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"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/callout.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
52 #include <sys/kernel.h>
54 #include <sys/rwlock.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
57 #include <sys/vimage.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/route.h>
65 #include <net/netisr.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 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
88 #include <netinet/ip_fw.h>
89 #include <netinet/ip_dummynet.h>
91 #include <security/mac/mac_framework.h>
94 CTASSERT(sizeof(struct ip) == 20);
98 #ifndef VIMAGE_GLOBALS
99 struct vnet_inet vnet_inet_0;
103 #ifdef VIMAGE_GLOBALS
104 static int ipsendredirects;
105 static int ip_checkinterface;
106 static int ip_keepfaith;
107 static int ip_sendsourcequench;
111 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
112 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
113 u_long in_ifaddrhmask; /* mask for hash table */
114 struct ipstat ipstat;
115 static int ip_rsvp_on;
116 struct socket *ip_rsvpd;
118 static struct ipqhead ipq[IPREASS_NHASH];
119 static int maxnipq; /* Administrative limit on # reass queues. */
120 static int maxfragsperpacket;
122 static int nipq; /* Total # of reass queues */
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
126 forwarding, CTLFLAG_RW, ipforwarding, 0,
127 "Enable IP forwarding between interfaces");
129 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
130 redirect, CTLFLAG_RW, ipsendredirects, 0,
131 "Enable sending IP redirects");
133 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
134 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
136 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
137 keepfaith, CTLFLAG_RW, ip_keepfaith, 0,
138 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
140 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
141 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
142 "Enable the transmission of source quench packets");
144 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
145 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values");
148 * XXX - Setting ip_checkinterface mostly implements the receive side of
149 * the Strong ES model described in RFC 1122, but since the routing table
150 * and transmit implementation do not implement the Strong ES model,
151 * setting this to 1 results in an odd hybrid.
153 * XXX - ip_checkinterface currently must be disabled if you use ipnat
154 * to translate the destination address to another local interface.
156 * XXX - ip_checkinterface must be disabled if you add IP aliases
157 * to the loopback interface instead of the interface where the
158 * packets for those addresses are received.
160 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
161 check_interface, CTLFLAG_RW, ip_checkinterface, 0,
162 "Verify packet arrives on correct interface");
164 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
166 static struct ifqueue ipintrq;
167 static int ipqmaxlen = IFQ_MAXLEN;
169 extern struct domain inetdomain;
170 extern struct protosw inetsw[];
171 u_char ip_protox[IPPROTO_MAX];
173 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
174 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
175 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
176 &ipintrq.ifq_drops, 0,
177 "Number of packets dropped from the IP input queue");
179 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
180 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
182 #ifdef VIMAGE_GLOBALS
183 static uma_zone_t ipq_zone;
185 static struct mtx ipqlock;
187 #define IPQ_LOCK() mtx_lock(&ipqlock)
188 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
189 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
190 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
192 static void maxnipq_update(void);
193 static void ipq_zone_change(void *);
195 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
197 "Current number of IPv4 fragment reassembly queue entries");
199 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
200 CTLFLAG_RW, maxfragsperpacket, 0,
201 "Maximum number of IPv4 fragments allowed per packet");
203 struct callout ipport_tick_callout;
206 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
207 &ip_mtu, 0, "Default MTU");
211 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
212 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
216 * ipfw_ether and ipfw_bridge hooks.
217 * XXX: Temporary until those are converted to pfil_hooks as well.
219 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
220 ip_dn_io_t *ip_dn_io_ptr = NULL;
221 #ifdef VIMAGE_GLOBALS
225 static void ip_freef(struct ipqhead *, struct ipq *);
228 * IP initialization: fill in IP protocol switch table.
229 * All protocols not implemented in kernel go to raw IP protocol handler.
