xref: /xnu-11417.140.69/bsd/netinet/ip_icmp.c (revision 43a90889846e00bfb5cf1d255cdc0a701a1e05a4)
1 /*
2  * Copyright (c) 2000-2024 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 		struct icmp oicp = {0};
252 		memcpy(&oicp, mtodo(n, oiphlen), ICMP_MINLEN);
253 		if (!ICMP_INFOTYPE(oicp.icmp_type)) {
254 			icmpstat.icps_oldicmp++;
255 			goto freeit;
256 		}
257 	}
258 
259 	/*
260 	 * Calculate the length to quote from original packet and prevent
261 	 * the ICMP mbuf from overflowing.
262 	 * Unfortunatly this is non-trivial since ip_forward()
263 	 * sends us truncated packets.
264 	 */
265 	nlen = m_length(n);
266 	if (oip->ip_p == IPPROTO_TCP) {
267 		struct tcphdr *th = NULL;
268 		u_int16_t tcphlen = 0;
269 
270 		/*
271 		 * If the packet got truncated and TCP header
272 		 * is not contained in the packet, send out
273 		 * standard reply with only IP header as payload
274 		 */
275 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
276 		    n->m_next == NULL) {
277 			goto stdreply;
278 		}
279 
280 		/*
281 		 * Otherwise, pull up to get IP and TCP headers
282 		 * together
283 		 */
284 		if (n->m_len < (oiphlen + sizeof(struct tcphdr)) &&
285 		    (n = m_pullup(n, (oiphlen + sizeof(struct tcphdr)))) == NULL) {
286 			goto freeit;
287 		}
288 
289 		/*
290 		 * Reinit pointers derived from mbuf data pointer
291 		 * as things might have moved around with m_pullup
292 		 */
293 		oip = mtod(n, struct ip *);
294 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
295 
296 		if (th != ((struct tcphdr *)P2ROUNDDOWN(th,
297 		    sizeof(u_int32_t))) ||
298 		    ((th->th_off << 2) > UINT16_MAX)) {
299 			goto freeit;
300 		}
301 		tcphlen = (uint16_t)(th->th_off << 2);
302 
303 		/* Sanity checks */
304 		if (tcphlen < sizeof(struct tcphdr)) {
305 			goto freeit;
306 		}
307 		if (oip->ip_len < (oiphlen + tcphlen)) {
308 			goto freeit;
309 		}
310 		if ((oiphlen + tcphlen) > n->m_len && n->m_next == NULL) {
311 			goto stdreply;
312 		}
313 		if (n->m_len < (oiphlen + tcphlen) &&
314 		    (n = m_pullup(n, (oiphlen + tcphlen))) == NULL) {
315 			goto freeit;
316 		}
317 
318 		/*
319 		 * Reinit pointers derived from mbuf data pointer
320 		 * as things might have moved around with m_pullup
321 		 */
322 		oip = mtod(n, struct ip *);
323 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
324 
325 		icmpelen = max(tcphlen, min(icmp_datalen,
326 		    (oip->ip_len - oiphlen)));
327 	} else {
328 stdreply:       icmpelen = max(ICMP_MINLEN, min(icmp_datalen,
329 		    (oip->ip_len - oiphlen)));
330 	}
331 
332 	icmplen = min(oiphlen + icmpelen, nlen);
333 	if (icmplen < sizeof(struct ip)) {
334 		goto freeit;
335 	}
336 
337 	/*
338 	 * First, formulate icmp message
339 	 * Allocate enough space for the IP header, ICMP header
340 	 * and the payload (part of the original message to be sent back).
341 	 */
342 	if (MHLEN > (sizeof(struct ip) + ICMP_MINLEN + icmplen)) {
343 		m = m_gethdr(M_DONTWAIT, MT_HEADER);    /* MAC-OK */
344 	} else {
345 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
346 	}
347 
348 	if (m == NULL) {
349 		goto freeit;
350 	}
351 
352 	/*
353 	 * Further refine the payload length to the space
354 	 * remaining in mbuf after including the IP header and ICMP
355 	 * header.
356 	 */
357 	icmplen = min(icmplen, (u_int)M_TRAILINGSPACE(m) -
358 	    (u_int)(sizeof(struct ip) - ICMP_MINLEN));
359 	m_align(m, ICMP_MINLEN + icmplen);
360 	m->m_len = ICMP_MINLEN + icmplen; /* for ICMP header and data */
361 
362 	icp = mtod(m, struct icmp *);
363 	icmpstat.icps_outhist[type]++;
364 	icp->icmp_type = (u_char)type;
365 	if (type == ICMP_REDIRECT) {
366 		icp->icmp_gwaddr.s_addr = dest;
367 	} else {
368 		icp->icmp_void = 0;
369 		/*
370 		 * The following assignments assume an overlay with the
371 		 * zeroed icmp_void field.
372 		 */
373 		if (type == ICMP_PARAMPROB) {
374 			icp->icmp_pptr = (u_char)code;
375 			code = 0;
376 		} else if (type == ICMP_UNREACH &&
377 		    code == ICMP_UNREACH_NEEDFRAG && nextmtu != 0) {
378 			icp->icmp_nextmtu = htons((uint16_t)nextmtu);
379 		}
380 	}
381 
382 	icp->icmp_code = (u_char)code;
383 
384 	/*
385 	 * Copy icmplen worth of content from original
386 	 * mbuf (n) to the new packet after ICMP header.
