xref: /freebsd/sys/net/if_me.c (revision f5463265955b829775bbb32e1fd0bc11dafc36ce)
1 /*-
2  * Copyright (c) 2014, 2018 Andrey V. Elsukov <ae@FreeBSD.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  *
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  */
26 
27 #include <sys/param.h>
28 #include <sys/systm.h>
29 #include <sys/jail.h>
30 #include <sys/kernel.h>
31 #include <sys/lock.h>
32 #include <sys/malloc.h>
33 #include <sys/module.h>
34 #include <sys/mbuf.h>
35 #include <sys/priv.h>
36 #include <sys/proc.h>
37 #include <sys/socket.h>
38 #include <sys/sockio.h>
39 #include <sys/sx.h>
40 #include <sys/sysctl.h>
41 #include <sys/syslog.h>
42 
43 #include <net/bpf.h>
44 #include <net/ethernet.h>
45 #include <net/if.h>
46 #include <net/if_var.h>
47 #include <net/if_private.h>
48 #include <net/if_clone.h>
49 #include <net/if_types.h>
50 #include <net/netisr.h>
51 #include <net/vnet.h>
52 #include <net/route.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip_encap.h>
60 
61 #include <machine/in_cksum.h>
62 #include <security/mac/mac_framework.h>
63 
64 #define	MEMTU			(1500 - sizeof(struct mobhdr))
65 static const char mename[] = "me";
66 static MALLOC_DEFINE(M_IFME, mename, "Minimal Encapsulation for IP");
67 /* Minimal forwarding header RFC 2004 */
68 struct mobhdr {
69 	uint8_t		mob_proto;	/* protocol */
70 	uint8_t		mob_flags;	/* flags */
71 #define	MOB_FLAGS_SP	0x80		/* source present */
72 	uint16_t	mob_csum;	/* header checksum */
73 	struct in_addr	mob_dst;	/* original destination address */
74 	struct in_addr	mob_src;	/* original source addr (optional) */
75 } __packed;
76 
77 struct me_softc {
78 	struct ifnet		*me_ifp;
79 	u_int			me_fibnum;
80 	struct in_addr		me_src;
81 	struct in_addr		me_dst;
82 
83 	CK_LIST_ENTRY(me_softc) chain;
84 	CK_LIST_ENTRY(me_softc) srchash;
85 };
86 CK_LIST_HEAD(me_list, me_softc);
87 #define	ME2IFP(sc)		((sc)->me_ifp)
88 #define	ME_READY(sc)		((sc)->me_src.s_addr != 0)
89 #define	ME_RLOCK_TRACKER	struct epoch_tracker me_et
90 #define	ME_RLOCK()		epoch_enter_preempt(net_epoch_preempt, &me_et)
91 #define	ME_RUNLOCK()		epoch_exit_preempt(net_epoch_preempt, &me_et)
92 #define	ME_WAIT()		epoch_wait_preempt(net_epoch_preempt)
93 
94 #ifndef ME_HASH_SIZE
95 #define	ME_HASH_SIZE	(1 << 4)
96 #endif
97 VNET_DEFINE_STATIC(struct me_list *, me_hashtbl) = NULL;
98 VNET_DEFINE_STATIC(struct me_list *, me_srchashtbl) = NULL;
99 #define	V_me_hashtbl		VNET(me_hashtbl)
100 #define	V_me_srchashtbl		VNET(me_srchashtbl)
101 #define	ME_HASH(src, dst)	(V_me_hashtbl[\
102     me_hashval((src), (dst)) & (ME_HASH_SIZE - 1)])
103 #define	ME_SRCHASH(src)		(V_me_srchashtbl[\
104     fnv_32_buf(&(src), sizeof(src), FNV1_32_INIT) & (ME_HASH_SIZE - 1)])
105 
106 static struct sx me_ioctl_sx;
107 SX_SYSINIT(me_ioctl_sx, &me_ioctl_sx, "me_ioctl");
108 
109 static int	me_clone_create(struct if_clone *, int, caddr_t);
110 static void	me_clone_destroy(struct ifnet *);
111 VNET_DEFINE_STATIC(struct if_clone *, me_cloner);
112 #define	V_me_cloner	VNET(me_cloner)
113 
114 #ifdef VIMAGE
115 static void	me_reassign(struct ifnet *, struct vnet *, char *);
116 #endif
117 static void	me_qflush(struct ifnet *);
118 static int	me_transmit(struct ifnet *, struct mbuf *);
119 static int	me_ioctl(struct ifnet *, u_long, caddr_t);