234 INIT_VNET_INET(curvnet);
238 V_ipsendredirects = 1; /* XXX */
239 V_ip_checkinterface = 0;
241 V_ip_sendsourcequench = 0;
243 V_ip_defttl = IPDEFTTL;
244 V_ip_do_randomid = 0;
247 V_nipq = 0; /* Total # of reass queues */
249 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
250 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
251 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */
252 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */
253 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
254 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
255 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
256 V_ipport_reservedlow = 0;
257 V_ipport_randomized = 1; /* user controlled via sysctl */
258 V_ipport_randomcps = 10; /* user controlled via sysctl */
259 V_ipport_randomtime = 45; /* user controlled via sysctl */
260 V_ipport_stoprandom = 0; /* toggled by ipport_tick */
265 /* XXX global static but not instantiated in this file */
266 V_ipfastforward_active = 0;
267 V_subnetsarelocal = 0;
268 V_sameprefixcarponly = 0;
271 TAILQ_INIT(&V_in_ifaddrhead);
272 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
273 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
275 panic("ip_init: PF_INET not found");
277 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
278 for (i = 0; i < IPPROTO_MAX; i++)
279 ip_protox[i] = pr - inetsw;
281 * Cycle through IP protocols and put them into the appropriate place
284 for (pr = inetdomain.dom_protosw;
285 pr < inetdomain.dom_protoswNPROTOSW; pr++)
286 if (pr->pr_domain->dom_family == PF_INET &&
287 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
288 /* Be careful to only index valid IP protocols. */
289 if (pr->pr_protocol < IPPROTO_MAX)
290 ip_protox[pr->pr_protocol] = pr - inetsw;
293 /* Initialize packet filter hooks. */
294 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
295 inet_pfil_hook.ph_af = AF_INET;
296 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
297 printf("%s: WARNING: unable to register pfil hook, "
298 "error %d\n", __func__, i);
300 /* Initialize IP reassembly queue. */
302 for (i = 0; i < IPREASS_NHASH; i++)
303 TAILQ_INIT(&V_ipq[i]);
304 V_maxnipq = nmbclusters / 32;
305 V_maxfragsperpacket = 16;
306 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
307 NULL, UMA_ALIGN_PTR, 0);
310 /* Start ipport_tick. */
311 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
313 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
314 SHUTDOWN_PRI_DEFAULT);
315 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
316 NULL, EVENTHANDLER_PRI_ANY);
318 /* Initialize various other remaining things. */
319 V_ip_id = time_second & 0xffff;
320 ipintrq.ifq_maxlen = ipqmaxlen;
321 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
322 netisr_register(NETISR_IP, ip_input, &ipintrq, 0);
329 callout_stop(&ipport_tick_callout);
333 * Ip input routine. Checksum and byte swap header. If fragmented
334 * try to reassemble. Process options. Pass to next level.
337 ip_input(struct mbuf *m)
339 INIT_VNET_INET(curvnet);
340 struct ip *ip = NULL;
341 struct in_ifaddr *ia = NULL;
343 int checkif, hlen = 0;
345 int dchg = 0; /* dest changed after fw */
346 struct in_addr odst; /* original dst address */
350 if (m->m_flags & M_FASTFWD_OURS) {
352 * Firewall or NAT changed destination to local.
353 * We expect ip_len and ip_off to be in host byte order.
355 m->m_flags &= ~M_FASTFWD_OURS;
356 /* Set up some basics that will be used later. */
357 ip = mtod(m, struct ip *);
358 hlen = ip->ip_hl << 2;
362 V_ipstat.ips_total++;
364 if (m->m_pkthdr.len < sizeof(struct ip))
367 if (m->m_len < sizeof (struct ip) &&
368 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
369 V_ipstat.ips_toosmall++;
372 ip = mtod(m, struct ip *);
374 if (ip->ip_v != IPVERSION) {
375 V_ipstat.ips_badvers++;
379 hlen = ip->ip_hl << 2;
380 if (hlen < sizeof(struct ip)) { /* minimum header length */
381 V_ipstat.ips_badhlen++;
384 if (hlen > m->m_len) {
385 if ((m = m_pullup(m, hlen)) == NULL) {
386 V_ipstat.ips_badhlen++;
389 ip = mtod(m, struct ip *);
392 /* 127/8 must not appear on wire - RFC1122 */
393 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
394 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
395 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
396 V_ipstat.ips_badaddr++;
401 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
402 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
404 if (hlen == sizeof(struct ip)) {
405 sum = in_cksum_hdr(ip);
407 sum = in_cksum(m, hlen);
411 V_ipstat.ips_badsum++;
416 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
417 /* packet is dropped by traffic conditioner */
422 * Convert fields to host representation.
