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_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | 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_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
104 &VNET_NAME(ipsendredirects), 0,
105 "Enable sending IP redirects");
107 VNET_DEFINE(int, ip_do_randomid);
108 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_VNET | CTLFLAG_RW,
109 &VNET_NAME(ip_do_randomid), 0,
110 "Assign random ip_id values");
113 * XXX - Setting ip_checkinterface mostly implements the receive side of
114 * the Strong ES model described in RFC 1122, but since the routing table
115 * and transmit implementation do not implement the Strong ES model,
116 * setting this to 1 results in an odd hybrid.
118 * XXX - ip_checkinterface currently must be disabled if you use ipnat
119 * to translate the destination address to another local interface.
121 * XXX - ip_checkinterface must be disabled if you add IP aliases
122 * to the loopback interface instead of the interface where the
123 * packets for those addresses are received.
125 static VNET_DEFINE(int, ip_checkinterface);
126 #define V_ip_checkinterface VNET(ip_checkinterface)
127 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
128 &VNET_NAME(ip_checkinterface), 0,
129 "Verify packet arrives on correct interface");
131 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
133 static struct netisr_handler ip_nh = {
135 .nh_handler = ip_input,
136 .nh_proto = NETISR_IP,
138 .nh_m2cpuid = rss_soft_m2cpuid,
139 .nh_policy = NETISR_POLICY_CPU,
140 .nh_dispatch = NETISR_DISPATCH_HYBRID,
142 .nh_policy = NETISR_POLICY_FLOW,
148 * Directly dispatched frames are currently assumed
149 * to have a flowid already calculated.
151 * It should likely have something that assert it
152 * actually has valid flow details.
154 static struct netisr_handler ip_direct_nh = {
155 .nh_name = "ip_direct",
156 .nh_handler = ip_direct_input,
157 .nh_proto = NETISR_IP_DIRECT,
158 .nh_m2cpuid = rss_m2cpuid,
159 .nh_policy = NETISR_POLICY_CPU,
160 .nh_dispatch = NETISR_DISPATCH_HYBRID,
164 extern struct domain inetdomain;
165 extern struct protosw inetsw[];
166 u_char ip_protox[IPPROTO_MAX];
167 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
168 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
169 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
171 static VNET_DEFINE(uma_zone_t, ipq_zone);
172 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
173 static struct mtx ipqlock;
175 #define V_ipq_zone VNET(ipq_zone)
176 #define V_ipq VNET(ipq)
178 #define IPQ_LOCK() mtx_lock(&ipqlock)
179 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
180 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
181 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
183 static void maxnipq_update(void);
184 static void ipq_zone_change(void *);
185 static void ip_drain_locked(void);
187 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
188 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
189 #define V_maxnipq VNET(maxnipq)
190 #define V_nipq VNET(nipq)
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_VNET | CTLFLAG_RD,
193 "Current number of IPv4 fragment reassembly queue entries");
195 static VNET_DEFINE(int, maxfragsperpacket);
196 #define V_maxfragsperpacket VNET(maxfragsperpacket)
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_VNET | CTLFLAG_RW,
198 &VNET_NAME(maxfragsperpacket), 0,
199 "Maximum number of IPv4 fragments allowed per packet");
202 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
203 &ip_mtu, 0, "Default MTU");
207 VNET_DEFINE(int, ipstealth);
208 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
209 &VNET_NAME(ipstealth), 0,
210 "IP stealth mode, no TTL decrementation on forwarding");
213 static void ip_freef(struct ipqhead *, struct ipq *);
216 * IP statistics are stored in the "array" of counter(9)s.
218 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
219 VNET_PCPUSTAT_SYSINIT(ipstat);
220 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
221 "IP statistics (struct ipstat, netinet/ip_var.h)");
224 VNET_PCPUSTAT_SYSUNINIT(ipstat);
228 * Kernel module interface for updating ipstat. The argument is an index
229 * into ipstat treated as an array.
232 kmod_ipstat_inc(int statnum)
235 counter_u64_add(VNET(ipstat)[statnum], 1);
239 kmod_ipstat_dec(int statnum)
242 counter_u64_add(VNET(ipstat)[statnum], -1);
246 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
250 netisr_getqlimit(&ip_nh, &qlimit);
251 error = sysctl_handle_int(oidp, &qlimit, 0, req);
252 if (error || !req->newptr)
256 return (netisr_setqlimit(&ip_nh, qlimit));
258 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
259 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
260 "Maximum size of the IP input queue");
263 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
265 u_int64_t qdrops_long;
268 netisr_getqdrops(&ip_nh, &qdrops_long);
269 qdrops = qdrops_long;
270 error = sysctl_handle_int(oidp, &qdrops, 0, req);
271 if (error || !req->newptr)
275 netisr_clearqdrops(&ip_nh);
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
280 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
281 "Number of packets dropped from the IP input queue");
285 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
289 netisr_getqlimit(&ip_direct_nh, &qlimit);
290 error = sysctl_handle_int(oidp, &qlimit, 0, req);
291 if (error || !req->newptr)
295 return (netisr_setqlimit(&ip_direct_nh, qlimit));
297 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
298 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
299 "Maximum size of the IP direct input queue");
302 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
304 u_int64_t qdrops_long;
307 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
308 qdrops = qdrops_long;
309 error = sysctl_handle_int(oidp, &qdrops, 0, req);
310 if (error || !req->newptr)
314 netisr_clearqdrops(&ip_direct_nh);
318 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
319 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
320 "Number of packets dropped from the IP direct input queue");
324 * IP initialization: fill in IP protocol switch table.
