xref: /xnu-11215.41.3/bsd/net/nat464_utils.c (revision 33de042d024d46de5ff4e89f2471de6608e37fa4)
1 /*
2  * Copyright (c) 2018-2023 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 /*
30  * Copyright (c) 2001 Daniel Hartmeier
31  * Copyright (c) 2002 - 2013 Henning Brauer
32  * NAT64 - Copyright (c) 2010 Viagenie Inc. (http://www.viagenie.ca)
33  * All rights reserved.
34  *
35  * Redistribution and use in source and binary forms, with or without
36  * modification, are permitted provided that the following conditions
37  * are met:
38  *
39  *    - Redistributions of source code must retain the above copyright
40  *	notice, this list of conditions and the following disclaimer.
41  *    - Redistributions in binary form must reproduce the above
42  *	copyright notice, this list of conditions and the following
43  *	disclaimer in the documentation and/or other materials provided
44  *	with the distribution.
45  *
46  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
47  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
48  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
49  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
50  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
51  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
52  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
53  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
54  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
56  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
57  * POSSIBILITY OF SUCH DAMAGE.
58  *
59  * Effort sponsored in part by the Defense Advanced Research Projects
60  * Agency (DARPA) and Air Force Research Laboratory, Air Force
61  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
62  *
63  */
64 #include <sys/param.h>
65 #include <sys/types.h>
66 #include <sys/mbuf.h>
67 
68 #include <net/if.h>
69 #include <net/if_types.h>
70 #include <net/dlil.h>
71 #include <net/nat464_utils.h>
72 #include <net/nwk_wq.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_var.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip6.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/icmp_var.h>
83 #include <netinet/icmp6.h>
84 #include <netinet/tcp.h>
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 #include <os/log.h>
88 
89 int clat_debug = 0;
90 
91 os_log_t nat_log_handle;
92 
93 static void
94 nat464_addr_cksum_fixup(uint16_t *, struct nat464_addr *, struct nat464_addr *,
95     protocol_family_t, protocol_family_t, uint8_t, boolean_t);
96 
97 /* Synthesize ipv6 from ipv4 */
98 int
nat464_synthesize_ipv6(ifnet_t ifp,const struct in_addr * addrv4,struct in6_addr * addr)99 nat464_synthesize_ipv6(ifnet_t ifp, const struct in_addr *addrv4, struct in6_addr *addr)
100 {
101 	static const struct in6_addr well_known_prefix = {
102 		.__u6_addr.__u6_addr8 = {0x00, 0x64, 0xff, 0x9b, 0x00, 0x00,
103 			                 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
104 			                 0x00, 0x00, 0x00, 0x00},
105 	};
106 
107 	struct ipv6_prefix nat64prefixes[NAT64_MAX_NUM_PREFIXES];
108 	int error = 0, i = 0;
109 	/* Below call is not optimized as it creates a copy of prefixes */
110 	if ((error = ifnet_get_nat64prefix(ifp, nat64prefixes)) != 0) {
111 		return error;
112 	}
113 
114 	for (i = 0; i < NAT64_MAX_NUM_PREFIXES; i++) {
115 		if (nat64prefixes[i].prefix_len != 0) {
116 			break;
117 		}
118 	}
119 
120 	VERIFY(i < NAT64_MAX_NUM_PREFIXES);
121 
122 	struct in6_addr prefix = nat64prefixes[i].ipv6_prefix;
123 	int prefix_len = nat64prefixes[i].prefix_len;
124 
125 	char *ptrv4 = (char *)__DECONST(struct in_addr *__indexable, addrv4);
126 	char *ptr = (char *)__DECONST(struct in6_addr *__indexable, addr);
127 
128 	if (IN_ZERONET(ntohl(addrv4->s_addr)) || // 0.0.0.0/8 Source hosts on local network
129 	    IN_LOOPBACK(ntohl(addrv4->s_addr)) || // 127.0.0.0/8 Loopback
130 	    IN_LINKLOCAL(ntohl(addrv4->s_addr)) || // 169.254.0.0/16 Link Local
131 	    IN_DS_LITE(ntohl(addrv4->s_addr)) || // 192.0.0.0/29 DS-Lite
132 	    IN_6TO4_RELAY_ANYCAST(ntohl(addrv4->s_addr)) || // 192.88.99.0/24 6to4 Relay Anycast
133 	    IN_MULTICAST(ntohl(addrv4->s_addr)) || // 224.0.0.0/4 Multicast
134 	    INADDR_BROADCAST == addrv4->s_addr) { // 255.255.255.255/32 Limited Broadcast
135 		return -1;
136 	}
137 
138 	/* Check for the well-known prefix */
139 	if (prefix_len == NAT64_PREFIX_LEN_96 &&
140 	    IN6_ARE_ADDR_EQUAL(&prefix, &well_known_prefix)) { // https://tools.ietf.org/html/rfc6052#section-3.1
141 		if (IN_PRIVATE(ntohl(addrv4->s_addr)) || // 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16 Private-Use
142 		    IN_SHARED_ADDRESS_SPACE(ntohl(addrv4->s_addr))) { // 100.64.0.0/10 Shared Address Space
143 			return -1;
144 		}
145 	}
146 
147 	memcpy(ptr, (char *)&prefix, prefix_len);
148 
149 	switch (prefix_len) {
150 	case NAT64_PREFIX_LEN_96:
151 		memcpy(ptr + 12, ptrv4, 4);
152 		break;
153 	case NAT64_PREFIX_LEN_64:
154 		memcpy(ptr + 9, ptrv4, 4);
155 		break;
156 	case NAT64_PREFIX_LEN_56:
157 		memcpy(ptr + 7, ptrv4, 1);
158 		memcpy(ptr + 9, ptrv4 + 1, 3);
159 		break;
160 	case NAT64_PREFIX_LEN_48:
161 		memcpy(ptr + 6, ptrv4, 2);
162 		memcpy(ptr + 9, ptrv4 + 2, 2);
163 		break;
164 	case NAT64_PREFIX_LEN_40:
165 		memcpy(ptr + 5, ptrv4, 3);
166 		memcpy(ptr + 9, ptrv4 + 3, 1);
167 		break;
168 	case NAT64_PREFIX_LEN_32:
169 		memcpy(ptr + 4, ptrv4, 4);
170 		break;
171 	default:
172 		panic("NAT64-prefix len is wrong: %u", prefix_len);
173 	}
174 
175 	if (clat_debug) {
176 		char buf[MAX_IPv6_STR_LEN];
177 		clat_log2((LOG_DEBUG, "%s synthesized  %s\n", __func__,
178 		    inet_ntop(AF_INET6, (void *)addr, buf, sizeof(buf))));
179 	}
180 
181 	return error;
182 }
183 
184 /* Synthesize ipv4 from ipv6 */
185 int
nat464_synthesize_ipv4(ifnet_t ifp,const struct in6_addr * addr,struct in_addr * addrv4)186 nat464_synthesize_ipv4(ifnet_t ifp, const struct in6_addr *addr, struct in_addr *addrv4)
187 {
188 	struct ipv6_prefix nat64prefixes[NAT64_MAX_NUM_PREFIXES];
189 	int error = 0, i = 0;
190 
191 	/* Below call is not optimized as it creates a copy of prefixes */
192 	if ((error = ifnet_get_nat64prefix(ifp, nat64prefixes)) != 0) {
193 		return error;
194 	}
195 
196 	for (i = 0; i < NAT64_MAX_NUM_PREFIXES; i++) {
197 		if (nat64prefixes[i].prefix_len != 0) {
198 			break;
199 		}
200 	}
201 
202 	VERIFY(i < NAT64_MAX_NUM_PREFIXES);
203 
204 	struct in6_addr prefix = nat64prefixes[i].ipv6_prefix;
205 	int prefix_len = nat64prefixes[i].prefix_len;
206 
207 	VERIFY(prefix_len < sizeof(prefix));
208 
209 	char *ptrv4 = (char *)__DECONST(struct in_addr *__indexable, addrv4);
210 	char *ptr = (char *)__DECONST(struct in6_addr *__indexable, addr);
211 
212 	/* -fbounds-safety:
213 	 * Override a warning about prefix_len being > 16 because
214 	 * we already checked that above.