424 ip->ip_len = ntohs(ip->ip_len);
425 if (ip->ip_len < hlen) {
426 V_ipstat.ips_badlen++;
429 ip->ip_off = ntohs(ip->ip_off);
432 * Check that the amount of data in the buffers
433 * is as at least much as the IP header would have us expect.
434 * Trim mbufs if longer than we expect.
435 * Drop packet if shorter than we expect.
437 if (m->m_pkthdr.len < ip->ip_len) {
439 V_ipstat.ips_tooshort++;
442 if (m->m_pkthdr.len > ip->ip_len) {
443 if (m->m_len == m->m_pkthdr.len) {
444 m->m_len = ip->ip_len;
445 m->m_pkthdr.len = ip->ip_len;
447 m_adj(m, ip->ip_len - m->m_pkthdr.len);
451 * Bypass packet filtering for packets from a tunnel (gif).
453 if (ip_ipsec_filtertunnel(m))
458 * Run through list of hooks for input packets.
460 * NB: Beware of the destination address changing (e.g.
461 * by NAT rewriting). When this happens, tell
462 * ip_forward to do the right thing.
465 /* Jump over all PFIL processing if hooks are not active. */
466 if (!PFIL_HOOKED(&inet_pfil_hook))
470 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
473 if (m == NULL) /* consumed by filter */
476 ip = mtod(m, struct ip *);
477 dchg = (odst.s_addr != ip->ip_dst.s_addr);
479 #ifdef IPFIREWALL_FORWARD
480 if (m->m_flags & M_FASTFWD_OURS) {
481 m->m_flags &= ~M_FASTFWD_OURS;
484 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
486 * Directly ship on the packet. This allows to forward packets
487 * that were destined for us to some other directly connected
493 #endif /* IPFIREWALL_FORWARD */
497 * Process options and, if not destined for us,
498 * ship it on. ip_dooptions returns 1 when an
499 * error was detected (causing an icmp message
500 * to be sent and the original packet to be freed).
502 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
505 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
506 * matter if it is destined to another node, or whether it is
507 * a multicast one, RSVP wants it! and prevents it from being forwarded
508 * anywhere else. Also checks if the rsvp daemon is running before
509 * grabbing the packet.
511 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
515 * Check our list of addresses, to see if the packet is for us.
516 * If we don't have any addresses, assume any unicast packet
517 * we receive might be for us (and let the upper layers deal
520 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
521 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
525 * Enable a consistency check between the destination address
526 * and the arrival interface for a unicast packet (the RFC 1122
527 * strong ES model) if IP forwarding is disabled and the packet
528 * is not locally generated and the packet is not subject to
531 * XXX - Checking also should be disabled if the destination
532 * address is ipnat'ed to a different interface.
534 * XXX - Checking is incompatible with IP aliases added
535 * to the loopback interface instead of the interface where
536 * the packets are received.
538 * XXX - This is the case for carp vhost IPs as well so we
539 * insert a workaround. If the packet got here, we already
540 * checked with carp_iamatch() and carp_forus().
542 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
543 m->m_pkthdr.rcvif != NULL &&
544 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
546 !m->m_pkthdr.rcvif->if_carp &&
551 * Check for exact addresses in the hash bucket.
553 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
555 * If the address matches, verify that the packet
556 * arrived via the correct interface if checking is
559 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
560 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
564 * Check for broadcast addresses.
566 * Only accept broadcast packets that arrive via the matching
567 * interface. Reception of forwarded directed broadcasts would
568 * be handled via ip_forward() and ether_output() with the loopback
569 * into the stack for SIMPLEX interfaces handled by ether_output().
571 if (m->m_pkthdr.rcvif != NULL &&
572 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
573 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
574 if (ifa->ifa_addr->sa_family != AF_INET)
577 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
580 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
583 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
588 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
589 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
590 V_ipstat.ips_cantforward++;
594 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
597 * If we are acting as a multicast router, all
598 * incoming multicast packets are passed to the
599 * kernel-level multicast forwarding function.
600 * The packet is returned (relatively) intact; if
601 * ip_mforward() returns a non-zero value, the packet
602 * must be discarded, else it may be accepted below.
605 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
606 V_ipstat.ips_cantforward++;
612 * The process-level routing daemon needs to receive
613 * all multicast IGMP packets, whether or not this
614 * host belongs to their destination groups.
