xref: /xnu-10063.121.3/bsd/netinet/ip_icmp.c (revision 2c2f96dc2b9a4408a43d3150ae9c105355ca3daa)
1 /*
2  * Copyright (c) 2000-2021 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * Copyright (c) 1982, 1986, 1988, 1993
30  *	The Regents of the University of California.  All rights reserved.
31  *
32  * Redistribution and use in source and binary forms, with or without
33  * modification, are permitted provided that the following conditions
34  * are met:
35  * 1. Redistributions of source code must retain the above copyright
36  *    notice, this list of conditions and the following disclaimer.
37  * 2. Redistributions in binary form must reproduce the above copyright
38  *    notice, this list of conditions and the following disclaimer in the
39  *    documentation and/or other materials provided with the distribution.
40  * 3. All advertising materials mentioning features or use of this software
41  *    must display the following acknowledgement:
42  *	This product includes software developed by the University of
43  *	California, Berkeley and its contributors.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
61  */
62 /*
63  * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
64  * support for mandatory and extensible security protections.  This notice
65  * is included in support of clause 2.2 (b) of the Apple Public License,
66  * Version 2.0.
67  */
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/mbuf.h>
72 #include <sys/mcache.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/sysctl.h>
78 
79 #include <machine/endian.h>
80 
81 #include <net/if.h>
82 #include <net/route.h>
83 #include <net/content_filter.h>
84 
85 #define _IP_VHL
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip.h>
90 #include <netinet/ip_icmp.h>
91 #include <netinet/ip_var.h>
92 #include <netinet/icmp_var.h>
93 #include <netinet/tcp.h>
94 #include <netinet/tcp_fsm.h>
95 #include <netinet/tcp_seq.h>
96 #include <netinet/tcp_timer.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet/tcpip.h>
99 
100 #if IPSEC
101 #include <netinet6/ipsec.h>
102 #include <netkey/key.h>
103 #endif
104 
105 #if NECP
106 #include <net/necp.h>
107 #endif /* NECP */
108 
109 #include <net/sockaddr_utils.h>
110 
111 /*
112  * ICMP routines: error generation, receive packet processing, and
113  * routines to turnaround packets back to the originator, and
114  * host table maintenance routines.
115  */
116 
117 struct  icmpstat icmpstat;
118 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats,
119     CTLFLAG_RD | CTLFLAG_LOCKED,
120     &icmpstat, icmpstat, "");
121 
122 static int      icmpmaskrepl = 0;
123 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
124     CTLFLAG_RW | CTLFLAG_LOCKED,
125     &icmpmaskrepl, 0, "");
126 
127 static int      icmptimestamp = 0;
128 SYSCTL_INT(_net_inet_icmp, ICMPCTL_TIMESTAMP, timestamp,
129     CTLFLAG_RW | CTLFLAG_LOCKED,
130     &icmptimestamp, 0, "");
131 
132 static int      drop_redirect = 1;
133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect,
134     CTLFLAG_RW | CTLFLAG_LOCKED,
135     &drop_redirect, 0, "");
136 
137 static int      log_redirect = 0;
138 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect,
139     CTLFLAG_RW | CTLFLAG_LOCKED,
140     &log_redirect, 0, "");
141 
142 const static int icmp_datalen = 8;
143 /*
144  * ICMP broadcast echo sysctl
145  */
146 static int      icmpbmcastecho = 1;
147 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW | CTLFLAG_LOCKED,
148     &icmpbmcastecho, 0, "");
149 
150 #if (DEBUG | DEVELOPMENT)
151 static int      icmpprintfs = 0;
152 SYSCTL_INT(_net_inet_icmp, OID_AUTO, verbose, CTLFLAG_RW | CTLFLAG_LOCKED,
153     &icmpprintfs, 0, "");
154 #endif
155 
156 static void     icmp_reflect(struct mbuf *);
157 static void     icmp_send(struct mbuf *, struct mbuf *);
158 
159 /*
160  * Generate packet gencount for ICMP for a given error type
161  * and code.
162  * We do it this way to ensure we only dedup the packets that belong
163  * to the same type, which is usually what port scanning and other such
164  * attack vectors depend on.
165  */
166 static uint32_t
icmp_error_packet_gencount(int type,int code)167 icmp_error_packet_gencount(int type, int code)
168 {
169 	return (PF_INET << 24) | (type << 16) | (code << 8);
170 }
171 
172 static int      suppress_icmp_port_unreach = 0;
173 SYSCTL_INT(_net_inet_icmp, OID_AUTO, suppress_icmp_port_unreach,
174     CTLFLAG_RW | CTLFLAG_LOCKED,
175     &suppress_icmp_port_unreach, 0,
176     "Suppress ICMP destination unreachable type with code port unreachable");
177 
178 /*
179  * Generate an error packet of type error
180  * in response to bad packet ip.
181  */
182 void
icmp_error(struct mbuf * n,int type,int code,u_int32_t dest,u_int32_t nextmtu)183 icmp_error(
184 	struct mbuf *n,
185 	int type,
186 	int code,
187 	u_int32_t dest,
188 	u_int32_t nextmtu)
189 {
190 	struct ip *oip = NULL;
191 	struct ip *nip = NULL;
192 	struct icmp *icp = NULL;
193 	struct mbuf *m = NULL;
194 	u_int32_t oiphlen = 0;
195 	u_int32_t icmplen = 0;
196 	u_int32_t icmpelen = 0;
197 	u_int32_t nlen = 0;
198 
199 	VERIFY((u_int)type <= ICMP_MAXTYPE);
200 	VERIFY(code <= UINT8_MAX);
201 
202 	/* Expect 32-bit aligned data pointer on strict-align platforms */
203 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(n);
204 
205 	if (type != ICMP_REDIRECT) {
206 		icmpstat.icps_error++;
207 	}
208 
209 	if (suppress_icmp_port_unreach &&
210 	    type == ICMP_UNREACH && code == ICMP_UNREACH_PORT) {
211 		goto freeit;
212 	}
213 	/*
214 	 * Don't send error:
215 	 *   if not the first fragment of message
216 	 *   if original packet was a multicast or broadcast packet
217 	 *   if the old packet protocol was ICMP
218 	 *   error message, only known informational types.
