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>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
49 #include <sys/kernel.h>
50 #include <sys/syslog.h>
51 #include <sys/sysctl.h>
55 #include <net/if_types.h>
56 #include <net/if_var.h>
57 #include <net/if_dl.h>
58 #include <net/route.h>
59 #include <net/netisr.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_icmp.h>
68 #include <netinet/ip_options.h>
69 #include <machine/in_cksum.h>
71 #include <netinet/ip_carp.h>
73 #if defined(IPSEC) || defined(FAST_IPSEC)
74 #include <netinet/ip_ipsec.h>
77 #include <sys/socketvar.h>
79 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
80 #include <netinet/ip_fw.h>
81 #include <netinet/ip_dummynet.h>
83 #include <security/mac/mac_framework.h>
88 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
89 &ipforwarding, 0, "Enable IP forwarding between interfaces");
91 static int ipsendredirects = 1; /* XXX */
92 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
93 &ipsendredirects, 0, "Enable sending IP redirects");
95 int ip_defttl = IPDEFTTL;
96 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
97 &ip_defttl, 0, "Maximum TTL on IP packets");
99 static int ip_keepfaith = 0;
100 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
102 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
104 static int ip_sendsourcequench = 0;
105 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
106 &ip_sendsourcequench, 0,
107 "Enable the transmission of source quench packets");
109 int ip_do_randomid = 0;
110 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
112 "Assign random ip_id values");
115 * XXX - Setting ip_checkinterface mostly implements the receive side of
116 * the Strong ES model described in RFC 1122, but since the routing table
117 * and transmit implementation do not implement the Strong ES model,
118 * setting this to 1 results in an odd hybrid.
120 * XXX - ip_checkinterface currently must be disabled if you use ipnat
121 * to translate the destination address to another local interface.
123 * XXX - ip_checkinterface must be disabled if you add IP aliases
124 * to the loopback interface instead of the interface where the
125 * packets for those addresses are received.
127 static int ip_checkinterface = 0;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
129 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
131 struct pfil_head inet_pfil_hook; /* Packet filter hooks */
133 static struct ifqueue ipintrq;
134 static int ipqmaxlen = IFQ_MAXLEN;
136 extern struct domain inetdomain;
137 extern struct protosw inetsw[];
138 u_char ip_protox[IPPROTO_MAX];
139 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
140 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
141 u_long in_ifaddrhmask; /* mask for hash table */
143 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
144 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
145 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
146 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
148 struct ipstat ipstat;
149 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
150 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
153 * IP datagram reassembly.
155 #define IPREASS_NHASH_LOG2 6
156 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
157 #define IPREASS_HMASK (IPREASS_NHASH - 1)
158 #define IPREASS_HASH(x,y) \
159 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
161 static uma_zone_t ipq_zone;
162 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
163 static struct mtx ipqlock;
165 #define IPQ_LOCK() mtx_lock(&ipqlock)
166 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
167 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
168 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
170 static void maxnipq_update(void);
171 static void ipq_zone_change(void *);
173 static int maxnipq; /* Administrative limit on # reass queues. */
174 static int nipq = 0; /* Total # of reass queues */
175 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD, &nipq, 0,
176 "Current number of IPv4 fragment reassembly queue entries");
178 static int maxfragsperpacket;
179 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
180 &maxfragsperpacket, 0,
181 "Maximum number of IPv4 fragments allowed per packet");
183 struct callout ipport_tick_callout;
186 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
187 &ip_mtu, 0, "Default MTU");
192 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
197 * ipfw_ether and ipfw_bridge hooks.
198 * XXX: Temporary until those are converted to pfil_hooks as well.
200 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
201 ip_dn_io_t *ip_dn_io_ptr = NULL;
204 static void ip_freef(struct ipqhead *, struct ipq *);
207 * IP initialization: fill in IP protocol switch table.
208 * All protocols not implemented in kernel go to raw IP protocol handler.