616 if (ip->ip_p == IPPROTO_IGMP)
618 V_ipstat.ips_forward++;
621 * Assume the packet is for us, to avoid prematurely taking
622 * a lock on the in_multi hash. Protocols must perform
623 * their own filtering and update statistics accordingly.
627 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
629 if (ip->ip_dst.s_addr == INADDR_ANY)
633 * FAITH(Firewall Aided Internet Translator)
635 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
636 if (V_ip_keepfaith) {
637 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
645 * Not for us; forward if possible and desirable.
647 if (V_ipforwarding == 0) {
648 V_ipstat.ips_cantforward++;
662 * IPSTEALTH: Process non-routing options only
663 * if the packet is destined for us.
665 if (V_ipstealth && hlen > sizeof (struct ip) &&
668 #endif /* IPSTEALTH */
670 /* Count the packet in the ip address stats */
672 ia->ia_ifa.if_ipackets++;
673 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
677 * Attempt reassembly; if it succeeds, proceed.
678 * ip_reass() will return a different mbuf.
680 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
684 ip = mtod(m, struct ip *);
685 /* Get the header length of the reassembled packet */
686 hlen = ip->ip_hl << 2;
690 * Further protocols expect the packet length to be w/o the
697 * enforce IPsec policy checking if we are seeing last header.
698 * note that we do not visit this with protocols with pcb layer
699 * code - like udp/tcp/raw ip.
701 if (ip_ipsec_input(m))
706 * Switch out to protocol's input routine.
708 V_ipstat.ips_delivered++;
710 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
717 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
718 * max has slightly different semantics than the sysctl, for historical
724 INIT_VNET_INET(curvnet);
727 * -1 for unlimited allocation.
730 uma_zone_set_max(V_ipq_zone, 0);
732 * Positive number for specific bound.
735 uma_zone_set_max(V_ipq_zone, V_maxnipq);
737 * Zero specifies no further fragment queue allocation -- set the
738 * bound very low, but rely on implementation elsewhere to actually
739 * prevent allocation and reclaim current queues.
742 uma_zone_set_max(V_ipq_zone, 1);
746 ipq_zone_change(void *tag)
748 INIT_VNET_INET(curvnet);
750 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
751 V_maxnipq = nmbclusters / 32;
757 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
759 INIT_VNET_INET(curvnet);
763 error = sysctl_handle_int(oidp, &i, 0, req);
764 if (error || !req->newptr)
768 * XXXRW: Might be a good idea to sanity check the argument and place
769 * an extreme upper bound.
778 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
779 NULL, 0, sysctl_maxnipq, "I",
780 "Maximum number of IPv4 fragment reassembly queue entries");
783 * Take incoming datagram fragment and try to reassemble it into
784 * whole datagram. If the argument is the first fragment or one
785 * in between the function will return NULL and store the mbuf
786 * in the fragment chain. If the argument is the last fragment
787 * the packet will be reassembled and the pointer to the new
788 * mbuf returned for further processing. Only m_tags attached
789 * to the first packet/fragment are preserved.
790 * The IP header is *NOT* adjusted out of iplen.
793 ip_reass(struct mbuf *m)
795 INIT_VNET_INET(curvnet);
797 struct mbuf *p, *q, *nq, *t;
798 struct ipq *fp = NULL;
799 struct ipqhead *head;
804 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
805 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
806 V_ipstat.ips_fragments++;
807 V_ipstat.ips_fragdropped++;
812 ip = mtod(m, struct ip *);
813 hlen = ip->ip_hl << 2;
815 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
820 * Look for queue of fragments
823 TAILQ_FOREACH(fp, head, ipq_list)
824 if (ip->ip_id == fp->ipq_id &&
825 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
826 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
828 mac_ipq_match(m, fp) &&
830 ip->ip_p == fp->ipq_p)
836 * Attempt to trim the number of allocated fragment queues if it
837 * exceeds the administrative limit.
839 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
841 * drop something from the tail of the current queue
842 * before proceeding further
844 struct ipq *q = TAILQ_LAST(head, ipqhead);
845 if (q == NULL) { /* gak */
846 for (i = 0; i < IPREASS_NHASH; i++) {
847 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
849 V_ipstat.ips_fragtimeout +=
851 ip_freef(&V_ipq[i], r);
856 V_ipstat.ips_fragtimeout += q->ipq_nfrags;
863 * Adjust ip_len to not reflect header,
864 * convert offset of this to bytes.
