xref: /titanic_41/usr/src/uts/common/inet/ip/icmp.c (revision 7800901e60d340b6af88e94a2149805dcfcaaf56)
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_RECVIF:
1770 			*ptr = icmp->icmp_recvif;
1771 			break;	/* goto sizeof (int) option return */
1772 		case IP_RECVPKTINFO:
1773 			/*
1774 			 * This also handles IP_PKTINFO.
1775 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
1776 			 * Differentiation is based on the size of the argument
1777 			 * passed in.
1778 			 * This option is handled in IP which will return an
1779 			 * error for IP_PKTINFO as it's not supported as a
1780 			 * sticky option.
1781 			 */
1782 			return (-EINVAL);
1783 		/*
1784 		 * Cannot "get" the value of following options
1785 		 * at this level. Action is same as "default" to
1786 		 * which we fallthrough so we keep them in comments.
1787 		 *
1788 		 * case IP_ADD_MEMBERSHIP:
1789 		 * case IP_DROP_MEMBERSHIP:
1790 		 * case IP_BLOCK_SOURCE:
1791 		 * case IP_UNBLOCK_SOURCE:
1792 		 * case IP_ADD_SOURCE_MEMBERSHIP:
1793 		 * case IP_DROP_SOURCE_MEMBERSHIP:
1794 		 * case MCAST_JOIN_GROUP:
1795 		 * case MCAST_LEAVE_GROUP:
1796 		 * case MCAST_BLOCK_SOURCE:
1797 		 * case MCAST_UNBLOCK_SOURCE:
1798 		 * case MCAST_JOIN_SOURCE_GROUP:
1799 		 * case MCAST_LEAVE_SOURCE_GROUP:
1800 		 * case MRT_INIT:
1801 		 * case MRT_DONE:
1802 		 * case MRT_ADD_VIF:
1803 		 * case MRT_DEL_VIF:
1804 		 * case MRT_ADD_MFC:
1805 		 * case MRT_DEL_MFC:
1806 		 * case MRT_VERSION:
1807 		 * case MRT_ASSERT:
1808 		 * case IP_SEC_OPT:
1809 		 * case IP_DONTFAILOVER_IF:
1810 		 * case IP_NEXTHOP:
1811 		 */
1812 		default:
1813 			return (-1);
1814 		}
1815 		break;
1816 	case IPPROTO_IPV6:
1817 		/*
1818 		 * Only allow IPv6 option processing on native IPv6 sockets.
1819 		 */
1820 		if (icmp->icmp_family != AF_INET6)
1821 			return (-1);
1822 		switch (name) {
1823 		case IPV6_UNICAST_HOPS:
1824 			*i1 = (unsigned int)icmp->icmp_ttl;
1825 			break;
1826 		case IPV6_MULTICAST_IF:
1827 			/* 0 index if not set */
1828 			*i1 = icmp->icmp_multicast_if_index;
1829 			break;
1830 		case IPV6_MULTICAST_HOPS:
1831 			*i1 = icmp->icmp_multicast_ttl;
1832 			break;
1833 		case IPV6_MULTICAST_LOOP:
1834 			*i1 = connp->conn_multicast_loop;
1835 			break;
1836 		case IPV6_BOUND_IF:
1837 			/* Zero if not set */
1838 			*i1 = icmp->icmp_bound_if;
1839 			break;
1840 		case IPV6_UNSPEC_SRC:
1841 			*i1 = icmp->icmp_unspec_source;
1842 			break;
1843 		case IPV6_CHECKSUM:
1844 			/*
1845 			 * Return offset or -1 if no checksum offset.
1846 			 * Does not apply to IPPROTO_ICMPV6
1847 			 */
1848 			if (icmp->icmp_proto == IPPROTO_ICMPV6)
1849 				return (-1);
1850 
1851 			if (icmp->icmp_raw_checksum) {
1852 				*i1 = icmp->icmp_checksum_off;
1853 			} else {
1854 				*i1 = -1;
1855 			}
1856 			break;
1857 		case IPV6_JOIN_GROUP:
1858 		case IPV6_LEAVE_GROUP:
1859 		case MCAST_JOIN_GROUP:
1860 		case MCAST_LEAVE_GROUP:
1861 		case MCAST_BLOCK_SOURCE:
1862 		case MCAST_UNBLOCK_SOURCE:
1863 		case MCAST_JOIN_SOURCE_GROUP:
1864 		case MCAST_LEAVE_SOURCE_GROUP:
1865 			/* cannot "get" the value for these */
1866 			return (-1);
1867 		case IPV6_RECVPKTINFO:
1868 			*i1 = icmp->icmp_ip_recvpktinfo;
1869 			break;
1870 		case IPV6_RECVTCLASS:
1871 			*i1 = icmp->icmp_ipv6_recvtclass;
1872 			break;
1873 		case IPV6_RECVPATHMTU:
1874 			*i1 = icmp->icmp_ipv6_recvpathmtu;
1875 			break;
1876 		case IPV6_V6ONLY:
1877 			*i1 = 1;
1878 			break;
1879 		case IPV6_RECVHOPLIMIT:
1880 			*i1 = icmp->icmp_ipv6_recvhoplimit;
1881 			break;
1882 		case IPV6_RECVHOPOPTS:
1883 			*i1 = icmp->icmp_ipv6_recvhopopts;
1884 			break;
1885 		case IPV6_RECVDSTOPTS:
1886 			*i1 = icmp->icmp_ipv6_recvdstopts;
1887 			break;
1888 		case _OLD_IPV6_RECVDSTOPTS:
1889 			*i1 = icmp->icmp_old_ipv6_recvdstopts;
1890 			break;
1891 		case IPV6_RECVRTHDRDSTOPTS:
1892 			*i1 = icmp->icmp_ipv6_recvrtdstopts;
1893 			break;
1894 		case IPV6_RECVRTHDR:
1895 			*i1 = icmp->icmp_ipv6_recvrthdr;
1896 			break;
1897 		case IPV6_PKTINFO: {
1898 			/* XXX assumes that caller has room for max size! */
1899 			struct in6_pktinfo *pkti;
1900 
1901 			pkti = (struct in6_pktinfo *)ptr;
1902 			if (ipp->ipp_fields & IPPF_IFINDEX)
1903 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
1904 			else
1905 				pkti->ipi6_ifindex = 0;
1906 			if (ipp->ipp_fields & IPPF_ADDR)
1907 				pkti->ipi6_addr = ipp->ipp_addr;
1908 			else
1909 				pkti->ipi6_addr = ipv6_all_zeros;
1910 			return (sizeof (struct in6_pktinfo));
1911 		}
1912 		case IPV6_NEXTHOP: {
1913 			sin6_t *sin6 = (sin6_t *)ptr;
1914 
1915 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
1916 				return (0);
1917 			*sin6 = sin6_null;
1918 			sin6->sin6_family = AF_INET6;
1919 			sin6->sin6_addr = ipp->ipp_nexthop;
1920 			return (sizeof (sin6_t));
1921 		}
1922 		case IPV6_HOPOPTS:
1923 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
1924 				return (0);
1925 			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
1926 				return (0);
1927 			bcopy((char *)ipp->ipp_hopopts +
1928 			    icmp->icmp_label_len_v6, ptr,
1929 			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1930 			if (icmp->icmp_label_len_v6 > 0) {
1931 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
1932 				ptr[1] = (ipp->ipp_hopoptslen -
1933 				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
1934 			}
1935 			return (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1936 		case IPV6_RTHDRDSTOPTS:
1937 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
1938 				return (0);
1939 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
1940 			return (ipp->ipp_rtdstoptslen);
1941 		case IPV6_RTHDR:
1942 			if (!(ipp->ipp_fields & IPPF_RTHDR))
1943 				return (0);
1944 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
1945 			return (ipp->ipp_rthdrlen);
1946 		case IPV6_DSTOPTS:
1947 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
1948 				return (0);
1949 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
1950 			return (ipp->ipp_dstoptslen);
1951 		case IPV6_PATHMTU:
1952 			if (!(ipp->ipp_fields & IPPF_PATHMTU))
1953 				return (0);
1954 
1955 			return (ip_fill_mtuinfo(&icmp->icmp_v6dst, 0,
1956 			    (struct ip6_mtuinfo *)ptr, is->is_netstack));
1957 		case IPV6_TCLASS:
1958 			if (ipp->ipp_fields & IPPF_TCLASS)
1959 				*i1 = ipp->ipp_tclass;
1960 			else
1961 				*i1 = IPV6_FLOW_TCLASS(
1962 				    IPV6_DEFAULT_VERS_AND_FLOW);
1963 			break;
1964 		default:
1965 			return (-1);
1966 		}
1967 		break;
1968 	case IPPROTO_ICMPV6:
1969 		/*
1970 		 * Only allow IPv6 option processing on native IPv6 sockets.
1971 		 */
1972 		if (icmp->icmp_family != AF_INET6)
1973 			return (-1);
1974 
1975 		if (icmp->icmp_proto != IPPROTO_ICMPV6)
1976 			return (-1);
1977 
1978 		switch (name) {
1979 		case ICMP6_FILTER:
1980 			if (icmp->icmp_filter == NULL) {
1981 				/* Make it look like "pass all" */
1982 				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1983 			} else {
1984 				(void) bcopy(icmp->icmp_filter, ptr,
1985 				    sizeof (icmp6_filter_t));
1986 			}
1987 			return (sizeof (icmp6_filter_t));
1988 		default:
1989 			return (-1);
1990 		}
1991 	default:
1992 		return (-1);
1993 	}
1994 	return (sizeof (int));
1995 }
1996 
1997 /*
1998  * This routine retrieves the current status of socket options.
1999  * It returns the size of the option retrieved.
2000  */
2001 int
2002 icmp_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
2003 {
2004 	icmp_t  *icmp = Q_TO_ICMP(q);
2005 	int 	err;
2006 
2007 	rw_enter(&icmp->icmp_rwlock, RW_READER);
2008 	err = icmp_opt_get_locked(q, level, name, ptr);
2009 	rw_exit(&icmp->icmp_rwlock);
2010 	return (err);
2011 }
2012 
2013 
2014 /* This routine sets socket options. */
2015 /* ARGSUSED */
2016 int
2017 icmp_opt_set_locked(queue_t *q, uint_t optset_context, int level, int name,
2018     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
2019     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
2020 {
2021 	conn_t	*connp = Q_TO_CONN(q);
2022 	icmp_t	*icmp = connp->conn_icmp;
2023 	icmp_stack_t *is = icmp->icmp_is;
2024 	int	*i1 = (int *)invalp;
2025 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2026 	boolean_t checkonly;
2027 	int	error;
2028 
2029 	switch (optset_context) {
2030 	case SETFN_OPTCOM_CHECKONLY:
2031 		checkonly = B_TRUE;
2032 		/*
2033 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
2034 		 * inlen != 0 implies value supplied and
2035 		 * 	we have to "pretend" to set it.
2036 		 * inlen == 0 implies that there is no
2037 		 * 	value part in T_CHECK request and just validation
2038 		 * done elsewhere should be enough, we just return here.
2039 		 */
2040 		if (inlen == 0) {
2041 			*outlenp = 0;
2042 			return (0);
2043 		}
2044 		break;
2045 	case SETFN_OPTCOM_NEGOTIATE:
2046 		checkonly = B_FALSE;
2047 		break;
2048 	case SETFN_UD_NEGOTIATE:
2049 	case SETFN_CONN_NEGOTIATE:
2050 		checkonly = B_FALSE;
2051 		/*
2052 		 * Negotiating local and "association-related" options
2053 		 * through T_UNITDATA_REQ.
2054 		 *
2055 		 * Following routine can filter out ones we do not
2056 		 * want to be "set" this way.
2057 		 */
2058 		if (!icmp_opt_allow_udr_set(level, name)) {
2059 			*outlenp = 0;
2060 			return (EINVAL);
2061 		}
2062 		break;
2063 	default:
2064 		/*
2065 		 * We should never get here
2066 		 */
2067 		*outlenp = 0;
2068 		return (EINVAL);
2069 	}
2070 
2071 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
2072 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
2073 
2074 	/*
2075 	 * For fixed length options, no sanity check
2076 	 * of passed in length is done. It is assumed *_optcom_req()
2077 	 * routines do the right thing.
2078 	 */
2079 
2080 	switch (level) {
2081 	case SOL_SOCKET:
2082 		switch (name) {
2083 		case SO_DEBUG:
2084 			if (!checkonly)
2085 				icmp->icmp_debug = onoff;
2086 			break;
2087 		case SO_PROTOTYPE:
2088 			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
2089 			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
2090 			    secpolicy_net_rawaccess(cr) != 0) {
2091 				*outlenp = 0;
2092 				return (EACCES);
2093 			}
2094 			/* Can't use IPPROTO_RAW with IPv6 */
2095 			if ((*i1 & 0xFF) == IPPROTO_RAW &&
2096 			    icmp->icmp_family == AF_INET6) {
2097 				*outlenp = 0;
2098 				return (EPROTONOSUPPORT);
2099 			}
2100 			if (checkonly) {
2101 				/* T_CHECK case */
2102 				*(int *)outvalp = (*i1 & 0xFF);
2103 				break;
2104 			}
2105 			icmp->icmp_proto = *i1 & 0xFF;
2106 			if ((icmp->icmp_proto == IPPROTO_RAW ||
2107 			    icmp->icmp_proto == IPPROTO_IGMP) &&
2108 			    icmp->icmp_family == AF_INET)
2109 				icmp->icmp_hdrincl = 1;
2110 			else
2111 				icmp->icmp_hdrincl = 0;
2112 
2113 			if (icmp->icmp_family == AF_INET6 &&
2114 			    icmp->icmp_proto == IPPROTO_ICMPV6) {
2115 				/* Set offset for icmp6_cksum */
2116 				icmp->icmp_raw_checksum = 0;
2117 				icmp->icmp_checksum_off = 2;
2118 			}
2119 			if (icmp->icmp_proto == IPPROTO_UDP ||
2120 			    icmp->icmp_proto == IPPROTO_TCP ||
2121 			    icmp->icmp_proto == IPPROTO_SCTP) {
2122 				icmp->icmp_no_tp_cksum = 1;
2123 				icmp->icmp_sticky_ipp.ipp_fields |=
2124 				    IPPF_NO_CKSUM;
2125 			} else {
2126 				icmp->icmp_no_tp_cksum = 0;
2127 				icmp->icmp_sticky_ipp.ipp_fields &=
2128 				    ~IPPF_NO_CKSUM;
2129 			}
2130 
2131 			if (icmp->icmp_filter != NULL &&
2132 			    icmp->icmp_proto != IPPROTO_ICMPV6) {
2133 				kmem_free(icmp->icmp_filter,
2134 				    sizeof (icmp6_filter_t));
2135 				icmp->icmp_filter = NULL;
2136 			}
2137 
2138 			/* Rebuild the header template */
2139 			error = icmp_build_hdrs(icmp);
2140 			if (error != 0) {
2141 				*outlenp = 0;
2142 				return (error);
2143 			}
2144 
2145 			/*
2146 			 * For SCTP, we don't use icmp_bind_proto() for
2147 			 * raw socket binding.  Note that we do not need
2148 			 * to set *outlenp.
2149 			 * FIXME: how does SCTP work?
2150 			 */
2151 			if (icmp->icmp_proto == IPPROTO_SCTP)
2152 				return (0);
2153 
2154 			*outlenp = sizeof (int);
2155 			*(int *)outvalp = *i1 & 0xFF;
2156 
2157 			/* Drop lock across the bind operation */
2158 			rw_exit(&icmp->icmp_rwlock);
2159 			icmp_bind_proto(q);
2160 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2161 			return (0);
2162 		case SO_REUSEADDR:
2163 			if (!checkonly)
2164 				icmp->icmp_reuseaddr = onoff;
2165 			break;
2166 
2167 		/*
2168 		 * The following three items are available here,
2169 		 * but are only meaningful to IP.
2170 		 */
2171 		case SO_DONTROUTE:
2172 			if (!checkonly)
2173 				icmp->icmp_dontroute = onoff;
2174 			break;
2175 		case SO_USELOOPBACK:
2176 			if (!checkonly)
2177 				icmp->icmp_useloopback = onoff;
2178 			break;
2179 		case SO_BROADCAST:
2180 			if (!checkonly)
2181 				icmp->icmp_broadcast = onoff;
2182 			break;
2183 
2184 		case SO_SNDBUF:
2185 			if (*i1 > is->is_max_buf) {
2186 				*outlenp = 0;
2187 				return (ENOBUFS);
2188 			}
2189 			if (!checkonly) {
2190 				q->q_hiwat = *i1;
2191 			}
2192 			break;
2193 		case SO_RCVBUF:
2194 			if (*i1 > is->is_max_buf) {
2195 				*outlenp = 0;
2196 				return (ENOBUFS);
2197 			}
2198 			if (!checkonly) {
2199 				RD(q)->q_hiwat = *i1;
2200 				rw_exit(&icmp->icmp_rwlock);
2201 				(void) mi_set_sth_hiwat(RD(q), *i1);
2202 				rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2203 			}
2204 			break;
2205 		case SO_DGRAM_ERRIND:
2206 			if (!checkonly)
2207 				icmp->icmp_dgram_errind = onoff;
2208 			break;
2209 		case SO_ALLZONES:
2210 			/*
2211 			 * "soft" error (negative)
2212 			 * option not handled at this level
2213 			 * Note: Do not modify *outlenp
2214 			 */
2215 			return (-EINVAL);
2216 		case SO_TIMESTAMP:
2217 			if (!checkonly) {
2218 				icmp->icmp_timestamp = onoff;
2219 			}
2220 			break;
2221 		case SO_MAC_EXEMPT:
2222 			if (secpolicy_net_mac_aware(cr) != 0 ||
2223 			    icmp->icmp_state != TS_UNBND)
2224 				return (EACCES);
2225 			if (!checkonly)
2226 				icmp->icmp_mac_exempt = onoff;
2227 			break;
2228 		/*
2229 		 * Following three not meaningful for icmp
2230 		 * Action is same as "default" so we keep them
2231 		 * in comments.
