xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 5bd73b51076b5cb5a2c9810f76c1d7ed20c4460e)
1 /*-
2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
28 
29 /*
30  * The FreeBSD IP packet firewall, main file
31  */
32 
33 #include "opt_ipfw.h"
34 #include "opt_ipdivert.h"
35 #include "opt_inet.h"
36 #ifndef INET
37 #error "IPFIREWALL requires INET"
38 #endif /* INET */
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/condvar.h>
45 #include <sys/counter.h>
46 #include <sys/eventhandler.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/jail.h>
52 #include <sys/module.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/rwlock.h>
56 #include <sys/rmlock.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/syslog.h>
61 #include <sys/ucred.h>
62 #include <net/ethernet.h> /* for ETHERTYPE_IP */
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #include <net/pfil.h>
67 #include <net/vnet.h>
68 
69 #include <netpfil/pf/pf_mtag.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet/pim.h>
80 #include <netinet/tcp_var.h>
81 #include <netinet/udp.h>
82 #include <netinet/udp_var.h>
83 #include <netinet/sctp.h>
84 
85 #include <netinet/ip6.h>
86 #include <netinet/icmp6.h>
87 #ifdef INET6
88 #include <netinet6/in6_pcb.h>
89 #include <netinet6/scope6_var.h>
90 #include <netinet6/ip6_var.h>
91 #endif
92 
93 #include <netpfil/ipfw/ip_fw_private.h>
94 
95 #include <machine/in_cksum.h>	/* XXX for in_cksum */
96 
97 #ifdef MAC
98 #include <security/mac/mac_framework.h>
99 #endif
100 
101 /*
102  * static variables followed by global ones.
103  * All ipfw global variables are here.
104  */
105 
106 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
107 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
108 
109 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
110 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
111 
112 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
113 static int default_to_accept = 1;
114 #else
115 static int default_to_accept;
116 #endif
117 
118 VNET_DEFINE(int, autoinc_step);
119 VNET_DEFINE(int, fw_one_pass) = 1;
120 
121 VNET_DEFINE(unsigned int, fw_tables_max);
122 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;	/* Don't use set-aware tables */
123 /* Use 128 tables by default */
124 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
125 
126 #ifndef LINEAR_SKIPTO
127 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
128     int tablearg, int jump_backwards);
129 #define	JUMP(ch, f, num, targ, back)	jump_fast(ch, f, num, targ, back)
130 #else
131 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
132     int tablearg, int jump_backwards);
133 #define	JUMP(ch, f, num, targ, back)	jump_linear(ch, f, num, targ, back)
134 #endif
135 
136 /*
137  * Each rule belongs to one of 32 different sets (0..31).
138  * The variable set_disable contains one bit per set.
139  * If the bit is set, all rules in the corresponding set
140  * are disabled. Set RESVD_SET(31) is reserved for the default rule
141  * and rules that are not deleted by the flush command,
142  * and CANNOT be disabled.
143  * Rules in set RESVD_SET can only be deleted individually.
144  */
145 VNET_DEFINE(u_int32_t, set_disable);
146 #define	V_set_disable			VNET(set_disable)
147 
148 VNET_DEFINE(int, fw_verbose);
149 /* counter for ipfw_log(NULL...) */
150 VNET_DEFINE(u_int64_t, norule_counter);
151 VNET_DEFINE(int, verbose_limit);
152 
153 /* layer3_chain contains the list of rules for layer 3 */
154 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
155 
156 /* ipfw_vnet_ready controls when we are open for business */
157 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
158 
159 VNET_DEFINE(int, ipfw_nat_ready) = 0;
160 
161 ipfw_nat_t *ipfw_nat_ptr = NULL;
162 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
163 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
164 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
165 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
166 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
167 
168 #ifdef SYSCTL_NODE
169 uint32_t dummy_def = IPFW_DEFAULT_RULE;
170 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
171 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
172 
173 SYSBEGIN(f3)
174 
175 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
176 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
177     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
178     "Only do a single pass through ipfw when using dummynet(4)");
179 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
180     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
181     "Rule number auto-increment step");
182 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
183     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
184     "Log matches to ipfw rules");
185 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
186     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
187     "Set upper limit of matches of ipfw rules logged");
188 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
189     &dummy_def, 0,
190     "The default/max possible rule number.");
191 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
192     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
193     "Maximum number of concurrently used tables");
194 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
195     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_tables_sets, "IU",
196     "Use per-set namespace for tables");
197 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
198     &default_to_accept, 0,
199     "Make the default rule accept all packets.");
200 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
201 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
202     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
203     "Number of static rules");
204 
205 #ifdef INET6
206 SYSCTL_DECL(_net_inet6_ip6);
207 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
208 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
209     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
210     "Deny packets with unknown IPv6 Extension Headers");
211 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
212     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
213     "Permit single packet IPv6 fragments");
214 #endif /* INET6 */
215 
216 SYSEND
217 
218 #endif /* SYSCTL_NODE */
219 
220 
221 /*
222  * Some macros used in the various matching options.
223  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
224  * Other macros just cast void * into the appropriate type
225  */
226 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
227 #define	TCP(p)		((struct tcphdr *)(p))
228 #define	SCTP(p)		((struct sctphdr *)(p))
229 #define	UDP(p)		((struct udphdr *)(p))
230 #define	ICMP(p)		((struct icmphdr *)(p))
231 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
232 
233 static __inline int
234 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
235 {
236 	int type = icmp->icmp_type;
237 
238 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
239 }
240 
241 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
242     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
243 
244 static int
245 is_icmp_query(struct icmphdr *icmp)
246 {
247 	int type = icmp->icmp_type;
248 
249 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
250 }
251 #undef TT
252 
253 /*
254  * The following checks use two arrays of 8 or 16 bits to store the
255  * bits that we want set or clear, respectively. They are in the
256  * low and high half of cmd->arg1 or cmd->d[0].
257  *
258  * We scan options and store the bits we find set. We succeed if
259  *
260  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
261  *
262  * The code is sometimes optimized not to store additional variables.
263  */
264 
265 static int
266 flags_match(ipfw_insn *cmd, u_int8_t bits)
267 {
268 	u_char want_clear;
269 	bits = ~bits;
270 
271 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
272 		return 0; /* some bits we want set were clear */
273 	want_clear = (cmd->arg1 >> 8) & 0xff;
274 	if ( (want_clear & bits) != want_clear)
275 		return 0; /* some bits we want clear were set */
276 	return 1;
277 }
278 
279 static int
280 ipopts_match(struct ip *ip, ipfw_insn *cmd)
281 {
282 	int optlen, bits = 0;
283 	u_char *cp = (u_char *)(ip + 1);
284 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
285 
286 	for (; x > 0; x -= optlen, cp += optlen) {
287 		int opt = cp[IPOPT_OPTVAL];
288 
289 		if (opt == IPOPT_EOL)
290 			break;
291 		if (opt == IPOPT_NOP)
292 			optlen = 1;
293 		else {
294 			optlen = cp[IPOPT_OLEN];
295 			if (optlen <= 0 || optlen > x)
296 				return 0; /* invalid or truncated */
297 		}
298 		switch (opt) {
299 
300 		default:
301 			break;
302 
303 		case IPOPT_LSRR:
304 			bits |= IP_FW_IPOPT_LSRR;
305 			break;
306 
307 		case IPOPT_SSRR:
308 			bits |= IP_FW_IPOPT_SSRR;
309 			break;
310 
311 		case IPOPT_RR:
312 			bits |= IP_FW_IPOPT_RR;
313 			break;
314 
315 		case IPOPT_TS:
316 			bits |= IP_FW_IPOPT_TS;
317 			break;
318 		}
319 	}
320 	return (flags_match(cmd, bits));
321 }
322 
323 static int
324 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
325 {
326 	int optlen, bits = 0;
327 	u_char *cp = (u_char *)(tcp + 1);
328 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
329 
330 	for (; x > 0; x -= optlen, cp += optlen) {
331 		int opt = cp[0];
332 		if (opt == TCPOPT_EOL)
333 			break;
334 		if (opt == TCPOPT_NOP)
335 			optlen = 1;
336 		else {
337 			optlen = cp[1];
338 			if (optlen <= 0)
339 				break;
340 		}
341 
342 		switch (opt) {
343 
344 		default:
345 			break;
346 
347 		case TCPOPT_MAXSEG:
348 			bits |= IP_FW_TCPOPT_MSS;
349 			break;
350 
351 		case TCPOPT_WINDOW:
352 			bits |= IP_FW_TCPOPT_WINDOW;
353 			break;
354 
355 		case TCPOPT_SACK_PERMITTED:
356 		case TCPOPT_SACK:
357 			bits |= IP_FW_TCPOPT_SACK;
358 			break;
359 
360 		case TCPOPT_TIMESTAMP:
361 			bits |= IP_FW_TCPOPT_TS;
362 			break;
363 
364 		}
365 	}
366 	return (flags_match(cmd, bits));
367 }
368 
369 static int
370 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
371     uint32_t *tablearg)
372 {
373 
374 	if (ifp == NULL)	/* no iface with this packet, match fails */
375 		return (0);
376 
377 	/* Check by name or by IP address */
378 	if (cmd->name[0] != '\0') { /* match by name */
379 		if (cmd->name[0] == '\1') /* use tablearg to match */
380 			return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0,
381 				&ifp->if_index, tablearg);
382 		/* Check name */
383 		if (cmd->p.glob) {
384 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
385 				return(1);
386 		} else {
387 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
388 				return(1);
389 		}
390 	} else {
391 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
392 		struct ifaddr *ia;
393 
394 		if_addr_rlock(ifp);
395 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
396 			if (ia->ifa_addr->sa_family != AF_INET)
397 				continue;
398 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
399 			    (ia->ifa_addr))->sin_addr.s_addr) {
400 				if_addr_runlock(ifp);
401 				return(1);	/* match */
402 			}
403 		}
404 		if_addr_runlock(ifp);
405 #endif /* __FreeBSD__ */
406 	}
407 	return(0);	/* no match, fail ... */
408 }
409 
410 /*
411  * The verify_path function checks if a route to the src exists and
412  * if it is reachable via ifp (when provided).
413  *
414  * The 'verrevpath' option checks that the interface that an IP packet
415  * arrives on is the same interface that traffic destined for the
416  * packet's source address would be routed out of.
417  * The 'versrcreach' option just checks that the source address is
418  * reachable via any route (except default) in the routing table.
419  * These two are a measure to block forged packets. This is also
420  * commonly known as "anti-spoofing" or Unicast Reverse Path
421  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
422  * is purposely reminiscent of the Cisco IOS command,
423  *
424  *   ip verify unicast reverse-path
425  *   ip verify unicast source reachable-via any
426  *
427  * which implements the same functionality. But note that the syntax
428  * is misleading, and the check may be performed on all IP packets
429  * whether unicast, multicast, or broadcast.
430  */
431 static int
432 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
433 {
434 #if defined(USERSPACE) || !defined(__FreeBSD__)
435 	return 0;
436 #else
437 	struct route ro;
438 	struct sockaddr_in *dst;
439 
440 	bzero(&ro, sizeof(ro));
441 
442 	dst = (struct sockaddr_in *)&(ro.ro_dst);
443 	dst->sin_family = AF_INET;
444 	dst->sin_len = sizeof(*dst);
445 	dst->sin_addr = src;
446 	in_rtalloc_ign(&ro, 0, fib);
447 
448 	if (ro.ro_rt == NULL)
449 		return 0;
450 
451 	/*
452 	 * If ifp is provided, check for equality with rtentry.
453 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
454 	 * in order to pass packets injected back by if_simloop():
455 	 * routing entry (via lo0) for our own address
456 	 * may exist, so we need to handle routing assymetry.