215 	 */
216 	if (memcmp((const struct in6_addr *__indexable)addr, &prefix, prefix_len) != 0) {
217 		return -1;
218 	}
219 
220 	switch (prefix_len) {
221 	case NAT64_PREFIX_LEN_96:
222 		memcpy(ptrv4, ptr + 12, 4);
223 		break;
224 	case NAT64_PREFIX_LEN_64:
225 		memcpy(ptrv4, ptr + 9, 4);
226 		break;
227 	case NAT64_PREFIX_LEN_56:
228 		memcpy(ptrv4, ptr + 7, 1);
229 		memcpy(ptrv4 + 1, ptr + 9, 3);
230 		break;
231 	case NAT64_PREFIX_LEN_48:
232 		memcpy(ptrv4, ptr + 6, 2);
233 		memcpy(ptrv4 + 2, ptr + 9, 2);
234 		break;
235 	case NAT64_PREFIX_LEN_40:
236 		memcpy(ptrv4, ptr + 5, 3);
237 		memcpy(ptrv4 + 3, ptr + 9, 1);
238 		break;
239 	case NAT64_PREFIX_LEN_32:
240 		memcpy(ptrv4, ptr + 4, 4);
241 		break;
242 	default:
243 		panic("NAT64-prefix len is wrong: %u",
244 		    prefix_len);
245 	}
246 
247 	if (clat_debug) {
248 		char buf[MAX_IPv4_STR_LEN];
249 		clat_log2((LOG_DEBUG, "%s desynthesized to %s\n", __func__,
250 		    inet_ntop(AF_INET, (void *)addrv4, buf, sizeof(buf))));
251 	}
252 	return error;
253 }
254 
255 #define PTR_IP(field)   ((int32_t)offsetof(struct ip, field))
256 #define PTR_IP6(field)  ((int32_t)offsetof(struct ip6_hdr, field))
257 
258 /*
259  *  Translate the ICMP header
260  */
261 int
nat464_translate_icmp(int naf,void * arg)262 nat464_translate_icmp(int naf, void *arg)
263 {
264 	struct icmp             *__single icmp4;
265 	struct icmp6_hdr        *__single icmp6;
266 	uint32_t                 mtu;
267 	int32_t                  ptr = -1;
268 	uint8_t          type;
269 	uint8_t          code;
270 
271 	switch (naf) {
272 	case AF_INET:
273 		icmp6 = arg;
274 		type  = icmp6->icmp6_type;
275 		code  = icmp6->icmp6_code;
276 		mtu   = ntohl(icmp6->icmp6_mtu);
277 
278 		switch (type) {
279 		case ICMP6_ECHO_REQUEST:
280 			type = ICMP_ECHO;
281 			break;
282 		case ICMP6_ECHO_REPLY:
283 			type = ICMP_ECHOREPLY;
284 			break;
285 		case ICMP6_DST_UNREACH:
286 			type = ICMP_UNREACH;
287 			switch (code) {
288 			case ICMP6_DST_UNREACH_NOROUTE:
289 			case ICMP6_DST_UNREACH_BEYONDSCOPE:
290 			case ICMP6_DST_UNREACH_ADDR:
291 				code = ICMP_UNREACH_HOST;
292 				break;
293 			case ICMP6_DST_UNREACH_ADMIN:
294 				code = ICMP_UNREACH_HOST_PROHIB;
295 				break;
296 			case ICMP6_DST_UNREACH_NOPORT:
297 				code = ICMP_UNREACH_PORT;
298 				break;
299 			default:
300 				return -1;
301 			}
302 			break;
303 		case ICMP6_PACKET_TOO_BIG:
304 			type = ICMP_UNREACH;
305 			code = ICMP_UNREACH_NEEDFRAG;
306 			mtu -= 20;
307 			break;
308 		case ICMP6_TIME_EXCEEDED:
309 			type = ICMP_TIMXCEED;
310 			break;
311 		case ICMP6_PARAM_PROB:
312 			switch (code) {
313 			case ICMP6_PARAMPROB_HEADER:
314 				type = ICMP_PARAMPROB;
315 				code = ICMP_PARAMPROB_ERRATPTR;
316 				ptr  = ntohl(icmp6->icmp6_pptr);
317 
318 				if (ptr == PTR_IP6(ip6_vfc)) {
319 					; /* preserve */
320 				} else if (ptr == PTR_IP6(ip6_vfc) + 1) {
321 					ptr = PTR_IP(ip_tos);
322 				} else if (ptr == PTR_IP6(ip6_plen) ||
323 				    ptr == PTR_IP6(ip6_plen) + 1) {
324 					ptr = PTR_IP(ip_len);
325 				} else if (ptr == PTR_IP6(ip6_nxt)) {
326 					ptr = PTR_IP(ip_p);
327 				} else if (ptr == PTR_IP6(ip6_hlim)) {
328 					ptr = PTR_IP(ip_ttl);
329 				} else if (ptr >= PTR_IP6(ip6_src) &&
330 				    ptr < PTR_IP6(ip6_dst)) {
331 					ptr = PTR_IP(ip_src);
332 				} else if (ptr >= PTR_IP6(ip6_dst) &&
333 				    ptr < (int32_t)sizeof(struct ip6_hdr)) {
334 					ptr = PTR_IP(ip_dst);
335 				} else {
336 					return -1;
337 				}
338 				break;
339 			case ICMP6_PARAMPROB_NEXTHEADER:
340 				type = ICMP_UNREACH;
341 				code = ICMP_UNREACH_PROTOCOL;
342 				break;
343 			default:
344 				return -1;
345 			}
346 			break;
347 		default:
348 			return -1;
349 		}
350 		icmp6->icmp6_type = type;
351 		icmp6->icmp6_code = code;
352 		/* aligns well with a icmpv4 nextmtu */
353 		icmp6->icmp6_mtu = htonl(mtu);
354 		/* icmpv4 pptr is a one most significant byte */
355 		if (ptr >= 0) {
356 			icmp6->icmp6_pptr = htonl(ptr << 24);
357 		}
358 		break;
359 
360 	case AF_INET6:
361 		icmp4 = arg;
362 		type  = icmp4->icmp_type;
363 		code  = icmp4->icmp_code;
364 		mtu   = ntohs(icmp4->icmp_nextmtu);
365 
366 		switch (type) {
367 		case ICMP_ECHO:
368 			type = ICMP6_ECHO_REQUEST;
369 			break;
370 		case ICMP_ECHOREPLY:
371 			type = ICMP6_ECHO_REPLY;
372 			break;
373 		case ICMP_UNREACH:
374 			type = ICMP6_DST_UNREACH;
375 			switch (code) {
376 			case ICMP_UNREACH_NET:
377 			case ICMP_UNREACH_HOST:
378 			case ICMP_UNREACH_NET_UNKNOWN:
379 			case ICMP_UNREACH_HOST_UNKNOWN:
380 			case ICMP_UNREACH_ISOLATED:
381 			case ICMP_UNREACH_TOSNET:
382 			case ICMP_UNREACH_TOSHOST:
383 				code = ICMP6_DST_UNREACH_NOROUTE;
384 				break;
385 			case ICMP_UNREACH_PORT:
386 				code = ICMP6_DST_UNREACH_NOPORT;
387 				break;
388 			case ICMP_UNREACH_NET_PROHIB:
389 			case ICMP_UNREACH_HOST_PROHIB:
390 			case ICMP_UNREACH_FILTER_PROHIB:
391 			case ICMP_UNREACH_PRECEDENCE_CUTOFF:
392 				code = ICMP6_DST_UNREACH_ADMIN;
393 				break;
394 			case ICMP_UNREACH_PROTOCOL:
395 				type = ICMP6_PARAM_PROB;
396 				code = ICMP6_PARAMPROB_NEXTHEADER;
397 				ptr  = offsetof(struct ip6_hdr, ip6_nxt);
398 				break;
399 			case ICMP_UNREACH_NEEDFRAG:
400 				type = ICMP6_PACKET_TOO_BIG;
401 				code = 0;
402 				/*
403 				 * Make sure we don't overflow adjusting for
404 				 * translation overhead.
