xref: /netbsd-src/sys/net/npf/npf_alg_icmp.c (revision a4ddc2c8fb9af816efe3b1c375a5530aef0e89e9)
1 /*	$NetBSD: npf_alg_icmp.c,v 1.16 2013/03/20 00:29:47 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 2010 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This material is based upon work partially supported by The
8  * NetBSD Foundation under a contract with Mindaugas Rasiukevicius.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * NPF ALG for ICMP and traceroute translations.
34  */
35 
36 #include <sys/cdefs.h>
37 __KERNEL_RCSID(0, "$NetBSD: npf_alg_icmp.c,v 1.16 2013/03/20 00:29:47 christos Exp $");
38 
39 #include <sys/param.h>
40 #include <sys/module.h>
41 #include <sys/pool.h>
42 
43 #include <netinet/in_systm.h>
44 #include <netinet/in.h>
45 #include <netinet/ip.h>
46 #include <netinet/tcp.h>
47 #include <netinet/udp.h>
48 #include <netinet/ip_icmp.h>
49 #include <netinet/icmp6.h>
50 #include <net/pfil.h>
51 
52 #include "npf_impl.h"
53 
54 MODULE(MODULE_CLASS_MISC, npf_alg_icmp, "npf");
55 
56 /*
57  * Traceroute criteria.
58  *
59  * IANA assigned base port: 33434.  However, common practice is to increase
60  * the port, thus monitor [33434-33484] range.  Additional filter is low TTL.
61  */
62 
63 #define	TR_BASE_PORT	33434
64 #define	TR_PORT_RANGE	33484
65 #define	TR_MAX_TTL	48
66 
67 static npf_alg_t *	alg_icmp	__read_mostly;
68 
69 static bool	npfa_icmp_match(npf_cache_t *, nbuf_t *, npf_nat_t *, int);
70 static bool	npfa_icmp_nat(npf_cache_t *, nbuf_t *, npf_nat_t *, int);
71 static npf_session_t *npfa_icmp_session(npf_cache_t *, nbuf_t *, int);
72 
73 /*
74  * npf_alg_icmp_{init,fini,modcmd}: ICMP ALG initialization, destruction
75  * and module interface.
76  */
77 
78 static int
79 npf_alg_icmp_init(void)
80 {
81 
82 	alg_icmp = npf_alg_register("icmp", npfa_icmp_match,
83 	    npfa_icmp_nat, npfa_icmp_session);
84 	KASSERT(alg_icmp != NULL);
85 	return 0;
86 }
87 
88 static int
89 npf_alg_icmp_fini(void)
90 {
91 
92 	KASSERT(alg_icmp != NULL);
93 	return npf_alg_unregister(alg_icmp);
94 }
95 
96 static int
97 npf_alg_icmp_modcmd(modcmd_t cmd, void *arg)
98 {
99 
100 	switch (cmd) {
101 	case MODULE_CMD_INIT:
102 		return npf_alg_icmp_init();
103 	case MODULE_CMD_FINI:
104 		return npf_alg_icmp_fini();
105 	case MODULE_CMD_AUTOUNLOAD:
106 		return EBUSY;
107 	default:
108 		return ENOTTY;
109 	}
110 	return 0;
111 }
112 
113 /*
114  * npfa_icmp_match: ALG matching inspector - determines ALG case and
115  * associates ALG with NAT entry.
