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"
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
49 #include <sys/kernel.h>
51 #include <sys/rwlock.h>
52 #include <sys/syslog.h>
53 #include <sys/sysctl.h>
57 #include <net/if_types.h>
58 #include <net/if_var.h>
59 #include <net/if_dl.h>
60 #include <net/route.h>
61 #include <net/netisr.h>
63 #include <net/flowtable.h>
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/in_var.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_fw.h>
72 #include <netinet/ip_icmp.h>
73 #include <netinet/ip_options.h>
74 #include <machine/in_cksum.h>
75 #include <netinet/ip_carp.h>
77 #include <netinet/ip_ipsec.h>
80 #include <sys/socketvar.h>
82 #include <security/mac/mac_framework.h>
85 CTASSERT(sizeof(struct ip) == 20);
88 struct rwlock in_ifaddr_lock;
89 RW_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
91 VNET_DEFINE(int, rsvp_on);
93 VNET_DEFINE(int, ipforwarding);
94 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
95 &VNET_NAME(ipforwarding), 0,
96 "Enable IP forwarding between interfaces");
98 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
99 #define V_ipsendredirects VNET(ipsendredirects)
100 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101 &VNET_NAME(ipsendredirects), 0,
102 "Enable sending IP redirects");
104 static VNET_DEFINE(int, ip_keepfaith);
105 #define V_ip_keepfaith VNET(ip_keepfaith)
106 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
107 &VNET_NAME(ip_keepfaith), 0,
108 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
110 static VNET_DEFINE(int, ip_sendsourcequench);
111 #define V_ip_sendsourcequench VNET(ip_sendsourcequench)
112 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
113 &VNET_NAME(ip_sendsourcequench), 0,
114 "Enable the transmission of source quench packets");
116 VNET_DEFINE(int, ip_do_randomid);
117 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
118 &VNET_NAME(ip_do_randomid), 0,
119 "Assign random ip_id values");
122 * XXX - Setting ip_checkinterface mostly implements the receive side of
123 * the Strong ES model described in RFC 1122, but since the routing table
124 * and transmit implementation do not implement the Strong ES model,
125 * setting this to 1 results in an odd hybrid.
127 * XXX - ip_checkinterface currently must be disabled if you use ipnat
128 * to translate the destination address to another local interface.
130 * XXX - ip_checkinterface must be disabled if you add IP aliases
131 * to the loopback interface instead of the interface where the
132 * packets for those addresses are received.
134 static VNET_DEFINE(int, ip_checkinterface);
135 #define V_ip_checkinterface VNET(ip_checkinterface)
136 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
137 &VNET_NAME(ip_checkinterface), 0,
138 "Verify packet arrives on correct interface");
140 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
142 static struct netisr_handler ip_nh = {
144 .nh_handler = ip_input,
145 .nh_proto = NETISR_IP,
146 .nh_policy = NETISR_POLICY_FLOW,
149 extern struct domain inetdomain;
150 extern struct protosw inetsw[];
151 u_char ip_protox[IPPROTO_MAX];
152 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
153 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
154 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
156 VNET_DEFINE(struct ipstat, ipstat);
157 SYSCTL_VNET_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
158 &VNET_NAME(ipstat), ipstat,
159 "IP statistics (struct ipstat, netinet/ip_var.h)");
161 static VNET_DEFINE(uma_zone_t, ipq_zone);
162 static VNET_DEFINE(TAILQ_HEAD(ipqhead, ipq), ipq[IPREASS_NHASH]);
163 static struct mtx ipqlock;
165 #define V_ipq_zone VNET(ipq_zone)
166 #define V_ipq VNET(ipq)
168 #define IPQ_LOCK() mtx_lock(&ipqlock)
169 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
170 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
171 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
173 static void maxnipq_update(void);
174 static void ipq_zone_change(void *);
175 static void ip_drain_locked(void);
177 static VNET_DEFINE(int, maxnipq); /* Administrative limit on # reass queues. */
178 static VNET_DEFINE(int, nipq); /* Total # of reass queues */
179 #define V_maxnipq VNET(maxnipq)
180 #define V_nipq VNET(nipq)
181 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
183 "Current number of IPv4 fragment reassembly queue entries");
185 static VNET_DEFINE(int, maxfragsperpacket);
186 #define V_maxfragsperpacket VNET(maxfragsperpacket)
187 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
188 &VNET_NAME(maxfragsperpacket), 0,
189 "Maximum number of IPv4 fragments allowed per packet");
192 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
193 &ip_mtu, 0, "Default MTU");
197 VNET_DEFINE(int, ipstealth);
198 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
199 &VNET_NAME(ipstealth), 0,
200 "IP stealth mode, no TTL decrementation on forwarding");
204 static VNET_DEFINE(int, ip_output_flowtable_size) = 2048;
205 VNET_DEFINE(struct flowtable *, ip_ft);
206 #define V_ip_output_flowtable_size VNET(ip_output_flowtable_size)
208 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, output_flowtable_size, CTLFLAG_RDTUN,
209 &VNET_NAME(ip_output_flowtable_size), 2048,
210 "number of entries in the per-cpu output flow caches");
213 static void ip_freef(struct ipqhead *, struct ipq *);
216 * Kernel module interface for updating ipstat. The argument is an index
217 * into ipstat treated as an array of u_long. While this encodes the general
218 * layout of ipstat into the caller, it doesn't encode its location, so that
219 * future changes to add, for example, per-CPU stats support won't cause
220 * binary compatibility problems for kernel modules.
