xref: /titanic_41/usr/src/uts/common/inet/ip/icmp.c (revision 9e86db79b7d1bbc5f2f04e99954cbd5eae0e22bb)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/stropts.h>
30 #include <sys/strlog.h>
31 #include <sys/strsun.h>
32 #define	_SUN_TPI_VERSION 2
33 #include <sys/tihdr.h>
34 #include <sys/timod.h>
35 #include <sys/ddi.h>
36 #include <sys/sunddi.h>
37 #include <sys/strsubr.h>
38 #include <sys/cmn_err.h>
39 #include <sys/debug.h>
40 #include <sys/kmem.h>
41 #include <sys/policy.h>
42 #include <sys/priv.h>
43 #include <sys/zone.h>
44 #include <sys/time.h>
45 
46 #include <sys/sockio.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/isa_defs.h>
50 #include <sys/suntpi.h>
51 #include <sys/xti_inet.h>
52 #include <sys/netstack.h>
53 
54 #include <net/route.h>
55 #include <net/if.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/ip6.h>
59 #include <netinet/icmp6.h>
60 #include <inet/common.h>
61 #include <inet/ip.h>
62 #include <inet/ip6.h>
63 #include <inet/proto_set.h>
64 #include <inet/nd.h>
65 #include <inet/optcom.h>
66 #include <inet/snmpcom.h>
67 #include <inet/kstatcom.h>
68 #include <inet/rawip_impl.h>
69 
70 #include <netinet/ip_mroute.h>
71 #include <inet/tcp.h>
72 #include <net/pfkeyv2.h>
73 #include <inet/ipsec_info.h>
74 #include <inet/ipclassifier.h>
75 
76 #include <sys/tsol/label.h>
77 #include <sys/tsol/tnet.h>
78 
79 #include <inet/ip_ire.h>
80 #include <inet/ip_if.h>
81 
82 #include <inet/ip_impl.h>
83 #include <sys/disp.h>
84 
85 /*
86  * Synchronization notes:
87  *
88  * RAWIP is MT and uses the usual kernel synchronization primitives. There is
89  * locks, which is icmp_rwlock. We also use conn_lock when updating things
90  * which affect the IP classifier lookup.
91  * The lock order is icmp_rwlock -> conn_lock.
92  *
93  * The icmp_rwlock:
94  * This protects most of the other fields in the icmp_t. The exact list of
95  * fields which are protected by each of the above locks is documented in
96  * the icmp_t structure definition.
97  *
98  * Plumbing notes:
99  * ICMP is always a device driver. For compatibility with mibopen() code
100  * it is possible to I_PUSH "icmp", but that results in pushing a passthrough
101  * dummy module.
102  */
103 
104 static void	icmp_addr_req(queue_t *q, mblk_t *mp);
105 static void	icmp_tpi_bind(queue_t *q, mblk_t *mp);
106 static int	icmp_bind_proto(conn_t *connp);
107 static int	icmp_build_hdrs(icmp_t *icmp);
108 static void	icmp_capability_req(queue_t *q, mblk_t *mp);
109 static int	icmp_close(queue_t *q, int flags);
110 static void	icmp_tpi_connect(queue_t *q, mblk_t *mp);
111 static void	icmp_tpi_disconnect(queue_t *q, mblk_t *mp);
112 static void	icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
113 		    int sys_error);
114 static void	icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
115 		    t_scalar_t t_error, int sys_error);
116 static void	icmp_icmp_error(conn_t *connp, mblk_t *mp);
117 static void	icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp);
118 static void	icmp_info_req(queue_t *q, mblk_t *mp);
119 static void	icmp_input(void *, mblk_t *, void *);
120 static conn_t 	*icmp_open(int family, cred_t *credp, int *err, int flags);
121 static int	icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
122 		    cred_t *credp);
123 static int	icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
124 		    cred_t *credp);
125 static int	icmp_unitdata_opt_process(queue_t *q, mblk_t *mp,
126 		    int *errorp, void *thisdg_attrs);
127 static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
128 int		icmp_opt_set(conn_t *connp, uint_t optset_context,
129 		    int level, int name, uint_t inlen,
130 		    uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
131 		    void *thisdg_attrs, cred_t *cr);
132 int		icmp_opt_get(conn_t *connp, int level, int name,
133 		    uchar_t *ptr);
134 static int	icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
135 static boolean_t icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt);
136 static int	icmp_param_set(queue_t *q, mblk_t *mp, char *value,
137 		    caddr_t cp, cred_t *cr);
138 static int	icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
139 		    uchar_t *ptr, int len);
140 static void	icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err);
141 static void	icmp_tpi_unbind(queue_t *q, mblk_t *mp);
142 static int	icmp_update_label(icmp_t *icmp, mblk_t *mp, ipaddr_t dst);
143 static void	icmp_wput(queue_t *q, mblk_t *mp);
144 static void	icmp_wput_fallback(queue_t *q, mblk_t *mp);
145 static int	raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp,
146 		    sin6_t *sin6, ip6_pkt_t *ipp);
147 static int	raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp,
148 		    ipaddr_t v4dst, ip4_pkt_t *pktinfop);
149 static void	icmp_wput_other(queue_t *q, mblk_t *mp);
150 static void	icmp_wput_iocdata(queue_t *q, mblk_t *mp);
151 static void	icmp_wput_restricted(queue_t *q, mblk_t *mp);
152 static void	icmp_ulp_recv(conn_t *, mblk_t *);
153 
154 static void	*rawip_stack_init(netstackid_t stackid, netstack_t *ns);
155 static void	rawip_stack_fini(netstackid_t stackid, void *arg);
156 
157 static void	*rawip_kstat_init(netstackid_t stackid);
158 static void	rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp);
159 static int	rawip_kstat_update(kstat_t *kp, int rw);
160 static void	rawip_stack_shutdown(netstackid_t stackid, void *arg);
161 static int	rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa,
162 		    uint_t *salenp);
163 static int	rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa,
164 		    uint_t *salenp);
165 
166 int		rawip_getsockname(sock_lower_handle_t, struct sockaddr *,
167 		    socklen_t *, cred_t *);
168 int		rawip_getpeername(sock_lower_handle_t, struct sockaddr *,
169 		    socklen_t *, cred_t *);
170 
171 static struct module_info icmp_mod_info =  {
172 	5707, "icmp", 1, INFPSZ, 512, 128
173 };
174 
175 /*
176  * Entry points for ICMP as a device.
177  * We have separate open functions for the /dev/icmp and /dev/icmp6 devices.
178  */
179 static struct qinit icmprinitv4 = {
180 	NULL, NULL, icmp_openv4, icmp_close, NULL, &icmp_mod_info
181 };
182 
183 static struct qinit icmprinitv6 = {
184 	NULL, NULL, icmp_openv6, icmp_close, NULL, &icmp_mod_info
185 };
186 
187 static struct qinit icmpwinit = {
188 	(pfi_t)icmp_wput, NULL, NULL, NULL, NULL, &icmp_mod_info
189 };
190 
191 /* ICMP entry point during fallback */
192 static struct qinit icmp_fallback_sock_winit = {
193 	(pfi_t)icmp_wput_fallback, NULL, NULL, NULL, NULL, &icmp_mod_info
194 };
195 
196 /* For AF_INET aka /dev/icmp */
197 struct streamtab icmpinfov4 = {
198 	&icmprinitv4, &icmpwinit
199 };
200 
201 /* For AF_INET6 aka /dev/icmp6 */
202 struct streamtab icmpinfov6 = {
203 	&icmprinitv6, &icmpwinit
204 };
205 
206 static sin_t	sin_null;	/* Zero address for quick clears */
207 static sin6_t	sin6_null;	/* Zero address for quick clears */
208 
209 /* Default structure copied into T_INFO_ACK messages */
210 static struct T_info_ack icmp_g_t_info_ack = {
211 	T_INFO_ACK,
212 	IP_MAXPACKET,	 /* TSDU_size.  icmp allows maximum size messages. */
213 	T_INVALID,	/* ETSDU_size.  icmp does not support expedited data. */
214 	T_INVALID,	/* CDATA_size. icmp does not support connect data. */
215 	T_INVALID,	/* DDATA_size. icmp does not support disconnect data. */
216 	0,		/* ADDR_size - filled in later. */
217 	0,		/* OPT_size - not initialized here */
218 	IP_MAXPACKET,	/* TIDU_size.  icmp allows maximum size messages. */
219 	T_CLTS,		/* SERV_type.  icmp supports connection-less. */
220 	TS_UNBND,	/* CURRENT_state.  This is set from icmp_state. */
221 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
222 };
223 
224 /*
225  * Table of ND variables supported by icmp.  These are loaded into is_nd
226  * when the stack instance is created.
227  * All of these are alterable, within the min/max values given, at run time.
228  */
229 static icmpparam_t	icmp_param_arr[] = {
230 	/* min	max	value	name */
231 	{ 0,	128,	32,	"icmp_wroff_extra" },
232 	{ 1,	255,	255,	"icmp_ipv4_ttl" },
233 	{ 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS,	"icmp_ipv6_hoplimit"},
234 	{ 0,	1,	1,	"icmp_bsd_compat" },
235 	{ 4096,	65536,	8192,	"icmp_xmit_hiwat"},
236 	{ 0,	65536,	1024,	"icmp_xmit_lowat"},
237 	{ 4096,	65536,	8192,	"icmp_recv_hiwat"},
238 	{ 65536, 1024*1024*1024, 256*1024,	"icmp_max_buf"},
239 };
240 #define	is_wroff_extra			is_param_arr[0].icmp_param_value
241 #define	is_ipv4_ttl			is_param_arr[1].icmp_param_value
242 #define	is_ipv6_hoplimit		is_param_arr[2].icmp_param_value
243 #define	is_bsd_compat			is_param_arr[3].icmp_param_value
244 #define	is_xmit_hiwat			is_param_arr[4].icmp_param_value
245 #define	is_xmit_lowat			is_param_arr[5].icmp_param_value
246 #define	is_recv_hiwat			is_param_arr[6].icmp_param_value
247 #define	is_max_buf			is_param_arr[7].icmp_param_value
248 
249 static int rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len);
250 static int rawip_do_connect(conn_t *connp, const struct sockaddr *sa,
251     socklen_t len, cred_t *cr);
252 static void rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error);
253 
254 /*
255  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
256  * passed to icmp_wput.
257  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP
258  * protocol type placed in the message following the address. A T_BIND_ACK
259  * message is returned by ip_bind_v4/v6.
260  */
261 static void
262 icmp_tpi_bind(queue_t *q, mblk_t *mp)
263 {
264 	int	error;
265 	struct sockaddr *sa;
266 	struct T_bind_req *tbr;
267 	socklen_t	len;
268 	sin_t	*sin;
269 	sin6_t	*sin6;
270 	icmp_t		*icmp;
271 	conn_t	*connp = Q_TO_CONN(q);
272 	mblk_t *mp1;
273 	cred_t *cr;
274 
275 	/*
276 	 * All Solaris components should pass a db_credp
277 	 * for this TPI message, hence we ASSERT.
278 	 * But in case there is some other M_PROTO that looks
279 	 * like a TPI message sent by some other kernel
280 	 * component, we check and return an error.
281 	 */
282 	cr = msg_getcred(mp, NULL);
283 	ASSERT(cr != NULL);
284 	if (cr == NULL) {
285 		icmp_err_ack(q, mp, TSYSERR, EINVAL);
286 		return;
287 	}
288 
289 	icmp = connp->conn_icmp;
290 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
291 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
292 		    "icmp_bind: bad req, len %u",
293 		    (uint_t)(mp->b_wptr - mp->b_rptr));
294 		icmp_err_ack(q, mp, TPROTO, 0);
295 		return;
296 	}
297 
298 	if (icmp->icmp_state != TS_UNBND) {
299 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
300 		    "icmp_bind: bad state, %d", icmp->icmp_state);
301 		icmp_err_ack(q, mp, TOUTSTATE, 0);
302 		return;
303 	}
304 
305 	/*
306 	 * Reallocate the message to make sure we have enough room for an
307 	 * address and the protocol type.
308 	 */
309 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
310 	if (!mp1) {
311 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
312 		return;
313 	}
314 	mp = mp1;
315 
316 	/* Reset the message type in preparation for shipping it back. */
317 	DB_TYPE(mp) = M_PCPROTO;
318 	tbr = (struct T_bind_req *)mp->b_rptr;
319 	len = tbr->ADDR_length;
320 	switch (len) {
321 	case 0:	/* request for a generic port */
322 		tbr->ADDR_offset = sizeof (struct T_bind_req);
323 		if (icmp->icmp_family == AF_INET) {
324 			tbr->ADDR_length = sizeof (sin_t);
325 			sin = (sin_t *)&tbr[1];
326 			*sin = sin_null;
327 			sin->sin_family = AF_INET;
328 			mp->b_wptr = (uchar_t *)&sin[1];
329 			sa = (struct sockaddr *)sin;
330 			len = sizeof (sin_t);
331 		} else {
332 			ASSERT(icmp->icmp_family == AF_INET6);
333 			tbr->ADDR_length = sizeof (sin6_t);
334 			sin6 = (sin6_t *)&tbr[1];
335 			*sin6 = sin6_null;
336 			sin6->sin6_family = AF_INET6;
337 			mp->b_wptr = (uchar_t *)&sin6[1];
338 			sa = (struct sockaddr *)sin6;
339 			len = sizeof (sin6_t);
340 		}
341 		break;
342 
343 	case sizeof (sin_t):	/* Complete IPv4 address */
344 		sa = (struct sockaddr *)mi_offset_param(mp, tbr->ADDR_offset,
345 		    sizeof (sin_t));
346 		break;
347 
348 	case sizeof (sin6_t):	/* Complete IPv6 address */
349 		sa = (struct sockaddr *)mi_offset_param(mp,
350 		    tbr->ADDR_offset, sizeof (sin6_t));
351 		break;
352 
353 	default:
354 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
355 		    "icmp_bind: bad ADDR_length %d", tbr->ADDR_length);
356 		icmp_err_ack(q, mp, TBADADDR, 0);
357 		return;
358 	}
359 
360 	error = rawip_do_bind(connp, sa, len);
361 done:
362 	ASSERT(mp->b_cont == NULL);
363 	if (error != 0) {
364 		if (error > 0) {
365 			icmp_err_ack(q, mp, TSYSERR, error);
366 		} else {
367 			icmp_err_ack(q, mp, -error, 0);
368 		}
369 	} else {
370 		tbr->PRIM_type = T_BIND_ACK;
371 		qreply(q, mp);
372 	}
373 }
374 
375 static int
376 rawip_do_bind(conn_t *connp, struct sockaddr *sa, socklen_t len)
377 {
378 	sin_t		*sin;
379 	sin6_t		*sin6;
380 	icmp_t		*icmp;
381 	int		error = 0;
382 	mblk_t		*ire_mp;
383 
384 
385 	icmp = connp->conn_icmp;
386 
387 	if (sa == NULL || !OK_32PTR((char *)sa)) {
388 		return (EINVAL);
389 	}
390 
391 	/*
392 	 * The state must be TS_UNBND. TPI mandates that users must send
393 	 * TPI primitives only 1 at a time and wait for the response before
394 	 * sending the next primitive.
395 	 */
396 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
397 	if (icmp->icmp_state != TS_UNBND || icmp->icmp_pending_op != -1) {
398 		error = -TOUTSTATE;
399 		goto done;
400 	}
401 
402 	ASSERT(len != 0);
403 	switch (len) {
404 	case sizeof (sin_t):    /* Complete IPv4 address */
405 		sin = (sin_t *)sa;
406 		if (sin->sin_family != AF_INET ||
407 		    icmp->icmp_family != AF_INET) {
408 			/* TSYSERR, EAFNOSUPPORT */
409 			error = EAFNOSUPPORT;
410 			goto done;
411 		}
412 		break;
413 	case sizeof (sin6_t): /* Complete IPv6 address */
414 		sin6 = (sin6_t *)sa;
415 		if (sin6->sin6_family != AF_INET6 ||
416 		    icmp->icmp_family != AF_INET6) {
417 			/* TSYSERR, EAFNOSUPPORT */
418 			error = EAFNOSUPPORT;
419 			goto done;
420 		}
421 		/* No support for mapped addresses on raw sockets */
422 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
423 			/* TSYSERR, EADDRNOTAVAIL */
424 			error = EADDRNOTAVAIL;
425 			goto done;
426 		}
427 		break;
428 
429 	default:
430 		/* TBADADDR */
431 		error = EADDRNOTAVAIL;
432 		goto done;
433 	}
434 
435 	icmp->icmp_pending_op = T_BIND_REQ;
436 	icmp->icmp_state = TS_IDLE;
437 
438 	/*
439 	 * Copy the source address into our icmp structure.  This address
440 	 * may still be zero; if so, ip will fill in the correct address
441 	 * each time an outbound packet is passed to it.
442 	 * If we are binding to a broadcast or multicast address then
443 	 * rawip_post_ip_bind_connect will clear the source address.
444 	 */
445 
446 	if (icmp->icmp_family == AF_INET) {
447 		ASSERT(sin != NULL);
448 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
449 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr,
450 		    &icmp->icmp_v6src);
451 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
452 		    icmp->icmp_ip_snd_options_len;
453 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
454 	} else {
455 		int error;
456 
457 		ASSERT(sin6 != NULL);
458 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
459 		icmp->icmp_v6src = sin6->sin6_addr;
460 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
461 		icmp->icmp_bound_v6src = icmp->icmp_v6src;
462 
463 		/* Rebuild the header template */
464 		error = icmp_build_hdrs(icmp);
465 		if (error != 0) {
466 			icmp->icmp_pending_op = -1;
467 			/*
468 			 * TSYSERR
469 			 */
470 			goto done;
471 		}
472 	}
473 
474 	ire_mp = NULL;
475 	if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) {
476 		/*
477 		 * request an IRE if src not 0 (INADDR_ANY)
478 		 */
479 		ire_mp = allocb(sizeof (ire_t), BPRI_HI);
480 		if (ire_mp == NULL) {
481 			icmp->icmp_pending_op = -1;
482 			error = ENOMEM;
483 			goto done;
484 		}
485 		DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;
486 	}
487 done:
488 	rw_exit(&icmp->icmp_rwlock);
489 	if (error != 0)
490 		return (error);
491 
492 	if (icmp->icmp_family == AF_INET6) {
493 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
494 		    &sin6->sin6_addr, sin6->sin6_port, B_TRUE);
495 	} else {
496 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
497 		    sin->sin_addr.s_addr, sin->sin_port, B_TRUE);
498 	}
499 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
500 	return (error);
501 }
502 
503 static void
504 rawip_post_ip_bind_connect(icmp_t *icmp, mblk_t *ire_mp, int error)
505 {
506 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
507 	if (icmp->icmp_state == TS_UNBND) {
508 		/*
509 		 * not yet bound - bind sent by icmp_bind_proto.
510 		 */
511 		rw_exit(&icmp->icmp_rwlock);
512 		return;
513 	}
514 	ASSERT(icmp->icmp_pending_op != -1);
515 	icmp->icmp_pending_op = -1;
516 
517 	if (error != 0) {
518 		if (icmp->icmp_state == TS_DATA_XFER) {
519 			/* Connect failed */
520 			/* Revert back to the bound source */
521 			icmp->icmp_v6src = icmp->icmp_bound_v6src;
522 			icmp->icmp_state = TS_IDLE;
523 			if (icmp->icmp_family == AF_INET6)
524 				(void) icmp_build_hdrs(icmp);
525 		} else {
526 			V6_SET_ZERO(icmp->icmp_v6src);
527 			V6_SET_ZERO(icmp->icmp_bound_v6src);
528 			icmp->icmp_state = TS_UNBND;
529 			if (icmp->icmp_family == AF_INET6)
530 				(void) icmp_build_hdrs(icmp);
531 		}
532 	} else {
533 		if (ire_mp != NULL && ire_mp->b_datap->db_type == IRE_DB_TYPE) {
534 			ire_t *ire;
535 
536 			ire = (ire_t *)ire_mp->b_rptr;
537 			/*
538 			 * If a broadcast/multicast address was bound set
539 			 * the source address to 0.
540 			 * This ensures no datagrams with broadcast address
541 			 * as source address are emitted (which would violate
542 			 * RFC1122 - Hosts requirements)
543 			 * Note: we get IRE_BROADCAST for IPv6
544 			 * to "mark" a multicast local address.
545 			 */
546 
547 
548 			if (ire->ire_type == IRE_BROADCAST &&
549 			    icmp->icmp_state != TS_DATA_XFER) {
550 				/*
551 				 * This was just a local bind to a
552 				 * MC/broadcast addr
553 				 */
554 				V6_SET_ZERO(icmp->icmp_v6src);
555 				if (icmp->icmp_family == AF_INET6)
556 					(void) icmp_build_hdrs(icmp);
557 			}
558 		}
559 
560 	}
561 	rw_exit(&icmp->icmp_rwlock);
562 	if (ire_mp != NULL)
563 		freeb(ire_mp);
564 }
565 
566 /*
567  * Send message to IP to just bind to the protocol.
568  */
569 static int
570 icmp_bind_proto(conn_t *connp)
571 {
572 	icmp_t	*icmp;
573 	int	error;
574 
575 	icmp = connp->conn_icmp;
576 
577 	if (icmp->icmp_family == AF_INET6)
578 		error = ip_proto_bind_laddr_v6(connp, NULL, icmp->icmp_proto,
579 		    &sin6_null.sin6_addr, 0, B_TRUE);
580 	else
581 		error = ip_proto_bind_laddr_v4(connp, NULL, icmp->icmp_proto,
582 		    sin_null.sin_addr.s_addr, 0, B_TRUE);
583 
584 	rawip_post_ip_bind_connect(icmp, NULL, error);
585 	return (error);
586 }
587 
588 static void
589 icmp_tpi_connect(queue_t *q, mblk_t *mp)
590 {
591 	conn_t	*connp = Q_TO_CONN(q);
592 	struct T_conn_req	*tcr;
593 	icmp_t	*icmp;
594 	struct sockaddr *sa;
595 	socklen_t len;
596 	int error;
597 	cred_t *cr;
598 
599 	/*
600 	 * All Solaris components should pass a db_credp
601 	 * for this TPI message, hence we ASSERT.
602 	 * But in case there is some other M_PROTO that looks
603 	 * like a TPI message sent by some other kernel
604 	 * component, we check and return an error.
605 	 */
606 	cr = msg_getcred(mp, NULL);
607 	ASSERT(cr != NULL);
608 	if (cr == NULL) {
609 		icmp_err_ack(q, mp, TSYSERR, EINVAL);
610 		return;
611 	}
612 
613 	icmp = connp->conn_icmp;
614 	tcr = (struct T_conn_req *)mp->b_rptr;
615 	/* Sanity checks */
616 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
617 		icmp_err_ack(q, mp, TPROTO, 0);
618 		return;
619 	}
620 
621 	if (tcr->OPT_length != 0) {
622 		icmp_err_ack(q, mp, TBADOPT, 0);
623 		return;
624 	}
625 
626 	len = tcr->DEST_length;
627 
628 	switch (len) {
629 	default:
630 		icmp_err_ack(q, mp, TBADADDR, 0);
631 		return;
632 	case sizeof (sin_t):
633 		sa = (struct sockaddr *)mi_offset_param(mp, tcr->DEST_offset,
634 		    sizeof (sin_t));
635 		break;
636 	case sizeof (sin6_t):
637 		sa = (struct sockaddr *)mi_offset_param(mp,
638 		    tcr->DEST_offset, sizeof (sin6_t));
639 		break;
640 	}
641 
642 	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
643 	if (error != 0) {
644 		icmp_err_ack(q, mp, TSYSERR, error);
645 		return;
646 	}
647 
648 	error = rawip_do_connect(connp, sa, len, cr);
649 	if (error != 0) {
650 		if (error < 0) {
651 			icmp_err_ack(q, mp, -error, 0);
652 		} else {
653 			icmp_err_ack(q, mp, 0, error);
654 		}
655 	} else {
656 		mblk_t *mp1;
657 
658 		/*
659 		 * We have to send a connection confirmation to
660 		 * keep TLI happy.
661 		 */
662 		if (icmp->icmp_family == AF_INET) {
663 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
664 			    sizeof (sin_t), NULL, 0);
665 		} else {
666 			ASSERT(icmp->icmp_family == AF_INET6);
667 			mp1 = mi_tpi_conn_con(NULL, (char *)sa,
668 			    sizeof (sin6_t), NULL, 0);
669 		}
670 		if (mp1 == NULL) {
671 			icmp_err_ack(q, mp, TSYSERR, ENOMEM);
672 			return;
673 		}
674 
675 		/*
676 		 * Send ok_ack for T_CONN_REQ
677 		 */
678 		mp = mi_tpi_ok_ack_alloc(mp);
679 		if (mp == NULL) {
680 			/* Unable to reuse the T_CONN_REQ for the ack. */
681 			icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
682 			return;
683 		}
684 		putnext(connp->conn_rq, mp);
685 		putnext(connp->conn_rq, mp1);
686 	}
687 }
688 
689 static int
690 rawip_do_connect(conn_t *connp, const struct sockaddr *sa, socklen_t len,
691     cred_t *cr)
692 {
693 	icmp_t	*icmp;
694 	sin_t	*sin;
695 	sin6_t	*sin6;
696 	mblk_t  *ire_mp;
697 	int	error;
698 	ipaddr_t	v4dst;
699 	in6_addr_t	v6dst;
700 
701 	icmp = connp->conn_icmp;
702 
703 	if (sa == NULL || !OK_32PTR((char *)sa)) {
704 		return (EINVAL);
705 	}
706 
707 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
708 	if (ire_mp == NULL)
709 		return (ENOMEM);
710 	DB_TYPE(ire_mp) = IRE_DB_REQ_TYPE;
711 
712 
713 	ASSERT(sa != NULL && len != 0);
714 
715 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
716 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
717 		rw_exit(&icmp->icmp_rwlock);
718 		freeb(ire_mp);
719 		return (-TOUTSTATE);
720 	}
721 
722 	switch (len) {
723 	case sizeof (sin_t):
724 		sin = (sin_t *)sa;
725 
726 		ASSERT(icmp->icmp_family == AF_INET);
727 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
728 
729 		v4dst = sin->sin_addr.s_addr;
730 		/*
731 		 * Interpret a zero destination to mean loopback.
732 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
733 		 * generate the T_CONN_CON.
734 		 */
735 		if (v4dst == INADDR_ANY) {
736 			v4dst = htonl(INADDR_LOOPBACK);
737 		}
738 
739 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
740 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
741 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
742 		    icmp->icmp_ip_snd_options_len;
743 		icmp->icmp_v6dst.sin6_addr = v6dst;
744 		icmp->icmp_v6dst.sin6_family = AF_INET6;
745 		icmp->icmp_v6dst.sin6_flowinfo = 0;
746 		icmp->icmp_v6dst.sin6_port = 0;
747 
748 		/*
749 		 * If the destination address is multicast and
750 		 * an outgoing multicast interface has been set,
751 		 * use the address of that interface as our
752 		 * source address if no source address has been set.
753 		 */
754 		if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY &&
755 		    CLASSD(v4dst) &&
756 		    icmp->icmp_multicast_if_addr != INADDR_ANY) {
757 			IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr,
758 			    &icmp->icmp_v6src);
759 		}
760 		break;
761 	case sizeof (sin6_t):
762 		sin6 = (sin6_t *)sa;
763 
764 		/* No support for mapped addresses on raw sockets */
765 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
766 			rw_exit(&icmp->icmp_rwlock);
767 			freeb(ire_mp);
768 			return (EADDRNOTAVAIL);
769 		}
770 
771 		ASSERT(icmp->icmp_ipversion == IPV6_VERSION);
772 		ASSERT(icmp->icmp_family == AF_INET6);
773 
774 		icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len;
775 
776 		icmp->icmp_v6dst = *sin6;
777 		icmp->icmp_v6dst.sin6_port = 0;
778 
779 		/*
780 		 * Interpret a zero destination to mean loopback.
781 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
782 		 * generate the T_CONN_CON.
783 		 */
784 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6dst.sin6_addr)) {
785 			icmp->icmp_v6dst.sin6_addr = ipv6_loopback;
786 		}
787 		/*
788 		 * If the destination address is multicast and
789 		 * an outgoing multicast interface has been set,
790 		 * then the ip bind logic will pick the correct source
791 		 * address (i.e. matching the outgoing multicast interface).
792 		 */
793 		break;
794 	}
795 
796 	icmp->icmp_pending_op = T_CONN_REQ;
797 
798 	if (icmp->icmp_state == TS_DATA_XFER) {
799 		/* Already connected - clear out state */
800 		icmp->icmp_v6src = icmp->icmp_bound_v6src;
801 		icmp->icmp_state = TS_IDLE;
802 	}
803 
804 	icmp->icmp_state = TS_DATA_XFER;
805 	rw_exit(&icmp->icmp_rwlock);
806 
807 	if (icmp->icmp_family == AF_INET6) {
808 		error = ip_proto_bind_connected_v6(connp, &ire_mp,
809 		    icmp->icmp_proto, &icmp->icmp_v6src, 0,
810 		    &icmp->icmp_v6dst.sin6_addr,
811 		    NULL, sin6->sin6_port, B_TRUE, B_TRUE, cr);
812 	} else {
813 		error = ip_proto_bind_connected_v4(connp, &ire_mp,
814 		    icmp->icmp_proto, &V4_PART_OF_V6(icmp->icmp_v6src), 0,
815 		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr), sin->sin_port,
816 		    B_TRUE, B_TRUE, cr);
817 	}
818 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
819 	return (error);
820 }
821 
822 static void
823 icmp_close_free(conn_t *connp)
824 {
825 	icmp_t *icmp = connp->conn_icmp;
826 
827 	/* If there are any options associated with the stream, free them. */
828 	if (icmp->icmp_ip_snd_options != NULL) {
829 		mi_free((char *)icmp->icmp_ip_snd_options);
830 		icmp->icmp_ip_snd_options = NULL;
831 		icmp->icmp_ip_snd_options_len = 0;
832 	}
833 
834 	if (icmp->icmp_filter != NULL) {
835 		kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t));
836 		icmp->icmp_filter = NULL;
837 	}
838 
839 	/* Free memory associated with sticky options */
840 	if (icmp->icmp_sticky_hdrs_len != 0) {
841 		kmem_free(icmp->icmp_sticky_hdrs,
842 		    icmp->icmp_sticky_hdrs_len);
843 		icmp->icmp_sticky_hdrs = NULL;
844 		icmp->icmp_sticky_hdrs_len = 0;
845 	}
846 
847 	if (icmp->icmp_last_cred != NULL) {
848 		crfree(icmp->icmp_last_cred);
849 		icmp->icmp_last_cred = NULL;
850 	}
851 
852 	if (icmp->icmp_effective_cred != NULL) {
853 		crfree(icmp->icmp_effective_cred);
854 		icmp->icmp_effective_cred = NULL;
855 	}
856 
857 	ip6_pkt_free(&icmp->icmp_sticky_ipp);
858 
859 	/*
860 	 * Clear any fields which the kmem_cache constructor clears.
861 	 * Only icmp_connp needs to be preserved.
862 	 * TBD: We should make this more efficient to avoid clearing
863 	 * everything.
864 	 */
865 	ASSERT(icmp->icmp_connp == connp);
866 	bzero(icmp, sizeof (icmp_t));
867 	icmp->icmp_connp = connp;
868 }
869 
870 static int
871 rawip_do_close(conn_t *connp)
872 {
873 	ASSERT(connp != NULL && IPCL_IS_RAWIP(connp));
874 
875 	ip_quiesce_conn(connp);
876 
877 	if (!IPCL_IS_NONSTR(connp)) {
878 		qprocsoff(connp->conn_rq);
879 	}
880 
881 	ASSERT(connp->conn_icmp->icmp_fallback_queue_head == NULL &&
882 	    connp->conn_icmp->icmp_fallback_queue_tail == NULL);
883 	icmp_close_free(connp);
884 
885 	/*
886 	 * Now we are truly single threaded on this stream, and can
887 	 * delete the things hanging off the connp, and finally the connp.
