xref: /illumos-gate/usr/src/uts/common/inet/ip/icmp.c (revision c9eab9d4e096bb9b983e9b007577edfa73c32eff)
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 		case SO_RCVTIMEO:
2304 		case SO_SNDTIMEO:
2305 			/*
2306 			 * Pass these two options in order for third part
2307 			 * protocol usage. Here just return directly.
2308 			 */
2309 			return (0);
2310 		/*
2311 		 * Following three not meaningful for icmp
2312 		 * Action is same as "default" so we keep them
2313 		 * in comments.
2314 		 * case SO_LINGER:
2315 		 * case SO_KEEPALIVE:
2316 		 * case SO_OOBINLINE:
2317 		 */
2318 		default:
2319 			*outlenp = 0;
2320 			return (EINVAL);
2321 		}
2322 		break;
2323 	case IPPROTO_IP:
2324 		/*
2325 		 * Only allow IPv4 option processing on IPv4 sockets.
2326 		 */
2327 		if (icmp->icmp_family != AF_INET) {
2328 			*outlenp = 0;
2329 			return (ENOPROTOOPT);
2330 		}
2331 		switch (name) {
2332 		case IP_OPTIONS:
2333 		case T_IP_OPTIONS:
2334 			/* Save options for use by IP. */
2335 			if ((inlen & 0x3) ||
2336 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2337 				*outlenp = 0;
2338 				return (EINVAL);
2339 			}
2340 			if (checkonly)
2341 				break;
2342 
2343 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2344 			    &icmp->icmp_ip_snd_options_len,
2345 			    icmp->icmp_label_len, invalp, inlen)) {
2346 				*outlenp = 0;
2347 				return (ENOMEM);
2348 			}
2349 
2350 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2351 			    icmp->icmp_ip_snd_options_len;
2352 			rw_exit(&icmp->icmp_rwlock);
2353 			(void) proto_set_tx_wroff(connp->conn_rq == NULL ? NULL:
2354 			    RD(connp->conn_rq), connp,
2355 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
2356 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2357 			break;
2358 		case IP_HDRINCL:
2359 			if (!checkonly)
2360 				icmp->icmp_hdrincl = onoff;
2361 			break;
2362 		case IP_TOS:
2363 		case T_IP_TOS:
2364 			if (!checkonly) {
2365 				icmp->icmp_type_of_service = (uint8_t)*i1;
2366 			}
2367 			break;
2368 		case IP_TTL:
2369 			if (!checkonly) {
2370 				icmp->icmp_ttl = (uint8_t)*i1;
2371 			}
2372 			break;
2373 		case IP_MULTICAST_IF:
2374 			/*
2375 			 * TODO should check OPTMGMT reply and undo this if
2376 			 * there is an error.
2377 			 */
2378 			if (!checkonly) {
2379 				icmp->icmp_multicast_if_addr = *i1;
2380 				PASS_OPT_TO_IP(connp);
2381 			}
2382 			break;
2383 		case IP_MULTICAST_TTL:
2384 			if (!checkonly)
2385 				icmp->icmp_multicast_ttl = *invalp;
2386 			break;
2387 		case IP_MULTICAST_LOOP:
2388 			if (!checkonly) {
2389 				connp->conn_multicast_loop =
2390 				    (*invalp == 0) ? 0 : 1;
2391 				PASS_OPT_TO_IP(connp);
2392 			}
2393 			break;
2394 		case IP_BOUND_IF:
2395 			if (!checkonly) {
2396 				icmp->icmp_bound_if = *i1;
2397 				PASS_OPT_TO_IP(connp);
2398 			}
2399 			break;
2400 		case IP_UNSPEC_SRC:
2401 			if (!checkonly) {
2402 				icmp->icmp_unspec_source = onoff;
2403 				PASS_OPT_TO_IP(connp);
2404 			}
2405 			break;
2406 		case IP_BROADCAST_TTL:
2407 			if (!checkonly)
2408 				connp->conn_broadcast_ttl = *invalp;
2409 			break;
2410 		case IP_RECVIF:
2411 			if (!checkonly) {
2412 				icmp->icmp_recvif = onoff;
2413 			}
2414 			/*
2415 			 * pass to ip
2416 			 */
2417 			return (-EINVAL);
2418 		case IP_PKTINFO: {
2419 			/*
2420 			 * This also handles IP_RECVPKTINFO.
2421 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2422 			 * Differentiation is based on the size of the argument
2423 			 * passed in.
2424 			 */
2425 			struct in_pktinfo *pktinfop;
2426 			ip4_pkt_t *attr_pktinfop;
2427 
2428 			if (checkonly)
2429 				break;
2430 
2431 			if (inlen == sizeof (int)) {
2432 				/*
2433 				 * This is IP_RECVPKTINFO option.
2434 				 * Keep a local copy of wether this option is
2435 				 * set or not and pass it down to IP for
2436 				 * processing.
2437 				 */
2438 				icmp->icmp_ip_recvpktinfo = onoff;
2439 				return (-EINVAL);
2440 			}
2441 
2442 
2443 			if (inlen != sizeof (struct in_pktinfo)) {
2444 				return (EINVAL);
2445 			}
2446 
2447 			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
2448 			    == NULL) {
2449 				/*
2450 				 * sticky option is not supported
2451 				 */
2452 				return (EINVAL);
2453 			}
2454 
2455 			pktinfop = (struct in_pktinfo *)invalp;
2456 
2457 			/*
2458 			 * Atleast one of the values should be specified
2459 			 */
2460 			if (pktinfop->ipi_ifindex == 0 &&
2461 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2462 				return (EINVAL);
2463 			}
2464 
2465 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2466 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2467 		}
2468 			break;
2469 		case IP_ADD_MEMBERSHIP:
2470 		case IP_DROP_MEMBERSHIP:
2471 		case IP_BLOCK_SOURCE:
2472 		case IP_UNBLOCK_SOURCE:
2473 		case IP_ADD_SOURCE_MEMBERSHIP:
2474 		case IP_DROP_SOURCE_MEMBERSHIP:
2475 		case MCAST_JOIN_GROUP:
2476 		case MCAST_LEAVE_GROUP:
2477 		case MCAST_BLOCK_SOURCE:
2478 		case MCAST_UNBLOCK_SOURCE:
2479 		case MCAST_JOIN_SOURCE_GROUP:
2480 		case MCAST_LEAVE_SOURCE_GROUP:
2481 		case MRT_INIT:
2482 		case MRT_DONE:
2483 		case MRT_ADD_VIF:
2484 		case MRT_DEL_VIF:
2485 		case MRT_ADD_MFC:
2486 		case MRT_DEL_MFC:
2487 		case MRT_VERSION:
2488 		case MRT_ASSERT:
2489 		case IP_SEC_OPT:
2490 		case IP_NEXTHOP:
2491 			/*
2492 			 * "soft" error (negative)
2493 			 * option not handled at this level
2494 			 * Note: Do not modify *outlenp
2495 			 */
2496 			return (-EINVAL);
2497 		default:
2498 			*outlenp = 0;
2499 			return (EINVAL);
2500 		}
2501 		break;
2502 	case IPPROTO_IPV6: {
2503 		ip6_pkt_t		*ipp;
2504 		boolean_t		sticky;
2505 
2506 		if (icmp->icmp_family != AF_INET6) {
2507 			*outlenp = 0;
2508 			return (ENOPROTOOPT);
2509 		}
2510 		/*
2511 		 * Deal with both sticky options and ancillary data
2512 		 */
2513 		if (thisdg_attrs == NULL) {
2514 			/* sticky options, or none */
2515 			ipp = &icmp->icmp_sticky_ipp;
2516 			sticky = B_TRUE;
2517 		} else {
2518 			/* ancillary data */
2519 			ipp = (ip6_pkt_t *)thisdg_attrs;
2520 			sticky = B_FALSE;
2521 		}
2522 
2523 		switch (name) {
2524 		case IPV6_MULTICAST_IF:
2525 			if (!checkonly) {
2526 				icmp->icmp_multicast_if_index = *i1;
2527 				PASS_OPT_TO_IP(connp);
2528 			}
2529 			break;
2530 		case IPV6_UNICAST_HOPS:
2531 			/* -1 means use default */
2532 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2533 				*outlenp = 0;
2534 				return (EINVAL);
2535 			}
2536 			if (!checkonly) {
2537 				if (*i1 == -1) {
2538 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2539 					    is->is_ipv6_hoplimit;
2540 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2541 					/* Pass modified value to IP. */
2542 					*i1 = ipp->ipp_hoplimit;
2543 				} else {
2544 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2545 					    (uint8_t)*i1;
2546 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2547 				}
2548 				/* Rebuild the header template */
2549 				error = icmp_build_hdrs(icmp);
2550 				if (error != 0) {
2551 					*outlenp = 0;
2552 					return (error);
2553 				}
2554 			}
2555 			break;
2556 		case IPV6_MULTICAST_HOPS:
2557 			/* -1 means use default */
2558 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2559 				*outlenp = 0;
2560 				return (EINVAL);
2561 			}
2562 			if (!checkonly) {
2563 				if (*i1 == -1) {
2564 					icmp->icmp_multicast_ttl =
2565 					    ipp->ipp_multicast_hops =
2566 					    IP_DEFAULT_MULTICAST_TTL;
2567 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2568 					/* Pass modified value to IP. */
2569 					*i1 = icmp->icmp_multicast_ttl;
2570 				} else {
2571 					icmp->icmp_multicast_ttl =
2572 					    ipp->ipp_multicast_hops =
2573 					    (uint8_t)*i1;
2574 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2575 				}
2576 			}
2577 			break;
2578 		case IPV6_MULTICAST_LOOP:
2579 			if (*i1 != 0 && *i1 != 1) {
2580 				*outlenp = 0;
2581 				return (EINVAL);
2582 			}
2583 			if (!checkonly) {
2584 				connp->conn_multicast_loop = *i1;
2585 				PASS_OPT_TO_IP(connp);
2586 			}
2587 			break;
2588 		case IPV6_CHECKSUM:
2589 			/*
2590 			 * Integer offset into the user data of where the
2591 			 * checksum is located.
2592 			 * Offset of -1 disables option.
2593 			 * Does not apply to IPPROTO_ICMPV6.
2594 			 */
2595 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2596 				*outlenp = 0;
2597 				return (EINVAL);
2598 			}
2599 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2600 				/* Negative or not 16 bit aligned offset */
2601 				*outlenp = 0;
2602 				return (EINVAL);
2603 			}
2604 			if (checkonly)
2605 				break;
2606 
2607 			if (*i1 == -1) {
2608 				icmp->icmp_raw_checksum = 0;
2609 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2610 			} else {
2611 				icmp->icmp_raw_checksum = 1;
2612 				icmp->icmp_checksum_off = *i1;
2613 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2614 			}
2615 			/* Rebuild the header template */
2616 			error = icmp_build_hdrs(icmp);
2617 			if (error != 0) {
2618 				*outlenp = 0;
2619 				return (error);
2620 			}
2621 			break;
2622 		case IPV6_JOIN_GROUP:
2623 		case IPV6_LEAVE_GROUP:
2624 		case MCAST_JOIN_GROUP:
2625 		case MCAST_LEAVE_GROUP:
2626 		case MCAST_BLOCK_SOURCE:
2627 		case MCAST_UNBLOCK_SOURCE:
2628 		case MCAST_JOIN_SOURCE_GROUP:
2629 		case MCAST_LEAVE_SOURCE_GROUP:
2630 			/*
2631 			 * "soft" error (negative)
2632 			 * option not handled at this level
2633 			 * Note: Do not modify *outlenp
2634 			 */
2635 			return (-EINVAL);
2636 		case IPV6_BOUND_IF:
2637 			if (!checkonly) {
2638 				icmp->icmp_bound_if = *i1;
2639 				PASS_OPT_TO_IP(connp);
2640 			}
2641 			break;
2642 		case IPV6_UNSPEC_SRC:
2643 			if (!checkonly) {
2644 				icmp->icmp_unspec_source = onoff;
2645 				PASS_OPT_TO_IP(connp);
2646 			}
2647 			break;
2648 		case IPV6_RECVTCLASS:
2649 			if (!checkonly) {
2650 				icmp->icmp_ipv6_recvtclass = onoff;
2651 				PASS_OPT_TO_IP(connp);
2652 			}
2653 			break;
2654 		/*
2655 		 * Set boolean switches for ancillary data delivery
2656 		 */
2657 		case IPV6_RECVPKTINFO:
2658 			if (!checkonly) {
2659 				icmp->icmp_ip_recvpktinfo = onoff;
2660 				PASS_OPT_TO_IP(connp);
2661 			}
2662 			break;
2663 		case IPV6_RECVPATHMTU:
2664 			if (!checkonly) {
2665 				icmp->icmp_ipv6_recvpathmtu = onoff;
2666 				PASS_OPT_TO_IP(connp);
2667 			}
2668 			break;
2669 		case IPV6_RECVHOPLIMIT:
2670 			if (!checkonly) {
2671 				icmp->icmp_ipv6_recvhoplimit = onoff;
2672 				PASS_OPT_TO_IP(connp);
2673 			}
2674 			break;
2675 		case IPV6_RECVHOPOPTS:
2676 			if (!checkonly) {
2677 				icmp->icmp_ipv6_recvhopopts = onoff;
2678 				PASS_OPT_TO_IP(connp);
2679 			}
2680 			break;
2681 		case IPV6_RECVDSTOPTS:
2682 			if (!checkonly) {
2683 				icmp->icmp_ipv6_recvdstopts = onoff;
2684 				PASS_OPT_TO_IP(connp);
2685 			}
2686 			break;
2687 		case _OLD_IPV6_RECVDSTOPTS:
2688 			if (!checkonly)
2689 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2690 			break;
2691 		case IPV6_RECVRTHDRDSTOPTS:
2692 			if (!checkonly) {
2693 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2694 				PASS_OPT_TO_IP(connp);
2695 			}
2696 			break;
2697 		case IPV6_RECVRTHDR:
2698 			if (!checkonly) {
2699 				icmp->icmp_ipv6_recvrthdr = onoff;
2700 				PASS_OPT_TO_IP(connp);
2701 			}
2702 			break;
2703 		/*
2704 		 * Set sticky options or ancillary data.
2705 		 * If sticky options, (re)build any extension headers
2706 		 * that might be needed as a result.
2707 		 */
2708 		case IPV6_PKTINFO:
2709 			/*
2710 			 * The source address and ifindex are verified
2711 			 * in ip_opt_set(). For ancillary data the
2712 			 * source address is checked in ip_wput_v6.
2713 			 */
2714 			if (inlen != 0 && inlen !=
2715 			    sizeof (struct in6_pktinfo)) {
2716 				return (EINVAL);
2717 			}
2718 			if (checkonly)
2719 				break;
2720 
2721 			if (inlen == 0) {
2722 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2723 				ipp->ipp_sticky_ignored |=
2724 				    (IPPF_IFINDEX|IPPF_ADDR);
2725 			} else {
2726 				struct in6_pktinfo *pkti;
2727 
2728 				pkti = (struct in6_pktinfo *)invalp;
2729 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2730 				ipp->ipp_addr = pkti->ipi6_addr;
2731 				if (ipp->ipp_ifindex != 0)
2732 					ipp->ipp_fields |= IPPF_IFINDEX;
2733 				else
2734 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2735 				if (!IN6_IS_ADDR_UNSPECIFIED(
2736 				    &ipp->ipp_addr))
2737 					ipp->ipp_fields |= IPPF_ADDR;
2738 				else
2739 					ipp->ipp_fields &= ~IPPF_ADDR;
2740 			}
2741 			if (sticky) {
2742 				error = icmp_build_hdrs(icmp);
2743 				if (error != 0)
2744 					return (error);
2745 				PASS_OPT_TO_IP(connp);
2746 			}
2747 			break;
2748 		case IPV6_HOPLIMIT:
2749 			/* This option can only be used as ancillary data. */
2750 			if (sticky)
2751 				return (EINVAL);
2752 			if (inlen != 0 && inlen != sizeof (int))
2753 				return (EINVAL);
2754 			if (checkonly)
2755 				break;
2756 
2757 			if (inlen == 0) {
2758 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2759 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2760 			} else {
2761 				if (*i1 > 255 || *i1 < -1)
2762 					return (EINVAL);
2763 				if (*i1 == -1)
2764 					ipp->ipp_hoplimit =
2765 					    is->is_ipv6_hoplimit;
2766 				else
2767 					ipp->ipp_hoplimit = *i1;
2768 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2769 			}
2770 			break;
2771 		case IPV6_TCLASS:
2772 			/*
2773 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2774 			 * and [0, 255] as the actualy traffic class.
2775 			 */
2776 			if (inlen != 0 && inlen != sizeof (int)) {
2777 				return (EINVAL);
2778 			}
2779 			if (checkonly)
2780 				break;
2781 
2782 			if (inlen == 0) {
2783 				ipp->ipp_fields &= ~IPPF_TCLASS;
2784 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2785 			} else {
2786 				if (*i1 >= 256 || *i1 < -1)
2787 					return (EINVAL);
2788 				if (*i1 == -1) {
2789 					ipp->ipp_tclass =
2790 					    IPV6_FLOW_TCLASS(
2791 					    IPV6_DEFAULT_VERS_AND_FLOW);
2792 				} else {
2793 					ipp->ipp_tclass = *i1;
2794 				}
2795 				ipp->ipp_fields |= IPPF_TCLASS;
2796 			}
2797 			if (sticky) {
2798 				error = icmp_build_hdrs(icmp);
2799 				if (error != 0)
2800 					return (error);
2801 			}
2802 			break;
2803 		case IPV6_NEXTHOP:
2804 			/*
2805 			 * IP will verify that the nexthop is reachable
2806 			 * and fail for sticky options.
2807 			 */
2808 			if (inlen != 0 && inlen != sizeof (sin6_t)) {
2809 				return (EINVAL);
2810 			}
2811 			if (checkonly)
2812 				break;
2813 
2814 			if (inlen == 0) {
2815 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2816 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2817 			} else {
2818 				sin6_t *sin6 = (sin6_t *)invalp;
2819 
2820 				if (sin6->sin6_family != AF_INET6) {
2821 					return (EAFNOSUPPORT);
2822 				}
2823 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
2824 					return (EADDRNOTAVAIL);
2825 				}
2826 				ipp->ipp_nexthop = sin6->sin6_addr;
2827 				if (!IN6_IS_ADDR_UNSPECIFIED(
2828 				    &ipp->ipp_nexthop))
2829 					ipp->ipp_fields |= IPPF_NEXTHOP;
2830 				else
2831 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2832 			}
2833 			if (sticky) {
2834 				error = icmp_build_hdrs(icmp);
2835 				if (error != 0)
2836 					return (error);
2837 				PASS_OPT_TO_IP(connp);
2838 			}
2839 			break;
2840 		case IPV6_HOPOPTS: {
2841 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2842 			/*
2843 			 * Sanity checks - minimum size, size a multiple of
2844 			 * eight bytes, and matching size passed in.