223 kmod_ipstat_inc(int statnum)
226 (*((u_long *)&V_ipstat + statnum))++;
230 kmod_ipstat_dec(int statnum)
233 (*((u_long *)&V_ipstat + statnum))--;
237 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
241 netisr_getqlimit(&ip_nh, &qlimit);
242 error = sysctl_handle_int(oidp, &qlimit, 0, req);
243 if (error || !req->newptr)
247 return (netisr_setqlimit(&ip_nh, qlimit));
249 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
250 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
251 "Maximum size of the IP input queue");
254 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
256 u_int64_t qdrops_long;
259 netisr_getqdrops(&ip_nh, &qdrops_long);
260 qdrops = qdrops_long;
261 error = sysctl_handle_int(oidp, &qdrops, 0, req);
262 if (error || !req->newptr)
266 netisr_clearqdrops(&ip_nh);
270 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
271 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
272 "Number of packets dropped from the IP input queue");
275 * IP initialization: fill in IP protocol switch table.
276 * All protocols not implemented in kernel go to raw IP protocol handler.
284 V_ip_id = time_second & 0xffff;
286 TAILQ_INIT(&V_in_ifaddrhead);
287 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
289 /* Initialize IP reassembly queue. */
290 for (i = 0; i < IPREASS_NHASH; i++)
291 TAILQ_INIT(&V_ipq[i]);
292 V_maxnipq = nmbclusters / 32;
293 V_maxfragsperpacket = 16;
294 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
295 NULL, UMA_ALIGN_PTR, 0);
298 /* Initialize packet filter hooks. */
299 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
300 V_inet_pfil_hook.ph_af = AF_INET;
301 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
302 printf("%s: WARNING: unable to register pfil hook, "
303 "error %d\n", __func__, i);
306 if (TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size",
307 &V_ip_output_flowtable_size)) {
308 if (V_ip_output_flowtable_size < 256)
309 V_ip_output_flowtable_size = 256;
310 if (!powerof2(V_ip_output_flowtable_size)) {
311 printf("flowtable must be power of 2 size\n");
312 V_ip_output_flowtable_size = 2048;
316 * round up to the next power of 2
318 V_ip_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
320 V_ip_ft = flowtable_alloc("ipv4", V_ip_output_flowtable_size, FL_PCPU);
323 /* Skip initialization of globals for non-default instances. */
324 if (!IS_DEFAULT_VNET(curvnet))
327 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
329 panic("ip_init: PF_INET not found");
331 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
332 for (i = 0; i < IPPROTO_MAX; i++)
333 ip_protox[i] = pr - inetsw;
335 * Cycle through IP protocols and put them into the appropriate place
338 for (pr = inetdomain.dom_protosw;
339 pr < inetdomain.dom_protoswNPROTOSW; pr++)
340 if (pr->pr_domain->dom_family == PF_INET &&
341 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
342 /* Be careful to only index valid IP protocols. */
343 if (pr->pr_protocol < IPPROTO_MAX)
344 ip_protox[pr->pr_protocol] = pr - inetsw;
347 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
348 NULL, EVENTHANDLER_PRI_ANY);
350 /* Initialize various other remaining things. */
352 netisr_register(&ip_nh);
360 /* Cleanup in_ifaddr hash table; should be empty. */
361 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
367 uma_zdestroy(V_ipq_zone);
372 * Ip input routine. Checksum and byte swap header. If fragmented
373 * try to reassemble. Process options. Pass to next level.