213 register struct protosw *pr;
216 TAILQ_INIT(&in_ifaddrhead);
217 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
218 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
220 panic("ip_init: PF_INET not found");
222 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
223 for (i = 0; i < IPPROTO_MAX; i++)
224 ip_protox[i] = pr - inetsw;
226 * Cycle through IP protocols and put them into the appropriate place
229 for (pr = inetdomain.dom_protosw;
230 pr < inetdomain.dom_protoswNPROTOSW; pr++)
231 if (pr->pr_domain->dom_family == PF_INET &&
232 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
233 /* Be careful to only index valid IP protocols. */
234 if (pr->pr_protocol < IPPROTO_MAX)
235 ip_protox[pr->pr_protocol] = pr - inetsw;
238 /* Initialize packet filter hooks. */
239 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
240 inet_pfil_hook.ph_af = AF_INET;
241 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
242 printf("%s: WARNING: unable to register pfil hook, "
243 "error %d\n", __func__, i);
245 /* Initialize IP reassembly queue. */
247 for (i = 0; i < IPREASS_NHASH; i++)
249 maxnipq = nmbclusters / 32;
250 maxfragsperpacket = 16;
251 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
252 NULL, UMA_ALIGN_PTR, 0);
255 /* Start ipport_tick. */
256 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
258 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
259 SHUTDOWN_PRI_DEFAULT);
260 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
261 NULL, EVENTHANDLER_PRI_ANY);
263 /* Initialize various other remaining things. */
264 ip_id = time_second & 0xffff;
265 ipintrq.ifq_maxlen = ipqmaxlen;
266 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
267 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
273 callout_stop(&ipport_tick_callout);
277 * Ip input routine. Checksum and byte swap header. If fragmented
278 * try to reassemble. Process options. Pass to next level.
281 ip_input(struct mbuf *m)
283 struct ip *ip = NULL;
284 struct in_ifaddr *ia = NULL;
286 int checkif, hlen = 0;
288 int dchg = 0; /* dest changed after fw */
289 struct in_addr odst; /* original dst address */
293 if (m->m_flags & M_FASTFWD_OURS) {
295 * Firewall or NAT changed destination to local.
296 * We expect ip_len and ip_off to be in host byte order.
298 m->m_flags &= ~M_FASTFWD_OURS;
299 /* Set up some basics that will be used later. */
300 ip = mtod(m, struct ip *);
301 hlen = ip->ip_hl << 2;
307 if (m->m_pkthdr.len < sizeof(struct ip))
310 if (m->m_len < sizeof (struct ip) &&
311 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
312 ipstat.ips_toosmall++;
315 ip = mtod(m, struct ip *);
317 if (ip->ip_v != IPVERSION) {
318 ipstat.ips_badvers++;
322 hlen = ip->ip_hl << 2;
323 if (hlen < sizeof(struct ip)) { /* minimum header length */
324 ipstat.ips_badhlen++;
327 if (hlen > m->m_len) {
328 if ((m = m_pullup(m, hlen)) == NULL) {
329 ipstat.ips_badhlen++;
332 ip = mtod(m, struct ip *);
335 /* 127/8 must not appear on wire - RFC1122 */
336 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
337 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
338 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
339 ipstat.ips_badaddr++;
344 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
345 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
347 if (hlen == sizeof(struct ip)) {
348 sum = in_cksum_hdr(ip);
350 sum = in_cksum(m, hlen);
359 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
360 /* packet is dropped by traffic conditioner */
365 * Convert fields to host representation.
367 ip->ip_len = ntohs(ip->ip_len);
368 if (ip->ip_len < hlen) {
372 ip->ip_off = ntohs(ip->ip_off);
375 * Check that the amount of data in the buffers
376 * is as at least much as the IP header would have us expect.
377 * Trim mbufs if longer than we expect.
378 * Drop packet if shorter than we expect.
380 if (m->m_pkthdr.len < ip->ip_len) {
382 ipstat.ips_tooshort++;
385 if (m->m_pkthdr.len > ip->ip_len) {
386 if (m->m_len == m->m_pkthdr.len) {
387 m->m_len = ip->ip_len;
388 m->m_pkthdr.len = ip->ip_len;
390 m_adj(m, ip->ip_len - m->m_pkthdr.len);
392 #if defined(IPSEC) || defined(FAST_IPSEC)
394 * Bypass packet filtering for packets from a tunnel (gif).
