1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/netinet/ip_input.c 180215 2008-07-03 12:44:36Z bz $");
34
35#include "opt_bootp.h"
36#include "opt_ipfw.h"
37#include "opt_ipstealth.h"
38#include "opt_ipsec.h"
39#include "opt_mac.h"
40#include "opt_carp.h"
41
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/callout.h>
45#include <sys/mbuf.h>
46#include <sys/malloc.h>
47#include <sys/domain.h>
48#include <sys/protosw.h>
49#include <sys/socket.h>
50#include <sys/time.h>
51#include <sys/kernel.h>
52#include <sys/syslog.h>
53#include <sys/sysctl.h>
54
55#include <net/pfil.h>
56#include <net/if.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>
62
63#include <netinet/in.h>
64#include <netinet/in_systm.h>
65#include <netinet/in_var.h>
66#include <netinet/ip.h>
67#include <netinet/in_pcb.h>
68#include <netinet/ip_var.h>
69#include <netinet/ip_icmp.h>
70#include <netinet/ip_options.h>
71#include <machine/in_cksum.h>
72#ifdef DEV_CARP
73#include <netinet/ip_carp.h>
74#endif
75#ifdef IPSEC
76#include <netinet/ip_ipsec.h>
77#endif /* IPSEC */
78
79#include <sys/socketvar.h>
80
81/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
82#include <netinet/ip_fw.h>
83#include <netinet/ip_dummynet.h>
84
85#include <security/mac/mac_framework.h>
86
87int rsvp_on = 0;
88
89int ipforwarding = 0;
90SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
91 &ipforwarding, 0, "Enable IP forwarding between interfaces");
92
93static int ipsendredirects = 1; /* XXX */
94SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
95 &ipsendredirects, 0, "Enable sending IP redirects");
96
97int ip_defttl = IPDEFTTL;
98SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
99 &ip_defttl, 0, "Maximum TTL on IP packets");
100
101static int ip_keepfaith = 0;
102SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
103 &ip_keepfaith, 0,
104 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
105
106static int ip_sendsourcequench = 0;
107SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
108 &ip_sendsourcequench, 0,
109 "Enable the transmission of source quench packets");
110
111int ip_do_randomid = 0;
112SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
113 &ip_do_randomid, 0,
114 "Assign random ip_id values");
115
116/*
117 * XXX - Setting ip_checkinterface mostly implements the receive side of
118 * the Strong ES model described in RFC 1122, but since the routing table
119 * and transmit implementation do not implement the Strong ES model,
120 * setting this to 1 results in an odd hybrid.
121 *
122 * XXX - ip_checkinterface currently must be disabled if you use ipnat
123 * to translate the destination address to another local interface.
124 *
125 * XXX - ip_checkinterface must be disabled if you add IP aliases
126 * to the loopback interface instead of the interface where the
127 * packets for those addresses are received.
128 */
129static int ip_checkinterface = 0;
130SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
131 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
132
133struct pfil_head inet_pfil_hook; /* Packet filter hooks */
134
135static struct ifqueue ipintrq;
136static int ipqmaxlen = IFQ_MAXLEN;
137
138extern struct domain inetdomain;
139extern struct protosw inetsw[];
140u_char ip_protox[IPPROTO_MAX];
141struct in_ifaddrhead in_ifaddrhead; /* first inet address */
142struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
143u_long in_ifaddrhmask; /* mask for hash table */
144
145SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
146 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
147SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
148 &ipintrq.ifq_drops, 0,
149 "Number of packets dropped from the IP input queue");
150
151struct ipstat ipstat;
152SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
153 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
154
155/*
156 * IP datagram reassembly.
157 */
158#define IPREASS_NHASH_LOG2 6
159#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
160#define IPREASS_HMASK (IPREASS_NHASH - 1)
161#define IPREASS_HASH(x,y) \
162 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
163
164static uma_zone_t ipq_zone;
165static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
166static struct mtx ipqlock;
167
168#define IPQ_LOCK() mtx_lock(&ipqlock)
169#define IPQ_UNLOCK() mtx_unlock(&ipqlock)
170#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
171#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
172
173static void maxnipq_update(void);
174static void ipq_zone_change(void *);
175
176static int maxnipq; /* Administrative limit on # reass queues. */
177static int nipq = 0; /* Total # of reass queues */
178SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
179 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
180
181static int maxfragsperpacket;
182SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
183 &maxfragsperpacket, 0,
184 "Maximum number of IPv4 fragments allowed per packet");
185
186struct callout ipport_tick_callout;
187
188#ifdef IPCTL_DEFMTU
189SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
190 &ip_mtu, 0, "Default MTU");
191#endif
192
193#ifdef IPSTEALTH
194int ipstealth = 0;
195SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
196 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
197#endif
198
199/*
200 * ipfw_ether and ipfw_bridge hooks.
201 * XXX: Temporary until those are converted to pfil_hooks as well.
202 */
203ip_fw_chk_t *ip_fw_chk_ptr = NULL;
204ip_dn_io_t *ip_dn_io_ptr = NULL;
205int fw_one_pass = 1;
206
207static void ip_freef(struct ipqhead *, struct ipq *);
208
209/*
210 * IP initialization: fill in IP protocol switch table.
211 * All protocols not implemented in kernel go to raw IP protocol handler.
212 */
213void
214ip_init(void)
215{
216 struct protosw *pr;
217 int i;
218
219 TAILQ_INIT(&in_ifaddrhead);
220 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
221 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
222 if (pr == NULL)
223 panic("ip_init: PF_INET not found");
224
225 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
226 for (i = 0; i < IPPROTO_MAX; i++)
227 ip_protox[i] = pr - inetsw;
228 /*
229 * Cycle through IP protocols and put them into the appropriate place
230 * in ip_protox[].
231 */
232 for (pr = inetdomain.dom_protosw;
233 pr < inetdomain.dom_protoswNPROTOSW; pr++)
234 if (pr->pr_domain->dom_family == PF_INET &&
235 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
236 /* Be careful to only index valid IP protocols. */
237 if (pr->pr_protocol < IPPROTO_MAX)
238 ip_protox[pr->pr_protocol] = pr - inetsw;
239 }
240
241 /* Initialize packet filter hooks. */
242 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
243 inet_pfil_hook.ph_af = AF_INET;
244 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
245 printf("%s: WARNING: unable to register pfil hook, "
246 "error %d\n", __func__, i);
247
248 /* Initialize IP reassembly queue. */
249 IPQ_LOCK_INIT();
250 for (i = 0; i < IPREASS_NHASH; i++)
251 TAILQ_INIT(&ipq[i]);
252 maxnipq = nmbclusters / 32;
253 maxfragsperpacket = 16;
254 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
255 NULL, UMA_ALIGN_PTR, 0);
256 maxnipq_update();
257
258 /* Start ipport_tick. */
259 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
260 ipport_tick(NULL);
261 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
262 SHUTDOWN_PRI_DEFAULT);
263 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
264 NULL, EVENTHANDLER_PRI_ANY);
265
266 /* Initialize various other remaining things. */
267 ip_id = time_second & 0xffff;
268 ipintrq.ifq_maxlen = ipqmaxlen;
269 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
270 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
271}
272
273void
274ip_fini(void *xtp)
275{
276
277 callout_stop(&ipport_tick_callout);
278}
279
280/*
281 * Ip input routine. Checksum and byte swap header. If fragmented
282 * try to reassemble. Process options. Pass to next level.
283 */
284void
285ip_input(struct mbuf *m)
286{
287 struct ip *ip = NULL;
288 struct in_ifaddr *ia = NULL;
289 struct ifaddr *ifa;
290 int checkif, hlen = 0;
291 u_short sum;
292 int dchg = 0; /* dest changed after fw */
293 struct in_addr odst; /* original dst address */
294
295 M_ASSERTPKTHDR(m);
296
297 if (m->m_flags & M_FASTFWD_OURS) {
298 /*
299 * Firewall or NAT changed destination to local.
300 * We expect ip_len and ip_off to be in host byte order.
301 */
302 m->m_flags &= ~M_FASTFWD_OURS;
303 /* Set up some basics that will be used later. */
304 ip = mtod(m, struct ip *);
305 hlen = ip->ip_hl << 2;
306 goto ours;
307 }
308
309 ipstat.ips_total++;
310
311 if (m->m_pkthdr.len < sizeof(struct ip))
312 goto tooshort;
313
314 if (m->m_len < sizeof (struct ip) &&
315 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
316 ipstat.ips_toosmall++;
317 return;
318 }
319 ip = mtod(m, struct ip *);
320
321 if (ip->ip_v != IPVERSION) {
322 ipstat.ips_badvers++;
323 goto bad;
324 }
325
326 hlen = ip->ip_hl << 2;
327 if (hlen < sizeof(struct ip)) { /* minimum header length */
328 ipstat.ips_badhlen++;
329 goto bad;
330 }
331 if (hlen > m->m_len) {
332 if ((m = m_pullup(m, hlen)) == NULL) {
333 ipstat.ips_badhlen++;
334 return;
335 }
336 ip = mtod(m, struct ip *);
337 }
338
339 /* 127/8 must not appear on wire - RFC1122 */
340 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
341 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
342 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
343 ipstat.ips_badaddr++;
344 goto bad;
345 }
346 }
347
348 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
349 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
350 } else {
351 if (hlen == sizeof(struct ip)) {
352 sum = in_cksum_hdr(ip);
353 } else {
354 sum = in_cksum(m, hlen);
355 }
356 }
357 if (sum) {
358 ipstat.ips_badsum++;
359 goto bad;
360 }
361
362#ifdef ALTQ
363 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
364 /* packet is dropped by traffic conditioner */
365 return;
366#endif
367
368 /*
369 * Convert fields to host representation.
370 */
371 ip->ip_len = ntohs(ip->ip_len);
372 if (ip->ip_len < hlen) {
373 ipstat.ips_badlen++;
374 goto bad;
375 }
376 ip->ip_off = ntohs(ip->ip_off);
377
378 /*
379 * Check that the amount of data in the buffers
380 * is as at least much as the IP header would have us expect.
381 * Trim mbufs if longer than we expect.
382 * Drop packet if shorter than we expect.
383 */
384 if (m->m_pkthdr.len < ip->ip_len) {
385tooshort:
386 ipstat.ips_tooshort++;
387 goto bad;
388 }
389 if (m->m_pkthdr.len > ip->ip_len) {
390 if (m->m_len == m->m_pkthdr.len) {
391 m->m_len = ip->ip_len;
392 m->m_pkthdr.len = ip->ip_len;
393 } else
394 m_adj(m, ip->ip_len - m->m_pkthdr.len);
395 }
396#ifdef IPSEC
397 /*
398 * Bypass packet filtering for packets from a tunnel (gif).
399 */
400 if (ip_ipsec_filtertunnel(m))
401 goto passin;
402#endif /* IPSEC */
403
404 /*
405 * Run through list of hooks for input packets.
406 *
407 * NB: Beware of the destination address changing (e.g.
408 * by NAT rewriting). When this happens, tell
409 * ip_forward to do the right thing.
410 */
411
412 /* Jump over all PFIL processing if hooks are not active. */
413 if (!PFIL_HOOKED(&inet_pfil_hook))
414 goto passin;
415
416 odst = ip->ip_dst;
417 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
418 PFIL_IN, NULL) != 0)
419 return;
420 if (m == NULL) /* consumed by filter */
421 return;
422
423 ip = mtod(m, struct ip *);
424 dchg = (odst.s_addr != ip->ip_dst.s_addr);
425
426#ifdef IPFIREWALL_FORWARD
427 if (m->m_flags & M_FASTFWD_OURS) {
428 m->m_flags &= ~M_FASTFWD_OURS;
429 goto ours;
430 }
431 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
432 /*
433 * Directly ship on the packet. This allows to forward packets
434 * that were destined for us to some other directly connected
435 * host.
