xref: /dflybsd-src/sys/netinet/if_ether.c (revision c03f08f36c0122373cab0d3832d20e8c3256afc1)
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 1.42 2007/08/27 16:15:42 hasso 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 
89 #include <net/if.h>
90 #include <net/if_dl.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 #include <net/netisr.h>
94 #include <net/if_llc.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 #include <netinet/if_ether.h>
99 
100 #include <sys/thread2.h>
101 #include <sys/msgport2.h>
102 #include <net/netmsg2.h>
103 
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 	u_short	la_preempt;	/* countdown for pre-expiry arps */
134 	u_short	la_asked;	/* #times we QUERIED following expiration */
135 };
136 
137 static	LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU];
138 
139 static int	arp_maxtries = 5;
140 static int	useloopback = 1; /* use loopback interface for local traffic */
141 static int	arp_proxyall = 0;
142 
143 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
144 	   &arp_maxtries, 0, "");
145 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
146 	   &useloopback, 0, "");
147 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
148 	   &arp_proxyall, 0, "");
149 
150 void 		arprequest_acces(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,  u_char *enaddr);
151 static void	arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
152 static void	arprequest (struct ifnet *,
153 			struct in_addr *, struct in_addr *, u_char *);
154 static void	arpintr(struct netmsg *);
155 static void	arptfree (struct llinfo_arp *);
156 static void	arptimer (void *);
157 static struct llinfo_arp
158 		*arplookup (in_addr_t addr, boolean_t create, boolean_t proxy);
159 #ifdef INET
160 static void	in_arpinput (struct mbuf *);
161 #endif
162 
163 static struct callout	arptimer_ch[MAXCPU];
164 
165 /*
166  * Timeout routine.  Age arp_tab entries periodically.
167  */
168 /* ARGSUSED */
169 static void
170 arptimer(void *ignored_arg)
171 {
172 	struct llinfo_arp *la, *nla;
173 
174 	crit_enter();
175 	LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) {
176 		if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second)
177 			arptfree(la);
178 	}
179 	callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL);
180 	crit_exit();
181 }
182 
183 /*
184  * Parallel to llc_rtrequest.
185  */
186 static void
187 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
188 {
189 	struct sockaddr *gate = rt->rt_gateway;
190 	struct llinfo_arp *la = rt->rt_llinfo;
191 
192 	struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK };
193 	static boolean_t arpinit_done[MAXCPU];
194 
195 	if (!arpinit_done[mycpuid]) {
196 		arpinit_done[mycpuid] = TRUE;
197 		callout_init(&arptimer_ch[mycpuid]);
198 		callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL);
199 	}
200 	if (rt->rt_flags & RTF_GATEWAY)
201 		return;
202 
203 	switch (req) {
204 	case RTM_ADD:
205 		/*
206 		 * XXX: If this is a manually added route to interface
207 		 * such as older version of routed or gated might provide,
208 		 * restore cloning bit.
209 		 */
210 		if (!(rt->rt_flags & RTF_HOST) &&
211 		    SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
212 			rt->rt_flags |= RTF_CLONING;
213 		if (rt->rt_flags & RTF_CLONING) {
214 			/*
215 			 * Case 1: This route should come from a route to iface.
216 			 */
217 			rt_setgate(rt, rt_key(rt),
218 				   (struct sockaddr *)&null_sdl);
219 			gate = rt->rt_gateway;
220 			SDL(gate)->sdl_type = rt->rt_ifp->if_type;
221 			SDL(gate)->sdl_index = rt->rt_ifp->if_index;
222 			rt->rt_expire = time_second;
223 			break;
224 		}
225 		/* Announce a new entry if requested. */
226 		if (rt->rt_flags & RTF_ANNOUNCE)
227 			arprequest(rt->rt_ifp,
228 			    &SIN(rt_key(rt))->sin_addr,
229 			    &SIN(rt_key(rt))->sin_addr,
230 			    LLADDR(SDL(gate)));
231 		/*FALLTHROUGH*/
232 	case RTM_RESOLVE:
233 		if (gate->sa_family != AF_LINK ||
234 		    gate->sa_len < sizeof(struct sockaddr_dl)) {
235 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
236 			break;
237 		}
238 		SDL(gate)->sdl_type = rt->rt_ifp->if_type;
239 		SDL(gate)->sdl_index = rt->rt_ifp->if_index;
240 		if (la != NULL)
241 			break; /* This happens on a route change */
242 		/*
243 		 * Case 2:  This route may come from cloning, or a manual route
244 		 * add with a LL address.
