xref: /src/sys/netinet6/in6_ifattach.c (revision b55bffeaaf9bae5dc7aa21eae441d89c999ebab8)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
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 project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: in6_ifattach.c,v 1.118 2001/05/24 07:44:00 itojun Exp $
32  */
33 
34 #include <sys/param.h>
35 #include <sys/eventhandler.h>
36 #include <sys/systm.h>
37 #include <sys/counter.h>
38 #include <sys/malloc.h>
39 #include <sys/socket.h>
40 #include <sys/sockio.h>
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/proc.h>
45 #include <sys/rmlock.h>
46 #include <sys/syslog.h>
47 #include <sys/md5.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/if_dl.h>
52 #include <net/if_private.h>
53 #include <net/if_types.h>
54 #include <net/if_llatbl.h>
55 #include <net/route.h>
56 #include <net/vnet.h>
57 
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/if_ether.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/udp.h>
64 #include <netinet/udp_var.h>
65 
66 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #include <netinet6/in6_var.h>
69 #include <netinet6/in6_pcb.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/nd6.h>
73 #include <netinet6/mld6_var.h>
74 #include <netinet6/scope6_var.h>
75 
76 #include <crypto/sha2/sha256.h>
77 #include <machine/atomic.h>
78 
79 #ifdef IP6_AUTO_LINKLOCAL
80 VNET_DEFINE(int, ip6_auto_linklocal) = IP6_AUTO_LINKLOCAL;
81 #else
82 VNET_DEFINE(int, ip6_auto_linklocal) = 1;	/* enabled by default */
83 #endif
84 
85 VNET_DEFINE(struct callout, in6_tmpaddrtimer_ch);
86 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
87 
88 VNET_DEFINE(int, ip6_stableaddr_netifsource) = IP6_STABLEADDR_NETIFSRC_NAME; /* Use interface name by default */
89 
90 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
91 #define	V_ripcbinfo			VNET(ripcbinfo)
92 
93 static int get_rand_ifid(struct ifnet *, struct in6_addr *);
94 static int in6_ifattach_linklocal(struct ifnet *, struct ifnet *);
95 static int in6_ifattach_loopback(struct ifnet *);
96 static void in6_purgemaddrs(struct ifnet *);
97 
98 #define EUI64_GBIT	0x01
99 #define EUI64_UBIT	0x02
100 #define EUI64_TO_IFID(in6)	do {(in6)->s6_addr[8] ^= EUI64_UBIT; } while (0)
101 #define EUI64_GROUP(in6)	((in6)->s6_addr[8] & EUI64_GBIT)
102 #define EUI64_INDIVIDUAL(in6)	(!EUI64_GROUP(in6))
103 #define EUI64_LOCAL(in6)	((in6)->s6_addr[8] & EUI64_UBIT)
104 #define EUI64_UNIVERSAL(in6)	(!EUI64_LOCAL(in6))
105 
106 #define IFID_LOCAL(in6)		(!EUI64_LOCAL(in6))
107 #define IFID_UNIVERSAL(in6)	(!EUI64_UNIVERSAL(in6))
108 
109 #define HMAC_IPAD	0x36
110 #define HMAC_OPAD	0x5C
111 
112 /*
113  * Generate a last-resort interface identifier, when the machine has no
114  * IEEE802/EUI64 address sources.
115  * The goal here is to get an interface identifier that is
116  * (1) random enough and (2) does not change across reboot.
117  * We currently use MD5(hostname) for it.
118  *
119  * in6 - upper 64bits are preserved
120  */
121 static int
get_rand_ifid(struct ifnet * ifp,struct in6_addr * in6)122 get_rand_ifid(struct ifnet *ifp, struct in6_addr *in6)
123 {
124 	MD5_CTX ctxt;
125 	struct prison *pr;
126 	u_int8_t digest[16];
127 	int hostnamelen;
128 
129 	pr = curthread->td_ucred->cr_prison;
130 	mtx_lock(&pr->pr_mtx);
131 	hostnamelen = strlen(pr->pr_hostname);
132 #if 0
133 	/* we need at least several letters as seed for ifid */
134 	if (hostnamelen < 3) {
135 		mtx_unlock(&pr->pr_mtx);
136 		return -1;
137 	}
138 #endif
139 
140 	/* generate 8 bytes of pseudo-random value. */
141 	bzero(&ctxt, sizeof(ctxt));
142 	MD5Init(&ctxt);
143 	MD5Update(&ctxt, pr->pr_hostname, hostnamelen);
144 	mtx_unlock(&pr->pr_mtx);
145 	MD5Final(digest, &ctxt);
146 
147 	/* assumes sizeof(digest) > sizeof(ifid) */
148 	bcopy(digest, &in6->s6_addr[8], 8);
149 
150 	/* make sure to set "u" bit to local, and "g" bit to individual. */
