xref: /NextBSD/sys/netinet/if_ether.c (revision 37e74d4f6151b6fae3b07141f988985cb55f2dfc)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  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  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ether.c	8.1 (Berkeley) 6/10/93
30  */
31 
32 /*
33  * Ethernet address resolution protocol.
34  * TODO:
35  *	add "inuse/lock" bit (or ref. count) along with valid bit
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/queue.h>
47 #include <sys/sysctl.h>
48 #include <sys/systm.h>
49 #include <sys/mbuf.h>
50 #include <sys/malloc.h>
51 #include <sys/proc.h>
52 #include <sys/rmlock.h>
53 #include <sys/socket.h>
54 #include <sys/syslog.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/netisr.h>
61 #include <net/ethernet.h>
62 #include <net/route.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_fib.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #include <netinet/if_ether.h>
70 #ifdef INET
71 #include <netinet/ip_carp.h>
72 #endif
73 
74 #include <security/mac/mac_framework.h>
75 
76 #define SIN(s) ((const struct sockaddr_in *)(s))
77 
78 static struct timeval arp_lastlog;
79 static int arp_curpps;
80 static int arp_maxpps = 1;
81 
82 /* Simple ARP state machine */
83 enum arp_llinfo_state {
84 	ARP_LLINFO_INCOMPLETE = 0, /* No LLE data */
85 	ARP_LLINFO_REACHABLE,	/* LLE is valid */
86 	ARP_LLINFO_VERIFY,	/* LLE is valid, need refresh */
87 	ARP_LLINFO_DELETED,	/* LLE is deleted */
88 };
89 
90 SYSCTL_DECL(_net_link_ether);
91 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
92 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
93 
94 /* timer values */
95 static VNET_DEFINE(int, arpt_keep) = (20*60);	/* once resolved, good for 20
96 						 * minutes */
97 static VNET_DEFINE(int, arp_maxtries) = 5;
98 static VNET_DEFINE(int, arp_proxyall) = 0;
99 static VNET_DEFINE(int, arpt_down) = 20;	/* keep incomplete entries for
100 						 * 20 seconds */
101 static VNET_DEFINE(int, arpt_rexmit) = 1;	/* retransmit arp entries, sec*/
102 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
103 VNET_PCPUSTAT_SYSINIT(arpstat);
104 
105 #ifdef VIMAGE
106 VNET_PCPUSTAT_SYSUNINIT(arpstat);
107 #endif /* VIMAGE */
108 
109 static VNET_DEFINE(int, arp_maxhold) = 1;
110 
111 #define	V_arpt_keep		VNET(arpt_keep)
112 #define	V_arpt_down		VNET(arpt_down)
113 #define	V_arpt_rexmit		VNET(arpt_rexmit)
114 #define	V_arp_maxtries		VNET(arp_maxtries)
115 #define	V_arp_proxyall		VNET(arp_proxyall)
116 #define	V_arp_maxhold		VNET(arp_maxhold)
117 
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW,
119 	&VNET_NAME(arpt_keep), 0,
120 	"ARP entry lifetime in seconds");
121 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW,
122 	&VNET_NAME(arp_maxtries), 0,
123 	"ARP resolution attempts before returning error");
124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW,
125 	&VNET_NAME(arp_proxyall), 0,
126 	"Enable proxy ARP for all suitable requests");
127 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW,
128 	&VNET_NAME(arpt_down), 0,
129 	"Incomplete ARP entry lifetime in seconds");
130 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
131     arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
132 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW,
133 	&VNET_NAME(arp_maxhold), 0,
134 	"Number of packets to hold per ARP entry");
135 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
136 	CTLFLAG_RW, &arp_maxpps, 0,
137 	"Maximum number of remotely triggered ARP messages that can be "
138 	"logged per second");
139 
140 #define	ARP_LOG(pri, ...)	do {					\
141 	if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps))	\
142 		log((pri), "arp: " __VA_ARGS__);			\
143 } while (0)
144 
145 
146 static void	arp_init(void);
147 static void	arpintr(struct mbuf *);
148 static void	arptimer(void *);
149 #ifdef INET
150 static void	in_arpinput(struct mbuf *);
151 #endif
152 
153 static void arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr,
154     struct ifnet *ifp, int bridged, struct llentry *la);
155 static void arp_mark_lle_reachable(struct llentry *la);
156 static void arp_iflladdr(void *arg __unused, struct ifnet *ifp);
157 
158 static eventhandler_tag iflladdr_tag;
159 
160 static const struct netisr_handler arp_nh = {
161 	.nh_name = "arp",
162 	.nh_handler = arpintr,
163 	.nh_proto = NETISR_ARP,
164 	.nh_policy = NETISR_POLICY_SOURCE,
165 };
166 
167 /*
168  * Timeout routine.  Age arp_tab entries periodically.
169  */
170 static void
arptimer(void * arg)171 arptimer(void *arg)
172 {
173 	struct llentry *lle = (struct llentry *)arg;
174 	struct ifnet *ifp;
175 	int r_skip_req;
176 
177 	if (lle->la_flags & LLE_STATIC) {
178 		return;
179 	}
180 	LLE_WLOCK(lle);
181 	if (callout_pending(&lle->lle_timer)) {
182 		/*
183 		 * Here we are a bit odd here in the treatment of
184 		 * active/pending. If the pending bit is set, it got
185 		 * rescheduled before I ran. The active
186 		 * bit we ignore, since if it was stopped
187 		 * in ll_tablefree() and was currently running
188 		 * it would have return 0 so the code would
189 		 * not have deleted it since the callout could
190 		 * not be stopped so we want to go through
191 		 * with the delete here now. If the callout
192 		 * was restarted, the pending bit will be back on and
193 		 * we just want to bail since the callout_reset would
194 		 * return 1 and our reference would have been removed
195 		 * by arpresolve() below.
