2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_bootp.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
50 #include <sys/kernel.h>
52 #include <sys/rwlock.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/netisr.h>
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_icmp.h>
75 #include <netinet/ip_options.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/ip_carp.h>
79 #include <netinet/ip_ipsec.h>
81 #include <netinet/in_rss.h>
83 #include <sys/socketvar.h>
85 #include <security/mac/mac_framework.h>
88 CTASSERT(sizeof(struct ip) == 20);
91 struct rwlock in_ifaddr_lock;
92 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
94 VNET_DEFINE(int, rsvp_on);
96 VNET_DEFINE(int, ipforwarding);
97 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
98 &VNET_NAME(ipforwarding), 0,
99 "Enable IP forwarding between interfaces");
101 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
102 #define V_ipsendredirects VNET(ipsendredirects)
103 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
104 &VNET_NAME(ipsendredirects), 0,
105 "Enable sending IP redirects");
107 static VNET_DEFINE(int, ip_keepfaith);
108 #define V_ip_keepfaith VNET(ip_keepfaith)
109 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
110 &VNET_NAME(ip_keepfaith), 0,
111 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
113 static VNET_DEFINE(int, ip_sendsourcequench);
114 #define V_ip_sendsourcequench VNET(ip_sendsourcequench)
115 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
116 &VNET_NAME(ip_sendsourcequench), 0,
117 "Enable the transmission of source quench packets");
119 VNET_DEFINE(int, ip_do_randomid);
120 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
121 &VNET_NAME(ip_do_randomid), 0,
122 "Assign random ip_id values");
125 * XXX - Setting ip_checkinterface mostly implements the receive side of
126 * the Strong ES model described in RFC 1122, but since the routing table
127 * and transmit implementation do not implement the Strong ES model,
128 * setting this to 1 results in an odd hybrid.
130 * XXX - ip_checkinterface currently must be disabled if you use ipnat
131 * to translate the destination address to another local interface.
133 * XXX - ip_checkinterface must be disabled if you add IP aliases
134 * to the loopback interface instead of the interface where the
135 * packets for those addresses are received.
137 static VNET_DEFINE(int, ip_checkinterface);
138 #define V_ip_checkinterface VNET(ip_checkinterface)
139 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
140 &VNET_NAME(ip_checkinterface), 0,
141 "Verify packet arrives on correct interface");
143 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
145 static struct netisr_handler ip_nh = {
147 .nh_handler = ip_input,
148 .nh_proto = NETISR_IP,
150 .nh_m2cpuid = rss_soft_m2cpuid,
151 .nh_policy = NETISR_POLICY_CPU,
152 .nh_dispatch = NETISR_DISPATCH_HYBRID,
154 .nh_policy = NETISR_POLICY_FLOW,
160 * Directly dispatched frames are currently assumed
161 * to have a flowid already calculated.
163 * It should likely have something that assert it
164 * actually has valid flow details.
166 static struct netisr_handler ip_direct_nh = {
167 .nh_name = "ip_direct",
168 .nh_handler = ip_direct_input,
169 .nh_proto = NETISR_IP_DIRECT,
170 .nh_m2cpuid = rss_m2cpuid,
171 .nh_policy = NETISR_POLICY_CPU,
172 .nh_dispatch = NETISR_DISPATCH_HYBRID,
176 extern struct domain inetdomain;
177 extern struct protosw inetsw[];
178 u_char ip_protox[IPPROTO_MAX];
179 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
180 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
181 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
183 static VNET_DEFINE(uma_zone_t, ipq_zone);
184 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
185 static struct mtx ipqlock;
187 #define V_ipq_zone VNET(ipq_zone)
188 #define V_ipq VNET(ipq)
190 #define IPQ_LOCK() mtx_lock(&ipqlock)
191 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
192 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
193 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
195 static void maxnipq_update(void);
196 static void ipq_zone_change(void *);
197 static void ip_drain_locked(void);
199 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
200 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
201 #define V_maxnipq VNET(maxnipq)
202 #define V_nipq VNET(nipq)
203 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
205 "Current number of IPv4 fragment reassembly queue entries");
207 static VNET_DEFINE(int, maxfragsperpacket);
208 #define V_maxfragsperpacket VNET(maxfragsperpacket)
209 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
210 &VNET_NAME(maxfragsperpacket), 0,
211 "Maximum number of IPv4 fragments allowed per packet");
214 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
215 &ip_mtu, 0, "Default MTU");
219 VNET_DEFINE(int, ipstealth);
220 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
221 &VNET_NAME(ipstealth), 0,
222 "IP stealth mode, no TTL decrementation on forwarding");
225 static void ip_freef(struct ipqhead *, struct ipq *);
228 * IP statistics are stored in the "array" of counter(9)s.
230 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
231 VNET_PCPUSTAT_SYSINIT(ipstat);
232 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
233 "IP statistics (struct ipstat, netinet/ip_var.h)");
236 VNET_PCPUSTAT_SYSUNINIT(ipstat);
240 * Kernel module interface for updating ipstat. The argument is an index
241 * into ipstat treated as an array.
