xref: /freebsd/sys/netinet/ip_options.c (revision ec0ea6efa1ad229d75c394c1a9b9cac33af2b1d3)
1 /*
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *      The Regents of the University of California.
6  * Copyright (c) 2005 Andre Oppermann, Internet Business Solutions AG.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ipstealth.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/mbuf.h>
42 #include <sys/domain.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/syslog.h>
48 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/if_types.h>
52 #include <net/if_var.h>
53 #include <net/if_dl.h>
54 #include <net/route.h>
55 #include <net/route/nhop.h>
56 #include <net/netisr.h>
57 #include <net/vnet.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_fib.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/ip_var.h>
66 #include <netinet/ip_options.h>
67 #include <netinet/ip_icmp.h>
68 #include <machine/in_cksum.h>
69 
70 #include <sys/socketvar.h>
71 
72 VNET_DEFINE_STATIC(int, ip_dosourceroute);
73 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute,
74     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_dosourceroute), 0,
75     "Enable forwarding source routed IP packets");
76 #define	V_ip_dosourceroute	VNET(ip_dosourceroute)
77 
78 VNET_DEFINE_STATIC(int,	ip_acceptsourceroute);
79 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
80     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_acceptsourceroute), 0,
81     "Enable accepting source routed IP packets");
82 #define	V_ip_acceptsourceroute	VNET(ip_acceptsourceroute)
83 
84 VNET_DEFINE(int, ip_doopts) = 1; /* 0 = ignore, 1 = process, 2 = reject */
85 SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_VNET | CTLFLAG_RW,
86     &VNET_NAME(ip_doopts), 0, "Enable IP options processing ([LS]SRR, RR, TS)");
87 
88 static void	save_rte(struct mbuf *m, u_char *, struct in_addr);
89 
90 /*
91  * Do option processing on a datagram, possibly discarding it if bad options
92  * are encountered, or forwarding it if source-routed.
93  *
94  * The pass argument is used when operating in the IPSTEALTH mode to tell
95  * what options to process: [LS]SRR (pass 0) or the others (pass 1).  The
96  * reason for as many as two passes is that when doing IPSTEALTH, non-routing
97  * options should be processed only if the packet is for us.
98  *
99  * Returns 1 if packet has been forwarded/freed, 0 if the packet should be
100  * processed further.
101  */
102 int
103 ip_dooptions(struct mbuf *m, int pass)
104 {
105 	struct ip *ip = mtod(m, struct ip *);
106 	u_char *cp;
107 	struct in_ifaddr *ia;
108 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
109 	struct in_addr *sin, dst;
110 	uint32_t ntime;
111 	struct nhop_object *nh;
112 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
113 
114 	NET_EPOCH_ASSERT();
115 
116 	/* Ignore or reject packets with IP options. */
117 	if (V_ip_doopts == 0)
118 		return 0;
119 	else if (V_ip_doopts == 2) {
120 		type = ICMP_UNREACH;
121 		code = ICMP_UNREACH_FILTER_PROHIB;
122 		goto bad;
123 	}
124 
125 	dst = ip->ip_dst;
126 	cp = (u_char *)(ip + 1);
127 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
128 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
129 		opt = cp[IPOPT_OPTVAL];
130 		if (opt == IPOPT_EOL)
131 			break;
132 		if (opt == IPOPT_NOP)
133 			optlen = 1;
134 		else {
135 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
136 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
137 				goto bad;
138 			}
139 			optlen = cp[IPOPT_OLEN];
140 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
141 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
142 				goto bad;
143 			}
144 		}
145 		switch (opt) {
146 		default:
147 			break;
148 
149 		/*
150 		 * Source routing with record.  Find interface with current
151 		 * destination address.  If none on this machine then drop if
152 		 * strictly routed, or do nothing if loosely routed.  Record
153 		 * interface address and bring up next address component.  If
154 		 * strictly routed make sure next address is on directly
155 		 * accessible net.
