1 /*
2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
4 *
5 * This code is derived from software contributed to The DragonFly Project
6 * by Jeffrey M. Hsu.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34 /*
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 *
62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
64 */
65
66 #define _IP_VHL
67
68 #include "opt_bootp.h"
69 #include "opt_ipdn.h"
70 #include "opt_ipdivert.h"
71 #include "opt_ipstealth.h"
72 #include "opt_ipsec.h"
73 #include "opt_rss.h"
74
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/mbuf.h>
78 #include <sys/malloc.h>
79 #include <sys/mpipe.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/time.h>
84 #include <sys/globaldata.h>
85 #include <sys/thread.h>
86 #include <sys/kernel.h>
87 #include <sys/syslog.h>
88 #include <sys/sysctl.h>
89 #include <sys/in_cksum.h>
90 #include <sys/lock.h>
91
92 #include <sys/mplock2.h>
93
94 #include <machine/stdarg.h>
95
96 #include <net/if.h>
97 #include <net/if_types.h>
98 #include <net/if_var.h>
99 #include <net/if_dl.h>
100 #include <net/pfil.h>
101 #include <net/route.h>
102 #include <net/netisr2.h>
103
104 #include <netinet/in.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_icmp.h>
111 #include <netinet/ip_divert.h>
112 #include <netinet/ip_flow.h>
113
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
116 #include <net/netmsg2.h>
117
118 #include <sys/socketvar.h>
119
120 #include <net/ipfw/ip_fw.h>
121 #include <net/dummynet/ip_dummynet.h>
122
123 #ifdef IPSEC
124 #include <netinet6/ipsec.h>
125 #include <netproto/key/key.h>
126 #endif
127
128 #ifdef FAST_IPSEC
129 #include <netproto/ipsec/ipsec.h>
130 #include <netproto/ipsec/key.h>
131 #endif
132
133 int rsvp_on = 0;
134 static int ip_rsvp_on;
135 struct socket *ip_rsvpd;
136
137 int ipforwarding = 0;
138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
139 &ipforwarding, 0, "Enable IP forwarding between interfaces");
140
141 static int ipsendredirects = 1; /* XXX */
142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
143 &ipsendredirects, 0, "Enable sending IP redirects");
144
145 int ip_defttl = IPDEFTTL;
146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
147 &ip_defttl, 0, "Maximum TTL on IP packets");
148
149 static int ip_dosourceroute = 0;
150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
152
153 static int ip_acceptsourceroute = 0;
154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
155 CTLFLAG_RW, &ip_acceptsourceroute, 0,
156 "Enable accepting source routed IP packets");
157
158 static int ip_keepfaith = 0;
159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
160 &ip_keepfaith, 0,
161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
162
163 static int nipq = 0; /* total # of reass queues */
164 static int maxnipq;
165 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
166 &maxnipq, 0,
167 "Maximum number of IPv4 fragment reassembly queue entries");
168
169 static int maxfragsperpacket;
170 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
171 &maxfragsperpacket, 0,
172 "Maximum number of IPv4 fragments allowed per packet");
173
174 static int ip_sendsourcequench = 0;
175 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
176 &ip_sendsourcequench, 0,
177 "Enable the transmission of source quench packets");
178
179 int ip_do_randomid = 1;
180 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
181 &ip_do_randomid, 0,
182 "Assign random ip_id values");
183 /*
184 * XXX - Setting ip_checkinterface mostly implements the receive side of
185 * the Strong ES model described in RFC 1122, but since the routing table
186 * and transmit implementation do not implement the Strong ES model,
187 * setting this to 1 results in an odd hybrid.
188 *
189 * XXX - ip_checkinterface currently must be disabled if you use ipnat
190 * to translate the destination address to another local interface.
191 *
192 * XXX - ip_checkinterface must be disabled if you add IP aliases
193 * to the loopback interface instead of the interface where the
194 * packets for those addresses are received.
195 */
196 static int ip_checkinterface = 0;
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
198 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
199
200 static u_long ip_hash_count = 0;
201 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
202 &ip_hash_count, 0, "Number of packets hashed by IP");
203
204 #ifdef RSS_DEBUG
205 static u_long ip_rehash_count = 0;
206 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
207 &ip_rehash_count, 0, "Number of packets rehashed by IP");
208
209 static u_long ip_dispatch_fast = 0;
210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
211 &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
212
213 static u_long ip_dispatch_slow = 0;
214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
215 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
216 #endif
217
218 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token);
219
220 #ifdef DIAGNOSTIC
221 static int ipprintfs = 0;
222 #endif
223
224 extern struct domain inetdomain;
225 extern struct protosw inetsw[];
226 u_char ip_protox[IPPROTO_MAX];
227 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
228 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
229 /* inet addr hash table */
230 u_long in_ifaddrhmask; /* mask for hash table */
231
232 static struct mbuf *ipforward_mtemp[MAXCPU];
233
234 struct ip_stats ipstats_percpu[MAXCPU] __cachealign;
235
236 static int
237 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
238 {
239 int cpu, error = 0;
240
241 for (cpu = 0; cpu < ncpus; ++cpu) {
242 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
243 sizeof(struct ip_stats))))
244 break;
245 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
246 sizeof(struct ip_stats))))
247 break;
248 }
249
250 return (error);
251 }
252 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
253 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
254
255 /* Packet reassembly stuff */
256 #define IPREASS_NHASH_LOG2 6
257 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
258 #define IPREASS_HMASK (IPREASS_NHASH - 1)
259 #define IPREASS_HASH(x,y) \
260 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
261
262 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
263
264 #ifdef IPCTL_DEFMTU
265 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
266 &ip_mtu, 0, "Default MTU");
267 #endif
268
269 #ifdef IPSTEALTH
270 static int ipstealth = 0;
271 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
272 #else
273 static const int ipstealth = 0;
274 #endif
275
276 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
277
278 struct pfil_head inet_pfil_hook;
279
280 /*
281 * struct ip_srcrt_opt is used to store packet state while it travels
282 * through the stack.
283 *
284 * XXX Note that the code even makes assumptions on the size and
285 * alignment of fields inside struct ip_srcrt so e.g. adding some
286 * fields will break the code. This needs to be fixed.
287 *
288 * We need to save the IP options in case a protocol wants to respond
289 * to an incoming packet over the same route if the packet got here
290 * using IP source routing. This allows connection establishment and
291 * maintenance when the remote end is on a network that is not known
292 * to us.
293 */
294 struct ip_srcrt {
295 struct in_addr dst; /* final destination */
296 char nop; /* one NOP to align */
297 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
298 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
299 };
300
301 struct ip_srcrt_opt {
302 int ip_nhops;
303 struct ip_srcrt ip_srcrt;
304 };
305
306 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
307 static struct malloc_pipe ipq_mpipe;
308
309 static void save_rte(struct mbuf *, u_char *, struct in_addr);
310 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
311 static void ip_freef(struct ipqhead *, struct ipq *);
312 static void ip_input_handler(netmsg_t);
313
314 /*
315 * IP initialization: fill in IP protocol switch table.
316 * All protocols not implemented in kernel go to raw IP protocol handler.
317 */
318 void
319 ip_init(void)
320 {
321 struct protosw *pr;
322 int i;
323 int cpu;
324
325 /*
326 * Make sure we can handle a reasonable number of fragments but
327 * cap it at 4000 (XXX).
328 */
329 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
330 IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL);
331 for (i = 0; i < ncpus; ++i) {
332 TAILQ_INIT(&in_ifaddrheads[i]);
333 in_ifaddrhashtbls[i] =
334 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
335 }
336 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
337 if (pr == NULL)
338 panic("ip_init");
339 for (i = 0; i < IPPROTO_MAX; i++)
340 ip_protox[i] = pr - inetsw;
341 for (pr = inetdomain.dom_protosw;
342 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
343 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
344 if (pr->pr_protocol != IPPROTO_RAW)
345 ip_protox[pr->pr_protocol] = pr - inetsw;
346 }
347 }
348
349 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
350 inet_pfil_hook.ph_af = AF_INET;
351 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
352 kprintf("%s: WARNING: unable to register pfil hook, "
353 "error %d\n", __func__, i);
354 }
355
356 for (i = 0; i < IPREASS_NHASH; i++)
357 TAILQ_INIT(&ipq[i]);
358
359 maxnipq = nmbclusters / 32;
360 maxfragsperpacket = 16;
361
362 ip_id = time_second & 0xffff; /* time_second survives reboots */
363
364 for (cpu = 0; cpu < ncpus; ++cpu) {
365 /*
366 * Initialize IP statistics counters for each CPU.