376 ip_input(struct mbuf *m)
378 struct ip *ip = NULL;
379 struct in_ifaddr *ia = NULL;
382 int checkif, hlen = 0;
383 uint16_t sum, ip_len;
384 int dchg = 0; /* dest changed after fw */
385 struct in_addr odst; /* original dst address */
389 if (m->m_flags & M_FASTFWD_OURS) {
390 m->m_flags &= ~M_FASTFWD_OURS;
391 /* Set up some basics that will be used later. */
392 ip = mtod(m, struct ip *);
393 ip->ip_len = ntohs(ip->ip_len);
394 ip->ip_off = ntohs(ip->ip_off);
395 hlen = ip->ip_hl << 2;
399 IPSTAT_INC(ips_total);
401 if (m->m_pkthdr.len < sizeof(struct ip))
404 if (m->m_len < sizeof (struct ip) &&
405 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
406 IPSTAT_INC(ips_toosmall);
409 ip = mtod(m, struct ip *);
411 if (ip->ip_v != IPVERSION) {
412 IPSTAT_INC(ips_badvers);
416 hlen = ip->ip_hl << 2;
417 if (hlen < sizeof(struct ip)) { /* minimum header length */
418 IPSTAT_INC(ips_badhlen);
421 if (hlen > m->m_len) {
422 if ((m = m_pullup(m, hlen)) == NULL) {
423 IPSTAT_INC(ips_badhlen);
426 ip = mtod(m, struct ip *);
429 /* 127/8 must not appear on wire - RFC1122 */
430 ifp = m->m_pkthdr.rcvif;
431 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
432 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
433 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
434 IPSTAT_INC(ips_badaddr);
439 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
440 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
442 if (hlen == sizeof(struct ip)) {
443 sum = in_cksum_hdr(ip);
445 sum = in_cksum(m, hlen);
449 IPSTAT_INC(ips_badsum);
454 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
455 /* packet is dropped by traffic conditioner */
459 ip_len = ntohs(ip->ip_len);
461 IPSTAT_INC(ips_badlen);
466 * Check that the amount of data in the buffers
467 * is as at least much as the IP header would have us expect.
468 * Trim mbufs if longer than we expect.
469 * Drop packet if shorter than we expect.
471 if (m->m_pkthdr.len < ip_len) {
473 IPSTAT_INC(ips_tooshort);
476 if (m->m_pkthdr.len > ip_len) {
477 if (m->m_len == m->m_pkthdr.len) {
479 m->m_pkthdr.len = ip_len;
481 m_adj(m, ip_len - m->m_pkthdr.len);
485 * Bypass packet filtering for packets previously handled by IPsec.
487 if (ip_ipsec_filtertunnel(m))
492 * Run through list of hooks for input packets.
494 * NB: Beware of the destination address changing (e.g.
495 * by NAT rewriting). When this happens, tell
496 * ip_forward to do the right thing.
499 /* Jump over all PFIL processing if hooks are not active. */
500 if (!PFIL_HOOKED(&V_inet_pfil_hook))
504 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
506 if (m == NULL) /* consumed by filter */
509 ip = mtod(m, struct ip *);
510 dchg = (odst.s_addr != ip->ip_dst.s_addr);
511 ifp = m->m_pkthdr.rcvif;
513 #ifdef IPFIREWALL_FORWARD
514 if (m->m_flags & M_FASTFWD_OURS) {
515 m->m_flags &= ~M_FASTFWD_OURS;
516 ip->ip_len = ntohs(ip->ip_len);
517 ip->ip_off = ntohs(ip->ip_off);
520 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
522 * Directly ship the packet on. This allows forwarding
523 * packets originally destined to us to some other directly
526 ip->ip_len = ntohs(ip->ip_len);
527 ip->ip_off = ntohs(ip->ip_off);
531 #endif /* IPFIREWALL_FORWARD */
535 * From now and up to output pfil(9) processing in ip_output()
536 * the header is in host byte order.
538 ip->ip_len = ntohs(ip->ip_len);
539 ip->ip_off = ntohs(ip->ip_off);
542 * Process options and, if not destined for us,
543 * ship it on. ip_dooptions returns 1 when an
544 * error was detected (causing an icmp message
545 * to be sent and the original packet to be freed).
547 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
550 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
551 * matter if it is destined to another node, or whether it is
552 * a multicast one, RSVP wants it! and prevents it from being forwarded
553 * anywhere else. Also checks if the rsvp daemon is running before
554 * grabbing the packet.
556 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
560 * Check our list of addresses, to see if the packet is for us.
561 * If we don't have any addresses, assume any unicast packet
562 * we receive might be for us (and let the upper layers deal
565 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
566 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
570 * Enable a consistency check between the destination address
571 * and the arrival interface for a unicast packet (the RFC 1122
572 * strong ES model) if IP forwarding is disabled and the packet
573 * is not locally generated and the packet is not subject to
576 * XXX - Checking also should be disabled if the destination
577 * address is ipnat'ed to a different interface.
579 * XXX - Checking is incompatible with IP aliases added
580 * to the loopback interface instead of the interface where
581 * the packets are received.
583 * XXX - This is the case for carp vhost IPs as well so we
584 * insert a workaround. If the packet got here, we already
585 * checked with carp_iamatch() and carp_forus().
587 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
588 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
589 ifp->if_carp == NULL && (dchg == 0);
592 * Check for exact addresses in the hash bucket.