2845 			 */
2846 			if (inlen != 0 &&
2847 			    inlen != (8 * (hopts->ip6h_len + 1))) {
2848 				return (EINVAL);
2849 			}
2850 
2851 			if (checkonly)
2852 				break;
2853 			error = optcom_pkt_set(invalp, inlen, sticky,
2854 			    (uchar_t **)&ipp->ipp_hopopts,
2855 			    &ipp->ipp_hopoptslen,
2856 			    sticky ? icmp->icmp_label_len_v6 : 0);
2857 			if (error != 0)
2858 				return (error);
2859 			if (ipp->ipp_hopoptslen == 0) {
2860 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2861 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2862 			} else {
2863 				ipp->ipp_fields |= IPPF_HOPOPTS;
2864 			}
2865 			if (sticky) {
2866 				error = icmp_build_hdrs(icmp);
2867 				if (error != 0)
2868 					return (error);
2869 			}
2870 			break;
2871 		}
2872 		case IPV6_RTHDRDSTOPTS: {
2873 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2874 
2875 			/*
2876 			 * Sanity checks - minimum size, size a multiple of
2877 			 * eight bytes, and matching size passed in.
2878 			 */
2879 			if (inlen != 0 &&
2880 			    inlen != (8 * (dopts->ip6d_len + 1)))
2881 				return (EINVAL);
2882 
2883 			if (checkonly)
2884 				break;
2885 
2886 			if (inlen == 0) {
2887 				if (sticky &&
2888 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2889 					kmem_free(ipp->ipp_rtdstopts,
2890 					    ipp->ipp_rtdstoptslen);
2891 					ipp->ipp_rtdstopts = NULL;
2892 					ipp->ipp_rtdstoptslen = 0;
2893 				}
2894 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2895 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2896 			} else {
2897 				error = optcom_pkt_set(invalp, inlen, sticky,
2898 				    (uchar_t **)&ipp->ipp_rtdstopts,
2899 				    &ipp->ipp_rtdstoptslen, 0);
2900 				if (error != 0)
2901 					return (error);
2902 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2903 			}
2904 			if (sticky) {
2905 				error = icmp_build_hdrs(icmp);
2906 				if (error != 0)
2907 					return (error);
2908 			}
2909 			break;
2910 		}
2911 		case IPV6_DSTOPTS: {
2912 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2913 
2914 			/*
2915 			 * Sanity checks - minimum size, size a multiple of
2916 			 * eight bytes, and matching size passed in.
2917 			 */
2918 			if (inlen != 0 &&
2919 			    inlen != (8 * (dopts->ip6d_len + 1)))
2920 				return (EINVAL);
2921 
2922 			if (checkonly)
2923 				break;
2924 
2925 			if (inlen == 0) {
2926 				if (sticky &&
2927 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2928 					kmem_free(ipp->ipp_dstopts,
2929 					    ipp->ipp_dstoptslen);
2930 					ipp->ipp_dstopts = NULL;
2931 					ipp->ipp_dstoptslen = 0;
2932 				}
2933 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2934 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2935 			} else {
2936 				error = optcom_pkt_set(invalp, inlen, sticky,
2937 				    (uchar_t **)&ipp->ipp_dstopts,
2938 				    &ipp->ipp_dstoptslen, 0);
2939 				if (error != 0)
2940 					return (error);
2941 				ipp->ipp_fields |= IPPF_DSTOPTS;
2942 			}
2943 			if (sticky) {
2944 				error = icmp_build_hdrs(icmp);
2945 				if (error != 0)
2946 					return (error);
2947 			}
2948 			break;
2949 		}
2950 		case IPV6_RTHDR: {
2951 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2952 
2953 			/*
2954 			 * Sanity checks - minimum size, size a multiple of
2955 			 * eight bytes, and matching size passed in.
2956 			 */
2957 			if (inlen != 0 &&
2958 			    inlen != (8 * (rt->ip6r_len + 1)))
2959 				return (EINVAL);
2960 
2961 			if (checkonly)
2962 				break;
2963 
2964 			if (inlen == 0) {
2965 				if (sticky &&
2966 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2967 					kmem_free(ipp->ipp_rthdr,
2968 					    ipp->ipp_rthdrlen);
2969 					ipp->ipp_rthdr = NULL;
2970 					ipp->ipp_rthdrlen = 0;
2971 				}
2972 				ipp->ipp_fields &= ~IPPF_RTHDR;
2973 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2974 			} else {
2975 				error = optcom_pkt_set(invalp, inlen, sticky,
2976 				    (uchar_t **)&ipp->ipp_rthdr,
2977 				    &ipp->ipp_rthdrlen, 0);
2978 				if (error != 0)
2979 					return (error);
2980 				ipp->ipp_fields |= IPPF_RTHDR;
2981 			}
2982 			if (sticky) {
2983 				error = icmp_build_hdrs(icmp);
2984 				if (error != 0)
2985 					return (error);
2986 			}
2987 			break;
2988 		}
2989 
2990 		case IPV6_DONTFRAG:
2991 			if (checkonly)
2992 				break;
2993 
2994 			if (onoff) {
2995 				ipp->ipp_fields |= IPPF_DONTFRAG;
2996 			} else {
2997 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2998 			}
2999 			break;
3000 
3001 		case IPV6_USE_MIN_MTU:
3002 			if (inlen != sizeof (int))
3003 				return (EINVAL);
3004 
3005 			if (*i1 < -1 || *i1 > 1)
3006 				return (EINVAL);
3007 
3008 			if (checkonly)
3009 				break;
3010 
3011 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3012 			ipp->ipp_use_min_mtu = *i1;
3013 			break;
3014 
3015 		/*
3016 		 * This option can't be set.  Its only returned via
3017 		 * getsockopt() or ancillary data.
3018 		 */
3019 		case IPV6_PATHMTU:
3020 			return (EINVAL);
3021 
3022 		case IPV6_SEC_OPT:
3023 		case IPV6_SRC_PREFERENCES:
3024 		case IPV6_V6ONLY:
3025 			/* Handled at IP level */
3026 			return (-EINVAL);
3027 		default:
3028 			*outlenp = 0;
3029 			return (EINVAL);
3030 		}
3031 		break;
3032 	}		/* end IPPROTO_IPV6 */
3033 
3034 	case IPPROTO_ICMPV6:
3035 		/*
3036 		 * Only allow IPv6 option processing on IPv6 sockets.
3037 		 */
3038 		if (icmp->icmp_family != AF_INET6) {
3039 			*outlenp = 0;
3040 			return (ENOPROTOOPT);
3041 		}
3042 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
3043 			*outlenp = 0;
3044 			return (ENOPROTOOPT);
3045 		}
3046 		switch (name) {
3047 		case ICMP6_FILTER:
3048 			if (!checkonly) {
3049 				if ((inlen != 0) &&
3050 				    (inlen != sizeof (icmp6_filter_t)))
3051 					return (EINVAL);
3052 
3053 				if (inlen == 0) {
3054 					if (icmp->icmp_filter != NULL) {
3055 						kmem_free(icmp->icmp_filter,
3056 						    sizeof (icmp6_filter_t));
3057 						icmp->icmp_filter = NULL;
3058 					}
3059 				} else {
3060 					if (icmp->icmp_filter == NULL) {
3061 						icmp->icmp_filter = kmem_alloc(
3062 						    sizeof (icmp6_filter_t),
3063 						    KM_NOSLEEP);
3064 						if (icmp->icmp_filter == NULL) {
3065 							*outlenp = 0;
3066 							return (ENOBUFS);
3067 						}
3068 					}
3069 					(void) bcopy(invalp, icmp->icmp_filter,
3070 					    inlen);
3071 				}
3072 			}
3073 			break;
3074 
3075 		default:
3076 			*outlenp = 0;
3077 			return (EINVAL);
3078 		}
3079 		break;
3080 	default:
3081 		*outlenp = 0;
3082 		return (EINVAL);
3083 	}
3084 	/*
3085 	 * Common case of OK return with outval same as inval.
3086 	 */
3087 	if (invalp != outvalp) {
3088 		/* don't trust bcopy for identical src/dst */
3089 		(void) bcopy(invalp, outvalp, inlen);
3090 	}
3091 	*outlenp = inlen;
3092 	return (0);
3093 }
3094 
3095 /* This routine sets socket options. */
3096 /* ARGSUSED */
3097 int
3098 icmp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3099     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3100     void *thisdg_attrs, cred_t *cr)
3101 {
3102 	boolean_t checkonly;
3103 	int	error;
3104 
3105 	error = 0;
3106 	switch (optset_context) {
3107 	case SETFN_OPTCOM_CHECKONLY:
3108 		checkonly = B_TRUE;
3109 		/*
3110 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3111 		 * inlen != 0 implies value supplied and
3112 		 * 	we have to "pretend" to set it.
3113 		 * inlen == 0 implies that there is no
3114 		 * 	value part in T_CHECK request and just validation
3115 		 * done elsewhere should be enough, we just return here.
3116 		 */
3117 		if (inlen == 0) {
3118 			*outlenp = 0;
3119 			error = 0;
3120 			goto done;
3121 		}
3122 		break;
3123 	case SETFN_OPTCOM_NEGOTIATE:
3124 		checkonly = B_FALSE;
3125 		break;
3126 	case SETFN_UD_NEGOTIATE:
3127 	case SETFN_CONN_NEGOTIATE:
3128 		checkonly = B_FALSE;
3129 		/*
3130 		 * Negotiating local and "association-related" options
3131 		 * through T_UNITDATA_REQ.
3132 		 *
3133 		 * Following routine can filter out ones we do not
3134 		 * want to be "set" this way.
3135 		 */
3136 		if (!icmp_opt_allow_udr_set(level, name)) {
3137 			*outlenp = 0;
3138 			error = EINVAL;
3139 			goto done;
3140 		}
3141 		break;
3142 	default:
3143 		/*
3144 		 * We should never get here
3145 		 */
3146 		*outlenp = 0;
3147 		error = EINVAL;
3148 		goto done;
3149 	}
3150 
3151 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3152 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3153 	error = icmp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3154 	    outvalp, cr, thisdg_attrs, checkonly);
3155 
3156 done:
3157 	return (error);
3158 }
3159 
3160 /* This routine sets socket options. */
3161 /* ARGSUSED */
3162 int
3163 icmp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3164     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3165     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3166 {
3167 	conn_t	*connp =  Q_TO_CONN(q);
3168 	icmp_t	*icmp;
3169 	int error;
3170 
3171 	icmp = connp->conn_icmp;
3172 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3173 	error = icmp_opt_set(connp, optset_context, level, name, inlen, invalp,
3174 	    outlenp, outvalp, thisdg_attrs, cr);
3175 	rw_exit(&icmp->icmp_rwlock);
3176 	return (error);
3177 }
3178 
3179 /*
3180  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
3181  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
3182  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
3183  * headers.
3184  * Returns failure if can't allocate memory.
3185  */
3186 static int
3187 icmp_build_hdrs(icmp_t *icmp)
3188 {
3189 	icmp_stack_t *is = icmp->icmp_is;
3190 	uchar_t	*hdrs;
3191 	uint_t	hdrs_len;
3192 	ip6_t	*ip6h;
3193 	ip6i_t	*ip6i;
3194 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
3195 
3196 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
3197 	hdrs_len = ip_total_hdrs_len_v6(ipp);
3198 	ASSERT(hdrs_len != 0);
3199 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
3200 		/* Need to reallocate */
3201 		if (hdrs_len != 0) {
3202 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3203 			if (hdrs == NULL)
3204 				return (ENOMEM);
3205 		} else {
3206 			hdrs = NULL;
3207 		}
3208 		if (icmp->icmp_sticky_hdrs_len != 0) {
3209 			kmem_free(icmp->icmp_sticky_hdrs,
3210 			    icmp->icmp_sticky_hdrs_len);
3211 		}
3212 		icmp->icmp_sticky_hdrs = hdrs;
3213 		icmp->icmp_sticky_hdrs_len = hdrs_len;
3214 	}
3215 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
3216 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
3217 
3218 	/* Set header fields not in ipp */
3219 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3220 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
3221 		ip6h = (ip6_t *)&ip6i[1];
3222 
3223 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
3224 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
3225 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
3226 		}
3227 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
3228 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
3229 		}
3230 	} else {
3231 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
3232 	}
3233 
3234 	if (!(ipp->ipp_fields & IPPF_ADDR))
3235 		ip6h->ip6_src = icmp->icmp_v6src;
3236 
3237 	/* Try to get everything in a single mblk */
3238 	if (hdrs_len > icmp->icmp_max_hdr_len) {
3239 		icmp->icmp_max_hdr_len = hdrs_len;
3240 		rw_exit(&icmp->icmp_rwlock);
3241 		(void) proto_set_tx_wroff(icmp->icmp_connp->conn_rq,
3242 		    icmp->icmp_connp,
3243 		    icmp->icmp_max_hdr_len + is->is_wroff_extra);
3244 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3245 	}
3246 	return (0);
3247 }
3248 
3249 /*
3250  * This routine retrieves the value of an ND variable in a icmpparam_t
3251  * structure.  It is called through nd_getset when a user reads the
3252  * variable.
3253  */
3254 /* ARGSUSED */
3255 static int
3256 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3257 {
3258 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3259 
3260 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
3261 	return (0);
3262 }
3263 
3264 /*
3265  * Walk through the param array specified registering each element with the
3266  * named dispatch (ND) handler.
3267  */
3268 static boolean_t
3269 icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt)
3270 {
3271 	for (; cnt-- > 0; icmppa++) {
3272 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
3273 			if (!nd_load(ndp, icmppa->icmp_param_name,
3274 			    icmp_param_get, icmp_param_set,
3275 			    (caddr_t)icmppa)) {
3276 				nd_free(ndp);
3277 				return (B_FALSE);
3278 			}
3279 		}
3280 	}
3281 	if (!nd_load(ndp, "icmp_status", icmp_status_report, NULL,
3282 	    NULL)) {
3283 		nd_free(ndp);
3284 		return (B_FALSE);
3285 	}
3286 	return (B_TRUE);
3287 }
3288 
3289 /* This routine sets an ND variable in a icmpparam_t structure. */
3290 /* ARGSUSED */
3291 static int
3292 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3293 {
3294 	long		new_value;
3295 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3296 
3297 	/*
3298 	 * Fail the request if the new value does not lie within the
3299 	 * required bounds.
3300 	 */
3301 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3302 	    new_value < icmppa->icmp_param_min ||
3303 	    new_value > icmppa->icmp_param_max) {
3304 		return (EINVAL);
3305 	}
3306 	/* Set the new value */
3307 	icmppa->icmp_param_value = new_value;
3308 	return (0);
3309 }
3310 static void
3311 icmp_queue_fallback(icmp_t *icmp, mblk_t *mp)
3312 {
3313 	ASSERT(MUTEX_HELD(&icmp->icmp_recv_lock));
3314 	if (IPCL_IS_NONSTR(icmp->icmp_connp)) {
3315 		/*
3316 		 * fallback has started but messages have not been moved yet
3317 		 */
3318 		if (icmp->icmp_fallback_queue_head == NULL) {
3319 			ASSERT(icmp->icmp_fallback_queue_tail == NULL);
3320 			icmp->icmp_fallback_queue_head = mp;
3321 			icmp->icmp_fallback_queue_tail = mp;
3322 		} else {
3323 			ASSERT(icmp->icmp_fallback_queue_tail != NULL);
3324 			icmp->icmp_fallback_queue_tail->b_next = mp;
3325 			icmp->icmp_fallback_queue_tail = mp;
3326 		}
3327 		mutex_exit(&icmp->icmp_recv_lock);
3328 	} else {
3329 		/*
3330 		 * no more fallbacks possible, ok to drop lock.
3331 		 */
3332 		mutex_exit(&icmp->icmp_recv_lock);
3333 		putnext(icmp->icmp_connp->conn_rq, mp);
3334 	}
3335 }
3336 
3337 /*ARGSUSED2*/
3338 static void
3339 icmp_input(void *arg1, mblk_t *mp, void *arg2)
3340 {
3341 	conn_t *connp = (conn_t *)arg1;
3342 	struct T_unitdata_ind	*tudi;
3343 	uchar_t			*rptr;
3344 	icmp_t			*icmp;
3345 	icmp_stack_t		*is;
3346 	sin_t			*sin;
3347 	sin6_t			*sin6;
3348 	ip6_t			*ip6h;
3349 	ip6i_t			*ip6i;
3350 	mblk_t			*mp1;
3351 	int			hdr_len;
3352 	ipha_t			*ipha;
3353 	int			udi_size;	/* Size of T_unitdata_ind */
3354 	uint_t			ipvers;
3355 	ip6_pkt_t		ipp;
3356 	uint8_t			nexthdr;
3357 	ip_pktinfo_t		*pinfo = NULL;
3358 	mblk_t			*options_mp = NULL;
3359 	uint_t			icmp_opt = 0;
3360 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
3361 	uint_t			hopstrip;
3362 	int			error;
3363 
3364 	ASSERT(connp->conn_flags & IPCL_RAWIPCONN);
3365 
3366 	icmp = connp->conn_icmp;
3367 	is = icmp->icmp_is;
3368 	rptr = mp->b_rptr;
3369 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3370 	ASSERT(OK_32PTR(rptr));
3371 
3372 	/*
3373 	 * IP should have prepended the options data in an M_CTL
3374 	 * Check M_CTL "type" to make sure are not here bcos of
3375 	 * a valid ICMP message
3376 	 */
3377 	if (DB_TYPE(mp) == M_CTL) {
3378 		/*
3379 		 * FIXME: does IP still do this?
3380 		 * IP sends up the IPSEC_IN message for handling IPSEC
3381 		 * policy at the TCP level. We don't need it here.