151 	in6->s6_addr[8] &= ~EUI64_GBIT;	/* g bit to "individual" */
152 	in6->s6_addr[8] |= EUI64_UBIT;	/* u bit to "local" */
153 
154 	/* convert EUI64 into IPv6 interface identifier */
155 	EUI64_TO_IFID(in6);
156 
157 	return 0;
158 }
159 
160 
161 /*
162  * Get interface link level sockaddr
163  */
164 static struct sockaddr_dl *
get_interface_link_level(struct ifnet * ifp)165 get_interface_link_level(struct ifnet *ifp)
166 {
167 	struct ifaddr *ifa;
168 	struct sockaddr_dl *sdl;
169 
170 	NET_EPOCH_ASSERT();
171 
172 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
173 		if (ifa->ifa_addr->sa_family != AF_LINK)
174 			continue;
175 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
176 		if (sdl == NULL)
177 			continue;
178 		if (sdl->sdl_alen == 0)
179 			continue;
180 
181 		return (sdl);
182 	}
183 
184 	return (NULL);
185 }
186 
187 /*
188  * Get hwaddr from link interface
189  */
190 static uint8_t *
in6_get_interface_hwaddr(struct ifnet * ifp,size_t * len)191 in6_get_interface_hwaddr(struct ifnet *ifp, size_t *len)
192 {
193 	struct sockaddr_dl *sdl;
194 	u_int8_t *addr;
195 	static u_int8_t allzero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
196 	static u_int8_t allone[8] =
197 		{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
198 
199 	sdl = get_interface_link_level(ifp);
200 	if (sdl == NULL)
201 		return (NULL);
202 
203 	addr = LLADDR(sdl);
204 	*len = sdl->sdl_alen;
205 
206 	/* get EUI64 */
207 	switch (ifp->if_type) {
208 	case IFT_BRIDGE:
209 	case IFT_ETHER:
210 	case IFT_L2VLAN:
211 	case IFT_ATM:
212 	case IFT_IEEE1394:
213 		/* IEEE802/EUI64 cases - what others? */
214 		/* IEEE1394 uses 16byte length address starting with EUI64 */
215 		if (*len > 8)
216 			*len = 8;
217 
218 		/* look at IEEE802/EUI64 only */
219 		if (*len != 8 && *len != 6)
220 			return (NULL);
221 
222 		/*
223 		 * check for invalid MAC address - on bsdi, we see it a lot
224 		 * since wildboar configures all-zero MAC on pccard before
225 		 * card insertion.
226 		 */
227 		if (memcmp(addr, allzero, *len) == 0 || memcmp(addr, allone, *len) == 0)
228 			return (NULL);
229 
230 		break;
231 
232 	case IFT_GIF:
233 	case IFT_STF:
234 		/*
235 		 * RFC2893 says: "SHOULD use IPv4 address as ifid source".
236 		 * however, IPv4 address is not very suitable as unique
237 		 * identifier source (can be renumbered).
238 		 * we don't do this.
239 		 */
240 		return (NULL);
241 
242 	case IFT_INFINIBAND:
243 		if (*len != 20)
244 			return (NULL);
245 		*len = 8;
246 		addr += 12;
247 		break;
248 
249 	default:
250 		return (NULL);
251 	}
252 
253 	return (addr);
254 }
255 
256 /*
257  * Get interface identifier for the specified interface.
258  * XXX assumes single sockaddr_dl (AF_LINK address) per an interface
259  *
260  * in6 - upper 64bits are preserved
261  */
262 int
in6_get_hw_ifid(struct ifnet * ifp,struct in6_addr * in6)263 in6_get_hw_ifid(struct ifnet *ifp, struct in6_addr *in6)
264 {
265 	size_t hwaddr_len;
266 	uint8_t *hwaddr;
267 	static u_int8_t allzero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
268 
269 	hwaddr = in6_get_interface_hwaddr(ifp, &hwaddr_len);
270 	if (hwaddr == NULL || (hwaddr_len != 6 && hwaddr_len != 8))
271 		return (-1);
272 
273 	/* make EUI64 address */
274 	if (hwaddr_len == 8)
275 		memcpy(&in6->s6_addr[8], hwaddr, 8);
276 	else if (hwaddr_len == 6) {
277 		in6->s6_addr[8] = hwaddr[0];
278 		in6->s6_addr[9] = hwaddr[1];
279 		in6->s6_addr[10] = hwaddr[2];
280 		in6->s6_addr[11] = 0xff;
281 		in6->s6_addr[12] = 0xfe;
282 		in6->s6_addr[13] = hwaddr[3];
283 		in6->s6_addr[14] = hwaddr[4];
284 		in6->s6_addr[15] = hwaddr[5];
285 	}
286 
287 	/* sanity check: g bit must not indicate "group" */
288 	if (EUI64_GROUP(in6))
289 		return -1;
290 
291 	/* convert EUI64 into IPv6 interface identifier */
292 	EUI64_TO_IFID(in6);
293 
294 	/*
295 	 * sanity check: ifid must not be all zero, avoid conflict with
