xref: /freebsd/sys/netinet/in.c (revision bc06c51419b338c3d6efa6543308f9905c12c79b)
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/route.h>
61 #include <net/route/nhop.h>
62 #include <net/route/route_ctl.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/if_ether.h>
66 #include <netinet/in.h>
67 #include <netinet/in_fib.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_carp.h>
72 #include <netinet/igmp_var.h>
73 #include <netinet/udp.h>
74 #include <netinet/udp_var.h>
75 
76 #ifdef MAC
77 #include <security/mac/mac_framework.h>
78 #endif
79 
80 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
81 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
83 
84 static void	in_socktrim(struct sockaddr_in *);
85 static void	in_purgemaddrs(struct ifnet *);
86 
87 static bool	ia_need_loopback_route(const struct in_ifaddr *);
88 
89 VNET_DEFINE_STATIC(int, nosameprefix);
90 #define	V_nosameprefix			VNET(nosameprefix)
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
92 	&VNET_NAME(nosameprefix), 0,
93 	"Refuse to create same prefixes on different interfaces");
94 
95 VNET_DEFINE_STATIC(bool, broadcast_lowest);
96 #define	V_broadcast_lowest		VNET(broadcast_lowest)
97 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
98 	&VNET_NAME(broadcast_lowest), 0,
99 	"Treat lowest address on a subnet (host 0) as broadcast");
100 
101 VNET_DEFINE(bool, ip_allow_net240) = false;
102 #define	V_ip_allow_net240		VNET(ip_allow_net240)
103 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
104 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
105 	"Allow use of Experimental addresses, aka Class E (240/4)");
106 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
107 
108 VNET_DEFINE(bool, ip_allow_net0) = false;
109 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
110 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
111 	"Allow use of addresses in network 0/8");
112 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
113 
114 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
115 #define	V_in_loopback_mask	VNET(in_loopback_mask)
116 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
117 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
118 	CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
119 	NULL, 0, sysctl_loopback_prefixlen, "I",
120 	"Prefix length of address space reserved for loopback");
121 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
122 
123 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
124 #define	V_ripcbinfo			VNET(ripcbinfo)
125 
126 static struct sx in_control_sx;
127 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
128 
129 /*
130  * Return 1 if an internet address is for a ``local'' host
131  * (one to which we have a connection).
132  */
133 int
in_localaddr(struct in_addr in)134 in_localaddr(struct in_addr in)
135 {
136 	u_long i = ntohl(in.s_addr);
137 	struct in_ifaddr *ia;
138 
139 	NET_EPOCH_ASSERT();
140 
141 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
142 		if ((i & ia->ia_subnetmask) == ia->ia_subnet)
143 			return (1);
144 	}
145 
146 	return (0);
147 }
148 
149 /*
150  * Return 1 if an internet address is for the local host and configured
151  * on one of its interfaces.
152  */
153 bool
in_localip(struct in_addr in)154 in_localip(struct in_addr in)
155 {
156 	struct in_ifaddr *ia;
157 
158 	NET_EPOCH_ASSERT();
159 
160 	CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
161 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
162 			return (true);
163 
164 	return (false);
165 }
166 
167 /*
168  * Like in_localip(), but FIB-aware and carp(4)-aware.
169  */
170 bool
in_localip_fib(struct in_addr in,uint16_t fib)171 in_localip_fib(struct in_addr in, uint16_t fib)
172 {
173 	struct in_ifaddr *ia;
174 
175 	NET_EPOCH_ASSERT();
176 
177 	CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
178 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
179 		    (ia->ia_ifa.ifa_carp == NULL ||
180 		    carp_master_p(&ia->ia_ifa)) &&
181 		    ia->ia_ifa.ifa_ifp->if_fib == fib)
182 			return (true);
183 
184 	return (false);
185 }
186 
187 /*
188  * Return 1 if an internet address is configured on an interface.
189  */
190 int
in_ifhasaddr(struct ifnet * ifp,struct in_addr in)191 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
192 {
193 	struct ifaddr *ifa;
194 	struct in_ifaddr *ia;
195 
196 	NET_EPOCH_ASSERT();
197 
198 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
199 		if (ifa->ifa_addr->sa_family != AF_INET)
200 			continue;
201 		ia = (struct in_ifaddr *)ifa;
202 		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
203 			return (1);
204 	}
205 
206 	return (0);
207 }
208 
209 /*
210  * Return a reference to the interface address which is different to
211  * the supplied one but with same IP address value.
212  */
213 static struct in_ifaddr *
in_localip_more(struct in_ifaddr * original_ia)214 in_localip_more(struct in_ifaddr *original_ia)
215 {
216 	struct epoch_tracker et;
217 	in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
218 	uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
219 	struct in_ifaddr *ia;
220 
221 	NET_EPOCH_ENTER(et);
222 	CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
223 		in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
224 		uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
225 		if (!V_rt_add_addr_allfibs && (original_fib != fib))
226 			continue;
227 		if ((original_ia != ia) && (original_addr == addr)) {
228 			ifa_ref(&ia->ia_ifa);
229 			NET_EPOCH_EXIT(et);
230 			return (ia);
231 		}
232 	}
233 	NET_EPOCH_EXIT(et);
234 
235 	return (NULL);
236 }
237 
238 /*
239  * Tries to find first IPv4 address in the provided fib.
240  * Prefers non-loopback addresses and return loopback IFF
241  * @loopback_ok is set.
242  *
243  * Returns ifa or NULL.
244  */
245 struct in_ifaddr *
in_findlocal(uint32_t fibnum,bool loopback_ok)246 in_findlocal(uint32_t fibnum, bool loopback_ok)
247 {
248 	struct in_ifaddr *ia = NULL, *ia_lo = NULL;
249 
250 	NET_EPOCH_ASSERT();
251 
252 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
253 		uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
254 		if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
255 			continue;
256 
257 		if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
258 			break;
259 		if (loopback_ok)
260 			ia_lo = ia;
261 	}
262 
263 	if (ia == NULL)
264 		ia = ia_lo;
265 
266 	return (ia);
267 }
268 
269 /*
270  * Determine whether an IP address is in a reserved set of addresses
271  * that may not be forwarded, or whether datagrams to that destination
272  * may be forwarded.
273  */
274 int
in_canforward(struct in_addr in)275 in_canforward(struct in_addr in)
276 {
277 	u_long i = ntohl(in.s_addr);
278 
279 	if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i))
280 		return (0);
281 	if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
282 		return (0);
283 	if (IN_ZERONET(i) && !V_ip_allow_net0)
284 		return (0);
285 	return (1);
286 }
287 
288 /*
289  * Sysctl to manage prefix of reserved loopback network; translate
290  * to/from mask.  The mask is always contiguous high-order 1 bits
291  * followed by all 0 bits.
