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>
44 #include <sys/callout.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
51 #include <sys/kernel.h>
53 #include <sys/rwlock.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/vimage.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
66 #include <net/flowtable.h>
68 #include <netinet/in.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_icmp.h>
75 #include <netinet/ip_options.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/vinet.h>
79 #include <netinet/ip_carp.h>
82 #include <netinet/ip_ipsec.h>
85 #include <sys/socketvar.h>
87 #include <security/mac/mac_framework.h>
90 CTASSERT(sizeof(struct ip) == 20);
94 #ifndef VIMAGE_GLOBALS
95 struct vnet_inet vnet_inet_0;
100 static int ipsendredirects;
101 static int ip_checkinterface;
102 static int ip_keepfaith;
103 static int ip_sendsourcequench;
107 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
108 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
109 u_long in_ifaddrhmask; /* mask for hash table */
110 struct ipstat ipstat;
111 static int ip_rsvp_on;
112 struct socket *ip_rsvpd;
114 static struct ipqhead ipq[IPREASS_NHASH];
115 static int maxnipq; /* Administrative limit on # reass queues. */
116 static int maxfragsperpacket;
118 static int nipq; /* Total # of reass queues */
121 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
122 forwarding, CTLFLAG_RW, ipforwarding, 0,
123 "Enable IP forwarding between interfaces");
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
126 redirect, CTLFLAG_RW, ipsendredirects, 0,
127 "Enable sending IP redirects");
129 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
130 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
132 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
133 keepfaith, CTLFLAG_RW, ip_keepfaith, 0,
134 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
136 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
137 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
138 "Enable the transmission of source quench packets");
140 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
141 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values");
144 * XXX - Setting ip_checkinterface mostly implements the receive side of
145 * the Strong ES model described in RFC 1122, but since the routing table
146 * and transmit implementation do not implement the Strong ES model,
147 * setting this to 1 results in an odd hybrid.
149 * XXX - ip_checkinterface currently must be disabled if you use ipnat
150 * to translate the destination address to another local interface.
152 * XXX - ip_checkinterface must be disabled if you add IP aliases
153 * to the loopback interface instead of the interface where the
154 * packets for those addresses are received.
156 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
157 check_interface, CTLFLAG_RW, ip_checkinterface, 0,
158 "Verify packet arrives on correct interface");
160 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
162 static struct netisr_handler ip_nh = {
164 .nh_handler = ip_input,
165 .nh_proto = NETISR_IP,
166 .nh_policy = NETISR_POLICY_FLOW,
169 extern struct domain inetdomain;
170 extern struct protosw inetsw[];
171 u_char ip_protox[IPPROTO_MAX];
174 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
177 #ifdef VIMAGE_GLOBALS
178 static uma_zone_t ipq_zone;
180 static struct mtx ipqlock;
182 #define IPQ_LOCK() mtx_lock(&ipqlock)
183 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
184 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
185 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
187 static void maxnipq_update(void);
188 static void ipq_zone_change(void *);
190 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
192 "Current number of IPv4 fragment reassembly queue entries");
194 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
195 CTLFLAG_RW, maxfragsperpacket, 0,
196 "Maximum number of IPv4 fragments allowed per packet");
198 struct callout ipport_tick_callout;
201 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
202 &ip_mtu, 0, "Default MTU");
206 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
207 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
210 static int ip_output_flowtable_size = 2048;
211 TUNABLE_INT("net.inet.ip.output_flowtable_size", &ip_output_flowtable_size);
212 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size,
213 CTLFLAG_RDTUN, ip_output_flowtable_size, 2048,
214 "number of entries in the per-cpu output flow caches");
216 struct flowtable *ip_ft;
219 #ifdef VIMAGE_GLOBALS
223 static void ip_freef(struct ipqhead *, struct ipq *);
225 #ifndef VIMAGE_GLOBALS
226 static void vnet_inet_register(void);
228 static const vnet_modinfo_t vnet_inet_modinfo = {
229 .vmi_id = VNET_MOD_INET,
231 .vmi_size = sizeof(struct vnet_inet)
234 static void vnet_inet_register()
237 vnet_mod_register(&vnet_inet_modinfo);
240 SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0);
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.
