xref: /dragonfly/sys/netinet/if_ether.c (revision ab709bfb)
1 /*
2  * Copyright (c) 2004, 2005 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Jeffrey M. Hsu.
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 DragonFly Project nor the names of its
16  *    contributors may be used to endorse or promote products derived
17  *    from this software without specific, prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
66  * $FreeBSD: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $
67  * $DragonFly: src/sys/netinet/if_ether.c,v 1.59 2008/11/22 11:03:35 sephe Exp $
68  */
69 
70 /*
71  * Ethernet address resolution protocol.
72  * TODO:
73  *	add "inuse/lock" bit (or ref. count) along with valid bit
74  */
75 
76 #include "opt_inet.h"
77 #include "opt_carp.h"
78 
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/queue.h>
82 #include <sys/sysctl.h>
83 #include <sys/systm.h>
84 #include <sys/mbuf.h>
85 #include <sys/malloc.h>
86 #include <sys/socket.h>
87 #include <sys/syslog.h>
88 #include <sys/lock.h>
89 
90 #include <net/if.h>
91 #include <net/if_dl.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #include <net/if_llc.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/if_ether.h>
100 
101 #include <sys/thread2.h>
102 #include <sys/msgport2.h>
103 #include <net/netmsg2.h>
104 
105 #ifdef CARP
106 #include <netinet/ip_carp.h>
107 #endif
108 
109 #define SIN(s) ((struct sockaddr_in *)s)
110 #define SDL(s) ((struct sockaddr_dl *)s)
111 
112 SYSCTL_DECL(_net_link_ether);
113 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
114 
115 /* timer values */
116 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */
117 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */
118 static int arpt_down = 20;	/* once declared down, don't send for 20 sec */
119 
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW,
121 	   &arpt_prune, 0, "");
122 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
123 	   &arpt_keep, 0, "");
124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW,
125 	   &arpt_down, 0, "");
126 
127 #define	rt_expire	rt_rmx.rmx_expire
128 
129 struct llinfo_arp {
130 	LIST_ENTRY(llinfo_arp) la_le;
131 	struct	rtentry *la_rt;
132 	struct	mbuf *la_hold;	/* last packet until resolved/timeout */
133 	struct	lwkt_port *la_msgport; /* last packet's msgport */
134 	u_short	la_preempt;	/* countdown for pre-expiry arps */
135 	u_short	la_asked;	/* #times we QUERIED following expiration */
136 };
137 
138 static	LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
139 
140 static int	arp_maxtries = 5;
141 static int	useloopback = 1; /* use loopback interface for local traffic */
142 static int	arp_proxyall = 0;
143 
144 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
145 	   &arp_maxtries, 0, "");
146 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
147 	   &useloopback, 0, "");
148 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
149 	   &arp_proxyall, 0, "");
150 
151 static int	arp_mpsafe = 0;
152 TUNABLE_INT("net.link.ether.inet.arp_mpsafe", &arp_mpsafe);
153 
154 static void	arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
155 static void	arprequest(struct ifnet *, const struct in_addr *,
156 			   const struct in_addr *, const u_char *);
157 static void	arprequest_async(struct ifnet *, const struct in_addr *,
158 				 const struct in_addr *, const u_char *);
159 static void	arpintr(struct netmsg *);
160 static void	arptfree(struct llinfo_arp *);
161 static void	arptimer(void *);
162 static struct llinfo_arp *
163 		arplookup(in_addr_t, boolean_t, boolean_t, boolean_t);
164 #ifdef INET
165 static void	in_arpinput(struct mbuf *);
166 #endif
167 
168 static struct callout	arptimer_ch[MAXCPU];
169 
170 /*
171  * Timeout routine.  Age arp_tab entries periodically.
172  */
173 /* ARGSUSED */
174 static void
175 arptimer(void *ignored_arg)
176 {
177 	struct llinfo_arp *la, *nla;
178 
179 	crit_enter();
180 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
181 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
182 			arptfree(la);
183 	}
184 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
185 	crit_exit();
186 }
187 
188 /*
189  * Parallel to llc_rtrequest.