219 	 */
220 	if (n->m_flags & (M_BCAST | M_MCAST)) {
221 		goto freeit;
222 	}
223 
224 	/*
225 	 * Drop if IP header plus ICMP_MINLEN bytes are not contiguous
226 	 * in first mbuf.
227 	 */
228 	if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) {
229 		goto freeit;
230 	}
231 
232 	oip = mtod(n, struct ip *);
233 	oiphlen = IP_VHL_HL(oip->ip_vhl) << 2;
234 	if (n->m_len < oiphlen + ICMP_MINLEN) {
235 		goto freeit;
236 	}
237 
238 #if (DEBUG | DEVELOPMENT)
239 	if (icmpprintfs > 1) {
240 		printf("icmp_error(0x%llx, %x, %d)\n",
241 		    (uint64_t)VM_KERNEL_ADDRPERM(oip), type, code);
242 	}
243 #endif
244 
245 	if (oip->ip_off & ~(IP_MF | IP_DF)) {
246 		goto freeit;
247 	}
248 
249 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
250 	    n->m_len >= oiphlen + ICMP_MINLEN &&
251 	    !ICMP_INFOTYPE(((struct icmp *)(void *)((caddr_t)oip + oiphlen))->
252 	    icmp_type)) {
253 		icmpstat.icps_oldicmp++;
254 		goto freeit;
255 	}
256 
257 	/*
258 	 * Calculate the length to quote from original packet and prevent
259 	 * the ICMP mbuf from overflowing.
260 	 * Unfortunatly this is non-trivial since ip_forward()
261 	 * sends us truncated packets.
262 	 */
263 	nlen = m_length(n);
264 	if (oip->ip_p == IPPROTO_TCP) {
265 		struct tcphdr *th = NULL;
266 		u_int16_t tcphlen = 0;
267 
268 		/*
269 		 * If the packet got truncated and TCP header
270 		 * is not contained in the packet, send out
271 		 * standard reply with only IP header as payload
272 		 */
273 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
274 		    n->m_next == NULL) {
275 			goto stdreply;
276 		}
277 
278 		/*
279 		 * Otherwise, pull up to get IP and TCP headers
280 		 * together
281 		 */
282 		if (n->m_len < (oiphlen + sizeof(struct tcphdr)) &&
283 		    (n = m_pullup(n, (oiphlen + sizeof(struct tcphdr)))) == NULL) {
284 			goto freeit;
285 		}
286 
287 		/*
288 		 * Reinit pointers derived from mbuf data pointer
289 		 * as things might have moved around with m_pullup
290 		 */
291 		oip = mtod(n, struct ip *);
292 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
293 
294 		if (th != ((struct tcphdr *)P2ROUNDDOWN(th,
295 		    sizeof(u_int32_t))) ||
296 		    ((th->th_off << 2) > UINT16_MAX)) {
297 			goto freeit;
298 		}
299 		tcphlen = (uint16_t)(th->th_off << 2);
300 
301 		/* Sanity checks */
302 		if (tcphlen < sizeof(struct tcphdr)) {
303 			goto freeit;
304 		}
305 		if (oip->ip_len < (oiphlen + tcphlen)) {
306 			goto freeit;
307 		}
308 		if ((oiphlen + tcphlen) > n->m_len && n->m_next == NULL) {
309 			goto stdreply;
310 		}
311 		if (n->m_len < (oiphlen + tcphlen) &&
312 		    (n = m_pullup(n, (oiphlen + tcphlen))) == NULL) {
313 			goto freeit;
314 		}
315 
316 		/*
317 		 * Reinit pointers derived from mbuf data pointer
318 		 * as things might have moved around with m_pullup
319 		 */
320 		oip = mtod(n, struct ip *);
321 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
322 
323 		icmpelen = max(tcphlen, min(icmp_datalen,
324 		    (oip->ip_len - oiphlen)));
325 	} else {
326 stdreply:       icmpelen = max(ICMP_MINLEN, min(icmp_datalen,
327 		    (oip->ip_len - oiphlen)));
328 	}
329 
330 	icmplen = min(oiphlen + icmpelen, nlen);
331 	if (icmplen < sizeof(struct ip)) {
332 		goto freeit;
333 	}
334 
335 	/*
336 	 * First, formulate icmp message
337 	 * Allocate enough space for the IP header, ICMP header
338 	 * and the payload (part of the original message to be sent back).
339 	 */
340 	if (MHLEN > (sizeof(struct ip) + ICMP_MINLEN + icmplen)) {
341 		m = m_gethdr(M_DONTWAIT, MT_HEADER);    /* MAC-OK */
342 	} else {
343 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
344 	}
345 
346 	if (m == NULL) {
347 		goto freeit;
348 	}
349 
350 	/*
351 	 * Further refine the payload length to the space
352 	 * remaining in mbuf after including the IP header and ICMP
353 	 * header.
354 	 */
355 	icmplen = min(icmplen, (u_int)M_TRAILINGSPACE(m) -
356 	    (u_int)(sizeof(struct ip) - ICMP_MINLEN));
357 	m_align(m, ICMP_MINLEN + icmplen);
358 	m->m_len = ICMP_MINLEN + icmplen; /* for ICMP header and data */
359 
360 	icp = mtod(m, struct icmp *);
361 	icmpstat.icps_outhist[type]++;
362 	icp->icmp_type = (u_char)type;
363 	if (type == ICMP_REDIRECT) {
364 		icp->icmp_gwaddr.s_addr = dest;
365 	} else {
366 		icp->icmp_void = 0;
367 		/*
368 		 * The following assignments assume an overlay with the
369 		 * zeroed icmp_void field.
