2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_bootp.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/callout.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
51 #include <sys/kernel.h>
52 #include <sys/syslog.h>
53 #include <sys/sysctl.h>
57 #include <net/if_types.h>
58 #include <net/if_var.h>
59 #include <net/if_dl.h>
60 #include <net/route.h>
61 #include <net/netisr.h>
63 #include <netinet/in.h>
64 #include <netinet/in_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>
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>
84 #include <security/mac/mac_framework.h>
89 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
90 &ipforwarding, 0, "Enable IP forwarding between interfaces");
92 static int ipsendredirects = 1; /* XXX */
93 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
94 &ipsendredirects, 0, "Enable sending IP redirects");
96 int ip_defttl = IPDEFTTL;
97 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
98 &ip_defttl, 0, "Maximum TTL on IP packets");
100 static int ip_keepfaith = 0;
101 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
103 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
105 static int ip_sendsourcequench = 0;
106 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
107 &ip_sendsourcequench, 0,
108 "Enable the transmission of source quench packets");
110 int ip_do_randomid = 0;
111 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
113 "Assign random ip_id values");
116 * XXX - Setting ip_checkinterface mostly implements the receive side of
117 * the Strong ES model described in RFC 1122, but since the routing table
118 * and transmit implementation do not implement the Strong ES model,
119 * setting this to 1 results in an odd hybrid.
121 * XXX - ip_checkinterface currently must be disabled if you use ipnat
122 * to translate the destination address to another local interface.
124 * XXX - ip_checkinterface must be disabled if you add IP aliases
125 * to the loopback interface instead of the interface where the
126 * packets for those addresses are received.
128 static int ip_checkinterface = 0;
129 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
130 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
132 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
134 static struct ifqueue ipintrq;
135 static int ipqmaxlen = IFQ_MAXLEN;
137 extern struct domain inetdomain;
138 extern struct protosw inetsw[];
139 u_char ip_protox[IPPROTO_MAX];
140 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
141 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
142 u_long in_ifaddrhmask; /* mask for hash table */
144 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
145 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
146 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
147 &ipintrq.ifq_drops, 0,
148 "Number of packets dropped from the IP input queue");
150 struct ipstat ipstat;
151 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
152 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
155 * IP datagram reassembly.
157 #define IPREASS_NHASH_LOG2 6
158 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
159 #define IPREASS_HMASK (IPREASS_NHASH - 1)
160 #define IPREASS_HASH(x,y) \
161 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
163 static uma_zone_t ipq_zone;
164 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
165 static struct mtx ipqlock;
167 #define IPQ_LOCK() mtx_lock(&ipqlock)
168 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
169 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
170 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
172 static void maxnipq_update(void);
173 static void ipq_zone_change(void *);
175 static int maxnipq; /* Administrative limit on # reass queues. */
176 static int nipq = 0; /* Total # of reass queues */
177 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
178 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
180 static int maxfragsperpacket;
181 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
182 &maxfragsperpacket, 0,
183 "Maximum number of IPv4 fragments allowed per packet");
185 struct callout ipport_tick_callout;
188 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
189 &ip_mtu, 0, "Default MTU");
194 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
195 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
199 * ipfw_ether and ipfw_bridge hooks.
200 * XXX: Temporary until those are converted to pfil_hooks as well.
202 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
203 ip_dn_io_t *ip_dn_io_ptr = NULL;
206 static void ip_freef(struct ipqhead *, struct ipq *);
209 * IP initialization: fill in IP protocol switch table.
210 * All protocols not implemented in kernel go to raw IP protocol handler.
