xref: /titanic_50/usr/src/uts/common/inet/ip/icmp.c (revision 910cba4f2f1e94daf355ee8635285732ac47326c)
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 2006 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 
54 #include <net/route.h>
55 #include <net/if.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/ip6.h>
59 #include <netinet/icmp6.h>
60 #include <inet/common.h>
61 #include <inet/ip.h>
62 #include <inet/ip6.h>
63 #include <inet/mi.h>
64 #include <inet/nd.h>
65 #include <inet/optcom.h>
66 #include <inet/snmpcom.h>
67 #include <inet/kstatcom.h>
68 #include <inet/rawip_impl.h>
69 
70 #include <netinet/ip_mroute.h>
71 #include <inet/tcp.h>
72 #include <net/pfkeyv2.h>
73 #include <inet/ipsec_info.h>
74 #include <inet/ipclassifier.h>
75 
76 #include <sys/tsol/label.h>
77 #include <sys/tsol/tnet.h>
78 
79 #define	ICMP6 "icmp6"
80 major_t	ICMP6_MAJ;
81 
82 /*
83  * Object to represent database of options to search passed to
84  * {sock,tpi}optcom_req() interface routine to take care of option
85  * management and associated methods.
86  * XXX These and other extern's should really move to a icmp header.
87  */
88 extern optdb_obj_t	icmp_opt_obj;
89 extern uint_t		icmp_max_optsize;
90 
91 /*
92  * Synchronization notes:
93  *
94  * At all points in this code where exclusive access is required, we
95  * pass a message to a subroutine by invoking qwriter(..., PERIM_OUTER)
96  * which will arrange to call the routine only after all threads have
97  * exited the shared resource.
98  */
99 
100 /* Named Dispatch Parameter Management Structure */
101 typedef struct icmpparam_s {
102 	uint_t	icmp_param_min;
103 	uint_t	icmp_param_max;
104 	uint_t	icmp_param_value;
105 	char	*icmp_param_name;
106 } icmpparam_t;
107 
108 static void	icmp_addr_req(queue_t *q, mblk_t *mp);
109 static void	icmp_bind(queue_t *q, mblk_t *mp);
110 static void	icmp_bind_proto(queue_t *q);
111 static int	icmp_build_hdrs(queue_t *q, icmp_t *icmp);
112 static void	icmp_capability_req(queue_t *q, mblk_t *mp);
113 static int	icmp_close(queue_t *q);
114 static void	icmp_connect(queue_t *q, mblk_t *mp);
115 static void	icmp_disconnect(queue_t *q, mblk_t *mp);
116 static void	icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
117 		    int sys_error);
118 static void	icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
119 		    t_scalar_t t_error, int sys_error);
120 static void	icmp_icmp_error(queue_t *q, mblk_t *mp);
121 static void	icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp);
122 static void	icmp_info_req(queue_t *q, mblk_t *mp);
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,
126 		    int sflag, cred_t *credp);
127 static int	icmp_unitdata_opt_process(queue_t *q, mblk_t *mp,
128 		    int *errorp, void *thisdg_attrs);
129 static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
130 int		icmp_opt_set(queue_t *q, uint_t optset_context,
131 		    int level, int name, uint_t inlen,
132 		    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
133 		    void *thisdg_attrs, cred_t *cr, mblk_t *mblk);
134 int		icmp_opt_get(queue_t *q, int level, int name,
135 		    uchar_t *ptr);
136 static int	icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
137 static boolean_t icmp_param_register(icmpparam_t *icmppa, int cnt);
138 static int	icmp_param_set(queue_t *q, mblk_t *mp, char *value,
139 		    caddr_t cp, cred_t *cr);
140 static void	icmp_rput(queue_t *q, mblk_t *mp);
141 static void	icmp_rput_bind_ack(queue_t *q, mblk_t *mp);
142 static int	icmp_snmp_get(queue_t *q, mblk_t *mpctl);
143 static int	icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
144 		    uchar_t *ptr, int len);
145 static int	icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
146 		    cred_t *cr);
147 static void	icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err);
148 static void	icmp_unbind(queue_t *q, mblk_t *mp);
149 static void	icmp_wput(queue_t *q, mblk_t *mp);
150 static void	icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6,
151 		    t_scalar_t tudr_optlen);
152 static void	icmp_wput_other(queue_t *q, mblk_t *mp);
153 static void	icmp_wput_iocdata(queue_t *q, mblk_t *mp);
154 static void	icmp_wput_restricted(queue_t *q, mblk_t *mp);
155 
156 static void	rawip_kstat_init(void);
157 static void	rawip_kstat_fini(void);
158 static int	rawip_kstat_update(kstat_t *kp, int rw);
159 
160 
161 static struct module_info info =  {
162 	5707, "icmp", 1, INFPSZ, 512, 128
163 };
164 
165 static struct qinit rinit = {
166 	(pfi_t)icmp_rput, NULL, icmp_open, icmp_close, NULL, &info
167 };
168 
169 static struct qinit winit = {
170 	(pfi_t)icmp_wput, NULL, NULL, NULL, NULL, &info
171 };
172 
173 struct streamtab icmpinfo = {
174 	&rinit, &winit
175 };
176 
177 static sin_t	sin_null;	/* Zero address for quick clears */
178 static sin6_t	sin6_null;	/* Zero address for quick clears */
179 static void	*icmp_g_head;	/* Head for list of open icmp streams. */
180 static IDP	icmp_g_nd;	/* Points to table of ICMP ND variables. */
181 
182 /* MIB-2 stuff for SNMP */
183 static mib2_rawip_t	rawip_mib;	/* SNMP fixed size info */
184 static kstat_t		*rawip_mibkp;	/* kstat exporting rawip_mib data */
185 
186 /* Default structure copied into T_INFO_ACK messages */
187 static struct T_info_ack icmp_g_t_info_ack = {
188 	T_INFO_ACK,
189 	IP_MAXPACKET,	 /* TSDU_size.  icmp allows maximum size messages. */
190 	T_INVALID,	/* ETSDU_size.  icmp does not support expedited data. */
191 	T_INVALID,	/* CDATA_size. icmp does not support connect data. */
192 	T_INVALID,	/* DDATA_size. icmp does not support disconnect data. */
193 	0,		/* ADDR_size - filled in later. */
194 	0,		/* OPT_size - not initialized here */
195 	IP_MAXPACKET,	/* TIDU_size.  icmp allows maximum size messages. */
196 	T_CLTS,		/* SERV_type.  icmp supports connection-less. */
197 	TS_UNBND,	/* CURRENT_state.  This is set from icmp_state. */
198 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
199 };
200 
201 /*
202  * Table of ND variables supported by icmp.  These are loaded into icmp_g_nd
203  * in icmp_open.
204  * All of these are alterable, within the min/max values given, at run time.
205  */
206 static icmpparam_t	icmp_param_arr[] = {
207 	/* min	max	value	name */
208 	{ 0,	128,	32,	"icmp_wroff_extra" },
209 	{ 1,	255,	255,	"icmp_ipv4_ttl" },
210 	{ 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS,	"icmp_ipv6_hoplimit"},
211 	{ 0,	1,	1,	"icmp_bsd_compat" },
212 	{ 4096,	65536,	8192,	"icmp_xmit_hiwat"},
213 	{ 0,	65536,	1024,	"icmp_xmit_lowat"},
214 	{ 4096,	65536,	8192,	"icmp_recv_hiwat"},
215 	{ 65536, 1024*1024*1024, 256*1024,	"icmp_max_buf"},
216 };
217 #define	icmp_wroff_extra		icmp_param_arr[0].icmp_param_value
218 #define	icmp_ipv4_ttl			icmp_param_arr[1].icmp_param_value
219 #define	icmp_ipv6_hoplimit		icmp_param_arr[2].icmp_param_value
220 #define	icmp_bsd_compat			icmp_param_arr[3].icmp_param_value
221 #define	icmp_xmit_hiwat			icmp_param_arr[4].icmp_param_value
222 #define	icmp_xmit_lowat			icmp_param_arr[5].icmp_param_value
223 #define	icmp_recv_hiwat			icmp_param_arr[6].icmp_param_value
224 #define	icmp_max_buf			icmp_param_arr[7].icmp_param_value
225 
226 /*
227  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
228  * passed to icmp_wput.
229  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP
230  * protocol type placed in the message following the address. A T_BIND_ACK
231  * message is passed upstream when ip acknowledges the request.
232  * (Called as writer.)
233  */
234 static void
235 icmp_bind(queue_t *q, mblk_t *mp)
236 {
237 	sin_t	*sin;
238 	sin6_t	*sin6;
239 	mblk_t	*mp1;
240 	struct T_bind_req	*tbr;
241 	icmp_t	*icmp;
242 
243 	icmp = (icmp_t *)q->q_ptr;
244 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
245 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
246 		    "icmp_bind: bad req, len %u",
247 		    (uint_t)(mp->b_wptr - mp->b_rptr));
248 		icmp_err_ack(q, mp, TPROTO, 0);
249 		return;
250 	}
251 	if (icmp->icmp_state != TS_UNBND) {
252 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
253 		    "icmp_bind: bad state, %d", icmp->icmp_state);
254 		icmp_err_ack(q, mp, TOUTSTATE, 0);
255 		return;
256 	}
257 	/*
258 	 * Reallocate the message to make sure we have enough room for an
259 	 * address and the protocol type.
260 	 */
261 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
262 	if (!mp1) {
263 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
264 		return;
265 	}
266 	mp = mp1;
267 	tbr = (struct T_bind_req *)mp->b_rptr;
268 	switch (tbr->ADDR_length) {
269 	case 0:			/* Generic request */
270 		tbr->ADDR_offset = sizeof (struct T_bind_req);
271 		if (icmp->icmp_family == AF_INET) {
272 			tbr->ADDR_length = sizeof (sin_t);
273 			sin = (sin_t *)&tbr[1];
274 			*sin = sin_null;
275 			sin->sin_family = AF_INET;
276 			mp->b_wptr = (uchar_t *)&sin[1];
277 		} else {
278 			ASSERT(icmp->icmp_family == AF_INET6);
279 			tbr->ADDR_length = sizeof (sin6_t);
280 			sin6 = (sin6_t *)&tbr[1];
281 			*sin6 = sin6_null;
282 			sin6->sin6_family = AF_INET6;
283 			mp->b_wptr = (uchar_t *)&sin6[1];
284 		}
285 		break;
286 	case sizeof (sin_t):	/* Complete IP address */
287 		sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset,
288 		    sizeof (sin_t));
289 		if (sin == NULL || !OK_32PTR((char *)sin)) {
290 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
291 			return;
292 		}
293 		if (icmp->icmp_family != AF_INET ||
294 		    sin->sin_family != AF_INET) {
295 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
296 			return;
297 		}
298 		break;
299 	case sizeof (sin6_t):	/* Complete IP address */
300 		sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset,
301 		    sizeof (sin6_t));
302 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
303 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
304 			return;
305 		}
306 		if (icmp->icmp_family != AF_INET6 ||
307 		    sin6->sin6_family != AF_INET6) {
308 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
309 			return;
310 		}
311 		/* No support for mapped addresses on raw sockets */
312 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
313 			icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL);
314 			return;
315 		}
316 		break;
317 	default:
318 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
319 		    "icmp_bind: bad ADDR_length %d", tbr->ADDR_length);
320 		icmp_err_ack(q, mp, TBADADDR, 0);
321 		return;
322 	}
323 	/*
324 	 * Copy the source address into our icmp structure.  This address
325 	 * may still be zero; if so, ip will fill in the correct address
326 	 * each time an outbound packet is passed to it.
327 	 * If we are binding to a broadcast or multicast address icmp_rput
328 	 * will clear the source address when it receives the T_BIND_ACK.
329 	 */
330 	icmp->icmp_state = TS_IDLE;
331 
332 	if (icmp->icmp_family == AF_INET) {
333 		ASSERT(sin != NULL);
334 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
335 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr,
336 		    &icmp->icmp_v6src);
337 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
338 		    icmp->icmp_ip_snd_options_len;
339 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
340 	} else {
341 		int error;
342 
343 		ASSERT(sin6 != NULL);
344 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
345 		icmp->icmp_v6src = sin6->sin6_addr;
346 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
347 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
348 
349 		/* Rebuild the header template */
350 		error = icmp_build_hdrs(q, icmp);
351 		if (error != 0) {
352 			icmp_err_ack(q, mp, TSYSERR, error);
353 			return;
354 		}
355 	}
356 	/*
357 	 * Place protocol type in the O_T_BIND_REQ/T_BIND_REQ following
358 	 * the address.
359 	 */
360 	*mp->b_wptr++ = icmp->icmp_proto;
361 	if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) {
362 		/*
363 		 * Append a request for an IRE if src not 0 (INADDR_ANY)
364 		 */
365 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
366 		if (!mp->b_cont) {
367 			icmp_err_ack(q, mp, TSYSERR, ENOMEM);
368 			return;
369 		}
370 		mp->b_cont->b_wptr += sizeof (ire_t);
371 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
372 	}
373 
374 	/* Pass the O_T_BIND_REQ/T_BIND_REQ to ip. */
375 	putnext(q, mp);
376 }
377 
378 /*
379  * Send message to IP to just bind to the protocol.
380  */
381 static void
382 icmp_bind_proto(queue_t *q)
383 {
384 	mblk_t	*mp;
385 	struct T_bind_req	*tbr;
386 	icmp_t	*icmp;
387 
388 	icmp = (icmp_t *)q->q_ptr;
389 	mp = allocb(sizeof (struct T_bind_req) + sizeof (sin6_t) + 1,
390 	    BPRI_MED);
391 	if (!mp) {
392 		return;
393 	}
394 	mp->b_datap->db_type = M_PROTO;
395 	tbr = (struct T_bind_req *)mp->b_rptr;
396 	tbr->PRIM_type = O_T_BIND_REQ; /* change to T_BIND_REQ ? */
397 	tbr->ADDR_offset = sizeof (struct T_bind_req);
398 	if (icmp->icmp_ipversion == IPV4_VERSION) {
399 		sin_t	*sin;
400 
401 		tbr->ADDR_length = sizeof (sin_t);
402 		sin = (sin_t *)&tbr[1];
403 		*sin = sin_null;
404 		sin->sin_family = AF_INET;
405 		mp->b_wptr = (uchar_t *)&sin[1];
406 	} else {
407 		sin6_t	*sin6;
408 
409 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
410 		tbr->ADDR_length = sizeof (sin6_t);
411 		sin6 = (sin6_t *)&tbr[1];
412 		*sin6 = sin6_null;
413 		sin6->sin6_family = AF_INET6;
414 		mp->b_wptr = (uchar_t *)&sin6[1];
415 	}
416 
417 	/* Place protocol type in the O_T_BIND_REQ following the address. */
418 	*mp->b_wptr++ = icmp->icmp_proto;
419 
420 	/* Pass the O_T_BIND_REQ to ip. */
421 	putnext(q, mp);
422 }
423 
424 /*
425  * This routine handles each T_CONN_REQ message passed to icmp.  It
426  * associates a default destination address with the stream.
427  *
428  * This routine sends down a T_BIND_REQ to IP with the following mblks:
429  *	T_BIND_REQ	- specifying local and remote address.
430  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
431  *	T_OK_ACK	- for the T_CONN_REQ
432  *	T_CONN_CON	- to keep the TPI user happy
433  *
434  * The connect completes in icmp_rput.
435  * When a T_BIND_ACK is received information is extracted from the IRE
436  * and the two appended messages are sent to the TPI user.
437  * Should icmp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert
438  * it to an error ack for the appropriate primitive.
439  */
440 static void
441 icmp_connect(queue_t *q, mblk_t *mp)
442 {
443 	sin_t	*sin;
444 	sin6_t	*sin6;
445 	mblk_t	*mp1, *mp2;
446 	struct T_conn_req	*tcr;
447 	icmp_t	*icmp;
448 	ipaddr_t	v4dst;
449 	in6_addr_t	v6dst;
450 	uint32_t	flowinfo;
451 
452 	icmp = (icmp_t *)q->q_ptr;
453 	tcr = (struct T_conn_req *)mp->b_rptr;
454 	/* Sanity checks */
455 	if ((mp->b_wptr - mp->b_rptr < sizeof (struct T_conn_req))) {
456 		icmp_err_ack(q, mp, TPROTO, 0);
457 		return;
458 	}
459 
460 	if (icmp->icmp_state == TS_DATA_XFER) {
461 		/* Already connected - clear out state */
462 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
463 		icmp->icmp_state = TS_IDLE;
464 	}
465 
466 
467 	if (tcr->OPT_length != 0) {
468 		icmp_err_ack(q, mp, TBADOPT, 0);
469 		return;
470 	}
471 	switch (tcr->DEST_length) {
472 	default:
473 		icmp_err_ack(q, mp, TBADADDR, 0);
474 		return;
475 
476 	case sizeof (sin_t):
477 		sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset,
478 		    sizeof (sin_t));
479 		if (sin == NULL || !OK_32PTR((char *)sin)) {
480 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
481 			return;
482 		}
483 		if (icmp->icmp_family != AF_INET ||
484 		    sin->sin_family != AF_INET) {
485 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
486 			return;
487 		}
488 		v4dst = sin->sin_addr.s_addr;
489 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
490 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
491 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
492 		    icmp->icmp_ip_snd_options_len;
493 		break;
494 
495 	case sizeof (sin6_t):
496 		sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset,
497 		    sizeof (sin6_t));
498 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
499 			icmp_err_ack(q, mp, TSYSERR, EINVAL);
500 			return;
501 		}
502 		if (icmp->icmp_family != AF_INET6 ||
503 		    sin6->sin6_family != AF_INET6) {
504 			icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
505 			return;
506 		}
507 		/* No support for mapped addresses on raw sockets */
508 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
509 			icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL);
510 			return;
511 		}
512 		v6dst = sin6->sin6_addr;
513 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
514 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
515 		flowinfo = sin6->sin6_flowinfo;
516 		break;
517 	}
518 	if (icmp->icmp_ipversion == IPV4_VERSION) {
519 		/*
520 		 * Interpret a zero destination to mean loopback.
521 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
522 		 * generate the T_CONN_CON.
523 		 */
524 		if (v4dst == INADDR_ANY) {
525 			v4dst = htonl(INADDR_LOOPBACK);
526 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
527 			if (icmp->icmp_family == AF_INET) {
528 				sin->sin_addr.s_addr = v4dst;
529 			} else {
530 				sin6->sin6_addr = v6dst;
531 			}
532 		}
533 		icmp->icmp_v6dst = v6dst;
534 		icmp->icmp_flowinfo = 0;
535 
536 		/*
537 		 * If the destination address is multicast and
538 		 * an outgoing multicast interface has been set,
539 		 * use the address of that interface as our
540 		 * source address if no source address has been set.
541 		 */
542 		if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY &&
543 		    CLASSD(v4dst) &&
544 		    icmp->icmp_multicast_if_addr != INADDR_ANY) {
545 			IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr,
546 			    &icmp->icmp_v6src);
547 		}
548 	} else {
549 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
550 		/*
551 		 * Interpret a zero destination to mean loopback.
552 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
553 		 * generate the T_CONN_CON.
554 		 */
555 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
556 			v6dst = ipv6_loopback;
557 			sin6->sin6_addr = v6dst;
558 		}
559 		icmp->icmp_v6dst = v6dst;
560 		icmp->icmp_flowinfo = flowinfo;
561 		/*
562 		 * If the destination address is multicast and
563 		 * an outgoing multicast interface has been set,
564 		 * then the ip bind logic will pick the correct source
565 		 * address (i.e. matching the outgoing multicast interface).
566 		 */
567 	}
568 
569 	/*
570 	 * Send down bind to IP to verify that there is a route
571 	 * and to determine the source address.
572 	 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput.
573 	 */
574 	if (icmp->icmp_family == AF_INET) {
575 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa_conn_t),
576 		    sin->sin_port);
577 	} else {
578 		ASSERT(icmp->icmp_family == AF_INET6);
579 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa6_conn_t),
580 		    sin6->sin6_port);
581 	}
582 	if (mp1 == NULL) {
583 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
584 		return;
585 	}
586 
587 	/*
588 	 * We also have to send a connection confirmation to
589 	 * keep TLI happy. Prepare it for icmp_rput.
590 	 */
591 	if (icmp->icmp_family == AF_INET) {
592 		mp2 = mi_tpi_conn_con(NULL, (char *)sin, sizeof (*sin), NULL,
593 		    0);
594 	} else {
595 		ASSERT(icmp->icmp_family == AF_INET6);
596 		mp2 = mi_tpi_conn_con(NULL, (char *)sin6, sizeof (*sin6), NULL,
597 		    0);
598 	}
599 	if (mp2 == NULL) {
600 		freemsg(mp1);
601 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
602 		return;
603 	}
604 
605 	mp = mi_tpi_ok_ack_alloc(mp);
606 	if (mp == NULL) {
607 		/* Unable to reuse the T_CONN_REQ for the ack. */
608 		freemsg(mp2);
609 		icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
610 		return;
611 	}
612 
613 	icmp->icmp_state = TS_DATA_XFER;
614 
615 	/* Hang onto the T_OK_ACK and T_CONN_CON for later. */
616 	linkb(mp1, mp);
617 	linkb(mp1, mp2);
618 
619 	mblk_setcred(mp1, icmp->icmp_credp);
620 	putnext(q, mp1);
621 }
622 
623 static int
624 icmp_close(queue_t *q)
625 {
626 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
627 	int	i1;
628 
629 	/* tell IP that if we're not here, he can't trust labels */
630 	if (is_system_labeled())
631 		putnext(WR(q), icmp->icmp_delabel);
632 
633 	qprocsoff(q);
634 
635 	/* If there are any options associated with the stream, free them. */
636 	if (icmp->icmp_ip_snd_options)
637 		mi_free((char *)icmp->icmp_ip_snd_options);
638 
639 	if (icmp->icmp_filter != NULL)
640 		kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t));
641 
642 	/* Free memory associated with sticky options */
643 	if (icmp->icmp_sticky_hdrs_len != 0) {
644 		kmem_free(icmp->icmp_sticky_hdrs,
645 		    icmp->icmp_sticky_hdrs_len);
646 		icmp->icmp_sticky_hdrs = NULL;
647 		icmp->icmp_sticky_hdrs_len = 0;
648 	}
649 
650 	ip6_pkt_free(&icmp->icmp_sticky_ipp);
651 
652 	crfree(icmp->icmp_credp);
653 
654 	/* Free the icmp structure and release the minor device number. */
655 	i1 = mi_close_comm(&icmp_g_head, q);
656 
657 	return (i1);
658 }
659 
660 /*
661  * This routine handles each T_DISCON_REQ message passed to icmp
662  * as an indicating that ICMP is no longer connected. This results
663  * in sending a T_BIND_REQ to IP to restore the binding to just
664  * the local address.
665  *
666  * This routine sends down a T_BIND_REQ to IP with the following mblks:
667  *	T_BIND_REQ	- specifying just the local address.