2232 		 * case SO_LINGER:
2233 		 * case SO_KEEPALIVE:
2234 		 * case SO_OOBINLINE:
2235 		 */
2236 		default:
2237 			*outlenp = 0;
2238 			return (EINVAL);
2239 		}
2240 		break;
2241 	case IPPROTO_IP:
2242 		/*
2243 		 * Only allow IPv4 option processing on IPv4 sockets.
2244 		 */
2245 		if (icmp->icmp_family != AF_INET) {
2246 			*outlenp = 0;
2247 			return (ENOPROTOOPT);
2248 		}
2249 		switch (name) {
2250 		case IP_OPTIONS:
2251 		case T_IP_OPTIONS:
2252 			/* Save options for use by IP. */
2253 			if ((inlen & 0x3) ||
2254 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2255 				*outlenp = 0;
2256 				return (EINVAL);
2257 			}
2258 			if (checkonly)
2259 				break;
2260 
2261 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2262 			    &icmp->icmp_ip_snd_options_len,
2263 			    icmp->icmp_label_len, invalp, inlen)) {
2264 				*outlenp = 0;
2265 				return (ENOMEM);
2266 			}
2267 
2268 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2269 			    icmp->icmp_ip_snd_options_len;
2270 			rw_exit(&icmp->icmp_rwlock);
2271 			(void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len +
2272 			    is->is_wroff_extra);
2273 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2274 			break;
2275 		case IP_HDRINCL:
2276 			if (!checkonly)
2277 				icmp->icmp_hdrincl = onoff;
2278 			break;
2279 		case IP_TOS:
2280 		case T_IP_TOS:
2281 			if (!checkonly) {
2282 				icmp->icmp_type_of_service = (uint8_t)*i1;
2283 			}
2284 			break;
2285 		case IP_TTL:
2286 			if (!checkonly) {
2287 				icmp->icmp_ttl = (uint8_t)*i1;
2288 			}
2289 			break;
2290 		case IP_MULTICAST_IF:
2291 			/*
2292 			 * TODO should check OPTMGMT reply and undo this if
2293 			 * there is an error.
2294 			 */
2295 			if (!checkonly)
2296 				icmp->icmp_multicast_if_addr = *i1;
2297 			break;
2298 		case IP_MULTICAST_TTL:
2299 			if (!checkonly)
2300 				icmp->icmp_multicast_ttl = *invalp;
2301 			break;
2302 		case IP_MULTICAST_LOOP:
2303 			if (!checkonly) {
2304 				connp->conn_multicast_loop =
2305 				    (*invalp == 0) ? 0 : 1;
2306 			}
2307 			break;
2308 		case IP_BOUND_IF:
2309 			if (!checkonly)
2310 				icmp->icmp_bound_if = *i1;
2311 			break;
2312 		case IP_UNSPEC_SRC:
2313 			if (!checkonly)
2314 				icmp->icmp_unspec_source = onoff;
2315 			break;
2316 		case IP_RECVIF:
2317 			if (!checkonly)
2318 				icmp->icmp_recvif = onoff;
2319 			/*
2320 			 * pass to ip
2321 			 */
2322 			return (-EINVAL);
2323 		case IP_PKTINFO: {
2324 			/*
2325 			 * This also handles IP_RECVPKTINFO.
2326 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2327 			 * Differentiation is based on the size of the argument
2328 			 * passed in.
2329 			 */
2330 			struct in_pktinfo *pktinfop;
2331 			ip4_pkt_t *attr_pktinfop;
2332 
2333 			if (checkonly)
2334 				break;
2335 
2336 			if (inlen == sizeof (int)) {
2337 				/*
2338 				 * This is IP_RECVPKTINFO option.
2339 				 * Keep a local copy of wether this option is
2340 				 * set or not and pass it down to IP for
2341 				 * processing.
2342 				 */
2343 				icmp->icmp_ip_recvpktinfo = onoff;
2344 				return (-EINVAL);
2345 			}
2346 
2347 
2348 			if (inlen != sizeof (struct in_pktinfo))
2349 				return (EINVAL);
2350 
2351 			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
2352 			    == NULL) {
2353 				/*
2354 				 * sticky option is not supported
2355 				 */
2356 				return (EINVAL);
2357 			}
2358 
2359 			pktinfop = (struct in_pktinfo *)invalp;
2360 
2361 			/*
2362 			 * Atleast one of the values should be specified
2363 			 */
2364 			if (pktinfop->ipi_ifindex == 0 &&
2365 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2366 				return (EINVAL);
2367 			}
2368 
2369 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2370 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2371 		}
2372 			break;
2373 		case IP_ADD_MEMBERSHIP:
2374 		case IP_DROP_MEMBERSHIP:
2375 		case IP_BLOCK_SOURCE:
2376 		case IP_UNBLOCK_SOURCE:
2377 		case IP_ADD_SOURCE_MEMBERSHIP:
2378 		case IP_DROP_SOURCE_MEMBERSHIP:
2379 		case MCAST_JOIN_GROUP:
2380 		case MCAST_LEAVE_GROUP:
2381 		case MCAST_BLOCK_SOURCE:
2382 		case MCAST_UNBLOCK_SOURCE:
2383 		case MCAST_JOIN_SOURCE_GROUP:
2384 		case MCAST_LEAVE_SOURCE_GROUP:
2385 		case MRT_INIT:
2386 		case MRT_DONE:
2387 		case MRT_ADD_VIF:
2388 		case MRT_DEL_VIF:
2389 		case MRT_ADD_MFC:
2390 		case MRT_DEL_MFC:
2391 		case MRT_VERSION:
2392 		case MRT_ASSERT:
2393 		case IP_SEC_OPT:
2394 		case IP_DONTFAILOVER_IF:
2395 		case IP_NEXTHOP:
2396 			/*
2397 			 * "soft" error (negative)
2398 			 * option not handled at this level
2399 			 * Note: Do not modify *outlenp
2400 			 */
2401 			return (-EINVAL);
2402 		default:
2403 			*outlenp = 0;
2404 			return (EINVAL);
2405 		}
2406 		break;
2407 	case IPPROTO_IPV6: {
2408 		ip6_pkt_t		*ipp;
2409 		boolean_t		sticky;
2410 
2411 		if (icmp->icmp_family != AF_INET6) {
2412 			*outlenp = 0;
2413 			return (ENOPROTOOPT);
2414 		}
2415 		/*
2416 		 * Deal with both sticky options and ancillary data
2417 		 */
2418 		if (thisdg_attrs == NULL) {
2419 			/* sticky options, or none */
2420 			ipp = &icmp->icmp_sticky_ipp;
2421 			sticky = B_TRUE;
2422 		} else {
2423 			/* ancillary data */
2424 			ipp = (ip6_pkt_t *)thisdg_attrs;
2425 			sticky = B_FALSE;
2426 		}
2427 
2428 		switch (name) {
2429 		case IPV6_MULTICAST_IF:
2430 			if (!checkonly)
2431 				icmp->icmp_multicast_if_index = *i1;
2432 			break;
2433 		case IPV6_UNICAST_HOPS:
2434 			/* -1 means use default */
2435 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2436 				*outlenp = 0;
2437 				return (EINVAL);
2438 			}
2439 			if (!checkonly) {
2440 				if (*i1 == -1) {
2441 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2442 					    is->is_ipv6_hoplimit;
2443 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2444 					/* Pass modified value to IP. */
2445 					*i1 = ipp->ipp_hoplimit;
2446 				} else {
2447 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2448 					    (uint8_t)*i1;
2449 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2450 				}
2451 				/* Rebuild the header template */
2452 				error = icmp_build_hdrs(icmp);
2453 				if (error != 0) {
2454 					*outlenp = 0;
2455 					return (error);
2456 				}
2457 			}
2458 			break;
2459 		case IPV6_MULTICAST_HOPS:
2460 			/* -1 means use default */
2461 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2462 				*outlenp = 0;
2463 				return (EINVAL);
2464 			}
2465 			if (!checkonly) {
2466 				if (*i1 == -1) {
2467 					icmp->icmp_multicast_ttl =
2468 					    ipp->ipp_multicast_hops =
2469 					    IP_DEFAULT_MULTICAST_TTL;
2470 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2471 					/* Pass modified value to IP. */
2472 					*i1 = icmp->icmp_multicast_ttl;
2473 				} else {
2474 					icmp->icmp_multicast_ttl =
2475 					    ipp->ipp_multicast_hops =
2476 					    (uint8_t)*i1;
2477 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2478 				}
2479 			}
2480 			break;
2481 		case IPV6_MULTICAST_LOOP:
2482 			if (*i1 != 0 && *i1 != 1) {
2483 				*outlenp = 0;
2484 				return (EINVAL);
2485 			}
2486 			if (!checkonly)
2487 				connp->conn_multicast_loop = *i1;
2488 			break;
2489 		case IPV6_CHECKSUM:
2490 			/*
2491 			 * Integer offset into the user data of where the
2492 			 * checksum is located.
2493 			 * Offset of -1 disables option.
2494 			 * Does not apply to IPPROTO_ICMPV6.
2495 			 */
2496 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2497 				*outlenp = 0;
2498 				return (EINVAL);
2499 			}
2500 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2501 				/* Negative or not 16 bit aligned offset */
2502 				*outlenp = 0;
2503 				return (EINVAL);
2504 			}
2505 			if (checkonly)
2506 				break;
2507 
2508 			if (*i1 == -1) {
2509 				icmp->icmp_raw_checksum = 0;
2510 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2511 			} else {
2512 				icmp->icmp_raw_checksum = 1;
2513 				icmp->icmp_checksum_off = *i1;
2514 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2515 			}
2516 			/* Rebuild the header template */
2517 			error = icmp_build_hdrs(icmp);
2518 			if (error != 0) {
2519 				*outlenp = 0;
2520 				return (error);
2521 			}
2522 			break;
2523 		case IPV6_JOIN_GROUP:
2524 		case IPV6_LEAVE_GROUP:
2525 		case MCAST_JOIN_GROUP:
2526 		case MCAST_LEAVE_GROUP:
2527 		case MCAST_BLOCK_SOURCE:
2528 		case MCAST_UNBLOCK_SOURCE:
2529 		case MCAST_JOIN_SOURCE_GROUP:
2530 		case MCAST_LEAVE_SOURCE_GROUP:
2531 			/*
2532 			 * "soft" error (negative)
2533 			 * option not handled at this level
2534 			 * Note: Do not modify *outlenp
2535 			 */
2536 			return (-EINVAL);
2537 		case IPV6_BOUND_IF:
2538 			if (!checkonly)
2539 				icmp->icmp_bound_if = *i1;
2540 			break;
2541 		case IPV6_UNSPEC_SRC:
2542 			if (!checkonly)
2543 				icmp->icmp_unspec_source = onoff;
2544 			break;
2545 		case IPV6_RECVTCLASS:
2546 			if (!checkonly)
2547 				icmp->icmp_ipv6_recvtclass = onoff;
2548 			break;
2549 		/*
2550 		 * Set boolean switches for ancillary data delivery
2551 		 */
2552 		case IPV6_RECVPKTINFO:
2553 			if (!checkonly)
2554 				icmp->icmp_ip_recvpktinfo = onoff;
2555 			break;
2556 		case IPV6_RECVPATHMTU:
2557 			if (!checkonly)
2558 				icmp->icmp_ipv6_recvpathmtu = onoff;
2559 			break;
2560 		case IPV6_RECVHOPLIMIT:
2561 			if (!checkonly)
2562 				icmp->icmp_ipv6_recvhoplimit = onoff;
2563 			break;
2564 		case IPV6_RECVHOPOPTS:
2565 			if (!checkonly)
2566 				icmp->icmp_ipv6_recvhopopts = onoff;
2567 			break;
2568 		case IPV6_RECVDSTOPTS:
2569 			if (!checkonly)
2570 				icmp->icmp_ipv6_recvdstopts = onoff;
2571 			break;
2572 		case _OLD_IPV6_RECVDSTOPTS:
2573 			if (!checkonly)
2574 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2575 			break;
2576 		case IPV6_RECVRTHDRDSTOPTS:
2577 			if (!checkonly)
2578 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2579 			break;
2580 		case IPV6_RECVRTHDR:
2581 			if (!checkonly)
2582 				icmp->icmp_ipv6_recvrthdr = onoff;
2583 			break;
2584 		/*
2585 		 * Set sticky options or ancillary data.
2586 		 * If sticky options, (re)build any extension headers
2587 		 * that might be needed as a result.
2588 		 */
2589 		case IPV6_PKTINFO:
2590 			/*
2591 			 * The source address and ifindex are verified
2592 			 * in ip_opt_set(). For ancillary data the
2593 			 * source address is checked in ip_wput_v6.
2594 			 */
2595 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2596 				return (EINVAL);
2597 			if (checkonly)
2598 				break;
2599 
2600 			if (inlen == 0) {
2601 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2602 				ipp->ipp_sticky_ignored |=
2603 				    (IPPF_IFINDEX|IPPF_ADDR);
2604 			} else {
2605 				struct in6_pktinfo *pkti;
2606 
2607 				pkti = (struct in6_pktinfo *)invalp;
2608 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2609 				ipp->ipp_addr = pkti->ipi6_addr;
2610 				if (ipp->ipp_ifindex != 0)
2611 					ipp->ipp_fields |= IPPF_IFINDEX;
2612 				else
2613 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2614 				if (!IN6_IS_ADDR_UNSPECIFIED(
2615 				    &ipp->ipp_addr))
2616 					ipp->ipp_fields |= IPPF_ADDR;
2617 				else
2618 					ipp->ipp_fields &= ~IPPF_ADDR;
2619 			}
2620 			if (sticky) {
2621 				error = icmp_build_hdrs(icmp);
2622 				if (error != 0)
2623 					return (error);
2624 			}
2625 			break;
2626 		case IPV6_HOPLIMIT:
2627 			/* This option can only be used as ancillary data. */
2628 			if (sticky)
2629 				return (EINVAL);
2630 			if (inlen != 0 && inlen != sizeof (int))
2631 				return (EINVAL);
2632 			if (checkonly)
2633 				break;
2634 
2635 			if (inlen == 0) {
2636 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2637 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2638 			} else {
2639 				if (*i1 > 255 || *i1 < -1)
2640 					return (EINVAL);
2641 				if (*i1 == -1)
2642 					ipp->ipp_hoplimit =
2643 					    is->is_ipv6_hoplimit;
2644 				else
2645 					ipp->ipp_hoplimit = *i1;
2646 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2647 			}
2648 			break;
2649 		case IPV6_TCLASS:
2650 			/*
2651 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2652 			 * and [0, 255] as the actualy traffic class.
2653 			 */
2654 			if (inlen != 0 && inlen != sizeof (int))
2655 				return (EINVAL);
2656 			if (checkonly)
2657 				break;
2658 
2659 			if (inlen == 0) {
2660 				ipp->ipp_fields &= ~IPPF_TCLASS;
2661 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2662 			} else {
2663 				if (*i1 >= 256 || *i1 < -1)
2664 					return (EINVAL);
2665 				if (*i1 == -1) {
2666 					ipp->ipp_tclass =
2667 					    IPV6_FLOW_TCLASS(
2668 					    IPV6_DEFAULT_VERS_AND_FLOW);
2669 				} else {
2670 					ipp->ipp_tclass = *i1;
2671 				}
2672 				ipp->ipp_fields |= IPPF_TCLASS;
2673 			}
2674 			if (sticky) {
2675 				error = icmp_build_hdrs(icmp);
2676 				if (error != 0)
2677 					return (error);
2678 			}
2679 			break;
2680 		case IPV6_NEXTHOP:
2681 			/*
2682 			 * IP will verify that the nexthop is reachable
2683 			 * and fail for sticky options.
2684 			 */
2685 			if (inlen != 0 && inlen != sizeof (sin6_t))
2686 				return (EINVAL);
2687 			if (checkonly)
2688 				break;
2689 
2690 			if (inlen == 0) {
2691 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2692 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2693 			} else {
2694 				sin6_t *sin6 = (sin6_t *)invalp;
2695 
2696 				if (sin6->sin6_family != AF_INET6)
2697 					return (EAFNOSUPPORT);
2698 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
2699 					return (EADDRNOTAVAIL);
2700 				ipp->ipp_nexthop = sin6->sin6_addr;
2701 				if (!IN6_IS_ADDR_UNSPECIFIED(
2702 				    &ipp->ipp_nexthop))
2703 					ipp->ipp_fields |= IPPF_NEXTHOP;
2704 				else
2705 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2706 			}
2707 			if (sticky) {
2708 				error = icmp_build_hdrs(icmp);
2709 				if (error != 0)
2710 					return (error);
2711 			}
2712 			break;
2713 		case IPV6_HOPOPTS: {
2714 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2715 			/*
2716 			 * Sanity checks - minimum size, size a multiple of
2717 			 * eight bytes, and matching size passed in.
