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_kdtrace.h"
40 #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>
65 #include <net/flowtable.h>
67 #include <netinet/in.h>
68 #include <netinet/in_kdtrace.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_fw.h>
75 #include <netinet/ip_icmp.h>
76 #include <netinet/ip_options.h>
77 #include <machine/in_cksum.h>
78 #include <netinet/ip_carp.h>
80 #include <netinet/ip_ipsec.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,
149 .nh_policy = NETISR_POLICY_FLOW,
152 extern struct domain inetdomain;
153 extern struct protosw inetsw[];
154 u_char ip_protox[IPPROTO_MAX];
155 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
156 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
157 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
159 static VNET_DEFINE(uma_zone_t, ipq_zone);
160 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
161 static struct mtx ipqlock;
163 #define V_ipq_zone VNET(ipq_zone)
164 #define V_ipq VNET(ipq)
166 #define IPQ_LOCK() mtx_lock(&ipqlock)
167 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
168 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
169 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
171 static void maxnipq_update(void);
172 static void ipq_zone_change(void *);
173 static void ip_drain_locked(void);
175 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
176 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
177 #define V_maxnipq VNET(maxnipq)
178 #define V_nipq VNET(nipq)
179 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
181 "Current number of IPv4 fragment reassembly queue entries");
183 static VNET_DEFINE(int, maxfragsperpacket);
184 #define V_maxfragsperpacket VNET(maxfragsperpacket)
185 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
186 &VNET_NAME(maxfragsperpacket), 0,
187 "Maximum number of IPv4 fragments allowed per packet");
190 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
191 &ip_mtu, 0, "Default MTU");
195 VNET_DEFINE(int, ipstealth);
196 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
197 &VNET_NAME(ipstealth), 0,
198 "IP stealth mode, no TTL decrementation on forwarding");
202 static VNET_DEFINE(int, ip_output_flowtable_size) = 2048;
203 VNET_DEFINE(struct flowtable *, ip_ft);
204 #define V_ip_output_flowtable_size VNET(ip_output_flowtable_size)
206 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN,
207 &VNET_NAME(ip_output_flowtable_size), 2048,
208 "number of entries in the per-cpu output flow caches");
211 static void ip_freef(struct ipqhead *, struct ipq *);
214 * IP statistics are stored in the "array" of counter(9)s.
216 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
217 VNET_PCPUSTAT_SYSINIT(ipstat);
218 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
219 "IP statistics (struct ipstat, netinet/ip_var.h)");
222 VNET_PCPUSTAT_SYSUNINIT(ipstat);
226 * Kernel module interface for updating ipstat. The argument is an index
227 * into ipstat treated as an array.
230 kmod_ipstat_inc(int statnum)
233 counter_u64_add(VNET(ipstat)[statnum], 1);
237 kmod_ipstat_dec(int statnum)
240 counter_u64_add(VNET(ipstat)[statnum], -1);
244 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
248 netisr_getqlimit(&ip_nh, &qlimit);
249 error = sysctl_handle_int(oidp, &qlimit, 0, req);
250 if (error || !req->newptr)
254 return (netisr_setqlimit(&ip_nh, qlimit));
256 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
257 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
258 "Maximum size of the IP input queue");
261 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
263 u_int64_t qdrops_long;
266 netisr_getqdrops(&ip_nh, &qdrops_long);
267 qdrops = qdrops_long;
268 error = sysctl_handle_int(oidp, &qdrops, 0, req);
269 if (error || !req->newptr)
273 netisr_clearqdrops(&ip_nh);
277 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
278 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
279 "Number of packets dropped from the IP input queue");
282 * IP initialization: fill in IP protocol switch table.
283 * All protocols not implemented in kernel go to raw IP protocol handler.