296 	 * subnet router anycast
297 	 */
298 	if ((in6->s6_addr[8] & ~(EUI64_GBIT | EUI64_UBIT)) == 0x00 &&
299 	    bcmp(&in6->s6_addr[9], allzero, 7) == 0)
300 		return -1;
301 
302 	return 0;
303 }
304 
305 /*
306  * Validate generated interface id to make sure it does not fall in any reserved range:
307  *
308  * https://www.iana.org/assignments/ipv6-interface-ids/ipv6-interface-ids.xhtml
309  */
310 static bool
validate_ifid(uint8_t * iid)311 validate_ifid(uint8_t *iid)
312 {
313 	static uint8_t allzero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
314 	static uint8_t reserved_eth[5] = { 0x02, 0x00, 0x5E, 0xFF, 0xFE };
315 	static uint8_t reserved_anycast[7] = { 0xFD, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF };
316 
317 	/* Subnet-Router Anycast (RFC 4291) */
318 	if (memcmp(iid, allzero, 8) == 0)
319 		return (false);
320 
321 	/*
322 	 * Reserved IPv6 Interface Identifiers corresponding to the IANA Ethernet Block (RFC 4291)
323 	 * and
324 	 * Proxy Mobile IPv6 (RFC 6543)
325 	 */
326 	if (memcmp(iid, reserved_eth, 5) == 0)
327 		return (false);
328 
329 	/* Reserved Subnet Anycast Addresses (RFC 2526) */
330 	if (memcmp(iid, reserved_anycast, 7) == 0 && iid[7] >= 0x80)
331 		return (false);
332 
333 	return (true);
334 }
335 
336 /*
337  * Get interface identifier for the specified interface, according to
338  * RFC 7217 Stable and Opaque IDs with SLAAC, using HMAC-SHA256 digest.
339  *
340  * in6 - upper 64bits are preserved
341  */
342 bool
in6_get_stableifid(struct ifnet * ifp,struct in6_addr * in6,int prefixlen)343 in6_get_stableifid(struct ifnet *ifp, struct in6_addr *in6, int prefixlen)
344 {
345 	struct sockaddr_dl *sdl;
346 	const uint8_t *netiface;
347 	size_t netiface_len, hostuuid_len;
348 	uint8_t hostuuid[HOSTUUIDLEN + 1], hmac_key[SHA256_BLOCK_LENGTH],
349 		hk_ipad[SHA256_BLOCK_LENGTH], hk_opad[SHA256_BLOCK_LENGTH];
350 	uint64_t dad_failures;
351 	SHA256_CTX ctxt;
352 
353 	switch (V_ip6_stableaddr_netifsource) {
354 		case IP6_STABLEADDR_NETIFSRC_ID:
355 			sdl = get_interface_link_level(ifp);
356 			if (sdl == NULL)
357 				return (false);
358 			netiface = (uint8_t *)&LLINDEX(sdl);
359 			netiface_len = sizeof(u_short); /* real return type of LLINDEX */
360 			break;
361 
362 		case IP6_STABLEADDR_NETIFSRC_MAC:
363 			netiface = in6_get_interface_hwaddr(ifp, &netiface_len);
364 			if (netiface == NULL)
365 				return (false);
366 			break;
367 
368 		case IP6_STABLEADDR_NETIFSRC_NAME:
369 		default:
370 			netiface = (const uint8_t *)if_name(ifp);
371 			netiface_len = strlen(netiface);
372 			break;
373 	}
374 
375 	/* Use hostuuid as constant "secret" key */
376 	getcredhostuuid(curthread->td_ucred, hostuuid, sizeof(hostuuid));
377 	if (strncmp(hostuuid, DEFAULT_HOSTUUID, sizeof(hostuuid)) == 0) {
378 		/* If hostuuid is not set, use a random value */
379 		arc4rand(hostuuid, HOSTUUIDLEN, 0);
380 		hostuuid[HOSTUUIDLEN] = '\0';
381 	}
382 	hostuuid_len = strlen(hostuuid);
383 
384 	dad_failures = atomic_load_int(&DAD_FAILURES(ifp));
385 
386 	/* RFC 7217 section 7, default max retries */
387 	if (dad_failures > V_ip6_stableaddr_maxretries)
388 		return (false);
389 
390 	/*
391 	 * Use hostuuid as basis for HMAC key
392 	 */
393 	memset(hmac_key, 0, sizeof(hmac_key));
394 	if (hostuuid_len <= SHA256_BLOCK_LENGTH) {
395 		/* copy to hmac key variable, zero padded */
396 		memcpy(hmac_key, hostuuid, hostuuid_len);
397 	} else {
398 		/* if longer than block length, use hash of the value, zero padded */
399 		SHA256_Init(&ctxt);
400 		SHA256_Update(&ctxt, hostuuid, hostuuid_len);
401 		SHA256_Final(hmac_key, &ctxt);
402 	}
403 	/* XOR key with ipad and opad values */
404 	for (uint16_t i = 0; i < sizeof(hmac_key); i++) {
405 		hk_ipad[i] = hmac_key[i] ^ HMAC_IPAD;
406 		hk_opad[i] = hmac_key[i] ^ HMAC_OPAD;
407 	}
408 
409 	/*
410 	 * Generate interface id in a loop, adding an offset to be factored in the hash function.