2718 			 */
2719 			if (inlen != 0 &&
2720 			    inlen != (8 * (hopts->ip6h_len + 1)))
2721 				return (EINVAL);
2722 
2723 			if (checkonly)
2724 				break;
2725 			error = optcom_pkt_set(invalp, inlen, sticky,
2726 			    (uchar_t **)&ipp->ipp_hopopts,
2727 			    &ipp->ipp_hopoptslen,
2728 			    sticky ? icmp->icmp_label_len_v6 : 0);
2729 			if (error != 0)
2730 				return (error);
2731 			if (ipp->ipp_hopoptslen == 0) {
2732 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2733 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2734 			} else {
2735 				ipp->ipp_fields |= IPPF_HOPOPTS;
2736 			}
2737 			if (sticky) {
2738 				error = icmp_build_hdrs(icmp);
2739 				if (error != 0)
2740 					return (error);
2741 			}
2742 			break;
2743 		}
2744 		case IPV6_RTHDRDSTOPTS: {
2745 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2746 
2747 			/*
2748 			 * Sanity checks - minimum size, size a multiple of
2749 			 * eight bytes, and matching size passed in.
2750 			 */
2751 			if (inlen != 0 &&
2752 			    inlen != (8 * (dopts->ip6d_len + 1)))
2753 				return (EINVAL);
2754 
2755 			if (checkonly)
2756 				break;
2757 
2758 			if (inlen == 0) {
2759 				if (sticky &&
2760 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2761 					kmem_free(ipp->ipp_rtdstopts,
2762 					    ipp->ipp_rtdstoptslen);
2763 					ipp->ipp_rtdstopts = NULL;
2764 					ipp->ipp_rtdstoptslen = 0;
2765 				}
2766 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2767 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2768 			} else {
2769 				error = optcom_pkt_set(invalp, inlen, sticky,
2770 				    (uchar_t **)&ipp->ipp_rtdstopts,
2771 				    &ipp->ipp_rtdstoptslen, 0);
2772 				if (error != 0)
2773 					return (error);
2774 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2775 			}
2776 			if (sticky) {
2777 				error = icmp_build_hdrs(icmp);
2778 				if (error != 0)
2779 					return (error);
2780 			}
2781 			break;
2782 		}
2783 		case IPV6_DSTOPTS: {
2784 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2785 
2786 			/*
2787 			 * Sanity checks - minimum size, size a multiple of
2788 			 * eight bytes, and matching size passed in.
2789 			 */
2790 			if (inlen != 0 &&
2791 			    inlen != (8 * (dopts->ip6d_len + 1)))
2792 				return (EINVAL);
2793 
2794 			if (checkonly)
2795 				break;
2796 
2797 			if (inlen == 0) {
2798 				if (sticky &&
2799 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2800 					kmem_free(ipp->ipp_dstopts,
2801 					    ipp->ipp_dstoptslen);
2802 					ipp->ipp_dstopts = NULL;
2803 					ipp->ipp_dstoptslen = 0;
2804 				}
2805 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2806 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2807 			} else {
2808 				error = optcom_pkt_set(invalp, inlen, sticky,
2809 				    (uchar_t **)&ipp->ipp_dstopts,
2810 				    &ipp->ipp_dstoptslen, 0);
2811 				if (error != 0)
2812 					return (error);
2813 				ipp->ipp_fields |= IPPF_DSTOPTS;
2814 			}
2815 			if (sticky) {
2816 				error = icmp_build_hdrs(icmp);
2817 				if (error != 0)
2818 					return (error);
2819 			}
2820 			break;
2821 		}
2822 		case IPV6_RTHDR: {
2823 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2824 
2825 			/*
2826 			 * Sanity checks - minimum size, size a multiple of
2827 			 * eight bytes, and matching size passed in.
2828 			 */
2829 			if (inlen != 0 &&
2830 			    inlen != (8 * (rt->ip6r_len + 1)))
2831 				return (EINVAL);
2832 
2833 			if (checkonly)
2834 				break;
2835 
2836 			if (inlen == 0) {
2837 				if (sticky &&
2838 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2839 					kmem_free(ipp->ipp_rthdr,
2840 					    ipp->ipp_rthdrlen);
2841 					ipp->ipp_rthdr = NULL;
2842 					ipp->ipp_rthdrlen = 0;
2843 				}
2844 				ipp->ipp_fields &= ~IPPF_RTHDR;
2845 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2846 			} else {
2847 				error = optcom_pkt_set(invalp, inlen, sticky,
2848 				    (uchar_t **)&ipp->ipp_rthdr,
2849 				    &ipp->ipp_rthdrlen, 0);
2850 				if (error != 0)
2851 					return (error);
2852 				ipp->ipp_fields |= IPPF_RTHDR;
2853 			}
2854 			if (sticky) {
2855 				error = icmp_build_hdrs(icmp);
2856 				if (error != 0)
2857 					return (error);
2858 			}
2859 			break;
2860 		}
2861 
2862 		case IPV6_DONTFRAG:
2863 			if (checkonly)
2864 				break;
2865 
2866 			if (onoff) {
2867 				ipp->ipp_fields |= IPPF_DONTFRAG;
2868 			} else {
2869 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2870 			}
2871 			break;
2872 
2873 		case IPV6_USE_MIN_MTU:
2874 			if (inlen != sizeof (int))
2875 				return (EINVAL);
2876 
2877 			if (*i1 < -1 || *i1 > 1)
2878 				return (EINVAL);
2879 
2880 			if (checkonly)
2881 				break;
2882 
2883 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
2884 			ipp->ipp_use_min_mtu = *i1;
2885 			break;
2886 
2887 		/*
2888 		 * This option can't be set.  Its only returned via
2889 		 * getsockopt() or ancillary data.
2890 		 */
2891 		case IPV6_PATHMTU:
2892 			return (EINVAL);
2893 
2894 		case IPV6_BOUND_PIF:
2895 		case IPV6_SEC_OPT:
2896 		case IPV6_DONTFAILOVER_IF:
2897 		case IPV6_SRC_PREFERENCES:
2898 		case IPV6_V6ONLY:
2899 			/* Handled at IP level */
2900 			return (-EINVAL);
2901 		default:
2902 			*outlenp = 0;
2903 			return (EINVAL);
2904 		}
2905 		break;
2906 	}		/* end IPPROTO_IPV6 */
2907 
2908 	case IPPROTO_ICMPV6:
2909 		/*
2910 		 * Only allow IPv6 option processing on IPv6 sockets.
2911 		 */
2912 		if (icmp->icmp_family != AF_INET6) {
2913 			*outlenp = 0;
2914 			return (ENOPROTOOPT);
2915 		}
2916 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
2917 			*outlenp = 0;
2918 			return (ENOPROTOOPT);
2919 		}
2920 		switch (name) {
2921 		case ICMP6_FILTER:
2922 			if (!checkonly) {
2923 				if ((inlen != 0) &&
2924 				    (inlen != sizeof (icmp6_filter_t)))
2925 					return (EINVAL);
2926 
2927 				if (inlen == 0) {
2928 					if (icmp->icmp_filter != NULL) {
2929 						kmem_free(icmp->icmp_filter,
2930 						    sizeof (icmp6_filter_t));
2931 						icmp->icmp_filter = NULL;
2932 					}
2933 				} else {
2934 					if (icmp->icmp_filter == NULL) {
2935 						icmp->icmp_filter = kmem_alloc(
2936 						    sizeof (icmp6_filter_t),
2937 						    KM_NOSLEEP);
2938 						if (icmp->icmp_filter == NULL) {
2939 							*outlenp = 0;
2940 							return (ENOBUFS);
2941 						}
2942 					}
2943 					(void) bcopy(invalp, icmp->icmp_filter,
2944 					    inlen);
2945 				}
2946 			}
2947 			break;
2948 
2949 		default:
2950 			*outlenp = 0;
2951 			return (EINVAL);
2952 		}
2953 		break;
2954 	default:
2955 		*outlenp = 0;
2956 		return (EINVAL);
2957 	}
2958 	/*
2959 	 * Common case of OK return with outval same as inval.
2960 	 */
2961 	if (invalp != outvalp) {
2962 		/* don't trust bcopy for identical src/dst */
2963 		(void) bcopy(invalp, outvalp, inlen);
2964 	}
2965 	*outlenp = inlen;
2966 	return (0);
2967 }
2968 /* This routine sets socket options. */
2969 /* ARGSUSED */
2970 int
2971 icmp_opt_set(queue_t *q, uint_t optset_context, int level, int name,
2972     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
2973     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
2974 {
2975 	icmp_t	*icmp;
2976 	int	err;
2977 
2978 	icmp = Q_TO_ICMP(q);
2979 
2980 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2981 	err = icmp_opt_set_locked(q, optset_context, level, name, inlen, invalp,
2982 	    outlenp, outvalp, thisdg_attrs, cr, mblk);
2983 	rw_exit(&icmp->icmp_rwlock);
2984 	return (err);
2985 }
2986 
2987 /*
2988  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
2989  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
2990  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
2991  * headers.
2992  * Returns failure if can't allocate memory.
2993  */
2994 static int
2995 icmp_build_hdrs(icmp_t *icmp)
2996 {
2997 	icmp_stack_t *is = icmp->icmp_is;
2998 	uchar_t	*hdrs;
2999 	uint_t	hdrs_len;
3000 	ip6_t	*ip6h;
3001 	ip6i_t	*ip6i;
3002 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
3003 
3004 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
3005 	hdrs_len = ip_total_hdrs_len_v6(ipp);
3006 	ASSERT(hdrs_len != 0);
3007 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
3008 		/* Need to reallocate */
3009 		if (hdrs_len != 0) {
3010 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3011 			if (hdrs == NULL)
3012 				return (ENOMEM);
3013 		} else {
3014 			hdrs = NULL;
3015 		}
3016 		if (icmp->icmp_sticky_hdrs_len != 0) {
3017 			kmem_free(icmp->icmp_sticky_hdrs,
3018 			    icmp->icmp_sticky_hdrs_len);
3019 		}
3020 		icmp->icmp_sticky_hdrs = hdrs;
3021 		icmp->icmp_sticky_hdrs_len = hdrs_len;
3022 	}
3023 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
3024 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
3025 
3026 	/* Set header fields not in ipp */
3027 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3028 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
3029 		ip6h = (ip6_t *)&ip6i[1];
3030 
3031 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
3032 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
3033 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
3034 		}
3035 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
3036 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
3037 		}
3038 	} else {
3039 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
3040 	}
3041 
3042 	if (!(ipp->ipp_fields & IPPF_ADDR))
3043 		ip6h->ip6_src = icmp->icmp_v6src;
3044 
3045 	/* Try to get everything in a single mblk */
3046 	if (hdrs_len > icmp->icmp_max_hdr_len) {
3047 		icmp->icmp_max_hdr_len = hdrs_len;
3048 		rw_exit(&icmp->icmp_rwlock);
3049 		(void) mi_set_sth_wroff(icmp->icmp_connp->conn_rq,
3050 		    icmp->icmp_max_hdr_len + is->is_wroff_extra);
3051 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3052 	}
3053 	return (0);
3054 }
3055 
3056 /*
3057  * This routine retrieves the value of an ND variable in a icmpparam_t
3058  * structure.  It is called through nd_getset when a user reads the
3059  * variable.
3060  */
3061 /* ARGSUSED */
3062 static int
3063 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3064 {
3065 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3066 
3067 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
3068 	return (0);
3069 }
3070 
3071 /*
3072  * Walk through the param array specified registering each element with the
3073  * named dispatch (ND) handler.
3074  */
3075 static boolean_t
3076 icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt)
3077 {
3078 	for (; cnt-- > 0; icmppa++) {
3079 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
3080 			if (!nd_load(ndp, icmppa->icmp_param_name,
3081 			    icmp_param_get, icmp_param_set,
3082 			    (caddr_t)icmppa)) {
3083 				nd_free(ndp);
3084 				return (B_FALSE);
3085 			}
3086 		}
3087 	}
3088 	if (!nd_load(ndp, "icmp_status", icmp_status_report, NULL,
3089 	    NULL)) {
3090 		nd_free(ndp);
3091 		return (B_FALSE);
3092 	}
3093 	return (B_TRUE);
3094 }
3095 
3096 /* This routine sets an ND variable in a icmpparam_t structure. */
3097 /* ARGSUSED */
3098 static int
3099 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3100 {
3101 	long		new_value;
3102 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3103 
3104 	/*
3105 	 * Fail the request if the new value does not lie within the
3106 	 * required bounds.
3107 	 */
3108 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3109 	    new_value < icmppa->icmp_param_min ||
3110 	    new_value > icmppa->icmp_param_max) {
3111 		return (EINVAL);
3112 	}
3113 	/* Set the new value */
3114 	icmppa->icmp_param_value = new_value;
3115 	return (0);
3116 }
3117 /*ARGSUSED2*/
3118 static void
3119 icmp_input(void *arg1, mblk_t *mp, void *arg2)
3120 {
3121 	conn_t *connp = (conn_t *)arg1;
3122 	struct T_unitdata_ind	*tudi;
3123 	uchar_t			*rptr;
3124 	icmp_t			*icmp;
3125 	icmp_stack_t		*is;
3126 	sin_t			*sin;
3127 	sin6_t			*sin6;
3128 	ip6_t			*ip6h;
3129 	ip6i_t			*ip6i;
3130 	mblk_t			*mp1;
3131 	int			hdr_len;
3132 	ipha_t			*ipha;
3133 	int			udi_size;	/* Size of T_unitdata_ind */
3134 	uint_t			ipvers;
3135 	ip6_pkt_t		ipp;
3136 	uint8_t			nexthdr;
3137 	ip_pktinfo_t		*pinfo = NULL;
3138 	mblk_t			*options_mp = NULL;
3139 	uint_t			icmp_opt = 0;
3140 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
3141 	uint_t			hopstrip;
3142 
3143 	ASSERT(connp->conn_flags & IPCL_RAWIPCONN);
3144 
3145 	icmp = connp->conn_icmp;
3146 	is = icmp->icmp_is;
3147 	rptr = mp->b_rptr;
3148 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3149 	ASSERT(OK_32PTR(rptr));
3150 
3151 	/*
3152 	 * IP should have prepended the options data in an M_CTL
3153 	 * Check M_CTL "type" to make sure are not here bcos of
3154 	 * a valid ICMP message
3155 	 */
3156 	if (DB_TYPE(mp) == M_CTL) {
3157 		/*
3158 		 * FIXME: does IP still do this?
3159 		 * IP sends up the IPSEC_IN message for handling IPSEC
3160 		 * policy at the TCP level. We don't need it here.
3161 		 */
3162 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
3163 			mp1 = mp->b_cont;
3164 			freeb(mp);
3165 			mp = mp1;
3166 			rptr = mp->b_rptr;
3167 		} else if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3168 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3169 		    IN_PKTINFO) {
3170 			/*
3171 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3172 			 * has been prepended to the packet by IP. We need to
3173 			 * extract the mblk and adjust the rptr
3174 			 */
3175 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3176 			options_mp = mp;
3177 			mp = mp->b_cont;
3178 			rptr = mp->b_rptr;
3179 		} else {
3180 			/*
3181 			 * ICMP messages.
3182 			 */
3183 			icmp_icmp_error(connp->conn_rq, mp);
3184 			return;
3185 		}
3186 	}
3187 
3188 	/*
3189 	 * Discard message if it is misaligned or smaller than the IP header.
3190 	 */
3191 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3192 		freemsg(mp);
3193 		if (options_mp != NULL)
3194 			freeb(options_mp);
3195 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3196 		return;
3197 	}
3198 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3199 
3200 	/* Handle M_DATA messages containing IP packets messages */
3201 	if (ipvers == IPV4_VERSION) {
3202 		/*
3203 		 * Special case where IP attaches
3204 		 * the IRE needs to be handled so that we don't send up
3205 		 * IRE to the user land.
3206 		 */
3207 		ipha = (ipha_t *)rptr;
3208 		hdr_len = IPH_HDR_LENGTH(ipha);
3209 
3210 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3211 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3212 
3213 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3214 			    TH_SYN) && mp->b_cont != NULL) {
3215 				mp1 = mp->b_cont;
3216 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3217 					freeb(mp1);
3218 					mp->b_cont = NULL;
3219 				}
3220 			}
3221 		}
3222 		if (is->is_bsd_compat) {
3223 			ushort_t len;
3224 			len = ntohs(ipha->ipha_length);
3225 
3226 			if (mp->b_datap->db_ref > 1) {
3227 				/*
3228 				 * Allocate a new IP header so that we can
3229 				 * modify ipha_length.
3230 				 */
3231 				mblk_t	*mp1;
3232 
3233 				mp1 = allocb(hdr_len, BPRI_MED);
3234 				if (!mp1) {
3235 					freemsg(mp);
3236 					if (options_mp != NULL)
3237 						freeb(options_mp);
3238 					BUMP_MIB(&is->is_rawip_mib,
3239 					    rawipInErrors);
3240 					return;
3241 				}
3242 				bcopy(rptr, mp1->b_rptr, hdr_len);
3243 				mp->b_rptr = rptr + hdr_len;
3244 				rptr = mp1->b_rptr;
3245 				ipha = (ipha_t *)rptr;
3246 				mp1->b_cont = mp;
3247 				mp1->b_wptr = rptr + hdr_len;
3248 				mp = mp1;
3249 			}
3250 			len -= hdr_len;
3251 			ipha->ipha_length = htons(len);
3252 		}
3253 	}
3254 
3255 	/*
3256 	 * This is the inbound data path.  Packets are passed upstream as
3257 	 * T_UNITDATA_IND messages with full IP headers still attached.