457 	 */
458 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
459 		RTFREE(ro.ro_rt);
460 		return 0;
461 	}
462 
463 	/* if no ifp provided, check if rtentry is not default route */
464 	if (ifp == NULL &&
465 	     satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
466 		RTFREE(ro.ro_rt);
467 		return 0;
468 	}
469 
470 	/* or if this is a blackhole/reject route */
471 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
472 		RTFREE(ro.ro_rt);
473 		return 0;
474 	}
475 
476 	/* found valid route */
477 	RTFREE(ro.ro_rt);
478 	return 1;
479 #endif /* __FreeBSD__ */
480 }
481 
482 #ifdef INET6
483 /*
484  * ipv6 specific rules here...
485  */
486 static __inline int
487 icmp6type_match (int type, ipfw_insn_u32 *cmd)
488 {
489 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
490 }
491 
492 static int
493 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
494 {
495 	int i;
496 	for (i=0; i <= cmd->o.arg1; ++i )
497 		if (curr_flow == cmd->d[i] )
498 			return 1;
499 	return 0;
500 }
501 
502 /* support for IP6_*_ME opcodes */
503 static int
504 search_ip6_addr_net (struct in6_addr * ip6_addr)
505 {
506 	struct ifnet *mdc;
507 	struct ifaddr *mdc2;
508 	struct in6_ifaddr *fdm;
509 	struct in6_addr copia;
510 
511 	TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
512 		if_addr_rlock(mdc);
513 		TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
514 			if (mdc2->ifa_addr->sa_family == AF_INET6) {
515 				fdm = (struct in6_ifaddr *)mdc2;
516 				copia = fdm->ia_addr.sin6_addr;
517 				/* need for leaving scope_id in the sock_addr */
518 				in6_clearscope(&copia);
519 				if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
520 					if_addr_runlock(mdc);
521 					return 1;
522 				}
523 			}
524 		}
525 		if_addr_runlock(mdc);
526 	}
527 	return 0;
528 }
529 
530 static int
531 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
532 {
533 	struct route_in6 ro;
534 	struct sockaddr_in6 *dst;
535 
536 	bzero(&ro, sizeof(ro));
537 
538 	dst = (struct sockaddr_in6 * )&(ro.ro_dst);
539 	dst->sin6_family = AF_INET6;
540 	dst->sin6_len = sizeof(*dst);
541 	dst->sin6_addr = *src;
542 
543 	in6_rtalloc_ign(&ro, 0, fib);
544 	if (ro.ro_rt == NULL)
545 		return 0;
546 
547 	/*
548 	 * if ifp is provided, check for equality with rtentry
549 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
550 	 * to support the case of sending packets to an address of our own.
551 	 * (where the former interface is the first argument of if_simloop()
552 	 *  (=ifp), the latter is lo0)
553 	 */
554 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
555 		RTFREE(ro.ro_rt);
556 		return 0;
557 	}
558 
559 	/* if no ifp provided, check if rtentry is not default route */
560 	if (ifp == NULL &&
561 	    IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
562 		RTFREE(ro.ro_rt);
563 		return 0;
564 	}
565 
566 	/* or if this is a blackhole/reject route */
567 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
568 		RTFREE(ro.ro_rt);
569 		return 0;
570 	}
571 
572 	/* found valid route */
573 	RTFREE(ro.ro_rt);
574 	return 1;
575 
576 }
577 
578 static int
579 is_icmp6_query(int icmp6_type)
580 {
581 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
582 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
583 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
584 	    icmp6_type == ICMP6_WRUREQUEST ||
585 	    icmp6_type == ICMP6_FQDN_QUERY ||
586 	    icmp6_type == ICMP6_NI_QUERY))
587 		return (1);
588 
589 	return (0);
590 }
591 
592 static void
593 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
594 {
595 	struct mbuf *m;
596 
597 	m = args->m;
598 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
599 		struct tcphdr *tcp;
600 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
601 
602 		if ((tcp->th_flags & TH_RST) == 0) {
603 			struct mbuf *m0;
604 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
605 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
606 			    tcp->th_flags | TH_RST);
607 			if (m0 != NULL)
608 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
609 				    NULL);
610 		}
611 		FREE_PKT(m);
612 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
613 #if 0
614 		/*
615 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
616 		 * as the contents are read. We need to m_adj() the
617 		 * needed amount.
618 		 * The mbuf will however be thrown away so we can adjust it.
619 		 * Remember we did an m_pullup on it already so we
620 		 * can make some assumptions about contiguousness.
621 		 */
622 		if (args->L3offset)
623 			m_adj(m, args->L3offset);
624 #endif
625 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
626 	} else
627 		FREE_PKT(m);
628 
629 	args->m = NULL;
630 }
631 
632 #endif /* INET6 */
633 
634 
635 /*
636  * sends a reject message, consuming the mbuf passed as an argument.
637  */
638 static void
639 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
640 {
641 
642 #if 0
643 	/* XXX When ip is not guaranteed to be at mtod() we will
644 	 * need to account for this */
645 	 * The mbuf will however be thrown away so we can adjust it.
646 	 * Remember we did an m_pullup on it already so we
647 	 * can make some assumptions about contiguousness.
648 	 */
649 	if (args->L3offset)
650 		m_adj(m, args->L3offset);
651 #endif
652 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
653 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
654 	} else if (args->f_id.proto == IPPROTO_TCP) {
655 		struct tcphdr *const tcp =
656 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
657 		if ( (tcp->th_flags & TH_RST) == 0) {
658 			struct mbuf *m;
659 			m = ipfw_send_pkt(args->m, &(args->f_id),
660 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
661 				tcp->th_flags | TH_RST);
662 			if (m != NULL)
663 				ip_output(m, NULL, NULL, 0, NULL, NULL);
664 		}
665 		FREE_PKT(args->m);
666 	} else
667 		FREE_PKT(args->m);
668 	args->m = NULL;
669 }
670 
671 /*
672  * Support for uid/gid/jail lookup. These tests are expensive
673  * (because we may need to look into the list of active sockets)
674  * so we cache the results. ugid_lookupp is 0 if we have not
675  * yet done a lookup, 1 if we succeeded, and -1 if we tried
676  * and failed. The function always returns the match value.
677  * We could actually spare the variable and use *uc, setting
678  * it to '(void *)check_uidgid if we have no info, NULL if
679  * we tried and failed, or any other value if successful.
680  */
681 static int
682 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
683     struct ucred **uc)
684 {
685 #if defined(USERSPACE)
686 	return 0;	// not supported in userspace
687 #else
688 #ifndef __FreeBSD__
689 	/* XXX */
690 	return cred_check(insn, proto, oif,
691 	    dst_ip, dst_port, src_ip, src_port,
692 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
693 #else  /* FreeBSD */
694 	struct in_addr src_ip, dst_ip;
695 	struct inpcbinfo *pi;
696 	struct ipfw_flow_id *id;
697 	struct inpcb *pcb, *inp;
698 	struct ifnet *oif;
699 	int lookupflags;
700 	int match;
701 
702 	id = &args->f_id;
703 	inp = args->inp;
704 	oif = args->oif;
705 
706 	/*
707 	 * Check to see if the UDP or TCP stack supplied us with
708 	 * the PCB. If so, rather then holding a lock and looking
709 	 * up the PCB, we can use the one that was supplied.
710 	 */
711 	if (inp && *ugid_lookupp == 0) {
712 		INP_LOCK_ASSERT(inp);
713 		if (inp->inp_socket != NULL) {
714 			*uc = crhold(inp->inp_cred);
715 			*ugid_lookupp = 1;
716 		} else
717 			*ugid_lookupp = -1;
718 	}
719 	/*
720 	 * If we have already been here and the packet has no
721 	 * PCB entry associated with it, then we can safely
722 	 * assume that this is a no match.
723 	 */
724 	if (*ugid_lookupp == -1)
725 		return (0);
726 	if (id->proto == IPPROTO_TCP) {
727 		lookupflags = 0;
728 		pi = &V_tcbinfo;
729 	} else if (id->proto == IPPROTO_UDP) {
730 		lookupflags = INPLOOKUP_WILDCARD;
731 		pi = &V_udbinfo;
732 	} else
733 		return 0;
734 	lookupflags |= INPLOOKUP_RLOCKPCB;
735 	match = 0;
736 	if (*ugid_lookupp == 0) {
737 		if (id->addr_type == 6) {
738 #ifdef INET6
739 			if (oif == NULL)
740 				pcb = in6_pcblookup_mbuf(pi,
741 				    &id->src_ip6, htons(id->src_port),
742 				    &id->dst_ip6, htons(id->dst_port),
743 				    lookupflags, oif, args->m);
744 			else
745 				pcb = in6_pcblookup_mbuf(pi,
746 				    &id->dst_ip6, htons(id->dst_port),
747 				    &id->src_ip6, htons(id->src_port),
748 				    lookupflags, oif, args->m);
749 #else
750 			*ugid_lookupp = -1;
751 			return (0);
752 #endif
753 		} else {
754 			src_ip.s_addr = htonl(id->src_ip);
755 			dst_ip.s_addr = htonl(id->dst_ip);
756 			if (oif == NULL)
757 				pcb = in_pcblookup_mbuf(pi,
758 				    src_ip, htons(id->src_port),
759 				    dst_ip, htons(id->dst_port),
760 				    lookupflags, oif, args->m);
761 			else
762 				pcb = in_pcblookup_mbuf(pi,
763 				    dst_ip, htons(id->dst_port),
764 				    src_ip, htons(id->src_port),
765 				    lookupflags, oif, args->m);
766 		}
767 		if (pcb != NULL) {
768 			INP_RLOCK_ASSERT(pcb);
769 			*uc = crhold(pcb->inp_cred);
770 			*ugid_lookupp = 1;
771 			INP_RUNLOCK(pcb);
772 		}
773 		if (*ugid_lookupp == 0) {
774 			/*
775 			 * We tried and failed, set the variable to -1
776 			 * so we will not try again on this packet.
777 			 */
778 			*ugid_lookupp = -1;
779 			return (0);
780 		}
781 	}
782 	if (insn->o.opcode == O_UID)
783 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
784 	else if (insn->o.opcode == O_GID)
785 		match = groupmember((gid_t)insn->d[0], *uc);
786 	else if (insn->o.opcode == O_JAIL)
787 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
788 	return (match);
789 #endif /* __FreeBSD__ */
790 #endif /* not supported in userspace */
791 }
792 
793 /*
794  * Helper function to set args with info on the rule after the matching
795  * one. slot is precise, whereas we guess rule_id as they are
796  * assigned sequentially.
797  */
798 static inline void
799 set_match(struct ip_fw_args *args, int slot,
800 	struct ip_fw_chain *chain)
801 {
802 	args->rule.chain_id = chain->id;
803 	args->rule.slot = slot + 1; /* we use 0 as a marker */
804 	args->rule.rule_id = 1 + chain->map[slot]->id;
805 	args->rule.rulenum = chain->map[slot]->rulenum;
806 }
807 
808 #ifndef LINEAR_SKIPTO
809 /*
810  * Helper function to enable cached rule lookups using
811  * cached_id and cached_pos fields in ipfw rule.
812  */
813 static int
814 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
815     int tablearg, int jump_backwards)
816 {
817 	int f_pos;
818 
819 	/* If possible use cached f_pos (in f->cached_pos),
820 	 * whose version is written in f->cached_id
821 	 * (horrible hacks to avoid changing the ABI).
822 	 */
823 	if (num != IP_FW_TARG && f->cached_id == chain->id)
824 		f_pos = f->cached_pos;
825 	else {
826 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
827 		/* make sure we do not jump backward */
828 		if (jump_backwards == 0 && i <= f->rulenum)
829 			i = f->rulenum + 1;
830 		if (chain->idxmap != NULL)
831 			f_pos = chain->idxmap[i];
832 		else
833 			f_pos = ipfw_find_rule(chain, i, 0);
834 		/* update the cache */
835 		if (num != IP_FW_TARG) {
836 			f->cached_id = chain->id;
837 			f->cached_pos = f_pos;
838 		}
839 	}
840 
841 	return (f_pos);
842 }
843 #else
844 /*
845  * Helper function to enable real fast rule lookups.