436 */
437 ip_forward(m, dchg);
438 return;
439 }
440#endif /* IPFIREWALL_FORWARD */
441
442passin:
443 /*
444 * Process options and, if not destined for us,
445 * ship it on. ip_dooptions returns 1 when an
446 * error was detected (causing an icmp message
447 * to be sent and the original packet to be freed).
448 */
449 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
450 return;
451
452 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
453 * matter if it is destined to another node, or whether it is
454 * a multicast one, RSVP wants it! and prevents it from being forwarded
455 * anywhere else. Also checks if the rsvp daemon is running before
456 * grabbing the packet.
457 */
458 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
459 goto ours;
460
461 /*
462 * Check our list of addresses, to see if the packet is for us.
463 * If we don't have any addresses, assume any unicast packet
464 * we receive might be for us (and let the upper layers deal
465 * with it).
466 */
467 if (TAILQ_EMPTY(&in_ifaddrhead) &&
468 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
469 goto ours;
470
471 /*
472 * Enable a consistency check between the destination address
473 * and the arrival interface for a unicast packet (the RFC 1122
474 * strong ES model) if IP forwarding is disabled and the packet
475 * is not locally generated and the packet is not subject to
476 * 'ipfw fwd'.
477 *
478 * XXX - Checking also should be disabled if the destination
479 * address is ipnat'ed to a different interface.
480 *
481 * XXX - Checking is incompatible with IP aliases added
482 * to the loopback interface instead of the interface where
483 * the packets are received.
484 *
485 * XXX - This is the case for carp vhost IPs as well so we
486 * insert a workaround. If the packet got here, we already
487 * checked with carp_iamatch() and carp_forus().
488 */
489 checkif = ip_checkinterface && (ipforwarding == 0) &&
490 m->m_pkthdr.rcvif != NULL &&
491 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
492#ifdef DEV_CARP
493 !m->m_pkthdr.rcvif->if_carp &&
494#endif
495 (dchg == 0);
496
497 /*
498 * Check for exact addresses in the hash bucket.
499 */
500 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
501 /*
502 * If the address matches, verify that the packet
503 * arrived via the correct interface if checking is
504 * enabled.
505 */
506 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
507 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
508 goto ours;
509 }
510 /*
511 * Check for broadcast addresses.
512 *
513 * Only accept broadcast packets that arrive via the matching
514 * interface. Reception of forwarded directed broadcasts would
515 * be handled via ip_forward() and ether_output() with the loopback
516 * into the stack for SIMPLEX interfaces handled by ether_output().
517 */
518 if (m->m_pkthdr.rcvif != NULL &&
519 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
520 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
521 if (ifa->ifa_addr->sa_family != AF_INET)
522 continue;
523 ia = ifatoia(ifa);
524 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
525 ip->ip_dst.s_addr)
526 goto ours;
527 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
528 goto ours;
529#ifdef BOOTP_COMPAT
530 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
531 goto ours;
532#endif
533 }
534 }
535 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
536 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
537 ipstat.ips_cantforward++;
538 m_freem(m);
539 return;
540 }
541 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
542 struct in_multi *inm;
543 if (ip_mrouter) {
544 /*
545 * If we are acting as a multicast router, all
546 * incoming multicast packets are passed to the
547 * kernel-level multicast forwarding function.
548 * The packet is returned (relatively) intact; if
549 * ip_mforward() returns a non-zero value, the packet
550 * must be discarded, else it may be accepted below.
551 */
552 if (ip_mforward &&
553 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
554 ipstat.ips_cantforward++;
555 m_freem(m);
556 return;
557 }
558
559 /*
560 * The process-level routing daemon needs to receive
561 * all multicast IGMP packets, whether or not this
562 * host belongs to their destination groups.
563 */
564 if (ip->ip_p == IPPROTO_IGMP)
565 goto ours;
566 ipstat.ips_forward++;
567 }
568 /*
569 * See if we belong to the destination multicast group on the
570 * arrival interface.
571 */
572 IN_MULTI_LOCK();
573 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
574 IN_MULTI_UNLOCK();
575 if (inm == NULL) {
576 ipstat.ips_notmember++;
577 m_freem(m);
578 return;
579 }
580 goto ours;
581 }
582 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
583 goto ours;
584 if (ip->ip_dst.s_addr == INADDR_ANY)
585 goto ours;
586
587 /*
588 * FAITH(Firewall Aided Internet Translator)
589 */
590 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
591 if (ip_keepfaith) {
592 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
593 goto ours;
594 }
595 m_freem(m);
596 return;
597 }
598
599 /*
600 * Not for us; forward if possible and desirable.
601 */
602 if (ipforwarding == 0) {
603 ipstat.ips_cantforward++;
604 m_freem(m);
605 } else {
606#ifdef IPSEC
607 if (ip_ipsec_fwd(m))
608 goto bad;
609#endif /* IPSEC */
610 ip_forward(m, dchg);
611 }
612 return;
613
614ours:
615#ifdef IPSTEALTH
616 /*
617 * IPSTEALTH: Process non-routing options only
618 * if the packet is destined for us.
619 */
620 if (ipstealth && hlen > sizeof (struct ip) &&
621 ip_dooptions(m, 1))
622 return;
623#endif /* IPSTEALTH */
624
625 /* Count the packet in the ip address stats */
626 if (ia != NULL) {
627 ia->ia_ifa.if_ipackets++;
628 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
629 }
630
631 /*
632 * Attempt reassembly; if it succeeds, proceed.
633 * ip_reass() will return a different mbuf.
634 */
635 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
636 m = ip_reass(m);
637 if (m == NULL)
638 return;
639 ip = mtod(m, struct ip *);
640 /* Get the header length of the reassembled packet */
641 hlen = ip->ip_hl << 2;
642 }
643
644 /*
645 * Further protocols expect the packet length to be w/o the
646 * IP header.
647 */
648 ip->ip_len -= hlen;
649
650#ifdef IPSEC
651 /*
652 * enforce IPsec policy checking if we are seeing last header.
653 * note that we do not visit this with protocols with pcb layer
654 * code - like udp/tcp/raw ip.
655 */
656 if (ip_ipsec_input(m))
657 goto bad;
658#endif /* IPSEC */
659
660 /*
661 * Switch out to protocol's input routine.
662 */
663 ipstat.ips_delivered++;
664
665 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
666 return;
667bad:
668 m_freem(m);
669}
670
671/*
672 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
673 * max has slightly different semantics than the sysctl, for historical
674 * reasons.
675 */
676static void
677maxnipq_update(void)
678{
679
680 /*
681 * -1 for unlimited allocation.
682 */
683 if (maxnipq < 0)
684 uma_zone_set_max(ipq_zone, 0);
685 /*
686 * Positive number for specific bound.
687 */
688 if (maxnipq > 0)
689 uma_zone_set_max(ipq_zone, maxnipq);
690 /*
691 * Zero specifies no further fragment queue allocation -- set the
692 * bound very low, but rely on implementation elsewhere to actually
693 * prevent allocation and reclaim current queues.
694 */
695 if (maxnipq == 0)
696 uma_zone_set_max(ipq_zone, 1);
697}
698
699static void
700ipq_zone_change(void *tag)
701{
702
703 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
704 maxnipq = nmbclusters / 32;
705 maxnipq_update();
706 }
707}
708
709static int
710sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
711{
712 int error, i;
713
714 i = maxnipq;
715 error = sysctl_handle_int(oidp, &i, 0, req);
716 if (error || !req->newptr)
717 return (error);
718
719 /*
720 * XXXRW: Might be a good idea to sanity check the argument and place
721 * an extreme upper bound.
722 */
723 if (i < -1)
724 return (EINVAL);
725 maxnipq = i;
726 maxnipq_update();
727 return (0);
728}
729
730SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
731 NULL, 0, sysctl_maxnipq, "I",
732 "Maximum number of IPv4 fragment reassembly queue entries");
733
734/*
735 * Take incoming datagram fragment and try to reassemble it into
736 * whole datagram. If the argument is the first fragment or one
737 * in between the function will return NULL and store the mbuf
738 * in the fragment chain. If the argument is the last fragment
739 * the packet will be reassembled and the pointer to the new
740 * mbuf returned for further processing. Only m_tags attached
741 * to the first packet/fragment are preserved.
742 * The IP header is *NOT* adjusted out of iplen.
743 */
744struct mbuf *
745ip_reass(struct mbuf *m)
746{
747 struct ip *ip;
748 struct mbuf *p, *q, *nq, *t;
749 struct ipq *fp = NULL;
750 struct ipqhead *head;
751 int i, hlen, next;
752 u_int8_t ecn, ecn0;
753 u_short hash;
754
755 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
756 if (maxnipq == 0 || maxfragsperpacket == 0) {
757 ipstat.ips_fragments++;
758 ipstat.ips_fragdropped++;
759 m_freem(m);
760 return (NULL);
761 }
762
763 ip = mtod(m, struct ip *);
764 hlen = ip->ip_hl << 2;
765
766 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
767 head = &ipq[hash];
768 IPQ_LOCK();
769
770 /*
771 * Look for queue of fragments
772 * of this datagram.
773 */
774 TAILQ_FOREACH(fp, head, ipq_list)
775 if (ip->ip_id == fp->ipq_id &&
776 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
777 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
778#ifdef MAC
779 mac_ipq_match(m, fp) &&
780#endif
781 ip->ip_p == fp->ipq_p)
782 goto found;
783
784 fp = NULL;
785
786 /*
787 * Attempt to trim the number of allocated fragment queues if it
788 * exceeds the administrative limit.
789 */
790 if ((nipq > maxnipq) && (maxnipq > 0)) {
791 /*
792 * drop something from the tail of the current queue
793 * before proceeding further
794 */
795 struct ipq *q = TAILQ_LAST(head, ipqhead);
796 if (q == NULL) { /* gak */
797 for (i = 0; i < IPREASS_NHASH; i++) {
798 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
799 if (r) {
800 ipstat.ips_fragtimeout += r->ipq_nfrags;
801 ip_freef(&ipq[i], r);
802 break;
803 }
804 }
805 } else {
806 ipstat.ips_fragtimeout += q->ipq_nfrags;
807 ip_freef(head, q);
808 }
809 }
810
811found:
812 /*
813 * Adjust ip_len to not reflect header,
814 * convert offset of this to bytes.
815 */
816 ip->ip_len -= hlen;
817 if (ip->ip_off & IP_MF) {
818 /*
819 * Make sure that fragments have a data length
820 * that's a non-zero multiple of 8 bytes.
821 */
822 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
823 ipstat.ips_toosmall++; /* XXX */
824 goto dropfrag;
825 }
826 m->m_flags |= M_FRAG;
827 } else
828 m->m_flags &= ~M_FRAG;
829 ip->ip_off <<= 3;
830
831
832 /*
833 * Attempt reassembly; if it succeeds, proceed.
834 * ip_reass() will return a different mbuf.
835 */
836 ipstat.ips_fragments++;
837 m->m_pkthdr.header = ip;
838
839 /* Previous ip_reass() started here. */
840 /*
841 * Presence of header sizes in mbufs
842 * would confuse code below.
843 */
844 m->m_data += hlen;
845 m->m_len -= hlen;
846
847 /*
848 * If first fragment to arrive, create a reassembly queue.
