xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 9ecd54f24fe9fa373e07c9fd7c052deb2188f545)
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/eventhandler.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/jail.h>
51 #include <sys/module.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/rwlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/syslog.h>
59 #include <sys/ucred.h>
60 #include <net/ethernet.h> /* for ETHERTYPE_IP */
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/route.h>
64 #include <net/pfil.h>
65 #include <net/vnet.h>
66 
67 #include <netpfil/pf/pf_mtag.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_icmp.h>
75 #include <netinet/ip_fw.h>
76 #include <netinet/ip_carp.h>
77 #include <netinet/pim.h>
78 #include <netinet/tcp_var.h>
79 #include <netinet/udp.h>
80 #include <netinet/udp_var.h>
81 #include <netinet/sctp.h>
82 
83 #include <netinet/ip6.h>
84 #include <netinet/icmp6.h>
85 #ifdef INET6
86 #include <netinet6/in6_pcb.h>
87 #include <netinet6/scope6_var.h>
88 #include <netinet6/ip6_var.h>
89 #endif
90 
91 #include <netpfil/ipfw/ip_fw_private.h>
92 
93 #include <machine/in_cksum.h>	/* XXX for in_cksum */
94 
95 #ifdef MAC
96 #include <security/mac/mac_framework.h>
97 #endif
98 
99 /*
100  * static variables followed by global ones.
101  * All ipfw global variables are here.
102  */
103 
104 /* ipfw_vnet_ready controls when we are open for business */
105 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
106 #define	V_ipfw_vnet_ready	VNET(ipfw_vnet_ready)
107 
108 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
109 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
110 
111 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
112 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
113 
114 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
115 static int default_to_accept = 1;
116 #else
117 static int default_to_accept;
118 #endif
119 
120 VNET_DEFINE(int, autoinc_step);
121 VNET_DEFINE(int, fw_one_pass) = 1;
122 
123 VNET_DEFINE(unsigned int, fw_tables_max);
124 /* Use 128 tables by default */
125 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
126 
127 /*
128  * Each rule belongs to one of 32 different sets (0..31).
129  * The variable set_disable contains one bit per set.
130  * If the bit is set, all rules in the corresponding set
131  * are disabled. Set RESVD_SET(31) is reserved for the default rule
132  * and rules that are not deleted by the flush command,
133  * and CANNOT be disabled.
134  * Rules in set RESVD_SET can only be deleted individually.
135  */
136 VNET_DEFINE(u_int32_t, set_disable);
137 #define	V_set_disable			VNET(set_disable)
138 
139 VNET_DEFINE(int, fw_verbose);
140 /* counter for ipfw_log(NULL...) */
141 VNET_DEFINE(u_int64_t, norule_counter);
142 VNET_DEFINE(int, verbose_limit);
143 
144 /* layer3_chain contains the list of rules for layer 3 */
145 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
146 
147 VNET_DEFINE(int, ipfw_nat_ready) = 0;
148 
149 ipfw_nat_t *ipfw_nat_ptr = NULL;
150 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
151 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
152 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
153 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
154 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
155 
156 #ifdef SYSCTL_NODE
157 uint32_t dummy_def = IPFW_DEFAULT_RULE;
158 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
159 
160 SYSBEGIN(f3)
161 
162 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
163 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
164     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
165     "Only do a single pass through ipfw when using dummynet(4)");
166 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
167     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
168     "Rule number auto-increment step");
169 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
170     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
171     "Log matches to ipfw rules");
172 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
173     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
174     "Set upper limit of matches of ipfw rules logged");
175 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
176     &dummy_def, 0,
177     "The default/max possible rule number.");
178 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
179     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
180     "Maximum number of tables");
181 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
182     &default_to_accept, 0,
183     "Make the default rule accept all packets.");
184 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
185 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
186 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
187     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
188     "Number of static rules");
189 
190 #ifdef INET6
191 SYSCTL_DECL(_net_inet6_ip6);
192 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
193 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
194     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
195     "Deny packets with unknown IPv6 Extension Headers");
196 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
197     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
198     "Permit single packet IPv6 fragments");
199 #endif /* INET6 */
200 
201 SYSEND
202 
203 #endif /* SYSCTL_NODE */
204 
205 
206 /*
207  * Some macros used in the various matching options.
208  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
209  * Other macros just cast void * into the appropriate type
210  */
211 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
212 #define	TCP(p)		((struct tcphdr *)(p))
213 #define	SCTP(p)		((struct sctphdr *)(p))
214 #define	UDP(p)		((struct udphdr *)(p))
215 #define	ICMP(p)		((struct icmphdr *)(p))
216 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
217 
218 static __inline int
219 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
220 {
221 	int type = icmp->icmp_type;
222 
223 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
224 }
225 
226 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
227     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
228 
229 static int
230 is_icmp_query(struct icmphdr *icmp)
231 {
232 	int type = icmp->icmp_type;
233 
234 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
235 }
236 #undef TT
237 
238 /*
239  * The following checks use two arrays of 8 or 16 bits to store the
240  * bits that we want set or clear, respectively. They are in the
241  * low and high half of cmd->arg1 or cmd->d[0].
242  *
243  * We scan options and store the bits we find set. We succeed if
244  *
245  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
246  *
247  * The code is sometimes optimized not to store additional variables.
248  */
249 
250 static int
251 flags_match(ipfw_insn *cmd, u_int8_t bits)
252 {
253 	u_char want_clear;
254 	bits = ~bits;
255 
256 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
257 		return 0; /* some bits we want set were clear */
258 	want_clear = (cmd->arg1 >> 8) & 0xff;
259 	if ( (want_clear & bits) != want_clear)
260 		return 0; /* some bits we want clear were set */
261 	return 1;
262 }
263 
264 static int
265 ipopts_match(struct ip *ip, ipfw_insn *cmd)
266 {
267 	int optlen, bits = 0;
268 	u_char *cp = (u_char *)(ip + 1);
269 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
270 
271 	for (; x > 0; x -= optlen, cp += optlen) {
272 		int opt = cp[IPOPT_OPTVAL];
273 
274 		if (opt == IPOPT_EOL)
275 			break;
276 		if (opt == IPOPT_NOP)
277 			optlen = 1;
278 		else {
279 			optlen = cp[IPOPT_OLEN];
280 			if (optlen <= 0 || optlen > x)
281 				return 0; /* invalid or truncated */
282 		}
283 		switch (opt) {
284 
285 		default:
286 			break;
287 
288 		case IPOPT_LSRR:
289 			bits |= IP_FW_IPOPT_LSRR;
290 			break;
291 
292 		case IPOPT_SSRR:
293 			bits |= IP_FW_IPOPT_SSRR;
294 			break;
295 
296 		case IPOPT_RR:
297 			bits |= IP_FW_IPOPT_RR;
298 			break;
299 
300 		case IPOPT_TS:
301 			bits |= IP_FW_IPOPT_TS;
302 			break;
303 		}
304 	}
305 	return (flags_match(cmd, bits));
306 }
307 
308 static int
309 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
310 {
311 	int optlen, bits = 0;
312 	u_char *cp = (u_char *)(tcp + 1);
313 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
314 
315 	for (; x > 0; x -= optlen, cp += optlen) {
316 		int opt = cp[0];
317 		if (opt == TCPOPT_EOL)
318 			break;
319 		if (opt == TCPOPT_NOP)
320 			optlen = 1;
321 		else {
322 			optlen = cp[1];
323 			if (optlen <= 0)
324 				break;
325 		}
326 
327 		switch (opt) {
328 
329 		default:
330 			break;
331 
332 		case TCPOPT_MAXSEG:
333 			bits |= IP_FW_TCPOPT_MSS;
334 			break;
335 
336 		case TCPOPT_WINDOW:
337 			bits |= IP_FW_TCPOPT_WINDOW;
338 			break;
339 
340 		case TCPOPT_SACK_PERMITTED:
341 		case TCPOPT_SACK:
342 			bits |= IP_FW_TCPOPT_SACK;
343 			break;
344 
345 		case TCPOPT_TIMESTAMP:
346 			bits |= IP_FW_TCPOPT_TS;
347 			break;
348 
349 		}
350 	}
351 	return (flags_match(cmd, bits));
352 }
353 
354 static int
355 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
356 {
357 	if (ifp == NULL)	/* no iface with this packet, match fails */
358 		return 0;
359 	/* Check by name or by IP address */
360 	if (cmd->name[0] != '\0') { /* match by name */
361 		if (cmd->name[0] == '\1') /* use tablearg to match */
362 			return ipfw_lookup_table_extended(chain, cmd->p.glob,
363 				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
364 		/* Check name */
365 		if (cmd->p.glob) {
366 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
367 				return(1);
368 		} else {
369 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
370 				return(1);
371 		}
372 	} else {
373 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
374 		struct ifaddr *ia;
375 
376 		if_addr_rlock(ifp);
377 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
378 			if (ia->ifa_addr->sa_family != AF_INET)
379 				continue;
380 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
381 			    (ia->ifa_addr))->sin_addr.s_addr) {
382 				if_addr_runlock(ifp);
383 				return(1);	/* match */
384 			}
385 		}
386 		if_addr_runlock(ifp);
387 #endif /* __FreeBSD__ */
388 	}
389 	return(0);	/* no match, fail ... */
390 }
391 
392 /*
393  * The verify_path function checks if a route to the src exists and
394  * if it is reachable via ifp (when provided).