846  */
847 static int
848 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
849     int tablearg, int jump_backwards)
850 {
851 	int f_pos;
852 
853 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
854 	/* make sure we do not jump backward */
855 	if (jump_backwards == 0 && num <= f->rulenum)
856 		num = f->rulenum + 1;
857 	f_pos = chain->idxmap[num];
858 
859 	return (f_pos);
860 }
861 #endif
862 
863 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
864 /*
865  * The main check routine for the firewall.
866  *
867  * All arguments are in args so we can modify them and return them
868  * back to the caller.
869  *
870  * Parameters:
871  *
872  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
873  *		Starts with the IP header.
874  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
875  *	args->L3offset	Number of bytes bypassed if we came from L2.
876  *			e.g. often sizeof(eh)  ** NOTYET **
877  *	args->oif	Outgoing interface, NULL if packet is incoming.
878  *		The incoming interface is in the mbuf. (in)
879  *	args->divert_rule (in/out)
880  *		Skip up to the first rule past this rule number;
881  *		upon return, non-zero port number for divert or tee.
882  *
883  *	args->rule	Pointer to the last matching rule (in/out)
884  *	args->next_hop	Socket we are forwarding to (out).
885  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
886  *	args->f_id	Addresses grabbed from the packet (out)
887  * 	args->rule.info	a cookie depending on rule action
888  *
889  * Return value:
890  *
891  *	IP_FW_PASS	the packet must be accepted
892  *	IP_FW_DENY	the packet must be dropped
893  *	IP_FW_DIVERT	divert packet, port in m_tag
894  *	IP_FW_TEE	tee packet, port in m_tag
895  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
896  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
897  *		args->rule contains the matching rule,
898  *		args->rule.info has additional information.
899  *
900  */
901 int
902 ipfw_chk(struct ip_fw_args *args)
903 {
904 
905 	/*
906 	 * Local variables holding state while processing a packet:
907 	 *
908 	 * IMPORTANT NOTE: to speed up the processing of rules, there
909 	 * are some assumption on the values of the variables, which
910 	 * are documented here. Should you change them, please check
911 	 * the implementation of the various instructions to make sure
912 	 * that they still work.
913 	 *
914 	 * args->eh	The MAC header. It is non-null for a layer2
915 	 *	packet, it is NULL for a layer-3 packet.
916 	 * **notyet**
917 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
918 	 *
919 	 * m | args->m	Pointer to the mbuf, as received from the caller.
920 	 *	It may change if ipfw_chk() does an m_pullup, or if it
921 	 *	consumes the packet because it calls send_reject().
922 	 *	XXX This has to change, so that ipfw_chk() never modifies
923 	 *	or consumes the buffer.
924 	 * ip	is the beginning of the ip(4 or 6) header.
925 	 *	Calculated by adding the L3offset to the start of data.
926 	 *	(Until we start using L3offset, the packet is
927 	 *	supposed to start with the ip header).
928 	 */
929 	struct mbuf *m = args->m;
930 	struct ip *ip = mtod(m, struct ip *);
931 
932 	/*
933 	 * For rules which contain uid/gid or jail constraints, cache
934 	 * a copy of the users credentials after the pcb lookup has been
935 	 * executed. This will speed up the processing of rules with
936 	 * these types of constraints, as well as decrease contention
937 	 * on pcb related locks.
938 	 */
939 #ifndef __FreeBSD__
940 	struct bsd_ucred ucred_cache;
941 #else
942 	struct ucred *ucred_cache = NULL;
943 #endif
944 	int ucred_lookup = 0;
945 
946 	/*
947 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
948 	 *	inbound path (ether_input, ip_input).
949 	 *	If non-NULL, ipfw_chk has been called on the outbound path
950 	 *	(ether_output, ip_output).
951 	 */
952 	struct ifnet *oif = args->oif;
953 
954 	int f_pos = 0;		/* index of current rule in the array */
955 	int retval = 0;
956 
957 	/*
958 	 * hlen	The length of the IP header.
959 	 */
960 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
961 
962 	/*
963 	 * offset	The offset of a fragment. offset != 0 means that
964 	 *	we have a fragment at this offset of an IPv4 packet.
965 	 *	offset == 0 means that (if this is an IPv4 packet)
966 	 *	this is the first or only fragment.
967 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
968 	 *	or there is a single packet fragement (fragement header added
969 	 *	without needed).  We will treat a single packet fragment as if
970 	 *	there was no fragment header (or log/block depending on the
971 	 *	V_fw_permit_single_frag6 sysctl setting).
972 	 */
973 	u_short offset = 0;
974 	u_short ip6f_mf = 0;
975 
976 	/*
977 	 * Local copies of addresses. They are only valid if we have
978 	 * an IP packet.
979 	 *
980 	 * proto	The protocol. Set to 0 for non-ip packets,
981 	 *	or to the protocol read from the packet otherwise.
982 	 *	proto != 0 means that we have an IPv4 packet.
983 	 *
984 	 * src_port, dst_port	port numbers, in HOST format. Only
985 	 *	valid for TCP and UDP packets.
986 	 *
987 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
988 	 *	Only valid for IPv4 packets.
989 	 */
990 	uint8_t proto;
991 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
992 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
993 	uint16_t iplen=0;
994 	int pktlen;
995 	uint16_t	etype = 0;	/* Host order stored ether type */
996 
997 	/*
998 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
999 	 * 	MATCH_NONE when checked and not matched (q = NULL),
1000 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1001 	 */
1002 	int dyn_dir = MATCH_UNKNOWN;
1003 	ipfw_dyn_rule *q = NULL;
1004 	struct ip_fw_chain *chain = &V_layer3_chain;
1005 
1006 	/*
1007 	 * We store in ulp a pointer to the upper layer protocol header.
1008 	 * In the ipv4 case this is easy to determine from the header,
1009 	 * but for ipv6 we might have some additional headers in the middle.
1010 	 * ulp is NULL if not found.
1011 	 */
1012 	void *ulp = NULL;		/* upper layer protocol pointer. */
1013 
1014 	/* XXX ipv6 variables */
1015 	int is_ipv6 = 0;
1016 	uint8_t	icmp6_type = 0;
1017 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1018 	/* end of ipv6 variables */
1019 
1020 	int is_ipv4 = 0;
1021 
1022 	int done = 0;		/* flag to exit the outer loop */
1023 	IPFW_RLOCK_TRACKER;
1024 
1025 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1026 		return (IP_FW_PASS);	/* accept */
1027 
1028 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1029 	src_ip.s_addr = 0;		/* make sure it is initialized */
1030 	pktlen = m->m_pkthdr.len;
1031 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1032 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
1033 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1034 
1035 /*
1036  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1037  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1038  * pointer might become stale after other pullups (but we never use it
1039  * this way).
1040  */
1041 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1042 #define PULLUP_LEN(_len, p, T)					\
1043 do {								\
1044 	int x = (_len) + T;					\
1045 	if ((m)->m_len < x) {					\
1046 		args->m = m = m_pullup(m, x);			\
1047 		if (m == NULL)					\
1048 			goto pullup_failed;			\
1049 	}							\
1050 	p = (mtod(m, char *) + (_len));				\
1051 } while (0)
1052 
1053 	/*
1054 	 * if we have an ether header,
1055 	 */
1056 	if (args->eh)
1057 		etype = ntohs(args->eh->ether_type);
1058 
1059 	/* Identify IP packets and fill up variables. */
1060 	if (pktlen >= sizeof(struct ip6_hdr) &&
1061 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1062 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1063 		is_ipv6 = 1;
1064 		args->f_id.addr_type = 6;
1065 		hlen = sizeof(struct ip6_hdr);
1066 		proto = ip6->ip6_nxt;
1067 
1068 		/* Search extension headers to find upper layer protocols */
1069 		while (ulp == NULL && offset == 0) {
1070 			switch (proto) {
1071 			case IPPROTO_ICMPV6:
1072 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1073 				icmp6_type = ICMP6(ulp)->icmp6_type;
1074 				break;
1075 
1076 			case IPPROTO_TCP:
1077 				PULLUP_TO(hlen, ulp, struct tcphdr);
1078 				dst_port = TCP(ulp)->th_dport;
1079 				src_port = TCP(ulp)->th_sport;
1080 				/* save flags for dynamic rules */
1081 				args->f_id._flags = TCP(ulp)->th_flags;
1082 				break;
1083 
1084 			case IPPROTO_SCTP:
1085 				PULLUP_TO(hlen, ulp, struct sctphdr);
1086 				src_port = SCTP(ulp)->src_port;
1087 				dst_port = SCTP(ulp)->dest_port;
1088 				break;
1089 
1090 			case IPPROTO_UDP:
1091 				PULLUP_TO(hlen, ulp, struct udphdr);
1092 				dst_port = UDP(ulp)->uh_dport;
1093 				src_port = UDP(ulp)->uh_sport;
1094 				break;
1095 
1096 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1097 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1098 				ext_hd |= EXT_HOPOPTS;
1099 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1100 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1101 				ulp = NULL;
1102 				break;
1103 
1104 			case IPPROTO_ROUTING:	/* RFC 2460 */
1105 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1106 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1107 				case 0:
1108 					ext_hd |= EXT_RTHDR0;
1109 					break;
1110 				case 2:
1111 					ext_hd |= EXT_RTHDR2;
1112 					break;
1113 				default:
1114 					if (V_fw_verbose)
1115 						printf("IPFW2: IPV6 - Unknown "
1116 						    "Routing Header type(%d)\n",
1117 						    ((struct ip6_rthdr *)
1118 						    ulp)->ip6r_type);
1119 					if (V_fw_deny_unknown_exthdrs)
1120 					    return (IP_FW_DENY);
1121 					break;
1122 				}
1123 				ext_hd |= EXT_ROUTING;
1124 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1125 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1126 				ulp = NULL;
1127 				break;
1128 
1129 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1130 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1131 				ext_hd |= EXT_FRAGMENT;
1132 				hlen += sizeof (struct ip6_frag);
1133 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1134 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1135 					IP6F_OFF_MASK;
1136 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1137 					IP6F_MORE_FRAG;
1138 				if (V_fw_permit_single_frag6 == 0 &&
1139 				    offset == 0 && ip6f_mf == 0) {
1140 					if (V_fw_verbose)
1141 						printf("IPFW2: IPV6 - Invalid "
1142 						    "Fragment Header\n");
1143 					if (V_fw_deny_unknown_exthdrs)
1144 					    return (IP_FW_DENY);
1145 					break;
1146 				}
1147 				args->f_id.extra =
1148 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1149 				ulp = NULL;
1150 				break;
1151 
1152 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1153 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1154 				ext_hd |= EXT_DSTOPTS;
1155 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1156 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1157 				ulp = NULL;
1158 				break;
1159 
1160 			case IPPROTO_AH:	/* RFC 2402 */
1161 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1162 				ext_hd |= EXT_AH;
1163 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1164 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1165 				ulp = NULL;
1166 				break;
1167 
1168 			case IPPROTO_ESP:	/* RFC 2406 */
1169 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1170 				/* Anything past Seq# is variable length and
1171 				 * data past this ext. header is encrypted. */
1172 				ext_hd |= EXT_ESP;
1173 				break;
1174 
1175 			case IPPROTO_NONE:	/* RFC 2460 */
1176 				/*
1177 				 * Packet ends here, and IPv6 header has
1178 				 * already been pulled up. If ip6e_len!=0
1179 				 * then octets must be ignored.
