xref: /original-bsd/sys/netinet/if_ether.c (revision 7cab520e)
1 /*
2  * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)if_ether.c	7.8 (Berkeley) 10/12/88
18  */
19 
20 /*
21  * Ethernet address resolution protocol.
22  * TODO:
23  *	run at splnet (add ARP protocol intr.)
24  *	link entries onto hash chains, keep free list
25  *	add "inuse/lock" bit (or ref. count) along with valid bit
26  */
27 
28 #include "param.h"
29 #include "systm.h"
30 #include "malloc.h"
31 #include "mbuf.h"
32 #include "socket.h"
33 #include "time.h"
34 #include "kernel.h"
35 #include "errno.h"
36 #include "ioctl.h"
37 #include "syslog.h"
38 
39 #include "../net/if.h"
40 #include "in.h"
41 #include "in_systm.h"
42 #include "ip.h"
43 #include "if_ether.h"
44 
45 #ifdef GATEWAY
46 #define	ARPTAB_BSIZ	16		/* bucket size */
47 #define	ARPTAB_NB	37		/* number of buckets */
48 #else
49 #define	ARPTAB_BSIZ	9		/* bucket size */
50 #define	ARPTAB_NB	19		/* number of buckets */
51 #endif
52 #define	ARPTAB_SIZE	(ARPTAB_BSIZ * ARPTAB_NB)
53 struct	arptab arptab[ARPTAB_SIZE];
54 int	arptab_size = ARPTAB_SIZE;	/* for arp command */
55 
56 /*
57  * ARP trailer negotiation.  Trailer protocol is not IP specific,
58  * but ARP request/response use IP addresses.
59  */
60 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
61 
62 #define	ARPTAB_HASH(a) \
63 	((u_long)(a) % ARPTAB_NB)
64 
65 #define	ARPTAB_LOOK(at,addr) { \
66 	register n; \
67 	at = &arptab[ARPTAB_HASH(addr) * ARPTAB_BSIZ]; \
68 	for (n = 0 ; n < ARPTAB_BSIZ ; n++,at++) \
69 		if (at->at_iaddr.s_addr == addr) \
70 			break; \
71 	if (n >= ARPTAB_BSIZ) \
72 		at = 0; \
73 }
74 
75 /* timer values */
76 #define	ARPT_AGE	(60*1)	/* aging timer, 1 min. */
77 #define	ARPT_KILLC	20	/* kill completed entry in 20 mins. */
78 #define	ARPT_KILLI	3	/* kill incomplete entry in 3 minutes */
79 
80 extern struct ifnet loif;
81 
82 /*
83  * Timeout routine.  Age arp_tab entries once a minute.
84  */
85 arptimer()
86 {
87 	register struct arptab *at;
88 	register i;
89 
90 	timeout(arptimer, (caddr_t)0, ARPT_AGE * hz);
91 	at = &arptab[0];
92 	for (i = 0; i < ARPTAB_SIZE; i++, at++) {
93 		if (at->at_flags == 0 || (at->at_flags & ATF_PERM))
94 			continue;
95 		if (++at->at_timer < ((at->at_flags&ATF_COM) ?
96 		    ARPT_KILLC : ARPT_KILLI))
97 			continue;
98 		/* timer has expired, clear entry */
99 		arptfree(at);
100 	}
101 }
102 
103 /*
104  * Broadcast an ARP packet, asking who has addr on interface ac.