370 		 */
371 		if (type == ICMP_PARAMPROB) {
372 			icp->icmp_pptr = (u_char)code;
373 			code = 0;
374 		} else if (type == ICMP_UNREACH &&
375 		    code == ICMP_UNREACH_NEEDFRAG && nextmtu != 0) {
376 			icp->icmp_nextmtu = htons((uint16_t)nextmtu);
377 		}
378 	}
379 
380 	icp->icmp_code = (u_char)code;
381 
382 	/*
383 	 * Copy icmplen worth of content from original
384 	 * mbuf (n) to the new packet after ICMP header.
385 	 */
386 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
387 	nip = &icp->icmp_ip;
388 
389 	/*
390 	 * Convert fields to network representation.
391 	 */
392 #if BYTE_ORDER != BIG_ENDIAN
393 	HTONS(nip->ip_len);
394 	HTONS(nip->ip_off);
395 #endif
396 	/*
397 	 * Set up ICMP message mbuf and copy old IP header (without options
398 	 * in front of ICMP message.
399 	 */
400 	m->m_data -= sizeof(struct ip);
401 	m->m_len += sizeof(struct ip);
402 	m->m_pkthdr.len = m->m_len;
403 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
404 	/*
405 	 * To avoid some flavors of port scanning and other attacks,
406 	 * use packet suppression without using any other sort of
407 	 * rate limiting with static bounds.
408 	 * XXX Not setting PKTF_FLOW_ID here because we were concerned
409 	 * about it triggering regression elsewhere outside of network stack
410 	 * where there might be an assumption around flow ID being non-zero.
411 	 * It should be noted though that previously if PKTF_FLOW_ID was not
412 	 * set, PF would have generated flow hash irrespective of ICMPv4/v6
413 	 * type. That doesn't happen now and PF only computes hash for ICMP
414 	 * types that need state creation (which is not true of error types).
415 	 * It would have been a problem because we really want all the ICMP
416 	 * error type packets to share the same flow ID for global suppression.
417 	 */
418 	m->m_pkthdr.comp_gencnt = icmp_error_packet_gencount(type, code);
419 
420 	nip = mtod(m, struct ip *);
421 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
422 	nip->ip_len = (uint16_t)m->m_len;
423 	nip->ip_vhl = IP_VHL_BORING;
424 	nip->ip_p = IPPROTO_ICMP;
425 	nip->ip_tos = 0;
426 	nip->ip_off = 0;
427 	icmp_reflect(m);
428 freeit:
429 	m_freem(n);
430 }
431 
432 /*
433  * Process a received ICMP message.
434  */
435 void
icmp_input(struct mbuf * m,int hlen)436 icmp_input(struct mbuf *m, int hlen)
437 {
438 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
439 	struct icmp *icp;
440 	struct ip *ip = mtod(m, struct ip *);
441 	int icmplen;
442 	int i;
443 	struct in_ifaddr *ia;
444 	void (*ctlfunc)(int, struct sockaddr *, void *, struct ifnet *);
445 	int code;
446 	boolean_t should_log_redirect = false;
447 
448 	/* Expect 32-bit aligned data pointer on strict-align platforms */
449 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
450 
451 	icmplen = ip->ip_len;
452 
453 	/*
454 	 * Locate icmp structure in mbuf, and check
455 	 * that not corrupted and of at least minimum length.
456 	 */
457 #if (DEBUG | DEVELOPMENT)
458 	if (icmpprintfs > 2) {
459 		char src_str[MAX_IPv4_STR_LEN];
460 		char dst_str[MAX_IPv4_STR_LEN];
461 
462 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
463 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
464 		printf("%s: from %s to %s, len %d\n",
465 		    __func__, src_str, dst_str, icmplen);
466 	}
467 #endif
468 	if (icmplen < ICMP_MINLEN) {
469 		icmpstat.icps_tooshort++;
470 		goto freeit;
471 	}
472 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
473 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
474 		icmpstat.icps_tooshort++;
475 		return;
476 	}
477 	/* Re-seat the pointers, since `m_pullup' might have moved `m'. `icp' is re-seated below. */
478 	ip = mtod(m, struct ip *);
479 
480 	m->m_len -= hlen;
481 	m->m_data += hlen;
482 	icp = mtod(m, struct icmp *);
483 	if (in_cksum(m, icmplen) != 0) {
484 		icmpstat.icps_checksum++;
485 		goto freeit;
486 	}
487 	m->m_len += hlen;
488 	m->m_data -= hlen;
489 
490 #if (DEBUG | DEVELOPMENT)
491 	if (icmpprintfs > 2) {
492 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
493 		    icp->icmp_code);
494 	}
495 #endif
496 
497 	/*
498 	 * Message type specific processing.
