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 /*
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 * Copyright 2017 OmniTI Computer Consulting, Inc. All rights reserved.
26 * Copyright 2018 Joyent, Inc.
27 */
28 /* Copyright (c) 1990 Mentat Inc. */
29
30 #include <sys/types.h>
31 #include <sys/stream.h>
32 #include <sys/strsubr.h>
33 #include <sys/dlpi.h>
34 #include <sys/strsun.h>
35 #include <sys/zone.h>
36 #include <sys/ddi.h>
37 #include <sys/sunddi.h>
38 #include <sys/cmn_err.h>
39 #include <sys/debug.h>
40 #include <sys/atomic.h>
41
42 #include <sys/systm.h>
43 #include <sys/param.h>
44 #include <sys/kmem.h>
45 #include <sys/sdt.h>
46 #include <sys/socket.h>
47 #include <sys/mac.h>
48 #include <net/if.h>
49 #include <net/if_arp.h>
50 #include <net/route.h>
51 #include <sys/sockio.h>
52 #include <netinet/in.h>
53 #include <net/if_dl.h>
54
55 #include <inet/common.h>
56 #include <inet/mi.h>
57 #include <inet/mib2.h>
58 #include <inet/nd.h>
59 #include <inet/arp.h>
60 #include <inet/snmpcom.h>
61 #include <inet/kstatcom.h>
62
63 #include <netinet/igmp_var.h>
64 #include <netinet/ip6.h>
65 #include <netinet/icmp6.h>
66 #include <netinet/sctp.h>
67
68 #include <inet/ip.h>
69 #include <inet/ip_impl.h>
70 #include <inet/ip6.h>
71 #include <inet/ip6_asp.h>
72 #include <inet/tcp.h>
73 #include <inet/ip_multi.h>
74 #include <inet/ip_if.h>
75 #include <inet/ip_ire.h>
76 #include <inet/ip_ftable.h>
77 #include <inet/ip_rts.h>
78 #include <inet/optcom.h>
79 #include <inet/ip_ndp.h>
80 #include <inet/ip_listutils.h>
81 #include <netinet/igmp.h>
82 #include <netinet/ip_mroute.h>
83 #include <inet/ipp_common.h>
84
85 #include <net/pfkeyv2.h>
86 #include <inet/sadb.h>
87 #include <inet/ipsec_impl.h>
88 #include <inet/ipdrop.h>
89 #include <inet/ip_netinfo.h>
90
91 #include <sys/pattr.h>
92 #include <inet/ipclassifier.h>
93 #include <inet/sctp_ip.h>
94 #include <inet/sctp/sctp_impl.h>
95 #include <inet/udp_impl.h>
96 #include <sys/sunddi.h>
97
98 #include <sys/tsol/label.h>
99 #include <sys/tsol/tnet.h>
100
101 #ifdef DEBUG
102 extern boolean_t skip_sctp_cksum;
103 #endif
104
105 int
ip_output_simple_v6(mblk_t * mp,ip_xmit_attr_t * ixa)106 ip_output_simple_v6(mblk_t *mp, ip_xmit_attr_t *ixa)
107 {
108 ip6_t *ip6h;
109 in6_addr_t firsthop; /* In IP header */
110 in6_addr_t dst; /* End of source route, or ip6_dst if none */
111 ire_t *ire;
112 in6_addr_t setsrc;
113 int error;
114 ill_t *ill = NULL;
115 dce_t *dce = NULL;
116 nce_t *nce;
117 iaflags_t ixaflags = ixa->ixa_flags;
118 ip_stack_t *ipst = ixa->ixa_ipst;
119 uint8_t *nexthdrp;
120 boolean_t repeat = B_FALSE;
121 boolean_t multirt = B_FALSE;
122 uint_t ifindex;
123 int64_t now;
124
125 ip6h = (ip6_t *)mp->b_rptr;
126 ASSERT(IPH_HDR_VERSION(ip6h) == IPV6_VERSION);
127
128 ASSERT(ixa->ixa_nce == NULL);
129
130 ixa->ixa_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
131 ASSERT(ixa->ixa_pktlen == msgdsize(mp));
132 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ixa->ixa_ip_hdr_length,
133 &nexthdrp)) {
134 /* Malformed packet */
135 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
136 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
137 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
138 freemsg(mp);
139 return (EINVAL);
140 }
141 ixa->ixa_protocol = *nexthdrp;
142
143 /*
144 * Assumes that source routed packets have already been massaged by
145 * the ULP (ip_massage_options_v6) and as a result ip6_dst is the next
146 * hop in the source route. The final destination is used for IPsec
147 * policy and DCE lookup.
148 */
149 firsthop = ip6h->ip6_dst;
150 dst = ip_get_dst_v6(ip6h, mp, NULL);
151
152 repeat_ire:
153 error = 0;
154 setsrc = ipv6_all_zeros;
155 ire = ip_select_route_v6(&firsthop, ip6h->ip6_src, ixa, NULL, &setsrc,
156 &error, &multirt);
157 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
158 if (error != 0) {
159 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
160 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
161 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
162 freemsg(mp);
163 goto done;
164 }
165
166 if (ire->ire_flags & (RTF_BLACKHOLE|RTF_REJECT)) {
167 /* ire_ill might be NULL hence need to skip some code */
168 if (ixaflags & IXAF_SET_SOURCE)
169 ip6h->ip6_src = ipv6_loopback;
170 ixa->ixa_fragsize = IP_MAXPACKET;
171 ire->ire_ob_pkt_count++;
172 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
173 /* No dce yet; use default one */
174 error = (ire->ire_sendfn)(ire, mp, ip6h, ixa,
175 &ipst->ips_dce_default->dce_ident);
176 goto done;
177 }
178
179 /* Note that ip6_dst is only used for IRE_MULTICAST */
180 nce = ire_to_nce(ire, INADDR_ANY, &ip6h->ip6_dst);
181 if (nce == NULL) {
182 /* Allocation failure? */
183 ip_drop_output("ire_to_nce", mp, ill);
184 freemsg(mp);
185 error = ENOBUFS;
186 goto done;
187 }
188 if (nce->nce_is_condemned) {
189 nce_t *nce1;
190
191 nce1 = ire_handle_condemned_nce(nce, ire, NULL, ip6h, B_TRUE);
192 nce_refrele(nce);
193 if (nce1 == NULL) {
194 if (!repeat) {
195 /* Try finding a better IRE */
196 repeat = B_TRUE;
197 ire_refrele(ire);
198 goto repeat_ire;
199 }
200 /* Tried twice - drop packet */
201 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
202 ip_drop_output("No nce", mp, ill);
203 freemsg(mp);
204 error = ENOBUFS;
205 goto done;
206 }
207 nce = nce1;
208 }
209 /*
210 * For multicast with multirt we have a flag passed back from
211 * ire_lookup_multi_ill_v6 since we don't have an IRE for each
212 * possible multicast address.
