xref: /xnu-8020.121.3/bsd/netinet/ip_icmp.c (revision fdd8201d7b966f0c3ea610489d29bd841d358941)
1 /*
2  * Copyright (c) 2000-2020 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 
110 /*
111  * ICMP routines: error generation, receive packet processing, and
112  * routines to turnaround packets back to the originator, and
113  * host table maintenance routines.
114  */
115 
116 struct  icmpstat icmpstat;
117 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats,
118     CTLFLAG_RD | CTLFLAG_LOCKED,
119     &icmpstat, icmpstat, "");
120 
121 static int      icmpmaskrepl = 0;
122 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
123     CTLFLAG_RW | CTLFLAG_LOCKED,
124     &icmpmaskrepl, 0, "");
125 
126 static int      icmptimestamp = 0;
127 SYSCTL_INT(_net_inet_icmp, ICMPCTL_TIMESTAMP, timestamp,
128     CTLFLAG_RW | CTLFLAG_LOCKED,
129     &icmptimestamp, 0, "");
130 
131 static int      drop_redirect = 1;
132 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect,
133     CTLFLAG_RW | CTLFLAG_LOCKED,
134     &drop_redirect, 0, "");
135 
136 static int      log_redirect = 0;
137 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect,
138     CTLFLAG_RW | CTLFLAG_LOCKED,
139     &log_redirect, 0, "");
140 
141 const static int icmp_datalen = 8;
142 
143 #if ICMP_BANDLIM
144 /* Default values in case CONFIG_ICMP_BANDLIM is not defined in the MASTER file */
145 #ifndef CONFIG_ICMP_BANDLIM
146 #if XNU_TARGET_OS_OSX
147 #define CONFIG_ICMP_BANDLIM 250
148 #else /* !XNU_TARGET_OS_OSX */
149 #define CONFIG_ICMP_BANDLIM 50
150 #endif /* !XNU_TARGET_OS_OSX */
151 #endif /* CONFIG_ICMP_BANDLIM */
152 
153 /*
154  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
155  *      variable content is -1 and read-only.
156  */
157 
158 static int      icmplim = CONFIG_ICMP_BANDLIM;
159 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW | CTLFLAG_LOCKED,
160     &icmplim, 0, "");
161 #else /* ICMP_BANDLIM */
162 static int      icmplim = -1;
163 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD | CTLFLAG_LOCKED,
164     &icmplim, 0, "");
165 #endif /* ICMP_BANDLIM */
166 
167 static int      icmplim_random_incr = CONFIG_ICMP_BANDLIM;
168 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM_INCR, icmplim_random_incr, CTLFLAG_RW | CTLFLAG_LOCKED,
169     &icmplim_random_incr, 0, "");
170 
171 /*
172  * ICMP broadcast echo sysctl
173  */
174 
175 static int      icmpbmcastecho = 1;
176 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW | CTLFLAG_LOCKED,
177     &icmpbmcastecho, 0, "");
178 
179 #if (DEBUG | DEVELOPMENT)
180 static int      icmpprintfs = 0;
181 SYSCTL_INT(_net_inet_icmp, OID_AUTO, verbose, CTLFLAG_RW | CTLFLAG_LOCKED,
182     &icmpprintfs, 0, "");
183 #endif
184 
185 static void     icmp_reflect(struct mbuf *);
186 static void     icmp_send(struct mbuf *, struct mbuf *);
187 
188 /*
189  * Generate an error packet of type error
190  * in response to bad packet ip.
191  */
192 void
icmp_error(struct mbuf * n,int type,int code,u_int32_t dest,u_int32_t nextmtu)193 icmp_error(
194 	struct mbuf *n,
195 	int type,
196 	int code,
197 	u_int32_t dest,
198 	u_int32_t nextmtu)
199 {
200 	struct ip *oip = NULL;
201 	struct ip *nip = NULL;
202 	struct icmp *icp = NULL;
203 	struct mbuf *m = NULL;
204 	u_int32_t oiphlen = 0;
205 	u_int32_t icmplen = 0;
206 	u_int32_t icmpelen = 0;
207 	u_int32_t nlen = 0;
208 
209 	VERIFY((u_int)type <= ICMP_MAXTYPE);
210 	VERIFY(code <= UINT8_MAX);
211 
212 	/* Expect 32-bit aligned data pointer on strict-align platforms */
213 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(n);
214 
215 	if (type != ICMP_REDIRECT) {
216 		icmpstat.icps_error++;
217 	}
218 	/*
219 	 * Don't send error:
220 	 *   if not the first fragment of message
221 	 *   if original packet was a multicast or broadcast packet
222 	 *   if the old packet protocol was ICMP
223 	 *   error message, only known informational types.
224 	 */
225 	if (n->m_flags & (M_BCAST | M_MCAST)) {
226 		goto freeit;
227 	}
228 
229 	/*
230 	 * Drop if IP header plus ICMP_MINLEN bytes are not contiguous
231 	 * in first mbuf.
232 	 */
233 	if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) {
234 		goto freeit;
235 	}
236 
237 	oip = mtod(n, struct ip *);
238 	oiphlen = IP_VHL_HL(oip->ip_vhl) << 2;
239 	if (n->m_len < oiphlen + ICMP_MINLEN) {
240 		goto freeit;
241 	}
242 
243 #if (DEBUG | DEVELOPMENT)
244 	if (icmpprintfs > 1) {
245 		printf("icmp_error(0x%llx, %x, %d)\n",
246 		    (uint64_t)VM_KERNEL_ADDRPERM(oip), type, code);
247 	}
248 #endif
249 
250 	if (oip->ip_off & ~(IP_MF | IP_DF)) {
251 		goto freeit;
252 	}
253 
254 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
255 	    n->m_len >= oiphlen + ICMP_MINLEN &&
256 	    !ICMP_INFOTYPE(((struct icmp *)(void *)((caddr_t)oip + oiphlen))->
257 	    icmp_type)) {
258 		icmpstat.icps_oldicmp++;
259 		goto freeit;
260 	}
261 
262 	/*
263 	 * Calculate the length to quote from original packet and prevent
264 	 * the ICMP mbuf from overflowing.