849 */
850 if (fp == NULL) {
851 fp = uma_zalloc(ipq_zone, M_NOWAIT);
852 if (fp == NULL)
853 goto dropfrag;
854#ifdef MAC
855 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
856 uma_zfree(ipq_zone, fp);
857 fp = NULL;
858 goto dropfrag;
859 }
860 mac_ipq_create(m, fp);
861#endif
862 TAILQ_INSERT_HEAD(head, fp, ipq_list);
863 nipq++;
864 fp->ipq_nfrags = 1;
865 fp->ipq_ttl = IPFRAGTTL;
866 fp->ipq_p = ip->ip_p;
867 fp->ipq_id = ip->ip_id;
868 fp->ipq_src = ip->ip_src;
869 fp->ipq_dst = ip->ip_dst;
870 fp->ipq_frags = m;
871 m->m_nextpkt = NULL;
872 goto done;
873 } else {
874 fp->ipq_nfrags++;
875#ifdef MAC
876 mac_ipq_update(m, fp);
877#endif
878 }
879
880#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
881
882 /*
883 * Handle ECN by comparing this segment with the first one;
884 * if CE is set, do not lose CE.
885 * drop if CE and not-ECT are mixed for the same packet.
886 */
887 ecn = ip->ip_tos & IPTOS_ECN_MASK;
888 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
889 if (ecn == IPTOS_ECN_CE) {
890 if (ecn0 == IPTOS_ECN_NOTECT)
891 goto dropfrag;
892 if (ecn0 != IPTOS_ECN_CE)
893 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
894 }
895 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
896 goto dropfrag;
897
898 /*
899 * Find a segment which begins after this one does.
900 */
901 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
902 if (GETIP(q)->ip_off > ip->ip_off)
903 break;
904
905 /*
906 * If there is a preceding segment, it may provide some of
907 * our data already. If so, drop the data from the incoming
908 * segment. If it provides all of our data, drop us, otherwise
909 * stick new segment in the proper place.
910 *
911 * If some of the data is dropped from the the preceding
912 * segment, then it's checksum is invalidated.
913 */
914 if (p) {
915 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
916 if (i > 0) {
917 if (i >= ip->ip_len)
918 goto dropfrag;
919 m_adj(m, i);
920 m->m_pkthdr.csum_flags = 0;
921 ip->ip_off += i;
922 ip->ip_len -= i;
923 }
924 m->m_nextpkt = p->m_nextpkt;
925 p->m_nextpkt = m;
926 } else {
927 m->m_nextpkt = fp->ipq_frags;
928 fp->ipq_frags = m;
929 }
930
931 /*
932 * While we overlap succeeding segments trim them or,
933 * if they are completely covered, dequeue them.
934 */
935 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
936 q = nq) {
937 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
938 if (i < GETIP(q)->ip_len) {
939 GETIP(q)->ip_len -= i;
940 GETIP(q)->ip_off += i;
941 m_adj(q, i);
942 q->m_pkthdr.csum_flags = 0;
943 break;
944 }
945 nq = q->m_nextpkt;
946 m->m_nextpkt = nq;
947 ipstat.ips_fragdropped++;
948 fp->ipq_nfrags--;
949 m_freem(q);
950 }
951
952 /*
953 * Check for complete reassembly and perform frag per packet
954 * limiting.
955 *
956 * Frag limiting is performed here so that the nth frag has
957 * a chance to complete the packet before we drop the packet.
958 * As a result, n+1 frags are actually allowed per packet, but
959 * only n will ever be stored. (n = maxfragsperpacket.)
960 *
961 */
962 next = 0;
963 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
964 if (GETIP(q)->ip_off != next) {
965 if (fp->ipq_nfrags > maxfragsperpacket) {
966 ipstat.ips_fragdropped += fp->ipq_nfrags;
967 ip_freef(head, fp);
968 }
969 goto done;
970 }
971 next += GETIP(q)->ip_len;
972 }
973 /* Make sure the last packet didn't have the IP_MF flag */
974 if (p->m_flags & M_FRAG) {
975 if (fp->ipq_nfrags > maxfragsperpacket) {
976 ipstat.ips_fragdropped += fp->ipq_nfrags;
977 ip_freef(head, fp);
978 }
979 goto done;
980 }
981
982 /*
983 * Reassembly is complete. Make sure the packet is a sane size.
984 */
985 q = fp->ipq_frags;
986 ip = GETIP(q);
987 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
988 ipstat.ips_toolong++;
989 ipstat.ips_fragdropped += fp->ipq_nfrags;
990 ip_freef(head, fp);
991 goto done;
992 }
993
994 /*
995 * Concatenate fragments.
996 */
997 m = q;
998 t = m->m_next;
999 m->m_next = NULL;
1000 m_cat(m, t);
1001 nq = q->m_nextpkt;
1002 q->m_nextpkt = NULL;
1003 for (q = nq; q != NULL; q = nq) {
1004 nq = q->m_nextpkt;
1005 q->m_nextpkt = NULL;
1006 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1007 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1008 m_cat(m, q);
1009 }
1010 /*
1011 * In order to do checksumming faster we do 'end-around carry' here
1012 * (and not in for{} loop), though it implies we are not going to
1013 * reassemble more than 64k fragments.
1014 */
1015 m->m_pkthdr.csum_data =
1016 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1017#ifdef MAC
1018 mac_ipq_reassemble(fp, m);
1019 mac_ipq_destroy(fp);
1020#endif
1021
1022 /*
1023 * Create header for new ip packet by modifying header of first
1024 * packet; dequeue and discard fragment reassembly header.
1025 * Make header visible.
1026 */
1027 ip->ip_len = (ip->ip_hl << 2) + next;
1028 ip->ip_src = fp->ipq_src;
1029 ip->ip_dst = fp->ipq_dst;
1030 TAILQ_REMOVE(head, fp, ipq_list);
1031 nipq--;
1032 uma_zfree(ipq_zone, fp);
1033 m->m_len += (ip->ip_hl << 2);
1034 m->m_data -= (ip->ip_hl << 2);
1035 /* some debugging cruft by sklower, below, will go away soon */
1036 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1037 m_fixhdr(m);
1038 ipstat.ips_reassembled++;
1039 IPQ_UNLOCK();
1040 return (m);
1041
1042dropfrag:
1043 ipstat.ips_fragdropped++;
1044 if (fp != NULL)
1045 fp->ipq_nfrags--;
1046 m_freem(m);
1047done:
1048 IPQ_UNLOCK();
1049 return (NULL);
1050
1051#undef GETIP
1052}
1053
1054/*
1055 * Free a fragment reassembly header and all
1056 * associated datagrams.
1057 */
1058static void
1059ip_freef(struct ipqhead *fhp, struct ipq *fp)
1060{
1061 struct mbuf *q;
1062
1063 IPQ_LOCK_ASSERT();
1064
1065 while (fp->ipq_frags) {
1066 q = fp->ipq_frags;
1067 fp->ipq_frags = q->m_nextpkt;
1068 m_freem(q);
1069 }
1070 TAILQ_REMOVE(fhp, fp, ipq_list);
1071 uma_zfree(ipq_zone, fp);
1072 nipq--;
1073}
1074
1075/*
1076 * IP timer processing;
1077 * if a timer expires on a reassembly
1078 * queue, discard it.
1079 */
1080void
1081ip_slowtimo(void)
1082{
1083 struct ipq *fp;
1084 int i;
1085
1086 IPQ_LOCK();
1087 for (i = 0; i < IPREASS_NHASH; i++) {
1088 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1089 struct ipq *fpp;
1090
1091 fpp = fp;
1092 fp = TAILQ_NEXT(fp, ipq_list);
1093 if(--fpp->ipq_ttl == 0) {
1094 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1095 ip_freef(&ipq[i], fpp);
1096 }
1097 }
1098 }
1099 /*
1100 * If we are over the maximum number of fragments
1101 * (due to the limit being lowered), drain off
1102 * enough to get down to the new limit.
1103 */
1104 if (maxnipq >= 0 && nipq > maxnipq) {
1105 for (i = 0; i < IPREASS_NHASH; i++) {
1106 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1107 ipstat.ips_fragdropped +=
1108 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1109 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1110 }
1111 }
1112 }
1113 IPQ_UNLOCK();
1114}
1115
1116/*
1117 * Drain off all datagram fragments.
1118 */
1119void
1120ip_drain(void)
1121{
1122 int i;
1123
1124 IPQ_LOCK();
1125 for (i = 0; i < IPREASS_NHASH; i++) {
1126 while(!TAILQ_EMPTY(&ipq[i])) {
1127 ipstat.ips_fragdropped +=
1128 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1129 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1130 }
1131 }
1132 IPQ_UNLOCK();
1133 in_rtqdrain();
1134}
1135
1136/*
1137 * The protocol to be inserted into ip_protox[] must be already registered
1138 * in inetsw[], either statically or through pf_proto_register().
1139 */
1140int
1141ipproto_register(u_char ipproto)
1142{
1143 struct protosw *pr;
1144
1145 /* Sanity checks. */
1146 if (ipproto == 0)
1147 return (EPROTONOSUPPORT);
1148
1149 /*
1150 * The protocol slot must not be occupied by another protocol
1151 * already. An index pointing to IPPROTO_RAW is unused.
1152 */
1153 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1154 if (pr == NULL)
1155 return (EPFNOSUPPORT);
1156 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1157 return (EEXIST);
1158
1159 /* Find the protocol position in inetsw[] and set the index. */
1160 for (pr = inetdomain.dom_protosw;
1161 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1162 if (pr->pr_domain->dom_family == PF_INET &&
1163 pr->pr_protocol && pr->pr_protocol == ipproto) {
1164 /* Be careful to only index valid IP protocols. */
1165 if (pr->pr_protocol < IPPROTO_MAX) {
1166 ip_protox[pr->pr_protocol] = pr - inetsw;
1167 return (0);
1168 } else
1169 return (EINVAL);
1170 }
1171 }
1172 return (EPROTONOSUPPORT);
1173}
1174
1175int
1176ipproto_unregister(u_char ipproto)
1177{
1178 struct protosw *pr;
1179
1180 /* Sanity checks. */
1181 if (ipproto == 0)
1182 return (EPROTONOSUPPORT);
1183
1184 /* Check if the protocol was indeed registered. */
1185 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1186 if (pr == NULL)
1187 return (EPFNOSUPPORT);
1188 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1189 return (ENOENT);
1190
1191 /* Reset the protocol slot to IPPROTO_RAW. */
1192 ip_protox[ipproto] = pr - inetsw;
1193 return (0);
1194}
1195
1196/*
1197 * Given address of next destination (final or next hop),
1198 * return internet address info of interface to be used to get there.
1199 */
1200struct in_ifaddr *
1201ip_rtaddr(struct in_addr dst, u_int fibnum)
1202{
1203 struct route sro;
1204 struct sockaddr_in *sin;
1205 struct in_ifaddr *ifa;
1206
1207 bzero(&sro, sizeof(sro));
1208 sin = (struct sockaddr_in *)&sro.ro_dst;
1209 sin->sin_family = AF_INET;
1210 sin->sin_len = sizeof(*sin);
1211 sin->sin_addr = dst;
1212 in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
1213
1214 if (sro.ro_rt == NULL)
1215 return (NULL);
1216
1217 ifa = ifatoia(sro.ro_rt->rt_ifa);
1218 RTFREE(sro.ro_rt);
1219 return (ifa);
1220}
1221
1222u_char inetctlerrmap[PRC_NCMDS] = {
1223 0, 0, 0, 0,
1224 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1225 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1226 EMSGSIZE, EHOSTUNREACH, 0, 0,
1227 0, 0, EHOSTUNREACH, 0,
1228 ENOPROTOOPT, ECONNREFUSED
1229};
1230
1231/*
1232 * Forward a packet. If some error occurs return the sender
1233 * an icmp packet. Note we can't always generate a meaningful
1234 * icmp message because icmp doesn't have a large enough repertoire
1235 * of codes and types.
