xref: /netbsd-src/sys/net/if_mpls.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: if_mpls.c,v 1.3 2010/06/27 13:39:11 kefren Exp $ */
2 
3 /*
4  * Copyright (c) 2010 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Mihai Chelaru <kefren@NetBSD.org>
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  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __KERNEL_RCSID(0, "$NetBSD: if_mpls.c,v 1.3 2010/06/27 13:39:11 kefren Exp $");
34 
35 #include "opt_inet.h"
36 #include "opt_mpls.h"
37 
38 #include <sys/param.h>
39 
40 #include <sys/errno.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/sysctl.h>
44 
45 #include <net/bpf.h>
46 #include <net/if.h>
47 #include <net/if_types.h>
48 #include <net/netisr.h>
49 #include <net/route.h>
50 
51 #ifdef INET
52 #include <netinet/in.h>
53 #include <netinet/in_systm.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip.h>
56 #endif
57 
58 #ifdef INET6
59 #include <netinet/ip6.h>
60 #include <netinet6/in6_var.h>
61 #include <netinet6/ip6_var.h>
62 #endif
63 
64 #include <netmpls/mpls.h>
65 #include <netmpls/mpls_var.h>
66 
67 #include "if_mpls.h"
68 
69 void ifmplsattach(int);
70 
71 static int mpls_clone_create(struct if_clone *, int);
72 static int mpls_clone_destroy(struct ifnet *);
73 
74 static struct if_clone mpls_if_cloner =
75 	IF_CLONE_INITIALIZER("mpls", mpls_clone_create, mpls_clone_destroy);
76 
77 
78 static void mpls_input(struct ifnet *, struct mbuf *);
79 static int mpls_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
80 	struct rtentry *);
81 static int mpls_ioctl(struct ifnet *, u_long, void *);
82 static int mpls_send_frame(struct mbuf *, struct ifnet *, struct rtentry *);
83 static int mpls_lse(struct mbuf *);
84 
85 #ifdef INET
86 static int mpls_unlabel_inet(struct mbuf *);
87 static struct mbuf *mpls_label_inet(struct mbuf *, union mpls_shim *);
88 #endif
89 
90 #ifdef INET6
91 static int mpls_unlabel_inet6(struct mbuf *);
92 static struct mbuf *mpls_label_inet6(struct mbuf *, union mpls_shim *);
93 #endif
94 
95 static struct mbuf *mpls_prepend_shim(struct mbuf *, union mpls_shim *);
96 
97 extern int mpls_defttl, mpls_mapttl_inet, mpls_mapttl_inet6, mpls_icmp_respond,
98 	mpls_forwarding, mpls_accept, mpls_mapprec_inet, mpls_mapclass_inet6;
99 
100 /* ARGSUSED */
101 void
102 ifmplsattach(int count)
103 {
104 	if_clone_attach(&mpls_if_cloner);
105 }
106 
107 static int
108 mpls_clone_create(struct if_clone *ifc, int unit)
109 {
110 	struct mpls_softc *sc;
111 
112 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
113 
114 	if_initname(&sc->sc_if, ifc->ifc_name, unit);
115 	sc->sc_if.if_softc = sc;
116 	sc->sc_if.if_type = IFT_MPLS;
117 	sc->sc_if.if_addrlen = 0;
118 	sc->sc_if.if_hdrlen = sizeof(union mpls_shim);
119 	sc->sc_if.if_dlt = DLT_NULL;
120 	sc->sc_if.if_mtu = 1500;