218 TAILQ_INIT(&in_ifaddrhead);
219 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
220 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
222 panic("ip_init: PF_INET not found");
224 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
225 for (i = 0; i < IPPROTO_MAX; i++)
226 ip_protox[i] = pr - inetsw;
228 * Cycle through IP protocols and put them into the appropriate place
231 for (pr = inetdomain.dom_protosw;
232 pr < inetdomain.dom_protoswNPROTOSW; pr++)
233 if (pr->pr_domain->dom_family == PF_INET &&
234 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
235 /* Be careful to only index valid IP protocols. */
236 if (pr->pr_protocol < IPPROTO_MAX)
237 ip_protox[pr->pr_protocol] = pr - inetsw;
240 /* Initialize packet filter hooks. */
241 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
242 inet_pfil_hook.ph_af = AF_INET;
243 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
244 printf("%s: WARNING: unable to register pfil hook, "
245 "error %d\n", __func__, i);
247 /* Initialize IP reassembly queue. */
249 for (i = 0; i < IPREASS_NHASH; i++)
251 maxnipq = nmbclusters / 32;
252 maxfragsperpacket = 16;
253 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
254 NULL, UMA_ALIGN_PTR, 0);
257 /* Start ipport_tick. */
258 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
260 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
261 SHUTDOWN_PRI_DEFAULT);
262 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
263 NULL, EVENTHANDLER_PRI_ANY);
265 /* Initialize various other remaining things. */
266 ip_id = time_second & 0xffff;
267 ipintrq.ifq_maxlen = ipqmaxlen;
268 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
269 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
276 callout_stop(&ipport_tick_callout);
280 * Ip input routine. Checksum and byte swap header. If fragmented
281 * try to reassemble. Process options. Pass to next level.
284 ip_input(struct mbuf *m)
286 struct ip *ip = NULL;
287 struct in_ifaddr *ia = NULL;
289 int checkif, hlen = 0;
291 int dchg = 0; /* dest changed after fw */
292 struct in_addr odst; /* original dst address */
296 if (m->m_flags & M_FASTFWD_OURS) {
298 * Firewall or NAT changed destination to local.
299 * We expect ip_len and ip_off to be in host byte order.
301 m->m_flags &= ~M_FASTFWD_OURS;
302 /* Set up some basics that will be used later. */
303 ip = mtod(m, struct ip *);
304 hlen = ip->ip_hl << 2;
310 if (m->m_pkthdr.len < sizeof(struct ip))
313 if (m->m_len < sizeof (struct ip) &&
314 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
315 ipstat.ips_toosmall++;
318 ip = mtod(m, struct ip *);
320 if (ip->ip_v != IPVERSION) {
321 ipstat.ips_badvers++;
325 hlen = ip->ip_hl << 2;
326 if (hlen < sizeof(struct ip)) { /* minimum header length */
327 ipstat.ips_badhlen++;
330 if (hlen > m->m_len) {
331 if ((m = m_pullup(m, hlen)) == NULL) {
332 ipstat.ips_badhlen++;
335 ip = mtod(m, struct ip *);
338 /* 127/8 must not appear on wire - RFC1122 */
339 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
340 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
341 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
342 ipstat.ips_badaddr++;
347 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
348 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
350 if (hlen == sizeof(struct ip)) {
351 sum = in_cksum_hdr(ip);
353 sum = in_cksum(m, hlen);
362 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
363 /* packet is dropped by traffic conditioner */
368 * Convert fields to host representation.
370 ip->ip_len = ntohs(ip->ip_len);
371 if (ip->ip_len < hlen) {
375 ip->ip_off = ntohs(ip->ip_off);
378 * Check that the amount of data in the buffers
379 * is as at least much as the IP header would have us expect.
380 * Trim mbufs if longer than we expect.
381 * Drop packet if shorter than we expect.
383 if (m->m_pkthdr.len < ip->ip_len) {
385 ipstat.ips_tooshort++;
388 if (m->m_pkthdr.len > ip->ip_len) {
389 if (m->m_len == m->m_pkthdr.len) {
390 m->m_len = ip->ip_len;
391 m->m_pkthdr.len = ip->ip_len;
393 m_adj(m, ip->ip_len - m->m_pkthdr.len);
397 * Bypass packet filtering for packets from a tunnel (gif).
399 if (ip_ipsec_filtertunnel(m))
404 * Run through list of hooks for input packets.
406 * NB: Beware of the destination address changing (e.g.
