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>
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 <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/ip_icmp.h>
70 #include <netinet/ip_options.h>
71 #include <machine/in_cksum.h>
73 #include <netinet/ip_carp.h>
76 #include <netinet/ip_ipsec.h>
79 #include <sys/socketvar.h>
81 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
82 #include <netinet/ip_fw.h>
83 #include <netinet/ip_dummynet.h>
85 #include <security/mac/mac_framework.h>
90 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
91 &ipforwarding, 0, "Enable IP forwarding between interfaces");
93 static int ipsendredirects = 1; /* XXX */
94 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
95 &ipsendredirects, 0, "Enable sending IP redirects");
97 int ip_defttl = IPDEFTTL;
98 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
99 &ip_defttl, 0, "Maximum TTL on IP packets");
101 static int ip_keepfaith = 0;
102 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
104 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
106 static int ip_sendsourcequench = 0;
107 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
108 &ip_sendsourcequench, 0,
109 "Enable the transmission of source quench packets");
111 int ip_do_randomid = 0;
112 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
114 "Assign random ip_id values");
117 * XXX - Setting ip_checkinterface mostly implements the receive side of
118 * the Strong ES model described in RFC 1122, but since the routing table
119 * and transmit implementation do not implement the Strong ES model,
120 * setting this to 1 results in an odd hybrid.
122 * XXX - ip_checkinterface currently must be disabled if you use ipnat
123 * to translate the destination address to another local interface.
125 * XXX - ip_checkinterface must be disabled if you add IP aliases
126 * to the loopback interface instead of the interface where the
127 * packets for those addresses are received.
129 static int ip_checkinterface = 0;
130 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
131 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
133 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
135 static struct ifqueue ipintrq;
136 static int ipqmaxlen = IFQ_MAXLEN;
138 extern struct domain inetdomain;
139 extern struct protosw inetsw[];
140 u_char ip_protox[IPPROTO_MAX];
141 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
142 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
143 u_long in_ifaddrhmask; /* mask for hash table */
145 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
146 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
147 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
148 &ipintrq.ifq_drops, 0,
149 "Number of packets dropped from the IP input queue");
151 struct ipstat ipstat;
152 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
153 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
156 * IP datagram reassembly.
158 #define IPREASS_NHASH_LOG2 6
159 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
160 #define IPREASS_HMASK (IPREASS_NHASH - 1)
161 #define IPREASS_HASH(x,y) \
162 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
164 static uma_zone_t ipq_zone;
165 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
166 static struct mtx ipqlock;
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 *);
176 static int maxnipq; /* Administrative limit on # reass queues. */
177 static int nipq = 0; /* Total # of reass queues */
178 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
179 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
181 static int maxfragsperpacket;
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
183 &maxfragsperpacket, 0,
184 "Maximum number of IPv4 fragments allowed per packet");
186 struct callout ipport_tick_callout;
189 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
190 &ip_mtu, 0, "Default MTU");
195 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
196 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
200 * ipfw_ether and ipfw_bridge hooks.
201 * XXX: Temporary until those are converted to pfil_hooks as well.
203 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
204 ip_dn_io_t *ip_dn_io_ptr = NULL;
207 static void ip_freef(struct ipqhead *, struct ipq *);
210 * IP initialization: fill in IP protocol switch table.
211 * All protocols not implemented in kernel go to raw IP protocol handler.
