallback, NULL, NULL, NULL, NULL, &icmp_mod_info
194 };
195 
196 /* For AF_INET aka /dev/icmp */
197 struct streamtab icmpinfov4 = {
198 	&icmprinitv4, &icmpwinit
199 };
200 
201 /* For AF_INET6 aka /dev/icmp6 */
202 struct streamtab icmpinfov6 = {
203 	&icmprinitv6, &icmpwinit
204 };
205 
206 static sin_t	sin_null;	/* Zero address for quick clears */
207 static sin6_t	sin6_null;	/* Zero address for quick clears */
208 
209 /* Default structure copied into T_INFO_ACK messages */
210 static struct T_info_ack icmp_g_t_info_ack = {
211 	T_INFO_ACK,
212 	IP_MAXPACKET,	 /* TSDU_size.  icmp allows maximum size messages. */
213 	T_INVALID,	/* ETSDU_size.  icmp does not support expedited data. */
214 	T_INVALID,	/* CDATA_size. icmp does not support connect data. */
215 	T_INVALID,	/* DDATA_size. icmp does not support disconnect data. */
216 	0,		/* ADDR_size - filled in later. */
217 	0,		/* OPT_size - not initialized here */
218 	IP_MAXPACKET,	/* TIDU_size.  icmp allows maximum size messages. */
219 	T_CLTS,		/* SERV_type.  icmp supports connection-less. */
220 	TS_UNBND,	/* CURRENT_state.  This is set from icmp_state. */
221 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
222 };
223 
224 /*
225  * Table of ND variables supported by icmp.  These are loaded into is_nd
226  * when the stack instance is created.
227  * All of these are alterable, within the min/max values given, at run time.
228  */
229 static icmpparam_t	icmp_param_arr[] = {
230 	/* min	max	value	name */
231 	{ 0,	128,	32,	"icmp_wroff_extra" },
232 	{ 1,	255,	255,	"icmp_ipv4_ttl" },
233 	{ 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS,	"icmp_ipv6_hoplimit"},
234 	{ 0,	1,	1,	"icmp_bsd_compat" },
235 	{ 4096,	65536,	8192,	"icmp_xmit_hiwat"},
236 	{ 0,	65536,	1024,	"icmp_xmit_lowat"},
237 	{ 4096,	65536,	8192,	"icmp_recv_hiwat"},
238 	{ 65536, 1024*1024*1024, 256*1024,	"icmp_max_buf"},
239 };
240 #define	is_wroff_extra			is_param_arr[0].icmp_param_value
241 #define	is_ipv4_ttl			is_param_arr[1].icmp_param_value
242 #define	is_ipv6_hoplimit		is_param_arr[2].icmp_param_value
243 #define	is_bsd_compat			is_param_arr[3].icmp_param_value
244 #define	is_xmit_hiwat			is_param_arr[4].icmp_param_value
245 #define	is_xmit_lowat			is_param_arr[5].icmp_param_value
246 #define	is_recv_hiwat			is_param_arr[6].icmp_param_value
247 #define	is_max_buf			is_param_arr[7].icmp_param_value
248 
249 static int rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len);
250 static int rawip_do_connect(conn_t *connp, const struct sockaddr *sa,
251     socklen_t len);
252 static void rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error);
253 
254 /*
255  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
256  * passed to icmp_wput.
257  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP
258  * protocol type placed in the message following the address. A T_BIND_ACK
259  * message is returned by ip_bind_v4/v6.
260  */
261 static void
262 icmp_tpi_bind(queue_t *q, mblk_t *mp)
263 {
264 	int	error;
265 	struct sockaddr *sa;
266 	struct T_bind_req *tbr;
267 	socklen_t	len;
268 	sin_t	*sin;
269 	sin6_t	*sin6;
270 	icmp_t		*icmp;
271 	conn_t	*connp = Q_TO_CONN(q);
272 	mblk_t *mp1;
273 
274 	icmp = connp->conn_icmp;
275 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
276 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
277 		    "icmp_bind: bad req, len %u",
278 		    (uint_t)(mp->b_wptr - mp->b_rptr));
279 		icmp_err_ack(q, mp, TPROTO, 0);
280 		return;
281 	}
282 
283 	if (icmp->icmp_state != TS_UNBND) {
284 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
285 		    "icmp_bind: bad state, %d", icmp->icmp_state);
286 		icmp_err_ack(q, mp, TOUTSTATE, 0);
287 		return;
288 	}
289 
290 	/*
291 	 * Reallocate the message to make sure we have enough room for an
292 	 * address and the protocol type.
293 	 */
294 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
295 	if (!mp1) {
296 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
297 		return;
298 	}
299 	mp = mp1;
300 
301 	/* Reset the message type in preparation for shipping it back. */
302 	DB_TYPE(mp) = M_PCPROTO;
303 	tbr = (struct T_bind_req *)mp->b_rptr;
304 	len = tbr->ADDR_length;
305 	switch (len) {
306 	case 0:	/* request for a generic port */
307 		tbr->ADDR_offset = sizeof (struct T_bind_req);
308 		if (icmp->icmp_family == AF_INET) {
309 			tbr->ADDR_length = sizeof (sin_t);
310 			sin = (sin_t *)&tbr[1];
311 			*sin = sin_null;
312 			sin->sin_family = AF_INET;
313 			mp->b_wptr = (uchar_t *)&sin[1];
314 			sa = (struct sockaddr *)sin;
315 			len = sizeof (sin_t);
316 		} else {
317 			ASSERT(icmp->icmp_family == AF_INET6);
318 			tbr->ADDR_length = sizeof (sin6_t);
319 			sin6 = (sin6_t *)&tbr[1];
320 			*sin6 = sin6_null;
321 			sin6->sin6_family = AF_INET6;
322 			mp->b_wptr = (uchar_t *)&sin6[1];
323 			sa = (struct sockaddr *)sin6;
324 			len = sizeof (sin6_t);
325 		}
326 		break;
327 
328 	case sizeof (sin_t):	/* Complete IPv4 address */
329 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
330 		    sizeof (sin_t));
331 		break;
332 
333 	case sizeof (sin6_t):	/* Complete IPv6 address */
334 		sa = (struct sockaddr *)mi_offset_param(mp,
335 		    tbr->ADDR_offset, sizeof (sin6_t));
336 		break;
337 
338 	default:
339 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
340 		    "icmp_bind: bad ADDR_length %d", tbr->ADDR_length);
341 		icmp_err_ack(q, mp, TBADADDR, 0);
342 		return;
343 	}
344 
345 	error = rawip_do_bind(connp, sa, len);
346 done:
347 	ASSERT(mp->b_cont == NULL);
348 	if (error != 0) {
349 		if (error > 0) {
350 			icmp_err_ack(q, mp, TSYSERR, error);
351 		} else {
352 			icmp_err_ack(q, mp, -error, 0);
353 		}
354 	} else {
355 		tbr->PRIM_type = T_BIND_ACK;
356 		qreply(q, mp);
357 	}
358 }
359 
360 static int
361 rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len)
362 {
363 	sin_t		*sin;
364 	sin6_t		*sin6;
365 	icmp_t		*icmp;
366 	int		error = 0;
367 	mblk_t		*ire_mp;
368 
369 
370 	icmp = connp->conn_icmp;
371 
372 	if (sa == NULL || !OK_32PTR((char *)sa)) {
373 		return (EINVAL);
374 	}
375 
376 	/*
377 	 * The state must be TS_UNBND. TPI mandates that users must send
378 	 * TPI primitives only 1 at a time and wait for the response before
379 	 * sending the next primitive.
380 	 */
381 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
382 	if (icmp->icmp_state != TS_UNBND || icmp->icmp_pending_op != -1) {
383 		error = -TOUTSTATE;
384 		goto done;
385 	}
386 
387 	ASSERT(len != 0);
388 	switch (len) {
389 	case sizeof (sin_t):    /* Complete IPv4 address */
390 		sin = (sin_t *)sa;
391 		if (sin->sin_family != AF_INET ||
392 		    icmp->icmp_family != AF_INET) {
393 			/* TSYSERR, EAFNOSUPPORT */
394 			error = EAFNOSUPPORT;
395 			goto done;
396 		}
397 		break;
398 	case sizeof (sin6_t): /* Complete IPv6 address */
399 		sin6 = (sin6_t *)sa;
400 		if (sin6->sin6_family != AF_INET6 ||
401 		    icmp->icmp_family != AF_INET6) {
402 			/* TSYSERR, EAFNOSUPPORT */
403 			error = EAFNOSUPPORT;
404 			goto done;
405 		}
406 		/* No support for mapped addresses on raw sockets */
407 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
408 			/* TSYSERR, EADDRNOTAVAIL */
409 			error = EADDRNOTAVAIL;
410 			goto done;
411 		}
412 		break;
413 
414 	default:
415 		/* TBADADDR */
416 		error = EADDRNOTAVAIL;
417 		goto done;
418 	}
419 
420 	icmp->icmp_pending_op = T_BIND_REQ;
421 	icmp->icmp_state = TS_IDLE;
422 
423 	/*
424 	 * Copy the source address into our icmp structure.  This address
425 	 * may still be zero; if so, ip will fill in the correct address
426 	 * each time an outbound packet is passed to it.
427 	 * If we are binding to a broadcast or multicast address then
428 	 * rawip_post_ip_bind_connect will clear the source address.
429 	 */
430 
431 	if (icmp->icmp_family == AF_INET) {
432 		ASSERT(sin != NULL);
433 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
434 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr,
435 		    &icmp->icmp_v6src);
436 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
437 		    icmp->icmp_ip_snd_options_len;
438 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
439 	} else {
440 		int error;
441 
442 		ASSERT(sin6 != NULL);
443 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
444 		icmp->icmp_v6src = sin6->sin6_addr;
445 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
446 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
447 
448 		/* Rebuild the header template */
449 		error = icmp_build_hdrs(icmp);
450 		if (error != 0) {
451 			icmp->icmp_pending_op = -1;
452 			/*
453 			 * TSYSERR
454 			 */
455 			goto done;
456 		}
457 	}
458 
459 	ire_mp = NULL;
460 	if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) {
461 		/*
462 		 * request an IRE if src not 0 (INADDR_ANY)
463 		 */
464 		ire_mp = allocb(sizeof (ire_t), BPRI_HI);
465 		if (ire_mp == NULL) {
466 			icmp->icmp_pending_op = -1;
467 			error = ENOMEM;
468 			goto done;
469 		}
470 		DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;
471 	}
472 done:
473 	rw_exit(&icmp->icmp_rwlock);
474 	if (error != 0)
475 		return (error);
476 
477 	if (icmp->icmp_family == AF_INET6) {
478 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
479 		    &sin6->sin6_addr, sin6->sin6_port, B_TRUE);
480 	} else {
481 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
482 		    sin->sin_addr.s_addr, sin->sin_port, B_TRUE);
483 	}
484 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
485 	return (error);
486 }
487 
488 static void
489 rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error)
490 {
491 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
492 	if (icmp->icmp_state == TS_UNBND) {
493 		/*
494 		 * not yet bound - bind sent by icmp_bind_proto.
495 		 */
496 		rw_exit(&icmp->icmp_rwlock);
497 		return;
498 	}
499 	ASSERT(icmp->icmp_pending_op != -1);
500 	icmp->icmp_pending_op = -1;
501 
502 	if (error != 0) {
503 		if (icmp->icmp_state == TS_DATA_XFER) {
504 			/* Connect failed */
505 			/* Revert back to the bound source */
506 			icmp->icmp_v6src = icmp->icmp_bound_v6src;
507 			icmp->icmp_state = TS_IDLE;
508 			if (icmp->icmp_family == AF_INET6)
509 				(void) icmp_build_hdrs(icmp);
510 		} else {
511 			V6_SET_ZERO(icmp->icmp_v6src);
512 			V6_SET_ZERO(icmp->icmp_bound_v6src);
513 			icmp->icmp_state = TS_UNBND;
514 			if (icmp->icmp_family == AF_INET6)
515 				(void) icmp_build_hdrs(icmp);
516 		}
517 	} else {
518 		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
519 			ire_t *ire;
520 
521 			ire = (ire_t *)ire_mp->b_rptr;
522 			/*
523 			 * If a broadcast/multicast address was bound set
524 			 * the source address to 0.
525 			 * This ensures no datagrams with broadcast address
526 			 * as source address are emitted (which would violate
527 			 * RFC1122 - Hosts requirements)
528 			 * Note: we get IRE_BROADCAST for IPv6
529 			 * to "mark" a multicast local address.
530 			 */
531 
532 
533 			if (ire->ire_type == IRE_BROADCAST &&
534 			    icmp->icmp_state != TS_DATA_XFER) {
535 				/*
536 				 * This was just a local bind to a
537 				 * MC/broadcast addr
538 				 */
539 				V6_SET_ZERO(icmp->icmp_v6src);
540 				if (icmp->icmp_family == AF_INET6)
541 					(void) icmp_build_hdrs(icmp);
542 			}
543 		}
544 
545 	}
546 	rw_exit(&icmp->icmp_rwlock);
547 	if (ire_mp != NULL)
548 		freeb(ire_mp);
549 }
550 
551 /*
552  * Send message to IP to just bind to the protocol.
553  */
554 static int
555 icmp_bind_proto(conn_t *connp)
556 {
557 	icmp_t	*icmp;
558 	int	error;
559 
560 	icmp = connp->conn_icmp;
561 
562 	if (icmp->icmp_family == AF_INET6)
563 		error = ip_proto_bind_laddr_v6(connp, NULL, icmp->icmp_proto,
564 		    &sin6_null.sin6_addr, 0, B_TRUE);
565 	else
566 		error = ip_proto_bind_laddr_v4(connp, NULL, icmp->icmp_proto,
567 		    sin_null.sin_addr.s_addr, 0, B_TRUE);
568 
569 	rawip_post_ip_bind_connect(icmp, NULL, error);
570 	return (error);
571 }
572 
573 static void
574 icmp_tpi_connect(queue_t *q, mblk_t *mp)
575 {
576 	conn_t	*connp = Q_TO_CONN(q);
577 	struct T_conn_req	*tcr;
578 	icmp_t	*icmp;
579 	struct sockaddr *sa;
580 	socklen_t len;
581 	int error;
582 
583 	icmp = connp->conn_icmp;
584 	tcr = (struct T_conn_req *)mp->b_rptr;
585 	/* Sanity checks */
586 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
587 		icmp_err_ack(q, mp, TPROTO, 0);
588 		return;
589 	}
590 
591 	if (tcr->OPT_length != 0) {
592 		icmp_err_ack(q, mp, TBADOPT, 0);
593 		return;
594 	}
595 
596 	len = tcr->DEST_length;
597 
598 	switch (len) {
599 	default:
600 		icmp_err_ack(q, mp, TBADADDR, 0);
601 		return;
602 	case sizeof (sin_t):
603 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
604 		    sizeof (sin_t));
605 		break;
606 	case sizeof (sin6_t):
607 		sa = (struct sockaddr *)mi_offset_param(mp,
608 		    tcr->DEST_offset, sizeof (sin6_t));
609 		break;
610 	}
611 
612 	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
613 	if (error != 0) {
614 		icmp_err_ack(q, mp, TSYSERR, error);
615 		return;
616 	}
617 
618 	error = rawip_do_connect(connp, sa, len);
619 	if (error != 0) {
620 		if (error < 0) {
621 			icmp_err_ack(q, mp, -error, 0);
622 		} else {
623 			icmp_err_ack(q, mp, 0, error);
624 		}
625 	} else {
626 		mblk_t *mp1;
627 
628 		/*
629 		 * We have to send a connection confirmation to
630 		 * keep TLI happy.
631 		 */
632 		if (icmp->icmp_family == AF_INET) {
633 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
634 			    sizeof (sin_t), NULL, 0);
635 		} else {
636 			ASSERT(icmp->icmp_family == AF_INET6);
637 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
638 			    sizeof (sin6_t), NULL, 0);
639 		}
640 		if (mp1 == NULL) {
641 			rw_exit(&icmp->icmp_rwlock);
642 			icmp_err_ack(q, mp, TSYSERR, ENOMEM);
643 			return;
644 		}
645 
646 		/*
647 		 * Send ok_ack for T_CONN_REQ
648 		 */
649 		mp = mi_tpi_ok_ack_alloc(mp);
650 		if (mp == NULL) {
651 			/* Unable to reuse the T_CONN_REQ for the ack. */
652 			freemsg(mp1);
653 			icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
654 			return;
655 		}
656 		putnext(connp->conn_rq, mp);
657 		putnext(connp->conn_rq, mp1);
658 	}
659 }
660 
661 static int
662 rawip_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len)
663 {
664 	icmp_t	*icmp;
665 	sin_t	*sin;
666 	sin6_t	*sin6;
667 	mblk_t  *ire_mp;
668 	int	error;
669 	ipaddr_t	v4dst;
670 	in6_addr_t	v6dst;
671 
672 	icmp = connp->conn_icmp;
673 
674 	if (sa == NULL || !OK_32PTR((char *)sa)) {
675 		return (EINVAL);
676 	}
677 
678 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
679 	if (ire_mp == NULL)
680 		return (ENOMEM);
681 	DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;
682 
683 
684 	ASSERT(sa != NULL && len != 0);
685 
686 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
687 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
688 		rw_exit(&icmp->icmp_rwlock);
689 		freeb(ire_mp);
690 		return (-TOUTSTATE);
691 	}
692 
693 	switch (len) {
694 	case sizeof (sin_t):
695 		sin = (sin_t *)sa;
696 
697 		ASSERT(icmp->icmp_family == AF_INET);
698 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
699 
700 		v4dst = sin->sin_addr.s_addr;
701 		/*
702 		 * Interpret a zero destination to mean loopback.
703 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
704 		 * generate the T_CONN_CON.
705 		 */
706 		if (v4dst == INADDR_ANY) {
707 			v4dst = htonl(INADDR_LOOPBACK);
708 		}
709 
710 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
711 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
712 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
713 		    icmp->icmp_ip_snd_options_len;
714 		icmp->icmp_v6dst.sin6_addr = v6dst;
715 		icmp->icmp_v6dst.sin6_family = AF_INET6;
716 		icmp->icmp_v6dst.sin6_flowinfo = 0;
717 		icmp->icmp_v6dst.sin6_port = 0;
718 
719 		/*
720 		 * If the destination address is multicast and
721 		 * an outgoing multicast interface has been set,
722 		 * use the address of that interface as our
723 		 * source address if no source address has been set.
724 		 */
725 		if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY &&
726 		    CLASSD(v4dst) &&
727 		    icmp->icmp_multicast_if_addr != INADDR_ANY) {
728 			IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr,
729 			    &icmp->icmp_v6src);
730 		}
731 		break;
732 	case sizeof (sin6_t):
733 		sin6 = (sin6_t *)sa;
734 
735 		/* No support for mapped addresses on raw sockets */
736 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
737 			rw_exit(&icmp->icmp_rwlock);
738 			freeb(ire_mp);
739 			return (EADDRNOTAVAIL);
740 		}
741 
742 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
743 		ASSERT(icmp->icmp_family == AF_INET6);
744 
745 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
746 
747 		icmp->icmp_v6dst = *sin6;
748 		icmp->icmp_v6dst.sin6_port = 0;
749 
750 		/*
751 		 * Interpret a zero destination to mean loopback.
752 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
753 		 * generate the T_CONN_CON.
754 		 */
755 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6dst.sin6_addr)) {
756 			icmp->icmp_v6dst.sin6_addr = ipv6_loopback;
757 		}
758 		/*
759 		 * If the destination address is multicast and
760 		 * an outgoing multicast interface has been set,
761 		 * then the ip bind logic will pick the correct source
762 		 * address (i.e. matching the outgoing multicast interface).
763 		 */
764 		break;
765 	}
766 
767 	icmp->icmp_pending_op = T_CONN_REQ;
768 
769 	if (icmp->icmp_state == TS_DATA_XFER) {
770 		/* Already connected - clear out state */
771 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
772 		icmp->icmp_state = TS_IDLE;
773 	}
774 
775 	icmp->icmp_state = TS_DATA_XFER;
776 	rw_exit(&icmp->icmp_rwlock);
777 
778 	if (icmp->icmp_family == AF_INET6) {
779 		error = ip_proto_bind_connected_v6(connp, &ire_mp,
780 		    icmp->icmp_proto, &icmp->icmp_v6src, 0,
781 		    &icmp->icmp_v6dst.sin6_addr,
782 		    NULL, sin6->sin6_port, B_TRUE, B_TRUE);
783 	} else {
784 		error = ip_proto_bind_connected_v4(connp, &ire_mp,
785 		    icmp->icmp_proto, &V4_PART_OF_V6(icmp->icmp_v6src), 0,
786 		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr), sin->sin_port,
787 		    B_TRUE, B_TRUE);
788 	}
789 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
790 	return (error);
791 }
792 
793 static void
794 icmp_close_free(conn_t *connp)
795 {
796 	icmp_t *icmp = connp->conn_icmp;
797 
798 	/* If there are any options associated with the stream, free them. */
799 	if (icmp->icmp_ip_snd_options != NULL) {
800 		mi_free((char *)icmp->icmp_ip_snd_options);
801 		icmp->icmp_ip_snd_options = NULL;
802 		icmp->icmp_ip_snd_options_len = 0;
803 	}
804 
805 	if (icmp->icmp_filter != NULL) {
806 		kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t));
807 		icmp->icmp_filter = NULL;
808 	}
809 
810 	/* Free memory associated with sticky options */
811 	if (icmp->icmp_sticky_hdrs_len != 0) {
812 		kmem_free(icmp->icmp_sticky_hdrs,
813 		    icmp->icmp_sticky_hdrs_len);
814 		icmp->icmp_sticky_hdrs = NULL;
815 		icmp->icmp_sticky_hdrs_len = 0;
816 	}
817 	ip6_pkt_free(&icmp->icmp_sticky_ipp);
818 
819 	/*
820 	 * Clear any fields which the kmem_cache constructor clears.
821 	 * Only icmp_connp needs to be preserved.
822 	 * TBD: We should make this more efficient to avoid clearing
823 	 * everything.
824 	 */
825 	ASSERT(icmp->icmp_connp == connp);
826 	bzero(icmp, sizeof (icmp_t));
827 	icmp->icmp_connp = connp;
828 }
829 
830 static int
831 rawip_do_close(conn_t *connp)
832 {
833 	ASSERT(connp != NULL && IPCL_IS_RAWIP(connp));
834 
835 	ip_quiesce_conn(connp);
836 
837 	if (!IPCL_IS_NONSTR(connp)) {
838 		qprocsoff(connp->conn_rq);
839 	}
840 
841 	ASSERT(connp->conn_icmp->icmp_fallback_queue_head == NULL &&
842 	    connp->conn_icmp->icmp_fallback_queue_tail == NULL);
843 	icmp_close_free(connp);
844 
845 	/*
846 	 * Now we are truly single threaded on this stream, and can
847 	 * delete the things hanging off the connp, and finally the connp.
848 	 * We removed this connp from the fanout list, it cannot be
849 	 * accessed thru the fanouts, and we already waited for the
850 	 * conn_ref to drop to 0. We are already in close, so
851 	 * there cannot be any other thread from the top. qprocsoff
852 	 * has completed, and service has completed or won't run in
853 	 * future.
854 	 */
855 	ASSERT(connp->conn_ref == 1);
856 
857 	if (!IPCL_IS_NONSTR(connp)) {
858 		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
859 	} else {
860 		ip_free_helper_stream(connp);
861 	}
862 
863 	connp->conn_ref--;
864 	ipcl_conn_destroy(connp);
865 
866 	return (0);
867 }
868 
869 static int
870 icmp_close(queue_t *q, int flags)
871 {
872 	conn_t  *connp;
873 
874 	if (flags & SO_FALLBACK) {
875 		/*
876 		 * stream is being closed while in fallback
877 		 * simply free the resources that were allocated
878 		 */
879 		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
880 		qprocsoff(q);
881 		goto done;
882 	}
883 
884 	connp = Q_TO_CONN(q);
885 	(void) rawip_do_close(connp);
886 done:
887 	q->q_ptr = WR(q)->q_ptr = NULL;
888 	return (0);
889 }
890 
891 /*
892  * This routine handles each T_DISCON_REQ message passed to icmp
893  * as an indicating that ICMP is no longer connected. This results
894  * in sending a T_BIND_REQ to IP to restore the binding to just
895  * the local address.
896  *
897  * The disconnect completes in rawip_post_ip_bind_connect.
898  */
899 static int
900 icmp_do_disconnect(conn_t *connp)
901 {
902 	icmp_t	*icmp;
903 	mblk_t	*ire_mp;
904 	int error;
905 
906 	icmp = connp->conn_icmp;
907 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
908 	if (icmp->icmp_state != TS_DATA_XFER || icmp->icmp_pending_op != -1) {
909 		rw_exit(&icmp->icmp_rwlock);
910 		return (-TOUTSTATE);
911 	}
912 	icmp->icmp_pending_op = T_DISCON_REQ;
913 	icmp->icmp_v6src = icmp->icmp_bound_v6src;
914 	icmp->icmp_state = TS_IDLE;
915 
916 
917 	if (icmp->icmp_family == AF_INET6) {
918 		/* Rebuild the header template */
919 		error = icmp_build_hdrs(icmp);
920 		if (error != 0) {
921 			icmp->icmp_pending_op = -1;
922 			rw_exit(&icmp->icmp_rwlock);
923 			return (error);
924 		}
925 	}
926 
927 	rw_exit(&icmp->icmp_rwlock);
928 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
929 	if (ire_mp == NULL) {
930 		return (ENOMEM);
931 	}
932 
933 	if (icmp->icmp_family == AF_INET6) {
934 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
935 		    &icmp->icmp_bound_v6src, 0, B_TRUE);
936 	} else {
937 
938 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
939 		    V4_PART_OF_V6(icmp->icmp_bound_v6src), 0, B_TRUE);
940 	}
941 
942 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
943 
944 	return (error);
945 }
946 
947 static void
948 icmp_tpi_disconnect(queue_t *q, mblk_t *mp)
949 {
950 	conn_t	*connp = Q_TO_CONN(q);
951 	int	error;
952 
953 	/*
954 	 * Allocate the largest primitive we need to send back
955 	 * T_error_ack is > than T_ok_ack
956 	 */
957 	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
958 	if (mp == NULL) {
959 		/* Unable to reuse the T_DISCON_REQ for the ack. */
960 		icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
961 		return;
962 	}
963 
964 	error = icmp_do_disconnect(connp);
965 
966 	if (error != 0) {
967 		if (error > 0) {
968 			icmp_err_ack(q, mp, 0, error);
969 		} else {
970 			icmp_err_ack(q, mp, -error, 0);
971 		}
972 	} else {
973 		mp = mi_tpi_ok_ack_alloc(mp);
974 		ASSERT(mp != NULL);
975 		qreply(q, mp);
976 	}
977 
978 }
979 
980 static int
981 icmp_disconnect(conn_t *connp)
982 {
983 	int	error;
984 	icmp_t	*icmp = connp->conn_icmp;
985 
986 	icmp->icmp_dgram_errind = B_FALSE;
987 
988 	error = icmp_do_disconnect(connp);
989 
990 	if (error < 0)
991 		error = proto_tlitosyserr(-error);
992 	return (error);
993 }
994 
995 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
996 static void
997 icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
998 {
999 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1000 		qreply(q, mp);
1001 }
1002 
1003 /* Shorthand to generate and send TPI error acks to our client */
1004 static void
1005 icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
1006     t_scalar_t t_error, int sys_error)
1007 {
1008 	struct T_error_ack	*teackp;
1009 
1010 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1011 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1012 		teackp = (struct T_error_ack *)mp->b_rptr;
1013 		teackp->ERROR_prim = primitive;
1014 		teackp->TLI_error = t_error;
1015 		teackp->UNIX_error = sys_error;
1016 		qreply(q, mp);
1017 	}
1018 }
1019 
1020 /*
1021  * icmp_icmp_error is called by icmp_input to process ICMP
1022  * messages passed up by IP.