190  */
191 static void
192 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
193 {
194 	struct sockaddr *gate = rt->rt_gateway;
195 	struct llinfo_arp *la = rt->rt_llinfo;
196 
197 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
198 	static boolean_t arpinit_done[MAXCPU];
199 
200 	if (!arpinit_done[mycpuid]) {
201 		arpinit_done[mycpuid] = TRUE;
202 		callout_init(&arptimer_ch[mycpuid]);
203 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
204 	}
205 	if (rt->rt_flags & RTF_GATEWAY)
206 		return;
207 
208 	switch (req) {
209 	case RTM_ADD:
210 		/*
211 		 * XXX: If this is a manually added route to interface
212 		 * such as older version of routed or gated might provide,
213 		 * restore cloning bit.
214 		 */
215 		if (!(rt->rt_flags & RTF_HOST) &&
216 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
217 			rt->rt_flags |= RTF_CLONING;
218 		if (rt->rt_flags & RTF_CLONING) {
219 			/*
220 			 * Case 1: This route should come from a route to iface.
221 			 */
222 			rt_setgate(rt, rt_key(rt),
223 				   (struct sockaddr *)&null_sdl,
224 				   RTL_DONTREPORT);
225 			gate = rt->rt_gateway;
226 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
227 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
228 			rt->rt_expire = time_second;
229 			break;
230 		}
231 		/* Announce a new entry if requested. */
232 		if (rt->rt_flags & RTF_ANNOUNCE) {
233 			arprequest_async(rt->rt_ifp,
234 			    &SIN(rt_key(rt))->sin_addr,
235 			    &SIN(rt_key(rt))->sin_addr,
236 			    LLADDR(SDL(gate)));
237 		}
238 		/*FALLTHROUGH*/
239 	case RTM_RESOLVE:
240 		if (gate->sa_family != AF_LINK ||
241 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
242 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
243 			break;
244 		}
245 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
246 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
247 		if (la != NULL)
248 			break; /* This happens on a route change */
249 		/*
250 		 * Case 2:  This route may come from cloning, or a manual route
251 		 * add with a LL address.
252 		 */
253 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
254 		rt->rt_llinfo = la;
255 		if (la == NULL) {
256 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
257 			break;
258 		}
259 		bzero(la, sizeof *la);
260 		la->la_rt = rt;
261 		rt->rt_flags |= RTF_LLINFO;
262 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
263 
264 #ifdef INET
265 		/*
266 		 * This keeps the multicast addresses from showing up
267 		 * in `arp -a' listings as unresolved.  It's not actually
268 		 * functional.  Then the same for broadcast.
269 		 */
270 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
271 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
272 					       LLADDR(SDL(gate)));
273 			SDL(gate)->sdl_alen = 6;
274 			rt->rt_expire = 0;
275 		}
276 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
277 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
278 			       rt->rt_ifp->if_addrlen);
279 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
280 			rt->rt_expire = 0;
281 		}
282 #endif
283 
284 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
285 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
286 			/*
287 			 * This test used to be
288 			 *	if (loif.if_flags & IFF_UP)
289 			 * It allowed local traffic to be forced
290 			 * through the hardware by configuring the
291 			 * loopback down.  However, it causes problems
292 			 * during network configuration for boards
293 			 * that can't receive packets they send.  It
294 			 * is now necessary to clear "useloopback" and
295 			 * remove the route to force traffic out to
296 			 * the hardware.
