xref: /freebsd/sys/netinet/ip_options.c (revision 640235e2c2ba32947f7c59d168437ffa1280f1e6)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *      The Regents of the University of California.
4  * Copyright (c) 2005 Andre Oppermann, Internet Business Solutions AG.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 4. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_ipstealth.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/domain.h>
41 #include <sys/protosw.h>
42 #include <sys/socket.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/syslog.h>
46 #include <sys/sysctl.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/if_var.h>
51 #include <net/if_dl.h>
52 #include <net/route.h>
53 #include <net/netisr.h>
54 #include <net/vnet.h>
55 
56 #include <netinet/in.h>
57 #include <netinet/in_fib.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_options.h>
64 #include <netinet/ip_icmp.h>
65 #include <machine/in_cksum.h>
66 
67 #include <sys/socketvar.h>
68 
69 static VNET_DEFINE(int, ip_dosourceroute);
70 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute,
71     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_dosourceroute), 0,
72     "Enable forwarding source routed IP packets");
73 #define	V_ip_dosourceroute	VNET(ip_dosourceroute)
74 
75 static VNET_DEFINE(int,	ip_acceptsourceroute);
76 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
77     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_acceptsourceroute), 0,
78     "Enable accepting source routed IP packets");
79 #define	V_ip_acceptsourceroute	VNET(ip_acceptsourceroute)
80 
81 VNET_DEFINE(int, ip_doopts) = 1; /* 0 = ignore, 1 = process, 2 = reject */
82 SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_VNET | CTLFLAG_RW,
83     &VNET_NAME(ip_doopts), 0, "Enable IP options processing ([LS]SRR, RR, TS)");
84 
85 static void	save_rte(struct mbuf *m, u_char *, struct in_addr);
86 
87 /*
88  * Do option processing on a datagram, possibly discarding it if bad options
89  * are encountered, or forwarding it if source-routed.
90  *
91  * The pass argument is used when operating in the IPSTEALTH mode to tell
92  * what options to process: [LS]SRR (pass 0) or the others (pass 1).  The
93  * reason for as many as two passes is that when doing IPSTEALTH, non-routing
94  * options should be processed only if the packet is for us.
95  *
96  * Returns 1 if packet has been forwarded/freed, 0 if the packet should be
97  * processed further.
98  */
99 int
100 ip_dooptions(struct mbuf *m, int pass)
101 {
102 	struct ip *ip = mtod(m, struct ip *);
103 	u_char *cp;
104 	struct in_ifaddr *ia;
105 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
106 	struct in_addr *sin, dst;
107 	uint32_t ntime;
108 	struct nhop4_extended nh_ext;
109 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
110 
111 	/* Ignore or reject packets with IP options. */
112 	if (V_ip_doopts == 0)
113 		return 0;
114 	else if (V_ip_doopts == 2) {
115 		type = ICMP_UNREACH;
116 		code = ICMP_UNREACH_FILTER_PROHIB;
117 		goto bad;
118 	}
119 
120 	dst = ip->ip_dst;
121 	cp = (u_char *)(ip + 1);
122 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
123 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
124 		opt = cp[IPOPT_OPTVAL];
125 		if (opt == IPOPT_EOL)
126 			break;
127 		if (opt == IPOPT_NOP)
128 			optlen = 1;
129 		else {
130 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
131 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
132 				goto bad;
133 			}
134 			optlen = cp[IPOPT_OLEN];
135 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
136 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
137 				goto bad;
138 			}
139 		}
140 		switch (opt) {
141 
142 		default:
143 			break;
144 
145 		/*
146 		 * Source routing with record.  Find interface with current
147 		 * destination address.  If none on this machine then drop if
148 		 * strictly routed, or do nothing if loosely routed.  Record
149 		 * interface address and bring up next address component.  If
150 		 * strictly routed make sure next address is on directly
151 		 * accessible net.
152 		 */
153 		case IPOPT_LSRR:
154 		case IPOPT_SSRR:
155 #ifdef IPSTEALTH
156 			if (V_ipstealth && pass > 0)
157 				break;
158 #endif
159 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
160 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
161 				goto bad;
162 			}
163 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
164 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
165 				goto bad;
166 			}
167 			ipaddr.sin_addr = ip->ip_dst;
168 			if (ifa_ifwithaddr_check((struct sockaddr *)&ipaddr)
169 			    == 0) {
170 				if (opt == IPOPT_SSRR) {
171 					type = ICMP_UNREACH;
172 					code = ICMP_UNREACH_SRCFAIL;
173 					goto bad;
174 				}
175 				if (!V_ip_dosourceroute)
176 					goto nosourcerouting;
177 				/*
178 				 * Loose routing, and not at next destination
179 				 * yet; nothing to do except forward.