291 V_ip_id = time_second & 0xffff;
293 TAILQ_INIT(&V_in_ifaddrhead);
294 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
296 /* Initialize IP reassembly queue. */
297 for (i = 0; i < IPREASS_NHASH; i++)
298 TAILQ_INIT(&V_ipq[i]);
299 V_maxnipq = nmbclusters / 32;
300 V_maxfragsperpacket = 16;
301 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
302 NULL, UMA_ALIGN_PTR, 0);
305 /* Initialize packet filter hooks. */
306 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
307 V_inet_pfil_hook.ph_af = AF_INET;
308 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
309 printf("%s: WARNING: unable to register pfil hook, "
310 "error %d\n", __func__, i);
313 if (TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size",
314 &V_ip_output_flowtable_size)) {
315 if (V_ip_output_flowtable_size < 256)
316 V_ip_output_flowtable_size = 256;
317 if (!powerof2(V_ip_output_flowtable_size)) {
318 printf("flowtable must be power of 2 size\n");
319 V_ip_output_flowtable_size = 2048;
323 * round up to the next power of 2
325 V_ip_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
327 V_ip_ft = flowtable_alloc("ipv4", V_ip_output_flowtable_size, FL_PCPU);
330 /* Skip initialization of globals for non-default instances. */
331 if (!IS_DEFAULT_VNET(curvnet))
334 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
336 panic("ip_init: PF_INET not found");
338 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
339 for (i = 0; i < IPPROTO_MAX; i++)
340 ip_protox[i] = pr - inetsw;
342 * Cycle through IP protocols and put them into the appropriate place
345 for (pr = inetdomain.dom_protosw;
346 pr < inetdomain.dom_protoswNPROTOSW; pr++)
347 if (pr->pr_domain->dom_family == PF_INET &&
348 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
349 /* Be careful to only index valid IP protocols. */
350 if (pr->pr_protocol < IPPROTO_MAX)
351 ip_protox[pr->pr_protocol] = pr - inetsw;
354 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
355 NULL, EVENTHANDLER_PRI_ANY);
357 /* Initialize various other remaining things. */
359 netisr_register(&ip_nh);
368 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
369 printf("%s: WARNING: unable to unregister pfil hook, "
370 "error %d\n", __func__, i);
372 /* Cleanup in_ifaddr hash table; should be empty. */
373 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
379 uma_zdestroy(V_ipq_zone);
384 * Ip input routine. Checksum and byte swap header. If fragmented
385 * try to reassemble. Process options. Pass to next level.
388 ip_input(struct mbuf *m)
390 struct ip *ip = NULL;
391 struct in_ifaddr *ia = NULL;
394 int checkif, hlen = 0;
395 uint16_t sum, ip_len;
396 int dchg = 0; /* dest changed after fw */
397 struct in_addr odst; /* original dst address */
401 if (m->m_flags & M_FASTFWD_OURS) {
402 m->m_flags &= ~M_FASTFWD_OURS;
403 /* Set up some basics that will be used later. */
404 ip = mtod(m, struct ip *);
405 hlen = ip->ip_hl << 2;
406 ip_len = ntohs(ip->ip_len);
410 IPSTAT_INC(ips_total);
412 if (m->m_pkthdr.len < sizeof(struct ip))
415 if (m->m_len < sizeof (struct ip) &&
416 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
417 IPSTAT_INC(ips_toosmall);
420 ip = mtod(m, struct ip *);
422 if (ip->ip_v != IPVERSION) {
423 IPSTAT_INC(ips_badvers);
427 hlen = ip->ip_hl << 2;
428 if (hlen < sizeof(struct ip)) { /* minimum header length */
429 IPSTAT_INC(ips_badhlen);
432 if (hlen > m->m_len) {
433 if ((m = m_pullup(m, hlen)) == NULL) {
434 IPSTAT_INC(ips_badhlen);
437 ip = mtod(m, struct ip *);
440 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
442 /* 127/8 must not appear on wire - RFC1122 */
443 ifp = m->m_pkthdr.rcvif;
444 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
445 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
446 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
447 IPSTAT_INC(ips_badaddr);
452 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
453 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
455 if (hlen == sizeof(struct ip)) {
456 sum = in_cksum_hdr(ip);
458 sum = in_cksum(m, hlen);
462 IPSTAT_INC(ips_badsum);
467 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
468 /* packet is dropped by traffic conditioner */
472 ip_len = ntohs(ip->ip_len);
474 IPSTAT_INC(ips_badlen);
479 * Check that the amount of data in the buffers
480 * is as at least much as the IP header would have us expect.
481 * Trim mbufs if longer than we expect.
482 * Drop packet if shorter than we expect.
484 if (m->m_pkthdr.len < ip_len) {
486 IPSTAT_INC(ips_tooshort);
489 if (m->m_pkthdr.len > ip_len) {
490 if (m->m_len == m->m_pkthdr.len) {
492 m->m_pkthdr.len = ip_len;
494 m_adj(m, ip_len - m->m_pkthdr.len);
498 * Bypass packet filtering for packets previously handled by IPsec.
