xref: /freebsd/sys/netinet/in.c (revision 0a1294f6c610948d7447ae276df74a6d5269b62e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * Copyright (C) 2001 WIDE Project.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 #include "opt_inet.h"
35 
36 #define IN_HISTORICAL_NETS		/* include class masks */
37 
38 #include <sys/param.h>
39 #include <sys/eventhandler.h>
40 #include <sys/systm.h>
41 #include <sys/sockio.h>
42 #include <sys/malloc.h>
43 #include <sys/priv.h>
44 #include <sys/socket.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/proc.h>
49 #include <sys/sysctl.h>
50 #include <sys/syslog.h>
51 #include <sys/sx.h>
52 
53 #include <net/if.h>
54 #include <net/if_var.h>
55 #include <net/if_arp.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_private.h>
59 #include <net/if_types.h>
60 #include <net/if_bridgevar.h>
61 #include <net/route.h>
62 #include <net/route/nhop.h>
63 #include <net/route/route_ctl.h>
64 #include <net/vnet.h>
65 
66 #include <netinet/if_ether.h>
67 #include <netinet/in.h>
68 #include <netinet/in_fib.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_carp.h>
73 #include <netinet/igmp_var.h>
74 #include <netinet/udp.h>
75 #include <netinet/udp_var.h>
76 
77 #ifdef MAC
78 #include <security/mac/mac_framework.h>
79 #endif
80 
81 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
83 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
84 
85 static void	in_socktrim(struct sockaddr_in *);
86 static void	in_purgemaddrs(struct ifnet *);
87 
88 static bool	ia_need_loopback_route(const struct in_ifaddr *);
89 
90 VNET_DEFINE_STATIC(int, nosameprefix);
91 #define	V_nosameprefix			VNET(nosameprefix)
92 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
93 	&VNET_NAME(nosameprefix), 0,
94 	"Refuse to create same prefixes on different interfaces");
95 
96 VNET_DEFINE_STATIC(bool, broadcast_lowest);
97 #define	V_broadcast_lowest		VNET(broadcast_lowest)
98 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
99 	&VNET_NAME(broadcast_lowest), 0,
100 	"Treat lowest address on a subnet (host 0) as broadcast");
101 
102 VNET_DEFINE(bool, ip_allow_net240) = false;
103 #define	V_ip_allow_net240		VNET(ip_allow_net240)
104 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
105 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
106 	"Allow forwarding of and ICMP response to Experimental addresses, aka Class E (240/4)");
107 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
108 
109 VNET_DEFINE(bool, ip_allow_net0) = false;
110 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
111 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
112 	"Allow forwarding of and ICMP response to addresses in network 0/8");
113 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
114 
115 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
116 #define	V_in_loopback_mask	VNET(in_loopback_mask)
117 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
118 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
119 	CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
120 	NULL, 0, sysctl_loopback_prefixlen, "I",
121 	"Prefix length of address space reserved for loopback");
122 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
123 
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 #define	V_ripcbinfo			VNET(ripcbinfo)
126 
127 static struct sx in_control_sx;
128 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
129 
130 /*
131  * Return true if an internet address is for a ``local'' host
132  * (one to which we have a connection).
133  */
134 bool
in_localaddr(struct in_addr in)135 in_localaddr(struct in_addr in)
136 {
137 	u_long i = ntohl(in.s_addr);
138 	struct in_ifaddr *ia;
139 
140 	NET_EPOCH_ASSERT();
141 
142 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
143 		if ((i & ia->ia_subnetmask) == ia->ia_subnet)
144 			return (true);
145 	}
146 
147 	return (false);
148 }
149 
150 /*
151  * Return true if an internet address is for the local host and configured
152  * on one of its interfaces.
153  */
154 bool
in_localip(struct in_addr in)155 in_localip(struct in_addr in)
156 {
157 	struct in_ifaddr *ia;
158 
159 	NET_EPOCH_ASSERT();
160 
161 	CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
162 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
163 			return (true);
164 
165 	return (false);
166 }
167 
168 /*
169  * Like in_localip(), but FIB-aware and carp(4)-aware.
170  */
171 bool
in_localip_fib(struct in_addr in,uint16_t fib)172 in_localip_fib(struct in_addr in, uint16_t fib)
173 {
174 	struct in_ifaddr *ia;
175 
176 	NET_EPOCH_ASSERT();
177 
178 	CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
179 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
180 		    (ia->ia_ifa.ifa_carp == NULL ||
181 		    carp_master_p(&ia->ia_ifa)) &&
182 		    ia->ia_ifa.ifa_ifp->if_fib == fib)
183 			return (true);
184 
185 	return (false);
186 }
187 
188 /*
189  * Return true if an internet address is configured on an interface.
190  */
191 bool
in_ifhasaddr(struct ifnet * ifp,struct in_addr in)192 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
193 {
194 	struct ifaddr *ifa;
195 	struct in_ifaddr *ia;
196 
197 	NET_EPOCH_ASSERT();
198 
199 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
200 		if (ifa->ifa_addr->sa_family != AF_INET)
201 			continue;
202 		ia = (struct in_ifaddr *)ifa;
203 		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
204 			return (true);
205 	}
206 
207 	return (false);
208 }
209 
210 /*
211  * Return a reference to the interface address which is different to
212  * the supplied one but with same IP address value.
213  */
214 static struct in_ifaddr *
in_localip_more(struct in_ifaddr * original_ia)215 in_localip_more(struct in_ifaddr *original_ia)
216 {
217 	struct epoch_tracker et;
218 	in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
219 	uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
220 	struct in_ifaddr *ia;
221 
222 	NET_EPOCH_ENTER(et);
223 	CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
224 		in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
225 		uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
226 		if (!V_rt_add_addr_allfibs && (original_fib != fib))
227 			continue;
228 		if ((original_ia != ia) && (original_addr == addr)) {
229 			ifa_ref(&ia->ia_ifa);
230 			NET_EPOCH_EXIT(et);
231 			return (ia);
232 		}
233 	}
234 	NET_EPOCH_EXIT(et);
235 
236 	return (NULL);
237 }
238 
239 /*
240  * Tries to find first IPv4 address in the provided fib.
241  * Prefers non-loopback addresses and return loopback IFF
242  * @loopback_ok is set.
243  *
244  * Returns ifa or NULL.
245  */
246 struct in_ifaddr *
in_findlocal(uint32_t fibnum,bool loopback_ok)247 in_findlocal(uint32_t fibnum, bool loopback_ok)
248 {
249 	struct in_ifaddr *ia = NULL, *ia_lo = NULL;
250 
251 	NET_EPOCH_ASSERT();
252 
253 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
254 		uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
255 		if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
256 			continue;
257 
258 		if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
259 			break;
260 		if (loopback_ok)
261 			ia_lo = ia;
262 	}
263 
264 	if (ia == NULL)
265 		ia = ia_lo;
266 
267 	return (ia);
268 }
269 
270 /*
271  * Determine whether an IP address is in a reserved set of addresses
272  * that may not be forwarded, or whether datagrams to that destination
273  * may be forwarded.
274  */
275 bool
in_canforward(struct in_addr in)276 in_canforward(struct in_addr in)
277 {
278 	u_long i = ntohl(in.s_addr);
279 
280 	if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i) ||
281 	    in_nullhost(in))
282 		return (false);
283 	if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
284 		return (false);
285 	if (IN_ZERONET(i) && !V_ip_allow_net0)
286 		return (false);
287 	return (true);
288 }
289 
290 /*
291  * Sysctl to manage prefix of reserved loopback network; translate
292  * to/from mask.  The mask is always contiguous high-order 1 bits
293  * followed by all 0 bits.
