xref: /titanic_50/usr/src/uts/common/inet/ip/icmp.c (revision a4e4e13f4001644f2f960e3be0056c22b3a40fd1)
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 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 
28 #pragma ident	"%Z%%M%	%I%	%E% SMI"
29 
30 #include <sys/types.h>
31 #include <sys/stream.h>
32 #include <sys/stropts.h>
33 #include <sys/strlog.h>
34 #include <sys/strsun.h>
35 #define	_SUN_TPI_VERSION 2
36 #include <sys/tihdr.h>
37 #include <sys/timod.h>
38 #include <sys/ddi.h>
39 #include <sys/sunddi.h>
40 #include <sys/strsubr.h>
41 #include <sys/cmn_err.h>
42 #include <sys/debug.h>
43 #include <sys/kmem.h>
44 #include <sys/policy.h>
45 #include <sys/priv.h>
46 #include <sys/zone.h>
47 #include <sys/time.h>
48 
49 #include <sys/socket.h>
50 #include <sys/isa_defs.h>
51 #include <sys/suntpi.h>
52 #include <sys/xti_inet.h>
53 #include <sys/netstack.h>
54 
55 #include <net/route.h>
56 #include <net/if.h>
57 
58 #include <netinet/in.h>
59 #include <netinet/ip6.h>
60 #include <netinet/icmp6.h>
61 #include <inet/common.h>
62 #include <inet/ip.h>
63 #include <inet/ip6.h>
64 #include <inet/mi.h>
65 #include <inet/nd.h>
66 #include <inet/optcom.h>
67 #include <inet/snmpcom.h>
68 #include <inet/kstatcom.h>
69 #include <inet/rawip_impl.h>
70 
71 #include <netinet/ip_mroute.h>
72 #include <inet/tcp.h>
73 #include <net/pfkeyv2.h>
74 #include <inet/ipsec_info.h>
75 #include <inet/ipclassifier.h>
76 
77 #include <sys/tsol/label.h>
78 #include <sys/tsol/tnet.h>
79 
80 #include <inet/ip_ire.h>
81 #include <inet/ip_if.h>
82 
83 #include <inet/ip_impl.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_bind(queue_t *q, mblk_t *mp);
106 static void	icmp_bind_proto(queue_t *q);
107 static void	icmp_bind_result(conn_t *, mblk_t *);
108 static void	icmp_bind_ack(conn_t *, mblk_t *mp);
109 static void	icmp_bind_error(conn_t *, mblk_t *mp);
110 static int	icmp_build_hdrs(icmp_t *icmp);
111 static void	icmp_capability_req(queue_t *q, mblk_t *mp);
112 static int	icmp_close(queue_t *q);
113 static void	icmp_connect(queue_t *q, mblk_t *mp);
114 static void	icmp_disconnect(queue_t *q, mblk_t *mp);
115 static void	icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
116 		    int sys_error);
117 static void	icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
118 		    t_scalar_t t_error, int sys_error);
119 static void	icmp_icmp_error(queue_t *q, mblk_t *mp);
120 static void	icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp);
121 static void	icmp_info_req(queue_t *q, mblk_t *mp);
122 static void	icmp_input(void *, mblk_t *, void *);
123 static mblk_t	*icmp_ip_bind_mp(icmp_t *icmp, t_scalar_t bind_prim,
124 		    t_scalar_t addr_length, in_port_t);
125 static int	icmp_open(queue_t *q, dev_t *devp, int flag, int sflag,
126 		    cred_t *credp, boolean_t isv6);
127 static int	icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
128 		    cred_t *credp);
129 static int	icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
130 		    cred_t *credp);
131 static void	icmp_output(queue_t *q, mblk_t *mp);
132 static int	icmp_unitdata_opt_process(queue_t *q, mblk_t *mp,
133 		    int *errorp, void *thisdg_attrs);
134 static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
135 int		icmp_opt_set(queue_t *q, uint_t optset_context,
136 		    int level, int name, uint_t inlen,
137 		    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
138 		    void *thisdg_attrs, cred_t *cr, mblk_t *mblk);
139 int		icmp_opt_get(queue_t *q, int level, int name,
140 		    uchar_t *ptr);
141 static int	icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
142 static boolean_t icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt);
143 static int	icmp_param_set(queue_t *q, mblk_t *mp, char *value,
144 		    caddr_t cp, cred_t *cr);
145 static int	icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
146 		    uchar_t *ptr, int len);
147 static int	icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
148 		    cred_t *cr);
149 static void	icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err);
150 static void	icmp_unbind(queue_t *q, mblk_t *mp);
151 static void	icmp_wput(queue_t *q, mblk_t *mp);
152 static void	icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6,
153 		    t_scalar_t tudr_optlen);
154 static void	icmp_wput_other(queue_t *q, mblk_t *mp);
155 static void	icmp_wput_iocdata(queue_t *q, mblk_t *mp);
156 static void	icmp_wput_restricted(queue_t *q, mblk_t *mp);
157 
158 static void	*rawip_stack_init(netstackid_t stackid, netstack_t *ns);
159 static void	rawip_stack_fini(netstackid_t stackid, void *arg);
160 
161 static void	*rawip_kstat_init(netstackid_t stackid);
162 static void	rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp);
163 static int	rawip_kstat_update(kstat_t *kp, int rw);
164 
165 
166 static struct module_info icmp_mod_info =  {
167 	5707, "icmp", 1, INFPSZ, 512, 128
168 };
169 
170 /*
171  * Entry points for ICMP as a device.
172  * We have separate open functions for the /dev/icmp and /dev/icmp6 devices.
173  */
174 static struct qinit icmprinitv4 = {
175 	NULL, NULL, icmp_openv4, icmp_close, NULL, &icmp_mod_info
176 };
177 
178 static struct qinit icmprinitv6 = {
179 	NULL, NULL, icmp_openv6, icmp_close, NULL, &icmp_mod_info
180 };
181 
182 static struct qinit icmpwinit = {
183 	(pfi_t)icmp_wput, (pfi_t)ip_wsrv, NULL, NULL, NULL, &icmp_mod_info
184 };
185 
186 /* For AF_INET aka /dev/icmp */
187 struct streamtab icmpinfov4 = {
188 	&icmprinitv4, &icmpwinit
189 };
190 
191 /* For AF_INET6 aka /dev/icmp6 */
192 struct streamtab icmpinfov6 = {
193 	&icmprinitv6, &icmpwinit
194 };
195 
196 static sin_t	sin_null;	/* Zero address for quick clears */
197 static sin6_t	sin6_null;	/* Zero address for quick clears */
198 
199 /* Default structure copied into T_INFO_ACK messages */
200 static struct T_info_ack icmp_g_t_info_ack = {
201 	T_INFO_ACK,
202 	IP_MAXPACKET,	 /* TSDU_size.  icmp allows maximum size messages. */
203 	T_INVALID,	/* ETSDU_size.  icmp does not support expedited data. */
204 	T_INVALID,	/* CDATA_size. icmp does not support connect data. */
205 	T_INVALID,	/* DDATA_size. icmp does not support disconnect data. */
206 	0,		/* ADDR_size - filled in later. */
207 	0,		/* OPT_size - not initialized here */
208 	IP_MAXPACKET,	/* TIDU_size.  icmp allows maximum size messages. */
209 	T_CLTS,		/* SERV_type.  icmp supports connection-less. */
210 	TS_UNBND,	/* CURRENT_state.  This is set from icmp_state. */
211 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
212 };
213 
214 /*
215  * Table of ND variables supported by icmp.  These are loaded into is_nd
216  * when the stack instance is created.
217  * All of these are alterable, within the min/max values given, at run time.
218  */
219 static icmpparam_t	icmp_param_arr[] = {
220 	/* min	max	value	name */
221 	{ 0,	128,	32,	"icmp_wroff_extra" },
222 	{ 1,	255,	255,	"icmp_ipv4_ttl" },
223 	{ 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS,	"icmp_ipv6_hoplimit"},
224 	{ 0,	1,	1,	"icmp_bsd_compat" },
225 	{ 4096,	65536,	8192,	"icmp_xmit_hiwat"},
226 	{ 0,	65536,	1024,	"icmp_xmit_lowat"},
227 	{ 4096,	65536,	8192,	"icmp_recv_hiwat"},
228 	{ 65536, 1024*1024*1024, 256*1024,	"icmp_max_buf"},
229 };
230 #define	is_wroff_extra			is_param_arr[0].icmp_param_value
231 #define	is_ipv4_ttl			is_param_arr[1].icmp_param_value
232 #define	is_ipv6_hoplimit		is_param_arr[2].icmp_param_value
233 #define	is_bsd_compat			is_param_arr[3].icmp_param_value
234 #define	is_xmit_hiwat			is_param_arr[4].icmp_param_value
235 #define	is_xmit_lowat			is_param_arr[5].icmp_param_value
236 #define	is_recv_hiwat			is_param_arr[6].icmp_param_value
237 #define	is_max_buf			is_param_arr[7].icmp_param_value
238 
239 /*
240  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
241  * passed to icmp_wput.
242  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP
243  * protocol type placed in the message following the address. A T_BIND_ACK
244  * message is returned by ip_bind_v4/v6.
245  */
246 static void
247 icmp_bind(queue_t *q, mblk_t *mp)
248 {
249 	sin_t	*sin;
250 	sin6_t	*sin6;
251 	mblk_t	*mp1;
252 	struct T_bind_req	*tbr;
253 	icmp_t	*icmp;
254 	conn_t	*connp = Q_TO_CONN(q);
255 
256 	icmp = connp->conn_icmp;
257 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
258 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
259 		    "icmp_bind: bad req, len %u",
260 		    (uint_t)(mp->b_wptr - mp->b_rptr));
261 		icmp_err_ack(q, mp, TPROTO, 0);
262 		return;
263 	}
264 	if (icmp->icmp_state != TS_UNBND) {
265 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
266 		    "icmp_bind: bad state, %d", icmp->icmp_state);
267 		icmp_err_ack(q, mp, TOUTSTATE, 0);
268 		return;
269 	}
270 	/*
271 	 * Reallocate the message to make sure we have enough room for an
272 	 * address and the protocol type.
273 	 */
274 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
275 	if (!mp1) {
276 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
277 		return;
278 	}
279 	mp = mp1;
280 	tbr = (struct T_bind_req *)mp->b_rptr;
281 	switch (tbr->ADDR_length) {
282 	case 0:			/* Generic request */
283 		tbr->ADDR_offset = sizeof (struct T_bind_req);
284 		if (icmp->icmp_family == AF_INET) {
285 			tbr->ADDR_length = sizeof (sin_t);
286 			sin = (sin_t *)&tbr[1];
287 			*sin = sin_null;
288 			sin->sin_family = AF_INET;
289 			mp->b_wptr = (uchar_t *)&sin[1];
290 		} else {
291 			ASSERT(icmp->icmp_family == AF_INET6);
292 			tbr->ADDR_length = sizeof (sin6_t);
293 			sin6 = (sin6_t *)&tbr[1];
294 			*sin6 = sin6_null;
295 			sin6->sin6_family = AF_INET6;
296 			mp->b_wptr = (uchar_t *)&sin6[1];
297 		}
298 		break;
299 	case sizeof (sin_t):	/* Complete IP address */
300 		sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset,
301 		    sizeof (sin_t));
302 		if (sin == NULL || !OK_32PTR((char *)sin)) {
303 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
304 			return;
305 		}
306 		if (icmp->icmp_family != AF_INET ||
307 		    sin->sin_family != AF_INET) {
308 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
309 			return;
310 		}
311 		break;
312 	case sizeof (sin6_t):	/* Complete IP address */
313 		sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset,
314 		    sizeof (sin6_t));
315 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
316 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
317 			return;
318 		}
319 		if (icmp->icmp_family != AF_INET6 ||
320 		    sin6->sin6_family != AF_INET6) {
321 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
322 			return;
323 		}
324 		/* No support for mapped addresses on raw sockets */
325 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
326 			icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL);
327 			return;
328 		}
329 		break;
330 	default:
331 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
332 		    "icmp_bind: bad ADDR_length %d", tbr->ADDR_length);
333 		icmp_err_ack(q, mp, TBADADDR, 0);
334 		return;
335 	}
336 
337 	/*
338 	 * The state must be TS_UNBND. TPI mandates that users must send
339 	 * TPI primitives only 1 at a time and wait for the response before
340 	 * sending the next primitive.
341 	 */
342 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
343 	if (icmp->icmp_state != TS_UNBND || icmp->icmp_pending_op != -1) {
344 		rw_exit(&icmp->icmp_rwlock);
345 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
346 		    "icmp_bind: bad state, %d", icmp->icmp_state);
347 		icmp_err_ack(q, mp, TOUTSTATE, 0);
348 		return;
349 	}
350 
351 	icmp->icmp_pending_op = tbr->PRIM_type;
352 
353 	/*
354 	 * Copy the source address into our icmp structure.  This address
355 	 * may still be zero; if so, ip will fill in the correct address
356 	 * each time an outbound packet is passed to it.
357 	 * If we are binding to a broadcast or multicast address then
358 	 * icmp_bind_ack will clear the source address when it receives
359 	 * the T_BIND_ACK.
360 	 */
361 	icmp->icmp_state = TS_IDLE;
362 
363 	if (icmp->icmp_family == AF_INET) {
364 		ASSERT(sin != NULL);
365 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
366 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr,
367 		    &icmp->icmp_v6src);
368 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
369 		    icmp->icmp_ip_snd_options_len;
370 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
371 	} else {
372 		int error;
373 
374 		ASSERT(sin6 != NULL);
375 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
376 		icmp->icmp_v6src = sin6->sin6_addr;
377 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
378 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
379 
380 		/* Rebuild the header template */
381 		error = icmp_build_hdrs(icmp);
382 		if (error != 0) {
383 			icmp->icmp_pending_op = -1;
384 			rw_exit(&icmp->icmp_rwlock);
385 			icmp_err_ack(q, mp, TSYSERR, error);
386 			return;
387 		}
388 	}
389 	/*
390 	 * Place protocol type in the O_T_BIND_REQ/T_BIND_REQ following
391 	 * the address.
392 	 */
393 	*mp->b_wptr++ = icmp->icmp_proto;
394 	if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) {
395 		/*
396 		 * Append a request for an IRE if src not 0 (INADDR_ANY)
397 		 */
398 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
399 		if (!mp->b_cont) {
400 			icmp->icmp_pending_op = -1;
401 			rw_exit(&icmp->icmp_rwlock);
402 			icmp_err_ack(q, mp, TSYSERR, ENOMEM);
403 			return;
404 		}
405 		mp->b_cont->b_wptr += sizeof (ire_t);
406 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
407 	}
408 	rw_exit(&icmp->icmp_rwlock);
409 
410 	/* Pass the O_T_BIND_REQ/T_BIND_REQ to ip. */
411 	if (icmp->icmp_family == AF_INET6)
412 		mp = ip_bind_v6(q, mp, connp, NULL);
413 	else
414 		mp = ip_bind_v4(q, mp, connp);
415 
416 	/* The above return NULL if the bind needs to be deferred */
417 	if (mp != NULL)
418 		icmp_bind_result(connp, mp);
419 	else
420 		CONN_INC_REF(connp);
421 }
422 
423 /*
424  * Send message to IP to just bind to the protocol.
425  */
426 static void
427 icmp_bind_proto(queue_t *q)
428 {
429 	mblk_t	*mp;
430 	struct T_bind_req	*tbr;
431 	icmp_t	*icmp;
432 	conn_t	*connp = Q_TO_CONN(q);
433 
434 	icmp = connp->conn_icmp;
435 
436 	mp = allocb(sizeof (struct T_bind_req) + sizeof (sin6_t) + 1,
437 	    BPRI_MED);
438 	if (!mp) {
439 		return;
440 	}
441 	mp->b_datap->db_type = M_PROTO;
442 	tbr = (struct T_bind_req *)mp->b_rptr;
443 	tbr->PRIM_type = O_T_BIND_REQ; /* change to T_BIND_REQ ? */
444 	tbr->ADDR_offset = sizeof (struct T_bind_req);
445 
446 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
447 	if (icmp->icmp_ipversion == IPV4_VERSION) {
448 		sin_t	*sin;
449 
450 		tbr->ADDR_length = sizeof (sin_t);
451 		sin = (sin_t *)&tbr[1];
452 		*sin = sin_null;
453 		sin->sin_family = AF_INET;
454 		mp->b_wptr = (uchar_t *)&sin[1];
455 	} else {
456 		sin6_t	*sin6;
457 
458 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
459 		tbr->ADDR_length = sizeof (sin6_t);
460 		sin6 = (sin6_t *)&tbr[1];
461 		*sin6 = sin6_null;
462 		sin6->sin6_family = AF_INET6;
463 		mp->b_wptr = (uchar_t *)&sin6[1];
464 	}
465 
466 	/* Place protocol type in the O_T_BIND_REQ following the address. */
467 	*mp->b_wptr++ = icmp->icmp_proto;
468 	rw_exit(&icmp->icmp_rwlock);
469 
470 	/* Pass the O_T_BIND_REQ to ip. */
471 	if (icmp->icmp_family == AF_INET6)
472 		mp = ip_bind_v6(q, mp, connp, NULL);
473 	else
474 		mp = ip_bind_v4(q, mp, connp);
475 
476 	/* The above return NULL if the bind needs to be deferred */
477 	if (mp != NULL)
478 		icmp_bind_result(connp, mp);
479 	else
480 		CONN_INC_REF(connp);
481 }
482 
483 /*
484  * This is called from ip_wput_nondata to handle the results of a
485  * deferred RAWIP bind.  It is called once the bind has been completed.
486  */
487 void
488 rawip_resume_bind(conn_t *connp, mblk_t *mp)
489 {
490 	ASSERT(connp != NULL && IPCL_IS_RAWIP(connp));
491 
492 	icmp_bind_result(connp, mp);
493 
494 	CONN_OPER_PENDING_DONE(connp);
495 }
496 
497 /*
498  * This routine handles each T_CONN_REQ message passed to icmp.  It
499  * associates a default destination address with the stream.
500  *
501  * This routine sends down a T_BIND_REQ to IP with the following mblks:
502  *	T_BIND_REQ	- specifying local and remote address.
503  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
504  *	T_OK_ACK	- for the T_CONN_REQ
505  *	T_CONN_CON	- to keep the TPI user happy
506  *
507  * The connect completes in icmp_bind_result.
508  * When a T_BIND_ACK is received information is extracted from the IRE
509  * and the two appended messages are sent to the TPI user.
510  * Should icmp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
511  * convert it to an error ack for the appropriate primitive.
512  */
513 static void
514 icmp_connect(queue_t *q, mblk_t *mp)
515 {
516 	sin_t	*sin;
517 	sin6_t	*sin6;
518 	mblk_t	*mp1, *mp2;
519 	struct T_conn_req	*tcr;
520 	icmp_t	*icmp;
521 	ipaddr_t	v4dst;
522 	in6_addr_t	v6dst;
523 	uint32_t	flowinfo;
524 	conn_t	*connp = Q_TO_CONN(q);
525 
526 	icmp = connp->conn_icmp;
527 	tcr = (struct T_conn_req *)mp->b_rptr;
528 	/* Sanity checks */
529 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
530 		icmp_err_ack(q, mp, TPROTO, 0);
531 		return;
532 	}
533 
534 	if (tcr->OPT_length != 0) {
535 		icmp_err_ack(q, mp, TBADOPT, 0);
536 		return;
537 	}
538 
539 	switch (tcr->DEST_length) {
540 	default:
541 		icmp_err_ack(q, mp, TBADADDR, 0);
542 		return;
543 
544 	case sizeof (sin_t):
545 		sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset,
546 		    sizeof (sin_t));
547 		if (sin == NULL || !OK_32PTR((char *)sin)) {
548 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
549 			return;
550 		}
551 		if (icmp->icmp_family != AF_INET ||
552 		    sin->sin_family != AF_INET) {
553 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
554 			return;
555 		}
556 		v4dst = sin->sin_addr.s_addr;
557 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
558 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
559 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
560 		    icmp->icmp_ip_snd_options_len;
561 		break;
562 
563 	case sizeof (sin6_t):
564 		sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset,
565 		    sizeof (sin6_t));
566 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
567 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
568 			return;
569 		}
570 		if (icmp->icmp_family != AF_INET6 ||
571 		    sin6->sin6_family != AF_INET6) {
572 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
573 			return;
574 		}
575 		/* No support for mapped addresses on raw sockets */
576 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
577 			icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL);
578 			return;
579 		}
580 		v6dst = sin6->sin6_addr;
581 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
582 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
583 		flowinfo = sin6->sin6_flowinfo;
584 		break;
585 	}
586 	if (icmp->icmp_ipversion == IPV4_VERSION) {
587 		/*
588 		 * Interpret a zero destination to mean loopback.
589 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
590 		 * generate the T_CONN_CON.
591 		 */
592 		if (v4dst == INADDR_ANY) {
593 			v4dst = htonl(INADDR_LOOPBACK);
594 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
595 			if (icmp->icmp_family == AF_INET) {
596 				sin->sin_addr.s_addr = v4dst;
597 			} else {
598 				sin6->sin6_addr = v6dst;
599 			}
600 		}
601 		icmp->icmp_v6dst = v6dst;
602 		icmp->icmp_flowinfo = 0;
603 
604 		/*
605 		 * If the destination address is multicast and
606 		 * an outgoing multicast interface has been set,
607 		 * use the address of that interface as our
608 		 * source address if no source address has been set.
609 		 */
610 		if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY &&
611 		    CLASSD(v4dst) &&
612 		    icmp->icmp_multicast_if_addr != INADDR_ANY) {
613 			IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr,
614 			    &icmp->icmp_v6src);
615 		}
616 	} else {
617 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
618 		/*
619 		 * Interpret a zero destination to mean loopback.
620 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
621 		 * generate the T_CONN_CON.
622 		 */
623 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
624 			v6dst = ipv6_loopback;
625 			sin6->sin6_addr = v6dst;
626 		}
627 		icmp->icmp_v6dst = v6dst;
628 		icmp->icmp_flowinfo = flowinfo;
629 		/*
630 		 * If the destination address is multicast and
631 		 * an outgoing multicast interface has been set,
632 		 * then the ip bind logic will pick the correct source
633 		 * address (i.e. matching the outgoing multicast interface).
634 		 */
635 	}
636 
637 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
638 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
639 		rw_exit(&icmp->icmp_rwlock);
640 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
641 		    "icmp_connect: bad state, %d", icmp->icmp_state);
642 		icmp_err_ack(q, mp, TOUTSTATE, 0);
643 		return;
644 	}
645 	icmp->icmp_pending_op = T_CONN_REQ;
646 
647 	if (icmp->icmp_state == TS_DATA_XFER) {
648 		/* Already connected - clear out state */
649 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
650 		icmp->icmp_state = TS_IDLE;
651 	}
652 
653 	/*
654 	 * Send down bind to IP to verify that there is a route
655 	 * and to determine the source address.
656 	 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput.
657 	 */
658 	if (icmp->icmp_family == AF_INET) {
659 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa_conn_t),
660 		    sin->sin_port);
661 	} else {
662 		ASSERT(icmp->icmp_family == AF_INET6);
663 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa6_conn_t),
664 		    sin6->sin6_port);
665 	}
666 	if (mp1 == NULL) {
667 		icmp->icmp_pending_op = -1;
668 		rw_exit(&icmp->icmp_rwlock);
669 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
670 		return;
671 	}
672 
673 	/*
674 	 * We also have to send a connection confirmation to
675 	 * keep TLI happy. Prepare it for icmp_bind_result.
676 	 */
677 	if (icmp->icmp_family == AF_INET) {
678 		mp2 = mi_tpi_conn_con(NULL, (char *)sin, sizeof (*sin), NULL,
679 		    0);
680 	} else {
681 		ASSERT(icmp->icmp_family == AF_INET6);
682 		mp2 = mi_tpi_conn_con(NULL, (char *)sin6, sizeof (*sin6), NULL,
683 		    0);
684 	}
685 	if (mp2 == NULL) {
686 		freemsg(mp1);
687 		icmp->icmp_pending_op = -1;
688 		rw_exit(&icmp->icmp_rwlock);
689 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
690 		return;
691 	}
692 
693 	mp = mi_tpi_ok_ack_alloc(mp);
694 	if (mp == NULL) {
695 		/* Unable to reuse the T_CONN_REQ for the ack. */
696 		freemsg(mp2);
697 		icmp->icmp_pending_op = -1;
698 		rw_exit(&icmp->icmp_rwlock);
699 		icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
700 		return;
701 	}
702 
703 	icmp->icmp_state = TS_DATA_XFER;
704 	rw_exit(&icmp->icmp_rwlock);
705 
706 	/* Hang onto the T_OK_ACK and T_CONN_CON for later. */
707 	linkb(mp1, mp);
708 	linkb(mp1, mp2);
709 
710 	mblk_setcred(mp1, connp->conn_cred);
711 	if (icmp->icmp_family == AF_INET)
712 		mp1 = ip_bind_v4(q, mp1, connp);
713 	else
714 		mp1 = ip_bind_v6(q, mp1, connp, NULL);
715 
716 	/* The above return NULL if the bind needs to be deferred */
717 	if (mp1 != NULL)
718 		icmp_bind_result(connp, mp1);
719 	else
720 		CONN_INC_REF(connp);
721 }
722 
723 static void
724 icmp_close_free(conn_t *connp)
725 {
726 	icmp_t *icmp = connp->conn_icmp;
727 
728 	/* If there are any options associated with the stream, free them. */
729 	if (icmp->icmp_ip_snd_options != NULL) {
730 		mi_free((char *)icmp->icmp_ip_snd_options);
731 		icmp->icmp_ip_snd_options = NULL;
732 		icmp->icmp_ip_snd_options_len = 0;
733 	}
734 
735 	if (icmp->icmp_filter != NULL) {
736 		kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t));
737 		icmp->icmp_filter = NULL;
738 	}
739 	/* Free memory associated with sticky options */
740 	if (icmp->icmp_sticky_hdrs_len != 0) {
741 		kmem_free(icmp->icmp_sticky_hdrs,
742 		    icmp->icmp_sticky_hdrs_len);
743 		icmp->icmp_sticky_hdrs = NULL;
744 		icmp->icmp_sticky_hdrs_len = 0;
745 	}
746 	ip6_pkt_free(&icmp->icmp_sticky_ipp);
747 
748 	/*
749 	 * Clear any fields which the kmem_cache constructor clears.