213 * We also need a flag for multicast since we can't check
214 * whether RTF_MULTIRT is set in ixa_ire for multicast.
215 */
216 if (multirt) {
217 ixa->ixa_postfragfn = ip_postfrag_multirt_v6;
218 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
219 } else {
220 ixa->ixa_postfragfn = ire->ire_postfragfn;
221 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
222 }
223 ASSERT(ixa->ixa_nce == NULL);
224 ixa->ixa_nce = nce;
225
226 /*
227 * Check for a dce_t with a path mtu.
228 */
229 ifindex = 0;
230 if (IN6_IS_ADDR_LINKSCOPE(&dst))
231 ifindex = nce->nce_common->ncec_ill->ill_phyint->phyint_ifindex;
232
233 dce = dce_lookup_v6(&dst, ifindex, ipst, NULL);
234 ASSERT(dce != NULL);
235
236 if (!(ixaflags & IXAF_PMTU_DISCOVERY)) {
237 ixa->ixa_fragsize = IPV6_MIN_MTU;
238 } else if (dce->dce_flags & DCEF_PMTU) {
239 /*
240 * To avoid a periodic timer to increase the path MTU we
241 * look at dce_last_change_time each time we send a packet.
242 */
243 now = ddi_get_lbolt64();
244 if (TICK_TO_SEC(now) - dce->dce_last_change_time >
245 ipst->ips_ip_pathmtu_interval) {
246 /*
247 * Older than 20 minutes. Drop the path MTU information.
248 */
249 mutex_enter(&dce->dce_lock);
250 dce->dce_flags &= ~DCEF_PMTU;
251 dce->dce_last_change_time = TICK_TO_SEC(now);
252 mutex_exit(&dce->dce_lock);
253 dce_increment_generation(dce);
254 ixa->ixa_fragsize = ip_get_base_mtu(nce->nce_ill, ire);
255 } else {
256 uint_t fragsize;
257
258 fragsize = ip_get_base_mtu(nce->nce_ill, ire);
259 if (fragsize > dce->dce_pmtu)
260 fragsize = dce->dce_pmtu;
261 ixa->ixa_fragsize = fragsize;
262 }
263 } else {
264 ixa->ixa_fragsize = ip_get_base_mtu(nce->nce_ill, ire);
265 }
266
267 /*
268 * We use use ire_nexthop_ill (and not ncec_ill) to avoid the under ipmp
269 * interface for source address selection.
270 */
271 ill = ire_nexthop_ill(ire);
272
273 if (ixaflags & IXAF_SET_SOURCE) {
274 in6_addr_t src;
275
276 /*
277 * We use the final destination to get
278 * correct selection for source routed packets
279 */
280
281 /* If unreachable we have no ill but need some source */
282 if (ill == NULL) {
283 src = ipv6_loopback;
284 error = 0;
285 } else {
286 error = ip_select_source_v6(ill, &setsrc, &dst,
287 ixa->ixa_zoneid, ipst, B_FALSE,
288 ixa->ixa_src_preferences, &src, NULL, NULL);
289 }
290 if (error != 0) {
291 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
292 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
293 ip_drop_output("ipIfStatsOutDiscards - no source",
294 mp, ill);
295 freemsg(mp);
296 goto done;
297 }
298 ip6h->ip6_src = src;
299 } else if (ixaflags & IXAF_VERIFY_SOURCE) {
300 /* Check if the IP source is assigned to the host. */
301 if (!ip_verify_src(mp, ixa, NULL)) {
302 /* Don't send a packet with a source that isn't ours */
303 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
304 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
305 ip_drop_output("ipIfStatsOutDiscards - invalid source",
306 mp, ill);
307 freemsg(mp);
308 error = EADDRNOTAVAIL;
309 goto done;
310 }
311 }
312
313 /*
314 * Check against global IPsec policy to set the AH/ESP attributes.
315 * IPsec will set IXAF_IPSEC_* and ixa_ipsec_* as appropriate.
316 */
317 if (!(ixaflags & (IXAF_NO_IPSEC|IXAF_IPSEC_SECURE))) {
318 ASSERT(ixa->ixa_ipsec_policy == NULL);
319 mp = ip_output_attach_policy(mp, NULL, ip6h, NULL, ixa);
320 if (mp == NULL) {
321 /* MIB and ip_drop_packet already done */
322 return (EHOSTUNREACH); /* IPsec policy failure */
323 }
324 }
325
326 if (ill != NULL) {
327 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
328 } else {
329 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsHCOutRequests);
330 }
331
332 /*
333 * We update the statistics on the most specific IRE i.e., the first
334 * one we found.
335 * We don't have an IRE when we fragment, hence ire_ob_pkt_count
336 * can only count the use prior to fragmentation. However the MIB
337 * counters on the ill will be incremented in post fragmentation.
338 */
339 ire->ire_ob_pkt_count++;
340
341 /*
342 * Based on ire_type and ire_flags call one of:
343 * ire_send_local_v6 - for IRE_LOCAL and IRE_LOOPBACK
344 * ire_send_multirt_v6 - if RTF_MULTIRT
345 * ire_send_noroute_v6 - if RTF_REJECT or RTF_BLACHOLE
346 * ire_send_multicast_v6 - for IRE_MULTICAST
347 * ire_send_wire_v6 - for the rest.
348 */
349 error = (ire->ire_sendfn)(ire, mp, ip6h, ixa, &dce->dce_ident);
350 done:
351 ire_refrele(ire);
352 if (dce != NULL)
353 dce_refrele(dce);
354 if (ill != NULL)
355 ill_refrele(ill);
356 if (ixa->ixa_nce != NULL)
357 nce_refrele(ixa->ixa_nce);
358 ixa->ixa_nce = NULL;
359 return (error);
360 }
361
362 /*
363 * ire_sendfn() functions.
364 * These functions use the following xmit_attr:
365 * - ixa_fragsize - read to determine whether or not to fragment
366 * - IXAF_IPSEC_SECURE - to determine whether or not to invoke IPsec
367 * - ixa_ipsec_* are used inside IPsec
368 * - IXAF_LOOPBACK_COPY - for multicast
369 */
370
371
372 /*
373 * ire_sendfn for IRE_LOCAL and IRE_LOOPBACK
374 *
375 * The checks for restrict_interzone_loopback are done in ire_route_recursive.
376 */
377 /* ARGSUSED4 */
378 int
ire_send_local_v6(ire_t * ire,mblk_t * mp,void * iph_arg,ip_xmit_attr_t * ixa,uint32_t * identp)379 ire_send_local_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
380 ip_xmit_attr_t *ixa, uint32_t *identp)
381 {
382 ip6_t *ip6h = (ip6_t *)iph_arg;
383 ip_stack_t *ipst = ixa->ixa_ipst;
384 ill_t *ill = ire->ire_ill;
385 ip_recv_attr_t iras; /* NOTE: No bzero for performance */
386 uint_t pktlen = ixa->ixa_pktlen;
387
388 /*
389 * No fragmentation, no nce, and no application of IPsec.