294  */
295 static int
sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)296 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
297 {
298 	int error, preflen;
299 
300 	/* ffs is 1-based; compensate. */
301 	preflen = 33 - ffs(V_in_loopback_mask);
302 	error = sysctl_handle_int(oidp, &preflen, 0, req);
303 	if (error || !req->newptr)
304 		return (error);
305 	if (preflen < 8 || preflen > 31)
306 		return (EINVAL);
307 	V_in_loopback_mask = 0xffffffff << (32 - preflen);
308 	return (0);
309 }
310 
311 /*
312  * Trim a mask in a sockaddr
313  */
314 static void
in_socktrim(struct sockaddr_in * ap)315 in_socktrim(struct sockaddr_in *ap)
316 {
317     char *cplim = (char *) &ap->sin_addr;
318     char *cp = (char *) (&ap->sin_addr + 1);
319 
320     ap->sin_len = 0;
321     while (--cp >= cplim)
322 	if (*cp) {
323 	    (ap)->sin_len = cp - (char *) (ap) + 1;
324 	    break;
325 	}
326 }
327 
328 /*
329  * Generic internet control operations (ioctl's).
330  */
331 int
in_control_ioctl(u_long cmd,void * data,struct ifnet * ifp,struct ucred * cred)332 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp,
333     struct ucred *cred)
334 {
335 	struct ifreq *ifr = (struct ifreq *)data;
336 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
337 	struct epoch_tracker et;
338 	struct ifaddr *ifa;
339 	struct in_ifaddr *ia;
340 	int error;
341 
342 	if (ifp == NULL)
343 		return (EADDRNOTAVAIL);
344 
345 	/*
346 	 * Filter out 4 ioctls we implement directly.  Forward the rest
347 	 * to specific functions and ifp->if_ioctl().
348 	 */
349 	switch (cmd) {
350 	case SIOCGIFADDR:
351 	case SIOCGIFBRDADDR:
352 	case SIOCGIFDSTADDR:
353 	case SIOCGIFNETMASK:
354 		break;
355 	case SIOCGIFALIAS:
356 		sx_xlock(&in_control_sx);
357 		error = in_gifaddr_ioctl(cmd, data, ifp, cred);
358 		sx_xunlock(&in_control_sx);
359 		return (error);
360 	case SIOCDIFADDR:
361 		sx_xlock(&in_control_sx);
362 		error = in_difaddr_ioctl(cmd, data, ifp, cred);
363 		sx_xunlock(&in_control_sx);
364 		return (error);
365 	case OSIOCAIFADDR:	/* 9.x compat */
366 	case SIOCAIFADDR:
367 		sx_xlock(&in_control_sx);
368 		error = in_aifaddr_ioctl(cmd, data, ifp, cred);
369 		sx_xunlock(&in_control_sx);
370 		return (error);
371 	case SIOCSIFADDR:
372 	case SIOCSIFBRDADDR:
373 	case SIOCSIFDSTADDR:
374 	case SIOCSIFNETMASK:
375 		/* We no longer support that old commands. */
376 		return (EINVAL);
377 	default:
378 		if (ifp->if_ioctl == NULL)
379 			return (EOPNOTSUPP);
380 		return ((*ifp->if_ioctl)(ifp, cmd, data));
381 	}
382 
383 	if (addr->sin_addr.s_addr != INADDR_ANY &&
384 	    prison_check_ip4(cred, &addr->sin_addr) != 0)
385 		return (EADDRNOTAVAIL);
386 
387 	/*
388 	 * Find address for this interface, if it exists.  If an
389 	 * address was specified, find that one instead of the
390 	 * first one on the interface, if possible.
391 	 */
392 	NET_EPOCH_ENTER(et);
393 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
394 		if (ifa->ifa_addr->sa_family != AF_INET)
395 			continue;
396 		ia = (struct in_ifaddr *)ifa;
397 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
398 			break;
399 	}
400 	if (ifa == NULL)
401 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
402 			if (ifa->ifa_addr->sa_family == AF_INET) {
403 				ia = (struct in_ifaddr *)ifa;
404 				if (prison_check_ip4(cred,
405 				    &ia->ia_addr.sin_addr) == 0)
406 					break;
407 			}
408 
409 	if (ifa == NULL) {
410 		NET_EPOCH_EXIT(et);
411 		return (EADDRNOTAVAIL);
412 	}
413 
414 	error = 0;
415 	switch (cmd) {
416 	case SIOCGIFADDR:
417 		*addr = ia->ia_addr;
418 		break;
419 
420 	case SIOCGIFBRDADDR:
421 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
422 			error = EINVAL;
423 			break;
424 		}
425 		*addr = ia->ia_broadaddr;
426 		break;
427 
428 	case SIOCGIFDSTADDR:
429 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
430 			error = EINVAL;
431 			break;
432 		}
433 		*addr = ia->ia_dstaddr;
434 		break;
435 
436 	case SIOCGIFNETMASK:
437 		*addr = ia->ia_sockmask;
438 		break;
439 	}
440 
441 	NET_EPOCH_EXIT(et);
442 
443 	return (error);
444 }
445 
446 int
in_mask2len(struct in_addr * mask)447 in_mask2len(struct in_addr *mask)
448 {
449 	int x, y;
450 	u_char *p;
451 
452 	p = (u_char *)mask;
453 	for (x = 0; x < sizeof(*mask); x++) {
454 		if (p[x] != 0xff)
455 			break;
456 	}
457 	y = 0;
458 	if (x < sizeof(*mask)) {
459 		for (y = 0; y < 8; y++) {
460 			if ((p[x] & (0x80 >> y)) == 0)
461 				break;
462 		}
463 	}
464 	return (x * 8 + y);
465 }
466 
467 int
in_control(struct socket * so,u_long cmd,void * data,struct ifnet * ifp,struct thread * td)468 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
469     struct thread *td)
470 {
471 	return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
472 }
473 
474 static int
in_aifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)475 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
476 {
477 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
478 	const struct sockaddr_in *addr = &ifra->ifra_addr;
479 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
480 	const struct sockaddr_in *mask = &ifra->ifra_mask;
481 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
482 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
483 	struct epoch_tracker et;
484 	struct ifaddr *ifa;
485 	struct in_ifaddr *ia;
486 	bool iaIsFirst;
487 	int error = 0;
488 
489 	error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
490 	if (error)
491 		return (error);
492 
493 	/*
494 	 * ifra_addr must be present and be of INET family.
495 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
496 	 */
497 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
498 	    addr->sin_family != AF_INET)
499 		return (EINVAL);
500 	if (broadaddr->sin_len != 0 &&
501 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
502 	    broadaddr->sin_family != AF_INET))
503 		return (EINVAL);
504 	if (mask->sin_len != 0 &&
505 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
506 	    mask->sin_family != AF_INET))
507 		return (EINVAL);
508 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
509 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
510 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
511 		return (EDESTADDRREQ);
512 	if (vhid != 0 && carp_attach_p == NULL)
513 		return (EPROTONOSUPPORT);
514 
515 #ifdef MAC
516 	/* Check if a MAC policy disallows setting the IPv4 address. */
517 	error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp);
518 	if (error != 0)
519 		return (error);
520 #endif
521 
522 	/*
523 	 * Check if bridge wants to allow adding addrs to member interfaces.
