787 Kernel panic in ip_input.c
[illumos-gate.git] / usr / src / uts / common / inet / ip / ip6_input.c
blobc7c241f9441c4b212c0a902f7749fb05257cdd12
1 /*
2 * CDDL HEADER START
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.
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.
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]
19 * CDDL HEADER END
23 * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved
25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
27 /* Copyright (c) 1990 Mentat Inc. */
29 #include <sys/types.h>
30 #include <sys/stream.h>
31 #include <sys/dlpi.h>
32 #include <sys/stropts.h>
33 #include <sys/sysmacros.h>
34 #include <sys/strsubr.h>
35 #include <sys/strlog.h>
36 #include <sys/strsun.h>
37 #include <sys/zone.h>
38 #define _SUN_TPI_VERSION 2
39 #include <sys/tihdr.h>
40 #include <sys/xti_inet.h>
41 #include <sys/ddi.h>
42 #include <sys/sunddi.h>
43 #include <sys/cmn_err.h>
44 #include <sys/debug.h>
45 #include <sys/kobj.h>
46 #include <sys/modctl.h>
47 #include <sys/atomic.h>
48 #include <sys/policy.h>
49 #include <sys/priv.h>
51 #include <sys/systm.h>
52 #include <sys/param.h>
53 #include <sys/kmem.h>
54 #include <sys/sdt.h>
55 #include <sys/socket.h>
56 #include <sys/vtrace.h>
57 #include <sys/isa_defs.h>
58 #include <sys/mac.h>
59 #include <net/if.h>
60 #include <net/if_arp.h>
61 #include <net/route.h>
62 #include <sys/sockio.h>
63 #include <netinet/in.h>
64 #include <net/if_dl.h>
66 #include <inet/common.h>
67 #include <inet/mi.h>
68 #include <inet/mib2.h>
69 #include <inet/nd.h>
70 #include <inet/arp.h>
71 #include <inet/snmpcom.h>
72 #include <inet/kstatcom.h>
74 #include <netinet/igmp_var.h>
75 #include <netinet/ip6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/sctp.h>
79 #include <inet/ip.h>
80 #include <inet/ip_impl.h>
81 #include <inet/ip6.h>
82 #include <inet/ip6_asp.h>
83 #include <inet/optcom.h>
84 #include <inet/tcp.h>
85 #include <inet/tcp_impl.h>
86 #include <inet/ip_multi.h>
87 #include <inet/ip_if.h>
88 #include <inet/ip_ire.h>
89 #include <inet/ip_ftable.h>
90 #include <inet/ip_rts.h>
91 #include <inet/ip_ndp.h>
92 #include <inet/ip_listutils.h>
93 #include <netinet/igmp.h>
94 #include <netinet/ip_mroute.h>
95 #include <inet/ipp_common.h>
97 #include <net/pfkeyv2.h>
98 #include <inet/sadb.h>
99 #include <inet/ipsec_impl.h>
100 #include <inet/ipdrop.h>
101 #include <inet/ip_netinfo.h>
102 #include <inet/ilb_ip.h>
103 #include <sys/squeue_impl.h>
104 #include <sys/squeue.h>
106 #include <sys/ethernet.h>
107 #include <net/if_types.h>
108 #include <sys/cpuvar.h>
110 #include <ipp/ipp.h>
111 #include <ipp/ipp_impl.h>
112 #include <ipp/ipgpc/ipgpc.h>
114 #include <sys/pattr.h>
115 #include <inet/ipclassifier.h>
116 #include <inet/sctp_ip.h>
117 #include <inet/sctp/sctp_impl.h>
118 #include <inet/udp_impl.h>
119 #include <sys/sunddi.h>
121 #include <sys/tsol/label.h>
122 #include <sys/tsol/tnet.h>
124 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
126 #ifdef DEBUG
127 extern boolean_t skip_sctp_cksum;
128 #endif
130 static void ip_input_local_v6(ire_t *, mblk_t *, ip6_t *, ip_recv_attr_t *);
132 static void ip_input_multicast_v6(ire_t *, mblk_t *, ip6_t *,
133 ip_recv_attr_t *);
135 #pragma inline(ip_input_common_v6, ip_input_local_v6, ip_forward_xmit_v6)
138 * Direct read side procedure capable of dealing with chains. GLDv3 based
139 * drivers call this function directly with mblk chains while STREAMS
140 * read side procedure ip_rput() calls this for single packet with ip_ring
141 * set to NULL to process one packet at a time.
143 * The ill will always be valid if this function is called directly from
144 * the driver.
146 * If ip_input_v6() is called from GLDv3:
148 * - This must be a non-VLAN IP stream.
149 * - 'mp' is either an untagged or a special priority-tagged packet.
150 * - Any VLAN tag that was in the MAC header has been stripped.
152 * If the IP header in packet is not 32-bit aligned, every message in the
153 * chain will be aligned before further operations. This is required on SPARC
154 * platform.
156 void
157 ip_input_v6(ill_t *ill, ill_rx_ring_t *ip_ring, mblk_t *mp_chain,
158 struct mac_header_info_s *mhip)
160 (void) ip_input_common_v6(ill, ip_ring, mp_chain, mhip, NULL, NULL,
161 NULL);
165 * ip_accept_tcp_v6() - This function is called by the squeue when it retrieves
166 * a chain of packets in the poll mode. The packets have gone through the
167 * data link processing but not IP processing. For performance and latency
168 * reasons, the squeue wants to process the chain in line instead of feeding
169 * it back via ip_input path.
171 * We set up the ip_recv_attr_t with IRAF_TARGET_SQP to that ip_fanout_v6
172 * will pass back any TCP packets matching the target sqp to
173 * ip_input_common_v6 using ira_target_sqp_mp. Other packets are handled by
174 * ip_input_v6 and ip_fanout_v6 as normal.
175 * The TCP packets that match the target squeue are returned to the caller
176 * as a b_next chain after each packet has been prepend with an mblk
177 * from ip_recv_attr_to_mblk.
179 mblk_t *
180 ip_accept_tcp_v6(ill_t *ill, ill_rx_ring_t *ip_ring, squeue_t *target_sqp,
181 mblk_t *mp_chain, mblk_t **last, uint_t *cnt)
183 return (ip_input_common_v6(ill, ip_ring, mp_chain, NULL, target_sqp,
184 last, cnt));
188 * Used by ip_input_v6 and ip_accept_tcp_v6
189 * The last three arguments are only used by ip_accept_tcp_v6, and mhip is
190 * only used by ip_input_v6.
192 mblk_t *
193 ip_input_common_v6(ill_t *ill, ill_rx_ring_t *ip_ring, mblk_t *mp_chain,
194 struct mac_header_info_s *mhip, squeue_t *target_sqp,
195 mblk_t **last, uint_t *cnt)
197 mblk_t *mp;
198 ip6_t *ip6h;
199 ip_recv_attr_t iras; /* Receive attributes */
200 rtc_t rtc;
201 iaflags_t chain_flags = 0; /* Fixed for chain */
202 mblk_t *ahead = NULL; /* Accepted head */
203 mblk_t *atail = NULL; /* Accepted tail */
204 uint_t acnt = 0; /* Accepted count */
206 ASSERT(mp_chain != NULL);
207 ASSERT(ill != NULL);
209 /* These ones do not change as we loop over packets */
210 iras.ira_ill = iras.ira_rill = ill;
211 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
212 iras.ira_rifindex = iras.ira_ruifindex;
213 iras.ira_sqp = NULL;
214 iras.ira_ring = ip_ring;
215 /* For ECMP and outbound transmit ring selection */
216 iras.ira_xmit_hint = ILL_RING_TO_XMIT_HINT(ip_ring);
218 iras.ira_target_sqp = target_sqp;
219 iras.ira_target_sqp_mp = NULL;
220 if (target_sqp != NULL)
221 chain_flags |= IRAF_TARGET_SQP;
224 * We try to have a mhip pointer when possible, but
225 * it might be NULL in some cases. In those cases we
226 * have to assume unicast.
228 iras.ira_mhip = mhip;
229 iras.ira_flags = 0;
230 if (mhip != NULL) {
231 switch (mhip->mhi_dsttype) {
232 case MAC_ADDRTYPE_MULTICAST :
233 chain_flags |= IRAF_L2DST_MULTICAST;
234 break;
235 case MAC_ADDRTYPE_BROADCAST :
236 chain_flags |= IRAF_L2DST_BROADCAST;
237 break;
242 * Initialize the one-element route cache.
244 * We do ire caching from one iteration to
245 * another. In the event the packet chain contains
246 * all packets from the same dst, this caching saves
247 * an ire_route_recursive for each of the succeeding
248 * packets in a packet chain.
250 rtc.rtc_ire = NULL;
251 rtc.rtc_ip6addr = ipv6_all_zeros;
253 /* Loop over b_next */
254 for (mp = mp_chain; mp != NULL; mp = mp_chain) {
255 mp_chain = mp->b_next;
256 mp->b_next = NULL;
259 * if db_ref > 1 then copymsg and free original. Packet
260 * may be changed and we do not want the other entity
261 * who has a reference to this message to trip over the
262 * changes. This is a blind change because trying to
263 * catch all places that might change the packet is too
264 * difficult.
266 * This corresponds to the fast path case, where we have
267 * a chain of M_DATA mblks. We check the db_ref count
268 * of only the 1st data block in the mblk chain. There
269 * doesn't seem to be a reason why a device driver would
270 * send up data with varying db_ref counts in the mblk
271 * chain. In any case the Fast path is a private
272 * interface, and our drivers don't do such a thing.
273 * Given the above assumption, there is no need to walk
274 * down the entire mblk chain (which could have a
275 * potential performance problem)
277 * The "(DB_REF(mp) > 1)" check was moved from ip_rput()
278 * to here because of exclusive ip stacks and vnics.
279 * Packets transmitted from exclusive stack over vnic
280 * can have db_ref > 1 and when it gets looped back to
281 * another vnic in a different zone, you have ip_input()
282 * getting dblks with db_ref > 1. So if someone
283 * complains of TCP performance under this scenario,
284 * take a serious look here on the impact of copymsg().
286 if (DB_REF(mp) > 1) {
287 if ((mp = ip_fix_dbref(mp, &iras)) == NULL)
288 continue;
292 * IP header ptr not aligned?
293 * OR IP header not complete in first mblk
295 ip6h = (ip6_t *)mp->b_rptr;
296 if (!OK_32PTR(ip6h) || MBLKL(mp) < IPV6_HDR_LEN) {
297 mp = ip_check_and_align_header(mp, IPV6_HDR_LEN, &iras);
298 if (mp == NULL)
299 continue;
300 ip6h = (ip6_t *)mp->b_rptr;
303 /* Protect against a mix of Ethertypes and IP versions */
304 if (IPH_HDR_VERSION(ip6h) != IPV6_VERSION) {
305 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
306 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
307 freemsg(mp);
308 /* mhip might point into 1st packet in the chain. */
309 iras.ira_mhip = NULL;
310 continue;
314 * Check for Martian addrs; we have to explicitly
315 * test for for zero dst since this is also used as
316 * an indication that the rtc is not used.
318 if (IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_dst)) {
319 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
320 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
321 freemsg(mp);
322 /* mhip might point into 1st packet in the chain. */
323 iras.ira_mhip = NULL;
324 continue;
327 * Keep L2SRC from a previous packet in chain since mhip
328 * might point into an earlier packet in the chain.
330 chain_flags |= (iras.ira_flags & IRAF_L2SRC_SET);
332 iras.ira_flags = IRAF_VERIFY_ULP_CKSUM | chain_flags;
333 iras.ira_free_flags = 0;
334 iras.ira_cred = NULL;
335 iras.ira_cpid = NOPID;
336 iras.ira_tsl = NULL;
337 iras.ira_zoneid = ALL_ZONES; /* Default for forwarding */
340 * We must count all incoming packets, even if they end
341 * up being dropped later on. Defer counting bytes until
342 * we have the whole IP header in first mblk.
344 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInReceives);
346 iras.ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
347 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCInOctets,
348 iras.ira_pktlen);
351 * Call one of:
352 * ill_input_full_v6
353 * ill_input_short_v6
354 * The former is used in the case of TX. See ill_set_inputfn().