244 kmod_ipstat_inc(int statnum)
247 counter_u64_add(VNET(ipstat)[statnum], 1);
251 kmod_ipstat_dec(int statnum)
254 counter_u64_add(VNET(ipstat)[statnum], -1);
258 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
262 netisr_getqlimit(&ip_nh, &qlimit);
263 error = sysctl_handle_int(oidp, &qlimit, 0, req);
264 if (error || !req->newptr)
268 return (netisr_setqlimit(&ip_nh, qlimit));
270 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
271 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
272 "Maximum size of the IP input queue");
275 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
277 u_int64_t qdrops_long;
280 netisr_getqdrops(&ip_nh, &qdrops_long);
281 qdrops = qdrops_long;
282 error = sysctl_handle_int(oidp, &qdrops, 0, req);
283 if (error || !req->newptr)
287 netisr_clearqdrops(&ip_nh);
291 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
292 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
293 "Number of packets dropped from the IP input queue");
297 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
301 netisr_getqlimit(&ip_direct_nh, &qlimit);
302 error = sysctl_handle_int(oidp, &qlimit, 0, req);
303 if (error || !req->newptr)
307 return (netisr_setqlimit(&ip_direct_nh, qlimit));
309 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
310 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
311 "Maximum size of the IP direct input queue");
314 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
316 u_int64_t qdrops_long;
319 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
320 qdrops = qdrops_long;
321 error = sysctl_handle_int(oidp, &qdrops, 0, req);
322 if (error || !req->newptr)
326 netisr_clearqdrops(&ip_direct_nh);
330 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
331 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
332 "Number of packets dropped from the IP direct input queue");
336 * IP initialization: fill in IP protocol switch table.
337 * All protocols not implemented in kernel go to raw IP protocol handler.
345 V_ip_id = time_second & 0xffff;
347 TAILQ_INIT(&V_in_ifaddrhead);
348 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
350 /* Initialize IP reassembly queue. */
351 for (i = 0; i < IPREASS_NHASH; i++)
352 TAILQ_INIT(&V_ipq[i]);
353 V_maxnipq = nmbclusters / 32;
354 V_maxfragsperpacket = 16;
355 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
356 NULL, UMA_ALIGN_PTR, 0);
359 /* Initialize packet filter hooks. */
360 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
361 V_inet_pfil_hook.ph_af = AF_INET;
362 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
363 printf("%s: WARNING: unable to register pfil hook, "
364 "error %d\n", __func__, i);
366 /* Skip initialization of globals for non-default instances. */
367 if (!IS_DEFAULT_VNET(curvnet))
370 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
372 panic("ip_init: PF_INET not found");
374 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
375 for (i = 0; i < IPPROTO_MAX; i++)
376 ip_protox[i] = pr - inetsw;
378 * Cycle through IP protocols and put them into the appropriate place
381 for (pr = inetdomain.dom_protosw;
382 pr < inetdomain.dom_protoswNPROTOSW; pr++)
383 if (pr->pr_domain->dom_family == PF_INET &&
384 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
385 /* Be careful to only index valid IP protocols. */
386 if (pr->pr_protocol < IPPROTO_MAX)
387 ip_protox[pr->pr_protocol] = pr - inetsw;
390 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
391 NULL, EVENTHANDLER_PRI_ANY);
393 /* Initialize various other remaining things. */
395 netisr_register(&ip_nh);
397 netisr_register(&ip_direct_nh);
407 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
408 printf("%s: WARNING: unable to unregister pfil hook, "
409 "error %d\n", __func__, i);
411 /* Cleanup in_ifaddr hash table; should be empty. */
412 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
418 uma_zdestroy(V_ipq_zone);
424 * IP direct input routine.
426 * This is called when reinjecting completed fragments where
427 * all of the previous checking and book-keeping has been done.
430 ip_direct_input(struct mbuf *m)
435 ip = mtod(m, struct ip *);
436 hlen = ip->ip_hl << 2;
438 IPSTAT_INC(ips_delivered);
439 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
445 * Ip input routine. Checksum and byte swap header. If fragmented
446 * try to reassemble. Process options. Pass to next level.