500 if (ip_ipsec_filtertunnel(m))
505 * Run through list of hooks for input packets.
507 * NB: Beware of the destination address changing (e.g.
508 * by NAT rewriting). When this happens, tell
509 * ip_forward to do the right thing.
512 /* Jump over all PFIL processing if hooks are not active. */
513 if (!PFIL_HOOKED(&V_inet_pfil_hook))
517 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
519 if (m == NULL) /* consumed by filter */
522 ip = mtod(m, struct ip *);
523 dchg = (odst.s_addr != ip->ip_dst.s_addr);
524 ifp = m->m_pkthdr.rcvif;
526 if (m->m_flags & M_FASTFWD_OURS) {
527 m->m_flags &= ~M_FASTFWD_OURS;
530 if (m->m_flags & M_IP_NEXTHOP) {
531 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
534 * Directly ship the packet on. This allows
535 * forwarding packets originally destined to us
536 * to some other directly connected host.
545 * Process options and, if not destined for us,
546 * ship it on. ip_dooptions returns 1 when an
547 * error was detected (causing an icmp message
548 * to be sent and the original packet to be freed).
550 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
553 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
554 * matter if it is destined to another node, or whether it is
555 * a multicast one, RSVP wants it! and prevents it from being forwarded
556 * anywhere else. Also checks if the rsvp daemon is running before
557 * grabbing the packet.
559 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
563 * Check our list of addresses, to see if the packet is for us.
564 * If we don't have any addresses, assume any unicast packet
565 * we receive might be for us (and let the upper layers deal
568 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
569 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
573 * Enable a consistency check between the destination address
574 * and the arrival interface for a unicast packet (the RFC 1122
575 * strong ES model) if IP forwarding is disabled and the packet
576 * is not locally generated and the packet is not subject to
579 * XXX - Checking also should be disabled if the destination
580 * address is ipnat'ed to a different interface.
582 * XXX - Checking is incompatible with IP aliases added
583 * to the loopback interface instead of the interface where
584 * the packets are received.
586 * XXX - This is the case for carp vhost IPs as well so we
587 * insert a workaround. If the packet got here, we already
588 * checked with carp_iamatch() and carp_forus().
590 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
591 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
592 ifp->if_carp == NULL && (dchg == 0);
595 * Check for exact addresses in the hash bucket.
597 /* IN_IFADDR_RLOCK(); */
598 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
600 * If the address matches, verify that the packet
601 * arrived via the correct interface if checking is
604 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
605 (!checkif || ia->ia_ifp == ifp)) {
606 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
607 counter_u64_add(ia->ia_ifa.ifa_ibytes,
609 /* IN_IFADDR_RUNLOCK(); */
613 /* IN_IFADDR_RUNLOCK(); */
616 * Check for broadcast addresses.
618 * Only accept broadcast packets that arrive via the matching
619 * interface. Reception of forwarded directed broadcasts would
620 * be handled via ip_forward() and ether_output() with the loopback
621 * into the stack for SIMPLEX interfaces handled by ether_output().
623 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
625 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
626 if (ifa->ifa_addr->sa_family != AF_INET)
629 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
631 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
632 counter_u64_add(ia->ia_ifa.ifa_ibytes,
634 IF_ADDR_RUNLOCK(ifp);
638 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
639 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
640 counter_u64_add(ia->ia_ifa.ifa_ibytes,
642 IF_ADDR_RUNLOCK(ifp);
647 IF_ADDR_RUNLOCK(ifp);
650 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
651 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
652 IPSTAT_INC(ips_cantforward);
656 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
659 * If we are acting as a multicast router, all
660 * incoming multicast packets are passed to the
661 * kernel-level multicast forwarding function.
662 * The packet is returned (relatively) intact; if
663 * ip_mforward() returns a non-zero value, the packet
664 * must be discarded, else it may be accepted below.
666 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
667 IPSTAT_INC(ips_cantforward);
673 * The process-level routing daemon needs to receive
674 * all multicast IGMP packets, whether or not this
675 * host belongs to their destination groups.
677 if (ip->ip_p == IPPROTO_IGMP)
679 IPSTAT_INC(ips_forward);
682 * Assume the packet is for us, to avoid prematurely taking
683 * a lock on the in_multi hash. Protocols must perform
684 * their own filtering and update statistics accordingly.