156 		 */
157 		case IPOPT_LSRR:
158 		case IPOPT_SSRR:
159 #ifdef IPSTEALTH
160 			if (V_ipstealth && pass > 0)
161 				break;
162 #endif
163 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
164 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
165 				goto bad;
166 			}
167 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
168 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
169 				goto bad;
170 			}
171 			ipaddr.sin_addr = ip->ip_dst;
172 			if (ifa_ifwithaddr_check((struct sockaddr *)&ipaddr)
173 			    == 0) {
174 				if (opt == IPOPT_SSRR) {
175 					type = ICMP_UNREACH;
176 					code = ICMP_UNREACH_SRCFAIL;
177 					goto bad;
178 				}
179 				if (!V_ip_dosourceroute)
180 					goto nosourcerouting;
181 				/*
182 				 * Loose routing, and not at next destination
183 				 * yet; nothing to do except forward.
184 				 */
185 				break;
186 			}
187 			off--;			/* 0 origin */
188 			if (off > optlen - (int)sizeof(struct in_addr)) {
189 				/*
190 				 * End of source route.  Should be for us.
191 				 */
192 				if (!V_ip_acceptsourceroute)
193 					goto nosourcerouting;
194 				save_rte(m, cp, ip->ip_src);
195 				break;
196 			}
197 #ifdef IPSTEALTH
198 			if (V_ipstealth)
199 				goto dropit;
200 #endif
201 			if (!V_ip_dosourceroute) {
202 				if (V_ipforwarding) {
203 					char srcbuf[INET_ADDRSTRLEN];
204 					char dstbuf[INET_ADDRSTRLEN];
205 
206 					/*
207 					 * Acting as a router, so generate
208 					 * ICMP
209 					 */
210 nosourcerouting:
211 					log(LOG_WARNING,
212 					    "attempted source route from %s "
213 					    "to %s\n",
214 					    inet_ntoa_r(ip->ip_src, srcbuf),
215 					    inet_ntoa_r(ip->ip_dst, dstbuf));
216 					type = ICMP_UNREACH;
217 					code = ICMP_UNREACH_SRCFAIL;
218 					goto bad;
219 				} else {
220 					/*
221 					 * Not acting as a router, so
222 					 * silently drop.
223 					 */
224 #ifdef IPSTEALTH
225 dropit:
226 #endif
227 					IPSTAT_INC(ips_cantforward);
228 					m_freem(m);
229 					return (1);
230 				}
231 			}
232 
233 			/*
234 			 * locate outgoing interface
235 			 */
236 			(void)memcpy(&ipaddr.sin_addr, cp + off,
237 			    sizeof(ipaddr.sin_addr));
238 
239 			type = ICMP_UNREACH;
240 			code = ICMP_UNREACH_SRCFAIL;
241 
242 			if (opt == IPOPT_SSRR) {
243 #define	INA	struct in_ifaddr *
244 #define	SA	struct sockaddr *
245 			    ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr,
246 					    RT_ALL_FIBS);
247 			    if (ia == NULL)
248 				    ia = (INA)ifa_ifwithnet((SA)&ipaddr, 0,
249 						    RT_ALL_FIBS);
250 				if (ia == NULL)
251 					goto bad;
252 
253 				memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
254 				    sizeof(struct in_addr));
255 			} else {
256 				/* XXX MRT 0 for routing */
257 				nh = fib4_lookup(M_GETFIB(m), ipaddr.sin_addr,
258 				     0, NHR_NONE, 0);
259 				if (nh == NULL)
260 					goto bad;
261 
262 				memcpy(cp + off, &(IA_SIN(nh->nh_ifa)->sin_addr),
263 				    sizeof(struct in_addr));
264 			}
265 
266 			ip->ip_dst = ipaddr.sin_addr;
267 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
268 			/*
269 			 * Let ip_intr's mcast routing check handle mcast pkts
270 			 */
271 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
272 			break;
273 
274 		case IPOPT_RR:
275 #ifdef IPSTEALTH
276 			if (V_ipstealth && pass == 0)
277 				break;
278 #endif
279 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
280 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
281 				goto bad;
282 			}
283 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
284 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
285 				goto bad;
286 			}
287 			/*
288 			 * If no space remains, ignore.
