xref: /freebsd/sys/netinet/ip_encap.c (revision c4fde87a3236b500246d4fce76dbca2a3d64f73a) 
1 /*	$KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 itojun Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * Copyright (c) 2018 Andrey V. Elsukov <ae@FreeBSD.org>
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 /*
35  * My grandfather said that there's a devil inside tunnelling technology...
36  *
37  * We have surprisingly many protocols that want packets with IP protocol
38  * #4 or #41.  Here's a list of protocols that want protocol #41:
39  *	RFC1933 configured tunnel
40  *	RFC1933 automatic tunnel
41  *	RFC2401 IPsec tunnel
42  *	RFC2473 IPv6 generic packet tunnelling
43  *	RFC2529 6over4 tunnel
44  *	mobile-ip6 (uses RFC2473)
45  *	RFC3056 6to4 tunnel
46  *	isatap tunnel
47  * Here's a list of protocol that want protocol #4:
48  *	RFC1853 IPv4-in-IPv4 tunnelling
49  *	RFC2003 IPv4 encapsulation within IPv4
50  *	RFC2344 reverse tunnelling for mobile-ip4
51  *	RFC2401 IPsec tunnel
52  * Well, what can I say.  They impose different en/decapsulation mechanism
53  * from each other, so they need separate protocol handler.  The only one
54  * we can easily determine by protocol # is IPsec, which always has
55  * AH/ESP/IPComp header right after outer IP header.
56  *
57  * So, clearly good old protosw does not work for protocol #4 and #41.
58  * The code will let you match protocol via src/dst address pair.
59  */
60 
61 #include "opt_inet.h"
62 #include "opt_inet6.h"
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/eventhandler.h>
67 #include <sys/kernel.h>
68 #include <sys/lock.h>
69 #include <sys/malloc.h>
70 #include <sys/mutex.h>
71 #include <sys/mbuf.h>
72 #include <sys/errno.h>
73 #include <sys/socket.h>
74 #include <sys/sysctl.h>
75 
76 #include <net/if.h>
77 #include <net/if_var.h>
78 #include <net/vnet.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/ip.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_encap.h>
84 #include <netinet/in_rss.h>
85 
86 #ifdef INET6
87 #include <netinet/ip6.h>
88 #include <netinet6/ip6_var.h>
89 #include <netinet6/in6_rss.h>
90 #endif
91 
92 static MALLOC_DEFINE(M_NETADDR, "encap_export_host",
93     "Export host address structure");
94 
95 SYSCTL_DECL(_net_route);
96 VNET_DEFINE(bool, encap_hash_inbound) = false;
97 SYSCTL_BOOL(_net_route, OID_AUTO, hash_inbound, CTLFLAG_RW | CTLFLAG_VNET,
98     &VNET_NAME(encap_hash_inbound), false,
99     "Compute flowid for packets received over tunnel interfaces");
100 
101 struct encaptab {
102 	CK_LIST_ENTRY(encaptab) chain;
103 	int		proto;
104 	int		min_length;
105 	int		exact_match;
106 	void		*arg;
107 
108 	encap_lookup_t	lookup;
109 	encap_check_t	check;
110 	encap_input_t	input;
111 };
112 
113 struct srcaddrtab {
114 	CK_LIST_ENTRY(srcaddrtab) chain;
115 
116 	encap_srcaddr_t	srcaddr;
117 	void		*arg;
118 };
119 
120 CK_LIST_HEAD(encaptab_head, encaptab);
121 CK_LIST_HEAD(srcaddrtab_head, srcaddrtab);
122 #ifdef INET