116  */
117 static bool
118 npfa_icmp_match(npf_cache_t *npc, nbuf_t *nbuf, npf_nat_t *nt, int di)
119 {
120 	const int proto = npc->npc_proto;
121 	const struct ip *ip = npc->npc_ip.v4;
122 	in_port_t dport;
123 
124 	KASSERT(npf_iscached(npc, NPC_IP46));
125 	KASSERT(npf_iscached(npc, NPC_LAYER4));
126 
127 	/* Check for low TTL. */
128 	if (ip->ip_ttl > TR_MAX_TTL) {
129 		return false;
130 	}
131 
132 	switch (proto) {
133 	case IPPROTO_TCP: {
134 		const struct tcphdr *th = npc->npc_l4.tcp;
135 		dport = ntohs(th->th_dport);
136 		break;
137 	}
138 	case IPPROTO_UDP: {
139 		const struct udphdr *uh = npc->npc_l4.udp;
140 		dport = ntohs(uh->uh_dport);
141 		break;
142 	}
143 	case IPPROTO_ICMP:
144 	case IPPROTO_ICMPV6:
145 		/* Just to pass the test below. */
146 		dport = TR_BASE_PORT;
147 		break;
148 	default:
149 		return false;
150 	}
151 
152 	/* Handle TCP/UDP traceroute - check for port range. */
153 	if (dport < TR_BASE_PORT || dport > TR_PORT_RANGE) {
154 		return false;
155 	}
156 
157 	/* Associate ALG with translation entry. */
158 	npf_nat_setalg(nt, alg_icmp, 0);
159 	return true;
160 }
161 
162 /*
163  * npfa_icmp{4,6}_inspect: retrieve unique identifiers - either ICMP query
164  * ID or TCP/UDP ports of the original packet, which is embedded.
165  */
166 
167 static bool
168 npfa_icmp4_inspect(const int type, npf_cache_t *npc, nbuf_t *nbuf)
169 {
170 	u_int offby;
171 
172 	/* Per RFC 792. */
173 	switch (type) {
174 	case ICMP_UNREACH:
175 	case ICMP_SOURCEQUENCH:
176 	case ICMP_REDIRECT:
177 	case ICMP_TIMXCEED:
178 	case ICMP_PARAMPROB:
179 		if (npc == NULL) {
180 			return false;
181 		}
182 		/* Should contain original IP header. */
183 		if (!nbuf_advance(nbuf, offsetof(struct icmp, icmp_ip), 0)) {
184 			return false;
185 		}
186 		return (npf_cache_all(npc, nbuf) & NPC_LAYER4) != 0;
187 
188 	case ICMP_ECHOREPLY:
189 	case ICMP_ECHO:
190 	case ICMP_TSTAMP:
191 	case ICMP_TSTAMPREPLY:
192 	case ICMP_IREQ:
193 	case ICMP_IREQREPLY:
194 		/* Should contain ICMP query ID - ensure. */
195 		offby = offsetof(struct icmp, icmp_id);
196 		if (!nbuf_advance(nbuf, offby, sizeof(uint16_t))) {
197 			return false;
198 		}
199 		npc->npc_info |= NPC_ICMP_ID;
200 		return true;
201 	default:
202 		break;
203 	}
204 	return false;
205 }
206 
207 static bool
208 npfa_icmp6_inspect(const int type, npf_cache_t *npc, nbuf_t *nbuf)
209 {
210 	u_int offby;
211 
212 	/* Per RFC 4443. */
213 	switch (type) {
214 	case ICMP6_DST_UNREACH:
215 	case ICMP6_PACKET_TOO_BIG:
216 	case ICMP6_TIME_EXCEEDED:
217 	case ICMP6_PARAM_PROB:
218 		if (npc == NULL) {
219 			return false;
220 		}
221 		/* Should contain original IP header. */
222 		if (!nbuf_advance(nbuf, sizeof(struct icmp6_hdr), 0)) {
223 			return false;
224 		}
225 		return (npf_cache_all(npc, nbuf) & NPC_LAYER4) != 0;
226 
227 	case ICMP6_ECHO_REQUEST:
228 	case ICMP6_ECHO_REPLY:
229 		/* Should contain ICMP query ID - ensure. */
230 		offby = offsetof(struct icmp6_hdr, icmp6_id);
231 		if (!nbuf_advance(nbuf, offby, sizeof(uint16_t))) {
232 			return false;
233 		}
234 		npc->npc_info |= NPC_ICMP_ID;
235 		return true;
236 	default:
237 		break;
238 	}
239 	return false;
240 }
241 
242 /*
243  * npfa_icmp_session: ALG ICMP inspector.
244  *
245  * => Returns true if "enpc" is filled.