265 	 * Unfortunatly this is non-trivial since ip_forward()
266 	 * sends us truncated packets.
267 	 */
268 	nlen = m_length(n);
269 	if (oip->ip_p == IPPROTO_TCP) {
270 		struct tcphdr *th = NULL;
271 		u_int16_t tcphlen = 0;
272 
273 		/*
274 		 * If the packet got truncated and TCP header
275 		 * is not contained in the packet, send out
276 		 * standard reply with only IP header as payload
277 		 */
278 		if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
279 		    n->m_next == NULL) {
280 			goto stdreply;
281 		}
282 
283 		/*
284 		 * Otherwise, pull up to get IP and TCP headers
285 		 * together
286 		 */
287 		if (n->m_len < (oiphlen + sizeof(struct tcphdr)) &&
288 		    (n = m_pullup(n, (oiphlen + sizeof(struct tcphdr)))) == NULL) {
289 			goto freeit;
290 		}
291 
292 		/*
293 		 * Reinit pointers derived from mbuf data pointer
294 		 * as things might have moved around with m_pullup
295 		 */
296 		oip = mtod(n, struct ip *);
297 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
298 
299 		if (th != ((struct tcphdr *)P2ROUNDDOWN(th,
300 		    sizeof(u_int32_t))) ||
301 		    ((th->th_off << 2) > UINT16_MAX)) {
302 			goto freeit;
303 		}
304 		tcphlen = (uint16_t)(th->th_off << 2);
305 
306 		/* Sanity checks */
307 		if (tcphlen < sizeof(struct tcphdr)) {
308 			goto freeit;
309 		}
310 		if (oip->ip_len < (oiphlen + tcphlen)) {
311 			goto freeit;
312 		}
313 		if ((oiphlen + tcphlen) > n->m_len && n->m_next == NULL) {
314 			goto stdreply;
315 		}
316 		if (n->m_len < (oiphlen + tcphlen) &&
317 		    (n = m_pullup(n, (oiphlen + tcphlen))) == NULL) {
318 			goto freeit;
319 		}
320 
321 		/*
322 		 * Reinit pointers derived from mbuf data pointer
323 		 * as things might have moved around with m_pullup
324 		 */
325 		oip = mtod(n, struct ip *);
326 		th = (struct tcphdr *)(void *)((caddr_t)oip + oiphlen);
327 
328 		icmpelen = max(tcphlen, min(icmp_datalen,
329 		    (oip->ip_len - oiphlen)));
330 	} else {
331 stdreply:       icmpelen = max(ICMP_MINLEN, min(icmp_datalen,
332 		    (oip->ip_len - oiphlen)));
333 	}
334 
335 	icmplen = min(oiphlen + icmpelen, nlen);
336 	if (icmplen < sizeof(struct ip)) {
337 		goto freeit;
338 	}
339 
340 	/*
341 	 * First, formulate icmp message
342 	 * Allocate enough space for the IP header, ICMP header
343 	 * and the payload (part of the original message to be sent back).
344 	 */
345 	if (MHLEN > (sizeof(struct ip) + ICMP_MINLEN + icmplen)) {
346 		m = m_gethdr(M_DONTWAIT, MT_HEADER);    /* MAC-OK */
347 	} else {
348 		m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
349 	}
350 
351 	if (m == NULL) {
352 		goto freeit;
353 	}
354 
355 	/*
356 	 * Further refine the payload length to the space
357 	 * remaining in mbuf after including the IP header and ICMP
358 	 * header.
359 	 */
360 	icmplen = min(icmplen, (u_int)M_TRAILINGSPACE(m) -
361 	    (u_int)(sizeof(struct ip) - ICMP_MINLEN));
362 	m_align(m, ICMP_MINLEN + icmplen);
363 	m->m_len = ICMP_MINLEN + icmplen; /* for ICMP header and data */
364 
365 	icp = mtod(m, struct icmp *);
366 	icmpstat.icps_outhist[type]++;
367 	icp->icmp_type = (u_char)type;
368 	if (type == ICMP_REDIRECT) {
369 		icp->icmp_gwaddr.s_addr = dest;
370 	} else {
371 		icp->icmp_void = 0;
372 		/*
373 		 * The following assignments assume an overlay with the
374 		 * zeroed icmp_void field.
375 		 */
376 		if (type == ICMP_PARAMPROB) {
377 			icp->icmp_pptr = (u_char)code;
378 			code = 0;
379 		} else if (type == ICMP_UNREACH &&
380 		    code == ICMP_UNREACH_NEEDFRAG && nextmtu != 0) {
381 			icp->icmp_nextmtu = htons((uint16_t)nextmtu);
382 		}
383 	}
384 
385 	icp->icmp_code = (u_char)code;
386 
387 	/*
388 	 * Copy icmplen worth of content from original
389 	 * mbuf (n) to the new packet after ICMP header.
390 	 */
391 	m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
392 	nip = &icp->icmp_ip;
393 
394 	/*
395 	 * Convert fields to network representation.
396 	 */
397 #if BYTE_ORDER != BIG_ENDIAN
398 	HTONS(nip->ip_len);
399 	HTONS(nip->ip_off);
400 #endif
401 	/*
402 	 * Set up ICMP message mbuf and copy old IP header (without options
403 	 * in front of ICMP message.
404 	 */
405 	m->m_data -= sizeof(struct ip);
406 	m->m_len += sizeof(struct ip);
407 	m->m_pkthdr.len = m->m_len;
408 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
409 	nip = mtod(m, struct ip *);
410 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
411 	nip->ip_len = (uint16_t)m->m_len;
412 	nip->ip_vhl = IP_VHL_BORING;
413 	nip->ip_p = IPPROTO_ICMP;
414 	nip->ip_tos = 0;
415 	nip->ip_off = 0;
416 	icmp_reflect(m);
417 freeit:
418 	m_freem(n);
419 }
420 
421 /*
422  * Process a received ICMP message.
