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"
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>
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_icmp.h>
74 #include <netinet/ip_options.h>
75 #include <machine/in_cksum.h>
76 #include <netinet/vinet.h>
78 #include <netinet/ip_carp.h>
81 #include <netinet/ip_ipsec.h>
84 #include <sys/socketvar.h>
86 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
87 #include <netinet/ip_fw.h>
88 #include <netinet/ip_dummynet.h>
90 #include <security/mac/mac_framework.h>
93 CTASSERT(sizeof(struct ip) == 20);
97 #ifndef VIMAGE_GLOBALS
98 struct vnet_inet vnet_inet_0;
102 #ifdef VIMAGE_GLOBALS
103 static int ipsendredirects;
104 static int ip_checkinterface;
105 static int ip_keepfaith;
106 static int ip_sendsourcequench;
110 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
111 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
112 u_long in_ifaddrhmask; /* mask for hash table */
113 struct ipstat ipstat;
114 static int ip_rsvp_on;
115 struct socket *ip_rsvpd;
117 static struct ipqhead ipq[IPREASS_NHASH];
118 static int maxnipq; /* Administrative limit on # reass queues. */
119 static int maxfragsperpacket;
121 static int nipq; /* Total # of reass queues */
124 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
125 forwarding, CTLFLAG_RW, ipforwarding, 0,
126 "Enable IP forwarding between interfaces");
128 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
129 redirect, CTLFLAG_RW, ipsendredirects, 0,
130 "Enable sending IP redirects");
132 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
133 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
135 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
136 keepfaith, CTLFLAG_RW, ip_keepfaith, 0,
137 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
139 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
140 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
141 "Enable the transmission of source quench packets");
143 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
144 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values");
147 * XXX - Setting ip_checkinterface mostly implements the receive side of
148 * the Strong ES model described in RFC 1122, but since the routing table
149 * and transmit implementation do not implement the Strong ES model,
150 * setting this to 1 results in an odd hybrid.
152 * XXX - ip_checkinterface currently must be disabled if you use ipnat
153 * to translate the destination address to another local interface.
155 * XXX - ip_checkinterface must be disabled if you add IP aliases
156 * to the loopback interface instead of the interface where the
157 * packets for those addresses are received.
159 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
160 check_interface, CTLFLAG_RW, ip_checkinterface, 0,
161 "Verify packet arrives on correct interface");
163 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
165 static struct ifqueue ipintrq;
166 static int ipqmaxlen = IFQ_MAXLEN;
168 extern struct domain inetdomain;
169 extern struct protosw inetsw[];
170 u_char ip_protox[IPPROTO_MAX];
172 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
173 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
174 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
175 &ipintrq.ifq_drops, 0,
176 "Number of packets dropped from the IP input queue");
178 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
179 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
181 #ifdef VIMAGE_GLOBALS
182 static uma_zone_t ipq_zone;
184 static struct mtx ipqlock;
186 #define IPQ_LOCK() mtx_lock(&ipqlock)
187 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
188 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
189 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
191 static void maxnipq_update(void);
192 static void ipq_zone_change(void *);
194 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
196 "Current number of IPv4 fragment reassembly queue entries");
198 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
199 CTLFLAG_RW, maxfragsperpacket, 0,
200 "Maximum number of IPv4 fragments allowed per packet");
202 struct callout ipport_tick_callout;
205 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
206 &ip_mtu, 0, "Default MTU");
210 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
211 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
215 * ipfw_ether and ipfw_bridge hooks.
216 * XXX: Temporary until those are converted to pfil_hooks as well.
218 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
219 ip_dn_io_t *ip_dn_io_ptr = NULL;
220 #ifdef VIMAGE_GLOBALS
224 static void ip_freef(struct ipqhead *, struct ipq *);
227 * IP initialization: fill in IP protocol switch table.
228 * All protocols not implemented in kernel go to raw IP protocol handler.
