xref: /freebsd/sys/netinet/ip_options.c (revision b3aaa0cc21c63d388230c7ef2a80abd631ff20d5)
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 #include "opt_mac.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/mbuf.h>
41 #include <sys/domain.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/syslog.h>
47 #include <sys/sysctl.h>
48 #include <sys/vimage.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/netisr.h>
56 
57 #include <netinet/in.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 #include <netinet/vinet.h>
67 
68 #include <sys/socketvar.h>
69 
70 #include <security/mac/mac_framework.h>
71 
72 static int	ip_dosourceroute = 0;
73 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
74     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
75 
76 static int	ip_acceptsourceroute = 0;
77 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
78     CTLFLAG_RW, &ip_acceptsourceroute, 0,
79     "Enable accepting source routed IP packets");
80 
81 int		ip_doopts = 1;	/* 0 = ignore, 1 = process, 2 = reject */
82 SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW,
83     &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 	INIT_VNET_INET(curvnet);
103 	struct ip *ip = mtod(m, struct ip *);
104 	u_char *cp;
105 	struct in_ifaddr *ia;
106 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
107 	struct in_addr *sin, dst;
108 	uint32_t ntime;
109 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
110 
111 	/* Ignore or reject packets with IP options. */
112 	if (ip_doopts == 0)
113 		return 0;
114 	else if (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 			ia = (struct in_ifaddr *)
169 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
170 			if (ia == NULL) {
171 				if (opt == IPOPT_SSRR) {
172 					type = ICMP_UNREACH;
173 					code = ICMP_UNREACH_SRCFAIL;
174 					goto bad;
175 				}
176 				if (!ip_dosourceroute)
177 					goto nosourcerouting;
178 				/*
179 				 * Loose routing, and not at next destination
180 				 * yet; nothing to do except forward.
181 				 */
182 				break;
183 			}
184 			off--;			/* 0 origin */
185 			if (off > optlen - (int)sizeof(struct in_addr)) {
186 				/*
187 				 * End of source route.  Should be for us.
188 				 */
189 				if (!ip_acceptsourceroute)
190 					goto nosourcerouting;
191 				save_rte(m, cp, ip->ip_src);
192 				break;
193 			}
194 #ifdef IPSTEALTH
195 			if (V_ipstealth)
196 				goto dropit;
197 #endif
198 			if (!ip_dosourceroute) {
199 				if (V_ipforwarding) {
200 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
201 					/*
202 					 * Acting as a router, so generate
203 					 * ICMP
204 					 */
205 nosourcerouting:
206 					strcpy(buf, inet_ntoa(ip->ip_dst));
207 					log(LOG_WARNING,
208 					    "attempted source route from %s to %s\n",
209 					    inet_ntoa(ip->ip_src), buf);
210 					type = ICMP_UNREACH;
211 					code = ICMP_UNREACH_SRCFAIL;
212 					goto bad;
213 				} else {
214 					/*
215 					 * Not acting as a router, so
216 					 * silently drop.
217 					 */
218 #ifdef IPSTEALTH
219 dropit:
220 #endif
221 					V_ipstat.ips_cantforward++;
222 					m_freem(m);
223 					return (1);
224 				}
225 			}
226 
227 			/*
228 			 * locate outgoing interface
229 			 */
230 			(void)memcpy(&ipaddr.sin_addr, cp + off,
231 			    sizeof(ipaddr.sin_addr));
232 
233 			if (opt == IPOPT_SSRR) {
234 #define	INA	struct in_ifaddr *
235 #define	SA	struct sockaddr *
236 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
237 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
238 			} else
239 /* XXX MRT 0 for routing */
240 				ia = ip_rtaddr(ipaddr.sin_addr, M_GETFIB(m));
241 			if (ia == NULL) {
242 				type = ICMP_UNREACH;
243 				code = ICMP_UNREACH_SRCFAIL;
244 				goto bad;
245 			}
246 			ip->ip_dst = ipaddr.sin_addr;
247 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
248 			    sizeof(struct in_addr));
249 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
250 			/*
251 			 * Let ip_intr's mcast routing check handle mcast pkts
252 			 */
253 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
254 			break;
255 
256 		case IPOPT_RR:
257 #ifdef IPSTEALTH
258 			if (V_ipstealth && pass == 0)
259 				break;
260 #endif
261 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
262 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
263 				goto bad;
264 			}
265 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
266 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
267 				goto bad;
268 			}
269 			/*
270 			 * If no space remains, ignore.