292  */
293 static int
sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)294 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
295 {
296 	int error, preflen;
297 
298 	/* ffs is 1-based; compensate. */
299 	preflen = 33 - ffs(V_in_loopback_mask);
300 	error = sysctl_handle_int(oidp, &preflen, 0, req);
301 	if (error || !req->newptr)
302 		return (error);
303 	if (preflen < 8 || preflen > 31)
304 		return (EINVAL);
305 	V_in_loopback_mask = 0xffffffff << (32 - preflen);
306 	return (0);
307 }
308 
309 /*
310  * Trim a mask in a sockaddr
311  */
312 static void
in_socktrim(struct sockaddr_in * ap)313 in_socktrim(struct sockaddr_in *ap)
314 {
315     char *cplim = (char *) &ap->sin_addr;
316     char *cp = (char *) (&ap->sin_addr + 1);
317 
318     ap->sin_len = 0;
319     while (--cp >= cplim)
320 	if (*cp) {
321 	    (ap)->sin_len = cp - (char *) (ap) + 1;
322 	    break;
323 	}
324 }
325 
326 /*
327  * Generic internet control operations (ioctl's).
328  */
329 int
in_control_ioctl(u_long cmd,void * data,struct ifnet * ifp,struct ucred * cred)330 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp,
331     struct ucred *cred)
332 {
333 	struct ifreq *ifr = (struct ifreq *)data;
334 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
335 	struct epoch_tracker et;
336 	struct ifaddr *ifa;
337 	struct in_ifaddr *ia;
338 	int error;
339 
340 	if (ifp == NULL)
341 		return (EADDRNOTAVAIL);
342 
343 	/*
344 	 * Filter out 4 ioctls we implement directly.  Forward the rest
345 	 * to specific functions and ifp->if_ioctl().
346 	 */
347 	switch (cmd) {
348 	case SIOCGIFADDR:
349 	case SIOCGIFBRDADDR:
350 	case SIOCGIFDSTADDR:
351 	case SIOCGIFNETMASK:
352 		break;
353 	case SIOCGIFALIAS:
354 		sx_xlock(&in_control_sx);
355 		error = in_gifaddr_ioctl(cmd, data, ifp, cred);
356 		sx_xunlock(&in_control_sx);
357 		return (error);
358 	case SIOCDIFADDR:
359 		sx_xlock(&in_control_sx);
360 		error = in_difaddr_ioctl(cmd, data, ifp, cred);
361 		sx_xunlock(&in_control_sx);
362 		return (error);
363 	case OSIOCAIFADDR:	/* 9.x compat */
364 	case SIOCAIFADDR:
365 		sx_xlock(&in_control_sx);
366 		error = in_aifaddr_ioctl(cmd, data, ifp, cred);
367 		sx_xunlock(&in_control_sx);
368 		return (error);
369 	case SIOCSIFADDR:
370 	case SIOCSIFBRDADDR:
371 	case SIOCSIFDSTADDR:
372 	case SIOCSIFNETMASK:
373 		/* We no longer support that old commands. */
374 		return (EINVAL);
375 	default:
376 		if (ifp->if_ioctl == NULL)
377 			return (EOPNOTSUPP);
378 		return ((*ifp->if_ioctl)(ifp, cmd, data));
379 	}
380 
381 	if (addr->sin_addr.s_addr != INADDR_ANY &&
382 	    prison_check_ip4(cred, &addr->sin_addr) != 0)
383 		return (EADDRNOTAVAIL);
384 
385 	/*
386 	 * Find address for this interface, if it exists.  If an
387 	 * address was specified, find that one instead of the
388 	 * first one on the interface, if possible.
389 	 */
390 	NET_EPOCH_ENTER(et);
391 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
392 		if (ifa->ifa_addr->sa_family != AF_INET)
393 			continue;
394 		ia = (struct in_ifaddr *)ifa;
395 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
396 			break;
397 	}
398 	if (ifa == NULL)
399 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
400 			if (ifa->ifa_addr->sa_family == AF_INET) {
401 				ia = (struct in_ifaddr *)ifa;
402 				if (prison_check_ip4(cred,
403 				    &ia->ia_addr.sin_addr) == 0)
404 					break;
405 			}
406 
407 	if (ifa == NULL) {
408 		NET_EPOCH_EXIT(et);
409 		return (EADDRNOTAVAIL);
410 	}
411 
412 	error = 0;
413 	switch (cmd) {
414 	case SIOCGIFADDR:
415 		*addr = ia->ia_addr;
416 		break;
417 
418 	case SIOCGIFBRDADDR:
419 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
420 			error = EINVAL;
421 			break;
422 		}
423 		*addr = ia->ia_broadaddr;
424 		break;
425 
426 	case SIOCGIFDSTADDR:
427 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
428 			error = EINVAL;
429 			break;
430 		}
431 		*addr = ia->ia_dstaddr;
432 		break;
433 
434 	case SIOCGIFNETMASK:
435 		*addr = ia->ia_sockmask;
436 		break;
437 	}
438 
439 	NET_EPOCH_EXIT(et);
440 
441 	return (error);
442 }
443 
444 int
in_control(struct socket * so,u_long cmd,void * data,struct ifnet * ifp,struct thread * td)445 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
446     struct thread *td)
447 {
448 	return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
449 }
450 
451 static int
in_aifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)452 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
453 {
454 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
455 	const struct sockaddr_in *addr = &ifra->ifra_addr;
456 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
457 	const struct sockaddr_in *mask = &ifra->ifra_mask;
458 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
459 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
460 	struct epoch_tracker et;
461 	struct ifaddr *ifa;
462 	struct in_ifaddr *ia;
463 	bool iaIsFirst;
464 	int error = 0;
465 
466 	error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
467 	if (error)
468 		return (error);
469 
470 	/*
471 	 * ifra_addr must be present and be of INET family.
472 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
473 	 */
474 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
475 	    addr->sin_family != AF_INET)
476 		return (EINVAL);
477 	if (broadaddr->sin_len != 0 &&
478 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
479 	    broadaddr->sin_family != AF_INET))
480 		return (EINVAL);
481 	if (mask->sin_len != 0 &&
482 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
483 	    mask->sin_family != AF_INET))
484 		return (EINVAL);
485 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
486 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
487 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
488 		return (EDESTADDRREQ);
489 	if (vhid != 0 && carp_attach_p == NULL)
490 		return (EPROTONOSUPPORT);
491 
492 #ifdef MAC
493 	/* Check if a MAC policy disallows setting the IPv4 address. */
494 	error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp);
495 	if (error != 0)
496 		return (error);
497 #endif
498 
499 	/*
500 	 * See whether address already exist.
501 	 */
502 	iaIsFirst = true;
503 	ia = NULL;
504 	NET_EPOCH_ENTER(et);
505 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
506 		struct in_ifaddr *it;
507 
508 		if (ifa->ifa_addr->sa_family != AF_INET)
509 			continue;
510 
511 		it = (struct in_ifaddr *)ifa;
512 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
513 		    prison_check_ip4(cred, &addr->sin_addr) == 0)
514 			ia = it;
515 		else
516 			iaIsFirst = false;
517 	}
518 	NET_EPOCH_EXIT(et);
519 
520 	if (ia != NULL)
521 		(void )in_difaddr_ioctl(cmd, data, ifp, cred);
522 
523 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
524 	ia = (struct in_ifaddr *)ifa;
525 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
526 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
527 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
528 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
529 	    CALLOUT_RETURNUNLOCKED);
530 
531 	ia->ia_ifp = ifp;
532 	ia->ia_addr = *addr;
533 	if (mask->sin_len != 0) {
534 		ia->ia_sockmask = *mask;
535 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
536 	} else {
537 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
538 
539 		/*
540 	 	 * If netmask isn't supplied, use historical default.