356 (*ill->ill_inputfn)(mp, ip6h, &ip6h->ip6_dst, &iras, &rtc);
358 /* Any references to clean up? No hold on ira_ill */
359 if (iras.ira_flags & (IRAF_IPSEC_SECURE|IRAF_SYSTEM_LABELED))
360 ira_cleanup(&iras, B_FALSE);
362 if (iras.ira_target_sqp_mp != NULL) {
363 /* Better be called from ip_accept_tcp */
364 ASSERT(target_sqp != NULL);
366 /* Found one packet to accept */
367 mp = iras.ira_target_sqp_mp;
368 iras.ira_target_sqp_mp = NULL;
369 ASSERT(ip_recv_attr_is_mblk(mp));
371 if (atail != NULL)
372 atail->b_next = mp;
373 else
374 ahead = mp;
375 atail = mp;
376 acnt++;
377 mp = NULL;
379 /* mhip might point into 1st packet in the chain. */
380 iras.ira_mhip = NULL;
382 /* Any remaining references to the route cache? */
383 if (rtc.rtc_ire != NULL) {
384 ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&rtc.rtc_ip6addr));
385 ire_refrele(rtc.rtc_ire);
388 if (ahead != NULL) {
389 /* Better be called from ip_accept_tcp */
390 ASSERT(target_sqp != NULL);
391 *last = atail;
392 *cnt = acnt;
393 return (ahead);
396 return (NULL);
400 * This input function is used when
401 * - is_system_labeled()
403 * Note that for IPv6 CGTP filtering is handled only when receiving fragment
404 * headers, and RSVP uses router alert options, thus we don't need anything
405 * extra for them.
407 void
408 ill_input_full_v6(mblk_t *mp, void *iph_arg, void *nexthop_arg,
409 ip_recv_attr_t *ira, rtc_t *rtc)
411 ip6_t *ip6h = (ip6_t *)iph_arg;
412 in6_addr_t *nexthop = (in6_addr_t *)nexthop_arg;
413 ill_t *ill = ira->ira_ill;
415 ASSERT(ira->ira_tsl == NULL);
418 * Attach any necessary label information to
419 * this packet
421 if (is_system_labeled()) {
422 ira->ira_flags |= IRAF_SYSTEM_LABELED;
425 * This updates ira_cred, ira_tsl and ira_free_flags based
426 * on the label.
428 if (!tsol_get_pkt_label(mp, IPV6_VERSION, ira)) {
429 if (ip6opt_ls != 0)
430 ip0dbg(("tsol_get_pkt_label v6 failed\n"));
431 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
432 ip_drop_input("ipIfStatsInDiscards", mp, ill);
433 freemsg(mp);
434 return;
436 /* Note that ira_tsl can be NULL here. */
438 /* tsol_get_pkt_label sometimes does pullupmsg */
439 ip6h = (ip6_t *)mp->b_rptr;
441 ill_input_short_v6(mp, ip6h, nexthop, ira, rtc);
445 * Check for IPv6 addresses that should not appear on the wire
446 * as either source or destination.
447 * If we ever implement Stateless IPv6 Translators (SIIT) we'd have
448 * to revisit the IPv4-mapped part.
450 static boolean_t
451 ip6_bad_address(in6_addr_t *addr, boolean_t is_src)
453 if (IN6_IS_ADDR_V4MAPPED(addr)) {
454 ip1dbg(("ip_input_v6: pkt with IPv4-mapped addr"));
455 return (B_TRUE);
457 if (IN6_IS_ADDR_LOOPBACK(addr)) {
458 ip1dbg(("ip_input_v6: pkt with loopback addr"));
459 return (B_TRUE);
461 if (!is_src && IN6_IS_ADDR_UNSPECIFIED(addr)) {
463 * having :: in the src is ok: it's used for DAD.
465 ip1dbg(("ip_input_v6: pkt with unspecified addr"));
466 return (B_TRUE);
468 return (B_FALSE);
472 * Routing lookup for IPv6 link-locals.
473 * First we look on the inbound interface, then we check for IPMP and
474 * look on the upper interface.
475 * We update ira_ruifindex if we find the IRE on the upper interface.
477 static ire_t *
478 ire_linklocal(const in6_addr_t *nexthop, ill_t *ill, ip_recv_attr_t *ira,
479 uint_t irr_flags, ip_stack_t *ipst)
481 int match_flags = MATCH_IRE_SECATTR | MATCH_IRE_ILL;
482 ire_t *ire;
484 ASSERT(IN6_IS_ADDR_LINKLOCAL(nexthop));
485 ire = ire_route_recursive_v6(nexthop, 0, ill, ALL_ZONES, ira->ira_tsl,
486 match_flags, irr_flags, ira->ira_xmit_hint, ipst, NULL, NULL, NULL);
487 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
488 !IS_UNDER_IPMP(ill))
489 return (ire);
492 * When we are using IMP we need to look for an IRE on both the
493 * under and upper interfaces since there are different
494 * link-local addresses for the under and upper.
496 ill = ipmp_ill_hold_ipmp_ill(ill);
497 if (ill == NULL)
498 return (ire);
500 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
502 ire_refrele(ire);
503 ire = ire_route_recursive_v6(nexthop, 0, ill, ALL_ZONES, ira->ira_tsl,
504 match_flags, irr_flags, ira->ira_xmit_hint, ipst, NULL, NULL, NULL);
505 ill_refrele(ill);
506 return (ire);
510 * This is the tail-end of the full receive side packet handling.
511 * It can be used directly when the configuration is simple.
513 void
514 ill_input_short_v6(mblk_t *mp, void *iph_arg, void *nexthop_arg,
515 ip_recv_attr_t *ira, rtc_t *rtc)
517 ire_t *ire;
518 ill_t *ill = ira->ira_ill;
519 ip_stack_t *ipst = ill->ill_ipst;
520 uint_t pkt_len;
521 ssize_t len;
522 ip6_t *ip6h = (ip6_t *)iph_arg;
523 in6_addr_t nexthop = *(in6_addr_t *)nexthop_arg;
524 ilb_stack_t *ilbs = ipst->ips_netstack->netstack_ilb;
525 uint_t irr_flags;
526 #define rptr ((uchar_t *)ip6h)
528 ASSERT(DB_TYPE(mp) == M_DATA);
531 * Check for source/dest being a bad address: loopback, any, or
532 * v4mapped. All of them start with a 64 bits of zero.
534 if (ip6h->ip6_src.s6_addr32[0] == 0 &&
535 ip6h->ip6_src.s6_addr32[1] == 0) {
536 if (ip6_bad_address(&ip6h->ip6_src, B_TRUE)) {
537 ip1dbg(("ip_input_v6: pkt with bad src addr\n"));
538 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
539 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
540 freemsg(mp);
541 return;
544 if (ip6h->ip6_dst.s6_addr32[0] == 0 &&
545 ip6h->ip6_dst.s6_addr32[1] == 0) {
546 if (ip6_bad_address(&ip6h->ip6_dst, B_FALSE)) {
547 ip1dbg(("ip_input_v6: pkt with bad dst addr\n"));
548 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
549 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
550 freemsg(mp);
551 return;
555 len = mp->b_wptr - rptr;
556 pkt_len = ira->ira_pktlen;
558 /* multiple mblk or too short */
559 len -= pkt_len;
560 if (len != 0) {
561 mp = ip_check_length(mp, rptr, len, pkt_len, IPV6_HDR_LEN, ira);
562 if (mp == NULL)
563 return;
564 ip6h = (ip6_t *)mp->b_rptr;
567 DTRACE_IP7(receive, mblk_t *, mp, conn_t *, NULL, void_ip_t *,
568 ip6h, __dtrace_ipsr_ill_t *, ill, ipha_t *, NULL, ip6_t *, ip6h,
569 int, 0);
571 * The event for packets being received from a 'physical'
572 * interface is placed after validation of the source and/or
573 * destination address as being local so that packets can be
574 * redirected to loopback addresses using ipnat.
576 DTRACE_PROBE4(ip6__physical__in__start,
577 ill_t *, ill, ill_t *, NULL,
578 ip6_t *, ip6h, mblk_t *, mp);
580 if (HOOKS6_INTERESTED_PHYSICAL_IN(ipst)) {
581 int ll_multicast = 0;
582 int error;
583 in6_addr_t orig_dst = ip6h->ip6_dst;
585 if (ira->ira_flags & IRAF_L2DST_MULTICAST)
586 ll_multicast = HPE_MULTICAST;
587 else if (ira->ira_flags & IRAF_L2DST_BROADCAST)
588 ll_multicast = HPE_BROADCAST;
590 FW_HOOKS6(ipst->ips_ip6_physical_in_event,
591 ipst->ips_ipv6firewall_physical_in,
592 ill, NULL, ip6h, mp, mp, ll_multicast, ipst, error);
594 DTRACE_PROBE1(ip6__physical__in__end, mblk_t *, mp);
596 if (mp == NULL)
597 return;
599 /* The length could have changed */
600 ip6h = (ip6_t *)mp->b_rptr;
601 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
602 pkt_len = ira->ira_pktlen;
605 * In case the destination changed we override any previous
606 * change to nexthop.
608 if (!IN6_ARE_ADDR_EQUAL(&orig_dst, &ip6h->ip6_dst))
609 nexthop = ip6h->ip6_dst;
611 if (IN6_IS_ADDR_UNSPECIFIED(&nexthop)) {
612 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
613 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
614 freemsg(mp);
615 return;
620 if (ipst->ips_ip6_observe.he_interested) {
621 zoneid_t dzone;
624 * On the inbound path the src zone will be unknown as
625 * this packet has come from the wire.
627 dzone = ip_get_zoneid_v6(&nexthop, mp, ill, ira, ALL_ZONES);
628 ipobs_hook(mp, IPOBS_HOOK_INBOUND, ALL_ZONES, dzone, ill, ipst);
631 if ((ip6h->ip6_vcf & IPV6_VERS_AND_FLOW_MASK) !=
632 IPV6_DEFAULT_VERS_AND_FLOW) {
633 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
634 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
635 ip_drop_input("ipIfStatsInWrongIPVersion", mp, ill);
636 freemsg(mp);
637 return;
641 * For IPv6 we update ira_ip_hdr_length and ira_protocol as
642 * we parse the headers, starting with the hop-by-hop options header.
644 ira->ira_ip_hdr_length = IPV6_HDR_LEN;
645 if ((ira->ira_protocol = ip6h->ip6_nxt) == IPPROTO_HOPOPTS) {
646 ip6_hbh_t *hbhhdr;
647 uint_t ehdrlen;
648 uint8_t *optptr;
650 if (pkt_len < IPV6_HDR_LEN + MIN_EHDR_LEN) {
651 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
652 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
653 freemsg(mp);
654 return;
656 if (mp->b_cont != NULL &&
657 rptr + IPV6_HDR_LEN + MIN_EHDR_LEN > mp->b_wptr) {
658 ip6h = ip_pullup(mp, IPV6_HDR_LEN + MIN_EHDR_LEN, ira);
659 if (ip6h == NULL) {
660 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
661 ip_drop_input("ipIfStatsInDiscards", mp, ill);
662 freemsg(mp);
663 return;
666 hbhhdr = (ip6_hbh_t *)&ip6h[1];
667 ehdrlen = 8 * (hbhhdr->ip6h_len + 1);
669 if (pkt_len < IPV6_HDR_LEN + ehdrlen) {
670 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
671 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
672 freemsg(mp);
673 return;
675 if (mp->b_cont != NULL &&
676 rptr + IPV6_HDR_LEN + ehdrlen > mp->b_wptr) {
677 ip6h = ip_pullup(mp, IPV6_HDR_LEN + ehdrlen, ira);
678 if (ip6h == NULL) {
679 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
680 ip_drop_input("ipIfStatsInDiscards", mp, ill);
681 freemsg(mp);
682 return;
684 hbhhdr = (ip6_hbh_t *)&ip6h[1];
688 * Update ira_ip_hdr_length to skip the hop-by-hop header
689 * once we get to ip_fanout_v6
691 ira->ira_ip_hdr_length += ehdrlen;
692 ira->ira_protocol = hbhhdr->ip6h_nxt;
694 optptr = (uint8_t *)&hbhhdr[1];
695 switch (ip_process_options_v6(mp, ip6h, optptr,
696 ehdrlen - 2, IPPROTO_HOPOPTS, ira)) {
697 case -1:
699 * Packet has been consumed and any
700 * needed ICMP messages sent.