390 *
391 *
392 * Note different order between IP provider and FW_HOOKS than in
393 * send_wire case.
394 */
395
396 /*
397 * DTrace this as ip:::send. A packet blocked by FW_HOOKS will fire the
398 * send probe, but not the receive probe.
399 */
400 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL, void_ip_t *,
401 ip6h, __dtrace_ipsr_ill_t *, ill, ipha_t *, NULL, ip6_t *, ip6h,
402 int, 1);
403
404 DTRACE_PROBE4(ip6__loopback__out__start,
405 ill_t *, NULL, ill_t *, ill,
406 ip6_t *, ip6h, mblk_t *, mp);
407
408 if (HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)) {
409 int error;
410
411 FW_HOOKS(ipst->ips_ip6_loopback_out_event,
412 ipst->ips_ipv6firewall_loopback_out,
413 NULL, ill, ip6h, mp, mp, 0, ipst, error);
414
415 DTRACE_PROBE1(ip6__loopback__out__end, mblk_t *, mp);
416 if (mp == NULL)
417 return (error);
418
419 /*
420 * Even if the destination was changed by the filter we use the
421 * forwarding decision that was made based on the address
422 * in ip_output/ip_set_destination.
423 */
424 /* Length could be different */
425 ip6h = (ip6_t *)mp->b_rptr;
426 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
427 }
428
429 /*
430 * If a callback is enabled then we need to know the
431 * source and destination zoneids for the packet. We already
432 * have those handy.
433 */
434 if (ipst->ips_ip6_observe.he_interested) {
435 zoneid_t szone, dzone;
436 zoneid_t stackzoneid;
437
438 stackzoneid = netstackid_to_zoneid(
439 ipst->ips_netstack->netstack_stackid);
440
441 if (stackzoneid == GLOBAL_ZONEID) {
442 /* Shared-IP zone */
443 dzone = ire->ire_zoneid;
444 szone = ixa->ixa_zoneid;
445 } else {
446 szone = dzone = stackzoneid;
447 }
448 ipobs_hook(mp, IPOBS_HOOK_LOCAL, szone, dzone, ill, ipst);
449 }
450
451 /* Handle lo0 stats */
452 ipst->ips_loopback_packets++;
453
454 /*
455 * Update output mib stats. Note that we can't move into the icmp
456 * sender (icmp_output etc) since they don't know the ill and the
457 * stats are per ill.
458 */
459 if (ixa->ixa_protocol == IPPROTO_ICMPV6) {
460 icmp6_t *icmp6;
461
462 icmp6 = (icmp6_t *)((uchar_t *)ip6h + ixa->ixa_ip_hdr_length);
463 icmp_update_out_mib_v6(ill, icmp6);
464 }
465
466 DTRACE_PROBE4(ip6__loopback__in__start,
467 ill_t *, ill, ill_t *, NULL,
468 ip6_t *, ip6h, mblk_t *, mp);
469
470 if (HOOKS6_INTERESTED_LOOPBACK_IN(ipst)) {
471 int error;
472
473 FW_HOOKS(ipst->ips_ip6_loopback_in_event,
474 ipst->ips_ipv6firewall_loopback_in,
475 ill, NULL, ip6h, mp, mp, 0, ipst, error);
476
477 DTRACE_PROBE1(ip6__loopback__in__end, mblk_t *, mp);
478 if (mp == NULL)
479 return (error);
480
481 /*
482 * Even if the destination was changed by the filter we use the
483 * forwarding decision that was made based on the address
484 * in ip_output/ip_set_destination.
485 */
486 /* Length could be different */
487 ip6h = (ip6_t *)mp->b_rptr;
488 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
489 }
490
491 DTRACE_IP7(receive, mblk_t *, mp, conn_t *, NULL, void_ip_t *,
492 ip6h, __dtrace_ipsr_ill_t *, ill, ipha_t *, NULL, ip6_t *, ip6h,
493 int, 1);
494
495 /* Map ixa to ira including IPsec policies */
496 ipsec_out_to_in(ixa, ill, &iras);
497 iras.ira_pktlen = pktlen;
498
499 ire->ire_ib_pkt_count++;
500 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInReceives);
501 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCInOctets, pktlen);
502
503 /* Destined to ire_zoneid - use that for fanout */
504 iras.ira_zoneid = ire->ire_zoneid;
505
506 if (is_system_labeled()) {
507 iras.ira_flags |= IRAF_SYSTEM_LABELED;
508
509 /*
510 * This updates ira_cred, ira_tsl and ira_free_flags based
511 * on the label. We don't expect this to ever fail for
512 * loopback packets, so we silently drop the packet should it
513 * fail.
514 */
515 if (!tsol_get_pkt_label(mp, IPV6_VERSION, &iras)) {
516 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
517 ip_drop_input("tsol_get_pkt_label", mp, ill);
518 freemsg(mp);
519 return (0);
520 }
521 ASSERT(iras.ira_tsl != NULL);
522
523 /* tsol_get_pkt_label sometimes does pullupmsg */
524 ip6h = (ip6_t *)mp->b_rptr;
525 }
526
527 ip_fanout_v6(mp, ip6h, &iras);
528
529 /* We moved any IPsec refs from ixa to iras */
530 ira_cleanup(&iras, B_FALSE);
531 return (0);
532 }
533
534 static void
multirt_check_v6(ire_t * ire,ip6_t * ip6h,ip_xmit_attr_t * ixa)535 multirt_check_v6(ire_t *ire, ip6_t *ip6h, ip_xmit_attr_t *ixa)
536 {
537 ip_stack_t *ipst = ixa->ixa_ipst;
538
539 /* Limit the TTL on multirt packets. Do this even if IPV6_HOPLIMIT */
540 if (ire->ire_type & IRE_MULTICAST) {
541 if (ip6h->ip6_hops > 1) {
542 ip2dbg(("ire_send_multirt_v6: forcing multicast "
543 "multirt TTL to 1 (was %d)\n", ip6h->ip6_hops));
544 ip6h->ip6_hops = 1;
545 }
546 ixa->ixa_flags |= IXAF_NO_TTL_CHANGE;
547 } else if ((ipst->ips_ip_multirt_ttl > 0) &&
548 (ip6h->ip6_hops > ipst->ips_ip_multirt_ttl)) {
549 ip6h->ip6_hops = ipst->ips_ip_multirt_ttl;
550 /*
551 * Need to ensure we don't increase the ttl should we go through
552 * ire_send_multicast.