407 * by NAT rewriting). When this happens, tell
408 * ip_forward to do the right thing.
411 /* Jump over all PFIL processing if hooks are not active. */
412 if (!PFIL_HOOKED(&inet_pfil_hook))
416 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
419 if (m == NULL) /* consumed by filter */
422 ip = mtod(m, struct ip *);
423 dchg = (odst.s_addr != ip->ip_dst.s_addr);
425 #ifdef IPFIREWALL_FORWARD
426 if (m->m_flags & M_FASTFWD_OURS) {
427 m->m_flags &= ~M_FASTFWD_OURS;
430 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
432 * Directly ship on the packet. This allows to forward packets
433 * that were destined for us to some other directly connected
439 #endif /* IPFIREWALL_FORWARD */
443 * Process options and, if not destined for us,
444 * ship it on. ip_dooptions returns 1 when an
445 * error was detected (causing an icmp message
446 * to be sent and the original packet to be freed).
448 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
451 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
452 * matter if it is destined to another node, or whether it is
453 * a multicast one, RSVP wants it! and prevents it from being forwarded
454 * anywhere else. Also checks if the rsvp daemon is running before
455 * grabbing the packet.
457 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
461 * Check our list of addresses, to see if the packet is for us.
462 * If we don't have any addresses, assume any unicast packet
463 * we receive might be for us (and let the upper layers deal
466 if (TAILQ_EMPTY(&in_ifaddrhead) &&
467 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
471 * Enable a consistency check between the destination address
472 * and the arrival interface for a unicast packet (the RFC 1122
473 * strong ES model) if IP forwarding is disabled and the packet
474 * is not locally generated and the packet is not subject to
477 * XXX - Checking also should be disabled if the destination
478 * address is ipnat'ed to a different interface.
480 * XXX - Checking is incompatible with IP aliases added
481 * to the loopback interface instead of the interface where
482 * the packets are received.
484 * XXX - This is the case for carp vhost IPs as well so we
485 * insert a workaround. If the packet got here, we already
486 * checked with carp_iamatch() and carp_forus().
488 checkif = ip_checkinterface && (ipforwarding == 0) &&
489 m->m_pkthdr.rcvif != NULL &&
490 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
492 !m->m_pkthdr.rcvif->if_carp &&
497 * Check for exact addresses in the hash bucket.
499 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
501 * If the address matches, verify that the packet
502 * arrived via the correct interface if checking is
505 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
506 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
510 * Check for broadcast addresses.
512 * Only accept broadcast packets that arrive via the matching
513 * interface. Reception of forwarded directed broadcasts would
514 * be handled via ip_forward() and ether_output() with the loopback
515 * into the stack for SIMPLEX interfaces handled by ether_output().
517 if (m->m_pkthdr.rcvif != NULL &&
518 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
519 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
520 if (ifa->ifa_addr->sa_family != AF_INET)
523 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
526 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
529 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
534 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
535 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
536 ipstat.ips_cantforward++;
540 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
541 struct in_multi *inm;
544 * If we are acting as a multicast router, all
545 * incoming multicast packets are passed to the
546 * kernel-level multicast forwarding function.
547 * The packet is returned (relatively) intact; if
548 * ip_mforward() returns a non-zero value, the packet
549 * must be discarded, else it may be accepted below.
552 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
553 ipstat.ips_cantforward++;
559 * The process-level routing daemon needs to receive
560 * all multicast IGMP packets, whether or not this
561 * host belongs to their destination groups.
563 if (ip->ip_p == IPPROTO_IGMP)
565 ipstat.ips_forward++;
568 * See if we belong to the destination multicast group on the
572 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
575 ipstat.ips_notmember++;
581 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
583 if (ip->ip_dst.s_addr == INADDR_ANY)
587 * FAITH(Firewall Aided Internet Translator)
589 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
591 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
599 * Not for us; forward if possible and desirable.
601 if (ipforwarding == 0) {
602 ipstat.ips_cantforward++;
616 * IPSTEALTH: Process non-routing options only
617 * if the packet is destined for us.