449 ip_input(struct mbuf *m)
451 struct ip *ip = NULL;
452 struct in_ifaddr *ia = NULL;
455 int checkif, hlen = 0;
456 uint16_t sum, ip_len;
457 int dchg = 0; /* dest changed after fw */
458 struct in_addr odst; /* original dst address */
462 if (m->m_flags & M_FASTFWD_OURS) {
463 m->m_flags &= ~M_FASTFWD_OURS;
464 /* Set up some basics that will be used later. */
465 ip = mtod(m, struct ip *);
466 hlen = ip->ip_hl << 2;
467 ip_len = ntohs(ip->ip_len);
471 IPSTAT_INC(ips_total);
473 if (m->m_pkthdr.len < sizeof(struct ip))
476 if (m->m_len < sizeof (struct ip) &&
477 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
478 IPSTAT_INC(ips_toosmall);
481 ip = mtod(m, struct ip *);
483 if (ip->ip_v != IPVERSION) {
484 IPSTAT_INC(ips_badvers);
488 hlen = ip->ip_hl << 2;
489 if (hlen < sizeof(struct ip)) { /* minimum header length */
490 IPSTAT_INC(ips_badhlen);
493 if (hlen > m->m_len) {
494 if ((m = m_pullup(m, hlen)) == NULL) {
495 IPSTAT_INC(ips_badhlen);
498 ip = mtod(m, struct ip *);
501 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
503 /* 127/8 must not appear on wire - RFC1122 */
504 ifp = m->m_pkthdr.rcvif;
505 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
506 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
507 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
508 IPSTAT_INC(ips_badaddr);
513 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
514 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
516 if (hlen == sizeof(struct ip)) {
517 sum = in_cksum_hdr(ip);
519 sum = in_cksum(m, hlen);
523 IPSTAT_INC(ips_badsum);
528 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
529 /* packet is dropped by traffic conditioner */
533 ip_len = ntohs(ip->ip_len);
535 IPSTAT_INC(ips_badlen);
540 * Check that the amount of data in the buffers
541 * is as at least much as the IP header would have us expect.
542 * Trim mbufs if longer than we expect.
543 * Drop packet if shorter than we expect.
545 if (m->m_pkthdr.len < ip_len) {
547 IPSTAT_INC(ips_tooshort);
550 if (m->m_pkthdr.len > ip_len) {
551 if (m->m_len == m->m_pkthdr.len) {
553 m->m_pkthdr.len = ip_len;
555 m_adj(m, ip_len - m->m_pkthdr.len);
560 * Bypass packet filtering for packets previously handled by IPsec.
562 if (ip_ipsec_filtertunnel(m))
567 * Run through list of hooks for input packets.
569 * NB: Beware of the destination address changing (e.g.
570 * by NAT rewriting). When this happens, tell
571 * ip_forward to do the right thing.
574 /* Jump over all PFIL processing if hooks are not active. */
575 if (!PFIL_HOOKED(&V_inet_pfil_hook))
579 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
581 if (m == NULL) /* consumed by filter */
584 ip = mtod(m, struct ip *);
585 dchg = (odst.s_addr != ip->ip_dst.s_addr);
586 ifp = m->m_pkthdr.rcvif;
588 if (m->m_flags & M_FASTFWD_OURS) {
589 m->m_flags &= ~M_FASTFWD_OURS;
592 if (m->m_flags & M_IP_NEXTHOP) {
593 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
596 * Directly ship the packet on. This allows
597 * forwarding packets originally destined to us
598 * to some other directly connected host.
607 * Process options and, if not destined for us,
608 * ship it on. ip_dooptions returns 1 when an
609 * error was detected (causing an icmp message
610 * to be sent and the original packet to be freed).
612 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
615 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
616 * matter if it is destined to another node, or whether it is
617 * a multicast one, RSVP wants it! and prevents it from being forwarded
618 * anywhere else. Also checks if the rsvp daemon is running before
619 * grabbing the packet.
621 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
625 * Check our list of addresses, to see if the packet is for us.
626 * If we don't have any addresses, assume any unicast packet
627 * we receive might be for us (and let the upper layers deal
630 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
631 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
635 * Enable a consistency check between the destination address
636 * and the arrival interface for a unicast packet (the RFC 1122
637 * strong ES model) if IP forwarding is disabled and the packet
638 * is not locally generated and the packet is not subject to
641 * XXX - Checking also should be disabled if the destination
642 * address is ipnat'ed to a different interface.
644 * XXX - Checking is incompatible with IP aliases added
645 * to the loopback interface instead of the interface where
646 * the packets are received.
648 * XXX - This is the case for carp vhost IPs as well so we
649 * insert a workaround. If the packet got here, we already
650 * checked with carp_iamatch() and carp_forus().
652 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
653 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
654 ifp->if_carp == NULL && (dchg == 0);
657 * Check for exact addresses in the hash bucket.
659 /* IN_IFADDR_RLOCK(); */
660 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
662 * If the address matches, verify that the packet
663 * arrived via the correct interface if checking is
666 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
667 (!checkif || ia->ia_ifp == ifp)) {
668 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
669 counter_u64_add(ia->ia_ifa.ifa_ibytes,
671 /* IN_IFADDR_RUNLOCK(); */
675 /* IN_IFADDR_RUNLOCK(); */
678 * Check for broadcast addresses.
680 * Only accept broadcast packets that arrive via the matching
681 * interface. Reception of forwarded directed broadcasts would
682 * be handled via ip_forward() and ether_output() with the loopback
683 * into the stack for SIMPLEX interfaces handled by ether_output().
685 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
687 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
688 if (ifa->ifa_addr->sa_family != AF_INET)
691 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
693 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
694 counter_u64_add(ia->ia_ifa.ifa_ibytes,
696 IF_ADDR_RUNLOCK(ifp);
700 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
701 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
702 counter_u64_add(ia->ia_ifa.ifa_ibytes,
704 IF_ADDR_RUNLOCK(ifp);
709 IF_ADDR_RUNLOCK(ifp);
712 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
713 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
714 IPSTAT_INC(ips_cantforward);
718 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
721 * If we are acting as a multicast router, all
722 * incoming multicast packets are passed to the
723 * kernel-level multicast forwarding function.