750 	 * Only icmp_connp needs to be preserved.
751 	 * TBD: We should make this more efficient to avoid clearing
752 	 * everything.
753 	 */
754 	ASSERT(icmp->icmp_connp == connp);
755 	bzero(icmp, sizeof (icmp_t));
756 	icmp->icmp_connp = connp;
757 }
758 
759 static int
760 icmp_close(queue_t *q)
761 {
762 	conn_t	*connp = (conn_t *)q->q_ptr;
763 
764 	ASSERT(connp != NULL && IPCL_IS_RAWIP(connp));
765 
766 	ip_quiesce_conn(connp);
767 
768 	qprocsoff(connp->conn_rq);
769 
770 	icmp_close_free(connp);
771 
772 	/*
773 	 * Now we are truly single threaded on this stream, and can
774 	 * delete the things hanging off the connp, and finally the connp.
775 	 * We removed this connp from the fanout list, it cannot be
776 	 * accessed thru the fanouts, and we already waited for the
777 	 * conn_ref to drop to 0. We are already in close, so
778 	 * there cannot be any other thread from the top. qprocsoff
779 	 * has completed, and service has completed or won't run in
780 	 * future.
781 	 */
782 	ASSERT(connp->conn_ref == 1);
783 
784 	inet_minor_free(ip_minor_arena, connp->conn_dev);
785 
786 	connp->conn_ref--;
787 	ipcl_conn_destroy(connp);
788 
789 	q->q_ptr = WR(q)->q_ptr = NULL;
790 	return (0);
791 }
792 
793 /*
794  * This routine handles each T_DISCON_REQ message passed to icmp
795  * as an indicating that ICMP is no longer connected. This results
796  * in sending a T_BIND_REQ to IP to restore the binding to just
797  * the local address.
798  *
799  * This routine sends down a T_BIND_REQ to IP with the following mblks:
800  *	T_BIND_REQ	- specifying just the local address.
801  *	T_OK_ACK	- for the T_DISCON_REQ
802  *
803  * The disconnect completes in icmp_bind_result.
804  * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user.
805  * Should icmp_bind_result receive T_ERROR_ACK for the T_BIND_REQ it will
806  * convert it to an error ack for the appropriate primitive.
807  */
808 static void
809 icmp_disconnect(queue_t *q, mblk_t *mp)
810 {
811 	icmp_t	*icmp;
812 	mblk_t	*mp1;
813 	conn_t	*connp = Q_TO_CONN(q);
814 
815 	icmp = connp->conn_icmp;
816 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
817 	if (icmp->icmp_state != TS_DATA_XFER || icmp->icmp_pending_op != -1) {
818 		rw_exit(&icmp->icmp_rwlock);
819 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
820 		    "icmp_disconnect: bad state, %d", icmp->icmp_state);
821 		icmp_err_ack(q, mp, TOUTSTATE, 0);
822 		return;
823 	}
824 	icmp->icmp_pending_op = T_DISCON_REQ;
825 	icmp->icmp_v6src = icmp->icmp_bound_v6src;
826 	icmp->icmp_state = TS_IDLE;
827 
828 	/*
829 	 * Send down bind to IP to remove the full binding and revert
830 	 * to the local address binding.
831 	 */
832 	if (icmp->icmp_family == AF_INET) {
833 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin_t), 0);
834 	} else {
835 		ASSERT(icmp->icmp_family == AF_INET6);
836 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin6_t), 0);
837 	}
838 	if (mp1 == NULL) {
839 		icmp->icmp_pending_op = -1;
840 		rw_exit(&icmp->icmp_rwlock);
841 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
842 		return;
843 	}
844 	mp = mi_tpi_ok_ack_alloc(mp);
845 	if (mp == NULL) {
846 		/* Unable to reuse the T_DISCON_REQ for the ack. */
847 		icmp->icmp_pending_op = -1;
848 		rw_exit(&icmp->icmp_rwlock);
849 		icmp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM);
850 		return;
851 	}
852 
853 	if (icmp->icmp_family == AF_INET6) {
854 		int error;
855 
856 		/* Rebuild the header template */
857 		error = icmp_build_hdrs(icmp);
858 		if (error != 0) {
859 			icmp->icmp_pending_op = -1;
860 			rw_exit(&icmp->icmp_rwlock);
861 			icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error);
862 			freemsg(mp1);
863 			return;
864 		}
865 	}
866 
867 	rw_exit(&icmp->icmp_rwlock);
868 	/* Append the T_OK_ACK to the T_BIND_REQ for icmp_bind_result */
869 	linkb(mp1, mp);
870 
871 	if (icmp->icmp_family == AF_INET6)
872 		mp1 = ip_bind_v6(q, mp1, connp, NULL);
873 	else
874 		mp1 = ip_bind_v4(q, mp1, connp);
875 
876 	/* The above return NULL if the bind needs to be deferred */
877 	if (mp1 != NULL)
878 		icmp_bind_result(connp, mp1);
879 	else
880 		CONN_INC_REF(connp);
881 }
882 
883 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
884 static void
885 icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
886 {
887 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
888 		qreply(q, mp);
889 }
890 
891 /* Shorthand to generate and send TPI error acks to our client */
892 static void
893 icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
894     t_scalar_t t_error, int sys_error)
895 {
896 	struct T_error_ack	*teackp;
897 
898 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
899 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
900 		teackp = (struct T_error_ack *)mp->b_rptr;
901 		teackp->ERROR_prim = primitive;
902 		teackp->TLI_error = t_error;
903 		teackp->UNIX_error = sys_error;
904 		qreply(q, mp);
905 	}
906 }
907 
908 /*
909  * icmp_icmp_error is called by icmp_input to process ICMP
910  * messages passed up by IP.
911  * Generates the appropriate T_UDERROR_IND for permanent
912  * (non-transient) errors.
913  * Assumes that IP has pulled up everything up to and including
914  * the ICMP header.
915  */
916 static void
917 icmp_icmp_error(queue_t *q, mblk_t *mp)
918 {
919 	icmph_t *icmph;
920 	ipha_t	*ipha;
921 	int	iph_hdr_length;
922 	sin_t	sin;
923 	sin6_t	sin6;
924 	mblk_t	*mp1;
925 	int	error = 0;
926 	icmp_t	*icmp = Q_TO_ICMP(q);
927 
928 	ipha = (ipha_t *)mp->b_rptr;
929 
930 	ASSERT(OK_32PTR(mp->b_rptr));
931 
932 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
933 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
934 		icmp_icmp_error_ipv6(q, mp);
935 		return;
936 	}
937 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
938 
939 	/* Skip past the outer IP and ICMP headers */
940 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
941 	icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]);
942 	ipha = (ipha_t *)&icmph[1];
943 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
944 
945 	switch (icmph->icmph_type) {
946 	case ICMP_DEST_UNREACHABLE:
947 		switch (icmph->icmph_code) {
948 		case ICMP_FRAGMENTATION_NEEDED:
949 			/*
950 			 * IP has already adjusted the path MTU.
951 			 */
952 			break;
953 		case ICMP_PORT_UNREACHABLE:
954 		case ICMP_PROTOCOL_UNREACHABLE:
955 			error = ECONNREFUSED;
956 			break;
957 		default:
958 			/* Transient errors */
959 			break;
960 		}
961 		break;
962 	default:
963 		/* Transient errors */
964 		break;
965 	}
966 	if (error == 0) {
967 		freemsg(mp);
968 		return;
969 	}
970 
971 	/*
972 	 * Deliver T_UDERROR_IND when the application has asked for it.
973 	 * The socket layer enables this automatically when connected.
974 	 */
975 	if (!icmp->icmp_dgram_errind) {
976 		freemsg(mp);
977 		return;
978 	}
979 
980 	switch (icmp->icmp_family) {
981 	case AF_INET:
982 		sin = sin_null;
983 		sin.sin_family = AF_INET;
984 		sin.sin_addr.s_addr = ipha->ipha_dst;
985 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0,
986 		    error);
987 		break;
988 	case AF_INET6:
989 		sin6 = sin6_null;
990 		sin6.sin6_family = AF_INET6;
991 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
992 
993 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
994 		    NULL, 0, error);
995 		break;
996 	}
997 	if (mp1)
998 		putnext(q, mp1);
999 	freemsg(mp);
1000 }
1001 
1002 /*
1003  * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6
1004  * for IPv6 packets.
1005  * Send permanent (non-transient) errors upstream.
1006  * Assumes that IP has pulled up all the extension headers as well
1007  * as the ICMPv6 header.
1008  */
1009 static void
1010 icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp)
1011 {
1012 	icmp6_t		*icmp6;
1013 	ip6_t		*ip6h, *outer_ip6h;
1014 	uint16_t	iph_hdr_length;
1015 	uint8_t		*nexthdrp;
1016 	sin6_t		sin6;
1017 	mblk_t		*mp1;
1018 	int		error = 0;
1019 	icmp_t		*icmp = Q_TO_ICMP(q);
1020 
1021 	outer_ip6h = (ip6_t *)mp->b_rptr;
1022 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1023 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1024 	else
1025 		iph_hdr_length = IPV6_HDR_LEN;
1026 
1027 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1028 	ip6h = (ip6_t *)&icmp6[1];
1029 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1030 		freemsg(mp);
1031 		return;
1032 	}
1033 
1034 	switch (icmp6->icmp6_type) {
1035 	case ICMP6_DST_UNREACH:
1036 		switch (icmp6->icmp6_code) {
1037 		case ICMP6_DST_UNREACH_NOPORT:
1038 			error = ECONNREFUSED;
1039 			break;
1040 		case ICMP6_DST_UNREACH_ADMIN:
1041 		case ICMP6_DST_UNREACH_NOROUTE:
1042 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1043 		case ICMP6_DST_UNREACH_ADDR:
1044 			/* Transient errors */
1045 			break;
1046 		default:
1047 			break;
1048 		}
1049 		break;
1050 	case ICMP6_PACKET_TOO_BIG: {
1051 		struct T_unitdata_ind	*tudi;
1052 		struct T_opthdr		*toh;
1053 		size_t			udi_size;
1054 		mblk_t			*newmp;
1055 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1056 		    sizeof (struct ip6_mtuinfo);
1057 		sin6_t			*sin6;
1058 		struct ip6_mtuinfo	*mtuinfo;
1059 
1060 		/*
1061 		 * If the application has requested to receive path mtu
1062 		 * information, send up an empty message containing an
1063 		 * IPV6_PATHMTU ancillary data item.
1064 		 */
1065 		if (!icmp->icmp_ipv6_recvpathmtu)
1066 			break;
1067 
1068 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1069 		    opt_length;
1070 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1071 			BUMP_MIB(&icmp->icmp_is->is_rawip_mib, rawipInErrors);
1072 			break;
1073 		}
1074 
1075 		/*
1076 		 * newmp->b_cont is left to NULL on purpose.  This is an
1077 		 * empty message containing only ancillary data.
1078 		 */
1079 		newmp->b_datap->db_type = M_PROTO;
1080 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1081 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1082 		tudi->PRIM_type = T_UNITDATA_IND;
1083 		tudi->SRC_length = sizeof (sin6_t);
1084 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1085 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1086 		tudi->OPT_length = opt_length;
1087 
1088 		sin6 = (sin6_t *)&tudi[1];
1089 		bzero(sin6, sizeof (sin6_t));
1090 		sin6->sin6_family = AF_INET6;
1091 		sin6->sin6_addr = icmp->icmp_v6dst;
1092 
1093 		toh = (struct T_opthdr *)&sin6[1];
1094 		toh->level = IPPROTO_IPV6;
1095 		toh->name = IPV6_PATHMTU;
1096 		toh->len = opt_length;
1097 		toh->status = 0;
1098 
1099 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1100 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1101 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1102 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1103 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1104 		/*
1105 		 * We've consumed everything we need from the original
1106 		 * message.  Free it, then send our empty message.
1107 		 */
1108 		freemsg(mp);
1109 		putnext(q, newmp);
1110 		return;
1111 	}
1112 	case ICMP6_TIME_EXCEEDED:
1113 		/* Transient errors */
1114 		break;
1115 	case ICMP6_PARAM_PROB:
1116 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1117 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1118 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1119 		    (uchar_t *)nexthdrp) {
1120 			error = ECONNREFUSED;
1121 			break;
1122 		}
1123 		break;
1124 	}
1125 	if (error == 0) {
1126 		freemsg(mp);
1127 		return;
1128 	}
1129 
1130 	/*
1131 	 * Deliver T_UDERROR_IND when the application has asked for it.
1132 	 * The socket layer enables this automatically when connected.
1133 	 */
1134 	if (!icmp->icmp_dgram_errind) {
1135 		freemsg(mp);
1136 		return;
1137 	}
1138 
1139 	sin6 = sin6_null;
1140 	sin6.sin6_family = AF_INET6;
1141 	sin6.sin6_addr = ip6h->ip6_dst;
1142 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1143 
1144 	mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0,
1145 	    error);
1146 	if (mp1)
1147 		putnext(q, mp1);
1148 	freemsg(mp);
1149 }
1150 
1151 /*
1152  * This routine responds to T_ADDR_REQ messages.  It is called by icmp_wput.
1153  * The local address is filled in if endpoint is bound. The remote address
1154  * is filled in if remote address has been precified ("connected endpoint")
1155  * (The concept of connected CLTS sockets is alien to published TPI
1156  *  but we support it anyway).
1157  */
1158 static void
1159 icmp_addr_req(queue_t *q, mblk_t *mp)
1160 {
1161 	icmp_t	*icmp = Q_TO_ICMP(q);
1162 	mblk_t	*ackmp;
1163 	struct T_addr_ack *taa;
1164 
1165 	/* Make it large enough for worst case */
1166 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1167 	    2 * sizeof (sin6_t), 1);
1168 	if (ackmp == NULL) {
1169 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
1170 		return;
1171 	}
1172 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1173 
1174 	bzero(taa, sizeof (struct T_addr_ack));
1175 	ackmp->b_wptr = (uchar_t *)&taa[1];
1176 
1177 	taa->PRIM_type = T_ADDR_ACK;
1178 	ackmp->b_datap->db_type = M_PCPROTO;
1179 	rw_enter(&icmp->icmp_rwlock, RW_READER);
1180 	/*
1181 	 * Note: Following code assumes 32 bit alignment of basic
1182 	 * data structures like sin_t and struct T_addr_ack.
1183 	 */
1184 	if (icmp->icmp_state != TS_UNBND) {
1185 		/*
1186 		 * Fill in local address
1187 		 */
1188 		taa->LOCADDR_offset = sizeof (*taa);
1189 		if (icmp->icmp_family == AF_INET) {
1190 			sin_t	*sin;
1191 
1192 			taa->LOCADDR_length = sizeof (sin_t);
1193 			sin = (sin_t *)&taa[1];
1194 			/* Fill zeroes and then intialize non-zero fields */
1195 			*sin = sin_null;
1196 			sin->sin_family = AF_INET;
1197 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
1198 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1199 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src,
1200 				    sin->sin_addr.s_addr);
1201 			} else {
1202 				/*
1203 				 * INADDR_ANY
1204 				 * icmp_v6src is not set, we might be bound to
1205 				 * broadcast/multicast. Use icmp_bound_v6src as
1206 				 * local address instead (that could
1207 				 * also still be INADDR_ANY)
1208 				 */
1209 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src,
1210 				    sin->sin_addr.s_addr);
1211 			}
1212 			ackmp->b_wptr = (uchar_t *)&sin[1];
1213 		} else {
1214 			sin6_t	*sin6;
1215 
1216 			ASSERT(icmp->icmp_family == AF_INET6);
1217 			taa->LOCADDR_length = sizeof (sin6_t);
1218 			sin6 = (sin6_t *)&taa[1];
1219 			/* Fill zeroes and then intialize non-zero fields */
1220 			*sin6 = sin6_null;
1221 			sin6->sin6_family = AF_INET6;
1222 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1223 				sin6->sin6_addr = icmp->icmp_v6src;
1224 			} else {
1225 				/*
1226 				 * UNSPECIFIED
1227 				 * icmp_v6src is not set, we might be bound to
1228 				 * broadcast/multicast. Use icmp_bound_v6src as
1229 				 * local address instead (that could
1230 				 * also still be UNSPECIFIED)
1231 				 */
1232 				sin6->sin6_addr = icmp->icmp_bound_v6src;
1233 			}
1234 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1235 		}
1236 	}
1237 	rw_exit(&icmp->icmp_rwlock);
1238 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1239 	qreply(q, ackmp);
1240 }
1241 
1242 static void
1243 icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp)
1244 {
1245 	*tap = icmp_g_t_info_ack;
1246 
1247 	if (icmp->icmp_family == AF_INET6)
1248 		tap->ADDR_size = sizeof (sin6_t);
1249 	else
1250 		tap->ADDR_size = sizeof (sin_t);
1251 	tap->CURRENT_state = icmp->icmp_state;
1252 	tap->OPT_size = icmp_max_optsize;
1253 }
1254 
1255 /*
1256  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1257  * icmp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1258  * icmp_g_t_info_ack.  The current state of the stream is copied from
1259  * icmp_state.
1260  */
1261 static void
1262 icmp_capability_req(queue_t *q, mblk_t *mp)
1263 {
1264 	icmp_t			*icmp = Q_TO_ICMP(q);
1265 	t_uscalar_t		cap_bits1;
1266 	struct T_capability_ack	*tcap;
1267 
1268 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1269 
1270 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1271 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1272 	if (!mp)
1273 		return;
1274 
1275 	tcap = (struct T_capability_ack *)mp->b_rptr;
1276 	tcap->CAP_bits1 = 0;
1277 
1278 	if (cap_bits1 & TC1_INFO) {
1279 		icmp_copy_info(&tcap->INFO_ack, icmp);
1280 		tcap->CAP_bits1 |= TC1_INFO;
1281 	}
1282 
1283 	qreply(q, mp);
1284 }
1285 
1286 /*
1287  * This routine responds to T_INFO_REQ messages.  It is called by icmp_wput.
1288  * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack.
1289  * The current state of the stream is copied from icmp_state.
1290  */
1291 static void
1292 icmp_info_req(queue_t *q, mblk_t *mp)
1293 {
1294 	icmp_t	*icmp = Q_TO_ICMP(q);
1295 
1296 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1297 	    T_INFO_ACK);
1298 	if (!mp)
1299 		return;
1300 	icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp);
1301 	qreply(q, mp);
1302 }
1303 
1304 /*
1305  * IP recognizes seven kinds of bind requests:
1306  *
1307  * - A zero-length address binds only to the protocol number.
1308  *
1309  * - A 4-byte address is treated as a request to
1310  * validate that the address is a valid local IPv4
1311  * address, appropriate for an application to bind to.
1312  * IP does the verification, but does not make any note
1313  * of the address at this time.
1314  *
1315  * - A 16-byte address contains is treated as a request
1316  * to validate a local IPv6 address, as the 4-byte
1317  * address case above.
1318  *
1319  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
1320  * use it for the inbound fanout of packets.
1321  *
1322  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
1323  * use it for the inbound fanout of packets.
1324  *
1325  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
1326  * information consisting of local and remote addresses
1327  * and ports (unused for raw sockets).  In this case, the addresses are both
1328  * validated as appropriate for this operation, and, if
1329  * so, the information is retained for use in the
1330  * inbound fanout.
1331  *
1332  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
1333  * fanout information, like the 12-byte case above.
1334  *
1335  * IP will also fill in the IRE request mblk with information
1336  * regarding our peer.  In all cases, we notify IP of our protocol
1337  * type by appending a single protocol byte to the bind request.
1338  */
1339 static mblk_t *
1340 icmp_ip_bind_mp(icmp_t *icmp, t_scalar_t bind_prim, t_scalar_t addr_length,
1341     in_port_t fport)
1342 {
1343 	char	*cp;
1344 	mblk_t	*mp;
1345 	struct T_bind_req *tbr;
1346 	ipa_conn_t	*ac;
1347 	ipa6_conn_t	*ac6;
1348 	sin_t		*sin;
1349 	sin6_t		*sin6;
1350 
1351 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
1352 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
1353 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
1354 	if (mp == NULL)
1355 		return (NULL);
1356 	mp->b_datap->db_type = M_PROTO;
1357 	tbr = (struct T_bind_req *)mp->b_rptr;
1358 	tbr->PRIM_type = bind_prim;
1359 	tbr->ADDR_offset = sizeof (*tbr);
1360 	tbr->CONIND_number = 0;
1361 	tbr->ADDR_length = addr_length;
1362 	cp = (char *)&tbr[1];
1363 	switch (addr_length) {
1364 	case sizeof (ipa_conn_t):
1365 		ASSERT(icmp->icmp_family == AF_INET);
1366 		/* Append a request for an IRE */
1367 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1368 		if (mp->b_cont == NULL) {
1369 			freemsg(mp);
1370 			return (NULL);
1371 		}
1372 		mp->b_cont->b_wptr += sizeof (ire_t);
1373 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1374 
1375 		/* cp known to be 32 bit aligned */
1376 		ac = (ipa_conn_t *)cp;
1377 		ac->ac_laddr = V4_PART_OF_V6(icmp->icmp_v6src);
1378 		ac->ac_faddr = V4_PART_OF_V6(icmp->icmp_v6dst);
1379 		ac->ac_fport = fport;
1380 		ac->ac_lport = 0;
1381 		break;
1382 
1383 	case sizeof (ipa6_conn_t):
1384 		ASSERT(icmp->icmp_family == AF_INET6);
1385 		/* Append a request for an IRE */
1386 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1387 		if (mp->b_cont == NULL) {
1388 			freemsg(mp);
1389 			return (NULL);
1390 		}
1391 		mp->b_cont->b_wptr += sizeof (ire_t);
1392 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1393 
1394 		/* cp known to be 32 bit aligned */
1395 		ac6 = (ipa6_conn_t *)cp;
1396 		ac6->ac6_laddr = icmp->icmp_v6src;
1397 		ac6->ac6_faddr = icmp->icmp_v6dst;
1398 		ac6->ac6_fport = fport;
1399 		ac6->ac6_lport = 0;
1400 		break;
1401 
1402 	case sizeof (sin_t):
1403 		ASSERT(icmp->icmp_family == AF_INET);
1404 		/* Append a request for an IRE */
1405 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1406 		if (!mp->b_cont) {
1407 			freemsg(mp);
1408 			return (NULL);
1409 		}
1410 		mp->b_cont->b_wptr += sizeof (ire_t);
1411 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1412 
1413 		sin = (sin_t *)cp;
1414 		*sin = sin_null;
1415 		sin->sin_family = AF_INET;
1416 		sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_bound_v6src);
1417 		break;
1418 
1419 	case sizeof (sin6_t):
1420 		ASSERT(icmp->icmp_family == AF_INET6);
1421 		/* Append a request for an IRE */
1422 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1423 		if (!mp->b_cont) {
1424 			freemsg(mp);
1425 			return (NULL);
1426 		}
1427 		mp->b_cont->b_wptr += sizeof (ire_t);
1428 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1429 
1430 		sin6 = (sin6_t *)cp;
1431 		*sin6 = sin6_null;
1432 		sin6->sin6_family = AF_INET6;
1433 		sin6->sin6_addr = icmp->icmp_bound_v6src;
1434 		break;
1435 	}
1436 	/* Add protocol number to end */
1437 	cp[addr_length] = icmp->icmp_proto;
1438 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
1439 	return (mp);
1440 }
1441 
1442 /* For /dev/icmp aka AF_INET open */
1443 static int
1444 icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1445 {
1446 	return (icmp_open(q, devp, flag, sflag, credp, B_FALSE));
1447 }
1448 
1449 /* For /dev/icmp6 aka AF_INET6 open */
1450 static int
1451 icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1452 {
1453 	return (icmp_open(q, devp, flag, sflag, credp, B_TRUE));
1454 }
1455 
1456 /*
1457  * This is the open routine for icmp.  It allocates a icmp_t structure for
1458  * the stream and, on the first open of the module, creates an ND table.