105  */
106 arpwhohas(ac, addr)
107 	register struct arpcom *ac;
108 	struct in_addr *addr;
109 {
110 	register struct mbuf *m;
111 	register struct ether_header *eh;
112 	register struct ether_arp *ea;
113 	struct sockaddr sa;
114 
115 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
116 		return;
117 	m->m_len = sizeof(*ea);
118 	m->m_pkthdr.len = sizeof(*ea);
119 	MH_ALIGN(m, sizeof(*ea));
120 	ea = mtod(m, struct ether_arp *);
121 	eh = (struct ether_header *)sa.sa_data;
122 	bzero((caddr_t)ea, sizeof (*ea));
123 	bcopy((caddr_t)etherbroadcastaddr, (caddr_t)eh->ether_dhost,
124 	    sizeof(eh->ether_dhost));
125 	eh->ether_type = ETHERTYPE_ARP;		/* if_output will swap */
126 	ea->arp_hrd = htons(ARPHRD_ETHER);
127 	ea->arp_pro = htons(ETHERTYPE_IP);
128 	ea->arp_hln = sizeof(ea->arp_sha);	/* hardware address length */
129 	ea->arp_pln = sizeof(ea->arp_spa);	/* protocol address length */
130 	ea->arp_op = htons(ARPOP_REQUEST);
131 	bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_sha,
132 	   sizeof(ea->arp_sha));
133 	bcopy((caddr_t)&ac->ac_ipaddr, (caddr_t)ea->arp_spa,
134 	   sizeof(ea->arp_spa));
135 	bcopy((caddr_t)addr, (caddr_t)ea->arp_tpa, sizeof(ea->arp_tpa));
136 	sa.sa_family = AF_UNSPEC;
137 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa);
138 }
139 
140 int	useloopback = 1;	/* use loopback interface for local traffic */
141 
142 /*
143  * Resolve an IP address into an ethernet address.  If success,
144  * desten is filled in.  If there is no entry in arptab,
145  * set one up and broadcast a request for the IP address.
146  * Hold onto this mbuf and resend it once the address
147  * is finally resolved.  A return value of 1 indicates
148  * that desten has been filled in and the packet should be sent
149  * normally; a 0 return indicates that the packet has been
150  * taken over here, either now or for later transmission.
151  *
152  * We do some (conservative) locking here at splimp, since
153  * arptab is also altered from input interrupt service (ecintr/ilintr
154  * calls arpinput when ETHERTYPE_ARP packets come in).
155  */
156 arpresolve(ac, m, destip, desten, usetrailers)
157 	register struct arpcom *ac;
158 	struct mbuf *m;
159 	register struct in_addr *destip;
160 	register u_char *desten;
161 	int *usetrailers;
162 {
163 	register struct arptab *at;
164 	struct sockaddr_in sin;
165 	u_long lna;
166 	int s;
167 
168 	*usetrailers = 0;
169 	if (m->m_flags & M_BCAST) {	/* broadcast */
170 		bcopy((caddr_t)etherbroadcastaddr, (caddr_t)desten,
171 		    sizeof(etherbroadcastaddr));
172 		return (1);
173 	}
174 	lna = in_lnaof(*destip);
175 	/* if for us, use software loopback driver if up */
176 	if (destip->s_addr == ac->ac_ipaddr.s_addr) {
177 		/*
178 		 * This test used to be
179 		 *	if (loif.if_flags & IFF_UP)
180 		 * It allowed local traffic to be forced
181 		 * through the hardware by configuring the loopback down.
182 		 * However, it causes problems during network configuration
183 		 * for boards that can't receive packets they send.
184 		 * It is now necessary to clear "useloopback"
185 		 * to force traffic out to the hardware.
186 		 */
187 		if (useloopback) {
188 			sin.sin_family = AF_INET;
189 			sin.sin_addr = *destip;
190 			(void) looutput(&loif, m, (struct sockaddr *)&sin);
191 			/*
192 			 * The packet has already been sent and freed.
193 			 */
194 			return (0);
195 		} else {
196 			bcopy((caddr_t)ac->ac_enaddr, (caddr_t)desten,
197 			    sizeof(ac->ac_enaddr));
198 			return (1);
199 		}
200 	}
201 	s = splimp();
202 	ARPTAB_LOOK(at, destip->s_addr);
203 	if (at == 0) {			/* not found */
204 		if (ac->ac_if.if_flags & IFF_NOARP) {
205 			bcopy((caddr_t)ac->ac_enaddr, (caddr_t)desten, 3);
206 			desten[3] = (lna >> 16) & 0x7f;
207 			desten[4] = (lna >> 8) & 0xff;
208 			desten[5] = lna & 0xff;
209 			splx(s);
210 			return (1);
211 		} else {
212 			at = arptnew(destip);
213 			if (at == 0)
214 				panic("arpresolve: no free entry");
215 			at->at_hold = m;
216 			arpwhohas(ac, destip);
217 			splx(s);
218 			return (0);
219 		}
220 	}
221 	at->at_timer = 0;		/* restart the timer */
222 	if (at->at_flags & ATF_COM) {	/* entry IS complete */
223 		bcopy((caddr_t)at->at_enaddr, (caddr_t)desten,
224 		    sizeof(at->at_enaddr));
225 		if (at->at_flags & ATF_USETRAILERS)
226 			*usetrailers = 1;
227 		splx(s);
228 		return (1);
229 	}
230 	/*
231 	 * There is an arptab entry, but no ethernet address
232 	 * response yet.  Replace the held mbuf with this
233 	 * latest one.