594 /* IN_IFADDR_RLOCK(); */
595 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
597 * If the address matches, verify that the packet
598 * arrived via the correct interface if checking is
601 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
602 (!checkif || ia->ia_ifp == ifp)) {
603 ifa_ref(&ia->ia_ifa);
604 /* IN_IFADDR_RUNLOCK(); */
608 /* IN_IFADDR_RUNLOCK(); */
611 * Check for broadcast addresses.
613 * Only accept broadcast packets that arrive via the matching
614 * interface. Reception of forwarded directed broadcasts would
615 * be handled via ip_forward() and ether_output() with the loopback
616 * into the stack for SIMPLEX interfaces handled by ether_output().
618 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
620 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
621 if (ifa->ifa_addr->sa_family != AF_INET)
624 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
627 IF_ADDR_RUNLOCK(ifp);
631 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
633 IF_ADDR_RUNLOCK(ifp);
638 IF_ADDR_RUNLOCK(ifp);
641 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
642 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
643 IPSTAT_INC(ips_cantforward);
647 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
650 * If we are acting as a multicast router, all
651 * incoming multicast packets are passed to the
652 * kernel-level multicast forwarding function.
653 * The packet is returned (relatively) intact; if
654 * ip_mforward() returns a non-zero value, the packet
655 * must be discarded, else it may be accepted below.
657 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
658 IPSTAT_INC(ips_cantforward);
664 * The process-level routing daemon needs to receive
665 * all multicast IGMP packets, whether or not this
666 * host belongs to their destination groups.
668 if (ip->ip_p == IPPROTO_IGMP)
670 IPSTAT_INC(ips_forward);
673 * Assume the packet is for us, to avoid prematurely taking
674 * a lock on the in_multi hash. Protocols must perform
675 * their own filtering and update statistics accordingly.
679 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
681 if (ip->ip_dst.s_addr == INADDR_ANY)
685 * FAITH(Firewall Aided Internet Translator)
687 if (ifp && ifp->if_type == IFT_FAITH) {
688 if (V_ip_keepfaith) {
689 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
697 * Not for us; forward if possible and desirable.
699 if (V_ipforwarding == 0) {
700 IPSTAT_INC(ips_cantforward);
714 * IPSTEALTH: Process non-routing options only
715 * if the packet is destined for us.
717 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) {
719 ifa_free(&ia->ia_ifa);
722 #endif /* IPSTEALTH */
724 /* Count the packet in the ip address stats */
726 ia->ia_ifa.if_ipackets++;
727 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
728 ifa_free(&ia->ia_ifa);
732 * Attempt reassembly; if it succeeds, proceed.
733 * ip_reass() will return a different mbuf.
735 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
739 ip = mtod(m, struct ip *);
740 /* Get the header length of the reassembled packet */
741 hlen = ip->ip_hl << 2;
745 * Further protocols expect the packet length to be w/o the
752 * enforce IPsec policy checking if we are seeing last header.
753 * note that we do not visit this with protocols with pcb layer
754 * code - like udp/tcp/raw ip.
756 if (ip_ipsec_input(m))
761 * Switch out to protocol's input routine.
763 IPSTAT_INC(ips_delivered);
765 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
772 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
773 * max has slightly different semantics than the sysctl, for historical
781 * -1 for unlimited allocation.
784 uma_zone_set_max(V_ipq_zone, 0);
786 * Positive number for specific bound.
789 uma_zone_set_max(V_ipq_zone, V_maxnipq);
791 * Zero specifies no further fragment queue allocation -- set the
792 * bound very low, but rely on implementation elsewhere to actually
793 * prevent allocation and reclaim current queues.
796 uma_zone_set_max(V_ipq_zone, 1);
800 ipq_zone_change(void *tag)
803 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
804 V_maxnipq = nmbclusters / 32;
810 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
815 error = sysctl_handle_int(oidp, &i, 0, req);
816 if (error || !req->newptr)
820 * XXXRW: Might be a good idea to sanity check the argument and place
821 * an extreme upper bound.
830 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
831 NULL, 0, sysctl_maxnipq, "I",
832 "Maximum number of IPv4 fragment reassembly queue entries");
835 * Take incoming datagram fragment and try to reassemble it into
836 * whole datagram. If the argument is the first fragment or one
837 * in between the function will return NULL and store the mbuf
838 * in the fragment chain. If the argument is the last fragment
839 * the packet will be reassembled and the pointer to the new
840 * mbuf returned for further processing. Only m_tags attached
841 * to the first packet/fragment are preserved.
842 * The IP header is *NOT* adjusted out of iplen.