688 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
690 if (ip->ip_dst.s_addr == INADDR_ANY)
694 * FAITH(Firewall Aided Internet Translator)
696 if (ifp && ifp->if_type == IFT_FAITH) {
697 if (V_ip_keepfaith) {
698 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
706 * Not for us; forward if possible and desirable.
708 if (V_ipforwarding == 0) {
709 IPSTAT_INC(ips_cantforward);
723 * IPSTEALTH: Process non-routing options only
724 * if the packet is destined for us.
726 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
728 #endif /* IPSTEALTH */
731 * Attempt reassembly; if it succeeds, proceed.
732 * ip_reass() will return a different mbuf.
734 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
738 ip = mtod(m, struct ip *);
739 /* Get the header length of the reassembled packet */
740 hlen = ip->ip_hl << 2;
745 * enforce IPsec policy checking if we are seeing last header.
746 * note that we do not visit this with protocols with pcb layer
747 * code - like udp/tcp/raw ip.
749 if (ip_ipsec_input(m))
754 * Switch out to protocol's input routine.
756 IPSTAT_INC(ips_delivered);
758 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
765 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
766 * max has slightly different semantics than the sysctl, for historical
774 * -1 for unlimited allocation.
777 uma_zone_set_max(V_ipq_zone, 0);
779 * Positive number for specific bound.
782 uma_zone_set_max(V_ipq_zone, V_maxnipq);
784 * Zero specifies no further fragment queue allocation -- set the
785 * bound very low, but rely on implementation elsewhere to actually
786 * prevent allocation and reclaim current queues.
789 uma_zone_set_max(V_ipq_zone, 1);
793 ipq_zone_change(void *tag)
796 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
797 V_maxnipq = nmbclusters / 32;
803 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
808 error = sysctl_handle_int(oidp, &i, 0, req);
809 if (error || !req->newptr)
813 * XXXRW: Might be a good idea to sanity check the argument and place
814 * an extreme upper bound.
823 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
824 NULL, 0, sysctl_maxnipq, "I",
825 "Maximum number of IPv4 fragment reassembly queue entries");
828 * Take incoming datagram fragment and try to reassemble it into
829 * whole datagram. If the argument is the first fragment or one
830 * in between the function will return NULL and store the mbuf
831 * in the fragment chain. If the argument is the last fragment
832 * the packet will be reassembled and the pointer to the new
833 * mbuf returned for further processing. Only m_tags attached
834 * to the first packet/fragment are preserved.
835 * The IP header is *NOT* adjusted out of iplen.
838 ip_reass(struct mbuf *m)
841 struct mbuf *p, *q, *nq, *t;
842 struct ipq *fp = NULL;
843 struct ipqhead *head;
848 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
849 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
850 IPSTAT_INC(ips_fragments);
851 IPSTAT_INC(ips_fragdropped);
856 ip = mtod(m, struct ip *);
857 hlen = ip->ip_hl << 2;
859 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
864 * Look for queue of fragments
867 TAILQ_FOREACH(fp, head, ipq_list)
868 if (ip->ip_id == fp->ipq_id &&
869 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
870 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
872 mac_ipq_match(m, fp) &&
874 ip->ip_p == fp->ipq_p)
880 * Attempt to trim the number of allocated fragment queues if it
881 * exceeds the administrative limit.
883 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
885 * drop something from the tail of the current queue
886 * before proceeding further
888 struct ipq *q = TAILQ_LAST(head, ipqhead);
889 if (q == NULL) { /* gak */
890 for (i = 0; i < IPREASS_NHASH; i++) {
891 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
893 IPSTAT_ADD(ips_fragtimeout,
895 ip_freef(&V_ipq[i], r);
900 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
907 * Adjust ip_len to not reflect header,
908 * convert offset of this to bytes.
910 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
911 if (ip->ip_off & htons(IP_MF)) {
913 * Make sure that fragments have a data length
914 * that's a non-zero multiple of 8 bytes.
916 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
917 IPSTAT_INC(ips_toosmall); /* XXX */
920 m->m_flags |= M_IP_FRAG;
922 m->m_flags &= ~M_IP_FRAG;
923 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
926 * Attempt reassembly; if it succeeds, proceed.
927 * ip_reass() will return a different mbuf.
929 IPSTAT_INC(ips_fragments);
930 m->m_pkthdr.PH_loc.ptr = ip;
932 /* Previous ip_reass() started here. */
934 * Presence of header sizes in mbufs
935 * would confuse code below.
