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>
83 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
84 &ipforwarding, 0, "Enable IP forwarding between interfaces");
86 static int ipsendredirects = 1; /* XXX */
87 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
88 &ipsendredirects, 0, "Enable sending IP redirects");
90 int ip_defttl = IPDEFTTL;
91 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
92 &ip_defttl, 0, "Maximum TTL on IP packets");
94 static int ip_keepfaith = 0;
95 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
97 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
99 static int ip_sendsourcequench = 0;
100 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
101 &ip_sendsourcequench, 0,
102 "Enable the transmission of source quench packets");
104 int ip_do_randomid = 0;
105 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
107 "Assign random ip_id values");
110 * XXX - Setting ip_checkinterface mostly implements the receive side of
111 * the Strong ES model described in RFC 1122, but since the routing table
112 * and transmit implementation do not implement the Strong ES model,
113 * setting this to 1 results in an odd hybrid.
115 * XXX - ip_checkinterface currently must be disabled if you use ipnat
116 * to translate the destination address to another local interface.
118 * XXX - ip_checkinterface must be disabled if you add IP aliases
119 * to the loopback interface instead of the interface where the
120 * packets for those addresses are received.
122 static int ip_checkinterface = 0;
123 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
124 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
126 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
128 static struct ifqueue ipintrq;
129 static int ipqmaxlen = IFQ_MAXLEN;
131 extern struct domain inetdomain;
132 extern struct protosw inetsw[];
133 u_char ip_protox[IPPROTO_MAX];
134 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
135 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
136 u_long in_ifaddrhmask; /* mask for hash table */
138 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
139 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
140 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
141 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
143 struct ipstat ipstat;
144 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
145 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
148 * IP datagram reassembly.
150 #define IPREASS_NHASH_LOG2 6
151 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
152 #define IPREASS_HMASK (IPREASS_NHASH - 1)
153 #define IPREASS_HASH(x,y) \
154 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
156 static uma_zone_t ipq_zone;
157 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
158 static struct mtx ipqlock;
160 #define IPQ_LOCK() mtx_lock(&ipqlock)
161 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
162 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
163 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
165 static void maxnipq_update(void);
166 static void ipq_zone_change(void *);
168 static int maxnipq; /* Administrative limit on # reass queues. */
169 static int nipq = 0; /* Total # of reass queues */
170 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0,
171 "Current number of IPv4 fragment reassembly queue entries");
173 static int maxfragsperpacket;
174 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
175 &maxfragsperpacket, 0,
176 "Maximum number of IPv4 fragments allowed per packet");
178 struct callout ipport_tick_callout;
181 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
182 &ip_mtu, 0, "Default MTU");
187 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
191 static void ip_freef(struct ipqhead *, struct ipq *);
194 * IP initialization: fill in IP protocol switch table.
195 * All protocols not implemented in kernel go to raw IP protocol handler.
200 register struct protosw *pr;
203 TAILQ_INIT(&in_ifaddrhead);
204 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
205 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
207 panic("ip_init: PF_INET not found");
209 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
210 for (i = 0; i < IPPROTO_MAX; i++)
211 ip_protox[i] = pr - inetsw;
213 * Cycle through IP protocols and put them into the appropriate place
216 for (pr = inetdomain.dom_protosw;
217 pr < inetdomain.dom_protoswNPROTOSW; pr++)
218 if (pr->pr_domain->dom_family == PF_INET &&
219 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
220 /* Be careful to only index valid IP protocols. */
221 if (pr->pr_protocol < IPPROTO_MAX)
222 ip_protox[pr->pr_protocol] = pr - inetsw;
225 /* Initialize packet filter hooks. */
226 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
227 inet_pfil_hook.ph_af = AF_INET;
228 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
229 printf("%s: WARNING: unable to register pfil hook, "
230 "error %d\n", __func__, i);
232 /* Initialize IP reassembly queue. */
234 for (i = 0; i < IPREASS_NHASH; i++)
236 maxnipq = nmbclusters / 32;
237 maxfragsperpacket = 16;
238 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
239 NULL, UMA_ALIGN_PTR, 0);
242 /* Start ipport_tick. */
243 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
245 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
246 SHUTDOWN_PRI_DEFAULT);
247 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
248 NULL, EVENTHANDLER_PRI_ANY);
250 /* Initialize various other remaining things. */
251 ip_id = time_second & 0xffff;
252 ipintrq.ifq_maxlen = ipqmaxlen;
253 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
254 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
260 callout_stop(&ipport_tick_callout);
264 * Ip input routine. Checksum and byte swap header. If fragmented
265 * try to reassemble. Process options. Pass to next level.