1236 *
1237 * If not forwarding, just drop the packet. This could be confusing
1238 * if ipforwarding was zero but some routing protocol was advancing
1239 * us as a gateway to somewhere. However, we must let the routing
1240 * protocol deal with that.
1241 *
1242 * The srcrt parameter indicates whether the packet is being forwarded
1243 * via a source route.
1244 */
1245void
1246ip_forward(struct mbuf *m, int srcrt)
1247{
1248 struct ip *ip = mtod(m, struct ip *);
1249 struct in_ifaddr *ia = NULL;
1250 struct mbuf *mcopy;
1251 struct in_addr dest;
1252 struct route ro;
1253 int error, type = 0, code = 0, mtu = 0;
1254
1255 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1256 ipstat.ips_cantforward++;
1257 m_freem(m);
1258 return;
1259 }
1260#ifdef IPSTEALTH
1261 if (!ipstealth) {
1262#endif
1263 if (ip->ip_ttl <= IPTTLDEC) {
1264 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1265 0, 0);
1266 return;
1267 }
1268#ifdef IPSTEALTH
1269 }
1270#endif
1271
1272 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1273 if (!srcrt && ia == NULL) {
1274 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1275 return;
1276 }
1277
1278 /*
1279 * Save the IP header and at most 8 bytes of the payload,
1280 * in case we need to generate an ICMP message to the src.
1281 *
1282 * XXX this can be optimized a lot by saving the data in a local
1283 * buffer on the stack (72 bytes at most), and only allocating the
1284 * mbuf if really necessary. The vast majority of the packets
1285 * are forwarded without having to send an ICMP back (either
1286 * because unnecessary, or because rate limited), so we are
1287 * really we are wasting a lot of work here.
1288 *
1289 * We don't use m_copy() because it might return a reference
1290 * to a shared cluster. Both this function and ip_output()
1291 * assume exclusive access to the IP header in `m', so any
1292 * data in a cluster may change before we reach icmp_error().
1293 */
1294 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1295 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1296 /*
1297 * It's probably ok if the pkthdr dup fails (because
1298 * the deep copy of the tag chain failed), but for now
1299 * be conservative and just discard the copy since
1300 * code below may some day want the tags.
1301 */
1302 m_free(mcopy);
1303 mcopy = NULL;
1304 }
1305 if (mcopy != NULL) {
1306 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1307 mcopy->m_pkthdr.len = mcopy->m_len;
1308 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1309 }
1310
1311#ifdef IPSTEALTH
1312 if (!ipstealth) {
1313#endif
1314 ip->ip_ttl -= IPTTLDEC;
1315#ifdef IPSTEALTH
1316 }
1317#endif
1318
1319 /*
1320 * If forwarding packet using same interface that it came in on,
1321 * perhaps should send a redirect to sender to shortcut a hop.
1322 * Only send redirect if source is sending directly to us,
1323 * and if packet was not source routed (or has any options).
1324 * Also, don't send redirect if forwarding using a default route
1325 * or a route modified by a redirect.
1326 */
1327 dest.s_addr = 0;
1328 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1329 struct sockaddr_in *sin;
1330 struct rtentry *rt;
1331
1332 bzero(&ro, sizeof(ro));
1333 sin = (struct sockaddr_in *)&ro.ro_dst;
1334 sin->sin_family = AF_INET;
1335 sin->sin_len = sizeof(*sin);
1336 sin->sin_addr = ip->ip_dst;
1337 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
1338
1339 rt = ro.ro_rt;
1340
1341 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1342 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1343#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1344 u_long src = ntohl(ip->ip_src.s_addr);
1345
1346 if (RTA(rt) &&
1347 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1348 if (rt->rt_flags & RTF_GATEWAY)
1349 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1350 else
1351 dest.s_addr = ip->ip_dst.s_addr;
1352 /* Router requirements says to only send host redirects */
1353 type = ICMP_REDIRECT;
1354 code = ICMP_REDIRECT_HOST;
1355 }
1356 }
1357 if (rt)
1358 RTFREE(rt);
1359 }
1360
1361 /*
1362 * Try to cache the route MTU from ip_output so we can consider it for
1363 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1364 */
1365 bzero(&ro, sizeof(ro));
1366
1367 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1368
1369 if (error == EMSGSIZE && ro.ro_rt)
1370 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1371 if (ro.ro_rt)
1372 RTFREE(ro.ro_rt);
1373
1374 if (error)
1375 ipstat.ips_cantforward++;
1376 else {
1377 ipstat.ips_forward++;
1378 if (type)
1379 ipstat.ips_redirectsent++;
1380 else {
1381 if (mcopy)
1382 m_freem(mcopy);
1383 return;
1384 }
1385 }
1386 if (mcopy == NULL)
1387 return;
1388
1389 switch (error) {
1390
1391 case 0: /* forwarded, but need redirect */
1392 /* type, code set above */
1393 break;
1394
1395 case ENETUNREACH: /* shouldn't happen, checked above */
1396 case EHOSTUNREACH:
1397 case ENETDOWN:
1398 case EHOSTDOWN:
1399 default:
1400 type = ICMP_UNREACH;
1401 code = ICMP_UNREACH_HOST;
1402 break;
1403
1404 case EMSGSIZE:
1405 type = ICMP_UNREACH;
1406 code = ICMP_UNREACH_NEEDFRAG;
1407
1408#ifdef IPSEC
1409 /*
1410 * If IPsec is configured for this path,
1411 * override any possibly mtu value set by ip_output.
1412 */
1413 mtu = ip_ipsec_mtu(m, mtu);
1414#endif /* IPSEC */
1415 /*
1416 * If the MTU was set before make sure we are below the
1417 * interface MTU.
1418 * If the MTU wasn't set before use the interface mtu or
1419 * fall back to the next smaller mtu step compared to the
1420 * current packet size.
1421 */
1422 if (mtu != 0) {
1423 if (ia != NULL)
1424 mtu = min(mtu, ia->ia_ifp->if_mtu);
1425 } else {
1426 if (ia != NULL)
1427 mtu = ia->ia_ifp->if_mtu;
1428 else
1429 mtu = ip_next_mtu(ip->ip_len, 0);
1430 }
1431 ipstat.ips_cantfrag++;
1432 break;
1433
1434 case ENOBUFS:
1435 /*
1436 * A router should not generate ICMP_SOURCEQUENCH as
1437 * required in RFC1812 Requirements for IP Version 4 Routers.
1438 * Source quench could be a big problem under DoS attacks,
1439 * or if the underlying interface is rate-limited.
1440 * Those who need source quench packets may re-enable them
1441 * via the net.inet.ip.sendsourcequench sysctl.
1442 */
1443 if (ip_sendsourcequench == 0) {
1444 m_freem(mcopy);
1445 return;
1446 } else {
1447 type = ICMP_SOURCEQUENCH;
1448 code = 0;
1449 }
1450 break;
1451
1452 case EACCES: /* ipfw denied packet */
1453 m_freem(mcopy);
1454 return;
1455 }
1456 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1457}
1458
1459void
1460ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1461 struct mbuf *m)
1462{
1463 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1464 struct bintime bt;
1465
1466 bintime(&bt);
1467 if (inp->inp_socket->so_options & SO_BINTIME) {
1468 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1469 SCM_BINTIME, SOL_SOCKET);
1470 if (*mp)
1471 mp = &(*mp)->m_next;
1472 }
1473 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1474 struct timeval tv;
1475
1476 bintime2timeval(&bt, &tv);
1477 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1478 SCM_TIMESTAMP, SOL_SOCKET);
1479 if (*mp)
1480 mp = &(*mp)->m_next;
1481 }
1482 }
1483 if (inp->inp_flags & INP_RECVDSTADDR) {
1484 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1485 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1486 if (*mp)
1487 mp = &(*mp)->m_next;
1488 }
1489 if (inp->inp_flags & INP_RECVTTL) {
1490 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1491 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1492 if (*mp)
1493 mp = &(*mp)->m_next;
1494 }
1495#ifdef notyet
1496 /* XXX
1497 * Moving these out of udp_input() made them even more broken
1498 * than they already were.
1499 */
1500 /* options were tossed already */
1501 if (inp->inp_flags & INP_RECVOPTS) {
1502 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1503 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1504 if (*mp)
1505 mp = &(*mp)->m_next;
1506 }
1507 /* ip_srcroute doesn't do what we want here, need to fix */
1508 if (inp->inp_flags & INP_RECVRETOPTS) {
1509 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1510 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1511 if (*mp)
1512 mp = &(*mp)->m_next;
1513 }
1514#endif
1515 if (inp->inp_flags & INP_RECVIF) {
1516 struct ifnet *ifp;
1517 struct sdlbuf {
1518 struct sockaddr_dl sdl;
1519 u_char pad[32];
1520 } sdlbuf;
1521 struct sockaddr_dl *sdp;
1522 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1523
1524 if (((ifp = m->m_pkthdr.rcvif))
1525 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1526 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1527 /*
1528 * Change our mind and don't try copy.
1529 */
1530 if ((sdp->sdl_family != AF_LINK)
1531 || (sdp->sdl_len > sizeof(sdlbuf))) {
1532 goto makedummy;
1533 }
1534 bcopy(sdp, sdl2, sdp->sdl_len);
1535 } else {
1536makedummy:
1537 sdl2->sdl_len
1538 = offsetof(struct sockaddr_dl, sdl_data[0]);
1539 sdl2->sdl_family = AF_LINK;
1540 sdl2->sdl_index = 0;
1541 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1542 }
1543 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1544 IP_RECVIF, IPPROTO_IP);
1545 if (*mp)
1546 mp = &(*mp)->m_next;
1547 }
1548}
1549
1550/*
1551 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1552 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1553 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1554 * compiled.
1555 */
1556static int ip_rsvp_on;
1557struct socket *ip_rsvpd;
1558int
1559ip_rsvp_init(struct socket *so)
1560{
1561 if (so->so_type != SOCK_RAW ||
1562 so->so_proto->pr_protocol != IPPROTO_RSVP)
1563 return EOPNOTSUPP;
1564
1565 if (ip_rsvpd != NULL)
1566 return EADDRINUSE;
1567
1568 ip_rsvpd = so;
1569 /*
1570 * This may seem silly, but we need to be sure we don't over-increment
1571 * the RSVP counter, in case something slips up.
1572 */
1573 if (!ip_rsvp_on) {
1574 ip_rsvp_on = 1;
1575 rsvp_on++;
1576 }
1577
1578 return 0;
1579}
1580
1581int
1582ip_rsvp_done(void)
1583{
1584 ip_rsvpd = NULL;
1585 /*
1586 * This may seem silly, but we need to be sure we don't over-decrement
1587 * the RSVP counter, in case something slips up.
1588 */
1589 if (ip_rsvp_on) {
1590 ip_rsvp_on = 0;
1591 rsvp_on--;
1592 }
1593 return 0;
1594}
1595
1596void
1597rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1598{
1599 if (rsvp_input_p) { /* call the real one if loaded */
1600 rsvp_input_p(m, off);
1601 return;
1602 }
1603
1604 /* Can still get packets with rsvp_on = 0 if there is a local member
1605 * of the group to which the RSVP packet is addressed. But in this
1606 * case we want to throw the packet away.