325 * All protocols not implemented in kernel go to raw IP protocol handler.
333 V_ip_id = time_second & 0xffff;
335 TAILQ_INIT(&V_in_ifaddrhead);
336 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
338 /* Initialize IP reassembly queue. */
339 for (i = 0; i < IPREASS_NHASH; i++)
340 TAILQ_INIT(&V_ipq[i]);
341 V_maxnipq = nmbclusters / 32;
342 V_maxfragsperpacket = 16;
343 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
344 NULL, UMA_ALIGN_PTR, 0);
347 /* Initialize packet filter hooks. */
348 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
349 V_inet_pfil_hook.ph_af = AF_INET;
350 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
351 printf("%s: WARNING: unable to register pfil hook, "
352 "error %d\n", __func__, i);
354 /* Skip initialization of globals for non-default instances. */
355 if (!IS_DEFAULT_VNET(curvnet))
358 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
360 panic("ip_init: PF_INET not found");
362 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
363 for (i = 0; i < IPPROTO_MAX; i++)
364 ip_protox[i] = pr - inetsw;
366 * Cycle through IP protocols and put them into the appropriate place
369 for (pr = inetdomain.dom_protosw;
370 pr < inetdomain.dom_protoswNPROTOSW; pr++)
371 if (pr->pr_domain->dom_family == PF_INET &&
372 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
373 /* Be careful to only index valid IP protocols. */
374 if (pr->pr_protocol < IPPROTO_MAX)
375 ip_protox[pr->pr_protocol] = pr - inetsw;
378 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
379 NULL, EVENTHANDLER_PRI_ANY);
381 /* Initialize various other remaining things. */
383 netisr_register(&ip_nh);
385 netisr_register(&ip_direct_nh);
395 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
396 printf("%s: WARNING: unable to unregister pfil hook, "
397 "error %d\n", __func__, i);
399 /* Cleanup in_ifaddr hash table; should be empty. */
400 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
406 uma_zdestroy(V_ipq_zone);
412 * IP direct input routine.
414 * This is called when reinjecting completed fragments where
415 * all of the previous checking and book-keeping has been done.
418 ip_direct_input(struct mbuf *m)
423 ip = mtod(m, struct ip *);
424 hlen = ip->ip_hl << 2;
426 IPSTAT_INC(ips_delivered);
427 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
433 * Ip input routine. Checksum and byte swap header. If fragmented
434 * try to reassemble. Process options. Pass to next level.
437 ip_input(struct mbuf *m)
439 struct ip *ip = NULL;
440 struct in_ifaddr *ia = NULL;
443 int checkif, hlen = 0;
444 uint16_t sum, ip_len;
445 int dchg = 0; /* dest changed after fw */
446 struct in_addr odst; /* original dst address */
450 if (m->m_flags & M_FASTFWD_OURS) {
451 m->m_flags &= ~M_FASTFWD_OURS;
452 /* Set up some basics that will be used later. */
453 ip = mtod(m, struct ip *);
454 hlen = ip->ip_hl << 2;
455 ip_len = ntohs(ip->ip_len);
459 IPSTAT_INC(ips_total);
461 if (m->m_pkthdr.len < sizeof(struct ip))
464 if (m->m_len < sizeof (struct ip) &&
465 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
466 IPSTAT_INC(ips_toosmall);
469 ip = mtod(m, struct ip *);
471 if (ip->ip_v != IPVERSION) {
472 IPSTAT_INC(ips_badvers);
476 hlen = ip->ip_hl << 2;
477 if (hlen < sizeof(struct ip)) { /* minimum header length */
478 IPSTAT_INC(ips_badhlen);
481 if (hlen > m->m_len) {
482 if ((m = m_pullup(m, hlen)) == NULL) {
483 IPSTAT_INC(ips_badhlen);
486 ip = mtod(m, struct ip *);
489 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
491 /* 127/8 must not appear on wire - RFC1122 */
492 ifp = m->m_pkthdr.rcvif;
493 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
494 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
495 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
496 IPSTAT_INC(ips_badaddr);
501 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
502 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
504 if (hlen == sizeof(struct ip)) {
505 sum = in_cksum_hdr(ip);
507 sum = in_cksum(m, hlen);
511 IPSTAT_INC(ips_badsum);
516 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
517 /* packet is dropped by traffic conditioner */
521 ip_len = ntohs(ip->ip_len);
523 IPSTAT_INC(ips_badlen);
528 * Check that the amount of data in the buffers
529 * is as at least much as the IP header would have us expect.
