2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_bootp.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_kdtrace.h"
40 #include "opt_route.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
50 #include <sys/kernel.h>
52 #include <sys/rwlock.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/netisr.h>
66 #include <netinet/in.h>
67 #include <netinet/in_kdtrace.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_icmp.h>
75 #include <netinet/ip_options.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/ip_carp.h>
79 #include <netinet/ip_ipsec.h>
82 #include <sys/socketvar.h>
84 #include <security/mac/mac_framework.h>
87 CTASSERT(sizeof(struct ip) == 20);
90 struct rwlock in_ifaddr_lock;
91 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
93 VNET_DEFINE(int, rsvp_on);
95 VNET_DEFINE(int, ipforwarding);
96 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97 &VNET_NAME(ipforwarding), 0,
98 "Enable IP forwarding between interfaces");
100 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
101 #define V_ipsendredirects VNET(ipsendredirects)
102 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
103 &VNET_NAME(ipsendredirects), 0,
104 "Enable sending IP redirects");
106 static VNET_DEFINE(int, ip_keepfaith);
107 #define V_ip_keepfaith VNET(ip_keepfaith)
108 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
109 &VNET_NAME(ip_keepfaith), 0,
110 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
112 static VNET_DEFINE(int, ip_sendsourcequench);
113 #define V_ip_sendsourcequench VNET(ip_sendsourcequench)
114 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
115 &VNET_NAME(ip_sendsourcequench), 0,
116 "Enable the transmission of source quench packets");
118 VNET_DEFINE(int, ip_do_randomid);
119 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
120 &VNET_NAME(ip_do_randomid), 0,
121 "Assign random ip_id values");
124 * XXX - Setting ip_checkinterface mostly implements the receive side of
125 * the Strong ES model described in RFC 1122, but since the routing table
126 * and transmit implementation do not implement the Strong ES model,
127 * setting this to 1 results in an odd hybrid.
129 * XXX - ip_checkinterface currently must be disabled if you use ipnat
130 * to translate the destination address to another local interface.
132 * XXX - ip_checkinterface must be disabled if you add IP aliases
133 * to the loopback interface instead of the interface where the
134 * packets for those addresses are received.
136 static VNET_DEFINE(int, ip_checkinterface);
137 #define V_ip_checkinterface VNET(ip_checkinterface)
138 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
139 &VNET_NAME(ip_checkinterface), 0,
140 "Verify packet arrives on correct interface");
142 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
144 static struct netisr_handler ip_nh = {
146 .nh_handler = ip_input,
147 .nh_proto = NETISR_IP,
148 .nh_policy = NETISR_POLICY_FLOW,
151 extern struct domain inetdomain;
152 extern struct protosw inetsw[];
153 u_char ip_protox[IPPROTO_MAX];
154 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
155 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
156 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
158 static VNET_DEFINE(uma_zone_t, ipq_zone);
159 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
160 static struct mtx ipqlock;
162 #define V_ipq_zone VNET(ipq_zone)
163 #define V_ipq VNET(ipq)
165 #define IPQ_LOCK() mtx_lock(&ipqlock)
166 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
167 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
168 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
170 static void maxnipq_update(void);
171 static void ipq_zone_change(void *);
172 static void ip_drain_locked(void);
174 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
175 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
176 #define V_maxnipq VNET(maxnipq)
177 #define V_nipq VNET(nipq)
178 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
180 "Current number of IPv4 fragment reassembly queue entries");
182 static VNET_DEFINE(int, maxfragsperpacket);
183 #define V_maxfragsperpacket VNET(maxfragsperpacket)
184 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
185 &VNET_NAME(maxfragsperpacket), 0,
186 "Maximum number of IPv4 fragments allowed per packet");
189 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
190 &ip_mtu, 0, "Default MTU");
194 VNET_DEFINE(int, ipstealth);
195 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
196 &VNET_NAME(ipstealth), 0,
197 "IP stealth mode, no TTL decrementation on forwarding");
200 static void ip_freef(struct ipqhead *, struct ipq *);
203 * IP statistics are stored in the "array" of counter(9)s.
205 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
206 VNET_PCPUSTAT_SYSINIT(ipstat);
207 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
208 "IP statistics (struct ipstat, netinet/ip_var.h)");
211 VNET_PCPUSTAT_SYSUNINIT(ipstat);
215 * Kernel module interface for updating ipstat. The argument is an index
216 * into ipstat treated as an array.