724 * The packet is returned (relatively) intact; if
725 * ip_mforward() returns a non-zero value, the packet
726 * must be discarded, else it may be accepted below.
728 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
729 IPSTAT_INC(ips_cantforward);
735 * The process-level routing daemon needs to receive
736 * all multicast IGMP packets, whether or not this
737 * host belongs to their destination groups.
739 if (ip->ip_p == IPPROTO_IGMP)
741 IPSTAT_INC(ips_forward);
744 * Assume the packet is for us, to avoid prematurely taking
745 * a lock on the in_multi hash. Protocols must perform
746 * their own filtering and update statistics accordingly.
750 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
752 if (ip->ip_dst.s_addr == INADDR_ANY)
756 * FAITH(Firewall Aided Internet Translator)
758 if (ifp && ifp->if_type == IFT_FAITH) {
759 if (V_ip_keepfaith) {
760 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
768 * Not for us; forward if possible and desirable.
770 if (V_ipforwarding == 0) {
771 IPSTAT_INC(ips_cantforward);
785 * IPSTEALTH: Process non-routing options only
786 * if the packet is destined for us.
788 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
790 #endif /* IPSTEALTH */
793 * Attempt reassembly; if it succeeds, proceed.
794 * ip_reass() will return a different mbuf.
796 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
797 /* XXXGL: shouldn't we save & set m_flags? */
801 ip = mtod(m, struct ip *);
802 /* Get the header length of the reassembled packet */
803 hlen = ip->ip_hl << 2;
808 * enforce IPsec policy checking if we are seeing last header.
809 * note that we do not visit this with protocols with pcb layer
810 * code - like udp/tcp/raw ip.
812 if (ip_ipsec_input(m))
817 * Switch out to protocol's input routine.
819 IPSTAT_INC(ips_delivered);
821 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
828 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
829 * max has slightly different semantics than the sysctl, for historical
837 * -1 for unlimited allocation.
840 uma_zone_set_max(V_ipq_zone, 0);
842 * Positive number for specific bound.
845 uma_zone_set_max(V_ipq_zone, V_maxnipq);
847 * Zero specifies no further fragment queue allocation -- set the
848 * bound very low, but rely on implementation elsewhere to actually
849 * prevent allocation and reclaim current queues.
852 uma_zone_set_max(V_ipq_zone, 1);
856 ipq_zone_change(void *tag)
859 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
860 V_maxnipq = nmbclusters / 32;
866 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
871 error = sysctl_handle_int(oidp, &i, 0, req);
872 if (error || !req->newptr)
876 * XXXRW: Might be a good idea to sanity check the argument and place
877 * an extreme upper bound.
886 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
887 NULL, 0, sysctl_maxnipq, "I",
888 "Maximum number of IPv4 fragment reassembly queue entries");
890 #define M_IP_FRAG M_PROTO9
893 * Take incoming datagram fragment and try to reassemble it into
894 * whole datagram. If the argument is the first fragment or one
895 * in between the function will return NULL and store the mbuf
896 * in the fragment chain. If the argument is the last fragment
897 * the packet will be reassembled and the pointer to the new
898 * mbuf returned for further processing. Only m_tags attached
899 * to the first packet/fragment are preserved.
900 * The IP header is *NOT* adjusted out of iplen.
903 ip_reass(struct mbuf *m)
906 struct mbuf *p, *q, *nq, *t;
907 struct ipq *fp = NULL;
908 struct ipqhead *head;
913 uint32_t rss_hash, rss_type;
916 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
917 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
918 IPSTAT_INC(ips_fragments);
919 IPSTAT_INC(ips_fragdropped);
924 ip = mtod(m, struct ip *);
925 hlen = ip->ip_hl << 2;
927 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
932 * Look for queue of fragments
935 TAILQ_FOREACH(fp, head, ipq_list)
936 if (ip->ip_id == fp->ipq_id &&
937 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
938 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
940 mac_ipq_match(m, fp) &&
942 ip->ip_p == fp->ipq_p)
948 * Attempt to trim the number of allocated fragment queues if it
949 * exceeds the administrative limit.
951 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
953 * drop something from the tail of the current queue
954 * before proceeding further
956 struct ipq *q = TAILQ_LAST(head, ipqhead);
957 if (q == NULL) { /* gak */
958 for (i = 0; i < IPREASS_NHASH; i++) {
959 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
961 IPSTAT_ADD(ips_fragtimeout,
963 ip_freef(&V_ipq[i], r);
968 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
975 * Adjust ip_len to not reflect header,
976 * convert offset of this to bytes.
978 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
979 if (ip->ip_off & htons(IP_MF)) {
981 * Make sure that fragments have a data length
982 * that's a non-zero multiple of 8 bytes.