411 	 * This is necessary, because if the generated interface id happens to be invalid we
412 	 * want to force the hash function to generate a different one, otherwise we would end up
413 	 * in an infinite loop trying the same invalid interface id over and over again.
414 	 *
415 	 * Using an uint8 counter for the offset, so limit iteration at UINT8_MAX. This is a safety
416 	 * measure, this will never iterate more than once or twice in practice.
417 	 */
418 	for(uint8_t offset = 0; offset < UINT8_MAX; offset++) {
419 		uint8_t digest[SHA256_DIGEST_LENGTH];
420 
421 		/* Calculate inner hash */
422 		SHA256_Init(&ctxt);
423 		SHA256_Update(&ctxt, hk_ipad, sizeof(hk_ipad));
424 		SHA256_Update(&ctxt, in6->s6_addr, prefixlen / 8);
425 		SHA256_Update(&ctxt, netiface, netiface_len);
426 		SHA256_Update(&ctxt, (uint8_t *)&dad_failures, 8);
427 		SHA256_Update(&ctxt, hostuuid, hostuuid_len);
428 		SHA256_Update(&ctxt, &offset, 1);
429 		SHA256_Final(digest, &ctxt);
430 
431 		/* Calculate outer hash */
432 		SHA256_Init(&ctxt);
433 		SHA256_Update(&ctxt, hk_opad, sizeof(hk_opad));
434 		SHA256_Update(&ctxt, digest, sizeof(digest));
435 		SHA256_Final(digest, &ctxt);
436 
437 		if (validate_ifid(digest)) {
438 			/* assumes sizeof(digest) > sizeof(ifid) */
439 			memcpy(&in6->s6_addr[8], digest, 8);
440 
441 			return (true);
442 		}
443 	}
444 
445 	return (false);
446 }
447 
448 /*
449  * Get interface identifier for the specified interface.  If it is not
450  * available on ifp0, borrow interface identifier from other information
451  * sources.
452  *
453  * altifp - secondary EUI64 source
454  */
455 int
in6_get_ifid(struct ifnet * ifp0,struct ifnet * altifp,struct in6_addr * in6)456 in6_get_ifid(struct ifnet *ifp0, struct ifnet *altifp,
457     struct in6_addr *in6)
458 {
459 	struct ifnet *ifp;
460 
461 	NET_EPOCH_ASSERT();
462 
463 	/* first, try to get it from the interface itself, with stable algorithm, if configured */
464 	if ((ifp0->if_inet6->nd_flags & ND6_IFF_STABLEADDR) && in6_get_stableifid(ifp0, in6, 64) == 0) {
465 		nd6log((LOG_DEBUG, "%s: got interface identifier from itself (stable private)\n",
466 		    if_name(ifp0)));
467 		goto success;
468 	}
469 
470 	/* then/otherwise try to get it from the interface itself */
471 	if (in6_get_hw_ifid(ifp0, in6) == 0) {
472 		nd6log((LOG_DEBUG, "%s: got interface identifier from itself\n",
473 		    if_name(ifp0)));
474 		goto success;
475 	}
476 
477 	/* try secondary EUI64 source. this basically is for ATM PVC */
478 	if (altifp && in6_get_hw_ifid(altifp, in6) == 0) {
479 		nd6log((LOG_DEBUG, "%s: got interface identifier from %s\n",
480 		    if_name(ifp0), if_name(altifp)));
481 		goto success;
482 	}
483 
484 	/* next, try to get it from some other hardware interface */
485 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
486 		if (ifp == ifp0)
487 			continue;
488 		if (in6_get_hw_ifid(ifp, in6) != 0)
489 			continue;
490 
491 		/*
492 		 * to borrow ifid from other interface, ifid needs to be
493 		 * globally unique
494 		 */
495 		if (IFID_UNIVERSAL(in6)) {
496 			nd6log((LOG_DEBUG,
497 			    "%s: borrow interface identifier from %s\n",
498 			    if_name(ifp0), if_name(ifp)));
499 			goto success;
500 		}
501 	}
502 
503 	/* last resort: get from random number source */
504 	if (get_rand_ifid(ifp, in6) == 0) {
505 		nd6log((LOG_DEBUG,
506 		    "%s: interface identifier generated by random number\n",
507 		    if_name(ifp0)));
508 		goto success;
509 	}
510 
511 	printf("%s: failed to get interface identifier\n", if_name(ifp0));
512 	return -1;
513 
514 success:
515 	nd6log((LOG_INFO, "%s: ifid: %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x\n",
516 	    if_name(ifp0), in6->s6_addr[8], in6->s6_addr[9], in6->s6_addr[10],
517 	    in6->s6_addr[11], in6->s6_addr[12], in6->s6_addr[13],
518 	    in6->s6_addr[14], in6->s6_addr[15]));
519 	return 0;
520 }
521 
522 /*
523  * altifp - secondary EUI64 source
524  */
525 static int
in6_ifattach_linklocal(struct ifnet * ifp,struct ifnet * altifp)526 in6_ifattach_linklocal(struct ifnet *ifp, struct ifnet *altifp)
527 {
528 	struct in6_ifaddr *ia;
529 	struct in6_aliasreq ifra;
530 	struct nd_prefixctl pr0;
531 	struct epoch_tracker et;
532 	struct nd_prefix *pr;
533 	int error;
534 
535 	/*
536 	 * configure link-local address.
