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>
54 #include <sys/vimage.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/route.h>
62 #include <net/netisr.h>
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_icmp.h>
71 #include <netinet/ip_options.h>
72 #include <machine/in_cksum.h>
74 #include <netinet/ip_carp.h>
77 #include <netinet/ip_ipsec.h>
80 #include <sys/socketvar.h>
82 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_dummynet.h>
86 #include <security/mac/mac_framework.h>
89 CTASSERT(sizeof(struct ip) == 20);
95 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
96 forwarding, CTLFLAG_RW, ipforwarding, 0,
97 "Enable IP forwarding between interfaces");
99 static int ipsendredirects = 1; /* XXX */
100 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
101 redirect, CTLFLAG_RW, ipsendredirects, 0,
102 "Enable sending IP redirects");
104 int ip_defttl = IPDEFTTL;
105 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
106 ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
108 static int ip_keepfaith = 0;
109 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
110 keepfaith, CTLFLAG_RW, ip_keepfaith, 0,
111 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
113 static int ip_sendsourcequench = 0;
114 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
115 sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
116 "Enable the transmission of source quench packets");
118 int ip_do_randomid = 0;
119 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
120 CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values");
123 * XXX - Setting ip_checkinterface mostly implements the receive side of
124 * the Strong ES model described in RFC 1122, but since the routing table
125 * and transmit implementation do not implement the Strong ES model,
126 * setting this to 1 results in an odd hybrid.
128 * XXX - ip_checkinterface currently must be disabled if you use ipnat
129 * to translate the destination address to another local interface.
131 * XXX - ip_checkinterface must be disabled if you add IP aliases
132 * to the loopback interface instead of the interface where the
133 * packets for those addresses are received.
135 static int ip_checkinterface = 0;
136 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
137 check_interface, CTLFLAG_RW, ip_checkinterface, 0,
138 "Verify packet arrives on correct interface");
140 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
142 static struct ifqueue ipintrq;
143 static int ipqmaxlen = IFQ_MAXLEN;
145 extern struct domain inetdomain;
146 extern struct protosw inetsw[];
147 u_char ip_protox[IPPROTO_MAX];
148 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
149 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
150 u_long in_ifaddrhmask; /* mask for hash table */
152 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
153 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
154 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
155 &ipintrq.ifq_drops, 0,
156 "Number of packets dropped from the IP input queue");
158 struct ipstat ipstat;
159 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
160 ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
163 * IP datagram reassembly.
165 #define IPREASS_NHASH_LOG2 6
166 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
167 #define IPREASS_HMASK (IPREASS_NHASH - 1)
168 #define IPREASS_HASH(x,y) \
169 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
171 static uma_zone_t ipq_zone;
172 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
173 static struct mtx ipqlock;
175 #define IPQ_LOCK() mtx_lock(&ipqlock)
176 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
177 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
178 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
180 static void maxnipq_update(void);
181 static void ipq_zone_change(void *);
183 static int maxnipq; /* Administrative limit on # reass queues. */
184 static int nipq = 0; /* Total # of reass queues */
185 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
187 "Current number of IPv4 fragment reassembly queue entries");
189 static int maxfragsperpacket;
190 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
191 CTLFLAG_RW, maxfragsperpacket, 0,
192 "Maximum number of IPv4 fragments allowed per packet");
194 struct callout ipport_tick_callout;
197 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
198 &ip_mtu, 0, "Default MTU");
203 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
204 ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
208 * ipfw_ether and ipfw_bridge hooks.
209 * XXX: Temporary until those are converted to pfil_hooks as well.
211 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
212 ip_dn_io_t *ip_dn_io_ptr = NULL;
215 static void ip_freef(struct ipqhead *, struct ipq *);
218 * IP initialization: fill in IP protocol switch table.
219 * All protocols not implemented in kernel go to raw IP protocol handler.