619 if (ipstealth && hlen > sizeof (struct ip) &&
622 #endif /* IPSTEALTH */
624 /* Count the packet in the ip address stats */
626 ia->ia_ifa.if_ipackets++;
627 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
631 * Attempt reassembly; if it succeeds, proceed.
632 * ip_reass() will return a different mbuf.
634 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
638 ip = mtod(m, struct ip *);
639 /* Get the header length of the reassembled packet */
640 hlen = ip->ip_hl << 2;
644 * Further protocols expect the packet length to be w/o the
651 * enforce IPsec policy checking if we are seeing last header.
652 * note that we do not visit this with protocols with pcb layer
653 * code - like udp/tcp/raw ip.
655 if (ip_ipsec_input(m))
660 * Switch out to protocol's input routine.
662 ipstat.ips_delivered++;
664 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
671 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
672 * max has slightly different semantics than the sysctl, for historical
680 * -1 for unlimited allocation.
683 uma_zone_set_max(ipq_zone, 0);
685 * Positive number for specific bound.
688 uma_zone_set_max(ipq_zone, maxnipq);
690 * Zero specifies no further fragment queue allocation -- set the
691 * bound very low, but rely on implementation elsewhere to actually
692 * prevent allocation and reclaim current queues.
695 uma_zone_set_max(ipq_zone, 1);
699 ipq_zone_change(void *tag)
702 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
703 maxnipq = nmbclusters / 32;
709 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
714 error = sysctl_handle_int(oidp, &i, 0, req);
715 if (error || !req->newptr)
719 * XXXRW: Might be a good idea to sanity check the argument and place
720 * an extreme upper bound.
729 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
730 NULL, 0, sysctl_maxnipq, "I",
731 "Maximum number of IPv4 fragment reassembly queue entries");
734 * Take incoming datagram fragment and try to reassemble it into
735 * whole datagram. If the argument is the first fragment or one
736 * in between the function will return NULL and store the mbuf
737 * in the fragment chain. If the argument is the last fragment
738 * the packet will be reassembled and the pointer to the new
739 * mbuf returned for further processing. Only m_tags attached
740 * to the first packet/fragment are preserved.
741 * The IP header is *NOT* adjusted out of iplen.
744 ip_reass(struct mbuf *m)
747 struct mbuf *p, *q, *nq, *t;
748 struct ipq *fp = NULL;
749 struct ipqhead *head;
754 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
755 if (maxnipq == 0 || maxfragsperpacket == 0) {
756 ipstat.ips_fragments++;
757 ipstat.ips_fragdropped++;
762 ip = mtod(m, struct ip *);
763 hlen = ip->ip_hl << 2;
765 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
770 * Look for queue of fragments
773 TAILQ_FOREACH(fp, head, ipq_list)
774 if (ip->ip_id == fp->ipq_id &&
775 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
776 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
778 mac_fragment_match(m, fp) &&
780 ip->ip_p == fp->ipq_p)
786 * Attempt to trim the number of allocated fragment queues if it
787 * exceeds the administrative limit.
789 if ((nipq > maxnipq) && (maxnipq > 0)) {
791 * drop something from the tail of the current queue
792 * before proceeding further
794 struct ipq *q = TAILQ_LAST(head, ipqhead);
795 if (q == NULL) { /* gak */
796 for (i = 0; i < IPREASS_NHASH; i++) {
797 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
799 ipstat.ips_fragtimeout += r->ipq_nfrags;
800 ip_freef(&ipq[i], r);
805 ipstat.ips_fragtimeout += q->ipq_nfrags;
812 * Adjust ip_len to not reflect header,
813 * convert offset of this to bytes.
816 if (ip->ip_off & IP_MF) {
818 * Make sure that fragments have a data length
819 * that's a non-zero multiple of 8 bytes.
821 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
822 ipstat.ips_toosmall++; /* XXX */
825 m->m_flags |= M_FRAG;
827 m->m_flags &= ~M_FRAG;
832 * Attempt reassembly; if it succeeds, proceed.