3258 	 */
3259 	if (icmp->icmp_family == AF_INET) {
3260 		ASSERT(ipvers == IPV4_VERSION);
3261 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3262 		if (icmp->icmp_recvif && (pinfo != NULL) &&
3263 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3264 			udi_size += sizeof (struct T_opthdr) +
3265 			    sizeof (uint_t);
3266 		}
3267 
3268 		if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3269 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3270 			udi_size += sizeof (struct T_opthdr) +
3271 			    sizeof (struct in_pktinfo);
3272 		}
3273 
3274 		/*
3275 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3276 		 * buffer. Since gethrestime() expects a pointer aligned
3277 		 * argument, we allocate space necessary for extra
3278 		 * alignment (even though it might not be used).
3279 		 */
3280 		if (icmp->icmp_timestamp) {
3281 			udi_size += sizeof (struct T_opthdr) +
3282 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3283 		}
3284 		mp1 = allocb(udi_size, BPRI_MED);
3285 		if (mp1 == NULL) {
3286 			freemsg(mp);
3287 			if (options_mp != NULL)
3288 				freeb(options_mp);
3289 			BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3290 			return;
3291 		}
3292 		mp1->b_cont = mp;
3293 		mp = mp1;
3294 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3295 		mp->b_datap->db_type = M_PROTO;
3296 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3297 		tudi->PRIM_type = T_UNITDATA_IND;
3298 		tudi->SRC_length = sizeof (sin_t);
3299 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3300 		sin = (sin_t *)&tudi[1];
3301 		*sin = sin_null;
3302 		sin->sin_family = AF_INET;
3303 		sin->sin_addr.s_addr = ipha->ipha_src;
3304 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3305 		    sizeof (sin_t);
3306 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3307 		tudi->OPT_length = udi_size;
3308 
3309 		/*
3310 		 * Add options if IP_RECVIF is set
3311 		 */
3312 		if (udi_size != 0) {
3313 			char *dstopt;
3314 
3315 			dstopt = (char *)&sin[1];
3316 			if (icmp->icmp_recvif && (pinfo != NULL) &&
3317 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3318 
3319 				struct T_opthdr *toh;
3320 				uint_t		*dstptr;
3321 
3322 				toh = (struct T_opthdr *)dstopt;
3323 				toh->level = IPPROTO_IP;
3324 				toh->name = IP_RECVIF;
3325 				toh->len = sizeof (struct T_opthdr) +
3326 				    sizeof (uint_t);
3327 				toh->status = 0;
3328 				dstopt += sizeof (struct T_opthdr);
3329 				dstptr = (uint_t *)dstopt;
3330 				*dstptr = pinfo->ip_pkt_ifindex;
3331 				dstopt += sizeof (uint_t);
3332 				udi_size -= toh->len;
3333 			}
3334 			if (icmp->icmp_timestamp) {
3335 				struct	T_opthdr *toh;
3336 
3337 				toh = (struct T_opthdr *)dstopt;
3338 				toh->level = SOL_SOCKET;
3339 				toh->name = SCM_TIMESTAMP;
3340 				toh->len = sizeof (struct T_opthdr) +
3341 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3342 				toh->status = 0;
3343 				dstopt += sizeof (struct T_opthdr);
3344 				/* Align for gethrestime() */
3345 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3346 				    sizeof (intptr_t));
3347 				gethrestime((timestruc_t *)dstopt);
3348 				dstopt = (char *)toh + toh->len;
3349 				udi_size -= toh->len;
3350 			}
3351 			if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3352 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3353 				struct	T_opthdr *toh;
3354 				struct	in_pktinfo *pktinfop;
3355 
3356 				toh = (struct T_opthdr *)dstopt;
3357 				toh->level = IPPROTO_IP;
3358 				toh->name = IP_PKTINFO;
3359 				toh->len = sizeof (struct T_opthdr) +
3360 				    sizeof (in_pktinfo_t);
3361 				toh->status = 0;
3362 				dstopt += sizeof (struct T_opthdr);
3363 				pktinfop = (struct in_pktinfo *)dstopt;
3364 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3365 				pktinfop->ipi_spec_dst =
3366 				    pinfo->ip_pkt_match_addr;
3367 
3368 				pktinfop->ipi_addr.s_addr = ipha->ipha_dst;
3369 
3370 				dstopt += sizeof (struct in_pktinfo);
3371 				udi_size -= toh->len;
3372 			}
3373 
3374 			/* Consumed all of allocated space */
3375 			ASSERT(udi_size == 0);
3376 		}
3377 
3378 		if (options_mp != NULL)
3379 			freeb(options_mp);
3380 
3381 		BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3382 		putnext(connp->conn_rq, mp);
3383 		return;
3384 	}
3385 
3386 	/*
3387 	 * We don't need options_mp in the IPv6 path.
3388 	 */
3389 	if (options_mp != NULL) {
3390 		freeb(options_mp);
3391 		options_mp = NULL;
3392 	}
3393 
3394 	/*
3395 	 * Discard message if it is smaller than the IPv6 header
3396 	 * or if the header is malformed.
3397 	 */
3398 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3399 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3400 	    icmp->icmp_family != AF_INET6) {
3401 		freemsg(mp);
3402 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3403 		return;
3404 	}
3405 
3406 	/* Initialize */
3407 	ipp.ipp_fields = 0;
3408 	hopstrip = 0;
3409 
3410 	ip6h = (ip6_t *)rptr;
3411 	/*
3412 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3413 	 * as well as individual lenghts of ext hdrs (and ptrs to
3414 	 * them).
3415 	 */
3416 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3417 		/* Look for ifindex information */
3418 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3419 			ip6i = (ip6i_t *)ip6h;
3420 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3421 				ASSERT(ip6i->ip6i_ifindex != 0);
3422 				ipp.ipp_fields |= IPPF_IFINDEX;
3423 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3424 			}
3425 			rptr = (uchar_t *)&ip6i[1];
3426 			mp->b_rptr = rptr;
3427 			if (rptr == mp->b_wptr) {
3428 				mp1 = mp->b_cont;
3429 				freeb(mp);
3430 				mp = mp1;
3431 				rptr = mp->b_rptr;
3432 			}
3433 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3434 			ip6h = (ip6_t *)rptr;
3435 		}
3436 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3437 
3438 		/*
3439 		 * We need to lie a bit to the user because users inside
3440 		 * labeled compartments should not see their own labels.  We
3441 		 * assume that in all other respects IP has checked the label,
3442 		 * and that the label is always first among the options.  (If
3443 		 * it's not first, then this code won't see it, and the option
3444 		 * will be passed along to the user.)
3445 		 *
3446 		 * If we had multilevel ICMP sockets, then the following code
3447 		 * should be skipped for them to allow the user to see the
3448 		 * label.
3449 		 *
3450 		 * Alignment restrictions in the definition of IP options
3451 		 * (namely, the requirement that the 4-octet DOI goes on a
3452 		 * 4-octet boundary) mean that we know exactly where the option
3453 		 * should start, but we're lenient for other hosts.
3454 		 *
3455 		 * Note that there are no multilevel ICMP or raw IP sockets
3456 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3457 		 */
3458 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3459 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3460 			const uchar_t *ucp =
3461 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3462 			int remlen = ipp.ipp_hopoptslen - 2;
3463 
3464 			while (remlen > 0) {
3465 				if (*ucp == IP6OPT_PAD1) {
3466 					remlen--;
3467 					ucp++;
3468 				} else if (*ucp == IP6OPT_PADN) {
3469 					remlen -= ucp[1] + 2;
3470 					ucp += ucp[1] + 2;
3471 				} else if (*ucp == ip6opt_ls) {
3472 					hopstrip = (ucp -
3473 					    (const uchar_t *)ipp.ipp_hopopts) +
3474 					    ucp[1] + 2;
3475 					hopstrip = (hopstrip + 7) & ~7;
3476 					break;
3477 				} else {
3478 					/* label option must be first */
3479 					break;
3480 				}
3481 			}
3482 		}
3483 	} else {
3484 		hdr_len = IPV6_HDR_LEN;
3485 		ip6i = NULL;
3486 		nexthdr = ip6h->ip6_nxt;
3487 	}
3488 	/*
3489 	 * One special case where IP attaches the IRE needs to
3490 	 * be handled so that we don't send up IRE to the user land.
3491 	 */
3492 	if (nexthdr == IPPROTO_TCP) {
3493 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3494 
3495 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3496 		    mp->b_cont != NULL) {
3497 			mp1 = mp->b_cont;
3498 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3499 				freeb(mp1);
3500 				mp->b_cont = NULL;
3501 			}
3502 		}
3503 	}
3504 	/*
3505 	 * Check a filter for ICMPv6 types if needed.
3506 	 * Verify raw checksums if needed.
3507 	 */
3508 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3509 		if (icmp->icmp_filter != NULL) {
3510 			int type;
3511 
3512 			/* Assumes that IP has done the pullupmsg */
3513 			type = mp->b_rptr[hdr_len];
3514 
3515 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3516 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3517 				freemsg(mp);
3518 				return;
3519 			}
3520 		} else {
3521 			/* Checksum */
3522 			uint16_t	*up;
3523 			uint32_t	sum;
3524 			int		remlen;
3525 
3526 			up = (uint16_t *)&ip6h->ip6_src;
3527 
3528 			remlen = msgdsize(mp) - hdr_len;
3529 			sum = htons(icmp->icmp_proto + remlen)
3530 			    + up[0] + up[1] + up[2] + up[3]
3531 			    + up[4] + up[5] + up[6] + up[7]
3532 			    + up[8] + up[9] + up[10] + up[11]
3533 			    + up[12] + up[13] + up[14] + up[15];
3534 			sum = (sum & 0xffff) + (sum >> 16);
3535 			sum = IP_CSUM(mp, hdr_len, sum);
3536 			if (sum != 0) {
3537 				/* IPv6 RAW checksum failed */
3538 				ip0dbg(("icmp_rput: RAW checksum "
3539 				    "failed %x\n", sum));
3540 				freemsg(mp);
3541 				BUMP_MIB(&is->is_rawip_mib,
3542 				    rawipInCksumErrs);
3543 				return;
3544 			}
3545 		}
3546 	}
3547 	/* Skip all the IPv6 headers per API */
3548 	mp->b_rptr += hdr_len;
3549 
3550 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3551 
3552 	/*
3553 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3554 	 * maintain state information, instead of relying on icmp_t
3555 	 * structure, since there arent any locks protecting these members
3556 	 * and there is a window where there might be a race between a
3557 	 * thread setting options on the write side and a thread reading
3558 	 * these options on the read size.
3559 	 */
3560 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3561 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3562 		if (icmp->icmp_ipv6_recvhopopts &&
3563 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3564 		    ipp.ipp_hopoptslen > hopstrip) {
3565 			udi_size += sizeof (struct T_opthdr) +
3566 			    ipp.ipp_hopoptslen - hopstrip;
3567 			icmp_opt |= IPPF_HOPOPTS;
3568 		}
3569 		if ((icmp->icmp_ipv6_recvdstopts ||
3570 		    icmp->icmp_old_ipv6_recvdstopts) &&
3571 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3572 			udi_size += sizeof (struct T_opthdr) +
3573 			    ipp.ipp_dstoptslen;
3574 			icmp_opt |= IPPF_DSTOPTS;
3575 		}
3576 		if (((icmp->icmp_ipv6_recvdstopts &&
3577 		    icmp->icmp_ipv6_recvrthdr &&
3578 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3579 		    icmp->icmp_ipv6_recvrtdstopts) &&
3580 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3581 			udi_size += sizeof (struct T_opthdr) +
3582 			    ipp.ipp_rtdstoptslen;
3583 			icmp_opt |= IPPF_RTDSTOPTS;
3584 		}
3585 		if (icmp->icmp_ipv6_recvrthdr &&
3586 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3587 			udi_size += sizeof (struct T_opthdr) +
3588 			    ipp.ipp_rthdrlen;
3589 			icmp_opt |= IPPF_RTHDR;
3590 		}
3591 		if (icmp->icmp_ip_recvpktinfo &&
3592 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3593 			udi_size += sizeof (struct T_opthdr) +
3594 			    sizeof (struct in6_pktinfo);
3595 			icmp_opt |= IPPF_IFINDEX;
3596 		}
3597 	}
3598 	if (icmp->icmp_ipv6_recvhoplimit) {
3599 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3600 		icmp_ipv6_recvhoplimit = B_TRUE;
3601 	}
3602 
3603 	if (icmp->icmp_ipv6_recvtclass)
3604 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3605 
3606 	/*
3607 	 * If SO_TIMESTAMP is set allocate the appropriate sized
3608 	 * buffer. Since gethrestime() expects a pointer aligned
3609 	 * argument, we allocate space necessary for extra
3610 	 * alignment (even though it might not be used).
3611 	 */
3612 	if (icmp->icmp_timestamp) {
3613 		udi_size += sizeof (struct T_opthdr) +
3614 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3615 	}
3616 
3617 	mp1 = allocb(udi_size, BPRI_MED);
3618 	if (mp1 == NULL) {
3619 		freemsg(mp);
3620 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3621 		return;
3622 	}
3623 	mp1->b_cont = mp;
3624 	mp = mp1;
3625 	mp->b_datap->db_type = M_PROTO;
3626 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3627 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3628 	tudi->PRIM_type = T_UNITDATA_IND;
3629 	tudi->SRC_length = sizeof (sin6_t);
3630 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3631 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3632 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3633 	tudi->OPT_length = udi_size;
3634 	sin6 = (sin6_t *)&tudi[1];
3635 	sin6->sin6_port = 0;
3636 	sin6->sin6_family = AF_INET6;
3637 
3638 	sin6->sin6_addr = ip6h->ip6_src;
3639 	/* No sin6_flowinfo per API */
3640 	sin6->sin6_flowinfo = 0;
3641 	/* For link-scope source pass up scope id */
3642 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3643 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3644 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3645 	else
3646 		sin6->sin6_scope_id = 0;
3647 
3648 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3649 	    icmp->icmp_zoneid, is->is_netstack);
3650 
3651 	if (udi_size != 0) {
3652 		uchar_t *dstopt;
3653 
3654 		dstopt = (uchar_t *)&sin6[1];
3655 		if (icmp_opt & IPPF_IFINDEX) {
3656 			struct T_opthdr *toh;
3657 			struct in6_pktinfo *pkti;
3658 
3659 			toh = (struct T_opthdr *)dstopt;
3660 			toh->level = IPPROTO_IPV6;
3661 			toh->name = IPV6_PKTINFO;
3662 			toh->len = sizeof (struct T_opthdr) +
3663 			    sizeof (*pkti);
3664 			toh->status = 0;
3665 			dstopt += sizeof (struct T_opthdr);
3666 			pkti = (struct in6_pktinfo *)dstopt;
3667 			pkti->ipi6_addr = ip6h->ip6_dst;
3668 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3669 			dstopt += sizeof (*pkti);
3670 			udi_size -= toh->len;
3671 		}
3672 		if (icmp_ipv6_recvhoplimit) {
3673 			struct T_opthdr *toh;
3674 
3675 			toh = (struct T_opthdr *)dstopt;
3676 			toh->level = IPPROTO_IPV6;
3677 			toh->name = IPV6_HOPLIMIT;
3678 			toh->len = sizeof (struct T_opthdr) +
3679 			    sizeof (uint_t);
3680 			toh->status = 0;
3681 			dstopt += sizeof (struct T_opthdr);
3682 			*(uint_t *)dstopt = ip6h->ip6_hops;
3683 			dstopt += sizeof (uint_t);
3684 			udi_size -= toh->len;
3685 		}
3686 		if (icmp->icmp_ipv6_recvtclass) {
3687 			struct T_opthdr *toh;
3688 
3689 			toh = (struct T_opthdr *)dstopt;
3690 			toh->level = IPPROTO_IPV6;
3691 			toh->name = IPV6_TCLASS;
3692 			toh->len = sizeof (struct T_opthdr) +
3693 			    sizeof (uint_t);
3694 			toh->status = 0;
3695 			dstopt += sizeof (struct T_opthdr);
3696 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3697 			dstopt += sizeof (uint_t);
3698 			udi_size -= toh->len;
3699 		}
3700 		if (icmp->icmp_timestamp) {
3701 			struct	T_opthdr *toh;
3702 
3703 			toh = (struct T_opthdr *)dstopt;
3704 			toh->level = SOL_SOCKET;
3705 			toh->name = SCM_TIMESTAMP;
3706 			toh->len = sizeof (struct T_opthdr) +
3707 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3708 			toh->status = 0;
3709 			dstopt += sizeof (struct T_opthdr);
3710 			/* Align for gethrestime() */
3711 			dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
3712 			    sizeof (intptr_t));
3713 			gethrestime((timestruc_t *)dstopt);
3714 			dstopt = (uchar_t *)toh + toh->len;
3715 			udi_size -= toh->len;
3716 		}
3717 		if (icmp_opt & IPPF_HOPOPTS) {
3718 			struct T_opthdr *toh;
3719 
3720 			toh = (struct T_opthdr *)dstopt;
3721 			toh->level = IPPROTO_IPV6;
3722 			toh->name = IPV6_HOPOPTS;
3723 			toh->len = sizeof (struct T_opthdr) +
3724 			    ipp.ipp_hopoptslen - hopstrip;
3725 			toh->status = 0;
3726 			dstopt += sizeof (struct T_opthdr);
3727 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
3728 			    ipp.ipp_hopoptslen - hopstrip);
3729 			if (hopstrip > 0) {
3730 				/* copy next header value and fake length */
3731 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
3732 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
3733 				    hopstrip / 8;
3734 			}
3735 			dstopt += ipp.ipp_hopoptslen - hopstrip;
3736 			udi_size -= toh->len;
3737 		}
3738 		if (icmp_opt & IPPF_RTDSTOPTS) {
3739 			struct T_opthdr *toh;
3740 
3741 			toh = (struct T_opthdr *)dstopt;
3742 			toh->level = IPPROTO_IPV6;
3743 			toh->name = IPV6_DSTOPTS;
3744 			toh->len = sizeof (struct T_opthdr) +
3745 			    ipp.ipp_rtdstoptslen;
3746 			toh->status = 0;
3747 			dstopt += sizeof (struct T_opthdr);
3748 			bcopy(ipp.ipp_rtdstopts, dstopt,
3749 			    ipp.ipp_rtdstoptslen);
3750 			dstopt += ipp.ipp_rtdstoptslen;
3751 			udi_size -= toh->len;
3752 		}
3753 		if (icmp_opt & IPPF_RTHDR) {
3754 			struct T_opthdr *toh;
3755 
3756 			toh = (struct T_opthdr *)dstopt;
3757 			toh->level = IPPROTO_IPV6;
3758 			toh->name = IPV6_RTHDR;
3759 			toh->len = sizeof (struct T_opthdr) +
3760 			    ipp.ipp_rthdrlen;
3761 			toh->status = 0;
3762 			dstopt += sizeof (struct T_opthdr);
3763 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
3764 			dstopt += ipp.ipp_rthdrlen;
3765 			udi_size -= toh->len;
3766 		}
3767 		if (icmp_opt & IPPF_DSTOPTS) {
3768 			struct T_opthdr *toh;
3769 
3770 			toh = (struct T_opthdr *)dstopt;
3771 			toh->level = IPPROTO_IPV6;
3772 			toh->name = IPV6_DSTOPTS;
3773 			toh->len = sizeof (struct T_opthdr) +
3774 			    ipp.ipp_dstoptslen;
3775 			toh->status = 0;
3776 			dstopt += sizeof (struct T_opthdr);
3777 			bcopy(ipp.ipp_dstopts, dstopt,
3778 			    ipp.ipp_dstoptslen);
3779 			dstopt += ipp.ipp_dstoptslen;
3780 			udi_size -= toh->len;
3781 		}
3782 		/* Consumed all of allocated space */
3783 		ASSERT(udi_size == 0);
3784 	}
3785 	BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3786 	putnext(connp->conn_rq, mp);
3787 }
3788 
3789 /*
3790  * Handle the results of a T_BIND_REQ whether deferred by IP or handled
3791  * immediately.