524 	 */
525 	if (ifp->if_bridge && bridge_member_ifaddrs_p &&
526 	    !bridge_member_ifaddrs_p())
527 		return (EINVAL);
528 
529 	/*
530 	 * See whether address already exist.
531 	 */
532 	iaIsFirst = true;
533 	ia = NULL;
534 	NET_EPOCH_ENTER(et);
535 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
536 		struct in_ifaddr *it;
537 
538 		if (ifa->ifa_addr->sa_family != AF_INET)
539 			continue;
540 
541 		it = (struct in_ifaddr *)ifa;
542 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
543 		    prison_check_ip4(cred, &addr->sin_addr) == 0)
544 			ia = it;
545 		else
546 			iaIsFirst = false;
547 	}
548 	NET_EPOCH_EXIT(et);
549 
550 	if (ia != NULL)
551 		(void )in_difaddr_ioctl(cmd, data, ifp, cred);
552 
553 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
554 	ia = (struct in_ifaddr *)ifa;
555 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
556 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
557 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
558 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
559 	    CALLOUT_RETURNUNLOCKED);
560 
561 	ia->ia_ifp = ifp;
562 	ia->ia_addr = *addr;
563 	if (mask->sin_len != 0) {
564 		ia->ia_sockmask = *mask;
565 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
566 	} else {
567 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
568 
569 		/*
570 	 	 * If netmask isn't supplied, use historical default.
571 		 * This is deprecated for interfaces other than loopback
572 		 * or point-to-point; warn in other cases.  In the future
573 		 * we should return an error rather than warning.
574 	 	 */
575 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
576 			printf("%s: set address: WARNING: network mask "
577 			     "should be specified; using historical default\n",
578 			     ifp->if_xname);
579 		if (IN_CLASSA(i))
580 			ia->ia_subnetmask = IN_CLASSA_NET;
581 		else if (IN_CLASSB(i))
582 			ia->ia_subnetmask = IN_CLASSB_NET;
583 		else
584 			ia->ia_subnetmask = IN_CLASSC_NET;
585 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
586 	}
587 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
588 	in_socktrim(&ia->ia_sockmask);
589 
590 	if (ifp->if_flags & IFF_BROADCAST) {
591 		if (broadaddr->sin_len != 0) {
592 			ia->ia_broadaddr = *broadaddr;
593 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
594 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
595 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
596 			ia->ia_broadaddr.sin_family = AF_INET;
597 		} else {
598 			ia->ia_broadaddr.sin_addr.s_addr =
599 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
600 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
601 			ia->ia_broadaddr.sin_family = AF_INET;
602 		}
603 	}
604 
605 	if (ifp->if_flags & IFF_POINTOPOINT)
606 		ia->ia_dstaddr = *dstaddr;
607 
608 	if (vhid != 0) {
609 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
610 		if (error)
611 			return (error);
612 	}
613 
614 	/* if_addrhead is already referenced by ifa_alloc() */
615 	IF_ADDR_WLOCK(ifp);
616 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
617 	IF_ADDR_WUNLOCK(ifp);
618 
619 	ifa_ref(ifa);			/* in_ifaddrhead */
620 	sx_assert(&in_control_sx, SA_XLOCKED);
621 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
622 	CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
623 	    ia_hash);
624 
625 	/*
626 	 * Give the interface a chance to initialize
627 	 * if this is its first address,
628 	 * and to validate the address if necessary.
629 	 */
630 	if (ifp->if_ioctl != NULL) {
631 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
632 		if (error)
633 			goto fail1;
634 	}
635 
636 	/*
637 	 * Add route for the network.
638 	 */
639 	if (vhid == 0) {
640 		error = in_addprefix(ia);
641 		if (error)
642 			goto fail1;
643 	}
644 
645 	/*
646 	 * Add a loopback route to self.
647 	 */
648 	if (vhid == 0 && ia_need_loopback_route(ia)) {
649 		struct in_ifaddr *eia;
650 
651 		eia = in_localip_more(ia);
652 
653 		if (eia == NULL) {
654 			error = ifa_add_loopback_route((struct ifaddr *)ia,
655 			    (struct sockaddr *)&ia->ia_addr);
656 			if (error)
657 				goto fail2;
658 		} else
659 			ifa_free(&eia->ia_ifa);
660 	}
661 
662 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
663 		struct in_addr allhosts_addr;
664 		struct in_ifinfo *ii;
665 
666 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
667 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
668 
669 		error = in_joingroup(ifp, &allhosts_addr, NULL,
670 			&ii->ii_allhosts);
671 	}
672 
673 	/*
674 	 * Note: we don't need extra reference for ifa, since we called
675 	 * with sx lock held, and ifaddr can not be deleted in concurrent
676 	 * thread.
677 	 */
678 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
679 
680 	return (error);
681 
682 fail2:
683 	if (vhid == 0)
684 		(void )in_scrubprefix(ia, LLE_STATIC);
685 
686 fail1:
687 	if (ia->ia_ifa.ifa_carp)
688 		(*carp_detach_p)(&ia->ia_ifa, false);
689 
690 	IF_ADDR_WLOCK(ifp);
691 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
692 	IF_ADDR_WUNLOCK(ifp);
693 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
694 
695 	sx_assert(&in_control_sx, SA_XLOCKED);
696 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
697 	CK_LIST_REMOVE(ia, ia_hash);
698 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
699 
700 	return (error);
701 }
702 
703 static int
in_difaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)704 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
705 {
706 	const struct ifreq *ifr = (struct ifreq *)data;
707 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
708 	    &ifr->ifr_addr;
709 	struct ifaddr *ifa;
710 	struct in_ifaddr *ia;
711 	bool deleteAny, iaIsLast;
712 	int error;
713 
714 	if (cred != NULL) {
715 		error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
716 		if (error)
717 			return (error);
718 	}
719 
720 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
721 	    addr->sin_family != AF_INET)
722 		deleteAny = true;
723 	else
724 		deleteAny = false;
725 
726 	iaIsLast = true;
727 	ia = NULL;
728 	IF_ADDR_WLOCK(ifp);
729 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
730 		struct in_ifaddr *it;
731 
732 		if (ifa->ifa_addr->sa_family != AF_INET)
733 			continue;
734 
735 		it = (struct in_ifaddr *)ifa;
736 		if (deleteAny && ia == NULL && (cred == NULL ||
737 		    prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
738 			ia = it;
739 
740 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
741 		    (cred == NULL || prison_check_ip4(cred,
742 		    &addr->sin_addr) == 0))
743 			ia = it;
744 
745 		if (it != ia)
746 			iaIsLast = false;
747 	}
748 
749 	if (ia == NULL) {
750 		IF_ADDR_WUNLOCK(ifp);
751 		return (EADDRNOTAVAIL);
752 	}
753 
754 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
755 	IF_ADDR_WUNLOCK(ifp);
756 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
757 
758 	sx_assert(&in_control_sx, SA_XLOCKED);
759 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
760 	CK_LIST_REMOVE(ia, ia_hash);
761 
762 	/*
763 	 * in_scrubprefix() kills the interface route.
764 	 */
765 	in_scrubprefix(ia, LLE_STATIC);
766 
767 	/*
768 	 * in_ifadown gets rid of all the rest of
769 	 * the routes.  This is not quite the right
770 	 * thing to do, but at least if we are running
771 	 * a routing process they will come back.
772 	 */
773 	in_ifadown(&ia->ia_ifa, 1);
774 
775 	if (ia->ia_ifa.ifa_carp)
776 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
777 
778 	/*
779 	 * If this is the last IPv4 address configured on this
780 	 * interface, leave the all-hosts group.
