xref: /titanic_41/usr/src/uts/common/inet/ip/ip6_if.c (revision 4bff34e37def8a90f9194d81bc345c52ba20086a)
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 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /*
26  * Copyright (c) 1990 Mentat Inc.
27  */
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 /*
32  * This file contains the interface control functions for IPv6.
33  */
34 
35 #include <sys/types.h>
36 #include <sys/sysmacros.h>
37 #include <sys/stream.h>
38 #include <sys/dlpi.h>
39 #include <sys/stropts.h>
40 #include <sys/ddi.h>
41 #include <sys/cmn_err.h>
42 #include <sys/kstat.h>
43 #include <sys/debug.h>
44 #include <sys/zone.h>
45 #include <sys/policy.h>
46 
47 #include <sys/systm.h>
48 #include <sys/param.h>
49 #include <sys/socket.h>
50 #include <sys/isa_defs.h>
51 #include <net/if.h>
52 #include <net/if_dl.h>
53 #include <net/route.h>
54 #include <netinet/in.h>
55 #include <netinet/igmp_var.h>
56 #include <netinet/ip6.h>
57 #include <netinet/icmp6.h>
58 #include <netinet/in.h>
59 
60 #include <inet/common.h>
61 #include <inet/nd.h>
62 #include <inet/mib2.h>
63 #include <inet/ip.h>
64 #include <inet/ip6.h>
65 #include <inet/ip_multi.h>
66 #include <inet/ip_ire.h>
67 #include <inet/ip_rts.h>
68 #include <inet/ip_ndp.h>
69 #include <inet/ip_if.h>
70 #include <inet/ip6_asp.h>
71 #include <inet/tun.h>
72 #include <inet/ipclassifier.h>
73 #include <inet/sctp_ip.h>
74 
75 #include <sys/tsol/tndb.h>
76 #include <sys/tsol/tnet.h>
77 
78 static in6_addr_t	ipv6_ll_template =
79 			{(uint32_t)V6_LINKLOCAL, 0x0, 0x0, 0x0};
80 
81 static ipif_t *
82 ipif_lookup_interface_v6(const in6_addr_t *if_addr, const in6_addr_t *dst,
83     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst);
84 
85 /*
86  * These two functions, ipif_lookup_group_v6() and ill_lookup_group_v6(),
87  * are called when an application does not specify an interface to be
88  * used for multicast traffic.  It calls ire_lookup_multi_v6() to look
89  * for an interface route for the specified multicast group.  Doing
90  * this allows the administrator to add prefix routes for multicast to
91  * indicate which interface to be used for multicast traffic in the above
92  * scenario.  The route could be for all multicast (ff00::/8), for a single
93  * multicast group (a /128 route) or anything in between.  If there is no
94  * such multicast route, we just find any multicast capable interface and
95  * return it.
96  */
97 ipif_t *
98 ipif_lookup_group_v6(const in6_addr_t *group, zoneid_t zoneid, ip_stack_t *ipst)
99 {
100 	ire_t	*ire;
101 	ipif_t	*ipif;
102 
103 	ire = ire_lookup_multi_v6(group, zoneid, ipst);
104 	if (ire != NULL) {
105 		ipif = ire->ire_ipif;
106 		ipif_refhold(ipif);
107 		ire_refrele(ire);
108 		return (ipif);
109 	}
110 
111 	return (ipif_lookup_multicast(ipst, zoneid, B_TRUE));
112 }
113 
114 ill_t *
115 ill_lookup_group_v6(const in6_addr_t *group, zoneid_t zoneid, ip_stack_t *ipst)
116 {
117 	ire_t	*ire;
118 	ill_t	*ill;
119 	ipif_t	*ipif;
120 
121 	ire = ire_lookup_multi_v6(group, zoneid, ipst);
122 	if (ire != NULL) {
123 		ill = ire->ire_ipif->ipif_ill;
124 		ill_refhold(ill);
125 		ire_refrele(ire);
126 		return (ill);
127 	}
128 
129 	ipif = ipif_lookup_multicast(ipst, zoneid, B_TRUE);
130 	if (ipif == NULL)
131 		return (NULL);
132 
133 	ill = ipif->ipif_ill;
134 	ill_refhold(ill);
135 	ipif_refrele(ipif);
136 	return (ill);
137 }
138 
139 /*
140  * Look for an ipif with the specified interface address and destination.
141  * The destination address is used only for matching point-to-point interfaces.
142  */
143 static ipif_t *
144 ipif_lookup_interface_v6(const in6_addr_t *if_addr, const in6_addr_t *dst,
145     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst)
146 {
147 	ipif_t	*ipif;
148 	ill_t	*ill;
149 	ipsq_t	*ipsq;
150 	ill_walk_context_t ctx;
151 
152 	if (error != NULL)
153 		*error = 0;
154 
155 	/*
156 	 * First match all the point-to-point interfaces
157 	 * before looking at non-point-to-point interfaces.
158 	 * This is done to avoid returning non-point-to-point
159 	 * ipif instead of unnumbered point-to-point ipif.
160 	 */
161 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
162 	ill = ILL_START_WALK_V6(&ctx, ipst);
163 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
164 		GRAB_CONN_LOCK(q);
165 		mutex_enter(&ill->ill_lock);
166 		for (ipif = ill->ill_ipif; ipif != NULL;
167 		    ipif = ipif->ipif_next) {
168 			/* Allow the ipif to be down */
169 			if ((ipif->ipif_flags & IPIF_POINTOPOINT) &&
170 			    (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr,
171 			    if_addr)) &&
172 			    (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
173 			    dst))) {
174 				if (IPIF_CAN_LOOKUP(ipif)) {
175 					ipif_refhold_locked(ipif);
176 					mutex_exit(&ill->ill_lock);
177 					RELEASE_CONN_LOCK(q);
178 					rw_exit(&ipst->ips_ill_g_lock);
179 					return (ipif);
180 				} else if (IPIF_CAN_WAIT(ipif, q)) {
181 					ipsq = ill->ill_phyint->phyint_ipsq;
182 					mutex_enter(&ipsq->ipsq_lock);
183 					mutex_exit(&ill->ill_lock);
184 					rw_exit(&ipst->ips_ill_g_lock);
185 					ipsq_enq(ipsq, q, mp, func, NEW_OP,
186 					    ill);
187 					mutex_exit(&ipsq->ipsq_lock);
188 					RELEASE_CONN_LOCK(q);
189 					if (error != NULL)
190 						*error = EINPROGRESS;
191 					return (NULL);
192 				}
193 			}
194 		}
195 		mutex_exit(&ill->ill_lock);
196 		RELEASE_CONN_LOCK(q);
197 	}
198 	rw_exit(&ipst->ips_ill_g_lock);
199 	/* lookup the ipif based on interface address */
200 	ipif = ipif_lookup_addr_v6(if_addr, NULL, ALL_ZONES, q, mp, func,
201 	    error, ipst);
202 	ASSERT(ipif == NULL || ipif->ipif_isv6);
203 	return (ipif);
204 }
205 
206 /*
207  * Look for an ipif with the specified address. For point-point links
208  * we look for matches on either the destination address and the local
209  * address, but we ignore the check on the local address if IPIF_UNNUMBERED
210  * is set.
211  * Matches on a specific ill if match_ill is set.
212  */
213 /* ARGSUSED */
214 ipif_t *
215 ipif_lookup_addr_v6(const in6_addr_t *addr, ill_t *match_ill, zoneid_t zoneid,
216     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst)
217 {
218 	ipif_t	*ipif;
219 	ill_t	*ill;
220 	boolean_t  ptp = B_FALSE;
221 	ipsq_t	*ipsq;
222 	ill_walk_context_t ctx;
223 
224 	if (error != NULL)
225 		*error = 0;
226 
227 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
228 	/*
229 	 * Repeat twice, first based on local addresses and
230 	 * next time for pointopoint.
231 	 */
232 repeat:
233 	ill = ILL_START_WALK_V6(&ctx, ipst);
234 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
235 		if (match_ill != NULL && ill != match_ill) {
236 			continue;
237 		}
238 		GRAB_CONN_LOCK(q);
239 		mutex_enter(&ill->ill_lock);
240 		for (ipif = ill->ill_ipif; ipif != NULL;
241 		    ipif = ipif->ipif_next) {
242 			if (zoneid != ALL_ZONES &&
243 			    ipif->ipif_zoneid != zoneid &&
244 			    ipif->ipif_zoneid != ALL_ZONES)
245 				continue;
246 			/* Allow the ipif to be down */
247 			if ((!ptp && (IN6_ARE_ADDR_EQUAL(
248 			    &ipif->ipif_v6lcl_addr, addr) &&
249 			    (ipif->ipif_flags & IPIF_UNNUMBERED) == 0)) ||
250 			    (ptp && (ipif->ipif_flags & IPIF_POINTOPOINT) &&
251 			    IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
252 			    addr))) {
253 				if (IPIF_CAN_LOOKUP(ipif)) {
254 					ipif_refhold_locked(ipif);
255 					mutex_exit(&ill->ill_lock);
256 					RELEASE_CONN_LOCK(q);
257 					rw_exit(&ipst->ips_ill_g_lock);
258 					return (ipif);
259 				} else if (IPIF_CAN_WAIT(ipif, q)) {
260 					ipsq = ill->ill_phyint->phyint_ipsq;
261 					mutex_enter(&ipsq->ipsq_lock);
262 					mutex_exit(&ill->ill_lock);
263 					rw_exit(&ipst->ips_ill_g_lock);
264 					ipsq_enq(ipsq, q, mp, func, NEW_OP,
265 					    ill);
266 					mutex_exit(&ipsq->ipsq_lock);
267 					RELEASE_CONN_LOCK(q);
268 					if (error != NULL)
269 						*error = EINPROGRESS;
270 					return (NULL);
271 				}
272 			}
273 		}
274 		mutex_exit(&ill->ill_lock);
275 		RELEASE_CONN_LOCK(q);
276 	}
277 
278 	/* If we already did the ptp case, then we are done */
279 	if (ptp) {
280 		rw_exit(&ipst->ips_ill_g_lock);
281 		if (error != NULL)
282 			*error = ENXIO;
283 		return (NULL);
284 	}
285 	ptp = B_TRUE;
286 	goto repeat;
287 }
288 
289 /*
290  * Look for an ipif with the specified address. For point-point links
291  * we look for matches on either the destination address and the local
292  * address, but we ignore the check on the local address if IPIF_UNNUMBERED
293  * is set.
294  * Matches on a specific ill if match_ill is set.
295  * Return the zoneid for the ipif. ALL_ZONES if none found.
296  */
297 zoneid_t
298 ipif_lookup_addr_zoneid_v6(const in6_addr_t *addr, ill_t *match_ill,
299     ip_stack_t *ipst)
300 {
301 	ipif_t	*ipif;
302 	ill_t	*ill;
303 	boolean_t  ptp = B_FALSE;
304 	ill_walk_context_t ctx;
305 	zoneid_t	zoneid;
306 
307 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
308 	/*
309 	 * Repeat twice, first based on local addresses and
310 	 * next time for pointopoint.
311 	 */
312 repeat:
313 	ill = ILL_START_WALK_V6(&ctx, ipst);
314 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
315 		if (match_ill != NULL && ill != match_ill) {
316 			continue;
317 		}
318 		mutex_enter(&ill->ill_lock);
319 		for (ipif = ill->ill_ipif; ipif != NULL;
320 		    ipif = ipif->ipif_next) {
321 			/* Allow the ipif to be down */
322 			if ((!ptp && (IN6_ARE_ADDR_EQUAL(
323 			    &ipif->ipif_v6lcl_addr, addr) &&
324 			    (ipif->ipif_flags & IPIF_UNNUMBERED) == 0)) ||
325 			    (ptp && (ipif->ipif_flags & IPIF_POINTOPOINT) &&
326 			    IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
327 			    addr)) &&
328 			    !(ipif->ipif_state_flags & IPIF_CONDEMNED)) {
329 				zoneid = ipif->ipif_zoneid;
330 				mutex_exit(&ill->ill_lock);
331 				rw_exit(&ipst->ips_ill_g_lock);
332 				/*
333 				 * If ipif_zoneid was ALL_ZONES then we have
334 				 * a trusted extensions shared IP address.
335 				 * In that case GLOBAL_ZONEID works to send.
336 				 */
337 				if (zoneid == ALL_ZONES)
338 					zoneid = GLOBAL_ZONEID;
339 				return (zoneid);
340 			}
341 		}
342 		mutex_exit(&ill->ill_lock);
343 	}
344 
345 	/* If we already did the ptp case, then we are done */
346 	if (ptp) {
347 		rw_exit(&ipst->ips_ill_g_lock);
348 		return (ALL_ZONES);
349 	}
350 	ptp = B_TRUE;
351 	goto repeat;
352 }
353 
354 /*
355  * Perform various checks to verify that an address would make sense as a local
356  * interface address.  This is currently only called when an attempt is made
357  * to set a local address.
358  *
359  * Does not allow a v4-mapped address, an address that equals the subnet
360  * anycast address, ... a multicast address, ...
361  */
362 boolean_t
363 ip_local_addr_ok_v6(const in6_addr_t *addr, const in6_addr_t *subnet_mask)
364 {
365 	in6_addr_t subnet;
366 
367 	if (IN6_IS_ADDR_UNSPECIFIED(addr))
368 		return (B_TRUE);	/* Allow all zeros */
369 
370 	/*
371 	 * Don't allow all zeroes or host part, but allow
372 	 * all ones netmask.
373 	 */
374 	V6_MASK_COPY(*addr, *subnet_mask, subnet);
375 	if (IN6_IS_ADDR_V4MAPPED(addr) ||
376 	    (IN6_ARE_ADDR_EQUAL(addr, &subnet) &&
377 	    !IN6_ARE_ADDR_EQUAL(subnet_mask, &ipv6_all_ones)) ||
378 	    (IN6_IS_ADDR_V4COMPAT(addr) && CLASSD(V4_PART_OF_V6((*addr)))) ||
379 	    IN6_IS_ADDR_MULTICAST(addr))
380 		return (B_FALSE);
381 
382 	return (B_TRUE);
383 }
384 
385 /*
386  * Perform various checks to verify that an address would make sense as a
387  * remote/subnet interface address.
388  */
389 boolean_t
390 ip_remote_addr_ok_v6(const in6_addr_t *addr, const in6_addr_t *subnet_mask)
391 {
392 	in6_addr_t subnet;
393 
394 	if (IN6_IS_ADDR_UNSPECIFIED(addr))
395 		return (B_TRUE);	/* Allow all zeros */
396 
397 	V6_MASK_COPY(*addr, *subnet_mask, subnet);
398 	if (IN6_IS_ADDR_V4MAPPED(addr) ||
399 	    (IN6_ARE_ADDR_EQUAL(addr, &subnet) &&
400 	    !IN6_ARE_ADDR_EQUAL(subnet_mask, &ipv6_all_ones)) ||
401 	    IN6_IS_ADDR_MULTICAST(addr) ||
402 	    (IN6_IS_ADDR_V4COMPAT(addr) && CLASSD(V4_PART_OF_V6((*addr)))))
403 		return (B_FALSE);
404 
405 	return (B_TRUE);
406 }
407 
408 /*
409  * ip_rt_add_v6 is called to add an IPv6 route to the forwarding table.
410  * ipif_arg is passed in to associate it with the correct interface
411  * (for link-local destinations and gateways).
412  */
413 /* ARGSUSED1 */
414 int
415 ip_rt_add_v6(const in6_addr_t *dst_addr, const in6_addr_t *mask,
416     const in6_addr_t *gw_addr, const in6_addr_t *src_addr, int flags,
417     ipif_t *ipif_arg, ire_t **ire_arg, queue_t *q, mblk_t *mp, ipsq_func_t func,
418     struct rtsa_s *sp, ip_stack_t *ipst)
419 {
420 	ire_t	*ire;
421 	ire_t	*gw_ire = NULL;
422 	ipif_t	*ipif;
423 	boolean_t ipif_refheld = B_FALSE;
424 	uint_t	type;
425 	int	match_flags = MATCH_IRE_TYPE;
426 	int	error;
427 	tsol_gc_t *gc = NULL;
428 	tsol_gcgrp_t *gcgrp = NULL;
429 	boolean_t gcgrp_xtraref = B_FALSE;
430 
431 	if (ire_arg != NULL)
432 		*ire_arg = NULL;
433 
434 	/*
435 	 * Prevent routes with a zero gateway from being created (since
436 	 * interfaces can currently be plumbed and brought up with no assigned
437 	 * address).
438 	 */
439 	if (IN6_IS_ADDR_UNSPECIFIED(gw_addr))
440 		return (ENETUNREACH);
441 
442 	/*
443 	 * If this is the case of RTF_HOST being set, then we set the netmask
444 	 * to all ones (regardless if one was supplied).
445 	 */
446 	if (flags & RTF_HOST)
447 		mask = &ipv6_all_ones;
448 
449 	/*
450 	 * Get the ipif, if any, corresponding to the gw_addr
451 	 */
452 	ipif = ipif_lookup_interface_v6(gw_addr, dst_addr, q, mp, func,
453 	    &error, ipst);
454 	if (ipif != NULL)
455 		ipif_refheld = B_TRUE;
456 	else if (error == EINPROGRESS) {
457 		ip1dbg(("ip_rt_add_v6: null and EINPROGRESS"));
458 		return (error);
459 	}
460 
461 	/*
462 	 * GateD will attempt to create routes with a loopback interface
463 	 * address as the gateway and with RTF_GATEWAY set.  We allow
464 	 * these routes to be added, but create them as interface routes
465 	 * since the gateway is an interface address.