499 	 */
500 	if (icp->icmp_type > ICMP_MAXTYPE) {
501 		goto raw;
502 	}
503 
504 	/* Initialize */
505 	SOCKADDR_ZERO(&icmpsrc, sizeof(icmpsrc));
506 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
507 	icmpsrc.sin_family = AF_INET;
508 	SOCKADDR_ZERO(&icmpdst, sizeof(icmpdst));
509 	icmpdst.sin_len = sizeof(struct sockaddr_in);
510 	icmpdst.sin_family = AF_INET;
511 	SOCKADDR_ZERO(&icmpgw, sizeof(icmpgw));
512 	icmpgw.sin_len = sizeof(struct sockaddr_in);
513 	icmpgw.sin_family = AF_INET;
514 
515 	icmpstat.icps_inhist[icp->icmp_type]++;
516 	code = icp->icmp_code;
517 	switch (icp->icmp_type) {
518 	case ICMP_UNREACH:
519 		switch (code) {
520 		case ICMP_UNREACH_NET:
521 		case ICMP_UNREACH_HOST:
522 		case ICMP_UNREACH_SRCFAIL:
523 		case ICMP_UNREACH_NET_UNKNOWN:
524 		case ICMP_UNREACH_HOST_UNKNOWN:
525 		case ICMP_UNREACH_ISOLATED:
526 		case ICMP_UNREACH_TOSNET:
527 		case ICMP_UNREACH_TOSHOST:
528 		case ICMP_UNREACH_HOST_PRECEDENCE:
529 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
530 			code = PRC_UNREACH_NET;
531 			break;
532 
533 		case ICMP_UNREACH_NEEDFRAG:
534 			code = PRC_MSGSIZE;
535 			break;
536 
537 		/*
538 		 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
539 		 * Treat subcodes 2,3 as immediate RST
540 		 */
541 		case ICMP_UNREACH_PROTOCOL:
542 		case ICMP_UNREACH_PORT:
543 			code = PRC_UNREACH_PORT;
544 			break;
545 
546 		case ICMP_UNREACH_NET_PROHIB:
547 		case ICMP_UNREACH_HOST_PROHIB:
548 		case ICMP_UNREACH_FILTER_PROHIB:
549 			code = PRC_UNREACH_ADMIN_PROHIB;
550 			break;
551 
552 		default:
553 			goto badcode;
554 		}
555 		goto deliver;
556 
557 	case ICMP_TIMXCEED:
558 		if (code > 1) {
559 			goto badcode;
560 		}
561 		code += PRC_TIMXCEED_INTRANS;
562 		goto deliver;
563 
564 	case ICMP_PARAMPROB:
565 		if (code > 1) {
566 			goto badcode;
567 		}
568 		code = PRC_PARAMPROB;
569 		goto deliver;
570 
571 	case ICMP_SOURCEQUENCH:
572 		if (code) {
573 			goto badcode;
574 		}
575 		code = PRC_QUENCH;
576 deliver:
577 		/*
578 		 * Problem with datagram; advise higher level routines.
579 		 */
580 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp)
581 		    || IP_VHL_HL(icp->icmp_ip.ip_vhl) <
582 		    (sizeof(struct ip) >> 2) ||
583 		    (m = m_pullup(m, hlen + ICMP_ADVLEN(icp))) == NULL) {
584 			icmpstat.icps_badlen++;
585 			goto freeit;
586 		}
587 
588 		/* Re-seat the pointers, since `m_pullup' might have moved `m'*/
589 		ip = mtod(m, struct ip *);
590 		icp = (struct icmp *)(void *)(mtod(m, uint8_t *) + hlen);
591 
592 #if BYTE_ORDER != BIG_ENDIAN
593 		NTOHS(icp->icmp_ip.ip_len);
594 #endif
595 
596 		/* Discard ICMP's in response to multicast packets */
597 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) {
598 			goto badcode;
599 		}
600 #if (DEBUG | DEVELOPMENT)
601 		if (icmpprintfs > 2) {
602 			printf("deliver to protocol %d\n",
603 			    icp->icmp_ip.ip_p);
604 		}
605 #endif
606 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
607 
608 		/*
609 		 * if the packet contains [IPv4 AH TCP], we can't make a
610 		 * notification to TCP layer.
611 		 */
612 		ctlfunc = ip_protox[icp->icmp_ip.ip_p]->pr_ctlinput;
613 
614 		if (ctlfunc) {
615 			struct ipctlparam ctl_param = {
616 				.ipc_m = m,
617 				.ipc_icmp = icp,
618 				.ipc_icmp_ip = &icp->icmp_ip,
619 				.ipc_off = hlen + offsetof(struct icmp, icmp_ip) + (IP_VHL_HL(icp->icmp_ip.ip_vhl) << 2)
620 			};
621 			LCK_MTX_ASSERT(inet_domain_mutex, LCK_MTX_ASSERT_OWNED);
622 
623 			lck_mtx_unlock(inet_domain_mutex);
624 
625 			(*ctlfunc)(code, SA(&icmpsrc),
626 			    (void *)&ctl_param, m->m_pkthdr.rcvif);
627 
628 			lck_mtx_lock(inet_domain_mutex);
629 		}
630 		break;
631 
632 badcode:
633 		icmpstat.icps_badcode++;
634 		break;
635 
636 	case ICMP_ECHO:
637 		if ((m->m_flags & (M_MCAST | M_BCAST))) {
638 			if (icmpbmcastecho == 0) {
639 				icmpstat.icps_bmcastecho++;
640 				break;
641 			}
642 		}
643 
644 		/*
645 		 * rdar://18644769
646 		 * Do not reply when the destination is link local multicast or broadcast
647 		 * and the source is not from a directly connected subnet
648 		 */
649 		if ((IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr)) ||
650 		    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) &&
651 		    in_localaddr(ip->ip_src) == 0) {
652 			icmpstat.icps_bmcastecho++;
653 #if (DEBUG | DEVELOPMENT)
654 			if (icmpprintfs > 0) {
655 				char src_str[MAX_IPv4_STR_LEN];
656 				char dst_str[MAX_IPv4_STR_LEN];
657 
658 				inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
659 				inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
660 				printf("%s: non local (B|M)CAST %s to %s, len %d\n",
661 				    __func__, src_str, dst_str, icmplen);
662 			}
663 #endif
664 			break;
665 		}
666 
667 		icp->icmp_type = ICMP_ECHOREPLY;
668 		goto reflect;
669 
670 	case ICMP_TSTAMP:
671 		if (icmptimestamp == 0) {
672 			break;
673 		}
674 
675 		if (!icmpbmcastecho
676 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
677 			icmpstat.icps_bmcasttstamp++;
678 			break;
679 		}
680 		if (icmplen < ICMP_TSLEN) {
681 			icmpstat.icps_badlen++;
682 			break;
683 		}
684 		icp->icmp_type = ICMP_TSTAMPREPLY;
685 		icp->icmp_rtime = iptime();
686 		icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
687 		goto reflect;
688 
689 	case ICMP_MASKREQ:
690 		if (icmpmaskrepl == 0) {
691 			break;
692 		}
693 		/*
694 		 * We are not able to respond with all ones broadcast
695 		 * unless we receive it over a point-to-point interface.