405 				 * If we do, just work with a lower mtu as is.
406 				 */
407 				if (mtu <= (UINT16_MAX - CLAT46_HDR_EXPANSION_OVERHD)) {
408 					mtu += CLAT46_HDR_EXPANSION_OVERHD;
409 				}
410 				break;
411 			default:
412 				return -1;
413 			}
414 			break;
415 		case ICMP_TIMXCEED:
416 			type = ICMP6_TIME_EXCEEDED;
417 			break;
418 		case ICMP_PARAMPROB:
419 			type = ICMP6_PARAM_PROB;
420 			switch (code) {
421 			case ICMP_PARAMPROB_ERRATPTR:
422 				code = ICMP6_PARAMPROB_HEADER;
423 				break;
424 			case ICMP_PARAMPROB_LENGTH:
425 				code = ICMP6_PARAMPROB_HEADER;
426 				break;
427 			default:
428 				return -1;
429 			}
430 
431 			ptr = icmp4->icmp_pptr;
432 			if (ptr == 0 || ptr == PTR_IP(ip_tos)) {
433 				; /* preserve */
434 			} else if (ptr == PTR_IP(ip_len) ||
435 			    ptr == PTR_IP(ip_len) + 1) {
436 				ptr = PTR_IP6(ip6_plen);
437 			} else if (ptr == PTR_IP(ip_ttl)) {
438 				ptr = PTR_IP6(ip6_hlim);
439 			} else if (ptr == PTR_IP(ip_p)) {
440 				ptr = PTR_IP6(ip6_nxt);
441 			} else if (ptr >= PTR_IP(ip_src) &&
442 			    ptr < PTR_IP(ip_dst)) {
443 				ptr = PTR_IP6(ip6_src);
444 			} else if (ptr >= PTR_IP(ip_dst) &&
445 			    ptr < (int32_t)sizeof(struct ip)) {
446 				ptr = PTR_IP6(ip6_dst);
447 			} else {
448 				return -1;
449 			}
450 			break;
451 		default:
452 			return -1;
453 		}
454 		icmp4->icmp_type = type;
455 		icmp4->icmp_code = code;
456 		icmp4->icmp_nextmtu = htons((uint16_t)mtu);
457 
458 		if (ptr >= 0) {
459 			icmp4->icmp_void = htonl(ptr);
460 		}
461 		break;
462 	}
463 
464 	return 0;
465 }
466 
467 /*
468  * @brief This routine is called to perform address family translation on the
469  *     inner IP header (that may come as payload) of an ICMP(v4/v6) error
470  *     response.
471  *
472  * @param pbuf Pointer to packet buffer
473  * @param off Points to end of ICMP header
474  * @param tot_len Pointer to total length of the outer IP header
475  * @param off2 Points to end of inner IP header
476  * @param proto2 Inner IP proto field
477  * @param ttl2 Inner IP ttl field
478  * @param tot_len2 Inner IP total length
479  * @param src Pointer to the generic v4/v6 src address
480  * @param dst Pointer to the generic v4/v6 dst address
481  * @param af Old protocol family
482  * @param naf New protocol family
483  *
484  * @return -1 on error and 0 on success
485  */
486 int
nat464_translate_icmp_ip(pbuf_t * pbuf,uint16_t off,uint16_t * tot_len,uint16_t * off2,uint8_t proto2,uint8_t ttl2,uint16_t tot_len2,struct nat464_addr * src,struct nat464_addr * dst,protocol_family_t af,protocol_family_t naf)487 nat464_translate_icmp_ip(pbuf_t *pbuf, uint16_t off, uint16_t *tot_len, uint16_t *off2,
488     uint8_t proto2, uint8_t ttl2, uint16_t tot_len2, struct nat464_addr *src,
489     struct nat464_addr *dst, protocol_family_t af, protocol_family_t naf)
490 {
491 	struct ip *__single ip4 = NULL;
492 	struct ip6_hdr *__single ip6 = NULL;
493 	void *__single hdr = NULL;
494 	int hlen = 0, olen = 0;
495 	uint64_t ipid_salt = (uint64_t)pbuf_get_packet_buffer_address(pbuf);
496 
497 	if (af == naf || (af != AF_INET && af != AF_INET6) ||
498 	    (naf != AF_INET && naf != AF_INET6)) {
499 		return -1;
500 	}
501 
502 	/* old header */
503 	olen = *off2 - off;
504 	/* new header */
505 	hlen = naf == PF_INET ? sizeof(*ip4) : sizeof(*ip6);
506 
507 	/* Modify the pbuf to accommodate the new header */
508 	hdr = pbuf_resize_segment(pbuf, off, olen, hlen);
509 	if (hdr == NULL) {
510 		return -1;
511 	}
512 
513 	/* translate inner ip/ip6 header */
514 	switch (naf) {
515 	case AF_INET:
516 		ip4 = hdr;
517 		bzero(ip4, sizeof(*ip4));
518 		ip4->ip_v = IPVERSION;
519 		ip4->ip_hl = sizeof(*ip4) >> 2;
520 		ip4->ip_len = htons((uint16_t)(sizeof(*ip4) + tot_len2 - olen));
521 		ip4->ip_id = rfc6864 ? 0 : htons(ip_randomid(ipid_salt));
522 		ip4->ip_off = htons(IP_DF);
523 		ip4->ip_ttl = ttl2;
524 		if (proto2 == IPPROTO_ICMPV6) {
525 			ip4->ip_p = IPPROTO_ICMP;
526 		} else {
527 			ip4->ip_p = proto2;
528 		}
529 		ip4->ip_src = src->natv4addr;
530 		ip4->ip_dst = dst->natv4addr;
531 		ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
532 
533 		if (clat_debug) {
534 			char buf[MAX_IPv4_STR_LEN];
535 			clat_log2((LOG_DEBUG, "%s translated to IPv4 (inner) "
536 			    "ip_len: %#x ip_p: %d ip_sum: %#x ip_src: %s ip_dst: %s \n",
537 			    __func__, ntohs(ip4->ip_len), ip4->ip_p, ntohs(ip4->ip_sum),
538 			    inet_ntop(AF_INET, (void *)&ip4->ip_src, buf, sizeof(buf)),
539 			    inet_ntop(AF_INET, (void *)&ip4->ip_dst, buf, sizeof(buf))));
540 		}
541 		break;
542 	case AF_INET6:
543 		ip6 = hdr;
544 		bzero(ip6, sizeof(*ip6));
545 		ip6->ip6_vfc  = IPV6_VERSION;
546 		ip6->ip6_plen = htons((uint16_t)(tot_len2 - olen));
547 		if (proto2 == IPPROTO_ICMP) {