466 	 */
467 	if ((ipif != NULL) && (ipif->ipif_ire_type == IRE_LOOPBACK)) {
468 		flags &= ~RTF_GATEWAY;
469 		if (IN6_ARE_ADDR_EQUAL(gw_addr, &ipv6_loopback) &&
470 		    IN6_ARE_ADDR_EQUAL(dst_addr, &ipv6_loopback) &&
471 		    IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones)) {
472 			ire = ire_ctable_lookup_v6(dst_addr, 0, IRE_LOOPBACK,
473 			    ipif, ALL_ZONES, NULL, match_flags, ipst);
474 			if (ire != NULL) {
475 				ire_refrele(ire);
476 				if (ipif_refheld)
477 					ipif_refrele(ipif);
478 				return (EEXIST);
479 			}
480 			ip1dbg(("ipif_up_done: 0x%p creating IRE 0x%x"
481 			    "for 0x%x\n", (void *)ipif,
482 			    ipif->ipif_ire_type,
483 			    ntohl(ipif->ipif_lcl_addr)));
484 			ire = ire_create_v6(
485 			    dst_addr,
486 			    mask,
487 			    &ipif->ipif_v6src_addr,
488 			    NULL,
489 			    &ipif->ipif_mtu,
490 			    NULL,
491 			    NULL,
492 			    NULL,
493 			    ipif->ipif_net_type,
494 			    ipif,
495 			    NULL,
496 			    0,
497 			    0,
498 			    flags,
499 			    &ire_uinfo_null,
500 			    NULL,
501 			    NULL,
502 			    ipst);
503 			if (ire == NULL) {
504 				if (ipif_refheld)
505 					ipif_refrele(ipif);
506 				return (ENOMEM);
507 			}
508 			error = ire_add(&ire, q, mp, func, B_FALSE);
509 			if (error == 0)
510 				goto save_ire;
511 			/*
512 			 * In the result of failure, ire_add() will have already
513 			 * deleted the ire in question, so there is no need to
514 			 * do that here.
515 			 */
516 			if (ipif_refheld)
517 				ipif_refrele(ipif);
518 			return (error);
519 		}
520 	}
521 
522 	/*
523 	 * Traditionally, interface routes are ones where RTF_GATEWAY isn't set
524 	 * and the gateway address provided is one of the system's interface
525 	 * addresses.  By using the routing socket interface and supplying an
526 	 * RTA_IFP sockaddr with an interface index, an alternate method of
527 	 * specifying an interface route to be created is available which uses
528 	 * the interface index that specifies the outgoing interface rather than
529 	 * the address of an outgoing interface (which may not be able to
530 	 * uniquely identify an interface).  When coupled with the RTF_GATEWAY
531 	 * flag, routes can be specified which not only specify the next-hop to
532 	 * be used when routing to a certain prefix, but also which outgoing
533 	 * interface should be used.
534 	 *
535 	 * Previously, interfaces would have unique addresses assigned to them
536 	 * and so the address assigned to a particular interface could be used
537 	 * to identify a particular interface.  One exception to this was the
538 	 * case of an unnumbered interface (where IPIF_UNNUMBERED was set).
539 	 *
540 	 * With the advent of IPv6 and its link-local addresses, this
541 	 * restriction was relaxed and interfaces could share addresses between
542 	 * themselves.  In fact, typically all of the link-local interfaces on
543 	 * an IPv6 node or router will have the same link-local address.  In
544 	 * order to differentiate between these interfaces, the use of an
545 	 * interface index is necessary and this index can be carried inside a
546 	 * RTA_IFP sockaddr (which is actually a sockaddr_dl).  One restriction
547 	 * of using the interface index, however, is that all of the ipif's that
548 	 * are part of an ill have the same index and so the RTA_IFP sockaddr
549 	 * cannot be used to differentiate between ipif's (or logical
550 	 * interfaces) that belong to the same ill (physical interface).
551 	 *
552 	 * For example, in the following case involving IPv4 interfaces and
553 	 * logical interfaces
554 	 *
555 	 *	192.0.2.32	255.255.255.224	192.0.2.33	U	if0
556 	 *	192.0.2.32	255.255.255.224	192.0.2.34	U	if0:1
557 	 *	192.0.2.32	255.255.255.224	192.0.2.35	U	if0:2
558 	 *
559 	 * the ipif's corresponding to each of these interface routes can be
560 	 * uniquely identified by the "gateway" (actually interface address).
561 	 *
562 	 * In this case involving multiple IPv6 default routes to a particular
563 	 * link-local gateway, the use of RTA_IFP is necessary to specify which
564 	 * default route is of interest:
565 	 *
566 	 *	default		fe80::123:4567:89ab:cdef	U	if0
567 	 *	default		fe80::123:4567:89ab:cdef	U	if1
568 	 */
569 
570 	/* RTF_GATEWAY not set */
571 	if (!(flags & RTF_GATEWAY)) {
572 		queue_t	*stq;
573 
574 		if (sp != NULL) {
575 			ip2dbg(("ip_rt_add_v6: gateway security attributes "
576 			    "cannot be set with interface route\n"));
577 			if (ipif_refheld)
578 				ipif_refrele(ipif);
579 			return (EINVAL);
580 		}
581 
582 		/*
583 		 * As the interface index specified with the RTA_IFP sockaddr is
584 		 * the same for all ipif's off of an ill, the matching logic
585 		 * below uses MATCH_IRE_ILL if such an index was specified.
586 		 * This means that routes sharing the same prefix when added
587 		 * using a RTA_IFP sockaddr must have distinct interface
588 		 * indices (namely, they must be on distinct ill's).
589 		 *
590 		 * On the other hand, since the gateway address will usually be
591 		 * different for each ipif on the system, the matching logic
592 		 * uses MATCH_IRE_IPIF in the case of a traditional interface
593 		 * route.  This means that interface routes for the same prefix
594 		 * can be created if they belong to distinct ipif's and if a
595 		 * RTA_IFP sockaddr is not present.
596 		 */
597 		if (ipif_arg != NULL) {
598 			if (ipif_refheld) {
599 				ipif_refrele(ipif);
600 				ipif_refheld = B_FALSE;
601 			}
602 			ipif = ipif_arg;
603 			match_flags |= MATCH_IRE_ILL;
604 		} else {
605 			/*
606 			 * Check the ipif corresponding to the gw_addr
607 			 */
608 			if (ipif == NULL)
609 				return (ENETUNREACH);
610 			match_flags |= MATCH_IRE_IPIF;
611 		}
612 
613 		ASSERT(ipif != NULL);
614 		/*
615 		 * We check for an existing entry at this point.
616 		 */
617 		match_flags |= MATCH_IRE_MASK;
618 		ire = ire_ftable_lookup_v6(dst_addr, mask, 0, IRE_INTERFACE,
619 		    ipif, NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
620 		if (ire != NULL) {
621 			ire_refrele(ire);
622 			if (ipif_refheld)
623 				ipif_refrele(ipif);
624 			return (EEXIST);
625 		}
626 
627 		stq = (ipif->ipif_net_type == IRE_IF_RESOLVER)
628 		    ? ipif->ipif_rq : ipif->ipif_wq;
629 
630 		/*
631 		 * Create a copy of the IRE_LOOPBACK, IRE_IF_NORESOLVER or
632 		 * IRE_IF_RESOLVER with the modified address and netmask.
633 		 */
634 		ire = ire_create_v6(
635 		    dst_addr,
636 		    mask,
637 		    &ipif->ipif_v6src_addr,
638 		    NULL,
639 		    &ipif->ipif_mtu,
640 		    NULL,
641 		    NULL,
642 		    stq,
643 		    ipif->ipif_net_type,
644 		    ipif,
645 		    NULL,
646 		    0,
647 		    0,
648 		    flags,
649 		    &ire_uinfo_null,
650 		    NULL,
651 		    NULL,
652 		    ipst);
653 		if (ire == NULL) {
654 			if (ipif_refheld)
655 				ipif_refrele(ipif);
656 			return (ENOMEM);
657 		}
658 
659 		/*
660 		 * Some software (for example, GateD and Sun Cluster) attempts
661 		 * to create (what amount to) IRE_PREFIX routes with the
662 		 * loopback address as the gateway.  This is primarily done to
663 		 * set up prefixes with the RTF_REJECT flag set (for example,
664 		 * when generating aggregate routes). We also OR in the
665 		 * RTF_BLACKHOLE flag as these interface routes, by
666 		 * definition, can only be that.
667 		 *
668 		 * If the IRE type (as defined by ipif->ipif_net_type) is
669 		 * IRE_LOOPBACK, then we map the request into a
670 		 * IRE_IF_NORESOLVER.
671 		 *
672 		 * Needless to say, the real IRE_LOOPBACK is NOT created by this
673 		 * routine, but rather using ire_create_v6() directly.
674 		 */
675 		if (ipif->ipif_net_type == IRE_LOOPBACK) {
676 			ire->ire_type = IRE_IF_NORESOLVER;
677 			ire->ire_flags |= RTF_BLACKHOLE;
678 		}
679 		error = ire_add(&ire, q, mp, func, B_FALSE);
680 		if (error == 0)
681 			goto save_ire;
682 		/*
683 		 * In the result of failure, ire_add() will have already
684 		 * deleted the ire in question, so there is no need to
685 		 * do that here.
686 		 */
687 		if (ipif_refheld)
688 			ipif_refrele(ipif);
689 		return (error);
690 	}
691 	if (ipif_refheld) {
692 		ipif_refrele(ipif);
693 		ipif_refheld = B_FALSE;
694 	}
695 
696 	/*
697 	 * Get an interface IRE for the specified gateway.
698 	 * If we don't have an IRE_IF_NORESOLVER or IRE_IF_RESOLVER for the
699 	 * gateway, it is currently unreachable and we fail the request
700 	 * accordingly.
701 	 */
702 	ipif = ipif_arg;
703 	if (ipif_arg != NULL)
704 		match_flags |= MATCH_IRE_ILL;
705 	gw_ire = ire_ftable_lookup_v6(gw_addr, 0, 0, IRE_INTERFACE, ipif_arg,
706 	    NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
707 	if (gw_ire == NULL)
708 		return (ENETUNREACH);
709 
710 	/*
711 	 * We create one of three types of IREs as a result of this request
712 	 * based on the netmask.  A netmask of all ones (which is automatically
713 	 * assumed when RTF_HOST is set) results in an IRE_HOST being created.
714 	 * An all zeroes netmask implies a default route so an IRE_DEFAULT is
715 	 * created.  Otherwise, an IRE_PREFIX route is created for the
716 	 * destination prefix.
717 	 */
718 	if (IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones))
719 		type = IRE_HOST;
720 	else if (IN6_IS_ADDR_UNSPECIFIED(mask))
721 		type = IRE_DEFAULT;
722 	else
723 		type = IRE_PREFIX;
724 
725 	/* check for a duplicate entry */
726 	ire = ire_ftable_lookup_v6(dst_addr, mask, gw_addr, type, ipif_arg,
727 	    NULL, ALL_ZONES, 0, NULL,
728 	    match_flags | MATCH_IRE_MASK | MATCH_IRE_GW, ipst);
729 	if (ire != NULL) {
730 		ire_refrele(gw_ire);
731 		ire_refrele(ire);
732 		return (EEXIST);
733 	}
734 
735 	/* Security attribute exists */
736 	if (sp != NULL) {
737 		tsol_gcgrp_addr_t ga;
738 
739 		/* find or create the gateway credentials group */
740 		ga.ga_af = AF_INET6;
741 		ga.ga_addr = *gw_addr;
742 
743 		/* we hold reference to it upon success */
744 		gcgrp = gcgrp_lookup(&ga, B_TRUE);
745 		if (gcgrp == NULL) {
746 			ire_refrele(gw_ire);
747 			return (ENOMEM);
748 		}
749 
750 		/*
751 		 * Create and add the security attribute to the group; a
752 		 * reference to the group is made upon allocating a new
753 		 * entry successfully.  If it finds an already-existing
754 		 * entry for the security attribute in the group, it simply
755 		 * returns it and no new reference is made to the group.
756 		 */
757 		gc = gc_create(sp, gcgrp, &gcgrp_xtraref);
758 		if (gc == NULL) {
759 			/* release reference held by gcgrp_lookup */
760 			GCGRP_REFRELE(gcgrp);
761 			ire_refrele(gw_ire);
762 			return (ENOMEM);
763 		}
764 	}
765 
766 	/* Create the IRE. */
767 	ire = ire_create_v6(
768 	    dst_addr,				/* dest address */
769 	    mask,				/* mask */
770 	    /* src address assigned by the caller? */
771 	    (((flags & RTF_SETSRC) && !IN6_IS_ADDR_UNSPECIFIED(src_addr)) ?
772 	    src_addr : NULL),
773 	    gw_addr,				/* gateway address */
774 	    &gw_ire->ire_max_frag,
775 	    NULL,				/* no src nce */
776 	    NULL,				/* no recv-from queue */
777 	    NULL,				/* no send-to queue */
778 	    (ushort_t)type,			/* IRE type */
779 	    ipif_arg,
780 	    NULL,
781 	    0,
782 	    0,
783 	    flags,
784 	    &gw_ire->ire_uinfo,			/* Inherit ULP info from gw */
785 	    gc,					/* security attribute */
786 	    NULL,
787 	    ipst);
788 
789 	/*
790 	 * The ire holds a reference to the 'gc' and the 'gc' holds a
791 	 * reference to the 'gcgrp'. We can now release the extra reference
792 	 * the 'gcgrp' acquired in the gcgrp_lookup, if it was not used.
793 	 */
794 	if (gcgrp_xtraref)
795 		GCGRP_REFRELE(gcgrp);
796 	if (ire == NULL) {
797 		if (gc != NULL)
798 			GC_REFRELE(gc);
799 		ire_refrele(gw_ire);
800 		return (ENOMEM);
801 	}
802 
803 	/*
804 	 * POLICY: should we allow an RTF_HOST with address INADDR_ANY?
805 	 * SUN/OS socket stuff does but do we really want to allow ::0 ?
806 	 */
807 
808 	/* Add the new IRE. */
809 	error = ire_add(&ire, q, mp, func, B_FALSE);
810 	/*
811 	 * In the result of failure, ire_add() will have already
812 	 * deleted the ire in question, so there is no need to
813 	 * do that here.
814 	 */
815 	if (error != 0) {
816 		ire_refrele(gw_ire);
817 		return (error);
818 	}
819 
820 	if (flags & RTF_MULTIRT) {
821 		/*
822 		 * Invoke the CGTP (multirouting) filtering module
823 		 * to add the dst address in the filtering database.
824 		 * Replicated inbound packets coming from that address
825 		 * will be filtered to discard the duplicates.
826 		 * It is not necessary to call the CGTP filter hook
827 		 * when the dst address is a multicast, because an
828 		 * IP source address cannot be a multicast.
829 		 */
830 		if (ipst->ips_ip_cgtp_filter_ops != NULL &&
831 		    !IN6_IS_ADDR_MULTICAST(&(ire->ire_addr_v6))) {
832 			int res;
833 
834 			res = ipst->ips_ip_cgtp_filter_ops->cfo_add_dest_v6(
835 			    ipst->ips_netstack->netstack_stackid,
836 			    &ire->ire_addr_v6,
837 			    &ire->ire_gateway_addr_v6,
838 			    &ire->ire_src_addr_v6,
839 			    &gw_ire->ire_src_addr_v6);
840 			if (res != 0) {
841 				ire_refrele(gw_ire);
842 				ire_delete(ire);
843 				return (res);
844 			}
845 		}
846 	}
847 
848 	/*
849 	 * Now that the prefix IRE entry has been created, delete any
850 	 * existing gateway IRE cache entries as well as any IRE caches
851 	 * using the gateway, and force them to be created through
852 	 * ip_newroute_v6.
853 	 */
854 	if (gc != NULL) {
855 		ASSERT(gcgrp != NULL);
856 		ire_clookup_delete_cache_gw_v6(gw_addr, ALL_ZONES, ipst);
857 	}
858 
859 save_ire:
860 	if (gw_ire != NULL) {
861 		ire_refrele(gw_ire);
862 	}
863 	if (ipif != NULL) {
864 		mblk_t	*save_mp;
865 
866 		/*
867 		 * Save enough information so that we can recreate the IRE if
868 		 * the interface goes down and then up.  The metrics associated
869 		 * with the route will be saved as well when rts_setmetrics() is
870 		 * called after the IRE has been created.  In the case where
871 		 * memory cannot be allocated, none of this information will be
872 		 * saved.
873 		 */
874 		save_mp = allocb(sizeof (ifrt_t), BPRI_MED);
875 		if (save_mp != NULL) {
876 			ifrt_t	*ifrt;
877 
878 			save_mp->b_wptr += sizeof (ifrt_t);
879 			ifrt = (ifrt_t *)save_mp->b_rptr;
880 			bzero(ifrt, sizeof (ifrt_t));
881 			ifrt->ifrt_type = ire->ire_type;
882 			ifrt->ifrt_v6addr = ire->ire_addr_v6;
883 			mutex_enter(&ire->ire_lock);
884 			ifrt->ifrt_v6gateway_addr = ire->ire_gateway_addr_v6;
885 			ifrt->ifrt_v6src_addr = ire->ire_src_addr_v6;
886 			mutex_exit(&ire->ire_lock);
887 			ifrt->ifrt_v6mask = ire->ire_mask_v6;
888 			ifrt->ifrt_flags = ire->ire_flags;
889 			ifrt->ifrt_max_frag = ire->ire_max_frag;
890 			mutex_enter(&ipif->ipif_saved_ire_lock);
891 			save_mp->b_cont = ipif->ipif_saved_ire_mp;
892 			ipif->ipif_saved_ire_mp = save_mp;
893 			ipif->ipif_saved_ire_cnt++;
894 			mutex_exit(&ipif->ipif_saved_ire_lock);
895 		}
896 	}
897 	if (ire_arg != NULL) {
898 		/*
899 		 * Store the ire that was successfully added into where ire_arg
900 		 * points to so that callers don't have to look it up
901 		 * themselves (but they are responsible for ire_refrele()ing
902 		 * the ire when they are finished with it).