1180 				 */
1181 				ulp = ip; /* non-NULL to get out of loop. */
1182 				break;
1183 
1184 			case IPPROTO_OSPFIGP:
1185 				/* XXX OSPF header check? */
1186 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1187 				break;
1188 
1189 			case IPPROTO_PIM:
1190 				/* XXX PIM header check? */
1191 				PULLUP_TO(hlen, ulp, struct pim);
1192 				break;
1193 
1194 			case IPPROTO_CARP:
1195 				PULLUP_TO(hlen, ulp, struct carp_header);
1196 				if (((struct carp_header *)ulp)->carp_version !=
1197 				    CARP_VERSION)
1198 					return (IP_FW_DENY);
1199 				if (((struct carp_header *)ulp)->carp_type !=
1200 				    CARP_ADVERTISEMENT)
1201 					return (IP_FW_DENY);
1202 				break;
1203 
1204 			case IPPROTO_IPV6:	/* RFC 2893 */
1205 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1206 				break;
1207 
1208 			case IPPROTO_IPV4:	/* RFC 2893 */
1209 				PULLUP_TO(hlen, ulp, struct ip);
1210 				break;
1211 
1212 			default:
1213 				if (V_fw_verbose)
1214 					printf("IPFW2: IPV6 - Unknown "
1215 					    "Extension Header(%d), ext_hd=%x\n",
1216 					     proto, ext_hd);
1217 				if (V_fw_deny_unknown_exthdrs)
1218 				    return (IP_FW_DENY);
1219 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1220 				break;
1221 			} /*switch */
1222 		}
1223 		ip = mtod(m, struct ip *);
1224 		ip6 = (struct ip6_hdr *)ip;
1225 		args->f_id.src_ip6 = ip6->ip6_src;
1226 		args->f_id.dst_ip6 = ip6->ip6_dst;
1227 		args->f_id.src_ip = 0;
1228 		args->f_id.dst_ip = 0;
1229 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1230 	} else if (pktlen >= sizeof(struct ip) &&
1231 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1232 	    	is_ipv4 = 1;
1233 		hlen = ip->ip_hl << 2;
1234 		args->f_id.addr_type = 4;
1235 
1236 		/*
1237 		 * Collect parameters into local variables for faster matching.
1238 		 */
1239 		proto = ip->ip_p;
1240 		src_ip = ip->ip_src;
1241 		dst_ip = ip->ip_dst;
1242 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1243 		iplen = ntohs(ip->ip_len);
1244 		pktlen = iplen < pktlen ? iplen : pktlen;
1245 
1246 		if (offset == 0) {
1247 			switch (proto) {
1248 			case IPPROTO_TCP:
1249 				PULLUP_TO(hlen, ulp, struct tcphdr);
1250 				dst_port = TCP(ulp)->th_dport;
1251 				src_port = TCP(ulp)->th_sport;
1252 				/* save flags for dynamic rules */
1253 				args->f_id._flags = TCP(ulp)->th_flags;
1254 				break;
1255 
1256 			case IPPROTO_SCTP:
1257 				PULLUP_TO(hlen, ulp, struct sctphdr);
1258 				src_port = SCTP(ulp)->src_port;
1259 				dst_port = SCTP(ulp)->dest_port;
1260 				break;
1261 
1262 			case IPPROTO_UDP:
1263 				PULLUP_TO(hlen, ulp, struct udphdr);
1264 				dst_port = UDP(ulp)->uh_dport;
1265 				src_port = UDP(ulp)->uh_sport;
1266 				break;
1267 
1268 			case IPPROTO_ICMP:
1269 				PULLUP_TO(hlen, ulp, struct icmphdr);
1270 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1271 				break;
1272 
1273 			default:
1274 				break;
1275 			}
1276 		}
1277 
1278 		ip = mtod(m, struct ip *);
1279 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1280 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1281 	}
1282 #undef PULLUP_TO
1283 	if (proto) { /* we may have port numbers, store them */
1284 		args->f_id.proto = proto;
1285 		args->f_id.src_port = src_port = ntohs(src_port);
1286 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1287 	}
1288 
1289 	IPFW_PF_RLOCK(chain);
1290 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1291 		IPFW_PF_RUNLOCK(chain);
1292 		return (IP_FW_PASS);	/* accept */
1293 	}
1294 	if (args->rule.slot) {
1295 		/*
1296 		 * Packet has already been tagged as a result of a previous
1297 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1298 		 * REASS, NETGRAPH, DIVERT/TEE...)
1299 		 * Validate the slot and continue from the next one
1300 		 * if still present, otherwise do a lookup.
1301 		 */
1302 		f_pos = (args->rule.chain_id == chain->id) ?
1303 		    args->rule.slot :
1304 		    ipfw_find_rule(chain, args->rule.rulenum,
1305 			args->rule.rule_id);
1306 	} else {
1307 		f_pos = 0;
1308 	}
1309 
1310 	/*
1311 	 * Now scan the rules, and parse microinstructions for each rule.
1312 	 * We have two nested loops and an inner switch. Sometimes we
1313 	 * need to break out of one or both loops, or re-enter one of
1314 	 * the loops with updated variables. Loop variables are:
1315 	 *
1316 	 *	f_pos (outer loop) points to the current rule.
1317 	 *		On output it points to the matching rule.
1318 	 *	done (outer loop) is used as a flag to break the loop.
1319 	 *	l (inner loop)	residual length of current rule.
1320 	 *		cmd points to the current microinstruction.
1321 	 *
1322 	 * We break the inner loop by setting l=0 and possibly
1323 	 * cmdlen=0 if we don't want to advance cmd.
1324 	 * We break the outer loop by setting done=1
1325 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1326 	 * as needed.
1327 	 */
1328 	for (; f_pos < chain->n_rules; f_pos++) {
1329 		ipfw_insn *cmd;
1330 		uint32_t tablearg = 0;
1331 		int l, cmdlen, skip_or; /* skip rest of OR block */
1332 		struct ip_fw *f;
1333 
1334 		f = chain->map[f_pos];
1335 		if (V_set_disable & (1 << f->set) )
1336 			continue;
1337 
1338 		skip_or = 0;
1339 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1340 		    l -= cmdlen, cmd += cmdlen) {
1341 			int match;
1342 
1343 			/*
1344 			 * check_body is a jump target used when we find a
1345 			 * CHECK_STATE, and need to jump to the body of
1346 			 * the target rule.
1347 			 */
1348 
1349 /* check_body: */
1350 			cmdlen = F_LEN(cmd);
1351 			/*
1352 			 * An OR block (insn_1 || .. || insn_n) has the
1353 			 * F_OR bit set in all but the last instruction.
1354 			 * The first match will set "skip_or", and cause
1355 			 * the following instructions to be skipped until
1356 			 * past the one with the F_OR bit clear.
1357 			 */
1358 			if (skip_or) {		/* skip this instruction */
1359 				if ((cmd->len & F_OR) == 0)
1360 					skip_or = 0;	/* next one is good */
1361 				continue;
1362 			}
1363 			match = 0; /* set to 1 if we succeed */
1364 
1365 			switch (cmd->opcode) {
1366 			/*
1367 			 * The first set of opcodes compares the packet's
1368 			 * fields with some pattern, setting 'match' if a
1369 			 * match is found. At the end of the loop there is
1370 			 * logic to deal with F_NOT and F_OR flags associated
1371 			 * with the opcode.
1372 			 */
1373 			case O_NOP:
1374 				match = 1;
1375 				break;
1376 
1377 			case O_FORWARD_MAC:
1378 				printf("ipfw: opcode %d unimplemented\n",
1379 				    cmd->opcode);
1380 				break;
1381 
1382 			case O_GID:
1383 			case O_UID:
1384 			case O_JAIL:
1385 				/*
1386 				 * We only check offset == 0 && proto != 0,
1387 				 * as this ensures that we have a
1388 				 * packet with the ports info.
1389 				 */
1390 				if (offset != 0)
1391 					break;
1392 				if (proto == IPPROTO_TCP ||
1393 				    proto == IPPROTO_UDP)
1394 					match = check_uidgid(
1395 						    (ipfw_insn_u32 *)cmd,
1396 						    args, &ucred_lookup,
1397 #ifdef __FreeBSD__
1398 						    &ucred_cache);
1399 #else
1400 						    (void *)&ucred_cache);
1401 #endif
1402 				break;
1403 
1404 			case O_RECV:
1405 				match = iface_match(m->m_pkthdr.rcvif,
1406 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1407 				break;
1408 
1409 			case O_XMIT:
1410 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1411 				    chain, &tablearg);
1412 				break;
1413 
1414 			case O_VIA:
1415 				match = iface_match(oif ? oif :
1416 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1417 				    chain, &tablearg);
1418 				break;
1419 
1420 			case O_MACADDR2:
1421 				if (args->eh != NULL) {	/* have MAC header */
1422 					u_int32_t *want = (u_int32_t *)
1423 						((ipfw_insn_mac *)cmd)->addr;
1424 					u_int32_t *mask = (u_int32_t *)
1425 						((ipfw_insn_mac *)cmd)->mask;
1426 					u_int32_t *hdr = (u_int32_t *)args->eh;
1427 
1428 					match =
1429 					    ( want[0] == (hdr[0] & mask[0]) &&
1430 					      want[1] == (hdr[1] & mask[1]) &&
1431 					      want[2] == (hdr[2] & mask[2]) );
1432 				}
1433 				break;
1434 
1435 			case O_MAC_TYPE:
1436 				if (args->eh != NULL) {
1437 					u_int16_t *p =
1438 					    ((ipfw_insn_u16 *)cmd)->ports;
1439 					int i;
1440 
1441 					for (i = cmdlen - 1; !match && i>0;
1442 					    i--, p += 2)
1443 						match = (etype >= p[0] &&
1444 						    etype <= p[1]);
1445 				}
1446 				break;
1447 
1448 			case O_FRAG:
1449 				match = (offset != 0);
1450 				break;
1451 
1452 			case O_IN:	/* "out" is "not in" */
1453 				match = (oif == NULL);
1454 				break;
1455 
1456 			case O_LAYER2:
1457 				match = (args->eh != NULL);
1458 				break;
1459 
1460 			case O_DIVERTED:
1461 			    {
1462 				/* For diverted packets, args->rule.info
1463 				 * contains the divert port (in host format)
1464 				 * reason and direction.
1465 				 */
1466 				uint32_t i = args->rule.info;
1467 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1468 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1469 			    }
1470 				break;
1471 
1472 			case O_PROTO:
1473 				/*
1474 				 * We do not allow an arg of 0 so the
1475 				 * check of "proto" only suffices.
1476 				 */
1477 				match = (proto == cmd->arg1);
1478 				break;
1479 
1480 			case O_IP_SRC:
1481 				match = is_ipv4 &&
1482 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1483 				    src_ip.s_addr);
1484 				break;
1485 
1486 			case O_IP_SRC_LOOKUP:
1487 			case O_IP_DST_LOOKUP:
1488 				if (is_ipv4) {
1489 				    uint32_t key =
1490 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1491 					    dst_ip.s_addr : src_ip.s_addr;
1492 				    uint32_t v = 0;
1493 
1494 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1495 					/* generic lookup. The key must be
1496 					 * in 32bit big-endian format.
