xref: /illumos-gate/usr/src/uts/common/inet/ip/ip6_if.c (revision 8b80e8cb6855118d46f605e91b5ed4ce83417395)
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 /*
30  * This file contains the interface control functions for IPv6.
31  */
32 
33 #include <sys/types.h>
34 #include <sys/sysmacros.h>
35 #include <sys/stream.h>
36 #include <sys/dlpi.h>
37 #include <sys/stropts.h>
38 #include <sys/ddi.h>
39 #include <sys/cmn_err.h>
40 #include <sys/kstat.h>
41 #include <sys/debug.h>
42 #include <sys/zone.h>
43 #include <sys/policy.h>
44 
45 #include <sys/systm.h>
46 #include <sys/param.h>
47 #include <sys/socket.h>
48 #include <sys/isa_defs.h>
49 #include <net/if.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 #include <netinet/in.h>
53 #include <netinet/igmp_var.h>
54 #include <netinet/ip6.h>
55 #include <netinet/icmp6.h>
56 #include <netinet/in.h>
57 
58 #include <inet/common.h>
59 #include <inet/nd.h>
60 #include <inet/mib2.h>
61 #include <inet/ip.h>
62 #include <inet/ip6.h>
63 #include <inet/ip_multi.h>
64 #include <inet/ip_ire.h>
65 #include <inet/ip_rts.h>
66 #include <inet/ip_ndp.h>
67 #include <inet/ip_if.h>
68 #include <inet/ip6_asp.h>
69 #include <inet/tun.h>
70 #include <inet/ipclassifier.h>
71 #include <inet/sctp_ip.h>
72 
73 #include <sys/tsol/tndb.h>
74 #include <sys/tsol/tnet.h>
75 
76 static in6_addr_t	ipv6_ll_template =
77 			{(uint32_t)V6_LINKLOCAL, 0x0, 0x0, 0x0};
78 
79 static ipif_t *
80 ipif_lookup_interface_v6(const in6_addr_t *if_addr, const in6_addr_t *dst,
81     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst);
82 
83 /*
84  * These two functions, ipif_lookup_group_v6() and ill_lookup_group_v6(),
85  * are called when an application does not specify an interface to be
86  * used for multicast traffic.  It calls ire_lookup_multi_v6() to look
87  * for an interface route for the specified multicast group.  Doing
88  * this allows the administrator to add prefix routes for multicast to
89  * indicate which interface to be used for multicast traffic in the above
90  * scenario.  The route could be for all multicast (ff00::/8), for a single
91  * multicast group (a /128 route) or anything in between.  If there is no
92  * such multicast route, we just find any multicast capable interface and
93  * return it.
94  */
95 ipif_t *
96 ipif_lookup_group_v6(const in6_addr_t *group, zoneid_t zoneid, ip_stack_t *ipst)
97 {
98 	ire_t	*ire;
99 	ipif_t	*ipif;
100 
101 	ire = ire_lookup_multi_v6(group, zoneid, ipst);
102 	if (ire != NULL) {
103 		ipif = ire->ire_ipif;
104 		ipif_refhold(ipif);
105 		ire_refrele(ire);
106 		return (ipif);
107 	}
108 
109 	return (ipif_lookup_multicast(ipst, zoneid, B_TRUE));
110 }
111 
112 ill_t *
113 ill_lookup_group_v6(const in6_addr_t *group, zoneid_t zoneid, ip_stack_t *ipst)
114 {
115 	ire_t	*ire;
116 	ill_t	*ill;
117 	ipif_t	*ipif;
118 
119 	ire = ire_lookup_multi_v6(group, zoneid, ipst);
120 	if (ire != NULL) {
121 		ill = ire->ire_ipif->ipif_ill;
122 		ill_refhold(ill);
123 		ire_refrele(ire);
124 		return (ill);
125 	}
126 
127 	ipif = ipif_lookup_multicast(ipst, zoneid, B_TRUE);
128 	if (ipif == NULL)
129 		return (NULL);
130 
131 	ill = ipif->ipif_ill;
132 	ill_refhold(ill);
133 	ipif_refrele(ipif);
134 	return (ill);
135 }
136 
137 /*
138  * Look for an ipif with the specified interface address and destination.
139  * The destination address is used only for matching point-to-point interfaces.
140  */
141 static ipif_t *
142 ipif_lookup_interface_v6(const in6_addr_t *if_addr, const in6_addr_t *dst,
143     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst)
144 {
145 	ipif_t	*ipif;
146 	ill_t	*ill;
147 	ipsq_t	*ipsq;
148 	ill_walk_context_t ctx;
149 
150 	if (error != NULL)
151 		*error = 0;
152 
153 	/*
154 	 * First match all the point-to-point interfaces
155 	 * before looking at non-point-to-point interfaces.
156 	 * This is done to avoid returning non-point-to-point
157 	 * ipif instead of unnumbered point-to-point ipif.
158 	 */
159 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
160 	ill = ILL_START_WALK_V6(&ctx, ipst);
161 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
162 		GRAB_CONN_LOCK(q);
163 		mutex_enter(&ill->ill_lock);
164 		for (ipif = ill->ill_ipif; ipif != NULL;
165 		    ipif = ipif->ipif_next) {
166 			/* Allow the ipif to be down */
167 			if ((ipif->ipif_flags & IPIF_POINTOPOINT) &&
168 			    (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr,
169 			    if_addr)) &&
170 			    (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
171 			    dst))) {
172 				if (IPIF_CAN_LOOKUP(ipif)) {
173 					ipif_refhold_locked(ipif);
174 					mutex_exit(&ill->ill_lock);
175 					RELEASE_CONN_LOCK(q);
176 					rw_exit(&ipst->ips_ill_g_lock);
177 					return (ipif);
178 				} else if (IPIF_CAN_WAIT(ipif, q)) {
179 					ipsq = ill->ill_phyint->phyint_ipsq;
180 					mutex_enter(&ipsq->ipsq_lock);
181 					mutex_exit(&ill->ill_lock);
182 					rw_exit(&ipst->ips_ill_g_lock);
183 					ipsq_enq(ipsq, q, mp, func, NEW_OP,
184 					    ill);
185 					mutex_exit(&ipsq->ipsq_lock);
186 					RELEASE_CONN_LOCK(q);
187 					if (error != NULL)
188 						*error = EINPROGRESS;
189 					return (NULL);
190 				}
191 			}
192 		}
193 		mutex_exit(&ill->ill_lock);
194 		RELEASE_CONN_LOCK(q);
195 	}
196 	rw_exit(&ipst->ips_ill_g_lock);
197 	/* lookup the ipif based on interface address */
198 	ipif = ipif_lookup_addr_v6(if_addr, NULL, ALL_ZONES, q, mp, func,
199 	    error, ipst);
200 	ASSERT(ipif == NULL || ipif->ipif_isv6);
201 	return (ipif);
202 }
203 
204 /*
205  * Look for an ipif with the specified address. For point-point links
206  * we look for matches on either the destination address and the local
207  * address, but we ignore the check on the local address if IPIF_UNNUMBERED
208  * is set.
209  * Matches on a specific ill if match_ill is set.
210  */
211 /* ARGSUSED */
212 ipif_t *
213 ipif_lookup_addr_v6(const in6_addr_t *addr, ill_t *match_ill, zoneid_t zoneid,
214     queue_t *q, mblk_t *mp, ipsq_func_t func, int *error, ip_stack_t *ipst)
215 {
216 	ipif_t	*ipif;
217 	ill_t	*ill;
218 	boolean_t  ptp = B_FALSE;
219 	ipsq_t	*ipsq;
220 	ill_walk_context_t ctx;
221 
222 	if (error != NULL)
223 		*error = 0;
224 
225 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
226 	/*
227 	 * Repeat twice, first based on local addresses and
228 	 * next time for pointopoint.
229 	 */
230 repeat:
231 	ill = ILL_START_WALK_V6(&ctx, ipst);
232 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
233 		if (match_ill != NULL && ill != match_ill) {
234 			continue;
235 		}
236 		GRAB_CONN_LOCK(q);
237 		mutex_enter(&ill->ill_lock);
238 		for (ipif = ill->ill_ipif; ipif != NULL;
239 		    ipif = ipif->ipif_next) {
240 			if (zoneid != ALL_ZONES &&
241 			    ipif->ipif_zoneid != zoneid &&
242 			    ipif->ipif_zoneid != ALL_ZONES)
243 				continue;
244 			/* Allow the ipif to be down */
245 			if ((!ptp && (IN6_ARE_ADDR_EQUAL(
246 			    &ipif->ipif_v6lcl_addr, addr) &&
247 			    (ipif->ipif_flags & IPIF_UNNUMBERED) == 0)) ||
248 			    (ptp && (ipif->ipif_flags & IPIF_POINTOPOINT) &&
249 			    IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
250 			    addr))) {
251 				if (IPIF_CAN_LOOKUP(ipif)) {
252 					ipif_refhold_locked(ipif);
253 					mutex_exit(&ill->ill_lock);
254 					RELEASE_CONN_LOCK(q);
255 					rw_exit(&ipst->ips_ill_g_lock);
256 					return (ipif);
257 				} else if (IPIF_CAN_WAIT(ipif, q)) {
258 					ipsq = ill->ill_phyint->phyint_ipsq;
259 					mutex_enter(&ipsq->ipsq_lock);
260 					mutex_exit(&ill->ill_lock);
261 					rw_exit(&ipst->ips_ill_g_lock);
262 					ipsq_enq(ipsq, q, mp, func, NEW_OP,
263 					    ill);
264 					mutex_exit(&ipsq->ipsq_lock);
265 					RELEASE_CONN_LOCK(q);
266 					if (error != NULL)
267 						*error = EINPROGRESS;
268 					return (NULL);
269 				}
270 			}
271 		}
272 		mutex_exit(&ill->ill_lock);
273 		RELEASE_CONN_LOCK(q);
274 	}
275 
276 	/* If we already did the ptp case, then we are done */
277 	if (ptp) {
278 		rw_exit(&ipst->ips_ill_g_lock);
279 		if (error != NULL)
280 			*error = ENXIO;
281 		return (NULL);
282 	}
283 	ptp = B_TRUE;
284 	goto repeat;
285 }
286 
287 /*
288  * Look for an ipif with the specified address. For point-point links
289  * we look for matches on either the destination address and the local
290  * address, but we ignore the check on the local address if IPIF_UNNUMBERED
291  * is set.
292  * Matches on a specific ill if match_ill is set.
293  * Return the zoneid for the ipif. ALL_ZONES if none found.
294  */
295 zoneid_t
296 ipif_lookup_addr_zoneid_v6(const in6_addr_t *addr, ill_t *match_ill,
297     ip_stack_t *ipst)
298 {
299 	ipif_t	*ipif;
300 	ill_t	*ill;
301 	boolean_t  ptp = B_FALSE;
302 	ill_walk_context_t ctx;
303 	zoneid_t	zoneid;
304 
305 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
306 	/*
307 	 * Repeat twice, first based on local addresses and
308 	 * next time for pointopoint.
309 	 */
310 repeat:
311 	ill = ILL_START_WALK_V6(&ctx, ipst);
312 	for (; ill != NULL; ill = ill_next(&ctx, ill)) {
313 		if (match_ill != NULL && ill != match_ill) {
314 			continue;
315 		}
316 		mutex_enter(&ill->ill_lock);
317 		for (ipif = ill->ill_ipif; ipif != NULL;
318 		    ipif = ipif->ipif_next) {
319 			/* Allow the ipif to be down */
320 			if ((!ptp && (IN6_ARE_ADDR_EQUAL(
321 			    &ipif->ipif_v6lcl_addr, addr) &&
322 			    (ipif->ipif_flags & IPIF_UNNUMBERED) == 0)) ||
323 			    (ptp && (ipif->ipif_flags & IPIF_POINTOPOINT) &&
324 			    IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6pp_dst_addr,
325 			    addr)) &&
326 			    !(ipif->ipif_state_flags & IPIF_CONDEMNED)) {
327 				zoneid = ipif->ipif_zoneid;
328 				mutex_exit(&ill->ill_lock);
329 				rw_exit(&ipst->ips_ill_g_lock);
330 				/*
331 				 * If ipif_zoneid was ALL_ZONES then we have
332 				 * a trusted extensions shared IP address.
333 				 * In that case GLOBAL_ZONEID works to send.
334 				 */
335 				if (zoneid == ALL_ZONES)
336 					zoneid = GLOBAL_ZONEID;
337 				return (zoneid);
338 			}
339 		}
340 		mutex_exit(&ill->ill_lock);
341 	}
342 
343 	/* If we already did the ptp case, then we are done */
344 	if (ptp) {
345 		rw_exit(&ipst->ips_ill_g_lock);
346 		return (ALL_ZONES);
347 	}
348 	ptp = B_TRUE;
349 	goto repeat;
350 }
351 
352 /*
353  * Perform various checks to verify that an address would make sense as a local
354  * interface address.  This is currently only called when an attempt is made
355  * to set a local address.
356  *
357  * Does not allow a v4-mapped address, an address that equals the subnet
358  * anycast address, ... a multicast address, ...
359  */
360 boolean_t
361 ip_local_addr_ok_v6(const in6_addr_t *addr, const in6_addr_t *subnet_mask)
362 {
363 	in6_addr_t subnet;
364 
365 	if (IN6_IS_ADDR_UNSPECIFIED(addr))
366 		return (B_TRUE);	/* Allow all zeros */
367 
368 	/*
369 	 * Don't allow all zeroes or host part, but allow
370 	 * all ones netmask.
371 	 */
372 	V6_MASK_COPY(*addr, *subnet_mask, subnet);
373 	if (IN6_IS_ADDR_V4MAPPED(addr) ||
374 	    (IN6_ARE_ADDR_EQUAL(addr, &subnet) &&
375 	    !IN6_ARE_ADDR_EQUAL(subnet_mask, &ipv6_all_ones)) ||
376 	    (IN6_IS_ADDR_V4COMPAT(addr) && CLASSD(V4_PART_OF_V6((*addr)))) ||
377 	    IN6_IS_ADDR_MULTICAST(addr))
378 		return (B_FALSE);
379 
380 	return (B_TRUE);
381 }
382 
383 /*
384  * Perform various checks to verify that an address would make sense as a
385  * remote/subnet interface address.
386  */
387 boolean_t
388 ip_remote_addr_ok_v6(const in6_addr_t *addr, const in6_addr_t *subnet_mask)
389 {
390 	in6_addr_t subnet;
391 
392 	if (IN6_IS_ADDR_UNSPECIFIED(addr))
393 		return (B_TRUE);	/* Allow all zeros */
394 
395 	V6_MASK_COPY(*addr, *subnet_mask, subnet);
396 	if (IN6_IS_ADDR_V4MAPPED(addr) ||
397 	    (IN6_ARE_ADDR_EQUAL(addr, &subnet) &&
398 	    !IN6_ARE_ADDR_EQUAL(subnet_mask, &ipv6_all_ones)) ||
399 	    IN6_IS_ADDR_MULTICAST(addr) ||
400 	    (IN6_IS_ADDR_V4COMPAT(addr) && CLASSD(V4_PART_OF_V6((*addr)))))
401 		return (B_FALSE);
402 
403 	return (B_TRUE);
404 }
405 
406 /*
407  * ip_rt_add_v6 is called to add an IPv6 route to the forwarding table.
408  * ipif_arg is passed in to associate it with the correct interface
409  * (for link-local destinations and gateways).
410  */
411 /* ARGSUSED1 */
412 int
413 ip_rt_add_v6(const in6_addr_t *dst_addr, const in6_addr_t *mask,
414     const in6_addr_t *gw_addr, const in6_addr_t *src_addr, int flags,
415     ipif_t *ipif_arg, ire_t **ire_arg, queue_t *q, mblk_t *mp, ipsq_func_t func,
416     struct rtsa_s *sp, ip_stack_t *ipst)
417 {
418 	ire_t	*ire;
419 	ire_t	*gw_ire = NULL;
420 	ipif_t	*ipif;
421 	boolean_t ipif_refheld = B_FALSE;
422 	uint_t	type;
423 	int	match_flags = MATCH_IRE_TYPE;
424 	int	error;
425 	tsol_gc_t *gc = NULL;
426 	tsol_gcgrp_t *gcgrp = NULL;
427 	boolean_t gcgrp_xtraref = B_FALSE;
428 
429 	if (ire_arg != NULL)
430 		*ire_arg = NULL;
431 
432 	/*
433 	 * Prevent routes with a zero gateway from being created (since
434 	 * interfaces can currently be plumbed and brought up with no assigned
435 	 * address).
436 	 */
437 	if (IN6_IS_ADDR_UNSPECIFIED(gw_addr))
438 		return (ENETUNREACH);
439 
440 	/*
441 	 * If this is the case of RTF_HOST being set, then we set the netmask
442 	 * to all ones (regardless if one was supplied).
443 	 */
444 	if (flags & RTF_HOST)
445 		mask = &ipv6_all_ones;
446 
447 	/*
448 	 * Get the ipif, if any, corresponding to the gw_addr
449 	 */
450 	ipif = ipif_lookup_interface_v6(gw_addr, dst_addr, q, mp, func,
451 	    &error, ipst);
452 	if (ipif != NULL)
453 		ipif_refheld = B_TRUE;
454 	else if (error == EINPROGRESS) {
455 		ip1dbg(("ip_rt_add_v6: null and EINPROGRESS"));
456 		return (error);
457 	}
458 
459 	/*
460 	 * GateD will attempt to create routes with a loopback interface
461 	 * address as the gateway and with RTF_GATEWAY set.  We allow
462 	 * these routes to be added, but create them as interface routes
463 	 * since the gateway is an interface address.