903 		 */
904 		*ire_arg = ire;
905 	} else {
906 		ire_refrele(ire);		/* Held in ire_add */
907 	}
908 	if (ipif_refheld)
909 		ipif_refrele(ipif);
910 	return (0);
911 }
912 
913 /*
914  * ip_rt_delete_v6 is called to delete an IPv6 route.
915  * ipif_arg is passed in to associate it with the correct interface
916  * (for link-local destinations and gateways).
917  */
918 /* ARGSUSED4 */
919 int
920 ip_rt_delete_v6(const in6_addr_t *dst_addr, const in6_addr_t *mask,
921     const in6_addr_t *gw_addr, uint_t rtm_addrs, int flags, ipif_t *ipif_arg,
922     queue_t *q, mblk_t *mp, ipsq_func_t func, ip_stack_t *ipst)
923 {
924 	ire_t	*ire = NULL;
925 	ipif_t	*ipif;
926 	uint_t	type;
927 	uint_t	match_flags = MATCH_IRE_TYPE;
928 	int	err = 0;
929 	boolean_t	ipif_refheld = B_FALSE;
930 
931 	/*
932 	 * If this is the case of RTF_HOST being set, then we set the netmask
933 	 * to all ones.  Otherwise, we use the netmask if one was supplied.
934 	 */
935 	if (flags & RTF_HOST) {
936 		mask = &ipv6_all_ones;
937 		match_flags |= MATCH_IRE_MASK;
938 	} else if (rtm_addrs & RTA_NETMASK) {
939 		match_flags |= MATCH_IRE_MASK;
940 	}
941 
942 	/*
943 	 * Note that RTF_GATEWAY is never set on a delete, therefore
944 	 * we check if the gateway address is one of our interfaces first,
945 	 * and fall back on RTF_GATEWAY routes.
946 	 *
947 	 * This makes it possible to delete an original
948 	 * IRE_IF_NORESOLVER/IRE_IF_RESOLVER - consistent with SunOS 4.1.
949 	 *
950 	 * As the interface index specified with the RTA_IFP sockaddr is the
951 	 * same for all ipif's off of an ill, the matching logic below uses
952 	 * MATCH_IRE_ILL if such an index was specified.  This means a route
953 	 * sharing the same prefix and interface index as the the route
954 	 * intended to be deleted might be deleted instead if a RTA_IFP sockaddr
955 	 * is specified in the request.
956 	 *
957 	 * On the other hand, since the gateway address will usually be
958 	 * different for each ipif on the system, the matching logic
959 	 * uses MATCH_IRE_IPIF in the case of a traditional interface
960 	 * route.  This means that interface routes for the same prefix can be
961 	 * uniquely identified if they belong to distinct ipif's and if a
962 	 * RTA_IFP sockaddr is not present.
963 	 *
964 	 * For more detail on specifying routes by gateway address and by
965 	 * interface index, see the comments in ip_rt_add_v6().
966 	 */
967 	ipif = ipif_lookup_interface_v6(gw_addr, dst_addr, q, mp, func, &err,
968 	    ipst);
969 	if (ipif != NULL) {
970 		ipif_refheld = B_TRUE;
971 		if (ipif_arg != NULL) {
972 			ipif_refrele(ipif);
973 			ipif_refheld = B_FALSE;
974 			ipif = ipif_arg;
975 			match_flags |= MATCH_IRE_ILL;
976 		} else {
977 			match_flags |= MATCH_IRE_IPIF;
978 		}
979 
980 		if (ipif->ipif_ire_type == IRE_LOOPBACK)
981 			ire = ire_ctable_lookup_v6(dst_addr, 0, IRE_LOOPBACK,
982 			    ipif, ALL_ZONES, NULL, match_flags, ipst);
983 		if (ire == NULL)
984 			ire = ire_ftable_lookup_v6(dst_addr, mask, 0,
985 			    IRE_INTERFACE, ipif, NULL, ALL_ZONES, 0, NULL,
986 			    match_flags, ipst);
987 	} else if (err == EINPROGRESS) {
988 		return (err);
989 	} else {
990 		err = 0;
991 	}
992 	if (ire == NULL) {
993 		/*
994 		 * At this point, the gateway address is not one of our own
995 		 * addresses or a matching interface route was not found.  We
996 		 * set the IRE type to lookup based on whether
997 		 * this is a host route, a default route or just a prefix.
998 		 *
999 		 * If an ipif_arg was passed in, then the lookup is based on an
1000 		 * interface index so MATCH_IRE_ILL is added to match_flags.
1001 		 * In any case, MATCH_IRE_IPIF is cleared and MATCH_IRE_GW is
1002 		 * set as the route being looked up is not a traditional
1003 		 * interface route.
1004 		 */
1005 		match_flags &= ~MATCH_IRE_IPIF;
1006 		match_flags |= MATCH_IRE_GW;
1007 		if (ipif_arg != NULL)
1008 			match_flags |= MATCH_IRE_ILL;
1009 		if (IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones))
1010 			type = IRE_HOST;
1011 		else if (IN6_IS_ADDR_UNSPECIFIED(mask))
1012 			type = IRE_DEFAULT;
1013 		else
1014 			type = IRE_PREFIX;
1015 		ire = ire_ftable_lookup_v6(dst_addr, mask, gw_addr, type,
1016 		    ipif_arg, NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
1017 	}
1018 
1019 	if (ipif_refheld) {
1020 		ipif_refrele(ipif);
1021 		ipif_refheld = B_FALSE;
1022 	}
1023 	if (ire == NULL)
1024 		return (ESRCH);
1025 
1026 	if (ire->ire_flags & RTF_MULTIRT) {
1027 		/*
1028 		 * Invoke the CGTP (multirouting) filtering module
1029 		 * to remove the dst address from the filtering database.
1030 		 * Packets coming from that address will no longer be
1031 		 * filtered to remove duplicates.
1032 		 */
1033 		if (ipst->ips_ip_cgtp_filter_ops != NULL) {
1034 			err = ipst->ips_ip_cgtp_filter_ops->cfo_del_dest_v6(
1035 			    ipst->ips_netstack->netstack_stackid,
1036 			    &ire->ire_addr_v6, &ire->ire_gateway_addr_v6);
1037 		}
1038 	}
1039 
1040 	ipif = ire->ire_ipif;
1041 	if (ipif != NULL) {
1042 		mblk_t		**mpp;
1043 		mblk_t		*mp;
1044 		ifrt_t		*ifrt;
1045 		in6_addr_t	gw_addr_v6;
1046 
1047 		/* Remove from ipif_saved_ire_mp list if it is there */
1048 		mutex_enter(&ire->ire_lock);
1049 		gw_addr_v6 = ire->ire_gateway_addr_v6;
1050 		mutex_exit(&ire->ire_lock);
1051 		mutex_enter(&ipif->ipif_saved_ire_lock);
1052 		for (mpp = &ipif->ipif_saved_ire_mp; *mpp != NULL;
1053 		    mpp = &(*mpp)->b_cont) {
1054 			/*
1055 			 * On a given ipif, the triple of address, gateway and
1056 			 * mask is unique for each saved IRE (in the case of
1057 			 * ordinary interface routes, the gateway address is
1058 			 * all-zeroes).
1059 			 */
1060 			mp = *mpp;
1061 			ifrt = (ifrt_t *)mp->b_rptr;
1062 			if (IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6addr,
1063 			    &ire->ire_addr_v6) &&
1064 			    IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6gateway_addr,
1065 			    &gw_addr_v6) &&
1066 			    IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6mask,
1067 			    &ire->ire_mask_v6)) {
1068 				*mpp = mp->b_cont;
1069 				ipif->ipif_saved_ire_cnt--;
1070 				freeb(mp);
1071 				break;
1072 			}
1073 		}
1074 		mutex_exit(&ipif->ipif_saved_ire_lock);
1075 	}
1076 	ire_delete(ire);
1077 	ire_refrele(ire);
1078 	return (err);
1079 }
1080 
1081 /*
1082  * Derive a token from the link layer address.
1083  */
1084 boolean_t
1085 ill_setdefaulttoken(ill_t *ill)
1086 {
1087 	int 		i;
1088 	in6_addr_t	v6addr, v6mask;
1089 
1090 	if (!MEDIA_V6INTFID(ill->ill_media, ill->ill_phys_addr_length,
1091 	    ill->ill_phys_addr, &v6addr))
1092 		return (B_FALSE);
1093 
1094 	(void) ip_plen_to_mask_v6(IPV6_TOKEN_LEN, &v6mask);
1095 
1096 	for (i = 0; i < 4; i++)
1097 		v6mask.s6_addr32[i] = v6mask.s6_addr32[i] ^
1098 		    (uint32_t)0xffffffff;
1099 
1100 	V6_MASK_COPY(v6addr, v6mask, ill->ill_token);
1101 	ill->ill_token_length = IPV6_TOKEN_LEN;
1102 	return (B_TRUE);
1103 }
1104 
1105 /*
1106  * Create a link-local address from a token.
1107  */
1108 static void
1109 ipif_get_linklocal(in6_addr_t *dest, const in6_addr_t *token)
1110 {
1111 	int i;
1112 
1113 	for (i = 0; i < 4; i++) {
1114 		dest->s6_addr32[i] =
1115 		    token->s6_addr32[i] | ipv6_ll_template.s6_addr32[i];
1116 	}
1117 }
1118 
1119 /*
1120  * Set a nice default address for either automatic tunnels tsrc/96 or
1121  * 6to4 tunnels 2002:<tsrc>::1/64
1122  */
1123 static void
1124 ipif_set_tun_auto_addr(ipif_t *ipif, struct iftun_req *ta)
1125 {
1126 	sin6_t	sin6;
1127 	sin_t	*sin;
1128 	ill_t 	*ill = ipif->ipif_ill;
1129 	tun_t *tp = (tun_t *)ill->ill_wq->q_next->q_ptr;
1130 
1131 	if (ta->ifta_saddr.ss_family != AF_INET ||
1132 	    (ipif->ipif_flags & IPIF_UP) || !ipif->ipif_isv6 ||
1133 	    (ta->ifta_flags & IFTUN_SRC) == 0)
1134 		return;
1135 
1136 	/*
1137 	 * Check the tunnel type by examining q_next->q_ptr
1138 	 */
1139 	if (tp->tun_flags & TUN_AUTOMATIC) {
1140 		/* this is an automatic tunnel */
1141 		(void) ip_plen_to_mask_v6(IPV6_ABITS - IP_ABITS,
1142 		    &ipif->ipif_v6net_mask);
1143 		bzero(&sin6, sizeof (sin6_t));
1144 		sin = (sin_t *)&ta->ifta_saddr;
1145 		V4_PART_OF_V6(sin6.sin6_addr) = sin->sin_addr.s_addr;
1146 		sin6.sin6_family = AF_INET6;
1147 		(void) ip_sioctl_addr(ipif, (sin_t *)&sin6,
1148 		    NULL, NULL, NULL, NULL);
1149 	} else if (tp->tun_flags & TUN_6TO4) {
1150 		/* this is a 6to4 tunnel */
1151 		(void) ip_plen_to_mask_v6(IPV6_PREFIX_LEN,
1152 		    &ipif->ipif_v6net_mask);
1153 		sin = (sin_t *)&ta->ifta_saddr;
1154 		/* create a 6to4 address from the IPv4 tsrc */
1155 		IN6_V4ADDR_TO_6TO4(&sin->sin_addr, &sin6.sin6_addr);
1156 		sin6.sin6_family = AF_INET6;
1157 		(void) ip_sioctl_addr(ipif, (sin_t *)&sin6,
1158 		    NULL, NULL, NULL, NULL);
1159 	} else {
1160 		ip1dbg(("ipif_set_tun_auto_addr: Unknown tunnel type"));
1161 		return;
1162 	}
1163 }
1164 
1165 /*
1166  * Set link local for ipif_id 0 of a configured tunnel based on the
1167  * tsrc or tdst parameter
1168  * For tunnels over IPv4 use the IPv4 address prepended with 32 zeros as
1169  * the token.
1170  * For tunnels over IPv6 use the low-order 64 bits of the "inner" IPv6 address
1171  * as the token for the "outer" link.
1172  */
1173 void
1174 ipif_set_tun_llink(ill_t *ill, struct iftun_req *ta)
1175 {
1176 	ipif_t		*ipif;
1177 	sin_t		*sin;
1178 	in6_addr_t	*s6addr;
1179 
1180 	ASSERT(IAM_WRITER_ILL(ill));
1181 
1182 	/* The first ipif must be id zero. */
1183 	ipif = ill->ill_ipif;
1184 	ASSERT(ipif->ipif_id == 0);
1185 
1186 	/* no link local for automatic tunnels */
1187 	if (!(ipif->ipif_flags & IPIF_POINTOPOINT)) {
1188 		ipif_set_tun_auto_addr(ipif, ta);
1189 		return;
1190 	}
1191 
1192 	if ((ta->ifta_flags & IFTUN_DST) &&
1193 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6pp_dst_addr)) {
1194 		sin6_t	sin6;
1195 
1196 		ASSERT(!(ipif->ipif_flags & IPIF_UP));
1197 		bzero(&sin6, sizeof (sin6_t));
1198 		if ((ta->ifta_saddr.ss_family == AF_INET)) {
1199 			sin = (sin_t *)&ta->ifta_daddr;
1200 			V4_PART_OF_V6(sin6.sin6_addr) =
1201 			    sin->sin_addr.s_addr;
1202 		} else {
1203 			s6addr =
1204 			    &((sin6_t *)&ta->ifta_daddr)->sin6_addr;
1205 			sin6.sin6_addr.s6_addr32[3] = s6addr->s6_addr32[3];
1206 			sin6.sin6_addr.s6_addr32[2] = s6addr->s6_addr32[2];
1207 		}
1208 		ipif_get_linklocal(&ipif->ipif_v6pp_dst_addr,
1209 		    &sin6.sin6_addr);
1210 		ipif->ipif_v6subnet = ipif->ipif_v6pp_dst_addr;
1211 	}
1212 	if ((ta->ifta_flags & IFTUN_SRC)) {
1213 		ASSERT(!(ipif->ipif_flags & IPIF_UP));
1214 
1215 		/* Set the token if it isn't already set */
1216 		if (IN6_IS_ADDR_UNSPECIFIED(&ill->ill_token)) {
1217 			if ((ta->ifta_saddr.ss_family == AF_INET)) {
1218 				sin = (sin_t *)&ta->ifta_saddr;
1219 				V4_PART_OF_V6(ill->ill_token) =
1220 				    sin->sin_addr.s_addr;
1221 			} else {
1222 				s6addr =
1223 				    &((sin6_t *)&ta->ifta_saddr)->sin6_addr;
1224 				ill->ill_token.s6_addr32[3] =
1225 				    s6addr->s6_addr32[3];
1226 				ill->ill_token.s6_addr32[2] =
1227 				    s6addr->s6_addr32[2];
1228 			}
1229 			ill->ill_token_length = IPV6_TOKEN_LEN;
1230 		}
1231 		/*
1232 		 * Attempt to set the link local address if it isn't set.
1233 		 */
1234 		if (IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr))
1235 			(void) ipif_setlinklocal(ipif);
1236 	}
1237 }
1238 
1239 /*
1240  * Is it not possible to set the link local address?
1241  * The address can be set if the token is set, and the token
1242  * isn't too long.
1243  * Return B_TRUE if the address can't be set, or B_FALSE if it can.
1244  */
1245 boolean_t
1246 ipif_cant_setlinklocal(ipif_t *ipif)
1247 {
1248 	ill_t *ill = ipif->ipif_ill;
1249 
1250 	if (IN6_IS_ADDR_UNSPECIFIED(&ill->ill_token) ||
1251 	    ill->ill_token_length > IPV6_ABITS - IPV6_LL_PREFIXLEN)
1252 		return (B_TRUE);
1253 
1254 	return (B_FALSE);
1255 }
1256 
1257 /*
1258  * Generate a link-local address from the token.
1259  * Return zero if the address was set, or non-zero if it couldn't be set.
1260  */
1261 int
1262 ipif_setlinklocal(ipif_t *ipif)
1263 {
1264 	ill_t		*ill = ipif->ipif_ill;
1265 	in6_addr_t	ov6addr;
1266 
1267 	ASSERT(IAM_WRITER_ILL(ill));
1268 
1269 	if (ipif_cant_setlinklocal(ipif))
1270 		return (-1);
1271 
1272 	ov6addr = ipif->ipif_v6lcl_addr;
1273 	ipif_get_linklocal(&ipif->ipif_v6lcl_addr, &ill->ill_token);
1274 	sctp_update_ipif_addr(ipif, ov6addr);
1275 	(void) ip_plen_to_mask_v6(IPV6_LL_PREFIXLEN, &ipif->ipif_v6net_mask);
1276 	V6_MASK_COPY(ipif->ipif_v6lcl_addr, ipif->ipif_v6net_mask,
1277 	    ipif->ipif_v6subnet);
1278 
1279 	if (ipif->ipif_flags & IPIF_NOLOCAL) {
1280 		ipif->ipif_v6src_addr = ipv6_all_zeros;
1281 	} else {
1282 		ipif->ipif_v6src_addr = ipif->ipif_v6lcl_addr;
1283 	}
1284 	return (0);
1285 }
1286 
1287 /*
1288  * This function sets up the multicast mappings in NDP.
1289  * Unlike ARP, there are no mapping_mps here. We delete the
1290  * mapping nces and add a new one.
1291  *
1292  * Returns non-zero on error and 0 on success.