781 	 * No state-change report need be transmitted.
782 	 */
783 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
784 		struct in_ifinfo *ii;
785 
786 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
787 		if (ii->ii_allhosts) {
788 			(void)in_leavegroup(ii->ii_allhosts, NULL);
789 			ii->ii_allhosts = NULL;
790 		}
791 	}
792 
793 	IF_ADDR_WLOCK(ifp);
794 	if (callout_stop(&ia->ia_garp_timer) == 1) {
795 		ifa_free(&ia->ia_ifa);
796 	}
797 	IF_ADDR_WUNLOCK(ifp);
798 
799 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
800 	    IFADDR_EVENT_DEL);
801 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
802 
803 	return (0);
804 }
805 
806 static int
in_gifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)807 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
808 {
809 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
810 	const struct sockaddr_in *addr = &ifra->ifra_addr;
811 	struct epoch_tracker et;
812 	struct ifaddr *ifa;
813 	struct in_ifaddr *ia;
814 
815 	/*
816 	 * ifra_addr must be present and be of INET family.
817 	 */
818 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
819 	    addr->sin_family != AF_INET)
820 		return (EINVAL);
821 
822 	/*
823 	 * See whether address exist.
824 	 */
825 	ia = NULL;
826 	NET_EPOCH_ENTER(et);
827 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
828 		struct in_ifaddr *it;
829 
830 		if (ifa->ifa_addr->sa_family != AF_INET)
831 			continue;
832 
833 		it = (struct in_ifaddr *)ifa;
834 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
835 		    prison_check_ip4(cred, &addr->sin_addr) == 0) {
836 			ia = it;
837 			break;
838 		}
839 	}
840 	if (ia == NULL) {
841 		NET_EPOCH_EXIT(et);
842 		return (EADDRNOTAVAIL);
843 	}
844 
845 	ifra->ifra_mask = ia->ia_sockmask;
846 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
847 	    ia->ia_dstaddr.sin_family == AF_INET)
848 		ifra->ifra_dstaddr = ia->ia_dstaddr;
849 	else if ((ifp->if_flags & IFF_BROADCAST) &&
850 	    ia->ia_broadaddr.sin_family == AF_INET)
851 		ifra->ifra_broadaddr = ia->ia_broadaddr;
852 	else
853 		memset(&ifra->ifra_broadaddr, 0,
854 		    sizeof(ifra->ifra_broadaddr));
855 
856 	NET_EPOCH_EXIT(et);
857 	return (0);
858 }
859 
860 static int
in_match_ifaddr(const struct rtentry * rt,const struct nhop_object * nh,void * arg)861 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
862 {
863 
864 	if (nh->nh_ifa == (struct ifaddr *)arg)
865 		return (1);
866 
867 	return (0);
868 }
869 
870 static int
in_handle_prefix_route(uint32_t fibnum,int cmd,struct sockaddr_in * dst,struct sockaddr_in * netmask,struct ifaddr * ifa,struct ifnet * ifp)871 in_handle_prefix_route(uint32_t fibnum, int cmd,
872     struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
873     struct ifnet *ifp)
874 {
875 
876 	NET_EPOCH_ASSERT();
877 
878 	/* Prepare gateway */
879 	struct sockaddr_dl_short sdl = {
880 		.sdl_family = AF_LINK,
881 		.sdl_len = sizeof(struct sockaddr_dl_short),
882 		.sdl_type = ifa->ifa_ifp->if_type,
883 		.sdl_index = ifa->ifa_ifp->if_index,
884 	};
885 
886 	struct rt_addrinfo info = {
887 		.rti_ifa = ifa,
888 		.rti_ifp = ifp,
889 		.rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
890 		.rti_info = {
891 			[RTAX_DST] = (struct sockaddr *)dst,
892 			[RTAX_NETMASK] = (struct sockaddr *)netmask,
893 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
894 		},
895 		/* Ensure we delete the prefix IFF prefix ifa matches */
896 		.rti_filter = in_match_ifaddr,
897 		.rti_filterdata = ifa,
898 	};
899 
900 	return (rib_handle_ifaddr_info(fibnum, cmd, &info));
901 }
902 
903 /*
904  * Routing table interaction with interface addresses.
905  *
906  * In general, two types of routes needs to be installed:
907  * a) "interface" or "prefix" route, telling user that the addresses
908  *   behind the ifa prefix are reached directly.
909  * b) "loopback" route installed for the ifa address, telling user that
910  *   the address belongs to local system.
911  *
912  * Handling for (a) and (b) differs in multi-fib aspects, hence they
913  *  are implemented in different functions below.
914  *
915  * The cases above may intersect - /32 interface aliases results in
916  *  the same prefix produced by (a) and (b). This blurs the definition
917  *  of the "loopback" route and complicate interactions. The interaction
918  *  table is defined below. The case numbers are used in the multiple
919  *  functions below to refer to the particular test case.
920  *
921  * There can be multiple options:
922  * 1) Adding address with prefix on non-p2p/non-loopback interface.
923  *  Example: 192.0.2.1/24. Action:
924  *  * add "prefix" route towards 192.0.2.0/24 via @ia interface,
925  *    using @ia as an address source.
926  *  * add "loopback" route towards 192.0.2.1 via V_loif, saving
927  *   @ia ifp in the gateway and using @ia as an address source.
928  *
929  * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
930  *  Example: 192.0.2.2/32. Action:
931  *  * add "prefix" host route via V_loif, using @ia as an address source.
932  *
933  * 3) Adding address with or without prefix to p2p interface.
934  *  Example: 10.0.0.1/24->10.0.0.2. Action:
935  *  * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
936  *    as an address source. Note: no sense in installing full /24 as the interface
937  *    is point-to-point.
938  *  * add "loopback" route towards 10.0.9.1 via V_loif, saving
939  *   @ia ifp in the gateway and using @ia as an address source.
940  *
941  * 4) Adding address with or without prefix to loopback interface.
942  *  Example: 192.0.2.1/24. Action:
943  *  * add "prefix" host route via @ia interface, using @ia as an address source.
944  *    Note: Skip installing /24 prefix as it would introduce TTL loop
945  *    for the traffic destined to these addresses.
946  */
947 
948 /*
949  * Checks if @ia needs to install loopback route to @ia address via
950  *  ifa_maintain_loopback_route().
951  *
952  * Return true on success.
953  */
954 static bool
ia_need_loopback_route(const struct in_ifaddr * ia)955 ia_need_loopback_route(const struct in_ifaddr *ia)
956 {
957 	struct ifnet *ifp = ia->ia_ifp;
958 
959 	/* Case 4: Skip loopback interfaces */
960 	if ((ifp->if_flags & IFF_LOOPBACK) ||
961 	    (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
962 		return (false);
963 
964 	/* Clash avoidance: Skip p2p interfaces with both addresses are equal */
965 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
966 	    ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
967 		return (false);
968 
969 	/* Case 2: skip /32 prefixes */
970 	if (!(ifp->if_flags & IFF_POINTOPOINT) &&
971 	    (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
972 		return (false);
973 
974 	return (true);
975 }
976 
977 /*
978  * Calculate "prefix" route corresponding to @ia.