530 * Trim mbufs if longer than we expect.
531 * Drop packet if shorter than we expect.
533 if (m->m_pkthdr.len < ip_len) {
535 IPSTAT_INC(ips_tooshort);
538 if (m->m_pkthdr.len > ip_len) {
539 if (m->m_len == m->m_pkthdr.len) {
541 m->m_pkthdr.len = ip_len;
543 m_adj(m, ip_len - m->m_pkthdr.len);
548 * Bypass packet filtering for packets previously handled by IPsec.
550 if (ip_ipsec_filtertunnel(m))
555 * Run through list of hooks for input packets.
557 * NB: Beware of the destination address changing (e.g.
558 * by NAT rewriting). When this happens, tell
559 * ip_forward to do the right thing.
562 /* Jump over all PFIL processing if hooks are not active. */
563 if (!PFIL_HOOKED(&V_inet_pfil_hook))
567 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
569 if (m == NULL) /* consumed by filter */
572 ip = mtod(m, struct ip *);
573 dchg = (odst.s_addr != ip->ip_dst.s_addr);
574 ifp = m->m_pkthdr.rcvif;
576 if (m->m_flags & M_FASTFWD_OURS) {
577 m->m_flags &= ~M_FASTFWD_OURS;
580 if (m->m_flags & M_IP_NEXTHOP) {
581 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
584 * Directly ship the packet on. This allows
585 * forwarding packets originally destined to us
586 * to some other directly connected host.
595 * Process options and, if not destined for us,
596 * ship it on. ip_dooptions returns 1 when an
597 * error was detected (causing an icmp message
598 * to be sent and the original packet to be freed).
600 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
603 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
604 * matter if it is destined to another node, or whether it is
605 * a multicast one, RSVP wants it! and prevents it from being forwarded
606 * anywhere else. Also checks if the rsvp daemon is running before
607 * grabbing the packet.
609 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
613 * Check our list of addresses, to see if the packet is for us.
614 * If we don't have any addresses, assume any unicast packet
615 * we receive might be for us (and let the upper layers deal
618 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
619 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
623 * Enable a consistency check between the destination address
624 * and the arrival interface for a unicast packet (the RFC 1122
625 * strong ES model) if IP forwarding is disabled and the packet
626 * is not locally generated and the packet is not subject to
629 * XXX - Checking also should be disabled if the destination
630 * address is ipnat'ed to a different interface.
632 * XXX - Checking is incompatible with IP aliases added
633 * to the loopback interface instead of the interface where
634 * the packets are received.
636 * XXX - This is the case for carp vhost IPs as well so we
637 * insert a workaround. If the packet got here, we already
638 * checked with carp_iamatch() and carp_forus().
640 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
641 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
642 ifp->if_carp == NULL && (dchg == 0);
645 * Check for exact addresses in the hash bucket.
647 /* IN_IFADDR_RLOCK(); */
648 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
650 * If the address matches, verify that the packet
651 * arrived via the correct interface if checking is
654 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
655 (!checkif || ia->ia_ifp == ifp)) {
656 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
657 counter_u64_add(ia->ia_ifa.ifa_ibytes,
659 /* IN_IFADDR_RUNLOCK(); */
663 /* IN_IFADDR_RUNLOCK(); */
666 * Check for broadcast addresses.
668 * Only accept broadcast packets that arrive via the matching
669 * interface. Reception of forwarded directed broadcasts would
670 * be handled via ip_forward() and ether_output() with the loopback
671 * into the stack for SIMPLEX interfaces handled by ether_output().
673 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
675 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
676 if (ifa->ifa_addr->sa_family != AF_INET)
679 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
681 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
682 counter_u64_add(ia->ia_ifa.ifa_ibytes,
684 IF_ADDR_RUNLOCK(ifp);
688 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
689 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
690 counter_u64_add(ia->ia_ifa.ifa_ibytes,
692 IF_ADDR_RUNLOCK(ifp);
697 IF_ADDR_RUNLOCK(ifp);
700 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
701 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
702 IPSTAT_INC(ips_cantforward);
706 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
709 * If we are acting as a multicast router, all
710 * incoming multicast packets are passed to the
711 * kernel-level multicast forwarding function.
712 * The packet is returned (relatively) intact; if
713 * ip_mforward() returns a non-zero value, the packet
714 * must be discarded, else it may be accepted below.
