in_proto.c revision 1.109 1 /* $NetBSD: in_proto.c,v 1.109 2014/05/18 14:46:16 rmind Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*
33 * Copyright (c) 1982, 1986, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)in_proto.c 8.2 (Berkeley) 2/9/95
61 */
62
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: in_proto.c,v 1.109 2014/05/18 14:46:16 rmind Exp $");
65
66 #include "opt_mrouting.h"
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 #include "opt_pim.h"
70 #include "opt_gateway.h"
71 #include "opt_compat_netbsd.h"
72
73 #include <sys/param.h>
74 #include <sys/socket.h>
75 #include <sys/protosw.h>
76 #include <sys/domain.h>
77 #include <sys/mbuf.h>
78
79 #include <net/if.h>
80 #include <net/radix.h>
81 #include <net/route.h>
82
83 #include <netinet/in.h>
84 #include <netinet/in_systm.h>
85 #include <netinet/ip.h>
86 #include <netinet/ip_var.h>
87 #include <netinet/ip_icmp.h>
88 #include <netinet/in_ifattach.h>
89 #include <netinet/in_pcb.h>
90 #include <netinet/in_proto.h>
91
92 #ifdef INET6
93 #ifndef INET
94 #include <netinet/in.h>
95 #endif
96 #include <netinet/ip6.h>
97 #endif
98
99 #include <netinet/igmp_var.h>
100 #ifdef PIM
101 #include <netinet/pim_var.h>
102 #endif
103 #include <netinet/tcp.h>
104 #include <netinet/tcp_fsm.h>
105 #include <netinet/tcp_seq.h>
106 #include <netinet/tcp_timer.h>
107 #include <netinet/tcp_var.h>
108 #include <netinet/tcpip.h>
109 #include <netinet/tcp_debug.h>
110 #include <netinet/udp.h>
111 #include <netinet/udp_var.h>
112 #include <netinet/ip_encap.h>
113
114 /*
115 * TCP/IP protocol family: IP, ICMP, UDP, TCP.
116 */
117
118 #ifdef IPSEC
119 #include <netipsec/ipsec.h>
120 #include <netipsec/key.h>
121 #endif /* IPSEC */
122
123 #include "carp.h"
124 #if NCARP > 0
125 #include <netinet/in_var.h>
126 #include <netinet/ip_carp.h>
127 #endif
128
129 #include "pfsync.h"
130 #if NPFSYNC > 0
131 #include <net/pfvar.h>
132 #include <net/if_pfsync.h>
133 #endif
134
135 #include "etherip.h"
136 #if NETHERIP > 0
137 #include <netinet/ip_etherip.h>
138 #endif
139
140 DOMAIN_DEFINE(inetdomain); /* forward declare and add to link set */
141
142 /* Wrappers to acquire kernel_lock. */
143
144 PR_WRAP_CTLINPUT(rip_ctlinput)
145 PR_WRAP_CTLINPUT(udp_ctlinput)
146 PR_WRAP_CTLINPUT(tcp_ctlinput)
147
148 #define rip_ctlinput rip_ctlinput_wrapper
149 #define udp_ctlinput udp_ctlinput_wrapper
150 #define tcp_ctlinput tcp_ctlinput_wrapper
151
152 PR_WRAP_CTLOUTPUT(rip_ctloutput)
153 PR_WRAP_CTLOUTPUT(udp_ctloutput)
154 PR_WRAP_CTLOUTPUT(tcp_ctloutput)
155
156 #define rip_ctloutput rip_ctloutput_wrapper
157 #define udp_ctloutput udp_ctloutput_wrapper
158 #define tcp_ctloutput tcp_ctloutput_wrapper
159
160 #if defined(IPSEC)
161 PR_WRAP_CTLINPUT(ah4_ctlinput)
162
163 #define ah4_ctlinput ah4_ctlinput_wrapper
164 PR_WRAP_CTLINPUT(esp4_ctlinput)
165
166 #define esp4_ctlinput esp4_ctlinput_wrapper
167 #endif
168
169 const struct protosw inetsw[] = {
170 { .pr_domain = &inetdomain,
171 .pr_init = ip_init,
172 .pr_output = ip_output,
173 .pr_fasttimo = ip_fasttimo,
174 .pr_slowtimo = ip_slowtimo,
175 .pr_drain = ip_drainstub,
176 },
177 { .pr_type = SOCK_DGRAM,
178 .pr_domain = &inetdomain,
179 .pr_protocol = IPPROTO_UDP,
180 .pr_flags = PR_ATOMIC|PR_ADDR|PR_PURGEIF,
181 .pr_input = udp_input,
182 .pr_ctlinput = udp_ctlinput,
183 .pr_ctloutput = udp_ctloutput,
184 .pr_usrreqs = &udp_usrreqs,
185 .pr_init = udp_init,
186 },
187 { .pr_type = SOCK_STREAM,
188 .pr_domain = &inetdomain,
189 .pr_protocol = IPPROTO_TCP,
190 .pr_flags = PR_CONNREQUIRED|PR_WANTRCVD|PR_LISTEN|PR_ABRTACPTDIS|PR_PURGEIF,