984 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
985 IPSTAT_INC(ips_toosmall); /* XXX */
988 m->m_flags |= M_IP_FRAG;
990 m->m_flags &= ~M_IP_FRAG;
991 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
994 * Attempt reassembly; if it succeeds, proceed.
995 * ip_reass() will return a different mbuf.
997 IPSTAT_INC(ips_fragments);
998 m->m_pkthdr.PH_loc.ptr = ip;
1000 /* Previous ip_reass() started here. */
1002 * Presence of header sizes in mbufs
1003 * would confuse code below.
1009 * If first fragment to arrive, create a reassembly queue.
1012 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
1016 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
1017 uma_zfree(V_ipq_zone, fp);
1021 mac_ipq_create(m, fp);
1023 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1026 fp->ipq_ttl = IPFRAGTTL;
1027 fp->ipq_p = ip->ip_p;
1028 fp->ipq_id = ip->ip_id;
1029 fp->ipq_src = ip->ip_src;
1030 fp->ipq_dst = ip->ip_dst;
1032 m->m_nextpkt = NULL;
1037 mac_ipq_update(m, fp);
1041 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
1044 * Handle ECN by comparing this segment with the first one;
1045 * if CE is set, do not lose CE.
1046 * drop if CE and not-ECT are mixed for the same packet.
1048 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1049 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1050 if (ecn == IPTOS_ECN_CE) {
1051 if (ecn0 == IPTOS_ECN_NOTECT)
1053 if (ecn0 != IPTOS_ECN_CE)
1054 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1056 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1060 * Find a segment which begins after this one does.
1062 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1063 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
1067 * If there is a preceding segment, it may provide some of
1068 * our data already. If so, drop the data from the incoming
1069 * segment. If it provides all of our data, drop us, otherwise
1070 * stick new segment in the proper place.
1072 * If some of the data is dropped from the preceding
1073 * segment, then it's checksum is invalidated.
1076 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
1079 if (i >= ntohs(ip->ip_len))
1082 m->m_pkthdr.csum_flags = 0;
1083 ip->ip_off = htons(ntohs(ip->ip_off) + i);
1084 ip->ip_len = htons(ntohs(ip->ip_len) - i);
1086 m->m_nextpkt = p->m_nextpkt;
1089 m->m_nextpkt = fp->ipq_frags;
1094 * While we overlap succeeding segments trim them or,
1095 * if they are completely covered, dequeue them.
1097 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1098 ntohs(GETIP(q)->ip_off); q = nq) {
1099 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1100 ntohs(GETIP(q)->ip_off);
1101 if (i < ntohs(GETIP(q)->ip_len)) {
1102 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1103 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1105 q->m_pkthdr.csum_flags = 0;
1110 IPSTAT_INC(ips_fragdropped);
1116 * Check for complete reassembly and perform frag per packet
1119 * Frag limiting is performed here so that the nth frag has
1120 * a chance to complete the packet before we drop the packet.
1121 * As a result, n+1 frags are actually allowed per packet, but
1122 * only n will ever be stored. (n = maxfragsperpacket.)
1126 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1127 if (ntohs(GETIP(q)->ip_off) != next) {
1128 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1129 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1134 next += ntohs(GETIP(q)->ip_len);
1136 /* Make sure the last packet didn't have the IP_MF flag */
1137 if (p->m_flags & M_IP_FRAG) {
1138 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1139 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1146 * Reassembly is complete. Make sure the packet is a sane size.
1150 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1151 IPSTAT_INC(ips_toolong);
1152 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1158 * Concatenate fragments.
1165 q->m_nextpkt = NULL;
1166 for (q = nq; q != NULL; q = nq) {
1168 q->m_nextpkt = NULL;
1169 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1170 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1174 * In order to do checksumming faster we do 'end-around carry' here
1175 * (and not in for{} loop), though it implies we are not going to
1176 * reassemble more than 64k fragments.
1178 while (m->m_pkthdr.csum_data & 0xffff0000)
1179 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1180 (m->m_pkthdr.csum_data >> 16);
1182 mac_ipq_reassemble(fp, m);
1183 mac_ipq_destroy(fp);
1187 * Create header for new ip packet by modifying header of first
1188 * packet; dequeue and discard fragment reassembly header.
1189 * Make header visible.
1191 ip->ip_len = htons((ip->ip_hl << 2) + next);
1192 ip->ip_src = fp->ipq_src;
1193 ip->ip_dst = fp->ipq_dst;
1194 TAILQ_REMOVE(head, fp, ipq_list);
1196 uma_zfree(V_ipq_zone, fp);
1197 m->m_len += (ip->ip_hl << 2);
1198 m->m_data -= (ip->ip_hl << 2);
1199 /* some debugging cruft by sklower, below, will go away soon */
1200 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1202 IPSTAT_INC(ips_reassembled);
1207 * Query the RSS layer for the flowid / flowtype for the
1210 * For now, just assume we have to calculate a new one.
1211 * Later on we should check to see if the assigned flowid matches
1212 * what RSS wants for the given IP protocol and if so, just keep it.