464 	 */
465 	if ((ipif != NULL) && (ipif->ipif_ire_type == IRE_LOOPBACK)) {
466 		flags &= ~RTF_GATEWAY;
467 		if (IN6_ARE_ADDR_EQUAL(gw_addr, &ipv6_loopback) &&
468 		    IN6_ARE_ADDR_EQUAL(dst_addr, &ipv6_loopback) &&
469 		    IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones)) {
470 			ire = ire_ctable_lookup_v6(dst_addr, 0, IRE_LOOPBACK,
471 			    ipif, ALL_ZONES, NULL, match_flags, ipst);
472 			if (ire != NULL) {
473 				ire_refrele(ire);
474 				if (ipif_refheld)
475 					ipif_refrele(ipif);
476 				return (EEXIST);
477 			}
478 			ip1dbg(("ipif_up_done: 0x%p creating IRE 0x%x"
479 			    "for 0x%x\n", (void *)ipif,
480 			    ipif->ipif_ire_type,
481 			    ntohl(ipif->ipif_lcl_addr)));
482 			ire = ire_create_v6(
483 			    dst_addr,
484 			    mask,
485 			    &ipif->ipif_v6src_addr,
486 			    NULL,
487 			    &ipif->ipif_mtu,
488 			    NULL,
489 			    NULL,
490 			    NULL,
491 			    ipif->ipif_net_type,
492 			    ipif,
493 			    NULL,
494 			    0,
495 			    0,
496 			    flags,
497 			    &ire_uinfo_null,
498 			    NULL,
499 			    NULL,
500 			    ipst);
501 			if (ire == NULL) {
502 				if (ipif_refheld)
503 					ipif_refrele(ipif);
504 				return (ENOMEM);
505 			}
506 			error = ire_add(&ire, q, mp, func, B_FALSE);
507 			if (error == 0)
508 				goto save_ire;
509 			/*
510 			 * In the result of failure, ire_add() will have already
511 			 * deleted the ire in question, so there is no need to
512 			 * do that here.
513 			 */
514 			if (ipif_refheld)
515 				ipif_refrele(ipif);
516 			return (error);
517 		}
518 	}
519 
520 	/*
521 	 * Traditionally, interface routes are ones where RTF_GATEWAY isn't set
522 	 * and the gateway address provided is one of the system's interface
523 	 * addresses.  By using the routing socket interface and supplying an
524 	 * RTA_IFP sockaddr with an interface index, an alternate method of
525 	 * specifying an interface route to be created is available which uses
526 	 * the interface index that specifies the outgoing interface rather than
527 	 * the address of an outgoing interface (which may not be able to
528 	 * uniquely identify an interface).  When coupled with the RTF_GATEWAY
529 	 * flag, routes can be specified which not only specify the next-hop to
530 	 * be used when routing to a certain prefix, but also which outgoing
531 	 * interface should be used.
532 	 *
533 	 * Previously, interfaces would have unique addresses assigned to them
534 	 * and so the address assigned to a particular interface could be used
535 	 * to identify a particular interface.  One exception to this was the
536 	 * case of an unnumbered interface (where IPIF_UNNUMBERED was set).
537 	 *
538 	 * With the advent of IPv6 and its link-local addresses, this
539 	 * restriction was relaxed and interfaces could share addresses between
540 	 * themselves.  In fact, typically all of the link-local interfaces on
541 	 * an IPv6 node or router will have the same link-local address.  In
542 	 * order to differentiate between these interfaces, the use of an
543 	 * interface index is necessary and this index can be carried inside a
544 	 * RTA_IFP sockaddr (which is actually a sockaddr_dl).  One restriction
545 	 * of using the interface index, however, is that all of the ipif's that
546 	 * are part of an ill have the same index and so the RTA_IFP sockaddr
547 	 * cannot be used to differentiate between ipif's (or logical
548 	 * interfaces) that belong to the same ill (physical interface).
549 	 *
550 	 * For example, in the following case involving IPv4 interfaces and
551 	 * logical interfaces
552 	 *
553 	 *	192.0.2.32	255.255.255.224	192.0.2.33	U	if0
554 	 *	192.0.2.32	255.255.255.224	192.0.2.34	U	if0:1
555 	 *	192.0.2.32	255.255.255.224	192.0.2.35	U	if0:2
556 	 *
557 	 * the ipif's corresponding to each of these interface routes can be
558 	 * uniquely identified by the "gateway" (actually interface address).
559 	 *
560 	 * In this case involving multiple IPv6 default routes to a particular
561 	 * link-local gateway, the use of RTA_IFP is necessary to specify which
562 	 * default route is of interest:
563 	 *
564 	 *	default		fe80::123:4567:89ab:cdef	U	if0
565 	 *	default		fe80::123:4567:89ab:cdef	U	if1
566 	 */
567 
568 	/* RTF_GATEWAY not set */
569 	if (!(flags & RTF_GATEWAY)) {
570 		queue_t	*stq;
571 
572 		if (sp != NULL) {
573 			ip2dbg(("ip_rt_add_v6: gateway security attributes "
574 			    "cannot be set with interface route\n"));
575 			if (ipif_refheld)
576 				ipif_refrele(ipif);
577 			return (EINVAL);
578 		}
579 
580 		/*
581 		 * As the interface index specified with the RTA_IFP sockaddr is
582 		 * the same for all ipif's off of an ill, the matching logic
583 		 * below uses MATCH_IRE_ILL if such an index was specified.
584 		 * This means that routes sharing the same prefix when added
585 		 * using a RTA_IFP sockaddr must have distinct interface
586 		 * indices (namely, they must be on distinct ill's).
587 		 *
588 		 * On the other hand, since the gateway address will usually be
589 		 * different for each ipif on the system, the matching logic
590 		 * uses MATCH_IRE_IPIF in the case of a traditional interface
591 		 * route.  This means that interface routes for the same prefix
592 		 * can be created if they belong to distinct ipif's and if a
593 		 * RTA_IFP sockaddr is not present.
594 		 */
595 		if (ipif_arg != NULL) {
596 			if (ipif_refheld) {
597 				ipif_refrele(ipif);
598 				ipif_refheld = B_FALSE;
599 			}
600 			ipif = ipif_arg;
601 			match_flags |= MATCH_IRE_ILL;
602 		} else {
603 			/*
604 			 * Check the ipif corresponding to the gw_addr
605 			 */
606 			if (ipif == NULL)
607 				return (ENETUNREACH);
608 			match_flags |= MATCH_IRE_IPIF;
609 		}
610 
611 		ASSERT(ipif != NULL);
612 		/*
613 		 * We check for an existing entry at this point.
614 		 */
615 		match_flags |= MATCH_IRE_MASK;
616 		ire = ire_ftable_lookup_v6(dst_addr, mask, 0, IRE_INTERFACE,
617 		    ipif, NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
618 		if (ire != NULL) {
619 			ire_refrele(ire);
620 			if (ipif_refheld)
621 				ipif_refrele(ipif);
622 			return (EEXIST);
623 		}
624 
625 		stq = (ipif->ipif_net_type == IRE_IF_RESOLVER)
626 		    ? ipif->ipif_rq : ipif->ipif_wq;
627 
628 		/*
629 		 * Create a copy of the IRE_LOOPBACK, IRE_IF_NORESOLVER or
630 		 * IRE_IF_RESOLVER with the modified address and netmask.
631 		 */
632 		ire = ire_create_v6(
633 		    dst_addr,
634 		    mask,
635 		    &ipif->ipif_v6src_addr,
636 		    NULL,
637 		    &ipif->ipif_mtu,
638 		    NULL,
639 		    NULL,
640 		    stq,
641 		    ipif->ipif_net_type,
642 		    ipif,
643 		    NULL,
644 		    0,
645 		    0,
646 		    flags,
647 		    &ire_uinfo_null,
648 		    NULL,
649 		    NULL,
650 		    ipst);
651 		if (ire == NULL) {
652 			if (ipif_refheld)
653 				ipif_refrele(ipif);
654 			return (ENOMEM);
655 		}
656 
657 		/*
658 		 * Some software (for example, GateD and Sun Cluster) attempts
659 		 * to create (what amount to) IRE_PREFIX routes with the
660 		 * loopback address as the gateway.  This is primarily done to
661 		 * set up prefixes with the RTF_REJECT flag set (for example,
662 		 * when generating aggregate routes). We also OR in the
663 		 * RTF_BLACKHOLE flag as these interface routes, by
664 		 * definition, can only be that.
665 		 *
666 		 * If the IRE type (as defined by ipif->ipif_net_type) is
667 		 * IRE_LOOPBACK, then we map the request into a
668 		 * IRE_IF_NORESOLVER.
669 		 *
670 		 * Needless to say, the real IRE_LOOPBACK is NOT created by this
671 		 * routine, but rather using ire_create_v6() directly.
672 		 */
673 		if (ipif->ipif_net_type == IRE_LOOPBACK) {
674 			ire->ire_type = IRE_IF_NORESOLVER;
675 			ire->ire_flags |= RTF_BLACKHOLE;
676 		}
677 		error = ire_add(&ire, q, mp, func, B_FALSE);
678 		if (error == 0)
679 			goto save_ire;
680 		/*
681 		 * In the result of failure, ire_add() will have already
682 		 * deleted the ire in question, so there is no need to
683 		 * do that here.
684 		 */
685 		if (ipif_refheld)
686 			ipif_refrele(ipif);
687 		return (error);
688 	}
689 	if (ipif_refheld) {
690 		ipif_refrele(ipif);
691 		ipif_refheld = B_FALSE;
692 	}
693 
694 	/*
695 	 * Get an interface IRE for the specified gateway.
696 	 * If we don't have an IRE_IF_NORESOLVER or IRE_IF_RESOLVER for the
697 	 * gateway, it is currently unreachable and we fail the request
698 	 * accordingly.
699 	 */
700 	ipif = ipif_arg;
701 	if (ipif_arg != NULL)
702 		match_flags |= MATCH_IRE_ILL;
703 	gw_ire = ire_ftable_lookup_v6(gw_addr, 0, 0, IRE_INTERFACE, ipif_arg,
704 	    NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
705 	if (gw_ire == NULL)
706 		return (ENETUNREACH);
707 
708 	/*
709 	 * We create one of three types of IREs as a result of this request
710 	 * based on the netmask.  A netmask of all ones (which is automatically
711 	 * assumed when RTF_HOST is set) results in an IRE_HOST being created.
712 	 * An all zeroes netmask implies a default route so an IRE_DEFAULT is
713 	 * created.  Otherwise, an IRE_PREFIX route is created for the
714 	 * destination prefix.
715 	 */
716 	if (IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones))
717 		type = IRE_HOST;
718 	else if (IN6_IS_ADDR_UNSPECIFIED(mask))
719 		type = IRE_DEFAULT;
720 	else
721 		type = IRE_PREFIX;
722 
723 	/* check for a duplicate entry */
724 	ire = ire_ftable_lookup_v6(dst_addr, mask, gw_addr, type, ipif_arg,
725 	    NULL, ALL_ZONES, 0, NULL,
726 	    match_flags | MATCH_IRE_MASK | MATCH_IRE_GW, ipst);
727 	if (ire != NULL) {
728 		ire_refrele(gw_ire);
729 		ire_refrele(ire);
730 		return (EEXIST);
731 	}
732 
733 	/* Security attribute exists */
734 	if (sp != NULL) {
735 		tsol_gcgrp_addr_t ga;
736 
737 		/* find or create the gateway credentials group */
738 		ga.ga_af = AF_INET6;
739 		ga.ga_addr = *gw_addr;
740 
741 		/* we hold reference to it upon success */
742 		gcgrp = gcgrp_lookup(&ga, B_TRUE);
743 		if (gcgrp == NULL) {
744 			ire_refrele(gw_ire);
745 			return (ENOMEM);
746 		}
747 
748 		/*
749 		 * Create and add the security attribute to the group; a
750 		 * reference to the group is made upon allocating a new
751 		 * entry successfully.  If it finds an already-existing
752 		 * entry for the security attribute in the group, it simply
753 		 * returns it and no new reference is made to the group.
754 		 */
755 		gc = gc_create(sp, gcgrp, &gcgrp_xtraref);
756 		if (gc == NULL) {
757 			/* release reference held by gcgrp_lookup */
758 			GCGRP_REFRELE(gcgrp);
759 			ire_refrele(gw_ire);
760 			return (ENOMEM);
761 		}
762 	}
763 
764 	/* Create the IRE. */
765 	ire = ire_create_v6(
766 	    dst_addr,				/* dest address */
767 	    mask,				/* mask */
768 	    /* src address assigned by the caller? */
769 	    (((flags & RTF_SETSRC) && !IN6_IS_ADDR_UNSPECIFIED(src_addr)) ?
770 	    src_addr : NULL),
771 	    gw_addr,				/* gateway address */
772 	    &gw_ire->ire_max_frag,
773 	    NULL,				/* no src nce */
774 	    NULL,				/* no recv-from queue */
775 	    NULL,				/* no send-to queue */
776 	    (ushort_t)type,			/* IRE type */
777 	    ipif_arg,
778 	    NULL,
779 	    0,
780 	    0,
781 	    flags,
782 	    &gw_ire->ire_uinfo,			/* Inherit ULP info from gw */
783 	    gc,					/* security attribute */
784 	    NULL,
785 	    ipst);
786 
787 	/*
788 	 * The ire holds a reference to the 'gc' and the 'gc' holds a
789 	 * reference to the 'gcgrp'. We can now release the extra reference
790 	 * the 'gcgrp' acquired in the gcgrp_lookup, if it was not used.
791 	 */
792 	if (gcgrp_xtraref)
793 		GCGRP_REFRELE(gcgrp);
794 	if (ire == NULL) {
795 		if (gc != NULL)
796 			GC_REFRELE(gc);
797 		ire_refrele(gw_ire);
798 		return (ENOMEM);
799 	}
800 
801 	/*
802 	 * POLICY: should we allow an RTF_HOST with address INADDR_ANY?
803 	 * SUN/OS socket stuff does but do we really want to allow ::0 ?
804 	 */
805 
806 	/* Add the new IRE. */
807 	error = ire_add(&ire, q, mp, func, B_FALSE);
808 	/*
809 	 * In the result of failure, ire_add() will have already
810 	 * deleted the ire in question, so there is no need to
811 	 * do that here.
812 	 */
813 	if (error != 0) {
814 		ire_refrele(gw_ire);
815 		return (error);
816 	}
817 
818 	if (flags & RTF_MULTIRT) {
819 		/*
820 		 * Invoke the CGTP (multirouting) filtering module
821 		 * to add the dst address in the filtering database.
822 		 * Replicated inbound packets coming from that address
823 		 * will be filtered to discard the duplicates.
824 		 * It is not necessary to call the CGTP filter hook
825 		 * when the dst address is a multicast, because an
826 		 * IP source address cannot be a multicast.
827 		 */
828 		if (ipst->ips_ip_cgtp_filter_ops != NULL &&
829 		    !IN6_IS_ADDR_MULTICAST(&(ire->ire_addr_v6))) {
830 			int res;
831 
832 			res = ipst->ips_ip_cgtp_filter_ops->cfo_add_dest_v6(
833 			    ipst->ips_netstack->netstack_stackid,
834 			    &ire->ire_addr_v6,
835 			    &ire->ire_gateway_addr_v6,
836 			    &ire->ire_src_addr_v6,
837 			    &gw_ire->ire_src_addr_v6);
838 			if (res != 0) {
839 				ire_refrele(gw_ire);
840 				ire_delete(ire);
841 				return (res);
842 			}
843 		}
844 	}
845 
846 	/*
847 	 * Now that the prefix IRE entry has been created, delete any
848 	 * existing gateway IRE cache entries as well as any IRE caches
849 	 * using the gateway, and force them to be created through
850 	 * ip_newroute_v6.
851 	 */
852 	if (gc != NULL) {
853 		ASSERT(gcgrp != NULL);
854 		ire_clookup_delete_cache_gw_v6(gw_addr, ALL_ZONES, ipst);
855 	}
856 
857 save_ire:
858 	if (gw_ire != NULL) {
859 		ire_refrele(gw_ire);
860 	}
861 	if (ipif != NULL) {
862 		mblk_t	*save_mp;
863 
864 		/*
865 		 * Save enough information so that we can recreate the IRE if
866 		 * the interface goes down and then up.  The metrics associated
867 		 * with the route will be saved as well when rts_setmetrics() is
868 		 * called after the IRE has been created.  In the case where
869 		 * memory cannot be allocated, none of this information will be
870 		 * saved.
871 		 */
872 		save_mp = allocb(sizeof (ifrt_t), BPRI_MED);
873 		if (save_mp != NULL) {
874 			ifrt_t	*ifrt;
875 
876 			save_mp->b_wptr += sizeof (ifrt_t);
877 			ifrt = (ifrt_t *)save_mp->b_rptr;
878 			bzero(ifrt, sizeof (ifrt_t));
879 			ifrt->ifrt_type = ire->ire_type;
880 			ifrt->ifrt_v6addr = ire->ire_addr_v6;
881 			mutex_enter(&ire->ire_lock);
882 			ifrt->ifrt_v6gateway_addr = ire->ire_gateway_addr_v6;
883 			ifrt->ifrt_v6src_addr = ire->ire_src_addr_v6;
884 			mutex_exit(&ire->ire_lock);
885 			ifrt->ifrt_v6mask = ire->ire_mask_v6;
886 			ifrt->ifrt_flags = ire->ire_flags;
887 			ifrt->ifrt_max_frag = ire->ire_max_frag;
888 			mutex_enter(&ipif->ipif_saved_ire_lock);
889 			save_mp->b_cont = ipif->ipif_saved_ire_mp;
890 			ipif->ipif_saved_ire_mp = save_mp;
891 			ipif->ipif_saved_ire_cnt++;
892 			mutex_exit(&ipif->ipif_saved_ire_lock);
893 		}
894 	}
895 	if (ire_arg != NULL) {
896 		/*
897 		 * Store the ire that was successfully added into where ire_arg
898 		 * points to so that callers don't have to look it up
899 		 * themselves (but they are responsible for ire_refrele()ing
900 		 * the ire when they are finished with it).