845 ip_reass(struct mbuf *m)
848 struct mbuf *p, *q, *nq, *t;
849 struct ipq *fp = NULL;
850 struct ipqhead *head;
855 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
856 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
857 IPSTAT_INC(ips_fragments);
858 IPSTAT_INC(ips_fragdropped);
863 ip = mtod(m, struct ip *);
864 hlen = ip->ip_hl << 2;
866 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
871 * Look for queue of fragments
874 TAILQ_FOREACH(fp, head, ipq_list)
875 if (ip->ip_id == fp->ipq_id &&
876 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
877 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
879 mac_ipq_match(m, fp) &&
881 ip->ip_p == fp->ipq_p)
887 * Attempt to trim the number of allocated fragment queues if it
888 * exceeds the administrative limit.
890 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
892 * drop something from the tail of the current queue
893 * before proceeding further
895 struct ipq *q = TAILQ_LAST(head, ipqhead);
896 if (q == NULL) { /* gak */
897 for (i = 0; i < IPREASS_NHASH; i++) {
898 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
900 IPSTAT_ADD(ips_fragtimeout,
902 ip_freef(&V_ipq[i], r);
907 IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
914 * Adjust ip_len to not reflect header,
915 * convert offset of this to bytes.
918 if (ip->ip_off & IP_MF) {
920 * Make sure that fragments have a data length
921 * that's a non-zero multiple of 8 bytes.
923 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
924 IPSTAT_INC(ips_toosmall); /* XXX */
927 m->m_flags |= M_FRAG;
929 m->m_flags &= ~M_FRAG;
934 * Attempt reassembly; if it succeeds, proceed.
935 * ip_reass() will return a different mbuf.
937 IPSTAT_INC(ips_fragments);
938 m->m_pkthdr.header = ip;
940 /* Previous ip_reass() started here. */
942 * Presence of header sizes in mbufs
943 * would confuse code below.
949 * If first fragment to arrive, create a reassembly queue.
952 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
956 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
957 uma_zfree(V_ipq_zone, fp);
961 mac_ipq_create(m, fp);
963 TAILQ_INSERT_HEAD(head, fp, ipq_list);
966 fp->ipq_ttl = IPFRAGTTL;
967 fp->ipq_p = ip->ip_p;
968 fp->ipq_id = ip->ip_id;
969 fp->ipq_src = ip->ip_src;
970 fp->ipq_dst = ip->ip_dst;
977 mac_ipq_update(m, fp);
981 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
984 * Handle ECN by comparing this segment with the first one;
985 * if CE is set, do not lose CE.
986 * drop if CE and not-ECT are mixed for the same packet.
988 ecn = ip->ip_tos & IPTOS_ECN_MASK;
989 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
990 if (ecn == IPTOS_ECN_CE) {
991 if (ecn0 == IPTOS_ECN_NOTECT)
993 if (ecn0 != IPTOS_ECN_CE)
994 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
996 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1000 * Find a segment which begins after this one does.
1002 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1003 if (GETIP(q)->ip_off > ip->ip_off)
1007 * If there is a preceding segment, it may provide some of
1008 * our data already. If so, drop the data from the incoming
1009 * segment. If it provides all of our data, drop us, otherwise
1010 * stick new segment in the proper place.
1012 * If some of the data is dropped from the preceding
1013 * segment, then it's checksum is invalidated.
1016 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1018 if (i >= ip->ip_len)
1021 m->m_pkthdr.csum_flags = 0;
1025 m->m_nextpkt = p->m_nextpkt;
1028 m->m_nextpkt = fp->ipq_frags;
1033 * While we overlap succeeding segments trim them or,
1034 * if they are completely covered, dequeue them.
1036 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1038 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1039 if (i < GETIP(q)->ip_len) {
1040 GETIP(q)->ip_len -= i;
1041 GETIP(q)->ip_off += i;
1043 q->m_pkthdr.csum_flags = 0;
1048 IPSTAT_INC(ips_fragdropped);
1054 * Check for complete reassembly and perform frag per packet
1057 * Frag limiting is performed here so that the nth frag has
1058 * a chance to complete the packet before we drop the packet.
1059 * As a result, n+1 frags are actually allowed per packet, but
1060 * only n will ever be stored. (n = maxfragsperpacket.)
1064 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1065 if (GETIP(q)->ip_off != next) {
1066 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1067 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1072 next += GETIP(q)->ip_len;
1074 /* Make sure the last packet didn't have the IP_MF flag */
1075 if (p->m_flags & M_FRAG) {
1076 if (fp->ipq_nfrags > V_maxfragsperpacket) {
1077 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1084 * Reassembly is complete. Make sure the packet is a sane size.
1088 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1089 IPSTAT_INC(ips_toolong);
1090 IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1096 * Concatenate fragments.
1103 q->m_nextpkt = NULL;
1104 for (q = nq; q != NULL; q = nq) {
1106 q->m_nextpkt = NULL;
1107 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1108 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1112 * In order to do checksumming faster we do 'end-around carry' here
1113 * (and not in for{} loop), though it implies we are not going to
1114 * reassemble more than 64k fragments.