548 			ip6->ip6_nxt = IPPROTO_ICMPV6;
549 		} else {
550 			ip6->ip6_nxt = proto2;
551 		}
552 		if (!ttl2 || ttl2 > IPV6_DEFHLIM) {
553 			ip6->ip6_hlim = IPV6_DEFHLIM;
554 		} else {
555 			ip6->ip6_hlim = ttl2;
556 		}
557 		ip6->ip6_src  = src->natv6addr;
558 		ip6->ip6_dst  = dst->natv6addr;
559 
560 		if (clat_debug) {
561 			char buf2[MAX_IPv6_STR_LEN];
562 			clat_log2((LOG_DEBUG, "%s translated to IPv6 (inner) "
563 			    "ip6_plen: %#x ip6_nxt: %d ip6_src: %s ip6_dst: %s \n",
564 			    __func__, ntohs(ip6->ip6_plen), ip6->ip6_nxt,
565 			    inet_ntop(AF_INET6, (void *)&ip6->ip6_src, buf2, sizeof(buf2)),
566 			    inet_ntop(AF_INET6, (void *)&ip6->ip6_dst, buf2, sizeof(buf2))));
567 		}
568 		break;
569 	}
570 
571 	/* adjust payload offset and total packet length */
572 	*off2 += hlen - olen;
573 	*tot_len += hlen - olen;
574 
575 	return 0;
576 }
577 /*
578  * @brief The function inserts IPv6 fragmentation header
579  *     and populates it with the passed parameters.
580  *
581  * @param pbuf Pointer to the packet buffer
582  * @param ip_id IP identifier (in network byte order)
583  * @param frag_offset Fragment offset (in network byte order)
584  * @param is_last_frag Boolean indicating if the fragment header is for
585  *     last fragment or not.
586  *
587  * @return -1 on error and 0 on success.
588  */
589 int
nat464_insert_frag46(pbuf_t * pbuf,uint16_t ip_id_val,uint16_t frag_offset,boolean_t is_last_frag)590 nat464_insert_frag46(pbuf_t *pbuf, uint16_t ip_id_val, uint16_t frag_offset,
591     boolean_t is_last_frag)
592 {
593 	struct ip6_frag *p_ip6_frag = NULL;
594 	struct ip6_hdr *p_ip6h = NULL;
595 
596 	/* Insert IPv6 fragmentation header */
597 	if (pbuf_resize_segment(pbuf, sizeof(struct ip6_hdr), 0,
598 	    sizeof(struct ip6_frag)) == NULL) {
599 		return -1;
600 	}
601 
602 	p_ip6h = mtod(pbuf->pb_mbuf, struct ip6_hdr *);
603 	p_ip6_frag = (struct ip6_frag *)pbuf_contig_segment(pbuf,
604 	    sizeof(struct ip6_hdr), sizeof(struct ip6_frag));
605 
606 	if (p_ip6_frag == NULL) {
607 		return -1;
608 	}
609 
610 	/* Populate IPv6 fragmentation header */
611 	p_ip6_frag->ip6f_nxt = p_ip6h->ip6_nxt;
612 	p_ip6_frag->ip6f_reserved = 0;
613 	p_ip6_frag->ip6f_offlg = (uint16_t)(frag_offset << 3);
614 	if (!is_last_frag) {
615 		p_ip6_frag->ip6f_offlg |= 0x1;
616 	}
617 	p_ip6_frag->ip6f_offlg = htons(p_ip6_frag->ip6f_offlg);
618 	p_ip6_frag->ip6f_ident = ip_id_val;
619 
620 	/* Update IPv6 header */
621 	p_ip6h->ip6_nxt = IPPROTO_FRAGMENT;
622 	p_ip6h->ip6_plen = htons(ntohs(p_ip6h->ip6_plen) +
623 	    sizeof(struct ip6_frag));
624 
625 	return 0;
626 }
627 
628 int
nat464_translate_64(pbuf_t * pbuf,int off,uint8_t tos,uint8_t * proto,uint8_t ttl,struct in_addr src_v4,struct in_addr dst_v4,uint64_t tot_len,boolean_t * p_is_first_frag)629 nat464_translate_64(pbuf_t *pbuf, int off, uint8_t tos,
630     uint8_t *proto, uint8_t ttl, struct in_addr src_v4,
631     struct in_addr dst_v4, uint64_t tot_len, boolean_t *p_is_first_frag)
632 {
633 	struct ip *ip4;
634 	struct ip6_frag *p_frag6 = NULL;
635 	struct ip6_frag frag6 = {};
636 	boolean_t is_frag = FALSE;
637 	uint16_t ip_frag_off = 0;
638 
639 	/*
640 	 * ip_input asserts for rcvif to be not NULL
641 	 * That may not be true for two corner cases
642 	 * 1. If for some reason a local app sends DNS
643 	 * AAAA query to local host
644 	 * 2. If IPv6 stack in kernel internally generates a
645 	 * message destined for a synthesized IPv6 end-point.
646 	 */
647 	if (pbuf->pb_ifp == NULL) {
648 		return NT_DROP;
649 	}
650 
651 	if (*proto == IPPROTO_FRAGMENT) {
652 		p_frag6 = (struct ip6_frag *)pbuf_contig_segment(pbuf,
653 		    sizeof(struct ip6_hdr), sizeof(struct ip6_frag));
654 		if (p_frag6 == NULL) {
655 			ip6stat.ip6s_clat464_in_64frag_transfail_drop++;
656 			return NT_DROP;
657 		}
658 
659 		frag6 = *p_frag6;
660 		p_frag6 = NULL;
661 		*proto = frag6.ip6f_nxt;
662 		off += sizeof(struct ip6_frag);
663 		is_frag = TRUE;
664 		ip_frag_off = (ntohs(frag6.ip6f_offlg & IP6F_OFF_MASK)) >> 3;
665 		if (ip_frag_off != 0) {
666 			*p_is_first_frag = FALSE;
667 		}
668 	}
669 
670 	ip4 = (struct ip *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip4));
671 	if (ip4 == NULL) {
672 		return NT_DROP;
673 	}
674 	ip4->ip_v   = 4;
675 	ip4->ip_hl  = 5;
676 	ip4->ip_tos = tos;
677 	ip4->ip_len = htons((uint16_t)(sizeof(*ip4) + (tot_len - off)));
678 	ip4->ip_id  = 0;
679 	ip4->ip_off = 0;
680 	ip4->ip_ttl = ttl;
681 	ip4->ip_p   = *proto;
682 	ip4->ip_sum = 0;
683 	ip4->ip_src = src_v4;
684 	ip4->ip_dst = dst_v4;
685 	if (is_frag) {
686 		/*
687 		 * https://tools.ietf.org/html/rfc7915#section-5.1.1
688 		 * Identification:  Copied from the low-order 16 bits in the
689 		 * Identification field in the Fragment Header.
