xref: /freebsd/sys/netpfil/pf/pf_lb.c (revision d10de21f)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002 - 2008 Henning Brauer
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  *    - Redistributions of source code must retain the above copyright
13  *      notice, this list of conditions and the following disclaimer.
14  *    - Redistributions in binary form must reproduce the above
15  *      copyright notice, this list of conditions and the following
16  *      disclaimer in the documentation and/or other materials provided
17  *      with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
27  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
29  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  *
32  * Effort sponsored in part by the Defense Advanced Research Projects
33  * Agency (DARPA) and Air Force Research Laboratory, Air Force
34  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
35  *
36  *	$OpenBSD: pf_lb.c,v 1.2 2009/02/12 02:13:15 sthen Exp $
37  */
38 
39 #include <sys/cdefs.h>
40 #include "opt_pf.h"
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 
44 #include <sys/param.h>
45 #include <sys/lock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/vnet.h>
52 #include <net/pfvar.h>
53 #include <net/if_pflog.h>
54 
55 #define DPFPRINTF(n, x)	if (V_pf_status.debug >= (n)) printf x
56 
57 static void		 pf_hash(struct pf_addr *, struct pf_addr *,
58 			    struct pf_poolhashkey *, sa_family_t);
59 static struct pf_krule	*pf_match_translation(struct pf_pdesc *, struct mbuf *,
60 			    int, struct pfi_kkif *,
61 			    struct pf_addr *, u_int16_t, struct pf_addr *,
62 			    uint16_t, int, struct pf_kanchor_stackframe *);
63 static int pf_get_sport(sa_family_t, uint8_t, struct pf_krule *,
64     struct pf_addr *, uint16_t, struct pf_addr *, uint16_t, struct pf_addr *,
65     uint16_t *, uint16_t, uint16_t, struct pf_ksrc_node **);
66 
67 #define mix(a,b,c) \
68 	do {					\
69 		a -= b; a -= c; a ^= (c >> 13);	\
70 		b -= c; b -= a; b ^= (a << 8);	\
71 		c -= a; c -= b; c ^= (b >> 13);	\
72 		a -= b; a -= c; a ^= (c >> 12);	\
73 		b -= c; b -= a; b ^= (a << 16);	\
74 		c -= a; c -= b; c ^= (b >> 5);	\
75 		a -= b; a -= c; a ^= (c >> 3);	\
76 		b -= c; b -= a; b ^= (a << 10);	\
77 		c -= a; c -= b; c ^= (b >> 15);	\
78 	} while (0)
79 
80 /*
81  * hash function based on bridge_hash in if_bridge.c
82  */
83 static void
pf_hash(struct pf_addr * inaddr,struct pf_addr * hash,struct pf_poolhashkey * key,sa_family_t af)84 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
85     struct pf_poolhashkey *key, sa_family_t af)
86 {
87 	u_int32_t	a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
88 
89 	switch (af) {
90 #ifdef INET
91 	case AF_INET:
92 		a += inaddr->addr32[0];
93 		b += key->key32[1];
94 		mix(a, b, c);
95 		hash->addr32[0] = c + key->key32[2];
96 		break;
97 #endif /* INET */
98 #ifdef INET6
99 	case AF_INET6:
100 		a += inaddr->addr32[0];
101 		b += inaddr->addr32[2];
102 		mix(a, b, c);
103 		hash->addr32[0] = c;
104 		a += inaddr->addr32[1];
105 		b += inaddr->addr32[3];
106 		c += key->key32[1];
107 		mix(a, b, c);
108 		hash->addr32[1] = c;
109 		a += inaddr->addr32[2];
110 		b += inaddr->addr32[1];
111 		c += key->key32[2];
112 		mix(a, b, c);
113 		hash->addr32[2] = c;
114 		a += inaddr->addr32[3];
115 		b += inaddr->addr32[0];
116 		c += key->key32[3];
117 		mix(a, b, c);
118 		hash->addr32[3] = c;
119 		break;
120 #endif /* INET6 */
121 	}
122 }
123 
124 static struct pf_krule *
pf_match_translation(struct pf_pdesc * pd,struct mbuf * m,int off,struct pfi_kkif * kif,struct pf_addr * saddr,u_int16_t sport,struct pf_addr * daddr,uint16_t dport,int rs_num,struct pf_kanchor_stackframe * anchor_stack)125 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
126     struct pfi_kkif *kif, struct pf_addr *saddr, u_int16_t sport,
127     struct pf_addr *daddr, uint16_t dport, int rs_num,
128     struct pf_kanchor_stackframe *anchor_stack)
129 {
130 	struct pf_krule		*r, *rm = NULL;
131 	struct pf_kruleset	*ruleset = NULL;
132 	int			 tag = -1;
133 	int			 rtableid = -1;
134 	int			 asd = 0;
135 
136 	r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