1023  * Generates the appropriate permanent (non-transient) errors.
1024  * Assumes that IP has pulled up everything up to and including
1025  * the ICMP header.
1026  */
1027 static void
1028 icmp_icmp_error(conn_t *connp, mblk_t *mp)
1029 {
1030 	icmph_t *icmph;
1031 	ipha_t	*ipha;
1032 	int	iph_hdr_length;
1033 	sin_t	sin;
1034 	mblk_t	*mp1;
1035 	int	error = 0;
1036 	icmp_t	*icmp = connp->conn_icmp;
1037 
1038 	ipha = (ipha_t *)mp->b_rptr;
1039 
1040 	ASSERT(OK_32PTR(mp->b_rptr));
1041 
1042 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1043 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1044 		icmp_icmp_error_ipv6(connp, mp);
1045 		return;
1046 	}
1047 
1048 	/*
1049 	 * icmp does not support v4 mapped addresses
1050 	 * so we can never be here for a V6 socket
1051 	 * i.e. icmp_family == AF_INET6
1052 	 */
1053 	ASSERT((IPH_HDR_VERSION(ipha) == IPV4_VERSION) &&
1054 	    (icmp->icmp_family == AF_INET));
1055 
1056 	ASSERT(icmp->icmp_family == AF_INET);
1057 
1058 	/* Skip past the outer IP and ICMP headers */
1059 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1060 	icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]);
1061 	ipha = (ipha_t *)&icmph[1];
1062 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1063 
1064 	switch (icmph->icmph_type) {
1065 	case ICMP_DEST_UNREACHABLE:
1066 		switch (icmph->icmph_code) {
1067 		case ICMP_FRAGMENTATION_NEEDED:
1068 			/*
1069 			 * IP has already adjusted the path MTU.
1070 			 */
1071 			break;
1072 		case ICMP_PORT_UNREACHABLE:
1073 		case ICMP_PROTOCOL_UNREACHABLE:
1074 			error = ECONNREFUSED;
1075 			break;
1076 		default:
1077 			/* Transient errors */
1078 			break;
1079 		}
1080 		break;
1081 	default:
1082 		/* Transient errors */
1083 		break;
1084 	}
1085 	if (error == 0) {
1086 		freemsg(mp);
1087 		return;
1088 	}
1089 
1090 	/*
1091 	 * Deliver T_UDERROR_IND when the application has asked for it.
1092 	 * The socket layer enables this automatically when connected.
1093 	 */
1094 	if (!icmp->icmp_dgram_errind) {
1095 		freemsg(mp);
1096 		return;
1097 	}
1098 
1099 	sin = sin_null;
1100 	sin.sin_family = AF_INET;
1101 	sin.sin_addr.s_addr = ipha->ipha_dst;
1102 	if (IPCL_IS_NONSTR(connp)) {
1103 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1104 		if (icmp->icmp_state == TS_DATA_XFER) {
1105 			if (sin.sin_addr.s_addr ==
1106 			    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr)) {
1107 				rw_exit(&icmp->icmp_rwlock);
1108 				(*connp->conn_upcalls->su_set_error)
1109 				    (connp->conn_upper_handle, error);
1110 				goto done;
1111 			}
1112 		} else {
1113 			icmp->icmp_delayed_error = error;
1114 			*((sin_t *)&icmp->icmp_delayed_addr) = sin;
1115 		}
1116 		rw_exit(&icmp->icmp_rwlock);
1117 	} else {
1118 
1119 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL,
1120 		    0, error);
1121 		if (mp1 != NULL)
1122 			putnext(connp->conn_rq, mp1);
1123 	}
1124 done:
1125 	freemsg(mp);
1126 }
1127 
1128 /*
1129  * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6
1130  * for IPv6 packets.
1131  * Send permanent (non-transient) errors upstream.
1132  * Assumes that IP has pulled up all the extension headers as well
1133  * as the ICMPv6 header.
1134  */
1135 static void
1136 icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
1137 {
1138 	icmp6_t		*icmp6;
1139 	ip6_t		*ip6h, *outer_ip6h;
1140 	uint16_t	iph_hdr_length;
1141 	uint8_t		*nexthdrp;
1142 	sin6_t		sin6;
1143 	mblk_t		*mp1;
1144 	int		error = 0;
1145 	icmp_t		*icmp = connp->conn_icmp;
1146 
1147 	outer_ip6h = (ip6_t *)mp->b_rptr;
1148 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1149 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1150 	else
1151 		iph_hdr_length = IPV6_HDR_LEN;
1152 
1153 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1154 	ip6h = (ip6_t *)&icmp6[1];
1155 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1156 		freemsg(mp);
1157 		return;
1158 	}
1159 
1160 	switch (icmp6->icmp6_type) {
1161 	case ICMP6_DST_UNREACH:
1162 		switch (icmp6->icmp6_code) {
1163 		case ICMP6_DST_UNREACH_NOPORT:
1164 			error = ECONNREFUSED;
1165 			break;
1166 		case ICMP6_DST_UNREACH_ADMIN:
1167 		case ICMP6_DST_UNREACH_NOROUTE:
1168 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1169 		case ICMP6_DST_UNREACH_ADDR:
1170 			/* Transient errors */
1171 			break;
1172 		default:
1173 			break;
1174 		}
1175 		break;
1176 	case ICMP6_PACKET_TOO_BIG: {
1177 		struct T_unitdata_ind	*tudi;
1178 		struct T_opthdr		*toh;
1179 		size_t			udi_size;
1180 		mblk_t			*newmp;
1181 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1182 		    sizeof (struct ip6_mtuinfo);
1183 		sin6_t			*sin6;
1184 		struct ip6_mtuinfo	*mtuinfo;
1185 
1186 		/*
1187 		 * If the application has requested to receive path mtu
1188 		 * information, send up an empty message containing an
1189 		 * IPV6_PATHMTU ancillary data item.
1190 		 */
1191 		if (!icmp->icmp_ipv6_recvpathmtu)
1192 			break;
1193 
1194 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1195 		    opt_length;
1196 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1197 			BUMP_MIB(&icmp->icmp_is->is_rawip_mib, rawipInErrors);
1198 			break;
1199 		}
1200 
1201 		/*
1202 		 * newmp->b_cont is left to NULL on purpose.  This is an
1203 		 * empty message containing only ancillary data.
1204 		 */
1205 		newmp->b_datap->db_type = M_PROTO;
1206 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1207 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1208 		tudi->PRIM_type = T_UNITDATA_IND;
1209 		tudi->SRC_length = sizeof (sin6_t);
1210 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1211 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1212 		tudi->OPT_length = opt_length;
1213 
1214 		sin6 = (sin6_t *)&tudi[1];
1215 		bzero(sin6, sizeof (sin6_t));
1216 		sin6->sin6_family = AF_INET6;
1217 		sin6->sin6_addr = icmp->icmp_v6dst.sin6_addr;
1218 
1219 		toh = (struct T_opthdr *)&sin6[1];
1220 		toh->level = IPPROTO_IPV6;
1221 		toh->name = IPV6_PATHMTU;
1222 		toh->len = opt_length;
1223 		toh->status = 0;
1224 
1225 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1226 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1227 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1228 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1229 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1230 		/*
1231 		 * We've consumed everything we need from the original
1232 		 * message.  Free it, then send our empty message.
1233 		 */
1234 		freemsg(mp);
1235 		if (!IPCL_IS_NONSTR(connp)) {
1236 			putnext(connp->conn_rq, newmp);
1237 		} else {
1238 			(*connp->conn_upcalls->su_recv)
1239 			    (connp->conn_upper_handle, newmp, 0, 0, &error,
1240 			    NULL);
1241 			ASSERT(error == 0);
1242 		}
1243 		return;
1244 	}
1245 	case ICMP6_TIME_EXCEEDED:
1246 		/* Transient errors */
1247 		break;
1248 	case ICMP6_PARAM_PROB:
1249 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1250 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1251 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1252 		    (uchar_t *)nexthdrp) {
1253 			error = ECONNREFUSED;
1254 			break;
1255 		}
1256 		break;
1257 	}
1258 	if (error == 0) {
1259 		freemsg(mp);
1260 		return;
1261 	}
1262 
1263 	/*
1264 	 * Deliver T_UDERROR_IND when the application has asked for it.
1265 	 * The socket layer enables this automatically when connected.
1266 	 */
1267 	if (!icmp->icmp_dgram_errind) {
1268 		freemsg(mp);
1269 		return;
1270 	}
1271 
1272 	sin6 = sin6_null;
1273 	sin6.sin6_family = AF_INET6;
1274 	sin6.sin6_addr = ip6h->ip6_dst;
1275 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1276 
1277 	if (IPCL_IS_NONSTR(connp)) {
1278 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1279 		if (icmp->icmp_state == TS_DATA_XFER) {
1280 			if (IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1281 			    &icmp->icmp_v6dst.sin6_addr)) {
1282 				rw_exit(&icmp->icmp_rwlock);
1283 				(*connp->conn_upcalls->su_set_error)
1284 				    (connp->conn_upper_handle, error);
1285 				goto done;
1286 			}
1287 		} else {
1288 			icmp->icmp_delayed_error = error;
1289 			*((sin6_t *)&icmp->icmp_delayed_addr) = sin6;
1290 		}
1291 		rw_exit(&icmp->icmp_rwlock);
1292 	} else {
1293 
1294 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1295 		    NULL, 0, error);
1296 		if (mp1 != NULL)
1297 			putnext(connp->conn_rq, mp1);
1298 	}
1299 done:
1300 	freemsg(mp);
1301 }
1302 
1303 /*
1304  * This routine responds to T_ADDR_REQ messages.  It is called by icmp_wput.
1305  * The local address is filled in if endpoint is bound. The remote address
1306  * is filled in if remote address has been precified ("connected endpoint")
1307  * (The concept of connected CLTS sockets is alien to published TPI
1308  *  but we support it anyway).
1309  */
1310 static void
1311 icmp_addr_req(queue_t *q, mblk_t *mp)
1312 {
1313 	icmp_t	*icmp = Q_TO_ICMP(q);
1314 	mblk_t	*ackmp;
1315 	struct T_addr_ack *taa;
1316 
1317 	/* Make it large enough for worst case */
1318 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1319 	    2 * sizeof (sin6_t), 1);
1320 	if (ackmp == NULL) {
1321 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
1322 		return;
1323 	}
1324 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1325 
1326 	bzero(taa, sizeof (struct T_addr_ack));
1327 	ackmp->b_wptr = (uchar_t *)&taa[1];
1328 
1329 	taa->PRIM_type = T_ADDR_ACK;
1330 	ackmp->b_datap->db_type = M_PCPROTO;
1331 	rw_enter(&icmp->icmp_rwlock, RW_READER);
1332 	/*
1333 	 * Note: Following code assumes 32 bit alignment of basic
1334 	 * data structures like sin_t and struct T_addr_ack.
1335 	 */
1336 	if (icmp->icmp_state != TS_UNBND) {
1337 		/*
1338 		 * Fill in local address
1339 		 */
1340 		taa->LOCADDR_offset = sizeof (*taa);
1341 		if (icmp->icmp_family == AF_INET) {
1342 			sin_t	*sin;
1343 
1344 			taa->LOCADDR_length = sizeof (sin_t);
1345 			sin = (sin_t *)&taa[1];
1346 			/* Fill zeroes and then intialize non-zero fields */
1347 			*sin = sin_null;
1348 			sin->sin_family = AF_INET;
1349 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
1350 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1351 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src,
1352 				    sin->sin_addr.s_addr);
1353 			} else {
1354 				/*
1355 				 * INADDR_ANY
1356 				 * icmp_v6src is not set, we might be bound to
1357 				 * broadcast/multicast. Use icmp_bound_v6src as
1358 				 * local address instead (that could
1359 				 * also still be INADDR_ANY)
1360 				 */
1361 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src,
1362 				    sin->sin_addr.s_addr);
1363 			}
1364 			ackmp->b_wptr = (uchar_t *)&sin[1];
1365 		} else {
1366 			sin6_t	*sin6;
1367 
1368 			ASSERT(icmp->icmp_family == AF_INET6);
1369 			taa->LOCADDR_length = sizeof (sin6_t);
1370 			sin6 = (sin6_t *)&taa[1];
1371 			/* Fill zeroes and then intialize non-zero fields */
1372 			*sin6 = sin6_null;
1373 			sin6->sin6_family = AF_INET6;
1374 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1375 				sin6->sin6_addr = icmp->icmp_v6src;
1376 			} else {
1377 				/*
1378 				 * UNSPECIFIED
1379 				 * icmp_v6src is not set, we might be bound to
1380 				 * broadcast/multicast. Use icmp_bound_v6src as
1381 				 * local address instead (that could
1382 				 * also still be UNSPECIFIED)
1383 				 */
1384 				sin6->sin6_addr = icmp->icmp_bound_v6src;
1385 			}
1386 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1387 		}
1388 	}
1389 	rw_exit(&icmp->icmp_rwlock);
1390 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1391 	qreply(q, ackmp);
1392 }
1393 
1394 static void
1395 icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp)
1396 {
1397 	*tap = icmp_g_t_info_ack;
1398 
1399 	if (icmp->icmp_family == AF_INET6)
1400 		tap->ADDR_size = sizeof (sin6_t);
1401 	else
1402 		tap->ADDR_size = sizeof (sin_t);
1403 	tap->CURRENT_state = icmp->icmp_state;
1404 	tap->OPT_size = icmp_max_optsize;
1405 }
1406 
1407 static void
1408 icmp_do_capability_ack(icmp_t *icmp, struct T_capability_ack *tcap,
1409     t_uscalar_t cap_bits1)
1410 {
1411 	tcap->CAP_bits1 = 0;
1412 
1413 	if (cap_bits1 & TC1_INFO) {
1414 		icmp_copy_info(&tcap->INFO_ack, icmp);
1415 		tcap->CAP_bits1 |= TC1_INFO;
1416 	}
1417 }
1418 
1419 /*
1420  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1421  * icmp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1422  * icmp_g_t_info_ack.  The current state of the stream is copied from
1423  * icmp_state.
1424  */
1425 static void
1426 icmp_capability_req(queue_t *q, mblk_t *mp)
1427 {
1428 	icmp_t			*icmp = Q_TO_ICMP(q);
1429 	t_uscalar_t		cap_bits1;
1430 	struct T_capability_ack	*tcap;
1431 
1432 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1433 
1434 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1435 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1436 	if (!mp)
1437 		return;
1438 
1439 	tcap = (struct T_capability_ack *)mp->b_rptr;
1440 
1441 	icmp_do_capability_ack(icmp, tcap, cap_bits1);
1442 
1443 	qreply(q, mp);
1444 }
1445 
1446 /*
1447  * This routine responds to T_INFO_REQ messages.  It is called by icmp_wput.
1448  * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack.
1449  * The current state of the stream is copied from icmp_state.
1450  */
1451 static void
1452 icmp_info_req(queue_t *q, mblk_t *mp)
1453 {
1454 	icmp_t	*icmp = Q_TO_ICMP(q);
1455 
1456 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1457 	    T_INFO_ACK);
1458 	if (!mp)
1459 		return;
1460 	icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp);
1461 	qreply(q, mp);
1462 }
1463 
1464 /* For /dev/icmp aka AF_INET open */
1465 static int
1466 icmp_tpi_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1467     int family)
1468 {
1469 	conn_t *connp;
1470 	dev_t	conn_dev;
1471 	icmp_stack_t *is;
1472 	int	error;
1473 
1474 	conn_dev = NULL;
1475 
1476 	/* If the stream is already open, return immediately. */
1477 	if (q->q_ptr != NULL)
1478 		return (0);
1479 
1480 	if (sflag == MODOPEN)
1481 		return (EINVAL);
1482 
1483 	/*
1484 	 * Since ICMP is not used so heavily, allocating from the small
1485 	 * arena should be sufficient.
1486 	 */
1487 	if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0) {
1488 		return (EBUSY);
1489 	}
1490 
1491 	if (flag & SO_FALLBACK) {
1492 		/*
1493 		 * Non streams socket needs a stream to fallback to
1494 		 */
1495 		RD(q)->q_ptr = (void *)conn_dev;
1496 		WR(q)->q_qinfo = &icmp_fallback_sock_winit;
1497 		WR(q)->q_ptr = (void *)ip_minor_arena_sa;
1498 		qprocson(q);
1499 		return (0);
1500 	}
1501 
1502 	connp = icmp_open(family, credp, &error, KM_SLEEP);
1503 	if (connp == NULL) {
1504 		ASSERT(error != NULL);
1505 		inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
1506 		return (error);
1507 	}
1508 
1509 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1510 	connp->conn_dev = conn_dev;
1511 	connp->conn_minor_arena = ip_minor_arena_sa;
1512 
1513 	is = connp->conn_icmp->icmp_is;
1514 
1515 	/*
1516 	 * Initialize the icmp_t structure for this stream.
1517 	 */
1518 	q->q_ptr = connp;
1519 	WR(q)->q_ptr = connp;
1520 	connp->conn_rq = q;
1521 	connp->conn_wq = WR(q);
1522 
1523 	if (connp->conn_icmp->icmp_family == AF_INET6) {
1524 		/* Build initial header template for transmit */
1525 		rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
1526 		if ((error = icmp_build_hdrs(connp->conn_icmp)) != 0) {
1527 			rw_exit(&connp->conn_icmp->icmp_rwlock);
1528 			inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
1529 			ipcl_conn_destroy(connp);
1530 			return (error);
1531 		}
1532 		rw_exit(&connp->conn_icmp->icmp_rwlock);
1533 	}
1534 
1535 
1536 	q->q_hiwat = is->is_recv_hiwat;
1537 	WR(q)->q_hiwat = is->is_xmit_hiwat;
1538 	WR(q)->q_lowat = is->is_xmit_lowat;
1539 
1540 	qprocson(q);
1541 
1542 	/* Set the Stream head write offset. */
1543 	(void) proto_set_tx_wroff(q, connp,
1544 	    connp->conn_icmp->icmp_max_hdr_len + is->is_wroff_extra);
1545 	(void) proto_set_rx_hiwat(connp->conn_rq, connp, q->q_hiwat);
1546 
1547 	mutex_enter(&connp->conn_lock);
1548 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1549 	mutex_exit(&connp->conn_lock);
1550 
1551 	return (0);
1552 }
1553 
1554 /* For /dev/icmp4 aka AF_INET open */
1555 static int
1556 icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1557 {
1558 	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET));
1559 }
1560 
1561 /* For /dev/icmp6 aka AF_INET6 open */
1562 static int
1563 icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1564 {
1565 	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET6));
1566 }
1567 
1568 /*
1569  * This is the open routine for icmp.  It allocates a icmp_t structure for
1570  * the stream and, on the first open of the module, creates an ND table.
1571  */
1572 /* ARGSUSED */
1573 static conn_t *
1574 icmp_open(int family, cred_t *credp, int *err, int flags)
1575 {
1576 	icmp_t	*icmp;
1577 	conn_t *connp;
1578 	zoneid_t zoneid;
1579 	netstack_t *ns;
1580 	icmp_stack_t *is;
1581 	boolean_t isv6 = B_FALSE;
1582 
1583 	*err = secpolicy_net_icmpaccess(credp);
1584 	if (*err != 0)
1585 		return (NULL);
1586 
1587 	if (family == AF_INET6)
1588 		isv6 = B_TRUE;
1589 	ns = netstack_find_by_cred(credp);
1590 	ASSERT(ns != NULL);
1591 	is = ns->netstack_icmp;
1592 	ASSERT(is != NULL);
1593 
1594 	/*
1595 	 * For exclusive stacks we set the zoneid to zero
1596 	 * to make ICMP operate as if in the global zone.
1597 	 */
1598 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
1599 		zoneid = GLOBAL_ZONEID;
1600 	else
1601 		zoneid = crgetzoneid(credp);
1602 
1603 	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);
1604 
1605 	connp = ipcl_conn_create(IPCL_RAWIPCONN, flags, ns);
1606 	icmp = connp->conn_icmp;
1607 	icmp->icmp_v6dst = sin6_null;
1608 
1609 	/*
1610 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
1611 	 * done by netstack_find_by_cred()
1612 	 */
1613 	netstack_rele(ns);
1614 
1615 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1616 	ASSERT(connp->conn_ulp == IPPROTO_ICMP);
1617 	ASSERT(connp->conn_icmp == icmp);
1618 	ASSERT(icmp->icmp_connp == connp);
1619 
1620 	/* Set the initial state of the stream and the privilege status. */
1621 	icmp->icmp_state = TS_UNBND;
1622 	if (isv6) {
1623 		icmp->icmp_ipversion = IPV6_VERSION;
1624 		icmp->icmp_family = AF_INET6;
1625 		connp->conn_ulp = IPPROTO_ICMPV6;
1626 		/* May be changed by a SO_PROTOTYPE socket option. */
1627 		icmp->icmp_proto = IPPROTO_ICMPV6;
1628 		icmp->icmp_checksum_off = 2;	/* Offset for icmp6_cksum */
1629 		icmp->icmp_max_hdr_len = IPV6_HDR_LEN;
1630 		icmp->icmp_ttl = (uint8_t)is->is_ipv6_hoplimit;
1631 		connp->conn_af_isv6 = B_TRUE;
1632 		connp->conn_flags |= IPCL_ISV6;
1633 	} else {
1634 		icmp->icmp_ipversion = IPV4_VERSION;
1635 		icmp->icmp_family = AF_INET;
1636 		/* May be changed by a SO_PROTOTYPE socket option. */
1637 		icmp->icmp_proto = IPPROTO_ICMP;
1638 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH;
1639 		icmp->icmp_ttl = (uint8_t)is->is_ipv4_ttl;
1640 		connp->conn_af_isv6 = B_FALSE;
1641 		connp->conn_flags &= ~IPCL_ISV6;
1642 	}
1643 	icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1644 	icmp->icmp_pending_op = -1;
1645 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1646 	connp->conn_zoneid = zoneid;
1647 
1648 	/*
1649 	 * If the caller has the process-wide flag set, then default to MAC
1650 	 * exempt mode.  This allows read-down to unlabeled hosts.
1651 	 */
1652 	if (getpflags(NET_MAC_AWARE, credp) != 0)
1653 		connp->conn_mac_exempt = B_TRUE;
1654 
1655 	connp->conn_ulp_labeled = is_system_labeled();
1656 
1657 	icmp->icmp_is = is;
1658 
1659 	connp->conn_recv = icmp_input;
1660 	crhold(credp);
1661 	connp->conn_cred = credp;
1662 
1663 	rw_exit(&icmp->icmp_rwlock);
1664 
1665 	connp->conn_flow_cntrld = B_FALSE;
1666 	return (connp);
1667 }
1668 
1669 /*
1670  * Which ICMP options OK to set through T_UNITDATA_REQ...
1671  */
1672 /* ARGSUSED */
1673 static boolean_t
1674 icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1675 {
1676 	return (B_TRUE);
1677 }
1678 
1679 /*
1680  * This routine gets default values of certain options whose default
1681  * values are maintained by protcol specific code
1682  */
1683 /* ARGSUSED */
1684 int
1685 icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr)
1686 {
1687 	icmp_t *icmp = Q_TO_ICMP(q);
1688 	icmp_stack_t *is = icmp->icmp_is;
1689 	int *i1 = (int *)ptr;
1690 
1691 	switch (level) {
1692 	case IPPROTO_IP:
1693 		switch (name) {
1694 		case IP_MULTICAST_TTL:
1695 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1696 			return (sizeof (uchar_t));
1697 		case IP_MULTICAST_LOOP:
1698 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1699 			return (sizeof (uchar_t));
1700 		}
1701 		break;
1702 	case IPPROTO_IPV6:
1703 		switch (name) {
1704 		case IPV6_MULTICAST_HOPS:
1705 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1706 			return (sizeof (int));
1707 		case IPV6_MULTICAST_LOOP:
1708 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1709 			return (sizeof (int));
1710 		case IPV6_UNICAST_HOPS:
1711 			*i1 = is->is_ipv6_hoplimit;
1712 			return (sizeof (int));
1713 		}
1714 		break;
1715 	case IPPROTO_ICMPV6:
1716 		switch (name) {
1717 		case ICMP6_FILTER:
1718 			/* Make it look like "pass all" */
1719 			ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1720 			return (sizeof (icmp6_filter_t));
1721 		}
1722 		break;
1723 	}
1724 	return (-1);
1725 }
1726 
1727 /*
1728  * This routine retrieves the current status of socket options.
1729  * It returns the size of the option retrieved.
1730  */
1731 int
1732 icmp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
1733 {
1734 	icmp_t		*icmp = connp->conn_icmp;
1735 	icmp_stack_t	*is = icmp->icmp_is;
1736 	int		*i1 = (int *)ptr;
1737 	ip6_pkt_t	*ipp = &icmp->icmp_sticky_ipp;
1738 	int		ret = 0;
1739 
1740 	ASSERT(RW_READ_HELD(&icmp->icmp_rwlock));
1741 	switch (level) {
1742 	case SOL_SOCKET:
1743 		switch (name) {
1744 		case SO_DEBUG:
1745 			*i1 = icmp->icmp_debug;
1746 			break;
1747 		case SO_TYPE:
1748 			*i1 = SOCK_RAW;
1749 			break;
1750 		case SO_PROTOTYPE:
1751 			*i1 = icmp->icmp_proto;
1752 			break;
1753 		case SO_REUSEADDR:
1754 			*i1 = icmp->icmp_reuseaddr;
1755 			break;
1756 
1757 		/*
1758 		 * The following three items are available here,
1759 		 * but are only meaningful to IP.
1760 		 */
1761 		case SO_DONTROUTE:
1762 			*i1 = icmp->icmp_dontroute;
1763 			break;
1764 		case SO_USELOOPBACK:
1765 			*i1 = icmp->icmp_useloopback;
1766 			break;
1767 		case SO_BROADCAST:
1768 			*i1 = icmp->icmp_broadcast;
1769 			break;
1770 
1771 		case SO_SNDBUF:
1772 			ASSERT(icmp->icmp_xmit_hiwat <= INT_MAX);
1773 			*i1 = icmp->icmp_xmit_hiwat;
1774 			break;
1775 		case SO_RCVBUF:
1776 			ASSERT(icmp->icmp_recv_hiwat <= INT_MAX);
1777 			*i1 = icmp->icmp_recv_hiwat;
1778 			break;
1779 		case SO_DGRAM_ERRIND:
1780 			*i1 = icmp->icmp_dgram_errind;
1781 			break;
1782 		case SO_TIMESTAMP:
1783 			*i1 = icmp->icmp_timestamp;
1784 			break;
1785 		case SO_MAC_EXEMPT:
1786 			*i1 = connp->conn_mac_exempt;
1787 			break;
1788 		case SO_DOMAIN:
1789 			*i1 = icmp->icmp_family;
1790 			break;
1791 
1792 		/*
1793 		 * Following four not meaningful for icmp
1794 		 * Action is same as "default" to which we fallthrough
1795 		 * so we keep them in comments.