541 		 * This is deprecated for interfaces other than loopback
542 		 * or point-to-point; warn in other cases.  In the future
543 		 * we should return an error rather than warning.
544 	 	 */
545 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
546 			printf("%s: set address: WARNING: network mask "
547 			     "should be specified; using historical default\n",
548 			     ifp->if_xname);
549 		if (IN_CLASSA(i))
550 			ia->ia_subnetmask = IN_CLASSA_NET;
551 		else if (IN_CLASSB(i))
552 			ia->ia_subnetmask = IN_CLASSB_NET;
553 		else
554 			ia->ia_subnetmask = IN_CLASSC_NET;
555 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
556 	}
557 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
558 	in_socktrim(&ia->ia_sockmask);
559 
560 	if (ifp->if_flags & IFF_BROADCAST) {
561 		if (broadaddr->sin_len != 0) {
562 			ia->ia_broadaddr = *broadaddr;
563 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
564 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
565 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
566 			ia->ia_broadaddr.sin_family = AF_INET;
567 		} else {
568 			ia->ia_broadaddr.sin_addr.s_addr =
569 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
570 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
571 			ia->ia_broadaddr.sin_family = AF_INET;
572 		}
573 	}
574 
575 	if (ifp->if_flags & IFF_POINTOPOINT)
576 		ia->ia_dstaddr = *dstaddr;
577 
578 	if (vhid != 0) {
579 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
580 		if (error)
581 			return (error);
582 	}
583 
584 	/* if_addrhead is already referenced by ifa_alloc() */
585 	IF_ADDR_WLOCK(ifp);
586 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
587 	IF_ADDR_WUNLOCK(ifp);
588 
589 	ifa_ref(ifa);			/* in_ifaddrhead */
590 	sx_assert(&in_control_sx, SA_XLOCKED);
591 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
592 	CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
593 	    ia_hash);
594 
595 	/*
596 	 * Give the interface a chance to initialize
597 	 * if this is its first address,
598 	 * and to validate the address if necessary.
599 	 */
600 	if (ifp->if_ioctl != NULL) {
601 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
602 		if (error)
603 			goto fail1;
604 	}
605 
606 	/*
607 	 * Add route for the network.
608 	 */
609 	if (vhid == 0) {
610 		error = in_addprefix(ia);
611 		if (error)
612 			goto fail1;
613 	}
614 
615 	/*
616 	 * Add a loopback route to self.
617 	 */
618 	if (vhid == 0 && ia_need_loopback_route(ia)) {
619 		struct in_ifaddr *eia;
620 
621 		eia = in_localip_more(ia);
622 
623 		if (eia == NULL) {
624 			error = ifa_add_loopback_route((struct ifaddr *)ia,
625 			    (struct sockaddr *)&ia->ia_addr);
626 			if (error)
627 				goto fail2;
628 		} else
629 			ifa_free(&eia->ia_ifa);
630 	}
631 
632 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
633 		struct in_addr allhosts_addr;
634 		struct in_ifinfo *ii;
635 
636 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
637 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
638 
639 		error = in_joingroup(ifp, &allhosts_addr, NULL,
640 			&ii->ii_allhosts);
641 	}
642 
643 	/*
644 	 * Note: we don't need extra reference for ifa, since we called
645 	 * with sx lock held, and ifaddr can not be deleted in concurrent
646 	 * thread.
647 	 */
648 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
649 
650 	return (error);
651 
652 fail2:
653 	if (vhid == 0)
654 		(void )in_scrubprefix(ia, LLE_STATIC);
655 
656 fail1:
657 	if (ia->ia_ifa.ifa_carp)
658 		(*carp_detach_p)(&ia->ia_ifa, false);
659 
660 	IF_ADDR_WLOCK(ifp);
661 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
662 	IF_ADDR_WUNLOCK(ifp);
663 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
664 
665 	sx_assert(&in_control_sx, SA_XLOCKED);
666 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
667 	CK_LIST_REMOVE(ia, ia_hash);
668 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
669 
670 	return (error);
671 }
672 
673 static int
in_difaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)674 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
675 {
676 	const struct ifreq *ifr = (struct ifreq *)data;
677 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
678 	    &ifr->ifr_addr;
679 	struct ifaddr *ifa;
680 	struct in_ifaddr *ia;
681 	bool deleteAny, iaIsLast;
682 	int error;
683 
684 	if (cred != NULL) {
685 		error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
686 		if (error)
687 			return (error);
688 	}
689 
690 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
691 	    addr->sin_family != AF_INET)
692 		deleteAny = true;
693 	else
694 		deleteAny = false;
695 
696 	iaIsLast = true;
697 	ia = NULL;
698 	IF_ADDR_WLOCK(ifp);
699 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
700 		struct in_ifaddr *it;
701 
702 		if (ifa->ifa_addr->sa_family != AF_INET)
703 			continue;
704 
705 		it = (struct in_ifaddr *)ifa;
706 		if (deleteAny && ia == NULL && (cred == NULL ||
707 		    prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
708 			ia = it;
709 
710 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
711 		    (cred == NULL || prison_check_ip4(cred,
712 		    &addr->sin_addr) == 0))
713 			ia = it;
714 
715 		if (it != ia)
716 			iaIsLast = false;
717 	}
718 
719 	if (ia == NULL) {
720 		IF_ADDR_WUNLOCK(ifp);
721 		return (EADDRNOTAVAIL);
722 	}
723 
724 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
725 	IF_ADDR_WUNLOCK(ifp);
726 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
727 
728 	sx_assert(&in_control_sx, SA_XLOCKED);
729 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
730 	CK_LIST_REMOVE(ia, ia_hash);
731 
732 	/*
733 	 * in_scrubprefix() kills the interface route.
734 	 */
735 	in_scrubprefix(ia, LLE_STATIC);
736 
737 	/*
738 	 * in_ifadown gets rid of all the rest of
739 	 * the routes.  This is not quite the right
740 	 * thing to do, but at least if we are running
741 	 * a routing process they will come back.
742 	 */
743 	in_ifadown(&ia->ia_ifa, 1);
744 
745 	if (ia->ia_ifa.ifa_carp)
746 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
747 
748 	/*
749 	 * If this is the last IPv4 address configured on this
750 	 * interface, leave the all-hosts group.
751 	 * No state-change report need be transmitted.
752 	 */
753 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
754 		struct in_ifinfo *ii;
755 
756 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
757 		if (ii->ii_allhosts) {
758 			(void)in_leavegroup(ii->ii_allhosts, NULL);
759 			ii->ii_allhosts = NULL;
760 		}
761 	}
762 
763 	IF_ADDR_WLOCK(ifp);
764 	if (callout_stop(&ia->ia_garp_timer) == 1) {
765 		ifa_free(&ia->ia_ifa);
766 	}
767 	IF_ADDR_WUNLOCK(ifp);
768 
769 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
770 	    IFADDR_EVENT_DEL);
771 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
772 
773 	return (0);
774 }
775 
776 static int
in_gifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)777 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
778 {
779 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
780 	const struct sockaddr_in *addr = &ifra->ifra_addr;
781 	struct epoch_tracker et;
782 	struct ifaddr *ifa;
783 	struct in_ifaddr *ia;
784 
785 	/*
786 	 * ifra_addr must be present and be of INET family.