702 return;
703 case 0:
704 /* no action needed */
705 break;
706 case 1:
708 * Known router alert. Make use handle it as local
709 * by setting the nexthop to be the all-host multicast
710 * address, and skip multicast membership filter by
711 * marking as a router alert.
713 ira->ira_flags |= IRAF_ROUTER_ALERT;
714 nexthop = ipv6_all_hosts_mcast;
715 break;
720 * Here we check to see if we machine is setup as
721 * L3 loadbalancer and if the incoming packet is for a VIP
723 * Check the following:
724 * - there is at least a rule
725 * - protocol of the packet is supported
727 * We don't load balance IPv6 link-locals.
729 if (ilb_has_rules(ilbs) && ILB_SUPP_L4(ira->ira_protocol) &&
730 !IN6_IS_ADDR_LINKLOCAL(&nexthop)) {
731 in6_addr_t lb_dst;
732 int lb_ret;
734 /* For convenience, we just pull up the mblk. */
735 if (mp->b_cont != NULL) {
736 if (pullupmsg(mp, -1) == 0) {
737 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
738 ip_drop_input("ipIfStatsInDiscards - pullupmsg",
739 mp, ill);
740 freemsg(mp);
741 return;
743 ip6h = (ip6_t *)mp->b_rptr;
745 lb_ret = ilb_check_v6(ilbs, ill, mp, ip6h, ira->ira_protocol,
746 (uint8_t *)ip6h + ira->ira_ip_hdr_length, &lb_dst);
747 if (lb_ret == ILB_DROPPED) {
748 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
749 ip_drop_input("ILB_DROPPED", mp, ill);
750 freemsg(mp);
751 return;
753 if (lb_ret == ILB_BALANCED) {
754 /* Set the dst to that of the chosen server */
755 nexthop = lb_dst;
756 DB_CKSUMFLAGS(mp) = 0;
760 if (ill->ill_flags & ILLF_ROUTER)
761 irr_flags = IRR_ALLOCATE;
762 else
763 irr_flags = IRR_NONE;
765 /* Can not use route cache with TX since the labels can differ */
766 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
767 if (IN6_IS_ADDR_MULTICAST(&nexthop)) {
768 ire = ire_multicast(ill);
769 } else if (IN6_IS_ADDR_LINKLOCAL(&nexthop)) {
770 ire = ire_linklocal(&nexthop, ill, ira, irr_flags,
771 ipst);
772 } else {
773 /* Match destination and label */
774 ire = ire_route_recursive_v6(&nexthop, 0, NULL,
775 ALL_ZONES, ira->ira_tsl, MATCH_IRE_SECATTR,
776 irr_flags, ira->ira_xmit_hint, ipst, NULL, NULL,
777 NULL);
779 /* Update the route cache so we do the ire_refrele */
780 ASSERT(ire != NULL);
781 if (rtc->rtc_ire != NULL)
782 ire_refrele(rtc->rtc_ire);
783 rtc->rtc_ire = ire;
784 rtc->rtc_ip6addr = nexthop;
785 } else if (IN6_ARE_ADDR_EQUAL(&nexthop, &rtc->rtc_ip6addr) &&
786 rtc->rtc_ire != NULL) {
787 /* Use the route cache */
788 ire = rtc->rtc_ire;
789 } else {
790 /* Update the route cache */
791 if (IN6_IS_ADDR_MULTICAST(&nexthop)) {
792 ire = ire_multicast(ill);
793 } else if (IN6_IS_ADDR_LINKLOCAL(&nexthop)) {
794 ire = ire_linklocal(&nexthop, ill, ira, irr_flags,
795 ipst);
796 } else {
797 ire = ire_route_recursive_dstonly_v6(&nexthop,
798 irr_flags, ira->ira_xmit_hint, ipst);
800 ASSERT(ire != NULL);
801 if (rtc->rtc_ire != NULL)
802 ire_refrele(rtc->rtc_ire);
803 rtc->rtc_ire = ire;
804 rtc->rtc_ip6addr = nexthop;
807 ire->ire_ib_pkt_count++;
810 * Based on ire_type and ire_flags call one of:
811 * ire_recv_local_v6 - for IRE_LOCAL
812 * ire_recv_loopback_v6 - for IRE_LOOPBACK
813 * ire_recv_multirt_v6 - if RTF_MULTIRT
814 * ire_recv_noroute_v6 - if RTF_REJECT or RTF_BLACHOLE
815 * ire_recv_multicast_v6 - for IRE_MULTICAST
816 * ire_recv_noaccept_v6 - for ire_noaccept ones
817 * ire_recv_forward_v6 - for the rest.
820 (*ire->ire_recvfn)(ire, mp, ip6h, ira);
822 #undef rptr
825 * ire_recvfn for IREs that need forwarding
827 void
828 ire_recv_forward_v6(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
830 ip6_t *ip6h = (ip6_t *)iph_arg;
831 ill_t *ill = ira->ira_ill;
832 ip_stack_t *ipst = ill->ill_ipst;
833 iaflags_t iraflags = ira->ira_flags;
834 ill_t *dst_ill;
835 nce_t *nce;
836 uint32_t added_tx_len;
837 uint32_t mtu, iremtu;
839 if (iraflags & (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST)) {
840 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
841 ip_drop_input("l2 multicast not forwarded", mp, ill);
842 freemsg(mp);
843 return;
846 if (!(ill->ill_flags & ILLF_ROUTER)) {
847 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
848 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
849 freemsg(mp);
850 return;
854 * Either ire_nce_capable or ire_dep_parent would be set for the IRE
855 * when it is found by ire_route_recursive, but that some other thread
856 * could have changed the routes with the effect of clearing
857 * ire_dep_parent. In that case we'd end up dropping the packet, or
858 * finding a new nce below.
859 * Get, allocate, or update the nce.
860 * We get a refhold on ire_nce_cache as a result of this to avoid races
861 * where ire_nce_cache is deleted.
863 * This ensures that we don't forward if the interface is down since
864 * ipif_down removes all the nces.
866 mutex_enter(&ire->ire_lock);
867 nce = ire->ire_nce_cache;
868 if (nce == NULL) {
869 /* Not yet set up - try to set one up */
870 mutex_exit(&ire->ire_lock);
871 (void) ire_revalidate_nce(ire);
872 mutex_enter(&ire->ire_lock);
873 nce = ire->ire_nce_cache;
874 if (nce == NULL) {
875 mutex_exit(&ire->ire_lock);
876 /* The ire_dep_parent chain went bad, or no memory */
877 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
878 ip_drop_input("No ire_dep_parent", mp, ill);
879 freemsg(mp);
880 return;
883 nce_refhold(nce);
884 mutex_exit(&ire->ire_lock);
886 if (nce->nce_is_condemned) {
887 nce_t *nce1;
889 nce1 = ire_handle_condemned_nce(nce, ire, NULL, ip6h, B_FALSE);
890 nce_refrele(nce);
891 if (nce1 == NULL) {
892 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
893 ip_drop_input("No nce", mp, ill);
894 freemsg(mp);
895 return;
897 nce = nce1;
899 dst_ill = nce->nce_ill;
902 * Unless we are forwarding, drop the packet.
903 * Unlike IPv4 we don't allow source routed packets out the same
904 * interface when we are not a router.
905 * Note that ill_forward_set() will set the ILLF_ROUTER on
906 * all the group members when it gets an ipmp-ill or under-ill.
908 if (!(dst_ill->ill_flags & ILLF_ROUTER)) {
909 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
910 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
911 freemsg(mp);
912 nce_refrele(nce);
913 return;
916 if (ire->ire_zoneid != GLOBAL_ZONEID && ire->ire_zoneid != ALL_ZONES) {
917 ire->ire_ib_pkt_count--;
919 * Should only use IREs that are visible from the
920 * global zone for forwarding.
921 * For IPv6 any source route would have already been
922 * advanced in ip_fanout_v6
924 ire = ire_route_recursive_v6(&ip6h->ip6_dst, 0, NULL,
925 GLOBAL_ZONEID, ira->ira_tsl, MATCH_IRE_SECATTR,
926 (ill->ill_flags & ILLF_ROUTER) ? IRR_ALLOCATE : IRR_NONE,
927 ira->ira_xmit_hint, ipst, NULL, NULL, NULL);
928 ire->ire_ib_pkt_count++;
929 (*ire->ire_recvfn)(ire, mp, ip6h, ira);
930 ire_refrele(ire);
931 nce_refrele(nce);
932 return;
935 * ipIfStatsHCInForwDatagrams should only be increment if there
936 * will be an attempt to forward the packet, which is why we
937 * increment after the above condition has been checked.
939 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInForwDatagrams);
941 /* Initiate Read side IPPF processing */
942 if (IPP_ENABLED(IPP_FWD_IN, ipst)) {
943 /* ip_process translates an IS_UNDER_IPMP */
944 mp = ip_process(IPP_FWD_IN, mp, ill, ill);
945 if (mp == NULL) {
946 /* ip_drop_packet and MIB done */
947 ip2dbg(("ire_recv_forward_v6: pkt dropped/deferred "
948 "during IPPF processing\n"));
949 nce_refrele(nce);
950 return;
954 DTRACE_PROBE4(ip6__forwarding__start,
955 ill_t *, ill, ill_t *, dst_ill, ip6_t *, ip6h, mblk_t *, mp);
957 if (HOOKS6_INTERESTED_FORWARDING(ipst)) {
958 int error;
960 FW_HOOKS(ipst->ips_ip6_forwarding_event,
961 ipst->ips_ipv6firewall_forwarding,
962 ill, dst_ill, ip6h, mp, mp, 0, ipst, error);
964 DTRACE_PROBE1(ip6__forwarding__end, mblk_t *, mp);
966 if (mp == NULL) {
967 nce_refrele(nce);
968 return;
971 * Even if the destination was changed by the filter we use the
972 * forwarding decision that was made based on the address
973 * in ip_input.
976 /* Might have changed */
977 ip6h = (ip6_t *)mp->b_rptr;
978 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
981 /* Packet is being forwarded. Turning off hwcksum flag. */
982 DB_CKSUMFLAGS(mp) = 0;
985 * Per RFC 3513 section 2.5.2, we must not forward packets with
986 * an unspecified source address.
987 * The loopback address check for both src and dst has already
988 * been checked in ip_input_v6
989 * In the future one can envision adding RPF checks using number 3.
991 switch (ipst->ips_src_check) {
992 case 0:
993 break;
994 case 1:
995 case 2:
996 if (IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src) ||
997 IN6_IS_ADDR_MULTICAST(&ip6h->ip6_src)) {
998 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
999 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
1000 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
1001 nce_refrele(nce);
1002 freemsg(mp);
1003 return;
1005 break;
1009 * Check to see if we're forwarding the packet to a
1010 * different link from which it came. If so, check the
1011 * source and destination addresses since routers must not
1012 * forward any packets with link-local source or
1013 * destination addresses to other links. Otherwise (if
1014 * we're forwarding onto the same link), conditionally send
1015 * a redirect message.
1017 if (!IS_ON_SAME_LAN(dst_ill, ill)) {
1018 if (IN6_IS_ADDR_LINKLOCAL(&ip6h->ip6_dst) ||
1019 IN6_IS_ADDR_LINKLOCAL(&ip6h->ip6_src)) {
1020 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
1021 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
1022 freemsg(mp);
1023 nce_refrele(nce);
1024 return;
1026 /* TBD add site-local check at site boundary? */
1027 } else if (ipst->ips_ipv6_send_redirects) {
1028 ip_send_potential_redirect_v6(mp, ip6h, ire, ira);
1031 added_tx_len = 0;
1032 if (iraflags & IRAF_SYSTEM_LABELED) {
1033 mblk_t *mp1;
1034 uint32_t old_pkt_len = ira->ira_pktlen;
1037 * Check if it can be forwarded and add/remove
1038 * CIPSO options as needed.
1040 if ((mp1 = tsol_ip_forward(ire, mp, ira)) == NULL) {
1041 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1042 ip_drop_input("tsol_ip_forward", mp, ill);
1043 freemsg(mp);
1044 nce_refrele(nce);
1045 return;
1048 * Size may have changed. Remember amount added in case
1049 * ip_fragment needs to send an ICMP too big.