219 TAILQ_INIT(&in_ifaddrhead);
220 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
221 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
223 panic("ip_init: PF_INET not found");
225 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
226 for (i = 0; i < IPPROTO_MAX; i++)
227 ip_protox[i] = pr - inetsw;
229 * Cycle through IP protocols and put them into the appropriate place
232 for (pr = inetdomain.dom_protosw;
233 pr < inetdomain.dom_protoswNPROTOSW; pr++)
234 if (pr->pr_domain->dom_family == PF_INET &&
235 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
236 /* Be careful to only index valid IP protocols. */
237 if (pr->pr_protocol < IPPROTO_MAX)
238 ip_protox[pr->pr_protocol] = pr - inetsw;
241 /* Initialize packet filter hooks. */
242 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
243 inet_pfil_hook.ph_af = AF_INET;
244 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
245 printf("%s: WARNING: unable to register pfil hook, "
246 "error %d\n", __func__, i);
248 /* Initialize IP reassembly queue. */
250 for (i = 0; i < IPREASS_NHASH; i++)
252 maxnipq = nmbclusters / 32;
253 maxfragsperpacket = 16;
254 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
255 NULL, UMA_ALIGN_PTR, 0);
258 /* Start ipport_tick. */
259 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
261 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
262 SHUTDOWN_PRI_DEFAULT);
263 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
264 NULL, EVENTHANDLER_PRI_ANY);
266 /* Initialize various other remaining things. */
267 ip_id = time_second & 0xffff;
268 ipintrq.ifq_maxlen = ipqmaxlen;
269 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
270 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
277 callout_stop(&ipport_tick_callout);
281 * Ip input routine. Checksum and byte swap header. If fragmented
282 * try to reassemble. Process options. Pass to next level.
285 ip_input(struct mbuf *m)
287 struct ip *ip = NULL;
288 struct in_ifaddr *ia = NULL;
290 int checkif, hlen = 0;
292 int dchg = 0; /* dest changed after fw */
293 struct in_addr odst; /* original dst address */
297 if (m->m_flags & M_FASTFWD_OURS) {
299 * Firewall or NAT changed destination to local.
300 * We expect ip_len and ip_off to be in host byte order.
302 m->m_flags &= ~M_FASTFWD_OURS;
303 /* Set up some basics that will be used later. */
304 ip = mtod(m, struct ip *);
305 hlen = ip->ip_hl << 2;
311 if (m->m_pkthdr.len < sizeof(struct ip))
314 if (m->m_len < sizeof (struct ip) &&
315 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
316 ipstat.ips_toosmall++;
319 ip = mtod(m, struct ip *);
321 if (ip->ip_v != IPVERSION) {
322 ipstat.ips_badvers++;
326 hlen = ip->ip_hl << 2;
327 if (hlen < sizeof(struct ip)) { /* minimum header length */
328 ipstat.ips_badhlen++;
331 if (hlen > m->m_len) {
332 if ((m = m_pullup(m, hlen)) == NULL) {
333 ipstat.ips_badhlen++;
336 ip = mtod(m, struct ip *);
339 /* 127/8 must not appear on wire - RFC1122 */
340 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
341 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
342 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
343 ipstat.ips_badaddr++;
348 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
349 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
351 if (hlen == sizeof(struct ip)) {
352 sum = in_cksum_hdr(ip);
354 sum = in_cksum(m, hlen);
363 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
364 /* packet is dropped by traffic conditioner */
369 * Convert fields to host representation.
371 ip->ip_len = ntohs(ip->ip_len);
372 if (ip->ip_len < hlen) {
376 ip->ip_off = ntohs(ip->ip_off);
379 * Check that the amount of data in the buffers
380 * is as at least much as the IP header would have us expect.
381 * Trim mbufs if longer than we expect.
382 * Drop packet if shorter than we expect.
384 if (m->m_pkthdr.len < ip->ip_len) {
386 ipstat.ips_tooshort++;
389 if (m->m_pkthdr.len > ip->ip_len) {
390 if (m->m_len == m->m_pkthdr.len) {
391 m->m_len = ip->ip_len;
392 m->m_pkthdr.len = ip->ip_len;
394 m_adj(m, ip->ip_len - m->m_pkthdr.len);
398 * Bypass packet filtering for packets from a tunnel (gif).
400 if (ip_ipsec_filtertunnel(m))
405 * Run through list of hooks for input packets.
407 * NB: Beware of the destination address changing (e.g.
408 * by NAT rewriting). When this happens, tell
409 * ip_forward to do the right thing.