1497 					 */
1498 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1499 					if (v == 0)
1500 					    key = dst_ip.s_addr;
1501 					else if (v == 1)
1502 					    key = src_ip.s_addr;
1503 					else if (v == 6) /* dscp */
1504 					    key = (ip->ip_tos >> 2) & 0x3f;
1505 					else if (offset != 0)
1506 					    break;
1507 					else if (proto != IPPROTO_TCP &&
1508 						proto != IPPROTO_UDP)
1509 					    break;
1510 					else if (v == 2)
1511 					    key = dst_port;
1512 					else if (v == 3)
1513 					    key = src_port;
1514 #ifndef USERSPACE
1515 					else if (v == 4 || v == 5) {
1516 					    check_uidgid(
1517 						(ipfw_insn_u32 *)cmd,
1518 						args, &ucred_lookup,
1519 #ifdef __FreeBSD__
1520 						&ucred_cache);
1521 					    if (v == 4 /* O_UID */)
1522 						key = ucred_cache->cr_uid;
1523 					    else if (v == 5 /* O_JAIL */)
1524 						key = ucred_cache->cr_prison->pr_id;
1525 #else /* !__FreeBSD__ */
1526 						(void *)&ucred_cache);
1527 					    if (v ==4 /* O_UID */)
1528 						key = ucred_cache.uid;
1529 					    else if (v == 5 /* O_JAIL */)
1530 						key = ucred_cache.xid;
1531 #endif /* !__FreeBSD__ */
1532 					} else
1533 #endif /* !USERSPACE */
1534 					    break;
1535 				    }
1536 				    match = ipfw_lookup_table(chain,
1537 					cmd->arg1, key, &v);
1538 				    if (!match)
1539 					break;
1540 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1541 					match =
1542 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1543 				    else
1544 					tablearg = v;
1545 				} else if (is_ipv6) {
1546 					uint32_t v = 0;
1547 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1548 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1549 					match = ipfw_lookup_table_extended(chain,
1550 							cmd->arg1,
1551 							sizeof(struct in6_addr),
1552 							pkey, &v);
1553 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1554 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1555 					if (match)
1556 						tablearg = v;
1557 				}
1558 				break;
1559 
1560 			case O_IP_FLOW_LOOKUP:
1561 				{
1562 					uint32_t v = 0;
1563 					match = ipfw_lookup_table_extended(chain,
1564 					    cmd->arg1, 0, &args->f_id, &v);
1565 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1566 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1567 					if (match)
1568 						tablearg = v;
1569 				}
1570 				break;
1571 			case O_IP_SRC_MASK:
1572 			case O_IP_DST_MASK:
1573 				if (is_ipv4) {
1574 				    uint32_t a =
1575 					(cmd->opcode == O_IP_DST_MASK) ?
1576 					    dst_ip.s_addr : src_ip.s_addr;
1577 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1578 				    int i = cmdlen-1;
1579 
1580 				    for (; !match && i>0; i-= 2, p+= 2)
1581 					match = (p[0] == (a & p[1]));
1582 				}
1583 				break;
1584 
1585 			case O_IP_SRC_ME:
1586 				if (is_ipv4) {
1587 					struct ifnet *tif;
1588 
1589 					INADDR_TO_IFP(src_ip, tif);
1590 					match = (tif != NULL);
1591 					break;
1592 				}
1593 #ifdef INET6
1594 				/* FALLTHROUGH */
1595 			case O_IP6_SRC_ME:
1596 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1597 #endif
1598 				break;
1599 
1600 			case O_IP_DST_SET:
1601 			case O_IP_SRC_SET:
1602 				if (is_ipv4) {
1603 					u_int32_t *d = (u_int32_t *)(cmd+1);
1604 					u_int32_t addr =
1605 					    cmd->opcode == O_IP_DST_SET ?
1606 						args->f_id.dst_ip :
1607 						args->f_id.src_ip;
1608 
1609 					    if (addr < d[0])
1610 						    break;
1611 					    addr -= d[0]; /* subtract base */
1612 					    match = (addr < cmd->arg1) &&
1613 						( d[ 1 + (addr>>5)] &
1614 						  (1<<(addr & 0x1f)) );
1615 				}
1616 				break;
1617 
1618 			case O_IP_DST:
1619 				match = is_ipv4 &&
1620 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1621 				    dst_ip.s_addr);
1622 				break;
1623 
1624 			case O_IP_DST_ME:
1625 				if (is_ipv4) {
1626 					struct ifnet *tif;
1627 
1628 					INADDR_TO_IFP(dst_ip, tif);
1629 					match = (tif != NULL);
1630 					break;
1631 				}
1632 #ifdef INET6
1633 				/* FALLTHROUGH */
1634 			case O_IP6_DST_ME:
1635 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1636 #endif
1637 				break;
1638 
1639 
1640 			case O_IP_SRCPORT:
1641 			case O_IP_DSTPORT:
1642 				/*
1643 				 * offset == 0 && proto != 0 is enough
1644 				 * to guarantee that we have a
1645 				 * packet with port info.
1646 				 */
1647 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1648 				    && offset == 0) {
1649 					u_int16_t x =
1650 					    (cmd->opcode == O_IP_SRCPORT) ?
1651 						src_port : dst_port ;
1652 					u_int16_t *p =
1653 					    ((ipfw_insn_u16 *)cmd)->ports;
1654 					int i;
1655 
1656 					for (i = cmdlen - 1; !match && i>0;
1657 					    i--, p += 2)
1658 						match = (x>=p[0] && x<=p[1]);
1659 				}
1660 				break;
1661 
1662 			case O_ICMPTYPE:
1663 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1664 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1665 				break;
1666 
1667 #ifdef INET6
1668 			case O_ICMP6TYPE:
1669 				match = is_ipv6 && offset == 0 &&
1670 				    proto==IPPROTO_ICMPV6 &&
1671 				    icmp6type_match(
1672 					ICMP6(ulp)->icmp6_type,
1673 					(ipfw_insn_u32 *)cmd);
1674 				break;
1675 #endif /* INET6 */
1676 
1677 			case O_IPOPT:
1678 				match = (is_ipv4 &&
1679 				    ipopts_match(ip, cmd) );
1680 				break;
1681 
1682 			case O_IPVER:
1683 				match = (is_ipv4 &&
1684 				    cmd->arg1 == ip->ip_v);
1685 				break;
1686 
1687 			case O_IPID:
1688 			case O_IPLEN:
1689 			case O_IPTTL:
1690 				if (is_ipv4) {	/* only for IP packets */
1691 				    uint16_t x;
1692 				    uint16_t *p;
1693 				    int i;
1694 
1695 				    if (cmd->opcode == O_IPLEN)
1696 					x = iplen;
1697 				    else if (cmd->opcode == O_IPTTL)
1698 					x = ip->ip_ttl;
1699 				    else /* must be IPID */
1700 					x = ntohs(ip->ip_id);
1701 				    if (cmdlen == 1) {
1702 					match = (cmd->arg1 == x);
1703 					break;
1704 				    }
1705 				    /* otherwise we have ranges */
1706 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1707 				    i = cmdlen - 1;
1708 				    for (; !match && i>0; i--, p += 2)
1709 					match = (x >= p[0] && x <= p[1]);
1710 				}
1711 				break;
1712 
1713 			case O_IPPRECEDENCE:
1714 				match = (is_ipv4 &&
1715 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1716 				break;
1717 
1718 			case O_IPTOS:
1719 				match = (is_ipv4 &&
1720 				    flags_match(cmd, ip->ip_tos));
1721 				break;
1722 
1723 			case O_DSCP:
1724 			    {
1725 				uint32_t *p;
1726 				uint16_t x;
1727 
1728 				p = ((ipfw_insn_u32 *)cmd)->d;
1729 
1730 				if (is_ipv4)
1731 					x = ip->ip_tos >> 2;
1732 				else if (is_ipv6) {
1733 					uint8_t *v;
1734 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
1735 					x = (*v & 0x0F) << 2;
1736 					v++;
1737 					x |= *v >> 6;
1738 				} else
1739 					break;
1740 
1741 				/* DSCP bitmask is stored as low_u32 high_u32 */
1742 				if (x > 32)
1743 					match = *(p + 1) & (1 << (x - 32));
1744 				else
1745 					match = *p & (1 << x);
1746 			    }
1747 				break;
1748 
1749 			case O_TCPDATALEN:
1750 				if (proto == IPPROTO_TCP && offset == 0) {
1751 				    struct tcphdr *tcp;
1752 				    uint16_t x;
1753 				    uint16_t *p;
1754 				    int i;
1755 
1756 				    tcp = TCP(ulp);
1757 				    x = iplen -
1758 					((ip->ip_hl + tcp->th_off) << 2);
1759 				    if (cmdlen == 1) {
1760 					match = (cmd->arg1 == x);
1761 					break;
1762 				    }
1763 				    /* otherwise we have ranges */
1764 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1765 				    i = cmdlen - 1;
1766 				    for (; !match && i>0; i--, p += 2)
1767 					match = (x >= p[0] && x <= p[1]);
1768 				}
1769 				break;
1770 
1771 			case O_TCPFLAGS:
1772 				match = (proto == IPPROTO_TCP && offset == 0 &&
1773 				    flags_match(cmd, TCP(ulp)->th_flags));
1774 				break;
1775 
1776 			case O_TCPOPTS:
1777 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
1778 					PULLUP_LEN(hlen, ulp,
1779 					    (TCP(ulp)->th_off << 2));
1780 					match = tcpopts_match(TCP(ulp), cmd);
1781 				}
1782 				break;
1783 
1784 			case O_TCPSEQ:
1785 				match = (proto == IPPROTO_TCP && offset == 0 &&
1786 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1787 					TCP(ulp)->th_seq);
1788 				break;
1789 
1790 			case O_TCPACK:
1791 				match = (proto == IPPROTO_TCP && offset == 0 &&
1792 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1793 					TCP(ulp)->th_ack);
1794 				break;
1795 
1796 			case O_TCPWIN:
1797 				if (proto == IPPROTO_TCP && offset == 0) {
1798 				    uint16_t x;
1799 				    uint16_t *p;
1800 				    int i;
1801 
1802 				    x = ntohs(TCP(ulp)->th_win);
1803 				    if (cmdlen == 1) {
1804 					match = (cmd->arg1 == x);
1805 					break;
1806 				    }
1807 				    /* Otherwise we have ranges. */
1808 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1809 				    i = cmdlen - 1;
1810 				    for (; !match && i > 0; i--, p += 2)
1811 					match = (x >= p[0] && x <= p[1]);
1812 				}
1813 				break;
1814 
1815 			case O_ESTAB:
1816 				/* reject packets which have SYN only */
1817 				/* XXX should i also check for TH_ACK ? */
1818 				match = (proto == IPPROTO_TCP && offset == 0 &&
1819 				    (TCP(ulp)->th_flags &
1820 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1821 				break;
1822 
1823 			case O_ALTQ: {
1824 				struct pf_mtag *at;
1825 				struct m_tag *mtag;
1826 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1827 
1828 				/*
1829 				 * ALTQ uses mbuf tags from another
1830 				 * packet filtering system - pf(4).
1831 				 * We allocate a tag in its format
1832 				 * and fill it in, pretending to be pf(4).
1833 				 */
1834 				match = 1;
1835 				at = pf_find_mtag(m);
1836 				if (at != NULL && at->qid != 0)
1837 					break;
1838 				mtag = m_tag_get(PACKET_TAG_PF,
1839 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1840 				if (mtag == NULL) {
1841 					/*
1842 					 * Let the packet fall back to the
1843 					 * default ALTQ.
1844 					 */
1845 					break;
1846 				}
1847 				m_tag_prepend(m, mtag);
1848 				at = (struct pf_mtag *)(mtag + 1);
1849 				at->qid = altq->qid;
1850 				at->hdr = ip;
1851 				break;
1852 			}
1853 
1854 			case O_LOG:
1855 				ipfw_log(chain, f, hlen, args, m,
1856 				    oif, offset | ip6f_mf, tablearg, ip);
1857 				match = 1;
1858 				break;
1859 
1860 			case O_PROB:
1861 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1862 				break;
1863 
1864 			case O_VERREVPATH:
1865 				/* Outgoing packets automatically pass/match */
1866 				match = ((oif != NULL) ||
1867 				    (m->m_pkthdr.rcvif == NULL) ||
1868 				    (
1869 #ifdef INET6
1870 				    is_ipv6 ?