219 kmod_ipstat_inc(int statnum)
222 counter_u64_add(VNET(ipstat)[statnum], 1);
226 kmod_ipstat_dec(int statnum)
229 counter_u64_add(VNET(ipstat)[statnum], -1);
233 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
237 netisr_getqlimit(&ip_nh, &qlimit);
238 error = sysctl_handle_int(oidp, &qlimit, 0, req);
239 if (error || !req->newptr)
243 return (netisr_setqlimit(&ip_nh, qlimit));
245 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
246 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
247 "Maximum size of the IP input queue");
250 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
252 u_int64_t qdrops_long;
255 netisr_getqdrops(&ip_nh, &qdrops_long);
256 qdrops = qdrops_long;
257 error = sysctl_handle_int(oidp, &qdrops, 0, req);
258 if (error || !req->newptr)
262 netisr_clearqdrops(&ip_nh);
266 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
267 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
268 "Number of packets dropped from the IP input queue");
271 * IP initialization: fill in IP protocol switch table.
272 * All protocols not implemented in kernel go to raw IP protocol handler.
280 V_ip_id = time_second & 0xffff;
282 TAILQ_INIT(&V_in_ifaddrhead);
283 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
285 /* Initialize IP reassembly queue. */
286 for (i = 0; i < IPREASS_NHASH; i++)
287 TAILQ_INIT(&V_ipq[i]);
288 V_maxnipq = nmbclusters / 32;
289 V_maxfragsperpacket = 16;
290 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
291 NULL, UMA_ALIGN_PTR, 0);
294 /* Initialize packet filter hooks. */
295 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
296 V_inet_pfil_hook.ph_af = AF_INET;
297 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
298 printf("%s: WARNING: unable to register pfil hook, "
299 "error %d\n", __func__, i);
301 /* Skip initialization of globals for non-default instances. */
302 if (!IS_DEFAULT_VNET(curvnet))
305 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
307 panic("ip_init: PF_INET not found");
309 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
310 for (i = 0; i < IPPROTO_MAX; i++)
311 ip_protox[i] = pr - inetsw;
313 * Cycle through IP protocols and put them into the appropriate place
316 for (pr = inetdomain.dom_protosw;
317 pr < inetdomain.dom_protoswNPROTOSW; pr++)
318 if (pr->pr_domain->dom_family == PF_INET &&
319 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
320 /* Be careful to only index valid IP protocols. */
321 if (pr->pr_protocol < IPPROTO_MAX)
322 ip_protox[pr->pr_protocol] = pr - inetsw;
325 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
326 NULL, EVENTHANDLER_PRI_ANY);
328 /* Initialize various other remaining things. */
330 netisr_register(&ip_nh);
339 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
340 printf("%s: WARNING: unable to unregister pfil hook, "
341 "error %d\n", __func__, i);
343 /* Cleanup in_ifaddr hash table; should be empty. */
344 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
350 uma_zdestroy(V_ipq_zone);
355 * Ip input routine. Checksum and byte swap header. If fragmented
356 * try to reassemble. Process options. Pass to next level.
359 ip_input(struct mbuf *m)
361 struct ip *ip = NULL;
362 struct in_ifaddr *ia = NULL;
365 int checkif, hlen = 0;
366 uint16_t sum, ip_len;
367 int dchg = 0; /* dest changed after fw */
368 struct in_addr odst; /* original dst address */
372 if (m->m_flags & M_FASTFWD_OURS) {
373 m->m_flags &= ~M_FASTFWD_OURS;
374 /* Set up some basics that will be used later. */
375 ip = mtod(m, struct ip *);
376 hlen = ip->ip_hl << 2;
377 ip_len = ntohs(ip->ip_len);
381 IPSTAT_INC(ips_total);
383 if (m->m_pkthdr.len < sizeof(struct ip))
386 if (m->m_len < sizeof (struct ip) &&
387 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
388 IPSTAT_INC(ips_toosmall);
391 ip = mtod(m, struct ip *);
393 if (ip->ip_v != IPVERSION) {
394 IPSTAT_INC(ips_badvers);
398 hlen = ip->ip_hl << 2;
399 if (hlen < sizeof(struct ip)) { /* minimum header length */
400 IPSTAT_INC(ips_badhlen);
403 if (hlen > m->m_len) {
404 if ((m = m_pullup(m, hlen)) == NULL) {
405 IPSTAT_INC(ips_badhlen);
408 ip = mtod(m, struct ip *);
411 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
413 /* 127/8 must not appear on wire - RFC1122 */
414 ifp = m->m_pkthdr.rcvif;
415 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
416 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
417 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
418 IPSTAT_INC(ips_badaddr);
423 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
424 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
426 if (hlen == sizeof(struct ip)) {
427 sum = in_cksum_hdr(ip);
429 sum = in_cksum(m, hlen);
433 IPSTAT_INC(ips_badsum);
438 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
439 /* packet is dropped by traffic conditioner */
443 ip_len = ntohs(ip->ip_len);
445 IPSTAT_INC(ips_badlen);
450 * Check that the amount of data in the buffers
451 * is as at least much as the IP header would have us expect.