288 INIT_VNET_INET(curvnet);
292 V_ipsendredirects = 1; /* XXX */
293 V_ip_checkinterface = 0;
295 V_ip_sendsourcequench = 0;
297 V_ip_defttl = IPDEFTTL;
298 V_ip_do_randomid = 0;
299 V_ip_id = time_second & 0xffff;
302 V_nipq = 0; /* Total # of reass queues */
304 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
305 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
306 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */
307 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */
308 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
309 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
310 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
311 V_ipport_reservedlow = 0;
312 V_ipport_randomized = 1; /* user controlled via sysctl */
313 V_ipport_randomcps = 10; /* user controlled via sysctl */
314 V_ipport_randomtime = 45; /* user controlled via sysctl */
315 V_ipport_stoprandom = 0; /* toggled by ipport_tick */
320 /* XXX global static but not instantiated in this file */
321 V_ipfastforward_active = 0;
322 V_subnetsarelocal = 0;
323 V_sameprefixcarponly = 0;
326 TAILQ_INIT(&V_in_ifaddrhead);
327 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
329 /* Initialize IP reassembly queue. */
330 for (i = 0; i < IPREASS_NHASH; i++)
331 TAILQ_INIT(&V_ipq[i]);
332 V_maxnipq = nmbclusters / 32;
333 V_maxfragsperpacket = 16;
334 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
335 NULL, UMA_ALIGN_PTR, 0);
338 /* Skip initialization of globals for non-default instances. */
339 if (!IS_DEFAULT_VNET(curvnet))
342 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
344 panic("ip_init: PF_INET not found");
346 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
347 for (i = 0; i < IPPROTO_MAX; i++)
348 ip_protox[i] = pr - inetsw;
350 * Cycle through IP protocols and put them into the appropriate place
353 for (pr = inetdomain.dom_protosw;
354 pr < inetdomain.dom_protoswNPROTOSW; pr++)
355 if (pr->pr_domain->dom_family == PF_INET &&
356 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
357 /* Be careful to only index valid IP protocols. */
358 if (pr->pr_protocol < IPPROTO_MAX)
359 ip_protox[pr->pr_protocol] = pr - inetsw;
362 /* Initialize packet filter hooks. */
363 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
364 inet_pfil_hook.ph_af = AF_INET;
365 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
366 printf("%s: WARNING: unable to register pfil hook, "
367 "error %d\n", __func__, i);
369 /* Start ipport_tick. */
370 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
371 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
372 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
373 SHUTDOWN_PRI_DEFAULT);
374 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
375 NULL, EVENTHANDLER_PRI_ANY);
377 /* Initialize various other remaining things. */
379 netisr_register(&ip_nh);
381 ip_ft = flowtable_alloc(ip_output_flowtable_size, FL_PCPU);
389 callout_stop(&ipport_tick_callout);
393 * Ip input routine. Checksum and byte swap header. If fragmented
394 * try to reassemble. Process options. Pass to next level.
397 ip_input(struct mbuf *m)
399 INIT_VNET_INET(curvnet);
400 struct ip *ip = NULL;
401 struct in_ifaddr *ia = NULL;
404 int checkif, hlen = 0;
406 int dchg = 0; /* dest changed after fw */
407 struct in_addr odst; /* original dst address */
411 if (m->m_flags & M_FASTFWD_OURS) {
413 * Firewall or NAT changed destination to local.
414 * We expect ip_len and ip_off to be in host byte order.
416 m->m_flags &= ~M_FASTFWD_OURS;
417 /* Set up some basics that will be used later. */
418 ip = mtod(m, struct ip *);
419 hlen = ip->ip_hl << 2;
423 IPSTAT_INC(ips_total);
425 if (m->m_pkthdr.len < sizeof(struct ip))
428 if (m->m_len < sizeof (struct ip) &&
429 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
430 IPSTAT_INC(ips_toosmall);
433 ip = mtod(m, struct ip *);
435 if (ip->ip_v != IPVERSION) {
436 IPSTAT_INC(ips_badvers);
440 hlen = ip->ip_hl << 2;
441 if (hlen < sizeof(struct ip)) { /* minimum header length */
442 IPSTAT_INC(ips_badhlen);
445 if (hlen > m->m_len) {
446 if ((m = m_pullup(m, hlen)) == NULL) {
447 IPSTAT_INC(ips_badhlen);
450 ip = mtod(m, struct ip *);
453 /* 127/8 must not appear on wire - RFC1122 */
454 ifp = m->m_pkthdr.rcvif;
455 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
456 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
457 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
458 IPSTAT_INC(ips_badaddr);
463 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
464 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
466 if (hlen == sizeof(struct ip)) {
467 sum = in_cksum_hdr(ip);
469 sum = in_cksum(m, hlen);
473 IPSTAT_INC(ips_badsum);
478 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
479 /* packet is dropped by traffic conditioner */
484 * Convert fields to host representation.