1116 m->m_pkthdr.csum_data =
1117 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1119 mac_ipq_reassemble(fp, m);
1120 mac_ipq_destroy(fp);
1124 * Create header for new ip packet by modifying header of first
1125 * packet; dequeue and discard fragment reassembly header.
1126 * Make header visible.
1128 ip->ip_len = (ip->ip_hl << 2) + next;
1129 ip->ip_src = fp->ipq_src;
1130 ip->ip_dst = fp->ipq_dst;
1131 TAILQ_REMOVE(head, fp, ipq_list);
1133 uma_zfree(V_ipq_zone, fp);
1134 m->m_len += (ip->ip_hl << 2);
1135 m->m_data -= (ip->ip_hl << 2);
1136 /* some debugging cruft by sklower, below, will go away soon */
1137 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1139 IPSTAT_INC(ips_reassembled);
1144 IPSTAT_INC(ips_fragdropped);
1156 * Free a fragment reassembly header and all
1157 * associated datagrams.
1160 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1166 while (fp->ipq_frags) {
1168 fp->ipq_frags = q->m_nextpkt;
1171 TAILQ_REMOVE(fhp, fp, ipq_list);
1172 uma_zfree(V_ipq_zone, fp);
1177 * IP timer processing;
1178 * if a timer expires on a reassembly
1179 * queue, discard it.
1184 VNET_ITERATOR_DECL(vnet_iter);
1188 VNET_LIST_RLOCK_NOSLEEP();
1190 VNET_FOREACH(vnet_iter) {
1191 CURVNET_SET(vnet_iter);
1192 for (i = 0; i < IPREASS_NHASH; i++) {
1193 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1197 fp = TAILQ_NEXT(fp, ipq_list);
1198 if(--fpp->ipq_ttl == 0) {
1199 IPSTAT_ADD(ips_fragtimeout,
1201 ip_freef(&V_ipq[i], fpp);
1206 * If we are over the maximum number of fragments
1207 * (due to the limit being lowered), drain off
1208 * enough to get down to the new limit.
1210 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1211 for (i = 0; i < IPREASS_NHASH; i++) {
1212 while (V_nipq > V_maxnipq &&
1213 !TAILQ_EMPTY(&V_ipq[i])) {
1214 IPSTAT_ADD(ips_fragdropped,
1215 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1217 TAILQ_FIRST(&V_ipq[i]));
1224 VNET_LIST_RUNLOCK_NOSLEEP();
1228 * Drain off all datagram fragments.
1231 ip_drain_locked(void)
1237 for (i = 0; i < IPREASS_NHASH; i++) {
1238 while(!TAILQ_EMPTY(&V_ipq[i])) {
1239 IPSTAT_ADD(ips_fragdropped,
1240 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1241 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1249 VNET_ITERATOR_DECL(vnet_iter);
1251 VNET_LIST_RLOCK_NOSLEEP();
1253 VNET_FOREACH(vnet_iter) {
1254 CURVNET_SET(vnet_iter);
1259 VNET_LIST_RUNLOCK_NOSLEEP();
1264 * The protocol to be inserted into ip_protox[] must be already registered
1265 * in inetsw[], either statically or through pf_proto_register().
1268 ipproto_register(short ipproto)
1272 /* Sanity checks. */
1273 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1274 return (EPROTONOSUPPORT);
1277 * The protocol slot must not be occupied by another protocol
1278 * already. An index pointing to IPPROTO_RAW is unused.
1280 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1282 return (EPFNOSUPPORT);
1283 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1286 /* Find the protocol position in inetsw[] and set the index. */
1287 for (pr = inetdomain.dom_protosw;
1288 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1289 if (pr->pr_domain->dom_family == PF_INET &&
1290 pr->pr_protocol && pr->pr_protocol == ipproto) {
1291 ip_protox[pr->pr_protocol] = pr - inetsw;
1295 return (EPROTONOSUPPORT);
1299 ipproto_unregister(short ipproto)
1303 /* Sanity checks. */
1304 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
1305 return (EPROTONOSUPPORT);
1307 /* Check if the protocol was indeed registered. */
1308 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1310 return (EPFNOSUPPORT);
1311 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1314 /* Reset the protocol slot to IPPROTO_RAW. */
1315 ip_protox[ipproto] = pr - inetsw;
1320 * Given address of next destination (final or next hop), return (referenced)
1321 * internet address info of interface to be used to get there.