941 * If first fragment to arrive, create a reassembly queue.
944 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
948 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
949 uma_zfree(V_ipq_zone, fp);
953 mac_ipq_create(m, fp);
955 TAILQ_INSERT_HEAD(head, fp, ipq_list);
958 fp->ipq_ttl = IPFRAGTTL;
959 fp->ipq_p = ip->ip_p;
960 fp->ipq_id = ip->ip_id;
961 fp->ipq_src = ip->ip_src;
962 fp->ipq_dst = ip->ip_dst;
969 mac_ipq_update(m, fp);
973 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
976 * Handle ECN by comparing this segment with the first one;
977 * if CE is set, do not lose CE.
978 * drop if CE and not-ECT are mixed for the same packet.
980 ecn = ip->ip_tos & IPTOS_ECN_MASK;
981 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
982 if (ecn == IPTOS_ECN_CE) {
983 if (ecn0 == IPTOS_ECN_NOTECT)
985 if (ecn0 != IPTOS_ECN_CE)
986 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
988 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
992 * Find a segment which begins after this one does.
994 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
995 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
999 * If there is a preceding segment, it may provide some of
1000 * our data already. If so, drop the data from the incoming
1001 * segment. If it provides all of our data, drop us, otherwise
1002 * stick new segment in the proper place.
1004 * If some of the data is dropped from the preceding
1005 * segment, then it's checksum is invalidated.
1008 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
1011 if (i >= ntohs(ip->ip_len))
1014 m->m_pkthdr.csum_flags = 0;
1015 ip->ip_off = htons(ntohs(ip->ip_off) + i);
1016 ip->ip_len = htons(ntohs(ip->ip_len) - i);
1018 m->m_nextpkt = p->m_nextpkt;
1021 m->m_nextpkt = fp->ipq_frags;
1026 * While we overlap succeeding segments trim them or,
1027 * if they are completely covered, dequeue them.
1029 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1030 ntohs(GETIP(q)->ip_off); q = nq) {
1031 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1032 ntohs(GETIP(q)->ip_off);
1033 if (i < ntohs(GETIP(q)->ip_len)) {
1034 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1035 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1037 q->m_pkthdr.csum_flags = 0;
1042 IPSTAT_INC(ips_fragdropped);
1048 * Check for complete reassembly and perform frag per packet
1051 * Frag limiting is performed here so that the nth frag has
1052 * a chance to complete the packet before we drop the packet.
1053 * As a result, n+1 frags are actually allowed per packet, but
1054 * only n will ever be stored. (n = maxfragsperpacket.)
1058 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1059 if (ntohs(GETIP(q)->ip_off) != next) {
1060 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1061 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1066 next += ntohs(GETIP(q)->ip_len);
1068 /* Make sure the last packet didn't have the IP_MF flag */
1069 if (p->m_flags & M_IP_FRAG) {
1070 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1071 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1078 * Reassembly is complete. Make sure the packet is a sane size.
1082 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1083 IPSTAT_INC(ips_toolong);
1084 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1090 * Concatenate fragments.
1097 q->m_nextpkt = NULL;
1098 for (q = nq; q != NULL; q = nq) {
1100 q->m_nextpkt = NULL;
1101 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1102 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1106 * In order to do checksumming faster we do 'end-around carry' here
1107 * (and not in for{} loop), though it implies we are not going to
1108 * reassemble more than 64k fragments.
1110 m->m_pkthdr.csum_data =
1111 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1113 mac_ipq_reassemble(fp, m);
1114 mac_ipq_destroy(fp);
1118 * Create header for new ip packet by modifying header of first
1119 * packet; dequeue and discard fragment reassembly header.
1120 * Make header visible.
1122 ip->ip_len = htons((ip->ip_hl << 2) + next);
1123 ip->ip_src = fp->ipq_src;
1124 ip->ip_dst = fp->ipq_dst;
1125 TAILQ_REMOVE(head, fp, ipq_list);
1127 uma_zfree(V_ipq_zone, fp);
1128 m->m_len += (ip->ip_hl << 2);
1129 m->m_data -= (ip->ip_hl << 2);
1130 /* some debugging cruft by sklower, below, will go away soon */
1131 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1133 IPSTAT_INC(ips_reassembled);
1138 IPSTAT_INC(ips_fragdropped);
1150 * Free a fragment reassembly header and all
1151 * associated datagrams.