234 	 */
235 	if (at->at_hold)
236 		m_freem(at->at_hold);
237 	at->at_hold = m;
238 	arpwhohas(ac, destip);		/* ask again */
239 	splx(s);
240 	return (0);
241 }
242 
243 /*
244  * Called from 10 Mb/s Ethernet interrupt handlers
245  * when ether packet type ETHERTYPE_ARP
246  * is received.  Common length and type checks are done here,
247  * then the protocol-specific routine is called.
248  */
249 arpinput(ac, m)
250 	struct arpcom *ac;
251 	struct mbuf *m;
252 {
253 	register struct arphdr *ar;
254 
255 	if (ac->ac_if.if_flags & IFF_NOARP)
256 		goto out;
257 	if (m->m_len < sizeof(struct arphdr))
258 		goto out;
259 	ar = mtod(m, struct arphdr *);
260 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
261 		goto out;
262 	if (m->m_len < sizeof(struct arphdr) + 2 * ar->ar_hln + 2 * ar->ar_pln)
263 		goto out;
264 
265 	switch (ntohs(ar->ar_pro)) {
266 
267 	case ETHERTYPE_IP:
268 	case ETHERTYPE_IPTRAILERS:
269 		in_arpinput(ac, m);
270 		return;
271 
272 	default:
273 		break;
274 	}
275 out:
276 	m_freem(m);
277 }
278 
279 /*
280  * ARP for Internet protocols on 10 Mb/s Ethernet.
281  * Algorithm is that given in RFC 826.
282  * In addition, a sanity check is performed on the sender
283  * protocol address, to catch impersonators.
284  * We also handle negotiations for use of trailer protocol:
285  * ARP replies for protocol type ETHERTYPE_TRAIL are sent
286  * along with IP replies if we want trailers sent to us,
287  * and also send them in response to IP replies.
288  * This allows either end to announce the desire to receive
289  * trailer packets.
290  * We reply to requests for ETHERTYPE_TRAIL protocol as well,
291  * but don't normally send requests.
292  */
293 in_arpinput(ac, m)
294 	register struct arpcom *ac;
295 	struct mbuf *m;
296 {
297 	register struct ether_arp *ea;
298 	struct ether_header *eh;
299 	register struct arptab *at;  /* same as "merge" flag */
300 	struct mbuf *mcopy = 0;
301 	struct sockaddr_in sin;
302 	struct sockaddr sa;
303 	struct in_addr isaddr, itaddr, myaddr;
304 	int proto, op, s, completed = 0;
305 
306 	myaddr = ac->ac_ipaddr;
307 	ea = mtod(m, struct ether_arp *);
308 	proto = ntohs(ea->arp_pro);
309 	op = ntohs(ea->arp_op);
310 	bcopy((caddr_t)ea->arp_spa, (caddr_t)&isaddr, sizeof (isaddr));
311 	bcopy((caddr_t)ea->arp_tpa, (caddr_t)&itaddr, sizeof (itaddr));
312 	if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)ac->ac_enaddr,
313 	    sizeof (ea->arp_sha)))
314 		goto out;	/* it's from me, ignore it. */
315 	if (!bcmp((caddr_t)ea->arp_sha, (caddr_t)etherbroadcastaddr,
316 	    sizeof (ea->arp_sha))) {
317 		log(LOG_ERR,
318 		    "arp: ether address is broadcast for IP address %x!\n",
319 		    ntohl(isaddr.s_addr));
320 		goto out;
321 	}
322 	if (isaddr.s_addr == myaddr.s_addr) {
323 		log(LOG_ERR, "%s: %s\n",
324 			"duplicate IP address!! sent from ethernet address",
325 			ether_sprintf(ea->arp_sha));
326 		itaddr = myaddr;
327 		if (op == ARPOP_REQUEST)
328 			goto reply;
329 		goto out;
330 	}
331 	s = splimp();
332 	ARPTAB_LOOK(at, isaddr.s_addr);
333 	if (at) {
334 		bcopy((caddr_t)ea->arp_sha, (caddr_t)at->at_enaddr,
335 		    sizeof(ea->arp_sha));
336 		if ((at->at_flags & ATF_COM) == 0)
337 			completed = 1;
338 		at->at_flags |= ATF_COM;
339 		if (at->at_hold) {
340 			sin.sin_family = AF_INET;
341 			sin.sin_addr = isaddr;
342 			(*ac->ac_if.if_output)(&ac->ac_if,
343 			    at->at_hold, (struct sockaddr *)&sin);
344 			at->at_hold = 0;
345 		}
346 	}
347 	if (at == 0 && itaddr.s_addr == myaddr.s_addr) {
348 		/* ensure we have a table entry */
349 		if (at = arptnew(&isaddr)) {
350 			bcopy((caddr_t)ea->arp_sha, (caddr_t)at->at_enaddr,
351 			    sizeof(ea->arp_sha));
352 			completed = 1;
353 			at->at_flags |= ATF_COM;
354 		}
355 	}
356 	splx(s);
357 reply:
358 	switch (proto) {
359 
360 	case ETHERTYPE_IPTRAILERS:
361 		/* partner says trailers are OK */
362 		if (at)
363 			at->at_flags |= ATF_USETRAILERS;
364 		/*
365 		 * Reply to request iff we want trailers.