246  */
247 static bool
248 npfa_icmp_inspect(npf_cache_t *npc, nbuf_t *nbuf, npf_cache_t *enpc)
249 {
250 	bool ret;
251 
252 	KASSERT(npf_iscached(npc, NPC_IP46));
253 	KASSERT(npf_iscached(npc, NPC_ICMP));
254 
255 	/* Advance to ICMP header. */
256 	nbuf_reset(nbuf);
257 	if (!nbuf_advance(nbuf, npc->npc_hlen, 0)) {
258 		return false;
259 	}
260 	enpc->npc_info = 0;
261 
262 	/*
263 	 * Inspect the ICMP packet.  The relevant data might be in the
264 	 * embedded packet.  Fill the "enpc" cache, if so.
265 	 */
266 	if (npf_iscached(npc, NPC_IP4)) {
267 		const struct icmp *ic = npc->npc_l4.icmp;
268 		ret = npfa_icmp4_inspect(ic->icmp_type, enpc, nbuf);
269 	} else if (npf_iscached(npc, NPC_IP6)) {
270 		const struct icmp6_hdr *ic6 = npc->npc_l4.icmp6;
271 		ret = npfa_icmp6_inspect(ic6->icmp6_type, enpc, nbuf);
272 	} else {
273 		ret = false;
274 	}
275 	if (!ret) {
276 		return false;
277 	}
278 
279 	/* ICMP ID is the original packet, just indicate it. */
280 	if (npf_iscached(enpc, NPC_ICMP_ID)) {
281 		npc->npc_info |= NPC_ICMP_ID;
282 		return false;
283 	}
284 
285 	/* Indicate that embedded packet is in the cache. */
286 	return true;
287 }
288 
289 static npf_session_t *
290 npfa_icmp_session(npf_cache_t *npc, nbuf_t *nbuf, int di)
291 {
292 	npf_cache_t enpc;
293 
294 	/* Inspect ICMP packet for an embedded packet. */
295 	if (!npf_iscached(npc, NPC_ICMP))
296 		return NULL;
297 	if (!npfa_icmp_inspect(npc, nbuf, &enpc))
298 		return NULL;
299 
300 	/*
301 	 * Invert the identifiers of the embedded packet.
302 	 * If it is ICMP, then ensure ICMP ID.
303 	 */
304 	union l4 {
305 		struct tcphdr th;
306 		struct udphdr uh;
307 	} l4;
308 	bool ret, forw;
309 
310 	#define	SWAP(type, x, y) { type tmp = x; x = y; y = tmp; }
311 	SWAP(npf_addr_t *, enpc.npc_srcip, enpc.npc_dstip);
312 
313 	switch (enpc.npc_proto) {
314 	case IPPROTO_TCP:
315 		l4.th.th_sport = enpc.npc_l4.tcp->th_dport;
316 		l4.th.th_dport = enpc.npc_l4.tcp->th_sport;
317 		enpc.npc_l4.tcp = &l4.th;
318 		break;
319 	case IPPROTO_UDP:
320 		l4.uh.uh_sport = enpc.npc_l4.udp->uh_dport;
321 		l4.uh.uh_dport = enpc.npc_l4.udp->uh_sport;
322 		enpc.npc_l4.udp = &l4.uh;
323 		break;
324 	case IPPROTO_ICMP: {
325 		const struct icmp *ic = enpc.npc_l4.icmp;
326 		ret = npfa_icmp4_inspect(ic->icmp_type, &enpc, nbuf);
327 		if (!ret || !npf_iscached(&enpc, NPC_ICMP_ID))
328 			return false;
329 		break;
330 	}
331 	case IPPROTO_ICMPV6: {
332 		const struct icmp6_hdr *ic6 = enpc.npc_l4.icmp6;
333 		ret = npfa_icmp6_inspect(ic6->icmp6_type, &enpc, nbuf);
334 		if (!ret || !npf_iscached(&enpc, NPC_ICMP_ID))
335 			return false;
336 		break;
337 	}
338 	default:
339 		return false;
340 	}
341 
342 	/* Lookup for a session using embedded packet. */
343 	return npf_session_lookup(&enpc, nbuf, di, &forw);
344 }
345 
346 /*
347  * npfa_icmp_nat: ALG inbound translation inspector, rewrite IP address
348  * in the IP header, which is embedded in ICMP packet.