553 */
554 ixa->ixa_flags |= IXAF_NO_TTL_CHANGE;
555 }
556
557 /* For IPv6 this also needs to insert a fragment header */
558 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
559 }
560
561 /*
562 * ire_sendfn for IRE_MULTICAST
563 *
564 * Note that we do path MTU discovery by default for IPv6 multicast. But
565 * since unconnected UDP and RAW sockets don't set IXAF_PMTU_DISCOVERY
566 * only connected sockets get this by default.
567 */
568 int
ire_send_multicast_v6(ire_t * ire,mblk_t * mp,void * iph_arg,ip_xmit_attr_t * ixa,uint32_t * identp)569 ire_send_multicast_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
570 ip_xmit_attr_t *ixa, uint32_t *identp)
571 {
572 ip6_t *ip6h = (ip6_t *)iph_arg;
573 ip_stack_t *ipst = ixa->ixa_ipst;
574 ill_t *ill = ire->ire_ill;
575 iaflags_t ixaflags = ixa->ixa_flags;
576
577 /*
578 * The IRE_MULTICAST is the same whether or not multirt is in use.
579 * Hence we need special-case code.
580 */
581 if (ixaflags & IXAF_MULTIRT_MULTICAST)
582 multirt_check_v6(ire, ip6h, ixa);
583
584 /*
585 * Check if anything in ip_input_v6 wants a copy of the transmitted
586 * packet (after IPsec and fragmentation)
587 *
588 * 1. Multicast routers always need a copy unless SO_DONTROUTE is set
589 * RSVP and the rsvp daemon is an example of a
590 * protocol and user level process that
591 * handles it's own routing. Hence, it uses the
592 * SO_DONTROUTE option to accomplish this.
593 * 2. If the sender has set IP_MULTICAST_LOOP, then we just
594 * check whether there are any receivers for the group on the ill
595 * (ignoring the zoneid).
596 * 3. If IP_MULTICAST_LOOP is not set, then we check if there are
597 * any members in other shared-IP zones.
598 * If such members exist, then we indicate that the sending zone
599 * shouldn't get a loopback copy to preserve the IP_MULTICAST_LOOP
600 * behavior.
601 *
602 * When we loopback we skip hardware checksum to make sure loopback
603 * copy is checksumed.
604 *
605 * Note that ire_ill is the upper in the case of IPMP.
606 */
607 ixa->ixa_flags &= ~(IXAF_LOOPBACK_COPY | IXAF_NO_HW_CKSUM);
608 if (ipst->ips_ip_g_mrouter && ill->ill_mrouter_cnt > 0 &&
609 !(ixaflags & IXAF_DONTROUTE)) {
610 ixa->ixa_flags |= IXAF_LOOPBACK_COPY | IXAF_NO_HW_CKSUM;
611 } else if (ixaflags & IXAF_MULTICAST_LOOP) {
612 /*
613 * If this zone or any other zone has members then loopback
614 * a copy.
615 */
616 if (ill_hasmembers_v6(ill, &ip6h->ip6_dst))
617 ixa->ixa_flags |= IXAF_LOOPBACK_COPY | IXAF_NO_HW_CKSUM;
618 } else if (ipst->ips_netstack->netstack_numzones > 1) {
619 /*
620 * This zone should not have a copy. But there are some other
621 * zones which might have members.
622 */
623 if (ill_hasmembers_otherzones_v6(ill, &ip6h->ip6_dst,
624 ixa->ixa_zoneid)) {
625 ixa->ixa_flags |= IXAF_NO_LOOP_ZONEID_SET;
626 ixa->ixa_no_loop_zoneid = ixa->ixa_zoneid;
627 ixa->ixa_flags |= IXAF_LOOPBACK_COPY | IXAF_NO_HW_CKSUM;
628 }
629 }
630
631 /*
632 * Unless IPV6_HOPLIMIT or ire_send_multirt_v6 already set a ttl,
633 * force the ttl to the IP_MULTICAST_TTL value
634 */
635 if (!(ixaflags & IXAF_NO_TTL_CHANGE)) {
636 ip6h->ip6_hops = ixa->ixa_multicast_ttl;
637 }
638
639 return (ire_send_wire_v6(ire, mp, ip6h, ixa, identp));
640 }
641
642 /*
643 * ire_sendfn for IREs with RTF_MULTIRT
644 */
645 int
ire_send_multirt_v6(ire_t * ire,mblk_t * mp,void * iph_arg,ip_xmit_attr_t * ixa,uint32_t * identp)646 ire_send_multirt_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
647 ip_xmit_attr_t *ixa, uint32_t *identp)
648 {
649 ip6_t *ip6h = (ip6_t *)iph_arg;
650
651 multirt_check_v6(ire, ip6h, ixa);
652
653 if (ire->ire_type & IRE_MULTICAST)
654 return (ire_send_multicast_v6(ire, mp, ip6h, ixa, identp));
655 else
656 return (ire_send_wire_v6(ire, mp, ip6h, ixa, identp));
657 }
658
659 /*
660 * ire_sendfn for IREs with RTF_REJECT/RTF_BLACKHOLE, including IRE_NOROUTE
661 */
662 /* ARGSUSED4 */
663 int
ire_send_noroute_v6(ire_t * ire,mblk_t * mp,void * iph_arg,ip_xmit_attr_t * ixa,uint32_t * identp)664 ire_send_noroute_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
665 ip_xmit_attr_t *ixa, uint32_t *identp)
666 {
667 ip6_t *ip6h = (ip6_t *)iph_arg;
668 ip_stack_t *ipst = ixa->ixa_ipst;
669 ill_t *ill;
670 ip_recv_attr_t iras;
671 boolean_t dummy;
672
673 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutNoRoutes);
674
675 if (ire->ire_type & IRE_NOROUTE) {
676 /* A lack of a route as opposed to RTF_REJECT|BLACKHOLE */
677 ip_rts_change_v6(RTM_MISS, &ip6h->ip6_dst, 0, 0, 0, 0, 0, 0,
678 RTA_DST, ipst);
679 }
680
681 if (ire->ire_flags & RTF_BLACKHOLE) {
682 ip_drop_output("ipIfStatsOutNoRoutes RTF_BLACKHOLE", mp, NULL);
683 freemsg(mp);
684 /* No error even for local senders - silent blackhole */
685 return (0);
686 }
687 ip_drop_output("ipIfStatsOutNoRoutes RTF_REJECT", mp, NULL);
688
689 /*
690 * We need an ill_t for the ip_recv_attr_t even though this packet
691 * was never received and icmp_unreachable doesn't currently use
692 * ira_ill.