137 	while (r != NULL) {
138 		struct pf_rule_addr	*src = NULL, *dst = NULL;
139 		struct pf_addr_wrap	*xdst = NULL;
140 
141 		if (r->action == PF_BINAT && pd->dir == PF_IN) {
142 			src = &r->dst;
143 			if (r->rpool.cur != NULL)
144 				xdst = &r->rpool.cur->addr;
145 		} else {
146 			src = &r->src;
147 			dst = &r->dst;
148 		}
149 
150 		pf_counter_u64_add(&r->evaluations, 1);
151 		if (pfi_kkif_match(r->kif, kif) == r->ifnot)
152 			r = r->skip[PF_SKIP_IFP].ptr;
153 		else if (r->direction && r->direction != pd->dir)
154 			r = r->skip[PF_SKIP_DIR].ptr;
155 		else if (r->af && r->af != pd->af)
156 			r = r->skip[PF_SKIP_AF].ptr;
157 		else if (r->proto && r->proto != pd->proto)
158 			r = r->skip[PF_SKIP_PROTO].ptr;
159 		else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
160 		    src->neg, kif, M_GETFIB(m)))
161 			r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
162 			    PF_SKIP_DST_ADDR].ptr;
163 		else if (src->port_op && !pf_match_port(src->port_op,
164 		    src->port[0], src->port[1], sport))
165 			r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
166 			    PF_SKIP_DST_PORT].ptr;
167 		else if (dst != NULL &&
168 		    PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL,
169 		    M_GETFIB(m)))
170 			r = r->skip[PF_SKIP_DST_ADDR].ptr;
171 		else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
172 		    0, NULL, M_GETFIB(m)))
173 			r = TAILQ_NEXT(r, entries);
174 		else if (dst != NULL && dst->port_op &&
175 		    !pf_match_port(dst->port_op, dst->port[0],
176 		    dst->port[1], dport))
177 			r = r->skip[PF_SKIP_DST_PORT].ptr;
178 		else if (r->match_tag && !pf_match_tag(m, r, &tag,
179 		    pd->pf_mtag ? pd->pf_mtag->tag : 0))
180 			r = TAILQ_NEXT(r, entries);
181 		else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
182 		    IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
183 		    off, &pd->hdr.tcp), r->os_fingerprint)))
184 			r = TAILQ_NEXT(r, entries);
185 		else {
186 			if (r->tag)
187 				tag = r->tag;
188 			if (r->rtableid >= 0)
189 				rtableid = r->rtableid;
190 			if (r->anchor == NULL) {
191 				rm = r;
192 				if (rm->action == PF_NONAT ||
193 				    rm->action == PF_NORDR ||
194 				    rm->action == PF_NOBINAT) {
195 					rm = NULL;
196 				}
197 				break;
198 			} else
199 				pf_step_into_anchor(anchor_stack, &asd,
200 				    &ruleset, rs_num, &r, NULL, NULL);
201 		}
202 		if (r == NULL)
203 			pf_step_out_of_anchor(anchor_stack, &asd, &ruleset,
204 			    rs_num, &r, NULL, NULL);
205 	}
206 
207 	if (tag > 0 && pf_tag_packet(m, pd, tag))
208 		return (NULL);
209 	if (rtableid >= 0)
210 		M_SETFIB(m, rtableid);
211 
212 	return (rm);
213 }
214 
215 static int
pf_get_sport(sa_family_t af,u_int8_t proto,struct pf_krule * r,struct pf_addr * saddr,uint16_t sport,struct pf_addr * daddr,uint16_t dport,struct pf_addr * naddr,uint16_t * nport,uint16_t low,uint16_t high,struct pf_ksrc_node ** sn)216 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_krule *r,
217     struct pf_addr *saddr, uint16_t sport, struct pf_addr *daddr,
218     uint16_t dport, struct pf_addr *naddr, uint16_t *nport, uint16_t low,
219     uint16_t high, struct pf_ksrc_node **sn)
220 {
221 	struct pf_state_key_cmp	key;
222 	struct pf_addr		init_addr;
223 
224 	bzero(&init_addr, sizeof(init_addr));
225 	if (pf_map_addr(af, r, saddr, naddr, NULL, &init_addr, sn))
226 		return (1);
227 
228 	bzero(&key, sizeof(key));
229 	key.af = af;
230 	key.proto = proto;
231 	key.port[0] = dport;
232 	PF_ACPY(&key.addr[0], daddr, key.af);
233 
234 	do {
235 		PF_ACPY(&key.addr[1], naddr, key.af);
236 
237 		/*
238 		 * port search; start random, step;
239 		 * similar 2 portloop in in_pcbbind
240 		 */
241 		if (proto == IPPROTO_SCTP) {
242 			key.port[1] = sport;
243 			if (!pf_find_state_all_exists(&key, PF_IN)) {
244 				*nport = sport;
245 				return (0);
246 			} else {
247 				return (1); /* Fail mapping. */
248 			}
249 		} else if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
250 		    proto == IPPROTO_ICMP) || (low == 0 && high == 0)) {
251 			/*
252 			 * XXX bug: icmp states don't use the id on both sides.