867 if (ip->ip_off & IP_MF) {
869 * Make sure that fragments have a data length
870 * that's a non-zero multiple of 8 bytes.
872 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
873 V_ipstat.ips_toosmall++; /* XXX */
876 m->m_flags |= M_FRAG;
878 m->m_flags &= ~M_FRAG;
883 * Attempt reassembly; if it succeeds, proceed.
884 * ip_reass() will return a different mbuf.
886 V_ipstat.ips_fragments++;
887 m->m_pkthdr.header = ip;
889 /* Previous ip_reass() started here. */
891 * Presence of header sizes in mbufs
892 * would confuse code below.
898 * If first fragment to arrive, create a reassembly queue.
901 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
905 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
906 uma_zfree(V_ipq_zone, fp);
910 mac_ipq_create(m, fp);
912 TAILQ_INSERT_HEAD(head, fp, ipq_list);
915 fp->ipq_ttl = IPFRAGTTL;
916 fp->ipq_p = ip->ip_p;
917 fp->ipq_id = ip->ip_id;
918 fp->ipq_src = ip->ip_src;
919 fp->ipq_dst = ip->ip_dst;
926 mac_ipq_update(m, fp);
930 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
933 * Handle ECN by comparing this segment with the first one;
934 * if CE is set, do not lose CE.
935 * drop if CE and not-ECT are mixed for the same packet.
937 ecn = ip->ip_tos & IPTOS_ECN_MASK;
938 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
939 if (ecn == IPTOS_ECN_CE) {
940 if (ecn0 == IPTOS_ECN_NOTECT)
942 if (ecn0 != IPTOS_ECN_CE)
943 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
945 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
949 * Find a segment which begins after this one does.
951 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
952 if (GETIP(q)->ip_off > ip->ip_off)
956 * If there is a preceding segment, it may provide some of
957 * our data already. If so, drop the data from the incoming
958 * segment. If it provides all of our data, drop us, otherwise
959 * stick new segment in the proper place.
961 * If some of the data is dropped from the the preceding
962 * segment, then it's checksum is invalidated.
965 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
970 m->m_pkthdr.csum_flags = 0;
974 m->m_nextpkt = p->m_nextpkt;
977 m->m_nextpkt = fp->ipq_frags;
982 * While we overlap succeeding segments trim them or,
983 * if they are completely covered, dequeue them.
985 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
987 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
988 if (i < GETIP(q)->ip_len) {
989 GETIP(q)->ip_len -= i;
990 GETIP(q)->ip_off += i;
992 q->m_pkthdr.csum_flags = 0;
997 V_ipstat.ips_fragdropped++;
1003 * Check for complete reassembly and perform frag per packet
1006 * Frag limiting is performed here so that the nth frag has
1007 * a chance to complete the packet before we drop the packet.
1008 * As a result, n+1 frags are actually allowed per packet, but
1009 * only n will ever be stored. (n = maxfragsperpacket.)
1013 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1014 if (GETIP(q)->ip_off != next) {
1015 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1016 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1021 next += GETIP(q)->ip_len;
1023 /* Make sure the last packet didn't have the IP_MF flag */
1024 if (p->m_flags & M_FRAG) {
1025 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1026 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1033 * Reassembly is complete. Make sure the packet is a sane size.
1037 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1038 V_ipstat.ips_toolong++;
1039 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1045 * Concatenate fragments.
1052 q->m_nextpkt = NULL;
1053 for (q = nq; q != NULL; q = nq) {
1055 q->m_nextpkt = NULL;
1056 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1057 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1061 * In order to do checksumming faster we do 'end-around carry' here
1062 * (and not in for{} loop), though it implies we are not going to
1063 * reassemble more than 64k fragments.
1065 m->m_pkthdr.csum_data =
1066 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1068 mac_ipq_reassemble(fp, m);
1069 mac_ipq_destroy(fp);
1073 * Create header for new ip packet by modifying header of first
1074 * packet; dequeue and discard fragment reassembly header.
1075 * Make header visible.