1796 		 * case SO_LINGER:
1797 		 * case SO_KEEPALIVE:
1798 		 * case SO_OOBINLINE:
1799 		 * case SO_ALLZONES:
1800 		 */
1801 		default:
1802 			ret = -1;
1803 			goto done;
1804 		}
1805 		break;
1806 	case IPPROTO_IP:
1807 		/*
1808 		 * Only allow IPv4 option processing on IPv4 sockets.
1809 		 */
1810 		if (icmp->icmp_family != AF_INET) {
1811 			ret = -1;
1812 			goto done;
1813 		}
1814 
1815 		switch (name) {
1816 		case IP_OPTIONS:
1817 		case T_IP_OPTIONS:
1818 			/* Options are passed up with each packet */
1819 			ret = 0;
1820 			goto done;
1821 		case IP_HDRINCL:
1822 			*i1 = (int)icmp->icmp_hdrincl;
1823 			break;
1824 		case IP_TOS:
1825 		case T_IP_TOS:
1826 			*i1 = (int)icmp->icmp_type_of_service;
1827 			break;
1828 		case IP_TTL:
1829 			*i1 = (int)icmp->icmp_ttl;
1830 			break;
1831 		case IP_MULTICAST_IF:
1832 			/* 0 address if not set */
1833 			*(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr;
1834 			ret = sizeof (ipaddr_t);
1835 			goto done;
1836 		case IP_MULTICAST_TTL:
1837 			*(uchar_t *)ptr = icmp->icmp_multicast_ttl;
1838 			ret = sizeof (uchar_t);
1839 			goto done;
1840 		case IP_MULTICAST_LOOP:
1841 			*ptr = connp->conn_multicast_loop;
1842 			ret = sizeof (uint8_t);
1843 			goto done;
1844 		case IP_BOUND_IF:
1845 			/* Zero if not set */
1846 			*i1 = icmp->icmp_bound_if;
1847 			break;	/* goto sizeof (int) option return */
1848 		case IP_UNSPEC_SRC:
1849 			*ptr = icmp->icmp_unspec_source;
1850 			break;	/* goto sizeof (int) option return */
1851 		case IP_RECVIF:
1852 			*ptr = icmp->icmp_recvif;
1853 			break;	/* goto sizeof (int) option return */
1854 		case IP_BROADCAST_TTL:
1855 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
1856 			return (sizeof (uchar_t));
1857 		case IP_RECVPKTINFO:
1858 			/*
1859 			 * This also handles IP_PKTINFO.
1860 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
1861 			 * Differentiation is based on the size of the argument
1862 			 * passed in.
1863 			 * This option is handled in IP which will return an
1864 			 * error for IP_PKTINFO as it's not supported as a
1865 			 * sticky option.
1866 			 */
1867 			ret = -EINVAL;
1868 			goto done;
1869 		/*
1870 		 * Cannot "get" the value of following options
1871 		 * at this level. Action is same as "default" to
1872 		 * which we fallthrough so we keep them in comments.
1873 		 *
1874 		 * case IP_ADD_MEMBERSHIP:
1875 		 * case IP_DROP_MEMBERSHIP:
1876 		 * case IP_BLOCK_SOURCE:
1877 		 * case IP_UNBLOCK_SOURCE:
1878 		 * case IP_ADD_SOURCE_MEMBERSHIP:
1879 		 * case IP_DROP_SOURCE_MEMBERSHIP:
1880 		 * case MCAST_JOIN_GROUP:
1881 		 * case MCAST_LEAVE_GROUP:
1882 		 * case MCAST_BLOCK_SOURCE:
1883 		 * case MCAST_UNBLOCK_SOURCE:
1884 		 * case MCAST_JOIN_SOURCE_GROUP:
1885 		 * case MCAST_LEAVE_SOURCE_GROUP:
1886 		 * case MRT_INIT:
1887 		 * case MRT_DONE:
1888 		 * case MRT_ADD_VIF:
1889 		 * case MRT_DEL_VIF:
1890 		 * case MRT_ADD_MFC:
1891 		 * case MRT_DEL_MFC:
1892 		 * case MRT_VERSION:
1893 		 * case MRT_ASSERT:
1894 		 * case IP_SEC_OPT:
1895 		 * case IP_NEXTHOP:
1896 		 */
1897 		default:
1898 			ret = -1;
1899 			goto done;
1900 		}
1901 		break;
1902 	case IPPROTO_IPV6:
1903 		/*
1904 		 * Only allow IPv6 option processing on native IPv6 sockets.
1905 		 */
1906 		if (icmp->icmp_family != AF_INET6) {
1907 			ret = -1;
1908 			goto done;
1909 		}
1910 		switch (name) {
1911 		case IPV6_UNICAST_HOPS:
1912 			*i1 = (unsigned int)icmp->icmp_ttl;
1913 			break;
1914 		case IPV6_MULTICAST_IF:
1915 			/* 0 index if not set */
1916 			*i1 = icmp->icmp_multicast_if_index;
1917 			break;
1918 		case IPV6_MULTICAST_HOPS:
1919 			*i1 = icmp->icmp_multicast_ttl;
1920 			break;
1921 		case IPV6_MULTICAST_LOOP:
1922 			*i1 = connp->conn_multicast_loop;
1923 			break;
1924 		case IPV6_BOUND_IF:
1925 			/* Zero if not set */
1926 			*i1 = icmp->icmp_bound_if;
1927 			break;
1928 		case IPV6_UNSPEC_SRC:
1929 			*i1 = icmp->icmp_unspec_source;
1930 			break;
1931 		case IPV6_CHECKSUM:
1932 			/*
1933 			 * Return offset or -1 if no checksum offset.
1934 			 * Does not apply to IPPROTO_ICMPV6
1935 			 */
1936 			if (icmp->icmp_proto == IPPROTO_ICMPV6) {
1937 				ret = -1;
1938 				goto done;
1939 			}
1940 
1941 			if (icmp->icmp_raw_checksum) {
1942 				*i1 = icmp->icmp_checksum_off;
1943 			} else {
1944 				*i1 = -1;
1945 			}
1946 			break;
1947 		case IPV6_JOIN_GROUP:
1948 		case IPV6_LEAVE_GROUP:
1949 		case MCAST_JOIN_GROUP:
1950 		case MCAST_LEAVE_GROUP:
1951 		case MCAST_BLOCK_SOURCE:
1952 		case MCAST_UNBLOCK_SOURCE:
1953 		case MCAST_JOIN_SOURCE_GROUP:
1954 		case MCAST_LEAVE_SOURCE_GROUP:
1955 			/* cannot "get" the value for these */
1956 			ret = -1;
1957 			goto done;
1958 		case IPV6_RECVPKTINFO:
1959 			*i1 = icmp->icmp_ip_recvpktinfo;
1960 			break;
1961 		case IPV6_RECVTCLASS:
1962 			*i1 = icmp->icmp_ipv6_recvtclass;
1963 			break;
1964 		case IPV6_RECVPATHMTU:
1965 			*i1 = icmp->icmp_ipv6_recvpathmtu;
1966 			break;
1967 		case IPV6_V6ONLY:
1968 			*i1 = 1;
1969 			break;
1970 		case IPV6_RECVHOPLIMIT:
1971 			*i1 = icmp->icmp_ipv6_recvhoplimit;
1972 			break;
1973 		case IPV6_RECVHOPOPTS:
1974 			*i1 = icmp->icmp_ipv6_recvhopopts;
1975 			break;
1976 		case IPV6_RECVDSTOPTS:
1977 			*i1 = icmp->icmp_ipv6_recvdstopts;
1978 			break;
1979 		case _OLD_IPV6_RECVDSTOPTS:
1980 			*i1 = icmp->icmp_old_ipv6_recvdstopts;
1981 			break;
1982 		case IPV6_RECVRTHDRDSTOPTS:
1983 			*i1 = icmp->icmp_ipv6_recvrtdstopts;
1984 			break;
1985 		case IPV6_RECVRTHDR:
1986 			*i1 = icmp->icmp_ipv6_recvrthdr;
1987 			break;
1988 		case IPV6_PKTINFO: {
1989 			/* XXX assumes that caller has room for max size! */
1990 			struct in6_pktinfo *pkti;
1991 
1992 			pkti = (struct in6_pktinfo *)ptr;
1993 			if (ipp->ipp_fields & IPPF_IFINDEX)
1994 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
1995 			else
1996 				pkti->ipi6_ifindex = 0;
1997 			if (ipp->ipp_fields & IPPF_ADDR)
1998 				pkti->ipi6_addr = ipp->ipp_addr;
1999 			else
2000 				pkti->ipi6_addr = ipv6_all_zeros;
2001 			ret = sizeof (struct in6_pktinfo);
2002 			goto done;
2003 		}
2004 		case IPV6_NEXTHOP: {
2005 			sin6_t *sin6 = (sin6_t *)ptr;
2006 
2007 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2008 				return (0);
2009 			*sin6 = sin6_null;
2010 			sin6->sin6_family = AF_INET6;
2011 			sin6->sin6_addr = ipp->ipp_nexthop;
2012 			ret = (sizeof (sin6_t));
2013 			goto done;
2014 		}
2015 		case IPV6_HOPOPTS:
2016 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2017 				return (0);
2018 			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
2019 				return (0);
2020 			bcopy((char *)ipp->ipp_hopopts +
2021 			    icmp->icmp_label_len_v6, ptr,
2022 			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
2023 			if (icmp->icmp_label_len_v6 > 0) {
2024 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2025 				ptr[1] = (ipp->ipp_hopoptslen -
2026 				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
2027 			}
2028 			ret = (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
2029 			goto done;
2030 		case IPV6_RTHDRDSTOPTS:
2031 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2032 				return (0);
2033 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2034 			ret = ipp->ipp_rtdstoptslen;
2035 			goto done;
2036 		case IPV6_RTHDR:
2037 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2038 				return (0);
2039 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2040 			ret = ipp->ipp_rthdrlen;
2041 			goto done;
2042 		case IPV6_DSTOPTS:
2043 			if (!(ipp->ipp_fields & IPPF_DSTOPTS)) {
2044 				ret = 0;
2045 				goto done;
2046 			}
2047 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2048 			ret = ipp->ipp_dstoptslen;
2049 			goto done;
2050 		case IPV6_PATHMTU:
2051 			if (!(ipp->ipp_fields & IPPF_PATHMTU)) {
2052 				ret = 0;
2053 			} else {
2054 				ret = ip_fill_mtuinfo(
2055 				    &icmp->icmp_v6dst.sin6_addr, 0,
2056 				    (struct ip6_mtuinfo *)ptr,
2057 				    is->is_netstack);
2058 			}
2059 			goto done;
2060 		case IPV6_TCLASS:
2061 			if (ipp->ipp_fields & IPPF_TCLASS)
2062 				*i1 = ipp->ipp_tclass;
2063 			else
2064 				*i1 = IPV6_FLOW_TCLASS(
2065 				    IPV6_DEFAULT_VERS_AND_FLOW);
2066 			break;
2067 		default:
2068 			ret = -1;
2069 			goto done;
2070 		}
2071 		break;
2072 	case IPPROTO_ICMPV6:
2073 		/*
2074 		 * Only allow IPv6 option processing on native IPv6 sockets.
2075 		 */
2076 		if (icmp->icmp_family != AF_INET6) {
2077 			ret = -1;
2078 		}
2079 
2080 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
2081 			ret = -1;
2082 		}
2083 
2084 		switch (name) {
2085 		case ICMP6_FILTER:
2086 			if (icmp->icmp_filter == NULL) {
2087 				/* Make it look like "pass all" */
2088 				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
2089 			} else {
2090 				(void) bcopy(icmp->icmp_filter, ptr,
2091 				    sizeof (icmp6_filter_t));
2092 			}
2093 			ret = sizeof (icmp6_filter_t);
2094 			goto done;
2095 		default:
2096 			ret = -1;
2097 			goto done;
2098 		}
2099 	default:
2100 		ret = -1;
2101 		goto done;
2102 	}
2103 	ret = sizeof (int);
2104 done:
2105 	return (ret);
2106 }
2107 
2108 /*
2109  * This routine retrieves the current status of socket options.
2110  * It returns the size of the option retrieved.
2111  */
2112 int
2113 icmp_tpi_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
2114 {
2115 	conn_t  *connp = Q_TO_CONN(q);
2116 	icmp_t	*icmp = connp->conn_icmp;
2117 	int 	err;
2118 
2119 	rw_enter(&icmp->icmp_rwlock, RW_READER);
2120 	err = icmp_opt_get(connp, level, name, ptr);
2121 	rw_exit(&icmp->icmp_rwlock);
2122 	return (err);
2123 }
2124 
2125 int
2126 icmp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
2127     uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
2128     void *thisdg_attrs, boolean_t checkonly)
2129 {
2130 
2131 	int	*i1 = (int *)invalp;
2132 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2133 	icmp_t *icmp = connp->conn_icmp;
2134 	icmp_stack_t *is = icmp->icmp_is;
2135 	int	error;
2136 
2137 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
2138 	/*
2139 	 * For fixed length options, no sanity check
2140 	 * of passed in length is done. It is assumed *_optcom_req()
2141 	 * routines do the right thing.
2142 	 */
2143 	switch (level) {
2144 	case SOL_SOCKET:
2145 		switch (name) {
2146 		case SO_DEBUG:
2147 			if (!checkonly)
2148 				icmp->icmp_debug = onoff;
2149 			break;
2150 		case SO_PROTOTYPE:
2151 			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
2152 			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
2153 			    secpolicy_net_rawaccess(cr) != 0) {
2154 				*outlenp = 0;
2155 				return (EACCES);
2156 			}
2157 			/* Can't use IPPROTO_RAW with IPv6 */
2158 			if ((*i1 & 0xFF) == IPPROTO_RAW &&
2159 			    icmp->icmp_family == AF_INET6) {
2160 				*outlenp = 0;
2161 				return (EPROTONOSUPPORT);
2162 			}
2163 			if (checkonly) {
2164 				/* T_CHECK case */
2165 				*(int *)outvalp = (*i1 & 0xFF);
2166 				break;
2167 			}
2168 			icmp->icmp_proto = *i1 & 0xFF;
2169 			if ((icmp->icmp_proto == IPPROTO_RAW ||
2170 			    icmp->icmp_proto == IPPROTO_IGMP) &&
2171 			    icmp->icmp_family == AF_INET)
2172 				icmp->icmp_hdrincl = 1;
2173 			else
2174 				icmp->icmp_hdrincl = 0;
2175 
2176 			if (icmp->icmp_family == AF_INET6 &&
2177 			    icmp->icmp_proto == IPPROTO_ICMPV6) {
2178 				/* Set offset for icmp6_cksum */
2179 				icmp->icmp_raw_checksum = 0;
2180 				icmp->icmp_checksum_off = 2;
2181 			}
2182 			if (icmp->icmp_proto == IPPROTO_UDP ||
2183 			    icmp->icmp_proto == IPPROTO_TCP ||
2184 			    icmp->icmp_proto == IPPROTO_SCTP) {
2185 				icmp->icmp_no_tp_cksum = 1;
2186 				icmp->icmp_sticky_ipp.ipp_fields |=
2187 				    IPPF_NO_CKSUM;
2188 			} else {
2189 				icmp->icmp_no_tp_cksum = 0;
2190 				icmp->icmp_sticky_ipp.ipp_fields &=
2191 				    ~IPPF_NO_CKSUM;
2192 			}
2193 
2194 			if (icmp->icmp_filter != NULL &&
2195 			    icmp->icmp_proto != IPPROTO_ICMPV6) {
2196 				kmem_free(icmp->icmp_filter,
2197 				    sizeof (icmp6_filter_t));
2198 				icmp->icmp_filter = NULL;
2199 			}
2200 
2201 			/* Rebuild the header template */
2202 			error = icmp_build_hdrs(icmp);
2203 			if (error != 0) {
2204 				*outlenp = 0;
2205 				return (error);
2206 			}
2207 
2208 			/*
2209 			 * For SCTP, we don't use icmp_bind_proto() for
2210 			 * raw socket binding.  Note that we do not need
2211 			 * to set *outlenp.
2212 			 * FIXME: how does SCTP work?
2213 			 */
2214 			if (icmp->icmp_proto == IPPROTO_SCTP)
2215 				return (0);
2216 
2217 			*outlenp = sizeof (int);
2218 			*(int *)outvalp = *i1 & 0xFF;
2219 
2220 			/* Drop lock across the bind operation */
2221 			rw_exit(&icmp->icmp_rwlock);
2222 			(void) icmp_bind_proto(connp);
2223 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2224 			return (0);
2225 		case SO_REUSEADDR:
2226 			if (!checkonly) {
2227 				icmp->icmp_reuseaddr = onoff;
2228 				PASS_OPT_TO_IP(connp);
2229 			}
2230 			break;
2231 
2232 		/*
2233 		 * The following three items are available here,
2234 		 * but are only meaningful to IP.
2235 		 */
2236 		case SO_DONTROUTE:
2237 			if (!checkonly) {
2238 				icmp->icmp_dontroute = onoff;
2239 				PASS_OPT_TO_IP(connp);
2240 			}
2241 			break;
2242 		case SO_USELOOPBACK:
2243 			if (!checkonly) {
2244 				icmp->icmp_useloopback = onoff;
2245 				PASS_OPT_TO_IP(connp);
2246 			}
2247 			break;
2248 		case SO_BROADCAST:
2249 			if (!checkonly) {
2250 				icmp->icmp_broadcast = onoff;
2251 				PASS_OPT_TO_IP(connp);
2252 			}
2253 			break;
2254 
2255 		case SO_SNDBUF:
2256 			if (*i1 > is->is_max_buf) {
2257 				*outlenp = 0;
2258 				return (ENOBUFS);
2259 			}
2260 			if (!checkonly) {
2261 				if (!IPCL_IS_NONSTR(connp)) {
2262 					connp->conn_wq->q_hiwat = *i1;
2263 				}
2264 				icmp->icmp_xmit_hiwat = *i1;
2265 			}
2266 			break;
2267 		case SO_RCVBUF:
2268 			if (*i1 > is->is_max_buf) {
2269 				*outlenp = 0;
2270 				return (ENOBUFS);
2271 			}
2272 			if (!checkonly) {
2273 				icmp->icmp_recv_hiwat = *i1;
2274 				rw_exit(&icmp->icmp_rwlock);
2275 				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
2276 				    *i1);
2277 				rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2278 			}
2279 			break;
2280 		case SO_DGRAM_ERRIND:
2281 			if (!checkonly)
2282 				icmp->icmp_dgram_errind = onoff;
2283 			break;
2284 		case SO_ALLZONES:
2285 			/*
2286 			 * "soft" error (negative)
2287 			 * option not handled at this level
2288 			 * Note: Do not modify *outlenp
2289 			 */
2290 			return (-EINVAL);
2291 		case SO_TIMESTAMP:
2292 			if (!checkonly) {
2293 				icmp->icmp_timestamp = onoff;
2294 			}
2295 			break;
2296 		case SO_MAC_EXEMPT:
2297 			/*
2298 			 * "soft" error (negative)
2299 			 * option not handled at this level
2300 			 * Note: Do not modify *outlenp
2301 			 */
2302 			return (-EINVAL);
2303 		/*
2304 		 * Following three not meaningful for icmp
2305 		 * Action is same as "default" so we keep them
2306 		 * in comments.
2307 		 * case SO_LINGER:
2308 		 * case SO_KEEPALIVE:
2309 		 * case SO_OOBINLINE:
2310 		 */
2311 		default:
2312 			*outlenp = 0;
2313 			return (EINVAL);
2314 		}
2315 		break;
2316 	case IPPROTO_IP:
2317 		/*
2318 		 * Only allow IPv4 option processing on IPv4 sockets.
2319 		 */
2320 		if (icmp->icmp_family != AF_INET) {
2321 			*outlenp = 0;
2322 			return (ENOPROTOOPT);
2323 		}
2324 		switch (name) {
2325 		case IP_OPTIONS:
2326 		case T_IP_OPTIONS:
2327 			/* Save options for use by IP. */
2328 			if ((inlen & 0x3) ||
2329 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2330 				*outlenp = 0;
2331 				return (EINVAL);
2332 			}
2333 			if (checkonly)
2334 				break;
2335 
2336 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2337 			    &icmp->icmp_ip_snd_options_len,
2338 			    icmp->icmp_label_len, invalp, inlen)) {
2339 				*outlenp = 0;
2340 				return (ENOMEM);
2341 			}
2342 
2343 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2344 			    icmp->icmp_ip_snd_options_len;
2345 			rw_exit(&icmp->icmp_rwlock);
2346 			(void) proto_set_tx_wroff(connp->conn_rq == NULL ? NULL:
2347 			    RD(connp->conn_rq), connp,
2348 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
2349 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2350 			break;
2351 		case IP_HDRINCL:
2352 			if (!checkonly)
2353 				icmp->icmp_hdrincl = onoff;
2354 			break;
2355 		case IP_TOS:
2356 		case T_IP_TOS:
2357 			if (!checkonly) {
2358 				icmp->icmp_type_of_service = (uint8_t)*i1;
2359 			}
2360 			break;
2361 		case IP_TTL:
2362 			if (!checkonly) {
2363 				icmp->icmp_ttl = (uint8_t)*i1;
2364 			}
2365 			break;
2366 		case IP_MULTICAST_IF:
2367 			/*
2368 			 * TODO should check OPTMGMT reply and undo this if
2369 			 * there is an error.
2370 			 */
2371 			if (!checkonly) {
2372 				icmp->icmp_multicast_if_addr = *i1;
2373 				PASS_OPT_TO_IP(connp);
2374 			}
2375 			break;
2376 		case IP_MULTICAST_TTL:
2377 			if (!checkonly)
2378 				icmp->icmp_multicast_ttl = *invalp;
2379 			break;
2380 		case IP_MULTICAST_LOOP:
2381 			if (!checkonly) {
2382 				connp->conn_multicast_loop =
2383 				    (*invalp == 0) ? 0 : 1;
2384 				PASS_OPT_TO_IP(connp);
2385 			}
2386 			break;
2387 		case IP_BOUND_IF:
2388 			if (!checkonly) {
2389 				icmp->icmp_bound_if = *i1;
2390 				PASS_OPT_TO_IP(connp);
2391 			}
2392 			break;
2393 		case IP_UNSPEC_SRC:
2394 			if (!checkonly) {
2395 				icmp->icmp_unspec_source = onoff;
2396 				PASS_OPT_TO_IP(connp);
2397 			}
2398 			break;
2399 		case IP_BROADCAST_TTL:
2400 			if (!checkonly)
2401 				connp->conn_broadcast_ttl = *invalp;
2402 			break;
2403 		case IP_RECVIF:
2404 			if (!checkonly) {
2405 				icmp->icmp_recvif = onoff;
2406 			}
2407 			/*
2408 			 * pass to ip
2409 			 */
2410 			return (-EINVAL);
2411 		case IP_PKTINFO: {
2412 			/*
2413 			 * This also handles IP_RECVPKTINFO.
2414 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2415 			 * Differentiation is based on the size of the argument
2416 			 * passed in.
2417 			 */
2418 			struct in_pktinfo *pktinfop;
2419 			ip4_pkt_t *attr_pktinfop;
2420 
2421 			if (checkonly)
2422 				break;
2423 
2424 			if (inlen == sizeof (int)) {
2425 				/*
2426 				 * This is IP_RECVPKTINFO option.
2427 				 * Keep a local copy of wether this option is
2428 				 * set or not and pass it down to IP for
2429 				 * processing.
2430 				 */
2431 				icmp->icmp_ip_recvpktinfo = onoff;
2432 				return (-EINVAL);
2433 			}
2434 
2435 
2436 			if (inlen != sizeof (struct in_pktinfo)) {
2437 				return (EINVAL);
2438 			}
2439 
2440 			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
2441 			    == NULL) {
2442 				/*
2443 				 * sticky option is not supported
2444 				 */
2445 				return (EINVAL);
2446 			}
2447 
2448 			pktinfop = (struct in_pktinfo *)invalp;
2449 
2450 			/*
2451 			 * Atleast one of the values should be specified
2452 			 */
2453 			if (pktinfop->ipi_ifindex == 0 &&
2454 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2455 				return (EINVAL);
2456 			}
2457 
2458 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2459 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2460 		}
2461 			break;
2462 		case IP_ADD_MEMBERSHIP:
2463 		case IP_DROP_MEMBERSHIP:
2464 		case IP_BLOCK_SOURCE:
2465 		case IP_UNBLOCK_SOURCE:
2466 		case IP_ADD_SOURCE_MEMBERSHIP:
2467 		case IP_DROP_SOURCE_MEMBERSHIP:
2468 		case MCAST_JOIN_GROUP:
2469 		case MCAST_LEAVE_GROUP:
2470 		case MCAST_BLOCK_SOURCE:
2471 		case MCAST_UNBLOCK_SOURCE:
2472 		case MCAST_JOIN_SOURCE_GROUP:
2473 		case MCAST_LEAVE_SOURCE_GROUP:
2474 		case MRT_INIT:
2475 		case MRT_DONE:
2476 		case MRT_ADD_VIF:
2477 		case MRT_DEL_VIF:
2478 		case MRT_ADD_MFC:
2479 		case MRT_DEL_MFC:
2480 		case MRT_VERSION:
2481 		case MRT_ASSERT:
2482 		case IP_SEC_OPT:
2483 		case IP_NEXTHOP:
2484 			/*
2485 			 * "soft" error (negative)
2486 			 * option not handled at this level
2487 			 * Note: Do not modify *outlenp
2488 			 */
2489 			return (-EINVAL);
2490 		default:
2491 			*outlenp = 0;
2492 			return (EINVAL);
2493 		}
2494 		break;
2495 	case IPPROTO_IPV6: {
2496 		ip6_pkt_t		*ipp;
2497 		boolean_t		sticky;
2498 
2499 		if (icmp->icmp_family != AF_INET6) {
2500 			*outlenp = 0;
2501 			return (ENOPROTOOPT);
2502 		}
2503 		/*
2504 		 * Deal with both sticky options and ancillary data
2505 		 */
2506 		if (thisdg_attrs == NULL) {
2507 			/* sticky options, or none */
2508 			ipp = &icmp->icmp_sticky_ipp;
2509 			sticky = B_TRUE;
2510 		} else {
2511 			/* ancillary data */
2512 			ipp = (ip6_pkt_t *)thisdg_attrs;
2513 			sticky = B_FALSE;
2514 		}
2515 
2516 		switch (name) {
2517 		case IPV6_MULTICAST_IF:
2518 			if (!checkonly) {
2519 				icmp->icmp_multicast_if_index = *i1;
2520 				PASS_OPT_TO_IP(connp);
2521 			}
2522 			break;
2523 		case IPV6_UNICAST_HOPS:
2524 			/* -1 means use default */
2525 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2526 				*outlenp = 0;
2527 				return (EINVAL);
2528 			}
2529 			if (!checkonly) {
2530 				if (*i1 == -1) {
2531 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2532 					    is->is_ipv6_hoplimit;
2533 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2534 					/* Pass modified value to IP. */
2535 					*i1 = ipp->ipp_hoplimit;
2536 				} else {
2537 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2538 					    (uint8_t)*i1;
2539 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2540 				}
2541 				/* Rebuild the header template */
2542 				error = icmp_build_hdrs(icmp);
2543 				if (error != 0) {
2544 					*outlenp = 0;
2545 					return (error);
2546 				}
2547 			}
2548 			break;
2549 		case IPV6_MULTICAST_HOPS:
2550 			/* -1 means use default */
2551 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2552 				*outlenp = 0;
2553 				return (EINVAL);
2554 			}
2555 			if (!checkonly) {
2556 				if (*i1 == -1) {
2557 					icmp->icmp_multicast_ttl =
2558 					    ipp->ipp_multicast_hops =
2559 					    IP_DEFAULT_MULTICAST_TTL;
2560 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2561 					/* Pass modified value to IP. */
2562 					*i1 = icmp->icmp_multicast_ttl;
2563 				} else {
2564 					icmp->icmp_multicast_ttl =
2565 					    ipp->ipp_multicast_hops =
2566 					    (uint8_t)*i1;
2567 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2568 				}
2569 			}
2570 			break;
2571 		case IPV6_MULTICAST_LOOP:
2572 			if (*i1 != 0 && *i1 != 1) {
2573 				*outlenp = 0;
2574 				return (EINVAL);
2575 			}
2576 			if (!checkonly) {
2577 				connp->conn_multicast_loop = *i1;
2578 				PASS_OPT_TO_IP(connp);
2579 			}
2580 			break;
2581 		case IPV6_CHECKSUM:
2582 			/*
2583 			 * Integer offset into the user data of where the
2584 			 * checksum is located.