268 ip_input(struct mbuf *m)
270 struct ip *ip = NULL;
271 struct in_ifaddr *ia = NULL;
273 int checkif, hlen = 0;
275 int dchg = 0; /* dest changed after fw */
276 struct in_addr odst; /* original dst address */
280 if (m->m_flags & M_FASTFWD_OURS) {
282 * Firewall or NAT changed destination to local.
283 * We expect ip_len and ip_off to be in host byte order.
285 m->m_flags &= ~M_FASTFWD_OURS;
286 /* Set up some basics that will be used later. */
287 ip = mtod(m, struct ip *);
288 hlen = ip->ip_hl << 2;
294 if (m->m_pkthdr.len < sizeof(struct ip))
297 if (m->m_len < sizeof (struct ip) &&
298 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
299 ipstat.ips_toosmall++;
302 ip = mtod(m, struct ip *);
304 if (ip->ip_v != IPVERSION) {
305 ipstat.ips_badvers++;
309 hlen = ip->ip_hl << 2;
310 if (hlen < sizeof(struct ip)) { /* minimum header length */
311 ipstat.ips_badhlen++;
314 if (hlen > m->m_len) {
315 if ((m = m_pullup(m, hlen)) == NULL) {
316 ipstat.ips_badhlen++;
319 ip = mtod(m, struct ip *);
322 /* 127/8 must not appear on wire - RFC1122 */
323 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
324 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
325 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
326 ipstat.ips_badaddr++;
331 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
332 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
334 if (hlen == sizeof(struct ip)) {
335 sum = in_cksum_hdr(ip);
337 sum = in_cksum(m, hlen);
346 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
347 /* packet is dropped by traffic conditioner */
352 * Convert fields to host representation.
354 ip->ip_len = ntohs(ip->ip_len);
355 if (ip->ip_len < hlen) {
359 ip->ip_off = ntohs(ip->ip_off);
362 * Check that the amount of data in the buffers
363 * is as at least much as the IP header would have us expect.
364 * Trim mbufs if longer than we expect.
365 * Drop packet if shorter than we expect.
367 if (m->m_pkthdr.len < ip->ip_len) {
369 ipstat.ips_tooshort++;
372 if (m->m_pkthdr.len > ip->ip_len) {
373 if (m->m_len == m->m_pkthdr.len) {
374 m->m_len = ip->ip_len;
375 m->m_pkthdr.len = ip->ip_len;
377 m_adj(m, ip->ip_len - m->m_pkthdr.len);
379 #if defined(IPSEC) || defined(FAST_IPSEC)
381 * Bypass packet filtering for packets from a tunnel (gif).
383 if (ip_ipsec_filtergif(m))
388 * Run through list of hooks for input packets.
390 * NB: Beware of the destination address changing (e.g.
391 * by NAT rewriting). When this happens, tell
392 * ip_forward to do the right thing.