1293  */
1294 int
1295 ipif_ndp_setup_multicast(ipif_t *ipif, nce_t **ret_nce)
1296 {
1297 	ill_t		*ill = ipif->ipif_ill;
1298 	in6_addr_t	v6_mcast_addr = {(uint32_t)V6_MCAST, 0, 0, 0};
1299 	in6_addr_t	v6_mcast_mask = {(uint32_t)V6_MCAST, 0, 0, 0};
1300 	in6_addr_t	v6_extract_mask;
1301 	uchar_t		*phys_addr, *bphys_addr, *alloc_phys;
1302 	nce_t		*mnce = NULL;
1303 	int		err = 0;
1304 	phyint_t	*phyi = ill->ill_phyint;
1305 	uint32_t	hw_extract_start;
1306 	dl_unitdata_req_t *dlur;
1307 	ip_stack_t	*ipst = ill->ill_ipst;
1308 
1309 	if (ret_nce != NULL)
1310 		*ret_nce = NULL;
1311 	/*
1312 	 * Delete the mapping nce. Normally these should not exist
1313 	 * as a previous ipif_down -> ipif_ndp_down should have deleted
1314 	 * all the nces. But they can exist if ip_rput_dlpi_writer
1315 	 * calls this when PHYI_MULTI_BCAST is set.
1316 	 */
1317 	mnce = ndp_lookup_v6(ill, &v6_mcast_addr, B_FALSE);
1318 	if (mnce != NULL) {
1319 		ndp_delete(mnce);
1320 		NCE_REFRELE(mnce);
1321 		mnce = NULL;
1322 	}
1323 
1324 	/*
1325 	 * Get media specific v6 mapping information. Note that
1326 	 * nd_lla_len can be 0 for tunnels.
1327 	 */
1328 	alloc_phys = kmem_alloc(ill->ill_nd_lla_len, KM_NOSLEEP);
1329 	if ((alloc_phys == NULL) && (ill->ill_nd_lla_len != 0))
1330 		return (ENOMEM);
1331 	/*
1332 	 * Determine the broadcast address.
1333 	 */
1334 	dlur = (dl_unitdata_req_t *)ill->ill_bcast_mp->b_rptr;
1335 	if (ill->ill_sap_length < 0)
1336 		bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset;
1337 	else
1338 		bphys_addr = (uchar_t *)dlur +
1339 		    dlur->dl_dest_addr_offset + ill->ill_sap_length;
1340 
1341 	/*
1342 	 * Check PHYI_MULTI_BCAST and possible length of physical
1343 	 * address to determine if we use the mapping or the
1344 	 * broadcast address.
1345 	 */
1346 	if ((phyi->phyint_flags & PHYI_MULTI_BCAST) ||
1347 	    (!MEDIA_V6MINFO(ill->ill_media, ill->ill_nd_lla_len,
1348 	    bphys_addr, alloc_phys, &hw_extract_start,
1349 	    &v6_extract_mask))) {
1350 		if (ill->ill_phys_addr_length > IP_MAX_HW_LEN) {
1351 			kmem_free(alloc_phys, ill->ill_nd_lla_len);
1352 			return (E2BIG);
1353 		}
1354 		/* Use the link-layer broadcast address for MULTI_BCAST */
1355 		phys_addr = bphys_addr;
1356 		bzero(&v6_extract_mask, sizeof (v6_extract_mask));
1357 		hw_extract_start = ill->ill_nd_lla_len;
1358 	} else {
1359 		phys_addr = alloc_phys;
1360 	}
1361 	if ((ipif->ipif_flags & IPIF_BROADCAST) ||
1362 	    (ill->ill_flags & ILLF_MULTICAST) ||
1363 	    (phyi->phyint_flags & PHYI_MULTI_BCAST)) {
1364 		mutex_enter(&ipst->ips_ndp6->ndp_g_lock);
1365 		err = ndp_add_v6(ill,
1366 		    phys_addr,
1367 		    &v6_mcast_addr,	/* v6 address */
1368 		    &v6_mcast_mask,	/* v6 mask */
1369 		    &v6_extract_mask,
1370 		    hw_extract_start,
1371 		    NCE_F_MAPPING | NCE_F_PERMANENT | NCE_F_NONUD,
1372 		    ND_REACHABLE,
1373 		    &mnce);
1374 		mutex_exit(&ipst->ips_ndp6->ndp_g_lock);
1375 		if (err == 0) {
1376 			if (ret_nce != NULL) {
1377 				*ret_nce = mnce;
1378 			} else {
1379 				NCE_REFRELE(mnce);
1380 			}
1381 		}
1382 	}
1383 	kmem_free(alloc_phys, ill->ill_nd_lla_len);
1384 	return (err);
1385 }
1386 
1387 /*
1388  * Get the resolver set up for a new ipif.  (Always called as writer.)
1389  */
1390 int
1391 ipif_ndp_up(ipif_t *ipif)
1392 {
1393 	ill_t		*ill = ipif->ipif_ill;
1394 	int		err = 0;
1395 	nce_t		*nce = NULL;
1396 	nce_t		*mnce = NULL;
1397 
1398 	ip1dbg(("ipif_ndp_up(%s:%u)\n", ill->ill_name, ipif->ipif_id));
1399 
1400 	/*
1401 	 * ND not supported on XRESOLV interfaces. If ND support (multicast)
1402 	 * added later, take out this check.
1403 	 */
1404 	if ((ill->ill_flags & ILLF_XRESOLV) ||
1405 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr) ||
1406 	    (!(ill->ill_net_type & IRE_INTERFACE))) {
1407 		ipif->ipif_addr_ready = 1;
1408 		return (0);
1409 	}
1410 
1411 	/*
1412 	 * Need to setup multicast mapping only when the first
1413 	 * interface is coming UP.
1414 	 */
1415 	if (ill->ill_ipif_up_count == 0 &&
1416 	    (ill->ill_flags & ILLF_MULTICAST)) {
1417 		/*
1418 		 * We set the multicast before setting up the mapping for
1419 		 * local address because ipif_ndp_setup_multicast does
1420 		 * ndp_walk to delete nces which will delete the mapping
1421 		 * for local address also if we added the mapping for
1422 		 * local address first.
1423 		 */
1424 		err = ipif_ndp_setup_multicast(ipif, &mnce);
1425 		if (err != 0)
1426 			return (err);
1427 	}
1428 
1429 	if ((ipif->ipif_flags & (IPIF_UNNUMBERED|IPIF_NOLOCAL)) == 0) {
1430 		uint16_t	flags;
1431 		uchar_t	*hw_addr = NULL;
1432 
1433 		/* Permanent entries don't need NUD */
1434 		flags = NCE_F_PERMANENT | NCE_F_NONUD;
1435 		if (ill->ill_flags & ILLF_ROUTER)
1436 			flags |= NCE_F_ISROUTER;
1437 
1438 		if (ipif->ipif_flags & IPIF_ANYCAST)
1439 			flags |= NCE_F_ANYCAST;
1440 
1441 		if (ill->ill_net_type == IRE_IF_RESOLVER) {
1442 			hw_addr = ill->ill_nd_lla;
1443 
1444 			if (ill->ill_move_in_progress) {
1445 				/*
1446 				 * Addresses are failing over to this ill.
1447 				 * Don't wait for NUD to see this change.
1448 				 * Publish our new link-layer address.
1449 				 */
1450 				flags |= NCE_F_UNSOL_ADV;
1451 			}
1452 		}
1453 		err = ndp_lookup_then_add_v6(ill,
1454 		    hw_addr,
1455 		    &ipif->ipif_v6lcl_addr,
1456 		    &ipv6_all_ones,
1457 		    &ipv6_all_zeros,
1458 		    0,
1459 		    flags,
1460 		    ND_PROBE,	/* Causes Duplicate Address Detection to run */
1461 		    &nce);
1462 		switch (err) {
1463 		case 0:
1464 			ip1dbg(("ipif_ndp_up: NCE created for %s\n",
1465 			    ill->ill_name));
1466 			ipif->ipif_addr_ready = 1;
1467 			break;
1468 		case EINPROGRESS:
1469 			ip1dbg(("ipif_ndp_up: running DAD now for %s\n",
1470 			    ill->ill_name));
1471 			break;
1472 		case EEXIST:
1473 			NCE_REFRELE(nce);
1474 			ip1dbg(("ipif_ndp_up: NCE already exists for %s\n",
1475 			    ill->ill_name));
1476 			if (mnce != NULL) {
1477 				ndp_delete(mnce);
1478 				NCE_REFRELE(mnce);
1479 			}
1480 			return (err);
1481 		default:
1482 			ip1dbg(("ipif_ndp_up: NCE creation failed %s\n",
1483 			    ill->ill_name));
1484 			if (mnce != NULL) {
1485 				ndp_delete(mnce);
1486 				NCE_REFRELE(mnce);
1487 			}
1488 			return (err);
1489 		}
1490 	} else {
1491 		/* No local NCE for this entry */
1492 		ipif->ipif_addr_ready = 1;
1493 	}
1494 	if (nce != NULL)
1495 		NCE_REFRELE(nce);
1496 	if (mnce != NULL)
1497 		NCE_REFRELE(mnce);
1498 	return (0);
1499 }
1500 
1501 /* Remove all cache entries for this logical interface */
1502 void
1503 ipif_ndp_down(ipif_t *ipif)
1504 {
1505 	nce_t	*nce;
1506 
1507 	if (ipif->ipif_isv6) {
1508 		nce = ndp_lookup_v6(ipif->ipif_ill, &ipif->ipif_v6lcl_addr,
1509 		    B_FALSE);
1510 		if (nce != NULL) {
1511 			ndp_delete(nce);
1512 			NCE_REFRELE(nce);
1513 		}
1514 	}
1515 	/*
1516 	 * Remove mapping and all other nces dependent on this ill
1517 	 * when the last ipif is going away.
1518 	 */
1519 	if (ipif->ipif_ill->ill_ipif_up_count == 0) {
1520 		ndp_walk(ipif->ipif_ill, (pfi_t)ndp_delete_per_ill,
1521 		    (uchar_t *)ipif->ipif_ill, ipif->ipif_ill->ill_ipst);
1522 	}
1523 }
1524 
1525 /*
1526  * Used when an interface comes up to recreate any extra routes on this
1527  * interface.
1528  */
1529 static ire_t **
1530 ipif_recover_ire_v6(ipif_t *ipif)
1531 {
1532 	mblk_t	*mp;
1533 	ire_t   **ipif_saved_irep;
1534 	ire_t   **irep;
1535 	ip_stack_t	*ipst = ipif->ipif_ill->ill_ipst;
1536 
1537 	ip1dbg(("ipif_recover_ire_v6(%s:%u)", ipif->ipif_ill->ill_name,
1538 	    ipif->ipif_id));
1539 
1540 	ASSERT(ipif->ipif_isv6);
1541 
1542 	mutex_enter(&ipif->ipif_saved_ire_lock);
1543 	ipif_saved_irep = (ire_t **)kmem_zalloc(sizeof (ire_t *) *
1544 	    ipif->ipif_saved_ire_cnt, KM_NOSLEEP);
1545 	if (ipif_saved_irep == NULL) {
1546 		mutex_exit(&ipif->ipif_saved_ire_lock);
1547 		return (NULL);
1548 	}
1549 
1550 	irep = ipif_saved_irep;
1551 
1552 	for (mp = ipif->ipif_saved_ire_mp; mp != NULL; mp = mp->b_cont) {
1553 		ire_t		*ire;
1554 		queue_t		*rfq;
1555 		queue_t		*stq;
1556 		ifrt_t		*ifrt;
1557 		in6_addr_t	*src_addr;
1558 		in6_addr_t	*gateway_addr;
1559 		char		buf[INET6_ADDRSTRLEN];
1560 		ushort_t	type;
1561 
1562 		/*
1563 		 * When the ire was initially created and then added in
1564 		 * ip_rt_add_v6(), it was created either using
1565 		 * ipif->ipif_net_type in the case of a traditional interface
1566 		 * route, or as one of the IRE_OFFSUBNET types (with the
1567 		 * exception of IRE_HOST type redirect ire which is created by
1568 		 * icmp_redirect_v6() and which we don't need to save or
1569 		 * recover).  In the case where ipif->ipif_net_type was
1570 		 * IRE_LOOPBACK, ip_rt_add_v6() will update the ire_type to
1571 		 * IRE_IF_NORESOLVER before calling ire_add_v6() to satisfy
1572 		 * software like GateD and Sun Cluster which creates routes
1573 		 * using the the loopback interface's address as a gateway.
1574 		 *
1575 		 * As ifrt->ifrt_type reflects the already updated ire_type,
1576 		 * ire_create_v6() will be called in the same way here as in
1577 		 * ip_rt_add_v6(), namely using ipif->ipif_net_type when the
1578 		 * route looks like a traditional interface route (where
1579 		 * ifrt->ifrt_type & IRE_INTERFACE is true) and otherwise
1580 		 * using the saved ifrt->ifrt_type.  This means that in
1581 		 * the case where ipif->ipif_net_type is IRE_LOOPBACK,
1582 		 * the ire created by ire_create_v6() will be an IRE_LOOPBACK,
1583 		 * it will then be turned into an IRE_IF_NORESOLVER and then
1584 		 * added by ire_add_v6().
1585 		 */
1586 		ifrt = (ifrt_t *)mp->b_rptr;
1587 		if (ifrt->ifrt_type & IRE_INTERFACE) {
1588 			rfq = NULL;
1589 			stq = (ipif->ipif_net_type == IRE_IF_RESOLVER)
1590 			    ? ipif->ipif_rq : ipif->ipif_wq;
1591 			src_addr = (ifrt->ifrt_flags & RTF_SETSRC)
1592 			    ? &ifrt->ifrt_v6src_addr
1593 			    : &ipif->ipif_v6src_addr;
1594 			gateway_addr = NULL;
1595 			type = ipif->ipif_net_type;
1596 		} else {
1597 			rfq = NULL;
1598 			stq = NULL;
1599 			src_addr = (ifrt->ifrt_flags & RTF_SETSRC)
1600 			    ? &ifrt->ifrt_v6src_addr : NULL;
1601 			gateway_addr = &ifrt->ifrt_v6gateway_addr;
1602 			type = ifrt->ifrt_type;
1603 		}
1604 
1605 		/*
1606 		 * Create a copy of the IRE with the saved address and netmask.
1607 		 */
1608 		ip1dbg(("ipif_recover_ire_v6: creating IRE %s (%d) for %s/%d\n",
1609 		    ip_nv_lookup(ire_nv_tbl, ifrt->ifrt_type), ifrt->ifrt_type,
1610 		    inet_ntop(AF_INET6, &ifrt->ifrt_v6addr, buf, sizeof (buf)),
1611 		    ip_mask_to_plen_v6(&ifrt->ifrt_v6mask)));
1612 		ire = ire_create_v6(
1613 		    &ifrt->ifrt_v6addr,
1614 		    &ifrt->ifrt_v6mask,
1615 		    src_addr,
1616 		    gateway_addr,
1617 		    &ifrt->ifrt_max_frag,
1618 		    NULL,
1619 		    rfq,
1620 		    stq,
1621 		    type,
1622 		    ipif,
1623 		    NULL,
1624 		    0,
1625 		    0,
1626 		    ifrt->ifrt_flags,
1627 		    &ifrt->ifrt_iulp_info,
1628 		    NULL,
1629 		    NULL,
1630 		    ipst);
1631 		if (ire == NULL) {
1632 			mutex_exit(&ipif->ipif_saved_ire_lock);
1633 			kmem_free(ipif_saved_irep,
1634 			    ipif->ipif_saved_ire_cnt * sizeof (ire_t *));
1635 			return (NULL);
1636 		}
1637 
1638 		/*
1639 		 * Some software (for example, GateD and Sun Cluster) attempts
1640 		 * to create (what amount to) IRE_PREFIX routes with the
1641 		 * loopback address as the gateway.  This is primarily done to
1642 		 * set up prefixes with the RTF_REJECT flag set (for example,
1643 		 * when generating aggregate routes.)
1644 		 *
1645 		 * If the IRE type (as defined by ipif->ipif_net_type) is
1646 		 * IRE_LOOPBACK, then we map the request into a
1647 		 * IRE_IF_NORESOLVER.
1648 		 */
1649 		if (ipif->ipif_net_type == IRE_LOOPBACK)
1650 			ire->ire_type = IRE_IF_NORESOLVER;
1651 		/*
1652 		 * ire held by ire_add, will be refreled' in ipif_up_done
1653 		 * towards the end
1654 		 */
1655 		(void) ire_add(&ire, NULL, NULL, NULL, B_FALSE);
1656 		*irep = ire;
1657 		irep++;
1658 		ip1dbg(("ipif_recover_ire_v6: added ire %p\n", (void *)ire));
1659 	}
1660 	mutex_exit(&ipif->ipif_saved_ire_lock);
1661 	return (ipif_saved_irep);
1662 }
1663 
1664 /*
1665  * Return the scope of the given IPv6 address.  If the address is an
1666  * IPv4 mapped IPv6 address, return the scope of the corresponding
1667  * IPv4 address.
1668  */
1669 in6addr_scope_t
1670 ip_addr_scope_v6(const in6_addr_t *addr)
1671 {
1672 	static in6_addr_t ipv6loopback = IN6ADDR_LOOPBACK_INIT;
1673 
1674 	if (IN6_IS_ADDR_V4MAPPED(addr)) {
1675 		in_addr_t v4addr_h = ntohl(V4_PART_OF_V6((*addr)));
1676 		if ((v4addr_h >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1677 		    (v4addr_h & IN_AUTOCONF_MASK) == IN_AUTOCONF_NET)
1678 			return (IP6_SCOPE_LINKLOCAL);
1679 		if ((v4addr_h & IN_PRIVATE8_MASK) == IN_PRIVATE8_NET ||
1680 		    (v4addr_h & IN_PRIVATE12_MASK) == IN_PRIVATE12_NET ||
1681 		    (v4addr_h & IN_PRIVATE16_MASK) == IN_PRIVATE16_NET)
1682 			return (IP6_SCOPE_SITELOCAL);
1683 		return (IP6_SCOPE_GLOBAL);
1684 	}
1685 
1686 	if (IN6_IS_ADDR_MULTICAST(addr))
1687 		return (IN6_ADDR_MC_SCOPE(addr));
1688 
1689 	/* link-local and loopback addresses are of link-local scope */
1690 	if (IN6_IS_ADDR_LINKLOCAL(addr) ||
1691 	    IN6_ARE_ADDR_EQUAL(addr, &ipv6loopback))
1692 		return (IP6_SCOPE_LINKLOCAL);
1693 	if (IN6_IS_ADDR_SITELOCAL(addr))
1694 		return (IP6_SCOPE_SITELOCAL);
1695 	return (IP6_SCOPE_GLOBAL);
1696 }
1697 
1698 
1699 /*
1700  * Returns the length of the common prefix of a1 and a2, as per
1701  * CommonPrefixLen() defined in RFC 3484.