690 		 */
691 		ip4->ip_id = ntohl(frag6.ip6f_ident) & 0xffff;
692 		ip4->ip_id = htons(ip4->ip_id);
693 		if (frag6.ip6f_offlg & IP6F_MORE_FRAG) {
694 			ip_frag_off |= IP_MF;
695 		}
696 		ip4->ip_off = htons(ip_frag_off);
697 	} else {
698 		ip4->ip_off |= htons(IP_DF);
699 	}
700 
701 	/*
702 	 * Defer calculating ip_sum for ICMPv6 as we do it
703 	 * later in Protocol translation
704 	 */
705 	if (*proto != IPPROTO_ICMPV6) {
706 		ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
707 	}
708 
709 	if (clat_debug) {
710 		char buf1[MAX_IPv4_STR_LEN], buf2[MAX_IPv4_STR_LEN];
711 		clat_log2((LOG_DEBUG, "%s translated to IPv4 ip_len: %#x "
712 		    "ip_p: %d ip_sum: %#x ip_src: %s ip_dst: %s \n", __func__,
713 		    ntohs(ip4->ip_len), ip4->ip_p, ntohs(ip4->ip_sum),
714 		    inet_ntop(AF_INET, (void *)&ip4->ip_src, buf1, sizeof(buf1)),
715 		    inet_ntop(AF_INET, (void *)&ip4->ip_dst, buf2, sizeof(buf2))));
716 	}
717 	return NT_NAT64;
718 }
719 /*
720  * @brief The routine translates the IPv4 header to IPv6 header.
721  *
722  * @param pbuf Pointer to the generic packet buffer
723  * @param off Offset to the end of IP header
724  * @param tos Type of service
725  * @param proto Protocol running over IP
726  * @param ttl Time to live
727  * @param src_v6 Source IPv6 address
728  * @param dst_v6 Destination IPv6 address
729  * @param tot_len Total payload length
730  *
731  * @return NT_NAT64 if IP header translation is successful, else error
732  */
733 int
nat464_translate_46(pbuf_t * pbuf,uint16_t off,uint8_t tos,uint8_t proto,uint8_t ttl,struct in6_addr src_v6,struct in6_addr dst_v6,uint16_t tot_len)734 nat464_translate_46(pbuf_t *pbuf, uint16_t off, uint8_t tos,
735     uint8_t proto, uint8_t ttl, struct in6_addr src_v6,
736     struct in6_addr dst_v6, uint16_t tot_len)
737 {
738 	struct ip6_hdr *ip6;
739 
740 	if (pbuf->pb_ifp == NULL) {
741 		return NT_DROP;
742 	}
743 
744 	/*
745 	 * Trim the buffer from head of size equal to to off (which is equal to
746 	 * the size of IP header and prepend IPv6 header length to the buffer
747 	 */
748 	ip6 = (struct ip6_hdr *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip6));
749 	if (ip6 == NULL) {
750 		return NT_DROP;
751 	}
752 	ip6->ip6_flow = htonl((6 << 28) | (tos << 20));
753 	ip6->ip6_plen = htons(tot_len - off);
754 	ip6->ip6_nxt  = proto;
755 	ip6->ip6_hlim = ttl;
756 	ip6->ip6_src = src_v6;
757 	ip6->ip6_dst = dst_v6;
758 
759 	if (clat_debug) {
760 		char buf1[MAX_IPv6_STR_LEN], buf2[MAX_IPv6_STR_LEN];
761 		clat_log2((LOG_DEBUG, "%s translated to IPv6 ip6_plen: %#x "
762 		    " ip6_nxt: %d ip6_src: %s ip6_dst: %s \n", __func__,
763 		    ntohs(ip6->ip6_plen), ip6->ip6_nxt,
764 		    inet_ntop(AF_INET6, (void *)&ip6->ip6_src, buf1, sizeof(buf1)),
765 		    inet_ntop(AF_INET6, (void *)&ip6->ip6_dst, buf2, sizeof(buf2))));
766 	}
767 	return NT_NAT64;
768 }
769 
770 /* Handle the next protocol checksum */
771 /*
772  * @brief This routine translates the Proto running over IP and updates the checksum
773  *     for IP header translation. It also updates pbuf checksum flags and related fields.
774  *
775  * @param pbuf Pointer to protocol buffer
776  * @param nsrc New source address
777  * @param ndst New destination address
778  * @param af Old family
779  * @param naf New family
780  *
781  * @return void
782  */
783 int
nat464_translate_proto(pbuf_t * pbuf,struct nat464_addr * osrc,struct nat464_addr * odst,uint8_t oproto,protocol_family_t af,protocol_family_t naf,int direction,boolean_t only_csum)784 nat464_translate_proto(pbuf_t *pbuf, struct nat464_addr *osrc,
785     struct nat464_addr *odst, uint8_t oproto, protocol_family_t af,
786     protocol_family_t naf, int direction, boolean_t only_csum)
787 {
788 	struct ip *iph = NULL;
789 	struct ip6_hdr *ip6h = NULL;
790 	uint16_t hlen = 0, plen = 0;
791 	uint16_t tot_len = 0;
792 	void *nsrc = NULL, *ndst = NULL;
793 	uint8_t *proto = 0;
794 	uint16_t *psum = NULL;
795 	boolean_t do_ones_complement = FALSE;
796 
797 	/* For now these routines only support 464 translations */
798 	VERIFY(af != naf);
799 	VERIFY(af == PF_INET || af == PF_INET6);
800 
801 	/*
802 	 * For now out must be for v4 to v6 translation
803 	 * and in must be for v6 to v4 translation.
804 	 */
805 	switch (naf) {
806 	case PF_INET: {
807 		iph = pbuf->pb_data;
808 		hlen = (uint16_t)(iph->ip_hl << 2);
809 		plen = ntohs(iph->ip_len) - hlen;
810 		tot_len = ntohs(iph->ip_len);
811 		nsrc = &iph->ip_src;
812 		ndst = &iph->ip_dst;
813 		proto = &iph->ip_p;
814 		break;
815 	}
816 	case PF_INET6: {
817 		ip6h = pbuf->pb_data;
818 		hlen = (uint16_t)sizeof(*ip6h);
819 		plen = ntohs(ip6h->ip6_plen);
820 		tot_len = hlen + plen;
821 		nsrc = &ip6h->ip6_src;
822 		ndst = &ip6h->ip6_dst;
823 		proto = &ip6h->ip6_nxt;
824 		break;
825 	}
826 	default:
827 		return NT_DROP; /* We should never come here */
828 	}
829 
830 	if (*proto != oproto) {
831 		return NT_DROP;
832 	}
833 
834 	/*
835 	 * We may want to manipulate csum flags in some cases
836 	 * and not act on the protocol header as it may not
837 	 * carry protocol checksums.