395  *
396  * The 'verrevpath' option checks that the interface that an IP packet
397  * arrives on is the same interface that traffic destined for the
398  * packet's source address would be routed out of.
399  * The 'versrcreach' option just checks that the source address is
400  * reachable via any route (except default) in the routing table.
401  * These two are a measure to block forged packets. This is also
402  * commonly known as "anti-spoofing" or Unicast Reverse Path
403  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
404  * is purposely reminiscent of the Cisco IOS command,
405  *
406  *   ip verify unicast reverse-path
407  *   ip verify unicast source reachable-via any
408  *
409  * which implements the same functionality. But note that the syntax
410  * is misleading, and the check may be performed on all IP packets
411  * whether unicast, multicast, or broadcast.
412  */
413 static int
414 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
415 {
416 #if defined(USERSPACE) || !defined(__FreeBSD__)
417 	return 0;
418 #else
419 	struct route ro;
420 	struct sockaddr_in *dst;
421 
422 	bzero(&ro, sizeof(ro));
423 
424 	dst = (struct sockaddr_in *)&(ro.ro_dst);
425 	dst->sin_family = AF_INET;
426 	dst->sin_len = sizeof(*dst);
427 	dst->sin_addr = src;
428 	in_rtalloc_ign(&ro, 0, fib);
429 
430 	if (ro.ro_rt == NULL)
431 		return 0;
432 
433 	/*
434 	 * If ifp is provided, check for equality with rtentry.
435 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
436 	 * in order to pass packets injected back by if_simloop():
437 	 * routing entry (via lo0) for our own address
438 	 * may exist, so we need to handle routing assymetry.
439 	 */
440 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
441 		RTFREE(ro.ro_rt);
442 		return 0;
443 	}
444 
445 	/* if no ifp provided, check if rtentry is not default route */
446 	if (ifp == NULL &&
447 	     satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
448 		RTFREE(ro.ro_rt);
449 		return 0;
450 	}
451 
452 	/* or if this is a blackhole/reject route */
453 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
454 		RTFREE(ro.ro_rt);
455 		return 0;
456 	}
457 
458 	/* found valid route */
459 	RTFREE(ro.ro_rt);
460 	return 1;
461 #endif /* __FreeBSD__ */
462 }
463 
464 #ifdef INET6
465 /*
466  * ipv6 specific rules here...
467  */
468 static __inline int
469 icmp6type_match (int type, ipfw_insn_u32 *cmd)
470 {
471 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
472 }
473 
474 static int
475 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
476 {
477 	int i;
478 	for (i=0; i <= cmd->o.arg1; ++i )
479 		if (curr_flow == cmd->d[i] )
480 			return 1;
481 	return 0;
482 }
483 
484 /* support for IP6_*_ME opcodes */
485 static int
486 search_ip6_addr_net (struct in6_addr * ip6_addr)
487 {
488 	struct ifnet *mdc;
489 	struct ifaddr *mdc2;
490 	struct in6_ifaddr *fdm;
491 	struct in6_addr copia;
492 
493 	TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
494 		if_addr_rlock(mdc);
495 		TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
496 			if (mdc2->ifa_addr->sa_family == AF_INET6) {
497 				fdm = (struct in6_ifaddr *)mdc2;
498 				copia = fdm->ia_addr.sin6_addr;
499 				/* need for leaving scope_id in the sock_addr */
500 				in6_clearscope(&copia);
501 				if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
502 					if_addr_runlock(mdc);
503 					return 1;
504 				}
505 			}
506 		}
507 		if_addr_runlock(mdc);
508 	}
509 	return 0;
510 }
511 
512 static int
513 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
514 {
515 	struct route_in6 ro;
516 	struct sockaddr_in6 *dst;
517 
518 	bzero(&ro, sizeof(ro));
519 
520 	dst = (struct sockaddr_in6 * )&(ro.ro_dst);
521 	dst->sin6_family = AF_INET6;
522 	dst->sin6_len = sizeof(*dst);
523 	dst->sin6_addr = *src;
524 
525 	in6_rtalloc_ign(&ro, 0, fib);
526 	if (ro.ro_rt == NULL)
527 		return 0;
528 
529 	/*
530 	 * if ifp is provided, check for equality with rtentry
531 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
532 	 * to support the case of sending packets to an address of our own.
533 	 * (where the former interface is the first argument of if_simloop()
534 	 *  (=ifp), the latter is lo0)
535 	 */
536 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
537 		RTFREE(ro.ro_rt);
538 		return 0;
539 	}
540 
541 	/* if no ifp provided, check if rtentry is not default route */
542 	if (ifp == NULL &&
543 	    IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
544 		RTFREE(ro.ro_rt);
545 		return 0;
546 	}
547 
548 	/* or if this is a blackhole/reject route */
549 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
550 		RTFREE(ro.ro_rt);
551 		return 0;
552 	}
553 
554 	/* found valid route */
555 	RTFREE(ro.ro_rt);
556 	return 1;
557 
558 }
559 
560 static int
561 is_icmp6_query(int icmp6_type)
562 {
563 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
564 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
565 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
566 	    icmp6_type == ICMP6_WRUREQUEST ||
567 	    icmp6_type == ICMP6_FQDN_QUERY ||
568 	    icmp6_type == ICMP6_NI_QUERY))
569 		return (1);
570 
571 	return (0);
572 }
573 
574 static void
575 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
576 {
577 	struct mbuf *m;
578 
579 	m = args->m;
580 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
581 		struct tcphdr *tcp;
582 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
583 
584 		if ((tcp->th_flags & TH_RST) == 0) {
585 			struct mbuf *m0;
586 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
587 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
588 			    tcp->th_flags | TH_RST);
589 			if (m0 != NULL)
590 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
591 				    NULL);
592 		}
593 		FREE_PKT(m);
594 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
595 #if 0
596 		/*
597 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
598 		 * as the contents are read. We need to m_adj() the
599 		 * needed amount.
600 		 * The mbuf will however be thrown away so we can adjust it.
601 		 * Remember we did an m_pullup on it already so we
602 		 * can make some assumptions about contiguousness.