668  *	T_OK_ACK	- for the T_DISCON_REQ
669  *
670  * The disconnect completes in icmp_rput.
671  * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user.
672  * Should icmp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert
673  * it to an error ack for the appropriate primitive.
674  */
675 static void
676 icmp_disconnect(queue_t *q, mblk_t *mp)
677 {
678 	icmp_t	*icmp;
679 	mblk_t	*mp1;
680 
681 	icmp = (icmp_t *)q->q_ptr;
682 
683 	if (icmp->icmp_state != TS_DATA_XFER) {
684 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
685 		    "icmp_disconnect: bad state, %d", icmp->icmp_state);
686 		icmp_err_ack(q, mp, TOUTSTATE, 0);
687 		return;
688 	}
689 	icmp->icmp_v6src = icmp->icmp_bound_v6src;
690 	icmp->icmp_state = TS_IDLE;
691 
692 	/*
693 	 * Send down bind to IP to remove the full binding and revert
694 	 * to the local address binding.
695 	 */
696 	if (icmp->icmp_family == AF_INET) {
697 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin_t), 0);
698 	} else {
699 		ASSERT(icmp->icmp_family == AF_INET6);
700 		mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin6_t), 0);
701 	}
702 	if (mp1 == NULL) {
703 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
704 		return;
705 	}
706 	mp = mi_tpi_ok_ack_alloc(mp);
707 	if (mp == NULL) {
708 		/* Unable to reuse the T_DISCON_REQ for the ack. */
709 		icmp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM);
710 		return;
711 	}
712 
713 	if (icmp->icmp_family == AF_INET6) {
714 		int error;
715 
716 		/* Rebuild the header template */
717 		error = icmp_build_hdrs(q, icmp);
718 		if (error != 0) {
719 			icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error);
720 			freemsg(mp1);
721 			return;
722 		}
723 	}
724 	icmp->icmp_discon_pending = 1;
725 
726 	/* Append the T_OK_ACK to the T_BIND_REQ for icmp_rput */
727 	linkb(mp1, mp);
728 	putnext(q, mp1);
729 }
730 
731 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
732 static void
733 icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
734 {
735 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
736 		qreply(q, mp);
737 }
738 
739 /* Shorthand to generate and send TPI error acks to our client */
740 static void
741 icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
742     t_scalar_t t_error, int sys_error)
743 {
744 	struct T_error_ack	*teackp;
745 
746 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
747 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
748 		teackp = (struct T_error_ack *)mp->b_rptr;
749 		teackp->ERROR_prim = primitive;
750 		teackp->TLI_error = t_error;
751 		teackp->UNIX_error = sys_error;
752 		qreply(q, mp);
753 	}
754 }
755 
756 /*
757  * icmp_icmp_error is called by icmp_rput to process ICMP
758  * messages passed up by IP.
759  * Generates the appropriate T_UDERROR_IND for permanent
760  * (non-transient) errors.
761  * Assumes that IP has pulled up everything up to and including
762  * the ICMP header.
763  */
764 static void
765 icmp_icmp_error(queue_t *q, mblk_t *mp)
766 {
767 	icmph_t *icmph;
768 	ipha_t	*ipha;
769 	int	iph_hdr_length;
770 	sin_t	sin;
771 	sin6_t	sin6;
772 	mblk_t	*mp1;
773 	int	error = 0;
774 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
775 
776 	/*
777 	 * Deliver T_UDERROR_IND when the application has asked for it.
778 	 * The socket layer enables this automatically when connected.
779 	 */
780 	if (!icmp->icmp_dgram_errind) {
781 		freemsg(mp);
782 		return;
783 	}
784 
785 	ipha = (ipha_t *)mp->b_rptr;
786 
787 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
788 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
789 		icmp_icmp_error_ipv6(q, mp);
790 		return;
791 	}
792 	ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
793 
794 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
795 	icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]);
796 	ipha = (ipha_t *)&icmph[1];
797 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
798 
799 	switch (icmph->icmph_type) {
800 	case ICMP_DEST_UNREACHABLE:
801 		switch (icmph->icmph_code) {
802 		case ICMP_FRAGMENTATION_NEEDED:
803 			/*
804 			 * IP has already adjusted the path MTU.
805 			 * XXX Somehow pass MTU indication to application?
806 			 */
807 			break;
808 		case ICMP_PORT_UNREACHABLE:
809 		case ICMP_PROTOCOL_UNREACHABLE:
810 			error = ECONNREFUSED;
811 			break;
812 		default:
813 			/* Transient errors */
814 			break;
815 		}
816 		break;
817 	default:
818 		/* Transient errors */
819 		break;
820 	}
821 	if (error == 0) {
822 		freemsg(mp);
823 		return;
824 	}
825 
826 	switch (icmp->icmp_family) {
827 	case AF_INET:
828 		sin = sin_null;
829 		sin.sin_family = AF_INET;
830 		sin.sin_addr.s_addr = ipha->ipha_dst;
831 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0,
832 		    error);
833 		break;
834 	case AF_INET6:
835 		sin6 = sin6_null;
836 		sin6.sin6_family = AF_INET6;
837 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
838 
839 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
840 		    NULL, 0, error);
841 		break;
842 	}
843 	if (mp1)
844 		putnext(q, mp1);
845 	freemsg(mp);
846 }
847 
848 /*
849  * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6
850  * for IPv6 packets.
851  * Send permanent (non-transient) errors upstream.
852  * Assumes that IP has pulled up all the extension headers as well
853  * as the ICMPv6 header.
854  */
855 static void
856 icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp)
857 {
858 	icmp6_t		*icmp6;
859 	ip6_t		*ip6h, *outer_ip6h;
860 	uint16_t	iph_hdr_length;
861 	uint8_t		*nexthdrp;
862 	sin6_t		sin6;
863 	mblk_t		*mp1;
864 	int		error = 0;
865 	icmp_t		*icmp = (icmp_t *)q->q_ptr;
866 
867 	outer_ip6h = (ip6_t *)mp->b_rptr;
868 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
869 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
870 	else
871 		iph_hdr_length = IPV6_HDR_LEN;
872 
873 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
874 	ip6h = (ip6_t *)&icmp6[1];
875 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
876 		freemsg(mp);
877 		return;
878 	}
879 	if (*nexthdrp != icmp->icmp_proto) {
880 		/*
881 		 * Could have switched icmp_proto after while ip did fanout of
882 		 * this message
883 		 */
884 		freemsg(mp);
885 		return;
886 	}
887 	switch (icmp6->icmp6_type) {
888 	case ICMP6_DST_UNREACH:
889 		switch (icmp6->icmp6_code) {
890 		case ICMP6_DST_UNREACH_NOPORT:
891 			error = ECONNREFUSED;
892 			break;
893 		case ICMP6_DST_UNREACH_ADMIN:
894 		case ICMP6_DST_UNREACH_NOROUTE:
895 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
896 		case ICMP6_DST_UNREACH_ADDR:
897 			/* Transient errors */
898 			break;
899 		default:
900 			break;
901 		}
902 		break;
903 	case ICMP6_PACKET_TOO_BIG: {
904 		struct T_unitdata_ind	*tudi;
905 		struct T_opthdr		*toh;
906 		size_t			udi_size;
907 		mblk_t			*newmp;
908 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
909 		    sizeof (struct ip6_mtuinfo);
910 		sin6_t			*sin6;
911 		struct ip6_mtuinfo	*mtuinfo;
912 
913 		/*
914 		 * If the application has requested to receive path mtu
915 		 * information, send up an empty message containing an
916 		 * IPV6_PATHMTU ancillary data item.
917 		 */
918 		if (!icmp->icmp_ipv6_recvpathmtu)
919 			break;
920 
921 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
922 		    opt_length;
923 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
924 			BUMP_MIB(&rawip_mib, rawipInErrors);
925 			break;
926 		}
927 
928 		/*
929 		 * newmp->b_cont is left to NULL on purpose.  This is an
930 		 * empty message containing only ancillary data.
931 		 */
932 		newmp->b_datap->db_type = M_PROTO;
933 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
934 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
935 		tudi->PRIM_type = T_UNITDATA_IND;
936 		tudi->SRC_length = sizeof (sin6_t);
937 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
938 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
939 		tudi->OPT_length = opt_length;
940 
941 		sin6 = (sin6_t *)&tudi[1];
942 		bzero(sin6, sizeof (sin6_t));
943 		sin6->sin6_family = AF_INET6;
944 		sin6->sin6_addr = icmp->icmp_v6dst;
945 
946 		toh = (struct T_opthdr *)&sin6[1];
947 		toh->level = IPPROTO_IPV6;
948 		toh->name = IPV6_PATHMTU;
949 		toh->len = opt_length;
950 		toh->status = 0;
951 
952 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
953 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
954 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
955 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
956 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
957 		/*
958 		 * We've consumed everything we need from the original
959 		 * message.  Free it, then send our empty message.
960 		 */
961 		freemsg(mp);
962 		putnext(q, newmp);
963 		return;
964 	}
965 	case ICMP6_TIME_EXCEEDED:
966 		/* Transient errors */
967 		break;
968 	case ICMP6_PARAM_PROB:
969 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
970 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
971 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
972 		    (uchar_t *)nexthdrp) {
973 			error = ECONNREFUSED;
974 			break;
975 		}
976 		break;
977 	}
978 	if (error == 0) {
979 		freemsg(mp);
980 		return;
981 	}
982 
983 	sin6 = sin6_null;
984 	sin6.sin6_family = AF_INET6;
985 	sin6.sin6_addr = ip6h->ip6_dst;
986 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
987 
988 	mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0,
989 	    error);
990 	if (mp1)
991 		putnext(q, mp1);
992 	freemsg(mp);
993 }
994 
995 /*
996  * This routine responds to T_ADDR_REQ messages.  It is called by icmp_wput.
997  * The local address is filled in if endpoint is bound. The remote address
998  * is filled in if remote address has been precified ("connected endpoint")
999  * (The concept of connected CLTS sockets is alien to published TPI
1000  *  but we support it anyway).
1001  */
1002 static void
1003 icmp_addr_req(queue_t *q, mblk_t *mp)
1004 {
1005 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
1006 	mblk_t	*ackmp;
1007 	struct T_addr_ack *taa;
1008 
1009 	/* Make it large enough for worst case */
1010 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1011 	    2 * sizeof (sin6_t), 1);
1012 	if (ackmp == NULL) {
1013 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
1014 		return;
1015 	}
1016 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1017 
1018 	bzero(taa, sizeof (struct T_addr_ack));
1019 	ackmp->b_wptr = (uchar_t *)&taa[1];
1020 
1021 	taa->PRIM_type = T_ADDR_ACK;
1022 	ackmp->b_datap->db_type = M_PCPROTO;
1023 
1024 	/*
1025 	 * Note: Following code assumes 32 bit alignment of basic
1026 	 * data structures like sin_t and struct T_addr_ack.
1027 	 */
1028 	if (icmp->icmp_state != TS_UNBND) {
1029 		/*
1030 		 * Fill in local address
1031 		 */
1032 		taa->LOCADDR_offset = sizeof (*taa);
1033 		if (icmp->icmp_family == AF_INET) {
1034 			sin_t	*sin;
1035 
1036 			taa->LOCADDR_length = sizeof (sin_t);
1037 			sin = (sin_t *)&taa[1];
1038 			/* Fill zeroes and then intialize non-zero fields */
1039 			*sin = sin_null;
1040 			sin->sin_family = AF_INET;
1041 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
1042 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1043 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src,
1044 				    sin->sin_addr.s_addr);
1045 			} else {
1046 				/*
1047 				 * INADDR_ANY
1048 				 * icmp_v6src is not set, we might be bound to
1049 				 * broadcast/multicast. Use icmp_bound_v6src as
1050 				 * local address instead (that could
1051 				 * also still be INADDR_ANY)
1052 				 */
1053 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src,
1054 				    sin->sin_addr.s_addr);
1055 			}
1056 			ackmp->b_wptr = (uchar_t *)&sin[1];
1057 		} else {
1058 			sin6_t	*sin6;
1059 
1060 			ASSERT(icmp->icmp_family == AF_INET6);
1061 			taa->LOCADDR_length = sizeof (sin6_t);
1062 			sin6 = (sin6_t *)&taa[1];
1063 			/* Fill zeroes and then intialize non-zero fields */
1064 			*sin6 = sin6_null;
1065 			sin6->sin6_family = AF_INET6;
1066 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1067 				sin6->sin6_addr = icmp->icmp_v6src;
1068 			} else {
1069 				/*
1070 				 * UNSPECIFIED
1071 				 * icmp_v6src is not set, we might be bound to
1072 				 * broadcast/multicast. Use icmp_bound_v6src as
1073 				 * local address instead (that could
1074 				 * also still be UNSPECIFIED)
1075 				 */
1076 				sin6->sin6_addr = icmp->icmp_bound_v6src;
1077 			}
1078 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1079 		}
1080 	}
1081 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1082 	qreply(q, ackmp);
1083 }
1084 
1085 static void
1086 icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp)
1087 {
1088 	*tap = icmp_g_t_info_ack;
1089 
1090 	if (icmp->icmp_family == AF_INET6)
1091 		tap->ADDR_size = sizeof (sin6_t);
1092 	else
1093 		tap->ADDR_size = sizeof (sin_t);
1094 	tap->CURRENT_state = icmp->icmp_state;
1095 	tap->OPT_size = icmp_max_optsize;
1096 }
1097 
1098 /*
1099  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1100  * icmp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1101  * icmp_g_t_info_ack.  The current state of the stream is copied from
1102  * icmp_state.
1103  */
1104 static void
1105 icmp_capability_req(queue_t *q, mblk_t *mp)
1106 {
1107 	icmp_t			*icmp = (icmp_t *)q->q_ptr;
1108 	t_uscalar_t		cap_bits1;
1109 	struct T_capability_ack	*tcap;
1110 
1111 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1112 
1113 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1114 		mp->b_datap->db_type, T_CAPABILITY_ACK);
1115 	if (!mp)
1116 		return;
1117 
1118 	tcap = (struct T_capability_ack *)mp->b_rptr;
1119 	tcap->CAP_bits1 = 0;
1120 
1121 	if (cap_bits1 & TC1_INFO) {
1122 		icmp_copy_info(&tcap->INFO_ack, icmp);
1123 		tcap->CAP_bits1 |= TC1_INFO;
1124 	}
1125 
1126 	qreply(q, mp);
1127 }
1128 
1129 /*
1130  * This routine responds to T_INFO_REQ messages.  It is called by icmp_wput.
1131  * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack.
1132  * The current state of the stream is copied from icmp_state.
1133  */
1134 static void
1135 icmp_info_req(queue_t *q, mblk_t *mp)
1136 {
1137 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
1138 
1139 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1140 	    T_INFO_ACK);
1141 	if (!mp)
1142 		return;
1143 	icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp);
1144 	qreply(q, mp);
1145 }
1146 
1147 /*
1148  * IP recognizes seven kinds of bind requests:
1149  *
1150  * - A zero-length address binds only to the protocol number.
1151  *
1152  * - A 4-byte address is treated as a request to
1153  * validate that the address is a valid local IPv4
1154  * address, appropriate for an application to bind to.
1155  * IP does the verification, but does not make any note
1156  * of the address at this time.
1157  *
1158  * - A 16-byte address contains is treated as a request
1159  * to validate a local IPv6 address, as the 4-byte
1160  * address case above.
1161  *
1162  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
1163  * use it for the inbound fanout of packets.
1164  *
1165  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
1166  * use it for the inbound fanout of packets.
1167  *
1168  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
1169  * information consisting of local and remote addresses
1170  * and ports (unused for raw sockets).  In this case, the addresses are both
1171  * validated as appropriate for this operation, and, if
1172  * so, the information is retained for use in the
1173  * inbound fanout.
1174  *
1175  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
1176  * fanout information, like the 12-byte case above.
1177  *
1178  * IP will also fill in the IRE request mblk with information
1179  * regarding our peer.  In all cases, we notify IP of our protocol
1180  * type by appending a single protocol byte to the bind request.
1181  */
1182 static mblk_t *
1183 icmp_ip_bind_mp(icmp_t *icmp, t_scalar_t bind_prim, t_scalar_t addr_length,
1184     in_port_t fport)
1185 {
1186 	char	*cp;
1187 	mblk_t	*mp;
1188 	struct T_bind_req *tbr;
1189 	ipa_conn_t	*ac;
1190 	ipa6_conn_t	*ac6;
1191 	sin_t		*sin;
1192 	sin6_t		*sin6;
1193 
1194 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
1195 
1196 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
1197 	if (mp == NULL)
1198 		return (NULL);
1199 	mp->b_datap->db_type = M_PROTO;
1200 	tbr = (struct T_bind_req *)mp->b_rptr;
1201 	tbr->PRIM_type = bind_prim;
1202 	tbr->ADDR_offset = sizeof (*tbr);
1203 	tbr->CONIND_number = 0;
1204 	tbr->ADDR_length = addr_length;
1205 	cp = (char *)&tbr[1];
1206 	switch (addr_length) {
1207 	case sizeof (ipa_conn_t):
1208 		ASSERT(icmp->icmp_family == AF_INET);
1209 		/* Append a request for an IRE */
1210 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1211 		if (mp->b_cont == NULL) {
1212 			freemsg(mp);
1213 			return (NULL);
1214 		}
1215 		mp->b_cont->b_wptr += sizeof (ire_t);
1216 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1217 
1218 		/* cp known to be 32 bit aligned */
1219 		ac = (ipa_conn_t *)cp;
1220 		ac->ac_laddr = V4_PART_OF_V6(icmp->icmp_v6src);
1221 		ac->ac_faddr = V4_PART_OF_V6(icmp->icmp_v6dst);
1222 		ac->ac_fport = fport;
1223 		ac->ac_lport = 0;
1224 		break;
1225 
1226 	case sizeof (ipa6_conn_t):
1227 		ASSERT(icmp->icmp_family == AF_INET6);
1228 		/* Append a request for an IRE */
1229 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1230 		if (mp->b_cont == NULL) {
1231 			freemsg(mp);
1232 			return (NULL);
1233 		}
1234 		mp->b_cont->b_wptr += sizeof (ire_t);
1235 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1236 
1237 		/* cp known to be 32 bit aligned */
1238 		ac6 = (ipa6_conn_t *)cp;
1239 		ac6->ac6_laddr = icmp->icmp_v6src;
1240 		ac6->ac6_faddr = icmp->icmp_v6dst;
1241 		ac6->ac6_fport = fport;
1242 		ac6->ac6_lport = 0;
1243 		break;
1244 
1245 	case sizeof (sin_t):
1246 		ASSERT(icmp->icmp_family == AF_INET);
1247 		/* Append a request for an IRE */
1248 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1249 		if (!mp->b_cont) {
1250 			freemsg(mp);
1251 			return (NULL);
1252 		}
1253 		mp->b_cont->b_wptr += sizeof (ire_t);
1254 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1255 
1256 		sin = (sin_t *)cp;
1257 		*sin = sin_null;
1258 		sin->sin_family = AF_INET;
1259 		sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_bound_v6src);
1260 		break;
1261 
1262 	case sizeof (sin6_t):
1263 		ASSERT(icmp->icmp_family == AF_INET6);
1264 		/* Append a request for an IRE */
1265 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1266 		if (!mp->b_cont) {
1267 			freemsg(mp);
1268 			return (NULL);
1269 		}
1270 		mp->b_cont->b_wptr += sizeof (ire_t);
1271 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1272 
1273 		sin6 = (sin6_t *)cp;
1274 		*sin6 = sin6_null;
1275 		sin6->sin6_family = AF_INET6;
1276 		sin6->sin6_addr = icmp->icmp_bound_v6src;
1277 		break;
1278 	}
1279 	/* Add protocol number to end */
1280 	cp[addr_length] = icmp->icmp_proto;
1281 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
1282 	return (mp);
1283 }
1284 
1285 /* ARGSUSED */
1286 static void
1287 dummy_func(void *arg)
1288 {
1289 }
1290 
1291 static mblk_t *
1292 alloc_wait(queue_t *q, size_t len, int pri, int *errp)
1293 {
1294 	mblk_t *mp;
1295 	bufcall_id_t id;
1296 	int retv;
1297 
1298 	while ((mp = allocb(len, pri)) == NULL) {
1299 		id = qbufcall(q, len, pri, dummy_func, NULL);
1300 		if (id == 0) {
1301 			*errp = ENOMEM;
1302 			break;
1303 		}
1304 		retv = qwait_sig(q);
1305 		qunbufcall(q, id);
1306 		if (retv == 0) {
1307 			*errp = EINTR;
1308 			break;
1309 		}
1310 	}
1311 	if (mp != NULL)
1312 		mp->b_wptr += len;
1313 	return (mp);
1314 }
1315 
1316 /*
1317  * This is the open routine for icmp.  It allocates a icmp_t structure for
1318  * the stream and, on the first open of the module, creates an ND table.
1319  */
1320 static int
1321 icmp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1322 {
1323 	int	err;
1324 	icmp_t	*icmp;
1325 	mblk_t	*mp;
1326 	out_labeled_t *olp;
1327 
1328 	/* If the stream is already open, return immediately. */
1329 	if (q->q_ptr != NULL)
1330 		return (0);
1331 
1332 	/* If this is not a push of icmp as a module, fail. */
1333 	if (sflag != MODOPEN)
1334 		return (EINVAL);
1335 
1336 	/*
1337 	 * Defer the qprocson until everything is initialized since
1338 	 * we are D_MTPERQ and after qprocson the rput routine can
1339 	 * run. (Could do qprocson earlier since icmp currently
1340 	 * has an outer perimeter.)
1341 	 */
1342 
1343 	/*
1344 	 * Create a icmp_t structure for this stream and link into the
1345 	 * list of open streams.
1346 	 */
1347 	err = mi_open_comm(&icmp_g_head, sizeof (icmp_t), q, devp,
1348 	    flag, sflag, credp);
1349 	if (err != 0)
1350 		return (err);
1351 
1352 	/*
1353 	 * The receive hiwat is only looked at on the stream head queue.
1354 	 * Store in q_hiwat in order to return on SO_RCVBUF getsockopts.
1355 	 */
1356 	q->q_hiwat = icmp_recv_hiwat;
1357 
1358 	/* Set the initial state of the stream and the privilege status. */
1359 	icmp = (icmp_t *)q->q_ptr;
1360 	icmp->icmp_state = TS_UNBND;
1361 	icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1362 	icmp->icmp_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1363 	icmp->icmp_filter = NULL;
1364 
1365 	icmp->icmp_credp = credp;
1366 	crhold(credp);
1367 
1368 	/*
1369 	 * If the caller has the process-wide flag set, then default to MAC
1370 	 * exempt mode.  This allows read-down to unlabeled hosts.