537 	 */
538 	in6_prepare_ifra(&ifra, NULL, &in6mask64);
539 
540 	ifra.ifra_addr.sin6_addr.s6_addr32[0] = htonl(0xfe800000);
541 	ifra.ifra_addr.sin6_addr.s6_addr32[1] = 0;
542 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
543 		ifra.ifra_addr.sin6_addr.s6_addr32[2] = 0;
544 		ifra.ifra_addr.sin6_addr.s6_addr32[3] = htonl(1);
545 	} else {
546 		NET_EPOCH_ENTER(et);
547 		error = in6_get_ifid(ifp, altifp, &ifra.ifra_addr.sin6_addr);
548 		NET_EPOCH_EXIT(et);
549 		if (error != 0) {
550 			nd6log((LOG_ERR,
551 			    "%s: no ifid available\n", if_name(ifp)));
552 			return (-1);
553 		}
554 	}
555 	if (in6_setscope(&ifra.ifra_addr.sin6_addr, ifp, NULL))
556 		return (-1);
557 
558 	/* link-local addresses should NEVER expire. */
559 	ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
560 	ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
561 
562 	/*
563 	 * Now call in6_update_ifa() to do a bunch of procedures to configure
564 	 * a link-local address. We can set the 3rd argument to NULL, because
565 	 * we know there's no other link-local address on the interface
566 	 * and therefore we are adding one (instead of updating one).
567 	 */
568 	if ((error = in6_update_ifa(ifp, &ifra, NULL,
569 				    IN6_IFAUPDATE_DADDELAY)) != 0) {
570 		/*
571 		 * XXX: When the interface does not support IPv6, this call
572 		 * would fail in the SIOCSIFADDR ioctl.  I believe the
573 		 * notification is rather confusing in this case, so just
574 		 * suppress it.  (jinmei@kame.net 20010130)
575 		 */
576 		if (error != EAFNOSUPPORT)
577 			nd6log((LOG_NOTICE, "in6_ifattach_linklocal: failed to "
578 			    "configure a link-local address on %s "
579 			    "(errno=%d)\n",
580 			    if_name(ifp), error));
581 		return (-1);
582 	}
583 
584 	NET_EPOCH_ENTER(et);
585 	ia = in6ifa_ifpforlinklocal(ifp, 0);
586 	NET_EPOCH_EXIT(et);
587 	if (ia == NULL) {
588 		/*
589 		 * Another thread removed the address that we just added.
590 		 * This should be rare, but it happens.
591 		 */
592 		nd6log((LOG_NOTICE, "%s: %s: new link-local address "
593 			"disappeared\n", __func__, if_name(ifp)));
594 		return (-1);
595 	}
596 	ifa_free(&ia->ia_ifa);
597 
598 	/*
599 	 * Make the link-local prefix (fe80::%link/64) as on-link.
600 	 * Since we'd like to manage prefixes separately from addresses,
601 	 * we make an ND6 prefix structure for the link-local prefix,
602 	 * and add it to the prefix list as a never-expire prefix.
603 	 * XXX: this change might affect some existing code base...
604 	 */
605 	bzero(&pr0, sizeof(pr0));
606 	pr0.ndpr_ifp = ifp;
607 	/* this should be 64 at this moment. */
608 	pr0.ndpr_plen = in6_mask2len(&ifra.ifra_prefixmask.sin6_addr, NULL);
609 	pr0.ndpr_prefix = ifra.ifra_addr;
610 	/* apply the mask for safety. (nd6_prelist_add will apply it again) */
611 	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr, &in6mask64);
612 	/*
613 	 * Initialize parameters.  The link-local prefix must always be
614 	 * on-link, and its lifetimes never expire.
615 	 */
616 	pr0.ndpr_raf_onlink = 1;
617 	pr0.ndpr_raf_auto = 1;	/* probably meaningless */
618 	pr0.ndpr_vltime = ND6_INFINITE_LIFETIME;
619 	pr0.ndpr_pltime = ND6_INFINITE_LIFETIME;
620 	/*
621 	 * Since there is no other link-local addresses, nd6_prefix_lookup()
622 	 * probably returns NULL.  However, we cannot always expect the result.
623 	 * For example, if we first remove the (only) existing link-local
624 	 * address, and then reconfigure another one, the prefix is still
625 	 * valid with referring to the old link-local address.