196 		 */
197 		LLE_WUNLOCK(lle);
198  		return;
199  	}
200 	ifp = lle->lle_tbl->llt_ifp;
201 	CURVNET_SET(ifp->if_vnet);
202 
203 	switch (lle->ln_state) {
204 	case ARP_LLINFO_REACHABLE:
205 
206 		/*
207 		 * Expiration time is approaching.
208 		 * Let's try to refresh entry if it is still
209 		 * in use.
210 		 *
211 		 * Set r_skip_req to get feedback from
212 		 * fast path. Change state and re-schedule
213 		 * ourselves.
214 		 */
215 		LLE_REQ_LOCK(lle);
216 		lle->r_skip_req = 1;
217 		LLE_REQ_UNLOCK(lle);
218 		lle->ln_state = ARP_LLINFO_VERIFY;
219 		callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
220 		LLE_WUNLOCK(lle);
221 		CURVNET_RESTORE();
222 		return;
223 	case ARP_LLINFO_VERIFY:
224 		LLE_REQ_LOCK(lle);
225 		r_skip_req = lle->r_skip_req;
226 		LLE_REQ_UNLOCK(lle);
227 
228 		if (r_skip_req == 0 && lle->la_preempt > 0) {
229 			/* Entry was used, issue refresh request */
230 			struct in_addr dst;
231 			dst = lle->r_l3addr.addr4;
232 			lle->la_preempt--;
233 			callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
234 			LLE_WUNLOCK(lle);
235 			arprequest(ifp, NULL, &dst, NULL);
236 			CURVNET_RESTORE();
237 			return;
238 		}
239 		/* Nothing happened. Reschedule if not too late */
240 		if (lle->la_expire > time_uptime) {
241 			callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
242 			LLE_WUNLOCK(lle);
243 			CURVNET_RESTORE();
244 			return;
245 		}
246 		break;
247 	case ARP_LLINFO_INCOMPLETE:
248 	case ARP_LLINFO_DELETED:
249 		break;
250 	}
251 
252 	if ((lle->la_flags & LLE_DELETED) == 0) {
253 		int evt;
254 
255 		if (lle->la_flags & LLE_VALID)
256 			evt = LLENTRY_EXPIRED;
257 		else
258 			evt = LLENTRY_TIMEDOUT;
259 		EVENTHANDLER_INVOKE(lle_event, lle, evt);
260 	}
261 
262 	callout_stop(&lle->lle_timer);
263 
264 	/* XXX: LOR avoidance. We still have ref on lle. */
265 	LLE_WUNLOCK(lle);
266 	IF_AFDATA_LOCK(ifp);
267 	LLE_WLOCK(lle);
268 
269 	/* Guard against race with other llentry_free(). */
270 	if (lle->la_flags & LLE_LINKED) {
271 		LLE_REMREF(lle);
272 		lltable_unlink_entry(lle->lle_tbl, lle);
273 	}
274 	IF_AFDATA_UNLOCK(ifp);
275 
276 	size_t pkts_dropped = llentry_free(lle);
277 
278 	ARPSTAT_ADD(dropped, pkts_dropped);
279 	ARPSTAT_INC(timeouts);
280 
281 	CURVNET_RESTORE();
282 }
283 
284 /*
285  * Stores link-layer header for @ifp in format suitable for if_output()
286  * into buffer @buf. Resulting header length is stored in @bufsize.
287  *
288  * Returns 0 on success.
289  */
290 static int
arp_fillheader(struct ifnet * ifp,struct arphdr * ah,int bcast,u_char * buf,size_t * bufsize)291 arp_fillheader(struct ifnet *ifp, struct arphdr *ah, int bcast, u_char *buf,
292     size_t *bufsize)
293 {
294 	struct if_encap_req ereq;
295 	int error;
296 
297 	bzero(buf, *bufsize);
298 	bzero(&ereq, sizeof(ereq));
299 	ereq.buf = buf;
300 	ereq.bufsize = *bufsize;
301 	ereq.rtype = IFENCAP_LL;
302 	ereq.family = AF_ARP;
303 	ereq.lladdr = ar_tha(ah);
304 	ereq.hdata = (u_char *)ah;
305 	if (bcast)
306 		ereq.flags = IFENCAP_FLAG_BROADCAST;
307 	error = ifp->if_requestencap(ifp, &ereq);
308 	if (error == 0)
309 		*bufsize = ereq.bufsize;
310 
311 	return (error);
312 }
313 
314 
315 /*
316  * Broadcast an ARP request. Caller specifies:
317  *	- arp header source ip address
318  *	- arp header target ip address
319  *	- arp header source ethernet address
320  */
321 void
arprequest(struct ifnet * ifp,const struct in_addr * sip,const struct in_addr * tip,u_char * enaddr)322 arprequest(struct ifnet *ifp, const struct in_addr *sip,
323     const struct in_addr *tip, u_char *enaddr)
324 {
325 	struct mbuf *m;
326 	struct arphdr *ah;
327 	struct sockaddr sa;
328 	u_char *carpaddr = NULL;
329 	uint8_t linkhdr[LLE_MAX_LINKHDR];
330 	size_t linkhdrsize;
331 	struct route ro;
332 	int error;
333 
334 	if (sip == NULL) {
335 		/*
336 		 * The caller did not supply a source address, try to find
337 		 * a compatible one among those assigned to this interface.
