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
33 #include "opt_bootp.h"
35 #include "opt_ipstealth.h"
36 #include "opt_ipsec.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/callout.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
50 #include <sys/kernel.h>
51 #include <sys/syslog.h>
52 #include <sys/sysctl.h>
56 #include <net/if_types.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/route.h>
60 #include <net/netisr.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/ip_icmp.h>
69 #include <netinet/ip_options.h>
70 #include <machine/in_cksum.h>
72 #include <netinet/ip_carp.h>
74 #if defined(IPSEC) || defined(FAST_IPSEC)
75 #include <netinet/ip_ipsec.h>
78 #include <sys/socketvar.h>
80 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
81 #include <netinet/ip_fw.h>
82 #include <netinet/ip_dummynet.h>
87 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
88 &ipforwarding, 0, "Enable IP forwarding between interfaces");
90 static int ipsendredirects = 1; /* XXX */
91 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
92 &ipsendredirects, 0, "Enable sending IP redirects");
94 int ip_defttl = IPDEFTTL;
95 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
96 &ip_defttl, 0, "Maximum TTL on IP packets");
98 static int ip_keepfaith = 0;
99 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
101 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
103 static int ip_sendsourcequench = 0;
104 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
105 &ip_sendsourcequench, 0,
106 "Enable the transmission of source quench packets");
108 int ip_do_randomid = 0;
109 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
111 "Assign random ip_id values");
114 * XXX - Setting ip_checkinterface mostly implements the receive side of
115 * the Strong ES model described in RFC 1122, but since the routing table
116 * and transmit implementation do not implement the Strong ES model,
117 * setting this to 1 results in an odd hybrid.
119 * XXX - ip_checkinterface currently must be disabled if you use ipnat
120 * to translate the destination address to another local interface.
122 * XXX - ip_checkinterface must be disabled if you add IP aliases
123 * to the loopback interface instead of the interface where the
124 * packets for those addresses are received.
126 static int ip_checkinterface = 0;
127 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
128 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
130 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
132 static struct ifqueue ipintrq;
133 static int ipqmaxlen = IFQ_MAXLEN;
135 extern struct domain inetdomain;
136 extern struct protosw inetsw[];
137 u_char ip_protox[IPPROTO_MAX];
138 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
139 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
140 u_long in_ifaddrhmask; /* mask for hash table */
142 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
143 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
144 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
145 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
147 struct ipstat ipstat;
148 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
149 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
152 * IP datagram reassembly.
154 #define IPREASS_NHASH_LOG2 6
155 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
156 #define IPREASS_HMASK (IPREASS_NHASH - 1)
157 #define IPREASS_HASH(x,y) \
158 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
160 static uma_zone_t ipq_zone;
161 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
162 static struct mtx ipqlock;
164 #define IPQ_LOCK() mtx_lock(&ipqlock)
165 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
166 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
167 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
169 static void maxnipq_update(void);
171 static int maxnipq; /* Administrative limit on # reass queues. */
172 static int nipq = 0; /* Total # of reass queues */
173 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0,
174 "Current number of IPv4 fragment reassembly queue entries");
176 static int maxfragsperpacket;
177 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
178 &maxfragsperpacket, 0,
179 "Maximum number of IPv4 fragments allowed per packet");
181 struct callout ipport_tick_callout;
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185 &ip_mtu, 0, "Default MTU");
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
195 * ipfw_ether and ipfw_bridge hooks.
196 * XXX: Temporary until those are converted to pfil_hooks as well.
198 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
199 ip_dn_io_t *ip_dn_io_ptr = NULL;
203 static void ip_freef(struct ipqhead *, struct ipq *);
206 * IP initialization: fill in IP protocol switch table.
207 * All protocols not implemented in kernel go to raw IP protocol handler.