1607 */
1608
1609 if (!rsvp_on) {
1610 m_freem(m);
1611 return;
1612 }
1613
1614 if (ip_rsvpd != NULL) {
1615 rip_input(m, off);
1616 return;
1617 }
1618 /* Drop the packet */
1619 m_freem(m);
1620}
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/netinet/ip_input.c 180215 2008-07-03 12:44:36Z bz $");
34
35#include "opt_bootp.h"
36#include "opt_ipfw.h"
37#include "opt_ipstealth.h"
38#include "opt_ipsec.h"
39#include "opt_mac.h"
40#include "opt_carp.h"
41
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/callout.h>
45#include <sys/mbuf.h>
46#include <sys/malloc.h>
47#include <sys/domain.h>
48#include <sys/protosw.h>
49#include <sys/socket.h>
50#include <sys/time.h>
51#include <sys/kernel.h>
52#include <sys/syslog.h>
53#include <sys/sysctl.h>
54
55#include <net/pfil.h>
56#include <net/if.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>
62
63#include <netinet/in.h>
64#include <netinet/in_systm.h>
65#include <netinet/in_var.h>
66#include <netinet/ip.h>
67#include <netinet/in_pcb.h>
68#include <netinet/ip_var.h>
69#include <netinet/ip_icmp.h>
70#include <netinet/ip_options.h>
71#include <machine/in_cksum.h>
72#ifdef DEV_CARP
73#include <netinet/ip_carp.h>
74#endif
75#ifdef IPSEC
76#include <netinet/ip_ipsec.h>
77#endif /* IPSEC */
78
79#include <sys/socketvar.h>
80
81/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
82#include <netinet/ip_fw.h>
83#include <netinet/ip_dummynet.h>
84
85#include <security/mac/mac_framework.h>
86
87int rsvp_on = 0;
88
89int ipforwarding = 0;
90SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
91 &ipforwarding, 0, "Enable IP forwarding between interfaces");
92
93static int ipsendredirects = 1; /* XXX */
94SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
95 &ipsendredirects, 0, "Enable sending IP redirects");
96
97int ip_defttl = IPDEFTTL;
98SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
99 &ip_defttl, 0, "Maximum TTL on IP packets");
100
101static int ip_keepfaith = 0;
102SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
103 &ip_keepfaith, 0,
104 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
105
106static int ip_sendsourcequench = 0;
107SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
108 &ip_sendsourcequench, 0,
109 "Enable the transmission of source quench packets");
110
111int ip_do_randomid = 0;
112SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
113 &ip_do_randomid, 0,
114 "Assign random ip_id values");
115
116/*
117 * XXX - Setting ip_checkinterface mostly implements the receive side of
118 * the Strong ES model described in RFC 1122, but since the routing table
119 * and transmit implementation do not implement the Strong ES model,
120 * setting this to 1 results in an odd hybrid.
121 *
122 * XXX - ip_checkinterface currently must be disabled if you use ipnat
123 * to translate the destination address to another local interface.
124 *
125 * XXX - ip_checkinterface must be disabled if you add IP aliases
126 * to the loopback interface instead of the interface where the
127 * packets for those addresses are received.
128 */
129static int ip_checkinterface = 0;
130SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
131 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
132
133struct pfil_head inet_pfil_hook; /* Packet filter hooks */
134
135static struct ifqueue ipintrq;
136static int ipqmaxlen = IFQ_MAXLEN;
137
138extern struct domain inetdomain;
139extern struct protosw inetsw[];
140u_char ip_protox[IPPROTO_MAX];
141struct in_ifaddrhead in_ifaddrhead; /* first inet address */
142struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
143u_long in_ifaddrhmask; /* mask for hash table */
144
145SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
146 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
147SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
148 &ipintrq.ifq_drops, 0,
149 "Number of packets dropped from the IP input queue");
150
151struct ipstat ipstat;
152SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
153 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
154
155/*
156 * IP datagram reassembly.
157 */
158#define IPREASS_NHASH_LOG2 6
159#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
160#define IPREASS_HMASK (IPREASS_NHASH - 1)
161#define IPREASS_HASH(x,y) \
162 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
163
164static uma_zone_t ipq_zone;
165static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
166static struct mtx ipqlock;
167
168#define IPQ_LOCK() mtx_lock(&ipqlock)
169#define IPQ_UNLOCK() mtx_unlock(&ipqlock)
170#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
171#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
172
173static void maxnipq_update(void);
174static void ipq_zone_change(void *);
175
176static int maxnipq; /* Administrative limit on # reass queues. */
177static int nipq = 0; /* Total # of reass queues */
178SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
179 &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
180
181static int maxfragsperpacket;
182SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
183 &maxfragsperpacket, 0,
184 "Maximum number of IPv4 fragments allowed per packet");
185
186struct callout ipport_tick_callout;
187
188#ifdef IPCTL_DEFMTU
189SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
190 &ip_mtu, 0, "Default MTU");
191#endif
192
193#ifdef IPSTEALTH
194int ipstealth = 0;
195SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
196 &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
197#endif
198
199/*
200 * ipfw_ether and ipfw_bridge hooks.
201 * XXX: Temporary until those are converted to pfil_hooks as well.
202 */
203ip_fw_chk_t *ip_fw_chk_ptr = NULL;
204ip_dn_io_t *ip_dn_io_ptr = NULL;
205int fw_one_pass = 1;
206
207static void ip_freef(struct ipqhead *, struct ipq *);
208
209/*
210 * IP initialization: fill in IP protocol switch table.
211 * All protocols not implemented in kernel go to raw IP protocol handler.
212 */
213void
214ip_init(void)
215{
216 struct protosw *pr;
217 int i;
218
219 TAILQ_INIT(&in_ifaddrhead);
220 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
221 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
222 if (pr == NULL)
223 panic("ip_init: PF_INET not found");
224
225 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
226 for (i = 0; i < IPPROTO_MAX; i++)
227 ip_protox[i] = pr - inetsw;
228 /*
229 * Cycle through IP protocols and put them into the appropriate place
230 * in ip_protox[].
231 */
232 for (pr = inetdomain.dom_protosw;
233 pr < inetdomain.dom_protoswNPROTOSW; pr++)
234 if (pr->pr_domain->dom_family == PF_INET &&
235 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
236 /* Be careful to only index valid IP protocols. */
237 if (pr->pr_protocol < IPPROTO_MAX)
238 ip_protox[pr->pr_protocol] = pr - inetsw;
239 }
240
241 /* Initialize packet filter hooks. */
242 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
243 inet_pfil_hook.ph_af = AF_INET;
244 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
245 printf("%s: WARNING: unable to register pfil hook, "
246 "error %d\n", __func__, i);
247
248 /* Initialize IP reassembly queue. */
249 IPQ_LOCK_INIT();
250 for (i = 0; i < IPREASS_NHASH; i++)
251 TAILQ_INIT(&ipq[i]);
252 maxnipq = nmbclusters / 32;
253 maxfragsperpacket = 16;
254 ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
255 NULL, UMA_ALIGN_PTR, 0);
256 maxnipq_update();
257
258 /* Start ipport_tick. */
259 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
260 ipport_tick(NULL);
261 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
262 SHUTDOWN_PRI_DEFAULT);
263 EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
264 NULL, EVENTHANDLER_PRI_ANY);
265
266 /* Initialize various other remaining things. */
267 ip_id = time_second & 0xffff;
268 ipintrq.ifq_maxlen = ipqmaxlen;
269 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
270 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
271}
272
273void
274ip_fini(void *xtp)
275{
276
277 callout_stop(&ipport_tick_callout);
278}
279
280/*
281 * Ip input routine. Checksum and byte swap header. If fragmented
282 * try to reassemble. Process options. Pass to next level.
283 */
284void
285ip_input(struct mbuf *m)
286{
287 struct ip *ip = NULL;
288 struct in_ifaddr *ia = NULL;
289 struct ifaddr *ifa;
290 int checkif, hlen = 0;
291 u_short sum;
292 int dchg = 0; /* dest changed after fw */
293 struct in_addr odst; /* original dst address */
294
295 M_ASSERTPKTHDR(m);
296
297 if (m->m_flags & M_FASTFWD_OURS) {
298 /*
299 * Firewall or NAT changed destination to local.
300 * We expect ip_len and ip_off to be in host byte order.
301 */
302 m->m_flags &= ~M_FASTFWD_OURS;
303 /* Set up some basics that will be used later. */
304 ip = mtod(m, struct ip *);
305 hlen = ip->ip_hl << 2;
306 goto ours;
307 }
308
309 ipstat.ips_total++;
310
311 if (m->m_pkthdr.len < sizeof(struct ip))
312 goto tooshort;
313
314 if (m->m_len < sizeof (struct ip) &&
315 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
316 ipstat.ips_toosmall++;
317 return;
318 }
319 ip = mtod(m, struct ip *);
320
321 if (ip->ip_v != IPVERSION) {
322 ipstat.ips_badvers++;
323 goto bad;
324 }
325
326 hlen = ip->ip_hl << 2;
327 if (hlen < sizeof(struct ip)) { /* minimum header length */
328 ipstat.ips_badhlen++;
329 goto bad;
330 }
331 if (hlen > m->m_len) {
332 if ((m = m_pullup(m, hlen)) == NULL) {
333 ipstat.ips_badhlen++;
334 return;
335 }
336 ip = mtod(m, struct ip *);
337 }
338
339 /* 127/8 must not appear on wire - RFC1122 */
340 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
341 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
342 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
343 ipstat.ips_badaddr++;
344 goto bad;
345 }
346 }
347
348 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
349 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
350 } else {
351 if (hlen == sizeof(struct ip)) {
352 sum = in_cksum_hdr(ip);
353 } else {
354 sum = in_cksum(m, hlen);
355 }
356 }
357 if (sum) {
358 ipstat.ips_badsum++;
359 goto bad;
360 }
361
362#ifdef ALTQ
363 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
364 /* packet is dropped by traffic conditioner */
365 return;
366#endif
367
368 /*
369 * Convert fields to host representation.
370 */
371 ip->ip_len = ntohs(ip->ip_len);
372 if (ip->ip_len < hlen) {
373 ipstat.ips_badlen++;
374 goto bad;
375 }
376 ip->ip_off = ntohs(ip->ip_off);
377
378 /*
379 * Check that the amount of data in the buffers
380 * is as at least much as the IP header would have us expect.
381 * Trim mbufs if longer than we expect.
382 * Drop packet if shorter than we expect.
383 */
384 if (m->m_pkthdr.len < ip->ip_len) {
385tooshort:
386 ipstat.ips_tooshort++;
387 goto bad;
388 }
389 if (m->m_pkthdr.len > ip->ip_len) {
390 if (m->m_len == m->m_pkthdr.len) {
391 m->m_len = ip->ip_len;
392 m->m_pkthdr.len = ip->ip_len;
393 } else
394 m_adj(m, ip->ip_len - m->m_pkthdr.len);
395 }
396#ifdef IPSEC
397 /*
398 * Bypass packet filtering for packets from a tunnel (gif).
399 */
400 if (ip_ipsec_filtertunnel(m))
401 goto passin;
402#endif /* IPSEC */
403
404 /*
405 * Run through list of hooks for input packets.
406 *
407 * NB: Beware of the destination address changing (e.g.
408 * by NAT rewriting). When this happens, tell
409 * ip_forward to do the right thing.
410 */
411
412 /* Jump over all PFIL processing if hooks are not active. */
413 if (!PFIL_HOOKED(&inet_pfil_hook))
414 goto passin;
415
416 odst = ip->ip_dst;
417 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
418 PFIL_IN, NULL) != 0)
419 return;
420 if (m == NULL) /* consumed by filter */
421 return;
422
423 ip = mtod(m, struct ip *);
424 dchg = (odst.s_addr != ip->ip_dst.s_addr);
425
426#ifdef IPFIREWALL_FORWARD
427 if (m->m_flags & M_FASTFWD_OURS) {
428 m->m_flags &= ~M_FASTFWD_OURS;
429 goto ours;
430 }
431 if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
432 /*
433 * Directly ship on the packet. This allows to forward packets
434 * that were destined for us to some other directly connected
435 * host.