123 static struct encaptab_head ipv4_encaptab = CK_LIST_HEAD_INITIALIZER();
124 static struct srcaddrtab_head ipv4_srcaddrtab = CK_LIST_HEAD_INITIALIZER();
125 #endif
126 #ifdef INET6
127 static struct encaptab_head ipv6_encaptab = CK_LIST_HEAD_INITIALIZER();
128 static struct srcaddrtab_head ipv6_srcaddrtab = CK_LIST_HEAD_INITIALIZER();
129 #endif
130 
131 static struct mtx encapmtx, srcaddrmtx;
132 MTX_SYSINIT(encapmtx, &encapmtx, "encapmtx", MTX_DEF);
133 MTX_SYSINIT(srcaddrmtx, &srcaddrmtx, "srcaddrmtx", MTX_DEF);
134 #define	ENCAP_WLOCK()		mtx_lock(&encapmtx)
135 #define	ENCAP_WUNLOCK()		mtx_unlock(&encapmtx)
136 #define	ENCAP_RLOCK_TRACKER	struct epoch_tracker encap_et
137 #define	ENCAP_RLOCK()		NET_EPOCH_ENTER(encap_et)
138 #define	ENCAP_RUNLOCK()		NET_EPOCH_EXIT(encap_et)
139 #define	ENCAP_WAIT()		NET_EPOCH_WAIT()
140 
141 #define	SRCADDR_WLOCK()		mtx_lock(&srcaddrmtx)
142 #define	SRCADDR_WUNLOCK()	mtx_unlock(&srcaddrmtx)
143 #define	SRCADDR_RLOCK_TRACKER	struct epoch_tracker srcaddr_et
144 #define	SRCADDR_RLOCK()		\
145     epoch_enter_preempt(net_epoch_preempt, &srcaddr_et)
146 #define	SRCADDR_RUNLOCK()	\
147     epoch_exit_preempt(net_epoch_preempt, &srcaddr_et)
148 #define	SRCADDR_WAIT()		epoch_wait_preempt(net_epoch_preempt)
149 
150 /*
151  * ifaddr_event_ext handler.
152  *
153  * Tunnelling interfaces may request the kernel to notify when
154  * some interface addresses appears or disappears. Usually tunnelling
155  * interface must use an address configured on the local machine as
156  * ingress address to be able receive datagramms and do not send
157  * spoofed packets.
158  */
159 static void
160 srcaddr_change_event(void *arg __unused, struct ifnet *ifp,
161     struct ifaddr *ifa, int event)
162 {
163 	SRCADDR_RLOCK_TRACKER;
164 	struct srcaddrtab_head *head;
165 	struct srcaddrtab *p;
166 
167 	/* Support for old ifaddr_event. */
168 	EVENTHANDLER_INVOKE(ifaddr_event, ifp);
169 
170 	switch (ifa->ifa_addr->sa_family) {
171 #ifdef INET
172 	case AF_INET:
173 		head = &ipv4_srcaddrtab;
174 		break;
175 #endif
176 #ifdef INET6
177 	case AF_INET6:
178 		head = &ipv6_srcaddrtab;
179 		break;
180 #endif
181 	default:
182 		/* ignore event */
183 		return;
184 	}
185 
186 	SRCADDR_RLOCK();
187 	CK_LIST_FOREACH(p, head, chain) {
188 		(*p->srcaddr)(p->arg, ifa->ifa_addr, event);
189 	}
190 	SRCADDR_RUNLOCK();
191 }
192 EVENTHANDLER_DEFINE(ifaddr_event_ext, srcaddr_change_event, NULL, 0);
193 
194 static struct srcaddrtab *
195 encap_register_srcaddr(struct srcaddrtab_head *head, encap_srcaddr_t func,
196     void *arg, int mflags)
197 {
198 	struct srcaddrtab *p, *tmp;
199 
200 	if (func == NULL)
201 		return (NULL);
202 	p = malloc(sizeof(*p), M_NETADDR, mflags);
203 	if (p == NULL)
204 		return (NULL);
205 	p->srcaddr = func;
206 	p->arg = arg;
207 
208 	SRCADDR_WLOCK();
209 	CK_LIST_FOREACH(tmp, head, chain) {
210 		if (func == tmp->srcaddr && arg == tmp->arg)
211 			break;
212 	}
213 	if (tmp == NULL)
214 		CK_LIST_INSERT_HEAD(head, p, chain);