603 		 */
604 		if (args->L3offset)
605 			m_adj(m, args->L3offset);
606 #endif
607 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
608 	} else
609 		FREE_PKT(m);
610 
611 	args->m = NULL;
612 }
613 
614 #endif /* INET6 */
615 
616 
617 /*
618  * sends a reject message, consuming the mbuf passed as an argument.
619  */
620 static void
621 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
622 {
623 
624 #if 0
625 	/* XXX When ip is not guaranteed to be at mtod() we will
626 	 * need to account for this */
627 	 * The mbuf will however be thrown away so we can adjust it.
628 	 * Remember we did an m_pullup on it already so we
629 	 * can make some assumptions about contiguousness.
630 	 */
631 	if (args->L3offset)
632 		m_adj(m, args->L3offset);
633 #endif
634 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
635 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
636 	} else if (args->f_id.proto == IPPROTO_TCP) {
637 		struct tcphdr *const tcp =
638 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
639 		if ( (tcp->th_flags & TH_RST) == 0) {
640 			struct mbuf *m;
641 			m = ipfw_send_pkt(args->m, &(args->f_id),
642 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
643 				tcp->th_flags | TH_RST);
644 			if (m != NULL)
645 				ip_output(m, NULL, NULL, 0, NULL, NULL);
646 		}
647 		FREE_PKT(args->m);
648 	} else
649 		FREE_PKT(args->m);
650 	args->m = NULL;
651 }
652 
653 /*
654  * Support for uid/gid/jail lookup. These tests are expensive
655  * (because we may need to look into the list of active sockets)
656  * so we cache the results. ugid_lookupp is 0 if we have not
657  * yet done a lookup, 1 if we succeeded, and -1 if we tried
658  * and failed. The function always returns the match value.
659  * We could actually spare the variable and use *uc, setting
660  * it to '(void *)check_uidgid if we have no info, NULL if
661  * we tried and failed, or any other value if successful.
662  */
663 static int
664 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
665     struct ucred **uc)
666 {
667 #if defined(USERSPACE)
668 	return 0;	// not supported in userspace
669 #else
670 #ifndef __FreeBSD__
671 	/* XXX */
672 	return cred_check(insn, proto, oif,
673 	    dst_ip, dst_port, src_ip, src_port,
674 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
675 #else  /* FreeBSD */
676 	struct in_addr src_ip, dst_ip;
677 	struct inpcbinfo *pi;
678 	struct ipfw_flow_id *id;
679 	struct inpcb *pcb, *inp;
680 	struct ifnet *oif;
681 	int lookupflags;
682 	int match;
683 
684 	id = &args->f_id;
685 	inp = args->inp;
686 	oif = args->oif;
687 
688 	/*
689 	 * Check to see if the UDP or TCP stack supplied us with
690 	 * the PCB. If so, rather then holding a lock and looking
691 	 * up the PCB, we can use the one that was supplied.
692 	 */
693 	if (inp && *ugid_lookupp == 0) {
694 		INP_LOCK_ASSERT(inp);
695 		if (inp->inp_socket != NULL) {
696 			*uc = crhold(inp->inp_cred);
697 			*ugid_lookupp = 1;
698 		} else
699 			*ugid_lookupp = -1;
700 	}
701 	/*
702 	 * If we have already been here and the packet has no
703 	 * PCB entry associated with it, then we can safely
704 	 * assume that this is a no match.
705 	 */
706 	if (*ugid_lookupp == -1)
707 		return (0);
708 	if (id->proto == IPPROTO_TCP) {
709 		lookupflags = 0;
710 		pi = &V_tcbinfo;
711 	} else if (id->proto == IPPROTO_UDP) {
712 		lookupflags = INPLOOKUP_WILDCARD;
713 		pi = &V_udbinfo;
714 	} else
715 		return 0;
716 	lookupflags |= INPLOOKUP_RLOCKPCB;
717 	match = 0;
718 	if (*ugid_lookupp == 0) {
719 		if (id->addr_type == 6) {
720 #ifdef INET6
721 			if (oif == NULL)
722 				pcb = in6_pcblookup_mbuf(pi,
723 				    &id->src_ip6, htons(id->src_port),
724 				    &id->dst_ip6, htons(id->dst_port),
725 				    lookupflags, oif, args->m);
726 			else
727 				pcb = in6_pcblookup_mbuf(pi,
728 				    &id->dst_ip6, htons(id->dst_port),
729 				    &id->src_ip6, htons(id->src_port),
730 				    lookupflags, oif, args->m);
731 #else
732 			*ugid_lookupp = -1;
733 			return (0);
734 #endif
735 		} else {
736 			src_ip.s_addr = htonl(id->src_ip);
737 			dst_ip.s_addr = htonl(id->dst_ip);
738 			if (oif == NULL)
739 				pcb = in_pcblookup_mbuf(pi,
740 				    src_ip, htons(id->src_port),
741 				    dst_ip, htons(id->dst_port),
742 				    lookupflags, oif, args->m);
743 			else
744 				pcb = in_pcblookup_mbuf(pi,
745 				    dst_ip, htons(id->dst_port),
746 				    src_ip, htons(id->src_port),
747 				    lookupflags, oif, args->m);
748 		}
749 		if (pcb != NULL) {
750 			INP_RLOCK_ASSERT(pcb);
751 			*uc = crhold(pcb->inp_cred);
752 			*ugid_lookupp = 1;
753 			INP_RUNLOCK(pcb);
754 		}
755 		if (*ugid_lookupp == 0) {
756 			/*
757 			 * We tried and failed, set the variable to -1
758 			 * so we will not try again on this packet.
759 			 */
760 			*ugid_lookupp = -1;
761 			return (0);
762 		}
763 	}
764 	if (insn->o.opcode == O_UID)
765 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
766 	else if (insn->o.opcode == O_GID)
767 		match = groupmember((gid_t)insn->d[0], *uc);
768 	else if (insn->o.opcode == O_JAIL)
769 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
770 	return (match);
771 #endif /* __FreeBSD__ */
772 #endif /* not supported in userspace */
773 }
774 
775 /*
776  * Helper function to set args with info on the rule after the matching
777  * one. slot is precise, whereas we guess rule_id as they are
778  * assigned sequentially.
779  */
780 static inline void
781 set_match(struct ip_fw_args *args, int slot,
782 	struct ip_fw_chain *chain)
783 {
784 	args->rule.chain_id = chain->id;
785 	args->rule.slot = slot + 1; /* we use 0 as a marker */
786 	args->rule.rule_id = 1 + chain->map[slot]->id;
787 	args->rule.rulenum = chain->map[slot]->rulenum;
788 }
789 
790 /*
791  * Helper function to enable cached rule lookups using
792  * x_next and next_rule fields in ipfw rule.
793  */
794 static int
795 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
796     int tablearg, int jump_backwards)
797 {
798 	int f_pos;
799 
800 	/* If possible use cached f_pos (in f->next_rule),
801 	 * whose version is written in f->next_rule
802 	 * (horrible hacks to avoid changing the ABI).
803 	 */
804 	if (num != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id)
805 		f_pos = (uintptr_t)f->next_rule;
806 	else {
807 		int i = IP_FW_ARG_TABLEARG(num);
808 		/* make sure we do not jump backward */
809 		if (jump_backwards == 0 && i <= f->rulenum)
810 			i = f->rulenum + 1;
811 		f_pos = ipfw_find_rule(chain, i, 0);
812 		/* update the cache */
813 		if (num != IP_FW_TABLEARG) {
814 			f->next_rule = (void *)(uintptr_t)f_pos;
815 			f->x_next = (void *)(uintptr_t)chain->id;
816 		}
817 	}
818 
819 	return (f_pos);
820 }
821 
822 /*
823  * The main check routine for the firewall.
824  *
825  * All arguments are in args so we can modify them and return them
826  * back to the caller.
827  *
828  * Parameters:
829  *
830  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
831  *		Starts with the IP header.