253 			 * (traceroute -I through nat)
254 			 */
255 			key.port[1] = sport;
256 			if (!pf_find_state_all_exists(&key, PF_IN)) {
257 				*nport = sport;
258 				return (0);
259 			}
260 		} else if (low == high) {
261 			key.port[1] = htons(low);
262 			if (!pf_find_state_all_exists(&key, PF_IN)) {
263 				*nport = htons(low);
264 				return (0);
265 			}
266 		} else {
267 			uint32_t tmp;
268 			uint16_t cut;
269 
270 			if (low > high) {
271 				tmp = low;
272 				low = high;
273 				high = tmp;
274 			}
275 			/* low < high */
276 			cut = arc4random() % (1 + high - low) + low;
277 			/* low <= cut <= high */
278 			for (tmp = cut; tmp <= high && tmp <= 0xffff; ++tmp) {
279 				key.port[1] = htons(tmp);
280 				if (!pf_find_state_all_exists(&key, PF_IN)) {
281 					*nport = htons(tmp);
282 					return (0);
283 				}
284 			}
285 			tmp = cut;
286 			for (tmp -= 1; tmp >= low && tmp <= 0xffff; --tmp) {
287 				key.port[1] = htons(tmp);
288 				if (!pf_find_state_all_exists(&key, PF_IN)) {
289 					*nport = htons(tmp);
290 					return (0);
291 				}
292 			}
293 		}
294 
295 		switch (r->rpool.opts & PF_POOL_TYPEMASK) {
296 		case PF_POOL_RANDOM:
297 		case PF_POOL_ROUNDROBIN:
298 			/*
299 			 * pick a different source address since we're out
300 			 * of free port choices for the current one.
301 			 */
302 			if (pf_map_addr(af, r, saddr, naddr, NULL, &init_addr, sn))
303 				return (1);
304 			break;
305 		case PF_POOL_NONE:
306 		case PF_POOL_SRCHASH:
307 		case PF_POOL_BITMASK:
308 		default:
309 			return (1);
310 		}
311 	} while (! PF_AEQ(&init_addr, naddr, af) );
312 	return (1);					/* none available */
313 }
314 
315 static int
pf_get_mape_sport(sa_family_t af,u_int8_t proto,struct pf_krule * r,struct pf_addr * saddr,uint16_t sport,struct pf_addr * daddr,uint16_t dport,struct pf_addr * naddr,uint16_t * nport,struct pf_ksrc_node ** sn)316 pf_get_mape_sport(sa_family_t af, u_int8_t proto, struct pf_krule *r,
317     struct pf_addr *saddr, uint16_t sport, struct pf_addr *daddr,
318     uint16_t dport, struct pf_addr *naddr, uint16_t *nport,
319     struct pf_ksrc_node **sn)
320 {
321 	uint16_t psmask, low, highmask;
322 	uint16_t i, ahigh, cut;
323 	int ashift, psidshift;
324 
325 	ashift = 16 - r->rpool.mape.offset;
326 	psidshift = ashift - r->rpool.mape.psidlen;
327 	psmask = r->rpool.mape.psid & ((1U << r->rpool.mape.psidlen) - 1);
328 	psmask = psmask << psidshift;
329 	highmask = (1U << psidshift) - 1;
330 
331 	ahigh = (1U << r->rpool.mape.offset) - 1;
332 	cut = arc4random() & ahigh;
333 	if (cut == 0)
334 		cut = 1;
335 
336 	for (i = cut; i <= ahigh; i++) {
337 		low = (i << ashift) | psmask;
338 		if (!pf_get_sport(af, proto, r, saddr, sport, daddr, dport,
339 		    naddr, nport, low, low | highmask, sn))
340 			return (0);
341 	}
342 	for (i = cut - 1; i > 0; i--) {
343 		low = (i << ashift) | psmask;
344 		if (!pf_get_sport(af, proto, r, saddr, sport, daddr, dport,
345 		    naddr, nport, low, low | highmask, sn))
346 			return (0);
347 	}
348 	return (1);
349 }
350 
351 u_short
pf_map_addr(sa_family_t af,struct pf_krule * r,struct pf_addr * saddr,struct pf_addr * naddr,struct pfi_kkif ** nkif,struct pf_addr * init_addr,struct pf_ksrc_node ** sn)352 pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr,