486 ip->ip_len = ntohs(ip->ip_len);
487 if (ip->ip_len < hlen) {
488 IPSTAT_INC(ips_badlen);
491 ip->ip_off = ntohs(ip->ip_off);
494 * Check that the amount of data in the buffers
495 * is as at least much as the IP header would have us expect.
496 * Trim mbufs if longer than we expect.
497 * Drop packet if shorter than we expect.
499 if (m->m_pkthdr.len < ip->ip_len) {
501 IPSTAT_INC(ips_tooshort);
504 if (m->m_pkthdr.len > ip->ip_len) {
505 if (m->m_len == m->m_pkthdr.len) {
506 m->m_len = ip->ip_len;
507 m->m_pkthdr.len = ip->ip_len;
509 m_adj(m, ip->ip_len - m->m_pkthdr.len);
513 * Bypass packet filtering for packets from a tunnel (gif).
515 if (ip_ipsec_filtertunnel(m))
520 * Run through list of hooks for input packets.
522 * NB: Beware of the destination address changing (e.g.
523 * by NAT rewriting). When this happens, tell
524 * ip_forward to do the right thing.
527 /* Jump over all PFIL processing if hooks are not active. */
528 if (!PFIL_HOOKED(&inet_pfil_hook))
532 if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
534 if (m == NULL) /* consumed by filter */
537 ip = mtod(m, struct ip *);
538 dchg = (odst.s_addr != ip->ip_dst.s_addr);
539 ifp = m->m_pkthdr.rcvif;
541 #ifdef IPFIREWALL_FORWARD
542 if (m->m_flags & M_FASTFWD_OURS) {
543 m->m_flags &= ~M_FASTFWD_OURS;
546 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
548 * Directly ship on the packet. This allows to forward packets
549 * that were destined for us to some other directly connected
555 #endif /* IPFIREWALL_FORWARD */
559 * Process options and, if not destined for us,
560 * ship it on. ip_dooptions returns 1 when an
561 * error was detected (causing an icmp message
562 * to be sent and the original packet to be freed).
564 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
567 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
568 * matter if it is destined to another node, or whether it is
569 * a multicast one, RSVP wants it! and prevents it from being forwarded
570 * anywhere else. Also checks if the rsvp daemon is running before
571 * grabbing the packet.
573 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
577 * Check our list of addresses, to see if the packet is for us.
578 * If we don't have any addresses, assume any unicast packet
579 * we receive might be for us (and let the upper layers deal
582 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
583 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
587 * Enable a consistency check between the destination address
588 * and the arrival interface for a unicast packet (the RFC 1122
589 * strong ES model) if IP forwarding is disabled and the packet
590 * is not locally generated and the packet is not subject to
593 * XXX - Checking also should be disabled if the destination
594 * address is ipnat'ed to a different interface.
596 * XXX - Checking is incompatible with IP aliases added
597 * to the loopback interface instead of the interface where
598 * the packets are received.
600 * XXX - This is the case for carp vhost IPs as well so we
601 * insert a workaround. If the packet got here, we already
602 * checked with carp_iamatch() and carp_forus().
604 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
605 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
612 * Check for exact addresses in the hash bucket.
614 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
616 * If the address matches, verify that the packet
617 * arrived via the correct interface if checking is
620 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
621 (!checkif || ia->ia_ifp == ifp))
625 * Check for broadcast addresses.
627 * Only accept broadcast packets that arrive via the matching
628 * interface. Reception of forwarded directed broadcasts would
629 * be handled via ip_forward() and ether_output() with the loopback
630 * into the stack for SIMPLEX interfaces handled by ether_output().
632 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
634 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
635 if (ifa->ifa_addr->sa_family != AF_INET)
638 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
643 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) {
648 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
656 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
657 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
658 IPSTAT_INC(ips_cantforward);
662 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
665 * If we are acting as a multicast router, all
666 * incoming multicast packets are passed to the
667 * kernel-level multicast forwarding function.
668 * The packet is returned (relatively) intact; if
669 * ip_mforward() returns a non-zero value, the packet
670 * must be discarded, else it may be accepted below.
