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 (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
24 *
25 * Copyright 2011 Nexenta Systems, 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/dlpi.h>
33 #include <sys/stropts.h>
34 #include <sys/sysmacros.h>
35 #include <sys/strsubr.h>
36 #include <sys/strlog.h>
37 #include <sys/strsun.h>
38 #include <sys/zone.h>
39 #define _SUN_TPI_VERSION 2
40 #include <sys/tihdr.h>
41 #include <sys/xti_inet.h>
42 #include <sys/ddi.h>
43 #include <sys/sunddi.h>
44 #include <sys/cmn_err.h>
45 #include <sys/debug.h>
46 #include <sys/kobj.h>
47 #include <sys/modctl.h>
48 #include <sys/atomic.h>
49 #include <sys/policy.h>
50 #include <sys/priv.h>
51
52 #include <sys/systm.h>
53 #include <sys/param.h>
54 #include <sys/kmem.h>
55 #include <sys/sdt.h>
56 #include <sys/socket.h>
57 #include <sys/vtrace.h>
58 #include <sys/isa_defs.h>
59 #include <sys/mac.h>
60 #include <net/if.h>
61 #include <net/if_arp.h>
62 #include <net/route.h>
63 #include <sys/sockio.h>
64 #include <netinet/in.h>
65 #include <net/if_dl.h>
66
67 #include <inet/common.h>
68 #include <inet/mi.h>
69 #include <inet/mib2.h>
70 #include <inet/nd.h>
71 #include <inet/arp.h>
72 #include <inet/snmpcom.h>
73 #include <inet/kstatcom.h>
74
75 #include <netinet/igmp_var.h>
76 #include <netinet/ip6.h>
77 #include <netinet/icmp6.h>
78 #include <netinet/sctp.h>
79
80 #include <inet/ip.h>
81 #include <inet/ip_impl.h>
82 #include <inet/ip6.h>
83 #include <inet/ip6_asp.h>
84 #include <inet/optcom.h>
85 #include <inet/tcp.h>
86 #include <inet/tcp_impl.h>
87 #include <inet/ip_multi.h>
88 #include <inet/ip_if.h>
89 #include <inet/ip_ire.h>
90 #include <inet/ip_ftable.h>
91 #include <inet/ip_rts.h>
92 #include <inet/ip_ndp.h>
93 #include <inet/ip_listutils.h>
94 #include <netinet/igmp.h>
95 #include <netinet/ip_mroute.h>
96 #include <inet/ipp_common.h>
97
98 #include <net/pfkeyv2.h>
99 #include <inet/sadb.h>
100 #include <inet/ipsec_impl.h>
101 #include <inet/ipdrop.h>
102 #include <inet/ip_netinfo.h>
103 #include <inet/ilb_ip.h>
104 #include <sys/squeue_impl.h>
105 #include <sys/squeue.h>
106
107 #include <sys/ethernet.h>
108 #include <net/if_types.h>
109 #include <sys/cpuvar.h>
110
111 #include <ipp/ipp.h>
112 #include <ipp/ipp_impl.h>
113 #include <ipp/ipgpc/ipgpc.h>
114
115 #include <sys/pattr.h>
116 #include <inet/ipclassifier.h>
117 #include <inet/sctp_ip.h>
118 #include <inet/sctp/sctp_impl.h>
119 #include <inet/udp_impl.h>
120 #include <sys/sunddi.h>
121
122 #include <sys/tsol/label.h>
123 #include <sys/tsol/tnet.h>
124
125 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
126
127 #ifdef DEBUG
128 extern boolean_t skip_sctp_cksum;
129 #endif
130
131 static void ip_input_local_v4(ire_t *, mblk_t *, ipha_t *,
132 ip_recv_attr_t *);
133
134 static void ip_input_broadcast_v4(ire_t *, mblk_t *, ipha_t *,
135 ip_recv_attr_t *);
136 static void ip_input_multicast_v4(ire_t *, mblk_t *, ipha_t *,
137 ip_recv_attr_t *);
138
139 #pragma inline(ip_input_common_v4, ip_input_local_v4, ip_forward_xmit_v4)
140
141 /*
142 * Direct read side procedure capable of dealing with chains. GLDv3 based
143 * drivers call this function directly with mblk chains while STREAMS
144 * read side procedure ip_rput() calls this for single packet with ip_ring
145 * set to NULL to process one packet at a time.
146 *
147 * The ill will always be valid if this function is called directly from
148 * the driver.
149 *
150 * If this chain is part of a VLAN stream, then the VLAN tag is
151 * stripped from the MAC header before being delivered to this
152 * function.
153 *
154 * If the IP header in packet is not 32-bit aligned, every message in the
155 * chain will be aligned before further operations. This is required on SPARC
156 * platform.
157 */
158 void
159 ip_input(ill_t *ill, ill_rx_ring_t *ip_ring, mblk_t *mp_chain,
160 struct mac_header_info_s *mhip)
161 {
162 (void) ip_input_common_v4(ill, ip_ring, mp_chain, mhip, NULL, NULL,
163 NULL);
164 }
165
166 /*
167 * ip_accept_tcp() - This function is called by the squeue when it retrieves
168 * a chain of packets in the poll mode. The packets have gone through the
169 * data link processing but not IP processing. For performance and latency
170 * reasons, the squeue wants to process the chain in line instead of feeding
171 * it back via ip_input path.
172 *
173 * We set up the ip_recv_attr_t with IRAF_TARGET_SQP to that ip_fanout_v4
174 * will pass back any TCP packets matching the target sqp to
175 * ip_input_common_v4 using ira_target_sqp_mp. Other packets are handled by
176 * ip_input_v4 and ip_fanout_v4 as normal.
177 * The TCP packets that match the target squeue are returned to the caller
178 * as a b_next chain after each packet has been prepend with an mblk
179 * from ip_recv_attr_to_mblk.
180 */
181 mblk_t *
182 ip_accept_tcp(ill_t *ill, ill_rx_ring_t *ip_ring, squeue_t *target_sqp,
183 mblk_t *mp_chain, mblk_t **last, uint_t *cnt)
184 {
185 return (ip_input_common_v4(ill, ip_ring, mp_chain, NULL, target_sqp,
186 last, cnt));
187 }
188
189 /*
190 * Used by ip_input and ip_accept_tcp
191 * The last three arguments are only used by ip_accept_tcp, and mhip is
192 * only used by ip_input.
193 */
194 mblk_t *
195 ip_input_common_v4(ill_t *ill, ill_rx_ring_t *ip_ring, mblk_t *mp_chain,
196 struct mac_header_info_s *mhip, squeue_t *target_sqp,
197 mblk_t **last, uint_t *cnt)
198 {
199 mblk_t *mp;
200 ipha_t *ipha;
201 ip_recv_attr_t iras; /* Receive attributes */
202 rtc_t rtc;
203 iaflags_t chain_flags = 0; /* Fixed for chain */
204 mblk_t *ahead = NULL; /* Accepted head */
205 mblk_t *atail = NULL; /* Accepted tail */
206 uint_t acnt = 0; /* Accepted count */
207
208 ASSERT(mp_chain != NULL);
209 ASSERT(ill != NULL);
210
211 /* These ones do not change as we loop over packets */
212 iras.ira_ill = iras.ira_rill = ill;
213 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
214 iras.ira_rifindex = iras.ira_ruifindex;
215 iras.ira_sqp = NULL;
216 iras.ira_ring = ip_ring;
217 /* For ECMP and outbound transmit ring selection */
218 iras.ira_xmit_hint = ILL_RING_TO_XMIT_HINT(ip_ring);
219
220 iras.ira_target_sqp = target_sqp;
221 iras.ira_target_sqp_mp = NULL;
222 if (target_sqp != NULL)
223 chain_flags |= IRAF_TARGET_SQP;
224
225 /*
226 * We try to have a mhip pointer when possible, but
227 * it might be NULL in some cases. In those cases we
228 * have to assume unicast.
229 */
230 iras.ira_mhip = mhip;
231 iras.ira_flags = 0;
232 if (mhip != NULL) {
233 switch (mhip->mhi_dsttype) {
234 case MAC_ADDRTYPE_MULTICAST :
235 chain_flags |= IRAF_L2DST_MULTICAST;
236 break;
237 case MAC_ADDRTYPE_BROADCAST :
238 chain_flags |= IRAF_L2DST_BROADCAST;
239 break;
240 }
241 }
242
243 /*
244 * Initialize the one-element route cache.
245 *
246 * We do ire caching from one iteration to
247 * another. In the event the packet chain contains
248 * all packets from the same dst, this caching saves
249 * an ire_route_recursive for each of the succeeding
250 * packets in a packet chain.
251 */
252 rtc.rtc_ire = NULL;
253 rtc.rtc_ipaddr = INADDR_ANY;
254
255 /* Loop over b_next */
256 for (mp = mp_chain; mp != NULL; mp = mp_chain) {
257 mp_chain = mp->b_next;
258 mp->b_next = NULL;
259
260 ASSERT(DB_TYPE(mp) == M_DATA);
261
262
263 /*
264 * if db_ref > 1 then copymsg and free original. Packet
265 * may be changed and we do not want the other entity
266 * who has a reference to this message to trip over the
267 * changes. This is a blind change because trying to
268 * catch all places that might change the packet is too
269 * difficult.
270 *
271 * This corresponds to the fast path case, where we have
272 * a chain of M_DATA mblks. We check the db_ref count
273 * of only the 1st data block in the mblk chain. There
274 * doesn't seem to be a reason why a device driver would
275 * send up data with varying db_ref counts in the mblk
276 * chain. In any case the Fast path is a private
277 * interface, and our drivers don't do such a thing.
278 * Given the above assumption, there is no need to walk
279 * down the entire mblk chain (which could have a
280 * potential performance problem)
281 *
282 * The "(DB_REF(mp) > 1)" check was moved from ip_rput()
283 * to here because of exclusive ip stacks and vnics.
284 * Packets transmitted from exclusive stack over vnic
285 * can have db_ref > 1 and when it gets looped back to
286 * another vnic in a different zone, you have ip_input()
287 * getting dblks with db_ref > 1. So if someone
288 * complains of TCP performance under this scenario,
289 * take a serious look here on the impact of copymsg().
290 */
291 if (DB_REF(mp) > 1) {
292 if ((mp = ip_fix_dbref(mp, &iras)) == NULL) {
293 /* mhip might point into 1st packet in chain */
294 iras.ira_mhip = NULL;
295 continue;
296 }
297 }
298
299 /*
300 * IP header ptr not aligned?
301 * OR IP header not complete in first mblk
302 */
303 ipha = (ipha_t *)mp->b_rptr;
304 if (!OK_32PTR(ipha) || MBLKL(mp) < IP_SIMPLE_HDR_LENGTH) {
305 mp = ip_check_and_align_header(mp, IP_SIMPLE_HDR_LENGTH,
306 &iras);
307 if (mp == NULL) {
308 /* mhip might point into 1st packet in chain */
309 iras.ira_mhip = NULL;
310 continue;
311 }
312 ipha = (ipha_t *)mp->b_rptr;
313 }
314
315 /* Protect against a mix of Ethertypes and IP versions */
316 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
317 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
318 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
319 freemsg(mp);
320 /* mhip might point into 1st packet in the chain. */
321 iras.ira_mhip = NULL;
322 continue;
323 }
324
325 /*
326 * Check for Martian addrs; we have to explicitly
327 * test for for zero dst since this is also used as
328 * an indication that the rtc is not used.
329 */
330 if (ipha->ipha_dst == INADDR_ANY) {
331 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
332 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
333 freemsg(mp);
334 /* mhip might point into 1st packet in the chain. */
335 iras.ira_mhip = NULL;
336 continue;
337 }
338
339 /*
340 * Keep L2SRC from a previous packet in chain since mhip
341 * might point into an earlier packet in the chain.
342 * Keep IRAF_VERIFIED_SRC to avoid redoing broadcast
343 * source check in forwarding path.
344 */
345 chain_flags |= (iras.ira_flags &
346 (IRAF_L2SRC_SET|IRAF_VERIFIED_SRC));
347
348 iras.ira_flags = IRAF_IS_IPV4 | IRAF_VERIFY_IP_CKSUM |
349 IRAF_VERIFY_ULP_CKSUM | chain_flags;
350 iras.ira_free_flags = 0;
351 iras.ira_cred = NULL;
352 iras.ira_cpid = NOPID;
353 iras.ira_tsl = NULL;
354 iras.ira_zoneid = ALL_ZONES; /* Default for forwarding */
355
356 /*
357 * We must count all incoming packets, even if they end
358 * up being dropped later on. Defer counting bytes until
359 * we have the whole IP header in first mblk.
360 */
361 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInReceives);
362
363 iras.ira_pktlen = ntohs(ipha->ipha_length);
364 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCInOctets,
365 iras.ira_pktlen);
366
367 /*
368 * Call one of:
369 * ill_input_full_v4
370 * ill_input_short_v4
371 * The former is used in unusual cases. See ill_set_inputfn().
372 */
373 (*ill->ill_inputfn)(mp, ipha, &ipha->ipha_dst, &iras, &rtc);
374
375 /* Any references to clean up? No hold on ira_ill */
376 if (iras.ira_flags & (IRAF_IPSEC_SECURE|IRAF_SYSTEM_LABELED))
377 ira_cleanup(&iras, B_FALSE);
378
379 if (iras.ira_target_sqp_mp != NULL) {
380 /* Better be called from ip_accept_tcp */
381 ASSERT(target_sqp != NULL);
382
383 /* Found one packet to accept */
384 mp = iras.ira_target_sqp_mp;
385 iras.ira_target_sqp_mp = NULL;
386 ASSERT(ip_recv_attr_is_mblk(mp));
387
388 if (atail != NULL)
389 atail->b_next = mp;
390 else
391 ahead = mp;
392 atail = mp;
393 acnt++;
394 mp = NULL;
395 }
396 /* mhip might point into 1st packet in the chain. */
397 iras.ira_mhip = NULL;
398 }
399 /* Any remaining references to the route cache? */
400 if (rtc.rtc_ire != NULL) {
401 ASSERT(rtc.rtc_ipaddr != INADDR_ANY);
402 ire_refrele(rtc.rtc_ire);
403 }
404
405 if (ahead != NULL) {
406 /* Better be called from ip_accept_tcp */
407 ASSERT(target_sqp != NULL);
408 *last = atail;
409 *cnt = acnt;
410 return (ahead);
411 }
412
413 return (NULL);
414 }
415
416 /*
417 * This input function is used when
418 * - is_system_labeled()
419 * - CGTP filtering
420 * - DHCP unicast before we have an IP address configured
421 * - there is an listener for IPPROTO_RSVP
422 */
423 void
424 ill_input_full_v4(mblk_t *mp, void *iph_arg, void *nexthop_arg,
425 ip_recv_attr_t *ira, rtc_t *rtc)
426 {
427 ipha_t *ipha = (ipha_t *)iph_arg;
428 ipaddr_t nexthop = *(ipaddr_t *)nexthop_arg;
429 ill_t *ill = ira->ira_ill;
430 ip_stack_t *ipst = ill->ill_ipst;
431 int cgtp_flt_pkt;
432
433 ASSERT(ira->ira_tsl == NULL);
434
435 /*
436 * Attach any necessary label information to
437 * this packet
438 */
439 if (is_system_labeled()) {
440 ira->ira_flags |= IRAF_SYSTEM_LABELED;
441
442 /*
443 * This updates ira_cred, ira_tsl and ira_free_flags based
444 * on the label.