395 /* Jump over all PFIL processing if hooks are not active. */
396 if (!PFIL_HOOKED(&inet_pfil_hook))
400 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
403 if (m == NULL) /* consumed by filter */
406 ip = mtod(m, struct ip *);
407 dchg = (odst.s_addr != ip->ip_dst.s_addr);
409 #ifdef IPFIREWALL_FORWARD
410 if (m->m_flags & M_FASTFWD_OURS) {
411 m->m_flags &= ~M_FASTFWD_OURS;
414 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
416 * Directly ship on the packet. This allows to forward packets
417 * that were destined for us to some other directly connected
423 #endif /* IPFIREWALL_FORWARD */
427 * Process options and, if not destined for us,
428 * ship it on. ip_dooptions returns 1 when an
429 * error was detected (causing an icmp message
430 * to be sent and the original packet to be freed).
432 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
435 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
436 * matter if it is destined to another node, or whether it is
437 * a multicast one, RSVP wants it! and prevents it from being forwarded
438 * anywhere else. Also checks if the rsvp daemon is running before
439 * grabbing the packet.
441 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
445 * Check our list of addresses, to see if the packet is for us.
446 * If we don't have any addresses, assume any unicast packet
447 * we receive might be for us (and let the upper layers deal
450 if (TAILQ_EMPTY(&in_ifaddrhead) &&
451 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
455 * Enable a consistency check between the destination address
456 * and the arrival interface for a unicast packet (the RFC 1122
457 * strong ES model) if IP forwarding is disabled and the packet
458 * is not locally generated and the packet is not subject to
461 * XXX - Checking also should be disabled if the destination
462 * address is ipnat'ed to a different interface.
464 * XXX - Checking is incompatible with IP aliases added
465 * to the loopback interface instead of the interface where
466 * the packets are received.
468 * XXX - This is the case for carp vhost IPs as well so we
469 * insert a workaround. If the packet got here, we already
470 * checked with carp_iamatch() and carp_forus().
472 checkif = ip_checkinterface && (ipforwarding == 0) &&
473 m->m_pkthdr.rcvif != NULL &&
474 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
476 !m->m_pkthdr.rcvif->if_carp &&
481 * Check for exact addresses in the hash bucket.
483 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
485 * If the address matches, verify that the packet
486 * arrived via the correct interface if checking is
489 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
490 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
494 * Check for broadcast addresses.
496 * Only accept broadcast packets that arrive via the matching
497 * interface. Reception of forwarded directed broadcasts would
498 * be handled via ip_forward() and ether_output() with the loopback
499 * into the stack for SIMPLEX interfaces handled by ether_output().
501 if (m->m_pkthdr.rcvif != NULL &&
502 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
503 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
504 if (ifa->ifa_addr->sa_family != AF_INET)
507 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
510 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
513 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
518 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
519 struct in_multi *inm;
522 * If we are acting as a multicast router, all
523 * incoming multicast packets are passed to the
524 * kernel-level multicast forwarding function.
525 * The packet is returned (relatively) intact; if
526 * ip_mforward() returns a non-zero value, the packet
527 * must be discarded, else it may be accepted below.
530 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
531 ipstat.ips_cantforward++;
537 * The process-level routing daemon needs to receive
538 * all multicast IGMP packets, whether or not this
539 * host belongs to their destination groups.
541 if (ip->ip_p == IPPROTO_IGMP)
543 ipstat.ips_forward++;
546 * See if we belong to the destination multicast group on the
550 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
553 ipstat.ips_notmember++;
559 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
561 if (ip->ip_dst.s_addr == INADDR_ANY)
565 * FAITH(Firewall Aided Internet Translator)
567 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
569 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
577 * Not for us; forward if possible and desirable.
579 if (ipforwarding == 0) {
580 ipstat.ips_cantforward++;
583 #if defined(IPSEC) || defined(FAST_IPSEC)
594 * IPSTEALTH: Process non-routing options only
595 * if the packet is destined for us.
597 if (ipstealth && hlen > sizeof (struct ip) &&
600 #endif /* IPSTEALTH */
602 /* Count the packet in the ip address stats */
604 ia->ia_ifa.if_ipackets++;
605 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
609 * Attempt reassembly; if it succeeds, proceed.
610 * ip_reass() will return a different mbuf.