832  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
833  *	args->L3offset	Number of bytes bypassed if we came from L2.
834  *			e.g. often sizeof(eh)  ** NOTYET **
835  *	args->oif	Outgoing interface, NULL if packet is incoming.
836  *		The incoming interface is in the mbuf. (in)
837  *	args->divert_rule (in/out)
838  *		Skip up to the first rule past this rule number;
839  *		upon return, non-zero port number for divert or tee.
840  *
841  *	args->rule	Pointer to the last matching rule (in/out)
842  *	args->next_hop	Socket we are forwarding to (out).
843  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
844  *	args->f_id	Addresses grabbed from the packet (out)
845  * 	args->rule.info	a cookie depending on rule action
846  *
847  * Return value:
848  *
849  *	IP_FW_PASS	the packet must be accepted
850  *	IP_FW_DENY	the packet must be dropped
851  *	IP_FW_DIVERT	divert packet, port in m_tag
852  *	IP_FW_TEE	tee packet, port in m_tag
853  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
854  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
855  *		args->rule contains the matching rule,
856  *		args->rule.info has additional information.
857  *
858  */
859 int
860 ipfw_chk(struct ip_fw_args *args)
861 {
862 
863 	/*
864 	 * Local variables holding state while processing a packet:
865 	 *
866 	 * IMPORTANT NOTE: to speed up the processing of rules, there
867 	 * are some assumption on the values of the variables, which
868 	 * are documented here. Should you change them, please check
869 	 * the implementation of the various instructions to make sure
870 	 * that they still work.
871 	 *
872 	 * args->eh	The MAC header. It is non-null for a layer2
873 	 *	packet, it is NULL for a layer-3 packet.
874 	 * **notyet**
875 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
876 	 *
877 	 * m | args->m	Pointer to the mbuf, as received from the caller.
878 	 *	It may change if ipfw_chk() does an m_pullup, or if it
879 	 *	consumes the packet because it calls send_reject().
880 	 *	XXX This has to change, so that ipfw_chk() never modifies
881 	 *	or consumes the buffer.
882 	 * ip	is the beginning of the ip(4 or 6) header.
883 	 *	Calculated by adding the L3offset to the start of data.
884 	 *	(Until we start using L3offset, the packet is
885 	 *	supposed to start with the ip header).
886 	 */
887 	struct mbuf *m = args->m;
888 	struct ip *ip = mtod(m, struct ip *);
889 
890 	/*
891 	 * For rules which contain uid/gid or jail constraints, cache
892 	 * a copy of the users credentials after the pcb lookup has been
893 	 * executed. This will speed up the processing of rules with
894 	 * these types of constraints, as well as decrease contention
895 	 * on pcb related locks.
896 	 */
897 #ifndef __FreeBSD__
898 	struct bsd_ucred ucred_cache;
899 #else
900 	struct ucred *ucred_cache = NULL;
901 #endif
902 	int ucred_lookup = 0;
903 
904 	/*
905 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
906 	 *	inbound path (ether_input, ip_input).
907 	 *	If non-NULL, ipfw_chk has been called on the outbound path
908 	 *	(ether_output, ip_output).
909 	 */
910 	struct ifnet *oif = args->oif;
911 
912 	int f_pos = 0;		/* index of current rule in the array */
913 	int retval = 0;
914 
915 	/*
916 	 * hlen	The length of the IP header.
917 	 */
918 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
919 
920 	/*
921 	 * offset	The offset of a fragment. offset != 0 means that
922 	 *	we have a fragment at this offset of an IPv4 packet.
923 	 *	offset == 0 means that (if this is an IPv4 packet)
924 	 *	this is the first or only fragment.
925 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
926 	 *	or there is a single packet fragement (fragement header added
927 	 *	without needed).  We will treat a single packet fragment as if
928 	 *	there was no fragment header (or log/block depending on the
929 	 *	V_fw_permit_single_frag6 sysctl setting).
930 	 */
931 	u_short offset = 0;
932 	u_short ip6f_mf = 0;
933 
934 	/*
935 	 * Local copies of addresses. They are only valid if we have
936 	 * an IP packet.
937 	 *
938 	 * proto	The protocol. Set to 0 for non-ip packets,
939 	 *	or to the protocol read from the packet otherwise.
940 	 *	proto != 0 means that we have an IPv4 packet.
941 	 *
942 	 * src_port, dst_port	port numbers, in HOST format. Only
943 	 *	valid for TCP and UDP packets.
944 	 *
945 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
946 	 *	Only valid for IPv4 packets.
947 	 */
948 	uint8_t proto;
949 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
950 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
951 	uint16_t iplen=0;
952 	int pktlen;
953 	uint16_t	etype = 0;	/* Host order stored ether type */
954 
955 	/*
956 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
957 	 * 	MATCH_NONE when checked and not matched (q = NULL),
958 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
959 	 */
960 	int dyn_dir = MATCH_UNKNOWN;
961 	ipfw_dyn_rule *q = NULL;
962 	struct ip_fw_chain *chain = &V_layer3_chain;
963 
964 	/*
965 	 * We store in ulp a pointer to the upper layer protocol header.
966 	 * In the ipv4 case this is easy to determine from the header,
967 	 * but for ipv6 we might have some additional headers in the middle.
968 	 * ulp is NULL if not found.
969 	 */
970 	void *ulp = NULL;		/* upper layer protocol pointer. */
971 
972 	/* XXX ipv6 variables */
973 	int is_ipv6 = 0;
974 	uint8_t	icmp6_type = 0;
975 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
976 	/* end of ipv6 variables */
977 
978 	int is_ipv4 = 0;
979 
980 	int done = 0;		/* flag to exit the outer loop */
981 
982 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
983 		return (IP_FW_PASS);	/* accept */
984 
985 	dst_ip.s_addr = 0;		/* make sure it is initialized */
986 	src_ip.s_addr = 0;		/* make sure it is initialized */
987 	pktlen = m->m_pkthdr.len;
988 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
989 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
990 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
991 
992 /*
993  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
994  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
995  * pointer might become stale after other pullups (but we never use it
996  * this way).