1154 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1160 while (fp->ipq_frags) {
1162 fp->ipq_frags = q->m_nextpkt;
1165 TAILQ_REMOVE(fhp, fp, ipq_list);
1166 uma_zfree(V_ipq_zone, fp);
1171 * IP timer processing;
1172 * if a timer expires on a reassembly
1173 * queue, discard it.
1178 VNET_ITERATOR_DECL(vnet_iter);
1182 VNET_LIST_RLOCK_NOSLEEP();
1184 VNET_FOREACH(vnet_iter) {
1185 CURVNET_SET(vnet_iter);
1186 for (i = 0; i < IPREASS_NHASH; i++) {
1187 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1191 fp = TAILQ_NEXT(fp, ipq_list);
1192 if(--fpp->ipq_ttl == 0) {
1193 IPSTAT_ADD(ips_fragtimeout,
1195 ip_freef(&V_ipq[i], fpp);
1200 * If we are over the maximum number of fragments
1201 * (due to the limit being lowered), drain off
1202 * enough to get down to the new limit.
1204 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1205 for (i = 0; i < IPREASS_NHASH; i++) {
1206 while (V_nipq > V_maxnipq &&
1207 !TAILQ_EMPTY(&V_ipq[i])) {
1208 IPSTAT_ADD(ips_fragdropped,
1209 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1211 TAILQ_FIRST(&V_ipq[i]));
1218 VNET_LIST_RUNLOCK_NOSLEEP();
1222 * Drain off all datagram fragments.
1225 ip_drain_locked(void)
1231 for (i = 0; i < IPREASS_NHASH; i++) {
1232 while(!TAILQ_EMPTY(&V_ipq[i])) {
1233 IPSTAT_ADD(ips_fragdropped,
1234 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1235 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1243 VNET_ITERATOR_DECL(vnet_iter);
1245 VNET_LIST_RLOCK_NOSLEEP();
1247 VNET_FOREACH(vnet_iter) {
1248 CURVNET_SET(vnet_iter);
1253 VNET_LIST_RUNLOCK_NOSLEEP();
1258 * The protocol to be inserted into ip_protox[] must be already registered
1259 * in inetsw[], either statically or through pf_proto_register().
1262 ipproto_register(short ipproto)
1266 /* Sanity checks. */
1267 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1268 return (EPROTONOSUPPORT);
1271 * The protocol slot must not be occupied by another protocol
1272 * already. An index pointing to IPPROTO_RAW is unused.
1274 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1276 return (EPFNOSUPPORT);
1277 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1280 /* Find the protocol position in inetsw[] and set the index. */
1281 for (pr = inetdomain.dom_protosw;
1282 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1283 if (pr->pr_domain->dom_family == PF_INET &&
1284 pr->pr_protocol && pr->pr_protocol == ipproto) {
1285 ip_protox[pr->pr_protocol] = pr - inetsw;
1289 return (EPROTONOSUPPORT);
1293 ipproto_unregister(short ipproto)
1297 /* Sanity checks. */
1298 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1299 return (EPROTONOSUPPORT);
1301 /* Check if the protocol was indeed registered. */
1302 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1304 return (EPFNOSUPPORT);
1305 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1308 /* Reset the protocol slot to IPPROTO_RAW. */
1309 ip_protox[ipproto] = pr - inetsw;
1314 * Given address of next destination (final or next hop), return (referenced)
1315 * internet address info of interface to be used to get there.
1318 ip_rtaddr(struct in_addr dst, u_int fibnum)
1321 struct sockaddr_in *sin;
1322 struct in_ifaddr *ia;
1324 bzero(&sro, sizeof(sro));
1325 sin = (struct sockaddr_in *)&sro.ro_dst;
1326 sin->sin_family = AF_INET;
1327 sin->sin_len = sizeof(*sin);
1328 sin->sin_addr = dst;
1329 in_rtalloc_ign(&sro, 0, fibnum);
1331 if (sro.ro_rt == NULL)
1334 ia = ifatoia(sro.ro_rt->rt_ifa);
1335 ifa_ref(&ia->ia_ifa);
1340 u_char inetctlerrmap[PRC_NCMDS] = {
1342 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1343 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1344 EMSGSIZE, EHOSTUNREACH, 0, 0,
1345 0, 0, EHOSTUNREACH, 0,
1346 ENOPROTOOPT, ECONNREFUSED
1350 * Forward a packet. If some error occurs return the sender
1351 * an icmp packet. Note we can't always generate a meaningful
1352 * icmp message because icmp doesn't have a large enough repertoire
1353 * of codes and types.