367 */
368 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
369
370 /*
371 * Preallocate mbuf template for forwarding
372 */
373 MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA);
374 }
375
376 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in);
377 netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
378 }
379
380 /* Do transport protocol processing. */
381 static void
382 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
383 {
384 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
385
386 /*
387 * Switch out to protocol's input routine.
388 */
389 PR_GET_MPLOCK(pr);
390 pr->pr_input(&m, &hlen, ip->ip_p);
391 PR_REL_MPLOCK(pr);
392 }
393
394 static void
395 transport_processing_handler(netmsg_t msg)
396 {
397 struct netmsg_packet *pmsg = &msg->packet;
398 struct ip *ip;
399 int hlen;
400
401 ip = mtod(pmsg->nm_packet, struct ip *);
402 hlen = pmsg->base.lmsg.u.ms_result;
403
404 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
405 /* msg was embedded in the mbuf, do not reply! */
406 }
407
408 static void
409 ip_input_handler(netmsg_t msg)
410 {
411 ip_input(msg->packet.nm_packet);
412 /* msg was embedded in the mbuf, do not reply! */
413 }
414
415 /*
416 * IP input routine. Checksum and byte swap header. If fragmented
417 * try to reassemble. Process options. Pass to next level.
418 */
419 void
420 ip_input(struct mbuf *m)
421 {
422 struct ip *ip;
423 struct in_ifaddr *ia = NULL;
424 struct in_ifaddr_container *iac;
425 int hlen, checkif;
426 u_short sum;
427 struct in_addr pkt_dst;
428 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
429 struct in_addr odst; /* original dst address(NAT) */
430 struct m_tag *mtag;
431 struct sockaddr_in *next_hop = NULL;
432 lwkt_port_t port;
433 #ifdef FAST_IPSEC
434 struct tdb_ident *tdbi;
435 struct secpolicy *sp;
436 int error;
437 #endif
438
439 M_ASSERTPKTHDR(m);
440
441 /*
442 * This routine is called from numerous places which may not have
443 * characterized the packet.
444 */
445 if ((m->m_flags & M_HASH) == 0) {
446 atomic_add_long(&ip_hash_count, 1);
447 ip_hashfn(&m, 0, IP_MPORT_IN);
448 if (m == NULL)
449 return;
450 KKASSERT(m->m_flags & M_HASH);
451 }
452 ip = mtod(m, struct ip *);
453
454 /*
455 * Pull out certain tags
456 */
457 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
458 /* Next hop */
459 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
460 KKASSERT(mtag != NULL);
461 next_hop = m_tag_data(mtag);
462 }
463
464 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
465 /* dummynet already filtered us */
466 ip = mtod(m, struct ip *);
467 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
468 goto iphack;
469 }
470
471 ipstat.ips_total++;
472
473 /* length checks already done in ip_hashfn() */
474 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
475
476 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
477 ipstat.ips_badvers++;
478 goto bad;
479 }
480
481 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
482 /* length checks already done in ip_hashfn() */
483 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
484 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
485
486 /* 127/8 must not appear on wire - RFC1122 */
487 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
488 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
489 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
490 ipstat.ips_badaddr++;
491 goto bad;
492 }
493 }
494
495 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
496 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
497 } else {
498 if (hlen == sizeof(struct ip))
499 sum = in_cksum_hdr(ip);
500 else
501 sum = in_cksum(m, hlen);
502 }
503 if (sum != 0) {
504 ipstat.ips_badsum++;
505 goto bad;
506 }
507
508 #ifdef ALTQ
509 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
510 /* packet is dropped by traffic conditioner */
511 return;
512 }
513 #endif
514 /*
515 * Convert fields to host representation.
516 */
517 ip->ip_len = ntohs(ip->ip_len);
518 ip->ip_off = ntohs(ip->ip_off);
519
520 /* length checks already done in ip_hashfn() */
521 KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
522 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
523
524 /*
525 * Trim mbufs if longer than the IP header would have us expect.
526 */
527 if (m->m_pkthdr.len > ip->ip_len) {
528 if (m->m_len == m->m_pkthdr.len) {
529 m->m_len = ip->ip_len;
530 m->m_pkthdr.len = ip->ip_len;
531 } else {
532 m_adj(m, ip->ip_len - m->m_pkthdr.len);
533 }
534 }
535 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
536 /*
537 * Bypass packet filtering for packets from a tunnel (gif).
538 */
539 if (ipsec_gethist(m, NULL))
540 goto pass;
541 #endif
542
543 /*
544 * IpHack's section.
545 * Right now when no processing on packet has done
546 * and it is still fresh out of network we do our black
547 * deals with it.
548 * - Firewall: deny/allow/divert
549 * - Xlate: translate packet's addr/port (NAT).
550 * - Pipe: pass pkt through dummynet.
551 * - Wrap: fake packet's addr/port <unimpl.>
552 * - Encapsulate: put it in another IP and send out. <unimp.>
553 */
554
555 iphack:
556 /*
557 * If we've been forwarded from the output side, then
558 * skip the firewall a second time
559 */
560 if (next_hop != NULL)
561 goto ours;
562
563 /* No pfil hooks */
564 if (!pfil_has_hooks(&inet_pfil_hook)) {
565 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
566 /*
567 * Strip dummynet tags from stranded packets
568 */
569 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
570 KKASSERT(mtag != NULL);
571 m_tag_delete(m, mtag);
572 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
573 }
574 goto pass;
575 }
576
577 /*
578 * Run through list of hooks for input packets.
579 *
580 * NOTE! If the packet is rewritten pf/ipfw/whoever must
581 * clear M_HASH.
582 */
583 odst = ip->ip_dst;
584 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
585 return;
586 if (m == NULL) /* consumed by filter */
587 return;
588 ip = mtod(m, struct ip *);
589 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
590 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
591
592 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
593 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
594 KKASSERT(mtag != NULL);
595 next_hop = m_tag_data(mtag);
596 }
597 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
598 ip_dn_queue(m);
599 return;
600 }
601 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
602 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
603 }
604 pass:
605 /*
606 * Process options and, if not destined for us,
607 * ship it on. ip_dooptions returns 1 when an
608 * error was detected (causing an icmp message
609 * to be sent and the original packet to be freed).
610 */
611 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
612 return;
613
614 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
615 * matter if it is destined to another node, or whether it is
616 * a multicast one, RSVP wants it! and prevents it from being forwarded
617 * anywhere else. Also checks if the rsvp daemon is running before
618 * grabbing the packet.
619 */
620 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
621 goto ours;
622
623 /*
624 * Check our list of addresses, to see if the packet is for us.
625 * If we don't have any addresses, assume any unicast packet
626 * we receive might be for us (and let the upper layers deal
627 * with it).
628 */
629 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
630 !(m->m_flags & (M_MCAST | M_BCAST)))
631 goto ours;
632
633 /*
634 * Cache the destination address of the packet; this may be
635 * changed by use of 'ipfw fwd'.
636 */
637 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
638
639 /*
640 * Enable a consistency check between the destination address
641 * and the arrival interface for a unicast packet (the RFC 1122
642 * strong ES model) if IP forwarding is disabled and the packet
643 * is not locally generated and the packet is not subject to
644 * 'ipfw fwd'.
645 *
646 * XXX - Checking also should be disabled if the destination
647 * address is ipnat'ed to a different interface.
648 *
649 * XXX - Checking is incompatible with IP aliases added
650 * to the loopback interface instead of the interface where
651 * the packets are received.
652 */
653 checkif = ip_checkinterface &&
654 !ipforwarding &&
655 m->m_pkthdr.rcvif != NULL &&
656 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
657 next_hop == NULL;
658
659 /*
660 * Check for exact addresses in the hash bucket.