191 .pr_input = tcp_input,
192 .pr_ctlinput = tcp_ctlinput,
193 .pr_ctloutput = tcp_ctloutput,
194 .pr_usrreqs = &tcp_usrreqs,
195 .pr_init = tcp_init,
196 .pr_fasttimo = tcp_fasttimo,
197 .pr_drain = tcp_drainstub,
198 },
199 { .pr_type = SOCK_RAW,
200 .pr_domain = &inetdomain,
201 .pr_protocol = IPPROTO_RAW,
202 .pr_flags = PR_ATOMIC|PR_ADDR|PR_PURGEIF,
203 .pr_input = rip_input,
204 .pr_output = rip_output,
205 .pr_ctlinput = rip_ctlinput,
206 .pr_ctloutput = rip_ctloutput,
207 .pr_usrreqs = &rip_usrreqs,
208 },
209 { .pr_type = SOCK_RAW,
210 .pr_domain = &inetdomain,
211 .pr_protocol = IPPROTO_ICMP,
212 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
213 .pr_input = icmp_input,
214 .pr_output = rip_output,
215 .pr_ctlinput = rip_ctlinput,
216 .pr_ctloutput = rip_ctloutput,
217 .pr_usrreqs = &rip_usrreqs,
218 .pr_init = icmp_init,
219 },
220 #ifdef GATEWAY
221 { .pr_domain = &inetdomain,
222 .pr_protocol = IPPROTO_IP,
223 .pr_slowtimo = ipflow_slowtimo,
224 .pr_init = ipflow_poolinit,
225 },
226 #endif /* GATEWAY */
227 #ifdef IPSEC
228 { .pr_type = SOCK_RAW,
229 .pr_domain = &inetdomain,
230 .pr_protocol = IPPROTO_AH,
231 .pr_flags = PR_ATOMIC|PR_ADDR,
232 .pr_input = ipsec4_common_input,
233 .pr_ctlinput = ah4_ctlinput,
234 },
235 { .pr_type = SOCK_RAW,
236 .pr_domain = &inetdomain,
237 .pr_protocol = IPPROTO_ESP,
238 .pr_flags = PR_ATOMIC|PR_ADDR,
239 .pr_input = ipsec4_common_input,
240 .pr_ctlinput = esp4_ctlinput,
241 },
242 { .pr_type = SOCK_RAW,
243 .pr_domain = &inetdomain,
244 .pr_protocol = IPPROTO_IPCOMP,
245 .pr_flags = PR_ATOMIC|PR_ADDR,
246 .pr_input = ipsec4_common_input,
247 },
248 #endif /* IPSEC */
249 { .pr_type = SOCK_RAW,
250 .pr_domain = &inetdomain,
251 .pr_protocol = IPPROTO_IPV4,
252 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
253 .pr_input = encap4_input,
254 .pr_output = rip_output,
255 .pr_ctlinput = rip_ctlinput,
256 .pr_ctloutput = rip_ctloutput,
257 .pr_usrreqs = &rip_usrreqs,
258 .pr_init = encap_init,
259 },
260 #ifdef INET6
261 { .pr_type = SOCK_RAW,
262 .pr_domain = &inetdomain,
263 .pr_protocol = IPPROTO_IPV6,
264 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
265 .pr_input = encap4_input,
266 .pr_output = rip_output,
267 .pr_ctlinput = rip_ctlinput,
268 .pr_ctloutput = rip_ctloutput,
269 .pr_usrreqs = &rip_usrreqs,
270 .pr_init = encap_init,
271 },
272 #endif /* INET6 */
273 #if NETHERIP > 0
274 { .pr_type = SOCK_RAW,
275 .pr_domain = &inetdomain,
276 .pr_protocol = IPPROTO_ETHERIP,
277 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
278 .pr_input = ip_etherip_input,
279 .pr_output = rip_output,
280 .pr_ctlinput = rip_ctlinput,
281 .pr_ctloutput = rip_ctloutput,
282 .pr_usrreqs = &rip_usrreqs,
283 },
284 #endif /* NETHERIP > 0 */
285 #if NCARP > 0
286 { .pr_type = SOCK_RAW,
287 .pr_domain = &inetdomain,
288 .pr_protocol = IPPROTO_CARP,
289 .pr_flags = PR_ATOMIC|PR_ADDR,
290 .pr_input = carp_proto_input,
291 .pr_output = rip_output,
292 .pr_ctloutput = rip_ctloutput,
293 .pr_usrreqs = &rip_usrreqs,
294 .pr_init = carp_init,
295 },
296 #endif /* NCARP > 0 */
297 #if NPFSYNC > 0
298 { .pr_type = SOCK_RAW,
299 .pr_domain = &inetdomain,
300 .pr_protocol = IPPROTO_PFSYNC,
301 .pr_flags = PR_ATOMIC|PR_ADDR,
302 .pr_input = pfsync_input,
303 .pr_output = rip_output,
304 .pr_ctloutput = rip_ctloutput,
305 .pr_usrreqs = &rip_usrreqs,
306 },
307 #endif /* NPFSYNC > 0 */
308 { .pr_type = SOCK_RAW,
309 .pr_domain = &inetdomain,
310 .pr_protocol = IPPROTO_IGMP,
311 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