1355 * If not forwarding, just drop the packet. This could be confusing
1356 * if ipforwarding was zero but some routing protocol was advancing
1357 * us as a gateway to somewhere. However, we must let the routing
1358 * protocol deal with that.
1360 * The srcrt parameter indicates whether the packet is being forwarded
1361 * via a source route.
1364 ip_forward(struct mbuf *m, int srcrt)
1366 struct ip *ip = mtod(m, struct ip *);
1367 struct in_ifaddr *ia;
1369 struct in_addr dest;
1371 int error, type = 0, code = 0, mtu = 0;
1373 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1374 IPSTAT_INC(ips_cantforward);
1381 if (ip->ip_ttl <= IPTTLDEC) {
1382 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1390 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1393 * 'ia' may be NULL if there is no route for this destination.
1394 * In case of IPsec, Don't discard it just yet, but pass it to
1395 * ip_output in case of outgoing IPsec policy.
1397 if (!srcrt && ia == NULL) {
1398 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1404 * Save the IP header and at most 8 bytes of the payload,
1405 * in case we need to generate an ICMP message to the src.
1407 * XXX this can be optimized a lot by saving the data in a local
1408 * buffer on the stack (72 bytes at most), and only allocating the
1409 * mbuf if really necessary. The vast majority of the packets
1410 * are forwarded without having to send an ICMP back (either
1411 * because unnecessary, or because rate limited), so we are
1412 * really we are wasting a lot of work here.
1414 * We don't use m_copy() because it might return a reference
1415 * to a shared cluster. Both this function and ip_output()
1416 * assume exclusive access to the IP header in `m', so any
1417 * data in a cluster may change before we reach icmp_error().
1419 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1420 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1422 * It's probably ok if the pkthdr dup fails (because
1423 * the deep copy of the tag chain failed), but for now
1424 * be conservative and just discard the copy since
1425 * code below may some day want the tags.
1430 if (mcopy != NULL) {
1431 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1432 mcopy->m_pkthdr.len = mcopy->m_len;
1433 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1439 ip->ip_ttl -= IPTTLDEC;
1445 * If forwarding packet using same interface that it came in on,
1446 * perhaps should send a redirect to sender to shortcut a hop.
1447 * Only send redirect if source is sending directly to us,
1448 * and if packet was not source routed (or has any options).
1449 * Also, don't send redirect if forwarding using a default route
1450 * or a route modified by a redirect.
1453 if (!srcrt && V_ipsendredirects &&
1454 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1455 struct sockaddr_in *sin;
1458 bzero(&ro, sizeof(ro));
1459 sin = (struct sockaddr_in *)&ro.ro_dst;
1460 sin->sin_family = AF_INET;
1461 sin->sin_len = sizeof(*sin);
1462 sin->sin_addr = ip->ip_dst;
1463 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1467 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1468 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1469 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1470 u_long src = ntohl(ip->ip_src.s_addr);
1473 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1474 if (rt->rt_flags & RTF_GATEWAY)
1475 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1477 dest.s_addr = ip->ip_dst.s_addr;
1478 /* Router requirements says to only send host redirects */
1479 type = ICMP_REDIRECT;
1480 code = ICMP_REDIRECT_HOST;
1488 * Try to cache the route MTU from ip_output so we can consider it for
1489 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1491 bzero(&ro, sizeof(ro));
1493 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1495 if (error == EMSGSIZE && ro.ro_rt)
1496 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1500 IPSTAT_INC(ips_cantforward);
1502 IPSTAT_INC(ips_forward);
1504 IPSTAT_INC(ips_redirectsent);
1509 ifa_free(&ia->ia_ifa);
1513 if (mcopy == NULL) {
1515 ifa_free(&ia->ia_ifa);
1521 case 0: /* forwarded, but need redirect */
1522 /* type, code set above */
1530 type = ICMP_UNREACH;
1531 code = ICMP_UNREACH_HOST;
1535 type = ICMP_UNREACH;
1536 code = ICMP_UNREACH_NEEDFRAG;
1540 * If IPsec is configured for this path,
1541 * override any possibly mtu value set by ip_output.