716 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
717 IPSTAT_INC(ips_cantforward);
723 * The process-level routing daemon needs to receive
724 * all multicast IGMP packets, whether or not this
725 * host belongs to their destination groups.
727 if (ip->ip_p == IPPROTO_IGMP)
729 IPSTAT_INC(ips_forward);
732 * Assume the packet is for us, to avoid prematurely taking
733 * a lock on the in_multi hash. Protocols must perform
734 * their own filtering and update statistics accordingly.
738 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
740 if (ip->ip_dst.s_addr == INADDR_ANY)
744 * Not for us; forward if possible and desirable.
746 if (V_ipforwarding == 0) {
747 IPSTAT_INC(ips_cantforward);
757 * IPSTEALTH: Process non-routing options only
758 * if the packet is destined for us.
760 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
762 #endif /* IPSTEALTH */
765 * Attempt reassembly; if it succeeds, proceed.
766 * ip_reass() will return a different mbuf.
768 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
769 /* XXXGL: shouldn't we save & set m_flags? */
773 ip = mtod(m, struct ip *);
774 /* Get the header length of the reassembled packet */
775 hlen = ip->ip_hl << 2;
780 * enforce IPsec policy checking if we are seeing last header.
781 * note that we do not visit this with protocols with pcb layer
782 * code - like udp/tcp/raw ip.
784 if (ip_ipsec_input(m, ip->ip_p) != 0)
789 * Switch out to protocol's input routine.
791 IPSTAT_INC(ips_delivered);
793 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
800 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
801 * max has slightly different semantics than the sysctl, for historical
809 * -1 for unlimited allocation.
812 uma_zone_set_max(V_ipq_zone, 0);
814 * Positive number for specific bound.
817 uma_zone_set_max(V_ipq_zone, V_maxnipq);
819 * Zero specifies no further fragment queue allocation -- set the
820 * bound very low, but rely on implementation elsewhere to actually
821 * prevent allocation and reclaim current queues.
824 uma_zone_set_max(V_ipq_zone, 1);
828 ipq_zone_change(void *tag)
831 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
832 V_maxnipq = nmbclusters / 32;
838 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
843 error = sysctl_handle_int(oidp, &i, 0, req);
844 if (error || !req->newptr)
848 * XXXRW: Might be a good idea to sanity check the argument and place
849 * an extreme upper bound.
858 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
859 NULL, 0, sysctl_maxnipq, "I",
860 "Maximum number of IPv4 fragment reassembly queue entries");
862 #define M_IP_FRAG M_PROTO9
865 * Take incoming datagram fragment and try to reassemble it into
866 * whole datagram. If the argument is the first fragment or one
867 * in between the function will return NULL and store the mbuf
868 * in the fragment chain. If the argument is the last fragment
869 * the packet will be reassembled and the pointer to the new
870 * mbuf returned for further processing. Only m_tags attached
871 * to the first packet/fragment are preserved.
872 * The IP header is *NOT* adjusted out of iplen.
875 ip_reass(struct mbuf *m)
878 struct mbuf *p, *q, *nq, *t;
879 struct ipq *fp = NULL;
880 struct ipqhead *head;
885 uint32_t rss_hash, rss_type;
888 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
889 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
890 IPSTAT_INC(ips_fragments);
891 IPSTAT_INC(ips_fragdropped);
896 ip = mtod(m, struct ip *);
897 hlen = ip->ip_hl << 2;
899 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
904 * Look for queue of fragments
907 TAILQ_FOREACH(fp, head, ipq_list)
908 if (ip->ip_id == fp->ipq_id &&
909 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
910 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
912 mac_ipq_match(m, fp) &&
914 ip->ip_p == fp->ipq_p)
920 * Attempt to trim the number of allocated fragment queues if it
921 * exceeds the administrative limit.
923 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
925 * drop something from the tail of the current queue
926 * before proceeding further
928 struct ipq *q = TAILQ_LAST(head, ipqhead);
929 if (q == NULL) { /* gak */
930 for (i = 0; i < IPREASS_NHASH; i++) {
931 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
933 IPSTAT_ADD(ips_fragtimeout,
935 ip_freef(&V_ipq[i], r);
940 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
947 * Adjust ip_len to not reflect header,
948 * convert offset of this to bytes.
950 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
951 if (ip->ip_off & htons(IP_MF)) {
953 * Make sure that fragments have a data length
954 * that's a non-zero multiple of 8 bytes.
956 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
957 IPSTAT_INC(ips_toosmall); /* XXX */
960 m->m_flags |= M_IP_FRAG;
962 m->m_flags &= ~M_IP_FRAG;
963 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
966 * Attempt reassembly; if it succeeds, proceed.
967 * ip_reass() will return a different mbuf.
969 IPSTAT_INC(ips_fragments);
970 m->m_pkthdr.PH_loc.ptr = ip;
972 /* Previous ip_reass() started here. */
974 * Presence of header sizes in mbufs
975 * would confuse code below.