661 */
662 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
663 ia = iac->ia;
664
665 /*
666 * If the address matches, verify that the packet
667 * arrived via the correct interface if checking is
668 * enabled.
669 */
670 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
671 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
672 goto ours;
673 }
674 ia = NULL;
675
676 /*
677 * Check for broadcast addresses.
678 *
679 * Only accept broadcast packets that arrive via the matching
680 * interface. Reception of forwarded directed broadcasts would
681 * be handled via ip_forward() and ether_output() with the loopback
682 * into the stack for SIMPLEX interfaces handled by ether_output().
683 */
684 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
685 struct ifaddr_container *ifac;
686
687 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
688 ifa_link) {
689 struct ifaddr *ifa = ifac->ifa;
690
691 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
692 continue;
693 if (ifa->ifa_addr->sa_family != AF_INET)
694 continue;
695 ia = ifatoia(ifa);
696 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
697 pkt_dst.s_addr)
698 goto ours;
699 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
700 goto ours;
701 #ifdef BOOTP_COMPAT
702 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
703 goto ours;
704 #endif
705 }
706 }
707 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
708 struct in_multi *inm;
709
710 /* XXX Multicast is not MPSAFE yet */
711 get_mplock();
712
713 if (ip_mrouter != NULL) {
714 /*
715 * If we are acting as a multicast router, all
716 * incoming multicast packets are passed to the
717 * kernel-level multicast forwarding function.
718 * The packet is returned (relatively) intact; if
719 * ip_mforward() returns a non-zero value, the packet
720 * must be discarded, else it may be accepted below.
721 */
722 if (ip_mforward != NULL &&
723 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
724 rel_mplock();
725 ipstat.ips_cantforward++;
726 m_freem(m);
727 return;
728 }
729
730 /*
731 * The process-level routing daemon needs to receive
732 * all multicast IGMP packets, whether or not this
733 * host belongs to their destination groups.
734 */
735 if (ip->ip_p == IPPROTO_IGMP) {
736 rel_mplock();
737 goto ours;
738 }
739 ipstat.ips_forward++;
740 }
741 /*
742 * See if we belong to the destination multicast group on the
743 * arrival interface.
744 */
745 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
746 if (inm == NULL) {
747 rel_mplock();
748 ipstat.ips_notmember++;
749 m_freem(m);
750 return;
751 }
752
753 rel_mplock();
754 goto ours;
755 }
756 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
757 goto ours;
758 if (ip->ip_dst.s_addr == INADDR_ANY)
759 goto ours;
760
761 /*
762 * FAITH(Firewall Aided Internet Translator)
763 */
764 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
765 if (ip_keepfaith) {
766 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
767 goto ours;
768 }
769 m_freem(m);
770 return;
771 }
772
773 /*
774 * Not for us; forward if possible and desirable.
775 */
776 if (!ipforwarding) {
777 ipstat.ips_cantforward++;
778 m_freem(m);
779 } else {
780 #ifdef IPSEC
781 /*
782 * Enforce inbound IPsec SPD.
783 */
784 if (ipsec4_in_reject(m, NULL)) {
785 ipsecstat.in_polvio++;
786 goto bad;
787 }
788 #endif
789 #ifdef FAST_IPSEC
790 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
791 crit_enter();
792 if (mtag != NULL) {
793 tdbi = (struct tdb_ident *)m_tag_data(mtag);
794 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
795 } else {
796 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
797 IP_FORWARDING, &error);
798 }
799 if (sp == NULL) { /* NB: can happen if error */
800 crit_exit();
801 /*XXX error stat???*/
802 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
803 goto bad;
804 }
805
806 /*
807 * Check security policy against packet attributes.
808 */
809 error = ipsec_in_reject(sp, m);
810 KEY_FREESP(&sp);
811 crit_exit();
812 if (error) {
813 ipstat.ips_cantforward++;
814 goto bad;
815 }
816 #endif
817 ip_forward(m, using_srcrt, next_hop);
818 }
819 return;
820
821 ours:
822
823 /*
824 * IPSTEALTH: Process non-routing options only
825 * if the packet is destined for us.
826 */
827 if (ipstealth &&
828 hlen > sizeof(struct ip) &&
829 ip_dooptions(m, 1, next_hop))
830 return;
831
832 /* Count the packet in the ip address stats */
833 if (ia != NULL) {
834 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1);
835 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len);
836 }
837
838 /*
839 * If offset or IP_MF are set, must reassemble.
840 * Otherwise, nothing need be done.
841 * (We could look in the reassembly queue to see
842 * if the packet was previously fragmented,
843 * but it's not worth the time; just let them time out.)
844 */
845 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
846 /*
847 * Attempt reassembly; if it succeeds, proceed. ip_reass()
848 * will return a different mbuf.
849 *
850 * NOTE: ip_reass() returns m with M_HASH cleared to force
851 * us to recharacterize the packet.
852 */
853 m = ip_reass(m);
854 if (m == NULL)
855 return;
856 ip = mtod(m, struct ip *);
857
858 /* Get the header length of the reassembled packet */
859 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
860 } else {
861 ip->ip_len -= hlen;
862 }
863
864 #ifdef IPSEC
865 /*
866 * enforce IPsec policy checking if we are seeing last header.
867 * note that we do not visit this with protocols with pcb layer
868 * code - like udp/tcp/raw ip.
869 */
870 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
871 ipsec4_in_reject(m, NULL)) {
872 ipsecstat.in_polvio++;
873 goto bad;
874 }
875 #endif
876 #if FAST_IPSEC
877 /*
878 * enforce IPsec policy checking if we are seeing last header.
879 * note that we do not visit this with protocols with pcb layer
880 * code - like udp/tcp/raw ip.
881 */
882 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
883 /*
884 * Check if the packet has already had IPsec processing
885 * done. If so, then just pass it along. This tag gets
886 * set during AH, ESP, etc. input handling, before the
887 * packet is returned to the ip input queue for delivery.
888 */
889 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
890 crit_enter();
891 if (mtag != NULL) {
892 tdbi = (struct tdb_ident *)m_tag_data(mtag);
893 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
894 } else {
895 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
896 IP_FORWARDING, &error);
897 }
898 if (sp != NULL) {
899 /*
900 * Check security policy against packet attributes.
901 */
902 error = ipsec_in_reject(sp, m);
903 KEY_FREESP(&sp);
904 } else {
905 /* XXX error stat??? */
906 error = EINVAL;
907 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
908 goto bad;
909 }
910 crit_exit();
911 if (error)
912 goto bad;
913 }
914 #endif /* FAST_IPSEC */
915
916 /*
917 * We must forward the packet to the correct protocol thread if
918 * we are not already in it.
919 *
920 * NOTE: ip_len is now in host form. ip_len is not adjusted
921 * further for protocol processing, instead we pass hlen
922 * to the protosw and let it deal with it.
923 */
924 ipstat.ips_delivered++;
925
926 if ((m->m_flags & M_HASH) == 0) {
927 #ifdef RSS_DEBUG
928 atomic_add_long(&ip_rehash_count, 1);
929 #endif
930 ip->ip_len = htons(ip->ip_len + hlen);
931 ip->ip_off = htons(ip->ip_off);
932
933 ip_hashfn(&m, 0, IP_MPORT_IN);
934 if (m == NULL)
935 return;
936
937 ip = mtod(m, struct ip *);
938 ip->ip_len = ntohs(ip->ip_len) - hlen;
939 ip->ip_off = ntohs(ip->ip_off);
940 KKASSERT(m->m_flags & M_HASH);
941 }
942 port = netisr_hashport(m->m_pkthdr.hash);
943
944 if (port != &curthread->td_msgport) {
945 struct netmsg_packet *pmsg;
946
947 #ifdef RSS_DEBUG
948 atomic_add_long(&ip_dispatch_slow, 1);
949 #endif
950
951 pmsg = &m->m_hdr.mh_netmsg;
952 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
953 0, transport_processing_handler);
954 pmsg->nm_packet = m;
955 pmsg->base.lmsg.u.ms_result = hlen;
956 lwkt_sendmsg(port, &pmsg->base.lmsg);
957 } else {
958 #ifdef RSS_DEBUG
959 atomic_add_long(&ip_dispatch_fast, 1);
960 #endif
961 transport_processing_oncpu(m, hlen, ip);
962 }
963 return;
964
965 bad:
966 m_freem(m);
967 }
968
969 /*
970 * Take incoming datagram fragment and try to reassemble it into
971 * whole datagram. If a chain for reassembly of this datagram already
972 * exists, then it is given as fp; otherwise have to make a chain.