787 	 */
788 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
789 	    addr->sin_family != AF_INET)
790 		return (EINVAL);
791 
792 	/*
793 	 * See whether address exist.
794 	 */
795 	ia = NULL;
796 	NET_EPOCH_ENTER(et);
797 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
798 		struct in_ifaddr *it;
799 
800 		if (ifa->ifa_addr->sa_family != AF_INET)
801 			continue;
802 
803 		it = (struct in_ifaddr *)ifa;
804 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
805 		    prison_check_ip4(cred, &addr->sin_addr) == 0) {
806 			ia = it;
807 			break;
808 		}
809 	}
810 	if (ia == NULL) {
811 		NET_EPOCH_EXIT(et);
812 		return (EADDRNOTAVAIL);
813 	}
814 
815 	ifra->ifra_mask = ia->ia_sockmask;
816 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
817 	    ia->ia_dstaddr.sin_family == AF_INET)
818 		ifra->ifra_dstaddr = ia->ia_dstaddr;
819 	else if ((ifp->if_flags & IFF_BROADCAST) &&
820 	    ia->ia_broadaddr.sin_family == AF_INET)
821 		ifra->ifra_broadaddr = ia->ia_broadaddr;
822 	else
823 		memset(&ifra->ifra_broadaddr, 0,
824 		    sizeof(ifra->ifra_broadaddr));
825 
826 	NET_EPOCH_EXIT(et);
827 	return (0);
828 }
829 
830 static int
in_match_ifaddr(const struct rtentry * rt,const struct nhop_object * nh,void * arg)831 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
832 {
833 
834 	if (nh->nh_ifa == (struct ifaddr *)arg)
835 		return (1);
836 
837 	return (0);
838 }
839 
840 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)841 in_handle_prefix_route(uint32_t fibnum, int cmd,
842     struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
843     struct ifnet *ifp)
844 {
845 
846 	NET_EPOCH_ASSERT();
847 
848 	/* Prepare gateway */
849 	struct sockaddr_dl_short sdl = {
850 		.sdl_family = AF_LINK,
851 		.sdl_len = sizeof(struct sockaddr_dl_short),
852 		.sdl_type = ifa->ifa_ifp->if_type,
853 		.sdl_index = ifa->ifa_ifp->if_index,
854 	};
855 
856 	struct rt_addrinfo info = {
857 		.rti_ifa = ifa,
858 		.rti_ifp = ifp,
859 		.rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
860 		.rti_info = {
861 			[RTAX_DST] = (struct sockaddr *)dst,
862 			[RTAX_NETMASK] = (struct sockaddr *)netmask,
863 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
864 		},
865 		/* Ensure we delete the prefix IFF prefix ifa matches */
866 		.rti_filter = in_match_ifaddr,
867 		.rti_filterdata = ifa,
868 	};
869 
870 	return (rib_handle_ifaddr_info(fibnum, cmd, &info));
871 }
872 
873 /*
874  * Routing table interaction with interface addresses.
875  *
876  * In general, two types of routes needs to be installed:
877  * a) "interface" or "prefix" route, telling user that the addresses
878  *   behind the ifa prefix are reached directly.
879  * b) "loopback" route installed for the ifa address, telling user that
880  *   the address belongs to local system.
881  *
882  * Handling for (a) and (b) differs in multi-fib aspects, hence they
883  *  are implemented in different functions below.
884  *
885  * The cases above may intersect - /32 interface aliases results in
886  *  the same prefix produced by (a) and (b). This blurs the definition
887  *  of the "loopback" route and complicate interactions. The interaction
888  *  table is defined below. The case numbers are used in the multiple
889  *  functions below to refer to the particular test case.
890  *
891  * There can be multiple options:
892  * 1) Adding address with prefix on non-p2p/non-loopback interface.
893  *  Example: 192.0.2.1/24. Action:
894  *  * add "prefix" route towards 192.0.2.0/24 via @ia interface,
895  *    using @ia as an address source.
896  *  * add "loopback" route towards 192.0.2.1 via V_loif, saving
897  *   @ia ifp in the gateway and using @ia as an address source.
898  *
899  * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
900  *  Example: 192.0.2.2/32. Action:
901  *  * add "prefix" host route via V_loif, using @ia as an address source.
902  *
903  * 3) Adding address with or without prefix to p2p interface.
904  *  Example: 10.0.0.1/24->10.0.0.2. Action:
905  *  * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
906  *    as an address source. Note: no sense in installing full /24 as the interface
907  *    is point-to-point.
908  *  * add "loopback" route towards 10.0.9.1 via V_loif, saving
909  *   @ia ifp in the gateway and using @ia as an address source.
910  *
911  * 4) Adding address with or without prefix to loopback interface.
912  *  Example: 192.0.2.1/24. Action:
913  *  * add "prefix" host route via @ia interface, using @ia as an address source.
914  *    Note: Skip installing /24 prefix as it would introduce TTL loop
915  *    for the traffic destined to these addresses.
916  */
917 
918 /*
919  * Checks if @ia needs to install loopback route to @ia address via
920  *  ifa_maintain_loopback_route().
921  *
922  * Return true on success.
923  */
924 static bool
ia_need_loopback_route(const struct in_ifaddr * ia)925 ia_need_loopback_route(const struct in_ifaddr *ia)
926 {
927 	struct ifnet *ifp = ia->ia_ifp;
928 
929 	/* Case 4: Skip loopback interfaces */
930 	if ((ifp->if_flags & IFF_LOOPBACK) ||
931 	    (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
932 		return (false);
933 
934 	/* Clash avoidance: Skip p2p interfaces with both addresses are equal */
935 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
936 	    ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
937 		return (false);
938 
939 	/* Case 2: skip /32 prefixes */
940 	if (!(ifp->if_flags & IFF_POINTOPOINT) &&
941 	    (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
942 		return (false);
943 
944 	return (true);
945 }
946 
947 /*
948  * Calculate "prefix" route corresponding to @ia.
949  */
950 static void
ia_getrtprefix(const struct in_ifaddr * ia,struct in_addr * prefix,struct in_addr * mask)951 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
952 {
953 
954 	if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
955 		/* Case 3: return host route for dstaddr */
956 		*prefix = ia->ia_dstaddr.sin_addr;
957 		mask->s_addr = INADDR_BROADCAST;
958 	} else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
959 		/* Case 4: return host route for ifaddr */
960 		*prefix = ia->ia_addr.sin_addr;
961 		mask->s_addr = INADDR_BROADCAST;
962 	} else {
963 		/* Cases 1,2: return actual ia prefix */
964 		*prefix = ia->ia_addr.sin_addr;
965 		*mask = ia->ia_sockmask.sin_addr;
966 		prefix->s_addr &= mask->s_addr;
967 	}
968 }
969 
970 /*
971  * Adds or delete interface "prefix" route corresponding to @ifa.
972  *  Returns 0 on success or errno.