2585 			 * Offset of -1 disables option.
2586 			 * Does not apply to IPPROTO_ICMPV6.
2587 			 */
2588 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2589 				*outlenp = 0;
2590 				return (EINVAL);
2591 			}
2592 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2593 				/* Negative or not 16 bit aligned offset */
2594 				*outlenp = 0;
2595 				return (EINVAL);
2596 			}
2597 			if (checkonly)
2598 				break;
2599 
2600 			if (*i1 == -1) {
2601 				icmp->icmp_raw_checksum = 0;
2602 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2603 			} else {
2604 				icmp->icmp_raw_checksum = 1;
2605 				icmp->icmp_checksum_off = *i1;
2606 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2607 			}
2608 			/* Rebuild the header template */
2609 			error = icmp_build_hdrs(icmp);
2610 			if (error != 0) {
2611 				*outlenp = 0;
2612 				return (error);
2613 			}
2614 			break;
2615 		case IPV6_JOIN_GROUP:
2616 		case IPV6_LEAVE_GROUP:
2617 		case MCAST_JOIN_GROUP:
2618 		case MCAST_LEAVE_GROUP:
2619 		case MCAST_BLOCK_SOURCE:
2620 		case MCAST_UNBLOCK_SOURCE:
2621 		case MCAST_JOIN_SOURCE_GROUP:
2622 		case MCAST_LEAVE_SOURCE_GROUP:
2623 			/*
2624 			 * "soft" error (negative)
2625 			 * option not handled at this level
2626 			 * Note: Do not modify *outlenp
2627 			 */
2628 			return (-EINVAL);
2629 		case IPV6_BOUND_IF:
2630 			if (!checkonly) {
2631 				icmp->icmp_bound_if = *i1;
2632 				PASS_OPT_TO_IP(connp);
2633 			}
2634 			break;
2635 		case IPV6_UNSPEC_SRC:
2636 			if (!checkonly) {
2637 				icmp->icmp_unspec_source = onoff;
2638 				PASS_OPT_TO_IP(connp);
2639 			}
2640 			break;
2641 		case IPV6_RECVTCLASS:
2642 			if (!checkonly) {
2643 				icmp->icmp_ipv6_recvtclass = onoff;
2644 				PASS_OPT_TO_IP(connp);
2645 			}
2646 			break;
2647 		/*
2648 		 * Set boolean switches for ancillary data delivery
2649 		 */
2650 		case IPV6_RECVPKTINFO:
2651 			if (!checkonly) {
2652 				icmp->icmp_ip_recvpktinfo = onoff;
2653 				PASS_OPT_TO_IP(connp);
2654 			}
2655 			break;
2656 		case IPV6_RECVPATHMTU:
2657 			if (!checkonly) {
2658 				icmp->icmp_ipv6_recvpathmtu = onoff;
2659 				PASS_OPT_TO_IP(connp);
2660 			}
2661 			break;
2662 		case IPV6_RECVHOPLIMIT:
2663 			if (!checkonly) {
2664 				icmp->icmp_ipv6_recvhoplimit = onoff;
2665 				PASS_OPT_TO_IP(connp);
2666 			}
2667 			break;
2668 		case IPV6_RECVHOPOPTS:
2669 			if (!checkonly) {
2670 				icmp->icmp_ipv6_recvhopopts = onoff;
2671 				PASS_OPT_TO_IP(connp);
2672 			}
2673 			break;
2674 		case IPV6_RECVDSTOPTS:
2675 			if (!checkonly) {
2676 				icmp->icmp_ipv6_recvdstopts = onoff;
2677 				PASS_OPT_TO_IP(connp);
2678 			}
2679 			break;
2680 		case _OLD_IPV6_RECVDSTOPTS:
2681 			if (!checkonly)
2682 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2683 			break;
2684 		case IPV6_RECVRTHDRDSTOPTS:
2685 			if (!checkonly) {
2686 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2687 				PASS_OPT_TO_IP(connp);
2688 			}
2689 			break;
2690 		case IPV6_RECVRTHDR:
2691 			if (!checkonly) {
2692 				icmp->icmp_ipv6_recvrthdr = onoff;
2693 				PASS_OPT_TO_IP(connp);
2694 			}
2695 			break;
2696 		/*
2697 		 * Set sticky options or ancillary data.
2698 		 * If sticky options, (re)build any extension headers
2699 		 * that might be needed as a result.
2700 		 */
2701 		case IPV6_PKTINFO:
2702 			/*
2703 			 * The source address and ifindex are verified
2704 			 * in ip_opt_set(). For ancillary data the
2705 			 * source address is checked in ip_wput_v6.
2706 			 */
2707 			if (inlen != 0 && inlen !=
2708 			    sizeof (struct in6_pktinfo)) {
2709 				return (EINVAL);
2710 			}
2711 			if (checkonly)
2712 				break;
2713 
2714 			if (inlen == 0) {
2715 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2716 				ipp->ipp_sticky_ignored |=
2717 				    (IPPF_IFINDEX|IPPF_ADDR);
2718 			} else {
2719 				struct in6_pktinfo *pkti;
2720 
2721 				pkti = (struct in6_pktinfo *)invalp;
2722 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2723 				ipp->ipp_addr = pkti->ipi6_addr;
2724 				if (ipp->ipp_ifindex != 0)
2725 					ipp->ipp_fields |= IPPF_IFINDEX;
2726 				else
2727 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2728 				if (!IN6_IS_ADDR_UNSPECIFIED(
2729 				    &ipp->ipp_addr))
2730 					ipp->ipp_fields |= IPPF_ADDR;
2731 				else
2732 					ipp->ipp_fields &= ~IPPF_ADDR;
2733 			}
2734 			if (sticky) {
2735 				error = icmp_build_hdrs(icmp);
2736 				if (error != 0)
2737 					return (error);
2738 				PASS_OPT_TO_IP(connp);
2739 			}
2740 			break;
2741 		case IPV6_HOPLIMIT:
2742 			/* This option can only be used as ancillary data. */
2743 			if (sticky)
2744 				return (EINVAL);
2745 			if (inlen != 0 && inlen != sizeof (int))
2746 				return (EINVAL);
2747 			if (checkonly)
2748 				break;
2749 
2750 			if (inlen == 0) {
2751 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2752 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2753 			} else {
2754 				if (*i1 > 255 || *i1 < -1)
2755 					return (EINVAL);
2756 				if (*i1 == -1)
2757 					ipp->ipp_hoplimit =
2758 					    is->is_ipv6_hoplimit;
2759 				else
2760 					ipp->ipp_hoplimit = *i1;
2761 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2762 			}
2763 			break;
2764 		case IPV6_TCLASS:
2765 			/*
2766 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2767 			 * and [0, 255] as the actualy traffic class.
2768 			 */
2769 			if (inlen != 0 && inlen != sizeof (int)) {
2770 				return (EINVAL);
2771 			}
2772 			if (checkonly)
2773 				break;
2774 
2775 			if (inlen == 0) {
2776 				ipp->ipp_fields &= ~IPPF_TCLASS;
2777 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2778 			} else {
2779 				if (*i1 >= 256 || *i1 < -1)
2780 					return (EINVAL);
2781 				if (*i1 == -1) {
2782 					ipp->ipp_tclass =
2783 					    IPV6_FLOW_TCLASS(
2784 					    IPV6_DEFAULT_VERS_AND_FLOW);
2785 				} else {
2786 					ipp->ipp_tclass = *i1;
2787 				}
2788 				ipp->ipp_fields |= IPPF_TCLASS;
2789 			}
2790 			if (sticky) {
2791 				error = icmp_build_hdrs(icmp);
2792 				if (error != 0)
2793 					return (error);
2794 			}
2795 			break;
2796 		case IPV6_NEXTHOP:
2797 			/*
2798 			 * IP will verify that the nexthop is reachable
2799 			 * and fail for sticky options.
2800 			 */
2801 			if (inlen != 0 && inlen != sizeof (sin6_t)) {
2802 				return (EINVAL);
2803 			}
2804 			if (checkonly)
2805 				break;
2806 
2807 			if (inlen == 0) {
2808 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2809 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2810 			} else {
2811 				sin6_t *sin6 = (sin6_t *)invalp;
2812 
2813 				if (sin6->sin6_family != AF_INET6) {
2814 					return (EAFNOSUPPORT);
2815 				}
2816 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
2817 					return (EADDRNOTAVAIL);
2818 				}
2819 				ipp->ipp_nexthop = sin6->sin6_addr;
2820 				if (!IN6_IS_ADDR_UNSPECIFIED(
2821 				    &ipp->ipp_nexthop))
2822 					ipp->ipp_fields |= IPPF_NEXTHOP;
2823 				else
2824 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2825 			}
2826 			if (sticky) {
2827 				error = icmp_build_hdrs(icmp);
2828 				if (error != 0)
2829 					return (error);
2830 				PASS_OPT_TO_IP(connp);
2831 			}
2832 			break;
2833 		case IPV6_HOPOPTS: {
2834 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2835 			/*
2836 			 * Sanity checks - minimum size, size a multiple of
2837 			 * eight bytes, and matching size passed in.
2838 			 */
2839 			if (inlen != 0 &&
2840 			    inlen != (8 * (hopts->ip6h_len + 1))) {
2841 				return (EINVAL);
2842 			}
2843 
2844 			if (checkonly)
2845 				break;
2846 			error = optcom_pkt_set(invalp, inlen, sticky,
2847 			    (uchar_t **)&ipp->ipp_hopopts,
2848 			    &ipp->ipp_hopoptslen,
2849 			    sticky ? icmp->icmp_label_len_v6 : 0);
2850 			if (error != 0)
2851 				return (error);
2852 			if (ipp->ipp_hopoptslen == 0) {
2853 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2854 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2855 			} else {
2856 				ipp->ipp_fields |= IPPF_HOPOPTS;
2857 			}
2858 			if (sticky) {
2859 				error = icmp_build_hdrs(icmp);
2860 				if (error != 0)
2861 					return (error);
2862 			}
2863 			break;
2864 		}
2865 		case IPV6_RTHDRDSTOPTS: {
2866 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2867 
2868 			/*
2869 			 * Sanity checks - minimum size, size a multiple of
2870 			 * eight bytes, and matching size passed in.
2871 			 */
2872 			if (inlen != 0 &&
2873 			    inlen != (8 * (dopts->ip6d_len + 1)))
2874 				return (EINVAL);
2875 
2876 			if (checkonly)
2877 				break;
2878 
2879 			if (inlen == 0) {
2880 				if (sticky &&
2881 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2882 					kmem_free(ipp->ipp_rtdstopts,
2883 					    ipp->ipp_rtdstoptslen);
2884 					ipp->ipp_rtdstopts = NULL;
2885 					ipp->ipp_rtdstoptslen = 0;
2886 				}
2887 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2888 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2889 			} else {
2890 				error = optcom_pkt_set(invalp, inlen, sticky,
2891 				    (uchar_t **)&ipp->ipp_rtdstopts,
2892 				    &ipp->ipp_rtdstoptslen, 0);
2893 				if (error != 0)
2894 					return (error);
2895 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2896 			}
2897 			if (sticky) {
2898 				error = icmp_build_hdrs(icmp);
2899 				if (error != 0)
2900 					return (error);
2901 			}
2902 			break;
2903 		}
2904 		case IPV6_DSTOPTS: {
2905 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2906 
2907 			/*
2908 			 * Sanity checks - minimum size, size a multiple of
2909 			 * eight bytes, and matching size passed in.
2910 			 */
2911 			if (inlen != 0 &&
2912 			    inlen != (8 * (dopts->ip6d_len + 1)))
2913 				return (EINVAL);
2914 
2915 			if (checkonly)
2916 				break;
2917 
2918 			if (inlen == 0) {
2919 				if (sticky &&
2920 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2921 					kmem_free(ipp->ipp_dstopts,
2922 					    ipp->ipp_dstoptslen);
2923 					ipp->ipp_dstopts = NULL;
2924 					ipp->ipp_dstoptslen = 0;
2925 				}
2926 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2927 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2928 			} else {
2929 				error = optcom_pkt_set(invalp, inlen, sticky,
2930 				    (uchar_t **)&ipp->ipp_dstopts,
2931 				    &ipp->ipp_dstoptslen, 0);
2932 				if (error != 0)
2933 					return (error);
2934 				ipp->ipp_fields |= IPPF_DSTOPTS;
2935 			}
2936 			if (sticky) {
2937 				error = icmp_build_hdrs(icmp);
2938 				if (error != 0)
2939 					return (error);
2940 			}
2941 			break;
2942 		}
2943 		case IPV6_RTHDR: {
2944 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2945 
2946 			/*
2947 			 * Sanity checks - minimum size, size a multiple of
2948 			 * eight bytes, and matching size passed in.
2949 			 */
2950 			if (inlen != 0 &&
2951 			    inlen != (8 * (rt->ip6r_len + 1)))
2952 				return (EINVAL);
2953 
2954 			if (checkonly)
2955 				break;
2956 
2957 			if (inlen == 0) {
2958 				if (sticky &&
2959 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2960 					kmem_free(ipp->ipp_rthdr,
2961 					    ipp->ipp_rthdrlen);
2962 					ipp->ipp_rthdr = NULL;
2963 					ipp->ipp_rthdrlen = 0;
2964 				}
2965 				ipp->ipp_fields &= ~IPPF_RTHDR;
2966 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2967 			} else {
2968 				error = optcom_pkt_set(invalp, inlen, sticky,
2969 				    (uchar_t **)&ipp->ipp_rthdr,
2970 				    &ipp->ipp_rthdrlen, 0);
2971 				if (error != 0)
2972 					return (error);
2973 				ipp->ipp_fields |= IPPF_RTHDR;
2974 			}
2975 			if (sticky) {
2976 				error = icmp_build_hdrs(icmp);
2977 				if (error != 0)
2978 					return (error);
2979 			}
2980 			break;
2981 		}
2982 
2983 		case IPV6_DONTFRAG:
2984 			if (checkonly)
2985 				break;
2986 
2987 			if (onoff) {
2988 				ipp->ipp_fields |= IPPF_DONTFRAG;
2989 			} else {
2990 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2991 			}
2992 			break;
2993 
2994 		case IPV6_USE_MIN_MTU:
2995 			if (inlen != sizeof (int))
2996 				return (EINVAL);
2997 
2998 			if (*i1 < -1 || *i1 > 1)
2999 				return (EINVAL);
3000 
3001 			if (checkonly)
3002 				break;
3003 
3004 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3005 			ipp->ipp_use_min_mtu = *i1;
3006 			break;
3007 
3008 		/*
3009 		 * This option can't be set.  Its only returned via
3010 		 * getsockopt() or ancillary data.
3011 		 */
3012 		case IPV6_PATHMTU:
3013 			return (EINVAL);
3014 
3015 		case IPV6_SEC_OPT:
3016 		case IPV6_SRC_PREFERENCES:
3017 		case IPV6_V6ONLY:
3018 			/* Handled at IP level */
3019 			return (-EINVAL);
3020 		default:
3021 			*outlenp = 0;
3022 			return (EINVAL);
3023 		}
3024 		break;
3025 	}		/* end IPPROTO_IPV6 */
3026 
3027 	case IPPROTO_ICMPV6:
3028 		/*
3029 		 * Only allow IPv6 option processing on IPv6 sockets.
3030 		 */
3031 		if (icmp->icmp_family != AF_INET6) {
3032 			*outlenp = 0;
3033 			return (ENOPROTOOPT);
3034 		}
3035 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
3036 			*outlenp = 0;
3037 			return (ENOPROTOOPT);
3038 		}
3039 		switch (name) {
3040 		case ICMP6_FILTER:
3041 			if (!checkonly) {
3042 				if ((inlen != 0) &&
3043 				    (inlen != sizeof (icmp6_filter_t)))
3044 					return (EINVAL);
3045 
3046 				if (inlen == 0) {
3047 					if (icmp->icmp_filter != NULL) {
3048 						kmem_free(icmp->icmp_filter,
3049 						    sizeof (icmp6_filter_t));
3050 						icmp->icmp_filter = NULL;
3051 					}
3052 				} else {
3053 					if (icmp->icmp_filter == NULL) {
3054 						icmp->icmp_filter = kmem_alloc(
3055 						    sizeof (icmp6_filter_t),
3056 						    KM_NOSLEEP);
3057 						if (icmp->icmp_filter == NULL) {
3058 							*outlenp = 0;
3059 							return (ENOBUFS);
3060 						}
3061 					}
3062 					(void) bcopy(invalp, icmp->icmp_filter,
3063 					    inlen);
3064 				}
3065 			}
3066 			break;
3067 
3068 		default:
3069 			*outlenp = 0;
3070 			return (EINVAL);
3071 		}
3072 		break;
3073 	default:
3074 		*outlenp = 0;
3075 		return (EINVAL);
3076 	}
3077 	/*
3078 	 * Common case of OK return with outval same as inval.
3079 	 */
3080 	if (invalp != outvalp) {
3081 		/* don't trust bcopy for identical src/dst */
3082 		(void) bcopy(invalp, outvalp, inlen);
3083 	}
3084 	*outlenp = inlen;
3085 	return (0);
3086 }
3087 
3088 /* This routine sets socket options. */
3089 /* ARGSUSED */
3090 int
3091 icmp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3092     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3093     void *thisdg_attrs, cred_t *cr)
3094 {
3095 	boolean_t checkonly;
3096 	int	error;
3097 
3098 	error = 0;
3099 	switch (optset_context) {
3100 	case SETFN_OPTCOM_CHECKONLY:
3101 		checkonly = B_TRUE;
3102 		/*
3103 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3104 		 * inlen != 0 implies value supplied and
3105 		 * 	we have to "pretend" to set it.
3106 		 * inlen == 0 implies that there is no
3107 		 * 	value part in T_CHECK request and just validation
3108 		 * done elsewhere should be enough, we just return here.
3109 		 */
3110 		if (inlen == 0) {
3111 			*outlenp = 0;
3112 			error = 0;
3113 			goto done;
3114 		}
3115 		break;
3116 	case SETFN_OPTCOM_NEGOTIATE:
3117 		checkonly = B_FALSE;
3118 		break;
3119 	case SETFN_UD_NEGOTIATE:
3120 	case SETFN_CONN_NEGOTIATE:
3121 		checkonly = B_FALSE;
3122 		/*
3123 		 * Negotiating local and "association-related" options
3124 		 * through T_UNITDATA_REQ.
3125 		 *
3126 		 * Following routine can filter out ones we do not
3127 		 * want to be "set" this way.
3128 		 */
3129 		if (!icmp_opt_allow_udr_set(level, name)) {
3130 			*outlenp = 0;
3131 			error = EINVAL;
3132 			goto done;
3133 		}
3134 		break;
3135 	default:
3136 		/*
3137 		 * We should never get here
3138 		 */
3139 		*outlenp = 0;
3140 		error = EINVAL;
3141 		goto done;
3142 	}
3143 
3144 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3145 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3146 	error = icmp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3147 	    outvalp, cr, thisdg_attrs, checkonly);
3148 
3149 done:
3150 	return (error);
3151 }
3152 
3153 /* This routine sets socket options. */
3154 /* ARGSUSED */
3155 int
3156 icmp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3157     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3158     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3159 {
3160 	conn_t	*connp =  Q_TO_CONN(q);
3161 	icmp_t	*icmp;
3162 	int error;
3163 
3164 	icmp = connp->conn_icmp;
3165 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3166 	error = icmp_opt_set(connp, optset_context, level, name, inlen, invalp,
3167 	    outlenp, outvalp, thisdg_attrs, cr);
3168 	rw_exit(&icmp->icmp_rwlock);
3169 	return (error);
3170 }
3171 
3172 /*
3173  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
3174  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
3175  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
3176  * headers.
3177  * Returns failure if can't allocate memory.
3178  */
3179 static int
3180 icmp_build_hdrs(icmp_t *icmp)
3181 {
3182 	icmp_stack_t *is = icmp->icmp_is;
3183 	uchar_t	*hdrs;
3184 	uint_t	hdrs_len;
3185 	ip6_t	*ip6h;
3186 	ip6i_t	*ip6i;
3187 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
3188 
3189 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
3190 	hdrs_len = ip_total_hdrs_len_v6(ipp);
3191 	ASSERT(hdrs_len != 0);
3192 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
3193 		/* Need to reallocate */
3194 		if (hdrs_len != 0) {
3195 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3196 			if (hdrs == NULL)
3197 				return (ENOMEM);
3198 		} else {
3199 			hdrs = NULL;
3200 		}
3201 		if (icmp->icmp_sticky_hdrs_len != 0) {
3202 			kmem_free(icmp->icmp_sticky_hdrs,
3203 			    icmp->icmp_sticky_hdrs_len);
3204 		}
3205 		icmp->icmp_sticky_hdrs = hdrs;
3206 		icmp->icmp_sticky_hdrs_len = hdrs_len;
3207 	}
3208 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
3209 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
3210 
3211 	/* Set header fields not in ipp */
3212 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3213 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
3214 		ip6h = (ip6_t *)&ip6i[1];
3215 
3216 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
3217 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
3218 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
3219 		}
3220 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
3221 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
3222 		}
3223 	} else {
3224 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
3225 	}
3226 
3227 	if (!(ipp->ipp_fields & IPPF_ADDR))
3228 		ip6h->ip6_src = icmp->icmp_v6src;
3229 
3230 	/* Try to get everything in a single mblk */
3231 	if (hdrs_len > icmp->icmp_max_hdr_len) {
3232 		icmp->icmp_max_hdr_len = hdrs_len;
3233 		rw_exit(&icmp->icmp_rwlock);
3234 		(void) proto_set_tx_wroff(icmp->icmp_connp->conn_rq,
3235 		    icmp->icmp_connp,
3236 		    icmp->icmp_max_hdr_len + is->is_wroff_extra);
3237 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3238 	}
3239 	return (0);
3240 }
3241 
3242 /*
3243  * This routine retrieves the value of an ND variable in a icmpparam_t
3244  * structure.  It is called through nd_getset when a user reads the
3245  * variable.
3246  */
3247 /* ARGSUSED */
3248 static int
3249 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3250 {
3251 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3252 
3253 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
3254 	return (0);
3255 }
3256 
3257 /*
3258  * Walk through the param array specified registering each element with the
3259  * named dispatch (ND) handler.
3260  */
3261 static boolean_t
3262 icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt)
3263 {
3264 	for (; cnt-- > 0; icmppa++) {
3265 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
3266 			if (!nd_load(ndp, icmppa->icmp_param_name,
3267 			    icmp_param_get, icmp_param_set,
3268 			    (caddr_t)icmppa)) {
3269 				nd_free(ndp);
3270 				return (B_FALSE);
3271 			}
3272 		}
3273 	}
3274 	if (!nd_load(ndp, "icmp_status", icmp_status_report, NULL,
3275 	    NULL)) {
3276 		nd_free(ndp);
3277 		return (B_FALSE);
3278 	}
3279 	return (B_TRUE);
3280 }
3281 
3282 /* This routine sets an ND variable in a icmpparam_t structure. */
3283 /* ARGSUSED */
3284 static int
3285 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3286 {
3287 	long		new_value;
3288 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3289 
3290 	/*
3291 	 * Fail the request if the new value does not lie within the
3292 	 * required bounds.
3293 	 */
3294 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3295 	    new_value < icmppa->icmp_param_min ||
3296 	    new_value > icmppa->icmp_param_max) {
3297 		return (EINVAL);
3298 	}
3299 	/* Set the new value */
3300 	icmppa->icmp_param_value = new_value;
3301 	return (0);
3302 }
3303 static void
3304 icmp_queue_fallback(icmp_t *icmp, mblk_t *mp)
3305 {
3306 	ASSERT(MUTEX_HELD(&icmp->icmp_recv_lock));
3307 	if (IPCL_IS_NONSTR(icmp->icmp_connp)) {
3308 		/*
3309 		 * fallback has started but messages have not been moved yet
3310 		 */
3311 		if (icmp->icmp_fallback_queue_head == NULL) {
3312 			ASSERT(icmp->icmp_fallback_queue_tail == NULL);
3313 			icmp->icmp_fallback_queue_head = mp;
3314 			icmp->icmp_fallback_queue_tail = mp;
3315 		} else {
3316 			ASSERT(icmp->icmp_fallback_queue_tail != NULL);
3317 			icmp->icmp_fallback_queue_tail->b_next = mp;
3318 			icmp->icmp_fallback_queue_tail = mp;
3319 		}
3320 		mutex_exit(&icmp->icmp_recv_lock);
3321 	} else {
3322 		/*
3323 		 * no more fallbacks possible, ok to drop lock.
3324 		 */
3325 		mutex_exit(&icmp->icmp_recv_lock);
3326 		putnext(icmp->icmp_connp->conn_rq, mp);
3327 	}
3328 }
3329 
3330 /*ARGSUSED2*/
3331 static void
3332 icmp_input(void *arg1, mblk_t *mp, void *arg2)
3333 {
3334 	conn_t *connp = (conn_t *)arg1;
3335 	struct T_unitdata_ind	*tudi;
3336 	uchar_t			*rptr;
3337 	icmp_t			*icmp;
3338 	icmp_stack_t		*is;
3339 	sin_t			*sin;
3340 	sin6_t			*sin6;
3341 	ip6_t			*ip6h;
3342 	ip6i_t			*ip6i;
3343 	mblk_t			*mp1;
3344 	int			hdr_len;
3345 	ipha_t			*ipha;
3346 	int			udi_size;	/* Size of T_unitdata_ind */
3347 	uint_t			ipvers;
3348 	ip6_pkt_t		ipp;
3349 	uint8_t			nexthdr;
3350 	ip_pktinfo_t		*pinfo = NULL;
3351 	mblk_t			*options_mp = NULL;
3352 	uint_t			icmp_opt = 0;
3353 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
3354 	uint_t			hopstrip;
3355 	int			error;
3356 
3357 	ASSERT(connp->conn_flags & IPCL_RAWIPCONN);
3358 
3359 	icmp = connp->conn_icmp;
3360 	is = icmp->icmp_is;
3361 	rptr = mp->b_rptr;
3362 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3363 	ASSERT(OK_32PTR(rptr));
3364 
3365 	/*
3366 	 * IP should have prepended the options data in an M_CTL
3367 	 * Check M_CTL "type" to make sure are not here bcos of
3368 	 * a valid ICMP message
3369 	 */
3370 	if (DB_TYPE(mp) == M_CTL) {
3371 		/*
3372 		 * FIXME: does IP still do this?
3373 		 * IP sends up the IPSEC_IN message for handling IPSEC
3374 		 * policy at the TCP level. We don't need it here.