672 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
673 IPSTAT_INC(ips_cantforward);
679 * The process-level routing daemon needs to receive
680 * all multicast IGMP packets, whether or not this
681 * host belongs to their destination groups.
683 if (ip->ip_p == IPPROTO_IGMP)
685 IPSTAT_INC(ips_forward);
688 * Assume the packet is for us, to avoid prematurely taking
689 * a lock on the in_multi hash. Protocols must perform
690 * their own filtering and update statistics accordingly.
694 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
696 if (ip->ip_dst.s_addr == INADDR_ANY)
700 * FAITH(Firewall Aided Internet Translator)
702 if (ifp && ifp->if_type == IFT_FAITH) {
703 if (V_ip_keepfaith) {
704 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
712 * Not for us; forward if possible and desirable.
714 if (V_ipforwarding == 0) {
715 IPSTAT_INC(ips_cantforward);
729 * IPSTEALTH: Process non-routing options only
730 * if the packet is destined for us.
732 if (V_ipstealth && hlen > sizeof (struct ip) &&
735 #endif /* IPSTEALTH */
737 /* Count the packet in the ip address stats */
739 ia->ia_ifa.if_ipackets++;
740 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
744 * Attempt reassembly; if it succeeds, proceed.
745 * ip_reass() will return a different mbuf.
747 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
751 ip = mtod(m, struct ip *);
752 /* Get the header length of the reassembled packet */
753 hlen = ip->ip_hl << 2;
757 * Further protocols expect the packet length to be w/o the
764 * enforce IPsec policy checking if we are seeing last header.
765 * note that we do not visit this with protocols with pcb layer
766 * code - like udp/tcp/raw ip.
768 if (ip_ipsec_input(m))
773 * Switch out to protocol's input routine.
775 IPSTAT_INC(ips_delivered);
777 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
784 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
785 * max has slightly different semantics than the sysctl, for historical
791 INIT_VNET_INET(curvnet);
794 * -1 for unlimited allocation.
797 uma_zone_set_max(V_ipq_zone, 0);
799 * Positive number for specific bound.
802 uma_zone_set_max(V_ipq_zone, V_maxnipq);
804 * Zero specifies no further fragment queue allocation -- set the
805 * bound very low, but rely on implementation elsewhere to actually
806 * prevent allocation and reclaim current queues.
809 uma_zone_set_max(V_ipq_zone, 1);
813 ipq_zone_change(void *tag)
815 INIT_VNET_INET(curvnet);
817 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
818 V_maxnipq = nmbclusters / 32;
824 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
826 INIT_VNET_INET(curvnet);
830 error = sysctl_handle_int(oidp, &i, 0, req);
831 if (error || !req->newptr)
835 * XXXRW: Might be a good idea to sanity check the argument and place
836 * an extreme upper bound.
845 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
846 NULL, 0, sysctl_maxnipq, "I",
847 "Maximum number of IPv4 fragment reassembly queue entries");
850 * Take incoming datagram fragment and try to reassemble it into
851 * whole datagram. If the argument is the first fragment or one
852 * in between the function will return NULL and store the mbuf
853 * in the fragment chain. If the argument is the last fragment
854 * the packet will be reassembled and the pointer to the new
855 * mbuf returned for further processing. Only m_tags attached
856 * to the first packet/fragment are preserved.
857 * The IP header is *NOT* adjusted out of iplen.
860 ip_reass(struct mbuf *m)
862 INIT_VNET_INET(curvnet);
864 struct mbuf *p, *q, *nq, *t;
865 struct ipq *fp = NULL;
866 struct ipqhead *head;
871 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
872 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
873 IPSTAT_INC(ips_fragments);
874 IPSTAT_INC(ips_fragdropped);
879 ip = mtod(m, struct ip *);
880 hlen = ip->ip_hl << 2;
882 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
887 * Look for queue of fragments
890 TAILQ_FOREACH(fp, head, ipq_list)
891 if (ip->ip_id == fp->ipq_id &&
892 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
893 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
895 mac_ipq_match(m, fp) &&
897 ip->ip_p == fp->ipq_p)
903 * Attempt to trim the number of allocated fragment queues if it
904 * exceeds the administrative limit.
906 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
908 * drop something from the tail of the current queue
909 * before proceeding further
911 struct ipq *q = TAILQ_LAST(head, ipqhead);
912 if (q == NULL) { /* gak */
913 for (i = 0; i < IPREASS_NHASH; i++) {
914 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
916 IPSTAT_ADD(ips_fragtimeout,
918 ip_freef(&V_ipq[i], r);
923 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
930 * Adjust ip_len to not reflect header,
931 * convert offset of this to bytes.