696 		 */
697 		if (icmplen < ICMP_MASKLEN) {
698 			break;
699 		}
700 		switch (ip->ip_dst.s_addr) {
701 		case INADDR_BROADCAST:
702 		case INADDR_ANY:
703 			icmpdst.sin_addr = ip->ip_src;
704 			break;
705 
706 		default:
707 			icmpdst.sin_addr = ip->ip_dst;
708 		}
709 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
710 			SA(&icmpdst), m->m_pkthdr.rcvif);
711 		if (ia == 0) {
712 			break;
713 		}
714 		IFA_LOCK(&ia->ia_ifa);
715 		if (ia->ia_ifp == 0) {
716 			IFA_UNLOCK(&ia->ia_ifa);
717 			ifa_remref(&ia->ia_ifa);
718 			ia = NULL;
719 			break;
720 		}
721 		icp->icmp_type = ICMP_MASKREPLY;
722 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
723 		if (ip->ip_src.s_addr == 0) {
724 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
725 				ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
726 			} else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
727 				ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
728 			}
729 		}
730 		IFA_UNLOCK(&ia->ia_ifa);
731 		ifa_remref(&ia->ia_ifa);
732 reflect:
733 		ip->ip_len += hlen;     /* since ip_input deducts this */
734 		icmpstat.icps_reflect++;
735 		icmpstat.icps_outhist[icp->icmp_type]++;
736 		icmp_reflect(m);
737 		return;
738 
739 	case ICMP_REDIRECT:
740 		if (drop_redirect) {
741 			break;
742 		}
743 		if (code > 3) {
744 			goto badcode;
745 		}
746 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
747 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
748 			icmpstat.icps_badlen++;
749 			break;
750 		}
751 
752 #if (DEBUG | DEVELOPMENT)
753 		should_log_redirect = log_redirect || (icmpprintfs > 0);
754 #else
755 		should_log_redirect = log_redirect;
756 #endif
757 		/*
758 		 * Short circuit routing redirects to force
759 		 * immediate change in the kernel's routing
760 		 * tables.  The message is also handed to anyone
761 		 * listening on a raw socket (e.g. the routing
762 		 * daemon for use in updating its tables).
763 		 */
764 		icmpgw.sin_addr = ip->ip_src;
765 		icmpdst.sin_addr = icp->icmp_gwaddr;
766 
767 		if (should_log_redirect) {
768 			char src_str[MAX_IPv4_STR_LEN];
769 			char dst_str[MAX_IPv4_STR_LEN];
770 			char gw_str[MAX_IPv4_STR_LEN];
771 
772 			inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
773 			inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, dst_str, sizeof(dst_str));
774 			inet_ntop(AF_INET, &icp->icmp_gwaddr, gw_str, sizeof(gw_str));
775 			printf("%s: redirect dst %s to %s from %s\n", __func__,
776 			    dst_str, gw_str, src_str);
777 		}
778 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
779 		rtredirect(m->m_pkthdr.rcvif, SA(&icmpsrc),
780 		    SA(&icmpdst), NULL, RTF_GATEWAY | RTF_HOST,
781 		    SA(&icmpgw), NULL);
782 		pfctlinput(PRC_REDIRECT_HOST, SA(&icmpsrc));
783 #if IPSEC
784 		key_sa_routechange(SA(&icmpsrc));
785 #endif
786 		break;
787 
788 	/*
789 	 * No kernel processing for the following;
790 	 * just fall through to send to raw listener.
791 	 */
792 	case ICMP_ECHOREPLY:
793 	case ICMP_ROUTERADVERT:
794 	case ICMP_ROUTERSOLICIT:
795 	case ICMP_TSTAMPREPLY:
796 	case ICMP_IREQREPLY:
797 	case ICMP_MASKREPLY:
798 	default:
799 		break;
800 	}
801 
802 raw:
803 	rip_input(m, hlen);
804 	return;
805 
806 freeit:
807 	m_freem(m);
808 }
809 
810 /*
811  * Reflect the ip packet back to the source
812  */
813 static void
icmp_reflect(struct mbuf * m)814 icmp_reflect(struct mbuf *m)
815 {
816 	struct ip *ip = mtod(m, struct ip *);
817 	struct sockaddr_in icmpdst;
818 	struct in_ifaddr *ia;
819 	struct in_addr t;
820 	struct mbuf *opts = NULL;
821 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
822 
823 	if (!in_canforward(ip->ip_src) &&
824 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
825 	    (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
826 		m_freem(m);     /* Bad return address */
827 		goto done;      /* Ip_output() will check for broadcast */
828 	}
829 	t = ip->ip_dst;
830 	ip->ip_dst = ip->ip_src;
831 	/*
832 	 * If the incoming packet was addressed directly to us,
833 	 * use dst as the src for the reply.  Otherwise (broadcast
834 	 * or anonymous), use the address which corresponds
835 	 * to the incoming interface.
836 	 */
837 	lck_rw_lock_shared(&in_ifaddr_rwlock);
838 	TAILQ_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
839 		IFA_LOCK(&ia->ia_ifa);
840 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
841 			ifa_addref(&ia->ia_ifa);
842 			IFA_UNLOCK(&ia->ia_ifa);
843 			goto match;
844 		}
845 		IFA_UNLOCK(&ia->ia_ifa);
846 	}
847 	/*
848 	 * Slow path; check for broadcast addresses.  Find a source
849 	 * IP address to use when replying to the broadcast request;
850 	 * let IP handle the source interface selection work.