121 	sc->sc_if.if_flags = 0;
122 	sc->sc_if.if_input = mpls_input;
123 	sc->sc_if.if_output = mpls_output;
124 	sc->sc_if.if_ioctl = mpls_ioctl;
125 
126 	if_attach(&sc->sc_if);
127 	if_alloc_sadl(&sc->sc_if);
128 	bpf_attach(&sc->sc_if, DLT_NULL, sizeof(uint32_t));
129 	return 0;
130 }
131 
132 static int
133 mpls_clone_destroy(struct ifnet *ifp)
134 {
135 	int s;
136 
137 	bpf_detach(ifp);
138 
139 	s = splnet();
140 	if_detach(ifp);
141 	splx(s);
142 
143 	free(ifp->if_softc, M_DEVBUF);
144 	return 0;
145 }
146 
147 static void
148 mpls_input(struct ifnet *ifp, struct mbuf *m)
149 {
150 #if 0
151 	/*
152 	 * TODO - kefren
153 	 * I'd love to unshim the packet, guess family
154 	 * and pass it to bpf
155 	 */
156 	bpf_mtap_af(ifp, AF_MPLS, m);
157 #endif
158 
159 	mpls_lse(m);
160 }
161 
162 void
163 mplsintr(void)
164 {
165 	struct mbuf *m;
166 	int s;
167 
168 	while (!IF_IS_EMPTY(&mplsintrq)) {
169 		s = splnet();
170 		IF_DEQUEUE(&mplsintrq, m);
171 		splx(s);
172 
173 		if (!m)
174 			return;
175 
176 		if (((m->m_flags & M_PKTHDR) == 0) ||
177 		    (m->m_pkthdr.rcvif == 0))
178 			panic("mplsintr(): no pkthdr or rcvif");
179 
180 #ifdef MBUFTRACE
181 		m_claimm(m, &mpls_owner);
182 #endif
183 		mpls_input(m->m_pkthdr.rcvif, m);
184 	}
185 }
186 
187 /*
188  * prepend shim and deliver
189  */
190 static int
191 mpls_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct rtentry *rt)
192 {
193 	union mpls_shim mh;
194 	struct rtentry *rt1;
195 	int err;
196 
197 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) {
198 		m_freem(m);
199 		return ENETDOWN;
200 	}
201 
202 	if (rt_gettag(rt) == NULL || rt_gettag(rt)->sa_family != AF_MPLS) {
203 		m_freem(m);
204 		return EINVAL;
205 	}
206 
207 	bpf_mtap_af(ifp, dst->sa_family, m);
208 
209 	mh.s_addr=MPLS_GETSADDR(rt);
210 	mh.shim.bos=1;
211 	mh.shim.exp=0;
212 	mh.shim.ttl=mpls_defttl;
213 
214 	switch(dst->sa_family) {
215 #ifdef INET
216 	case AF_INET:
217 		m = mpls_label_inet(m, &mh);
218 		break;
219 #endif
220 #ifdef INET6
221 	case AF_INET6:
222 		m = mpls_label_inet6(m, &mh);
223 		break;
224 #endif
225 	default:
226 		m = mpls_prepend_shim(m, &mh);
227 		break;
228 	}
229 
230 	if (m == NULL) {
231 		IF_DROP(&ifp->if_snd);
232 		ifp->if_oerrors++;
233 		return ENOBUFS;
234 	}
235 
236 	ifp->if_opackets++;
237 	ifp->if_obytes += m->m_pkthdr.len;
238 
239 	if ((rt1=rtalloc1(rt->rt_gateway, 1)) == NULL) {
240 		m_freem(m);
241 		return EHOSTUNREACH;
242 	}
243 
244 	err = mpls_send_frame(m, rt1->rt_ifp, rt);
245 	RTFREE(rt1);
246 	return err;
247 }
248 
249 static int
250 mpls_ioctl(struct ifnet *ifp, u_long cmd, void *data)
251 {
252 	int error = 0, s = splnet();
253 	struct ifreq *ifr = data;
254 
255 	switch(cmd) {
256 	case SIOCINITIFADDR:
257 		ifp->if_flags |= IFF_UP | IFF_RUNNING;