396 if (ip_ipsec_filtergif(m))
401 * Run through list of hooks for input packets.
403 * NB: Beware of the destination address changing (e.g.
404 * by NAT rewriting). When this happens, tell
405 * ip_forward to do the right thing.
408 /* Jump over all PFIL processing if hooks are not active. */
409 if (!PFIL_HOOKED(&inet_pfil_hook))
413 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
416 if (m == NULL) /* consumed by filter */
419 ip = mtod(m, struct ip *);
420 dchg = (odst.s_addr != ip->ip_dst.s_addr);
422 #ifdef IPFIREWALL_FORWARD
423 if (m->m_flags & M_FASTFWD_OURS) {
424 m->m_flags &= ~M_FASTFWD_OURS;
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 */
440 * Process options and, if not destined for us,
441 * ship it on. ip_dooptions returns 1 when an
442 * error was detected (causing an icmp message
443 * to be sent and the original packet to be freed).
445 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
448 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
449 * matter if it is destined to another node, or whether it is
450 * a multicast one, RSVP wants it! and prevents it from being forwarded
451 * anywhere else. Also checks if the rsvp daemon is running before
452 * grabbing the packet.
454 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
458 * Check our list of addresses, to see if the packet is for us.
459 * If we don't have any addresses, assume any unicast packet
460 * we receive might be for us (and let the upper layers deal
463 if (TAILQ_EMPTY(&in_ifaddrhead) &&
464 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
468 * Enable a consistency check between the destination address
469 * and the arrival interface for a unicast packet (the RFC 1122
470 * strong ES model) if IP forwarding is disabled and the packet
471 * is not locally generated and the packet is not subject to
474 * XXX - Checking also should be disabled if the destination
475 * address is ipnat'ed to a different interface.
477 * XXX - Checking is incompatible with IP aliases added
478 * to the loopback interface instead of the interface where
479 * the packets are received.
481 * XXX - This is the case for carp vhost IPs as well so we
482 * insert a workaround. If the packet got here, we already
483 * checked with carp_iamatch() and carp_forus().
485 checkif = ip_checkinterface && (ipforwarding == 0) &&
486 m->m_pkthdr.rcvif != NULL &&
487 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
489 !m->m_pkthdr.rcvif->if_carp &&
494 * Check for exact addresses in the hash bucket.
496 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
498 * If the address matches, verify that the packet
499 * arrived via the correct interface if checking is
502 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
503 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
507 * Check for broadcast addresses.
509 * Only accept broadcast packets that arrive via the matching
510 * interface. Reception of forwarded directed broadcasts would
511 * be handled via ip_forward() and ether_output() with the loopback
512 * into the stack for SIMPLEX interfaces handled by ether_output().
514 if (m->m_pkthdr.rcvif != NULL &&
515 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
516 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
517 if (ifa->ifa_addr->sa_family != AF_INET)
520 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
523 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
526 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
531 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
532 struct in_multi *inm;
535 * If we are acting as a multicast router, all
536 * incoming multicast packets are passed to the
537 * kernel-level multicast forwarding function.
538 * The packet is returned (relatively) intact; if
539 * ip_mforward() returns a non-zero value, the packet
540 * must be discarded, else it may be accepted below.
543 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
544 ipstat.ips_cantforward++;
550 * The process-level routing daemon needs to receive
551 * all multicast IGMP packets, whether or not this
552 * host belongs to their destination groups.
554 if (ip->ip_p == IPPROTO_IGMP)
556 ipstat.ips_forward++;
559 * See if we belong to the destination multicast group on the
563 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
566 ipstat.ips_notmember++;
572 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
574 if (ip->ip_dst.s_addr == INADDR_ANY)
578 * FAITH(Firewall Aided Internet Translator)
580 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
582 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
590 * Not for us; forward if possible and desirable.