387 	 */
388 	m_copydata(n, 0, icmplen, mtod(m, caddr_t) + offsetof(struct icmp, icmp_ip));
389 	nip = &icp->icmp_ip;
390 
391 	/*
392 	 * Convert fields to network representation.
393 	 */
394 #if BYTE_ORDER != BIG_ENDIAN
395 	HTONS(nip->ip_len);
396 	HTONS(nip->ip_off);
397 #endif
398 	/*
399 	 * Set up ICMP message mbuf and copy old IP header (without options
400 	 * in front of ICMP message.
401 	 */
402 	m->m_data -= sizeof(struct ip);
403 	m->m_len += sizeof(struct ip);
404 	m->m_pkthdr.len = m->m_len;
405 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
406 	/*
407 	 * To avoid some flavors of port scanning and other attacks,
408 	 * use packet suppression without using any other sort of
409 	 * rate limiting with static bounds.
410 	 * XXX Not setting PKTF_FLOW_ID here because we were concerned
411 	 * about it triggering regression elsewhere outside of network stack
412 	 * where there might be an assumption around flow ID being non-zero.
413 	 * It should be noted though that previously if PKTF_FLOW_ID was not
414 	 * set, PF would have generated flow hash irrespective of ICMPv4/v6
415 	 * type. That doesn't happen now and PF only computes hash for ICMP
416 	 * types that need state creation (which is not true of error types).
417 	 * It would have been a problem because we really want all the ICMP
418 	 * error type packets to share the same flow ID for global suppression.
419 	 */
420 	m->m_pkthdr.comp_gencnt = icmp_error_packet_gencount(type, code);
421 
422 	nip = mtod(m, struct ip *);
423 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
424 	nip->ip_len = (uint16_t)m->m_len;
425 	nip->ip_vhl = IP_VHL_BORING;
426 	nip->ip_p = IPPROTO_ICMP;
427 	nip->ip_tos = 0;
428 	nip->ip_off = 0;
429 	icmp_reflect(m);
430 freeit:
431 	m_freem(n);
432 }
433 
434 /*
435  * Process a received ICMP message.
436  */
437 void
icmp_input(struct mbuf * m,int hlen)438 icmp_input(struct mbuf *m, int hlen)
439 {
440 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
441 	struct icmp *icp;
442 	struct ip *ip = mtod(m, struct ip *);
443 	int icmplen;
444 	int i;
445 	struct in_ifaddr *ia;
446 	void (*ctlfunc)(int, struct sockaddr *, void *, struct ifnet *);
447 	int code;
448 	boolean_t should_log_redirect = false;
449 
450 	/* Expect 32-bit aligned data pointer on strict-align platforms */
451 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
452 
453 	icmplen = ip->ip_len;
454 
455 	/*
456 	 * Locate icmp structure in mbuf, and check
457 	 * that not corrupted and of at least minimum length.
458 	 */
459 #if (DEBUG | DEVELOPMENT)
460 	if (icmpprintfs > 2) {
461 		char src_str[MAX_IPv4_STR_LEN];
462 		char dst_str[MAX_IPv4_STR_LEN];
463 
464 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
465 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
466 		printf("%s: from %s to %s, len %d\n",
467 		    __func__, src_str, dst_str, icmplen);
468 	}
469 #endif
470 	if (icmplen < ICMP_MINLEN) {
471 		icmpstat.icps_tooshort++;
472 		goto freeit;
473 	}
474 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
475 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
476 		icmpstat.icps_tooshort++;
477 		return;
478 	}
479 	/* Reset the pointers, since `m_pullup' might have moved `m'. `icp' is reset below. */
480 	ip = mtod(m, struct ip *);
481 
482 	m->m_len -= hlen;
483 	m->m_data += hlen;
484 	// Forging because we might not have the full size of one struct icmp,
485 	// but we have enough bytes to work with
486 	icp = __unsafe_forge_single(struct icmp *, mtod(m, struct icmp *));
487 	if (in_cksum(m, icmplen) != 0) {
488 		icmpstat.icps_checksum++;
489 		goto freeit;
490 	}
491 	m->m_len += hlen;
492 	m->m_data -= hlen;
493 
494 #if (DEBUG | DEVELOPMENT)
495 	if (icmpprintfs > 2) {
496 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
497 		    icp->icmp_code);
498 	}
499 #endif
500 
501 	/*
502 	 * Message type specific processing.