979  */
980 static void
ia_getrtprefix(const struct in_ifaddr * ia,struct in_addr * prefix,struct in_addr * mask)981 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
982 {
983 
984 	if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
985 		/* Case 3: return host route for dstaddr */
986 		*prefix = ia->ia_dstaddr.sin_addr;
987 		mask->s_addr = INADDR_BROADCAST;
988 	} else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
989 		/* Case 4: return host route for ifaddr */
990 		*prefix = ia->ia_addr.sin_addr;
991 		mask->s_addr = INADDR_BROADCAST;
992 	} else {
993 		/* Cases 1,2: return actual ia prefix */
994 		*prefix = ia->ia_addr.sin_addr;
995 		*mask = ia->ia_sockmask.sin_addr;
996 		prefix->s_addr &= mask->s_addr;
997 	}
998 }
999 
1000 /*
1001  * Adds or delete interface "prefix" route corresponding to @ifa.
1002  *  Returns 0 on success or errno.
1003  */
1004 static int
in_handle_ifaddr_route(int cmd,struct in_ifaddr * ia)1005 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
1006 {
1007 	struct ifaddr *ifa = &ia->ia_ifa;
1008 	struct in_addr daddr, maddr;
1009 	struct sockaddr_in *pmask;
1010 	struct epoch_tracker et;
1011 	int error;
1012 
1013 	ia_getrtprefix(ia, &daddr, &maddr);
1014 
1015 	struct sockaddr_in mask = {
1016 		.sin_family = AF_INET,
1017 		.sin_len = sizeof(struct sockaddr_in),
1018 		.sin_addr = maddr,
1019 	};
1020 
1021 	pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
1022 
1023 	struct sockaddr_in dst = {
1024 		.sin_family = AF_INET,
1025 		.sin_len = sizeof(struct sockaddr_in),
1026 		.sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
1027 	};
1028 
1029 	struct ifnet *ifp = ia->ia_ifp;
1030 
1031 	if ((maddr.s_addr == INADDR_BROADCAST) &&
1032 	    (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
1033 		/* Case 2: host route on broadcast interface */
1034 		ifp = V_loif;
1035 	}
1036 
1037 	uint32_t fibnum = ifa->ifa_ifp->if_fib;
1038 	NET_EPOCH_ENTER(et);
1039 	error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
1040 	NET_EPOCH_EXIT(et);
1041 
1042 	return (error);
1043 }
1044 
1045 /*
1046  * Check if we have a route for the given prefix already.
1047  */
1048 static bool
in_hasrtprefix(struct in_ifaddr * target)1049 in_hasrtprefix(struct in_ifaddr *target)
1050 {
1051 	struct epoch_tracker et;
1052 	struct in_ifaddr *ia;
1053 	struct in_addr prefix, mask, p, m;
1054 	bool result = false;
1055 
1056 	ia_getrtprefix(target, &prefix, &mask);
1057 
1058 	/* Look for an existing address with the same prefix, mask, and fib */
1059 	NET_EPOCH_ENTER(et);
1060 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1061 		ia_getrtprefix(ia, &p, &m);
1062 
1063 		if (prefix.s_addr != p.s_addr ||
1064 		    mask.s_addr != m.s_addr)
1065 			continue;
1066 
1067 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1068 			continue;
1069 
1070 		/*
1071 		 * If we got a matching prefix route inserted by other
1072 		 * interface address, we are done here.
1073 		 */
1074 		if (ia->ia_flags & IFA_ROUTE) {
1075 			result = true;
1076 			break;
1077 		}
1078 	}
1079 	NET_EPOCH_EXIT(et);
1080 
1081 	return (result);
1082 }
1083 
1084 int
in_addprefix(struct in_ifaddr * target)1085 in_addprefix(struct in_ifaddr *target)
1086 {
1087 	int error;
1088 
1089 	if (in_hasrtprefix(target)) {
1090 		if (V_nosameprefix)
1091 			return (EEXIST);
1092 		else {
1093 			rt_addrmsg(RTM_ADD, &target->ia_ifa,
1094 			    target->ia_ifp->if_fib);
1095 			return (0);
1096 		}
1097 	}
1098 
1099 	/*
1100 	 * No-one seem to have this prefix route, so we try to insert it.
1101 	 */
1102 	rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1103 	error = in_handle_ifaddr_route(RTM_ADD, target);
1104 	if (!error)
1105 		target->ia_flags |= IFA_ROUTE;
1106 	return (error);
1107 }
1108 
1109 /*
1110  * Removes either all lle entries for given @ia, or lle
1111  * corresponding to @ia address.
1112  */
1113 static void
in_scrubprefixlle(struct in_ifaddr * ia,int all,u_int flags)1114 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1115 {
1116 	struct sockaddr_in addr, mask;
1117 	struct sockaddr *saddr, *smask;
1118 	struct ifnet *ifp;
1119 
1120 	saddr = (struct sockaddr *)&addr;
1121 	bzero(&addr, sizeof(addr));
1122 	addr.sin_len = sizeof(addr);
1123 	addr.sin_family = AF_INET;
1124 	smask = (struct sockaddr *)&mask;
1125 	bzero(&mask, sizeof(mask));
1126 	mask.sin_len = sizeof(mask);
1127 	mask.sin_family = AF_INET;
1128 	mask.sin_addr.s_addr = ia->ia_subnetmask;
1129 	ifp = ia->ia_ifp;
1130 
1131 	if (all) {
1132 		/*
1133 		 * Remove all L2 entries matching given prefix.
1134 		 * Convert address to host representation to avoid
1135 		 * doing this on every callback. ia_subnetmask is already
1136 		 * stored in host representation.
1137 		 */
1138 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1139 		lltable_prefix_free(AF_INET, saddr, smask, flags);
1140 	} else {
1141 		/* Remove interface address only */
1142 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1143 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1144 	}
1145 }
1146 
1147 /*
1148  * If there is no other address in the system that can serve a route to the
1149  * same prefix, remove the route.  Hand over the route to the new address
1150  * otherwise.
1151  */
1152 int
in_scrubprefix(struct in_ifaddr * target,u_int flags)1153 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1154 {
1155 	struct epoch_tracker et;
1156 	struct in_ifaddr *ia;
1157 	struct in_addr prefix, mask, p, m;
1158 	int error = 0;
1159 
1160 	/*
1161 	 * Remove the loopback route to the interface address.
1162 	 */
1163 	if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1164 		struct in_ifaddr *eia;
1165 
1166 		eia = in_localip_more(target);
1167 
1168 		if (eia != NULL) {
1169 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
1170 			    (struct sockaddr *)&target->ia_addr);
1171 			ifa_free(&eia->ia_ifa);
1172 		} else {
1173 			error = ifa_del_loopback_route((struct ifaddr *)target,
1174 			    (struct sockaddr *)&target->ia_addr);
1175 		}
1176 	}
1177 
1178 	ia_getrtprefix(target, &prefix, &mask);
1179 
1180 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1181 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1182 
1183 		/*
1184 		 * Removing address from !IFF_UP interface or
1185 		 * prefix which exists on other interface (along with route).
1186 		 * No entries should exist here except target addr.
1187 		 * Given that, delete this entry only.
1188 		 */
1189 		in_scrubprefixlle(target, 0, flags);
1190 		return (0);
1191 	}
1192 
1193 	NET_EPOCH_ENTER(et);
1194 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1195 		ia_getrtprefix(ia, &p, &m);
1196 
1197 		if (prefix.s_addr != p.s_addr ||
1198 		    mask.s_addr != m.s_addr)
1199 			continue;
1200 
1201 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1202 			continue;
1203 
1204 		/*
1205 		 * If we got a matching prefix address, move IFA_ROUTE and
1206 		 * the route itself to it.  Make sure that routing daemons
1207 		 * get a heads-up.