349  */
350 static bool
351 npfa_icmp_nat(npf_cache_t *npc, nbuf_t *nbuf, npf_nat_t *nt, int di)
352 {
353 	npf_cache_t enpc;
354 
355 	if (di != PFIL_IN || !npf_iscached(npc, NPC_ICMP))
356 		return false;
357 	if (!npfa_icmp_inspect(npc, nbuf, &enpc))
358 		return false;
359 
360 	KASSERT(npf_iscached(&enpc, NPC_IP46));
361 	KASSERT(npf_iscached(&enpc, NPC_LAYER4));
362 
363 	struct icmp *ic = npc->npc_l4.icmp;
364 	uint16_t cksum = ic->icmp_cksum;
365 
366 	CTASSERT(offsetof(struct icmp, icmp_cksum) ==
367 	    offsetof(struct icmp6_hdr, icmp6_cksum));
368 
369 	/*
370 	 * Retrieve the original address and port, then calculate ICMP
371 	 * checksum for these changes in the embedded packet.  While data
372 	 * is not rewritten in the cache, save IP and TCP/UDP checksums.
373 	 */
374 	const int proto = enpc.npc_proto;
375 	uint16_t ipcksum = 0, l4cksum = 0;
376 	npf_addr_t *addr;
377 	in_port_t port;
378 
379 	npf_nat_getorig(nt, &addr, &port);
380 
381 	if (npf_iscached(&enpc, NPC_IP4)) {
382 		const struct ip *eip = enpc.npc_ip.v4;
383 		ipcksum = eip->ip_sum;
384 	}
385 	cksum = npf_addr_cksum(cksum, enpc.npc_alen, enpc.npc_srcip, addr);
386 
387 	switch (proto) {
388 	case IPPROTO_TCP: {
389 		const struct tcphdr *th = enpc.npc_l4.tcp;
390 		cksum = npf_fixup16_cksum(cksum, th->th_sport, port);
391 		l4cksum = th->th_sum;
392 		break;
393 	}
394 	case IPPROTO_UDP: {
395 		const struct udphdr *uh = enpc.npc_l4.udp;
396 		cksum = npf_fixup16_cksum(cksum, uh->uh_sport, port);
397 		l4cksum = uh->uh_sum;
398 		break;
399 	}
400 	case IPPROTO_ICMP:
401 	case IPPROTO_ICMPV6:
402 		break;
403 	default:
404 		return false;
405 	}
406 
407 	/*
408 	 * Rewrite the source IP address and port of the embedded IP header,
409 	 * which represents the original packet, therefore passing PFIL_OUT.
410 	 * This updates the checksums in the embedded packet.
411 	 */
412 	if (npf_nat_translate(&enpc, nbuf, nt, false, PFIL_OUT)) {
413 		return false;
414 	}
415 
416 	/*
417 	 * Finish calculation of the ICMP checksum: include the checksum
418 	 * change in the embedded packet.
419 	 */
420 	if (npf_iscached(&enpc, NPC_IP4)) {
421 		const struct ip *eip = enpc.npc_ip.v4;
422 		cksum = npf_fixup16_cksum(cksum, ipcksum, eip->ip_sum);
423 	}
424 	switch (proto) {
425 	case IPPROTO_TCP: {
426 		const struct tcphdr *th = enpc.npc_l4.tcp;
427 		cksum = npf_fixup16_cksum(cksum, l4cksum, th->th_sum);
428 		break;
429 	}
430 	case IPPROTO_UDP:
431 		if (l4cksum) {
432 			const struct udphdr *uh = enpc.npc_l4.udp;
433 			cksum = npf_fixup16_cksum(cksum, l4cksum, uh->uh_sum);
434 		}
435 		break;
436 	}
437 	ic->icmp_cksum = cksum;
438 	return true;
439 }
440