1051 mp = mp1;
1052 ip6h = (ip6_t *)mp->b_rptr;
1053 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
1054 ira->ira_ip_hdr_length = IPV6_HDR_LEN;
1055 if (ira->ira_pktlen > old_pkt_len)
1056 added_tx_len = ira->ira_pktlen - old_pkt_len;
1059 mtu = dst_ill->ill_mtu;
1060 if ((iremtu = ire->ire_metrics.iulp_mtu) != 0 && iremtu < mtu)
1061 mtu = iremtu;
1062 ip_forward_xmit_v6(nce, mp, ip6h, ira, mtu, added_tx_len);
1063 nce_refrele(nce);
1064 return;
1069 * Used for sending out unicast and multicast packets that are
1070 * forwarded.
1072 void
1073 ip_forward_xmit_v6(nce_t *nce, mblk_t *mp, ip6_t *ip6h, ip_recv_attr_t *ira,
1074 uint32_t mtu, uint32_t added_tx_len)
1076 ill_t *dst_ill = nce->nce_ill;
1077 uint32_t pkt_len;
1078 iaflags_t iraflags = ira->ira_flags;
1079 ip_stack_t *ipst = dst_ill->ill_ipst;
1081 if (ip6h->ip6_hops-- <= 1) {
1082 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
1083 ip_drop_input("ICMP6_TIME_EXCEED_TRANSIT", mp, ira->ira_ill);
1084 icmp_time_exceeded_v6(mp, ICMP6_TIME_EXCEED_TRANSIT, B_FALSE,
1085 ira);
1086 return;
1089 /* Initiate Write side IPPF processing before any fragmentation */
1090 if (IPP_ENABLED(IPP_FWD_OUT, ipst)) {
1091 /* ip_process translates an IS_UNDER_IPMP */
1092 mp = ip_process(IPP_FWD_OUT, mp, dst_ill, dst_ill);
1093 if (mp == NULL) {
1094 /* ip_drop_packet and MIB done */
1095 ip2dbg(("ire_recv_forward_v6: pkt dropped/deferred" \
1096 " during IPPF processing\n"));
1097 return;
1101 pkt_len = ira->ira_pktlen;
1103 BUMP_MIB(dst_ill->ill_ip_mib, ipIfStatsHCOutForwDatagrams);
1105 if (pkt_len > mtu) {
1106 BUMP_MIB(dst_ill->ill_ip_mib, ipIfStatsOutFragFails);
1107 ip_drop_output("ipIfStatsOutFragFails", mp, dst_ill);
1108 if (iraflags & IRAF_SYSTEM_LABELED) {
1110 * Remove any CIPSO option added by
1111 * tsol_ip_forward, and make sure we report
1112 * a path MTU so that there
1113 * is room to add such a CIPSO option for future
1114 * packets.
1116 mtu = tsol_pmtu_adjust(mp, mtu, added_tx_len, AF_INET6);
1118 icmp_pkt2big_v6(mp, mtu, B_TRUE, ira);
1119 return;
1122 ASSERT(pkt_len ==
1123 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
1125 if (iraflags & IRAF_LOOPBACK_COPY) {
1127 * IXAF_NO_LOOP_ZONEID is not set hence 6th arg
1128 * is don't care
1130 (void) ip_postfrag_loopcheck(mp, nce,
1131 (IXAF_LOOPBACK_COPY | IXAF_NO_DEV_FLOW_CTL),
1132 pkt_len, ira->ira_xmit_hint, GLOBAL_ZONEID, 0, NULL);
1133 } else {
1134 (void) ip_xmit(mp, nce, IXAF_NO_DEV_FLOW_CTL,
1135 pkt_len, ira->ira_xmit_hint, GLOBAL_ZONEID, 0, NULL);
1140 * ire_recvfn for RTF_REJECT and RTF_BLACKHOLE routes, including IRE_NOROUTE,
1141 * which is what ire_route_recursive returns when there is no matching ire.
1142 * Send ICMP unreachable unless blackhole.
1144 void
1145 ire_recv_noroute_v6(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1147 ip6_t *ip6h = (ip6_t *)iph_arg;
1148 ill_t *ill = ira->ira_ill;
1149 ip_stack_t *ipst = ill->ill_ipst;
1151 /* Would we have forwarded this packet if we had a route? */
1152 if (ira->ira_flags & (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST)) {
1153 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1154 ip_drop_input("l2 multicast not forwarded", mp, ill);
1155 freemsg(mp);
1156 return;
1159 if (!(ill->ill_flags & ILLF_ROUTER)) {
1160 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1161 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
1162 freemsg(mp);
1163 return;
1166 * If we had a route this could have been forwarded. Count as such.
1168 * ipIfStatsHCInForwDatagrams should only be increment if there
1169 * will be an attempt to forward the packet, which is why we
1170 * increment after the above condition has been checked.
1172 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInForwDatagrams);
1174 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInNoRoutes);
1176 ip_rts_change_v6(RTM_MISS, &ip6h->ip6_dst, 0, 0, 0, 0, 0, 0, RTA_DST,
1177 ipst);
1179 if (ire->ire_flags & RTF_BLACKHOLE) {
1180 ip_drop_input("ipIfStatsInNoRoutes RTF_BLACKHOLE", mp, ill);
1181 freemsg(mp);
1182 } else {
1183 ip_drop_input("ipIfStatsInNoRoutes RTF_REJECT", mp, ill);
1185 icmp_unreachable_v6(mp, ICMP6_DST_UNREACH_NOROUTE, B_FALSE,
1186 ira);
1191 * ire_recvfn for IRE_LOCALs marked with ire_noaccept. Such IREs are used for
1192 * VRRP when in noaccept mode.
1193 * We silently drop packets except for Neighbor Solicitations and
1194 * Neighbor Advertisements.
1196 void
1197 ire_recv_noaccept_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
1198 ip_recv_attr_t *ira)
1200 ip6_t *ip6h = (ip6_t *)iph_arg;
1201 ill_t *ill = ira->ira_ill;
1202 icmp6_t *icmp6;
1203 int ip_hdr_length;
1205 if (ip6h->ip6_nxt != IPPROTO_ICMPV6) {
1206 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1207 ip_drop_input("ipIfStatsInDiscards - noaccept", mp, ill);
1208 freemsg(mp);
1209 return;
1211 ip_hdr_length = ira->ira_ip_hdr_length;
1212 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMP6_MINLEN)) {
1213 if (ira->ira_pktlen < (ip_hdr_length + ICMP6_MINLEN)) {
1214 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1215 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1216 freemsg(mp);
1217 return;
1219 ip6h = ip_pullup(mp, ip_hdr_length + ICMP6_MINLEN, ira);
1220 if (ip6h == NULL) {
1221 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInErrors);
1222 freemsg(mp);
1223 return;
1226 icmp6 = (icmp6_t *)(&mp->b_rptr[ip_hdr_length]);
1228 if (icmp6->icmp6_type != ND_NEIGHBOR_SOLICIT &&
1229 icmp6->icmp6_type != ND_NEIGHBOR_ADVERT) {
1230 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1231 ip_drop_input("ipIfStatsInDiscards - noaccept", mp, ill);
1232 freemsg(mp);
1233 return;
1235 ire_recv_local_v6(ire, mp, ip6h, ira);
1239 * ire_recvfn for IRE_MULTICAST.
1241 void
1242 ire_recv_multicast_v6(ire_t *ire, mblk_t *mp, void *iph_arg,
1243 ip_recv_attr_t *ira)
1245 ip6_t *ip6h = (ip6_t *)iph_arg;
1246 ill_t *ill = ira->ira_ill;
1248 ASSERT(ire->ire_ill == ira->ira_ill);
1250 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInMcastPkts);
1251 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCInMcastOctets, ira->ira_pktlen);
1253 /* Tag for higher-level protocols */
1254 ira->ira_flags |= IRAF_MULTICAST;
1257 * So that we don't end up with dups, only one ill an IPMP group is
1258 * nominated to receive multicast traffic.
1259 * If we have no cast_ill we are liberal and accept everything.
1261 if (IS_UNDER_IPMP(ill)) {
1262 ip_stack_t *ipst = ill->ill_ipst;
1264 /* For an under ill_grp can change under lock */
1265 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
1266 if (!ill->ill_nom_cast && ill->ill_grp != NULL &&
1267 ill->ill_grp->ig_cast_ill != NULL) {
1268 rw_exit(&ipst->ips_ill_g_lock);
1269 ip_drop_input("not on cast ill", mp, ill);
1270 freemsg(mp);
1271 return;
1273 rw_exit(&ipst->ips_ill_g_lock);
1275 * We switch to the upper ill so that mrouter and hasmembers
1276 * can operate on upper here and in ip_input_multicast.
1278 ill = ipmp_ill_hold_ipmp_ill(ill);
1279 if (ill != NULL) {
1280 ASSERT(ill != ira->ira_ill);
1281 ASSERT(ire->ire_ill == ira->ira_ill);
1282 ira->ira_ill = ill;
1283 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1284 } else {
1285 ill = ira->ira_ill;
1289 #ifdef notdef
1291 * Check if we are a multicast router - send ip_mforward a copy of
1292 * the packet.
1293 * Due to mroute_decap tunnels we consider forwarding packets even if
1294 * mrouted has not joined the allmulti group on this interface.
1296 if (ipst->ips_ip_g_mrouter) {
1297 int retval;
1300 * Clear the indication that this may have hardware
1301 * checksum as we are not using it for forwarding.
1303 DB_CKSUMFLAGS(mp) = 0;
1306 * ip_mforward helps us make these distinctions: If received
1307 * on tunnel and not IGMP, then drop.
1308 * If IGMP packet, then don't check membership
1309 * If received on a phyint and IGMP or PIM, then
1310 * don't check membership
1312 retval = ip_mforward_v6(mp, ira);
1313 /* ip_mforward updates mib variables if needed */
1315 switch (retval) {
1316 case 0:
1318 * pkt is okay and arrived on phyint.
1320 break;
1321 case -1:
1322 /* pkt is mal-formed, toss it */
1323 freemsg(mp);
1324 goto done;
1325 case 1:
1327 * pkt is okay and arrived on a tunnel
1329 * If we are running a multicast router
1330 * we need to see all mld packets, which
1331 * are marked with router alerts.
1333 if (ira->ira_flags & IRAF_ROUTER_ALERT)
1334 goto forus;
1335 ip_drop_input("Multicast on tunnel ignored", mp, ill);
1336 freemsg(mp);
1337 goto done;
1340 #endif /* notdef */
1343 * If this was a router alert we skip the group membership check.
1345 if (ira->ira_flags & IRAF_ROUTER_ALERT)
1346 goto forus;
1349 * Check if we have members on this ill. This is not necessary for
1350 * correctness because even if the NIC/GLD had a leaky filter, we
1351 * filter before passing to each conn_t.
1353 if (!ill_hasmembers_v6(ill, &ip6h->ip6_dst)) {
1355 * Nobody interested
1357 * This might just be caused by the fact that
1358 * multiple IP Multicast addresses map to the same
1359 * link layer multicast - no need to increment counter!
1361 ip_drop_input("Multicast with no members", mp, ill);
1362 freemsg(mp);
1363 goto done;
1365 forus:
1366 ip2dbg(("ire_recv_multicast_v6: multicast for us\n"));
1369 * After reassembly and IPsec we will need to duplicate the
1370 * multicast packet for all matching zones on the ill.
1372 ira->ira_zoneid = ALL_ZONES;
1374 /* Reassemble on the ill on which the packet arrived */
1375 ip_input_local_v6(ire, mp, ip6h, ira);
1376 done:
1377 if (ill != ire->ire_ill) {
1378 ill_refrele(ill);
1379 ira->ira_ill = ire->ire_ill;
1380 ira->ira_ruifindex = ira->ira_ill->ill_phyint->phyint_ifindex;
1385 * ire_recvfn for IRE_OFFLINK with RTF_MULTIRT.
1386 * Drop packets since we don't forward out multirt routes.
1388 /* ARGSUSED */
1389 void
1390 ire_recv_multirt_v6(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1392 ill_t *ill = ira->ira_ill;
1394 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInNoRoutes);
1395 ip_drop_input("Not forwarding out MULTIRT", mp, ill);
1396 freemsg(mp);
1400 * ire_recvfn for IRE_LOOPBACK. This is only used when a FW_HOOK
1401 * has rewritten the packet to have a loopback destination address (We
1402 * filter out packet with a loopback destination from arriving over the wire).