1208 		 */
1209 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1210 			ifa_ref(&ia->ia_ifa);
1211 			NET_EPOCH_EXIT(et);
1212 			error = in_handle_ifaddr_route(RTM_DELETE, target);
1213 			if (error == 0)
1214 				target->ia_flags &= ~IFA_ROUTE;
1215 			else
1216 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1217 					error);
1218 			/* Scrub all entries IFF interface is different */
1219 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1220 			    flags);
1221 			error = in_handle_ifaddr_route(RTM_ADD, ia);
1222 			if (error == 0)
1223 				ia->ia_flags |= IFA_ROUTE;
1224 			else
1225 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1226 					error);
1227 			ifa_free(&ia->ia_ifa);
1228 			return (error);
1229 		}
1230 	}
1231 	NET_EPOCH_EXIT(et);
1232 
1233 	/*
1234 	 * remove all L2 entries on the given prefix
1235 	 */
1236 	in_scrubprefixlle(target, 1, flags);
1237 
1238 	/*
1239 	 * As no-one seem to have this prefix, we can remove the route.
1240 	 */
1241 	rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1242 	error = in_handle_ifaddr_route(RTM_DELETE, target);
1243 	if (error == 0)
1244 		target->ia_flags &= ~IFA_ROUTE;
1245 	else
1246 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1247 	return (error);
1248 }
1249 
1250 void
in_ifscrub_all(void)1251 in_ifscrub_all(void)
1252 {
1253 	struct ifnet *ifp;
1254 	struct ifaddr *ifa, *nifa;
1255 	struct ifreq ifr;
1256 
1257 	IFNET_RLOCK();
1258 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1259 		/* Cannot lock here - lock recursion. */
1260 		/* NET_EPOCH_ENTER(et); */
1261 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1262 			if (ifa->ifa_addr->sa_family != AF_INET)
1263 				continue;
1264 
1265 			/*
1266 			 * This is ugly but the only way for legacy IP to
1267 			 * cleanly remove addresses and everything attached.
1268 			 */
1269 			bzero(&ifr, sizeof(ifr));
1270 			ifr.ifr_addr = *ifa->ifa_addr;
1271 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1272 			    ifp, NULL);
1273 		}
1274 		/* NET_EPOCH_EXIT(et); */
1275 		in_purgemaddrs(ifp);
1276 		igmp_domifdetach(ifp);
1277 	}
1278 	IFNET_RUNLOCK();
1279 }
1280 
1281 bool
in_ifaddr_broadcast(struct in_addr in,struct in_ifaddr * ia)1282 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1283 {
1284 
1285 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1286 	     /*
1287 	      * Optionally check for old-style (host 0) broadcast, but
1288 	      * taking into account that RFC 3021 obsoletes it.
1289 	      */
1290 	    __predict_false(V_broadcast_lowest &&
1291 	    ia->ia_subnetmask != IN_RFC3021_MASK &&
1292 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
1293 	     /*
1294 	      * Check for an all one subnetmask. These
1295 	      * only exist when an interface gets a secondary
1296 	      * address.
1297 	      */
1298 	    ia->ia_subnetmask != (u_long)0xffffffff);
1299 }
1300 
1301 /*
1302  * Return true if the address might be a local broadcast address.
1303  */
1304 bool
in_ifnet_broadcast(struct in_addr in,struct ifnet * ifp)1305 in_ifnet_broadcast(struct in_addr in, struct ifnet *ifp)
1306 {
1307 	struct ifaddr *ifa;
1308 
1309 	NET_EPOCH_ASSERT();
1310 
1311 	if (in_broadcast(in))
1312 		return (true);
1313 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1314 		return (false);
1315 	/*
1316 	 * Look through the list of addresses for a match
1317 	 * with a broadcast address.
1318 	 */
1319 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1320 		if (ifa->ifa_addr->sa_family == AF_INET &&
1321 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa))
1322 			return (true);
1323 	return (false);
1324 }
1325 
1326 /*
1327  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1328  */
1329 void
in_ifdetach(struct ifnet * ifp)1330 in_ifdetach(struct ifnet *ifp)
1331 {
1332 	IN_MULTI_LOCK();
1333 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1334 	in_pcbpurgeif0(&V_udbinfo, ifp);
1335 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1336 	in_purgemaddrs(ifp);
1337 	IN_MULTI_UNLOCK();
1338 
1339 	/*
1340 	 * Make sure all multicast deletions invoking if_ioctl() are
1341 	 * completed before returning. Else we risk accessing a freed
1342 	 * ifnet structure pointer.
1343 	 */
1344 	inm_release_wait(NULL);
1345 }
1346 
1347 static void
in_ifnet_event(void * arg __unused,struct ifnet * ifp,int event)1348 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1349 {
1350 	struct epoch_tracker et;
1351 	struct ifaddr *ifa;
1352 	struct in_ifaddr *ia;
1353 	int error;
1354 
1355 	NET_EPOCH_ENTER(et);
1356 	switch (event) {
1357 	case IFNET_EVENT_DOWN:
1358 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1359 			if (ifa->ifa_addr->sa_family != AF_INET)
1360 				continue;
1361 			ia = (struct in_ifaddr *)ifa;
1362 			if ((ia->ia_flags & IFA_ROUTE) == 0)
1363 				continue;
1364 			ifa_ref(ifa);
1365 			/*
1366 			 * in_scrubprefix() kills the interface route.
1367 			 */
1368 			in_scrubprefix(ia, 0);
1369 			/*
1370 			 * in_ifadown gets rid of all the rest of the
1371 			 * routes.  This is not quite the right thing
1372 			 * to do, but at least if we are running a
1373 			 * routing process they will come back.
1374 			 */
1375 			in_ifadown(ifa, 0);
1376 			ifa_free(ifa);
1377 		}
1378 		break;
1379 
1380 	case IFNET_EVENT_UP:
1381 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1382 			if (ifa->ifa_addr->sa_family != AF_INET)
1383 				continue;
1384 			ia = (struct in_ifaddr *)ifa;
1385 			if (ia->ia_flags & IFA_ROUTE)
1386 				continue;
1387 			ifa_ref(ifa);
1388 			error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1389 			rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1390 			error = in_handle_ifaddr_route(RTM_ADD, ia);
1391 			if (error == 0)
1392 				ia->ia_flags |= IFA_ROUTE;
1393 			error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1394 			ifa_free(ifa);
1395 		}
1396 		break;
1397 	}
1398 	NET_EPOCH_EXIT(et);
1399 }
1400 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1401 
1402 /*
1403  * Delete all IPv4 multicast address records, and associated link-layer
1404  * multicast address records, associated with ifp.
1405  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1406  * XXX This should not race with ifma_protospec being set during
1407  * a new allocation, if it does, we have bigger problems.
1408  */
1409 static void
in_purgemaddrs(struct ifnet * ifp)1410 in_purgemaddrs(struct ifnet *ifp)
1411 {
1412 	struct epoch_tracker	 et;
1413 	struct in_multi_head purgeinms;
1414 	struct in_multi		*inm;
1415 	struct ifmultiaddr	*ifma;
1416 
1417 	SLIST_INIT(&purgeinms);
1418 	IN_MULTI_LIST_LOCK();
1419 
1420 	/*
1421 	 * Extract list of in_multi associated with the detaching ifp
1422 	 * which the PF_INET layer is about to release.
1423 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1424 	 * by code further down.