934 if (ip->ip_off & IP_MF) {
936 * Make sure that fragments have a data length
937 * that's a non-zero multiple of 8 bytes.
939 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
940 IPSTAT_INC(ips_toosmall); /* XXX */
943 m->m_flags |= M_FRAG;
945 m->m_flags &= ~M_FRAG;
950 * Attempt reassembly; if it succeeds, proceed.
951 * ip_reass() will return a different mbuf.
953 IPSTAT_INC(ips_fragments);
954 m->m_pkthdr.header = ip;
956 /* Previous ip_reass() started here. */
958 * Presence of header sizes in mbufs
959 * would confuse code below.
965 * If first fragment to arrive, create a reassembly queue.
968 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
972 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
973 uma_zfree(V_ipq_zone, fp);
977 mac_ipq_create(m, fp);
979 TAILQ_INSERT_HEAD(head, fp, ipq_list);
982 fp->ipq_ttl = IPFRAGTTL;
983 fp->ipq_p = ip->ip_p;
984 fp->ipq_id = ip->ip_id;
985 fp->ipq_src = ip->ip_src;
986 fp->ipq_dst = ip->ip_dst;
993 mac_ipq_update(m, fp);
997 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1000 * Handle ECN by comparing this segment with the first one;
1001 * if CE is set, do not lose CE.
1002 * drop if CE and not-ECT are mixed for the same packet.
1004 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1005 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1006 if (ecn == IPTOS_ECN_CE) {
1007 if (ecn0 == IPTOS_ECN_NOTECT)
1009 if (ecn0 != IPTOS_ECN_CE)
1010 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1012 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1016 * Find a segment which begins after this one does.
1018 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1019 if (GETIP(q)->ip_off > ip->ip_off)
1023 * If there is a preceding segment, it may provide some of
1024 * our data already. If so, drop the data from the incoming
1025 * segment. If it provides all of our data, drop us, otherwise
1026 * stick new segment in the proper place.
1028 * If some of the data is dropped from the the preceding
1029 * segment, then it's checksum is invalidated.
1032 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1034 if (i >= ip->ip_len)
1037 m->m_pkthdr.csum_flags = 0;
1041 m->m_nextpkt = p->m_nextpkt;
1044 m->m_nextpkt = fp->ipq_frags;
1049 * While we overlap succeeding segments trim them or,
1050 * if they are completely covered, dequeue them.
1052 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1054 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1055 if (i < GETIP(q)->ip_len) {
1056 GETIP(q)->ip_len -= i;
1057 GETIP(q)->ip_off += i;
1059 q->m_pkthdr.csum_flags = 0;
1064 IPSTAT_INC(ips_fragdropped);
1070 * Check for complete reassembly and perform frag per packet
1073 * Frag limiting is performed here so that the nth frag has
1074 * a chance to complete the packet before we drop the packet.
1075 * As a result, n+1 frags are actually allowed per packet, but
1076 * only n will ever be stored. (n = maxfragsperpacket.)
1080 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1081 if (GETIP(q)->ip_off != next) {
1082 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1083 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1088 next += GETIP(q)->ip_len;
1090 /* Make sure the last packet didn't have the IP_MF flag */
1091 if (p->m_flags & M_FRAG) {
1092 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1093 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1100 * Reassembly is complete. Make sure the packet is a sane size.
1104 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1105 IPSTAT_INC(ips_toolong);
1106 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1112 * Concatenate fragments.
1119 q->m_nextpkt = NULL;
1120 for (q = nq; q != NULL; q = nq) {
1122 q->m_nextpkt = NULL;
1123 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1124 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1128 * In order to do checksumming faster we do 'end-around carry' here
1129 * (and not in for{} loop), though it implies we are not going to
1130 * reassemble more than 64k fragments.
1132 m->m_pkthdr.csum_data =
1133 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1135 mac_ipq_reassemble(fp, m);
1136 mac_ipq_destroy(fp);
1140 * Create header for new ip packet by modifying header of first
1141 * packet; dequeue and discard fragment reassembly header.
1142 * Make header visible.