851 	 */
852 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
853 		IFA_LOCK(&ia->ia_ifa);
854 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
855 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr) {
856 			ifa_addref(&ia->ia_ifa);
857 			IFA_UNLOCK(&ia->ia_ifa);
858 			break;
859 		}
860 		IFA_UNLOCK(&ia->ia_ifa);
861 	}
862 match:
863 	lck_rw_done(&in_ifaddr_rwlock);
864 
865 	/* Initialize */
866 	SOCKADDR_ZERO(&icmpdst, sizeof(icmpdst));
867 	icmpdst.sin_len = sizeof(struct sockaddr_in);
868 	icmpdst.sin_family = AF_INET;
869 	icmpdst.sin_addr = t;
870 	if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif) {
871 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
872 			SA(&icmpdst), m->m_pkthdr.rcvif);
873 	}
874 	/*
875 	 * The following happens if the packet was not addressed to us,
876 	 * and was received on an interface with no IP address.
877 	 */
878 	if (ia == (struct in_ifaddr *)0) {
879 		lck_rw_lock_shared(&in_ifaddr_rwlock);
880 		ia = in_ifaddrhead.tqh_first;
881 		if (ia == (struct in_ifaddr *)0) {/* no address yet, bail out */
882 			lck_rw_done(&in_ifaddr_rwlock);
883 			m_freem(m);
884 			goto done;
885 		}
886 		ifa_addref(&ia->ia_ifa);
887 		lck_rw_done(&in_ifaddr_rwlock);
888 	}
889 	IFA_LOCK_SPIN(&ia->ia_ifa);
890 	t = IA_SIN(ia)->sin_addr;
891 	IFA_UNLOCK(&ia->ia_ifa);
892 	ip->ip_src = t;
893 	ip->ip_ttl = (u_char)ip_defttl;
894 	ifa_remref(&ia->ia_ifa);
895 	ia = NULL;
896 
897 	if (optlen > 0) {
898 		u_char *cp;
899 		int opt, cnt;
900 		u_int len;
901 
902 		/*
903 		 * Retrieve any source routing from the incoming packet;
904 		 * add on any record-route or timestamp options.
905 		 */
906 		cp = (u_char *) (ip + 1);
907 		if ((opts = ip_srcroute()) == 0 &&
908 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { /* MAC-OK */
909 			opts->m_len = sizeof(struct in_addr);
910 			mtod(opts, struct in_addr *)->s_addr = 0;
911 		}
912 		if (opts) {
913 #if (DEBUG | DEVELOPMENT)
914 			if (icmpprintfs > 1) {
915 				printf("icmp_reflect optlen %d rt %d => ",
916 				    optlen, opts->m_len);
917 			}
918 #endif
919 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
920 				opt = cp[IPOPT_OPTVAL];
921 				if (opt == IPOPT_EOL) {
922 					break;
923 				}
924 				if (opt == IPOPT_NOP) {
925 					len = 1;
926 				} else {
927 					if (cnt < IPOPT_OLEN + sizeof(*cp)) {
928 						break;
929 					}
930 					len = cp[IPOPT_OLEN];
931 					if (len < IPOPT_OLEN + sizeof(*cp) ||
932 					    len > cnt) {
933 						break;
934 					}
935 				}
936 				/*
937 				 * Should check for overflow, but it "can't happen"
938 				 */
939 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
940 				    opt == IPOPT_SECURITY) {
941 					bcopy((caddr_t)cp,
942 					    mtod(opts, caddr_t) + opts->m_len, len);
943 					opts->m_len += len;
944 				}
945 			}
946 			/* Terminate & pad, if necessary */
947 			cnt = opts->m_len % 4;
948 			if (cnt) {
949 				for (; cnt < 4; cnt++) {
950 					*(mtod(opts, caddr_t) + opts->m_len) =
951 					    IPOPT_EOL;
952 					opts->m_len++;
953 				}
954 			}
955 #if (DEBUG | DEVELOPMENT)
956 			if (icmpprintfs > 1) {
957 				printf("%d\n", opts->m_len);
958 			}
959 #endif
960 		}
961 		/*
962 		 * Now strip out original options by copying rest of first
963 		 * mbuf's data back, and adjust the IP length.
964 		 */
965 		ip->ip_len -= optlen;
966 		ip->ip_vhl = IP_VHL_BORING;
967 		m->m_len -= optlen;
968 		if (m->m_flags & M_PKTHDR) {
969 			m->m_pkthdr.len -= optlen;
970 		}
971 		optlen += sizeof(struct ip);
972 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
973 		    (unsigned)(m->m_len - sizeof(struct ip)));
974 	}
975 	m->m_flags &= ~(M_BCAST | M_MCAST);
976 	icmp_send(m, opts);
977 done:
978 	if (opts) {
979 		(void)m_free(opts);
980 	}
981 }
982 
983 /*
984  * Send an icmp packet back to the ip level,
985  * after supplying a checksum.