436 */
437 ip_forward(m, dchg);
438 return;
439 }
440#endif /* IPFIREWALL_FORWARD */
441
442passin:
443 /*
444 * Process options and, if not destined for us,
445 * ship it on. ip_dooptions returns 1 when an
446 * error was detected (causing an icmp message
447 * to be sent and the original packet to be freed).
448 */
449 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
450 return;
451
452 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
453 * matter if it is destined to another node, or whether it is
454 * a multicast one, RSVP wants it! and prevents it from being forwarded
455 * anywhere else. Also checks if the rsvp daemon is running before
456 * grabbing the packet.
457 */
458 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
459 goto ours;
460
461 /*
462 * Check our list of addresses, to see if the packet is for us.
463 * If we don't have any addresses, assume any unicast packet
464 * we receive might be for us (and let the upper layers deal
465 * with it).
466 */
467 if (TAILQ_EMPTY(&in_ifaddrhead) &&
468 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
469 goto ours;
470
471 /*
472 * Enable a consistency check between the destination address
473 * and the arrival interface for a unicast packet (the RFC 1122
474 * strong ES model) if IP forwarding is disabled and the packet
475 * is not locally generated and the packet is not subject to
476 * 'ipfw fwd'.
477 *
478 * XXX - Checking also should be disabled if the destination
479 * address is ipnat'ed to a different interface.
480 *
481 * XXX - Checking is incompatible with IP aliases added
482 * to the loopback interface instead of the interface where
483 * the packets are received.
484 *
485 * XXX - This is the case for carp vhost IPs as well so we
486 * insert a workaround. If the packet got here, we already
487 * checked with carp_iamatch() and carp_forus().
488 */
489 checkif = ip_checkinterface && (ipforwarding == 0) &&
490 m->m_pkthdr.rcvif != NULL &&
491 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
492#ifdef DEV_CARP
493 !m->m_pkthdr.rcvif->if_carp &&
494#endif
495 (dchg == 0);
496
497 /*
498 * Check for exact addresses in the hash bucket.
499 */
500 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
501 /*
502 * If the address matches, verify that the packet
503 * arrived via the correct interface if checking is
504 * enabled.
505 */
506 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
507 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
508 goto ours;
509 }
510 /*
511 * Check for broadcast addresses.
512 *
513 * Only accept broadcast packets that arrive via the matching
514 * interface. Reception of forwarded directed broadcasts would
515 * be handled via ip_forward() and ether_output() with the loopback
516 * into the stack for SIMPLEX interfaces handled by ether_output().
517 */
518 if (m->m_pkthdr.rcvif != NULL &&
519 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
520 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
521 if (ifa->ifa_addr->sa_family != AF_INET)
522 continue;
523 ia = ifatoia(ifa);
524 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
525 ip->ip_dst.s_addr)
526 goto ours;
527 if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
528 goto ours;
529#ifdef BOOTP_COMPAT
530 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
531 goto ours;
532#endif
533 }
534 }
535 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
536 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
537 ipstat.ips_cantforward++;
538 m_freem(m);
539 return;
540 }
541 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
542 struct in_multi *inm;
543 if (ip_mrouter) {
544 /*
545 * If we are acting as a multicast router, all
546 * incoming multicast packets are passed to the
547 * kernel-level multicast forwarding function.
548 * The packet is returned (relatively) intact; if
549 * ip_mforward() returns a non-zero value, the packet
550 * must be discarded, else it may be accepted below.
551 */
552 if (ip_mforward &&
553 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
554 ipstat.ips_cantforward++;
555 m_freem(m);
556 return;
557 }
558
559 /*
560 * The process-level routing daemon needs to receive
561 * all multicast IGMP packets, whether or not this
562 * host belongs to their destination groups.
563 */
564 if (ip->ip_p == IPPROTO_IGMP)
565 goto ours;
566 ipstat.ips_forward++;
567 }
568 /*
569 * See if we belong to the destination multicast group on the
570 * arrival interface.
571 */
572 IN_MULTI_LOCK();
573 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
574 IN_MULTI_UNLOCK();
575 if (inm == NULL) {
576 ipstat.ips_notmember++;
577 m_freem(m);
578 return;
579 }
580 goto ours;
581 }
582 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
583 goto ours;
584 if (ip->ip_dst.s_addr == INADDR_ANY)
585 goto ours;
586
587 /*
588 * FAITH(Firewall Aided Internet Translator)
589 */
590 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
591 if (ip_keepfaith) {
592 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
593 goto ours;
594 }
595 m_freem(m);
596 return;
597 }
598
599 /*
600 * Not for us; forward if possible and desirable.
601 */
602 if (ipforwarding == 0) {
603 ipstat.ips_cantforward++;
604 m_freem(m);
605 } else {
606#ifdef IPSEC
607 if (ip_ipsec_fwd(m))
608 goto bad;
609#endif /* IPSEC */
610 ip_forward(m, dchg);
611 }
612 return;
613
614ours:
615#ifdef IPSTEALTH
616 /*
617 * IPSTEALTH: Process non-routing options only
618 * if the packet is destined for us.
619 */
620 if (ipstealth && hlen > sizeof (struct ip) &&
621 ip_dooptions(m, 1))
622 return;
623#endif /* IPSTEALTH */
624
625 /* Count the packet in the ip address stats */
626 if (ia != NULL) {
627 ia->ia_ifa.if_ipackets++;
628 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
629 }
630
631 /*
632 * Attempt reassembly; if it succeeds, proceed.
633 * ip_reass() will return a different mbuf.
634 */
635 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
636 m = ip_reass(m);
637 if (m == NULL)
638 return;
639 ip = mtod(m, struct ip *);
640 /* Get the header length of the reassembled packet */
641 hlen = ip->ip_hl << 2;
642 }
643
644 /*
645 * Further protocols expect the packet length to be w/o the
646 * IP header.
647 */
648 ip->ip_len -= hlen;
649
650#ifdef IPSEC
651 /*
652 * enforce IPsec policy checking if we are seeing last header.
653 * note that we do not visit this with protocols with pcb layer
654 * code - like udp/tcp/raw ip.
655 */
656 if (ip_ipsec_input(m))
657 goto bad;
658#endif /* IPSEC */
659
660 /*
661 * Switch out to protocol's input routine.
662 */
663 ipstat.ips_delivered++;
664
665 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
666 return;
667bad:
668 m_freem(m);
669}
670
671/*
672 * After maxnipq has been updated, propagate the change to UMA. The UMA zone
673 * max has slightly different semantics than the sysctl, for historical
674 * reasons.
675 */
676static void
677maxnipq_update(void)
678{
679
680 /*
681 * -1 for unlimited allocation.
682 */
683 if (maxnipq < 0)
684 uma_zone_set_max(ipq_zone, 0);
685 /*
686 * Positive number for specific bound.
687 */
688 if (maxnipq > 0)
689 uma_zone_set_max(ipq_zone, maxnipq);
690 /*
691 * Zero specifies no further fragment queue allocation -- set the
692 * bound very low, but rely on implementation elsewhere to actually
693 * prevent allocation and reclaim current queues.
694 */
695 if (maxnipq == 0)
696 uma_zone_set_max(ipq_zone, 1);
697}
698
699static void
700ipq_zone_change(void *tag)
701{
702
703 if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
704 maxnipq = nmbclusters / 32;
705 maxnipq_update();
706 }
707}
708
709static int
710sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
711{
712 int error, i;
713
714 i = maxnipq;
715 error = sysctl_handle_int(oidp, &i, 0, req);
716 if (error || !req->newptr)
717 return (error);
718
719 /*
720 * XXXRW: Might be a good idea to sanity check the argument and place
721 * an extreme upper bound.
722 */
723 if (i < -1)
724 return (EINVAL);
725 maxnipq = i;
726 maxnipq_update();
727 return (0);
728}
729
730SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
731 NULL, 0, sysctl_maxnipq, "I",
732 "Maximum number of IPv4 fragment reassembly queue entries");
733
734/*
735 * Take incoming datagram fragment and try to reassemble it into
736 * whole datagram. If the argument is the first fragment or one
737 * in between the function will return NULL and store the mbuf
738 * in the fragment chain. If the argument is the last fragment
739 * the packet will be reassembled and the pointer to the new
740 * mbuf returned for further processing. Only m_tags attached
741 * to the first packet/fragment are preserved.
742 * The IP header is *NOT* adjusted out of iplen.
743 */
744struct mbuf *
745ip_reass(struct mbuf *m)
746{
747 struct ip *ip;
748 struct mbuf *p, *q, *nq, *t;
749 struct ipq *fp = NULL;
750 struct ipqhead *head;
751 int i, hlen, next;
752 u_int8_t ecn, ecn0;
753 u_short hash;
754
755 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
756 if (maxnipq == 0 || maxfragsperpacket == 0) {
757 ipstat.ips_fragments++;
758 ipstat.ips_fragdropped++;
759 m_freem(m);
760 return (NULL);
761 }
762
763 ip = mtod(m, struct ip *);
764 hlen = ip->ip_hl << 2;
765
766 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
767 head = &ipq[hash];
768 IPQ_LOCK();
769
770 /*
771 * Look for queue of fragments
772 * of this datagram.
773 */
774 TAILQ_FOREACH(fp, head, ipq_list)
775 if (ip->ip_id == fp->ipq_id &&
776 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
777 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
778#ifdef MAC
779 mac_ipq_match(m, fp) &&
780#endif
781 ip->ip_p == fp->ipq_p)
782 goto found;
783
784 fp = NULL;
785
786 /*
787 * Attempt to trim the number of allocated fragment queues if it
788 * exceeds the administrative limit.
789 */
790 if ((nipq > maxnipq) && (maxnipq > 0)) {
791 /*
792 * drop something from the tail of the current queue
793 * before proceeding further
794 */
795 struct ipq *q = TAILQ_LAST(head, ipqhead);
796 if (q == NULL) { /* gak */
797 for (i = 0; i < IPREASS_NHASH; i++) {
798 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
799 if (r) {
800 ipstat.ips_fragtimeout += r->ipq_nfrags;
801 ip_freef(&ipq[i], r);
802 break;
803 }
804 }
805 } else {
806 ipstat.ips_fragtimeout += q->ipq_nfrags;
807 ip_freef(head, q);
808 }
809 }
810
811found:
812 /*
813 * Adjust ip_len to not reflect header,
814 * convert offset of this to bytes.
815 */
816 ip->ip_len -= hlen;
817 if (ip->ip_off & IP_MF) {
818 /*
819 * Make sure that fragments have a data length
820 * that's a non-zero multiple of 8 bytes.
821 */
822 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
823 ipstat.ips_toosmall++; /* XXX */
824 goto dropfrag;
825 }
826 m->m_flags |= M_FRAG;
827 } else
828 m->m_flags &= ~M_FRAG;
829 ip->ip_off <<= 3;
830
831
832 /*
833 * Attempt reassembly; if it succeeds, proceed.
834 * ip_reass() will return a different mbuf.
835 */
836 ipstat.ips_fragments++;
837 m->m_pkthdr.header = ip;
838
839 /* Previous ip_reass() started here. */
840 /*
841 * Presence of header sizes in mbufs
842 * would confuse code below.
843 */
844 m->m_data += hlen;
845 m->m_len -= hlen;
846
847 /*
848 * If first fragment to arrive, create a reassembly queue.