412 /* Jump over all PFIL processing if hooks are not active. */
413 if (!PFIL_HOOKED(&inet_pfil_hook))
417 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
420 if (m == NULL) /* consumed by filter */
423 ip = mtod(m, struct ip *);
424 dchg = (odst.s_addr != ip->ip_dst.s_addr);
426 #ifdef IPFIREWALL_FORWARD
427 if (m->m_flags & M_FASTFWD_OURS) {
428 m->m_flags &= ~M_FASTFWD_OURS;
431 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
433 * Directly ship on the packet. This allows to forward packets
434 * that were destined for us to some other directly connected
440 #endif /* IPFIREWALL_FORWARD */
444 * Process options and, if not destined for us,
445 * ship it on. ip_dooptions returns 1 when an
446 * error was detected (causing an icmp message
447 * to be sent and the original packet to be freed).
449 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
452 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
453 * matter if it is destined to another node, or whether it is
454 * a multicast one, RSVP wants it! and prevents it from being forwarded
455 * anywhere else. Also checks if the rsvp daemon is running before
456 * grabbing the packet.
458 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
462 * Check our list of addresses, to see if the packet is for us.
463 * If we don't have any addresses, assume any unicast packet
464 * we receive might be for us (and let the upper layers deal
467 if (TAILQ_EMPTY(&in_ifaddrhead) &&
468 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
472 * Enable a consistency check between the destination address
473 * and the arrival interface for a unicast packet (the RFC 1122
474 * strong ES model) if IP forwarding is disabled and the packet
475 * is not locally generated and the packet is not subject to
478 * XXX - Checking also should be disabled if the destination
479 * address is ipnat'ed to a different interface.
481 * XXX - Checking is incompatible with IP aliases added
482 * to the loopback interface instead of the interface where
483 * the packets are received.
485 * XXX - This is the case for carp vhost IPs as well so we
486 * insert a workaround. If the packet got here, we already
487 * checked with carp_iamatch() and carp_forus().
489 checkif = ip_checkinterface && (ipforwarding == 0) &&
490 m->m_pkthdr.rcvif != NULL &&
491 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
493 !m->m_pkthdr.rcvif->if_carp &&
498 * Check for exact addresses in the hash bucket.
500 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
502 * If the address matches, verify that the packet
503 * arrived via the correct interface if checking is
506 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
507 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
511 * Check for broadcast addresses.
513 * Only accept broadcast packets that arrive via the matching
514 * interface. Reception of forwarded directed broadcasts would
515 * be handled via ip_forward() and ether_output() with the loopback
516 * into the stack for SIMPLEX interfaces handled by ether_output().
518 if (m->m_pkthdr.rcvif != NULL &&
519 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
520 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
521 if (ifa->ifa_addr->sa_family != AF_INET)
524 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
527 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
530 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
535 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
536 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
537 ipstat.ips_cantforward++;
541 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
542 struct in_multi *inm;
545 * If we are acting as a multicast router, all
546 * incoming multicast packets are passed to the
547 * kernel-level multicast forwarding function.
548 * The packet is returned (relatively) intact; if
549 * ip_mforward() returns a non-zero value, the packet
550 * must be discarded, else it may be accepted below.
553 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
554 ipstat.ips_cantforward++;
560 * The process-level routing daemon needs to receive
561 * all multicast IGMP packets, whether or not this
562 * host belongs to their destination groups.
564 if (ip->ip_p == IPPROTO_IGMP)
566 ipstat.ips_forward++;
569 * See if we belong to the destination multicast group on the
573 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
576 ipstat.ips_notmember++;
582 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
584 if (ip->ip_dst.s_addr == INADDR_ANY)
588 * FAITH(Firewall Aided Internet Translator)
590 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
592 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
600 * Not for us; forward if possible and desirable.
602 if (ipforwarding == 0) {
603 ipstat.ips_cantforward++;
617 * IPSTEALTH: Process non-routing options only
618 * if the packet is destined for us.