997  */
998 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
999 #define PULLUP_LEN(_len, p, T)					\
1000 do {								\
1001 	int x = (_len) + T;					\
1002 	if ((m)->m_len < x) {					\
1003 		args->m = m = m_pullup(m, x);			\
1004 		if (m == NULL)					\
1005 			goto pullup_failed;			\
1006 	}							\
1007 	p = (mtod(m, char *) + (_len));				\
1008 } while (0)
1009 
1010 	/*
1011 	 * if we have an ether header,
1012 	 */
1013 	if (args->eh)
1014 		etype = ntohs(args->eh->ether_type);
1015 
1016 	/* Identify IP packets and fill up variables. */
1017 	if (pktlen >= sizeof(struct ip6_hdr) &&
1018 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1019 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1020 		is_ipv6 = 1;
1021 		args->f_id.addr_type = 6;
1022 		hlen = sizeof(struct ip6_hdr);
1023 		proto = ip6->ip6_nxt;
1024 
1025 		/* Search extension headers to find upper layer protocols */
1026 		while (ulp == NULL && offset == 0) {
1027 			switch (proto) {
1028 			case IPPROTO_ICMPV6:
1029 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1030 				icmp6_type = ICMP6(ulp)->icmp6_type;
1031 				break;
1032 
1033 			case IPPROTO_TCP:
1034 				PULLUP_TO(hlen, ulp, struct tcphdr);
1035 				dst_port = TCP(ulp)->th_dport;
1036 				src_port = TCP(ulp)->th_sport;
1037 				/* save flags for dynamic rules */
1038 				args->f_id._flags = TCP(ulp)->th_flags;
1039 				break;
1040 
1041 			case IPPROTO_SCTP:
1042 				PULLUP_TO(hlen, ulp, struct sctphdr);
1043 				src_port = SCTP(ulp)->src_port;
1044 				dst_port = SCTP(ulp)->dest_port;
1045 				break;
1046 
1047 			case IPPROTO_UDP:
1048 				PULLUP_TO(hlen, ulp, struct udphdr);
1049 				dst_port = UDP(ulp)->uh_dport;
1050 				src_port = UDP(ulp)->uh_sport;
1051 				break;
1052 
1053 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1054 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1055 				ext_hd |= EXT_HOPOPTS;
1056 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1057 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1058 				ulp = NULL;
1059 				break;
1060 
1061 			case IPPROTO_ROUTING:	/* RFC 2460 */
1062 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1063 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1064 				case 0:
1065 					ext_hd |= EXT_RTHDR0;
1066 					break;
1067 				case 2:
1068 					ext_hd |= EXT_RTHDR2;
1069 					break;
1070 				default:
1071 					if (V_fw_verbose)
1072 						printf("IPFW2: IPV6 - Unknown "
1073 						    "Routing Header type(%d)\n",
1074 						    ((struct ip6_rthdr *)
1075 						    ulp)->ip6r_type);
1076 					if (V_fw_deny_unknown_exthdrs)
1077 					    return (IP_FW_DENY);
1078 					break;
1079 				}
1080 				ext_hd |= EXT_ROUTING;
1081 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1082 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1083 				ulp = NULL;
1084 				break;
1085 
1086 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1087 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1088 				ext_hd |= EXT_FRAGMENT;
1089 				hlen += sizeof (struct ip6_frag);
1090 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1091 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1092 					IP6F_OFF_MASK;
1093 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1094 					IP6F_MORE_FRAG;
1095 				if (V_fw_permit_single_frag6 == 0 &&
1096 				    offset == 0 && ip6f_mf == 0) {
1097 					if (V_fw_verbose)
1098 						printf("IPFW2: IPV6 - Invalid "
1099 						    "Fragment Header\n");
1100 					if (V_fw_deny_unknown_exthdrs)
1101 					    return (IP_FW_DENY);
1102 					break;
1103 				}
1104 				args->f_id.extra =
1105 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1106 				ulp = NULL;
1107 				break;
1108 
1109 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1110 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1111 				ext_hd |= EXT_DSTOPTS;
1112 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1113 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1114 				ulp = NULL;
1115 				break;
1116 
1117 			case IPPROTO_AH:	/* RFC 2402 */
1118 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1119 				ext_hd |= EXT_AH;
1120 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1121 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1122 				ulp = NULL;
1123 				break;
1124 
1125 			case IPPROTO_ESP:	/* RFC 2406 */
1126 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1127 				/* Anything past Seq# is variable length and
1128 				 * data past this ext. header is encrypted. */
1129 				ext_hd |= EXT_ESP;
1130 				break;
1131 
1132 			case IPPROTO_NONE:	/* RFC 2460 */
1133 				/*
1134 				 * Packet ends here, and IPv6 header has
1135 				 * already been pulled up. If ip6e_len!=0
1136 				 * then octets must be ignored.
1137 				 */
1138 				ulp = ip; /* non-NULL to get out of loop. */
1139 				break;
1140 
1141 			case IPPROTO_OSPFIGP:
1142 				/* XXX OSPF header check? */
1143 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1144 				break;
1145 
1146 			case IPPROTO_PIM:
1147 				/* XXX PIM header check? */
1148 				PULLUP_TO(hlen, ulp, struct pim);
1149 				break;
1150 
1151 			case IPPROTO_CARP:
1152 				PULLUP_TO(hlen, ulp, struct carp_header);
1153 				if (((struct carp_header *)ulp)->carp_version !=
1154 				    CARP_VERSION)
1155 					return (IP_FW_DENY);
1156 				if (((struct carp_header *)ulp)->carp_type !=
1157 				    CARP_ADVERTISEMENT)
1158 					return (IP_FW_DENY);
1159 				break;
1160 
1161 			case IPPROTO_IPV6:	/* RFC 2893 */
1162 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1163 				break;
1164 
1165 			case IPPROTO_IPV4:	/* RFC 2893 */
1166 				PULLUP_TO(hlen, ulp, struct ip);
1167 				break;
1168 
1169 			default:
1170 				if (V_fw_verbose)
1171 					printf("IPFW2: IPV6 - Unknown "
1172 					    "Extension Header(%d), ext_hd=%x\n",
1173 					     proto, ext_hd);
1174 				if (V_fw_deny_unknown_exthdrs)
1175 				    return (IP_FW_DENY);
1176 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1177 				break;
1178 			} /*switch */
1179 		}
1180 		ip = mtod(m, struct ip *);
1181 		ip6 = (struct ip6_hdr *)ip;
1182 		args->f_id.src_ip6 = ip6->ip6_src;
1183 		args->f_id.dst_ip6 = ip6->ip6_dst;
1184 		args->f_id.src_ip = 0;
1185 		args->f_id.dst_ip = 0;
1186 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1187 	} else if (pktlen >= sizeof(struct ip) &&
1188 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1189 	    	is_ipv4 = 1;
1190 		hlen = ip->ip_hl << 2;
1191 		args->f_id.addr_type = 4;
1192 
1193 		/*
1194 		 * Collect parameters into local variables for faster matching.
1195 		 */
1196 		proto = ip->ip_p;
1197 		src_ip = ip->ip_src;
1198 		dst_ip = ip->ip_dst;
1199 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1200 		iplen = ntohs(ip->ip_len);
1201 		pktlen = iplen < pktlen ? iplen : pktlen;
1202 
1203 		if (offset == 0) {
1204 			switch (proto) {
1205 			case IPPROTO_TCP:
1206 				PULLUP_TO(hlen, ulp, struct tcphdr);
1207 				dst_port = TCP(ulp)->th_dport;
1208 				src_port = TCP(ulp)->th_sport;
1209 				/* save flags for dynamic rules */
1210 				args->f_id._flags = TCP(ulp)->th_flags;
1211 				break;
1212 
1213 			case IPPROTO_SCTP:
1214 				PULLUP_TO(hlen, ulp, struct sctphdr);
1215 				src_port = SCTP(ulp)->src_port;
1216 				dst_port = SCTP(ulp)->dest_port;
1217 				break;
1218 
1219 			case IPPROTO_UDP:
1220 				PULLUP_TO(hlen, ulp, struct udphdr);
1221 				dst_port = UDP(ulp)->uh_dport;
1222 				src_port = UDP(ulp)->uh_sport;
1223 				break;
1224 
1225 			case IPPROTO_ICMP:
1226 				PULLUP_TO(hlen, ulp, struct icmphdr);
1227 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1228 				break;
1229 
1230 			default:
1231 				break;
1232 			}
1233 		}
1234 
1235 		ip = mtod(m, struct ip *);
1236 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1237 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1238 	}
1239 #undef PULLUP_TO
1240 	if (proto) { /* we may have port numbers, store them */
1241 		args->f_id.proto = proto;
1242 		args->f_id.src_port = src_port = ntohs(src_port);
1243 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1244 	}
1245 
1246 	IPFW_PF_RLOCK(chain);
1247 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1248 		IPFW_PF_RUNLOCK(chain);
1249 		return (IP_FW_PASS);	/* accept */
1250 	}
1251 	if (args->rule.slot) {
1252 		/*
1253 		 * Packet has already been tagged as a result of a previous
1254 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1255 		 * REASS, NETGRAPH, DIVERT/TEE...)
1256 		 * Validate the slot and continue from the next one
1257 		 * if still present, otherwise do a lookup.
1258 		 */
1259 		f_pos = (args->rule.chain_id == chain->id) ?
1260 		    args->rule.slot :
1261 		    ipfw_find_rule(chain, args->rule.rulenum,
1262 			args->rule.rule_id);
1263 	} else {
1264 		f_pos = 0;
1265 	}
1266 
1267 	/*
1268 	 * Now scan the rules, and parse microinstructions for each rule.