626 	 */
627 	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
628 		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
629 			return (error);
630 		/* Reference prefix */
631 		ia->ia6_ndpr = pr;
632 		pr->ndpr_addrcnt++;
633 	}
634 	nd6_prefix_rele(pr);
635 
636 	return 0;
637 }
638 
639 /*
640  * ifp - must be IFT_LOOP
641  */
642 static int
in6_ifattach_loopback(struct ifnet * ifp)643 in6_ifattach_loopback(struct ifnet *ifp)
644 {
645 	struct in6_aliasreq ifra;
646 	int error;
647 
648 	in6_prepare_ifra(&ifra, &in6addr_loopback, &in6mask128);
649 
650 	/*
651 	 * Always initialize ia_dstaddr (= broadcast address) to loopback
652 	 * address.  Follows IPv4 practice - see in_ifinit().
653 	 */
654 	ifra.ifra_dstaddr.sin6_len = sizeof(struct sockaddr_in6);
655 	ifra.ifra_dstaddr.sin6_family = AF_INET6;
656 	ifra.ifra_dstaddr.sin6_addr = in6addr_loopback;
657 
658 	/* the loopback  address should NEVER expire. */
659 	ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
660 	ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
661 
662 	/*
663 	 * We are sure that this is a newly assigned address, so we can set
664 	 * NULL to the 3rd arg.
665 	 */
666 	if ((error = in6_update_ifa(ifp, &ifra, NULL, 0)) != 0) {
667 		nd6log((LOG_ERR, "in6_ifattach_loopback: failed to configure "
668 		    "the loopback address on %s (errno=%d)\n",
669 		    if_name(ifp), error));
670 		return (-1);
671 	}
672 
673 	return 0;
674 }
675 
676 /*
677  * compute NI group address, based on the current hostname setting.
678  * see RFC 4620.
679  *
680  * when ifp == NULL, the caller is responsible for filling scopeid.
681  *
682  * If oldmcprefix == 1, FF02:0:0:0:0:2::/96 is used for NI group address
683  * while it is FF02:0:0:0:0:2:FF00::/104 in RFC 4620.
684  */
685 static int
in6_nigroup0(struct ifnet * ifp,const char * name,int namelen,struct in6_addr * in6,int oldmcprefix)686 in6_nigroup0(struct ifnet *ifp, const char *name, int namelen,
687     struct in6_addr *in6, int oldmcprefix)
688 {
689 	struct prison *pr;
690 	const char *p;
691 	u_char *q;
692 	MD5_CTX ctxt;
693 	u_int8_t digest[16];
694 	char l;
695 	char n[64];	/* a single label must not exceed 63 chars */
696 
697 	/*
698 	 * If no name is given and namelen is -1,
699 	 * we try to do the hostname lookup ourselves.
700 	 */
701 	if (!name && namelen == -1) {
702 		pr = curthread->td_ucred->cr_prison;
703 		mtx_lock(&pr->pr_mtx);
704 		name = pr->pr_hostname;
705 		namelen = strlen(name);
706 	} else
707 		pr = NULL;
708 	if (!name || !namelen) {
709 		if (pr != NULL)
710 			mtx_unlock(&pr->pr_mtx);
711 		return -1;
712 	}
713 
714 	p = name;
715 	while (p && *p && *p != '.' && p - name < namelen)
716 		p++;
717 	if (p == name || p - name > sizeof(n) - 1) {
718 		if (pr != NULL)
719 			mtx_unlock(&pr->pr_mtx);
720 		return -1;	/* label too long */
721 	}
722 	l = p - name;
723 	strncpy(n, name, l);
724 	if (pr != NULL)
725 		mtx_unlock(&pr->pr_mtx);
726 	n[(int)l] = '\0';
727 	for (q = n; *q; q++) {
728 		if ('A' <= *q && *q <= 'Z')
729 			*q = *q - 'A' + 'a';
730 	}
731 
732 	/* generate 16 bytes of pseudo-random value. */
733 	bzero(&ctxt, sizeof(ctxt));
734 	MD5Init(&ctxt);
735 	MD5Update(&ctxt, &l, sizeof(l));
736 	MD5Update(&ctxt, n, l);
737 	MD5Final(digest, &ctxt);
738 
739 	bzero(in6, sizeof(*in6));
740 	in6->s6_addr16[0] = IPV6_ADDR_INT16_MLL;
741 	in6->s6_addr8[11] = 2;
742 	if (oldmcprefix == 0) {
743 		in6->s6_addr8[12] = 0xff;
744 	 	/* Copy the first 24 bits of 128-bit hash into the address. */
745 		bcopy(digest, &in6->s6_addr8[13], 3);
746 	} else {
747 	 	/* Copy the first 32 bits of 128-bit hash into the address. */
748 		bcopy(digest, &in6->s6_addr32[3], sizeof(in6->s6_addr32[3]));
749 	}
750 	if (in6_setscope(in6, ifp, NULL))
751 		return (-1); /* XXX: should not fail */
752 
753 	return 0;
754 }
755 
756 int
in6_nigroup(struct ifnet * ifp,const char * name,int namelen,struct in6_addr * in6)757 in6_nigroup(struct ifnet *ifp, const char *name, int namelen,
758     struct in6_addr *in6)
759 {
760 
761 	return (in6_nigroup0(ifp, name, namelen, in6, 0));
762 }
763 
764 int
in6_nigroup_oldmcprefix(struct ifnet * ifp,const char * name,int namelen,struct in6_addr * in6)765 in6_nigroup_oldmcprefix(struct ifnet *ifp, const char *name, int namelen,
766     struct in6_addr *in6)
767 {
768 
769 	return (in6_nigroup0(ifp, name, namelen, in6, 1));
770 }
771 
772 /*
773  * XXX multiple loopback interface needs more care.  for instance,
774  * nodelocal address needs to be configured onto only one of them.