503 	 */
504 	if (icp->icmp_type > ICMP_MAXTYPE) {
505 		goto raw;
506 	}
507 
508 	/* Initialize */
509 	SOCKADDR_ZERO(&icmpsrc, sizeof(icmpsrc));
510 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
511 	icmpsrc.sin_family = AF_INET;
512 	SOCKADDR_ZERO(&icmpdst, sizeof(icmpdst));
513 	icmpdst.sin_len = sizeof(struct sockaddr_in);
514 	icmpdst.sin_family = AF_INET;
515 	SOCKADDR_ZERO(&icmpgw, sizeof(icmpgw));
516 	icmpgw.sin_len = sizeof(struct sockaddr_in);
517 	icmpgw.sin_family = AF_INET;
518 
519 	icmpstat.icps_inhist[icp->icmp_type]++;
520 	code = icp->icmp_code;
521 	switch (icp->icmp_type) {
522 	case ICMP_UNREACH:
523 		switch (code) {
524 		case ICMP_UNREACH_NET:
525 		case ICMP_UNREACH_HOST:
526 		case ICMP_UNREACH_SRCFAIL:
527 		case ICMP_UNREACH_NET_UNKNOWN:
528 		case ICMP_UNREACH_HOST_UNKNOWN:
529 		case ICMP_UNREACH_ISOLATED:
530 		case ICMP_UNREACH_TOSNET:
531 		case ICMP_UNREACH_TOSHOST:
532 		case ICMP_UNREACH_HOST_PRECEDENCE:
533 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
534 			code = PRC_UNREACH_NET;
535 			break;
536 
537 		case ICMP_UNREACH_NEEDFRAG:
538 			code = PRC_MSGSIZE;
539 			break;
540 
541 		/*
542 		 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
543 		 * Treat subcodes 2,3 as immediate RST
544 		 */
545 		case ICMP_UNREACH_PROTOCOL:
546 		case ICMP_UNREACH_PORT:
547 			code = PRC_UNREACH_PORT;
548 			break;
549 
550 		case ICMP_UNREACH_NET_PROHIB:
551 		case ICMP_UNREACH_HOST_PROHIB:
552 		case ICMP_UNREACH_FILTER_PROHIB:
553 			code = PRC_UNREACH_ADMIN_PROHIB;
554 			break;
555 
556 		default:
557 			goto badcode;
558 		}
559 		goto deliver;
560 
561 	case ICMP_TIMXCEED:
562 		if (code > 1) {
563 			goto badcode;
564 		}
565 		code += PRC_TIMXCEED_INTRANS;
566 		goto deliver;
567 
568 	case ICMP_PARAMPROB:
569 		if (code > 1) {
570 			goto badcode;
571 		}
572 		code = PRC_PARAMPROB;
573 		goto deliver;
574 
575 	case ICMP_SOURCEQUENCH:
576 		if (code) {
577 			goto badcode;
578 		}
579 		code = PRC_QUENCH;
580 deliver:
581 		/*
582 		 * Problem with datagram; advise higher level routines.
583 		 */
584 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp)
585 		    || IP_VHL_HL(icp->icmp_ip.ip_vhl) <
586 		    (sizeof(struct ip) >> 2) ||
587 		    (m = m_pullup(m, hlen + ICMP_ADVLEN(icp))) == NULL) {
588 			icmpstat.icps_badlen++;
589 			goto freeit;
590 		}
591 
592 		/* Reset the pointers, since `m_pullup' might have moved `m'*/
593 		ip = mtod(m, struct ip *);
594 		icp = __unsafe_forge_single(struct icmp *, mtodo(m, hlen));
595 
596 #if BYTE_ORDER != BIG_ENDIAN
597 		NTOHS(icp->icmp_ip.ip_len);
598 #endif
599 
600 		/* Discard ICMP's in response to multicast packets */
601 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) {
602 			goto badcode;
603 		}
604 #if (DEBUG | DEVELOPMENT)
605 		if (icmpprintfs > 2) {
606 			printf("deliver to protocol %d\n",
607 			    icp->icmp_ip.ip_p);
608 		}
609 #endif
610 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
611 
612 		/*
613 		 * if the packet contains [IPv4 AH TCP], we can't make a
614 		 * notification to TCP layer.