833 * ip_reass() will return a different mbuf.
835 ipstat.ips_fragments++;
836 m->m_pkthdr.header = ip;
838 /* Previous ip_reass() started here. */
840 * Presence of header sizes in mbufs
841 * would confuse code below.
847 * If first fragment to arrive, create a reassembly queue.
850 fp = uma_zalloc(ipq_zone, M_NOWAIT);
854 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
855 uma_zfree(ipq_zone, fp);
859 mac_create_ipq(m, fp);
861 TAILQ_INSERT_HEAD(head, fp, ipq_list);
864 fp->ipq_ttl = IPFRAGTTL;
865 fp->ipq_p = ip->ip_p;
866 fp->ipq_id = ip->ip_id;
867 fp->ipq_src = ip->ip_src;
868 fp->ipq_dst = ip->ip_dst;
875 mac_update_ipq(m, fp);
879 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
882 * Handle ECN by comparing this segment with the first one;
883 * if CE is set, do not lose CE.
884 * drop if CE and not-ECT are mixed for the same packet.
886 ecn = ip->ip_tos & IPTOS_ECN_MASK;
887 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
888 if (ecn == IPTOS_ECN_CE) {
889 if (ecn0 == IPTOS_ECN_NOTECT)
891 if (ecn0 != IPTOS_ECN_CE)
892 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
894 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
898 * Find a segment which begins after this one does.
900 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
901 if (GETIP(q)->ip_off > ip->ip_off)
905 * If there is a preceding segment, it may provide some of
906 * our data already. If so, drop the data from the incoming
907 * segment. If it provides all of our data, drop us, otherwise
908 * stick new segment in the proper place.
910 * If some of the data is dropped from the the preceding
911 * segment, then it's checksum is invalidated.
914 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
919 m->m_pkthdr.csum_flags = 0;
923 m->m_nextpkt = p->m_nextpkt;
926 m->m_nextpkt = fp->ipq_frags;
931 * While we overlap succeeding segments trim them or,
932 * if they are completely covered, dequeue them.
934 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
936 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
937 if (i < GETIP(q)->ip_len) {
938 GETIP(q)->ip_len -= i;
939 GETIP(q)->ip_off += i;
941 q->m_pkthdr.csum_flags = 0;
946 ipstat.ips_fragdropped++;
952 * Check for complete reassembly and perform frag per packet
955 * Frag limiting is performed here so that the nth frag has
956 * a chance to complete the packet before we drop the packet.
957 * As a result, n+1 frags are actually allowed per packet, but
958 * only n will ever be stored. (n = maxfragsperpacket.)
962 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
963 if (GETIP(q)->ip_off != next) {
964 if (fp->ipq_nfrags > maxfragsperpacket) {
965 ipstat.ips_fragdropped += fp->ipq_nfrags;
970 next += GETIP(q)->ip_len;
972 /* Make sure the last packet didn't have the IP_MF flag */
973 if (p->m_flags & M_FRAG) {
974 if (fp->ipq_nfrags > maxfragsperpacket) {
975 ipstat.ips_fragdropped += fp->ipq_nfrags;
982 * Reassembly is complete. Make sure the packet is a sane size.
986 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
987 ipstat.ips_toolong++;
988 ipstat.ips_fragdropped += fp->ipq_nfrags;
994 * Concatenate fragments.
1001 q->m_nextpkt = NULL;
1002 for (q = nq; q != NULL; q = nq) {
1004 q->m_nextpkt = NULL;
1005 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1006 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1010 * In order to do checksumming faster we do 'end-around carry' here
1011 * (and not in for{} loop), though it implies we are not going to
1012 * reassemble more than 64k fragments.
1014 m->m_pkthdr.csum_data =
1015 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1017 mac_create_datagram_from_ipq(fp, m);
1018 mac_destroy_ipq(fp);
1022 * Create header for new ip packet by modifying header of first
1023 * packet; dequeue and discard fragment reassembly header.
1024 * Make header visible.