1702  */
1703 static int
1704 ip_common_prefix_v6(const in6_addr_t *a1, const in6_addr_t *a2)
1705 {
1706 	int i;
1707 	uint32_t a1val, a2val, mask;
1708 
1709 	for (i = 0; i < 4; i++) {
1710 		if ((a1val = a1->s6_addr32[i]) != (a2val = a2->s6_addr32[i])) {
1711 			a1val ^= a2val;
1712 			i *= 32;
1713 			mask = 0x80000000u;
1714 			while (!(a1val & mask)) {
1715 				mask >>= 1;
1716 				i++;
1717 			}
1718 			return (i);
1719 		}
1720 	}
1721 	return (IPV6_ABITS);
1722 }
1723 
1724 #define	IPIF_VALID_IPV6_SOURCE(ipif) \
1725 	(((ipif)->ipif_flags & IPIF_UP) && \
1726 	!((ipif)->ipif_flags & (IPIF_NOLOCAL|IPIF_ANYCAST)) && \
1727 	(ipif)->ipif_addr_ready)
1728 
1729 /* source address candidate */
1730 typedef struct candidate {
1731 	ipif_t		*cand_ipif;
1732 	/* The properties of this candidate */
1733 	boolean_t	cand_isdst;
1734 	boolean_t	cand_isdst_set;
1735 	in6addr_scope_t	cand_scope;
1736 	boolean_t	cand_scope_set;
1737 	boolean_t	cand_isdeprecated;
1738 	boolean_t	cand_isdeprecated_set;
1739 	boolean_t	cand_ispreferred;
1740 	boolean_t	cand_ispreferred_set;
1741 	boolean_t	cand_matchedinterface;
1742 	boolean_t	cand_matchedinterface_set;
1743 	boolean_t	cand_matchedlabel;
1744 	boolean_t	cand_matchedlabel_set;
1745 	boolean_t	cand_istmp;
1746 	boolean_t	cand_istmp_set;
1747 	int		cand_common_pref;
1748 	boolean_t	cand_common_pref_set;
1749 	boolean_t	cand_pref_eq;
1750 	boolean_t	cand_pref_eq_set;
1751 	int		cand_pref_len;
1752 	boolean_t	cand_pref_len_set;
1753 } cand_t;
1754 #define	cand_srcaddr	cand_ipif->ipif_v6lcl_addr
1755 #define	cand_mask	cand_ipif->ipif_v6net_mask
1756 #define	cand_flags	cand_ipif->ipif_flags
1757 #define	cand_ill	cand_ipif->ipif_ill
1758 #define	cand_zoneid	cand_ipif->ipif_zoneid
1759 
1760 /* information about the destination for source address selection */
1761 typedef struct dstinfo {
1762 	const in6_addr_t	*dst_addr;
1763 	ill_t			*dst_ill;
1764 	uint_t			dst_restrict_ill;
1765 	boolean_t		dst_prefer_src_tmp;
1766 	in6addr_scope_t		dst_scope;
1767 	char			*dst_label;
1768 } dstinfo_t;
1769 
1770 /*
1771  * The following functions are rules used to select a source address in
1772  * ipif_select_source_v6().  Each rule compares a current candidate (cc)
1773  * against the best candidate (bc).  Each rule has three possible outcomes;
1774  * the candidate is preferred over the best candidate (CAND_PREFER), the
1775  * candidate is not preferred over the best candidate (CAND_AVOID), or the
1776  * candidate is of equal value as the best candidate (CAND_TIE).
1777  *
1778  * These rules are part of a greater "Default Address Selection for IPv6"
1779  * sheme, which is standards based work coming out of the IETF ipv6 working
1780  * group.  The IETF document defines both IPv6 source address selection and
1781  * destination address ordering.  The rules defined here implement the IPv6
1782  * source address selection.  Destination address ordering is done by
1783  * libnsl, and uses a similar set of rules to implement the sorting.
1784  *
1785  * Most of the rules are defined by the RFC and are not typically altered.  The
1786  * last rule, number 8, has language that allows for local preferences.  In the
1787  * scheme below, this means that new Solaris rules should normally go between
1788  * rule_ifprefix and rule_prefix.
1789  */
1790 typedef enum {CAND_AVOID, CAND_TIE, CAND_PREFER} rule_res_t;
1791 typedef	rule_res_t (*rulef_t)(cand_t *, cand_t *, const dstinfo_t *,
1792     ip_stack_t *);
1793 
1794 /* Prefer an address if it is equal to the destination address. */
1795 /* ARGSUSED3 */
1796 static rule_res_t
1797 rule_isdst(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1798 {
1799 	if (!bc->cand_isdst_set) {
1800 		bc->cand_isdst =
1801 		    IN6_ARE_ADDR_EQUAL(&bc->cand_srcaddr, dstinfo->dst_addr);
1802 		bc->cand_isdst_set = B_TRUE;
1803 	}
1804 
1805 	cc->cand_isdst =
1806 	    IN6_ARE_ADDR_EQUAL(&cc->cand_srcaddr, dstinfo->dst_addr);
1807 	cc->cand_isdst_set = B_TRUE;
1808 
1809 	if (cc->cand_isdst == bc->cand_isdst)
1810 		return (CAND_TIE);
1811 	else if (cc->cand_isdst)
1812 		return (CAND_PREFER);
1813 	else
1814 		return (CAND_AVOID);
1815 }
1816 
1817 /*
1818  * Prefer addresses that are of closest scope to the destination.  Always
1819  * prefer addresses that are of greater scope than the destination over
1820  * those that are of lesser scope than the destination.
1821  */
1822 /* ARGSUSED3 */
1823 static rule_res_t
1824 rule_scope(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1825 {
1826 	if (!bc->cand_scope_set) {
1827 		bc->cand_scope = ip_addr_scope_v6(&bc->cand_srcaddr);
1828 		bc->cand_scope_set = B_TRUE;
1829 	}
1830 
1831 	cc->cand_scope = ip_addr_scope_v6(&cc->cand_srcaddr);
1832 	cc->cand_scope_set = B_TRUE;
1833 
1834 	if (cc->cand_scope < bc->cand_scope) {
1835 		if (cc->cand_scope < dstinfo->dst_scope)
1836 			return (CAND_AVOID);
1837 		else
1838 			return (CAND_PREFER);
1839 	} else if (bc->cand_scope < cc->cand_scope) {
1840 		if (bc->cand_scope < dstinfo->dst_scope)
1841 			return (CAND_PREFER);
1842 		else
1843 			return (CAND_AVOID);
1844 	} else {
1845 		return (CAND_TIE);
1846 	}
1847 }
1848 
1849 /*
1850  * Prefer non-deprecated source addresses.
1851  */
1852 /* ARGSUSED2 */
1853 static rule_res_t
1854 rule_deprecated(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1855     ip_stack_t *ipst)
1856 {
1857 	if (!bc->cand_isdeprecated_set) {
1858 		bc->cand_isdeprecated =
1859 		    ((bc->cand_flags & IPIF_DEPRECATED) != 0);
1860 		bc->cand_isdeprecated_set = B_TRUE;
1861 	}
1862 
1863 	cc->cand_isdeprecated = ((cc->cand_flags & IPIF_DEPRECATED) != 0);
1864 	cc->cand_isdeprecated_set = B_TRUE;
1865 
1866 	if (bc->cand_isdeprecated == cc->cand_isdeprecated)
1867 		return (CAND_TIE);
1868 	else if (cc->cand_isdeprecated)
1869 		return (CAND_AVOID);
1870 	else
1871 		return (CAND_PREFER);
1872 }
1873 
1874 /*
1875  * Prefer source addresses that have the IPIF_PREFERRED flag set.  This
1876  * rule must be before rule_interface because the flag could be set on any
1877  * interface, not just the interface being used for outgoing packets (for
1878  * example, the IFF_PREFERRED could be set on an address assigned to the
1879  * loopback interface).
1880  */
1881 /* ARGSUSED2 */
1882 static rule_res_t
1883 rule_preferred(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1884     ip_stack_t *ipst)
1885 {
1886 	if (!bc->cand_ispreferred_set) {
1887 		bc->cand_ispreferred = ((bc->cand_flags & IPIF_PREFERRED) != 0);
1888 		bc->cand_ispreferred_set = B_TRUE;
1889 	}
1890 
1891 	cc->cand_ispreferred = ((cc->cand_flags & IPIF_PREFERRED) != 0);
1892 	cc->cand_ispreferred_set = B_TRUE;
1893 
1894 	if (bc->cand_ispreferred == cc->cand_ispreferred)
1895 		return (CAND_TIE);
1896 	else if (cc->cand_ispreferred)
1897 		return (CAND_PREFER);
1898 	else
1899 		return (CAND_AVOID);
1900 }
1901 
1902 /*
1903  * Prefer source addresses that are assigned to the outgoing interface, or
1904  * to an interface that is in the same IPMP group as the outgoing
1905  * interface.
1906  */
1907 /* ARGSUSED3 */
1908 static rule_res_t
1909 rule_interface(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1910     ip_stack_t *ipst)
1911 {
1912 	ill_t *dstill = dstinfo->dst_ill;
1913 
1914 	/*
1915 	 * If dstinfo->dst_restrict_ill is set, this rule is unnecessary
1916 	 * since we know all candidates will be on the same link.
1917 	 */
1918 	if (dstinfo->dst_restrict_ill)
1919 		return (CAND_TIE);
1920 
1921 	if (!bc->cand_matchedinterface_set) {
1922 		bc->cand_matchedinterface = (bc->cand_ill == dstill ||
1923 		    (dstill->ill_group != NULL &&
1924 		    dstill->ill_group == bc->cand_ill->ill_group));
1925 		bc->cand_matchedinterface_set = B_TRUE;
1926 	}
1927 
1928 	cc->cand_matchedinterface = (cc->cand_ill == dstill ||
1929 	    (dstill->ill_group != NULL &&
1930 	    dstill->ill_group == cc->cand_ill->ill_group));
1931 	cc->cand_matchedinterface_set = B_TRUE;
1932 
1933 	if (bc->cand_matchedinterface == cc->cand_matchedinterface)
1934 		return (CAND_TIE);
1935 	else if (cc->cand_matchedinterface)
1936 		return (CAND_PREFER);
1937 	else
1938 		return (CAND_AVOID);
1939 }
1940 
1941 /*
1942  * Prefer source addresses whose label matches the destination's label.
1943  */
1944 static rule_res_t
1945 rule_label(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1946 {
1947 	char *label;
1948 
1949 	if (!bc->cand_matchedlabel_set) {
1950 		label = ip6_asp_lookup(&bc->cand_srcaddr, NULL, ipst);
1951 		bc->cand_matchedlabel =
1952 		    ip6_asp_labelcmp(label, dstinfo->dst_label);
1953 		bc->cand_matchedlabel_set = B_TRUE;
1954 	}
1955 
1956 	label = ip6_asp_lookup(&cc->cand_srcaddr, NULL, ipst);
1957 	cc->cand_matchedlabel = ip6_asp_labelcmp(label, dstinfo->dst_label);
1958 	cc->cand_matchedlabel_set = B_TRUE;
1959 
1960 	if (bc->cand_matchedlabel == cc->cand_matchedlabel)
1961 		return (CAND_TIE);
1962 	else if (cc->cand_matchedlabel)
1963 		return (CAND_PREFER);
1964 	else
1965 		return (CAND_AVOID);
1966 }
1967 
1968 /*
1969  * Prefer public addresses over temporary ones.  An application can reverse
1970  * the logic of this rule and prefer temporary addresses by using the
1971  * IPV6_SRC_PREFERENCES socket option.
1972  */
1973 /* ARGSUSED3 */
1974 static rule_res_t
1975 rule_temporary(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1976     ip_stack_t *ipst)
1977 {
1978 	if (!bc->cand_istmp_set) {
1979 		bc->cand_istmp = ((bc->cand_flags & IPIF_TEMPORARY) != 0);
1980 		bc->cand_istmp_set = B_TRUE;
1981 	}
1982 
1983 	cc->cand_istmp = ((cc->cand_flags & IPIF_TEMPORARY) != 0);
1984 	cc->cand_istmp_set = B_TRUE;
1985 
1986 	if (bc->cand_istmp == cc->cand_istmp)
1987 		return (CAND_TIE);
1988 
1989 	if (dstinfo->dst_prefer_src_tmp && cc->cand_istmp)
1990 		return (CAND_PREFER);
1991 	else if (!dstinfo->dst_prefer_src_tmp && !cc->cand_istmp)
1992 		return (CAND_PREFER);
1993 	else
1994 		return (CAND_AVOID);
1995 }
1996 
1997 /*
1998  * Prefer source addresses with longer matching prefix with the destination
1999  * under the interface mask.  This gets us on the same subnet before applying
2000  * any Solaris-specific rules.
2001  */
2002 /* ARGSUSED3 */
2003 static rule_res_t
2004 rule_ifprefix(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2005     ip_stack_t *ipst)
2006 {
2007 	if (!bc->cand_pref_eq_set) {
2008 		bc->cand_pref_eq = V6_MASK_EQ_2(bc->cand_srcaddr,
2009 		    bc->cand_mask, *dstinfo->dst_addr);
2010 		bc->cand_pref_eq_set = B_TRUE;
2011 	}
2012 
2013 	cc->cand_pref_eq = V6_MASK_EQ_2(cc->cand_srcaddr, cc->cand_mask,
2014 	    *dstinfo->dst_addr);
2015 	cc->cand_pref_eq_set = B_TRUE;
2016 
2017 	if (bc->cand_pref_eq) {
2018 		if (cc->cand_pref_eq) {
2019 			if (!bc->cand_pref_len_set) {
2020 				bc->cand_pref_len =
2021 				    ip_mask_to_plen_v6(&bc->cand_mask);
2022 				bc->cand_pref_len_set = B_TRUE;
2023 			}
2024 			cc->cand_pref_len = ip_mask_to_plen_v6(&cc->cand_mask);
2025 			cc->cand_pref_len_set = B_TRUE;
2026 			if (bc->cand_pref_len == cc->cand_pref_len)
2027 				return (CAND_TIE);
2028 			else if (bc->cand_pref_len > cc->cand_pref_len)
2029 				return (CAND_AVOID);
2030 			else
2031 				return (CAND_PREFER);
2032 		} else {
2033 			return (CAND_AVOID);
2034 		}
2035 	} else {
2036 		if (cc->cand_pref_eq)
2037 			return (CAND_PREFER);
2038 		else
2039 			return (CAND_TIE);
2040 	}
2041 }
2042 
2043 /*
2044  * Prefer to use zone-specific addresses when possible instead of all-zones
2045  * addresses.
2046  */
2047 /* ARGSUSED2 */
2048 static rule_res_t
2049 rule_zone_specific(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2050     ip_stack_t *ipst)
2051 {
2052 	if ((bc->cand_zoneid == ALL_ZONES) ==
2053 	    (cc->cand_zoneid == ALL_ZONES))
2054 		return (CAND_TIE);
2055 	else if (cc->cand_zoneid == ALL_ZONES)
2056 		return (CAND_AVOID);
2057 	else
2058 		return (CAND_PREFER);
2059 }
2060 
2061 /*
2062  * Prefer to use DHCPv6 (first) and static addresses (second) when possible
2063  * instead of statelessly autoconfigured addresses.
2064  *
2065  * This is done after trying all other preferences (and before the final tie
2066  * breaker) so that, if all else is equal, we select addresses configured by
2067  * DHCPv6 over other addresses.  We presume that DHCPv6 addresses, unlike
2068  * stateless autoconfigured addresses, are deliberately configured by an
2069  * administrator, and thus are correctly set up in DNS and network packet
2070  * filters.
2071  */
2072 /* ARGSUSED2 */
2073 static rule_res_t
2074 rule_addr_type(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2075     ip_stack_t *ipst)
2076 {
2077 #define	ATYPE(x)	\
2078 	((x) & IPIF_DHCPRUNNING) ? 1 : ((x) & IPIF_ADDRCONF) ? 3 : 2
2079 	int bcval = ATYPE(bc->cand_flags);
2080 	int ccval = ATYPE(cc->cand_flags);
2081 #undef ATYPE
2082 
2083 	if (bcval == ccval)
2084 		return (CAND_TIE);
2085 	else if (ccval < bcval)
2086 		return (CAND_PREFER);
2087 	else
2088 		return (CAND_AVOID);
2089 }
2090 
2091 /*
2092  * Prefer source addresses with longer matching prefix with the destination.
2093  * We do the longest matching prefix calculation by doing an xor of both
2094  * addresses with the destination, and pick the address with the longest string
2095  * of leading zeros, as per CommonPrefixLen() defined in RFC 3484.