592 if (ipforwarding == 0) {
593 ipstat.ips_cantforward++;
596 #if defined(IPSEC) || defined(FAST_IPSEC)
607 * IPSTEALTH: Process non-routing options only
608 * if the packet is destined for us.
610 if (ipstealth && hlen > sizeof (struct ip) &&
613 #endif /* IPSTEALTH */
615 /* Count the packet in the ip address stats */
617 ia->ia_ifa.if_ipackets++;
618 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
622 * Attempt reassembly; if it succeeds, proceed.
623 * ip_reass() will return a different mbuf.
625 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
629 ip = mtod(m, struct ip *);
630 /* Get the header length of the reassembled packet */
631 hlen = ip->ip_hl << 2;
635 * Further protocols expect the packet length to be w/o the
640 #if defined(IPSEC) || defined(FAST_IPSEC)
642 * enforce IPsec policy checking if we are seeing last header.
643 * note that we do not visit this with protocols with pcb layer
644 * code - like udp/tcp/raw ip.
646 if (ip_ipsec_input(m))
651 * Switch out to protocol's input routine.
653 ipstat.ips_delivered++;
655 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
662 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
663 * max has slightly different semantics than the sysctl, for historical
671 * -1 for unlimited allocation.
674 uma_zone_set_max(ipq_zone, 0);
676 * Positive number for specific bound.
679 uma_zone_set_max(ipq_zone, maxnipq);
681 * Zero specifies no further fragment queue allocation -- set the
682 * bound very low, but rely on implementation elsewhere to actually
683 * prevent allocation and reclaim current queues.
686 uma_zone_set_max(ipq_zone, 1);
690 ipq_zone_change(void *tag)
693 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
694 maxnipq = nmbclusters / 32;
700 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
705 error = sysctl_handle_int(oidp, &i, 0, req);
706 if (error || !req->newptr)
710 * XXXRW: Might be a good idea to sanity check the argument and place
711 * an extreme upper bound.
720 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
721 NULL, 0, sysctl_maxnipq, "I",
722 "Maximum number of IPv4 fragment reassembly queue entries");
725 * Take incoming datagram fragment and try to reassemble it into
726 * whole datagram. If the argument is the first fragment or one
727 * in between the function will return NULL and store the mbuf
728 * in the fragment chain. If the argument is the last fragment
729 * the packet will be reassembled and the pointer to the new
730 * mbuf returned for further processing. Only m_tags attached
731 * to the first packet/fragment are preserved.
732 * The IP header is *NOT* adjusted out of iplen.
736 ip_reass(struct mbuf *m)
739 struct mbuf *p, *q, *nq, *t;
740 struct ipq *fp = NULL;
741 struct ipqhead *head;
746 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
747 if (maxnipq == 0 || maxfragsperpacket == 0) {
748 ipstat.ips_fragments++;
749 ipstat.ips_fragdropped++;
754 ip = mtod(m, struct ip *);
755 hlen = ip->ip_hl << 2;
757 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
762 * Look for queue of fragments
765 TAILQ_FOREACH(fp, head, ipq_list)
766 if (ip->ip_id == fp->ipq_id &&
767 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
768 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
770 mac_fragment_match(m, fp) &&
772 ip->ip_p == fp->ipq_p)
778 * Attempt to trim the number of allocated fragment queues if it
779 * exceeds the administrative limit.
781 if ((nipq > maxnipq) && (maxnipq > 0)) {
783 * drop something from the tail of the current queue
784 * before proceeding further
786 struct ipq *q = TAILQ_LAST(head, ipqhead);
787 if (q == NULL) { /* gak */
788 for (i = 0; i < IPREASS_NHASH; i++) {
789 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
791 ipstat.ips_fragtimeout += r->ipq_nfrags;
792 ip_freef(&ipq[i], r);
797 ipstat.ips_fragtimeout += q->ipq_nfrags;
804 * Adjust ip_len to not reflect header,
805 * convert offset of this to bytes.
808 if (ip->ip_off & IP_MF) {
810 * Make sure that fragments have a data length
811 * that's a non-zero multiple of 8 bytes.