233 INIT_VNET_INET(curvnet);
237 V_ipsendredirects = 1; /* XXX */
238 V_ip_checkinterface = 0;
240 V_ip_sendsourcequench = 0;
242 V_ip_defttl = IPDEFTTL;
243 V_ip_do_randomid = 0;
246 V_nipq = 0; /* Total # of reass queues */
248 V_ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
249 V_ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
250 V_ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */
251 V_ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */
252 V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
253 V_ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
254 V_ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
255 V_ipport_reservedlow = 0;
256 V_ipport_randomized = 1; /* user controlled via sysctl */
257 V_ipport_randomcps = 10; /* user controlled via sysctl */
258 V_ipport_randomtime = 45; /* user controlled via sysctl */
259 V_ipport_stoprandom = 0; /* toggled by ipport_tick */
264 /* XXX global static but not instantiated in this file */
265 V_ipfastforward_active = 0;
266 V_subnetsarelocal = 0;
267 V_sameprefixcarponly = 0;
270 TAILQ_INIT(&V_in_ifaddrhead);
271 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
272 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
274 panic("ip_init: PF_INET not found");
276 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
277 for (i = 0; i < IPPROTO_MAX; i++)
278 ip_protox[i] = pr - inetsw;
280 * Cycle through IP protocols and put them into the appropriate place
283 for (pr = inetdomain.dom_protosw;
284 pr < inetdomain.dom_protoswNPROTOSW; pr++)
285 if (pr->pr_domain->dom_family == PF_INET &&
286 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
287 /* Be careful to only index valid IP protocols. */
288 if (pr->pr_protocol < IPPROTO_MAX)
289 ip_protox[pr->pr_protocol] = pr - inetsw;
292 /* Initialize packet filter hooks. */
293 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
294 inet_pfil_hook.ph_af = AF_INET;
295 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
296 printf("%s: WARNING: unable to register pfil hook, "
297 "error %d\n", __func__, i);
299 /* Initialize IP reassembly queue. */
301 for (i = 0; i < IPREASS_NHASH; i++)
302 TAILQ_INIT(&V_ipq[i]);
303 V_maxnipq = nmbclusters / 32;
304 V_maxfragsperpacket = 16;
305 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
306 NULL, UMA_ALIGN_PTR, 0);
309 /* Start ipport_tick. */
310 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
312 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
313 SHUTDOWN_PRI_DEFAULT);
314 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
315 NULL, EVENTHANDLER_PRI_ANY);
317 /* Initialize various other remaining things. */
318 V_ip_id = time_second & 0xffff;
319 ipintrq.ifq_maxlen = ipqmaxlen;
320 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
321 netisr_register(NETISR_IP, ip_input, &ipintrq, 0);
328 callout_stop(&ipport_tick_callout);
332 * Ip input routine. Checksum and byte swap header. If fragmented
333 * try to reassemble. Process options. Pass to next level.
336 ip_input(struct mbuf *m)
338 INIT_VNET_INET(curvnet);
339 struct ip *ip = NULL;
340 struct in_ifaddr *ia = NULL;
342 int checkif, hlen = 0;
344 int dchg = 0; /* dest changed after fw */
345 struct in_addr odst; /* original dst address */
349 if (m->m_flags & M_FASTFWD_OURS) {
351 * Firewall or NAT changed destination to local.
352 * We expect ip_len and ip_off to be in host byte order.
354 m->m_flags &= ~M_FASTFWD_OURS;
355 /* Set up some basics that will be used later. */
356 ip = mtod(m, struct ip *);
357 hlen = ip->ip_hl << 2;
361 V_ipstat.ips_total++;
363 if (m->m_pkthdr.len < sizeof(struct ip))
366 if (m->m_len < sizeof (struct ip) &&
367 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
368 V_ipstat.ips_toosmall++;
371 ip = mtod(m, struct ip *);
373 if (ip->ip_v != IPVERSION) {
374 V_ipstat.ips_badvers++;
378 hlen = ip->ip_hl << 2;
379 if (hlen < sizeof(struct ip)) { /* minimum header length */
380 V_ipstat.ips_badhlen++;
383 if (hlen > m->m_len) {
384 if ((m = m_pullup(m, hlen)) == NULL) {
385 V_ipstat.ips_badhlen++;
388 ip = mtod(m, struct ip *);
391 /* 127/8 must not appear on wire - RFC1122 */
392 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
393 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
394 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
395 V_ipstat.ips_badaddr++;
400 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
401 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
403 if (hlen == sizeof(struct ip)) {
404 sum = in_cksum_hdr(ip);
406 sum = in_cksum(m, hlen);
410 V_ipstat.ips_badsum++;
415 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
416 /* packet is dropped by traffic conditioner */
421 * Convert fields to host representation.