1543 mtu = ip_ipsec_mtu(mcopy, mtu);
1546 * If the MTU was set before make sure we are below the
1548 * If the MTU wasn't set before use the interface mtu or
1549 * fall back to the next smaller mtu step compared to the
1550 * current packet size.
1554 mtu = min(mtu, ia->ia_ifp->if_mtu);
1557 mtu = ia->ia_ifp->if_mtu;
1559 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1561 IPSTAT_INC(ips_cantfrag);
1566 * A router should not generate ICMP_SOURCEQUENCH as
1567 * required in RFC1812 Requirements for IP Version 4 Routers.
1568 * Source quench could be a big problem under DoS attacks,
1569 * or if the underlying interface is rate-limited.
1570 * Those who need source quench packets may re-enable them
1571 * via the net.inet.ip.sendsourcequench sysctl.
1573 if (V_ip_sendsourcequench == 0) {
1576 ifa_free(&ia->ia_ifa);
1579 type = ICMP_SOURCEQUENCH;
1584 case EACCES: /* ipfw denied packet */
1587 ifa_free(&ia->ia_ifa);
1591 ifa_free(&ia->ia_ifa);
1592 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1596 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1600 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1604 if (inp->inp_socket->so_options & SO_BINTIME) {
1605 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1606 SCM_BINTIME, SOL_SOCKET);
1608 mp = &(*mp)->m_next;
1610 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1613 bintime2timeval(&bt, &tv);
1614 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1615 SCM_TIMESTAMP, SOL_SOCKET);
1617 mp = &(*mp)->m_next;
1620 if (inp->inp_flags & INP_RECVDSTADDR) {
1621 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1622 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1624 mp = &(*mp)->m_next;
1626 if (inp->inp_flags & INP_RECVTTL) {
1627 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1628 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1630 mp = &(*mp)->m_next;
1634 * Moving these out of udp_input() made them even more broken
1635 * than they already were.
1637 /* options were tossed already */
1638 if (inp->inp_flags & INP_RECVOPTS) {
1639 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1640 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1642 mp = &(*mp)->m_next;
1644 /* ip_srcroute doesn't do what we want here, need to fix */
1645 if (inp->inp_flags & INP_RECVRETOPTS) {
1646 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1647 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1649 mp = &(*mp)->m_next;
1652 if (inp->inp_flags & INP_RECVIF) {
1655 struct sockaddr_dl sdl;
1658 struct sockaddr_dl *sdp;
1659 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1661 if ((ifp = m->m_pkthdr.rcvif) &&
1662 ifp->if_index && ifp->if_index <= V_if_index) {
1663 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1665 * Change our mind and don't try copy.
1667 if (sdp->sdl_family != AF_LINK ||
1668 sdp->sdl_len > sizeof(sdlbuf)) {
1671 bcopy(sdp, sdl2, sdp->sdl_len);
1675 offsetof(struct sockaddr_dl, sdl_data[0]);
1676 sdl2->sdl_family = AF_LINK;
1677 sdl2->sdl_index = 0;
1678 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1680 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1681 IP_RECVIF, IPPROTO_IP);
1683 mp = &(*mp)->m_next;
1685 if (inp->inp_flags & INP_RECVTOS) {
1686 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1687 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1689 mp = &(*mp)->m_next;
1694 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1695 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1696 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1699 static VNET_DEFINE(int, ip_rsvp_on);
1700 VNET_DEFINE(struct socket *, ip_rsvpd);
1702 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1705 ip_rsvp_init(struct socket *so)
1708 if (so->so_type != SOCK_RAW ||
1709 so->so_proto->pr_protocol != IPPROTO_RSVP)
1712 if (V_ip_rsvpd != NULL)
1717 * This may seem silly, but we need to be sure we don't over-increment
1718 * the RSVP counter, in case something slips up.
1720 if (!V_ip_rsvp_on) {
1734 * This may seem silly, but we need to be sure we don't over-decrement
1735 * the RSVP counter, in case something slips up.
1745 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1748 if (rsvp_input_p) { /* call the real one if loaded */
1749 rsvp_input_p(m, off);
1753 /* Can still get packets with rsvp_on = 0 if there is a local member
1754 * of the group to which the RSVP packet is addressed. But in this
1755 * case we want to throw the packet away.
1763 if (V_ip_rsvpd != NULL) {
1767 /* Drop the packet */