620 if (ipstealth && hlen > sizeof (struct ip) &&
623 #endif /* IPSTEALTH */
625 /* Count the packet in the ip address stats */
627 ia->ia_ifa.if_ipackets++;
628 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
632 * Attempt reassembly; if it succeeds, proceed.
633 * ip_reass() will return a different mbuf.
635 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
639 ip = mtod(m, struct ip *);
640 /* Get the header length of the reassembled packet */
641 hlen = ip->ip_hl << 2;
645 * Further protocols expect the packet length to be w/o the
652 * enforce IPsec policy checking if we are seeing last header.
653 * note that we do not visit this with protocols with pcb layer
654 * code - like udp/tcp/raw ip.
656 if (ip_ipsec_input(m))
661 * Switch out to protocol's input routine.
663 ipstat.ips_delivered++;
665 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
672 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
673 * max has slightly different semantics than the sysctl, for historical
681 * -1 for unlimited allocation.
684 uma_zone_set_max(ipq_zone, 0);
686 * Positive number for specific bound.
689 uma_zone_set_max(ipq_zone, maxnipq);
691 * Zero specifies no further fragment queue allocation -- set the
692 * bound very low, but rely on implementation elsewhere to actually
693 * prevent allocation and reclaim current queues.
696 uma_zone_set_max(ipq_zone, 1);
700 ipq_zone_change(void *tag)
703 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
704 maxnipq = nmbclusters / 32;
710 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
715 error = sysctl_handle_int(oidp, &i, 0, req);
716 if (error || !req->newptr)
720 * XXXRW: Might be a good idea to sanity check the argument and place
721 * an extreme upper bound.
730 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
731 NULL, 0, sysctl_maxnipq, "I",
732 "Maximum number of IPv4 fragment reassembly queue entries");
735 * Take incoming datagram fragment and try to reassemble it into
736 * whole datagram. If the argument is the first fragment or one
737 * in between the function will return NULL and store the mbuf
738 * in the fragment chain. If the argument is the last fragment
739 * the packet will be reassembled and the pointer to the new
740 * mbuf returned for further processing. Only m_tags attached
741 * to the first packet/fragment are preserved.
742 * The IP header is *NOT* adjusted out of iplen.
745 ip_reass(struct mbuf *m)
748 struct mbuf *p, *q, *nq, *t;
749 struct ipq *fp = NULL;
750 struct ipqhead *head;
755 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
756 if (maxnipq == 0 || maxfragsperpacket == 0) {
757 ipstat.ips_fragments++;
758 ipstat.ips_fragdropped++;
763 ip = mtod(m, struct ip *);
764 hlen = ip->ip_hl << 2;
766 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
771 * Look for queue of fragments
774 TAILQ_FOREACH(fp, head, ipq_list)
775 if (ip->ip_id == fp->ipq_id &&
776 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
777 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
779 mac_ipq_match(m, fp) &&
781 ip->ip_p == fp->ipq_p)
787 * Attempt to trim the number of allocated fragment queues if it
788 * exceeds the administrative limit.
790 if ((nipq > maxnipq) && (maxnipq > 0)) {
792 * drop something from the tail of the current queue
793 * before proceeding further
795 struct ipq *q = TAILQ_LAST(head, ipqhead);
796 if (q == NULL) { /* gak */
797 for (i = 0; i < IPREASS_NHASH; i++) {
798 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
800 ipstat.ips_fragtimeout += r->ipq_nfrags;
801 ip_freef(&ipq[i], r);
806 ipstat.ips_fragtimeout += q->ipq_nfrags;
813 * Adjust ip_len to not reflect header,
814 * convert offset of this to bytes.
817 if (ip->ip_off & IP_MF) {
819 * Make sure that fragments have a data length
820 * that's a non-zero multiple of 8 bytes.
822 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
823 ipstat.ips_toosmall++; /* XXX */
826 m->m_flags |= M_FRAG;
828 m->m_flags &= ~M_FRAG;
833 * Attempt reassembly; if it succeeds, proceed.