3382 		 */
3383 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
3384 			mp1 = mp->b_cont;
3385 			freeb(mp);
3386 			mp = mp1;
3387 			rptr = mp->b_rptr;
3388 		} else if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3389 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3390 		    IN_PKTINFO) {
3391 			/*
3392 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3393 			 * has been prepended to the packet by IP. We need to
3394 			 * extract the mblk and adjust the rptr
3395 			 */
3396 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3397 			options_mp = mp;
3398 			mp = mp->b_cont;
3399 			rptr = mp->b_rptr;
3400 		} else {
3401 			/*
3402 			 * ICMP messages.
3403 			 */
3404 			icmp_icmp_error(connp, mp);
3405 			return;
3406 		}
3407 	}
3408 
3409 	/*
3410 	 * Discard message if it is misaligned or smaller than the IP header.
3411 	 */
3412 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3413 		freemsg(mp);
3414 		if (options_mp != NULL)
3415 			freeb(options_mp);
3416 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3417 		return;
3418 	}
3419 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3420 
3421 	/* Handle M_DATA messages containing IP packets messages */
3422 	if (ipvers == IPV4_VERSION) {
3423 		/*
3424 		 * Special case where IP attaches
3425 		 * the IRE needs to be handled so that we don't send up
3426 		 * IRE to the user land.
3427 		 */
3428 		ipha = (ipha_t *)rptr;
3429 		hdr_len = IPH_HDR_LENGTH(ipha);
3430 
3431 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3432 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3433 
3434 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3435 			    TH_SYN) && mp->b_cont != NULL) {
3436 				mp1 = mp->b_cont;
3437 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3438 					freeb(mp1);
3439 					mp->b_cont = NULL;
3440 				}
3441 			}
3442 		}
3443 		if (is->is_bsd_compat) {
3444 			ushort_t len;
3445 			len = ntohs(ipha->ipha_length);
3446 
3447 			if (mp->b_datap->db_ref > 1) {
3448 				/*
3449 				 * Allocate a new IP header so that we can
3450 				 * modify ipha_length.
3451 				 */
3452 				mblk_t	*mp1;
3453 
3454 				mp1 = allocb(hdr_len, BPRI_MED);
3455 				if (!mp1) {
3456 					freemsg(mp);
3457 					if (options_mp != NULL)
3458 						freeb(options_mp);
3459 					BUMP_MIB(&is->is_rawip_mib,
3460 					    rawipInErrors);
3461 					return;
3462 				}
3463 				bcopy(rptr, mp1->b_rptr, hdr_len);
3464 				mp->b_rptr = rptr + hdr_len;
3465 				rptr = mp1->b_rptr;
3466 				ipha = (ipha_t *)rptr;
3467 				mp1->b_cont = mp;
3468 				mp1->b_wptr = rptr + hdr_len;
3469 				mp = mp1;
3470 			}
3471 			len -= hdr_len;
3472 			ipha->ipha_length = htons(len);
3473 		}
3474 	}
3475 
3476 	/*
3477 	 * This is the inbound data path.  Packets are passed upstream as
3478 	 * T_UNITDATA_IND messages with full IP headers still attached.
3479 	 */
3480 	if (icmp->icmp_family == AF_INET) {
3481 		ASSERT(ipvers == IPV4_VERSION);
3482 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3483 		if (icmp->icmp_recvif && (pinfo != NULL) &&
3484 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3485 			udi_size += sizeof (struct T_opthdr) +
3486 			    sizeof (uint_t);
3487 		}
3488 
3489 		if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3490 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3491 			udi_size += sizeof (struct T_opthdr) +
3492 			    sizeof (struct in_pktinfo);
3493 		}
3494 
3495 		/*
3496 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3497 		 * buffer. Since gethrestime() expects a pointer aligned
3498 		 * argument, we allocate space necessary for extra
3499 		 * alignment (even though it might not be used).
3500 		 */
3501 		if (icmp->icmp_timestamp) {
3502 			udi_size += sizeof (struct T_opthdr) +
3503 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3504 		}
3505 		mp1 = allocb(udi_size, BPRI_MED);
3506 		if (mp1 == NULL) {
3507 			freemsg(mp);
3508 			if (options_mp != NULL)
3509 				freeb(options_mp);
3510 			BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3511 			return;
3512 		}
3513 		mp1->b_cont = mp;
3514 		mp = mp1;
3515 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3516 		mp->b_datap->db_type = M_PROTO;
3517 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3518 		tudi->PRIM_type = T_UNITDATA_IND;
3519 		tudi->SRC_length = sizeof (sin_t);
3520 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3521 		sin = (sin_t *)&tudi[1];
3522 		*sin = sin_null;
3523 		sin->sin_family = AF_INET;
3524 		sin->sin_addr.s_addr = ipha->ipha_src;
3525 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3526 		    sizeof (sin_t);
3527 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3528 		tudi->OPT_length = udi_size;
3529 
3530 		/*
3531 		 * Add options if IP_RECVIF is set
3532 		 */
3533 		if (udi_size != 0) {
3534 			char *dstopt;
3535 
3536 			dstopt = (char *)&sin[1];
3537 			if (icmp->icmp_recvif && (pinfo != NULL) &&
3538 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3539 
3540 				struct T_opthdr *toh;
3541 				uint_t		*dstptr;
3542 
3543 				toh = (struct T_opthdr *)dstopt;
3544 				toh->level = IPPROTO_IP;
3545 				toh->name = IP_RECVIF;
3546 				toh->len = sizeof (struct T_opthdr) +
3547 				    sizeof (uint_t);
3548 				toh->status = 0;
3549 				dstopt += sizeof (struct T_opthdr);
3550 				dstptr = (uint_t *)dstopt;
3551 				*dstptr = pinfo->ip_pkt_ifindex;
3552 				dstopt += sizeof (uint_t);
3553 				udi_size -= toh->len;
3554 			}
3555 			if (icmp->icmp_timestamp) {
3556 				struct	T_opthdr *toh;
3557 
3558 				toh = (struct T_opthdr *)dstopt;
3559 				toh->level = SOL_SOCKET;
3560 				toh->name = SCM_TIMESTAMP;
3561 				toh->len = sizeof (struct T_opthdr) +
3562 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3563 				toh->status = 0;
3564 				dstopt += sizeof (struct T_opthdr);
3565 				/* Align for gethrestime() */
3566 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3567 				    sizeof (intptr_t));
3568 				gethrestime((timestruc_t *)dstopt);
3569 				dstopt = (char *)toh + toh->len;
3570 				udi_size -= toh->len;
3571 			}
3572 			if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3573 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3574 				struct	T_opthdr *toh;
3575 				struct	in_pktinfo *pktinfop;
3576 
3577 				toh = (struct T_opthdr *)dstopt;
3578 				toh->level = IPPROTO_IP;
3579 				toh->name = IP_PKTINFO;
3580 				toh->len = sizeof (struct T_opthdr) +
3581 				    sizeof (in_pktinfo_t);
3582 				toh->status = 0;
3583 				dstopt += sizeof (struct T_opthdr);
3584 				pktinfop = (struct in_pktinfo *)dstopt;
3585 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3586 				pktinfop->ipi_spec_dst =
3587 				    pinfo->ip_pkt_match_addr;
3588 
3589 				pktinfop->ipi_addr.s_addr = ipha->ipha_dst;
3590 
3591 				dstopt += sizeof (struct in_pktinfo);
3592 				udi_size -= toh->len;
3593 			}
3594 
3595 			/* Consumed all of allocated space */
3596 			ASSERT(udi_size == 0);
3597 		}
3598 
3599 		if (options_mp != NULL)
3600 			freeb(options_mp);
3601 
3602 		BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3603 		goto deliver;
3604 	}
3605 
3606 	/*
3607 	 * We don't need options_mp in the IPv6 path.
3608 	 */
3609 	if (options_mp != NULL) {
3610 		freeb(options_mp);
3611 		options_mp = NULL;
3612 	}
3613 
3614 	/*
3615 	 * Discard message if it is smaller than the IPv6 header
3616 	 * or if the header is malformed.
3617 	 */
3618 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3619 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3620 	    icmp->icmp_family != AF_INET6) {
3621 		freemsg(mp);
3622 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3623 		return;
3624 	}
3625 
3626 	/* Initialize */
3627 	ipp.ipp_fields = 0;
3628 	hopstrip = 0;
3629 
3630 	ip6h = (ip6_t *)rptr;
3631 	/*
3632 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3633 	 * as well as individual lenghts of ext hdrs (and ptrs to
3634 	 * them).
3635 	 */
3636 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3637 		/* Look for ifindex information */
3638 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3639 			ip6i = (ip6i_t *)ip6h;
3640 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3641 				ASSERT(ip6i->ip6i_ifindex != 0);
3642 				ipp.ipp_fields |= IPPF_IFINDEX;
3643 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3644 			}
3645 			rptr = (uchar_t *)&ip6i[1];
3646 			mp->b_rptr = rptr;
3647 			if (rptr == mp->b_wptr) {
3648 				mp1 = mp->b_cont;
3649 				freeb(mp);
3650 				mp = mp1;
3651 				rptr = mp->b_rptr;
3652 			}
3653 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3654 			ip6h = (ip6_t *)rptr;
3655 		}
3656 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3657 
3658 		/*
3659 		 * We need to lie a bit to the user because users inside
3660 		 * labeled compartments should not see their own labels.  We
3661 		 * assume that in all other respects IP has checked the label,
3662 		 * and that the label is always first among the options.  (If
3663 		 * it's not first, then this code won't see it, and the option
3664 		 * will be passed along to the user.)
3665 		 *
3666 		 * If we had multilevel ICMP sockets, then the following code
3667 		 * should be skipped for them to allow the user to see the
3668 		 * label.
3669 		 *
3670 		 * Alignment restrictions in the definition of IP options
3671 		 * (namely, the requirement that the 4-octet DOI goes on a
3672 		 * 4-octet boundary) mean that we know exactly where the option
3673 		 * should start, but we're lenient for other hosts.
3674 		 *
3675 		 * Note that there are no multilevel ICMP or raw IP sockets
3676 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3677 		 */
3678 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3679 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3680 			const uchar_t *ucp =
3681 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3682 			int remlen = ipp.ipp_hopoptslen - 2;
3683 
3684 			while (remlen > 0) {
3685 				if (*ucp == IP6OPT_PAD1) {
3686 					remlen--;
3687 					ucp++;
3688 				} else if (*ucp == IP6OPT_PADN) {
3689 					remlen -= ucp[1] + 2;
3690 					ucp += ucp[1] + 2;
3691 				} else if (*ucp == ip6opt_ls) {
3692 					hopstrip = (ucp -
3693 					    (const uchar_t *)ipp.ipp_hopopts) +
3694 					    ucp[1] + 2;
3695 					hopstrip = (hopstrip + 7) & ~7;
3696 					break;
3697 				} else {
3698 					/* label option must be first */
3699 					break;
3700 				}
3701 			}
3702 		}
3703 	} else {
3704 		hdr_len = IPV6_HDR_LEN;
3705 		ip6i = NULL;
3706 		nexthdr = ip6h->ip6_nxt;
3707 	}
3708 	/*
3709 	 * One special case where IP attaches the IRE needs to
3710 	 * be handled so that we don't send up IRE to the user land.
3711 	 */
3712 	if (nexthdr == IPPROTO_TCP) {
3713 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3714 
3715 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3716 		    mp->b_cont != NULL) {
3717 			mp1 = mp->b_cont;
3718 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3719 				freeb(mp1);
3720 				mp->b_cont = NULL;
3721 			}
3722 		}
3723 	}
3724 	/*
3725 	 * Check a filter for ICMPv6 types if needed.
3726 	 * Verify raw checksums if needed.
3727 	 */
3728 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3729 		if (icmp->icmp_filter != NULL) {
3730 			int type;
3731 
3732 			/* Assumes that IP has done the pullupmsg */
3733 			type = mp->b_rptr[hdr_len];
3734 
3735 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3736 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3737 				freemsg(mp);
3738 				return;
3739 			}
3740 		} else {
3741 			/* Checksum */
3742 			uint16_t	*up;
3743 			uint32_t	sum;
3744 			int		remlen;
3745 
3746 			up = (uint16_t *)&ip6h->ip6_src;
3747 
3748 			remlen = msgdsize(mp) - hdr_len;
3749 			sum = htons(icmp->icmp_proto + remlen)
3750 			    + up[0] + up[1] + up[2] + up[3]
3751 			    + up[4] + up[5] + up[6] + up[7]
3752 			    + up[8] + up[9] + up[10] + up[11]
3753 			    + up[12] + up[13] + up[14] + up[15];
3754 			sum = (sum & 0xffff) + (sum >> 16);
3755 			sum = IP_CSUM(mp, hdr_len, sum);
3756 			if (sum != 0) {
3757 				/* IPv6 RAW checksum failed */
3758 				ip0dbg(("icmp_rput: RAW checksum "
3759 				    "failed %x\n", sum));
3760 				freemsg(mp);
3761 				BUMP_MIB(&is->is_rawip_mib,
3762 				    rawipInCksumErrs);
3763 				return;
3764 			}
3765 		}
3766 	}
3767 	/* Skip all the IPv6 headers per API */
3768 	mp->b_rptr += hdr_len;
3769 
3770 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3771 
3772 	/*
3773 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3774 	 * maintain state information, instead of relying on icmp_t
3775 	 * structure, since there arent any locks protecting these members
3776 	 * and there is a window where there might be a race between a
3777 	 * thread setting options on the write side and a thread reading
3778 	 * these options on the read size.
3779 	 */
3780 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3781 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3782 		if (icmp->icmp_ipv6_recvhopopts &&
3783 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3784 		    ipp.ipp_hopoptslen > hopstrip) {
3785 			udi_size += sizeof (struct T_opthdr) +
3786 			    ipp.ipp_hopoptslen - hopstrip;
3787 			icmp_opt |= IPPF_HOPOPTS;
3788 		}
3789 		if ((icmp->icmp_ipv6_recvdstopts ||
3790 		    icmp->icmp_old_ipv6_recvdstopts) &&
3791 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3792 			udi_size += sizeof (struct T_opthdr) +
3793 			    ipp.ipp_dstoptslen;
3794 			icmp_opt |= IPPF_DSTOPTS;
3795 		}
3796 		if (((icmp->icmp_ipv6_recvdstopts &&
3797 		    icmp->icmp_ipv6_recvrthdr &&
3798 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3799 		    icmp->icmp_ipv6_recvrtdstopts) &&
3800 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3801 			udi_size += sizeof (struct T_opthdr) +
3802 			    ipp.ipp_rtdstoptslen;
3803 			icmp_opt |= IPPF_RTDSTOPTS;
3804 		}
3805 		if (icmp->icmp_ipv6_recvrthdr &&
3806 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3807 			udi_size += sizeof (struct T_opthdr) +
3808 			    ipp.ipp_rthdrlen;
3809 			icmp_opt |= IPPF_RTHDR;
3810 		}
3811 		if (icmp->icmp_ip_recvpktinfo &&
3812 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3813 			udi_size += sizeof (struct T_opthdr) +
3814 			    sizeof (struct in6_pktinfo);
3815 			icmp_opt |= IPPF_IFINDEX;
3816 		}
3817 	}
3818 	if (icmp->icmp_ipv6_recvhoplimit) {
3819 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3820 		icmp_ipv6_recvhoplimit = B_TRUE;
3821 	}
3822 
3823 	if (icmp->icmp_ipv6_recvtclass)
3824 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3825 
3826 	/*
3827 	 * If SO_TIMESTAMP is set allocate the appropriate sized
3828 	 * buffer. Since gethrestime() expects a pointer aligned
3829 	 * argument, we allocate space necessary for extra
3830 	 * alignment (even though it might not be used).