615 		 */
616 		ctlfunc = ip_protox[icp->icmp_ip.ip_p]->pr_ctlinput;
617 
618 		if (ctlfunc) {
619 			struct ipctlparam ctl_param = {
620 				.ipc_m = m,
621 				.ipc_icmp = icp,
622 				.ipc_icmp_ip = &icp->icmp_ip,
623 				.ipc_off = hlen + offsetof(struct icmp, icmp_ip) + (IP_VHL_HL(icp->icmp_ip.ip_vhl) << 2)
624 			};
625 			LCK_MTX_ASSERT(inet_domain_mutex, LCK_MTX_ASSERT_OWNED);
626 
627 			lck_mtx_unlock(inet_domain_mutex);
628 
629 			(*ctlfunc)(code, SA(&icmpsrc),
630 			    (void *)&ctl_param, m->m_pkthdr.rcvif);
631 
632 			lck_mtx_lock(inet_domain_mutex);
633 		}
634 		break;
635 
636 badcode:
637 		icmpstat.icps_badcode++;
638 		break;
639 
640 	case ICMP_ECHO:
641 		if ((m->m_flags & (M_MCAST | M_BCAST))) {
642 			if (icmpbmcastecho == 0) {
643 				icmpstat.icps_bmcastecho++;
644 				break;
645 			}
646 		}
647 
648 		/*
649 		 * rdar://18644769
650 		 * Do not reply when the destination is link local multicast or broadcast
651 		 * and the source is not from a directly connected subnet
652 		 */
653 		if ((IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr)) ||
654 		    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) &&
655 		    in_localaddr(ip->ip_src) == 0) {
656 			icmpstat.icps_bmcastecho++;
657 #if (DEBUG | DEVELOPMENT)
658 			if (icmpprintfs > 0) {
659 				char src_str[MAX_IPv4_STR_LEN];
660 				char dst_str[MAX_IPv4_STR_LEN];
661 
662 				inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
663 				inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
664 				printf("%s: non local (B|M)CAST %s to %s, len %d\n",
665 				    __func__, src_str, dst_str, icmplen);
666 			}
667 #endif
668 			break;
669 		}
670 
671 		icp->icmp_type = ICMP_ECHOREPLY;
672 		goto reflect;
673 
674 	case ICMP_TSTAMP:
675 		if (icmptimestamp == 0) {
676 			break;
677 		}
678 
679 		if (!icmpbmcastecho
680 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
681 			icmpstat.icps_bmcasttstamp++;
682 			break;
683 		}
684 		if (icmplen < ICMP_TSLEN) {
685 			icmpstat.icps_badlen++;
686 			break;
687 		}
688 		icp->icmp_type = ICMP_TSTAMPREPLY;
689 		icp->icmp_rtime = iptime();
690 		icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
691 		goto reflect;
692 
693 	case ICMP_MASKREQ:
694 		if (icmpmaskrepl == 0) {
695 			break;
696 		}
697 		/*
698 		 * We are not able to respond with all ones broadcast
699 		 * unless we receive it over a point-to-point interface.
700 		 */
701 		if (icmplen < ICMP_MASKLEN) {
702 			break;
703 		}
704 		switch (ip->ip_dst.s_addr) {
705 		case INADDR_BROADCAST:
706 		case INADDR_ANY:
707 			icmpdst.sin_addr = ip->ip_src;
708 			break;
709 
710 		default:
711 			icmpdst.sin_addr = ip->ip_dst;
712 		}
713 		ia = ifatoia(ifaof_ifpforaddr(SA(&icmpdst),
714 		    m->m_pkthdr.rcvif));
715 		if (ia == 0) {
716 			break;
717 		}
718 		IFA_LOCK(&ia->ia_ifa);
719 		if (ia->ia_ifp == 0) {
720 			IFA_UNLOCK(&ia->ia_ifa);
721 			ifa_remref(&ia->ia_ifa);
722 			ia = NULL;
723 			break;
724 		}
725 		icp->icmp_type = ICMP_MASKREPLY;
726 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
727 		if (ip->ip_src.s_addr == 0) {
728 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
729 				ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
730 			} else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
731 				ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
732 			}
733 		}
734 		IFA_UNLOCK(&ia->ia_ifa);
735 		ifa_remref(&ia->ia_ifa);
736 reflect:
737 		ip->ip_len += hlen;     /* since ip_input deducts this */
738 		icmpstat.icps_reflect++;
739 		icmpstat.icps_outhist[icp->icmp_type]++;
740 		icmp_reflect(m);
741 		return;
742 
743 	case ICMP_REDIRECT:
744 		if (drop_redirect) {
745 			break;
746 		}
747 		if (code > 3) {
748 			goto badcode;
749 		}
750 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
751 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
752 			icmpstat.icps_badlen++;
753 			break;
754 		}
755 
756 #if (DEBUG | DEVELOPMENT)
757 		should_log_redirect = log_redirect || (icmpprintfs > 0);
758 #else
759 		should_log_redirect = log_redirect;
760 #endif
761 		/*
762 		 * Short circuit routing redirects to force
763 		 * immediate change in the kernel's routing
764 		 * tables.  The message is also handed to anyone
765 		 * listening on a raw socket (e.g. the routing
766 		 * daemon for use in updating its tables).
767 		 */
768 		icmpgw.sin_addr = ip->ip_src;
769 		icmpdst.sin_addr = icp->icmp_gwaddr;
770 
771 		if (should_log_redirect) {
772 			char src_str[MAX_IPv4_STR_LEN];
773 			char dst_str[MAX_IPv4_STR_LEN];
774 			char gw_str[MAX_IPv4_STR_LEN];
775 
776 			inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
777 			inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, dst_str, sizeof(dst_str));
778 			inet_ntop(AF_INET, &icp->icmp_gwaddr, gw_str, sizeof(gw_str));
779 			printf("%s: redirect dst %s to %s from %s\n", __func__,
780 			    dst_str, gw_str, src_str);
781 		}
782 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
783 		rtredirect(m->m_pkthdr.rcvif, SA(&icmpsrc),
784 		    SA(&icmpdst), NULL, RTF_GATEWAY | RTF_HOST,
785 		    SA(&icmpgw), NULL);
786 		pfctlinput(PRC_REDIRECT_HOST, SA(&icmpsrc));
787 #if IPSEC
788 		key_sa_routechange(SA(&icmpsrc));
789 #endif
790 		break;
791 
792 	/*
793 	 * No kernel processing for the following;
794 	 * just fall through to send to raw listener.