973 */
974 struct mbuf *
975 ip_reass(struct mbuf *m)
976 {
977 struct ip *ip = mtod(m, struct ip *);
978 struct mbuf *p = NULL, *q, *nq;
979 struct mbuf *n;
980 struct ipq *fp = NULL;
981 struct ipqhead *head;
982 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
983 int i, next;
984 u_short sum;
985
986 /* If maxnipq is 0, never accept fragments. */
987 if (maxnipq == 0) {
988 ipstat.ips_fragments++;
989 ipstat.ips_fragdropped++;
990 m_freem(m);
991 return NULL;
992 }
993
994 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
995 /*
996 * Look for queue of fragments of this datagram.
997 */
998 lwkt_gettoken(&ipq_token);
999 head = &ipq[sum];
1000 TAILQ_FOREACH(fp, head, ipq_list) {
1001 if (ip->ip_id == fp->ipq_id &&
1002 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1003 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1004 ip->ip_p == fp->ipq_p)
1005 goto found;
1006 }
1007
1008 fp = NULL;
1009
1010 /*
1011 * Enforce upper bound on number of fragmented packets
1012 * for which we attempt reassembly;
1013 * If maxnipq is -1, accept all fragments without limitation.
1014 */
1015 if (nipq > maxnipq && maxnipq > 0) {
1016 /*
1017 * drop something from the tail of the current queue
1018 * before proceeding further
1019 */
1020 struct ipq *q = TAILQ_LAST(head, ipqhead);
1021 if (q == NULL) {
1022 /*
1023 * The current queue is empty,
1024 * so drop from one of the others.
1025 */
1026 for (i = 0; i < IPREASS_NHASH; i++) {
1027 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
1028 if (r) {
1029 ipstat.ips_fragtimeout += r->ipq_nfrags;
1030 ip_freef(&ipq[i], r);
1031 break;
1032 }
1033 }
1034 } else {
1035 ipstat.ips_fragtimeout += q->ipq_nfrags;
1036 ip_freef(head, q);
1037 }
1038 }
1039 found:
1040 /*
1041 * Adjust ip_len to not reflect header,
1042 * convert offset of this to bytes.
1043 */
1044 ip->ip_len -= hlen;
1045 if (ip->ip_off & IP_MF) {
1046 /*
1047 * Make sure that fragments have a data length
1048 * that's a non-zero multiple of 8 bytes.
1049 */
1050 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1051 ipstat.ips_toosmall++; /* XXX */
1052 m_freem(m);
1053 goto done;
1054 }
1055 m->m_flags |= M_FRAG;
1056 } else {
1057 m->m_flags &= ~M_FRAG;
1058 }
1059 ip->ip_off <<= 3;
1060
1061 ipstat.ips_fragments++;
1062 m->m_pkthdr.header = ip;
1063
1064 /*
1065 * If the hardware has not done csum over this fragment
1066 * then csum_data is not valid at all.
1067 */
1068 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1069 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1070 m->m_pkthdr.csum_data = 0;
1071 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1072 }
1073
1074 /*
1075 * Presence of header sizes in mbufs
1076 * would confuse code below.
1077 */
1078 m->m_data += hlen;
1079 m->m_len -= hlen;
1080
1081 /*
1082 * If first fragment to arrive, create a reassembly queue.
1083 */
1084 if (fp == NULL) {
1085 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1086 goto dropfrag;
1087 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1088 nipq++;
1089 fp->ipq_nfrags = 1;
1090 fp->ipq_ttl = IPFRAGTTL;
1091 fp->ipq_p = ip->ip_p;
1092 fp->ipq_id = ip->ip_id;
1093 fp->ipq_src = ip->ip_src;
1094 fp->ipq_dst = ip->ip_dst;
1095 fp->ipq_frags = m;
1096 m->m_nextpkt = NULL;
1097 goto inserted;
1098 } else {
1099 fp->ipq_nfrags++;
1100 }
1101
1102 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1103
1104 /*
1105 * Find a segment which begins after this one does.
1106 */
1107 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1108 if (GETIP(q)->ip_off > ip->ip_off)
1109 break;
1110 }
1111
1112 /*
1113 * If there is a preceding segment, it may provide some of
1114 * our data already. If so, drop the data from the incoming
1115 * segment. If it provides all of our data, drop us, otherwise
1116 * stick new segment in the proper place.
1117 *
1118 * If some of the data is dropped from the the preceding
1119 * segment, then it's checksum is invalidated.
1120 */
1121 if (p) {
1122 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1123 if (i > 0) {
1124 if (i >= ip->ip_len)
1125 goto dropfrag;
1126 m_adj(m, i);
1127 m->m_pkthdr.csum_flags = 0;
1128 ip->ip_off += i;
1129 ip->ip_len -= i;
1130 }
1131 m->m_nextpkt = p->m_nextpkt;
1132 p->m_nextpkt = m;
1133 } else {
1134 m->m_nextpkt = fp->ipq_frags;
1135 fp->ipq_frags = m;
1136 }
1137
1138 /*
1139 * While we overlap succeeding segments trim them or,
1140 * if they are completely covered, dequeue them.
1141 */
1142 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1143 q = nq) {
1144 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1145 if (i < GETIP(q)->ip_len) {
1146 GETIP(q)->ip_len -= i;
1147 GETIP(q)->ip_off += i;
1148 m_adj(q, i);
1149 q->m_pkthdr.csum_flags = 0;
1150 break;
1151 }
1152 nq = q->m_nextpkt;
1153 m->m_nextpkt = nq;
1154 ipstat.ips_fragdropped++;
1155 fp->ipq_nfrags--;
1156 q->m_nextpkt = NULL;
1157 m_freem(q);
1158 }
1159
1160 inserted:
1161 /*
1162 * Check for complete reassembly and perform frag per packet
1163 * limiting.
1164 *
1165 * Frag limiting is performed here so that the nth frag has
1166 * a chance to complete the packet before we drop the packet.
1167 * As a result, n+1 frags are actually allowed per packet, but
1168 * only n will ever be stored. (n = maxfragsperpacket.)
1169 *
1170 */
1171 next = 0;
1172 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1173 if (GETIP(q)->ip_off != next) {
1174 if (fp->ipq_nfrags > maxfragsperpacket) {
1175 ipstat.ips_fragdropped += fp->ipq_nfrags;
1176 ip_freef(head, fp);
1177 }
1178 goto done;
1179 }
1180 next += GETIP(q)->ip_len;
1181 }
1182 /* Make sure the last packet didn't have the IP_MF flag */
1183 if (p->m_flags & M_FRAG) {
1184 if (fp->ipq_nfrags > maxfragsperpacket) {
1185 ipstat.ips_fragdropped += fp->ipq_nfrags;
1186 ip_freef(head, fp);
1187 }
1188 goto done;
1189 }
1190
1191 /*
1192 * Reassembly is complete. Make sure the packet is a sane size.
1193 */
1194 q = fp->ipq_frags;
1195 ip = GETIP(q);
1196 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1197 ipstat.ips_toolong++;
1198 ipstat.ips_fragdropped += fp->ipq_nfrags;
1199 ip_freef(head, fp);
1200 goto done;
1201 }
1202
1203 /*
1204 * Concatenate fragments.
1205 */
1206 m = q;
1207 n = m->m_next;
1208 m->m_next = NULL;
1209 m_cat(m, n);
1210 nq = q->m_nextpkt;
1211 q->m_nextpkt = NULL;
1212 for (q = nq; q != NULL; q = nq) {
1213 nq = q->m_nextpkt;
1214 q->m_nextpkt = NULL;
1215 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1216 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1217 m_cat(m, q);
1218 }
1219
1220 /*
1221 * Clean up the 1's complement checksum. Carry over 16 bits must
1222 * be added back. This assumes no more then 65535 packet fragments
1223 * were reassembled. A second carry can also occur (but not a third).