1077 ip->ip_len = (ip->ip_hl << 2) + next;
1078 ip->ip_src = fp->ipq_src;
1079 ip->ip_dst = fp->ipq_dst;
1080 TAILQ_REMOVE(head, fp, ipq_list);
1082 uma_zfree(V_ipq_zone, fp);
1083 m->m_len += (ip->ip_hl << 2);
1084 m->m_data -= (ip->ip_hl << 2);
1085 /* some debugging cruft by sklower, below, will go away soon */
1086 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1088 V_ipstat.ips_reassembled++;
1093 V_ipstat.ips_fragdropped++;
1105 * Free a fragment reassembly header and all
1106 * associated datagrams.
1109 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1111 INIT_VNET_INET(curvnet);
1116 while (fp->ipq_frags) {
1118 fp->ipq_frags = q->m_nextpkt;
1121 TAILQ_REMOVE(fhp, fp, ipq_list);
1122 uma_zfree(V_ipq_zone, fp);
1127 * IP timer processing;
1128 * if a timer expires on a reassembly
1129 * queue, discard it.
1134 VNET_ITERATOR_DECL(vnet_iter);
1140 VNET_FOREACH(vnet_iter) {
1141 CURVNET_SET(vnet_iter);
1142 INIT_VNET_INET(vnet_iter);
1143 for (i = 0; i < IPREASS_NHASH; i++) {
1144 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1148 fp = TAILQ_NEXT(fp, ipq_list);
1149 if(--fpp->ipq_ttl == 0) {
1150 V_ipstat.ips_fragtimeout +=
1152 ip_freef(&V_ipq[i], fpp);
1157 * If we are over the maximum number of fragments
1158 * (due to the limit being lowered), drain off
1159 * enough to get down to the new limit.
1161 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1162 for (i = 0; i < IPREASS_NHASH; i++) {
1163 while (V_nipq > V_maxnipq &&
1164 !TAILQ_EMPTY(&V_ipq[i])) {
1165 V_ipstat.ips_fragdropped +=
1166 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1168 TAILQ_FIRST(&V_ipq[i]));
1174 VNET_LIST_RUNLOCK();
1179 * Drain off all datagram fragments.
1184 VNET_ITERATOR_DECL(vnet_iter);
1189 VNET_FOREACH(vnet_iter) {
1190 CURVNET_SET(vnet_iter);
1191 INIT_VNET_INET(vnet_iter);
1192 for (i = 0; i < IPREASS_NHASH; i++) {
1193 while(!TAILQ_EMPTY(&V_ipq[i])) {
1194 V_ipstat.ips_fragdropped +=
1195 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1196 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1201 VNET_LIST_RUNLOCK();
1207 * The protocol to be inserted into ip_protox[] must be already registered
1208 * in inetsw[], either statically or through pf_proto_register().
1211 ipproto_register(u_char ipproto)
1215 /* Sanity checks. */
1217 return (EPROTONOSUPPORT);
1220 * The protocol slot must not be occupied by another protocol
1221 * already. An index pointing to IPPROTO_RAW is unused.
1223 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1225 return (EPFNOSUPPORT);
1226 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1229 /* Find the protocol position in inetsw[] and set the index. */
1230 for (pr = inetdomain.dom_protosw;
1231 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1232 if (pr->pr_domain->dom_family == PF_INET &&
1233 pr->pr_protocol && pr->pr_protocol == ipproto) {
1234 /* Be careful to only index valid IP protocols. */
1235 if (pr->pr_protocol < IPPROTO_MAX) {
1236 ip_protox[pr->pr_protocol] = pr - inetsw;
1242 return (EPROTONOSUPPORT);
1246 ipproto_unregister(u_char ipproto)
1250 /* Sanity checks. */
1252 return (EPROTONOSUPPORT);
1254 /* Check if the protocol was indeed registered. */
1255 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1257 return (EPFNOSUPPORT);
1258 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1261 /* Reset the protocol slot to IPPROTO_RAW. */
1262 ip_protox[ipproto] = pr - inetsw;
1267 * Given address of next destination (final or next hop),
1268 * return internet address info of interface to be used to get there.