423  */
424 void
icmp_input(struct mbuf * m,int hlen)425 icmp_input(struct mbuf *m, int hlen)
426 {
427 	struct sockaddr_in icmpsrc, icmpdst, icmpgw;
428 	struct icmp *icp;
429 	struct ip *ip = mtod(m, struct ip *);
430 	int icmplen;
431 	int i;
432 	struct in_ifaddr *ia;
433 	void (*ctlfunc)(int, struct sockaddr *, void *, struct ifnet *);
434 	int code;
435 	boolean_t should_log_redirect = false;
436 
437 	/* Expect 32-bit aligned data pointer on strict-align platforms */
438 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
439 
440 	icmplen = ip->ip_len;
441 
442 	/*
443 	 * Locate icmp structure in mbuf, and check
444 	 * that not corrupted and of at least minimum length.
445 	 */
446 #if (DEBUG | DEVELOPMENT)
447 	if (icmpprintfs > 2) {
448 		char src_str[MAX_IPv4_STR_LEN];
449 		char dst_str[MAX_IPv4_STR_LEN];
450 
451 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
452 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
453 		printf("%s: from %s to %s, len %d\n",
454 		    __func__, src_str, dst_str, icmplen);
455 	}
456 #endif
457 	if (icmplen < ICMP_MINLEN) {
458 		icmpstat.icps_tooshort++;
459 		goto freeit;
460 	}
461 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
462 	if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
463 		icmpstat.icps_tooshort++;
464 		return;
465 	}
466 	/* Re-seat the pointers, since `m_pullup' might have moved `m'. `icp' is re-seated below. */
467 	ip = mtod(m, struct ip *);
468 
469 	m->m_len -= hlen;
470 	m->m_data += hlen;
471 	icp = mtod(m, struct icmp *);
472 	if (in_cksum(m, icmplen) != 0) {
473 		icmpstat.icps_checksum++;
474 		goto freeit;
475 	}
476 	m->m_len += hlen;
477 	m->m_data -= hlen;
478 
479 #if (DEBUG | DEVELOPMENT)
480 	if (icmpprintfs > 2) {
481 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
482 		    icp->icmp_code);
483 	}
484 #endif
485 
486 	/*
487 	 * Message type specific processing.
488 	 */
489 	if (icp->icmp_type > ICMP_MAXTYPE) {
490 		goto raw;
491 	}
492 
493 	/* Initialize */
494 	bzero(&icmpsrc, sizeof(icmpsrc));
495 	icmpsrc.sin_len = sizeof(struct sockaddr_in);
496 	icmpsrc.sin_family = AF_INET;
497 	bzero(&icmpdst, sizeof(icmpdst));
498 	icmpdst.sin_len = sizeof(struct sockaddr_in);
499 	icmpdst.sin_family = AF_INET;
500 	bzero(&icmpgw, sizeof(icmpgw));
501 	icmpgw.sin_len = sizeof(struct sockaddr_in);
502 	icmpgw.sin_family = AF_INET;
503 
504 	icmpstat.icps_inhist[icp->icmp_type]++;
505 	code = icp->icmp_code;
506 	switch (icp->icmp_type) {
507 	case ICMP_UNREACH:
508 		switch (code) {
509 		case ICMP_UNREACH_NET:
510 		case ICMP_UNREACH_HOST:
511 		case ICMP_UNREACH_SRCFAIL:
512 		case ICMP_UNREACH_NET_UNKNOWN:
513 		case ICMP_UNREACH_HOST_UNKNOWN:
514 		case ICMP_UNREACH_ISOLATED:
515 		case ICMP_UNREACH_TOSNET:
516 		case ICMP_UNREACH_TOSHOST:
517 		case ICMP_UNREACH_HOST_PRECEDENCE:
518 		case ICMP_UNREACH_PRECEDENCE_CUTOFF:
519 			code = PRC_UNREACH_NET;
520 			break;
521 
522 		case ICMP_UNREACH_NEEDFRAG:
523 			code = PRC_MSGSIZE;
524 			break;
525 
526 		/*
527 		 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
528 		 * Treat subcodes 2,3 as immediate RST
529 		 */
530 		case ICMP_UNREACH_PROTOCOL:
531 		case ICMP_UNREACH_PORT:
532 			code = PRC_UNREACH_PORT;
533 			break;
534 
535 		case ICMP_UNREACH_NET_PROHIB:
536 		case ICMP_UNREACH_HOST_PROHIB:
537 		case ICMP_UNREACH_FILTER_PROHIB:
538 			code = PRC_UNREACH_ADMIN_PROHIB;
539 			break;
540 
541 		default:
542 			goto badcode;
543 		}
544 		goto deliver;
545 
546 	case ICMP_TIMXCEED:
547 		if (code > 1) {
548 			goto badcode;
549 		}
550 		code += PRC_TIMXCEED_INTRANS;
551 		goto deliver;
552 
553 	case ICMP_PARAMPROB:
554 		if (code > 1) {
555 			goto badcode;
556 		}
557 		code = PRC_PARAMPROB;
558 		goto deliver;
559 
560 	case ICMP_SOURCEQUENCH:
561 		if (code) {
562 			goto badcode;
563 		}
564 		code = PRC_QUENCH;
565 deliver:
566 		/*
567 		 * Problem with datagram; advise higher level routines.
568 		 */
569 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp)
570 		    || IP_VHL_HL(icp->icmp_ip.ip_vhl) <
571 		    (sizeof(struct ip) >> 2) ||
572 		    (m = m_pullup(m, hlen + ICMP_ADVLEN(icp))) == NULL) {
573 			icmpstat.icps_badlen++;
574 			goto freeit;
575 		}
576 
577 		/* Re-seat the pointers, since `m_pullup' might have moved `m'*/
578 		ip = mtod(m, struct ip *);
579 		icp = (struct icmp *)(void *)(mtod(m, uint8_t *) + hlen);
580 
581 #if BYTE_ORDER != BIG_ENDIAN
582 		NTOHS(icp->icmp_ip.ip_len);
583 #endif
584 
585 		/* Discard ICMP's in response to multicast packets */
586 		if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) {
587 			goto badcode;
588 		}
589 #if (DEBUG | DEVELOPMENT)
590 		if (icmpprintfs > 2) {
591 			printf("deliver to protocol %d\n",
592 			    icp->icmp_ip.ip_p);
593 		}
594 #endif
595 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
596 
597 		/*
598 		 * if the packet contains [IPv4 AH TCP], we can't make a
599 		 * notification to TCP layer.