2096  */
2097 /* ARGSUSED3 */
2098 static rule_res_t
2099 rule_prefix(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
2100 {
2101 	if (!bc->cand_common_pref_set) {
2102 		bc->cand_common_pref = ip_common_prefix_v6(&bc->cand_srcaddr,
2103 		    dstinfo->dst_addr);
2104 		bc->cand_common_pref_set = B_TRUE;
2105 	}
2106 
2107 	cc->cand_common_pref = ip_common_prefix_v6(&cc->cand_srcaddr,
2108 	    dstinfo->dst_addr);
2109 	cc->cand_common_pref_set = B_TRUE;
2110 
2111 	if (bc->cand_common_pref == cc->cand_common_pref)
2112 		return (CAND_TIE);
2113 	else if (bc->cand_common_pref > cc->cand_common_pref)
2114 		return (CAND_AVOID);
2115 	else
2116 		return (CAND_PREFER);
2117 }
2118 
2119 /*
2120  * Last rule: we must pick something, so just prefer the current best
2121  * candidate.
2122  */
2123 /* ARGSUSED */
2124 static rule_res_t
2125 rule_must_be_last(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2126     ip_stack_t *ipst)
2127 {
2128 	return (CAND_AVOID);
2129 }
2130 
2131 /*
2132  * Determine the best source address given a destination address and a
2133  * destination ill.  If no suitable source address is found, it returns
2134  * NULL. If there is a usable address pointed to by the usesrc
2135  * (i.e ill_usesrc_ifindex != 0) then return that first since it is more
2136  * fine grained (i.e per interface)
2137  *
2138  * This implementation is based on the "Default Address Selection for IPv6"
2139  * specification produced by the IETF IPv6 working group.  It has been
2140  * implemented so that the list of addresses is only traversed once (the
2141  * specification's algorithm could traverse the list of addresses once for
2142  * every rule).
2143  *
2144  * The restrict_ill argument restricts the algorithm to chose a source
2145  * address that is assigned to the destination ill or an ill in the same
2146  * IPMP group as the destination ill.  This is used when the destination
2147  * address is a link-local or multicast address, and when
2148  * ipv6_strict_dst_multihoming is turned on.
2149  *
2150  * src_prefs is the caller's set of source address preferences.  If source
2151  * address selection is being called to determine the source address of a
2152  * connected socket (from ip_bind_connected_v6()), then the preferences are
2153  * taken from conn_src_preferences.  These preferences can be set on a
2154  * per-socket basis using the IPV6_SRC_PREFERENCES socket option.  The only
2155  * preference currently implemented is for rfc3041 temporary addresses.
2156  */
2157 ipif_t *
2158 ipif_select_source_v6(ill_t *dstill, const in6_addr_t *dst,
2159     uint_t restrict_ill, uint32_t src_prefs, zoneid_t zoneid)
2160 {
2161 	dstinfo_t	dstinfo;
2162 	char		dstr[INET6_ADDRSTRLEN];
2163 	char		sstr[INET6_ADDRSTRLEN];
2164 	ipif_t		*ipif;
2165 	ill_t		*ill, *usesrc_ill = NULL;
2166 	ill_walk_context_t	ctx;
2167 	cand_t		best_c;	/* The best candidate */
2168 	cand_t		curr_c;	/* The current candidate */
2169 	uint_t		index;
2170 	boolean_t	first_candidate = B_TRUE;
2171 	rule_res_t	rule_result;
2172 	tsol_tpc_t	*src_rhtp, *dst_rhtp;
2173 	ip_stack_t	*ipst = dstill->ill_ipst;
2174 
2175 	/*
2176 	 * The list of ordering rules.  They are applied in the order they
2177 	 * appear in the list.
2178 	 *
2179 	 * Solaris doesn't currently support Mobile IPv6, so there's no
2180 	 * rule_mipv6 corresponding to rule 4 in the specification.
2181 	 */
2182 	rulef_t	rules[] = {
2183 		rule_isdst,
2184 		rule_scope,
2185 		rule_deprecated,
2186 		rule_preferred,
2187 		rule_interface,
2188 		rule_label,
2189 		rule_temporary,
2190 		rule_ifprefix,			/* local rules after this */
2191 		rule_zone_specific,
2192 		rule_addr_type,
2193 		rule_prefix,			/* local rules before this */
2194 		rule_must_be_last,		/* must always be last */
2195 		NULL
2196 	};
2197 
2198 	ASSERT(dstill->ill_isv6);
2199 	ASSERT(!IN6_IS_ADDR_V4MAPPED(dst));
2200 
2201 	/*
2202 	 * Check if there is a usable src address pointed to by the
2203 	 * usesrc ifindex. This has higher precedence since it is
2204 	 * finer grained (i.e per interface) v/s being system wide.
2205 	 */
2206 	if (dstill->ill_usesrc_ifindex != 0) {
2207 		if ((usesrc_ill =
2208 		    ill_lookup_on_ifindex(dstill->ill_usesrc_ifindex, B_TRUE,
2209 		    NULL, NULL, NULL, NULL, ipst)) != NULL) {
2210 			dstinfo.dst_ill = usesrc_ill;
2211 		} else {
2212 			return (NULL);
2213 		}
2214 	} else {
2215 		dstinfo.dst_ill = dstill;
2216 	}
2217 
2218 	/*
2219 	 * If we're dealing with an unlabeled destination on a labeled system,
2220 	 * make sure that we ignore source addresses that are incompatible with
2221 	 * the destination's default label.  That destination's default label
2222 	 * must dominate the minimum label on the source address.
2223 	 *
2224 	 * (Note that this has to do with Trusted Solaris.  It's not related to
2225 	 * the labels described by ip6_asp_lookup.)
2226 	 */
2227 	dst_rhtp = NULL;
2228 	if (is_system_labeled()) {
2229 		dst_rhtp = find_tpc(dst, IPV6_VERSION, B_FALSE);
2230 		if (dst_rhtp == NULL)
2231 			return (NULL);
2232 		if (dst_rhtp->tpc_tp.host_type != UNLABELED) {
2233 			TPC_RELE(dst_rhtp);
2234 			dst_rhtp = NULL;
2235 		}
2236 	}
2237 
2238 	dstinfo.dst_addr = dst;
2239 	dstinfo.dst_scope = ip_addr_scope_v6(dst);
2240 	dstinfo.dst_label = ip6_asp_lookup(dst, NULL, ipst);
2241 	dstinfo.dst_prefer_src_tmp = ((src_prefs & IPV6_PREFER_SRC_TMP) != 0);
2242 
2243 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
2244 	/*
2245 	 * Section three of the I-D states that for multicast and
2246 	 * link-local destinations, the candidate set must be restricted to
2247 	 * an interface that is on the same link as the outgoing interface.
2248 	 * Also, when ipv6_strict_dst_multihoming is turned on, always
2249 	 * restrict the source address to the destination link as doing
2250 	 * otherwise will almost certainly cause problems.
2251 	 */
2252 	if (IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst) ||
2253 	    ipst->ips_ipv6_strict_dst_multihoming || usesrc_ill != NULL) {
2254 		if (restrict_ill == RESTRICT_TO_NONE)
2255 			dstinfo.dst_restrict_ill = RESTRICT_TO_GROUP;
2256 		else
2257 			dstinfo.dst_restrict_ill = restrict_ill;
2258 	} else {
2259 		dstinfo.dst_restrict_ill = restrict_ill;
2260 	}
2261 
2262 	bzero(&best_c, sizeof (cand_t));
2263 
2264 	/*
2265 	 * Take a pass through the list of IPv6 interfaces to chose the
2266 	 * best possible source address.  If restrict_ill is true, we only
2267 	 * iterate through the ill's that are in the same IPMP group as the
2268 	 * destination's outgoing ill.  If restrict_ill is false, we walk
2269 	 * the entire list of IPv6 ill's.
2270 	 */
2271 	if (dstinfo.dst_restrict_ill != RESTRICT_TO_NONE) {
2272 		if (dstinfo.dst_ill->ill_group != NULL &&
2273 		    dstinfo.dst_restrict_ill == RESTRICT_TO_GROUP) {
2274 			ill = dstinfo.dst_ill->ill_group->illgrp_ill;
2275 		} else {
2276 			ill = dstinfo.dst_ill;
2277 		}
2278 	} else {
2279 		ill = ILL_START_WALK_V6(&ctx, ipst);
2280 	}
2281 
2282 	while (ill != NULL) {
2283 		ASSERT(ill->ill_isv6);
2284 
2285 		/*
2286 		 * Avoid FAILED/OFFLINE ills.
2287 		 * Global and site local addresses will failover and
2288 		 * will be available on the new ill.
2289 		 * But link local addresses don't move.
2290 		 */
2291 		if (dstinfo.dst_restrict_ill != RESTRICT_TO_ILL &&
2292 		    ill->ill_phyint->phyint_flags &
2293 		    (PHYI_OFFLINE | PHYI_FAILED))
2294 			goto next_ill;
2295 
2296 		for (ipif = ill->ill_ipif; ipif != NULL;
2297 		    ipif = ipif->ipif_next) {
2298 
2299 			if (!IPIF_VALID_IPV6_SOURCE(ipif))
2300 				continue;
2301 
2302 			if (zoneid != ALL_ZONES &&
2303 			    ipif->ipif_zoneid != zoneid &&
2304 			    ipif->ipif_zoneid != ALL_ZONES)
2305 				continue;
2306 
2307 			/*
2308 			 * Check compatibility of local address for
2309 			 * destination's default label if we're on a labeled
2310 			 * system.  Incompatible addresses can't be used at
2311 			 * all and must be skipped over.
2312 			 */
2313 			if (dst_rhtp != NULL) {
2314 				boolean_t incompat;
2315 
2316 				src_rhtp = find_tpc(&ipif->ipif_v6lcl_addr,
2317 				    IPV6_VERSION, B_FALSE);
2318 				if (src_rhtp == NULL)
2319 					continue;
2320 				incompat =
2321 				    src_rhtp->tpc_tp.host_type != SUN_CIPSO ||
2322 				    src_rhtp->tpc_tp.tp_doi !=
2323 				    dst_rhtp->tpc_tp.tp_doi ||
2324 				    (!_blinrange(&dst_rhtp->tpc_tp.tp_def_label,
2325 				    &src_rhtp->tpc_tp.tp_sl_range_cipso) &&
2326 				    !blinlset(&dst_rhtp->tpc_tp.tp_def_label,
2327 				    src_rhtp->tpc_tp.tp_sl_set_cipso));
2328 				TPC_RELE(src_rhtp);
2329 				if (incompat)
2330 					continue;
2331 			}
2332 
2333 			if (first_candidate) {
2334 				/*
2335 				 * This is first valid address in the list.
2336 				 * It is automatically the best candidate
2337 				 * so far.
2338 				 */
2339 				best_c.cand_ipif = ipif;
2340 				first_candidate = B_FALSE;
2341 				continue;
2342 			}
2343 
2344 			bzero(&curr_c, sizeof (cand_t));
2345 			curr_c.cand_ipif = ipif;
2346 
2347 			/*
2348 			 * Compare this current candidate (curr_c) with the
2349 			 * best candidate (best_c) by applying the
2350 			 * comparison rules in order until one breaks the
2351 			 * tie.
2352 			 */
2353 			for (index = 0; rules[index] != NULL; index++) {
2354 				/* Apply a comparison rule. */
2355 				rule_result =
2356 				    (rules[index])(&best_c, &curr_c, &dstinfo,
2357 				    ipst);
2358 				if (rule_result == CAND_AVOID) {
2359 					/*
2360 					 * The best candidate is still the
2361 					 * best candidate.  Forget about
2362 					 * this current candidate and go on
2363 					 * to the next one.
2364 					 */
2365 					break;
2366 				} else if (rule_result == CAND_PREFER) {
2367 					/*
2368 					 * This candidate is prefered.  It
2369 					 * becomes the best candidate so
2370 					 * far.  Go on to the next address.
2371 					 */
2372 					best_c = curr_c;
2373 					break;
2374 				}
2375 				/* We have a tie, apply the next rule. */
2376 			}
2377 
2378 			/*
2379 			 * The last rule must be a tie breaker rule and
2380 			 * must never produce a tie.  At this point, the
2381 			 * candidate should have either been rejected, or
2382 			 * have been prefered as the best candidate so far.
2383 			 */
2384 			ASSERT(rule_result != CAND_TIE);
2385 		}
2386 
2387 		/*
2388 		 * We may be walking the linked-list of ill's in an
2389 		 * IPMP group or traversing the IPv6 ill avl tree. If it is a
2390 		 * usesrc ILL then it can't be part of IPMP group and we
2391 		 * will exit the while loop.
2392 		 */
2393 next_ill:
2394 		if (dstinfo.dst_restrict_ill == RESTRICT_TO_ILL)
2395 			ill = NULL;
2396 		else if (dstinfo.dst_restrict_ill == RESTRICT_TO_GROUP)
2397 			ill = ill->ill_group_next;
2398 		else
2399 			ill = ill_next(&ctx, ill);
2400 	}
2401 
2402 	ipif = best_c.cand_ipif;
2403 	ip1dbg(("ipif_select_source_v6(%s, %s) -> %s\n",
2404 	    dstinfo.dst_ill->ill_name,
2405 	    inet_ntop(AF_INET6, dstinfo.dst_addr, dstr, sizeof (dstr)),
2406 	    (ipif == NULL ? "NULL" :
2407 	    inet_ntop(AF_INET6, &ipif->ipif_v6lcl_addr, sstr, sizeof (sstr)))));
2408 
2409 	if (usesrc_ill != NULL)
2410 		ill_refrele(usesrc_ill);
2411 
2412 	if (dst_rhtp != NULL)
2413 		TPC_RELE(dst_rhtp);
2414 
2415 	if (ipif == NULL) {
2416 		rw_exit(&ipst->ips_ill_g_lock);
2417 		return (NULL);
2418 	}
2419 
2420 	mutex_enter(&ipif->ipif_ill->ill_lock);
2421 	if (IPIF_CAN_LOOKUP(ipif)) {
2422 		ipif_refhold_locked(ipif);
2423 		mutex_exit(&ipif->ipif_ill->ill_lock);
2424 		rw_exit(&ipst->ips_ill_g_lock);
2425 		return (ipif);
2426 	}
2427 	mutex_exit(&ipif->ipif_ill->ill_lock);
2428 	rw_exit(&ipst->ips_ill_g_lock);
2429 	ip1dbg(("ipif_select_source_v6 cannot lookup ipif %p"
2430 	    " returning null \n", (void *)ipif));
2431 
2432 	return (NULL);
2433 }
2434 
2435 /*
2436  * If old_ipif is not NULL, see if ipif was derived from old
2437  * ipif and if so, recreate the interface route by re-doing
2438  * source address selection. This happens when ipif_down ->
2439  * ipif_update_other_ipifs calls us.
2440  *
2441  * If old_ipif is NULL, just redo the source address selection
2442  * if needed. This happens when illgrp_insert or ipif_up_done_v6
2443  * calls us.
2444  */
2445 void
2446 ipif_recreate_interface_routes_v6(ipif_t *old_ipif, ipif_t *ipif)
2447 {
2448 	ire_t *ire;
2449 	ire_t *ipif_ire;
2450 	queue_t *stq;
2451 	ill_t *ill;
2452 	ipif_t *nipif = NULL;
2453 	boolean_t nipif_refheld = B_FALSE;
2454 	boolean_t ip6_asp_table_held = B_FALSE;
2455 	ip_stack_t	*ipst = ipif->ipif_ill->ill_ipst;
2456 
2457 	ill = ipif->ipif_ill;
2458 
2459 	if (!(ipif->ipif_flags &
2460 	    (IPIF_NOLOCAL|IPIF_ANYCAST|IPIF_DEPRECATED))) {
2461 		/*
2462 		 * Can't possibly have borrowed the source
2463 		 * from old_ipif.
2464 		 */
2465 		return;
2466 	}
2467 
2468 	/*
2469 	 * Is there any work to be done? No work if the address
2470 	 * is INADDR_ANY, loopback or NOLOCAL or ANYCAST (
2471 	 * ipif_select_source_v6() does not borrow addresses from
2472 	 * NOLOCAL and ANYCAST interfaces).
2473 	 */
2474 	if ((old_ipif != NULL) &&
2475 	    ((IN6_IS_ADDR_UNSPECIFIED(&old_ipif->ipif_v6lcl_addr)) ||
2476 	    (old_ipif->ipif_ill->ill_wq == NULL) ||
2477 	    (old_ipif->ipif_flags &
2478 	    (IPIF_NOLOCAL|IPIF_ANYCAST)))) {
2479 		return;
2480 	}
2481 
2482 	/*
2483 	 * Perform the same checks as when creating the
2484 	 * IRE_INTERFACE in ipif_up_done_v6.
2485 	 */
2486 	if (!(ipif->ipif_flags & IPIF_UP))
2487 		return;
2488 
2489 	if ((ipif->ipif_flags & IPIF_NOXMIT))
2490 		return;
2491 
2492 	if (IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6subnet) &&
2493 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6net_mask))
2494 		return;
2495 
2496 	/*
2497 	 * We know that ipif uses some other source for its
2498 	 * IRE_INTERFACE. Is it using the source of this
2499 	 * old_ipif?
2500 	 */
2501 	ipif_ire = ipif_to_ire_v6(ipif);
2502 	if (ipif_ire == NULL)
2503 		return;
2504 
2505 	if (old_ipif != NULL &&
2506 	    !IN6_ARE_ADDR_EQUAL(&old_ipif->ipif_v6lcl_addr,
2507 	    &ipif_ire->ire_src_addr_v6)) {
2508 		ire_refrele(ipif_ire);
2509 		return;
2510 	}
2511 
2512 	if (ip_debug > 2) {
2513 		/* ip1dbg */
2514 		pr_addr_dbg("ipif_recreate_interface_routes_v6: deleting IRE"
2515 		    " for src %s\n", AF_INET6, &ipif_ire->ire_src_addr_v6);
2516 	}
2517 
2518 	stq = ipif_ire->ire_stq;
2519 
2520 	/*
2521 	 * Can't use our source address. Select a different source address
2522 	 * for the IRE_INTERFACE.  We restrict interface route source
2523 	 * address selection to ipif's assigned to the same link as the
2524 	 * interface.