423 ip->ip_len = ntohs(ip->ip_len);
424 if (ip->ip_len < hlen) {
425 V_ipstat.ips_badlen++;
428 ip->ip_off = ntohs(ip->ip_off);
431 * Check that the amount of data in the buffers
432 * is as at least much as the IP header would have us expect.
433 * Trim mbufs if longer than we expect.
434 * Drop packet if shorter than we expect.
436 if (m->m_pkthdr.len < ip->ip_len) {
438 V_ipstat.ips_tooshort++;
441 if (m->m_pkthdr.len > ip->ip_len) {
442 if (m->m_len == m->m_pkthdr.len) {
443 m->m_len = ip->ip_len;
444 m->m_pkthdr.len = ip->ip_len;
446 m_adj(m, ip->ip_len - m->m_pkthdr.len);
450 * Bypass packet filtering for packets from a tunnel (gif).
452 if (ip_ipsec_filtertunnel(m))
457 * Run through list of hooks for input packets.
459 * NB: Beware of the destination address changing (e.g.
460 * by NAT rewriting). When this happens, tell
461 * ip_forward to do the right thing.
464 /* Jump over all PFIL processing if hooks are not active. */
465 if (!PFIL_HOOKED(&inet_pfil_hook))
469 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
472 if (m == NULL) /* consumed by filter */
475 ip = mtod(m, struct ip *);
476 dchg = (odst.s_addr != ip->ip_dst.s_addr);
478 #ifdef IPFIREWALL_FORWARD
479 if (m->m_flags & M_FASTFWD_OURS) {
480 m->m_flags &= ~M_FASTFWD_OURS;
483 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
485 * Directly ship on the packet. This allows to forward packets
486 * that were destined for us to some other directly connected
492 #endif /* IPFIREWALL_FORWARD */
496 * Process options and, if not destined for us,
497 * ship it on. ip_dooptions returns 1 when an
498 * error was detected (causing an icmp message
499 * to be sent and the original packet to be freed).
501 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
504 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
505 * matter if it is destined to another node, or whether it is
506 * a multicast one, RSVP wants it! and prevents it from being forwarded
507 * anywhere else. Also checks if the rsvp daemon is running before
508 * grabbing the packet.
510 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
514 * Check our list of addresses, to see if the packet is for us.
515 * If we don't have any addresses, assume any unicast packet
516 * we receive might be for us (and let the upper layers deal
519 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
520 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
524 * Enable a consistency check between the destination address
525 * and the arrival interface for a unicast packet (the RFC 1122
526 * strong ES model) if IP forwarding is disabled and the packet
527 * is not locally generated and the packet is not subject to
530 * XXX - Checking also should be disabled if the destination
531 * address is ipnat'ed to a different interface.
533 * XXX - Checking is incompatible with IP aliases added
534 * to the loopback interface instead of the interface where
535 * the packets are received.
537 * XXX - This is the case for carp vhost IPs as well so we
538 * insert a workaround. If the packet got here, we already
539 * checked with carp_iamatch() and carp_forus().
541 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
542 m->m_pkthdr.rcvif != NULL &&
543 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
545 !m->m_pkthdr.rcvif->if_carp &&
550 * Check for exact addresses in the hash bucket.
552 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
554 * If the address matches, verify that the packet
555 * arrived via the correct interface if checking is
558 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
559 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
563 * Check for broadcast addresses.
565 * Only accept broadcast packets that arrive via the matching
566 * interface. Reception of forwarded directed broadcasts would
567 * be handled via ip_forward() and ether_output() with the loopback
568 * into the stack for SIMPLEX interfaces handled by ether_output().
570 if (m->m_pkthdr.rcvif != NULL &&
571 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
572 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
573 if (ifa->ifa_addr->sa_family != AF_INET)
576 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
579 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
582 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
587 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
588 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
589 V_ipstat.ips_cantforward++;
593 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
594 struct in_multi *inm;
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 * See if we belong to the destination multicast group on the
625 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
628 V_ipstat.ips_notmember++;
634 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
636 if (ip->ip_dst.s_addr == INADDR_ANY)
640 * FAITH(Firewall Aided Internet Translator)
642 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
643 if (V_ip_keepfaith) {
644 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
652 * Not for us; forward if possible and desirable.