1144 ip->ip_len = (ip->ip_hl << 2) + next;
1145 ip->ip_src = fp->ipq_src;
1146 ip->ip_dst = fp->ipq_dst;
1147 TAILQ_REMOVE(head, fp, ipq_list);
1149 uma_zfree(V_ipq_zone, fp);
1150 m->m_len += (ip->ip_hl << 2);
1151 m->m_data -= (ip->ip_hl << 2);
1152 /* some debugging cruft by sklower, below, will go away soon */
1153 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1155 IPSTAT_INC(ips_reassembled);
1160 IPSTAT_INC(ips_fragdropped);
1172 * Free a fragment reassembly header and all
1173 * associated datagrams.
1176 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1178 INIT_VNET_INET(curvnet);
1183 while (fp->ipq_frags) {
1185 fp->ipq_frags = q->m_nextpkt;
1188 TAILQ_REMOVE(fhp, fp, ipq_list);
1189 uma_zfree(V_ipq_zone, fp);
1194 * IP timer processing;
1195 * if a timer expires on a reassembly
1196 * queue, discard it.
1201 VNET_ITERATOR_DECL(vnet_iter);
1207 VNET_FOREACH(vnet_iter) {
1208 CURVNET_SET(vnet_iter);
1209 INIT_VNET_INET(vnet_iter);
1210 for (i = 0; i < IPREASS_NHASH; i++) {
1211 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1215 fp = TAILQ_NEXT(fp, ipq_list);
1216 if(--fpp->ipq_ttl == 0) {
1217 IPSTAT_ADD(ips_fragtimeout,
1219 ip_freef(&V_ipq[i], fpp);
1224 * If we are over the maximum number of fragments
1225 * (due to the limit being lowered), drain off
1226 * enough to get down to the new limit.
1228 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1229 for (i = 0; i < IPREASS_NHASH; i++) {
1230 while (V_nipq > V_maxnipq &&
1231 !TAILQ_EMPTY(&V_ipq[i])) {
1232 IPSTAT_ADD(ips_fragdropped,
1233 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1235 TAILQ_FIRST(&V_ipq[i]));
1241 VNET_LIST_RUNLOCK();
1246 * Drain off all datagram fragments.
1251 VNET_ITERATOR_DECL(vnet_iter);
1256 VNET_FOREACH(vnet_iter) {
1257 CURVNET_SET(vnet_iter);
1258 INIT_VNET_INET(vnet_iter);
1259 for (i = 0; i < IPREASS_NHASH; i++) {
1260 while(!TAILQ_EMPTY(&V_ipq[i])) {
1261 IPSTAT_ADD(ips_fragdropped,
1262 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1263 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1268 VNET_LIST_RUNLOCK();
1274 * The protocol to be inserted into ip_protox[] must be already registered
1275 * in inetsw[], either statically or through pf_proto_register().
1278 ipproto_register(u_char ipproto)
1282 /* Sanity checks. */
1284 return (EPROTONOSUPPORT);
1287 * The protocol slot must not be occupied by another protocol
1288 * already. An index pointing to IPPROTO_RAW is unused.
1290 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1292 return (EPFNOSUPPORT);
1293 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1296 /* Find the protocol position in inetsw[] and set the index. */
1297 for (pr = inetdomain.dom_protosw;
1298 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1299 if (pr->pr_domain->dom_family == PF_INET &&
1300 pr->pr_protocol && pr->pr_protocol == ipproto) {
1301 /* Be careful to only index valid IP protocols. */
1302 if (pr->pr_protocol < IPPROTO_MAX) {
1303 ip_protox[pr->pr_protocol] = pr - inetsw;
1309 return (EPROTONOSUPPORT);
1313 ipproto_unregister(u_char ipproto)
1317 /* Sanity checks. */
1319 return (EPROTONOSUPPORT);
1321 /* Check if the protocol was indeed registered. */
1322 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1324 return (EPFNOSUPPORT);
1325 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1328 /* Reset the protocol slot to IPPROTO_RAW. */
1329 ip_protox[ipproto] = pr - inetsw;
1334 * Given address of next destination (final or next hop),
1335 * return internet address info of interface to be used to get there.