775  * XXX multiple link-local address case
776  *
777  * altifp - secondary EUI64 source
778  */
779 void
in6_ifattach(struct ifnet * ifp,struct ifnet * altifp)780 in6_ifattach(struct ifnet *ifp, struct ifnet *altifp)
781 {
782 	struct in6_ifaddr *ia;
783 
784 	/* XXXGL: can this happen after IFT_PFLOG and IFT_PFSYNC are gone? */
785 	if (ifp->if_inet6 == NULL)
786 		return;
787 	/*
788 	 * quirks based on interface type
789 	 */
790 	switch (ifp->if_type) {
791 	case IFT_STF:
792 		/*
793 		 * 6to4 interface is a very special kind of beast.
794 		 * no multicast, no linklocal.  RFC2529 specifies how to make
795 		 * linklocals for 6to4 interface, but there's no use and
796 		 * it is rather harmful to have one.
797 		 */
798 		ifp->if_inet6->nd_flags &= ~ND6_IFF_AUTO_LINKLOCAL;
799 		ifp->if_inet6->nd_flags |= ND6_IFF_NO_DAD;
800 		break;
801 	default:
802 		break;
803 	}
804 
805 	/*
806 	 * usually, we require multicast capability to the interface
807 	 */
808 	if ((ifp->if_flags & IFF_MULTICAST) == 0) {
809 		nd6log((LOG_INFO, "in6_ifattach: "
810 		    "%s is not multicast capable, IPv6 not enabled\n",
811 		    if_name(ifp)));
812 		return;
813 	}
814 
815 	/*
816 	 * assign loopback address for loopback interface.
817 	 */
818 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
819 		/*
820 		 * check that loopback address doesn't exist yet.
821 		 */
822 		ia = in6ifa_ifwithaddr(&in6addr_loopback, 0, false);
823 		if (ia == NULL)
824 			in6_ifattach_loopback(ifp);
825 	}
826 
827 	/*
828 	 * assign a link-local address, if there's none.
829 	 */
830 	if (!(ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) &&
831 	    ifp->if_inet6->nd_flags & ND6_IFF_AUTO_LINKLOCAL) {
832 		struct epoch_tracker et;
833 
834 		NET_EPOCH_ENTER(et);
835 		ia = in6ifa_ifpforlinklocal(ifp, 0);
836 		NET_EPOCH_EXIT(et);
837 		if (ia == NULL)
838 			in6_ifattach_linklocal(ifp, altifp);
839 		else
840 			ifa_free(&ia->ia_ifa);
841 	}
842 }
843 
844 /*
845  * NOTE: in6_ifdetach() does not support loopback if at this moment.
846  *
847  * When shutting down a VNET we clean up layers top-down.  In that case
848  * upper layer protocols (ulp) are cleaned up already and locks are destroyed
849  * and we must not call into these cleanup functions anymore, thus purgeulp
850  * is set to 0 in that case by in6_ifdetach_destroy().
851  * The normal case of destroying a (cloned) interface still needs to cleanup
852  * everything related to the interface and will have purgeulp set to 1.
853  */
854 static void
_in6_ifdetach(struct ifnet * ifp,int purgeulp)855 _in6_ifdetach(struct ifnet *ifp, int purgeulp)
856 {
857 	struct ifaddr *ifa, *next;
858 
859 	/* XXXGL: can this happen after IFT_PFLOG and IFT_PFSYNC are gone? */
860 	if (ifp->if_inet6 == NULL)
861 		return;
862 
863 	/*
864 	 * nuke any of IPv6 addresses we have
865 	 */
866 	CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
867 		if (ifa->ifa_addr->sa_family != AF_INET6)
868 			continue;
869 		in6_purgeaddr(ifa);
870 	}
871 	if (purgeulp) {
872 		IN6_MULTI_LOCK();
873 		in6_pcbpurgeif0(&V_udbinfo, ifp);
874 		in6_pcbpurgeif0(&V_ulitecbinfo, ifp);
875 		in6_pcbpurgeif0(&V_ripcbinfo, ifp);
876 		IN6_MULTI_UNLOCK();
877 	}
878 	/* leave from all multicast groups joined */
879 	in6_purgemaddrs(ifp);
880 
881 	/*
882 	 * Remove neighbor management table.