366 		 */
367 		if (op != ARPOP_REQUEST || ac->ac_if.if_flags & IFF_NOTRAILERS)
368 			goto out;
369 		break;
370 
371 	case ETHERTYPE_IP:
372 		/*
373 		 * Reply if this is an IP request,
374 		 * or if we want to send a trailer response.
375 		 * Send the latter only to the IP response
376 		 * that completes the current ARP entry.
377 		 */
378 		if (op != ARPOP_REQUEST &&
379 		    (completed == 0 || ac->ac_if.if_flags & IFF_NOTRAILERS))
380 			goto out;
381 	}
382 	if (itaddr.s_addr == myaddr.s_addr) {
383 		/* I am the target */
384 		bcopy((caddr_t)ea->arp_sha, (caddr_t)ea->arp_tha,
385 		    sizeof(ea->arp_sha));
386 		bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->arp_sha,
387 		    sizeof(ea->arp_sha));
388 	} else {
389 		ARPTAB_LOOK(at, itaddr.s_addr);
390 		if (at == NULL || (at->at_flags & ATF_PUBL) == 0)
391 			goto out;
392 		bcopy((caddr_t)ea->arp_sha, (caddr_t)ea->arp_tha,
393 		    sizeof(ea->arp_sha));
394 		bcopy((caddr_t)at->at_enaddr, (caddr_t)ea->arp_sha,
395 		    sizeof(ea->arp_sha));
396 	}
397 
398 	bcopy((caddr_t)ea->arp_spa, (caddr_t)ea->arp_tpa,
399 	    sizeof(ea->arp_spa));
400 	bcopy((caddr_t)&itaddr, (caddr_t)ea->arp_spa,
401 	    sizeof(ea->arp_spa));
402 	ea->arp_op = htons(ARPOP_REPLY);
403 	/*
404 	 * If incoming packet was an IP reply,
405 	 * we are sending a reply for type IPTRAILERS.
406 	 * If we are sending a reply for type IP
407 	 * and we want to receive trailers,
408 	 * send a trailer reply as well.
409 	 */
410 	if (op == ARPOP_REPLY)
411 		ea->arp_pro = htons(ETHERTYPE_IPTRAILERS);
412 	else if (proto == ETHERTYPE_IP &&
413 	    (ac->ac_if.if_flags & IFF_NOTRAILERS) == 0)
414 		mcopy = m_copy(m, 0, (int)M_COPYALL);
415 	eh = (struct ether_header *)sa.sa_data;
416 	bcopy((caddr_t)ea->arp_tha, (caddr_t)eh->ether_dhost,
417 	    sizeof(eh->ether_dhost));
418 	eh->ether_type = ETHERTYPE_ARP;
419 	sa.sa_family = AF_UNSPEC;
420 	(*ac->ac_if.if_output)(&ac->ac_if, m, &sa);
421 	if (mcopy) {
422 		ea = mtod(mcopy, struct ether_arp *);
423 		ea->arp_pro = htons(ETHERTYPE_IPTRAILERS);
424 		(*ac->ac_if.if_output)(&ac->ac_if, mcopy, &sa);
425 	}
426 	return;
427 out:
428 	m_freem(m);
429 	return;
430 }
431 
432 /*
433  * Free an arptab entry.