3792  */
3793 static void
3794 icmp_bind_result(conn_t *connp, mblk_t *mp)
3795 {
3796 	struct T_error_ack	*tea;
3797 
3798 	switch (mp->b_datap->db_type) {
3799 	case M_PROTO:
3800 	case M_PCPROTO:
3801 		/* M_PROTO messages contain some type of TPI message. */
3802 		if ((mp->b_wptr - mp->b_rptr) < sizeof (t_scalar_t)) {
3803 			freemsg(mp);
3804 			return;
3805 		}
3806 		tea = (struct T_error_ack *)mp->b_rptr;
3807 
3808 		switch (tea->PRIM_type) {
3809 		case T_ERROR_ACK:
3810 			switch (tea->ERROR_prim) {
3811 			case O_T_BIND_REQ:
3812 			case T_BIND_REQ:
3813 				icmp_bind_error(connp, mp);
3814 				return;
3815 			default:
3816 				break;
3817 			}
3818 			ASSERT(0);
3819 			freemsg(mp);
3820 			return;
3821 
3822 		case T_BIND_ACK:
3823 			icmp_bind_ack(connp, mp);
3824 			return;
3825 
3826 		default:
3827 			break;
3828 		}
3829 		freemsg(mp);
3830 		return;
3831 	default:
3832 		/* FIXME: other cases? */
3833 		ASSERT(0);
3834 		freemsg(mp);
3835 		return;
3836 	}
3837 }
3838 
3839 /*
3840  * Process a T_BIND_ACK
3841  */
3842 static void
3843 icmp_bind_ack(conn_t *connp, mblk_t *mp)
3844 {
3845 	icmp_t	*icmp = connp->conn_icmp;
3846 	mblk_t	*mp1;
3847 	ire_t	*ire;
3848 	struct T_bind_ack *tba;
3849 	uchar_t *addrp;
3850 	ipa_conn_t	*ac;
3851 	ipa6_conn_t	*ac6;
3852 
3853 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3854 	/*
3855 	 * We know if headers are included or not so we can
3856 	 * safely do this.
3857 	 */
3858 	if (icmp->icmp_state == TS_UNBND) {
3859 		/*
3860 		 * TPI has not yet bound - bind sent by
3861 		 * icmp_bind_proto.
3862 		 */
3863 		freemsg(mp);
3864 		rw_exit(&icmp->icmp_rwlock);
3865 		return;
3866 	}
3867 	ASSERT(icmp->icmp_pending_op != -1);
3868 
3869 	/*
3870 	 * If a broadcast/multicast address was bound set
3871 	 * the source address to 0.
3872 	 * This ensures no datagrams with broadcast address
3873 	 * as source address are emitted (which would violate
3874 	 * RFC1122 - Hosts requirements)
3875 	 *
3876 	 * Note that when connecting the returned IRE is
3877 	 * for the destination address and we only perform
3878 	 * the broadcast check for the source address (it
3879 	 * is OK to connect to a broadcast/multicast address.)
3880 	 */
3881 	mp1 = mp->b_cont;
3882 	if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) {
3883 		ire = (ire_t *)mp1->b_rptr;
3884 
3885 		/*
3886 		 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast
3887 		 * local address.
3888 		 */
3889 		if (ire->ire_type == IRE_BROADCAST &&
3890 		    icmp->icmp_state != TS_DATA_XFER) {
3891 			ASSERT(icmp->icmp_pending_op == T_BIND_REQ ||
3892 			    icmp->icmp_pending_op == O_T_BIND_REQ);
3893 			/* This was just a local bind to a MC/broadcast addr */
3894 			V6_SET_ZERO(icmp->icmp_v6src);
3895 			if (icmp->icmp_family == AF_INET6)
3896 				(void) icmp_build_hdrs(icmp);
3897 		} else if (V6_OR_V4_INADDR_ANY(icmp->icmp_v6src)) {
3898 			/*
3899 			 * Local address not yet set - pick it from the
3900 			 * T_bind_ack
3901 			 */
3902 			tba = (struct T_bind_ack *)mp->b_rptr;
3903 			addrp = &mp->b_rptr[tba->ADDR_offset];
3904 			switch (icmp->icmp_family) {
3905 			case AF_INET:
3906 				if (tba->ADDR_length == sizeof (ipa_conn_t)) {
3907 					ac = (ipa_conn_t *)addrp;
3908 				} else {
3909 					ASSERT(tba->ADDR_length ==
3910 					    sizeof (ipa_conn_x_t));
3911 					ac = &((ipa_conn_x_t *)addrp)->acx_conn;
3912 				}
3913 				IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr,
3914 				    &icmp->icmp_v6src);
3915 				break;
3916 			case AF_INET6:
3917 				if (tba->ADDR_length == sizeof (ipa6_conn_t)) {
3918 					ac6 = (ipa6_conn_t *)addrp;
3919 				} else {
3920 					ASSERT(tba->ADDR_length ==
3921 					    sizeof (ipa6_conn_x_t));
3922 					ac6 = &((ipa6_conn_x_t *)
3923 					    addrp)->ac6x_conn;
3924 				}
3925 				icmp->icmp_v6src = ac6->ac6_laddr;
3926 				(void) icmp_build_hdrs(icmp);
3927 			}
3928 		}
3929 		mp1 = mp1->b_cont;
3930 	}
3931 	icmp->icmp_pending_op = -1;
3932 	rw_exit(&icmp->icmp_rwlock);
3933 	/*
3934 	 * Look for one or more appended ACK message added by
3935 	 * icmp_connect or icmp_disconnect.
3936 	 * If none found just send up the T_BIND_ACK.
3937 	 * icmp_connect has appended a T_OK_ACK and a
3938 	 * T_CONN_CON.
3939 	 * icmp_disconnect has appended a T_OK_ACK.
3940 	 */
3941 	if (mp1 != NULL) {
3942 		if (mp->b_cont == mp1)
3943 			mp->b_cont = NULL;
3944 		else {
3945 			ASSERT(mp->b_cont->b_cont == mp1);
3946 			mp->b_cont->b_cont = NULL;
3947 		}
3948 		freemsg(mp);
3949 		mp = mp1;
3950 		while (mp != NULL) {
3951 			mp1 = mp->b_cont;
3952 			mp->b_cont = NULL;
3953 			putnext(connp->conn_rq, mp);
3954 			mp = mp1;
3955 		}
3956 		return;
3957 	}
3958 	freemsg(mp->b_cont);
3959 	mp->b_cont = NULL;
3960 	putnext(connp->conn_rq, mp);
3961 }
3962 
3963 static void
3964 icmp_bind_error(conn_t *connp, mblk_t *mp)
3965 {
3966 	icmp_t	*icmp = connp->conn_icmp;
3967 	struct T_error_ack *tea;
3968 
3969 	tea = (struct T_error_ack *)mp->b_rptr;
3970 	/*
3971 	 * If our O_T_BIND_REQ/T_BIND_REQ fails,
3972 	 * clear out the source address before
3973 	 * passing the message upstream.
3974 	 * If this was caused by a T_CONN_REQ
3975 	 * revert back to bound state.
3976 	 */
3977 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3978 	if (icmp->icmp_state == TS_UNBND) {
3979 		/*
3980 		 * TPI has not yet bound - bind sent by icmp_bind_proto.
3981 		 */
3982 		freemsg(mp);
3983 		rw_exit(&icmp->icmp_rwlock);
3984 		return;
3985 	}
3986 	ASSERT(icmp->icmp_pending_op != -1);
3987 	tea->ERROR_prim = icmp->icmp_pending_op;
3988 	icmp->icmp_pending_op = -1;
3989 
3990 	switch (tea->ERROR_prim) {
3991 	case T_CONN_REQ:
3992 		ASSERT(icmp->icmp_state == TS_DATA_XFER);
3993 		/* Connect failed */
3994 		/* Revert back to the bound source */
3995 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
3996 		icmp->icmp_state = TS_IDLE;
3997 		if (icmp->icmp_family == AF_INET6)
3998 			(void) icmp_build_hdrs(icmp);
3999 		break;
4000 
4001 	case T_DISCON_REQ:
4002 	case T_BIND_REQ:
4003 	case O_T_BIND_REQ:
4004 		V6_SET_ZERO(icmp->icmp_v6src);
4005 		V6_SET_ZERO(icmp->icmp_bound_v6src);
4006 		icmp->icmp_state = TS_UNBND;
4007 		if (icmp->icmp_family == AF_INET6)
4008 			(void) icmp_build_hdrs(icmp);
4009 		break;
4010 	default:
4011 		break;
4012 	}
4013 	rw_exit(&icmp->icmp_rwlock);
4014 	putnext(connp->conn_rq, mp);
4015 }
4016 
4017 /*
4018  * return SNMP stuff in buffer in mpdata
4019  */
4020 mblk_t *
4021 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
4022 {
4023 	mblk_t			*mpdata;
4024 	struct opthdr		*optp;
4025 	conn_t			*connp = Q_TO_CONN(q);
4026 	icmp_stack_t		*is = connp->conn_netstack->netstack_icmp;
4027 	mblk_t			*mp2ctl;
4028 
4029 	/*
4030 	 * make a copy of the original message
4031 	 */
4032 	mp2ctl = copymsg(mpctl);
4033 
4034 	if (mpctl == NULL ||
4035 	    (mpdata = mpctl->b_cont) == NULL) {
4036 		freemsg(mpctl);
4037 		freemsg(mp2ctl);
4038 		return (0);
4039 	}
4040 
4041 	/* fixed length structure for IPv4 and IPv6 counters */
4042 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4043 	optp->level = EXPER_RAWIP;
4044 	optp->name = 0;
4045 	(void) snmp_append_data(mpdata, (char *)&is->is_rawip_mib,
4046 	    sizeof (is->is_rawip_mib));
4047 	optp->len = msgdsize(mpdata);
4048 	qreply(q, mpctl);
4049 
4050 	return (mp2ctl);
4051 }
4052 
4053 /*
4054  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
4055  * TODO:  If this ever actually tries to set anything, it needs to be
4056  * to do the appropriate locking.
4057  */
4058 /* ARGSUSED */
4059 int
4060 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4061     uchar_t *ptr, int len)
4062 {
4063 	switch (level) {
4064 	case EXPER_RAWIP:
4065 		return (0);
4066 	default:
4067 		return (1);
4068 	}
4069 }
4070 
4071 /* Report for ndd "icmp_status" */
4072 /* ARGSUSED */
4073 static int
4074 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4075 {
4076 	conn_t  *connp;
4077 	ip_stack_t *ipst;
4078 	char	laddrbuf[INET6_ADDRSTRLEN];
4079 	char	faddrbuf[INET6_ADDRSTRLEN];
4080 	int	i;
4081 
4082 	(void) mi_mpprintf(mp,
4083 	    "RAWIP    " MI_COL_HDRPAD_STR
4084 	/*   01234567[89ABCDEF] */
4085 	    "  src addr        dest addr       state");
4086 	/*   xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */
4087 
4088 	connp = Q_TO_CONN(q);
4089 	ipst = connp->conn_netstack->netstack_ip;
4090 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
4091 		connf_t *connfp;
4092 		char	*state;
4093 
4094 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
4095 		connp = NULL;
4096 
4097 		while ((connp = ipcl_get_next_conn(connfp, connp,
4098 		    IPCL_RAWIPCONN)) != NULL) {
4099 			icmp_t  *icmp;
4100 
4101 			mutex_enter(&(connp)->conn_lock);
4102 			icmp = connp->conn_icmp;
4103 
4104 			if (icmp->icmp_state == TS_UNBND)
4105 				state = "UNBOUND";
4106 			else if (icmp->icmp_state == TS_IDLE)
4107 				state = "IDLE";
4108 			else if (icmp->icmp_state == TS_DATA_XFER)
4109 				state = "CONNECTED";
4110 			else
4111 				state = "UnkState";
4112 
4113 			(void) mi_mpprintf(mp, MI_COL_PTRFMT_STR "%s %s %s",
4114 			    (void *)icmp,
4115 			    inet_ntop(AF_INET6, &icmp->icmp_v6dst, faddrbuf,
4116 			    sizeof (faddrbuf)),
4117 			    inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf,
4118 			    sizeof (laddrbuf)),
4119 			    state);
4120 			mutex_exit(&(connp)->conn_lock);
4121 		}
4122 	}
4123 	return (0);
4124 }
4125 
4126 /*
4127  * This routine creates a T_UDERROR_IND message and passes it upstream.
4128  * The address and options are copied from the T_UNITDATA_REQ message
4129  * passed in mp.  This message is freed.
4130  */
4131 static void
4132 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
4133 {
4134 	mblk_t	*mp1;
4135 	uchar_t	*rptr = mp->b_rptr;
4136 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
4137 
4138 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
4139 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
4140 	    tudr->OPT_length, err);
4141 	if (mp1)
4142 		qreply(q, mp1);
4143 	freemsg(mp);
4144 }
4145 
4146 /*
4147  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
4148  * After some error checking, the message is passed downstream to ip.
4149  */
4150 static void
4151 icmp_unbind(queue_t *q, mblk_t *mp)
4152 {
4153 	icmp_t	*icmp = Q_TO_ICMP(q);
4154 
4155 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4156 	/* If a bind has not been done, we can't unbind. */
4157 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
4158 		rw_exit(&icmp->icmp_rwlock);
4159 		icmp_err_ack(q, mp, TOUTSTATE, 0);
4160 		return;
4161 	}
4162 	icmp->icmp_pending_op = T_UNBIND_REQ;
4163 	rw_exit(&icmp->icmp_rwlock);
4164 
4165 	/*
4166 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
4167 	 * and therefore ip_unbind must never return NULL.
4168 	 */
4169 	mp = ip_unbind(q, mp);
4170 	ASSERT(mp != NULL);
4171 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4172 
4173 	/*
4174 	 * Once we're unbound from IP, the pending operation may be cleared
4175 	 * here.
4176 	 */
4177 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4178 	V6_SET_ZERO(icmp->icmp_v6src);
4179 	V6_SET_ZERO(icmp->icmp_bound_v6src);
4180 	icmp->icmp_pending_op = -1;
4181 	icmp->icmp_state = TS_UNBND;
4182 	if (icmp->icmp_family == AF_INET6)
4183 		(void) icmp_build_hdrs(icmp);
4184 	rw_exit(&icmp->icmp_rwlock);
4185 
4186 	qreply(q, mp);
4187 }
4188 
4189 /*
4190  * Process IPv4 packets that already include an IP header.