1425 	 */
1426 	IF_ADDR_WLOCK(ifp);
1427 	NET_EPOCH_ENTER(et);
1428 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1429 		inm = inm_ifmultiaddr_get_inm(ifma);
1430 		if (inm == NULL)
1431 			continue;
1432 		inm_rele_locked(&purgeinms, inm);
1433 	}
1434 	NET_EPOCH_EXIT(et);
1435 	IF_ADDR_WUNLOCK(ifp);
1436 
1437 	inm_release_list_deferred(&purgeinms);
1438 	igmp_ifdetach(ifp);
1439 	IN_MULTI_LIST_UNLOCK();
1440 }
1441 
1442 struct in_llentry {
1443 	struct llentry		base;
1444 };
1445 
1446 #define	IN_LLTBL_DEFAULT_HSIZE	32
1447 #define	IN_LLTBL_HASH(k, h) \
1448 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1449 
1450 /*
1451  * Do actual deallocation of @lle.
1452  */
1453 static void
in_lltable_destroy_lle_unlocked(epoch_context_t ctx)1454 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1455 {
1456 	struct llentry *lle;
1457 
1458 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1459 	LLE_LOCK_DESTROY(lle);
1460 	LLE_REQ_DESTROY(lle);
1461 	free(lle, M_LLTABLE);
1462 }
1463 
1464 /*
1465  * Called by LLE_FREE_LOCKED when number of references
1466  * drops to zero.
1467  */
1468 static void
in_lltable_destroy_lle(struct llentry * lle)1469 in_lltable_destroy_lle(struct llentry *lle)
1470 {
1471 
1472 	LLE_WUNLOCK(lle);
1473 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1474 }
1475 
1476 static struct llentry *
in_lltable_new(struct in_addr addr4,u_int flags)1477 in_lltable_new(struct in_addr addr4, u_int flags)
1478 {
1479 	struct in_llentry *lle;
1480 
1481 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1482 	if (lle == NULL)		/* NB: caller generates msg */
1483 		return NULL;
1484 
1485 	/*
1486 	 * For IPv4 this will trigger "arpresolve" to generate
1487 	 * an ARP request.
1488 	 */
1489 	lle->base.la_expire = time_uptime; /* mark expired */
1490 	lle->base.r_l3addr.addr4 = addr4;
1491 	lle->base.lle_refcnt = 1;
1492 	lle->base.lle_free = in_lltable_destroy_lle;
1493 	LLE_LOCK_INIT(&lle->base);
1494 	LLE_REQ_INIT(&lle->base);
1495 	callout_init(&lle->base.lle_timer, 1);
1496 
1497 	return (&lle->base);
1498 }
1499 
1500 static int
in_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)1501 in_lltable_match_prefix(const struct sockaddr *saddr,
1502     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1503 {
1504 	struct in_addr addr, mask, lle_addr;
1505 
1506 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1507 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1508 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1509 
1510 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1511 		return (0);
1512 
1513 	if (lle->la_flags & LLE_IFADDR) {
1514 		/*
1515 		 * Delete LLE_IFADDR records IFF address & flag matches.
1516 		 * Note that addr is the interface address within prefix
1517 		 * being matched.
1518 		 * Note also we should handle 'ifdown' cases without removing
1519 		 * ifaddr macs.
1520 		 */
1521 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1522 			return (1);
1523 		return (0);
1524 	}
1525 
1526 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1527 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1528 		return (1);
1529 
1530 	return (0);
1531 }
1532 
1533 static void
in_lltable_free_entry(struct lltable * llt,struct llentry * lle)1534 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1535 {
1536 	size_t pkts_dropped;
1537 
1538 	LLE_WLOCK_ASSERT(lle);
1539 	KASSERT(llt != NULL, ("lltable is NULL"));
1540 
1541 	/* Unlink entry from table if not already */
1542 	if ((lle->la_flags & LLE_LINKED) != 0) {
1543 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1544 		lltable_unlink_entry(llt, lle);
1545 	}
1546 
1547 	/* Drop hold queue */
1548 	pkts_dropped = llentry_free(lle);
1549 	ARPSTAT_ADD(dropped, pkts_dropped);
1550 }
1551 
1552 static int
in_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)1553 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1554 {
1555 	struct nhop_object *nh;
1556 	struct in_addr addr;
1557 
1558 	KASSERT(l3addr->sa_family == AF_INET,
1559 	    ("sin_family %d", l3addr->sa_family));
1560 
1561 	addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1562 
1563 	nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1564 	if (nh == NULL)
1565 		return (EINVAL);
1566 
1567 	/*
1568 	 * If the gateway for an existing host route matches the target L3
1569 	 * address, which is a special route inserted by some implementation
1570 	 * such as MANET, and the interface is of the correct type, then
1571 	 * allow for ARP to proceed.
1572 	 */
1573 	if (nh->nh_flags & NHF_GATEWAY) {
1574 		if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1575 		    (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1576 		    memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1577 		    sizeof(in_addr_t)) != 0) {
1578 			return (EINVAL);
1579 		}
1580 	}
1581 
1582 	/*
1583 	 * Make sure that at least the destination address is covered
1584 	 * by the route. This is for handling the case where 2 or more
1585 	 * interfaces have the same prefix. An incoming packet arrives
1586 	 * on one interface and the corresponding outgoing packet leaves
1587 	 * another interface.
1588 	 */
1589 	if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1590 		struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1591 		struct in_addr dst_addr, mask_addr;
1592 
1593 		if (ia == NULL)
1594 			return (EINVAL);
1595 
1596 		/*
1597 		 * ifaof_ifpforaddr() returns _best matching_ IFA.
1598 		 * It is possible that ifa prefix does not cover our address.
1599 		 * Explicitly verify and fail if that's the case.
1600 		 */
1601 		dst_addr = IA_SIN(ia)->sin_addr;
1602 		mask_addr.s_addr = htonl(ia->ia_subnetmask);
1603 
1604 		if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1605 			return (EINVAL);
1606 	}
1607 
1608 	return (0);
1609 }
1610 
1611 static inline uint32_t
in_lltable_hash_dst(const struct in_addr dst,uint32_t hsize)1612 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1613 {
1614 
1615 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1616 }
1617 
1618 static uint32_t
in_lltable_hash(const struct llentry * lle,uint32_t hsize)1619 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1620 {
1621 
1622 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1623 }
1624 
1625 static void
in_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)1626 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1627 {
1628 	struct sockaddr_in *sin;
1629 
1630 	sin = (struct sockaddr_in *)sa;
1631 	bzero(sin, sizeof(*sin));
1632 	sin->sin_family = AF_INET;
1633 	sin->sin_len = sizeof(*sin);
1634 	sin->sin_addr = lle->r_l3addr.addr4;
1635 }
1636 
1637 static inline struct llentry *
in_lltable_find_dst(struct lltable * llt,struct in_addr dst)1638 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1639 {
1640 	struct llentry *lle;
1641 	struct llentries *lleh;
1642 	u_int hashidx;
1643 
1644 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1645 	lleh = &llt->lle_head[hashidx];
1646 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1647 		if (lle->la_flags & LLE_DELETED)
1648 			continue;
1649 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1650 			break;
1651 	}
1652 
1653 	return (lle);
1654 }
1655 
1656 static void
in_lltable_delete_entry(struct lltable * llt,struct llentry * lle)1657 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1658 {
1659 
1660 	lle->la_flags |= LLE_DELETED;
1661 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1662 #ifdef DIAGNOSTIC
1663 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1664 #endif
1665 	llentry_free(lle);
1666 }
1667 
1668 static struct llentry *
in_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1669 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1670 {
1671 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1672 	struct ifnet *ifp = llt->llt_ifp;
1673 	struct llentry *lle;
1674 	char linkhdr[LLE_MAX_LINKHDR];
1675 	size_t linkhdrsize;
1676 	int lladdr_off;
1677 
1678 	KASSERT(l3addr->sa_family == AF_INET,
1679 	    ("sin_family %d", l3addr->sa_family));
1680 
1681 	/*
1682 	 * A route that covers the given address must have
1683 	 * been installed 1st because we are doing a resolution,
1684 	 * verify this.