612 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
616 ip = mtod(m, struct ip *);
617 /* Get the header length of the reassembled packet */
618 hlen = ip->ip_hl << 2;
622 * Further protocols expect the packet length to be w/o the
627 #if defined(IPSEC) || defined(FAST_IPSEC)
629 * enforce IPsec policy checking if we are seeing last header.
630 * note that we do not visit this with protocols with pcb layer
631 * code - like udp/tcp/raw ip.
633 if (ip_ipsec_input(m))
638 * Switch out to protocol's input routine.
640 ipstat.ips_delivered++;
642 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
649 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
650 * max has slightly different semantics than the sysctl, for historical
658 * -1 for unlimited allocation.
661 uma_zone_set_max(ipq_zone, 0);
663 * Positive number for specific bound.
666 uma_zone_set_max(ipq_zone, maxnipq);
668 * Zero specifies no further fragment queue allocation -- set the
669 * bound very low, but rely on implementation elsewhere to actually
670 * prevent allocation and reclaim current queues.
673 uma_zone_set_max(ipq_zone, 1);
677 ipq_zone_change(void *tag)
680 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
681 maxnipq = nmbclusters / 32;
687 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
692 error = sysctl_handle_int(oidp, &i, 0, req);
693 if (error || !req->newptr)
697 * XXXRW: Might be a good idea to sanity check the argument and place
698 * an extreme upper bound.
707 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
708 NULL, 0, sysctl_maxnipq, "I",
709 "Maximum number of IPv4 fragment reassembly queue entries");
712 * Take incoming datagram fragment and try to reassemble it into
713 * whole datagram. If the argument is the first fragment or one
714 * in between the function will return NULL and store the mbuf
715 * in the fragment chain. If the argument is the last fragment
716 * the packet will be reassembled and the pointer to the new
717 * mbuf returned for further processing. Only m_tags attached
718 * to the first packet/fragment are preserved.
719 * The IP header is *NOT* adjusted out of iplen.
723 ip_reass(struct mbuf *m)
726 struct mbuf *p, *q, *nq, *t;
727 struct ipq *fp = NULL;
728 struct ipqhead *head;
733 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
734 if (maxnipq == 0 || maxfragsperpacket == 0) {
735 ipstat.ips_fragments++;
736 ipstat.ips_fragdropped++;
741 ip = mtod(m, struct ip *);
742 hlen = ip->ip_hl << 2;
744 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
749 * Look for queue of fragments
752 TAILQ_FOREACH(fp, head, ipq_list)
753 if (ip->ip_id == fp->ipq_id &&
754 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
755 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
757 mac_fragment_match(m, fp) &&
759 ip->ip_p == fp->ipq_p)
765 * Attempt to trim the number of allocated fragment queues if it
766 * exceeds the administrative limit.
768 if ((nipq > maxnipq) && (maxnipq > 0)) {
770 * drop something from the tail of the current queue
771 * before proceeding further
773 struct ipq *q = TAILQ_LAST(head, ipqhead);
774 if (q == NULL) { /* gak */
775 for (i = 0; i < IPREASS_NHASH; i++) {
776 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
778 ipstat.ips_fragtimeout += r->ipq_nfrags;
779 ip_freef(&ipq[i], r);
784 ipstat.ips_fragtimeout += q->ipq_nfrags;
791 * Adjust ip_len to not reflect header,
792 * convert offset of this to bytes.
795 if (ip->ip_off & IP_MF) {
797 * Make sure that fragments have a data length
798 * that's a non-zero multiple of 8 bytes.
800 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
801 ipstat.ips_toosmall++; /* XXX */
804 m->m_flags |= M_FRAG;
806 m->m_flags &= ~M_FRAG;
811 * Attempt reassembly; if it succeeds, proceed.
812 * ip_reass() will return a different mbuf.
814 ipstat.ips_fragments++;
815 m->m_pkthdr.header = ip;
817 /* Previous ip_reass() started here. */
819 * Presence of header sizes in mbufs
820 * would confuse code below.