452 * Trim mbufs if longer than we expect.
453 * Drop packet if shorter than we expect.
455 if (m->m_pkthdr.len < ip_len) {
457 IPSTAT_INC(ips_tooshort);
460 if (m->m_pkthdr.len > ip_len) {
461 if (m->m_len == m->m_pkthdr.len) {
463 m->m_pkthdr.len = ip_len;
465 m_adj(m, ip_len - m->m_pkthdr.len);
469 * Bypass packet filtering for packets previously handled by IPsec.
471 if (ip_ipsec_filtertunnel(m))
476 * Run through list of hooks for input packets.
478 * NB: Beware of the destination address changing (e.g.
479 * by NAT rewriting). When this happens, tell
480 * ip_forward to do the right thing.
483 /* Jump over all PFIL processing if hooks are not active. */
484 if (!PFIL_HOOKED(&V_inet_pfil_hook))
488 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
490 if (m == NULL) /* consumed by filter */
493 ip = mtod(m, struct ip *);
494 dchg = (odst.s_addr != ip->ip_dst.s_addr);
495 ifp = m->m_pkthdr.rcvif;
497 if (m->m_flags & M_FASTFWD_OURS) {
498 m->m_flags &= ~M_FASTFWD_OURS;
501 if (m->m_flags & M_IP_NEXTHOP) {
502 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
505 * Directly ship the packet on. This allows
506 * forwarding packets originally destined to us
507 * to some other directly connected host.
516 * Process options and, if not destined for us,
517 * ship it on. ip_dooptions returns 1 when an
518 * error was detected (causing an icmp message
519 * to be sent and the original packet to be freed).
521 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
524 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
525 * matter if it is destined to another node, or whether it is
526 * a multicast one, RSVP wants it! and prevents it from being forwarded
527 * anywhere else. Also checks if the rsvp daemon is running before
528 * grabbing the packet.
530 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
534 * Check our list of addresses, to see if the packet is for us.
535 * If we don't have any addresses, assume any unicast packet
536 * we receive might be for us (and let the upper layers deal
539 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
540 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
544 * Enable a consistency check between the destination address
545 * and the arrival interface for a unicast packet (the RFC 1122
546 * strong ES model) if IP forwarding is disabled and the packet
547 * is not locally generated and the packet is not subject to
550 * XXX - Checking also should be disabled if the destination
551 * address is ipnat'ed to a different interface.
553 * XXX - Checking is incompatible with IP aliases added
554 * to the loopback interface instead of the interface where
555 * the packets are received.
557 * XXX - This is the case for carp vhost IPs as well so we
558 * insert a workaround. If the packet got here, we already
559 * checked with carp_iamatch() and carp_forus().
561 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
562 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
563 ifp->if_carp == NULL && (dchg == 0);
566 * Check for exact addresses in the hash bucket.
568 /* IN_IFADDR_RLOCK(); */
569 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
571 * If the address matches, verify that the packet
572 * arrived via the correct interface if checking is
575 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
576 (!checkif || ia->ia_ifp == ifp)) {
577 ifa_ref(&ia->ia_ifa);
578 /* IN_IFADDR_RUNLOCK(); */
582 /* IN_IFADDR_RUNLOCK(); */
585 * Check for broadcast addresses.
587 * Only accept broadcast packets that arrive via the matching
588 * interface. Reception of forwarded directed broadcasts would
589 * be handled via ip_forward() and ether_output() with the loopback
590 * into the stack for SIMPLEX interfaces handled by ether_output().
592 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
594 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
595 if (ifa->ifa_addr->sa_family != AF_INET)
598 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
601 IF_ADDR_RUNLOCK(ifp);
605 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
607 IF_ADDR_RUNLOCK(ifp);
612 IF_ADDR_RUNLOCK(ifp);
615 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
616 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
617 IPSTAT_INC(ips_cantforward);
621 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
624 * If we are acting as a multicast router, all
625 * incoming multicast packets are passed to the
626 * kernel-level multicast forwarding function.
627 * The packet is returned (relatively) intact; if
628 * ip_mforward() returns a non-zero value, the packet
629 * must be discarded, else it may be accepted below.
631 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
632 IPSTAT_INC(ips_cantforward);
638 * The process-level routing daemon needs to receive
639 * all multicast IGMP packets, whether or not this
640 * host belongs to their destination groups.