120 static int	me_output(struct ifnet *, struct mbuf *,
121 		    const struct sockaddr *, struct route *);
122 static int	me_input(struct mbuf *, int, int, void *);
123 
124 static int	me_set_tunnel(struct me_softc *, in_addr_t, in_addr_t);
125 static void	me_delete_tunnel(struct me_softc *);
126 
127 SYSCTL_DECL(_net_link);
128 static SYSCTL_NODE(_net_link, IFT_TUNNEL, me, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
129     "Minimal Encapsulation for IP (RFC 2004)");
130 #ifndef MAX_ME_NEST
131 #define MAX_ME_NEST 1
132 #endif
133 
134 VNET_DEFINE_STATIC(int, max_me_nesting) = MAX_ME_NEST;
135 #define	V_max_me_nesting	VNET(max_me_nesting)
136 SYSCTL_INT(_net_link_me, OID_AUTO, max_nesting, CTLFLAG_RW | CTLFLAG_VNET,
137     &VNET_NAME(max_me_nesting), 0, "Max nested tunnels");
138 
139 static uint32_t
140 me_hashval(in_addr_t src, in_addr_t dst)
141 {
142 	uint32_t ret;
143 
144 	ret = fnv_32_buf(&src, sizeof(src), FNV1_32_INIT);
145 	return (fnv_32_buf(&dst, sizeof(dst), ret));
146 }
147 
148 static struct me_list *
149 me_hashinit(void)
150 {
151 	struct me_list *hash;
152 	int i;
153 
154 	hash = malloc(sizeof(struct me_list) * ME_HASH_SIZE,
155 	    M_IFME, M_WAITOK);
156 	for (i = 0; i < ME_HASH_SIZE; i++)
157 		CK_LIST_INIT(&hash[i]);
158 
159 	return (hash);
160 }
161 
162 static void
163 vnet_me_init(const void *unused __unused)
164 {
165 
166 	V_me_cloner = if_clone_simple(mename, me_clone_create,
167 	    me_clone_destroy, 0);
168 }
169 VNET_SYSINIT(vnet_me_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
170     vnet_me_init, NULL);
171 
172 static void
173 vnet_me_uninit(const void *unused __unused)
174 {
175 
176 	if (V_me_hashtbl != NULL) {
177 		free(V_me_hashtbl, M_IFME);
178 		V_me_hashtbl = NULL;
179 		ME_WAIT();
180 		free(V_me_srchashtbl, M_IFME);
181 	}
182 	if_clone_detach(V_me_cloner);
183 }
184 VNET_SYSUNINIT(vnet_me_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
185     vnet_me_uninit, NULL);
186 
187 static int
188 me_clone_create(struct if_clone *ifc, int unit, caddr_t params)
189 {
190 	struct me_softc *sc;
191 
192 	sc = malloc(sizeof(struct me_softc), M_IFME, M_WAITOK | M_ZERO);
193 	sc->me_fibnum = curthread->td_proc->p_fibnum;
194 	ME2IFP(sc) = if_alloc(IFT_TUNNEL);
195 	ME2IFP(sc)->if_softc = sc;
196 	if_initname(ME2IFP(sc), mename, unit);
197 
198 	ME2IFP(sc)->if_mtu = MEMTU;
199 	ME2IFP(sc)->if_flags = IFF_POINTOPOINT|IFF_MULTICAST;
200 	ME2IFP(sc)->if_output = me_output;
201 	ME2IFP(sc)->if_ioctl = me_ioctl;
202 	ME2IFP(sc)->if_transmit = me_transmit;
203 	ME2IFP(sc)->if_qflush = me_qflush;
204 #ifdef VIMAGE
205 	ME2IFP(sc)->if_reassign = me_reassign;
206 #endif
207 	ME2IFP(sc)->if_capabilities |= IFCAP_LINKSTATE;
208 	ME2IFP(sc)->if_capenable |= IFCAP_LINKSTATE;
209 	if_attach(ME2IFP(sc));
210 	bpfattach(ME2IFP(sc), DLT_NULL, sizeof(u_int32_t));
211 	return (0);
212 }
213 
214 #ifdef VIMAGE
215 static void
216 me_reassign(struct ifnet *ifp, struct vnet *new_vnet __unused,
217     char *unused __unused)
218 {
219 	struct me_softc *sc;
220 
221 	sx_xlock(&me_ioctl_sx);
222 	sc = ifp->if_softc;
223 	if (sc != NULL)
224 		me_delete_tunnel(sc);
225 	sx_xunlock(&me_ioctl_sx);
226 }
227 #endif /* VIMAGE */
228 
229 static void
230 me_clone_destroy(struct ifnet *ifp)
231 {
232 	struct me_softc *sc;
233 