813 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
814 ipstat.ips_toosmall++; /* XXX */
817 m->m_flags |= M_FRAG;
819 m->m_flags &= ~M_FRAG;
824 * Attempt reassembly; if it succeeds, proceed.
825 * ip_reass() will return a different mbuf.
827 ipstat.ips_fragments++;
828 m->m_pkthdr.header = ip;
830 /* Previous ip_reass() started here. */
832 * Presence of header sizes in mbufs
833 * would confuse code below.
839 * If first fragment to arrive, create a reassembly queue.
842 fp = uma_zalloc(ipq_zone, M_NOWAIT);
846 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
847 uma_zfree(ipq_zone, fp);
851 mac_create_ipq(m, fp);
853 TAILQ_INSERT_HEAD(head, fp, ipq_list);
856 fp->ipq_ttl = IPFRAGTTL;
857 fp->ipq_p = ip->ip_p;
858 fp->ipq_id = ip->ip_id;
859 fp->ipq_src = ip->ip_src;
860 fp->ipq_dst = ip->ip_dst;
867 mac_update_ipq(m, fp);
871 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
874 * Handle ECN by comparing this segment with the first one;
875 * if CE is set, do not lose CE.
876 * drop if CE and not-ECT are mixed for the same packet.
878 ecn = ip->ip_tos & IPTOS_ECN_MASK;
879 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
880 if (ecn == IPTOS_ECN_CE) {
881 if (ecn0 == IPTOS_ECN_NOTECT)
883 if (ecn0 != IPTOS_ECN_CE)
884 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
886 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
890 * Find a segment which begins after this one does.
892 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
893 if (GETIP(q)->ip_off > ip->ip_off)
897 * If there is a preceding segment, it may provide some of
898 * our data already. If so, drop the data from the incoming
899 * segment. If it provides all of our data, drop us, otherwise
900 * stick new segment in the proper place.
902 * If some of the data is dropped from the the preceding
903 * segment, then it's checksum is invalidated.
906 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
911 m->m_pkthdr.csum_flags = 0;
915 m->m_nextpkt = p->m_nextpkt;
918 m->m_nextpkt = fp->ipq_frags;
923 * While we overlap succeeding segments trim them or,
924 * if they are completely covered, dequeue them.
926 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
928 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
929 if (i < GETIP(q)->ip_len) {
930 GETIP(q)->ip_len -= i;
931 GETIP(q)->ip_off += i;
933 q->m_pkthdr.csum_flags = 0;
938 ipstat.ips_fragdropped++;
944 * Check for complete reassembly and perform frag per packet
947 * Frag limiting is performed here so that the nth frag has
948 * a chance to complete the packet before we drop the packet.
949 * As a result, n+1 frags are actually allowed per packet, but
950 * only n will ever be stored. (n = maxfragsperpacket.)
954 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
955 if (GETIP(q)->ip_off != next) {
956 if (fp->ipq_nfrags > maxfragsperpacket) {
957 ipstat.ips_fragdropped += fp->ipq_nfrags;
962 next += GETIP(q)->ip_len;
964 /* Make sure the last packet didn't have the IP_MF flag */
965 if (p->m_flags & M_FRAG) {
966 if (fp->ipq_nfrags > maxfragsperpacket) {
967 ipstat.ips_fragdropped += fp->ipq_nfrags;
974 * Reassembly is complete. Make sure the packet is a sane size.
978 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
979 ipstat.ips_toolong++;
980 ipstat.ips_fragdropped += fp->ipq_nfrags;
986 * Concatenate fragments.
994 for (q = nq; q != NULL; q = nq) {
997 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
998 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1002 * In order to do checksumming faster we do 'end-around carry' here
1003 * (and not in for{} loop), though it implies we are not going to
1004 * reassemble more than 64k fragments.
1006 m->m_pkthdr.csum_data =
1007 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1009 mac_create_datagram_from_ipq(fp, m);
1010 mac_destroy_ipq(fp);
1014 * Create header for new ip packet by modifying header of first
1015 * packet; dequeue and discard fragment reassembly header.
1016 * Make header visible.