215 	SRCADDR_WUNLOCK();
216 
217 	if (tmp != NULL) {
218 		free(p, M_NETADDR);
219 		p = tmp;
220 	}
221 	return (p);
222 }
223 
224 static int
225 encap_unregister_srcaddr(struct srcaddrtab_head *head,
226     const struct srcaddrtab *cookie)
227 {
228 	struct srcaddrtab *p;
229 
230 	SRCADDR_WLOCK();
231 	CK_LIST_FOREACH(p, head, chain) {
232 		if (p == cookie) {
233 			CK_LIST_REMOVE(p, chain);
234 			SRCADDR_WUNLOCK();
235 			SRCADDR_WAIT();
236 			free(p, M_NETADDR);
237 			return (0);
238 		}
239 	}
240 	SRCADDR_WUNLOCK();
241 	return (EINVAL);
242 }
243 
244 static struct encaptab *
245 encap_attach(struct encaptab_head *head, const struct encap_config *cfg,
246     void *arg, int mflags)
247 {
248 	struct encaptab *ep, *tmp;
249 
250 	if (cfg == NULL || cfg->input == NULL ||
251 	    (cfg->check == NULL && cfg->lookup == NULL) ||
252 	    (cfg->lookup != NULL && cfg->exact_match != ENCAP_DRV_LOOKUP) ||
253 	    (cfg->exact_match == ENCAP_DRV_LOOKUP && cfg->lookup == NULL))
254 		return (NULL);
255 
256 	ep = malloc(sizeof(*ep), M_NETADDR, mflags);
257 	if (ep == NULL)
258 		return (NULL);
259 
260 	ep->proto = cfg->proto;
261 	ep->min_length = cfg->min_length;
262 	ep->exact_match = cfg->exact_match;
263 	ep->arg = arg;
264 	ep->lookup = cfg->exact_match == ENCAP_DRV_LOOKUP ? cfg->lookup: NULL;
265 	ep->check = cfg->exact_match != ENCAP_DRV_LOOKUP ? cfg->check: NULL;
266 	ep->input = cfg->input;
267 
268 	ENCAP_WLOCK();
269 	CK_LIST_FOREACH(tmp, head, chain) {
270 		if (tmp->exact_match <= ep->exact_match)
271 			break;
272 	}
273 	if (tmp == NULL)
274 		CK_LIST_INSERT_HEAD(head, ep, chain);
275 	else
276 		CK_LIST_INSERT_BEFORE(tmp, ep, chain);
277 	ENCAP_WUNLOCK();
278 	return (ep);
279 }
280 
281 static int
282 encap_detach(struct encaptab_head *head, const struct encaptab *cookie)
283 {
284 	struct encaptab *ep;
285 
286 	ENCAP_WLOCK();
287 	CK_LIST_FOREACH(ep, head, chain) {
288 		if (ep == cookie) {
289 			CK_LIST_REMOVE(ep, chain);
290 			ENCAP_WUNLOCK();
291 			ENCAP_WAIT();
292 			free(ep, M_NETADDR);
293 			return (0);
294 		}
295 	}
296 	ENCAP_WUNLOCK();
297 	return (EINVAL);
298 }
299 
300 static int
301 encap_input(struct encaptab_head *head, struct mbuf *m, int off, int proto)
302 {
303 	ENCAP_RLOCK_TRACKER;
304 	struct encaptab *ep, *match;
305 	void *arg;
306 	int matchprio, ret;
307 
308 	match = NULL;
309 	matchprio = 0;
310 
311 	ENCAP_RLOCK();
312 	CK_LIST_FOREACH(ep, head, chain) {
313 		if (ep->proto >= 0 && ep->proto != proto)
314 			continue;
315 		if (ep->min_length > m->m_pkthdr.len)
316 			continue;
317 		if (ep->exact_match == ENCAP_DRV_LOOKUP)
318 			ret = (*ep->lookup)(m, off, proto, &arg);
319 		else
320 			ret = (*ep->check)(m, off, proto, ep->arg);
321 		if (ret <= 0)
322 			continue;
323 		if (ret > matchprio) {
324 			match = ep;
325 			if (ep->exact_match != ENCAP_DRV_LOOKUP)
326 				arg = ep->arg;
327 			/*
328 			 * No need to continue the search, we got the
329 			 * exact match.