693 */
694 ill = ill_lookup_on_name("lo0", B_FALSE,
695 !(ixa->ixa_flags & IRAF_IS_IPV4), &dummy, ipst);
696 if (ill == NULL) {
697 freemsg(mp);
698 return (EHOSTUNREACH);
699 }
700
701 bzero(&iras, sizeof (iras));
702 /* Map ixa to ira including IPsec policies */
703 ipsec_out_to_in(ixa, ill, &iras);
704
705 icmp_unreachable_v6(mp, ICMP6_DST_UNREACH_NOROUTE, B_FALSE, &iras);
706 /* We moved any IPsec refs from ixa to iras */
707 ira_cleanup(&iras, B_FALSE);
708
709 ill_refrele(ill);
710 return (EHOSTUNREACH);
711 }
712
713 /*
714 * Calculate a checksum ignoring any hardware capabilities
715 *
716 * Returns B_FALSE if the packet was too short for the checksum. Caller
717 * should free and do stats.
718 */
719 static boolean_t
ip_output_sw_cksum_v6(mblk_t * mp,ip6_t * ip6h,ip_xmit_attr_t * ixa)720 ip_output_sw_cksum_v6(mblk_t *mp, ip6_t *ip6h, ip_xmit_attr_t *ixa)
721 {
722 ip_stack_t *ipst = ixa->ixa_ipst;
723 uint_t pktlen = ixa->ixa_pktlen;
724 uint16_t *cksump;
725 uint32_t cksum;
726 uint8_t protocol = ixa->ixa_protocol;
727 uint16_t ip_hdr_length = ixa->ixa_ip_hdr_length;
728
729 #define iphs ((uint16_t *)ip6h)
730
731 /* Just in case it contained garbage */
732 DB_CKSUMFLAGS(mp) &= ~HCK_FLAGS;
733
734 /*
735 * Calculate ULP checksum
736 */
737 if (protocol == IPPROTO_TCP) {
738 cksump = IPH_TCPH_CHECKSUMP(ip6h, ip_hdr_length);
739 cksum = IP_TCP_CSUM_COMP;
740 } else if (protocol == IPPROTO_UDP) {
741 cksump = IPH_UDPH_CHECKSUMP(ip6h, ip_hdr_length);
742 cksum = IP_UDP_CSUM_COMP;
743 } else if (protocol == IPPROTO_SCTP) {
744 sctp_hdr_t *sctph;
745
746 ASSERT(MBLKL(mp) >= (ip_hdr_length + sizeof (*sctph)));
747 sctph = (sctp_hdr_t *)(mp->b_rptr + ip_hdr_length);
748 /*
749 * Zero out the checksum field to ensure proper
750 * checksum calculation.
751 */
752 sctph->sh_chksum = 0;
753 #ifdef DEBUG
754 if (!skip_sctp_cksum)
755 #endif
756 sctph->sh_chksum = sctp_cksum(mp, ip_hdr_length);
757 return (B_TRUE);
758 } else if (ixa->ixa_flags & IXAF_SET_RAW_CKSUM) {
759 /*
760 * icmp has placed length and routing
761 * header adjustment in the checksum field.
762 */
763 cksump = (uint16_t *)(((uint8_t *)ip6h) + ip_hdr_length +
764 ixa->ixa_raw_cksum_offset);
765 cksum = htons(protocol);
766 } else if (protocol == IPPROTO_ICMPV6) {
767 cksump = IPH_ICMPV6_CHECKSUMP(ip6h, ip_hdr_length);
768 cksum = IP_ICMPV6_CSUM_COMP; /* Pseudo-header cksum */
769 } else {
770 return (B_TRUE);
771 }
772
773 /* ULP puts the checksum field is in the first mblk */
774 ASSERT(((uchar_t *)cksump) + sizeof (uint16_t) <= mp->b_wptr);
775
776 /*
777 * We accumulate the pseudo header checksum in cksum.
778 * This is pretty hairy code, so watch close. One
779 * thing to keep in mind is that UDP and TCP have
780 * stored their respective datagram lengths in their
781 * checksum fields. This lines things up real nice.
782 */
783 cksum += iphs[4] + iphs[5] + iphs[6] + iphs[7] +
784 iphs[8] + iphs[9] + iphs[10] + iphs[11] +
785 iphs[12] + iphs[13] + iphs[14] + iphs[15] +
786 iphs[16] + iphs[17] + iphs[18] + iphs[19];
787 cksum = IP_CSUM(mp, ip_hdr_length, cksum);
788
789 /*
790 * For UDP/IPv6 a zero UDP checksum is not allowed.
791 * Change to 0xffff
792 */
793 if (protocol == IPPROTO_UDP && cksum == 0)
794 *cksump = ~cksum;
795 else
796 *cksump = cksum;
797
798 IP6_STAT(ipst, ip6_out_sw_cksum);
799 IP6_STAT_UPDATE(ipst, ip6_out_sw_cksum_bytes, pktlen);
800
801 /* No IP header checksum for IPv6 */
802
803 return (B_TRUE);
804 #undef iphs
805 }
806
807 /* There are drivers that can't do partial checksum for ICMPv6 */
808 int nxge_cksum_workaround = 1;
809
810 /*
811 * Calculate the ULP checksum - try to use hardware.
812 * In the case of MULTIRT or multicast the
813 * IXAF_NO_HW_CKSUM is set in which case we use software.
814 *
815 * Returns B_FALSE if the packet was too short for the checksum. Caller
816 * should free and do stats.
817 */
818 static boolean_t
ip_output_cksum_v6(iaflags_t ixaflags,mblk_t * mp,ip6_t * ip6h,ip_xmit_attr_t * ixa,ill_t * ill)819 ip_output_cksum_v6(iaflags_t ixaflags, mblk_t *mp, ip6_t *ip6h,
820 ip_xmit_attr_t *ixa, ill_t *ill)
821 {
822 uint_t pktlen = ixa->ixa_pktlen;
823 uint16_t *cksump;
824 uint16_t hck_flags;
825 uint32_t cksum;
826 uint8_t protocol = ixa->ixa_protocol;
827 uint16_t ip_hdr_length = ixa->ixa_ip_hdr_length;
828
829 #define iphs ((uint16_t *)ip6h)
830
831 if ((ixaflags & IXAF_NO_HW_CKSUM) || !ILL_HCKSUM_CAPABLE(ill) ||
832 !dohwcksum) {
833 return (ip_output_sw_cksum_v6(mp, ip6h, ixa));
834 }
835
836 /*
837 * Calculate ULP checksum. Note that we don't use cksump and cksum
838 * if the ill has FULL support.
839 */
840 if (protocol == IPPROTO_TCP) {
841 cksump = IPH_TCPH_CHECKSUMP(ip6h, ip_hdr_length);
842 cksum = IP_TCP_CSUM_COMP; /* Pseudo-header cksum */
843 } else if (protocol == IPPROTO_UDP) {
844 cksump = IPH_UDPH_CHECKSUMP(ip6h, ip_hdr_length);
845 cksum = IP_UDP_CSUM_COMP; /* Pseudo-header cksum */
846 } else if (protocol == IPPROTO_SCTP) {
847 sctp_hdr_t *sctph;
848
849 ASSERT(MBLKL(mp) >= (ip_hdr_length + sizeof (*sctph)));
850 sctph = (sctp_hdr_t *)(mp->b_rptr + ip_hdr_length);
851 /*
852 * Zero out the checksum field to ensure proper
853 * checksum calculation.