212 register struct protosw *pr;
215 TAILQ_INIT(&in_ifaddrhead);
216 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
217 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
219 panic("ip_init: PF_INET not found");
221 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
222 for (i = 0; i < IPPROTO_MAX; i++)
223 ip_protox[i] = pr - inetsw;
225 * Cycle through IP protocols and put them into the appropriate place
228 for (pr = inetdomain.dom_protosw;
229 pr < inetdomain.dom_protoswNPROTOSW; pr++)
230 if (pr->pr_domain->dom_family == PF_INET &&
231 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
232 /* Be careful to only index valid IP protocols. */
233 if (pr->pr_protocol < IPPROTO_MAX)
234 ip_protox[pr->pr_protocol] = pr - inetsw;
237 /* Initialize packet filter hooks. */
238 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
239 inet_pfil_hook.ph_af = AF_INET;
240 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
241 printf("%s: WARNING: unable to register pfil hook, "
242 "error %d\n", __func__, i);
244 /* Initialize IP reassembly queue. */
246 for (i = 0; i < IPREASS_NHASH; i++)
248 maxnipq = nmbclusters / 32;
249 maxfragsperpacket = 16;
250 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
251 NULL, UMA_ALIGN_PTR, 0);
254 /* Start ipport_tick. */
255 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
257 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
258 SHUTDOWN_PRI_DEFAULT);
260 /* Initialize various other remaining things. */
261 ip_id = time_second & 0xffff;
262 ipintrq.ifq_maxlen = ipqmaxlen;
263 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
264 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
270 callout_stop(&ipport_tick_callout);
274 * Ip input routine. Checksum and byte swap header. If fragmented
275 * try to reassemble. Process options. Pass to next level.
278 ip_input(struct mbuf *m)
280 struct ip *ip = NULL;
281 struct in_ifaddr *ia = NULL;
283 int checkif, hlen = 0;
285 int dchg = 0; /* dest changed after fw */
286 struct in_addr odst; /* original dst address */
290 if (m->m_flags & M_FASTFWD_OURS) {
292 * Firewall or NAT changed destination to local.
293 * We expect ip_len and ip_off to be in host byte order.
295 m->m_flags &= ~M_FASTFWD_OURS;
296 /* Set up some basics that will be used later. */
297 ip = mtod(m, struct ip *);
298 hlen = ip->ip_hl << 2;
304 if (m->m_pkthdr.len < sizeof(struct ip))
307 if (m->m_len < sizeof (struct ip) &&
308 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
309 ipstat.ips_toosmall++;
312 ip = mtod(m, struct ip *);
314 if (ip->ip_v != IPVERSION) {
315 ipstat.ips_badvers++;
319 hlen = ip->ip_hl << 2;
320 if (hlen < sizeof(struct ip)) { /* minimum header length */
321 ipstat.ips_badhlen++;
324 if (hlen > m->m_len) {
325 if ((m = m_pullup(m, hlen)) == NULL) {
326 ipstat.ips_badhlen++;
329 ip = mtod(m, struct ip *);
332 /* 127/8 must not appear on wire - RFC1122 */
333 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
334 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
335 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
336 ipstat.ips_badaddr++;
341 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
342 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
344 if (hlen == sizeof(struct ip)) {
345 sum = in_cksum_hdr(ip);
347 sum = in_cksum(m, hlen);
356 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
357 /* packet is dropped by traffic conditioner */
362 * Convert fields to host representation.
364 ip->ip_len = ntohs(ip->ip_len);
365 if (ip->ip_len < hlen) {
369 ip->ip_off = ntohs(ip->ip_off);
372 * Check that the amount of data in the buffers
373 * is as at least much as the IP header would have us expect.
374 * Trim mbufs if longer than we expect.
375 * Drop packet if shorter than we expect.
377 if (m->m_pkthdr.len < ip->ip_len) {
379 ipstat.ips_tooshort++;
382 if (m->m_pkthdr.len > ip->ip_len) {
383 if (m->m_len == m->m_pkthdr.len) {
384 m->m_len = ip->ip_len;
385 m->m_pkthdr.len = ip->ip_len;
387 m_adj(m, ip->ip_len - m->m_pkthdr.len);
389 #if defined(IPSEC) || defined(FAST_IPSEC)
391 * Bypass packet filtering for packets from a tunnel (gif).