1224 */
1225 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1226 (m->m_pkthdr.csum_data >> 16);
1227 if (m->m_pkthdr.csum_data > 0xFFFF)
1228 m->m_pkthdr.csum_data -= 0xFFFF;
1229
1230 /*
1231 * Create header for new ip packet by
1232 * modifying header of first packet;
1233 * dequeue and discard fragment reassembly header.
1234 * Make header visible.
1235 */
1236 ip->ip_len = next;
1237 ip->ip_src = fp->ipq_src;
1238 ip->ip_dst = fp->ipq_dst;
1239 TAILQ_REMOVE(head, fp, ipq_list);
1240 nipq--;
1241 mpipe_free(&ipq_mpipe, fp);
1242 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1243 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1244 /* some debugging cruft by sklower, below, will go away soon */
1245 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1246 int plen = 0;
1247
1248 for (n = m; n; n = n->m_next)
1249 plen += n->m_len;
1250 m->m_pkthdr.len = plen;
1251 }
1252
1253 /*
1254 * Reassembly complete, return the next protocol.
1255 *
1256 * Be sure to clear M_HASH to force the packet
1257 * to be re-characterized.
1258 *
1259 * Clear M_FRAG, we are no longer a fragment.
1260 */
1261 m->m_flags &= ~(M_HASH | M_FRAG);
1262
1263 ipstat.ips_reassembled++;
1264 lwkt_reltoken(&ipq_token);
1265 return (m);
1266
1267 dropfrag:
1268 ipstat.ips_fragdropped++;
1269 if (fp != NULL)
1270 fp->ipq_nfrags--;
1271 m_freem(m);
1272 done:
1273 lwkt_reltoken(&ipq_token);
1274 return (NULL);
1275
1276 #undef GETIP
1277 }
1278
1279 /*
1280 * Free a fragment reassembly header and all
1281 * associated datagrams.
1282 *
1283 * Called with ipq_token held.
1284 */
1285 static void
1286 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1287 {
1288 struct mbuf *q;
1289
1290 /*
1291 * Remove first to protect against blocking
1292 */
1293 TAILQ_REMOVE(fhp, fp, ipq_list);
1294
1295 /*
1296 * Clean out at our leisure
1297 */
1298 while (fp->ipq_frags) {
1299 q = fp->ipq_frags;
1300 fp->ipq_frags = q->m_nextpkt;
1301 q->m_nextpkt = NULL;
1302 m_freem(q);
1303 }
1304 mpipe_free(&ipq_mpipe, fp);
1305 nipq--;
1306 }
1307
1308 /*
1309 * IP timer processing;
1310 * if a timer expires on a reassembly
1311 * queue, discard it.
1312 */
1313 void
1314 ip_slowtimo(void)
1315 {
1316 struct ipq *fp, *fp_temp;
1317 struct ipqhead *head;
1318 int i;
1319
1320 lwkt_gettoken(&ipq_token);
1321 for (i = 0; i < IPREASS_NHASH; i++) {
1322 head = &ipq[i];
1323 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1324 if (--fp->ipq_ttl == 0) {
1325 ipstat.ips_fragtimeout += fp->ipq_nfrags;
1326 ip_freef(head, fp);
1327 }
1328 }
1329 }
1330 /*
1331 * If we are over the maximum number of fragments
1332 * (due to the limit being lowered), drain off
1333 * enough to get down to the new limit.
1334 */
1335 if (maxnipq >= 0 && nipq > maxnipq) {
1336 for (i = 0; i < IPREASS_NHASH; i++) {
1337 head = &ipq[i];
1338 while (nipq > maxnipq && !TAILQ_EMPTY(head)) {
1339 ipstat.ips_fragdropped +=
1340 TAILQ_FIRST(head)->ipq_nfrags;
1341 ip_freef(head, TAILQ_FIRST(head));
1342 }
1343 }
1344 }
1345 lwkt_reltoken(&ipq_token);
1346 ipflow_slowtimo();
1347 }
1348
1349 /*
1350 * Drain off all datagram fragments.
1351 */
1352 void
1353 ip_drain(void)
1354 {
1355 struct ipqhead *head;
1356 int i;
1357
1358 lwkt_gettoken(&ipq_token);
1359 for (i = 0; i < IPREASS_NHASH; i++) {
1360 head = &ipq[i];
1361 while (!TAILQ_EMPTY(head)) {
1362 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1363 ip_freef(head, TAILQ_FIRST(head));
1364 }
1365 }
1366 lwkt_reltoken(&ipq_token);
1367 in_rtqdrain();
1368 }
1369
1370 /*
1371 * Do option processing on a datagram,
1372 * possibly discarding it if bad options are encountered,
1373 * or forwarding it if source-routed.
1374 * The pass argument is used when operating in the IPSTEALTH
1375 * mode to tell what options to process:
1376 * [LS]SRR (pass 0) or the others (pass 1).
1377 * The reason for as many as two passes is that when doing IPSTEALTH,
1378 * non-routing options should be processed only if the packet is for us.
1379 * Returns 1 if packet has been forwarded/freed,
1380 * 0 if the packet should be processed further.
1381 */
1382 static int
1383 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1384 {
1385 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1386 struct ip *ip = mtod(m, struct ip *);
1387 u_char *cp;
1388 struct in_ifaddr *ia;
1389 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1390 boolean_t forward = FALSE;
1391 struct in_addr *sin, dst;
1392 n_time ntime;
1393
1394 dst = ip->ip_dst;
1395 cp = (u_char *)(ip + 1);
1396 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1397 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1398 opt = cp[IPOPT_OPTVAL];
1399 if (opt == IPOPT_EOL)
1400 break;
1401 if (opt == IPOPT_NOP)
1402 optlen = 1;
1403 else {
1404 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1405 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1406 goto bad;
1407 }
1408 optlen = cp[IPOPT_OLEN];
1409 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1410 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1411 goto bad;
1412 }
1413 }
1414 switch (opt) {
1415
1416 default:
1417 break;
1418
1419 /*
1420 * Source routing with record.
1421 * Find interface with current destination address.
1422 * If none on this machine then drop if strictly routed,
1423 * or do nothing if loosely routed.
1424 * Record interface address and bring up next address
1425 * component. If strictly routed make sure next
1426 * address is on directly accessible net.
1427 */
1428 case IPOPT_LSRR:
1429 case IPOPT_SSRR:
1430 if (ipstealth && pass > 0)
1431 break;
1432 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1433 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1434 goto bad;
1435 }
1436 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1437 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1438 goto bad;
1439 }
1440 ipaddr.sin_addr = ip->ip_dst;
1441 ia = (struct in_ifaddr *)
1442 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1443 if (ia == NULL) {
1444 if (opt == IPOPT_SSRR) {
1445 type = ICMP_UNREACH;
1446 code = ICMP_UNREACH_SRCFAIL;
1447 goto bad;
1448 }
1449 if (!ip_dosourceroute)
1450 goto nosourcerouting;
1451 /*
1452 * Loose routing, and not at next destination
1453 * yet; nothing to do except forward.
1454 */
1455 break;
1456 }
1457 off--; /* 0 origin */
1458 if (off > optlen - (int)sizeof(struct in_addr)) {
1459 /*
1460 * End of source route. Should be for us.
1461 */
1462 if (!ip_acceptsourceroute)
1463 goto nosourcerouting;
1464 save_rte(m, cp, ip->ip_src);
1465 break;
1466 }
1467 if (ipstealth)
1468 goto dropit;
1469 if (!ip_dosourceroute) {
1470 if (ipforwarding) {
1471 char buf[sizeof "aaa.bbb.ccc.ddd"];
1472
1473 /*
1474 * Acting as a router, so generate ICMP
1475 */
1476 nosourcerouting:
1477 strcpy(buf, inet_ntoa(ip->ip_dst));
1478 log(LOG_WARNING,
1479 "attempted source route from %s to %s\n",
1480 inet_ntoa(ip->ip_src), buf);
1481 type = ICMP_UNREACH;
1482 code = ICMP_UNREACH_SRCFAIL;
1483 goto bad;
1484 } else {
1485 /*
1486 * Not acting as a router,
1487 * so silently drop.