986  */
987 static void
icmp_send(struct mbuf * m,struct mbuf * opts)988 icmp_send(struct mbuf *m, struct mbuf *opts)
989 {
990 	struct ip *ip = mtod(m, struct ip *);
991 	int hlen;
992 	struct icmp *icp;
993 	struct route ro;
994 	struct ip_out_args ipoa;
995 
996 	bzero(&ipoa, sizeof(ipoa));
997 	ipoa.ipoa_boundif = IFSCOPE_NONE;
998 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR;
999 	ipoa.ipoa_sotc = SO_TC_UNSPEC;
1000 	ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1001 
1002 	if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) && m->m_pkthdr.rcvif != NULL) {
1003 		ipoa.ipoa_boundif = m->m_pkthdr.rcvif->if_index;
1004 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
1005 	}
1006 
1007 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1008 	m->m_data += hlen;
1009 	m->m_len -= hlen;
1010 	icp = mtod(m, struct icmp *);
1011 	icp->icmp_cksum = 0;
1012 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
1013 	m->m_data -= hlen;
1014 	m->m_len += hlen;
1015 	m->m_pkthdr.rcvif = NULL;
1016 	m->m_pkthdr.csum_data = 0;
1017 	m->m_pkthdr.csum_flags = 0;
1018 #if (DEBUG | DEVELOPMENT)
1019 	if (icmpprintfs > 2) {
1020 		char src_str[MAX_IPv4_STR_LEN];
1021 		char dst_str[MAX_IPv4_STR_LEN];
1022 
1023 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
1024 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
1025 		printf("%s: dst %s src %s\n", __func__, dst_str, src_str);
1026 	}
1027 #endif
1028 	bzero(&ro, sizeof ro);
1029 	(void) ip_output(m, opts, &ro, IP_OUTARGS, NULL, &ipoa);
1030 	ROUTE_RELEASE(&ro);
1031 }
1032 
1033 u_int32_t
iptime(void)1034 iptime(void)
1035 {
1036 	struct timeval atv;
1037 	u_int32_t t;
1038 
1039 	getmicrotime(&atv);
1040 	t = (atv.tv_sec % (24 * 60 * 60)) * 1000 + atv.tv_usec / 1000;
1041 	return htonl(t);
1042 }
1043 
1044 #if 1
1045 /*
1046  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1047  * given current value MTU.  If DIR is less than zero, a larger plateau
1048  * is returned; otherwise, a smaller value is returned.
1049  */
1050 int
ip_next_mtu(int mtu,int dir)1051 ip_next_mtu(int mtu, int dir)
1052 {
1053 	static int mtutab[] = {
1054 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
1055 		68, 0
1056 	};
1057 	int i;
1058 
1059 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
1060 		if (mtu >= mtutab[i]) {
1061 			break;
1062 		}
1063 	}
1064 
1065 	if (dir < 0) {
1066 		if (i == 0) {
1067 			return 0;
1068 		} else {
1069 			return mtutab[i - 1];
1070 		}
1071 	} else {
1072 		if (mtutab[i] == 0) {
1073 			return 0;
1074 		} else if (mtu > mtutab[i]) {
1075 			return mtutab[i];
1076 		} else {
1077 			return mtutab[i + 1];
1078 		}
1079 	}
1080 }
1081 #endif
1082 
1083 #if __APPLE__
1084 
1085 /*
1086  * Non-privileged ICMP socket operations
1087  * - send ICMP echo request
1088  * - all ICMP
1089  * - limited socket options
1090  */
1091 
1092 #include <netinet/ip_icmp.h>
1093 #include <netinet/in_pcb.h>
1094 
1095 extern u_int32_t rip_sendspace;
1096 extern u_int32_t rip_recvspace;
1097 extern struct inpcbinfo ripcbinfo;
1098 
1099 int rip_abort(struct socket *);
1100 int rip_bind(struct socket *, struct sockaddr *, struct proc *);
1101 int rip_connect(struct socket *, struct sockaddr *, struct proc *);
1102 int rip_detach(struct socket *);
1103 int rip_disconnect(struct socket *);
1104 int rip_shutdown(struct socket *);
1105 
1106 __private_extern__ int icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct proc *p);
1107 __private_extern__ int icmp_dgram_attach(struct socket *so, int proto, struct proc *p);
1108 __private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt);
1109 
1110 __private_extern__ struct pr_usrreqs icmp_dgram_usrreqs = {
1111 	.pru_abort =            rip_abort,
1112 	.pru_attach =           icmp_dgram_attach,
1113 	.pru_bind =             rip_bind,
1114 	.pru_connect =          rip_connect,
1115 	.pru_control =          in_control,
1116 	.pru_detach =           rip_detach,
1117 	.pru_disconnect =       rip_disconnect,
1118 	.pru_peeraddr =         in_getpeeraddr,
1119 	.pru_send =             icmp_dgram_send,
1120 	.pru_shutdown =         rip_shutdown,
1121 	.pru_sockaddr =         in_getsockaddr,
1122 	.pru_sosend =           sosend,
1123 	.pru_soreceive =        soreceive,
1124 };
1125 
1126 /* Like rip_attach but without root privilege enforcement */
1127 __private_extern__ int
icmp_dgram_attach(struct socket * so,__unused int proto,struct proc * p)1128 icmp_dgram_attach(struct socket *so, __unused int proto, struct proc *p)
1129 {
1130 	struct inpcb *inp;
1131 	int error;
1132 
1133 	inp = sotoinpcb(so);
1134 	if (inp) {
1135 		panic("icmp_dgram_attach");
1136 	}
1137 
1138 	error = soreserve(so, rip_sendspace, rip_recvspace);
1139 	if (error) {
1140 		return error;
1141 	}
1142 	error = in_pcballoc(so, &ripcbinfo, p);
1143 	if (error) {
1144 		return error;
1145 	}
1146 	inp = (struct inpcb *)so->so_pcb;
1147 	inp->inp_vflag |= INP_IPV4;
1148 	inp->inp_ip_p = IPPROTO_ICMP;
1149 	inp->inp_ip_ttl = (u_char)ip_defttl;
1150 	return 0;
1151 }
1152 
1153 /*
1154  * Raw IP socket option processing.