2525 	 */
2526 	if (ip6_asp_can_lookup(ipst)) {
2527 		ip6_asp_table_held = B_TRUE;
2528 		nipif = ipif_select_source_v6(ill, &ipif->ipif_v6subnet,
2529 		    RESTRICT_TO_GROUP, IPV6_PREFER_SRC_DEFAULT,
2530 		    ipif->ipif_zoneid);
2531 	}
2532 	if (nipif == NULL) {
2533 		/* Last resort - all ipif's have IPIF_NOLOCAL */
2534 		nipif = ipif;
2535 	} else {
2536 		nipif_refheld = B_TRUE;
2537 	}
2538 
2539 	ire = ire_create_v6(
2540 	    &ipif->ipif_v6subnet,	/* dest pref */
2541 	    &ipif->ipif_v6net_mask,	/* mask */
2542 	    &nipif->ipif_v6src_addr,	/* src addr */
2543 	    NULL,			/* no gateway */
2544 	    &ipif->ipif_mtu,		/* max frag */
2545 	    NULL,			/* no src nce */
2546 	    NULL,			/* no recv from queue */
2547 	    stq,			/* send-to queue */
2548 	    ill->ill_net_type,		/* IF_[NO]RESOLVER */
2549 	    ipif,
2550 	    NULL,
2551 	    0,
2552 	    0,
2553 	    0,
2554 	    &ire_uinfo_null,
2555 	    NULL,
2556 	    NULL,
2557 	    ipst);
2558 
2559 	if (ire != NULL) {
2560 		ire_t *ret_ire;
2561 		int   error;
2562 
2563 		/*
2564 		 * We don't need ipif_ire anymore. We need to delete
2565 		 * before we add so that ire_add does not detect
2566 		 * duplicates.
2567 		 */
2568 		ire_delete(ipif_ire);
2569 		ret_ire = ire;
2570 		error = ire_add(&ret_ire, NULL, NULL, NULL, B_FALSE);
2571 		ASSERT(error == 0);
2572 		ASSERT(ret_ire == ire);
2573 		if (ret_ire != NULL) {
2574 			/* Held in ire_add */
2575 			ire_refrele(ret_ire);
2576 		}
2577 	}
2578 	/*
2579 	 * Either we are falling through from above or could not
2580 	 * allocate a replacement.
2581 	 */
2582 	ire_refrele(ipif_ire);
2583 	if (ip6_asp_table_held)
2584 		ip6_asp_table_refrele(ipst);
2585 	if (nipif_refheld)
2586 		ipif_refrele(nipif);
2587 }
2588 
2589 /*
2590  * This old_ipif is going away.
2591  *
2592  * Determine if any other ipif's are using our address as
2593  * ipif_v6lcl_addr (due to those being IPIF_NOLOCAL, IPIF_ANYCAST, or
2594  * IPIF_DEPRECATED).
2595  * Find the IRE_INTERFACE for such ipif's and recreate them
2596  * to use an different source address following the rules in
2597  * ipif_up_done_v6.
2598  *
2599  * This function takes an illgrp as an argument so that illgrp_delete
2600  * can call this to update source address even after deleting the
2601  * old_ipif->ipif_ill from the ill group.
2602  */
2603 void
2604 ipif_update_other_ipifs_v6(ipif_t *old_ipif, ill_group_t *illgrp)
2605 {
2606 	ipif_t	*ipif;
2607 	ill_t	*ill;
2608 	char	buf[INET6_ADDRSTRLEN];
2609 
2610 	ASSERT(IAM_WRITER_IPIF(old_ipif));
2611 
2612 	ill = old_ipif->ipif_ill;
2613 
2614 	ip1dbg(("ipif_update_other_ipifs_v6(%s, %s)\n",
2615 	    ill->ill_name,
2616 	    inet_ntop(AF_INET6, &old_ipif->ipif_v6lcl_addr,
2617 	    buf, sizeof (buf))));
2618 
2619 	/*
2620 	 * If this part of a group, look at all ills as ipif_select_source
2621 	 * borrows a source address across all the ills in the group.
2622 	 */
2623 	if (illgrp != NULL)
2624 		ill = illgrp->illgrp_ill;
2625 
2626 	/* Don't need a lock since this is a writer */
2627 	for (; ill != NULL; ill = ill->ill_group_next) {
2628 		for (ipif = ill->ill_ipif; ipif != NULL;
2629 		    ipif = ipif->ipif_next) {
2630 
2631 			if (ipif == old_ipif)
2632 				continue;
2633 
2634 			ipif_recreate_interface_routes_v6(old_ipif, ipif);
2635 		}
2636 	}
2637 }
2638 
2639 /*
2640  * Perform an attach and bind to get phys addr plus info_req for
2641  * the physical device.
2642  * q and mp represents an ioctl which will be queued waiting for
2643  * completion of the DLPI message exchange.
2644  * MUST be called on an ill queue. Can not set conn_pending_ill for that
2645  * reason thus the DL_PHYS_ADDR_ACK code does not assume ill_pending_q.
2646  *
2647  * Returns EINPROGRESS when mp has been consumed by queueing it on
2648  * ill_pending_mp and the ioctl will complete in ip_rput.
2649  */
2650 int
2651 ill_dl_phys(ill_t *ill, ipif_t *ipif, mblk_t *mp, queue_t *q)
2652 {
2653 	mblk_t	*v6token_mp = NULL;
2654 	mblk_t	*v6lla_mp = NULL;
2655 	mblk_t	*phys_mp = NULL;
2656 	mblk_t	*info_mp = NULL;
2657 	mblk_t	*attach_mp = NULL;
2658 	mblk_t	*bind_mp = NULL;
2659 	mblk_t	*unbind_mp = NULL;
2660 	mblk_t	*notify_mp = NULL;
2661 
2662 	ip1dbg(("ill_dl_phys(%s:%u)\n", ill->ill_name, ipif->ipif_id));
2663 	ASSERT(ill->ill_dlpi_style_set);
2664 	ASSERT(WR(q)->q_next != NULL);
2665 
2666 	if (ill->ill_isv6) {
2667 		v6token_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2668 		    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2669 		if (v6token_mp == NULL)
2670 			goto bad;
2671 		((dl_phys_addr_req_t *)v6token_mp->b_rptr)->dl_addr_type =
2672 		    DL_IPV6_TOKEN;
2673 
2674 		v6lla_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2675 		    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2676 		if (v6lla_mp == NULL)
2677 			goto bad;
2678 		((dl_phys_addr_req_t *)v6lla_mp->b_rptr)->dl_addr_type =
2679 		    DL_IPV6_LINK_LAYER_ADDR;
2680 	}
2681 
2682 	/*
2683 	 * Allocate a DL_NOTIFY_REQ and set the notifications we want.
2684 	 */
2685 	notify_mp = ip_dlpi_alloc(sizeof (dl_notify_req_t) + sizeof (long),
2686 	    DL_NOTIFY_REQ);
2687 	if (notify_mp == NULL)
2688 		goto bad;
2689 	((dl_notify_req_t *)notify_mp->b_rptr)->dl_notifications =
2690 	    (DL_NOTE_PHYS_ADDR | DL_NOTE_SDU_SIZE | DL_NOTE_FASTPATH_FLUSH |
2691 	    DL_NOTE_LINK_UP | DL_NOTE_LINK_DOWN | DL_NOTE_CAPAB_RENEG);
2692 
2693 	phys_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2694 	    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2695 	if (phys_mp == NULL)
2696 		goto bad;
2697 	((dl_phys_addr_req_t *)phys_mp->b_rptr)->dl_addr_type =
2698 	    DL_CURR_PHYS_ADDR;
2699 
2700 	info_mp = ip_dlpi_alloc(
2701 	    sizeof (dl_info_req_t) + sizeof (dl_info_ack_t),
2702 	    DL_INFO_REQ);
2703 	if (info_mp == NULL)
2704 		goto bad;
2705 
2706 	bind_mp = ip_dlpi_alloc(sizeof (dl_bind_req_t) + sizeof (long),
2707 	    DL_BIND_REQ);
2708 	if (bind_mp == NULL)
2709 		goto bad;
2710 	((dl_bind_req_t *)bind_mp->b_rptr)->dl_sap = ill->ill_sap;
2711 	((dl_bind_req_t *)bind_mp->b_rptr)->dl_service_mode = DL_CLDLS;
2712 
2713 	unbind_mp = ip_dlpi_alloc(sizeof (dl_unbind_req_t), DL_UNBIND_REQ);
2714 	if (unbind_mp == NULL)
2715 		goto bad;
2716 
2717 	/* If we need to attach, pre-alloc and initialize the mblk */
2718 	if (ill->ill_needs_attach) {
2719 		attach_mp = ip_dlpi_alloc(sizeof (dl_attach_req_t),
2720 		    DL_ATTACH_REQ);
2721 		if (attach_mp == NULL)
2722 			goto bad;
2723 		((dl_attach_req_t *)attach_mp->b_rptr)->dl_ppa = ill->ill_ppa;
2724 	}
2725 
2726 	/*
2727 	 * Here we are going to delay the ioctl ack until after
2728 	 * ACKs from DL_PHYS_ADDR_REQ. So need to save the
2729 	 * original ioctl message before sending the requests
2730 	 */
2731 	mutex_enter(&ill->ill_lock);
2732 	/* ipsq_pending_mp_add won't fail since we pass in a NULL connp */
2733 	(void) ipsq_pending_mp_add(NULL, ipif, ill->ill_wq, mp, 0);
2734 	/*
2735 	 * Set ill_phys_addr_pend to zero. It will be set to the addr_type of
2736 	 * the DL_PHYS_ADDR_REQ in ill_dlpi_send() and ill_dlpi_done(). It will
2737 	 * be used to track which DL_PHYS_ADDR_REQ is being ACK'd/NAK'd.
2738 	 */
2739 	ill->ill_phys_addr_pend = 0;
2740 	mutex_exit(&ill->ill_lock);
2741 
2742 	if (attach_mp != NULL) {
2743 		ip1dbg(("ill_dl_phys: attach\n"));
2744 		ill_dlpi_send(ill, attach_mp);
2745 	}
2746 	ill_dlpi_send(ill, bind_mp);
2747 	ill_dlpi_send(ill, info_mp);
2748 	if (ill->ill_isv6) {
2749 		ill_dlpi_send(ill, v6token_mp);
2750 		ill_dlpi_send(ill, v6lla_mp);
2751 	}
2752 	ill_dlpi_send(ill, phys_mp);
2753 	ill_dlpi_send(ill, notify_mp);
2754 	ill_dlpi_send(ill, unbind_mp);
2755 
2756 	/*
2757 	 * This operation will complete in ip_rput_dlpi_writer with either
2758 	 * a DL_PHYS_ADDR_ACK or DL_ERROR_ACK.
2759 	 */
2760 	return (EINPROGRESS);
2761 bad:
2762 	freemsg(v6token_mp);
2763 	freemsg(v6lla_mp);
2764 	freemsg(phys_mp);
2765 	freemsg(info_mp);
2766 	freemsg(attach_mp);
2767 	freemsg(bind_mp);
2768 	freemsg(unbind_mp);
2769 	freemsg(notify_mp);
2770 	return (ENOMEM);
2771 }
2772 
2773 uint_t ip_loopback_mtu_v6plus = IP_LOOPBACK_MTU + IPV6_HDR_LEN + 20;
2774 
2775 /*
2776  * DLPI is up.
2777  * Create all the IREs associated with an interface bring up multicast.
2778  * Set the interface flag and finish other initialization
2779  * that potentially had to be differed to after DL_BIND_ACK.
2780  */
2781 int
2782 ipif_up_done_v6(ipif_t *ipif)
2783 {
2784 	ire_t	*ire_array[20];
2785 	ire_t	**irep = ire_array;
2786 	ire_t	**irep1;
2787 	ill_t	*ill = ipif->ipif_ill;
2788 	queue_t	*stq;
2789 	in6_addr_t	v6addr;
2790 	in6_addr_t	route_mask;
2791 	ipif_t	 *src_ipif = NULL;
2792 	ipif_t   *tmp_ipif;
2793 	boolean_t	flush_ire_cache = B_TRUE;
2794 	int	err;
2795 	char	buf[INET6_ADDRSTRLEN];
2796 	phyint_t *phyi;
2797 	ire_t	**ipif_saved_irep = NULL;
2798 	int ipif_saved_ire_cnt;
2799 	int cnt;
2800 	boolean_t src_ipif_held = B_FALSE;
2801 	boolean_t ire_added = B_FALSE;
2802 	boolean_t loopback = B_FALSE;
2803 	boolean_t ip6_asp_table_held = B_FALSE;
2804 	ip_stack_t	*ipst = ill->ill_ipst;
2805 
2806 	ip1dbg(("ipif_up_done_v6(%s:%u)\n",
2807 	    ipif->ipif_ill->ill_name, ipif->ipif_id));
2808 
2809 	/* Check if this is a loopback interface */
2810 	if (ipif->ipif_ill->ill_wq == NULL)
2811 		loopback = B_TRUE;
2812 
2813 	ASSERT(ipif->ipif_isv6);
2814 	ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
2815 
2816 	/*
2817 	 * If all other interfaces for this ill are down or DEPRECATED,
2818 	 * or otherwise unsuitable for source address selection, remove
2819 	 * any IRE_CACHE entries for this ill to make sure source
2820 	 * address selection gets to take this new ipif into account.
2821 	 * No need to hold ill_lock while traversing the ipif list since
2822 	 * we are writer
2823 	 */
2824 	for (tmp_ipif = ill->ill_ipif; tmp_ipif;
2825 	    tmp_ipif = tmp_ipif->ipif_next) {
2826 		if (((tmp_ipif->ipif_flags &
2827 		    (IPIF_NOXMIT|IPIF_ANYCAST|IPIF_NOLOCAL|IPIF_DEPRECATED)) ||
2828 		    !(tmp_ipif->ipif_flags & IPIF_UP)) ||
2829 		    (tmp_ipif == ipif))
2830 			continue;
2831 		/* first useable pre-existing interface */
2832 		flush_ire_cache = B_FALSE;
2833 		break;
2834 	}
2835 	if (flush_ire_cache)
2836 		ire_walk_ill_v6(MATCH_IRE_ILL_GROUP | MATCH_IRE_TYPE,
2837 		    IRE_CACHE, ill_ipif_cache_delete, (char *)ill, ill);
2838 
2839 	/*
2840 	 * Figure out which way the send-to queue should go.  Only
2841 	 * IRE_IF_RESOLVER or IRE_IF_NORESOLVER should show up here.
2842 	 */
2843 	switch (ill->ill_net_type) {
2844 	case IRE_IF_RESOLVER:
2845 		stq = ill->ill_rq;
2846 		break;
2847 	case IRE_IF_NORESOLVER:
2848 	case IRE_LOOPBACK:
2849 		stq = ill->ill_wq;
2850 		break;
2851 	default:
2852 		return (EINVAL);
2853 	}
2854 
2855 	if (IS_LOOPBACK(ill)) {
2856 		/*
2857 		 * lo0:1 and subsequent ipifs were marked IRE_LOCAL in
2858 		 * ipif_lookup_on_name(), but in the case of zones we can have
2859 		 * several loopback addresses on lo0. So all the interfaces with
2860 		 * loopback addresses need to be marked IRE_LOOPBACK.
2861 		 */
2862 		if (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr, &ipv6_loopback))
2863 			ipif->ipif_ire_type = IRE_LOOPBACK;
2864 		else
2865 			ipif->ipif_ire_type = IRE_LOCAL;
2866 	}
2867 
2868 	if (ipif->ipif_flags & (IPIF_NOLOCAL|IPIF_ANYCAST|IPIF_DEPRECATED)) {
2869 		/*
2870 		 * Can't use our source address. Select a different
2871 		 * source address for the IRE_INTERFACE and IRE_LOCAL
2872 		 */
2873 		if (ip6_asp_can_lookup(ipst)) {
2874 			ip6_asp_table_held = B_TRUE;
2875 			src_ipif = ipif_select_source_v6(ipif->ipif_ill,
2876 			    &ipif->ipif_v6subnet, RESTRICT_TO_NONE,
2877 			    IPV6_PREFER_SRC_DEFAULT, ipif->ipif_zoneid);
2878 		}
2879 		if (src_ipif == NULL)
2880 			src_ipif = ipif;	/* Last resort */
2881 		else
2882 			src_ipif_held = B_TRUE;
2883 	} else {
2884 		src_ipif = ipif;
2885 	}
2886 
2887 	if (!IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr) &&
2888 	    !(ipif->ipif_flags & IPIF_NOLOCAL)) {
2889 
2890 		/*
2891 		 * If we're on a labeled system then make sure that zone-
2892 		 * private addresses have proper remote host database entries.
2893 		 */
2894 		if (is_system_labeled() &&
2895 		    ipif->ipif_ire_type != IRE_LOOPBACK) {
2896 			if (ip6opt_ls == 0) {
2897 				cmn_err(CE_WARN, "IPv6 not enabled "
2898 				    "via /etc/system");
2899 				return (EINVAL);
2900 			}
2901 			if (!tsol_check_interface_address(ipif))
2902 				return (EINVAL);
2903 		}
2904 
2905 		/* Register the source address for __sin6_src_id */
2906 		err = ip_srcid_insert(&ipif->ipif_v6lcl_addr,
2907 		    ipif->ipif_zoneid, ipst);
2908 		if (err != 0) {
2909 			ip0dbg(("ipif_up_done_v6: srcid_insert %d\n", err));
2910 			if (src_ipif_held)
2911 				ipif_refrele(src_ipif);
2912 			if (ip6_asp_table_held)
2913 				ip6_asp_table_refrele(ipst);
2914 			return (err);
2915 		}
2916 		/*
2917 		 * If the interface address is set, create the LOCAL
2918 		 * or LOOPBACK IRE.