1338 ip_rtaddr(struct in_addr dst, u_int fibnum)
1341 struct sockaddr_in *sin;
1342 struct in_ifaddr *ifa;
1344 bzero(&sro, sizeof(sro));
1345 sin = (struct sockaddr_in *)&sro.ro_dst;
1346 sin->sin_family = AF_INET;
1347 sin->sin_len = sizeof(*sin);
1348 sin->sin_addr = dst;
1349 in_rtalloc_ign(&sro, 0, fibnum);
1351 if (sro.ro_rt == NULL)
1354 ifa = ifatoia(sro.ro_rt->rt_ifa);
1359 u_char inetctlerrmap[PRC_NCMDS] = {
1361 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1362 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1363 EMSGSIZE, EHOSTUNREACH, 0, 0,
1364 0, 0, EHOSTUNREACH, 0,
1365 ENOPROTOOPT, ECONNREFUSED
1369 * Forward a packet. If some error occurs return the sender
1370 * an icmp packet. Note we can't always generate a meaningful
1371 * icmp message because icmp doesn't have a large enough repertoire
1372 * of codes and types.
1374 * If not forwarding, just drop the packet. This could be confusing
1375 * if ipforwarding was zero but some routing protocol was advancing
1376 * us as a gateway to somewhere. However, we must let the routing
1377 * protocol deal with that.
1379 * The srcrt parameter indicates whether the packet is being forwarded
1380 * via a source route.
1383 ip_forward(struct mbuf *m, int srcrt)
1385 INIT_VNET_INET(curvnet);
1386 struct ip *ip = mtod(m, struct ip *);
1387 struct in_ifaddr *ia;
1389 struct in_addr dest;
1391 int error, type = 0, code = 0, mtu = 0;
1393 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1394 IPSTAT_INC(ips_cantforward);
1401 if (ip->ip_ttl <= IPTTLDEC) {
1402 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1410 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1413 * 'ia' may be NULL if there is no route for this destination.
1414 * In case of IPsec, Don't discard it just yet, but pass it to
1415 * ip_output in case of outgoing IPsec policy.
1417 if (!srcrt && ia == NULL) {
1418 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1424 * Save the IP header and at most 8 bytes of the payload,
1425 * in case we need to generate an ICMP message to the src.
1427 * XXX this can be optimized a lot by saving the data in a local
1428 * buffer on the stack (72 bytes at most), and only allocating the
1429 * mbuf if really necessary. The vast majority of the packets
1430 * are forwarded without having to send an ICMP back (either
1431 * because unnecessary, or because rate limited), so we are
1432 * really we are wasting a lot of work here.
1434 * We don't use m_copy() because it might return a reference
1435 * to a shared cluster. Both this function and ip_output()
1436 * assume exclusive access to the IP header in `m', so any
1437 * data in a cluster may change before we reach icmp_error().
1439 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1440 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1442 * It's probably ok if the pkthdr dup fails (because
1443 * the deep copy of the tag chain failed), but for now
1444 * be conservative and just discard the copy since
1445 * code below may some day want the tags.
1450 if (mcopy != NULL) {
1451 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1452 mcopy->m_pkthdr.len = mcopy->m_len;
1453 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1459 ip->ip_ttl -= IPTTLDEC;
1465 * If forwarding packet using same interface that it came in on,
1466 * perhaps should send a redirect to sender to shortcut a hop.
1467 * Only send redirect if source is sending directly to us,
1468 * and if packet was not source routed (or has any options).
1469 * Also, don't send redirect if forwarding using a default route
1470 * or a route modified by a redirect.
1473 if (!srcrt && V_ipsendredirects &&
1474 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1475 struct sockaddr_in *sin;
1478 bzero(&ro, sizeof(ro));
1479 sin = (struct sockaddr_in *)&ro.ro_dst;
1480 sin->sin_family = AF_INET;
1481 sin->sin_len = sizeof(*sin);
1482 sin->sin_addr = ip->ip_dst;
1483 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1487 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1488 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1489 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1490 u_long src = ntohl(ip->ip_src.s_addr);
1493 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1494 if (rt->rt_flags & RTF_GATEWAY)
1495 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1497 dest.s_addr = ip->ip_dst.s_addr;
1498 /* Router requirements says to only send host redirects */
1499 type = ICMP_REDIRECT;
1500 code = ICMP_REDIRECT_HOST;
1508 * Try to cache the route MTU from ip_output so we can consider it for
1509 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1511 bzero(&ro, sizeof(ro));
1513 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1515 if (error == EMSGSIZE && ro.ro_rt)
1516 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1521 IPSTAT_INC(ips_cantforward);
1523 IPSTAT_INC(ips_forward);
1525 IPSTAT_INC(ips_redirectsent);
1537 case 0: /* forwarded, but need redirect */
1538 /* type, code set above */
1546 type = ICMP_UNREACH;
1547 code = ICMP_UNREACH_HOST;
1551 type = ICMP_UNREACH;
1552 code = ICMP_UNREACH_NEEDFRAG;
1556 * If IPsec is configured for this path,
1557 * override any possibly mtu value set by ip_output.