258 		break;
259 	case SIOCSIFMTU:
260 		if (ifr != NULL && ifr->ifr_mtu < 576) {
261 			error = EINVAL;
262 			break;
263 		}
264 		/* FALLTHROUGH */
265 	case SIOCGIFMTU:
266 		if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
267 			error = 0;
268 		break;
269 	case SIOCSIFFLAGS:
270 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
271 			break;
272 		if (ifp->if_flags & IFF_UP)
273 			ifp->if_flags |= IFF_RUNNING;
274 		break;
275 	default:
276 		error = ifioctl_common(ifp, cmd, data);
277 		break;
278 	}
279 	splx(s);
280 	return error;
281 }
282 
283 /*
284  * MPLS Label Switch Engine
285  */
286 static int
287 mpls_lse(struct mbuf *m)
288 {
289 	struct sockaddr_mpls dst;
290 	union mpls_shim tshim, *htag;
291 	struct rtentry *rt = NULL;
292 	int error = ENOBUFS;
293 
294 	if (m->m_len < sizeof(union mpls_shim) &&
295 	    (m = m_pullup(m, sizeof(union mpls_shim))) == NULL)
296 		goto done;
297 
298 	dst.smpls_len = sizeof(struct sockaddr_mpls);
299 	dst.smpls_family = AF_MPLS;
300 	dst.smpls_addr.s_addr = ntohl(mtod(m, union mpls_shim *)->s_addr);
301 
302 	/* Check if we're accepting MPLS Frames */
303 	error = EINVAL;
304 	if (!mpls_accept)
305 		goto done;
306 
307 	/* TTL decrement */
308 	if ((m = mpls_ttl_dec(m)) == NULL)
309 		goto done;
310 
311 	if (dst.smpls_addr.shim.label <= MPLS_LABEL_RESMAX) {
312 		/* Don't swap reserved labels */
313 		switch (dst.smpls_addr.shim.label) {
314 #ifdef INET
315 		case MPLS_LABEL_IPV4NULL:
316 			/* Pop shim and push mbuf to IP stack */
317 			if (dst.smpls_addr.shim.bos)
318 				error = mpls_unlabel_inet(m);
319 			break;
320 #endif
321 #ifdef INET6
322 		case MPLS_LABEL_IPV6NULL:
323 			/* Pop shim and push mbuf to IPv6 stack */
324 			if (dst.smpls_addr.shim.bos)
325 				error = mpls_unlabel_inet6(m);
326 			break;
327 #endif
328 		case MPLS_LABEL_RTALERT:	/* Yeah, I'm all alerted */
329 		case MPLS_LABEL_IMPLNULL:	/* This is logical only */
330 		default:			/* Rest are not allowed */
331 			break;
332 		}
333 		goto done;
334 	}
335 
336 	/* Check if we should do MPLS forwarding */
337 	error = EHOSTUNREACH;
338 	if (!mpls_forwarding)
339 		goto done;
340 
341 	/* Get a route to dst */
342 	dst.smpls_addr.shim.ttl =
343 	    dst.smpls_addr.shim.bos =
344 	    dst.smpls_addr.shim.exp = 0;
345 	dst.smpls_addr.s_addr = htonl(dst.smpls_addr.s_addr);
346 	if ((rt = rtalloc1((const struct sockaddr*)&dst, 1)) == NULL)
347 		goto done;
348 
349 	/* MPLS packet with no MPLS tagged route ? */
350 	if ((rt->rt_flags & RTF_GATEWAY) == 0 ||
351 	     rt_gettag(rt) == NULL ||
352 	     rt_gettag(rt)->sa_family != AF_MPLS)
353 		goto done;
354 
355 	tshim.s_addr = MPLS_GETSADDR(rt);
356 
357 	/* Swap labels */
358 	if ((m->m_len < sizeof(union mpls_shim)) &&
359 	    (m = m_pullup(m, sizeof(union mpls_shim))) == 0) {