600 		 */
601 		ctlfunc = ip_protox[icp->icmp_ip.ip_p]->pr_ctlinput;
602 
603 		if (ctlfunc) {
604 			struct ipctlparam ctl_param = {
605 				.ipc_m = m,
606 				.ipc_icmp = icp,
607 				.ipc_icmp_ip = &icp->icmp_ip,
608 				.ipc_off = hlen + offsetof(struct icmp, icmp_ip) + (IP_VHL_HL(icp->icmp_ip.ip_vhl) << 2)
609 			};
610 			LCK_MTX_ASSERT(inet_domain_mutex, LCK_MTX_ASSERT_OWNED);
611 
612 			lck_mtx_unlock(inet_domain_mutex);
613 
614 			(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
615 			    (void *)&ctl_param, m->m_pkthdr.rcvif);
616 
617 			lck_mtx_lock(inet_domain_mutex);
618 		}
619 		break;
620 
621 badcode:
622 		icmpstat.icps_badcode++;
623 		break;
624 
625 	case ICMP_ECHO:
626 		if ((m->m_flags & (M_MCAST | M_BCAST))) {
627 			if (icmpbmcastecho == 0) {
628 				icmpstat.icps_bmcastecho++;
629 				break;
630 			}
631 		}
632 
633 		/*
634 		 * rdar://18644769
635 		 * Do not reply when the destination is link local multicast or broadcast
636 		 * and the source is not from a directly connected subnet
637 		 */
638 		if ((IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr)) ||
639 		    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) &&
640 		    in_localaddr(ip->ip_src) == 0) {
641 			icmpstat.icps_bmcastecho++;
642 #if (DEBUG | DEVELOPMENT)
643 			if (icmpprintfs > 0) {
644 				char src_str[MAX_IPv4_STR_LEN];
645 				char dst_str[MAX_IPv4_STR_LEN];
646 
647 				inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
648 				inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
649 				printf("%s: non local (B|M)CAST %s to %s, len %d\n",
650 				    __func__, src_str, dst_str, icmplen);
651 			}
652 #endif
653 			break;
654 		}
655 
656 		icp->icmp_type = ICMP_ECHOREPLY;
657 #if ICMP_BANDLIM
658 		if (badport_bandlim(BANDLIM_ICMP_ECHO)) {
659 			goto freeit;
660 		} else
661 #endif
662 		goto reflect;
663 
664 	case ICMP_TSTAMP:
665 		if (icmptimestamp == 0) {
666 			break;
667 		}
668 
669 		if (!icmpbmcastecho
670 		    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
671 			icmpstat.icps_bmcasttstamp++;
672 			break;
673 		}
674 		if (icmplen < ICMP_TSLEN) {
675 			icmpstat.icps_badlen++;
676 			break;
677 		}
678 		icp->icmp_type = ICMP_TSTAMPREPLY;
679 		icp->icmp_rtime = iptime();
680 		icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
681 #if ICMP_BANDLIM
682 		if (badport_bandlim(BANDLIM_ICMP_TSTAMP)) {
683 			goto freeit;
684 		} else
685 #endif
686 		goto reflect;
687 
688 	case ICMP_MASKREQ:
689 		if (icmpmaskrepl == 0) {
690 			break;
691 		}
692 		/*
693 		 * We are not able to respond with all ones broadcast
694 		 * unless we receive it over a point-to-point interface.
695 		 */
696 		if (icmplen < ICMP_MASKLEN) {
697 			break;
698 		}
699 		switch (ip->ip_dst.s_addr) {
700 		case INADDR_BROADCAST:
701 		case INADDR_ANY:
702 			icmpdst.sin_addr = ip->ip_src;
703 			break;
704 
705 		default:
706 			icmpdst.sin_addr = ip->ip_dst;
707 		}
708 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
709 			(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
710 		if (ia == 0) {
711 			break;
712 		}
713 		IFA_LOCK(&ia->ia_ifa);
714 		if (ia->ia_ifp == 0) {
715 			IFA_UNLOCK(&ia->ia_ifa);
716 			IFA_REMREF(&ia->ia_ifa);
717 			ia = NULL;
718 			break;
719 		}
720 		icp->icmp_type = ICMP_MASKREPLY;
721 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
722 		if (ip->ip_src.s_addr == 0) {
723 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
724 				ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
725 			} else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
726 				ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
727 			}
728 		}
729 		IFA_UNLOCK(&ia->ia_ifa);
730 		IFA_REMREF(&ia->ia_ifa);
731 reflect:
732 		ip->ip_len += hlen;     /* since ip_input deducts this */
733 		icmpstat.icps_reflect++;
734 		icmpstat.icps_outhist[icp->icmp_type]++;
735 		icmp_reflect(m);
736 		return;
737 
738 	case ICMP_REDIRECT:
739 		if (drop_redirect) {
740 			break;
741 		}
742 		if (code > 3) {
743 			goto badcode;
744 		}
745 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
746 		    IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
747 			icmpstat.icps_badlen++;
748 			break;
749 		}
750 
751 #if (DEBUG | DEVELOPMENT)
752 		should_log_redirect = log_redirect || (icmpprintfs > 0);
753 #else
754 		should_log_redirect = log_redirect;
755 #endif
756 		/*
757 		 * Short circuit routing redirects to force
758 		 * immediate change in the kernel's routing
759 		 * tables.  The message is also handed to anyone
760 		 * listening on a raw socket (e.g. the routing
761 		 * daemon for use in updating its tables).