445 */
446 if (!tsol_get_pkt_label(mp, IPV4_VERSION, ira)) {
447 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
448 ip_drop_input("ipIfStatsInDiscards", mp, ill);
449 freemsg(mp);
450 return;
451 }
452 /* Note that ira_tsl can be NULL here. */
453
454 /* tsol_get_pkt_label sometimes does pullupmsg */
455 ipha = (ipha_t *)mp->b_rptr;
456 }
457
458 /*
459 * Invoke the CGTP (multirouting) filtering module to process
460 * the incoming packet. Packets identified as duplicates
461 * must be discarded. Filtering is active only if the
462 * the ip_cgtp_filter ndd variable is non-zero.
463 */
464 cgtp_flt_pkt = CGTP_IP_PKT_NOT_CGTP;
465 if (ipst->ips_ip_cgtp_filter &&
466 ipst->ips_ip_cgtp_filter_ops != NULL) {
467 netstackid_t stackid;
468
469 stackid = ipst->ips_netstack->netstack_stackid;
470 /*
471 * CGTP and IPMP are mutually exclusive so
472 * phyint_ifindex is fine here.
473 */
474 cgtp_flt_pkt =
475 ipst->ips_ip_cgtp_filter_ops->cfo_filter(stackid,
476 ill->ill_phyint->phyint_ifindex, mp);
477 if (cgtp_flt_pkt == CGTP_IP_PKT_DUPLICATE) {
478 ip_drop_input("CGTP_IP_PKT_DUPLICATE", mp, ill);
479 freemsg(mp);
480 return;
481 }
482 }
483
484 /*
485 * Brutal hack for DHCPv4 unicast: RFC2131 allows a DHCP
486 * server to unicast DHCP packets to a DHCP client using the
487 * IP address it is offering to the client. This can be
488 * disabled through the "broadcast bit", but not all DHCP
489 * servers honor that bit. Therefore, to interoperate with as
490 * many DHCP servers as possible, the DHCP client allows the
491 * server to unicast, but we treat those packets as broadcast
492 * here. Note that we don't rewrite the packet itself since
493 * (a) that would mess up the checksums and (b) the DHCP
494 * client conn is bound to INADDR_ANY so ip_fanout_udp() will
495 * hand it the packet regardless.
496 */
497 if (ill->ill_dhcpinit != 0 &&
498 ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION &&
499 ipha->ipha_protocol == IPPROTO_UDP) {
500 udpha_t *udpha;
501
502 ipha = ip_pullup(mp, sizeof (ipha_t) + sizeof (udpha_t), ira);
503 if (ipha == NULL) {
504 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
505 ip_drop_input("ipIfStatsInDiscards - dhcp", mp, ill);
506 freemsg(mp);
507 return;
508 }
509 /* Reload since pullupmsg() can change b_rptr. */
510 udpha = (udpha_t *)&ipha[1];
511
512 if (ntohs(udpha->uha_dst_port) == IPPORT_BOOTPC) {
513 DTRACE_PROBE2(ip4__dhcpinit__pkt, ill_t *, ill,
514 mblk_t *, mp);
515 /*
516 * This assumes that we deliver to all conns for
517 * multicast and broadcast packets.
518 */
519 nexthop = INADDR_BROADCAST;
520 ira->ira_flags |= IRAF_DHCP_UNICAST;
521 }
522 }
523
524 /*
525 * If rsvpd is running, let RSVP daemon handle its processing
526 * and forwarding of RSVP multicast/unicast packets.
527 * If rsvpd is not running but mrouted is running, RSVP
528 * multicast packets are forwarded as multicast traffic
529 * and RSVP unicast packets are forwarded by unicast router.
530 * If neither rsvpd nor mrouted is running, RSVP multicast
531 * packets are not forwarded, but the unicast packets are
532 * forwarded like unicast traffic.
533 */
534 if (ipha->ipha_protocol == IPPROTO_RSVP &&
535 ipst->ips_ipcl_proto_fanout_v4[IPPROTO_RSVP].connf_head != NULL) {
536 /* RSVP packet and rsvpd running. Treat as ours */
537 ip2dbg(("ip_input: RSVP for us: 0x%x\n", ntohl(nexthop)));
538 /*
539 * We use a multicast address to get the packet to
540 * ire_recv_multicast_v4. There will not be a membership
541 * check since we set IRAF_RSVP
542 */
543 nexthop = htonl(INADDR_UNSPEC_GROUP);
544 ira->ira_flags |= IRAF_RSVP;
545 }
546
547 ill_input_short_v4(mp, ipha, &nexthop, ira, rtc);
548 }
549
550 /*
551 * This is the tail-end of the full receive side packet handling.
552 * It can be used directly when the configuration is simple.
553 */
554 void
555 ill_input_short_v4(mblk_t *mp, void *iph_arg, void *nexthop_arg,
556 ip_recv_attr_t *ira, rtc_t *rtc)
557 {
558 ire_t *ire;
559 uint_t opt_len;
560 ill_t *ill = ira->ira_ill;
561 ip_stack_t *ipst = ill->ill_ipst;
562 uint_t pkt_len;
563 ssize_t len;
564 ipha_t *ipha = (ipha_t *)iph_arg;
565 ipaddr_t nexthop = *(ipaddr_t *)nexthop_arg;
566 ilb_stack_t *ilbs = ipst->ips_netstack->netstack_ilb;
567 uint_t irr_flags;
568 #define rptr ((uchar_t *)ipha)
569
570 ASSERT(DB_TYPE(mp) == M_DATA);
571
572 /*
573 * The following test for loopback is faster than
574 * IP_LOOPBACK_ADDR(), because it avoids any bitwise
575 * operations.
576 * Note that these addresses are always in network byte order
577 */
578 if (((*(uchar_t *)&ipha->ipha_dst) == IN_LOOPBACKNET) ||
579 ((*(uchar_t *)&ipha->ipha_src) == IN_LOOPBACKNET)) {
580 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
581 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
582 freemsg(mp);
583 return;
584 }
585
586 len = mp->b_wptr - rptr;
587 pkt_len = ira->ira_pktlen;
588
589 /* multiple mblk or too short */
590 len -= pkt_len;
591 if (len != 0) {
592 mp = ip_check_length(mp, rptr, len, pkt_len,
593 IP_SIMPLE_HDR_LENGTH, ira);
594 if (mp == NULL)
595 return;
596 ipha = (ipha_t *)mp->b_rptr;
597 }
598
599 DTRACE_IP7(receive, mblk_t *, mp, conn_t *, NULL, void_ip_t *,
600 ipha, __dtrace_ipsr_ill_t *, ill, ipha_t *, ipha, ip6_t *, NULL,
601 int, 0);
602
603 /*
604 * The event for packets being received from a 'physical'
605 * interface is placed after validation of the source and/or
606 * destination address as being local so that packets can be
607 * redirected to loopback addresses using ipnat.
608 */
609 DTRACE_PROBE4(ip4__physical__in__start,
610 ill_t *, ill, ill_t *, NULL,
611 ipha_t *, ipha, mblk_t *, mp);
612
613 if (HOOKS4_INTERESTED_PHYSICAL_IN(ipst)) {
614 int ll_multicast = 0;
615 int error;
616 ipaddr_t orig_dst = ipha->ipha_dst;
617
618 if (ira->ira_flags & IRAF_L2DST_MULTICAST)
619 ll_multicast = HPE_MULTICAST;
620 else if (ira->ira_flags & IRAF_L2DST_BROADCAST)
621 ll_multicast = HPE_BROADCAST;
622
623 FW_HOOKS(ipst->ips_ip4_physical_in_event,
624 ipst->ips_ipv4firewall_physical_in,
625 ill, NULL, ipha, mp, mp, ll_multicast, ipst, error);
626
627 DTRACE_PROBE1(ip4__physical__in__end, mblk_t *, mp);
628
629 if (mp == NULL)
630 return;
631 /* The length could have changed */
632 ipha = (ipha_t *)mp->b_rptr;
633 ira->ira_pktlen = ntohs(ipha->ipha_length);
634 pkt_len = ira->ira_pktlen;
635
636 /*
637 * In case the destination changed we override any previous
638 * change to nexthop.
639 */
640 if (orig_dst != ipha->ipha_dst)
641 nexthop = ipha->ipha_dst;
642 if (nexthop == INADDR_ANY) {
643 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInAddrErrors);
644 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
645 freemsg(mp);
646 return;
647 }
648 }
649
650 if (ipst->ips_ip4_observe.he_interested) {
651 zoneid_t dzone;
652
653 /*
654 * On the inbound path the src zone will be unknown as
655 * this packet has come from the wire.
656 */
657 dzone = ip_get_zoneid_v4(nexthop, mp, ira, ALL_ZONES);
658 ipobs_hook(mp, IPOBS_HOOK_INBOUND, ALL_ZONES, dzone, ill, ipst);
659 }
660
661 /*
662 * If there is a good HW IP header checksum we clear the need
663 * look at the IP header checksum.
664 */
665 if ((DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM) &&
666 ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
667 /* Header checksum was ok. Clear the flag */
668 DB_CKSUMFLAGS(mp) &= ~HCK_IPV4_HDRCKSUM;
669 ira->ira_flags &= ~IRAF_VERIFY_IP_CKSUM;
670 }
671
672 /*
673 * Here we check to see if we machine is setup as
674 * L3 loadbalancer and if the incoming packet is for a VIP
675 *
676 * Check the following:
677 * - there is at least a rule
678 * - protocol of the packet is supported
679 */
680 if (ilb_has_rules(ilbs) && ILB_SUPP_L4(ipha->ipha_protocol)) {
681 ipaddr_t lb_dst;
682 int lb_ret;
683
684 /* For convenience, we pull up the mblk. */
685 if (mp->b_cont != NULL) {
686 if (pullupmsg(mp, -1) == 0) {
687 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
688 ip_drop_input("ipIfStatsInDiscards - pullupmsg",
689 mp, ill);
690 freemsg(mp);
691 return;
692 }
693 ipha = (ipha_t *)mp->b_rptr;
694 }
695
696 /*
697 * We just drop all fragments going to any VIP, at
698 * least for now....
699 */
700 if (ntohs(ipha->ipha_fragment_offset_and_flags) &
701 (IPH_MF | IPH_OFFSET)) {
702 if (!ilb_rule_match_vip_v4(ilbs, nexthop, NULL)) {
703 goto after_ilb;
704 }
705
706 ILB_KSTAT_UPDATE(ilbs, ip_frag_in, 1);
707 ILB_KSTAT_UPDATE(ilbs, ip_frag_dropped, 1);
708 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
709 ip_drop_input("ILB fragment", mp, ill);
710 freemsg(mp);
711 return;
712 }
713 lb_ret = ilb_check_v4(ilbs, ill, mp, ipha, ipha->ipha_protocol,
714 (uint8_t *)ipha + IPH_HDR_LENGTH(ipha), &lb_dst);
715
716 if (lb_ret == ILB_DROPPED) {
717 /* Is this the right counter to increase? */
718 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
719 ip_drop_input("ILB_DROPPED", mp, ill);
720 freemsg(mp);
721 return;
722 }
723 if (lb_ret == ILB_BALANCED) {
724 /* Set the dst to that of the chosen server */
725 nexthop = lb_dst;
726 DB_CKSUMFLAGS(mp) = 0;
727 }
728 }
729
730 after_ilb:
731 opt_len = ipha->ipha_version_and_hdr_length - IP_SIMPLE_HDR_VERSION;
732 ira->ira_ip_hdr_length = IP_SIMPLE_HDR_LENGTH;
733 if (opt_len != 0) {
734 int error = 0;
735
736 ira->ira_ip_hdr_length += (opt_len << 2);
737 ira->ira_flags |= IRAF_IPV4_OPTIONS;
738
739 /* IP Options present! Validate the length. */
740 mp = ip_check_optlen(mp, ipha, opt_len, pkt_len, ira);
741 if (mp == NULL)
742 return;
743
744 /* Might have changed */
745 ipha = (ipha_t *)mp->b_rptr;
746
747 /* Verify IP header checksum before parsing the options */
748 if ((ira->ira_flags & IRAF_VERIFY_IP_CKSUM) &&
749 ip_csum_hdr(ipha)) {
750 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
751 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
752 freemsg(mp);
753 return;
754 }
755 ira->ira_flags &= ~IRAF_VERIFY_IP_CKSUM;
756
757 /*
758 * Go off to ip_input_options which returns the next hop
759 * destination address, which may have been affected
760 * by source routing.
761 */
762 IP_STAT(ipst, ip_opt);
763
764 nexthop = ip_input_options(ipha, nexthop, mp, ira, &error);
765 if (error != 0) {
766 /*
767 * An ICMP error has been sent and the packet has
768 * been dropped.
769 */
770 return;
771 }
772 }
773
774 if (ill->ill_flags & ILLF_ROUTER)
775 irr_flags = IRR_ALLOCATE;
776 else
777 irr_flags = IRR_NONE;
778
779 /* Can not use route cache with TX since the labels can differ */
780 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
781 if (CLASSD(nexthop)) {
782 ire = ire_multicast(ill);
783 } else {
784 /* Match destination and label */
785 ire = ire_route_recursive_v4(nexthop, 0, NULL,
786 ALL_ZONES, ira->ira_tsl, MATCH_IRE_SECATTR,
787 irr_flags, ira->ira_xmit_hint, ipst, NULL, NULL,
788 NULL);
789 }
790 /* Update the route cache so we do the ire_refrele */
791 ASSERT(ire != NULL);
792 if (rtc->rtc_ire != NULL)
793 ire_refrele(rtc->rtc_ire);
794 rtc->rtc_ire = ire;
795 rtc->rtc_ipaddr = nexthop;
796 } else if (nexthop == rtc->rtc_ipaddr && rtc->rtc_ire != NULL) {
797 /* Use the route cache */
798 ire = rtc->rtc_ire;
799 } else {
800 /* Update the route cache */
801 if (CLASSD(nexthop)) {
802 ire = ire_multicast(ill);
803 } else {
804 /* Just match the destination */
805 ire = ire_route_recursive_dstonly_v4(nexthop, irr_flags,
806 ira->ira_xmit_hint, ipst);
807 }
808 ASSERT(ire != NULL);
809 if (rtc->rtc_ire != NULL)
810 ire_refrele(rtc->rtc_ire);
811 rtc->rtc_ire = ire;
812 rtc->rtc_ipaddr = nexthop;
813 }
814
815 ire->ire_ib_pkt_count++;
816
817 /*
818 * Based on ire_type and ire_flags call one of:
819 * ire_recv_local_v4 - for IRE_LOCAL
820 * ire_recv_loopback_v4 - for IRE_LOOPBACK
821 * ire_recv_multirt_v4 - if RTF_MULTIRT
822 * ire_recv_noroute_v4 - if RTF_REJECT or RTF_BLACHOLE
823 * ire_recv_multicast_v4 - for IRE_MULTICAST
824 * ire_recv_broadcast_v4 - for IRE_BROADCAST
825 * ire_recv_noaccept_v4 - for ire_noaccept ones
826 * ire_recv_forward_v4 - for the rest.