1026 ip->ip_len = (ip->ip_hl << 2) + next;
1027 ip->ip_src = fp->ipq_src;
1028 ip->ip_dst = fp->ipq_dst;
1029 TAILQ_REMOVE(head, fp, ipq_list);
1031 uma_zfree(ipq_zone, fp);
1032 m->m_len += (ip->ip_hl << 2);
1033 m->m_data -= (ip->ip_hl << 2);
1034 /* some debugging cruft by sklower, below, will go away soon */
1035 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1037 ipstat.ips_reassembled++;
1042 ipstat.ips_fragdropped++;
1054 * Free a fragment reassembly header and all
1055 * associated datagrams.
1058 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1064 while (fp->ipq_frags) {
1066 fp->ipq_frags = q->m_nextpkt;
1069 TAILQ_REMOVE(fhp, fp, ipq_list);
1070 uma_zfree(ipq_zone, fp);
1075 * IP timer processing;
1076 * if a timer expires on a reassembly
1077 * queue, discard it.
1086 for (i = 0; i < IPREASS_NHASH; i++) {
1087 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1091 fp = TAILQ_NEXT(fp, ipq_list);
1092 if(--fpp->ipq_ttl == 0) {
1093 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1094 ip_freef(&ipq[i], fpp);
1099 * If we are over the maximum number of fragments
1100 * (due to the limit being lowered), drain off
1101 * enough to get down to the new limit.
1103 if (maxnipq >= 0 && nipq > maxnipq) {
1104 for (i = 0; i < IPREASS_NHASH; i++) {
1105 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1106 ipstat.ips_fragdropped +=
1107 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1108 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1116 * Drain off all datagram fragments.
1124 for (i = 0; i < IPREASS_NHASH; i++) {
1125 while(!TAILQ_EMPTY(&ipq[i])) {
1126 ipstat.ips_fragdropped +=
1127 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1128 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1136 * The protocol to be inserted into ip_protox[] must be already registered
1137 * in inetsw[], either statically or through pf_proto_register().
1140 ipproto_register(u_char ipproto)
1144 /* Sanity checks. */
1146 return (EPROTONOSUPPORT);
1149 * The protocol slot must not be occupied by another protocol
1150 * already. An index pointing to IPPROTO_RAW is unused.
1152 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1154 return (EPFNOSUPPORT);
1155 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1158 /* Find the protocol position in inetsw[] and set the index. */
1159 for (pr = inetdomain.dom_protosw;
1160 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1161 if (pr->pr_domain->dom_family == PF_INET &&
1162 pr->pr_protocol && pr->pr_protocol == ipproto) {
1163 /* Be careful to only index valid IP protocols. */
1164 if (pr->pr_protocol < IPPROTO_MAX) {
1165 ip_protox[pr->pr_protocol] = pr - inetsw;
1171 return (EPROTONOSUPPORT);
1175 ipproto_unregister(u_char ipproto)
1179 /* Sanity checks. */
1181 return (EPROTONOSUPPORT);
1183 /* Check if the protocol was indeed registered. */
1184 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1186 return (EPFNOSUPPORT);
1187 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1190 /* Reset the protocol slot to IPPROTO_RAW. */
1191 ip_protox[ipproto] = pr - inetsw;
1196 * Given address of next destination (final or next hop),
1197 * return internet address info of interface to be used to get there.
1200 ip_rtaddr(struct in_addr dst, u_int fibnum)
1203 struct sockaddr_in *sin;
1204 struct in_ifaddr *ifa;
1206 bzero(&sro, sizeof(sro));
1207 sin = (struct sockaddr_in *)&sro.ro_dst;
1208 sin->sin_family = AF_INET;
1209 sin->sin_len = sizeof(*sin);
1210 sin->sin_addr = dst;
1211 in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
1213 if (sro.ro_rt == NULL)
1216 ifa = ifatoia(sro.ro_rt->rt_ifa);
1221 u_char inetctlerrmap[PRC_NCMDS] = {
1223 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1224 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1225 EMSGSIZE, EHOSTUNREACH, 0, 0,
1226 0, 0, EHOSTUNREACH, 0,
1227 ENOPROTOOPT, ECONNREFUSED
1231 * Forward a packet. If some error occurs return the sender
1232 * an icmp packet. Note we can't always generate a meaningful
1233 * icmp message because icmp doesn't have a large enough repertoire
1234 * of codes and types.