4191  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
4192  * IPPROTO_IGMP).
4193  */
4194 static void
4195 icmp_wput_hdrincl(queue_t *q, mblk_t *mp, icmp_t *icmp, ip4_pkt_t *pktinfop)
4196 {
4197 	icmp_stack_t *is = icmp->icmp_is;
4198 	ipha_t	*ipha;
4199 	int	ip_hdr_length;
4200 	int	tp_hdr_len;
4201 	mblk_t	*mp1;
4202 	uint_t	pkt_len;
4203 	ip_opt_info_t optinfo;
4204 	conn_t	*connp = icmp->icmp_connp;
4205 
4206 	optinfo.ip_opt_flags = 0;
4207 	optinfo.ip_opt_ill_index = 0;
4208 	ipha = (ipha_t *)mp->b_rptr;
4209 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4210 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
4211 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4212 			ASSERT(icmp != NULL);
4213 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4214 			freemsg(mp);
4215 			return;
4216 		}
4217 		ipha = (ipha_t *)mp->b_rptr;
4218 	}
4219 	ipha->ipha_version_and_hdr_length =
4220 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
4221 
4222 	/*
4223 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4224 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4225 	 * tell IP that the application has sent a complete IP header and not
4226 	 * to compute the transport checksum nor change the DF flag.
4227 	 */
4228 	ipha->ipha_ident = IP_HDR_INCLUDED;
4229 	ipha->ipha_hdr_checksum = 0;
4230 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
4231 	/* Insert options if any */
4232 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4233 		/*
4234 		 * Put the IP header plus any transport header that is
4235 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
4236 		 * that at least the checksum field is in the first mblk.)
4237 		 */
4238 		switch (ipha->ipha_protocol) {
4239 		case IPPROTO_UDP:
4240 			tp_hdr_len = 8;
4241 			break;
4242 		case IPPROTO_TCP:
4243 			tp_hdr_len = 20;
4244 			break;
4245 		default:
4246 			tp_hdr_len = 0;
4247 			break;
4248 		}
4249 		/*
4250 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
4251 		 * tp_hdr_len bytes will be in a single mblk.
4252 		 */
4253 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
4254 		    tp_hdr_len)) {
4255 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
4256 			    tp_hdr_len)) {
4257 				BUMP_MIB(&is->is_rawip_mib,
4258 				    rawipOutErrors);
4259 				freemsg(mp);
4260 				return;
4261 			}
4262 			ipha = (ipha_t *)mp->b_rptr;
4263 		}
4264 
4265 		/*
4266 		 * if the length is larger then the max allowed IP packet,
4267 		 * then send an error and abort the processing.
4268 		 */
4269 		pkt_len = ntohs(ipha->ipha_length)
4270 		    + icmp->icmp_ip_snd_options_len;
4271 		if (pkt_len > IP_MAXPACKET) {
4272 			icmp_ud_err(q, mp, EMSGSIZE);
4273 			return;
4274 		}
4275 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra +
4276 		    tp_hdr_len, BPRI_LO))) {
4277 			icmp_ud_err(q, mp, ENOMEM);
4278 			return;
4279 		}
4280 		mp1->b_rptr += is->is_wroff_extra;
4281 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
4282 
4283 		ipha->ipha_length = htons((uint16_t)pkt_len);
4284 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
4285 
4286 		/* Copy transport header if any */
4287 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
4288 		mp1->b_wptr += tp_hdr_len;
4289 
4290 		/* Add options */
4291 		ipha = (ipha_t *)mp1->b_rptr;
4292 		bcopy(icmp->icmp_ip_snd_options, &ipha[1],
4293 		    icmp->icmp_ip_snd_options_len);
4294 
4295 		/* Drop IP header and transport header from original */
4296 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
4297 
4298 		mp1->b_cont = mp;
4299 		mp = mp1;
4300 		/*
4301 		 * Massage source route putting first source
4302 		 * route in ipha_dst.
4303 		 */
4304 		(void) ip_massage_options(ipha, is->is_netstack);
4305 	}
4306 
4307 	if (pktinfop != NULL) {
4308 		/*
4309 		 * Over write the source address provided in the header
4310 		 */
4311 		if (pktinfop->ip4_addr != INADDR_ANY) {
4312 			ipha->ipha_src = pktinfop->ip4_addr;
4313 			optinfo.ip_opt_flags = IP_VERIFY_SRC;
4314 		}
4315 
4316 		if (pktinfop->ip4_ill_index != 0) {
4317 			optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4318 		}
4319 	}
4320 
4321 	mblk_setcred(mp, connp->conn_cred);
4322 	ip_output_options(connp, mp, q, IP_WPUT,
4323 	    &optinfo);
4324 }
4325 
4326 static boolean_t
4327 icmp_update_label(queue_t *q, icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
4328 {
4329 	int err;
4330 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4331 	icmp_stack_t		*is = icmp->icmp_is;
4332 	conn_t	*connp = icmp->icmp_connp;
4333 
4334 	err = tsol_compute_label(DB_CREDDEF(mp, connp->conn_cred), dst,
4335 	    opt_storage, icmp->icmp_mac_exempt,
4336 	    is->is_netstack->netstack_ip);
4337 	if (err == 0) {
4338 		err = tsol_update_options(&icmp->icmp_ip_snd_options,
4339 		    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
4340 		    opt_storage);
4341 	}
4342 	if (err != 0) {
4343 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4344 		DTRACE_PROBE4(
4345 		    tx__ip__log__drop__updatelabel__icmp,
4346 		    char *, "queue(1) failed to update options(2) on mp(3)",
4347 		    queue_t *, q, char *, opt_storage, mblk_t *, mp);
4348 		icmp_ud_err(q, mp, err);
4349 		return (B_FALSE);
4350 	}
4351 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
4352 	return (B_TRUE);
4353 }
4354 
4355 /*
4356  * This routine handles all messages passed downstream.  It either
4357  * consumes the message or passes it downstream; it never queues a
4358  * a message.
4359  */
4360 static void
4361 icmp_wput(queue_t *q, mblk_t *mp)
4362 {
4363 	uchar_t	*rptr = mp->b_rptr;
4364 	ipha_t	*ipha;
4365 	mblk_t	*mp1;
4366 	int	ip_hdr_length;
4367 #define	tudr ((struct T_unitdata_req *)rptr)
4368 	size_t	ip_len;
4369 	conn_t	*connp = Q_TO_CONN(q);
4370 	icmp_t	*icmp = connp->conn_icmp;
4371 	icmp_stack_t *is = icmp->icmp_is;
4372 	sin6_t	*sin6;
4373 	sin_t	*sin;
4374 	ipaddr_t	v4dst;
4375 	ip4_pkt_t	pktinfo;
4376 	ip4_pkt_t	*pktinfop = &pktinfo;
4377 	ip_opt_info_t	optinfo;
4378 
4379 	switch (mp->b_datap->db_type) {
4380 	case M_DATA:
4381 		if (icmp->icmp_hdrincl) {
4382 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
4383 			ipha = (ipha_t *)mp->b_rptr;
4384 			if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH) {
4385 				if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4386 					BUMP_MIB(&is->is_rawip_mib,
4387 					    rawipOutErrors);
4388 					freemsg(mp);
4389 					return;
4390 				}
4391 				ipha = (ipha_t *)mp->b_rptr;
4392 			}
4393 			/*
4394 			 * If this connection was used for v6 (inconceivable!)
4395 			 * or if we have a new destination, then it's time to
4396 			 * figure a new label.
4397 			 */
4398 			if (is_system_labeled() &&
4399 			    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4400 			    V4_PART_OF_V6(icmp->icmp_v6lastdst) !=
4401 			    ipha->ipha_dst) &&
4402 			    !icmp_update_label(q, icmp, mp, ipha->ipha_dst)) {
4403 				return;
4404 			}
4405 			icmp_wput_hdrincl(q, mp, icmp, NULL);
4406 			return;
4407 		}
4408 		freemsg(mp);
4409 		return;
4410 	case M_PROTO:
4411 	case M_PCPROTO:
4412 		ip_len = mp->b_wptr - rptr;
4413 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4414 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4415 			if (((union T_primitives *)rptr)->type
4416 			    == T_UNITDATA_REQ)
4417 				break;
4418 		}
4419 		/* FALLTHRU */
4420 	default:
4421 		icmp_wput_other(q, mp);
4422 		return;
4423 	}
4424 
4425 	/* Handle T_UNITDATA_REQ messages here. */
4426 
4427 
4428 
4429 	if (icmp->icmp_state == TS_UNBND) {
4430 		/* If a port has not been bound to the stream, fail. */
4431 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4432 		icmp_ud_err(q, mp, EPROTO);
4433 		return;
4434 	}
4435 	mp1 = mp->b_cont;
4436 	if (mp1 == NULL) {
4437 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4438 		icmp_ud_err(q, mp, EPROTO);
4439 		return;
4440 	}
4441 
4442 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4443 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4444 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4445 		return;
4446 	}
4447 
4448 	switch (icmp->icmp_family) {
4449 	case AF_INET6:
4450 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4451 		if (!OK_32PTR((char *)sin6) ||
4452 		    tudr->DEST_length != sizeof (sin6_t) ||
4453 		    sin6->sin6_family != AF_INET6) {
4454 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4455 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4456 			return;
4457 		}
4458 
4459 		/* No support for mapped addresses on raw sockets */
4460 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4461 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4462 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4463 			return;
4464 		}
4465 
4466 		/*
4467 		 * Destination is a native IPv6 address.
4468 		 * Send out an IPv6 format packet.
4469 		 */
4470 		icmp_wput_ipv6(q, mp, sin6, tudr->OPT_length);
4471 		return;
4472 
4473 	case AF_INET:
4474 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4475 		if (!OK_32PTR((char *)sin) ||
4476 		    tudr->DEST_length != sizeof (sin_t) ||
4477 		    sin->sin_family != AF_INET) {
4478 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4479 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4480 			return;
4481 		}
4482 		/* Extract and ipaddr */
4483 		v4dst = sin->sin_addr.s_addr;
4484 		break;
4485 
4486 	default:
4487 		ASSERT(0);
4488 	}
4489 
4490 	pktinfop->ip4_ill_index = 0;
4491 	pktinfop->ip4_addr = INADDR_ANY;
4492 	optinfo.ip_opt_flags = 0;
4493 	optinfo.ip_opt_ill_index = 0;
4494 
4495 
4496 	/*
4497 	 * If options passed in, feed it for verification and handling
4498 	 */
4499 	if (tudr->OPT_length != 0) {
4500 		int error;
4501 
4502 		error = 0;
4503 		if (icmp_unitdata_opt_process(q, mp, &error,
4504 		    (void *)pktinfop) < 0) {
4505 			/* failure */
4506 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4507 			icmp_ud_err(q, mp, error);
4508 			return;
4509 		}
4510 		ASSERT(error == 0);
4511 		/*
4512 		 * Note: Success in processing options.
4513 		 * mp option buffer represented by
4514 		 * OPT_length/offset now potentially modified
4515 		 * and contain option setting results
4516 		 */
4517 
4518 	}
4519 
4520 	if (v4dst == INADDR_ANY)
4521 		v4dst = htonl(INADDR_LOOPBACK);
4522 
4523 	/* Check if our saved options are valid; update if not */
4524 	if (is_system_labeled() &&
4525 	    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4526 	    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst) &&
4527 	    !icmp_update_label(q, icmp, mp, v4dst)) {
4528 		return;
4529 	}
4530 
4531 	/* Protocol 255 contains full IP headers */
4532 	if (icmp->icmp_hdrincl) {
4533 		freeb(mp);
4534 		icmp_wput_hdrincl(q, mp1, icmp, pktinfop);
4535 		return;
4536 	}
4537 
4538 
4539 	/* Add an IP header */
4540 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4541 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
4542 	if ((uchar_t *)ipha < mp1->b_datap->db_base ||
4543 	    mp1->b_datap->db_ref != 1 ||
4544 	    !OK_32PTR(ipha)) {
4545 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra,
4546 		    BPRI_LO))) {
4547 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4548 			icmp_ud_err(q, mp, ENOMEM);
4549 			return;
4550 		}
4551 		mp1->b_cont = mp->b_cont;
4552 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4553 		mp1->b_wptr = (uchar_t *)ipha;
4554 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4555 	}
4556 #ifdef	_BIG_ENDIAN
4557 	/* Set version, header length, and tos */
4558 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4559 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4560 	    icmp->icmp_type_of_service);
4561 	/* Set ttl and protocol */
4562 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4563 #else
4564 	/* Set version, header length, and tos */
4565 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4566 	    ((icmp->icmp_type_of_service << 8) |
4567 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4568 	/* Set ttl and protocol */
4569 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4570 #endif
4571 	if (pktinfop->ip4_addr != INADDR_ANY) {
4572 		ipha->ipha_src = pktinfop->ip4_addr;
4573 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4574 	} else {
4575 
4576 		/*
4577 		 * Copy our address into the packet.  If this is zero,
4578 		 * ip will fill in the real source address.
4579 		 */
4580 		IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4581 	}
4582 
4583 	ipha->ipha_fragment_offset_and_flags = 0;
4584 
4585 	if (pktinfop->ip4_ill_index != 0) {
4586 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4587 	}
4588 
4589 
4590 	/*
4591 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4592 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4593 	 * tell IP that the application has sent a complete IP header and not
4594 	 * to compute the transport checksum nor change the DF flag.
4595 	 */
4596 	ipha->ipha_ident = IP_HDR_INCLUDED;
4597 
4598 	/* Finish common formatting of the packet. */
4599 	mp1->b_rptr = (uchar_t *)ipha;
4600 
4601 	ip_len = mp1->b_wptr - (uchar_t *)ipha;
4602 	if (mp1->b_cont != NULL)
4603 		ip_len += msgdsize(mp1->b_cont);
4604 
4605 	/*
4606 	 * Set the length into the IP header.
4607 	 * If the length is greater than the maximum allowed by IP,
4608 	 * then free the message and return. Do not try and send it
4609 	 * as this can cause problems in layers below.
4610 	 */
4611 	if (ip_len > IP_MAXPACKET) {
4612 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4613 		icmp_ud_err(q, mp, EMSGSIZE);
4614 		return;
4615 	}
4616 	ipha->ipha_length = htons((uint16_t)ip_len);
4617 	/*
4618 	 * Copy in the destination address from the T_UNITDATA
4619 	 * request
4620 	 */
4621 	ipha->ipha_dst = v4dst;
4622 
4623 	/*
4624 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4625 	 */
4626 	if (CLASSD(v4dst))
4627 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4628 
4629 	/* Copy in options if any */
4630 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4631 		bcopy(icmp->icmp_ip_snd_options,
4632 		    &ipha[1], icmp->icmp_ip_snd_options_len);
4633 		/*
4634 		 * Massage source route putting first source route in ipha_dst.
4635 		 * Ignore the destination in the T_unitdata_req.
4636 		 */
4637 		(void) ip_massage_options(ipha, is->is_netstack);
4638 	}
4639 
4640 	freeb(mp);
4641 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
4642 	mblk_setcred(mp1, connp->conn_cred);
4643 	ip_output_options(Q_TO_CONN(q), mp1, q, IP_WPUT, &optinfo);
4644 #undef	ipha
4645 #undef tudr
4646 }
4647 
4648 static boolean_t
4649 icmp_update_label_v6(queue_t *wq, icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4650 {
4651 	int err;
4652 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4653 	icmp_stack_t		*is = icmp->icmp_is;
4654 	conn_t	*connp = icmp->icmp_connp;
4655 
4656 	err = tsol_compute_label_v6(DB_CREDDEF(mp, connp->conn_cred), dst,
4657 	    opt_storage, icmp->icmp_mac_exempt,
4658 	    is->is_netstack->netstack_ip);
4659 	if (err == 0) {
4660 		err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4661 		    &icmp->icmp_label_len_v6, opt_storage);
4662 	}
4663 	if (err != 0) {
4664 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4665 		DTRACE_PROBE4(
4666 		    tx__ip__log__drop__updatelabel__icmp6,
4667 		    char *, "queue(1) failed to update options(2) on mp(3)",
4668 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4669 		icmp_ud_err(wq, mp, err);
4670 		return (B_FALSE);
4671 	}
4672 
4673 	icmp->icmp_v6lastdst = *dst;
4674 	return (B_TRUE);
4675 }
4676 
4677 /*
4678  * icmp_wput_ipv6():
4679  * Assumes that icmp_wput did some sanity checking on the destination
4680  * address, but that the label may not yet be correct.
4681  */
4682 void
4683 icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6, t_scalar_t tudr_optlen)
4684 {
4685 	ip6_t			*ip6h;
4686 	ip6i_t			*ip6i;	/* mp1->b_rptr even if no ip6i_t */
4687 	mblk_t			*mp1;
4688 	int			ip_hdr_len = IPV6_HDR_LEN;
4689 	size_t			ip_len;
4690 	icmp_t			*icmp = Q_TO_ICMP(q);
4691 	icmp_stack_t		*is = icmp->icmp_is;
4692 	ip6_pkt_t		ipp_s;	/* For ancillary data options */
4693 	ip6_pkt_t		*ipp = &ipp_s;
4694 	ip6_pkt_t		*tipp;
4695 	uint32_t		csum = 0;
4696 	uint_t			ignore = 0;
4697 	uint_t			option_exists = 0, is_sticky = 0;
4698 	uint8_t			*cp;
4699 	uint8_t			*nxthdr_ptr;
4700 	in6_addr_t		ip6_dst;
4701 
4702 	/*
4703 	 * If the local address is a mapped address return
4704 	 * an error.