901 		 */
902 		*ire_arg = ire;
903 	} else {
904 		ire_refrele(ire);		/* Held in ire_add */
905 	}
906 	if (ipif_refheld)
907 		ipif_refrele(ipif);
908 	return (0);
909 }
910 
911 /*
912  * ip_rt_delete_v6 is called to delete an IPv6 route.
913  * ipif_arg is passed in to associate it with the correct interface
914  * (for link-local destinations and gateways).
915  */
916 /* ARGSUSED4 */
917 int
918 ip_rt_delete_v6(const in6_addr_t *dst_addr, const in6_addr_t *mask,
919     const in6_addr_t *gw_addr, uint_t rtm_addrs, int flags, ipif_t *ipif_arg,
920     queue_t *q, mblk_t *mp, ipsq_func_t func, ip_stack_t *ipst)
921 {
922 	ire_t	*ire = NULL;
923 	ipif_t	*ipif;
924 	uint_t	type;
925 	uint_t	match_flags = MATCH_IRE_TYPE;
926 	int	err = 0;
927 	boolean_t	ipif_refheld = B_FALSE;
928 
929 	/*
930 	 * If this is the case of RTF_HOST being set, then we set the netmask
931 	 * to all ones.  Otherwise, we use the netmask if one was supplied.
932 	 */
933 	if (flags & RTF_HOST) {
934 		mask = &ipv6_all_ones;
935 		match_flags |= MATCH_IRE_MASK;
936 	} else if (rtm_addrs & RTA_NETMASK) {
937 		match_flags |= MATCH_IRE_MASK;
938 	}
939 
940 	/*
941 	 * Note that RTF_GATEWAY is never set on a delete, therefore
942 	 * we check if the gateway address is one of our interfaces first,
943 	 * and fall back on RTF_GATEWAY routes.
944 	 *
945 	 * This makes it possible to delete an original
946 	 * IRE_IF_NORESOLVER/IRE_IF_RESOLVER - consistent with SunOS 4.1.
947 	 *
948 	 * As the interface index specified with the RTA_IFP sockaddr is the
949 	 * same for all ipif's off of an ill, the matching logic below uses
950 	 * MATCH_IRE_ILL if such an index was specified.  This means a route
951 	 * sharing the same prefix and interface index as the the route
952 	 * intended to be deleted might be deleted instead if a RTA_IFP sockaddr
953 	 * is specified in the request.
954 	 *
955 	 * On the other hand, since the gateway address will usually be
956 	 * different for each ipif on the system, the matching logic
957 	 * uses MATCH_IRE_IPIF in the case of a traditional interface
958 	 * route.  This means that interface routes for the same prefix can be
959 	 * uniquely identified if they belong to distinct ipif's and if a
960 	 * RTA_IFP sockaddr is not present.
961 	 *
962 	 * For more detail on specifying routes by gateway address and by
963 	 * interface index, see the comments in ip_rt_add_v6().
964 	 */
965 	ipif = ipif_lookup_interface_v6(gw_addr, dst_addr, q, mp, func, &err,
966 	    ipst);
967 	if (ipif != NULL) {
968 		ipif_refheld = B_TRUE;
969 		if (ipif_arg != NULL) {
970 			ipif_refrele(ipif);
971 			ipif_refheld = B_FALSE;
972 			ipif = ipif_arg;
973 			match_flags |= MATCH_IRE_ILL;
974 		} else {
975 			match_flags |= MATCH_IRE_IPIF;
976 		}
977 
978 		if (ipif->ipif_ire_type == IRE_LOOPBACK)
979 			ire = ire_ctable_lookup_v6(dst_addr, 0, IRE_LOOPBACK,
980 			    ipif, ALL_ZONES, NULL, match_flags, ipst);
981 		if (ire == NULL)
982 			ire = ire_ftable_lookup_v6(dst_addr, mask, 0,
983 			    IRE_INTERFACE, ipif, NULL, ALL_ZONES, 0, NULL,
984 			    match_flags, ipst);
985 	} else if (err == EINPROGRESS) {
986 		return (err);
987 	} else {
988 		err = 0;
989 	}
990 	if (ire == NULL) {
991 		/*
992 		 * At this point, the gateway address is not one of our own
993 		 * addresses or a matching interface route was not found.  We
994 		 * set the IRE type to lookup based on whether
995 		 * this is a host route, a default route or just a prefix.
996 		 *
997 		 * If an ipif_arg was passed in, then the lookup is based on an
998 		 * interface index so MATCH_IRE_ILL is added to match_flags.
999 		 * In any case, MATCH_IRE_IPIF is cleared and MATCH_IRE_GW is
1000 		 * set as the route being looked up is not a traditional
1001 		 * interface route.
1002 		 */
1003 		match_flags &= ~MATCH_IRE_IPIF;
1004 		match_flags |= MATCH_IRE_GW;
1005 		if (ipif_arg != NULL)
1006 			match_flags |= MATCH_IRE_ILL;
1007 		if (IN6_ARE_ADDR_EQUAL(mask, &ipv6_all_ones))
1008 			type = IRE_HOST;
1009 		else if (IN6_IS_ADDR_UNSPECIFIED(mask))
1010 			type = IRE_DEFAULT;
1011 		else
1012 			type = IRE_PREFIX;
1013 		ire = ire_ftable_lookup_v6(dst_addr, mask, gw_addr, type,
1014 		    ipif_arg, NULL, ALL_ZONES, 0, NULL, match_flags, ipst);
1015 	}
1016 
1017 	if (ipif_refheld) {
1018 		ipif_refrele(ipif);
1019 		ipif_refheld = B_FALSE;
1020 	}
1021 	if (ire == NULL)
1022 		return (ESRCH);
1023 
1024 	if (ire->ire_flags & RTF_MULTIRT) {
1025 		/*
1026 		 * Invoke the CGTP (multirouting) filtering module
1027 		 * to remove the dst address from the filtering database.
1028 		 * Packets coming from that address will no longer be
1029 		 * filtered to remove duplicates.
1030 		 */
1031 		if (ipst->ips_ip_cgtp_filter_ops != NULL) {
1032 			err = ipst->ips_ip_cgtp_filter_ops->cfo_del_dest_v6(
1033 			    ipst->ips_netstack->netstack_stackid,
1034 			    &ire->ire_addr_v6, &ire->ire_gateway_addr_v6);
1035 		}
1036 	}
1037 
1038 	ipif = ire->ire_ipif;
1039 	if (ipif != NULL) {
1040 		mblk_t		**mpp;
1041 		mblk_t		*mp;
1042 		ifrt_t		*ifrt;
1043 		in6_addr_t	gw_addr_v6;
1044 
1045 		/* Remove from ipif_saved_ire_mp list if it is there */
1046 		mutex_enter(&ire->ire_lock);
1047 		gw_addr_v6 = ire->ire_gateway_addr_v6;
1048 		mutex_exit(&ire->ire_lock);
1049 		mutex_enter(&ipif->ipif_saved_ire_lock);
1050 		for (mpp = &ipif->ipif_saved_ire_mp; *mpp != NULL;
1051 		    mpp = &(*mpp)->b_cont) {
1052 			/*
1053 			 * On a given ipif, the triple of address, gateway and
1054 			 * mask is unique for each saved IRE (in the case of
1055 			 * ordinary interface routes, the gateway address is
1056 			 * all-zeroes).
1057 			 */
1058 			mp = *mpp;
1059 			ifrt = (ifrt_t *)mp->b_rptr;
1060 			if (IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6addr,
1061 			    &ire->ire_addr_v6) &&
1062 			    IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6gateway_addr,
1063 			    &gw_addr_v6) &&
1064 			    IN6_ARE_ADDR_EQUAL(&ifrt->ifrt_v6mask,
1065 			    &ire->ire_mask_v6)) {
1066 				*mpp = mp->b_cont;
1067 				ipif->ipif_saved_ire_cnt--;
1068 				freeb(mp);
1069 				break;
1070 			}
1071 		}
1072 		mutex_exit(&ipif->ipif_saved_ire_lock);
1073 	}
1074 	ire_delete(ire);
1075 	ire_refrele(ire);
1076 	return (err);
1077 }
1078 
1079 /*
1080  * Derive a token from the link layer address.
1081  */
1082 boolean_t
1083 ill_setdefaulttoken(ill_t *ill)
1084 {
1085 	int 		i;
1086 	in6_addr_t	v6addr, v6mask;
1087 
1088 	if (!MEDIA_V6INTFID(ill->ill_media, ill->ill_phys_addr_length,
1089 	    ill->ill_phys_addr, &v6addr))
1090 		return (B_FALSE);
1091 
1092 	(void) ip_plen_to_mask_v6(IPV6_TOKEN_LEN, &v6mask);
1093 
1094 	for (i = 0; i < 4; i++)
1095 		v6mask.s6_addr32[i] = v6mask.s6_addr32[i] ^
1096 		    (uint32_t)0xffffffff;
1097 
1098 	V6_MASK_COPY(v6addr, v6mask, ill->ill_token);
1099 	ill->ill_token_length = IPV6_TOKEN_LEN;
1100 	return (B_TRUE);
1101 }
1102 
1103 /*
1104  * Create a link-local address from a token.
1105  */
1106 static void
1107 ipif_get_linklocal(in6_addr_t *dest, const in6_addr_t *token)
1108 {
1109 	int i;
1110 
1111 	for (i = 0; i < 4; i++) {
1112 		dest->s6_addr32[i] =
1113 		    token->s6_addr32[i] | ipv6_ll_template.s6_addr32[i];
1114 	}
1115 }
1116 
1117 /*
1118  * Set a nice default address for either automatic tunnels tsrc/96 or
1119  * 6to4 tunnels 2002:<tsrc>::1/64
1120  */
1121 static void
1122 ipif_set_tun_auto_addr(ipif_t *ipif, struct iftun_req *ta)
1123 {
1124 	sin6_t	sin6;
1125 	sin_t	*sin;
1126 	ill_t 	*ill = ipif->ipif_ill;
1127 	tun_t *tp = (tun_t *)ill->ill_wq->q_next->q_ptr;
1128 
1129 	if (ta->ifta_saddr.ss_family != AF_INET ||
1130 	    (ipif->ipif_flags & IPIF_UP) || !ipif->ipif_isv6 ||
1131 	    (ta->ifta_flags & IFTUN_SRC) == 0)
1132 		return;
1133 
1134 	/*
1135 	 * Check the tunnel type by examining q_next->q_ptr
1136 	 */
1137 	if (tp->tun_flags & TUN_AUTOMATIC) {
1138 		/* this is an automatic tunnel */
1139 		(void) ip_plen_to_mask_v6(IPV6_ABITS - IP_ABITS,
1140 		    &ipif->ipif_v6net_mask);
1141 		bzero(&sin6, sizeof (sin6_t));
1142 		sin = (sin_t *)&ta->ifta_saddr;
1143 		V4_PART_OF_V6(sin6.sin6_addr) = sin->sin_addr.s_addr;
1144 		sin6.sin6_family = AF_INET6;
1145 		(void) ip_sioctl_addr(ipif, (sin_t *)&sin6,
1146 		    NULL, NULL, NULL, NULL);
1147 	} else if (tp->tun_flags & TUN_6TO4) {
1148 		/* this is a 6to4 tunnel */
1149 		(void) ip_plen_to_mask_v6(IPV6_PREFIX_LEN,
1150 		    &ipif->ipif_v6net_mask);
1151 		sin = (sin_t *)&ta->ifta_saddr;
1152 		/* create a 6to4 address from the IPv4 tsrc */
1153 		IN6_V4ADDR_TO_6TO4(&sin->sin_addr, &sin6.sin6_addr);
1154 		sin6.sin6_family = AF_INET6;
1155 		(void) ip_sioctl_addr(ipif, (sin_t *)&sin6,
1156 		    NULL, NULL, NULL, NULL);
1157 	} else {
1158 		ip1dbg(("ipif_set_tun_auto_addr: Unknown tunnel type"));
1159 		return;
1160 	}
1161 }
1162 
1163 /*
1164  * Set link local for ipif_id 0 of a configured tunnel based on the
1165  * tsrc or tdst parameter
1166  * For tunnels over IPv4 use the IPv4 address prepended with 32 zeros as
1167  * the token.
1168  * For tunnels over IPv6 use the low-order 64 bits of the "inner" IPv6 address
1169  * as the token for the "outer" link.
1170  */
1171 void
1172 ipif_set_tun_llink(ill_t *ill, struct iftun_req *ta)
1173 {
1174 	ipif_t		*ipif;
1175 	sin_t		*sin;
1176 	in6_addr_t	*s6addr;
1177 
1178 	ASSERT(IAM_WRITER_ILL(ill));
1179 
1180 	/* The first ipif must be id zero. */
1181 	ipif = ill->ill_ipif;
1182 	ASSERT(ipif->ipif_id == 0);
1183 
1184 	/* no link local for automatic tunnels */
1185 	if (!(ipif->ipif_flags & IPIF_POINTOPOINT)) {
1186 		ipif_set_tun_auto_addr(ipif, ta);
1187 		return;
1188 	}
1189 
1190 	if ((ta->ifta_flags & IFTUN_DST) &&
1191 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6pp_dst_addr)) {
1192 		sin6_t	sin6;
1193 
1194 		ASSERT(!(ipif->ipif_flags & IPIF_UP));
1195 		bzero(&sin6, sizeof (sin6_t));
1196 		if ((ta->ifta_saddr.ss_family == AF_INET)) {
1197 			sin = (sin_t *)&ta->ifta_daddr;
1198 			V4_PART_OF_V6(sin6.sin6_addr) =
1199 			    sin->sin_addr.s_addr;
1200 		} else {
1201 			s6addr =
1202 			    &((sin6_t *)&ta->ifta_daddr)->sin6_addr;
1203 			sin6.sin6_addr.s6_addr32[3] = s6addr->s6_addr32[3];
1204 			sin6.sin6_addr.s6_addr32[2] = s6addr->s6_addr32[2];
1205 		}
1206 		ipif_get_linklocal(&ipif->ipif_v6pp_dst_addr,
1207 		    &sin6.sin6_addr);
1208 		ipif->ipif_v6subnet = ipif->ipif_v6pp_dst_addr;
1209 	}
1210 	if ((ta->ifta_flags & IFTUN_SRC)) {
1211 		ASSERT(!(ipif->ipif_flags & IPIF_UP));
1212 
1213 		/* Set the token if it isn't already set */
1214 		if (IN6_IS_ADDR_UNSPECIFIED(&ill->ill_token)) {
1215 			if ((ta->ifta_saddr.ss_family == AF_INET)) {
1216 				sin = (sin_t *)&ta->ifta_saddr;
1217 				V4_PART_OF_V6(ill->ill_token) =
1218 				    sin->sin_addr.s_addr;
1219 			} else {
1220 				s6addr =
1221 				    &((sin6_t *)&ta->ifta_saddr)->sin6_addr;
1222 				ill->ill_token.s6_addr32[3] =
1223 				    s6addr->s6_addr32[3];
1224 				ill->ill_token.s6_addr32[2] =
1225 				    s6addr->s6_addr32[2];
1226 			}
1227 			ill->ill_token_length = IPV6_TOKEN_LEN;
1228 		}
1229 		/*
1230 		 * Attempt to set the link local address if it isn't set.
1231 		 */
1232 		if (IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr))
1233 			(void) ipif_setlinklocal(ipif);
1234 	}
1235 }
1236 
1237 /*
1238  * Is it not possible to set the link local address?
1239  * The address can be set if the token is set, and the token
1240  * isn't too long.
1241  * Return B_TRUE if the address can't be set, or B_FALSE if it can.
1242  */
1243 boolean_t
1244 ipif_cant_setlinklocal(ipif_t *ipif)
1245 {
1246 	ill_t *ill = ipif->ipif_ill;
1247 
1248 	if (IN6_IS_ADDR_UNSPECIFIED(&ill->ill_token) ||
1249 	    ill->ill_token_length > IPV6_ABITS - IPV6_LL_PREFIXLEN)
1250 		return (B_TRUE);
1251 
1252 	return (B_FALSE);
1253 }
1254 
1255 /*
1256  * Generate a link-local address from the token.
1257  * Return zero if the address was set, or non-zero if it couldn't be set.
1258  */
1259 int
1260 ipif_setlinklocal(ipif_t *ipif)
1261 {
1262 	ill_t		*ill = ipif->ipif_ill;
1263 	in6_addr_t	ov6addr;
1264 
1265 	ASSERT(IAM_WRITER_ILL(ill));
1266 
1267 	if (ipif_cant_setlinklocal(ipif))
1268 		return (-1);
1269 
1270 	ov6addr = ipif->ipif_v6lcl_addr;
1271 	ipif_get_linklocal(&ipif->ipif_v6lcl_addr, &ill->ill_token);
1272 	sctp_update_ipif_addr(ipif, ov6addr);
1273 	(void) ip_plen_to_mask_v6(IPV6_LL_PREFIXLEN, &ipif->ipif_v6net_mask);
1274 	V6_MASK_COPY(ipif->ipif_v6lcl_addr, ipif->ipif_v6net_mask,
1275 	    ipif->ipif_v6subnet);
1276 
1277 	if (ipif->ipif_flags & IPIF_NOLOCAL) {
1278 		ipif->ipif_v6src_addr = ipv6_all_zeros;
1279 	} else {
1280 		ipif->ipif_v6src_addr = ipif->ipif_v6lcl_addr;
1281 	}
1282 	return (0);
1283 }
1284 
1285 /*
1286  * This function sets up the multicast mappings in NDP.
1287  * Unlike ARP, there are no mapping_mps here. We delete the
1288  * mapping nces and add a new one.
1289  *
1290  * Returns non-zero on error and 0 on success.