1371 	 */
1372 	if (getpflags(NET_MAC_AWARE, credp) != 0)
1373 		icmp->icmp_mac_exempt = B_TRUE;
1374 
1375 	icmp->icmp_zoneid = getzoneid();
1376 
1377 	if (getmajor(*devp) == (major_t)ICMP6_MAJ) {
1378 		icmp->icmp_ipversion = IPV6_VERSION;
1379 		icmp->icmp_family = AF_INET6;
1380 		/* May be changed by a SO_PROTOTYPE socket option. */
1381 		icmp->icmp_proto = IPPROTO_ICMPV6;
1382 		icmp->icmp_checksum_off = 2;	/* Offset for icmp6_cksum */
1383 		icmp->icmp_max_hdr_len = IPV6_HDR_LEN;
1384 		icmp->icmp_ttl = (uint8_t)icmp_ipv6_hoplimit;
1385 	} else {
1386 		icmp->icmp_ipversion = IPV4_VERSION;
1387 		icmp->icmp_family = AF_INET;
1388 		/* May be changed by a SO_PROTOTYPE socket option. */
1389 		icmp->icmp_proto = IPPROTO_ICMP;
1390 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH;
1391 		icmp->icmp_ttl = (uint8_t)icmp_ipv4_ttl;
1392 	}
1393 	qprocson(q);
1394 
1395 	/*
1396 	 * Check if icmp is being I_PUSHed by a non-privileged user.
1397 	 * If so, we set icmp_restricted to indicate that only MIB
1398 	 * traffic may pass.
1399 	 */
1400 	if (secpolicy_net_icmpaccess(credp) != 0) {
1401 		icmp->icmp_restricted = 1;
1402 	}
1403 
1404 	/*
1405 	 * The transmit hiwat is only looked at on IP's queue.
1406 	 * Store in q_hiwat in order to return on SO_SNDBUF
1407 	 * getsockopts.
1408 	 */
1409 	WR(q)->q_hiwat = icmp_xmit_hiwat;
1410 	WR(q)->q_next->q_hiwat = WR(q)->q_hiwat;
1411 	WR(q)->q_lowat = icmp_xmit_lowat;
1412 	WR(q)->q_next->q_lowat = WR(q)->q_lowat;
1413 
1414 	if (icmp->icmp_family == AF_INET6) {
1415 		/* Build initial header template for transmit */
1416 		err = icmp_build_hdrs(q, icmp);
1417 		if (err != 0)
1418 			goto open_error;
1419 	}
1420 	/* Set the Stream head write offset. */
1421 	(void) mi_set_sth_wroff(q, icmp->icmp_max_hdr_len + icmp_wroff_extra);
1422 	(void) mi_set_sth_hiwat(q, q->q_hiwat);
1423 
1424 	if (is_system_labeled()) {
1425 		/* notify IP that we know about labeling */
1426 		mp = alloc_wait(q, sizeof (*olp), BPRI_MED, &err);
1427 		if (mp == NULL)
1428 			goto open_error;
1429 		mp->b_datap->db_type = M_CTL;
1430 		olp = (out_labeled_t *)mp->b_rptr;
1431 		olp->out_labeled_type = IP_ULP_OUT_LABELED;
1432 		olp->out_qnext = WR(q)->q_next;
1433 		putnext(WR(q), mp);
1434 
1435 		/* save off a copy for closing */
1436 		mp = alloc_wait(q, sizeof (*olp), BPRI_MED, &err);
1437 		if (mp == NULL)
1438 			goto open_error;
1439 		mp->b_datap->db_type = M_CTL;
1440 		olp = (out_labeled_t *)mp->b_rptr;
1441 		olp->out_labeled_type = IP_ULP_OUT_LABELED;
1442 		olp->out_qnext = NULL;
1443 		icmp->icmp_delabel = mp;
1444 	}
1445 
1446 	return (0);
1447 
1448 open_error:
1449 	qprocsoff(q);
1450 	crfree(credp);
1451 	(void) mi_close_comm(&icmp_g_head, q);
1452 	return (err);
1453 }
1454 
1455 /*
1456  * Which ICMP options OK to set through T_UNITDATA_REQ...
1457  */
1458 /* ARGSUSED */
1459 static boolean_t
1460 icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1461 {
1462 	return (B_TRUE);
1463 }
1464 
1465 /*
1466  * This routine gets default values of certain options whose default
1467  * values are maintained by protcol specific code
1468  */
1469 /* ARGSUSED */
1470 int
1471 icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr)
1472 {
1473 	int *i1 = (int *)ptr;
1474 
1475 	switch (level) {
1476 	case IPPROTO_IP:
1477 		switch (name) {
1478 		case IP_MULTICAST_TTL:
1479 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1480 			return (sizeof (uchar_t));
1481 		case IP_MULTICAST_LOOP:
1482 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1483 			return (sizeof (uchar_t));
1484 		}
1485 		break;
1486 	case IPPROTO_IPV6:
1487 		switch (name) {
1488 		case IPV6_MULTICAST_HOPS:
1489 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1490 			return (sizeof (int));
1491 		case IPV6_MULTICAST_LOOP:
1492 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1493 			return (sizeof (int));
1494 		case IPV6_UNICAST_HOPS:
1495 			*i1 = icmp_ipv6_hoplimit;
1496 			return (sizeof (int));
1497 		}
1498 		break;
1499 	case IPPROTO_ICMPV6:
1500 		switch (name) {
1501 		case ICMP6_FILTER:
1502 			/* Make it look like "pass all" */
1503 			ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1504 			return (sizeof (icmp6_filter_t));
1505 		}
1506 		break;
1507 	}
1508 	return (-1);
1509 }
1510 
1511 /*
1512  * This routine retrieves the current status of socket options.
1513  * It returns the size of the option retrieved.
1514  */
1515 int
1516 icmp_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
1517 {
1518 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
1519 	int	*i1 = (int *)ptr;
1520 	ip6_pkt_t	*ipp = &icmp->icmp_sticky_ipp;
1521 
1522 	switch (level) {
1523 	case SOL_SOCKET:
1524 		switch (name) {
1525 		case SO_DEBUG:
1526 			*i1 = icmp->icmp_debug;
1527 			break;
1528 		case SO_TYPE:
1529 			*i1 = SOCK_RAW;
1530 			break;
1531 		case SO_PROTOTYPE:
1532 			*i1 = icmp->icmp_proto;
1533 			break;
1534 		case SO_REUSEADDR:
1535 			*i1 = icmp->icmp_reuseaddr;
1536 			break;
1537 
1538 		/*
1539 		 * The following three items are available here,
1540 		 * but are only meaningful to IP.
1541 		 */
1542 		case SO_DONTROUTE:
1543 			*i1 = icmp->icmp_dontroute;
1544 			break;
1545 		case SO_USELOOPBACK:
1546 			*i1 = icmp->icmp_useloopback;
1547 			break;
1548 		case SO_BROADCAST:
1549 			*i1 = icmp->icmp_broadcast;
1550 			break;
1551 
1552 		case SO_SNDBUF:
1553 			ASSERT(q->q_hiwat <= INT_MAX);
1554 			*i1 = (int)q->q_hiwat;
1555 			break;
1556 		case SO_RCVBUF:
1557 			ASSERT(RD(q)->q_hiwat <= INT_MAX);
1558 			*i1 = (int)RD(q)->q_hiwat;
1559 			break;
1560 		case SO_DGRAM_ERRIND:
1561 			*i1 = icmp->icmp_dgram_errind;
1562 			break;
1563 		case SO_TIMESTAMP:
1564 			*i1 = icmp->icmp_timestamp;
1565 			break;
1566 		case SO_MAC_EXEMPT:
1567 			*i1 = icmp->icmp_mac_exempt;
1568 			break;
1569 		/*
1570 		 * Following three not meaningful for icmp
1571 		 * Action is same as "default" to which we fallthrough
1572 		 * so we keep them in comments.
1573 		 * case SO_LINGER:
1574 		 * case SO_KEEPALIVE:
1575 		 * case SO_OOBINLINE:
1576 		 */
1577 		default:
1578 			return (-1);
1579 		}
1580 		break;
1581 	case IPPROTO_IP:
1582 		/*
1583 		 * Only allow IPv4 option processing on IPv4 sockets.
1584 		 */
1585 		if (icmp->icmp_family != AF_INET)
1586 			return (-1);
1587 
1588 		switch (name) {
1589 		case IP_OPTIONS:
1590 		case T_IP_OPTIONS:
1591 			/* Options are passed up with each packet */
1592 			return (0);
1593 		case IP_HDRINCL:
1594 			*i1 = (int)icmp->icmp_hdrincl;
1595 			break;
1596 		case IP_TOS:
1597 		case T_IP_TOS:
1598 			*i1 = (int)icmp->icmp_type_of_service;
1599 			break;
1600 		case IP_TTL:
1601 			*i1 = (int)icmp->icmp_ttl;
1602 			break;
1603 		case IP_MULTICAST_IF:
1604 			/* 0 address if not set */
1605 			*(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr;
1606 			return (sizeof (ipaddr_t));
1607 		case IP_MULTICAST_TTL:
1608 			*(uchar_t *)ptr = icmp->icmp_multicast_ttl;
1609 			return (sizeof (uchar_t));
1610 		case IP_MULTICAST_LOOP:
1611 			*ptr = icmp->icmp_multicast_loop;
1612 			return (sizeof (uint8_t));
1613 		case IP_BOUND_IF:
1614 			/* Zero if not set */
1615 			*i1 = icmp->icmp_bound_if;
1616 			break;	/* goto sizeof (int) option return */
1617 		case IP_UNSPEC_SRC:
1618 			*ptr = icmp->icmp_unspec_source;
1619 			break;	/* goto sizeof (int) option return */
1620 		case IP_XMIT_IF:
1621 			*i1 = icmp->icmp_xmit_if;
1622 			break;	/* goto sizeof (int) option return */
1623 		case IP_RECVIF:
1624 			*ptr = icmp->icmp_recvif;
1625 			break;	/* goto sizeof (int) option return */
1626 		/*
1627 		 * Cannot "get" the value of following options
1628 		 * at this level. Action is same as "default" to
1629 		 * which we fallthrough so we keep them in comments.
1630 		 *
1631 		 * case IP_ADD_MEMBERSHIP:
1632 		 * case IP_DROP_MEMBERSHIP:
1633 		 * case IP_BLOCK_SOURCE:
1634 		 * case IP_UNBLOCK_SOURCE:
1635 		 * case IP_ADD_SOURCE_MEMBERSHIP:
1636 		 * case IP_DROP_SOURCE_MEMBERSHIP:
1637 		 * case MCAST_JOIN_GROUP:
1638 		 * case MCAST_LEAVE_GROUP:
1639 		 * case MCAST_BLOCK_SOURCE:
1640 		 * case MCAST_UNBLOCK_SOURCE:
1641 		 * case MCAST_JOIN_SOURCE_GROUP:
1642 		 * case MCAST_LEAVE_SOURCE_GROUP:
1643 		 * case MRT_INIT:
1644 		 * case MRT_DONE:
1645 		 * case MRT_ADD_VIF:
1646 		 * case MRT_DEL_VIF:
1647 		 * case MRT_ADD_MFC:
1648 		 * case MRT_DEL_MFC:
1649 		 * case MRT_VERSION:
1650 		 * case MRT_ASSERT:
1651 		 * case IP_SEC_OPT:
1652 		 * case IP_DONTFAILOVER_IF:
1653 		 * case IP_NEXTHOP:
1654 		 */
1655 		default:
1656 			return (-1);
1657 		}
1658 		break;
1659 	case IPPROTO_IPV6:
1660 		/*
1661 		 * Only allow IPv6 option processing on native IPv6 sockets.
1662 		 */
1663 		if (icmp->icmp_family != AF_INET6)
1664 			return (-1);
1665 		switch (name) {
1666 		case IPV6_UNICAST_HOPS:
1667 			*i1 = (unsigned int)icmp->icmp_ttl;
1668 			break;
1669 		case IPV6_MULTICAST_IF:
1670 			/* 0 index if not set */
1671 			*i1 = icmp->icmp_multicast_if_index;
1672 			break;
1673 		case IPV6_MULTICAST_HOPS:
1674 			*i1 = icmp->icmp_multicast_ttl;
1675 			break;
1676 		case IPV6_MULTICAST_LOOP:
1677 			*i1 = icmp->icmp_multicast_loop;
1678 			break;
1679 		case IPV6_BOUND_IF:
1680 			/* Zero if not set */
1681 			*i1 = icmp->icmp_bound_if;
1682 			break;
1683 		case IPV6_UNSPEC_SRC:
1684 			*i1 = icmp->icmp_unspec_source;
1685 			break;
1686 		case IPV6_CHECKSUM:
1687 			/*
1688 			 * Return offset or -1 if no checksum offset.
1689 			 * Does not apply to IPPROTO_ICMPV6
1690 			 */
1691 			if (icmp->icmp_proto == IPPROTO_ICMPV6)
1692 				return (-1);
1693 
1694 			if (icmp->icmp_raw_checksum) {
1695 				*i1 = icmp->icmp_checksum_off;
1696 			} else {
1697 				*i1 = -1;
1698 			}
1699 			break;
1700 		case IPV6_JOIN_GROUP:
1701 		case IPV6_LEAVE_GROUP:
1702 		case MCAST_JOIN_GROUP:
1703 		case MCAST_LEAVE_GROUP:
1704 		case MCAST_BLOCK_SOURCE:
1705 		case MCAST_UNBLOCK_SOURCE:
1706 		case MCAST_JOIN_SOURCE_GROUP:
1707 		case MCAST_LEAVE_SOURCE_GROUP:
1708 			/* cannot "get" the value for these */
1709 			return (-1);
1710 		case IPV6_RECVPKTINFO:
1711 			*i1 = icmp->icmp_ipv6_recvpktinfo;
1712 			break;
1713 		case IPV6_RECVTCLASS:
1714 			*i1 = icmp->icmp_ipv6_recvtclass;
1715 			break;
1716 		case IPV6_RECVPATHMTU:
1717 			*i1 = icmp->icmp_ipv6_recvpathmtu;
1718 			break;
1719 		case IPV6_V6ONLY:
1720 			*i1 = 1;
1721 			break;
1722 		case IPV6_RECVHOPLIMIT:
1723 			*i1 = icmp->icmp_ipv6_recvhoplimit;
1724 			break;
1725 		case IPV6_RECVHOPOPTS:
1726 			*i1 = icmp->icmp_ipv6_recvhopopts;
1727 			break;
1728 		case IPV6_RECVDSTOPTS:
1729 			*i1 = icmp->icmp_ipv6_recvdstopts;
1730 			break;
1731 		case _OLD_IPV6_RECVDSTOPTS:
1732 			*i1 = icmp->icmp_old_ipv6_recvdstopts;
1733 			break;
1734 		case IPV6_RECVRTHDRDSTOPTS:
1735 			*i1 = icmp->icmp_ipv6_recvrtdstopts;
1736 			break;
1737 		case IPV6_RECVRTHDR:
1738 			*i1 = icmp->icmp_ipv6_recvrthdr;
1739 			break;
1740 		case IPV6_PKTINFO: {
1741 			/* XXX assumes that caller has room for max size! */
1742 			struct in6_pktinfo *pkti;
1743 
1744 			pkti = (struct in6_pktinfo *)ptr;
1745 			if (ipp->ipp_fields & IPPF_IFINDEX)
1746 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
1747 			else
1748 				pkti->ipi6_ifindex = 0;
1749 			if (ipp->ipp_fields & IPPF_ADDR)
1750 				pkti->ipi6_addr = ipp->ipp_addr;
1751 			else
1752 				pkti->ipi6_addr = ipv6_all_zeros;
1753 			return (sizeof (struct in6_pktinfo));
1754 		}
1755 		case IPV6_NEXTHOP: {
1756 			sin6_t *sin6 = (sin6_t *)ptr;
1757 
1758 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
1759 				return (0);
1760 			*sin6 = sin6_null;
1761 			sin6->sin6_family = AF_INET6;
1762 			sin6->sin6_addr = ipp->ipp_nexthop;
1763 			return (sizeof (sin6_t));
1764 		}
1765 		case IPV6_HOPOPTS:
1766 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
1767 				return (0);
1768 			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
1769 				return (0);
1770 			bcopy((char *)ipp->ipp_hopopts +
1771 			    icmp->icmp_label_len_v6, ptr,
1772 			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1773 			if (icmp->icmp_label_len_v6 > 0) {
1774 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
1775 				ptr[1] = (ipp->ipp_hopoptslen -
1776 				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
1777 			}
1778 			return (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
1779 		case IPV6_RTHDRDSTOPTS:
1780 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
1781 				return (0);
1782 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
1783 			return (ipp->ipp_rtdstoptslen);
1784 		case IPV6_RTHDR:
1785 			if (!(ipp->ipp_fields & IPPF_RTHDR))
1786 				return (0);
1787 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
1788 			return (ipp->ipp_rthdrlen);
1789 		case IPV6_DSTOPTS:
1790 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
1791 				return (0);
1792 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
1793 			return (ipp->ipp_dstoptslen);
1794 		case IPV6_PATHMTU:
1795 			if (!(ipp->ipp_fields & IPPF_PATHMTU))
1796 				return (0);
1797 
1798 			return (ip_fill_mtuinfo(&icmp->icmp_v6dst, 0,
1799 				(struct ip6_mtuinfo *)ptr));
1800 		case IPV6_TCLASS:
1801 			if (ipp->ipp_fields & IPPF_TCLASS)
1802 				*i1 = ipp->ipp_tclass;
1803 			else
1804 				*i1 = IPV6_FLOW_TCLASS(
1805 				    IPV6_DEFAULT_VERS_AND_FLOW);
1806 			break;
1807 		default:
1808 			return (-1);
1809 		}
1810 		break;
1811 	case IPPROTO_ICMPV6:
1812 		/*
1813 		 * Only allow IPv6 option processing on native IPv6 sockets.
1814 		 */
1815 		if (icmp->icmp_family != AF_INET6)
1816 			return (-1);
1817 
1818 		if (icmp->icmp_proto != IPPROTO_ICMPV6)
1819 			return (-1);
1820 
1821 		switch (name) {
1822 		case ICMP6_FILTER:
1823 			if (icmp->icmp_filter == NULL) {
1824 				/* Make it look like "pass all" */
1825 				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1826 			} else {
1827 				(void) bcopy(icmp->icmp_filter, ptr,
1828 				    sizeof (icmp6_filter_t));
1829 			}
1830 			return (sizeof (icmp6_filter_t));
1831 		default:
1832 			return (-1);
1833 		}
1834 	default:
1835 		return (-1);
1836 	}
1837 	return (sizeof (int));
1838 }
1839 
1840 /* This routine sets socket options. */
1841 /* ARGSUSED */
1842 int
1843 icmp_opt_set(queue_t *q, uint_t optset_context, int level, int name,
1844     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
1845     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
1846 {
1847 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
1848 	int	*i1 = (int *)invalp;
1849 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
1850 	boolean_t checkonly;
1851 	int	error;
1852 
1853 	switch (optset_context) {
1854 	case SETFN_OPTCOM_CHECKONLY:
1855 		checkonly = B_TRUE;
1856 		/*
1857 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
1858 		 * inlen != 0 implies value supplied and
1859 		 * 	we have to "pretend" to set it.
1860 		 * inlen == 0 implies that there is no
1861 		 * 	value part in T_CHECK request and just validation
1862 		 * done elsewhere should be enough, we just return here.
1863 		 */
1864 		if (inlen == 0) {
1865 			*outlenp = 0;
1866 			return (0);
1867 		}
1868 		break;
1869 	case SETFN_OPTCOM_NEGOTIATE:
1870 		checkonly = B_FALSE;
1871 		break;
1872 	case SETFN_UD_NEGOTIATE:
1873 	case SETFN_CONN_NEGOTIATE:
1874 		checkonly = B_FALSE;
1875 		/*
1876 		 * Negotiating local and "association-related" options
1877 		 * through T_UNITDATA_REQ.
1878 		 *
1879 		 * Following routine can filter out ones we do not
1880 		 * want to be "set" this way.
1881 		 */
1882 		if (!icmp_opt_allow_udr_set(level, name)) {
1883 			*outlenp = 0;
1884 			return (EINVAL);
1885 		}
1886 		break;
1887 	default:
1888 		/*
1889 		 * We should never get here
1890 		 */
1891 		*outlenp = 0;
1892 		return (EINVAL);
1893 	}
1894 
1895 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
1896 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
1897 
1898 	/*
1899 	 * For fixed length options, no sanity check
1900 	 * of passed in length is done. It is assumed *_optcom_req()
1901 	 * routines do the right thing.
1902 	 */
1903 
1904 	switch (level) {
1905 	case SOL_SOCKET:
1906 		switch (name) {
1907 		case SO_DEBUG:
1908 			if (!checkonly)
1909 				icmp->icmp_debug = onoff;
1910 			break;
1911 		case SO_PROTOTYPE:
1912 			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
1913 			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
1914 			    secpolicy_net_rawaccess(cr) != 0) {
1915 				*outlenp = 0;
1916 				return (EACCES);
1917 			}
1918 			/* Can't use IPPROTO_RAW with IPv6 */
1919 			if ((*i1 & 0xFF) == IPPROTO_RAW &&
1920 			    icmp->icmp_family == AF_INET6) {
1921 				*outlenp = 0;
1922 				return (EPROTONOSUPPORT);
1923 			}
1924 			if (checkonly) {
1925 				/* T_CHECK case */
1926 				*(int *)outvalp = (*i1 & 0xFF);
1927 				break;
1928 			}
1929 			icmp->icmp_proto = *i1 & 0xFF;
1930 			if ((icmp->icmp_proto == IPPROTO_RAW ||
1931 			    icmp->icmp_proto == IPPROTO_IGMP) &&
1932 			    icmp->icmp_family == AF_INET)
1933 				icmp->icmp_hdrincl = 1;
1934 			else
1935 				icmp->icmp_hdrincl = 0;
1936 
1937 			if (icmp->icmp_family == AF_INET6 &&
1938 			    icmp->icmp_proto == IPPROTO_ICMPV6) {
1939 				/* Set offset for icmp6_cksum */
1940 				icmp->icmp_raw_checksum = 0;
1941 				icmp->icmp_checksum_off = 2;
1942 			}
1943 			if (icmp->icmp_proto == IPPROTO_UDP ||
1944 			    icmp->icmp_proto == IPPROTO_TCP ||
1945 			    icmp->icmp_proto == IPPROTO_SCTP) {
1946 				icmp->icmp_no_tp_cksum = 1;
1947 				icmp->icmp_sticky_ipp.ipp_fields |=
1948 				    IPPF_NO_CKSUM;
1949 			} else {
1950 				icmp->icmp_no_tp_cksum = 0;
1951 				icmp->icmp_sticky_ipp.ipp_fields &=
1952 				    ~IPPF_NO_CKSUM;
1953 			}
1954 
1955 			if (icmp->icmp_filter != NULL &&
1956 			    icmp->icmp_proto != IPPROTO_ICMPV6) {
1957 				kmem_free(icmp->icmp_filter,
1958 				    sizeof (icmp6_filter_t));
1959 				icmp->icmp_filter = NULL;
1960 			}
1961 
1962 			/* Rebuild the header template */
1963 			error = icmp_build_hdrs(q, icmp);
1964 			if (error != 0) {
1965 				*outlenp = 0;
1966 				return (error);
1967 			}
1968 
1969 			/*
1970 			 * For SCTP, we don't use icmp_bind_proto() for
1971 			 * raw socket binding.  Note that we do not need
1972 			 * to set *outlenp.