826 * If first fragment to arrive, create a reassembly queue.
829 fp = uma_zalloc(ipq_zone, M_NOWAIT);
833 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
834 uma_zfree(ipq_zone, fp);
838 mac_create_ipq(m, fp);
840 TAILQ_INSERT_HEAD(head, fp, ipq_list);
843 fp->ipq_ttl = IPFRAGTTL;
844 fp->ipq_p = ip->ip_p;
845 fp->ipq_id = ip->ip_id;
846 fp->ipq_src = ip->ip_src;
847 fp->ipq_dst = ip->ip_dst;
854 mac_update_ipq(m, fp);
858 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
861 * Handle ECN by comparing this segment with the first one;
862 * if CE is set, do not lose CE.
863 * drop if CE and not-ECT are mixed for the same packet.
865 ecn = ip->ip_tos & IPTOS_ECN_MASK;
866 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
867 if (ecn == IPTOS_ECN_CE) {
868 if (ecn0 == IPTOS_ECN_NOTECT)
870 if (ecn0 != IPTOS_ECN_CE)
871 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
873 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
877 * Find a segment which begins after this one does.
879 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
880 if (GETIP(q)->ip_off > ip->ip_off)
884 * If there is a preceding segment, it may provide some of
885 * our data already. If so, drop the data from the incoming
886 * segment. If it provides all of our data, drop us, otherwise
887 * stick new segment in the proper place.
889 * If some of the data is dropped from the the preceding
890 * segment, then it's checksum is invalidated.
893 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
898 m->m_pkthdr.csum_flags = 0;
902 m->m_nextpkt = p->m_nextpkt;
905 m->m_nextpkt = fp->ipq_frags;
910 * While we overlap succeeding segments trim them or,
911 * if they are completely covered, dequeue them.
913 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
915 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
916 if (i < GETIP(q)->ip_len) {
917 GETIP(q)->ip_len -= i;
918 GETIP(q)->ip_off += i;
920 q->m_pkthdr.csum_flags = 0;
925 ipstat.ips_fragdropped++;
931 * Check for complete reassembly and perform frag per packet
934 * Frag limiting is performed here so that the nth frag has
935 * a chance to complete the packet before we drop the packet.
936 * As a result, n+1 frags are actually allowed per packet, but
937 * only n will ever be stored. (n = maxfragsperpacket.)
941 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
942 if (GETIP(q)->ip_off != next) {
943 if (fp->ipq_nfrags > maxfragsperpacket) {
944 ipstat.ips_fragdropped += fp->ipq_nfrags;
949 next += GETIP(q)->ip_len;
951 /* Make sure the last packet didn't have the IP_MF flag */
952 if (p->m_flags & M_FRAG) {
953 if (fp->ipq_nfrags > maxfragsperpacket) {
954 ipstat.ips_fragdropped += fp->ipq_nfrags;
961 * Reassembly is complete. Make sure the packet is a sane size.
965 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
966 ipstat.ips_toolong++;
967 ipstat.ips_fragdropped += fp->ipq_nfrags;
973 * Concatenate fragments.
981 for (q = nq; q != NULL; q = nq) {
984 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
985 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
989 * In order to do checksumming faster we do 'end-around carry' here
990 * (and not in for{} loop), though it implies we are not going to
991 * reassemble more than 64k fragments.
993 m->m_pkthdr.csum_data =
994 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
996 mac_create_datagram_from_ipq(fp, m);
1001 * Create header for new ip packet by modifying header of first
1002 * packet; dequeue and discard fragment reassembly header.
1003 * Make header visible.