1236 * If not forwarding, just drop the packet. This could be confusing
1237 * if ipforwarding was zero but some routing protocol was advancing
1238 * us as a gateway to somewhere. However, we must let the routing
1239 * protocol deal with that.
1241 * The srcrt parameter indicates whether the packet is being forwarded
1242 * via a source route.
1245 ip_forward(struct mbuf *m, int srcrt)
1247 struct ip *ip = mtod(m, struct ip *);
1248 struct in_ifaddr *ia = NULL;
1250 struct in_addr dest;
1252 int error, type = 0, code = 0, mtu = 0;
1254 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1255 ipstat.ips_cantforward++;
1262 if (ip->ip_ttl <= IPTTLDEC) {
1263 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1271 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1272 if (!srcrt && ia == NULL) {
1273 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1278 * Save the IP header and at most 8 bytes of the payload,
1279 * in case we need to generate an ICMP message to the src.
1281 * XXX this can be optimized a lot by saving the data in a local
1282 * buffer on the stack (72 bytes at most), and only allocating the
1283 * mbuf if really necessary. The vast majority of the packets
1284 * are forwarded without having to send an ICMP back (either
1285 * because unnecessary, or because rate limited), so we are
1286 * really we are wasting a lot of work here.
1288 * We don't use m_copy() because it might return a reference
1289 * to a shared cluster. Both this function and ip_output()
1290 * assume exclusive access to the IP header in `m', so any
1291 * data in a cluster may change before we reach icmp_error().
1293 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1294 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1296 * It's probably ok if the pkthdr dup fails (because
1297 * the deep copy of the tag chain failed), but for now
1298 * be conservative and just discard the copy since
1299 * code below may some day want the tags.
1304 if (mcopy != NULL) {
1305 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1306 mcopy->m_pkthdr.len = mcopy->m_len;
1307 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1313 ip->ip_ttl -= IPTTLDEC;
1319 * If forwarding packet using same interface that it came in on,
1320 * perhaps should send a redirect to sender to shortcut a hop.
1321 * Only send redirect if source is sending directly to us,
1322 * and if packet was not source routed (or has any options).
1323 * Also, don't send redirect if forwarding using a default route
1324 * or a route modified by a redirect.
1327 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1328 struct sockaddr_in *sin;
1331 bzero(&ro, sizeof(ro));
1332 sin = (struct sockaddr_in *)&ro.ro_dst;
1333 sin->sin_family = AF_INET;
1334 sin->sin_len = sizeof(*sin);
1335 sin->sin_addr = ip->ip_dst;
1336 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
1340 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1341 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1342 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1343 u_long src = ntohl(ip->ip_src.s_addr);
1346 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1347 if (rt->rt_flags & RTF_GATEWAY)
1348 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1350 dest.s_addr = ip->ip_dst.s_addr;
1351 /* Router requirements says to only send host redirects */
1352 type = ICMP_REDIRECT;
1353 code = ICMP_REDIRECT_HOST;
1361 * Try to cache the route MTU from ip_output so we can consider it for
1362 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1364 bzero(&ro, sizeof(ro));
1366 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1368 if (error == EMSGSIZE && ro.ro_rt)
1369 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1374 ipstat.ips_cantforward++;
1376 ipstat.ips_forward++;
1378 ipstat.ips_redirectsent++;
1390 case 0: /* forwarded, but need redirect */
1391 /* type, code set above */
1394 case ENETUNREACH: /* shouldn't happen, checked above */
1399 type = ICMP_UNREACH;
1400 code = ICMP_UNREACH_HOST;
1404 type = ICMP_UNREACH;
1405 code = ICMP_UNREACH_NEEDFRAG;
1409 * If IPsec is configured for this path,
1410 * override any possibly mtu value set by ip_output.