234 	sx_xlock(&me_ioctl_sx);
235 	sc = ifp->if_softc;
236 	me_delete_tunnel(sc);
237 	bpfdetach(ifp);
238 	if_detach(ifp);
239 	ifp->if_softc = NULL;
240 	sx_xunlock(&me_ioctl_sx);
241 
242 	ME_WAIT();
243 	if_free(ifp);
244 	free(sc, M_IFME);
245 }
246 
247 static int
248 me_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
249 {
250 	struct ifreq *ifr = (struct ifreq *)data;
251 	struct sockaddr_in *src, *dst;
252 	struct me_softc *sc;
253 	int error;
254 
255 	switch (cmd) {
256 	case SIOCSIFMTU:
257 		if (ifr->ifr_mtu < 576)
258 			return (EINVAL);
259 		ifp->if_mtu = ifr->ifr_mtu;
260 		return (0);
261 	case SIOCSIFADDR:
262 		ifp->if_flags |= IFF_UP;
263 	case SIOCSIFFLAGS:
264 	case SIOCADDMULTI:
265 	case SIOCDELMULTI:
266 		return (0);
267 	}
268 	sx_xlock(&me_ioctl_sx);
269 	sc = ifp->if_softc;
270 	if (sc == NULL) {
271 		error = ENXIO;
272 		goto end;
273 	}
274 	error = 0;
275 	switch (cmd) {
276 	case SIOCSIFPHYADDR:
277 		src = &((struct in_aliasreq *)data)->ifra_addr;
278 		dst = &((struct in_aliasreq *)data)->ifra_dstaddr;
279 		if (src->sin_family != dst->sin_family ||
280 		    src->sin_family != AF_INET ||
281 		    src->sin_len != dst->sin_len ||
282 		    src->sin_len != sizeof(struct sockaddr_in)) {
283 			error = EINVAL;
284 			break;
285 		}
286 		if (src->sin_addr.s_addr == INADDR_ANY ||
287 		    dst->sin_addr.s_addr == INADDR_ANY) {
288 			error = EADDRNOTAVAIL;
289 			break;
290 		}
291 		error = me_set_tunnel(sc, src->sin_addr.s_addr,
292 		    dst->sin_addr.s_addr);
293 		break;
294 	case SIOCDIFPHYADDR:
295 		me_delete_tunnel(sc);
296 		break;
297 	case SIOCGIFPSRCADDR:
298 	case SIOCGIFPDSTADDR:
299 		if (!ME_READY(sc)) {
300 			error = EADDRNOTAVAIL;
301 			break;
302 		}
303 		src = (struct sockaddr_in *)&ifr->ifr_addr;
304 		memset(src, 0, sizeof(*src));
305 		src->sin_family = AF_INET;
306 		src->sin_len = sizeof(*src);
307 		switch (cmd) {
308 		case SIOCGIFPSRCADDR:
309 			src->sin_addr = sc->me_src;
310 			break;
311 		case SIOCGIFPDSTADDR:
312 			src->sin_addr = sc->me_dst;
313 			break;
314 		}
315 		error = prison_if(curthread->td_ucred, sintosa(src));
316 		if (error != 0)
317 			memset(src, 0, sizeof(*src));
318 		break;
319 	case SIOCGTUNFIB:
320 		ifr->ifr_fib = sc->me_fibnum;
321 		break;
322 	case SIOCSTUNFIB:
323 		if ((error = priv_check(curthread, PRIV_NET_ME)) != 0)
324 			break;
325 		if (ifr->ifr_fib >= rt_numfibs)
326 			error = EINVAL;
327 		else
328 			sc->me_fibnum = ifr->ifr_fib;
329 		break;
330 	default:
331 		error = EINVAL;
332 		break;
333 	}
334 end:
335 	sx_xunlock(&me_ioctl_sx);
336 	return (error);
337 }
338 
339 static int
340 me_lookup(const struct mbuf *m, int off, int proto, void **arg)
341 {
342 	const struct ip *ip;
343 	struct me_softc *sc;
344 
345 	if (V_me_hashtbl == NULL)
346 		return (0);
347 
348 	NET_EPOCH_ASSERT();
349 	ip = mtod(m, const struct ip *);
350 	CK_LIST_FOREACH(sc, &ME_HASH(ip->ip_dst.s_addr,
351 	    ip->ip_src.s_addr), chain) {
352 		if (sc->me_src.s_addr == ip->ip_dst.s_addr &&
353 		    sc->me_dst.s_addr == ip->ip_src.s_addr) {
354 			if ((ME2IFP(sc)->if_flags & IFF_UP) == 0)
355 				return (0);
356 			*arg = sc;
357 			return (ENCAP_DRV_LOOKUP);
358 		}
359 	}
360 	return (0);
361 }
362 
363 /*
364  * Check that ingress address belongs to local host.