1403 * We don't know what zone to use, thus we always use the GLOBAL_ZONEID.
1405 void
1406 ire_recv_loopback_v6(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1408 ip6_t *ip6h = (ip6_t *)iph_arg;
1409 ill_t *ill = ira->ira_ill;
1410 ill_t *ire_ill = ire->ire_ill;
1412 ira->ira_zoneid = GLOBAL_ZONEID;
1414 /* Switch to the lo0 ill for further processing */
1415 if (ire_ill != ill) {
1417 * Update ira_ill to be the ILL on which the IP address
1418 * is hosted.
1419 * No need to hold the ill since we have a hold on the ire
1421 ASSERT(ira->ira_ill == ira->ira_rill);
1422 ira->ira_ill = ire_ill;
1424 ip_input_local_v6(ire, mp, ip6h, ira);
1426 /* Restore */
1427 ASSERT(ira->ira_ill == ire_ill);
1428 ira->ira_ill = ill;
1429 return;
1432 ip_input_local_v6(ire, mp, ip6h, ira);
1436 * ire_recvfn for IRE_LOCAL.
1438 void
1439 ire_recv_local_v6(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1441 ip6_t *ip6h = (ip6_t *)iph_arg;
1442 ill_t *ill = ira->ira_ill;
1443 ill_t *ire_ill = ire->ire_ill;
1445 /* Make a note for DAD that this address is in use */
1446 ire->ire_last_used_time = LBOLT_FASTPATH;
1448 /* Only target the IRE_LOCAL with the right zoneid. */
1449 ira->ira_zoneid = ire->ire_zoneid;
1452 * If the packet arrived on the wrong ill, we check that
1453 * this is ok.
1454 * If it is, then we ensure that we do the reassembly on
1455 * the ill on which the address is hosted. We keep ira_rill as
1456 * the one on which the packet arrived, so that IP_PKTINFO and
1457 * friends can report this.
1459 if (ire_ill != ill) {
1460 ire_t *new_ire;
1462 new_ire = ip_check_multihome(&ip6h->ip6_dst, ire, ill);
1463 if (new_ire == NULL) {
1464 /* Drop packet */
1465 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1466 ip_drop_input("ipIfStatsInForwProhibits", mp, ill);
1467 freemsg(mp);
1468 return;
1471 * Update ira_ill to be the ILL on which the IP address
1472 * is hosted. No need to hold the ill since we have a
1473 * hold on the ire. Note that we do the switch even if
1474 * new_ire == ire (for IPMP, ire would be the one corresponding
1475 * to the IPMP ill).
1477 ASSERT(ira->ira_ill == ira->ira_rill);
1478 ira->ira_ill = new_ire->ire_ill;
1480 /* ira_ruifindex tracks the upper for ira_rill */
1481 if (IS_UNDER_IPMP(ill))
1482 ira->ira_ruifindex = ill_get_upper_ifindex(ill);
1484 ip_input_local_v6(new_ire, mp, ip6h, ira);
1486 /* Restore */
1487 ASSERT(ira->ira_ill == new_ire->ire_ill);
1488 ira->ira_ill = ill;
1489 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1491 if (new_ire != ire)
1492 ire_refrele(new_ire);
1493 return;
1496 ip_input_local_v6(ire, mp, ip6h, ira);
1500 * Common function for packets arriving for the host. Handles
1501 * checksum verification, reassembly checks, etc.
1503 static void
1504 ip_input_local_v6(ire_t *ire, mblk_t *mp, ip6_t *ip6h, ip_recv_attr_t *ira)
1506 iaflags_t iraflags = ira->ira_flags;
1509 * For multicast we need some extra work before
1510 * we call ip_fanout_v6(), since in the case of shared-IP zones
1511 * we need to pretend that a packet arrived for each zoneid.
1513 if (iraflags & IRAF_MULTICAST) {
1514 ip_input_multicast_v6(ire, mp, ip6h, ira);
1515 return;
1517 ip_fanout_v6(mp, ip6h, ira);
1521 * Handle multiple zones which want to receive the same multicast packets
1522 * on this ill by delivering a packet to each of them.
1524 * Note that for packets delivered to transports we could instead do this
1525 * as part of the fanout code, but since we need to handle icmp_inbound
1526 * it is simpler to have multicast work the same as IPv4 broadcast.
1528 * The ip_fanout matching for multicast matches based on ilm independent of
1529 * zoneid since the zoneid restriction is applied when joining a multicast
1530 * group.
1532 /* ARGSUSED */
1533 static void
1534 ip_input_multicast_v6(ire_t *ire, mblk_t *mp, ip6_t *ip6h, ip_recv_attr_t *ira)
1536 ill_t *ill = ira->ira_ill;
1537 iaflags_t iraflags = ira->ira_flags;
1538 ip_stack_t *ipst = ill->ill_ipst;
1539 netstack_t *ns = ipst->ips_netstack;
1540 zoneid_t zoneid;
1541 mblk_t *mp1;
1542 ip6_t *ip6h1;
1543 uint_t ira_pktlen = ira->ira_pktlen;
1544 uint16_t ira_ip_hdr_length = ira->ira_ip_hdr_length;
1546 /* ire_recv_multicast has switched to the upper ill for IPMP */
1547 ASSERT(!IS_UNDER_IPMP(ill));
1550 * If we don't have more than one shared-IP zone, or if
1551 * there are no members in anything but the global zone,
1552 * then just set the zoneid and proceed.
1554 if (ns->netstack_numzones == 1 ||
1555 !ill_hasmembers_otherzones_v6(ill, &ip6h->ip6_dst,
1556 GLOBAL_ZONEID)) {
1557 ira->ira_zoneid = GLOBAL_ZONEID;
1559 /* If sender didn't want this zone to receive it, drop */
1560 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1561 ira->ira_no_loop_zoneid == ira->ira_zoneid) {
1562 ip_drop_input("Multicast but wrong zoneid", mp, ill);
1563 freemsg(mp);
1564 return;
1566 ip_fanout_v6(mp, ip6h, ira);
1567 return;
1571 * Here we loop over all zoneids that have members in the group
1572 * and deliver a packet to ip_fanout for each zoneid.
1574 * First find any members in the lowest numeric zoneid by looking for
1575 * first zoneid larger than -1 (ALL_ZONES).
1576 * We terminate the loop when we receive -1 (ALL_ZONES).
1578 zoneid = ill_hasmembers_nextzone_v6(ill, &ip6h->ip6_dst, ALL_ZONES);
1579 for (; zoneid != ALL_ZONES;
1580 zoneid = ill_hasmembers_nextzone_v6(ill, &ip6h->ip6_dst, zoneid)) {
1582 * Avoid an extra copymsg/freemsg by skipping global zone here
1583 * and doing that at the end.
1585 if (zoneid == GLOBAL_ZONEID)
1586 continue;
1588 ira->ira_zoneid = zoneid;
1590 /* If sender didn't want this zone to receive it, skip */
1591 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1592 ira->ira_no_loop_zoneid == ira->ira_zoneid)
1593 continue;
1595 mp1 = copymsg(mp);
1596 if (mp1 == NULL) {
1597 /* Failed to deliver to one zone */
1598 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1599 ip_drop_input("ipIfStatsInDiscards", mp, ill);
1600 continue;
1602 ip6h1 = (ip6_t *)mp1->b_rptr;
1603 ip_fanout_v6(mp1, ip6h1, ira);
1605 * IPsec might have modified ira_pktlen and ira_ip_hdr_length
1606 * so we restore them for a potential next iteration
1608 ira->ira_pktlen = ira_pktlen;
1609 ira->ira_ip_hdr_length = ira_ip_hdr_length;
1612 /* Do the main ire */
1613 ira->ira_zoneid = GLOBAL_ZONEID;
1614 /* If sender didn't want this zone to receive it, drop */
1615 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1616 ira->ira_no_loop_zoneid == ira->ira_zoneid) {
1617 ip_drop_input("Multicast but wrong zoneid", mp, ill);
1618 freemsg(mp);
1619 } else {
1620 ip_fanout_v6(mp, ip6h, ira);
1626 * Determine the zoneid and IRAF_TX_MAC_EXEMPTABLE if trusted extensions
1627 * is in use. Updates ira_zoneid and ira_flags as a result.
1629 static void
1630 ip_fanout_tx_v6(mblk_t *mp, ip6_t *ip6h, uint8_t protocol, uint_t ip_hdr_length,
1631 ip_recv_attr_t *ira)
1633 uint16_t *up;
1634 uint16_t lport;
1635 zoneid_t zoneid;
1637 ASSERT(ira->ira_flags & IRAF_SYSTEM_LABELED);
1640 * If the packet is unlabeled we might allow read-down
1641 * for MAC_EXEMPT. Below we clear this if it is a multi-level
1642 * port (MLP).
1643 * Note that ira_tsl can be NULL here.
1645 if (ira->ira_tsl != NULL && ira->ira_tsl->tsl_flags & TSLF_UNLABELED)
1646 ira->ira_flags |= IRAF_TX_MAC_EXEMPTABLE;
1648 if (ira->ira_zoneid != ALL_ZONES)
1649 return;
1651 ira->ira_flags |= IRAF_TX_SHARED_ADDR;
1653 up = (uint16_t *)((uchar_t *)ip6h + ip_hdr_length);
1654 switch (protocol) {
1655 case IPPROTO_TCP:
1656 case IPPROTO_SCTP:
1657 case IPPROTO_UDP:
1658 /* Caller ensures this */
1659 ASSERT(((uchar_t *)ip6h) + ip_hdr_length +4 <= mp->b_wptr);
1662 * Only these transports support MLP.
1663 * We know their destination port numbers is in
1664 * the same place in the header.
1666 lport = up[1];
1669 * No need to handle exclusive-stack zones
1670 * since ALL_ZONES only applies to the shared IP instance.
1672 zoneid = tsol_mlp_findzone(protocol, lport);
1674 * If no shared MLP is found, tsol_mlp_findzone returns
1675 * ALL_ZONES. In that case, we assume it's SLP, and
1676 * search for the zone based on the packet label.
1678 * If there is such a zone, we prefer to find a
1679 * connection in it. Otherwise, we look for a
1680 * MAC-exempt connection in any zone whose label
1681 * dominates the default label on the packet.