1271 ip_rtaddr(struct in_addr dst, u_int fibnum)
1274 struct sockaddr_in *sin;
1275 struct in_ifaddr *ifa;
1277 bzero(&sro, sizeof(sro));
1278 sin = (struct sockaddr_in *)&sro.ro_dst;
1279 sin->sin_family = AF_INET;
1280 sin->sin_len = sizeof(*sin);
1281 sin->sin_addr = dst;
1282 in_rtalloc_ign(&sro, 0, fibnum);
1284 if (sro.ro_rt == NULL)
1287 ifa = ifatoia(sro.ro_rt->rt_ifa);
1292 u_char inetctlerrmap[PRC_NCMDS] = {
1294 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1295 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1296 EMSGSIZE, EHOSTUNREACH, 0, 0,
1297 0, 0, EHOSTUNREACH, 0,
1298 ENOPROTOOPT, ECONNREFUSED
1302 * Forward a packet. If some error occurs return the sender
1303 * an icmp packet. Note we can't always generate a meaningful
1304 * icmp message because icmp doesn't have a large enough repertoire
1305 * of codes and types.
1307 * If not forwarding, just drop the packet. This could be confusing
1308 * if ipforwarding was zero but some routing protocol was advancing
1309 * us as a gateway to somewhere. However, we must let the routing
1310 * protocol deal with that.
1312 * The srcrt parameter indicates whether the packet is being forwarded
1313 * via a source route.
1316 ip_forward(struct mbuf *m, int srcrt)
1318 INIT_VNET_INET(curvnet);
1319 struct ip *ip = mtod(m, struct ip *);
1320 struct in_ifaddr *ia = NULL;
1322 struct in_addr dest;
1324 int error, type = 0, code = 0, mtu = 0;
1326 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1327 V_ipstat.ips_cantforward++;
1334 if (ip->ip_ttl <= IPTTLDEC) {
1335 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1343 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1344 if (!srcrt && ia == NULL) {
1345 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1350 * Save the IP header and at most 8 bytes of the payload,
1351 * in case we need to generate an ICMP message to the src.
1353 * XXX this can be optimized a lot by saving the data in a local
1354 * buffer on the stack (72 bytes at most), and only allocating the
1355 * mbuf if really necessary. The vast majority of the packets
1356 * are forwarded without having to send an ICMP back (either
1357 * because unnecessary, or because rate limited), so we are
1358 * really we are wasting a lot of work here.
1360 * We don't use m_copy() because it might return a reference
1361 * to a shared cluster. Both this function and ip_output()
1362 * assume exclusive access to the IP header in `m', so any
1363 * data in a cluster may change before we reach icmp_error().
1365 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1366 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1368 * It's probably ok if the pkthdr dup fails (because
1369 * the deep copy of the tag chain failed), but for now
1370 * be conservative and just discard the copy since
1371 * code below may some day want the tags.
1376 if (mcopy != NULL) {
1377 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1378 mcopy->m_pkthdr.len = mcopy->m_len;
1379 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1385 ip->ip_ttl -= IPTTLDEC;
1391 * If forwarding packet using same interface that it came in on,
1392 * perhaps should send a redirect to sender to shortcut a hop.
1393 * Only send redirect if source is sending directly to us,
1394 * and if packet was not source routed (or has any options).
1395 * Also, don't send redirect if forwarding using a default route
1396 * or a route modified by a redirect.
1399 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1400 struct sockaddr_in *sin;
1403 bzero(&ro, sizeof(ro));
1404 sin = (struct sockaddr_in *)&ro.ro_dst;
1405 sin->sin_family = AF_INET;
1406 sin->sin_len = sizeof(*sin);
1407 sin->sin_addr = ip->ip_dst;
1408 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1412 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1413 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1414 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1415 u_long src = ntohl(ip->ip_src.s_addr);
1418 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1419 if (rt->rt_flags & RTF_GATEWAY)
1420 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1422 dest.s_addr = ip->ip_dst.s_addr;
1423 /* Router requirements says to only send host redirects */
1424 type = ICMP_REDIRECT;
1425 code = ICMP_REDIRECT_HOST;
1433 * Try to cache the route MTU from ip_output so we can consider it for
1434 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1436 bzero(&ro, sizeof(ro));
1438 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1440 if (error == EMSGSIZE && ro.ro_rt)
1441 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1446 V_ipstat.ips_cantforward++;
1448 V_ipstat.ips_forward++;
1450 V_ipstat.ips_redirectsent++;
1462 case 0: /* forwarded, but need redirect */
1463 /* type, code set above */
1466 case ENETUNREACH: /* shouldn't happen, checked above */
1471 type = ICMP_UNREACH;
1472 code = ICMP_UNREACH_HOST;
1476 type = ICMP_UNREACH;
1477 code = ICMP_UNREACH_NEEDFRAG;
1481 * If IPsec is configured for this path,
1482 * override any possibly mtu value set by ip_output.