353     struct pf_addr *naddr, struct pfi_kkif **nkif, struct pf_addr *init_addr,
354     struct pf_ksrc_node **sn)
355 {
356 	u_short			 reason = 0;
357 	struct pf_kpool		*rpool = &r->rpool;
358 	struct pf_addr		*raddr = NULL, *rmask = NULL;
359 	struct pf_srchash	*sh = NULL;
360 
361 	/* Try to find a src_node if none was given and this
362 	   is a sticky-address rule. */
363 	if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
364 	    (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE)
365 		*sn = pf_find_src_node(saddr, r, af, &sh, false);
366 
367 	/* If a src_node was found or explicitly given and it has a non-zero
368 	   route address, use this address. A zeroed address is found if the
369 	   src node was created just a moment ago in pf_create_state and it
370 	   needs to be filled in with routing decision calculated here. */
371 	if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
372 		/* If the supplied address is the same as the current one we've
373 		 * been asked before, so tell the caller that there's no other
374 		 * address to be had. */
375 		if (PF_AEQ(naddr, &(*sn)->raddr, af)) {
376 			reason = PFRES_MAPFAILED;
377 			goto done;
378 		}
379 
380 		PF_ACPY(naddr, &(*sn)->raddr, af);
381 		if (nkif)
382 			*nkif = (*sn)->rkif;
383 		if (V_pf_status.debug >= PF_DEBUG_NOISY) {
384 			printf("pf_map_addr: src tracking maps ");
385 			pf_print_host(saddr, 0, af);
386 			printf(" to ");
387 			pf_print_host(naddr, 0, af);
388 			if (nkif)
389 				printf("@%s", (*nkif)->pfik_name);
390 			printf("\n");
391 		}
392 		goto done;
393 	}
394 
395 	mtx_lock(&rpool->mtx);
396 	/* Find the route using chosen algorithm. Store the found route
397 	   in src_node if it was given or found. */
398 	if (rpool->cur->addr.type == PF_ADDR_NOROUTE) {
399 		reason = PFRES_MAPFAILED;
400 		goto done_pool_mtx;
401 	}
402 	if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
403 		switch (af) {
404 #ifdef INET
405 		case AF_INET:
406 			if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
407 			    (rpool->opts & PF_POOL_TYPEMASK) !=
408 			    PF_POOL_ROUNDROBIN) {
409 				reason = PFRES_MAPFAILED;
410 				goto done_pool_mtx;
411 			}
412 			raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
413 			rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
414 			break;
415 #endif /* INET */
416 #ifdef INET6
417 		case AF_INET6:
418 			if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
419 			    (rpool->opts & PF_POOL_TYPEMASK) !=
420 			    PF_POOL_ROUNDROBIN) {
421 				reason = PFRES_MAPFAILED;
422 				goto done_pool_mtx;
423 			}
424 			raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
425 			rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
426 			break;
427 #endif /* INET6 */
428 		}
429 	} else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
430 		if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) {
431 			reason = PFRES_MAPFAILED;
432 			goto done_pool_mtx; /* unsupported */
433 		}
434 	} else {
435 		raddr = &rpool->cur->addr.v.a.addr;
436 		rmask = &rpool->cur->addr.v.a.mask;
437 	}
438 
439 	switch (rpool->opts & PF_POOL_TYPEMASK) {
440 	case PF_POOL_NONE:
441 		PF_ACPY(naddr, raddr, af);
442 		break;
443 	case PF_POOL_BITMASK:
444 		PF_POOLMASK(naddr, raddr, rmask, saddr, af);