1683 if (zoneid == ALL_ZONES)
1684 zoneid = tsol_attr_to_zoneid(ira);
1685 else
1686 ira->ira_flags &= ~IRAF_TX_MAC_EXEMPTABLE;
1687 break;
1688 default:
1689 /* Handle shared address for other protocols */
1690 zoneid = tsol_attr_to_zoneid(ira);
1691 break;
1693 ira->ira_zoneid = zoneid;
1697 * Increment checksum failure statistics
1699 static void
1700 ip_input_cksum_err_v6(uint8_t protocol, uint16_t hck_flags, ill_t *ill)
1702 ip_stack_t *ipst = ill->ill_ipst;
1704 switch (protocol) {
1705 case IPPROTO_TCP:
1706 BUMP_MIB(ill->ill_ip_mib, tcpIfStatsInErrs);
1708 if (hck_flags & HCK_FULLCKSUM)
1709 IP6_STAT(ipst, ip6_tcp_in_full_hw_cksum_err);
1710 else if (hck_flags & HCK_PARTIALCKSUM)
1711 IP6_STAT(ipst, ip6_tcp_in_part_hw_cksum_err);
1712 else
1713 IP6_STAT(ipst, ip6_tcp_in_sw_cksum_err);
1714 break;
1715 case IPPROTO_UDP:
1716 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInCksumErrs);
1717 if (hck_flags & HCK_FULLCKSUM)
1718 IP6_STAT(ipst, ip6_udp_in_full_hw_cksum_err);
1719 else if (hck_flags & HCK_PARTIALCKSUM)
1720 IP6_STAT(ipst, ip6_udp_in_part_hw_cksum_err);
1721 else
1722 IP6_STAT(ipst, ip6_udp_in_sw_cksum_err);
1723 break;
1724 case IPPROTO_ICMPV6:
1725 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInMsgs);
1726 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInErrors);
1727 break;
1728 default:
1729 ASSERT(0);
1730 break;
1734 /* Calculate the IPv6 pseudo-header checksum for TCP, UDP, and ICMPV6 */
1735 uint32_t
1736 ip_input_cksum_pseudo_v6(ip6_t *ip6h, ip_recv_attr_t *ira)
1738 uint_t ulp_len;
1739 uint32_t cksum;
1740 uint8_t protocol = ira->ira_protocol;
1741 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
1743 #define iphs ((uint16_t *)ip6h)
1745 switch (protocol) {
1746 case IPPROTO_TCP:
1747 ulp_len = ira->ira_pktlen - ip_hdr_length;
1749 /* Protocol and length */
1750 cksum = htons(ulp_len) + IP_TCP_CSUM_COMP;
1751 /* IP addresses */
1752 cksum += iphs[4] + iphs[5] + iphs[6] + iphs[7] +
1753 iphs[8] + iphs[9] + iphs[10] + iphs[11] +
1754 iphs[12] + iphs[13] + iphs[14] + iphs[15] +
1755 iphs[16] + iphs[17] + iphs[18] + iphs[19];
1756 break;
1758 case IPPROTO_UDP: {
1759 udpha_t *udpha;
1761 udpha = (udpha_t *)((uchar_t *)ip6h + ip_hdr_length);
1763 /* Protocol and length */
1764 cksum = udpha->uha_length + IP_UDP_CSUM_COMP;
1765 /* IP addresses */
1766 cksum += iphs[4] + iphs[5] + iphs[6] + iphs[7] +
1767 iphs[8] + iphs[9] + iphs[10] + iphs[11] +
1768 iphs[12] + iphs[13] + iphs[14] + iphs[15] +
1769 iphs[16] + iphs[17] + iphs[18] + iphs[19];
1770 break;
1772 case IPPROTO_ICMPV6:
1773 ulp_len = ira->ira_pktlen - ip_hdr_length;
1775 /* Protocol and length */
1776 cksum = htons(ulp_len) + IP_ICMPV6_CSUM_COMP;
1777 /* IP addresses */
1778 cksum += iphs[4] + iphs[5] + iphs[6] + iphs[7] +
1779 iphs[8] + iphs[9] + iphs[10] + iphs[11] +
1780 iphs[12] + iphs[13] + iphs[14] + iphs[15] +
1781 iphs[16] + iphs[17] + iphs[18] + iphs[19];
1782 break;
1783 default:
1784 cksum = 0;
1785 break;
1787 #undef iphs
1788 return (cksum);
1793 * Software verification of the ULP checksums.
1794 * Returns B_TRUE if ok.
1795 * Increments statistics of failed.
1797 static boolean_t
1798 ip_input_sw_cksum_v6(mblk_t *mp, ip6_t *ip6h, ip_recv_attr_t *ira)
1800 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1801 uint32_t cksum;
1802 uint8_t protocol = ira->ira_protocol;
1803 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
1805 IP6_STAT(ipst, ip6_in_sw_cksum);
1807 ASSERT(protocol == IPPROTO_TCP || protocol == IPPROTO_UDP ||
1808 protocol == IPPROTO_ICMPV6);
1810 cksum = ip_input_cksum_pseudo_v6(ip6h, ira);
1811 cksum = IP_CSUM(mp, ip_hdr_length, cksum);
1812 if (cksum == 0)
1813 return (B_TRUE);
1815 ip_input_cksum_err_v6(protocol, 0, ira->ira_ill);
1816 return (B_FALSE);
1820 * Verify the ULP checksums.
1821 * Returns B_TRUE if ok, or if the ULP doesn't have a well-defined checksum
1822 * algorithm.
1823 * Increments statistics if failed.
1825 static boolean_t
1826 ip_input_cksum_v6(iaflags_t iraflags, mblk_t *mp, ip6_t *ip6h,
1827 ip_recv_attr_t *ira)
1829 ill_t *ill = ira->ira_rill;
1830 uint16_t hck_flags;
1831 uint32_t cksum;
1832 mblk_t *mp1;
1833 uint_t len;
1834 uint8_t protocol = ira->ira_protocol;
1835 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
1838 switch (protocol) {
1839 case IPPROTO_TCP:
1840 case IPPROTO_ICMPV6:
1841 break;
1843 case IPPROTO_UDP: {
1844 udpha_t *udpha;
1846 udpha = (udpha_t *)((uchar_t *)ip6h + ip_hdr_length);
1848 * Before going through the regular checksum
1849 * calculation, make sure the received checksum
1850 * is non-zero. RFC 2460 says, a 0x0000 checksum
1851 * in a UDP packet (within IPv6 packet) is invalid
1852 * and should be replaced by 0xffff. This makes
1853 * sense as regular checksum calculation will
1854 * pass for both the cases i.e. 0x0000 and 0xffff.
1855 * Removing one of the case makes error detection
1856 * stronger.
1858 if (udpha->uha_checksum == 0) {
1859 /* 0x0000 checksum is invalid */
1860 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInCksumErrs);
1861 return (B_FALSE);
1863 break;
1865 case IPPROTO_SCTP: {
1866 sctp_hdr_t *sctph;
1867 uint32_t pktsum;
1869 sctph = (sctp_hdr_t *)((uchar_t *)ip6h + ip_hdr_length);
1870 #ifdef DEBUG
1871 if (skip_sctp_cksum)
1872 return (B_TRUE);
1873 #endif
1874 pktsum = sctph->sh_chksum;
1875 sctph->sh_chksum = 0;
1876 cksum = sctp_cksum(mp, ip_hdr_length);
1877 sctph->sh_chksum = pktsum;
1878 if (cksum == pktsum)
1879 return (B_TRUE);
1882 * Defer until later whether a bad checksum is ok
1883 * in order to allow RAW sockets to use Adler checksum
1884 * with SCTP.
1886 ira->ira_flags |= IRAF_SCTP_CSUM_ERR;
1887 return (B_TRUE);
1890 default:
1891 /* No ULP checksum to verify. */
1892 return (B_TRUE);
1896 * Revert to software checksum calculation if the interface
1897 * isn't capable of checksum offload.
1898 * We clear DB_CKSUMFLAGS when going through IPsec in ip_fanout.
1899 * Note: IRAF_NO_HW_CKSUM is not currently used.
1901 ASSERT(!IS_IPMP(ill));
1902 if ((iraflags & IRAF_NO_HW_CKSUM) || !ILL_HCKSUM_CAPABLE(ill) ||
1903 !dohwcksum) {
1904 return (ip_input_sw_cksum_v6(mp, ip6h, ira));
1908 * We apply this for all ULP protocols. Does the HW know to
1909 * not set the flags for SCTP and other protocols.
1912 hck_flags = DB_CKSUMFLAGS(mp);
1914 if (hck_flags & HCK_FULLCKSUM_OK) {
1916 * Hardware has already verified the checksum.
1918 return (B_TRUE);
1921 if (hck_flags & HCK_FULLCKSUM) {
1923 * Full checksum has been computed by the hardware
1924 * and has been attached. If the driver wants us to
1925 * verify the correctness of the attached value, in
1926 * order to protect against faulty hardware, compare
1927 * it against -0 (0xFFFF) to see if it's valid.
1929 cksum = DB_CKSUM16(mp);
1930 if (cksum == 0xFFFF)
1931 return (B_TRUE);
1932 ip_input_cksum_err_v6(protocol, hck_flags, ira->ira_ill);
1933 return (B_FALSE);
1936 mp1 = mp->b_cont;
1937 if ((hck_flags & HCK_PARTIALCKSUM) &&
1938 (mp1 == NULL || mp1->b_cont == NULL) &&
1939 ip_hdr_length >= DB_CKSUMSTART(mp) &&
1940 ((len = ip_hdr_length - DB_CKSUMSTART(mp)) & 1) == 0) {
1941 uint32_t adj;
1942 uchar_t *cksum_start;
1944 cksum = ip_input_cksum_pseudo_v6(ip6h, ira);
1946 cksum_start = ((uchar_t *)ip6h + DB_CKSUMSTART(mp));
1949 * Partial checksum has been calculated by hardware
1950 * and attached to the packet; in addition, any
1951 * prepended extraneous data is even byte aligned,
1952 * and there are at most two mblks associated with
1953 * the packet. If any such data exists, we adjust
1954 * the checksum; also take care any postpended data.
1956 IP_ADJCKSUM_PARTIAL(cksum_start, mp, mp1, len, adj);
1958 * One's complement subtract extraneous checksum
1960 cksum += DB_CKSUM16(mp);
1961 if (adj >= cksum)
1962 cksum = ~(adj - cksum) & 0xFFFF;
1963 else
1964 cksum -= adj;
1965 cksum = (cksum & 0xFFFF) + ((int)cksum >> 16);
1966 cksum = (cksum & 0xFFFF) + ((int)cksum >> 16);
1967 if (!(~cksum & 0xFFFF))
1968 return (B_TRUE);
1970 ip_input_cksum_err_v6(protocol, hck_flags, ira->ira_ill);
1971 return (B_FALSE);
1973 return (ip_input_sw_cksum_v6(mp, ip6h, ira));
1978 * Handle fanout of received packets.
1979 * Unicast packets that are looped back (from ire_send_local_v6) and packets
1980 * from the wire are differentiated by checking IRAF_VERIFY_ULP_CKSUM.
1982 * IPQoS Notes
1983 * Before sending it to the client, invoke IPPF processing. Policy processing
1984 * takes place only if the callout_position, IPP_LOCAL_IN, is enabled.
1986 void
1987 ip_fanout_v6(mblk_t *mp, ip6_t *ip6h, ip_recv_attr_t *ira)
1989 ill_t *ill = ira->ira_ill;
1990 iaflags_t iraflags = ira->ira_flags;
1991 ip_stack_t *ipst = ill->ill_ipst;
1992 uint8_t protocol;
1993 conn_t *connp;
1994 #define rptr ((uchar_t *)ip6h)
1995 uint_t ip_hdr_length;
1996 uint_t min_ulp_header_length;
1997 int offset;
1998 ssize_t len;
1999 netstack_t *ns = ipst->ips_netstack;
2000 ipsec_stack_t *ipss = ns->netstack_ipsec;
2001 ill_t *rill = ira->ira_rill;
2003 ASSERT(ira->ira_pktlen == ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN);
2006 * We repeat this as we parse over destination options header and
2007 * fragment headers (earlier we've handled any hop-by-hop options
2008 * header.)
2009 * We update ira_protocol and ira_ip_hdr_length as we skip past
2010 * the intermediate headers; they already point past any
2011 * hop-by-hop header.
2013 repeat:
2014 protocol = ira->ira_protocol;
2015 ip_hdr_length = ira->ira_ip_hdr_length;
2018 * Time for IPP once we've done reassembly and IPsec.
2019 * We skip this for loopback packets since we don't do IPQoS
2020 * on loopback.
2022 if (IPP_ENABLED(IPP_LOCAL_IN, ipst) &&
2023 !(iraflags & IRAF_LOOPBACK) &&
2024 (protocol != IPPROTO_ESP || protocol != IPPROTO_AH ||
2025 protocol != IPPROTO_DSTOPTS || protocol != IPPROTO_ROUTING ||
2026 protocol != IPPROTO_FRAGMENT)) {
2028 * Use the interface on which the packet arrived - not where
2029 * the IP address is hosted.
2031 /* ip_process translates an IS_UNDER_IPMP */
2032 mp = ip_process(IPP_LOCAL_IN, mp, rill, ill);
2033 if (mp == NULL) {
2034 /* ip_drop_packet and MIB done */
2035 return;
2039 /* Determine the minimum required size of the upper-layer header */
2040 /* Need to do this for at least the set of ULPs that TX handles. */
2041 switch (protocol) {
2042 case IPPROTO_TCP:
2043 min_ulp_header_length = TCP_MIN_HEADER_LENGTH;
2044 break;
2045 case IPPROTO_SCTP:
2046 min_ulp_header_length = SCTP_COMMON_HDR_LENGTH;
2047 break;
2048 case IPPROTO_UDP:
2049 min_ulp_header_length = UDPH_SIZE;
2050 break;
2051 case IPPROTO_ICMP:
2052 case IPPROTO_ICMPV6:
2053 min_ulp_header_length = ICMPH_SIZE;
2054 break;
2055 case IPPROTO_FRAGMENT:
2056 case IPPROTO_DSTOPTS:
2057 case IPPROTO_ROUTING:
2058 min_ulp_header_length = MIN_EHDR_LEN;
2059 break;
2060 default:
2061 min_ulp_header_length = 0;
2062 break;
2064 /* Make sure we have the min ULP header length */
2065 len = mp->b_wptr - rptr;
2066 if (len < ip_hdr_length + min_ulp_header_length) {
2067 if (ira->ira_pktlen < ip_hdr_length + min_ulp_header_length)
2068 goto pkt_too_short;
2070 IP6_STAT(ipst, ip6_recv_pullup);
2071 ip6h = ip_pullup(mp, ip_hdr_length + min_ulp_header_length,
2072 ira);
2073 if (ip6h == NULL)
2074 goto discard;
2075 len = mp->b_wptr - rptr;
2079 * If trusted extensions then determine the zoneid and TX specific
2080 * ira_flags.