888 	 * We removed this connp from the fanout list, it cannot be
889 	 * accessed thru the fanouts, and we already waited for the
890 	 * conn_ref to drop to 0. We are already in close, so
891 	 * there cannot be any other thread from the top. qprocsoff
892 	 * has completed, and service has completed or won't run in
893 	 * future.
894 	 */
895 	ASSERT(connp->conn_ref == 1);
896 
897 	if (!IPCL_IS_NONSTR(connp)) {
898 		inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
899 	} else {
900 		ip_free_helper_stream(connp);
901 	}
902 
903 	connp->conn_ref--;
904 	ipcl_conn_destroy(connp);
905 
906 	return (0);
907 }
908 
909 static int
910 icmp_close(queue_t *q, int flags)
911 {
912 	conn_t  *connp;
913 
914 	if (flags & SO_FALLBACK) {
915 		/*
916 		 * stream is being closed while in fallback
917 		 * simply free the resources that were allocated
918 		 */
919 		inet_minor_free(WR(q)->q_ptr, (dev_t)(RD(q)->q_ptr));
920 		qprocsoff(q);
921 		goto done;
922 	}
923 
924 	connp = Q_TO_CONN(q);
925 	(void) rawip_do_close(connp);
926 done:
927 	q->q_ptr = WR(q)->q_ptr = NULL;
928 	return (0);
929 }
930 
931 /*
932  * This routine handles each T_DISCON_REQ message passed to icmp
933  * as an indicating that ICMP is no longer connected. This results
934  * in sending a T_BIND_REQ to IP to restore the binding to just
935  * the local address.
936  *
937  * The disconnect completes in rawip_post_ip_bind_connect.
938  */
939 static int
940 icmp_do_disconnect(conn_t *connp)
941 {
942 	icmp_t	*icmp;
943 	mblk_t	*ire_mp;
944 	int error;
945 
946 	icmp = connp->conn_icmp;
947 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
948 	if (icmp->icmp_state != TS_DATA_XFER || icmp->icmp_pending_op != -1) {
949 		rw_exit(&icmp->icmp_rwlock);
950 		return (-TOUTSTATE);
951 	}
952 	icmp->icmp_pending_op = T_DISCON_REQ;
953 	icmp->icmp_v6src = icmp->icmp_bound_v6src;
954 	icmp->icmp_state = TS_IDLE;
955 
956 
957 	if (icmp->icmp_family == AF_INET6) {
958 		/* Rebuild the header template */
959 		error = icmp_build_hdrs(icmp);
960 		if (error != 0) {
961 			icmp->icmp_pending_op = -1;
962 			rw_exit(&icmp->icmp_rwlock);
963 			return (error);
964 		}
965 	}
966 
967 	rw_exit(&icmp->icmp_rwlock);
968 	ire_mp = allocb(sizeof (ire_t), BPRI_HI);
969 	if (ire_mp == NULL) {
970 		return (ENOMEM);
971 	}
972 
973 	if (icmp->icmp_family == AF_INET6) {
974 		error = ip_proto_bind_laddr_v6(connp, &ire_mp, icmp->icmp_proto,
975 		    &icmp->icmp_bound_v6src, 0, B_TRUE);
976 	} else {
977 
978 		error = ip_proto_bind_laddr_v4(connp, &ire_mp, icmp->icmp_proto,
979 		    V4_PART_OF_V6(icmp->icmp_bound_v6src), 0, B_TRUE);
980 	}
981 
982 	rawip_post_ip_bind_connect(icmp, ire_mp, error);
983 
984 	return (error);
985 }
986 
987 static void
988 icmp_tpi_disconnect(queue_t *q, mblk_t *mp)
989 {
990 	conn_t	*connp = Q_TO_CONN(q);
991 	int	error;
992 
993 	/*
994 	 * Allocate the largest primitive we need to send back
995 	 * T_error_ack is > than T_ok_ack
996 	 */
997 	mp = reallocb(mp, sizeof (struct T_error_ack), 1);
998 	if (mp == NULL) {
999 		/* Unable to reuse the T_DISCON_REQ for the ack. */
1000 		icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, ENOMEM);
1001 		return;
1002 	}
1003 
1004 	error = icmp_do_disconnect(connp);
1005 
1006 	if (error != 0) {
1007 		if (error > 0) {
1008 			icmp_err_ack(q, mp, 0, error);
1009 		} else {
1010 			icmp_err_ack(q, mp, -error, 0);
1011 		}
1012 	} else {
1013 		mp = mi_tpi_ok_ack_alloc(mp);
1014 		ASSERT(mp != NULL);
1015 		qreply(q, mp);
1016 	}
1017 
1018 }
1019 
1020 static int
1021 icmp_disconnect(conn_t *connp)
1022 {
1023 	int	error;
1024 	icmp_t	*icmp = connp->conn_icmp;
1025 
1026 	icmp->icmp_dgram_errind = B_FALSE;
1027 
1028 	error = icmp_do_disconnect(connp);
1029 
1030 	if (error < 0)
1031 		error = proto_tlitosyserr(-error);
1032 	return (error);
1033 }
1034 
1035 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
1036 static void
1037 icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
1038 {
1039 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
1040 		qreply(q, mp);
1041 }
1042 
1043 /* Shorthand to generate and send TPI error acks to our client */
1044 static void
1045 icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive,
1046     t_scalar_t t_error, int sys_error)
1047 {
1048 	struct T_error_ack	*teackp;
1049 
1050 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
1051 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
1052 		teackp = (struct T_error_ack *)mp->b_rptr;
1053 		teackp->ERROR_prim = primitive;
1054 		teackp->TLI_error = t_error;
1055 		teackp->UNIX_error = sys_error;
1056 		qreply(q, mp);
1057 	}
1058 }
1059 
1060 /*
1061  * icmp_icmp_error is called by icmp_input to process ICMP
1062  * messages passed up by IP.
1063  * Generates the appropriate permanent (non-transient) errors.
1064  * Assumes that IP has pulled up everything up to and including
1065  * the ICMP header.
1066  */
1067 static void
1068 icmp_icmp_error(conn_t *connp, mblk_t *mp)
1069 {
1070 	icmph_t *icmph;
1071 	ipha_t	*ipha;
1072 	int	iph_hdr_length;
1073 	sin_t	sin;
1074 	mblk_t	*mp1;
1075 	int	error = 0;
1076 	icmp_t	*icmp = connp->conn_icmp;
1077 
1078 	ipha = (ipha_t *)mp->b_rptr;
1079 
1080 	ASSERT(OK_32PTR(mp->b_rptr));
1081 
1082 	if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
1083 		ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
1084 		icmp_icmp_error_ipv6(connp, mp);
1085 		return;
1086 	}
1087 
1088 	/*
1089 	 * icmp does not support v4 mapped addresses
1090 	 * so we can never be here for a V6 socket
1091 	 * i.e. icmp_family == AF_INET6
1092 	 */
1093 	ASSERT((IPH_HDR_VERSION(ipha) == IPV4_VERSION) &&
1094 	    (icmp->icmp_family == AF_INET));
1095 
1096 	ASSERT(icmp->icmp_family == AF_INET);
1097 
1098 	/* Skip past the outer IP and ICMP headers */
1099 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1100 	icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]);
1101 	ipha = (ipha_t *)&icmph[1];
1102 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
1103 
1104 	switch (icmph->icmph_type) {
1105 	case ICMP_DEST_UNREACHABLE:
1106 		switch (icmph->icmph_code) {
1107 		case ICMP_FRAGMENTATION_NEEDED:
1108 			/*
1109 			 * IP has already adjusted the path MTU.
1110 			 */
1111 			break;
1112 		case ICMP_PORT_UNREACHABLE:
1113 		case ICMP_PROTOCOL_UNREACHABLE:
1114 			error = ECONNREFUSED;
1115 			break;
1116 		default:
1117 			/* Transient errors */
1118 			break;
1119 		}
1120 		break;
1121 	default:
1122 		/* Transient errors */
1123 		break;
1124 	}
1125 	if (error == 0) {
1126 		freemsg(mp);
1127 		return;
1128 	}
1129 
1130 	/*
1131 	 * Deliver T_UDERROR_IND when the application has asked for it.
1132 	 * The socket layer enables this automatically when connected.
1133 	 */
1134 	if (!icmp->icmp_dgram_errind) {
1135 		freemsg(mp);
1136 		return;
1137 	}
1138 
1139 	sin = sin_null;
1140 	sin.sin_family = AF_INET;
1141 	sin.sin_addr.s_addr = ipha->ipha_dst;
1142 
1143 	if (IPCL_IS_NONSTR(connp)) {
1144 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1145 		if (icmp->icmp_state == TS_DATA_XFER) {
1146 			if (sin.sin_addr.s_addr ==
1147 			    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr)) {
1148 				rw_exit(&icmp->icmp_rwlock);
1149 				(*connp->conn_upcalls->su_set_error)
1150 				    (connp->conn_upper_handle, error);
1151 				goto done;
1152 			}
1153 		} else {
1154 			icmp->icmp_delayed_error = error;
1155 			*((sin_t *)&icmp->icmp_delayed_addr) = sin;
1156 		}
1157 		rw_exit(&icmp->icmp_rwlock);
1158 	} else {
1159 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL,
1160 		    0, error);
1161 		if (mp1 != NULL)
1162 			putnext(connp->conn_rq, mp1);
1163 	}
1164 done:
1165 	ASSERT(!RW_ISWRITER(&icmp->icmp_rwlock));
1166 	freemsg(mp);
1167 }
1168 
1169 /*
1170  * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6
1171  * for IPv6 packets.
1172  * Send permanent (non-transient) errors upstream.
1173  * Assumes that IP has pulled up all the extension headers as well
1174  * as the ICMPv6 header.
1175  */
1176 static void
1177 icmp_icmp_error_ipv6(conn_t *connp, mblk_t *mp)
1178 {
1179 	icmp6_t		*icmp6;
1180 	ip6_t		*ip6h, *outer_ip6h;
1181 	uint16_t	iph_hdr_length;
1182 	uint8_t		*nexthdrp;
1183 	sin6_t		sin6;
1184 	mblk_t		*mp1;
1185 	int		error = 0;
1186 	icmp_t		*icmp = connp->conn_icmp;
1187 
1188 	outer_ip6h = (ip6_t *)mp->b_rptr;
1189 	if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6)
1190 		iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h);
1191 	else
1192 		iph_hdr_length = IPV6_HDR_LEN;
1193 
1194 	icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length];
1195 	ip6h = (ip6_t *)&icmp6[1];
1196 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) {
1197 		freemsg(mp);
1198 		return;
1199 	}
1200 
1201 	switch (icmp6->icmp6_type) {
1202 	case ICMP6_DST_UNREACH:
1203 		switch (icmp6->icmp6_code) {
1204 		case ICMP6_DST_UNREACH_NOPORT:
1205 			error = ECONNREFUSED;
1206 			break;
1207 		case ICMP6_DST_UNREACH_ADMIN:
1208 		case ICMP6_DST_UNREACH_NOROUTE:
1209 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
1210 		case ICMP6_DST_UNREACH_ADDR:
1211 			/* Transient errors */
1212 			break;
1213 		default:
1214 			break;
1215 		}
1216 		break;
1217 	case ICMP6_PACKET_TOO_BIG: {
1218 		struct T_unitdata_ind	*tudi;
1219 		struct T_opthdr		*toh;
1220 		size_t			udi_size;
1221 		mblk_t			*newmp;
1222 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
1223 		    sizeof (struct ip6_mtuinfo);
1224 		sin6_t			*sin6;
1225 		struct ip6_mtuinfo	*mtuinfo;
1226 
1227 		/*
1228 		 * If the application has requested to receive path mtu
1229 		 * information, send up an empty message containing an
1230 		 * IPV6_PATHMTU ancillary data item.
1231 		 */
1232 		if (!icmp->icmp_ipv6_recvpathmtu)
1233 			break;
1234 
1235 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
1236 		    opt_length;
1237 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
1238 			BUMP_MIB(&icmp->icmp_is->is_rawip_mib, rawipInErrors);
1239 			break;
1240 		}
1241 
1242 		/*
1243 		 * newmp->b_cont is left to NULL on purpose.  This is an
1244 		 * empty message containing only ancillary data.
1245 		 */
1246 		newmp->b_datap->db_type = M_PROTO;
1247 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
1248 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
1249 		tudi->PRIM_type = T_UNITDATA_IND;
1250 		tudi->SRC_length = sizeof (sin6_t);
1251 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
1252 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
1253 		tudi->OPT_length = opt_length;
1254 
1255 		sin6 = (sin6_t *)&tudi[1];
1256 		bzero(sin6, sizeof (sin6_t));
1257 		sin6->sin6_family = AF_INET6;
1258 		sin6->sin6_addr = icmp->icmp_v6dst.sin6_addr;
1259 
1260 		toh = (struct T_opthdr *)&sin6[1];
1261 		toh->level = IPPROTO_IPV6;
1262 		toh->name = IPV6_PATHMTU;
1263 		toh->len = opt_length;
1264 		toh->status = 0;
1265 
1266 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
1267 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
1268 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
1269 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
1270 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
1271 		/*
1272 		 * We've consumed everything we need from the original
1273 		 * message.  Free it, then send our empty message.
1274 		 */
1275 		freemsg(mp);
1276 		icmp_ulp_recv(connp, newmp);
1277 
1278 		return;
1279 	}
1280 	case ICMP6_TIME_EXCEEDED:
1281 		/* Transient errors */
1282 		break;
1283 	case ICMP6_PARAM_PROB:
1284 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
1285 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
1286 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
1287 		    (uchar_t *)nexthdrp) {
1288 			error = ECONNREFUSED;
1289 			break;
1290 		}
1291 		break;
1292 	}
1293 	if (error == 0) {
1294 		freemsg(mp);
1295 		return;
1296 	}
1297 
1298 	/*
1299 	 * Deliver T_UDERROR_IND when the application has asked for it.
1300 	 * The socket layer enables this automatically when connected.
1301 	 */
1302 	if (!icmp->icmp_dgram_errind) {
1303 		freemsg(mp);
1304 		return;
1305 	}
1306 
1307 	sin6 = sin6_null;
1308 	sin6.sin6_family = AF_INET6;
1309 	sin6.sin6_addr = ip6h->ip6_dst;
1310 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
1311 
1312 	if (IPCL_IS_NONSTR(connp)) {
1313 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1314 		if (icmp->icmp_state == TS_DATA_XFER) {
1315 			if (IN6_ARE_ADDR_EQUAL(&sin6.sin6_addr,
1316 			    &icmp->icmp_v6dst.sin6_addr)) {
1317 				rw_exit(&icmp->icmp_rwlock);
1318 				(*connp->conn_upcalls->su_set_error)
1319 				    (connp->conn_upper_handle, error);
1320 				goto done;
1321 			}
1322 		} else {
1323 			icmp->icmp_delayed_error = error;
1324 			*((sin6_t *)&icmp->icmp_delayed_addr) = sin6;
1325 		}
1326 		rw_exit(&icmp->icmp_rwlock);
1327 	} else {
1328 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
1329 		    NULL, 0, error);
1330 		if (mp1 != NULL)
1331 			putnext(connp->conn_rq, mp1);
1332 	}
1333 done:
1334 	ASSERT(!RW_ISWRITER(&icmp->icmp_rwlock));
1335 	freemsg(mp);
1336 }
1337 
1338 /*
1339  * This routine responds to T_ADDR_REQ messages.  It is called by icmp_wput.
1340  * The local address is filled in if endpoint is bound. The remote address
1341  * is filled in if remote address has been precified ("connected endpoint")
1342  * (The concept of connected CLTS sockets is alien to published TPI
1343  *  but we support it anyway).
1344  */
1345 static void
1346 icmp_addr_req(queue_t *q, mblk_t *mp)
1347 {
1348 	icmp_t	*icmp = Q_TO_ICMP(q);
1349 	mblk_t	*ackmp;
1350 	struct T_addr_ack *taa;
1351 
1352 	/* Make it large enough for worst case */
1353 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
1354 	    2 * sizeof (sin6_t), 1);
1355 	if (ackmp == NULL) {
1356 		icmp_err_ack(q, mp, TSYSERR, ENOMEM);
1357 		return;
1358 	}
1359 	taa = (struct T_addr_ack *)ackmp->b_rptr;
1360 
1361 	bzero(taa, sizeof (struct T_addr_ack));
1362 	ackmp->b_wptr = (uchar_t *)&taa[1];
1363 
1364 	taa->PRIM_type = T_ADDR_ACK;
1365 	ackmp->b_datap->db_type = M_PCPROTO;
1366 	rw_enter(&icmp->icmp_rwlock, RW_READER);
1367 	/*
1368 	 * Note: Following code assumes 32 bit alignment of basic
1369 	 * data structures like sin_t and struct T_addr_ack.
1370 	 */
1371 	if (icmp->icmp_state != TS_UNBND) {
1372 		/*
1373 		 * Fill in local address
1374 		 */
1375 		taa->LOCADDR_offset = sizeof (*taa);
1376 		if (icmp->icmp_family == AF_INET) {
1377 			sin_t	*sin;
1378 
1379 			taa->LOCADDR_length = sizeof (sin_t);
1380 			sin = (sin_t *)&taa[1];
1381 			/* Fill zeroes and then intialize non-zero fields */
1382 			*sin = sin_null;
1383 			sin->sin_family = AF_INET;
1384 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
1385 			    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1386 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src,
1387 				    sin->sin_addr.s_addr);
1388 			} else {
1389 				/*
1390 				 * INADDR_ANY
1391 				 * icmp_v6src is not set, we might be bound to
1392 				 * broadcast/multicast. Use icmp_bound_v6src as
1393 				 * local address instead (that could
1394 				 * also still be INADDR_ANY)
1395 				 */
1396 				IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src,
1397 				    sin->sin_addr.s_addr);
1398 			}
1399 			ackmp->b_wptr = (uchar_t *)&sin[1];
1400 		} else {
1401 			sin6_t	*sin6;
1402 
1403 			ASSERT(icmp->icmp_family == AF_INET6);
1404 			taa->LOCADDR_length = sizeof (sin6_t);
1405 			sin6 = (sin6_t *)&taa[1];
1406 			/* Fill zeroes and then intialize non-zero fields */
1407 			*sin6 = sin6_null;
1408 			sin6->sin6_family = AF_INET6;
1409 			if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
1410 				sin6->sin6_addr = icmp->icmp_v6src;
1411 			} else {
1412 				/*
1413 				 * UNSPECIFIED
1414 				 * icmp_v6src is not set, we might be bound to
1415 				 * broadcast/multicast. Use icmp_bound_v6src as
1416 				 * local address instead (that could
1417 				 * also still be UNSPECIFIED)
1418 				 */
1419 				sin6->sin6_addr = icmp->icmp_bound_v6src;
1420 			}
1421 			ackmp->b_wptr = (uchar_t *)&sin6[1];
1422 		}
1423 	}
1424 	rw_exit(&icmp->icmp_rwlock);
1425 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
1426 	qreply(q, ackmp);
1427 }
1428 
1429 static void
1430 icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp)
1431 {
1432 	*tap = icmp_g_t_info_ack;
1433 
1434 	if (icmp->icmp_family == AF_INET6)
1435 		tap->ADDR_size = sizeof (sin6_t);
1436 	else
1437 		tap->ADDR_size = sizeof (sin_t);
1438 	tap->CURRENT_state = icmp->icmp_state;
1439 	tap->OPT_size = icmp_max_optsize;
1440 }
1441 
1442 static void
1443 icmp_do_capability_ack(icmp_t *icmp, struct T_capability_ack *tcap,
1444     t_uscalar_t cap_bits1)
1445 {
1446 	tcap->CAP_bits1 = 0;
1447 
1448 	if (cap_bits1 & TC1_INFO) {
1449 		icmp_copy_info(&tcap->INFO_ack, icmp);
1450 		tcap->CAP_bits1 |= TC1_INFO;
1451 	}
1452 }
1453 
1454 /*
1455  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
1456  * icmp_wput.  Much of the T_CAPABILITY_ACK information is copied from
1457  * icmp_g_t_info_ack.  The current state of the stream is copied from
1458  * icmp_state.
1459  */
1460 static void
1461 icmp_capability_req(queue_t *q, mblk_t *mp)
1462 {
1463 	icmp_t			*icmp = Q_TO_ICMP(q);
1464 	t_uscalar_t		cap_bits1;
1465 	struct T_capability_ack	*tcap;
1466 
1467 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
1468 
1469 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
1470 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
1471 	if (!mp)
1472 		return;
1473 
1474 	tcap = (struct T_capability_ack *)mp->b_rptr;
1475 
1476 	icmp_do_capability_ack(icmp, tcap, cap_bits1);
1477 
1478 	qreply(q, mp);
1479 }
1480 
1481 /*
1482  * This routine responds to T_INFO_REQ messages.  It is called by icmp_wput.
1483  * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack.
1484  * The current state of the stream is copied from icmp_state.
1485  */
1486 static void
1487 icmp_info_req(queue_t *q, mblk_t *mp)
1488 {
1489 	icmp_t	*icmp = Q_TO_ICMP(q);
1490 
1491 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
1492 	    T_INFO_ACK);
1493 	if (!mp)
1494 		return;
1495 	icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp);
1496 	qreply(q, mp);
1497 }
1498 
1499 /* For /dev/icmp aka AF_INET open */
1500 static int
1501 icmp_tpi_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
1502     int family)
1503 {
1504 	conn_t *connp;
1505 	dev_t	conn_dev;
1506 	icmp_stack_t *is;
1507 	int	error;
1508 
1509 	conn_dev = NULL;
1510 
1511 	/* If the stream is already open, return immediately. */
1512 	if (q->q_ptr != NULL)
1513 		return (0);
1514 
1515 	if (sflag == MODOPEN)
1516 		return (EINVAL);
1517 
1518 	/*
1519 	 * Since ICMP is not used so heavily, allocating from the small
1520 	 * arena should be sufficient.
1521 	 */
1522 	if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0) {
1523 		return (EBUSY);
1524 	}
1525 
1526 	if (flag & SO_FALLBACK) {
1527 		/*
1528 		 * Non streams socket needs a stream to fallback to
1529 		 */
1530 		RD(q)->q_ptr = (void *)conn_dev;
1531 		WR(q)->q_qinfo = &icmp_fallback_sock_winit;
1532 		WR(q)->q_ptr = (void *)ip_minor_arena_sa;
1533 		qprocson(q);
1534 		return (0);
1535 	}
1536 
1537 	connp = icmp_open(family, credp, &error, KM_SLEEP);
1538 	if (connp == NULL) {
1539 		ASSERT(error != NULL);
1540 		inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
1541 		return (error);
1542 	}
1543 
1544 	*devp = makedevice(getemajor(*devp), (minor_t)conn_dev);
1545 	connp->conn_dev = conn_dev;
1546 	connp->conn_minor_arena = ip_minor_arena_sa;
1547 
1548 	is = connp->conn_icmp->icmp_is;
1549 
1550 	/*
1551 	 * Initialize the icmp_t structure for this stream.
1552 	 */
1553 	q->q_ptr = connp;
1554 	WR(q)->q_ptr = connp;
1555 	connp->conn_rq = q;
1556 	connp->conn_wq = WR(q);
1557 
1558 	if (connp->conn_icmp->icmp_family == AF_INET6) {
1559 		/* Build initial header template for transmit */
1560 		rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
1561 		if ((error = icmp_build_hdrs(connp->conn_icmp)) != 0) {
1562 			rw_exit(&connp->conn_icmp->icmp_rwlock);
1563 			inet_minor_free(ip_minor_arena_sa, connp->conn_dev);
1564 			ipcl_conn_destroy(connp);
1565 			return (error);
1566 		}
1567 		rw_exit(&connp->conn_icmp->icmp_rwlock);
1568 	}
1569 
1570 
1571 	q->q_hiwat = is->is_recv_hiwat;
1572 	WR(q)->q_hiwat = is->is_xmit_hiwat;
1573 	WR(q)->q_lowat = is->is_xmit_lowat;
1574 
1575 	qprocson(q);
1576 
1577 	/* Set the Stream head write offset. */
1578 	(void) proto_set_tx_wroff(q, connp,
1579 	    connp->conn_icmp->icmp_max_hdr_len + is->is_wroff_extra);
1580 	(void) proto_set_rx_hiwat(connp->conn_rq, connp, q->q_hiwat);
1581 
1582 	mutex_enter(&connp->conn_lock);
1583 	connp->conn_state_flags &= ~CONN_INCIPIENT;
1584 	mutex_exit(&connp->conn_lock);
1585 
1586 	return (0);
1587 }
1588 
1589 /* For /dev/icmp4 aka AF_INET open */
1590 static int
1591 icmp_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1592 {
1593 	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET));
1594 }
1595 
1596 /* For /dev/icmp6 aka AF_INET6 open */
1597 static int
1598 icmp_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1599 {
1600 	return (icmp_tpi_open(q, devp, flag, sflag, credp, AF_INET6));
1601 }
1602 
1603 /*
1604  * This is the open routine for icmp.  It allocates a icmp_t structure for
1605  * the stream and, on the first open of the module, creates an ND table.
1606  */
1607 /* ARGSUSED */
1608 static conn_t *
1609 icmp_open(int family, cred_t *credp, int *err, int flags)
1610 {
1611 	icmp_t	*icmp;
1612 	conn_t *connp;
1613 	zoneid_t zoneid;
1614 	netstack_t *ns;
1615 	icmp_stack_t *is;
1616 	boolean_t isv6 = B_FALSE;
1617 
1618 	*err = secpolicy_net_icmpaccess(credp);
1619 	if (*err != 0)
1620 		return (NULL);
1621 
1622 	if (family == AF_INET6)
1623 		isv6 = B_TRUE;
1624 	ns = netstack_find_by_cred(credp);
1625 	ASSERT(ns != NULL);
1626 	is = ns->netstack_icmp;
1627 	ASSERT(is != NULL);
1628 
1629 	/*
1630 	 * For exclusive stacks we set the zoneid to zero
1631 	 * to make ICMP operate as if in the global zone.
1632 	 */
1633 	if (ns->netstack_stackid != GLOBAL_NETSTACKID)
1634 		zoneid = GLOBAL_ZONEID;
1635 	else
1636 		zoneid = crgetzoneid(credp);
1637 
1638 	ASSERT(flags == KM_SLEEP || flags == KM_NOSLEEP);
1639 
1640 	connp = ipcl_conn_create(IPCL_RAWIPCONN, flags, ns);
1641 	icmp = connp->conn_icmp;
1642 	icmp->icmp_v6dst = sin6_null;
1643 
1644 	/*
1645 	 * ipcl_conn_create did a netstack_hold. Undo the hold that was
1646 	 * done by netstack_find_by_cred()
1647 	 */
1648 	netstack_rele(ns);
1649 
1650 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
1651 	ASSERT(connp->conn_ulp == IPPROTO_ICMP);
1652 	ASSERT(connp->conn_icmp == icmp);
1653 	ASSERT(icmp->icmp_connp == connp);
1654 
1655 	/* Set the initial state of the stream and the privilege status. */
1656 	icmp->icmp_state = TS_UNBND;
1657 	if (isv6) {
1658 		icmp->icmp_ipversion = IPV6_VERSION;
1659 		icmp->icmp_family = AF_INET6;
1660 		connp->conn_ulp = IPPROTO_ICMPV6;
1661 		/* May be changed by a SO_PROTOTYPE socket option. */
1662 		icmp->icmp_proto = IPPROTO_ICMPV6;
1663 		icmp->icmp_checksum_off = 2;	/* Offset for icmp6_cksum */
1664 		icmp->icmp_max_hdr_len = IPV6_HDR_LEN;
1665 		icmp->icmp_ttl = (uint8_t)is->is_ipv6_hoplimit;
1666 		connp->conn_af_isv6 = B_TRUE;
1667 	} else {
1668 		icmp->icmp_ipversion = IPV4_VERSION;
1669 		icmp->icmp_family = AF_INET;
1670 		/* May be changed by a SO_PROTOTYPE socket option. */
1671 		icmp->icmp_proto = IPPROTO_ICMP;
1672 		icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH;
1673 		icmp->icmp_ttl = (uint8_t)is->is_ipv4_ttl;
1674 		connp->conn_af_isv6 = B_FALSE;
1675 	}
1676 	icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1677 	icmp->icmp_pending_op = -1;
1678 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1679 	connp->conn_zoneid = zoneid;
1680 
1681 	/*
1682 	 * If the caller has the process-wide flag set, then default to MAC
1683 	 * exempt mode.  This allows read-down to unlabeled hosts.
1684 	 */
1685 	if (getpflags(NET_MAC_AWARE, credp) != 0)
1686 		connp->conn_mac_exempt = B_TRUE;
1687 
1688 	connp->conn_ulp_labeled = is_system_labeled();
1689 
1690 	icmp->icmp_is = is;
1691 
1692 	connp->conn_recv = icmp_input;
1693 	crhold(credp);
1694 	connp->conn_cred = credp;
1695 
1696 	rw_exit(&icmp->icmp_rwlock);
1697 
1698 	connp->conn_flow_cntrld = B_FALSE;
1699 	return (connp);
1700 }
1701 
1702 /*
1703  * Which ICMP options OK to set through T_UNITDATA_REQ...
1704  */
1705 /* ARGSUSED */
1706 static boolean_t
1707 icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
1708 {
1709 	return (B_TRUE);
1710 }
1711 
1712 /*
1713  * This routine gets default values of certain options whose default
1714  * values are maintained by protcol specific code
1715  */
1716 /* ARGSUSED */
1717 int
1718 icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr)
1719 {
1720 	icmp_t *icmp = Q_TO_ICMP(q);
1721 	icmp_stack_t *is = icmp->icmp_is;
1722 	int *i1 = (int *)ptr;
1723 
1724 	switch (level) {
1725 	case IPPROTO_IP:
1726 		switch (name) {
1727 		case IP_MULTICAST_TTL:
1728 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
1729 			return (sizeof (uchar_t));
1730 		case IP_MULTICAST_LOOP:
1731 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
1732 			return (sizeof (uchar_t));
1733 		}
1734 		break;
1735 	case IPPROTO_IPV6:
1736 		switch (name) {
1737 		case IPV6_MULTICAST_HOPS:
1738 			*i1 = IP_DEFAULT_MULTICAST_TTL;
1739 			return (sizeof (int));
1740 		case IPV6_MULTICAST_LOOP:
1741 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
1742 			return (sizeof (int));
1743 		case IPV6_UNICAST_HOPS:
1744 			*i1 = is->is_ipv6_hoplimit;
1745 			return (sizeof (int));
1746 		}
1747 		break;
1748 	case IPPROTO_ICMPV6:
1749 		switch (name) {
1750 		case ICMP6_FILTER:
1751 			/* Make it look like "pass all" */
1752 			ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
1753 			return (sizeof (icmp6_filter_t));
1754 		}
1755 		break;
1756 	}
1757 	return (-1);
1758 }
1759 
1760 /*
1761  * This routine retrieves the current status of socket options.
1762  * It returns the size of the option retrieved.