312 .pr_input = igmp_input,
313 .pr_output = rip_output,
314 .pr_ctloutput = rip_ctloutput,
315 .pr_ctlinput = rip_ctlinput,
316 .pr_usrreqs = &rip_usrreqs,
317 .pr_fasttimo = igmp_fasttimo,
318 .pr_slowtimo = igmp_slowtimo,
319 .pr_init = igmp_init,
320 },
321 #ifdef PIM
322 { .pr_type = SOCK_RAW,
323 .pr_domain = &inetdomain,
324 .pr_protocol = IPPROTO_PIM,
325 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
326 .pr_input = pim_input,
327 .pr_output = rip_output,
328 .pr_ctloutput = rip_ctloutput,
329 .pr_ctlinput = rip_ctlinput,
330 .pr_usrreqs = &rip_usrreqs,
331 },
332 #endif /* PIM */
333 /* raw wildcard */
334 { .pr_type = SOCK_RAW,
335 .pr_domain = &inetdomain,
336 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
337 .pr_input = rip_input,
338 .pr_output = rip_output,
339 .pr_ctloutput = rip_ctloutput,
340 .pr_ctlinput = rip_ctlinput,
341 .pr_usrreqs = &rip_usrreqs,
342 .pr_init = rip_init,
343 },
344 };
345
346 extern struct ifqueue ipintrq;
347
348 const struct sockaddr_in in_any = {
349 .sin_len = sizeof(struct sockaddr_in)
350 , .sin_family = AF_INET
351 , .sin_port = 0
352 , .sin_addr = {.s_addr = 0 /* INADDR_ANY */}
353 };
354
355 struct domain inetdomain = {
356 .dom_family = PF_INET, .dom_name = "internet", .dom_init = NULL,
357 .dom_externalize = NULL, .dom_dispose = NULL,
358 .dom_protosw = inetsw,
359 .dom_protoswNPROTOSW = &inetsw[__arraycount(inetsw)],
360 .dom_rtattach = rt_inithead,
361 .dom_rtoffset = 32,
362 .dom_maxrtkey = sizeof(struct ip_pack4),
363 #ifdef IPSELSRC
364 .dom_ifattach = in_domifattach,
365 .dom_ifdetach = in_domifdetach,
366 #else
367 .dom_ifattach = NULL,
368 .dom_ifdetach = NULL,
369 #endif
370 .dom_ifqueues = { &ipintrq, NULL },
371 .dom_link = { NULL },
372 .dom_mowner = MOWNER_INIT("",""),
373 .dom_sa_cmpofs = offsetof(struct sockaddr_in, sin_addr),
374 .dom_sa_cmplen = sizeof(struct in_addr),
375 .dom_sa_any = (const struct sockaddr *)&in_any,
376 .dom_sockaddr_const_addr = sockaddr_in_const_addr,
377 .dom_sockaddr_addr = sockaddr_in_addr,
378 .dom_rtcache = LIST_HEAD_INITIALIZER(inetdomain.dom_rtcache)
379 };
380
381 u_char ip_protox[IPPROTO_MAX];
382
383 int icmperrppslim = 100; /* 100pps */
384
385 static void
386 sockaddr_in_addrlen(const struct sockaddr *sa, socklen_t *slenp)
387 {
388 socklen_t slen;
389
390 if (slenp == NULL)
391 return;
392
393 slen = sockaddr_getlen(sa);
394 *slenp = (socklen_t)MIN(sizeof(struct in_addr),
395 slen - MIN(slen, offsetof(struct sockaddr_in, sin_addr)));
396 }
397
398 const void *
399 sockaddr_in_const_addr(const struct sockaddr *sa, socklen_t *slenp)
400 {
401 const struct sockaddr_in *sin;
402
403 sockaddr_in_addrlen(sa, slenp);
404 sin = (const struct sockaddr_in *)sa;
405 return &sin->sin_addr;
406 }
407
408 void *
409 sockaddr_in_addr(struct sockaddr *sa, socklen_t *slenp)
410 {
411 struct sockaddr_in *sin;
412
413 sockaddr_in_addrlen(sa, slenp);
414 sin = (struct sockaddr_in *)sa;
415 return &sin->sin_addr;
416 }
417
418 int
419 sockaddr_in_cmp(const struct sockaddr *sa1, const struct sockaddr *sa2)
420 {
421 uint_fast8_t len;
422 const uint_fast8_t addrofs = offsetof(struct sockaddr_in, sin_addr),
423 addrend = addrofs + sizeof(struct in_addr);
424 int rc;
425 const struct sockaddr_in *sin1, *sin2;
426
427 sin1 = satocsin(sa1);
428 sin2 = satocsin(sa2);
429
430 len = MIN(addrend, MIN(sin1->sin_len, sin2->sin_len));
431
432 if (len > addrofs &&
433 (rc = memcmp(&sin1->sin_addr, &sin2->sin_addr,
434 len - addrofs)) != 0)
435 return rc;
436
437 return sin1->sin_len - sin2->sin_len;
438 }
439