834 * ip_reass() will return a different mbuf.
836 ipstat.ips_fragments++;
837 m->m_pkthdr.header = ip;
839 /* Previous ip_reass() started here. */
841 * Presence of header sizes in mbufs
842 * would confuse code below.
848 * If first fragment to arrive, create a reassembly queue.
851 fp = uma_zalloc(ipq_zone, M_NOWAIT);
855 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
856 uma_zfree(ipq_zone, fp);
860 mac_ipq_create(m, fp);
862 TAILQ_INSERT_HEAD(head, fp, ipq_list);
865 fp->ipq_ttl = IPFRAGTTL;
866 fp->ipq_p = ip->ip_p;
867 fp->ipq_id = ip->ip_id;
868 fp->ipq_src = ip->ip_src;
869 fp->ipq_dst = ip->ip_dst;
876 mac_ipq_update(m, fp);
880 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
883 * Handle ECN by comparing this segment with the first one;
884 * if CE is set, do not lose CE.
885 * drop if CE and not-ECT are mixed for the same packet.
887 ecn = ip->ip_tos & IPTOS_ECN_MASK;
888 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
889 if (ecn == IPTOS_ECN_CE) {
890 if (ecn0 == IPTOS_ECN_NOTECT)
892 if (ecn0 != IPTOS_ECN_CE)
893 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
895 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
899 * Find a segment which begins after this one does.
901 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
902 if (GETIP(q)->ip_off > ip->ip_off)
906 * If there is a preceding segment, it may provide some of
907 * our data already. If so, drop the data from the incoming
908 * segment. If it provides all of our data, drop us, otherwise
909 * stick new segment in the proper place.
911 * If some of the data is dropped from the the preceding
912 * segment, then it's checksum is invalidated.
915 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
920 m->m_pkthdr.csum_flags = 0;
924 m->m_nextpkt = p->m_nextpkt;
927 m->m_nextpkt = fp->ipq_frags;
932 * While we overlap succeeding segments trim them or,
933 * if they are completely covered, dequeue them.
935 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
937 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
938 if (i < GETIP(q)->ip_len) {
939 GETIP(q)->ip_len -= i;
940 GETIP(q)->ip_off += i;
942 q->m_pkthdr.csum_flags = 0;
947 ipstat.ips_fragdropped++;
953 * Check for complete reassembly and perform frag per packet
956 * Frag limiting is performed here so that the nth frag has
957 * a chance to complete the packet before we drop the packet.
958 * As a result, n+1 frags are actually allowed per packet, but
959 * only n will ever be stored. (n = maxfragsperpacket.)
963 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
964 if (GETIP(q)->ip_off != next) {
965 if (fp->ipq_nfrags > maxfragsperpacket) {
966 ipstat.ips_fragdropped += fp->ipq_nfrags;
971 next += GETIP(q)->ip_len;
973 /* Make sure the last packet didn't have the IP_MF flag */
974 if (p->m_flags & M_FRAG) {
975 if (fp->ipq_nfrags > maxfragsperpacket) {
976 ipstat.ips_fragdropped += fp->ipq_nfrags;
983 * Reassembly is complete. Make sure the packet is a sane size.
987 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
988 ipstat.ips_toolong++;
989 ipstat.ips_fragdropped += fp->ipq_nfrags;
995 * Concatenate fragments.
1002 q->m_nextpkt = NULL;
1003 for (q = nq; q != NULL; q = nq) {
1005 q->m_nextpkt = NULL;
1006 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1007 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1011 * In order to do checksumming faster we do 'end-around carry' here
1012 * (and not in for{} loop), though it implies we are not going to
1013 * reassemble more than 64k fragments.
1015 m->m_pkthdr.csum_data =
1016 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1018 mac_ipq_reassemble(fp, m);
1019 mac_ipq_destroy(fp);
1023 * Create header for new ip packet by modifying header of first
1024 * packet; dequeue and discard fragment reassembly header.