1291  */
1292 int
1293 ipif_ndp_setup_multicast(ipif_t *ipif, nce_t **ret_nce)
1294 {
1295 	ill_t		*ill = ipif->ipif_ill;
1296 	in6_addr_t	v6_mcast_addr = {(uint32_t)V6_MCAST, 0, 0, 0};
1297 	in6_addr_t	v6_mcast_mask = {(uint32_t)V6_MCAST, 0, 0, 0};
1298 	in6_addr_t	v6_extract_mask;
1299 	uchar_t		*phys_addr, *bphys_addr, *alloc_phys;
1300 	nce_t		*mnce = NULL;
1301 	int		err = 0;
1302 	phyint_t	*phyi = ill->ill_phyint;
1303 	uint32_t	hw_extract_start;
1304 	dl_unitdata_req_t *dlur;
1305 	ip_stack_t	*ipst = ill->ill_ipst;
1306 
1307 	if (ret_nce != NULL)
1308 		*ret_nce = NULL;
1309 	/*
1310 	 * Delete the mapping nce. Normally these should not exist
1311 	 * as a previous ipif_down -> ipif_ndp_down should have deleted
1312 	 * all the nces. But they can exist if ip_rput_dlpi_writer
1313 	 * calls this when PHYI_MULTI_BCAST is set.
1314 	 */
1315 	mnce = ndp_lookup_v6(ill, &v6_mcast_addr, B_FALSE);
1316 	if (mnce != NULL) {
1317 		ndp_delete(mnce);
1318 		NCE_REFRELE(mnce);
1319 		mnce = NULL;
1320 	}
1321 
1322 	/*
1323 	 * Get media specific v6 mapping information. Note that
1324 	 * nd_lla_len can be 0 for tunnels.
1325 	 */
1326 	alloc_phys = kmem_alloc(ill->ill_nd_lla_len, KM_NOSLEEP);
1327 	if ((alloc_phys == NULL) && (ill->ill_nd_lla_len != 0))
1328 		return (ENOMEM);
1329 	/*
1330 	 * Determine the broadcast address.
1331 	 */
1332 	dlur = (dl_unitdata_req_t *)ill->ill_bcast_mp->b_rptr;
1333 	if (ill->ill_sap_length < 0)
1334 		bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset;
1335 	else
1336 		bphys_addr = (uchar_t *)dlur +
1337 		    dlur->dl_dest_addr_offset + ill->ill_sap_length;
1338 
1339 	/*
1340 	 * Check PHYI_MULTI_BCAST and possible length of physical
1341 	 * address to determine if we use the mapping or the
1342 	 * broadcast address.
1343 	 */
1344 	if ((phyi->phyint_flags & PHYI_MULTI_BCAST) ||
1345 	    (!MEDIA_V6MINFO(ill->ill_media, ill->ill_nd_lla_len,
1346 	    bphys_addr, alloc_phys, &hw_extract_start,
1347 	    &v6_extract_mask))) {
1348 		if (ill->ill_phys_addr_length > IP_MAX_HW_LEN) {
1349 			kmem_free(alloc_phys, ill->ill_nd_lla_len);
1350 			return (E2BIG);
1351 		}
1352 		/* Use the link-layer broadcast address for MULTI_BCAST */
1353 		phys_addr = bphys_addr;
1354 		bzero(&v6_extract_mask, sizeof (v6_extract_mask));
1355 		hw_extract_start = ill->ill_nd_lla_len;
1356 	} else {
1357 		phys_addr = alloc_phys;
1358 	}
1359 	if ((ipif->ipif_flags & IPIF_BROADCAST) ||
1360 	    (ill->ill_flags & ILLF_MULTICAST) ||
1361 	    (phyi->phyint_flags & PHYI_MULTI_BCAST)) {
1362 		mutex_enter(&ipst->ips_ndp6->ndp_g_lock);
1363 		err = ndp_add_v6(ill,
1364 		    phys_addr,
1365 		    &v6_mcast_addr,	/* v6 address */
1366 		    &v6_mcast_mask,	/* v6 mask */
1367 		    &v6_extract_mask,
1368 		    hw_extract_start,
1369 		    NCE_F_MAPPING | NCE_F_PERMANENT | NCE_F_NONUD,
1370 		    ND_REACHABLE,
1371 		    &mnce);
1372 		mutex_exit(&ipst->ips_ndp6->ndp_g_lock);
1373 		if (err == 0) {
1374 			if (ret_nce != NULL) {
1375 				*ret_nce = mnce;
1376 			} else {
1377 				NCE_REFRELE(mnce);
1378 			}
1379 		}
1380 	}
1381 	kmem_free(alloc_phys, ill->ill_nd_lla_len);
1382 	return (err);
1383 }
1384 
1385 /*
1386  * Get the resolver set up for a new ipif.  (Always called as writer.)
1387  */
1388 int
1389 ipif_ndp_up(ipif_t *ipif)
1390 {
1391 	ill_t		*ill = ipif->ipif_ill;
1392 	int		err = 0;
1393 	nce_t		*nce = NULL;
1394 	nce_t		*mnce = NULL;
1395 
1396 	ip1dbg(("ipif_ndp_up(%s:%u)\n", ill->ill_name, ipif->ipif_id));
1397 
1398 	/*
1399 	 * ND not supported on XRESOLV interfaces. If ND support (multicast)
1400 	 * added later, take out this check.
1401 	 */
1402 	if ((ill->ill_flags & ILLF_XRESOLV) ||
1403 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr) ||
1404 	    (!(ill->ill_net_type & IRE_INTERFACE))) {
1405 		ipif->ipif_addr_ready = 1;
1406 		return (0);
1407 	}
1408 
1409 	/*
1410 	 * Need to setup multicast mapping only when the first
1411 	 * interface is coming UP.
1412 	 */
1413 	if (ill->ill_ipif_up_count == 0 &&
1414 	    (ill->ill_flags & ILLF_MULTICAST)) {
1415 		/*
1416 		 * We set the multicast before setting up the mapping for
1417 		 * local address because ipif_ndp_setup_multicast does
1418 		 * ndp_walk to delete nces which will delete the mapping
1419 		 * for local address also if we added the mapping for
1420 		 * local address first.
1421 		 */
1422 		err = ipif_ndp_setup_multicast(ipif, &mnce);
1423 		if (err != 0)
1424 			return (err);
1425 	}
1426 
1427 	if ((ipif->ipif_flags & (IPIF_UNNUMBERED|IPIF_NOLOCAL)) == 0) {
1428 		uint16_t	flags;
1429 		uchar_t	*hw_addr = NULL;
1430 
1431 		/* Permanent entries don't need NUD */
1432 		flags = NCE_F_PERMANENT | NCE_F_NONUD;
1433 		if (ill->ill_flags & ILLF_ROUTER)
1434 			flags |= NCE_F_ISROUTER;
1435 
1436 		if (ipif->ipif_flags & IPIF_ANYCAST)
1437 			flags |= NCE_F_ANYCAST;
1438 
1439 		if (ill->ill_net_type == IRE_IF_RESOLVER) {
1440 			hw_addr = ill->ill_nd_lla;
1441 
1442 			if (ill->ill_move_in_progress) {
1443 				/*
1444 				 * Addresses are failing over to this ill.
1445 				 * Don't wait for NUD to see this change.
1446 				 * Publish our new link-layer address.
1447 				 */
1448 				flags |= NCE_F_UNSOL_ADV;
1449 			}
1450 		}
1451 		err = ndp_lookup_then_add_v6(ill,
1452 		    hw_addr,
1453 		    &ipif->ipif_v6lcl_addr,
1454 		    &ipv6_all_ones,
1455 		    &ipv6_all_zeros,
1456 		    0,
1457 		    flags,
1458 		    ND_PROBE,	/* Causes Duplicate Address Detection to run */
1459 		    &nce);
1460 		switch (err) {
1461 		case 0:
1462 			ip1dbg(("ipif_ndp_up: NCE created for %s\n",
1463 			    ill->ill_name));
1464 			ipif->ipif_addr_ready = 1;
1465 			break;
1466 		case EINPROGRESS:
1467 			ip1dbg(("ipif_ndp_up: running DAD now for %s\n",
1468 			    ill->ill_name));
1469 			break;
1470 		case EEXIST:
1471 			NCE_REFRELE(nce);
1472 			ip1dbg(("ipif_ndp_up: NCE already exists for %s\n",
1473 			    ill->ill_name));
1474 			if (mnce != NULL) {
1475 				ndp_delete(mnce);
1476 				NCE_REFRELE(mnce);
1477 			}
1478 			return (err);
1479 		default:
1480 			ip1dbg(("ipif_ndp_up: NCE creation failed %s\n",
1481 			    ill->ill_name));
1482 			if (mnce != NULL) {
1483 				ndp_delete(mnce);
1484 				NCE_REFRELE(mnce);
1485 			}
1486 			return (err);
1487 		}
1488 	} else {
1489 		/* No local NCE for this entry */
1490 		ipif->ipif_addr_ready = 1;
1491 	}
1492 	if (nce != NULL)
1493 		NCE_REFRELE(nce);
1494 	if (mnce != NULL)
1495 		NCE_REFRELE(mnce);
1496 	return (0);
1497 }
1498 
1499 /* Remove all cache entries for this logical interface */
1500 void
1501 ipif_ndp_down(ipif_t *ipif)
1502 {
1503 	nce_t	*nce;
1504 
1505 	if (ipif->ipif_isv6) {
1506 		nce = ndp_lookup_v6(ipif->ipif_ill, &ipif->ipif_v6lcl_addr,
1507 		    B_FALSE);
1508 		if (nce != NULL) {
1509 			ndp_delete(nce);
1510 			NCE_REFRELE(nce);
1511 		}
1512 	}
1513 	/*
1514 	 * Remove mapping and all other nces dependent on this ill
1515 	 * when the last ipif is going away.
1516 	 */
1517 	if (ipif->ipif_ill->ill_ipif_up_count == 0) {
1518 		ndp_walk(ipif->ipif_ill, (pfi_t)ndp_delete_per_ill,
1519 		    (uchar_t *)ipif->ipif_ill, ipif->ipif_ill->ill_ipst);
1520 	}
1521 }
1522 
1523 /*
1524  * Used when an interface comes up to recreate any extra routes on this
1525  * interface.
1526  */
1527 static ire_t **
1528 ipif_recover_ire_v6(ipif_t *ipif)
1529 {
1530 	mblk_t	*mp;
1531 	ire_t   **ipif_saved_irep;
1532 	ire_t   **irep;
1533 	ip_stack_t	*ipst = ipif->ipif_ill->ill_ipst;
1534 
1535 	ip1dbg(("ipif_recover_ire_v6(%s:%u)", ipif->ipif_ill->ill_name,
1536 	    ipif->ipif_id));
1537 
1538 	ASSERT(ipif->ipif_isv6);
1539 
1540 	mutex_enter(&ipif->ipif_saved_ire_lock);
1541 	ipif_saved_irep = (ire_t **)kmem_zalloc(sizeof (ire_t *) *
1542 	    ipif->ipif_saved_ire_cnt, KM_NOSLEEP);
1543 	if (ipif_saved_irep == NULL) {
1544 		mutex_exit(&ipif->ipif_saved_ire_lock);
1545 		return (NULL);
1546 	}
1547 
1548 	irep = ipif_saved_irep;
1549 
1550 	for (mp = ipif->ipif_saved_ire_mp; mp != NULL; mp = mp->b_cont) {
1551 		ire_t		*ire;
1552 		queue_t		*rfq;
1553 		queue_t		*stq;
1554 		ifrt_t		*ifrt;
1555 		in6_addr_t	*src_addr;
1556 		in6_addr_t	*gateway_addr;
1557 		char		buf[INET6_ADDRSTRLEN];
1558 		ushort_t	type;
1559 
1560 		/*
1561 		 * When the ire was initially created and then added in
1562 		 * ip_rt_add_v6(), it was created either using
1563 		 * ipif->ipif_net_type in the case of a traditional interface
1564 		 * route, or as one of the IRE_OFFSUBNET types (with the
1565 		 * exception of IRE_HOST type redirect ire which is created by
1566 		 * icmp_redirect_v6() and which we don't need to save or
1567 		 * recover).  In the case where ipif->ipif_net_type was
1568 		 * IRE_LOOPBACK, ip_rt_add_v6() will update the ire_type to
1569 		 * IRE_IF_NORESOLVER before calling ire_add_v6() to satisfy
1570 		 * software like GateD and Sun Cluster which creates routes
1571 		 * using the the loopback interface's address as a gateway.
1572 		 *
1573 		 * As ifrt->ifrt_type reflects the already updated ire_type,
1574 		 * ire_create_v6() will be called in the same way here as in
1575 		 * ip_rt_add_v6(), namely using ipif->ipif_net_type when the
1576 		 * route looks like a traditional interface route (where
1577 		 * ifrt->ifrt_type & IRE_INTERFACE is true) and otherwise
1578 		 * using the saved ifrt->ifrt_type.  This means that in
1579 		 * the case where ipif->ipif_net_type is IRE_LOOPBACK,
1580 		 * the ire created by ire_create_v6() will be an IRE_LOOPBACK,
1581 		 * it will then be turned into an IRE_IF_NORESOLVER and then
1582 		 * added by ire_add_v6().
1583 		 */
1584 		ifrt = (ifrt_t *)mp->b_rptr;
1585 		if (ifrt->ifrt_type & IRE_INTERFACE) {
1586 			rfq = NULL;
1587 			stq = (ipif->ipif_net_type == IRE_IF_RESOLVER)
1588 			    ? ipif->ipif_rq : ipif->ipif_wq;
1589 			src_addr = (ifrt->ifrt_flags & RTF_SETSRC)
1590 			    ? &ifrt->ifrt_v6src_addr
1591 			    : &ipif->ipif_v6src_addr;
1592 			gateway_addr = NULL;
1593 			type = ipif->ipif_net_type;
1594 		} else {
1595 			rfq = NULL;
1596 			stq = NULL;
1597 			src_addr = (ifrt->ifrt_flags & RTF_SETSRC)
1598 			    ? &ifrt->ifrt_v6src_addr : NULL;
1599 			gateway_addr = &ifrt->ifrt_v6gateway_addr;
1600 			type = ifrt->ifrt_type;
1601 		}
1602 
1603 		/*
1604 		 * Create a copy of the IRE with the saved address and netmask.
1605 		 */
1606 		ip1dbg(("ipif_recover_ire_v6: creating IRE %s (%d) for %s/%d\n",
1607 		    ip_nv_lookup(ire_nv_tbl, ifrt->ifrt_type), ifrt->ifrt_type,
1608 		    inet_ntop(AF_INET6, &ifrt->ifrt_v6addr, buf, sizeof (buf)),
1609 		    ip_mask_to_plen_v6(&ifrt->ifrt_v6mask)));
1610 		ire = ire_create_v6(
1611 		    &ifrt->ifrt_v6addr,
1612 		    &ifrt->ifrt_v6mask,
1613 		    src_addr,
1614 		    gateway_addr,
1615 		    &ifrt->ifrt_max_frag,
1616 		    NULL,
1617 		    rfq,
1618 		    stq,
1619 		    type,
1620 		    ipif,
1621 		    NULL,
1622 		    0,
1623 		    0,
1624 		    ifrt->ifrt_flags,
1625 		    &ifrt->ifrt_iulp_info,
1626 		    NULL,
1627 		    NULL,
1628 		    ipst);
1629 		if (ire == NULL) {
1630 			mutex_exit(&ipif->ipif_saved_ire_lock);
1631 			kmem_free(ipif_saved_irep,
1632 			    ipif->ipif_saved_ire_cnt * sizeof (ire_t *));
1633 			return (NULL);
1634 		}
1635 
1636 		/*
1637 		 * Some software (for example, GateD and Sun Cluster) attempts
1638 		 * to create (what amount to) IRE_PREFIX routes with the
1639 		 * loopback address as the gateway.  This is primarily done to
1640 		 * set up prefixes with the RTF_REJECT flag set (for example,
1641 		 * when generating aggregate routes.)
1642 		 *
1643 		 * If the IRE type (as defined by ipif->ipif_net_type) is
1644 		 * IRE_LOOPBACK, then we map the request into a
1645 		 * IRE_IF_NORESOLVER.
1646 		 */
1647 		if (ipif->ipif_net_type == IRE_LOOPBACK)
1648 			ire->ire_type = IRE_IF_NORESOLVER;
1649 		/*
1650 		 * ire held by ire_add, will be refreled' in ipif_up_done
1651 		 * towards the end
1652 		 */
1653 		(void) ire_add(&ire, NULL, NULL, NULL, B_FALSE);
1654 		*irep = ire;
1655 		irep++;
1656 		ip1dbg(("ipif_recover_ire_v6: added ire %p\n", (void *)ire));
1657 	}
1658 	mutex_exit(&ipif->ipif_saved_ire_lock);
1659 	return (ipif_saved_irep);
1660 }
1661 
1662 /*
1663  * Return the scope of the given IPv6 address.  If the address is an
1664  * IPv4 mapped IPv6 address, return the scope of the corresponding
1665  * IPv4 address.
1666  */
1667 in6addr_scope_t
1668 ip_addr_scope_v6(const in6_addr_t *addr)
1669 {
1670 	static in6_addr_t ipv6loopback = IN6ADDR_LOOPBACK_INIT;
1671 
1672 	if (IN6_IS_ADDR_V4MAPPED(addr)) {
1673 		in_addr_t v4addr_h = ntohl(V4_PART_OF_V6((*addr)));
1674 		if ((v4addr_h >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
1675 		    (v4addr_h & IN_AUTOCONF_MASK) == IN_AUTOCONF_NET)
1676 			return (IP6_SCOPE_LINKLOCAL);
1677 		if ((v4addr_h & IN_PRIVATE8_MASK) == IN_PRIVATE8_NET ||
1678 		    (v4addr_h & IN_PRIVATE12_MASK) == IN_PRIVATE12_NET ||
1679 		    (v4addr_h & IN_PRIVATE16_MASK) == IN_PRIVATE16_NET)
1680 			return (IP6_SCOPE_SITELOCAL);
1681 		return (IP6_SCOPE_GLOBAL);
1682 	}
1683 
1684 	if (IN6_IS_ADDR_MULTICAST(addr))
1685 		return (IN6_ADDR_MC_SCOPE(addr));
1686 
1687 	/* link-local and loopback addresses are of link-local scope */
1688 	if (IN6_IS_ADDR_LINKLOCAL(addr) ||
1689 	    IN6_ARE_ADDR_EQUAL(addr, &ipv6loopback))
1690 		return (IP6_SCOPE_LINKLOCAL);
1691 	if (IN6_IS_ADDR_SITELOCAL(addr))
1692 		return (IP6_SCOPE_SITELOCAL);
1693 	return (IP6_SCOPE_GLOBAL);
1694 }
1695 
1696 
1697 /*
1698  * Returns the length of the common prefix of a1 and a2, as per
1699  * CommonPrefixLen() defined in RFC 3484.