3831 	 */
3832 	if (icmp->icmp_timestamp) {
3833 		udi_size += sizeof (struct T_opthdr) +
3834 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3835 	}
3836 
3837 	mp1 = allocb(udi_size, BPRI_MED);
3838 	if (mp1 == NULL) {
3839 		freemsg(mp);
3840 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3841 		return;
3842 	}
3843 	mp1->b_cont = mp;
3844 	mp = mp1;
3845 	mp->b_datap->db_type = M_PROTO;
3846 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3847 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3848 	tudi->PRIM_type = T_UNITDATA_IND;
3849 	tudi->SRC_length = sizeof (sin6_t);
3850 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3851 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3852 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3853 	tudi->OPT_length = udi_size;
3854 	sin6 = (sin6_t *)&tudi[1];
3855 	sin6->sin6_port = 0;
3856 	sin6->sin6_family = AF_INET6;
3857 
3858 	sin6->sin6_addr = ip6h->ip6_src;
3859 	/* No sin6_flowinfo per API */
3860 	sin6->sin6_flowinfo = 0;
3861 	/* For link-scope source pass up scope id */
3862 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3863 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3864 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3865 	else
3866 		sin6->sin6_scope_id = 0;
3867 
3868 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3869 	    icmp->icmp_zoneid, is->is_netstack);
3870 
3871 	if (udi_size != 0) {
3872 		uchar_t *dstopt;
3873 
3874 		dstopt = (uchar_t *)&sin6[1];
3875 		if (icmp_opt & IPPF_IFINDEX) {
3876 			struct T_opthdr *toh;
3877 			struct in6_pktinfo *pkti;
3878 
3879 			toh = (struct T_opthdr *)dstopt;
3880 			toh->level = IPPROTO_IPV6;
3881 			toh->name = IPV6_PKTINFO;
3882 			toh->len = sizeof (struct T_opthdr) +
3883 			    sizeof (*pkti);
3884 			toh->status = 0;
3885 			dstopt += sizeof (struct T_opthdr);
3886 			pkti = (struct in6_pktinfo *)dstopt;
3887 			pkti->ipi6_addr = ip6h->ip6_dst;
3888 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3889 			dstopt += sizeof (*pkti);
3890 			udi_size -= toh->len;
3891 		}
3892 		if (icmp_ipv6_recvhoplimit) {
3893 			struct T_opthdr *toh;
3894 
3895 			toh = (struct T_opthdr *)dstopt;
3896 			toh->level = IPPROTO_IPV6;
3897 			toh->name = IPV6_HOPLIMIT;
3898 			toh->len = sizeof (struct T_opthdr) +
3899 			    sizeof (uint_t);
3900 			toh->status = 0;
3901 			dstopt += sizeof (struct T_opthdr);
3902 			*(uint_t *)dstopt = ip6h->ip6_hops;
3903 			dstopt += sizeof (uint_t);
3904 			udi_size -= toh->len;
3905 		}
3906 		if (icmp->icmp_ipv6_recvtclass) {
3907 			struct T_opthdr *toh;
3908 
3909 			toh = (struct T_opthdr *)dstopt;
3910 			toh->level = IPPROTO_IPV6;
3911 			toh->name = IPV6_TCLASS;
3912 			toh->len = sizeof (struct T_opthdr) +
3913 			    sizeof (uint_t);
3914 			toh->status = 0;
3915 			dstopt += sizeof (struct T_opthdr);
3916 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3917 			dstopt += sizeof (uint_t);
3918 			udi_size -= toh->len;
3919 		}
3920 		if (icmp->icmp_timestamp) {
3921 			struct  T_opthdr *toh;
3922 
3923 			toh = (struct T_opthdr *)dstopt;
3924 			toh->level = SOL_SOCKET;
3925 			toh->name = SCM_TIMESTAMP;
3926 			toh->len = sizeof (struct T_opthdr) +
3927 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3928 			toh->status = 0;
3929 			dstopt += sizeof (struct T_opthdr);
3930 			/* Align for gethrestime() */
3931 			dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
3932 			    sizeof (intptr_t));
3933 			gethrestime((timestruc_t *)dstopt);
3934 			dstopt = (uchar_t *)toh + toh->len;
3935 			udi_size -= toh->len;
3936 		}
3937 
3938 		if (icmp_opt & IPPF_HOPOPTS) {
3939 			struct T_opthdr *toh;
3940 
3941 			toh = (struct T_opthdr *)dstopt;
3942 			toh->level = IPPROTO_IPV6;
3943 			toh->name = IPV6_HOPOPTS;
3944 			toh->len = sizeof (struct T_opthdr) +
3945 			    ipp.ipp_hopoptslen - hopstrip;
3946 			toh->status = 0;
3947 			dstopt += sizeof (struct T_opthdr);
3948 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
3949 			    ipp.ipp_hopoptslen - hopstrip);
3950 			if (hopstrip > 0) {
3951 				/* copy next header value and fake length */
3952 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
3953 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
3954 				    hopstrip / 8;
3955 			}
3956 			dstopt += ipp.ipp_hopoptslen - hopstrip;
3957 			udi_size -= toh->len;
3958 		}
3959 		if (icmp_opt & IPPF_RTDSTOPTS) {
3960 			struct T_opthdr *toh;
3961 
3962 			toh = (struct T_opthdr *)dstopt;
3963 			toh->level = IPPROTO_IPV6;
3964 			toh->name = IPV6_DSTOPTS;
3965 			toh->len = sizeof (struct T_opthdr) +
3966 			    ipp.ipp_rtdstoptslen;
3967 			toh->status = 0;
3968 			dstopt += sizeof (struct T_opthdr);
3969 			bcopy(ipp.ipp_rtdstopts, dstopt,
3970 			    ipp.ipp_rtdstoptslen);
3971 			dstopt += ipp.ipp_rtdstoptslen;
3972 			udi_size -= toh->len;
3973 		}
3974 		if (icmp_opt & IPPF_RTHDR) {
3975 			struct T_opthdr *toh;
3976 
3977 			toh = (struct T_opthdr *)dstopt;
3978 			toh->level = IPPROTO_IPV6;
3979 			toh->name = IPV6_RTHDR;
3980 			toh->len = sizeof (struct T_opthdr) +
3981 			    ipp.ipp_rthdrlen;
3982 			toh->status = 0;
3983 			dstopt += sizeof (struct T_opthdr);
3984 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
3985 			dstopt += ipp.ipp_rthdrlen;
3986 			udi_size -= toh->len;
3987 		}
3988 		if (icmp_opt & IPPF_DSTOPTS) {
3989 			struct T_opthdr *toh;
3990 
3991 			toh = (struct T_opthdr *)dstopt;
3992 			toh->level = IPPROTO_IPV6;
3993 			toh->name = IPV6_DSTOPTS;
3994 			toh->len = sizeof (struct T_opthdr) +
3995 			    ipp.ipp_dstoptslen;
3996 			toh->status = 0;
3997 			dstopt += sizeof (struct T_opthdr);
3998 			bcopy(ipp.ipp_dstopts, dstopt,
3999 			    ipp.ipp_dstoptslen);
4000 			dstopt += ipp.ipp_dstoptslen;
4001 			udi_size -= toh->len;
4002 		}
4003 		/* Consumed all of allocated space */
4004 		ASSERT(udi_size == 0);
4005 	}
4006 	BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
4007 
4008 deliver:
4009 	if (IPCL_IS_NONSTR(connp)) {
4010 		if ((*connp->conn_upcalls->su_recv)
4011 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
4012 		    NULL) < 0) {
4013 			mutex_enter(&icmp->icmp_recv_lock);
4014 			if (error == ENOSPC) {
4015 				/*
4016 				 * let's confirm while holding the lock
4017 				 */
4018 				if ((*connp->conn_upcalls->su_recv)
4019 				    (connp->conn_upper_handle, NULL, 0, 0,
4020 				    &error, NULL) < 0) {
4021 					if (error == ENOSPC) {
4022 						connp->conn_flow_cntrld =
4023 						    B_TRUE;
4024 					} else {
4025 						ASSERT(error == EOPNOTSUPP);
4026 					}
4027 				}
4028 				mutex_exit(&icmp->icmp_recv_lock);
4029 			} else {
4030 				ASSERT(error == EOPNOTSUPP);
4031 				icmp_queue_fallback(icmp, mp);
4032 			}
4033 		}
4034 	} else {
4035 		putnext(connp->conn_rq, mp);
4036 	}
4037 	ASSERT(MUTEX_NOT_HELD(&icmp->icmp_recv_lock));
4038 }
4039 
4040 /*
4041  * return SNMP stuff in buffer in mpdata
4042  */
4043 mblk_t *
4044 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
4045 {
4046 	mblk_t			*mpdata;
4047 	struct opthdr		*optp;
4048 	conn_t			*connp = Q_TO_CONN(q);
4049 	icmp_stack_t		*is = connp->conn_netstack->netstack_icmp;
4050 	mblk_t			*mp2ctl;
4051 
4052 	/*
4053 	 * make a copy of the original message
4054 	 */
4055 	mp2ctl = copymsg(mpctl);
4056 
4057 	if (mpctl == NULL ||
4058 	    (mpdata = mpctl->b_cont) == NULL) {
4059 		freemsg(mpctl);
4060 		freemsg(mp2ctl);
4061 		return (0);
4062 	}
4063 
4064 	/* fixed length structure for IPv4 and IPv6 counters */
4065 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4066 	optp->level = EXPER_RAWIP;
4067 	optp->name = 0;
4068 	(void) snmp_append_data(mpdata, (char *)&is->is_rawip_mib,
4069 	    sizeof (is->is_rawip_mib));
4070 	optp->len = msgdsize(mpdata);
4071 	qreply(q, mpctl);
4072 
4073 	return (mp2ctl);
4074 }
4075 
4076 /*
4077  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
4078  * TODO:  If this ever actually tries to set anything, it needs to be
4079  * to do the appropriate locking.
4080  */
4081 /* ARGSUSED */
4082 int
4083 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4084     uchar_t *ptr, int len)
4085 {
4086 	switch (level) {
4087 	case EXPER_RAWIP:
4088 		return (0);
4089 	default:
4090 		return (1);
4091 	}
4092 }
4093 
4094 /* Report for ndd "icmp_status" */
4095 /* ARGSUSED */
4096 static int
4097 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4098 {
4099 	conn_t  *connp;
4100 	ip_stack_t *ipst;
4101 	char	laddrbuf[INET6_ADDRSTRLEN];
4102 	char	faddrbuf[INET6_ADDRSTRLEN];
4103 	int	i;
4104 
4105 	(void) mi_mpprintf(mp,
4106 	    "RAWIP    " MI_COL_HDRPAD_STR
4107 	/*   01234567[89ABCDEF] */
4108 	    "  src addr        dest addr       state");
4109 	/*   xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */
4110 
4111 	connp = Q_TO_CONN(q);
4112 	ipst = connp->conn_netstack->netstack_ip;
4113 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4114 		connf_t *connfp;
4115 		char	*state;
4116 
4117 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4118 		connp = NULL;
4119 
4120 		while ((connp = ipcl_get_next_conn(connfp, connp,
4121 		    IPCL_RAWIPCONN)) != NULL) {
4122 			icmp_t  *icmp;
4123 
4124 			mutex_enter(&(connp)->conn_lock);
4125 			icmp = connp->conn_icmp;
4126 
4127 			if (icmp->icmp_state == TS_UNBND)
4128 				state = "UNBOUND";
4129 			else if (icmp->icmp_state == TS_IDLE)
4130 				state = "IDLE";
4131 			else if (icmp->icmp_state == TS_DATA_XFER)
4132 				state = "CONNECTED";
4133 			else
4134 				state = "UnkState";
4135 
4136 			(void) mi_mpprintf(mp, MI_COL_PTRFMT_STR "%s %s %s",
4137 			    (void *)icmp,
4138 			    inet_ntop(AF_INET6, &icmp->icmp_v6dst.sin6_addr,
4139 			    faddrbuf,
4140 			    sizeof (faddrbuf)),
4141 			    inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf,
4142 			    sizeof (laddrbuf)),
4143 			    state);
4144 			mutex_exit(&(connp)->conn_lock);
4145 		}
4146 	}
4147 	return (0);
4148 }
4149 
4150 /*
4151  * This routine creates a T_UDERROR_IND message and passes it upstream.
4152  * The address and options are copied from the T_UNITDATA_REQ message
4153  * passed in mp.  This message is freed.
4154  */
4155 static void
4156 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
4157 {
4158 	mblk_t	*mp1;
4159 	uchar_t	*rptr = mp->b_rptr;
4160 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
4161 
4162 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
4163 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
4164 	    tudr->OPT_length, err);
4165 	if (mp1)
4166 		qreply(q, mp1);
4167 	freemsg(mp);
4168 }
4169 
4170 
4171 static int
4172 rawip_do_unbind(conn_t *connp)
4173 {
4174 	icmp_t *icmp = connp->conn_icmp;
4175 
4176 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4177 	/* If a bind has not been done, we can't unbind. */
4178 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
4179 		rw_exit(&icmp->icmp_rwlock);
4180 		return (-TOUTSTATE);
4181 	}
4182 	icmp->icmp_pending_op = T_UNBIND_REQ;
4183 	rw_exit(&icmp->icmp_rwlock);
4184 
4185 	/*
4186 	 * Call ip to unbind
4187 	 */
4188 
4189 	ip_unbind(connp);
4190 
4191 	/*
4192 	 * Once we're unbound from IP, the pending operation may be cleared
4193 	 * here.
4194 	 */
4195 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4196 	V6_SET_ZERO(icmp->icmp_v6src);
4197 	V6_SET_ZERO(icmp->icmp_bound_v6src);
4198 	icmp->icmp_pending_op = -1;
4199 	icmp->icmp_state = TS_UNBND;
4200 	if (icmp->icmp_family == AF_INET6)
4201 		(void) icmp_build_hdrs(icmp);
4202 	rw_exit(&icmp->icmp_rwlock);
4203 	return (0);
4204 }
4205 
4206 /*
4207  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
4208  * After some error checking, the message is passed downstream to ip.
4209  */
4210 static void
4211 icmp_tpi_unbind(queue_t *q, mblk_t *mp)
4212 {
4213 	conn_t	*connp = Q_TO_CONN(q);
4214 	int	error;
4215 
4216 	ASSERT(mp->b_cont == NULL);
4217 	error = rawip_do_unbind(connp);
4218 	if (error) {
4219 		if (error < 0) {
4220 			icmp_err_ack(q, mp, -error, 0);
4221 		} else {
4222 			icmp_err_ack(q, mp, 0, error);
4223 		}
4224 		return;
4225 	}
4226 
4227 	/*
4228 	 * Convert mp into a T_OK_ACK
4229 	 */
4230 
4231 	mp = mi_tpi_ok_ack_alloc(mp);
4232 
4233 	/*
4234 	 * should not happen in practice... T_OK_ACK is smaller than the
4235 	 * original message.
4236 	 */
4237 	ASSERT(mp != NULL);
4238 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4239 	qreply(q, mp);
4240 }
4241 
4242 
4243 /*
4244  * Process IPv4 packets that already include an IP header.
4245  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
4246  * IPPROTO_IGMP).
4247  */
4248 static int
4249 icmp_wput_hdrincl(queue_t *q, conn_t *connp, mblk_t *mp, icmp_t *icmp,
4250     ip4_pkt_t *pktinfop)
4251 {
4252 	icmp_stack_t *is = icmp->icmp_is;
4253 	ipha_t	*ipha;
4254 	int	ip_hdr_length;
4255 	int	tp_hdr_len;
4256 	mblk_t	*mp1;
4257 	uint_t	pkt_len;
4258 	ip_opt_info_t optinfo;
4259 
4260 	optinfo.ip_opt_flags = 0;
4261 	optinfo.ip_opt_ill_index = 0;
4262 	ipha = (ipha_t *)mp->b_rptr;
4263 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4264 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
4265 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4266 			ASSERT(icmp != NULL);
4267 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4268 			freemsg(mp);
4269 			return (0);
4270 		}
4271 		ipha = (ipha_t *)mp->b_rptr;
4272 	}
4273 	ipha->ipha_version_and_hdr_length =
4274 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
4275 
4276 	/*
4277 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4278 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4279 	 * tell IP that the application has sent a complete IP header and not
4280 	 * to compute the transport checksum nor change the DF flag.
4281 	 */
4282 	ipha->ipha_ident = IP_HDR_INCLUDED;
4283 	ipha->ipha_hdr_checksum = 0;
4284 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
4285 	/* Insert options if any */
4286 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4287 		/*
4288 		 * Put the IP header plus any transport header that is
4289 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
4290 		 * that at least the checksum field is in the first mblk.)
4291 		 */
4292 		switch (ipha->ipha_protocol) {
4293 		case IPPROTO_UDP:
4294 			tp_hdr_len = 8;
4295 			break;
4296 		case IPPROTO_TCP:
4297 			tp_hdr_len = 20;
4298 			break;
4299 		default:
4300 			tp_hdr_len = 0;
4301 			break;
4302 		}
4303 		/*
4304 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
4305 		 * tp_hdr_len bytes will be in a single mblk.
4306 		 */
4307 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
4308 		    tp_hdr_len)) {
4309 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
4310 			    tp_hdr_len)) {
4311 				BUMP_MIB(&is->is_rawip_mib,
4312 				    rawipOutErrors);
4313 				freemsg(mp);
4314 				return (0);
4315 			}
4316 			ipha = (ipha_t *)mp->b_rptr;
4317 		}
4318 
4319 		/*
4320 		 * if the length is larger then the max allowed IP packet,
4321 		 * then send an error and abort the processing.
4322 		 */
4323 		pkt_len = ntohs(ipha->ipha_length)
4324 		    + icmp->icmp_ip_snd_options_len;
4325 		if (pkt_len > IP_MAXPACKET) {
4326 			return (EMSGSIZE);
4327 		}
4328 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra +
4329 		    tp_hdr_len, BPRI_LO))) {
4330 			return (ENOMEM);
4331 		}
4332 		mp1->b_rptr += is->is_wroff_extra;
4333 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
4334 
4335 		ipha->ipha_length = htons((uint16_t)pkt_len);
4336 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
4337 
4338 		/* Copy transport header if any */
4339 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
4340 		mp1->b_wptr += tp_hdr_len;
4341 
4342 		/* Add options */
4343 		ipha = (ipha_t *)mp1->b_rptr;
4344 		bcopy(icmp->icmp_ip_snd_options, &ipha[1],
4345 		    icmp->icmp_ip_snd_options_len);
4346 
4347 		/* Drop IP header and transport header from original */
4348 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
4349 
4350 		mp1->b_cont = mp;
4351 		mp = mp1;
4352 		/*
4353 		 * Massage source route putting first source
4354 		 * route in ipha_dst.
4355 		 */
4356 		(void) ip_massage_options(ipha, is->is_netstack);
4357 	}
4358 
4359 	if (pktinfop != NULL) {
4360 		/*
4361 		 * Over write the source address provided in the header
4362 		 */
4363 		if (pktinfop->ip4_addr != INADDR_ANY) {
4364 			ipha->ipha_src = pktinfop->ip4_addr;
4365 			optinfo.ip_opt_flags = IP_VERIFY_SRC;
4366 		}
4367 
4368 		if (pktinfop->ip4_ill_index != 0) {
4369 			optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4370 		}
4371 	}
4372 
4373 	mblk_setcred(mp, connp->conn_cred);
4374 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4375 	return (0);
4376 }
4377 
4378 static int
4379 icmp_update_label(icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
4380 {
4381 	int err;
4382 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4383 	icmp_stack_t		*is = icmp->icmp_is;
4384 	conn_t	*connp = icmp->icmp_connp;
4385 
4386 	err = tsol_compute_label(DB_CREDDEF(mp, connp->conn_cred), dst,
4387 	    opt_storage, connp->conn_mac_exempt,
4388 	    is->is_netstack->netstack_ip);
4389 	if (err == 0) {
4390 		err = tsol_update_options(&icmp->icmp_ip_snd_options,
4391 		    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
4392 		    opt_storage);
4393 	}
4394 	if (err != 0) {
4395 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4396 		DTRACE_PROBE4(
4397 		    tx__ip__log__drop__updatelabel__icmp,
4398 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4399 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4400 		return (err);
4401 	}
4402 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
4403 	return (0);
4404 }
4405 
4406 /*
4407  * This routine handles all messages passed downstream.  It either
4408  * consumes the message or passes it downstream; it never queues a
4409  * a message.
4410  */
4411 static void
4412 icmp_wput(queue_t *q, mblk_t *mp)
4413 {
4414 	uchar_t	*rptr = mp->b_rptr;
4415 	ipha_t	*ipha;
4416 	mblk_t	*mp1;
4417 #define	tudr ((struct T_unitdata_req *)rptr)
4418 	size_t	ip_len;
4419 	conn_t	*connp = Q_TO_CONN(q);
4420 	icmp_t	*icmp = connp->conn_icmp;
4421 	icmp_stack_t *is = icmp->icmp_is;
4422 	sin6_t	*sin6;
4423 	sin_t	*sin;
4424 	ipaddr_t	v4dst;
4425 	ip4_pkt_t	pktinfo;
4426 	ip4_pkt_t	*pktinfop = &pktinfo;
4427 	ip6_pkt_t	ipp_s;  /* For ancillary data options */
4428 	ip6_pkt_t	*ipp = &ipp_s;
4429 	int error;
4430 
4431 	ipp->ipp_fields = 0;
4432 	ipp->ipp_sticky_ignored = 0;
4433 
4434 	switch (mp->b_datap->db_type) {
4435 	case M_DATA:
4436 		if (icmp->icmp_hdrincl) {
4437 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
4438 			ipha = (ipha_t *)mp->b_rptr;
4439 			if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH) {
4440 				if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4441 					BUMP_MIB(&is->is_rawip_mib,
4442 					    rawipOutErrors);
4443 					freemsg(mp);
4444 					return;
4445 				}
4446 				ipha = (ipha_t *)mp->b_rptr;
4447 			}
4448 			/*
4449 			 * If this connection was used for v6 (inconceivable!)