1269 	 * We have two nested loops and an inner switch. Sometimes we
1270 	 * need to break out of one or both loops, or re-enter one of
1271 	 * the loops with updated variables. Loop variables are:
1272 	 *
1273 	 *	f_pos (outer loop) points to the current rule.
1274 	 *		On output it points to the matching rule.
1275 	 *	done (outer loop) is used as a flag to break the loop.
1276 	 *	l (inner loop)	residual length of current rule.
1277 	 *		cmd points to the current microinstruction.
1278 	 *
1279 	 * We break the inner loop by setting l=0 and possibly
1280 	 * cmdlen=0 if we don't want to advance cmd.
1281 	 * We break the outer loop by setting done=1
1282 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1283 	 * as needed.
1284 	 */
1285 	for (; f_pos < chain->n_rules; f_pos++) {
1286 		ipfw_insn *cmd;
1287 		uint32_t tablearg = 0;
1288 		int l, cmdlen, skip_or; /* skip rest of OR block */
1289 		struct ip_fw *f;
1290 
1291 		f = chain->map[f_pos];
1292 		if (V_set_disable & (1 << f->set) )
1293 			continue;
1294 
1295 		skip_or = 0;
1296 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1297 		    l -= cmdlen, cmd += cmdlen) {
1298 			int match;
1299 
1300 			/*
1301 			 * check_body is a jump target used when we find a
1302 			 * CHECK_STATE, and need to jump to the body of
1303 			 * the target rule.
1304 			 */
1305 
1306 /* check_body: */
1307 			cmdlen = F_LEN(cmd);
1308 			/*
1309 			 * An OR block (insn_1 || .. || insn_n) has the
1310 			 * F_OR bit set in all but the last instruction.
1311 			 * The first match will set "skip_or", and cause
1312 			 * the following instructions to be skipped until
1313 			 * past the one with the F_OR bit clear.
1314 			 */
1315 			if (skip_or) {		/* skip this instruction */
1316 				if ((cmd->len & F_OR) == 0)
1317 					skip_or = 0;	/* next one is good */
1318 				continue;
1319 			}
1320 			match = 0; /* set to 1 if we succeed */
1321 
1322 			switch (cmd->opcode) {
1323 			/*
1324 			 * The first set of opcodes compares the packet's
1325 			 * fields with some pattern, setting 'match' if a
1326 			 * match is found. At the end of the loop there is
1327 			 * logic to deal with F_NOT and F_OR flags associated
1328 			 * with the opcode.
1329 			 */
1330 			case O_NOP:
1331 				match = 1;
1332 				break;
1333 
1334 			case O_FORWARD_MAC:
1335 				printf("ipfw: opcode %d unimplemented\n",
1336 				    cmd->opcode);
1337 				break;
1338 
1339 			case O_GID:
1340 			case O_UID:
1341 			case O_JAIL:
1342 				/*
1343 				 * We only check offset == 0 && proto != 0,
1344 				 * as this ensures that we have a
1345 				 * packet with the ports info.
1346 				 */
1347 				if (offset != 0)
1348 					break;
1349 				if (proto == IPPROTO_TCP ||
1350 				    proto == IPPROTO_UDP)
1351 					match = check_uidgid(
1352 						    (ipfw_insn_u32 *)cmd,
1353 						    args, &ucred_lookup,
1354 #ifdef __FreeBSD__
1355 						    &ucred_cache);
1356 #else
1357 						    (void *)&ucred_cache);
1358 #endif
1359 				break;
1360 
1361 			case O_RECV:
1362 				match = iface_match(m->m_pkthdr.rcvif,
1363 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1364 				break;
1365 
1366 			case O_XMIT:
1367 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1368 				    chain, &tablearg);
1369 				break;
1370 
1371 			case O_VIA:
1372 				match = iface_match(oif ? oif :
1373 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1374 				    chain, &tablearg);
1375 				break;
1376 
1377 			case O_MACADDR2:
1378 				if (args->eh != NULL) {	/* have MAC header */
1379 					u_int32_t *want = (u_int32_t *)
1380 						((ipfw_insn_mac *)cmd)->addr;
1381 					u_int32_t *mask = (u_int32_t *)
1382 						((ipfw_insn_mac *)cmd)->mask;
1383 					u_int32_t *hdr = (u_int32_t *)args->eh;
1384 
1385 					match =
1386 					    ( want[0] == (hdr[0] & mask[0]) &&
1387 					      want[1] == (hdr[1] & mask[1]) &&
1388 					      want[2] == (hdr[2] & mask[2]) );
1389 				}
1390 				break;
1391 
1392 			case O_MAC_TYPE:
1393 				if (args->eh != NULL) {
1394 					u_int16_t *p =
1395 					    ((ipfw_insn_u16 *)cmd)->ports;
1396 					int i;
1397 
1398 					for (i = cmdlen - 1; !match && i>0;
1399 					    i--, p += 2)
1400 						match = (etype >= p[0] &&
1401 						    etype <= p[1]);
1402 				}
1403 				break;
1404 
1405 			case O_FRAG:
1406 				match = (offset != 0);
1407 				break;
1408 
1409 			case O_IN:	/* "out" is "not in" */
1410 				match = (oif == NULL);
1411 				break;
1412 
1413 			case O_LAYER2:
1414 				match = (args->eh != NULL);
1415 				break;
1416 
1417 			case O_DIVERTED:
1418 			    {
1419 				/* For diverted packets, args->rule.info
1420 				 * contains the divert port (in host format)
1421 				 * reason and direction.
1422 				 */
1423 				uint32_t i = args->rule.info;
1424 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1425 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1426 			    }
1427 				break;
1428 
1429 			case O_PROTO:
1430 				/*
1431 				 * We do not allow an arg of 0 so the
1432 				 * check of "proto" only suffices.
1433 				 */
1434 				match = (proto == cmd->arg1);
1435 				break;
1436 
1437 			case O_IP_SRC:
1438 				match = is_ipv4 &&
1439 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1440 				    src_ip.s_addr);
1441 				break;
1442 
1443 			case O_IP_SRC_LOOKUP:
1444 			case O_IP_DST_LOOKUP:
1445 				if (is_ipv4) {
1446 				    uint32_t key =
1447 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1448 					    dst_ip.s_addr : src_ip.s_addr;
1449 				    uint32_t v = 0;
1450 
1451 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1452 					/* generic lookup. The key must be
1453 					 * in 32bit big-endian format.
1454 					 */
1455 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1456 					if (v == 0)
1457 					    key = dst_ip.s_addr;
1458 					else if (v == 1)
1459 					    key = src_ip.s_addr;
1460 					else if (v == 6) /* dscp */
1461 					    key = (ip->ip_tos >> 2) & 0x3f;
1462 					else if (offset != 0)
1463 					    break;
1464 					else if (proto != IPPROTO_TCP &&
1465 						proto != IPPROTO_UDP)
1466 					    break;
1467 					else if (v == 2)
1468 					    key = htonl(dst_port);
1469 					else if (v == 3)
1470 					    key = htonl(src_port);
1471 #ifndef USERSPACE
1472 					else if (v == 4 || v == 5) {
1473 					    check_uidgid(
1474 						(ipfw_insn_u32 *)cmd,
1475 						args, &ucred_lookup,
1476 #ifdef __FreeBSD__
1477 						&ucred_cache);
1478 					    if (v == 4 /* O_UID */)
1479 						key = ucred_cache->cr_uid;
1480 					    else if (v == 5 /* O_JAIL */)
1481 						key = ucred_cache->cr_prison->pr_id;
1482 #else /* !__FreeBSD__ */
1483 						(void *)&ucred_cache);
1484 					    if (v ==4 /* O_UID */)
1485 						key = ucred_cache.uid;
1486 					    else if (v == 5 /* O_JAIL */)
1487 						key = ucred_cache.xid;
1488 #endif /* !__FreeBSD__ */
1489 					    key = htonl(key);
1490 					} else
1491 #endif /* !USERSPACE */
1492 					    break;
1493 				    }
1494 				    match = ipfw_lookup_table(chain,
1495 					cmd->arg1, key, &v);
1496 				    if (!match)
1497 					break;
1498 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1499 					match =
1500 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1501 				    else
1502 					tablearg = v;
1503 				} else if (is_ipv6) {
1504 					uint32_t v = 0;
1505 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1506 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1507 					match = ipfw_lookup_table_extended(chain,
1508 							cmd->arg1, pkey, &v,
1509 							IPFW_TABLE_CIDR);
1510 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1511 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1512 					if (match)
1513 						tablearg = v;
1514 				}
1515 				break;
1516 
1517 			case O_IP_SRC_MASK:
1518 			case O_IP_DST_MASK:
1519 				if (is_ipv4) {
1520 				    uint32_t a =
1521 					(cmd->opcode == O_IP_DST_MASK) ?