297 			 */
298 			rt->rt_expire = 0;
299 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
300 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
301 			if (useloopback)
302 				rt->rt_ifp = loif;
303 		}
304 		break;
305 
306 	case RTM_DELETE:
307 		if (la == NULL)
308 			break;
309 		LIST_REMOVE(la, la_le);
310 		rt->rt_llinfo = NULL;
311 		rt->rt_flags &= ~RTF_LLINFO;
312 		if (la->la_hold != NULL)
313 			m_freem(la->la_hold);
314 		Free(la);
315 		break;
316 	}
317 }
318 
319 static struct mbuf *
320 arpreq_alloc(struct ifnet *ifp, const struct in_addr *sip,
321 	     const struct in_addr *tip, const u_char *enaddr)
322 {
323 	struct mbuf *m;
324 	struct arphdr *ah;
325 	u_short ar_hrd;
326 
327 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
328 		return NULL;
329 	m->m_pkthdr.rcvif = NULL;
330 
331 	switch (ifp->if_type) {
332 	case IFT_ETHER:
333 		/*
334 		 * This may not be correct for types not explicitly
335 		 * listed, but this is our best guess
336 		 */
337 	default:
338 		ar_hrd = htons(ARPHRD_ETHER);
339 
340 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
341 		m->m_pkthdr.len = m->m_len;
342 		MH_ALIGN(m, m->m_len);
343 
344 		ah = mtod(m, struct arphdr *);
345 		break;
346 	}
347 
348 	ah->ar_hrd = ar_hrd;
349 	ah->ar_pro = htons(ETHERTYPE_IP);
350 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
351 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
352 	ah->ar_op = htons(ARPOP_REQUEST);
353 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
354 	memset(ar_tha(ah), 0, ah->ar_hln);
355 	memcpy(ar_spa(ah), sip, ah->ar_pln);
356 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
357 
358 	return m;
359 }
360 
361 static void
362 arpreq_send(struct ifnet *ifp, struct mbuf *m)
363 {
364 	struct sockaddr sa;
365 	struct ether_header *eh;
366 
367 	switch (ifp->if_type) {
368 	case IFT_ETHER:
369 		/*
370 		 * This may not be correct for types not explicitly
371 		 * listed, but this is our best guess
372 		 */
373 	default:
374 		eh = (struct ether_header *)sa.sa_data;
375 		/* if_output() will not swap */
376 		eh->ether_type = htons(ETHERTYPE_ARP);
377 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
378 		break;
379 	}
380 
381 	sa.sa_family = AF_UNSPEC;
382 	sa.sa_len = sizeof(sa);
383 	ifp->if_output(ifp, m, &sa, NULL);
384 }
385 
386 static void
387 arpreq_send_handler(struct netmsg *nmsg)
388 {
389 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
390 	struct ifnet *ifp = nmsg->nm_lmsg.u.ms_resultp;
391 
392 	arpreq_send(ifp, m);
393 	/* nmsg was embedded in the mbuf, do not reply! */
394 }
395 
396 /*
397  * Broadcast an ARP request. Caller specifies:
398  *	- arp header source ip address
399  *	- arp header target ip address
400  *	- arp header source ethernet address
401  *
402  * NOTE: Caller MUST NOT hold ifp's serializer
403  */
404 static void
405 arprequest(struct ifnet *ifp, const struct in_addr *sip,
406 	   const struct in_addr *tip, const u_char *enaddr)
407 {
408 	struct mbuf *m;
409 
410 	m = arpreq_alloc(ifp, sip, tip, enaddr);
411 	if (m == NULL)
412 		return;
413 	arpreq_send(ifp, m);
414 }
415 
416 /*
417  * Same as arprequest(), except:
418  * - Caller is allowed to hold ifp's serializer
419  * - Network output is done in TDF_NETWORK kernel thread
420  */
421 static void
422 arprequest_async(struct ifnet *ifp, const struct in_addr *sip,
423 		 const struct in_addr *tip, const u_char *enaddr)
424 {
425 	struct mbuf *m;
426 	struct netmsg_packet *pmsg;
427 
428 	m = arpreq_alloc(ifp, sip, tip, enaddr);
429 	if (m == NULL)
430 		return;
431 
432 	pmsg = &m->m_hdr.mh_netmsg;
433 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
434 		    arpreq_send_handler);
435 	pmsg->nm_packet = m;
436 	pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = ifp;
437 
438 	lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->nm_netmsg.nm_lmsg);
439 }
440 
441 /*
442  * Resolve an IP address into an ethernet address.  If success,
443  * desten is filled in.  If there is no entry in arptab,
444  * set one up and broadcast a request for the IP address.
445  * Hold onto this mbuf and resend it once the address
446  * is finally resolved.  A return value of 1 indicates
447  * that desten has been filled in and the packet should be sent
448  * normally; a 0 return indicates that the packet has been
449  * taken over here, either now or for later transmission.