1324 ip_rtaddr(struct in_addr dst, u_int fibnum)
1327 struct sockaddr_in *sin;
1328 struct in_ifaddr *ia;
1330 bzero(&sro, sizeof(sro));
1331 sin = (struct sockaddr_in *)&sro.ro_dst;
1332 sin->sin_family = AF_INET;
1333 sin->sin_len = sizeof(*sin);
1334 sin->sin_addr = dst;
1335 in_rtalloc_ign(&sro, 0, fibnum);
1337 if (sro.ro_rt == NULL)
1340 ia = ifatoia(sro.ro_rt->rt_ifa);
1341 ifa_ref(&ia->ia_ifa);
1346 u_char inetctlerrmap[PRC_NCMDS] = {
1348 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1349 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1350 EMSGSIZE, EHOSTUNREACH, 0, 0,
1351 0, 0, EHOSTUNREACH, 0,
1352 ENOPROTOOPT, ECONNREFUSED
1356 * Forward a packet. If some error occurs return the sender
1357 * an icmp packet. Note we can't always generate a meaningful
1358 * icmp message because icmp doesn't have a large enough repertoire
1359 * of codes and types.
1361 * If not forwarding, just drop the packet. This could be confusing
1362 * if ipforwarding was zero but some routing protocol was advancing
1363 * us as a gateway to somewhere. However, we must let the routing
1364 * protocol deal with that.
1366 * The srcrt parameter indicates whether the packet is being forwarded
1367 * via a source route.
1369 * IP header in host byte order.
1372 ip_forward(struct mbuf *m, int srcrt)
1374 struct ip *ip = mtod(m, struct ip *);
1375 struct in_ifaddr *ia;
1377 struct in_addr dest;
1379 int error, type = 0, code = 0, mtu = 0;
1381 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1382 IPSTAT_INC(ips_cantforward);
1389 if (ip->ip_ttl <= IPTTLDEC) {
1390 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1398 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1401 * 'ia' may be NULL if there is no route for this destination.
1402 * In case of IPsec, Don't discard it just yet, but pass it to
1403 * ip_output in case of outgoing IPsec policy.
1405 if (!srcrt && ia == NULL) {
1406 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1412 * Save the IP header and at most 8 bytes of the payload,
1413 * in case we need to generate an ICMP message to the src.
1415 * XXX this can be optimized a lot by saving the data in a local
1416 * buffer on the stack (72 bytes at most), and only allocating the
1417 * mbuf if really necessary. The vast majority of the packets
1418 * are forwarded without having to send an ICMP back (either
1419 * because unnecessary, or because rate limited), so we are
1420 * really we are wasting a lot of work here.
1422 * We don't use m_copy() because it might return a reference
1423 * to a shared cluster. Both this function and ip_output()
1424 * assume exclusive access to the IP header in `m', so any
1425 * data in a cluster may change before we reach icmp_error().
1427 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1428 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1430 * It's probably ok if the pkthdr dup fails (because
1431 * the deep copy of the tag chain failed), but for now
1432 * be conservative and just discard the copy since
1433 * code below may some day want the tags.
1438 if (mcopy != NULL) {
1439 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1440 mcopy->m_pkthdr.len = mcopy->m_len;
1441 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1447 ip->ip_ttl -= IPTTLDEC;
1453 * If forwarding packet using same interface that it came in on,
1454 * perhaps should send a redirect to sender to shortcut a hop.
1455 * Only send redirect if source is sending directly to us,
1456 * and if packet was not source routed (or has any options).
1457 * Also, don't send redirect if forwarding using a default route
1458 * or a route modified by a redirect.
1461 if (!srcrt && V_ipsendredirects &&
1462 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1463 struct sockaddr_in *sin;
1466 bzero(&ro, sizeof(ro));
1467 sin = (struct sockaddr_in *)&ro.ro_dst;
1468 sin->sin_family = AF_INET;
1469 sin->sin_len = sizeof(*sin);
1470 sin->sin_addr = ip->ip_dst;
1471 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1475 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1476 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1477 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1478 u_long src = ntohl(ip->ip_src.s_addr);
1481 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1482 if (rt->rt_flags & RTF_GATEWAY)
1483 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1485 dest.s_addr = ip->ip_dst.s_addr;
1486 /* Router requirements says to only send host redirects */
1487 type = ICMP_REDIRECT;
1488 code = ICMP_REDIRECT_HOST;
1496 * Try to cache the route MTU from ip_output so we can consider it for
1497 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1499 bzero(&ro, sizeof(ro));
1501 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1503 if (error == EMSGSIZE && ro.ro_rt)
1504 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1508 IPSTAT_INC(ips_cantforward);
1510 IPSTAT_INC(ips_forward);
1512 IPSTAT_INC(ips_redirectsent);
1517 ifa_free(&ia->ia_ifa);
1521 if (mcopy == NULL) {
1523 ifa_free(&ia->ia_ifa);
1529 case 0: /* forwarded, but need redirect */
1530 /* type, code set above */
1538 type = ICMP_UNREACH;
1539 code = ICMP_UNREACH_HOST;
1543 type = ICMP_UNREACH;
1544 code = ICMP_UNREACH_NEEDFRAG;
1548 * If IPsec is configured for this path,
1549 * override any possibly mtu value set by ip_output.