1459  */
1460 /*ARGSUSED2*/
1461 static int
1462 icmp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1463     boolean_t isv6)
1464 {
1465 	int	err;
1466 	icmp_t	*icmp;
1467 	conn_t *connp;
1468 	dev_t	conn_dev;
1469 	zoneid_t zoneid;
1470 	netstack_t *ns;
1471 	icmp_stack_t *is;
1472 
1473 	/* If the stream is already open, return immediately. */
1474 	if (q->q_ptr != NULL)
1475 		return (0);
1476 
1477 	if (sflag == MODOPEN)
1478 		return (EINVAL);
1479 
1480 	ns = netstack_find_by_cred(credp);
1481 	ASSERT(ns != NULL);
1482 	is = ns->netstack_icmp;
1483 	ASSERT(is != NULL);
1484 
1485 	/*
1486 	 * For exclusive stacks we set the zoneid to zero
1487 	 * to make ICMP operate as if in the global zone.
1488 	 */
1489 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
1490 		zoneid = GLOBAL_ZONEID;
1491 	else
1492 		zoneid = crgetzoneid(credp);
1493 
1494 	if ((conn_dev = inet_minor_alloc(ip_minor_arena)) == 0) {
1495 		netstack_rele(ns);
1496 		return (EBUSY);
1497 	}
1498 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1499 
1500 	connp = ipcl_conn_create(IPCL_RAWIPCONN, KM_SLEEP, ns);
1501 	connp->conn_dev = conn_dev;
1502 	icmp = connp->conn_icmp;
1503 
1504 	/*
1505 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
1506 	 * done by netstack_find_by_cred()
1507 	 */
1508 	netstack_rele(ns);
1509 
1510 	/*
1511 	 * Initialize the icmp_t structure for this stream.
1512 	 */
1513 	q->q_ptr = connp;
1514 	WR(q)->q_ptr = connp;
1515 	connp->conn_rq = q;
1516 	connp->conn_wq = WR(q);
1517 
1518 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1519 	ASSERT(connp->conn_ulp == IPPROTO_ICMP);
1520 	ASSERT(connp->conn_icmp == icmp);
1521 	ASSERT(icmp->icmp_connp == connp);
1522 
1523 	/* Set the initial state of the stream and the privilege status. */
1524 	icmp->icmp_state = TS_UNBND;
1525 	if (isv6) {
1526 		icmp->icmp_ipversion = IPV6_VERSION;
1527 		icmp->icmp_family = AF_INET6;
1528 		connp->conn_ulp = IPPROTO_ICMPV6;
1529 		/* May be changed by a SO_PROTOTYPE socket option. */
1530 		icmp->icmp_proto = IPPROTO_ICMPV6;
1531 		icmp->icmp_checksum_off = 2;	/* Offset for icmp6_cksum */
1532 		icmp->icmp_max_hdr_len = IPV6_HDR_LEN;
1533 		icmp->icmp_ttl = (uint8_t)is->is_ipv6_hoplimit;
1534 		connp->conn_af_isv6 = B_TRUE;
1535 		connp->conn_flags |= IPCL_ISV6;
1536 	} else {
1537 		icmp->icmp_ipversion = IPV4_VERSION;
1538 		icmp->icmp_family = AF_INET;
1539 		/* May be changed by a SO_PROTOTYPE socket option. */
1540 		icmp->icmp_proto = IPPROTO_ICMP;
1541 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH;
1542 		icmp->icmp_ttl = (uint8_t)is->is_ipv4_ttl;
1543 		connp->conn_af_isv6 = B_FALSE;
1544 		connp->conn_flags &= ~IPCL_ISV6;
1545 	}
1546 	icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1547 	icmp->icmp_pending_op = -1;
1548 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1549 	connp->conn_zoneid = zoneid;
1550 
1551 	/*
1552 	 * If the caller has the process-wide flag set, then default to MAC
1553 	 * exempt mode.  This allows read-down to unlabeled hosts.
1554 	 */
1555 	if (getpflags(NET_MAC_AWARE, credp) != 0)
1556 		icmp->icmp_mac_exempt = B_TRUE;
1557 
1558 	connp->conn_ulp_labeled = is_system_labeled();
1559 
1560 	icmp->icmp_is = is;
1561 
1562 	q->q_hiwat = is->is_recv_hiwat;
1563 	WR(q)->q_hiwat = is->is_xmit_hiwat;
1564 	WR(q)->q_lowat = is->is_xmit_lowat;
1565 
1566 	connp->conn_recv = icmp_input;
1567 	crhold(credp);
1568 	connp->conn_cred = credp;
1569 
1570 	mutex_enter(&connp->conn_lock);
1571 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1572 	mutex_exit(&connp->conn_lock);
1573 
1574 	qprocson(q);
1575 
1576 	if (icmp->icmp_family == AF_INET6) {
1577 		/* Build initial header template for transmit */
1578 		if ((err = icmp_build_hdrs(icmp)) != 0) {
1579 			rw_exit(&icmp->icmp_rwlock);
1580 			qprocsoff(q);
1581 			ipcl_conn_destroy(connp);
1582 			return (err);
1583 		}
1584 	}
1585 	rw_exit(&icmp->icmp_rwlock);
1586 
1587 	/* Set the Stream head write offset. */
1588 	(void) mi_set_sth_wroff(q,
1589 	    icmp->icmp_max_hdr_len + is->is_wroff_extra);
1590 	(void) mi_set_sth_hiwat(q, q->q_hiwat);
1591 
1592 	return (0);
1593 }
1594 
1595 /*
1596  * Which ICMP options OK to set through T_UNITDATA_REQ...
1597  */
1598 /* ARGSUSED */
1599 static boolean_t
1600 icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1601 {
1602 	return (B_TRUE);
1603 }
1604 
1605 /*
1606  * This routine gets default values of certain options whose default
1607  * values are maintained by protcol specific code
1608  */
1609 /* ARGSUSED */
1610 int
1611 icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr)
1612 {
1613 	icmp_t *icmp = Q_TO_ICMP(q);
1614 	icmp_stack_t *is = icmp->icmp_is;
1615 	int *i1 = (int *)ptr;
1616 
1617 	switch (level) {
1618 	case IPPROTO_IP:
1619 		switch (name) {
1620 		case IP_MULTICAST_TTL:
1621 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1622 			return (sizeof (uchar_t));
1623 		case IP_MULTICAST_LOOP:
1624 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1625 			return (sizeof (uchar_t));
1626 		}
1627 		break;
1628 	case IPPROTO_IPV6:
1629 		switch (name) {
1630 		case IPV6_MULTICAST_HOPS:
1631 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1632 			return (sizeof (int));
1633 		case IPV6_MULTICAST_LOOP:
1634 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1635 			return (sizeof (int));
1636 		case IPV6_UNICAST_HOPS:
1637 			*i1 = is->is_ipv6_hoplimit;
1638 			return (sizeof (int));
1639 		}
1640 		break;
1641 	case IPPROTO_ICMPV6:
1642 		switch (name) {
1643 		case ICMP6_FILTER:
1644 			/* Make it look like "pass all" */
1645 			ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1646 			return (sizeof (icmp6_filter_t));
1647 		}
1648 		break;
1649 	}
1650 	return (-1);
1651 }
1652 
1653 /*
1654  * This routine retrieves the current status of socket options.
1655  * It returns the size of the option retrieved.
1656  */
1657 int
1658 icmp_opt_get_locked(queue_t *q, int level, int name, uchar_t *ptr)
1659 {
1660 	conn_t	*connp = Q_TO_CONN(q);
1661 	icmp_t	*icmp = connp->conn_icmp;
1662 	icmp_stack_t *is = icmp->icmp_is;
1663 	int	*i1 = (int *)ptr;
1664 	ip6_pkt_t	*ipp = &icmp->icmp_sticky_ipp;
1665 
1666 	switch (level) {
1667 	case SOL_SOCKET:
1668 		switch (name) {
1669 		case SO_DEBUG:
1670 			*i1 = icmp->icmp_debug;
1671 			break;
1672 		case SO_TYPE:
1673 			*i1 = SOCK_RAW;
1674 			break;
1675 		case SO_PROTOTYPE:
1676 			*i1 = icmp->icmp_proto;
1677 			break;
1678 		case SO_REUSEADDR:
1679 			*i1 = icmp->icmp_reuseaddr;
1680 			break;
1681 
1682 		/*
1683 		 * The following three items are available here,
1684 		 * but are only meaningful to IP.
1685 		 */
1686 		case SO_DONTROUTE:
1687 			*i1 = icmp->icmp_dontroute;
1688 			break;
1689 		case SO_USELOOPBACK:
1690 			*i1 = icmp->icmp_useloopback;
1691 			break;
1692 		case SO_BROADCAST:
1693 			*i1 = icmp->icmp_broadcast;
1694 			break;
1695 
1696 		case SO_SNDBUF:
1697 			ASSERT(q->q_hiwat <= INT_MAX);
1698 			*i1 = (int)q->q_hiwat;
1699 			break;
1700 		case SO_RCVBUF:
1701 			ASSERT(RD(q)->q_hiwat <= INT_MAX);
1702 			*i1 = (int)RD(q)->q_hiwat;
1703 			break;
1704 		case SO_DGRAM_ERRIND:
1705 			*i1 = icmp->icmp_dgram_errind;
1706 			break;
1707 		case SO_TIMESTAMP:
1708 			*i1 = icmp->icmp_timestamp;
1709 			break;
1710 		case SO_MAC_EXEMPT:
1711 			*i1 = icmp->icmp_mac_exempt;
1712 			break;
1713 		case SO_DOMAIN:
1714 			*i1 = icmp->icmp_family;
1715 			break;
1716 
1717 		/*
1718 		 * Following four not meaningful for icmp
1719 		 * Action is same as "default" to which we fallthrough
1720 		 * so we keep them in comments.
1721 		 * case SO_LINGER:
1722 		 * case SO_KEEPALIVE:
1723 		 * case SO_OOBINLINE:
1724 		 * case SO_ALLZONES:
1725 		 */
1726 		default:
1727 			return (-1);
1728 		}
1729 		break;
1730 	case IPPROTO_IP:
1731 		/*
1732 		 * Only allow IPv4 option processing on IPv4 sockets.
1733 		 */
1734 		if (icmp->icmp_family != AF_INET)
1735 			return (-1);
1736 
1737 		switch (name) {
1738 		case IP_OPTIONS:
1739 		case T_IP_OPTIONS:
1740 			/* Options are passed up with each packet */
1741 			return (0);
1742 		case IP_HDRINCL:
1743 			*i1 = (int)icmp->icmp_hdrincl;
1744 			break;
1745 		case IP_TOS:
1746 		case T_IP_TOS:
1747 			*i1 = (int)icmp->icmp_type_of_service;
1748 			break;
1749 		case IP_TTL:
1750 			*i1 = (int)icmp->icmp_ttl;
1751 			break;
1752 		case IP_MULTICAST_IF:
1753 			/* 0 address if not set */
1754 			*(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr;
1755 			return (sizeof (ipaddr_t));
1756 		case IP_MULTICAST_TTL:
1757 			*(uchar_t *)ptr = icmp->icmp_multicast_ttl;
1758 			return (sizeof (uchar_t));
1759 		case IP_MULTICAST_LOOP:
1760 			*ptr = connp->conn_multicast_loop;
1761 			return (sizeof (uint8_t));
1762 		case IP_BOUND_IF:
1763 			/* Zero if not set */
1764 			*i1 = icmp->icmp_bound_if;
1765 			break;	/* goto sizeof (int) option return */
1766 		case IP_UNSPEC_SRC:
1767 			*ptr = icmp->icmp_unspec_source;
1768 			break;	/* goto sizeof (int) option return */
1769 		case IP_BROADCAST_TTL:
1770 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
1771 			return (sizeof (uchar_t));
1772 		case IP_RECVIF:
1773 			*ptr = icmp->icmp_recvif;
1774 			break;	/* goto sizeof (int) option return */
1775 		case IP_RECVPKTINFO:
1776 			/*
1777 			 * This also handles IP_PKTINFO.
1778 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
1779 			 * Differentiation is based on the size of the argument
1780 			 * passed in.
1781 			 * This option is handled in IP which will return an
1782 			 * error for IP_PKTINFO as it's not supported as a
1783 			 * sticky option.
1784 			 */
1785 			return (-EINVAL);
1786 		/*
1787 		 * Cannot "get" the value of following options
1788 		 * at this level. Action is same as "default" to
1789 		 * which we fallthrough so we keep them in comments.
1790 		 *
1791 		 * case IP_ADD_MEMBERSHIP:
1792 		 * case IP_DROP_MEMBERSHIP:
1793 		 * case IP_BLOCK_SOURCE:
1794 		 * case IP_UNBLOCK_SOURCE:
1795 		 * case IP_ADD_SOURCE_MEMBERSHIP:
1796 		 * case IP_DROP_SOURCE_MEMBERSHIP:
1797 		 * case MCAST_JOIN_GROUP:
1798 		 * case MCAST_LEAVE_GROUP:
1799 		 * case MCAST_BLOCK_SOURCE:
1800 		 * case MCAST_UNBLOCK_SOURCE:
1801 		 * case MCAST_JOIN_SOURCE_GROUP:
1802 		 * case MCAST_LEAVE_SOURCE_GROUP:
1803 		 * case MRT_INIT:
1804 		 * case MRT_DONE:
1805 		 * case MRT_ADD_VIF:
1806 		 * case MRT_DEL_VIF:
1807 		 * case MRT_ADD_MFC:
1808 		 * case MRT_DEL_MFC:
1809 		 * case MRT_VERSION:
1810 		 * case MRT_ASSERT:
1811 		 * case IP_SEC_OPT:
1812 		 * case IP_DONTFAILOVER_IF:
1813 		 * case IP_NEXTHOP:
1814 		 */
1815 		default:
1816 			return (-1);
1817 		}
1818 		break;
1819 	case IPPROTO_IPV6:
1820 		/*
1821 		 * Only allow IPv6 option processing on native IPv6 sockets.
1822 		 */
1823 		if (icmp->icmp_family != AF_INET6)
1824 			return (-1);
1825 		switch (name) {
1826 		case IPV6_UNICAST_HOPS:
1827 			*i1 = (unsigned int)icmp->icmp_ttl;
1828 			break;
1829 		case IPV6_MULTICAST_IF:
1830 			/* 0 index if not set */
1831 			*i1 = icmp->icmp_multicast_if_index;
1832 			break;
1833 		case IPV6_MULTICAST_HOPS:
1834 			*i1 = icmp->icmp_multicast_ttl;
1835 			break;
1836 		case IPV6_MULTICAST_LOOP:
1837 			*i1 = connp->conn_multicast_loop;
1838 			break;
1839 		case IPV6_BOUND_IF:
1840 			/* Zero if not set */
1841 			*i1 = icmp->icmp_bound_if;
1842 			break;
1843 		case IPV6_UNSPEC_SRC:
1844 			*i1 = icmp->icmp_unspec_source;
1845 			break;
1846 		case IPV6_CHECKSUM:
1847 			/*
1848 			 * Return offset or -1 if no checksum offset.
1849 			 * Does not apply to IPPROTO_ICMPV6
1850 			 */
1851 			if (icmp->icmp_proto == IPPROTO_ICMPV6)
1852 				return (-1);
1853 
1854 			if (icmp->icmp_raw_checksum) {
1855 				*i1 = icmp->icmp_checksum_off;
1856 			} else {
1857 				*i1 = -1;
1858 			}
1859 			break;
1860 		case IPV6_JOIN_GROUP:
1861 		case IPV6_LEAVE_GROUP:
1862 		case MCAST_JOIN_GROUP:
1863 		case MCAST_LEAVE_GROUP:
1864 		case MCAST_BLOCK_SOURCE:
1865 		case MCAST_UNBLOCK_SOURCE:
1866 		case MCAST_JOIN_SOURCE_GROUP:
1867 		case MCAST_LEAVE_SOURCE_GROUP:
1868 			/* cannot "get" the value for these */
1869 			return (-1);
1870 		case IPV6_RECVPKTINFO:
1871 			*i1 = icmp->icmp_ip_recvpktinfo;
1872 			break;
1873 		case IPV6_RECVTCLASS:
1874 			*i1 = icmp->icmp_ipv6_recvtclass;
1875 			break;
1876 		case IPV6_RECVPATHMTU:
1877 			*i1 = icmp->icmp_ipv6_recvpathmtu;
1878 			break;
1879 		case IPV6_V6ONLY:
1880 			*i1 = 1;
1881 			break;
1882 		case IPV6_RECVHOPLIMIT:
1883 			*i1 = icmp->icmp_ipv6_recvhoplimit;
1884 			break;
1885 		case IPV6_RECVHOPOPTS:
1886 			*i1 = icmp->icmp_ipv6_recvhopopts;
1887 			break;
1888 		case IPV6_RECVDSTOPTS:
1889 			*i1 = icmp->icmp_ipv6_recvdstopts;
1890 			break;
1891 		case _OLD_IPV6_RECVDSTOPTS:
1892 			*i1 = icmp->icmp_old_ipv6_recvdstopts;
1893 			break;
1894 		case IPV6_RECVRTHDRDSTOPTS:
1895 			*i1 = icmp->icmp_ipv6_recvrtdstopts;
1896 			break;
1897 		case IPV6_RECVRTHDR:
1898 			*i1 = icmp->icmp_ipv6_recvrthdr;
1899 			break;
1900 		case IPV6_PKTINFO: {
1901 			/* XXX assumes that caller has room for max size! */
1902 			struct in6_pktinfo *pkti;
1903 
1904 			pkti = (struct in6_pktinfo *)ptr;
1905 			if (ipp->ipp_fields & IPPF_IFINDEX)
1906 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
1907 			else
1908 				pkti->ipi6_ifindex = 0;
1909 			if (ipp->ipp_fields & IPPF_ADDR)
1910 				pkti->ipi6_addr = ipp->ipp_addr;
1911 			else
1912 				pkti->ipi6_addr = ipv6_all_zeros;
1913 			return (sizeof (struct in6_pktinfo));
1914 		}
1915 		case IPV6_NEXTHOP: {
1916 			sin6_t *sin6 = (sin6_t *)ptr;
1917 
1918 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
1919 				return (0);
1920 			*sin6 = sin6_null;
1921 			sin6->sin6_family = AF_INET6;
1922 			sin6->sin6_addr = ipp->ipp_nexthop;
1923 			return (sizeof (sin6_t));
1924 		}
1925 		case IPV6_HOPOPTS:
1926 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
1927 				return (0);
1928 			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
1929 				return (0);
1930 			bcopy((char *)ipp->ipp_hopopts +
1931 			    icmp->icmp_label_len_v6, ptr,
1932 			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1933 			if (icmp->icmp_label_len_v6 > 0) {
1934 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
1935 				ptr[1] = (ipp->ipp_hopoptslen -
1936 				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
1937 			}
1938 			return (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1939 		case IPV6_RTHDRDSTOPTS:
1940 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
1941 				return (0);
1942 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
1943 			return (ipp->ipp_rtdstoptslen);
1944 		case IPV6_RTHDR:
1945 			if (!(ipp->ipp_fields & IPPF_RTHDR))
1946 				return (0);
1947 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
1948 			return (ipp->ipp_rthdrlen);
1949 		case IPV6_DSTOPTS:
1950 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
1951 				return (0);
1952 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
1953 			return (ipp->ipp_dstoptslen);
1954 		case IPV6_PATHMTU:
1955 			if (!(ipp->ipp_fields & IPPF_PATHMTU))
1956 				return (0);
1957 
1958 			return (ip_fill_mtuinfo(&icmp->icmp_v6dst, 0,
1959 			    (struct ip6_mtuinfo *)ptr, is->is_netstack));
1960 		case IPV6_TCLASS:
1961 			if (ipp->ipp_fields & IPPF_TCLASS)
1962 				*i1 = ipp->ipp_tclass;
1963 			else
1964 				*i1 = IPV6_FLOW_TCLASS(
1965 				    IPV6_DEFAULT_VERS_AND_FLOW);
1966 			break;
1967 		default:
1968 			return (-1);
1969 		}
1970 		break;
1971 	case IPPROTO_ICMPV6:
1972 		/*
1973 		 * Only allow IPv6 option processing on native IPv6 sockets.
1974 		 */
1975 		if (icmp->icmp_family != AF_INET6)
1976 			return (-1);
1977 
1978 		if (icmp->icmp_proto != IPPROTO_ICMPV6)
1979 			return (-1);
1980 
1981 		switch (name) {
1982 		case ICMP6_FILTER:
1983 			if (icmp->icmp_filter == NULL) {
1984 				/* Make it look like "pass all" */
1985 				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1986 			} else {
1987 				(void) bcopy(icmp->icmp_filter, ptr,
1988 				    sizeof (icmp6_filter_t));
1989 			}
1990 			return (sizeof (icmp6_filter_t));
1991 		default:
1992 			return (-1);
1993 		}
1994 	default:
1995 		return (-1);
1996 	}
1997 	return (sizeof (int));
1998 }
1999 
2000 /*
2001  * This routine retrieves the current status of socket options.
2002  * It returns the size of the option retrieved.
2003  */
2004 int
2005 icmp_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
2006 {
2007 	icmp_t  *icmp = Q_TO_ICMP(q);
2008 	int 	err;
2009 
2010 	rw_enter(&icmp->icmp_rwlock, RW_READER);
2011 	err = icmp_opt_get_locked(q, level, name, ptr);
2012 	rw_exit(&icmp->icmp_rwlock);
2013 	return (err);
2014 }
2015 
2016 
2017 /* This routine sets socket options. */
2018 /* ARGSUSED */
2019 int
2020 icmp_opt_set_locked(queue_t *q, uint_t optset_context, int level, int name,
2021     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
2022     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
2023 {
2024 	conn_t	*connp = Q_TO_CONN(q);
2025 	icmp_t	*icmp = connp->conn_icmp;
2026 	icmp_stack_t *is = icmp->icmp_is;
2027 	int	*i1 = (int *)invalp;
2028 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2029 	boolean_t checkonly;
2030 	int	error;
2031 
2032 	switch (optset_context) {
2033 	case SETFN_OPTCOM_CHECKONLY:
2034 		checkonly = B_TRUE;
2035 		/*
2036 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
2037 		 * inlen != 0 implies value supplied and
2038 		 * 	we have to "pretend" to set it.
2039 		 * inlen == 0 implies that there is no
2040 		 * 	value part in T_CHECK request and just validation
2041 		 * done elsewhere should be enough, we just return here.
2042 		 */
2043 		if (inlen == 0) {
2044 			*outlenp = 0;
2045 			return (0);
2046 		}
2047 		break;
2048 	case SETFN_OPTCOM_NEGOTIATE:
2049 		checkonly = B_FALSE;
2050 		break;
2051 	case SETFN_UD_NEGOTIATE:
2052 	case SETFN_CONN_NEGOTIATE:
2053 		checkonly = B_FALSE;
2054 		/*
2055 		 * Negotiating local and "association-related" options
2056 		 * through T_UNITDATA_REQ.
2057 		 *
2058 		 * Following routine can filter out ones we do not
2059 		 * want to be "set" this way.
2060 		 */
2061 		if (!icmp_opt_allow_udr_set(level, name)) {
2062 			*outlenp = 0;
2063 			return (EINVAL);
2064 		}
2065 		break;
2066 	default:
2067 		/*
2068 		 * We should never get here
2069 		 */
2070 		*outlenp = 0;
2071 		return (EINVAL);
2072 	}
2073 
2074 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
2075 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
2076 
2077 	/*
2078 	 * For fixed length options, no sanity check
2079 	 * of passed in length is done. It is assumed *_optcom_req()
2080 	 * routines do the right thing.