762 		 */
763 		icmpgw.sin_addr = ip->ip_src;
764 		icmpdst.sin_addr = icp->icmp_gwaddr;
765 
766 		if (should_log_redirect) {
767 			char src_str[MAX_IPv4_STR_LEN];
768 			char dst_str[MAX_IPv4_STR_LEN];
769 			char gw_str[MAX_IPv4_STR_LEN];
770 
771 			inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
772 			inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, dst_str, sizeof(dst_str));
773 			inet_ntop(AF_INET, &icp->icmp_gwaddr, gw_str, sizeof(gw_str));
774 			printf("%s: redirect dst %s to %s from %s\n", __func__,
775 			    dst_str, gw_str, src_str);
776 		}
777 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
778 		rtredirect(m->m_pkthdr.rcvif, (struct sockaddr *)&icmpsrc,
779 		    (struct sockaddr *)&icmpdst, NULL, RTF_GATEWAY | RTF_HOST,
780 		    (struct sockaddr *)&icmpgw, NULL);
781 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
782 #if IPSEC
783 		key_sa_routechange((struct sockaddr *)&icmpsrc);
784 #endif
785 		break;
786 
787 	/*
788 	 * No kernel processing for the following;
789 	 * just fall through to send to raw listener.
790 	 */
791 	case ICMP_ECHOREPLY:
792 	case ICMP_ROUTERADVERT:
793 	case ICMP_ROUTERSOLICIT:
794 	case ICMP_TSTAMPREPLY:
795 	case ICMP_IREQREPLY:
796 	case ICMP_MASKREPLY:
797 	default:
798 		break;
799 	}
800 
801 raw:
802 	rip_input(m, hlen);
803 	return;
804 
805 freeit:
806 	m_freem(m);
807 }
808 
809 /*
810  * Reflect the ip packet back to the source
811  */
812 static void
icmp_reflect(struct mbuf * m)813 icmp_reflect(struct mbuf *m)
814 {
815 	struct ip *ip = mtod(m, struct ip *);
816 	struct sockaddr_in icmpdst;
817 	struct in_ifaddr *ia;
818 	struct in_addr t;
819 	struct mbuf *opts = NULL;
820 	int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
821 
822 	if (!in_canforward(ip->ip_src) &&
823 	    ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
824 	    (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
825 		m_freem(m);     /* Bad return address */
826 		goto done;      /* Ip_output() will check for broadcast */
827 	}
828 	t = ip->ip_dst;
829 	ip->ip_dst = ip->ip_src;
830 	/*
831 	 * If the incoming packet was addressed directly to us,
832 	 * use dst as the src for the reply.  Otherwise (broadcast
833 	 * or anonymous), use the address which corresponds
834 	 * to the incoming interface.
835 	 */
836 	lck_rw_lock_shared(&in_ifaddr_rwlock);
837 	TAILQ_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
838 		IFA_LOCK(&ia->ia_ifa);
839 		if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
840 			IFA_ADDREF_LOCKED(&ia->ia_ifa);
841 			IFA_UNLOCK(&ia->ia_ifa);
842 			goto match;
843 		}
844 		IFA_UNLOCK(&ia->ia_ifa);
845 	}
846 	/*
847 	 * Slow path; check for broadcast addresses.  Find a source
848 	 * IP address to use when replying to the broadcast request;
849 	 * let IP handle the source interface selection work.
850 	 */
851 	for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
852 		IFA_LOCK(&ia->ia_ifa);
853 		if (ia->ia_ifp && (ia->ia_ifp->if_flags & IFF_BROADCAST) &&
854 		    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr) {
855 			IFA_ADDREF_LOCKED(&ia->ia_ifa);
856 			IFA_UNLOCK(&ia->ia_ifa);
857 			break;
858 		}
859 		IFA_UNLOCK(&ia->ia_ifa);
860 	}
861 match:
862 	lck_rw_done(&in_ifaddr_rwlock);
863 
864 	/* Initialize */
865 	bzero(&icmpdst, sizeof(icmpdst));
866 	icmpdst.sin_len = sizeof(struct sockaddr_in);
867 	icmpdst.sin_family = AF_INET;
868 	icmpdst.sin_addr = t;
869 	if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif) {
870 		ia = (struct in_ifaddr *)ifaof_ifpforaddr(
871 			(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
872 	}
873 	/*
874 	 * The following happens if the packet was not addressed to us,
875 	 * and was received on an interface with no IP address.
876 	 */
877 	if (ia == (struct in_ifaddr *)0) {
878 		lck_rw_lock_shared(&in_ifaddr_rwlock);
879 		ia = in_ifaddrhead.tqh_first;
880 		if (ia == (struct in_ifaddr *)0) {/* no address yet, bail out */
881 			lck_rw_done(&in_ifaddr_rwlock);
882 			m_freem(m);
883 			goto done;
884 		}
885 		IFA_ADDREF(&ia->ia_ifa);
886 		lck_rw_done(&in_ifaddr_rwlock);
887 	}
888 	IFA_LOCK_SPIN(&ia->ia_ifa);
889 	t = IA_SIN(ia)->sin_addr;
890 	IFA_UNLOCK(&ia->ia_ifa);
891 	ip->ip_src = t;
892 	ip->ip_ttl = (u_char)ip_defttl;
893 	IFA_REMREF(&ia->ia_ifa);
894 	ia = NULL;
895 
896 	if (optlen > 0) {
897 		u_char *cp;
898 		int opt, cnt;
899 		u_int len;
900 
901 		/*
902 		 * Retrieve any source routing from the incoming packet;
903 		 * add on any record-route or timestamp options.