445 		break;
446 	case PF_POOL_RANDOM:
447 		if (init_addr != NULL && PF_AZERO(init_addr, af)) {
448 			switch (af) {
449 #ifdef INET
450 			case AF_INET:
451 				rpool->counter.addr32[0] = htonl(arc4random());
452 				break;
453 #endif /* INET */
454 #ifdef INET6
455 			case AF_INET6:
456 				if (rmask->addr32[3] != 0xffffffff)
457 					rpool->counter.addr32[3] =
458 					    htonl(arc4random());
459 				else
460 					break;
461 				if (rmask->addr32[2] != 0xffffffff)
462 					rpool->counter.addr32[2] =
463 					    htonl(arc4random());
464 				else
465 					break;
466 				if (rmask->addr32[1] != 0xffffffff)
467 					rpool->counter.addr32[1] =
468 					    htonl(arc4random());
469 				else
470 					break;
471 				if (rmask->addr32[0] != 0xffffffff)
472 					rpool->counter.addr32[0] =
473 					    htonl(arc4random());
474 				break;
475 #endif /* INET6 */
476 			}
477 			PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
478 			PF_ACPY(init_addr, naddr, af);
479 
480 		} else {
481 			PF_AINC(&rpool->counter, af);
482 			PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
483 		}
484 		break;
485 	case PF_POOL_SRCHASH:
486 	    {
487 		unsigned char hash[16];
488 
489 		pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
490 		PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
491 		break;
492 	    }
493 	case PF_POOL_ROUNDROBIN:
494 	    {
495 		struct pf_kpooladdr *acur = rpool->cur;
496 
497 		if (rpool->cur->addr.type == PF_ADDR_TABLE) {
498 			if (!pfr_pool_get(rpool->cur->addr.p.tbl,
499 			    &rpool->tblidx, &rpool->counter, af))
500 				goto get_addr;
501 		} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
502 			if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
503 			    &rpool->tblidx, &rpool->counter, af))
504 				goto get_addr;
505 		} else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
506 			goto get_addr;
507 
508 	try_next:
509 		if (TAILQ_NEXT(rpool->cur, entries) == NULL)
510 			rpool->cur = TAILQ_FIRST(&rpool->list);
511 		else
512 			rpool->cur = TAILQ_NEXT(rpool->cur, entries);
513 		if (rpool->cur->addr.type == PF_ADDR_TABLE) {
514 			rpool->tblidx = -1;
515 			if (pfr_pool_get(rpool->cur->addr.p.tbl,
516 			    &rpool->tblidx, &rpool->counter, af)) {
517 				/* table contains no address of type 'af' */
518 				if (rpool->cur != acur)
519 					goto try_next;
520 				reason = PFRES_MAPFAILED;
521 				goto done_pool_mtx;
522 			}
523 		} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
524 			rpool->tblidx = -1;
525 			if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
526 			    &rpool->tblidx, &rpool->counter, af)) {
527 				/* table contains no address of type 'af' */
528 				if (rpool->cur != acur)
529 					goto try_next;
530 				reason = PFRES_MAPFAILED;
531 				goto done_pool_mtx;
532 			}
533 		} else {
534 			raddr = &rpool->cur->addr.v.a.addr;
535 			rmask = &rpool->cur->addr.v.a.mask;
536 			PF_ACPY(&rpool->counter, raddr, af);
537 		}
538 
539 	get_addr:
540 		PF_ACPY(naddr, &rpool->counter, af);
541 		if (init_addr != NULL && PF_AZERO(init_addr, af))
542 			PF_ACPY(init_addr, naddr, af);
543 		PF_AINC(&rpool->counter, af);
544 		break;
545 	    }
546 	}
547 
548 	if (nkif)
549 		*nkif = rpool->cur->kif;
550 
551 	if (*sn != NULL) {
552 		PF_ACPY(&(*sn)->raddr, naddr, af);