981 * If first fragment to arrive, create a reassembly queue.
984 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
988 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
989 uma_zfree(V_ipq_zone, fp);
993 mac_ipq_create(m, fp);
995 TAILQ_INSERT_HEAD(head, fp, ipq_list);
998 fp->ipq_ttl = IPFRAGTTL;
999 fp->ipq_p = ip->ip_p;
1000 fp->ipq_id = ip->ip_id;
1001 fp->ipq_src = ip->ip_src;
1002 fp->ipq_dst = ip->ip_dst;
1004 m->m_nextpkt = NULL;
1009 mac_ipq_update(m, fp);
1013 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
1016 * Handle ECN by comparing this segment with the first one;
1017 * if CE is set, do not lose CE.
1018 * drop if CE and not-ECT are mixed for the same packet.
1020 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1021 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1022 if (ecn == IPTOS_ECN_CE) {
1023 if (ecn0 == IPTOS_ECN_NOTECT)
1025 if (ecn0 != IPTOS_ECN_CE)
1026 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1028 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1032 * Find a segment which begins after this one does.
1034 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1035 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
1039 * If there is a preceding segment, it may provide some of
1040 * our data already. If so, drop the data from the incoming
1041 * segment. If it provides all of our data, drop us, otherwise
1042 * stick new segment in the proper place.
1044 * If some of the data is dropped from the preceding
1045 * segment, then it's checksum is invalidated.
1048 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
1051 if (i >= ntohs(ip->ip_len))
1054 m->m_pkthdr.csum_flags = 0;
1055 ip->ip_off = htons(ntohs(ip->ip_off) + i);
1056 ip->ip_len = htons(ntohs(ip->ip_len) - i);
1058 m->m_nextpkt = p->m_nextpkt;
1061 m->m_nextpkt = fp->ipq_frags;
1066 * While we overlap succeeding segments trim them or,
1067 * if they are completely covered, dequeue them.
1069 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1070 ntohs(GETIP(q)->ip_off); q = nq) {
1071 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1072 ntohs(GETIP(q)->ip_off);
1073 if (i < ntohs(GETIP(q)->ip_len)) {
1074 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1075 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1077 q->m_pkthdr.csum_flags = 0;
1082 IPSTAT_INC(ips_fragdropped);
1088 * Check for complete reassembly and perform frag per packet
1091 * Frag limiting is performed here so that the nth frag has
1092 * a chance to complete the packet before we drop the packet.
1093 * As a result, n+1 frags are actually allowed per packet, but
1094 * only n will ever be stored. (n = maxfragsperpacket.)
1098 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1099 if (ntohs(GETIP(q)->ip_off) != next) {
1100 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1101 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1106 next += ntohs(GETIP(q)->ip_len);
1108 /* Make sure the last packet didn't have the IP_MF flag */
1109 if (p->m_flags & M_IP_FRAG) {
1110 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1111 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1118 * Reassembly is complete. Make sure the packet is a sane size.
1122 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1123 IPSTAT_INC(ips_toolong);
1124 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1130 * Concatenate fragments.
1137 q->m_nextpkt = NULL;
1138 for (q = nq; q != NULL; q = nq) {
1140 q->m_nextpkt = NULL;
1141 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1142 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1146 * In order to do checksumming faster we do 'end-around carry' here
1147 * (and not in for{} loop), though it implies we are not going to
1148 * reassemble more than 64k fragments.
1150 while (m->m_pkthdr.csum_data & 0xffff0000)
1151 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1152 (m->m_pkthdr.csum_data >> 16);
1154 mac_ipq_reassemble(fp, m);
1155 mac_ipq_destroy(fp);
1159 * Create header for new ip packet by modifying header of first
1160 * packet; dequeue and discard fragment reassembly header.
1161 * Make header visible.
1163 ip->ip_len = htons((ip->ip_hl << 2) + next);
1164 ip->ip_src = fp->ipq_src;
1165 ip->ip_dst = fp->ipq_dst;
1166 TAILQ_REMOVE(head, fp, ipq_list);
1168 uma_zfree(V_ipq_zone, fp);
1169 m->m_len += (ip->ip_hl << 2);
1170 m->m_data -= (ip->ip_hl << 2);
1171 /* some debugging cruft by sklower, below, will go away soon */
1172 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1174 IPSTAT_INC(ips_reassembled);
1179 * Query the RSS layer for the flowid / flowtype for the
1182 * For now, just assume we have to calculate a new one.
1183 * Later on we should check to see if the assigned flowid matches
1184 * what RSS wants for the given IP protocol and if so, just keep it.
1186 * We then queue into the relevant netisr so it can be dispatched
1187 * to the correct CPU.
1189 * Note - this may return 1, which means the flowid in the mbuf
1190 * is correct for the configured RSS hash types and can be used.