393 if (ip_ipsec_filtergif(m))
398 * Run through list of hooks for input packets.
400 * NB: Beware of the destination address changing (e.g.
401 * by NAT rewriting). When this happens, tell
402 * ip_forward to do the right thing.
405 /* Jump over all PFIL processing if hooks are not active. */
406 if (!PFIL_HOOKED(&inet_pfil_hook))
410 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
413 if (m == NULL) /* consumed by filter */
416 ip = mtod(m, struct ip *);
417 dchg = (odst.s_addr != ip->ip_dst.s_addr);
419 #ifdef IPFIREWALL_FORWARD
420 if (m->m_flags & M_FASTFWD_OURS) {
421 m->m_flags &= ~M_FASTFWD_OURS;
424 #ifndef IPFIREWALL_FORWARD_EXTENDED
425 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
427 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
429 * Directly ship on the packet. This allows to forward packets
430 * that were destined for us to some other directly connected
436 #endif /* IPFIREWALL_FORWARD_EXTENDED */
437 #endif /* IPFIREWALL_FORWARD */
441 * Process options and, if not destined for us,
442 * ship it on. ip_dooptions returns 1 when an
443 * error was detected (causing an icmp message
444 * to be sent and the original packet to be freed).
446 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
449 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
450 * matter if it is destined to another node, or whether it is
451 * a multicast one, RSVP wants it! and prevents it from being forwarded
452 * anywhere else. Also checks if the rsvp daemon is running before
453 * grabbing the packet.
455 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
459 * Check our list of addresses, to see if the packet is for us.
460 * If we don't have any addresses, assume any unicast packet
461 * we receive might be for us (and let the upper layers deal
464 if (TAILQ_EMPTY(&in_ifaddrhead) &&
465 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
469 * Enable a consistency check between the destination address
470 * and the arrival interface for a unicast packet (the RFC 1122
471 * strong ES model) if IP forwarding is disabled and the packet
472 * is not locally generated and the packet is not subject to
475 * XXX - Checking also should be disabled if the destination
476 * address is ipnat'ed to a different interface.
478 * XXX - Checking is incompatible with IP aliases added
479 * to the loopback interface instead of the interface where
480 * the packets are received.
482 * XXX - This is the case for carp vhost IPs as well so we
483 * insert a workaround. If the packet got here, we already
484 * checked with carp_iamatch() and carp_forus().
486 checkif = ip_checkinterface && (ipforwarding == 0) &&
487 m->m_pkthdr.rcvif != NULL &&
488 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
490 !m->m_pkthdr.rcvif->if_carp &&
495 * Check for exact addresses in the hash bucket.
497 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
499 * If the address matches, verify that the packet
500 * arrived via the correct interface if checking is
503 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
504 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
508 * Check for broadcast addresses.
510 * Only accept broadcast packets that arrive via the matching
511 * interface. Reception of forwarded directed broadcasts would
512 * be handled via ip_forward() and ether_output() with the loopback
513 * into the stack for SIMPLEX interfaces handled by ether_output().
515 if (m->m_pkthdr.rcvif != NULL &&
516 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
517 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
518 if (ifa->ifa_addr->sa_family != AF_INET)
521 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
524 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
527 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
532 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
533 struct in_multi *inm;
536 * If we are acting as a multicast router, all
537 * incoming multicast packets are passed to the
538 * kernel-level multicast forwarding function.
539 * The packet is returned (relatively) intact; if
540 * ip_mforward() returns a non-zero value, the packet
541 * must be discarded, else it may be accepted below.
544 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
545 ipstat.ips_cantforward++;
551 * The process-level routing daemon needs to receive
552 * all multicast IGMP packets, whether or not this
553 * host belongs to their destination groups.
555 if (ip->ip_p == IPPROTO_IGMP)
557 ipstat.ips_forward++;
560 * See if we belong to the destination multicast group on the
564 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
567 ipstat.ips_notmember++;
573 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
575 if (ip->ip_dst.s_addr == INADDR_ANY)
579 * FAITH(Firewall Aided Internet Translator)
581 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
583 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
591 * Not for us; forward if possible and desirable.