642 if (ip->ip_p == IPPROTO_IGMP)
644 IPSTAT_INC(ips_forward);
647 * Assume the packet is for us, to avoid prematurely taking
648 * a lock on the in_multi hash. Protocols must perform
649 * their own filtering and update statistics accordingly.
653 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
655 if (ip->ip_dst.s_addr == INADDR_ANY)
659 * FAITH(Firewall Aided Internet Translator)
661 if (ifp && ifp->if_type == IFT_FAITH) {
662 if (V_ip_keepfaith) {
663 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
671 * Not for us; forward if possible and desirable.
673 if (V_ipforwarding == 0) {
674 IPSTAT_INC(ips_cantforward);
688 * IPSTEALTH: Process non-routing options only
689 * if the packet is destined for us.
691 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) {
693 ifa_free(&ia->ia_ifa);
696 #endif /* IPSTEALTH */
698 /* Count the packet in the ip address stats */
700 ia->ia_ifa.if_ipackets++;
701 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
702 ifa_free(&ia->ia_ifa);
706 * Attempt reassembly; if it succeeds, proceed.
707 * ip_reass() will return a different mbuf.
709 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
710 /* XXXGL: shouldn't we save & set m_flags? */
714 ip = mtod(m, struct ip *);
715 /* Get the header length of the reassembled packet */
716 hlen = ip->ip_hl << 2;
721 * enforce IPsec policy checking if we are seeing last header.
722 * note that we do not visit this with protocols with pcb layer
723 * code - like udp/tcp/raw ip.
725 if (ip_ipsec_input(m))
730 * Switch out to protocol's input routine.
732 IPSTAT_INC(ips_delivered);
734 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
741 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
742 * max has slightly different semantics than the sysctl, for historical
750 * -1 for unlimited allocation.
753 uma_zone_set_max(V_ipq_zone, 0);
755 * Positive number for specific bound.
758 uma_zone_set_max(V_ipq_zone, V_maxnipq);
760 * Zero specifies no further fragment queue allocation -- set the
761 * bound very low, but rely on implementation elsewhere to actually
762 * prevent allocation and reclaim current queues.
765 uma_zone_set_max(V_ipq_zone, 1);
769 ipq_zone_change(void *tag)
772 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
773 V_maxnipq = nmbclusters / 32;
779 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
784 error = sysctl_handle_int(oidp, &i, 0, req);
785 if (error || !req->newptr)
789 * XXXRW: Might be a good idea to sanity check the argument and place
790 * an extreme upper bound.
799 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
800 NULL, 0, sysctl_maxnipq, "I",
801 "Maximum number of IPv4 fragment reassembly queue entries");
803 #define M_IP_FRAG M_PROTO9
806 * Take incoming datagram fragment and try to reassemble it into
807 * whole datagram. If the argument is the first fragment or one
808 * in between the function will return NULL and store the mbuf
809 * in the fragment chain. If the argument is the last fragment
810 * the packet will be reassembled and the pointer to the new
811 * mbuf returned for further processing. Only m_tags attached
812 * to the first packet/fragment are preserved.
813 * The IP header is *NOT* adjusted out of iplen.
816 ip_reass(struct mbuf *m)
819 struct mbuf *p, *q, *nq, *t;
820 struct ipq *fp = NULL;
821 struct ipqhead *head;
826 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
827 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
828 IPSTAT_INC(ips_fragments);
829 IPSTAT_INC(ips_fragdropped);
834 ip = mtod(m, struct ip *);
835 hlen = ip->ip_hl << 2;
837 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
842 * Look for queue of fragments
845 TAILQ_FOREACH(fp, head, ipq_list)
846 if (ip->ip_id == fp->ipq_id &&
847 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
848 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
850 mac_ipq_match(m, fp) &&
852 ip->ip_p == fp->ipq_p)
858 * Attempt to trim the number of allocated fragment queues if it
859 * exceeds the administrative limit.
861 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
863 * drop something from the tail of the current queue
864 * before proceeding further
866 struct ipq *q = TAILQ_LAST(head, ipqhead);
867 if (q == NULL) { /* gak */
868 for (i = 0; i < IPREASS_NHASH; i++) {
869 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
871 IPSTAT_ADD(ips_fragtimeout,
873 ip_freef(&V_ipq[i], r);
878 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
885 * Adjust ip_len to not reflect header,
886 * convert offset of this to bytes.
888 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
889 if (ip->ip_off & htons(IP_MF)) {
891 * Make sure that fragments have a data length
892 * that's a non-zero multiple of 8 bytes.
894 if (ip->ip_len == htons(0) || (ntohs(ip->ip_len) & 0x7) != 0) {
895 IPSTAT_INC(ips_toosmall); /* XXX */
898 m->m_flags |= M_IP_FRAG;
900 m->m_flags &= ~M_IP_FRAG;
901 ip->ip_off = htons(ntohs(ip->ip_off) << 3);
904 * Attempt reassembly; if it succeeds, proceed.