795 	 */
796 	case ICMP_ECHOREPLY:
797 	case ICMP_ROUTERADVERT:
798 	case ICMP_ROUTERSOLICIT:
799 	case ICMP_TSTAMPREPLY:
800 	case ICMP_IREQREPLY:
801 	case ICMP_MASKREPLY:
802 	default:
803 		break;
804 	}
805 
806 raw:
807 	rip_input(m, hlen);
808 	return;
809 
810 freeit:
811 	m_freem(m);
812 }
813 
814 /*
815  * Reflect the ip packet back to the source
816  */
817 static void
icmp_reflect(struct mbuf * m)818 icmp_reflect(struct mbuf *m)
819 {
820 	struct ip *ip = mtod(m, struct ip *);
821 	struct sockaddr_in icmpdst;
822 	struct in_ifaddr *ia;
823 	struct in_addr t;
824 	struct mbuf *opts = NULL;
825 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
826 
827 	if (!in_canforward(ip->ip_src) &&
828 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
829 	    (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
830 		m_freem(m);     /* Bad return address */
831 		goto done;      /* Ip_output() will check for broadcast */
832 	}
833 	t = ip->ip_dst;
834 	ip->ip_dst = ip->ip_src;
835 	/*
836 	 * If the incoming packet was addressed directly to us,
837 	 * use dst as the src for the reply.  Otherwise (broadcast
838 	 * or anonymous), use the address which corresponds
839 	 * to the incoming interface.
840 	 */
841 	lck_rw_lock_shared(&in_ifaddr_rwlock);
842 	TAILQ_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
843 		IFA_LOCK(&ia->ia_ifa);
844 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
845 			ifa_addref(&ia->ia_ifa);
846 			IFA_UNLOCK(&ia->ia_ifa);
847 			goto match;
848 		}
849 		IFA_UNLOCK(&ia->ia_ifa);
850 	}
851 	/*
852 	 * Slow path; check for broadcast addresses.  Find a source
853 	 * IP address to use when replying to the broadcast request;
854 	 * let IP handle the source interface selection work.
855 	 */
856 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
857 		IFA_LOCK(&ia->ia_ifa);
858 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
859 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr) {
860 			ifa_addref(&ia->ia_ifa);
861 			IFA_UNLOCK(&ia->ia_ifa);
862 			break;
863 		}
864 		IFA_UNLOCK(&ia->ia_ifa);
865 	}
866 match:
867 	lck_rw_done(&in_ifaddr_rwlock);
868 
869 	/* Initialize */
870 	SOCKADDR_ZERO(&icmpdst, sizeof(icmpdst));
871 	icmpdst.sin_len = sizeof(struct sockaddr_in);
872 	icmpdst.sin_family = AF_INET;
873 	icmpdst.sin_addr = t;
874 	if ((ia == NULL) && m->m_pkthdr.rcvif) {
875 		ia = ifatoia(ifaof_ifpforaddr(SA(&icmpdst),
876 		    m->m_pkthdr.rcvif));
877 	}
878 	/*
879 	 * The following happens if the packet was not addressed to us,
880 	 * and was received on an interface with no IP address.
881 	 */
882 	if (ia == NULL) {
883 		lck_rw_lock_shared(&in_ifaddr_rwlock);
884 		ia = in_ifaddrhead.tqh_first;
885 		if (ia == NULL) {/* no address yet, bail out */
886 			lck_rw_done(&in_ifaddr_rwlock);
887 			m_freem(m);
888 			goto done;
889 		}
890 		ifa_addref(&ia->ia_ifa);
891 		lck_rw_done(&in_ifaddr_rwlock);
892 	}
893 	IFA_LOCK_SPIN(&ia->ia_ifa);
894 	t = IA_SIN(ia)->sin_addr;
895 	IFA_UNLOCK(&ia->ia_ifa);
896 	ip->ip_src = t;
897 	ip->ip_ttl = (u_char)ip_defttl;
898 	ifa_remref(&ia->ia_ifa);
899 	ia = NULL;
900 
901 	if (optlen > 0) {
902 		u_char *cp;
903 		int opt, cnt;
904 		u_int len;
905 
906 		/*
907 		 * Retrieve any source routing from the incoming packet;
908 		 * add on any record-route or timestamp options.