338 		 */
339 		struct ifaddr *ifa;
340 
341 		IF_ADDR_RLOCK(ifp);
342 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
343 			if (ifa->ifa_addr->sa_family != AF_INET)
344 				continue;
345 
346 			if (ifa->ifa_carp) {
347 				if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
348 					continue;
349 				sip = &IA_SIN(ifa)->sin_addr;
350 			} else {
351 				carpaddr = NULL;
352 				sip = &IA_SIN(ifa)->sin_addr;
353 			}
354 
355 			if (0 == ((sip->s_addr ^ tip->s_addr) &
356 			    IA_MASKSIN(ifa)->sin_addr.s_addr))
357 				break;  /* found it. */
358 		}
359 		IF_ADDR_RUNLOCK(ifp);
360 		if (sip == NULL) {
361 			printf("%s: cannot find matching address\n", __func__);
362 			return;
363 		}
364 	}
365 	if (enaddr == NULL)
366 		enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
367 
368 	if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
369 		return;
370 	m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
371 		2 * ifp->if_addrlen;
372 	m->m_pkthdr.len = m->m_len;
373 	M_ALIGN(m, m->m_len);
374 	ah = mtod(m, struct arphdr *);
375 	bzero((caddr_t)ah, m->m_len);
376 #ifdef MAC
377 	mac_netinet_arp_send(ifp, m);
378 #endif
379 	ah->ar_pro = htons(ETHERTYPE_IP);
380 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
381 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
382 	ah->ar_op = htons(ARPOP_REQUEST);
383 	bcopy(enaddr, ar_sha(ah), ah->ar_hln);
384 	bcopy(sip, ar_spa(ah), ah->ar_pln);
385 	bcopy(tip, ar_tpa(ah), ah->ar_pln);
386 	sa.sa_family = AF_ARP;
387 	sa.sa_len = 2;
388 
389 	/* Calculate link header for sending frame */
390 	bzero(&ro, sizeof(ro));
391 	linkhdrsize = sizeof(linkhdr);
392 	error = arp_fillheader(ifp, ah, 1, linkhdr, &linkhdrsize);
393 	if (error != 0 && error != EAFNOSUPPORT) {
394 		ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
395 		    if_name(ifp), error);
396 		return;
397 	}
398 
399 	ro.ro_prepend = linkhdr;
400 	ro.ro_plen = linkhdrsize;
401 	ro.ro_flags = 0;
402 
403 	m->m_flags |= M_BCAST;
404 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
405 	(*ifp->if_output)(ifp, m, &sa, &ro);
406 	ARPSTAT_INC(txrequests);
407 }
408 
409 
410 /*
411  * Resolve an IP address into an ethernet address - heavy version.
412  * Used internally by arpresolve().
413  * We have already checked than  we can't use existing lle without
414  * modification so we have to acquire LLE_EXCLUSIVE lle lock.
415  *
416  * On success, desten and flags are filled in and the function returns 0;
417  * If the packet must be held pending resolution, we return EWOULDBLOCK
418  * On other errors, we return the corresponding error code.
419  * Note that m_freem() handles NULL.
420  */
421 static int
arpresolve_full(struct ifnet * ifp,int is_gw,int flags,struct mbuf * m,const struct sockaddr * dst,u_char * desten,uint32_t * pflags)422 arpresolve_full(struct ifnet *ifp, int is_gw, int flags, struct mbuf *m,
423 	const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
424 {
425 	struct llentry *la = NULL, *la_tmp;
426 	struct mbuf *curr = NULL;
427 	struct mbuf *next = NULL;
428 	int error, renew;
429 	char *lladdr;
430 	int ll_len;
431 
432 	if (pflags != NULL)
433 		*pflags = 0;
434 
435 	if ((flags & LLE_CREATE) == 0) {
436 		IF_AFDATA_RLOCK(ifp);
437 		la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
438 		IF_AFDATA_RUNLOCK(ifp);
439 	}
440 	if (la == NULL && (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) {
441 		la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
442 		if (la == NULL) {
443 			log(LOG_DEBUG,
444 			    "arpresolve: can't allocate llinfo for %s on %s\n",
445 			    inet_ntoa(SIN(dst)->sin_addr), if_name(ifp));
446 			m_freem(m);
447 			return (EINVAL);
448 		}
449 
450 		IF_AFDATA_WLOCK(ifp);
451 		LLE_WLOCK(la);
452 		la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
453 		/* Prefer ANY existing lle over newly-created one */
454 		if (la_tmp == NULL)
455 			lltable_link_entry(LLTABLE(ifp), la);
456 		IF_AFDATA_WUNLOCK(ifp);
457 		if (la_tmp != NULL) {
458 			lltable_free_entry(LLTABLE(ifp), la);
459 			la = la_tmp;
460 		}
461 	}
462 	if (la == NULL) {
463 		m_freem(m);
464 		return (EINVAL);
465 	}
466 
467 	if ((la->la_flags & LLE_VALID) &&
468 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
469 		if (flags & LLE_ADDRONLY) {
470 			lladdr = la->ll_addr;
471 			ll_len = ifp->if_addrlen;
472 		} else {
473 			lladdr = la->r_linkdata;
474 			ll_len = la->r_hdrlen;
475 		}
476 		bcopy(lladdr, desten, ll_len);
477 
478 		/* Check if we have feedback request from arptimer() */
479 		if (la->r_skip_req != 0) {
480 			LLE_REQ_LOCK(la);
481 			la->r_skip_req = 0; /* Notify that entry was used */
482 			LLE_REQ_UNLOCK(la);
483 		}
484 		if (pflags != NULL)
485 			*pflags = la->la_flags & (LLE_VALID|LLE_IFADDR);
486 		LLE_WUNLOCK(la);
487 		return (0);
488 	}
489 
490 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
491 	/*
492 	 * There is an arptab entry, but no ethernet address
493 	 * response yet.  Add the mbuf to the list, dropping
494 	 * the oldest packet if we have exceeded the system
495 	 * setting.
496 	 */
497 	if (m != NULL) {
498 		if (la->la_numheld >= V_arp_maxhold) {
499 			if (la->la_hold != NULL) {
500 				next = la->la_hold->m_nextpkt;
501 				m_freem(la->la_hold);
502 				la->la_hold = next;
503 				la->la_numheld--;
504 				ARPSTAT_INC(dropped);
505 			}
506 		}
507 		if (la->la_hold != NULL) {
508 			curr = la->la_hold;
509 			while (curr->m_nextpkt != NULL)
510 				curr = curr->m_nextpkt;
511 			curr->m_nextpkt = m;
512 		} else
513 			la->la_hold = m;
514 		la->la_numheld++;
515 	}
516 	/*
517 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
518 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
519 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
520 	 * ARP request, but not faster than one request per second.