1488 */
1489 dropit:
1490 ipstat.ips_cantforward++;
1491 m_freem(m);
1492 return (1);
1493 }
1494 }
1495
1496 /*
1497 * locate outgoing interface
1498 */
1499 memcpy(&ipaddr.sin_addr, cp + off,
1500 sizeof ipaddr.sin_addr);
1501
1502 if (opt == IPOPT_SSRR) {
1503 #define INA struct in_ifaddr *
1504 #define SA struct sockaddr *
1505 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1506 == NULL)
1507 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1508 } else {
1509 ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1510 }
1511 if (ia == NULL) {
1512 type = ICMP_UNREACH;
1513 code = ICMP_UNREACH_SRCFAIL;
1514 goto bad;
1515 }
1516 ip->ip_dst = ipaddr.sin_addr;
1517 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1518 sizeof(struct in_addr));
1519 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1520 /*
1521 * Let ip_intr's mcast routing check handle mcast pkts
1522 */
1523 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1524 break;
1525
1526 case IPOPT_RR:
1527 if (ipstealth && pass == 0)
1528 break;
1529 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1530 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1531 goto bad;
1532 }
1533 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1534 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1535 goto bad;
1536 }
1537 /*
1538 * If no space remains, ignore.
1539 */
1540 off--; /* 0 origin */
1541 if (off > optlen - (int)sizeof(struct in_addr))
1542 break;
1543 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1544 sizeof ipaddr.sin_addr);
1545 /*
1546 * locate outgoing interface; if we're the destination,
1547 * use the incoming interface (should be same).
1548 */
1549 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1550 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1551 type = ICMP_UNREACH;
1552 code = ICMP_UNREACH_HOST;
1553 goto bad;
1554 }
1555 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1556 sizeof(struct in_addr));
1557 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1558 break;
1559
1560 case IPOPT_TS:
1561 if (ipstealth && pass == 0)
1562 break;
1563 code = cp - (u_char *)ip;
1564 if (optlen < 4 || optlen > 40) {
1565 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1566 goto bad;
1567 }
1568 if ((off = cp[IPOPT_OFFSET]) < 5) {
1569 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1570 goto bad;
1571 }
1572 if (off > optlen - (int)sizeof(int32_t)) {
1573 cp[IPOPT_OFFSET + 1] += (1 << 4);
1574 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1575 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1576 goto bad;
1577 }
1578 break;
1579 }
1580 off--; /* 0 origin */
1581 sin = (struct in_addr *)(cp + off);
1582 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1583
1584 case IPOPT_TS_TSONLY:
1585 break;
1586
1587 case IPOPT_TS_TSANDADDR:
1588 if (off + sizeof(n_time) +
1589 sizeof(struct in_addr) > optlen) {
1590 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1591 goto bad;
1592 }
1593 ipaddr.sin_addr = dst;
1594 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1595 m->m_pkthdr.rcvif);
1596 if (ia == NULL)
1597 continue;
1598 memcpy(sin, &IA_SIN(ia)->sin_addr,
1599 sizeof(struct in_addr));
1600 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1601 off += sizeof(struct in_addr);
1602 break;
1603
1604 case IPOPT_TS_PRESPEC:
1605 if (off + sizeof(n_time) +
1606 sizeof(struct in_addr) > optlen) {
1607 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1608 goto bad;
1609 }
1610 memcpy(&ipaddr.sin_addr, sin,
1611 sizeof(struct in_addr));
1612 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1613 continue;
1614 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1615 off += sizeof(struct in_addr);
1616 break;
1617
1618 default:
1619 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1620 goto bad;
1621 }
1622 ntime = iptime();
1623 memcpy(cp + off, &ntime, sizeof(n_time));
1624 cp[IPOPT_OFFSET] += sizeof(n_time);
1625 }
1626 }
1627 if (forward && ipforwarding) {
1628 ip_forward(m, TRUE, next_hop);
1629 return (1);
1630 }
1631 return (0);
1632 bad:
1633 icmp_error(m, type, code, 0, 0);
1634 ipstat.ips_badoptions++;
1635 return (1);
1636 }
1637
1638 /*
1639 * Given address of next destination (final or next hop),
1640 * return internet address info of interface to be used to get there.
1641 */
1642 struct in_ifaddr *
1643 ip_rtaddr(struct in_addr dst, struct route *ro0)
1644 {
1645 struct route sro, *ro;
1646 struct sockaddr_in *sin;
1647 struct in_ifaddr *ia;
1648
1649 if (ro0 != NULL) {
1650 ro = ro0;
1651 } else {
1652 bzero(&sro, sizeof(sro));
1653 ro = &sro;
1654 }
1655
1656 sin = (struct sockaddr_in *)&ro->ro_dst;
1657
1658 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1659 if (ro->ro_rt != NULL) {
1660 RTFREE(ro->ro_rt);
1661 ro->ro_rt = NULL;
1662 }
1663 sin->sin_family = AF_INET;
1664 sin->sin_len = sizeof *sin;
1665 sin->sin_addr = dst;
1666 rtalloc_ign(ro, RTF_PRCLONING);
1667 }
1668
1669 if (ro->ro_rt == NULL)
1670 return (NULL);
1671
1672 ia = ifatoia(ro->ro_rt->rt_ifa);
1673
1674 if (ro == &sro)
1675 RTFREE(ro->ro_rt);
1676 return ia;
1677 }
1678
1679 /*
1680 * Save incoming source route for use in replies,
1681 * to be picked up later by ip_srcroute if the receiver is interested.
1682 */
1683 static void
1684 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1685 {
1686 struct m_tag *mtag;
1687 struct ip_srcrt_opt *opt;
1688 unsigned olen;
1689
1690 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1691 if (mtag == NULL)
1692 return;
1693 opt = m_tag_data(mtag);
1694
1695 olen = option[IPOPT_OLEN];
1696 #ifdef DIAGNOSTIC
1697 if (ipprintfs)
1698 kprintf("save_rte: olen %d\n", olen);
1699 #endif
1700 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1701 m_tag_free(mtag);
1702 return;
1703 }
1704 bcopy(option, opt->ip_srcrt.srcopt, olen);
1705 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1706 opt->ip_srcrt.dst = dst;
1707 m_tag_prepend(m, mtag);
1708 }
1709
1710 /*
1711 * Retrieve incoming source route for use in replies,
1712 * in the same form used by setsockopt.
1713 * The first hop is placed before the options, will be removed later.
1714 */
1715 struct mbuf *
1716 ip_srcroute(struct mbuf *m0)
1717 {
1718 struct in_addr *p, *q;
1719 struct mbuf *m;
1720 struct m_tag *mtag;
1721 struct ip_srcrt_opt *opt;
1722
1723 if (m0 == NULL)
1724 return NULL;
1725
1726 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1727 if (mtag == NULL)
1728 return NULL;
1729 opt = m_tag_data(mtag);
1730
1731 if (opt->ip_nhops == 0)
1732 return (NULL);
1733 m = m_get(MB_DONTWAIT, MT_HEADER);
1734 if (m == NULL)
1735 return (NULL);
1736
1737 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1738
1739 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1740 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1741 sizeof(struct in_addr) + OPTSIZ;
1742 #ifdef DIAGNOSTIC
1743 if (ipprintfs) {
1744 kprintf("ip_srcroute: nhops %d mlen %d",
1745 opt->ip_nhops, m->m_len);
1746 }
1747 #endif
1748
1749 /*
1750 * First save first hop for return route
1751 */
1752 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1753 *(mtod(m, struct in_addr *)) = *p--;
1754 #ifdef DIAGNOSTIC
1755 if (ipprintfs)
1756 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1757 #endif
1758
1759 /*
1760 * Copy option fields and padding (nop) to mbuf.
1761 */
1762 opt->ip_srcrt.nop = IPOPT_NOP;
1763 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1764 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1765 OPTSIZ);
1766 q = (struct in_addr *)(mtod(m, caddr_t) +
1767 sizeof(struct in_addr) + OPTSIZ);
1768 #undef OPTSIZ
1769 /*
1770 * Record return path as an IP source route,
1771 * reversing the path (pointers are now aligned).