973  */
974 static int
in_handle_ifaddr_route(int cmd,struct in_ifaddr * ia)975 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
976 {
977 	struct ifaddr *ifa = &ia->ia_ifa;
978 	struct in_addr daddr, maddr;
979 	struct sockaddr_in *pmask;
980 	struct epoch_tracker et;
981 	int error;
982 
983 	ia_getrtprefix(ia, &daddr, &maddr);
984 
985 	struct sockaddr_in mask = {
986 		.sin_family = AF_INET,
987 		.sin_len = sizeof(struct sockaddr_in),
988 		.sin_addr = maddr,
989 	};
990 
991 	pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
992 
993 	struct sockaddr_in dst = {
994 		.sin_family = AF_INET,
995 		.sin_len = sizeof(struct sockaddr_in),
996 		.sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
997 	};
998 
999 	struct ifnet *ifp = ia->ia_ifp;
1000 
1001 	if ((maddr.s_addr == INADDR_BROADCAST) &&
1002 	    (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
1003 		/* Case 2: host route on broadcast interface */
1004 		ifp = V_loif;
1005 	}
1006 
1007 	uint32_t fibnum = ifa->ifa_ifp->if_fib;
1008 	NET_EPOCH_ENTER(et);
1009 	error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
1010 	NET_EPOCH_EXIT(et);
1011 
1012 	return (error);
1013 }
1014 
1015 /*
1016  * Check if we have a route for the given prefix already.
1017  */
1018 static bool
in_hasrtprefix(struct in_ifaddr * target)1019 in_hasrtprefix(struct in_ifaddr *target)
1020 {
1021 	struct epoch_tracker et;
1022 	struct in_ifaddr *ia;
1023 	struct in_addr prefix, mask, p, m;
1024 	bool result = false;
1025 
1026 	ia_getrtprefix(target, &prefix, &mask);
1027 
1028 	/* Look for an existing address with the same prefix, mask, and fib */
1029 	NET_EPOCH_ENTER(et);
1030 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1031 		ia_getrtprefix(ia, &p, &m);
1032 
1033 		if (prefix.s_addr != p.s_addr ||
1034 		    mask.s_addr != m.s_addr)
1035 			continue;
1036 
1037 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1038 			continue;
1039 
1040 		/*
1041 		 * If we got a matching prefix route inserted by other
1042 		 * interface address, we are done here.
1043 		 */
1044 		if (ia->ia_flags & IFA_ROUTE) {
1045 			result = true;
1046 			break;
1047 		}
1048 	}
1049 	NET_EPOCH_EXIT(et);
1050 
1051 	return (result);
1052 }
1053 
1054 int
in_addprefix(struct in_ifaddr * target)1055 in_addprefix(struct in_ifaddr *target)
1056 {
1057 	int error;
1058 
1059 	if (in_hasrtprefix(target)) {
1060 		if (V_nosameprefix)
1061 			return (EEXIST);
1062 		else {
1063 			rt_addrmsg(RTM_ADD, &target->ia_ifa,
1064 			    target->ia_ifp->if_fib);
1065 			return (0);
1066 		}
1067 	}
1068 
1069 	/*
1070 	 * No-one seem to have this prefix route, so we try to insert it.
1071 	 */
1072 	rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1073 	error = in_handle_ifaddr_route(RTM_ADD, target);
1074 	if (!error)
1075 		target->ia_flags |= IFA_ROUTE;
1076 	return (error);
1077 }
1078 
1079 /*
1080  * Removes either all lle entries for given @ia, or lle
1081  * corresponding to @ia address.
1082  */
1083 static void
in_scrubprefixlle(struct in_ifaddr * ia,int all,u_int flags)1084 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1085 {
1086 	struct sockaddr_in addr, mask;
1087 	struct sockaddr *saddr, *smask;
1088 	struct ifnet *ifp;
1089 
1090 	saddr = (struct sockaddr *)&addr;
1091 	bzero(&addr, sizeof(addr));
1092 	addr.sin_len = sizeof(addr);
1093 	addr.sin_family = AF_INET;
1094 	smask = (struct sockaddr *)&mask;
1095 	bzero(&mask, sizeof(mask));
1096 	mask.sin_len = sizeof(mask);
1097 	mask.sin_family = AF_INET;
1098 	mask.sin_addr.s_addr = ia->ia_subnetmask;
1099 	ifp = ia->ia_ifp;
1100 
1101 	if (all) {
1102 		/*
1103 		 * Remove all L2 entries matching given prefix.
1104 		 * Convert address to host representation to avoid
1105 		 * doing this on every callback. ia_subnetmask is already
1106 		 * stored in host representation.
1107 		 */
1108 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1109 		lltable_prefix_free(AF_INET, saddr, smask, flags);
1110 	} else {
1111 		/* Remove interface address only */
1112 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1113 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1114 	}
1115 }
1116 
1117 /*
1118  * If there is no other address in the system that can serve a route to the
1119  * same prefix, remove the route.  Hand over the route to the new address
1120  * otherwise.
1121  */
1122 int
in_scrubprefix(struct in_ifaddr * target,u_int flags)1123 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1124 {
1125 	struct epoch_tracker et;
1126 	struct in_ifaddr *ia;
1127 	struct in_addr prefix, mask, p, m;
1128 	int error = 0;
1129 
1130 	/*
1131 	 * Remove the loopback route to the interface address.
1132 	 */
1133 	if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1134 		struct in_ifaddr *eia;
1135 
1136 		eia = in_localip_more(target);
1137 
1138 		if (eia != NULL) {
1139 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
1140 			    (struct sockaddr *)&target->ia_addr);
1141 			ifa_free(&eia->ia_ifa);
1142 		} else {
1143 			error = ifa_del_loopback_route((struct ifaddr *)target,
1144 			    (struct sockaddr *)&target->ia_addr);
1145 		}
1146 	}
1147 
1148 	ia_getrtprefix(target, &prefix, &mask);
1149 
1150 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1151 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1152 
1153 		/*
1154 		 * Removing address from !IFF_UP interface or
1155 		 * prefix which exists on other interface (along with route).
1156 		 * No entries should exist here except target addr.
1157 		 * Given that, delete this entry only.
1158 		 */
1159 		in_scrubprefixlle(target, 0, flags);
1160 		return (0);
1161 	}
1162 
1163 	NET_EPOCH_ENTER(et);
1164 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1165 		ia_getrtprefix(ia, &p, &m);
1166 
1167 		if (prefix.s_addr != p.s_addr ||
1168 		    mask.s_addr != m.s_addr)
1169 			continue;
1170 
1171 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1172 			continue;
1173 
1174 		/*
1175 		 * If we got a matching prefix address, move IFA_ROUTE and
1176 		 * the route itself to it.  Make sure that routing daemons
1177 		 * get a heads-up.
1178 		 */
1179 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1180 			ifa_ref(&ia->ia_ifa);
1181 			NET_EPOCH_EXIT(et);
1182 			error = in_handle_ifaddr_route(RTM_DELETE, target);
1183 			if (error == 0)
1184 				target->ia_flags &= ~IFA_ROUTE;
1185 			else
1186 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1187 					error);
1188 			/* Scrub all entries IFF interface is different */
1189 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1190 			    flags);
1191 			error = in_handle_ifaddr_route(RTM_ADD, ia);
1192 			if (error == 0)
1193 				ia->ia_flags |= IFA_ROUTE;
1194 			else
1195 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1196 					error);
1197 			ifa_free(&ia->ia_ifa);
1198 			return (error);
1199 		}
1200 	}
1201 	NET_EPOCH_EXIT(et);
1202 
1203 	/*
1204 	 * remove all L2 entries on the given prefix
1205 	 */
1206 	in_scrubprefixlle(target, 1, flags);
1207 
1208 	/*
1209 	 * As no-one seem to have this prefix, we can remove the route.