1700  */
1701 static int
1702 ip_common_prefix_v6(const in6_addr_t *a1, const in6_addr_t *a2)
1703 {
1704 	int i;
1705 	uint32_t a1val, a2val, mask;
1706 
1707 	for (i = 0; i < 4; i++) {
1708 		if ((a1val = a1->s6_addr32[i]) != (a2val = a2->s6_addr32[i])) {
1709 			a1val ^= a2val;
1710 			i *= 32;
1711 			mask = 0x80000000u;
1712 			while (!(a1val & mask)) {
1713 				mask >>= 1;
1714 				i++;
1715 			}
1716 			return (i);
1717 		}
1718 	}
1719 	return (IPV6_ABITS);
1720 }
1721 
1722 #define	IPIF_VALID_IPV6_SOURCE(ipif) \
1723 	(((ipif)->ipif_flags & IPIF_UP) && \
1724 	!((ipif)->ipif_flags & (IPIF_NOLOCAL|IPIF_ANYCAST)) && \
1725 	(ipif)->ipif_addr_ready)
1726 
1727 /* source address candidate */
1728 typedef struct candidate {
1729 	ipif_t		*cand_ipif;
1730 	/* The properties of this candidate */
1731 	boolean_t	cand_isdst;
1732 	boolean_t	cand_isdst_set;
1733 	in6addr_scope_t	cand_scope;
1734 	boolean_t	cand_scope_set;
1735 	boolean_t	cand_isdeprecated;
1736 	boolean_t	cand_isdeprecated_set;
1737 	boolean_t	cand_ispreferred;
1738 	boolean_t	cand_ispreferred_set;
1739 	boolean_t	cand_matchedinterface;
1740 	boolean_t	cand_matchedinterface_set;
1741 	boolean_t	cand_matchedlabel;
1742 	boolean_t	cand_matchedlabel_set;
1743 	boolean_t	cand_istmp;
1744 	boolean_t	cand_istmp_set;
1745 	int		cand_common_pref;
1746 	boolean_t	cand_common_pref_set;
1747 	boolean_t	cand_pref_eq;
1748 	boolean_t	cand_pref_eq_set;
1749 	int		cand_pref_len;
1750 	boolean_t	cand_pref_len_set;
1751 } cand_t;
1752 #define	cand_srcaddr	cand_ipif->ipif_v6lcl_addr
1753 #define	cand_mask	cand_ipif->ipif_v6net_mask
1754 #define	cand_flags	cand_ipif->ipif_flags
1755 #define	cand_ill	cand_ipif->ipif_ill
1756 #define	cand_zoneid	cand_ipif->ipif_zoneid
1757 
1758 /* information about the destination for source address selection */
1759 typedef struct dstinfo {
1760 	const in6_addr_t	*dst_addr;
1761 	ill_t			*dst_ill;
1762 	uint_t			dst_restrict_ill;
1763 	boolean_t		dst_prefer_src_tmp;
1764 	in6addr_scope_t		dst_scope;
1765 	char			*dst_label;
1766 } dstinfo_t;
1767 
1768 /*
1769  * The following functions are rules used to select a source address in
1770  * ipif_select_source_v6().  Each rule compares a current candidate (cc)
1771  * against the best candidate (bc).  Each rule has three possible outcomes;
1772  * the candidate is preferred over the best candidate (CAND_PREFER), the
1773  * candidate is not preferred over the best candidate (CAND_AVOID), or the
1774  * candidate is of equal value as the best candidate (CAND_TIE).
1775  *
1776  * These rules are part of a greater "Default Address Selection for IPv6"
1777  * sheme, which is standards based work coming out of the IETF ipv6 working
1778  * group.  The IETF document defines both IPv6 source address selection and
1779  * destination address ordering.  The rules defined here implement the IPv6
1780  * source address selection.  Destination address ordering is done by
1781  * libnsl, and uses a similar set of rules to implement the sorting.
1782  *
1783  * Most of the rules are defined by the RFC and are not typically altered.  The
1784  * last rule, number 8, has language that allows for local preferences.  In the
1785  * scheme below, this means that new Solaris rules should normally go between
1786  * rule_ifprefix and rule_prefix.
1787  */
1788 typedef enum {CAND_AVOID, CAND_TIE, CAND_PREFER} rule_res_t;
1789 typedef	rule_res_t (*rulef_t)(cand_t *, cand_t *, const dstinfo_t *,
1790     ip_stack_t *);
1791 
1792 /* Prefer an address if it is equal to the destination address. */
1793 /* ARGSUSED3 */
1794 static rule_res_t
1795 rule_isdst(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1796 {
1797 	if (!bc->cand_isdst_set) {
1798 		bc->cand_isdst =
1799 		    IN6_ARE_ADDR_EQUAL(&bc->cand_srcaddr, dstinfo->dst_addr);
1800 		bc->cand_isdst_set = B_TRUE;
1801 	}
1802 
1803 	cc->cand_isdst =
1804 	    IN6_ARE_ADDR_EQUAL(&cc->cand_srcaddr, dstinfo->dst_addr);
1805 	cc->cand_isdst_set = B_TRUE;
1806 
1807 	if (cc->cand_isdst == bc->cand_isdst)
1808 		return (CAND_TIE);
1809 	else if (cc->cand_isdst)
1810 		return (CAND_PREFER);
1811 	else
1812 		return (CAND_AVOID);
1813 }
1814 
1815 /*
1816  * Prefer addresses that are of closest scope to the destination.  Always
1817  * prefer addresses that are of greater scope than the destination over
1818  * those that are of lesser scope than the destination.
1819  */
1820 /* ARGSUSED3 */
1821 static rule_res_t
1822 rule_scope(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1823 {
1824 	if (!bc->cand_scope_set) {
1825 		bc->cand_scope = ip_addr_scope_v6(&bc->cand_srcaddr);
1826 		bc->cand_scope_set = B_TRUE;
1827 	}
1828 
1829 	cc->cand_scope = ip_addr_scope_v6(&cc->cand_srcaddr);
1830 	cc->cand_scope_set = B_TRUE;
1831 
1832 	if (cc->cand_scope < bc->cand_scope) {
1833 		if (cc->cand_scope < dstinfo->dst_scope)
1834 			return (CAND_AVOID);
1835 		else
1836 			return (CAND_PREFER);
1837 	} else if (bc->cand_scope < cc->cand_scope) {
1838 		if (bc->cand_scope < dstinfo->dst_scope)
1839 			return (CAND_PREFER);
1840 		else
1841 			return (CAND_AVOID);
1842 	} else {
1843 		return (CAND_TIE);
1844 	}
1845 }
1846 
1847 /*
1848  * Prefer non-deprecated source addresses.
1849  */
1850 /* ARGSUSED2 */
1851 static rule_res_t
1852 rule_deprecated(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1853     ip_stack_t *ipst)
1854 {
1855 	if (!bc->cand_isdeprecated_set) {
1856 		bc->cand_isdeprecated =
1857 		    ((bc->cand_flags & IPIF_DEPRECATED) != 0);
1858 		bc->cand_isdeprecated_set = B_TRUE;
1859 	}
1860 
1861 	cc->cand_isdeprecated = ((cc->cand_flags & IPIF_DEPRECATED) != 0);
1862 	cc->cand_isdeprecated_set = B_TRUE;
1863 
1864 	if (bc->cand_isdeprecated == cc->cand_isdeprecated)
1865 		return (CAND_TIE);
1866 	else if (cc->cand_isdeprecated)
1867 		return (CAND_AVOID);
1868 	else
1869 		return (CAND_PREFER);
1870 }
1871 
1872 /*
1873  * Prefer source addresses that have the IPIF_PREFERRED flag set.  This
1874  * rule must be before rule_interface because the flag could be set on any
1875  * interface, not just the interface being used for outgoing packets (for
1876  * example, the IFF_PREFERRED could be set on an address assigned to the
1877  * loopback interface).
1878  */
1879 /* ARGSUSED2 */
1880 static rule_res_t
1881 rule_preferred(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1882     ip_stack_t *ipst)
1883 {
1884 	if (!bc->cand_ispreferred_set) {
1885 		bc->cand_ispreferred = ((bc->cand_flags & IPIF_PREFERRED) != 0);
1886 		bc->cand_ispreferred_set = B_TRUE;
1887 	}
1888 
1889 	cc->cand_ispreferred = ((cc->cand_flags & IPIF_PREFERRED) != 0);
1890 	cc->cand_ispreferred_set = B_TRUE;
1891 
1892 	if (bc->cand_ispreferred == cc->cand_ispreferred)
1893 		return (CAND_TIE);
1894 	else if (cc->cand_ispreferred)
1895 		return (CAND_PREFER);
1896 	else
1897 		return (CAND_AVOID);
1898 }
1899 
1900 /*
1901  * Prefer source addresses that are assigned to the outgoing interface, or
1902  * to an interface that is in the same IPMP group as the outgoing
1903  * interface.
1904  */
1905 /* ARGSUSED3 */
1906 static rule_res_t
1907 rule_interface(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1908     ip_stack_t *ipst)
1909 {
1910 	ill_t *dstill = dstinfo->dst_ill;
1911 
1912 	/*
1913 	 * If dstinfo->dst_restrict_ill is set, this rule is unnecessary
1914 	 * since we know all candidates will be on the same link.
1915 	 */
1916 	if (dstinfo->dst_restrict_ill)
1917 		return (CAND_TIE);
1918 
1919 	if (!bc->cand_matchedinterface_set) {
1920 		bc->cand_matchedinterface = (bc->cand_ill == dstill ||
1921 		    (dstill->ill_group != NULL &&
1922 		    dstill->ill_group == bc->cand_ill->ill_group));
1923 		bc->cand_matchedinterface_set = B_TRUE;
1924 	}
1925 
1926 	cc->cand_matchedinterface = (cc->cand_ill == dstill ||
1927 	    (dstill->ill_group != NULL &&
1928 	    dstill->ill_group == cc->cand_ill->ill_group));
1929 	cc->cand_matchedinterface_set = B_TRUE;
1930 
1931 	if (bc->cand_matchedinterface == cc->cand_matchedinterface)
1932 		return (CAND_TIE);
1933 	else if (cc->cand_matchedinterface)
1934 		return (CAND_PREFER);
1935 	else
1936 		return (CAND_AVOID);
1937 }
1938 
1939 /*
1940  * Prefer source addresses whose label matches the destination's label.
1941  */
1942 static rule_res_t
1943 rule_label(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
1944 {
1945 	char *label;
1946 
1947 	if (!bc->cand_matchedlabel_set) {
1948 		label = ip6_asp_lookup(&bc->cand_srcaddr, NULL, ipst);
1949 		bc->cand_matchedlabel =
1950 		    ip6_asp_labelcmp(label, dstinfo->dst_label);
1951 		bc->cand_matchedlabel_set = B_TRUE;
1952 	}
1953 
1954 	label = ip6_asp_lookup(&cc->cand_srcaddr, NULL, ipst);
1955 	cc->cand_matchedlabel = ip6_asp_labelcmp(label, dstinfo->dst_label);
1956 	cc->cand_matchedlabel_set = B_TRUE;
1957 
1958 	if (bc->cand_matchedlabel == cc->cand_matchedlabel)
1959 		return (CAND_TIE);
1960 	else if (cc->cand_matchedlabel)
1961 		return (CAND_PREFER);
1962 	else
1963 		return (CAND_AVOID);
1964 }
1965 
1966 /*
1967  * Prefer public addresses over temporary ones.  An application can reverse
1968  * the logic of this rule and prefer temporary addresses by using the
1969  * IPV6_SRC_PREFERENCES socket option.
1970  */
1971 /* ARGSUSED3 */
1972 static rule_res_t
1973 rule_temporary(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
1974     ip_stack_t *ipst)
1975 {
1976 	if (!bc->cand_istmp_set) {
1977 		bc->cand_istmp = ((bc->cand_flags & IPIF_TEMPORARY) != 0);
1978 		bc->cand_istmp_set = B_TRUE;
1979 	}
1980 
1981 	cc->cand_istmp = ((cc->cand_flags & IPIF_TEMPORARY) != 0);
1982 	cc->cand_istmp_set = B_TRUE;
1983 
1984 	if (bc->cand_istmp == cc->cand_istmp)
1985 		return (CAND_TIE);
1986 
1987 	if (dstinfo->dst_prefer_src_tmp && cc->cand_istmp)
1988 		return (CAND_PREFER);
1989 	else if (!dstinfo->dst_prefer_src_tmp && !cc->cand_istmp)
1990 		return (CAND_PREFER);
1991 	else
1992 		return (CAND_AVOID);
1993 }
1994 
1995 /*
1996  * Prefer source addresses with longer matching prefix with the destination
1997  * under the interface mask.  This gets us on the same subnet before applying
1998  * any Solaris-specific rules.
1999  */
2000 /* ARGSUSED3 */
2001 static rule_res_t
2002 rule_ifprefix(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2003     ip_stack_t *ipst)
2004 {
2005 	if (!bc->cand_pref_eq_set) {
2006 		bc->cand_pref_eq = V6_MASK_EQ_2(bc->cand_srcaddr,
2007 		    bc->cand_mask, *dstinfo->dst_addr);
2008 		bc->cand_pref_eq_set = B_TRUE;
2009 	}
2010 
2011 	cc->cand_pref_eq = V6_MASK_EQ_2(cc->cand_srcaddr, cc->cand_mask,
2012 	    *dstinfo->dst_addr);
2013 	cc->cand_pref_eq_set = B_TRUE;
2014 
2015 	if (bc->cand_pref_eq) {
2016 		if (cc->cand_pref_eq) {
2017 			if (!bc->cand_pref_len_set) {
2018 				bc->cand_pref_len =
2019 				    ip_mask_to_plen_v6(&bc->cand_mask);
2020 				bc->cand_pref_len_set = B_TRUE;
2021 			}
2022 			cc->cand_pref_len = ip_mask_to_plen_v6(&cc->cand_mask);
2023 			cc->cand_pref_len_set = B_TRUE;
2024 			if (bc->cand_pref_len == cc->cand_pref_len)
2025 				return (CAND_TIE);
2026 			else if (bc->cand_pref_len > cc->cand_pref_len)
2027 				return (CAND_AVOID);
2028 			else
2029 				return (CAND_PREFER);
2030 		} else {
2031 			return (CAND_AVOID);
2032 		}
2033 	} else {
2034 		if (cc->cand_pref_eq)
2035 			return (CAND_PREFER);
2036 		else
2037 			return (CAND_TIE);
2038 	}
2039 }
2040 
2041 /*
2042  * Prefer to use zone-specific addresses when possible instead of all-zones
2043  * addresses.
2044  */
2045 /* ARGSUSED2 */
2046 static rule_res_t
2047 rule_zone_specific(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2048     ip_stack_t *ipst)
2049 {
2050 	if ((bc->cand_zoneid == ALL_ZONES) ==
2051 	    (cc->cand_zoneid == ALL_ZONES))
2052 		return (CAND_TIE);
2053 	else if (cc->cand_zoneid == ALL_ZONES)
2054 		return (CAND_AVOID);
2055 	else
2056 		return (CAND_PREFER);
2057 }
2058 
2059 /*
2060  * Prefer to use DHCPv6 (first) and static addresses (second) when possible
2061  * instead of statelessly autoconfigured addresses.
2062  *
2063  * This is done after trying all other preferences (and before the final tie
2064  * breaker) so that, if all else is equal, we select addresses configured by
2065  * DHCPv6 over other addresses.  We presume that DHCPv6 addresses, unlike
2066  * stateless autoconfigured addresses, are deliberately configured by an
2067  * administrator, and thus are correctly set up in DNS and network packet
2068  * filters.
2069  */
2070 /* ARGSUSED2 */
2071 static rule_res_t
2072 rule_addr_type(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2073     ip_stack_t *ipst)
2074 {
2075 #define	ATYPE(x)	\
2076 	((x) & IPIF_DHCPRUNNING) ? 1 : ((x) & IPIF_ADDRCONF) ? 3 : 2
2077 	int bcval = ATYPE(bc->cand_flags);
2078 	int ccval = ATYPE(cc->cand_flags);
2079 #undef ATYPE
2080 
2081 	if (bcval == ccval)
2082 		return (CAND_TIE);
2083 	else if (ccval < bcval)
2084 		return (CAND_PREFER);
2085 	else
2086 		return (CAND_AVOID);
2087 }
2088 
2089 /*
2090  * Prefer source addresses with longer matching prefix with the destination.
2091  * We do the longest matching prefix calculation by doing an xor of both
2092  * addresses with the destination, and pick the address with the longest string
2093  * of leading zeros, as per CommonPrefixLen() defined in RFC 3484.
2094  */
2095 /* ARGSUSED3 */
2096 static rule_res_t
2097 rule_prefix(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo, ip_stack_t *ipst)
2098 {
2099 	if (!bc->cand_common_pref_set) {
2100 		bc->cand_common_pref = ip_common_prefix_v6(&bc->cand_srcaddr,
2101 		    dstinfo->dst_addr);
2102 		bc->cand_common_pref_set = B_TRUE;
2103 	}
2104 
2105 	cc->cand_common_pref = ip_common_prefix_v6(&cc->cand_srcaddr,
2106 	    dstinfo->dst_addr);
2107 	cc->cand_common_pref_set = B_TRUE;
2108 
2109 	if (bc->cand_common_pref == cc->cand_common_pref)
2110 		return (CAND_TIE);
2111 	else if (bc->cand_common_pref > cc->cand_common_pref)
2112 		return (CAND_AVOID);
2113 	else
2114 		return (CAND_PREFER);
2115 }
2116 
2117 /*
2118  * Last rule: we must pick something, so just prefer the current best
2119  * candidate.