1018 ip->ip_len = (ip->ip_hl << 2) + next;
1019 ip->ip_src = fp->ipq_src;
1020 ip->ip_dst = fp->ipq_dst;
1021 TAILQ_REMOVE(head, fp, ipq_list);
1023 uma_zfree(ipq_zone, fp);
1024 m->m_len += (ip->ip_hl << 2);
1025 m->m_data -= (ip->ip_hl << 2);
1026 /* some debugging cruft by sklower, below, will go away soon */
1027 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1029 ipstat.ips_reassembled++;
1034 ipstat.ips_fragdropped++;
1046 * Free a fragment reassembly header and all
1047 * associated datagrams.
1051 struct ipqhead *fhp;
1054 register struct mbuf *q;
1058 while (fp->ipq_frags) {
1060 fp->ipq_frags = q->m_nextpkt;
1063 TAILQ_REMOVE(fhp, fp, ipq_list);
1064 uma_zfree(ipq_zone, fp);
1069 * IP timer processing;
1070 * if a timer expires on a reassembly
1071 * queue, discard it.
1076 register struct ipq *fp;
1080 for (i = 0; i < IPREASS_NHASH; i++) {
1081 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1085 fp = TAILQ_NEXT(fp, ipq_list);
1086 if(--fpp->ipq_ttl == 0) {
1087 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1088 ip_freef(&ipq[i], fpp);
1093 * If we are over the maximum number of fragments
1094 * (due to the limit being lowered), drain off
1095 * enough to get down to the new limit.
1097 if (maxnipq >= 0 && nipq > maxnipq) {
1098 for (i = 0; i < IPREASS_NHASH; i++) {
1099 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1100 ipstat.ips_fragdropped +=
1101 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1102 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1110 * Drain off all datagram fragments.
1118 for (i = 0; i < IPREASS_NHASH; i++) {
1119 while(!TAILQ_EMPTY(&ipq[i])) {
1120 ipstat.ips_fragdropped +=
1121 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1122 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1130 * The protocol to be inserted into ip_protox[] must be already registered
1131 * in inetsw[], either statically or through pf_proto_register().
1134 ipproto_register(u_char ipproto)
1138 /* Sanity checks. */
1140 return (EPROTONOSUPPORT);
1143 * The protocol slot must not be occupied by another protocol
1144 * already. An index pointing to IPPROTO_RAW is unused.
1146 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1148 return (EPFNOSUPPORT);
1149 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1152 /* Find the protocol position in inetsw[] and set the index. */
1153 for (pr = inetdomain.dom_protosw;
1154 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1155 if (pr->pr_domain->dom_family == PF_INET &&
1156 pr->pr_protocol && pr->pr_protocol == ipproto) {
1157 /* Be careful to only index valid IP protocols. */
1158 if (pr->pr_protocol < IPPROTO_MAX) {
1159 ip_protox[pr->pr_protocol] = pr - inetsw;
1165 return (EPROTONOSUPPORT);
1169 ipproto_unregister(u_char ipproto)
1173 /* Sanity checks. */
1175 return (EPROTONOSUPPORT);
1177 /* Check if the protocol was indeed registered. */
1178 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1180 return (EPFNOSUPPORT);
1181 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1184 /* Reset the protocol slot to IPPROTO_RAW. */
1185 ip_protox[ipproto] = pr - inetsw;
1190 * Given address of next destination (final or next hop),
1191 * return internet address info of interface to be used to get there.
1198 struct sockaddr_in *sin;
1199 struct in_ifaddr *ifa;
1201 bzero(&sro, sizeof(sro));
1202 sin = (struct sockaddr_in *)&sro.ro_dst;
1203 sin->sin_family = AF_INET;
1204 sin->sin_len = sizeof(*sin);
1205 sin->sin_addr = dst;
1206 rtalloc_ign(&sro, RTF_CLONING);
1208 if (sro.ro_rt == NULL)
1211 ifa = ifatoia(sro.ro_rt->rt_ifa);
1216 u_char inetctlerrmap[PRC_NCMDS] = {
1218 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1219 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1220 EMSGSIZE, EHOSTUNREACH, 0, 0,
1221 0, 0, EHOSTUNREACH, 0,
1222 ENOPROTOOPT, ECONNREFUSED
1226 * Forward a packet. If some error occurs return the sender
1227 * an icmp packet. Note we can't always generate a meaningful
1228 * icmp message because icmp doesn't have a large enough repertoire
1229 * of codes and types.