289 			 */
290 			off--;			/* 0 origin */
291 			if (off > optlen - (int)sizeof(struct in_addr))
292 				break;
293 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
294 			    sizeof(ipaddr.sin_addr));
295 			/*
296 			 * Locate outgoing interface; if we're the
297 			 * destination, use the incoming interface (should be
298 			 * same).
299 			 */
300 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) != NULL) {
301 				memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
302 				    sizeof(struct in_addr));
303 			} else if ((nh = fib4_lookup(M_GETFIB(m),
304 			    ipaddr.sin_addr, 0, NHR_NONE, 0)) != NULL) {
305 				memcpy(cp + off, &(IA_SIN(nh->nh_ifa)->sin_addr),
306 				    sizeof(struct in_addr));
307 			} else {
308 				type = ICMP_UNREACH;
309 				code = ICMP_UNREACH_HOST;
310 				goto bad;
311 			}
312 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
313 			break;
314 
315 		case IPOPT_TS:
316 #ifdef IPSTEALTH
317 			if (V_ipstealth && pass == 0)
318 				break;
319 #endif
320 			code = cp - (u_char *)ip;
321 			if (optlen < 4 || optlen > 40) {
322 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
323 				goto bad;
324 			}
325 			if ((off = cp[IPOPT_OFFSET]) < 5) {
326 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
327 				goto bad;
328 			}
329 			if (off > optlen - (int)sizeof(int32_t)) {
330 				cp[IPOPT_OFFSET + 1] += (1 << 4);
331 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
332 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
333 					goto bad;
334 				}
335 				break;
336 			}
337 			off--;				/* 0 origin */
338 			sin = (struct in_addr *)(cp + off);
339 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
340 			case IPOPT_TS_TSONLY:
341 				break;
342 
343 			case IPOPT_TS_TSANDADDR:
344 				if (off + sizeof(uint32_t) +
345 				    sizeof(struct in_addr) > optlen) {
346 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
347 					goto bad;
348 				}
349 				ipaddr.sin_addr = dst;
350 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
351 							    m->m_pkthdr.rcvif);
352 				if (ia == NULL)
353 					continue;
354 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
355 				    sizeof(struct in_addr));
356 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
357 				off += sizeof(struct in_addr);
358 				break;
359 
360 			case IPOPT_TS_PRESPEC:
361 				if (off + sizeof(uint32_t) +
362 				    sizeof(struct in_addr) > optlen) {
363 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
364 					goto bad;
365 				}
366 				(void)memcpy(&ipaddr.sin_addr, sin,
367 				    sizeof(struct in_addr));
368 				if (ifa_ifwithaddr_check((SA)&ipaddr) == 0)
369 					continue;
370 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
371 				off += sizeof(struct in_addr);
372 				break;
373 
374 			default:
375 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
376 				goto bad;
377 			}
378 			ntime = iptime();
379 			(void)memcpy(cp + off, &ntime, sizeof(uint32_t));
380 			cp[IPOPT_OFFSET] += sizeof(uint32_t);
381 		}
382 	}
383 	if (forward && V_ipforwarding) {
384 		ip_forward(m, 1);
385 		return (1);
386 	}
387 	return (0);
388 bad:
389 	icmp_error(m, type, code, 0, 0);
390 	IPSTAT_INC(ips_badoptions);
391 	return (1);
392 }
393 
394 /*
395  * Save incoming source route for use in replies, to be picked up later by
396  * ip_srcroute if the receiver is interested.