2120  */
2121 /* ARGSUSED */
2122 static rule_res_t
2123 rule_must_be_last(cand_t *bc, cand_t *cc, const dstinfo_t *dstinfo,
2124     ip_stack_t *ipst)
2125 {
2126 	return (CAND_AVOID);
2127 }
2128 
2129 /*
2130  * Determine the best source address given a destination address and a
2131  * destination ill.  If no suitable source address is found, it returns
2132  * NULL. If there is a usable address pointed to by the usesrc
2133  * (i.e ill_usesrc_ifindex != 0) then return that first since it is more
2134  * fine grained (i.e per interface)
2135  *
2136  * This implementation is based on the "Default Address Selection for IPv6"
2137  * specification produced by the IETF IPv6 working group.  It has been
2138  * implemented so that the list of addresses is only traversed once (the
2139  * specification's algorithm could traverse the list of addresses once for
2140  * every rule).
2141  *
2142  * The restrict_ill argument restricts the algorithm to chose a source
2143  * address that is assigned to the destination ill or an ill in the same
2144  * IPMP group as the destination ill.  This is used when the destination
2145  * address is a link-local or multicast address, and when
2146  * ipv6_strict_dst_multihoming is turned on.
2147  *
2148  * src_prefs is the caller's set of source address preferences.  If source
2149  * address selection is being called to determine the source address of a
2150  * connected socket (from ip_bind_connected_v6()), then the preferences are
2151  * taken from conn_src_preferences.  These preferences can be set on a
2152  * per-socket basis using the IPV6_SRC_PREFERENCES socket option.  The only
2153  * preference currently implemented is for rfc3041 temporary addresses.
2154  */
2155 ipif_t *
2156 ipif_select_source_v6(ill_t *dstill, const in6_addr_t *dst,
2157     uint_t restrict_ill, uint32_t src_prefs, zoneid_t zoneid)
2158 {
2159 	dstinfo_t	dstinfo;
2160 	char		dstr[INET6_ADDRSTRLEN];
2161 	char		sstr[INET6_ADDRSTRLEN];
2162 	ipif_t		*ipif;
2163 	ill_t		*ill, *usesrc_ill = NULL;
2164 	ill_walk_context_t	ctx;
2165 	cand_t		best_c;	/* The best candidate */
2166 	cand_t		curr_c;	/* The current candidate */
2167 	uint_t		index;
2168 	boolean_t	first_candidate = B_TRUE;
2169 	rule_res_t	rule_result;
2170 	tsol_tpc_t	*src_rhtp, *dst_rhtp;
2171 	ip_stack_t	*ipst = dstill->ill_ipst;
2172 
2173 	/*
2174 	 * The list of ordering rules.  They are applied in the order they
2175 	 * appear in the list.
2176 	 *
2177 	 * Solaris doesn't currently support Mobile IPv6, so there's no
2178 	 * rule_mipv6 corresponding to rule 4 in the specification.
2179 	 */
2180 	rulef_t	rules[] = {
2181 		rule_isdst,
2182 		rule_scope,
2183 		rule_deprecated,
2184 		rule_preferred,
2185 		rule_interface,
2186 		rule_label,
2187 		rule_temporary,
2188 		rule_ifprefix,			/* local rules after this */
2189 		rule_zone_specific,
2190 		rule_addr_type,
2191 		rule_prefix,			/* local rules before this */
2192 		rule_must_be_last,		/* must always be last */
2193 		NULL
2194 	};
2195 
2196 	ASSERT(dstill->ill_isv6);
2197 	ASSERT(!IN6_IS_ADDR_V4MAPPED(dst));
2198 
2199 	/*
2200 	 * Check if there is a usable src address pointed to by the
2201 	 * usesrc ifindex. This has higher precedence since it is
2202 	 * finer grained (i.e per interface) v/s being system wide.
2203 	 */
2204 	if (dstill->ill_usesrc_ifindex != 0) {
2205 		if ((usesrc_ill =
2206 		    ill_lookup_on_ifindex(dstill->ill_usesrc_ifindex, B_TRUE,
2207 		    NULL, NULL, NULL, NULL, ipst)) != NULL) {
2208 			dstinfo.dst_ill = usesrc_ill;
2209 		} else {
2210 			return (NULL);
2211 		}
2212 	} else {
2213 		dstinfo.dst_ill = dstill;
2214 	}
2215 
2216 	/*
2217 	 * If we're dealing with an unlabeled destination on a labeled system,
2218 	 * make sure that we ignore source addresses that are incompatible with
2219 	 * the destination's default label.  That destination's default label
2220 	 * must dominate the minimum label on the source address.
2221 	 *
2222 	 * (Note that this has to do with Trusted Solaris.  It's not related to
2223 	 * the labels described by ip6_asp_lookup.)
2224 	 */
2225 	dst_rhtp = NULL;
2226 	if (is_system_labeled()) {
2227 		dst_rhtp = find_tpc(dst, IPV6_VERSION, B_FALSE);
2228 		if (dst_rhtp == NULL)
2229 			return (NULL);
2230 		if (dst_rhtp->tpc_tp.host_type != UNLABELED) {
2231 			TPC_RELE(dst_rhtp);
2232 			dst_rhtp = NULL;
2233 		}
2234 	}
2235 
2236 	dstinfo.dst_addr = dst;
2237 	dstinfo.dst_scope = ip_addr_scope_v6(dst);
2238 	dstinfo.dst_label = ip6_asp_lookup(dst, NULL, ipst);
2239 	dstinfo.dst_prefer_src_tmp = ((src_prefs & IPV6_PREFER_SRC_TMP) != 0);
2240 
2241 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
2242 	/*
2243 	 * Section three of the I-D states that for multicast and
2244 	 * link-local destinations, the candidate set must be restricted to
2245 	 * an interface that is on the same link as the outgoing interface.
2246 	 * Also, when ipv6_strict_dst_multihoming is turned on, always
2247 	 * restrict the source address to the destination link as doing
2248 	 * otherwise will almost certainly cause problems.
2249 	 */
2250 	if (IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst) ||
2251 	    ipst->ips_ipv6_strict_dst_multihoming || usesrc_ill != NULL) {
2252 		if (restrict_ill == RESTRICT_TO_NONE)
2253 			dstinfo.dst_restrict_ill = RESTRICT_TO_GROUP;
2254 		else
2255 			dstinfo.dst_restrict_ill = restrict_ill;
2256 	} else {
2257 		dstinfo.dst_restrict_ill = restrict_ill;
2258 	}
2259 
2260 	bzero(&best_c, sizeof (cand_t));
2261 
2262 	/*
2263 	 * Take a pass through the list of IPv6 interfaces to chose the
2264 	 * best possible source address.  If restrict_ill is true, we only
2265 	 * iterate through the ill's that are in the same IPMP group as the
2266 	 * destination's outgoing ill.  If restrict_ill is false, we walk
2267 	 * the entire list of IPv6 ill's.
2268 	 */
2269 	if (dstinfo.dst_restrict_ill != RESTRICT_TO_NONE) {
2270 		if (dstinfo.dst_ill->ill_group != NULL &&
2271 		    dstinfo.dst_restrict_ill == RESTRICT_TO_GROUP) {
2272 			ill = dstinfo.dst_ill->ill_group->illgrp_ill;
2273 		} else {
2274 			ill = dstinfo.dst_ill;
2275 		}
2276 	} else {
2277 		ill = ILL_START_WALK_V6(&ctx, ipst);
2278 	}
2279 
2280 	while (ill != NULL) {
2281 		ASSERT(ill->ill_isv6);
2282 
2283 		/*
2284 		 * Avoid FAILED/OFFLINE ills.
2285 		 * Global and site local addresses will failover and
2286 		 * will be available on the new ill.
2287 		 * But link local addresses don't move.
2288 		 */
2289 		if (dstinfo.dst_restrict_ill != RESTRICT_TO_ILL &&
2290 		    ill->ill_phyint->phyint_flags &
2291 		    (PHYI_OFFLINE | PHYI_FAILED))
2292 			goto next_ill;
2293 
2294 		for (ipif = ill->ill_ipif; ipif != NULL;
2295 		    ipif = ipif->ipif_next) {
2296 
2297 			if (!IPIF_VALID_IPV6_SOURCE(ipif))
2298 				continue;
2299 
2300 			if (zoneid != ALL_ZONES &&
2301 			    ipif->ipif_zoneid != zoneid &&
2302 			    ipif->ipif_zoneid != ALL_ZONES)
2303 				continue;
2304 
2305 			/*
2306 			 * Check compatibility of local address for
2307 			 * destination's default label if we're on a labeled
2308 			 * system.  Incompatible addresses can't be used at
2309 			 * all and must be skipped over.
2310 			 */
2311 			if (dst_rhtp != NULL) {
2312 				boolean_t incompat;
2313 
2314 				src_rhtp = find_tpc(&ipif->ipif_v6lcl_addr,
2315 				    IPV6_VERSION, B_FALSE);
2316 				if (src_rhtp == NULL)
2317 					continue;
2318 				incompat =
2319 				    src_rhtp->tpc_tp.host_type != SUN_CIPSO ||
2320 				    src_rhtp->tpc_tp.tp_doi !=
2321 				    dst_rhtp->tpc_tp.tp_doi ||
2322 				    (!_blinrange(&dst_rhtp->tpc_tp.tp_def_label,
2323 				    &src_rhtp->tpc_tp.tp_sl_range_cipso) &&
2324 				    !blinlset(&dst_rhtp->tpc_tp.tp_def_label,
2325 				    src_rhtp->tpc_tp.tp_sl_set_cipso));
2326 				TPC_RELE(src_rhtp);
2327 				if (incompat)
2328 					continue;
2329 			}
2330 
2331 			if (first_candidate) {
2332 				/*
2333 				 * This is first valid address in the list.
2334 				 * It is automatically the best candidate
2335 				 * so far.
2336 				 */
2337 				best_c.cand_ipif = ipif;
2338 				first_candidate = B_FALSE;
2339 				continue;
2340 			}
2341 
2342 			bzero(&curr_c, sizeof (cand_t));
2343 			curr_c.cand_ipif = ipif;
2344 
2345 			/*
2346 			 * Compare this current candidate (curr_c) with the
2347 			 * best candidate (best_c) by applying the
2348 			 * comparison rules in order until one breaks the
2349 			 * tie.
2350 			 */
2351 			for (index = 0; rules[index] != NULL; index++) {
2352 				/* Apply a comparison rule. */
2353 				rule_result =
2354 				    (rules[index])(&best_c, &curr_c, &dstinfo,
2355 				    ipst);
2356 				if (rule_result == CAND_AVOID) {
2357 					/*
2358 					 * The best candidate is still the
2359 					 * best candidate.  Forget about
2360 					 * this current candidate and go on
2361 					 * to the next one.
2362 					 */
2363 					break;
2364 				} else if (rule_result == CAND_PREFER) {
2365 					/*
2366 					 * This candidate is prefered.  It
2367 					 * becomes the best candidate so
2368 					 * far.  Go on to the next address.
2369 					 */
2370 					best_c = curr_c;
2371 					break;
2372 				}
2373 				/* We have a tie, apply the next rule. */
2374 			}
2375 
2376 			/*
2377 			 * The last rule must be a tie breaker rule and
2378 			 * must never produce a tie.  At this point, the
2379 			 * candidate should have either been rejected, or
2380 			 * have been prefered as the best candidate so far.
2381 			 */
2382 			ASSERT(rule_result != CAND_TIE);
2383 		}
2384 
2385 		/*
2386 		 * We may be walking the linked-list of ill's in an
2387 		 * IPMP group or traversing the IPv6 ill avl tree. If it is a
2388 		 * usesrc ILL then it can't be part of IPMP group and we
2389 		 * will exit the while loop.
2390 		 */
2391 next_ill:
2392 		if (dstinfo.dst_restrict_ill == RESTRICT_TO_ILL)
2393 			ill = NULL;
2394 		else if (dstinfo.dst_restrict_ill == RESTRICT_TO_GROUP)
2395 			ill = ill->ill_group_next;
2396 		else
2397 			ill = ill_next(&ctx, ill);
2398 	}
2399 
2400 	ipif = best_c.cand_ipif;
2401 	ip1dbg(("ipif_select_source_v6(%s, %s) -> %s\n",
2402 	    dstinfo.dst_ill->ill_name,
2403 	    inet_ntop(AF_INET6, dstinfo.dst_addr, dstr, sizeof (dstr)),
2404 	    (ipif == NULL ? "NULL" :
2405 	    inet_ntop(AF_INET6, &ipif->ipif_v6lcl_addr, sstr, sizeof (sstr)))));
2406 
2407 	if (usesrc_ill != NULL)
2408 		ill_refrele(usesrc_ill);
2409 
2410 	if (dst_rhtp != NULL)
2411 		TPC_RELE(dst_rhtp);
2412 
2413 	if (ipif == NULL) {
2414 		rw_exit(&ipst->ips_ill_g_lock);
2415 		return (NULL);
2416 	}
2417 
2418 	mutex_enter(&ipif->ipif_ill->ill_lock);
2419 	if (IPIF_CAN_LOOKUP(ipif)) {
2420 		ipif_refhold_locked(ipif);
2421 		mutex_exit(&ipif->ipif_ill->ill_lock);
2422 		rw_exit(&ipst->ips_ill_g_lock);
2423 		return (ipif);
2424 	}
2425 	mutex_exit(&ipif->ipif_ill->ill_lock);
2426 	rw_exit(&ipst->ips_ill_g_lock);
2427 	ip1dbg(("ipif_select_source_v6 cannot lookup ipif %p"
2428 	    " returning null \n", (void *)ipif));
2429 
2430 	return (NULL);
2431 }
2432 
2433 /*
2434  * If old_ipif is not NULL, see if ipif was derived from old
2435  * ipif and if so, recreate the interface route by re-doing
2436  * source address selection. This happens when ipif_down ->
2437  * ipif_update_other_ipifs calls us.
2438  *
2439  * If old_ipif is NULL, just redo the source address selection
2440  * if needed. This happens when illgrp_insert or ipif_up_done_v6
2441  * calls us.
2442  */
2443 void
2444 ipif_recreate_interface_routes_v6(ipif_t *old_ipif, ipif_t *ipif)
2445 {
2446 	ire_t *ire;
2447 	ire_t *ipif_ire;
2448 	queue_t *stq;
2449 	ill_t *ill;
2450 	ipif_t *nipif = NULL;
2451 	boolean_t nipif_refheld = B_FALSE;
2452 	boolean_t ip6_asp_table_held = B_FALSE;
2453 	ip_stack_t	*ipst = ipif->ipif_ill->ill_ipst;
2454 
2455 	ill = ipif->ipif_ill;
2456 
2457 	if (!(ipif->ipif_flags &
2458 	    (IPIF_NOLOCAL|IPIF_ANYCAST|IPIF_DEPRECATED))) {
2459 		/*
2460 		 * Can't possibly have borrowed the source
2461 		 * from old_ipif.
2462 		 */
2463 		return;
2464 	}
2465 
2466 	/*
2467 	 * Is there any work to be done? No work if the address
2468 	 * is INADDR_ANY, loopback or NOLOCAL or ANYCAST (
2469 	 * ipif_select_source_v6() does not borrow addresses from
2470 	 * NOLOCAL and ANYCAST interfaces).
2471 	 */
2472 	if ((old_ipif != NULL) &&
2473 	    ((IN6_IS_ADDR_UNSPECIFIED(&old_ipif->ipif_v6lcl_addr)) ||
2474 	    (old_ipif->ipif_ill->ill_wq == NULL) ||
2475 	    (old_ipif->ipif_flags &
2476 	    (IPIF_NOLOCAL|IPIF_ANYCAST)))) {
2477 		return;
2478 	}
2479 
2480 	/*
2481 	 * Perform the same checks as when creating the
2482 	 * IRE_INTERFACE in ipif_up_done_v6.
2483 	 */
2484 	if (!(ipif->ipif_flags & IPIF_UP))
2485 		return;
2486 
2487 	if ((ipif->ipif_flags & IPIF_NOXMIT))
2488 		return;
2489 
2490 	if (IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6subnet) &&
2491 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6net_mask))
2492 		return;
2493 
2494 	/*
2495 	 * We know that ipif uses some other source for its
2496 	 * IRE_INTERFACE. Is it using the source of this
2497 	 * old_ipif?
2498 	 */
2499 	ipif_ire = ipif_to_ire_v6(ipif);
2500 	if (ipif_ire == NULL)
2501 		return;
2502 
2503 	if (old_ipif != NULL &&
2504 	    !IN6_ARE_ADDR_EQUAL(&old_ipif->ipif_v6lcl_addr,
2505 	    &ipif_ire->ire_src_addr_v6)) {
2506 		ire_refrele(ipif_ire);
2507 		return;
2508 	}
2509 
2510 	if (ip_debug > 2) {
2511 		/* ip1dbg */
2512 		pr_addr_dbg("ipif_recreate_interface_routes_v6: deleting IRE"
2513 		    " for src %s\n", AF_INET6, &ipif_ire->ire_src_addr_v6);
2514 	}
2515 
2516 	stq = ipif_ire->ire_stq;
2517 
2518 	/*
2519 	 * Can't use our source address. Select a different source address
2520 	 * for the IRE_INTERFACE.  We restrict interface route source
2521 	 * address selection to ipif's assigned to the same link as the
2522 	 * interface.