4705 	 * It would be possible to send an IPv6 packet but the
4706 	 * response would never make it back to the application
4707 	 * since it is bound to a mapped address.
4708 	 */
4709 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4710 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4711 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4712 		return;
4713 	}
4714 
4715 	ipp->ipp_fields = 0;
4716 	ipp->ipp_sticky_ignored = 0;
4717 
4718 	/*
4719 	 * If TPI options passed in, feed it for verification and handling
4720 	 */
4721 	if (tudr_optlen != 0) {
4722 		int error;
4723 
4724 		if (icmp_unitdata_opt_process(q, mp, &error,
4725 		    (void *)ipp) < 0) {
4726 			/* failure */
4727 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4728 			icmp_ud_err(q, mp, error);
4729 			return;
4730 		}
4731 		ignore = ipp->ipp_sticky_ignored;
4732 		ASSERT(error == 0);
4733 	}
4734 
4735 	if (sin6->sin6_scope_id != 0 &&
4736 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4737 		/*
4738 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4739 		 * option nor ancillary data.  It needs to be
4740 		 * explicitly set in options_exists.
4741 		 */
4742 		option_exists |= IPPF_SCOPE_ID;
4743 	}
4744 
4745 	/*
4746 	 * Compute the destination address
4747 	 */
4748 	ip6_dst = sin6->sin6_addr;
4749 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
4750 		ip6_dst = ipv6_loopback;
4751 
4752 	/*
4753 	 * If we're not going to the same destination as last time, then
4754 	 * recompute the label required.  This is done in a separate routine to
4755 	 * avoid blowing up our stack here.
4756 	 */
4757 	if (is_system_labeled() &&
4758 	    !IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst) &&
4759 	    !icmp_update_label_v6(q, icmp, mp, &ip6_dst)) {
4760 		return;
4761 	}
4762 
4763 	/*
4764 	 * If there's a security label here, then we ignore any options the
4765 	 * user may try to set.  We keep the peer's label as a hidden sticky
4766 	 * option.
4767 	 */
4768 	if (icmp->icmp_label_len_v6 > 0) {
4769 		ignore &= ~IPPF_HOPOPTS;
4770 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
4771 	}
4772 
4773 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
4774 	    (ipp->ipp_fields == 0)) {
4775 		/* No sticky options nor ancillary data. */
4776 		goto no_options;
4777 	}
4778 
4779 	/*
4780 	 * Go through the options figuring out where each is going to
4781 	 * come from and build two masks.  The first mask indicates if
4782 	 * the option exists at all.  The second mask indicates if the
4783 	 * option is sticky or ancillary.
4784 	 */
4785 	if (!(ignore & IPPF_HOPOPTS)) {
4786 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
4787 			option_exists |= IPPF_HOPOPTS;
4788 			ip_hdr_len += ipp->ipp_hopoptslen;
4789 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
4790 			option_exists |= IPPF_HOPOPTS;
4791 			is_sticky |= IPPF_HOPOPTS;
4792 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen;
4793 		}
4794 	}
4795 
4796 	if (!(ignore & IPPF_RTHDR)) {
4797 		if (ipp->ipp_fields & IPPF_RTHDR) {
4798 			option_exists |= IPPF_RTHDR;
4799 			ip_hdr_len += ipp->ipp_rthdrlen;
4800 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
4801 			option_exists |= IPPF_RTHDR;
4802 			is_sticky |= IPPF_RTHDR;
4803 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
4804 		}
4805 	}
4806 
4807 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
4808 		/*
4809 		 * Need to have a router header to use these.
4810 		 */
4811 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
4812 			option_exists |= IPPF_RTDSTOPTS;
4813 			ip_hdr_len += ipp->ipp_rtdstoptslen;
4814 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
4815 			option_exists |= IPPF_RTDSTOPTS;
4816 			is_sticky |= IPPF_RTDSTOPTS;
4817 			ip_hdr_len +=
4818 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
4819 		}
4820 	}
4821 
4822 	if (!(ignore & IPPF_DSTOPTS)) {
4823 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
4824 			option_exists |= IPPF_DSTOPTS;
4825 			ip_hdr_len += ipp->ipp_dstoptslen;
4826 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
4827 			option_exists |= IPPF_DSTOPTS;
4828 			is_sticky |= IPPF_DSTOPTS;
4829 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
4830 		}
4831 	}
4832 
4833 	if (!(ignore & IPPF_IFINDEX)) {
4834 		if (ipp->ipp_fields & IPPF_IFINDEX) {
4835 			option_exists |= IPPF_IFINDEX;
4836 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
4837 			option_exists |= IPPF_IFINDEX;
4838 			is_sticky |= IPPF_IFINDEX;
4839 		}
4840 	}
4841 
4842 	if (!(ignore & IPPF_ADDR)) {
4843 		if (ipp->ipp_fields & IPPF_ADDR) {
4844 			option_exists |= IPPF_ADDR;
4845 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
4846 			option_exists |= IPPF_ADDR;
4847 			is_sticky |= IPPF_ADDR;
4848 		}
4849 	}
4850 
4851 	if (!(ignore & IPPF_DONTFRAG)) {
4852 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
4853 			option_exists |= IPPF_DONTFRAG;
4854 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
4855 			option_exists |= IPPF_DONTFRAG;
4856 			is_sticky |= IPPF_DONTFRAG;
4857 		}
4858 	}
4859 
4860 	if (!(ignore & IPPF_USE_MIN_MTU)) {
4861 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
4862 			option_exists |= IPPF_USE_MIN_MTU;
4863 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
4864 		    IPPF_USE_MIN_MTU) {
4865 			option_exists |= IPPF_USE_MIN_MTU;
4866 			is_sticky |= IPPF_USE_MIN_MTU;
4867 		}
4868 	}
4869 
4870 	if (!(ignore & IPPF_NEXTHOP)) {
4871 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
4872 			option_exists |= IPPF_NEXTHOP;
4873 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
4874 			option_exists |= IPPF_NEXTHOP;
4875 			is_sticky |= IPPF_NEXTHOP;
4876 		}
4877 	}
4878 
4879 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
4880 		option_exists |= IPPF_HOPLIMIT;
4881 	/* IPV6_HOPLIMIT can never be sticky */
4882 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
4883 
4884 	if (!(ignore & IPPF_UNICAST_HOPS) &&
4885 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
4886 		option_exists |= IPPF_UNICAST_HOPS;
4887 		is_sticky |= IPPF_UNICAST_HOPS;
4888 	}
4889 
4890 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
4891 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
4892 		option_exists |= IPPF_MULTICAST_HOPS;
4893 		is_sticky |= IPPF_MULTICAST_HOPS;
4894 	}
4895 
4896 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
4897 		/* This is a sticky socket option only */
4898 		option_exists |= IPPF_NO_CKSUM;
4899 		is_sticky |= IPPF_NO_CKSUM;
4900 	}
4901 
4902 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
4903 		/* This is a sticky socket option only */
4904 		option_exists |= IPPF_RAW_CKSUM;
4905 		is_sticky |= IPPF_RAW_CKSUM;
4906 	}
4907 
4908 	if (!(ignore & IPPF_TCLASS)) {
4909 		if (ipp->ipp_fields & IPPF_TCLASS) {
4910 			option_exists |= IPPF_TCLASS;
4911 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
4912 			option_exists |= IPPF_TCLASS;
4913 			is_sticky |= IPPF_TCLASS;
4914 		}
4915 	}
4916 
4917 no_options:
4918 
4919 	/*
4920 	 * If any options carried in the ip6i_t were specified, we
4921 	 * need to account for the ip6i_t in the data we'll be sending
4922 	 * down.
4923 	 */
4924 	if (option_exists & IPPF_HAS_IP6I)
4925 		ip_hdr_len += sizeof (ip6i_t);
4926 
4927 	/* check/fix buffer config, setup pointers into it */
4928 	mp1 = mp->b_cont;
4929 	ip6h = (ip6_t *)&mp1->b_rptr[-ip_hdr_len];
4930 	if ((mp1->b_datap->db_ref != 1) ||
4931 	    ((unsigned char *)ip6h < mp1->b_datap->db_base) ||
4932 	    !OK_32PTR(ip6h)) {
4933 		/* Try to get everything in a single mblk next time */
4934 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
4935 			icmp->icmp_max_hdr_len = ip_hdr_len;
4936 			(void) mi_set_sth_wroff(RD(q),
4937 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
4938 		}
4939 		mp1 = allocb(ip_hdr_len + is->is_wroff_extra, BPRI_LO);
4940 		if (!mp1) {
4941 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4942 			icmp_ud_err(q, mp, ENOMEM);
4943 			return;
4944 		}
4945 		mp1->b_cont = mp->b_cont;
4946 		mp1->b_wptr = mp1->b_datap->db_lim;
4947 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
4948 	}
4949 	mp1->b_rptr = (unsigned char *)ip6h;
4950 	ip6i = (ip6i_t *)ip6h;
4951 
4952 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
4953 	if (option_exists & IPPF_HAS_IP6I) {
4954 		ip6h = (ip6_t *)&ip6i[1];
4955 		ip6i->ip6i_flags = 0;
4956 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
4957 
4958 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
4959 		if (option_exists & IPPF_SCOPE_ID) {
4960 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4961 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
4962 		} else if (option_exists & IPPF_IFINDEX) {
4963 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
4964 			ASSERT(tipp->ipp_ifindex != 0);
4965 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4966 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
4967 		}
4968 
4969 		if (option_exists & IPPF_RAW_CKSUM) {
4970 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
4971 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
4972 		}
4973 
4974 		if (option_exists & IPPF_NO_CKSUM) {
4975 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
4976 		}
4977 
4978 		if (option_exists & IPPF_ADDR) {
4979 			/*
4980 			 * Enable per-packet source address verification if
4981 			 * IPV6_PKTINFO specified the source address.
4982 			 * ip6_src is set in the transport's _wput function.
4983 			 */
4984 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
4985 		}
4986 
4987 		if (option_exists & IPPF_DONTFRAG) {
4988 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
4989 		}
4990 
4991 		if (option_exists & IPPF_USE_MIN_MTU) {
4992 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
4993 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
4994 		}
4995 
4996 		if (option_exists & IPPF_NEXTHOP) {
4997 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
4998 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
4999 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
5000 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
5001 		}
5002 
5003 		/*
5004 		 * tell IP this is an ip6i_t private header
5005 		 */
5006 		ip6i->ip6i_nxt = IPPROTO_RAW;
5007 	}
5008 
5009 	/* Initialize IPv6 header */
5010 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5011 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
5012 
5013 	/* Set the hoplimit of the outgoing packet. */
5014 	if (option_exists & IPPF_HOPLIMIT) {
5015 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
5016 		ip6h->ip6_hops = ipp->ipp_hoplimit;
5017 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5018 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
5019 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
5020 		if (option_exists & IPPF_MULTICAST_HOPS)
5021 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5022 	} else {
5023 		ip6h->ip6_hops = icmp->icmp_ttl;
5024 		if (option_exists & IPPF_UNICAST_HOPS)
5025 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5026 	}
5027 
5028 	if (option_exists & IPPF_ADDR) {
5029 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
5030 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
5031 		ip6h->ip6_src = tipp->ipp_addr;
5032 	} else {
5033 		/*
5034 		 * The source address was not set using IPV6_PKTINFO.
5035 		 * First look at the bound source.
5036 		 * If unspecified fallback to __sin6_src_id.
5037 		 */
5038 		ip6h->ip6_src = icmp->icmp_v6src;
5039 		if (sin6->__sin6_src_id != 0 &&
5040 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
5041 			ip_srcid_find_id(sin6->__sin6_src_id,
5042 			    &ip6h->ip6_src, icmp->icmp_zoneid,
5043 			    is->is_netstack);
5044 		}
5045 	}
5046 
5047 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
5048 	cp = (uint8_t *)&ip6h[1];
5049 
5050 	/*
5051 	 * Here's where we have to start stringing together
5052 	 * any extension headers in the right order:
5053 	 * Hop-by-hop, destination, routing, and final destination opts.
5054 	 */
5055 	if (option_exists & IPPF_HOPOPTS) {
5056 		/* Hop-by-hop options */
5057 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
5058 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
5059 
5060 		*nxthdr_ptr = IPPROTO_HOPOPTS;
5061 		nxthdr_ptr = &hbh->ip6h_nxt;
5062 
5063 		bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
5064 		cp += tipp->ipp_hopoptslen;
5065 	}
5066 	/*
5067 	 * En-route destination options
5068 	 * Only do them if there's a routing header as well
5069 	 */
5070 	if (option_exists & IPPF_RTDSTOPTS) {
5071 		ip6_dest_t *dst = (ip6_dest_t *)cp;
5072 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
5073 
5074 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5075 		nxthdr_ptr = &dst->ip6d_nxt;
5076 
5077 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
5078 		cp += tipp->ipp_rtdstoptslen;
5079 	}
5080 	/*
5081 	 * Routing header next
5082 	 */
5083 	if (option_exists & IPPF_RTHDR) {
5084 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
5085 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
5086 
5087 		*nxthdr_ptr = IPPROTO_ROUTING;
5088 		nxthdr_ptr = &rt->ip6r_nxt;
5089 
5090 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
5091 		cp += tipp->ipp_rthdrlen;
5092 	}
5093 	/*
5094 	 * Do ultimate destination options
5095 	 */
5096 	if (option_exists & IPPF_DSTOPTS) {
5097 		ip6_dest_t *dest = (ip6_dest_t *)cp;
5098 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
5099 
5100 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5101 		nxthdr_ptr = &dest->ip6d_nxt;
5102 
5103 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
5104 		cp += tipp->ipp_dstoptslen;
5105 	}
5106 
5107 	/*
5108 	 * Now set the last header pointer to the proto passed in
5109 	 */
5110 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
5111 	*nxthdr_ptr = icmp->icmp_proto;
5112 
5113 	/*
5114 	 * Copy in the destination address
5115 	 */
5116 	ip6h->ip6_dst = ip6_dst;
5117 
5118 	ip6h->ip6_vcf =
5119 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
5120 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
5121 
5122 	if (option_exists & IPPF_TCLASS) {
5123 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
5124 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
5125 		    tipp->ipp_tclass);
5126 	}
5127 	if (option_exists & IPPF_RTHDR) {
5128 		ip6_rthdr_t	*rth;
5129 
5130 		/*
5131 		 * Perform any processing needed for source routing.
5132 		 * We know that all extension headers will be in the same mblk
5133 		 * as the IPv6 header.
5134 		 */
5135 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
5136 		if (rth != NULL && rth->ip6r_segleft != 0) {
5137 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
5138 				/*
5139 				 * Drop packet - only support Type 0 routing.
5140 				 * Notify the application as well.
5141 				 */
5142 				icmp_ud_err(q, mp, EPROTO);
5143 				BUMP_MIB(&is->is_rawip_mib,
5144 				    rawipOutErrors);
5145 				return;
5146 			}
5147 			/*
5148 			 * rth->ip6r_len is twice the number of
5149 			 * addresses in the header
5150 			 */
5151 			if (rth->ip6r_len & 0x1) {
5152 				icmp_ud_err(q, mp, EPROTO);
5153 				BUMP_MIB(&is->is_rawip_mib,
5154 				    rawipOutErrors);
5155 				return;
5156 			}
5157 			/*
5158 			 * Shuffle the routing header and ip6_dst
5159 			 * addresses, and get the checksum difference
5160 			 * between the first hop (in ip6_dst) and
5161 			 * the destination (in the last routing hdr entry).
5162 			 */
5163 			csum = ip_massage_options_v6(ip6h, rth,
5164 			    is->is_netstack);
5165 			/*
5166 			 * Verify that the first hop isn't a mapped address.
5167 			 * Routers along the path need to do this verification
5168 			 * for subsequent hops.
5169 			 */
5170 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
5171 				icmp_ud_err(q, mp, EADDRNOTAVAIL);
5172 				BUMP_MIB(&is->is_rawip_mib,
5173 				    rawipOutErrors);
5174 				return;
5175 			}
5176 		}
5177 	}
5178 
5179 	ip_len = mp1->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
5180 	if (mp1->b_cont != NULL)
5181 		ip_len += msgdsize(mp1->b_cont);
5182 
5183 	/*
5184 	 * Set the length into the IP header.
5185 	 * If the length is greater than the maximum allowed by IP,
5186 	 * then free the message and return. Do not try and send it
5187 	 * as this can cause problems in layers below.
5188 	 */
5189 	if (ip_len > IP_MAXPACKET) {
5190 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5191 		icmp_ud_err(q, mp, EMSGSIZE);
5192 		return;
5193 	}
5194 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
5195 		uint_t	cksum_off;	/* From ip6i == mp1->b_rptr */
5196 		uint16_t *cksum_ptr;
5197 		uint_t	ext_hdrs_len;
5198 
5199 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
5200 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
5201 		    icmp->icmp_checksum_off == 2);
5202 
5203 		/*
5204 		 * We make it easy for IP to include our pseudo header
5205 		 * by putting our length in uh_checksum, modified (if
5206 		 * we have a routing header) by the checksum difference
5207 		 * between the ultimate destination and first hop addresses.