4450 			 * or if we have a new destination, then it's time to
4451 			 * figure a new label.
4452 			 */
4453 			if (is_system_labeled() &&
4454 			    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4455 			    V4_PART_OF_V6(icmp->icmp_v6lastdst) !=
4456 			    ipha->ipha_dst)) {
4457 				error = icmp_update_label(icmp, mp,
4458 				    ipha->ipha_dst);
4459 				if (error != 0) {
4460 					icmp_ud_err(q, mp, error);
4461 					return;
4462 				}
4463 			}
4464 			error = icmp_wput_hdrincl(q, connp, mp, icmp, NULL);
4465 			if (error != 0)
4466 				icmp_ud_err(q, mp, error);
4467 			return;
4468 		}
4469 		freemsg(mp);
4470 		return;
4471 	case M_PROTO:
4472 	case M_PCPROTO:
4473 		ip_len = mp->b_wptr - rptr;
4474 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4475 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4476 			if (((union T_primitives *)rptr)->type
4477 			    == T_UNITDATA_REQ)
4478 				break;
4479 		}
4480 		/* FALLTHRU */
4481 	default:
4482 		icmp_wput_other(q, mp);
4483 		return;
4484 	}
4485 
4486 	/* Handle T_UNITDATA_REQ messages here. */
4487 
4488 	mp1 = mp->b_cont;
4489 	if (mp1 == NULL) {
4490 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4491 		icmp_ud_err(q, mp, EPROTO);
4492 		return;
4493 	}
4494 
4495 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4496 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4497 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4498 		return;
4499 	}
4500 
4501 	switch (icmp->icmp_family) {
4502 	case AF_INET6:
4503 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4504 		if (!OK_32PTR((char *)sin6) ||
4505 		    tudr->DEST_length != sizeof (sin6_t) ||
4506 		    sin6->sin6_family != AF_INET6) {
4507 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4508 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4509 			return;
4510 		}
4511 
4512 		/* No support for mapped addresses on raw sockets */
4513 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4514 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4515 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4516 			return;
4517 		}
4518 
4519 		/*
4520 		 * Destination is a native IPv6 address.
4521 		 * Send out an IPv6 format packet.
4522 		 */
4523 		if (tudr->OPT_length != 0) {
4524 			int error;
4525 
4526 			error = 0;
4527 			if (icmp_unitdata_opt_process(q, mp, &error,
4528 			    (void *)ipp) < 0) {
4529 				/* failure */
4530 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4531 				icmp_ud_err(q, mp, error);
4532 				return;
4533 			}
4534 			ASSERT(error == 0);
4535 		}
4536 
4537 		error = raw_ip_send_data_v6(q, connp, mp1, sin6, ipp);
4538 		goto done;
4539 
4540 	case AF_INET:
4541 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4542 		if (!OK_32PTR((char *)sin) ||
4543 		    tudr->DEST_length != sizeof (sin_t) ||
4544 		    sin->sin_family != AF_INET) {
4545 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4546 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4547 			return;
4548 		}
4549 		/* Extract and ipaddr */
4550 		v4dst = sin->sin_addr.s_addr;
4551 		break;
4552 
4553 	default:
4554 		ASSERT(0);
4555 	}
4556 
4557 	pktinfop->ip4_ill_index = 0;
4558 	pktinfop->ip4_addr = INADDR_ANY;
4559 
4560 	/*
4561 	 * If options passed in, feed it for verification and handling
4562 	 */
4563 	if (tudr->OPT_length != 0) {
4564 		int error;
4565 
4566 		error = 0;
4567 		if (icmp_unitdata_opt_process(q, mp, &error,
4568 		    (void *)pktinfop) < 0) {
4569 			/* failure */
4570 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4571 			icmp_ud_err(q, mp, error);
4572 			return;
4573 		}
4574 		ASSERT(error == 0);
4575 		/*
4576 		 * Note: Success in processing options.
4577 		 * mp option buffer represented by
4578 		 * OPT_length/offset now potentially modified
4579 		 * and contain option setting results
4580 		 */
4581 	}
4582 
4583 	error = raw_ip_send_data_v4(q, connp, mp1, v4dst, pktinfop);
4584 done:
4585 	if (error != 0) {
4586 		icmp_ud_err(q, mp, error);
4587 		return;
4588 	} else {
4589 		mp->b_cont = NULL;
4590 		freeb(mp);
4591 	}
4592 }
4593 
4594 
4595 /* ARGSUSED */
4596 static void
4597 icmp_wput_fallback(queue_t *q, mblk_t *mp)
4598 {
4599 #ifdef DEBUG
4600 	cmn_err(CE_CONT, "icmp_wput_fallback: Message during fallback \n");
4601 #endif
4602 	freemsg(mp);
4603 }
4604 
4605 static int
4606 raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp, ipaddr_t v4dst,
4607     ip4_pkt_t *pktinfop)
4608 {
4609 	ipha_t	*ipha;
4610 	size_t	ip_len;
4611 	icmp_t	*icmp = connp->conn_icmp;
4612 	icmp_stack_t *is = icmp->icmp_is;
4613 	int	ip_hdr_length;
4614 	ip_opt_info_t	optinfo;
4615 
4616 	optinfo.ip_opt_flags = 0;
4617 	optinfo.ip_opt_ill_index = 0;
4618 
4619 	if (icmp->icmp_state == TS_UNBND) {
4620 		/* If a port has not been bound to the stream, fail. */
4621 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4622 		return (EPROTO);
4623 	}
4624 
4625 	if (v4dst == INADDR_ANY)
4626 		v4dst = htonl(INADDR_LOOPBACK);
4627 
4628 	/* Check if our saved options are valid; update if not */
4629 	if (is_system_labeled() &&
4630 	    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4631 	    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst)) {
4632 		int error = icmp_update_label(icmp, mp, v4dst);
4633 
4634 		if (error != 0)
4635 			return (error);
4636 	}
4637 
4638 	/* Protocol 255 contains full IP headers */
4639 	if (icmp->icmp_hdrincl)
4640 		return (icmp_wput_hdrincl(q, connp, mp, icmp, pktinfop));
4641 
4642 	/* Add an IP header */
4643 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4644 	ipha = (ipha_t *)&mp->b_rptr[-ip_hdr_length];
4645 	if ((uchar_t *)ipha < mp->b_datap->db_base ||
4646 	    mp->b_datap->db_ref != 1 ||
4647 	    !OK_32PTR(ipha)) {
4648 		mblk_t	*mp1;
4649 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra,
4650 		    BPRI_LO))) {
4651 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4652 			return (ENOMEM);
4653 		}
4654 		mp1->b_cont = mp;
4655 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4656 		mp1->b_wptr = (uchar_t *)ipha;
4657 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4658 		mp = mp1;
4659 	}
4660 #ifdef	_BIG_ENDIAN
4661 	/* Set version, header length, and tos */
4662 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4663 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4664 	    icmp->icmp_type_of_service);
4665 	/* Set ttl and protocol */
4666 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4667 #else
4668 	/* Set version, header length, and tos */
4669 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4670 	    ((icmp->icmp_type_of_service << 8) |
4671 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4672 	/* Set ttl and protocol */
4673 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4674 #endif
4675 	if (pktinfop->ip4_addr != INADDR_ANY) {
4676 		ipha->ipha_src = pktinfop->ip4_addr;
4677 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4678 	} else {
4679 
4680 		/*
4681 		 * Copy our address into the packet.  If this is zero,
4682 		 * ip will fill in the real source address.
4683 		 */
4684 		IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4685 	}
4686 
4687 	ipha->ipha_fragment_offset_and_flags = 0;
4688 
4689 	if (pktinfop->ip4_ill_index != 0) {
4690 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4691 	}
4692 
4693 
4694 	/*
4695 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4696 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4697 	 * tell IP that the application has sent a complete IP header and not
4698 	 * to compute the transport checksum nor change the DF flag.
4699 	 */
4700 	ipha->ipha_ident = IP_HDR_INCLUDED;
4701 
4702 	/* Finish common formatting of the packet. */
4703 	mp->b_rptr = (uchar_t *)ipha;
4704 
4705 	ip_len = mp->b_wptr - (uchar_t *)ipha;
4706 	if (mp->b_cont != NULL)
4707 		ip_len += msgdsize(mp->b_cont);
4708 
4709 	/*
4710 	 * Set the length into the IP header.
4711 	 * If the length is greater than the maximum allowed by IP,
4712 	 * then free the message and return. Do not try and send it
4713 	 * as this can cause problems in layers below.
4714 	 */
4715 	if (ip_len > IP_MAXPACKET) {
4716 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4717 		return (EMSGSIZE);
4718 	}
4719 	ipha->ipha_length = htons((uint16_t)ip_len);
4720 	/*
4721 	 * Copy in the destination address request
4722 	 */
4723 	ipha->ipha_dst = v4dst;
4724 
4725 	/*
4726 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4727 	 */
4728 	if (CLASSD(v4dst))
4729 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4730 
4731 	/* Copy in options if any */
4732 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4733 		bcopy(icmp->icmp_ip_snd_options,
4734 		    &ipha[1], icmp->icmp_ip_snd_options_len);
4735 		/*
4736 		 * Massage source route putting first source route in ipha_dst.
4737 		 * Ignore the destination in the T_unitdata_req.
4738 		 */
4739 		(void) ip_massage_options(ipha, is->is_netstack);
4740 	}
4741 
4742 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
4743 	mblk_setcred(mp, connp->conn_cred);
4744 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4745 	return (0);
4746 }
4747 
4748 static int
4749 icmp_update_label_v6(icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4750 {
4751 	int err;
4752 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4753 	icmp_stack_t		*is = icmp->icmp_is;
4754 	conn_t	*connp = icmp->icmp_connp;
4755 
4756 	err = tsol_compute_label_v6(DB_CREDDEF(mp, connp->conn_cred), dst,
4757 	    opt_storage, connp->conn_mac_exempt,
4758 	    is->is_netstack->netstack_ip);
4759 	if (err == 0) {
4760 		err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4761 		    &icmp->icmp_label_len_v6, opt_storage);
4762 	}
4763 	if (err != 0) {
4764 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4765 		DTRACE_PROBE4(
4766 		    tx__ip__log__drop__updatelabel__icmp6,
4767 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4768 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4769 		return (err);
4770 	}
4771 
4772 	icmp->icmp_v6lastdst = *dst;
4773 	return (0);
4774 }
4775 
4776 /*
4777  * raw_ip_send_data_v6():
4778  * Assumes that icmp_wput did some sanity checking on the destination
4779  * address, but that the label may not yet be correct.
4780  */
4781 static int
4782 raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp, sin6_t *sin6,
4783     ip6_pkt_t *ipp)
4784 {
4785 	ip6_t			*ip6h;
4786 	ip6i_t			*ip6i;	/* mp->b_rptr even if no ip6i_t */
4787 	int			ip_hdr_len = IPV6_HDR_LEN;
4788 	size_t			ip_len;
4789 	icmp_t			*icmp = connp->conn_icmp;
4790 	icmp_stack_t		*is = icmp->icmp_is;
4791 	ip6_pkt_t		*tipp;
4792 	uint32_t		csum = 0;
4793 	uint_t			ignore = 0;
4794 	uint_t			option_exists = 0, is_sticky = 0;
4795 	uint8_t			*cp;
4796 	uint8_t			*nxthdr_ptr;
4797 	in6_addr_t		ip6_dst;
4798 
4799 	/*
4800 	 * If the local address is a mapped address return
4801 	 * an error.
4802 	 * It would be possible to send an IPv6 packet but the
4803 	 * response would never make it back to the application
4804 	 * since it is bound to a mapped address.
4805 	 */
4806 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4807 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4808 		return (EADDRNOTAVAIL);
4809 	}
4810 
4811 	ignore = ipp->ipp_sticky_ignored;
4812 	if (sin6->sin6_scope_id != 0 &&
4813 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4814 		/*
4815 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4816 		 * option nor ancillary data.  It needs to be
4817 		 * explicitly set in options_exists.
4818 		 */
4819 		option_exists |= IPPF_SCOPE_ID;
4820 	}
4821 
4822 	/*
4823 	 * Compute the destination address
4824 	 */
4825 	ip6_dst = sin6->sin6_addr;
4826 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
4827 		ip6_dst = ipv6_loopback;
4828 
4829 	/*
4830 	 * If we're not going to the same destination as last time, then
4831 	 * recompute the label required.  This is done in a separate routine to
4832 	 * avoid blowing up our stack here.
4833 	 */
4834 	if (is_system_labeled() &&
4835 	    !IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst)) {
4836 		int error = 0;
4837 
4838 		error = icmp_update_label_v6(icmp, mp, &ip6_dst);
4839 		if (error != 0)
4840 			return (error);
4841 	}
4842 
4843 	/*
4844 	 * If there's a security label here, then we ignore any options the
4845 	 * user may try to set.  We keep the peer's label as a hidden sticky
4846 	 * option.
4847 	 */
4848 	if (icmp->icmp_label_len_v6 > 0) {
4849 		ignore &= ~IPPF_HOPOPTS;
4850 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
4851 	}
4852 
4853 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
4854 	    (ipp->ipp_fields == 0)) {
4855 		/* No sticky options nor ancillary data. */
4856 		goto no_options;
4857 	}
4858 
4859 	/*
4860 	 * Go through the options figuring out where each is going to
4861 	 * come from and build two masks.  The first mask indicates if
4862 	 * the option exists at all.  The second mask indicates if the
4863 	 * option is sticky or ancillary.
4864 	 */
4865 	if (!(ignore & IPPF_HOPOPTS)) {
4866 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
4867 			option_exists |= IPPF_HOPOPTS;
4868 			ip_hdr_len += ipp->ipp_hopoptslen;
4869 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
4870 			option_exists |= IPPF_HOPOPTS;
4871 			is_sticky |= IPPF_HOPOPTS;
4872 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen;
4873 		}
4874 	}
4875 
4876 	if (!(ignore & IPPF_RTHDR)) {
4877 		if (ipp->ipp_fields & IPPF_RTHDR) {
4878 			option_exists |= IPPF_RTHDR;
4879 			ip_hdr_len += ipp->ipp_rthdrlen;
4880 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
4881 			option_exists |= IPPF_RTHDR;
4882 			is_sticky |= IPPF_RTHDR;
4883 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
4884 		}
4885 	}
4886 
4887 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
4888 		/*
4889 		 * Need to have a router header to use these.
4890 		 */
4891 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
4892 			option_exists |= IPPF_RTDSTOPTS;
4893 			ip_hdr_len += ipp->ipp_rtdstoptslen;
4894 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
4895 			option_exists |= IPPF_RTDSTOPTS;
4896 			is_sticky |= IPPF_RTDSTOPTS;
4897 			ip_hdr_len +=
4898 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
4899 		}
4900 	}
4901 
4902 	if (!(ignore & IPPF_DSTOPTS)) {
4903 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
4904 			option_exists |= IPPF_DSTOPTS;
4905 			ip_hdr_len += ipp->ipp_dstoptslen;
4906 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
4907 			option_exists |= IPPF_DSTOPTS;
4908 			is_sticky |= IPPF_DSTOPTS;
4909 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
4910 		}
4911 	}
4912 
4913 	if (!(ignore & IPPF_IFINDEX)) {
4914 		if (ipp->ipp_fields & IPPF_IFINDEX) {
4915 			option_exists |= IPPF_IFINDEX;
4916 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
4917 			option_exists |= IPPF_IFINDEX;
4918 			is_sticky |= IPPF_IFINDEX;
4919 		}
4920 	}
4921 
4922 	if (!(ignore & IPPF_ADDR)) {
4923 		if (ipp->ipp_fields & IPPF_ADDR) {
4924 			option_exists |= IPPF_ADDR;
4925 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
4926 			option_exists |= IPPF_ADDR;
4927 			is_sticky |= IPPF_ADDR;
4928 		}
4929 	}
4930 
4931 	if (!(ignore & IPPF_DONTFRAG)) {
4932 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
4933 			option_exists |= IPPF_DONTFRAG;
4934 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
4935 			option_exists |= IPPF_DONTFRAG;
4936 			is_sticky |= IPPF_DONTFRAG;
4937 		}
4938 	}
4939 
4940 	if (!(ignore & IPPF_USE_MIN_MTU)) {
4941 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
4942 			option_exists |= IPPF_USE_MIN_MTU;
4943 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
4944 		    IPPF_USE_MIN_MTU) {
4945 			option_exists |= IPPF_USE_MIN_MTU;
4946 			is_sticky |= IPPF_USE_MIN_MTU;
4947 		}
4948 	}
4949 
4950 	if (!(ignore & IPPF_NEXTHOP)) {
4951 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
4952 			option_exists |= IPPF_NEXTHOP;
4953 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
4954 			option_exists |= IPPF_NEXTHOP;
4955 			is_sticky |= IPPF_NEXTHOP;
4956 		}
4957 	}
4958 
4959 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
4960 		option_exists |= IPPF_HOPLIMIT;
4961 	/* IPV6_HOPLIMIT can never be sticky */
4962 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
4963 
4964 	if (!(ignore & IPPF_UNICAST_HOPS) &&
4965 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
4966 		option_exists |= IPPF_UNICAST_HOPS;
4967 		is_sticky |= IPPF_UNICAST_HOPS;
4968 	}
4969 
4970 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
4971 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
4972 		option_exists |= IPPF_MULTICAST_HOPS;
4973 		is_sticky |= IPPF_MULTICAST_HOPS;
4974 	}
4975 
4976 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
4977 		/* This is a sticky socket option only */
4978 		option_exists |= IPPF_NO_CKSUM;
4979 		is_sticky |= IPPF_NO_CKSUM;
4980 	}
4981 
4982 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
4983 		/* This is a sticky socket option only */
4984 		option_exists |= IPPF_RAW_CKSUM;
4985 		is_sticky |= IPPF_RAW_CKSUM;
4986 	}
4987 
4988 	if (!(ignore & IPPF_TCLASS)) {
4989 		if (ipp->ipp_fields & IPPF_TCLASS) {
4990 			option_exists |= IPPF_TCLASS;
4991 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
4992 			option_exists |= IPPF_TCLASS;
4993 			is_sticky |= IPPF_TCLASS;
4994 		}
4995 	}
4996 
4997 no_options:
4998 
4999 	/*
5000 	 * If any options carried in the ip6i_t were specified, we
5001 	 * need to account for the ip6i_t in the data we'll be sending
5002 	 * down.