553 		if (nkif)
554 			(*sn)->rkif = *nkif;
555 	}
556 
557 	if (V_pf_status.debug >= PF_DEBUG_NOISY &&
558 	    (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
559 		printf("pf_map_addr: selected address ");
560 		pf_print_host(naddr, 0, af);
561 		if (nkif)
562 			printf("@%s", (*nkif)->pfik_name);
563 		printf("\n");
564 	}
565 
566 done_pool_mtx:
567 	mtx_unlock(&rpool->mtx);
568 
569 done:
570 	if (reason) {
571 		counter_u64_add(V_pf_status.counters[reason], 1);
572 	}
573 
574 	return (reason);
575 }
576 
577 struct pf_krule *
pf_get_translation(struct pf_pdesc * pd,struct mbuf * m,int off,struct pfi_kkif * kif,struct pf_ksrc_node ** sn,struct pf_state_key ** skp,struct pf_state_key ** nkp,struct pf_addr * saddr,struct pf_addr * daddr,uint16_t sport,uint16_t dport,struct pf_kanchor_stackframe * anchor_stack)578 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
579     struct pfi_kkif *kif, struct pf_ksrc_node **sn,
580     struct pf_state_key **skp, struct pf_state_key **nkp,
581     struct pf_addr *saddr, struct pf_addr *daddr,
582     uint16_t sport, uint16_t dport, struct pf_kanchor_stackframe *anchor_stack)
583 {
584 	struct pf_krule	*r = NULL;
585 	struct pf_addr	*naddr;
586 	uint16_t	*nport;
587 	uint16_t	 low, high;
588 
589 	PF_RULES_RASSERT();
590 	KASSERT(*skp == NULL, ("*skp not NULL"));
591 	KASSERT(*nkp == NULL, ("*nkp not NULL"));
592 
593 	if (pd->dir == PF_OUT) {
594 		r = pf_match_translation(pd, m, off, kif, saddr,
595 		    sport, daddr, dport, PF_RULESET_BINAT, anchor_stack);
596 		if (r == NULL)
597 			r = pf_match_translation(pd, m, off, kif,
598 			    saddr, sport, daddr, dport, PF_RULESET_NAT,
599 			    anchor_stack);
600 	} else {
601 		r = pf_match_translation(pd, m, off, kif, saddr,
602 		    sport, daddr, dport, PF_RULESET_RDR, anchor_stack);
603 		if (r == NULL)
604 			r = pf_match_translation(pd, m, off, kif,
605 			    saddr, sport, daddr, dport, PF_RULESET_BINAT,
606 			    anchor_stack);
607 	}
608 
609 	if (r == NULL)
610 		return (NULL);
611 
612 	switch (r->action) {
613 	case PF_NONAT:
614 	case PF_NOBINAT:
615 	case PF_NORDR:
616 		return (NULL);
617 	}
618 
619 	*skp = pf_state_key_setup(pd, saddr, daddr, sport, dport);
620 	if (*skp == NULL)
621 		return (NULL);
622 	*nkp = pf_state_key_clone(*skp);
623 	if (*nkp == NULL) {
624 		uma_zfree(V_pf_state_key_z, *skp);
625 		*skp = NULL;
626 		return (NULL);
627 	}
628 
629 	/* XXX We only modify one side for now. */
630 	naddr = &(*nkp)->addr[1];
631 	nport = &(*nkp)->port[1];
632 
633 	switch (r->action) {
634 	case PF_NAT:
635 		if (pd->proto == IPPROTO_ICMP) {
636 			low  = 1;
637 			high = 65535;
638 		} else {
639 			low  = r->rpool.proxy_port[0];
640 			high = r->rpool.proxy_port[1];
641 		}
642 		if (r->rpool.mape.offset > 0) {
643 			if (pf_get_mape_sport(pd->af, pd->proto, r, saddr,
644 			    sport, daddr, dport, naddr, nport, sn)) {
645 				DPFPRINTF(PF_DEBUG_MISC,
646 				    ("pf: MAP-E port allocation (%u/%u/%u)"
647 				    " failed\n",
648 				    r->rpool.mape.offset,
649 				    r->rpool.mape.psidlen,
650 				    r->rpool.mape.psid));
651 				goto notrans;
652 			}
653 		} else if (pf_get_sport(pd->af, pd->proto, r, saddr, sport,
654 		    daddr, dport, naddr, nport, low, high, sn)) {
655 			DPFPRINTF(PF_DEBUG_MISC,