1551 mtu = ip_ipsec_mtu(mcopy, mtu);
1554 * If the MTU was set before make sure we are below the
1556 * If the MTU wasn't set before use the interface mtu or
1557 * fall back to the next smaller mtu step compared to the
1558 * current packet size.
1562 mtu = min(mtu, ia->ia_ifp->if_mtu);
1565 mtu = ia->ia_ifp->if_mtu;
1567 mtu = ip_next_mtu(ip->ip_len, 0);
1569 IPSTAT_INC(ips_cantfrag);
1574 * A router should not generate ICMP_SOURCEQUENCH as
1575 * required in RFC1812 Requirements for IP Version 4 Routers.
1576 * Source quench could be a big problem under DoS attacks,
1577 * or if the underlying interface is rate-limited.
1578 * Those who need source quench packets may re-enable them
1579 * via the net.inet.ip.sendsourcequench sysctl.
1581 if (V_ip_sendsourcequench == 0) {
1584 ifa_free(&ia->ia_ifa);
1587 type = ICMP_SOURCEQUENCH;
1592 case EACCES: /* ipfw denied packet */
1595 ifa_free(&ia->ia_ifa);
1599 ifa_free(&ia->ia_ifa);
1600 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1604 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1608 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1612 if (inp->inp_socket->so_options & SO_BINTIME) {
1613 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1614 SCM_BINTIME, SOL_SOCKET);
1616 mp = &(*mp)->m_next;
1618 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1621 bintime2timeval(&bt, &tv);
1622 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1623 SCM_TIMESTAMP, SOL_SOCKET);
1625 mp = &(*mp)->m_next;
1628 if (inp->inp_flags & INP_RECVDSTADDR) {
1629 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1630 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1632 mp = &(*mp)->m_next;
1634 if (inp->inp_flags & INP_RECVTTL) {
1635 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1636 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1638 mp = &(*mp)->m_next;
1642 * Moving these out of udp_input() made them even more broken
1643 * than they already were.
1645 /* options were tossed already */
1646 if (inp->inp_flags & INP_RECVOPTS) {
1647 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1648 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1650 mp = &(*mp)->m_next;
1652 /* ip_srcroute doesn't do what we want here, need to fix */
1653 if (inp->inp_flags & INP_RECVRETOPTS) {
1654 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1655 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1657 mp = &(*mp)->m_next;
1660 if (inp->inp_flags & INP_RECVIF) {
1663 struct sockaddr_dl sdl;
1666 struct sockaddr_dl *sdp;
1667 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1669 if (((ifp = m->m_pkthdr.rcvif))
1670 && ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1671 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1673 * Change our mind and don't try copy.
1675 if ((sdp->sdl_family != AF_LINK)
1676 || (sdp->sdl_len > sizeof(sdlbuf))) {
1679 bcopy(sdp, sdl2, sdp->sdl_len);
1683 = offsetof(struct sockaddr_dl, sdl_data[0]);
1684 sdl2->sdl_family = AF_LINK;
1685 sdl2->sdl_index = 0;
1686 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1688 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1689 IP_RECVIF, IPPROTO_IP);
1691 mp = &(*mp)->m_next;
1693 if (inp->inp_flags & INP_RECVTOS) {
1694 *mp = sbcreatecontrol((caddr_t) &ip->ip_tos,
1695 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1697 mp = &(*mp)->m_next;
1702 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1703 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1704 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1707 static VNET_DEFINE(int, ip_rsvp_on);
1708 VNET_DEFINE(struct socket *, ip_rsvpd);
1710 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1713 ip_rsvp_init(struct socket *so)
1716 if (so->so_type != SOCK_RAW ||
1717 so->so_proto->pr_protocol != IPPROTO_RSVP)
1720 if (V_ip_rsvpd != NULL)
1725 * This may seem silly, but we need to be sure we don't over-increment
1726 * the RSVP counter, in case something slips up.
1728 if (!V_ip_rsvp_on) {
1742 * This may seem silly, but we need to be sure we don't over-decrement
1743 * the RSVP counter, in case something slips up.
1753 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1756 if (rsvp_input_p) { /* call the real one if loaded */
1757 rsvp_input_p(m, off);
1761 /* Can still get packets with rsvp_on = 0 if there is a local member
1762 * of the group to which the RSVP packet is addressed. But in this
1763 * case we want to throw the packet away.
1771 if (V_ip_rsvpd != NULL) {
1775 /* Drop the packet */