1025 * Make header visible.
1027 ip->ip_len = (ip->ip_hl << 2) + next;
1028 ip->ip_src = fp->ipq_src;
1029 ip->ip_dst = fp->ipq_dst;
1030 TAILQ_REMOVE(head, fp, ipq_list);
1032 uma_zfree(ipq_zone, fp);
1033 m->m_len += (ip->ip_hl << 2);
1034 m->m_data -= (ip->ip_hl << 2);
1035 /* some debugging cruft by sklower, below, will go away soon */
1036 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1038 ipstat.ips_reassembled++;
1043 ipstat.ips_fragdropped++;
1055 * Free a fragment reassembly header and all
1056 * associated datagrams.
1059 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1065 while (fp->ipq_frags) {
1067 fp->ipq_frags = q->m_nextpkt;
1070 TAILQ_REMOVE(fhp, fp, ipq_list);
1071 uma_zfree(ipq_zone, fp);
1076 * IP timer processing;
1077 * if a timer expires on a reassembly
1078 * queue, discard it.
1087 for (i = 0; i < IPREASS_NHASH; i++) {
1088 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1092 fp = TAILQ_NEXT(fp, ipq_list);
1093 if(--fpp->ipq_ttl == 0) {
1094 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1095 ip_freef(&ipq[i], fpp);
1100 * If we are over the maximum number of fragments
1101 * (due to the limit being lowered), drain off
1102 * enough to get down to the new limit.
1104 if (maxnipq >= 0 && nipq > maxnipq) {
1105 for (i = 0; i < IPREASS_NHASH; i++) {
1106 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1107 ipstat.ips_fragdropped +=
1108 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1109 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1117 * Drain off all datagram fragments.
1125 for (i = 0; i < IPREASS_NHASH; i++) {
1126 while(!TAILQ_EMPTY(&ipq[i])) {
1127 ipstat.ips_fragdropped +=
1128 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1129 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1137 * The protocol to be inserted into ip_protox[] must be already registered
1138 * in inetsw[], either statically or through pf_proto_register().
1141 ipproto_register(u_char ipproto)
1145 /* Sanity checks. */
1147 return (EPROTONOSUPPORT);
1150 * The protocol slot must not be occupied by another protocol
1151 * already. An index pointing to IPPROTO_RAW is unused.
1153 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1155 return (EPFNOSUPPORT);
1156 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1159 /* Find the protocol position in inetsw[] and set the index. */
1160 for (pr = inetdomain.dom_protosw;
1161 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1162 if (pr->pr_domain->dom_family == PF_INET &&
1163 pr->pr_protocol && pr->pr_protocol == ipproto) {
1164 /* Be careful to only index valid IP protocols. */
1165 if (pr->pr_protocol < IPPROTO_MAX) {
1166 ip_protox[pr->pr_protocol] = pr - inetsw;
1172 return (EPROTONOSUPPORT);
1176 ipproto_unregister(u_char ipproto)
1180 /* Sanity checks. */
1182 return (EPROTONOSUPPORT);
1184 /* Check if the protocol was indeed registered. */
1185 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1187 return (EPFNOSUPPORT);
1188 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1191 /* Reset the protocol slot to IPPROTO_RAW. */
1192 ip_protox[ipproto] = pr - inetsw;
1197 * Given address of next destination (final or next hop),
1198 * return internet address info of interface to be used to get there.
1201 ip_rtaddr(struct in_addr dst)
1204 struct sockaddr_in *sin;
1205 struct in_ifaddr *ifa;
1207 bzero(&sro, sizeof(sro));
1208 sin = (struct sockaddr_in *)&sro.ro_dst;
1209 sin->sin_family = AF_INET;
1210 sin->sin_len = sizeof(*sin);
1211 sin->sin_addr = dst;
1212 rtalloc_ign(&sro, RTF_CLONING);
1214 if (sro.ro_rt == NULL)
1217 ifa = ifatoia(sro.ro_rt->rt_ifa);
1222 u_char inetctlerrmap[PRC_NCMDS] = {
1224 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1225 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1226 EMSGSIZE, EHOSTUNREACH, 0, 0,
1227 0, 0, EHOSTUNREACH, 0,
1228 ENOPROTOOPT, ECONNREFUSED
1232 * Forward a packet. If some error occurs return the sender
1233 * an icmp packet. Note we can't always generate a meaningful
1234 * icmp message because icmp doesn't have a large enough repertoire
1235 * of codes and types.