1685 	 */
1686 	if (!(flags & LLE_IFADDR) &&
1687 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1688 		return (NULL);
1689 
1690 	lle = in_lltable_new(sin->sin_addr, flags);
1691 	if (lle == NULL) {
1692 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1693 		return (NULL);
1694 	}
1695 	lle->la_flags = flags;
1696 	if (flags & LLE_STATIC)
1697 		lle->r_flags |= RLLE_VALID;
1698 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1699 		linkhdrsize = LLE_MAX_LINKHDR;
1700 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1701 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1702 			in_lltable_free_entry(llt, lle);
1703 			return (NULL);
1704 		}
1705 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1706 		    lladdr_off);
1707 		lle->la_flags |= LLE_STATIC;
1708 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1709 		lle->la_expire = 0;
1710 	}
1711 
1712 	return (lle);
1713 }
1714 
1715 /*
1716  * Return NULL if not found or marked for deletion.
1717  * If found return lle read locked.
1718  */
1719 static struct llentry *
in_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1720 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1721 {
1722 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1723 	struct llentry *lle;
1724 
1725 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1726 	KASSERT(l3addr->sa_family == AF_INET,
1727 	    ("sin_family %d", l3addr->sa_family));
1728 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1729 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1730 	    ("wrong lle request flags: %#x", flags));
1731 
1732 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1733 	if (lle == NULL)
1734 		return (NULL);
1735 	if (flags & LLE_UNLOCKED)
1736 		return (lle);
1737 
1738 	if (flags & LLE_EXCLUSIVE)
1739 		LLE_WLOCK(lle);
1740 	else
1741 		LLE_RLOCK(lle);
1742 
1743 	/*
1744 	 * If the afdata lock is not held, the LLE may have been unlinked while
1745 	 * we were blocked on the LLE lock.  Check for this case.
1746 	 */
1747 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1748 		if (flags & LLE_EXCLUSIVE)
1749 			LLE_WUNLOCK(lle);
1750 		else
1751 			LLE_RUNLOCK(lle);
1752 		return (NULL);
1753 	}
1754 	return (lle);
1755 }
1756 
1757 static int
in_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)1758 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1759     struct sysctl_req *wr)
1760 {
1761 	struct ifnet *ifp = llt->llt_ifp;
1762 	/* XXX stack use */
1763 	struct {
1764 		struct rt_msghdr	rtm;
1765 		struct sockaddr_in	sin;
1766 		struct sockaddr_dl	sdl;
1767 	} arpc;
1768 	struct sockaddr_dl *sdl;
1769 	int error;
1770 
1771 	bzero(&arpc, sizeof(arpc));
1772 	/* skip deleted entries */
1773 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1774 		return (0);
1775 	/* Skip if jailed and not a valid IP of the prison. */
1776 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1777 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1778 		return (0);
1779 	/*
1780 	 * produce a msg made of:
1781 	 *  struct rt_msghdr;
1782 	 *  struct sockaddr_in; (IPv4)
1783 	 *  struct sockaddr_dl;
1784 	 */
1785 	arpc.rtm.rtm_msglen = sizeof(arpc);
1786 	arpc.rtm.rtm_version = RTM_VERSION;
1787 	arpc.rtm.rtm_type = RTM_GET;
1788 	arpc.rtm.rtm_flags = RTF_UP;
1789 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1790 
1791 	/* publish */
1792 	if (lle->la_flags & LLE_PUB)
1793 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1794 
1795 	sdl = &arpc.sdl;
1796 	sdl->sdl_family = AF_LINK;
1797 	sdl->sdl_len = sizeof(*sdl);
1798 	sdl->sdl_index = ifp->if_index;
1799 	sdl->sdl_type = ifp->if_type;
1800 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1801 		sdl->sdl_alen = ifp->if_addrlen;
1802 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1803 	} else {
1804 		sdl->sdl_alen = 0;
1805 		bzero(LLADDR(sdl), ifp->if_addrlen);
1806 	}
1807 
1808 	arpc.rtm.rtm_rmx.rmx_expire =
1809 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1810 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1811 	if (lle->la_flags & LLE_STATIC)
1812 		arpc.rtm.rtm_flags |= RTF_STATIC;
1813 	if (lle->la_flags & LLE_IFADDR)
1814 		arpc.rtm.rtm_flags |= RTF_PINNED;
1815 	arpc.rtm.rtm_index = ifp->if_index;
1816 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1817 
1818 	return (error);
1819 }
1820 
1821 static void
in_lltable_post_resolved(struct lltable * llt,struct llentry * lle)1822 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1823 {
1824 	struct ifnet *ifp = llt->llt_ifp;
1825 
1826 	/* gratuitous ARP */
1827 	if ((lle->la_flags & LLE_PUB) != 0)
1828 		arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1829 		    lle->ll_addr);
1830 }
1831 
1832 static struct lltable *
in_lltattach(struct ifnet * ifp)1833 in_lltattach(struct ifnet *ifp)
1834 {
1835 	struct lltable *llt;
1836 
1837 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1838  	llt->llt_af = AF_INET;
1839  	llt->llt_ifp = ifp;
1840 
1841 	llt->llt_lookup = in_lltable_lookup;
1842 	llt->llt_alloc_entry = in_lltable_alloc;
1843 	llt->llt_delete_entry = in_lltable_delete_entry;
1844 	llt->llt_dump_entry = in_lltable_dump_entry;
1845 	llt->llt_hash = in_lltable_hash;
1846 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1847 	llt->llt_free_entry = in_lltable_free_entry;
1848 	llt->llt_match_prefix = in_lltable_match_prefix;
1849 	llt->llt_mark_used = llentry_mark_used;
1850 	llt->llt_post_resolved = in_lltable_post_resolved;
1851  	lltable_link(llt);
1852 
1853 	return (llt);
1854 }
1855 
1856 struct lltable *
in_lltable_get(struct ifnet * ifp)1857 in_lltable_get(struct ifnet *ifp)
1858 {
1859 	struct lltable *llt = NULL;
1860 
1861 	void *afdata_ptr = ifp->if_afdata[AF_INET];
1862 	if (afdata_ptr != NULL)
1863 		llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1864 	return (llt);
1865 }
1866 
1867 void *
in_domifattach(struct ifnet * ifp)1868 in_domifattach(struct ifnet *ifp)
1869 {
1870 	struct in_ifinfo *ii;
1871 
1872 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1873 
1874 	ii->ii_llt = in_lltattach(ifp);
1875 	ii->ii_igmp = igmp_domifattach(ifp);
1876 
1877 	return (ii);
1878 }
1879 
1880 void
in_domifdetach(struct ifnet * ifp,void * aux)1881 in_domifdetach(struct ifnet *ifp, void *aux)
1882 {
1883 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1884 
1885 	igmp_domifdetach(ifp);
1886 	lltable_free(ii->ii_llt);
1887 	free(ii, M_IFADDR);
1888 }
1889