360 		error = ENOBUFS;
361 		goto done;
362 	}
363 
364 	/* Replace only the label */
365 	htag = mtod(m, union mpls_shim *);
366 	htag->s_addr = ntohl(htag->s_addr);
367 	htag->shim.label = tshim.shim.label;
368 	htag->s_addr = htonl(htag->s_addr);
369 
370 	error = mpls_send_frame(m, rt->rt_ifp, rt);
371 
372 done:
373 	if (error != 0 && m != NULL)
374 		m_freem(m);
375 	if (rt != NULL)
376 		RTFREE(rt);
377 
378 	return error;
379 }
380 
381 static int
382 mpls_send_frame(struct mbuf *m, struct ifnet *ifp, struct rtentry *rt)
383 {
384 	union mpls_shim msh;
385 
386 	if ((rt->rt_flags & RTF_GATEWAY) == 0)
387 		return EHOSTUNREACH;
388 
389 	rt->rt_use++;
390 
391 	msh.s_addr = MPLS_GETSADDR(rt);
392 	if (msh.shim.label == MPLS_LABEL_IMPLNULL) {
393 		m_adj(m, sizeof(union mpls_shim));
394 		m->m_pkthdr.csum_flags = 0;
395 	}
396 
397 	switch(ifp->if_type) {
398 	/* only these two are supported for now */
399 	case IFT_ETHER:
400 	case IFT_TUNNEL:
401 		return (*ifp->if_output)(ifp, m, rt->rt_gateway, rt);
402 	case IFT_LOOP:
403 		break;
404 	default:
405 		return ENETUNREACH;
406 	}
407 	return 0;
408 }
409 
410 
411 
412 #ifdef INET
413 static int
414 mpls_unlabel_inet(struct mbuf *m)
415 {
416 	int s, iphlen;
417 	struct ip *iph;
418 	union mpls_shim *ms;
419 	struct ifqueue *inq;
420 
421 	if (mpls_mapttl_inet || mpls_mapprec_inet) {
422 
423 		/* get shim info */
424 		ms = mtod(m, union mpls_shim *);
425 		ms->s_addr = ntohl(ms->s_addr);
426 
427 		/* and get rid of it */
428 		m_adj(m, sizeof(union mpls_shim));
429 
430 		/* get ip header */
431 		if (m->m_len < sizeof (struct ip) &&
432 		    (m = m_pullup(m, sizeof(struct ip))) == NULL)
433 			return ENOBUFS;
434 		iph = mtod(m, struct ip *);
435 		iphlen = iph->ip_hl << 2;
436 
437 		/* get it all */
438 		if (m->m_len < iphlen) {
439 			if ((m = m_pullup(m, iphlen)) == NULL)
440 				return ENOBUFS;
441 			iph = mtod(m, struct ip *);
442 		}
443 
444 		/* check ipsum */
445 		if (in_cksum(m, iphlen) != 0) {
446 			m_freem(m);
447 			return EINVAL;
448 		}
449 
450 		/* set IP ttl from MPLS ttl */
451 		if (mpls_mapttl_inet)
452 			iph->ip_ttl = ms->shim.ttl;
453 
454 		/* set IP Precedence from MPLS Exp */
455 		if (mpls_mapprec_inet) {
456 			iph->ip_tos = (iph->ip_tos << 3) >> 3;
457 			iph->ip_tos |= ms->shim.exp << 5;
458 		}
459 
460 		/* reset ipsum because we modified TTL and TOS */
461 		iph->ip_sum = 0;
462 		iph->ip_sum = in_cksum(m, iphlen);
463 	} else
464 		m_adj(m, sizeof(union mpls_shim));
465 
466 	/* Put it on IP queue */
467 	inq = &ipintrq;
468 	s = splnet();
469 	if (IF_QFULL(inq)) {
470 		IF_DROP(inq);
471 		splx(s);
472 		m_freem(m);
473 		return ENOBUFS;
474 	}
475 	IF_ENQUEUE(inq, m);
476 	splx(s);
477 	schednetisr(NETISR_IP);
478 
479 	return 0;
480 }
481 