593 if (ipforwarding == 0) {
594 ipstat.ips_cantforward++;
597 #if defined(IPSEC) || defined(FAST_IPSEC)
608 * IPSTEALTH: Process non-routing options only
609 * if the packet is destined for us.
611 if (ipstealth && hlen > sizeof (struct ip) &&
614 #endif /* IPSTEALTH */
616 /* Count the packet in the ip address stats */
618 ia->ia_ifa.if_ipackets++;
619 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
623 * Attempt reassembly; if it succeeds, proceed.
624 * ip_reass() will return a different mbuf.
626 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
630 ip = mtod(m, struct ip *);
631 /* Get the header length of the reassembled packet */
632 hlen = ip->ip_hl << 2;
636 * Further protocols expect the packet length to be w/o the
641 #if defined(IPSEC) || defined(FAST_IPSEC)
643 * enforce IPsec policy checking if we are seeing last header.
644 * note that we do not visit this with protocols with pcb layer
645 * code - like udp/tcp/raw ip.
647 if (ip_ipsec_input(m))
652 * Switch out to protocol's input routine.
654 ipstat.ips_delivered++;
656 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
663 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
664 * max has slightly different semantics than the sysctl, for historical
672 * -1 for unlimited allocation.
675 uma_zone_set_max(ipq_zone, 0);
677 * Positive number for specific bound.
680 uma_zone_set_max(ipq_zone, maxnipq);
682 * Zero specifies no further fragment queue allocation -- set the
683 * bound very low, but rely on implementation elsewhere to actually
684 * prevent allocation and reclaim current queues.
687 uma_zone_set_max(ipq_zone, 1);
691 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
696 error = sysctl_handle_int(oidp, &i, 0, req);
697 if (error || !req->newptr)
701 * XXXRW: Might be a good idea to sanity check the argument and place
702 * an extreme upper bound.
711 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
712 NULL, 0, sysctl_maxnipq, "I",
713 "Maximum number of IPv4 fragment reassembly queue entries");
716 * Take incoming datagram fragment and try to reassemble it into
717 * whole datagram. If the argument is the first fragment or one
718 * in between the function will return NULL and store the mbuf
719 * in the fragment chain. If the argument is the last fragment
720 * the packet will be reassembled and the pointer to the new
721 * mbuf returned for further processing. Only m_tags attached
722 * to the first packet/fragment are preserved.
723 * The IP header is *NOT* adjusted out of iplen.
727 ip_reass(struct mbuf *m)
730 struct mbuf *p, *q, *nq, *t;
731 struct ipq *fp = NULL;
732 struct ipqhead *head;
737 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
738 if (maxnipq == 0 || maxfragsperpacket == 0) {
739 ipstat.ips_fragments++;
740 ipstat.ips_fragdropped++;
745 ip = mtod(m, struct ip *);
746 hlen = ip->ip_hl << 2;
748 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
753 * Look for queue of fragments
756 TAILQ_FOREACH(fp, head, ipq_list)
757 if (ip->ip_id == fp->ipq_id &&
758 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
759 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
761 mac_fragment_match(m, fp) &&
763 ip->ip_p == fp->ipq_p)
769 * Attempt to trim the number of allocated fragment queues if it
770 * exceeds the administrative limit.
772 if ((nipq > maxnipq) && (maxnipq > 0)) {
774 * drop something from the tail of the current queue
775 * before proceeding further
777 struct ipq *q = TAILQ_LAST(head, ipqhead);
778 if (q == NULL) { /* gak */
779 for (i = 0; i < IPREASS_NHASH; i++) {
780 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
782 ipstat.ips_fragtimeout += r->ipq_nfrags;
783 ip_freef(&ipq[i], r);
788 ipstat.ips_fragtimeout += q->ipq_nfrags;
795 * Adjust ip_len to not reflect header,
796 * convert offset of this to bytes.
799 if (ip->ip_off & IP_MF) {
801 * Make sure that fragments have a data length
802 * that's a non-zero multiple of 8 bytes.