434  */
435 arptfree(at)
436 	register struct arptab *at;
437 {
438 	int s = splimp();
439 
440 	if (at->at_hold)
441 		m_freem(at->at_hold);
442 	at->at_hold = 0;
443 	at->at_timer = at->at_flags = 0;
444 	at->at_iaddr.s_addr = 0;
445 	splx(s);
446 }
447 
448 /*
449  * Enter a new address in arptab, pushing out the oldest entry
450  * from the bucket if there is no room.
451  * This always succeeds since no bucket can be completely filled
452  * with permanent entries (except from arpioctl when testing whether
453  * another permanent entry will fit).
454  * MUST BE CALLED AT SPLIMP.
455  */
456 struct arptab *
457 arptnew(addr)
458 	struct in_addr *addr;
459 {
460 	register n;
461 	int oldest = -1;
462 	register struct arptab *at, *ato = NULL;
463 	static int first = 1;
464 
465 	if (first) {
466 		first = 0;
467 		timeout(arptimer, (caddr_t)0, hz);
468 	}
469 	at = &arptab[ARPTAB_HASH(addr->s_addr) * ARPTAB_BSIZ];
470 	for (n = 0; n < ARPTAB_BSIZ; n++,at++) {
471 		if (at->at_flags == 0)
472 			goto out;	 /* found an empty entry */
473 		if (at->at_flags & ATF_PERM)
474 			continue;
475 		if ((int) at->at_timer > oldest) {
476 			oldest = at->at_timer;
477 			ato = at;
478 		}
479 	}
480 	if (ato == NULL)
481 		return (NULL);
482 	at = ato;
483 	arptfree(at);
484 out:
485 	at->at_iaddr = *addr;
486 	at->at_flags = ATF_INUSE;
487 	return (at);
488 }
489 
490 arpioctl(cmd, data)
491 	int cmd;
492 	caddr_t data;
493 {
494 	register struct arpreq *ar = (struct arpreq *)data;
495 	register struct arptab *at;
496 	register struct sockaddr_in *sin;
497 	int s;
498 
499 	if (ar->arp_pa.sa_family != AF_INET ||
500 	    ar->arp_ha.sa_family != AF_UNSPEC)
501 		return (EAFNOSUPPORT);
502 	sin = (struct sockaddr_in *)&ar->arp_pa;
503 	s = splimp();
504 	ARPTAB_LOOK(at, sin->sin_addr.s_addr);
505 	if (at == NULL) {		/* not found */
506 		if (cmd != SIOCSARP) {
507 			splx(s);
508 			return (ENXIO);
509 		}
510 		if (ifa_ifwithnet(&ar->arp_pa) == NULL) {
511 			splx(s);
512 			return (ENETUNREACH);
513 		}
514 	}
515 	switch (cmd) {
516 
517 	case SIOCSARP:		/* set entry */
518 		if (at == NULL) {
519 			at = arptnew(&sin->sin_addr);
520 			if (at == NULL) {
521 				splx(s);
522 				return (EADDRNOTAVAIL);
523 			}
524 			if (ar->arp_flags & ATF_PERM) {
525 			/* never make all entries in a bucket permanent */
526 				register struct arptab *tat;
527 
528 				/* try to re-allocate */
529 				tat = arptnew(&sin->sin_addr);
530 				if (tat == NULL) {
531 					arptfree(at);
532 					splx(s);
533 					return (EADDRNOTAVAIL);
534 				}
535 				arptfree(tat);
536 			}
537 		}
538 		bcopy((caddr_t)ar->arp_ha.sa_data, (caddr_t)at->at_enaddr,
539 		    sizeof(at->at_enaddr));
540 		at->at_flags = ATF_COM | ATF_INUSE |
541 			(ar->arp_flags & (ATF_PERM|ATF_PUBL|ATF_USETRAILERS));
542 		at->at_timer = 0;
543 		break;
544 
545 	case SIOCDARP:		/* delete entry */
546 		arptfree(at);
547 		break;
548 
549 	case SIOCGARP:		/* get entry */
550 		bcopy((caddr_t)at->at_enaddr, (caddr_t)ar->arp_ha.sa_data,
551 		    sizeof(at->at_enaddr));
552 		ar->arp_flags = at->at_flags;
553 		break;
554 	}
555 	splx(s);
556 	return (0);
557 }
558