245 		 */
246 		R_Malloc(la, struct llinfo_arp *, sizeof *la);
247 		rt->rt_llinfo = la;
248 		if (la == NULL) {
249 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
250 			break;
251 		}
252 		bzero(la, sizeof *la);
253 		la->la_rt = rt;
254 		rt->rt_flags |= RTF_LLINFO;
255 		LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le);
256 
257 #ifdef INET
258 		/*
259 		 * This keeps the multicast addresses from showing up
260 		 * in `arp -a' listings as unresolved.  It's not actually
261 		 * functional.  Then the same for broadcast.
262 		 */
263 		if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) {
264 			ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr,
265 					       LLADDR(SDL(gate)));
266 			SDL(gate)->sdl_alen = 6;
267 			rt->rt_expire = 0;
268 		}
269 		if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) {
270 			memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr,
271 			       rt->rt_ifp->if_addrlen);
272 			SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen;
273 			rt->rt_expire = 0;
274 		}
275 #endif
276 
277 		if (SIN(rt_key(rt))->sin_addr.s_addr ==
278 		    (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) {
279 			/*
280 			 * This test used to be
281 			 *	if (loif.if_flags & IFF_UP)
282 			 * It allowed local traffic to be forced
283 			 * through the hardware by configuring the
284 			 * loopback down.  However, it causes problems
285 			 * during network configuration for boards
286 			 * that can't receive packets they send.  It
287 			 * is now necessary to clear "useloopback" and
288 			 * remove the route to force traffic out to
289 			 * the hardware.
290 			 */
291 			rt->rt_expire = 0;
292 			bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)),
293 			      SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen);
294 			if (useloopback)
295 				rt->rt_ifp = loif;
296 		}
297 		break;
298 
299 	case RTM_DELETE:
300 		if (la == NULL)
301 			break;
302 		LIST_REMOVE(la, la_le);
303 		rt->rt_llinfo = NULL;
304 		rt->rt_flags &= ~RTF_LLINFO;
305 		if (la->la_hold != NULL)
306 			m_freem(la->la_hold);
307 		Free(la);
308 	}
309 }
310 
311 /*
312  * Broadcast an ARP request. Caller specifies:
313  *	- arp header source ip address
314  *	- arp header target ip address
315  *	- arp header source ethernet address
316  */
317 static void
318 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip,
319 	   u_char *enaddr)
320 {
321 	struct mbuf *m;
322 	struct ether_header *eh;
323 	struct arphdr *ah;
324 	struct sockaddr sa;
325 	static u_char llcx[] = { 0x82, 0x40, LLC_SNAP_LSAP, LLC_SNAP_LSAP,
326 				 LLC_UI, 0x00, 0x00, 0x00, 0x08, 0x06 };
327 	u_short ar_hrd;
328 
329 	if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL)
330 		return;
331 	m->m_pkthdr.rcvif = (struct ifnet *)NULL;
332 
333 	switch (ifp->if_type) {
334 	case IFT_ETHER:
335 		/*
336 		 * This may not be correct for types not explicitly
337 		 * listed, but this is our best guess
338 		 */
339 	default:
340 		ar_hrd = htons(ARPHRD_ETHER);
341 
342 		m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
343 		m->m_pkthdr.len = m->m_len;
344 		MH_ALIGN(m, m->m_len);
345 
346 		eh = (struct ether_header *)sa.sa_data;
347 		/* if_output() will not swap */
348 		eh->ether_type = htons(ETHERTYPE_ARP);
349 		memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen);
350 
351 		ah = mtod(m, struct arphdr *);
352 		break;
353 	}
354 
355 	ah->ar_hrd = ar_hrd;
356 	ah->ar_pro = htons(ETHERTYPE_IP);
357 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
358 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
359 	ah->ar_op = htons(ARPOP_REQUEST);
360 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
361 	memset(ar_tha(ah), 0, ah->ar_hln);
362 	memcpy(ar_spa(ah), sip, ah->ar_pln);
363 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
364 
365 	sa.sa_family = AF_UNSPEC;
366 	sa.sa_len = sizeof sa;
367 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)NULL);
368 }
369 
370 /*
371  * Resolve an IP address into an ethernet address.  If success,
372  * desten is filled in.  If there is no entry in arptab,
373  * set one up and broadcast a request for the IP address.