450  */
451 int
452 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
453 	   struct sockaddr *dst, u_char *desten)
454 {
455 	struct rtentry *rt;
456 	struct llinfo_arp *la = NULL;
457 	struct sockaddr_dl *sdl;
458 
459 	if (m->m_flags & M_BCAST) {	/* broadcast */
460 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
461 		return (1);
462 	}
463 	if (m->m_flags & M_MCAST) {/* multicast */
464 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
465 		return (1);
466 	}
467 	if (rt0 != NULL) {
468 		if (rt_llroute(dst, rt0, &rt) != 0) {
469 			m_freem(m);
470 			return 0;
471 		}
472 		la = rt->rt_llinfo;
473 	}
474 	if (la == NULL) {
475 		la = arplookup(SIN(dst)->sin_addr.s_addr,
476 			       TRUE, RTL_REPORTMSG, FALSE);
477 		if (la != NULL)
478 			rt = la->la_rt;
479 	}
480 	if (la == NULL || rt == NULL) {
481 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
482 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
483 		    rt ? "rt" : "");
484 		m_freem(m);
485 		return (0);
486 	}
487 	sdl = SDL(rt->rt_gateway);
488 	/*
489 	 * Check the address family and length is valid, the address
490 	 * is resolved; otherwise, try to resolve.
491 	 */
492 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
493 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
494 		/*
495 		 * If entry has an expiry time and it is approaching,
496 		 * see if we need to send an ARP request within this
497 		 * arpt_down interval.
498 		 */
499 		if ((rt->rt_expire != 0) &&
500 		    (time_second + la->la_preempt > rt->rt_expire)) {
501 			arprequest(ifp,
502 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
503 				   &SIN(dst)->sin_addr,
504 				   IF_LLADDR(ifp));
505 			la->la_preempt--;
506 		}
507 
508 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
509 		return 1;
510 	}
511 	/*
512 	 * If ARP is disabled on this interface, stop.
513 	 * XXX
514 	 * Probably should not allocate empty llinfo struct if we are
515 	 * not going to be sending out an arp request.
516 	 */
517 	if (ifp->if_flags & IFF_NOARP) {
518 		m_freem(m);
519 		return (0);
520 	}
521 	/*
522 	 * There is an arptab entry, but no ethernet address
523 	 * response yet.  Replace the held mbuf with this
524 	 * latest one.
525 	 */
526 	if (la->la_hold != NULL)
527 		m_freem(la->la_hold);
528 	la->la_hold = m;
529 	la->la_msgport = curnetport;
530 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
531 		rt->rt_flags &= ~RTF_REJECT;
532 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
533 			rt->rt_expire = time_second;
534 			if (la->la_asked++ < arp_maxtries) {
535 				arprequest(ifp,
536 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
537 					   &SIN(dst)->sin_addr,
538 					   IF_LLADDR(ifp));
539 			} else {
540 				rt->rt_flags |= RTF_REJECT;
541 				rt->rt_expire += arpt_down;
542 				la->la_asked = 0;
543 				la->la_preempt = arp_maxtries;
544 			}
545 		}
546 	}
547 	return (0);
548 }
549 
550 /*
551  * Common length and type checks are done here,
552  * then the protocol-specific routine is called.
553  */
554 static void
555 arpintr(struct netmsg *msg)
556 {
557 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
558 	struct arphdr *ar;
559 	u_short ar_hrd;
560 
561 	if (m->m_len < sizeof(struct arphdr) &&
562 	    (m = m_pullup(m, sizeof(struct arphdr))) == NULL) {
563 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
564 		return;
565 	}
566 	ar = mtod(m, struct arphdr *);
567 
568 	ar_hrd = ntohs(ar->ar_hrd);
569 	if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) {
570 		log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n",
571 		    (unsigned char *)&ar->ar_hrd, "");
572 		m_freem(m);
573 		return;
574 	}
575 
576 	if (m->m_pkthdr.len < arphdr_len(ar)) {
577 		if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
578 			log(LOG_ERR, "arp: runt packet\n");
579 			return;
580 		}
581 		ar = mtod(m, struct arphdr *);
582 	}
583 
584 	switch (ntohs(ar->ar_pro)) {
585 #ifdef INET
586 	case ETHERTYPE_IP:
587 		in_arpinput(m);
588 		return;
589 #endif
590 	}
591 	m_freem(m);
592 	/* msg was embedded in the mbuf, do not reply! */
593 }
594 
595 #ifdef INET
596 /*
597  * ARP for Internet protocols on 10 Mb/s Ethernet.