827 */
828 (*ire->ire_recvfn)(ire, mp, ipha, ira);
829 }
830 #undef rptr
831
832 /*
833 * ire_recvfn for IREs that need forwarding
834 */
835 void
836 ire_recv_forward_v4(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
837 {
838 ipha_t *ipha = (ipha_t *)iph_arg;
839 ill_t *ill = ira->ira_ill;
840 ip_stack_t *ipst = ill->ill_ipst;
841 ill_t *dst_ill;
842 nce_t *nce;
843 ipaddr_t src = ipha->ipha_src;
844 uint32_t added_tx_len;
845 uint32_t mtu, iremtu;
846
847 if (ira->ira_flags & (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST)) {
848 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
849 ip_drop_input("l2 multicast not forwarded", mp, ill);
850 freemsg(mp);
851 return;
852 }
853
854 if (!(ill->ill_flags & ILLF_ROUTER) && !ip_source_routed(ipha, ipst)) {
855 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
856 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
857 freemsg(mp);
858 return;
859 }
860
861 /*
862 * Either ire_nce_capable or ire_dep_parent would be set for the IRE
863 * when it is found by ire_route_recursive, but that some other thread
864 * could have changed the routes with the effect of clearing
865 * ire_dep_parent. In that case we'd end up dropping the packet, or
866 * finding a new nce below.
867 * Get, allocate, or update the nce.
868 * We get a refhold on ire_nce_cache as a result of this to avoid races
869 * where ire_nce_cache is deleted.
870 *
871 * This ensures that we don't forward if the interface is down since
872 * ipif_down removes all the nces.
873 */
874 mutex_enter(&ire->ire_lock);
875 nce = ire->ire_nce_cache;
876 if (nce == NULL) {
877 /* Not yet set up - try to set one up */
878 mutex_exit(&ire->ire_lock);
879 (void) ire_revalidate_nce(ire);
880 mutex_enter(&ire->ire_lock);
881 nce = ire->ire_nce_cache;
882 if (nce == NULL) {
883 mutex_exit(&ire->ire_lock);
884 /* The ire_dep_parent chain went bad, or no memory */
885 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
886 ip_drop_input("No ire_dep_parent", mp, ill);
887 freemsg(mp);
888 return;
889 }
890 }
891 nce_refhold(nce);
892 mutex_exit(&ire->ire_lock);
893
894 if (nce->nce_is_condemned) {
895 nce_t *nce1;
896
897 nce1 = ire_handle_condemned_nce(nce, ire, ipha, NULL, B_FALSE);
898 nce_refrele(nce);
899 if (nce1 == NULL) {
900 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
901 ip_drop_input("No nce", mp, ill);
902 freemsg(mp);
903 return;
904 }
905 nce = nce1;
906 }
907 dst_ill = nce->nce_ill;
908
909 /*
910 * Unless we are forwarding, drop the packet.
911 * We have to let source routed packets through if they go out
912 * the same interface i.e., they are 'ping -l' packets.
913 */
914 if (!(dst_ill->ill_flags & ILLF_ROUTER) &&
915 !(ip_source_routed(ipha, ipst) && dst_ill == ill)) {
916 if (ip_source_routed(ipha, ipst)) {
917 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
918 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
919 nce_refrele(nce);
920 return;
921 }
922 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
923 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
924 freemsg(mp);
925 nce_refrele(nce);
926 return;
927 }
928
929 if (ire->ire_zoneid != GLOBAL_ZONEID && ire->ire_zoneid != ALL_ZONES) {
930 ipaddr_t dst = ipha->ipha_dst;
931
932 ire->ire_ib_pkt_count--;
933 /*
934 * Should only use IREs that are visible from the
935 * global zone for forwarding.
936 * Take a source route into account the same way as ip_input
937 * did.
938 */
939 if (ira->ira_flags & IRAF_IPV4_OPTIONS) {
940 int error = 0;
941
942 dst = ip_input_options(ipha, dst, mp, ira, &error);
943 ASSERT(error == 0); /* ip_input checked */
944 }
945 ire = ire_route_recursive_v4(dst, 0, NULL, GLOBAL_ZONEID,
946 ira->ira_tsl, MATCH_IRE_SECATTR,
947 (ill->ill_flags & ILLF_ROUTER) ? IRR_ALLOCATE : IRR_NONE,
948 ira->ira_xmit_hint, ipst, NULL, NULL, NULL);
949 ire->ire_ib_pkt_count++;
950 (*ire->ire_recvfn)(ire, mp, ipha, ira);
951 ire_refrele(ire);
952 nce_refrele(nce);
953 return;
954 }
955
956 /*
957 * ipIfStatsHCInForwDatagrams should only be increment if there
958 * will be an attempt to forward the packet, which is why we
959 * increment after the above condition has been checked.
960 */
961 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInForwDatagrams);
962
963 /* Initiate Read side IPPF processing */
964 if (IPP_ENABLED(IPP_FWD_IN, ipst)) {
965 /* ip_process translates an IS_UNDER_IPMP */
966 mp = ip_process(IPP_FWD_IN, mp, ill, ill);
967 if (mp == NULL) {
968 /* ip_drop_packet and MIB done */
969 ip2dbg(("ire_recv_forward_v4: pkt dropped/deferred "
970 "during IPPF processing\n"));
971 nce_refrele(nce);
972 return;
973 }
974 }
975
976 DTRACE_PROBE4(ip4__forwarding__start,
977 ill_t *, ill, ill_t *, dst_ill, ipha_t *, ipha, mblk_t *, mp);
978
979 if (HOOKS4_INTERESTED_FORWARDING(ipst)) {
980 int error;
981
982 FW_HOOKS(ipst->ips_ip4_forwarding_event,
983 ipst->ips_ipv4firewall_forwarding,
984 ill, dst_ill, ipha, mp, mp, 0, ipst, error);
985
986 DTRACE_PROBE1(ip4__forwarding__end, mblk_t *, mp);
987
988 if (mp == NULL) {
989 nce_refrele(nce);
990 return;
991 }
992 /*
993 * Even if the destination was changed by the filter we use the
994 * forwarding decision that was made based on the address
995 * in ip_input.
996 */
997
998 /* Might have changed */
999 ipha = (ipha_t *)mp->b_rptr;
1000 ira->ira_pktlen = ntohs(ipha->ipha_length);
1001 }
1002
1003 /* Packet is being forwarded. Turning off hwcksum flag. */
1004 DB_CKSUMFLAGS(mp) = 0;
1005
1006 /*
1007 * Martian Address Filtering [RFC 1812, Section 5.3.7]
1008 * The loopback address check for both src and dst has already
1009 * been checked in ip_input
1010 * In the future one can envision adding RPF checks using number 3.
1011 * If we already checked the same source address we can skip this.
1012 */
1013 if (!(ira->ira_flags & IRAF_VERIFIED_SRC) ||
1014 src != ira->ira_verified_src) {
1015 switch (ipst->ips_src_check) {
1016 case 0:
1017 break;
1018 case 2:
1019 if (ip_type_v4(src, ipst) == IRE_BROADCAST) {
1020 BUMP_MIB(ill->ill_ip_mib,
1021 ipIfStatsForwProhibits);
1022 BUMP_MIB(ill->ill_ip_mib,
1023 ipIfStatsInAddrErrors);
1024 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
1025 freemsg(mp);
1026 nce_refrele(nce);
1027 return;
1028 }
1029 /* FALLTHRU */
1030
1031 case 1:
1032 if (CLASSD(src)) {
1033 BUMP_MIB(ill->ill_ip_mib,
1034 ipIfStatsForwProhibits);
1035 BUMP_MIB(ill->ill_ip_mib,
1036 ipIfStatsInAddrErrors);
1037 ip_drop_input("ipIfStatsInAddrErrors", mp, ill);
1038 freemsg(mp);
1039 nce_refrele(nce);
1040 return;
1041 }
1042 break;
1043 }
1044 /* Remember for next packet */
1045 ira->ira_flags |= IRAF_VERIFIED_SRC;
1046 ira->ira_verified_src = src;
1047 }
1048
1049 /*
1050 * Check if packet is going out the same link on which it arrived.
1051 * Means we might need to send a redirect.
1052 */
1053 if (IS_ON_SAME_LAN(dst_ill, ill) && ipst->ips_ip_g_send_redirects) {
1054 ip_send_potential_redirect_v4(mp, ipha, ire, ira);
1055 }
1056
1057 added_tx_len = 0;
1058 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
1059 mblk_t *mp1;
1060 uint32_t old_pkt_len = ira->ira_pktlen;
1061
1062 /* Verify IP header checksum before adding/removing options */
1063 if ((ira->ira_flags & IRAF_VERIFY_IP_CKSUM) &&
1064 ip_csum_hdr(ipha)) {
1065 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
1066 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
1067 freemsg(mp);
1068 nce_refrele(nce);
1069 return;
1070 }
1071 ira->ira_flags &= ~IRAF_VERIFY_IP_CKSUM;
1072
1073 /*
1074 * Check if it can be forwarded and add/remove
1075 * CIPSO options as needed.
1076 */
1077 if ((mp1 = tsol_ip_forward(ire, mp, ira)) == NULL) {
1078 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1079 ip_drop_input("tsol_ip_forward", mp, ill);
1080 freemsg(mp);
1081 nce_refrele(nce);
1082 return;
1083 }
1084 /*
1085 * Size may have changed. Remember amount added in case
1086 * IP needs to send an ICMP too big.
1087 */
1088 mp = mp1;
1089 ipha = (ipha_t *)mp->b_rptr;
1090 ira->ira_pktlen = ntohs(ipha->ipha_length);
1091 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
1092 if (ira->ira_pktlen > old_pkt_len)
1093 added_tx_len = ira->ira_pktlen - old_pkt_len;
1094
1095 /* Options can have been added or removed */
1096 if (ira->ira_ip_hdr_length != IP_SIMPLE_HDR_LENGTH)
1097 ira->ira_flags |= IRAF_IPV4_OPTIONS;
1098 else
1099 ira->ira_flags &= ~IRAF_IPV4_OPTIONS;
1100 }
1101
1102 mtu = dst_ill->ill_mtu;
1103 if ((iremtu = ire->ire_metrics.iulp_mtu) != 0 && iremtu < mtu)
1104 mtu = iremtu;
1105 ip_forward_xmit_v4(nce, ill, mp, ipha, ira, mtu, added_tx_len);
1106 nce_refrele(nce);
1107 }
1108
1109 /*
1110 * Used for sending out unicast and multicast packets that are
1111 * forwarded.
1112 */
1113 void
1114 ip_forward_xmit_v4(nce_t *nce, ill_t *ill, mblk_t *mp, ipha_t *ipha,
1115 ip_recv_attr_t *ira, uint32_t mtu, uint32_t added_tx_len)
1116 {
1117 ill_t *dst_ill = nce->nce_ill;
1118 uint32_t pkt_len;
1119 uint32_t sum;
1120 iaflags_t iraflags = ira->ira_flags;
1121 ip_stack_t *ipst = ill->ill_ipst;
1122 iaflags_t ixaflags;
1123
1124 if (ipha->ipha_ttl <= 1) {
1125 /* Perhaps the checksum was bad */
1126 if ((iraflags & IRAF_VERIFY_IP_CKSUM) && ip_csum_hdr(ipha)) {
1127 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
1128 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
1129 freemsg(mp);
1130 return;
1131 }
1132 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1133 ip_drop_input("ICMP_TTL_EXCEEDED", mp, ill);
1134 icmp_time_exceeded(mp, ICMP_TTL_EXCEEDED, ira);
1135 return;
1136 }
1137 ipha->ipha_ttl--;
1138 /* Adjust the checksum to reflect the ttl decrement. */
1139 sum = (int)ipha->ipha_hdr_checksum + IP_HDR_CSUM_TTL_ADJUST;
1140 ipha->ipha_hdr_checksum = (uint16_t)(sum + (sum >> 16));
1141
1142 /* Check if there are options to update */
1143 if (iraflags & IRAF_IPV4_OPTIONS) {
1144 ASSERT(ipha->ipha_version_and_hdr_length !=
1145 IP_SIMPLE_HDR_VERSION);
1146 ASSERT(!(iraflags & IRAF_VERIFY_IP_CKSUM));
1147
1148 if (!ip_forward_options(mp, ipha, dst_ill, ira)) {
1149 /* ipIfStatsForwProhibits and ip_drop_input done */
1150 return;
1151 }
1152
1153 ipha->ipha_hdr_checksum = 0;
1154 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
1155 }
1156
1157 /* Initiate Write side IPPF processing before any fragmentation */
1158 if (IPP_ENABLED(IPP_FWD_OUT, ipst)) {
1159 /* ip_process translates an IS_UNDER_IPMP */
1160 mp = ip_process(IPP_FWD_OUT, mp, dst_ill, dst_ill);
1161 if (mp == NULL) {
1162 /* ip_drop_packet and MIB done */
1163 ip2dbg(("ire_recv_forward_v4: pkt dropped/deferred" \
1164 " during IPPF processing\n"));
1165 return;
1166 }
1167 }
1168
1169 pkt_len = ira->ira_pktlen;
1170
1171 BUMP_MIB(dst_ill->ill_ip_mib, ipIfStatsHCOutForwDatagrams);
1172
1173 ixaflags = IXAF_IS_IPV4 | IXAF_NO_DEV_FLOW_CTL;
1174
1175 if (pkt_len > mtu) {
1176 /*
1177 * It needs fragging on its way out. If we haven't
1178 * verified the header checksum yet we do it now since
1179 * are going to put a surely good checksum in the
1180 * outgoing header, we have to make sure that it
1181 * was good coming in.