904 		 */
905 		cp = (u_char *) (ip + 1);
906 		if ((opts = ip_srcroute()) == 0 &&
907 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { /* MAC-OK */
908 			opts->m_len = sizeof(struct in_addr);
909 			mtod(opts, struct in_addr *)->s_addr = 0;
910 		}
911 		if (opts) {
912 #if (DEBUG | DEVELOPMENT)
913 			if (icmpprintfs > 1) {
914 				printf("icmp_reflect optlen %d rt %d => ",
915 				    optlen, opts->m_len);
916 			}
917 #endif
918 			for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
919 				opt = cp[IPOPT_OPTVAL];
920 				if (opt == IPOPT_EOL) {
921 					break;
922 				}
923 				if (opt == IPOPT_NOP) {
924 					len = 1;
925 				} else {
926 					if (cnt < IPOPT_OLEN + sizeof(*cp)) {
927 						break;
928 					}
929 					len = cp[IPOPT_OLEN];
930 					if (len < IPOPT_OLEN + sizeof(*cp) ||
931 					    len > cnt) {
932 						break;
933 					}
934 				}
935 				/*
936 				 * Should check for overflow, but it "can't happen"
937 				 */
938 				if (opt == IPOPT_RR || opt == IPOPT_TS ||
939 				    opt == IPOPT_SECURITY) {
940 					bcopy((caddr_t)cp,
941 					    mtod(opts, caddr_t) + opts->m_len, len);
942 					opts->m_len += len;
943 				}
944 			}
945 			/* Terminate & pad, if necessary */
946 			cnt = opts->m_len % 4;
947 			if (cnt) {
948 				for (; cnt < 4; cnt++) {
949 					*(mtod(opts, caddr_t) + opts->m_len) =
950 					    IPOPT_EOL;
951 					opts->m_len++;
952 				}
953 			}
954 #if (DEBUG | DEVELOPMENT)
955 			if (icmpprintfs > 1) {
956 				printf("%d\n", opts->m_len);
957 			}
958 #endif
959 		}
960 		/*
961 		 * Now strip out original options by copying rest of first
962 		 * mbuf's data back, and adjust the IP length.
963 		 */
964 		ip->ip_len -= optlen;
965 		ip->ip_vhl = IP_VHL_BORING;
966 		m->m_len -= optlen;
967 		if (m->m_flags & M_PKTHDR) {
968 			m->m_pkthdr.len -= optlen;
969 		}
970 		optlen += sizeof(struct ip);
971 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
972 		    (unsigned)(m->m_len - sizeof(struct ip)));
973 	}
974 	m->m_flags &= ~(M_BCAST | M_MCAST);
975 	icmp_send(m, opts);
976 done:
977 	if (opts) {
978 		(void)m_free(opts);
979 	}
980 }
981 
982 /*
983  * Send an icmp packet back to the ip level,
984  * after supplying a checksum.
985  */
986 static void
icmp_send(struct mbuf * m,struct mbuf * opts)987 icmp_send(struct mbuf *m, struct mbuf *opts)
988 {
989 	struct ip *ip = mtod(m, struct ip *);
990 	int hlen;
991 	struct icmp *icp;
992 	struct route ro;
993 	struct ip_out_args ipoa;
994 
995 	bzero(&ipoa, sizeof(ipoa));
996 	ipoa.ipoa_boundif = IFSCOPE_NONE;
997 	ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR;
998 	ipoa.ipoa_sotc = SO_TC_UNSPEC;
999 	ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1000 
1001 	if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) && m->m_pkthdr.rcvif != NULL) {
1002 		ipoa.ipoa_boundif = m->m_pkthdr.rcvif->if_index;
1003 		ipoa.ipoa_flags |= IPOAF_BOUND_IF;
1004 	}
1005 
1006 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1007 	m->m_data += hlen;
1008 	m->m_len -= hlen;
1009 	icp = mtod(m, struct icmp *);
1010 	icp->icmp_cksum = 0;
1011 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
1012 	m->m_data -= hlen;
1013 	m->m_len += hlen;
1014 	m->m_pkthdr.rcvif = NULL;
1015 	m->m_pkthdr.csum_data = 0;
1016 	m->m_pkthdr.csum_flags = 0;
1017 #if (DEBUG | DEVELOPMENT)
1018 	if (icmpprintfs > 2) {
1019 		char src_str[MAX_IPv4_STR_LEN];
1020 		char dst_str[MAX_IPv4_STR_LEN];
1021 
1022 		inet_ntop(AF_INET, &ip->ip_src, src_str, sizeof(src_str));
1023 		inet_ntop(AF_INET, &ip->ip_dst, dst_str, sizeof(dst_str));
1024 		printf("%s: dst %s src %s\n", __func__, dst_str, src_str);
1025 	}
1026 #endif
1027 	bzero(&ro, sizeof ro);
1028 	(void) ip_output(m, opts, &ro, IP_OUTARGS, NULL, &ipoa);
1029 	ROUTE_RELEASE(&ro);
1030 }
1031 
1032 u_int32_t
iptime(void)1033 iptime(void)
1034 {
1035 	struct timeval atv;
1036 	u_int32_t t;
1037 
1038 	getmicrotime(&atv);
1039 	t = (atv.tv_sec % (24 * 60 * 60)) * 1000 + atv.tv_usec / 1000;
1040 	return htonl(t);
1041 }
1042 
1043 #if 1
1044 /*
1045  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1046  * given current value MTU.  If DIR is less than zero, a larger plateau
1047  * is returned; otherwise, a smaller value is returned.
1048  */
1049 int
ip_next_mtu(int mtu,int dir)1050 ip_next_mtu(int mtu, int dir)
1051 {
1052 	static int mtutab[] = {
1053 		65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
1054 		68, 0
1055 	};
1056 	int i;
1057 
1058 	for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
1059 		if (mtu >= mtutab[i]) {
1060 			break;
1061 		}
1062 	}
1063 
1064 	if (dir < 0) {
1065 		if (i == 0) {
1066 			return 0;
1067 		} else {
1068 			return mtutab[i - 1];
1069 		}
1070 	} else {
1071 		if (mtutab[i] == 0) {
1072 			return 0;
1073 		} else if (mtu > mtutab[i]) {
1074 			return mtutab[i];
1075 		} else {
1076 			return mtutab[i + 1];
1077 		}
1078 	}
1079 }
1080 #endif
1081 
1082 #if ICMP_BANDLIM
1083 
1084 /*
1085  * badport_bandlim() - check for ICMP bandwidth limit
1086  *	Returns false when it is ok to send ICMP error and true to limit sending
1087  *	of ICMP error.
1088  *
1089  *	For now we separate the TCP and UDP subsystems w/ different 'which'
1090  *	values.  We may eventually remove this separation (and simplify the
1091  *	code further).
1092  *
1093  *	Note that the printing of the error message is delayed so we can
1094  *	properly print the icmp error rate that the system was trying to do
1095  *	(i.e. 22000/100 pps, etc...).  This can cause long delays in printing
1096  *	the 'final' error, but it doesn't make sense to solve the printing
1097  *	delay with more complex code.