1973 			 */
1974 			if (icmp->icmp_proto == IPPROTO_SCTP)
1975 				return (0);
1976 
1977 			icmp_bind_proto(q);
1978 			*outlenp = sizeof (int);
1979 			*(int *)outvalp = *i1 & 0xFF;
1980 			return (0);
1981 		case SO_REUSEADDR:
1982 			if (!checkonly)
1983 				icmp->icmp_reuseaddr = onoff;
1984 			break;
1985 
1986 		/*
1987 		 * The following three items are available here,
1988 		 * but are only meaningful to IP.
1989 		 */
1990 		case SO_DONTROUTE:
1991 			if (!checkonly)
1992 				icmp->icmp_dontroute = onoff;
1993 			break;
1994 		case SO_USELOOPBACK:
1995 			if (!checkonly)
1996 				icmp->icmp_useloopback = onoff;
1997 			break;
1998 		case SO_BROADCAST:
1999 			if (!checkonly)
2000 				icmp->icmp_broadcast = onoff;
2001 			break;
2002 
2003 		case SO_SNDBUF:
2004 			if (*i1 > icmp_max_buf) {
2005 				*outlenp = 0;
2006 				return (ENOBUFS);
2007 			}
2008 			if (!checkonly) {
2009 				q->q_hiwat = *i1;
2010 				q->q_next->q_hiwat = *i1;
2011 			}
2012 			break;
2013 		case SO_RCVBUF:
2014 			if (*i1 > icmp_max_buf) {
2015 				*outlenp = 0;
2016 				return (ENOBUFS);
2017 			}
2018 			if (!checkonly) {
2019 				RD(q)->q_hiwat = *i1;
2020 				(void) mi_set_sth_hiwat(RD(q), *i1);
2021 			}
2022 			break;
2023 		case SO_DGRAM_ERRIND:
2024 			if (!checkonly)
2025 				icmp->icmp_dgram_errind = onoff;
2026 			break;
2027 		case SO_TIMESTAMP:
2028 			if (!checkonly) {
2029 				icmp->icmp_timestamp = onoff;
2030 			}
2031 			break;
2032 		case SO_MAC_EXEMPT:
2033 			if (secpolicy_net_mac_aware(cr) != 0 ||
2034 			    icmp->icmp_state != TS_UNBND)
2035 				return (EACCES);
2036 			if (!checkonly)
2037 				icmp->icmp_mac_exempt = onoff;
2038 			break;
2039 		/*
2040 		 * Following three not meaningful for icmp
2041 		 * Action is same as "default" so we keep them
2042 		 * in comments.
2043 		 * case SO_LINGER:
2044 		 * case SO_KEEPALIVE:
2045 		 * case SO_OOBINLINE:
2046 		 */
2047 		default:
2048 			*outlenp = 0;
2049 			return (EINVAL);
2050 		}
2051 		break;
2052 	case IPPROTO_IP:
2053 		/*
2054 		 * Only allow IPv4 option processing on IPv4 sockets.
2055 		 */
2056 		if (icmp->icmp_family != AF_INET) {
2057 			*outlenp = 0;
2058 			return (ENOPROTOOPT);
2059 		}
2060 		switch (name) {
2061 		case IP_OPTIONS:
2062 		case T_IP_OPTIONS:
2063 			/* Save options for use by IP. */
2064 			if ((inlen & 0x3) ||
2065 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2066 				*outlenp = 0;
2067 				return (EINVAL);
2068 			}
2069 			if (checkonly)
2070 				break;
2071 
2072 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2073 			    &icmp->icmp_ip_snd_options_len,
2074 			    icmp->icmp_label_len, invalp, inlen)) {
2075 				*outlenp = 0;
2076 				return (ENOMEM);
2077 			}
2078 
2079 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2080 			    icmp->icmp_ip_snd_options_len;
2081 			(void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len +
2082 						icmp_wroff_extra);
2083 			break;
2084 		case IP_HDRINCL:
2085 			if (!checkonly)
2086 				icmp->icmp_hdrincl = onoff;
2087 			break;
2088 		case IP_TOS:
2089 		case T_IP_TOS:
2090 			if (!checkonly) {
2091 				icmp->icmp_type_of_service = (uint8_t)*i1;
2092 			}
2093 			break;
2094 		case IP_TTL:
2095 			if (!checkonly) {
2096 				icmp->icmp_ttl = (uint8_t)*i1;
2097 			}
2098 			break;
2099 		case IP_MULTICAST_IF:
2100 			/*
2101 			 * TODO should check OPTMGMT reply and undo this if
2102 			 * there is an error.
2103 			 */
2104 			if (!checkonly)
2105 				icmp->icmp_multicast_if_addr = *i1;
2106 			break;
2107 		case IP_MULTICAST_TTL:
2108 			if (!checkonly)
2109 				icmp->icmp_multicast_ttl = *invalp;
2110 			break;
2111 		case IP_MULTICAST_LOOP:
2112 			if (!checkonly) {
2113 				icmp->icmp_multicast_loop =
2114 				    (*invalp == 0) ? 0 : 1;
2115 			}
2116 			break;
2117 		case IP_BOUND_IF:
2118 			if (!checkonly)
2119 				icmp->icmp_bound_if = *i1;
2120 			break;
2121 		case IP_UNSPEC_SRC:
2122 			if (!checkonly)
2123 				icmp->icmp_unspec_source = onoff;
2124 			break;
2125 		case IP_XMIT_IF:
2126 			if (!checkonly)
2127 				icmp->icmp_xmit_if = *i1;
2128 			break;
2129 		case IP_RECVIF:
2130 			if (!checkonly)
2131 				icmp->icmp_recvif = onoff;
2132 			break;
2133 		case IP_ADD_MEMBERSHIP:
2134 		case IP_DROP_MEMBERSHIP:
2135 		case IP_BLOCK_SOURCE:
2136 		case IP_UNBLOCK_SOURCE:
2137 		case IP_ADD_SOURCE_MEMBERSHIP:
2138 		case IP_DROP_SOURCE_MEMBERSHIP:
2139 		case MCAST_JOIN_GROUP:
2140 		case MCAST_LEAVE_GROUP:
2141 		case MCAST_BLOCK_SOURCE:
2142 		case MCAST_UNBLOCK_SOURCE:
2143 		case MCAST_JOIN_SOURCE_GROUP:
2144 		case MCAST_LEAVE_SOURCE_GROUP:
2145 		case MRT_INIT:
2146 		case MRT_DONE:
2147 		case MRT_ADD_VIF:
2148 		case MRT_DEL_VIF:
2149 		case MRT_ADD_MFC:
2150 		case MRT_DEL_MFC:
2151 		case MRT_VERSION:
2152 		case MRT_ASSERT:
2153 		case IP_SEC_OPT:
2154 		case IP_DONTFAILOVER_IF:
2155 		case IP_NEXTHOP:
2156 			/*
2157 			 * "soft" error (negative)
2158 			 * option not handled at this level
2159 			 * Note: Do not modify *outlenp
2160 			 */
2161 			return (-EINVAL);
2162 		default:
2163 			*outlenp = 0;
2164 			return (EINVAL);
2165 		}
2166 		break;
2167 	case IPPROTO_IPV6: {
2168 		ip6_pkt_t		*ipp;
2169 		boolean_t		sticky;
2170 
2171 		if (icmp->icmp_family != AF_INET6) {
2172 			*outlenp = 0;
2173 			return (ENOPROTOOPT);
2174 		}
2175 		/*
2176 		 * Deal with both sticky options and ancillary data
2177 		 */
2178 		if (thisdg_attrs == NULL) {
2179 			/* sticky options, or none */
2180 			ipp = &icmp->icmp_sticky_ipp;
2181 			sticky = B_TRUE;
2182 		} else {
2183 			/* ancillary data */
2184 			ipp = (ip6_pkt_t *)thisdg_attrs;
2185 			sticky = B_FALSE;
2186 		}
2187 
2188 		switch (name) {
2189 		case IPV6_MULTICAST_IF:
2190 			if (!checkonly)
2191 				icmp->icmp_multicast_if_index = *i1;
2192 			break;
2193 		case IPV6_UNICAST_HOPS:
2194 			/* -1 means use default */
2195 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2196 				*outlenp = 0;
2197 				return (EINVAL);
2198 			}
2199 			if (!checkonly) {
2200 				if (*i1 == -1) {
2201 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2202 					    icmp_ipv6_hoplimit;
2203 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2204 					/* Pass modified value to IP. */
2205 					*i1 = ipp->ipp_hoplimit;
2206 				} else {
2207 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2208 					    (uint8_t)*i1;
2209 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2210 				}
2211 				/* Rebuild the header template */
2212 				error = icmp_build_hdrs(q, icmp);
2213 				if (error != 0) {
2214 					*outlenp = 0;
2215 					return (error);
2216 				}
2217 			}
2218 			break;
2219 		case IPV6_MULTICAST_HOPS:
2220 			/* -1 means use default */
2221 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2222 				*outlenp = 0;
2223 				return (EINVAL);
2224 			}
2225 			if (!checkonly) {
2226 				if (*i1 == -1) {
2227 					icmp->icmp_multicast_ttl =
2228 					    ipp->ipp_multicast_hops =
2229 					    IP_DEFAULT_MULTICAST_TTL;
2230 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2231 					/* Pass modified value to IP. */
2232 					*i1 = icmp->icmp_multicast_ttl;
2233 				} else {
2234 					icmp->icmp_multicast_ttl =
2235 					    ipp->ipp_multicast_hops =
2236 					    (uint8_t)*i1;
2237 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2238 				}
2239 			}
2240 			break;
2241 		case IPV6_MULTICAST_LOOP:
2242 			if (*i1 != 0 && *i1 != 1) {
2243 				*outlenp = 0;
2244 				return (EINVAL);
2245 			}
2246 			if (!checkonly)
2247 				icmp->icmp_multicast_loop = *i1;
2248 			break;
2249 		case IPV6_CHECKSUM:
2250 			/*
2251 			 * Integer offset into the user data of where the
2252 			 * checksum is located.
2253 			 * Offset of -1 disables option.
2254 			 * Does not apply to IPPROTO_ICMPV6.
2255 			 */
2256 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2257 				*outlenp = 0;
2258 				return (EINVAL);
2259 			}
2260 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2261 				/* Negative or not 16 bit aligned offset */
2262 				*outlenp = 0;
2263 				return (EINVAL);
2264 			}
2265 			if (checkonly)
2266 				break;
2267 
2268 			if (*i1 == -1) {
2269 				icmp->icmp_raw_checksum = 0;
2270 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2271 			} else {
2272 				icmp->icmp_raw_checksum = 1;
2273 				icmp->icmp_checksum_off = *i1;
2274 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2275 			}
2276 			/* Rebuild the header template */
2277 			error = icmp_build_hdrs(q, icmp);
2278 			if (error != 0) {
2279 				*outlenp = 0;
2280 				return (error);
2281 			}
2282 			break;
2283 		case IPV6_JOIN_GROUP:
2284 		case IPV6_LEAVE_GROUP:
2285 		case MCAST_JOIN_GROUP:
2286 		case MCAST_LEAVE_GROUP:
2287 		case MCAST_BLOCK_SOURCE:
2288 		case MCAST_UNBLOCK_SOURCE:
2289 		case MCAST_JOIN_SOURCE_GROUP:
2290 		case MCAST_LEAVE_SOURCE_GROUP:
2291 			/*
2292 			 * "soft" error (negative)
2293 			 * option not handled at this level
2294 			 * Note: Do not modify *outlenp
2295 			 */
2296 			return (-EINVAL);
2297 		case IPV6_BOUND_IF:
2298 			if (!checkonly)
2299 				icmp->icmp_bound_if = *i1;
2300 			break;
2301 		case IPV6_UNSPEC_SRC:
2302 			if (!checkonly)
2303 				icmp->icmp_unspec_source = onoff;
2304 			break;
2305 		case IPV6_RECVTCLASS:
2306 			if (!checkonly)
2307 				icmp->icmp_ipv6_recvtclass = onoff;
2308 			break;
2309 		/*
2310 		 * Set boolean switches for ancillary data delivery
2311 		 */
2312 		case IPV6_RECVPKTINFO:
2313 			if (!checkonly)
2314 				icmp->icmp_ipv6_recvpktinfo = onoff;
2315 			break;
2316 		case IPV6_RECVPATHMTU:
2317 			if (!checkonly)
2318 				icmp->icmp_ipv6_recvpathmtu = onoff;
2319 			break;
2320 		case IPV6_RECVHOPLIMIT:
2321 			if (!checkonly)
2322 				icmp->icmp_ipv6_recvhoplimit = onoff;
2323 			break;
2324 		case IPV6_RECVHOPOPTS:
2325 			if (!checkonly)
2326 				icmp->icmp_ipv6_recvhopopts = onoff;
2327 			break;
2328 		case IPV6_RECVDSTOPTS:
2329 			if (!checkonly)
2330 				icmp->icmp_ipv6_recvdstopts = onoff;
2331 			break;
2332 		case _OLD_IPV6_RECVDSTOPTS:
2333 			if (!checkonly)
2334 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2335 			break;
2336 		case IPV6_RECVRTHDRDSTOPTS:
2337 			if (!checkonly)
2338 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2339 			break;
2340 		case IPV6_RECVRTHDR:
2341 			if (!checkonly)
2342 				icmp->icmp_ipv6_recvrthdr = onoff;
2343 			break;
2344 		/*
2345 		 * Set sticky options or ancillary data.
2346 		 * If sticky options, (re)build any extension headers
2347 		 * that might be needed as a result.
2348 		 */
2349 		case IPV6_PKTINFO:
2350 			/*
2351 			 * The source address and ifindex are verified
2352 			 * in ip_opt_set(). For ancillary data the
2353 			 * source address is checked in ip_wput_v6.
2354 			 */
2355 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
2356 				return (EINVAL);
2357 			if (checkonly)
2358 				break;
2359 
2360 			if (inlen == 0) {
2361 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2362 				ipp->ipp_sticky_ignored |=
2363 				    (IPPF_IFINDEX|IPPF_ADDR);
2364 			} else {
2365 				struct in6_pktinfo *pkti;
2366 
2367 				pkti = (struct in6_pktinfo *)invalp;
2368 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2369 				ipp->ipp_addr = pkti->ipi6_addr;
2370 				if (ipp->ipp_ifindex != 0)
2371 					ipp->ipp_fields |= IPPF_IFINDEX;
2372 				else
2373 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2374 				if (!IN6_IS_ADDR_UNSPECIFIED(
2375 				    &ipp->ipp_addr))
2376 					ipp->ipp_fields |= IPPF_ADDR;
2377 				else
2378 					ipp->ipp_fields &= ~IPPF_ADDR;
2379 			}
2380 			if (sticky) {
2381 				error = icmp_build_hdrs(q, icmp);
2382 				if (error != 0)
2383 					return (error);
2384 			}
2385 			break;
2386 		case IPV6_HOPLIMIT:
2387 			/* This option can only be used as ancillary data. */
2388 			if (sticky)
2389 				return (EINVAL);
2390 			if (inlen != 0 && inlen != sizeof (int))
2391 				return (EINVAL);
2392 			if (checkonly)
2393 				break;
2394 
2395 			if (inlen == 0) {
2396 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2397 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2398 			} else {
2399 				if (*i1 > 255 || *i1 < -1)
2400 					return (EINVAL);
2401 				if (*i1 == -1)
2402 					ipp->ipp_hoplimit = icmp_ipv6_hoplimit;
2403 				else
2404 					ipp->ipp_hoplimit = *i1;
2405 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2406 			}
2407 			break;
2408 		case IPV6_TCLASS:
2409 			/*
2410 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2411 			 * and [0, 255] as the actualy traffic class.
2412 			 */
2413 			if (inlen != 0 && inlen != sizeof (int))
2414 				return (EINVAL);
2415 			if (checkonly)
2416 				break;
2417 
2418 			if (inlen == 0) {
2419 				ipp->ipp_fields &= ~IPPF_TCLASS;
2420 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2421 			} else {
2422 				if (*i1 >= 256 || *i1 < -1)
2423 					return (EINVAL);
2424 				if (*i1 == -1) {
2425 					ipp->ipp_tclass =
2426 					    IPV6_FLOW_TCLASS(
2427 					    IPV6_DEFAULT_VERS_AND_FLOW);
2428 				} else {
2429 					ipp->ipp_tclass = *i1;
2430 				}
2431 				ipp->ipp_fields |= IPPF_TCLASS;
2432 			}
2433 			if (sticky) {
2434 				error = icmp_build_hdrs(q, icmp);
2435 				if (error != 0)
2436 					return (error);
2437 			}
2438 			break;
2439 		case IPV6_NEXTHOP:
2440 			/*
2441 			 * IP will verify that the nexthop is reachable
2442 			 * and fail for sticky options.
2443 			 */
2444 			if (inlen != 0 && inlen != sizeof (sin6_t))
2445 				return (EINVAL);
2446 			if (checkonly)
2447 				break;
2448 
2449 			if (inlen == 0) {
2450 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2451 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2452 			} else {
2453 				sin6_t *sin6 = (sin6_t *)invalp;
2454 
2455 				if (sin6->sin6_family != AF_INET6)
2456 					return (EAFNOSUPPORT);
2457 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
2458 					return (EADDRNOTAVAIL);
2459 				ipp->ipp_nexthop = sin6->sin6_addr;
2460 				if (!IN6_IS_ADDR_UNSPECIFIED(
2461 				    &ipp->ipp_nexthop))
2462 					ipp->ipp_fields |= IPPF_NEXTHOP;
2463 				else
2464 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2465 			}
2466 			if (sticky) {
2467 				error = icmp_build_hdrs(q, icmp);
2468 				if (error != 0)
2469 					return (error);
2470 			}
2471 			break;
2472 		case IPV6_HOPOPTS: {
2473 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2474 			/*
2475 			 * Sanity checks - minimum size, size a multiple of
2476 			 * eight bytes, and matching size passed in.
2477 			 */
2478 			if (inlen != 0 &&
2479 			    inlen != (8 * (hopts->ip6h_len + 1)))
2480 				return (EINVAL);
2481 
2482 			if (checkonly)
2483 				break;
2484 			error = optcom_pkt_set(invalp, inlen, sticky,
2485 			    (uchar_t **)&ipp->ipp_hopopts,
2486 			    &ipp->ipp_hopoptslen,
2487 			    sticky ? icmp->icmp_label_len_v6 : 0);
2488 			if (error != 0)
2489 				return (error);
2490 			if (ipp->ipp_hopoptslen == 0) {
2491 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2492 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2493 			} else {
2494 				ipp->ipp_fields |= IPPF_HOPOPTS;
2495 			}
2496 			if (sticky) {
2497 				error = icmp_build_hdrs(q, icmp);
2498 				if (error != 0)
2499 					return (error);
2500 			}
2501 			break;
2502 		}
2503 		case IPV6_RTHDRDSTOPTS: {
2504 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2505 
2506 			/*
2507 			 * Sanity checks - minimum size, size a multiple of
2508 			 * eight bytes, and matching size passed in.
2509 			 */
2510 			if (inlen != 0 &&
2511 			    inlen != (8 * (dopts->ip6d_len + 1)))
2512 				return (EINVAL);
2513 
2514 			if (checkonly)
2515 				break;
2516 
2517 			if (inlen == 0) {
2518 				if (sticky &&
2519 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2520 					kmem_free(ipp->ipp_rtdstopts,
2521 					    ipp->ipp_rtdstoptslen);
2522 					ipp->ipp_rtdstopts = NULL;
2523 					ipp->ipp_rtdstoptslen = 0;
2524 				}
2525 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2526 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2527 			} else {
2528 				error = optcom_pkt_set(invalp, inlen, sticky,
2529 				    (uchar_t **)&ipp->ipp_rtdstopts,
2530 				    &ipp->ipp_rtdstoptslen, 0);
2531 				if (error != 0)
2532 					return (error);
2533 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2534 			}
2535 			if (sticky) {
2536 				error = icmp_build_hdrs(q, icmp);
2537 				if (error != 0)
2538 					return (error);
2539 			}
2540 			break;
2541 		}
2542 		case IPV6_DSTOPTS: {
2543 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2544 
2545 			/*
2546 			 * Sanity checks - minimum size, size a multiple of
2547 			 * eight bytes, and matching size passed in.
2548 			 */
2549 			if (inlen != 0 &&
2550 			    inlen != (8 * (dopts->ip6d_len + 1)))
2551 				return (EINVAL);
2552 
2553 			if (checkonly)
2554 				break;
2555 
2556 			if (inlen == 0) {
2557 				if (sticky &&
2558 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2559 					kmem_free(ipp->ipp_dstopts,
2560 					    ipp->ipp_dstoptslen);
2561 					ipp->ipp_dstopts = NULL;
2562 					ipp->ipp_dstoptslen = 0;
2563 				}
2564 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2565 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2566 			} else {
2567 				error = optcom_pkt_set(invalp, inlen, sticky,
2568 				    (uchar_t **)&ipp->ipp_dstopts,
2569 				    &ipp->ipp_dstoptslen, 0);
2570 				if (error != 0)
2571 					return (error);
2572 				ipp->ipp_fields |= IPPF_DSTOPTS;
2573 			}
2574 			if (sticky) {
2575 				error = icmp_build_hdrs(q, icmp);
2576 				if (error != 0)
2577 					return (error);
2578 			}
2579 			break;
2580 		}
2581 		case IPV6_RTHDR: {
2582 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2583 
2584 			/*
2585 			 * Sanity checks - minimum size, size a multiple of
2586 			 * eight bytes, and matching size passed in.
2587 			 */
2588 			if (inlen != 0 &&
2589 			    inlen != (8 * (rt->ip6r_len + 1)))
2590 				return (EINVAL);
2591 
2592 			if (checkonly)
2593 				break;
2594 
2595 			if (inlen == 0) {
2596 				if (sticky &&
2597 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
2598 					kmem_free(ipp->ipp_rthdr,
2599 					    ipp->ipp_rthdrlen);
2600 					ipp->ipp_rthdr = NULL;
2601 					ipp->ipp_rthdrlen = 0;
2602 				}
2603 				ipp->ipp_fields &= ~IPPF_RTHDR;
2604 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
2605 			} else {
2606 				error = optcom_pkt_set(invalp, inlen, sticky,
2607 				    (uchar_t **)&ipp->ipp_rthdr,
2608 				    &ipp->ipp_rthdrlen, 0);
2609 				if (error != 0)
2610 					return (error);
2611 				ipp->ipp_fields |= IPPF_RTHDR;
2612 			}
2613 			if (sticky) {
2614 				error = icmp_build_hdrs(q, icmp);
2615 				if (error != 0)
2616 					return (error);
2617 			}
2618 			break;
2619 		}
2620 
2621 		case IPV6_DONTFRAG:
2622 			if (checkonly)
2623 				break;
2624 
2625 			if (onoff) {
2626 				ipp->ipp_fields |= IPPF_DONTFRAG;
2627 			} else {
2628 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
2629 			}
2630 			break;
2631 
2632 		case IPV6_USE_MIN_MTU:
2633 			if (inlen != sizeof (int))
2634 				return (EINVAL);
2635 
2636 			if (*i1 < -1 || *i1 > 1)
2637 				return (EINVAL);
2638 
2639 			if (checkonly)
2640 				break;
2641 
2642 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
2643 			ipp->ipp_use_min_mtu = *i1;
2644 			break;
2645 
2646 		/*
2647 		 * This option can't be set.  Its only returned via
2648 		 * getsockopt() or ancillary data.