1182 */
1183 if ((iraflags & IRAF_VERIFY_IP_CKSUM) && ip_csum_hdr(ipha)) {
1184 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
1185 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
1186 freemsg(mp);
1187 return;
1188 }
1189 if (ipha->ipha_fragment_offset_and_flags & IPH_DF_HTONS) {
1190 BUMP_MIB(dst_ill->ill_ip_mib, ipIfStatsOutFragFails);
1191 ip_drop_output("ipIfStatsOutFragFails", mp, dst_ill);
1192 if (iraflags & IRAF_SYSTEM_LABELED) {
1193 /*
1194 * Remove any CIPSO option added by
1195 * tsol_ip_forward, and make sure we report
1196 * a path MTU so that there
1197 * is room to add such a CIPSO option for future
1198 * packets.
1199 */
1200 mtu = tsol_pmtu_adjust(mp, mtu, added_tx_len,
1201 AF_INET);
1202 }
1203
1204 icmp_frag_needed(mp, mtu, ira);
1205 return;
1206 }
1207
1208 (void) ip_fragment_v4(mp, nce, ixaflags, pkt_len, mtu,
1209 ira->ira_xmit_hint, GLOBAL_ZONEID, 0, ip_xmit, NULL);
1210 return;
1211 }
1212
1213 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
1214 if (iraflags & IRAF_LOOPBACK_COPY) {
1215 /*
1216 * IXAF_NO_LOOP_ZONEID is not set hence 7th arg
1217 * is don't care
1218 */
1219 (void) ip_postfrag_loopcheck(mp, nce,
1220 ixaflags | IXAF_LOOPBACK_COPY,
1221 pkt_len, ira->ira_xmit_hint, GLOBAL_ZONEID, 0, NULL);
1222 } else {
1223 (void) ip_xmit(mp, nce, ixaflags, pkt_len, ira->ira_xmit_hint,
1224 GLOBAL_ZONEID, 0, NULL);
1225 }
1226 }
1227
1228 /*
1229 * ire_recvfn for RTF_REJECT and RTF_BLACKHOLE routes, including IRE_NOROUTE,
1230 * which is what ire_route_recursive returns when there is no matching ire.
1231 * Send ICMP unreachable unless blackhole.
1232 */
1233 void
1234 ire_recv_noroute_v4(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1235 {
1236 ipha_t *ipha = (ipha_t *)iph_arg;
1237 ill_t *ill = ira->ira_ill;
1238 ip_stack_t *ipst = ill->ill_ipst;
1239
1240 /* Would we have forwarded this packet if we had a route? */
1241 if (ira->ira_flags & (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST)) {
1242 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1243 ip_drop_input("l2 multicast not forwarded", mp, ill);
1244 freemsg(mp);
1245 return;
1246 }
1247
1248 if (!(ill->ill_flags & ILLF_ROUTER)) {
1249 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1250 ip_drop_input("ipIfStatsForwProhibits", mp, ill);
1251 freemsg(mp);
1252 return;
1253 }
1254 /*
1255 * If we had a route this could have been forwarded. Count as such.
1256 *
1257 * ipIfStatsHCInForwDatagrams should only be increment if there
1258 * will be an attempt to forward the packet, which is why we
1259 * increment after the above condition has been checked.
1260 */
1261 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInForwDatagrams);
1262
1263 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInNoRoutes);
1264
1265 ip_rts_change(RTM_MISS, ipha->ipha_dst, 0, 0, 0, 0, 0, 0, RTA_DST,
1266 ipst);
1267
1268 if (ire->ire_flags & RTF_BLACKHOLE) {
1269 ip_drop_input("ipIfStatsInNoRoutes RTF_BLACKHOLE", mp, ill);
1270 freemsg(mp);
1271 } else {
1272 ip_drop_input("ipIfStatsInNoRoutes RTF_REJECT", mp, ill);
1273
1274 if (ip_source_routed(ipha, ipst)) {
1275 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
1276 } else {
1277 icmp_unreachable(mp, ICMP_HOST_UNREACHABLE, ira);
1278 }
1279 }
1280 }
1281
1282 /*
1283 * ire_recvfn for IRE_LOCALs marked with ire_noaccept. Such IREs are used for
1284 * VRRP when in noaccept mode.
1285 * We silently drop the packet. ARP handles packets even if noaccept is set.
1286 */
1287 /* ARGSUSED */
1288 void
1289 ire_recv_noaccept_v4(ire_t *ire, mblk_t *mp, void *iph_arg,
1290 ip_recv_attr_t *ira)
1291 {
1292 ill_t *ill = ira->ira_ill;
1293
1294 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1295 ip_drop_input("ipIfStatsInDiscards - noaccept", mp, ill);
1296 freemsg(mp);
1297 }
1298
1299 /*
1300 * ire_recvfn for IRE_BROADCAST.
1301 */
1302 void
1303 ire_recv_broadcast_v4(ire_t *ire, mblk_t *mp, void *iph_arg,
1304 ip_recv_attr_t *ira)
1305 {
1306 ipha_t *ipha = (ipha_t *)iph_arg;
1307 ill_t *ill = ira->ira_ill;
1308 ill_t *dst_ill = ire->ire_ill;
1309 ip_stack_t *ipst = ill->ill_ipst;
1310 ire_t *alt_ire;
1311 nce_t *nce;
1312 ipaddr_t ipha_dst;
1313
1314 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInBcastPkts);
1315
1316 /* Tag for higher-level protocols */
1317 ira->ira_flags |= IRAF_BROADCAST;
1318
1319 /*
1320 * Whether local or directed broadcast forwarding: don't allow
1321 * for TCP.
1322 */
1323 if (ipha->ipha_protocol == IPPROTO_TCP) {
1324 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1325 ip_drop_input("ipIfStatsInDiscards", mp, ill);
1326 freemsg(mp);
1327 return;
1328 }
1329
1330 /*
1331 * So that we don't end up with dups, only one ill an IPMP group is
1332 * nominated to receive broadcast traffic.
1333 * If we have no cast_ill we are liberal and accept everything.
1334 */
1335 if (IS_UNDER_IPMP(ill)) {
1336 /* For an under ill_grp can change under lock */
1337 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
1338 if (!ill->ill_nom_cast && ill->ill_grp != NULL &&
1339 ill->ill_grp->ig_cast_ill != NULL) {
1340 rw_exit(&ipst->ips_ill_g_lock);
1341 /* No MIB since this is normal operation */
1342 ip_drop_input("not nom_cast", mp, ill);
1343 freemsg(mp);
1344 return;
1345 }
1346 rw_exit(&ipst->ips_ill_g_lock);
1347
1348 ira->ira_ruifindex = ill_get_upper_ifindex(ill);
1349 }
1350
1351 /*
1352 * After reassembly and IPsec we will need to duplicate the
1353 * broadcast packet for all matching zones on the ill.
1354 */
1355 ira->ira_zoneid = ALL_ZONES;
1356
1357 /*
1358 * Check for directed broadcast i.e. ire->ire_ill is different than
1359 * the incoming ill.
1360 * The same broadcast address can be assigned to multiple interfaces
1361 * so have to check explicitly for that case by looking up the alt_ire
1362 */
1363 if (dst_ill == ill && !(ire->ire_flags & RTF_MULTIRT)) {
1364 /* Reassemble on the ill on which the packet arrived */
1365 ip_input_local_v4(ire, mp, ipha, ira);
1366 /* Restore */
1367 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1368 return;
1369 }
1370
1371 /* Is there an IRE_BROADCAST on the incoming ill? */
1372 ipha_dst = ((ira->ira_flags & IRAF_DHCP_UNICAST) ? INADDR_BROADCAST :
1373 ipha->ipha_dst);
1374 alt_ire = ire_ftable_lookup_v4(ipha_dst, 0, 0, IRE_BROADCAST, ill,
1375 ALL_ZONES, ira->ira_tsl,
1376 MATCH_IRE_TYPE|MATCH_IRE_ILL|MATCH_IRE_SECATTR, 0, ipst, NULL);
1377 if (alt_ire != NULL) {
1378 /* Not a directed broadcast */
1379 /*
1380 * In the special case of multirouted broadcast
1381 * packets, we unconditionally need to "gateway"
1382 * them to the appropriate interface here so that reassembly
1383 * works. We know that the IRE_BROADCAST on cgtp0 doesn't
1384 * have RTF_MULTIRT set so we look for such an IRE in the
1385 * bucket.
1386 */
1387 if (alt_ire->ire_flags & RTF_MULTIRT) {
1388 irb_t *irb;
1389 ire_t *ire1;
1390
1391 irb = ire->ire_bucket;
1392 irb_refhold(irb);
1393 for (ire1 = irb->irb_ire; ire1 != NULL;
1394 ire1 = ire1->ire_next) {
1395 if (IRE_IS_CONDEMNED(ire1))
1396 continue;
1397 if (!(ire1->ire_type & IRE_BROADCAST) ||
1398 (ire1->ire_flags & RTF_MULTIRT))
1399 continue;
1400 ill = ire1->ire_ill;
1401 ill_refhold(ill);
1402 break;
1403 }
1404 irb_refrele(irb);
1405 if (ire1 != NULL) {
1406 ill_t *orig_ill = ira->ira_ill;
1407
1408 ire_refrele(alt_ire);
1409 /* Reassemble on the new ill */
1410 ira->ira_ill = ill;
1411 ip_input_local_v4(ire, mp, ipha, ira);
1412 ill_refrele(ill);
1413 /* Restore */
1414 ira->ira_ill = orig_ill;
1415 ira->ira_ruifindex =
1416 orig_ill->ill_phyint->phyint_ifindex;
1417 return;
1418 }
1419 }
1420 ire_refrele(alt_ire);
1421 /* Reassemble on the ill on which the packet arrived */
1422 ip_input_local_v4(ire, mp, ipha, ira);
1423 goto done;
1424 }
1425
1426 /*
1427 * This is a directed broadcast
1428 *
1429 * If directed broadcast is allowed, then forward the packet out
1430 * the destination interface with IXAF_LOOPBACK_COPY set. That will
1431 * result in ip_input() receiving a copy of the packet on the
1432 * appropriate ill. (We could optimize this to avoid the extra trip
1433 * via ip_input(), but since directed broadcasts are normally disabled
1434 * it doesn't make sense to optimize it.)
1435 */
1436 if (!ipst->ips_ip_g_forward_directed_bcast ||
1437 (ira->ira_flags & (IRAF_L2DST_MULTICAST|IRAF_L2DST_BROADCAST))) {
1438 ip_drop_input("directed broadcast not allowed", mp, ill);
1439 freemsg(mp);
1440 goto done;
1441 }
1442 if ((ira->ira_flags & IRAF_VERIFY_IP_CKSUM) && ip_csum_hdr(ipha)) {
1443 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
1444 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
1445 freemsg(mp);
1446 goto done;
1447 }
1448
1449 /*
1450 * Clear the indication that this may have hardware
1451 * checksum as we are not using it for forwarding.
1452 */
1453 DB_CKSUMFLAGS(mp) = 0;
1454
1455 /*
1456 * Adjust ttl to 2 (1+1 - the forward engine will decrement it by one.
1457 */
1458 ipha->ipha_ttl = ipst->ips_ip_broadcast_ttl + 1;
1459 ipha->ipha_hdr_checksum = 0;
1460 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
1461
1462 /*
1463 * We use ip_forward_xmit to do any fragmentation.
1464 * and loopback copy on the outbound interface.
1465 *
1466 * Make it so that IXAF_LOOPBACK_COPY to be set on transmit side.
1467 */
1468 ira->ira_flags |= IRAF_LOOPBACK_COPY;
1469
1470 nce = arp_nce_init(dst_ill, ipha->ipha_dst, IRE_BROADCAST);
1471 if (nce == NULL) {
1472 BUMP_MIB(dst_ill->ill_ip_mib, ipIfStatsOutDiscards);
1473 ip_drop_output("No nce", mp, dst_ill);
1474 freemsg(mp);
1475 goto done;
1476 }
1477
1478 ip_forward_xmit_v4(nce, ill, mp, ipha, ira, dst_ill->ill_mc_mtu, 0);
1479 nce_refrele(nce);
1480 done:
1481 /* Restore */
1482 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1483 }
1484
1485 /*
1486 * ire_recvfn for IRE_MULTICAST.
1487 */
1488 void
1489 ire_recv_multicast_v4(ire_t *ire, mblk_t *mp, void *iph_arg,
1490 ip_recv_attr_t *ira)
1491 {
1492 ipha_t *ipha = (ipha_t *)iph_arg;
1493 ill_t *ill = ira->ira_ill;
1494 ip_stack_t *ipst = ill->ill_ipst;
1495
1496 ASSERT(ire->ire_ill == ira->ira_ill);
1497
1498 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInMcastPkts);
1499 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCInMcastOctets, ira->ira_pktlen);
1500
1501 /* RSVP hook */
1502 if (ira->ira_flags & IRAF_RSVP)
1503 goto forus;
1504
1505 /* Tag for higher-level protocols */
1506 ira->ira_flags |= IRAF_MULTICAST;
1507
1508 /*
1509 * So that we don't end up with dups, only one ill an IPMP group is
1510 * nominated to receive multicast traffic.
1511 * If we have no cast_ill we are liberal and accept everything.
1512 */
1513 if (IS_UNDER_IPMP(ill)) {
1514 ip_stack_t *ipst = ill->ill_ipst;
1515
1516 /* For an under ill_grp can change under lock */
1517 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
1518 if (!ill->ill_nom_cast && ill->ill_grp != NULL &&
1519 ill->ill_grp->ig_cast_ill != NULL) {
1520 rw_exit(&ipst->ips_ill_g_lock);
1521 ip_drop_input("not on cast ill", mp, ill);
1522 freemsg(mp);
1523 return;
1524 }
1525 rw_exit(&ipst->ips_ill_g_lock);
1526 /*
1527 * We switch to the upper ill so that mrouter and hasmembers
1528 * can operate on upper here and in ip_input_multicast.
1529 */
1530 ill = ipmp_ill_hold_ipmp_ill(ill);
1531 if (ill != NULL) {
1532 ASSERT(ill != ira->ira_ill);
1533 ASSERT(ire->ire_ill == ira->ira_ill);
1534 ira->ira_ill = ill;
1535 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1536 } else {
1537 ill = ira->ira_ill;
1538 }
1539 }
1540
1541 /*
1542 * Check if we are a multicast router - send ip_mforward a copy of
1543 * the packet.
1544 * Due to mroute_decap tunnels we consider forwarding packets even if
1545 * mrouted has not joined the allmulti group on this interface.
1546 */
1547 if (ipst->ips_ip_g_mrouter) {
1548 int retval;
1549
1550 /*
1551 * Clear the indication that this may have hardware
1552 * checksum as we are not using it for forwarding.
1553 */
1554 DB_CKSUMFLAGS(mp) = 0;
1555
1556 /*
1557 * ip_mforward helps us make these distinctions: If received
1558 * on tunnel and not IGMP, then drop.