598  * Algorithm is that given in RFC 826.
599  * In addition, a sanity check is performed on the sender
600  * protocol address, to catch impersonators.
601  * We no longer handle negotiations for use of trailer protocol:
602  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
603  * along with IP replies if we wanted trailers sent to us,
604  * and also sent them in response to IP replies.
605  * This allowed either end to announce the desire to receive
606  * trailer packets.
607  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
608  * but formerly didn't normally send requests.
609  */
610 static int log_arp_wrong_iface = 1;
611 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
612 	   &log_arp_wrong_iface, 0,
613 	   "log arp packets arriving on the wrong interface");
614 
615 static void
616 arp_hold_output(struct netmsg *nmsg)
617 {
618 	struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet;
619 	struct rtentry *rt;
620 	struct ifnet *ifp;
621 
622 	rt = nmsg->nm_lmsg.u.ms_resultp;
623 	ifp = m->m_pkthdr.rcvif;
624 	m->m_pkthdr.rcvif = NULL;
625 
626 	ifp->if_output(ifp, m, rt_key(rt), rt);
627 
628 	/* Drop the reference count bumped by the sender */
629 	RTFREE(rt);
630 
631 	/* nmsg was embedded in the mbuf, do not reply! */
632 }
633 
634 static void
635 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
636 		 boolean_t generate_report, boolean_t dologging)
637 {
638 	struct arphdr *ah = mtod(m, struct arphdr *);
639 	struct ifnet *ifp = m->m_pkthdr.rcvif;
640 	struct llinfo_arp *la;
641 	struct sockaddr_dl *sdl;
642 	struct rtentry *rt;
643 
644 	la = arplookup(saddr, create, generate_report, FALSE);
645 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
646 		struct in_addr isaddr = { saddr };
647 
648 		/* the following is not an error when doing bridging */
649 		if (rt->rt_ifp != ifp) {
650 			if (dologging && log_arp_wrong_iface) {
651 				log(LOG_ERR,
652 				    "arp: %s is on %s "
653 				    "but got reply from %*D on %s\n",
654 				    inet_ntoa(isaddr),
655 				    rt->rt_ifp->if_xname,
656 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
657 				    ifp->if_xname);
658 			}
659 			return;
660 		}
661 		if (sdl->sdl_alen &&
662 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
663 			if (rt->rt_expire != 0) {
664 				if (dologging) {
665 			    		log(LOG_INFO,
666 			    		"arp: %s moved from %*D to %*D on %s\n",
667 			    		inet_ntoa(isaddr),
668 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
669 			    		":", ifp->if_addrlen,
670 			    		(u_char *)ar_sha(ah), ":",
671 			    		ifp->if_xname);
672 				}
673 			} else {
674 				if (dologging) {
675 					log(LOG_ERR,
676 					"arp: %*D attempts to modify "
677 					"permanent entry for %s on %s\n",
678 					ifp->if_addrlen, (u_char *)ar_sha(ah),
679 					":", inet_ntoa(isaddr), ifp->if_xname);
680 				}
681 				return;
682 			}
683 		}
684 		/*
685 		 * sanity check for the address length.
686 		 * XXX this does not work for protocols with variable address
687 		 * length. -is
688 		 */
689 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
690 			log(LOG_WARNING,
691 			    "arp from %*D: new addr len %d, was %d",
692 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
693 			    ah->ar_hln, sdl->sdl_alen);
694 		}
695 		if (ifp->if_addrlen != ah->ar_hln) {
696 			if (dologging) {
697 				log(LOG_WARNING,
698 				"arp from %*D: addr len: new %d, i/f %d "
699 				"(ignored)",
700 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
701 				ah->ar_hln, ifp->if_addrlen);
702 			}
703 			return;
704 		}
705 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
706 		if (rt->rt_expire != 0)
707 			rt->rt_expire = time_second + arpt_keep;
708 		rt->rt_flags &= ~RTF_REJECT;
709 		la->la_asked = 0;
710 		la->la_preempt = arp_maxtries;
711 
712 		/*
713 		 * This particular cpu might have been holding an mbuf
714 		 * pending ARP resolution.  If so, transmit the mbuf now.