1214 * We then queue into the relevant netisr so it can be dispatched
1215 * to the correct CPU.
1217 * Note - this may return 1, which means the flowid in the mbuf
1218 * is correct for the configured RSS hash types and can be used.
1220 if (rss_mbuf_software_hash_v4(m, 0, &rss_hash, &rss_type) == 0) {
1221 m->m_pkthdr.flowid = rss_hash;
1222 M_HASHTYPE_SET(m, rss_type);
1223 m->m_flags |= M_FLOWID;
1227 * Queue/dispatch for reprocessing.
1229 * Note: this is much slower than just handling the frame in the
1230 * current receive context. It's likely worth investigating
1233 netisr_dispatch(NETISR_IP_DIRECT, m);
1237 /* Handle in-line */
1241 IPSTAT_INC(ips_fragdropped);
1253 * Free a fragment reassembly header and all
1254 * associated datagrams.
1257 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1263 while (fp->ipq_frags) {
1265 fp->ipq_frags = q->m_nextpkt;
1268 TAILQ_REMOVE(fhp, fp, ipq_list);
1269 uma_zfree(V_ipq_zone, fp);
1274 * IP timer processing;
1275 * if a timer expires on a reassembly
1276 * queue, discard it.
1281 VNET_ITERATOR_DECL(vnet_iter);
1285 VNET_LIST_RLOCK_NOSLEEP();
1287 VNET_FOREACH(vnet_iter) {
1288 CURVNET_SET(vnet_iter);
1289 for (i = 0; i < IPREASS_NHASH; i++) {
1290 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1294 fp = TAILQ_NEXT(fp, ipq_list);
1295 if(--fpp->ipq_ttl == 0) {
1296 IPSTAT_ADD(ips_fragtimeout,
1298 ip_freef(&V_ipq[i], fpp);
1303 * If we are over the maximum number of fragments
1304 * (due to the limit being lowered), drain off
1305 * enough to get down to the new limit.
1307 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1308 for (i = 0; i < IPREASS_NHASH; i++) {
1309 while (V_nipq > V_maxnipq &&
1310 !TAILQ_EMPTY(&V_ipq[i])) {
1311 IPSTAT_ADD(ips_fragdropped,
1312 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1314 TAILQ_FIRST(&V_ipq[i]));
1321 VNET_LIST_RUNLOCK_NOSLEEP();
1325 * Drain off all datagram fragments.
1328 ip_drain_locked(void)
1334 for (i = 0; i < IPREASS_NHASH; i++) {
1335 while(!TAILQ_EMPTY(&V_ipq[i])) {
1336 IPSTAT_ADD(ips_fragdropped,
1337 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1338 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1346 VNET_ITERATOR_DECL(vnet_iter);
1348 VNET_LIST_RLOCK_NOSLEEP();
1350 VNET_FOREACH(vnet_iter) {
1351 CURVNET_SET(vnet_iter);
1356 VNET_LIST_RUNLOCK_NOSLEEP();
1361 * The protocol to be inserted into ip_protox[] must be already registered
1362 * in inetsw[], either statically or through pf_proto_register().
1365 ipproto_register(short ipproto)
1369 /* Sanity checks. */
1370 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1371 return (EPROTONOSUPPORT);
1374 * The protocol slot must not be occupied by another protocol
1375 * already. An index pointing to IPPROTO_RAW is unused.
1377 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1379 return (EPFNOSUPPORT);
1380 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1383 /* Find the protocol position in inetsw[] and set the index. */
1384 for (pr = inetdomain.dom_protosw;
1385 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1386 if (pr->pr_domain->dom_family == PF_INET &&
1387 pr->pr_protocol && pr->pr_protocol == ipproto) {
1388 ip_protox[pr->pr_protocol] = pr - inetsw;
1392 return (EPROTONOSUPPORT);
1396 ipproto_unregister(short ipproto)
1400 /* Sanity checks. */
1401 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1402 return (EPROTONOSUPPORT);
1404 /* Check if the protocol was indeed registered. */
1405 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1407 return (EPFNOSUPPORT);
1408 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1411 /* Reset the protocol slot to IPPROTO_RAW. */
1412 ip_protox[ipproto] = pr - inetsw;
1417 * Given address of next destination (final or next hop), return (referenced)
1418 * internet address info of interface to be used to get there.
1421 ip_rtaddr(struct in_addr dst, u_int fibnum)
1424 struct sockaddr_in *sin;
1425 struct in_ifaddr *ia;
1427 bzero(&sro, sizeof(sro));
1428 sin = (struct sockaddr_in *)&sro.ro_dst;
1429 sin->sin_family = AF_INET;
1430 sin->sin_len = sizeof(*sin);
1431 sin->sin_addr = dst;
1432 in_rtalloc_ign(&sro, 0, fibnum);
1434 if (sro.ro_rt == NULL)
1437 ia = ifatoia(sro.ro_rt->rt_ifa);
1438 ifa_ref(&ia->ia_ifa);
1443 u_char inetctlerrmap[PRC_NCMDS] = {
1445 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1446 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1447 EMSGSIZE, EHOSTUNREACH, 0, 0,
1448 0, 0, EHOSTUNREACH, 0,
1449 ENOPROTOOPT, ECONNREFUSED
1453 * Forward a packet. If some error occurs return the sender
1454 * an icmp packet. Note we can't always generate a meaningful
1455 * icmp message because icmp doesn't have a large enough repertoire
1456 * of codes and types.