1098  */
1099 
1100 boolean_t
badport_bandlim(int which)1101 badport_bandlim(int which)
1102 {
1103 	static uint64_t lticks[BANDLIM_MAX + 1];
1104 	static int lpackets[BANDLIM_MAX + 1];
1105 	uint64_t time;
1106 	uint64_t secs;
1107 	static boolean_t is_initialized = FALSE;
1108 	static int icmplim_random;
1109 	const char *bandlimittype[] = {
1110 		"Limiting icmp unreach response",
1111 		"Limiting icmp ping response",
1112 		"Limiting icmp tstamp response",
1113 		"Limiting closed port RST response",
1114 		"Limiting open port RST response"
1115 	};
1116 
1117 	/* Return ok status if feature disabled or argument out of range. */
1118 
1119 	if (icmplim <= 0 || which > BANDLIM_MAX || which < 0) {
1120 		return false;
1121 	}
1122 
1123 	if (is_initialized == FALSE) {
1124 		if (icmplim_random_incr > 0 &&
1125 		    icmplim <= INT32_MAX - (icmplim_random_incr + 1)) {
1126 			icmplim_random = icmplim + (random() % icmplim_random_incr) + 1;
1127 		}
1128 		is_initialized = TRUE;
1129 	}
1130 
1131 	time = net_uptime();
1132 	secs = time - lticks[which];
1133 
1134 	/*
1135 	 * reset stats when cumulative delta exceeds one second.
1136 	 */
1137 
1138 	if (secs > 1) {
1139 		if (lpackets[which] > icmplim_random) {
1140 			printf("%s from %d to %d packets per second\n",
1141 			    bandlimittype[which],
1142 			    lpackets[which],
1143 			    icmplim_random
1144 			    );
1145 		}
1146 		lticks[which] = time;
1147 		lpackets[which] = 0;
1148 	}
1149 
1150 	/*
1151 	 * bump packet count
1152 	 */
1153 	if (++lpackets[which] > icmplim_random) {
1154 		/*
1155 		 * After hitting the randomized limit, we further randomize the
1156 		 * behavior of how we apply rate limitation.
1157 		 * We rate limit based on probability that increases with the
1158 		 * increase in lpackets[which] count.
1159 		 */
1160 		if ((random() % (lpackets[which] - icmplim_random)) != 0) {
1161 			return true;
1162 		}
1163 	}
1164 	return false;
1165 }
1166 
1167 #endif
1168 
1169 #if __APPLE__
1170 
1171 /*
1172  * Non-privileged ICMP socket operations
1173  * - send ICMP echo request
1174  * - all ICMP
1175  * - limited socket options
1176  */
1177 
1178 #include <netinet/ip_icmp.h>
1179 #include <netinet/in_pcb.h>
1180 
1181 extern u_int32_t rip_sendspace;
1182 extern u_int32_t rip_recvspace;
1183 extern struct inpcbinfo ripcbinfo;
1184 
1185 int rip_abort(struct socket *);
1186 int rip_bind(struct socket *, struct sockaddr *, struct proc *);
1187 int rip_connect(struct socket *, struct sockaddr *, struct proc *);
1188 int rip_detach(struct socket *);
1189 int rip_disconnect(struct socket *);
1190 int rip_shutdown(struct socket *);
1191 
1192 __private_extern__ int icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct proc *p);
1193 __private_extern__ int icmp_dgram_attach(struct socket *so, int proto, struct proc *p);
1194 __private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt);
1195 
1196 __private_extern__ struct pr_usrreqs icmp_dgram_usrreqs = {
1197 	.pru_abort =            rip_abort,
1198 	.pru_attach =           icmp_dgram_attach,
1199 	.pru_bind =             rip_bind,
1200 	.pru_connect =          rip_connect,
1201 	.pru_control =          in_control,
1202 	.pru_detach =           rip_detach,
1203 	.pru_disconnect =       rip_disconnect,
1204 	.pru_peeraddr =         in_getpeeraddr,
1205 	.pru_send =             icmp_dgram_send,
1206 	.pru_shutdown =         rip_shutdown,
1207 	.pru_sockaddr =         in_getsockaddr,
1208 	.pru_sosend =           sosend,
1209 	.pru_soreceive =        soreceive,
1210 };
1211 
1212 /* Like rip_attach but without root privilege enforcement */
1213 __private_extern__ int
icmp_dgram_attach(struct socket * so,__unused int proto,struct proc * p)1214 icmp_dgram_attach(struct socket *so, __unused int proto, struct proc *p)
1215 {
1216 	struct inpcb *inp;
1217 	int error;
1218 
1219 	inp = sotoinpcb(so);
1220 	if (inp) {
1221 		panic("icmp_dgram_attach");
1222 	}
1223 
1224 	error = soreserve(so, rip_sendspace, rip_recvspace);
1225 	if (error) {
1226 		return error;
1227 	}
1228 	error = in_pcballoc(so, &ripcbinfo, p);
1229 	if (error) {
1230 		return error;
1231 	}
1232 	inp = (struct inpcb *)so->so_pcb;
1233 	inp->inp_vflag |= INP_IPV4;
1234 	inp->inp_ip_p = IPPROTO_ICMP;
1235 	inp->inp_ip_ttl = (u_char)ip_defttl;
1236 	return 0;
1237 }
1238 
1239 /*
1240  * Raw IP socket option processing.