5003 	 */
5004 	if (option_exists & IPPF_HAS_IP6I)
5005 		ip_hdr_len += sizeof (ip6i_t);
5006 
5007 	/* check/fix buffer config, setup pointers into it */
5008 	ip6h = (ip6_t *)&mp->b_rptr[-ip_hdr_len];
5009 	if ((mp->b_datap->db_ref != 1) ||
5010 	    ((unsigned char *)ip6h < mp->b_datap->db_base) ||
5011 	    !OK_32PTR(ip6h)) {
5012 		mblk_t	*mp1;
5013 
5014 		/* Try to get everything in a single mblk next time */
5015 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
5016 			icmp->icmp_max_hdr_len = ip_hdr_len;
5017 
5018 			(void) proto_set_tx_wroff(q == NULL ? NULL:RD(q), connp,
5019 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
5020 		}
5021 		mp1 = allocb(ip_hdr_len + is->is_wroff_extra, BPRI_LO);
5022 		if (!mp1) {
5023 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5024 			return (ENOMEM);
5025 		}
5026 		mp1->b_cont = mp;
5027 		mp1->b_wptr = mp1->b_datap->db_lim;
5028 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
5029 		mp = mp1;
5030 	}
5031 	mp->b_rptr = (unsigned char *)ip6h;
5032 	ip6i = (ip6i_t *)ip6h;
5033 
5034 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
5035 	if (option_exists & IPPF_HAS_IP6I) {
5036 		ip6h = (ip6_t *)&ip6i[1];
5037 		ip6i->ip6i_flags = 0;
5038 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5039 
5040 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
5041 		if (option_exists & IPPF_SCOPE_ID) {
5042 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5043 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
5044 		} else if (option_exists & IPPF_IFINDEX) {
5045 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
5046 			ASSERT(tipp->ipp_ifindex != 0);
5047 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5048 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
5049 		}
5050 
5051 		if (option_exists & IPPF_RAW_CKSUM) {
5052 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
5053 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
5054 		}
5055 
5056 		if (option_exists & IPPF_NO_CKSUM) {
5057 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
5058 		}
5059 
5060 		if (option_exists & IPPF_ADDR) {
5061 			/*
5062 			 * Enable per-packet source address verification if
5063 			 * IPV6_PKTINFO specified the source address.
5064 			 * ip6_src is set in the transport's _wput function.
5065 			 */
5066 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
5067 		}
5068 
5069 		if (option_exists & IPPF_DONTFRAG) {
5070 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
5071 		}
5072 
5073 		if (option_exists & IPPF_USE_MIN_MTU) {
5074 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
5075 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
5076 		}
5077 
5078 		if (option_exists & IPPF_NEXTHOP) {
5079 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
5080 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
5081 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
5082 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
5083 		}
5084 
5085 		/*
5086 		 * tell IP this is an ip6i_t private header
5087 		 */
5088 		ip6i->ip6i_nxt = IPPROTO_RAW;
5089 	}
5090 
5091 	/* Initialize IPv6 header */
5092 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5093 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
5094 
5095 	/* Set the hoplimit of the outgoing packet. */
5096 	if (option_exists & IPPF_HOPLIMIT) {
5097 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
5098 		ip6h->ip6_hops = ipp->ipp_hoplimit;
5099 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5100 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
5101 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
5102 		if (option_exists & IPPF_MULTICAST_HOPS)
5103 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5104 	} else {
5105 		ip6h->ip6_hops = icmp->icmp_ttl;
5106 		if (option_exists & IPPF_UNICAST_HOPS)
5107 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5108 	}
5109 
5110 	if (option_exists & IPPF_ADDR) {
5111 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
5112 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
5113 		ip6h->ip6_src = tipp->ipp_addr;
5114 	} else {
5115 		/*
5116 		 * The source address was not set using IPV6_PKTINFO.
5117 		 * First look at the bound source.
5118 		 * If unspecified fallback to __sin6_src_id.
5119 		 */
5120 		ip6h->ip6_src = icmp->icmp_v6src;
5121 		if (sin6->__sin6_src_id != 0 &&
5122 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
5123 			ip_srcid_find_id(sin6->__sin6_src_id,
5124 			    &ip6h->ip6_src, icmp->icmp_zoneid,
5125 			    is->is_netstack);
5126 		}
5127 	}
5128 
5129 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
5130 	cp = (uint8_t *)&ip6h[1];
5131 
5132 	/*
5133 	 * Here's where we have to start stringing together
5134 	 * any extension headers in the right order:
5135 	 * Hop-by-hop, destination, routing, and final destination opts.
5136 	 */
5137 	if (option_exists & IPPF_HOPOPTS) {
5138 		/* Hop-by-hop options */
5139 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
5140 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
5141 
5142 		*nxthdr_ptr = IPPROTO_HOPOPTS;
5143 		nxthdr_ptr = &hbh->ip6h_nxt;
5144 
5145 		bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
5146 		cp += tipp->ipp_hopoptslen;
5147 	}
5148 	/*
5149 	 * En-route destination options
5150 	 * Only do them if there's a routing header as well
5151 	 */
5152 	if (option_exists & IPPF_RTDSTOPTS) {
5153 		ip6_dest_t *dst = (ip6_dest_t *)cp;
5154 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
5155 
5156 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5157 		nxthdr_ptr = &dst->ip6d_nxt;
5158 
5159 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
5160 		cp += tipp->ipp_rtdstoptslen;
5161 	}
5162 	/*
5163 	 * Routing header next
5164 	 */
5165 	if (option_exists & IPPF_RTHDR) {
5166 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
5167 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
5168 
5169 		*nxthdr_ptr = IPPROTO_ROUTING;
5170 		nxthdr_ptr = &rt->ip6r_nxt;
5171 
5172 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
5173 		cp += tipp->ipp_rthdrlen;
5174 	}
5175 	/*
5176 	 * Do ultimate destination options
5177 	 */
5178 	if (option_exists & IPPF_DSTOPTS) {
5179 		ip6_dest_t *dest = (ip6_dest_t *)cp;
5180 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
5181 
5182 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5183 		nxthdr_ptr = &dest->ip6d_nxt;
5184 
5185 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
5186 		cp += tipp->ipp_dstoptslen;
5187 	}
5188 
5189 	/*
5190 	 * Now set the last header pointer to the proto passed in
5191 	 */
5192 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
5193 	*nxthdr_ptr = icmp->icmp_proto;
5194 
5195 	/*
5196 	 * Copy in the destination address
5197 	 */
5198 	ip6h->ip6_dst = ip6_dst;
5199 
5200 	ip6h->ip6_vcf =
5201 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
5202 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
5203 
5204 	if (option_exists & IPPF_TCLASS) {
5205 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
5206 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
5207 		    tipp->ipp_tclass);
5208 	}
5209 	if (option_exists & IPPF_RTHDR) {
5210 		ip6_rthdr_t	*rth;
5211 
5212 		/*
5213 		 * Perform any processing needed for source routing.
5214 		 * We know that all extension headers will be in the same mblk
5215 		 * as the IPv6 header.
5216 		 */
5217 		rth = ip_find_rthdr_v6(ip6h, mp->b_wptr);
5218 		if (rth != NULL && rth->ip6r_segleft != 0) {
5219 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
5220 				/*
5221 				 * Drop packet - only support Type 0 routing.
5222 				 * Notify the application as well.
5223 				 */
5224 				BUMP_MIB(&is->is_rawip_mib,
5225 				    rawipOutErrors);
5226 				return (EPROTO);
5227 			}
5228 			/*
5229 			 * rth->ip6r_len is twice the number of
5230 			 * addresses in the header
5231 			 */
5232 			if (rth->ip6r_len & 0x1) {
5233 				BUMP_MIB(&is->is_rawip_mib,
5234 				    rawipOutErrors);
5235 				return (EPROTO);
5236 			}
5237 			/*
5238 			 * Shuffle the routing header and ip6_dst
5239 			 * addresses, and get the checksum difference
5240 			 * between the first hop (in ip6_dst) and
5241 			 * the destination (in the last routing hdr entry).
5242 			 */
5243 			csum = ip_massage_options_v6(ip6h, rth,
5244 			    is->is_netstack);
5245 			/*
5246 			 * Verify that the first hop isn't a mapped address.
5247 			 * Routers along the path need to do this verification
5248 			 * for subsequent hops.
5249 			 */
5250 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
5251 				BUMP_MIB(&is->is_rawip_mib,
5252 				    rawipOutErrors);
5253 				return (EADDRNOTAVAIL);
5254 			}
5255 		}
5256 	}
5257 
5258 	ip_len = mp->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
5259 	if (mp->b_cont != NULL)
5260 		ip_len += msgdsize(mp->b_cont);
5261 
5262 	/*
5263 	 * Set the length into the IP header.
5264 	 * If the length is greater than the maximum allowed by IP,
5265 	 * then free the message and return. Do not try and send it
5266 	 * as this can cause problems in layers below.
5267 	 */
5268 	if (ip_len > IP_MAXPACKET) {
5269 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5270 		return (EMSGSIZE);
5271 	}
5272 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
5273 		uint_t	cksum_off;	/* From ip6i == mp->b_rptr */
5274 		uint16_t *cksum_ptr;
5275 		uint_t	ext_hdrs_len;
5276 
5277 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
5278 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
5279 		    icmp->icmp_checksum_off == 2);
5280 
5281 		/*
5282 		 * We make it easy for IP to include our pseudo header
5283 		 * by putting our length in uh_checksum, modified (if
5284 		 * we have a routing header) by the checksum difference
5285 		 * between the ultimate destination and first hop addresses.
5286 		 * Note: ICMPv6 must always checksum the packet.
5287 		 */
5288 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
5289 		if (cksum_off + sizeof (uint16_t) > mp->b_wptr - mp->b_rptr) {
5290 			if (!pullupmsg(mp, cksum_off + sizeof (uint16_t))) {
5291 				BUMP_MIB(&is->is_rawip_mib,
5292 				    rawipOutErrors);
5293 				freemsg(mp);
5294 				return (0);
5295 			}
5296 			ip6i = (ip6i_t *)mp->b_rptr;
5297 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
5298 				ip6h = (ip6_t *)&ip6i[1];
5299 			else
5300 				ip6h = (ip6_t *)ip6i;
5301 		}
5302 		/* Add payload length to checksum */
5303 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
5304 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
5305 		csum += htons(ip_len - ext_hdrs_len);
5306 
5307 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
5308 		csum = (csum & 0xFFFF) + (csum >> 16);
5309 		*cksum_ptr = (uint16_t)csum;
5310 	}
5311 
5312 #ifdef _LITTLE_ENDIAN
5313 	ip_len = htons(ip_len);
5314 #endif
5315 	ip6h->ip6_plen = (uint16_t)ip_len;
5316 
5317 	/* We're done. Pass the packet to IP */
5318 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
5319 	ip_output_v6(icmp->icmp_connp, mp, q, IP_WPUT);
5320 	return (0);
5321 }
5322 
5323 static void
5324 icmp_wput_other(queue_t *q, mblk_t *mp)
5325 {
5326 	uchar_t	*rptr = mp->b_rptr;
5327 	struct iocblk *iocp;
5328 #define	tudr ((struct T_unitdata_req *)rptr)
5329 	conn_t	*connp = Q_TO_CONN(q);
5330 	icmp_t	*icmp = connp->conn_icmp;
5331 	icmp_stack_t *is = icmp->icmp_is;
5332 	cred_t *cr;
5333 
5334 	cr = DB_CREDDEF(mp, connp->conn_cred);
5335 
5336 	switch (mp->b_datap->db_type) {
5337 	case M_PROTO:
5338 	case M_PCPROTO:
5339 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5340 			/*
5341 			 * If the message does not contain a PRIM_type,
5342 			 * throw it away.
5343 			 */
5344 			freemsg(mp);
5345 			return;
5346 		}
5347 		switch (((union T_primitives *)rptr)->type) {
5348 		case T_ADDR_REQ:
5349 			icmp_addr_req(q, mp);
5350 			return;
5351 		case O_T_BIND_REQ:
5352 		case T_BIND_REQ:
5353 			icmp_tpi_bind(q, mp);
5354 			return;
5355 		case T_CONN_REQ:
5356 			icmp_tpi_connect(q, mp);
5357 			return;
5358 		case T_CAPABILITY_REQ:
5359 			icmp_capability_req(q, mp);
5360 			return;
5361 		case T_INFO_REQ:
5362 			icmp_info_req(q, mp);
5363 			return;
5364 		case T_UNITDATA_REQ:
5365 			/*
5366 			 * If a T_UNITDATA_REQ gets here, the address must
5367 			 * be bad.  Valid T_UNITDATA_REQs are found above
5368 			 * and break to below this switch.
5369 			 */
5370 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
5371 			return;
5372 		case T_UNBIND_REQ:
5373 			icmp_tpi_unbind(q, mp);
5374 			return;
5375 
5376 		case T_SVR4_OPTMGMT_REQ:
5377 			if (!snmpcom_req(q, mp, icmp_snmp_set, ip_snmp_get,
5378 			    cr)) {
5379 				/* Only IP can return anything meaningful */
5380 				(void) svr4_optcom_req(q, mp, cr,
5381 				    &icmp_opt_obj, B_TRUE);
5382 			}
5383 			return;
5384 
5385 		case T_OPTMGMT_REQ:
5386 			/* Only IP can return anything meaningful */
5387 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj, B_TRUE);
5388 			return;
5389 
5390 		case T_DISCON_REQ:
5391 			icmp_tpi_disconnect(q, mp);
5392 			return;
5393 
5394 		/* The following TPI message is not supported by icmp. */
5395 		case O_T_CONN_RES:
5396 		case T_CONN_RES:
5397 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
5398 			return;
5399 
5400 		/* The following 3 TPI requests are illegal for icmp. */
5401 		case T_DATA_REQ:
5402 		case T_EXDATA_REQ:
5403 		case T_ORDREL_REQ:
5404 			freemsg(mp);
5405 			(void) putctl1(RD(q), M_ERROR, EPROTO);
5406 			return;
5407 		default:
5408 			break;
5409 		}
5410 		break;
5411 	case M_IOCTL:
5412 		iocp = (struct iocblk *)mp->b_rptr;
5413 		switch (iocp->ioc_cmd) {
5414 		case TI_GETPEERNAME:
5415 			if (icmp->icmp_state != TS_DATA_XFER) {
5416 				/*
5417 				 * If a default destination address has not
5418 				 * been associated with the stream, then we
5419 				 * don't know the peer's name.
5420 				 */
5421 				iocp->ioc_error = ENOTCONN;
5422 		err_ret:;
5423 				iocp->ioc_count = 0;
5424 				mp->b_datap->db_type = M_IOCACK;
5425 				qreply(q, mp);
5426 				return;
5427 			}
5428 			/* FALLTHRU */
5429 		case TI_GETMYNAME:
5430 			/*
5431 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
5432 			 * need to copyin the user's strbuf structure.
5433 			 * Processing will continue in the M_IOCDATA case
5434 			 * below.
5435 			 */
5436 			mi_copyin(q, mp, NULL,
5437 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
5438 			return;
5439 		case ND_SET:
5440 			/* nd_getset performs the necessary error checking */
5441 		case ND_GET:
5442 			if (nd_getset(q, is->is_nd, mp)) {
5443 				qreply(q, mp);
5444 				return;
5445 			}
5446 			break;
5447 		case _SIOCSOCKFALLBACK:
5448 			/*
5449 			 * socket is falling back to be a
5450 			 * streams socket. Nothing  to do
5451 			 */
5452 			iocp->ioc_count = 0;
5453 			iocp->ioc_rval = 0;
5454 			qreply(q, mp);
5455 			return;
5456 		default:
5457 			break;
5458 		}
5459 		break;
5460 	case M_IOCDATA:
5461 		icmp_wput_iocdata(q, mp);
5462 		return;
5463 	default:
5464 		break;
5465 	}
5466 	ip_wput(q, mp);
5467 }
5468 
5469 /*
5470  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
5471  * messages.
5472  */
5473 static void
5474 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
5475 {
5476 	mblk_t	*mp1;
5477 	STRUCT_HANDLE(strbuf, sb);
5478 	icmp_t	*icmp;
5479 	uint_t	addrlen;
5480 	uint_t	error;
5481 
5482 	/* Make sure it is one of ours. */
5483 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5484 	case TI_GETMYNAME:
5485 	case TI_GETPEERNAME:
5486 		break;
5487 	default:
5488 		icmp = Q_TO_ICMP(q);
5489 		ip_output(icmp->icmp_connp, mp, q, IP_WPUT);
5490 		return;
5491 	}
5492 	switch (mi_copy_state(q, mp, &mp1)) {
5493 	case -1:
5494 		return;
5495 	case MI_COPY_CASE(MI_COPY_IN, 1):
5496 		break;
5497 	case MI_COPY_CASE(MI_COPY_OUT, 1):
5498 		/*
5499 		 * The address has been copied out, so now
5500 		 * copyout the strbuf.
5501 		 */
5502 		mi_copyout(q, mp);
5503 		return;
5504 	case MI_COPY_CASE(MI_COPY_OUT, 2):
5505 		/*
5506 		 * The address and strbuf have been copied out.
5507 		 * We're done, so just acknowledge the original
5508 		 * M_IOCTL.
5509 		 */
5510 		mi_copy_done(q, mp, 0);
5511 		return;
5512 	default:
5513 		/*
5514 		 * Something strange has happened, so acknowledge
5515 		 * the original M_IOCTL with an EPROTO error.
5516 		 */
5517 		mi_copy_done(q, mp, EPROTO);
5518 		return;
5519 	}
5520 	/*
5521 	 * Now we have the strbuf structure for TI_GETMYNAME
5522 	 * and TI_GETPEERNAME.  Next we copyout the requested
5523 	 * address and then we'll copyout the strbuf.