804 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
805 ipstat.ips_toosmall++; /* XXX */
808 m->m_flags |= M_FRAG;
810 m->m_flags &= ~M_FRAG;
815 * Attempt reassembly; if it succeeds, proceed.
816 * ip_reass() will return a different mbuf.
818 ipstat.ips_fragments++;
819 m->m_pkthdr.header = ip;
821 /* Previous ip_reass() started here. */
823 * Presence of header sizes in mbufs
824 * would confuse code below.
830 * If first fragment to arrive, create a reassembly queue.
833 fp = uma_zalloc(ipq_zone, M_NOWAIT);
837 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
838 uma_zfree(ipq_zone, fp);
841 mac_create_ipq(m, fp);
843 TAILQ_INSERT_HEAD(head, fp, ipq_list);
846 fp->ipq_ttl = IPFRAGTTL;
847 fp->ipq_p = ip->ip_p;
848 fp->ipq_id = ip->ip_id;
849 fp->ipq_src = ip->ip_src;
850 fp->ipq_dst = ip->ip_dst;
857 mac_update_ipq(m, fp);
861 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
864 * Handle ECN by comparing this segment with the first one;
865 * if CE is set, do not lose CE.
866 * drop if CE and not-ECT are mixed for the same packet.
868 ecn = ip->ip_tos & IPTOS_ECN_MASK;
869 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
870 if (ecn == IPTOS_ECN_CE) {
871 if (ecn0 == IPTOS_ECN_NOTECT)
873 if (ecn0 != IPTOS_ECN_CE)
874 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
876 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
880 * Find a segment which begins after this one does.
882 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
883 if (GETIP(q)->ip_off > ip->ip_off)
887 * If there is a preceding segment, it may provide some of
888 * our data already. If so, drop the data from the incoming
889 * segment. If it provides all of our data, drop us, otherwise
890 * stick new segment in the proper place.
892 * If some of the data is dropped from the the preceding
893 * segment, then it's checksum is invalidated.
896 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
901 m->m_pkthdr.csum_flags = 0;
905 m->m_nextpkt = p->m_nextpkt;
908 m->m_nextpkt = fp->ipq_frags;
913 * While we overlap succeeding segments trim them or,
914 * if they are completely covered, dequeue them.
916 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
918 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
919 if (i < GETIP(q)->ip_len) {
920 GETIP(q)->ip_len -= i;
921 GETIP(q)->ip_off += i;
923 q->m_pkthdr.csum_flags = 0;
928 ipstat.ips_fragdropped++;
934 * Check for complete reassembly and perform frag per packet
937 * Frag limiting is performed here so that the nth frag has
938 * a chance to complete the packet before we drop the packet.
939 * As a result, n+1 frags are actually allowed per packet, but
940 * only n will ever be stored. (n = maxfragsperpacket.)
944 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
945 if (GETIP(q)->ip_off != next) {
946 if (fp->ipq_nfrags > maxfragsperpacket) {
947 ipstat.ips_fragdropped += fp->ipq_nfrags;
952 next += GETIP(q)->ip_len;
954 /* Make sure the last packet didn't have the IP_MF flag */
955 if (p->m_flags & M_FRAG) {
956 if (fp->ipq_nfrags > maxfragsperpacket) {
957 ipstat.ips_fragdropped += fp->ipq_nfrags;
964 * Reassembly is complete. Make sure the packet is a sane size.
968 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
969 ipstat.ips_toolong++;
970 ipstat.ips_fragdropped += fp->ipq_nfrags;
976 * Concatenate fragments.
984 for (q = nq; q != NULL; q = nq) {
987 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
988 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
992 * In order to do checksumming faster we do 'end-around carry' here
993 * (and not in for{} loop), though it implies we are not going to
994 * reassemble more than 64k fragments.
996 m->m_pkthdr.csum_data =
997 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
999 mac_create_datagram_from_ipq(fp, m);
1000 mac_destroy_ipq(fp);
1004 * Create header for new ip packet by modifying header of first
1005 * packet; dequeue and discard fragment reassembly header.
1006 * Make header visible.