905 * ip_reass() will return a different mbuf.
907 IPSTAT_INC(ips_fragments);
908 m->m_pkthdr.PH_loc.ptr = ip;
910 /* Previous ip_reass() started here. */
912 * Presence of header sizes in mbufs
913 * would confuse code below.
919 * If first fragment to arrive, create a reassembly queue.
922 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
926 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
927 uma_zfree(V_ipq_zone, fp);
931 mac_ipq_create(m, fp);
933 TAILQ_INSERT_HEAD(head, fp, ipq_list);
936 fp->ipq_ttl = IPFRAGTTL;
937 fp->ipq_p = ip->ip_p;
938 fp->ipq_id = ip->ip_id;
939 fp->ipq_src = ip->ip_src;
940 fp->ipq_dst = ip->ip_dst;
947 mac_ipq_update(m, fp);
951 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.PH_loc.ptr))
954 * Handle ECN by comparing this segment with the first one;
955 * if CE is set, do not lose CE.
956 * drop if CE and not-ECT are mixed for the same packet.
958 ecn = ip->ip_tos & IPTOS_ECN_MASK;
959 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
960 if (ecn == IPTOS_ECN_CE) {
961 if (ecn0 == IPTOS_ECN_NOTECT)
963 if (ecn0 != IPTOS_ECN_CE)
964 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
966 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
970 * Find a segment which begins after this one does.
972 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
973 if (ntohs(GETIP(q)->ip_off) > ntohs(ip->ip_off))
977 * If there is a preceding segment, it may provide some of
978 * our data already. If so, drop the data from the incoming
979 * segment. If it provides all of our data, drop us, otherwise
980 * stick new segment in the proper place.
982 * If some of the data is dropped from the preceding
983 * segment, then it's checksum is invalidated.
986 i = ntohs(GETIP(p)->ip_off) + ntohs(GETIP(p)->ip_len) -
989 if (i >= ntohs(ip->ip_len))
992 m->m_pkthdr.csum_flags = 0;
993 ip->ip_off = htons(ntohs(ip->ip_off) + i);
994 ip->ip_len = htons(ntohs(ip->ip_len) - i);
996 m->m_nextpkt = p->m_nextpkt;
999 m->m_nextpkt = fp->ipq_frags;
1004 * While we overlap succeeding segments trim them or,
1005 * if they are completely covered, dequeue them.
1007 for (; q != NULL && ntohs(ip->ip_off) + ntohs(ip->ip_len) >
1008 ntohs(GETIP(q)->ip_off); q = nq) {
1009 i = (ntohs(ip->ip_off) + ntohs(ip->ip_len)) -
1010 ntohs(GETIP(q)->ip_off);
1011 if (i < ntohs(GETIP(q)->ip_len)) {
1012 GETIP(q)->ip_len = htons(ntohs(GETIP(q)->ip_len) - i);
1013 GETIP(q)->ip_off = htons(ntohs(GETIP(q)->ip_off) + i);
1015 q->m_pkthdr.csum_flags = 0;
1020 IPSTAT_INC(ips_fragdropped);
1026 * Check for complete reassembly and perform frag per packet
1029 * Frag limiting is performed here so that the nth frag has
1030 * a chance to complete the packet before we drop the packet.
1031 * As a result, n+1 frags are actually allowed per packet, but
1032 * only n will ever be stored. (n = maxfragsperpacket.)
1036 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1037 if (ntohs(GETIP(q)->ip_off) != next) {
1038 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1039 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1044 next += ntohs(GETIP(q)->ip_len);
1046 /* Make sure the last packet didn't have the IP_MF flag */
1047 if (p->m_flags & M_IP_FRAG) {
1048 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1049 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1056 * Reassembly is complete. Make sure the packet is a sane size.
1060 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1061 IPSTAT_INC(ips_toolong);
1062 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1068 * Concatenate fragments.
1075 q->m_nextpkt = NULL;
1076 for (q = nq; q != NULL; q = nq) {
1078 q->m_nextpkt = NULL;
1079 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1080 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1084 * In order to do checksumming faster we do 'end-around carry' here
1085 * (and not in for{} loop), though it implies we are not going to
1086 * reassemble more than 64k fragments.
1088 while (m->m_pkthdr.csum_data & 0xffff0000)
1089 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1090 (m->m_pkthdr.csum_data >> 16);
1092 mac_ipq_reassemble(fp, m);
1093 mac_ipq_destroy(fp);
1097 * Create header for new ip packet by modifying header of first
1098 * packet; dequeue and discard fragment reassembly header.