5524 	 */
5525 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5526 	    (void *)mp1->b_rptr);
5527 	icmp = Q_TO_ICMP(q);
5528 	if (icmp->icmp_family == AF_INET)
5529 		addrlen = sizeof (sin_t);
5530 	else
5531 		addrlen = sizeof (sin6_t);
5532 
5533 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5534 		mi_copy_done(q, mp, EINVAL);
5535 		return;
5536 	}
5537 
5538 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5539 
5540 	if (mp1 == NULL)
5541 		return;
5542 
5543 	rw_enter(&icmp->icmp_rwlock, RW_READER);
5544 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5545 	case TI_GETMYNAME:
5546 		error = rawip_do_getsockname(icmp, (void *)mp1->b_rptr,
5547 		    &addrlen);
5548 		break;
5549 	case TI_GETPEERNAME:
5550 		error = rawip_do_getpeername(icmp, (void *)mp1->b_rptr,
5551 		    &addrlen);
5552 		break;
5553 	}
5554 	rw_exit(&icmp->icmp_rwlock);
5555 
5556 	if (error != 0) {
5557 		mi_copy_done(q, mp, error);
5558 	} else {
5559 		mp1->b_wptr += addrlen;
5560 		STRUCT_FSET(sb, len, addrlen);
5561 
5562 		/* Copy out the address */
5563 		mi_copyout(q, mp);
5564 	}
5565 }
5566 
5567 static int
5568 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5569     void *thisdg_attrs)
5570 {
5571 	conn_t	*connp = Q_TO_CONN(q);
5572 	struct T_unitdata_req *udreqp;
5573 	int is_absreq_failure;
5574 	cred_t *cr;
5575 
5576 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5577 	*errorp = 0;
5578 
5579 	cr = DB_CREDDEF(mp, connp->conn_cred);
5580 
5581 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5582 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5583 	    thisdg_attrs, &is_absreq_failure);
5584 
5585 	if (*errorp != 0) {
5586 		/*
5587 		 * Note: No special action needed in this
5588 		 * module for "is_absreq_failure"
5589 		 */
5590 		return (-1);		/* failure */
5591 	}
5592 	ASSERT(is_absreq_failure == 0);
5593 	return (0);	/* success */
5594 }
5595 
5596 void
5597 icmp_ddi_g_init(void)
5598 {
5599 	icmp_max_optsize = optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5600 	    icmp_opt_obj.odb_opt_arr_cnt);
5601 
5602 	/*
5603 	 * We want to be informed each time a stack is created or
5604 	 * destroyed in the kernel, so we can maintain the
5605 	 * set of icmp_stack_t's.
5606 	 */
5607 	netstack_register(NS_ICMP, rawip_stack_init, NULL, rawip_stack_fini);
5608 }
5609 
5610 void
5611 icmp_ddi_g_destroy(void)
5612 {
5613 	netstack_unregister(NS_ICMP);
5614 }
5615 
5616 #define	INET_NAME	"ip"
5617 
5618 /*
5619  * Initialize the ICMP stack instance.
5620  */
5621 static void *
5622 rawip_stack_init(netstackid_t stackid, netstack_t *ns)
5623 {
5624 	icmp_stack_t	*is;
5625 	icmpparam_t	*pa;
5626 	int		error = 0;
5627 	major_t		major;
5628 
5629 	is = (icmp_stack_t *)kmem_zalloc(sizeof (*is), KM_SLEEP);
5630 	is->is_netstack = ns;
5631 
5632 	pa = (icmpparam_t *)kmem_alloc(sizeof (icmp_param_arr), KM_SLEEP);
5633 	is->is_param_arr = pa;
5634 	bcopy(icmp_param_arr, is->is_param_arr, sizeof (icmp_param_arr));
5635 
5636 	(void) icmp_param_register(&is->is_nd,
5637 	    is->is_param_arr, A_CNT(icmp_param_arr));
5638 	is->is_ksp = rawip_kstat_init(stackid);
5639 
5640 	major = mod_name_to_major(INET_NAME);
5641 	error = ldi_ident_from_major(major, &is->is_ldi_ident);
5642 	ASSERT(error == 0);
5643 	return (is);
5644 }
5645 
5646 /*
5647  * Free the ICMP stack instance.
5648  */
5649 static void
5650 rawip_stack_fini(netstackid_t stackid, void *arg)
5651 {
5652 	icmp_stack_t *is = (icmp_stack_t *)arg;
5653 
5654 	nd_free(&is->is_nd);
5655 	kmem_free(is->is_param_arr, sizeof (icmp_param_arr));
5656 	is->is_param_arr = NULL;
5657 
5658 	rawip_kstat_fini(stackid, is->is_ksp);
5659 	is->is_ksp = NULL;
5660 	ldi_ident_release(is->is_ldi_ident);
5661 	kmem_free(is, sizeof (*is));
5662 }
5663 
5664 static void *
5665 rawip_kstat_init(netstackid_t stackid) {
5666 	kstat_t	*ksp;
5667 
5668 	rawip_named_kstat_t template = {
5669 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5670 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5671 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5672 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5673 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5674 	};
5675 
5676 	ksp = kstat_create_netstack("icmp", 0, "rawip", "mib2",
5677 					KSTAT_TYPE_NAMED,
5678 					NUM_OF_FIELDS(rawip_named_kstat_t),
5679 					0, stackid);
5680 	if (ksp == NULL || ksp->ks_data == NULL)
5681 		return (NULL);
5682 
5683 	bcopy(&template, ksp->ks_data, sizeof (template));
5684 	ksp->ks_update = rawip_kstat_update;
5685 	ksp->ks_private = (void *)(uintptr_t)stackid;
5686 
5687 	kstat_install(ksp);
5688 	return (ksp);
5689 }
5690 
5691 static void
5692 rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
5693 {
5694 	if (ksp != NULL) {
5695 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
5696 		kstat_delete_netstack(ksp, stackid);
5697 	}
5698 }
5699 
5700 static int
5701 rawip_kstat_update(kstat_t *ksp, int rw)
5702 {
5703 	rawip_named_kstat_t *rawipkp;
5704 	netstackid_t	stackid = (netstackid_t)(uintptr_t)ksp->ks_private;
5705 	netstack_t	*ns;
5706 	icmp_stack_t	*is;
5707 
5708 	if ((ksp == NULL) || (ksp->ks_data == NULL))
5709 		return (EIO);
5710 
5711 	if (rw == KSTAT_WRITE)
5712 		return (EACCES);
5713 
5714 	rawipkp = (rawip_named_kstat_t *)ksp->ks_data;
5715 
5716 	ns = netstack_find_by_stackid(stackid);
5717 	if (ns == NULL)
5718 		return (-1);
5719 	is = ns->netstack_icmp;
5720 	if (is == NULL) {
5721 		netstack_rele(ns);
5722 		return (-1);
5723 	}
5724 	rawipkp->inDatagrams.value.ui32 =  is->is_rawip_mib.rawipInDatagrams;
5725 	rawipkp->inCksumErrs.value.ui32 =  is->is_rawip_mib.rawipInCksumErrs;
5726 	rawipkp->inErrors.value.ui32 =	   is->is_rawip_mib.rawipInErrors;
5727 	rawipkp->outDatagrams.value.ui32 = is->is_rawip_mib.rawipOutDatagrams;
5728 	rawipkp->outErrors.value.ui32 =	   is->is_rawip_mib.rawipOutErrors;
5729 	netstack_rele(ns);
5730 	return (0);
5731 }
5732 
5733 /* ARGSUSED */
5734 int
5735 rawip_accept(sock_lower_handle_t lproto_handle,
5736     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
5737     cred_t *cr)
5738 {
5739 	return (EOPNOTSUPP);
5740 }
5741 
5742 /* ARGSUSED */
5743 int
5744 rawip_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
5745     socklen_t len, cred_t *cr)
5746 {
5747 	conn_t  *connp = (conn_t *)proto_handle;
5748 	int error;
5749 
5750 	/* Binding to a NULL address really means unbind */
5751 	if (sa == NULL)
5752 		error = rawip_do_unbind(connp);
5753 	else
5754 		error = rawip_do_bind(connp, sa, len);
5755 
5756 	if (error < 0) {
5757 		if (error == -TOUTSTATE)
5758 			error = EINVAL;
5759 		else
5760 			error = proto_tlitosyserr(-error);
5761 	}
5762 	return (error);
5763 }
5764 
5765 static int
5766 rawip_implicit_bind(conn_t *connp)
5767 {
5768 	sin6_t sin6addr;
5769 	sin_t *sin;
5770 	sin6_t *sin6;
5771 	socklen_t len;
5772 	int error;
5773 
5774 	if (connp->conn_icmp->icmp_family == AF_INET) {
5775 		len = sizeof (struct sockaddr_in);
5776 		sin = (sin_t *)&sin6addr;
5777 		*sin = sin_null;
5778 		sin->sin_family = AF_INET;
5779 		sin->sin_addr.s_addr = INADDR_ANY;
5780 	} else {
5781 		ASSERT(connp->conn_icmp->icmp_family == AF_INET6);
5782 		len = sizeof (sin6_t);
5783 		sin6 = (sin6_t *)&sin6addr;
5784 		*sin6 = sin6_null;
5785 		sin6->sin6_family = AF_INET6;
5786 		V6_SET_ZERO(sin6->sin6_addr);
5787 	}
5788 
5789 	error = rawip_do_bind(connp, (struct sockaddr *)&sin6addr, len);
5790 
5791 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
5792 }
5793 
5794 static int
5795 rawip_unbind(conn_t *connp)
5796 {
5797 	int error;
5798 
5799 	error = rawip_do_unbind(connp);
5800 	if (error < 0) {
5801 		error = proto_tlitosyserr(-error);
5802 	}
5803 	return (error);
5804 }
5805 
5806 /* ARGSUSED */
5807 int
5808 rawip_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
5809 {
5810 	return (EOPNOTSUPP);
5811 }
5812 
5813 /* ARGSUSED */
5814 int
5815 rawip_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
5816     socklen_t len, sock_connid_t *id, cred_t *cr)
5817 {
5818 	conn_t	*connp = (conn_t *)proto_handle;
5819 	icmp_t *icmp = connp->conn_icmp;
5820 	int	error;
5821 	boolean_t did_bind = B_FALSE;
5822 
5823 	if (sa == NULL) {
5824 		/*
5825 		 * Disconnect
5826 		 * Make sure we are connected
5827 		 */
5828 		if (icmp->icmp_state != TS_DATA_XFER)
5829 			return (EINVAL);
5830 
5831 		error = icmp_disconnect(connp);
5832 		return (error);
5833 	}
5834 
5835 	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
5836 	if (error != 0)
5837 		return (error);
5838 
5839 	/* do an implicit bind if necessary */
5840 	if (icmp->icmp_state == TS_UNBND) {
5841 		error = rawip_implicit_bind(connp);
5842 		/*
5843 		 * We could be racing with an actual bind, in which case
5844 		 * we would see EPROTO. We cross our fingers and try
5845 		 * to connect.
5846 		 */
5847 		if (!(error == 0 || error == EPROTO))
5848 			return (error);
5849 		did_bind = B_TRUE;
5850 	}
5851 
5852 	/*
5853 	 * set SO_DGRAM_ERRIND
5854 	 */
5855 	icmp->icmp_dgram_errind = B_TRUE;
5856 
5857 	error = rawip_do_connect(connp, sa, len);
5858 
5859 	if (error != 0 && did_bind) {
5860 		int unbind_err;
5861 
5862 		unbind_err = rawip_unbind(connp);
5863 		ASSERT(unbind_err == 0);
5864 	}
5865 
5866 	if (error == 0) {
5867 		*id = 0;
5868 		(*connp->conn_upcalls->su_connected)
5869 		    (connp->conn_upper_handle, 0, NULL, -1);
5870 	} else if (error < 0) {
5871 		error = proto_tlitosyserr(-error);
5872 	}
5873 	return (error);
5874 }
5875 
5876 /* ARGSUSED */
5877 void
5878 rawip_fallback(sock_lower_handle_t proto_handle, queue_t *q,
5879     boolean_t direct_sockfs, so_proto_quiesced_cb_t quiesced_cb)
5880 {
5881 	conn_t  *connp = (conn_t *)proto_handle;
5882 	icmp_t	*icmp;
5883 	struct T_capability_ack tca;
5884 	struct sockaddr_in6 laddr, faddr;
5885 	socklen_t laddrlen, faddrlen;
5886 	short opts;
5887 	struct stroptions *stropt;
5888 	mblk_t *stropt_mp;
5889 	int error;
5890 
5891 	icmp = connp->conn_icmp;
5892 
5893 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
5894 
5895 	/*
5896 	 * setup the fallback stream that was allocated
5897 	 */
5898 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
5899 	connp->conn_minor_arena = WR(q)->q_ptr;
5900 
5901 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
5902 
5903 	WR(q)->q_qinfo = &icmpwinit;
5904 
5905 	connp->conn_rq = RD(q);
5906 	connp->conn_wq = WR(q);
5907 
5908 	/* Notify stream head about options before sending up data */
5909 	stropt_mp->b_datap->db_type = M_SETOPTS;
5910 	stropt_mp->b_wptr += sizeof (*stropt);
5911 	stropt = (struct stroptions *)stropt_mp->b_rptr;
5912 	stropt->so_flags = SO_WROFF | SO_HIWAT;
5913 	stropt->so_wroff =
5914 	    (ushort_t)(icmp->icmp_max_hdr_len + icmp->icmp_is->is_wroff_extra);
5915 	stropt->so_hiwat = icmp->icmp_recv_hiwat;
5916 	putnext(RD(q), stropt_mp);
5917 
5918 	/*
5919 	 * free helper stream
5920 	 */
5921 	ip_free_helper_stream(connp);
5922 
5923 	/*
5924 	 * Collect the information needed to sync with the sonode
5925 	 */
5926 	icmp_do_capability_ack(icmp, &tca, TC1_INFO);
5927 
5928 	laddrlen = faddrlen = sizeof (sin6_t);
5929 	(void) rawip_getsockname((sock_lower_handle_t)connp,
5930 	    (struct sockaddr *)&laddr, &laddrlen, NULL);
5931 	error = rawip_getpeername((sock_lower_handle_t)connp,
5932 	    (struct sockaddr *)&faddr, &faddrlen, NULL);
5933 	if (error != 0)
5934 		faddrlen = 0;
5935 	opts = 0;
5936 	if (icmp->icmp_dgram_errind)
5937 		opts |= SO_DGRAM_ERRIND;
5938 	if (icmp->icmp_dontroute)
5939 		opts |= SO_DONTROUTE;
5940 
5941 	/*
5942 	 * Once we grab the drain lock, no data will be send up
5943 	 * to the socket. So we notify the socket that the endpoint
5944 	 * is quiescent and it's therefore safe move data from
5945 	 * the socket to the stream head.