1763  */
1764 int
1765 icmp_opt_get(conn_t *connp, int level, int name, uchar_t *ptr)
1766 {
1767 	icmp_t		*icmp = connp->conn_icmp;
1768 	icmp_stack_t	*is = icmp->icmp_is;
1769 	int		*i1 = (int *)ptr;
1770 	ip6_pkt_t	*ipp = &icmp->icmp_sticky_ipp;
1771 	int		ret = 0;
1772 
1773 	ASSERT(RW_READ_HELD(&icmp->icmp_rwlock));
1774 	switch (level) {
1775 	case SOL_SOCKET:
1776 		switch (name) {
1777 		case SO_DEBUG:
1778 			*i1 = icmp->icmp_debug;
1779 			break;
1780 		case SO_TYPE:
1781 			*i1 = SOCK_RAW;
1782 			break;
1783 		case SO_PROTOTYPE:
1784 			*i1 = icmp->icmp_proto;
1785 			break;
1786 		case SO_REUSEADDR:
1787 			*i1 = icmp->icmp_reuseaddr;
1788 			break;
1789 
1790 		/*
1791 		 * The following three items are available here,
1792 		 * but are only meaningful to IP.
1793 		 */
1794 		case SO_DONTROUTE:
1795 			*i1 = icmp->icmp_dontroute;
1796 			break;
1797 		case SO_USELOOPBACK:
1798 			*i1 = icmp->icmp_useloopback;
1799 			break;
1800 		case SO_BROADCAST:
1801 			*i1 = icmp->icmp_broadcast;
1802 			break;
1803 
1804 		case SO_SNDBUF:
1805 			ASSERT(icmp->icmp_xmit_hiwat <= INT_MAX);
1806 			*i1 = icmp->icmp_xmit_hiwat;
1807 			break;
1808 		case SO_RCVBUF:
1809 			ASSERT(icmp->icmp_recv_hiwat <= INT_MAX);
1810 			*i1 = icmp->icmp_recv_hiwat;
1811 			break;
1812 		case SO_DGRAM_ERRIND:
1813 			*i1 = icmp->icmp_dgram_errind;
1814 			break;
1815 		case SO_TIMESTAMP:
1816 			*i1 = icmp->icmp_timestamp;
1817 			break;
1818 		case SO_MAC_EXEMPT:
1819 			*i1 = connp->conn_mac_exempt;
1820 			break;
1821 		case SO_DOMAIN:
1822 			*i1 = icmp->icmp_family;
1823 			break;
1824 
1825 		/*
1826 		 * Following four not meaningful for icmp
1827 		 * Action is same as "default" to which we fallthrough
1828 		 * so we keep them in comments.
1829 		 * case SO_LINGER:
1830 		 * case SO_KEEPALIVE:
1831 		 * case SO_OOBINLINE:
1832 		 * case SO_ALLZONES:
1833 		 */
1834 		default:
1835 			ret = -1;
1836 			goto done;
1837 		}
1838 		break;
1839 	case IPPROTO_IP:
1840 		/*
1841 		 * Only allow IPv4 option processing on IPv4 sockets.
1842 		 */
1843 		if (icmp->icmp_family != AF_INET) {
1844 			ret = -1;
1845 			goto done;
1846 		}
1847 
1848 		switch (name) {
1849 		case IP_OPTIONS:
1850 		case T_IP_OPTIONS:
1851 			/* Options are passed up with each packet */
1852 			ret = 0;
1853 			goto done;
1854 		case IP_HDRINCL:
1855 			*i1 = (int)icmp->icmp_hdrincl;
1856 			break;
1857 		case IP_TOS:
1858 		case T_IP_TOS:
1859 			*i1 = (int)icmp->icmp_type_of_service;
1860 			break;
1861 		case IP_TTL:
1862 			*i1 = (int)icmp->icmp_ttl;
1863 			break;
1864 		case IP_MULTICAST_IF:
1865 			/* 0 address if not set */
1866 			*(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr;
1867 			ret = sizeof (ipaddr_t);
1868 			goto done;
1869 		case IP_MULTICAST_TTL:
1870 			*(uchar_t *)ptr = icmp->icmp_multicast_ttl;
1871 			ret = sizeof (uchar_t);
1872 			goto done;
1873 		case IP_MULTICAST_LOOP:
1874 			*ptr = connp->conn_multicast_loop;
1875 			ret = sizeof (uint8_t);
1876 			goto done;
1877 		case IP_BOUND_IF:
1878 			/* Zero if not set */
1879 			*i1 = icmp->icmp_bound_if;
1880 			break;	/* goto sizeof (int) option return */
1881 		case IP_UNSPEC_SRC:
1882 			*ptr = icmp->icmp_unspec_source;
1883 			break;	/* goto sizeof (int) option return */
1884 		case IP_RECVIF:
1885 			*ptr = icmp->icmp_recvif;
1886 			break;	/* goto sizeof (int) option return */
1887 		case IP_BROADCAST_TTL:
1888 			*(uchar_t *)ptr = connp->conn_broadcast_ttl;
1889 			return (sizeof (uchar_t));
1890 		case IP_RECVPKTINFO:
1891 			/*
1892 			 * This also handles IP_PKTINFO.
1893 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
1894 			 * Differentiation is based on the size of the argument
1895 			 * passed in.
1896 			 * This option is handled in IP which will return an
1897 			 * error for IP_PKTINFO as it's not supported as a
1898 			 * sticky option.
1899 			 */
1900 			ret = -EINVAL;
1901 			goto done;
1902 		/*
1903 		 * Cannot "get" the value of following options
1904 		 * at this level. Action is same as "default" to
1905 		 * which we fallthrough so we keep them in comments.
1906 		 *
1907 		 * case IP_ADD_MEMBERSHIP:
1908 		 * case IP_DROP_MEMBERSHIP:
1909 		 * case IP_BLOCK_SOURCE:
1910 		 * case IP_UNBLOCK_SOURCE:
1911 		 * case IP_ADD_SOURCE_MEMBERSHIP:
1912 		 * case IP_DROP_SOURCE_MEMBERSHIP:
1913 		 * case MCAST_JOIN_GROUP:
1914 		 * case MCAST_LEAVE_GROUP:
1915 		 * case MCAST_BLOCK_SOURCE:
1916 		 * case MCAST_UNBLOCK_SOURCE:
1917 		 * case MCAST_JOIN_SOURCE_GROUP:
1918 		 * case MCAST_LEAVE_SOURCE_GROUP:
1919 		 * case MRT_INIT:
1920 		 * case MRT_DONE:
1921 		 * case MRT_ADD_VIF:
1922 		 * case MRT_DEL_VIF:
1923 		 * case MRT_ADD_MFC:
1924 		 * case MRT_DEL_MFC:
1925 		 * case MRT_VERSION:
1926 		 * case MRT_ASSERT:
1927 		 * case IP_SEC_OPT:
1928 		 * case IP_NEXTHOP:
1929 		 */
1930 		default:
1931 			ret = -1;
1932 			goto done;
1933 		}
1934 		break;
1935 	case IPPROTO_IPV6:
1936 		/*
1937 		 * Only allow IPv6 option processing on native IPv6 sockets.
1938 		 */
1939 		if (icmp->icmp_family != AF_INET6) {
1940 			ret = -1;
1941 			goto done;
1942 		}
1943 		switch (name) {
1944 		case IPV6_UNICAST_HOPS:
1945 			*i1 = (unsigned int)icmp->icmp_ttl;
1946 			break;
1947 		case IPV6_MULTICAST_IF:
1948 			/* 0 index if not set */
1949 			*i1 = icmp->icmp_multicast_if_index;
1950 			break;
1951 		case IPV6_MULTICAST_HOPS:
1952 			*i1 = icmp->icmp_multicast_ttl;
1953 			break;
1954 		case IPV6_MULTICAST_LOOP:
1955 			*i1 = connp->conn_multicast_loop;
1956 			break;
1957 		case IPV6_BOUND_IF:
1958 			/* Zero if not set */
1959 			*i1 = icmp->icmp_bound_if;
1960 			break;
1961 		case IPV6_UNSPEC_SRC:
1962 			*i1 = icmp->icmp_unspec_source;
1963 			break;
1964 		case IPV6_CHECKSUM:
1965 			/*
1966 			 * Return offset or -1 if no checksum offset.
1967 			 * Does not apply to IPPROTO_ICMPV6
1968 			 */
1969 			if (icmp->icmp_proto == IPPROTO_ICMPV6) {
1970 				ret = -1;
1971 				goto done;
1972 			}
1973 
1974 			if (icmp->icmp_raw_checksum) {
1975 				*i1 = icmp->icmp_checksum_off;
1976 			} else {
1977 				*i1 = -1;
1978 			}
1979 			break;
1980 		case IPV6_JOIN_GROUP:
1981 		case IPV6_LEAVE_GROUP:
1982 		case MCAST_JOIN_GROUP:
1983 		case MCAST_LEAVE_GROUP:
1984 		case MCAST_BLOCK_SOURCE:
1985 		case MCAST_UNBLOCK_SOURCE:
1986 		case MCAST_JOIN_SOURCE_GROUP:
1987 		case MCAST_LEAVE_SOURCE_GROUP:
1988 			/* cannot "get" the value for these */
1989 			ret = -1;
1990 			goto done;
1991 		case IPV6_RECVPKTINFO:
1992 			*i1 = icmp->icmp_ip_recvpktinfo;
1993 			break;
1994 		case IPV6_RECVTCLASS:
1995 			*i1 = icmp->icmp_ipv6_recvtclass;
1996 			break;
1997 		case IPV6_RECVPATHMTU:
1998 			*i1 = icmp->icmp_ipv6_recvpathmtu;
1999 			break;
2000 		case IPV6_V6ONLY:
2001 			*i1 = 1;
2002 			break;
2003 		case IPV6_RECVHOPLIMIT:
2004 			*i1 = icmp->icmp_ipv6_recvhoplimit;
2005 			break;
2006 		case IPV6_RECVHOPOPTS:
2007 			*i1 = icmp->icmp_ipv6_recvhopopts;
2008 			break;
2009 		case IPV6_RECVDSTOPTS:
2010 			*i1 = icmp->icmp_ipv6_recvdstopts;
2011 			break;
2012 		case _OLD_IPV6_RECVDSTOPTS:
2013 			*i1 = icmp->icmp_old_ipv6_recvdstopts;
2014 			break;
2015 		case IPV6_RECVRTHDRDSTOPTS:
2016 			*i1 = icmp->icmp_ipv6_recvrtdstopts;
2017 			break;
2018 		case IPV6_RECVRTHDR:
2019 			*i1 = icmp->icmp_ipv6_recvrthdr;
2020 			break;
2021 		case IPV6_PKTINFO: {
2022 			/* XXX assumes that caller has room for max size! */
2023 			struct in6_pktinfo *pkti;
2024 
2025 			pkti = (struct in6_pktinfo *)ptr;
2026 			if (ipp->ipp_fields & IPPF_IFINDEX)
2027 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
2028 			else
2029 				pkti->ipi6_ifindex = 0;
2030 			if (ipp->ipp_fields & IPPF_ADDR)
2031 				pkti->ipi6_addr = ipp->ipp_addr;
2032 			else
2033 				pkti->ipi6_addr = ipv6_all_zeros;
2034 			ret = sizeof (struct in6_pktinfo);
2035 			goto done;
2036 		}
2037 		case IPV6_NEXTHOP: {
2038 			sin6_t *sin6 = (sin6_t *)ptr;
2039 
2040 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
2041 				return (0);
2042 			*sin6 = sin6_null;
2043 			sin6->sin6_family = AF_INET6;
2044 			sin6->sin6_addr = ipp->ipp_nexthop;
2045 			ret = (sizeof (sin6_t));
2046 			goto done;
2047 		}
2048 		case IPV6_HOPOPTS:
2049 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
2050 				return (0);
2051 			if (ipp->ipp_hopoptslen <= icmp->icmp_label_len_v6)
2052 				return (0);
2053 			bcopy((char *)ipp->ipp_hopopts +
2054 			    icmp->icmp_label_len_v6, ptr,
2055 			    ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
2056 			if (icmp->icmp_label_len_v6 > 0) {
2057 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
2058 				ptr[1] = (ipp->ipp_hopoptslen -
2059 				    icmp->icmp_label_len_v6 + 7) / 8 - 1;
2060 			}
2061 			ret = (ipp->ipp_hopoptslen - icmp->icmp_label_len_v6);
2062 			goto done;
2063 		case IPV6_RTHDRDSTOPTS:
2064 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
2065 				return (0);
2066 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
2067 			ret = ipp->ipp_rtdstoptslen;
2068 			goto done;
2069 		case IPV6_RTHDR:
2070 			if (!(ipp->ipp_fields & IPPF_RTHDR))
2071 				return (0);
2072 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
2073 			ret = ipp->ipp_rthdrlen;
2074 			goto done;
2075 		case IPV6_DSTOPTS:
2076 			if (!(ipp->ipp_fields & IPPF_DSTOPTS)) {
2077 				ret = 0;
2078 				goto done;
2079 			}
2080 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
2081 			ret = ipp->ipp_dstoptslen;
2082 			goto done;
2083 		case IPV6_PATHMTU:
2084 			if (!(ipp->ipp_fields & IPPF_PATHMTU)) {
2085 				ret = 0;
2086 			} else {
2087 				ret = ip_fill_mtuinfo(
2088 				    &icmp->icmp_v6dst.sin6_addr, 0,
2089 				    (struct ip6_mtuinfo *)ptr,
2090 				    is->is_netstack);
2091 			}
2092 			goto done;
2093 		case IPV6_TCLASS:
2094 			if (ipp->ipp_fields & IPPF_TCLASS)
2095 				*i1 = ipp->ipp_tclass;
2096 			else
2097 				*i1 = IPV6_FLOW_TCLASS(
2098 				    IPV6_DEFAULT_VERS_AND_FLOW);
2099 			break;
2100 		default:
2101 			ret = -1;
2102 			goto done;
2103 		}
2104 		break;
2105 	case IPPROTO_ICMPV6:
2106 		/*
2107 		 * Only allow IPv6 option processing on native IPv6 sockets.
2108 		 */
2109 		if (icmp->icmp_family != AF_INET6) {
2110 			ret = -1;
2111 		}
2112 
2113 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
2114 			ret = -1;
2115 		}
2116 
2117 		switch (name) {
2118 		case ICMP6_FILTER:
2119 			if (icmp->icmp_filter == NULL) {
2120 				/* Make it look like "pass all" */
2121 				ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr);
2122 			} else {
2123 				(void) bcopy(icmp->icmp_filter, ptr,
2124 				    sizeof (icmp6_filter_t));
2125 			}
2126 			ret = sizeof (icmp6_filter_t);
2127 			goto done;
2128 		default:
2129 			ret = -1;
2130 			goto done;
2131 		}
2132 	default:
2133 		ret = -1;
2134 		goto done;
2135 	}
2136 	ret = sizeof (int);
2137 done:
2138 	return (ret);
2139 }
2140 
2141 /*
2142  * This routine retrieves the current status of socket options.
2143  * It returns the size of the option retrieved.
2144  */
2145 int
2146 icmp_tpi_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
2147 {
2148 	conn_t  *connp = Q_TO_CONN(q);
2149 	icmp_t	*icmp = connp->conn_icmp;
2150 	int 	err;
2151 
2152 	rw_enter(&icmp->icmp_rwlock, RW_READER);
2153 	err = icmp_opt_get(connp, level, name, ptr);
2154 	rw_exit(&icmp->icmp_rwlock);
2155 	return (err);
2156 }
2157 
2158 int
2159 icmp_do_opt_set(conn_t *connp, int level, int name, uint_t inlen,
2160     uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr,
2161     void *thisdg_attrs, boolean_t checkonly)
2162 {
2163 
2164 	int	*i1 = (int *)invalp;
2165 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
2166 	icmp_t *icmp = connp->conn_icmp;
2167 	icmp_stack_t *is = icmp->icmp_is;
2168 	int	error;
2169 
2170 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
2171 	/*
2172 	 * For fixed length options, no sanity check
2173 	 * of passed in length is done. It is assumed *_optcom_req()
2174 	 * routines do the right thing.
2175 	 */
2176 	switch (level) {
2177 	case SOL_SOCKET:
2178 		switch (name) {
2179 		case SO_DEBUG:
2180 			if (!checkonly)
2181 				icmp->icmp_debug = onoff;
2182 			break;
2183 		case SO_PROTOTYPE:
2184 			if ((*i1 & 0xFF) != IPPROTO_ICMP &&
2185 			    (*i1 & 0xFF) != IPPROTO_ICMPV6 &&
2186 			    secpolicy_net_rawaccess(cr) != 0) {
2187 				*outlenp = 0;
2188 				return (EACCES);
2189 			}
2190 			/* Can't use IPPROTO_RAW with IPv6 */
2191 			if ((*i1 & 0xFF) == IPPROTO_RAW &&
2192 			    icmp->icmp_family == AF_INET6) {
2193 				*outlenp = 0;
2194 				return (EPROTONOSUPPORT);
2195 			}
2196 			if (checkonly) {
2197 				/* T_CHECK case */
2198 				*(int *)outvalp = (*i1 & 0xFF);
2199 				break;
2200 			}
2201 			icmp->icmp_proto = *i1 & 0xFF;
2202 			if ((icmp->icmp_proto == IPPROTO_RAW ||
2203 			    icmp->icmp_proto == IPPROTO_IGMP) &&
2204 			    icmp->icmp_family == AF_INET)
2205 				icmp->icmp_hdrincl = 1;
2206 			else
2207 				icmp->icmp_hdrincl = 0;
2208 
2209 			if (icmp->icmp_family == AF_INET6 &&
2210 			    icmp->icmp_proto == IPPROTO_ICMPV6) {
2211 				/* Set offset for icmp6_cksum */
2212 				icmp->icmp_raw_checksum = 0;
2213 				icmp->icmp_checksum_off = 2;
2214 			}
2215 			if (icmp->icmp_proto == IPPROTO_UDP ||
2216 			    icmp->icmp_proto == IPPROTO_TCP ||
2217 			    icmp->icmp_proto == IPPROTO_SCTP) {
2218 				icmp->icmp_no_tp_cksum = 1;
2219 				icmp->icmp_sticky_ipp.ipp_fields |=
2220 				    IPPF_NO_CKSUM;
2221 			} else {
2222 				icmp->icmp_no_tp_cksum = 0;
2223 				icmp->icmp_sticky_ipp.ipp_fields &=
2224 				    ~IPPF_NO_CKSUM;
2225 			}
2226 
2227 			if (icmp->icmp_filter != NULL &&
2228 			    icmp->icmp_proto != IPPROTO_ICMPV6) {
2229 				kmem_free(icmp->icmp_filter,
2230 				    sizeof (icmp6_filter_t));
2231 				icmp->icmp_filter = NULL;
2232 			}
2233 
2234 			/* Rebuild the header template */
2235 			error = icmp_build_hdrs(icmp);
2236 			if (error != 0) {
2237 				*outlenp = 0;
2238 				return (error);
2239 			}
2240 
2241 			/*
2242 			 * For SCTP, we don't use icmp_bind_proto() for
2243 			 * raw socket binding.  Note that we do not need
2244 			 * to set *outlenp.
2245 			 * FIXME: how does SCTP work?
2246 			 */
2247 			if (icmp->icmp_proto == IPPROTO_SCTP)
2248 				return (0);
2249 
2250 			*outlenp = sizeof (int);
2251 			*(int *)outvalp = *i1 & 0xFF;
2252 
2253 			/* Drop lock across the bind operation */
2254 			rw_exit(&icmp->icmp_rwlock);
2255 			(void) icmp_bind_proto(connp);
2256 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2257 			return (0);
2258 		case SO_REUSEADDR:
2259 			if (!checkonly) {
2260 				icmp->icmp_reuseaddr = onoff;
2261 				PASS_OPT_TO_IP(connp);
2262 			}
2263 			break;
2264 
2265 		/*
2266 		 * The following three items are available here,
2267 		 * but are only meaningful to IP.
2268 		 */
2269 		case SO_DONTROUTE:
2270 			if (!checkonly) {
2271 				icmp->icmp_dontroute = onoff;
2272 				PASS_OPT_TO_IP(connp);
2273 			}
2274 			break;
2275 		case SO_USELOOPBACK:
2276 			if (!checkonly) {
2277 				icmp->icmp_useloopback = onoff;
2278 				PASS_OPT_TO_IP(connp);
2279 			}
2280 			break;
2281 		case SO_BROADCAST:
2282 			if (!checkonly) {
2283 				icmp->icmp_broadcast = onoff;
2284 				PASS_OPT_TO_IP(connp);
2285 			}
2286 			break;
2287 
2288 		case SO_SNDBUF:
2289 			if (*i1 > is->is_max_buf) {
2290 				*outlenp = 0;
2291 				return (ENOBUFS);
2292 			}
2293 			if (!checkonly) {
2294 				if (!IPCL_IS_NONSTR(connp)) {
2295 					connp->conn_wq->q_hiwat = *i1;
2296 				}
2297 				icmp->icmp_xmit_hiwat = *i1;
2298 			}
2299 			break;
2300 		case SO_RCVBUF:
2301 			if (*i1 > is->is_max_buf) {
2302 				*outlenp = 0;
2303 				return (ENOBUFS);
2304 			}
2305 			if (!checkonly) {
2306 				icmp->icmp_recv_hiwat = *i1;
2307 				rw_exit(&icmp->icmp_rwlock);
2308 				(void) proto_set_rx_hiwat(connp->conn_rq, connp,
2309 				    *i1);
2310 				rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2311 			}
2312 			break;
2313 		case SO_DGRAM_ERRIND:
2314 			if (!checkonly)
2315 				icmp->icmp_dgram_errind = onoff;
2316 			break;
2317 		case SO_ALLZONES:
2318 			/*
2319 			 * "soft" error (negative)
2320 			 * option not handled at this level
2321 			 * Note: Do not modify *outlenp
2322 			 */
2323 			return (-EINVAL);
2324 		case SO_TIMESTAMP:
2325 			if (!checkonly) {
2326 				icmp->icmp_timestamp = onoff;
2327 			}
2328 			break;
2329 		case SO_MAC_EXEMPT:
2330 			/*
2331 			 * "soft" error (negative)
2332 			 * option not handled at this level
2333 			 * Note: Do not modify *outlenp
2334 			 */
2335 			return (-EINVAL);
2336 		case SO_RCVTIMEO:
2337 		case SO_SNDTIMEO:
2338 			/*
2339 			 * Pass these two options in order for third part
2340 			 * protocol usage. Here just return directly.
2341 			 */
2342 			return (0);
2343 		/*
2344 		 * Following three not meaningful for icmp
2345 		 * Action is same as "default" so we keep them
2346 		 * in comments.
2347 		 * case SO_LINGER:
2348 		 * case SO_KEEPALIVE:
2349 		 * case SO_OOBINLINE:
2350 		 */
2351 		default:
2352 			*outlenp = 0;
2353 			return (EINVAL);
2354 		}
2355 		break;
2356 	case IPPROTO_IP:
2357 		/*
2358 		 * Only allow IPv4 option processing on IPv4 sockets.
2359 		 */
2360 		if (icmp->icmp_family != AF_INET) {
2361 			*outlenp = 0;
2362 			return (ENOPROTOOPT);
2363 		}
2364 		switch (name) {
2365 		case IP_OPTIONS:
2366 		case T_IP_OPTIONS:
2367 			/* Save options for use by IP. */
2368 			if ((inlen & 0x3) ||
2369 			    inlen + icmp->icmp_label_len > IP_MAX_OPT_LENGTH) {
2370 				*outlenp = 0;
2371 				return (EINVAL);
2372 			}
2373 			if (checkonly)
2374 				break;
2375 
2376 			if (!tsol_option_set(&icmp->icmp_ip_snd_options,
2377 			    &icmp->icmp_ip_snd_options_len,
2378 			    icmp->icmp_label_len, invalp, inlen)) {
2379 				*outlenp = 0;
2380 				return (ENOMEM);
2381 			}
2382 
2383 			icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
2384 			    icmp->icmp_ip_snd_options_len;
2385 			rw_exit(&icmp->icmp_rwlock);
2386 			(void) proto_set_tx_wroff(connp->conn_rq == NULL ? NULL:
2387 			    RD(connp->conn_rq), connp,
2388 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
2389 			rw_enter(&icmp->icmp_rwlock, RW_WRITER);
2390 			break;
2391 		case IP_HDRINCL:
2392 			if (!checkonly)
2393 				icmp->icmp_hdrincl = onoff;
2394 			break;
2395 		case IP_TOS:
2396 		case T_IP_TOS:
2397 			if (!checkonly) {
2398 				icmp->icmp_type_of_service = (uint8_t)*i1;
2399 			}
2400 			break;
2401 		case IP_TTL:
2402 			if (!checkonly) {
2403 				icmp->icmp_ttl = (uint8_t)*i1;
2404 			}
2405 			break;
2406 		case IP_MULTICAST_IF:
2407 			/*
2408 			 * TODO should check OPTMGMT reply and undo this if
2409 			 * there is an error.
2410 			 */
2411 			if (!checkonly) {
2412 				icmp->icmp_multicast_if_addr = *i1;
2413 				PASS_OPT_TO_IP(connp);
2414 			}
2415 			break;
2416 		case IP_MULTICAST_TTL:
2417 			if (!checkonly)
2418 				icmp->icmp_multicast_ttl = *invalp;
2419 			break;
2420 		case IP_MULTICAST_LOOP:
2421 			if (!checkonly) {
2422 				connp->conn_multicast_loop =
2423 				    (*invalp == 0) ? 0 : 1;
2424 				PASS_OPT_TO_IP(connp);
2425 			}
2426 			break;
2427 		case IP_BOUND_IF:
2428 			if (!checkonly) {
2429 				icmp->icmp_bound_if = *i1;
2430 				PASS_OPT_TO_IP(connp);
2431 			}
2432 			break;
2433 		case IP_UNSPEC_SRC:
2434 			if (!checkonly) {
2435 				icmp->icmp_unspec_source = onoff;
2436 				PASS_OPT_TO_IP(connp);
2437 			}
2438 			break;
2439 		case IP_BROADCAST_TTL:
2440 			if (!checkonly)
2441 				connp->conn_broadcast_ttl = *invalp;
2442 			break;
2443 		case IP_RECVIF:
2444 			if (!checkonly) {
2445 				icmp->icmp_recvif = onoff;
2446 			}
2447 			/*
2448 			 * pass to ip
2449 			 */
2450 			return (-EINVAL);
2451 		case IP_PKTINFO: {
2452 			/*
2453 			 * This also handles IP_RECVPKTINFO.
2454 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
2455 			 * Differentiation is based on the size of the argument
2456 			 * passed in.
2457 			 */
2458 			struct in_pktinfo *pktinfop;
2459 			ip4_pkt_t *attr_pktinfop;
2460 
2461 			if (checkonly)
2462 				break;
2463 
2464 			if (inlen == sizeof (int)) {
2465 				/*
2466 				 * This is IP_RECVPKTINFO option.
2467 				 * Keep a local copy of wether this option is
2468 				 * set or not and pass it down to IP for
2469 				 * processing.
2470 				 */
2471 				icmp->icmp_ip_recvpktinfo = onoff;
2472 				return (-EINVAL);
2473 			}
2474 
2475 
2476 			if (inlen != sizeof (struct in_pktinfo)) {
2477 				return (EINVAL);
2478 			}
2479 
2480 			if ((attr_pktinfop = (ip4_pkt_t *)thisdg_attrs)
2481 			    == NULL) {
2482 				/*
2483 				 * sticky option is not supported
2484 				 */
2485 				return (EINVAL);
2486 			}
2487 
2488 			pktinfop = (struct in_pktinfo *)invalp;
2489 
2490 			/*
2491 			 * Atleast one of the values should be specified
2492 			 */
2493 			if (pktinfop->ipi_ifindex == 0 &&
2494 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
2495 				return (EINVAL);
2496 			}
2497 
2498 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
2499 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
2500 		}
2501 			break;
2502 		case IP_ADD_MEMBERSHIP:
2503 		case IP_DROP_MEMBERSHIP:
2504 		case IP_BLOCK_SOURCE:
2505 		case IP_UNBLOCK_SOURCE:
2506 		case IP_ADD_SOURCE_MEMBERSHIP:
2507 		case IP_DROP_SOURCE_MEMBERSHIP:
2508 		case MCAST_JOIN_GROUP:
2509 		case MCAST_LEAVE_GROUP:
2510 		case MCAST_BLOCK_SOURCE:
2511 		case MCAST_UNBLOCK_SOURCE:
2512 		case MCAST_JOIN_SOURCE_GROUP:
2513 		case MCAST_LEAVE_SOURCE_GROUP:
2514 		case MRT_INIT:
2515 		case MRT_DONE:
2516 		case MRT_ADD_VIF:
2517 		case MRT_DEL_VIF:
2518 		case MRT_ADD_MFC:
2519 		case MRT_DEL_MFC:
2520 		case MRT_VERSION:
2521 		case MRT_ASSERT:
2522 		case IP_SEC_OPT:
2523 		case IP_NEXTHOP:
2524 			/*
2525 			 * "soft" error (negative)
2526 			 * option not handled at this level
2527 			 * Note: Do not modify *outlenp
2528 			 */
2529 			return (-EINVAL);
2530 		default:
2531 			*outlenp = 0;
2532 			return (EINVAL);
2533 		}
2534 		break;
2535 	case IPPROTO_IPV6: {
2536 		ip6_pkt_t		*ipp;
2537 		boolean_t		sticky;
2538 
2539 		if (icmp->icmp_family != AF_INET6) {
2540 			*outlenp = 0;
2541 			return (ENOPROTOOPT);
2542 		}
2543 		/*
2544 		 * Deal with both sticky options and ancillary data
2545 		 */
2546 		if (thisdg_attrs == NULL) {
2547 			/* sticky options, or none */
2548 			ipp = &icmp->icmp_sticky_ipp;
2549 			sticky = B_TRUE;
2550 		} else {
2551 			/* ancillary data */
2552 			ipp = (ip6_pkt_t *)thisdg_attrs;
2553 			sticky = B_FALSE;
2554 		}
2555 
2556 		switch (name) {
2557 		case IPV6_MULTICAST_IF:
2558 			if (!checkonly) {
2559 				icmp->icmp_multicast_if_index = *i1;
2560 				PASS_OPT_TO_IP(connp);
2561 			}
2562 			break;
2563 		case IPV6_UNICAST_HOPS:
2564 			/* -1 means use default */
2565 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2566 				*outlenp = 0;
2567 				return (EINVAL);
2568 			}
2569 			if (!checkonly) {
2570 				if (*i1 == -1) {
2571 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2572 					    is->is_ipv6_hoplimit;
2573 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
2574 					/* Pass modified value to IP. */
2575 					*i1 = ipp->ipp_hoplimit;
2576 				} else {
2577 					icmp->icmp_ttl = ipp->ipp_unicast_hops =
2578 					    (uint8_t)*i1;
2579 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
2580 				}
2581 				/* Rebuild the header template */
2582 				error = icmp_build_hdrs(icmp);
2583 				if (error != 0) {
2584 					*outlenp = 0;
2585 					return (error);
2586 				}
2587 			}
2588 			break;
2589 		case IPV6_MULTICAST_HOPS:
2590 			/* -1 means use default */
2591 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
2592 				*outlenp = 0;
2593 				return (EINVAL);
2594 			}
2595 			if (!checkonly) {
2596 				if (*i1 == -1) {
2597 					icmp->icmp_multicast_ttl =
2598 					    ipp->ipp_multicast_hops =
2599 					    IP_DEFAULT_MULTICAST_TTL;
2600 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
2601 					/* Pass modified value to IP. */
2602 					*i1 = icmp->icmp_multicast_ttl;
2603 				} else {
2604 					icmp->icmp_multicast_ttl =
2605 					    ipp->ipp_multicast_hops =
2606 					    (uint8_t)*i1;
2607 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
2608 				}
2609 			}
2610 			break;
2611 		case IPV6_MULTICAST_LOOP:
2612 			if (*i1 != 0 && *i1 != 1) {
2613 				*outlenp = 0;
2614 				return (EINVAL);
2615 			}
2616 			if (!checkonly) {
2617 				connp->conn_multicast_loop = *i1;
2618 				PASS_OPT_TO_IP(connp);
2619 			}
2620 			break;
2621 		case IPV6_CHECKSUM:
2622 			/*
2623 			 * Integer offset into the user data of where the
2624 			 * checksum is located.