224 INIT_VNET_INET(curvnet);
228 TAILQ_INIT(&V_in_ifaddrhead);
229 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
230 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
232 panic("ip_init: PF_INET not found");
234 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
235 for (i = 0; i < IPPROTO_MAX; i++)
236 ip_protox[i] = pr - inetsw;
238 * Cycle through IP protocols and put them into the appropriate place
241 for (pr = inetdomain.dom_protosw;
242 pr < inetdomain.dom_protoswNPROTOSW; pr++)
243 if (pr->pr_domain->dom_family == PF_INET &&
244 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
245 /* Be careful to only index valid IP protocols. */
246 if (pr->pr_protocol < IPPROTO_MAX)
247 ip_protox[pr->pr_protocol] = pr - inetsw;
250 /* Initialize packet filter hooks. */
251 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
252 inet_pfil_hook.ph_af = AF_INET;
253 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
254 printf("%s: WARNING: unable to register pfil hook, "
255 "error %d\n", __func__, i);
257 /* Initialize IP reassembly queue. */
259 for (i = 0; i < IPREASS_NHASH; i++)
260 TAILQ_INIT(&V_ipq[i]);
261 V_maxnipq = nmbclusters / 32;
262 V_maxfragsperpacket = 16;
263 V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
264 NULL, UMA_ALIGN_PTR, 0);
267 /* Start ipport_tick. */
268 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
270 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
271 SHUTDOWN_PRI_DEFAULT);
272 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
273 NULL, EVENTHANDLER_PRI_ANY);
275 /* Initialize various other remaining things. */
276 V_ip_id = time_second & 0xffff;
277 ipintrq.ifq_maxlen = ipqmaxlen;
278 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
279 netisr_register(NETISR_IP, ip_input, &ipintrq, 0);
286 callout_stop(&ipport_tick_callout);
290 * Ip input routine. Checksum and byte swap header. If fragmented
291 * try to reassemble. Process options. Pass to next level.
294 ip_input(struct mbuf *m)
296 INIT_VNET_INET(curvnet);
297 struct ip *ip = NULL;
298 struct in_ifaddr *ia = NULL;
300 int checkif, hlen = 0;
302 int dchg = 0; /* dest changed after fw */
303 struct in_addr odst; /* original dst address */
307 if (m->m_flags & M_FASTFWD_OURS) {
309 * Firewall or NAT changed destination to local.
310 * We expect ip_len and ip_off to be in host byte order.
312 m->m_flags &= ~M_FASTFWD_OURS;
313 /* Set up some basics that will be used later. */
314 ip = mtod(m, struct ip *);
315 hlen = ip->ip_hl << 2;
319 V_ipstat.ips_total++;
321 if (m->m_pkthdr.len < sizeof(struct ip))
324 if (m->m_len < sizeof (struct ip) &&
325 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
326 V_ipstat.ips_toosmall++;
329 ip = mtod(m, struct ip *);
331 if (ip->ip_v != IPVERSION) {
332 V_ipstat.ips_badvers++;
336 hlen = ip->ip_hl << 2;
337 if (hlen < sizeof(struct ip)) { /* minimum header length */
338 V_ipstat.ips_badhlen++;
341 if (hlen > m->m_len) {
342 if ((m = m_pullup(m, hlen)) == NULL) {
343 V_ipstat.ips_badhlen++;
346 ip = mtod(m, struct ip *);
349 /* 127/8 must not appear on wire - RFC1122 */
350 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
351 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
352 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
353 V_ipstat.ips_badaddr++;
358 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
359 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
361 if (hlen == sizeof(struct ip)) {
362 sum = in_cksum_hdr(ip);
364 sum = in_cksum(m, hlen);
368 V_ipstat.ips_badsum++;
373 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
374 /* packet is dropped by traffic conditioner */
379 * Convert fields to host representation.
381 ip->ip_len = ntohs(ip->ip_len);
382 if (ip->ip_len < hlen) {
383 V_ipstat.ips_badlen++;
386 ip->ip_off = ntohs(ip->ip_off);
389 * Check that the amount of data in the buffers
390 * is as at least much as the IP header would have us expect.
391 * Trim mbufs if longer than we expect.
392 * Drop packet if shorter than we expect.
394 if (m->m_pkthdr.len < ip->ip_len) {
396 V_ipstat.ips_tooshort++;
399 if (m->m_pkthdr.len > ip->ip_len) {
400 if (m->m_len == m->m_pkthdr.len) {
401 m->m_len = ip->ip_len;
402 m->m_pkthdr.len = ip->ip_len;
404 m_adj(m, ip->ip_len - m->m_pkthdr.len);
408 * Bypass packet filtering for packets from a tunnel (gif).