1522 					    dst_ip.s_addr : src_ip.s_addr;
1523 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1524 				    int i = cmdlen-1;
1525 
1526 				    for (; !match && i>0; i-= 2, p+= 2)
1527 					match = (p[0] == (a & p[1]));
1528 				}
1529 				break;
1530 
1531 			case O_IP_SRC_ME:
1532 				if (is_ipv4) {
1533 					struct ifnet *tif;
1534 
1535 					INADDR_TO_IFP(src_ip, tif);
1536 					match = (tif != NULL);
1537 					break;
1538 				}
1539 #ifdef INET6
1540 				/* FALLTHROUGH */
1541 			case O_IP6_SRC_ME:
1542 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1543 #endif
1544 				break;
1545 
1546 			case O_IP_DST_SET:
1547 			case O_IP_SRC_SET:
1548 				if (is_ipv4) {
1549 					u_int32_t *d = (u_int32_t *)(cmd+1);
1550 					u_int32_t addr =
1551 					    cmd->opcode == O_IP_DST_SET ?
1552 						args->f_id.dst_ip :
1553 						args->f_id.src_ip;
1554 
1555 					    if (addr < d[0])
1556 						    break;
1557 					    addr -= d[0]; /* subtract base */
1558 					    match = (addr < cmd->arg1) &&
1559 						( d[ 1 + (addr>>5)] &
1560 						  (1<<(addr & 0x1f)) );
1561 				}
1562 				break;
1563 
1564 			case O_IP_DST:
1565 				match = is_ipv4 &&
1566 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1567 				    dst_ip.s_addr);
1568 				break;
1569 
1570 			case O_IP_DST_ME:
1571 				if (is_ipv4) {
1572 					struct ifnet *tif;
1573 
1574 					INADDR_TO_IFP(dst_ip, tif);
1575 					match = (tif != NULL);
1576 					break;
1577 				}
1578 #ifdef INET6
1579 				/* FALLTHROUGH */
1580 			case O_IP6_DST_ME:
1581 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1582 #endif
1583 				break;
1584 
1585 
1586 			case O_IP_SRCPORT:
1587 			case O_IP_DSTPORT:
1588 				/*
1589 				 * offset == 0 && proto != 0 is enough
1590 				 * to guarantee that we have a
1591 				 * packet with port info.
1592 				 */
1593 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1594 				    && offset == 0) {
1595 					u_int16_t x =
1596 					    (cmd->opcode == O_IP_SRCPORT) ?
1597 						src_port : dst_port ;
1598 					u_int16_t *p =
1599 					    ((ipfw_insn_u16 *)cmd)->ports;
1600 					int i;
1601 
1602 					for (i = cmdlen - 1; !match && i>0;
1603 					    i--, p += 2)
1604 						match = (x>=p[0] && x<=p[1]);
1605 				}
1606 				break;
1607 
1608 			case O_ICMPTYPE:
1609 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1610 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1611 				break;
1612 
1613 #ifdef INET6
1614 			case O_ICMP6TYPE:
1615 				match = is_ipv6 && offset == 0 &&
1616 				    proto==IPPROTO_ICMPV6 &&
1617 				    icmp6type_match(
1618 					ICMP6(ulp)->icmp6_type,
1619 					(ipfw_insn_u32 *)cmd);
1620 				break;
1621 #endif /* INET6 */
1622 
1623 			case O_IPOPT:
1624 				match = (is_ipv4 &&
1625 				    ipopts_match(ip, cmd) );
1626 				break;
1627 
1628 			case O_IPVER:
1629 				match = (is_ipv4 &&
1630 				    cmd->arg1 == ip->ip_v);
1631 				break;
1632 
1633 			case O_IPID:
1634 			case O_IPLEN:
1635 			case O_IPTTL:
1636 				if (is_ipv4) {	/* only for IP packets */
1637 				    uint16_t x;
1638 				    uint16_t *p;
1639 				    int i;
1640 
1641 				    if (cmd->opcode == O_IPLEN)
1642 					x = iplen;
1643 				    else if (cmd->opcode == O_IPTTL)
1644 					x = ip->ip_ttl;
1645 				    else /* must be IPID */
1646 					x = ntohs(ip->ip_id);
1647 				    if (cmdlen == 1) {
1648 					match = (cmd->arg1 == x);
1649 					break;
1650 				    }
1651 				    /* otherwise we have ranges */
1652 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1653 				    i = cmdlen - 1;
1654 				    for (; !match && i>0; i--, p += 2)
1655 					match = (x >= p[0] && x <= p[1]);
1656 				}
1657 				break;
1658 
1659 			case O_IPPRECEDENCE:
1660 				match = (is_ipv4 &&
1661 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1662 				break;
1663 
1664 			case O_IPTOS:
1665 				match = (is_ipv4 &&
1666 				    flags_match(cmd, ip->ip_tos));
1667 				break;
1668 
1669 			case O_DSCP:
1670 			    {
1671 				uint32_t *p;
1672 				uint16_t x;
1673 
1674 				p = ((ipfw_insn_u32 *)cmd)->d;
1675 
1676 				if (is_ipv4)
1677 					x = ip->ip_tos >> 2;
1678 				else if (is_ipv6) {
1679 					uint8_t *v;
1680 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
1681 					x = (*v & 0x0F) << 2;
1682 					v++;
1683 					x |= *v >> 6;
1684 				} else
1685 					break;
1686 
1687 				/* DSCP bitmask is stored as low_u32 high_u32 */
1688 				if (x > 32)
1689 					match = *(p + 1) & (1 << (x - 32));
1690 				else
1691 					match = *p & (1 << x);
1692 			    }
1693 				break;
1694 
1695 			case O_TCPDATALEN:
1696 				if (proto == IPPROTO_TCP && offset == 0) {
1697 				    struct tcphdr *tcp;
1698 				    uint16_t x;
1699 				    uint16_t *p;
1700 				    int i;
1701 
1702 				    tcp = TCP(ulp);
1703 				    x = iplen -
1704 					((ip->ip_hl + tcp->th_off) << 2);
1705 				    if (cmdlen == 1) {
1706 					match = (cmd->arg1 == x);
1707 					break;
1708 				    }
1709 				    /* otherwise we have ranges */
1710 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1711 				    i = cmdlen - 1;
1712 				    for (; !match && i>0; i--, p += 2)
1713 					match = (x >= p[0] && x <= p[1]);
1714 				}
1715 				break;
1716 
1717 			case O_TCPFLAGS:
1718 				match = (proto == IPPROTO_TCP && offset == 0 &&
1719 				    flags_match(cmd, TCP(ulp)->th_flags));
1720 				break;
1721 
1722 			case O_TCPOPTS:
1723 				PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1724 				match = (proto == IPPROTO_TCP && offset == 0 &&
1725 				    tcpopts_match(TCP(ulp), cmd));
1726 				break;
1727 
1728 			case O_TCPSEQ:
1729 				match = (proto == IPPROTO_TCP && offset == 0 &&
1730 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1731 					TCP(ulp)->th_seq);
1732 				break;
1733 
1734 			case O_TCPACK:
1735 				match = (proto == IPPROTO_TCP && offset == 0 &&
1736 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1737 					TCP(ulp)->th_ack);
1738 				break;
1739 
1740 			case O_TCPWIN:
1741 				if (proto == IPPROTO_TCP && offset == 0) {
1742 				    uint16_t x;
1743 				    uint16_t *p;
1744 				    int i;
1745 
1746 				    x = ntohs(TCP(ulp)->th_win);
1747 				    if (cmdlen == 1) {
1748 					match = (cmd->arg1 == x);
1749 					break;
1750 				    }
1751 				    /* Otherwise we have ranges. */
1752 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1753 				    i = cmdlen - 1;
1754 				    for (; !match && i > 0; i--, p += 2)
1755 					match = (x >= p[0] && x <= p[1]);
1756 				}
1757 				break;
1758 
1759 			case O_ESTAB:
1760 				/* reject packets which have SYN only */
1761 				/* XXX should i also check for TH_ACK ? */
1762 				match = (proto == IPPROTO_TCP && offset == 0 &&
1763 				    (TCP(ulp)->th_flags &
1764 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1765 				break;
1766 
1767 			case O_ALTQ: {
1768 				struct pf_mtag *at;
1769 				struct m_tag *mtag;
1770 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1771 
1772 				/*
1773 				 * ALTQ uses mbuf tags from another
1774 				 * packet filtering system - pf(4).