1210 	 */
1211 	rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
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, prefix delete failed\n", error);
1217 	return (error);
1218 }
1219 
1220 void
in_ifscrub_all(void)1221 in_ifscrub_all(void)
1222 {
1223 	struct ifnet *ifp;
1224 	struct ifaddr *ifa, *nifa;
1225 	struct ifreq ifr;
1226 
1227 	IFNET_RLOCK();
1228 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1229 		/* Cannot lock here - lock recursion. */
1230 		/* NET_EPOCH_ENTER(et); */
1231 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1232 			if (ifa->ifa_addr->sa_family != AF_INET)
1233 				continue;
1234 
1235 			/*
1236 			 * This is ugly but the only way for legacy IP to
1237 			 * cleanly remove addresses and everything attached.
1238 			 */
1239 			bzero(&ifr, sizeof(ifr));
1240 			ifr.ifr_addr = *ifa->ifa_addr;
1241 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1242 			    ifp, NULL);
1243 		}
1244 		/* NET_EPOCH_EXIT(et); */
1245 		in_purgemaddrs(ifp);
1246 		igmp_domifdetach(ifp);
1247 	}
1248 	IFNET_RUNLOCK();
1249 }
1250 
1251 int
in_ifaddr_broadcast(struct in_addr in,struct in_ifaddr * ia)1252 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1253 {
1254 
1255 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1256 	     /*
1257 	      * Optionally check for old-style (host 0) broadcast, but
1258 	      * taking into account that RFC 3021 obsoletes it.
1259 	      */
1260 	    (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK &&
1261 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
1262 	     /*
1263 	      * Check for an all one subnetmask. These
1264 	      * only exist when an interface gets a secondary
1265 	      * address.
1266 	      */
1267 	    ia->ia_subnetmask != (u_long)0xffffffff);
1268 }
1269 
1270 /*
1271  * Return 1 if the address might be a local broadcast address.
1272  */
1273 int
in_broadcast(struct in_addr in,struct ifnet * ifp)1274 in_broadcast(struct in_addr in, struct ifnet *ifp)
1275 {
1276 	struct ifaddr *ifa;
1277 	int found;
1278 
1279 	NET_EPOCH_ASSERT();
1280 
1281 	if (in.s_addr == INADDR_BROADCAST ||
1282 	    in.s_addr == INADDR_ANY)
1283 		return (1);
1284 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1285 		return (0);
1286 	found = 0;
1287 	/*
1288 	 * Look through the list of addresses for a match
1289 	 * with a broadcast address.
1290 	 */
1291 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1292 		if (ifa->ifa_addr->sa_family == AF_INET &&
1293 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
1294 			found = 1;
1295 			break;
1296 		}
1297 	return (found);
1298 }
1299 
1300 /*
1301  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1302  */
1303 void
in_ifdetach(struct ifnet * ifp)1304 in_ifdetach(struct ifnet *ifp)
1305 {
1306 	IN_MULTI_LOCK();
1307 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1308 	in_pcbpurgeif0(&V_udbinfo, ifp);
1309 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1310 	in_purgemaddrs(ifp);
1311 	IN_MULTI_UNLOCK();
1312 
1313 	/*
1314 	 * Make sure all multicast deletions invoking if_ioctl() are
1315 	 * completed before returning. Else we risk accessing a freed
1316 	 * ifnet structure pointer.
1317 	 */
1318 	inm_release_wait(NULL);
1319 }
1320 
1321 static void
in_ifnet_event(void * arg __unused,struct ifnet * ifp,int event)1322 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1323 {
1324 	struct epoch_tracker et;
1325 	struct ifaddr *ifa;
1326 	struct in_ifaddr *ia;
1327 	int error;
1328 
1329 	NET_EPOCH_ENTER(et);
1330 	switch (event) {
1331 	case IFNET_EVENT_DOWN:
1332 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1333 			if (ifa->ifa_addr->sa_family != AF_INET)
1334 				continue;
1335 			ia = (struct in_ifaddr *)ifa;
1336 			if ((ia->ia_flags & IFA_ROUTE) == 0)
1337 				continue;
1338 			ifa_ref(ifa);
1339 			/*
1340 			 * in_scrubprefix() kills the interface route.
1341 			 */
1342 			in_scrubprefix(ia, 0);
1343 			/*
1344 			 * in_ifadown gets rid of all the rest of the
1345 			 * routes.  This is not quite the right thing
1346 			 * to do, but at least if we are running a
1347 			 * routing process they will come back.
1348 			 */
1349 			in_ifadown(ifa, 0);
1350 			ifa_free(ifa);
1351 		}
1352 		break;
1353 
1354 	case IFNET_EVENT_UP:
1355 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1356 			if (ifa->ifa_addr->sa_family != AF_INET)
1357 				continue;
1358 			ia = (struct in_ifaddr *)ifa;
1359 			if (ia->ia_flags & IFA_ROUTE)
1360 				continue;
1361 			ifa_ref(ifa);
1362 			error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1363 			rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1364 			error = in_handle_ifaddr_route(RTM_ADD, ia);
1365 			if (error == 0)
1366 				ia->ia_flags |= IFA_ROUTE;
1367 			error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1368 			ifa_free(ifa);
1369 		}
1370 		break;
1371 	}
1372 	NET_EPOCH_EXIT(et);
1373 }
1374 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1375 
1376 /*
1377  * Delete all IPv4 multicast address records, and associated link-layer
1378  * multicast address records, associated with ifp.
1379  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1380  * XXX This should not race with ifma_protospec being set during
1381  * a new allocation, if it does, we have bigger problems.
1382  */
1383 static void
in_purgemaddrs(struct ifnet * ifp)1384 in_purgemaddrs(struct ifnet *ifp)
1385 {
1386 	struct epoch_tracker	 et;
1387 	struct in_multi_head purgeinms;
1388 	struct in_multi		*inm;
1389 	struct ifmultiaddr	*ifma;
1390 
1391 	SLIST_INIT(&purgeinms);
1392 	IN_MULTI_LIST_LOCK();
1393 
1394 	/*
1395 	 * Extract list of in_multi associated with the detaching ifp
1396 	 * which the PF_INET layer is about to release.
1397 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1398 	 * by code further down.
1399 	 */
1400 	IF_ADDR_WLOCK(ifp);
1401 	NET_EPOCH_ENTER(et);
1402 	CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1403 		inm = inm_ifmultiaddr_get_inm(ifma);
1404 		if (inm == NULL)
1405 			continue;
1406 		inm_rele_locked(&purgeinms, inm);
1407 	}
1408 	NET_EPOCH_EXIT(et);
1409 	IF_ADDR_WUNLOCK(ifp);
1410 
1411 	inm_release_list_deferred(&purgeinms);
1412 	igmp_ifdetach(ifp);
1413 	IN_MULTI_LIST_UNLOCK();
1414 }
1415 
1416 struct in_llentry {
1417 	struct llentry		base;
1418 };
1419 
1420 #define	IN_LLTBL_DEFAULT_HSIZE	32
1421 #define	IN_LLTBL_HASH(k, h) \
1422 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1423 
1424 /*
1425  * Do actual deallocation of @lle.