374  * Hold onto this mbuf and resend it once the address
375  * is finally resolved.  A return value of 1 indicates
376  * that desten has been filled in and the packet should be sent
377  * normally; a 0 return indicates that the packet has been
378  * taken over here, either now or for later transmission.
379  */
380 int
381 arpresolve(
382 	struct ifnet *ifp,
383 	struct rtentry *rt0,
384 	struct mbuf *m,
385 	struct sockaddr *dst,
386 	u_char *desten)
387 {
388 	struct rtentry *rt;
389 	struct llinfo_arp *la = NULL;
390 	struct sockaddr_dl *sdl;
391 
392 	if (m->m_flags & M_BCAST) {	/* broadcast */
393 		memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen);
394 		return (1);
395 	}
396 	if (m->m_flags & M_MCAST) {/* multicast */
397 		ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
398 		return (1);
399 	}
400 	if (rt0 != NULL) {
401 		if (rt_llroute(dst, rt0, &rt) != 0) {
402 			m_freem(m);
403 			return 0;
404 		}
405 		la = rt->rt_llinfo;
406 	}
407 	if (la == NULL) {
408 		la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE);
409 		if (la != NULL)
410 			rt = la->la_rt;
411 	}
412 	if (la == NULL || rt == NULL) {
413 		log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n",
414 		    inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ",
415 		    rt ? "rt" : "");
416 		m_freem(m);
417 		return (0);
418 	}
419 	sdl = SDL(rt->rt_gateway);
420 	/*
421 	 * Check the address family and length is valid, the address
422 	 * is resolved; otherwise, try to resolve.
423 	 */
424 	if ((rt->rt_expire == 0 || rt->rt_expire > time_second) &&
425 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
426 		/*
427 		 * If entry has an expiry time and it is approaching,
428 		 * see if we need to send an ARP request within this
429 		 * arpt_down interval.
430 		 */
431 		if ((rt->rt_expire != 0) &&
432 		    (time_second + la->la_preempt > rt->rt_expire)) {
433 			arprequest(ifp,
434 				   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
435 				   &SIN(dst)->sin_addr,
436 				   IF_LLADDR(ifp));
437 			la->la_preempt--;
438 		}
439 
440 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
441 		return 1;
442 	}
443 	/*
444 	 * If ARP is disabled on this interface, stop.
445 	 * XXX
446 	 * Probably should not allocate empty llinfo struct if we are
447 	 * not going to be sending out an arp request.
448 	 */
449 	if (ifp->if_flags & IFF_NOARP) {
450 		m_freem(m);
451 		return (0);
452 	}
453 	/*
454 	 * There is an arptab entry, but no ethernet address
455 	 * response yet.  Replace the held mbuf with this
456 	 * latest one.
457 	 */
458 	if (la->la_hold != NULL)
459 		m_freem(la->la_hold);
460 	la->la_hold = m;
461 	if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) {
462 		rt->rt_flags &= ~RTF_REJECT;
463 		if (la->la_asked == 0 || rt->rt_expire != time_second) {
464 			rt->rt_expire = time_second;
465 			if (la->la_asked++ < arp_maxtries) {
466 				arprequest(ifp,
467 					   &SIN(rt->rt_ifa->ifa_addr)->sin_addr,
468 					   &SIN(dst)->sin_addr,
469 					   IF_LLADDR(ifp));
470 			} else {
471 				rt->rt_flags |= RTF_REJECT;
472 				rt->rt_expire += arpt_down;
473 				la->la_asked = 0;
474 				la->la_preempt = arp_maxtries;
475 			}
476 
477 		}
478 	}
479 	return (0);
480 }
481 
482 /*
483  * Common length and type checks are done here,
484  * then the protocol-specific routine is called.