715 		 */
716 		if (la->la_hold != NULL) {
717 			struct mbuf *m = la->la_hold;
718 			struct lwkt_port *port = la->la_msgport;
719 			struct netmsg_packet *pmsg;
720 
721 			la->la_hold = NULL;
722 			la->la_msgport = NULL;
723 
724 			m_adj(m, sizeof(struct ether_header));
725 
726 			/*
727 			 * Make sure that this rtentry will not be freed
728 			 * before the packet is processed on the target
729 			 * msgport.  The reference count will be dropped
730 			 * in the handler associated with this packet.
731 			 */
732 			rt->rt_refcnt++;
733 
734 			pmsg = &m->m_hdr.mh_netmsg;
735 			netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport,
736 				    MSGF_MPSAFE | MSGF_PRIORITY,
737 				    arp_hold_output);
738 			pmsg->nm_packet = m;
739 
740 			/* Record necessary information */
741 			m->m_pkthdr.rcvif = ifp;
742 			pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = rt;
743 
744 			lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
745 		}
746 	}
747 }
748 
749 #ifdef SMP
750 
751 struct netmsg_arp_update {
752 	struct netmsg	netmsg;
753 	struct mbuf	*m;
754 	in_addr_t	saddr;
755 	boolean_t	create;
756 };
757 
758 static void arp_update_msghandler(struct netmsg *);
759 
760 #endif
761 
762 /*
763  * Called from arpintr() - this routine is run from a single cpu.
764  */
765 static void
766 in_arpinput(struct mbuf *m)
767 {
768 	struct arphdr *ah;
769 	struct ifnet *ifp = m->m_pkthdr.rcvif;
770 	struct ether_header *eh;
771 	struct rtentry *rt;
772 	struct ifaddr_container *ifac;
773 	struct in_ifaddr_container *iac;
774 	struct in_ifaddr *ia = NULL;
775 	struct sockaddr sa;
776 	struct in_addr isaddr, itaddr, myaddr;
777 #ifdef SMP
778 	struct netmsg_arp_update msg;
779 #endif
780 	uint8_t *enaddr = NULL;
781 	int op;
782 	int req_len;
783 
784 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
785 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
786 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
787 		return;
788 	}
789 
790 	ah = mtod(m, struct arphdr *);
791 	op = ntohs(ah->ar_op);
792 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
793 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
794 
795 	myaddr.s_addr = INADDR_ANY;
796 #ifdef CARP
797 	if (ifp->if_carp != NULL) {
798 		get_mplock();
799 		if (ifp->if_carp != NULL &&
800 		    carp_iamatch(ifp->if_carp, &itaddr, &isaddr, &enaddr)) {
801 			rel_mplock();
802 			myaddr = itaddr;
803 			goto match;
804 		}
805 		rel_mplock();
806 	}
807 #endif
808 
809 	/*
810 	 * Check both target and sender IP addresses:
811 	 *
812 	 * If we receive the packet on the interface owning the address,
813 	 * then accept the address.
814 	 *
815 	 * For a bridge, we accept the address if the receive interface and
816 	 * the interface owning the address are on the same bridge.
817 	 * (This will change slightly when we have clusters of interfaces).
818 	 */
819 	LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) {
820 		ia = iac->ia;
821 
822 		/* Skip all ia's which don't match */
823 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
824 			continue;
825 #ifdef CARP
826 		if (ia->ia_ifp->if_type == IFT_CARP)
827 			continue;
828 #endif
829 		if (ia->ia_ifp == ifp)
830 			goto match;
831 
832 		if (ifp->if_bridge && ia->ia_ifp &&
833 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
834 			goto match;
835 	}
836 	LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) {
837 		ia = iac->ia;
838 
839 		/* Skip all ia's which don't match */
840 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
841 			continue;
842 #ifdef CARP
843 		if (ia->ia_ifp->if_type == IFT_CARP)
844 			continue;
845 #endif
846 		if (ia->ia_ifp == ifp)
847 			goto match;
848 
849 		if (ifp->if_bridge && ia->ia_ifp &&
850 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
851 			goto match;
852 	}
853 	/*
854 	 * No match, use the first inet address on the receive interface
855 	 * as a dummy address for the rest of the function.