397  */
398 static void
399 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
400 {
401 	unsigned olen;
402 	struct ipopt_tag *opts;
403 
404 	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
405 	    sizeof(struct ipopt_tag), M_NOWAIT);
406 	if (opts == NULL)
407 		return;
408 
409 	olen = option[IPOPT_OLEN];
410 	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) {
411 		m_tag_free((struct m_tag *)opts);
412 		return;
413 	}
414 	bcopy(option, opts->ip_srcrt.srcopt, olen);
415 	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
416 	opts->ip_srcrt.dst = dst;
417 	m_tag_prepend(m, (struct m_tag *)opts);
418 }
419 
420 /*
421  * Retrieve incoming source route for use in replies, in the same form used
422  * by setsockopt.  The first hop is placed before the options, will be
423  * removed later.
424  */
425 struct mbuf *
426 ip_srcroute(struct mbuf *m0)
427 {
428 	struct in_addr *p, *q;
429 	struct mbuf *m;
430 	struct ipopt_tag *opts;
431 
432 	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
433 	if (opts == NULL)
434 		return (NULL);
435 
436 	if (opts->ip_nhops == 0)
437 		return (NULL);
438 	m = m_get(M_NOWAIT, MT_DATA);
439 	if (m == NULL)
440 		return (NULL);
441 
442 #define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
443 
444 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
445 	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
446 	    sizeof(struct in_addr) + OPTSIZ;
447 
448 	/*
449 	 * First, save first hop for return route.
450 	 */
451 	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
452 	*(mtod(m, struct in_addr *)) = *p--;
453 
454 	/*
455 	 * Copy option fields and padding (nop) to mbuf.
456 	 */
457 	opts->ip_srcrt.nop = IPOPT_NOP;
458 	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
459 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
460 	    &(opts->ip_srcrt.nop), OPTSIZ);
461 	q = (struct in_addr *)(mtod(m, caddr_t) +
462 	    sizeof(struct in_addr) + OPTSIZ);
463 #undef OPTSIZ
464 	/*
465 	 * Record return path as an IP source route, reversing the path
466 	 * (pointers are now aligned).
467 	 */
468 	while (p >= opts->ip_srcrt.route) {
469 		*q++ = *p--;
470 	}
471 	/*
472 	 * Last hop goes to final destination.
473 	 */
474 	*q = opts->ip_srcrt.dst;
475 	m_tag_delete(m0, (struct m_tag *)opts);
476 	return (m);
477 }
478 
479 /*
480  * Strip out IP options, at higher level protocol in the kernel.
481  */
482 void
483 ip_stripoptions(struct mbuf *m)
484 {
485 	struct ip *ip = mtod(m, struct ip *);
486 	int olen;
487 
488 	olen = (ip->ip_hl << 2) - sizeof(struct ip);
489 	m->m_len -= olen;
490 	if (m->m_flags & M_PKTHDR)
491 		m->m_pkthdr.len -= olen;
492 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
493 	ip->ip_hl = sizeof(struct ip) >> 2;
494 
495 	bcopy((char *)ip + sizeof(struct ip) + olen, (ip + 1),
496 	    (size_t )(m->m_len - sizeof(struct ip)));
497 }
498 
499 /*
500  * Insert IP options into preformed packet.  Adjust IP destination as
501  * required for IP source routing, as indicated by a non-zero in_addr at the
502  * start of the options.
503  *
504  * XXX This routine assumes that the packet has no options in place.