485  */
486 static void
487 arpintr(struct netmsg *msg)
488 {
489 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
490 	struct arphdr *ar;
491 	u_short ar_hrd;
492 
493 	if (m->m_len < sizeof(struct arphdr) &&
494 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
495 		log(LOG_ERR, "arp: runt packet -- m_pullup failed\n");
496 		goto out2;
497 	}
498 	ar = mtod(m, struct arphdr *);
499 
500 	ar_hrd = ntohs(ar->ar_hrd);
501 	if (ar_hrd != ARPHRD_ETHER &&
502 	    ar_hrd != ARPHRD_IEEE802) {
503 		log(LOG_ERR,
504 		    "arp: unknown hardware address format (0x%2D)\n",
505 		    (unsigned char *)&ar->ar_hrd, "");
506 		goto out1;
507 	}
508 
509 	if (m->m_pkthdr.len < arphdr_len(ar) &&
510 	    (m = m_pullup(m, arphdr_len(ar))) == NULL) {
511 		log(LOG_ERR, "arp: runt packet\n");
512 		goto out1;
513 	}
514 
515 	switch (ntohs(ar->ar_pro)) {
516 #ifdef INET
517 		case ETHERTYPE_IP:
518 			in_arpinput(m);
519 			goto out2;
520 #endif
521 	}
522 out1:
523 	m_freem(m);
524 out2:
525 	;
526 	/* msg was embedded in the mbuf, do not reply! */
527 }
528 
529 #ifdef INET
530 /*
531  * ARP for Internet protocols on 10 Mb/s Ethernet.
532  * Algorithm is that given in RFC 826.
533  * In addition, a sanity check is performed on the sender
534  * protocol address, to catch impersonators.
535  * We no longer handle negotiations for use of trailer protocol:
536  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
537  * along with IP replies if we wanted trailers sent to us,
538  * and also sent them in response to IP replies.
539  * This allowed either end to announce the desire to receive
540  * trailer packets.
541  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
542  * but formerly didn't normally send requests.
543  */
544 static int log_arp_wrong_iface = 1;
545 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
546 	&log_arp_wrong_iface, 0,
547 	"log arp packets arriving on the wrong interface");
548 
549 static void
550 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create,
551 		 boolean_t dologging)
552 {
553 	struct arphdr *ah = mtod(m, struct arphdr *);
554 	struct ifnet *ifp = m->m_pkthdr.rcvif;
555 	struct llinfo_arp *la;
556 	struct sockaddr_dl *sdl;
557 	struct rtentry *rt;
558 	int cpu = mycpuid;
559 
560 	la = arplookup(saddr, create, FALSE);
561 	if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) {
562 		struct in_addr isaddr = { saddr };
563 
564 		/* the following is not an error when doing bridging */
565 		if (rt->rt_ifp != ifp) {
566 			if (dologging && log_arp_wrong_iface && cpu == 0) {
567 				log(LOG_ERR,
568 				    "arp: %s is on %s "
569 				    "but got reply from %*D on %s\n",
570 				    inet_ntoa(isaddr),
571 				    rt->rt_ifp->if_xname,
572 				    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
573 				    ifp->if_xname);
574 			}
575 			return;
576 		}
577 		if (sdl->sdl_alen &&
578 		    bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) {
579 			if (rt->rt_expire != 0) {
580 				if (dologging && cpu == 0) {
581 			    		log(LOG_INFO,
582 			    		"arp: %s moved from %*D to %*D on %s\n",
583 			    		inet_ntoa(isaddr),
584 			    		ifp->if_addrlen, (u_char *)LLADDR(sdl),
585 			    		":", ifp->if_addrlen,
586 			    		(u_char *)ar_sha(ah), ":",
587 			    		ifp->if_xname);
588 				}
589 			} else {
590 				if (dologging && cpu == 0) {
591 					log(LOG_ERR,
592 					"arp: %*D attempts to modify permanent entry for %s on %s\n",
593 					ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
594 					inet_ntoa(isaddr), ifp->if_xname);
595 				}
596 				return;
597 			}
598 		}
599 		/*
600 		 * sanity check for the address length.