410 if (ip_ipsec_filtertunnel(m))
415 * Run through list of hooks for input packets.
417 * NB: Beware of the destination address changing (e.g.
418 * by NAT rewriting). When this happens, tell
419 * ip_forward to do the right thing.
422 /* Jump over all PFIL processing if hooks are not active. */
423 if (!PFIL_HOOKED(&inet_pfil_hook))
427 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
430 if (m == NULL) /* consumed by filter */
433 ip = mtod(m, struct ip *);
434 dchg = (odst.s_addr != ip->ip_dst.s_addr);
436 #ifdef IPFIREWALL_FORWARD
437 if (m->m_flags & M_FASTFWD_OURS) {
438 m->m_flags &= ~M_FASTFWD_OURS;
441 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
443 * Directly ship on the packet. This allows to forward packets
444 * that were destined for us to some other directly connected
450 #endif /* IPFIREWALL_FORWARD */
454 * Process options and, if not destined for us,
455 * ship it on. ip_dooptions returns 1 when an
456 * error was detected (causing an icmp message
457 * to be sent and the original packet to be freed).
459 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
462 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
463 * matter if it is destined to another node, or whether it is
464 * a multicast one, RSVP wants it! and prevents it from being forwarded
465 * anywhere else. Also checks if the rsvp daemon is running before
466 * grabbing the packet.
468 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
472 * Check our list of addresses, to see if the packet is for us.
473 * If we don't have any addresses, assume any unicast packet
474 * we receive might be for us (and let the upper layers deal
477 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
478 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
482 * Enable a consistency check between the destination address
483 * and the arrival interface for a unicast packet (the RFC 1122
484 * strong ES model) if IP forwarding is disabled and the packet
485 * is not locally generated and the packet is not subject to
488 * XXX - Checking also should be disabled if the destination
489 * address is ipnat'ed to a different interface.
491 * XXX - Checking is incompatible with IP aliases added
492 * to the loopback interface instead of the interface where
493 * the packets are received.
495 * XXX - This is the case for carp vhost IPs as well so we
496 * insert a workaround. If the packet got here, we already
497 * checked with carp_iamatch() and carp_forus().
499 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
500 m->m_pkthdr.rcvif != NULL &&
501 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
503 !m->m_pkthdr.rcvif->if_carp &&
508 * Check for exact addresses in the hash bucket.
510 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
512 * If the address matches, verify that the packet
513 * arrived via the correct interface if checking is
516 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
517 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
521 * Check for broadcast addresses.
523 * Only accept broadcast packets that arrive via the matching
524 * interface. Reception of forwarded directed broadcasts would
525 * be handled via ip_forward() and ether_output() with the loopback
526 * into the stack for SIMPLEX interfaces handled by ether_output().
528 if (m->m_pkthdr.rcvif != NULL &&
529 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
530 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
531 if (ifa->ifa_addr->sa_family != AF_INET)
534 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
537 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
540 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
545 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
546 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
547 V_ipstat.ips_cantforward++;
551 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
552 struct in_multi *inm;
555 * If we are acting as a multicast router, all
556 * incoming multicast packets are passed to the
557 * kernel-level multicast forwarding function.
558 * The packet is returned (relatively) intact; if
559 * ip_mforward() returns a non-zero value, the packet
560 * must be discarded, else it may be accepted below.
563 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
564 V_ipstat.ips_cantforward++;
570 * The process-level routing daemon needs to receive
571 * all multicast IGMP packets, whether or not this
572 * host belongs to their destination groups.
574 if (ip->ip_p == IPPROTO_IGMP)
576 V_ipstat.ips_forward++;
579 * See if we belong to the destination multicast group on the
583 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
586 V_ipstat.ips_notmember++;
592 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
594 if (ip->ip_dst.s_addr == INADDR_ANY)
598 * FAITH(Firewall Aided Internet Translator)
600 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
601 if (V_ip_keepfaith) {
602 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
610 * Not for us; forward if possible and desirable.
612 if (V_ipforwarding == 0) {
613 V_ipstat.ips_cantforward++;
627 * IPSTEALTH: Process non-routing options only
628 * if the packet is destined for us.