2625 			 * Offset of -1 disables option.
2626 			 * Does not apply to IPPROTO_ICMPV6.
2627 			 */
2628 			if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) {
2629 				*outlenp = 0;
2630 				return (EINVAL);
2631 			}
2632 			if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) {
2633 				/* Negative or not 16 bit aligned offset */
2634 				*outlenp = 0;
2635 				return (EINVAL);
2636 			}
2637 			if (checkonly)
2638 				break;
2639 
2640 			if (*i1 == -1) {
2641 				icmp->icmp_raw_checksum = 0;
2642 				ipp->ipp_fields &= ~IPPF_RAW_CKSUM;
2643 			} else {
2644 				icmp->icmp_raw_checksum = 1;
2645 				icmp->icmp_checksum_off = *i1;
2646 				ipp->ipp_fields |= IPPF_RAW_CKSUM;
2647 			}
2648 			/* Rebuild the header template */
2649 			error = icmp_build_hdrs(icmp);
2650 			if (error != 0) {
2651 				*outlenp = 0;
2652 				return (error);
2653 			}
2654 			break;
2655 		case IPV6_JOIN_GROUP:
2656 		case IPV6_LEAVE_GROUP:
2657 		case MCAST_JOIN_GROUP:
2658 		case MCAST_LEAVE_GROUP:
2659 		case MCAST_BLOCK_SOURCE:
2660 		case MCAST_UNBLOCK_SOURCE:
2661 		case MCAST_JOIN_SOURCE_GROUP:
2662 		case MCAST_LEAVE_SOURCE_GROUP:
2663 			/*
2664 			 * "soft" error (negative)
2665 			 * option not handled at this level
2666 			 * Note: Do not modify *outlenp
2667 			 */
2668 			return (-EINVAL);
2669 		case IPV6_BOUND_IF:
2670 			if (!checkonly) {
2671 				icmp->icmp_bound_if = *i1;
2672 				PASS_OPT_TO_IP(connp);
2673 			}
2674 			break;
2675 		case IPV6_UNSPEC_SRC:
2676 			if (!checkonly) {
2677 				icmp->icmp_unspec_source = onoff;
2678 				PASS_OPT_TO_IP(connp);
2679 			}
2680 			break;
2681 		case IPV6_RECVTCLASS:
2682 			if (!checkonly) {
2683 				icmp->icmp_ipv6_recvtclass = onoff;
2684 				PASS_OPT_TO_IP(connp);
2685 			}
2686 			break;
2687 		/*
2688 		 * Set boolean switches for ancillary data delivery
2689 		 */
2690 		case IPV6_RECVPKTINFO:
2691 			if (!checkonly) {
2692 				icmp->icmp_ip_recvpktinfo = onoff;
2693 				PASS_OPT_TO_IP(connp);
2694 			}
2695 			break;
2696 		case IPV6_RECVPATHMTU:
2697 			if (!checkonly) {
2698 				icmp->icmp_ipv6_recvpathmtu = onoff;
2699 				PASS_OPT_TO_IP(connp);
2700 			}
2701 			break;
2702 		case IPV6_RECVHOPLIMIT:
2703 			if (!checkonly) {
2704 				icmp->icmp_ipv6_recvhoplimit = onoff;
2705 				PASS_OPT_TO_IP(connp);
2706 			}
2707 			break;
2708 		case IPV6_RECVHOPOPTS:
2709 			if (!checkonly) {
2710 				icmp->icmp_ipv6_recvhopopts = onoff;
2711 				PASS_OPT_TO_IP(connp);
2712 			}
2713 			break;
2714 		case IPV6_RECVDSTOPTS:
2715 			if (!checkonly) {
2716 				icmp->icmp_ipv6_recvdstopts = onoff;
2717 				PASS_OPT_TO_IP(connp);
2718 			}
2719 			break;
2720 		case _OLD_IPV6_RECVDSTOPTS:
2721 			if (!checkonly)
2722 				icmp->icmp_old_ipv6_recvdstopts = onoff;
2723 			break;
2724 		case IPV6_RECVRTHDRDSTOPTS:
2725 			if (!checkonly) {
2726 				icmp->icmp_ipv6_recvrtdstopts = onoff;
2727 				PASS_OPT_TO_IP(connp);
2728 			}
2729 			break;
2730 		case IPV6_RECVRTHDR:
2731 			if (!checkonly) {
2732 				icmp->icmp_ipv6_recvrthdr = onoff;
2733 				PASS_OPT_TO_IP(connp);
2734 			}
2735 			break;
2736 		/*
2737 		 * Set sticky options or ancillary data.
2738 		 * If sticky options, (re)build any extension headers
2739 		 * that might be needed as a result.
2740 		 */
2741 		case IPV6_PKTINFO:
2742 			/*
2743 			 * The source address and ifindex are verified
2744 			 * in ip_opt_set(). For ancillary data the
2745 			 * source address is checked in ip_wput_v6.
2746 			 */
2747 			if (inlen != 0 && inlen !=
2748 			    sizeof (struct in6_pktinfo)) {
2749 				return (EINVAL);
2750 			}
2751 			if (checkonly)
2752 				break;
2753 
2754 			if (inlen == 0) {
2755 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
2756 				ipp->ipp_sticky_ignored |=
2757 				    (IPPF_IFINDEX|IPPF_ADDR);
2758 			} else {
2759 				struct in6_pktinfo *pkti;
2760 
2761 				pkti = (struct in6_pktinfo *)invalp;
2762 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
2763 				ipp->ipp_addr = pkti->ipi6_addr;
2764 				if (ipp->ipp_ifindex != 0)
2765 					ipp->ipp_fields |= IPPF_IFINDEX;
2766 				else
2767 					ipp->ipp_fields &= ~IPPF_IFINDEX;
2768 				if (!IN6_IS_ADDR_UNSPECIFIED(
2769 				    &ipp->ipp_addr))
2770 					ipp->ipp_fields |= IPPF_ADDR;
2771 				else
2772 					ipp->ipp_fields &= ~IPPF_ADDR;
2773 			}
2774 			if (sticky) {
2775 				error = icmp_build_hdrs(icmp);
2776 				if (error != 0)
2777 					return (error);
2778 				PASS_OPT_TO_IP(connp);
2779 			}
2780 			break;
2781 		case IPV6_HOPLIMIT:
2782 			/* This option can only be used as ancillary data. */
2783 			if (sticky)
2784 				return (EINVAL);
2785 			if (inlen != 0 && inlen != sizeof (int))
2786 				return (EINVAL);
2787 			if (checkonly)
2788 				break;
2789 
2790 			if (inlen == 0) {
2791 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
2792 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
2793 			} else {
2794 				if (*i1 > 255 || *i1 < -1)
2795 					return (EINVAL);
2796 				if (*i1 == -1)
2797 					ipp->ipp_hoplimit =
2798 					    is->is_ipv6_hoplimit;
2799 				else
2800 					ipp->ipp_hoplimit = *i1;
2801 				ipp->ipp_fields |= IPPF_HOPLIMIT;
2802 			}
2803 			break;
2804 		case IPV6_TCLASS:
2805 			/*
2806 			 * IPV6_RECVTCLASS accepts -1 as use kernel default
2807 			 * and [0, 255] as the actualy traffic class.
2808 			 */
2809 			if (inlen != 0 && inlen != sizeof (int)) {
2810 				return (EINVAL);
2811 			}
2812 			if (checkonly)
2813 				break;
2814 
2815 			if (inlen == 0) {
2816 				ipp->ipp_fields &= ~IPPF_TCLASS;
2817 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
2818 			} else {
2819 				if (*i1 >= 256 || *i1 < -1)
2820 					return (EINVAL);
2821 				if (*i1 == -1) {
2822 					ipp->ipp_tclass =
2823 					    IPV6_FLOW_TCLASS(
2824 					    IPV6_DEFAULT_VERS_AND_FLOW);
2825 				} else {
2826 					ipp->ipp_tclass = *i1;
2827 				}
2828 				ipp->ipp_fields |= IPPF_TCLASS;
2829 			}
2830 			if (sticky) {
2831 				error = icmp_build_hdrs(icmp);
2832 				if (error != 0)
2833 					return (error);
2834 			}
2835 			break;
2836 		case IPV6_NEXTHOP:
2837 			/*
2838 			 * IP will verify that the nexthop is reachable
2839 			 * and fail for sticky options.
2840 			 */
2841 			if (inlen != 0 && inlen != sizeof (sin6_t)) {
2842 				return (EINVAL);
2843 			}
2844 			if (checkonly)
2845 				break;
2846 
2847 			if (inlen == 0) {
2848 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
2849 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
2850 			} else {
2851 				sin6_t *sin6 = (sin6_t *)invalp;
2852 
2853 				if (sin6->sin6_family != AF_INET6) {
2854 					return (EAFNOSUPPORT);
2855 				}
2856 				if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
2857 					return (EADDRNOTAVAIL);
2858 				}
2859 				ipp->ipp_nexthop = sin6->sin6_addr;
2860 				if (!IN6_IS_ADDR_UNSPECIFIED(
2861 				    &ipp->ipp_nexthop))
2862 					ipp->ipp_fields |= IPPF_NEXTHOP;
2863 				else
2864 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
2865 			}
2866 			if (sticky) {
2867 				error = icmp_build_hdrs(icmp);
2868 				if (error != 0)
2869 					return (error);
2870 				PASS_OPT_TO_IP(connp);
2871 			}
2872 			break;
2873 		case IPV6_HOPOPTS: {
2874 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
2875 			/*
2876 			 * Sanity checks - minimum size, size a multiple of
2877 			 * eight bytes, and matching size passed in.
2878 			 */
2879 			if (inlen != 0 &&
2880 			    inlen != (8 * (hopts->ip6h_len + 1))) {
2881 				return (EINVAL);
2882 			}
2883 
2884 			if (checkonly)
2885 				break;
2886 			error = optcom_pkt_set(invalp, inlen, sticky,
2887 			    (uchar_t **)&ipp->ipp_hopopts,
2888 			    &ipp->ipp_hopoptslen,
2889 			    sticky ? icmp->icmp_label_len_v6 : 0);
2890 			if (error != 0)
2891 				return (error);
2892 			if (ipp->ipp_hopoptslen == 0) {
2893 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
2894 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
2895 			} else {
2896 				ipp->ipp_fields |= IPPF_HOPOPTS;
2897 			}
2898 			if (sticky) {
2899 				error = icmp_build_hdrs(icmp);
2900 				if (error != 0)
2901 					return (error);
2902 			}
2903 			break;
2904 		}
2905 		case IPV6_RTHDRDSTOPTS: {
2906 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2907 
2908 			/*
2909 			 * Sanity checks - minimum size, size a multiple of
2910 			 * eight bytes, and matching size passed in.
2911 			 */
2912 			if (inlen != 0 &&
2913 			    inlen != (8 * (dopts->ip6d_len + 1)))
2914 				return (EINVAL);
2915 
2916 			if (checkonly)
2917 				break;
2918 
2919 			if (inlen == 0) {
2920 				if (sticky &&
2921 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
2922 					kmem_free(ipp->ipp_rtdstopts,
2923 					    ipp->ipp_rtdstoptslen);
2924 					ipp->ipp_rtdstopts = NULL;
2925 					ipp->ipp_rtdstoptslen = 0;
2926 				}
2927 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
2928 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
2929 			} else {
2930 				error = optcom_pkt_set(invalp, inlen, sticky,
2931 				    (uchar_t **)&ipp->ipp_rtdstopts,
2932 				    &ipp->ipp_rtdstoptslen, 0);
2933 				if (error != 0)
2934 					return (error);
2935 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
2936 			}
2937 			if (sticky) {
2938 				error = icmp_build_hdrs(icmp);
2939 				if (error != 0)
2940 					return (error);
2941 			}
2942 			break;
2943 		}
2944 		case IPV6_DSTOPTS: {
2945 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
2946 
2947 			/*
2948 			 * Sanity checks - minimum size, size a multiple of
2949 			 * eight bytes, and matching size passed in.
2950 			 */
2951 			if (inlen != 0 &&
2952 			    inlen != (8 * (dopts->ip6d_len + 1)))
2953 				return (EINVAL);
2954 
2955 			if (checkonly)
2956 				break;
2957 
2958 			if (inlen == 0) {
2959 				if (sticky &&
2960 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
2961 					kmem_free(ipp->ipp_dstopts,
2962 					    ipp->ipp_dstoptslen);
2963 					ipp->ipp_dstopts = NULL;
2964 					ipp->ipp_dstoptslen = 0;
2965 				}
2966 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
2967 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
2968 			} else {
2969 				error = optcom_pkt_set(invalp, inlen, sticky,
2970 				    (uchar_t **)&ipp->ipp_dstopts,
2971 				    &ipp->ipp_dstoptslen, 0);
2972 				if (error != 0)
2973 					return (error);
2974 				ipp->ipp_fields |= IPPF_DSTOPTS;
2975 			}
2976 			if (sticky) {
2977 				error = icmp_build_hdrs(icmp);
2978 				if (error != 0)
2979 					return (error);
2980 			}
2981 			break;
2982 		}
2983 		case IPV6_RTHDR: {
2984 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
2985 
2986 			/*
2987 			 * Sanity checks - minimum size, size a multiple of
2988 			 * eight bytes, and matching size passed in.
2989 			 */
2990 			if (inlen != 0 &&
2991 			    inlen != (8 * (rt->ip6r_len + 1)))
2992 				return (EINVAL);
2993 
2994 			if (checkonly)
2995 				break;
2996 
2997 			if (inlen == 0) {
2998 				if (sticky &&
2999 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
3000 					kmem_free(ipp->ipp_rthdr,
3001 					    ipp->ipp_rthdrlen);
3002 					ipp->ipp_rthdr = NULL;
3003 					ipp->ipp_rthdrlen = 0;
3004 				}
3005 				ipp->ipp_fields &= ~IPPF_RTHDR;
3006 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
3007 			} else {
3008 				error = optcom_pkt_set(invalp, inlen, sticky,
3009 				    (uchar_t **)&ipp->ipp_rthdr,
3010 				    &ipp->ipp_rthdrlen, 0);
3011 				if (error != 0)
3012 					return (error);
3013 				ipp->ipp_fields |= IPPF_RTHDR;
3014 			}
3015 			if (sticky) {
3016 				error = icmp_build_hdrs(icmp);
3017 				if (error != 0)
3018 					return (error);
3019 			}
3020 			break;
3021 		}
3022 
3023 		case IPV6_DONTFRAG:
3024 			if (checkonly)
3025 				break;
3026 
3027 			if (onoff) {
3028 				ipp->ipp_fields |= IPPF_DONTFRAG;
3029 			} else {
3030 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
3031 			}
3032 			break;
3033 
3034 		case IPV6_USE_MIN_MTU:
3035 			if (inlen != sizeof (int))
3036 				return (EINVAL);
3037 
3038 			if (*i1 < -1 || *i1 > 1)
3039 				return (EINVAL);
3040 
3041 			if (checkonly)
3042 				break;
3043 
3044 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
3045 			ipp->ipp_use_min_mtu = *i1;
3046 			break;
3047 
3048 		/*
3049 		 * This option can't be set.  Its only returned via
3050 		 * getsockopt() or ancillary data.
3051 		 */
3052 		case IPV6_PATHMTU:
3053 			return (EINVAL);
3054 
3055 		case IPV6_SEC_OPT:
3056 		case IPV6_SRC_PREFERENCES:
3057 		case IPV6_V6ONLY:
3058 			/* Handled at IP level */
3059 			return (-EINVAL);
3060 		default:
3061 			*outlenp = 0;
3062 			return (EINVAL);
3063 		}
3064 		break;
3065 	}		/* end IPPROTO_IPV6 */
3066 
3067 	case IPPROTO_ICMPV6:
3068 		/*
3069 		 * Only allow IPv6 option processing on IPv6 sockets.
3070 		 */
3071 		if (icmp->icmp_family != AF_INET6) {
3072 			*outlenp = 0;
3073 			return (ENOPROTOOPT);
3074 		}
3075 		if (icmp->icmp_proto != IPPROTO_ICMPV6) {
3076 			*outlenp = 0;
3077 			return (ENOPROTOOPT);
3078 		}
3079 		switch (name) {
3080 		case ICMP6_FILTER:
3081 			if (!checkonly) {
3082 				if ((inlen != 0) &&
3083 				    (inlen != sizeof (icmp6_filter_t)))
3084 					return (EINVAL);
3085 
3086 				if (inlen == 0) {
3087 					if (icmp->icmp_filter != NULL) {
3088 						kmem_free(icmp->icmp_filter,
3089 						    sizeof (icmp6_filter_t));
3090 						icmp->icmp_filter = NULL;
3091 					}
3092 				} else {
3093 					if (icmp->icmp_filter == NULL) {
3094 						icmp->icmp_filter = kmem_alloc(
3095 						    sizeof (icmp6_filter_t),
3096 						    KM_NOSLEEP);
3097 						if (icmp->icmp_filter == NULL) {
3098 							*outlenp = 0;
3099 							return (ENOBUFS);
3100 						}
3101 					}
3102 					(void) bcopy(invalp, icmp->icmp_filter,
3103 					    inlen);
3104 				}
3105 			}
3106 			break;
3107 
3108 		default:
3109 			*outlenp = 0;
3110 			return (EINVAL);
3111 		}
3112 		break;
3113 	default:
3114 		*outlenp = 0;
3115 		return (EINVAL);
3116 	}
3117 	/*
3118 	 * Common case of OK return with outval same as inval.
3119 	 */
3120 	if (invalp != outvalp) {
3121 		/* don't trust bcopy for identical src/dst */
3122 		(void) bcopy(invalp, outvalp, inlen);
3123 	}
3124 	*outlenp = inlen;
3125 	return (0);
3126 }
3127 
3128 /* This routine sets socket options. */
3129 /* ARGSUSED */
3130 int
3131 icmp_opt_set(conn_t *connp, uint_t optset_context, int level, int name,
3132     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3133     void *thisdg_attrs, cred_t *cr)
3134 {
3135 	boolean_t checkonly;
3136 	int	error;
3137 
3138 	error = 0;
3139 	switch (optset_context) {
3140 	case SETFN_OPTCOM_CHECKONLY:
3141 		checkonly = B_TRUE;
3142 		/*
3143 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3144 		 * inlen != 0 implies value supplied and
3145 		 * 	we have to "pretend" to set it.
3146 		 * inlen == 0 implies that there is no
3147 		 * 	value part in T_CHECK request and just validation
3148 		 * done elsewhere should be enough, we just return here.
3149 		 */
3150 		if (inlen == 0) {
3151 			*outlenp = 0;
3152 			error = 0;
3153 			goto done;
3154 		}
3155 		break;
3156 	case SETFN_OPTCOM_NEGOTIATE:
3157 		checkonly = B_FALSE;
3158 		break;
3159 	case SETFN_UD_NEGOTIATE:
3160 	case SETFN_CONN_NEGOTIATE:
3161 		checkonly = B_FALSE;
3162 		/*
3163 		 * Negotiating local and "association-related" options
3164 		 * through T_UNITDATA_REQ.
3165 		 *
3166 		 * Following routine can filter out ones we do not
3167 		 * want to be "set" this way.
3168 		 */
3169 		if (!icmp_opt_allow_udr_set(level, name)) {
3170 			*outlenp = 0;
3171 			error = EINVAL;
3172 			goto done;
3173 		}
3174 		break;
3175 	default:
3176 		/*
3177 		 * We should never get here
3178 		 */
3179 		*outlenp = 0;
3180 		error = EINVAL;
3181 		goto done;
3182 	}
3183 
3184 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3185 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3186 	error = icmp_do_opt_set(connp, level, name, inlen, invalp, outlenp,
3187 	    outvalp, cr, thisdg_attrs, checkonly);
3188 
3189 done:
3190 	return (error);
3191 }
3192 
3193 /* This routine sets socket options. */
3194 /* ARGSUSED */
3195 int
3196 icmp_tpi_opt_set(queue_t *q, uint_t optset_context, int level, int name,
3197     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3198     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3199 {
3200 	conn_t	*connp =  Q_TO_CONN(q);
3201 	icmp_t	*icmp;
3202 	int error;
3203 
3204 	icmp = connp->conn_icmp;
3205 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3206 	error = icmp_opt_set(connp, optset_context, level, name, inlen, invalp,
3207 	    outlenp, outvalp, thisdg_attrs, cr);
3208 	rw_exit(&icmp->icmp_rwlock);
3209 	return (error);
3210 }
3211 
3212 /*
3213  * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl,
3214  * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum.
3215  * The headers include ip6i_t (if needed), ip6_t, and any sticky extension
3216  * headers.
3217  * Returns failure if can't allocate memory.
3218  */
3219 static int
3220 icmp_build_hdrs(icmp_t *icmp)
3221 {
3222 	icmp_stack_t *is = icmp->icmp_is;
3223 	uchar_t	*hdrs;
3224 	uint_t	hdrs_len;
3225 	ip6_t	*ip6h;
3226 	ip6i_t	*ip6i;
3227 	ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp;
3228 
3229 	ASSERT(RW_WRITE_HELD(&icmp->icmp_rwlock));
3230 	hdrs_len = ip_total_hdrs_len_v6(ipp);
3231 	ASSERT(hdrs_len != 0);
3232 	if (hdrs_len != icmp->icmp_sticky_hdrs_len) {
3233 		/* Need to reallocate */
3234 		if (hdrs_len != 0) {
3235 			hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
3236 			if (hdrs == NULL)
3237 				return (ENOMEM);
3238 		} else {
3239 			hdrs = NULL;
3240 		}
3241 		if (icmp->icmp_sticky_hdrs_len != 0) {
3242 			kmem_free(icmp->icmp_sticky_hdrs,
3243 			    icmp->icmp_sticky_hdrs_len);
3244 		}
3245 		icmp->icmp_sticky_hdrs = hdrs;
3246 		icmp->icmp_sticky_hdrs_len = hdrs_len;
3247 	}
3248 	ip_build_hdrs_v6(icmp->icmp_sticky_hdrs,
3249 	    icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto);
3250 
3251 	/* Set header fields not in ipp */
3252 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
3253 		ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs;
3254 		ip6h = (ip6_t *)&ip6i[1];
3255 
3256 		if (ipp->ipp_fields & IPPF_RAW_CKSUM) {
3257 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
3258 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
3259 		}
3260 		if (ipp->ipp_fields & IPPF_NO_CKSUM) {
3261 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
3262 		}
3263 	} else {
3264 		ip6h = (ip6_t *)icmp->icmp_sticky_hdrs;
3265 	}
3266 
3267 	if (!(ipp->ipp_fields & IPPF_ADDR))
3268 		ip6h->ip6_src = icmp->icmp_v6src;
3269 
3270 	/* Try to get everything in a single mblk */
3271 	if (hdrs_len > icmp->icmp_max_hdr_len) {
3272 		icmp->icmp_max_hdr_len = hdrs_len;
3273 		rw_exit(&icmp->icmp_rwlock);
3274 		(void) proto_set_tx_wroff(icmp->icmp_connp->conn_rq,
3275 		    icmp->icmp_connp,
3276 		    icmp->icmp_max_hdr_len + is->is_wroff_extra);
3277 		rw_enter(&icmp->icmp_rwlock, RW_WRITER);
3278 	}
3279 	return (0);
3280 }
3281 
3282 /*
3283  * This routine retrieves the value of an ND variable in a icmpparam_t
3284  * structure.  It is called through nd_getset when a user reads the
3285  * variable.
3286  */
3287 /* ARGSUSED */
3288 static int
3289 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
3290 {
3291 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3292 
3293 	(void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value);
3294 	return (0);
3295 }
3296 
3297 /*
3298  * Walk through the param array specified registering each element with the
3299  * named dispatch (ND) handler.
3300  */
3301 static boolean_t
3302 icmp_param_register(IDP *ndp, icmpparam_t *icmppa, int cnt)
3303 {
3304 	for (; cnt-- > 0; icmppa++) {
3305 		if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) {
3306 			if (!nd_load(ndp, icmppa->icmp_param_name,
3307 			    icmp_param_get, icmp_param_set,
3308 			    (caddr_t)icmppa)) {
3309 				nd_free(ndp);
3310 				return (B_FALSE);
3311 			}
3312 		}
3313 	}
3314 	return (B_TRUE);
3315 }
3316 
3317 /* This routine sets an ND variable in a icmpparam_t structure. */
3318 /* ARGSUSED */
3319 static int
3320 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
3321 {
3322 	long		new_value;
3323 	icmpparam_t	*icmppa = (icmpparam_t *)cp;
3324 
3325 	/*
3326 	 * Fail the request if the new value does not lie within the
3327 	 * required bounds.
3328 	 */
3329 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
3330 	    new_value < icmppa->icmp_param_min ||
3331 	    new_value > icmppa->icmp_param_max) {
3332 		return (EINVAL);
3333 	}
3334 	/* Set the new value */
3335 	icmppa->icmp_param_value = new_value;
3336 	return (0);
3337 }
3338 
3339 static mblk_t *
3340 icmp_queue_fallback(icmp_t *icmp, mblk_t *mp)
3341 {
3342 	ASSERT(MUTEX_HELD(&icmp->icmp_recv_lock));
3343 	if (IPCL_IS_NONSTR(icmp->icmp_connp)) {
3344 		/*
3345 		 * fallback has started but messages have not been moved yet
3346 		 */
3347 		if (icmp->icmp_fallback_queue_head == NULL) {
3348 			ASSERT(icmp->icmp_fallback_queue_tail == NULL);
3349 			icmp->icmp_fallback_queue_head = mp;
3350 			icmp->icmp_fallback_queue_tail = mp;
3351 		} else {
3352 			ASSERT(icmp->icmp_fallback_queue_tail != NULL);
3353 			icmp->icmp_fallback_queue_tail->b_next = mp;
3354 			icmp->icmp_fallback_queue_tail = mp;
3355 		}
3356 		return (NULL);
3357 	} else {
3358 		/*
3359 		 * Fallback completed, let the caller putnext() the mblk.
3360 		 */
3361 		return (mp);
3362 	}
3363 }
3364 
3365 /*
3366  * Deliver data to ULP. In case we have a socket, and it's falling back to
3367  * TPI, then we'll queue the mp for later processing.
3368  */
3369 static void
3370 icmp_ulp_recv(conn_t *connp, mblk_t *mp)
3371 {
3372 
3373 	if (IPCL_IS_NONSTR(connp)) {
3374 		icmp_t *icmp = connp->conn_icmp;
3375 		int error;
3376 
3377 		if ((*connp->conn_upcalls->su_recv)
3378 		    (connp->conn_upper_handle, mp, msgdsize(mp), 0, &error,
3379 		    NULL) < 0) {
3380 			mutex_enter(&icmp->icmp_recv_lock);
3381 			if (error == ENOSPC) {
3382 				/*
3383 				 * let's confirm while holding the lock
3384 				 */
3385 				if ((*connp->conn_upcalls->su_recv)
3386 				    (connp->conn_upper_handle, NULL, 0, 0,
3387 				    &error, NULL) < 0) {
3388 					ASSERT(error == ENOSPC);
3389 					if (error == ENOSPC) {
3390 						connp->conn_flow_cntrld =
3391 						    B_TRUE;
3392 					}
3393 				}
3394 				mutex_exit(&icmp->icmp_recv_lock);
3395 			} else {
3396 				ASSERT(error == EOPNOTSUPP);
3397 				mp = icmp_queue_fallback(icmp, mp);
3398 				mutex_exit(&icmp->icmp_recv_lock);
3399 				if (mp != NULL)
3400 					putnext(connp->conn_rq, mp);
3401 			}
3402 		}
3403 		ASSERT(MUTEX_NOT_HELD(&icmp->icmp_recv_lock));
3404 	} else {
3405 		putnext(connp->conn_rq, mp);
3406 	}
3407 }
3408 
3409 /*ARGSUSED2*/
3410 static void
3411 icmp_input(void *arg1, mblk_t *mp, void *arg2)
3412 {
3413 	conn_t *connp = (conn_t *)arg1;
3414 	struct T_unitdata_ind	*tudi;
3415 	uchar_t			*rptr;
3416 	icmp_t			*icmp;
3417 	icmp_stack_t		*is;
3418 	sin_t			*sin;
3419 	sin6_t			*sin6;
3420 	ip6_t			*ip6h;
3421 	ip6i_t			*ip6i;
3422 	mblk_t			*mp1;
3423 	int			hdr_len;
3424 	ipha_t			*ipha;
3425 	int			udi_size;	/* Size of T_unitdata_ind */
3426 	uint_t			ipvers;
3427 	ip6_pkt_t		ipp;
3428 	uint8_t			nexthdr;
3429 	ip_pktinfo_t		*pinfo = NULL;
3430 	mblk_t			*options_mp = NULL;
3431 	uint_t			icmp_opt = 0;
3432 	boolean_t		icmp_ipv6_recvhoplimit = B_FALSE;
3433 	uint_t			hopstrip;
3434 
3435 	ASSERT(connp->conn_flags & IPCL_RAWIPCONN);
3436 
3437 	icmp = connp->conn_icmp;
3438 	is = icmp->icmp_is;
3439 	rptr = mp->b_rptr;
3440 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
3441 	ASSERT(OK_32PTR(rptr));
3442 
3443 	/*
3444 	 * IP should have prepended the options data in an M_CTL
3445 	 * Check M_CTL "type" to make sure are not here bcos of
3446 	 * a valid ICMP message
3447 	 */
3448 	if (DB_TYPE(mp) == M_CTL) {
3449 		/*
3450 		 * FIXME: does IP still do this?
3451 		 * IP sends up the IPSEC_IN message for handling IPSEC
3452 		 * policy at the TCP level. We don't need it here.
3453 		 */
3454 		if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) {
3455 			mp1 = mp->b_cont;
3456 			freeb(mp);
3457 			mp = mp1;
3458 			rptr = mp->b_rptr;
3459 		} else if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
3460 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
3461 		    IN_PKTINFO) {
3462 			/*
3463 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
3464 			 * has been prepended to the packet by IP. We need to
3465 			 * extract the mblk and adjust the rptr
3466 			 */
3467 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
3468 			options_mp = mp;
3469 			mp = mp->b_cont;
3470 			rptr = mp->b_rptr;
3471 		} else {
3472 			/*
3473 			 * ICMP messages.
3474 			 */
3475 			icmp_icmp_error(connp, mp);
3476 			return;
3477 		}
3478 	}
3479 
3480 	/*
3481 	 * Discard message if it is misaligned or smaller than the IP header.
3482 	 */
3483 	if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) {
3484 		freemsg(mp);
3485 		if (options_mp != NULL)
3486 			freeb(options_mp);
3487 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3488 		return;
3489 	}
3490 	ipvers = IPH_HDR_VERSION((ipha_t *)rptr);
3491 
3492 	/* Handle M_DATA messages containing IP packets messages */
3493 	if (ipvers == IPV4_VERSION) {
3494 		/*
3495 		 * Special case where IP attaches
3496 		 * the IRE needs to be handled so that we don't send up
3497 		 * IRE to the user land.