1005 ip->ip_len = (ip->ip_hl << 2) + next;
1006 ip->ip_src = fp->ipq_src;
1007 ip->ip_dst = fp->ipq_dst;
1008 TAILQ_REMOVE(head, fp, ipq_list);
1010 uma_zfree(ipq_zone, fp);
1011 m->m_len += (ip->ip_hl << 2);
1012 m->m_data -= (ip->ip_hl << 2);
1013 /* some debugging cruft by sklower, below, will go away soon */
1014 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1016 ipstat.ips_reassembled++;
1021 ipstat.ips_fragdropped++;
1033 * Free a fragment reassembly header and all
1034 * associated datagrams.
1038 struct ipqhead *fhp;
1041 register struct mbuf *q;
1045 while (fp->ipq_frags) {
1047 fp->ipq_frags = q->m_nextpkt;
1050 TAILQ_REMOVE(fhp, fp, ipq_list);
1051 uma_zfree(ipq_zone, fp);
1056 * IP timer processing;
1057 * if a timer expires on a reassembly
1058 * queue, discard it.
1063 register struct ipq *fp;
1067 for (i = 0; i < IPREASS_NHASH; i++) {
1068 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1072 fp = TAILQ_NEXT(fp, ipq_list);
1073 if(--fpp->ipq_ttl == 0) {
1074 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1075 ip_freef(&ipq[i], fpp);
1080 * If we are over the maximum number of fragments
1081 * (due to the limit being lowered), drain off
1082 * enough to get down to the new limit.
1084 if (maxnipq >= 0 && nipq > maxnipq) {
1085 for (i = 0; i < IPREASS_NHASH; i++) {
1086 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1087 ipstat.ips_fragdropped +=
1088 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1089 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1097 * Drain off all datagram fragments.
1105 for (i = 0; i < IPREASS_NHASH; i++) {
1106 while(!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 * The protocol to be inserted into ip_protox[] must be already registered
1118 * in inetsw[], either statically or through pf_proto_register().
1121 ipproto_register(u_char ipproto)
1125 /* Sanity checks. */
1127 return (EPROTONOSUPPORT);
1130 * The protocol slot must not be occupied by another protocol
1131 * already. An index pointing to IPPROTO_RAW is unused.
1133 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1135 return (EPFNOSUPPORT);
1136 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1139 /* Find the protocol position in inetsw[] and set the index. */
1140 for (pr = inetdomain.dom_protosw;
1141 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1142 if (pr->pr_domain->dom_family == PF_INET &&
1143 pr->pr_protocol && pr->pr_protocol == ipproto) {
1144 /* Be careful to only index valid IP protocols. */
1145 if (pr->pr_protocol < IPPROTO_MAX) {
1146 ip_protox[pr->pr_protocol] = pr - inetsw;
1152 return (EPROTONOSUPPORT);
1156 ipproto_unregister(u_char ipproto)
1160 /* Sanity checks. */
1162 return (EPROTONOSUPPORT);
1164 /* Check if the protocol was indeed registered. */
1165 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1167 return (EPFNOSUPPORT);
1168 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1171 /* Reset the protocol slot to IPPROTO_RAW. */
1172 ip_protox[ipproto] = pr - inetsw;
1177 * Given address of next destination (final or next hop),
1178 * return internet address info of interface to be used to get there.
1185 struct sockaddr_in *sin;
1186 struct in_ifaddr *ifa;
1188 bzero(&sro, sizeof(sro));
1189 sin = (struct sockaddr_in *)&sro.ro_dst;
1190 sin->sin_family = AF_INET;
1191 sin->sin_len = sizeof(*sin);
1192 sin->sin_addr = dst;
1193 rtalloc_ign(&sro, RTF_CLONING);
1195 if (sro.ro_rt == NULL)
1198 ifa = ifatoia(sro.ro_rt->rt_ifa);
1203 u_char inetctlerrmap[PRC_NCMDS] = {
1205 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1206 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1207 EMSGSIZE, EHOSTUNREACH, 0, 0,
1208 0, 0, EHOSTUNREACH, 0,
1209 ENOPROTOOPT, ECONNREFUSED
1213 * Forward a packet. If some error occurs return the sender
1214 * an icmp packet. Note we can't always generate a meaningful
1215 * icmp message because icmp doesn't have a large enough repertoire
1216 * of codes and types.