365  */
366 static void
367 me_set_running(struct me_softc *sc)
368 {
369 
370 	if (in_localip(sc->me_src))
371 		ME2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING;
372 	else
373 		ME2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
374 }
375 
376 /*
377  * ifaddr_event handler.
378  * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
379  * source address spoofing.
380  */
381 static void
382 me_srcaddr(void *arg __unused, const struct sockaddr *sa,
383     int event __unused)
384 {
385 	const struct sockaddr_in *sin;
386 	struct me_softc *sc;
387 
388 	/* Check that VNET is ready */
389 	if (V_me_hashtbl == NULL)
390 		return;
391 
392 	NET_EPOCH_ASSERT();
393 	sin = (const struct sockaddr_in *)sa;
394 	CK_LIST_FOREACH(sc, &ME_SRCHASH(sin->sin_addr.s_addr), srchash) {
395 		if (sc->me_src.s_addr != sin->sin_addr.s_addr)
396 			continue;
397 		me_set_running(sc);
398 	}
399 }
400 
401 static int
402 me_set_tunnel(struct me_softc *sc, in_addr_t src, in_addr_t dst)
403 {
404 	struct epoch_tracker et;
405 	struct me_softc *tmp;
406 
407 	sx_assert(&me_ioctl_sx, SA_XLOCKED);
408 
409 	if (V_me_hashtbl == NULL) {
410 		V_me_hashtbl = me_hashinit();
411 		V_me_srchashtbl = me_hashinit();
412 	}
413 
414 	if (sc->me_src.s_addr == src && sc->me_dst.s_addr == dst)
415 		return (0);
416 
417 	CK_LIST_FOREACH(tmp, &ME_HASH(src, dst), chain) {
418 		if (tmp == sc)
419 			continue;
420 		if (tmp->me_src.s_addr == src &&
421 		    tmp->me_dst.s_addr == dst)
422 			return (EADDRNOTAVAIL);
423 	}
424 
425 	me_delete_tunnel(sc);
426 	sc->me_dst.s_addr = dst;
427 	sc->me_src.s_addr = src;
428 	CK_LIST_INSERT_HEAD(&ME_HASH(src, dst), sc, chain);
429 	CK_LIST_INSERT_HEAD(&ME_SRCHASH(src), sc, srchash);
430 
431 	NET_EPOCH_ENTER(et);
432 	me_set_running(sc);
433 	NET_EPOCH_EXIT(et);
434 	if_link_state_change(ME2IFP(sc), LINK_STATE_UP);
435 	return (0);
436 }
437 
438 static void
439 me_delete_tunnel(struct me_softc *sc)
440 {
441 
442 	sx_assert(&me_ioctl_sx, SA_XLOCKED);
443 	if (ME_READY(sc)) {
444 		CK_LIST_REMOVE(sc, chain);
445 		CK_LIST_REMOVE(sc, srchash);
446 		ME_WAIT();
447 
448 		sc->me_src.s_addr = 0;
449 		sc->me_dst.s_addr = 0;
450 		ME2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
451 		if_link_state_change(ME2IFP(sc), LINK_STATE_DOWN);
452 	}
453 }
454 
455 static uint16_t
456 me_in_cksum(uint16_t *p, int nwords)
457 {
458 	uint32_t sum = 0;
459 
460 	while (nwords-- > 0)
461 		sum += *p++;
462 	sum = (sum >> 16) + (sum & 0xffff);
463 	sum += (sum >> 16);
464 	return (~sum);
465 }
466 
467 static int
468 me_input(struct mbuf *m, int off, int proto, void *arg)
469 {
470 	struct me_softc *sc = arg;
471 	struct mobhdr *mh;
472 	struct ifnet *ifp;
473 	struct ip *ip;
474 	int hlen;
475 
476 	NET_EPOCH_ASSERT();
477 
478 	ifp = ME2IFP(sc);
479 	/* checks for short packets */
480 	hlen = sizeof(struct mobhdr);
481 	if (m->m_pkthdr.len < sizeof(struct ip) + hlen)