1484 mtu = ip_ipsec_mtu(m, mtu);
1487 * If the MTU was set before make sure we are below the
1489 * If the MTU wasn't set before use the interface mtu or
1490 * fall back to the next smaller mtu step compared to the
1491 * current packet size.
1495 mtu = min(mtu, ia->ia_ifp->if_mtu);
1498 mtu = ia->ia_ifp->if_mtu;
1500 mtu = ip_next_mtu(ip->ip_len, 0);
1502 V_ipstat.ips_cantfrag++;
1507 * A router should not generate ICMP_SOURCEQUENCH as
1508 * required in RFC1812 Requirements for IP Version 4 Routers.
1509 * Source quench could be a big problem under DoS attacks,
1510 * or if the underlying interface is rate-limited.
1511 * Those who need source quench packets may re-enable them
1512 * via the net.inet.ip.sendsourcequench sysctl.
1514 if (V_ip_sendsourcequench == 0) {
1518 type = ICMP_SOURCEQUENCH;
1523 case EACCES: /* ipfw denied packet */
1527 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1531 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1534 INIT_VNET_NET(inp->inp_vnet);
1536 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1540 if (inp->inp_socket->so_options & SO_BINTIME) {
1541 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1542 SCM_BINTIME, SOL_SOCKET);
1544 mp = &(*mp)->m_next;
1546 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1549 bintime2timeval(&bt, &tv);
1550 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1551 SCM_TIMESTAMP, SOL_SOCKET);
1553 mp = &(*mp)->m_next;
1556 if (inp->inp_flags & INP_RECVDSTADDR) {
1557 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1558 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1560 mp = &(*mp)->m_next;
1562 if (inp->inp_flags & INP_RECVTTL) {
1563 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1564 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1566 mp = &(*mp)->m_next;
1570 * Moving these out of udp_input() made them even more broken
1571 * than they already were.
1573 /* options were tossed already */
1574 if (inp->inp_flags & INP_RECVOPTS) {
1575 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1576 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1578 mp = &(*mp)->m_next;
1580 /* ip_srcroute doesn't do what we want here, need to fix */
1581 if (inp->inp_flags & INP_RECVRETOPTS) {
1582 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1583 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1585 mp = &(*mp)->m_next;
1588 if (inp->inp_flags & INP_RECVIF) {
1591 struct sockaddr_dl sdl;
1594 struct sockaddr_dl *sdp;
1595 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1597 if (((ifp = m->m_pkthdr.rcvif))
1598 && ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1599 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1601 * Change our mind and don't try copy.
1603 if ((sdp->sdl_family != AF_LINK)
1604 || (sdp->sdl_len > sizeof(sdlbuf))) {
1607 bcopy(sdp, sdl2, sdp->sdl_len);
1611 = offsetof(struct sockaddr_dl, sdl_data[0]);
1612 sdl2->sdl_family = AF_LINK;
1613 sdl2->sdl_index = 0;
1614 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1616 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1617 IP_RECVIF, IPPROTO_IP);
1619 mp = &(*mp)->m_next;
1624 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1625 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1626 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1630 ip_rsvp_init(struct socket *so)
1632 INIT_VNET_INET(so->so_vnet);
1634 if (so->so_type != SOCK_RAW ||
1635 so->so_proto->pr_protocol != IPPROTO_RSVP)
1638 if (V_ip_rsvpd != NULL)
1643 * This may seem silly, but we need to be sure we don't over-increment
1644 * the RSVP counter, in case something slips up.
1646 if (!V_ip_rsvp_on) {
1657 INIT_VNET_INET(curvnet);
1661 * This may seem silly, but we need to be sure we don't over-decrement
1662 * the RSVP counter, in case something slips up.
1672 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1674 INIT_VNET_INET(curvnet);
1676 if (rsvp_input_p) { /* call the real one if loaded */
1677 rsvp_input_p(m, off);
1681 /* Can still get packets with rsvp_on = 0 if there is a local member
1682 * of the group to which the RSVP packet is addressed. But in this
1683 * case we want to throw the packet away.
1691 if (V_ip_rsvpd != NULL) {
1695 /* Drop the packet */