1237 * If not forwarding, just drop the packet. This could be confusing
1238 * if ipforwarding was zero but some routing protocol was advancing
1239 * us as a gateway to somewhere. However, we must let the routing
1240 * protocol deal with that.
1242 * The srcrt parameter indicates whether the packet is being forwarded
1243 * via a source route.
1246 ip_forward(struct mbuf *m, int srcrt)
1248 struct ip *ip = mtod(m, struct ip *);
1249 struct in_ifaddr *ia = NULL;
1251 struct in_addr dest;
1252 int error, type = 0, code = 0, mtu = 0;
1254 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1255 ipstat.ips_cantforward++;
1262 if (ip->ip_ttl <= IPTTLDEC) {
1263 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1271 ia = ip_rtaddr(ip->ip_dst);
1272 if (!srcrt && ia == NULL) {
1273 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1278 * Save the IP header and at most 8 bytes of the payload,
1279 * in case we need to generate an ICMP message to the src.
1281 * XXX this can be optimized a lot by saving the data in a local
1282 * buffer on the stack (72 bytes at most), and only allocating the
1283 * mbuf if really necessary. The vast majority of the packets
1284 * are forwarded without having to send an ICMP back (either
1285 * because unnecessary, or because rate limited), so we are
1286 * really we are wasting a lot of work here.
1288 * We don't use m_copy() because it might return a reference
1289 * to a shared cluster. Both this function and ip_output()
1290 * assume exclusive access to the IP header in `m', so any
1291 * data in a cluster may change before we reach icmp_error().
1293 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1294 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1296 * It's probably ok if the pkthdr dup fails (because
1297 * the deep copy of the tag chain failed), but for now
1298 * be conservative and just discard the copy since
1299 * code below may some day want the tags.
1304 if (mcopy != NULL) {
1305 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1306 mcopy->m_pkthdr.len = mcopy->m_len;
1307 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1313 ip->ip_ttl -= IPTTLDEC;
1319 * If forwarding packet using same interface that it came in on,
1320 * perhaps should send a redirect to sender to shortcut a hop.
1321 * Only send redirect if source is sending directly to us,
1322 * and if packet was not source routed (or has any options).
1323 * Also, don't send redirect if forwarding using a default route
1324 * or a route modified by a redirect.
1327 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1328 struct sockaddr_in *sin;
1332 bzero(&ro, sizeof(ro));
1333 sin = (struct sockaddr_in *)&ro.ro_dst;
1334 sin->sin_family = AF_INET;
1335 sin->sin_len = sizeof(*sin);
1336 sin->sin_addr = ip->ip_dst;
1337 rtalloc_ign(&ro, RTF_CLONING);
1341 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1342 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1343 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1344 u_long src = ntohl(ip->ip_src.s_addr);
1347 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1348 if (rt->rt_flags & RTF_GATEWAY)
1349 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1351 dest.s_addr = ip->ip_dst.s_addr;
1352 /* Router requirements says to only send host redirects */
1353 type = ICMP_REDIRECT;
1354 code = ICMP_REDIRECT_HOST;
1361 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1363 ipstat.ips_cantforward++;
1365 ipstat.ips_forward++;
1367 ipstat.ips_redirectsent++;
1379 case 0: /* forwarded, but need redirect */
1380 /* type, code set above */
1383 case ENETUNREACH: /* shouldn't happen, checked above */
1388 type = ICMP_UNREACH;
1389 code = ICMP_UNREACH_HOST;
1393 type = ICMP_UNREACH;
1394 code = ICMP_UNREACH_NEEDFRAG;
1397 mtu = ip_ipsec_mtu(m);
1400 * If the MTU wasn't set before use the interface mtu or
1401 * fall back to the next smaller mtu step compared to the
1402 * current packet size.