482 /*
483  * Prepend MPLS label
484  */
485 static struct mbuf *
486 mpls_label_inet(struct mbuf *m, union mpls_shim *ms)
487 {
488 	struct ip *iphdr;
489 
490 	if (mpls_mapttl_inet || mpls_mapprec_inet) {
491 		if ((m->m_len < sizeof(struct ip)) &&
492 		    (m = m_pullup(m, sizeof(struct ip))) == 0)
493 			return NULL;
494 		iphdr = mtod(m, struct ip *);
495 
496 		/* Map TTL */
497 		if (mpls_mapttl_inet)
498 			ms->shim.ttl = iphdr->ip_ttl;
499 
500 		/* Copy IP precedence to EXP */
501 		if (mpls_mapprec_inet)
502 			ms->shim.exp = ((u_int8_t)iphdr->ip_tos) >> 5;
503 	}
504 
505 	if ((m = mpls_prepend_shim(m, ms)) == NULL)
506 		return NULL;
507 
508 	return m;
509 }
510 
511 #endif	/* INET */
512 
513 #ifdef INET6
514 
515 static int
516 mpls_unlabel_inet6(struct mbuf *m)
517 {
518 	struct ip6_hdr *ip6hdr;
519 	union mpls_shim ms;
520 	struct ifqueue *inq;
521 	int s;
522 
523 	/* TODO: mapclass */
524 	if (mpls_mapttl_inet6) {
525 		ms.s_addr = ntohl(mtod(m, union mpls_shim *)->s_addr);
526 		m_adj(m, sizeof(union mpls_shim));
527 
528 		if (m->m_len < sizeof (struct ip6_hdr) &&
529 		    (m = m_pullup(m, sizeof(struct ip6_hdr))) == 0)
530 			return ENOBUFS;
531 		ip6hdr = mtod(m, struct ip6_hdr *);
532 
533 		/* Because we just decremented this in mpls_lse */
534 		ip6hdr->ip6_hlim = ms.shim.ttl + 1;
535 	} else
536 		m_adj(m, sizeof(union mpls_shim));
537 
538 	/* Put it back on IPv6 stack */
539 	schednetisr(NETISR_IPV6);
540 	inq = &ip6intrq;
541 	s = splnet();
542 	if (IF_QFULL(inq)) {
543 		IF_DROP(inq);
544 		splx(s);
545 		m_freem(m);
546 		return ENOBUFS;
547 	}
548 
549 	IF_ENQUEUE(inq, m);
550 	splx(s);
551 
552 	return 0;
553 }
554 
555 static struct mbuf *
556 mpls_label_inet6(struct mbuf *m, union mpls_shim *ms)
557 {
558 	struct ip6_hdr *ip6h;
559 
560 	if (mpls_mapttl_inet6 || mpls_mapclass_inet6) {
561 		if (m->m_len < sizeof(struct ip6_hdr) &&
562 		    (m = m_pullup(m, sizeof(struct ip6_hdr))) == 0)
563 			return NULL;
564 		ip6h = mtod(m, struct ip6_hdr *);
565 
566 		if (mpls_mapttl_inet6)
567 			ms->shim.ttl = ip6h->ip6_hlim;
568 
569 		if (mpls_mapclass_inet6)
570 			ms->shim.exp = ip6h->ip6_vfc << 1 >> 5;
571 	}
572 
573 	if ((m = mpls_prepend_shim(m, ms)) == NULL)
574 		return NULL;
575 
576 	return m;
577 }
578 
579 #endif	/* INET6 */
580 
581 static struct mbuf *
582 mpls_prepend_shim(struct mbuf *m, union mpls_shim *ms)
583 {
584 	union mpls_shim *shim;
585 
586 	M_PREPEND(m, sizeof(*ms), M_DONTWAIT);
587 	if (m == NULL)
588 		return NULL;
589 
590 	if (m->m_len < sizeof(union mpls_shim) &&
591 	    (m = m_pullup(m, sizeof(union mpls_shim))) == 0)
592 		return NULL;
593 
594 	shim = mtod(m, union mpls_shim *);
595 
596 	memcpy(shim, ms, sizeof(*shim));
597 	shim->s_addr = htonl(shim->s_addr);
598 
599 	return m;
600 }
601