1218 * If not forwarding, just drop the packet. This could be confusing
1219 * if ipforwarding was zero but some routing protocol was advancing
1220 * us as a gateway to somewhere. However, we must let the routing
1221 * protocol deal with that.
1223 * The srcrt parameter indicates whether the packet is being forwarded
1224 * via a source route.
1227 ip_forward(struct mbuf *m, int srcrt)
1229 struct ip *ip = mtod(m, struct ip *);
1230 struct in_ifaddr *ia = NULL;
1232 struct in_addr dest;
1233 int error, type = 0, code = 0, mtu = 0;
1235 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1236 ipstat.ips_cantforward++;
1243 if (ip->ip_ttl <= IPTTLDEC) {
1244 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1252 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1253 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1258 * Save the IP header and at most 8 bytes of the payload,
1259 * in case we need to generate an ICMP message to the src.
1261 * XXX this can be optimized a lot by saving the data in a local
1262 * buffer on the stack (72 bytes at most), and only allocating the
1263 * mbuf if really necessary. The vast majority of the packets
1264 * are forwarded without having to send an ICMP back (either
1265 * because unnecessary, or because rate limited), so we are
1266 * really we are wasting a lot of work here.
1268 * We don't use m_copy() because it might return a reference
1269 * to a shared cluster. Both this function and ip_output()
1270 * assume exclusive access to the IP header in `m', so any
1271 * data in a cluster may change before we reach icmp_error().
1273 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1274 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1276 * It's probably ok if the pkthdr dup fails (because
1277 * the deep copy of the tag chain failed), but for now
1278 * be conservative and just discard the copy since
1279 * code below may some day want the tags.
1284 if (mcopy != NULL) {
1285 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1286 mcopy->m_pkthdr.len = mcopy->m_len;
1287 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1293 ip->ip_ttl -= IPTTLDEC;
1299 * If forwarding packet using same interface that it came in on,
1300 * perhaps should send a redirect to sender to shortcut a hop.
1301 * Only send redirect if source is sending directly to us,
1302 * and if packet was not source routed (or has any options).
1303 * Also, don't send redirect if forwarding using a default route
1304 * or a route modified by a redirect.
1307 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1308 struct sockaddr_in *sin;
1312 bzero(&ro, sizeof(ro));
1313 sin = (struct sockaddr_in *)&ro.ro_dst;
1314 sin->sin_family = AF_INET;
1315 sin->sin_len = sizeof(*sin);
1316 sin->sin_addr = ip->ip_dst;
1317 rtalloc_ign(&ro, RTF_CLONING);
1321 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1322 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1323 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1324 u_long src = ntohl(ip->ip_src.s_addr);
1327 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1328 if (rt->rt_flags & RTF_GATEWAY)
1329 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1331 dest.s_addr = ip->ip_dst.s_addr;
1332 /* Router requirements says to only send host redirects */
1333 type = ICMP_REDIRECT;
1334 code = ICMP_REDIRECT_HOST;
1341 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1343 ipstat.ips_cantforward++;
1345 ipstat.ips_forward++;
1347 ipstat.ips_redirectsent++;
1359 case 0: /* forwarded, but need redirect */
1360 /* type, code set above */
1363 case ENETUNREACH: /* shouldn't happen, checked above */
1368 type = ICMP_UNREACH;
1369 code = ICMP_UNREACH_HOST;
1373 type = ICMP_UNREACH;
1374 code = ICMP_UNREACH_NEEDFRAG;
1376 #if defined(IPSEC) || defined(FAST_IPSEC)
1377 mtu = ip_ipsec_mtu(m);
1380 * If the MTU wasn't set before use the interface mtu or
1381 * fall back to the next smaller mtu step compared to the
1382 * current packet size.