271 			 */
272 			off--;			/* 0 origin */
273 			if (off > optlen - (int)sizeof(struct in_addr))
274 				break;
275 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
276 			    sizeof(ipaddr.sin_addr));
277 			/*
278 			 * Locate outgoing interface; if we're the
279 			 * destination, use the incoming interface (should be
280 			 * same).
281 			 */
282 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
283 			    (ia = ip_rtaddr(ipaddr.sin_addr, M_GETFIB(m))) == NULL) {
284 				type = ICMP_UNREACH;
285 				code = ICMP_UNREACH_HOST;
286 				goto bad;
287 			}
288 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
289 			    sizeof(struct in_addr));
290 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
291 			break;
292 
293 		case IPOPT_TS:
294 #ifdef IPSTEALTH
295 			if (V_ipstealth && pass == 0)
296 				break;
297 #endif
298 			code = cp - (u_char *)ip;
299 			if (optlen < 4 || optlen > 40) {
300 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
301 				goto bad;
302 			}
303 			if ((off = cp[IPOPT_OFFSET]) < 5) {
304 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
305 				goto bad;
306 			}
307 			if (off > optlen - (int)sizeof(int32_t)) {
308 				cp[IPOPT_OFFSET + 1] += (1 << 4);
309 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
310 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
311 					goto bad;
312 				}
313 				break;
314 			}
315 			off--;				/* 0 origin */
316 			sin = (struct in_addr *)(cp + off);
317 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
318 
319 			case IPOPT_TS_TSONLY:
320 				break;
321 
322 			case IPOPT_TS_TSANDADDR:
323 				if (off + sizeof(uint32_t) +
324 				    sizeof(struct in_addr) > optlen) {
325 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
326 					goto bad;
327 				}
328 				ipaddr.sin_addr = dst;
329 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
330 							    m->m_pkthdr.rcvif);
331 				if (ia == NULL)
332 					continue;
333 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
334 				    sizeof(struct in_addr));
335 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
336 				off += sizeof(struct in_addr);
337 				break;
338 
339 			case IPOPT_TS_PRESPEC:
340 				if (off + sizeof(uint32_t) +
341 				    sizeof(struct in_addr) > optlen) {
342 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
343 					goto bad;
344 				}
345 				(void)memcpy(&ipaddr.sin_addr, sin,
346 				    sizeof(struct in_addr));
347 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
348 					continue;
349 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
350 				off += sizeof(struct in_addr);
351 				break;
352 
353 			default:
354 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
355 				goto bad;
356 			}
357 			ntime = iptime();
358 			(void)memcpy(cp + off, &ntime, sizeof(uint32_t));
359 			cp[IPOPT_OFFSET] += sizeof(uint32_t);
360 		}
361 	}
362 	if (forward && V_ipforwarding) {
363 		ip_forward(m, 1);
364 		return (1);
365 	}
366 	return (0);
367 bad:
368 	icmp_error(m, type, code, 0, 0);
369 	V_ipstat.ips_badoptions++;
370 	return (1);
371 }
372 
373 /*
374  * Save incoming source route for use in replies, to be picked up later by
375  * ip_srcroute if the receiver is interested.
376  */
377 static void
378 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
379 {
380 	unsigned olen;
381 	struct ipopt_tag *opts;
382 
383 	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
384 	    sizeof(struct ipopt_tag), M_NOWAIT);
385 	if (opts == NULL)
386 		return;
387 
388 	olen = option[IPOPT_OLEN];
389 	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) {
390 		m_tag_free((struct m_tag *)opts);
391 		return;
392 	}
393 	bcopy(option, opts->ip_srcrt.srcopt, olen);
394 	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
395 	opts->ip_srcrt.dst = dst;
396 	m_tag_prepend(m, (struct m_tag *)opts);
397 }
398 
399 /*
400  * Retrieve incoming source route for use in replies, in the same form used
401  * by setsockopt.  The first hop is placed before the options, will be
402  * removed later.
403  */
404 struct mbuf *
405 ip_srcroute(struct mbuf *m0)
406 {
407 	struct in_addr *p, *q;
408 	struct mbuf *m;
409 	struct ipopt_tag *opts;
410 
411 	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
412 	if (opts == NULL)
413 		return (NULL);
414 
415 	if (opts->ip_nhops == 0)
416 		return (NULL);
417 	m = m_get(M_DONTWAIT, MT_DATA);
418 	if (m == NULL)
419 		return (NULL);
420 
421 #define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
422 
423 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
424 	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
425 	    sizeof(struct in_addr) + OPTSIZ;
426 
427 	/*
428 	 * First, save first hop for return route.