1871 					verify_path6(&(args->f_id.src_ip6),
1872 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1873 #endif
1874 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1875 				        args->f_id.fib)));
1876 				break;
1877 
1878 			case O_VERSRCREACH:
1879 				/* Outgoing packets automatically pass/match */
1880 				match = (hlen > 0 && ((oif != NULL) ||
1881 #ifdef INET6
1882 				    is_ipv6 ?
1883 				        verify_path6(&(args->f_id.src_ip6),
1884 				            NULL, args->f_id.fib) :
1885 #endif
1886 				    verify_path(src_ip, NULL, args->f_id.fib)));
1887 				break;
1888 
1889 			case O_ANTISPOOF:
1890 				/* Outgoing packets automatically pass/match */
1891 				if (oif == NULL && hlen > 0 &&
1892 				    (  (is_ipv4 && in_localaddr(src_ip))
1893 #ifdef INET6
1894 				    || (is_ipv6 &&
1895 				        in6_localaddr(&(args->f_id.src_ip6)))
1896 #endif
1897 				    ))
1898 					match =
1899 #ifdef INET6
1900 					    is_ipv6 ? verify_path6(
1901 					        &(args->f_id.src_ip6),
1902 					        m->m_pkthdr.rcvif,
1903 						args->f_id.fib) :
1904 #endif
1905 					    verify_path(src_ip,
1906 					    	m->m_pkthdr.rcvif,
1907 					        args->f_id.fib);
1908 				else
1909 					match = 1;
1910 				break;
1911 
1912 			case O_IPSEC:
1913 #ifdef IPSEC
1914 				match = (m_tag_find(m,
1915 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1916 #endif
1917 				/* otherwise no match */
1918 				break;
1919 
1920 #ifdef INET6
1921 			case O_IP6_SRC:
1922 				match = is_ipv6 &&
1923 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1924 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1925 				break;
1926 
1927 			case O_IP6_DST:
1928 				match = is_ipv6 &&
1929 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1930 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1931 				break;
1932 			case O_IP6_SRC_MASK:
1933 			case O_IP6_DST_MASK:
1934 				if (is_ipv6) {
1935 					int i = cmdlen - 1;
1936 					struct in6_addr p;
1937 					struct in6_addr *d =
1938 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1939 
1940 					for (; !match && i > 0; d += 2,
1941 					    i -= F_INSN_SIZE(struct in6_addr)
1942 					    * 2) {
1943 						p = (cmd->opcode ==
1944 						    O_IP6_SRC_MASK) ?
1945 						    args->f_id.src_ip6:
1946 						    args->f_id.dst_ip6;
1947 						APPLY_MASK(&p, &d[1]);
1948 						match =
1949 						    IN6_ARE_ADDR_EQUAL(&d[0],
1950 						    &p);
1951 					}
1952 				}
1953 				break;
1954 
1955 			case O_FLOW6ID:
1956 				match = is_ipv6 &&
1957 				    flow6id_match(args->f_id.flow_id6,
1958 				    (ipfw_insn_u32 *) cmd);
1959 				break;
1960 
1961 			case O_EXT_HDR:
1962 				match = is_ipv6 &&
1963 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1964 				break;
1965 
1966 			case O_IP6:
1967 				match = is_ipv6;
1968 				break;
1969 #endif
1970 
1971 			case O_IP4:
1972 				match = is_ipv4;
1973 				break;
1974 
1975 			case O_TAG: {
1976 				struct m_tag *mtag;
1977 				uint32_t tag = TARG(cmd->arg1, tag);
1978 
1979 				/* Packet is already tagged with this tag? */
1980 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1981 
1982 				/* We have `untag' action when F_NOT flag is
1983 				 * present. And we must remove this mtag from
1984 				 * mbuf and reset `match' to zero (`match' will
1985 				 * be inversed later).
1986 				 * Otherwise we should allocate new mtag and
1987 				 * push it into mbuf.
1988 				 */
1989 				if (cmd->len & F_NOT) { /* `untag' action */
1990 					if (mtag != NULL)
1991 						m_tag_delete(m, mtag);
1992 					match = 0;
1993 				} else {
1994 					if (mtag == NULL) {
1995 						mtag = m_tag_alloc( MTAG_IPFW,
1996 						    tag, 0, M_NOWAIT);
1997 						if (mtag != NULL)
1998 							m_tag_prepend(m, mtag);
1999 					}
2000 					match = 1;
2001 				}
2002 				break;
2003 			}
2004 
2005 			case O_FIB: /* try match the specified fib */
2006 				if (args->f_id.fib == cmd->arg1)
2007 					match = 1;
2008 				break;
2009 
2010 			case O_SOCKARG:	{
2011 #ifndef USERSPACE	/* not supported in userspace */
2012 				struct inpcb *inp = args->inp;
2013 				struct inpcbinfo *pi;
2014 
2015 				if (is_ipv6) /* XXX can we remove this ? */
2016 					break;
2017 
2018 				if (proto == IPPROTO_TCP)
2019 					pi = &V_tcbinfo;
2020 				else if (proto == IPPROTO_UDP)
2021 					pi = &V_udbinfo;
2022 				else
2023 					break;
2024 
2025 				/*
2026 				 * XXXRW: so_user_cookie should almost
2027 				 * certainly be inp_user_cookie?
2028 				 */
2029 
2030 				/* For incomming packet, lookup up the
2031 				inpcb using the src/dest ip/port tuple */
2032 				if (inp == NULL) {
2033 					inp = in_pcblookup(pi,
2034 						src_ip, htons(src_port),
2035 						dst_ip, htons(dst_port),
2036 						INPLOOKUP_RLOCKPCB, NULL);
2037 					if (inp != NULL) {
2038 						tablearg =
2039 						    inp->inp_socket->so_user_cookie;
2040 						if (tablearg)
2041 							match = 1;
2042 						INP_RUNLOCK(inp);
2043 					}
2044 				} else {
2045 					if (inp->inp_socket) {
2046 						tablearg =
2047 						    inp->inp_socket->so_user_cookie;
2048 						if (tablearg)
2049 							match = 1;
2050 					}
2051 				}
2052 #endif /* !USERSPACE */
2053 				break;
2054 			}
2055 
2056 			case O_TAGGED: {
2057 				struct m_tag *mtag;
2058 				uint32_t tag = TARG(cmd->arg1, tag);
2059 
2060 				if (cmdlen == 1) {
2061 					match = m_tag_locate(m, MTAG_IPFW,
2062 					    tag, NULL) != NULL;
2063 					break;
2064 				}
2065 
2066 				/* we have ranges */
2067 				for (mtag = m_tag_first(m);
2068 				    mtag != NULL && !match;
2069 				    mtag = m_tag_next(m, mtag)) {
2070 					uint16_t *p;
2071 					int i;
2072 
2073 					if (mtag->m_tag_cookie != MTAG_IPFW)
2074 						continue;
2075 
2076 					p = ((ipfw_insn_u16 *)cmd)->ports;
2077 					i = cmdlen - 1;
2078 					for(; !match && i > 0; i--, p += 2)
2079 						match =
2080 						    mtag->m_tag_id >= p[0] &&
2081 						    mtag->m_tag_id <= p[1];
2082 				}
2083 				break;
2084 			}
2085 
2086 			/*
2087 			 * The second set of opcodes represents 'actions',
2088 			 * i.e. the terminal part of a rule once the packet
2089 			 * matches all previous patterns.
2090 			 * Typically there is only one action for each rule,
2091 			 * and the opcode is stored at the end of the rule
2092 			 * (but there are exceptions -- see below).
2093 			 *
2094 			 * In general, here we set retval and terminate the
2095 			 * outer loop (would be a 'break 3' in some language,
2096 			 * but we need to set l=0, done=1)
2097 			 *
2098 			 * Exceptions:
2099 			 * O_COUNT and O_SKIPTO actions:
2100 			 *   instead of terminating, we jump to the next rule
2101 			 *   (setting l=0), or to the SKIPTO target (setting
2102 			 *   f/f_len, cmd and l as needed), respectively.
2103 			 *
2104 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2105 			 *   perform some action and set match = 1;
2106 			 *
2107 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2108 			 *   not real 'actions', and are stored right
2109 			 *   before the 'action' part of the rule.
2110 			 *   These opcodes try to install an entry in the
2111 			 *   state tables; if successful, we continue with
2112 			 *   the next opcode (match=1; break;), otherwise
2113 			 *   the packet must be dropped (set retval,
2114 			 *   break loops with l=0, done=1)
2115 			 *
2116 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2117 			 *   cause a lookup of the state table, and a jump
2118 			 *   to the 'action' part of the parent rule
2119 			 *   if an entry is found, or
2120 			 *   (CHECK_STATE only) a jump to the next rule if
2121 			 *   the entry is not found.
2122 			 *   The result of the lookup is cached so that
2123 			 *   further instances of these opcodes become NOPs.
2124 			 *   The jump to the next rule is done by setting
2125 			 *   l=0, cmdlen=0.
2126 			 */
2127 			case O_LIMIT:
2128 			case O_KEEP_STATE:
2129 				if (ipfw_install_state(chain, f,
2130 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2131 					/* error or limit violation */
2132 					retval = IP_FW_DENY;
2133 					l = 0;	/* exit inner loop */
2134 					done = 1; /* exit outer loop */
2135 				}
2136 				match = 1;
2137 				break;
2138 
2139 			case O_PROBE_STATE:
2140 			case O_CHECK_STATE:
2141 				/*
2142 				 * dynamic rules are checked at the first
2143 				 * keep-state or check-state occurrence,
2144 				 * with the result being stored in dyn_dir.
2145 				 * The compiler introduces a PROBE_STATE
2146 				 * instruction for us when we have a
2147 				 * KEEP_STATE (because PROBE_STATE needs
2148 				 * to be run first).
2149 				 */
2150 				if (dyn_dir == MATCH_UNKNOWN &&
2151 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2152 				     &dyn_dir, proto == IPPROTO_TCP ?
2153 					TCP(ulp) : NULL))
2154 					!= NULL) {
2155 					/*
2156 					 * Found dynamic entry, update stats
2157 					 * and jump to the 'action' part of
2158 					 * the parent rule by setting
2159 					 * f, cmd, l and clearing cmdlen.
2160 					 */
2161 					IPFW_INC_DYN_COUNTER(q, pktlen);
2162 					/* XXX we would like to have f_pos
2163 					 * readily accessible in the dynamic
2164 				         * rule, instead of having to
2165 					 * lookup q->rule.
2166 					 */
2167 					f = q->rule;
2168 					f_pos = ipfw_find_rule(chain,
2169 						f->rulenum, f->id);
2170 					cmd = ACTION_PTR(f);
2171 					l = f->cmd_len - f->act_ofs;
2172 					ipfw_dyn_unlock(q);
2173 					cmdlen = 0;
2174 					match = 1;
2175 					break;
2176 				}
2177 				/*
2178 				 * Dynamic entry not found. If CHECK_STATE,
2179 				 * skip to next rule, if PROBE_STATE just
2180 				 * ignore and continue with next opcode.
2181 				 */
2182 				if (cmd->opcode == O_CHECK_STATE)
2183 					l = 0;	/* exit inner loop */
2184 				match = 1;
2185 				break;
2186 
2187 			case O_ACCEPT:
2188 				retval = 0;	/* accept */
2189 				l = 0;		/* exit inner loop */
2190 				done = 1;	/* exit outer loop */
2191 				break;
2192 
2193 			case O_PIPE:
2194 			case O_QUEUE:
2195 				set_match(args, f_pos, chain);
2196 				args->rule.info = TARG(cmd->arg1, pipe);
2197 				if (cmd->opcode == O_PIPE)
2198 					args->rule.info |= IPFW_IS_PIPE;
2199 				if (V_fw_one_pass)
2200 					args->rule.info |= IPFW_ONEPASS;
2201 				retval = IP_FW_DUMMYNET;
2202 				l = 0;          /* exit inner loop */
2203 				done = 1;       /* exit outer loop */
2204 				break;
2205 
2206 			case O_DIVERT:
2207 			case O_TEE:
2208 				if (args->eh) /* not on layer 2 */
2209 				    break;
2210 				/* otherwise this is terminal */
2211 				l = 0;		/* exit inner loop */
2212 				done = 1;	/* exit outer loop */
2213 				retval = (cmd->opcode == O_DIVERT) ?