521 	 */
522 	if (la->la_asked < V_arp_maxtries)
523 		error = EWOULDBLOCK;	/* First request. */
524 	else
525 		error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN;
526 
527 	if (renew) {
528 		int canceled;
529 
530 		LLE_ADDREF(la);
531 		la->la_expire = time_uptime;
532 		canceled = callout_reset(&la->lle_timer, hz * V_arpt_down,
533 		    arptimer, la);
534 		if (canceled)
535 			LLE_REMREF(la);
536 		la->la_asked++;
537 		LLE_WUNLOCK(la);
538 		arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
539 		return (error);
540 	}
541 
542 	LLE_WUNLOCK(la);
543 	return (error);
544 }
545 
546 /*
547  * Resolve an IP address into an ethernet address.
548  */
549 int
arpresolve_addr(struct ifnet * ifp,int flags,const struct sockaddr * dst,char * desten,uint32_t * pflags)550 arpresolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
551     char *desten, uint32_t *pflags)
552 {
553 	int error;
554 
555 	flags |= LLE_ADDRONLY;
556 	error = arpresolve_full(ifp, 0, flags, NULL, dst, desten, pflags);
557 	return (error);
558 }
559 
560 /*
561  * Lookups link header based on an IP address.
562  * On input:
563  *    ifp is the interface we use
564  *    is_gw != 0 if @dst represents gateway to some destination
565  *    m is the mbuf. May be NULL if we don't have a packet.
566  *    dst is the next hop,
567  *    desten is the storage to put LL header.
568  *    flags returns lle entry flags.
569  *    flags returns subset of lle flags: LLE_VALID | LLE_IFADDR
570  *
571  * On success, full/partial link header and flags are filled in and
572  * the function returns 0.
573  * If the packet must be held pending resolution, we return EWOULDBLOCK
574  * On other errors, we return the corresponding error code.
575  * Note that m_freem() handles NULL.
576  */
577 int
arpresolve(struct ifnet * ifp,int is_gw,struct mbuf * m,const struct sockaddr * dst,u_char * desten,uint32_t * pflags)578 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
579 	const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
580 {
581 	struct llentry *la = 0;
582 
583 	if (pflags != NULL)
584 		*pflags = 0;
585 
586 	if (m != NULL) {
587 		if (m->m_flags & M_BCAST) {
588 			/* broadcast */
589 			(void)memcpy(desten,
590 			    ifp->if_broadcastaddr, ifp->if_addrlen);
591 			return (0);
592 		}
593 		if (m->m_flags & M_MCAST) {
594 			/* multicast */
595 			ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
596 			return (0);
597 		}
598 	}
599 
600 	IF_AFDATA_RLOCK(ifp);
601 	la = lla_lookup(LLTABLE(ifp), LLE_UNLOCKED, dst);
602 	if (la != NULL && (la->r_flags & RLLE_VALID) != 0) {
603 		/* Entry found, let's copy lle info */
604 		bcopy(la->r_linkdata, desten, la->r_hdrlen);
605 		if (pflags != NULL)
606 			*pflags = LLE_VALID | (la->r_flags & RLLE_IFADDR);
607 		/* Check if we have feedback request from arptimer() */
608 		if (la->r_skip_req != 0) {
609 			LLE_REQ_LOCK(la);
610 			la->r_skip_req = 0; /* Notify that entry was used */
611 			LLE_REQ_UNLOCK(la);
612 		}
613 		IF_AFDATA_RUNLOCK(ifp);
614 
615 		return (0);
616 	}
617 	IF_AFDATA_RUNLOCK(ifp);
618 
619 	return (arpresolve_full(ifp, is_gw, la == NULL ? LLE_CREATE : 0, m, dst,
620 	    desten, pflags));
621 }
622 
623 /*
624  * Common length and type checks are done here,
625  * then the protocol-specific routine is called.
626  */
627 static void
arpintr(struct mbuf * m)628 arpintr(struct mbuf *m)
629 {
630 	struct arphdr *ar;
631 	struct ifnet *ifp;
632 	char *layer;
633 	int hlen;
634 
635 	ifp = m->m_pkthdr.rcvif;
636 
637 	if (m->m_len < sizeof(struct arphdr) &&
638 	    ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
639 		ARP_LOG(LOG_NOTICE, "packet with short header received on %s\n",
640 		    if_name(ifp));
641 		return;
642 	}
643 	ar = mtod(m, struct arphdr *);
644 
645 	/* Check if length is sufficient */
646 	if (m->m_len <  arphdr_len(ar)) {
647 		m = m_pullup(m, arphdr_len(ar));
648 		if (m == NULL) {
649 			ARP_LOG(LOG_NOTICE, "short packet received on %s\n",
650 			    if_name(ifp));
651 			return;
652 		}
653 		ar = mtod(m, struct arphdr *);
654 	}
655 
656 	hlen = 0;
657 	layer = "";
658 	switch (ntohs(ar->ar_hrd)) {
659 	case ARPHRD_ETHER:
660 		hlen = ETHER_ADDR_LEN; /* RFC 826 */
661 		layer = "ethernet";
662 		break;
663 	case ARPHRD_IEEE802:
664 		hlen = 6; /* RFC 1390, FDDI_ADDR_LEN */
665 		layer = "fddi";
666 		break;
667 	case ARPHRD_ARCNET:
668 		hlen = 1; /* RFC 1201, ARC_ADDR_LEN */
669 		layer = "arcnet";
670 		break;
671 	case ARPHRD_INFINIBAND:
672 		hlen = 20;	/* RFC 4391, INFINIBAND_ALEN */
673 		layer = "infiniband";
674 		break;
675 	case ARPHRD_IEEE1394:
676 		hlen = 0; /* SHALL be 16 */ /* RFC 2734 */
677 		layer = "firewire";
678 
679 		/*
680 		 * Restrict too long hardware addresses.