854 */
855 sctph->sh_chksum = 0;
856 #ifdef DEBUG
857 if (!skip_sctp_cksum)
858 #endif
859 sctph->sh_chksum = sctp_cksum(mp, ip_hdr_length);
860 goto ip_hdr_cksum;
861 } else if (ixa->ixa_flags & IXAF_SET_RAW_CKSUM) {
862 /*
863 * icmp has placed length and routing
864 * header adjustment in the checksum field.
865 */
866 cksump = (uint16_t *)(((uint8_t *)ip6h) + ip_hdr_length +
867 ixa->ixa_raw_cksum_offset);
868 cksum = htons(protocol);
869 } else if (protocol == IPPROTO_ICMPV6) {
870 /*
871 * Currently we assume no HW support for ICMP checksum calc.
872 *
873 * When HW support is advertised for ICMP, we'll want the
874 * following to be set:
875 * cksump = IPH_ICMPV6_CHECKSUMP(ip6h, ip_hdr_length);
876 * cksum = IP_ICMPV6_CSUM_COMP; Pseudo-header cksum
877 */
878
879 return (ip_output_sw_cksum_v6(mp, ip6h, ixa));
880 } else {
881 ip_hdr_cksum:
882 /* No IP header checksum for IPv6 */
883 return (B_TRUE);
884 }
885
886 /* ULP puts the checksum field is in the first mblk */
887 ASSERT(((uchar_t *)cksump) + sizeof (uint16_t) <= mp->b_wptr);
888
889 /*
890 * Underlying interface supports hardware checksum offload for
891 * the payload; leave the payload checksum for the hardware to
892 * calculate. N.B: We only need to set up checksum info on the
893 * first mblk.
894 */
895 hck_flags = ill->ill_hcksum_capab->ill_hcksum_txflags;
896
897 DB_CKSUMFLAGS(mp) &= ~HCK_FLAGS;
898 if (hck_flags & HCKSUM_INET_FULL_V6) {
899 /*
900 * Hardware calculates pseudo-header, header and the
901 * payload checksums, so clear the checksum field in
902 * the protocol header.
903 */
904 *cksump = 0;
905 DB_CKSUMFLAGS(mp) |= HCK_FULLCKSUM;
906 return (B_TRUE);
907 }
908 if (((hck_flags) & HCKSUM_INET_PARTIAL) &&
909 (protocol != IPPROTO_ICMPV6 || !nxge_cksum_workaround)) {
910 /*
911 * Partial checksum offload has been enabled. Fill
912 * the checksum field in the protocol header with the
913 * pseudo-header checksum value.
914 *
915 * We accumulate the pseudo header checksum in cksum.
916 * This is pretty hairy code, so watch close. One
917 * thing to keep in mind is that UDP and TCP have
918 * stored their respective datagram lengths in their
919 * checksum fields. This lines things up real nice.
920 */
921 cksum += iphs[4] + iphs[5] + iphs[6] + iphs[7] +
922 iphs[8] + iphs[9] + iphs[10] + iphs[11] +
923 iphs[12] + iphs[13] + iphs[14] + iphs[15] +
924 iphs[16] + iphs[17] + iphs[18] + iphs[19];
925 cksum += *(cksump);
926 cksum = (cksum & 0xFFFF) + (cksum >> 16);
927 *(cksump) = (cksum & 0xFFFF) + (cksum >> 16);
928
929 /*
930 * Offsets are relative to beginning of IP header.
931 */
932 DB_CKSUMSTART(mp) = ip_hdr_length;
933 DB_CKSUMSTUFF(mp) = (uint8_t *)cksump - (uint8_t *)ip6h;
934 DB_CKSUMEND(mp) = pktlen;
935 DB_CKSUMFLAGS(mp) |= HCK_PARTIALCKSUM;
936 return (B_TRUE);
937 }
938 /* Hardware capabilities include neither full nor partial IPv6 */
939 return (ip_output_sw_cksum_v6(mp, ip6h, ixa));
940 #undef iphs
941 }
942
943 /*
944 * ire_sendfn for offlink and onlink destinations.
945 * Also called from the multicast, and multirt send functions.
946 *
947 * Assumes that the caller has a hold on the ire.
948 *
949 * This function doesn't care if the IRE just became condemned since that
950 * can happen at any time.
951 */
952 /* ARGSUSED */
953 int
ire_send_wire_v6(ire_t * ire,mblk_t * mp,void * iph_arg,ip_xmit_attr_t * ixa,uint32_t * identp)954 ire_send_wire_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
955 ip_xmit_attr_t *ixa, uint32_t *identp)
956 {
957 ip_stack_t *ipst = ixa->ixa_ipst;
958 ip6_t *ip6h = (ip6_t *)iph_arg;
959 iaflags_t ixaflags = ixa->ixa_flags;
960 ill_t *ill;
961 uint32_t pktlen = ixa->ixa_pktlen;
962
963 ASSERT(ixa->ixa_nce != NULL);
964 ill = ixa->ixa_nce->nce_ill;
965
966 /*
967 * Update output mib stats. Note that we can't move into the icmp
968 * sender (icmp_output etc) since they don't know the ill and the
969 * stats are per ill.
970 *
971 * With IPMP we record the stats on the upper ill.
972 */
973 if (ixa->ixa_protocol == IPPROTO_ICMPV6) {
974 icmp6_t *icmp6;
975
976 icmp6 = (icmp6_t *)((uchar_t *)ip6h + ixa->ixa_ip_hdr_length);
977 icmp_update_out_mib_v6(ixa->ixa_nce->nce_common->ncec_ill,
978 icmp6);
979 }
980
981 if (ixaflags & IXAF_DONTROUTE)
982 ip6h->ip6_hops = 1;
983
984 /*
985 * This might set b_band, thus the IPsec and fragmentation
986 * code in IP ensures that b_band is updated in the first mblk.