1099 * Make header visible.
1101 ip->ip_len = htons((ip->ip_hl << 2) + next);
1102 ip->ip_src = fp->ipq_src;
1103 ip->ip_dst = fp->ipq_dst;
1104 TAILQ_REMOVE(head, fp, ipq_list);
1106 uma_zfree(V_ipq_zone, fp);
1107 m->m_len += (ip->ip_hl << 2);
1108 m->m_data -= (ip->ip_hl << 2);
1109 /* some debugging cruft by sklower, below, will go away soon */
1110 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1112 IPSTAT_INC(ips_reassembled);
1117 IPSTAT_INC(ips_fragdropped);
1129 * Free a fragment reassembly header and all
1130 * associated datagrams.
1133 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1139 while (fp->ipq_frags) {
1141 fp->ipq_frags = q->m_nextpkt;
1144 TAILQ_REMOVE(fhp, fp, ipq_list);
1145 uma_zfree(V_ipq_zone, fp);
1150 * IP timer processing;
1151 * if a timer expires on a reassembly
1152 * queue, discard it.
1157 VNET_ITERATOR_DECL(vnet_iter);
1161 VNET_LIST_RLOCK_NOSLEEP();
1163 VNET_FOREACH(vnet_iter) {
1164 CURVNET_SET(vnet_iter);
1165 for (i = 0; i < IPREASS_NHASH; i++) {
1166 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1170 fp = TAILQ_NEXT(fp, ipq_list);
1171 if(--fpp->ipq_ttl == 0) {
1172 IPSTAT_ADD(ips_fragtimeout,
1174 ip_freef(&V_ipq[i], fpp);
1179 * If we are over the maximum number of fragments
1180 * (due to the limit being lowered), drain off
1181 * enough to get down to the new limit.
1183 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1184 for (i = 0; i < IPREASS_NHASH; i++) {
1185 while (V_nipq > V_maxnipq &&
1186 !TAILQ_EMPTY(&V_ipq[i])) {
1187 IPSTAT_ADD(ips_fragdropped,
1188 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1190 TAILQ_FIRST(&V_ipq[i]));
1197 VNET_LIST_RUNLOCK_NOSLEEP();
1201 * Drain off all datagram fragments.
1204 ip_drain_locked(void)
1210 for (i = 0; i < IPREASS_NHASH; i++) {
1211 while(!TAILQ_EMPTY(&V_ipq[i])) {
1212 IPSTAT_ADD(ips_fragdropped,
1213 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1214 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1222 VNET_ITERATOR_DECL(vnet_iter);
1224 VNET_LIST_RLOCK_NOSLEEP();
1226 VNET_FOREACH(vnet_iter) {
1227 CURVNET_SET(vnet_iter);
1232 VNET_LIST_RUNLOCK_NOSLEEP();
1237 * The protocol to be inserted into ip_protox[] must be already registered
1238 * in inetsw[], either statically or through pf_proto_register().
1241 ipproto_register(short ipproto)
1245 /* Sanity checks. */
1246 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1247 return (EPROTONOSUPPORT);
1250 * The protocol slot must not be occupied by another protocol
1251 * already. An index pointing to IPPROTO_RAW is unused.
1253 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1255 return (EPFNOSUPPORT);
1256 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1259 /* Find the protocol position in inetsw[] and set the index. */
1260 for (pr = inetdomain.dom_protosw;
1261 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1262 if (pr->pr_domain->dom_family == PF_INET &&
1263 pr->pr_protocol && pr->pr_protocol == ipproto) {
1264 ip_protox[pr->pr_protocol] = pr - inetsw;
1268 return (EPROTONOSUPPORT);
1272 ipproto_unregister(short ipproto)
1276 /* Sanity checks. */
1277 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1278 return (EPROTONOSUPPORT);
1280 /* Check if the protocol was indeed registered. */
1281 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1283 return (EPFNOSUPPORT);
1284 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1287 /* Reset the protocol slot to IPPROTO_RAW. */
1288 ip_protox[ipproto] = pr - inetsw;
1293 * Given address of next destination (final or next hop), return (referenced)
1294 * internet address info of interface to be used to get there.