601 		 * XXX this does not work for protocols with variable address
602 		 * length. -is
603 		 */
604 		if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln &&
605 		    cpu == 0)
606 		{
607 			log(LOG_WARNING,
608 			    "arp from %*D: new addr len %d, was %d",
609 			    ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
610 			    ah->ar_hln, sdl->sdl_alen);
611 		}
612 		if (ifp->if_addrlen != ah->ar_hln) {
613 			if (dologging && cpu == 0) {
614 				log(LOG_WARNING,
615 				"arp from %*D: addr len: new %d, i/f %d (ignored)",
616 				ifp->if_addrlen, (u_char *) ar_sha(ah), ":",
617 				ah->ar_hln, ifp->if_addrlen);
618 			}
619 			return;
620 		}
621 		memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln);
622 		if (rt->rt_expire != 0)
623 			rt->rt_expire = time_second + arpt_keep;
624 		rt->rt_flags &= ~RTF_REJECT;
625 		la->la_asked = 0;
626 		la->la_preempt = arp_maxtries;
627 
628 		/*
629 		 * This particular cpu might have been holding an mbuf
630 		 * pending ARP resolution.  If so, transmit the mbuf now.
631 		 */
632 		if (la->la_hold != NULL) {
633 			m_adj(la->la_hold, sizeof(struct ether_header));
634 			lwkt_serialize_enter(ifp->if_serializer);
635 			(*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt);
636 			lwkt_serialize_exit(ifp->if_serializer);
637 			la->la_hold = NULL;
638 		}
639 	}
640 }
641 
642 #ifdef SMP
643 
644 struct netmsg_arp_update {
645 	struct netmsg	netmsg;
646 	struct mbuf	*m;
647 	in_addr_t	saddr;
648 	boolean_t	create;
649 };
650 
651 static void arp_update_msghandler(struct netmsg *netmsg);
652 
653 #endif
654 
655 /*
656  * Called from arpintr() - this routine is run from a single cpu.
657  */
658 static void
659 in_arpinput(struct mbuf *m)
660 {
661 	struct arphdr *ah;
662 	struct ifnet *ifp = m->m_pkthdr.rcvif;
663 	struct ether_header *eh;
664 	struct rtentry *rt;
665 	struct ifaddr *ifa;
666 	struct in_ifaddr *ia;
667 	struct sockaddr sa;
668 	struct in_addr isaddr, itaddr, myaddr;
669 #ifdef SMP
670 	struct netmsg_arp_update msg;
671 #endif
672 	u_int8_t *enaddr = NULL;
673 	int op;
674 	int req_len;
675 
676 	req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
677 	if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
678 		log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n");
679 		return;
680 	}
681 
682 	ah = mtod(m, struct arphdr *);
683 	op = ntohs(ah->ar_op);
684 	memcpy(&isaddr, ar_spa(ah), sizeof isaddr);
685 	memcpy(&itaddr, ar_tpa(ah), sizeof itaddr);
686 	/*
687 	 * Check both target and sender IP addresses:
688 	 *
689 	 * If we receive the packet on the interface owning the address,
690 	 * then accept the address.
691 	 *
692 	 * For a bridge, we accept the address if the receive interface and
693 	 * the interface owning the address are on the same bridge.
694 	 * (This will change slightly when we have clusters of interfaces).
695 	 */
696 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
697 		/* Skip all ia's which don't match */
698 		if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
699 			continue;
700 
701 		if (ia->ia_ifp == ifp)
702 			goto match;
703 
704 		if (ifp->if_bridge && ia->ia_ifp &&
705 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
706 			goto match;
707 
708 #ifdef CARP
709 	/*
710 	 * If the interface does not match, but the recieving interface
711 	 * is part of carp, we call carp_iamatch to see if this is a
712 	 * request for the virtual host ip.
713 	 * XXX: This is really ugly!