630 if (V_ipstealth && hlen > sizeof (struct ip) &&
633 #endif /* IPSTEALTH */
635 /* Count the packet in the ip address stats */
637 ia->ia_ifa.if_ipackets++;
638 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
642 * Attempt reassembly; if it succeeds, proceed.
643 * ip_reass() will return a different mbuf.
645 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
649 ip = mtod(m, struct ip *);
650 /* Get the header length of the reassembled packet */
651 hlen = ip->ip_hl << 2;
655 * Further protocols expect the packet length to be w/o the
662 * enforce IPsec policy checking if we are seeing last header.
663 * note that we do not visit this with protocols with pcb layer
664 * code - like udp/tcp/raw ip.
666 if (ip_ipsec_input(m))
671 * Switch out to protocol's input routine.
673 V_ipstat.ips_delivered++;
675 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
682 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
683 * max has slightly different semantics than the sysctl, for historical
689 INIT_VNET_INET(curvnet);
692 * -1 for unlimited allocation.
695 uma_zone_set_max(V_ipq_zone, 0);
697 * Positive number for specific bound.
700 uma_zone_set_max(V_ipq_zone, V_maxnipq);
702 * Zero specifies no further fragment queue allocation -- set the
703 * bound very low, but rely on implementation elsewhere to actually
704 * prevent allocation and reclaim current queues.
707 uma_zone_set_max(V_ipq_zone, 1);
711 ipq_zone_change(void *tag)
713 INIT_VNET_INET(curvnet);
715 if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
716 V_maxnipq = nmbclusters / 32;
722 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
724 INIT_VNET_INET(curvnet);
728 error = sysctl_handle_int(oidp, &i, 0, req);
729 if (error || !req->newptr)
733 * XXXRW: Might be a good idea to sanity check the argument and place
734 * an extreme upper bound.
743 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
744 NULL, 0, sysctl_maxnipq, "I",
745 "Maximum number of IPv4 fragment reassembly queue entries");
748 * Take incoming datagram fragment and try to reassemble it into
749 * whole datagram. If the argument is the first fragment or one
750 * in between the function will return NULL and store the mbuf
751 * in the fragment chain. If the argument is the last fragment
752 * the packet will be reassembled and the pointer to the new
753 * mbuf returned for further processing. Only m_tags attached
754 * to the first packet/fragment are preserved.
755 * The IP header is *NOT* adjusted out of iplen.
758 ip_reass(struct mbuf *m)
760 INIT_VNET_INET(curvnet);
762 struct mbuf *p, *q, *nq, *t;
763 struct ipq *fp = NULL;
764 struct ipqhead *head;
769 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
770 if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
771 V_ipstat.ips_fragments++;
772 V_ipstat.ips_fragdropped++;
777 ip = mtod(m, struct ip *);
778 hlen = ip->ip_hl << 2;
780 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
785 * Look for queue of fragments
788 TAILQ_FOREACH(fp, head, ipq_list)
789 if (ip->ip_id == fp->ipq_id &&
790 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
791 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
793 mac_ipq_match(m, fp) &&
795 ip->ip_p == fp->ipq_p)
801 * Attempt to trim the number of allocated fragment queues if it
802 * exceeds the administrative limit.
804 if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
806 * drop something from the tail of the current queue
807 * before proceeding further
809 struct ipq *q = TAILQ_LAST(head, ipqhead);
810 if (q == NULL) { /* gak */
811 for (i = 0; i < IPREASS_NHASH; i++) {
812 struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
814 V_ipstat.ips_fragtimeout +=
816 ip_freef(&V_ipq[i], r);
821 V_ipstat.ips_fragtimeout += q->ipq_nfrags;
828 * Adjust ip_len to not reflect header,
829 * convert offset of this to bytes.
832 if (ip->ip_off & IP_MF) {
834 * Make sure that fragments have a data length
835 * that's a non-zero multiple of 8 bytes.
837 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
838 V_ipstat.ips_toosmall++; /* XXX */
841 m->m_flags |= M_FRAG;
843 m->m_flags &= ~M_FRAG;
848 * Attempt reassembly; if it succeeds, proceed.
849 * ip_reass() will return a different mbuf.