1458 * If not forwarding, just drop the packet. This could be confusing
1459 * if ipforwarding was zero but some routing protocol was advancing
1460 * us as a gateway to somewhere. However, we must let the routing
1461 * protocol deal with that.
1463 * The srcrt parameter indicates whether the packet is being forwarded
1464 * via a source route.
1467 ip_forward(struct mbuf *m, int srcrt)
1469 struct ip *ip = mtod(m, struct ip *);
1470 struct in_ifaddr *ia;
1472 struct in_addr dest;
1474 int error, type = 0, code = 0, mtu = 0;
1476 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1477 IPSTAT_INC(ips_cantforward);
1484 if (ip->ip_ttl <= IPTTLDEC) {
1485 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1493 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1496 * 'ia' may be NULL if there is no route for this destination.
1497 * In case of IPsec, Don't discard it just yet, but pass it to
1498 * ip_output in case of outgoing IPsec policy.
1500 if (!srcrt && ia == NULL) {
1501 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1507 * Save the IP header and at most 8 bytes of the payload,
1508 * in case we need to generate an ICMP message to the src.
1510 * XXX this can be optimized a lot by saving the data in a local
1511 * buffer on the stack (72 bytes at most), and only allocating the
1512 * mbuf if really necessary. The vast majority of the packets
1513 * are forwarded without having to send an ICMP back (either
1514 * because unnecessary, or because rate limited), so we are
1515 * really we are wasting a lot of work here.
1517 * We don't use m_copy() because it might return a reference
1518 * to a shared cluster. Both this function and ip_output()
1519 * assume exclusive access to the IP header in `m', so any
1520 * data in a cluster may change before we reach icmp_error().
1522 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1523 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1525 * It's probably ok if the pkthdr dup fails (because
1526 * the deep copy of the tag chain failed), but for now
1527 * be conservative and just discard the copy since
1528 * code below may some day want the tags.
1533 if (mcopy != NULL) {
1534 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1535 mcopy->m_pkthdr.len = mcopy->m_len;
1536 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1542 ip->ip_ttl -= IPTTLDEC;
1548 * If forwarding packet using same interface that it came in on,
1549 * perhaps should send a redirect to sender to shortcut a hop.
1550 * Only send redirect if source is sending directly to us,
1551 * and if packet was not source routed (or has any options).
1552 * Also, don't send redirect if forwarding using a default route
1553 * or a route modified by a redirect.
1556 if (!srcrt && V_ipsendredirects &&
1557 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1558 struct sockaddr_in *sin;
1561 bzero(&ro, sizeof(ro));
1562 sin = (struct sockaddr_in *)&ro.ro_dst;
1563 sin->sin_family = AF_INET;
1564 sin->sin_len = sizeof(*sin);
1565 sin->sin_addr = ip->ip_dst;
1566 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1570 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1571 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1572 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1573 u_long src = ntohl(ip->ip_src.s_addr);
1576 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1577 if (rt->rt_flags & RTF_GATEWAY)
1578 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1580 dest.s_addr = ip->ip_dst.s_addr;
1581 /* Router requirements says to only send host redirects */
1582 type = ICMP_REDIRECT;
1583 code = ICMP_REDIRECT_HOST;
1591 * Try to cache the route MTU from ip_output so we can consider it for
1592 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1594 bzero(&ro, sizeof(ro));
1596 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1598 if (error == EMSGSIZE && ro.ro_rt)
1599 mtu = ro.ro_rt->rt_mtu;
1603 IPSTAT_INC(ips_cantforward);
1605 IPSTAT_INC(ips_forward);
1607 IPSTAT_INC(ips_redirectsent);
1612 ifa_free(&ia->ia_ifa);
1616 if (mcopy == NULL) {
1618 ifa_free(&ia->ia_ifa);
1624 case 0: /* forwarded, but need redirect */
1625 /* type, code set above */
1633 type = ICMP_UNREACH;
1634 code = ICMP_UNREACH_HOST;
1638 type = ICMP_UNREACH;
1639 code = ICMP_UNREACH_NEEDFRAG;
1643 * If IPsec is configured for this path,
1644 * override any possibly mtu value set by ip_output.
1646 mtu = ip_ipsec_mtu(mcopy, mtu);
1649 * If the MTU was set before make sure we are below the
1651 * If the MTU wasn't set before use the interface mtu or
1652 * fall back to the next smaller mtu step compared to the
1653 * current packet size.
1657 mtu = min(mtu, ia->ia_ifp->if_mtu);
1660 mtu = ia->ia_ifp->if_mtu;
1662 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1664 IPSTAT_INC(ips_cantfrag);
1669 * A router should not generate ICMP_SOURCEQUENCH as
1670 * required in RFC1812 Requirements for IP Version 4 Routers.