1386 mtu = ia->ia_ifp->if_mtu;
1388 mtu = ip_next_mtu(ip->ip_len, 0);
1390 ipstat.ips_cantfrag++;
1395 * A router should not generate ICMP_SOURCEQUENCH as
1396 * required in RFC1812 Requirements for IP Version 4 Routers.
1397 * Source quench could be a big problem under DoS attacks,
1398 * or if the underlying interface is rate-limited.
1399 * Those who need source quench packets may re-enable them
1400 * via the net.inet.ip.sendsourcequench sysctl.
1402 if (ip_sendsourcequench == 0) {
1406 type = ICMP_SOURCEQUENCH;
1411 case EACCES: /* ipfw denied packet */
1415 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1419 ip_savecontrol(inp, mp, ip, m)
1420 register struct inpcb *inp;
1421 register struct mbuf **mp;
1422 register struct ip *ip;
1423 register struct mbuf *m;
1425 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1429 if (inp->inp_socket->so_options & SO_BINTIME) {
1430 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1431 SCM_BINTIME, SOL_SOCKET);
1433 mp = &(*mp)->m_next;
1435 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1438 bintime2timeval(&bt, &tv);
1439 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1440 SCM_TIMESTAMP, SOL_SOCKET);
1442 mp = &(*mp)->m_next;
1445 if (inp->inp_flags & INP_RECVDSTADDR) {
1446 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1447 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1449 mp = &(*mp)->m_next;
1451 if (inp->inp_flags & INP_RECVTTL) {
1452 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1453 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1455 mp = &(*mp)->m_next;
1459 * Moving these out of udp_input() made them even more broken
1460 * than they already were.
1462 /* options were tossed already */
1463 if (inp->inp_flags & INP_RECVOPTS) {
1464 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1465 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1467 mp = &(*mp)->m_next;
1469 /* ip_srcroute doesn't do what we want here, need to fix */
1470 if (inp->inp_flags & INP_RECVRETOPTS) {
1471 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1472 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1474 mp = &(*mp)->m_next;
1477 if (inp->inp_flags & INP_RECVIF) {
1480 struct sockaddr_dl sdl;
1483 struct sockaddr_dl *sdp;
1484 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1486 if (((ifp = m->m_pkthdr.rcvif))
1487 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1488 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1490 * Change our mind and don't try copy.
1492 if ((sdp->sdl_family != AF_LINK)
1493 || (sdp->sdl_len > sizeof(sdlbuf))) {
1496 bcopy(sdp, sdl2, sdp->sdl_len);
1500 = offsetof(struct sockaddr_dl, sdl_data[0]);
1501 sdl2->sdl_family = AF_LINK;
1502 sdl2->sdl_index = 0;
1503 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1505 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1506 IP_RECVIF, IPPROTO_IP);
1508 mp = &(*mp)->m_next;
1513 * XXX these routines are called from the upper part of the kernel.
1514 * They need to be locked when we remove Giant.
1516 * They could also be moved to ip_mroute.c, since all the RSVP
1517 * handling is done there already.
1519 static int ip_rsvp_on;
1520 struct socket *ip_rsvpd;
1522 ip_rsvp_init(struct socket *so)
1524 if (so->so_type != SOCK_RAW ||
1525 so->so_proto->pr_protocol != IPPROTO_RSVP)
1528 if (ip_rsvpd != NULL)
1533 * This may seem silly, but we need to be sure we don't over-increment
1534 * the RSVP counter, in case something slips up.
1549 * This may seem silly, but we need to be sure we don't over-decrement
1550 * the RSVP counter, in case something slips up.
1560 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1562 if (rsvp_input_p) { /* call the real one if loaded */
1563 rsvp_input_p(m, off);
1567 /* Can still get packets with rsvp_on = 0 if there is a local member
1568 * of the group to which the RSVP packet is addressed. But in this
1569 * case we want to throw the packet away.
1577 if (ip_rsvpd != NULL) {
1581 /* Drop the packet */