1008 ip->ip_len = (ip->ip_hl << 2) + next;
1009 ip->ip_src = fp->ipq_src;
1010 ip->ip_dst = fp->ipq_dst;
1011 TAILQ_REMOVE(head, fp, ipq_list);
1013 uma_zfree(ipq_zone, fp);
1014 m->m_len += (ip->ip_hl << 2);
1015 m->m_data -= (ip->ip_hl << 2);
1016 /* some debugging cruft by sklower, below, will go away soon */
1017 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1019 ipstat.ips_reassembled++;
1024 ipstat.ips_fragdropped++;
1036 * Free a fragment reassembly header and all
1037 * associated datagrams.
1041 struct ipqhead *fhp;
1044 register struct mbuf *q;
1048 while (fp->ipq_frags) {
1050 fp->ipq_frags = q->m_nextpkt;
1053 TAILQ_REMOVE(fhp, fp, ipq_list);
1054 uma_zfree(ipq_zone, fp);
1059 * IP timer processing;
1060 * if a timer expires on a reassembly
1061 * queue, discard it.
1066 register struct ipq *fp;
1070 for (i = 0; i < IPREASS_NHASH; i++) {
1071 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1075 fp = TAILQ_NEXT(fp, ipq_list);
1076 if(--fpp->ipq_ttl == 0) {
1077 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1078 ip_freef(&ipq[i], fpp);
1083 * If we are over the maximum number of fragments
1084 * (due to the limit being lowered), drain off
1085 * enough to get down to the new limit.
1087 if (maxnipq >= 0 && nipq > maxnipq) {
1088 for (i = 0; i < IPREASS_NHASH; i++) {
1089 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1090 ipstat.ips_fragdropped +=
1091 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1092 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1100 * Drain off all datagram fragments.
1108 for (i = 0; i < IPREASS_NHASH; i++) {
1109 while(!TAILQ_EMPTY(&ipq[i])) {
1110 ipstat.ips_fragdropped +=
1111 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1112 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1120 * The protocol to be inserted into ip_protox[] must be already registered
1121 * in inetsw[], either statically or through pf_proto_register().
1124 ipproto_register(u_char ipproto)
1128 /* Sanity checks. */
1130 return (EPROTONOSUPPORT);
1133 * The protocol slot must not be occupied by another protocol
1134 * already. An index pointing to IPPROTO_RAW is unused.
1136 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1138 return (EPFNOSUPPORT);
1139 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1142 /* Find the protocol position in inetsw[] and set the index. */
1143 for (pr = inetdomain.dom_protosw;
1144 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1145 if (pr->pr_domain->dom_family == PF_INET &&
1146 pr->pr_protocol && pr->pr_protocol == ipproto) {
1147 /* Be careful to only index valid IP protocols. */
1148 if (pr->pr_protocol < IPPROTO_MAX) {
1149 ip_protox[pr->pr_protocol] = pr - inetsw;
1155 return (EPROTONOSUPPORT);
1159 ipproto_unregister(u_char ipproto)
1163 /* Sanity checks. */
1165 return (EPROTONOSUPPORT);
1167 /* Check if the protocol was indeed registered. */
1168 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1170 return (EPFNOSUPPORT);
1171 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1174 /* Reset the protocol slot to IPPROTO_RAW. */
1175 ip_protox[ipproto] = pr - inetsw;
1180 * Given address of next destination (final or next hop),
1181 * return internet address info of interface to be used to get there.
1188 struct sockaddr_in *sin;
1189 struct in_ifaddr *ifa;
1191 bzero(&sro, sizeof(sro));
1192 sin = (struct sockaddr_in *)&sro.ro_dst;
1193 sin->sin_family = AF_INET;
1194 sin->sin_len = sizeof(*sin);
1195 sin->sin_addr = dst;
1196 rtalloc_ign(&sro, RTF_CLONING);
1198 if (sro.ro_rt == NULL)
1201 ifa = ifatoia(sro.ro_rt->rt_ifa);
1206 u_char inetctlerrmap[PRC_NCMDS] = {
1208 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1209 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1210 EMSGSIZE, EHOSTUNREACH, 0, 0,
1211 0, 0, EHOSTUNREACH, 0,
1212 ENOPROTOOPT, ECONNREFUSED
1216 * Forward a packet. If some error occurs return the sender
1217 * an icmp packet. Note we can't always generate a meaningful
1218 * icmp message because icmp doesn't have a large enough repertoire
1219 * of codes and types.