656 			    ("pf: NAT proxy port allocation (%u-%u) failed\n",
657 			    r->rpool.proxy_port[0], r->rpool.proxy_port[1]));
658 			goto notrans;
659 		}
660 		break;
661 	case PF_BINAT:
662 		switch (pd->dir) {
663 		case PF_OUT:
664 			if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
665 				switch (pd->af) {
666 #ifdef INET
667 				case AF_INET:
668 					if (r->rpool.cur->addr.p.dyn->
669 					    pfid_acnt4 < 1)
670 						goto notrans;
671 					PF_POOLMASK(naddr,
672 					    &r->rpool.cur->addr.p.dyn->
673 					    pfid_addr4,
674 					    &r->rpool.cur->addr.p.dyn->
675 					    pfid_mask4, saddr, AF_INET);
676 					break;
677 #endif /* INET */
678 #ifdef INET6
679 				case AF_INET6:
680 					if (r->rpool.cur->addr.p.dyn->
681 					    pfid_acnt6 < 1)
682 						goto notrans;
683 					PF_POOLMASK(naddr,
684 					    &r->rpool.cur->addr.p.dyn->
685 					    pfid_addr6,
686 					    &r->rpool.cur->addr.p.dyn->
687 					    pfid_mask6, saddr, AF_INET6);
688 					break;
689 #endif /* INET6 */
690 				}
691 			} else
692 				PF_POOLMASK(naddr,
693 				    &r->rpool.cur->addr.v.a.addr,
694 				    &r->rpool.cur->addr.v.a.mask, saddr,
695 				    pd->af);
696 			break;
697 		case PF_IN:
698 			if (r->src.addr.type == PF_ADDR_DYNIFTL) {
699 				switch (pd->af) {
700 #ifdef INET
701 				case AF_INET:
702 					if (r->src.addr.p.dyn-> pfid_acnt4 < 1)
703 						goto notrans;
704 					PF_POOLMASK(naddr,
705 					    &r->src.addr.p.dyn->pfid_addr4,
706 					    &r->src.addr.p.dyn->pfid_mask4,
707 					    daddr, AF_INET);
708 					break;
709 #endif /* INET */
710 #ifdef INET6
711 				case AF_INET6:
712 					if (r->src.addr.p.dyn->pfid_acnt6 < 1)
713 						goto notrans;
714 					PF_POOLMASK(naddr,
715 					    &r->src.addr.p.dyn->pfid_addr6,
716 					    &r->src.addr.p.dyn->pfid_mask6,
717 					    daddr, AF_INET6);
718 					break;
719 #endif /* INET6 */
720 				}
721 			} else
722 				PF_POOLMASK(naddr, &r->src.addr.v.a.addr,
723 				    &r->src.addr.v.a.mask, daddr, pd->af);
724 			break;
725 		}
726 		break;
727 	case PF_RDR: {
728 		if (pf_map_addr(pd->af, r, saddr, naddr, NULL, NULL, sn))
729 			goto notrans;
730 		if ((r->rpool.opts & PF_POOL_TYPEMASK) == PF_POOL_BITMASK)
731 			PF_POOLMASK(naddr, naddr, &r->rpool.cur->addr.v.a.mask,
732 			    daddr, pd->af);
733 
734 		/* Do not change SCTP ports. */
735 		if (pd->proto == IPPROTO_SCTP)
736 			break;
737 
738 		if (r->rpool.proxy_port[1]) {
739 			uint32_t	tmp_nport;
740 
741 			tmp_nport = ((ntohs(dport) - ntohs(r->dst.port[0])) %
742 			    (r->rpool.proxy_port[1] - r->rpool.proxy_port[0] +
743 			    1)) + r->rpool.proxy_port[0];
744 
745 			/* Wrap around if necessary. */
746 			if (tmp_nport > 65535)
747 				tmp_nport -= 65535;
748 			*nport = htons((uint16_t)tmp_nport);
749 		} else if (r->rpool.proxy_port[0])
750 			*nport = htons(r->rpool.proxy_port[0]);
751 		break;
752 	}
753 	default:
754 		panic("%s: unknown action %u", __func__, r->action);
755 	}
756 
757 	/* Return success only if translation really happened. */
758 	if (bcmp(*skp, *nkp, sizeof(struct pf_state_key_cmp)))
759 		return (r);
760 
761 notrans:
762 	uma_zfree(V_pf_state_key_z, *nkp);
763 	uma_zfree(V_pf_state_key_z, *skp);
764 	*skp = *nkp = NULL;
765 	*sn = NULL;
766 
767 	return (NULL);
768 }
769