5946 	 */
5947 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
5948 	    (struct sockaddr *)&laddr, laddrlen,
5949 	    (struct sockaddr *)&faddr, faddrlen, opts);
5950 
5951 	/*
5952 	 * push up any packets that were queued in icmp_t
5953 	 */
5954 
5955 	mutex_enter(&icmp->icmp_recv_lock);
5956 	while (icmp->icmp_fallback_queue_head != NULL) {
5957 		mblk_t	*mp;
5958 
5959 		mp = icmp->icmp_fallback_queue_head;
5960 		icmp->icmp_fallback_queue_head = mp->b_next;
5961 		mp->b_next = NULL;
5962 		mutex_exit(&icmp->icmp_recv_lock);
5963 		putnext(RD(q), mp);
5964 		mutex_enter(&icmp->icmp_recv_lock);
5965 	}
5966 	icmp->icmp_fallback_queue_tail = icmp->icmp_fallback_queue_head;
5967 	/*
5968 	 * No longer a streams less socket
5969 	 */
5970 	connp->conn_flags &= ~IPCL_NONSTR;
5971 	mutex_exit(&icmp->icmp_recv_lock);
5972 	ASSERT(icmp->icmp_fallback_queue_head == NULL &&
5973 	    icmp->icmp_fallback_queue_tail == NULL);
5974 
5975 	ASSERT(connp->conn_ref >= 1);
5976 }
5977 
5978 /* ARGSUSED */
5979 sock_lower_handle_t
5980 rawip_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
5981     uint_t *smodep, int *errorp, int flags, cred_t *credp)
5982 {
5983 	conn_t *connp;
5984 
5985 	if (type != SOCK_RAW || (family != AF_INET && family != AF_INET6)) {
5986 		*errorp = EPROTONOSUPPORT;
5987 		return (NULL);
5988 	}
5989 
5990 	connp = icmp_open(family, credp, errorp, flags);
5991 	if (connp != NULL) {
5992 		icmp_stack_t *is;
5993 
5994 		is = connp->conn_icmp->icmp_is;
5995 		connp->conn_flags |= IPCL_NONSTR;
5996 
5997 		if (connp->conn_icmp->icmp_family == AF_INET6) {
5998 			/* Build initial header template for transmit */
5999 			rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
6000 			if ((*errorp =
6001 			    icmp_build_hdrs(connp->conn_icmp)) != 0) {
6002 				rw_exit(&connp->conn_icmp->icmp_rwlock);
6003 				ipcl_conn_destroy(connp);
6004 				return (NULL);
6005 			}
6006 			rw_exit(&connp->conn_icmp->icmp_rwlock);
6007 		}
6008 
6009 		connp->conn_icmp->icmp_recv_hiwat = is->is_recv_hiwat;
6010 		connp->conn_icmp->icmp_xmit_hiwat = is->is_xmit_hiwat;
6011 
6012 		if ((*errorp = ip_create_helper_stream(connp,
6013 		    is->is_ldi_ident)) != 0) {
6014 			cmn_err(CE_CONT, "create of IP helper stream failed\n");
6015 			(void) rawip_do_close(connp);
6016 			return (NULL);
6017 		}
6018 
6019 		mutex_enter(&connp->conn_lock);
6020 		connp->conn_state_flags &= ~CONN_INCIPIENT;
6021 		mutex_exit(&connp->conn_lock);
6022 		*sock_downcalls = &sock_rawip_downcalls;
6023 		*smodep = SM_ATOMIC;
6024 	} else {
6025 		ASSERT(*errorp != 0);
6026 	}
6027 
6028 	return ((sock_lower_handle_t)connp);
6029 }
6030 
6031 /* ARGSUSED */
6032 void
6033 rawip_activate(sock_lower_handle_t proto_handle,
6034     sock_upper_handle_t sock_handle, sock_upcalls_t *sock_upcalls, int flags,
6035     cred_t *cr)
6036 {
6037 	conn_t 			*connp = (conn_t *)proto_handle;
6038 	icmp_stack_t 		*is = connp->conn_icmp->icmp_is;
6039 	struct sock_proto_props sopp;
6040 
6041 	connp->conn_upcalls = sock_upcalls;
6042 	connp->conn_upper_handle = sock_handle;
6043 
6044 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT | SOCKOPT_RCVLOWAT |
6045 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
6046 	sopp.sopp_wroff = connp->conn_icmp->icmp_max_hdr_len +
6047 	    is->is_wroff_extra;
6048 	sopp.sopp_rxhiwat = is->is_recv_hiwat;
6049 	sopp.sopp_rxlowat = icmp_mod_info.mi_lowat;
6050 	sopp.sopp_maxblk = INFPSZ;
6051 	sopp.sopp_maxpsz = IP_MAXPACKET;
6052 	sopp.sopp_minpsz = (icmp_mod_info.mi_minpsz == 1) ? 0 :
6053 	    icmp_mod_info.mi_minpsz;
6054 
6055 	(*connp->conn_upcalls->su_set_proto_props)
6056 	    (connp->conn_upper_handle, &sopp);
6057 }
6058 
6059 static int
6060 rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6061 {
6062 	sin_t	*sin = (sin_t *)sa;
6063 	sin6_t	*sin6 = (sin6_t *)sa;
6064 
6065 	ASSERT(icmp != NULL);
6066 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6067 
6068 	switch (icmp->icmp_family) {
6069 	case AF_INET:
6070 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6071 		if (*salenp < sizeof (sin_t))
6072 			return (EINVAL);
6073 
6074 		*salenp = sizeof (sin_t);
6075 		*sin = sin_null;
6076 		sin->sin_family = AF_INET;
6077 		if (icmp->icmp_state == TS_UNBND) {
6078 			break;
6079 		}
6080 
6081 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
6082 		    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6083 			sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_v6src);
6084 		} else {
6085 			/*
6086 			 * INADDR_ANY
6087 			 * icmp_v6src is not set, we might be bound to
6088 			 * broadcast/multicast. Use icmp_bound_v6src as
6089 			 * local address instead (that could
6090 			 * also still be INADDR_ANY)
6091 			 */
6092 			sin->sin_addr.s_addr =
6093 			    V4_PART_OF_V6(icmp->icmp_bound_v6src);
6094 		}
6095 		break;
6096 	case AF_INET6:
6097 
6098 		if (*salenp < sizeof (sin6_t))
6099 			return (EINVAL);
6100 
6101 		*salenp = sizeof (sin6_t);
6102 		*sin6 = sin6_null;
6103 		sin6->sin6_family = AF_INET6;
6104 		if (icmp->icmp_state == TS_UNBND) {
6105 			break;
6106 		}
6107 		if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6108 			sin6->sin6_addr = icmp->icmp_v6src;
6109 		} else {
6110 			/*
6111 			 * UNSPECIFIED
6112 			 * icmp_v6src is not set, we might be bound to
6113 			 * broadcast/multicast. Use icmp_bound_v6src as
6114 			 * local address instead (that could
6115 			 * also still be UNSPECIFIED)
6116 			 */
6117 
6118 			sin6->sin6_addr = icmp->icmp_bound_v6src;
6119 		}
6120 		break;
6121 	}
6122 	return (0);
6123 }
6124 
6125 static int
6126 rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6127 {
6128 	sin_t   *sin = (sin_t *)sa;
6129 	sin6_t  *sin6 = (sin6_t *)sa;
6130 
6131 	ASSERT(icmp != NULL);
6132 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6133 
6134 	if (icmp->icmp_state != TS_DATA_XFER)
6135 		return (ENOTCONN);
6136 
6137 	sa->sa_family = icmp->icmp_family;
6138 	switch (icmp->icmp_family) {
6139 	case AF_INET:
6140 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6141 
6142 		if (*salenp < sizeof (sin_t))
6143 			return (EINVAL);
6144 
6145 		*salenp = sizeof (sin_t);
6146 		*sin = sin_null;
6147 		sin->sin_family = AF_INET;
6148 		sin->sin_addr.s_addr =
6149 		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6150 		break;
6151 	case AF_INET6:
6152 		if (*salenp < sizeof (sin6_t))
6153 			return (EINVAL);
6154 
6155 		*salenp = sizeof (sin6_t);
6156 		*sin6 = sin6_null;
6157 		*sin6 = icmp->icmp_v6dst;
6158 		break;
6159 	}
6160 	return (0);
6161 }
6162 
6163 /* ARGSUSED */
6164 int
6165 rawip_getpeername(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6166     socklen_t *salenp, cred_t *cr)
6167 {
6168 	conn_t  *connp = (conn_t *)proto_handle;
6169 	icmp_t  *icmp = connp->conn_icmp;
6170 	int	error;
6171 
6172 	ASSERT(icmp != NULL);
6173 
6174 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6175 
6176 	error = rawip_do_getpeername(icmp, sa, salenp);
6177 
6178 	rw_exit(&icmp->icmp_rwlock);
6179 
6180 	return (error);
6181 }
6182 
6183 /* ARGSUSED */
6184 int
6185 rawip_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6186     socklen_t *salenp, cred_t *cr)
6187 {
6188 	conn_t  *connp = (conn_t *)proto_handle;
6189 	icmp_t	*icmp = connp->conn_icmp;
6190 	int	error;
6191 
6192 	ASSERT(icmp != NULL);
6193 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6194 
6195 	error = rawip_do_getsockname(icmp, sa, salenp);
6196 
6197 	rw_exit(&icmp->icmp_rwlock);
6198 
6199 	return (error);
6200 }
6201 
6202 int
6203 rawip_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6204     const void *optvalp, socklen_t optlen, cred_t *cr)
6205 {
6206 	conn_t	*connp = (conn_t *)proto_handle;
6207 	icmp_t *icmp = connp->conn_icmp;
6208 	int error;
6209 
6210 	error = proto_opt_check(level, option_name, optlen, NULL,
6211 	    icmp_opt_obj.odb_opt_des_arr,
6212 	    icmp_opt_obj.odb_opt_arr_cnt,
6213 	    icmp_opt_obj.odb_topmost_tpiprovider,
6214 	    B_TRUE, B_FALSE, cr);
6215 
6216 	if (error != 0) {
6217 		/*
6218 		 * option not recognized
6219 		 */
6220 		if (error < 0) {
6221 			error = proto_tlitosyserr(-error);
6222 		}
6223 		return (error);
6224 	}
6225 
6226 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6227 	error = icmp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level,
6228 	    option_name, optlen, (uchar_t *)optvalp, (uint_t *)&optlen,
6229 	    (uchar_t *)optvalp, NULL, cr);
6230 	rw_exit(&icmp->icmp_rwlock);
6231 
6232 	if (error < 0) {
6233 		/*
6234 		 * Pass on to ip
6235 		 */
6236 		error = ip_set_options(connp, level, option_name, optvalp,
6237 		    optlen, cr);
6238 	}
6239 
6240 	ASSERT(error >= 0);
6241 
6242 	return (error);
6243 }
6244 
6245 int
6246 rawip_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6247     void *optvalp, socklen_t *optlen, cred_t *cr)
6248 {
6249 	int		error;
6250 	conn_t		*connp = (conn_t *)proto_handle;
6251 	icmp_t		*icmp = connp->conn_icmp;
6252 	t_uscalar_t	max_optbuf_len;
6253 	void		*optvalp_buf;
6254 	int		len;
6255 
6256 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
6257 	    icmp_opt_obj.odb_opt_des_arr,
6258 	    icmp_opt_obj.odb_opt_arr_cnt,
6259 	    icmp_opt_obj.odb_topmost_tpiprovider,
6260 	    B_FALSE, B_TRUE, cr);
6261 
6262 	if (error != 0) {
6263 		if (error < 0) {
6264 			error = proto_tlitosyserr(-error);
6265 		}
6266 		return (error);
6267 	}
6268 
6269 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
6270 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6271 	len = icmp_opt_get(connp, level, option_name, optvalp_buf);
6272 	rw_exit(&icmp->icmp_rwlock);
6273 
6274 	if (len < 0) {
6275 		/*
6276 		 * Pass on to IP
6277 		 */
6278 		kmem_free(optvalp_buf, max_optbuf_len);
6279 		return (ip_get_options(connp, level, option_name, optvalp,
6280 		    optlen, cr));
6281 	} else {
6282 		/*
6283 		 * update optlen and copy option value
6284 		 */
6285 		t_uscalar_t size = MIN(len, *optlen);
6286 		bcopy(optvalp_buf, optvalp, size);
6287 		bcopy(&size, optlen, sizeof (size));
6288 
6289 		kmem_free(optvalp_buf, max_optbuf_len);
6290 		return (0);
6291 	}
6292 }
6293 
6294 /* ARGSUSED */
6295 int
6296 rawip_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
6297 {
6298 	conn_t	*connp = (conn_t *)proto_handle;
6299 	(void) rawip_do_close(connp);
6300 	return (0);
6301 }
6302 
6303 /* ARGSUSED */
6304 int
6305 rawip_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
6306 {
6307 	conn_t  *connp = (conn_t *)proto_handle;
6308 
6309 	/* shut down the send side */
6310 	if (how != SHUT_RD)
6311 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6312 		    SOCK_OPCTL_SHUT_SEND, 0);
6313 	/* shut down the recv side */
6314 	if (how != SHUT_WR)
6315 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6316 		    SOCK_OPCTL_SHUT_RECV, 0);
6317 	return (0);
6318 }
6319 
6320 void
6321 rawip_clr_flowctrl(sock_lower_handle_t proto_handle)
6322 {
6323 	conn_t  *connp = (conn_t *)proto_handle;
6324 	icmp_t	*icmp = connp->conn_icmp;
6325 
6326 	mutex_enter(&icmp->icmp_recv_lock);
6327 	connp->conn_flow_cntrld = B_FALSE;
6328 	mutex_exit(&icmp->icmp_recv_lock);
6329 }
6330 
6331 int
6332 rawip_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
6333     int mode, int32_t *rvalp, cred_t *cr)
6334 {
6335 	conn_t  	*connp = (conn_t *)proto_handle;
6336 	int		error;
6337 
6338 	switch (cmd) {
6339 	case ND_SET:
6340 	case ND_GET:
6341 	case _SIOCSOCKFALLBACK:
6342 	case TI_GETPEERNAME:
6343 	case TI_GETMYNAME:
6344 #ifdef DEBUG
6345 		cmn_err(CE_CONT, "icmp_ioctl cmd 0x%x on non streams"
6346 		    " socket", cmd);
6347 #endif
6348 		error = EINVAL;
6349 		break;
6350 	default:
6351 		/*
6352 		 * Pass on to IP using helper stream
6353 		 */
6354 		error = ldi_ioctl(connp->conn_helper_info->iphs_handle,
6355 		    cmd, arg, mode, cr, rvalp);
6356 		break;
6357 	}
6358 	return (error);
6359 }
6360 
6361 /* ARGSUSED */
6362 int
6363 rawip_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
6364     cred_t *cr)
6365 {
6366 	conn_t *connp = (conn_t *)proto_handle;
6367 	icmp_t	*icmp = connp->conn_icmp;
6368 	icmp_stack_t *is = icmp->icmp_is;
6369 	int error = 0;
6370 	boolean_t bypass_dgram_errind = B_FALSE;
6371 
6372 	ASSERT(DB_TYPE(mp) == M_DATA);
6373 
6374 	if (is_system_labeled())
6375 		msg_setcredpid(mp, cr, curproc->p_pid);
6376 
6377 	/* do an implicit bind if necessary */
6378 	if (icmp->icmp_state == TS_UNBND) {
6379 		error = rawip_implicit_bind(connp);
6380 		/*
6381 		 * We could be racing with an actual bind, in which case
6382 		 * we would see EPROTO. We cross our fingers and try
6383 		 * to connect.
6384 		 */
6385 		if (!(error == 0 || error == EPROTO)) {
6386 			freemsg(mp);
6387 			return (error);
6388 		}
6389 	}
6390 
6391 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6392 
6393 	if (msg->msg_name != NULL && icmp->icmp_state == TS_DATA_XFER) {
6394 		error = EISCONN;
6395 		goto done_lock;
6396 	}
6397 
6398 	switch (icmp->icmp_family) {
6399 	case AF_INET6: {
6400 		sin6_t	*sin6;
6401 		ip6_pkt_t	ipp_s;	/* For ancillary data options */
6402 		ip6_pkt_t	*ipp = &ipp_s;
6403 
6404 		sin6 = (sin6_t *)msg->msg_name;
6405 		if (sin6 != NULL) {
6406 			error = proto_verify_ip_addr(icmp->icmp_family,
6407 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6408 			if (error != 0) {
6409 				bypass_dgram_errind = B_TRUE;
6410 				goto done_lock;
6411 			}
6412 			if (icmp->icmp_delayed_error != 0) {
6413 				sin6_t  *sin1 = (sin6_t *)msg->msg_name;
6414 				sin6_t  *sin2 = (sin6_t *)
6415 				    &icmp->icmp_delayed_addr;
6416 
6417 				error = icmp->icmp_delayed_error;
6418 				icmp->icmp_delayed_error = 0;
6419 
6420 				/* Compare IP address and port */
6421 
6422 				if (sin1->sin6_port == sin2->sin6_port &&
6423 				    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
6424 				    &sin2->sin6_addr)) {
6425 					goto done_lock;
6426 				}
6427 			}
6428 		} else {
6429 			/*
6430 			 * Use connected address
6431 			 */
6432 			if (icmp->icmp_state != TS_DATA_XFER) {
6433 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6434 				error = EDESTADDRREQ;
6435 				bypass_dgram_errind = B_TRUE;
6436 				goto done_lock;
6437 			}
6438 			sin6 = &icmp->icmp_v6dst;
6439 		}
6440 
6441 		/* No support for mapped addresses on raw sockets */
6442 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
6443 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6444 			error = EADDRNOTAVAIL;
6445 			goto done_lock;
6446 		}
6447 
6448 		ipp->ipp_fields = 0;
6449 		ipp->ipp_sticky_ignored = 0;
6450 
6451 		/*
6452 		 * If options passed in, feed it for verification and handling
6453 		 */
6454 		if (msg->msg_controllen != 0) {
6455 			error = process_auxiliary_options(connp,
6456 			    msg->msg_control, msg->msg_controllen,
6457 			    ipp, &icmp_opt_obj, icmp_opt_set);
6458 			if (error != 0) {
6459 				goto done_lock;
6460 			}
6461 		}
6462 
6463 		rw_exit(&icmp->icmp_rwlock);
6464 
6465 		/*
6466 		 * Destination is a native IPv6 address.
6467 		 * Send out an IPv6 format packet.
6468 		 */
6469 
6470 		error = raw_ip_send_data_v6(connp->conn_wq, connp, mp, sin6,
6471 		    ipp);
6472 	}
6473 		break;
6474 	case AF_INET: {
6475 		sin_t	*sin;
6476 		ip4_pkt_t pktinfo;
6477 		ip4_pkt_t *pktinfop = &pktinfo;
6478 		ipaddr_t	v4dst;
6479 
6480 		sin = (sin_t *)msg->msg_name;
6481 		if (sin != NULL) {
6482 			error = proto_verify_ip_addr(icmp->icmp_family,
6483 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6484 			if (error != 0) {
6485 				bypass_dgram_errind = B_TRUE;
6486 				goto done_lock;
6487 			}
6488 			v4dst = sin->sin_addr.s_addr;
6489 			if (icmp->icmp_delayed_error != 0) {
6490 				sin_t *sin1 = (sin_t *)msg->msg_name;
6491 				sin_t *sin2 = (sin_t *)&icmp->icmp_delayed_addr;
6492 
6493 				error = icmp->icmp_delayed_error;
6494 				icmp->icmp_delayed_error = 0;
6495 
6496 				/* Compare IP address and port */
6497 				if (sin1->sin_port == sin2->sin_port &&
6498 				    sin1->sin_addr.s_addr ==
6499 				    sin2->sin_addr.s_addr) {
6500 					goto done_lock;
6501 				}
6502 
6503 			}
6504 		} else {
6505 			/*
6506 			 * Use connected address
6507 			 */
6508 			if (icmp->icmp_state != TS_DATA_XFER) {
6509 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6510 				error = EDESTADDRREQ;
6511 				bypass_dgram_errind = B_TRUE;
6512 				goto done_lock;
6513 			}
6514 			v4dst = V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6515 		}
6516 
6517 
6518 		pktinfop->ip4_ill_index = 0;
6519 		pktinfop->ip4_addr = INADDR_ANY;
6520 
6521 		/*
6522 		 * If options passed in, feed it for verification and handling
6523 		 */
6524 		if (msg->msg_controllen != 0) {
6525 			error = process_auxiliary_options(connp,
6526 			    msg->msg_control, msg->msg_controllen,
6527 			    pktinfop, &icmp_opt_obj, icmp_opt_set);
6528 			if (error != 0) {
6529 				goto done_lock;
6530 			}
6531 		}
6532 		rw_exit(&icmp->icmp_rwlock);
6533 
6534 		error = raw_ip_send_data_v4(connp->conn_wq, connp, mp,
6535 		    v4dst, pktinfop);
6536 		break;
6537 	}
6538 
6539 	default:
6540 		ASSERT(0);
6541 	}
6542 
6543 	goto done;
6544 
6545 done_lock:
6546 	rw_exit(&icmp->icmp_rwlock);
6547 	if (error != 0) {
6548 		ASSERT(mp != NULL);
6549 		freemsg(mp);
6550 	}
6551 done:
6552 	if (bypass_dgram_errind)
6553 		return (error);
6554 	return (icmp->icmp_dgram_errind ? error : 0);
6555 }
6556 
6557 sock_downcalls_t sock_rawip_downcalls = {
6558 	rawip_activate,
6559 	rawip_accept,
6560 	rawip_bind,
6561 	rawip_listen,
6562 	rawip_connect,
6563 	rawip_getpeername,
6564 	rawip_getsockname,
6565 	rawip_getsockopt,
6566 	rawip_setsockopt,
6567 	rawip_send,
6568 	NULL,
6569 	NULL,
6570 	NULL,
6571 	rawip_shutdown,
6572 	rawip_clr_flowctrl,
6573 	rawip_ioctl,
6574 	rawip_close
6575 };
6576