2081 	 */
2082 
2083 	switch (level) {
2084 	case SOL_SOCKET:
2085 		switch (name) {
2086 		case SO_DEBUG:
2087 			if (!checkonly)
2088 				icmp->icmp_debug = onoff;
2089 			break;
2090 		case SO_PROTOTYPE:
2091 			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
2092 			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
2093 			    secpolicy_net_rawaccess(cr) != 0) {
2094 				*outlenp = 0;
2095 				return (EACCES);
2096 			}
2097 			/* Can't use IPPROTO_RAW with IPv6 */
2098 			if ((*i1 & 0xFF) == IPPROTO_RAW &&
2099 			    icmp->icmp_family == AF_INET6) {
2100 				*outlenp = 0;
2101 				return (EPROTONOSUPPORT);
2102 			}
2103 			if (checkonly) {
2104 				/* T_CHECK case */
2105 				*(int *)outvalp = (*i1 & 0xFF);
2106 				break;
2107 			}
2108 			icmp->icmp_proto = *i1 & 0xFF;
2109 			if ((icmp->icmp_proto == IPPROTO_RAW ||
2110 			    icmp->icmp_proto == IPPROTO_IGMP) &&
2111 			    icmp->icmp_family == AF_INET)
2112 				icmp->icmp_hdrincl = 1;
2113 			else
2114 				icmp->icmp_hdrincl = 0;
2115 
2116 			if (icmp->icmp_family == AF_INET6 &&
2117 			    icmp->icmp_proto == IPPROTO_ICMPV6) {
2118 				/* Set offset for icmp6_cksum */
2119 				icmp->icmp_raw_checksum = 0;
2120 				icmp->icmp_checksum_off = 2;
2121 			}
2122 			if (icmp->icmp_proto == IPPROTO_UDP ||
2123 			    icmp->icmp_proto == IPPROTO_TCP ||
2124 			    icmp->icmp_proto == IPPROTO_SCTP) {
2125 				icmp->icmp_no_tp_cksum = 1;
2126 				icmp->icmp_sticky_ipp.ipp_fields |=
2127 				    IPPF_NO_CKSUM;
2128 			} else {
2129 				icmp->icmp_no_tp_cksum = 0;
2130 				icmp->icmp_sticky_ipp.ipp_fields &=
2131 				    ~IPPF_NO_CKSUM;
2132 			}
2133 
2134 			if (icmp->icmp_filter != NULL &&
2135 			    icmp->icmp_proto != IPPROTO_ICMPV6) {
2136 				kmem_free(icmp->icmp_filter,
2137 				    sizeof (icmp6_filter_t));
2138 				icmp->icmp_filter = NULL;
2139 			}
2140 
2141 			/* Rebuild the header template */
2142 			error = icmp_build_hdrs(icmp);
2143 			if (error != 0) {
2144 				*outlenp = 0;
2145 				return (error);
2146 			}
2147 
2148 			/*
2149 			 * For SCTP, we don't use icmp_bind_proto() for
2150 			 * raw socket binding.  Note that we do not need
2151 			 * to set *outlenp.
2152 			 * FIXME: how does SCTP work?
2153 			 */
2154 			if (icmp->icmp_proto == IPPROTO_SCTP)
2155 				return (0);
2156 
2157 			*outlenp = sizeof (int);
2158 			*(int *)outvalp = *i1 & 0xFF;
2159 
2160 			/* Drop lock across the bind operation */
2161 			rw_exit(&icmp->icmp_rwlock);
2162 			icmp_bind_proto(q);
2163 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2164 			return (0);
2165 		case SO_REUSEADDR:
2166 			if (!checkonly)
2167 				icmp->icmp_reuseaddr = onoff;
2168 			break;
2169 
2170 		/*
2171 		 * The following three items are available here,
2172 		 * but are only meaningful to IP.
2173 		 */
2174 		case SO_DONTROUTE:
2175 			if (!checkonly)
2176 				icmp->icmp_dontroute = onoff;
2177 			break;
2178 		case SO_USELOOPBACK:
2179 			if (!checkonly)
2180 				icmp->icmp_useloopback = onoff;
2181 			break;
2182 		case SO_BROADCAST:
2183 			if (!checkonly)
2184 				icmp->icmp_broadcast = onoff;
2185 			break;
2186 
2187 		case SO_SNDBUF:
2188 			if (*i1 > is->is_max_buf) {
2189 				*outlenp = 0;
2190 				return (ENOBUFS);
2191 			}
2192 			if (!checkonly) {
2193 				q->q_hiwat = *i1;
2194 			}
2195 			break;
2196 		case SO_RCVBUF:
2197 			if (*i1 > is->is_max_buf) {
2198 				*outlenp = 0;
2199 				return (ENOBUFS);
2200 			}
2201 			if (!checkonly) {
2202 				RD(q)->q_hiwat = *i1;
2203 				rw_exit(&icmp->icmp_rwlock);
2204 				(void) mi_set_sth_hiwat(RD(q), *i1);
2205 				rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2206 			}
2207 			break;
2208 		case SO_DGRAM_ERRIND:
2209 			if (!checkonly)
2210 				icmp->icmp_dgram_errind = onoff;
2211 			break;
2212 		case SO_ALLZONES:
2213 			/*
2214 			 * "soft" error (negative)
2215 			 * option not handled at this level
2216 			 * Note: Do not modify *outlenp
2217 			 */
2218 			return (-EINVAL);
2219 		case SO_TIMESTAMP:
2220 			if (!checkonly) {
2221 				icmp->icmp_timestamp = onoff;
2222 			}
2223 			break;
2224 		case SO_MAC_EXEMPT:
2225 			if (secpolicy_net_mac_aware(cr) != 0 ||
2226 			    icmp->icmp_state != TS_UNBND)
2227 				return (EACCES);
2228 			if (!checkonly)
2229 				icmp->icmp_mac_exempt = onoff;
2230 			break;
2231 		/*
2232 		 * Following three not meaningful for icmp
2233 		 * Action is same as "default" so we keep them
2234 		 * in comments.
2235 		 * case SO_LINGER:
2236 		 * case SO_KEEPALIVE:
2237 		 * case SO_OOBINLINE:
2238 		 */
2239 		default:
2240 			*outlenp = 0;
2241 			return (EINVAL);
2242 		}
2243 		break;
2244 	case IPPROTO_IP:
2245 		/*
2246 		 * Only allow IPv4 option processing on IPv4 sockets.
2247 		 */
2248 		if (icmp->icmp_family != AF_INET) {
2249 			*outlenp = 0;
2250 			return (ENOPROTOOPT);
2251 		}
2252 		switch (name) {
2253 		case IP_OPTIONS:
2254 		case T_IP_OPTIONS:
2255 			/* Save options for use by IP. */
2256 			if ((inlen & 0x3) ||
2257 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2258 				*outlenp = 0;
2259 				return (EINVAL);
2260 			}
2261 			if (checkonly)
2262 				break;
2263 
2264 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2265 			    &icmp->icmp_ip_snd_options_len,
2266 			    icmp->icmp_label_len, invalp, inlen)) {
2267 				*outlenp = 0;
2268 				return (ENOMEM);
2269 			}
2270 
2271 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2272 			    icmp->icmp_ip_snd_options_len;
2273 			rw_exit(&icmp->icmp_rwlock);
2274 			(void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len +
2275 			    is->is_wroff_extra);
2276 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2277 			break;
2278 		case IP_HDRINCL:
2279 			if (!checkonly)
2280 				icmp->icmp_hdrincl = onoff;
2281 			break;
2282 		case IP_TOS:
2283 		case T_IP_TOS:
2284 			if (!checkonly) {
2285 				icmp->icmp_type_of_service = (uint8_t)*i1;
2286 			}
2287 			break;
2288 		case IP_TTL:
2289 			if (!checkonly) {
2290 				icmp->icmp_ttl = (uint8_t)*i1;
2291 			}
2292 			break;
2293 		case IP_MULTICAST_IF:
2294 			/*
2295 			 * TODO should check OPTMGMT reply and undo this if
2296 			 * there is an error.
2297 			 */
2298 			if (!checkonly)
2299 				icmp->icmp_multicast_if_addr = *i1;
2300 			break;
2301 		case IP_MULTICAST_TTL:
2302 			if (!checkonly)
2303 				icmp->icmp_multicast_ttl = *invalp;
2304 			break;
2305 		case IP_MULTICAST_LOOP:
2306 			if (!checkonly) {
2307 				connp->conn_multicast_loop =
2308 				    (*invalp == 0) ? 0 : 1;
2309 			}
2310 			break;
2311 		case IP_BOUND_IF:
2312 			if (!checkonly)
2313 				icmp->icmp_bound_if = *i1;
2314 			break;
2315 		case IP_UNSPEC_SRC:
2316 			if (!checkonly)
2317 				icmp->icmp_unspec_source = onoff;
2318 			break;
2319 		case IP_BROADCAST_TTL:
2320 			if (!checkonly)
2321 				connp->conn_broadcast_ttl = *invalp;
2322 			break;
2323 		case IP_RECVIF:
2324 			if (!checkonly)
2325 				icmp->icmp_recvif = onoff;
2326 			/*
2327 			 * pass to ip
2328 			 */
2329 			return (-EINVAL);
2330 		case IP_PKTINFO: {
2331 			/*
2332 			 * This also handles IP_RECVPKTINFO.
2333 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2334 			 * Differentiation is based on the size of the argument
2335 			 * passed in.
2336 			 */
2337 			struct in_pktinfo *pktinfop;
2338 			ip4_pkt_t *attr_pktinfop;
2339 
2340 			if (checkonly)
2341 				break;
2342 
2343 			if (inlen == sizeof (int)) {
2344 				/*
2345 				 * This is IP_RECVPKTINFO option.
2346 				 * Keep a local copy of wether this option is
2347 				 * set or not and pass it down to IP for
2348 				 * processing.
2349 				 */
2350 				icmp->icmp_ip_recvpktinfo = onoff;
2351 				return (-EINVAL);
2352 			}
2353 
2354 
2355 			if (inlen != sizeof (struct in_pktinfo))
2356 				return (EINVAL);
2357 
2358 			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
2359 			    == NULL) {
2360 				/*
2361 				 * sticky option is not supported
2362 				 */
2363 				return (EINVAL);
2364 			}
2365 
2366 			pktinfop = (struct in_pktinfo *)invalp;
2367 
2368 			/*
2369 			 * Atleast one of the values should be specified
2370 			 */
2371 			if (pktinfop->ipi_ifindex == 0 &&
2372 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2373 				return (EINVAL);
2374 			}
2375 
2376 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2377 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2378 		}
2379 			break;
2380 		case IP_ADD_MEMBERSHIP:
2381 		case IP_DROP_MEMBERSHIP:
2382 		case IP_BLOCK_SOURCE:
2383 		case IP_UNBLOCK_SOURCE:
2384 		case IP_ADD_SOURCE_MEMBERSHIP:
2385 		case IP_DROP_SOURCE_MEMBERSHIP:
2386 		case MCAST_JOIN_GROUP:
2387 		case MCAST_LEAVE_GROUP:
2388 		case MCAST_BLOCK_SOURCE:
2389 		case MCAST_UNBLOCK_SOURCE:
2390 		case MCAST_JOIN_SOURCE_GROUP:
2391 		case MCAST_LEAVE_SOURCE_GROUP:
2392 		case MRT_INIT:
2393 		case MRT_DONE:
2394 		case MRT_ADD_VIF:
2395 		case MRT_DEL_VIF:
2396 		case MRT_ADD_MFC:
2397 		case MRT_DEL_MFC:
2398 		case MRT_VERSION:
2399 		case MRT_ASSERT:
2400 		case IP_SEC_OPT:
2401 		case IP_DONTFAILOVER_IF:
2402 		case IP_NEXTHOP:
2403 			/*
2404 			 * "soft" error (negative)
2405 			 * option not handled at this level
2406 			 * Note: Do not modify *outlenp
2407 			 */
2408 			return (-EINVAL);
2409 		default:
2410 			*outlenp = 0;
2411 			return (EINVAL);
2412 		}
2413 		break;
2414 	case IPPROTO_IPV6: {
2415 		ip6_pkt_t		*ipp;
2416 		boolean_t		sticky;
2417 
2418 		if (icmp->icmp_family != AF_INET6) {
2419 			*outlenp = 0;
2420 			return (ENOPROTOOPT);
2421 		}
2422 		/*
2423 		 * Deal with both sticky options and ancillary data
2424 		 */
2425 		if (thisdg_attrs == NULL) {
2426 			/* sticky options, or none */
2427 			ipp = &icmp->icmp_sticky_ipp;
2428 			sticky = B_TRUE;
2429 		} else {
2430 			/* ancillary data */
2431 			ipp = (ip6_pkt_t *)thisdg_attrs;
2432 			sticky = B_FALSE;
2433 		}
2434 
2435 		switch (name) {
2436 		case IPV6_MULTICAST_IF:
2437 			if (!checkonly)
2438 				icmp->icmp_multicast_if_index = *i1;
2439 			break;
2440 		case IPV6_UNICAST_HOPS:
2441 			/* -1 means use default */
2442 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2443 				*outlenp = 0;
2444 				return (EINVAL);
2445 			}
2446 			if (!checkonly) {
2447 				if (*i1 == -1) {
2448 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2449 					    is->is_ipv6_hoplimit;
2450 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2451 					/* Pass modified value to IP. */
2452 					*i1 = ipp->ipp_hoplimit;
2453 				} else {
2454 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2455 					    (uint8_t)*i1;
2456 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2457 				}
2458 				/* Rebuild the header template */
2459 				error = icmp_build_hdrs(icmp);
2460 				if (error != 0) {
2461 					*outlenp = 0;
2462 					return (error);
2463 				}
2464 			}
2465 			break;
2466 		case IPV6_MULTICAST_HOPS:
2467 			/* -1 means use default */
2468 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2469 				*outlenp = 0;
2470 				return (EINVAL);
2471 			}
2472 			if (!checkonly) {
2473 				if (*i1 == -1) {
2474 					icmp->icmp_multicast_ttl =
2475 					    ipp->ipp_multicast_hops =
2476 					    IP_DEFAULT_MULTICAST_TTL;
2477 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2478 					/* Pass modified value to IP. */
2479 					*i1 = icmp->icmp_multicast_ttl;
2480 				} else {
2481 					icmp->icmp_multicast_ttl =
2482 					    ipp->ipp_multicast_hops =
2483 					    (uint8_t)*i1;
2484 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2485 				}
2486 			}
2487 			break;
2488 		case IPV6_MULTICAST_LOOP:
2489 			if (*i1 != 0 && *i1 != 1) {
2490 				*outlenp = 0;
2491 				return (EINVAL);
2492 			}
2493 			if (!checkonly)
2494 				connp->conn_multicast_loop = *i1;
2495 			break;
2496 		case IPV6_CHECKSUM:
2497 			/*
2498 			 * Integer offset into the user data of where the
2499 			 * checksum is located.
2500 			 * Offset of -1 disables option.
2501 			 * Does not apply to IPPROTO_ICMPV6.
2502 			 */
2503 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2504 				*outlenp = 0;
2505 				return (EINVAL);
2506 			}
2507 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2508 				/* Negative or not 16 bit aligned offset */
2509 				*outlenp = 0;
2510 				return (EINVAL);
2511 			}
2512 			if (checkonly)
2513 				break;
2514 
2515 			if (*i1 == -1) {
2516 				icmp->icmp_raw_checksum = 0;
2517 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2518 			} else {
2519 				icmp->icmp_raw_checksum = 1;
2520 				icmp->icmp_checksum_off = *i1;
2521 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2522 			}
2523 			/* Rebuild the header template */
2524 			error = icmp_build_hdrs(icmp);
2525 			if (error != 0) {
2526 				*outlenp = 0;
2527 				return (error);
2528 			}
2529 			break;
2530 		case IPV6_JOIN_GROUP:
2531 		case IPV6_LEAVE_GROUP:
2532 		case MCAST_JOIN_GROUP:
2533 		case MCAST_LEAVE_GROUP:
2534 		case MCAST_BLOCK_SOURCE:
2535 		case MCAST_UNBLOCK_SOURCE:
2536 		case MCAST_JOIN_SOURCE_GROUP:
2537 		case MCAST_LEAVE_SOURCE_GROUP:
2538 			/*
2539 			 * "soft" error (negative)
2540 			 * option not handled at this level
2541 			 * Note: Do not modify *outlenp
2542 			 */
2543 			return (-EINVAL);
2544 		case IPV6_BOUND_IF:
2545 			if (!checkonly)
2546 				icmp->icmp_bound_if = *i1;
2547 			break;
2548 		case IPV6_UNSPEC_SRC:
2549 			if (!checkonly)
2550 				icmp->icmp_unspec_source = onoff;
2551 			break;
2552 		case IPV6_RECVTCLASS:
2553 			if (!checkonly)
2554 				icmp->icmp_ipv6_recvtclass = onoff;
2555 			break;
2556 		/*
2557 		 * Set boolean switches for ancillary data delivery
2558 		 */
2559 		case IPV6_RECVPKTINFO:
2560 			if (!checkonly)
2561 				icmp->icmp_ip_recvpktinfo = onoff;
2562 			break;
2563 		case IPV6_RECVPATHMTU:
2564 			if (!checkonly)
2565 				icmp->icmp_ipv6_recvpathmtu = onoff;
2566 			break;
2567 		case IPV6_RECVHOPLIMIT:
2568 			if (!checkonly)
2569 				icmp->icmp_ipv6_recvhoplimit = onoff;
2570 			break;
2571 		case IPV6_RECVHOPOPTS:
2572 			if (!checkonly)
2573 				icmp->icmp_ipv6_recvhopopts = onoff;
2574 			break;
2575 		case IPV6_RECVDSTOPTS:
2576 			if (!checkonly)
2577 				icmp->icmp_ipv6_recvdstopts = onoff;
2578 			break;
2579 		case _OLD_IPV6_RECVDSTOPTS:
2580 			if (!checkonly)
2581 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2582 			break;
2583 		case IPV6_RECVRTHDRDSTOPTS:
2584 			if (!checkonly)
2585 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2586 			break;
2587 		case IPV6_RECVRTHDR:
2588 			if (!checkonly)
2589 				icmp->icmp_ipv6_recvrthdr = onoff;
2590 			break;
2591 		/*
2592 		 * Set sticky options or ancillary data.
2593 		 * If sticky options, (re)build any extension headers
2594 		 * that might be needed as a result.
2595 		 */
2596 		case IPV6_PKTINFO:
2597 			/*
2598 			 * The source address and ifindex are verified
2599 			 * in ip_opt_set(). For ancillary data the
2600 			 * source address is checked in ip_wput_v6.
2601 			 */
2602 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2603 				return (EINVAL);
2604 			if (checkonly)
2605 				break;
2606 
2607 			if (inlen == 0) {
2608 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2609 				ipp->ipp_sticky_ignored |=
2610 				    (IPPF_IFINDEX|IPPF_ADDR);
2611 			} else {
2612 				struct in6_pktinfo *pkti;
2613 
2614 				pkti = (struct in6_pktinfo *)invalp;
2615 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2616 				ipp->ipp_addr = pkti->ipi6_addr;
2617 				if (ipp->ipp_ifindex != 0)
2618 					ipp->ipp_fields |= IPPF_IFINDEX;
2619 				else
2620 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2621 				if (!IN6_IS_ADDR_UNSPECIFIED(
2622 				    &ipp->ipp_addr))
2623 					ipp->ipp_fields |= IPPF_ADDR;
2624 				else
2625 					ipp->ipp_fields &= ~IPPF_ADDR;
2626 			}
2627 			if (sticky) {
2628 				error = icmp_build_hdrs(icmp);
2629 				if (error != 0)
2630 					return (error);
2631 			}
2632 			break;
2633 		case IPV6_HOPLIMIT:
2634 			/* This option can only be used as ancillary data. */
2635 			if (sticky)
2636 				return (EINVAL);
2637 			if (inlen != 0 && inlen != sizeof (int))
2638 				return (EINVAL);
2639 			if (checkonly)
2640 				break;
2641 
2642 			if (inlen == 0) {
2643 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2644 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2645 			} else {
2646 				if (*i1 > 255 || *i1 < -1)
2647 					return (EINVAL);
2648 				if (*i1 == -1)
2649 					ipp->ipp_hoplimit =
2650 					    is->is_ipv6_hoplimit;
2651 				else
2652 					ipp->ipp_hoplimit = *i1;
2653 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2654 			}
2655 			break;
2656 		case IPV6_TCLASS:
2657 			/*
2658 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2659 			 * and [0, 255] as the actualy traffic class.
2660 			 */
2661 			if (inlen != 0 && inlen != sizeof (int))
2662 				return (EINVAL);
2663 			if (checkonly)
2664 				break;
2665 
2666 			if (inlen == 0) {
2667 				ipp->ipp_fields &= ~IPPF_TCLASS;
2668 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2669 			} else {
2670 				if (*i1 >= 256 || *i1 < -1)
2671 					return (EINVAL);
2672 				if (*i1 == -1) {
2673 					ipp->ipp_tclass =
2674 					    IPV6_FLOW_TCLASS(
2675 					    IPV6_DEFAULT_VERS_AND_FLOW);
2676 				} else {
2677 					ipp->ipp_tclass = *i1;
2678 				}
2679 				ipp->ipp_fields |= IPPF_TCLASS;
2680 			}
2681 			if (sticky) {
2682 				error = icmp_build_hdrs(icmp);
2683 				if (error != 0)
2684 					return (error);
2685 			}
2686 			break;
2687 		case IPV6_NEXTHOP:
2688 			/*
2689 			 * IP will verify that the nexthop is reachable
2690 			 * and fail for sticky options.
2691 			 */
2692 			if (inlen != 0 && inlen != sizeof (sin6_t))
2693 				return (EINVAL);
2694 			if (checkonly)
2695 				break;
2696 
2697 			if (inlen == 0) {
2698 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2699 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2700 			} else {
2701 				sin6_t *sin6 = (sin6_t *)invalp;
2702 
2703 				if (sin6->sin6_family != AF_INET6)
2704 					return (EAFNOSUPPORT);
2705 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
2706 					return (EADDRNOTAVAIL);
2707 				ipp->ipp_nexthop = sin6->sin6_addr;
2708 				if (!IN6_IS_ADDR_UNSPECIFIED(
2709 				    &ipp->ipp_nexthop))
2710 					ipp->ipp_fields |= IPPF_NEXTHOP;
2711 				else
2712 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2713 			}
2714 			if (sticky) {
2715 				error = icmp_build_hdrs(icmp);
2716 				if (error != 0)
2717 					return (error);
2718 			}
2719 			break;
2720 		case IPV6_HOPOPTS: {
2721 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2722 			/*
2723 			 * Sanity checks - minimum size, size a multiple of
2724 			 * eight bytes, and matching size passed in.
2725 			 */
2726 			if (inlen != 0 &&
2727 			    inlen != (8 * (hopts->ip6h_len + 1)))
2728 				return (EINVAL);
2729 
2730 			if (checkonly)
2731 				break;
2732 			error = optcom_pkt_set(invalp, inlen, sticky,
2733 			    (uchar_t **)&ipp->ipp_hopopts,
2734 			    &ipp->ipp_hopoptslen,
2735 			    sticky ? icmp->icmp_label_len_v6 : 0);
2736 			if (error != 0)
2737 				return (error);
2738 			if (ipp->ipp_hopoptslen == 0) {
2739 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2740 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2741 			} else {
2742 				ipp->ipp_fields |= IPPF_HOPOPTS;
2743 			}
2744 			if (sticky) {
2745 				error = icmp_build_hdrs(icmp);
2746 				if (error != 0)
2747 					return (error);
2748 			}
2749 			break;
2750 		}
2751 		case IPV6_RTHDRDSTOPTS: {
2752 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2753 
2754 			/*
2755 			 * Sanity checks - minimum size, size a multiple of
2756 			 * eight bytes, and matching size passed in.
2757 			 */
2758 			if (inlen != 0 &&
2759 			    inlen != (8 * (dopts->ip6d_len + 1)))
2760 				return (EINVAL);
2761 
2762 			if (checkonly)
2763 				break;
2764 
2765 			if (inlen == 0) {
2766 				if (sticky &&
2767 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2768 					kmem_free(ipp->ipp_rtdstopts,
2769 					    ipp->ipp_rtdstoptslen);
2770 					ipp->ipp_rtdstopts = NULL;
2771 					ipp->ipp_rtdstoptslen = 0;
2772 				}
2773 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2774 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2775 			} else {
2776 				error = optcom_pkt_set(invalp, inlen, sticky,
2777 				    (uchar_t **)&ipp->ipp_rtdstopts,
2778 				    &ipp->ipp_rtdstoptslen, 0);
2779 				if (error != 0)
2780 					return (error);
2781 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2782 			}
2783 			if (sticky) {
2784 				error = icmp_build_hdrs(icmp);
2785 				if (error != 0)
2786 					return (error);
2787 			}
2788 			break;
2789 		}
2790 		case IPV6_DSTOPTS: {
2791 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2792 
2793 			/*
2794 			 * Sanity checks - minimum size, size a multiple of
2795 			 * eight bytes, and matching size passed in.
2796 			 */
2797 			if (inlen != 0 &&
2798 			    inlen != (8 * (dopts->ip6d_len + 1)))
2799 				return (EINVAL);
2800 
2801 			if (checkonly)
2802 				break;
2803 
2804 			if (inlen == 0) {
2805 				if (sticky &&
2806 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2807 					kmem_free(ipp->ipp_dstopts,
2808 					    ipp->ipp_dstoptslen);
2809 					ipp->ipp_dstopts = NULL;
2810 					ipp->ipp_dstoptslen = 0;
2811 				}
2812 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2813 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2814 			} else {
2815 				error = optcom_pkt_set(invalp, inlen, sticky,
2816 				    (uchar_t **)&ipp->ipp_dstopts,
2817 				    &ipp->ipp_dstoptslen, 0);
2818 				if (error != 0)
2819 					return (error);
2820 				ipp->ipp_fields |= IPPF_DSTOPTS;
2821 			}
2822 			if (sticky) {
2823 				error = icmp_build_hdrs(icmp);
2824 				if (error != 0)
2825 					return (error);
2826 			}
2827 			break;
2828 		}
2829 		case IPV6_RTHDR: {
2830 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2831 
2832 			/*
2833 			 * Sanity checks - minimum size, size a multiple of
2834 			 * eight bytes, and matching size passed in.