856 	 */
857 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
858 		struct ifaddr *ifa = ifac->ifa;
859 
860 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
861 			ia = ifatoia(ifa);
862 			goto match;
863 		}
864 	}
865 	/*
866 	 * If we got here, we didn't find any suitable interface,
867 	 * so drop the packet.
868 	 */
869 	m_freem(m);
870 	return;
871 
872 match:
873 	if (!enaddr)
874 		enaddr = (uint8_t *)IF_LLADDR(ifp);
875 	if (myaddr.s_addr == INADDR_ANY)
876 		myaddr = ia->ia_addr.sin_addr;
877 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
878 		m_freem(m);	/* it's from me, ignore it. */
879 		return;
880 	}
881 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
882 		log(LOG_ERR,
883 		    "arp: link address is broadcast for IP address %s!\n",
884 		    inet_ntoa(isaddr));
885 		m_freem(m);
886 		return;
887 	}
888 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
889 		log(LOG_ERR,
890 		   "arp: %*D is using my IP address %s!\n",
891 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
892 		   inet_ntoa(isaddr));
893 		itaddr = myaddr;
894 		goto reply;
895 	}
896 #ifdef SMP
897 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
898 		    arp_update_msghandler);
899 	msg.m = m;
900 	msg.saddr = isaddr.s_addr;
901 	msg.create = (itaddr.s_addr == myaddr.s_addr);
902 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
903 #else
904 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
905 			 RTL_REPORTMSG, TRUE);
906 #endif
907 reply:
908 	if (op != ARPOP_REQUEST) {
909 		m_freem(m);
910 		return;
911 	}
912 	if (itaddr.s_addr == myaddr.s_addr) {
913 		/* I am the target */
914 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
915 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
916 	} else {
917 		struct llinfo_arp *la;
918 
919 		la = arplookup(itaddr.s_addr, FALSE, RTL_DONTREPORT, SIN_PROXY);
920 		if (la == NULL) {
921 			struct sockaddr_in sin;
922 
923 			if (!arp_proxyall) {
924 				m_freem(m);
925 				return;
926 			}
927 
928 			bzero(&sin, sizeof sin);
929 			sin.sin_family = AF_INET;
930 			sin.sin_len = sizeof sin;
931 			sin.sin_addr = itaddr;
932 
933 			rt = rtpurelookup((struct sockaddr *)&sin);
934 			if (rt == NULL) {
935 				m_freem(m);
936 				return;
937 			}
938 			--rt->rt_refcnt;
939 			/*
940 			 * Don't send proxies for nodes on the same interface
941 			 * as this one came out of, or we'll get into a fight
942 			 * over who claims what Ether address.
943 			 */
944 			if (rt->rt_ifp == ifp) {
945 				m_freem(m);
946 				return;
947 			}
948 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
949 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
950 #ifdef DEBUG_PROXY
951 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
952 #endif
953 		} else {
954 			struct sockaddr_dl *sdl;
955 
956 			rt = la->la_rt;
957 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
958 			sdl = SDL(rt->rt_gateway);
959 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
960 		}
961 	}
962 
963 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
964 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
965 	ah->ar_op = htons(ARPOP_REPLY);
966 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
967 	switch (ifp->if_type) {
968 	case IFT_ETHER:
969 	/*
970 	 * May not be correct for types not explictly
971 	 * listed, but it is our best guess.
972 	 */
973 	default:
974 		eh = (struct ether_header *)sa.sa_data;
975 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
976 		eh->ether_type = htons(ETHERTYPE_ARP);
977 		break;
978 	}
979 	sa.sa_family = AF_UNSPEC;
980 	sa.sa_len = sizeof sa;
981 	ifp->if_output(ifp, m, &sa, NULL);
982 }
983 
984 #ifdef SMP
985 
986 static void
987 arp_update_msghandler(struct netmsg *netmsg)
988 {
989 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
990 	int nextcpu;
991 
992 	/*
993 	 * This message handler will be called on all of the CPUs,
994 	 * however, we only need to generate rtmsg on CPU0.