3375 		 */
3376 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
3377 			mp1 = mp->b_cont;
3378 			freeb(mp);
3379 			mp = mp1;
3380 			rptr = mp->b_rptr;
3381 		} else if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3382 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3383 		    IN_PKTINFO) {
3384 			/*
3385 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3386 			 * has been prepended to the packet by IP. We need to
3387 			 * extract the mblk and adjust the rptr
3388 			 */
3389 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3390 			options_mp = mp;
3391 			mp = mp->b_cont;
3392 			rptr = mp->b_rptr;
3393 		} else {
3394 			/*
3395 			 * ICMP messages.
3396 			 */
3397 			icmp_icmp_error(connp, mp);
3398 			return;
3399 		}
3400 	}
3401 
3402 	/*
3403 	 * Discard message if it is misaligned or smaller than the IP header.
3404 	 */
3405 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3406 		freemsg(mp);
3407 		if (options_mp != NULL)
3408 			freeb(options_mp);
3409 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3410 		return;
3411 	}
3412 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3413 
3414 	/* Handle M_DATA messages containing IP packets messages */
3415 	if (ipvers == IPV4_VERSION) {
3416 		/*
3417 		 * Special case where IP attaches
3418 		 * the IRE needs to be handled so that we don't send up
3419 		 * IRE to the user land.
3420 		 */
3421 		ipha = (ipha_t *)rptr;
3422 		hdr_len = IPH_HDR_LENGTH(ipha);
3423 
3424 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3425 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3426 
3427 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3428 			    TH_SYN) && mp->b_cont != NULL) {
3429 				mp1 = mp->b_cont;
3430 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3431 					freeb(mp1);
3432 					mp->b_cont = NULL;
3433 				}
3434 			}
3435 		}
3436 		if (is->is_bsd_compat) {
3437 			ushort_t len;
3438 			len = ntohs(ipha->ipha_length);
3439 
3440 			if (mp->b_datap->db_ref > 1) {
3441 				/*
3442 				 * Allocate a new IP header so that we can
3443 				 * modify ipha_length.
3444 				 */
3445 				mblk_t	*mp1;
3446 
3447 				mp1 = allocb(hdr_len, BPRI_MED);
3448 				if (!mp1) {
3449 					freemsg(mp);
3450 					if (options_mp != NULL)
3451 						freeb(options_mp);
3452 					BUMP_MIB(&is->is_rawip_mib,
3453 					    rawipInErrors);
3454 					return;
3455 				}
3456 				bcopy(rptr, mp1->b_rptr, hdr_len);
3457 				mp->b_rptr = rptr + hdr_len;
3458 				rptr = mp1->b_rptr;
3459 				ipha = (ipha_t *)rptr;
3460 				mp1->b_cont = mp;
3461 				mp1->b_wptr = rptr + hdr_len;
3462 				mp = mp1;
3463 			}
3464 			len -= hdr_len;
3465 			ipha->ipha_length = htons(len);
3466 		}
3467 	}
3468 
3469 	/*
3470 	 * This is the inbound data path.  Packets are passed upstream as
3471 	 * T_UNITDATA_IND messages with full IP headers still attached.
3472 	 */
3473 	if (icmp->icmp_family == AF_INET) {
3474 		ASSERT(ipvers == IPV4_VERSION);
3475 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3476 		if (icmp->icmp_recvif && (pinfo != NULL) &&
3477 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3478 			udi_size += sizeof (struct T_opthdr) +
3479 			    sizeof (uint_t);
3480 		}
3481 
3482 		if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3483 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3484 			udi_size += sizeof (struct T_opthdr) +
3485 			    sizeof (struct in_pktinfo);
3486 		}
3487 
3488 		/*
3489 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3490 		 * buffer. Since gethrestime() expects a pointer aligned
3491 		 * argument, we allocate space necessary for extra
3492 		 * alignment (even though it might not be used).
3493 		 */
3494 		if (icmp->icmp_timestamp) {
3495 			udi_size += sizeof (struct T_opthdr) +
3496 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3497 		}
3498 		mp1 = allocb(udi_size, BPRI_MED);
3499 		if (mp1 == NULL) {
3500 			freemsg(mp);
3501 			if (options_mp != NULL)
3502 				freeb(options_mp);
3503 			BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3504 			return;
3505 		}
3506 		mp1->b_cont = mp;
3507 		mp = mp1;
3508 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3509 		mp->b_datap->db_type = M_PROTO;
3510 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3511 		tudi->PRIM_type = T_UNITDATA_IND;
3512 		tudi->SRC_length = sizeof (sin_t);
3513 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3514 		sin = (sin_t *)&tudi[1];
3515 		*sin = sin_null;
3516 		sin->sin_family = AF_INET;
3517 		sin->sin_addr.s_addr = ipha->ipha_src;
3518 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3519 		    sizeof (sin_t);
3520 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3521 		tudi->OPT_length = udi_size;
3522 
3523 		/*
3524 		 * Add options if IP_RECVIF is set
3525 		 */
3526 		if (udi_size != 0) {
3527 			char *dstopt;
3528 
3529 			dstopt = (char *)&sin[1];
3530 			if (icmp->icmp_recvif && (pinfo != NULL) &&
3531 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3532 
3533 				struct T_opthdr *toh;
3534 				uint_t		*dstptr;
3535 
3536 				toh = (struct T_opthdr *)dstopt;
3537 				toh->level = IPPROTO_IP;
3538 				toh->name = IP_RECVIF;
3539 				toh->len = sizeof (struct T_opthdr) +
3540 				    sizeof (uint_t);
3541 				toh->status = 0;
3542 				dstopt += sizeof (struct T_opthdr);
3543 				dstptr = (uint_t *)dstopt;
3544 				*dstptr = pinfo->ip_pkt_ifindex;
3545 				dstopt += sizeof (uint_t);
3546 				udi_size -= toh->len;
3547 			}
3548 			if (icmp->icmp_timestamp) {
3549 				struct	T_opthdr *toh;
3550 
3551 				toh = (struct T_opthdr *)dstopt;
3552 				toh->level = SOL_SOCKET;
3553 				toh->name = SCM_TIMESTAMP;
3554 				toh->len = sizeof (struct T_opthdr) +
3555 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3556 				toh->status = 0;
3557 				dstopt += sizeof (struct T_opthdr);
3558 				/* Align for gethrestime() */
3559 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3560 				    sizeof (intptr_t));
3561 				gethrestime((timestruc_t *)dstopt);
3562 				dstopt = (char *)toh + toh->len;
3563 				udi_size -= toh->len;
3564 			}
3565 			if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3566 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3567 				struct	T_opthdr *toh;
3568 				struct	in_pktinfo *pktinfop;
3569 
3570 				toh = (struct T_opthdr *)dstopt;
3571 				toh->level = IPPROTO_IP;
3572 				toh->name = IP_PKTINFO;
3573 				toh->len = sizeof (struct T_opthdr) +
3574 				    sizeof (in_pktinfo_t);
3575 				toh->status = 0;
3576 				dstopt += sizeof (struct T_opthdr);
3577 				pktinfop = (struct in_pktinfo *)dstopt;
3578 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3579 				pktinfop->ipi_spec_dst =
3580 				    pinfo->ip_pkt_match_addr;
3581 
3582 				pktinfop->ipi_addr.s_addr = ipha->ipha_dst;
3583 
3584 				dstopt += sizeof (struct in_pktinfo);
3585 				udi_size -= toh->len;
3586 			}
3587 
3588 			/* Consumed all of allocated space */
3589 			ASSERT(udi_size == 0);
3590 		}
3591 
3592 		if (options_mp != NULL)
3593 			freeb(options_mp);
3594 
3595 		BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3596 		goto deliver;
3597 	}
3598 
3599 	/*
3600 	 * We don't need options_mp in the IPv6 path.
3601 	 */
3602 	if (options_mp != NULL) {
3603 		freeb(options_mp);
3604 		options_mp = NULL;
3605 	}
3606 
3607 	/*
3608 	 * Discard message if it is smaller than the IPv6 header
3609 	 * or if the header is malformed.
3610 	 */
3611 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3612 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3613 	    icmp->icmp_family != AF_INET6) {
3614 		freemsg(mp);
3615 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3616 		return;
3617 	}
3618 
3619 	/* Initialize */
3620 	ipp.ipp_fields = 0;
3621 	hopstrip = 0;
3622 
3623 	ip6h = (ip6_t *)rptr;
3624 	/*
3625 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3626 	 * as well as individual lenghts of ext hdrs (and ptrs to
3627 	 * them).
3628 	 */
3629 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3630 		/* Look for ifindex information */
3631 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3632 			ip6i = (ip6i_t *)ip6h;
3633 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3634 				ASSERT(ip6i->ip6i_ifindex != 0);
3635 				ipp.ipp_fields |= IPPF_IFINDEX;
3636 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3637 			}
3638 			rptr = (uchar_t *)&ip6i[1];
3639 			mp->b_rptr = rptr;
3640 			if (rptr == mp->b_wptr) {
3641 				mp1 = mp->b_cont;
3642 				freeb(mp);
3643 				mp = mp1;
3644 				rptr = mp->b_rptr;
3645 			}
3646 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3647 			ip6h = (ip6_t *)rptr;
3648 		}
3649 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3650 
3651 		/*
3652 		 * We need to lie a bit to the user because users inside
3653 		 * labeled compartments should not see their own labels.  We
3654 		 * assume that in all other respects IP has checked the label,
3655 		 * and that the label is always first among the options.  (If
3656 		 * it's not first, then this code won't see it, and the option
3657 		 * will be passed along to the user.)
3658 		 *
3659 		 * If we had multilevel ICMP sockets, then the following code
3660 		 * should be skipped for them to allow the user to see the
3661 		 * label.
3662 		 *
3663 		 * Alignment restrictions in the definition of IP options
3664 		 * (namely, the requirement that the 4-octet DOI goes on a
3665 		 * 4-octet boundary) mean that we know exactly where the option
3666 		 * should start, but we're lenient for other hosts.
3667 		 *
3668 		 * Note that there are no multilevel ICMP or raw IP sockets
3669 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3670 		 */
3671 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3672 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3673 			const uchar_t *ucp =
3674 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3675 			int remlen = ipp.ipp_hopoptslen - 2;
3676 
3677 			while (remlen > 0) {
3678 				if (*ucp == IP6OPT_PAD1) {
3679 					remlen--;
3680 					ucp++;
3681 				} else if (*ucp == IP6OPT_PADN) {
3682 					remlen -= ucp[1] + 2;
3683 					ucp += ucp[1] + 2;
3684 				} else if (*ucp == ip6opt_ls) {
3685 					hopstrip = (ucp -
3686 					    (const uchar_t *)ipp.ipp_hopopts) +
3687 					    ucp[1] + 2;
3688 					hopstrip = (hopstrip + 7) & ~7;
3689 					break;
3690 				} else {
3691 					/* label option must be first */
3692 					break;
3693 				}
3694 			}
3695 		}
3696 	} else {
3697 		hdr_len = IPV6_HDR_LEN;
3698 		ip6i = NULL;
3699 		nexthdr = ip6h->ip6_nxt;
3700 	}
3701 	/*
3702 	 * One special case where IP attaches the IRE needs to
3703 	 * be handled so that we don't send up IRE to the user land.
3704 	 */
3705 	if (nexthdr == IPPROTO_TCP) {
3706 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3707 
3708 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3709 		    mp->b_cont != NULL) {
3710 			mp1 = mp->b_cont;
3711 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3712 				freeb(mp1);
3713 				mp->b_cont = NULL;
3714 			}
3715 		}
3716 	}
3717 	/*
3718 	 * Check a filter for ICMPv6 types if needed.
3719 	 * Verify raw checksums if needed.
3720 	 */
3721 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3722 		if (icmp->icmp_filter != NULL) {
3723 			int type;
3724 
3725 			/* Assumes that IP has done the pullupmsg */
3726 			type = mp->b_rptr[hdr_len];
3727 
3728 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3729 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3730 				freemsg(mp);
3731 				return;
3732 			}
3733 		} else {
3734 			/* Checksum */
3735 			uint16_t	*up;
3736 			uint32_t	sum;
3737 			int		remlen;
3738 
3739 			up = (uint16_t *)&ip6h->ip6_src;
3740 
3741 			remlen = msgdsize(mp) - hdr_len;
3742 			sum = htons(icmp->icmp_proto + remlen)
3743 			    + up[0] + up[1] + up[2] + up[3]
3744 			    + up[4] + up[5] + up[6] + up[7]
3745 			    + up[8] + up[9] + up[10] + up[11]
3746 			    + up[12] + up[13] + up[14] + up[15];
3747 			sum = (sum & 0xffff) + (sum >> 16);
3748 			sum = IP_CSUM(mp, hdr_len, sum);
3749 			if (sum != 0) {
3750 				/* IPv6 RAW checksum failed */
3751 				ip0dbg(("icmp_rput: RAW checksum "
3752 				    "failed %x\n", sum));
3753 				freemsg(mp);
3754 				BUMP_MIB(&is->is_rawip_mib,
3755 				    rawipInCksumErrs);
3756 				return;
3757 			}
3758 		}
3759 	}
3760 	/* Skip all the IPv6 headers per API */
3761 	mp->b_rptr += hdr_len;
3762 
3763 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3764 
3765 	/*
3766 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3767 	 * maintain state information, instead of relying on icmp_t
3768 	 * structure, since there arent any locks protecting these members
3769 	 * and there is a window where there might be a race between a
3770 	 * thread setting options on the write side and a thread reading
3771 	 * these options on the read size.
3772 	 */
3773 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3774 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3775 		if (icmp->icmp_ipv6_recvhopopts &&
3776 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3777 		    ipp.ipp_hopoptslen > hopstrip) {
3778 			udi_size += sizeof (struct T_opthdr) +
3779 			    ipp.ipp_hopoptslen - hopstrip;
3780 			icmp_opt |= IPPF_HOPOPTS;
3781 		}
3782 		if ((icmp->icmp_ipv6_recvdstopts ||
3783 		    icmp->icmp_old_ipv6_recvdstopts) &&
3784 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3785 			udi_size += sizeof (struct T_opthdr) +
3786 			    ipp.ipp_dstoptslen;
3787 			icmp_opt |= IPPF_DSTOPTS;
3788 		}
3789 		if (((icmp->icmp_ipv6_recvdstopts &&
3790 		    icmp->icmp_ipv6_recvrthdr &&
3791 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3792 		    icmp->icmp_ipv6_recvrtdstopts) &&
3793 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3794 			udi_size += sizeof (struct T_opthdr) +
3795 			    ipp.ipp_rtdstoptslen;
3796 			icmp_opt |= IPPF_RTDSTOPTS;
3797 		}
3798 		if (icmp->icmp_ipv6_recvrthdr &&
3799 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3800 			udi_size += sizeof (struct T_opthdr) +
3801 			    ipp.ipp_rthdrlen;
3802 			icmp_opt |= IPPF_RTHDR;
3803 		}
3804 		if (icmp->icmp_ip_recvpktinfo &&
3805 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3806 			udi_size += sizeof (struct T_opthdr) +
3807 			    sizeof (struct in6_pktinfo);
3808 			icmp_opt |= IPPF_IFINDEX;
3809 		}
3810 	}
3811 	if (icmp->icmp_ipv6_recvhoplimit) {
3812 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3813 		icmp_ipv6_recvhoplimit = B_TRUE;
3814 	}
3815 
3816 	if (icmp->icmp_ipv6_recvtclass)
3817 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3818 
3819 	/*
3820 	 * If SO_TIMESTAMP is set allocate the appropriate sized
3821 	 * buffer. Since gethrestime() expects a pointer aligned
3822 	 * argument, we allocate space necessary for extra
3823 	 * alignment (even though it might not be used).
3824 	 */
3825 	if (icmp->icmp_timestamp) {
3826 		udi_size += sizeof (struct T_opthdr) +
3827 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3828 	}
3829 
3830 	mp1 = allocb(udi_size, BPRI_MED);
3831 	if (mp1 == NULL) {
3832 		freemsg(mp);
3833 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3834 		return;
3835 	}
3836 	mp1->b_cont = mp;
3837 	mp = mp1;
3838 	mp->b_datap->db_type = M_PROTO;
3839 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3840 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3841 	tudi->PRIM_type = T_UNITDATA_IND;
3842 	tudi->SRC_length = sizeof (sin6_t);
3843 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3844 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3845 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3846 	tudi->OPT_length = udi_size;
3847 	sin6 = (sin6_t *)&tudi[1];
3848 	sin6->sin6_port = 0;
3849 	sin6->sin6_family = AF_INET6;
3850 
3851 	sin6->sin6_addr = ip6h->ip6_src;
3852 	/* No sin6_flowinfo per API */
3853 	sin6->sin6_flowinfo = 0;
3854 	/* For link-scope source pass up scope id */
3855 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3856 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3857 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3858 	else
3859 		sin6->sin6_scope_id = 0;
3860 
3861 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3862 	    icmp->icmp_zoneid, is->is_netstack);
3863 
3864 	if (udi_size != 0) {
3865 		uchar_t *dstopt;
3866 
3867 		dstopt = (uchar_t *)&sin6[1];
3868 		if (icmp_opt & IPPF_IFINDEX) {
3869 			struct T_opthdr *toh;
3870 			struct in6_pktinfo *pkti;
3871 
3872 			toh = (struct T_opthdr *)dstopt;
3873 			toh->level = IPPROTO_IPV6;
3874 			toh->name = IPV6_PKTINFO;
3875 			toh->len = sizeof (struct T_opthdr) +
3876 			    sizeof (*pkti);
3877 			toh->status = 0;
3878 			dstopt += sizeof (struct T_opthdr);
3879 			pkti = (struct in6_pktinfo *)dstopt;
3880 			pkti->ipi6_addr = ip6h->ip6_dst;
3881 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3882 			dstopt += sizeof (*pkti);
3883 			udi_size -= toh->len;
3884 		}
3885 		if (icmp_ipv6_recvhoplimit) {
3886 			struct T_opthdr *toh;
3887 
3888 			toh = (struct T_opthdr *)dstopt;
3889 			toh->level = IPPROTO_IPV6;
3890 			toh->name = IPV6_HOPLIMIT;
3891 			toh->len = sizeof (struct T_opthdr) +
3892 			    sizeof (uint_t);
3893 			toh->status = 0;
3894 			dstopt += sizeof (struct T_opthdr);
3895 			*(uint_t *)dstopt = ip6h->ip6_hops;
3896 			dstopt += sizeof (uint_t);
3897 			udi_size -= toh->len;
3898 		}
3899 		if (icmp->icmp_ipv6_recvtclass) {
3900 			struct T_opthdr *toh;
3901 
3902 			toh = (struct T_opthdr *)dstopt;
3903 			toh->level = IPPROTO_IPV6;
3904 			toh->name = IPV6_TCLASS;
3905 			toh->len = sizeof (struct T_opthdr) +
3906 			    sizeof (uint_t);
3907 			toh->status = 0;
3908 			dstopt += sizeof (struct T_opthdr);
3909 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3910 			dstopt += sizeof (uint_t);
3911 			udi_size -= toh->len;
3912 		}
3913 		if (icmp->icmp_timestamp) {
3914 			struct  T_opthdr *toh;
3915 
3916 			toh = (struct T_opthdr *)dstopt;
3917 			toh->level = SOL_SOCKET;
3918 			toh->name = SCM_TIMESTAMP;
3919 			toh->len = sizeof (struct T_opthdr) +
3920 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3921 			toh->status = 0;
3922 			dstopt += sizeof (struct T_opthdr);
3923 			/* Align for gethrestime() */
3924 			dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
3925 			    sizeof (intptr_t));
3926 			gethrestime((timestruc_t *)dstopt);
3927 			dstopt = (uchar_t *)toh + toh->len;
3928 			udi_size -= toh->len;
3929 		}
3930 
3931 		if (icmp_opt & IPPF_HOPOPTS) {
3932 			struct T_opthdr *toh;
3933 
3934 			toh = (struct T_opthdr *)dstopt;
3935 			toh->level = IPPROTO_IPV6;
3936 			toh->name = IPV6_HOPOPTS;
3937 			toh->len = sizeof (struct T_opthdr) +
3938 			    ipp.ipp_hopoptslen - hopstrip;
3939 			toh->status = 0;
3940 			dstopt += sizeof (struct T_opthdr);
3941 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
3942 			    ipp.ipp_hopoptslen - hopstrip);
3943 			if (hopstrip > 0) {
3944 				/* copy next header value and fake length */
3945 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
3946 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
3947 				    hopstrip / 8;
3948 			}
3949 			dstopt += ipp.ipp_hopoptslen - hopstrip;
3950 			udi_size -= toh->len;
3951 		}
3952 		if (icmp_opt & IPPF_RTDSTOPTS) {
3953 			struct T_opthdr *toh;
3954 
3955 			toh = (struct T_opthdr *)dstopt;
3956 			toh->level = IPPROTO_IPV6;
3957 			toh->name = IPV6_DSTOPTS;
3958 			toh->len = sizeof (struct T_opthdr) +
3959 			    ipp.ipp_rtdstoptslen;
3960 			toh->status = 0;
3961 			dstopt += sizeof (struct T_opthdr);
3962 			bcopy(ipp.ipp_rtdstopts, dstopt,
3963 			    ipp.ipp_rtdstoptslen);
3964 			dstopt += ipp.ipp_rtdstoptslen;
3965 			udi_size -= toh->len;
3966 		}
3967 		if (icmp_opt & IPPF_RTHDR) {
3968 			struct T_opthdr *toh;
3969 
3970 			toh = (struct T_opthdr *)dstopt;
3971 			toh->level = IPPROTO_IPV6;
3972 			toh->name = IPV6_RTHDR;
3973 			toh->len = sizeof (struct T_opthdr) +
3974 			    ipp.ipp_rthdrlen;
3975 			toh->status = 0;
3976 			dstopt += sizeof (struct T_opthdr);
3977 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
3978 			dstopt += ipp.ipp_rthdrlen;
3979 			udi_size -= toh->len;
3980 		}
3981 		if (icmp_opt & IPPF_DSTOPTS) {
3982 			struct T_opthdr *toh;
3983 
3984 			toh = (struct T_opthdr *)dstopt;
3985 			toh->level = IPPROTO_IPV6;
3986 			toh->name = IPV6_DSTOPTS;
3987 			toh->len = sizeof (struct T_opthdr) +
3988 			    ipp.ipp_dstoptslen;
3989 			toh->status = 0;
3990 			dstopt += sizeof (struct T_opthdr);
3991 			bcopy(ipp.ipp_dstopts, dstopt,
3992 			    ipp.ipp_dstoptslen);
3993 			dstopt += ipp.ipp_dstoptslen;
3994 			udi_size -= toh->len;
3995 		}
3996 		/* Consumed all of allocated space */
3997 		ASSERT(udi_size == 0);
3998 	}
3999 	BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
4000 
4001 deliver:
4002 	if (IPCL_IS_NONSTR(connp)) {
4003 		if ((*connp->conn_upcalls->su_recv)
4004 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
4005 		    NULL) < 0) {
4006 			mutex_enter(&icmp->icmp_recv_lock);
4007 			if (error == ENOSPC) {
4008 				/*
4009 				 * let's confirm while holding the lock
4010 				 */
4011 				if ((*connp->conn_upcalls->su_recv)
4012 				    (connp->conn_upper_handle, NULL, 0, 0,
4013 				    &error, NULL) < 0) {
4014 					if (error == ENOSPC) {
4015 						connp->conn_flow_cntrld =
4016 						    B_TRUE;
4017 					} else {
4018 						ASSERT(error == EOPNOTSUPP);
4019 					}
4020 				}
4021 				mutex_exit(&icmp->icmp_recv_lock);
4022 			} else {
4023 				ASSERT(error == EOPNOTSUPP);
4024 				icmp_queue_fallback(icmp, mp);
4025 			}
4026 		}
4027 	} else {
4028 		putnext(connp->conn_rq, mp);
4029 	}
4030 	ASSERT(MUTEX_NOT_HELD(&icmp->icmp_recv_lock));
4031 }
4032 
4033 /*
4034  * return SNMP stuff in buffer in mpdata
4035  */
4036 mblk_t *
4037 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
4038 {
4039 	mblk_t			*mpdata;
4040 	struct opthdr		*optp;
4041 	conn_t			*connp = Q_TO_CONN(q);
4042 	icmp_stack_t		*is = connp->conn_netstack->netstack_icmp;
4043 	mblk_t			*mp2ctl;
4044 
4045 	/*
4046 	 * make a copy of the original message
4047 	 */
4048 	mp2ctl = copymsg(mpctl);
4049 
4050 	if (mpctl == NULL ||
4051 	    (mpdata = mpctl->b_cont) == NULL) {
4052 		freemsg(mpctl);
4053 		freemsg(mp2ctl);
4054 		return (0);
4055 	}
4056 
4057 	/* fixed length structure for IPv4 and IPv6 counters */
4058 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4059 	optp->level = EXPER_RAWIP;
4060 	optp->name = 0;
4061 	(void) snmp_append_data(mpdata, (char *)&is->is_rawip_mib,
4062 	    sizeof (is->is_rawip_mib));
4063 	optp->len = msgdsize(mpdata);
4064 	qreply(q, mpctl);
4065 
4066 	return (mp2ctl);
4067 }
4068 
4069 /*
4070  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
4071  * TODO:  If this ever actually tries to set anything, it needs to be
4072  * to do the appropriate locking.
4073  */
4074 /* ARGSUSED */
4075 int
4076 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4077     uchar_t *ptr, int len)
4078 {
4079 	switch (level) {
4080 	case EXPER_RAWIP:
4081 		return (0);
4082 	default:
4083 		return (1);
4084 	}
4085 }
4086 
4087 /* Report for ndd "icmp_status" */
4088 /* ARGSUSED */
4089 static int
4090 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4091 {
4092 	conn_t  *connp;
4093 	ip_stack_t *ipst;
4094 	char	laddrbuf[INET6_ADDRSTRLEN];
4095 	char	faddrbuf[INET6_ADDRSTRLEN];
4096 	int	i;
4097 
4098 	(void) mi_mpprintf(mp,
4099 	    "RAWIP    " MI_COL_HDRPAD_STR
4100 	/*   01234567[89ABCDEF] */
4101 	    "  src addr        dest addr       state");
4102 	/*   xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */
4103 
4104 	connp = Q_TO_CONN(q);
4105 	ipst = connp->conn_netstack->netstack_ip;
4106 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4107 		connf_t *connfp;
4108 		char	*state;
4109 
4110 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4111 		connp = NULL;
4112 
4113 		while ((connp = ipcl_get_next_conn(connfp, connp,
4114 		    IPCL_RAWIPCONN)) != NULL) {
4115 			icmp_t  *icmp;
4116 
4117 			mutex_enter(&(connp)->conn_lock);
4118 			icmp = connp->conn_icmp;
4119 
4120 			if (icmp->icmp_state == TS_UNBND)
4121 				state = "UNBOUND";
4122 			else if (icmp->icmp_state == TS_IDLE)
4123 				state = "IDLE";
4124 			else if (icmp->icmp_state == TS_DATA_XFER)
4125 				state = "CONNECTED";
4126 			else
4127 				state = "UnkState";
4128 
4129 			(void) mi_mpprintf(mp, MI_COL_PTRFMT_STR "%s %s %s",
4130 			    (void *)icmp,
4131 			    inet_ntop(AF_INET6, &icmp->icmp_v6dst.sin6_addr,
4132 			    faddrbuf,
4133 			    sizeof (faddrbuf)),
4134 			    inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf,
4135 			    sizeof (laddrbuf)),
4136 			    state);
4137 			mutex_exit(&(connp)->conn_lock);
4138 		}
4139 	}
4140 	return (0);
4141 }
4142 
4143 /*
4144  * This routine creates a T_UDERROR_IND message and passes it upstream.