838 	 * For example, fragments other than the first one would
839 	 * not carry protocol headers.
840 	 */
841 	if (only_csum) {
842 		/*
843 		 * Only translate ICMP proto in the header
844 		 * and adjust checksums
845 		 */
846 		if (*proto == IPPROTO_ICMP) {
847 			if (naf != PF_INET6) {
848 				return NT_DROP;
849 			}
850 
851 			*proto = IPPROTO_ICMPV6;
852 		} else if (*proto == IPPROTO_ICMPV6) {
853 			if (naf != PF_INET) {
854 				return NT_DROP;
855 			}
856 
857 			*proto = IPPROTO_ICMP;
858 			/* Recalculate IP checksum as proto field has changed */
859 			iph->ip_sum = 0;
860 			iph->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, hlen);
861 		}
862 		goto done;
863 	}
864 
865 	switch (*proto) {
866 	case IPPROTO_UDP: {
867 		struct udphdr *uh = (struct udphdr *)pbuf_contig_segment(pbuf, hlen,
868 		    sizeof(*uh));
869 
870 		if (uh == NULL) {
871 			return NT_DROP;
872 		}
873 
874 		if (!(*pbuf->pb_csum_flags & (CSUM_UDP | CSUM_PARTIAL)) &&
875 		    uh->uh_sum == 0 && af == PF_INET && naf == PF_INET6) {
876 			uh->uh_sum = pbuf_inet6_cksum(pbuf, IPPROTO_UDP,
877 			    hlen, ntohs(ip6h->ip6_plen));
878 			if (uh->uh_sum == 0) {
879 				uh->uh_sum = 0xffff;
880 			}
881 			goto done;
882 		}
883 
884 		psum = &uh->uh_sum;
885 		break;
886 	}
887 	case IPPROTO_TCP: {
888 		struct tcphdr *th = (struct tcphdr *)pbuf_contig_segment(pbuf, hlen,
889 		    sizeof(*th));
890 
891 		if (th == NULL) {
892 			return NT_DROP;
893 		}
894 
895 		psum = &th->th_sum;
896 		break;
897 	}
898 	}
899 
900 	/*
901 	 * Translate the protocol header, update IP header if needed,
902 	 * calculate checksums and update the checksum flags.
903 	 */
904 	switch (*proto) {
905 	case IPPROTO_UDP:
906 	/* Fall through */
907 	case IPPROTO_TCP:
908 	{
909 		/*
910 		 * If it is a locally generated and has CSUM flags set
911 		 * for TCP and UDP it means we have pseudo header checksum
912 		 * that has not yet been one's complemented.
913 		 */
914 		if (direction == NT_OUT &&
915 		    (*pbuf->pb_csum_flags & CSUM_PARTIAL)) {
916 			do_ones_complement = TRUE;
917 		}
918 
919 		nat464_addr_cksum_fixup(psum, osrc, (struct nat464_addr *)nsrc,
920 		    af, naf, (*proto == IPPROTO_UDP) ? 1 : 0, do_ones_complement);
921 		nat464_addr_cksum_fixup(psum, odst, (struct nat464_addr *)ndst,
922 		    af, naf, (*proto == IPPROTO_UDP) ? 1 : 0, do_ones_complement);
923 
924 		break;
925 	}
926 	case IPPROTO_ICMP: {
927 		if (naf != PF_INET6) {  /* allow only v6 as naf for ICMP */
928 			return NT_DROP;
929 		}
930 
931 		struct icmp *__single icmph = NULL;
932 		struct icmp6_hdr *__single icmp6h = NULL;
933 		uint16_t ip2off = 0, hlen2 = 0, tot_len2 = 0;
934 
935 		icmph = (struct icmp*) pbuf_contig_segment(pbuf, hlen,
936 		    ICMP_MINLEN);
937 		if (icmph == NULL) {
938 			return NT_DROP;
939 		}
940 
941 		/* Translate the ICMP header */
942 		if (nat464_translate_icmp(PF_INET6, icmph) != 0) {
943 			return NT_DROP;
944 		}
945 
946 		*proto = IPPROTO_ICMPV6;
947 		icmp6h = (struct icmp6_hdr *__single)(void *)icmph;
948 		pbuf_copy_back(pbuf, hlen, sizeof(struct icmp6_hdr),
949 		    icmp6h, sizeof(*icmp6h));
950 
951 		/*Translate the inner IP header only for error messages */
952 		if (ICMP6_ERRORTYPE(icmp6h->icmp6_type)) {
953 			ip2off = (uint16_t)(hlen + sizeof(*icmp6h));
954 			struct ip *iph2 = NULL;
955 			iph2 = (struct ip*) pbuf_contig_segment(pbuf, ip2off,
956 			    sizeof(*iph2));
957 			if (iph2 == NULL) {
958 				return NT_DROP;
959 			}
960 
961 			hlen2 = (uint16_t)(ip2off + (iph2->ip_hl << 2));
962 			tot_len2 = ntohs(iph2->ip_len);
963 
964 			/* Destination in outer IP should be Source in inner IP */
965 			VERIFY(IN_ARE_ADDR_EQUAL(&odst->natv4addr, &iph2->ip_src));
966 			if (nat464_translate_icmp_ip(pbuf, ip2off, &tot_len,
967 			    &hlen2, iph2->ip_p, iph2->ip_ttl, tot_len2,
968 			    (struct nat464_addr *)ndst, (struct nat464_addr *)nsrc,
969 			    PF_INET, PF_INET6) != 0) {
970 				return NT_DROP;
971 			}
972 			/* Update total length/payload length for outer header */
973 			switch (naf) {
974 			case PF_INET:
975 				iph->ip_len = htons(tot_len);
976 				break;
977 			case PF_INET6:
978 				ip6h->ip6_plen = htons(tot_len - hlen);
979 				break;
980 			}
981 			iph2 = NULL;
982 		}
983 
984 		icmp6h->icmp6_cksum = 0;
985 		icmp6h->icmp6_cksum = pbuf_inet6_cksum(pbuf, IPPROTO_ICMPV6, hlen,
986 		    ntohs(ip6h->ip6_plen));
987 
988 		clat_log2((LOG_DEBUG, "%s translated to ICMPV6 type: %d "
989 		    "code: %d checksum: %#x \n", __func__, icmp6h->icmp6_type,
990 		    icmp6h->icmp6_code, icmp6h->icmp6_cksum));
991 
992 		icmph = NULL;
993 		icmp6h = NULL;
994 		break;
995 	}
996 	case IPPROTO_ICMPV6:
997 	{       if (naf != PF_INET) {           /* allow only v4 as naf for ICMPV6 */
998 			return NT_DROP;
999 		}
1000 
1001 		struct icmp6_hdr *__single icmp6h = NULL;
1002 		struct icmp *__single icmph = NULL;
1003 		uint16_t ip2off = 0, hlen2 = 0, tot_len2 = 0;
1004 
1005 		icmp6h = (struct icmp6_hdr*) pbuf_contig_segment(pbuf, hlen,
1006 		    sizeof(*icmp6h));
1007 		if (icmp6h == NULL) {
1008 			return NT_DROP;
1009 		}
1010 
1011 		/* Translate the ICMP header */
1012 		if (nat464_translate_icmp(PF_INET, icmp6h) != 0) {
1013 			return NT_DROP;
1014 		}
1015 
1016 		*proto = IPPROTO_ICMP;
1017 		icmph = (struct icmp *__single)(void *)icmp6h;
1018 		pbuf_copy_back(pbuf, hlen, ICMP_MINLEN,
1019 		    icmph, sizeof(*icmph));
1020 
1021 		/*Translate the inner IP header only for error messages */
1022 		if (ICMP_ERRORTYPE(icmph->icmp_type)) {
1023 			ip2off = hlen + ICMP_MINLEN;
1024 			struct ip6_hdr *iph2 = NULL;