1297 ip_rtaddr(struct in_addr dst, u_int fibnum)
1300 struct sockaddr_in *sin;
1301 struct in_ifaddr *ia;
1303 bzero(&sro, sizeof(sro));
1304 sin = (struct sockaddr_in *)&sro.ro_dst;
1305 sin->sin_family = AF_INET;
1306 sin->sin_len = sizeof(*sin);
1307 sin->sin_addr = dst;
1308 in_rtalloc_ign(&sro, 0, fibnum);
1310 if (sro.ro_rt == NULL)
1313 ia = ifatoia(sro.ro_rt->rt_ifa);
1314 ifa_ref(&ia->ia_ifa);
1319 u_char inetctlerrmap[PRC_NCMDS] = {
1321 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1322 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1323 EMSGSIZE, EHOSTUNREACH, 0, 0,
1324 0, 0, EHOSTUNREACH, 0,
1325 ENOPROTOOPT, ECONNREFUSED
1329 * Forward a packet. If some error occurs return the sender
1330 * an icmp packet. Note we can't always generate a meaningful
1331 * icmp message because icmp doesn't have a large enough repertoire
1332 * of codes and types.
1334 * If not forwarding, just drop the packet. This could be confusing
1335 * if ipforwarding was zero but some routing protocol was advancing
1336 * us as a gateway to somewhere. However, we must let the routing
1337 * protocol deal with that.
1339 * The srcrt parameter indicates whether the packet is being forwarded
1340 * via a source route.
1343 ip_forward(struct mbuf *m, int srcrt)
1345 struct ip *ip = mtod(m, struct ip *);
1346 struct in_ifaddr *ia;
1348 struct sockaddr_in *sin;
1349 struct in_addr dest;
1351 int error, type = 0, code = 0, mtu = 0;
1353 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1354 IPSTAT_INC(ips_cantforward);
1361 if (ip->ip_ttl <= IPTTLDEC) {
1362 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1370 bzero(&ro, sizeof(ro));
1371 sin = (struct sockaddr_in *)&ro.ro_dst;
1372 sin->sin_family = AF_INET;
1373 sin->sin_len = sizeof(*sin);
1374 sin->sin_addr = ip->ip_dst;
1376 rtalloc_mpath_fib(&ro,
1377 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
1380 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1382 if (ro.ro_rt != NULL) {
1383 ia = ifatoia(ro.ro_rt->rt_ifa);
1384 ifa_ref(&ia->ia_ifa);
1389 * 'ia' may be NULL if there is no route for this destination.
1390 * In case of IPsec, Don't discard it just yet, but pass it to
1391 * ip_output in case of outgoing IPsec policy.
1393 if (!srcrt && ia == NULL) {
1394 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1401 * Save the IP header and at most 8 bytes of the payload,
1402 * in case we need to generate an ICMP message to the src.
1404 * XXX this can be optimized a lot by saving the data in a local
1405 * buffer on the stack (72 bytes at most), and only allocating the
1406 * mbuf if really necessary. The vast majority of the packets
1407 * are forwarded without having to send an ICMP back (either
1408 * because unnecessary, or because rate limited), so we are
1409 * really we are wasting a lot of work here.
1411 * We don't use m_copy() because it might return a reference
1412 * to a shared cluster. Both this function and ip_output()
1413 * assume exclusive access to the IP header in `m', so any
1414 * data in a cluster may change before we reach icmp_error().
1416 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1417 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1419 * It's probably ok if the pkthdr dup fails (because
1420 * the deep copy of the tag chain failed), but for now
1421 * be conservative and just discard the copy since
1422 * code below may some day want the tags.
1427 if (mcopy != NULL) {
1428 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1429 mcopy->m_pkthdr.len = mcopy->m_len;
1430 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1436 ip->ip_ttl -= IPTTLDEC;
1442 * If forwarding packet using same interface that it came in on,
1443 * perhaps should send a redirect to sender to shortcut a hop.
1444 * Only send redirect if source is sending directly to us,
1445 * and if packet was not source routed (or has any options).
1446 * Also, don't send redirect if forwarding using a default route
1447 * or a route modified by a redirect.
1450 if (!srcrt && V_ipsendredirects &&
1451 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1456 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1457 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1458 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1459 u_long src = ntohl(ip->ip_src.s_addr);
1462 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1463 if (rt->rt_flags & RTF_GATEWAY)
1464 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1466 dest.s_addr = ip->ip_dst.s_addr;
1467 /* Router requirements says to only send host redirects */
1468 type = ICMP_REDIRECT;
1469 code = ICMP_REDIRECT_HOST;
1474 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1476 if (error == EMSGSIZE && ro.ro_rt)
1477 mtu = ro.ro_rt->rt_mtu;
1481 IPSTAT_INC(ips_cantforward);
1483 IPSTAT_INC(ips_forward);
1485 IPSTAT_INC(ips_redirectsent);
1490 ifa_free(&ia->ia_ifa);
1494 if (mcopy == NULL) {
1496 ifa_free(&ia->ia_ifa);
1502 case 0: /* forwarded, but need redirect */
1503 /* type, code set above */
1511 type = ICMP_UNREACH;
1512 code = ICMP_UNREACH_HOST;
1516 type = ICMP_UNREACH;
1517 code = ICMP_UNREACH_NEEDFRAG;
1521 * If IPsec is configured for this path,
1522 * override any possibly mtu value set by ip_output.