1412 mtu = ip_ipsec_mtu(m, mtu);
1415 * If the MTU was set before make sure we are below the
1417 * If the MTU wasn't set before use the interface mtu or
1418 * fall back to the next smaller mtu step compared to the
1419 * current packet size.
1423 mtu = min(mtu, ia->ia_ifp->if_mtu);
1426 mtu = ia->ia_ifp->if_mtu;
1428 mtu = ip_next_mtu(ip->ip_len, 0);
1430 ipstat.ips_cantfrag++;
1435 * A router should not generate ICMP_SOURCEQUENCH as
1436 * required in RFC1812 Requirements for IP Version 4 Routers.
1437 * Source quench could be a big problem under DoS attacks,
1438 * or if the underlying interface is rate-limited.
1439 * Those who need source quench packets may re-enable them
1440 * via the net.inet.ip.sendsourcequench sysctl.
1442 if (ip_sendsourcequench == 0) {
1446 type = ICMP_SOURCEQUENCH;
1451 case EACCES: /* ipfw denied packet */
1455 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1459 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1462 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1466 if (inp->inp_socket->so_options & SO_BINTIME) {
1467 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1468 SCM_BINTIME, SOL_SOCKET);
1470 mp = &(*mp)->m_next;
1472 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1475 bintime2timeval(&bt, &tv);
1476 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1477 SCM_TIMESTAMP, SOL_SOCKET);
1479 mp = &(*mp)->m_next;
1482 if (inp->inp_flags & INP_RECVDSTADDR) {
1483 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1484 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1486 mp = &(*mp)->m_next;
1488 if (inp->inp_flags & INP_RECVTTL) {
1489 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1490 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1492 mp = &(*mp)->m_next;
1496 * Moving these out of udp_input() made them even more broken
1497 * than they already were.
1499 /* options were tossed already */
1500 if (inp->inp_flags & INP_RECVOPTS) {
1501 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1502 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1504 mp = &(*mp)->m_next;
1506 /* ip_srcroute doesn't do what we want here, need to fix */
1507 if (inp->inp_flags & INP_RECVRETOPTS) {
1508 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1509 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1511 mp = &(*mp)->m_next;
1514 if (inp->inp_flags & INP_RECVIF) {
1517 struct sockaddr_dl sdl;
1520 struct sockaddr_dl *sdp;
1521 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1523 if (((ifp = m->m_pkthdr.rcvif))
1524 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1525 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1527 * Change our mind and don't try copy.
1529 if ((sdp->sdl_family != AF_LINK)
1530 || (sdp->sdl_len > sizeof(sdlbuf))) {
1533 bcopy(sdp, sdl2, sdp->sdl_len);
1537 = offsetof(struct sockaddr_dl, sdl_data[0]);
1538 sdl2->sdl_family = AF_LINK;
1539 sdl2->sdl_index = 0;
1540 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1542 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1543 IP_RECVIF, IPPROTO_IP);
1545 mp = &(*mp)->m_next;
1550 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1551 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1552 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1555 static int ip_rsvp_on;
1556 struct socket *ip_rsvpd;
1558 ip_rsvp_init(struct socket *so)
1560 if (so->so_type != SOCK_RAW ||
1561 so->so_proto->pr_protocol != IPPROTO_RSVP)
1564 if (ip_rsvpd != NULL)
1569 * This may seem silly, but we need to be sure we don't over-increment
1570 * the RSVP counter, in case something slips up.
1585 * This may seem silly, but we need to be sure we don't over-decrement
1586 * the RSVP counter, in case something slips up.
1596 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1598 if (rsvp_input_p) { /* call the real one if loaded */
1599 rsvp_input_p(m, off);
1603 /* Can still get packets with rsvp_on = 0 if there is a local member
1604 * of the group to which the RSVP packet is addressed. But in this
1605 * case we want to throw the packet away.
1613 if (ip_rsvpd != NULL) {
1617 /* Drop the packet */