909 		 */
910 		cp = (u_char *) (ip + 1);
911 		if ((opts = ip_srcroute()) == 0 &&
912 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { /* MAC-OK */
913 			opts->m_len = sizeof(struct in_addr);
914 			mtod(opts, struct in_addr *)->s_addr = 0;
915 		}
916 		if (opts) {
917 #if (DEBUG | DEVELOPMENT)
918 			if (icmpprintfs > 1) {
919 				printf("icmp_reflect optlen %d rt %d => ",
920 				    optlen, opts->m_len);
921 			}
922 #endif
923 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
924 				opt = cp[IPOPT_OPTVAL];
925 				if (opt == IPOPT_EOL) {
926 					break;
927 				}
928 				if (opt == IPOPT_NOP) {
929 					len = 1;
930 				} else {
931 					if (cnt < IPOPT_OLEN + sizeof(*cp)) {
932 						break;
933 					}
934 					len = cp[IPOPT_OLEN];
935 					if (len < IPOPT_OLEN + sizeof(*cp) ||
936 					    len > cnt) {
937 						break;
938 					}
939 				}
940 				/*
941 				 * Should check for overflow, but it "can't happen"
942 				 */
943 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
944 				    opt == IPOPT_SECURITY) {
945 					bcopy((caddr_t)cp,
946 					    mtod(opts, caddr_t) + opts->m_len, len);
947 					opts->m_len += len;
948 				}
949 			}
950 			/* Terminate & pad, if necessary */
951 			cnt = opts->m_len % 4;
952 			if (cnt) {
953 				for (; cnt < 4; cnt++) {
954 					*(mtod(opts, caddr_t) + opts->m_len) =
955 					    IPOPT_EOL;
956 					opts->m_len++;
957 				}
958 			}
959 #if (DEBUG | DEVELOPMENT)
960 			if (icmpprintfs > 1) {
961 				printf("%d\n", opts->m_len);
962 			}
963 #endif
964 		}
965 		/*
966 		 * Now strip out original options by copying rest of first
967 		 * mbuf's data back, and adjust the IP length.
968 		 */
969 		ip->ip_len -= optlen;
970 		ip->ip_vhl = IP_VHL_BORING;
971 		m->m_len -= optlen;
972 		if (m->m_flags & M_PKTHDR) {
973 			m->m_pkthdr.len -= optlen;
974 		}
975 		optlen += sizeof(struct ip);
976 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
977 		    (unsigned)(m->m_len - sizeof(struct ip)));
978 	}
979 	m->m_flags &= ~(M_BCAST | M_MCAST);
980 	icmp_send(m, opts);
981 done:
982 	if (opts) {
983 		(void)m_free(opts);
984 	}
985 }
986 
987 /*
988  * Send an icmp packet back to the ip level,
989  * after supplying a checksum.
990  */
991 static void
icmp_send(struct mbuf * m,struct mbuf * opts)992 icmp_send(struct mbuf *m, struct mbuf *opts)
993 {
994 	struct ip *ip = mtod(m, struct ip *);
995 	int hlen;
996 	struct icmp *icp;
997 	struct route ro;
998 	struct ip_out_args ipoa;
999 
1000 	bzero(&ipoa, sizeof(ipoa));
1001 	ipoa.ipoa_boundif = IFSCOPE_NONE;
1002 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR;
1003 	ipoa.ipoa_sotc = SO_TC_UNSPEC;
1004 	ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1005 
1006 	if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) && m->m_pkthdr.rcvif != NULL) {
1007 		ipoa.ipoa_boundif = m->m_pkthdr.rcvif->if_index;
1008 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
1009 	}
1010 
1011 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1012 	m->m_data += hlen;
1013 	m->m_len -= hlen;
1014 	icp = __unsafe_forge_single(struct icmp *, mtod(m, struct icmp *));
1015 	icp->icmp_cksum = 0;
1016 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
1017 	m->m_data -= hlen;
1018 	m->m_len += hlen;
1019 	m->m_pkthdr.rcvif = NULL;
1020 	m->m_pkthdr.csum_data = 0;
1021 	m->m_pkthdr.csum_flags = 0;
1022 #if (DEBUG | DEVELOPMENT)
1023 	if (icmpprintfs > 2) {
1024 		char src_str[MAX_IPv4_STR_LEN];
1025 		char dst_str[MAX_IPv4_STR_LEN];
1026 
1027 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
1028 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
1029 		printf("%s: dst %s src %s\n", __func__, dst_str, src_str);
1030 	}
1031 #endif
1032 	bzero(&ro, sizeof ro);
1033 	(void) ip_output(m, opts, &ro, IP_OUTARGS, NULL, &ipoa);
1034 	ROUTE_RELEASE(&ro);
1035 }
1036 
1037 u_int32_t
iptime(void)1038 iptime(void)
1039 {
1040 	struct timeval atv;
1041 	u_int32_t t;
1042 
1043 	getmicrotime(&atv);
1044 	t = (atv.tv_sec % (24 * 60 * 60)) * 1000 + atv.tv_usec / 1000;
1045 	return htonl(t);
1046 }
1047 
1048 #if 1
1049 /*
1050  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1051  * given current value MTU.  If DIR is less than zero, a larger plateau
1052  * is returned; otherwise, a smaller value is returned.