2919 		 */
2920 		ip1dbg(("ipif_up_done_v6: creating IRE %d for %s\n",
2921 		    ipif->ipif_ire_type,
2922 		    inet_ntop(AF_INET6, &ipif->ipif_v6lcl_addr,
2923 		    buf, sizeof (buf))));
2924 
2925 		*irep++ = ire_create_v6(
2926 		    &ipif->ipif_v6lcl_addr,		/* dest address */
2927 		    &ipv6_all_ones,			/* mask */
2928 		    &src_ipif->ipif_v6src_addr,		/* source address */
2929 		    NULL,				/* no gateway */
2930 		    &ip_loopback_mtu_v6plus,		/* max frag size */
2931 		    NULL,
2932 		    ipif->ipif_rq,			/* recv-from queue */
2933 		    NULL,				/* no send-to queue */
2934 		    ipif->ipif_ire_type,		/* LOCAL or LOOPBACK */
2935 		    ipif,				/* interface */
2936 		    NULL,
2937 		    0,
2938 		    0,
2939 		    (ipif->ipif_flags & IPIF_PRIVATE) ? RTF_PRIVATE : 0,
2940 		    &ire_uinfo_null,
2941 		    NULL,
2942 		    NULL,
2943 		    ipst);
2944 	}
2945 
2946 	/*
2947 	 * Set up the IRE_IF_RESOLVER or IRE_IF_NORESOLVER, as appropriate.
2948 	 * Note that atun interfaces have an all-zero ipif_v6subnet.
2949 	 * Thus we allow a zero subnet as long as the mask is non-zero.
2950 	 */
2951 	if (stq != NULL && !(ipif->ipif_flags & IPIF_NOXMIT) &&
2952 	    !(IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6subnet) &&
2953 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6net_mask))) {
2954 		/* ipif_v6subnet is ipif_v6pp_dst_addr for pt-pt */
2955 		v6addr = ipif->ipif_v6subnet;
2956 
2957 		if (ipif->ipif_flags & IPIF_POINTOPOINT) {
2958 			route_mask = ipv6_all_ones;
2959 		} else {
2960 			route_mask = ipif->ipif_v6net_mask;
2961 		}
2962 
2963 		ip1dbg(("ipif_up_done_v6: creating if IRE %d for %s\n",
2964 		    ill->ill_net_type,
2965 		    inet_ntop(AF_INET6, &v6addr, buf, sizeof (buf))));
2966 
2967 		*irep++ = ire_create_v6(
2968 		    &v6addr,			/* dest pref */
2969 		    &route_mask,		/* mask */
2970 		    &src_ipif->ipif_v6src_addr,	/* src addr */
2971 		    NULL,			/* no gateway */
2972 		    &ipif->ipif_mtu,		/* max frag */
2973 		    NULL,			/* no src nce */
2974 		    NULL,			/* no recv from queue */
2975 		    stq,			/* send-to queue */
2976 		    ill->ill_net_type,		/* IF_[NO]RESOLVER */
2977 		    ipif,
2978 		    NULL,
2979 		    0,
2980 		    0,
2981 		    (ipif->ipif_flags & IPIF_PRIVATE) ? RTF_PRIVATE : 0,
2982 		    &ire_uinfo_null,
2983 		    NULL,
2984 		    NULL,
2985 		    ipst);
2986 	}
2987 
2988 	/*
2989 	 * Setup 2002::/16 route, if this interface is a 6to4 tunnel
2990 	 */
2991 	if (IN6_IS_ADDR_6TO4(&ipif->ipif_v6lcl_addr) &&
2992 	    (ill->ill_is_6to4tun)) {
2993 		/*
2994 		 * Destination address is 2002::/16
2995 		 */
2996 #ifdef	_BIG_ENDIAN
2997 		const in6_addr_t prefix_addr = { 0x20020000U, 0, 0, 0 };
2998 		const in6_addr_t prefix_mask = { 0xffff0000U, 0, 0, 0 };
2999 #else
3000 		const in6_addr_t prefix_addr = { 0x00000220U, 0, 0, 0 };
3001 		const in6_addr_t prefix_mask = { 0x0000ffffU, 0, 0, 0 };
3002 #endif /* _BIG_ENDIAN */
3003 		char	buf2[INET6_ADDRSTRLEN];
3004 		ire_t *isdup;
3005 		in6_addr_t *first_addr = &ill->ill_ipif->ipif_v6lcl_addr;
3006 
3007 		/*
3008 		 * check to see if this route has already been added for
3009 		 * this tunnel interface.
3010 		 */
3011 		isdup = ire_ftable_lookup_v6(first_addr, &prefix_mask, 0,
3012 		    IRE_IF_NORESOLVER, ill->ill_ipif, NULL, ALL_ZONES, 0, NULL,
3013 		    (MATCH_IRE_SRC | MATCH_IRE_MASK), ipst);
3014 
3015 		if (isdup == NULL) {
3016 			ip1dbg(("ipif_up_done_v6: creating if IRE %d for %s",
3017 			    IRE_IF_NORESOLVER, inet_ntop(AF_INET6, &v6addr,
3018 			    buf2, sizeof (buf2))));
3019 
3020 			*irep++ = ire_create_v6(
3021 			    &prefix_addr,		/* 2002:: */
3022 			    &prefix_mask,		/* ffff:: */
3023 			    &ipif->ipif_v6lcl_addr, 	/* src addr */
3024 			    NULL, 			/* gateway */
3025 			    &ipif->ipif_mtu, 		/* max_frag */
3026 			    NULL, 			/* no src nce */
3027 			    NULL, 			/* no rfq */
3028 			    ill->ill_wq, 		/* stq */
3029 			    IRE_IF_NORESOLVER,		/* type */
3030 			    ipif,			/* interface */
3031 			    NULL,			/* v6cmask */
3032 			    0,
3033 			    0,
3034 			    RTF_UP,
3035 			    &ire_uinfo_null,
3036 			    NULL,
3037 			    NULL,
3038 			    ipst);
3039 		} else {
3040 			ire_refrele(isdup);
3041 		}
3042 	}
3043 
3044 	/* If an earlier ire_create failed, get out now */
3045 	for (irep1 = irep; irep1 > ire_array; ) {
3046 		irep1--;
3047 		if (*irep1 == NULL) {
3048 			ip1dbg(("ipif_up_done_v6: NULL ire found in"
3049 			    " ire_array\n"));
3050 			err = ENOMEM;
3051 			goto bad;
3052 		}
3053 	}
3054 
3055 	ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
3056 
3057 	/*
3058 	 * Need to atomically check for ip_addr_availablity_check
3059 	 * now under ill_g_lock, and if it fails got bad, and remove
3060 	 * from group also
3061 	 */
3062 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
3063 	mutex_enter(&ipst->ips_ip_addr_avail_lock);
3064 	ill->ill_ipif_up_count++;
3065 	ipif->ipif_flags |= IPIF_UP;
3066 	err = ip_addr_availability_check(ipif);
3067 	mutex_exit(&ipst->ips_ip_addr_avail_lock);
3068 	rw_exit(&ipst->ips_ill_g_lock);
3069 
3070 	if (err != 0) {
3071 		/*
3072 		 * Our address may already be up on the same ill. In this case,
3073 		 * the external resolver entry for our ipif replaced the one for
3074 		 * the other ipif. So we don't want to delete it (otherwise the
3075 		 * other ipif would be unable to send packets).
3076 		 * ip_addr_availability_check() identifies this case for us and
3077 		 * returns EADDRINUSE; we need to turn it into EADDRNOTAVAIL
3078 		 * which is the expected error code.
3079 		 */
3080 		if (err == EADDRINUSE) {
3081 			if (ipif->ipif_ill->ill_flags & ILLF_XRESOLV) {
3082 				freemsg(ipif->ipif_arp_del_mp);
3083 				ipif->ipif_arp_del_mp = NULL;
3084 			}
3085 			err = EADDRNOTAVAIL;
3086 		}
3087 		ill->ill_ipif_up_count--;
3088 		ipif->ipif_flags &= ~IPIF_UP;
3089 		goto bad;
3090 	}
3091 
3092 	/*
3093 	 * Add in all newly created IREs. We want to add before
3094 	 * we call ifgrp_insert which wants to know whether
3095 	 * IRE_IF_RESOLVER exists or not.
3096 	 *
3097 	 * NOTE : We refrele the ire though we may branch to "bad"
3098 	 *	  later on where we do ire_delete. This is okay
3099 	 *	  because nobody can delete it as we are running
3100 	 *	  exclusively.
3101 	 */
3102 	for (irep1 = irep; irep1 > ire_array; ) {
3103 		irep1--;
3104 		/* Shouldn't be adding any bcast ire's */
3105 		ASSERT((*irep1)->ire_type != IRE_BROADCAST);
3106 		ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
3107 		/*
3108 		 * refheld by ire_add. refele towards the end of the func
3109 		 */
3110 		(void) ire_add(irep1, NULL, NULL, NULL, B_FALSE);
3111 	}
3112 	if (ip6_asp_table_held) {
3113 		ip6_asp_table_refrele(ipst);
3114 		ip6_asp_table_held = B_FALSE;
3115 	}
3116 	ire_added = B_TRUE;
3117 
3118 	/*
3119 	 * Form groups if possible.
3120 	 *
3121 	 * If we are supposed to be in a ill_group with a name, insert it
3122 	 * now as we know that at least one ipif is UP. Otherwise form
3123 	 * nameless groups.
3124 	 *
3125 	 * If ip_enable_group_ifs is set and ipif address is not ::0, insert
3126 	 * this ipif into the appropriate interface group, or create a
3127 	 * new one. If this is already in a nameless group, we try to form
3128 	 * a bigger group looking at other ills potentially sharing this
3129 	 * ipif's prefix.
3130 	 */
3131 	phyi = ill->ill_phyint;
3132 	if (phyi->phyint_groupname_len != 0) {
3133 		ASSERT(phyi->phyint_groupname != NULL);
3134 		if (ill->ill_ipif_up_count == 1) {
3135 			ASSERT(ill->ill_group == NULL);
3136 			err = illgrp_insert(&ipst->ips_illgrp_head_v6, ill,
3137 			    phyi->phyint_groupname, NULL, B_TRUE);
3138 			if (err != 0) {
3139 				ip1dbg(("ipif_up_done_v6: illgrp allocation "
3140 				    "failed, error %d\n", err));
3141 				goto bad;
3142 			}
3143 		}
3144 		ASSERT(ill->ill_group != NULL);
3145 	}
3146 
3147 	/* Recover any additional IRE_IF_[NO]RESOLVER entries for this ipif */
3148 	ipif_saved_ire_cnt = ipif->ipif_saved_ire_cnt;
3149 	ipif_saved_irep = ipif_recover_ire_v6(ipif);
3150 
3151 	if (ipif->ipif_ipif_up_count == 1 && !loopback) {
3152 		/*
3153 		 * Need to recover all multicast memberships in the driver.
3154 		 * This had to be deferred until we had attached.
3155 		 */
3156 		ill_recover_multicast(ill);
3157 	}
3158 	/* Join the allhosts multicast address and the solicited node MC */
3159 	ipif_multicast_up(ipif);
3160 
3161 	if (!loopback) {
3162 		/*
3163 		 * See whether anybody else would benefit from the
3164 		 * new ipif that we added. We call this always rather
3165 		 * than while adding a non-IPIF_NOLOCAL/DEPRECATED/ANYCAST
3166 		 * ipif for the benefit of illgrp_insert (done above)
3167 		 * which does not do source address selection as it does
3168 		 * not want to re-create interface routes that we are
3169 		 * having reference to it here.
3170 		 */
3171 		ill_update_source_selection(ill);
3172 	}
3173 
3174 	for (irep1 = irep; irep1 > ire_array; ) {
3175 		irep1--;
3176 		if (*irep1 != NULL) {
3177 			/* was held in ire_add */
3178 			ire_refrele(*irep1);
3179 		}
3180 	}
3181 
3182 	cnt = ipif_saved_ire_cnt;
3183 	for (irep1 = ipif_saved_irep; cnt > 0; irep1++, cnt--) {
3184 		if (*irep1 != NULL) {
3185 			/* was held in ire_add */
3186 			ire_refrele(*irep1);
3187 		}
3188 	}
3189 
3190 	if (ipif->ipif_addr_ready) {
3191 		ip_rts_ifmsg(ipif);
3192 		ip_rts_newaddrmsg(RTM_ADD, 0, ipif);
3193 		sctp_update_ipif(ipif, SCTP_IPIF_UP);
3194 	}
3195 
3196 	if (ipif_saved_irep != NULL) {
3197 		kmem_free(ipif_saved_irep,
3198 		    ipif_saved_ire_cnt * sizeof (ire_t *));
3199 	}
3200 
3201 	if (src_ipif_held)
3202 		ipif_refrele(src_ipif);
3203 	return (0);
3204 
3205 bad:
3206 	if (ip6_asp_table_held)
3207 		ip6_asp_table_refrele(ipst);
3208 	/*
3209 	 * We don't have to bother removing from ill groups because
3210 	 *
3211 	 * 1) For groups with names, we insert only when the first ipif
3212 	 *    comes up. In that case if it fails, it will not be in any
3213 	 *    group. So, we need not try to remove for that case.
3214 	 *
3215 	 * 2) For groups without names, either we tried to insert ipif_ill
3216 	 *    in a group as singleton or found some other group to become
3217 	 *    a bigger group. For the former, if it fails we don't have
3218 	 *    anything to do as ipif_ill is not in the group and for the
3219 	 *    latter, there are no failures in illgrp_insert/illgrp_delete
3220 	 *    (ENOMEM can't occur for this. Check ifgrp_insert).
3221 	 */
3222 
3223 	while (irep > ire_array) {
3224 		irep--;
3225 		if (*irep != NULL) {
3226 			ire_delete(*irep);
3227 			if (ire_added)
3228 				ire_refrele(*irep);
3229 		}
3230 
3231 	}
3232 	(void) ip_srcid_remove(&ipif->ipif_v6lcl_addr, ipif->ipif_zoneid, ipst);
3233 
3234 	if (ipif_saved_irep != NULL) {
3235 		kmem_free(ipif_saved_irep,
3236 		    ipif_saved_ire_cnt * sizeof (ire_t *));
3237 	}
3238 	if (src_ipif_held)
3239 		ipif_refrele(src_ipif);
3240 
3241 	ipif_ndp_down(ipif);
3242 	if (ipif->ipif_ill->ill_flags & ILLF_XRESOLV)
3243 		ipif_arp_down(ipif);
3244 
3245 	return (err);
3246 }
3247 
3248 /*
3249  * Delete an ND entry and the corresponding IRE_CACHE entry if it exists.
3250  */
3251 /* ARGSUSED */
3252 int
3253 ip_siocdelndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3254     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3255 {
3256 	in6_addr_t	addr;
3257 	sin6_t		*sin6;
3258 	nce_t		*nce;
3259 	struct lifreq	*lifr;
3260 	lif_nd_req_t	*lnr;
3261 	mblk_t	*mp1;
3262 
3263 	mp1 = mp->b_cont->b_cont;
3264 	lifr = (struct lifreq *)mp1->b_rptr;
3265 	lnr = &lifr->lifr_nd;
3266 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3267 	if (ipif->ipif_id != 0)
3268 		return (EINVAL);
3269 
3270 	if (!ipif->ipif_isv6)
3271 		return (EINVAL);
3272 
3273 	if (lnr->lnr_addr.ss_family != AF_INET6)
3274 		return (EAFNOSUPPORT);
3275 
3276 	sin6 = (sin6_t *)&lnr->lnr_addr;
3277 	addr = sin6->sin6_addr;
3278 	nce = ndp_lookup_v6(ipif->ipif_ill, &addr, B_FALSE);
3279 	if (nce == NULL)
3280 		return (ESRCH);
3281 	ndp_delete(nce);
3282 	NCE_REFRELE(nce);
3283 	return (0);
3284 }
3285 
3286 /*
3287  * Return nbr cache info.
3288  */
3289 /* ARGSUSED */
3290 int
3291 ip_siocqueryndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3292     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3293 {
3294 	ill_t		*ill = ipif->ipif_ill;
3295 	struct lifreq	*lifr;
3296 	lif_nd_req_t	*lnr;
3297 
3298 	lifr = (struct lifreq *)mp->b_cont->b_cont->b_rptr;
3299 	lnr = &lifr->lifr_nd;
3300 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3301 	if (ipif->ipif_id != 0)
3302 		return (EINVAL);
3303 
3304 	if (!ipif->ipif_isv6)
3305 		return (EINVAL);
3306 
3307 	if (lnr->lnr_addr.ss_family != AF_INET6)
3308 		return (EAFNOSUPPORT);
3309 
3310 	if (ill->ill_phys_addr_length > sizeof (lnr->lnr_hdw_addr))
3311 		return (EINVAL);
3312 
3313 	return (ndp_query(ill, lnr));
3314 }
3315 
3316 /*
3317  * Perform an update of the nd entry for the specified address.
3318  */
3319 /* ARGSUSED */
3320 int
3321 ip_siocsetndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3322     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3323 {
3324 	ill_t		*ill = ipif->ipif_ill;
3325 	struct	lifreq	*lifr;
3326 	lif_nd_req_t	*lnr;
3327 
3328 	ASSERT(!(q->q_flag & QREADR) && q->q_next == NULL);
3329 
3330 	lifr = (struct lifreq *)mp->b_cont->b_cont->b_rptr;
3331 	lnr = &lifr->lifr_nd;
3332 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3333 	if (ipif->ipif_id != 0)
3334 		return (EINVAL);
3335 
3336 	if (!ipif->ipif_isv6)
3337 		return (EINVAL);
3338 
3339 	if (lnr->lnr_addr.ss_family != AF_INET6)
3340 		return (EAFNOSUPPORT);
3341 
3342 	return (ndp_sioc_update(ill, lnr));
3343 }
3344