2082 if (iraflags & IRAF_SYSTEM_LABELED) {
2083 /* This can update ira->ira_flags and ira->ira_zoneid */
2084 ip_fanout_tx_v6(mp, ip6h, protocol, ip_hdr_length, ira);
2085 iraflags = ira->ira_flags;
2089 /* Verify ULP checksum. Handles TCP, UDP, and SCTP */
2090 if (iraflags & IRAF_VERIFY_ULP_CKSUM) {
2091 if (!ip_input_cksum_v6(iraflags, mp, ip6h, ira)) {
2092 /* Bad checksum. Stats are already incremented */
2093 ip_drop_input("Bad ULP checksum", mp, ill);
2094 freemsg(mp);
2095 return;
2097 /* IRAF_SCTP_CSUM_ERR could have been set */
2098 iraflags = ira->ira_flags;
2100 switch (protocol) {
2101 case IPPROTO_TCP:
2102 /* For TCP, discard multicast packets. */
2103 if (iraflags & IRAF_MULTIBROADCAST)
2104 goto discard;
2106 /* First mblk contains IP+TCP headers per above check */
2107 ASSERT(len >= ip_hdr_length + TCP_MIN_HEADER_LENGTH);
2109 /* TCP options present? */
2110 offset = ((uchar_t *)ip6h)[ip_hdr_length + 12] >> 4;
2111 if (offset != 5) {
2112 if (offset < 5)
2113 goto discard;
2116 * There must be TCP options.
2117 * Make sure we can grab them.
2119 offset <<= 2;
2120 offset += ip_hdr_length;
2121 if (len < offset) {
2122 if (ira->ira_pktlen < offset)
2123 goto pkt_too_short;
2125 IP6_STAT(ipst, ip6_recv_pullup);
2126 ip6h = ip_pullup(mp, offset, ira);
2127 if (ip6h == NULL)
2128 goto discard;
2129 len = mp->b_wptr - rptr;
2134 * Pass up a squeue hint to tcp.
2135 * If ira_sqp is already set (this is loopback) we leave it
2136 * alone.
2138 if (ira->ira_sqp == NULL) {
2139 ira->ira_sqp = ip_squeue_get(ira->ira_ring);
2142 /* Look for AF_INET or AF_INET6 that matches */
2143 connp = ipcl_classify_v6(mp, IPPROTO_TCP, ip_hdr_length,
2144 ira, ipst);
2145 if (connp == NULL) {
2146 /* Send the TH_RST */
2147 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2148 tcp_xmit_listeners_reset(mp, ira, ipst, NULL);
2149 return;
2151 if (connp->conn_incoming_ifindex != 0 &&
2152 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2153 CONN_DEC_REF(connp);
2155 /* Send the TH_RST */
2156 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2157 tcp_xmit_listeners_reset(mp, ira, ipst, NULL);
2158 return;
2160 if (CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss) ||
2161 (iraflags & IRAF_IPSEC_SECURE)) {
2162 mp = ipsec_check_inbound_policy(mp, connp,
2163 NULL, ip6h, ira);
2164 if (mp == NULL) {
2165 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2166 /* Note that mp is NULL */
2167 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2168 CONN_DEC_REF(connp);
2169 return;
2172 /* Found a client; up it goes */
2173 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2174 ira->ira_ill = ira->ira_rill = NULL;
2175 if (!IPCL_IS_TCP(connp)) {
2176 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2177 (connp->conn_recv)(connp, mp, NULL, ira);
2178 CONN_DEC_REF(connp);
2179 ira->ira_ill = ill;
2180 ira->ira_rill = rill;
2181 return;
2185 * We do different processing whether called from
2186 * ip_accept_tcp and we match the target, don't match
2187 * the target, and when we are called by ip_input.
2189 if (iraflags & IRAF_TARGET_SQP) {
2190 if (ira->ira_target_sqp == connp->conn_sqp) {
2191 mblk_t *attrmp;
2193 attrmp = ip_recv_attr_to_mblk(ira);
2194 if (attrmp == NULL) {
2195 BUMP_MIB(ill->ill_ip_mib,
2196 ipIfStatsInDiscards);
2197 ip_drop_input("ipIfStatsInDiscards",
2198 mp, ill);
2199 freemsg(mp);
2200 CONN_DEC_REF(connp);
2201 } else {
2202 SET_SQUEUE(attrmp, connp->conn_recv,
2203 connp);
2204 attrmp->b_cont = mp;
2205 ASSERT(ira->ira_target_sqp_mp == NULL);
2206 ira->ira_target_sqp_mp = attrmp;
2208 * Conn ref release when drained from
2209 * the squeue.
2212 } else {
2213 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2214 connp->conn_recv, connp, ira, SQ_FILL,
2215 SQTAG_IP6_TCP_INPUT);
2217 } else {
2218 SQUEUE_ENTER_ONE(connp->conn_sqp, mp, connp->conn_recv,
2219 connp, ira, ip_squeue_flag, SQTAG_IP6_TCP_INPUT);
2221 ira->ira_ill = ill;
2222 ira->ira_rill = rill;
2223 return;
2225 case IPPROTO_SCTP: {
2226 sctp_hdr_t *sctph;
2227 uint32_t ports; /* Source and destination ports */
2228 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
2230 /* For SCTP, discard multicast packets. */
2231 if (iraflags & IRAF_MULTIBROADCAST)
2232 goto discard;
2235 * Since there is no SCTP h/w cksum support yet, just
2236 * clear the flag.
2238 DB_CKSUMFLAGS(mp) = 0;
2240 /* Length ensured above */
2241 ASSERT(MBLKL(mp) >= ip_hdr_length + SCTP_COMMON_HDR_LENGTH);
2242 sctph = (sctp_hdr_t *)(rptr + ip_hdr_length);
2244 /* get the ports */
2245 ports = *(uint32_t *)&sctph->sh_sport;
2247 if (iraflags & IRAF_SCTP_CSUM_ERR) {
2249 * No potential sctp checksum errors go to the Sun
2250 * sctp stack however they might be Adler-32 summed
2251 * packets a userland stack bound to a raw IP socket
2252 * could reasonably use. Note though that Adler-32 is
2253 * a long deprecated algorithm and customer sctp
2254 * networks should eventually migrate to CRC-32 at
2255 * which time this facility should be removed.
2257 ip_fanout_sctp_raw(mp, NULL, ip6h, ports, ira);
2258 return;
2260 connp = sctp_fanout(&ip6h->ip6_src, &ip6h->ip6_dst, ports,
2261 ira, mp, sctps, sctph);
2262 if (connp == NULL) {
2263 /* Check for raw socket or OOTB handling */
2264 ip_fanout_sctp_raw(mp, NULL, ip6h, ports, ira);
2265 return;
2267 if (connp->conn_incoming_ifindex != 0 &&
2268 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2269 CONN_DEC_REF(connp);
2271 /* Check for raw socket or OOTB handling */
2272 ip_fanout_sctp_raw(mp, NULL, ip6h, ports, ira);
2273 return;
2276 /* Found a client; up it goes */
2277 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2278 sctp_input(connp, NULL, ip6h, mp, ira);
2279 /* sctp_input does a rele of the sctp_t */
2280 return;
2283 case IPPROTO_UDP:
2284 /* First mblk contains IP+UDP headers as checked above */
2285 ASSERT(MBLKL(mp) >= ip_hdr_length + UDPH_SIZE);
2287 if (iraflags & IRAF_MULTIBROADCAST) {
2288 uint16_t *up; /* Pointer to ports in ULP header */
2290 up = (uint16_t *)((uchar_t *)ip6h + ip_hdr_length);
2292 ip_fanout_udp_multi_v6(mp, ip6h, up[1], up[0], ira);
2293 return;
2296 /* Look for AF_INET or AF_INET6 that matches */
2297 connp = ipcl_classify_v6(mp, IPPROTO_UDP, ip_hdr_length,
2298 ira, ipst);
2299 if (connp == NULL) {
2300 no_udp_match:
2301 if (ipst->ips_ipcl_proto_fanout_v6[IPPROTO_UDP].
2302 connf_head != NULL) {
2303 ASSERT(ira->ira_protocol == IPPROTO_UDP);
2304 ip_fanout_proto_v6(mp, ip6h, ira);
2305 } else {
2306 ip_fanout_send_icmp_v6(mp, ICMP6_DST_UNREACH,
2307 ICMP6_DST_UNREACH_NOPORT, ira);
2309 return;
2312 if (connp->conn_incoming_ifindex != 0 &&
2313 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2314 CONN_DEC_REF(connp);
2315 goto no_udp_match;
2317 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
2318 !canputnext(connp->conn_rq)) {
2319 CONN_DEC_REF(connp);
2320 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
2321 ip_drop_input("udpIfStatsInOverflows", mp, ill);
2322 freemsg(mp);
2323 return;
2325 if (CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss) ||
2326 (iraflags & IRAF_IPSEC_SECURE)) {
2327 mp = ipsec_check_inbound_policy(mp, connp,
2328 NULL, ip6h, ira);
2329 if (mp == NULL) {
2330 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2331 /* Note that mp is NULL */
2332 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2333 CONN_DEC_REF(connp);
2334 return;
2338 /* Found a client; up it goes */
2339 IP6_STAT(ipst, ip6_udp_fannorm);
2340 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2341 ira->ira_ill = ira->ira_rill = NULL;
2342 (connp->conn_recv)(connp, mp, NULL, ira);
2343 CONN_DEC_REF(connp);
2344 ira->ira_ill = ill;
2345 ira->ira_rill = rill;
2346 return;
2347 default:
2348 break;
2352 * Clear hardware checksumming flag as it is currently only
2353 * used by TCP and UDP.
2355 DB_CKSUMFLAGS(mp) = 0;
2357 switch (protocol) {
2358 case IPPROTO_ICMPV6:
2359 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInMsgs);
2361 /* Check variable for testing applications */
2362 if (ipst->ips_ipv6_drop_inbound_icmpv6) {
2363 ip_drop_input("ipv6_drop_inbound_icmpv6", mp, ill);
2364 freemsg(mp);
2365 return;
2368 * We need to accomodate icmp messages coming in clear
2369 * until we get everything secure from the wire. If
2370 * icmp_accept_clear_messages is zero we check with
2371 * the global policy and act accordingly. If it is
2372 * non-zero, we accept the message without any checks.
2373 * But *this does not mean* that this will be delivered
2374 * to RAW socket clients. By accepting we might send
2375 * replies back, change our MTU value etc.,
2376 * but delivery to the ULP/clients depends on their
2377 * policy dispositions.
2379 if (ipst->ips_icmp_accept_clear_messages == 0) {
2380 mp = ipsec_check_global_policy(mp, NULL,
2381 NULL, ip6h, ira, ns);
2382 if (mp == NULL)
2383 return;
2387 * On a labeled system, we have to check whether the zone
2388 * itself is permitted to receive raw traffic.
2390 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
2391 if (!tsol_can_accept_raw(mp, ira, B_FALSE)) {
2392 BUMP_MIB(ill->ill_icmp6_mib,
2393 ipv6IfIcmpInErrors);
2394 ip_drop_input("tsol_can_accept_raw", mp, ill);
2395 freemsg(mp);
2396 return;
2400 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2401 mp = icmp_inbound_v6(mp, ira);
2402 if (mp == NULL) {
2403 /* No need to pass to RAW sockets */
2404 return;
2406 break;
2408 case IPPROTO_DSTOPTS: {
2409 ip6_dest_t *desthdr;
2410 uint_t ehdrlen;
2411 uint8_t *optptr;
2413 /* We already check for MIN_EHDR_LEN above */
2415 /* Check if AH is present and needs to be processed. */
2416 mp = ipsec_early_ah_v6(mp, ira);
2417 if (mp == NULL)
2418 return;
2421 * Reinitialize pointers, as ipsec_early_ah_v6() does
2422 * complete pullups. We don't have to do more pullups
2423 * as a result.