429 	 */
430 	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
431 	*(mtod(m, struct in_addr *)) = *p--;
432 
433 	/*
434 	 * Copy option fields and padding (nop) to mbuf.
435 	 */
436 	opts->ip_srcrt.nop = IPOPT_NOP;
437 	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
438 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
439 	    &(opts->ip_srcrt.nop), OPTSIZ);
440 	q = (struct in_addr *)(mtod(m, caddr_t) +
441 	    sizeof(struct in_addr) + OPTSIZ);
442 #undef OPTSIZ
443 	/*
444 	 * Record return path as an IP source route, reversing the path
445 	 * (pointers are now aligned).
446 	 */
447 	while (p >= opts->ip_srcrt.route) {
448 		*q++ = *p--;
449 	}
450 	/*
451 	 * Last hop goes to final destination.
452 	 */
453 	*q = opts->ip_srcrt.dst;
454 	m_tag_delete(m0, (struct m_tag *)opts);
455 	return (m);
456 }
457 
458 /*
459  * Strip out IP options, at higher level protocol in the kernel.  Second
460  * argument is buffer to which options will be moved, and return value is
461  * their length.
462  *
463  * XXX should be deleted; last arg currently ignored.
464  */
465 void
466 ip_stripoptions(struct mbuf *m, struct mbuf *mopt)
467 {
468 	int i;
469 	struct ip *ip = mtod(m, struct ip *);
470 	caddr_t opts;
471 	int olen;
472 
473 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
474 	opts = (caddr_t)(ip + 1);
475 	i = m->m_len - (sizeof (struct ip) + olen);
476 	bcopy(opts + olen, opts, (unsigned)i);
477 	m->m_len -= olen;
478 	if (m->m_flags & M_PKTHDR)
479 		m->m_pkthdr.len -= olen;
480 	ip->ip_v = IPVERSION;
481 	ip->ip_hl = sizeof(struct ip) >> 2;
482 }
483 
484 /*
485  * Insert IP options into preformed packet.  Adjust IP destination as
486  * required for IP source routing, as indicated by a non-zero in_addr at the
487  * start of the options.
488  *
489  * XXX This routine assumes that the packet has no options in place.
490  */
491 struct mbuf *
492 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
493 {
494 	struct ipoption *p = mtod(opt, struct ipoption *);
495 	struct mbuf *n;
496 	struct ip *ip = mtod(m, struct ip *);
497 	unsigned optlen;
498 
499 	optlen = opt->m_len - sizeof(p->ipopt_dst);
500 	if (optlen + ip->ip_len > IP_MAXPACKET) {
501 		*phlen = 0;
502 		return (m);		/* XXX should fail */
503 	}
504 	if (p->ipopt_dst.s_addr)
505 		ip->ip_dst = p->ipopt_dst;
506 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
507 		MGETHDR(n, M_DONTWAIT, MT_DATA);
508 		if (n == NULL) {
509 			*phlen = 0;
510 			return (m);
511 		}
512 		M_MOVE_PKTHDR(n, m);
513 		n->m_pkthdr.rcvif = NULL;
514 		n->m_pkthdr.len += optlen;
515 		m->m_len -= sizeof(struct ip);
516 		m->m_data += sizeof(struct ip);
517 		n->m_next = m;
518 		m = n;
519 		m->m_len = optlen + sizeof(struct ip);
520 		m->m_data += max_linkhdr;
521 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
522 	} else {
523 		m->m_data -= optlen;
524 		m->m_len += optlen;
525 		m->m_pkthdr.len += optlen;
526 		bcopy(ip, mtod(m, void *), sizeof(struct ip));
527 	}
528 	ip = mtod(m, struct ip *);
529 	bcopy(p->ipopt_list, ip + 1, optlen);
530 	*phlen = sizeof(struct ip) + optlen;
531 	ip->ip_v = IPVERSION;
532 	ip->ip_hl = *phlen >> 2;
533 	ip->ip_len += optlen;
534 	return (m);
535 }
536 
537 /*
538  * Copy options from ip to jp, omitting those not copied during
539  * fragmentation.