4145  * The address and options are copied from the T_UNITDATA_REQ message
4146  * passed in mp.  This message is freed.
4147  */
4148 static void
4149 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
4150 {
4151 	mblk_t	*mp1;
4152 	uchar_t	*rptr = mp->b_rptr;
4153 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
4154 
4155 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
4156 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
4157 	    tudr->OPT_length, err);
4158 	if (mp1)
4159 		qreply(q, mp1);
4160 	freemsg(mp);
4161 }
4162 
4163 
4164 static int
4165 rawip_do_unbind(conn_t *connp)
4166 {
4167 	icmp_t *icmp = connp->conn_icmp;
4168 
4169 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4170 	/* If a bind has not been done, we can't unbind. */
4171 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
4172 		rw_exit(&icmp->icmp_rwlock);
4173 		return (-TOUTSTATE);
4174 	}
4175 	icmp->icmp_pending_op = T_UNBIND_REQ;
4176 	rw_exit(&icmp->icmp_rwlock);
4177 
4178 	/*
4179 	 * Call ip to unbind
4180 	 */
4181 
4182 	ip_unbind(connp);
4183 
4184 	/*
4185 	 * Once we're unbound from IP, the pending operation may be cleared
4186 	 * here.
4187 	 */
4188 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4189 	V6_SET_ZERO(icmp->icmp_v6src);
4190 	V6_SET_ZERO(icmp->icmp_bound_v6src);
4191 	icmp->icmp_pending_op = -1;
4192 	icmp->icmp_state = TS_UNBND;
4193 	if (icmp->icmp_family == AF_INET6)
4194 		(void) icmp_build_hdrs(icmp);
4195 	rw_exit(&icmp->icmp_rwlock);
4196 	return (0);
4197 }
4198 
4199 /*
4200  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
4201  * After some error checking, the message is passed downstream to ip.
4202  */
4203 static void
4204 icmp_tpi_unbind(queue_t *q, mblk_t *mp)
4205 {
4206 	conn_t	*connp = Q_TO_CONN(q);
4207 	int	error;
4208 
4209 	ASSERT(mp->b_cont == NULL);
4210 	error = rawip_do_unbind(connp);
4211 	if (error) {
4212 		if (error < 0) {
4213 			icmp_err_ack(q, mp, -error, 0);
4214 		} else {
4215 			icmp_err_ack(q, mp, 0, error);
4216 		}
4217 		return;
4218 	}
4219 
4220 	/*
4221 	 * Convert mp into a T_OK_ACK
4222 	 */
4223 
4224 	mp = mi_tpi_ok_ack_alloc(mp);
4225 
4226 	/*
4227 	 * should not happen in practice... T_OK_ACK is smaller than the
4228 	 * original message.
4229 	 */
4230 	ASSERT(mp != NULL);
4231 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4232 	qreply(q, mp);
4233 }
4234 
4235 
4236 /*
4237  * Process IPv4 packets that already include an IP header.
4238  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
4239  * IPPROTO_IGMP).
4240  */
4241 static int
4242 icmp_wput_hdrincl(queue_t *q, conn_t *connp, mblk_t *mp, icmp_t *icmp,
4243     ip4_pkt_t *pktinfop)
4244 {
4245 	icmp_stack_t *is = icmp->icmp_is;
4246 	ipha_t	*ipha;
4247 	int	ip_hdr_length;
4248 	int	tp_hdr_len;
4249 	mblk_t	*mp1;
4250 	uint_t	pkt_len;
4251 	ip_opt_info_t optinfo;
4252 
4253 	optinfo.ip_opt_flags = 0;
4254 	optinfo.ip_opt_ill_index = 0;
4255 	ipha = (ipha_t *)mp->b_rptr;
4256 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4257 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
4258 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4259 			ASSERT(icmp != NULL);
4260 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4261 			freemsg(mp);
4262 			return (0);
4263 		}
4264 		ipha = (ipha_t *)mp->b_rptr;
4265 	}
4266 	ipha->ipha_version_and_hdr_length =
4267 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
4268 
4269 	/*
4270 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4271 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4272 	 * tell IP that the application has sent a complete IP header and not
4273 	 * to compute the transport checksum nor change the DF flag.
4274 	 */
4275 	ipha->ipha_ident = IP_HDR_INCLUDED;
4276 	ipha->ipha_hdr_checksum = 0;
4277 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
4278 	/* Insert options if any */
4279 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4280 		/*
4281 		 * Put the IP header plus any transport header that is
4282 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
4283 		 * that at least the checksum field is in the first mblk.)
4284 		 */
4285 		switch (ipha->ipha_protocol) {
4286 		case IPPROTO_UDP:
4287 			tp_hdr_len = 8;
4288 			break;
4289 		case IPPROTO_TCP:
4290 			tp_hdr_len = 20;
4291 			break;
4292 		default:
4293 			tp_hdr_len = 0;
4294 			break;
4295 		}
4296 		/*
4297 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
4298 		 * tp_hdr_len bytes will be in a single mblk.
4299 		 */
4300 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
4301 		    tp_hdr_len)) {
4302 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
4303 			    tp_hdr_len)) {
4304 				BUMP_MIB(&is->is_rawip_mib,
4305 				    rawipOutErrors);
4306 				freemsg(mp);
4307 				return (0);
4308 			}
4309 			ipha = (ipha_t *)mp->b_rptr;
4310 		}
4311 
4312 		/*
4313 		 * if the length is larger then the max allowed IP packet,
4314 		 * then send an error and abort the processing.
4315 		 */
4316 		pkt_len = ntohs(ipha->ipha_length)
4317 		    + icmp->icmp_ip_snd_options_len;
4318 		if (pkt_len > IP_MAXPACKET) {
4319 			return (EMSGSIZE);
4320 		}
4321 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra +
4322 		    tp_hdr_len, BPRI_LO))) {
4323 			return (ENOMEM);
4324 		}
4325 		mp1->b_rptr += is->is_wroff_extra;
4326 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
4327 
4328 		ipha->ipha_length = htons((uint16_t)pkt_len);
4329 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
4330 
4331 		/* Copy transport header if any */
4332 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
4333 		mp1->b_wptr += tp_hdr_len;
4334 
4335 		/* Add options */
4336 		ipha = (ipha_t *)mp1->b_rptr;
4337 		bcopy(icmp->icmp_ip_snd_options, &ipha[1],
4338 		    icmp->icmp_ip_snd_options_len);
4339 
4340 		/* Drop IP header and transport header from original */
4341 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
4342 
4343 		mp1->b_cont = mp;
4344 		mp = mp1;
4345 		/*
4346 		 * Massage source route putting first source
4347 		 * route in ipha_dst.
4348 		 */
4349 		(void) ip_massage_options(ipha, is->is_netstack);
4350 	}
4351 
4352 	if (pktinfop != NULL) {
4353 		/*
4354 		 * Over write the source address provided in the header
4355 		 */
4356 		if (pktinfop->ip4_addr != INADDR_ANY) {
4357 			ipha->ipha_src = pktinfop->ip4_addr;
4358 			optinfo.ip_opt_flags = IP_VERIFY_SRC;
4359 		}
4360 
4361 		if (pktinfop->ip4_ill_index != 0) {
4362 			optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4363 		}
4364 	}
4365 
4366 	mblk_setcred(mp, connp->conn_cred);
4367 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4368 	return (0);
4369 }
4370 
4371 static int
4372 icmp_update_label(icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
4373 {
4374 	int err;
4375 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4376 	icmp_stack_t		*is = icmp->icmp_is;
4377 	conn_t	*connp = icmp->icmp_connp;
4378 
4379 	err = tsol_compute_label(DB_CREDDEF(mp, connp->conn_cred), dst,
4380 	    opt_storage, connp->conn_mac_exempt,
4381 	    is->is_netstack->netstack_ip);
4382 	if (err == 0) {
4383 		err = tsol_update_options(&icmp->icmp_ip_snd_options,
4384 		    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
4385 		    opt_storage);
4386 	}
4387 	if (err != 0) {
4388 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4389 		DTRACE_PROBE4(
4390 		    tx__ip__log__drop__updatelabel__icmp,
4391 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4392 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4393 		return (err);
4394 	}
4395 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
4396 	return (0);
4397 }
4398 
4399 /*
4400  * This routine handles all messages passed downstream.  It either
4401  * consumes the message or passes it downstream; it never queues a
4402  * a message.
4403  */
4404 static void
4405 icmp_wput(queue_t *q, mblk_t *mp)
4406 {
4407 	uchar_t	*rptr = mp->b_rptr;
4408 	ipha_t	*ipha;
4409 	mblk_t	*mp1;
4410 #define	tudr ((struct T_unitdata_req *)rptr)
4411 	size_t	ip_len;
4412 	conn_t	*connp = Q_TO_CONN(q);
4413 	icmp_t	*icmp = connp->conn_icmp;
4414 	icmp_stack_t *is = icmp->icmp_is;
4415 	sin6_t	*sin6;
4416 	sin_t	*sin;
4417 	ipaddr_t	v4dst;
4418 	ip4_pkt_t	pktinfo;
4419 	ip4_pkt_t	*pktinfop = &pktinfo;
4420 	ip6_pkt_t	ipp_s;  /* For ancillary data options */
4421 	ip6_pkt_t	*ipp = &ipp_s;
4422 	int error;
4423 
4424 	ipp->ipp_fields = 0;
4425 	ipp->ipp_sticky_ignored = 0;
4426 
4427 	switch (mp->b_datap->db_type) {
4428 	case M_DATA:
4429 		if (icmp->icmp_hdrincl) {
4430 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
4431 			ipha = (ipha_t *)mp->b_rptr;
4432 			if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH) {
4433 				if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4434 					BUMP_MIB(&is->is_rawip_mib,
4435 					    rawipOutErrors);
4436 					freemsg(mp);
4437 					return;
4438 				}
4439 				ipha = (ipha_t *)mp->b_rptr;
4440 			}
4441 			/*
4442 			 * If this connection was used for v6 (inconceivable!)
4443 			 * or if we have a new destination, then it's time to
4444 			 * figure a new label.
4445 			 */
4446 			if (is_system_labeled() &&
4447 			    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4448 			    V4_PART_OF_V6(icmp->icmp_v6lastdst) !=
4449 			    ipha->ipha_dst)) {
4450 				error = icmp_update_label(icmp, mp,
4451 				    ipha->ipha_dst);
4452 				if (error != 0) {
4453 					icmp_ud_err(q, mp, error);
4454 					return;
4455 				}
4456 			}
4457 			error = icmp_wput_hdrincl(q, connp, mp, icmp, NULL);
4458 			if (error != 0)
4459 				icmp_ud_err(q, mp, error);
4460 			return;
4461 		}
4462 		freemsg(mp);
4463 		return;
4464 	case M_PROTO:
4465 	case M_PCPROTO:
4466 		ip_len = mp->b_wptr - rptr;
4467 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4468 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4469 			if (((union T_primitives *)rptr)->type
4470 			    == T_UNITDATA_REQ)
4471 				break;
4472 		}
4473 		/* FALLTHRU */
4474 	default:
4475 		icmp_wput_other(q, mp);
4476 		return;
4477 	}
4478 
4479 	/* Handle T_UNITDATA_REQ messages here. */
4480 
4481 	mp1 = mp->b_cont;
4482 	if (mp1 == NULL) {
4483 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4484 		icmp_ud_err(q, mp, EPROTO);
4485 		return;
4486 	}
4487 
4488 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4489 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4490 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4491 		return;
4492 	}
4493 
4494 	switch (icmp->icmp_family) {
4495 	case AF_INET6:
4496 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4497 		if (!OK_32PTR((char *)sin6) ||
4498 		    tudr->DEST_length != sizeof (sin6_t) ||
4499 		    sin6->sin6_family != AF_INET6) {
4500 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4501 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4502 			return;
4503 		}
4504 
4505 		/* No support for mapped addresses on raw sockets */
4506 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4507 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4508 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4509 			return;
4510 		}
4511 
4512 		/*
4513 		 * Destination is a native IPv6 address.
4514 		 * Send out an IPv6 format packet.
4515 		 */
4516 		if (tudr->OPT_length != 0) {
4517 			int error;
4518 
4519 			error = 0;
4520 			if (icmp_unitdata_opt_process(q, mp, &error,
4521 			    (void *)ipp) < 0) {
4522 				/* failure */
4523 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4524 				icmp_ud_err(q, mp, error);
4525 				return;
4526 			}
4527 			ASSERT(error == 0);
4528 		}
4529 
4530 		error = raw_ip_send_data_v6(q, connp, mp1, sin6, ipp);
4531 		goto done;
4532 
4533 	case AF_INET:
4534 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4535 		if (!OK_32PTR((char *)sin) ||
4536 		    tudr->DEST_length != sizeof (sin_t) ||
4537 		    sin->sin_family != AF_INET) {
4538 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4539 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4540 			return;
4541 		}
4542 		/* Extract and ipaddr */
4543 		v4dst = sin->sin_addr.s_addr;
4544 		break;
4545 
4546 	default:
4547 		ASSERT(0);
4548 	}
4549 
4550 	pktinfop->ip4_ill_index = 0;
4551 	pktinfop->ip4_addr = INADDR_ANY;
4552 
4553 	/*
4554 	 * If options passed in, feed it for verification and handling
4555 	 */
4556 	if (tudr->OPT_length != 0) {
4557 		int error;
4558 
4559 		error = 0;
4560 		if (icmp_unitdata_opt_process(q, mp, &error,
4561 		    (void *)pktinfop) < 0) {
4562 			/* failure */
4563 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4564 			icmp_ud_err(q, mp, error);
4565 			return;
4566 		}
4567 		ASSERT(error == 0);
4568 		/*
4569 		 * Note: Success in processing options.
4570 		 * mp option buffer represented by
4571 		 * OPT_length/offset now potentially modified
4572 		 * and contain option setting results
4573 		 */
4574 	}
4575 
4576 	error = raw_ip_send_data_v4(q, connp, mp1, v4dst, pktinfop);
4577 done:
4578 	if (error != 0) {
4579 		icmp_ud_err(q, mp, error);
4580 		return;
4581 	} else {
4582 		mp->b_cont = NULL;
4583 		freeb(mp);
4584 	}
4585 }
4586 
4587 
4588 /* ARGSUSED */
4589 static void
4590 icmp_wput_fallback(queue_t *q, mblk_t *mp)
4591 {
4592 #ifdef DEBUG
4593 	cmn_err(CE_CONT, "icmp_wput_fallback: Message during fallback \n");
4594 #endif
4595 	freemsg(mp);
4596 }
4597 
4598 static int
4599 raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp, ipaddr_t v4dst,
4600     ip4_pkt_t *pktinfop)
4601 {
4602 	ipha_t	*ipha;
4603 	size_t	ip_len;
4604 	icmp_t	*icmp = connp->conn_icmp;
4605 	icmp_stack_t *is = icmp->icmp_is;
4606 	int	ip_hdr_length;
4607 	ip_opt_info_t	optinfo;
4608 
4609 	optinfo.ip_opt_flags = 0;
4610 	optinfo.ip_opt_ill_index = 0;
4611 
4612 	if (icmp->icmp_state == TS_UNBND) {
4613 		/* If a port has not been bound to the stream, fail. */
4614 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4615 		return (EPROTO);
4616 	}
4617 
4618 	if (v4dst == INADDR_ANY)
4619 		v4dst = htonl(INADDR_LOOPBACK);
4620 
4621 	/* Check if our saved options are valid; update if not */
4622 	if (is_system_labeled() &&
4623 	    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4624 	    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst)) {
4625 		int error = icmp_update_label(icmp, mp, v4dst);
4626 
4627 		if (error != 0)
4628 			return (error);
4629 	}
4630 
4631 	/* Protocol 255 contains full IP headers */
4632 	if (icmp->icmp_hdrincl)
4633 		return (icmp_wput_hdrincl(q, connp, mp, icmp, pktinfop));
4634 
4635 	/* Add an IP header */
4636 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4637 	ipha = (ipha_t *)&mp->b_rptr[-ip_hdr_length];
4638 	if ((uchar_t *)ipha < mp->b_datap->db_base ||
4639 	    mp->b_datap->db_ref != 1 ||
4640 	    !OK_32PTR(ipha)) {
4641 		mblk_t	*mp1;
4642 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra,
4643 		    BPRI_LO))) {
4644 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4645 			return (ENOMEM);
4646 		}
4647 		mp1->b_cont = mp;
4648 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4649 		mp1->b_wptr = (uchar_t *)ipha;
4650 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4651 		mp = mp1;
4652 	}
4653 #ifdef	_BIG_ENDIAN
4654 	/* Set version, header length, and tos */
4655 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4656 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4657 	    icmp->icmp_type_of_service);
4658 	/* Set ttl and protocol */
4659 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4660 #else
4661 	/* Set version, header length, and tos */
4662 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4663 	    ((icmp->icmp_type_of_service << 8) |
4664 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4665 	/* Set ttl and protocol */
4666 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4667 #endif
4668 	if (pktinfop->ip4_addr != INADDR_ANY) {
4669 		ipha->ipha_src = pktinfop->ip4_addr;
4670 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4671 	} else {
4672 
4673 		/*
4674 		 * Copy our address into the packet.  If this is zero,
4675 		 * ip will fill in the real source address.
4676 		 */
4677 		IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4678 	}
4679 
4680 	ipha->ipha_fragment_offset_and_flags = 0;
4681 
4682 	if (pktinfop->ip4_ill_index != 0) {
4683 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4684 	}
4685 
4686 
4687 	/*
4688 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4689 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4690 	 * tell IP that the application has sent a complete IP header and not
4691 	 * to compute the transport checksum nor change the DF flag.
4692 	 */
4693 	ipha->ipha_ident = IP_HDR_INCLUDED;
4694 
4695 	/* Finish common formatting of the packet. */
4696 	mp->b_rptr = (uchar_t *)ipha;
4697 
4698 	ip_len = mp->b_wptr - (uchar_t *)ipha;
4699 	if (mp->b_cont != NULL)
4700 		ip_len += msgdsize(mp->b_cont);
4701 
4702 	/*
4703 	 * Set the length into the IP header.
4704 	 * If the length is greater than the maximum allowed by IP,
4705 	 * then free the message and return. Do not try and send it
4706 	 * as this can cause problems in layers below.
4707 	 */
4708 	if (ip_len > IP_MAXPACKET) {
4709 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4710 		return (EMSGSIZE);
4711 	}
4712 	ipha->ipha_length = htons((uint16_t)ip_len);
4713 	/*
4714 	 * Copy in the destination address request
4715 	 */
4716 	ipha->ipha_dst = v4dst;
4717 
4718 	/*
4719 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4720 	 */
4721 	if (CLASSD(v4dst))
4722 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4723 
4724 	/* Copy in options if any */
4725 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4726 		bcopy(icmp->icmp_ip_snd_options,
4727 		    &ipha[1], icmp->icmp_ip_snd_options_len);
4728 		/*
4729 		 * Massage source route putting first source route in ipha_dst.
4730 		 * Ignore the destination in the T_unitdata_req.
4731 		 */
4732 		(void) ip_massage_options(ipha, is->is_netstack);
4733 	}
4734 
4735 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
4736 	mblk_setcred(mp, connp->conn_cred);
4737 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4738 	return (0);
4739 }
4740 
4741 static int
4742 icmp_update_label_v6(icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4743 {
4744 	int err;
4745 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4746 	icmp_stack_t		*is = icmp->icmp_is;
4747 	conn_t	*connp = icmp->icmp_connp;
4748 
4749 	err = tsol_compute_label_v6(DB_CREDDEF(mp, connp->conn_cred), dst,
4750 	    opt_storage, connp->conn_mac_exempt,
4751 	    is->is_netstack->netstack_ip);
4752 	if (err == 0) {
4753 		err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4754 		    &icmp->icmp_label_len_v6, opt_storage);
4755 	}
4756 	if (err != 0) {
4757 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4758 		DTRACE_PROBE4(
4759 		    tx__ip__log__drop__updatelabel__icmp6,
4760 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4761 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4762 		return (err);
4763 	}
4764 
4765 	icmp->icmp_v6lastdst = *dst;
4766 	return (0);
4767 }
4768 
4769 /*
4770  * raw_ip_send_data_v6():
4771  * Assumes that icmp_wput did some sanity checking on the destination
4772  * address, but that the label may not yet be correct.
4773  */
4774 static int
4775 raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp, sin6_t *sin6,
4776     ip6_pkt_t *ipp)
4777 {
4778 	ip6_t			*ip6h;
4779 	ip6i_t			*ip6i;	/* mp->b_rptr even if no ip6i_t */
4780 	int			ip_hdr_len = IPV6_HDR_LEN;
4781 	size_t			ip_len;
4782 	icmp_t			*icmp = connp->conn_icmp;
4783 	icmp_stack_t		*is = icmp->icmp_is;
4784 	ip6_pkt_t		*tipp;
4785 	uint32_t		csum = 0;
4786 	uint_t			ignore = 0;
4787 	uint_t			option_exists = 0, is_sticky = 0;
4788 	uint8_t			*cp;
4789 	uint8_t			*nxthdr_ptr;
4790 	in6_addr_t		ip6_dst;
4791 
4792 	/*
4793 	 * If the local address is a mapped address return
4794 	 * an error.
4795 	 * It would be possible to send an IPv6 packet but the
4796 	 * response would never make it back to the application
4797 	 * since it is bound to a mapped address.
4798 	 */
4799 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4800 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4801 		return (EADDRNOTAVAIL);
4802 	}
4803 
4804 	ignore = ipp->ipp_sticky_ignored;
4805 	if (sin6->sin6_scope_id != 0 &&
4806 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4807 		/*
4808 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4809 		 * option nor ancillary data.  It needs to be
4810 		 * explicitly set in options_exists.
4811 		 */
4812 		option_exists |= IPPF_SCOPE_ID;
4813 	}
4814 
4815 	/*
4816 	 * Compute the destination address
4817 	 */
4818 	ip6_dst = sin6->sin6_addr;
4819 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
4820 		ip6_dst = ipv6_loopback;
4821 
4822 	/*
4823 	 * If we're not going to the same destination as last time, then
4824 	 * recompute the label required.  This is done in a separate routine to
4825 	 * avoid blowing up our stack here.
4826 	 */
4827 	if (is_system_labeled() &&
4828 	    !IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst)) {
4829 		int error = 0;
4830 
4831 		error = icmp_update_label_v6(icmp, mp, &ip6_dst);
4832 		if (error != 0)
4833 			return (error);
4834 	}
4835 
4836 	/*
4837 	 * If there's a security label here, then we ignore any options the
4838 	 * user may try to set.  We keep the peer's label as a hidden sticky
4839 	 * option.
4840 	 */
4841 	if (icmp->icmp_label_len_v6 > 0) {
4842 		ignore &= ~IPPF_HOPOPTS;
4843 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
4844 	}
4845 
4846 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
4847 	    (ipp->ipp_fields == 0)) {
4848 		/* No sticky options nor ancillary data. */
4849 		goto no_options;
4850 	}
4851 
4852 	/*
4853 	 * Go through the options figuring out where each is going to
4854 	 * come from and build two masks.  The first mask indicates if
4855 	 * the option exists at all.  The second mask indicates if the
4856 	 * option is sticky or ancillary.
4857 	 */
4858 	if (!(ignore & IPPF_HOPOPTS)) {
4859 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
4860 			option_exists |= IPPF_HOPOPTS;
4861 			ip_hdr_len += ipp->ipp_hopoptslen;
4862 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
4863 			option_exists |= IPPF_HOPOPTS;
4864 			is_sticky |= IPPF_HOPOPTS;
4865 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen;
4866 		}
4867 	}
4868 
4869 	if (!(ignore & IPPF_RTHDR)) {
4870 		if (ipp->ipp_fields & IPPF_RTHDR) {
4871 			option_exists |= IPPF_RTHDR;
4872 			ip_hdr_len += ipp->ipp_rthdrlen;
4873 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
4874 			option_exists |= IPPF_RTHDR;
4875 			is_sticky |= IPPF_RTHDR;
4876 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
4877 		}
4878 	}
4879 
4880 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
4881 		/*
4882 		 * Need to have a router header to use these.
4883 		 */
4884 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
4885 			option_exists |= IPPF_RTDSTOPTS;
4886 			ip_hdr_len += ipp->ipp_rtdstoptslen;
4887 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
4888 			option_exists |= IPPF_RTDSTOPTS;
4889 			is_sticky |= IPPF_RTDSTOPTS;
4890 			ip_hdr_len +=
4891 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
4892 		}
4893 	}
4894 
4895 	if (!(ignore & IPPF_DSTOPTS)) {
4896 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
4897 			option_exists |= IPPF_DSTOPTS;
4898 			ip_hdr_len += ipp->ipp_dstoptslen;
4899 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
4900 			option_exists |= IPPF_DSTOPTS;
4901 			is_sticky |= IPPF_DSTOPTS;
4902 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
4903 		}
4904 	}
4905 
4906 	if (!(ignore & IPPF_IFINDEX)) {
4907 		if (ipp->ipp_fields & IPPF_IFINDEX) {
4908 			option_exists |= IPPF_IFINDEX;
4909 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
4910 			option_exists |= IPPF_IFINDEX;
4911 			is_sticky |= IPPF_IFINDEX;
4912 		}
4913 	}
4914 
4915 	if (!(ignore & IPPF_ADDR)) {
4916 		if (ipp->ipp_fields & IPPF_ADDR) {
4917 			option_exists |= IPPF_ADDR;
4918 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
4919 			option_exists |= IPPF_ADDR;
4920 			is_sticky |= IPPF_ADDR;
4921 		}
4922 	}
4923 
4924 	if (!(ignore & IPPF_DONTFRAG)) {
4925 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
4926 			option_exists |= IPPF_DONTFRAG;
4927 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
4928 			option_exists |= IPPF_DONTFRAG;
4929 			is_sticky |= IPPF_DONTFRAG;
4930 		}
4931 	}
4932 
4933 	if (!(ignore & IPPF_USE_MIN_MTU)) {
4934 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
4935 			option_exists |= IPPF_USE_MIN_MTU;
4936 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
4937 		    IPPF_USE_MIN_MTU) {
4938 			option_exists |= IPPF_USE_MIN_MTU;
4939 			is_sticky |= IPPF_USE_MIN_MTU;
4940 		}
4941 	}
4942 
4943 	if (!(ignore & IPPF_NEXTHOP)) {
4944 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
4945 			option_exists |= IPPF_NEXTHOP;
4946 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
4947 			option_exists |= IPPF_NEXTHOP;
4948 			is_sticky |= IPPF_NEXTHOP;
4949 		}
4950 	}
4951 
4952 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
4953 		option_exists |= IPPF_HOPLIMIT;
4954 	/* IPV6_HOPLIMIT can never be sticky */
4955 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
4956 
4957 	if (!(ignore & IPPF_UNICAST_HOPS) &&
4958 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
4959 		option_exists |= IPPF_UNICAST_HOPS;
4960 		is_sticky |= IPPF_UNICAST_HOPS;
4961 	}
4962 
4963 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
4964 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
4965 		option_exists |= IPPF_MULTICAST_HOPS;
4966 		is_sticky |= IPPF_MULTICAST_HOPS;
4967 	}
4968 
4969 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
4970 		/* This is a sticky socket option only */
4971 		option_exists |= IPPF_NO_CKSUM;
4972 		is_sticky |= IPPF_NO_CKSUM;
4973 	}
4974 
4975 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
4976 		/* This is a sticky socket option only */
4977 		option_exists |= IPPF_RAW_CKSUM;
4978 		is_sticky |= IPPF_RAW_CKSUM;
4979 	}
4980 
4981 	if (!(ignore & IPPF_TCLASS)) {
4982 		if (ipp->ipp_fields & IPPF_TCLASS) {
4983 			option_exists |= IPPF_TCLASS;
4984 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
4985 			option_exists |= IPPF_TCLASS;
4986 			is_sticky |= IPPF_TCLASS;
4987 		}
4988 	}
4989 
4990 no_options:
4991 
4992 	/*
4993 	 * If any options carried in the ip6i_t were specified, we
4994 	 * need to account for the ip6i_t in the data we'll be sending
4995 	 * down.