3498 		 */
3499 		ipha = (ipha_t *)rptr;
3500 		hdr_len = IPH_HDR_LENGTH(ipha);
3501 
3502 		if (ipha->ipha_protocol == IPPROTO_TCP) {
3503 			tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3504 
3505 			if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) ==
3506 			    TH_SYN) && mp->b_cont != NULL) {
3507 				mp1 = mp->b_cont;
3508 				if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3509 					freeb(mp1);
3510 					mp->b_cont = NULL;
3511 				}
3512 			}
3513 		}
3514 		if (is->is_bsd_compat) {
3515 			ushort_t len;
3516 			len = ntohs(ipha->ipha_length);
3517 
3518 			if (mp->b_datap->db_ref > 1) {
3519 				/*
3520 				 * Allocate a new IP header so that we can
3521 				 * modify ipha_length.
3522 				 */
3523 				mblk_t	*mp1;
3524 
3525 				mp1 = allocb(hdr_len, BPRI_MED);
3526 				if (!mp1) {
3527 					freemsg(mp);
3528 					if (options_mp != NULL)
3529 						freeb(options_mp);
3530 					BUMP_MIB(&is->is_rawip_mib,
3531 					    rawipInErrors);
3532 					return;
3533 				}
3534 				bcopy(rptr, mp1->b_rptr, hdr_len);
3535 				mp->b_rptr = rptr + hdr_len;
3536 				rptr = mp1->b_rptr;
3537 				ipha = (ipha_t *)rptr;
3538 				mp1->b_cont = mp;
3539 				mp1->b_wptr = rptr + hdr_len;
3540 				mp = mp1;
3541 			}
3542 			len -= hdr_len;
3543 			ipha->ipha_length = htons(len);
3544 		}
3545 	}
3546 
3547 	/*
3548 	 * This is the inbound data path.  Packets are passed upstream as
3549 	 * T_UNITDATA_IND messages with full IP headers still attached.
3550 	 */
3551 	if (icmp->icmp_family == AF_INET) {
3552 		ASSERT(ipvers == IPV4_VERSION);
3553 		udi_size =  sizeof (struct T_unitdata_ind) + sizeof (sin_t);
3554 		if (icmp->icmp_recvif && (pinfo != NULL) &&
3555 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3556 			udi_size += sizeof (struct T_opthdr) +
3557 			    sizeof (uint_t);
3558 		}
3559 
3560 		if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3561 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3562 			udi_size += sizeof (struct T_opthdr) +
3563 			    sizeof (struct in_pktinfo);
3564 		}
3565 
3566 		/*
3567 		 * If SO_TIMESTAMP is set allocate the appropriate sized
3568 		 * buffer. Since gethrestime() expects a pointer aligned
3569 		 * argument, we allocate space necessary for extra
3570 		 * alignment (even though it might not be used).
3571 		 */
3572 		if (icmp->icmp_timestamp) {
3573 			udi_size += sizeof (struct T_opthdr) +
3574 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3575 		}
3576 		mp1 = allocb(udi_size, BPRI_MED);
3577 		if (mp1 == NULL) {
3578 			freemsg(mp);
3579 			if (options_mp != NULL)
3580 				freeb(options_mp);
3581 			BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3582 			return;
3583 		}
3584 		mp1->b_cont = mp;
3585 		mp = mp1;
3586 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
3587 		mp->b_datap->db_type = M_PROTO;
3588 		mp->b_wptr = (uchar_t *)tudi + udi_size;
3589 		tudi->PRIM_type = T_UNITDATA_IND;
3590 		tudi->SRC_length = sizeof (sin_t);
3591 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3592 		sin = (sin_t *)&tudi[1];
3593 		*sin = sin_null;
3594 		sin->sin_family = AF_INET;
3595 		sin->sin_addr.s_addr = ipha->ipha_src;
3596 		tudi->OPT_offset =  sizeof (struct T_unitdata_ind) +
3597 		    sizeof (sin_t);
3598 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
3599 		tudi->OPT_length = udi_size;
3600 
3601 		/*
3602 		 * Add options if IP_RECVIF is set
3603 		 */
3604 		if (udi_size != 0) {
3605 			char *dstopt;
3606 
3607 			dstopt = (char *)&sin[1];
3608 			if (icmp->icmp_recvif && (pinfo != NULL) &&
3609 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
3610 
3611 				struct T_opthdr *toh;
3612 				uint_t		*dstptr;
3613 
3614 				toh = (struct T_opthdr *)dstopt;
3615 				toh->level = IPPROTO_IP;
3616 				toh->name = IP_RECVIF;
3617 				toh->len = sizeof (struct T_opthdr) +
3618 				    sizeof (uint_t);
3619 				toh->status = 0;
3620 				dstopt += sizeof (struct T_opthdr);
3621 				dstptr = (uint_t *)dstopt;
3622 				*dstptr = pinfo->ip_pkt_ifindex;
3623 				dstopt += sizeof (uint_t);
3624 				udi_size -= toh->len;
3625 			}
3626 			if (icmp->icmp_timestamp) {
3627 				struct	T_opthdr *toh;
3628 
3629 				toh = (struct T_opthdr *)dstopt;
3630 				toh->level = SOL_SOCKET;
3631 				toh->name = SCM_TIMESTAMP;
3632 				toh->len = sizeof (struct T_opthdr) +
3633 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3634 				toh->status = 0;
3635 				dstopt += sizeof (struct T_opthdr);
3636 				/* Align for gethrestime() */
3637 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
3638 				    sizeof (intptr_t));
3639 				gethrestime((timestruc_t *)dstopt);
3640 				dstopt = (char *)toh + toh->len;
3641 				udi_size -= toh->len;
3642 			}
3643 			if (icmp->icmp_ip_recvpktinfo && (pinfo != NULL) &&
3644 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
3645 				struct	T_opthdr *toh;
3646 				struct	in_pktinfo *pktinfop;
3647 
3648 				toh = (struct T_opthdr *)dstopt;
3649 				toh->level = IPPROTO_IP;
3650 				toh->name = IP_PKTINFO;
3651 				toh->len = sizeof (struct T_opthdr) +
3652 				    sizeof (in_pktinfo_t);
3653 				toh->status = 0;
3654 				dstopt += sizeof (struct T_opthdr);
3655 				pktinfop = (struct in_pktinfo *)dstopt;
3656 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
3657 				pktinfop->ipi_spec_dst =
3658 				    pinfo->ip_pkt_match_addr;
3659 
3660 				pktinfop->ipi_addr.s_addr = ipha->ipha_dst;
3661 
3662 				dstopt += sizeof (struct in_pktinfo);
3663 				udi_size -= toh->len;
3664 			}
3665 
3666 			/* Consumed all of allocated space */
3667 			ASSERT(udi_size == 0);
3668 		}
3669 
3670 		if (options_mp != NULL)
3671 			freeb(options_mp);
3672 
3673 		BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
3674 		goto deliver;
3675 	}
3676 
3677 	/*
3678 	 * We don't need options_mp in the IPv6 path.
3679 	 */
3680 	if (options_mp != NULL) {
3681 		freeb(options_mp);
3682 		options_mp = NULL;
3683 	}
3684 
3685 	/*
3686 	 * Discard message if it is smaller than the IPv6 header
3687 	 * or if the header is malformed.
3688 	 */
3689 	if ((mp->b_wptr - rptr) < sizeof (ip6_t) ||
3690 	    IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION ||
3691 	    icmp->icmp_family != AF_INET6) {
3692 		freemsg(mp);
3693 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3694 		return;
3695 	}
3696 
3697 	/* Initialize */
3698 	ipp.ipp_fields = 0;
3699 	hopstrip = 0;
3700 
3701 	ip6h = (ip6_t *)rptr;
3702 	/*
3703 	 * Call on ip_find_hdr_v6 which gets the total hdr len
3704 	 * as well as individual lenghts of ext hdrs (and ptrs to
3705 	 * them).
3706 	 */
3707 	if (ip6h->ip6_nxt != icmp->icmp_proto) {
3708 		/* Look for ifindex information */
3709 		if (ip6h->ip6_nxt == IPPROTO_RAW) {
3710 			ip6i = (ip6i_t *)ip6h;
3711 			if (ip6i->ip6i_flags & IP6I_IFINDEX) {
3712 				ASSERT(ip6i->ip6i_ifindex != 0);
3713 				ipp.ipp_fields |= IPPF_IFINDEX;
3714 				ipp.ipp_ifindex = ip6i->ip6i_ifindex;
3715 			}
3716 			rptr = (uchar_t *)&ip6i[1];
3717 			mp->b_rptr = rptr;
3718 			if (rptr == mp->b_wptr) {
3719 				mp1 = mp->b_cont;
3720 				freeb(mp);
3721 				mp = mp1;
3722 				rptr = mp->b_rptr;
3723 			}
3724 			ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN);
3725 			ip6h = (ip6_t *)rptr;
3726 		}
3727 		hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr);
3728 
3729 		/*
3730 		 * We need to lie a bit to the user because users inside
3731 		 * labeled compartments should not see their own labels.  We
3732 		 * assume that in all other respects IP has checked the label,
3733 		 * and that the label is always first among the options.  (If
3734 		 * it's not first, then this code won't see it, and the option
3735 		 * will be passed along to the user.)
3736 		 *
3737 		 * If we had multilevel ICMP sockets, then the following code
3738 		 * should be skipped for them to allow the user to see the
3739 		 * label.
3740 		 *
3741 		 * Alignment restrictions in the definition of IP options
3742 		 * (namely, the requirement that the 4-octet DOI goes on a
3743 		 * 4-octet boundary) mean that we know exactly where the option
3744 		 * should start, but we're lenient for other hosts.
3745 		 *
3746 		 * Note that there are no multilevel ICMP or raw IP sockets
3747 		 * yet, thus nobody ever sees the IP6OPT_LS option.
3748 		 */
3749 		if ((ipp.ipp_fields & IPPF_HOPOPTS) &&
3750 		    ipp.ipp_hopoptslen > 5 && is_system_labeled()) {
3751 			const uchar_t *ucp =
3752 			    (const uchar_t *)ipp.ipp_hopopts + 2;
3753 			int remlen = ipp.ipp_hopoptslen - 2;
3754 
3755 			while (remlen > 0) {
3756 				if (*ucp == IP6OPT_PAD1) {
3757 					remlen--;
3758 					ucp++;
3759 				} else if (*ucp == IP6OPT_PADN) {
3760 					remlen -= ucp[1] + 2;
3761 					ucp += ucp[1] + 2;
3762 				} else if (*ucp == ip6opt_ls) {
3763 					hopstrip = (ucp -
3764 					    (const uchar_t *)ipp.ipp_hopopts) +
3765 					    ucp[1] + 2;
3766 					hopstrip = (hopstrip + 7) & ~7;
3767 					break;
3768 				} else {
3769 					/* label option must be first */
3770 					break;
3771 				}
3772 			}
3773 		}
3774 	} else {
3775 		hdr_len = IPV6_HDR_LEN;
3776 		ip6i = NULL;
3777 		nexthdr = ip6h->ip6_nxt;
3778 	}
3779 	/*
3780 	 * One special case where IP attaches the IRE needs to
3781 	 * be handled so that we don't send up IRE to the user land.
3782 	 */
3783 	if (nexthdr == IPPROTO_TCP) {
3784 		tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len];
3785 
3786 		if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) &&
3787 		    mp->b_cont != NULL) {
3788 			mp1 = mp->b_cont;
3789 			if (mp1->b_datap->db_type == IRE_DB_TYPE) {
3790 				freeb(mp1);
3791 				mp->b_cont = NULL;
3792 			}
3793 		}
3794 	}
3795 	/*
3796 	 * Check a filter for ICMPv6 types if needed.
3797 	 * Verify raw checksums if needed.
3798 	 */
3799 	if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) {
3800 		if (icmp->icmp_filter != NULL) {
3801 			int type;
3802 
3803 			/* Assumes that IP has done the pullupmsg */
3804 			type = mp->b_rptr[hdr_len];
3805 
3806 			ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr);
3807 			if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) {
3808 				freemsg(mp);
3809 				return;
3810 			}
3811 		} else {
3812 			/* Checksum */
3813 			uint16_t	*up;
3814 			uint32_t	sum;
3815 			int		remlen;
3816 
3817 			up = (uint16_t *)&ip6h->ip6_src;
3818 
3819 			remlen = msgdsize(mp) - hdr_len;
3820 			sum = htons(icmp->icmp_proto + remlen)
3821 			    + up[0] + up[1] + up[2] + up[3]
3822 			    + up[4] + up[5] + up[6] + up[7]
3823 			    + up[8] + up[9] + up[10] + up[11]
3824 			    + up[12] + up[13] + up[14] + up[15];
3825 			sum = (sum & 0xffff) + (sum >> 16);
3826 			sum = IP_CSUM(mp, hdr_len, sum);
3827 			if (sum != 0) {
3828 				/* IPv6 RAW checksum failed */
3829 				ip0dbg(("icmp_rput: RAW checksum "
3830 				    "failed %x\n", sum));
3831 				freemsg(mp);
3832 				BUMP_MIB(&is->is_rawip_mib,
3833 				    rawipInCksumErrs);
3834 				return;
3835 			}
3836 		}
3837 	}
3838 	/* Skip all the IPv6 headers per API */
3839 	mp->b_rptr += hdr_len;
3840 
3841 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3842 
3843 	/*
3844 	 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to
3845 	 * maintain state information, instead of relying on icmp_t
3846 	 * structure, since there arent any locks protecting these members
3847 	 * and there is a window where there might be a race between a
3848 	 * thread setting options on the write side and a thread reading
3849 	 * these options on the read size.
3850 	 */
3851 	if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
3852 	    IPPF_RTHDR|IPPF_IFINDEX)) {
3853 		if (icmp->icmp_ipv6_recvhopopts &&
3854 		    (ipp.ipp_fields & IPPF_HOPOPTS) &&
3855 		    ipp.ipp_hopoptslen > hopstrip) {
3856 			udi_size += sizeof (struct T_opthdr) +
3857 			    ipp.ipp_hopoptslen - hopstrip;
3858 			icmp_opt |= IPPF_HOPOPTS;
3859 		}
3860 		if ((icmp->icmp_ipv6_recvdstopts ||
3861 		    icmp->icmp_old_ipv6_recvdstopts) &&
3862 		    (ipp.ipp_fields & IPPF_DSTOPTS)) {
3863 			udi_size += sizeof (struct T_opthdr) +
3864 			    ipp.ipp_dstoptslen;
3865 			icmp_opt |= IPPF_DSTOPTS;
3866 		}
3867 		if (((icmp->icmp_ipv6_recvdstopts &&
3868 		    icmp->icmp_ipv6_recvrthdr &&
3869 		    (ipp.ipp_fields & IPPF_RTHDR)) ||
3870 		    icmp->icmp_ipv6_recvrtdstopts) &&
3871 		    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
3872 			udi_size += sizeof (struct T_opthdr) +
3873 			    ipp.ipp_rtdstoptslen;
3874 			icmp_opt |= IPPF_RTDSTOPTS;
3875 		}
3876 		if (icmp->icmp_ipv6_recvrthdr &&
3877 		    (ipp.ipp_fields & IPPF_RTHDR)) {
3878 			udi_size += sizeof (struct T_opthdr) +
3879 			    ipp.ipp_rthdrlen;
3880 			icmp_opt |= IPPF_RTHDR;
3881 		}
3882 		if (icmp->icmp_ip_recvpktinfo &&
3883 		    (ipp.ipp_fields & IPPF_IFINDEX)) {
3884 			udi_size += sizeof (struct T_opthdr) +
3885 			    sizeof (struct in6_pktinfo);
3886 			icmp_opt |= IPPF_IFINDEX;
3887 		}
3888 	}
3889 	if (icmp->icmp_ipv6_recvhoplimit) {
3890 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3891 		icmp_ipv6_recvhoplimit = B_TRUE;
3892 	}
3893 
3894 	if (icmp->icmp_ipv6_recvtclass)
3895 		udi_size += sizeof (struct T_opthdr) + sizeof (int);
3896 
3897 	/*
3898 	 * If SO_TIMESTAMP is set allocate the appropriate sized
3899 	 * buffer. Since gethrestime() expects a pointer aligned
3900 	 * argument, we allocate space necessary for extra
3901 	 * alignment (even though it might not be used).
3902 	 */
3903 	if (icmp->icmp_timestamp) {
3904 		udi_size += sizeof (struct T_opthdr) +
3905 		    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3906 	}
3907 
3908 	mp1 = allocb(udi_size, BPRI_MED);
3909 	if (mp1 == NULL) {
3910 		freemsg(mp);
3911 		BUMP_MIB(&is->is_rawip_mib, rawipInErrors);
3912 		return;
3913 	}
3914 	mp1->b_cont = mp;
3915 	mp = mp1;
3916 	mp->b_datap->db_type = M_PROTO;
3917 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
3918 	mp->b_wptr = (uchar_t *)tudi + udi_size;
3919 	tudi->PRIM_type = T_UNITDATA_IND;
3920 	tudi->SRC_length = sizeof (sin6_t);
3921 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
3922 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
3923 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
3924 	tudi->OPT_length = udi_size;
3925 	sin6 = (sin6_t *)&tudi[1];
3926 	sin6->sin6_port = 0;
3927 	sin6->sin6_family = AF_INET6;
3928 
3929 	sin6->sin6_addr = ip6h->ip6_src;
3930 	/* No sin6_flowinfo per API */
3931 	sin6->sin6_flowinfo = 0;
3932 	/* For link-scope source pass up scope id */
3933 	if ((ipp.ipp_fields & IPPF_IFINDEX) &&
3934 	    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
3935 		sin6->sin6_scope_id = ipp.ipp_ifindex;
3936 	else
3937 		sin6->sin6_scope_id = 0;
3938 
3939 	sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst,
3940 	    icmp->icmp_zoneid, is->is_netstack);
3941 
3942 	if (udi_size != 0) {
3943 		uchar_t *dstopt;
3944 
3945 		dstopt = (uchar_t *)&sin6[1];
3946 		if (icmp_opt & IPPF_IFINDEX) {
3947 			struct T_opthdr *toh;
3948 			struct in6_pktinfo *pkti;
3949 
3950 			toh = (struct T_opthdr *)dstopt;
3951 			toh->level = IPPROTO_IPV6;
3952 			toh->name = IPV6_PKTINFO;
3953 			toh->len = sizeof (struct T_opthdr) +
3954 			    sizeof (*pkti);
3955 			toh->status = 0;
3956 			dstopt += sizeof (struct T_opthdr);
3957 			pkti = (struct in6_pktinfo *)dstopt;
3958 			pkti->ipi6_addr = ip6h->ip6_dst;
3959 			pkti->ipi6_ifindex = ipp.ipp_ifindex;
3960 			dstopt += sizeof (*pkti);
3961 			udi_size -= toh->len;
3962 		}
3963 		if (icmp_ipv6_recvhoplimit) {
3964 			struct T_opthdr *toh;
3965 
3966 			toh = (struct T_opthdr *)dstopt;
3967 			toh->level = IPPROTO_IPV6;
3968 			toh->name = IPV6_HOPLIMIT;
3969 			toh->len = sizeof (struct T_opthdr) +
3970 			    sizeof (uint_t);
3971 			toh->status = 0;
3972 			dstopt += sizeof (struct T_opthdr);
3973 			*(uint_t *)dstopt = ip6h->ip6_hops;
3974 			dstopt += sizeof (uint_t);
3975 			udi_size -= toh->len;
3976 		}
3977 		if (icmp->icmp_ipv6_recvtclass) {
3978 			struct T_opthdr *toh;
3979 
3980 			toh = (struct T_opthdr *)dstopt;
3981 			toh->level = IPPROTO_IPV6;
3982 			toh->name = IPV6_TCLASS;
3983 			toh->len = sizeof (struct T_opthdr) +
3984 			    sizeof (uint_t);
3985 			toh->status = 0;
3986 			dstopt += sizeof (struct T_opthdr);
3987 			*(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow);
3988 			dstopt += sizeof (uint_t);
3989 			udi_size -= toh->len;
3990 		}
3991 		if (icmp->icmp_timestamp) {
3992 			struct  T_opthdr *toh;
3993 
3994 			toh = (struct T_opthdr *)dstopt;
3995 			toh->level = SOL_SOCKET;
3996 			toh->name = SCM_TIMESTAMP;
3997 			toh->len = sizeof (struct T_opthdr) +
3998 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
3999 			toh->status = 0;
4000 			dstopt += sizeof (struct T_opthdr);
4001 			/* Align for gethrestime() */
4002 			dstopt = (uchar_t *)P2ROUNDUP((intptr_t)dstopt,
4003 			    sizeof (intptr_t));
4004 			gethrestime((timestruc_t *)dstopt);
4005 			dstopt = (uchar_t *)toh + toh->len;
4006 			udi_size -= toh->len;
4007 		}
4008 
4009 		if (icmp_opt & IPPF_HOPOPTS) {
4010 			struct T_opthdr *toh;
4011 
4012 			toh = (struct T_opthdr *)dstopt;
4013 			toh->level = IPPROTO_IPV6;
4014 			toh->name = IPV6_HOPOPTS;
4015 			toh->len = sizeof (struct T_opthdr) +
4016 			    ipp.ipp_hopoptslen - hopstrip;
4017 			toh->status = 0;
4018 			dstopt += sizeof (struct T_opthdr);
4019 			bcopy((char *)ipp.ipp_hopopts + hopstrip, dstopt,
4020 			    ipp.ipp_hopoptslen - hopstrip);
4021 			if (hopstrip > 0) {
4022 				/* copy next header value and fake length */
4023 				dstopt[0] = ((uchar_t *)ipp.ipp_hopopts)[0];
4024 				dstopt[1] = ((uchar_t *)ipp.ipp_hopopts)[1] -
4025 				    hopstrip / 8;
4026 			}
4027 			dstopt += ipp.ipp_hopoptslen - hopstrip;
4028 			udi_size -= toh->len;
4029 		}
4030 		if (icmp_opt & IPPF_RTDSTOPTS) {
4031 			struct T_opthdr *toh;
4032 
4033 			toh = (struct T_opthdr *)dstopt;
4034 			toh->level = IPPROTO_IPV6;
4035 			toh->name = IPV6_DSTOPTS;
4036 			toh->len = sizeof (struct T_opthdr) +
4037 			    ipp.ipp_rtdstoptslen;
4038 			toh->status = 0;
4039 			dstopt += sizeof (struct T_opthdr);
4040 			bcopy(ipp.ipp_rtdstopts, dstopt,
4041 			    ipp.ipp_rtdstoptslen);
4042 			dstopt += ipp.ipp_rtdstoptslen;
4043 			udi_size -= toh->len;
4044 		}
4045 		if (icmp_opt & IPPF_RTHDR) {
4046 			struct T_opthdr *toh;
4047 
4048 			toh = (struct T_opthdr *)dstopt;
4049 			toh->level = IPPROTO_IPV6;
4050 			toh->name = IPV6_RTHDR;
4051 			toh->len = sizeof (struct T_opthdr) +
4052 			    ipp.ipp_rthdrlen;
4053 			toh->status = 0;
4054 			dstopt += sizeof (struct T_opthdr);
4055 			bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
4056 			dstopt += ipp.ipp_rthdrlen;
4057 			udi_size -= toh->len;
4058 		}
4059 		if (icmp_opt & IPPF_DSTOPTS) {
4060 			struct T_opthdr *toh;
4061 
4062 			toh = (struct T_opthdr *)dstopt;
4063 			toh->level = IPPROTO_IPV6;
4064 			toh->name = IPV6_DSTOPTS;
4065 			toh->len = sizeof (struct T_opthdr) +
4066 			    ipp.ipp_dstoptslen;
4067 			toh->status = 0;
4068 			dstopt += sizeof (struct T_opthdr);
4069 			bcopy(ipp.ipp_dstopts, dstopt,
4070 			    ipp.ipp_dstoptslen);
4071 			dstopt += ipp.ipp_dstoptslen;
4072 			udi_size -= toh->len;
4073 		}
4074 		/* Consumed all of allocated space */
4075 		ASSERT(udi_size == 0);
4076 	}
4077 	BUMP_MIB(&is->is_rawip_mib, rawipInDatagrams);
4078 
4079 deliver:
4080 	icmp_ulp_recv(connp, mp);
4081 
4082 }
4083 
4084 /*
4085  * return SNMP stuff in buffer in mpdata
4086  */
4087 mblk_t *
4088 icmp_snmp_get(queue_t *q, mblk_t *mpctl)
4089 {
4090 	mblk_t			*mpdata;
4091 	struct opthdr		*optp;
4092 	conn_t			*connp = Q_TO_CONN(q);
4093 	icmp_stack_t		*is = connp->conn_netstack->netstack_icmp;
4094 	mblk_t			*mp2ctl;
4095 
4096 	/*
4097 	 * make a copy of the original message
4098 	 */
4099 	mp2ctl = copymsg(mpctl);
4100 
4101 	if (mpctl == NULL ||
4102 	    (mpdata = mpctl->b_cont) == NULL) {
4103 		freemsg(mpctl);
4104 		freemsg(mp2ctl);
4105 		return (0);
4106 	}
4107 
4108 	/* fixed length structure for IPv4 and IPv6 counters */
4109 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
4110 	optp->level = EXPER_RAWIP;
4111 	optp->name = 0;
4112 	(void) snmp_append_data(mpdata, (char *)&is->is_rawip_mib,
4113 	    sizeof (is->is_rawip_mib));
4114 	optp->len = msgdsize(mpdata);
4115 	qreply(q, mpctl);
4116 
4117 	return (mp2ctl);
4118 }
4119 
4120 /*
4121  * Return 0 if invalid set request, 1 otherwise, including non-rawip requests.
4122  * TODO:  If this ever actually tries to set anything, it needs to be
4123  * to do the appropriate locking.
4124  */
4125 /* ARGSUSED */
4126 int
4127 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
4128     uchar_t *ptr, int len)
4129 {
4130 	switch (level) {
4131 	case EXPER_RAWIP:
4132 		return (0);
4133 	default:
4134 		return (1);
4135 	}
4136 }
4137 
4138 /*
4139  * This routine creates a T_UDERROR_IND message and passes it upstream.
4140  * The address and options are copied from the T_UNITDATA_REQ message
4141  * passed in mp.  This message is freed.
4142  */
4143 static void
4144 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err)
4145 {
4146 	mblk_t	*mp1;
4147 	uchar_t	*rptr = mp->b_rptr;
4148 	struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr;
4149 
4150 	mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset],
4151 	    tudr->DEST_length, (char *)&rptr[tudr->OPT_offset],
4152 	    tudr->OPT_length, err);
4153 	if (mp1)
4154 		qreply(q, mp1);
4155 	freemsg(mp);
4156 }
4157 
4158 
4159 static int
4160 rawip_do_unbind(conn_t *connp)
4161 {
4162 	icmp_t *icmp = connp->conn_icmp;
4163 
4164 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4165 	/* If a bind has not been done, we can't unbind. */
4166 	if (icmp->icmp_state == TS_UNBND || icmp->icmp_pending_op != -1) {
4167 		rw_exit(&icmp->icmp_rwlock);
4168 		return (-TOUTSTATE);
4169 	}
4170 	icmp->icmp_pending_op = T_UNBIND_REQ;
4171 	rw_exit(&icmp->icmp_rwlock);
4172 
4173 	/*
4174 	 * Call ip to unbind
4175 	 */
4176 
4177 	ip_unbind(connp);
4178 
4179 	/*
4180 	 * Once we're unbound from IP, the pending operation may be cleared
4181 	 * here.
4182 	 */
4183 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
4184 	V6_SET_ZERO(icmp->icmp_v6src);
4185 	V6_SET_ZERO(icmp->icmp_bound_v6src);
4186 	icmp->icmp_pending_op = -1;
4187 	icmp->icmp_state = TS_UNBND;
4188 	if (icmp->icmp_family == AF_INET6)
4189 		(void) icmp_build_hdrs(icmp);
4190 	rw_exit(&icmp->icmp_rwlock);
4191 	return (0);
4192 }
4193 
4194 /*
4195  * This routine is called by icmp_wput to handle T_UNBIND_REQ messages.
4196  * After some error checking, the message is passed downstream to ip.
4197  */
4198 static void
4199 icmp_tpi_unbind(queue_t *q, mblk_t *mp)
4200 {
4201 	conn_t	*connp = Q_TO_CONN(q);
4202 	int	error;
4203 
4204 	ASSERT(mp->b_cont == NULL);
4205 	error = rawip_do_unbind(connp);
4206 	if (error) {
4207 		if (error < 0) {
4208 			icmp_err_ack(q, mp, -error, 0);
4209 		} else {
4210 			icmp_err_ack(q, mp, 0, error);
4211 		}
4212 		return;
4213 	}
4214 
4215 	/*
4216 	 * Convert mp into a T_OK_ACK
4217 	 */
4218 
4219 	mp = mi_tpi_ok_ack_alloc(mp);
4220 
4221 	/*
4222 	 * should not happen in practice... T_OK_ACK is smaller than the
4223 	 * original message.
4224 	 */
4225 	ASSERT(mp != NULL);
4226 	ASSERT(((struct T_ok_ack *)mp->b_rptr)->PRIM_type == T_OK_ACK);
4227 	qreply(q, mp);
4228 }
4229 
4230 
4231 /*
4232  * Process IPv4 packets that already include an IP header.
4233  * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and
4234  * IPPROTO_IGMP).
4235  */
4236 static int
4237 icmp_wput_hdrincl(queue_t *q, conn_t *connp, mblk_t *mp, icmp_t *icmp,
4238     ip4_pkt_t *pktinfop)
4239 {
4240 	icmp_stack_t *is = icmp->icmp_is;
4241 	ipha_t	*ipha;
4242 	int	ip_hdr_length;
4243 	int	tp_hdr_len;
4244 	int	error;
4245 	uchar_t	ip_snd_opt[IP_MAX_OPT_LENGTH];
4246 	uint32_t ip_snd_opt_len = 0;
4247 	mblk_t	*mp1;
4248 	uint_t	pkt_len;
4249 	ip_opt_info_t optinfo;
4250 	pid_t	cpid;
4251 	cred_t	*cr;
4252 
4253 	rw_enter(&icmp->icmp_rwlock, RW_READER);
4254 
4255 	optinfo.ip_opt_flags = 0;
4256 	optinfo.ip_opt_ill_index = 0;
4257 	ipha = (ipha_t *)mp->b_rptr;
4258 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len;
4259 	if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
4260 		if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
4261 			ASSERT(icmp != NULL);
4262 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4263 			freemsg(mp);
4264 			rw_exit(&icmp->icmp_rwlock);
4265 			return (0);
4266 		}
4267 		ipha = (ipha_t *)mp->b_rptr;
4268 	}
4269 	ipha->ipha_version_and_hdr_length =
4270 	    (IP_VERSION<<4) | (ip_hdr_length>>2);
4271 
4272 	/*
4273 	 * Check if our saved options are valid; update if not.
4274 	 * TSOL Note: Since we are not in WRITER mode, ICMP packets
4275 	 * to different destination may require different labels,
4276 	 * or worse, ICMP packets to same IP address may require
4277 	 * different labels due to use of shared all-zones address.
4278 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
4279 	 * and ip_snd_options_len are consistent for the current
4280 	 * destination and are updated atomically.
4281 	 */
4282 	mutex_enter(&connp->conn_lock);
4283 	if (is_system_labeled()) {
4284 		/*
4285 		 * Recompute the Trusted Extensions security label if
4286 		 * we're not going to the same destination as last
4287 		 * time or the cred attached to the received mblk
4288 		 * changed.