1192 if (rss_mbuf_software_hash_v4(m, 0, &rss_hash, &rss_type) == 0) {
1193 m->m_pkthdr.flowid = rss_hash;
1194 M_HASHTYPE_SET(m, rss_type);
1198 * Queue/dispatch for reprocessing.
1200 * Note: this is much slower than just handling the frame in the
1201 * current receive context. It's likely worth investigating
1204 netisr_dispatch(NETISR_IP_DIRECT, m);
1208 /* Handle in-line */
1212 IPSTAT_INC(ips_fragdropped);
1224 * Free a fragment reassembly header and all
1225 * associated datagrams.
1228 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1234 while (fp->ipq_frags) {
1236 fp->ipq_frags = q->m_nextpkt;
1239 TAILQ_REMOVE(fhp, fp, ipq_list);
1240 uma_zfree(V_ipq_zone, fp);
1245 * IP timer processing;
1246 * if a timer expires on a reassembly
1247 * queue, discard it.
1252 VNET_ITERATOR_DECL(vnet_iter);
1256 VNET_LIST_RLOCK_NOSLEEP();
1258 VNET_FOREACH(vnet_iter) {
1259 CURVNET_SET(vnet_iter);
1260 for (i = 0; i < IPREASS_NHASH; i++) {
1261 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1265 fp = TAILQ_NEXT(fp, ipq_list);
1266 if(--fpp->ipq_ttl == 0) {
1267 IPSTAT_ADD(ips_fragtimeout,
1269 ip_freef(&V_ipq[i], fpp);
1274 * If we are over the maximum number of fragments
1275 * (due to the limit being lowered), drain off
1276 * enough to get down to the new limit.
1278 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1279 for (i = 0; i < IPREASS_NHASH; i++) {
1280 while (V_nipq > V_maxnipq &&
1281 !TAILQ_EMPTY(&V_ipq[i])) {
1282 IPSTAT_ADD(ips_fragdropped,
1283 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1285 TAILQ_FIRST(&V_ipq[i]));
1292 VNET_LIST_RUNLOCK_NOSLEEP();
1296 * Drain off all datagram fragments.
1299 ip_drain_locked(void)
1305 for (i = 0; i < IPREASS_NHASH; i++) {
1306 while(!TAILQ_EMPTY(&V_ipq[i])) {
1307 IPSTAT_ADD(ips_fragdropped,
1308 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1309 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1317 VNET_ITERATOR_DECL(vnet_iter);
1319 VNET_LIST_RLOCK_NOSLEEP();
1321 VNET_FOREACH(vnet_iter) {
1322 CURVNET_SET(vnet_iter);
1327 VNET_LIST_RUNLOCK_NOSLEEP();
1331 * The protocol to be inserted into ip_protox[] must be already registered
1332 * in inetsw[], either statically or through pf_proto_register().
1335 ipproto_register(short ipproto)
1339 /* Sanity checks. */
1340 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1341 return (EPROTONOSUPPORT);
1344 * The protocol slot must not be occupied by another protocol
1345 * already. An index pointing to IPPROTO_RAW is unused.
1347 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1349 return (EPFNOSUPPORT);
1350 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1353 /* Find the protocol position in inetsw[] and set the index. */
1354 for (pr = inetdomain.dom_protosw;
1355 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1356 if (pr->pr_domain->dom_family == PF_INET &&
1357 pr->pr_protocol && pr->pr_protocol == ipproto) {
1358 ip_protox[pr->pr_protocol] = pr - inetsw;
1362 return (EPROTONOSUPPORT);
1366 ipproto_unregister(short ipproto)
1370 /* Sanity checks. */
1371 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1372 return (EPROTONOSUPPORT);
1374 /* Check if the protocol was indeed registered. */
1375 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1377 return (EPFNOSUPPORT);
1378 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1381 /* Reset the protocol slot to IPPROTO_RAW. */
1382 ip_protox[ipproto] = pr - inetsw;
1387 * Given address of next destination (final or next hop), return (referenced)
1388 * internet address info of interface to be used to get there.
1391 ip_rtaddr(struct in_addr dst, u_int fibnum)
1394 struct sockaddr_in *sin;
1395 struct in_ifaddr *ia;
1397 bzero(&sro, sizeof(sro));
1398 sin = (struct sockaddr_in *)&sro.ro_dst;
1399 sin->sin_family = AF_INET;
1400 sin->sin_len = sizeof(*sin);
1401 sin->sin_addr = dst;
1402 in_rtalloc_ign(&sro, 0, fibnum);
1404 if (sro.ro_rt == NULL)
1407 ia = ifatoia(sro.ro_rt->rt_ifa);
1408 ifa_ref(&ia->ia_ifa);
1413 u_char inetctlerrmap[PRC_NCMDS] = {
1415 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1416 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1417 EMSGSIZE, EHOSTUNREACH, 0, 0,
1418 0, 0, EHOSTUNREACH, 0,
1419 ENOPROTOOPT, ECONNREFUSED
1423 * Forward a packet. If some error occurs return the sender
1424 * an icmp packet. Note we can't always generate a meaningful
1425 * icmp message because icmp doesn't have a large enough repertoire
1426 * of codes and types.