2523 	 */
2524 	if (ip6_asp_can_lookup(ipst)) {
2525 		ip6_asp_table_held = B_TRUE;
2526 		nipif = ipif_select_source_v6(ill, &ipif->ipif_v6subnet,
2527 		    RESTRICT_TO_GROUP, IPV6_PREFER_SRC_DEFAULT,
2528 		    ipif->ipif_zoneid);
2529 	}
2530 	if (nipif == NULL) {
2531 		/* Last resort - all ipif's have IPIF_NOLOCAL */
2532 		nipif = ipif;
2533 	} else {
2534 		nipif_refheld = B_TRUE;
2535 	}
2536 
2537 	ire = ire_create_v6(
2538 	    &ipif->ipif_v6subnet,	/* dest pref */
2539 	    &ipif->ipif_v6net_mask,	/* mask */
2540 	    &nipif->ipif_v6src_addr,	/* src addr */
2541 	    NULL,			/* no gateway */
2542 	    &ipif->ipif_mtu,		/* max frag */
2543 	    NULL,			/* no src nce */
2544 	    NULL,			/* no recv from queue */
2545 	    stq,			/* send-to queue */
2546 	    ill->ill_net_type,		/* IF_[NO]RESOLVER */
2547 	    ipif,
2548 	    NULL,
2549 	    0,
2550 	    0,
2551 	    0,
2552 	    &ire_uinfo_null,
2553 	    NULL,
2554 	    NULL,
2555 	    ipst);
2556 
2557 	if (ire != NULL) {
2558 		ire_t *ret_ire;
2559 		int   error;
2560 
2561 		/*
2562 		 * We don't need ipif_ire anymore. We need to delete
2563 		 * before we add so that ire_add does not detect
2564 		 * duplicates.
2565 		 */
2566 		ire_delete(ipif_ire);
2567 		ret_ire = ire;
2568 		error = ire_add(&ret_ire, NULL, NULL, NULL, B_FALSE);
2569 		ASSERT(error == 0);
2570 		ASSERT(ret_ire == ire);
2571 		if (ret_ire != NULL) {
2572 			/* Held in ire_add */
2573 			ire_refrele(ret_ire);
2574 		}
2575 	}
2576 	/*
2577 	 * Either we are falling through from above or could not
2578 	 * allocate a replacement.
2579 	 */
2580 	ire_refrele(ipif_ire);
2581 	if (ip6_asp_table_held)
2582 		ip6_asp_table_refrele(ipst);
2583 	if (nipif_refheld)
2584 		ipif_refrele(nipif);
2585 }
2586 
2587 /*
2588  * This old_ipif is going away.
2589  *
2590  * Determine if any other ipif's are using our address as
2591  * ipif_v6lcl_addr (due to those being IPIF_NOLOCAL, IPIF_ANYCAST, or
2592  * IPIF_DEPRECATED).
2593  * Find the IRE_INTERFACE for such ipif's and recreate them
2594  * to use an different source address following the rules in
2595  * ipif_up_done_v6.
2596  *
2597  * This function takes an illgrp as an argument so that illgrp_delete
2598  * can call this to update source address even after deleting the
2599  * old_ipif->ipif_ill from the ill group.
2600  */
2601 void
2602 ipif_update_other_ipifs_v6(ipif_t *old_ipif, ill_group_t *illgrp)
2603 {
2604 	ipif_t	*ipif;
2605 	ill_t	*ill;
2606 	char	buf[INET6_ADDRSTRLEN];
2607 
2608 	ASSERT(IAM_WRITER_IPIF(old_ipif));
2609 
2610 	ill = old_ipif->ipif_ill;
2611 
2612 	ip1dbg(("ipif_update_other_ipifs_v6(%s, %s)\n",
2613 	    ill->ill_name,
2614 	    inet_ntop(AF_INET6, &old_ipif->ipif_v6lcl_addr,
2615 	    buf, sizeof (buf))));
2616 
2617 	/*
2618 	 * If this part of a group, look at all ills as ipif_select_source
2619 	 * borrows a source address across all the ills in the group.
2620 	 */
2621 	if (illgrp != NULL)
2622 		ill = illgrp->illgrp_ill;
2623 
2624 	/* Don't need a lock since this is a writer */
2625 	for (; ill != NULL; ill = ill->ill_group_next) {
2626 		for (ipif = ill->ill_ipif; ipif != NULL;
2627 		    ipif = ipif->ipif_next) {
2628 
2629 			if (ipif == old_ipif)
2630 				continue;
2631 
2632 			ipif_recreate_interface_routes_v6(old_ipif, ipif);
2633 		}
2634 	}
2635 }
2636 
2637 /*
2638  * Perform an attach and bind to get phys addr plus info_req for
2639  * the physical device.
2640  * q and mp represents an ioctl which will be queued waiting for
2641  * completion of the DLPI message exchange.
2642  * MUST be called on an ill queue. Can not set conn_pending_ill for that
2643  * reason thus the DL_PHYS_ADDR_ACK code does not assume ill_pending_q.
2644  *
2645  * Returns EINPROGRESS when mp has been consumed by queueing it on
2646  * ill_pending_mp and the ioctl will complete in ip_rput.
2647  */
2648 int
2649 ill_dl_phys(ill_t *ill, ipif_t *ipif, mblk_t *mp, queue_t *q)
2650 {
2651 	mblk_t	*v6token_mp = NULL;
2652 	mblk_t	*v6lla_mp = NULL;
2653 	mblk_t	*phys_mp = NULL;
2654 	mblk_t	*info_mp = NULL;
2655 	mblk_t	*attach_mp = NULL;
2656 	mblk_t	*bind_mp = NULL;
2657 	mblk_t	*unbind_mp = NULL;
2658 	mblk_t	*notify_mp = NULL;
2659 
2660 	ip1dbg(("ill_dl_phys(%s:%u)\n", ill->ill_name, ipif->ipif_id));
2661 	ASSERT(ill->ill_dlpi_style_set);
2662 	ASSERT(WR(q)->q_next != NULL);
2663 
2664 	if (ill->ill_isv6) {
2665 		v6token_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2666 		    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2667 		if (v6token_mp == NULL)
2668 			goto bad;
2669 		((dl_phys_addr_req_t *)v6token_mp->b_rptr)->dl_addr_type =
2670 		    DL_IPV6_TOKEN;
2671 
2672 		v6lla_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2673 		    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2674 		if (v6lla_mp == NULL)
2675 			goto bad;
2676 		((dl_phys_addr_req_t *)v6lla_mp->b_rptr)->dl_addr_type =
2677 		    DL_IPV6_LINK_LAYER_ADDR;
2678 	}
2679 
2680 	/*
2681 	 * Allocate a DL_NOTIFY_REQ and set the notifications we want.
2682 	 */
2683 	notify_mp = ip_dlpi_alloc(sizeof (dl_notify_req_t) + sizeof (long),
2684 	    DL_NOTIFY_REQ);
2685 	if (notify_mp == NULL)
2686 		goto bad;
2687 	((dl_notify_req_t *)notify_mp->b_rptr)->dl_notifications =
2688 	    (DL_NOTE_PHYS_ADDR | DL_NOTE_SDU_SIZE | DL_NOTE_FASTPATH_FLUSH |
2689 	    DL_NOTE_LINK_UP | DL_NOTE_LINK_DOWN | DL_NOTE_CAPAB_RENEG);
2690 
2691 	phys_mp = ip_dlpi_alloc(sizeof (dl_phys_addr_req_t) +
2692 	    sizeof (t_scalar_t), DL_PHYS_ADDR_REQ);
2693 	if (phys_mp == NULL)
2694 		goto bad;
2695 	((dl_phys_addr_req_t *)phys_mp->b_rptr)->dl_addr_type =
2696 	    DL_CURR_PHYS_ADDR;
2697 
2698 	info_mp = ip_dlpi_alloc(
2699 	    sizeof (dl_info_req_t) + sizeof (dl_info_ack_t),
2700 	    DL_INFO_REQ);
2701 	if (info_mp == NULL)
2702 		goto bad;
2703 
2704 	bind_mp = ip_dlpi_alloc(sizeof (dl_bind_req_t) + sizeof (long),
2705 	    DL_BIND_REQ);
2706 	if (bind_mp == NULL)
2707 		goto bad;
2708 	((dl_bind_req_t *)bind_mp->b_rptr)->dl_sap = ill->ill_sap;
2709 	((dl_bind_req_t *)bind_mp->b_rptr)->dl_service_mode = DL_CLDLS;
2710 
2711 	unbind_mp = ip_dlpi_alloc(sizeof (dl_unbind_req_t), DL_UNBIND_REQ);
2712 	if (unbind_mp == NULL)
2713 		goto bad;
2714 
2715 	/* If we need to attach, pre-alloc and initialize the mblk */
2716 	if (ill->ill_needs_attach) {
2717 		attach_mp = ip_dlpi_alloc(sizeof (dl_attach_req_t),
2718 		    DL_ATTACH_REQ);
2719 		if (attach_mp == NULL)
2720 			goto bad;
2721 		((dl_attach_req_t *)attach_mp->b_rptr)->dl_ppa = ill->ill_ppa;
2722 	}
2723 
2724 	/*
2725 	 * Here we are going to delay the ioctl ack until after
2726 	 * ACKs from DL_PHYS_ADDR_REQ. So need to save the
2727 	 * original ioctl message before sending the requests
2728 	 */
2729 	mutex_enter(&ill->ill_lock);
2730 	/* ipsq_pending_mp_add won't fail since we pass in a NULL connp */
2731 	(void) ipsq_pending_mp_add(NULL, ipif, ill->ill_wq, mp, 0);
2732 	/*
2733 	 * Set ill_phys_addr_pend to zero. It will be set to the addr_type of
2734 	 * the DL_PHYS_ADDR_REQ in ill_dlpi_send() and ill_dlpi_done(). It will
2735 	 * be used to track which DL_PHYS_ADDR_REQ is being ACK'd/NAK'd.
2736 	 */
2737 	ill->ill_phys_addr_pend = 0;
2738 	mutex_exit(&ill->ill_lock);
2739 
2740 	if (attach_mp != NULL) {
2741 		ip1dbg(("ill_dl_phys: attach\n"));
2742 		ill_dlpi_send(ill, attach_mp);
2743 	}
2744 	ill_dlpi_send(ill, bind_mp);
2745 	ill_dlpi_send(ill, info_mp);
2746 	if (ill->ill_isv6) {
2747 		ill_dlpi_send(ill, v6token_mp);
2748 		ill_dlpi_send(ill, v6lla_mp);
2749 	}
2750 	ill_dlpi_send(ill, phys_mp);
2751 	ill_dlpi_send(ill, notify_mp);
2752 	ill_dlpi_send(ill, unbind_mp);
2753 
2754 	/*
2755 	 * This operation will complete in ip_rput_dlpi_writer with either
2756 	 * a DL_PHYS_ADDR_ACK or DL_ERROR_ACK.
2757 	 */
2758 	return (EINPROGRESS);
2759 bad:
2760 	freemsg(v6token_mp);
2761 	freemsg(v6lla_mp);
2762 	freemsg(phys_mp);
2763 	freemsg(info_mp);
2764 	freemsg(attach_mp);
2765 	freemsg(bind_mp);
2766 	freemsg(unbind_mp);
2767 	freemsg(notify_mp);
2768 	return (ENOMEM);
2769 }
2770 
2771 uint_t ip_loopback_mtu_v6plus = IP_LOOPBACK_MTU + IPV6_HDR_LEN + 20;
2772 
2773 /*
2774  * DLPI is up.
2775  * Create all the IREs associated with an interface bring up multicast.
2776  * Set the interface flag and finish other initialization
2777  * that potentially had to be differed to after DL_BIND_ACK.
2778  */
2779 int
2780 ipif_up_done_v6(ipif_t *ipif)
2781 {
2782 	ire_t	*ire_array[20];
2783 	ire_t	**irep = ire_array;
2784 	ire_t	**irep1;
2785 	ill_t	*ill = ipif->ipif_ill;
2786 	queue_t	*stq;
2787 	in6_addr_t	v6addr;
2788 	in6_addr_t	route_mask;
2789 	ipif_t	 *src_ipif = NULL;
2790 	ipif_t   *tmp_ipif;
2791 	boolean_t	flush_ire_cache = B_TRUE;
2792 	int	err;
2793 	char	buf[INET6_ADDRSTRLEN];
2794 	phyint_t *phyi;
2795 	ire_t	**ipif_saved_irep = NULL;
2796 	int ipif_saved_ire_cnt;
2797 	int cnt;
2798 	boolean_t src_ipif_held = B_FALSE;
2799 	boolean_t ire_added = B_FALSE;
2800 	boolean_t loopback = B_FALSE;
2801 	boolean_t ip6_asp_table_held = B_FALSE;
2802 	ip_stack_t	*ipst = ill->ill_ipst;
2803 
2804 	ip1dbg(("ipif_up_done_v6(%s:%u)\n",
2805 	    ipif->ipif_ill->ill_name, ipif->ipif_id));
2806 
2807 	/* Check if this is a loopback interface */
2808 	if (ipif->ipif_ill->ill_wq == NULL)
2809 		loopback = B_TRUE;
2810 
2811 	ASSERT(ipif->ipif_isv6);
2812 	ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
2813 
2814 	/*
2815 	 * If all other interfaces for this ill are down or DEPRECATED,
2816 	 * or otherwise unsuitable for source address selection, remove
2817 	 * any IRE_CACHE entries for this ill to make sure source
2818 	 * address selection gets to take this new ipif into account.
2819 	 * No need to hold ill_lock while traversing the ipif list since
2820 	 * we are writer
2821 	 */
2822 	for (tmp_ipif = ill->ill_ipif; tmp_ipif;
2823 	    tmp_ipif = tmp_ipif->ipif_next) {
2824 		if (((tmp_ipif->ipif_flags &
2825 		    (IPIF_NOXMIT|IPIF_ANYCAST|IPIF_NOLOCAL|IPIF_DEPRECATED)) ||
2826 		    !(tmp_ipif->ipif_flags & IPIF_UP)) ||
2827 		    (tmp_ipif == ipif))
2828 			continue;
2829 		/* first useable pre-existing interface */
2830 		flush_ire_cache = B_FALSE;
2831 		break;
2832 	}
2833 	if (flush_ire_cache)
2834 		ire_walk_ill_v6(MATCH_IRE_ILL_GROUP | MATCH_IRE_TYPE,
2835 		    IRE_CACHE, ill_ipif_cache_delete, (char *)ill, ill);
2836 
2837 	/*
2838 	 * Figure out which way the send-to queue should go.  Only
2839 	 * IRE_IF_RESOLVER or IRE_IF_NORESOLVER should show up here.
2840 	 */
2841 	switch (ill->ill_net_type) {
2842 	case IRE_IF_RESOLVER:
2843 		stq = ill->ill_rq;
2844 		break;
2845 	case IRE_IF_NORESOLVER:
2846 	case IRE_LOOPBACK:
2847 		stq = ill->ill_wq;
2848 		break;
2849 	default:
2850 		return (EINVAL);
2851 	}
2852 
2853 	if (IS_LOOPBACK(ill)) {
2854 		/*
2855 		 * lo0:1 and subsequent ipifs were marked IRE_LOCAL in
2856 		 * ipif_lookup_on_name(), but in the case of zones we can have
2857 		 * several loopback addresses on lo0. So all the interfaces with
2858 		 * loopback addresses need to be marked IRE_LOOPBACK.
2859 		 */
2860 		if (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr, &ipv6_loopback))
2861 			ipif->ipif_ire_type = IRE_LOOPBACK;
2862 		else
2863 			ipif->ipif_ire_type = IRE_LOCAL;
2864 	}
2865 
2866 	if (ipif->ipif_flags & (IPIF_NOLOCAL|IPIF_ANYCAST|IPIF_DEPRECATED)) {
2867 		/*
2868 		 * Can't use our source address. Select a different
2869 		 * source address for the IRE_INTERFACE and IRE_LOCAL
2870 		 */
2871 		if (ip6_asp_can_lookup(ipst)) {
2872 			ip6_asp_table_held = B_TRUE;
2873 			src_ipif = ipif_select_source_v6(ipif->ipif_ill,
2874 			    &ipif->ipif_v6subnet, RESTRICT_TO_NONE,
2875 			    IPV6_PREFER_SRC_DEFAULT, ipif->ipif_zoneid);
2876 		}
2877 		if (src_ipif == NULL)
2878 			src_ipif = ipif;	/* Last resort */
2879 		else
2880 			src_ipif_held = B_TRUE;
2881 	} else {
2882 		src_ipif = ipif;
2883 	}
2884 
2885 	if (!IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6lcl_addr) &&
2886 	    !(ipif->ipif_flags & IPIF_NOLOCAL)) {
2887 
2888 		/*
2889 		 * If we're on a labeled system then make sure that zone-
2890 		 * private addresses have proper remote host database entries.
2891 		 */
2892 		if (is_system_labeled() &&
2893 		    ipif->ipif_ire_type != IRE_LOOPBACK) {
2894 			if (ip6opt_ls == 0) {
2895 				cmn_err(CE_WARN, "IPv6 not enabled "
2896 				    "via /etc/system");
2897 				return (EINVAL);
2898 			}
2899 			if (!tsol_check_interface_address(ipif))
2900 				return (EINVAL);
2901 		}
2902 
2903 		/* Register the source address for __sin6_src_id */
2904 		err = ip_srcid_insert(&ipif->ipif_v6lcl_addr,
2905 		    ipif->ipif_zoneid, ipst);
2906 		if (err != 0) {
2907 			ip0dbg(("ipif_up_done_v6: srcid_insert %d\n", err));
2908 			if (src_ipif_held)
2909 				ipif_refrele(src_ipif);
2910 			if (ip6_asp_table_held)
2911 				ip6_asp_table_refrele(ipst);
2912 			return (err);
2913 		}
2914 		/*
2915 		 * If the interface address is set, create the LOCAL
2916 		 * or LOOPBACK IRE.