654 if (V_ipforwarding == 0) {
655 V_ipstat.ips_cantforward++;
669 * IPSTEALTH: Process non-routing options only
670 * if the packet is destined for us.
672 if (V_ipstealth && hlen > sizeof (struct ip) &&
675 #endif /* IPSTEALTH */
677 /* Count the packet in the ip address stats */
679 ia->ia_ifa.if_ipackets++;
680 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
684 * Attempt reassembly; if it succeeds, proceed.
685 * ip_reass() will return a different mbuf.
687 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
691 ip = mtod(m, struct ip *);
692 /* Get the header length of the reassembled packet */
693 hlen = ip->ip_hl << 2;
697 * Further protocols expect the packet length to be w/o the
704 * enforce IPsec policy checking if we are seeing last header.
705 * note that we do not visit this with protocols with pcb layer
706 * code - like udp/tcp/raw ip.
708 if (ip_ipsec_input(m))
713 * Switch out to protocol's input routine.
715 V_ipstat.ips_delivered++;
717 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
724 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
725 * max has slightly different semantics than the sysctl, for historical
731 INIT_VNET_INET(curvnet);
734 * -1 for unlimited allocation.
737 uma_zone_set_max(V_ipq_zone, 0);
739 * Positive number for specific bound.
742 uma_zone_set_max(V_ipq_zone, V_maxnipq);
744 * Zero specifies no further fragment queue allocation -- set the
745 * bound very low, but rely on implementation elsewhere to actually
746 * prevent allocation and reclaim current queues.
749 uma_zone_set_max(V_ipq_zone, 1);
753 ipq_zone_change(void *tag)
755 INIT_VNET_INET(curvnet);
757 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
758 V_maxnipq = nmbclusters / 32;
764 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
766 INIT_VNET_INET(curvnet);
770 error = sysctl_handle_int(oidp, &i, 0, req);
771 if (error || !req->newptr)
775 * XXXRW: Might be a good idea to sanity check the argument and place
776 * an extreme upper bound.
785 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
786 NULL, 0, sysctl_maxnipq, "I",
787 "Maximum number of IPv4 fragment reassembly queue entries");
790 * Take incoming datagram fragment and try to reassemble it into
791 * whole datagram. If the argument is the first fragment or one
792 * in between the function will return NULL and store the mbuf
793 * in the fragment chain. If the argument is the last fragment
794 * the packet will be reassembled and the pointer to the new
795 * mbuf returned for further processing. Only m_tags attached
796 * to the first packet/fragment are preserved.
797 * The IP header is *NOT* adjusted out of iplen.
800 ip_reass(struct mbuf *m)
802 INIT_VNET_INET(curvnet);
804 struct mbuf *p, *q, *nq, *t;
805 struct ipq *fp = NULL;
806 struct ipqhead *head;
811 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
812 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
813 V_ipstat.ips_fragments++;
814 V_ipstat.ips_fragdropped++;
819 ip = mtod(m, struct ip *);
820 hlen = ip->ip_hl << 2;
822 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
827 * Look for queue of fragments
830 TAILQ_FOREACH(fp, head, ipq_list)
831 if (ip->ip_id == fp->ipq_id &&
832 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
833 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
835 mac_ipq_match(m, fp) &&
837 ip->ip_p == fp->ipq_p)
843 * Attempt to trim the number of allocated fragment queues if it
844 * exceeds the administrative limit.
846 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
848 * drop something from the tail of the current queue
849 * before proceeding further
851 struct ipq *q = TAILQ_LAST(head, ipqhead);
852 if (q == NULL) { /* gak */
853 for (i = 0; i < IPREASS_NHASH; i++) {
854 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
856 V_ipstat.ips_fragtimeout +=
858 ip_freef(&V_ipq[i], r);
863 V_ipstat.ips_fragtimeout += q->ipq_nfrags;
870 * Adjust ip_len to not reflect header,
871 * convert offset of this to bytes.
874 if (ip->ip_off & IP_MF) {
876 * Make sure that fragments have a data length
877 * that's a non-zero multiple of 8 bytes.