4289 		 */
4290 		cr = msg_getcred(mp, &cpid);
4291 		if (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4292 		    V4_PART_OF_V6(icmp->icmp_v6lastdst) != ipha->ipha_dst ||
4293 		    cr != icmp->icmp_last_cred) {
4294 			error = icmp_update_label(icmp, mp, ipha->ipha_dst);
4295 			if (error != 0) {
4296 				mutex_exit(&connp->conn_lock);
4297 				rw_exit(&icmp->icmp_rwlock);
4298 				return (error);
4299 			}
4300 		}
4301 		/*
4302 		 * Apply credentials with modified security label if they
4303 		 * exist. icmp_update_label() may have generated these
4304 		 * credentials for packets to unlabeled remote nodes.
4305 		 */
4306 		if (icmp->icmp_effective_cred != NULL)
4307 			mblk_setcred(mp, icmp->icmp_effective_cred, cpid);
4308 	}
4309 
4310 	if (icmp->icmp_ip_snd_options_len > 0) {
4311 		ip_snd_opt_len = icmp->icmp_ip_snd_options_len;
4312 		bcopy(icmp->icmp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
4313 	}
4314 	mutex_exit(&connp->conn_lock);
4315 
4316 	/*
4317 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4318 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4319 	 * tell IP that the application has sent a complete IP header and not
4320 	 * to compute the transport checksum nor change the DF flag.
4321 	 */
4322 	ipha->ipha_ident = IP_HDR_INCLUDED;
4323 	ipha->ipha_hdr_checksum = 0;
4324 	ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF);
4325 	/* Insert options if any */
4326 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4327 		/*
4328 		 * Put the IP header plus any transport header that is
4329 		 * checksumed by ip_wput into the first mblk. (ip_wput assumes
4330 		 * that at least the checksum field is in the first mblk.)
4331 		 */
4332 		switch (ipha->ipha_protocol) {
4333 		case IPPROTO_UDP:
4334 			tp_hdr_len = 8;
4335 			break;
4336 		case IPPROTO_TCP:
4337 			tp_hdr_len = 20;
4338 			break;
4339 		default:
4340 			tp_hdr_len = 0;
4341 			break;
4342 		}
4343 		/*
4344 		 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus
4345 		 * tp_hdr_len bytes will be in a single mblk.
4346 		 */
4347 		if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH +
4348 		    tp_hdr_len)) {
4349 			if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH +
4350 			    tp_hdr_len)) {
4351 				BUMP_MIB(&is->is_rawip_mib,
4352 				    rawipOutErrors);
4353 				freemsg(mp);
4354 				rw_exit(&icmp->icmp_rwlock);
4355 				return (0);
4356 			}
4357 			ipha = (ipha_t *)mp->b_rptr;
4358 		}
4359 
4360 		/*
4361 		 * if the length is larger then the max allowed IP packet,
4362 		 * then send an error and abort the processing.
4363 		 */
4364 		pkt_len = ntohs(ipha->ipha_length)
4365 		    + ip_snd_opt_len;
4366 		if (pkt_len > IP_MAXPACKET) {
4367 			rw_exit(&icmp->icmp_rwlock);
4368 			return (EMSGSIZE);
4369 		}
4370 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra +
4371 		    tp_hdr_len, BPRI_LO))) {
4372 			rw_exit(&icmp->icmp_rwlock);
4373 			return (ENOMEM);
4374 		}
4375 		mp1->b_rptr += is->is_wroff_extra;
4376 		mp1->b_wptr = mp1->b_rptr + ip_hdr_length;
4377 
4378 		ipha->ipha_length = htons((uint16_t)pkt_len);
4379 		bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH);
4380 
4381 		/* Copy transport header if any */
4382 		bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len);
4383 		mp1->b_wptr += tp_hdr_len;
4384 
4385 		/* Add options */
4386 		ipha = (ipha_t *)mp1->b_rptr;
4387 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
4388 
4389 		/* Drop IP header and transport header from original */
4390 		(void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len);
4391 
4392 		mp1->b_cont = mp;
4393 		mp = mp1;
4394 		/*
4395 		 * Massage source route putting first source
4396 		 * route in ipha_dst.
4397 		 */
4398 		(void) ip_massage_options(ipha, is->is_netstack);
4399 	}
4400 
4401 	if (pktinfop != NULL) {
4402 		/*
4403 		 * Over write the source address provided in the header
4404 		 */
4405 		if (pktinfop->ip4_addr != INADDR_ANY) {
4406 			ipha->ipha_src = pktinfop->ip4_addr;
4407 			optinfo.ip_opt_flags = IP_VERIFY_SRC;
4408 		}
4409 
4410 		if (pktinfop->ip4_ill_index != 0) {
4411 			optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4412 		}
4413 	}
4414 
4415 	rw_exit(&icmp->icmp_rwlock);
4416 
4417 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4418 	return (0);
4419 }
4420 
4421 static int
4422 icmp_update_label(icmp_t *icmp, mblk_t *mp, ipaddr_t dst)
4423 {
4424 	int err;
4425 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
4426 	icmp_stack_t		*is = icmp->icmp_is;
4427 	conn_t			*connp = icmp->icmp_connp;
4428 	cred_t	*cred;
4429 	cred_t	*msg_cred;
4430 	cred_t	*effective_cred;
4431 
4432 	/*
4433 	 * All Solaris components should pass a db_credp
4434 	 * for this message, hence we ASSERT.
4435 	 * On production kernels we return an error to be robust against
4436 	 * random streams modules sitting on top of us.
4437 	 */
4438 	cred = msg_cred = msg_getcred(mp, NULL);
4439 	ASSERT(cred != NULL);
4440 	if (cred == NULL)
4441 		return (EINVAL);
4442 
4443 	/*
4444 	 * Verify the destination is allowed to receive packets at
4445 	 * the security label of the message data. check_dest()
4446 	 * may create a new effective cred for this message
4447 	 * with a modified label or label flags.
4448 	 */
4449 	if ((err = tsol_check_dest(cred, &dst, IPV4_VERSION,
4450 	    connp->conn_mac_exempt, &effective_cred)) != 0)
4451 		goto done;
4452 	if (effective_cred != NULL)
4453 		cred = effective_cred;
4454 
4455 	/*
4456 	 * Calculate the security label to be placed in the text
4457 	 * of the message (if any).
4458 	 */
4459 	if ((err = tsol_compute_label(cred, dst, opt_storage,
4460 	    is->is_netstack->netstack_ip)) != 0)
4461 		goto done;
4462 
4463 	/*
4464 	 * Insert the security label in the cached ip options,
4465 	 * removing any old label that may exist.
4466 	 */
4467 	if ((err = tsol_update_options(&icmp->icmp_ip_snd_options,
4468 	    &icmp->icmp_ip_snd_options_len, &icmp->icmp_label_len,
4469 	    opt_storage)) != 0)
4470 		goto done;
4471 
4472 	/*
4473 	 * Save the destination address and cred we used to generate
4474 	 * the security label text.
4475 	 */
4476 	IN6_IPADDR_TO_V4MAPPED(dst, &icmp->icmp_v6lastdst);
4477 	if (cred != icmp->icmp_effective_cred) {
4478 		if (icmp->icmp_effective_cred != NULL)
4479 			crfree(icmp->icmp_effective_cred);
4480 		crhold(cred);
4481 		icmp->icmp_effective_cred = cred;
4482 	}
4483 
4484 	if (msg_cred != icmp->icmp_last_cred) {
4485 		if (icmp->icmp_last_cred != NULL)
4486 			crfree(icmp->icmp_last_cred);
4487 		crhold(msg_cred);
4488 		icmp->icmp_last_cred = msg_cred;
4489 	}
4490 
4491 done:
4492 	if (effective_cred != NULL)
4493 		crfree(effective_cred);
4494 
4495 	if (err != 0) {
4496 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4497 		DTRACE_PROBE4(
4498 		    tx__ip__log__drop__updatelabel__icmp,
4499 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4500 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4501 		return (err);
4502 	}
4503 	return (0);
4504 }
4505 
4506 /*
4507  * This routine handles all messages passed downstream.  It either
4508  * consumes the message or passes it downstream; it never queues a
4509  * a message.
4510  */
4511 static void
4512 icmp_wput(queue_t *q, mblk_t *mp)
4513 {
4514 	uchar_t	*rptr = mp->b_rptr;
4515 	mblk_t	*mp1;
4516 #define	tudr ((struct T_unitdata_req *)rptr)
4517 	size_t	ip_len;
4518 	conn_t	*connp = Q_TO_CONN(q);
4519 	icmp_t	*icmp = connp->conn_icmp;
4520 	icmp_stack_t *is = icmp->icmp_is;
4521 	sin6_t	*sin6;
4522 	sin_t	*sin;
4523 	ipaddr_t	v4dst;
4524 	ip4_pkt_t	pktinfo;
4525 	ip4_pkt_t	*pktinfop = &pktinfo;
4526 	ip6_pkt_t	ipp_s;  /* For ancillary data options */
4527 	ip6_pkt_t	*ipp = &ipp_s;
4528 	int error;
4529 
4530 	ipp->ipp_fields = 0;
4531 	ipp->ipp_sticky_ignored = 0;
4532 
4533 	switch (mp->b_datap->db_type) {
4534 	case M_DATA:
4535 		if (icmp->icmp_hdrincl) {
4536 			ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
4537 			error = icmp_wput_hdrincl(q, connp, mp, icmp, NULL);
4538 			if (error != 0)
4539 				icmp_ud_err(q, mp, error);
4540 			return;
4541 		}
4542 		freemsg(mp);
4543 		return;
4544 	case M_PROTO:
4545 	case M_PCPROTO:
4546 		ip_len = mp->b_wptr - rptr;
4547 		if (ip_len >= sizeof (struct T_unitdata_req)) {
4548 			/* Expedite valid T_UNITDATA_REQ to below the switch */
4549 			if (((union T_primitives *)rptr)->type
4550 			    == T_UNITDATA_REQ)
4551 				break;
4552 		}
4553 		/* FALLTHRU */
4554 	default:
4555 		icmp_wput_other(q, mp);
4556 		return;
4557 	}
4558 
4559 	/* Handle T_UNITDATA_REQ messages here. */
4560 
4561 	mp1 = mp->b_cont;
4562 	if (mp1 == NULL) {
4563 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4564 		icmp_ud_err(q, mp, EPROTO);
4565 		return;
4566 	}
4567 
4568 	if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) {
4569 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4570 		icmp_ud_err(q, mp, EADDRNOTAVAIL);
4571 		return;
4572 	}
4573 
4574 	switch (icmp->icmp_family) {
4575 	case AF_INET6:
4576 		sin6 = (sin6_t *)&rptr[tudr->DEST_offset];
4577 		if (!OK_32PTR((char *)sin6) ||
4578 		    tudr->DEST_length != sizeof (sin6_t) ||
4579 		    sin6->sin6_family != AF_INET6) {
4580 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4581 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4582 			return;
4583 		}
4584 
4585 		/* No support for mapped addresses on raw sockets */
4586 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
4587 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4588 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4589 			return;
4590 		}
4591 
4592 		/*
4593 		 * Destination is a native IPv6 address.
4594 		 * Send out an IPv6 format packet.
4595 		 */
4596 		if (tudr->OPT_length != 0) {
4597 			int error;
4598 
4599 			error = 0;
4600 			if (icmp_unitdata_opt_process(q, mp, &error,
4601 			    (void *)ipp) < 0) {
4602 				/* failure */
4603 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4604 				icmp_ud_err(q, mp, error);
4605 				return;
4606 			}
4607 			ASSERT(error == 0);
4608 		}
4609 
4610 		error = raw_ip_send_data_v6(q, connp, mp1, sin6, ipp);
4611 		goto done;
4612 
4613 	case AF_INET:
4614 		sin = (sin_t *)&rptr[tudr->DEST_offset];
4615 		if (!OK_32PTR((char *)sin) ||
4616 		    tudr->DEST_length != sizeof (sin_t) ||
4617 		    sin->sin_family != AF_INET) {
4618 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4619 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
4620 			return;
4621 		}
4622 		/* Extract and ipaddr */
4623 		v4dst = sin->sin_addr.s_addr;
4624 		break;
4625 
4626 	default:
4627 		ASSERT(0);
4628 	}
4629 
4630 	pktinfop->ip4_ill_index = 0;
4631 	pktinfop->ip4_addr = INADDR_ANY;
4632 
4633 	/*
4634 	 * If options passed in, feed it for verification and handling
4635 	 */
4636 	if (tudr->OPT_length != 0) {
4637 		int error;
4638 
4639 		error = 0;
4640 		if (icmp_unitdata_opt_process(q, mp, &error,
4641 		    (void *)pktinfop) < 0) {
4642 			/* failure */
4643 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4644 			icmp_ud_err(q, mp, error);
4645 			return;
4646 		}
4647 		ASSERT(error == 0);
4648 		/*
4649 		 * Note: Success in processing options.
4650 		 * mp option buffer represented by
4651 		 * OPT_length/offset now potentially modified
4652 		 * and contain option setting results
4653 		 */
4654 	}
4655 
4656 	error = raw_ip_send_data_v4(q, connp, mp1, v4dst, pktinfop);
4657 done:
4658 	if (error != 0) {
4659 		icmp_ud_err(q, mp, error);
4660 		return;
4661 	} else {
4662 		mp->b_cont = NULL;
4663 		freeb(mp);
4664 	}
4665 }
4666 
4667 
4668 /* ARGSUSED */
4669 static void
4670 icmp_wput_fallback(queue_t *q, mblk_t *mp)
4671 {
4672 #ifdef DEBUG
4673 	cmn_err(CE_CONT, "icmp_wput_fallback: Message during fallback \n");
4674 #endif
4675 	freemsg(mp);
4676 }
4677 
4678 static int
4679 raw_ip_send_data_v4(queue_t *q, conn_t *connp, mblk_t *mp, ipaddr_t v4dst,
4680     ip4_pkt_t *pktinfop)
4681 {
4682 	ipha_t	*ipha;
4683 	size_t	ip_len;
4684 	icmp_t	*icmp = connp->conn_icmp;
4685 	icmp_stack_t *is = icmp->icmp_is;
4686 	int	ip_hdr_length;
4687 	ip_opt_info_t	optinfo;
4688 	uchar_t	ip_snd_opt[IP_MAX_OPT_LENGTH];
4689 	uint32_t ip_snd_opt_len = 0;
4690 	pid_t	cpid;
4691 	cred_t	*cr;
4692 
4693 	optinfo.ip_opt_flags = 0;
4694 	optinfo.ip_opt_ill_index = 0;
4695 
4696 	if (icmp->icmp_state == TS_UNBND) {
4697 		/* If a port has not been bound to the stream, fail. */
4698 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4699 		return (EPROTO);
4700 	}
4701 
4702 	if (v4dst == INADDR_ANY)
4703 		v4dst = htonl(INADDR_LOOPBACK);
4704 
4705 	/* Protocol 255 contains full IP headers */
4706 	if (icmp->icmp_hdrincl)
4707 		return (icmp_wput_hdrincl(q, connp, mp, icmp, pktinfop));
4708 
4709 	rw_enter(&icmp->icmp_rwlock, RW_READER);
4710 
4711 	/*
4712 	 * Check if our saved options are valid; update if not.
4713 	 * TSOL Note: Since we are not in WRITER mode, ICMP packets
4714 	 * to different destination may require different labels,
4715 	 * or worse, ICMP packets to same IP address may require
4716 	 * different labels due to use of shared all-zones address.
4717 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
4718 	 * and ip_snd_options_len are consistent for the current
4719 	 * destination and are updated atomically.
4720 	 */
4721 	mutex_enter(&connp->conn_lock);
4722 	if (is_system_labeled()) {
4723 
4724 		/*
4725 		 * Recompute the Trusted Extensions security label if we're not
4726 		 * going to the same destination as last time or the cred
4727 		 * attached to the received mblk changed.
4728 		 */
4729 		cr = msg_getcred(mp, &cpid);
4730 		if (!IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6lastdst) ||
4731 		    V4_PART_OF_V6(icmp->icmp_v6lastdst) != v4dst ||
4732 		    cr != icmp->icmp_last_cred) {
4733 			int error = icmp_update_label(icmp, mp, v4dst);
4734 			if (error != 0) {
4735 				mutex_exit(&connp->conn_lock);
4736 				rw_exit(&icmp->icmp_rwlock);
4737 				return (error);
4738 			}
4739 		}
4740 		/*
4741 		 * Apply credentials with modified security label if they
4742 		 * exist. icmp_update_label() may have generated these
4743 		 * credentials for packets to unlabeled remote nodes.
4744 		 */
4745 		if (icmp->icmp_effective_cred != NULL)
4746 			mblk_setcred(mp, icmp->icmp_effective_cred, cpid);
4747 	}
4748 
4749 	if (icmp->icmp_ip_snd_options_len > 0) {
4750 		ip_snd_opt_len = icmp->icmp_ip_snd_options_len;
4751 		bcopy(icmp->icmp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
4752 	}
4753 	mutex_exit(&connp->conn_lock);
4754 
4755 	/* Add an IP header */
4756 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + ip_snd_opt_len;
4757 	ipha = (ipha_t *)&mp->b_rptr[-ip_hdr_length];
4758 	if ((uchar_t *)ipha < mp->b_datap->db_base ||
4759 	    mp->b_datap->db_ref != 1 ||
4760 	    !OK_32PTR(ipha)) {
4761 		mblk_t	*mp1;
4762 		if (!(mp1 = allocb(ip_hdr_length + is->is_wroff_extra,
4763 		    BPRI_LO))) {
4764 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4765 			rw_exit(&icmp->icmp_rwlock);
4766 			return (ENOMEM);
4767 		}
4768 		mp1->b_cont = mp;
4769 		ipha = (ipha_t *)mp1->b_datap->db_lim;
4770 		mp1->b_wptr = (uchar_t *)ipha;
4771 		ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length);
4772 		mp = mp1;
4773 	}
4774 #ifdef	_BIG_ENDIAN
4775 	/* Set version, header length, and tos */
4776 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4777 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
4778 	    icmp->icmp_type_of_service);
4779 	/* Set ttl and protocol */
4780 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto;
4781 #else
4782 	/* Set version, header length, and tos */
4783 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
4784 	    ((icmp->icmp_type_of_service << 8) |
4785 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
4786 	/* Set ttl and protocol */
4787 	*(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl;
4788 #endif
4789 	if (pktinfop->ip4_addr != INADDR_ANY) {
4790 		ipha->ipha_src = pktinfop->ip4_addr;
4791 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
4792 	} else {
4793 
4794 		/*
4795 		 * Copy our address into the packet.  If this is zero,
4796 		 * ip will fill in the real source address.
4797 		 */
4798 		IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src);
4799 	}
4800 
4801 	ipha->ipha_fragment_offset_and_flags = 0;
4802 
4803 	if (pktinfop->ip4_ill_index != 0) {
4804 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
4805 	}
4806 
4807 
4808 	/*
4809 	 * For the socket of SOCK_RAW type, the checksum is provided in the
4810 	 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to
4811 	 * tell IP that the application has sent a complete IP header and not
4812 	 * to compute the transport checksum nor change the DF flag.
4813 	 */
4814 	ipha->ipha_ident = IP_HDR_INCLUDED;
4815 
4816 	/* Finish common formatting of the packet. */
4817 	mp->b_rptr = (uchar_t *)ipha;
4818 
4819 	ip_len = mp->b_wptr - (uchar_t *)ipha;
4820 	if (mp->b_cont != NULL)
4821 		ip_len += msgdsize(mp->b_cont);
4822 
4823 	/*
4824 	 * Set the length into the IP header.
4825 	 * If the length is greater than the maximum allowed by IP,
4826 	 * then free the message and return. Do not try and send it
4827 	 * as this can cause problems in layers below.
4828 	 */
4829 	if (ip_len > IP_MAXPACKET) {
4830 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4831 		rw_exit(&icmp->icmp_rwlock);
4832 		return (EMSGSIZE);
4833 	}
4834 	ipha->ipha_length = htons((uint16_t)ip_len);
4835 	/*
4836 	 * Copy in the destination address request
4837 	 */
4838 	ipha->ipha_dst = v4dst;
4839 
4840 	/*
4841 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
4842 	 */
4843 	if (CLASSD(v4dst))
4844 		ipha->ipha_ttl = icmp->icmp_multicast_ttl;
4845 
4846 	/* Copy in options if any */
4847 	if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) {
4848 		bcopy(ip_snd_opt,
4849 		    &ipha[1], ip_snd_opt_len);
4850 		/*
4851 		 * Massage source route putting first source route in ipha_dst.
4852 		 * Ignore the destination in the T_unitdata_req.
4853 		 */
4854 		(void) ip_massage_options(ipha, is->is_netstack);
4855 	}
4856 
4857 	rw_exit(&icmp->icmp_rwlock);
4858 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
4859 
4860 	ip_output_options(connp, mp, q, IP_WPUT, &optinfo);
4861 	return (0);
4862 }
4863 
4864 static int
4865 icmp_update_label_v6(icmp_t *icmp, mblk_t *mp, in6_addr_t *dst)
4866 {
4867 	int err;
4868 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
4869 	icmp_stack_t		*is = icmp->icmp_is;
4870 	conn_t			*connp = icmp->icmp_connp;
4871 	cred_t	*cred;
4872 	cred_t	*msg_cred;
4873 	cred_t	*effective_cred;
4874 
4875 	/*
4876 	 * All Solaris components should pass a db_credp
4877 	 * for this message, hence we ASSERT.
4878 	 * On production kernels we return an error to be robust against
4879 	 * random streams modules sitting on top of us.
4880 	 */
4881 	cred = msg_cred = msg_getcred(mp, NULL);
4882 	ASSERT(cred != NULL);
4883 	if (cred == NULL)
4884 		return (EINVAL);
4885 
4886 	/*
4887 	 * Verify the destination is allowed to receive packets at
4888 	 * the security label of the message data. check_dest()
4889 	 * may create a new effective cred for this message
4890 	 * with a modified label or label flags.
4891 	 */
4892 	if ((err = tsol_check_dest(cred, dst, IPV6_VERSION,
4893 	    connp->conn_mac_exempt, &effective_cred)) != 0)
4894 		goto done;
4895 	if (effective_cred != NULL)
4896 		cred = effective_cred;
4897 
4898 	/*
4899 	 * Calculate the security label to be placed in the text
4900 	 * of the message (if any).
4901 	 */
4902 	if ((err = tsol_compute_label_v6(cred, dst, opt_storage,
4903 	    is->is_netstack->netstack_ip)) != 0)
4904 		goto done;
4905 
4906 	/*
4907 	 * Insert the security label in the cached ip options,
4908 	 * removing any old label that may exist.
4909 	 */
4910 	if ((err = tsol_update_sticky(&icmp->icmp_sticky_ipp,
4911 	    &icmp->icmp_label_len_v6, opt_storage)) != 0)
4912 		goto done;
4913 
4914 	/*
4915 	 * Save the destination address and cred we used to generate
4916 	 * the security label text.
4917 	 */
4918 	icmp->icmp_v6lastdst = *dst;
4919 	if (cred != icmp->icmp_effective_cred) {
4920 		if (icmp->icmp_effective_cred != NULL)
4921 			crfree(icmp->icmp_effective_cred);
4922 		crhold(cred);
4923 		icmp->icmp_effective_cred = cred;
4924 	}
4925 
4926 	if (msg_cred != icmp->icmp_last_cred) {
4927 		if (icmp->icmp_last_cred != NULL)
4928 			crfree(icmp->icmp_last_cred);
4929 		crhold(msg_cred);
4930 		icmp->icmp_last_cred = msg_cred;
4931 	}
4932 
4933 done:
4934 	if (effective_cred != NULL)
4935 		crfree(effective_cred);
4936 
4937 	if (err != 0) {
4938 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4939 		DTRACE_PROBE4(
4940 		    tx__ip__log__drop__updatelabel__icmp6,
4941 		    char *, "icmp(1) failed to update options(2) on mp(3)",
4942 		    icmp_t *, icmp, char *, opt_storage, mblk_t *, mp);
4943 		return (err);
4944 	}
4945 	return (0);
4946 }
4947 
4948 /*
4949  * raw_ip_send_data_v6():
4950  * Assumes that icmp_wput did some sanity checking on the destination
4951  * address, but that the label may not yet be correct.
4952  */
4953 static int
4954 raw_ip_send_data_v6(queue_t *q, conn_t *connp, mblk_t *mp, sin6_t *sin6,
4955     ip6_pkt_t *ipp)
4956 {
4957 	ip6_t			*ip6h;
4958 	ip6i_t			*ip6i;	/* mp->b_rptr even if no ip6i_t */
4959 	int			ip_hdr_len = IPV6_HDR_LEN;
4960 	size_t			ip_len;
4961 	icmp_t			*icmp = connp->conn_icmp;
4962 	icmp_stack_t		*is = icmp->icmp_is;
4963 	ip6_pkt_t		*tipp;
4964 	ip6_hbh_t		*hopoptsptr = NULL;
4965 	uint_t			hopoptslen = 0;
4966 	uint32_t		csum = 0;
4967 	uint_t			ignore = 0;
4968 	uint_t			option_exists = 0, is_sticky = 0;
4969 	uint8_t			*cp;
4970 	uint8_t			*nxthdr_ptr;
4971 	in6_addr_t		ip6_dst;
4972 	pid_t			cpid;
4973 	cred_t			*cr;
4974 
4975 	rw_enter(&icmp->icmp_rwlock, RW_READER);
4976 
4977 	/*
4978 	 * If the local address is a mapped address return
4979 	 * an error.
4980 	 * It would be possible to send an IPv6 packet but the
4981 	 * response would never make it back to the application
4982 	 * since it is bound to a mapped address.
4983 	 */
4984 	if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) {
4985 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
4986 		rw_exit(&icmp->icmp_rwlock);
4987 		return (EADDRNOTAVAIL);
4988 	}
4989 
4990 	ignore = ipp->ipp_sticky_ignored;
4991 	if (sin6->sin6_scope_id != 0 &&
4992 	    IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
4993 		/*
4994 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
4995 		 * option nor ancillary data.  It needs to be
4996 		 * explicitly set in options_exists.
4997 		 */
4998 		option_exists |= IPPF_SCOPE_ID;
4999 	}
5000 
5001 	/*
5002 	 * Compute the destination address
5003 	 */
5004 	ip6_dst = sin6->sin6_addr;
5005 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
5006 		ip6_dst = ipv6_loopback;
5007 
5008 	/*
5009 	 * Check if our saved options are valid; update if not.
5010 	 * TSOL Note: Since we are not in WRITER mode, ICMP packets
5011 	 * to different destination may require different labels,
5012 	 * or worse, ICMP packets to same IP address may require
5013 	 * different labels due to use of shared all-zones address.
5014 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
5015 	 * and sticky ipp_hopoptslen are consistent for the current
5016 	 * destination and are updated atomically.
5017 	 */
5018 	mutex_enter(&connp->conn_lock);
5019 	if (is_system_labeled()) {
5020 		/*
5021 		 * Recompute the Trusted Extensions security label if we're
5022 		 * not going to the same destination as last time or the cred
5023 		 * attached to the received mblk changed. This is done in a
5024 		 * separate routine to avoid blowing up our stack here.
5025 		 */
5026 		cr = msg_getcred(mp, &cpid);
5027 		if (!IN6_ARE_ADDR_EQUAL(&icmp->icmp_v6lastdst, &ip6_dst) ||
5028 		    cr != icmp->icmp_last_cred) {
5029 			int error = 0;
5030 			error = icmp_update_label_v6(icmp, mp, &ip6_dst);
5031 			if (error != 0) {
5032 				mutex_exit(&connp->conn_lock);
5033 				rw_exit(&icmp->icmp_rwlock);
5034 				return (error);
5035 			}
5036 		}
5037 
5038 		/*
5039 		 * Apply credentials with modified security label if they exist.
5040 		 * icmp_update_label_v6() may have generated these credentials
5041 		 * for MAC-Exempt connections.
5042 		 */
5043 		if (icmp->icmp_effective_cred != NULL)
5044 			mblk_setcred(mp, icmp->icmp_effective_cred, cpid);
5045 	}
5046 
5047 	/*
5048 	 * If there's a security label here, then we ignore any options the
5049 	 * user may try to set.  We keep the peer's label as a hidden sticky
5050 	 * option.
5051 	 */
5052 	if (icmp->icmp_label_len_v6 > 0) {
5053 		ignore &= ~IPPF_HOPOPTS;
5054 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
5055 	}
5056 
5057 	if ((icmp->icmp_sticky_ipp.ipp_fields == 0) &&
5058 	    (ipp->ipp_fields == 0)) {
5059 		/* No sticky options nor ancillary data. */
5060 		mutex_exit(&connp->conn_lock);
5061 		goto no_options;
5062 	}
5063 
5064 	/*
5065 	 * Go through the options figuring out where each is going to
5066 	 * come from and build two masks.  The first mask indicates if
5067 	 * the option exists at all.  The second mask indicates if the
5068 	 * option is sticky or ancillary.
5069 	 */
5070 	if (!(ignore & IPPF_HOPOPTS)) {
5071 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
5072 			option_exists |= IPPF_HOPOPTS;
5073 			ip_hdr_len += ipp->ipp_hopoptslen;
5074 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
5075 			option_exists |= IPPF_HOPOPTS;
5076 			is_sticky |= IPPF_HOPOPTS;
5077 			ASSERT(icmp->icmp_sticky_ipp.ipp_hopoptslen != 0);
5078 			hopoptsptr = kmem_alloc(
5079 			    icmp->icmp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
5080 			if (hopoptsptr == NULL) {
5081 				mutex_exit(&connp->conn_lock);
5082 				rw_exit(&icmp->icmp_rwlock);
5083 				return (ENOMEM);
5084 			}
5085 			hopoptslen = icmp->icmp_sticky_ipp.ipp_hopoptslen;
5086 			bcopy(icmp->icmp_sticky_ipp.ipp_hopopts, hopoptsptr,
5087 			    hopoptslen);
5088 			ip_hdr_len += hopoptslen;
5089 		}
5090 	}
5091 	mutex_exit(&connp->conn_lock);
5092 
5093 	if (!(ignore & IPPF_RTHDR)) {
5094 		if (ipp->ipp_fields & IPPF_RTHDR) {
5095 			option_exists |= IPPF_RTHDR;
5096 			ip_hdr_len += ipp->ipp_rthdrlen;
5097 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
5098 			option_exists |= IPPF_RTHDR;
5099 			is_sticky |= IPPF_RTHDR;
5100 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen;
5101 		}
5102 	}
5103 
5104 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
5105 		/*
5106 		 * Need to have a router header to use these.