1426  */
1427 static void
in_lltable_destroy_lle_unlocked(epoch_context_t ctx)1428 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1429 {
1430 	struct llentry *lle;
1431 
1432 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1433 	LLE_LOCK_DESTROY(lle);
1434 	LLE_REQ_DESTROY(lle);
1435 	free(lle, M_LLTABLE);
1436 }
1437 
1438 /*
1439  * Called by LLE_FREE_LOCKED when number of references
1440  * drops to zero.
1441  */
1442 static void
in_lltable_destroy_lle(struct llentry * lle)1443 in_lltable_destroy_lle(struct llentry *lle)
1444 {
1445 
1446 	LLE_WUNLOCK(lle);
1447 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1448 }
1449 
1450 static struct llentry *
in_lltable_new(struct in_addr addr4,u_int flags)1451 in_lltable_new(struct in_addr addr4, u_int flags)
1452 {
1453 	struct in_llentry *lle;
1454 
1455 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1456 	if (lle == NULL)		/* NB: caller generates msg */
1457 		return NULL;
1458 
1459 	/*
1460 	 * For IPv4 this will trigger "arpresolve" to generate
1461 	 * an ARP request.
1462 	 */
1463 	lle->base.la_expire = time_uptime; /* mark expired */
1464 	lle->base.r_l3addr.addr4 = addr4;
1465 	lle->base.lle_refcnt = 1;
1466 	lle->base.lle_free = in_lltable_destroy_lle;
1467 	LLE_LOCK_INIT(&lle->base);
1468 	LLE_REQ_INIT(&lle->base);
1469 	callout_init(&lle->base.lle_timer, 1);
1470 
1471 	return (&lle->base);
1472 }
1473 
1474 static int
in_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)1475 in_lltable_match_prefix(const struct sockaddr *saddr,
1476     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1477 {
1478 	struct in_addr addr, mask, lle_addr;
1479 
1480 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1481 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1482 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1483 
1484 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1485 		return (0);
1486 
1487 	if (lle->la_flags & LLE_IFADDR) {
1488 		/*
1489 		 * Delete LLE_IFADDR records IFF address & flag matches.
1490 		 * Note that addr is the interface address within prefix
1491 		 * being matched.
1492 		 * Note also we should handle 'ifdown' cases without removing
1493 		 * ifaddr macs.
1494 		 */
1495 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1496 			return (1);
1497 		return (0);
1498 	}
1499 
1500 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1501 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1502 		return (1);
1503 
1504 	return (0);
1505 }
1506 
1507 static void
in_lltable_free_entry(struct lltable * llt,struct llentry * lle)1508 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1509 {
1510 	size_t pkts_dropped;
1511 
1512 	LLE_WLOCK_ASSERT(lle);
1513 	KASSERT(llt != NULL, ("lltable is NULL"));
1514 
1515 	/* Unlink entry from table if not already */
1516 	if ((lle->la_flags & LLE_LINKED) != 0) {
1517 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1518 		lltable_unlink_entry(llt, lle);
1519 	}
1520 
1521 	/* Drop hold queue */
1522 	pkts_dropped = llentry_free(lle);
1523 	ARPSTAT_ADD(dropped, pkts_dropped);
1524 }
1525 
1526 static int
in_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)1527 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1528 {
1529 	struct nhop_object *nh;
1530 	struct in_addr addr;
1531 
1532 	KASSERT(l3addr->sa_family == AF_INET,
1533 	    ("sin_family %d", l3addr->sa_family));
1534 
1535 	addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1536 
1537 	nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1538 	if (nh == NULL)
1539 		return (EINVAL);
1540 
1541 	/*
1542 	 * If the gateway for an existing host route matches the target L3
1543 	 * address, which is a special route inserted by some implementation
1544 	 * such as MANET, and the interface is of the correct type, then
1545 	 * allow for ARP to proceed.
1546 	 */
1547 	if (nh->nh_flags & NHF_GATEWAY) {
1548 		if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1549 		    (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1550 		    memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1551 		    sizeof(in_addr_t)) != 0) {
1552 			return (EINVAL);
1553 		}
1554 	}
1555 
1556 	/*
1557 	 * Make sure that at least the destination address is covered
1558 	 * by the route. This is for handling the case where 2 or more
1559 	 * interfaces have the same prefix. An incoming packet arrives
1560 	 * on one interface and the corresponding outgoing packet leaves
1561 	 * another interface.
1562 	 */
1563 	if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1564 		struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1565 		struct in_addr dst_addr, mask_addr;
1566 
1567 		if (ia == NULL)
1568 			return (EINVAL);
1569 
1570 		/*
1571 		 * ifaof_ifpforaddr() returns _best matching_ IFA.
1572 		 * It is possible that ifa prefix does not cover our address.
1573 		 * Explicitly verify and fail if that's the case.
1574 		 */
1575 		dst_addr = IA_SIN(ia)->sin_addr;
1576 		mask_addr.s_addr = htonl(ia->ia_subnetmask);
1577 
1578 		if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1579 			return (EINVAL);
1580 	}
1581 
1582 	return (0);
1583 }
1584 
1585 static inline uint32_t
in_lltable_hash_dst(const struct in_addr dst,uint32_t hsize)1586 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1587 {
1588 
1589 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1590 }
1591 
1592 static uint32_t
in_lltable_hash(const struct llentry * lle,uint32_t hsize)1593 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1594 {
1595 
1596 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1597 }
1598 
1599 static void
in_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)1600 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1601 {
1602 	struct sockaddr_in *sin;
1603 
1604 	sin = (struct sockaddr_in *)sa;
1605 	bzero(sin, sizeof(*sin));
1606 	sin->sin_family = AF_INET;
1607 	sin->sin_len = sizeof(*sin);
1608 	sin->sin_addr = lle->r_l3addr.addr4;
1609 }
1610 
1611 static inline struct llentry *
in_lltable_find_dst(struct lltable * llt,struct in_addr dst)1612 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1613 {
1614 	struct llentry *lle;
1615 	struct llentries *lleh;
1616 	u_int hashidx;
1617 
1618 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1619 	lleh = &llt->lle_head[hashidx];
1620 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1621 		if (lle->la_flags & LLE_DELETED)
1622 			continue;
1623 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1624 			break;
1625 	}
1626 
1627 	return (lle);
1628 }
1629 
1630 static void
in_lltable_delete_entry(struct lltable * llt,struct llentry * lle)1631 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1632 {
1633 
1634 	lle->la_flags |= LLE_DELETED;
1635 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1636 #ifdef DIAGNOSTIC
1637 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1638 #endif
1639 	llentry_free(lle);
1640 }
1641 
1642 static struct llentry *
in_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1643 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1644 {
1645 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1646 	struct ifnet *ifp = llt->llt_ifp;
1647 	struct llentry *lle;
1648 	char linkhdr[LLE_MAX_LINKHDR];
1649 	size_t linkhdrsize;
1650 	int lladdr_off;
1651 
1652 	KASSERT(l3addr->sa_family == AF_INET,
1653 	    ("sin_family %d", l3addr->sa_family));
1654 
1655 	/*
1656 	 * A route that covers the given address must have
1657 	 * been installed 1st because we are doing a resolution,
1658 	 * verify this.