180 				 */
181 				break;
182 			}
183 			off--;			/* 0 origin */
184 			if (off > optlen - (int)sizeof(struct in_addr)) {
185 				/*
186 				 * End of source route.  Should be for us.
187 				 */
188 				if (!V_ip_acceptsourceroute)
189 					goto nosourcerouting;
190 				save_rte(m, cp, ip->ip_src);
191 				break;
192 			}
193 #ifdef IPSTEALTH
194 			if (V_ipstealth)
195 				goto dropit;
196 #endif
197 			if (!V_ip_dosourceroute) {
198 				if (V_ipforwarding) {
199 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
200 					/*
201 					 * Acting as a router, so generate
202 					 * ICMP
203 					 */
204 nosourcerouting:
205 					strcpy(buf, inet_ntoa(ip->ip_dst));
206 					log(LOG_WARNING,
207 					    "attempted source route from %s to %s\n",
208 					    inet_ntoa(ip->ip_src), buf);
209 					type = ICMP_UNREACH;
210 					code = ICMP_UNREACH_SRCFAIL;
211 					goto bad;
212 				} else {
213 					/*
214 					 * Not acting as a router, so
215 					 * silently drop.
216 					 */
217 #ifdef IPSTEALTH
218 dropit:
219 #endif
220 					IPSTAT_INC(ips_cantforward);
221 					m_freem(m);
222 					return (1);
223 				}
224 			}
225 
226 			/*
227 			 * locate outgoing interface
228 			 */
229 			(void)memcpy(&ipaddr.sin_addr, cp + off,
230 			    sizeof(ipaddr.sin_addr));
231 
232 			type = ICMP_UNREACH;
233 			code = ICMP_UNREACH_SRCFAIL;
234 
235 			if (opt == IPOPT_SSRR) {
236 #define	INA	struct in_ifaddr *
237 #define	SA	struct sockaddr *
238 			    ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr,
239 					    RT_ALL_FIBS);
240 			    if (ia == NULL)
241 				    ia = (INA)ifa_ifwithnet((SA)&ipaddr, 0,
242 						    RT_ALL_FIBS);
243 				if (ia == NULL)
244 					goto bad;
245 
246 				memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
247 				    sizeof(struct in_addr));
248 				ifa_free(&ia->ia_ifa);
249 			} else {
250 				/* XXX MRT 0 for routing */
251 				if (fib4_lookup_nh_ext(M_GETFIB(m),
252 				    ipaddr.sin_addr, 0, 0, &nh_ext) != 0)
253 					goto bad;
254 
255 				memcpy(cp + off, &nh_ext.nh_src,
256 				    sizeof(struct in_addr));
257 			}
258 
259 			ip->ip_dst = ipaddr.sin_addr;
260 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
261 			/*
262 			 * Let ip_intr's mcast routing check handle mcast pkts
263 			 */
264 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
265 			break;
266 
267 		case IPOPT_RR:
268 #ifdef IPSTEALTH
269 			if (V_ipstealth && pass == 0)
270 				break;
271 #endif
272 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
273 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
274 				goto bad;
275 			}
276 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
277 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
278 				goto bad;
279 			}
280 			/*
281 			 * If no space remains, ignore.
282 			 */
283 			off--;			/* 0 origin */
284 			if (off > optlen - (int)sizeof(struct in_addr))
285 				break;
286 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
287 			    sizeof(ipaddr.sin_addr));
288 			/*
289 			 * Locate outgoing interface; if we're the
290 			 * destination, use the incoming interface (should be
291 			 * same).