2917 		 */
2918 		ip1dbg(("ipif_up_done_v6: creating IRE %d for %s\n",
2919 		    ipif->ipif_ire_type,
2920 		    inet_ntop(AF_INET6, &ipif->ipif_v6lcl_addr,
2921 		    buf, sizeof (buf))));
2922 
2923 		*irep++ = ire_create_v6(
2924 		    &ipif->ipif_v6lcl_addr,		/* dest address */
2925 		    &ipv6_all_ones,			/* mask */
2926 		    &src_ipif->ipif_v6src_addr,		/* source address */
2927 		    NULL,				/* no gateway */
2928 		    &ip_loopback_mtu_v6plus,		/* max frag size */
2929 		    NULL,
2930 		    ipif->ipif_rq,			/* recv-from queue */
2931 		    NULL,				/* no send-to queue */
2932 		    ipif->ipif_ire_type,		/* LOCAL or LOOPBACK */
2933 		    ipif,				/* interface */
2934 		    NULL,
2935 		    0,
2936 		    0,
2937 		    (ipif->ipif_flags & IPIF_PRIVATE) ? RTF_PRIVATE : 0,
2938 		    &ire_uinfo_null,
2939 		    NULL,
2940 		    NULL,
2941 		    ipst);
2942 	}
2943 
2944 	/*
2945 	 * Set up the IRE_IF_RESOLVER or IRE_IF_NORESOLVER, as appropriate.
2946 	 * Note that atun interfaces have an all-zero ipif_v6subnet.
2947 	 * Thus we allow a zero subnet as long as the mask is non-zero.
2948 	 */
2949 	if (stq != NULL && !(ipif->ipif_flags & IPIF_NOXMIT) &&
2950 	    !(IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6subnet) &&
2951 	    IN6_IS_ADDR_UNSPECIFIED(&ipif->ipif_v6net_mask))) {
2952 		/* ipif_v6subnet is ipif_v6pp_dst_addr for pt-pt */
2953 		v6addr = ipif->ipif_v6subnet;
2954 
2955 		if (ipif->ipif_flags & IPIF_POINTOPOINT) {
2956 			route_mask = ipv6_all_ones;
2957 		} else {
2958 			route_mask = ipif->ipif_v6net_mask;
2959 		}
2960 
2961 		ip1dbg(("ipif_up_done_v6: creating if IRE %d for %s\n",
2962 		    ill->ill_net_type,
2963 		    inet_ntop(AF_INET6, &v6addr, buf, sizeof (buf))));
2964 
2965 		*irep++ = ire_create_v6(
2966 		    &v6addr,			/* dest pref */
2967 		    &route_mask,		/* mask */
2968 		    &src_ipif->ipif_v6src_addr,	/* src addr */
2969 		    NULL,			/* no gateway */
2970 		    &ipif->ipif_mtu,		/* max frag */
2971 		    NULL,			/* no src nce */
2972 		    NULL,			/* no recv from queue */
2973 		    stq,			/* send-to queue */
2974 		    ill->ill_net_type,		/* IF_[NO]RESOLVER */
2975 		    ipif,
2976 		    NULL,
2977 		    0,
2978 		    0,
2979 		    (ipif->ipif_flags & IPIF_PRIVATE) ? RTF_PRIVATE : 0,
2980 		    &ire_uinfo_null,
2981 		    NULL,
2982 		    NULL,
2983 		    ipst);
2984 	}
2985 
2986 	/*
2987 	 * Setup 2002::/16 route, if this interface is a 6to4 tunnel
2988 	 */
2989 	if (IN6_IS_ADDR_6TO4(&ipif->ipif_v6lcl_addr) &&
2990 	    (ill->ill_is_6to4tun)) {
2991 		/*
2992 		 * Destination address is 2002::/16
2993 		 */
2994 #ifdef	_BIG_ENDIAN
2995 		const in6_addr_t prefix_addr = { 0x20020000U, 0, 0, 0 };
2996 		const in6_addr_t prefix_mask = { 0xffff0000U, 0, 0, 0 };
2997 #else
2998 		const in6_addr_t prefix_addr = { 0x00000220U, 0, 0, 0 };
2999 		const in6_addr_t prefix_mask = { 0x0000ffffU, 0, 0, 0 };
3000 #endif /* _BIG_ENDIAN */
3001 		char	buf2[INET6_ADDRSTRLEN];
3002 		ire_t *isdup;
3003 		in6_addr_t *first_addr = &ill->ill_ipif->ipif_v6lcl_addr;
3004 
3005 		/*
3006 		 * check to see if this route has already been added for
3007 		 * this tunnel interface.
3008 		 */
3009 		isdup = ire_ftable_lookup_v6(first_addr, &prefix_mask, 0,
3010 		    IRE_IF_NORESOLVER, ill->ill_ipif, NULL, ALL_ZONES, 0, NULL,
3011 		    (MATCH_IRE_SRC | MATCH_IRE_MASK), ipst);
3012 
3013 		if (isdup == NULL) {
3014 			ip1dbg(("ipif_up_done_v6: creating if IRE %d for %s",
3015 			    IRE_IF_NORESOLVER, inet_ntop(AF_INET6, &v6addr,
3016 			    buf2, sizeof (buf2))));
3017 
3018 			*irep++ = ire_create_v6(
3019 			    &prefix_addr,		/* 2002:: */
3020 			    &prefix_mask,		/* ffff:: */
3021 			    &ipif->ipif_v6lcl_addr, 	/* src addr */
3022 			    NULL, 			/* gateway */
3023 			    &ipif->ipif_mtu, 		/* max_frag */
3024 			    NULL, 			/* no src nce */
3025 			    NULL, 			/* no rfq */
3026 			    ill->ill_wq, 		/* stq */
3027 			    IRE_IF_NORESOLVER,		/* type */
3028 			    ipif,			/* interface */
3029 			    NULL,			/* v6cmask */
3030 			    0,
3031 			    0,
3032 			    RTF_UP,
3033 			    &ire_uinfo_null,
3034 			    NULL,
3035 			    NULL,
3036 			    ipst);
3037 		} else {
3038 			ire_refrele(isdup);
3039 		}
3040 	}
3041 
3042 	/* If an earlier ire_create failed, get out now */
3043 	for (irep1 = irep; irep1 > ire_array; ) {
3044 		irep1--;
3045 		if (*irep1 == NULL) {
3046 			ip1dbg(("ipif_up_done_v6: NULL ire found in"
3047 			    " ire_array\n"));
3048 			err = ENOMEM;
3049 			goto bad;
3050 		}
3051 	}
3052 
3053 	ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
3054 
3055 	/*
3056 	 * Need to atomically check for ip_addr_availablity_check
3057 	 * now under ill_g_lock, and if it fails got bad, and remove
3058 	 * from group also
3059 	 */
3060 	rw_enter(&ipst->ips_ill_g_lock, RW_READER);
3061 	mutex_enter(&ipst->ips_ip_addr_avail_lock);
3062 	ill->ill_ipif_up_count++;
3063 	ipif->ipif_flags |= IPIF_UP;
3064 	err = ip_addr_availability_check(ipif);
3065 	mutex_exit(&ipst->ips_ip_addr_avail_lock);
3066 	rw_exit(&ipst->ips_ill_g_lock);
3067 
3068 	if (err != 0) {
3069 		/*
3070 		 * Our address may already be up on the same ill. In this case,
3071 		 * the external resolver entry for our ipif replaced the one for
3072 		 * the other ipif. So we don't want to delete it (otherwise the
3073 		 * other ipif would be unable to send packets).
3074 		 * ip_addr_availability_check() identifies this case for us and
3075 		 * returns EADDRINUSE; we need to turn it into EADDRNOTAVAIL
3076 		 * which is the expected error code.
3077 		 */
3078 		if (err == EADDRINUSE) {
3079 			if (ipif->ipif_ill->ill_flags & ILLF_XRESOLV) {
3080 				freemsg(ipif->ipif_arp_del_mp);
3081 				ipif->ipif_arp_del_mp = NULL;
3082 			}
3083 			err = EADDRNOTAVAIL;
3084 		}
3085 		ill->ill_ipif_up_count--;
3086 		ipif->ipif_flags &= ~IPIF_UP;
3087 		goto bad;
3088 	}
3089 
3090 	/*
3091 	 * Add in all newly created IREs. We want to add before
3092 	 * we call ifgrp_insert which wants to know whether
3093 	 * IRE_IF_RESOLVER exists or not.
3094 	 *
3095 	 * NOTE : We refrele the ire though we may branch to "bad"
3096 	 *	  later on where we do ire_delete. This is okay
3097 	 *	  because nobody can delete it as we are running
3098 	 *	  exclusively.
3099 	 */
3100 	for (irep1 = irep; irep1 > ire_array; ) {
3101 		irep1--;
3102 		/* Shouldn't be adding any bcast ire's */
3103 		ASSERT((*irep1)->ire_type != IRE_BROADCAST);
3104 		ASSERT(!MUTEX_HELD(&ipif->ipif_ill->ill_lock));
3105 		/*
3106 		 * refheld by ire_add. refele towards the end of the func
3107 		 */
3108 		(void) ire_add(irep1, NULL, NULL, NULL, B_FALSE);
3109 	}
3110 	if (ip6_asp_table_held) {
3111 		ip6_asp_table_refrele(ipst);
3112 		ip6_asp_table_held = B_FALSE;
3113 	}
3114 	ire_added = B_TRUE;
3115 
3116 	/*
3117 	 * Form groups if possible.
3118 	 *
3119 	 * If we are supposed to be in a ill_group with a name, insert it
3120 	 * now as we know that at least one ipif is UP. Otherwise form
3121 	 * nameless groups.
3122 	 *
3123 	 * If ip_enable_group_ifs is set and ipif address is not ::0, insert
3124 	 * this ipif into the appropriate interface group, or create a
3125 	 * new one. If this is already in a nameless group, we try to form
3126 	 * a bigger group looking at other ills potentially sharing this
3127 	 * ipif's prefix.
3128 	 */
3129 	phyi = ill->ill_phyint;
3130 	if (phyi->phyint_groupname_len != 0) {
3131 		ASSERT(phyi->phyint_groupname != NULL);
3132 		if (ill->ill_ipif_up_count == 1) {
3133 			ASSERT(ill->ill_group == NULL);
3134 			err = illgrp_insert(&ipst->ips_illgrp_head_v6, ill,
3135 			    phyi->phyint_groupname, NULL, B_TRUE);
3136 			if (err != 0) {
3137 				ip1dbg(("ipif_up_done_v6: illgrp allocation "
3138 				    "failed, error %d\n", err));
3139 				goto bad;
3140 			}
3141 		}
3142 		ASSERT(ill->ill_group != NULL);
3143 	}
3144 
3145 	/* Recover any additional IRE_IF_[NO]RESOLVER entries for this ipif */
3146 	ipif_saved_ire_cnt = ipif->ipif_saved_ire_cnt;
3147 	ipif_saved_irep = ipif_recover_ire_v6(ipif);
3148 
3149 	if (ill->ill_need_recover_multicast) {
3150 		/*
3151 		 * Need to recover all multicast memberships in the driver.
3152 		 * This had to be deferred until we had attached.
3153 		 */
3154 		ill_recover_multicast(ill);
3155 	}
3156 	/* Join the allhosts multicast address and the solicited node MC */
3157 	ipif_multicast_up(ipif);
3158 
3159 	if (!loopback) {
3160 		/*
3161 		 * See whether anybody else would benefit from the
3162 		 * new ipif that we added. We call this always rather
3163 		 * than while adding a non-IPIF_NOLOCAL/DEPRECATED/ANYCAST
3164 		 * ipif for the benefit of illgrp_insert (done above)
3165 		 * which does not do source address selection as it does
3166 		 * not want to re-create interface routes that we are
3167 		 * having reference to it here.
3168 		 */
3169 		ill_update_source_selection(ill);
3170 	}
3171 
3172 	for (irep1 = irep; irep1 > ire_array; ) {
3173 		irep1--;
3174 		if (*irep1 != NULL) {
3175 			/* was held in ire_add */
3176 			ire_refrele(*irep1);
3177 		}
3178 	}
3179 
3180 	cnt = ipif_saved_ire_cnt;
3181 	for (irep1 = ipif_saved_irep; cnt > 0; irep1++, cnt--) {
3182 		if (*irep1 != NULL) {
3183 			/* was held in ire_add */
3184 			ire_refrele(*irep1);
3185 		}
3186 	}
3187 
3188 	if (ipif->ipif_addr_ready)
3189 		ipif_up_notify(ipif);
3190 
3191 	if (ipif_saved_irep != NULL) {
3192 		kmem_free(ipif_saved_irep,
3193 		    ipif_saved_ire_cnt * sizeof (ire_t *));
3194 	}
3195 
3196 	if (src_ipif_held)
3197 		ipif_refrele(src_ipif);
3198 
3199 	return (0);
3200 
3201 bad:
3202 	if (ip6_asp_table_held)
3203 		ip6_asp_table_refrele(ipst);
3204 	/*
3205 	 * We don't have to bother removing from ill groups because
3206 	 *
3207 	 * 1) For groups with names, we insert only when the first ipif
3208 	 *    comes up. In that case if it fails, it will not be in any
3209 	 *    group. So, we need not try to remove for that case.
3210 	 *
3211 	 * 2) For groups without names, either we tried to insert ipif_ill
3212 	 *    in a group as singleton or found some other group to become
3213 	 *    a bigger group. For the former, if it fails we don't have
3214 	 *    anything to do as ipif_ill is not in the group and for the
3215 	 *    latter, there are no failures in illgrp_insert/illgrp_delete
3216 	 *    (ENOMEM can't occur for this. Check ifgrp_insert).
3217 	 */
3218 
3219 	while (irep > ire_array) {
3220 		irep--;
3221 		if (*irep != NULL) {
3222 			ire_delete(*irep);
3223 			if (ire_added)
3224 				ire_refrele(*irep);
3225 		}
3226 
3227 	}
3228 	(void) ip_srcid_remove(&ipif->ipif_v6lcl_addr, ipif->ipif_zoneid, ipst);
3229 
3230 	if (ipif_saved_irep != NULL) {
3231 		kmem_free(ipif_saved_irep,
3232 		    ipif_saved_ire_cnt * sizeof (ire_t *));
3233 	}
3234 	if (src_ipif_held)
3235 		ipif_refrele(src_ipif);
3236 
3237 	ipif_ndp_down(ipif);
3238 	if (ipif->ipif_ill->ill_flags & ILLF_XRESOLV)
3239 		ipif_arp_down(ipif);
3240 
3241 	return (err);
3242 }
3243 
3244 /*
3245  * Delete an ND entry and the corresponding IRE_CACHE entry if it exists.
3246  */
3247 /* ARGSUSED */
3248 int
3249 ip_siocdelndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3250     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3251 {
3252 	in6_addr_t	addr;
3253 	sin6_t		*sin6;
3254 	nce_t		*nce;
3255 	struct lifreq	*lifr;
3256 	lif_nd_req_t	*lnr;
3257 	mblk_t	*mp1;
3258 
3259 	mp1 = mp->b_cont->b_cont;
3260 	lifr = (struct lifreq *)mp1->b_rptr;
3261 	lnr = &lifr->lifr_nd;
3262 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3263 	if (ipif->ipif_id != 0)
3264 		return (EINVAL);
3265 
3266 	if (!ipif->ipif_isv6)
3267 		return (EINVAL);
3268 
3269 	if (lnr->lnr_addr.ss_family != AF_INET6)
3270 		return (EAFNOSUPPORT);
3271 
3272 	sin6 = (sin6_t *)&lnr->lnr_addr;
3273 	addr = sin6->sin6_addr;
3274 	nce = ndp_lookup_v6(ipif->ipif_ill, &addr, B_FALSE);
3275 	if (nce == NULL)
3276 		return (ESRCH);
3277 	ndp_delete(nce);
3278 	NCE_REFRELE(nce);
3279 	return (0);
3280 }
3281 
3282 /*
3283  * Return nbr cache info.
3284  */
3285 /* ARGSUSED */
3286 int
3287 ip_siocqueryndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3288     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3289 {
3290 	ill_t		*ill = ipif->ipif_ill;
3291 	struct lifreq	*lifr;
3292 	lif_nd_req_t	*lnr;
3293 
3294 	lifr = (struct lifreq *)mp->b_cont->b_cont->b_rptr;
3295 	lnr = &lifr->lifr_nd;
3296 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3297 	if (ipif->ipif_id != 0)
3298 		return (EINVAL);
3299 
3300 	if (!ipif->ipif_isv6)
3301 		return (EINVAL);
3302 
3303 	if (lnr->lnr_addr.ss_family != AF_INET6)
3304 		return (EAFNOSUPPORT);
3305 
3306 	if (ill->ill_phys_addr_length > sizeof (lnr->lnr_hdw_addr))
3307 		return (EINVAL);
3308 
3309 	return (ndp_query(ill, lnr));
3310 }
3311 
3312 /*
3313  * Perform an update of the nd entry for the specified address.
3314  */
3315 /* ARGSUSED */
3316 int
3317 ip_siocsetndp_v6(ipif_t *ipif, sin_t *dummy_sin, queue_t *q, mblk_t *mp,
3318     ip_ioctl_cmd_t *ipip, void *dummy_ifreq)
3319 {
3320 	ill_t		*ill = ipif->ipif_ill;
3321 	struct	lifreq	*lifr;
3322 	lif_nd_req_t	*lnr;
3323 
3324 	ASSERT(!(q->q_flag & QREADR) && q->q_next == NULL);
3325 
3326 	lifr = (struct lifreq *)mp->b_cont->b_cont->b_rptr;
3327 	lnr = &lifr->lifr_nd;
3328 	/* Only allow for logical unit zero i.e. not on "le0:17" */
3329 	if (ipif->ipif_id != 0)
3330 		return (EINVAL);
3331 
3332 	if (!ipif->ipif_isv6)
3333 		return (EINVAL);
3334 
3335 	if (lnr->lnr_addr.ss_family != AF_INET6)
3336 		return (EAFNOSUPPORT);
3337 
3338 	return (ndp_sioc_update(ill, lnr));
3339 }
3340