879 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
880 V_ipstat.ips_toosmall++; /* XXX */
883 m->m_flags |= M_FRAG;
885 m->m_flags &= ~M_FRAG;
890 * Attempt reassembly; if it succeeds, proceed.
891 * ip_reass() will return a different mbuf.
893 V_ipstat.ips_fragments++;
894 m->m_pkthdr.header = ip;
896 /* Previous ip_reass() started here. */
898 * Presence of header sizes in mbufs
899 * would confuse code below.
905 * If first fragment to arrive, create a reassembly queue.
908 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
912 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
913 uma_zfree(V_ipq_zone, fp);
917 mac_ipq_create(m, fp);
919 TAILQ_INSERT_HEAD(head, fp, ipq_list);
922 fp->ipq_ttl = IPFRAGTTL;
923 fp->ipq_p = ip->ip_p;
924 fp->ipq_id = ip->ip_id;
925 fp->ipq_src = ip->ip_src;
926 fp->ipq_dst = ip->ip_dst;
933 mac_ipq_update(m, fp);
937 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
940 * Handle ECN by comparing this segment with the first one;
941 * if CE is set, do not lose CE.
942 * drop if CE and not-ECT are mixed for the same packet.
944 ecn = ip->ip_tos & IPTOS_ECN_MASK;
945 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
946 if (ecn == IPTOS_ECN_CE) {
947 if (ecn0 == IPTOS_ECN_NOTECT)
949 if (ecn0 != IPTOS_ECN_CE)
950 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
952 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
956 * Find a segment which begins after this one does.
958 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
959 if (GETIP(q)->ip_off > ip->ip_off)
963 * If there is a preceding segment, it may provide some of
964 * our data already. If so, drop the data from the incoming
965 * segment. If it provides all of our data, drop us, otherwise
966 * stick new segment in the proper place.
968 * If some of the data is dropped from the the preceding
969 * segment, then it's checksum is invalidated.
972 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
977 m->m_pkthdr.csum_flags = 0;
981 m->m_nextpkt = p->m_nextpkt;
984 m->m_nextpkt = fp->ipq_frags;
989 * While we overlap succeeding segments trim them or,
990 * if they are completely covered, dequeue them.
992 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
994 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
995 if (i < GETIP(q)->ip_len) {
996 GETIP(q)->ip_len -= i;
997 GETIP(q)->ip_off += i;
999 q->m_pkthdr.csum_flags = 0;
1004 V_ipstat.ips_fragdropped++;
1010 * Check for complete reassembly and perform frag per packet
1013 * Frag limiting is performed here so that the nth frag has
1014 * a chance to complete the packet before we drop the packet.
1015 * As a result, n+1 frags are actually allowed per packet, but
1016 * only n will ever be stored. (n = maxfragsperpacket.)
1020 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1021 if (GETIP(q)->ip_off != next) {
1022 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1023 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1028 next += GETIP(q)->ip_len;
1030 /* Make sure the last packet didn't have the IP_MF flag */
1031 if (p->m_flags & M_FRAG) {
1032 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1033 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1040 * Reassembly is complete. Make sure the packet is a sane size.
1044 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1045 V_ipstat.ips_toolong++;
1046 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1052 * Concatenate fragments.
1059 q->m_nextpkt = NULL;
1060 for (q = nq; q != NULL; q = nq) {
1062 q->m_nextpkt = NULL;
1063 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1064 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1068 * In order to do checksumming faster we do 'end-around carry' here
1069 * (and not in for{} loop), though it implies we are not going to
1070 * reassemble more than 64k fragments.
1072 m->m_pkthdr.csum_data =
1073 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1075 mac_ipq_reassemble(fp, m);
1076 mac_ipq_destroy(fp);
1080 * Create header for new ip packet by modifying header of first
1081 * packet; dequeue and discard fragment reassembly header.
1082 * Make header visible.
1084 ip->ip_len = (ip->ip_hl << 2) + next;
1085 ip->ip_src = fp->ipq_src;
1086 ip->ip_dst = fp->ipq_dst;
1087 TAILQ_REMOVE(head, fp, ipq_list);
1089 uma_zfree(V_ipq_zone, fp);
1090 m->m_len += (ip->ip_hl << 2);
1091 m->m_data -= (ip->ip_hl << 2);
1092 /* some debugging cruft by sklower, below, will go away soon */
1093 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1095 V_ipstat.ips_reassembled++;
1100 V_ipstat.ips_fragdropped++;
1112 * Free a fragment reassembly header and all
1113 * associated datagrams.