5107 		 */
5108 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
5109 			option_exists |= IPPF_RTDSTOPTS;
5110 			ip_hdr_len += ipp->ipp_rtdstoptslen;
5111 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
5112 			option_exists |= IPPF_RTDSTOPTS;
5113 			is_sticky |= IPPF_RTDSTOPTS;
5114 			ip_hdr_len +=
5115 			    icmp->icmp_sticky_ipp.ipp_rtdstoptslen;
5116 		}
5117 	}
5118 
5119 	if (!(ignore & IPPF_DSTOPTS)) {
5120 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
5121 			option_exists |= IPPF_DSTOPTS;
5122 			ip_hdr_len += ipp->ipp_dstoptslen;
5123 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
5124 			option_exists |= IPPF_DSTOPTS;
5125 			is_sticky |= IPPF_DSTOPTS;
5126 			ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen;
5127 		}
5128 	}
5129 
5130 	if (!(ignore & IPPF_IFINDEX)) {
5131 		if (ipp->ipp_fields & IPPF_IFINDEX) {
5132 			option_exists |= IPPF_IFINDEX;
5133 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
5134 			option_exists |= IPPF_IFINDEX;
5135 			is_sticky |= IPPF_IFINDEX;
5136 		}
5137 	}
5138 
5139 	if (!(ignore & IPPF_ADDR)) {
5140 		if (ipp->ipp_fields & IPPF_ADDR) {
5141 			option_exists |= IPPF_ADDR;
5142 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) {
5143 			option_exists |= IPPF_ADDR;
5144 			is_sticky |= IPPF_ADDR;
5145 		}
5146 	}
5147 
5148 	if (!(ignore & IPPF_DONTFRAG)) {
5149 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
5150 			option_exists |= IPPF_DONTFRAG;
5151 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
5152 			option_exists |= IPPF_DONTFRAG;
5153 			is_sticky |= IPPF_DONTFRAG;
5154 		}
5155 	}
5156 
5157 	if (!(ignore & IPPF_USE_MIN_MTU)) {
5158 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
5159 			option_exists |= IPPF_USE_MIN_MTU;
5160 		} else if (icmp->icmp_sticky_ipp.ipp_fields &
5161 		    IPPF_USE_MIN_MTU) {
5162 			option_exists |= IPPF_USE_MIN_MTU;
5163 			is_sticky |= IPPF_USE_MIN_MTU;
5164 		}
5165 	}
5166 
5167 	if (!(ignore & IPPF_NEXTHOP)) {
5168 		if (ipp->ipp_fields & IPPF_NEXTHOP) {
5169 			option_exists |= IPPF_NEXTHOP;
5170 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) {
5171 			option_exists |= IPPF_NEXTHOP;
5172 			is_sticky |= IPPF_NEXTHOP;
5173 		}
5174 	}
5175 
5176 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
5177 		option_exists |= IPPF_HOPLIMIT;
5178 	/* IPV6_HOPLIMIT can never be sticky */
5179 	ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
5180 
5181 	if (!(ignore & IPPF_UNICAST_HOPS) &&
5182 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
5183 		option_exists |= IPPF_UNICAST_HOPS;
5184 		is_sticky |= IPPF_UNICAST_HOPS;
5185 	}
5186 
5187 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
5188 	    (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
5189 		option_exists |= IPPF_MULTICAST_HOPS;
5190 		is_sticky |= IPPF_MULTICAST_HOPS;
5191 	}
5192 
5193 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) {
5194 		/* This is a sticky socket option only */
5195 		option_exists |= IPPF_NO_CKSUM;
5196 		is_sticky |= IPPF_NO_CKSUM;
5197 	}
5198 
5199 	if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) {
5200 		/* This is a sticky socket option only */
5201 		option_exists |= IPPF_RAW_CKSUM;
5202 		is_sticky |= IPPF_RAW_CKSUM;
5203 	}
5204 
5205 	if (!(ignore & IPPF_TCLASS)) {
5206 		if (ipp->ipp_fields & IPPF_TCLASS) {
5207 			option_exists |= IPPF_TCLASS;
5208 		} else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
5209 			option_exists |= IPPF_TCLASS;
5210 			is_sticky |= IPPF_TCLASS;
5211 		}
5212 	}
5213 
5214 no_options:
5215 
5216 	/*
5217 	 * If any options carried in the ip6i_t were specified, we
5218 	 * need to account for the ip6i_t in the data we'll be sending
5219 	 * down.
5220 	 */
5221 	if (option_exists & IPPF_HAS_IP6I)
5222 		ip_hdr_len += sizeof (ip6i_t);
5223 
5224 	/* check/fix buffer config, setup pointers into it */
5225 	ip6h = (ip6_t *)&mp->b_rptr[-ip_hdr_len];
5226 	if ((mp->b_datap->db_ref != 1) ||
5227 	    ((unsigned char *)ip6h < mp->b_datap->db_base) ||
5228 	    !OK_32PTR(ip6h)) {
5229 		mblk_t	*mp1;
5230 
5231 		/* Try to get everything in a single mblk next time */
5232 		if (ip_hdr_len > icmp->icmp_max_hdr_len) {
5233 			icmp->icmp_max_hdr_len = ip_hdr_len;
5234 
5235 			(void) proto_set_tx_wroff(q == NULL ? NULL:RD(q), connp,
5236 			    icmp->icmp_max_hdr_len + is->is_wroff_extra);
5237 		}
5238 		mp1 = allocb(ip_hdr_len + is->is_wroff_extra, BPRI_LO);
5239 		if (!mp1) {
5240 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5241 			kmem_free(hopoptsptr, hopoptslen);
5242 			rw_exit(&icmp->icmp_rwlock);
5243 			return (ENOMEM);
5244 		}
5245 		mp1->b_cont = mp;
5246 		mp1->b_wptr = mp1->b_datap->db_lim;
5247 		ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len);
5248 		mp = mp1;
5249 	}
5250 	mp->b_rptr = (unsigned char *)ip6h;
5251 	ip6i = (ip6i_t *)ip6h;
5252 
5253 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp)
5254 	if (option_exists & IPPF_HAS_IP6I) {
5255 		ip6h = (ip6_t *)&ip6i[1];
5256 		ip6i->ip6i_flags = 0;
5257 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5258 
5259 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
5260 		if (option_exists & IPPF_SCOPE_ID) {
5261 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5262 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
5263 		} else if (option_exists & IPPF_IFINDEX) {
5264 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
5265 			ASSERT(tipp->ipp_ifindex != 0);
5266 			ip6i->ip6i_flags |= IP6I_IFINDEX;
5267 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
5268 		}
5269 
5270 		if (option_exists & IPPF_RAW_CKSUM) {
5271 			ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM;
5272 			ip6i->ip6i_checksum_off = icmp->icmp_checksum_off;
5273 		}
5274 
5275 		if (option_exists & IPPF_NO_CKSUM) {
5276 			ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM;
5277 		}
5278 
5279 		if (option_exists & IPPF_ADDR) {
5280 			/*
5281 			 * Enable per-packet source address verification if
5282 			 * IPV6_PKTINFO specified the source address.
5283 			 * ip6_src is set in the transport's _wput function.
5284 			 */
5285 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
5286 		}
5287 
5288 		if (option_exists & IPPF_DONTFRAG) {
5289 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
5290 		}
5291 
5292 		if (option_exists & IPPF_USE_MIN_MTU) {
5293 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
5294 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
5295 		}
5296 
5297 		if (option_exists & IPPF_NEXTHOP) {
5298 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
5299 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
5300 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
5301 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
5302 		}
5303 
5304 		/*
5305 		 * tell IP this is an ip6i_t private header
5306 		 */
5307 		ip6i->ip6i_nxt = IPPROTO_RAW;
5308 	}
5309 
5310 	/* Initialize IPv6 header */
5311 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
5312 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
5313 
5314 	/* Set the hoplimit of the outgoing packet. */
5315 	if (option_exists & IPPF_HOPLIMIT) {
5316 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
5317 		ip6h->ip6_hops = ipp->ipp_hoplimit;
5318 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5319 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
5320 		ip6h->ip6_hops = icmp->icmp_multicast_ttl;
5321 		if (option_exists & IPPF_MULTICAST_HOPS)
5322 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5323 	} else {
5324 		ip6h->ip6_hops = icmp->icmp_ttl;
5325 		if (option_exists & IPPF_UNICAST_HOPS)
5326 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
5327 	}
5328 
5329 	if (option_exists & IPPF_ADDR) {
5330 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
5331 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
5332 		ip6h->ip6_src = tipp->ipp_addr;
5333 	} else {
5334 		/*
5335 		 * The source address was not set using IPV6_PKTINFO.
5336 		 * First look at the bound source.
5337 		 * If unspecified fallback to __sin6_src_id.
5338 		 */
5339 		ip6h->ip6_src = icmp->icmp_v6src;
5340 		if (sin6->__sin6_src_id != 0 &&
5341 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
5342 			ip_srcid_find_id(sin6->__sin6_src_id,
5343 			    &ip6h->ip6_src, icmp->icmp_zoneid,
5344 			    is->is_netstack);
5345 		}
5346 	}
5347 
5348 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
5349 	cp = (uint8_t *)&ip6h[1];
5350 
5351 	/*
5352 	 * Here's where we have to start stringing together
5353 	 * any extension headers in the right order:
5354 	 * Hop-by-hop, destination, routing, and final destination opts.
5355 	 */
5356 	if (option_exists & IPPF_HOPOPTS) {
5357 		/* Hop-by-hop options */
5358 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
5359 
5360 		*nxthdr_ptr = IPPROTO_HOPOPTS;
5361 		nxthdr_ptr = &hbh->ip6h_nxt;
5362 
5363 		if (hopoptslen == 0) {
5364 			tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
5365 			bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen);
5366 			cp += tipp->ipp_hopoptslen;
5367 		} else {
5368 			bcopy(hopoptsptr, cp, hopoptslen);
5369 			cp += hopoptslen;
5370 			kmem_free(hopoptsptr, hopoptslen);
5371 		}
5372 	}
5373 	/*
5374 	 * En-route destination options
5375 	 * Only do them if there's a routing header as well
5376 	 */
5377 	if (option_exists & IPPF_RTDSTOPTS) {
5378 		ip6_dest_t *dst = (ip6_dest_t *)cp;
5379 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
5380 
5381 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5382 		nxthdr_ptr = &dst->ip6d_nxt;
5383 
5384 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
5385 		cp += tipp->ipp_rtdstoptslen;
5386 	}
5387 	/*
5388 	 * Routing header next
5389 	 */
5390 	if (option_exists & IPPF_RTHDR) {
5391 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
5392 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
5393 
5394 		*nxthdr_ptr = IPPROTO_ROUTING;
5395 		nxthdr_ptr = &rt->ip6r_nxt;
5396 
5397 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
5398 		cp += tipp->ipp_rthdrlen;
5399 	}
5400 	/*
5401 	 * Do ultimate destination options
5402 	 */
5403 	if (option_exists & IPPF_DSTOPTS) {
5404 		ip6_dest_t *dest = (ip6_dest_t *)cp;
5405 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
5406 
5407 		*nxthdr_ptr = IPPROTO_DSTOPTS;
5408 		nxthdr_ptr = &dest->ip6d_nxt;
5409 
5410 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
5411 		cp += tipp->ipp_dstoptslen;
5412 	}
5413 
5414 	/*
5415 	 * Now set the last header pointer to the proto passed in
5416 	 */
5417 	ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len);
5418 	*nxthdr_ptr = icmp->icmp_proto;
5419 
5420 	/*
5421 	 * Copy in the destination address
5422 	 */
5423 	ip6h->ip6_dst = ip6_dst;
5424 
5425 	ip6h->ip6_vcf =
5426 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
5427 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
5428 
5429 	if (option_exists & IPPF_TCLASS) {
5430 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
5431 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
5432 		    tipp->ipp_tclass);
5433 	}
5434 	if (option_exists & IPPF_RTHDR) {
5435 		ip6_rthdr_t	*rth;
5436 
5437 		/*
5438 		 * Perform any processing needed for source routing.
5439 		 * We know that all extension headers will be in the same mblk
5440 		 * as the IPv6 header.
5441 		 */
5442 		rth = ip_find_rthdr_v6(ip6h, mp->b_wptr);
5443 		if (rth != NULL && rth->ip6r_segleft != 0) {
5444 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
5445 				/*
5446 				 * Drop packet - only support Type 0 routing.
5447 				 * Notify the application as well.
5448 				 */
5449 				BUMP_MIB(&is->is_rawip_mib,
5450 				    rawipOutErrors);
5451 				rw_exit(&icmp->icmp_rwlock);
5452 				return (EPROTO);
5453 			}
5454 			/*
5455 			 * rth->ip6r_len is twice the number of
5456 			 * addresses in the header
5457 			 */
5458 			if (rth->ip6r_len & 0x1) {
5459 				BUMP_MIB(&is->is_rawip_mib,
5460 				    rawipOutErrors);
5461 				rw_exit(&icmp->icmp_rwlock);
5462 				return (EPROTO);
5463 			}
5464 			/*
5465 			 * Shuffle the routing header and ip6_dst
5466 			 * addresses, and get the checksum difference
5467 			 * between the first hop (in ip6_dst) and
5468 			 * the destination (in the last routing hdr entry).
5469 			 */
5470 			csum = ip_massage_options_v6(ip6h, rth,
5471 			    is->is_netstack);
5472 			/*
5473 			 * Verify that the first hop isn't a mapped address.
5474 			 * Routers along the path need to do this verification
5475 			 * for subsequent hops.
5476 			 */
5477 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
5478 				BUMP_MIB(&is->is_rawip_mib,
5479 				    rawipOutErrors);
5480 				rw_exit(&icmp->icmp_rwlock);
5481 				return (EADDRNOTAVAIL);
5482 			}
5483 		}
5484 	}
5485 
5486 	ip_len = mp->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN;
5487 	if (mp->b_cont != NULL)
5488 		ip_len += msgdsize(mp->b_cont);
5489 
5490 	/*
5491 	 * Set the length into the IP header.
5492 	 * If the length is greater than the maximum allowed by IP,
5493 	 * then free the message and return. Do not try and send it
5494 	 * as this can cause problems in layers below.
5495 	 */
5496 	if (ip_len > IP_MAXPACKET) {
5497 		BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
5498 		rw_exit(&icmp->icmp_rwlock);
5499 		return (EMSGSIZE);
5500 	}
5501 	if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) {
5502 		uint_t	cksum_off;	/* From ip6i == mp->b_rptr */
5503 		uint16_t *cksum_ptr;
5504 		uint_t	ext_hdrs_len;
5505 
5506 		/* ICMPv6 must have an offset matching icmp6_cksum offset */
5507 		ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 ||
5508 		    icmp->icmp_checksum_off == 2);
5509 
5510 		/*
5511 		 * We make it easy for IP to include our pseudo header
5512 		 * by putting our length in uh_checksum, modified (if
5513 		 * we have a routing header) by the checksum difference
5514 		 * between the ultimate destination and first hop addresses.
5515 		 * Note: ICMPv6 must always checksum the packet.
5516 		 */
5517 		cksum_off = ip_hdr_len + icmp->icmp_checksum_off;
5518 		if (cksum_off + sizeof (uint16_t) > mp->b_wptr - mp->b_rptr) {
5519 			if (!pullupmsg(mp, cksum_off + sizeof (uint16_t))) {
5520 				BUMP_MIB(&is->is_rawip_mib,
5521 				    rawipOutErrors);
5522 				freemsg(mp);
5523 				rw_exit(&icmp->icmp_rwlock);
5524 				return (0);
5525 			}
5526 			ip6i = (ip6i_t *)mp->b_rptr;
5527 			if (ip6i->ip6i_nxt == IPPROTO_RAW)
5528 				ip6h = (ip6_t *)&ip6i[1];
5529 			else
5530 				ip6h = (ip6_t *)ip6i;
5531 		}
5532 		/* Add payload length to checksum */
5533 		ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN -
5534 		    (int)((uchar_t *)ip6h - (uchar_t *)ip6i);
5535 		csum += htons(ip_len - ext_hdrs_len);
5536 
5537 		cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off);
5538 		csum = (csum & 0xFFFF) + (csum >> 16);
5539 		*cksum_ptr = (uint16_t)csum;
5540 	}
5541 
5542 #ifdef _LITTLE_ENDIAN
5543 	ip_len = htons(ip_len);
5544 #endif
5545 	ip6h->ip6_plen = (uint16_t)ip_len;
5546 
5547 	/* We're done. Pass the packet to IP */
5548 	rw_exit(&icmp->icmp_rwlock);
5549 	BUMP_MIB(&is->is_rawip_mib, rawipOutDatagrams);
5550 	ip_output_v6(icmp->icmp_connp, mp, q, IP_WPUT);
5551 	return (0);
5552 }
5553 
5554 static void
5555 icmp_wput_other(queue_t *q, mblk_t *mp)
5556 {
5557 	uchar_t	*rptr = mp->b_rptr;
5558 	struct iocblk *iocp;
5559 #define	tudr ((struct T_unitdata_req *)rptr)
5560 	conn_t	*connp = Q_TO_CONN(q);
5561 	icmp_t	*icmp = connp->conn_icmp;
5562 	icmp_stack_t *is = icmp->icmp_is;
5563 	cred_t *cr;
5564 
5565 	switch (mp->b_datap->db_type) {
5566 	case M_PROTO:
5567 	case M_PCPROTO:
5568 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5569 			/*
5570 			 * If the message does not contain a PRIM_type,
5571 			 * throw it away.
5572 			 */
5573 			freemsg(mp);
5574 			return;
5575 		}
5576 		switch (((union T_primitives *)rptr)->type) {
5577 		case T_ADDR_REQ:
5578 			icmp_addr_req(q, mp);
5579 			return;
5580 		case O_T_BIND_REQ:
5581 		case T_BIND_REQ:
5582 			icmp_tpi_bind(q, mp);
5583 			return;
5584 		case T_CONN_REQ:
5585 			icmp_tpi_connect(q, mp);
5586 			return;
5587 		case T_CAPABILITY_REQ:
5588 			icmp_capability_req(q, mp);
5589 			return;
5590 		case T_INFO_REQ:
5591 			icmp_info_req(q, mp);
5592 			return;
5593 		case T_UNITDATA_REQ:
5594 			/*
5595 			 * If a T_UNITDATA_REQ gets here, the address must
5596 			 * be bad.  Valid T_UNITDATA_REQs are found above
5597 			 * and break to below this switch.
5598 			 */
5599 			icmp_ud_err(q, mp, EADDRNOTAVAIL);
5600 			return;
5601 		case T_UNBIND_REQ:
5602 			icmp_tpi_unbind(q, mp);
5603 			return;
5604 
5605 		case T_SVR4_OPTMGMT_REQ:
5606 			/*
5607 			 * All Solaris components should pass a db_credp
5608 			 * for this TPI message, hence we ASSERT.
5609 			 * But in case there is some other M_PROTO that looks
5610 			 * like a TPI message sent by some other kernel
5611 			 * component, we check and return an error.
5612 			 */
5613 			cr = msg_getcred(mp, NULL);
5614 			ASSERT(cr != NULL);
5615 			if (cr == NULL) {
5616 				icmp_err_ack(q, mp, TSYSERR, EINVAL);
5617 				return;
5618 			}
5619 
5620 			if (!snmpcom_req(q, mp, icmp_snmp_set, ip_snmp_get,
5621 			    cr)) {
5622 				/* Only IP can return anything meaningful */
5623 				(void) svr4_optcom_req(q, mp, cr,
5624 				    &icmp_opt_obj, B_TRUE);
5625 			}
5626 			return;
5627 
5628 		case T_OPTMGMT_REQ:
5629 			/*
5630 			 * All Solaris components should pass a db_credp
5631 			 * for this TPI message, hence we ASSERT.
5632 			 * But in case there is some other M_PROTO that looks
5633 			 * like a TPI message sent by some other kernel
5634 			 * component, we check and return an error.
5635 			 */
5636 			cr = msg_getcred(mp, NULL);
5637 			ASSERT(cr != NULL);
5638 			if (cr == NULL) {
5639 				icmp_err_ack(q, mp, TSYSERR, EINVAL);
5640 				return;
5641 			}
5642 			/* Only IP can return anything meaningful */
5643 			(void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj, B_TRUE);
5644 			return;
5645 
5646 		case T_DISCON_REQ:
5647 			icmp_tpi_disconnect(q, mp);
5648 			return;
5649 
5650 		/* The following TPI message is not supported by icmp. */
5651 		case O_T_CONN_RES:
5652 		case T_CONN_RES:
5653 			icmp_err_ack(q, mp, TNOTSUPPORT, 0);
5654 			return;
5655 
5656 		/* The following 3 TPI requests are illegal for icmp. */
5657 		case T_DATA_REQ:
5658 		case T_EXDATA_REQ:
5659 		case T_ORDREL_REQ:
5660 			freemsg(mp);
5661 			(void) putctl1(RD(q), M_ERROR, EPROTO);
5662 			return;
5663 		default:
5664 			break;
5665 		}
5666 		break;
5667 	case M_IOCTL:
5668 		iocp = (struct iocblk *)mp->b_rptr;
5669 		switch (iocp->ioc_cmd) {
5670 		case TI_GETPEERNAME:
5671 			if (icmp->icmp_state != TS_DATA_XFER) {
5672 				/*
5673 				 * If a default destination address has not
5674 				 * been associated with the stream, then we
5675 				 * don't know the peer's name.
5676 				 */
5677 				iocp->ioc_error = ENOTCONN;
5678 		err_ret:;
5679 				iocp->ioc_count = 0;
5680 				mp->b_datap->db_type = M_IOCACK;
5681 				qreply(q, mp);
5682 				return;
5683 			}
5684 			/* FALLTHRU */
5685 		case TI_GETMYNAME:
5686 			/*
5687 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
5688 			 * need to copyin the user's strbuf structure.
5689 			 * Processing will continue in the M_IOCDATA case
5690 			 * below.
5691 			 */
5692 			mi_copyin(q, mp, NULL,
5693 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
5694 			return;
5695 		case ND_SET:
5696 			/* nd_getset performs the necessary error checking */
5697 		case ND_GET:
5698 			if (nd_getset(q, is->is_nd, mp)) {
5699 				qreply(q, mp);
5700 				return;
5701 			}
5702 			break;
5703 		case _SIOCSOCKFALLBACK:
5704 			/*
5705 			 * socket is falling back to be a
5706 			 * streams socket. Nothing  to do
5707 			 */
5708 			iocp->ioc_count = 0;
5709 			iocp->ioc_rval = 0;
5710 			qreply(q, mp);
5711 			return;
5712 		default:
5713 			break;
5714 		}
5715 		break;
5716 	case M_IOCDATA:
5717 		icmp_wput_iocdata(q, mp);
5718 		return;
5719 	default:
5720 		break;
5721 	}
5722 	ip_wput(q, mp);
5723 }
5724 
5725 /*
5726  * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA
5727  * messages.
5728  */
5729 static void
5730 icmp_wput_iocdata(queue_t *q, mblk_t *mp)
5731 {
5732 	mblk_t	*mp1;
5733 	STRUCT_HANDLE(strbuf, sb);
5734 	icmp_t	*icmp;
5735 	uint_t	addrlen;
5736 	uint_t	error;
5737 
5738 	/* Make sure it is one of ours. */
5739 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5740 	case TI_GETMYNAME:
5741 	case TI_GETPEERNAME:
5742 		break;
5743 	default:
5744 		icmp = Q_TO_ICMP(q);
5745 		ip_output(icmp->icmp_connp, mp, q, IP_WPUT);
5746 		return;
5747 	}
5748 	switch (mi_copy_state(q, mp, &mp1)) {
5749 	case -1:
5750 		return;
5751 	case MI_COPY_CASE(MI_COPY_IN, 1):
5752 		break;
5753 	case MI_COPY_CASE(MI_COPY_OUT, 1):
5754 		/*
5755 		 * The address has been copied out, so now
5756 		 * copyout the strbuf.
5757 		 */
5758 		mi_copyout(q, mp);
5759 		return;
5760 	case MI_COPY_CASE(MI_COPY_OUT, 2):
5761 		/*
5762 		 * The address and strbuf have been copied out.
5763 		 * We're done, so just acknowledge the original
5764 		 * M_IOCTL.
5765 		 */
5766 		mi_copy_done(q, mp, 0);
5767 		return;
5768 	default:
5769 		/*
5770 		 * Something strange has happened, so acknowledge
5771 		 * the original M_IOCTL with an EPROTO error.
5772 		 */
5773 		mi_copy_done(q, mp, EPROTO);
5774 		return;
5775 	}
5776 	/*
5777 	 * Now we have the strbuf structure for TI_GETMYNAME
5778 	 * and TI_GETPEERNAME.  Next we copyout the requested
5779 	 * address and then we'll copyout the strbuf.
5780 	 */
5781 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
5782 	    (void *)mp1->b_rptr);
5783 	icmp = Q_TO_ICMP(q);
5784 	if (icmp->icmp_family == AF_INET)
5785 		addrlen = sizeof (sin_t);
5786 	else
5787 		addrlen = sizeof (sin6_t);
5788 
5789 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
5790 		mi_copy_done(q, mp, EINVAL);
5791 		return;
5792 	}
5793 
5794 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
5795 
5796 	if (mp1 == NULL)
5797 		return;
5798 
5799 	rw_enter(&icmp->icmp_rwlock, RW_READER);
5800 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
5801 	case TI_GETMYNAME:
5802 		error = rawip_do_getsockname(icmp, (void *)mp1->b_rptr,
5803 		    &addrlen);
5804 		break;
5805 	case TI_GETPEERNAME:
5806 		error = rawip_do_getpeername(icmp, (void *)mp1->b_rptr,
5807 		    &addrlen);
5808 		break;
5809 	}
5810 	rw_exit(&icmp->icmp_rwlock);
5811 
5812 	if (error != 0) {
5813 		mi_copy_done(q, mp, error);
5814 	} else {
5815 		mp1->b_wptr += addrlen;
5816 		STRUCT_FSET(sb, len, addrlen);
5817 
5818 		/* Copy out the address */
5819 		mi_copyout(q, mp);
5820 	}
5821 }
5822 
5823 static int
5824 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
5825     void *thisdg_attrs)
5826 {
5827 	struct T_unitdata_req *udreqp;
5828 	int is_absreq_failure;
5829 	cred_t *cr;
5830 
5831 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
5832 	*errorp = 0;
5833 
5834 	/*
5835 	 * All Solaris components should pass a db_credp
5836 	 * for this TPI message, hence we ASSERT.
5837 	 * But in case there is some other M_PROTO that looks
5838 	 * like a TPI message sent by some other kernel
5839 	 * component, we check and return an error.
5840 	 */
5841 	cr = msg_getcred(mp, NULL);
5842 	ASSERT(cr != NULL);
5843 	if (cr == NULL)
5844 		return (-1);
5845 
5846 	*errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length,
5847 	    udreqp->OPT_offset, cr, &icmp_opt_obj,
5848 	    thisdg_attrs, &is_absreq_failure);
5849 
5850 	if (*errorp != 0) {
5851 		/*
5852 		 * Note: No special action needed in this
5853 		 * module for "is_absreq_failure"
5854 		 */
5855 		return (-1);		/* failure */
5856 	}
5857 	ASSERT(is_absreq_failure == 0);
5858 	return (0);	/* success */
5859 }
5860 
5861 void
5862 icmp_ddi_g_init(void)
5863 {
5864 	icmp_max_optsize = optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr,
5865 	    icmp_opt_obj.odb_opt_arr_cnt);
5866 
5867 	/*
5868 	 * We want to be informed each time a stack is created or
5869 	 * destroyed in the kernel, so we can maintain the
5870 	 * set of icmp_stack_t's.
5871 	 */
5872 	netstack_register(NS_ICMP, rawip_stack_init, NULL, rawip_stack_fini);
5873 }
5874 
5875 void
5876 icmp_ddi_g_destroy(void)
5877 {
5878 	netstack_unregister(NS_ICMP);
5879 }
5880 
5881 #define	INET_NAME	"ip"
5882 
5883 /*
5884  * Initialize the ICMP stack instance.
5885  */
5886 static void *
5887 rawip_stack_init(netstackid_t stackid, netstack_t *ns)
5888 {
5889 	icmp_stack_t	*is;
5890 	icmpparam_t	*pa;
5891 	int		error = 0;
5892 	major_t		major;
5893 
5894 	is = (icmp_stack_t *)kmem_zalloc(sizeof (*is), KM_SLEEP);
5895 	is->is_netstack = ns;
5896 
5897 	pa = (icmpparam_t *)kmem_alloc(sizeof (icmp_param_arr), KM_SLEEP);
5898 	is->is_param_arr = pa;
5899 	bcopy(icmp_param_arr, is->is_param_arr, sizeof (icmp_param_arr));
5900 
5901 	(void) icmp_param_register(&is->is_nd,
5902 	    is->is_param_arr, A_CNT(icmp_param_arr));
5903 	is->is_ksp = rawip_kstat_init(stackid);
5904 
5905 	major = mod_name_to_major(INET_NAME);
5906 	error = ldi_ident_from_major(major, &is->is_ldi_ident);
5907 	ASSERT(error == 0);
5908 	return (is);
5909 }
5910 
5911 /*
5912  * Free the ICMP stack instance.
5913  */
5914 static void
5915 rawip_stack_fini(netstackid_t stackid, void *arg)
5916 {
5917 	icmp_stack_t *is = (icmp_stack_t *)arg;
5918 
5919 	nd_free(&is->is_nd);
5920 	kmem_free(is->is_param_arr, sizeof (icmp_param_arr));
5921 	is->is_param_arr = NULL;
5922 
5923 	rawip_kstat_fini(stackid, is->is_ksp);
5924 	is->is_ksp = NULL;
5925 	ldi_ident_release(is->is_ldi_ident);
5926 	kmem_free(is, sizeof (*is));
5927 }
5928 
5929 static void *
5930 rawip_kstat_init(netstackid_t stackid) {
5931 	kstat_t	*ksp;
5932 
5933 	rawip_named_kstat_t template = {
5934 		{ "inDatagrams",	KSTAT_DATA_UINT32, 0 },
5935 		{ "inCksumErrs",	KSTAT_DATA_UINT32, 0 },
5936 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
5937 		{ "outDatagrams",	KSTAT_DATA_UINT32, 0 },
5938 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
5939 	};
5940 
5941 	ksp = kstat_create_netstack("icmp", 0, "rawip", "mib2",
5942 					KSTAT_TYPE_NAMED,
5943 					NUM_OF_FIELDS(rawip_named_kstat_t),
5944 					0, stackid);
5945 	if (ksp == NULL || ksp->ks_data == NULL)
5946 		return (NULL);
5947 
5948 	bcopy(&template, ksp->ks_data, sizeof (template));
5949 	ksp->ks_update = rawip_kstat_update;
5950 	ksp->ks_private = (void *)(uintptr_t)stackid;
5951 
5952 	kstat_install(ksp);
5953 	return (ksp);
5954 }
5955 
5956 static void
5957 rawip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
5958 {
5959 	if (ksp != NULL) {
5960 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
5961 		kstat_delete_netstack(ksp, stackid);
5962 	}
5963 }
5964 
5965 static int
5966 rawip_kstat_update(kstat_t *ksp, int rw)
5967 {
5968 	rawip_named_kstat_t *rawipkp;
5969 	netstackid_t	stackid = (netstackid_t)(uintptr_t)ksp->ks_private;
5970 	netstack_t	*ns;
5971 	icmp_stack_t	*is;
5972 
5973 	if ((ksp == NULL) || (ksp->ks_data == NULL))
5974 		return (EIO);
5975 
5976 	if (rw == KSTAT_WRITE)
5977 		return (EACCES);
5978 
5979 	rawipkp = (rawip_named_kstat_t *)ksp->ks_data;
5980 
5981 	ns = netstack_find_by_stackid(stackid);
5982 	if (ns == NULL)
5983 		return (-1);
5984 	is = ns->netstack_icmp;
5985 	if (is == NULL) {
5986 		netstack_rele(ns);
5987 		return (-1);
5988 	}
5989 	rawipkp->inDatagrams.value.ui32 =  is->is_rawip_mib.rawipInDatagrams;
5990 	rawipkp->inCksumErrs.value.ui32 =  is->is_rawip_mib.rawipInCksumErrs;
5991 	rawipkp->inErrors.value.ui32 =	   is->is_rawip_mib.rawipInErrors;
5992 	rawipkp->outDatagrams.value.ui32 = is->is_rawip_mib.rawipOutDatagrams;
5993 	rawipkp->outErrors.value.ui32 =	   is->is_rawip_mib.rawipOutErrors;
5994 	netstack_rele(ns);
5995 	return (0);
5996 }
5997 
5998 /* ARGSUSED */
5999 int
6000 rawip_accept(sock_lower_handle_t lproto_handle,
6001     sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle,
6002     cred_t *cr)
6003 {
6004 	return (EOPNOTSUPP);
6005 }
6006 
6007 /* ARGSUSED */
6008 int
6009 rawip_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6010     socklen_t len, cred_t *cr)
6011 {
6012 	conn_t  *connp = (conn_t *)proto_handle;
6013 	int error;
6014 
6015 	/* All Solaris components should pass a cred for this operation. */
6016 	ASSERT(cr != NULL);
6017 
6018 	/* Binding to a NULL address really means unbind */
6019 	if (sa == NULL)
6020 		error = rawip_do_unbind(connp);
6021 	else
6022 		error = rawip_do_bind(connp, sa, len);
6023 
6024 	if (error < 0) {
6025 		if (error == -TOUTSTATE)
6026 			error = EINVAL;
6027 		else
6028 			error = proto_tlitosyserr(-error);
6029 	}
6030 	return (error);
6031 }
6032 
6033 static int
6034 rawip_implicit_bind(conn_t *connp)
6035 {
6036 	sin6_t sin6addr;
6037 	sin_t *sin;
6038 	sin6_t *sin6;
6039 	socklen_t len;
6040 	int error;
6041 
6042 	if (connp->conn_icmp->icmp_family == AF_INET) {
6043 		len = sizeof (struct sockaddr_in);
6044 		sin = (sin_t *)&sin6addr;
6045 		*sin = sin_null;
6046 		sin->sin_family = AF_INET;
6047 		sin->sin_addr.s_addr = INADDR_ANY;
6048 	} else {
6049 		ASSERT(connp->conn_icmp->icmp_family == AF_INET6);
6050 		len = sizeof (sin6_t);
6051 		sin6 = (sin6_t *)&sin6addr;
6052 		*sin6 = sin6_null;
6053 		sin6->sin6_family = AF_INET6;
6054 		V6_SET_ZERO(sin6->sin6_addr);
6055 	}
6056 
6057 	error = rawip_do_bind(connp, (struct sockaddr *)&sin6addr, len);
6058 
6059 	return ((error < 0) ? proto_tlitosyserr(-error) : error);
6060 }
6061 
6062 static int
6063 rawip_unbind(conn_t *connp)
6064 {
6065 	int error;
6066 
6067 	error = rawip_do_unbind(connp);
6068 	if (error < 0) {
6069 		error = proto_tlitosyserr(-error);
6070 	}
6071 	return (error);
6072 }
6073 
6074 /* ARGSUSED */
6075 int
6076 rawip_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr)
6077 {
6078 	return (EOPNOTSUPP);
6079 }
6080 
6081 /* ARGSUSED */
6082 int
6083 rawip_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa,
6084     socklen_t len, sock_connid_t *id, cred_t *cr)
6085 {
6086 	conn_t	*connp = (conn_t *)proto_handle;
6087 	icmp_t *icmp = connp->conn_icmp;
6088 	int	error;
6089 	boolean_t did_bind = B_FALSE;
6090 
6091 	/* All Solaris components should pass a cred for this operation. */
6092 	ASSERT(cr != NULL);
6093 
6094 	if (sa == NULL) {
6095 		/*
6096 		 * Disconnect
6097 		 * Make sure we are connected
6098 		 */
6099 		if (icmp->icmp_state != TS_DATA_XFER)
6100 			return (EINVAL);
6101 
6102 		error = icmp_disconnect(connp);
6103 		return (error);
6104 	}
6105 
6106 	error = proto_verify_ip_addr(icmp->icmp_family, sa, len);
6107 	if (error != 0)
6108 		return (error);
6109 
6110 	/* do an implicit bind if necessary */
6111 	if (icmp->icmp_state == TS_UNBND) {
6112 		error = rawip_implicit_bind(connp);
6113 		/*
6114 		 * We could be racing with an actual bind, in which case
6115 		 * we would see EPROTO. We cross our fingers and try
6116 		 * to connect.