2214 					IP_FW_DIVERT : IP_FW_TEE;
2215 				set_match(args, f_pos, chain);
2216 				args->rule.info = TARG(cmd->arg1, divert);
2217 				break;
2218 
2219 			case O_COUNT:
2220 				IPFW_INC_RULE_COUNTER(f, pktlen);
2221 				l = 0;		/* exit inner loop */
2222 				break;
2223 
2224 			case O_SKIPTO:
2225 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2226 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2227 			    /*
2228 			     * Skip disabled rules, and re-enter
2229 			     * the inner loop with the correct
2230 			     * f_pos, f, l and cmd.
2231 			     * Also clear cmdlen and skip_or
2232 			     */
2233 			    for (; f_pos < chain->n_rules - 1 &&
2234 				    (V_set_disable &
2235 				     (1 << chain->map[f_pos]->set));
2236 				    f_pos++)
2237 				;
2238 			    /* Re-enter the inner loop at the skipto rule. */
2239 			    f = chain->map[f_pos];
2240 			    l = f->cmd_len;
2241 			    cmd = f->cmd;
2242 			    match = 1;
2243 			    cmdlen = 0;
2244 			    skip_or = 0;
2245 			    continue;
2246 			    break;	/* not reached */
2247 
2248 			case O_CALLRETURN: {
2249 				/*
2250 				 * Implementation of `subroutine' call/return,
2251 				 * in the stack carried in an mbuf tag. This
2252 				 * is different from `skipto' in that any call
2253 				 * address is possible (`skipto' must prevent
2254 				 * backward jumps to avoid endless loops).
2255 				 * We have `return' action when F_NOT flag is
2256 				 * present. The `m_tag_id' field is used as
2257 				 * stack pointer.
2258 				 */
2259 				struct m_tag *mtag;
2260 				uint16_t jmpto, *stack;
2261 
2262 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2263 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2264 				/*
2265 				 * Hand-rolled version of m_tag_locate() with
2266 				 * wildcard `type'.
2267 				 * If not already tagged, allocate new tag.
2268 				 */
2269 				mtag = m_tag_first(m);
2270 				while (mtag != NULL) {
2271 					if (mtag->m_tag_cookie ==
2272 					    MTAG_IPFW_CALL)
2273 						break;
2274 					mtag = m_tag_next(m, mtag);
2275 				}
2276 				if (mtag == NULL && IS_CALL) {
2277 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2278 					    IPFW_CALLSTACK_SIZE *
2279 					    sizeof(uint16_t), M_NOWAIT);
2280 					if (mtag != NULL)
2281 						m_tag_prepend(m, mtag);
2282 				}
2283 
2284 				/*
2285 				 * On error both `call' and `return' just
2286 				 * continue with next rule.
2287 				 */
2288 				if (IS_RETURN && (mtag == NULL ||
2289 				    mtag->m_tag_id == 0)) {
2290 					l = 0;		/* exit inner loop */
2291 					break;
2292 				}
2293 				if (IS_CALL && (mtag == NULL ||
2294 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2295 					printf("ipfw: call stack error, "
2296 					    "go to next rule\n");
2297 					l = 0;		/* exit inner loop */
2298 					break;
2299 				}
2300 
2301 				IPFW_INC_RULE_COUNTER(f, pktlen);
2302 				stack = (uint16_t *)(mtag + 1);
2303 
2304 				/*
2305 				 * The `call' action may use cached f_pos
2306 				 * (in f->next_rule), whose version is written
2307 				 * in f->next_rule.
2308 				 * The `return' action, however, doesn't have
2309 				 * fixed jump address in cmd->arg1 and can't use
2310 				 * cache.
2311 				 */
2312 				if (IS_CALL) {
2313 					stack[mtag->m_tag_id] = f->rulenum;
2314 					mtag->m_tag_id++;
2315 			    		f_pos = JUMP(chain, f, cmd->arg1,
2316 					    tablearg, 1);
2317 				} else {	/* `return' action */
2318 					mtag->m_tag_id--;
2319 					jmpto = stack[mtag->m_tag_id] + 1;
2320 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2321 				}
2322 
2323 				/*
2324 				 * Skip disabled rules, and re-enter
2325 				 * the inner loop with the correct
2326 				 * f_pos, f, l and cmd.
2327 				 * Also clear cmdlen and skip_or
2328 				 */
2329 				for (; f_pos < chain->n_rules - 1 &&
2330 				    (V_set_disable &
2331 				    (1 << chain->map[f_pos]->set)); f_pos++)
2332 					;
2333 				/* Re-enter the inner loop at the dest rule. */
2334 				f = chain->map[f_pos];
2335 				l = f->cmd_len;
2336 				cmd = f->cmd;
2337 				cmdlen = 0;
2338 				skip_or = 0;
2339 				continue;
2340 				break;	/* NOTREACHED */
2341 			}
2342 #undef IS_CALL
2343 #undef IS_RETURN
2344 
2345 			case O_REJECT:
2346 				/*
2347 				 * Drop the packet and send a reject notice
2348 				 * if the packet is not ICMP (or is an ICMP
2349 				 * query), and it is not multicast/broadcast.
2350 				 */
2351 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2352 				    (proto != IPPROTO_ICMP ||
2353 				     is_icmp_query(ICMP(ulp))) &&
2354 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2355 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2356 					send_reject(args, cmd->arg1, iplen, ip);
2357 					m = args->m;
2358 				}
2359 				/* FALLTHROUGH */
2360 #ifdef INET6
2361 			case O_UNREACH6:
2362 				if (hlen > 0 && is_ipv6 &&
2363 				    ((offset & IP6F_OFF_MASK) == 0) &&
2364 				    (proto != IPPROTO_ICMPV6 ||
2365 				     (is_icmp6_query(icmp6_type) == 1)) &&
2366 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2367 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2368 					send_reject6(
2369 					    args, cmd->arg1, hlen,
2370 					    (struct ip6_hdr *)ip);
2371 					m = args->m;
2372 				}
2373 				/* FALLTHROUGH */
2374 #endif
2375 			case O_DENY:
2376 				retval = IP_FW_DENY;
2377 				l = 0;		/* exit inner loop */
2378 				done = 1;	/* exit outer loop */
2379 				break;
2380 
2381 			case O_FORWARD_IP:
2382 				if (args->eh)	/* not valid on layer2 pkts */
2383 					break;
2384 				if (q == NULL || q->rule != f ||
2385 				    dyn_dir == MATCH_FORWARD) {
2386 				    struct sockaddr_in *sa;
2387 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2388 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2389 					bcopy(sa, &args->hopstore,
2390 							sizeof(*sa));
2391 					args->hopstore.sin_addr.s_addr =
2392 						    htonl(tablearg);
2393 					args->next_hop = &args->hopstore;
2394 				    } else {
2395 					args->next_hop = sa;
2396 				    }
2397 				}
2398 				retval = IP_FW_PASS;
2399 				l = 0;          /* exit inner loop */
2400 				done = 1;       /* exit outer loop */
2401 				break;
2402 
2403 #ifdef INET6
2404 			case O_FORWARD_IP6:
2405 				if (args->eh)	/* not valid on layer2 pkts */
2406 					break;
2407 				if (q == NULL || q->rule != f ||
2408 				    dyn_dir == MATCH_FORWARD) {
2409 					struct sockaddr_in6 *sin6;
2410 
2411 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2412 					args->next_hop6 = sin6;
2413 				}
2414 				retval = IP_FW_PASS;
2415 				l = 0;		/* exit inner loop */
2416 				done = 1;	/* exit outer loop */
2417 				break;
2418 #endif
2419 
2420 			case O_NETGRAPH:
2421 			case O_NGTEE:
2422 				set_match(args, f_pos, chain);
2423 				args->rule.info = TARG(cmd->arg1, netgraph);
2424 				if (V_fw_one_pass)
2425 					args->rule.info |= IPFW_ONEPASS;
2426 				retval = (cmd->opcode == O_NETGRAPH) ?
2427 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2428 				l = 0;          /* exit inner loop */
2429 				done = 1;       /* exit outer loop */
2430 				break;
2431 
2432 			case O_SETFIB: {
2433 				uint32_t fib;
2434 
2435 				IPFW_INC_RULE_COUNTER(f, pktlen);
2436 				fib = TARG(cmd->arg1, fib) & 0x7FFFF;
2437 				if (fib >= rt_numfibs)
2438 					fib = 0;
2439 				M_SETFIB(m, fib);
2440 				args->f_id.fib = fib;
2441 				l = 0;		/* exit inner loop */
2442 				break;
2443 		        }
2444 
2445 			case O_SETDSCP: {
2446 				uint16_t code;
2447 
2448 				code = TARG(cmd->arg1, dscp) & 0x3F;
2449 				l = 0;		/* exit inner loop */
2450 				if (is_ipv4) {
2451 					uint16_t a;
2452 
2453 					a = ip->ip_tos;
2454 					ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2455 					a += ntohs(ip->ip_sum) - ip->ip_tos;
2456 					ip->ip_sum = htons(a);
2457 				} else if (is_ipv6) {
2458 					uint8_t *v;
2459 
2460 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2461 					*v = (*v & 0xF0) | (code >> 2);
2462 					v++;
2463 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2464 				} else
2465 					break;
2466 
2467 				IPFW_INC_RULE_COUNTER(f, pktlen);
2468 				break;
2469 			}
2470 
2471 			case O_NAT:
2472 				l = 0;          /* exit inner loop */
2473 				done = 1;       /* exit outer loop */
2474  				if (!IPFW_NAT_LOADED) {
2475 				    retval = IP_FW_DENY;
2476 				    break;
2477 				}
2478 
2479 				struct cfg_nat *t;
2480 				int nat_id;
2481 
2482 				set_match(args, f_pos, chain);
2483 				/* Check if this is 'global' nat rule */
2484 				if (cmd->arg1 == 0) {
2485 					retval = ipfw_nat_ptr(args, NULL, m);
2486 					break;
2487 				}
2488 				t = ((ipfw_insn_nat *)cmd)->nat;
2489 				if (t == NULL) {
2490 					nat_id = TARG(cmd->arg1, nat);
2491 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2492 
2493 					if (t == NULL) {
2494 					    retval = IP_FW_DENY;
2495 					    break;
2496 					}
2497 					if (cmd->arg1 != IP_FW_TARG)
2498 					    ((ipfw_insn_nat *)cmd)->nat = t;
2499 				}
2500 				retval = ipfw_nat_ptr(args, t, m);
2501 				break;
2502 
2503 			case O_REASS: {
2504 				int ip_off;
2505 
2506 				IPFW_INC_RULE_COUNTER(f, pktlen);
2507 				l = 0;	/* in any case exit inner loop */
2508 				ip_off = ntohs(ip->ip_off);
2509 
2510 				/* if not fragmented, go to next rule */
2511 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2512 				    break;
2513 
2514 				args->m = m = ip_reass(m);
2515 
2516 				/*
2517 				 * do IP header checksum fixup.