2835 			 */
2836 			if (inlen != 0 &&
2837 			    inlen != (8 * (rt->ip6r_len + 1)))
2838 				return (EINVAL);
2839 
2840 			if (checkonly)
2841 				break;
2842 
2843 			if (inlen == 0) {
2844 				if (sticky &&
2845 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2846 					kmem_free(ipp->ipp_rthdr,
2847 					    ipp->ipp_rthdrlen);
2848 					ipp->ipp_rthdr = NULL;
2849 					ipp->ipp_rthdrlen = 0;
2850 				}
2851 				ipp->ipp_fields &= ~IPPF_RTHDR;
2852 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2853 			} else {
2854 				error = optcom_pkt_set(invalp, inlen, sticky,
2855 				    (uchar_t **)&ipp->ipp_rthdr,
2856 				    &ipp->ipp_rthdrlen, 0);
2857 				if (error != 0)
2858 					return (error);
2859 				ipp->ipp_fields |= IPPF_RTHDR;
2860 			}
2861 			if (sticky) {
2862 				error = icmp_build_hdrs(icmp);
2863 				if (error != 0)
2864 					return (error);
2865 			}
2866 			break;
2867 		}
2868 
2869 		case IPV6_DONTFRAG:
2870 			if (checkonly)
2871 				break;
2872 
2873 			if (onoff) {
2874 				ipp->ipp_fields |= IPPF_DONTFRAG;
2875 			} else {
2876 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2877 			}
2878 			break;
2879 
2880 		case IPV6_USE_MIN_MTU:
2881 			if (inlen != sizeof (int))
2882 				return (EINVAL);
2883 
2884 			if (*i1 < -1 || *i1 > 1)
2885 				return (EINVAL);
2886 
2887 			if (checkonly)
2888 				break;
2889 
2890 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
2891 			ipp->ipp_use_min_mtu = *i1;
2892 			break;
2893 
2894 		/*
2895 		 * This option can't be set.  Its only returned via
2896 		 * getsockopt() or ancillary data.
2897 		 */
2898 		case IPV6_PATHMTU:
2899 			return (EINVAL);
2900 
2901 		case IPV6_BOUND_PIF:
2902 		case IPV6_SEC_OPT:
2903 		case IPV6_DONTFAILOVER_IF:
2904 		case IPV6_SRC_PREFERENCES:
2905 		case IPV6_V6ONLY:
2906 			/* Handled at IP level */
2907 			return (-EINVAL);
2908 		default:
2909 			*outlenp = 0;
2910 			return (EINVAL);
2911 		}
2912 		break;
2913 	}		/* end IPPROTO_IPV6 */
2914 
2915 	case IPPROTO_ICMPV6:
2916 		/*
2917 		 * Only allow IPv6 option processing on IPv6 sockets.
2918 		 */
2919 		if (icmp->icmp_family != AF_INET6) {
2920 			*outlenp = 0;
2921 			return (ENOPROTOOPT);
2922 		}
2923 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
2924 			*outlenp = 0;
2925 			return (ENOPROTOOPT);
2926 		}
2927 		switch (name) {
2928 		case ICMP6_FILTER:
2929 			if (!checkonly) {
2930 				if ((inlen != 0) &&
2931 				    (inlen != sizeof (icmp6_filter_t)))
2932 					return (EINVAL);
2933 
2934 				if (inlen == 0) {
2935 					if (icmp->icmp_filter != NULL) {
2936 						kmem_free(icmp->icmp_filter,
2937 						    sizeof (icmp6_filter_t));
2938 						icmp->icmp_filter = NULL;
2939 					}
2940 				} else {
2941 					if (icmp->icmp_filter == NULL) {
2942 						icmp->icmp_filter = kmem_alloc(
2943 						    sizeof (icmp6_filter_t),
2944 						    KM_NOSLEEP);
2945 						if (icmp->icmp_filter == NULL) {
2946 							*outlenp = 0;
2947 							return (ENOBUFS);
2948 						}
2949 					}
2950 					(void) bcopy(invalp, icmp->icmp_filter,
2951 					    inlen);
2952 				}
2953 			}
2954 			break;
2955 
2956 		default:
2957 			*outlenp = 0;
2958 			return (EINVAL);
2959 		}
2960 		break;
2961 	default:
2962 		*outlenp = 0;
2963 		return (EINVAL);
2964 	}
2965 	/*
2966 	 * Common case of OK return with outval same as inval.
2967 	 */
2968 	if (invalp != outvalp) {
2969 		/* don't trust bcopy for identical src/dst */
2970 		(void) bcopy(invalp, outvalp, inlen);
2971 	}
2972 	*outlenp = inlen;
2973 	return (0);
2974 }
2975 /* This routine sets socket options. */
2976 /* ARGSUSED */
2977 int
2978 icmp_opt_set(queue_t *q, uint_t optset_context, int level, int name,
2979     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
2980     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
2981 {
2982 	icmp_t	*icmp;
2983 	int	err;
2984 
2985 	icmp = Q_TO_ICMP(q);
2986 
2987 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2988 	err = icmp_opt_set_locked(q, optset_context, level, name, inlen, invalp,
2989 	    outlenp, outvalp, thisdg_attrs, cr, mblk);
2990 	rw_exit(&icmp->icmp_rwlock);
2991 	return (err);
2992 }
2993 
2994 /*
2995  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
2996  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
2997  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
2998  * headers.
2999  * Returns failure if can't allocate memory.
3000  */
3001 static int
3002 icmp_build_hdrs(icmp_t *icmp)
3003 {
3004 	icmp_stack_t *is = icmp->icmp_is;
3005 	uchar_t	*hdrs;
3006 	uint_t	hdrs_len;
3007 	ip6_t	*ip6h;
3008 	ip6i_t	*ip6i;
3009 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
3010 
3011 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
3012 	hdrs_len = ip_total_hdrs_len_v6(ipp);
3013 	ASSERT(hdrs_len != 0);
3014 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
3015 		/* Need to reallocate */
3016 		if (hdrs_len != 0) {
3017 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3018 			if (hdrs == NULL)
3019 				return (ENOMEM);
3020 		} else {
3021 			hdrs = NULL;
3022 		}
3023 		if (icmp->icmp_sticky_hdrs_len != 0) {
3024 			kmem_free(icmp->icmp_sticky_hdrs,
3025 			    icmp->icmp_sticky_hdrs_len);
3026 		}
3027 		icmp->icmp_sticky_hdrs = hdrs;
3028 		icmp->icmp_sticky_hdrs_len = hdrs_len;
3029 	}
3030 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
3031 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
3032 
3033 	/* Set header fields not in ipp */
3034 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3035 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
3036 		ip6h = (ip6_t *)&ip6i[1];
3037 
3038 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
3039 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
3040 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
3041 		}
3042 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
3043 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
3044 		}
3045 	} else {
3046 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
3047 	}
3048 
3049 	if (!(ipp->ipp_fields & IPPF_ADDR))
3050 		ip6h->ip6_src = icmp->icmp_v6src;
3051 
3052 	/* Try to get everything in a single mblk */
3053 	if (hdrs_len > icmp->icmp_max_hdr_len) {
3054 		icmp->icmp_max_hdr_len = hdrs_len;
3055 		rw_exit(&icmp->icmp_rwlock);
3056 		(void) mi_set_sth_wroff(icmp->icmp_connp->conn_rq,
3057 		    icmp->icmp_max_hdr_len + is->is_wroff_extra);
3058 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3059 	}
3060 	return (0);
3061 }
3062 
3063 /*
3064  * This routine retrieves the value of an ND variable in a icmpparam_t
3065  * structure.  It is called through nd_getset when a user reads the
3066  * variable.
3067  */
3068 /* ARGSUSED */
3069 static int
3070 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3071 {
3072 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3073 
3074 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
3075 	return (0);
3076 }
3077 
3078 /*
3079  * Walk through the param array specified registering each element with the
3080  * named dispatch (ND) handler.
3081  */
3082 static boolean_t
3083 icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt)
3084 {
3085 	for (; cnt-- > 0; icmppa++) {
3086 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
3087 			if (!nd_load(ndp, icmppa->icmp_param_name,
3088 			    icmp_param_get, icmp_param_set,
3089 			    (caddr_t)icmppa)) {
3090 				nd_free(ndp);
3091 				return (B_FALSE);
3092 			}
3093 		}
3094 	}
3095 	if (!nd_load(ndp, "icmp_status", icmp_status_report, NULL,
3096 	    NULL)) {
3097 		nd_free(ndp);
3098 		return (B_FALSE);
3099 	}
3100 	return (B_TRUE);
3101 }
3102 
3103 /* This routine sets an ND variable in a icmpparam_t structure. */
3104 /* ARGSUSED */
3105 static int
3106 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3107 {
3108 	long		new_value;
3109 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3110 
3111 	/*
3112 	 * Fail the request if the new value does not lie within the
3113 	 * required bounds.
3114 	 */
3115 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3116 	    new_value < icmppa->icmp_param_min ||
3117 	    new_value > icmppa->icmp_param_max) {
3118 		return (EINVAL);
3119 	}
3120 	/* Set the new value */
3121 	icmppa->icmp_param_value = new_value;
3122 	return (0);
3123 }
3124 /*ARGSUSED2*/
3125 static void
3126 icmp_input(void *arg1, mblk_t *mp, void *arg2)
3127 {
3128 	conn_t *connp = (conn_t *)arg1;
3129 	struct T_unitdata_ind	*tudi;
3130 	uchar_t			*rptr;
3131 	icmp_t			*icmp;
3132 	icmp_stack_t		*is;
3133 	sin_t			*sin;
3134 	sin6_t			*sin6;
3135 	ip6_t			*ip6h;
3136 	ip6i_t			*ip6i;
3137 	mblk_t			*mp1;
3138 	int			hdr_len;
3139 	ipha_t			*ipha;
3140 	int			udi_size;	/* Size of T_unitdata_ind */
3141 	uint_t			ipvers;
3142 	ip6_pkt_t		ipp;
3143 	uint8_t			nexthdr;
3144 	ip_pktinfo_t		*pinfo = NULL;
3145 	mblk_t			*options_mp = NULL;
3146 	uint_t			icmp_opt = 0;
3147 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
3148 	uint_t			hopstrip;
3149 
3150 	ASSERT(connp->conn_flags & IPCL_RAWIPCONN);
3151 
3152 	icmp = connp->conn_icmp;
3153 	is = icmp->icmp_is;
3154 	rptr = mp->b_rptr;
3155 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3156 	ASSERT(OK_32PTR(rptr));
3157 
3158 	/*
3159 	 * IP should have prepended the options data in an M_CTL
3160 	 * Check M_CTL "type" to make sure are not here bcos of
3161 	 * a valid ICMP message
3162 	 */
3163 	if (DB_TYPE(mp) == M_CTL) {
3164 		/*
3165 		 * FIXME: does IP still do this?
3166 		 * IP sends up the IPSEC_IN message for handling IPSEC
3167 		 * policy at the TCP level. We don't need it here.
3168 		 */
3169 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
3170 			mp1 = mp->b_cont;
3171 			freeb(mp);
3172 			mp = mp1;
3173 			rptr = mp->b_rptr;
3174 		} else if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3175 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3176 		    IN_PKTINFO) {
3177 			/*
3178 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3179 			 * has been prepended to the packet by IP. We need to
3180 			 * extract the mblk and adjust the rptr
3181 			 */
3182 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3183 			options_mp = mp;
3184 			mp = mp->b_cont;
3185 			rptr = mp->b_rptr;
3186 		} else {
3187 			/*
3188 			 * ICMP messages.
3189 			 */
3190 			icmp_icmp_error(connp->conn_rq, mp);
3191 			return;
3192 		}
3193 	}
3194 
3195 	/*
3196 	 * Discard message if it is misaligned or smaller than the IP header.
3197 	 */
3198 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3199 		freemsg(mp);
3200 		if (options_mp != NULL)
3201 			freeb(options_mp);
3202 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3203 		return;
3204 	}
3205 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3206 
3207 	/* Handle M_DATA messages containing IP packets messages */
3208 	if (ipvers == IPV4_VERSION) {
3209 		/*
3210 		 * Special case where IP attaches
3211 		 * the IRE needs to be handled so that we don't send up
3212 		 * IRE to the user land.
3213 		 */
3214 		ipha = (ipha_t *)rptr;
3215 		hdr_len = IPH_HDR_LENGTH(ipha);
3216 
3217 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3218 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3219 
3220 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3221 			    TH_SYN) && mp->b_cont != NULL) {
3222 				mp1 = mp->b_cont;
3223 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3224 					freeb(mp1);
3225 					mp->b_cont = NULL;
3226 				}
3227 			}
3228 		}
3229 		if (is->is_bsd_compat) {
3230 			ushort_t len;
3231 			len = ntohs(ipha->ipha_length);
3232 
3233 			if (mp->b_datap->db_ref > 1) {
3234 				/*
3235 				 * Allocate a new IP header so that we can
3236 				 * modify ipha_length.
3237 				 */
3238 				mblk_t	*mp1;
3239 
3240 				mp1 = allocb(hdr_len, BPRI_MED);
3241 				if (!mp1) {
3242 					freemsg(mp);
3243 					if (options_mp != NULL)
3244 						freeb(options_mp);
3245 					BUMP_MIB(&is->is_rawip_mib,
3246 					    rawipInErrors);
3247 					return;
3248 				}
3249 				bcopy(rptr, mp1->b_rptr, hdr_len);
3250 				mp->b_rptr = rptr + hdr_len;
3251 				rptr = mp1->b_rptr;
3252 				ipha = (ipha_t *)rptr;
3253 				mp1->b_cont = mp;
3254 				mp1->b_wptr = rptr + hdr_len;
3255 				mp = mp1;
3256 			}
3257 			len -= hdr_len;
3258 			ipha->ipha_length = htons(len);
3259 		}
3260 	}
3261 
3262 	/*
3263 	 * This is the inbound data path.  Packets are passed upstream as
3264 	 * T_UNITDATA_IND messages with full IP headers still attached.
3265 	 */
3266 	if (icmp->icmp_family == AF_INET) {
3267 		ASSERT(ipvers == IPV4_VERSION);
3268 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3269 		if (icmp->icmp_recvif && (pinfo != NULL) &&
3270 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3271 			udi_size += sizeof (struct T_opthdr) +
3272 			    sizeof (uint_t);
3273 		}
3274 
3275 		if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3276 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3277 			udi_size += sizeof (struct T_opthdr) +
3278 			    sizeof (struct in_pktinfo);
3279 		}
3280 
3281 		/*
3282 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3283 		 * buffer. Since gethrestime() expects a pointer aligned
3284 		 * argument, we allocate space necessary for extra
3285 		 * alignment (even though it might not be used).
3286 		 */
3287 		if (icmp->icmp_timestamp) {
3288 			udi_size += sizeof (struct T_opthdr) +
3289 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3290 		}
3291 		mp1 = allocb(udi_size, BPRI_MED);
3292 		if (mp1 == NULL) {
3293 			freemsg(mp);
3294 			if (options_mp != NULL)
3295 				freeb(options_mp);
3296 			BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3297 			return;
3298 		}
3299 		mp1->b_cont = mp;
3300 		mp = mp1;
3301 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3302 		mp->b_datap->db_type = M_PROTO;
3303 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3304 		tudi->PRIM_type = T_UNITDATA_IND;
3305 		tudi->SRC_length = sizeof (sin_t);
3306 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3307 		sin = (sin_t *)&tudi[1];
3308 		*sin = sin_null;
3309 		sin->sin_family = AF_INET;
3310 		sin->sin_addr.s_addr = ipha->ipha_src;
3311 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3312 		    sizeof (sin_t);
3313 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3314 		tudi->OPT_length = udi_size;
3315 
3316 		/*
3317 		 * Add options if IP_RECVIF is set
3318 		 */
3319 		if (udi_size != 0) {
3320 			char *dstopt;
3321 
3322 			dstopt = (char *)&sin[1];
3323 			if (icmp->icmp_recvif && (pinfo != NULL) &&
3324 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3325 
3326 				struct T_opthdr *toh;
3327 				uint_t		*dstptr;
3328 
3329 				toh = (struct T_opthdr *)dstopt;
3330 				toh->level = IPPROTO_IP;
3331 				toh->name = IP_RECVIF;
3332 				toh->len = sizeof (struct T_opthdr) +
3333 				    sizeof (uint_t);
3334 				toh->status = 0;
3335 				dstopt += sizeof (struct T_opthdr);
3336 				dstptr = (uint_t *)dstopt;
3337 				*dstptr = pinfo->ip_pkt_ifindex;
3338 				dstopt += sizeof (uint_t);
3339 				udi_size -= toh->len;
3340 			}
3341 			if (icmp->icmp_timestamp) {
3342 				struct	T_opthdr *toh;
3343 
3344 				toh = (struct T_opthdr *)dstopt;
3345 				toh->level = SOL_SOCKET;
3346 				toh->name = SCM_TIMESTAMP;
3347 				toh->len = sizeof (struct T_opthdr) +
3348 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3349 				toh->status = 0;
3350 				dstopt += sizeof (struct T_opthdr);
3351 				/* Align for gethrestime() */
3352 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3353 				    sizeof (intptr_t));
3354 				gethrestime((timestruc_t *)dstopt);
3355 				dstopt = (char *)toh + toh->len;
3356 				udi_size -= toh->len;
3357 			}
3358 			if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3359 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3360 				struct	T_opthdr *toh;
3361 				struct	in_pktinfo *pktinfop;
3362 
3363 				toh = (struct T_opthdr *)dstopt;
3364 				toh->level = IPPROTO_IP;
3365 				toh->name = IP_PKTINFO;
3366 				toh->len = sizeof (struct T_opthdr) +
3367 				    sizeof (in_pktinfo_t);
3368 				toh->status = 0;
3369 				dstopt += sizeof (struct T_opthdr);
3370 				pktinfop = (struct in_pktinfo *)dstopt;
3371 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3372 				pktinfop->ipi_spec_dst =
3373 				    pinfo->ip_pkt_match_addr;
3374 
3375 				pktinfop->ipi_addr.s_addr = ipha->ipha_dst;
3376 
3377 				dstopt += sizeof (struct in_pktinfo);
3378 				udi_size -= toh->len;
3379 			}
3380 
3381 			/* Consumed all of allocated space */
3382 			ASSERT(udi_size == 0);
3383 		}
3384 
3385 		if (options_mp != NULL)
3386 			freeb(options_mp);
3387 
3388 		BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3389 		putnext(connp->conn_rq, mp);
3390 		return;
3391 	}
3392 
3393 	/*
3394 	 * We don't need options_mp in the IPv6 path.
3395 	 */
3396 	if (options_mp != NULL) {
3397 		freeb(options_mp);
3398 		options_mp = NULL;
3399 	}
3400 
3401 	/*
3402 	 * Discard message if it is smaller than the IPv6 header
3403 	 * or if the header is malformed.
3404 	 */
3405 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3406 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3407 	    icmp->icmp_family != AF_INET6) {
3408 		freemsg(mp);
3409 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3410 		return;
3411 	}
3412 
3413 	/* Initialize */
3414 	ipp.ipp_fields = 0;
3415 	hopstrip = 0;
3416 
3417 	ip6h = (ip6_t *)rptr;
3418 	/*
3419 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3420 	 * as well as individual lenghts of ext hdrs (and ptrs to
3421 	 * them).
3422 	 */
3423 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3424 		/* Look for ifindex information */
3425 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3426 			ip6i = (ip6i_t *)ip6h;
3427 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3428 				ASSERT(ip6i->ip6i_ifindex != 0);
3429 				ipp.ipp_fields |= IPPF_IFINDEX;
3430 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3431 			}
3432 			rptr = (uchar_t *)&ip6i[1];
3433 			mp->b_rptr = rptr;
3434 			if (rptr == mp->b_wptr) {
3435 				mp1 = mp->b_cont;
3436 				freeb(mp);
3437 				mp = mp1;
3438 				rptr = mp->b_rptr;
3439 			}
3440 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3441 			ip6h = (ip6_t *)rptr;
3442 		}
3443 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3444 
3445 		/*
3446 		 * We need to lie a bit to the user because users inside
3447 		 * labeled compartments should not see their own labels.  We
3448 		 * assume that in all other respects IP has checked the label,
3449 		 * and that the label is always first among the options.  (If
3450 		 * it's not first, then this code won't see it, and the option
3451 		 * will be passed along to the user.)
3452 		 *
3453 		 * If we had multilevel ICMP sockets, then the following code
3454 		 * should be skipped for them to allow the user to see the
3455 		 * label.
3456 		 *
3457 		 * Alignment restrictions in the definition of IP options
3458 		 * (namely, the requirement that the 4-octet DOI goes on a
3459 		 * 4-octet boundary) mean that we know exactly where the option
3460 		 * should start, but we're lenient for other hosts.
3461 		 *
3462 		 * Note that there are no multilevel ICMP or raw IP sockets
3463 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3464 		 */
3465 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3466 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3467 			const uchar_t *ucp =
3468 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3469 			int remlen = ipp.ipp_hopoptslen - 2;
3470 
3471 			while (remlen > 0) {
3472 				if (*ucp == IP6OPT_PAD1) {
3473 					remlen--;
3474 					ucp++;
3475 				} else if (*ucp == IP6OPT_PADN) {
3476 					remlen -= ucp[1] + 2;
3477 					ucp += ucp[1] + 2;
3478 				} else if (*ucp == ip6opt_ls) {
3479 					hopstrip = (ucp -
3480 					    (const uchar_t *)ipp.ipp_hopopts) +
3481 					    ucp[1] + 2;
3482 					hopstrip = (hopstrip + 7) & ~7;
3483 					break;
3484 				} else {
3485 					/* label option must be first */
3486 					break;
3487 				}
3488 			}
3489 		}
3490 	} else {
3491 		hdr_len = IPV6_HDR_LEN;
3492 		ip6i = NULL;
3493 		nexthdr = ip6h->ip6_nxt;
3494 	}
3495 	/*
3496 	 * One special case where IP attaches the IRE needs to
3497 	 * be handled so that we don't send up IRE to the user land.
3498 	 */
3499 	if (nexthdr == IPPROTO_TCP) {
3500 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3501 
3502 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3503 		    mp->b_cont != NULL) {
3504 			mp1 = mp->b_cont;
3505 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3506 				freeb(mp1);
3507 				mp->b_cont = NULL;
3508 			}
3509 		}
3510 	}
3511 	/*
3512 	 * Check a filter for ICMPv6 types if needed.
3513 	 * Verify raw checksums if needed.
3514 	 */
3515 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3516 		if (icmp->icmp_filter != NULL) {
3517 			int type;
3518 
3519 			/* Assumes that IP has done the pullupmsg */
3520 			type = mp->b_rptr[hdr_len];
3521 
3522 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3523 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3524 				freemsg(mp);
3525 				return;
3526 			}
3527 		} else {
3528 			/* Checksum */
3529 			uint16_t	*up;
3530 			uint32_t	sum;
3531 			int		remlen;
3532 
3533 			up = (uint16_t *)&ip6h->ip6_src;
3534 
3535 			remlen = msgdsize(mp) - hdr_len;
3536 			sum = htons(icmp->icmp_proto + remlen)
3537 			    + up[0] + up[1] + up[2] + up[3]
3538 			    + up[4] + up[5] + up[6] + up[7]
3539 			    + up[8] + up[9] + up[10] + up[11]
3540 			    + up[12] + up[13] + up[14] + up[15];
3541 			sum = (sum & 0xffff) + (sum >> 16);
3542 			sum = IP_CSUM(mp, hdr_len, sum);
3543 			if (sum != 0) {
3544 				/* IPv6 RAW checksum failed */
3545 				ip0dbg(("icmp_rput: RAW checksum "
3546 				    "failed %x\n", sum));
3547 				freemsg(mp);
3548 				BUMP_MIB(&is->is_rawip_mib,
3549 				    rawipInCksumErrs);
3550 				return;
3551 			}
3552 		}
3553 	}
3554 	/* Skip all the IPv6 headers per API */
3555 	mp->b_rptr += hdr_len;
3556 
3557 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3558 
3559 	/*
3560 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3561 	 * maintain state information, instead of relying on icmp_t
3562 	 * structure, since there arent any locks protecting these members
3563 	 * and there is a window where there might be a race between a
3564 	 * thread setting options on the write side and a thread reading
3565 	 * these options on the read size.