482 		hlen -= sizeof(struct in_addr);
483 	if (m->m_len < sizeof(struct ip) + hlen)
484 		m = m_pullup(m, sizeof(struct ip) + hlen);
485 	if (m == NULL)
486 		goto drop;
487 	mh = (struct mobhdr *)mtodo(m, sizeof(struct ip));
488 	/* check for wrong flags */
489 	if (mh->mob_flags & (~MOB_FLAGS_SP)) {
490 		m_freem(m);
491 		goto drop;
492 	}
493 	if (mh->mob_flags) {
494 	       if (hlen != sizeof(struct mobhdr)) {
495 			m_freem(m);
496 			goto drop;
497 	       }
498 	} else
499 		hlen = sizeof(struct mobhdr) - sizeof(struct in_addr);
500 	/* check mobile header checksum */
501 	if (me_in_cksum((uint16_t *)mh, hlen / sizeof(uint16_t)) != 0) {
502 		m_freem(m);
503 		goto drop;
504 	}
505 #ifdef MAC
506 	mac_ifnet_create_mbuf(ifp, m);
507 #endif
508 	ip = mtod(m, struct ip *);
509 	ip->ip_dst = mh->mob_dst;
510 	ip->ip_p = mh->mob_proto;
511 	ip->ip_sum = 0;
512 	ip->ip_len = htons(m->m_pkthdr.len - hlen);
513 	if (mh->mob_flags)
514 		ip->ip_src = mh->mob_src;
515 	memmove(mtodo(m, hlen), ip, sizeof(struct ip));
516 	m_adj(m, hlen);
517 	m_clrprotoflags(m);
518 	m->m_pkthdr.rcvif = ifp;
519 	m->m_pkthdr.csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
520 	M_SETFIB(m, ifp->if_fib);
521 	hlen = AF_INET;
522 	BPF_MTAP2(ifp, &hlen, sizeof(hlen), m);
523 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
524 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
525 	if ((ifp->if_flags & IFF_MONITOR) != 0)
526 		m_freem(m);
527 	else
528 		netisr_dispatch(NETISR_IP, m);
529 	return (IPPROTO_DONE);
530 drop:
531 	if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
532 	return (IPPROTO_DONE);
533 }
534 
535 static int
536 me_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
537    struct route *ro)
538 {
539 	uint32_t af;
540 
541 	if (dst->sa_family == AF_UNSPEC)
542 		bcopy(dst->sa_data, &af, sizeof(af));
543 	else
544 		af = RO_GET_FAMILY(ro, dst);
545 	m->m_pkthdr.csum_data = af;
546 	return (ifp->if_transmit(ifp, m));
547 }
548 
549 #define	MTAG_ME	1414491977
550 static int
551 me_transmit(struct ifnet *ifp, struct mbuf *m)
552 {
553 	ME_RLOCK_TRACKER;
554 	struct mobhdr mh;
555 	struct me_softc *sc;
556 	struct ip *ip;
557 	uint32_t af;
558 	int error, hlen, plen;
559 
560 	ME_RLOCK();
561 #ifdef MAC
562 	error = mac_ifnet_check_transmit(ifp, m);
563 	if (error != 0)
564 		goto drop;
565 #endif
566 	error = ENETDOWN;
567 	sc = ifp->if_softc;
568 	if (sc == NULL || !ME_READY(sc) ||
569 	    (ifp->if_flags & IFF_MONITOR) != 0 ||
570 	    (ifp->if_flags & IFF_UP) == 0 ||
571 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
572 	    (error = if_tunnel_check_nesting(ifp, m, MTAG_ME,
573 		V_max_me_nesting)) != 0) {
574 		m_freem(m);
575 		goto drop;
576 	}
577 	af = m->m_pkthdr.csum_data;
578 	if (af != AF_INET) {
579 		error = EAFNOSUPPORT;
580 		m_freem(m);
581 		goto drop;
582 	}
583 	if (m->m_len < sizeof(struct ip))