1559 mtu = ip_ipsec_mtu(m, mtu);
1562 * If the MTU was set before make sure we are below the
1564 * If the MTU wasn't set before use the interface mtu or
1565 * fall back to the next smaller mtu step compared to the
1566 * current packet size.
1570 mtu = min(mtu, ia->ia_ifp->if_mtu);
1573 mtu = ia->ia_ifp->if_mtu;
1575 mtu = ip_next_mtu(ip->ip_len, 0);
1577 IPSTAT_INC(ips_cantfrag);
1582 * A router should not generate ICMP_SOURCEQUENCH as
1583 * required in RFC1812 Requirements for IP Version 4 Routers.
1584 * Source quench could be a big problem under DoS attacks,
1585 * or if the underlying interface is rate-limited.
1586 * Those who need source quench packets may re-enable them
1587 * via the net.inet.ip.sendsourcequench sysctl.
1589 if (V_ip_sendsourcequench == 0) {
1593 type = ICMP_SOURCEQUENCH;
1598 case EACCES: /* ipfw denied packet */
1602 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1606 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1609 INIT_VNET_NET(inp->inp_vnet);
1611 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1615 if (inp->inp_socket->so_options & SO_BINTIME) {
1616 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1617 SCM_BINTIME, SOL_SOCKET);
1619 mp = &(*mp)->m_next;
1621 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1624 bintime2timeval(&bt, &tv);
1625 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1626 SCM_TIMESTAMP, SOL_SOCKET);
1628 mp = &(*mp)->m_next;
1631 if (inp->inp_flags & INP_RECVDSTADDR) {
1632 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1633 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1635 mp = &(*mp)->m_next;
1637 if (inp->inp_flags & INP_RECVTTL) {
1638 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1639 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1641 mp = &(*mp)->m_next;
1645 * Moving these out of udp_input() made them even more broken
1646 * than they already were.
1648 /* options were tossed already */
1649 if (inp->inp_flags & INP_RECVOPTS) {
1650 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1651 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1653 mp = &(*mp)->m_next;
1655 /* ip_srcroute doesn't do what we want here, need to fix */
1656 if (inp->inp_flags & INP_RECVRETOPTS) {
1657 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1658 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1660 mp = &(*mp)->m_next;
1663 if (inp->inp_flags & INP_RECVIF) {
1666 struct sockaddr_dl sdl;
1669 struct sockaddr_dl *sdp;
1670 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1672 if (((ifp = m->m_pkthdr.rcvif))
1673 && ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1674 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1676 * Change our mind and don't try copy.
1678 if ((sdp->sdl_family != AF_LINK)
1679 || (sdp->sdl_len > sizeof(sdlbuf))) {
1682 bcopy(sdp, sdl2, sdp->sdl_len);
1686 = offsetof(struct sockaddr_dl, sdl_data[0]);
1687 sdl2->sdl_family = AF_LINK;
1688 sdl2->sdl_index = 0;
1689 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1691 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1692 IP_RECVIF, IPPROTO_IP);
1694 mp = &(*mp)->m_next;
1699 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1700 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1701 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1705 ip_rsvp_init(struct socket *so)
1707 INIT_VNET_INET(so->so_vnet);
1709 if (so->so_type != SOCK_RAW ||
1710 so->so_proto->pr_protocol != IPPROTO_RSVP)
1713 if (V_ip_rsvpd != NULL)
1718 * This may seem silly, but we need to be sure we don't over-increment
1719 * the RSVP counter, in case something slips up.
1721 if (!V_ip_rsvp_on) {
1732 INIT_VNET_INET(curvnet);
1736 * This may seem silly, but we need to be sure we don't over-decrement
1737 * the RSVP counter, in case something slips up.
1747 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1749 INIT_VNET_INET(curvnet);
1751 if (rsvp_input_p) { /* call the real one if loaded */
1752 rsvp_input_p(m, off);
1756 /* Can still get packets with rsvp_on = 0 if there is a local member
1757 * of the group to which the RSVP packet is addressed. But in this
1758 * case we want to throw the packet away.
1766 if (V_ip_rsvpd != NULL) {
1770 /* Drop the packet */