3566 	 */
3567 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3568 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3569 		if (icmp->icmp_ipv6_recvhopopts &&
3570 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3571 		    ipp.ipp_hopoptslen > hopstrip) {
3572 			udi_size += sizeof (struct T_opthdr) +
3573 			    ipp.ipp_hopoptslen - hopstrip;
3574 			icmp_opt |= IPPF_HOPOPTS;
3575 		}
3576 		if ((icmp->icmp_ipv6_recvdstopts ||
3577 		    icmp->icmp_old_ipv6_recvdstopts) &&
3578 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3579 			udi_size += sizeof (struct T_opthdr) +
3580 			    ipp.ipp_dstoptslen;
3581 			icmp_opt |= IPPF_DSTOPTS;
3582 		}
3583 		if (((icmp->icmp_ipv6_recvdstopts &&
3584 		    icmp->icmp_ipv6_recvrthdr &&
3585 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3586 		    icmp->icmp_ipv6_recvrtdstopts) &&
3587 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3588 			udi_size += sizeof (struct T_opthdr) +
3589 			    ipp.ipp_rtdstoptslen;
3590 			icmp_opt |= IPPF_RTDSTOPTS;
3591 		}
3592 		if (icmp->icmp_ipv6_recvrthdr &&
3593 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3594 			udi_size += sizeof (struct T_opthdr) +
3595 			    ipp.ipp_rthdrlen;
3596 			icmp_opt |= IPPF_RTHDR;
3597 		}
3598 		if (icmp->icmp_ip_recvpktinfo &&
3599 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3600 			udi_size += sizeof (struct T_opthdr) +
3601 			    sizeof (struct in6_pktinfo);
3602 			icmp_opt |= IPPF_IFINDEX;
3603 		}
3604 	}
3605 	if (icmp->icmp_ipv6_recvhoplimit) {
3606 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3607 		icmp_ipv6_recvhoplimit = B_TRUE;
3608 	}
3609 
3610 	if (icmp->icmp_ipv6_recvtclass)
3611 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3612 
3613 	/*
3614 	 * If SO_TIMESTAMP is set allocate the appropriate sized
3615 	 * buffer. Since gethrestime() expects a pointer aligned
3616 	 * argument, we allocate space necessary for extra
3617 	 * alignment (even though it might not be used).
3618 	 */
3619 	if (icmp->icmp_timestamp) {
3620 		udi_size += sizeof (struct T_opthdr) +
3621 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3622 	}
3623 
3624 	mp1 = allocb(udi_size, BPRI_MED);
3625 	if (mp1 == NULL) {
3626 		freemsg(mp);
3627 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3628 		return;
3629 	}
3630 	mp1->b_cont = mp;
3631 	mp = mp1;
3632 	mp->b_datap->db_type = M_PROTO;
3633 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3634 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3635 	tudi->PRIM_type = T_UNITDATA_IND;
3636 	tudi->SRC_length = sizeof (sin6_t);
3637 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3638 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3639 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3640 	tudi->OPT_length = udi_size;
3641 	sin6 = (sin6_t *)&tudi[1];
3642 	sin6->sin6_port = 0;
3643 	sin6->sin6_family = AF_INET6;
3644 
3645 	sin6->sin6_addr = ip6h->ip6_src;
3646 	/* No sin6_flowinfo per API */
3647 	sin6->sin6_flowinfo = 0;
3648 	/* For link-scope source pass up scope id */
3649 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3650 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3651 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3652 	else
3653 		sin6->sin6_scope_id = 0;
3654 
3655 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3656 	    icmp->icmp_zoneid, is->is_netstack);
3657 
3658 	if (udi_size != 0) {
3659 		uchar_t *dstopt;
3660 
3661 		dstopt = (uchar_t *)&sin6[1];
3662 		if (icmp_opt & IPPF_IFINDEX) {
3663 			struct T_opthdr *toh;
3664 			struct in6_pktinfo *pkti;
3665 
3666 			toh = (struct T_opthdr *)dstopt;
3667 			toh->level = IPPROTO_IPV6;
3668 			toh->name = IPV6_PKTINFO;
3669 			toh->len = sizeof (struct T_opthdr) +
3670 			    sizeof (*pkti);
3671 			toh->status = 0;
3672 			dstopt += sizeof (struct T_opthdr);
3673 			pkti = (struct in6_pktinfo *)dstopt;
3674 			pkti->ipi6_addr = ip6h->ip6_dst;
3675 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3676 			dstopt += sizeof (*pkti);
3677 			udi_size -= toh->len;
3678 		}
3679 		if (icmp_ipv6_recvhoplimit) {
3680 			struct T_opthdr *toh;
3681 
3682 			toh = (struct T_opthdr *)dstopt;
3683 			toh->level = IPPROTO_IPV6;
3684 			toh->name = IPV6_HOPLIMIT;
3685 			toh->len = sizeof (struct T_opthdr) +
3686 			    sizeof (uint_t);
3687 			toh->status = 0;
3688 			dstopt += sizeof (struct T_opthdr);
3689 			*(uint_t *)dstopt = ip6h->ip6_hops;
3690 			dstopt += sizeof (uint_t);
3691 			udi_size -= toh->len;
3692 		}
3693 		if (icmp->icmp_ipv6_recvtclass) {
3694 			struct T_opthdr *toh;
3695 
3696 			toh = (struct T_opthdr *)dstopt;
3697 			toh->level = IPPROTO_IPV6;
3698 			toh->name = IPV6_TCLASS;
3699 			toh->len = sizeof (struct T_opthdr) +
3700 			    sizeof (uint_t);
3701 			toh->status = 0;
3702 			dstopt += sizeof (struct T_opthdr);
3703 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3704 			dstopt += sizeof (uint_t);
3705 			udi_size -= toh->len;
3706 		}
3707 		if (icmp->icmp_timestamp) {
3708 			struct	T_opthdr *toh;
3709 
3710 			toh = (struct T_opthdr *)dstopt;
3711 			toh->level = SOL_SOCKET;
3712 			toh->name = SCM_TIMESTAMP;
3713 			toh->len = sizeof (struct T_opthdr) +
3714 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3715 			toh->status = 0;
3716 			dstopt += sizeof (struct T_opthdr);
3717 			/* Align for gethrestime() */
3718 			dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
3719 			    sizeof (intptr_t));
3720 			gethrestime((timestruc_t *)dstopt);
3721 			dstopt = (uchar_t *)toh + toh->len;
3722 			udi_size -= toh->len;
3723 		}
3724 		if (icmp_opt & IPPF_HOPOPTS) {
3725 			struct T_opthdr *toh;
3726 
3727 			toh = (struct T_opthdr *)dstopt;
3728 			toh->level = IPPROTO_IPV6;
3729 			toh->name = IPV6_HOPOPTS;
3730 			toh->len = sizeof (struct T_opthdr) +
3731 			    ipp.ipp_hopoptslen - hopstrip;
3732 			toh->status = 0;
3733 			dstopt += sizeof (struct T_opthdr);
3734 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
3735 			    ipp.ipp_hopoptslen - hopstrip);
3736 			if (hopstrip > 0) {
3737 				/* copy next header value and fake length */
3738 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
3739 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
3740 				    hopstrip / 8;
3741 			}
3742 			dstopt += ipp.ipp_hopoptslen - hopstrip;
3743 			udi_size -= toh->len;
3744 		}
3745 		if (icmp_opt & IPPF_RTDSTOPTS) {
3746 			struct T_opthdr *toh;
3747 
3748 			toh = (struct T_opthdr *)dstopt;
3749 			toh->level = IPPROTO_IPV6;
3750 			toh->name = IPV6_DSTOPTS;
3751 			toh->len = sizeof (struct T_opthdr) +
3752 			    ipp.ipp_rtdstoptslen;
3753 			toh->status = 0;
3754 			dstopt += sizeof (struct T_opthdr);
3755 			bcopy(ipp.ipp_rtdstopts, dstopt,
3756 			    ipp.ipp_rtdstoptslen);
3757 			dstopt += ipp.ipp_rtdstoptslen;
3758 			udi_size -= toh->len;
3759 		}
3760 		if (icmp_opt & IPPF_RTHDR) {
3761 			struct T_opthdr *toh;
3762 
3763 			toh = (struct T_opthdr *)dstopt;
3764 			toh->level = IPPROTO_IPV6;
3765 			toh->name = IPV6_RTHDR;
3766 			toh->len = sizeof (struct T_opthdr) +
3767 			    ipp.ipp_rthdrlen;
3768 			toh->status = 0;
3769 			dstopt += sizeof (struct T_opthdr);
3770 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
3771 			dstopt += ipp.ipp_rthdrlen;
3772 			udi_size -= toh->len;
3773 		}
3774 		if (icmp_opt & IPPF_DSTOPTS) {
3775 			struct T_opthdr *toh;
3776 
3777 			toh = (struct T_opthdr *)dstopt;
3778 			toh->level = IPPROTO_IPV6;
3779 			toh->name = IPV6_DSTOPTS;
3780 			toh->len = sizeof (struct T_opthdr) +
3781 			    ipp.ipp_dstoptslen;
3782 			toh->status = 0;
3783 			dstopt += sizeof (struct T_opthdr);
3784 			bcopy(ipp.ipp_dstopts, dstopt,
3785 			    ipp.ipp_dstoptslen);
3786 			dstopt += ipp.ipp_dstoptslen;
3787 			udi_size -= toh->len;
3788 		}
3789 		/* Consumed all of allocated space */
3790 		ASSERT(udi_size == 0);
3791 	}
3792 	BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3793 	putnext(connp->conn_rq, mp);
3794 }
3795 
3796 /*
3797  * Handle the results of a T_BIND_REQ whether deferred by IP or handled
3798  * immediately.
3799  */
3800 static void
3801 icmp_bind_result(conn_t *connp, mblk_t *mp)
3802 {
3803 	struct T_error_ack	*tea;
3804 
3805 	switch (mp->b_datap->db_type) {
3806 	case M_PROTO:
3807 	case M_PCPROTO:
3808 		/* M_PROTO messages contain some type of TPI message. */
3809 		if ((mp->b_wptr - mp->b_rptr) < sizeof (t_scalar_t)) {
3810 			freemsg(mp);
3811 			return;
3812 		}
3813 		tea = (struct T_error_ack *)mp->b_rptr;
3814 
3815 		switch (tea->PRIM_type) {
3816 		case T_ERROR_ACK:
3817 			switch (tea->ERROR_prim) {
3818 			case O_T_BIND_REQ:
3819 			case T_BIND_REQ:
3820 				icmp_bind_error(connp, mp);
3821 				return;
3822 			default:
3823 				break;
3824 			}
3825 			ASSERT(0);
3826 			freemsg(mp);
3827 			return;
3828 
3829 		case T_BIND_ACK:
3830 			icmp_bind_ack(connp, mp);
3831 			return;
3832 
3833 		default:
3834 			break;
3835 		}
3836 		freemsg(mp);
3837 		return;
3838 	default:
3839 		/* FIXME: other cases? */
3840 		ASSERT(0);
3841 		freemsg(mp);
3842 		return;
3843 	}
3844 }
3845 
3846 /*
3847  * Process a T_BIND_ACK
3848  */
3849 static void
3850 icmp_bind_ack(conn_t *connp, mblk_t *mp)
3851 {
3852 	icmp_t	*icmp = connp->conn_icmp;
3853 	mblk_t	*mp1;
3854 	ire_t	*ire;
3855 	struct T_bind_ack *tba;
3856 	uchar_t *addrp;
3857 	ipa_conn_t	*ac;
3858 	ipa6_conn_t	*ac6;
3859 
3860 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3861 	/*
3862 	 * We know if headers are included or not so we can
3863 	 * safely do this.
3864 	 */
3865 	if (icmp->icmp_state == TS_UNBND) {
3866 		/*
3867 		 * TPI has not yet bound - bind sent by
3868 		 * icmp_bind_proto.
3869 		 */
3870 		freemsg(mp);
3871 		rw_exit(&icmp->icmp_rwlock);
3872 		return;
3873 	}
3874 	ASSERT(icmp->icmp_pending_op != -1);
3875 
3876 	/*
3877 	 * If a broadcast/multicast address was bound set
3878 	 * the source address to 0.
3879 	 * This ensures no datagrams with broadcast address
3880 	 * as source address are emitted (which would violate
3881 	 * RFC1122 - Hosts requirements)
3882 	 *
3883 	 * Note that when connecting the returned IRE is
3884 	 * for the destination address and we only perform
3885 	 * the broadcast check for the source address (it
3886 	 * is OK to connect to a broadcast/multicast address.)
3887 	 */
3888 	mp1 = mp->b_cont;
3889 	if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) {
3890 		ire = (ire_t *)mp1->b_rptr;
3891 
3892 		/*
3893 		 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast
3894 		 * local address.
3895 		 */
3896 		if (ire->ire_type == IRE_BROADCAST &&
3897 		    icmp->icmp_state != TS_DATA_XFER) {
3898 			ASSERT(icmp->icmp_pending_op == T_BIND_REQ ||
3899 			    icmp->icmp_pending_op == O_T_BIND_REQ);
3900 			/* This was just a local bind to a MC/broadcast addr */
3901 			V6_SET_ZERO(icmp->icmp_v6src);
3902 			if (icmp->icmp_family == AF_INET6)
3903 				(void) icmp_build_hdrs(icmp);
3904 		} else if (V6_OR_V4_INADDR_ANY(icmp->icmp_v6src)) {
3905 			/*
3906 			 * Local address not yet set - pick it from the
3907 			 * T_bind_ack
3908 			 */
3909 			tba = (struct T_bind_ack *)mp->b_rptr;
3910 			addrp = &mp->b_rptr[tba->ADDR_offset];
3911 			switch (icmp->icmp_family) {
3912 			case AF_INET:
3913 				if (tba->ADDR_length == sizeof (ipa_conn_t)) {
3914 					ac = (ipa_conn_t *)addrp;
3915 				} else {
3916 					ASSERT(tba->ADDR_length ==
3917 					    sizeof (ipa_conn_x_t));
3918 					ac = &((ipa_conn_x_t *)addrp)->acx_conn;
3919 				}
3920 				IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr,
3921 				    &icmp->icmp_v6src);
3922 				break;
3923 			case AF_INET6:
3924 				if (tba->ADDR_length == sizeof (ipa6_conn_t)) {
3925 					ac6 = (ipa6_conn_t *)addrp;
3926 				} else {
3927 					ASSERT(tba->ADDR_length ==
3928 					    sizeof (ipa6_conn_x_t));
3929 					ac6 = &((ipa6_conn_x_t *)
3930 					    addrp)->ac6x_conn;
3931 				}
3932 				icmp->icmp_v6src = ac6->ac6_laddr;
3933 				(void) icmp_build_hdrs(icmp);
3934 			}
3935 		}
3936 		mp1 = mp1->b_cont;
3937 	}
3938 	icmp->icmp_pending_op = -1;
3939 	rw_exit(&icmp->icmp_rwlock);
3940 	/*
3941 	 * Look for one or more appended ACK message added by
3942 	 * icmp_connect or icmp_disconnect.
3943 	 * If none found just send up the T_BIND_ACK.
3944 	 * icmp_connect has appended a T_OK_ACK and a
3945 	 * T_CONN_CON.
3946 	 * icmp_disconnect has appended a T_OK_ACK.
3947 	 */
3948 	if (mp1 != NULL) {
3949 		if (mp->b_cont == mp1)
3950 			mp->b_cont = NULL;
3951 		else {
3952 			ASSERT(mp->b_cont->b_cont == mp1);
3953 			mp->b_cont->b_cont = NULL;
3954 		}
3955 		freemsg(mp);
3956 		mp = mp1;
3957 		while (mp != NULL) {
3958 			mp1 = mp->b_cont;
3959 			mp->b_cont = NULL;
3960 			putnext(connp->conn_rq, mp);
3961 			mp = mp1;
3962 		}
3963 		return;
3964 	}
3965 	freemsg(mp->b_cont);
3966 	mp->b_cont = NULL;
3967 	putnext(connp->conn_rq, mp);
3968 }
3969 
3970 static void
3971 icmp_bind_error(conn_t *connp, mblk_t *mp)
3972 {
3973 	icmp_t	*icmp = connp->conn_icmp;
3974 	struct T_error_ack *tea;
3975 
3976 	tea = (struct T_error_ack *)mp->b_rptr;
3977 	/*
3978 	 * If our O_T_BIND_REQ/T_BIND_REQ fails,
3979 	 * clear out the source address before
3980 	 * passing the message upstream.
3981 	 * If this was caused by a T_CONN_REQ
3982 	 * revert back to bound state.
3983 	 */
3984 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3985 	if (icmp->icmp_state == TS_UNBND) {
3986 		/*
3987 		 * TPI has not yet bound - bind sent by icmp_bind_proto.
3988 		 */
3989 		freemsg(mp);
3990 		rw_exit(&icmp->icmp_rwlock);
3991 		return;
3992 	}
3993 	ASSERT(icmp->icmp_pending_op != -1);
3994 	tea->ERROR_prim = icmp->icmp_pending_op;
3995 	icmp->icmp_pending_op = -1;
3996 
3997 	switch (tea->ERROR_prim) {
3998 	case T_CONN_REQ:
3999 		ASSERT(icmp->icmp_state == TS_DATA_XFER);
4000 		/* Connect failed */
4001 		/* Revert back to the bound source */
4002 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
4003 		icmp->icmp_state = TS_IDLE;
4004 		if (icmp->icmp_family == AF_INET6)
4005 			(void) icmp_build_hdrs(icmp);
4006 		break;
4007 
4008 	case T_DISCON_REQ:
4009 	case T_BIND_REQ:
4010 	case O_T_BIND_REQ:
4011 		V6_SET_ZERO(icmp->icmp_v6src);
4012 		V6_SET_ZERO(icmp->icmp_bound_v6src);
4013 		icmp->icmp_state = TS_UNBND;
4014 		if (icmp->icmp_family == AF_INET6)
4015 			(void) icmp_build_hdrs(icmp);
4016 		break;
4017 	default:
4018 		break;
4019 	}
4020 	rw_exit(&icmp->icmp_rwlock);
4021 	putnext(connp->conn_rq, mp);
4022 }
4023 
4024 /*
4025  * return SNMP stuff in buffer in mpdata
4026  */
4027 mblk_t *
4028 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
4029 {
4030 	mblk_t			*mpdata;
4031 	struct opthdr		*optp;
4032 	conn_t			*connp = Q_TO_CONN(q);
4033 	icmp_stack_t		*is = connp->conn_netstack->netstack_icmp;
4034 	mblk_t			*mp2ctl;
4035 
4036 	/*
4037 	 * make a copy of the original message
4038 	 */
4039 	mp2ctl = copymsg(mpctl);
4040 
4041 	if (mpctl == NULL ||
4042 	    (mpdata = mpctl->b_cont) == NULL) {
4043 		freemsg(mpctl);
4044 		freemsg(mp2ctl);
4045 		return (0);
4046 	}
4047 
4048 	/* fixed length structure for IPv4 and IPv6 counters */
4049 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4050 	optp->level = EXPER_RAWIP;
4051 	optp->name = 0;
4052 	(void) snmp_append_data(mpdata, (char *)&is->is_rawip_mib,
4053 	    sizeof (is->is_rawip_mib));
4054 	optp->len = msgdsize(mpdata);
4055 	qreply(q, mpctl);
4056 
4057 	return (mp2ctl);
4058 }
4059 
4060 /*
4061  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
4062  * TODO:  If this ever actually tries to set anything, it needs to be
4063  * to do the appropriate locking.
4064  */
4065 /* ARGSUSED */
4066 int
4067 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4068     uchar_t *ptr, int len)
4069 {
4070 	switch (level) {
4071 	case EXPER_RAWIP:
4072 		return (0);
4073 	default:
4074 		return (1);
4075 	}
4076 }
4077 
4078 /* Report for ndd "icmp_status" */
4079 /* ARGSUSED */
4080 static int
4081 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4082 {
4083 	conn_t  *connp;
4084 	ip_stack_t *ipst;
4085 	char	laddrbuf[INET6_ADDRSTRLEN];
4086 	char	faddrbuf[INET6_ADDRSTRLEN];
4087 	int	i;
4088 
4089 	(void) mi_mpprintf(mp,
4090 	    "RAWIP    " MI_COL_HDRPAD_STR
4091 	/*   01234567[89ABCDEF] */
4092 	    "  src addr        dest addr       state");
4093 	/*   xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */
4094 
4095 	connp = Q_TO_CONN(q);
4096 	ipst = connp->conn_netstack->netstack_ip;
4097 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4098 		connf_t *connfp;
4099 		char	*state;
4100 
4101 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4102 		connp = NULL;
4103 
4104 		while ((connp = ipcl_get_next_conn(connfp, connp,
4105 		    IPCL_RAWIPCONN)) != NULL) {
4106 			icmp_t  *icmp;
4107 
4108 			mutex_enter(&(connp)->conn_lock);
4109 			icmp = connp->conn_icmp;
4110 
4111 			if (icmp->icmp_state == TS_UNBND)
4112 				state = "UNBOUND";
4113 			else if (icmp->icmp_state == TS_IDLE)
4114 				state = "IDLE";
4115 			else if (icmp->icmp_state == TS_DATA_XFER)
4116 				state = "CONNECTED";
4117 			else
4118 				state = "UnkState";
4119 
4120 			(void) mi_mpprintf(mp, MI_COL_PTRFMT_STR "%s %s %s",
4121 			    (void *)icmp,
4122 			    inet_ntop(AF_INET6, &icmp->icmp_v6dst, faddrbuf,
4123 			    sizeof (faddrbuf)),
4124 			    inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf,
4125 			    sizeof (laddrbuf)),
4126 			    state);
4127 			mutex_exit(&(connp)->conn_lock);
4128 		}
4129 	}
4130 	return (0);
4131 }
4132 
4133 /*
4134  * This routine creates a T_UDERROR_IND message and passes it upstream.
4135  * The address and options are copied from the T_UNITDATA_REQ message
4136  * passed in mp.  This message is freed.
4137  */
4138 static void
4139 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
4140 {
4141 	mblk_t	*mp1;
4142 	uchar_t	*rptr = mp->b_rptr;
4143 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
4144 
4145 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
4146 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
4147 	    tudr->OPT_length, err);
4148 	if (mp1)
4149 		qreply(q, mp1);
4150 	freemsg(mp);
4151 }
4152 
4153 /*
4154  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
4155  * After some error checking, the message is passed downstream to ip.
4156  */
4157 static void
4158 icmp_unbind(queue_t *q, mblk_t *mp)
4159 {
4160 	icmp_t	*icmp = Q_TO_ICMP(q);
4161 
4162 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4163 	/* If a bind has not been done, we can't unbind. */
4164 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
4165 		rw_exit(&icmp->icmp_rwlock);
4166 		icmp_err_ack(q, mp, TOUTSTATE, 0);
4167 		return;
4168 	}
4169 	icmp->icmp_pending_op = T_UNBIND_REQ;
4170 	rw_exit(&icmp->icmp_rwlock);
4171 
4172 	/*
4173 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
4174 	 * and therefore ip_unbind must never return NULL.
4175 	 */
4176 	mp = ip_unbind(q, mp);
4177 	ASSERT(mp != NULL);
4178 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4179 
4180 	/*
4181 	 * Once we're unbound from IP, the pending operation may be cleared
4182 	 * here.
4183 	 */
4184 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4185 	V6_SET_ZERO(icmp->icmp_v6src);
4186 	V6_SET_ZERO(icmp->icmp_bound_v6src);
4187 	icmp->icmp_pending_op = -1;
4188 	icmp->icmp_state = TS_UNBND;
4189 	if (icmp->icmp_family == AF_INET6)
4190 		(void) icmp_build_hdrs(icmp);
4191 	rw_exit(&icmp->icmp_rwlock);
4192 
4193 	qreply(q, mp);
4194 }
4195 
4196 /*
4197  * Process IPv4 packets that already include an IP header.
4198  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
4199  * IPPROTO_IGMP).
4200  */
4201 static void
4202 icmp_wput_hdrincl(queue_t *q, mblk_t *mp, icmp_t *icmp, ip4_pkt_t *pktinfop)
4203 {
4204 	icmp_stack_t *is = icmp->icmp_is;
4205 	ipha_t	*ipha;
4206 	int	ip_hdr_length;
4207 	int	tp_hdr_len;
4208 	mblk_t	*mp1;
4209 	uint_t	pkt_len;
4210 	ip_opt_info_t optinfo;
4211 	conn_t	*connp = icmp->icmp_connp;
4212 
4213 	optinfo.ip_opt_flags = 0;
4214 	optinfo.ip_opt_ill_index = 0;
4215 	ipha = (ipha_t *)mp->b_rptr;
4216 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4217 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
4218 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4219 			ASSERT(icmp != NULL);
4220 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4221 			freemsg(mp);
4222 			return;
4223 		}
4224 		ipha = (ipha_t *)mp->b_rptr;
4225 	}
4226 	ipha->ipha_version_and_hdr_length =
4227 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
4228 
4229 	/*
4230 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4231 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4232 	 * tell IP that the application has sent a complete IP header and not
4233 	 * to compute the transport checksum nor change the DF flag.
4234 	 */
4235 	ipha->ipha_ident = IP_HDR_INCLUDED;
4236 	ipha->ipha_hdr_checksum = 0;
4237 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
4238 	/* Insert options if any */
4239 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4240 		/*
4241 		 * Put the IP header plus any transport header that is
4242 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
4243 		 * that at least the checksum field is in the first mblk.)