1116 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1118 INIT_VNET_INET(curvnet);
1123 while (fp->ipq_frags) {
1125 fp->ipq_frags = q->m_nextpkt;
1128 TAILQ_REMOVE(fhp, fp, ipq_list);
1129 uma_zfree(V_ipq_zone, fp);
1134 * IP timer processing;
1135 * if a timer expires on a reassembly
1136 * queue, discard it.
1141 VNET_ITERATOR_DECL(vnet_iter);
1147 VNET_FOREACH(vnet_iter) {
1148 CURVNET_SET(vnet_iter);
1149 INIT_VNET_INET(vnet_iter);
1150 for (i = 0; i < IPREASS_NHASH; i++) {
1151 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1155 fp = TAILQ_NEXT(fp, ipq_list);
1156 if(--fpp->ipq_ttl == 0) {
1157 V_ipstat.ips_fragtimeout +=
1159 ip_freef(&V_ipq[i], fpp);
1164 * If we are over the maximum number of fragments
1165 * (due to the limit being lowered), drain off
1166 * enough to get down to the new limit.
1168 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1169 for (i = 0; i < IPREASS_NHASH; i++) {
1170 while (V_nipq > V_maxnipq &&
1171 !TAILQ_EMPTY(&V_ipq[i])) {
1172 V_ipstat.ips_fragdropped +=
1173 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1175 TAILQ_FIRST(&V_ipq[i]));
1181 VNET_LIST_RUNLOCK();
1186 * Drain off all datagram fragments.
1191 VNET_ITERATOR_DECL(vnet_iter);
1196 VNET_FOREACH(vnet_iter) {
1197 CURVNET_SET(vnet_iter);
1198 INIT_VNET_INET(vnet_iter);
1199 for (i = 0; i < IPREASS_NHASH; i++) {
1200 while(!TAILQ_EMPTY(&V_ipq[i])) {
1201 V_ipstat.ips_fragdropped +=
1202 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1203 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1208 VNET_LIST_RUNLOCK();
1214 * The protocol to be inserted into ip_protox[] must be already registered
1215 * in inetsw[], either statically or through pf_proto_register().
1218 ipproto_register(u_char ipproto)
1222 /* Sanity checks. */
1224 return (EPROTONOSUPPORT);
1227 * The protocol slot must not be occupied by another protocol
1228 * already. An index pointing to IPPROTO_RAW is unused.
1230 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1232 return (EPFNOSUPPORT);
1233 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1236 /* Find the protocol position in inetsw[] and set the index. */
1237 for (pr = inetdomain.dom_protosw;
1238 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1239 if (pr->pr_domain->dom_family == PF_INET &&
1240 pr->pr_protocol && pr->pr_protocol == ipproto) {
1241 /* Be careful to only index valid IP protocols. */
1242 if (pr->pr_protocol < IPPROTO_MAX) {
1243 ip_protox[pr->pr_protocol] = pr - inetsw;
1249 return (EPROTONOSUPPORT);
1253 ipproto_unregister(u_char ipproto)
1257 /* Sanity checks. */
1259 return (EPROTONOSUPPORT);
1261 /* Check if the protocol was indeed registered. */
1262 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1264 return (EPFNOSUPPORT);
1265 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1268 /* Reset the protocol slot to IPPROTO_RAW. */
1269 ip_protox[ipproto] = pr - inetsw;
1274 * Given address of next destination (final or next hop),
1275 * return internet address info of interface to be used to get there.
1278 ip_rtaddr(struct in_addr dst, u_int fibnum)
1281 struct sockaddr_in *sin;
1282 struct in_ifaddr *ifa;
1284 bzero(&sro, sizeof(sro));
1285 sin = (struct sockaddr_in *)&sro.ro_dst;
1286 sin->sin_family = AF_INET;
1287 sin->sin_len = sizeof(*sin);
1288 sin->sin_addr = dst;
1289 in_rtalloc_ign(&sro, 0, fibnum);
1291 if (sro.ro_rt == NULL)
1294 ifa = ifatoia(sro.ro_rt->rt_ifa);
1299 u_char inetctlerrmap[PRC_NCMDS] = {
1301 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1302 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1303 EMSGSIZE, EHOSTUNREACH, 0, 0,
1304 0, 0, EHOSTUNREACH, 0,
1305 ENOPROTOOPT, ECONNREFUSED
1309 * Forward a packet. If some error occurs return the sender
1310 * an icmp packet. Note we can't always generate a meaningful
1311 * icmp message because icmp doesn't have a large enough repertoire
1312 * of codes and types.