540  */
541 int
542 ip_optcopy(struct ip *ip, struct ip *jp)
543 {
544 	u_char *cp, *dp;
545 	int opt, optlen, cnt;
546 
547 	cp = (u_char *)(ip + 1);
548 	dp = (u_char *)(jp + 1);
549 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
550 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
551 		opt = cp[0];
552 		if (opt == IPOPT_EOL)
553 			break;
554 		if (opt == IPOPT_NOP) {
555 			/* Preserve for IP mcast tunnel's LSRR alignment. */
556 			*dp++ = IPOPT_NOP;
557 			optlen = 1;
558 			continue;
559 		}
560 
561 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
562 		    ("ip_optcopy: malformed ipv4 option"));
563 		optlen = cp[IPOPT_OLEN];
564 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
565 		    ("ip_optcopy: malformed ipv4 option"));
566 
567 		/* Bogus lengths should have been caught by ip_dooptions. */
568 		if (optlen > cnt)
569 			optlen = cnt;
570 		if (IPOPT_COPIED(opt)) {
571 			bcopy(cp, dp, optlen);
572 			dp += optlen;
573 		}
574 	}
575 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
576 		*dp++ = IPOPT_EOL;
577 	return (optlen);
578 }
579 
580 /*
581  * Set up IP options in pcb for insertion in output packets.  Store in mbuf
582  * with pointer in pcbopt, adding pseudo-option with destination address if
583  * source routed.
584  */
585 int
586 ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m)
587 {
588 	int cnt, optlen;
589 	u_char *cp;
590 	struct mbuf **pcbopt;
591 	u_char opt;
592 
593 	INP_WLOCK_ASSERT(inp);
594 
595 	pcbopt = &inp->inp_options;
596 
597 	/* turn off any old options */
598 	if (*pcbopt)
599 		(void)m_free(*pcbopt);
600 	*pcbopt = 0;
601 	if (m == NULL || m->m_len == 0) {
602 		/*
603 		 * Only turning off any previous options.
604 		 */
605 		if (m != NULL)
606 			(void)m_free(m);
607 		return (0);
608 	}
609 
610 	if (m->m_len % sizeof(int32_t))
611 		goto bad;
612 	/*
613 	 * IP first-hop destination address will be stored before actual
614 	 * options; move other options back and clear it when none present.
615 	 */
616 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
617 		goto bad;
618 	cnt = m->m_len;
619 	m->m_len += sizeof(struct in_addr);
620 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
621 	bcopy(mtod(m, void *), cp, (unsigned)cnt);
622 	bzero(mtod(m, void *), sizeof(struct in_addr));
623 
624 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
625 		opt = cp[IPOPT_OPTVAL];
626 		if (opt == IPOPT_EOL)
627 			break;
628 		if (opt == IPOPT_NOP)
629 			optlen = 1;
630 		else {
631 			if (cnt < IPOPT_OLEN + sizeof(*cp))
632 				goto bad;
633 			optlen = cp[IPOPT_OLEN];
634 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
635 				goto bad;
636 		}
637 		switch (opt) {
638 
639 		default:
640 			break;
641 
642 		case IPOPT_LSRR:
643 		case IPOPT_SSRR:
644 			/*
645 			 * User process specifies route as:
646 			 *
647 			 *	->A->B->C->D
648 			 *
649 			 * D must be our final destination (but we can't
650 			 * check that since we may not have connected yet).
651 			 * A is first hop destination, which doesn't appear
652 			 * in actual IP option, but is stored before the
653 			 * options.
654 			 */
655 			/* XXX-BZ PRIV_NETINET_SETHDROPTS? */
656 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
657 				goto bad;
658 			m->m_len -= sizeof(struct in_addr);
659 			cnt -= sizeof(struct in_addr);
660 			optlen -= sizeof(struct in_addr);
661 			cp[IPOPT_OLEN] = optlen;
662 			/*
663 			 * Move first hop before start of options.
664 			 */
665 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
666 			    sizeof(struct in_addr));
667 			/*
668 			 * Then copy rest of options back
669 			 * to close up the deleted entry.
670 			 */
671 			bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
672 			    &cp[IPOPT_OFFSET+1],
673 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
674 			break;
675 		}
676 	}
677 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
678 		goto bad;
679 	*pcbopt = m;
680 	return (0);
681 
682 bad:
683 	(void)m_free(m);
684 	return (EINVAL);
685 }
686