505  */
506 struct mbuf *
507 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
508 {
509 	struct ipoption *p = mtod(opt, struct ipoption *);
510 	struct mbuf *n;
511 	struct ip *ip = mtod(m, struct ip *);
512 	unsigned optlen;
513 
514 	optlen = opt->m_len - sizeof(p->ipopt_dst);
515 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) {
516 		*phlen = 0;
517 		return (m);		/* XXX should fail */
518 	}
519 	if (p->ipopt_dst.s_addr)
520 		ip->ip_dst = p->ipopt_dst;
521 	if (!M_WRITABLE(m) || M_LEADINGSPACE(m) < optlen) {
522 		n = m_gethdr(M_NOWAIT, MT_DATA);
523 		if (n == NULL) {
524 			*phlen = 0;
525 			return (m);
526 		}
527 		m_move_pkthdr(n, m);
528 		n->m_pkthdr.rcvif = NULL;
529 		n->m_pkthdr.len += optlen;
530 		m->m_len -= sizeof(struct ip);
531 		m->m_data += sizeof(struct ip);
532 		n->m_next = m;
533 		m = n;
534 		m->m_len = optlen + sizeof(struct ip);
535 		m->m_data += max_linkhdr;
536 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
537 	} else {
538 		m->m_data -= optlen;
539 		m->m_len += optlen;
540 		m->m_pkthdr.len += optlen;
541 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
542 	}
543 	ip = mtod(m, struct ip *);
544 	bcopy(p->ipopt_list, ip + 1, optlen);
545 	*phlen = sizeof(struct ip) + optlen;
546 	ip->ip_v = IPVERSION;
547 	ip->ip_hl = *phlen >> 2;
548 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
549 	return (m);
550 }
551 
552 /*
553  * Copy options from ip to jp, omitting those not copied during
554  * fragmentation.
555  */
556 int
557 ip_optcopy(struct ip *ip, struct ip *jp)
558 {
559 	u_char *cp, *dp;
560 	int opt, optlen, cnt;
561 
562 	cp = (u_char *)(ip + 1);
563 	dp = (u_char *)(jp + 1);
564 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
565 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
566 		opt = cp[0];
567 		if (opt == IPOPT_EOL)
568 			break;
569 		if (opt == IPOPT_NOP) {
570 			/* Preserve for IP mcast tunnel's LSRR alignment. */
571 			*dp++ = IPOPT_NOP;
572 			optlen = 1;
573 			continue;
574 		}
575 
576 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
577 		    ("ip_optcopy: malformed ipv4 option"));
578 		optlen = cp[IPOPT_OLEN];
579 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
580 		    ("ip_optcopy: malformed ipv4 option"));
581 
582 		/* Bogus lengths should have been caught by ip_dooptions. */
583 		if (optlen > cnt)
584 			optlen = cnt;
585 		if (IPOPT_COPIED(opt)) {
586 			bcopy(cp, dp, optlen);
587 			dp += optlen;
588 		}
589 	}
590 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
591 		*dp++ = IPOPT_EOL;
592 	return (optlen);
593 }
594 
595 /*
596  * Set up IP options in pcb for insertion in output packets.  Store in mbuf
597  * with pointer in pcbopt, adding pseudo-option with destination address if
598  * source routed.
599  */
600 int
601 ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m)
602 {
603 	int cnt, optlen;
604 	u_char *cp;
605 	struct mbuf **pcbopt;
606 	u_char opt;
607 
608 	INP_WLOCK_ASSERT(inp);
609 
610 	pcbopt = &inp->inp_options;
611 
612 	/* turn off any old options */
613 	if (*pcbopt)
614 		(void)m_free(*pcbopt);
615 	*pcbopt = NULL;
616 	if (m == NULL || m->m_len == 0) {
617 		/*
618 		 * Only turning off any previous options.
619 		 */
620 		if (m != NULL)
621 			(void)m_free(m);
622 		return (0);
623 	}
624 
625 	if (m->m_len % sizeof(int32_t))
626 		goto bad;
627 	/*
628 	 * IP first-hop destination address will be stored before actual
629 	 * options; move other options back and clear it when none present.