681 		 * Currently we are capable of handling 20-byte
682 		 * addresses ( sizeof(lle->ll_addr) )
683 		 */
684 		if (ar->ar_hln >= 20)
685 			hlen = 16;
686 		break;
687 	default:
688 		ARP_LOG(LOG_NOTICE,
689 		    "packet with unknown hardware format 0x%02d received on "
690 		    "%s\n", ntohs(ar->ar_hrd), if_name(ifp));
691 		m_freem(m);
692 		return;
693 	}
694 
695 	if (hlen != 0 && hlen != ar->ar_hln) {
696 		ARP_LOG(LOG_NOTICE,
697 		    "packet with invalid %s address length %d received on %s\n",
698 		    layer, ar->ar_hln, if_name(ifp));
699 		m_freem(m);
700 		return;
701 	}
702 
703 	ARPSTAT_INC(received);
704 	switch (ntohs(ar->ar_pro)) {
705 #ifdef INET
706 	case ETHERTYPE_IP:
707 		in_arpinput(m);
708 		return;
709 #endif
710 	}
711 	m_freem(m);
712 }
713 
714 #ifdef INET
715 /*
716  * ARP for Internet protocols on 10 Mb/s Ethernet.
717  * Algorithm is that given in RFC 826.
718  * In addition, a sanity check is performed on the sender
719  * protocol address, to catch impersonators.
720  * We no longer handle negotiations for use of trailer protocol:
721  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
722  * along with IP replies if we wanted trailers sent to us,
723  * and also sent them in response to IP replies.
724  * This allowed either end to announce the desire to receive
725  * trailer packets.
726  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
727  * but formerly didn't normally send requests.
728  */
729 static int log_arp_wrong_iface = 1;
730 static int log_arp_movements = 1;
731 static int log_arp_permanent_modify = 1;
732 static int allow_multicast = 0;
733 
734 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
735 	&log_arp_wrong_iface, 0,
736 	"log arp packets arriving on the wrong interface");
737 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
738 	&log_arp_movements, 0,
739 	"log arp replies from MACs different than the one in the cache");
740 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
741 	&log_arp_permanent_modify, 0,
742 	"log arp replies from MACs different than the one in the permanent arp entry");
743 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
744 	&allow_multicast, 0, "accept multicast addresses");
745 
746 static void
in_arpinput(struct mbuf * m)747 in_arpinput(struct mbuf *m)
748 {
749 	struct rm_priotracker in_ifa_tracker;
750 	struct arphdr *ah;
751 	struct ifnet *ifp = m->m_pkthdr.rcvif;
752 	struct llentry *la = NULL, *la_tmp;
753 	struct ifaddr *ifa;
754 	struct in_ifaddr *ia;
755 	struct sockaddr sa;
756 	struct in_addr isaddr, itaddr, myaddr;
757 	u_int8_t *enaddr = NULL;
758 	int op;
759 	int bridged = 0, is_bridge = 0;
760 	int carped;
761 	struct sockaddr_in sin;
762 	struct sockaddr *dst;
763 	struct nhop4_basic nh4;
764 	uint8_t linkhdr[LLE_MAX_LINKHDR];
765 	struct route ro;
766 	size_t linkhdrsize;
767 	int lladdr_off;
768 	int error;
769 
770 	sin.sin_len = sizeof(struct sockaddr_in);
771 	sin.sin_family = AF_INET;
772 	sin.sin_addr.s_addr = 0;
773 
774 	if (ifp->if_bridge)
775 		bridged = 1;
776 	if (ifp->if_type == IFT_BRIDGE)
777 		is_bridge = 1;
778 
779 	/*
780 	 * We already have checked that mbuf contains enough contiguous data
781 	 * to hold entire arp message according to the arp header.
782 	 */
783 	ah = mtod(m, struct arphdr *);
784 
785 	/*
786 	 * ARP is only for IPv4 so we can reject packets with
787 	 * a protocol length not equal to an IPv4 address.
788 	 */
789 	if (ah->ar_pln != sizeof(struct in_addr)) {
790 		ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
791 		    sizeof(struct in_addr));
792 		goto drop;
793 	}
794 
795 	if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
796 		ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
797 		    ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
798 		goto drop;
799 	}
800 
801 	op = ntohs(ah->ar_op);
802 	(void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
803 	(void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
804 
805 	if (op == ARPOP_REPLY)
806 		ARPSTAT_INC(rxreplies);
807 
808 	/*
809 	 * For a bridge, we want to check the address irrespective
810 	 * of the receive interface. (This will change slightly
811 	 * when we have clusters of interfaces).
812 	 */
813 	IN_IFADDR_RLOCK(&in_ifa_tracker);
814 	LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
815 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
816 		    ia->ia_ifp == ifp) &&
817 		    itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
818 		    (ia->ia_ifa.ifa_carp == NULL ||
819 		    (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
820 			ifa_ref(&ia->ia_ifa);
821 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
822 			goto match;
823 		}
824 	}
825 	LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
826 		if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
827 		    ia->ia_ifp == ifp) &&
828 		    isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
829 			ifa_ref(&ia->ia_ifa);
830 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
831 			goto match;
832 		}
833 
834 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia)				\
835   (ia->ia_ifp->if_bridge == ifp->if_softc &&				\
836   !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) &&	\
837   addr == ia->ia_addr.sin_addr.s_addr)
838 	/*
839 	 * Check the case when bridge shares its MAC address with
840 	 * some of its children, so packets are claimed by bridge
841 	 * itself (bridge_input() does it first), but they are really
842 	 * meant to be destined to the bridge member.
843 	 */
844 	if (is_bridge) {
845 		LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
846 			if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
847 				ifa_ref(&ia->ia_ifa);
848 				ifp = ia->ia_ifp;
849 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
850 				goto match;
851 			}
852 		}
853 	}
854 #undef BDG_MEMBER_MATCHES_ARP
855 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
856 
857 	/*
858 	 * No match, use the first inet address on the receive interface
859 	 * as a dummy address for the rest of the function.