1659 	 */
1660 	if (!(flags & LLE_IFADDR) &&
1661 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1662 		return (NULL);
1663 
1664 	lle = in_lltable_new(sin->sin_addr, flags);
1665 	if (lle == NULL) {
1666 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1667 		return (NULL);
1668 	}
1669 	lle->la_flags = flags;
1670 	if (flags & LLE_STATIC)
1671 		lle->r_flags |= RLLE_VALID;
1672 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1673 		linkhdrsize = LLE_MAX_LINKHDR;
1674 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1675 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1676 			in_lltable_free_entry(llt, lle);
1677 			return (NULL);
1678 		}
1679 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1680 		    lladdr_off);
1681 		lle->la_flags |= LLE_STATIC;
1682 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1683 		lle->la_expire = 0;
1684 	}
1685 
1686 	return (lle);
1687 }
1688 
1689 /*
1690  * Return NULL if not found or marked for deletion.
1691  * If found return lle read locked.
1692  */
1693 static struct llentry *
in_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1694 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1695 {
1696 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1697 	struct llentry *lle;
1698 
1699 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1700 	KASSERT(l3addr->sa_family == AF_INET,
1701 	    ("sin_family %d", l3addr->sa_family));
1702 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1703 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1704 	    ("wrong lle request flags: %#x", flags));
1705 
1706 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1707 	if (lle == NULL)
1708 		return (NULL);
1709 	if (flags & LLE_UNLOCKED)
1710 		return (lle);
1711 
1712 	if (flags & LLE_EXCLUSIVE)
1713 		LLE_WLOCK(lle);
1714 	else
1715 		LLE_RLOCK(lle);
1716 
1717 	/*
1718 	 * If the afdata lock is not held, the LLE may have been unlinked while
1719 	 * we were blocked on the LLE lock.  Check for this case.
1720 	 */
1721 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1722 		if (flags & LLE_EXCLUSIVE)
1723 			LLE_WUNLOCK(lle);
1724 		else
1725 			LLE_RUNLOCK(lle);
1726 		return (NULL);
1727 	}
1728 	return (lle);
1729 }
1730 
1731 static int
in_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)1732 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1733     struct sysctl_req *wr)
1734 {
1735 	struct ifnet *ifp = llt->llt_ifp;
1736 	/* XXX stack use */
1737 	struct {
1738 		struct rt_msghdr	rtm;
1739 		struct sockaddr_in	sin;
1740 		struct sockaddr_dl	sdl;
1741 	} arpc;
1742 	struct sockaddr_dl *sdl;
1743 	int error;
1744 
1745 	bzero(&arpc, sizeof(arpc));
1746 	/* skip deleted entries */
1747 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1748 		return (0);
1749 	/* Skip if jailed and not a valid IP of the prison. */
1750 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1751 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1752 		return (0);
1753 	/*
1754 	 * produce a msg made of:
1755 	 *  struct rt_msghdr;
1756 	 *  struct sockaddr_in; (IPv4)
1757 	 *  struct sockaddr_dl;
1758 	 */
1759 	arpc.rtm.rtm_msglen = sizeof(arpc);
1760 	arpc.rtm.rtm_version = RTM_VERSION;
1761 	arpc.rtm.rtm_type = RTM_GET;
1762 	arpc.rtm.rtm_flags = RTF_UP;
1763 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1764 
1765 	/* publish */
1766 	if (lle->la_flags & LLE_PUB)
1767 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1768 
1769 	sdl = &arpc.sdl;
1770 	sdl->sdl_family = AF_LINK;
1771 	sdl->sdl_len = sizeof(*sdl);
1772 	sdl->sdl_index = ifp->if_index;
1773 	sdl->sdl_type = ifp->if_type;
1774 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1775 		sdl->sdl_alen = ifp->if_addrlen;
1776 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1777 	} else {
1778 		sdl->sdl_alen = 0;
1779 		bzero(LLADDR(sdl), ifp->if_addrlen);
1780 	}
1781 
1782 	arpc.rtm.rtm_rmx.rmx_expire =
1783 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1784 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1785 	if (lle->la_flags & LLE_STATIC)
1786 		arpc.rtm.rtm_flags |= RTF_STATIC;
1787 	if (lle->la_flags & LLE_IFADDR)
1788 		arpc.rtm.rtm_flags |= RTF_PINNED;
1789 	arpc.rtm.rtm_index = ifp->if_index;
1790 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1791 
1792 	return (error);
1793 }
1794 
1795 static void
in_lltable_post_resolved(struct lltable * llt,struct llentry * lle)1796 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1797 {
1798 	struct ifnet *ifp = llt->llt_ifp;
1799 
1800 	/* gratuitous ARP */
1801 	if ((lle->la_flags & LLE_PUB) != 0)
1802 		arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1803 		    lle->ll_addr);
1804 }
1805 
1806 static struct lltable *
in_lltattach(struct ifnet * ifp)1807 in_lltattach(struct ifnet *ifp)
1808 {
1809 	struct lltable *llt;
1810 
1811 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1812  	llt->llt_af = AF_INET;
1813  	llt->llt_ifp = ifp;
1814 
1815 	llt->llt_lookup = in_lltable_lookup;
1816 	llt->llt_alloc_entry = in_lltable_alloc;
1817 	llt->llt_delete_entry = in_lltable_delete_entry;
1818 	llt->llt_dump_entry = in_lltable_dump_entry;
1819 	llt->llt_hash = in_lltable_hash;
1820 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1821 	llt->llt_free_entry = in_lltable_free_entry;
1822 	llt->llt_match_prefix = in_lltable_match_prefix;
1823 	llt->llt_mark_used = llentry_mark_used;
1824 	llt->llt_post_resolved = in_lltable_post_resolved;
1825  	lltable_link(llt);
1826 
1827 	return (llt);
1828 }
1829 
1830 struct lltable *
in_lltable_get(struct ifnet * ifp)1831 in_lltable_get(struct ifnet *ifp)
1832 {
1833 	struct lltable *llt = NULL;
1834 
1835 	void *afdata_ptr = ifp->if_afdata[AF_INET];
1836 	if (afdata_ptr != NULL)
1837 		llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1838 	return (llt);
1839 }
1840 
1841 void *
in_domifattach(struct ifnet * ifp)1842 in_domifattach(struct ifnet *ifp)
1843 {
1844 	struct in_ifinfo *ii;
1845 
1846 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1847 
1848 	ii->ii_llt = in_lltattach(ifp);
1849 	ii->ii_igmp = igmp_domifattach(ifp);
1850 
1851 	return (ii);
1852 }
1853 
1854 void
in_domifdetach(struct ifnet * ifp,void * aux)1855 in_domifdetach(struct ifnet *ifp, void *aux)
1856 {
1857 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1858 
1859 	igmp_domifdetach(ifp);
1860 	lltable_free(ii->ii_llt);
1861 	free(ii, M_IFADDR);
1862 }
1863