6117 		 */
6118 		if (!(error == 0 || error == EPROTO))
6119 			return (error);
6120 		did_bind = B_TRUE;
6121 	}
6122 
6123 	/*
6124 	 * set SO_DGRAM_ERRIND
6125 	 */
6126 	icmp->icmp_dgram_errind = B_TRUE;
6127 
6128 	error = rawip_do_connect(connp, sa, len, cr);
6129 
6130 	if (error != 0 && did_bind) {
6131 		int unbind_err;
6132 
6133 		unbind_err = rawip_unbind(connp);
6134 		ASSERT(unbind_err == 0);
6135 	}
6136 
6137 	if (error == 0) {
6138 		*id = 0;
6139 		(*connp->conn_upcalls->su_connected)
6140 		    (connp->conn_upper_handle, 0, NULL, -1);
6141 	} else if (error < 0) {
6142 		error = proto_tlitosyserr(-error);
6143 	}
6144 	return (error);
6145 }
6146 
6147 /* ARGSUSED */
6148 int
6149 rawip_fallback(sock_lower_handle_t proto_handle, queue_t *q,
6150     boolean_t direct_sockfs, so_proto_quiesced_cb_t quiesced_cb)
6151 {
6152 	conn_t  *connp = (conn_t *)proto_handle;
6153 	icmp_t	*icmp;
6154 	struct T_capability_ack tca;
6155 	struct sockaddr_in6 laddr, faddr;
6156 	socklen_t laddrlen, faddrlen;
6157 	short opts;
6158 	struct stroptions *stropt;
6159 	mblk_t *stropt_mp;
6160 	int error;
6161 
6162 	icmp = connp->conn_icmp;
6163 
6164 	stropt_mp = allocb_wait(sizeof (*stropt), BPRI_HI, STR_NOSIG, NULL);
6165 
6166 	/*
6167 	 * setup the fallback stream that was allocated
6168 	 */
6169 	connp->conn_dev = (dev_t)RD(q)->q_ptr;
6170 	connp->conn_minor_arena = WR(q)->q_ptr;
6171 
6172 	RD(q)->q_ptr = WR(q)->q_ptr = connp;
6173 
6174 	WR(q)->q_qinfo = &icmpwinit;
6175 
6176 	connp->conn_rq = RD(q);
6177 	connp->conn_wq = WR(q);
6178 
6179 	/* Notify stream head about options before sending up data */
6180 	stropt_mp->b_datap->db_type = M_SETOPTS;
6181 	stropt_mp->b_wptr += sizeof (*stropt);
6182 	stropt = (struct stroptions *)stropt_mp->b_rptr;
6183 	stropt->so_flags = SO_WROFF | SO_HIWAT;
6184 	stropt->so_wroff =
6185 	    (ushort_t)(icmp->icmp_max_hdr_len + icmp->icmp_is->is_wroff_extra);
6186 	stropt->so_hiwat = icmp->icmp_recv_hiwat;
6187 	putnext(RD(q), stropt_mp);
6188 
6189 	/*
6190 	 * free helper stream
6191 	 */
6192 	ip_free_helper_stream(connp);
6193 
6194 	/*
6195 	 * Collect the information needed to sync with the sonode
6196 	 */
6197 	icmp_do_capability_ack(icmp, &tca, TC1_INFO);
6198 
6199 	laddrlen = faddrlen = sizeof (sin6_t);
6200 	(void) rawip_getsockname((sock_lower_handle_t)connp,
6201 	    (struct sockaddr *)&laddr, &laddrlen, CRED());
6202 	error = rawip_getpeername((sock_lower_handle_t)connp,
6203 	    (struct sockaddr *)&faddr, &faddrlen, CRED());
6204 	if (error != 0)
6205 		faddrlen = 0;
6206 	opts = 0;
6207 	if (icmp->icmp_dgram_errind)
6208 		opts |= SO_DGRAM_ERRIND;
6209 	if (icmp->icmp_dontroute)
6210 		opts |= SO_DONTROUTE;
6211 
6212 	(*quiesced_cb)(connp->conn_upper_handle, q, &tca,
6213 	    (struct sockaddr *)&laddr, laddrlen,
6214 	    (struct sockaddr *)&faddr, faddrlen, opts);
6215 
6216 	/*
6217 	 * Attempts to send data up during fallback will result in it being
6218 	 * queued in udp_t. Now we push up any queued packets.
6219 	 */
6220 	mutex_enter(&icmp->icmp_recv_lock);
6221 	while (icmp->icmp_fallback_queue_head != NULL) {
6222 		mblk_t	*mp;
6223 
6224 		mp = icmp->icmp_fallback_queue_head;
6225 		icmp->icmp_fallback_queue_head = mp->b_next;
6226 		mp->b_next = NULL;
6227 		mutex_exit(&icmp->icmp_recv_lock);
6228 		putnext(RD(q), mp);
6229 		mutex_enter(&icmp->icmp_recv_lock);
6230 	}
6231 	icmp->icmp_fallback_queue_tail = icmp->icmp_fallback_queue_head;
6232 
6233 	/*
6234 	 * No longer a streams less socket
6235 	 */
6236 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6237 	connp->conn_flags &= ~IPCL_NONSTR;
6238 	rw_exit(&icmp->icmp_rwlock);
6239 
6240 	mutex_exit(&icmp->icmp_recv_lock);
6241 
6242 	ASSERT(icmp->icmp_fallback_queue_head == NULL &&
6243 	    icmp->icmp_fallback_queue_tail == NULL);
6244 
6245 	ASSERT(connp->conn_ref >= 1);
6246 
6247 	return (0);
6248 }
6249 
6250 /* ARGSUSED */
6251 sock_lower_handle_t
6252 rawip_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls,
6253     uint_t *smodep, int *errorp, int flags, cred_t *credp)
6254 {
6255 	conn_t *connp;
6256 
6257 	if (type != SOCK_RAW || (family != AF_INET && family != AF_INET6)) {
6258 		*errorp = EPROTONOSUPPORT;
6259 		return (NULL);
6260 	}
6261 
6262 	connp = icmp_open(family, credp, errorp, flags);
6263 	if (connp != NULL) {
6264 		icmp_stack_t *is;
6265 
6266 		is = connp->conn_icmp->icmp_is;
6267 		connp->conn_flags |= IPCL_NONSTR;
6268 
6269 		if (connp->conn_icmp->icmp_family == AF_INET6) {
6270 			/* Build initial header template for transmit */
6271 			rw_enter(&connp->conn_icmp->icmp_rwlock, RW_WRITER);
6272 			if ((*errorp =
6273 			    icmp_build_hdrs(connp->conn_icmp)) != 0) {
6274 				rw_exit(&connp->conn_icmp->icmp_rwlock);
6275 				ipcl_conn_destroy(connp);
6276 				return (NULL);
6277 			}
6278 			rw_exit(&connp->conn_icmp->icmp_rwlock);
6279 		}
6280 
6281 		connp->conn_icmp->icmp_recv_hiwat = is->is_recv_hiwat;
6282 		connp->conn_icmp->icmp_xmit_hiwat = is->is_xmit_hiwat;
6283 
6284 		if ((*errorp = ip_create_helper_stream(connp,
6285 		    is->is_ldi_ident)) != 0) {
6286 			cmn_err(CE_CONT, "create of IP helper stream failed\n");
6287 			(void) rawip_do_close(connp);
6288 			return (NULL);
6289 		}
6290 
6291 		mutex_enter(&connp->conn_lock);
6292 		connp->conn_state_flags &= ~CONN_INCIPIENT;
6293 		mutex_exit(&connp->conn_lock);
6294 		*sock_downcalls = &sock_rawip_downcalls;
6295 		*smodep = SM_ATOMIC;
6296 	} else {
6297 		ASSERT(*errorp != 0);
6298 	}
6299 
6300 	return ((sock_lower_handle_t)connp);
6301 }
6302 
6303 /* ARGSUSED */
6304 void
6305 rawip_activate(sock_lower_handle_t proto_handle,
6306     sock_upper_handle_t sock_handle, sock_upcalls_t *sock_upcalls, int flags,
6307     cred_t *cr)
6308 {
6309 	conn_t 			*connp = (conn_t *)proto_handle;
6310 	icmp_stack_t 		*is = connp->conn_icmp->icmp_is;
6311 	struct sock_proto_props sopp;
6312 
6313 	/* All Solaris components should pass a cred for this operation. */
6314 	ASSERT(cr != NULL);
6315 
6316 	connp->conn_upcalls = sock_upcalls;
6317 	connp->conn_upper_handle = sock_handle;
6318 
6319 	sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT | SOCKOPT_RCVLOWAT |
6320 	    SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ;
6321 	sopp.sopp_wroff = connp->conn_icmp->icmp_max_hdr_len +
6322 	    is->is_wroff_extra;
6323 	sopp.sopp_rxhiwat = is->is_recv_hiwat;
6324 	sopp.sopp_rxlowat = icmp_mod_info.mi_lowat;
6325 	sopp.sopp_maxblk = INFPSZ;
6326 	sopp.sopp_maxpsz = IP_MAXPACKET;
6327 	sopp.sopp_minpsz = (icmp_mod_info.mi_minpsz == 1) ? 0 :
6328 	    icmp_mod_info.mi_minpsz;
6329 
6330 	(*connp->conn_upcalls->su_set_proto_props)
6331 	    (connp->conn_upper_handle, &sopp);
6332 }
6333 
6334 static int
6335 rawip_do_getsockname(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6336 {
6337 	sin_t	*sin = (sin_t *)sa;
6338 	sin6_t	*sin6 = (sin6_t *)sa;
6339 
6340 	ASSERT(icmp != NULL);
6341 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6342 
6343 	switch (icmp->icmp_family) {
6344 	case AF_INET:
6345 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6346 		if (*salenp < sizeof (sin_t))
6347 			return (EINVAL);
6348 
6349 		*salenp = sizeof (sin_t);
6350 		*sin = sin_null;
6351 		sin->sin_family = AF_INET;
6352 		if (icmp->icmp_state == TS_UNBND) {
6353 			break;
6354 		}
6355 
6356 		if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) &&
6357 		    !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6358 			sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_v6src);
6359 		} else {
6360 			/*
6361 			 * INADDR_ANY
6362 			 * icmp_v6src is not set, we might be bound to
6363 			 * broadcast/multicast. Use icmp_bound_v6src as
6364 			 * local address instead (that could
6365 			 * also still be INADDR_ANY)
6366 			 */
6367 			sin->sin_addr.s_addr =
6368 			    V4_PART_OF_V6(icmp->icmp_bound_v6src);
6369 		}
6370 		break;
6371 	case AF_INET6:
6372 
6373 		if (*salenp < sizeof (sin6_t))
6374 			return (EINVAL);
6375 
6376 		*salenp = sizeof (sin6_t);
6377 		*sin6 = sin6_null;
6378 		sin6->sin6_family = AF_INET6;
6379 		if (icmp->icmp_state == TS_UNBND) {
6380 			break;
6381 		}
6382 		if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) {
6383 			sin6->sin6_addr = icmp->icmp_v6src;
6384 		} else {
6385 			/*
6386 			 * UNSPECIFIED
6387 			 * icmp_v6src is not set, we might be bound to
6388 			 * broadcast/multicast. Use icmp_bound_v6src as
6389 			 * local address instead (that could
6390 			 * also still be UNSPECIFIED)
6391 			 */
6392 
6393 			sin6->sin6_addr = icmp->icmp_bound_v6src;
6394 		}
6395 		break;
6396 	}
6397 	return (0);
6398 }
6399 
6400 static int
6401 rawip_do_getpeername(icmp_t *icmp, struct sockaddr *sa, uint_t *salenp)
6402 {
6403 	sin_t   *sin = (sin_t *)sa;
6404 	sin6_t  *sin6 = (sin6_t *)sa;
6405 
6406 	ASSERT(icmp != NULL);
6407 	ASSERT(RW_LOCK_HELD(&icmp->icmp_rwlock));
6408 
6409 	if (icmp->icmp_state != TS_DATA_XFER)
6410 		return (ENOTCONN);
6411 
6412 	sa->sa_family = icmp->icmp_family;
6413 	switch (icmp->icmp_family) {
6414 	case AF_INET:
6415 		ASSERT(icmp->icmp_ipversion == IPV4_VERSION);
6416 
6417 		if (*salenp < sizeof (sin_t))
6418 			return (EINVAL);
6419 
6420 		*salenp = sizeof (sin_t);
6421 		*sin = sin_null;
6422 		sin->sin_family = AF_INET;
6423 		sin->sin_addr.s_addr =
6424 		    V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6425 		break;
6426 	case AF_INET6:
6427 		if (*salenp < sizeof (sin6_t))
6428 			return (EINVAL);
6429 
6430 		*salenp = sizeof (sin6_t);
6431 		*sin6 = sin6_null;
6432 		*sin6 = icmp->icmp_v6dst;
6433 		break;
6434 	}
6435 	return (0);
6436 }
6437 
6438 /* ARGSUSED */
6439 int
6440 rawip_getpeername(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6441     socklen_t *salenp, cred_t *cr)
6442 {
6443 	conn_t  *connp = (conn_t *)proto_handle;
6444 	icmp_t  *icmp = connp->conn_icmp;
6445 	int	error;
6446 
6447 	/* All Solaris components should pass a cred for this operation. */
6448 	ASSERT(cr != NULL);
6449 
6450 	ASSERT(icmp != NULL);
6451 
6452 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6453 
6454 	error = rawip_do_getpeername(icmp, sa, salenp);
6455 
6456 	rw_exit(&icmp->icmp_rwlock);
6457 
6458 	return (error);
6459 }
6460 
6461 /* ARGSUSED */
6462 int
6463 rawip_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *sa,
6464     socklen_t *salenp, cred_t *cr)
6465 {
6466 	conn_t  *connp = (conn_t *)proto_handle;
6467 	icmp_t	*icmp = connp->conn_icmp;
6468 	int	error;
6469 
6470 	/* All Solaris components should pass a cred for this operation. */
6471 	ASSERT(cr != NULL);
6472 
6473 	ASSERT(icmp != NULL);
6474 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6475 
6476 	error = rawip_do_getsockname(icmp, sa, salenp);
6477 
6478 	rw_exit(&icmp->icmp_rwlock);
6479 
6480 	return (error);
6481 }
6482 
6483 int
6484 rawip_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6485     const void *optvalp, socklen_t optlen, cred_t *cr)
6486 {
6487 	conn_t	*connp = (conn_t *)proto_handle;
6488 	icmp_t *icmp = connp->conn_icmp;
6489 	int error;
6490 
6491 	/* All Solaris components should pass a cred for this operation. */
6492 	ASSERT(cr != NULL);
6493 
6494 	error = proto_opt_check(level, option_name, optlen, NULL,
6495 	    icmp_opt_obj.odb_opt_des_arr,
6496 	    icmp_opt_obj.odb_opt_arr_cnt,
6497 	    icmp_opt_obj.odb_topmost_tpiprovider,
6498 	    B_TRUE, B_FALSE, cr);
6499 
6500 	if (error != 0) {
6501 		/*
6502 		 * option not recognized
6503 		 */
6504 		if (error < 0) {
6505 			error = proto_tlitosyserr(-error);
6506 		}
6507 		return (error);
6508 	}
6509 
6510 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6511 	error = icmp_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level,
6512 	    option_name, optlen, (uchar_t *)optvalp, (uint_t *)&optlen,
6513 	    (uchar_t *)optvalp, NULL, cr);
6514 	rw_exit(&icmp->icmp_rwlock);
6515 
6516 	if (error < 0) {
6517 		/*
6518 		 * Pass on to ip
6519 		 */
6520 		error = ip_set_options(connp, level, option_name, optvalp,
6521 		    optlen, cr);
6522 	}
6523 
6524 	ASSERT(error >= 0);
6525 
6526 	return (error);
6527 }
6528 
6529 int
6530 rawip_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name,
6531     void *optvalp, socklen_t *optlen, cred_t *cr)
6532 {
6533 	int		error;
6534 	conn_t		*connp = (conn_t *)proto_handle;
6535 	icmp_t		*icmp = connp->conn_icmp;
6536 	t_uscalar_t	max_optbuf_len;
6537 	void		*optvalp_buf;
6538 	int		len;
6539 
6540 	/* All Solaris components should pass a cred for this operation. */
6541 	ASSERT(cr != NULL);
6542 
6543 	error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len,
6544 	    icmp_opt_obj.odb_opt_des_arr,
6545 	    icmp_opt_obj.odb_opt_arr_cnt,
6546 	    icmp_opt_obj.odb_topmost_tpiprovider,
6547 	    B_FALSE, B_TRUE, cr);
6548 
6549 	if (error != 0) {
6550 		if (error < 0) {
6551 			error = proto_tlitosyserr(-error);
6552 		}
6553 		return (error);
6554 	}
6555 
6556 	optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP);
6557 	rw_enter(&icmp->icmp_rwlock, RW_READER);
6558 	len = icmp_opt_get(connp, level, option_name, optvalp_buf);
6559 	rw_exit(&icmp->icmp_rwlock);
6560 
6561 	if (len < 0) {
6562 		/*
6563 		 * Pass on to IP
6564 		 */
6565 		kmem_free(optvalp_buf, max_optbuf_len);
6566 		return (ip_get_options(connp, level, option_name, optvalp,
6567 		    optlen, cr));
6568 	} else {
6569 		/*
6570 		 * update optlen and copy option value
6571 		 */
6572 		t_uscalar_t size = MIN(len, *optlen);
6573 		bcopy(optvalp_buf, optvalp, size);
6574 		bcopy(&size, optlen, sizeof (size));
6575 
6576 		kmem_free(optvalp_buf, max_optbuf_len);
6577 		return (0);
6578 	}
6579 }
6580 
6581 /* ARGSUSED */
6582 int
6583 rawip_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr)
6584 {
6585 	conn_t	*connp = (conn_t *)proto_handle;
6586 
6587 	/* All Solaris components should pass a cred for this operation. */
6588 	ASSERT(cr != NULL);
6589 
6590 	(void) rawip_do_close(connp);
6591 	return (0);
6592 }
6593 
6594 /* ARGSUSED */
6595 int
6596 rawip_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr)
6597 {
6598 	conn_t  *connp = (conn_t *)proto_handle;
6599 
6600 	/* All Solaris components should pass a cred for this operation. */
6601 	ASSERT(cr != NULL);
6602 
6603 	/* shut down the send side */
6604 	if (how != SHUT_RD)
6605 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6606 		    SOCK_OPCTL_SHUT_SEND, 0);
6607 	/* shut down the recv side */
6608 	if (how != SHUT_WR)
6609 		(*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle,
6610 		    SOCK_OPCTL_SHUT_RECV, 0);
6611 	return (0);
6612 }
6613 
6614 void
6615 rawip_clr_flowctrl(sock_lower_handle_t proto_handle)
6616 {
6617 	conn_t  *connp = (conn_t *)proto_handle;
6618 	icmp_t	*icmp = connp->conn_icmp;
6619 
6620 	mutex_enter(&icmp->icmp_recv_lock);
6621 	connp->conn_flow_cntrld = B_FALSE;
6622 	mutex_exit(&icmp->icmp_recv_lock);
6623 }
6624 
6625 int
6626 rawip_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg,
6627     int mode, int32_t *rvalp, cred_t *cr)
6628 {
6629 	conn_t  	*connp = (conn_t *)proto_handle;
6630 	int		error;
6631 
6632 	/* All Solaris components should pass a cred for this operation. */
6633 	ASSERT(cr != NULL);
6634 
6635 	switch (cmd) {
6636 	case ND_SET:
6637 	case ND_GET:
6638 	case _SIOCSOCKFALLBACK:
6639 	case TI_GETPEERNAME:
6640 	case TI_GETMYNAME:
6641 #ifdef DEBUG
6642 		cmn_err(CE_CONT, "icmp_ioctl cmd 0x%x on non streams"
6643 		    " socket", cmd);
6644 #endif
6645 		error = EINVAL;
6646 		break;
6647 	default:
6648 		/*
6649 		 * Pass on to IP using helper stream
6650 		 */
6651 		error = ldi_ioctl(connp->conn_helper_info->iphs_handle,
6652 		    cmd, arg, mode, cr, rvalp);
6653 		break;
6654 	}
6655 	return (error);
6656 }
6657 
6658 /* ARGSUSED */
6659 int
6660 rawip_send(sock_lower_handle_t proto_handle, mblk_t *mp, struct nmsghdr *msg,
6661     cred_t *cr)
6662 {
6663 	conn_t *connp = (conn_t *)proto_handle;
6664 	icmp_t	*icmp = connp->conn_icmp;
6665 	icmp_stack_t *is = icmp->icmp_is;
6666 	int error = 0;
6667 	boolean_t bypass_dgram_errind = B_FALSE;
6668 
6669 	ASSERT(DB_TYPE(mp) == M_DATA);
6670 
6671 	/* All Solaris components should pass a cred for this operation. */
6672 	ASSERT(cr != NULL);
6673 
6674 	/* If labeled then sockfs should have already set db_credp */
6675 	ASSERT(!is_system_labeled() || msg_getcred(mp, NULL) != NULL);
6676 
6677 	/* do an implicit bind if necessary */
6678 	if (icmp->icmp_state == TS_UNBND) {
6679 		error = rawip_implicit_bind(connp);
6680 		/*
6681 		 * We could be racing with an actual bind, in which case
6682 		 * we would see EPROTO. We cross our fingers and try
6683 		 * to connect.
6684 		 */
6685 		if (!(error == 0 || error == EPROTO)) {
6686 			freemsg(mp);
6687 			return (error);
6688 		}
6689 	}
6690 
6691 	rw_enter(&icmp->icmp_rwlock, RW_WRITER);
6692 
6693 	if (msg->msg_name != NULL && icmp->icmp_state == TS_DATA_XFER) {
6694 		error = EISCONN;
6695 		goto done_lock;
6696 	}
6697 
6698 	switch (icmp->icmp_family) {
6699 	case AF_INET6: {
6700 		sin6_t	*sin6;
6701 		ip6_pkt_t	ipp_s;	/* For ancillary data options */
6702 		ip6_pkt_t	*ipp = &ipp_s;
6703 
6704 		sin6 = (sin6_t *)msg->msg_name;
6705 		if (sin6 != NULL) {
6706 			error = proto_verify_ip_addr(icmp->icmp_family,
6707 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6708 			if (error != 0) {
6709 				bypass_dgram_errind = B_TRUE;
6710 				goto done_lock;
6711 			}
6712 			if (icmp->icmp_delayed_error != 0) {
6713 				sin6_t  *sin1 = (sin6_t *)msg->msg_name;
6714 				sin6_t  *sin2 = (sin6_t *)
6715 				    &icmp->icmp_delayed_addr;
6716 
6717 				error = icmp->icmp_delayed_error;
6718 				icmp->icmp_delayed_error = 0;
6719 
6720 				/* Compare IP address and port */
6721 
6722 				if (sin1->sin6_port == sin2->sin6_port &&
6723 				    IN6_ARE_ADDR_EQUAL(&sin1->sin6_addr,
6724 				    &sin2->sin6_addr)) {
6725 					goto done_lock;
6726 				}
6727 			}
6728 		} else {
6729 			/*
6730 			 * Use connected address
6731 			 */
6732 			if (icmp->icmp_state != TS_DATA_XFER) {
6733 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6734 				error = EDESTADDRREQ;
6735 				bypass_dgram_errind = B_TRUE;
6736 				goto done_lock;
6737 			}
6738 			sin6 = &icmp->icmp_v6dst;
6739 		}
6740 
6741 		/* No support for mapped addresses on raw sockets */
6742 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
6743 			BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6744 			error = EADDRNOTAVAIL;
6745 			goto done_lock;
6746 		}
6747 
6748 		ipp->ipp_fields = 0;
6749 		ipp->ipp_sticky_ignored = 0;
6750 
6751 		/*
6752 		 * If options passed in, feed it for verification and handling
6753 		 */
6754 		if (msg->msg_controllen != 0) {
6755 			error = process_auxiliary_options(connp,
6756 			    msg->msg_control, msg->msg_controllen,
6757 			    ipp, &icmp_opt_obj, icmp_opt_set, cr);
6758 			if (error != 0) {
6759 				goto done_lock;
6760 			}
6761 		}
6762 
6763 		rw_exit(&icmp->icmp_rwlock);
6764 
6765 		/*
6766 		 * Destination is a native IPv6 address.
6767 		 * Send out an IPv6 format packet.
6768 		 */
6769 
6770 		error = raw_ip_send_data_v6(connp->conn_wq, connp, mp, sin6,
6771 		    ipp);
6772 	}
6773 		break;
6774 	case AF_INET: {
6775 		sin_t	*sin;
6776 		ip4_pkt_t pktinfo;
6777 		ip4_pkt_t *pktinfop = &pktinfo;
6778 		ipaddr_t	v4dst;
6779 
6780 		sin = (sin_t *)msg->msg_name;
6781 		if (sin != NULL) {
6782 			error = proto_verify_ip_addr(icmp->icmp_family,
6783 			    (struct sockaddr *)msg->msg_name, msg->msg_namelen);
6784 			if (error != 0) {
6785 				bypass_dgram_errind = B_TRUE;
6786 				goto done_lock;
6787 			}
6788 			v4dst = sin->sin_addr.s_addr;
6789 			if (icmp->icmp_delayed_error != 0) {
6790 				sin_t *sin1 = (sin_t *)msg->msg_name;
6791 				sin_t *sin2 = (sin_t *)&icmp->icmp_delayed_addr;
6792 
6793 				error = icmp->icmp_delayed_error;
6794 				icmp->icmp_delayed_error = 0;
6795 
6796 				/* Compare IP address and port */
6797 				if (sin1->sin_port == sin2->sin_port &&
6798 				    sin1->sin_addr.s_addr ==
6799 				    sin2->sin_addr.s_addr) {
6800 					goto done_lock;
6801 				}
6802 
6803 			}
6804 		} else {
6805 			/*
6806 			 * Use connected address
6807 			 */
6808 			if (icmp->icmp_state != TS_DATA_XFER) {
6809 				BUMP_MIB(&is->is_rawip_mib, rawipOutErrors);
6810 				error = EDESTADDRREQ;
6811 				bypass_dgram_errind = B_TRUE;
6812 				goto done_lock;
6813 			}
6814 			v4dst = V4_PART_OF_V6(icmp->icmp_v6dst.sin6_addr);
6815 		}
6816 
6817 
6818 		pktinfop->ip4_ill_index = 0;
6819 		pktinfop->ip4_addr = INADDR_ANY;
6820 
6821 		/*
6822 		 * If options passed in, feed it for verification and handling
6823 		 */
6824 		if (msg->msg_controllen != 0) {
6825 			error = process_auxiliary_options(connp,
6826 			    msg->msg_control, msg->msg_controllen,
6827 			    pktinfop, &icmp_opt_obj, icmp_opt_set, cr);
6828 			if (error != 0) {
6829 				goto done_lock;
6830 			}
6831 		}
6832 		rw_exit(&icmp->icmp_rwlock);
6833 
6834 		error = raw_ip_send_data_v4(connp->conn_wq, connp, mp,
6835 		    v4dst, pktinfop);
6836 		break;
6837 	}
6838 
6839 	default:
6840 		ASSERT(0);
6841 	}
6842 
6843 	goto done;
6844 
6845 done_lock:
6846 	rw_exit(&icmp->icmp_rwlock);
6847 	if (error != 0) {
6848 		ASSERT(mp != NULL);
6849 		freemsg(mp);
6850 	}
6851 done:
6852 	if (bypass_dgram_errind)
6853 		return (error);
6854 	return (icmp->icmp_dgram_errind ? error : 0);
6855 }
6856 
6857 sock_downcalls_t sock_rawip_downcalls = {
6858 	rawip_activate,
6859 	rawip_accept,
6860 	rawip_bind,
6861 	rawip_listen,
6862 	rawip_connect,
6863 	rawip_getpeername,
6864 	rawip_getsockname,
6865 	rawip_getsockopt,
6866 	rawip_setsockopt,
6867 	rawip_send,
6868 	NULL,
6869 	NULL,
6870 	NULL,
6871 	rawip_shutdown,
6872 	rawip_clr_flowctrl,
6873 	rawip_ioctl,
6874 	rawip_close
6875 };
6876