1559 * If IGMP packet, then don't check membership
1560 * If received on a phyint and IGMP or PIM, then
1561 * don't check membership
1562 */
1563 retval = ip_mforward(mp, ira);
1564 /* ip_mforward updates mib variables if needed */
1565
1566 switch (retval) {
1567 case 0:
1568 /*
1569 * pkt is okay and arrived on phyint.
1570 *
1571 * If we are running as a multicast router
1572 * we need to see all IGMP and/or PIM packets.
1573 */
1574 if ((ipha->ipha_protocol == IPPROTO_IGMP) ||
1575 (ipha->ipha_protocol == IPPROTO_PIM)) {
1576 goto forus;
1577 }
1578 break;
1579 case -1:
1580 /* pkt is mal-formed, toss it */
1581 freemsg(mp);
1582 goto done;
1583 case 1:
1584 /*
1585 * pkt is okay and arrived on a tunnel
1586 *
1587 * If we are running a multicast router
1588 * we need to see all igmp packets.
1589 */
1590 if (ipha->ipha_protocol == IPPROTO_IGMP) {
1591 goto forus;
1592 }
1593 ip_drop_input("Multicast on tunnel ignored", mp, ill);
1594 freemsg(mp);
1595 goto done;
1596 }
1597 }
1598
1599 /*
1600 * Check if we have members on this ill. This is not necessary for
1601 * correctness because even if the NIC/GLD had a leaky filter, we
1602 * filter before passing to each conn_t.
1603 */
1604 if (!ill_hasmembers_v4(ill, ipha->ipha_dst)) {
1605 /*
1606 * Nobody interested
1607 *
1608 * This might just be caused by the fact that
1609 * multiple IP Multicast addresses map to the same
1610 * link layer multicast - no need to increment counter!
1611 */
1612 ip_drop_input("Multicast with no members", mp, ill);
1613 freemsg(mp);
1614 goto done;
1615 }
1616 forus:
1617 ip2dbg(("ire_recv_multicast_v4: multicast for us: 0x%x\n",
1618 ntohl(ipha->ipha_dst)));
1619
1620 /*
1621 * After reassembly and IPsec we will need to duplicate the
1622 * multicast packet for all matching zones on the ill.
1623 */
1624 ira->ira_zoneid = ALL_ZONES;
1625
1626 /* Reassemble on the ill on which the packet arrived */
1627 ip_input_local_v4(ire, mp, ipha, ira);
1628 done:
1629 if (ill != ire->ire_ill) {
1630 ill_refrele(ill);
1631 ira->ira_ill = ire->ire_ill;
1632 ira->ira_ruifindex = ira->ira_ill->ill_phyint->phyint_ifindex;
1633 }
1634 }
1635
1636 /*
1637 * ire_recvfn for IRE_OFFLINK with RTF_MULTIRT.
1638 * Drop packets since we don't forward out multirt routes.
1639 */
1640 /* ARGSUSED */
1641 void
1642 ire_recv_multirt_v4(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1643 {
1644 ill_t *ill = ira->ira_ill;
1645
1646 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInNoRoutes);
1647 ip_drop_input("Not forwarding out MULTIRT", mp, ill);
1648 freemsg(mp);
1649 }
1650
1651 /*
1652 * ire_recvfn for IRE_LOOPBACK. This is only used when a FW_HOOK
1653 * has rewritten the packet to have a loopback destination address (We
1654 * filter out packet with a loopback destination from arriving over the wire).
1655 * We don't know what zone to use, thus we always use the GLOBAL_ZONEID.
1656 */
1657 void
1658 ire_recv_loopback_v4(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1659 {
1660 ipha_t *ipha = (ipha_t *)iph_arg;
1661 ill_t *ill = ira->ira_ill;
1662 ill_t *ire_ill = ire->ire_ill;
1663
1664 ira->ira_zoneid = GLOBAL_ZONEID;
1665
1666 /* Switch to the lo0 ill for further processing */
1667 if (ire_ill != ill) {
1668 /*
1669 * Update ira_ill to be the ILL on which the IP address
1670 * is hosted.
1671 * No need to hold the ill since we have a hold on the ire
1672 */
1673 ASSERT(ira->ira_ill == ira->ira_rill);
1674 ira->ira_ill = ire_ill;
1675
1676 ip_input_local_v4(ire, mp, ipha, ira);
1677
1678 /* Restore */
1679 ASSERT(ira->ira_ill == ire_ill);
1680 ira->ira_ill = ill;
1681 return;
1682
1683 }
1684 ip_input_local_v4(ire, mp, ipha, ira);
1685 }
1686
1687 /*
1688 * ire_recvfn for IRE_LOCAL.
1689 */
1690 void
1691 ire_recv_local_v4(ire_t *ire, mblk_t *mp, void *iph_arg, ip_recv_attr_t *ira)
1692 {
1693 ipha_t *ipha = (ipha_t *)iph_arg;
1694 ill_t *ill = ira->ira_ill;
1695 ill_t *ire_ill = ire->ire_ill;
1696
1697 /* Make a note for DAD that this address is in use */
1698 ire->ire_last_used_time = LBOLT_FASTPATH;
1699
1700 /* Only target the IRE_LOCAL with the right zoneid. */
1701 ira->ira_zoneid = ire->ire_zoneid;
1702
1703 /*
1704 * If the packet arrived on the wrong ill, we check that
1705 * this is ok.
1706 * If it is, then we ensure that we do the reassembly on
1707 * the ill on which the address is hosted. We keep ira_rill as
1708 * the one on which the packet arrived, so that IP_PKTINFO and
1709 * friends can report this.
1710 */
1711 if (ire_ill != ill) {
1712 ire_t *new_ire;
1713
1714 new_ire = ip_check_multihome(&ipha->ipha_dst, ire, ill);
1715 if (new_ire == NULL) {
1716 /* Drop packet */
1717 BUMP_MIB(ill->ill_ip_mib, ipIfStatsForwProhibits);
1718 ip_drop_input("ipIfStatsInForwProhibits", mp, ill);
1719 freemsg(mp);
1720 return;
1721 }
1722 /*
1723 * Update ira_ill to be the ILL on which the IP address
1724 * is hosted. No need to hold the ill since we have a
1725 * hold on the ire. Note that we do the switch even if
1726 * new_ire == ire (for IPMP, ire would be the one corresponding
1727 * to the IPMP ill).
1728 */
1729 ASSERT(ira->ira_ill == ira->ira_rill);
1730 ira->ira_ill = new_ire->ire_ill;
1731
1732 /* ira_ruifindex tracks the upper for ira_rill */
1733 if (IS_UNDER_IPMP(ill))
1734 ira->ira_ruifindex = ill_get_upper_ifindex(ill);
1735
1736 ip_input_local_v4(new_ire, mp, ipha, ira);
1737
1738 /* Restore */
1739 ASSERT(ira->ira_ill == new_ire->ire_ill);
1740 ira->ira_ill = ill;
1741 ira->ira_ruifindex = ill->ill_phyint->phyint_ifindex;
1742
1743 if (new_ire != ire)
1744 ire_refrele(new_ire);
1745 return;
1746 }
1747
1748 ip_input_local_v4(ire, mp, ipha, ira);
1749 }
1750
1751 /*
1752 * Common function for packets arriving for the host. Handles
1753 * checksum verification, reassembly checks, etc.
1754 */
1755 static void
1756 ip_input_local_v4(ire_t *ire, mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
1757 {
1758 ill_t *ill = ira->ira_ill;
1759 iaflags_t iraflags = ira->ira_flags;
1760
1761 /*
1762 * Verify IP header checksum. If the packet was AH or ESP then
1763 * this flag has already been cleared. Likewise if the packet
1764 * had a hardware checksum.
1765 */
1766 if ((iraflags & IRAF_VERIFY_IP_CKSUM) && ip_csum_hdr(ipha)) {
1767 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInCksumErrs);
1768 ip_drop_input("ipIfStatsInCksumErrs", mp, ill);
1769 freemsg(mp);
1770 return;
1771 }
1772
1773 if (iraflags & IRAF_IPV4_OPTIONS) {
1774 if (!ip_input_local_options(mp, ipha, ira)) {
1775 /* Error has been sent and mp consumed */
1776 return;
1777 }
1778 /*
1779 * Some old hardware does partial checksum by including the
1780 * whole IP header, so the partial checksum value might have
1781 * become invalid if any option in the packet have been
1782 * updated. Always clear partial checksum flag here.
1783 */
1784 DB_CKSUMFLAGS(mp) &= ~HCK_PARTIALCKSUM;
1785 }
1786
1787 /*
1788 * Is packet part of fragmented IP packet?
1789 * We compare against defined values in network byte order
1790 */
1791 if (ipha->ipha_fragment_offset_and_flags &
1792 (IPH_MF_HTONS | IPH_OFFSET_HTONS)) {
1793 /*
1794 * Make sure we have ira_l2src before we loose the original
1795 * mblk
1796 */
1797 if (!(ira->ira_flags & IRAF_L2SRC_SET))
1798 ip_setl2src(mp, ira, ira->ira_rill);
1799
1800 mp = ip_input_fragment(mp, ipha, ira);
1801 if (mp == NULL)
1802 return;
1803 /* Completed reassembly */
1804 ipha = (ipha_t *)mp->b_rptr;
1805 }
1806
1807 /*
1808 * For broadcast and multicast we need some extra work before
1809 * we call ip_fanout_v4(), since in the case of shared-IP zones
1810 * we need to pretend that a packet arrived for each zoneid.
1811 */
1812 if (iraflags & IRAF_MULTIBROADCAST) {
1813 if (iraflags & IRAF_BROADCAST)
1814 ip_input_broadcast_v4(ire, mp, ipha, ira);
1815 else
1816 ip_input_multicast_v4(ire, mp, ipha, ira);
1817 return;
1818 }
1819 ip_fanout_v4(mp, ipha, ira);
1820 }
1821
1822
1823 /*
1824 * Handle multiple zones which match the same broadcast address
1825 * and ill by delivering a packet to each of them.
1826 * Walk the bucket and look for different ire_zoneid but otherwise
1827 * the same IRE (same ill/addr/mask/type).
1828 * Note that ire_add() tracks IREs that are identical in all
1829 * fields (addr/mask/type/gw/ill/zoneid) within a single IRE by
1830 * increasing ire_identical_cnt. Thus we don't need to be concerned
1831 * about those.
1832 */
1833 static void
1834 ip_input_broadcast_v4(ire_t *ire, mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
1835 {
1836 ill_t *ill = ira->ira_ill;
1837 ip_stack_t *ipst = ill->ill_ipst;
1838 netstack_t *ns = ipst->ips_netstack;
1839 irb_t *irb;
1840 ire_t *ire1;
1841 mblk_t *mp1;
1842 ipha_t *ipha1;
1843 uint_t ira_pktlen = ira->ira_pktlen;
1844 uint16_t ira_ip_hdr_length = ira->ira_ip_hdr_length;
1845
1846 irb = ire->ire_bucket;
1847
1848 /*
1849 * If we don't have more than one shared-IP zone, or if
1850 * there can't be more than one IRE_BROADCAST for this
1851 * IP address, then just set the zoneid and proceed.
1852 */
1853 if (ns->netstack_numzones == 1 || irb->irb_ire_cnt == 1) {
1854 ira->ira_zoneid = ire->ire_zoneid;
1855
1856 ip_fanout_v4(mp, ipha, ira);
1857 return;
1858 }
1859 irb_refhold(irb);
1860 for (ire1 = irb->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) {
1861 /* We do the main IRE after the end of the loop */
1862 if (ire1 == ire)
1863 continue;
1864
1865 /*
1866 * Only IREs for the same IP address should be in the same
1867 * bucket.
1868 * But could have IRE_HOSTs in the case of CGTP.
1869 */
1870 ASSERT(ire1->ire_addr == ire->ire_addr);
1871 if (!(ire1->ire_type & IRE_BROADCAST))
1872 continue;
1873
1874 if (IRE_IS_CONDEMNED(ire1))
1875 continue;
1876
1877 mp1 = copymsg(mp);
1878 if (mp1 == NULL) {
1879 /* Failed to deliver to one zone */
1880 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1881 ip_drop_input("ipIfStatsInDiscards", mp, ill);
1882 continue;
1883 }
1884 ira->ira_zoneid = ire1->ire_zoneid;
1885 ipha1 = (ipha_t *)mp1->b_rptr;
1886 ip_fanout_v4(mp1, ipha1, ira);
1887 /*
1888 * IPsec might have modified ira_pktlen and ira_ip_hdr_length
1889 * so we restore them for a potential next iteration
1890 */
1891 ira->ira_pktlen = ira_pktlen;
1892 ira->ira_ip_hdr_length = ira_ip_hdr_length;
1893 }
1894 irb_refrele(irb);
1895 /* Do the main ire */
1896 ira->ira_zoneid = ire->ire_zoneid;
1897 ip_fanout_v4(mp, ipha, ira);
1898 }
1899
1900 /*
1901 * Handle multiple zones which want to receive the same multicast packets
1902 * on this ill by delivering a packet to each of them.
1903 *
1904 * Note that for packets delivered to transports we could instead do this
1905 * as part of the fanout code, but since we need to handle icmp_inbound
1906 * it is simpler to have multicast work the same as broadcast.
1907 *
1908 * The ip_fanout matching for multicast matches based on ilm independent of
1909 * zoneid since the zoneid restriction is applied when joining a multicast
1910 * group.
1911 */
1912 /* ARGSUSED */
1913 static void
1914 ip_input_multicast_v4(ire_t *ire, mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
1915 {
1916 ill_t *ill = ira->ira_ill;
1917 iaflags_t iraflags = ira->ira_flags;
1918 ip_stack_t *ipst = ill->ill_ipst;
1919 netstack_t *ns = ipst->ips_netstack;
1920 zoneid_t zoneid;
1921 mblk_t *mp1;
1922 ipha_t *ipha1;
1923 uint_t ira_pktlen = ira->ira_pktlen;
1924 uint16_t ira_ip_hdr_length = ira->ira_ip_hdr_length;
1925
1926 /* ire_recv_multicast has switched to the upper ill for IPMP */
1927 ASSERT(!IS_UNDER_IPMP(ill));
1928
1929 /*
1930 * If we don't have more than one shared-IP zone, or if
1931 * there are no members in anything but the global zone,
1932 * then just set the zoneid and proceed.
1933 */
1934 if (ns->netstack_numzones == 1 ||
1935 !ill_hasmembers_otherzones_v4(ill, ipha->ipha_dst,
1936 GLOBAL_ZONEID)) {
1937 ira->ira_zoneid = GLOBAL_ZONEID;
1938
1939 /* If sender didn't want this zone to receive it, drop */
1940 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1941 ira->ira_no_loop_zoneid == ira->ira_zoneid) {
1942 ip_drop_input("Multicast but wrong zoneid", mp, ill);
1943 freemsg(mp);
1944 return;
1945 }
1946 ip_fanout_v4(mp, ipha, ira);
1947 return;
1948 }
1949
1950 /*
1951 * Here we loop over all zoneids that have members in the group
1952 * and deliver a packet to ip_fanout for each zoneid.