883 	 * Enabling the nd6_purge will panic on vmove for interfaces on VNET
884 	 * teardown as the IPv6 layer is cleaned up already and the locks
885 	 * are destroyed.
886 	 */
887 	if (purgeulp)
888 		nd6_purge(ifp);
889 }
890 
891 void
in6_ifdetach(struct ifnet * ifp)892 in6_ifdetach(struct ifnet *ifp)
893 {
894 
895 	_in6_ifdetach(ifp, 1);
896 }
897 
898 static void
in6_ifextra_free(epoch_context_t ctx)899 in6_ifextra_free(epoch_context_t ctx)
900 {
901 	struct in6_ifextra *ext =
902 	     __containerof(ctx, struct in6_ifextra, epoch_ctx);
903 
904 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
905 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
906 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
907 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
908 	free(ext, M_IFADDR);
909 }
910 
911 static void
in6_ifdeparture(void * arg __unused,struct ifnet * ifp)912 in6_ifdeparture(void *arg __unused, struct ifnet *ifp)
913 {
914 	struct in6_ifextra *ext = ifp->if_inet6;
915 
916 	/* XXXGL: can this happen after IFT_PFLOG and IFT_PFSYNC are gone? */
917 	if (ifp->if_inet6 == NULL)
918 		return;
919 
920 #ifdef VIMAGE
921 	/*
922 	 * On VNET shutdown abort here as the stack teardown will do all
923 	 * the work top-down for us.  XXXGL: see comment in in.c:in_ifdetach().
924 	 */
925 	if (!VNET_IS_SHUTTING_DOWN(ifp->if_vnet))
926 #endif
927 		_in6_ifdetach(ifp, 1);
928 	/*
929 	 * XXXGL: mld and nd bits are left in a consistent state after
930 	 * destructors, but I'm not sure if it safe to call lltable_free() here.
931 	 * Individual lle entries are epoch(9) protected, but the table itself
932 	 * isn't.
933 	 */
934 	mld_domifdetach(ifp);
935 	nd6_ifdetach(ifp);
936 	lltable_free(ext->lltable);
937 	NET_EPOCH_CALL(in6_ifextra_free, &ext->epoch_ctx);
938 }
939 EVENTHANDLER_DEFINE(ifnet_departure_event, in6_ifdeparture, NULL,
940     EVENTHANDLER_PRI_ANY);
941 
942 void
in6_ifdetach_destroy(struct ifnet * ifp)943 in6_ifdetach_destroy(struct ifnet *ifp)
944 {
945 
946 	_in6_ifdetach(ifp, 0);
947 }
948 
949 void
in6_tmpaddrtimer(void * arg)950 in6_tmpaddrtimer(void *arg)
951 {
952 	CURVNET_SET((struct vnet *) arg);
953 
954 	callout_reset(&V_in6_tmpaddrtimer_ch,
955 	    (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
956 	    V_ip6_temp_regen_advance) * hz, in6_tmpaddrtimer, curvnet);
957 
958 	CURVNET_RESTORE();
959 }
960 
961 static void
in6_purgemaddrs(struct ifnet * ifp)962 in6_purgemaddrs(struct ifnet *ifp)
963 {
964 	struct in6_multi_head inmh;
965 
966 	SLIST_INIT(&inmh);
967 	IN6_MULTI_LOCK();
968 	IN6_MULTI_LIST_LOCK();
969 	mld_ifdetach(ifp, &inmh);
970 	IN6_MULTI_LIST_UNLOCK();
971 	IN6_MULTI_UNLOCK();
972 	in6m_release_list_deferred(&inmh);
973 
974 	/*
975 	 * Make sure all multicast deletions invoking if_ioctl() are
976 	 * completed before returning. Else we risk accessing a freed
977 	 * ifnet structure pointer.
978 	 */
979 	in6m_release_wait(NULL);
980 }
981 
982 void
in6_ifattach_destroy(void)983 in6_ifattach_destroy(void)
984 {
985 
986 	callout_drain(&V_in6_tmpaddrtimer_ch);
987 }
988 
989 static void
in6_ifattach_init(void * dummy)990 in6_ifattach_init(void *dummy)
991 {
992 
993 	/* Timer for regeneranation of temporary addresses randomize ID. */
994 	callout_init(&V_in6_tmpaddrtimer_ch, 1);
995 	callout_reset(&V_in6_tmpaddrtimer_ch,
996 	    (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
997 	    V_ip6_temp_regen_advance) * hz,
998 	    in6_tmpaddrtimer, curvnet);
999 }
1000 
1001 /*
1002  * Cheat.
1003  * This must be after route_init(), which is now SI_ORDER_THIRD.
1004  */
1005 SYSINIT(in6_ifattach_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
1006     in6_ifattach_init, NULL);
1007