1231 * If not forwarding, just drop the packet. This could be confusing
1232 * if ipforwarding was zero but some routing protocol was advancing
1233 * us as a gateway to somewhere. However, we must let the routing
1234 * protocol deal with that.
1236 * The srcrt parameter indicates whether the packet is being forwarded
1237 * via a source route.
1240 ip_forward(struct mbuf *m, int srcrt)
1242 struct ip *ip = mtod(m, struct ip *);
1243 struct in_ifaddr *ia = NULL;
1245 struct in_addr dest;
1246 int error, type = 0, code = 0, mtu = 0;
1248 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1249 ipstat.ips_cantforward++;
1256 if (ip->ip_ttl <= IPTTLDEC) {
1257 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1265 if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1266 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1271 * Save the IP header and at most 8 bytes of the payload,
1272 * in case we need to generate an ICMP message to the src.
1274 * XXX this can be optimized a lot by saving the data in a local
1275 * buffer on the stack (72 bytes at most), and only allocating the
1276 * mbuf if really necessary. The vast majority of the packets
1277 * are forwarded without having to send an ICMP back (either
1278 * because unnecessary, or because rate limited), so we are
1279 * really we are wasting a lot of work here.
1281 * We don't use m_copy() because it might return a reference
1282 * to a shared cluster. Both this function and ip_output()
1283 * assume exclusive access to the IP header in `m', so any
1284 * data in a cluster may change before we reach icmp_error().
1286 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1287 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1289 * It's probably ok if the pkthdr dup fails (because
1290 * the deep copy of the tag chain failed), but for now
1291 * be conservative and just discard the copy since
1292 * code below may some day want the tags.
1297 if (mcopy != NULL) {
1298 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1299 mcopy->m_pkthdr.len = mcopy->m_len;
1300 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1306 ip->ip_ttl -= IPTTLDEC;
1312 * If forwarding packet using same interface that it came in on,
1313 * perhaps should send a redirect to sender to shortcut a hop.
1314 * Only send redirect if source is sending directly to us,
1315 * and if packet was not source routed (or has any options).
1316 * Also, don't send redirect if forwarding using a default route
1317 * or a route modified by a redirect.
1320 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1321 struct sockaddr_in *sin;
1325 bzero(&ro, sizeof(ro));
1326 sin = (struct sockaddr_in *)&ro.ro_dst;
1327 sin->sin_family = AF_INET;
1328 sin->sin_len = sizeof(*sin);
1329 sin->sin_addr = ip->ip_dst;
1330 rtalloc_ign(&ro, RTF_CLONING);
1334 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1335 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1336 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1337 u_long src = ntohl(ip->ip_src.s_addr);
1340 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1341 if (rt->rt_flags & RTF_GATEWAY)
1342 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1344 dest.s_addr = ip->ip_dst.s_addr;
1345 /* Router requirements says to only send host redirects */
1346 type = ICMP_REDIRECT;
1347 code = ICMP_REDIRECT_HOST;
1354 error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1356 ipstat.ips_cantforward++;
1358 ipstat.ips_forward++;
1360 ipstat.ips_redirectsent++;
1372 case 0: /* forwarded, but need redirect */
1373 /* type, code set above */
1376 case ENETUNREACH: /* shouldn't happen, checked above */
1381 type = ICMP_UNREACH;
1382 code = ICMP_UNREACH_HOST;
1386 type = ICMP_UNREACH;
1387 code = ICMP_UNREACH_NEEDFRAG;
1389 #if defined(IPSEC) || defined(FAST_IPSEC)
1390 mtu = ip_ipsec_mtu(m);
1393 * If the MTU wasn't set before use the interface mtu or
1394 * fall back to the next smaller mtu step compared to the
1395 * current packet size.