851 V_ipstat.ips_fragments++;
852 m->m_pkthdr.header = ip;
854 /* Previous ip_reass() started here. */
856 * Presence of header sizes in mbufs
857 * would confuse code below.
863 * If first fragment to arrive, create a reassembly queue.
866 fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
870 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
871 uma_zfree(V_ipq_zone, fp);
875 mac_ipq_create(m, fp);
877 TAILQ_INSERT_HEAD(head, fp, ipq_list);
880 fp->ipq_ttl = IPFRAGTTL;
881 fp->ipq_p = ip->ip_p;
882 fp->ipq_id = ip->ip_id;
883 fp->ipq_src = ip->ip_src;
884 fp->ipq_dst = ip->ip_dst;
891 mac_ipq_update(m, fp);
895 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
898 * Handle ECN by comparing this segment with the first one;
899 * if CE is set, do not lose CE.
900 * drop if CE and not-ECT are mixed for the same packet.
902 ecn = ip->ip_tos & IPTOS_ECN_MASK;
903 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
904 if (ecn == IPTOS_ECN_CE) {
905 if (ecn0 == IPTOS_ECN_NOTECT)
907 if (ecn0 != IPTOS_ECN_CE)
908 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
910 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
914 * Find a segment which begins after this one does.
916 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
917 if (GETIP(q)->ip_off > ip->ip_off)
921 * If there is a preceding segment, it may provide some of
922 * our data already. If so, drop the data from the incoming
923 * segment. If it provides all of our data, drop us, otherwise
924 * stick new segment in the proper place.
926 * If some of the data is dropped from the the preceding
927 * segment, then it's checksum is invalidated.
930 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
935 m->m_pkthdr.csum_flags = 0;
939 m->m_nextpkt = p->m_nextpkt;
942 m->m_nextpkt = fp->ipq_frags;
947 * While we overlap succeeding segments trim them or,
948 * if they are completely covered, dequeue them.
950 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
952 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
953 if (i < GETIP(q)->ip_len) {
954 GETIP(q)->ip_len -= i;
955 GETIP(q)->ip_off += i;
957 q->m_pkthdr.csum_flags = 0;
962 V_ipstat.ips_fragdropped++;
968 * Check for complete reassembly and perform frag per packet
971 * Frag limiting is performed here so that the nth frag has
972 * a chance to complete the packet before we drop the packet.
973 * As a result, n+1 frags are actually allowed per packet, but
974 * only n will ever be stored. (n = maxfragsperpacket.)
978 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
979 if (GETIP(q)->ip_off != next) {
980 if (fp->ipq_nfrags > V_maxfragsperpacket) {
981 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
986 next += GETIP(q)->ip_len;
988 /* Make sure the last packet didn't have the IP_MF flag */
989 if (p->m_flags & M_FRAG) {
990 if (fp->ipq_nfrags > V_maxfragsperpacket) {
991 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
998 * Reassembly is complete. Make sure the packet is a sane size.
1002 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1003 V_ipstat.ips_toolong++;
1004 V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1010 * Concatenate fragments.
1017 q->m_nextpkt = NULL;
1018 for (q = nq; q != NULL; q = nq) {
1020 q->m_nextpkt = NULL;
1021 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1022 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1026 * In order to do checksumming faster we do 'end-around carry' here
1027 * (and not in for{} loop), though it implies we are not going to
1028 * reassemble more than 64k fragments.
1030 m->m_pkthdr.csum_data =
1031 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1033 mac_ipq_reassemble(fp, m);
1034 mac_ipq_destroy(fp);
1038 * Create header for new ip packet by modifying header of first
1039 * packet; dequeue and discard fragment reassembly header.
1040 * Make header visible.
1042 ip->ip_len = (ip->ip_hl << 2) + next;
1043 ip->ip_src = fp->ipq_src;
1044 ip->ip_dst = fp->ipq_dst;
1045 TAILQ_REMOVE(head, fp, ipq_list);
1047 uma_zfree(V_ipq_zone, fp);
1048 m->m_len += (ip->ip_hl << 2);
1049 m->m_data -= (ip->ip_hl << 2);
1050 /* some debugging cruft by sklower, below, will go away soon */
1051 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1053 V_ipstat.ips_reassembled++;
1058 V_ipstat.ips_fragdropped++;
1070 * Free a fragment reassembly header and all
1071 * associated datagrams.