1772 */
1773 while (p >= opt->ip_srcrt.route) {
1774 #ifdef DIAGNOSTIC
1775 if (ipprintfs)
1776 kprintf(" %x", ntohl(q->s_addr));
1777 #endif
1778 *q++ = *p--;
1779 }
1780 /*
1781 * Last hop goes to final destination.
1782 */
1783 *q = opt->ip_srcrt.dst;
1784 m_tag_delete(m0, mtag);
1785 #ifdef DIAGNOSTIC
1786 if (ipprintfs)
1787 kprintf(" %x\n", ntohl(q->s_addr));
1788 #endif
1789 return (m);
1790 }
1791
1792 /*
1793 * Strip out IP options.
1794 */
1795 void
1796 ip_stripoptions(struct mbuf *m)
1797 {
1798 int datalen;
1799 struct ip *ip = mtod(m, struct ip *);
1800 caddr_t opts;
1801 int optlen;
1802
1803 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1804 opts = (caddr_t)(ip + 1);
1805 datalen = m->m_len - (sizeof(struct ip) + optlen);
1806 bcopy(opts + optlen, opts, datalen);
1807 m->m_len -= optlen;
1808 if (m->m_flags & M_PKTHDR)
1809 m->m_pkthdr.len -= optlen;
1810 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1811 }
1812
1813 u_char inetctlerrmap[PRC_NCMDS] = {
1814 0, 0, 0, 0,
1815 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1816 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1817 EMSGSIZE, EHOSTUNREACH, 0, 0,
1818 0, 0, 0, 0,
1819 ENOPROTOOPT, ECONNREFUSED
1820 };
1821
1822 /*
1823 * Forward a packet. If some error occurs return the sender
1824 * an icmp packet. Note we can't always generate a meaningful
1825 * icmp message because icmp doesn't have a large enough repertoire
1826 * of codes and types.
1827 *
1828 * If not forwarding, just drop the packet. This could be confusing
1829 * if ipforwarding was zero but some routing protocol was advancing
1830 * us as a gateway to somewhere. However, we must let the routing
1831 * protocol deal with that.
1832 *
1833 * The using_srcrt parameter indicates whether the packet is being forwarded
1834 * via a source route.
1835 */
1836 void
1837 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1838 {
1839 struct ip *ip = mtod(m, struct ip *);
1840 struct rtentry *rt;
1841 struct route fwd_ro;
1842 int error, type = 0, code = 0, destmtu = 0;
1843 struct mbuf *mcopy, *mtemp = NULL;
1844 n_long dest;
1845 struct in_addr pkt_dst;
1846
1847 dest = INADDR_ANY;
1848 /*
1849 * Cache the destination address of the packet; this may be
1850 * changed by use of 'ipfw fwd'.
1851 */
1852 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1853
1854 #ifdef DIAGNOSTIC
1855 if (ipprintfs)
1856 kprintf("forward: src %x dst %x ttl %x\n",
1857 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1858 #endif
1859
1860 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1861 ipstat.ips_cantforward++;
1862 m_freem(m);
1863 return;
1864 }
1865 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1866 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1867 return;
1868 }
1869
1870 bzero(&fwd_ro, sizeof(fwd_ro));
1871 ip_rtaddr(pkt_dst, &fwd_ro);
1872 if (fwd_ro.ro_rt == NULL) {
1873 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1874 return;
1875 }
1876 rt = fwd_ro.ro_rt;
1877
1878 if (curthread->td_type == TD_TYPE_NETISR) {
1879 /*
1880 * Save the IP header and at most 8 bytes of the payload,
1881 * in case we need to generate an ICMP message to the src.
1882 */
1883 mtemp = ipforward_mtemp[mycpuid];
1884 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1885 mtemp->m_data == mtemp->m_pktdat &&
1886 m_tag_first(mtemp) == NULL,
1887 ("ip_forward invalid mtemp1"));
1888
1889 if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) {
1890 /*
1891 * It's probably ok if the pkthdr dup fails (because
1892 * the deep copy of the tag chain failed), but for now
1893 * be conservative and just discard the copy since
1894 * code below may some day want the tags.
1895 */
1896 mtemp = NULL;
1897 } else {
1898 mtemp->m_type = m->m_type;
1899 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1900 (int)ip->ip_len);
1901 mtemp->m_pkthdr.len = mtemp->m_len;
1902 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t));
1903 }
1904 }
1905
1906 if (!ipstealth)
1907 ip->ip_ttl -= IPTTLDEC;
1908
1909 /*
1910 * If forwarding packet using same interface that it came in on,
1911 * perhaps should send a redirect to sender to shortcut a hop.
1912 * Only send redirect if source is sending directly to us,
1913 * and if packet was not source routed (or has any options).
1914 * Also, don't send redirect if forwarding using a default route
1915 * or a route modified by a redirect.
1916 */
1917 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1918 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1919 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1920 ipsendredirects && !using_srcrt && next_hop == NULL) {
1921 u_long src = ntohl(ip->ip_src.s_addr);
1922 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1923
1924 if (rt_ifa != NULL &&
1925 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1926 if (rt->rt_flags & RTF_GATEWAY)
1927 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1928 else
1929 dest = pkt_dst.s_addr;
1930 /*
1931 * Router requirements says to only send
1932 * host redirects.
1933 */
1934 type = ICMP_REDIRECT;
1935 code = ICMP_REDIRECT_HOST;
1936 #ifdef DIAGNOSTIC
1937 if (ipprintfs)
1938 kprintf("redirect (%d) to %x\n", code, dest);
1939 #endif
1940 }
1941 }
1942
1943 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1944 if (error == 0) {
1945 ipstat.ips_forward++;
1946 if (type == 0) {
1947 if (mtemp)
1948 ipflow_create(&fwd_ro, mtemp);
1949 goto done;
1950 } else {
1951 ipstat.ips_redirectsent++;
1952 }
1953 } else {
1954 ipstat.ips_cantforward++;
1955 }
1956
1957 if (mtemp == NULL)
1958 goto done;
1959
1960 /*
1961 * Errors that do not require generating ICMP message
1962 */
1963 switch (error) {
1964 case ENOBUFS:
1965 /*
1966 * A router should not generate ICMP_SOURCEQUENCH as
1967 * required in RFC1812 Requirements for IP Version 4 Routers.
1968 * Source quench could be a big problem under DoS attacks,
1969 * or if the underlying interface is rate-limited.
1970 * Those who need source quench packets may re-enable them
1971 * via the net.inet.ip.sendsourcequench sysctl.
1972 */
1973 if (!ip_sendsourcequench)
1974 goto done;
1975 break;
1976
1977 case EACCES: /* ipfw denied packet */
1978 goto done;
1979 }
1980
1981 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1982 mtemp->m_data == mtemp->m_pktdat,
1983 ("ip_forward invalid mtemp2"));
1984 mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT);
1985 if (mcopy == NULL)
1986 goto done;
1987
1988 /*
1989 * Send ICMP message.
1990 */
1991 switch (error) {
1992 case 0: /* forwarded, but need redirect */
1993 /* type, code set above */
1994 break;
1995
1996 case ENETUNREACH: /* shouldn't happen, checked above */
1997 case EHOSTUNREACH:
1998 case ENETDOWN:
1999 case EHOSTDOWN:
2000 default:
2001 type = ICMP_UNREACH;
2002 code = ICMP_UNREACH_HOST;
2003 break;
2004
2005 case EMSGSIZE:
2006 type = ICMP_UNREACH;
2007 code = ICMP_UNREACH_NEEDFRAG;
2008 #ifdef IPSEC
2009 /*
2010 * If the packet is routed over IPsec tunnel, tell the
2011 * originator the tunnel MTU.
2012 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2013 * XXX quickhack!!!