1953 *
1954 * First find any members in the lowest numeric zoneid by looking for
1955 * first zoneid larger than -1 (ALL_ZONES).
1956 * We terminate the loop when we receive -1 (ALL_ZONES).
1957 */
1958 zoneid = ill_hasmembers_nextzone_v4(ill, ipha->ipha_dst, ALL_ZONES);
1959 for (; zoneid != ALL_ZONES;
1960 zoneid = ill_hasmembers_nextzone_v4(ill, ipha->ipha_dst, zoneid)) {
1961 /*
1962 * Avoid an extra copymsg/freemsg by skipping global zone here
1963 * and doing that at the end.
1964 */
1965 if (zoneid == GLOBAL_ZONEID)
1966 continue;
1967
1968 ira->ira_zoneid = zoneid;
1969
1970 /* If sender didn't want this zone to receive it, skip */
1971 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1972 ira->ira_no_loop_zoneid == ira->ira_zoneid)
1973 continue;
1974
1975 mp1 = copymsg(mp);
1976 if (mp1 == NULL) {
1977 /* Failed to deliver to one zone */
1978 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1979 ip_drop_input("ipIfStatsInDiscards", mp, ill);
1980 continue;
1981 }
1982 ipha1 = (ipha_t *)mp1->b_rptr;
1983 ip_fanout_v4(mp1, ipha1, ira);
1984 /*
1985 * IPsec might have modified ira_pktlen and ira_ip_hdr_length
1986 * so we restore them for a potential next iteration
1987 */
1988 ira->ira_pktlen = ira_pktlen;
1989 ira->ira_ip_hdr_length = ira_ip_hdr_length;
1990 }
1991
1992 /* Do the main ire */
1993 ira->ira_zoneid = GLOBAL_ZONEID;
1994 /* If sender didn't want this zone to receive it, drop */
1995 if ((iraflags & IRAF_NO_LOOP_ZONEID_SET) &&
1996 ira->ira_no_loop_zoneid == ira->ira_zoneid) {
1997 ip_drop_input("Multicast but wrong zoneid", mp, ill);
1998 freemsg(mp);
1999 } else {
2000 ip_fanout_v4(mp, ipha, ira);
2001 }
2002 }
2003
2004
2005 /*
2006 * Determine the zoneid and IRAF_TX_* flags if trusted extensions
2007 * is in use. Updates ira_zoneid and ira_flags as a result.
2008 */
2009 static void
2010 ip_fanout_tx_v4(mblk_t *mp, ipha_t *ipha, uint8_t protocol,
2011 uint_t ip_hdr_length, ip_recv_attr_t *ira)
2012 {
2013 uint16_t *up;
2014 uint16_t lport;
2015 zoneid_t zoneid;
2016
2017 ASSERT(ira->ira_flags & IRAF_SYSTEM_LABELED);
2018
2019 /*
2020 * If the packet is unlabeled we might allow read-down
2021 * for MAC_EXEMPT. Below we clear this if it is a multi-level
2022 * port (MLP).
2023 * Note that ira_tsl can be NULL here.
2024 */
2025 if (ira->ira_tsl != NULL && ira->ira_tsl->tsl_flags & TSLF_UNLABELED)
2026 ira->ira_flags |= IRAF_TX_MAC_EXEMPTABLE;
2027
2028 if (ira->ira_zoneid != ALL_ZONES)
2029 return;
2030
2031 ira->ira_flags |= IRAF_TX_SHARED_ADDR;
2032
2033 up = (uint16_t *)((uchar_t *)ipha + ip_hdr_length);
2034 switch (protocol) {
2035 case IPPROTO_TCP:
2036 case IPPROTO_SCTP:
2037 case IPPROTO_UDP:
2038 /* Caller ensures this */
2039 ASSERT(((uchar_t *)ipha) + ip_hdr_length +4 <= mp->b_wptr);
2040
2041 /*
2042 * Only these transports support MLP.
2043 * We know their destination port numbers is in
2044 * the same place in the header.
2045 */
2046 lport = up[1];
2047
2048 /*
2049 * No need to handle exclusive-stack zones
2050 * since ALL_ZONES only applies to the shared IP instance.
2051 */
2052 zoneid = tsol_mlp_findzone(protocol, lport);
2053 /*
2054 * If no shared MLP is found, tsol_mlp_findzone returns
2055 * ALL_ZONES. In that case, we assume it's SLP, and
2056 * search for the zone based on the packet label.
2057 *
2058 * If there is such a zone, we prefer to find a
2059 * connection in it. Otherwise, we look for a
2060 * MAC-exempt connection in any zone whose label
2061 * dominates the default label on the packet.
2062 */
2063 if (zoneid == ALL_ZONES)
2064 zoneid = tsol_attr_to_zoneid(ira);
2065 else
2066 ira->ira_flags &= ~IRAF_TX_MAC_EXEMPTABLE;
2067 break;
2068 default:
2069 /* Handle shared address for other protocols */
2070 zoneid = tsol_attr_to_zoneid(ira);
2071 break;
2072 }
2073 ira->ira_zoneid = zoneid;
2074 }
2075
2076 /*
2077 * Increment checksum failure statistics
2078 */
2079 static void
2080 ip_input_cksum_err_v4(uint8_t protocol, uint16_t hck_flags, ill_t *ill)
2081 {
2082 ip_stack_t *ipst = ill->ill_ipst;
2083
2084 switch (protocol) {
2085 case IPPROTO_TCP:
2086 BUMP_MIB(ill->ill_ip_mib, tcpIfStatsInErrs);
2087
2088 if (hck_flags & HCK_FULLCKSUM)
2089 IP_STAT(ipst, ip_tcp_in_full_hw_cksum_err);
2090 else if (hck_flags & HCK_PARTIALCKSUM)
2091 IP_STAT(ipst, ip_tcp_in_part_hw_cksum_err);
2092 else
2093 IP_STAT(ipst, ip_tcp_in_sw_cksum_err);
2094 break;
2095 case IPPROTO_UDP:
2096 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInCksumErrs);
2097 if (hck_flags & HCK_FULLCKSUM)
2098 IP_STAT(ipst, ip_udp_in_full_hw_cksum_err);
2099 else if (hck_flags & HCK_PARTIALCKSUM)
2100 IP_STAT(ipst, ip_udp_in_part_hw_cksum_err);
2101 else
2102 IP_STAT(ipst, ip_udp_in_sw_cksum_err);
2103 break;
2104 case IPPROTO_ICMP:
2105 BUMP_MIB(&ipst->ips_icmp_mib, icmpInCksumErrs);
2106 break;
2107 default:
2108 ASSERT(0);
2109 break;
2110 }
2111 }
2112
2113 /* Calculate the IPv4 pseudo-header checksum */
2114 uint32_t
2115 ip_input_cksum_pseudo_v4(ipha_t *ipha, ip_recv_attr_t *ira)
2116 {
2117 uint_t ulp_len;
2118 uint32_t cksum;
2119 uint8_t protocol = ira->ira_protocol;
2120 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
2121
2122 #define iphs ((uint16_t *)ipha)
2123
2124 switch (protocol) {
2125 case IPPROTO_TCP:
2126 ulp_len = ira->ira_pktlen - ip_hdr_length;
2127
2128 /* Protocol and length */
2129 cksum = htons(ulp_len) + IP_TCP_CSUM_COMP;
2130 /* IP addresses */
2131 cksum += iphs[6] + iphs[7] + iphs[8] + iphs[9];
2132 break;
2133
2134 case IPPROTO_UDP: {
2135 udpha_t *udpha;
2136
2137 udpha = (udpha_t *)((uchar_t *)ipha + ip_hdr_length);
2138
2139 /* Protocol and length */
2140 cksum = udpha->uha_length + IP_UDP_CSUM_COMP;
2141 /* IP addresses */
2142 cksum += iphs[6] + iphs[7] + iphs[8] + iphs[9];
2143 break;
2144 }
2145
2146 default:
2147 cksum = 0;
2148 break;
2149 }
2150 #undef iphs
2151 return (cksum);
2152 }
2153
2154
2155 /*
2156 * Software verification of the ULP checksums.
2157 * Returns B_TRUE if ok.
2158 * Increments statistics of failed.
2159 */
2160 static boolean_t
2161 ip_input_sw_cksum_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
2162 {
2163 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2164 uint32_t cksum;
2165 uint8_t protocol = ira->ira_protocol;
2166 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
2167
2168 IP_STAT(ipst, ip_in_sw_cksum);
2169
2170 ASSERT(protocol == IPPROTO_TCP || protocol == IPPROTO_UDP);
2171
2172 cksum = ip_input_cksum_pseudo_v4(ipha, ira);
2173 cksum = IP_CSUM(mp, ip_hdr_length, cksum);
2174 if (cksum == 0)
2175 return (B_TRUE);
2176
2177 ip_input_cksum_err_v4(protocol, 0, ira->ira_ill);
2178 return (B_FALSE);
2179 }
2180
2181 /*
2182 * Verify the ULP checksums.
2183 * Returns B_TRUE if ok, or if the ULP doesn't have a well-defined checksum
2184 * algorithm.
2185 * Increments statistics if failed.
2186 */
2187 static boolean_t
2188 ip_input_cksum_v4(iaflags_t iraflags, mblk_t *mp, ipha_t *ipha,
2189 ip_recv_attr_t *ira)
2190 {
2191 ill_t *ill = ira->ira_rill;
2192 uint16_t hck_flags;
2193 uint32_t cksum;
2194 mblk_t *mp1;
2195 int32_t len;
2196 uint8_t protocol = ira->ira_protocol;
2197 uint16_t ip_hdr_length = ira->ira_ip_hdr_length;
2198
2199
2200 switch (protocol) {
2201 case IPPROTO_TCP:
2202 break;
2203
2204 case IPPROTO_UDP: {
2205 udpha_t *udpha;
2206
2207 udpha = (udpha_t *)((uchar_t *)ipha + ip_hdr_length);
2208 if (udpha->uha_checksum == 0) {
2209 /* Packet doesn't have a UDP checksum */
2210 return (B_TRUE);
2211 }
2212 break;
2213 }
2214 case IPPROTO_SCTP: {
2215 sctp_hdr_t *sctph;
2216 uint32_t pktsum;
2217
2218 sctph = (sctp_hdr_t *)((uchar_t *)ipha + ip_hdr_length);
2219 #ifdef DEBUG
2220 if (skip_sctp_cksum)
2221 return (B_TRUE);
2222 #endif
2223 pktsum = sctph->sh_chksum;
2224 sctph->sh_chksum = 0;
2225 cksum = sctp_cksum(mp, ip_hdr_length);
2226 sctph->sh_chksum = pktsum;
2227 if (cksum == pktsum)
2228 return (B_TRUE);
2229
2230 /*
2231 * Defer until later whether a bad checksum is ok
2232 * in order to allow RAW sockets to use Adler checksum
2233 * with SCTP.
2234 */
2235 ira->ira_flags |= IRAF_SCTP_CSUM_ERR;
2236 return (B_TRUE);
2237 }
2238
2239 default:
2240 /* No ULP checksum to verify. */
2241 return (B_TRUE);
2242 }
2243 /*
2244 * Revert to software checksum calculation if the interface
2245 * isn't capable of checksum offload.
2246 * We clear DB_CKSUMFLAGS when going through IPsec in ip_fanout.
2247 * Note: IRAF_NO_HW_CKSUM is not currently used.
2248 */
2249 ASSERT(!IS_IPMP(ill));
2250 if ((iraflags & IRAF_NO_HW_CKSUM) || !ILL_HCKSUM_CAPABLE(ill) ||
2251 !dohwcksum) {
2252 return (ip_input_sw_cksum_v4(mp, ipha, ira));
2253 }
2254
2255 /*
2256 * We apply this for all ULP protocols. Does the HW know to
2257 * not set the flags for SCTP and other protocols.
2258 */
2259
2260 hck_flags = DB_CKSUMFLAGS(mp);
2261
2262 if (hck_flags & HCK_FULLCKSUM_OK) {
2263 /*
2264 * Hardware has already verified the checksum.
2265 */
2266 return (B_TRUE);
2267 }
2268
2269 if (hck_flags & HCK_FULLCKSUM) {
2270 /*
2271 * Full checksum has been computed by the hardware
2272 * and has been attached. If the driver wants us to
2273 * verify the correctness of the attached value, in
2274 * order to protect against faulty hardware, compare
2275 * it against -0 (0xFFFF) to see if it's valid.
2276 */
2277 cksum = DB_CKSUM16(mp);
2278 if (cksum == 0xFFFF)
2279 return (B_TRUE);
2280 ip_input_cksum_err_v4(protocol, hck_flags, ira->ira_ill);
2281 return (B_FALSE);
2282 }
2283
2284 mp1 = mp->b_cont;
2285 if ((hck_flags & HCK_PARTIALCKSUM) &&
2286 (mp1 == NULL || mp1->b_cont == NULL) &&
2287 ip_hdr_length >= DB_CKSUMSTART(mp) &&
2288 ((len = ip_hdr_length - DB_CKSUMSTART(mp)) & 1) == 0) {
2289 uint32_t adj;
2290 uchar_t *cksum_start;
2291
2292 cksum = ip_input_cksum_pseudo_v4(ipha, ira);
2293
2294 cksum_start = ((uchar_t *)ipha + DB_CKSUMSTART(mp));
2295
2296 /*
2297 * Partial checksum has been calculated by hardware
2298 * and attached to the packet; in addition, any
2299 * prepended extraneous data is even byte aligned,
2300 * and there are at most two mblks associated with
2301 * the packet. If any such data exists, we adjust
2302 * the checksum; also take care any postpended data.
2303 */
2304 IP_ADJCKSUM_PARTIAL(cksum_start, mp, mp1, len, adj);
2305 /*
2306 * One's complement subtract extraneous checksum
2307 */
2308 cksum += DB_CKSUM16(mp);
2309 if (adj >= cksum)
2310 cksum = ~(adj - cksum) & 0xFFFF;
2311 else
2312 cksum -= adj;
2313 cksum = (cksum & 0xFFFF) + ((int)cksum >> 16);
2314 cksum = (cksum & 0xFFFF) + ((int)cksum >> 16);
2315 if (!(~cksum & 0xFFFF))
2316 return (B_TRUE);
2317
2318 ip_input_cksum_err_v4(protocol, hck_flags, ira->ira_ill);
2319 return (B_FALSE);
2320 }
2321 return (ip_input_sw_cksum_v4(mp, ipha, ira));
2322 }
2323
2324
2325 /*
2326 * Handle fanout of received packets.