1314 * If not forwarding, just drop the packet. This could be confusing
1315 * if ipforwarding was zero but some routing protocol was advancing
1316 * us as a gateway to somewhere. However, we must let the routing
1317 * protocol deal with that.
1319 * The srcrt parameter indicates whether the packet is being forwarded
1320 * via a source route.
1323 ip_forward(struct mbuf *m, int srcrt)
1325 INIT_VNET_INET(curvnet);
1326 struct ip *ip = mtod(m, struct ip *);
1327 struct in_ifaddr *ia = NULL;
1329 struct in_addr dest;
1331 int error, type = 0, code = 0, mtu = 0;
1333 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1334 V_ipstat.ips_cantforward++;
1341 if (ip->ip_ttl <= IPTTLDEC) {
1342 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1350 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1351 if (!srcrt && ia == NULL) {
1352 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1357 * Save the IP header and at most 8 bytes of the payload,
1358 * in case we need to generate an ICMP message to the src.
1360 * XXX this can be optimized a lot by saving the data in a local
1361 * buffer on the stack (72 bytes at most), and only allocating the
1362 * mbuf if really necessary. The vast majority of the packets
1363 * are forwarded without having to send an ICMP back (either
1364 * because unnecessary, or because rate limited), so we are
1365 * really we are wasting a lot of work here.
1367 * We don't use m_copy() because it might return a reference
1368 * to a shared cluster. Both this function and ip_output()
1369 * assume exclusive access to the IP header in `m', so any
1370 * data in a cluster may change before we reach icmp_error().
1372 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1373 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1375 * It's probably ok if the pkthdr dup fails (because
1376 * the deep copy of the tag chain failed), but for now
1377 * be conservative and just discard the copy since
1378 * code below may some day want the tags.
1383 if (mcopy != NULL) {
1384 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1385 mcopy->m_pkthdr.len = mcopy->m_len;
1386 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1392 ip->ip_ttl -= IPTTLDEC;
1398 * If forwarding packet using same interface that it came in on,
1399 * perhaps should send a redirect to sender to shortcut a hop.
1400 * Only send redirect if source is sending directly to us,
1401 * and if packet was not source routed (or has any options).
1402 * Also, don't send redirect if forwarding using a default route
1403 * or a route modified by a redirect.
1406 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1407 struct sockaddr_in *sin;
1410 bzero(&ro, sizeof(ro));
1411 sin = (struct sockaddr_in *)&ro.ro_dst;
1412 sin->sin_family = AF_INET;
1413 sin->sin_len = sizeof(*sin);
1414 sin->sin_addr = ip->ip_dst;
1415 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1419 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1420 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1421 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1422 u_long src = ntohl(ip->ip_src.s_addr);
1425 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1426 if (rt->rt_flags & RTF_GATEWAY)
1427 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1429 dest.s_addr = ip->ip_dst.s_addr;
1430 /* Router requirements says to only send host redirects */
1431 type = ICMP_REDIRECT;
1432 code = ICMP_REDIRECT_HOST;
1440 * Try to cache the route MTU from ip_output so we can consider it for
1441 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1443 bzero(&ro, sizeof(ro));
1445 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1447 if (error == EMSGSIZE && ro.ro_rt)
1448 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1453 V_ipstat.ips_cantforward++;
1455 V_ipstat.ips_forward++;
1457 V_ipstat.ips_redirectsent++;
1469 case 0: /* forwarded, but need redirect */
1470 /* type, code set above */
1473 case ENETUNREACH: /* shouldn't happen, checked above */
1478 type = ICMP_UNREACH;
1479 code = ICMP_UNREACH_HOST;
1483 type = ICMP_UNREACH;
1484 code = ICMP_UNREACH_NEEDFRAG;
1488 * If IPsec is configured for this path,
1489 * override any possibly mtu value set by ip_output.