1524 mtu = ip_ipsec_mtu(mcopy, mtu);
1527 * If the MTU was set before make sure we are below the
1529 * If the MTU wasn't set before use the interface mtu or
1530 * fall back to the next smaller mtu step compared to the
1531 * current packet size.
1535 mtu = min(mtu, ia->ia_ifp->if_mtu);
1538 mtu = ia->ia_ifp->if_mtu;
1540 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1542 IPSTAT_INC(ips_cantfrag);
1547 * A router should not generate ICMP_SOURCEQUENCH as
1548 * required in RFC1812 Requirements for IP Version 4 Routers.
1549 * Source quench could be a big problem under DoS attacks,
1550 * or if the underlying interface is rate-limited.
1551 * Those who need source quench packets may re-enable them
1552 * via the net.inet.ip.sendsourcequench sysctl.
1554 if (V_ip_sendsourcequench == 0) {
1557 ifa_free(&ia->ia_ifa);
1560 type = ICMP_SOURCEQUENCH;
1565 case EACCES: /* ipfw denied packet */
1568 ifa_free(&ia->ia_ifa);
1572 ifa_free(&ia->ia_ifa);
1573 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1577 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1581 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1585 if (inp->inp_socket->so_options & SO_BINTIME) {
1586 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1587 SCM_BINTIME, SOL_SOCKET);
1589 mp = &(*mp)->m_next;
1591 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1594 bintime2timeval(&bt, &tv);
1595 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1596 SCM_TIMESTAMP, SOL_SOCKET);
1598 mp = &(*mp)->m_next;
1601 if (inp->inp_flags & INP_RECVDSTADDR) {
1602 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1603 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1605 mp = &(*mp)->m_next;
1607 if (inp->inp_flags & INP_RECVTTL) {
1608 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1609 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1611 mp = &(*mp)->m_next;
1615 * Moving these out of udp_input() made them even more broken
1616 * than they already were.
1618 /* options were tossed already */
1619 if (inp->inp_flags & INP_RECVOPTS) {
1620 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1621 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1623 mp = &(*mp)->m_next;
1625 /* ip_srcroute doesn't do what we want here, need to fix */
1626 if (inp->inp_flags & INP_RECVRETOPTS) {
1627 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1628 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1630 mp = &(*mp)->m_next;
1633 if (inp->inp_flags & INP_RECVIF) {
1636 struct sockaddr_dl sdl;
1639 struct sockaddr_dl *sdp;
1640 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1642 if ((ifp = m->m_pkthdr.rcvif) &&
1643 ifp->if_index && ifp->if_index <= V_if_index) {
1644 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1646 * Change our mind and don't try copy.
1648 if (sdp->sdl_family != AF_LINK ||
1649 sdp->sdl_len > sizeof(sdlbuf)) {
1652 bcopy(sdp, sdl2, sdp->sdl_len);
1656 offsetof(struct sockaddr_dl, sdl_data[0]);
1657 sdl2->sdl_family = AF_LINK;
1658 sdl2->sdl_index = 0;
1659 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1661 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1662 IP_RECVIF, IPPROTO_IP);
1664 mp = &(*mp)->m_next;
1666 if (inp->inp_flags & INP_RECVTOS) {
1667 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1668 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1670 mp = &(*mp)->m_next;
1675 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1676 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1677 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1680 static VNET_DEFINE(int, ip_rsvp_on);
1681 VNET_DEFINE(struct socket *, ip_rsvpd);
1683 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1686 ip_rsvp_init(struct socket *so)
1689 if (so->so_type != SOCK_RAW ||
1690 so->so_proto->pr_protocol != IPPROTO_RSVP)
1693 if (V_ip_rsvpd != NULL)
1698 * This may seem silly, but we need to be sure we don't over-increment
1699 * the RSVP counter, in case something slips up.
1701 if (!V_ip_rsvp_on) {
1715 * This may seem silly, but we need to be sure we don't over-decrement
1716 * the RSVP counter, in case something slips up.
1726 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1729 if (rsvp_input_p) { /* call the real one if loaded */
1730 rsvp_input_p(m, off);
1734 /* Can still get packets with rsvp_on = 0 if there is a local member
1735 * of the group to which the RSVP packet is addressed. But in this
1736 * case we want to throw the packet away.
1744 if (V_ip_rsvpd != NULL) {
1748 /* Drop the packet */