1053  */
1054 int
ip_next_mtu(int mtu,int dir)1055 ip_next_mtu(int mtu, int dir)
1056 {
1057 	static int mtutab[] = {
1058 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
1059 		68, 0
1060 	};
1061 	int i;
1062 
1063 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
1064 		if (mtu >= mtutab[i]) {
1065 			break;
1066 		}
1067 	}
1068 
1069 	if (dir < 0) {
1070 		if (i == 0) {
1071 			return 0;
1072 		} else {
1073 			return mtutab[i - 1];
1074 		}
1075 	} else {
1076 		if (mtutab[i] == 0) {
1077 			return 0;
1078 		} else if (mtu > mtutab[i]) {
1079 			return mtutab[i];
1080 		} else {
1081 			return mtutab[i + 1];
1082 		}
1083 	}
1084 }
1085 #endif
1086 
1087 #if __APPLE__
1088 
1089 /*
1090  * Non-privileged ICMP socket operations
1091  * - send ICMP echo request
1092  * - all ICMP
1093  * - limited socket options
1094  */
1095 
1096 #include <netinet/ip_icmp.h>
1097 #include <netinet/in_pcb.h>
1098 
1099 extern u_int32_t rip_sendspace;
1100 extern u_int32_t rip_recvspace;
1101 extern struct inpcbinfo ripcbinfo;
1102 
1103 int rip_abort(struct socket *);
1104 int rip_bind(struct socket *, struct sockaddr *, struct proc *);
1105 int rip_connect(struct socket *, struct sockaddr *, struct proc *);
1106 int rip_detach(struct socket *);
1107 int rip_disconnect(struct socket *);
1108 int rip_shutdown(struct socket *);
1109 
1110 __private_extern__ int icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct proc *p);
1111 __private_extern__ int icmp_dgram_attach(struct socket *so, int proto, struct proc *p);
1112 __private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt);
1113 
1114 __private_extern__ struct pr_usrreqs icmp_dgram_usrreqs = {
1115 	.pru_abort =            rip_abort,
1116 	.pru_attach =           icmp_dgram_attach,
1117 	.pru_bind =             rip_bind,
1118 	.pru_connect =          rip_connect,
1119 	.pru_control =          in_control,
1120 	.pru_detach =           rip_detach,
1121 	.pru_disconnect =       rip_disconnect,
1122 	.pru_peeraddr =         in_getpeeraddr,
1123 	.pru_send =             icmp_dgram_send,
1124 	.pru_shutdown =         rip_shutdown,
1125 	.pru_sockaddr =         in_getsockaddr,
1126 	.pru_sosend =           sosend,
1127 	.pru_soreceive =        soreceive,
1128 };
1129 
1130 /* Like rip_attach but without root privilege enforcement */
1131 __private_extern__ int
icmp_dgram_attach(struct socket * so,__unused int proto,struct proc * p)1132 icmp_dgram_attach(struct socket *so, __unused int proto, struct proc *p)
1133 {
1134 	struct inpcb *inp;
1135 	int error;
1136 
1137 	inp = sotoinpcb(so);
1138 	if (inp) {
1139 		panic("icmp_dgram_attach");
1140 	}
1141 
1142 	error = soreserve(so, rip_sendspace, rip_recvspace);
1143 	if (error) {
1144 		return error;
1145 	}
1146 	error = in_pcballoc(so, &ripcbinfo, p);
1147 	if (error) {
1148 		return error;
1149 	}
1150 	inp = (struct inpcb *)so->so_pcb;
1151 	inp->inp_vflag |= INP_IPV4;
1152 	inp->inp_ip_p = IPPROTO_ICMP;
1153 	inp->inp_ip_ttl = (u_char)ip_defttl;
1154 	return 0;
1155 }
1156 
1157 /*
1158  * Raw IP socket option processing.
1159  */
1160 __private_extern__ int
icmp_dgram_ctloutput(struct socket * so,struct sockopt * sopt)1161 icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt)
1162 {
1163 	int     error;
1164 
1165 	/* Allow <SOL_SOCKET,SO_BINDTODEVICE> at this level */
1166 	if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_BINDTODEVICE) {
1167 		return rip_ctloutput(so, sopt);;
1168 	}
1169 
1170 	if (sopt->sopt_level != IPPROTO_IP) {
1171 		return EINVAL;
1172 	}
1173 
1174 	switch (sopt->sopt_name) {
1175 	case IP_OPTIONS:
1176 	case IP_HDRINCL:
1177 	case IP_TOS:
1178 	case IP_TTL:
1179 	case IP_RECVOPTS:
1180 	case IP_RECVRETOPTS:
1181 	case IP_RECVDSTADDR:
1182 	case IP_RETOPTS:
1183 	case IP_MULTICAST_IF:
1184 	case IP_MULTICAST_IFINDEX:
1185 	case IP_MULTICAST_TTL:
1186 	case IP_MULTICAST_LOOP:
1187 	case IP_ADD_MEMBERSHIP:
1188 	case IP_DROP_MEMBERSHIP:
1189 	case IP_MULTICAST_VIF:
1190 	case IP_PORTRANGE:
1191 	case IP_RECVIF:
1192 	case IP_IPSEC_POLICY:
1193 	case IP_STRIPHDR:
1194 	case IP_RECVTTL:
1195 	case IP_BOUND_IF:
1196 	case IP_DONTFRAG:
1197 	case IP_NO_IFT_CELLULAR:
1198 		error = rip_ctloutput(so, sopt);
1199 		break;
1200 
1201 	default:
1202 		error = EINVAL;
1203 		break;
1204 	}
1205 
1206 	return error;
1207 }
1208 
1209 __private_extern__ int
icmp_dgram_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1210 icmp_dgram_send(struct socket *so, int flags, struct mbuf *m,
1211     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1212 {
1213 	struct ip *ip;
1214 	struct inpcb *inp = sotoinpcb(so);
1215 	int hlen;
1216 	struct icmp *icp;
1217 	struct in_ifaddr *ia = NULL;
1218 	int icmplen;
1219 	int error = EINVAL;
1220 	int inp_flags = inp ? inp->inp_flags : 0;
1221 
1222 	if (inp == NULL
1223 #if NECP
1224 	    || (necp_socket_should_use_flow_divert(inp))
1225 #endif /* NECP */
1226 	    ) {
1227 		if (inp != NULL) {
1228 			error = EPROTOTYPE;
1229 		}
1230 		goto bad;
1231 	}
1232 
1233 #if CONTENT_FILTER
1234 	/*
1235 	 * If socket is subject to Content Filter, get inp_flags from saved state
1236 	 */
1237 	if (CFIL_DGRAM_FILTERED(so) && nam == NULL) {
1238 		cfil_dgram_peek_socket_state(m, &inp_flags);
1239 	}
1240 #endif
1241 
1242 	if ((inp_flags & INP_HDRINCL) != 0) {
1243 		/* Expect 32-bit aligned data ptr on strict-align platforms */
1244 		MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
1245 
1246 		if (m->m_pkthdr.len < sizeof(struct ip)) {
1247 			goto bad;
1248 		}
1249 
1250 		/*
1251 		 * This is not raw IP, we liberal only for fields TOS,
1252 		 * id and TTL.