2649 		 */
2650 		case IPV6_PATHMTU:
2651 			return (EINVAL);
2652 
2653 		case IPV6_BOUND_PIF:
2654 		case IPV6_SEC_OPT:
2655 		case IPV6_DONTFAILOVER_IF:
2656 		case IPV6_SRC_PREFERENCES:
2657 		case IPV6_V6ONLY:
2658 			/* Handled at IP level */
2659 			return (-EINVAL);
2660 		default:
2661 			*outlenp = 0;
2662 			return (EINVAL);
2663 		}
2664 		break;
2665 	}		/* end IPPROTO_IPV6 */
2666 
2667 	case IPPROTO_ICMPV6:
2668 		/*
2669 		 * Only allow IPv6 option processing on IPv6 sockets.
2670 		 */
2671 		if (icmp->icmp_family != AF_INET6) {
2672 			*outlenp = 0;
2673 			return (ENOPROTOOPT);
2674 		}
2675 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
2676 			*outlenp = 0;
2677 			return (ENOPROTOOPT);
2678 		}
2679 		switch (name) {
2680 		case ICMP6_FILTER:
2681 			if (!checkonly) {
2682 				if ((inlen != 0) &&
2683 				    (inlen != sizeof (icmp6_filter_t)))
2684 					return (EINVAL);
2685 
2686 				if (inlen == 0) {
2687 					if (icmp->icmp_filter != NULL) {
2688 						kmem_free(icmp->icmp_filter,
2689 						    sizeof (icmp6_filter_t));
2690 						icmp->icmp_filter = NULL;
2691 					}
2692 				} else {
2693 					if (icmp->icmp_filter == NULL) {
2694 						icmp->icmp_filter = kmem_alloc(
2695 						    sizeof (icmp6_filter_t),
2696 						    KM_NOSLEEP);
2697 						if (icmp->icmp_filter == NULL) {
2698 							*outlenp = 0;
2699 							return (ENOBUFS);
2700 						}
2701 					}
2702 					(void) bcopy(invalp, icmp->icmp_filter,
2703 					    inlen);
2704 				}
2705 			}
2706 			break;
2707 
2708 		default:
2709 			*outlenp = 0;
2710 			return (EINVAL);
2711 		}
2712 		break;
2713 	default:
2714 		*outlenp = 0;
2715 		return (EINVAL);
2716 	}
2717 	/*
2718 	 * Common case of OK return with outval same as inval.
2719 	 */
2720 	if (invalp != outvalp) {
2721 		/* don't trust bcopy for identical src/dst */
2722 		(void) bcopy(invalp, outvalp, inlen);
2723 	}
2724 	*outlenp = inlen;
2725 	return (0);
2726 }
2727 
2728 /*
2729  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
2730  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
2731  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
2732  * headers.
2733  * Returns failure if can't allocate memory.
2734  */
2735 static int
2736 icmp_build_hdrs(queue_t *q, icmp_t *icmp)
2737 {
2738 	uchar_t	*hdrs;
2739 	uint_t	hdrs_len;
2740 	ip6_t	*ip6h;
2741 	ip6i_t	*ip6i;
2742 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
2743 
2744 	hdrs_len = ip_total_hdrs_len_v6(ipp);
2745 	ASSERT(hdrs_len != 0);
2746 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
2747 		/* Need to reallocate */
2748 		if (hdrs_len != 0) {
2749 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
2750 			if (hdrs == NULL)
2751 				return (ENOMEM);
2752 		} else {
2753 			hdrs = NULL;
2754 		}
2755 		if (icmp->icmp_sticky_hdrs_len != 0) {
2756 			kmem_free(icmp->icmp_sticky_hdrs,
2757 			    icmp->icmp_sticky_hdrs_len);
2758 		}
2759 		icmp->icmp_sticky_hdrs = hdrs;
2760 		icmp->icmp_sticky_hdrs_len = hdrs_len;
2761 	}
2762 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
2763 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
2764 
2765 	/* Set header fields not in ipp */
2766 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
2767 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
2768 		ip6h = (ip6_t *)&ip6i[1];
2769 
2770 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
2771 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
2772 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
2773 		}
2774 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
2775 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
2776 		}
2777 	} else {
2778 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
2779 	}
2780 
2781 	if (!(ipp->ipp_fields & IPPF_ADDR))
2782 		ip6h->ip6_src = icmp->icmp_v6src;
2783 
2784 	/* Try to get everything in a single mblk */
2785 	if (hdrs_len > icmp->icmp_max_hdr_len) {
2786 		icmp->icmp_max_hdr_len = hdrs_len;
2787 		(void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len +
2788 		    icmp_wroff_extra);
2789 	}
2790 	return (0);
2791 }
2792 
2793 /*
2794  * This routine retrieves the value of an ND variable in a icmpparam_t
2795  * structure.  It is called through nd_getset when a user reads the
2796  * variable.
2797  */
2798 /* ARGSUSED */
2799 static int
2800 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
2801 {
2802 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
2803 
2804 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
2805 	return (0);
2806 }
2807 
2808 /*
2809  * Walk through the param array specified registering each element with the
2810  * named dispatch (ND) handler.
2811  */
2812 static boolean_t
2813 icmp_param_register(icmpparam_t *icmppa, int cnt)
2814 {
2815 	for (; cnt-- > 0; icmppa++) {
2816 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
2817 			if (!nd_load(&icmp_g_nd, icmppa->icmp_param_name,
2818 			    icmp_param_get, icmp_param_set,
2819 			    (caddr_t)icmppa)) {
2820 				nd_free(&icmp_g_nd);
2821 				return (B_FALSE);
2822 			}
2823 		}
2824 	}
2825 	if (!nd_load(&icmp_g_nd, "icmp_status", icmp_status_report, NULL,
2826 	    NULL)) {
2827 		nd_free(&icmp_g_nd);
2828 		return (B_FALSE);
2829 	}
2830 	return (B_TRUE);
2831 }
2832 
2833 /* This routine sets an ND variable in a icmpparam_t structure. */
2834 /* ARGSUSED */
2835 static int
2836 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
2837 {
2838 	long		new_value;
2839 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
2840 
2841 	/*
2842 	 * Fail the request if the new value does not lie within the
2843 	 * required bounds.
2844 	 */
2845 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
2846 	    new_value < icmppa->icmp_param_min ||
2847 	    new_value > icmppa->icmp_param_max) {
2848 		return (EINVAL);
2849 	}
2850 	/* Set the new value */
2851 	icmppa->icmp_param_value = new_value;
2852 	return (0);
2853 }
2854 
2855 static void
2856 icmp_rput(queue_t *q, mblk_t *mp)
2857 {
2858 	struct T_unitdata_ind	*tudi;
2859 	uchar_t			*rptr;
2860 	struct T_error_ack	*tea;
2861 	icmp_t			*icmp;
2862 	sin_t			*sin;
2863 	sin6_t			*sin6;
2864 	ip6_t			*ip6h;
2865 	ip6i_t			*ip6i;
2866 	mblk_t			*mp1;
2867 	int			hdr_len;
2868 	ipha_t			*ipha;
2869 	int			udi_size;	/* Size of T_unitdata_ind */
2870 	uint_t			ipvers;
2871 	ip6_pkt_t		ipp;
2872 	uint8_t			nexthdr;
2873 	boolean_t		recvif = B_FALSE;
2874 	in_pktinfo_t		*pinfo;
2875 	mblk_t			*options_mp = NULL;
2876 	uint_t			icmp_opt = 0;
2877 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
2878 	uint_t			hopstrip;
2879 
2880 	icmp = (icmp_t *)q->q_ptr;
2881 	if (icmp->icmp_restricted) {
2882 		putnext(q, mp);
2883 		return;
2884 	}
2885 
2886 	if (mp->b_datap->db_type == M_CTL) {
2887 		/*
2888 		 * IP sends up the IPSEC_IN message for handling IPSEC
2889 		 * policy at the TCP level. We don't need it here.
2890 		 */
2891 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
2892 			mp1 = mp->b_cont;
2893 			freeb(mp);
2894 			mp = mp1;
2895 		} else {
2896 			pinfo = (in_pktinfo_t *)mp->b_rptr;
2897 			if ((icmp->icmp_recvif != 0) &&
2898 			    (pinfo->in_pkt_ulp_type == IN_PKTINFO)) {
2899 				/*
2900 				 * IP has passed the options in mp and the
2901 				 * actual data is in b_cont.
2902 				 */
2903 				recvif = B_TRUE;
2904 				/*
2905 				 * We are here bcos IP_RECVIF is set so we need
2906 				 * to extract the options mblk and adjust the
2907 				 * rptr
2908 				 */
2909 				options_mp = mp;
2910 				mp = mp->b_cont;
2911 			}
2912 		}
2913 	}
2914 
2915 	rptr = mp->b_rptr;
2916 	switch (mp->b_datap->db_type) {
2917 	case M_DATA:
2918 		/*
2919 		 * M_DATA messages contain IP packets.  They are handled
2920 		 * following the switch.
2921 		 */
2922 		break;
2923 	case M_PROTO:
2924 	case M_PCPROTO:
2925 		/* M_PROTO messages contain some type of TPI message. */
2926 		if ((mp->b_wptr - rptr) < sizeof (t_scalar_t)) {
2927 			freemsg(mp);
2928 			return;
2929 		}
2930 		tea = (struct T_error_ack *)rptr;
2931 		switch (tea->PRIM_type) {
2932 		case T_ERROR_ACK:
2933 			switch (tea->ERROR_prim) {
2934 			case O_T_BIND_REQ:
2935 			case T_BIND_REQ:
2936 				/*
2937 				 * If our O_T_BIND_REQ/T_BIND_REQ fails,
2938 				 * clear out the source address before
2939 				 * passing the message upstream.
2940 				 * If this was caused by a T_CONN_REQ
2941 				 * revert back to bound state.
2942 				 */
2943 				if (icmp->icmp_state == TS_UNBND) {
2944 					/*
2945 					 * TPI has not yet bound - bind sent by
2946 					 * icmp_bind_proto.
2947 					 */
2948 					freemsg(mp);
2949 					return;
2950 				}
2951 				if (icmp->icmp_state == TS_DATA_XFER) {
2952 					/* Connect failed */
2953 					tea->ERROR_prim = T_CONN_REQ;
2954 					icmp->icmp_v6src =
2955 					    icmp->icmp_bound_v6src;
2956 					icmp->icmp_state = TS_IDLE;
2957 					if (icmp->icmp_family == AF_INET6)
2958 						(void) icmp_build_hdrs(q, icmp);
2959 					break;
2960 				}
2961 
2962 				if (icmp->icmp_discon_pending) {
2963 					tea->ERROR_prim = T_DISCON_REQ;
2964 					icmp->icmp_discon_pending = 0;
2965 				}
2966 				V6_SET_ZERO(icmp->icmp_v6src);
2967 				V6_SET_ZERO(icmp->icmp_bound_v6src);
2968 				icmp->icmp_state = TS_UNBND;
2969 				if (icmp->icmp_family == AF_INET6)
2970 					(void) icmp_build_hdrs(q, icmp);
2971 				break;
2972 			default:
2973 				break;
2974 			}
2975 			break;
2976 		case T_BIND_ACK:
2977 			icmp_rput_bind_ack(q, mp);
2978 			return;
2979 
2980 		case T_OPTMGMT_ACK:
2981 		case T_OK_ACK:
2982 			if (tea->PRIM_type == T_OK_ACK) {
2983 				struct T_ok_ack *toa;
2984 				toa = (struct T_ok_ack *)rptr;
2985 				if (toa->CORRECT_prim == T_UNBIND_REQ) {
2986 					/*
2987 					 * If somebody sets IPSEC options, IP
2988 					 * sends some IPSEC info which is used
2989 					 * by the TCP for detached connections.
2990 					 * We don't need it here.
2991 					 */
2992 					if ((mp1 = mp->b_cont) != NULL) {
2993 						freemsg(mp1);
2994 						mp->b_cont = NULL;
2995 					}
2996 				}
2997 			}
2998 			break;
2999 		default:
3000 			freemsg(mp);
3001 			return;
3002 		}
3003 		putnext(q, mp);
3004 		return;
3005 	case M_CTL:
3006 		if (recvif) {
3007 			/*
3008 			 * IP has passed the options in mp and the actual data
3009 			 * is in b_cont. Jump to normal data processing.
3010 			 */
3011 			break;
3012 		}
3013 
3014 		/* Contains ICMP packet from IP */
3015 		icmp_icmp_error(q, mp);
3016 		return;
3017 	default:
3018 		putnext(q, mp);
3019 		return;
3020 	}
3021 
3022 	/*
3023 	 * Discard message if it is misaligned or smaller than the IP header.
3024 	 */
3025 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3026 		freemsg(mp);
3027 		if (options_mp != NULL)
3028 			freeb(options_mp);
3029 		BUMP_MIB(&rawip_mib, rawipInErrors);
3030 		return;
3031 	}
3032 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3033 
3034 	/* Handle M_DATA messages containing IP packets messages */
3035 	if (ipvers == IPV4_VERSION) {
3036 		/*
3037 		 * Special case where IP attaches
3038 		 * the IRE needs to be handled so that we don't send up
3039 		 * IRE to the user land.
3040 		 */
3041 		ipha = (ipha_t *)rptr;
3042 		hdr_len = IPH_HDR_LENGTH(ipha);
3043 
3044 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3045 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3046 
3047 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3048 			    TH_SYN) && mp->b_cont != NULL) {
3049 				mp1 = mp->b_cont;
3050 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3051 					freeb(mp1);
3052 					mp->b_cont = NULL;
3053 				}
3054 			}
3055 		}
3056 		if (icmp_bsd_compat) {
3057 			ushort_t len;
3058 			len = ntohs(ipha->ipha_length);
3059 
3060 			if (mp->b_datap->db_ref > 1) {
3061 				/*
3062 				 * Allocate a new IP header so that we can
3063 				 * modify ipha_length.
3064 				 */
3065 				mblk_t	*mp1;
3066 
3067 				mp1 = allocb(hdr_len, BPRI_MED);
3068 				if (!mp1) {
3069 					freemsg(mp);
3070 					if (options_mp != NULL)
3071 						freeb(options_mp);
3072 					BUMP_MIB(&rawip_mib, rawipInErrors);
3073 					return;
3074 				}
3075 				bcopy(rptr, mp1->b_rptr, hdr_len);
3076 				mp->b_rptr = rptr + hdr_len;
3077 				rptr = mp1->b_rptr;
3078 				ipha = (ipha_t *)rptr;
3079 				mp1->b_cont = mp;
3080 				mp1->b_wptr = rptr + hdr_len;
3081 				mp = mp1;
3082 			}
3083 			len -= hdr_len;
3084 			ipha->ipha_length = htons(len);
3085 		}
3086 	}
3087 
3088 	/*
3089 	 * This is the inbound data path.  Packets are passed upstream as
3090 	 * T_UNITDATA_IND messages with full IP headers still attached.
3091 	 */
3092 	if (icmp->icmp_family == AF_INET) {
3093 		ASSERT(ipvers == IPV4_VERSION);
3094 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3095 		if (recvif) {
3096 			udi_size += sizeof (struct T_opthdr) +
3097 			    sizeof (uint_t);
3098 		}
3099 		/*
3100 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3101 		 * buffer. Since gethrestime() expects a pointer aligned
3102 		 * argument, we allocate space necessary for extra
3103 		 * alignment (even though it might not be used).
3104 		 */
3105 		if (icmp->icmp_timestamp) {
3106 			udi_size += sizeof (struct T_opthdr) +
3107 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3108 		}
3109 		mp1 = allocb(udi_size, BPRI_MED);
3110 		if (mp1 == NULL) {
3111 			freemsg(mp);
3112 			if (options_mp != NULL)
3113 				freeb(options_mp);
3114 			BUMP_MIB(&rawip_mib, rawipInErrors);
3115 			return;
3116 		}
3117 		mp1->b_cont = mp;
3118 		mp = mp1;
3119 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3120 		mp->b_datap->db_type = M_PROTO;
3121 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3122 		tudi->PRIM_type = T_UNITDATA_IND;
3123 		tudi->SRC_length = sizeof (sin_t);
3124 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3125 		sin = (sin_t *)&tudi[1];
3126 		*sin = sin_null;
3127 		sin->sin_family = AF_INET;
3128 		sin->sin_addr.s_addr = ipha->ipha_src;
3129 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3130 		    sizeof (sin_t);
3131 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3132 		tudi->OPT_length = udi_size;
3133 
3134 		/*
3135 		 * Add options if IP_RECVIF is set
3136 		 */
3137 		if (udi_size != 0) {
3138 			char *dstopt;
3139 
3140 			dstopt = (char *)&sin[1];
3141 			if (recvif) {
3142 
3143 				struct T_opthdr *toh;
3144 				uint_t		*dstptr;
3145 
3146 				toh = (struct T_opthdr *)dstopt;
3147 				toh->level = IPPROTO_IP;
3148 				toh->name = IP_RECVIF;
3149 				toh->len = sizeof (struct T_opthdr) +
3150 					sizeof (uint_t);
3151 				toh->status = 0;
3152 				dstopt += sizeof (struct T_opthdr);
3153 				dstptr = (uint_t *)dstopt;
3154 				*dstptr = pinfo->in_pkt_ifindex;
3155 				dstopt += sizeof (uint_t);
3156 				freeb(options_mp);
3157 				udi_size -= toh->len;
3158 			}
3159 			if (icmp->icmp_timestamp) {
3160 				struct	T_opthdr *toh;
3161 
3162 				toh = (struct T_opthdr *)dstopt;
3163 				toh->level = SOL_SOCKET;
3164 				toh->name = SCM_TIMESTAMP;
3165 				toh->len = sizeof (struct T_opthdr) +
3166 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3167 				toh->status = 0;
3168 				dstopt += sizeof (struct T_opthdr);
3169 				/* Align for gethrestime() */
3170 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3171 				    sizeof (intptr_t));
3172 				gethrestime((timestruc_t *)dstopt);
3173 				dstopt += sizeof (timestruc_t);
3174 				udi_size -= toh->len;
3175 			}
3176 
3177 			/* Consumed all of allocated space */
3178 			ASSERT(udi_size == 0);
3179 		}
3180 
3181 		BUMP_MIB(&rawip_mib, rawipInDatagrams);
3182 		putnext(q, mp);
3183 		return;
3184 	}
3185 
3186 	/*
3187 	 * We don't need options_mp in the IPv6 path.
3188 	 */
3189 	if (options_mp != NULL) {
3190 		freeb(options_mp);
3191 		options_mp = NULL;
3192 	}
3193 
3194 	/*
3195 	 * Discard message if it is smaller than the IPv6 header
3196 	 * or if the header is malformed.
3197 	 */
3198 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3199 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3200 	    icmp->icmp_family != AF_INET6) {
3201 		freemsg(mp);
3202 		BUMP_MIB(&rawip_mib, rawipInErrors);
3203 		return;
3204 	}
3205 
3206 	/* Initialize */
3207 	ipp.ipp_fields = 0;
3208 	hopstrip = 0;
3209 
3210 	ip6h = (ip6_t *)rptr;
3211 	/*
3212 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3213 	 * as well as individual lenghts of ext hdrs (and ptrs to
3214 	 * them).
3215 	 */
3216 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3217 		/* Look for ifindex information */
3218 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3219 			ip6i = (ip6i_t *)ip6h;
3220 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3221 				ASSERT(ip6i->ip6i_ifindex != 0);
3222 				ipp.ipp_fields |= IPPF_IFINDEX;
3223 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3224 			}
3225 			rptr = (uchar_t *)&ip6i[1];
3226 			mp->b_rptr = rptr;
3227 			if (rptr == mp->b_wptr) {
3228 				mp1 = mp->b_cont;
3229 				freeb(mp);
3230 				mp = mp1;
3231 				rptr = mp->b_rptr;
3232 			}
3233 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3234 			ip6h = (ip6_t *)rptr;
3235 		}
3236 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3237 
3238 		/*
3239 		 * We need to lie a bit to the user because users inside
3240 		 * labeled compartments should not see their own labels.  We
3241 		 * assume that in all other respects IP has checked the label,
3242 		 * and that the label is always first among the options.  (If
3243 		 * it's not first, then this code won't see it, and the option
3244 		 * will be passed along to the user.)
3245 		 *
3246 		 * If we had multilevel ICMP sockets, then the following code
3247 		 * should be skipped for them to allow the user to see the
3248 		 * label.
3249 		 *
3250 		 * Alignment restrictions in the definition of IP options
3251 		 * (namely, the requirement that the 4-octet DOI goes on a
3252 		 * 4-octet boundary) mean that we know exactly where the option
3253 		 * should start, but we're lenient for other hosts.
3254 		 *
3255 		 * Note that there are no multilevel ICMP or raw IP sockets
3256 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3257 		 */
3258 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3259 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3260 			const uchar_t *ucp =
3261 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3262 			int remlen = ipp.ipp_hopoptslen - 2;
3263 
3264 			while (remlen > 0) {
3265 				if (*ucp == IP6OPT_PAD1) {
3266 					remlen--;
3267 					ucp++;
3268 				} else if (*ucp == IP6OPT_PADN) {
3269 					remlen -= ucp[1] + 2;
3270 					ucp += ucp[1] + 2;
3271 				} else if (*ucp == ip6opt_ls) {
3272 					hopstrip = (ucp -
3273 					    (const uchar_t *)ipp.ipp_hopopts) +
3274 					    ucp[1] + 2;
3275 					hopstrip = (hopstrip + 7) & ~7;
3276 					break;
3277 				} else {
3278 					/* label option must be first */
3279 					break;
3280 				}
3281 			}
3282 		}
3283 	} else {
3284 		hdr_len = IPV6_HDR_LEN;
3285 		ip6i = NULL;
3286 		nexthdr = ip6h->ip6_nxt;
3287 	}
3288 	/*
3289 	 * One special case where IP attaches the IRE needs to
3290 	 * be handled so that we don't send up IRE to the user land.
3291 	 */
3292 	if (nexthdr == IPPROTO_TCP) {
3293 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3294 
3295 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3296 		    mp->b_cont != NULL) {
3297 			mp1 = mp->b_cont;
3298 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3299 				freeb(mp1);
3300 				mp->b_cont = NULL;
3301 			}
3302 		}
3303 	}
3304 	/*
3305 	 * Check a filter for ICMPv6 types if needed.