987 */
988 if (IPP_ENABLED(IPP_LOCAL_OUT, ipst)) {
989 /* ip_process translates an IS_UNDER_IPMP */
990 mp = ip_process(IPP_LOCAL_OUT, mp, ill, ill);
991 if (mp == NULL) {
992 /* ip_drop_packet and MIB done */
993 return (0); /* Might just be delayed */
994 }
995 }
996
997 /*
998 * To handle IPsec/iptun's labeling needs we need to tag packets
999 * while we still have ixa_tsl
1000 */
1001 if (is_system_labeled() && ixa->ixa_tsl != NULL &&
1002 (ill->ill_mactype == DL_6TO4 || ill->ill_mactype == DL_IPV4 ||
1003 ill->ill_mactype == DL_IPV6)) {
1004 cred_t *newcr;
1005
1006 newcr = copycred_from_tslabel(ixa->ixa_cred, ixa->ixa_tsl,
1007 KM_NOSLEEP);
1008 if (newcr == NULL) {
1009 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
1010 ip_drop_output("ipIfStatsOutDiscards - newcr",
1011 mp, ill);
1012 freemsg(mp);
1013 return (ENOBUFS);
1014 }
1015 mblk_setcred(mp, newcr, NOPID);
1016 crfree(newcr); /* mblk_setcred did its own crhold */
1017 }
1018
1019 /*
1020 * IXAF_IPV6_ADD_FRAGHDR is set for CGTP so that we will add a
1021 * fragment header without fragmenting. CGTP on the receiver will
1022 * filter duplicates on the ident field.
1023 */
1024 if (pktlen > ixa->ixa_fragsize ||
1025 (ixaflags & (IXAF_IPSEC_SECURE|IXAF_IPV6_ADD_FRAGHDR))) {
1026 uint32_t ident = 0;
1027
1028 if (ixaflags & IXAF_IPSEC_SECURE)
1029 pktlen += ipsec_out_extra_length(ixa);
1030
1031 if (pktlen > IP_MAXPACKET)
1032 return (EMSGSIZE);
1033
1034 if (ixaflags & IXAF_SET_ULP_CKSUM) {
1035 /*
1036 * Compute ULP checksum using software
1037 */
1038 if (!ip_output_sw_cksum_v6(mp, ip6h, ixa)) {
1039 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
1040 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
1041 freemsg(mp);
1042 return (EINVAL);
1043 }
1044 /* Avoid checksum again below if we only add fraghdr */
1045 ixaflags &= ~IXAF_SET_ULP_CKSUM;
1046 }
1047
1048 /*
1049 * If we need a fragment header, pick the ident and insert
1050 * the header before IPsec to we have a place to store
1051 * the ident value.
1052 */
1053 if ((ixaflags & IXAF_IPV6_ADD_FRAGHDR) ||
1054 pktlen > ixa->ixa_fragsize) {
1055 /*
1056 * If this packet would generate a icmp_frag_needed
1057 * message, we need to handle it before we do the IPsec
1058 * processing. Otherwise, we need to strip the IPsec
1059 * headers before we send up the message to the ULPs
1060 * which becomes messy and difficult.
1061 */
1062 if ((pktlen > ixa->ixa_fragsize) &&
1063 (ixaflags & IXAF_DONTFRAG)) {
1064 /* Generate ICMP and return error */
1065 ip_recv_attr_t iras;
1066
1067 DTRACE_PROBE4(ip6__fragsize__fail,
1068 uint_t, pktlen, uint_t, ixa->ixa_fragsize,
1069 uint_t, ixa->ixa_pktlen,
1070 uint_t, ixa->ixa_pmtu);
1071
1072 bzero(&iras, sizeof (iras));
1073 /* Map ixa to ira including IPsec policies */
1074 ipsec_out_to_in(ixa, ill, &iras);
1075
1076 ip_drop_output("ICMP6_PKT_TOO_BIG", mp, ill);
1077 icmp_pkt2big_v6(mp, ixa->ixa_fragsize, B_TRUE,
1078 &iras);
1079 /* We moved any IPsec refs from ixa to iras */
1080 ira_cleanup(&iras, B_FALSE);
1081 return (EMSGSIZE);
1082 }
1083 DTRACE_PROBE4(ip6__fragsize__ok, uint_t, pktlen,
1084 uint_t, ixa->ixa_fragsize, uint_t, ixa->ixa_pktlen,
1085 uint_t, ixa->ixa_pmtu);
1086 /*
1087 * Assign an ident value for this packet. There could
1088 * be other threads targeting the same destination, so
1089 * we have to arrange for a atomic increment.
1090 * Normally ixa_extra_ident is 0, but in the case of
1091 * LSO it will be the number of TCP segments that the
1092 * driver/hardware will extraly construct.
1093 *
1094 * Note that cl_inet_ipident has only been used for
1095 * IPv4. We don't use it here.
1096 */
1097 ident = atomic_add_32_nv(identp, ixa->ixa_extra_ident +
1098 1);
1099 ixa->ixa_ident = ident; /* In case we do IPsec */
1100 }
1101 if (ixaflags & IXAF_IPSEC_SECURE) {
1102 /*
1103 * Pass in sufficient information so that
1104 * IPsec can determine whether to fragment, and
1105 * which function to call after fragmentation.
1106 */
1107 return (ipsec_out_process(mp, ixa));
1108 }
1109
1110 mp = ip_fraghdr_add_v6(mp, ident, ixa);
1111 if (mp == NULL) {
1112 /* MIB and ip_drop_output already done */
1113 return (ENOMEM);
1114 }
1115 ASSERT(pktlen == ixa->ixa_pktlen);
1116 pktlen += sizeof (ip6_frag_t);
1117
1118 if (pktlen > ixa->ixa_fragsize) {
1119 return (ip_fragment_v6(mp, ixa->ixa_nce, ixaflags,
1120 pktlen, ixa->ixa_fragsize,
1121 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
1122 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
1123 &ixa->ixa_cookie));
1124 }
1125 }
1126 if (ixaflags & IXAF_SET_ULP_CKSUM) {
1127 /* Compute ULP checksum and IP header checksum */
1128 /* An IS_UNDER_IPMP ill is ok here */
1129 if (!ip_output_cksum_v6(ixaflags, mp, ip6h, ixa, ill)) {
1130 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
1131 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
1132 freemsg(mp);
1133 return (EINVAL);
1134 }
1135 }
1136 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixaflags,
1137 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
1138 ixa->ixa_no_loop_zoneid, &ixa->ixa_cookie));
1139 }
1140
1141 /*
1142 * Post fragmentation function for RTF_MULTIRT routes.
1143 * Since IRE_MULTICASTs might have RTF_MULTIRT, this function
1144 * checks IXAF_LOOPBACK_COPY.
1145 *
1146 * If no packet is sent due to failures then we return an errno, but if at
1147 * least one succeeded we return zero.