1399 mtu = ia->ia_ifp->if_mtu;
1401 mtu = ip_next_mtu(ip->ip_len, 0);
1403 ipstat.ips_cantfrag++;
1408 * A router should not generate ICMP_SOURCEQUENCH as
1409 * required in RFC1812 Requirements for IP Version 4 Routers.
1410 * Source quench could be a big problem under DoS attacks,
1411 * or if the underlying interface is rate-limited.
1412 * Those who need source quench packets may re-enable them
1413 * via the net.inet.ip.sendsourcequench sysctl.
1415 if (ip_sendsourcequench == 0) {
1419 type = ICMP_SOURCEQUENCH;
1424 case EACCES: /* ipfw denied packet */
1428 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1432 ip_savecontrol(inp, mp, ip, m)
1433 register struct inpcb *inp;
1434 register struct mbuf **mp;
1435 register struct ip *ip;
1436 register struct mbuf *m;
1438 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1442 if (inp->inp_socket->so_options & SO_BINTIME) {
1443 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1444 SCM_BINTIME, SOL_SOCKET);
1446 mp = &(*mp)->m_next;
1448 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1451 bintime2timeval(&bt, &tv);
1452 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1453 SCM_TIMESTAMP, SOL_SOCKET);
1455 mp = &(*mp)->m_next;
1458 if (inp->inp_flags & INP_RECVDSTADDR) {
1459 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1460 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1462 mp = &(*mp)->m_next;
1464 if (inp->inp_flags & INP_RECVTTL) {
1465 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1466 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1468 mp = &(*mp)->m_next;
1472 * Moving these out of udp_input() made them even more broken
1473 * than they already were.
1475 /* options were tossed already */
1476 if (inp->inp_flags & INP_RECVOPTS) {
1477 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1478 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1480 mp = &(*mp)->m_next;
1482 /* ip_srcroute doesn't do what we want here, need to fix */
1483 if (inp->inp_flags & INP_RECVRETOPTS) {
1484 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1485 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1487 mp = &(*mp)->m_next;
1490 if (inp->inp_flags & INP_RECVIF) {
1493 struct sockaddr_dl sdl;
1496 struct sockaddr_dl *sdp;
1497 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1499 if (((ifp = m->m_pkthdr.rcvif))
1500 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1501 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1503 * Change our mind and don't try copy.
1505 if ((sdp->sdl_family != AF_LINK)
1506 || (sdp->sdl_len > sizeof(sdlbuf))) {
1509 bcopy(sdp, sdl2, sdp->sdl_len);
1513 = offsetof(struct sockaddr_dl, sdl_data[0]);
1514 sdl2->sdl_family = AF_LINK;
1515 sdl2->sdl_index = 0;
1516 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1518 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1519 IP_RECVIF, IPPROTO_IP);
1521 mp = &(*mp)->m_next;
1526 * XXX these routines are called from the upper part of the kernel.
1527 * They need to be locked when we remove Giant.
1529 * They could also be moved to ip_mroute.c, since all the RSVP
1530 * handling is done there already.
1532 static int ip_rsvp_on;
1533 struct socket *ip_rsvpd;
1535 ip_rsvp_init(struct socket *so)
1537 if (so->so_type != SOCK_RAW ||
1538 so->so_proto->pr_protocol != IPPROTO_RSVP)
1541 if (ip_rsvpd != NULL)
1546 * This may seem silly, but we need to be sure we don't over-increment
1547 * the RSVP counter, in case something slips up.
1562 * This may seem silly, but we need to be sure we don't over-decrement
1563 * the RSVP counter, in case something slips up.
1573 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1575 if (rsvp_input_p) { /* call the real one if loaded */
1576 rsvp_input_p(m, off);
1580 /* Can still get packets with rsvp_on = 0 if there is a local member
1581 * of the group to which the RSVP packet is addressed. But in this
1582 * case we want to throw the packet away.
1590 if (ip_rsvpd != NULL) {
1594 /* Drop the packet */