1428 * If not forwarding, just drop the packet. This could be confusing
1429 * if ipforwarding was zero but some routing protocol was advancing
1430 * us as a gateway to somewhere. However, we must let the routing
1431 * protocol deal with that.
1433 * The srcrt parameter indicates whether the packet is being forwarded
1434 * via a source route.
1437 ip_forward(struct mbuf *m, int srcrt)
1439 struct ip *ip = mtod(m, struct ip *);
1440 struct in_ifaddr *ia;
1442 struct in_addr dest;
1444 int error, type = 0, code = 0, mtu = 0;
1446 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1447 IPSTAT_INC(ips_cantforward);
1452 if (ip_ipsec_fwd(m) != 0) {
1453 IPSTAT_INC(ips_cantforward);
1461 if (ip->ip_ttl <= IPTTLDEC) {
1462 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1470 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1473 * 'ia' may be NULL if there is no route for this destination.
1474 * In case of IPsec, Don't discard it just yet, but pass it to
1475 * ip_output in case of outgoing IPsec policy.
1477 if (!srcrt && ia == NULL) {
1478 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1484 * Save the IP header and at most 8 bytes of the payload,
1485 * in case we need to generate an ICMP message to the src.
1487 * XXX this can be optimized a lot by saving the data in a local
1488 * buffer on the stack (72 bytes at most), and only allocating the
1489 * mbuf if really necessary. The vast majority of the packets
1490 * are forwarded without having to send an ICMP back (either
1491 * because unnecessary, or because rate limited), so we are
1492 * really we are wasting a lot of work here.
1494 * We don't use m_copy() because it might return a reference
1495 * to a shared cluster. Both this function and ip_output()
1496 * assume exclusive access to the IP header in `m', so any
1497 * data in a cluster may change before we reach icmp_error().
1499 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1500 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1502 * It's probably ok if the pkthdr dup fails (because
1503 * the deep copy of the tag chain failed), but for now
1504 * be conservative and just discard the copy since
1505 * code below may some day want the tags.
1510 if (mcopy != NULL) {
1511 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1512 mcopy->m_pkthdr.len = mcopy->m_len;
1513 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1519 ip->ip_ttl -= IPTTLDEC;
1525 * If forwarding packet using same interface that it came in on,
1526 * perhaps should send a redirect to sender to shortcut a hop.
1527 * Only send redirect if source is sending directly to us,
1528 * and if packet was not source routed (or has any options).
1529 * Also, don't send redirect if forwarding using a default route
1530 * or a route modified by a redirect.
1533 if (!srcrt && V_ipsendredirects &&
1534 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1535 struct sockaddr_in *sin;
1538 bzero(&ro, sizeof(ro));
1539 sin = (struct sockaddr_in *)&ro.ro_dst;
1540 sin->sin_family = AF_INET;
1541 sin->sin_len = sizeof(*sin);
1542 sin->sin_addr = ip->ip_dst;
1543 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1547 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1548 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1549 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1550 u_long src = ntohl(ip->ip_src.s_addr);
1553 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1554 if (rt->rt_flags & RTF_GATEWAY)
1555 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1557 dest.s_addr = ip->ip_dst.s_addr;
1558 /* Router requirements says to only send host redirects */
1559 type = ICMP_REDIRECT;
1560 code = ICMP_REDIRECT_HOST;
1568 * Try to cache the route MTU from ip_output so we can consider it for
1569 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1571 bzero(&ro, sizeof(ro));
1573 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1575 if (error == EMSGSIZE && ro.ro_rt)
1576 mtu = ro.ro_rt->rt_mtu;
1580 IPSTAT_INC(ips_cantforward);
1582 IPSTAT_INC(ips_forward);
1584 IPSTAT_INC(ips_redirectsent);
1589 ifa_free(&ia->ia_ifa);
1593 if (mcopy == NULL) {
1595 ifa_free(&ia->ia_ifa);
1601 case 0: /* forwarded, but need redirect */
1602 /* type, code set above */
1610 type = ICMP_UNREACH;
1611 code = ICMP_UNREACH_HOST;
1615 type = ICMP_UNREACH;
1616 code = ICMP_UNREACH_NEEDFRAG;
1620 * If IPsec is configured for this path,
1621 * override any possibly mtu value set by ip_output.
1623 mtu = ip_ipsec_mtu(mcopy, mtu);
1626 * If the MTU was set before make sure we are below the
1628 * If the MTU wasn't set before use the interface mtu or
1629 * fall back to the next smaller mtu step compared to the
1630 * current packet size.