630 	 */
631 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
632 		goto bad;
633 	cnt = m->m_len;
634 	m->m_len += sizeof(struct in_addr);
635 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
636 	bcopy(mtod(m, void *), cp, (unsigned)cnt);
637 	bzero(mtod(m, void *), sizeof(struct in_addr));
638 
639 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
640 		opt = cp[IPOPT_OPTVAL];
641 		if (opt == IPOPT_EOL)
642 			break;
643 		if (opt == IPOPT_NOP)
644 			optlen = 1;
645 		else {
646 			if (cnt < IPOPT_OLEN + sizeof(*cp))
647 				goto bad;
648 			optlen = cp[IPOPT_OLEN];
649 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
650 				goto bad;
651 		}
652 		switch (opt) {
653 		default:
654 			break;
655 
656 		case IPOPT_LSRR:
657 		case IPOPT_SSRR:
658 			/*
659 			 * User process specifies route as:
660 			 *
661 			 *	->A->B->C->D
662 			 *
663 			 * D must be our final destination (but we can't
664 			 * check that since we may not have connected yet).
665 			 * A is first hop destination, which doesn't appear
666 			 * in actual IP option, but is stored before the
667 			 * options.
668 			 */
669 			/* XXX-BZ PRIV_NETINET_SETHDROPTS? */
670 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
671 				goto bad;
672 			m->m_len -= sizeof(struct in_addr);
673 			cnt -= sizeof(struct in_addr);
674 			optlen -= sizeof(struct in_addr);
675 			cp[IPOPT_OLEN] = optlen;
676 			/*
677 			 * Move first hop before start of options.
678 			 */
679 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
680 			    sizeof(struct in_addr));
681 			/*
682 			 * Then copy rest of options back
683 			 * to close up the deleted entry.
684 			 */
685 			bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
686 			    &cp[IPOPT_OFFSET+1],
687 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
688 			break;
689 		}
690 	}
691 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
692 		goto bad;
693 	*pcbopt = m;
694 	return (0);
695 
696 bad:
697 	(void)m_free(m);
698 	return (EINVAL);
699 }
700 
701 /*
702  * Check for the presence of the IP Router Alert option [RFC2113]
703  * in the header of an IPv4 datagram.
704  *
705  * This call is not intended for use from the forwarding path; it is here
706  * so that protocol domains may check for the presence of the option.
707  * Given how FreeBSD's IPv4 stack is currently structured, the Router Alert
708  * option does not have much relevance to the implementation, though this
709  * may change in future.
710  * Router alert options SHOULD be passed if running in IPSTEALTH mode and
711  * we are not the endpoint.
712  * Length checks on individual options should already have been performed
713  * by ip_dooptions() therefore they are folded under INVARIANTS here.
714  *
715  * Return zero if not present or options are invalid, non-zero if present.
716  */
717 int
718 ip_checkrouteralert(struct mbuf *m)
719 {
720 	struct ip *ip = mtod(m, struct ip *);
721 	u_char *cp;
722 	int opt, optlen, cnt, found_ra;
723 
724 	found_ra = 0;
725 	cp = (u_char *)(ip + 1);
726 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
727 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
728 		opt = cp[IPOPT_OPTVAL];
729 		if (opt == IPOPT_EOL)
730 			break;
731 		if (opt == IPOPT_NOP)
732 			optlen = 1;
733 		else {
734 #ifdef INVARIANTS
735 			if (cnt < IPOPT_OLEN + sizeof(*cp))
736 				break;
737 #endif
738 			optlen = cp[IPOPT_OLEN];
739 #ifdef INVARIANTS
740 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
741 				break;
742 #endif
743 		}
744 		switch (opt) {
745 		case IPOPT_RA:
746 #ifdef INVARIANTS
747 			if (optlen != IPOPT_OFFSET + sizeof(uint16_t) ||
748 			    (*((uint16_t *)&cp[IPOPT_OFFSET]) != 0))
749 			    break;
750 			else
751 #endif
752 			found_ra = 1;
753 			break;
754 		default:
755 			break;
756 		}
757 	}
758 
759 	return (found_ra);
760 }
761