292 			 */
293 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) != NULL) {
294 				memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
295 				    sizeof(struct in_addr));
296 				ifa_free(&ia->ia_ifa);
297 			} else if (fib4_lookup_nh_ext(M_GETFIB(m),
298 			    ipaddr.sin_addr, 0, 0, &nh_ext) == 0) {
299 				memcpy(cp + off, &nh_ext.nh_src,
300 				    sizeof(struct in_addr));
301 			} else {
302 				type = ICMP_UNREACH;
303 				code = ICMP_UNREACH_HOST;
304 				goto bad;
305 			}
306 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
307 			break;
308 
309 		case IPOPT_TS:
310 #ifdef IPSTEALTH
311 			if (V_ipstealth && pass == 0)
312 				break;
313 #endif
314 			code = cp - (u_char *)ip;
315 			if (optlen < 4 || optlen > 40) {
316 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
317 				goto bad;
318 			}
319 			if ((off = cp[IPOPT_OFFSET]) < 5) {
320 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
321 				goto bad;
322 			}
323 			if (off > optlen - (int)sizeof(int32_t)) {
324 				cp[IPOPT_OFFSET + 1] += (1 << 4);
325 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
326 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
327 					goto bad;
328 				}
329 				break;
330 			}
331 			off--;				/* 0 origin */
332 			sin = (struct in_addr *)(cp + off);
333 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
334 
335 			case IPOPT_TS_TSONLY:
336 				break;
337 
338 			case IPOPT_TS_TSANDADDR:
339 				if (off + sizeof(uint32_t) +
340 				    sizeof(struct in_addr) > optlen) {
341 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
342 					goto bad;
343 				}
344 				ipaddr.sin_addr = dst;
345 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
346 							    m->m_pkthdr.rcvif);
347 				if (ia == NULL)
348 					continue;
349 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
350 				    sizeof(struct in_addr));
351 				ifa_free(&ia->ia_ifa);
352 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
353 				off += sizeof(struct in_addr);
354 				break;
355 
356 			case IPOPT_TS_PRESPEC:
357 				if (off + sizeof(uint32_t) +
358 				    sizeof(struct in_addr) > optlen) {
359 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
360 					goto bad;
361 				}
362 				(void)memcpy(&ipaddr.sin_addr, sin,
363 				    sizeof(struct in_addr));
364 				if (ifa_ifwithaddr_check((SA)&ipaddr) == 0)
365 					continue;
366 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
367 				off += sizeof(struct in_addr);
368 				break;
369 
370 			default:
371 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
372 				goto bad;
373 			}
374 			ntime = iptime();
375 			(void)memcpy(cp + off, &ntime, sizeof(uint32_t));
376 			cp[IPOPT_OFFSET] += sizeof(uint32_t);
377 		}
378 	}
379 	if (forward && V_ipforwarding) {
380 		ip_forward(m, 1);
381 		return (1);
382 	}
383 	return (0);
384 bad:
385 	icmp_error(m, type, code, 0, 0);
386 	IPSTAT_INC(ips_badoptions);
387 	return (1);
388 }
389 
390 /*
391  * Save incoming source route for use in replies, to be picked up later by
392  * ip_srcroute if the receiver is interested.
393  */
394 static void
395 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
396 {
397 	unsigned olen;
398 	struct ipopt_tag *opts;
399 
400 	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
401 	    sizeof(struct ipopt_tag), M_NOWAIT);
402 	if (opts == NULL)
403 		return;
404 
405 	olen = option[IPOPT_OLEN];
406 	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) {
407 		m_tag_free((struct m_tag *)opts);
408 		return;
409 	}
410 	bcopy(option, opts->ip_srcrt.srcopt, olen);
411 	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
412 	opts->ip_srcrt.dst = dst;
413 	m_tag_prepend(m, (struct m_tag *)opts);
414 }
415 
416 /*
417  * Retrieve incoming source route for use in replies, in the same form used
418  * by setsockopt.  The first hop is placed before the options, will be
419  * removed later.
420  */
421 struct mbuf *
422 ip_srcroute(struct mbuf *m0)
423 {
424 	struct in_addr *p, *q;
425 	struct mbuf *m;
426 	struct ipopt_tag *opts;
427 
428 	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
429 	if (opts == NULL)
430 		return (NULL);
431 
432 	if (opts->ip_nhops == 0)
433 		return (NULL);
434 	m = m_get(M_NOWAIT, MT_DATA);
435 	if (m == NULL)
436 		return (NULL);
437 
438 #define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
439 
440 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
441 	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
442 	    sizeof(struct in_addr) + OPTSIZ;
443 
444 	/*
445 	 * First, save first hop for return route.
446 	 */
447 	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
448 	*(mtod(m, struct in_addr *)) = *p--;
449 
450 	/*
451 	 * Copy option fields and padding (nop) to mbuf.
452 	 */
453 	opts->ip_srcrt.nop = IPOPT_NOP;
454 	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
455 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
456 	    &(opts->ip_srcrt.nop), OPTSIZ);
457 	q = (struct in_addr *)(mtod(m, caddr_t) +
458 	    sizeof(struct in_addr) + OPTSIZ);
459 #undef OPTSIZ
460 	/*
461 	 * Record return path as an IP source route, reversing the path
462 	 * (pointers are now aligned).
463 	 */
464 	while (p >= opts->ip_srcrt.route) {
465 		*q++ = *p--;
466 	}
467 	/*
468 	 * Last hop goes to final destination.
469 	 */
470 	*q = opts->ip_srcrt.dst;
471 	m_tag_delete(m0, (struct m_tag *)opts);
472 	return (m);
473 }
474 
475 /*
476  * Strip out IP options, at higher level protocol in the kernel.