1491 mtu = ip_ipsec_mtu(m, mtu);
1494 * If the MTU was set before make sure we are below the
1496 * If the MTU wasn't set before use the interface mtu or
1497 * fall back to the next smaller mtu step compared to the
1498 * current packet size.
1502 mtu = min(mtu, ia->ia_ifp->if_mtu);
1505 mtu = ia->ia_ifp->if_mtu;
1507 mtu = ip_next_mtu(ip->ip_len, 0);
1509 V_ipstat.ips_cantfrag++;
1514 * A router should not generate ICMP_SOURCEQUENCH as
1515 * required in RFC1812 Requirements for IP Version 4 Routers.
1516 * Source quench could be a big problem under DoS attacks,
1517 * or if the underlying interface is rate-limited.
1518 * Those who need source quench packets may re-enable them
1519 * via the net.inet.ip.sendsourcequench sysctl.
1521 if (V_ip_sendsourcequench == 0) {
1525 type = ICMP_SOURCEQUENCH;
1530 case EACCES: /* ipfw denied packet */
1534 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1538 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1541 INIT_VNET_NET(inp->inp_vnet);
1543 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1547 if (inp->inp_socket->so_options & SO_BINTIME) {
1548 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1549 SCM_BINTIME, SOL_SOCKET);
1551 mp = &(*mp)->m_next;
1553 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1556 bintime2timeval(&bt, &tv);
1557 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1558 SCM_TIMESTAMP, SOL_SOCKET);
1560 mp = &(*mp)->m_next;
1563 if (inp->inp_flags & INP_RECVDSTADDR) {
1564 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1565 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1567 mp = &(*mp)->m_next;
1569 if (inp->inp_flags & INP_RECVTTL) {
1570 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1571 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1573 mp = &(*mp)->m_next;
1577 * Moving these out of udp_input() made them even more broken
1578 * than they already were.
1580 /* options were tossed already */
1581 if (inp->inp_flags & INP_RECVOPTS) {
1582 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1583 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1585 mp = &(*mp)->m_next;
1587 /* ip_srcroute doesn't do what we want here, need to fix */
1588 if (inp->inp_flags & INP_RECVRETOPTS) {
1589 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1590 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1592 mp = &(*mp)->m_next;
1595 if (inp->inp_flags & INP_RECVIF) {
1598 struct sockaddr_dl sdl;
1601 struct sockaddr_dl *sdp;
1602 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1604 if (((ifp = m->m_pkthdr.rcvif))
1605 && ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1606 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1608 * Change our mind and don't try copy.
1610 if ((sdp->sdl_family != AF_LINK)
1611 || (sdp->sdl_len > sizeof(sdlbuf))) {
1614 bcopy(sdp, sdl2, sdp->sdl_len);
1618 = offsetof(struct sockaddr_dl, sdl_data[0]);
1619 sdl2->sdl_family = AF_LINK;
1620 sdl2->sdl_index = 0;
1621 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1623 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1624 IP_RECVIF, IPPROTO_IP);
1626 mp = &(*mp)->m_next;
1631 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1632 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1633 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1637 ip_rsvp_init(struct socket *so)
1639 INIT_VNET_INET(so->so_vnet);
1641 if (so->so_type != SOCK_RAW ||
1642 so->so_proto->pr_protocol != IPPROTO_RSVP)
1645 if (V_ip_rsvpd != NULL)
1650 * This may seem silly, but we need to be sure we don't over-increment
1651 * the RSVP counter, in case something slips up.
1653 if (!V_ip_rsvp_on) {
1664 INIT_VNET_INET(curvnet);
1668 * This may seem silly, but we need to be sure we don't over-decrement
1669 * the RSVP counter, in case something slips up.
1679 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1681 INIT_VNET_INET(curvnet);
1683 if (rsvp_input_p) { /* call the real one if loaded */
1684 rsvp_input_p(m, off);
1688 /* Can still get packets with rsvp_on = 0 if there is a local member
1689 * of the group to which the RSVP packet is addressed. But in this
1690 * case we want to throw the packet away.
1698 if (V_ip_rsvpd != NULL) {
1702 /* Drop the packet */