1241  */
1242 __private_extern__ int
icmp_dgram_ctloutput(struct socket * so,struct sockopt * sopt)1243 icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt)
1244 {
1245 	int     error;
1246 
1247 	if (sopt->sopt_level != IPPROTO_IP) {
1248 		return EINVAL;
1249 	}
1250 
1251 	switch (sopt->sopt_name) {
1252 	case IP_OPTIONS:
1253 	case IP_HDRINCL:
1254 	case IP_TOS:
1255 	case IP_TTL:
1256 	case IP_RECVOPTS:
1257 	case IP_RECVRETOPTS:
1258 	case IP_RECVDSTADDR:
1259 	case IP_RETOPTS:
1260 	case IP_MULTICAST_IF:
1261 	case IP_MULTICAST_IFINDEX:
1262 	case IP_MULTICAST_TTL:
1263 	case IP_MULTICAST_LOOP:
1264 	case IP_ADD_MEMBERSHIP:
1265 	case IP_DROP_MEMBERSHIP:
1266 	case IP_MULTICAST_VIF:
1267 	case IP_PORTRANGE:
1268 	case IP_RECVIF:
1269 	case IP_IPSEC_POLICY:
1270 	case IP_STRIPHDR:
1271 	case IP_RECVTTL:
1272 	case IP_BOUND_IF:
1273 	case IP_DONTFRAG:
1274 	case IP_NO_IFT_CELLULAR:
1275 		error = rip_ctloutput(so, sopt);
1276 		break;
1277 
1278 	default:
1279 		error = EINVAL;
1280 		break;
1281 	}
1282 
1283 	return error;
1284 }
1285 
1286 __private_extern__ int
icmp_dgram_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1287 icmp_dgram_send(struct socket *so, int flags, struct mbuf *m,
1288     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1289 {
1290 	struct ip *ip;
1291 	struct inpcb *inp = sotoinpcb(so);
1292 	int hlen;
1293 	struct icmp *icp;
1294 	struct in_ifaddr *ia = NULL;
1295 	int icmplen;
1296 	int error = EINVAL;
1297 	int inp_flags = inp ? inp->inp_flags : 0;
1298 
1299 	if (inp == NULL
1300 #if NECP
1301 	    || (necp_socket_should_use_flow_divert(inp))
1302 #endif /* NECP */
1303 	    ) {
1304 		if (inp != NULL) {
1305 			error = EPROTOTYPE;
1306 		}
1307 		goto bad;
1308 	}
1309 
1310 #if CONTENT_FILTER
1311 	/*
1312 	 * If socket is subject to Content Filter, get inp_flags from saved state
1313 	 */
1314 	if (CFIL_DGRAM_FILTERED(so) && nam == NULL) {
1315 		cfil_dgram_peek_socket_state(m, &inp_flags);
1316 	}
1317 #endif
1318 
1319 	if ((inp_flags & INP_HDRINCL) != 0) {
1320 		/* Expect 32-bit aligned data ptr on strict-align platforms */
1321 		MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
1322 		/*
1323 		 * This is not raw IP, we liberal only for fields TOS,
1324 		 * id and TTL.
1325 		 */
1326 		ip = mtod(m, struct ip *);
1327 
1328 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1329 		/* Some sanity checks */
1330 		if (m->m_pkthdr.len < hlen + ICMP_MINLEN) {
1331 			goto bad;
1332 		}
1333 		/* Only IPv4 */
1334 		if (IP_VHL_V(ip->ip_vhl) != 4) {
1335 			goto bad;
1336 		}
1337 		if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len) {
1338 			goto bad;
1339 		}
1340 		/* Bogus fragments can tie up peer resources */
1341 		if ((ip->ip_off & ~IP_DF) != 0) {
1342 			goto bad;
1343 		}
1344 		/* Allow only ICMP even for user provided IP header */
1345 		if (ip->ip_p != IPPROTO_ICMP) {
1346 			goto bad;
1347 		}
1348 		/*
1349 		 * To prevent spoofing, specified source address must
1350 		 * be one of ours.
1351 		 */
1352 		if (ip->ip_src.s_addr != INADDR_ANY) {
1353 			socket_unlock(so, 0);
1354 			lck_rw_lock_shared(&in_ifaddr_rwlock);
1355 			if (TAILQ_EMPTY(&in_ifaddrhead)) {
1356 				lck_rw_done(&in_ifaddr_rwlock);
1357 				socket_lock(so, 0);
1358 				goto bad;
1359 			}
1360 			TAILQ_FOREACH(ia, INADDR_HASH(ip->ip_src.s_addr),
1361 			    ia_hash) {
1362 				IFA_LOCK(&ia->ia_ifa);
1363 				if (IA_SIN(ia)->sin_addr.s_addr ==
1364 				    ip->ip_src.s_addr) {
1365 					IFA_UNLOCK(&ia->ia_ifa);
1366 					lck_rw_done(&in_ifaddr_rwlock);
1367 					socket_lock(so, 0);
1368 					goto ours;
1369 				}
1370 				IFA_UNLOCK(&ia->ia_ifa);
1371 			}
1372 			lck_rw_done(&in_ifaddr_rwlock);
1373 			socket_lock(so, 0);
1374 			goto bad;
1375 		}
1376 ours:
1377 		/* Do not trust we got a valid checksum */
1378 		ip->ip_sum = 0;
1379 
1380 		icp = (struct icmp *)(void *)(((char *)m->m_data) + hlen);
1381 		icmplen = m->m_pkthdr.len - hlen;
1382 	} else {
1383 		if ((icmplen = m->m_pkthdr.len) < ICMP_MINLEN) {
1384 			goto bad;
1385 		}
1386 		icp = mtod(m, struct icmp *);
1387 	}
1388 	/*
1389 	 * Allow only to send request types with code 0
1390 	 */
1391 	if (icp->icmp_code != 0) {
1392 		goto bad;
1393 	}
1394 	switch (icp->icmp_type) {
1395 	case ICMP_ECHO:
1396 		break;
1397 	case ICMP_TSTAMP:
1398 		if (icmplen != 20) {
1399 			goto bad;
1400 		}
1401 		break;
1402 	case ICMP_MASKREQ:
1403 		if (icmplen != 12) {
1404 			goto bad;
1405 		}
1406 		break;
1407 	default:
1408 		goto bad;
1409 	}
1410 	return rip_send(so, flags, m, nam, control, p);
1411 bad:
1412 	VERIFY(error != 0);
1413 
1414 	if (m != NULL) {
1415 		m_freem(m);
1416 	}
1417 	if (control != NULL) {
1418 		m_freem(control);
1419 	}
1420 
1421 	return error;
1422 }
1423 
1424 #endif /* __APPLE__ */
1425