714 	 */
715         	if (ifp->if_carp != NULL &&
716 		    carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) &&
717 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr)
718                         goto match;
719 #endif
720 	}
721 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) {
722 		/* Skip all ia's which don't match */
723 		if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr)
724 			continue;
725 
726 		if (ia->ia_ifp == ifp)
727 			goto match;
728 
729 		if (ifp->if_bridge && ia->ia_ifp &&
730 		    ifp->if_bridge == ia->ia_ifp->if_bridge)
731 			goto match;
732 	}
733 	/*
734 	 * No match, use the first inet address on the receive interface
735 	 * as a dummy address for the rest of the function.
736 	 */
737 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
738 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
739 			ia = ifatoia(ifa);
740 			goto match;
741 		}
742 	}
743 	/*
744 	 * If we got here, we didn't find any suitable interface,
745 	 * so drop the packet.
746 	 */
747 	m_freem(m);
748 	return;
749 
750 match:
751 	if (!enaddr)
752 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
753 	myaddr = ia->ia_addr.sin_addr;
754 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) {
755 		m_freem(m);	/* it's from me, ignore it. */
756 		return;
757 	}
758 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
759 		log(LOG_ERR,
760 		    "arp: link address is broadcast for IP address %s!\n",
761 		    inet_ntoa(isaddr));
762 		m_freem(m);
763 		return;
764 	}
765 	if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) {
766 		log(LOG_ERR,
767 		   "arp: %*D is using my IP address %s!\n",
768 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
769 		   inet_ntoa(isaddr));
770 		itaddr = myaddr;
771 		goto reply;
772 	}
773 #ifdef SMP
774 	netmsg_init(&msg.netmsg, &curthread->td_msgport, 0,
775 		    arp_update_msghandler);
776 	msg.m = m;
777 	msg.saddr = isaddr.s_addr;
778 	msg.create = (itaddr.s_addr == myaddr.s_addr);
779 	lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0);
780 #endif
781 	arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr),
782 			 TRUE);
783 reply:
784 	if (op != ARPOP_REQUEST) {
785 		m_freem(m);
786 		return;
787 	}
788 	if (itaddr.s_addr == myaddr.s_addr) {
789 		/* I am the target */
790 		memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
791 		memcpy(ar_sha(ah), enaddr, ah->ar_hln);
792 	} else {
793 		struct llinfo_arp *la;
794 
795 		la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY);
796 		if (la == NULL) {
797 			struct sockaddr_in sin;
798 
799 			if (!arp_proxyall) {
800 				m_freem(m);
801 				return;
802 			}
803 
804 			bzero(&sin, sizeof sin);
805 			sin.sin_family = AF_INET;
806 			sin.sin_len = sizeof sin;
807 			sin.sin_addr = itaddr;
808 
809 			rt = rtpurelookup((struct sockaddr *)&sin);
810 			if (rt == NULL) {
811 				m_freem(m);
812 				return;
813 			}
814 			--rt->rt_refcnt;
815 			/*
816 			 * Don't send proxies for nodes on the same interface
817 			 * as this one came out of, or we'll get into a fight
818 			 * over who claims what Ether address.
819 			 */
820 			if (rt->rt_ifp == ifp) {
821 				m_freem(m);
822 				return;
823 			}
824 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
825 			memcpy(ar_sha(ah), enaddr, ah->ar_hln);
826 #ifdef DEBUG_PROXY
827 			kprintf("arp: proxying for %s\n", inet_ntoa(itaddr));
828 #endif
829 		} else {
830 			struct sockaddr_dl *sdl;
831 
832 			rt = la->la_rt;
833 			memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
834 			sdl = SDL(rt->rt_gateway);
835 			memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln);
836 		}
837 	}
838 
839 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
840 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
841 	ah->ar_op = htons(ARPOP_REPLY);
842 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
843 	switch (ifp->if_type) {
844 	case IFT_ETHER:
845 	/*
846 	 * May not be correct for types not explictly
847 	 * listed, but it is our best guess.