849 */
850 if (fp == NULL) {
851 fp = uma_zalloc(ipq_zone, M_NOWAIT);
852 if (fp == NULL)
853 goto dropfrag;
854#ifdef MAC
855 if (mac_ipq_init(fp, M_NOWAIT) != 0) {
856 uma_zfree(ipq_zone, fp);
857 fp = NULL;
858 goto dropfrag;
859 }
860 mac_ipq_create(m, fp);
861#endif
862 TAILQ_INSERT_HEAD(head, fp, ipq_list);
863 nipq++;
864 fp->ipq_nfrags = 1;
865 fp->ipq_ttl = IPFRAGTTL;
866 fp->ipq_p = ip->ip_p;
867 fp->ipq_id = ip->ip_id;
868 fp->ipq_src = ip->ip_src;
869 fp->ipq_dst = ip->ip_dst;
870 fp->ipq_frags = m;
871 m->m_nextpkt = NULL;
872 goto done;
873 } else {
874 fp->ipq_nfrags++;
875#ifdef MAC
876 mac_ipq_update(m, fp);
877#endif
878 }
879
880#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
881
882 /*
883 * Handle ECN by comparing this segment with the first one;
884 * if CE is set, do not lose CE.
885 * drop if CE and not-ECT are mixed for the same packet.
886 */
887 ecn = ip->ip_tos & IPTOS_ECN_MASK;
888 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
889 if (ecn == IPTOS_ECN_CE) {
890 if (ecn0 == IPTOS_ECN_NOTECT)
891 goto dropfrag;
892 if (ecn0 != IPTOS_ECN_CE)
893 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
894 }
895 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
896 goto dropfrag;
897
898 /*
899 * Find a segment which begins after this one does.
900 */
901 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
902 if (GETIP(q)->ip_off > ip->ip_off)
903 break;
904
905 /*
906 * If there is a preceding segment, it may provide some of
907 * our data already. If so, drop the data from the incoming
908 * segment. If it provides all of our data, drop us, otherwise
909 * stick new segment in the proper place.
910 *
911 * If some of the data is dropped from the the preceding
912 * segment, then it's checksum is invalidated.
913 */
914 if (p) {
915 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
916 if (i > 0) {
917 if (i >= ip->ip_len)
918 goto dropfrag;
919 m_adj(m, i);
920 m->m_pkthdr.csum_flags = 0;
921 ip->ip_off += i;
922 ip->ip_len -= i;
923 }
924 m->m_nextpkt = p->m_nextpkt;
925 p->m_nextpkt = m;
926 } else {
927 m->m_nextpkt = fp->ipq_frags;
928 fp->ipq_frags = m;
929 }
930
931 /*
932 * While we overlap succeeding segments trim them or,
933 * if they are completely covered, dequeue them.
934 */
935 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
936 q = nq) {
937 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
938 if (i < GETIP(q)->ip_len) {
939 GETIP(q)->ip_len -= i;
940 GETIP(q)->ip_off += i;
941 m_adj(q, i);
942 q->m_pkthdr.csum_flags = 0;
943 break;
944 }
945 nq = q->m_nextpkt;
946 m->m_nextpkt = nq;
947 ipstat.ips_fragdropped++;
948 fp->ipq_nfrags--;
949 m_freem(q);
950 }
951
952 /*
953 * Check for complete reassembly and perform frag per packet
954 * limiting.
955 *
956 * Frag limiting is performed here so that the nth frag has
957 * a chance to complete the packet before we drop the packet.
958 * As a result, n+1 frags are actually allowed per packet, but
959 * only n will ever be stored. (n = maxfragsperpacket.)
960 *
961 */
962 next = 0;
963 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
964 if (GETIP(q)->ip_off != next) {
965 if (fp->ipq_nfrags > maxfragsperpacket) {
966 ipstat.ips_fragdropped += fp->ipq_nfrags;
967 ip_freef(head, fp);
968 }
969 goto done;
970 }
971 next += GETIP(q)->ip_len;
972 }
973 /* Make sure the last packet didn't have the IP_MF flag */
974 if (p->m_flags & M_FRAG) {
975 if (fp->ipq_nfrags > maxfragsperpacket) {
976 ipstat.ips_fragdropped += fp->ipq_nfrags;
977 ip_freef(head, fp);
978 }
979 goto done;
980 }
981
982 /*
983 * Reassembly is complete. Make sure the packet is a sane size.
984 */
985 q = fp->ipq_frags;
986 ip = GETIP(q);
987 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
988 ipstat.ips_toolong++;
989 ipstat.ips_fragdropped += fp->ipq_nfrags;
990 ip_freef(head, fp);
991 goto done;
992 }
993
994 /*
995 * Concatenate fragments.
996 */
997 m = q;
998 t = m->m_next;
999 m->m_next = NULL;
1000 m_cat(m, t);
1001 nq = q->m_nextpkt;
1002 q->m_nextpkt = NULL;
1003 for (q = nq; q != NULL; q = nq) {
1004 nq = q->m_nextpkt;
1005 q->m_nextpkt = NULL;
1006 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1007 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1008 m_cat(m, q);
1009 }
1010 /*
1011 * In order to do checksumming faster we do 'end-around carry' here
1012 * (and not in for{} loop), though it implies we are not going to
1013 * reassemble more than 64k fragments.
1014 */
1015 m->m_pkthdr.csum_data =
1016 (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1017#ifdef MAC
1018 mac_ipq_reassemble(fp, m);
1019 mac_ipq_destroy(fp);
1020#endif
1021
1022 /*
1023 * Create header for new ip packet by modifying header of first
1024 * packet; dequeue and discard fragment reassembly header.
1025 * Make header visible.
1026 */
1027 ip->ip_len = (ip->ip_hl << 2) + next;
1028 ip->ip_src = fp->ipq_src;
1029 ip->ip_dst = fp->ipq_dst;
1030 TAILQ_REMOVE(head, fp, ipq_list);
1031 nipq--;
1032 uma_zfree(ipq_zone, fp);
1033 m->m_len += (ip->ip_hl << 2);
1034 m->m_data -= (ip->ip_hl << 2);
1035 /* some debugging cruft by sklower, below, will go away soon */
1036 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1037 m_fixhdr(m);
1038 ipstat.ips_reassembled++;
1039 IPQ_UNLOCK();
1040 return (m);
1041
1042dropfrag:
1043 ipstat.ips_fragdropped++;
1044 if (fp != NULL)
1045 fp->ipq_nfrags--;
1046 m_freem(m);
1047done:
1048 IPQ_UNLOCK();
1049 return (NULL);
1050
1051#undef GETIP
1052}
1053
1054/*
1055 * Free a fragment reassembly header and all
1056 * associated datagrams.
1057 */
1058static void
1059ip_freef(struct ipqhead *fhp, struct ipq *fp)
1060{
1061 struct mbuf *q;
1062
1063 IPQ_LOCK_ASSERT();
1064
1065 while (fp->ipq_frags) {
1066 q = fp->ipq_frags;
1067 fp->ipq_frags = q->m_nextpkt;
1068 m_freem(q);
1069 }
1070 TAILQ_REMOVE(fhp, fp, ipq_list);
1071 uma_zfree(ipq_zone, fp);
1072 nipq--;
1073}
1074
1075/*
1076 * IP timer processing;
1077 * if a timer expires on a reassembly
1078 * queue, discard it.
1079 */
1080void
1081ip_slowtimo(void)
1082{
1083 struct ipq *fp;
1084 int i;
1085
1086 IPQ_LOCK();
1087 for (i = 0; i < IPREASS_NHASH; i++) {
1088 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1089 struct ipq *fpp;
1090
1091 fpp = fp;
1092 fp = TAILQ_NEXT(fp, ipq_list);
1093 if(--fpp->ipq_ttl == 0) {
1094 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1095 ip_freef(&ipq[i], fpp);
1096 }
1097 }
1098 }
1099 /*
1100 * If we are over the maximum number of fragments
1101 * (due to the limit being lowered), drain off
1102 * enough to get down to the new limit.
1103 */
1104 if (maxnipq >= 0 && nipq > maxnipq) {
1105 for (i = 0; i < IPREASS_NHASH; i++) {
1106 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1107 ipstat.ips_fragdropped +=
1108 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1109 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1110 }
1111 }
1112 }
1113 IPQ_UNLOCK();
1114}
1115
1116/*
1117 * Drain off all datagram fragments.
1118 */
1119void
1120ip_drain(void)
1121{
1122 int i;
1123
1124 IPQ_LOCK();
1125 for (i = 0; i < IPREASS_NHASH; i++) {
1126 while(!TAILQ_EMPTY(&ipq[i])) {
1127 ipstat.ips_fragdropped +=
1128 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1129 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1130 }
1131 }
1132 IPQ_UNLOCK();
1133 in_rtqdrain();
1134}
1135
1136/*
1137 * The protocol to be inserted into ip_protox[] must be already registered
1138 * in inetsw[], either statically or through pf_proto_register().
1139 */
1140int
1141ipproto_register(u_char ipproto)
1142{
1143 struct protosw *pr;
1144
1145 /* Sanity checks. */
1146 if (ipproto == 0)
1147 return (EPROTONOSUPPORT);
1148
1149 /*
1150 * The protocol slot must not be occupied by another protocol
1151 * already. An index pointing to IPPROTO_RAW is unused.
1152 */
1153 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1154 if (pr == NULL)
1155 return (EPFNOSUPPORT);
1156 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
1157 return (EEXIST);
1158
1159 /* Find the protocol position in inetsw[] and set the index. */
1160 for (pr = inetdomain.dom_protosw;
1161 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1162 if (pr->pr_domain->dom_family == PF_INET &&
1163 pr->pr_protocol && pr->pr_protocol == ipproto) {
1164 /* Be careful to only index valid IP protocols. */
1165 if (pr->pr_protocol < IPPROTO_MAX) {
1166 ip_protox[pr->pr_protocol] = pr - inetsw;
1167 return (0);
1168 } else
1169 return (EINVAL);
1170 }
1171 }
1172 return (EPROTONOSUPPORT);
1173}
1174
1175int
1176ipproto_unregister(u_char ipproto)
1177{
1178 struct protosw *pr;
1179
1180 /* Sanity checks. */
1181 if (ipproto == 0)
1182 return (EPROTONOSUPPORT);
1183
1184 /* Check if the protocol was indeed registered. */
1185 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1186 if (pr == NULL)
1187 return (EPFNOSUPPORT);
1188 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
1189 return (ENOENT);
1190
1191 /* Reset the protocol slot to IPPROTO_RAW. */
1192 ip_protox[ipproto] = pr - inetsw;
1193 return (0);
1194}
1195
1196/*
1197 * Given address of next destination (final or next hop),
1198 * return internet address info of interface to be used to get there.
1199 */
1200struct in_ifaddr *
1201ip_rtaddr(struct in_addr dst, u_int fibnum)
1202{
1203 struct route sro;
1204 struct sockaddr_in *sin;
1205 struct in_ifaddr *ifa;
1206
1207 bzero(&sro, sizeof(sro));
1208 sin = (struct sockaddr_in *)&sro.ro_dst;
1209 sin->sin_family = AF_INET;
1210 sin->sin_len = sizeof(*sin);
1211 sin->sin_addr = dst;
1212 in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
1213
1214 if (sro.ro_rt == NULL)
1215 return (NULL);
1216
1217 ifa = ifatoia(sro.ro_rt->rt_ifa);
1218 RTFREE(sro.ro_rt);
1219 return (ifa);
1220}
1221
1222u_char inetctlerrmap[PRC_NCMDS] = {
1223 0, 0, 0, 0,
1224 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1225 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1226 EMSGSIZE, EHOSTUNREACH, 0, 0,
1227 0, 0, EHOSTUNREACH, 0,
1228 ENOPROTOOPT, ECONNREFUSED
1229};
1230
1231/*
1232 * Forward a packet. If some error occurs return the sender
1233 * an icmp packet. Note we can't always generate a meaningful
1234 * icmp message because icmp doesn't have a large enough repertoire
1235 * of codes and types.