1671 * Source quench could be a big problem under DoS attacks,
1672 * or if the underlying interface is rate-limited.
1673 * Those who need source quench packets may re-enable them
1674 * via the net.inet.ip.sendsourcequench sysctl.
1676 if (V_ip_sendsourcequench == 0) {
1679 ifa_free(&ia->ia_ifa);
1682 type = ICMP_SOURCEQUENCH;
1687 case EACCES: /* ipfw denied packet */
1690 ifa_free(&ia->ia_ifa);
1694 ifa_free(&ia->ia_ifa);
1695 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1699 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1703 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1707 if (inp->inp_socket->so_options & SO_BINTIME) {
1708 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1709 SCM_BINTIME, SOL_SOCKET);
1711 mp = &(*mp)->m_next;
1713 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1716 bintime2timeval(&bt, &tv);
1717 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1718 SCM_TIMESTAMP, SOL_SOCKET);
1720 mp = &(*mp)->m_next;
1723 if (inp->inp_flags & INP_RECVDSTADDR) {
1724 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1725 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1727 mp = &(*mp)->m_next;
1729 if (inp->inp_flags & INP_RECVTTL) {
1730 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1731 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1733 mp = &(*mp)->m_next;
1737 * Moving these out of udp_input() made them even more broken
1738 * than they already were.
1740 /* options were tossed already */
1741 if (inp->inp_flags & INP_RECVOPTS) {
1742 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1743 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1745 mp = &(*mp)->m_next;
1747 /* ip_srcroute doesn't do what we want here, need to fix */
1748 if (inp->inp_flags & INP_RECVRETOPTS) {
1749 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1750 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1752 mp = &(*mp)->m_next;
1755 if (inp->inp_flags & INP_RECVIF) {
1758 struct sockaddr_dl sdl;
1761 struct sockaddr_dl *sdp;
1762 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1764 if ((ifp = m->m_pkthdr.rcvif) &&
1765 ifp->if_index && ifp->if_index <= V_if_index) {
1766 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1768 * Change our mind and don't try copy.
1770 if (sdp->sdl_family != AF_LINK ||
1771 sdp->sdl_len > sizeof(sdlbuf)) {
1774 bcopy(sdp, sdl2, sdp->sdl_len);
1778 offsetof(struct sockaddr_dl, sdl_data[0]);
1779 sdl2->sdl_family = AF_LINK;
1780 sdl2->sdl_index = 0;
1781 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1783 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1784 IP_RECVIF, IPPROTO_IP);
1786 mp = &(*mp)->m_next;
1788 if (inp->inp_flags & INP_RECVTOS) {
1789 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1790 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1792 mp = &(*mp)->m_next;
1795 if (inp->inp_flags2 & INP_RECVFLOWID) {
1796 uint32_t flowid, flow_type;
1798 flowid = m->m_pkthdr.flowid;
1799 flow_type = M_HASHTYPE_GET(m);
1802 * XXX should handle the failure of one or the
1803 * other - don't populate both?
1805 *mp = sbcreatecontrol((caddr_t) &flowid,
1806 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1808 mp = &(*mp)->m_next;
1809 *mp = sbcreatecontrol((caddr_t) &flow_type,
1810 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1812 mp = &(*mp)->m_next;
1816 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1817 uint32_t flowid, flow_type;
1818 uint32_t rss_bucketid;
1820 flowid = m->m_pkthdr.flowid;
1821 flow_type = M_HASHTYPE_GET(m);
1823 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1824 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1825 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1827 mp = &(*mp)->m_next;
1834 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1835 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1836 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1839 static VNET_DEFINE(int, ip_rsvp_on);
1840 VNET_DEFINE(struct socket *, ip_rsvpd);
1842 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1845 ip_rsvp_init(struct socket *so)
1848 if (so->so_type != SOCK_RAW ||
1849 so->so_proto->pr_protocol != IPPROTO_RSVP)
1852 if (V_ip_rsvpd != NULL)
1857 * This may seem silly, but we need to be sure we don't over-increment
1858 * the RSVP counter, in case something slips up.
1860 if (!V_ip_rsvp_on) {
1874 * This may seem silly, but we need to be sure we don't over-decrement
1875 * the RSVP counter, in case something slips up.
1885 rsvp_input(struct mbuf **mp, int *offp, int proto)
1892 if (rsvp_input_p) { /* call the real one if loaded */
1894 rsvp_input_p(mp, offp, proto);
1895 return (IPPROTO_DONE);
1898 /* Can still get packets with rsvp_on = 0 if there is a local member
1899 * of the group to which the RSVP packet is addressed. But in this
1900 * case we want to throw the packet away.
1905 return (IPPROTO_DONE);
1908 if (V_ip_rsvpd != NULL) {
1910 rip_input(mp, offp, proto);
1911 return (IPPROTO_DONE);
1913 /* Drop the packet */
1915 return (IPPROTO_DONE);