3306 	 * Verify raw checksums if needed.
3307 	 */
3308 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3309 		if (icmp->icmp_filter != NULL) {
3310 			int type;
3311 
3312 			/* Assumes that IP has done the pullupmsg */
3313 			type = mp->b_rptr[hdr_len];
3314 
3315 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3316 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3317 				freemsg(mp);
3318 				return;
3319 			}
3320 		} else {
3321 			/* Checksum */
3322 			uint16_t	*up;
3323 			uint32_t	sum;
3324 			int		remlen;
3325 
3326 			up = (uint16_t *)&ip6h->ip6_src;
3327 
3328 			remlen = msgdsize(mp) - hdr_len;
3329 			sum = htons(icmp->icmp_proto + remlen)
3330 			    + up[0] + up[1] + up[2] + up[3]
3331 			    + up[4] + up[5] + up[6] + up[7]
3332 			    + up[8] + up[9] + up[10] + up[11]
3333 			    + up[12] + up[13] + up[14] + up[15];
3334 			sum = (sum & 0xffff) + (sum >> 16);
3335 			sum = IP_CSUM(mp, hdr_len, sum);
3336 			if (sum != 0) {
3337 				/* IPv6 RAW checksum failed */
3338 				ip0dbg(("icmp_rput: RAW checksum "
3339 				    "failed %x\n", sum));
3340 				freemsg(mp);
3341 				BUMP_MIB(&rawip_mib, rawipInCksumErrs);
3342 				return;
3343 			}
3344 		}
3345 	}
3346 	/* Skip all the IPv6 headers per API */
3347 	mp->b_rptr += hdr_len;
3348 
3349 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3350 
3351 	/*
3352 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3353 	 * maintain state information, instead of relying on icmp_t
3354 	 * structure, since there arent any locks protecting these members
3355 	 * and there is a window where there might be a race between a
3356 	 * thread setting options on the write side and a thread reading
3357 	 * these options on the read size.
3358 	 */
3359 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3360 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3361 		if (icmp->icmp_ipv6_recvhopopts &&
3362 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3363 		    ipp.ipp_hopoptslen > hopstrip) {
3364 			udi_size += sizeof (struct T_opthdr) +
3365 			    ipp.ipp_hopoptslen - hopstrip;
3366 			icmp_opt |= IPPF_HOPOPTS;
3367 		}
3368 		if ((icmp->icmp_ipv6_recvdstopts ||
3369 			icmp->icmp_old_ipv6_recvdstopts) &&
3370 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3371 			udi_size += sizeof (struct T_opthdr) +
3372 			    ipp.ipp_dstoptslen;
3373 			icmp_opt |= IPPF_DSTOPTS;
3374 		}
3375 		if (((icmp->icmp_ipv6_recvdstopts &&
3376 		    icmp->icmp_ipv6_recvrthdr &&
3377 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3378 		    icmp->icmp_ipv6_recvrtdstopts) &&
3379 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3380 			udi_size += sizeof (struct T_opthdr) +
3381 			    ipp.ipp_rtdstoptslen;
3382 			icmp_opt |= IPPF_RTDSTOPTS;
3383 		}
3384 		if (icmp->icmp_ipv6_recvrthdr &&
3385 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3386 			udi_size += sizeof (struct T_opthdr) +
3387 			    ipp.ipp_rthdrlen;
3388 			icmp_opt |= IPPF_RTHDR;
3389 		}
3390 		if (icmp->icmp_ipv6_recvpktinfo &&
3391 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3392 			udi_size += sizeof (struct T_opthdr) +
3393 			    sizeof (struct in6_pktinfo);
3394 			icmp_opt |= IPPF_IFINDEX;
3395 		}
3396 	}
3397 	if (icmp->icmp_ipv6_recvhoplimit) {
3398 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3399 		icmp_ipv6_recvhoplimit = B_TRUE;
3400 	}
3401 
3402 	if (icmp->icmp_ipv6_recvtclass)
3403 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3404 
3405 	mp1 = allocb(udi_size, BPRI_MED);
3406 	if (mp1 == NULL) {
3407 		freemsg(mp);
3408 		BUMP_MIB(&rawip_mib, rawipInErrors);
3409 		return;
3410 	}
3411 	mp1->b_cont = mp;
3412 	mp = mp1;
3413 	mp->b_datap->db_type = M_PROTO;
3414 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3415 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3416 	tudi->PRIM_type = T_UNITDATA_IND;
3417 	tudi->SRC_length = sizeof (sin6_t);
3418 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3419 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3420 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3421 	tudi->OPT_length = udi_size;
3422 	sin6 = (sin6_t *)&tudi[1];
3423 	sin6->sin6_port = 0;
3424 	sin6->sin6_family = AF_INET6;
3425 
3426 	sin6->sin6_addr = ip6h->ip6_src;
3427 	/* No sin6_flowinfo per API */
3428 	sin6->sin6_flowinfo = 0;
3429 	/* For link-scope source pass up scope id */
3430 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3431 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3432 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3433 	else
3434 		sin6->sin6_scope_id = 0;
3435 
3436 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3437 	    icmp->icmp_zoneid);
3438 
3439 	if (udi_size != 0) {
3440 		uchar_t *dstopt;
3441 
3442 		dstopt = (uchar_t *)&sin6[1];
3443 		if (icmp_opt & IPPF_IFINDEX) {
3444 			struct T_opthdr *toh;
3445 			struct in6_pktinfo *pkti;
3446 
3447 			toh = (struct T_opthdr *)dstopt;
3448 			toh->level = IPPROTO_IPV6;
3449 			toh->name = IPV6_PKTINFO;
3450 			toh->len = sizeof (struct T_opthdr) +
3451 			    sizeof (*pkti);
3452 			toh->status = 0;
3453 			dstopt += sizeof (struct T_opthdr);
3454 			pkti = (struct in6_pktinfo *)dstopt;
3455 			pkti->ipi6_addr = ip6h->ip6_dst;
3456 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3457 			dstopt += sizeof (*pkti);
3458 			udi_size -= toh->len;
3459 		}
3460 		if (icmp_ipv6_recvhoplimit) {
3461 			struct T_opthdr *toh;
3462 
3463 			toh = (struct T_opthdr *)dstopt;
3464 			toh->level = IPPROTO_IPV6;
3465 			toh->name = IPV6_HOPLIMIT;
3466 			toh->len = sizeof (struct T_opthdr) +
3467 			    sizeof (uint_t);
3468 			toh->status = 0;
3469 			dstopt += sizeof (struct T_opthdr);
3470 			*(uint_t *)dstopt = ip6h->ip6_hops;
3471 			dstopt += sizeof (uint_t);
3472 			udi_size -= toh->len;
3473 		}
3474 		if (icmp->icmp_ipv6_recvtclass) {
3475 			struct T_opthdr *toh;
3476 
3477 			toh = (struct T_opthdr *)dstopt;
3478 			toh->level = IPPROTO_IPV6;
3479 			toh->name = IPV6_TCLASS;
3480 			toh->len = sizeof (struct T_opthdr) +
3481 			    sizeof (uint_t);
3482 			toh->status = 0;
3483 			dstopt += sizeof (struct T_opthdr);
3484 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3485 			dstopt += sizeof (uint_t);
3486 			udi_size -= toh->len;
3487 		}
3488 		if (icmp_opt & IPPF_HOPOPTS) {
3489 			struct T_opthdr *toh;
3490 
3491 			toh = (struct T_opthdr *)dstopt;
3492 			toh->level = IPPROTO_IPV6;
3493 			toh->name = IPV6_HOPOPTS;
3494 			toh->len = sizeof (struct T_opthdr) +
3495 			    ipp.ipp_hopoptslen - hopstrip;
3496 			toh->status = 0;
3497 			dstopt += sizeof (struct T_opthdr);
3498 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
3499 			    ipp.ipp_hopoptslen - hopstrip);
3500 			if (hopstrip > 0) {
3501 				/* copy next header value and fake length */
3502 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
3503 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
3504 				    hopstrip / 8;
3505 			}
3506 			dstopt += ipp.ipp_hopoptslen - hopstrip;
3507 			udi_size -= toh->len;
3508 		}
3509 		if (icmp_opt & IPPF_RTDSTOPTS) {
3510 			struct T_opthdr *toh;
3511 
3512 			toh = (struct T_opthdr *)dstopt;
3513 			toh->level = IPPROTO_IPV6;
3514 			toh->name = IPV6_DSTOPTS;
3515 			toh->len = sizeof (struct T_opthdr) +
3516 			    ipp.ipp_rtdstoptslen;
3517 			toh->status = 0;
3518 			dstopt += sizeof (struct T_opthdr);
3519 			bcopy(ipp.ipp_rtdstopts, dstopt,
3520 			    ipp.ipp_rtdstoptslen);
3521 			dstopt += ipp.ipp_rtdstoptslen;
3522 			udi_size -= toh->len;
3523 		}
3524 		if (icmp_opt & IPPF_RTHDR) {
3525 			struct T_opthdr *toh;
3526 
3527 			toh = (struct T_opthdr *)dstopt;
3528 			toh->level = IPPROTO_IPV6;
3529 			toh->name = IPV6_RTHDR;
3530 			toh->len = sizeof (struct T_opthdr) +
3531 			    ipp.ipp_rthdrlen;
3532 			toh->status = 0;
3533 			dstopt += sizeof (struct T_opthdr);
3534 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
3535 			dstopt += ipp.ipp_rthdrlen;
3536 			udi_size -= toh->len;
3537 		}
3538 		if (icmp_opt & IPPF_DSTOPTS) {
3539 			struct T_opthdr *toh;
3540 
3541 			toh = (struct T_opthdr *)dstopt;
3542 			toh->level = IPPROTO_IPV6;
3543 			toh->name = IPV6_DSTOPTS;
3544 			toh->len = sizeof (struct T_opthdr) +
3545 			    ipp.ipp_dstoptslen;
3546 			toh->status = 0;
3547 			dstopt += sizeof (struct T_opthdr);
3548 			bcopy(ipp.ipp_dstopts, dstopt,
3549 			    ipp.ipp_dstoptslen);
3550 			dstopt += ipp.ipp_dstoptslen;
3551 			udi_size -= toh->len;
3552 		}
3553 		/* Consumed all of allocated space */
3554 		ASSERT(udi_size == 0);
3555 	}
3556 	BUMP_MIB(&rawip_mib, rawipInDatagrams);
3557 	putnext(q, mp);
3558 }
3559 
3560 /*
3561  * Process a T_BIND_ACK
3562  */
3563 static void
3564 icmp_rput_bind_ack(queue_t *q, mblk_t *mp)
3565 {
3566 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
3567 	mblk_t	*mp1;
3568 	ire_t	*ire;
3569 	struct T_bind_ack *tba;
3570 	uchar_t *addrp;
3571 	ipa_conn_t	*ac;
3572 	ipa6_conn_t	*ac6;
3573 
3574 	/*
3575 	 * We know if headers are included or not so we can
3576 	 * safely do this.
3577 	 */
3578 	if (icmp->icmp_state == TS_UNBND) {
3579 		/*
3580 		 * TPI has not yet bound - bind sent by
3581 		 * icmp_bind_proto.
3582 		 */
3583 		freemsg(mp);
3584 		return;
3585 	}
3586 	if (icmp->icmp_discon_pending)
3587 		icmp->icmp_discon_pending = 0;
3588 
3589 	/*
3590 	 * If a broadcast/multicast address was bound set
3591 	 * the source address to 0.
3592 	 * This ensures no datagrams with broadcast address
3593 	 * as source address are emitted (which would violate
3594 	 * RFC1122 - Hosts requirements)
3595 	 *
3596 	 * Note that when connecting the returned IRE is
3597 	 * for the destination address and we only perform
3598 	 * the broadcast check for the source address (it
3599 	 * is OK to connect to a broadcast/multicast address.)
3600 	 */
3601 	mp1 = mp->b_cont;
3602 	if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) {
3603 		ire = (ire_t *)mp1->b_rptr;
3604 
3605 		/*
3606 		 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast
3607 		 * local address.
3608 		 */
3609 		if (ire->ire_type == IRE_BROADCAST &&
3610 		    icmp->icmp_state != TS_DATA_XFER) {
3611 			/* This was just a local bind to a MC/broadcast addr */
3612 			V6_SET_ZERO(icmp->icmp_v6src);
3613 			if (icmp->icmp_family == AF_INET6)
3614 				(void) icmp_build_hdrs(q, icmp);
3615 		} else if (V6_OR_V4_INADDR_ANY(icmp->icmp_v6src)) {
3616 			/*
3617 			 * Local address not yet set - pick it from the
3618 			 * T_bind_ack
3619 			 */
3620 			tba = (struct T_bind_ack *)mp->b_rptr;
3621 			addrp = &mp->b_rptr[tba->ADDR_offset];
3622 			switch (icmp->icmp_family) {
3623 			case AF_INET:
3624 				if (tba->ADDR_length == sizeof (ipa_conn_t)) {
3625 					ac = (ipa_conn_t *)addrp;
3626 				} else {
3627 					ASSERT(tba->ADDR_length ==
3628 					    sizeof (ipa_conn_x_t));
3629 					ac = &((ipa_conn_x_t *)addrp)->acx_conn;
3630 				}
3631 				IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr,
3632 				    &icmp->icmp_v6src);
3633 				break;
3634 			case AF_INET6:
3635 				if (tba->ADDR_length == sizeof (ipa6_conn_t)) {
3636 					ac6 = (ipa6_conn_t *)addrp;
3637 				} else {
3638 					ASSERT(tba->ADDR_length ==
3639 					    sizeof (ipa6_conn_x_t));
3640 					ac6 = &((ipa6_conn_x_t *)
3641 					    addrp)->ac6x_conn;
3642 				}
3643 				icmp->icmp_v6src = ac6->ac6_laddr;
3644 				(void) icmp_build_hdrs(q, icmp);
3645 			}
3646 		}
3647 		mp1 = mp1->b_cont;
3648 	}
3649 	/*
3650 	 * Look for one or more appended ACK message added by
3651 	 * icmp_connect or icmp_disconnect.
3652 	 * If none found just send up the T_BIND_ACK.
3653 	 * icmp_connect has appended a T_OK_ACK and a
3654 	 * T_CONN_CON.
3655 	 * icmp_disconnect has appended a T_OK_ACK.
3656 	 */
3657 	if (mp1 != NULL) {
3658 		if (mp->b_cont == mp1)
3659 			mp->b_cont = NULL;
3660 		else {
3661 			ASSERT(mp->b_cont->b_cont == mp1);
3662 			mp->b_cont->b_cont = NULL;
3663 		}
3664 		freemsg(mp);
3665 		mp = mp1;
3666 		while (mp != NULL) {
3667 			mp1 = mp->b_cont;
3668 			mp->b_cont = NULL;
3669 			putnext(q, mp);
3670 			mp = mp1;
3671 		}
3672 		return;
3673 	}
3674 	freemsg(mp->b_cont);
3675 	mp->b_cont = NULL;
3676 	putnext(q, mp);
3677 }
3678 
3679 /*
3680  * return SNMP stuff in buffer in mpdata
3681  */
3682 static int
3683 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
3684 {
3685 	mblk_t			*mpdata;
3686 	struct opthdr		*optp;
3687 
3688 	if (mpctl == NULL ||
3689 	    (mpdata = mpctl->b_cont) == NULL) {
3690 		return (0);
3691 	}
3692 
3693 	/* fixed length structure for IPv4 and IPv6 counters */
3694 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
3695 	optp->level = EXPER_RAWIP;
3696 	optp->name = 0;
3697 	(void) snmp_append_data(mpdata, (char *)&rawip_mib, sizeof (rawip_mib));
3698 	optp->len = msgdsize(mpdata);
3699 	qreply(q, mpctl);
3700 
3701 	return (1);
3702 }
3703 
3704 /*
3705  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
3706  * TODO:  If this ever actually tries to set anything, it needs to be
3707  * to do the appropriate locking.
3708  */
3709 /* ARGSUSED */
3710 static int
3711 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
3712     uchar_t *ptr, int len)
3713 {
3714 	switch (level) {
3715 	case EXPER_RAWIP:
3716 		return (0);
3717 	default:
3718 		return (1);
3719 	}
3720 }
3721 
3722 /* Report for ndd "icmp_status" */
3723 /* ARGSUSED */
3724 static int
3725 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3726 {
3727 	IDP	idp;
3728 	icmp_t	*icmp;
3729 	char	*state;
3730 	char	laddrbuf[INET6_ADDRSTRLEN];
3731 	char	faddrbuf[INET6_ADDRSTRLEN];
3732 
3733 	(void) mi_mpprintf(mp,
3734 	    "RAWIP    " MI_COL_HDRPAD_STR
3735 	/*   01234567[89ABCDEF] */
3736 	    "  src addr        dest addr       state");
3737 	/*   xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */
3738 
3739 
3740 	for (idp = mi_first_ptr(&icmp_g_head);
3741 	    (icmp = (icmp_t *)idp) != NULL;
3742 	    idp = mi_next_ptr(&icmp_g_head, idp)) {
3743 		if (icmp->icmp_state == TS_UNBND)
3744 			state = "UNBOUND";
3745 		else if (icmp->icmp_state == TS_IDLE)
3746 			state = "IDLE";
3747 		else if (icmp->icmp_state == TS_DATA_XFER)
3748 			state = "CONNECTED";
3749 		else
3750 			state = "UnkState";
3751 
3752 		(void) mi_mpprintf(mp,
3753 		    MI_COL_PTRFMT_STR "%s %s %s",
3754 		    (void *)icmp,
3755 		    inet_ntop(AF_INET6, &icmp->icmp_v6dst, faddrbuf,
3756 		    sizeof (faddrbuf)),
3757 		    inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf,
3758 		    sizeof (laddrbuf)),
3759 		    state);
3760 	}
3761 	return (0);
3762 }
3763 
3764 /*
3765  * This routine creates a T_UDERROR_IND message and passes it upstream.
3766  * The address and options are copied from the T_UNITDATA_REQ message
3767  * passed in mp.  This message is freed.
3768  */
3769 static void
3770 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
3771 {
3772 	mblk_t	*mp1;
3773 	uchar_t	*rptr = mp->b_rptr;
3774 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
3775 
3776 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
3777 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
3778 	    tudr->OPT_length, err);
3779 	if (mp1)
3780 		qreply(q, mp1);
3781 	freemsg(mp);
3782 }
3783 
3784 /*
3785  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
3786  * After some error checking, the message is passed downstream to ip.
3787  */
3788 static void
3789 icmp_unbind(queue_t *q, mblk_t *mp)
3790 {
3791 	icmp_t	*icmp = (icmp_t *)q->q_ptr;
3792 
3793 	/* If a bind has not been done, we can't unbind. */
3794 	if (icmp->icmp_state == TS_UNBND) {
3795 		icmp_err_ack(q, mp, TOUTSTATE, 0);
3796 		return;
3797 	}
3798 	V6_SET_ZERO(icmp->icmp_v6src);
3799 	V6_SET_ZERO(icmp->icmp_bound_v6src);
3800 	icmp->icmp_state = TS_UNBND;
3801 
3802 	if (icmp->icmp_family == AF_INET6) {
3803 		int error;
3804 
3805 		/* Rebuild the header template */
3806 		error = icmp_build_hdrs(q, icmp);
3807 		if (error != 0) {
3808 			icmp_err_ack(q, mp, TSYSERR, error);
3809 			return;
3810 		}
3811 	}
3812 	/* Pass the unbind to IP. */
3813 	putnext(q, mp);
3814 }
3815 
3816 /*
3817  * Process IPv4 packets that already include an IP header.
3818  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
3819  * IPPROTO_IGMP).
3820  */
3821 static void
3822 icmp_wput_hdrincl(queue_t *q, mblk_t *mp, icmp_t *icmp)
3823 {
3824 	ipha_t	*ipha;
3825 	int	ip_hdr_length;
3826 	int	tp_hdr_len;
3827 	mblk_t	*mp1;
3828 	uint_t	pkt_len;
3829 
3830 	ipha = (ipha_t *)mp->b_rptr;
3831 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
3832 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
3833 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
3834 			BUMP_MIB(&rawip_mib, rawipOutErrors);
3835 			freemsg(mp);
3836 			return;
3837 		}
3838 		ipha = (ipha_t *)mp->b_rptr;
3839 	}
3840 	ipha->ipha_version_and_hdr_length =
3841 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
3842 
3843 	/*
3844 	 * For the socket of SOCK_RAW type, the checksum is provided in the
3845 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
3846 	 * tell IP that the application has sent a complete IP header and not
3847 	 * to compute the transport checksum nor change the DF flag.
3848 	 */
3849 	ipha->ipha_ident = IP_HDR_INCLUDED;
3850 	ipha->ipha_hdr_checksum = 0;
3851 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
3852 	/* Insert options if any */
3853 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
3854 		/*
3855 		 * Put the IP header plus any transport header that is
3856 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
3857 		 * that at least the checksum field is in the first mblk.)
3858 		 */
3859 		switch (ipha->ipha_protocol) {
3860 		case IPPROTO_UDP:
3861 			tp_hdr_len = 8;
3862 			break;
3863 		case IPPROTO_TCP:
3864 			tp_hdr_len = 20;
3865 			break;
3866 		default:
3867 			tp_hdr_len = 0;
3868 			break;
3869 		}
3870 		/*
3871 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
3872 		 * tp_hdr_len bytes will be in a single mblk.
3873 		 */
3874 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
3875 		    tp_hdr_len)) {
3876 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
3877 			    tp_hdr_len)) {
3878 				BUMP_MIB(&rawip_mib, rawipOutErrors);
3879 				freemsg(mp);
3880 				return;
3881 			}
3882 			ipha = (ipha_t *)mp->b_rptr;
3883 		}
3884 
3885 		/*
3886 		 * if the length is larger then the max allowed IP packet,
3887 		 * then send an error and abort the processing.
3888 		 */
3889 		pkt_len = ntohs(ipha->ipha_length)
3890 		    + icmp->icmp_ip_snd_options_len;
3891 		if (pkt_len > IP_MAXPACKET) {
3892 			icmp_ud_err(q, mp, EMSGSIZE);
3893 			return;
3894 		}
3895 		if (!(mp1 = allocb(ip_hdr_length + icmp_wroff_extra +
3896 		    tp_hdr_len, BPRI_LO))) {
3897 			icmp_ud_err(q, mp, ENOMEM);
3898 			return;
3899 		}
3900 		mp1->b_rptr += icmp_wroff_extra;
3901 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
3902 
3903 		ipha->ipha_length = htons((uint16_t)pkt_len);
3904 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
3905 
3906 		/* Copy transport header if any */
3907 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
3908 		mp1->b_wptr += tp_hdr_len;
3909 
3910 		/* Add options */
3911 		ipha = (ipha_t *)mp1->b_rptr;
3912 		bcopy(icmp->icmp_ip_snd_options, &ipha[1],
3913 		    icmp->icmp_ip_snd_options_len);
3914 
3915 		/* Drop IP header and transport header from original */
3916 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
3917 
3918 		mp1->b_cont = mp;
3919 		mp = mp1;
3920 		/*
3921 		 * Massage source route putting first source
3922 		 * route in ipha_dst.