1236 *
1237 * If not forwarding, just drop the packet. This could be confusing
1238 * if ipforwarding was zero but some routing protocol was advancing
1239 * us as a gateway to somewhere. However, we must let the routing
1240 * protocol deal with that.
1241 *
1242 * The srcrt parameter indicates whether the packet is being forwarded
1243 * via a source route.
1244 */
1245void
1246ip_forward(struct mbuf *m, int srcrt)
1247{
1248 struct ip *ip = mtod(m, struct ip *);
1249 struct in_ifaddr *ia = NULL;
1250 struct mbuf *mcopy;
1251 struct in_addr dest;
1252 struct route ro;
1253 int error, type = 0, code = 0, mtu = 0;
1254
1255 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1256 ipstat.ips_cantforward++;
1257 m_freem(m);
1258 return;
1259 }
1260#ifdef IPSTEALTH
1261 if (!ipstealth) {
1262#endif
1263 if (ip->ip_ttl <= IPTTLDEC) {
1264 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1265 0, 0);
1266 return;
1267 }
1268#ifdef IPSTEALTH
1269 }
1270#endif
1271
1272 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1273 if (!srcrt && ia == NULL) {
1274 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1275 return;
1276 }
1277
1278 /*
1279 * Save the IP header and at most 8 bytes of the payload,
1280 * in case we need to generate an ICMP message to the src.
1281 *
1282 * XXX this can be optimized a lot by saving the data in a local
1283 * buffer on the stack (72 bytes at most), and only allocating the
1284 * mbuf if really necessary. The vast majority of the packets
1285 * are forwarded without having to send an ICMP back (either
1286 * because unnecessary, or because rate limited), so we are
1287 * really we are wasting a lot of work here.
1288 *
1289 * We don't use m_copy() because it might return a reference
1290 * to a shared cluster. Both this function and ip_output()
1291 * assume exclusive access to the IP header in `m', so any
1292 * data in a cluster may change before we reach icmp_error().
1293 */
1294 MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1295 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1296 /*
1297 * It's probably ok if the pkthdr dup fails (because
1298 * the deep copy of the tag chain failed), but for now
1299 * be conservative and just discard the copy since
1300 * code below may some day want the tags.
1301 */
1302 m_free(mcopy);
1303 mcopy = NULL;
1304 }
1305 if (mcopy != NULL) {
1306 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1307 mcopy->m_pkthdr.len = mcopy->m_len;
1308 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1309 }
1310
1311#ifdef IPSTEALTH
1312 if (!ipstealth) {
1313#endif
1314 ip->ip_ttl -= IPTTLDEC;
1315#ifdef IPSTEALTH
1316 }
1317#endif
1318
1319 /*
1320 * If forwarding packet using same interface that it came in on,
1321 * perhaps should send a redirect to sender to shortcut a hop.
1322 * Only send redirect if source is sending directly to us,
1323 * and if packet was not source routed (or has any options).
1324 * Also, don't send redirect if forwarding using a default route
1325 * or a route modified by a redirect.
1326 */
1327 dest.s_addr = 0;
1328 if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1329 struct sockaddr_in *sin;
1330 struct rtentry *rt;
1331
1332 bzero(&ro, sizeof(ro));
1333 sin = (struct sockaddr_in *)&ro.ro_dst;
1334 sin->sin_family = AF_INET;
1335 sin->sin_len = sizeof(*sin);
1336 sin->sin_addr = ip->ip_dst;
1337 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
1338
1339 rt = ro.ro_rt;
1340
1341 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1342 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1343#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1344 u_long src = ntohl(ip->ip_src.s_addr);
1345
1346 if (RTA(rt) &&
1347 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1348 if (rt->rt_flags & RTF_GATEWAY)
1349 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1350 else
1351 dest.s_addr = ip->ip_dst.s_addr;
1352 /* Router requirements says to only send host redirects */
1353 type = ICMP_REDIRECT;
1354 code = ICMP_REDIRECT_HOST;
1355 }
1356 }
1357 if (rt)
1358 RTFREE(rt);
1359 }
1360
1361 /*
1362 * Try to cache the route MTU from ip_output so we can consider it for
1363 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1364 */
1365 bzero(&ro, sizeof(ro));
1366
1367 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1368
1369 if (error == EMSGSIZE && ro.ro_rt)
1370 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1371 if (ro.ro_rt)
1372 RTFREE(ro.ro_rt);
1373
1374 if (error)
1375 ipstat.ips_cantforward++;
1376 else {
1377 ipstat.ips_forward++;
1378 if (type)
1379 ipstat.ips_redirectsent++;
1380 else {
1381 if (mcopy)
1382 m_freem(mcopy);
1383 return;
1384 }
1385 }
1386 if (mcopy == NULL)
1387 return;
1388
1389 switch (error) {
1390
1391 case 0: /* forwarded, but need redirect */
1392 /* type, code set above */
1393 break;
1394
1395 case ENETUNREACH: /* shouldn't happen, checked above */
1396 case EHOSTUNREACH:
1397 case ENETDOWN:
1398 case EHOSTDOWN:
1399 default:
1400 type = ICMP_UNREACH;
1401 code = ICMP_UNREACH_HOST;
1402 break;
1403
1404 case EMSGSIZE:
1405 type = ICMP_UNREACH;
1406 code = ICMP_UNREACH_NEEDFRAG;
1407
1408#ifdef IPSEC
1409 /*
1410 * If IPsec is configured for this path,
1411 * override any possibly mtu value set by ip_output.
1412 */
1413 mtu = ip_ipsec_mtu(m, mtu);
1414#endif /* IPSEC */
1415 /*
1416 * If the MTU was set before make sure we are below the
1417 * interface MTU.
1418 * If the MTU wasn't set before use the interface mtu or
1419 * fall back to the next smaller mtu step compared to the
1420 * current packet size.
1421 */
1422 if (mtu != 0) {
1423 if (ia != NULL)
1424 mtu = min(mtu, ia->ia_ifp->if_mtu);
1425 } else {
1426 if (ia != NULL)
1427 mtu = ia->ia_ifp->if_mtu;
1428 else
1429 mtu = ip_next_mtu(ip->ip_len, 0);
1430 }
1431 ipstat.ips_cantfrag++;
1432 break;
1433
1434 case ENOBUFS:
1435 /*
1436 * A router should not generate ICMP_SOURCEQUENCH as
1437 * required in RFC1812 Requirements for IP Version 4 Routers.
1438 * Source quench could be a big problem under DoS attacks,
1439 * or if the underlying interface is rate-limited.
1440 * Those who need source quench packets may re-enable them
1441 * via the net.inet.ip.sendsourcequench sysctl.
1442 */
1443 if (ip_sendsourcequench == 0) {
1444 m_freem(mcopy);
1445 return;
1446 } else {
1447 type = ICMP_SOURCEQUENCH;
1448 code = 0;
1449 }
1450 break;
1451
1452 case EACCES: /* ipfw denied packet */
1453 m_freem(mcopy);
1454 return;
1455 }
1456 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1457}
1458
1459void
1460ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1461 struct mbuf *m)
1462{
1463 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1464 struct bintime bt;
1465
1466 bintime(&bt);
1467 if (inp->inp_socket->so_options & SO_BINTIME) {
1468 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1469 SCM_BINTIME, SOL_SOCKET);
1470 if (*mp)
1471 mp = &(*mp)->m_next;
1472 }
1473 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1474 struct timeval tv;
1475
1476 bintime2timeval(&bt, &tv);
1477 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1478 SCM_TIMESTAMP, SOL_SOCKET);
1479 if (*mp)
1480 mp = &(*mp)->m_next;
1481 }
1482 }
1483 if (inp->inp_flags & INP_RECVDSTADDR) {
1484 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1485 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1486 if (*mp)
1487 mp = &(*mp)->m_next;
1488 }
1489 if (inp->inp_flags & INP_RECVTTL) {
1490 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1491 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1492 if (*mp)
1493 mp = &(*mp)->m_next;
1494 }
1495#ifdef notyet
1496 /* XXX
1497 * Moving these out of udp_input() made them even more broken
1498 * than they already were.
1499 */
1500 /* options were tossed already */
1501 if (inp->inp_flags & INP_RECVOPTS) {
1502 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1503 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1504 if (*mp)
1505 mp = &(*mp)->m_next;
1506 }
1507 /* ip_srcroute doesn't do what we want here, need to fix */
1508 if (inp->inp_flags & INP_RECVRETOPTS) {
1509 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1510 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1511 if (*mp)
1512 mp = &(*mp)->m_next;
1513 }
1514#endif
1515 if (inp->inp_flags & INP_RECVIF) {
1516 struct ifnet *ifp;
1517 struct sdlbuf {
1518 struct sockaddr_dl sdl;
1519 u_char pad[32];
1520 } sdlbuf;
1521 struct sockaddr_dl *sdp;
1522 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1523
1524 if (((ifp = m->m_pkthdr.rcvif))
1525 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1526 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1527 /*
1528 * Change our mind and don't try copy.
1529 */
1530 if ((sdp->sdl_family != AF_LINK)
1531 || (sdp->sdl_len > sizeof(sdlbuf))) {
1532 goto makedummy;
1533 }
1534 bcopy(sdp, sdl2, sdp->sdl_len);
1535 } else {
1536makedummy:
1537 sdl2->sdl_len
1538 = offsetof(struct sockaddr_dl, sdl_data[0]);
1539 sdl2->sdl_family = AF_LINK;
1540 sdl2->sdl_index = 0;
1541 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1542 }
1543 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1544 IP_RECVIF, IPPROTO_IP);
1545 if (*mp)
1546 mp = &(*mp)->m_next;
1547 }
1548}
1549
1550/*
1551 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1552 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1553 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1554 * compiled.
1555 */
1556static int ip_rsvp_on;
1557struct socket *ip_rsvpd;
1558int
1559ip_rsvp_init(struct socket *so)
1560{
1561 if (so->so_type != SOCK_RAW ||
1562 so->so_proto->pr_protocol != IPPROTO_RSVP)
1563 return EOPNOTSUPP;
1564
1565 if (ip_rsvpd != NULL)
1566 return EADDRINUSE;
1567
1568 ip_rsvpd = so;
1569 /*
1570 * This may seem silly, but we need to be sure we don't over-increment
1571 * the RSVP counter, in case something slips up.
1572 */
1573 if (!ip_rsvp_on) {
1574 ip_rsvp_on = 1;
1575 rsvp_on++;
1576 }
1577
1578 return 0;
1579}
1580
1581int
1582ip_rsvp_done(void)
1583{
1584 ip_rsvpd = NULL;
1585 /*
1586 * This may seem silly, but we need to be sure we don't over-decrement
1587 * the RSVP counter, in case something slips up.
1588 */
1589 if (ip_rsvp_on) {
1590 ip_rsvp_on = 0;
1591 rsvp_on--;
1592 }
1593 return 0;
1594}
1595
1596void
1597rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
1598{
1599 if (rsvp_input_p) { /* call the real one if loaded */
1600 rsvp_input_p(m, off);
1601 return;
1602 }
1603
1604 /* Can still get packets with rsvp_on = 0 if there is a local member
1605 * of the group to which the RSVP packet is addressed. But in this
1606 * case we want to throw the packet away.
1607 */
1608
1609 if (!rsvp_on) {
1610 m_freem(m);
1611 return;
1612 }
1613
1614 if (ip_rsvpd != NULL) {
1615 rip_input(m, off);
1616 return;
1617 }
1618 /* Drop the packet */
1619 m_freem(m);
1620}