1148 */
1149 int
ip_postfrag_multirt_v6(mblk_t * mp,nce_t * nce,iaflags_t ixaflags,uint_t pkt_len,uint32_t xmit_hint,zoneid_t szone,zoneid_t nolzid,uintptr_t * ixacookie)1150 ip_postfrag_multirt_v6(mblk_t *mp, nce_t *nce, iaflags_t ixaflags,
1151 uint_t pkt_len, uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid,
1152 uintptr_t *ixacookie)
1153 {
1154 irb_t *irb;
1155 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
1156 ire_t *ire;
1157 ire_t *ire1;
1158 mblk_t *mp1;
1159 nce_t *nce1;
1160 ill_t *ill = nce->nce_ill;
1161 ill_t *ill1;
1162 ip_stack_t *ipst = ill->ill_ipst;
1163 int error = 0;
1164 int num_sent = 0;
1165 int err;
1166 uint_t ire_type;
1167 in6_addr_t nexthop;
1168
1169 ASSERT(!(ixaflags & IXAF_IS_IPV4));
1170
1171 /* Check for IXAF_LOOPBACK_COPY */
1172 if (ixaflags & IXAF_LOOPBACK_COPY) {
1173 mblk_t *mp1;
1174
1175 mp1 = copymsg(mp);
1176 if (mp1 == NULL) {
1177 /* Failed to deliver the loopback copy. */
1178 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
1179 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
1180 error = ENOBUFS;
1181 } else {
1182 ip_postfrag_loopback(mp1, nce, ixaflags, pkt_len,
1183 nolzid);
1184 }
1185 }
1186
1187 /*
1188 * Loop over RTF_MULTIRT for ip6_dst in the same bucket. Send
1189 * a copy to each one.
1190 * Use the nce (nexthop) and ip6_dst to find the ire.
1191 *
1192 * MULTIRT is not designed to work with shared-IP zones thus we don't
1193 * need to pass a zoneid or a label to the IRE lookup.
1194 */
1195 if (IN6_ARE_ADDR_EQUAL(&nce->nce_addr, &ip6h->ip6_dst)) {
1196 /* Broadcast and multicast case */
1197 ire = ire_ftable_lookup_v6(&ip6h->ip6_dst, 0, 0, 0, NULL,
1198 ALL_ZONES, NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
1199 } else {
1200 /* Unicast case */
1201 ire = ire_ftable_lookup_v6(&ip6h->ip6_dst, 0, &nce->nce_addr,
1202 0, NULL, ALL_ZONES, NULL, MATCH_IRE_GW, 0, ipst, NULL);
1203 }
1204
1205 if (ire == NULL ||
1206 (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
1207 !(ire->ire_flags & RTF_MULTIRT)) {
1208 /* Drop */
1209 ip_drop_output("ip_postfrag_multirt didn't find route",
1210 mp, nce->nce_ill);
1211 if (ire != NULL)
1212 ire_refrele(ire);
1213 return (ENETUNREACH);
1214 }
1215
1216 irb = ire->ire_bucket;
1217 irb_refhold(irb);
1218 for (ire1 = irb->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) {
1219 if (IRE_IS_CONDEMNED(ire1) ||
1220 !(ire1->ire_flags & RTF_MULTIRT))
1221 continue;
1222
1223 /* Note: When IPv6 uses radix tree we don't need this check */
1224 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, &ire1->ire_addr_v6))
1225 continue;
1226
1227 /* Do the ire argument one after the loop */
1228 if (ire1 == ire)
1229 continue;
1230
1231 ill1 = ire_nexthop_ill(ire1);
1232 if (ill1 == NULL) {
1233 /*
1234 * This ire might not have been picked by
1235 * ire_route_recursive, in which case ire_dep might
1236 * not have been setup yet.
1237 * We kick ire_route_recursive to try to resolve
1238 * starting at ire1.
1239 */
1240 ire_t *ire2;
1241 uint_t match_flags = MATCH_IRE_DSTONLY;
1242
1243 if (ire1->ire_ill != NULL)
1244 match_flags |= MATCH_IRE_ILL;
1245 ire2 = ire_route_recursive_impl_v6(ire1,
1246 &ire1->ire_addr_v6, ire1->ire_type, ire1->ire_ill,
1247 ire1->ire_zoneid, NULL, match_flags,
1248 IRR_ALLOCATE, 0, ipst, NULL, NULL, NULL);
1249 if (ire2 != NULL)
1250 ire_refrele(ire2);
1251 ill1 = ire_nexthop_ill(ire1);
1252 }
1253 if (ill1 == NULL) {
1254 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
1255 ip_drop_output("ipIfStatsOutDiscards - no ill",
1256 mp, ill);
1257 error = ENETUNREACH;
1258 continue;
1259 }
1260 /* Pick the addr and type to use for ndp_nce_init */
1261 if (nce->nce_common->ncec_flags & NCE_F_MCAST) {
1262 ire_type = IRE_MULTICAST;
1263 nexthop = ip6h->ip6_dst;
1264 } else {
1265 ire_type = ire1->ire_type; /* Doesn't matter */
1266 nexthop = ire1->ire_gateway_addr_v6;
1267 }
1268
1269 /* If IPMP meta or under, then we just drop */
1270 if (ill1->ill_grp != NULL) {
1271 BUMP_MIB(ill1->ill_ip_mib, ipIfStatsOutDiscards);
1272 ip_drop_output("ipIfStatsOutDiscards - IPMP",
1273 mp, ill1);
1274 ill_refrele(ill1);
1275 error = ENETUNREACH;
1276 continue;
1277 }
1278
1279 nce1 = ndp_nce_init(ill1, &nexthop, ire_type);
1280 if (nce1 == NULL) {
1281 BUMP_MIB(ill1->ill_ip_mib, ipIfStatsOutDiscards);
1282 ip_drop_output("ipIfStatsOutDiscards - no nce",
1283 mp, ill1);
1284 ill_refrele(ill1);
1285 error = ENOBUFS;
1286 continue;
1287 }
1288 mp1 = copymsg(mp);
1289 if (mp1 == NULL) {
1290 BUMP_MIB(ill1->ill_ip_mib, ipIfStatsOutDiscards);
1291 ip_drop_output("ipIfStatsOutDiscards", mp, ill1);
1292 nce_refrele(nce1);
1293 ill_refrele(ill1);
1294 error = ENOBUFS;
1295 continue;
1296 }
1297 /* Preserve HW checksum for this copy */
1298 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
1299 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
1300 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
1301 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
1302 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
1303
1304 ire1->ire_ob_pkt_count++;
1305 err = ip_xmit(mp1, nce1, ixaflags, pkt_len, xmit_hint, szone,
1306 0, ixacookie);
1307 if (err == 0)
1308 num_sent++;
1309 else
1310 error = err;
1311 nce_refrele(nce1);
1312 ill_refrele(ill1);
1313 }
1314 irb_refrele(irb);
1315 ire_refrele(ire);
1316 /* Finally, the main one */
1317 err = ip_xmit(mp, nce, ixaflags, pkt_len, xmit_hint, szone, 0,
1318 ixacookie);
1319 if (err == 0)
1320 num_sent++;
1321 else
1322 error = err;
1323 if (num_sent > 0)
1324 return (0);
1325 else
1326 return (error);
1327 }
1328