1221 * If not forwarding, just drop the packet. This could be confusing
1222 * if ipforwarding was zero but some routing protocol was advancing
1223 * us as a gateway to somewhere. However, we must let the routing
1224 * protocol deal with that.
1226 * The srcrt parameter indicates whether the packet is being forwarded
1227 * via a source route.
1230 ip_forward(struct mbuf *m, int srcrt)
1232 struct ip *ip = mtod(m, struct ip *);
1233 struct in_ifaddr *ia = NULL;
1235 struct in_addr dest;
1236 int error, type = 0, code = 0, mtu = 0;
1238 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1239 ipstat.ips_cantforward++;
1246 if (ip->ip_ttl <= IPTTLDEC) {
1247 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1255 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1256 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1261 * Save the IP header and at most 8 bytes of the payload,
1262 * in case we need to generate an ICMP message to the src.
1264 * XXX this can be optimized a lot by saving the data in a local
1265 * buffer on the stack (72 bytes at most), and only allocating the
1266 * mbuf if really necessary. The vast majority of the packets
1267 * are forwarded without having to send an ICMP back (either
1268 * because unnecessary, or because rate limited), so we are
1269 * really we are wasting a lot of work here.
1271 * We don't use m_copy() because it might return a reference
1272 * to a shared cluster. Both this function and ip_output()
1273 * assume exclusive access to the IP header in `m', so any
1274 * data in a cluster may change before we reach icmp_error().
1276 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1277 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1279 * It's probably ok if the pkthdr dup fails (because
1280 * the deep copy of the tag chain failed), but for now
1281 * be conservative and just discard the copy since
1282 * code below may some day want the tags.
1287 if (mcopy != NULL) {
1288 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1289 mcopy->m_pkthdr.len = mcopy->m_len;
1290 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1296 ip->ip_ttl -= IPTTLDEC;
1302 * If forwarding packet using same interface that it came in on,
1303 * perhaps should send a redirect to sender to shortcut a hop.
1304 * Only send redirect if source is sending directly to us,
1305 * and if packet was not source routed (or has any options).
1306 * Also, don't send redirect if forwarding using a default route
1307 * or a route modified by a redirect.
1310 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1311 struct sockaddr_in *sin;
1315 bzero(&ro, sizeof(ro));
1316 sin = (struct sockaddr_in *)&ro.ro_dst;
1317 sin->sin_family = AF_INET;
1318 sin->sin_len = sizeof(*sin);
1319 sin->sin_addr = ip->ip_dst;
1320 rtalloc_ign(&ro, RTF_CLONING);
1324 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1325 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1326 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1327 u_long src = ntohl(ip->ip_src.s_addr);
1330 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1331 if (rt->rt_flags & RTF_GATEWAY)
1332 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1334 dest.s_addr = ip->ip_dst.s_addr;
1335 /* Router requirements says to only send host redirects */
1336 type = ICMP_REDIRECT;
1337 code = ICMP_REDIRECT_HOST;
1344 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1346 ipstat.ips_cantforward++;
1348 ipstat.ips_forward++;
1350 ipstat.ips_redirectsent++;
1362 case 0: /* forwarded, but need redirect */
1363 /* type, code set above */
1366 case ENETUNREACH: /* shouldn't happen, checked above */
1371 type = ICMP_UNREACH;
1372 code = ICMP_UNREACH_HOST;
1376 type = ICMP_UNREACH;
1377 code = ICMP_UNREACH_NEEDFRAG;
1379 #if defined(IPSEC) || defined(FAST_IPSEC)
1380 mtu = ip_ipsec_mtu(m);
1383 * If the MTU wasn't set before use the interface mtu or
1384 * fall back to the next smaller mtu step compared to the
1385 * current packet size.