1074 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1076 INIT_VNET_INET(curvnet);
1081 while (fp->ipq_frags) {
1083 fp->ipq_frags = q->m_nextpkt;
1086 TAILQ_REMOVE(fhp, fp, ipq_list);
1087 uma_zfree(V_ipq_zone, fp);
1092 * IP timer processing;
1093 * if a timer expires on a reassembly
1094 * queue, discard it.
1099 VNET_ITERATOR_DECL(vnet_iter);
1105 VNET_FOREACH(vnet_iter) {
1106 CURVNET_SET(vnet_iter);
1107 INIT_VNET_INET(vnet_iter);
1108 for (i = 0; i < IPREASS_NHASH; i++) {
1109 for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1113 fp = TAILQ_NEXT(fp, ipq_list);
1114 if(--fpp->ipq_ttl == 0) {
1115 V_ipstat.ips_fragtimeout +=
1117 ip_freef(&V_ipq[i], fpp);
1122 * If we are over the maximum number of fragments
1123 * (due to the limit being lowered), drain off
1124 * enough to get down to the new limit.
1126 if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1127 for (i = 0; i < IPREASS_NHASH; i++) {
1128 while (V_nipq > V_maxnipq &&
1129 !TAILQ_EMPTY(&V_ipq[i])) {
1130 V_ipstat.ips_fragdropped +=
1131 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1133 TAILQ_FIRST(&V_ipq[i]));
1139 VNET_LIST_RUNLOCK();
1144 * Drain off all datagram fragments.
1149 VNET_ITERATOR_DECL(vnet_iter);
1154 VNET_FOREACH(vnet_iter) {
1155 CURVNET_SET(vnet_iter);
1156 INIT_VNET_INET(vnet_iter);
1157 for (i = 0; i < IPREASS_NHASH; i++) {
1158 while(!TAILQ_EMPTY(&V_ipq[i])) {
1159 V_ipstat.ips_fragdropped +=
1160 TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1161 ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1166 VNET_LIST_RUNLOCK();
1172 * The protocol to be inserted into ip_protox[] must be already registered
1173 * in inetsw[], either statically or through pf_proto_register().
1176 ipproto_register(u_char ipproto)
1180 /* Sanity checks. */
1182 return (EPROTONOSUPPORT);
1185 * The protocol slot must not be occupied by another protocol
1186 * already. An index pointing to IPPROTO_RAW is unused.
1188 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1190 return (EPFNOSUPPORT);
1191 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1194 /* Find the protocol position in inetsw[] and set the index. */
1195 for (pr = inetdomain.dom_protosw;
1196 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1197 if (pr->pr_domain->dom_family == PF_INET &&
1198 pr->pr_protocol && pr->pr_protocol == ipproto) {
1199 /* Be careful to only index valid IP protocols. */
1200 if (pr->pr_protocol < IPPROTO_MAX) {
1201 ip_protox[pr->pr_protocol] = pr - inetsw;
1207 return (EPROTONOSUPPORT);
1211 ipproto_unregister(u_char ipproto)
1215 /* Sanity checks. */
1217 return (EPROTONOSUPPORT);
1219 /* Check if the protocol was indeed registered. */
1220 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1222 return (EPFNOSUPPORT);
1223 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1226 /* Reset the protocol slot to IPPROTO_RAW. */
1227 ip_protox[ipproto] = pr - inetsw;
1232 * Given address of next destination (final or next hop),
1233 * return internet address info of interface to be used to get there.
1236 ip_rtaddr(struct in_addr dst, u_int fibnum)
1239 struct sockaddr_in *sin;
1240 struct in_ifaddr *ifa;
1242 bzero(&sro, sizeof(sro));
1243 sin = (struct sockaddr_in *)&sro.ro_dst;
1244 sin->sin_family = AF_INET;
1245 sin->sin_len = sizeof(*sin);
1246 sin->sin_addr = dst;
1247 in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
1249 if (sro.ro_rt == NULL)
1252 ifa = ifatoia(sro.ro_rt->rt_ifa);
1257 u_char inetctlerrmap[PRC_NCMDS] = {
1259 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1260 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1261 EMSGSIZE, EHOSTUNREACH, 0, 0,
1262 0, 0, EHOSTUNREACH, 0,
1263 ENOPROTOOPT, ECONNREFUSED
1267 * Forward a packet. If some error occurs return the sender
1268 * an icmp packet. Note we can't always generate a meaningful
1269 * icmp message because icmp doesn't have a large enough repertoire
1270 * of codes and types.