2518 				 */
2519 				if (m == NULL) { /* fragment got swallowed */
2520 				    retval = IP_FW_DENY;
2521 				} else { /* good, packet complete */
2522 				    int hlen;
2523 
2524 				    ip = mtod(m, struct ip *);
2525 				    hlen = ip->ip_hl << 2;
2526 				    ip->ip_sum = 0;
2527 				    if (hlen == sizeof(struct ip))
2528 					ip->ip_sum = in_cksum_hdr(ip);
2529 				    else
2530 					ip->ip_sum = in_cksum(m, hlen);
2531 				    retval = IP_FW_REASS;
2532 				    set_match(args, f_pos, chain);
2533 				}
2534 				done = 1;	/* exit outer loop */
2535 				break;
2536 			}
2537 
2538 			default:
2539 				panic("-- unknown opcode %d\n", cmd->opcode);
2540 			} /* end of switch() on opcodes */
2541 			/*
2542 			 * if we get here with l=0, then match is irrelevant.
2543 			 */
2544 
2545 			if (cmd->len & F_NOT)
2546 				match = !match;
2547 
2548 			if (match) {
2549 				if (cmd->len & F_OR)
2550 					skip_or = 1;
2551 			} else {
2552 				if (!(cmd->len & F_OR)) /* not an OR block, */
2553 					break;		/* try next rule    */
2554 			}
2555 
2556 		}	/* end of inner loop, scan opcodes */
2557 #undef PULLUP_LEN
2558 
2559 		if (done)
2560 			break;
2561 
2562 /* next_rule:; */	/* try next rule		*/
2563 
2564 	}		/* end of outer for, scan rules */
2565 
2566 	if (done) {
2567 		struct ip_fw *rule = chain->map[f_pos];
2568 		/* Update statistics */
2569 		IPFW_INC_RULE_COUNTER(rule, pktlen);
2570 	} else {
2571 		retval = IP_FW_DENY;
2572 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2573 	}
2574 	IPFW_PF_RUNLOCK(chain);
2575 #ifdef __FreeBSD__
2576 	if (ucred_cache != NULL)
2577 		crfree(ucred_cache);
2578 #endif
2579 	return (retval);
2580 
2581 pullup_failed:
2582 	if (V_fw_verbose)
2583 		printf("ipfw: pullup failed\n");
2584 	return (IP_FW_DENY);
2585 }
2586 
2587 /*
2588  * Set maximum number of tables that can be used in given VNET ipfw instance.
2589  */
2590 #ifdef SYSCTL_NODE
2591 static int
2592 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2593 {
2594 	int error;
2595 	unsigned int ntables;
2596 
2597 	ntables = V_fw_tables_max;
2598 
2599 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2600 	/* Read operation or some error */
2601 	if ((error != 0) || (req->newptr == NULL))
2602 		return (error);
2603 
2604 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2605 }
2606 
2607 /*
2608  * Switches table namespace between global and per-set.
2609  */
2610 static int
2611 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2612 {
2613 	int error;
2614 	unsigned int sets;
2615 
2616 	sets = V_fw_tables_sets;
2617 
2618 	error = sysctl_handle_int(oidp, &sets, 0, req);
2619 	/* Read operation or some error */
2620 	if ((error != 0) || (req->newptr == NULL))
2621 		return (error);
2622 
2623 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2624 }
2625 #endif
2626 
2627 /*
2628  * Module and VNET glue
2629  */
2630 
2631 /*
2632  * Stuff that must be initialised only on boot or module load
2633  */
2634 static int
2635 ipfw_init(void)
2636 {
2637 	int error = 0;
2638 
2639 	/*
2640  	 * Only print out this stuff the first time around,
2641 	 * when called from the sysinit code.
2642 	 */
2643 	printf("ipfw2 "
2644 #ifdef INET6
2645 		"(+ipv6) "
2646 #endif
2647 		"initialized, divert %s, nat %s, "
2648 		"default to %s, logging ",
2649 #ifdef IPDIVERT
2650 		"enabled",
2651 #else
2652 		"loadable",
2653 #endif
2654 #ifdef IPFIREWALL_NAT
2655 		"enabled",
2656 #else
2657 		"loadable",
2658 #endif
2659 		default_to_accept ? "accept" : "deny");
2660 
2661 	/*
2662 	 * Note: V_xxx variables can be accessed here but the vnet specific
2663 	 * initializer may not have been called yet for the VIMAGE case.
2664 	 * Tuneables will have been processed. We will print out values for
2665 	 * the default vnet.
2666 	 * XXX This should all be rationalized AFTER 8.0
2667 	 */
2668 	if (V_fw_verbose == 0)
2669 		printf("disabled\n");
2670 	else if (V_verbose_limit == 0)
2671 		printf("unlimited\n");
2672 	else
2673 		printf("limited to %d packets/entry by default\n",
2674 		    V_verbose_limit);
2675 
2676 	/* Check user-supplied table count for validness */
2677 	if (default_fw_tables > IPFW_TABLES_MAX)
2678 	  default_fw_tables = IPFW_TABLES_MAX;
2679 
2680 	ipfw_init_sopt_handler();
2681 	ipfw_log_bpf(1); /* init */
2682 	ipfw_iface_init();
2683 	return (error);
2684 }
2685 
2686 /*
2687  * Called for the removal of the last instance only on module unload.
2688  */
2689 static void
2690 ipfw_destroy(void)
2691 {
2692 
2693 	ipfw_iface_destroy();
2694 	ipfw_log_bpf(0); /* uninit */
2695 	ipfw_destroy_sopt_handler();
2696 	printf("IP firewall unloaded\n");
2697 }
2698 
2699 /*
2700  * Stuff that must be initialized for every instance
2701  * (including the first of course).
2702  */
2703 static int
2704 vnet_ipfw_init(const void *unused)
2705 {
2706 	int error, first;
2707 	struct ip_fw *rule = NULL;
2708 	struct ip_fw_chain *chain;
2709 
2710 	chain = &V_layer3_chain;
2711 
2712 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2713 
2714 	/* First set up some values that are compile time options */
2715 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2716 	V_fw_deny_unknown_exthdrs = 1;
2717 #ifdef IPFIREWALL_VERBOSE
2718 	V_fw_verbose = 1;
2719 #endif
2720 #ifdef IPFIREWALL_VERBOSE_LIMIT
2721 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2722 #endif
2723 #ifdef IPFIREWALL_NAT
2724 	LIST_INIT(&chain->nat);
2725 #endif
2726 
2727 	ipfw_init_counters();
2728 	/* insert the default rule and create the initial map */
2729 	chain->n_rules = 1;
2730 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2731 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2732 
2733 	/* Set initial number of tables */
2734 	V_fw_tables_max = default_fw_tables;
2735 	error = ipfw_init_tables(chain, first);
2736 	if (error) {
2737 		printf("ipfw2: setting up tables failed\n");
2738 		free(chain->map, M_IPFW);
2739 		free(rule, M_IPFW);
2740 		return (ENOSPC);
2741 	}
2742 
2743 	/* fill and insert the default rule */
2744 	rule->act_ofs = 0;
2745 	rule->rulenum = IPFW_DEFAULT_RULE;
2746 	rule->cmd_len = 1;
2747 	rule->set = RESVD_SET;
2748 	rule->cmd[0].len = 1;
2749 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2750 	chain->default_rule = chain->map[0] = rule;
2751 	chain->id = rule->id = 1;
2752 	/* Pre-calculate rules length for legacy dump format */
2753 	chain->static_len = sizeof(struct ip_fw_rule0);
2754 
2755 	IPFW_LOCK_INIT(chain);
2756 	ipfw_dyn_init(chain);
2757 #ifdef LINEAR_SKIPTO
2758 	ipfw_init_skipto_cache(chain);
2759 #endif
2760 
2761 	/* First set up some values that are compile time options */
2762 	V_ipfw_vnet_ready = 1;		/* Open for business */
2763 
2764 	/*
2765 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2766 	 * Even if the latter two fail we still keep the module alive
2767 	 * because the sockopt and layer2 paths are still useful.
2768 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2769 	 * so we can ignore the exact return value and just set a flag.
2770 	 *
2771 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2772 	 * changes in the underlying (per-vnet) variables trigger
2773 	 * immediate hook()/unhook() calls.
2774 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2775 	 * is checked on each packet because there are no pfil hooks.
2776 	 */
2777 	V_ip_fw_ctl_ptr = ipfw_ctl3;
2778 	error = ipfw_attach_hooks(1);
2779 	return (error);
2780 }
2781 
2782 /*
2783  * Called for the removal of each instance.
2784  */
2785 static int
2786 vnet_ipfw_uninit(const void *unused)
2787 {
2788 	struct ip_fw *reap;
2789 	struct ip_fw_chain *chain = &V_layer3_chain;
2790 	int i, last;
2791 
2792 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2793 	/*
2794 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2795 	 * Then grab, release and grab again the WLOCK so we make
2796 	 * sure the update is propagated and nobody will be in.
2797 	 */
2798 	(void)ipfw_attach_hooks(0 /* detach */);
2799 	V_ip_fw_ctl_ptr = NULL;
2800 
2801 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2802 
2803 	IPFW_UH_WLOCK(chain);
2804 	IPFW_UH_WUNLOCK(chain);
2805 	IPFW_UH_WLOCK(chain);
2806 
2807 	IPFW_WLOCK(chain);
2808 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2809 	IPFW_WUNLOCK(chain);
2810 
2811 	reap = NULL;
2812 	IPFW_WLOCK(chain);
2813 	for (i = 0; i < chain->n_rules; i++)
2814 		ipfw_reap_add(chain, &reap, chain->map[i]);
2815 	free(chain->map, M_IPFW);
2816 #ifdef LINEAR_SKIPTO
2817 	ipfw_destroy_skipto_cache(chain);
2818 #endif
2819 	IPFW_WUNLOCK(chain);
2820 	IPFW_UH_WUNLOCK(chain);
2821 	ipfw_destroy_tables(chain, last);
2822 	if (reap != NULL)
2823 		ipfw_reap_rules(reap);
2824 	vnet_ipfw_iface_destroy(chain);
2825 	IPFW_LOCK_DESTROY(chain);
2826 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2827 	ipfw_destroy_counters();
2828 	return (0);
2829 }
2830 
2831 /*
2832  * Module event handler.
2833  * In general we have the choice of handling most of these events by the
2834  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2835  * use the SYSINIT handlers as they are more capable of expressing the
2836  * flow of control during module and vnet operations, so this is just
2837  * a skeleton. Note there is no SYSINIT equivalent of the module
2838  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2839  */
2840 static int
2841 ipfw_modevent(module_t mod, int type, void *unused)
2842 {
2843 	int err = 0;
2844 
2845 	switch (type) {
2846 	case MOD_LOAD:
2847 		/* Called once at module load or
2848 	 	 * system boot if compiled in. */
2849 		break;
2850 	case MOD_QUIESCE:
2851 		/* Called before unload. May veto unloading. */
2852 		break;
2853 	case MOD_UNLOAD:
2854 		/* Called during unload. */
2855 		break;
2856 	case MOD_SHUTDOWN:
2857 		/* Called during system shutdown. */
2858 		break;
2859 	default:
2860 		err = EOPNOTSUPP;
2861 		break;
2862 	}
2863 	return err;
2864 }
2865 
2866 static moduledata_t ipfwmod = {
2867 	"ipfw",
2868 	ipfw_modevent,
2869 	0
2870 };
2871 
2872 /* Define startup order. */
2873 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2874 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2875 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2876 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2877 
2878 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2879 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2880 MODULE_VERSION(ipfw, 3);
2881 /* should declare some dependencies here */
2882 
2883 /*
2884  * Starting up. Done in order after ipfwmod() has been called.
2885  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2886  */
2887 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2888 	    ipfw_init, NULL);
2889 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2890 	    vnet_ipfw_init, NULL);
2891 
2892 /*
2893  * Closing up shop. These are done in REVERSE ORDER, but still
2894  * after ipfwmod() has been called. Not called on reboot.
2895  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2896  * or when the module is unloaded.
2897  */
2898 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2899 	    ipfw_destroy, NULL);
2900 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2901 	    vnet_ipfw_uninit, NULL);
2902 /* end of file */
2903