848 	 */
849 	default:
850 		eh = (struct ether_header *)sa.sa_data;
851 		memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost);
852 		eh->ether_type = htons(ETHERTYPE_ARP);
853 		break;
854 	}
855 	sa.sa_family = AF_UNSPEC;
856 	sa.sa_len = sizeof sa;
857 	lwkt_serialize_enter(ifp->if_serializer);
858 	(*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0);
859 	lwkt_serialize_exit(ifp->if_serializer);
860 	return;
861 }
862 
863 #ifdef SMP
864 
865 static
866 void
867 arp_update_msghandler(struct netmsg *netmsg)
868 {
869 	struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg;
870 	int nextcpu;
871 
872 	arp_update_oncpu(msg->m, msg->saddr, msg->create, FALSE);
873 
874 	nextcpu = mycpuid + 1;
875 	if (nextcpu < ncpus) {
876 		lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg);
877 	} else {
878 		lwkt_replymsg(&msg->netmsg.nm_lmsg, 0);
879 	}
880 }
881 
882 #endif
883 
884 #endif
885 
886 /*
887  * Free an arp entry.  If the arp entry is actively referenced or represents
888  * a static entry we only clear it back to an unresolved state, otherwise
889  * we destroy the entry entirely.
890  *
891  * Note that static entries are created when route add ... -interface is used
892  * to create an interface route to a (direct) destination.
893  */
894 static void
895 arptfree(struct llinfo_arp *la)
896 {
897 	struct rtentry *rt = la->la_rt;
898 	struct sockaddr_dl *sdl;
899 
900 	if (rt == NULL)
901 		panic("arptfree");
902 	sdl = SDL(rt->rt_gateway);
903 	if (sdl != NULL &&
904 	    ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) ||
905 	     (rt->rt_flags & RTF_STATIC))) {
906 		sdl->sdl_alen = 0;
907 		la->la_preempt = la->la_asked = 0;
908 		rt->rt_flags &= ~RTF_REJECT;
909 		return;
910 	}
911 	rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL);
912 }
913 
914 /*
915  * Lookup or enter a new address in arptab.
916  */
917 static struct llinfo_arp *
918 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy)
919 {
920 	struct rtentry *rt;
921 	struct sockaddr_inarp sin = { sizeof sin, AF_INET };
922 	const char *why = NULL;
923 
924 	sin.sin_addr.s_addr = addr;
925 	sin.sin_other = proxy ? SIN_PROXY : 0;
926 	if (create)
927 		rt = rtlookup((struct sockaddr *)&sin);
928 	else
929 		rt = rtpurelookup((struct sockaddr *)&sin);
930 	if (rt == NULL)
931 		return (NULL);
932 	rt->rt_refcnt--;
933 
934 	if (rt->rt_flags & RTF_GATEWAY)
935 		why = "host is not on local network";
936 	else if (!(rt->rt_flags & RTF_LLINFO))
937 		why = "could not allocate llinfo";
938 	else if (rt->rt_gateway->sa_family != AF_LINK)
939 		why = "gateway route is not ours";
940 
941 	if (why) {
942 		if (create) {
943 			log(LOG_DEBUG, "arplookup %s failed: %s\n",
944 			    inet_ntoa(sin.sin_addr), why);
945 		}
946 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) {
947 			/* No references to this route.  Purge it. */
948 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
949 				  rt_mask(rt), rt->rt_flags, NULL);
950 		}
951 		return (NULL);
952 	}
953 	return (rt->rt_llinfo);
954 }
955 
956 void
957 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
958 {
959 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
960 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
961 			   IF_LLADDR(ifp));
962 	ifa->ifa_rtrequest = arp_rtrequest;
963 	ifa->ifa_flags |= RTF_CLONING;
964 }
965 
966 void
967 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
968 {
969 	if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY)
970 		arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr,
971 			   enaddr);
972 	ifa->ifa_rtrequest = arp_rtrequest;
973 	ifa->ifa_flags |= RTF_CLONING;
974 }
975 
976 static void
977 arp_init(void)
978 {
979 	int cpu;
980 
981 	for (cpu = 0; cpu < ncpus2; cpu++)
982 		LIST_INIT(&llinfo_arp_list[cpu]);
983 	netisr_register(NETISR_ARP, cpu0_portfn, arpintr);
984 }
985 
986 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
987