4244 		 */
4245 		switch (ipha->ipha_protocol) {
4246 		case IPPROTO_UDP:
4247 			tp_hdr_len = 8;
4248 			break;
4249 		case IPPROTO_TCP:
4250 			tp_hdr_len = 20;
4251 			break;
4252 		default:
4253 			tp_hdr_len = 0;
4254 			break;
4255 		}
4256 		/*
4257 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
4258 		 * tp_hdr_len bytes will be in a single mblk.
4259 		 */
4260 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
4261 		    tp_hdr_len)) {
4262 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
4263 			    tp_hdr_len)) {
4264 				BUMP_MIB(&is->is_rawip_mib,
4265 				    rawipOutErrors);
4266 				freemsg(mp);
4267 				return;
4268 			}
4269 			ipha = (ipha_t *)mp->b_rptr;
4270 		}
4271 
4272 		/*
4273 		 * if the length is larger then the max allowed IP packet,
4274 		 * then send an error and abort the processing.
4275 		 */
4276 		pkt_len = ntohs(ipha->ipha_length)
4277 		    + icmp->icmp_ip_snd_options_len;
4278 		if (pkt_len > IP_MAXPACKET) {
4279 			icmp_ud_err(q, mp, EMSGSIZE);
4280 			return;
4281 		}
4282 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra +
4283 		    tp_hdr_len, BPRI_LO))) {
4284 			icmp_ud_err(q, mp, ENOMEM);
4285 			return;
4286 		}
4287 		mp1->b_rptr += is->is_wroff_extra;
4288 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
4289 
4290 		ipha->ipha_length = htons((uint16_t)pkt_len);
4291 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
4292 
4293 		/* Copy transport header if any */
4294 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
4295 		mp1->b_wptr += tp_hdr_len;
4296 
4297 		/* Add options */
4298 		ipha = (ipha_t *)mp1->b_rptr;
4299 		bcopy(icmp->icmp_ip_snd_options, &ipha[1],
4300 		    icmp->icmp_ip_snd_options_len);
4301 
4302 		/* Drop IP header and transport header from original */
4303 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
4304 
4305 		mp1->b_cont = mp;
4306 		mp = mp1;
4307 		/*
4308 		 * Massage source route putting first source
4309 		 * route in ipha_dst.
4310 		 */
4311 		(void) ip_massage_options(ipha, is->is_netstack);
4312 	}
4313 
4314 	if (pktinfop != NULL) {
4315 		/*
4316 		 * Over write the source address provided in the header
4317 		 */
4318 		if (pktinfop->ip4_addr != INADDR_ANY) {
4319 			ipha->ipha_src = pktinfop->ip4_addr;
4320 			optinfo.ip_opt_flags = IP_VERIFY_SRC;
4321 		}
4322 
4323 		if (pktinfop->ip4_ill_index != 0) {
4324 			optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4325 		}
4326 	}
4327 
4328 	mblk_setcred(mp, connp->conn_cred);
4329 	ip_output_options(connp, mp, q, IP_WPUT,
4330 	    &optinfo);
4331 }
4332 
4333 static boolean_t
4334 icmp_update_label(queue_t *q, icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
4335 {
4336 	int err;
4337 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4338 	icmp_stack_t		*is = icmp->icmp_is;
4339 	conn_t	*connp = icmp->icmp_connp;
4340 
4341 	err = tsol_compute_label(DB_CREDDEF(mp, connp->conn_cred), dst,
4342 	    opt_storage, icmp->icmp_mac_exempt,
4343 	    is->is_netstack->netstack_ip);
4344 	if (err == 0) {
4345 		err = tsol_update_options(&icmp->icmp_ip_snd_options,
4346 		    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
4347 		    opt_storage);
4348 	}
4349 	if (err != 0) {
4350 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4351 		DTRACE_PROBE4(
4352 		    tx__ip__log__drop__updatelabel__icmp,
4353 		    char *, "queue(1) failed to update options(2) on mp(3)",
4354 		    queue_t *, q, char *, opt_storage, mblk_t *, mp);
4355 		icmp_ud_err(q, mp, err);
4356 		return (B_FALSE);
4357 	}
4358 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
4359 	return (B_TRUE);
4360 }
4361 
4362 /*
4363  * This routine handles all messages passed downstream.  It either
4364  * consumes the message or passes it downstream; it never queues a
4365  * a message.
4366  */
4367 static void
4368 icmp_wput(queue_t *q, mblk_t *mp)
4369 {
4370 	uchar_t	*rptr = mp->b_rptr;
4371 	ipha_t	*ipha;
4372 	mblk_t	*mp1;
4373 	int	ip_hdr_length;
4374 #define	tudr ((struct T_unitdata_req *)rptr)
4375 	size_t	ip_len;
4376 	conn_t	*connp = Q_TO_CONN(q);
4377 	icmp_t	*icmp = connp->conn_icmp;
4378 	icmp_stack_t *is = icmp->icmp_is;
4379 	sin6_t	*sin6;
4380 	sin_t	*sin;
4381 	ipaddr_t	v4dst;
4382 	ip4_pkt_t	pktinfo;
4383 	ip4_pkt_t	*pktinfop = &pktinfo;
4384 	ip_opt_info_t	optinfo;
4385 
4386 	switch (mp->b_datap->db_type) {
4387 	case M_DATA:
4388 		if (icmp->icmp_hdrincl) {
4389 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
4390 			ipha = (ipha_t *)mp->b_rptr;
4391 			if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH) {
4392 				if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4393 					BUMP_MIB(&is->is_rawip_mib,
4394 					    rawipOutErrors);
4395 					freemsg(mp);
4396 					return;
4397 				}
4398 				ipha = (ipha_t *)mp->b_rptr;
4399 			}
4400 			/*
4401 			 * If this connection was used for v6 (inconceivable!)
4402 			 * or if we have a new destination, then it's time to
4403 			 * figure a new label.
4404 			 */
4405 			if (is_system_labeled() &&
4406 			    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4407 			    V4_PART_OF_V6(icmp->icmp_v6lastdst) !=
4408 			    ipha->ipha_dst) &&
4409 			    !icmp_update_label(q, icmp, mp, ipha->ipha_dst)) {
4410 				return;
4411 			}
4412 			icmp_wput_hdrincl(q, mp, icmp, NULL);
4413 			return;
4414 		}
4415 		freemsg(mp);
4416 		return;
4417 	case M_PROTO:
4418 	case M_PCPROTO:
4419 		ip_len = mp->b_wptr - rptr;
4420 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4421 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4422 			if (((union T_primitives *)rptr)->type
4423 			    == T_UNITDATA_REQ)
4424 				break;
4425 		}
4426 		/* FALLTHRU */
4427 	default:
4428 		icmp_wput_other(q, mp);
4429 		return;
4430 	}
4431 
4432 	/* Handle T_UNITDATA_REQ messages here. */
4433 
4434 
4435 
4436 	if (icmp->icmp_state == TS_UNBND) {
4437 		/* If a port has not been bound to the stream, fail. */
4438 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4439 		icmp_ud_err(q, mp, EPROTO);
4440 		return;
4441 	}
4442 	mp1 = mp->b_cont;
4443 	if (mp1 == NULL) {
4444 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4445 		icmp_ud_err(q, mp, EPROTO);
4446 		return;
4447 	}
4448 
4449 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4450 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4451 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4452 		return;
4453 	}
4454 
4455 	switch (icmp->icmp_family) {
4456 	case AF_INET6:
4457 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4458 		if (!OK_32PTR((char *)sin6) ||
4459 		    tudr->DEST_length != sizeof (sin6_t) ||
4460 		    sin6->sin6_family != AF_INET6) {
4461 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4462 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4463 			return;
4464 		}
4465 
4466 		/* No support for mapped addresses on raw sockets */
4467 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4468 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4469 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4470 			return;
4471 		}
4472 
4473 		/*
4474 		 * Destination is a native IPv6 address.
4475 		 * Send out an IPv6 format packet.
4476 		 */
4477 		icmp_wput_ipv6(q, mp, sin6, tudr->OPT_length);
4478 		return;
4479 
4480 	case AF_INET:
4481 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4482 		if (!OK_32PTR((char *)sin) ||
4483 		    tudr->DEST_length != sizeof (sin_t) ||
4484 		    sin->sin_family != AF_INET) {
4485 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4486 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4487 			return;
4488 		}
4489 		/* Extract and ipaddr */
4490 		v4dst = sin->sin_addr.s_addr;
4491 		break;
4492 
4493 	default:
4494 		ASSERT(0);
4495 	}
4496 
4497 	pktinfop->ip4_ill_index = 0;
4498 	pktinfop->ip4_addr = INADDR_ANY;
4499 	optinfo.ip_opt_flags = 0;
4500 	optinfo.ip_opt_ill_index = 0;
4501 
4502 
4503 	/*
4504 	 * If options passed in, feed it for verification and handling
4505 	 */
4506 	if (tudr->OPT_length != 0) {
4507 		int error;
4508 
4509 		error = 0;
4510 		if (icmp_unitdata_opt_process(q, mp, &error,
4511 		    (void *)pktinfop) < 0) {
4512 			/* failure */
4513 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4514 			icmp_ud_err(q, mp, error);
4515 			return;
4516 		}
4517 		ASSERT(error == 0);
4518 		/*
4519 		 * Note: Success in processing options.
4520 		 * mp option buffer represented by
4521 		 * OPT_length/offset now potentially modified
4522 		 * and contain option setting results
4523 		 */
4524 
4525 	}
4526 
4527 	if (v4dst == INADDR_ANY)
4528 		v4dst = htonl(INADDR_LOOPBACK);
4529 
4530 	/* Check if our saved options are valid; update if not */
4531 	if (is_system_labeled() &&
4532 	    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4533 	    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst) &&
4534 	    !icmp_update_label(q, icmp, mp, v4dst)) {
4535 		return;
4536 	}
4537 
4538 	/* Protocol 255 contains full IP headers */
4539 	if (icmp->icmp_hdrincl) {
4540 		freeb(mp);
4541 		icmp_wput_hdrincl(q, mp1, icmp, pktinfop);
4542 		return;
4543 	}
4544 
4545 
4546 	/* Add an IP header */
4547 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4548 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
4549 	if ((uchar_t *)ipha < mp1->b_datap->db_base ||
4550 	    mp1->b_datap->db_ref != 1 ||
4551 	    !OK_32PTR(ipha)) {
4552 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra,
4553 		    BPRI_LO))) {
4554 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4555 			icmp_ud_err(q, mp, ENOMEM);
4556 			return;
4557 		}
4558 		mp1->b_cont = mp->b_cont;
4559 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4560 		mp1->b_wptr = (uchar_t *)ipha;
4561 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4562 	}
4563 #ifdef	_BIG_ENDIAN
4564 	/* Set version, header length, and tos */
4565 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4566 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4567 	    icmp->icmp_type_of_service);
4568 	/* Set ttl and protocol */
4569 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4570 #else
4571 	/* Set version, header length, and tos */
4572 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4573 	    ((icmp->icmp_type_of_service << 8) |
4574 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4575 	/* Set ttl and protocol */
4576 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4577 #endif
4578 	if (pktinfop->ip4_addr != INADDR_ANY) {
4579 		ipha->ipha_src = pktinfop->ip4_addr;
4580 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4581 	} else {
4582 
4583 		/*
4584 		 * Copy our address into the packet.  If this is zero,
4585 		 * ip will fill in the real source address.
4586 		 */
4587 		IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4588 	}
4589 
4590 	ipha->ipha_fragment_offset_and_flags = 0;
4591 
4592 	if (pktinfop->ip4_ill_index != 0) {
4593 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4594 	}
4595 
4596 
4597 	/*
4598 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4599 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4600 	 * tell IP that the application has sent a complete IP header and not
4601 	 * to compute the transport checksum nor change the DF flag.
4602 	 */
4603 	ipha->ipha_ident = IP_HDR_INCLUDED;
4604 
4605 	/* Finish common formatting of the packet. */
4606 	mp1->b_rptr = (uchar_t *)ipha;
4607 
4608 	ip_len = mp1->b_wptr - (uchar_t *)ipha;
4609 	if (mp1->b_cont != NULL)
4610 		ip_len += msgdsize(mp1->b_cont);
4611 
4612 	/*
4613 	 * Set the length into the IP header.
4614 	 * If the length is greater than the maximum allowed by IP,
4615 	 * then free the message and return. Do not try and send it
4616 	 * as this can cause problems in layers below.
4617 	 */
4618 	if (ip_len > IP_MAXPACKET) {
4619 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4620 		icmp_ud_err(q, mp, EMSGSIZE);
4621 		return;
4622 	}
4623 	ipha->ipha_length = htons((uint16_t)ip_len);
4624 	/*
4625 	 * Copy in the destination address from the T_UNITDATA
4626 	 * request
4627 	 */
4628 	ipha->ipha_dst = v4dst;
4629 
4630 	/*
4631 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4632 	 */
4633 	if (CLASSD(v4dst))
4634 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4635 
4636 	/* Copy in options if any */
4637 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4638 		bcopy(icmp->icmp_ip_snd_options,
4639 		    &ipha[1], icmp->icmp_ip_snd_options_len);
4640 		/*
4641 		 * Massage source route putting first source route in ipha_dst.
4642 		 * Ignore the destination in the T_unitdata_req.
4643 		 */
4644 		(void) ip_massage_options(ipha, is->is_netstack);
4645 	}
4646 
4647 	freeb(mp);
4648 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
4649 	mblk_setcred(mp1, connp->conn_cred);
4650 	ip_output_options(Q_TO_CONN(q), mp1, q, IP_WPUT, &optinfo);
4651 #undef	ipha
4652 #undef tudr
4653 }
4654 
4655 static boolean_t
4656 icmp_update_label_v6(queue_t *wq, icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4657 {
4658 	int err;
4659 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4660 	icmp_stack_t		*is = icmp->icmp_is;
4661 	conn_t	*connp = icmp->icmp_connp;
4662 
4663 	err = tsol_compute_label_v6(DB_CREDDEF(mp, connp->conn_cred), dst,
4664 	    opt_storage, icmp->icmp_mac_exempt,
4665 	    is->is_netstack->netstack_ip);
4666 	if (err == 0) {
4667 		err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4668 		    &icmp->icmp_label_len_v6, opt_storage);
4669 	}
4670 	if (err != 0) {
4671 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4672 		DTRACE_PROBE4(
4673 		    tx__ip__log__drop__updatelabel__icmp6,
4674 		    char *, "queue(1) failed to update options(2) on mp(3)",
4675 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4676 		icmp_ud_err(wq, mp, err);
4677 		return (B_FALSE);
4678 	}
4679 
4680 	icmp->icmp_v6lastdst = *dst;
4681 	return (B_TRUE);
4682 }
4683 
4684 /*
4685  * icmp_wput_ipv6():
4686  * Assumes that icmp_wput did some sanity checking on the destination
4687  * address, but that the label may not yet be correct.
4688  */
4689 void
4690 icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6, t_scalar_t tudr_optlen)
4691 {
4692 	ip6_t			*ip6h;
4693 	ip6i_t			*ip6i;	/* mp1->b_rptr even if no ip6i_t */
4694 	mblk_t			*mp1;
4695 	int			ip_hdr_len = IPV6_HDR_LEN;
4696 	size_t			ip_len;
4697 	icmp_t			*icmp = Q_TO_ICMP(q);
4698 	icmp_stack_t		*is = icmp->icmp_is;
4699 	ip6_pkt_t		ipp_s;	/* For ancillary data options */
4700 	ip6_pkt_t		*ipp = &ipp_s;
4701 	ip6_pkt_t		*tipp;
4702 	uint32_t		csum = 0;
4703 	uint_t			ignore = 0;
4704 	uint_t			option_exists = 0, is_sticky = 0;
4705 	uint8_t			*cp;
4706 	uint8_t			*nxthdr_ptr;
4707 	in6_addr_t		ip6_dst;
4708 
4709 	/*
4710 	 * If the local address is a mapped address return
4711 	 * an error.
4712 	 * It would be possible to send an IPv6 packet but the
4713 	 * response would never make it back to the application
4714 	 * since it is bound to a mapped address.
4715 	 */
4716 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4717 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4718 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4719 		return;
4720 	}
4721 
4722 	ipp->ipp_fields = 0;
4723 	ipp->ipp_sticky_ignored = 0;
4724 
4725 	/*
4726 	 * If TPI options passed in, feed it for verification and handling
4727 	 */
4728 	if (tudr_optlen != 0) {
4729 		int error;
4730 
4731 		if (icmp_unitdata_opt_process(q, mp, &error,
4732 		    (void *)ipp) < 0) {
4733 			/* failure */
4734 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4735 			icmp_ud_err(q, mp, error);
4736 			return;
4737 		}
4738 		ignore = ipp->ipp_sticky_ignored;
4739 		ASSERT(error == 0);
4740 	}
4741 
4742 	if (sin6->sin6_scope_id != 0 &&
4743 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4744 		/*
4745 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4746 		 * option nor ancillary data.  It needs to be
4747 		 * explicitly set in options_exists.
4748 		 */
4749 		option_exists |= IPPF_SCOPE_ID;
4750 	}
4751 
4752 	/*
4753 	 * Compute the destination address
4754 	 */
4755 	ip6_dst = sin6->sin6_addr;
4756 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
4757 		ip6_dst = ipv6_loopback;
4758 
4759 	/*
4760 	 * If we're not going to the same destination as last time, then
4761 	 * recompute the label required.  This is done in a separate routine to
4762 	 * avoid blowing up our stack here.
4763 	 */
4764 	if (is_system_labeled() &&
4765 	    !IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst) &&
4766 	    !icmp_update_label_v6(q, icmp, mp, &ip6_dst)) {
4767 		return;
4768 	}
4769 
4770 	/*
4771 	 * If there's a security label here, then we ignore any options the
4772 	 * user may try to set.  We keep the peer's label as a hidden sticky
4773 	 * option.
4774 	 */
4775 	if (icmp->icmp_label_len_v6 > 0) {
4776 		ignore &= ~IPPF_HOPOPTS;
4777 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
4778 	}
4779 
4780 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
4781 	    (ipp->ipp_fields == 0)) {
4782 		/* No sticky options nor ancillary data. */
4783 		goto no_options;
4784 	}
4785 
4786 	/*
4787 	 * Go through the options figuring out where each is going to
4788 	 * come from and build two masks.  The first mask indicates if
4789 	 * the option exists at all.  The second mask indicates if the
4790 	 * option is sticky or ancillary.
4791 	 */
4792 	if (!(ignore & IPPF_HOPOPTS)) {
4793 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
4794 			option_exists |= IPPF_HOPOPTS;
4795 			ip_hdr_len += ipp->ipp_hopoptslen;
4796 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
4797 			option_exists |= IPPF_HOPOPTS;
4798 			is_sticky |= IPPF_HOPOPTS;
4799 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen;
4800 		}
4801 	}
4802 
4803 	if (!(ignore & IPPF_RTHDR)) {
4804 		if (ipp->ipp_fields & IPPF_RTHDR) {
4805 			option_exists |= IPPF_RTHDR;
4806 			ip_hdr_len += ipp->ipp_rthdrlen;
4807 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
4808 			option_exists |= IPPF_RTHDR;
4809 			is_sticky |= IPPF_RTHDR;
4810 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
4811 		}
4812 	}
4813 
4814 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
4815 		/*
4816 		 * Need to have a router header to use these.
4817 		 */
4818 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
4819 			option_exists |= IPPF_RTDSTOPTS;
4820 			ip_hdr_len += ipp->ipp_rtdstoptslen;
4821 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
4822 			option_exists |= IPPF_RTDSTOPTS;
4823 			is_sticky |= IPPF_RTDSTOPTS;
4824 			ip_hdr_len +=
4825 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
4826 		}
4827 	}
4828 
4829 	if (!(ignore & IPPF_DSTOPTS)) {
4830 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
4831 			option_exists |= IPPF_DSTOPTS;
4832 			ip_hdr_len += ipp->ipp_dstoptslen;
4833 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
4834 			option_exists |= IPPF_DSTOPTS;
4835 			is_sticky |= IPPF_DSTOPTS;
4836 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
4837 		}
4838 	}
4839 
4840 	if (!(ignore & IPPF_IFINDEX)) {
4841 		if (ipp->ipp_fields & IPPF_IFINDEX) {
4842 			option_exists |= IPPF_IFINDEX;
4843 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
4844 			option_exists |= IPPF_IFINDEX;
4845 			is_sticky |= IPPF_IFINDEX;
4846 		}
4847 	}
4848 
4849 	if (!(ignore & IPPF_ADDR)) {
4850 		if (ipp->ipp_fields & IPPF_ADDR) {
4851 			option_exists |= IPPF_ADDR;
4852 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
4853 			option_exists |= IPPF_ADDR;
4854 			is_sticky |= IPPF_ADDR;
4855 		}
4856 	}
4857 
4858 	if (!(ignore & IPPF_DONTFRAG)) {
4859 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
4860 			option_exists |= IPPF_DONTFRAG;
4861 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
4862 			option_exists |= IPPF_DONTFRAG;
4863 			is_sticky |= IPPF_DONTFRAG;
4864 		}
4865 	}
4866 
4867 	if (!(ignore & IPPF_USE_MIN_MTU)) {
4868 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
4869 			option_exists |= IPPF_USE_MIN_MTU;
4870 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
4871 		    IPPF_USE_MIN_MTU) {
4872 			option_exists |= IPPF_USE_MIN_MTU;
4873 			is_sticky |= IPPF_USE_MIN_MTU;
4874 		}
4875 	}
4876 
4877 	if (!(ignore & IPPF_NEXTHOP)) {
4878 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
4879 			option_exists |= IPPF_NEXTHOP;
4880 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
4881 			option_exists |= IPPF_NEXTHOP;
4882 			is_sticky |= IPPF_NEXTHOP;
4883 		}
4884 	}
4885 
4886 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
4887 		option_exists |= IPPF_HOPLIMIT;
4888 	/* IPV6_HOPLIMIT can never be sticky */
4889 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
4890 
4891 	if (!(ignore & IPPF_UNICAST_HOPS) &&
4892 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
4893 		option_exists |= IPPF_UNICAST_HOPS;
4894 		is_sticky |= IPPF_UNICAST_HOPS;
4895 	}
4896 
4897 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
4898 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
4899 		option_exists |= IPPF_MULTICAST_HOPS;
4900 		is_sticky |= IPPF_MULTICAST_HOPS;
4901 	}
4902 
4903 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
4904 		/* This is a sticky socket option only */
4905 		option_exists |= IPPF_NO_CKSUM;
4906 		is_sticky |= IPPF_NO_CKSUM;
4907 	}
4908 
4909 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
4910 		/* This is a sticky socket option only */
4911 		option_exists |= IPPF_RAW_CKSUM;
4912 		is_sticky |= IPPF_RAW_CKSUM;
4913 	}
4914 
4915 	if (!(ignore & IPPF_TCLASS)) {
4916 		if (ipp->ipp_fields & IPPF_TCLASS) {
4917 			option_exists |= IPPF_TCLASS;
4918 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
4919 			option_exists |= IPPF_TCLASS;
4920 			is_sticky |= IPPF_TCLASS;
4921 		}
4922 	}
4923 
4924 no_options:
4925 
4926 	/*
4927 	 * If any options carried in the ip6i_t were specified, we
4928 	 * need to account for the ip6i_t in the data we'll be sending
4929 	 * down.
4930 	 */
4931 	if (option_exists & IPPF_HAS_IP6I)
4932 		ip_hdr_len += sizeof (ip6i_t);
4933 
4934 	/* check/fix buffer config, setup pointers into it */
4935 	mp1 = mp->b_cont;
4936 	ip6h = (ip6_t *)&mp1->b_rptr[-ip_hdr_len];
4937 	if ((mp1->b_datap->db_ref != 1) ||
4938 	    ((unsigned char *)ip6h < mp1->b_datap->db_base) ||
4939 	    !OK_32PTR(ip6h)) {
4940 		/* Try to get everything in a single mblk next time */
4941 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
4942 			icmp->icmp_max_hdr_len = ip_hdr_len;
4943 			(void) mi_set_sth_wroff(RD(q),
4944 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
4945 		}
4946 		mp1 = allocb(ip_hdr_len + is->is_wroff_extra, BPRI_LO);
4947 		if (!mp1) {
4948 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4949 			icmp_ud_err(q, mp, ENOMEM);
4950 			return;
4951 		}
4952 		mp1->b_cont = mp->b_cont;
4953 		mp1->b_wptr = mp1->b_datap->db_lim;
4954 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
4955 	}
4956 	mp1->b_rptr = (unsigned char *)ip6h;
4957 	ip6i = (ip6i_t *)ip6h;
4958 
4959 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
4960 	if (option_exists & IPPF_HAS_IP6I) {
4961 		ip6h = (ip6_t *)&ip6i[1];
4962 		ip6i->ip6i_flags = 0;
4963 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
4964 
4965 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
4966 		if (option_exists & IPPF_SCOPE_ID) {
4967 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4968 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
4969 		} else if (option_exists & IPPF_IFINDEX) {
4970 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
4971 			ASSERT(tipp->ipp_ifindex != 0);
4972 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4973 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
4974 		}
4975 
4976 		if (option_exists & IPPF_RAW_CKSUM) {
4977 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
4978 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
4979 		}
4980 
4981 		if (option_exists & IPPF_NO_CKSUM) {
4982 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
4983 		}
4984 
4985 		if (option_exists & IPPF_ADDR) {
4986 			/*
4987 			 * Enable per-packet source address verification if
4988 			 * IPV6_PKTINFO specified the source address.