4996 	 */
4997 	if (option_exists & IPPF_HAS_IP6I)
4998 		ip_hdr_len += sizeof (ip6i_t);
4999 
5000 	/* check/fix buffer config, setup pointers into it */
5001 	ip6h = (ip6_t *)&mp->b_rptr[-ip_hdr_len];
5002 	if ((mp->b_datap->db_ref != 1) ||
5003 	    ((unsigned char *)ip6h < mp->b_datap->db_base) ||
5004 	    !OK_32PTR(ip6h)) {
5005 		mblk_t	*mp1;
5006 
5007 		/* Try to get everything in a single mblk next time */
5008 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
5009 			icmp->icmp_max_hdr_len = ip_hdr_len;
5010 
5011 			(void) proto_set_tx_wroff(q == NULL ? NULL:RD(q), connp,
5012 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
5013 		}
5014 		mp1 = allocb(ip_hdr_len + is->is_wroff_extra, BPRI_LO);
5015 		if (!mp1) {
5016 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5017 			return (ENOMEM);
5018 		}
5019 		mp1->b_cont = mp;
5020 		mp1->b_wptr = mp1->b_datap->db_lim;
5021 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
5022 		mp = mp1;
5023 	}
5024 	mp->b_rptr = (unsigned char *)ip6h;
5025 	ip6i = (ip6i_t *)ip6h;
5026 
5027 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
5028 	if (option_exists & IPPF_HAS_IP6I) {
5029 		ip6h = (ip6_t *)&ip6i[1];
5030 		ip6i->ip6i_flags = 0;
5031 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5032 
5033 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
5034 		if (option_exists & IPPF_SCOPE_ID) {
5035 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5036 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
5037 		} else if (option_exists & IPPF_IFINDEX) {
5038 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
5039 			ASSERT(tipp->ipp_ifindex != 0);
5040 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5041 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
5042 		}
5043 
5044 		if (option_exists & IPPF_RAW_CKSUM) {
5045 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
5046 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
5047 		}
5048 
5049 		if (option_exists & IPPF_NO_CKSUM) {
5050 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
5051 		}
5052 
5053 		if (option_exists & IPPF_ADDR) {
5054 			/*
5055 			 * Enable per-packet source address verification if
5056 			 * IPV6_PKTINFO specified the source address.
5057 			 * ip6_src is set in the transport's _wput function.
5058 			 */
5059 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
5060 		}
5061 
5062 		if (option_exists & IPPF_DONTFRAG) {
5063 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
5064 		}
5065 
5066 		if (option_exists & IPPF_USE_MIN_MTU) {
5067 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
5068 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
5069 		}
5070 
5071 		if (option_exists & IPPF_NEXTHOP) {
5072 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
5073 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
5074 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
5075 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
5076 		}
5077 
5078 		/*
5079 		 * tell IP this is an ip6i_t private header
5080 		 */
5081 		ip6i->ip6i_nxt = IPPROTO_RAW;
5082 	}
5083 
5084 	/* Initialize IPv6 header */
5085 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5086 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
5087 
5088 	/* Set the hoplimit of the outgoing packet. */
5089 	if (option_exists & IPPF_HOPLIMIT) {
5090 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
5091 		ip6h->ip6_hops = ipp->ipp_hoplimit;
5092 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5093 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
5094 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
5095 		if (option_exists & IPPF_MULTICAST_HOPS)
5096 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5097 	} else {
5098 		ip6h->ip6_hops = icmp->icmp_ttl;
5099 		if (option_exists & IPPF_UNICAST_HOPS)
5100 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5101 	}
5102 
5103 	if (option_exists & IPPF_ADDR) {
5104 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
5105 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
5106 		ip6h->ip6_src = tipp->ipp_addr;
5107 	} else {
5108 		/*
5109 		 * The source address was not set using IPV6_PKTINFO.
5110 		 * First look at the bound source.
5111 		 * If unspecified fallback to __sin6_src_id.
5112 		 */
5113 		ip6h->ip6_src = icmp->icmp_v6src;
5114 		if (sin6->__sin6_src_id != 0 &&
5115 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
5116 			ip_srcid_find_id(sin6->__sin6_src_id,
5117 			    &ip6h->ip6_src, icmp->icmp_zoneid,
5118 			    is->is_netstack);
5119 		}
5120 	}
5121 
5122 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
5123 	cp = (uint8_t *)&ip6h[1];
5124 
5125 	/*
5126 	 * Here's where we have to start stringing together
5127 	 * any extension headers in the right order:
5128 	 * Hop-by-hop, destination, routing, and final destination opts.
5129 	 */
5130 	if (option_exists & IPPF_HOPOPTS) {
5131 		/* Hop-by-hop options */
5132 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
5133 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
5134 
5135 		*nxthdr_ptr = IPPROTO_HOPOPTS;
5136 		nxthdr_ptr = &hbh->ip6h_nxt;
5137 
5138 		bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
5139 		cp += tipp->ipp_hopoptslen;
5140 	}
5141 	/*
5142 	 * En-route destination options
5143 	 * Only do them if there's a routing header as well
5144 	 */
5145 	if (option_exists & IPPF_RTDSTOPTS) {
5146 		ip6_dest_t *dst = (ip6_dest_t *)cp;
5147 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
5148 
5149 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5150 		nxthdr_ptr = &dst->ip6d_nxt;
5151 
5152 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
5153 		cp += tipp->ipp_rtdstoptslen;
5154 	}
5155 	/*
5156 	 * Routing header next
5157 	 */
5158 	if (option_exists & IPPF_RTHDR) {
5159 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
5160 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
5161 
5162 		*nxthdr_ptr = IPPROTO_ROUTING;
5163 		nxthdr_ptr = &rt->ip6r_nxt;
5164 
5165 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
5166 		cp += tipp->ipp_rthdrlen;
5167 	}
5168 	/*
5169 	 * Do ultimate destination options
5170 	 */
5171 	if (option_exists & IPPF_DSTOPTS) {
5172 		ip6_dest_t *dest = (ip6_dest_t *)cp;
5173 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
5174 
5175 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5176 		nxthdr_ptr = &dest->ip6d_nxt;
5177 
5178 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
5179 		cp += tipp->ipp_dstoptslen;
5180 	}
5181 
5182 	/*
5183 	 * Now set the last header pointer to the proto passed in
5184 	 */
5185 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
5186 	*nxthdr_ptr = icmp->icmp_proto;
5187 
5188 	/*
5189 	 * Copy in the destination address
5190 	 */
5191 	ip6h->ip6_dst = ip6_dst;
5192 
5193 	ip6h->ip6_vcf =
5194 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
5195 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
5196 
5197 	if (option_exists & IPPF_TCLASS) {
5198 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
5199 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
5200 		    tipp->ipp_tclass);
5201 	}
5202 	if (option_exists & IPPF_RTHDR) {
5203 		ip6_rthdr_t	*rth;
5204 
5205 		/*
5206 		 * Perform any processing needed for source routing.
5207 		 * We know that all extension headers will be in the same mblk
5208 		 * as the IPv6 header.
5209 		 */
5210 		rth = ip_find_rthdr_v6(ip6h, mp->b_wptr);
5211 		if (rth != NULL && rth->ip6r_segleft != 0) {
5212 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
5213 				/*
5214 				 * Drop packet - only support Type 0 routing.
5215 				 * Notify the application as well.
5216 				 */
5217 				BUMP_MIB(&is->is_rawip_mib,
5218 				    rawipOutErrors);
5219 				return (EPROTO);
5220 			}
5221 			/*
5222 			 * rth->ip6r_len is twice the number of
5223 			 * addresses in the header
5224 			 */
5225 			if (rth->ip6r_len & 0x1) {
5226 				BUMP_MIB(&is->is_rawip_mib,
5227 				    rawipOutErrors);
5228 				return (EPROTO);
5229 			}
5230 			/*
5231 			 * Shuffle the routing header and ip6_dst
5232 			 * addresses, and get the checksum difference
5233 			 * between the first hop (in ip6_dst) and
5234 			 * the destination (in the last routing hdr entry).
5235 			 */
5236 			csum = ip_massage_options_v6(ip6h, rth,
5237 			    is->is_netstack);
5238 			/*
5239 			 * Verify that the first hop isn't a mapped address.
5240 			 * Routers along the path need to do this verification
5241 			 * for subsequent hops.
5242 			 */
5243 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
5244 				BUMP_MIB(&is->is_rawip_mib,
5245 				    rawipOutErrors);
5246 				return (EADDRNOTAVAIL);
5247 			}
5248 		}
5249 	}
5250 
5251 	ip_len = mp->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
5252 	if (mp->b_cont != NULL)
5253 		ip_len += msgdsize(mp->b_cont);
5254 
5255 	/*
5256 	 * Set the length into the IP header.
5257 	 * If the length is greater than the maximum allowed by IP,
5258 	 * then free the message and return. Do not try and send it
5259 	 * as this can cause problems in layers below.
5260 	 */
5261 	if (ip_len > IP_MAXPACKET) {
5262 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5263 		return (EMSGSIZE);
5264 	}
5265 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
5266 		uint_t	cksum_off;	/* From ip6i == mp->b_rptr */
5267 		uint16_t *cksum_ptr;
5268 		uint_t	ext_hdrs_len;
5269 
5270 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
5271 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
5272 		    icmp->icmp_checksum_off == 2);
5273 
5274 		/*
5275 		 * We make it easy for IP to include our pseudo header
5276 		 * by putting our length in uh_checksum, modified (if
5277 		 * we have a routing header) by the checksum difference
5278 		 * between the ultimate destination and first hop addresses.
5279 		 * Note: ICMPv6 must always checksum the packet.
5280 		 */
5281 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
5282 		if (cksum_off + sizeof (uint16_t) > mp->b_wptr - mp->b_rptr) {
5283 			if (!pullupmsg(mp, cksum_off + sizeof (uint16_t))) {
5284 				BUMP_MIB(&is->is_rawip_mib,
5285 				    rawipOutErrors);
5286 				freemsg(mp);
5287 				return (0);
5288 			}
5289 			ip6i = (ip6i_t *)mp->b_rptr;
5290 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
5291 				ip6h = (ip6_t *)&ip6i[1];
5292 			else
5293 				ip6h = (ip6_t *)ip6i;
5294 		}
5295 		/* Add payload length to checksum */
5296 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
5297 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
5298 		csum += htons(ip_len - ext_hdrs_len);
5299 
5300 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
5301 		csum = (csum & 0xFFFF) + (csum >> 16);
5302 		*cksum_ptr = (uint16_t)csum;
5303 	}
5304 
5305 #ifdef _LITTLE_ENDIAN
5306 	ip_len = htons(ip_len);
5307 #endif
5308 	ip6h->ip6_plen = (uint16_t)ip_len;
5309 
5310 	/* We're done. Pass the packet to IP */
5311 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
5312 	ip_output_v6(icmp->icmp_connp, mp, q, IP_WPUT);
5313 	return (0);
5314 }
5315 
5316 static void
5317 icmp_wput_other(queue_t *q, mblk_t *mp)
5318 {
5319 	uchar_t	*rptr = mp->b_rptr;
5320 	struct iocblk *iocp;
5321 #define	tudr ((struct T_unitdata_req *)rptr)
5322 	conn_t	*connp = Q_TO_CONN(q);
5323 	icmp_t	*icmp = connp->conn_icmp;
5324 	icmp_stack_t *is = icmp->icmp_is;
5325 	cred_t *cr;
5326 
5327 	cr = DB_CREDDEF(mp, connp->conn_cred);
5328 
5329 	switch (mp->b_datap->db_type) {
5330 	case M_PROTO:
5331 	case M_PCPROTO:
5332 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5333 			/*
5334 			 * If the message does not contain a PRIM_type,
5335 			 * throw it away.
5336 			 */
5337 			freemsg(mp);
5338 			return;
5339 		}
5340 		switch (((union T_primitives *)rptr)->type) {
5341 		case T_ADDR_REQ:
5342 			icmp_addr_req(q, mp);
5343 			return;
5344 		case O_T_BIND_REQ:
5345 		case T_BIND_REQ:
5346 			icmp_tpi_bind(q, mp);
5347 			return;
5348 		case T_CONN_REQ:
5349 			icmp_tpi_connect(q, mp);
5350 			return;
5351 		case T_CAPABILITY_REQ:
5352 			icmp_capability_req(q, mp);
5353 			return;
5354 		case T_INFO_REQ:
5355 			icmp_info_req(q, mp);
5356 			return;
5357 		case T_UNITDATA_REQ:
5358 			/*
5359 			 * If a T_UNITDATA_REQ gets here, the address must
5360 			 * be bad.  Valid T_UNITDATA_REQs are found above
5361 			 * and break to below this switch.
5362 			 */
5363 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
5364 			return;
5365 		case T_UNBIND_REQ:
5366 			icmp_tpi_unbind(q, mp);
5367 			return;
5368 
5369 		case T_SVR4_OPTMGMT_REQ:
5370 			if (!snmpcom_req(q, mp, icmp_snmp_set, ip_snmp_get,
5371 			    cr)) {
5372 				/* Only IP can return anything meaningful */
5373 				(void) svr4_optcom_req(q, mp, cr,
5374 				    &icmp_opt_obj, B_TRUE);
5375 			}
5376 			return;
5377 
5378 		case T_OPTMGMT_REQ:
5379 			/* Only IP can return anything meaningful */
5380 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj, B_TRUE);
5381 			return;
5382 
5383 		case T_DISCON_REQ:
5384 			icmp_tpi_disconnect(q, mp);
5385 			return;
5386 
5387 		/* The following TPI message is not supported by icmp. */
5388 		case O_T_CONN_RES:
5389 		case T_CONN_RES:
5390 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
5391 			return;
5392 
5393 		/* The following 3 TPI requests are illegal for icmp. */
5394 		case T_DATA_REQ:
5395 		case T_EXDATA_REQ:
5396 		case T_ORDREL_REQ:
5397 			freemsg(mp);
5398 			(void) putctl1(RD(q), M_ERROR, EPROTO);
5399 			return;
5400 		default:
5401 			break;
5402 		}
5403 		break;
5404 	case M_IOCTL:
5405 		iocp = (struct iocblk *)mp->b_rptr;
5406 		switch (iocp->ioc_cmd) {
5407 		case TI_GETPEERNAME:
5408 			if (icmp->icmp_state != TS_DATA_XFER) {
5409 				/*
5410 				 * If a default destination address has not
5411 				 * been associated with the stream, then we
5412 				 * don't know the peer's name.
5413 				 */
5414 				iocp->ioc_error = ENOTCONN;
5415 		err_ret:;
5416 				iocp->ioc_count = 0;
5417 				mp->b_datap->db_type = M_IOCACK;
5418 				qreply(q, mp);
5419 				return;
5420 			}
5421 			/* FALLTHRU */
5422 		case TI_GETMYNAME:
5423 			/*
5424 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
5425 			 * need to copyin the user's strbuf structure.
5426 			 * Processing will continue in the M_IOCDATA case
5427 			 * below.
5428 			 */
5429 			mi_copyin(q, mp, NULL,
5430 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
5431 			return;
5432 		case ND_SET:
5433 			/* nd_getset performs the necessary error checking */
5434 		case ND_GET:
5435 			if (nd_getset(q, is->is_nd, mp)) {
5436 				qreply(q, mp);
5437 				return;
5438 			}
5439 			break;
5440 		case _SIOCSOCKFALLBACK:
5441 			/*
5442 			 * socket is falling back to be a
5443 			 * streams socket. Nothing  to do
5444 			 */
5445 			iocp->ioc_count = 0;
5446 			iocp->ioc_rval = 0;
5447 			qreply(q, mp);
5448 			return;
5449 		default:
5450 			break;
5451 		}
5452 		break;
5453 	case M_IOCDATA:
5454 		icmp_wput_iocdata(q, mp);
5455 		return;
5456 	default:
5457 		break;
5458 	}
5459 	ip_wput(q, mp);
5460 }
5461 
5462 /*
5463  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
5464  * messages.
5465  */
5466 static void
5467 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
5468 {
5469 	mblk_t	*mp1;
5470 	STRUCT_HANDLE(strbuf, sb);
5471 	icmp_t	*icmp;
5472 	uint_t	addrlen;
5473 	uint_t	error;
5474 
5475 	/* Make sure it is one of ours. */
5476 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5477 	case TI_GETMYNAME:
5478 	case TI_GETPEERNAME:
5479 		break;
5480 	default:
5481 		icmp = Q_TO_ICMP(q);
5482 		ip_output(icmp->icmp_connp, mp, q, IP_WPUT);
5483 		return;
5484 	}
5485 	switch (mi_copy_state(q, mp, &mp1)) {
5486 	case -1:
5487 		return;
5488 	case MI_COPY_CASE(MI_COPY_IN, 1):
5489 		break;
5490 	case MI_COPY_CASE(MI_COPY_OUT, 1):
5491 		/*
5492 		 * The address has been copied out, so now
5493 		 * copyout the strbuf.
5494 		 */
5495 		mi_copyout(q, mp);
5496 		return;
5497 	case MI_COPY_CASE(MI_COPY_OUT, 2):
5498 		/*
5499 		 * The address and strbuf have been copied out.
5500 		 * We're done, so just acknowledge the original
5501 		 * M_IOCTL.
5502 		 */
5503 		mi_copy_done(q, mp, 0);
5504 		return;
5505 	default:
5506 		/*
5507 		 * Something strange has happened, so acknowledge
5508 		 * the original M_IOCTL with an EPROTO error.
5509 		 */
5510 		mi_copy_done(q, mp, EPROTO);
5511 		return;
5512 	}
5513 	/*
5514 	 * Now we have the strbuf structure for TI_GETMYNAME
5515 	 * and TI_GETPEERNAME.  Next we copyout the requested
5516 	 * address and then we'll copyout the strbuf.
5517 	 */
5518 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5519 	    (void *)mp1->b_rptr);
5520 	icmp = Q_TO_ICMP(q);
5521 	if (icmp->icmp_family == AF_INET)
5522 		addrlen = sizeof (sin_t);
5523 	else
5524 		addrlen = sizeof (sin6_t);
5525 
5526 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5527 		mi_copy_done(q, mp, EINVAL);
5528 		return;
5529 	}
5530 
5531 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5532 
5533 	if (mp1 == NULL)
5534 		return;
5535 
5536 	rw_enter(&icmp->icmp_rwlock, RW_READER);
5537 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5538 	case TI_GETMYNAME:
5539 		error = rawip_do_getsockname(icmp, (void *)mp1->b_rptr,
5540 		    &addrlen);
5541 		break;
5542 	case TI_GETPEERNAME:
5543 		error = rawip_do_getpeername(icmp, (void *)mp1->b_rptr,
5544 		    &addrlen);
5545 		break;
5546 	}
5547 	rw_exit(&icmp->icmp_rwlock);
5548 
5549 	if (error != 0) {
5550 		mi_copy_done(q, mp, error);
5551 	} else {
5552 		mp1->b_wptr += addrlen;
5553 		STRUCT_FSET(sb, len, addrlen);
5554 
5555 		/* Copy out the address */
5556 		mi_copyout(q, mp);
5557 	}
5558 }
5559 
5560 static int
5561 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5562     void *thisdg_attrs)
5563 {
5564 	conn_t	*connp = Q_TO_CONN(q);
5565 	struct T_unitdata_req *udreqp;
5566 	int is_absreq_failure;
5567 	cred_t *cr;
5568 
5569 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5570 	*errorp = 0;
5571 
5572 	cr = DB_CREDDEF(mp, connp->conn_cred);
5573 
5574 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5575 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5576 	    thisdg_attrs, &is_absreq_failure);
5577 
5578 	if (*errorp != 0) {
5579 		/*
5580 		 * Note: No special action needed in this
5581 		 * module for "is_absreq_failure"
5582 		 */
5583 		return (-1);		/* failure */
5584 	}
5585 	ASSERT(is_absreq_failure == 0);
5586 	return (0);	/* success */
5587 }
5588 
5589 void
5590 icmp_ddi_g_init(void)
5591 {
5592 	icmp_max_optsize = optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5593 	    icmp_opt_obj.odb_opt_arr_cnt);
5594 
5595 	/*
5596 	 * We want to be informed each time a stack is created or
5597 	 * destroyed in the kernel, so we can maintain the
5598 	 * set of icmp_stack_t's.
5599 	 */
5600 	netstack_register(NS_ICMP, rawip_stack_init, NULL, rawip_stack_fini);
5601 }
5602 
5603 void
5604 icmp_ddi_g_destroy(void)
5605 {
5606 	netstack_unregister(NS_ICMP);
5607 }
5608 
5609 #define	INET_NAME	"ip"
5610 
5611 /*
5612  * Initialize the ICMP stack instance.
5613  */
5614 static void *
5615 rawip_stack_init(netstackid_t stackid, netstack_t *ns)
5616 {
5617 	icmp_stack_t	*is;
5618 	icmpparam_t	*pa;
5619 	int		error = 0;
5620 	major_t		major;
5621 
5622 	is = (icmp_stack_t *)kmem_zalloc(sizeof (*is), KM_SLEEP);
5623 	is->is_netstack = ns;
5624 
5625 	pa = (icmpparam_t *)kmem_alloc(sizeof (icmp_param_arr), KM_SLEEP);
5626 	is->is_param_arr = pa;
5627 	bcopy(icmp_param_arr, is->is_param_arr, sizeof (icmp_param_arr));
5628 
5629 	(void) icmp_param_register(&is->is_nd,
5630 	    is->is_param_arr, A_CNT(icmp_param_arr));
5631 	is->is_ksp = rawip_kstat_init(stackid);
5632 
5633 	major = mod_name_to_major(INET_NAME);
5634 	error = ldi_ident_from_major(major, &is->is_ldi_ident);
5635 	ASSERT(error == 0);
5636 	return (is);
5637 }
5638 
5639 /*
5640  * Free the ICMP stack instance.
5641  */
5642 static void
5643 rawip_stack_fini(netstackid_t stackid, void *arg)
5644 {
5645 	icmp_stack_t *is = (icmp_stack_t *)arg;
5646 
5647 	nd_free(&is->is_nd);
5648 	kmem_free(is->is_param_arr, sizeof (icmp_param_arr));
5649 	is->is_param_arr = NULL;
5650 
5651 	rawip_kstat_fini(stackid, is->is_ksp);
5652 	is->is_ksp = NULL;
5653 	ldi_ident_release(is->is_ldi_ident);
5654 	kmem_free(is, sizeof (*is));
5655 }
5656 
5657 static void *
5658 rawip_kstat_init(netstackid_t stackid) {
5659 	kstat_t	*ksp;
5660 
5661 	rawip_named_kstat_t template = {
5662 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5663 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5664 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5665 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5666 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5667 	};
5668 
5669 	ksp = kstat_create_netstack("icmp", 0, "rawip", "mib2",
5670 					KSTAT_TYPE_NAMED,
5671 					NUM_OF_FIELDS(rawip_named_kstat_t),
5672 					0, stackid);
5673 	if (ksp == NULL || ksp->ks_data == NULL)
5674 		return (NULL);
5675 
5676 	bcopy(&template, ksp->ks_data, sizeof (template));
5677 	ksp->ks_update = rawip_kstat_update;
5678 	ksp->ks_private = (void *)(uintptr_t)stackid;
5679 
5680 	kstat_install(ksp);
5681 	return (ksp);
5682 }
5683 
5684 static void
5685 rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
5686 {
5687 	if (ksp != NULL) {
5688 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
5689 		kstat_delete_netstack(ksp, stackid);
5690 	}
5691 }
5692 
5693 static int
5694 rawip_kstat_update(kstat_t *ksp, int rw)
5695 {
5696 	rawip_named_kstat_t *rawipkp;
5697 	netstackid_t	stackid = (netstackid_t)(uintptr_t)ksp->ks_private;
5698 	netstack_t	*ns;
5699 	icmp_stack_t	*is;
5700 
5701 	if ((ksp == NULL) || (ksp->ks_data == NULL))
5702 		return (EIO);
5703 
5704 	if (rw == KSTAT_WRITE)
5705 		return (EACCES);
5706 
5707 	rawipkp = (rawip_named_kstat_t *)ksp->ks_data;
5708 
5709 	ns = netstack_find_by_stackid(stackid);
5710 	if (ns == NULL)
5711 		return (-1);
5712 	is = ns->netstack_icmp;
5713 	if (is == NULL) {
5714 		netstack_rele(ns);
5715 		return (-1);
5716 	}
5717 	rawipkp->inDatagrams.value.ui32 =  is->is_rawip_mib.rawipInDatagrams;
5718 	rawipkp->inCksumErrs.value.ui32 =  is->is_rawip_mib.rawipInCksumErrs;
5719 	rawipkp->inErrors.value.ui32 =	   is->is_rawip_mib.rawipInErrors;
5720 	rawipkp->outDatagrams.value.ui32 = is->is_rawip_mib.rawipOutDatagrams;
5721 	rawipkp->outErrors.value.ui32 =	   is->is_rawip_mib.rawipOutErrors;
5722 	netstack_rele(ns);
5723 	return (0);
5724 }
5725 
5726 /* ARGSUSED */
5727 int
5728 rawip_accept(sock_lower_handle_t lproto_handle,
5729     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
5730     cred_t *cr)
5731 {
5732 	return (EOPNOTSUPP);
5733 }
5734 
5735 /* ARGSUSED */
5736 int
5737 rawip_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
5738     socklen_t len, cred_t *cr)
5739 {
5740 	conn_t  *connp = (conn_t *)proto_handle;
5741 	int error;
5742 
5743 	/* Binding to a NULL address really means unbind */
5744 	if (sa == NULL)
5745 		error = rawip_do_unbind(connp);
5746 	else
5747 		error = rawip_do_bind(connp, sa, len);
5748 
5749 	if (error < 0) {
5750 		if (error == -TOUTSTATE)
5751 			error = EINVAL;
5752 		else
5753 			error = proto_tlitosyserr(-error);
5754 	}
5755 	return (error);
5756 }
5757 
5758 static int
5759 rawip_implicit_bind(conn_t *connp)
5760 {
5761 	sin6_t sin6addr;
5762 	sin_t *sin;
5763 	sin6_t *sin6;
5764 	socklen_t len;
5765 	int error;
5766 
5767 	if (connp->conn_icmp->icmp_family == AF_INET) {
5768 		len = sizeof (struct sockaddr_in);
5769 		sin = (sin_t *)&sin6addr;
5770 		*sin = sin_null;
5771 		sin->sin_family = AF_INET;
5772 		sin->sin_addr.s_addr = INADDR_ANY;
5773 	} else {
5774 		ASSERT(connp->conn_icmp->icmp_family == AF_INET6);
5775 		len = sizeof (sin6_t);
5776 		sin6 = (sin6_t *)&sin6addr;
5777 		*sin6 = sin6_null;
5778 		sin6->sin6_family = AF_INET6;
5779 		V6_SET_ZERO(sin6->sin6_addr);
5780 	}
5781 
5782 	error = rawip_do_bind(connp, (struct sockaddr *)&sin6addr, len);
5783 
5784 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
5785 }
5786 
5787 static int
5788 rawip_unbind(conn_t *connp)
5789 {
5790 	int error;
5791 
5792 	error = rawip_do_unbind(connp);
5793 	if (error < 0) {
5794 		error = proto_tlitosyserr(-error);
5795 	}
5796 	return (error);
5797 }
5798 
5799 /* ARGSUSED */
5800 int
5801 rawip_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
5802 {
5803 	return (EOPNOTSUPP);
5804 }
5805 
5806 /* ARGSUSED */
5807 int
5808 rawip_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
5809     socklen_t len, sock_connid_t *id, cred_t *cr)
5810 {
5811 	conn_t	*connp = (conn_t *)proto_handle;
5812 	icmp_t *icmp = connp->conn_icmp;
5813 	int	error;
5814 	boolean_t did_bind = B_FALSE;
5815 
5816 	if (sa == NULL) {
5817 		/*
5818 		 * Disconnect
5819 		 * Make sure we are connected
5820 		 */
5821 		if (icmp->icmp_state != TS_DATA_XFER)
5822 			return (EINVAL);
5823 
5824 		error = icmp_disconnect(connp);
5825 		return (error);
5826 	}
5827 
5828 	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
5829 	if (error != 0)
5830 		return (error);
5831 
5832 	/* do an implicit bind if necessary */
5833 	if (icmp->icmp_state == TS_UNBND) {
5834 		error = rawip_implicit_bind(connp);
5835 		/*
5836 		 * We could be racing with an actual bind, in which case
5837 		 * we would see EPROTO. We cross our fingers and try
5838 		 * to connect.