860 	 */
861 	IF_ADDR_RLOCK(ifp);
862 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
863 		if (ifa->ifa_addr->sa_family == AF_INET &&
864 		    (ifa->ifa_carp == NULL ||
865 		    (*carp_iamatch_p)(ifa, &enaddr))) {
866 			ia = ifatoia(ifa);
867 			ifa_ref(ifa);
868 			IF_ADDR_RUNLOCK(ifp);
869 			goto match;
870 		}
871 	IF_ADDR_RUNLOCK(ifp);
872 
873 	/*
874 	 * If bridging, fall back to using any inet address.
875 	 */
876 	IN_IFADDR_RLOCK(&in_ifa_tracker);
877 	if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
878 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
879 		goto drop;
880 	}
881 	ifa_ref(&ia->ia_ifa);
882 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
883 match:
884 	if (!enaddr)
885 		enaddr = (u_int8_t *)IF_LLADDR(ifp);
886 	carped = (ia->ia_ifa.ifa_carp != NULL);
887 	myaddr = ia->ia_addr.sin_addr;
888 	ifa_free(&ia->ia_ifa);
889 	if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
890 		goto drop;	/* it's from me, ignore it. */
891 	if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
892 		ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
893 		    "%s!\n", inet_ntoa(isaddr));
894 		goto drop;
895 	}
896 
897 	if (ifp->if_addrlen != ah->ar_hln) {
898 		ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
899 		    "i/f %d (ignored)\n", ifp->if_addrlen,
900 		    (u_char *) ar_sha(ah), ":", ah->ar_hln,
901 		    ifp->if_addrlen);
902 		goto drop;
903 	}
904 
905 	/*
906 	 * Warn if another host is using the same IP address, but only if the
907 	 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
908 	 * case we suppress the warning to avoid false positive complaints of
909 	 * potential misconfiguration.
910 	 */
911 	if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
912 	    myaddr.s_addr != 0) {
913 		ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
914 		   ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
915 		   inet_ntoa(isaddr), ifp->if_xname);
916 		itaddr = myaddr;
917 		ARPSTAT_INC(dupips);
918 		goto reply;
919 	}
920 	if (ifp->if_flags & IFF_STATICARP)
921 		goto reply;
922 
923 	bzero(&sin, sizeof(sin));
924 	sin.sin_len = sizeof(struct sockaddr_in);
925 	sin.sin_family = AF_INET;
926 	sin.sin_addr = isaddr;
927 	dst = (struct sockaddr *)&sin;
928 	IF_AFDATA_RLOCK(ifp);
929 	la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
930 	IF_AFDATA_RUNLOCK(ifp);
931 	if (la != NULL)
932 		arp_check_update_lle(ah, isaddr, ifp, bridged, la);
933 	else if (itaddr.s_addr == myaddr.s_addr) {
934 		/*
935 		 * Request/reply to our address, but no lle exists yet.
936 		 * Calculate full link prepend to use in lle.
937 		 */
938 		linkhdrsize = sizeof(linkhdr);
939 		if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
940 		    &linkhdrsize, &lladdr_off) != 0)
941 			goto reply;
942 
943 		/* Allocate new entry */
944 		la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
945 		if (la == NULL) {
946 
947 			/*
948 			 * lle creation may fail if source address belongs
949 			 * to non-directly connected subnet. However, we
950 			 * will try to answer the request instead of dropping
951 			 * frame.
952 			 */
953 			goto reply;
954 		}
955 		lltable_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
956 		    lladdr_off);
957 
958 		IF_AFDATA_WLOCK(ifp);
959 		LLE_WLOCK(la);
960 		la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
961 
962 		/*
963 		 * Check if lle still does not exists.
964 		 * If it does, that means that we either
965 		 * 1) have configured it explicitly, via
966 		 * 1a) 'arp -s' static entry or
967 		 * 1b) interface address static record
968 		 * or
969 		 * 2) it was the result of sending first packet to-host
970 		 * or
971 		 * 3) it was another arp reply packet we handled in
972 		 * different thread.
973 		 *
974 		 * In all cases except 3) we definitely need to prefer
975 		 * existing lle. For the sake of simplicity, prefer any
976 		 * existing lle over newly-create one.
977 		 */
978 		if (la_tmp == NULL)
979 			lltable_link_entry(LLTABLE(ifp), la);
980 		IF_AFDATA_WUNLOCK(ifp);
981 
982 		if (la_tmp == NULL) {
983 			arp_mark_lle_reachable(la);
984 			LLE_WUNLOCK(la);
985 		} else {
986 			/* Free newly-create entry and handle packet */
987 			lltable_free_entry(LLTABLE(ifp), la);
988 			la = la_tmp;
989 			la_tmp = NULL;
990 			arp_check_update_lle(ah, isaddr, ifp, bridged, la);
991 			/* arp_check_update_lle() returns @la unlocked */
992 		}
993 		la = NULL;
994 	}
995 reply:
996 	if (op != ARPOP_REQUEST)
997 		goto drop;
998 	ARPSTAT_INC(rxrequests);
999 
1000 	if (itaddr.s_addr == myaddr.s_addr) {
1001 		/* Shortcut.. the receiving interface is the target. */
1002 		(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1003 		(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1004 	} else {
1005 		struct llentry *lle = NULL;
1006 
1007 		sin.sin_addr = itaddr;
1008 		IF_AFDATA_RLOCK(ifp);
1009 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1010 		IF_AFDATA_RUNLOCK(ifp);
1011 
1012 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1013 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1014 			(void)memcpy(ar_sha(ah), lle->ll_addr, ah->ar_hln);
1015 			LLE_RUNLOCK(lle);
1016 		} else {
1017 
1018 			if (lle != NULL)
1019 				LLE_RUNLOCK(lle);
1020 
1021 			if (!V_arp_proxyall)
1022 				goto drop;
1023 
1024 			/* XXX MRT use table 0 for arp reply  */
1025 			if (fib4_lookup_nh_basic(0, itaddr, 0, 0, &nh4) != 0)
1026 				goto drop;
1027 
1028 			/*
1029 			 * Don't send proxies for nodes on the same interface
1030 			 * as this one came out of, or we'll get into a fight
1031 			 * over who claims what Ether address.