1272 * If not forwarding, just drop the packet. This could be confusing
1273 * if ipforwarding was zero but some routing protocol was advancing
1274 * us as a gateway to somewhere. However, we must let the routing
1275 * protocol deal with that.
1277 * The srcrt parameter indicates whether the packet is being forwarded
1278 * via a source route.
1281 ip_forward(struct mbuf *m, int srcrt)
1283 INIT_VNET_INET(curvnet);
1284 struct ip *ip = mtod(m, struct ip *);
1285 struct in_ifaddr *ia = NULL;
1287 struct in_addr dest;
1289 int error, type = 0, code = 0, mtu = 0;
1291 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1292 V_ipstat.ips_cantforward++;
1299 if (ip->ip_ttl <= IPTTLDEC) {
1300 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1308 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1309 if (!srcrt && ia == NULL) {
1310 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1315 * Save the IP header and at most 8 bytes of the payload,
1316 * in case we need to generate an ICMP message to the src.
1318 * XXX this can be optimized a lot by saving the data in a local
1319 * buffer on the stack (72 bytes at most), and only allocating the
1320 * mbuf if really necessary. The vast majority of the packets
1321 * are forwarded without having to send an ICMP back (either
1322 * because unnecessary, or because rate limited), so we are
1323 * really we are wasting a lot of work here.
1325 * We don't use m_copy() because it might return a reference
1326 * to a shared cluster. Both this function and ip_output()
1327 * assume exclusive access to the IP header in `m', so any
1328 * data in a cluster may change before we reach icmp_error().
1330 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1331 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1333 * It's probably ok if the pkthdr dup fails (because
1334 * the deep copy of the tag chain failed), but for now
1335 * be conservative and just discard the copy since
1336 * code below may some day want the tags.
1341 if (mcopy != NULL) {
1342 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1343 mcopy->m_pkthdr.len = mcopy->m_len;
1344 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1350 ip->ip_ttl -= IPTTLDEC;
1356 * If forwarding packet using same interface that it came in on,
1357 * perhaps should send a redirect to sender to shortcut a hop.
1358 * Only send redirect if source is sending directly to us,
1359 * and if packet was not source routed (or has any options).
1360 * Also, don't send redirect if forwarding using a default route
1361 * or a route modified by a redirect.
1364 if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1365 struct sockaddr_in *sin;
1368 bzero(&ro, sizeof(ro));
1369 sin = (struct sockaddr_in *)&ro.ro_dst;
1370 sin->sin_family = AF_INET;
1371 sin->sin_len = sizeof(*sin);
1372 sin->sin_addr = ip->ip_dst;
1373 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
1377 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1378 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1379 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1380 u_long src = ntohl(ip->ip_src.s_addr);
1383 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1384 if (rt->rt_flags & RTF_GATEWAY)
1385 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1387 dest.s_addr = ip->ip_dst.s_addr;
1388 /* Router requirements says to only send host redirects */
1389 type = ICMP_REDIRECT;
1390 code = ICMP_REDIRECT_HOST;
1398 * Try to cache the route MTU from ip_output so we can consider it for
1399 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1401 bzero(&ro, sizeof(ro));
1403 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1405 if (error == EMSGSIZE && ro.ro_rt)
1406 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1411 V_ipstat.ips_cantforward++;
1413 V_ipstat.ips_forward++;
1415 V_ipstat.ips_redirectsent++;
1427 case 0: /* forwarded, but need redirect */
1428 /* type, code set above */
1431 case ENETUNREACH: /* shouldn't happen, checked above */
1436 type = ICMP_UNREACH;
1437 code = ICMP_UNREACH_HOST;
1441 type = ICMP_UNREACH;
1442 code = ICMP_UNREACH_NEEDFRAG;
1446 * If IPsec is configured for this path,
1447 * override any possibly mtu value set by ip_output.