4989 			 * ip6_src is set in the transport's _wput function.
4990 			 */
4991 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
4992 		}
4993 
4994 		if (option_exists & IPPF_DONTFRAG) {
4995 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
4996 		}
4997 
4998 		if (option_exists & IPPF_USE_MIN_MTU) {
4999 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
5000 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
5001 		}
5002 
5003 		if (option_exists & IPPF_NEXTHOP) {
5004 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
5005 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
5006 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
5007 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
5008 		}
5009 
5010 		/*
5011 		 * tell IP this is an ip6i_t private header
5012 		 */
5013 		ip6i->ip6i_nxt = IPPROTO_RAW;
5014 	}
5015 
5016 	/* Initialize IPv6 header */
5017 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5018 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
5019 
5020 	/* Set the hoplimit of the outgoing packet. */
5021 	if (option_exists & IPPF_HOPLIMIT) {
5022 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
5023 		ip6h->ip6_hops = ipp->ipp_hoplimit;
5024 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5025 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
5026 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
5027 		if (option_exists & IPPF_MULTICAST_HOPS)
5028 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5029 	} else {
5030 		ip6h->ip6_hops = icmp->icmp_ttl;
5031 		if (option_exists & IPPF_UNICAST_HOPS)
5032 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5033 	}
5034 
5035 	if (option_exists & IPPF_ADDR) {
5036 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
5037 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
5038 		ip6h->ip6_src = tipp->ipp_addr;
5039 	} else {
5040 		/*
5041 		 * The source address was not set using IPV6_PKTINFO.
5042 		 * First look at the bound source.
5043 		 * If unspecified fallback to __sin6_src_id.
5044 		 */
5045 		ip6h->ip6_src = icmp->icmp_v6src;
5046 		if (sin6->__sin6_src_id != 0 &&
5047 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
5048 			ip_srcid_find_id(sin6->__sin6_src_id,
5049 			    &ip6h->ip6_src, icmp->icmp_zoneid,
5050 			    is->is_netstack);
5051 		}
5052 	}
5053 
5054 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
5055 	cp = (uint8_t *)&ip6h[1];
5056 
5057 	/*
5058 	 * Here's where we have to start stringing together
5059 	 * any extension headers in the right order:
5060 	 * Hop-by-hop, destination, routing, and final destination opts.
5061 	 */
5062 	if (option_exists & IPPF_HOPOPTS) {
5063 		/* Hop-by-hop options */
5064 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
5065 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
5066 
5067 		*nxthdr_ptr = IPPROTO_HOPOPTS;
5068 		nxthdr_ptr = &hbh->ip6h_nxt;
5069 
5070 		bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
5071 		cp += tipp->ipp_hopoptslen;
5072 	}
5073 	/*
5074 	 * En-route destination options
5075 	 * Only do them if there's a routing header as well
5076 	 */
5077 	if (option_exists & IPPF_RTDSTOPTS) {
5078 		ip6_dest_t *dst = (ip6_dest_t *)cp;
5079 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
5080 
5081 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5082 		nxthdr_ptr = &dst->ip6d_nxt;
5083 
5084 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
5085 		cp += tipp->ipp_rtdstoptslen;
5086 	}
5087 	/*
5088 	 * Routing header next
5089 	 */
5090 	if (option_exists & IPPF_RTHDR) {
5091 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
5092 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
5093 
5094 		*nxthdr_ptr = IPPROTO_ROUTING;
5095 		nxthdr_ptr = &rt->ip6r_nxt;
5096 
5097 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
5098 		cp += tipp->ipp_rthdrlen;
5099 	}
5100 	/*
5101 	 * Do ultimate destination options
5102 	 */
5103 	if (option_exists & IPPF_DSTOPTS) {
5104 		ip6_dest_t *dest = (ip6_dest_t *)cp;
5105 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
5106 
5107 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5108 		nxthdr_ptr = &dest->ip6d_nxt;
5109 
5110 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
5111 		cp += tipp->ipp_dstoptslen;
5112 	}
5113 
5114 	/*
5115 	 * Now set the last header pointer to the proto passed in
5116 	 */
5117 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
5118 	*nxthdr_ptr = icmp->icmp_proto;
5119 
5120 	/*
5121 	 * Copy in the destination address
5122 	 */
5123 	ip6h->ip6_dst = ip6_dst;
5124 
5125 	ip6h->ip6_vcf =
5126 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
5127 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
5128 
5129 	if (option_exists & IPPF_TCLASS) {
5130 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
5131 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
5132 		    tipp->ipp_tclass);
5133 	}
5134 	if (option_exists & IPPF_RTHDR) {
5135 		ip6_rthdr_t	*rth;
5136 
5137 		/*
5138 		 * Perform any processing needed for source routing.
5139 		 * We know that all extension headers will be in the same mblk
5140 		 * as the IPv6 header.
5141 		 */
5142 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
5143 		if (rth != NULL && rth->ip6r_segleft != 0) {
5144 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
5145 				/*
5146 				 * Drop packet - only support Type 0 routing.
5147 				 * Notify the application as well.
5148 				 */
5149 				icmp_ud_err(q, mp, EPROTO);
5150 				BUMP_MIB(&is->is_rawip_mib,
5151 				    rawipOutErrors);
5152 				return;
5153 			}
5154 			/*
5155 			 * rth->ip6r_len is twice the number of
5156 			 * addresses in the header
5157 			 */
5158 			if (rth->ip6r_len & 0x1) {
5159 				icmp_ud_err(q, mp, EPROTO);
5160 				BUMP_MIB(&is->is_rawip_mib,
5161 				    rawipOutErrors);
5162 				return;
5163 			}
5164 			/*
5165 			 * Shuffle the routing header and ip6_dst
5166 			 * addresses, and get the checksum difference
5167 			 * between the first hop (in ip6_dst) and
5168 			 * the destination (in the last routing hdr entry).
5169 			 */
5170 			csum = ip_massage_options_v6(ip6h, rth,
5171 			    is->is_netstack);
5172 			/*
5173 			 * Verify that the first hop isn't a mapped address.
5174 			 * Routers along the path need to do this verification
5175 			 * for subsequent hops.
5176 			 */
5177 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
5178 				icmp_ud_err(q, mp, EADDRNOTAVAIL);
5179 				BUMP_MIB(&is->is_rawip_mib,
5180 				    rawipOutErrors);
5181 				return;
5182 			}
5183 		}
5184 	}
5185 
5186 	ip_len = mp1->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
5187 	if (mp1->b_cont != NULL)
5188 		ip_len += msgdsize(mp1->b_cont);
5189 
5190 	/*
5191 	 * Set the length into the IP header.
5192 	 * If the length is greater than the maximum allowed by IP,
5193 	 * then free the message and return. Do not try and send it
5194 	 * as this can cause problems in layers below.
5195 	 */
5196 	if (ip_len > IP_MAXPACKET) {
5197 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5198 		icmp_ud_err(q, mp, EMSGSIZE);
5199 		return;
5200 	}
5201 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
5202 		uint_t	cksum_off;	/* From ip6i == mp1->b_rptr */
5203 		uint16_t *cksum_ptr;
5204 		uint_t	ext_hdrs_len;
5205 
5206 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
5207 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
5208 		    icmp->icmp_checksum_off == 2);
5209 
5210 		/*
5211 		 * We make it easy for IP to include our pseudo header
5212 		 * by putting our length in uh_checksum, modified (if
5213 		 * we have a routing header) by the checksum difference
5214 		 * between the ultimate destination and first hop addresses.
5215 		 * Note: ICMPv6 must always checksum the packet.
5216 		 */
5217 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
5218 		if (cksum_off + sizeof (uint16_t) > mp1->b_wptr - mp1->b_rptr) {
5219 			if (!pullupmsg(mp1, cksum_off + sizeof (uint16_t))) {
5220 				BUMP_MIB(&is->is_rawip_mib,
5221 				    rawipOutErrors);
5222 				freemsg(mp);
5223 				return;
5224 			}
5225 			ip6i = (ip6i_t *)mp1->b_rptr;
5226 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
5227 				ip6h = (ip6_t *)&ip6i[1];
5228 			else
5229 				ip6h = (ip6_t *)ip6i;
5230 		}
5231 		/* Add payload length to checksum */
5232 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
5233 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
5234 		csum += htons(ip_len - ext_hdrs_len);
5235 
5236 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
5237 		csum = (csum & 0xFFFF) + (csum >> 16);
5238 		*cksum_ptr = (uint16_t)csum;
5239 	}
5240 
5241 #ifdef _LITTLE_ENDIAN
5242 	ip_len = htons(ip_len);
5243 #endif
5244 	ip6h->ip6_plen = (uint16_t)ip_len;
5245 
5246 	freeb(mp);
5247 
5248 	/* We're done. Pass the packet to IP */
5249 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
5250 	ip_output_v6(icmp->icmp_connp, mp1, q, IP_WPUT);
5251 }
5252 
5253 static void
5254 icmp_wput_other(queue_t *q, mblk_t *mp)
5255 {
5256 	uchar_t	*rptr = mp->b_rptr;
5257 	struct iocblk *iocp;
5258 #define	tudr ((struct T_unitdata_req *)rptr)
5259 	conn_t	*connp = Q_TO_CONN(q);
5260 	icmp_t	*icmp = connp->conn_icmp;
5261 	icmp_stack_t *is = icmp->icmp_is;
5262 	cred_t *cr;
5263 
5264 	cr = DB_CREDDEF(mp, connp->conn_cred);
5265 
5266 	switch (mp->b_datap->db_type) {
5267 	case M_PROTO:
5268 	case M_PCPROTO:
5269 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5270 			/*
5271 			 * If the message does not contain a PRIM_type,
5272 			 * throw it away.
5273 			 */
5274 			freemsg(mp);
5275 			return;
5276 		}
5277 		switch (((union T_primitives *)rptr)->type) {
5278 		case T_ADDR_REQ:
5279 			icmp_addr_req(q, mp);
5280 			return;
5281 		case O_T_BIND_REQ:
5282 		case T_BIND_REQ:
5283 			icmp_bind(q, mp);
5284 			return;
5285 		case T_CONN_REQ:
5286 			icmp_connect(q, mp);
5287 			return;
5288 		case T_CAPABILITY_REQ:
5289 			icmp_capability_req(q, mp);
5290 			return;
5291 		case T_INFO_REQ:
5292 			icmp_info_req(q, mp);
5293 			return;
5294 		case T_UNITDATA_REQ:
5295 			/*
5296 			 * If a T_UNITDATA_REQ gets here, the address must
5297 			 * be bad.  Valid T_UNITDATA_REQs are found above
5298 			 * and break to below this switch.
5299 			 */
5300 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
5301 			return;
5302 		case T_UNBIND_REQ:
5303 			icmp_unbind(q, mp);
5304 			return;
5305 
5306 		case T_SVR4_OPTMGMT_REQ:
5307 			if (!snmpcom_req(q, mp, icmp_snmp_set, ip_snmp_get,
5308 			    cr)) {
5309 				/* Only IP can return anything meaningful */
5310 				(void) svr4_optcom_req(q, mp, cr,
5311 				    &icmp_opt_obj, B_TRUE);
5312 			}
5313 			return;
5314 
5315 		case T_OPTMGMT_REQ:
5316 			/* Only IP can return anything meaningful */
5317 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj, B_TRUE);
5318 			return;
5319 
5320 		case T_DISCON_REQ:
5321 			icmp_disconnect(q, mp);
5322 			return;
5323 
5324 		/* The following TPI message is not supported by icmp. */
5325 		case O_T_CONN_RES:
5326 		case T_CONN_RES:
5327 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
5328 			return;
5329 
5330 		/* The following 3 TPI requests are illegal for icmp. */
5331 		case T_DATA_REQ:
5332 		case T_EXDATA_REQ:
5333 		case T_ORDREL_REQ:
5334 			freemsg(mp);
5335 			(void) putctl1(RD(q), M_ERROR, EPROTO);
5336 			return;
5337 		default:
5338 			break;
5339 		}
5340 		break;
5341 	case M_IOCTL:
5342 		iocp = (struct iocblk *)mp->b_rptr;
5343 		switch (iocp->ioc_cmd) {
5344 		case TI_GETPEERNAME:
5345 			if (icmp->icmp_state != TS_DATA_XFER) {
5346 				/*
5347 				 * If a default destination address has not
5348 				 * been associated with the stream, then we
5349 				 * don't know the peer's name.
5350 				 */
5351 				iocp->ioc_error = ENOTCONN;
5352 		err_ret:;
5353 				iocp->ioc_count = 0;
5354 				mp->b_datap->db_type = M_IOCACK;
5355 				qreply(q, mp);
5356 				return;
5357 			}
5358 			/* FALLTHRU */
5359 		case TI_GETMYNAME:
5360 			/*
5361 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
5362 			 * need to copyin the user's strbuf structure.
5363 			 * Processing will continue in the M_IOCDATA case
5364 			 * below.
5365 			 */
5366 			mi_copyin(q, mp, NULL,
5367 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
5368 			return;
5369 		case ND_SET:
5370 			/* nd_getset performs the necessary error checking */
5371 		case ND_GET:
5372 			if (nd_getset(q, is->is_nd, mp)) {
5373 				qreply(q, mp);
5374 				return;
5375 			}
5376 			break;
5377 		default:
5378 			break;
5379 		}
5380 		break;
5381 	case M_IOCDATA:
5382 		icmp_wput_iocdata(q, mp);
5383 		return;
5384 	default:
5385 		break;
5386 	}
5387 	ip_wput(q, mp);
5388 }
5389 
5390 /*
5391  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
5392  * messages.
5393  */
5394 static void
5395 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
5396 {
5397 	mblk_t	*mp1;
5398 	STRUCT_HANDLE(strbuf, sb);
5399 	icmp_t	*icmp;
5400 	in6_addr_t	v6addr;
5401 	ipaddr_t	v4addr;
5402 	uint32_t	flowinfo = 0;
5403 	int		addrlen;
5404 
5405 	/* Make sure it is one of ours. */
5406 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5407 	case TI_GETMYNAME:
5408 	case TI_GETPEERNAME:
5409 		break;
5410 	default:
5411 		icmp = Q_TO_ICMP(q);
5412 		ip_output(icmp->icmp_connp, mp, q, IP_WPUT);
5413 		return;
5414 	}
5415 	switch (mi_copy_state(q, mp, &mp1)) {
5416 	case -1:
5417 		return;
5418 	case MI_COPY_CASE(MI_COPY_IN, 1):
5419 		break;
5420 	case MI_COPY_CASE(MI_COPY_OUT, 1):
5421 		/*
5422 		 * The address has been copied out, so now
5423 		 * copyout the strbuf.
5424 		 */
5425 		mi_copyout(q, mp);
5426 		return;
5427 	case MI_COPY_CASE(MI_COPY_OUT, 2):
5428 		/*
5429 		 * The address and strbuf have been copied out.
5430 		 * We're done, so just acknowledge the original
5431 		 * M_IOCTL.
5432 		 */
5433 		mi_copy_done(q, mp, 0);
5434 		return;
5435 	default:
5436 		/*
5437 		 * Something strange has happened, so acknowledge
5438 		 * the original M_IOCTL with an EPROTO error.
5439 		 */
5440 		mi_copy_done(q, mp, EPROTO);
5441 		return;
5442 	}
5443 	/*
5444 	 * Now we have the strbuf structure for TI_GETMYNAME
5445 	 * and TI_GETPEERNAME.  Next we copyout the requested
5446 	 * address and then we'll copyout the strbuf.
5447 	 */
5448 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5449 	    (void *)mp1->b_rptr);
5450 	icmp = Q_TO_ICMP(q);
5451 	if (icmp->icmp_family == AF_INET)
5452 		addrlen = sizeof (sin_t);
5453 	else
5454 		addrlen = sizeof (sin6_t);
5455 
5456 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5457 		mi_copy_done(q, mp, EINVAL);
5458 		return;
5459 	}
5460 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5461 	case TI_GETMYNAME:
5462 		if (icmp->icmp_family == AF_INET) {
5463 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5464 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
5465 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5466 				v4addr = V4_PART_OF_V6(icmp->icmp_v6src);
5467 			} else {
5468 				/*
5469 				 * INADDR_ANY
5470 				 * icmp_v6src is not set, we might be bound to
5471 				 * broadcast/multicast. Use icmp_bound_v6src as
5472 				 * local address instead (that could
5473 				 * also still be INADDR_ANY)
5474 				 */
5475 				v4addr = V4_PART_OF_V6(icmp->icmp_bound_v6src);
5476 			}
5477 		} else {
5478 			/* icmp->icmp_family == AF_INET6 */
5479 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5480 				v6addr = icmp->icmp_v6src;
5481 			} else {
5482 				/*
5483 				 * UNSPECIFIED
5484 				 * icmp_v6src is not set, we might be bound to
5485 				 * broadcast/multicast. Use icmp_bound_v6src as
5486 				 * local address instead (that could
5487 				 * also still be UNSPECIFIED)
5488 				 */
5489 				v6addr = icmp->icmp_bound_v6src;
5490 			}
5491 		}
5492 		break;
5493 	case TI_GETPEERNAME:
5494 		if (icmp->icmp_family == AF_INET) {
5495 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5496 			v4addr = V4_PART_OF_V6(icmp->icmp_v6dst);
5497 		} else {
5498 			/* icmp->icmp_family == AF_INET6) */
5499 			v6addr = icmp->icmp_v6dst;
5500 			flowinfo = icmp->icmp_flowinfo;
5501 		}
5502 		break;
5503 	default:
5504 		mi_copy_done(q, mp, EPROTO);
5505 		return;
5506 	}
5507 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5508 	if (!mp1)
5509 		return;
5510 
5511 	if (icmp->icmp_family == AF_INET) {
5512 		sin_t *sin;
5513 
5514 		STRUCT_FSET(sb, len, (int)sizeof (sin_t));
5515 		sin = (sin_t *)mp1->b_rptr;
5516 		mp1->b_wptr = (uchar_t *)&sin[1];
5517 		*sin = sin_null;
5518 		sin->sin_family = AF_INET;
5519 		sin->sin_addr.s_addr = v4addr;
5520 	} else {
5521 		/* icmp->icmp_family == AF_INET6 */
5522 		sin6_t *sin6;
5523 
5524 		ASSERT(icmp->icmp_family == AF_INET6);
5525 		STRUCT_FSET(sb, len, (int)sizeof (sin6_t));
5526 		sin6 = (sin6_t *)mp1->b_rptr;
5527 		mp1->b_wptr = (uchar_t *)&sin6[1];
5528 		*sin6 = sin6_null;
5529 		sin6->sin6_family = AF_INET6;
5530 		sin6->sin6_flowinfo = flowinfo;
5531 		sin6->sin6_addr = v6addr;
5532 	}
5533 	/* Copy out the address */
5534 	mi_copyout(q, mp);
5535 }
5536 
5537 static int
5538 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5539     void *thisdg_attrs)
5540 {
5541 	conn_t	*connp = Q_TO_CONN(q);
5542 	struct T_unitdata_req *udreqp;
5543 	int is_absreq_failure;
5544 	cred_t *cr;
5545 
5546 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5547 	*errorp = 0;
5548 
5549 	cr = DB_CREDDEF(mp, connp->conn_cred);
5550 
5551 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5552 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5553 	    thisdg_attrs, &is_absreq_failure);
5554 
5555 	if (*errorp != 0) {
5556 		/*
5557 		 * Note: No special action needed in this
5558 		 * module for "is_absreq_failure"
5559 		 */
5560 		return (-1);		/* failure */
5561 	}
5562 	ASSERT(is_absreq_failure == 0);
5563 	return (0);	/* success */
5564 }
5565 
5566 void
5567 icmp_ddi_init(void)
5568 {
5569 	icmp_max_optsize = optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5570 	    icmp_opt_obj.odb_opt_arr_cnt);
5571 
5572 	/*
5573 	 * We want to be informed each time a stack is created or
5574 	 * destroyed in the kernel, so we can maintain the
5575 	 * set of icmp_stack_t's.
5576 	 */
5577 	netstack_register(NS_ICMP, rawip_stack_init, NULL, rawip_stack_fini);
5578 }
5579 
5580 void
5581 icmp_ddi_destroy(void)
5582 {
5583 	netstack_unregister(NS_ICMP);
5584 }
5585 
5586 /*
5587  * Initialize the ICMP stack instance.
5588  */
5589 static void *
5590 rawip_stack_init(netstackid_t stackid, netstack_t *ns)
5591 {
5592 	icmp_stack_t	*is;
5593 	icmpparam_t	*pa;
5594 
5595 	is = (icmp_stack_t *)kmem_zalloc(sizeof (*is), KM_SLEEP);
5596 	is->is_netstack = ns;
5597 
5598 	pa = (icmpparam_t *)kmem_alloc(sizeof (icmp_param_arr), KM_SLEEP);
5599 	is->is_param_arr = pa;
5600 	bcopy(icmp_param_arr, is->is_param_arr, sizeof (icmp_param_arr));
5601 
5602 	(void) icmp_param_register(&is->is_nd,
5603 	    is->is_param_arr, A_CNT(icmp_param_arr));
5604 	is->is_ksp = rawip_kstat_init(stackid);
5605 	return (is);
5606 }
5607 
5608 /*
5609  * Free the ICMP stack instance.
5610  */
5611 static void
5612 rawip_stack_fini(netstackid_t stackid, void *arg)
5613 {
5614 	icmp_stack_t *is = (icmp_stack_t *)arg;
5615 
5616 	nd_free(&is->is_nd);
5617 	kmem_free(is->is_param_arr, sizeof (icmp_param_arr));
5618 	is->is_param_arr = NULL;
5619 
5620 	rawip_kstat_fini(stackid, is->is_ksp);
5621 	is->is_ksp = NULL;
5622 	kmem_free(is, sizeof (*is));
5623 }
5624 
5625 static void *
5626 rawip_kstat_init(netstackid_t stackid) {
5627 	kstat_t	*ksp;
5628 
5629 	rawip_named_kstat_t template = {
5630 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5631 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5632 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5633 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5634 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5635 	};
5636 
5637 	ksp = kstat_create_netstack("icmp", 0, "rawip", "mib2",
5638 					KSTAT_TYPE_NAMED,
5639 					NUM_OF_FIELDS(rawip_named_kstat_t),
5640 					0, stackid);
5641 	if (ksp == NULL || ksp->ks_data == NULL)
5642 		return (NULL);
5643 
5644 	bcopy(&template, ksp->ks_data, sizeof (template));
5645 	ksp->ks_update = rawip_kstat_update;
5646 	ksp->ks_private = (void *)(uintptr_t)stackid;
5647 
5648 	kstat_install(ksp);
5649 	return (ksp);
5650 }
5651 
5652 static void
5653 rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
5654 {
5655 	if (ksp != NULL) {
5656 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
5657 		kstat_delete_netstack(ksp, stackid);
5658 	}
5659 }
5660 
5661 static int
5662 rawip_kstat_update(kstat_t *ksp, int rw)
5663 {
5664 	rawip_named_kstat_t *rawipkp;
5665 	netstackid_t	stackid = (netstackid_t)(uintptr_t)ksp->ks_private;
5666 	netstack_t	*ns;
5667 	icmp_stack_t	*is;
5668 
5669 	if ((ksp == NULL) || (ksp->ks_data == NULL))
5670 		return (EIO);
5671 
5672 	if (rw == KSTAT_WRITE)
5673 		return (EACCES);
5674 
5675 	rawipkp = (rawip_named_kstat_t *)ksp->ks_data;
5676 
5677 	ns = netstack_find_by_stackid(stackid);
5678 	if (ns == NULL)
5679 		return (-1);
5680 	is = ns->netstack_icmp;
5681 	if (is == NULL) {
5682 		netstack_rele(ns);
5683 		return (-1);
5684 	}
5685 	rawipkp->inDatagrams.value.ui32 =  is->is_rawip_mib.rawipInDatagrams;
5686 	rawipkp->inCksumErrs.value.ui32 =  is->is_rawip_mib.rawipInCksumErrs;
5687 	rawipkp->inErrors.value.ui32 =	   is->is_rawip_mib.rawipInErrors;
5688 	rawipkp->outDatagrams.value.ui32 = is->is_rawip_mib.rawipOutDatagrams;
5689 	rawipkp->outErrors.value.ui32 =	   is->is_rawip_mib.rawipOutErrors;
5690 	netstack_rele(ns);
5691 	return (0);
5692 }
5693