584 		m = m_pullup(m, sizeof(struct ip));
585 	if (m == NULL) {
586 		error = ENOBUFS;
587 		goto drop;
588 	}
589 	ip = mtod(m, struct ip *);
590 	/* Fragmented datagramms shouldn't be encapsulated */
591 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
592 		error = EINVAL;
593 		m_freem(m);
594 		goto drop;
595 	}
596 	mh.mob_proto = ip->ip_p;
597 	mh.mob_src = ip->ip_src;
598 	mh.mob_dst = ip->ip_dst;
599 	if (in_hosteq(sc->me_src, ip->ip_src)) {
600 		hlen = sizeof(struct mobhdr) - sizeof(struct in_addr);
601 		mh.mob_flags = 0;
602 	} else {
603 		hlen = sizeof(struct mobhdr);
604 		mh.mob_flags = MOB_FLAGS_SP;
605 	}
606 	BPF_MTAP2(ifp, &af, sizeof(af), m);
607 	plen = m->m_pkthdr.len;
608 	ip->ip_src = sc->me_src;
609 	ip->ip_dst = sc->me_dst;
610 	m->m_flags &= ~(M_BCAST|M_MCAST);
611 	M_SETFIB(m, sc->me_fibnum);
612 	M_PREPEND(m, hlen, M_NOWAIT);
613 	if (m == NULL) {
614 		error = ENOBUFS;
615 		goto drop;
616 	}
617 	if (m->m_len < sizeof(struct ip) + hlen)
618 		m = m_pullup(m, sizeof(struct ip) + hlen);
619 	if (m == NULL) {
620 		error = ENOBUFS;
621 		goto drop;
622 	}
623 	memmove(mtod(m, void *), mtodo(m, hlen), sizeof(struct ip));
624 	ip = mtod(m, struct ip *);
625 	ip->ip_len = htons(m->m_pkthdr.len);
626 	ip->ip_p = IPPROTO_MOBILE;
627 	ip->ip_sum = 0;
628 	mh.mob_csum = 0;
629 	mh.mob_csum = me_in_cksum((uint16_t *)&mh, hlen / sizeof(uint16_t));
630 	bcopy(&mh, mtodo(m, sizeof(struct ip)), hlen);
631 	error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
632 drop:
633 	if (error)
634 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
635 	else {
636 		if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
637 		if_inc_counter(ifp, IFCOUNTER_OBYTES, plen);
638 	}
639 	ME_RUNLOCK();
640 	return (error);
641 }
642 
643 static void
644 me_qflush(struct ifnet *ifp __unused)
645 {
646 
647 }
648 
649 static const struct srcaddrtab *me_srcaddrtab = NULL;
650 static const struct encaptab *ecookie = NULL;
651 static const struct encap_config me_encap_cfg = {
652 	.proto = IPPROTO_MOBILE,
653 	.min_length = sizeof(struct ip) + sizeof(struct mobhdr) -
654 	    sizeof(in_addr_t),
655 	.exact_match = ENCAP_DRV_LOOKUP,
656 	.lookup = me_lookup,
657 	.input = me_input
658 };
659 
660 static int
661 memodevent(module_t mod, int type, void *data)
662 {
663 
664 	switch (type) {
665 	case MOD_LOAD:
666 		me_srcaddrtab = ip_encap_register_srcaddr(me_srcaddr,
667 		    NULL, M_WAITOK);
668 		ecookie = ip_encap_attach(&me_encap_cfg, NULL, M_WAITOK);
669 		break;
670 	case MOD_UNLOAD:
671 		ip_encap_detach(ecookie);
672 		ip_encap_unregister_srcaddr(me_srcaddrtab);
673 		break;
674 	default:
675 		return (EOPNOTSUPP);
676 	}
677 	return (0);
678 }
679 
680 static moduledata_t me_mod = {
681 	"if_me",
682 	memodevent,
683 	0
684 };
685 
686 DECLARE_MODULE(if_me, me_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
687 MODULE_VERSION(if_me, 1);
688