1449 mtu = ip_ipsec_mtu(m, mtu);
1452 * If the MTU was set before make sure we are below the
1454 * If the MTU wasn't set before use the interface mtu or
1455 * fall back to the next smaller mtu step compared to the
1456 * current packet size.
1460 mtu = min(mtu, ia->ia_ifp->if_mtu);
1463 mtu = ia->ia_ifp->if_mtu;
1465 mtu = ip_next_mtu(ip->ip_len, 0);
1467 V_ipstat.ips_cantfrag++;
1472 * A router should not generate ICMP_SOURCEQUENCH as
1473 * required in RFC1812 Requirements for IP Version 4 Routers.
1474 * Source quench could be a big problem under DoS attacks,
1475 * or if the underlying interface is rate-limited.
1476 * Those who need source quench packets may re-enable them
1477 * via the net.inet.ip.sendsourcequench sysctl.
1479 if (V_ip_sendsourcequench == 0) {
1483 type = ICMP_SOURCEQUENCH;
1488 case EACCES: /* ipfw denied packet */
1492 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1496 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1499 INIT_VNET_NET(inp->inp_vnet);
1501 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1505 if (inp->inp_socket->so_options & SO_BINTIME) {
1506 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1507 SCM_BINTIME, SOL_SOCKET);
1509 mp = &(*mp)->m_next;
1511 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1514 bintime2timeval(&bt, &tv);
1515 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1516 SCM_TIMESTAMP, SOL_SOCKET);
1518 mp = &(*mp)->m_next;
1521 if (inp->inp_flags & INP_RECVDSTADDR) {
1522 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1523 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1525 mp = &(*mp)->m_next;
1527 if (inp->inp_flags & INP_RECVTTL) {
1528 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1529 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1531 mp = &(*mp)->m_next;
1535 * Moving these out of udp_input() made them even more broken
1536 * than they already were.
1538 /* options were tossed already */
1539 if (inp->inp_flags & INP_RECVOPTS) {
1540 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1541 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1543 mp = &(*mp)->m_next;
1545 /* ip_srcroute doesn't do what we want here, need to fix */
1546 if (inp->inp_flags & INP_RECVRETOPTS) {
1547 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1548 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1550 mp = &(*mp)->m_next;
1553 if (inp->inp_flags & INP_RECVIF) {
1556 struct sockaddr_dl sdl;
1559 struct sockaddr_dl *sdp;
1560 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1562 if (((ifp = m->m_pkthdr.rcvif))
1563 && ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1564 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1566 * Change our mind and don't try copy.
1568 if ((sdp->sdl_family != AF_LINK)
1569 || (sdp->sdl_len > sizeof(sdlbuf))) {
1572 bcopy(sdp, sdl2, sdp->sdl_len);
1576 = offsetof(struct sockaddr_dl, sdl_data[0]);
1577 sdl2->sdl_family = AF_LINK;
1578 sdl2->sdl_index = 0;
1579 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1581 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1582 IP_RECVIF, IPPROTO_IP);
1584 mp = &(*mp)->m_next;
1589 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1590 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1591 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1594 static int ip_rsvp_on;
1595 struct socket *ip_rsvpd;
1597 ip_rsvp_init(struct socket *so)
1599 INIT_VNET_INET(so->so_vnet);
1601 if (so->so_type != SOCK_RAW ||
1602 so->so_proto->pr_protocol != IPPROTO_RSVP)
1605 if (V_ip_rsvpd != NULL)
1610 * This may seem silly, but we need to be sure we don't over-increment
1611 * the RSVP counter, in case something slips up.
1613 if (!V_ip_rsvp_on) {
1624 INIT_VNET_INET(curvnet);
1628 * This may seem silly, but we need to be sure we don't over-decrement
1629 * the RSVP counter, in case something slips up.
1639 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1641 INIT_VNET_INET(curvnet);
1643 if (rsvp_input_p) { /* call the real one if loaded */
1644 rsvp_input_p(m, off);
1648 /* Can still get packets with rsvp_on = 0 if there is a local member
1649 * of the group to which the RSVP packet is addressed. But in this
1650 * case we want to throw the packet away.
1658 if (V_ip_rsvpd != NULL) {
1662 /* Drop the packet */