2425 ip6h = (ip6_t *)mp->b_rptr;
2427 if (ira->ira_pktlen - ip_hdr_length < MIN_EHDR_LEN)
2428 goto pkt_too_short;
2430 if (mp->b_cont != NULL &&
2431 rptr + ip_hdr_length + MIN_EHDR_LEN > mp->b_wptr) {
2432 ip6h = ip_pullup(mp, ip_hdr_length + MIN_EHDR_LEN, ira);
2433 if (ip6h == NULL)
2434 goto discard;
2436 desthdr = (ip6_dest_t *)(rptr + ip_hdr_length);
2437 ehdrlen = 8 * (desthdr->ip6d_len + 1);
2438 if (ira->ira_pktlen - ip_hdr_length < ehdrlen)
2439 goto pkt_too_short;
2440 if (mp->b_cont != NULL &&
2441 rptr + IPV6_HDR_LEN + ehdrlen > mp->b_wptr) {
2442 ip6h = ip_pullup(mp, IPV6_HDR_LEN + ehdrlen, ira);
2443 if (ip6h == NULL)
2444 goto discard;
2446 desthdr = (ip6_dest_t *)(rptr + ip_hdr_length);
2448 optptr = (uint8_t *)&desthdr[1];
2451 * Update ira_ip_hdr_length to skip the destination header
2452 * when we repeat.
2454 ira->ira_ip_hdr_length += ehdrlen;
2456 ira->ira_protocol = desthdr->ip6d_nxt;
2459 * Note: XXX This code does not seem to make
2460 * distinction between Destination Options Header
2461 * being before/after Routing Header which can
2462 * happen if we are at the end of source route.
2463 * This may become significant in future.
2464 * (No real significant Destination Options are
2465 * defined/implemented yet ).
2467 switch (ip_process_options_v6(mp, ip6h, optptr,
2468 ehdrlen - 2, IPPROTO_DSTOPTS, ira)) {
2469 case -1:
2471 * Packet has been consumed and any needed
2472 * ICMP errors sent.
2474 return;
2475 case 0:
2476 /* No action needed continue */
2477 break;
2478 case 1:
2480 * Unnexpected return value
2481 * (Router alert is a Hop-by-Hop option)
2483 #ifdef DEBUG
2484 panic("ip_fanout_v6: router "
2485 "alert hbh opt indication in dest opt");
2486 /*NOTREACHED*/
2487 #else
2488 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2489 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2490 freemsg(mp);
2491 return;
2492 #endif
2494 goto repeat;
2496 case IPPROTO_FRAGMENT: {
2497 ip6_frag_t *fraghdr;
2499 if (ira->ira_pktlen - ip_hdr_length < sizeof (ip6_frag_t))
2500 goto pkt_too_short;
2502 if (mp->b_cont != NULL &&
2503 rptr + ip_hdr_length + sizeof (ip6_frag_t) > mp->b_wptr) {
2504 ip6h = ip_pullup(mp,
2505 ip_hdr_length + sizeof (ip6_frag_t), ira);
2506 if (ip6h == NULL)
2507 goto discard;
2510 fraghdr = (ip6_frag_t *)(rptr + ip_hdr_length);
2511 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
2514 * Invoke the CGTP (multirouting) filtering module to
2515 * process the incoming packet. Packets identified as
2516 * duplicates must be discarded. Filtering is active
2517 * only if the ip_cgtp_filter ndd variable is
2518 * non-zero.
2520 if (ipst->ips_ip_cgtp_filter &&
2521 ipst->ips_ip_cgtp_filter_ops != NULL) {
2522 int cgtp_flt_pkt;
2523 netstackid_t stackid;
2525 stackid = ipst->ips_netstack->netstack_stackid;
2528 * CGTP and IPMP are mutually exclusive so
2529 * phyint_ifindex is fine here.
2531 cgtp_flt_pkt =
2532 ipst->ips_ip_cgtp_filter_ops->cfo_filter_v6(
2533 stackid, ill->ill_phyint->phyint_ifindex,
2534 ip6h, fraghdr);
2535 if (cgtp_flt_pkt == CGTP_IP_PKT_DUPLICATE) {
2536 ip_drop_input("CGTP_IP_PKT_DUPLICATE", mp, ill);
2537 freemsg(mp);
2538 return;
2543 * Update ip_hdr_length to skip the frag header
2544 * ip_input_fragment_v6 will determine the extension header
2545 * prior to the fragment header and update its nexthdr value,
2546 * and also set ira_protocol to the nexthdr that follows the
2547 * completed fragment.
2549 ip_hdr_length += sizeof (ip6_frag_t);
2552 * Make sure we have ira_l2src before we loose the original
2553 * mblk
2555 if (!(ira->ira_flags & IRAF_L2SRC_SET))
2556 ip_setl2src(mp, ira, ira->ira_rill);
2558 mp = ip_input_fragment_v6(mp, ip6h, fraghdr,
2559 ira->ira_pktlen - ip_hdr_length, ira);
2560 if (mp == NULL) {
2561 /* Reassembly is still pending */
2562 return;
2564 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
2567 * The mblk chain has the frag header removed and
2568 * ira_protocol, ira_pktlen, ira_ip_hdr_length as well as the
2569 * IP header has been updated to refleact the result.
2571 ip6h = (ip6_t *)mp->b_rptr;
2572 ip_hdr_length = ira->ira_ip_hdr_length;
2573 goto repeat;
2575 case IPPROTO_HOPOPTS:
2577 * Illegal header sequence.
2578 * (Hop-by-hop headers are processed above
2579 * and required to immediately follow IPv6 header)
2581 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ill);
2582 icmp_param_problem_nexthdr_v6(mp, B_FALSE, ira);
2583 return;
2585 case IPPROTO_ROUTING: {
2586 uint_t ehdrlen;
2587 ip6_rthdr_t *rthdr;
2589 /* Check if AH is present and needs to be processed. */
2590 mp = ipsec_early_ah_v6(mp, ira);
2591 if (mp == NULL)
2592 return;
2595 * Reinitialize pointers, as ipsec_early_ah_v6() does
2596 * complete pullups. We don't have to do more pullups
2597 * as a result.
2599 ip6h = (ip6_t *)mp->b_rptr;
2601 if (ira->ira_pktlen - ip_hdr_length < MIN_EHDR_LEN)
2602 goto pkt_too_short;
2604 if (mp->b_cont != NULL &&
2605 rptr + ip_hdr_length + MIN_EHDR_LEN > mp->b_wptr) {
2606 ip6h = ip_pullup(mp, ip_hdr_length + MIN_EHDR_LEN, ira);
2607 if (ip6h == NULL)
2608 goto discard;
2610 rthdr = (ip6_rthdr_t *)(rptr + ip_hdr_length);
2611 protocol = ira->ira_protocol = rthdr->ip6r_nxt;
2612 ehdrlen = 8 * (rthdr->ip6r_len + 1);
2613 if (ira->ira_pktlen - ip_hdr_length < ehdrlen)
2614 goto pkt_too_short;
2615 if (mp->b_cont != NULL &&
2616 rptr + IPV6_HDR_LEN + ehdrlen > mp->b_wptr) {
2617 ip6h = ip_pullup(mp, IPV6_HDR_LEN + ehdrlen, ira);
2618 if (ip6h == NULL)
2619 goto discard;
2620 rthdr = (ip6_rthdr_t *)(rptr + ip_hdr_length);
2622 if (rthdr->ip6r_segleft != 0) {
2623 /* Not end of source route */
2624 if (ira->ira_flags &
2625 (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST)) {
2626 BUMP_MIB(ill->ill_ip_mib,
2627 ipIfStatsForwProhibits);
2628 ip_drop_input("ipIfStatsInForwProhibits",
2629 mp, ill);
2630 freemsg(mp);
2631 return;
2633 ip_process_rthdr(mp, ip6h, rthdr, ira);
2634 return;
2636 ira->ira_ip_hdr_length += ehdrlen;
2637 goto repeat;
2640 case IPPROTO_AH:
2641 case IPPROTO_ESP: {
2643 * Fast path for AH/ESP.
2645 netstack_t *ns = ipst->ips_netstack;
2646 ipsec_stack_t *ipss = ns->netstack_ipsec;
2648 IP_STAT(ipst, ipsec_proto_ahesp);
2650 if (!ipsec_loaded(ipss)) {
2651 ip_proto_not_sup(mp, ira);
2652 return;
2655 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2656 /* select inbound SA and have IPsec process the pkt */
2657 if (protocol == IPPROTO_ESP) {
2658 esph_t *esph;
2660 mp = ipsec_inbound_esp_sa(mp, ira, &esph);
2661 if (mp == NULL)
2662 return;
2664 ASSERT(esph != NULL);
2665 ASSERT(ira->ira_flags & IRAF_IPSEC_SECURE);
2666 ASSERT(ira->ira_ipsec_esp_sa != NULL);
2667 ASSERT(ira->ira_ipsec_esp_sa->ipsa_input_func != NULL);
2669 mp = ira->ira_ipsec_esp_sa->ipsa_input_func(mp, esph,
2670 ira);
2671 } else {
2672 ah_t *ah;
2674 mp = ipsec_inbound_ah_sa(mp, ira, &ah);
2675 if (mp == NULL)
2676 return;
2678 ASSERT(ah != NULL);
2679 ASSERT(ira->ira_flags & IRAF_IPSEC_SECURE);
2680 ASSERT(ira->ira_ipsec_ah_sa != NULL);
2681 ASSERT(ira->ira_ipsec_ah_sa->ipsa_input_func != NULL);
2682 mp = ira->ira_ipsec_ah_sa->ipsa_input_func(mp, ah,
2683 ira);
2686 if (mp == NULL) {
2688 * Either it failed or is pending. In the former case
2689 * ipIfStatsInDiscards was increased.
2691 return;
2693 /* we're done with IPsec processing, send it up */
2694 ip_input_post_ipsec(mp, ira);
2695 return;
2697 case IPPROTO_NONE:
2698 /* All processing is done. Count as "delivered". */
2699 freemsg(mp);
2700 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2701 return;
2703 case IPPROTO_ENCAP:
2704 case IPPROTO_IPV6:
2705 /* iptun will verify trusted label */
2706 connp = ipcl_classify_v6(mp, protocol, ip_hdr_length,
2707 ira, ipst);
2708 if (connp != NULL) {
2709 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2710 ira->ira_ill = ira->ira_rill = NULL;
2711 connp->conn_recv(connp, mp, NULL, ira);
2712 CONN_DEC_REF(connp);
2713 ira->ira_ill = ill;
2714 ira->ira_rill = rill;
2715 return;
2717 /* FALLTHRU */
2718 default:
2720 * On a labeled system, we have to check whether the zone
2721 * itself is permitted to receive raw traffic.
2723 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
2724 if (!tsol_can_accept_raw(mp, ira, B_FALSE)) {
2725 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2726 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2727 freemsg(mp);
2728 return;
2731 break;
2735 * The above input functions may have returned the pulled up message.
2736 * So ip6h need to be reinitialized.
2738 ip6h = (ip6_t *)mp->b_rptr;
2739 ira->ira_protocol = protocol;
2740 if (ipst->ips_ipcl_proto_fanout_v6[protocol].connf_head == NULL) {
2741 /* No user-level listener for these packets packets */
2742 ip_proto_not_sup(mp, ira);
2743 return;
2747 * Handle fanout to raw sockets. There
2748 * can be more than one stream bound to a particular
2749 * protocol. When this is the case, each one gets a copy
2750 * of any incoming packets.
2752 ASSERT(ira->ira_protocol == protocol);
2753 ip_fanout_proto_v6(mp, ip6h, ira);
2754 return;
2756 pkt_too_short:
2757 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2758 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2759 freemsg(mp);
2760 return;
2762 discard:
2763 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2764 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2765 freemsg(mp);
2766 #undef rptr