1155  */
1156 __private_extern__ int
icmp_dgram_ctloutput(struct socket * so,struct sockopt * sopt)1157 icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt)
1158 {
1159 	int     error;
1160 
1161 	if (sopt->sopt_level != IPPROTO_IP) {
1162 		return EINVAL;
1163 	}
1164 
1165 	switch (sopt->sopt_name) {
1166 	case IP_OPTIONS:
1167 	case IP_HDRINCL:
1168 	case IP_TOS:
1169 	case IP_TTL:
1170 	case IP_RECVOPTS:
1171 	case IP_RECVRETOPTS:
1172 	case IP_RECVDSTADDR:
1173 	case IP_RETOPTS:
1174 	case IP_MULTICAST_IF:
1175 	case IP_MULTICAST_IFINDEX:
1176 	case IP_MULTICAST_TTL:
1177 	case IP_MULTICAST_LOOP:
1178 	case IP_ADD_MEMBERSHIP:
1179 	case IP_DROP_MEMBERSHIP:
1180 	case IP_MULTICAST_VIF:
1181 	case IP_PORTRANGE:
1182 	case IP_RECVIF:
1183 	case IP_IPSEC_POLICY:
1184 	case IP_STRIPHDR:
1185 	case IP_RECVTTL:
1186 	case IP_BOUND_IF:
1187 	case IP_DONTFRAG:
1188 	case IP_NO_IFT_CELLULAR:
1189 		error = rip_ctloutput(so, sopt);
1190 		break;
1191 
1192 	default:
1193 		error = EINVAL;
1194 		break;
1195 	}
1196 
1197 	return error;
1198 }
1199 
1200 __private_extern__ int
icmp_dgram_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1201 icmp_dgram_send(struct socket *so, int flags, struct mbuf *m,
1202     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1203 {
1204 	struct ip *ip;
1205 	struct inpcb *inp = sotoinpcb(so);
1206 	int hlen;
1207 	struct icmp *icp;
1208 	struct in_ifaddr *ia = NULL;
1209 	int icmplen;
1210 	int error = EINVAL;
1211 	int inp_flags = inp ? inp->inp_flags : 0;
1212 
1213 	if (inp == NULL
1214 #if NECP
1215 	    || (necp_socket_should_use_flow_divert(inp))
1216 #endif /* NECP */
1217 	    ) {
1218 		if (inp != NULL) {
1219 			error = EPROTOTYPE;
1220 		}
1221 		goto bad;
1222 	}
1223 
1224 #if CONTENT_FILTER
1225 	/*
1226 	 * If socket is subject to Content Filter, get inp_flags from saved state
1227 	 */
1228 	if (CFIL_DGRAM_FILTERED(so) && nam == NULL) {
1229 		cfil_dgram_peek_socket_state(m, &inp_flags);
1230 	}
1231 #endif
1232 
1233 	if ((inp_flags & INP_HDRINCL) != 0) {
1234 		/* Expect 32-bit aligned data ptr on strict-align platforms */
1235 		MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
1236 		/*
1237 		 * This is not raw IP, we liberal only for fields TOS,
1238 		 * id and TTL.
1239 		 */
1240 		ip = mtod(m, struct ip *);
1241 
1242 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1243 		/* Some sanity checks */
1244 		if (m->m_pkthdr.len < hlen + ICMP_MINLEN) {
1245 			goto bad;
1246 		}
1247 		/* Only IPv4 */
1248 		if (IP_VHL_V(ip->ip_vhl) != 4) {
1249 			goto bad;
1250 		}
1251 		if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len) {
1252 			goto bad;
1253 		}
1254 		/* Bogus fragments can tie up peer resources */
1255 		if ((ip->ip_off & ~IP_DF) != 0) {
1256 			goto bad;
1257 		}
1258 		/* Allow only ICMP even for user provided IP header */
1259 		if (ip->ip_p != IPPROTO_ICMP) {
1260 			goto bad;
1261 		}
1262 		/*
1263 		 * To prevent spoofing, specified source address must
1264 		 * be one of ours.
1265 		 */
1266 		if (ip->ip_src.s_addr != INADDR_ANY) {
1267 			socket_unlock(so, 0);
1268 			lck_rw_lock_shared(&in_ifaddr_rwlock);
1269 			if (TAILQ_EMPTY(&in_ifaddrhead)) {
1270 				lck_rw_done(&in_ifaddr_rwlock);
1271 				socket_lock(so, 0);
1272 				goto bad;
1273 			}
1274 			TAILQ_FOREACH(ia, INADDR_HASH(ip->ip_src.s_addr),
1275 			    ia_hash) {
1276 				IFA_LOCK(&ia->ia_ifa);
1277 				if (IA_SIN(ia)->sin_addr.s_addr ==
1278 				    ip->ip_src.s_addr) {
1279 					IFA_UNLOCK(&ia->ia_ifa);
1280 					lck_rw_done(&in_ifaddr_rwlock);
1281 					socket_lock(so, 0);
1282 					goto ours;
1283 				}
1284 				IFA_UNLOCK(&ia->ia_ifa);
1285 			}
1286 			lck_rw_done(&in_ifaddr_rwlock);
1287 			socket_lock(so, 0);
1288 			goto bad;
1289 		}
1290 ours:
1291 		/* Do not trust we got a valid checksum */
1292 		ip->ip_sum = 0;
1293 
1294 		icp = (struct icmp *)(void *)(((char *)m->m_data) + hlen);
1295 		icmplen = m->m_pkthdr.len - hlen;
1296 	} else {
1297 		if ((icmplen = m->m_pkthdr.len) < ICMP_MINLEN) {
1298 			goto bad;
1299 		}
1300 		icp = mtod(m, struct icmp *);
1301 	}
1302 	/*
1303 	 * Allow only to send request types with code 0
1304 	 */
1305 	if (icp->icmp_code != 0) {
1306 		goto bad;
1307 	}
1308 	switch (icp->icmp_type) {
1309 	case ICMP_ECHO:
1310 		break;
1311 	case ICMP_TSTAMP:
1312 		if (icmplen != 20) {
1313 			goto bad;
1314 		}
1315 		break;
1316 	case ICMP_MASKREQ:
1317 		if (icmplen != 12) {
1318 			goto bad;
1319 		}
1320 		break;
1321 	default:
1322 		goto bad;
1323 	}
1324 	return rip_send(so, flags, m, nam, control, p);
1325 bad:
1326 	VERIFY(error != 0);
1327 
1328 	if (m != NULL) {
1329 		m_freem(m);
1330 	}
1331 	if (control != NULL) {
1332 		m_freem(control);
1333 	}
1334 
1335 	return error;
1336 }
1337 
1338 #endif /* __APPLE__ */
1339