2327 * Unicast packets that are looped back (from ire_send_local_v4) and packets
2328 * from the wire are differentiated by checking IRAF_VERIFY_ULP_CKSUM.
2329 *
2330 * IPQoS Notes
2331 * Before sending it to the client, invoke IPPF processing. Policy processing
2332 * takes place only if the callout_position, IPP_LOCAL_IN, is enabled.
2333 */
2334 void
2335 ip_fanout_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
2336 {
2337 ill_t *ill = ira->ira_ill;
2338 iaflags_t iraflags = ira->ira_flags;
2339 ip_stack_t *ipst = ill->ill_ipst;
2340 uint8_t protocol = ipha->ipha_protocol;
2341 conn_t *connp;
2342 #define rptr ((uchar_t *)ipha)
2343 uint_t ip_hdr_length;
2344 uint_t min_ulp_header_length;
2345 int offset;
2346 ssize_t len;
2347 netstack_t *ns = ipst->ips_netstack;
2348 ipsec_stack_t *ipss = ns->netstack_ipsec;
2349 ill_t *rill = ira->ira_rill;
2350
2351 ASSERT(ira->ira_pktlen == ntohs(ipha->ipha_length));
2352
2353 ip_hdr_length = ira->ira_ip_hdr_length;
2354 ira->ira_protocol = protocol;
2355
2356 /*
2357 * Time for IPP once we've done reassembly and IPsec.
2358 * We skip this for loopback packets since we don't do IPQoS
2359 * on loopback.
2360 */
2361 if (IPP_ENABLED(IPP_LOCAL_IN, ipst) &&
2362 !(iraflags & IRAF_LOOPBACK) &&
2363 (protocol != IPPROTO_ESP && protocol != IPPROTO_AH)) {
2364 /*
2365 * Use the interface on which the packet arrived - not where
2366 * the IP address is hosted.
2367 */
2368 /* ip_process translates an IS_UNDER_IPMP */
2369 mp = ip_process(IPP_LOCAL_IN, mp, rill, ill);
2370 if (mp == NULL) {
2371 /* ip_drop_packet and MIB done */
2372 return;
2373 }
2374 }
2375
2376 /* Determine the minimum required size of the upper-layer header */
2377 /* Need to do this for at least the set of ULPs that TX handles. */
2378 switch (protocol) {
2379 case IPPROTO_TCP:
2380 min_ulp_header_length = TCP_MIN_HEADER_LENGTH;
2381 break;
2382 case IPPROTO_SCTP:
2383 min_ulp_header_length = SCTP_COMMON_HDR_LENGTH;
2384 break;
2385 case IPPROTO_UDP:
2386 min_ulp_header_length = UDPH_SIZE;
2387 break;
2388 case IPPROTO_ICMP:
2389 min_ulp_header_length = ICMPH_SIZE;
2390 break;
2391 default:
2392 min_ulp_header_length = 0;
2393 break;
2394 }
2395 /* Make sure we have the min ULP header length */
2396 len = mp->b_wptr - rptr;
2397 if (len < ip_hdr_length + min_ulp_header_length) {
2398 if (ira->ira_pktlen < ip_hdr_length + min_ulp_header_length) {
2399 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2400 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2401 freemsg(mp);
2402 return;
2403 }
2404 IP_STAT(ipst, ip_recv_pullup);
2405 ipha = ip_pullup(mp, ip_hdr_length + min_ulp_header_length,
2406 ira);
2407 if (ipha == NULL)
2408 goto discard;
2409 len = mp->b_wptr - rptr;
2410 }
2411
2412 /*
2413 * If trusted extensions then determine the zoneid and TX specific
2414 * ira_flags.
2415 */
2416 if (iraflags & IRAF_SYSTEM_LABELED) {
2417 /* This can update ira->ira_flags and ira->ira_zoneid */
2418 ip_fanout_tx_v4(mp, ipha, protocol, ip_hdr_length, ira);
2419 iraflags = ira->ira_flags;
2420 }
2421
2422
2423 /* Verify ULP checksum. Handles TCP, UDP, and SCTP */
2424 if (iraflags & IRAF_VERIFY_ULP_CKSUM) {
2425 if (!ip_input_cksum_v4(iraflags, mp, ipha, ira)) {
2426 /* Bad checksum. Stats are already incremented */
2427 ip_drop_input("Bad ULP checksum", mp, ill);
2428 freemsg(mp);
2429 return;
2430 }
2431 /* IRAF_SCTP_CSUM_ERR could have been set */
2432 iraflags = ira->ira_flags;
2433 }
2434 switch (protocol) {
2435 case IPPROTO_TCP:
2436 /* For TCP, discard broadcast and multicast packets. */
2437 if (iraflags & IRAF_MULTIBROADCAST)
2438 goto discard;
2439
2440 /* First mblk contains IP+TCP headers per above check */
2441 ASSERT(len >= ip_hdr_length + TCP_MIN_HEADER_LENGTH);
2442
2443 /* TCP options present? */
2444 offset = ((uchar_t *)ipha)[ip_hdr_length + 12] >> 4;
2445 if (offset != 5) {
2446 if (offset < 5)
2447 goto discard;
2448
2449 /*
2450 * There must be TCP options.
2451 * Make sure we can grab them.
2452 */
2453 offset <<= 2;
2454 offset += ip_hdr_length;
2455 if (len < offset) {
2456 if (ira->ira_pktlen < offset) {
2457 BUMP_MIB(ill->ill_ip_mib,
2458 ipIfStatsInTruncatedPkts);
2459 ip_drop_input(
2460 "ipIfStatsInTruncatedPkts",
2461 mp, ill);
2462 freemsg(mp);
2463 return;
2464 }
2465 IP_STAT(ipst, ip_recv_pullup);
2466 ipha = ip_pullup(mp, offset, ira);
2467 if (ipha == NULL)
2468 goto discard;
2469 len = mp->b_wptr - rptr;
2470 }
2471 }
2472
2473 /*
2474 * Pass up a squeue hint to tcp.
2475 * If ira_sqp is already set (this is loopback) we leave it
2476 * alone.
2477 */
2478 if (ira->ira_sqp == NULL) {
2479 ira->ira_sqp = ip_squeue_get(ira->ira_ring);
2480 }
2481
2482 /* Look for AF_INET or AF_INET6 that matches */
2483 connp = ipcl_classify_v4(mp, IPPROTO_TCP, ip_hdr_length,
2484 ira, ipst);
2485 if (connp == NULL) {
2486 /* Send the TH_RST */
2487 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2488 tcp_xmit_listeners_reset(mp, ira, ipst, NULL);
2489 return;
2490 }
2491 if (connp->conn_incoming_ifindex != 0 &&
2492 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2493 CONN_DEC_REF(connp);
2494
2495 /* Send the TH_RST */
2496 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2497 tcp_xmit_listeners_reset(mp, ira, ipst, NULL);
2498 return;
2499 }
2500 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2501 (iraflags & IRAF_IPSEC_SECURE)) {
2502 mp = ipsec_check_inbound_policy(mp, connp,
2503 ipha, NULL, ira);
2504 if (mp == NULL) {
2505 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2506 /* Note that mp is NULL */
2507 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2508 CONN_DEC_REF(connp);
2509 return;
2510 }
2511 }
2512 /* Found a client; up it goes */
2513 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2514 ira->ira_ill = ira->ira_rill = NULL;
2515 if (!IPCL_IS_TCP(connp)) {
2516 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2517 (connp->conn_recv)(connp, mp, NULL, ira);
2518 CONN_DEC_REF(connp);
2519 ira->ira_ill = ill;
2520 ira->ira_rill = rill;
2521 return;
2522 }
2523
2524 /*
2525 * We do different processing whether called from
2526 * ip_accept_tcp and we match the target, don't match
2527 * the target, and when we are called by ip_input.
2528 */
2529 if (iraflags & IRAF_TARGET_SQP) {
2530 if (ira->ira_target_sqp == connp->conn_sqp) {
2531 mblk_t *attrmp;
2532
2533 attrmp = ip_recv_attr_to_mblk(ira);
2534 if (attrmp == NULL) {
2535 BUMP_MIB(ill->ill_ip_mib,
2536 ipIfStatsInDiscards);
2537 ip_drop_input("ipIfStatsInDiscards",
2538 mp, ill);
2539 freemsg(mp);
2540 CONN_DEC_REF(connp);
2541 } else {
2542 SET_SQUEUE(attrmp, connp->conn_recv,
2543 connp);
2544 attrmp->b_cont = mp;
2545 ASSERT(ira->ira_target_sqp_mp == NULL);
2546 ira->ira_target_sqp_mp = attrmp;
2547 /*
2548 * Conn ref release when drained from
2549 * the squeue.
2550 */
2551 }
2552 } else {
2553 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2554 connp->conn_recv, connp, ira, SQ_FILL,
2555 SQTAG_IP_TCP_INPUT);
2556 }
2557 } else {
2558 SQUEUE_ENTER_ONE(connp->conn_sqp, mp, connp->conn_recv,
2559 connp, ira, ip_squeue_flag, SQTAG_IP_TCP_INPUT);
2560 }
2561 ira->ira_ill = ill;
2562 ira->ira_rill = rill;
2563 return;
2564
2565 case IPPROTO_SCTP: {
2566 sctp_hdr_t *sctph;
2567 in6_addr_t map_src, map_dst;
2568 uint32_t ports; /* Source and destination ports */
2569 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
2570
2571 /* For SCTP, discard broadcast and multicast packets. */
2572 if (iraflags & IRAF_MULTIBROADCAST)
2573 goto discard;
2574
2575 /*
2576 * Since there is no SCTP h/w cksum support yet, just
2577 * clear the flag.
2578 */
2579 DB_CKSUMFLAGS(mp) = 0;
2580
2581 /* Length ensured above */
2582 ASSERT(MBLKL(mp) >= ip_hdr_length + SCTP_COMMON_HDR_LENGTH);
2583 sctph = (sctp_hdr_t *)(rptr + ip_hdr_length);
2584
2585 /* get the ports */
2586 ports = *(uint32_t *)&sctph->sh_sport;
2587
2588 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &map_dst);
2589 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &map_src);
2590 if (iraflags & IRAF_SCTP_CSUM_ERR) {
2591 /*
2592 * No potential sctp checksum errors go to the Sun
2593 * sctp stack however they might be Adler-32 summed
2594 * packets a userland stack bound to a raw IP socket
2595 * could reasonably use. Note though that Adler-32 is
2596 * a long deprecated algorithm and customer sctp
2597 * networks should eventually migrate to CRC-32 at
2598 * which time this facility should be removed.
2599 */
2600 ip_fanout_sctp_raw(mp, ipha, NULL, ports, ira);
2601 return;
2602 }
2603 connp = sctp_fanout(&map_src, &map_dst, ports, ira, mp,
2604 sctps, sctph);
2605 if (connp == NULL) {
2606 /* Check for raw socket or OOTB handling */
2607 ip_fanout_sctp_raw(mp, ipha, NULL, ports, ira);
2608 return;
2609 }
2610 if (connp->conn_incoming_ifindex != 0 &&
2611 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2612 CONN_DEC_REF(connp);
2613 /* Check for raw socket or OOTB handling */
2614 ip_fanout_sctp_raw(mp, ipha, NULL, ports, ira);
2615 return;
2616 }
2617
2618 /* Found a client; up it goes */
2619 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2620 sctp_input(connp, ipha, NULL, mp, ira);
2621 /* sctp_input does a rele of the sctp_t */
2622 return;
2623 }
2624
2625 case IPPROTO_UDP:
2626 /* First mblk contains IP+UDP headers as checked above */
2627 ASSERT(MBLKL(mp) >= ip_hdr_length + UDPH_SIZE);
2628
2629 if (iraflags & IRAF_MULTIBROADCAST) {
2630 uint16_t *up; /* Pointer to ports in ULP header */
2631
2632 up = (uint16_t *)((uchar_t *)ipha + ip_hdr_length);
2633 ip_fanout_udp_multi_v4(mp, ipha, up[1], up[0], ira);
2634 return;
2635 }
2636
2637 /* Look for AF_INET or AF_INET6 that matches */
2638 connp = ipcl_classify_v4(mp, IPPROTO_UDP, ip_hdr_length,
2639 ira, ipst);
2640 if (connp == NULL) {
2641 no_udp_match:
2642 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].
2643 connf_head != NULL) {
2644 ASSERT(ira->ira_protocol == IPPROTO_UDP);
2645 ip_fanout_proto_v4(mp, ipha, ira);
2646 } else {
2647 ip_fanout_send_icmp_v4(mp,
2648 ICMP_DEST_UNREACHABLE,
2649 ICMP_PORT_UNREACHABLE, ira);
2650 }
2651 return;
2652
2653 }
2654 if (connp->conn_incoming_ifindex != 0 &&
2655 connp->conn_incoming_ifindex != ira->ira_ruifindex) {
2656 CONN_DEC_REF(connp);
2657 goto no_udp_match;
2658 }
2659 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
2660 !canputnext(connp->conn_rq)) {
2661 CONN_DEC_REF(connp);
2662 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
2663 ip_drop_input("udpIfStatsInOverflows", mp, ill);
2664 freemsg(mp);
2665 return;
2666 }
2667 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2668 (iraflags & IRAF_IPSEC_SECURE)) {
2669 mp = ipsec_check_inbound_policy(mp, connp,
2670 ipha, NULL, ira);
2671 if (mp == NULL) {
2672 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2673 /* Note that mp is NULL */
2674 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2675 CONN_DEC_REF(connp);
2676 return;
2677 }
2678 }
2679 /*
2680 * Remove 0-spi if it's 0, or move everything behind
2681 * the UDP header over it and forward to ESP via
2682 * ip_fanout_v4().
2683 */
2684 if (connp->conn_udp->udp_nat_t_endpoint) {
2685 if (iraflags & IRAF_IPSEC_SECURE) {
2686 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
2687 DROPPER(ipss, ipds_esp_nat_t_ipsec),
2688 &ipss->ipsec_dropper);
2689 CONN_DEC_REF(connp);
2690 return;
2691 }
2692
2693 mp = zero_spi_check(mp, ira);
2694 if (mp == NULL) {
2695 /*
2696 * Packet was consumed - probably sent to
2697 * ip_fanout_v4.
2698 */
2699 CONN_DEC_REF(connp);
2700 return;
2701 }
2702 /* Else continue like a normal UDP packet. */
2703 ipha = (ipha_t *)mp->b_rptr;
2704 protocol = ipha->ipha_protocol;
2705 ira->ira_protocol = protocol;
2706 }
2707 /* Found a client; up it goes */
2708 IP_STAT(ipst, ip_udp_fannorm);
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;
2716 default:
2717 break;
2718 }
2719
2720 /*
2721 * Clear hardware checksumming flag as it is currently only
2722 * used by TCP and UDP.