2014 */
2015 if (fwd_ro.ro_rt != NULL) {
2016 struct secpolicy *sp = NULL;
2017 int ipsecerror;
2018 int ipsechdr;
2019 struct route *ro;
2020
2021 sp = ipsec4_getpolicybyaddr(mcopy,
2022 IPSEC_DIR_OUTBOUND,
2023 IP_FORWARDING,
2024 &ipsecerror);
2025
2026 if (sp == NULL)
2027 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2028 else {
2029 /* count IPsec header size */
2030 ipsechdr = ipsec4_hdrsiz(mcopy,
2031 IPSEC_DIR_OUTBOUND,
2032 NULL);
2033
2034 /*
2035 * find the correct route for outer IPv4
2036 * header, compute tunnel MTU.
2037 *
2038 */
2039 if (sp->req != NULL && sp->req->sav != NULL &&
2040 sp->req->sav->sah != NULL) {
2041 ro = &sp->req->sav->sah->sa_route;
2042 if (ro->ro_rt != NULL &&
2043 ro->ro_rt->rt_ifp != NULL) {
2044 destmtu =
2045 ro->ro_rt->rt_ifp->if_mtu;
2046 destmtu -= ipsechdr;
2047 }
2048 }
2049
2050 key_freesp(sp);
2051 }
2052 }
2053 #elif FAST_IPSEC
2054 /*
2055 * If the packet is routed over IPsec tunnel, tell the
2056 * originator the tunnel MTU.
2057 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2058 * XXX quickhack!!!
2059 */
2060 if (fwd_ro.ro_rt != NULL) {
2061 struct secpolicy *sp = NULL;
2062 int ipsecerror;
2063 int ipsechdr;
2064 struct route *ro;
2065
2066 sp = ipsec_getpolicybyaddr(mcopy,
2067 IPSEC_DIR_OUTBOUND,
2068 IP_FORWARDING,
2069 &ipsecerror);
2070
2071 if (sp == NULL)
2072 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2073 else {
2074 /* count IPsec header size */
2075 ipsechdr = ipsec4_hdrsiz(mcopy,
2076 IPSEC_DIR_OUTBOUND,
2077 NULL);
2078
2079 /*
2080 * find the correct route for outer IPv4
2081 * header, compute tunnel MTU.
2082 */
2083
2084 if (sp->req != NULL &&
2085 sp->req->sav != NULL &&
2086 sp->req->sav->sah != NULL) {
2087 ro = &sp->req->sav->sah->sa_route;
2088 if (ro->ro_rt != NULL &&
2089 ro->ro_rt->rt_ifp != NULL) {
2090 destmtu =
2091 ro->ro_rt->rt_ifp->if_mtu;
2092 destmtu -= ipsechdr;
2093 }
2094 }
2095
2096 KEY_FREESP(&sp);
2097 }
2098 }
2099 #else /* !IPSEC && !FAST_IPSEC */
2100 if (fwd_ro.ro_rt != NULL)
2101 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2102 #endif /*IPSEC*/
2103 ipstat.ips_cantfrag++;
2104 break;
2105
2106 case ENOBUFS:
2107 type = ICMP_SOURCEQUENCH;
2108 code = 0;
2109 break;
2110
2111 case EACCES: /* ipfw denied packet */
2112 panic("ip_forward EACCES should not reach");
2113 }
2114 icmp_error(mcopy, type, code, dest, destmtu);
2115 done:
2116 if (mtemp != NULL)
2117 m_tag_delete_chain(mtemp);
2118 if (fwd_ro.ro_rt != NULL)
2119 RTFREE(fwd_ro.ro_rt);
2120 }
2121
2122 void
2123 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2124 struct mbuf *m)
2125 {
2126 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2127 struct timeval tv;
2128
2129 microtime(&tv);
2130 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2131 SCM_TIMESTAMP, SOL_SOCKET);
2132 if (*mp)
2133 mp = &(*mp)->m_next;
2134 }
2135 if (inp->inp_flags & INP_RECVDSTADDR) {
2136 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2137 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2138 if (*mp)
2139 mp = &(*mp)->m_next;
2140 }
2141 if (inp->inp_flags & INP_RECVTTL) {
2142 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2143 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2144 if (*mp)
2145 mp = &(*mp)->m_next;
2146 }
2147 #ifdef notyet
2148 /* XXX
2149 * Moving these out of udp_input() made them even more broken
2150 * than they already were.
2151 */
2152 /* options were tossed already */
2153 if (inp->inp_flags & INP_RECVOPTS) {
2154 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2155 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2156 if (*mp)
2157 mp = &(*mp)->m_next;
2158 }
2159 /* ip_srcroute doesn't do what we want here, need to fix */
2160 if (inp->inp_flags & INP_RECVRETOPTS) {
2161 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2162 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2163 if (*mp)
2164 mp = &(*mp)->m_next;
2165 }
2166 #endif
2167 if (inp->inp_flags & INP_RECVIF) {
2168 struct ifnet *ifp;
2169 struct sdlbuf {
2170 struct sockaddr_dl sdl;
2171 u_char pad[32];
2172 } sdlbuf;
2173 struct sockaddr_dl *sdp;
2174 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2175
2176 if (((ifp = m->m_pkthdr.rcvif)) &&
2177 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2178 sdp = IF_LLSOCKADDR(ifp);
2179 /*
2180 * Change our mind and don't try copy.
2181 */
2182 if ((sdp->sdl_family != AF_LINK) ||
2183 (sdp->sdl_len > sizeof(sdlbuf))) {
2184 goto makedummy;
2185 }
2186 bcopy(sdp, sdl2, sdp->sdl_len);
2187 } else {
2188 makedummy:
2189 sdl2->sdl_len =
2190 offsetof(struct sockaddr_dl, sdl_data[0]);
2191 sdl2->sdl_family = AF_LINK;
2192 sdl2->sdl_index = 0;
2193 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2194 }
2195 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2196 IP_RECVIF, IPPROTO_IP);
2197 if (*mp)
2198 mp = &(*mp)->m_next;
2199 }
2200 }
2201
2202 /*
2203 * XXX these routines are called from the upper part of the kernel.
2204 *
2205 * They could also be moved to ip_mroute.c, since all the RSVP
2206 * handling is done there already.
2207 */
2208 int
2209 ip_rsvp_init(struct socket *so)
2210 {
2211 if (so->so_type != SOCK_RAW ||
2212 so->so_proto->pr_protocol != IPPROTO_RSVP)
2213 return EOPNOTSUPP;
2214
2215 if (ip_rsvpd != NULL)
2216 return EADDRINUSE;
2217
2218 ip_rsvpd = so;
2219 /*
2220 * This may seem silly, but we need to be sure we don't over-increment
2221 * the RSVP counter, in case something slips up.
2222 */
2223 if (!ip_rsvp_on) {
2224 ip_rsvp_on = 1;
2225 rsvp_on++;
2226 }
2227
2228 return 0;
2229 }
2230
2231 int
2232 ip_rsvp_done(void)
2233 {
2234 ip_rsvpd = NULL;
2235 /*
2236 * This may seem silly, but we need to be sure we don't over-decrement
2237 * the RSVP counter, in case something slips up.
2238 */
2239 if (ip_rsvp_on) {
2240 ip_rsvp_on = 0;
2241 rsvp_on--;
2242 }
2243 return 0;
2244 }
2245
2246 int
2247 rsvp_input(struct mbuf **mp, int *offp, int proto)
2248 {
2249 struct mbuf *m = *mp;
2250
2251 *mp = NULL;
2252
2253 if (rsvp_input_p) { /* call the real one if loaded */
2254 *mp = m;
2255 rsvp_input_p(mp, offp, proto);
2256 return(IPPROTO_DONE);
2257 }
2258
2259 /* Can still get packets with rsvp_on = 0 if there is a local member
2260 * of the group to which the RSVP packet is addressed. But in this
2261 * case we want to throw the packet away.
2262 */
2263
2264 if (!rsvp_on) {
2265 m_freem(m);
2266 return(IPPROTO_DONE);
2267 }
2268
2269 if (ip_rsvpd != NULL) {
2270 *mp = m;
2271 rip_input(mp, offp, proto);
2272 return(IPPROTO_DONE);
2273 }
2274 /* Drop the packet */
2275 m_freem(m);
2276 return(IPPROTO_DONE);
2277 }
Cache object: 8a4ecc9932d7b87c10e7a1807d76b696
|