5839 		 */
5840 		if (!(error == 0 || error == EPROTO))
5841 			return (error);
5842 		did_bind = B_TRUE;
5843 	}
5844 
5845 	/*
5846 	 * set SO_DGRAM_ERRIND
5847 	 */
5848 	icmp->icmp_dgram_errind = B_TRUE;
5849 
5850 	error = rawip_do_connect(connp, sa, len);
5851 
5852 	if (error != 0 && did_bind) {
5853 		int unbind_err;
5854 
5855 		unbind_err = rawip_unbind(connp);
5856 		ASSERT(unbind_err == 0);
5857 	}
5858 
5859 	if (error == 0) {
5860 		*id = 0;
5861 		(*connp->conn_upcalls->su_connected)
5862 		    (connp->conn_upper_handle, 0, NULL, -1);
5863 	} else if (error < 0) {
5864 		error = proto_tlitosyserr(-error);
5865 	}
5866 	return (error);
5867 }
5868 
5869 /* ARGSUSED */
5870 void
5871 rawip_fallback(sock_lower_handle_t proto_handle, queue_t *q,
5872     boolean_t direct_sockfs, so_proto_quiesced_cb_t quiesced_cb)
5873 {
5874 	conn_t  *connp = (conn_t *)proto_handle;
5875 	icmp_t	*icmp;
5876 	struct T_capability_ack tca;
5877 	struct sockaddr_in6 laddr, faddr;
5878 	socklen_t laddrlen, faddrlen;
5879 	short opts;
5880 	struct stroptions *stropt;
5881 	mblk_t *stropt_mp;
5882 	int error;
5883 
5884 	icmp = connp->conn_icmp;
5885 
5886 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
5887 
5888 	/*
5889 	 * setup the fallback stream that was allocated
5890 	 */
5891 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
5892 	connp->conn_minor_arena = WR(q)->q_ptr;
5893 
5894 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
5895 
5896 	WR(q)->q_qinfo = &icmpwinit;
5897 
5898 	connp->conn_rq = RD(q);
5899 	connp->conn_wq = WR(q);
5900 
5901 	/* Notify stream head about options before sending up data */
5902 	stropt_mp->b_datap->db_type = M_SETOPTS;
5903 	stropt_mp->b_wptr += sizeof (*stropt);
5904 	stropt = (struct stroptions *)stropt_mp->b_rptr;
5905 	stropt->so_flags = SO_WROFF | SO_HIWAT;
5906 	stropt->so_wroff =
5907 	    (ushort_t)(icmp->icmp_max_hdr_len + icmp->icmp_is->is_wroff_extra);
5908 	stropt->so_hiwat = icmp->icmp_recv_hiwat;
5909 	putnext(RD(q), stropt_mp);
5910 
5911 	/*
5912 	 * free helper stream
5913 	 */
5914 	ip_free_helper_stream(connp);
5915 
5916 	/*
5917 	 * Collect the information needed to sync with the sonode
5918 	 */
5919 	icmp_do_capability_ack(icmp, &tca, TC1_INFO);
5920 
5921 	laddrlen = faddrlen = sizeof (sin6_t);
5922 	(void) rawip_getsockname((sock_lower_handle_t)connp,
5923 	    (struct sockaddr *)&laddr, &laddrlen, NULL);
5924 	error = rawip_getpeername((sock_lower_handle_t)connp,
5925 	    (struct sockaddr *)&faddr, &faddrlen, NULL);
5926 	if (error != 0)
5927 		faddrlen = 0;
5928 	opts = 0;
5929 	if (icmp->icmp_dgram_errind)
5930 		opts |= SO_DGRAM_ERRIND;
5931 	if (icmp->icmp_dontroute)
5932 		opts |= SO_DONTROUTE;
5933 
5934 	/*
5935 	 * Once we grab the drain lock, no data will be send up
5936 	 * to the socket. So we notify the socket that the endpoint
5937 	 * is quiescent and it's therefore safe move data from
5938 	 * the socket to the stream head.
5939 	 */
5940 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
5941 	    (struct sockaddr *)&laddr, laddrlen,
5942 	    (struct sockaddr *)&faddr, faddrlen, opts);
5943 
5944 	/*
5945 	 * push up any packets that were queued in icmp_t
5946 	 */
5947 
5948 	mutex_enter(&icmp->icmp_recv_lock);
5949 	while (icmp->icmp_fallback_queue_head != NULL) {
5950 		mblk_t	*mp;
5951 
5952 		mp = icmp->icmp_fallback_queue_head;
5953 		icmp->icmp_fallback_queue_head = mp->b_next;
5954 		mp->b_next = NULL;
5955 		mutex_exit(&icmp->icmp_recv_lock);
5956 		putnext(RD(q), mp);
5957 		mutex_enter(&icmp->icmp_recv_lock);
5958 	}
5959 	icmp->icmp_fallback_queue_tail = icmp->icmp_fallback_queue_head;
5960 	/*
5961 	 * No longer a streams less socket
5962 	 */
5963 	connp->conn_flags &= ~IPCL_NONSTR;
5964 	mutex_exit(&icmp->icmp_recv_lock);
5965 	ASSERT(icmp->icmp_fallback_queue_head == NULL &&
5966 	    icmp->icmp_fallback_queue_tail == NULL);
5967 
5968 	ASSERT(connp->conn_ref >= 1);
5969 }
5970 
5971 /* ARGSUSED */
5972 sock_lower_handle_t
5973 rawip_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
5974     uint_t *smodep, int *errorp, int flags, cred_t *credp)
5975 {
5976 	conn_t *connp;
5977 
5978 	if (type != SOCK_RAW || (family != AF_INET && family != AF_INET6)) {
5979 		*errorp = EPROTONOSUPPORT;
5980 		return (NULL);
5981 	}
5982 
5983 	connp = icmp_open(family, credp, errorp, flags);
5984 	if (connp != NULL) {
5985 		icmp_stack_t *is;
5986 
5987 		is = connp->conn_icmp->icmp_is;
5988 		connp->conn_flags |= IPCL_NONSTR;
5989 
5990 		if (connp->conn_icmp->icmp_family == AF_INET6) {
5991 			/* Build initial header template for transmit */
5992 			rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
5993 			if ((*errorp =
5994 			    icmp_build_hdrs(connp->conn_icmp)) != 0) {
5995 				rw_exit(&connp->conn_icmp->icmp_rwlock);
5996 				ipcl_conn_destroy(connp);
5997 				return (NULL);
5998 			}
5999 			rw_exit(&connp->conn_icmp->icmp_rwlock);
6000 		}
6001 
6002 		connp->conn_icmp->icmp_recv_hiwat = is->is_recv_hiwat;
6003 		connp->conn_icmp->icmp_xmit_hiwat = is->is_xmit_hiwat;
6004 
6005 		if ((*errorp = ip_create_helper_stream(connp,
6006 		    is->is_ldi_ident)) != 0) {
6007 			cmn_err(CE_CONT, "create of IP helper stream failed\n");
6008 			(void) rawip_do_close(connp);
6009 			return (NULL);
6010 		}
6011 
6012 		mutex_enter(&connp->conn_lock);
6013 		connp->conn_state_flags &= ~CONN_INCIPIENT;
6014 		mutex_exit(&connp->conn_lock);
6015 		*sock_downcalls = &sock_rawip_downcalls;
6016 		*smodep = SM_ATOMIC;
6017 	} else {
6018 		ASSERT(*errorp != 0);
6019 	}
6020 
6021 	return ((sock_lower_handle_t)connp);
6022 }
6023 
6024 /* ARGSUSED */
6025 void
6026 rawip_activate(sock_lower_handle_t proto_handle,
6027     sock_upper_handle_t sock_handle, sock_upcalls_t *sock_upcalls, int flags,
6028     cred_t *cr)
6029 {
6030 	conn_t 			*connp = (conn_t *)proto_handle;
6031 	icmp_stack_t 		*is = connp->conn_icmp->icmp_is;
6032 	struct sock_proto_props sopp;
6033 
6034 	connp->conn_upcalls = sock_upcalls;
6035 	connp->conn_upper_handle = sock_handle;
6036 
6037 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT | SOCKOPT_RCVLOWAT |
6038 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
6039 	sopp.sopp_wroff = connp->conn_icmp->icmp_max_hdr_len +
6040 	    is->is_wroff_extra;
6041 	sopp.sopp_rxhiwat = is->is_recv_hiwat;
6042 	sopp.sopp_rxlowat = icmp_mod_info.mi_lowat;
6043 	sopp.sopp_maxblk = INFPSZ;
6044 	sopp.sopp_maxpsz = IP_MAXPACKET;
6045 	sopp.sopp_minpsz = (icmp_mod_info.mi_minpsz == 1) ? 0 :
6046 	    icmp_mod_info.mi_minpsz;
6047 
6048 	(*connp->conn_upcalls->su_set_proto_props)
6049 	    (connp->conn_upper_handle, &sopp);
6050 }
6051 
6052 static int
6053 rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6054 {
6055 	sin_t	*sin = (sin_t *)sa;
6056 	sin6_t	*sin6 = (sin6_t *)sa;
6057 
6058 	ASSERT(icmp != NULL);
6059 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6060 
6061 	switch (icmp->icmp_family) {
6062 	case AF_INET:
6063 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6064 		if (*salenp < sizeof (sin_t))
6065 			return (EINVAL);
6066 
6067 		*salenp = sizeof (sin_t);
6068 		*sin = sin_null;
6069 		sin->sin_family = AF_INET;
6070 		if (icmp->icmp_state == TS_UNBND) {
6071 			break;
6072 		}
6073 
6074 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
6075 		    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6076 			sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_v6src);
6077 		} else {
6078 			/*
6079 			 * INADDR_ANY
6080 			 * icmp_v6src is not set, we might be bound to
6081 			 * broadcast/multicast. Use icmp_bound_v6src as
6082 			 * local address instead (that could
6083 			 * also still be INADDR_ANY)
6084 			 */
6085 			sin->sin_addr.s_addr =
6086 			    V4_PART_OF_V6(icmp->icmp_bound_v6src);
6087 		}
6088 		break;
6089 	case AF_INET6:
6090 
6091 		if (*salenp < sizeof (sin6_t))
6092 			return (EINVAL);
6093 
6094 		*salenp = sizeof (sin6_t);
6095 		*sin6 = sin6_null;
6096 		sin6->sin6_family = AF_INET6;
6097 		if (icmp->icmp_state == TS_UNBND) {
6098 			break;
6099 		}
6100 		if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6101 			sin6->sin6_addr = icmp->icmp_v6src;
6102 		} else {
6103 			/*
6104 			 * UNSPECIFIED
6105 			 * icmp_v6src is not set, we might be bound to
6106 			 * broadcast/multicast. Use icmp_bound_v6src as
6107 			 * local address instead (that could
6108 			 * also still be UNSPECIFIED)
6109 			 */
6110 
6111 			sin6->sin6_addr = icmp->icmp_bound_v6src;
6112 		}
6113 		break;
6114 	}
6115 	return (0);
6116 }
6117 
6118 static int
6119 rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6120 {
6121 	sin_t   *sin = (sin_t *)sa;
6122 	sin6_t  *sin6 = (sin6_t *)sa;
6123 
6124 	ASSERT(icmp != NULL);
6125 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6126 
6127 	if (icmp->icmp_state != TS_DATA_XFER)
6128 		return (ENOTCONN);
6129 
6130 	sa->sa_family = icmp->icmp_family;
6131 	switch (icmp->icmp_family) {
6132 	case AF_INET:
6133 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6134 
6135 		if (*salenp < sizeof (sin_t))
6136 			return (EINVAL);
6137 
6138 		*salenp = sizeof (sin_t);
6139 		*sin = sin_null;
6140 		sin->sin_family = AF_INET;
6141 		sin->sin_addr.s_addr =
6142 		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6143 		break;
6144 	case AF_INET6:
6145 		if (*salenp < sizeof (sin6_t))
6146 			return (EINVAL);
6147 
6148 		*salenp = sizeof (sin6_t);
6149 		*sin6 = sin6_null;
6150 		*sin6 = icmp->icmp_v6dst;
6151 		break;
6152 	}
6153 	return (0);
6154 }
6155 
6156 /* ARGSUSED */
6157 int
6158 rawip_getpeername(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6159     socklen_t *salenp, cred_t *cr)
6160 {
6161 	conn_t  *connp = (conn_t *)proto_handle;
6162 	icmp_t  *icmp = connp->conn_icmp;
6163 	int	error;
6164 
6165 	ASSERT(icmp != NULL);
6166 
6167 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6168 
6169 	error = rawip_do_getpeername(icmp, sa, salenp);
6170 
6171 	rw_exit(&icmp->icmp_rwlock);
6172 
6173 	return (error);
6174 }
6175 
6176 /* ARGSUSED */
6177 int
6178 rawip_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6179     socklen_t *salenp, cred_t *cr)
6180 {
6181 	conn_t  *connp = (conn_t *)proto_handle;
6182 	icmp_t	*icmp = connp->conn_icmp;
6183 	int	error;
6184 
6185 	ASSERT(icmp != NULL);
6186 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6187 
6188 	error = rawip_do_getsockname(icmp, sa, salenp);
6189 
6190 	rw_exit(&icmp->icmp_rwlock);
6191 
6192 	return (error);
6193 }
6194 
6195 int
6196 rawip_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6197     const void *optvalp, socklen_t optlen, cred_t *cr)
6198 {
6199 	conn_t	*connp = (conn_t *)proto_handle;
6200 	icmp_t *icmp = connp->conn_icmp;
6201 	int error;
6202 
6203 	error = proto_opt_check(level, option_name, optlen, NULL,
6204 	    icmp_opt_obj.odb_opt_des_arr,
6205 	    icmp_opt_obj.odb_opt_arr_cnt,
6206 	    icmp_opt_obj.odb_topmost_tpiprovider,
6207 	    B_TRUE, B_FALSE, cr);
6208 
6209 	if (error != 0) {
6210 		/*
6211 		 * option not recognized
6212 		 */
6213 		if (error < 0) {
6214 			error = proto_tlitosyserr(-error);
6215 		}
6216 		return (error);
6217 	}
6218 
6219 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6220 	error = icmp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level,
6221 	    option_name, optlen, (uchar_t *)optvalp, (uint_t *)&optlen,
6222 	    (uchar_t *)optvalp, NULL, cr);
6223 	rw_exit(&icmp->icmp_rwlock);
6224 
6225 	if (error < 0) {
6226 		/*
6227 		 * Pass on to ip
6228 		 */
6229 		error = ip_set_options(connp, level, option_name, optvalp,
6230 		    optlen, cr);
6231 	}
6232 
6233 	ASSERT(error >= 0);
6234 
6235 	return (error);
6236 }
6237 
6238 int
6239 rawip_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6240     void *optvalp, socklen_t *optlen, cred_t *cr)
6241 {
6242 	int		error;
6243 	conn_t		*connp = (conn_t *)proto_handle;
6244 	icmp_t		*icmp = connp->conn_icmp;
6245 	t_uscalar_t	max_optbuf_len;
6246 	void		*optvalp_buf;
6247 	int		len;
6248 
6249 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
6250 	    icmp_opt_obj.odb_opt_des_arr,
6251 	    icmp_opt_obj.odb_opt_arr_cnt,
6252 	    icmp_opt_obj.odb_topmost_tpiprovider,
6253 	    B_FALSE, B_TRUE, cr);
6254 
6255 	if (error != 0) {
6256 		if (error < 0) {
6257 			error = proto_tlitosyserr(-error);
6258 		}
6259 		return (error);
6260 	}
6261 
6262 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
6263 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6264 	len = icmp_opt_get(connp, level, option_name, optvalp_buf);
6265 	rw_exit(&icmp->icmp_rwlock);
6266 
6267 	if (len < 0) {
6268 		/*
6269 		 * Pass on to IP
6270 		 */
6271 		kmem_free(optvalp_buf, max_optbuf_len);
6272 		return (ip_get_options(connp, level, option_name, optvalp,
6273 		    optlen, cr));
6274 	} else {
6275 		/*
6276 		 * update optlen and copy option value
6277 		 */
6278 		t_uscalar_t size = MIN(len, *optlen);
6279 		bcopy(optvalp_buf, optvalp, size);
6280 		bcopy(&size, optlen, sizeof (size));
6281 
6282 		kmem_free(optvalp_buf, max_optbuf_len);
6283 		return (0);
6284 	}
6285 }
6286 
6287 /* ARGSUSED */
6288 int
6289 rawip_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
6290 {
6291 	conn_t	*connp = (conn_t *)proto_handle;
6292 	(void) rawip_do_close(connp);
6293 	return (0);
6294 }
6295 
6296 /* ARGSUSED */
6297 int
6298 rawip_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
6299 {
6300 	conn_t  *connp = (conn_t *)proto_handle;
6301 
6302 	/* shut down the send side */
6303 	if (how != SHUT_RD)
6304 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6305 		    SOCK_OPCTL_SHUT_SEND, 0);
6306 	/* shut down the recv side */
6307 	if (how != SHUT_WR)
6308 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6309 		    SOCK_OPCTL_SHUT_RECV, 0);
6310 	return (0);
6311 }
6312 
6313 void
6314 rawip_clr_flowctrl(sock_lower_handle_t proto_handle)
6315 {
6316 	conn_t  *connp = (conn_t *)proto_handle;
6317 	icmp_t	*icmp = connp->conn_icmp;
6318 
6319 	mutex_enter(&icmp->icmp_recv_lock);
6320 	connp->conn_flow_cntrld = B_FALSE;
6321 	mutex_exit(&icmp->icmp_recv_lock);
6322 }
6323 
6324 int
6325 rawip_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
6326     int mode, int32_t *rvalp, cred_t *cr)
6327 {
6328 	conn_t  	*connp = (conn_t *)proto_handle;
6329 	int		error;
6330 
6331 	switch (cmd) {
6332 	case ND_SET:
6333 	case ND_GET:
6334 	case _SIOCSOCKFALLBACK:
6335 	case TI_GETPEERNAME:
6336 	case TI_GETMYNAME:
6337 #ifdef DEBUG
6338 		cmn_err(CE_CONT, "icmp_ioctl cmd 0x%x on non streams"
6339 		    " socket", cmd);
6340 #endif
6341 		error = EINVAL;
6342 		break;
6343 	default:
6344 		/*
6345 		 * Pass on to IP using helper stream
6346 		 */
6347 		error = ldi_ioctl(connp->conn_helper_info->iphs_handle,
6348 		    cmd, arg, mode, cr, rvalp);
6349 		break;
6350 	}
6351 	return (error);
6352 }
6353 
6354 /* ARGSUSED */
6355 int
6356 rawip_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
6357     cred_t *cr)
6358 {
6359 	conn_t *connp = (conn_t *)proto_handle;
6360 	icmp_t	*icmp = connp->conn_icmp;
6361 	icmp_stack_t *is = icmp->icmp_is;
6362 	int error = 0;
6363 	boolean_t bypass_dgram_errind = B_FALSE;
6364 
6365 	ASSERT(DB_TYPE(mp) == M_DATA);
6366 
6367 	if (is_system_labeled())
6368 		msg_setcredpid(mp, cr, curproc->p_pid);
6369 
6370 	/* do an implicit bind if necessary */
6371 	if (icmp->icmp_state == TS_UNBND) {
6372 		error = rawip_implicit_bind(connp);
6373 		/*
6374 		 * We could be racing with an actual bind, in which case
6375 		 * we would see EPROTO. We cross our fingers and try
6376 		 * to connect.
6377 		 */
6378 		if (!(error == 0 || error == EPROTO)) {
6379 			freemsg(mp);
6380 			return (error);
6381 		}
6382 	}
6383 
6384 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6385 
6386 	if (msg->msg_name != NULL && icmp->icmp_state == TS_DATA_XFER) {
6387 		error = EISCONN;
6388 		goto done_lock;
6389 	}
6390 
6391 	switch (icmp->icmp_family) {
6392 	case AF_INET6: {
6393 		sin6_t	*sin6;
6394 		ip6_pkt_t	ipp_s;	/* For ancillary data options */
6395 		ip6_pkt_t	*ipp = &ipp_s;
6396 
6397 		sin6 = (sin6_t *)msg->msg_name;
6398 		if (sin6 != NULL) {
6399 			error = proto_verify_ip_addr(icmp->icmp_family,
6400 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6401 			if (error != 0) {
6402 				bypass_dgram_errind = B_TRUE;
6403 				goto done_lock;
6404 			}
6405 			if (icmp->icmp_delayed_error != 0) {
6406 				sin6_t  *sin1 = (sin6_t *)msg->msg_name;
6407 				sin6_t  *sin2 = (sin6_t *)
6408 				    &icmp->icmp_delayed_addr;
6409 
6410 				error = icmp->icmp_delayed_error;
6411 				icmp->icmp_delayed_error = 0;
6412 
6413 				/* Compare IP address and port */
6414 
6415 				if (sin1->sin6_port == sin2->sin6_port &&
6416 				    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
6417 				    &sin2->sin6_addr)) {
6418 					goto done_lock;
6419 				}
6420 			}
6421 		} else {
6422 			/*
6423 			 * Use connected address
6424 			 */
6425 			if (icmp->icmp_state != TS_DATA_XFER) {
6426 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6427 				error = EDESTADDRREQ;
6428 				bypass_dgram_errind = B_TRUE;
6429 				goto done_lock;
6430 			}
6431 			sin6 = &icmp->icmp_v6dst;
6432 		}
6433 
6434 		/* No support for mapped addresses on raw sockets */
6435 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
6436 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6437 			error = EADDRNOTAVAIL;
6438 			goto done_lock;
6439 		}
6440 
6441 		ipp->ipp_fields = 0;
6442 		ipp->ipp_sticky_ignored = 0;
6443 
6444 		/*
6445 		 * If options passed in, feed it for verification and handling
6446 		 */
6447 		if (msg->msg_controllen != 0) {
6448 			error = process_auxiliary_options(connp,
6449 			    msg->msg_control, msg->msg_controllen,
6450 			    ipp, &icmp_opt_obj, icmp_opt_set);
6451 			if (error != 0) {
6452 				goto done_lock;
6453 			}
6454 		}
6455 
6456 		rw_exit(&icmp->icmp_rwlock);
6457 
6458 		/*
6459 		 * Destination is a native IPv6 address.
6460 		 * Send out an IPv6 format packet.
6461 		 */
6462 
6463 		error = raw_ip_send_data_v6(connp->conn_wq, connp, mp, sin6,
6464 		    ipp);
6465 	}
6466 		break;
6467 	case AF_INET: {
6468 		sin_t	*sin;
6469 		ip4_pkt_t pktinfo;
6470 		ip4_pkt_t *pktinfop = &pktinfo;
6471 		ipaddr_t	v4dst;
6472 
6473 		sin = (sin_t *)msg->msg_name;
6474 		if (sin != NULL) {
6475 			error = proto_verify_ip_addr(icmp->icmp_family,
6476 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6477 			if (error != 0) {
6478 				bypass_dgram_errind = B_TRUE;
6479 				goto done_lock;
6480 			}
6481 			v4dst = sin->sin_addr.s_addr;
6482 			if (icmp->icmp_delayed_error != 0) {
6483 				sin_t *sin1 = (sin_t *)msg->msg_name;
6484 				sin_t *sin2 = (sin_t *)&icmp->icmp_delayed_addr;
6485 
6486 				error = icmp->icmp_delayed_error;
6487 				icmp->icmp_delayed_error = 0;
6488 
6489 				/* Compare IP address and port */
6490 				if (sin1->sin_port == sin2->sin_port &&
6491 				    sin1->sin_addr.s_addr ==
6492 				    sin2->sin_addr.s_addr) {
6493 					goto done_lock;
6494 				}
6495 
6496 			}
6497 		} else {
6498 			/*
6499 			 * Use connected address
6500 			 */
6501 			if (icmp->icmp_state != TS_DATA_XFER) {
6502 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6503 				error = EDESTADDRREQ;
6504 				bypass_dgram_errind = B_TRUE;
6505 				goto done_lock;
6506 			}
6507 			v4dst = V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6508 		}
6509 
6510 
6511 		pktinfop->ip4_ill_index = 0;
6512 		pktinfop->ip4_addr = INADDR_ANY;
6513 
6514 		/*
6515 		 * If options passed in, feed it for verification and handling
6516 		 */
6517 		if (msg->msg_controllen != 0) {
6518 			error = process_auxiliary_options(connp,
6519 			    msg->msg_control, msg->msg_controllen,
6520 			    pktinfop, &icmp_opt_obj, icmp_opt_set);
6521 			if (error != 0) {
6522 				goto done_lock;
6523 			}
6524 		}
6525 		rw_exit(&icmp->icmp_rwlock);
6526 
6527 		error = raw_ip_send_data_v4(connp->conn_wq, connp, mp,
6528 		    v4dst, pktinfop);
6529 		break;
6530 	}
6531 
6532 	default:
6533 		ASSERT(0);
6534 	}
6535 
6536 	goto done;
6537 
6538 done_lock:
6539 	rw_exit(&icmp->icmp_rwlock);
6540 	if (error != 0) {
6541 		ASSERT(mp != NULL);
6542 		freemsg(mp);
6543 	}
6544 done:
6545 	if (bypass_dgram_errind)
6546 		return (error);
6547 	return (icmp->icmp_dgram_errind ? error : 0);
6548 }
6549 
6550 sock_downcalls_t sock_rawip_downcalls = {
6551 	rawip_activate,
6552 	rawip_accept,
6553 	rawip_bind,
6554 	rawip_listen,
6555 	rawip_connect,
6556 	rawip_getpeername,
6557 	rawip_getsockname,
6558 	rawip_getsockopt,
6559 	rawip_setsockopt,
6560 	rawip_send,
6561 	NULL,
6562 	NULL,
6563 	NULL,
6564 	rawip_shutdown,
6565 	rawip_clr_flowctrl,
6566 	rawip_ioctl,
6567 	rawip_close
6568 };
6569