3923 		 */
3924 		(void) ip_massage_options(ipha);
3925 	}
3926 	mblk_setcred(mp, icmp->icmp_credp);
3927 	putnext(q, mp);
3928 }
3929 
3930 static boolean_t
3931 icmp_update_label(queue_t *q, icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
3932 {
3933 	int err;
3934 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
3935 
3936 	err = tsol_compute_label(DB_CREDDEF(mp, icmp->icmp_credp), dst,
3937 	    opt_storage, icmp->icmp_mac_exempt);
3938 	if (err == 0) {
3939 		err = tsol_update_options(&icmp->icmp_ip_snd_options,
3940 		    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
3941 		    opt_storage);
3942 	}
3943 	if (err != 0) {
3944 		BUMP_MIB(&rawip_mib, rawipOutErrors);
3945 		DTRACE_PROBE4(
3946 		    tx__ip__log__drop__updatelabel__icmp,
3947 		    char *, "queue(1) failed to update options(2) on mp(3)",
3948 		    queue_t *, q, char *, opt_storage, mblk_t *, mp);
3949 		icmp_ud_err(q, mp, err);
3950 		return (B_FALSE);
3951 	}
3952 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
3953 	return (B_TRUE);
3954 }
3955 
3956 /*
3957  * This routine handles all messages passed downstream.  It either
3958  * consumes the message or passes it downstream; it never queues a
3959  * a message.
3960  */
3961 static void
3962 icmp_wput(queue_t *q, mblk_t *mp)
3963 {
3964 	uchar_t	*rptr = mp->b_rptr;
3965 	ipha_t	*ipha;
3966 	mblk_t	*mp1;
3967 	int	ip_hdr_length;
3968 #define	tudr ((struct T_unitdata_req *)rptr)
3969 	size_t	ip_len;
3970 	icmp_t	*icmp;
3971 	sin6_t	*sin6;
3972 	sin_t	*sin;
3973 	ipaddr_t	v4dst;
3974 
3975 	icmp = (icmp_t *)q->q_ptr;
3976 	if (icmp->icmp_restricted) {
3977 		icmp_wput_restricted(q, mp);
3978 		return;
3979 	}
3980 
3981 	switch (mp->b_datap->db_type) {
3982 	case M_DATA:
3983 		if (icmp->icmp_hdrincl) {
3984 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
3985 			ipha = (ipha_t *)mp->b_rptr;
3986 			if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH) {
3987 				if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
3988 					BUMP_MIB(&rawip_mib, rawipOutErrors);
3989 					freemsg(mp);
3990 					return;
3991 				}
3992 				ipha = (ipha_t *)mp->b_rptr;
3993 			}
3994 			/*
3995 			 * If this connection was used for v6 (inconceivable!)
3996 			 * or if we have a new destination, then it's time to
3997 			 * figure a new label.
3998 			 */
3999 			if (is_system_labeled() &&
4000 			    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4001 			    V4_PART_OF_V6(icmp->icmp_v6lastdst) !=
4002 			    ipha->ipha_dst) &&
4003 			    !icmp_update_label(q, icmp, mp, ipha->ipha_dst)) {
4004 				return;
4005 			}
4006 			icmp_wput_hdrincl(q, mp, icmp);
4007 			return;
4008 		}
4009 		freemsg(mp);
4010 		return;
4011 	case M_PROTO:
4012 	case M_PCPROTO:
4013 		ip_len = mp->b_wptr - rptr;
4014 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4015 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4016 			if (((union T_primitives *)rptr)->type
4017 			    == T_UNITDATA_REQ)
4018 				break;
4019 		}
4020 		/* FALLTHRU */
4021 	default:
4022 		icmp_wput_other(q, mp);
4023 		return;
4024 	}
4025 
4026 	/* Handle T_UNITDATA_REQ messages here. */
4027 
4028 	if (icmp->icmp_state == TS_UNBND) {
4029 		/* If a port has not been bound to the stream, fail. */
4030 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4031 		icmp_ud_err(q, mp, EPROTO);
4032 		return;
4033 	}
4034 	mp1 = mp->b_cont;
4035 	if (mp1 == NULL) {
4036 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4037 		icmp_ud_err(q, mp, EPROTO);
4038 		return;
4039 	}
4040 
4041 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4042 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4043 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4044 		return;
4045 	}
4046 
4047 	switch (icmp->icmp_family) {
4048 	case AF_INET6:
4049 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4050 		if (!OK_32PTR((char *)sin6) ||
4051 		    tudr->DEST_length != sizeof (sin6_t) ||
4052 		    sin6->sin6_family != AF_INET6) {
4053 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4054 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4055 			return;
4056 		}
4057 
4058 		/* No support for mapped addresses on raw sockets */
4059 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4060 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4061 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4062 			return;
4063 		}
4064 
4065 		/*
4066 		 * Destination is a native IPv6 address.
4067 		 * Send out an IPv6 format packet.
4068 		 */
4069 		icmp_wput_ipv6(q, mp, sin6, tudr->OPT_length);
4070 		return;
4071 
4072 	case AF_INET:
4073 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4074 		if (!OK_32PTR((char *)sin) ||
4075 		    tudr->DEST_length != sizeof (sin_t) ||
4076 		    sin->sin_family != AF_INET) {
4077 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4078 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4079 			return;
4080 		}
4081 		/* Extract and ipaddr */
4082 		v4dst = sin->sin_addr.s_addr;
4083 		break;
4084 
4085 	default:
4086 		ASSERT(0);
4087 	}
4088 
4089 	/*
4090 	 * If options passed in, feed it for verification and handling
4091 	 */
4092 	if (tudr->OPT_length != 0) {
4093 		int error;
4094 
4095 		if (icmp_unitdata_opt_process(q, mp, &error,
4096 		    (uchar_t *)0) < 0) {
4097 			/* failure */
4098 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4099 			icmp_ud_err(q, mp, error);
4100 			return;
4101 		}
4102 		/*
4103 		 * Note: Success in processing options.
4104 		 * mp option buffer represented by
4105 		 * OPT_length/offset now potentially modified
4106 		 * and contain option setting results
4107 		 */
4108 	}
4109 
4110 	if (v4dst == INADDR_ANY)
4111 		v4dst = htonl(INADDR_LOOPBACK);
4112 
4113 	/* Check if our saved options are valid; update if not */
4114 	if (is_system_labeled() &&
4115 	    (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4116 	    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst) &&
4117 	    !icmp_update_label(q, icmp, mp, v4dst)) {
4118 		return;
4119 	}
4120 
4121 	/* Protocol 255 contains full IP headers */
4122 	if (icmp->icmp_hdrincl) {
4123 		freeb(mp);
4124 		icmp_wput_hdrincl(q, mp1, icmp);
4125 		return;
4126 	}
4127 
4128 	/* Add an IP header */
4129 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4130 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
4131 	if ((uchar_t *)ipha < mp1->b_datap->db_base ||
4132 	    mp1->b_datap->db_ref != 1 ||
4133 	    !OK_32PTR(ipha)) {
4134 		if (!(mp1 = allocb(ip_hdr_length + icmp_wroff_extra,
4135 		    BPRI_LO))) {
4136 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4137 			icmp_ud_err(q, mp, ENOMEM);
4138 			return;
4139 		}
4140 		mp1->b_cont = mp->b_cont;
4141 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4142 		mp1->b_wptr = (uchar_t *)ipha;
4143 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4144 	}
4145 #ifdef	_BIG_ENDIAN
4146 	/* Set version, header length, and tos */
4147 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4148 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4149 		icmp->icmp_type_of_service);
4150 	/* Set ttl and protocol */
4151 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4152 #else
4153 	/* Set version, header length, and tos */
4154 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4155 	    ((icmp->icmp_type_of_service << 8) |
4156 		((IP_VERSION << 4) | (ip_hdr_length>>2)));
4157 	/* Set ttl and protocol */
4158 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4159 #endif
4160 	/*
4161 	 * Copy our address into the packet.  If this is zero,
4162 	 * ip will fill in the real source address.
4163 	 */
4164 	IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4165 	ipha->ipha_fragment_offset_and_flags = 0;
4166 
4167 	/*
4168 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4169 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4170 	 * tell IP that the application has sent a complete IP header and not
4171 	 * to compute the transport checksum nor change the DF flag.
4172 	 */
4173 	ipha->ipha_ident = IP_HDR_INCLUDED;
4174 
4175 	/* Finish common formatting of the packet. */
4176 	mp1->b_rptr = (uchar_t *)ipha;
4177 
4178 	ip_len = mp1->b_wptr - (uchar_t *)ipha;
4179 	if (mp1->b_cont != NULL)
4180 		ip_len += msgdsize(mp1->b_cont);
4181 
4182 	/*
4183 	 * Set the length into the IP header.
4184 	 * If the length is greater than the maximum allowed by IP,
4185 	 * then free the message and return. Do not try and send it
4186 	 * as this can cause problems in layers below.
4187 	 */
4188 	if (ip_len > IP_MAXPACKET) {
4189 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4190 		icmp_ud_err(q, mp, EMSGSIZE);
4191 		return;
4192 	}
4193 	ipha->ipha_length = htons((uint16_t)ip_len);
4194 	/*
4195 	 * Copy in the destination address from the T_UNITDATA
4196 	 * request
4197 	 */
4198 	ipha->ipha_dst = v4dst;
4199 
4200 	/*
4201 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4202 	 */
4203 	if (CLASSD(v4dst))
4204 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4205 
4206 	/* Copy in options if any */
4207 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4208 		bcopy(icmp->icmp_ip_snd_options,
4209 		    &ipha[1], icmp->icmp_ip_snd_options_len);
4210 		/*
4211 		 * Massage source route putting first source route in ipha_dst.
4212 		 * Ignore the destination in the T_unitdata_req.
4213 		 */
4214 		(void) ip_massage_options(ipha);
4215 	}
4216 	freeb(mp);
4217 	BUMP_MIB(&rawip_mib, rawipOutDatagrams);
4218 	mblk_setcred(mp1, icmp->icmp_credp);
4219 	putnext(q, mp1);
4220 #undef	ipha
4221 #undef tudr
4222 }
4223 
4224 static boolean_t
4225 icmp_update_label_v6(queue_t *wq, icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4226 {
4227 	int err;
4228 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4229 
4230 	err = tsol_compute_label_v6(DB_CREDDEF(mp, icmp->icmp_credp), dst,
4231 	    opt_storage, icmp->icmp_mac_exempt);
4232 	if (err == 0) {
4233 		err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4234 		    &icmp->icmp_label_len_v6, opt_storage);
4235 	}
4236 	if (err != 0) {
4237 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4238 		DTRACE_PROBE4(
4239 		    tx__ip__log__drop__updatelabel__icmp6,
4240 		    char *, "queue(1) failed to update options(2) on mp(3)",
4241 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
4242 		icmp_ud_err(wq, mp, err);
4243 		return (B_FALSE);
4244 	}
4245 
4246 	icmp->icmp_v6lastdst = *dst;
4247 	return (B_TRUE);
4248 }
4249 
4250 /*
4251  * icmp_wput_ipv6():
4252  * Assumes that icmp_wput did some sanity checking on the destination
4253  * address, but that the label may not yet be correct.
4254  */
4255 void
4256 icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6, t_scalar_t tudr_optlen)
4257 {
4258 	ip6_t			*ip6h;
4259 	ip6i_t			*ip6i;	/* mp1->b_rptr even if no ip6i_t */
4260 	mblk_t			*mp1;
4261 	int			ip_hdr_len = IPV6_HDR_LEN;
4262 	size_t			ip_len;
4263 	icmp_t			*icmp;
4264 	ip6_pkt_t		ipp_s;	/* For ancillary data options */
4265 	ip6_pkt_t		*ipp = &ipp_s;
4266 	ip6_pkt_t		*tipp;
4267 	uint32_t		csum = 0;
4268 	uint_t			ignore = 0;
4269 	uint_t			option_exists = 0, is_sticky = 0;
4270 	uint8_t			*cp;
4271 	uint8_t			*nxthdr_ptr;
4272 	in6_addr_t		ip6_dst;
4273 
4274 	icmp = (icmp_t *)q->q_ptr;
4275 
4276 	/*
4277 	 * If the local address is a mapped address return
4278 	 * an error.
4279 	 * It would be possible to send an IPv6 packet but the
4280 	 * response would never make it back to the application
4281 	 * since it is bound to a mapped address.
4282 	 */
4283 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4284 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4285 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4286 		return;
4287 	}
4288 
4289 	ipp->ipp_fields = 0;
4290 	ipp->ipp_sticky_ignored = 0;
4291 
4292 	/*
4293 	 * If TPI options passed in, feed it for verification and handling
4294 	 */
4295 	if (tudr_optlen != 0) {
4296 		int error;
4297 
4298 		if (icmp_unitdata_opt_process(q, mp, &error,
4299 		    (void *)ipp) < 0) {
4300 			/* failure */
4301 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4302 			icmp_ud_err(q, mp, error);
4303 			return;
4304 		}
4305 		ignore = ipp->ipp_sticky_ignored;
4306 		ASSERT(error == 0);
4307 	}
4308 
4309 	if (sin6->sin6_scope_id != 0 &&
4310 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4311 		/*
4312 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4313 		 * option nor ancillary data.  It needs to be
4314 		 * explicitly set in options_exists.
4315 		 */
4316 		option_exists |= IPPF_SCOPE_ID;
4317 	}
4318 
4319 	/*
4320 	 * Compute the destination address
4321 	 */
4322 	ip6_dst = sin6->sin6_addr;
4323 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
4324 		ip6_dst = ipv6_loopback;
4325 
4326 	/*
4327 	 * If we're not going to the same destination as last time, then
4328 	 * recompute the label required.  This is done in a separate routine to
4329 	 * avoid blowing up our stack here.
4330 	 */
4331 	if (is_system_labeled() &&
4332 	    !IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst) &&
4333 	    !icmp_update_label_v6(q, icmp, mp, &ip6_dst)) {
4334 		return;
4335 	}
4336 
4337 	/*
4338 	 * If there's a security label here, then we ignore any options the
4339 	 * user may try to set.  We keep the peer's label as a hidden sticky
4340 	 * option.
4341 	 */
4342 	if (icmp->icmp_label_len_v6 > 0) {
4343 		ignore &= ~IPPF_HOPOPTS;
4344 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
4345 	}
4346 
4347 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
4348 	    (ipp->ipp_fields == 0)) {
4349 		/* No sticky options nor ancillary data. */
4350 		goto no_options;
4351 	}
4352 
4353 	/*
4354 	 * Go through the options figuring out where each is going to
4355 	 * come from and build two masks.  The first mask indicates if
4356 	 * the option exists at all.  The second mask indicates if the
4357 	 * option is sticky or ancillary.
4358 	 */
4359 	if (!(ignore & IPPF_HOPOPTS)) {
4360 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
4361 			option_exists |= IPPF_HOPOPTS;
4362 			ip_hdr_len += ipp->ipp_hopoptslen;
4363 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
4364 			option_exists |= IPPF_HOPOPTS;
4365 			is_sticky |= IPPF_HOPOPTS;
4366 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen;
4367 		}
4368 	}
4369 
4370 	if (!(ignore & IPPF_RTHDR)) {
4371 		if (ipp->ipp_fields & IPPF_RTHDR) {
4372 			option_exists |= IPPF_RTHDR;
4373 			ip_hdr_len += ipp->ipp_rthdrlen;
4374 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
4375 			option_exists |= IPPF_RTHDR;
4376 			is_sticky |= IPPF_RTHDR;
4377 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
4378 		}
4379 	}
4380 
4381 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
4382 		/*
4383 		 * Need to have a router header to use these.
4384 		 */
4385 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
4386 			option_exists |= IPPF_RTDSTOPTS;
4387 			ip_hdr_len += ipp->ipp_rtdstoptslen;
4388 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
4389 			option_exists |= IPPF_RTDSTOPTS;
4390 			is_sticky |= IPPF_RTDSTOPTS;
4391 			ip_hdr_len +=
4392 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
4393 		}
4394 	}
4395 
4396 	if (!(ignore & IPPF_DSTOPTS)) {
4397 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
4398 			option_exists |= IPPF_DSTOPTS;
4399 			ip_hdr_len += ipp->ipp_dstoptslen;
4400 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
4401 			option_exists |= IPPF_DSTOPTS;
4402 			is_sticky |= IPPF_DSTOPTS;
4403 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
4404 		}
4405 	}
4406 
4407 	if (!(ignore & IPPF_IFINDEX)) {
4408 		if (ipp->ipp_fields & IPPF_IFINDEX) {
4409 			option_exists |= IPPF_IFINDEX;
4410 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
4411 			option_exists |= IPPF_IFINDEX;
4412 			is_sticky |= IPPF_IFINDEX;
4413 		}
4414 	}
4415 
4416 	if (!(ignore & IPPF_ADDR)) {
4417 		if (ipp->ipp_fields & IPPF_ADDR) {
4418 			option_exists |= IPPF_ADDR;
4419 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
4420 			option_exists |= IPPF_ADDR;
4421 			is_sticky |= IPPF_ADDR;
4422 		}
4423 	}
4424 
4425 	if (!(ignore & IPPF_DONTFRAG)) {
4426 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
4427 			option_exists |= IPPF_DONTFRAG;
4428 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
4429 			option_exists |= IPPF_DONTFRAG;
4430 			is_sticky |= IPPF_DONTFRAG;
4431 		}
4432 	}
4433 
4434 	if (!(ignore & IPPF_USE_MIN_MTU)) {
4435 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
4436 			option_exists |= IPPF_USE_MIN_MTU;
4437 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
4438 		    IPPF_USE_MIN_MTU) {
4439 			option_exists |= IPPF_USE_MIN_MTU;
4440 			is_sticky |= IPPF_USE_MIN_MTU;
4441 		}
4442 	}
4443 
4444 	if (!(ignore & IPPF_NEXTHOP)) {
4445 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
4446 			option_exists |= IPPF_NEXTHOP;
4447 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
4448 			option_exists |= IPPF_NEXTHOP;
4449 			is_sticky |= IPPF_NEXTHOP;
4450 		}
4451 	}
4452 
4453 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
4454 		option_exists |= IPPF_HOPLIMIT;
4455 	/* IPV6_HOPLIMIT can never be sticky */
4456 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
4457 
4458 	if (!(ignore & IPPF_UNICAST_HOPS) &&
4459 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
4460 		option_exists |= IPPF_UNICAST_HOPS;
4461 		is_sticky |= IPPF_UNICAST_HOPS;
4462 	}
4463 
4464 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
4465 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
4466 		option_exists |= IPPF_MULTICAST_HOPS;
4467 		is_sticky |= IPPF_MULTICAST_HOPS;
4468 	}
4469 
4470 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
4471 		/* This is a sticky socket option only */
4472 		option_exists |= IPPF_NO_CKSUM;
4473 		is_sticky |= IPPF_NO_CKSUM;
4474 	}
4475 
4476 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
4477 		/* This is a sticky socket option only */
4478 		option_exists |= IPPF_RAW_CKSUM;
4479 		is_sticky |= IPPF_RAW_CKSUM;
4480 	}
4481 
4482 	if (!(ignore & IPPF_TCLASS)) {
4483 		if (ipp->ipp_fields & IPPF_TCLASS) {
4484 			option_exists |= IPPF_TCLASS;
4485 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
4486 			option_exists |= IPPF_TCLASS;
4487 			is_sticky |= IPPF_TCLASS;
4488 		}
4489 	}
4490 
4491 no_options:
4492 
4493 	/*
4494 	 * If any options carried in the ip6i_t were specified, we
4495 	 * need to account for the ip6i_t in the data we'll be sending
4496 	 * down.
4497 	 */
4498 	if (option_exists & IPPF_HAS_IP6I)
4499 		ip_hdr_len += sizeof (ip6i_t);
4500 
4501 	/* check/fix buffer config, setup pointers into it */
4502 	mp1 = mp->b_cont;
4503 	ip6h = (ip6_t *)&mp1->b_rptr[-ip_hdr_len];
4504 	if ((mp1->b_datap->db_ref != 1) ||
4505 	    ((unsigned char *)ip6h < mp1->b_datap->db_base) ||
4506 	    !OK_32PTR(ip6h)) {
4507 		/* Try to get everything in a single mblk next time */
4508 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
4509 			icmp->icmp_max_hdr_len = ip_hdr_len;
4510 			(void) mi_set_sth_wroff(RD(q),
4511 			    icmp->icmp_max_hdr_len + icmp_wroff_extra);
4512 		}
4513 		mp1 = allocb(ip_hdr_len + icmp_wroff_extra, BPRI_LO);
4514 		if (!mp1) {
4515 			BUMP_MIB(&rawip_mib, rawipOutErrors);
4516 			icmp_ud_err(q, mp, ENOMEM);
4517 			return;
4518 		}
4519 		mp1->b_cont = mp->b_cont;
4520 		mp1->b_wptr = mp1->b_datap->db_lim;
4521 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
4522 	}
4523 	mp1->b_rptr = (unsigned char *)ip6h;
4524 	ip6i = (ip6i_t *)ip6h;
4525 
4526 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
4527 	if (option_exists & IPPF_HAS_IP6I) {
4528 		ip6h = (ip6_t *)&ip6i[1];
4529 		ip6i->ip6i_flags = 0;
4530 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
4531 
4532 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
4533 		if (option_exists & IPPF_SCOPE_ID) {
4534 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4535 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
4536 		} else if (option_exists & IPPF_IFINDEX) {
4537 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
4538 			ASSERT(tipp->ipp_ifindex != 0);
4539 			ip6i->ip6i_flags |= IP6I_IFINDEX;
4540 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
4541 		}
4542 
4543 		if (option_exists & IPPF_RAW_CKSUM) {
4544 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
4545 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
4546 		}
4547 
4548 		if (option_exists & IPPF_NO_CKSUM) {
4549 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
4550 		}
4551 
4552 		if (option_exists & IPPF_ADDR) {
4553 			/*
4554 			 * Enable per-packet source address verification if
4555 			 * IPV6_PKTINFO specified the source address.
4556 			 * ip6_src is set in the transport's _wput function.
4557 			 */
4558 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
4559 		}
4560 
4561 		if (option_exists & IPPF_DONTFRAG) {
4562 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
4563 		}
4564 
4565 		if (option_exists & IPPF_USE_MIN_MTU) {
4566 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
4567 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
4568 		}
4569 
4570 		if (option_exists & IPPF_NEXTHOP) {
4571 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
4572 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
4573 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
4574 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
4575 		}
4576 
4577 		/*
4578 		 * tell IP this is an ip6i_t private header
4579 		 */
4580 		ip6i->ip6i_nxt = IPPROTO_RAW;
4581 	}
4582 
4583 	/* Initialize IPv6 header */
4584 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
4585 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
4586 
4587 	/* Set the hoplimit of the outgoing packet. */
4588 	if (option_exists & IPPF_HOPLIMIT) {
4589 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
4590 		ip6h->ip6_hops = ipp->ipp_hoplimit;
4591 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
4592 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4593 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
4594 		if (option_exists & IPPF_MULTICAST_HOPS)
4595 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
4596 	} else {
4597 		ip6h->ip6_hops = icmp->icmp_ttl;
4598 		if (option_exists & IPPF_UNICAST_HOPS)
4599 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
4600 	}
4601 
4602 	if (option_exists & IPPF_ADDR) {
4603 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
4604 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
4605 		ip6h->ip6_src = tipp->ipp_addr;
4606 	} else {
4607 		/*
4608 		 * The source address was not set using IPV6_PKTINFO.