477  */
478 void
479 ip_stripoptions(struct mbuf *m)
480 {
481 	struct ip *ip = mtod(m, struct ip *);
482 	int olen;
483 
484 	olen = (ip->ip_hl << 2) - sizeof(struct ip);
485 	m->m_len -= olen;
486 	if (m->m_flags & M_PKTHDR)
487 		m->m_pkthdr.len -= olen;
488 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
489 	ip->ip_hl = sizeof(struct ip) >> 2;
490 
491 	bcopy((char *)ip + sizeof(struct ip) + olen, (ip + 1),
492 	    (size_t )(m->m_len - sizeof(struct ip)));
493 }
494 
495 /*
496  * Insert IP options into preformed packet.  Adjust IP destination as
497  * required for IP source routing, as indicated by a non-zero in_addr at the
498  * start of the options.
499  *
500  * XXX This routine assumes that the packet has no options in place.
501  */
502 struct mbuf *
503 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
504 {
505 	struct ipoption *p = mtod(opt, struct ipoption *);
506 	struct mbuf *n;
507 	struct ip *ip = mtod(m, struct ip *);
508 	unsigned optlen;
509 
510 	optlen = opt->m_len - sizeof(p->ipopt_dst);
511 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) {
512 		*phlen = 0;
513 		return (m);		/* XXX should fail */
514 	}
515 	if (p->ipopt_dst.s_addr)
516 		ip->ip_dst = p->ipopt_dst;
517 	if (!M_WRITABLE(m) || M_LEADINGSPACE(m) < optlen) {
518 		n = m_gethdr(M_NOWAIT, MT_DATA);
519 		if (n == NULL) {
520 			*phlen = 0;
521 			return (m);
522 		}
523 		m_move_pkthdr(n, m);
524 		n->m_pkthdr.rcvif = NULL;
525 		n->m_pkthdr.len += optlen;
526 		m->m_len -= sizeof(struct ip);
527 		m->m_data += sizeof(struct ip);
528 		n->m_next = m;
529 		m = n;
530 		m->m_len = optlen + sizeof(struct ip);
531 		m->m_data += max_linkhdr;
532 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
533 	} else {
534 		m->m_data -= optlen;
535 		m->m_len += optlen;
536 		m->m_pkthdr.len += optlen;
537 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
538 	}
539 	ip = mtod(m, struct ip *);
540 	bcopy(p->ipopt_list, ip + 1, optlen);
541 	*phlen = sizeof(struct ip) + optlen;
542 	ip->ip_v = IPVERSION;
543 	ip->ip_hl = *phlen >> 2;
544 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
545 	return (m);
546 }
547 
548 /*
549  * Copy options from ip to jp, omitting those not copied during
550  * fragmentation.
551  */
552 int
553 ip_optcopy(struct ip *ip, struct ip *jp)
554 {
555 	u_char *cp, *dp;
556 	int opt, optlen, cnt;
557 
558 	cp = (u_char *)(ip + 1);
559 	dp = (u_char *)(jp + 1);
560 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
561 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
562 		opt = cp[0];
563 		if (opt == IPOPT_EOL)
564 			break;
565 		if (opt == IPOPT_NOP) {
566 			/* Preserve for IP mcast tunnel's LSRR alignment. */
567 			*dp++ = IPOPT_NOP;
568 			optlen = 1;
569 			continue;
570 		}
571 
572 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
573 		    ("ip_optcopy: malformed ipv4 option"));
574 		optlen = cp[IPOPT_OLEN];
575 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
576 		    ("ip_optcopy: malformed ipv4 option"));
577 
578 		/* Bogus lengths should have been caught by ip_dooptions. */
579 		if (optlen > cnt)
580 			optlen = cnt;
581 		if (IPOPT_COPIED(opt)) {
582 			bcopy(cp, dp, optlen);
583 			dp += optlen;
584 		}
585 	}
586 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
587 		*dp++ = IPOPT_EOL;
588 	return (optlen);
589 }
590 
591 /*
592  * Set up IP options in pcb for insertion in output packets.  Store in mbuf
593  * with pointer in pcbopt, adding pseudo-option with destination address if
594  * source routed.
595  */
596 int
597 ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m)
598 {
599 	int cnt, optlen;
600 	u_char *cp;
601 	struct mbuf **pcbopt;
602 	u_char opt;
603 
604 	INP_WLOCK_ASSERT(inp);
605 
606 	pcbopt = &inp->inp_options;
607 
608 	/* turn off any old options */
609 	if (*pcbopt)
610 		(void)m_free(*pcbopt);
611 	*pcbopt = NULL;
612 	if (m == NULL || m->m_len == 0) {
613 		/*
614 		 * Only turning off any previous options.