995 	 */
996 	arp_update_oncpu(msg->m, msg->saddr, msg->create,
997 			 mycpuid == 0 ? RTL_REPORTMSG : RTL_DONTREPORT,
998 			 mycpuid == 0);
999 
1000 	nextcpu = mycpuid + 1;
1001 	if (nextcpu < ncpus)
1002 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
1003 	else
1004 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
1005 }
1006 
1007 #endif	/* SMP */
1008 
1009 #endif	/* INET */
1010 
1011 /*
1012  * Free an arp entry.  If the arp entry is actively referenced or represents
1013  * a static entry we only clear it back to an unresolved state, otherwise
1014  * we destroy the entry entirely.
1015  *
1016  * Note that static entries are created when route add ... -interface is used
1017  * to create an interface route to a (direct) destination.
1018  */
1019 static void
1020 arptfree(struct llinfo_arp *la)
1021 {
1022 	struct rtentry *rt = la->la_rt;
1023 	struct sockaddr_dl *sdl;
1024 
1025 	if (rt == NULL)
1026 		panic("arptfree");
1027 	sdl = SDL(rt->rt_gateway);
1028 	if (sdl != NULL &&
1029 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
1030 	     (rt->rt_flags & RTF_STATIC))) {
1031 		sdl->sdl_alen = 0;
1032 		la->la_preempt = la->la_asked = 0;
1033 		rt->rt_flags &= ~RTF_REJECT;
1034 		return;
1035 	}
1036 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
1037 }
1038 
1039 /*
1040  * Lookup or enter a new address in arptab.
1041  */
1042 static struct llinfo_arp *
1043 arplookup(in_addr_t addr, boolean_t create, boolean_t generate_report,
1044 	  boolean_t proxy)
1045 {
1046 	struct rtentry *rt;
1047 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
1048 	const char *why = NULL;
1049 
1050 	sin.sin_addr.s_addr = addr;
1051 	sin.sin_other = proxy ? SIN_PROXY : 0;
1052 	if (create) {
1053 		rt = _rtlookup((struct sockaddr *)&sin,
1054 			       generate_report, RTL_DOCLONE);
1055 	} else {
1056 		rt = rtpurelookup((struct sockaddr *)&sin);
1057 	}
1058 	if (rt == NULL)
1059 		return (NULL);
1060 	rt->rt_refcnt--;
1061 
1062 	if (rt->rt_flags & RTF_GATEWAY)
1063 		why = "host is not on local network";
1064 	else if (!(rt->rt_flags & RTF_LLINFO))
1065 		why = "could not allocate llinfo";
1066 	else if (rt->rt_gateway->sa_family != AF_LINK)
1067 		why = "gateway route is not ours";
1068 
1069 	if (why) {
1070 		if (create) {
1071 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
1072 			    inet_ntoa(sin.sin_addr), why);
1073 		}
1074 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
1075 			/* No references to this route.  Purge it. */
1076 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1077 				  rt_mask(rt), rt->rt_flags, NULL);
1078 		}
1079 		return (NULL);
1080 	}
1081 	return (rt->rt_llinfo);
1082 }
1083 
1084 void
1085 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1086 {
1087 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) {
1088 		arprequest_async(ifp, &IA_SIN(ifa)->sin_addr,
1089 				 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1090 	}
1091 	ifa->ifa_rtrequest = arp_rtrequest;
1092 	ifa->ifa_flags |= RTF_CLONING;
1093 }
1094 
1095 void
1096 arp_iainit(struct ifnet *ifp, const struct in_addr *addr, const u_char *enaddr)
1097 {
1098 	if (addr->s_addr != INADDR_ANY)
1099 		arprequest_async(ifp, addr, addr, enaddr);
1100 }
1101 
1102 static void
1103 arp_init(void)
1104 {
1105 	uint32_t flags;
1106 	int cpu;
1107 
1108 	for (cpu = 0; cpu < ncpus2; cpu++)
1109 		LIST_INIT(&llinfo_arp_list[cpu]);
1110 
1111 	if (arp_mpsafe) {
1112 		flags = NETISR_FLAG_MPSAFE;
1113 		kprintf("arp: MPSAFE\n");
1114 	} else {
1115 		flags = NETISR_FLAG_NOTMPSAFE;
1116 	}
1117 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr, flags);
1118 }
1119 
1120 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1121