4609 		 * First look at the bound source.
4610 		 * If unspecified fallback to __sin6_src_id.
4611 		 */
4612 		ip6h->ip6_src = icmp->icmp_v6src;
4613 		if (sin6->__sin6_src_id != 0 &&
4614 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
4615 			ip_srcid_find_id(sin6->__sin6_src_id,
4616 			    &ip6h->ip6_src, icmp->icmp_zoneid);
4617 		}
4618 	}
4619 
4620 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
4621 	cp = (uint8_t *)&ip6h[1];
4622 
4623 	/*
4624 	 * Here's where we have to start stringing together
4625 	 * any extension headers in the right order:
4626 	 * Hop-by-hop, destination, routing, and final destination opts.
4627 	 */
4628 	if (option_exists & IPPF_HOPOPTS) {
4629 		/* Hop-by-hop options */
4630 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
4631 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
4632 
4633 		*nxthdr_ptr = IPPROTO_HOPOPTS;
4634 		nxthdr_ptr = &hbh->ip6h_nxt;
4635 
4636 		bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
4637 		cp += tipp->ipp_hopoptslen;
4638 	}
4639 	/*
4640 	 * En-route destination options
4641 	 * Only do them if there's a routing header as well
4642 	 */
4643 	if (option_exists & IPPF_RTDSTOPTS) {
4644 		ip6_dest_t *dst = (ip6_dest_t *)cp;
4645 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
4646 
4647 		*nxthdr_ptr = IPPROTO_DSTOPTS;
4648 		nxthdr_ptr = &dst->ip6d_nxt;
4649 
4650 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
4651 		cp += tipp->ipp_rtdstoptslen;
4652 	}
4653 	/*
4654 	 * Routing header next
4655 	 */
4656 	if (option_exists & IPPF_RTHDR) {
4657 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
4658 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
4659 
4660 		*nxthdr_ptr = IPPROTO_ROUTING;
4661 		nxthdr_ptr = &rt->ip6r_nxt;
4662 
4663 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
4664 		cp += tipp->ipp_rthdrlen;
4665 	}
4666 	/*
4667 	 * Do ultimate destination options
4668 	 */
4669 	if (option_exists & IPPF_DSTOPTS) {
4670 		ip6_dest_t *dest = (ip6_dest_t *)cp;
4671 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
4672 
4673 		*nxthdr_ptr = IPPROTO_DSTOPTS;
4674 		nxthdr_ptr = &dest->ip6d_nxt;
4675 
4676 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
4677 		cp += tipp->ipp_dstoptslen;
4678 	}
4679 
4680 	/*
4681 	 * Now set the last header pointer to the proto passed in
4682 	 */
4683 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
4684 	*nxthdr_ptr = icmp->icmp_proto;
4685 
4686 	/*
4687 	 * Copy in the destination address
4688 	 */
4689 	ip6h->ip6_dst = ip6_dst;
4690 
4691 	ip6h->ip6_vcf =
4692 		(IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
4693 		(sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
4694 
4695 	if (option_exists & IPPF_TCLASS) {
4696 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
4697 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
4698 		    tipp->ipp_tclass);
4699 	}
4700 	if (option_exists & IPPF_RTHDR) {
4701 		ip6_rthdr_t	*rth;
4702 
4703 		/*
4704 		 * Perform any processing needed for source routing.
4705 		 * We know that all extension headers will be in the same mblk
4706 		 * as the IPv6 header.
4707 		 */
4708 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
4709 		if (rth != NULL && rth->ip6r_segleft != 0) {
4710 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
4711 				/*
4712 				 * Drop packet - only support Type 0 routing.
4713 				 * Notify the application as well.
4714 				 */
4715 				icmp_ud_err(q, mp, EPROTO);
4716 				BUMP_MIB(&rawip_mib, rawipOutErrors);
4717 				return;
4718 			}
4719 			/*
4720 			 * rth->ip6r_len is twice the number of
4721 			 * addresses in the header
4722 			 */
4723 			if (rth->ip6r_len & 0x1) {
4724 				icmp_ud_err(q, mp, EPROTO);
4725 				BUMP_MIB(&rawip_mib, rawipOutErrors);
4726 				return;
4727 			}
4728 			/*
4729 			 * Shuffle the routing header and ip6_dst
4730 			 * addresses, and get the checksum difference
4731 			 * between the first hop (in ip6_dst) and
4732 			 * the destination (in the last routing hdr entry).
4733 			 */
4734 			csum = ip_massage_options_v6(ip6h, rth);
4735 			/*
4736 			 * Verify that the first hop isn't a mapped address.
4737 			 * Routers along the path need to do this verification
4738 			 * for subsequent hops.
4739 			 */
4740 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
4741 				icmp_ud_err(q, mp, EADDRNOTAVAIL);
4742 				BUMP_MIB(&rawip_mib, rawipOutErrors);
4743 				return;
4744 			}
4745 		}
4746 	}
4747 
4748 	ip_len = mp1->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
4749 	if (mp1->b_cont != NULL)
4750 		ip_len += msgdsize(mp1->b_cont);
4751 
4752 	/*
4753 	 * Set the length into the IP header.
4754 	 * If the length is greater than the maximum allowed by IP,
4755 	 * then free the message and return. Do not try and send it
4756 	 * as this can cause problems in layers below.
4757 	 */
4758 	if (ip_len > IP_MAXPACKET) {
4759 		BUMP_MIB(&rawip_mib, rawipOutErrors);
4760 		icmp_ud_err(q, mp, EMSGSIZE);
4761 		return;
4762 	}
4763 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
4764 		uint_t	cksum_off;	/* From ip6i == mp1->b_rptr */
4765 		uint16_t *cksum_ptr;
4766 		uint_t	ext_hdrs_len;
4767 
4768 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
4769 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
4770 		    icmp->icmp_checksum_off == 2);
4771 
4772 		/*
4773 		 * We make it easy for IP to include our pseudo header
4774 		 * by putting our length in uh_checksum, modified (if
4775 		 * we have a routing header) by the checksum difference
4776 		 * between the ultimate destination and first hop addresses.
4777 		 * Note: ICMPv6 must always checksum the packet.
4778 		 */
4779 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
4780 		if (cksum_off + sizeof (uint16_t) > mp1->b_wptr - mp1->b_rptr) {
4781 			if (!pullupmsg(mp1, cksum_off + sizeof (uint16_t))) {
4782 				BUMP_MIB(&rawip_mib, rawipOutErrors);
4783 				freemsg(mp);
4784 				return;
4785 			}
4786 			ip6i = (ip6i_t *)mp1->b_rptr;
4787 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
4788 				ip6h = (ip6_t *)&ip6i[1];
4789 			else
4790 				ip6h = (ip6_t *)ip6i;
4791 		}
4792 		/* Add payload length to checksum */
4793 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
4794 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
4795 		csum += htons(ip_len - ext_hdrs_len);
4796 
4797 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
4798 		csum = (csum & 0xFFFF) + (csum >> 16);
4799 		*cksum_ptr = (uint16_t)csum;
4800 	}
4801 
4802 #ifdef _LITTLE_ENDIAN
4803 	ip_len = htons(ip_len);
4804 #endif
4805 	ip6h->ip6_plen = (uint16_t)ip_len;
4806 
4807 	freeb(mp);
4808 
4809 	/* We're done. Pass the packet to IP */
4810 	BUMP_MIB(&rawip_mib, rawipOutDatagrams);
4811 	mblk_setcred(mp1, icmp->icmp_credp);
4812 	putnext(q, mp1);
4813 }
4814 
4815 static void
4816 icmp_wput_other(queue_t *q, mblk_t *mp)
4817 {
4818 	uchar_t	*rptr = mp->b_rptr;
4819 	struct iocblk *iocp;
4820 #define	tudr ((struct T_unitdata_req *)rptr)
4821 	icmp_t	*icmp;
4822 	cred_t *cr;
4823 
4824 	icmp = (icmp_t *)q->q_ptr;
4825 
4826 	cr = DB_CREDDEF(mp, icmp->icmp_credp);
4827 
4828 	switch (mp->b_datap->db_type) {
4829 	case M_PROTO:
4830 	case M_PCPROTO:
4831 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
4832 			/*
4833 			 * If the message does not contain a PRIM_type,
4834 			 * throw it away.
4835 			 */
4836 			freemsg(mp);
4837 			return;
4838 		}
4839 		switch (((union T_primitives *)rptr)->type) {
4840 		case T_ADDR_REQ:
4841 			icmp_addr_req(q, mp);
4842 			return;
4843 		case O_T_BIND_REQ:
4844 		case T_BIND_REQ:
4845 			qwriter(q, mp, icmp_bind, PERIM_OUTER);
4846 			return;
4847 		case T_CONN_REQ:
4848 			icmp_connect(q, mp);
4849 			return;
4850 		case T_CAPABILITY_REQ:
4851 			icmp_capability_req(q, mp);
4852 			return;
4853 		case T_INFO_REQ:
4854 			icmp_info_req(q, mp);
4855 			return;
4856 		case T_UNITDATA_REQ:
4857 			/*
4858 			 * If a T_UNITDATA_REQ gets here, the address must
4859 			 * be bad.  Valid T_UNITDATA_REQs are found above
4860 			 * and break to below this switch.
4861 			 */
4862 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4863 			return;
4864 		case T_UNBIND_REQ:
4865 			icmp_unbind(q, mp);
4866 			return;
4867 
4868 		case T_SVR4_OPTMGMT_REQ:
4869 			if (!snmpcom_req(q, mp, icmp_snmp_set, icmp_snmp_get,
4870 			    cr))
4871 				/* Only IP can return anything meaningful */
4872 				(void) svr4_optcom_req(q, mp, cr,
4873 				    &icmp_opt_obj);
4874 			return;
4875 
4876 		case T_OPTMGMT_REQ:
4877 			/* Only IP can return anything meaningful */
4878 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj);
4879 			return;
4880 
4881 		case T_DISCON_REQ:
4882 			icmp_disconnect(q, mp);
4883 			return;
4884 
4885 		/* The following TPI message is not supported by icmp. */
4886 		case O_T_CONN_RES:
4887 		case T_CONN_RES:
4888 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
4889 			return;
4890 
4891 		/* The following 3 TPI requests are illegal for icmp. */
4892 		case T_DATA_REQ:
4893 		case T_EXDATA_REQ:
4894 		case T_ORDREL_REQ:
4895 			freemsg(mp);
4896 			(void) putctl1(RD(q), M_ERROR, EPROTO);
4897 			return;
4898 		default:
4899 			break;
4900 		}
4901 		break;
4902 	case M_IOCTL:
4903 		iocp = (struct iocblk *)mp->b_rptr;
4904 		switch (iocp->ioc_cmd) {
4905 		case TI_GETPEERNAME:
4906 			if (icmp->icmp_state != TS_DATA_XFER) {
4907 				/*
4908 				 * If a default destination address has not
4909 				 * been associated with the stream, then we
4910 				 * don't know the peer's name.
4911 				 */
4912 				iocp->ioc_error = ENOTCONN;
4913 			    err_ret:;
4914 				iocp->ioc_count = 0;
4915 				mp->b_datap->db_type = M_IOCACK;
4916 				qreply(q, mp);
4917 				return;
4918 			}
4919 			/* FALLTHRU */
4920 		case TI_GETMYNAME:
4921 			/*
4922 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
4923 			 * need to copyin the user's strbuf structure.
4924 			 * Processing will continue in the M_IOCDATA case
4925 			 * below.
4926 			 */
4927 			mi_copyin(q, mp, NULL,
4928 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
4929 			return;
4930 		case ND_SET:
4931 			/* nd_getset performs the necessary error checking */
4932 		case ND_GET:
4933 			if (nd_getset(q, icmp_g_nd, mp)) {
4934 				qreply(q, mp);
4935 				return;
4936 			}
4937 			break;
4938 		default:
4939 			break;
4940 		}
4941 		break;
4942 	case M_IOCDATA:
4943 		icmp_wput_iocdata(q, mp);
4944 		return;
4945 	default:
4946 		break;
4947 	}
4948 	putnext(q, mp);
4949 }
4950 
4951 /*
4952  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
4953  * messages.
4954  */
4955 static void
4956 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
4957 {
4958 	mblk_t	*mp1;
4959 	STRUCT_HANDLE(strbuf, sb);
4960 	icmp_t	*icmp;
4961 	in6_addr_t	v6addr;
4962 	ipaddr_t	v4addr;
4963 	uint32_t	flowinfo = 0;
4964 	int		addrlen;
4965 
4966 	/* Make sure it is one of ours. */
4967 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
4968 	case TI_GETMYNAME:
4969 	case TI_GETPEERNAME:
4970 		break;
4971 	default:
4972 		putnext(q, mp);
4973 		return;
4974 	}
4975 	switch (mi_copy_state(q, mp, &mp1)) {
4976 	case -1:
4977 		return;
4978 	case MI_COPY_CASE(MI_COPY_IN, 1):
4979 		break;
4980 	case MI_COPY_CASE(MI_COPY_OUT, 1):
4981 		/*
4982 		 * The address has been copied out, so now
4983 		 * copyout the strbuf.
4984 		 */
4985 		mi_copyout(q, mp);
4986 		return;
4987 	case MI_COPY_CASE(MI_COPY_OUT, 2):
4988 		/*
4989 		 * The address and strbuf have been copied out.
4990 		 * We're done, so just acknowledge the original
4991 		 * M_IOCTL.
4992 		 */
4993 		mi_copy_done(q, mp, 0);
4994 		return;
4995 	default:
4996 		/*
4997 		 * Something strange has happened, so acknowledge
4998 		 * the original M_IOCTL with an EPROTO error.
4999 		 */
5000 		mi_copy_done(q, mp, EPROTO);
5001 		return;
5002 	}
5003 	/*
5004 	 * Now we have the strbuf structure for TI_GETMYNAME
5005 	 * and TI_GETPEERNAME.  Next we copyout the requested
5006 	 * address and then we'll copyout the strbuf.
5007 	 */
5008 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5009 	    (void *)mp1->b_rptr);
5010 	icmp = (icmp_t *)q->q_ptr;
5011 	if (icmp->icmp_family == AF_INET)
5012 		addrlen = sizeof (sin_t);
5013 	else
5014 		addrlen = sizeof (sin6_t);
5015 
5016 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5017 		mi_copy_done(q, mp, EINVAL);
5018 		return;
5019 	}
5020 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5021 	case TI_GETMYNAME:
5022 		if (icmp->icmp_family == AF_INET) {
5023 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5024 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
5025 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5026 				v4addr = V4_PART_OF_V6(icmp->icmp_v6src);
5027 			} else {
5028 				/*
5029 				 * INADDR_ANY
5030 				 * icmp_v6src is not set, we might be bound to
5031 				 * broadcast/multicast. Use icmp_bound_v6src as
5032 				 * local address instead (that could
5033 				 * also still be INADDR_ANY)
5034 				 */
5035 				v4addr = V4_PART_OF_V6(icmp->icmp_bound_v6src);
5036 			}
5037 		} else {
5038 			/* icmp->icmp_family == AF_INET6 */
5039 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
5040 				v6addr = icmp->icmp_v6src;
5041 			} else {
5042 				/*
5043 				 * UNSPECIFIED
5044 				 * icmp_v6src is not set, we might be bound to
5045 				 * broadcast/multicast. Use icmp_bound_v6src as
5046 				 * local address instead (that could
5047 				 * also still be UNSPECIFIED)
5048 				 */
5049 				v6addr = icmp->icmp_bound_v6src;
5050 			}
5051 		}
5052 		break;
5053 	case TI_GETPEERNAME:
5054 		if (icmp->icmp_family == AF_INET) {
5055 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
5056 			v4addr = V4_PART_OF_V6(icmp->icmp_v6dst);
5057 		} else {
5058 			/* icmp->icmp_family == AF_INET6) */
5059 			v6addr = icmp->icmp_v6dst;
5060 			flowinfo = icmp->icmp_flowinfo;
5061 		}
5062 		break;
5063 	default:
5064 		mi_copy_done(q, mp, EPROTO);
5065 		return;
5066 	}
5067 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5068 	if (!mp1)
5069 		return;
5070 
5071 	if (icmp->icmp_family == AF_INET) {
5072 		sin_t *sin;
5073 
5074 		STRUCT_FSET(sb, len, (int)sizeof (sin_t));
5075 		sin = (sin_t *)mp1->b_rptr;
5076 		mp1->b_wptr = (uchar_t *)&sin[1];
5077 		*sin = sin_null;
5078 		sin->sin_family = AF_INET;
5079 		sin->sin_addr.s_addr = v4addr;
5080 	} else {
5081 		/* icmp->icmp_family == AF_INET6 */
5082 		sin6_t *sin6;
5083 
5084 		ASSERT(icmp->icmp_family == AF_INET6);
5085 		STRUCT_FSET(sb, len, (int)sizeof (sin6_t));
5086 		sin6 = (sin6_t *)mp1->b_rptr;
5087 		mp1->b_wptr = (uchar_t *)&sin6[1];
5088 		*sin6 = sin6_null;
5089 		sin6->sin6_family = AF_INET6;
5090 		sin6->sin6_flowinfo = flowinfo;
5091 		sin6->sin6_addr = v6addr;
5092 	}
5093 	/* Copy out the address */
5094 	mi_copyout(q, mp);
5095 }
5096 
5097 /*
5098  * Only allow MIB requests and M_FLUSHes to pass.
5099  * All other messages are nacked or dropped.
5100  */
5101 static void
5102 icmp_wput_restricted(queue_t *q, mblk_t *mp)
5103 {
5104 	cred_t	*cr;
5105 	icmp_t	*icmp;
5106 
5107 	switch (DB_TYPE(mp)) {
5108 	case M_PROTO:
5109 	case M_PCPROTO:
5110 		if (MBLKL(mp) < sizeof (t_scalar_t)) {
5111 			freemsg(mp);
5112 			return;
5113 		}
5114 		icmp = (icmp_t *)q->q_ptr;
5115 		cr = DB_CREDDEF(mp, icmp->icmp_credp);
5116 
5117 		switch (((union T_primitives *)mp->b_rptr)->type) {
5118 		case T_SVR4_OPTMGMT_REQ:
5119 			if (!snmpcom_req(q, mp,
5120 			    icmp_snmp_set, icmp_snmp_get, cr))
5121 				(void) svr4_optcom_req(q, mp, cr,
5122 				    &icmp_opt_obj);
5123 			return;
5124 		case T_OPTMGMT_REQ:
5125 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj);
5126 			return;
5127 		default:
5128 			icmp_err_ack(q, mp, TSYSERR, ENOTSUP);
5129 			return;
5130 		}
5131 		/* NOTREACHED */
5132 	case M_IOCTL:
5133 		miocnak(q, mp, 0, ENOTSUP);
5134 		break;
5135 	case M_FLUSH:
5136 		putnext(q, mp);
5137 		break;
5138 	default:
5139 		freemsg(mp);
5140 		break;
5141 	}
5142 }
5143 
5144 static int
5145 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5146     void *thisdg_attrs)
5147 {
5148 	icmp_t	*icmp;
5149 	struct T_unitdata_req *udreqp;
5150 	int is_absreq_failure;
5151 	cred_t *cr;
5152 
5153 	icmp = (icmp_t *)q->q_ptr;
5154 
5155 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5156 	*errorp = 0;
5157 
5158 	cr = DB_CREDDEF(mp, icmp->icmp_credp);
5159 
5160 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5161 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5162 	    thisdg_attrs, &is_absreq_failure);
5163 
5164 	if (*errorp != 0) {
5165 		/*
5166 		 * Note: No special action needed in this
5167 		 * module for "is_absreq_failure"
5168 		 */
5169 		return (-1);		/* failure */
5170 	}
5171 	ASSERT(is_absreq_failure == 0);
5172 	return (0);	/* success */
5173 }
5174 
5175 void
5176 icmp_ddi_init(void)
5177 {
5178 	ICMP6_MAJ = ddi_name_to_major(ICMP6);
5179 	icmp_max_optsize =
5180 	    optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5181 		icmp_opt_obj.odb_opt_arr_cnt);
5182 
5183 	(void) icmp_param_register(icmp_param_arr, A_CNT(icmp_param_arr));
5184 
5185 	rawip_kstat_init();
5186 }
5187 
5188 void
5189 icmp_ddi_destroy(void)
5190 {
5191 	nd_free(&icmp_g_nd);
5192 
5193 	rawip_kstat_fini();
5194 }
5195 
5196 static void
5197 rawip_kstat_init(void) {
5198 
5199 	rawip_named_kstat_t template = {
5200 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5201 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5202 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5203 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5204 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5205 	};
5206 
5207 	rawip_mibkp = kstat_create("icmp", 0, "rawip", "mib2",
5208 					KSTAT_TYPE_NAMED,
5209 					NUM_OF_FIELDS(rawip_named_kstat_t),
5210 					0);
5211 	if (rawip_mibkp == NULL)
5212 		return;
5213 
5214 	bcopy(&template, rawip_mibkp->ks_data, sizeof (template));
5215 
5216 	rawip_mibkp->ks_update = rawip_kstat_update;
5217 
5218 	kstat_install(rawip_mibkp);
5219 }
5220 
5221 static void
5222 rawip_kstat_fini(void) {
5223 	if (rawip_mibkp) {
5224 		kstat_delete(rawip_mibkp);
5225 		rawip_mibkp = NULL;
5226 	}
5227 }
5228 
5229 static int
5230 rawip_kstat_update(kstat_t *kp, int rw) {
5231 	rawip_named_kstat_t *rawipkp;
5232 
5233 	if ((kp == NULL) || (kp->ks_data == NULL))
5234 		return (EIO);
5235 
5236 	if (rw == KSTAT_WRITE)
5237 		return (EACCES);
5238 
5239 	rawipkp = (rawip_named_kstat_t *)kp->ks_data;
5240 
5241 	rawipkp->inDatagrams.value.ui32 =	rawip_mib.rawipInDatagrams;
5242 	rawipkp->inCksumErrs.value.ui32 =	rawip_mib.rawipInCksumErrs;
5243 	rawipkp->inErrors.value.ui32 =		rawip_mib.rawipInErrors;
5244 	rawipkp->outDatagrams.value.ui32 =	rawip_mib.rawipOutDatagrams;
5245 	rawipkp->outErrors.value.ui32 =		rawip_mib.rawipOutErrors;
5246 
5247 	return (0);
5248 }
5249