615 		 */
616 		if (m != NULL)
617 			(void)m_free(m);
618 		return (0);
619 	}
620 
621 	if (m->m_len % sizeof(int32_t))
622 		goto bad;
623 	/*
624 	 * IP first-hop destination address will be stored before actual
625 	 * options; move other options back and clear it when none present.
626 	 */
627 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
628 		goto bad;
629 	cnt = m->m_len;
630 	m->m_len += sizeof(struct in_addr);
631 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
632 	bcopy(mtod(m, void *), cp, (unsigned)cnt);
633 	bzero(mtod(m, void *), sizeof(struct in_addr));
634 
635 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
636 		opt = cp[IPOPT_OPTVAL];
637 		if (opt == IPOPT_EOL)
638 			break;
639 		if (opt == IPOPT_NOP)
640 			optlen = 1;
641 		else {
642 			if (cnt < IPOPT_OLEN + sizeof(*cp))
643 				goto bad;
644 			optlen = cp[IPOPT_OLEN];
645 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
646 				goto bad;
647 		}
648 		switch (opt) {
649 
650 		default:
651 			break;
652 
653 		case IPOPT_LSRR:
654 		case IPOPT_SSRR:
655 			/*
656 			 * User process specifies route as:
657 			 *
658 			 *	->A->B->C->D
659 			 *
660 			 * D must be our final destination (but we can't
661 			 * check that since we may not have connected yet).
662 			 * A is first hop destination, which doesn't appear
663 			 * in actual IP option, but is stored before the
664 			 * options.
665 			 */
666 			/* XXX-BZ PRIV_NETINET_SETHDROPTS? */
667 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
668 				goto bad;
669 			m->m_len -= sizeof(struct in_addr);
670 			cnt -= sizeof(struct in_addr);
671 			optlen -= sizeof(struct in_addr);
672 			cp[IPOPT_OLEN] = optlen;
673 			/*
674 			 * Move first hop before start of options.
675 			 */
676 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
677 			    sizeof(struct in_addr));
678 			/*
679 			 * Then copy rest of options back
680 			 * to close up the deleted entry.
681 			 */
682 			bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
683 			    &cp[IPOPT_OFFSET+1],
684 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
685 			break;
686 		}
687 	}
688 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
689 		goto bad;
690 	*pcbopt = m;
691 	return (0);
692 
693 bad:
694 	(void)m_free(m);
695 	return (EINVAL);
696 }
697 
698 /*
699  * Check for the presence of the IP Router Alert option [RFC2113]
700  * in the header of an IPv4 datagram.
701  *
702  * This call is not intended for use from the forwarding path; it is here
703  * so that protocol domains may check for the presence of the option.
704  * Given how FreeBSD's IPv4 stack is currently structured, the Router Alert
705  * option does not have much relevance to the implementation, though this
706  * may change in future.
707  * Router alert options SHOULD be passed if running in IPSTEALTH mode and
708  * we are not the endpoint.
709  * Length checks on individual options should already have been performed
710  * by ip_dooptions() therefore they are folded under INVARIANTS here.
711  *
712  * Return zero if not present or options are invalid, non-zero if present.
713  */
714 int
715 ip_checkrouteralert(struct mbuf *m)
716 {
717 	struct ip *ip = mtod(m, struct ip *);
718 	u_char *cp;
719 	int opt, optlen, cnt, found_ra;
720 
721 	found_ra = 0;
722 	cp = (u_char *)(ip + 1);
723 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
724 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
725 		opt = cp[IPOPT_OPTVAL];
726 		if (opt == IPOPT_EOL)
727 			break;
728 		if (opt == IPOPT_NOP)
729 			optlen = 1;
730 		else {
731 #ifdef INVARIANTS
732 			if (cnt < IPOPT_OLEN + sizeof(*cp))
733 				break;
734 #endif
735 			optlen = cp[IPOPT_OLEN];
736 #ifdef INVARIANTS
737 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
738 				break;
739 #endif
740 		}
741 		switch (opt) {
742 		case IPOPT_RA:
743 #ifdef INVARIANTS
744 			if (optlen != IPOPT_OFFSET + sizeof(uint16_t) ||
745 			    (*((uint16_t *)&cp[IPOPT_OFFSET]) != 0))
746 			    break;
747 			else
748 #endif
749 			found_ra = 1;
750 			break;
751 		default:
752 			break;
753 		}
754 	}
755 
756 	return (found_ra);
757 }
758