1389 mtu = ia->ia_ifp->if_mtu;
1391 mtu = ip_next_mtu(ip->ip_len, 0);
1393 ipstat.ips_cantfrag++;
1398 * A router should not generate ICMP_SOURCEQUENCH as
1399 * required in RFC1812 Requirements for IP Version 4 Routers.
1400 * Source quench could be a big problem under DoS attacks,
1401 * or if the underlying interface is rate-limited.
1402 * Those who need source quench packets may re-enable them
1403 * via the net.inet.ip.sendsourcequench sysctl.
1405 if (ip_sendsourcequench == 0) {
1409 type = ICMP_SOURCEQUENCH;
1414 case EACCES: /* ipfw denied packet */
1418 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1422 ip_savecontrol(inp, mp, ip, m)
1423 register struct inpcb *inp;
1424 register struct mbuf **mp;
1425 register struct ip *ip;
1426 register struct mbuf *m;
1428 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1432 if (inp->inp_socket->so_options & SO_BINTIME) {
1433 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1434 SCM_BINTIME, SOL_SOCKET);
1436 mp = &(*mp)->m_next;
1438 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1441 bintime2timeval(&bt, &tv);
1442 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1443 SCM_TIMESTAMP, SOL_SOCKET);
1445 mp = &(*mp)->m_next;
1448 if (inp->inp_flags & INP_RECVDSTADDR) {
1449 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1450 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1452 mp = &(*mp)->m_next;
1454 if (inp->inp_flags & INP_RECVTTL) {
1455 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1456 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1458 mp = &(*mp)->m_next;
1462 * Moving these out of udp_input() made them even more broken
1463 * than they already were.
1465 /* options were tossed already */
1466 if (inp->inp_flags & INP_RECVOPTS) {
1467 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1468 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1470 mp = &(*mp)->m_next;
1472 /* ip_srcroute doesn't do what we want here, need to fix */
1473 if (inp->inp_flags & INP_RECVRETOPTS) {
1474 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1475 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1477 mp = &(*mp)->m_next;
1480 if (inp->inp_flags & INP_RECVIF) {
1483 struct sockaddr_dl sdl;
1486 struct sockaddr_dl *sdp;
1487 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1489 if (((ifp = m->m_pkthdr.rcvif))
1490 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1491 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1493 * Change our mind and don't try copy.
1495 if ((sdp->sdl_family != AF_LINK)
1496 || (sdp->sdl_len > sizeof(sdlbuf))) {
1499 bcopy(sdp, sdl2, sdp->sdl_len);
1503 = offsetof(struct sockaddr_dl, sdl_data[0]);
1504 sdl2->sdl_family = AF_LINK;
1505 sdl2->sdl_index = 0;
1506 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1508 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1509 IP_RECVIF, IPPROTO_IP);
1511 mp = &(*mp)->m_next;
1516 * XXX these routines are called from the upper part of the kernel.
1517 * They need to be locked when we remove Giant.
1519 * They could also be moved to ip_mroute.c, since all the RSVP
1520 * handling is done there already.
1522 static int ip_rsvp_on;
1523 struct socket *ip_rsvpd;
1525 ip_rsvp_init(struct socket *so)
1527 if (so->so_type != SOCK_RAW ||
1528 so->so_proto->pr_protocol != IPPROTO_RSVP)
1531 if (ip_rsvpd != NULL)
1536 * This may seem silly, but we need to be sure we don't over-increment
1537 * the RSVP counter, in case something slips up.
1552 * This may seem silly, but we need to be sure we don't over-decrement
1553 * the RSVP counter, in case something slips up.
1563 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1565 if (rsvp_input_p) { /* call the real one if loaded */
1566 rsvp_input_p(m, off);
1570 /* Can still get packets with rsvp_on = 0 if there is a local member
1571 * of the group to which the RSVP packet is addressed. But in this
1572 * case we want to throw the packet away.
1580 if (ip_rsvpd != NULL) {
1584 /* Drop the packet */