5208 		 * Note: ICMPv6 must always checksum the packet.
5209 		 */
5210 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
5211 		if (cksum_off + sizeof (uint16_t) > mp1->b_wptr - mp1->b_rptr) {
5212 			if (!pullupmsg(mp1, cksum_off + sizeof (uint16_t))) {
5213 				BUMP_MIB(&is->is_rawip_mib,
5214 				    rawipOutErrors);
5215 				freemsg(mp);
5216 				return;
5217 			}
5218 			ip6i = (ip6i_t *)mp1->b_rptr;
5219 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
5220 				ip6h = (ip6_t *)&ip6i[1];
5221 			else
5222 				ip6h = (ip6_t *)ip6i;
5223 		}
5224 		/* Add payload length to checksum */
5225 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
5226 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
5227 		csum += htons(ip_len - ext_hdrs_len);
5228 
5229 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
5230 		csum = (csum & 0xFFFF) + (csum >> 16);
5231 		*cksum_ptr = (uint16_t)csum;
5232 	}
5233 
5234 #ifdef _LITTLE_ENDIAN
5235 	ip_len = htons(ip_len);
5236 #endif
5237 	ip6h->ip6_plen = (uint16_t)ip_len;
5238 
5239 	freeb(mp);
5240 
5241 	/* We're done. Pass the packet to IP */
5242 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
5243 	ip_output_v6(icmp->icmp_connp, mp1, q, IP_WPUT);
5244 }
5245 
5246 static void
5247 icmp_wput_other(queue_t *q, mblk_t *mp)
5248 {
5249 	uchar_t	*rptr = mp->b_rptr;
5250 	struct iocblk *iocp;
5251 #define	tudr ((struct T_unitdata_req *)rptr)
5252 	conn_t	*connp = Q_TO_CONN(q);
5253 	icmp_t	*icmp = connp->conn_icmp;
5254 	icmp_stack_t *is = icmp->icmp_is;
5255 	cred_t *cr;
5256 
5257 	cr = DB_CREDDEF(mp, connp->conn_cred);
5258 
5259 	switch (mp->b_datap->db_type) {
5260 	case M_PROTO:
5261 	case M_PCPROTO:
5262 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5263 			/*
5264 			 * If the message does not contain a PRIM_type,
5265 			 * throw it away.
5266 			 */
5267 			freemsg(mp);
5268 			return;
5269 		}
5270 		switch (((union T_primitives *)rptr)->type) {
5271 		case T_ADDR_REQ:
5272 			icmp_addr_req(q, mp);
5273 			return;
5274 		case O_T_BIND_REQ:
5275 		case T_BIND_REQ:
5276 			icmp_bind(q, mp);
5277 			return;
5278 		case T_CONN_REQ:
5279 			icmp_connect(q, mp);
5280 			return;
5281 		case T_CAPABILITY_REQ:
5282 			icmp_capability_req(q, mp);
5283 			return;
5284 		case T_INFO_REQ:
5285 			icmp_info_req(q, mp);
5286 			return;
5287 		case T_UNITDATA_REQ:
5288 			/*
5289 			 * If a T_UNITDATA_REQ gets here, the address must
5290 			 * be bad.  Valid T_UNITDATA_REQs are found above
5291 			 * and break to below this switch.
5292 			 */
5293 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
5294 			return;
5295 		case T_UNBIND_REQ:
5296 			icmp_unbind(q, mp);
5297 			return;
5298 
5299 		case T_SVR4_OPTMGMT_REQ:
5300 			if (!snmpcom_req(q, mp, icmp_snmp_set, ip_snmp_get,
5301 			    cr)) {
5302 				/* Only IP can return anything meaningful */
5303 				(void) svr4_optcom_req(q, mp, cr,
5304 				    &icmp_opt_obj, B_TRUE);
5305 			}
5306 			return;
5307 
5308 		case T_OPTMGMT_REQ:
5309 			/* Only IP can return anything meaningful */
5310 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj, B_TRUE);
5311 			return;
5312 
5313 		case T_DISCON_REQ:
5314 			icmp_disconnect(q, mp);
5315 			return;
5316 
5317 		/* The following TPI message is not supported by icmp. */
5318 		case O_T_CONN_RES:
5319 		case T_CONN_RES:
5320 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
5321 			return;
5322 
5323 		/* The following 3 TPI requests are illegal for icmp. */
5324 		case T_DATA_REQ:
5325 		case T_EXDATA_REQ:
5326 		case T_ORDREL_REQ:
5327 			freemsg(mp);
5328 			(void) putctl1(RD(q), M_ERROR, EPROTO);
5329 			return;
5330 		default:
5331 			break;
5332 		}
5333 		break;
5334 	case M_IOCTL:
5335 		iocp = (struct iocblk *)mp->b_rptr;
5336 		switch (iocp->ioc_cmd) {
5337 		case TI_GETPEERNAME:
5338 			if (icmp->icmp_state != TS_DATA_XFER) {
5339 				/*
5340 				 * If a default destination address has not
5341 				 * been associated with the stream, then we
5342 				 * don't know the peer's name.
5343 				 */
5344 				iocp->ioc_error = ENOTCONN;
5345 		err_ret:;
5346 				iocp->ioc_count = 0;
5347 				mp->b_datap->db_type = M_IOCACK;
5348 				qreply(q, mp);
5349 				return;
5350 			}
5351 			/* FALLTHRU */
5352 		case TI_GETMYNAME:
5353 			/*
5354 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
5355 			 * need to copyin the user's strbuf structure.
5356 			 * Processing will continue in the M_IOCDATA case
5357 			 * below.
5358 			 */
5359 			mi_copyin(q, mp, NULL,
5360 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
5361 			return;
5362 		case ND_SET:
5363 			/* nd_getset performs the necessary error checking */
5364 		case ND_GET:
5365 			if (nd_getset(q, is->is_nd, mp)) {
5366 				qreply(q, mp);
5367 				return;
5368 			}
5369 			break;
5370 		default:
5371 			break;
5372 		}
5373 		break;
5374 	case M_IOCDATA:
5375 		icmp_wput_iocdata(q, mp);
5376 		return;
5377 	default:
5378 		break;
5379 	}
5380 	ip_wput(q, mp);
5381 }
5382 
5383 /*
5384  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
5385  * messages.
5386  */
5387 static void
5388 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
5389 {
5390 	mblk_t	*mp1;
5391 	STRUCT_HANDLE(strbuf, sb);
5392 	icmp_t	*icmp;
5393 	in6_addr_t	v6addr;
5394 	ipaddr_t	v4addr;
5395 	uint32_t	flowinfo = 0;
5396 	int		addrlen;
5397 
5398 	/* Make sure it is one of ours. */
5399 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5400 	case TI_GETMYNAME:
5401 	case TI_GETPEERNAME:
5402 		break;
5403 	default:
5404 		icmp = Q_TO_ICMP(q);
5405 		ip_output(icmp->icmp_connp, mp, q, IP_WPUT);
5406 		return;
5407 	}
5408 	switch (mi_copy_state(q, mp, &mp1)) {
5409 	case -1:
5410 		return;
5411 	case MI_COPY_CASE(MI_COPY_IN, 1):
5412 		break;
5413 	case MI_COPY_CASE(MI_COPY_OUT, 1):
5414 		/*
5415 		 * The address has been copied out, so now
5416 		 * copyout the strbuf.
5417 		 */
5418 		mi_copyout(q, mp);
5419 		return;
5420 	case MI_COPY_CASE(MI_COPY_OUT, 2):
5421 		/*
5422 		 * The address and strbuf have been copied out.
5423 		 * We're done, so just acknowledge the original
5424 		 * M_IOCTL.
5425 		 */
5426 		mi_copy_done(q, mp, 0);
5427 		return;
5428 	default:
5429 		/*
5430 		 * Something strange has happened, so acknowledge
5431 		 * the original M_IOCTL with an EPROTO error.
5432 		 */
5433 		mi_copy_done(q, mp, EPROTO);
5434 		return;
5435 	}
5436 	/*
5437 	 * Now we have the strbuf structure for TI_GETMYNAME
5438 	 * and TI_GETPEERNAME.  Next we copyout the requested
5439 	 * address and then we'll copyout the strbuf.
5440 	 */
5441 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5442 	    (void *)mp1->b_rptr);
5443 	icmp = Q_TO_ICMP(q);
5444 	if (icmp->icmp_family == AF_INET)
5445 		addrlen = sizeof (sin_t);
5446 	else
5447 		addrlen = sizeof (sin6_t);
5448 
5449 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5450 		mi_copy_done(q, mp, EINVAL);
5451 		return;
5452 	}
5453 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5454 	case TI_GETMYNAME:
5455 		if (icmp->icmp_family == AF_INET) {
5456 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5457 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
5458 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5459 				v4addr = V4_PART_OF_V6(icmp->icmp_v6src);
5460 			} else {
5461 				/*
5462 				 * INADDR_ANY
5463 				 * icmp_v6src is not set, we might be bound to
5464 				 * broadcast/multicast. Use icmp_bound_v6src as
5465 				 * local address instead (that could
5466 				 * also still be INADDR_ANY)
5467 				 */
5468 				v4addr = V4_PART_OF_V6(icmp->icmp_bound_v6src);
5469 			}
5470 		} else {
5471 			/* icmp->icmp_family == AF_INET6 */
5472 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5473 				v6addr = icmp->icmp_v6src;
5474 			} else {
5475 				/*
5476 				 * UNSPECIFIED
5477 				 * icmp_v6src is not set, we might be bound to
5478 				 * broadcast/multicast. Use icmp_bound_v6src as
5479 				 * local address instead (that could
5480 				 * also still be UNSPECIFIED)
5481 				 */
5482 				v6addr = icmp->icmp_bound_v6src;
5483 			}
5484 		}
5485 		break;
5486 	case TI_GETPEERNAME:
5487 		if (icmp->icmp_family == AF_INET) {
5488 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5489 			v4addr = V4_PART_OF_V6(icmp->icmp_v6dst);
5490 		} else {
5491 			/* icmp->icmp_family == AF_INET6) */
5492 			v6addr = icmp->icmp_v6dst;
5493 			flowinfo = icmp->icmp_flowinfo;
5494 		}
5495 		break;
5496 	default:
5497 		mi_copy_done(q, mp, EPROTO);
5498 		return;
5499 	}
5500 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5501 	if (!mp1)
5502 		return;
5503 
5504 	if (icmp->icmp_family == AF_INET) {
5505 		sin_t *sin;
5506 
5507 		STRUCT_FSET(sb, len, (int)sizeof (sin_t));
5508 		sin = (sin_t *)mp1->b_rptr;
5509 		mp1->b_wptr = (uchar_t *)&sin[1];
5510 		*sin = sin_null;
5511 		sin->sin_family = AF_INET;
5512 		sin->sin_addr.s_addr = v4addr;
5513 	} else {
5514 		/* icmp->icmp_family == AF_INET6 */
5515 		sin6_t *sin6;
5516 
5517 		ASSERT(icmp->icmp_family == AF_INET6);
5518 		STRUCT_FSET(sb, len, (int)sizeof (sin6_t));
5519 		sin6 = (sin6_t *)mp1->b_rptr;
5520 		mp1->b_wptr = (uchar_t *)&sin6[1];
5521 		*sin6 = sin6_null;
5522 		sin6->sin6_family = AF_INET6;
5523 		sin6->sin6_flowinfo = flowinfo;
5524 		sin6->sin6_addr = v6addr;
5525 	}
5526 	/* Copy out the address */
5527 	mi_copyout(q, mp);
5528 }
5529 
5530 static int
5531 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5532     void *thisdg_attrs)
5533 {
5534 	conn_t	*connp = Q_TO_CONN(q);
5535 	struct T_unitdata_req *udreqp;
5536 	int is_absreq_failure;
5537 	cred_t *cr;
5538 
5539 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5540 	*errorp = 0;
5541 
5542 	cr = DB_CREDDEF(mp, connp->conn_cred);
5543 
5544 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5545 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5546 	    thisdg_attrs, &is_absreq_failure);
5547 
5548 	if (*errorp != 0) {
5549 		/*
5550 		 * Note: No special action needed in this
5551 		 * module for "is_absreq_failure"
5552 		 */
5553 		return (-1);		/* failure */
5554 	}
5555 	ASSERT(is_absreq_failure == 0);
5556 	return (0);	/* success */
5557 }
5558 
5559 void
5560 icmp_ddi_init(void)
5561 {
5562 	icmp_max_optsize = optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5563 	    icmp_opt_obj.odb_opt_arr_cnt);
5564 
5565 	/*
5566 	 * We want to be informed each time a stack is created or
5567 	 * destroyed in the kernel, so we can maintain the
5568 	 * set of icmp_stack_t's.
5569 	 */
5570 	netstack_register(NS_ICMP, rawip_stack_init, NULL, rawip_stack_fini);
5571 }
5572 
5573 void
5574 icmp_ddi_destroy(void)
5575 {
5576 	netstack_unregister(NS_ICMP);
5577 }
5578 
5579 /*
5580  * Initialize the ICMP stack instance.
5581  */
5582 static void *
5583 rawip_stack_init(netstackid_t stackid, netstack_t *ns)
5584 {
5585 	icmp_stack_t	*is;
5586 	icmpparam_t	*pa;
5587 
5588 	is = (icmp_stack_t *)kmem_zalloc(sizeof (*is), KM_SLEEP);
5589 	is->is_netstack = ns;
5590 
5591 	pa = (icmpparam_t *)kmem_alloc(sizeof (icmp_param_arr), KM_SLEEP);
5592 	is->is_param_arr = pa;
5593 	bcopy(icmp_param_arr, is->is_param_arr, sizeof (icmp_param_arr));
5594 
5595 	(void) icmp_param_register(&is->is_nd,
5596 	    is->is_param_arr, A_CNT(icmp_param_arr));
5597 	is->is_ksp = rawip_kstat_init(stackid);
5598 	return (is);
5599 }
5600 
5601 /*
5602  * Free the ICMP stack instance.
5603  */
5604 static void
5605 rawip_stack_fini(netstackid_t stackid, void *arg)
5606 {
5607 	icmp_stack_t *is = (icmp_stack_t *)arg;
5608 
5609 	nd_free(&is->is_nd);
5610 	kmem_free(is->is_param_arr, sizeof (icmp_param_arr));
5611 	is->is_param_arr = NULL;
5612 
5613 	rawip_kstat_fini(stackid, is->is_ksp);
5614 	is->is_ksp = NULL;
5615 	kmem_free(is, sizeof (*is));
5616 }
5617 
5618 static void *
5619 rawip_kstat_init(netstackid_t stackid) {
5620 	kstat_t	*ksp;
5621 
5622 	rawip_named_kstat_t template = {
5623 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5624 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5625 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5626 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5627 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5628 	};
5629 
5630 	ksp = kstat_create_netstack("icmp", 0, "rawip", "mib2",
5631 					KSTAT_TYPE_NAMED,
5632 					NUM_OF_FIELDS(rawip_named_kstat_t),
5633 					0, stackid);
5634 	if (ksp == NULL || ksp->ks_data == NULL)
5635 		return (NULL);
5636 
5637 	bcopy(&template, ksp->ks_data, sizeof (template));
5638 	ksp->ks_update = rawip_kstat_update;
5639 	ksp->ks_private = (void *)(uintptr_t)stackid;
5640 
5641 	kstat_install(ksp);
5642 	return (ksp);
5643 }
5644 
5645 static void
5646 rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
5647 {
5648 	if (ksp != NULL) {
5649 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
5650 		kstat_delete_netstack(ksp, stackid);
5651 	}
5652 }
5653 
5654 static int
5655 rawip_kstat_update(kstat_t *ksp, int rw)
5656 {
5657 	rawip_named_kstat_t *rawipkp;
5658 	netstackid_t	stackid = (netstackid_t)(uintptr_t)ksp->ks_private;
5659 	netstack_t	*ns;
5660 	icmp_stack_t	*is;
5661 
5662 	if ((ksp == NULL) || (ksp->ks_data == NULL))
5663 		return (EIO);
5664 
5665 	if (rw == KSTAT_WRITE)
5666 		return (EACCES);
5667 
5668 	rawipkp = (rawip_named_kstat_t *)ksp->ks_data;
5669 
5670 	ns = netstack_find_by_stackid(stackid);
5671 	if (ns == NULL)
5672 		return (-1);
5673 	is = ns->netstack_icmp;
5674 	if (is == NULL) {
5675 		netstack_rele(ns);
5676 		return (-1);
5677 	}
5678 	rawipkp->inDatagrams.value.ui32 =  is->is_rawip_mib.rawipInDatagrams;
5679 	rawipkp->inCksumErrs.value.ui32 =  is->is_rawip_mib.rawipInCksumErrs;
5680 	rawipkp->inErrors.value.ui32 =	   is->is_rawip_mib.rawipInErrors;
5681 	rawipkp->outDatagrams.value.ui32 = is->is_rawip_mib.rawipOutDatagrams;
5682 	rawipkp->outErrors.value.ui32 =	   is->is_rawip_mib.rawipOutErrors;
5683 	netstack_rele(ns);
5684 	return (0);
5685 }
5686