1032 			 */
1033 			if (nh4.nh_ifp == ifp)
1034 				goto drop;
1035 
1036 			(void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1037 			(void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1038 
1039 			/*
1040 			 * Also check that the node which sent the ARP packet
1041 			 * is on the interface we expect it to be on. This
1042 			 * avoids ARP chaos if an interface is connected to the
1043 			 * wrong network.
1044 			 */
1045 
1046 			/* XXX MRT use table 0 for arp checks */
1047 			if (fib4_lookup_nh_basic(0, isaddr, 0, 0, &nh4) != 0)
1048 				goto drop;
1049 			if (nh4.nh_ifp != ifp) {
1050 				ARP_LOG(LOG_INFO, "proxy: ignoring request"
1051 				    " from %s via %s\n",
1052 				    inet_ntoa(isaddr), ifp->if_xname);
1053 				goto drop;
1054 			}
1055 
1056 #ifdef DEBUG_PROXY
1057 			printf("arp: proxying for %s\n", inet_ntoa(itaddr));
1058 #endif
1059 		}
1060 	}
1061 
1062 	if (itaddr.s_addr == myaddr.s_addr &&
1063 	    IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
1064 		/* RFC 3927 link-local IPv4; always reply by broadcast. */
1065 #ifdef DEBUG_LINKLOCAL
1066 		printf("arp: sending reply for link-local addr %s\n",
1067 		    inet_ntoa(itaddr));
1068 #endif
1069 		m->m_flags |= M_BCAST;
1070 		m->m_flags &= ~M_MCAST;
1071 	} else {
1072 		/* default behaviour; never reply by broadcast. */
1073 		m->m_flags &= ~(M_BCAST|M_MCAST);
1074 	}
1075 	(void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1076 	(void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1077 	ah->ar_op = htons(ARPOP_REPLY);
1078 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1079 	m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1080 	m->m_pkthdr.len = m->m_len;
1081 	m->m_pkthdr.rcvif = NULL;
1082 	sa.sa_family = AF_ARP;
1083 	sa.sa_len = 2;
1084 
1085 	/* Calculate link header for sending frame */
1086 	bzero(&ro, sizeof(ro));
1087 	linkhdrsize = sizeof(linkhdr);
1088 	error = arp_fillheader(ifp, ah, 0, linkhdr, &linkhdrsize);
1089 
1090 	/*
1091 	 * arp_fillheader() may fail due to lack of support inside encap request
1092 	 * routing. This is not necessary an error, AF_ARP can/should be handled
1093 	 * by if_output().
1094 	 */
1095 	if (error != 0 && error != EAFNOSUPPORT) {
1096 		ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
1097 		    if_name(ifp), error);
1098 		return;
1099 	}
1100 
1101 	ro.ro_prepend = linkhdr;
1102 	ro.ro_plen = linkhdrsize;
1103 	ro.ro_flags = 0;
1104 
1105 	m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1106 	(*ifp->if_output)(ifp, m, &sa, &ro);
1107 	ARPSTAT_INC(txreplies);
1108 	return;
1109 
1110 drop:
1111 	m_freem(m);
1112 }
1113 #endif
1114 
1115 /*
1116  * Checks received arp data against existing @la.
1117  * Updates lle state/performs notification if necessary.
1118  */
1119 static void
arp_check_update_lle(struct arphdr * ah,struct in_addr isaddr,struct ifnet * ifp,int bridged,struct llentry * la)1120 arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, struct ifnet *ifp,
1121     int bridged, struct llentry *la)
1122 {
1123 	struct sockaddr sa;
1124 	struct mbuf *m_hold, *m_hold_next;
1125 	uint8_t linkhdr[LLE_MAX_LINKHDR];
1126 	size_t linkhdrsize;
1127 	int lladdr_off;
1128 
1129 	LLE_WLOCK_ASSERT(la);
1130 
1131 	/* the following is not an error when doing bridging */
1132 	if (!bridged && la->lle_tbl->llt_ifp != ifp) {
1133 		if (log_arp_wrong_iface)
1134 			ARP_LOG(LOG_WARNING, "%s is on %s "
1135 			    "but got reply from %*D on %s\n",
1136 			    inet_ntoa(isaddr),
1137 			    la->lle_tbl->llt_ifp->if_xname,
1138 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1139 			    ifp->if_xname);
1140 		LLE_WUNLOCK(la);
1141 		return;
1142 	}
1143 	if ((la->la_flags & LLE_VALID) &&
1144 	    bcmp(ar_sha(ah), la->ll_addr, ifp->if_addrlen)) {
1145 		if (la->la_flags & LLE_STATIC) {
1146 			LLE_WUNLOCK(la);
1147 			if (log_arp_permanent_modify)
1148 				ARP_LOG(LOG_ERR,
1149 				    "%*D attempts to modify "
1150 				    "permanent entry for %s on %s\n",
1151 				    ifp->if_addrlen,
1152 				    (u_char *)ar_sha(ah), ":",
1153 				    inet_ntoa(isaddr), ifp->if_xname);
1154 			return;
1155 		}
1156 		if (log_arp_movements) {
1157 			ARP_LOG(LOG_INFO, "%s moved from %*D "
1158 			    "to %*D on %s\n",
1159 			    inet_ntoa(isaddr),
1160 			    ifp->if_addrlen,
1161 			    (u_char *)&la->ll_addr, ":",
1162 			    ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1163 			    ifp->if_xname);
1164 		}
1165 	}
1166 
1167 	/* Calculate full link prepend to use in lle */
1168 	linkhdrsize = sizeof(linkhdr);
1169 	if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
1170 	    &linkhdrsize, &lladdr_off) != 0)
1171 		return;
1172 
1173 	/* Check if something has changed */
1174 	if (memcmp(la->r_linkdata, linkhdr, linkhdrsize) != 0 ||
1175 	    (la->la_flags & LLE_VALID) == 0) {
1176 		/* Try to perform LLE update */
1177 		if (lltable_try_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
1178 		    lladdr_off) == 0)
1179 			return;
1180 
1181 		/* Clear fast path feedback request if set */
1182 		la->r_skip_req = 0;
1183 	}
1184 
1185 	arp_mark_lle_reachable(la);
1186 
1187 	/*
1188 	 * The packets are all freed within the call to the output
1189 	 * routine.