1025 			iph2 = (struct ip6_hdr*) pbuf_contig_segment(pbuf, ip2off,
1026 			    sizeof(*iph2));
1027 			if (iph2 == NULL) {
1028 				return NT_DROP;
1029 			}
1030 
1031 			/* hlen2 points to end of inner IP header from the beginning */
1032 			hlen2 = ip2off + sizeof(struct ip6_hdr);
1033 			tot_len2 = ntohs(iph2->ip6_plen) + sizeof(struct ip6_hdr);
1034 
1035 			if (nat464_translate_icmp_ip(pbuf, ip2off, &tot_len,
1036 			    &hlen2, iph2->ip6_nxt, iph2->ip6_hlim, tot_len2,
1037 			    (struct nat464_addr *)ndst, (struct nat464_addr *)nsrc,
1038 			    PF_INET6, PF_INET) != 0) {
1039 				return NT_DROP;
1040 			}
1041 
1042 			/* Update total length for outer header */
1043 			switch (naf) {
1044 			case PF_INET:
1045 				iph->ip_len = htons(tot_len);
1046 				break;
1047 			case PF_INET6:
1048 				ip6h->ip6_plen = htons(tot_len - hlen);
1049 				break;
1050 			}
1051 			iph2 = NULL;
1052 		}
1053 		/* Recalculate IP checksum as some IP fields might have changed */
1054 		iph->ip_sum = 0;
1055 		iph->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, iph->ip_hl << 2);
1056 		icmph->icmp_cksum = 0;
1057 		icmph->icmp_cksum = pbuf_inet_cksum(pbuf, 0, hlen,
1058 		    ntohs(iph->ip_len) - hlen);
1059 
1060 		clat_log2((LOG_DEBUG, "%s translated to ICMP type: %d "
1061 		    "code: %d checksum: %#x \n", __func__, icmph->icmp_type,
1062 		    icmph->icmp_code, icmph->icmp_cksum));
1063 
1064 		icmp6h = NULL;
1065 		icmph = NULL;
1066 		break;}
1067 
1068 	/*
1069 	 * https://tools.ietf.org/html/rfc7915#section-5.1.1
1070 	 * If the Next Header field of the Fragment Header is an
1071 	 * extension header (except ESP, but including the Authentication
1072 	 * Header (AH)), then the packet SHOULD be dropped and logged.
1073 	 */
1074 	case IPPROTO_HOPOPTS:
1075 	case IPPROTO_ROUTING:
1076 	case IPPROTO_DSTOPTS:
1077 	case IPPROTO_AH:
1078 		return NT_DROP;
1079 
1080 	case IPPROTO_FRAGMENT:
1081 		/*
1082 		 * The fragment header is appended after or removed before
1083 		 * calling into this routine.
1084 		 */
1085 		VERIFY(FALSE);
1086 	case IPPROTO_ESP:
1087 		break;
1088 
1089 	default:
1090 		return NT_DROP;
1091 	}
1092 
1093 done:
1094 	/* Update checksum flags and offsets based on direction */
1095 	if (direction == NT_OUT) {
1096 		if ((*pbuf->pb_csum_flags & (CSUM_DATA_VALID | CSUM_PARTIAL)) ==
1097 		    (CSUM_DATA_VALID | CSUM_PARTIAL)) {
1098 			(pbuf->pb_mbuf)->m_pkthdr.csum_tx_start += CLAT46_HDR_EXPANSION_OVERHD;
1099 			(pbuf->pb_mbuf)->m_pkthdr.csum_tx_stuff += CLAT46_HDR_EXPANSION_OVERHD;
1100 		}
1101 
1102 		if (*pbuf->pb_csum_flags & CSUM_TCP) {
1103 			*pbuf->pb_csum_flags |= CSUM_TCPIPV6;
1104 		}
1105 		if (*pbuf->pb_csum_flags & CSUM_UDP) {
1106 			*pbuf->pb_csum_flags |= CSUM_UDPIPV6;
1107 		}
1108 		if (*pbuf->pb_csum_flags & CSUM_FRAGMENT) {
1109 			*pbuf->pb_csum_flags |= CSUM_FRAGMENT_IPV6;
1110 		}
1111 
1112 		/* Clear IPv4 checksum flags */
1113 		*pbuf->pb_csum_flags &= ~(CSUM_IP | CSUM_IP_FRAGS | CSUM_DELAY_DATA | CSUM_FRAGMENT);
1114 		/*
1115 		 * If the packet requires TCP segmentation due to TSO offload,
1116 		 * then change the checksum flag to indicate that an IPv6
1117 		 * TCP segmentation is needed now.
1118 		 */
1119 		if (*pbuf->pb_csum_flags & CSUM_TSO_IPV4) {
1120 			*pbuf->pb_csum_flags &= ~CSUM_TSO_IPV4;
1121 			*pbuf->pb_csum_flags |= CSUM_TSO_IPV6;
1122 		}
1123 	} else if (direction == NT_IN) {
1124 		/* XXX On input just reset csum flags */
1125 		*pbuf->pb_csum_flags = 0; /* Reset all flags for now */
1126 #if 0
1127 		/* Update csum flags and offsets for rx */
1128 		if (*pbuf->pb_csum_flags & CSUM_PARTIAL) {
1129 			(pbuf->pb_mbuf)->m_pkthdr.csum_rx_start -= CLAT46_HDR_EXPANSION_OVERHD;
1130 		}
1131 #endif
1132 	}
1133 	return NT_NAT64;
1134 }
1135 
1136 /* Fix the proto checksum for address change */
1137 static void
nat464_addr_cksum_fixup(uint16_t * pc,struct nat464_addr * ao,struct nat464_addr * an,protocol_family_t af,protocol_family_t naf,uint8_t u,boolean_t do_ones_complement)1138 nat464_addr_cksum_fixup(uint16_t *pc, struct nat464_addr *ao, struct nat464_addr *an,
1139     protocol_family_t af, protocol_family_t naf, uint8_t u, boolean_t do_ones_complement)
1140 {
1141 	/* Currently we only support v4 to v6 and vice versa */
1142 	VERIFY(af != naf);
1143 
1144 	switch (af) {
1145 	case PF_INET:
1146 		switch (naf) {
1147 		case PF_INET6:
1148 			if (do_ones_complement) {
1149 				*pc = ~nat464_cksum_fixup(nat464_cksum_fixup(
1150 					    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(
1151 						    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(~*pc,
1152 						    ao->nataddr16[0], an->nataddr16[0], u),
1153 						    ao->nataddr16[1], an->nataddr16[1], u),
1154 						    0, an->nataddr16[2], u),
1155 						    0, an->nataddr16[3], u),
1156 					    0, an->nataddr16[4], u),
1157 					    0, an->nataddr16[5], u),
1158 					    0, an->nataddr16[6], u),
1159 				    0, an->nataddr16[7], u);
1160 			} else {
1161 				*pc = nat464_cksum_fixup(nat464_cksum_fixup(
1162 					    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(
1163 						    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(*pc,
1164 						    ao->nataddr16[0], an->nataddr16[0], u),
1165 						    ao->nataddr16[1], an->nataddr16[1], u),
1166 						    0, an->nataddr16[2], u),
1167 						    0, an->nataddr16[3], u),
1168 					    0, an->nataddr16[4], u),
1169 					    0, an->nataddr16[5], u),
1170 					    0, an->nataddr16[6], u),
1171 				    0, an->nataddr16[7], u);
1172 			}
1173 			break;
1174 		}
1175 		break;
1176 	case PF_INET6:
1177 		/*
1178 		 * XXX For NAT464 this only applies to the incoming path.