1253 		 */
1254 		ip = mtod(m, struct ip *);
1255 
1256 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1257 		/* Some sanity checks */
1258 		if (m->m_pkthdr.len < hlen + ICMP_MINLEN) {
1259 			goto bad;
1260 		}
1261 		/* Only IPv4 */
1262 		if (IP_VHL_V(ip->ip_vhl) != 4) {
1263 			goto bad;
1264 		}
1265 		if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len) {
1266 			goto bad;
1267 		}
1268 		/* Bogus fragments can tie up peer resources */
1269 		if ((ip->ip_off & ~IP_DF) != 0) {
1270 			goto bad;
1271 		}
1272 		/* Allow only ICMP even for user provided IP header */
1273 		if (ip->ip_p != IPPROTO_ICMP) {
1274 			goto bad;
1275 		}
1276 		/*
1277 		 * To prevent spoofing, specified source address must
1278 		 * be one of ours.
1279 		 */
1280 		if (ip->ip_src.s_addr != INADDR_ANY) {
1281 			socket_unlock(so, 0);
1282 			lck_rw_lock_shared(&in_ifaddr_rwlock);
1283 			if (TAILQ_EMPTY(&in_ifaddrhead)) {
1284 				lck_rw_done(&in_ifaddr_rwlock);
1285 				socket_lock(so, 0);
1286 				goto bad;
1287 			}
1288 			TAILQ_FOREACH(ia, INADDR_HASH(ip->ip_src.s_addr),
1289 			    ia_hash) {
1290 				IFA_LOCK(&ia->ia_ifa);
1291 				if (IA_SIN(ia)->sin_addr.s_addr ==
1292 				    ip->ip_src.s_addr) {
1293 					IFA_UNLOCK(&ia->ia_ifa);
1294 					lck_rw_done(&in_ifaddr_rwlock);
1295 					socket_lock(so, 0);
1296 					goto ours;
1297 				}
1298 				IFA_UNLOCK(&ia->ia_ifa);
1299 			}
1300 			lck_rw_done(&in_ifaddr_rwlock);
1301 			socket_lock(so, 0);
1302 			goto bad;
1303 		}
1304 ours:
1305 		/* Do not trust we got a valid checksum */
1306 		ip->ip_sum = 0;
1307 
1308 		icp = __unsafe_forge_single(struct icmp *, mtodo(m, hlen));
1309 		icmplen = m->m_pkthdr.len - hlen;
1310 	} else {
1311 		if ((icmplen = m->m_pkthdr.len) < ICMP_MINLEN) {
1312 			goto bad;
1313 		}
1314 		icp = __unsafe_forge_single(struct icmp *, mtod(m, struct icmp *));
1315 	}
1316 	/*
1317 	 * Allow only to send request types with code 0
1318 	 */
1319 	if (icp->icmp_code != 0) {
1320 		goto bad;
1321 	}
1322 	switch (icp->icmp_type) {
1323 	case ICMP_ECHO:
1324 		break;
1325 	case ICMP_TSTAMP:
1326 		if (icmplen != 20) {
1327 			goto bad;
1328 		}
1329 		break;
1330 	case ICMP_MASKREQ:
1331 		if (icmplen != 12) {
1332 			goto bad;
1333 		}
1334 		break;
1335 	default:
1336 		goto bad;
1337 	}
1338 	return rip_send(so, flags, m, nam, control, p);
1339 bad:
1340 	VERIFY(error != 0);
1341 
1342 	if (m != NULL) {
1343 		m_freem(m);
1344 	}
1345 	if (control != NULL) {
1346 		m_freem(control);
1347 	}
1348 
1349 	return error;
1350 }
1351 
1352 #endif /* __APPLE__ */
1353