2723 */
2724 DB_CKSUMFLAGS(mp) = 0;
2725
2726 switch (protocol) {
2727 case IPPROTO_ICMP:
2728 /*
2729 * We need to accomodate icmp messages coming in clear
2730 * until we get everything secure from the wire. If
2731 * icmp_accept_clear_messages is zero we check with
2732 * the global policy and act accordingly. If it is
2733 * non-zero, we accept the message without any checks.
2734 * But *this does not mean* that this will be delivered
2735 * to RAW socket clients. By accepting we might send
2736 * replies back, change our MTU value etc.,
2737 * but delivery to the ULP/clients depends on their
2738 * policy dispositions.
2739 */
2740 if (ipst->ips_icmp_accept_clear_messages == 0) {
2741 mp = ipsec_check_global_policy(mp, NULL,
2742 ipha, NULL, ira, ns);
2743 if (mp == NULL)
2744 return;
2745 }
2746
2747 /*
2748 * On a labeled system, we have to check whether the zone
2749 * itself is permitted to receive raw traffic.
2750 */
2751 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
2752 if (!tsol_can_accept_raw(mp, ira, B_FALSE)) {
2753 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
2754 ip_drop_input("tsol_can_accept_raw", mp, ill);
2755 freemsg(mp);
2756 return;
2757 }
2758 }
2759
2760 /*
2761 * ICMP header checksum, including checksum field,
2762 * should be zero.
2763 */
2764 if (IP_CSUM(mp, ip_hdr_length, 0)) {
2765 BUMP_MIB(&ipst->ips_icmp_mib, icmpInCksumErrs);
2766 ip_drop_input("icmpInCksumErrs", mp, ill);
2767 freemsg(mp);
2768 return;
2769 }
2770 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2771 mp = icmp_inbound_v4(mp, ira);
2772 if (mp == NULL) {
2773 /* No need to pass to RAW sockets */
2774 return;
2775 }
2776 break;
2777
2778 case IPPROTO_IGMP:
2779 /*
2780 * If we are not willing to accept IGMP packets in clear,
2781 * then check with global policy.
2782 */
2783 if (ipst->ips_igmp_accept_clear_messages == 0) {
2784 mp = ipsec_check_global_policy(mp, NULL,
2785 ipha, NULL, ira, ns);
2786 if (mp == NULL)
2787 return;
2788 }
2789 if ((ira->ira_flags & IRAF_SYSTEM_LABELED) &&
2790 !tsol_can_accept_raw(mp, ira, B_TRUE)) {
2791 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2792 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2793 freemsg(mp);
2794 return;
2795 }
2796 /*
2797 * Validate checksum
2798 */
2799 if (IP_CSUM(mp, ip_hdr_length, 0)) {
2800 ++ipst->ips_igmpstat.igps_rcv_badsum;
2801 ip_drop_input("igps_rcv_badsum", mp, ill);
2802 freemsg(mp);
2803 return;
2804 }
2805
2806 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2807 mp = igmp_input(mp, ira);
2808 if (mp == NULL) {
2809 /* Bad packet - discarded by igmp_input */
2810 return;
2811 }
2812 break;
2813 case IPPROTO_PIM:
2814 /*
2815 * If we are not willing to accept PIM packets in clear,
2816 * then check with global policy.
2817 */
2818 if (ipst->ips_pim_accept_clear_messages == 0) {
2819 mp = ipsec_check_global_policy(mp, NULL,
2820 ipha, NULL, ira, ns);
2821 if (mp == NULL)
2822 return;
2823 }
2824 if ((ira->ira_flags & IRAF_SYSTEM_LABELED) &&
2825 !tsol_can_accept_raw(mp, ira, B_TRUE)) {
2826 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2827 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2828 freemsg(mp);
2829 return;
2830 }
2831 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2832
2833 /* Checksum is verified in pim_input */
2834 mp = pim_input(mp, ira);
2835 if (mp == NULL) {
2836 /* Bad packet - discarded by pim_input */
2837 return;
2838 }
2839 break;
2840 case IPPROTO_AH:
2841 case IPPROTO_ESP: {
2842 /*
2843 * Fast path for AH/ESP.
2844 */
2845 netstack_t *ns = ipst->ips_netstack;
2846 ipsec_stack_t *ipss = ns->netstack_ipsec;
2847
2848 IP_STAT(ipst, ipsec_proto_ahesp);
2849
2850 if (!ipsec_loaded(ipss)) {
2851 ip_proto_not_sup(mp, ira);
2852 return;
2853 }
2854
2855 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
2856 /* select inbound SA and have IPsec process the pkt */
2857 if (protocol == IPPROTO_ESP) {
2858 esph_t *esph;
2859 boolean_t esp_in_udp_sa;
2860 boolean_t esp_in_udp_packet;
2861
2862 mp = ipsec_inbound_esp_sa(mp, ira, &esph);
2863 if (mp == NULL)
2864 return;
2865
2866 ASSERT(esph != NULL);
2867 ASSERT(ira->ira_flags & IRAF_IPSEC_SECURE);
2868 ASSERT(ira->ira_ipsec_esp_sa != NULL);
2869 ASSERT(ira->ira_ipsec_esp_sa->ipsa_input_func != NULL);
2870
2871 esp_in_udp_sa = ((ira->ira_ipsec_esp_sa->ipsa_flags &
2872 IPSA_F_NATT) != 0);
2873 esp_in_udp_packet =
2874 (ira->ira_flags & IRAF_ESP_UDP_PORTS) != 0;
2875
2876 /*
2877 * The following is a fancy, but quick, way of saying:
2878 * ESP-in-UDP SA and Raw ESP packet --> drop
2879 * OR
2880 * ESP SA and ESP-in-UDP packet --> drop
2881 */
2882 if (esp_in_udp_sa != esp_in_udp_packet) {
2883 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2884 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
2885 DROPPER(ipss, ipds_esp_no_sa),
2886 &ipss->ipsec_dropper);
2887 return;
2888 }
2889 mp = ira->ira_ipsec_esp_sa->ipsa_input_func(mp, esph,
2890 ira);
2891 } else {
2892 ah_t *ah;
2893
2894 mp = ipsec_inbound_ah_sa(mp, ira, &ah);
2895 if (mp == NULL)
2896 return;
2897
2898 ASSERT(ah != NULL);
2899 ASSERT(ira->ira_flags & IRAF_IPSEC_SECURE);
2900 ASSERT(ira->ira_ipsec_ah_sa != NULL);
2901 ASSERT(ira->ira_ipsec_ah_sa->ipsa_input_func != NULL);
2902 mp = ira->ira_ipsec_ah_sa->ipsa_input_func(mp, ah,
2903 ira);
2904 }
2905
2906 if (mp == NULL) {
2907 /*
2908 * Either it failed or is pending. In the former case
2909 * ipIfStatsInDiscards was increased.
2910 */
2911 return;
2912 }
2913 /* we're done with IPsec processing, send it up */
2914 ip_input_post_ipsec(mp, ira);
2915 return;
2916 }
2917 case IPPROTO_ENCAP: {
2918 ipha_t *inner_ipha;
2919
2920 /*
2921 * Handle self-encapsulated packets (IP-in-IP where
2922 * the inner addresses == the outer addresses).
2923 */
2924 if ((uchar_t *)ipha + ip_hdr_length + sizeof (ipha_t) >
2925 mp->b_wptr) {
2926 if (ira->ira_pktlen <
2927 ip_hdr_length + sizeof (ipha_t)) {
2928 BUMP_MIB(ill->ill_ip_mib,
2929 ipIfStatsInTruncatedPkts);
2930 ip_drop_input("ipIfStatsInTruncatedPkts",
2931 mp, ill);
2932 freemsg(mp);
2933 return;
2934 }
2935 ipha = ip_pullup(mp, (uchar_t *)ipha + ip_hdr_length +
2936 sizeof (ipha_t) - mp->b_rptr, ira);
2937 if (ipha == NULL) {
2938 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2939 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2940 freemsg(mp);
2941 return;
2942 }
2943 }
2944 inner_ipha = (ipha_t *)((uchar_t *)ipha + ip_hdr_length);
2945 /*
2946 * Check the sanity of the inner IP header.
2947 */
2948 if ((IPH_HDR_VERSION(inner_ipha) != IPV4_VERSION)) {
2949 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2950 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2951 freemsg(mp);
2952 return;
2953 }
2954 if (IPH_HDR_LENGTH(inner_ipha) < sizeof (ipha_t)) {
2955 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2956 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2957 freemsg(mp);
2958 return;
2959 }
2960 if (inner_ipha->ipha_src != ipha->ipha_src ||
2961 inner_ipha->ipha_dst != ipha->ipha_dst) {
2962 /* We fallthru to iptun fanout below */
2963 goto iptun;
2964 }
2965
2966 /*
2967 * Self-encapsulated tunnel packet. Remove
2968 * the outer IP header and fanout again.
2969 * We also need to make sure that the inner
2970 * header is pulled up until options.
2971 */
2972 mp->b_rptr = (uchar_t *)inner_ipha;
2973 ipha = inner_ipha;
2974 ip_hdr_length = IPH_HDR_LENGTH(ipha);
2975 if ((uchar_t *)ipha + ip_hdr_length > mp->b_wptr) {
2976 if (ira->ira_pktlen <
2977 (uchar_t *)ipha + ip_hdr_length - mp->b_rptr) {
2978 BUMP_MIB(ill->ill_ip_mib,
2979 ipIfStatsInTruncatedPkts);
2980 ip_drop_input("ipIfStatsInTruncatedPkts",
2981 mp, ill);
2982 freemsg(mp);
2983 return;
2984 }
2985 ipha = ip_pullup(mp,
2986 (uchar_t *)ipha + ip_hdr_length - mp->b_rptr, ira);
2987 if (ipha == NULL) {
2988 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2989 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2990 freemsg(mp);
2991 return;
2992 }
2993 }
2994 if (ip_hdr_length > sizeof (ipha_t)) {
2995 /* We got options on the inner packet. */
2996 ipaddr_t dst = ipha->ipha_dst;
2997 int error = 0;
2998
2999 dst = ip_input_options(ipha, dst, mp, ira, &error);
3000 if (error != 0) {
3001 /*
3002 * An ICMP error has been sent and the packet
3003 * has been dropped.
3004 */
3005 return;
3006 }
3007 if (dst != ipha->ipha_dst) {
3008 /*
3009 * Someone put a source-route in
3010 * the inside header of a self-
3011 * encapsulated packet. Drop it
3012 * with extreme prejudice and let
3013 * the sender know.
3014 */
3015 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
3016 mp, ill);
3017 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
3018 ira);
3019 return;
3020 }
3021 }
3022 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
3023 /*
3024 * This means that somebody is sending
3025 * Self-encapsualted packets without AH/ESP.
3026 *
3027 * Send this packet to find a tunnel endpoint.
3028 * if I can't find one, an ICMP
3029 * PROTOCOL_UNREACHABLE will get sent.
3030 */
3031 protocol = ipha->ipha_protocol;
3032 ira->ira_protocol = protocol;
3033 goto iptun;
3034 }
3035
3036 /* Update based on removed IP header */
3037 ira->ira_ip_hdr_length = ip_hdr_length;
3038 ira->ira_pktlen = ntohs(ipha->ipha_length);
3039
3040 if (ira->ira_flags & IRAF_IPSEC_DECAPS) {
3041 /*
3042 * This packet is self-encapsulated multiple
3043 * times. We don't want to recurse infinitely.
3044 * To keep it simple, drop the packet.
3045 */
3046 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
3047 ip_drop_input("ipIfStatsInDiscards", mp, ill);
3048 freemsg(mp);
3049 return;
3050 }
3051 ASSERT(ira->ira_flags & IRAF_IPSEC_SECURE);
3052 ira->ira_flags |= IRAF_IPSEC_DECAPS;
3053
3054 ip_input_post_ipsec(mp, ira);
3055 return;
3056 }
3057
3058 iptun: /* IPPROTO_ENCAPS that is not self-encapsulated */
3059 case IPPROTO_IPV6:
3060 /* iptun will verify trusted label */
3061 connp = ipcl_classify_v4(mp, protocol, ip_hdr_length,
3062 ira, ipst);
3063 if (connp != NULL) {
3064 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
3065 ira->ira_ill = ira->ira_rill = NULL;
3066 (connp->conn_recv)(connp, mp, NULL, ira);
3067 CONN_DEC_REF(connp);
3068 ira->ira_ill = ill;
3069 ira->ira_rill = rill;
3070 return;
3071 }
3072 /* FALLTHRU */
3073 default:
3074 /*
3075 * On a labeled system, we have to check whether the zone
3076 * itself is permitted to receive raw traffic.
3077 */
3078 if (ira->ira_flags & IRAF_SYSTEM_LABELED) {
3079 if (!tsol_can_accept_raw(mp, ira, B_FALSE)) {
3080 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
3081 ip_drop_input("ipIfStatsInDiscards", mp, ill);
3082 freemsg(mp);
3083 return;
3084 }
3085 }
3086 break;
3087 }
3088
3089 /*
3090 * The above input functions may have returned the pulled up message.
3091 * So ipha need to be reinitialized.
3092 */
3093 ipha = (ipha_t *)mp->b_rptr;
3094 ira->ira_protocol = protocol = ipha->ipha_protocol;
3095 if (ipst->ips_ipcl_proto_fanout_v4[protocol].connf_head == NULL) {
3096 /*
3097 * No user-level listener for these packets packets.
3098 * Check for IPPROTO_ENCAP...
3099 */
3100 if (protocol == IPPROTO_ENCAP && ipst->ips_ip_g_mrouter) {
3101 /*
3102 * Check policy here,
3103 * THEN ship off to ip_mroute_decap().
3104 *
3105 * BTW, If I match a configured IP-in-IP
3106 * tunnel above, this path will not be reached, and
3107 * ip_mroute_decap will never be called.
3108 */
3109 mp = ipsec_check_global_policy(mp, connp,
3110 ipha, NULL, ira, ns);
3111 if (mp != NULL) {
3112 ip_mroute_decap(mp, ira);
3113 } /* Else we already freed everything! */
3114 } else {
3115 ip_proto_not_sup(mp, ira);
3116 }
3117 return;
3118 }
3119
3120 /*
3121 * Handle fanout to raw sockets. There
3122 * can be more than one stream bound to a particular
3123 * protocol. When this is the case, each one gets a copy
3124 * of any incoming packets.
3125 */
3126 ASSERT(ira->ira_protocol == ipha->ipha_protocol);
3127 ip_fanout_proto_v4(mp, ipha, ira);
3128 return;
3129
3130 discard:
3131 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
3132 ip_drop_input("ipIfStatsInDiscards", mp, ill);
3133 freemsg(mp);
3134 #undef rptr
3135 }