330 			 */
331 			if (ret >= ep->exact_match)
332 				break;
333 			matchprio = ret;
334 		}
335 	}
336 
337 	if (match != NULL) {
338 		/* found a match, "match" has the best one */
339 		ret = (*match->input)(m, off, proto, arg);
340 		ENCAP_RUNLOCK();
341 		MPASS(ret == IPPROTO_DONE);
342 		return (IPPROTO_DONE);
343 	}
344 	ENCAP_RUNLOCK();
345 	return (0);
346 }
347 
348 #ifdef INET
349 const struct srcaddrtab *
350 ip_encap_register_srcaddr(encap_srcaddr_t func, void *arg, int mflags)
351 {
352 
353 	return (encap_register_srcaddr(&ipv4_srcaddrtab, func, arg, mflags));
354 }
355 
356 int
357 ip_encap_unregister_srcaddr(const struct srcaddrtab *cookie)
358 {
359 
360 	return (encap_unregister_srcaddr(&ipv4_srcaddrtab, cookie));
361 }
362 
363 const struct encaptab *
364 ip_encap_attach(const struct encap_config *cfg, void *arg, int mflags)
365 {
366 
367 	return (encap_attach(&ipv4_encaptab, cfg, arg, mflags));
368 }
369 
370 int
371 ip_encap_detach(const struct encaptab *cookie)
372 {
373 
374 	return (encap_detach(&ipv4_encaptab, cookie));
375 }
376 
377 int
378 encap4_input(struct mbuf **mp, int *offp, int proto)
379 {
380 
381 	if (encap_input(&ipv4_encaptab, *mp, *offp, proto) != IPPROTO_DONE)
382 		return (rip_input(mp, offp, proto));
383 	return (IPPROTO_DONE);
384 }
385 #endif /* INET */
386 
387 #ifdef INET6
388 const struct srcaddrtab *
389 ip6_encap_register_srcaddr(encap_srcaddr_t func, void *arg, int mflags)
390 {
391 
392 	return (encap_register_srcaddr(&ipv6_srcaddrtab, func, arg, mflags));
393 }
394 
395 int
396 ip6_encap_unregister_srcaddr(const struct srcaddrtab *cookie)
397 {
398 
399 	return (encap_unregister_srcaddr(&ipv6_srcaddrtab, cookie));
400 }
401 
402 const struct encaptab *
403 ip6_encap_attach(const struct encap_config *cfg, void *arg, int mflags)
404 {
405 
406 	return (encap_attach(&ipv6_encaptab, cfg, arg, mflags));
407 }
408 
409 int
410 ip6_encap_detach(const struct encaptab *cookie)
411 {
412 
413 	return (encap_detach(&ipv6_encaptab, cookie));
414 }
415 
416 int
417 encap6_input(struct mbuf **mp, int *offp, int proto)
418 {
419 
420 	if (encap_input(&ipv6_encaptab, *mp, *offp, proto) != IPPROTO_DONE)
421 		return (rip6_input(mp, offp, proto));
422 	return (IPPROTO_DONE);
423 }
424 #endif /* INET6 */
425 
426 /*
427  * Set the flowid of a decapsulated IP packet from its inner addresses.
428  *
429  * flowid of an incoming packet computed over the outer tunnel tuple,
430  * which is the same for every flow a peer. Hashing the inner addresses
431  * instead to make the mpath (and RSS, if enabled) work properly.
432  *
433  * Callers must check the V_encap_hash_inbound sysctl before calling.
434  */
435 bool
436 encap_set_flowid(struct mbuf *m, int af)
437 {
438 	uint32_t rss_hash, rss_type;
439 
440 	M_ASSERTPKTHDR(m);
441 
442 	switch (af) {
443 #ifdef INET
444 	case AF_INET: {
445 		if (rss_mbuf_software_hash_v4(m, 0, &rss_hash, &rss_type) != 0)
446 			return (false);
447 		break;
448 	}
449 #endif
450 #ifdef INET6
451 	case AF_INET6: {
452 		if (rss_mbuf_software_hash_v6(m, 0, &rss_hash, &rss_type) != 0)
453 			return (false);
454 		break;
455 	}
456 #endif
457 	default:
458 		return (false);
459 	}
460 
461 	m->m_pkthdr.flowid = rss_hash;
462 	M_HASHTYPE_SET(m, rss_type);
463 	return (true);
464 }
465