1179 		 * The checksum therefore is already ones complemented.
1180 		 * Therefore we just perform normal fixup.
1181 		 */
1182 		switch (naf) {
1183 		case PF_INET:
1184 			*pc = nat464_cksum_fixup(nat464_cksum_fixup(
1185 				    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(
1186 					    nat464_cksum_fixup(nat464_cksum_fixup(nat464_cksum_fixup(*pc,
1187 					    ao->nataddr16[0], an->nataddr16[0], u),
1188 					    ao->nataddr16[1], an->nataddr16[1], u),
1189 					    ao->nataddr16[2], 0, u),
1190 					    ao->nataddr16[3], 0, u),
1191 				    ao->nataddr16[4], 0, u),
1192 				    ao->nataddr16[5], 0, u),
1193 				    ao->nataddr16[6], 0, u),
1194 			    ao->nataddr16[7], 0, u);
1195 			break;
1196 		}
1197 		break;
1198 	}
1199 }
1200 
1201 uint16_t
nat464_cksum_fixup(uint16_t cksum,uint16_t old,uint16_t new,uint8_t udp)1202 nat464_cksum_fixup(uint16_t cksum, uint16_t old, uint16_t new, uint8_t udp)
1203 {
1204 	uint32_t l;
1205 
1206 	if (udp && !cksum) {
1207 		return 0;
1208 	}
1209 	l = cksum + old - new;
1210 	l = (l >> 16) + (l & 0xffff);
1211 	l = l & 0xffff;
1212 	if (udp && !l) {
1213 		return 0xffff;
1214 	}
1215 	return (uint16_t)l;
1216 }
1217 
1218 /* CLAT46 event handlers */
1219 void
in6_clat46_eventhdlr_callback(struct eventhandler_entry_arg arg0 __unused,in6_clat46_evhdlr_code_t in6_clat46_ev_code,pid_t epid,uuid_t euuid)1220 in6_clat46_eventhdlr_callback(struct eventhandler_entry_arg arg0 __unused,
1221     in6_clat46_evhdlr_code_t in6_clat46_ev_code, pid_t epid, uuid_t euuid)
1222 {
1223 	struct kev_msg ev_msg;
1224 	struct kev_netevent_clat46_data clat46_event_data;
1225 
1226 	bzero(&ev_msg, sizeof(ev_msg));
1227 	bzero(&clat46_event_data, sizeof(clat46_event_data));
1228 
1229 	ev_msg.vendor_code      = KEV_VENDOR_APPLE;
1230 	ev_msg.kev_class        = KEV_NETWORK_CLASS;
1231 	ev_msg.kev_subclass     = KEV_NETEVENT_SUBCLASS;
1232 	ev_msg.event_code       = KEV_NETEVENT_CLAT46_EVENT;
1233 
1234 	bzero(&clat46_event_data, sizeof(clat46_event_data));
1235 	clat46_event_data.clat46_event_code = in6_clat46_ev_code;
1236 	clat46_event_data.epid = epid;
1237 	uuid_copy(clat46_event_data.euuid, euuid);
1238 
1239 	ev_msg.dv[0].data_ptr = &clat46_event_data;
1240 	ev_msg.dv[0].data_length = sizeof(clat46_event_data);
1241 
1242 	kev_post_msg(&ev_msg);
1243 }
1244 
1245 struct in6_clat46_event_nwk_wq_entry {
1246 	struct nwk_wq_entry nwk_wqe;
1247 	struct kev_netevent_clat46_data in6_clat46_ev_arg;
1248 };
1249 
1250 static void
in6_clat46_event_callback(struct nwk_wq_entry * nwk_item)1251 in6_clat46_event_callback(struct nwk_wq_entry *nwk_item)
1252 {
1253 	struct in6_clat46_event_nwk_wq_entry *p_ev;
1254 
1255 	p_ev = __container_of(nwk_item,
1256 	    struct in6_clat46_event_nwk_wq_entry, nwk_wqe);
1257 
1258 	EVENTHANDLER_INVOKE(&in6_clat46_evhdlr_ctxt, in6_clat46_event,
1259 	    p_ev->in6_clat46_ev_arg.clat46_event_code, p_ev->in6_clat46_ev_arg.epid,
1260 	    p_ev->in6_clat46_ev_arg.euuid);
1261 
1262 	kfree_type(struct in6_clat46_event_nwk_wq_entry, p_ev);
1263 }
1264 
1265 void
in6_clat46_event_enqueue_nwk_wq_entry(in6_clat46_evhdlr_code_t in6_clat46_event_code,pid_t epid,uuid_t euuid)1266 in6_clat46_event_enqueue_nwk_wq_entry(in6_clat46_evhdlr_code_t in6_clat46_event_code,
1267     pid_t epid, uuid_t euuid)
1268 {
1269 	struct in6_clat46_event_nwk_wq_entry *p_ev = NULL;
1270 
1271 	p_ev = kalloc_type(struct in6_clat46_event_nwk_wq_entry,
1272 	    Z_WAITOK | Z_ZERO | Z_NOFAIL);
1273 
1274 	p_ev->nwk_wqe.func = in6_clat46_event_callback;
1275 	p_ev->in6_clat46_ev_arg.clat46_event_code = in6_clat46_event_code;
1276 	p_ev->in6_clat46_ev_arg.epid = epid;
1277 	uuid_copy(p_ev->in6_clat46_ev_arg.euuid, euuid);
1278 
1279 	evhlog(debug, "%s: eventhandler enqueuing event of type=in6_clat46_event event_code=%s",
1280 	    __func__, in6_clat46_evhdlr_code2str(in6_clat46_event_code));
1281 
1282 	nwk_wq_enqueue(&p_ev->nwk_wqe);
1283 }
1284 
1285 extern const char*
in6_clat46_evhdlr_code2str(enum in6_clat46_evhdlr_code_t code)1286 in6_clat46_evhdlr_code2str(enum in6_clat46_evhdlr_code_t code)
1287 {
1288 	switch (code) {
1289 #define CLAT46_CODE_TO_STRING(type) case type: return #type;
1290 		CLAT46_CODE_TO_STRING(IN6_CLAT46_EVENT_V4_FLOW)
1291 		CLAT46_CODE_TO_STRING(IN6_CLAT46_EVENT_V6_ADDR_CONFFAIL)
1292 #undef CLAT46_CODE_TO_STRING
1293 	}
1294 	return "UNKNOWN_IN6_CLAT46_EVHDLR_CODE";
1295 }
1296