1634 mtu = min(mtu, ia->ia_ifp->if_mtu);
1637 mtu = ia->ia_ifp->if_mtu;
1639 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1641 IPSTAT_INC(ips_cantfrag);
1645 case EACCES: /* ipfw denied packet */
1648 ifa_free(&ia->ia_ifa);
1652 ifa_free(&ia->ia_ifa);
1653 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1657 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1661 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1665 if (inp->inp_socket->so_options & SO_BINTIME) {
1666 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1667 SCM_BINTIME, SOL_SOCKET);
1669 mp = &(*mp)->m_next;
1671 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1674 bintime2timeval(&bt, &tv);
1675 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1676 SCM_TIMESTAMP, SOL_SOCKET);
1678 mp = &(*mp)->m_next;
1681 if (inp->inp_flags & INP_RECVDSTADDR) {
1682 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1683 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1685 mp = &(*mp)->m_next;
1687 if (inp->inp_flags & INP_RECVTTL) {
1688 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1689 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1691 mp = &(*mp)->m_next;
1695 * Moving these out of udp_input() made them even more broken
1696 * than they already were.
1698 /* options were tossed already */
1699 if (inp->inp_flags & INP_RECVOPTS) {
1700 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1701 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1703 mp = &(*mp)->m_next;
1705 /* ip_srcroute doesn't do what we want here, need to fix */
1706 if (inp->inp_flags & INP_RECVRETOPTS) {
1707 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1708 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1710 mp = &(*mp)->m_next;
1713 if (inp->inp_flags & INP_RECVIF) {
1716 struct sockaddr_dl sdl;
1719 struct sockaddr_dl *sdp;
1720 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1722 if ((ifp = m->m_pkthdr.rcvif) &&
1723 ifp->if_index && ifp->if_index <= V_if_index) {
1724 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1726 * Change our mind and don't try copy.
1728 if (sdp->sdl_family != AF_LINK ||
1729 sdp->sdl_len > sizeof(sdlbuf)) {
1732 bcopy(sdp, sdl2, sdp->sdl_len);
1736 offsetof(struct sockaddr_dl, sdl_data[0]);
1737 sdl2->sdl_family = AF_LINK;
1738 sdl2->sdl_index = 0;
1739 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1741 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1742 IP_RECVIF, IPPROTO_IP);
1744 mp = &(*mp)->m_next;
1746 if (inp->inp_flags & INP_RECVTOS) {
1747 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1748 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1750 mp = &(*mp)->m_next;
1753 if (inp->inp_flags2 & INP_RECVFLOWID) {
1754 uint32_t flowid, flow_type;
1756 flowid = m->m_pkthdr.flowid;
1757 flow_type = M_HASHTYPE_GET(m);
1760 * XXX should handle the failure of one or the
1761 * other - don't populate both?
1763 *mp = sbcreatecontrol((caddr_t) &flowid,
1764 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1766 mp = &(*mp)->m_next;
1767 *mp = sbcreatecontrol((caddr_t) &flow_type,
1768 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1770 mp = &(*mp)->m_next;
1774 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1775 uint32_t flowid, flow_type;
1776 uint32_t rss_bucketid;
1778 flowid = m->m_pkthdr.flowid;
1779 flow_type = M_HASHTYPE_GET(m);
1781 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1782 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1783 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1785 mp = &(*mp)->m_next;
1792 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1793 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1794 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1797 static VNET_DEFINE(int, ip_rsvp_on);
1798 VNET_DEFINE(struct socket *, ip_rsvpd);
1800 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1803 ip_rsvp_init(struct socket *so)
1806 if (so->so_type != SOCK_RAW ||
1807 so->so_proto->pr_protocol != IPPROTO_RSVP)
1810 if (V_ip_rsvpd != NULL)
1815 * This may seem silly, but we need to be sure we don't over-increment
1816 * the RSVP counter, in case something slips up.
1818 if (!V_ip_rsvp_on) {
1832 * This may seem silly, but we need to be sure we don't over-decrement
1833 * the RSVP counter, in case something slips up.
1843 rsvp_input(struct mbuf **mp, int *offp, int proto)
1850 if (rsvp_input_p) { /* call the real one if loaded */
1852 rsvp_input_p(mp, offp, proto);
1853 return (IPPROTO_DONE);
1856 /* Can still get packets with rsvp_on = 0 if there is a local member
1857 * of the group to which the RSVP packet is addressed. But in this
1858 * case we want to throw the packet away.
1863 return (IPPROTO_DONE);
1866 if (V_ip_rsvpd != NULL) {
1868 rip_input(mp, offp, proto);
1869 return (IPPROTO_DONE);
1871 /* Drop the packet */
1873 return (IPPROTO_DONE);