1406 mtu = ia->ia_ifp->if_mtu;
1408 mtu = ip_next_mtu(ip->ip_len, 0);
1410 ipstat.ips_cantfrag++;
1415 * A router should not generate ICMP_SOURCEQUENCH as
1416 * required in RFC1812 Requirements for IP Version 4 Routers.
1417 * Source quench could be a big problem under DoS attacks,
1418 * or if the underlying interface is rate-limited.
1419 * Those who need source quench packets may re-enable them
1420 * via the net.inet.ip.sendsourcequench sysctl.
1422 if (ip_sendsourcequench == 0) {
1426 type = ICMP_SOURCEQUENCH;
1431 case EACCES: /* ipfw denied packet */
1435 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1439 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1442 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1446 if (inp->inp_socket->so_options & SO_BINTIME) {
1447 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1448 SCM_BINTIME, SOL_SOCKET);
1450 mp = &(*mp)->m_next;
1452 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1455 bintime2timeval(&bt, &tv);
1456 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1457 SCM_TIMESTAMP, SOL_SOCKET);
1459 mp = &(*mp)->m_next;
1462 if (inp->inp_flags & INP_RECVDSTADDR) {
1463 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1464 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1466 mp = &(*mp)->m_next;
1468 if (inp->inp_flags & INP_RECVTTL) {
1469 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1470 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1472 mp = &(*mp)->m_next;
1476 * Moving these out of udp_input() made them even more broken
1477 * than they already were.
1479 /* options were tossed already */
1480 if (inp->inp_flags & INP_RECVOPTS) {
1481 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1482 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1484 mp = &(*mp)->m_next;
1486 /* ip_srcroute doesn't do what we want here, need to fix */
1487 if (inp->inp_flags & INP_RECVRETOPTS) {
1488 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1489 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1491 mp = &(*mp)->m_next;
1494 if (inp->inp_flags & INP_RECVIF) {
1497 struct sockaddr_dl sdl;
1500 struct sockaddr_dl *sdp;
1501 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1503 if (((ifp = m->m_pkthdr.rcvif))
1504 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1505 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1507 * Change our mind and don't try copy.
1509 if ((sdp->sdl_family != AF_LINK)
1510 || (sdp->sdl_len > sizeof(sdlbuf))) {
1513 bcopy(sdp, sdl2, sdp->sdl_len);
1517 = offsetof(struct sockaddr_dl, sdl_data[0]);
1518 sdl2->sdl_family = AF_LINK;
1519 sdl2->sdl_index = 0;
1520 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1522 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1523 IP_RECVIF, IPPROTO_IP);
1525 mp = &(*mp)->m_next;
1530 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1531 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1532 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1535 static int ip_rsvp_on;
1536 struct socket *ip_rsvpd;
1538 ip_rsvp_init(struct socket *so)
1540 if (so->so_type != SOCK_RAW ||
1541 so->so_proto->pr_protocol != IPPROTO_RSVP)
1544 if (ip_rsvpd != NULL)
1549 * This may seem silly, but we need to be sure we don't over-increment
1550 * the RSVP counter, in case something slips up.
1565 * This may seem silly, but we need to be sure we don't over-decrement
1566 * the RSVP counter, in case something slips up.
1576 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1578 if (rsvp_input_p) { /* call the real one if loaded */
1579 rsvp_input_p(m, off);
1583 /* Can still get packets with rsvp_on = 0 if there is a local member
1584 * of the group to which the RSVP packet is addressed. But in this
1585 * case we want to throw the packet away.
1593 if (ip_rsvpd != NULL) {
1597 /* Drop the packet */