1190 	 *
1191 	 * NB: The lock MUST be released before the call to the
1192 	 * output routine.
1193 	 */
1194 	if (la->la_hold != NULL) {
1195 		m_hold = la->la_hold;
1196 		la->la_hold = NULL;
1197 		la->la_numheld = 0;
1198 		lltable_fill_sa_entry(la, &sa);
1199 		LLE_WUNLOCK(la);
1200 		for (; m_hold != NULL; m_hold = m_hold_next) {
1201 			m_hold_next = m_hold->m_nextpkt;
1202 			m_hold->m_nextpkt = NULL;
1203 			/* Avoid confusing lower layers. */
1204 			m_clrprotoflags(m_hold);
1205 			(*ifp->if_output)(ifp, m_hold, &sa, NULL);
1206 		}
1207 	} else
1208 		LLE_WUNLOCK(la);
1209 }
1210 
1211 static void
arp_mark_lle_reachable(struct llentry * la)1212 arp_mark_lle_reachable(struct llentry *la)
1213 {
1214 	int canceled, wtime;
1215 
1216 	LLE_WLOCK_ASSERT(la);
1217 
1218 	la->ln_state = ARP_LLINFO_REACHABLE;
1219 	EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
1220 
1221 	if (!(la->la_flags & LLE_STATIC)) {
1222 		LLE_ADDREF(la);
1223 		la->la_expire = time_uptime + V_arpt_keep;
1224 		wtime = V_arpt_keep - V_arp_maxtries * V_arpt_rexmit;
1225 		if (wtime < 0)
1226 			wtime = V_arpt_keep;
1227 		canceled = callout_reset(&la->lle_timer,
1228 		    hz * wtime, arptimer, la);
1229 		if (canceled)
1230 			LLE_REMREF(la);
1231 	}
1232 	la->la_asked = 0;
1233 	la->la_preempt = V_arp_maxtries;
1234 }
1235 
1236 /*
1237  * Add pernament link-layer record for given interface address.
1238  */
1239 static __noinline void
arp_add_ifa_lle(struct ifnet * ifp,const struct sockaddr * dst)1240 arp_add_ifa_lle(struct ifnet *ifp, const struct sockaddr *dst)
1241 {
1242 	struct llentry *lle, *lle_tmp;
1243 
1244 	/*
1245 	 * Interface address LLE record is considered static
1246 	 * because kernel code relies on LLE_STATIC flag to check
1247 	 * if these entries can be rewriten by arp updates.
1248 	 */
1249 	lle = lltable_alloc_entry(LLTABLE(ifp), LLE_IFADDR | LLE_STATIC, dst);
1250 	if (lle == NULL) {
1251 		log(LOG_INFO, "arp_ifinit: cannot create arp "
1252 		    "entry for interface address\n");
1253 		return;
1254 	}
1255 
1256 	IF_AFDATA_WLOCK(ifp);
1257 	LLE_WLOCK(lle);
1258 	/* Unlink any entry if exists */
1259 	lle_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
1260 	if (lle_tmp != NULL)
1261 		lltable_unlink_entry(LLTABLE(ifp), lle_tmp);
1262 
1263 	lltable_link_entry(LLTABLE(ifp), lle);
1264 	IF_AFDATA_WUNLOCK(ifp);
1265 
1266 	if (lle_tmp != NULL)
1267 		EVENTHANDLER_INVOKE(lle_event, lle_tmp, LLENTRY_EXPIRED);
1268 
1269 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED);
1270 	LLE_WUNLOCK(lle);
1271 	if (lle_tmp != NULL)
1272 		lltable_free_entry(LLTABLE(ifp), lle_tmp);
1273 }
1274 
1275 void
arp_ifinit(struct ifnet * ifp,struct ifaddr * ifa)1276 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1277 {
1278 	const struct sockaddr_in *dst_in;
1279 	const struct sockaddr *dst;
1280 
1281 	if (ifa->ifa_carp != NULL)
1282 		return;
1283 
1284 	dst = ifa->ifa_addr;
1285 	dst_in = (const struct sockaddr_in *)dst;
1286 
1287 	if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY)
1288 		return;
1289 	arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp));
1290 
1291 	arp_add_ifa_lle(ifp, dst);
1292 }
1293 
1294 void
arp_announce_ifaddr(struct ifnet * ifp,struct in_addr addr,u_char * enaddr)1295 arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr)
1296 {
1297 
1298 	if (ntohl(addr.s_addr) != INADDR_ANY)
1299 		arprequest(ifp, &addr, &addr, enaddr);
1300 }
1301 
1302 /*
1303  * Sends gratuitous ARPs for each ifaddr to notify other
1304  * nodes about the address change.
1305  */
1306 static __noinline void
arp_handle_ifllchange(struct ifnet * ifp)1307 arp_handle_ifllchange(struct ifnet *ifp)
1308 {
1309 	struct ifaddr *ifa;
1310 
1311 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1312 		if (ifa->ifa_addr->sa_family == AF_INET)
1313 			arp_ifinit(ifp, ifa);
1314 	}
1315 }
1316 
1317 
1318 /*
1319  * A handler for interface link layer address change event.
1320  */
1321 static void
arp_iflladdr(void * arg __unused,struct ifnet * ifp)1322 arp_iflladdr(void *arg __unused, struct ifnet *ifp)
1323 {
1324 
1325 	lltable_update_ifaddr(LLTABLE(ifp));
1326 
1327 	if ((ifp->if_flags & IFF_UP) != 0)
1328 		arp_handle_ifllchange(ifp);
1329 }
1330 
1331 static void
arp_init(void)1332 arp_init(void)
1333 {
1334 
1335 	netisr_register(&arp_nh);
1336 	if (IS_DEFAULT_VNET(curvnet))
1337 		iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event,
1338 		    arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
1339 }
1340 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
1341