1775 				 * We allocate a tag in its format
1776 				 * and fill it in, pretending to be pf(4).
1777 				 */
1778 				match = 1;
1779 				at = pf_find_mtag(m);
1780 				if (at != NULL && at->qid != 0)
1781 					break;
1782 				mtag = m_tag_get(PACKET_TAG_PF,
1783 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1784 				if (mtag == NULL) {
1785 					/*
1786 					 * Let the packet fall back to the
1787 					 * default ALTQ.
1788 					 */
1789 					break;
1790 				}
1791 				m_tag_prepend(m, mtag);
1792 				at = (struct pf_mtag *)(mtag + 1);
1793 				at->qid = altq->qid;
1794 				at->hdr = ip;
1795 				break;
1796 			}
1797 
1798 			case O_LOG:
1799 				ipfw_log(f, hlen, args, m,
1800 				    oif, offset | ip6f_mf, tablearg, ip);
1801 				match = 1;
1802 				break;
1803 
1804 			case O_PROB:
1805 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1806 				break;
1807 
1808 			case O_VERREVPATH:
1809 				/* Outgoing packets automatically pass/match */
1810 				match = ((oif != NULL) ||
1811 				    (m->m_pkthdr.rcvif == NULL) ||
1812 				    (
1813 #ifdef INET6
1814 				    is_ipv6 ?
1815 					verify_path6(&(args->f_id.src_ip6),
1816 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1817 #endif
1818 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1819 				        args->f_id.fib)));
1820 				break;
1821 
1822 			case O_VERSRCREACH:
1823 				/* Outgoing packets automatically pass/match */
1824 				match = (hlen > 0 && ((oif != NULL) ||
1825 #ifdef INET6
1826 				    is_ipv6 ?
1827 				        verify_path6(&(args->f_id.src_ip6),
1828 				            NULL, args->f_id.fib) :
1829 #endif
1830 				    verify_path(src_ip, NULL, args->f_id.fib)));
1831 				break;
1832 
1833 			case O_ANTISPOOF:
1834 				/* Outgoing packets automatically pass/match */
1835 				if (oif == NULL && hlen > 0 &&
1836 				    (  (is_ipv4 && in_localaddr(src_ip))
1837 #ifdef INET6
1838 				    || (is_ipv6 &&
1839 				        in6_localaddr(&(args->f_id.src_ip6)))
1840 #endif
1841 				    ))
1842 					match =
1843 #ifdef INET6
1844 					    is_ipv6 ? verify_path6(
1845 					        &(args->f_id.src_ip6),
1846 					        m->m_pkthdr.rcvif,
1847 						args->f_id.fib) :
1848 #endif
1849 					    verify_path(src_ip,
1850 					    	m->m_pkthdr.rcvif,
1851 					        args->f_id.fib);
1852 				else
1853 					match = 1;
1854 				break;
1855 
1856 			case O_IPSEC:
1857 #ifdef IPSEC
1858 				match = (m_tag_find(m,
1859 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1860 #endif
1861 				/* otherwise no match */
1862 				break;
1863 
1864 #ifdef INET6
1865 			case O_IP6_SRC:
1866 				match = is_ipv6 &&
1867 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1868 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1869 				break;
1870 
1871 			case O_IP6_DST:
1872 				match = is_ipv6 &&
1873 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1874 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1875 				break;
1876 			case O_IP6_SRC_MASK:
1877 			case O_IP6_DST_MASK:
1878 				if (is_ipv6) {
1879 					int i = cmdlen - 1;
1880 					struct in6_addr p;
1881 					struct in6_addr *d =
1882 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1883 
1884 					for (; !match && i > 0; d += 2,
1885 					    i -= F_INSN_SIZE(struct in6_addr)
1886 					    * 2) {
1887 						p = (cmd->opcode ==
1888 						    O_IP6_SRC_MASK) ?
1889 						    args->f_id.src_ip6:
1890 						    args->f_id.dst_ip6;
1891 						APPLY_MASK(&p, &d[1]);
1892 						match =
1893 						    IN6_ARE_ADDR_EQUAL(&d[0],
1894 						    &p);
1895 					}
1896 				}
1897 				break;
1898 
1899 			case O_FLOW6ID:
1900 				match = is_ipv6 &&
1901 				    flow6id_match(args->f_id.flow_id6,
1902 				    (ipfw_insn_u32 *) cmd);
1903 				break;
1904 
1905 			case O_EXT_HDR:
1906 				match = is_ipv6 &&
1907 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1908 				break;
1909 
1910 			case O_IP6:
1911 				match = is_ipv6;
1912 				break;
1913 #endif
1914 
1915 			case O_IP4:
1916 				match = is_ipv4;
1917 				break;
1918 
1919 			case O_TAG: {
1920 				struct m_tag *mtag;
1921 				uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1922 
1923 				/* Packet is already tagged with this tag? */
1924 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1925 
1926 				/* We have `untag' action when F_NOT flag is
1927 				 * present. And we must remove this mtag from
1928 				 * mbuf and reset `match' to zero (`match' will
1929 				 * be inversed later).
1930 				 * Otherwise we should allocate new mtag and
1931 				 * push it into mbuf.
1932 				 */
1933 				if (cmd->len & F_NOT) { /* `untag' action */
1934 					if (mtag != NULL)
1935 						m_tag_delete(m, mtag);
1936 					match = 0;
1937 				} else {
1938 					if (mtag == NULL) {
1939 						mtag = m_tag_alloc( MTAG_IPFW,
1940 						    tag, 0, M_NOWAIT);
1941 						if (mtag != NULL)
1942 							m_tag_prepend(m, mtag);
1943 					}
1944 					match = 1;
1945 				}
1946 				break;
1947 			}
1948 
1949 			case O_FIB: /* try match the specified fib */
1950 				if (args->f_id.fib == cmd->arg1)
1951 					match = 1;
1952 				break;
1953 
1954 			case O_SOCKARG:	{
1955 #ifndef USERSPACE	/* not supported in userspace */
1956 				struct inpcb *inp = args->inp;
1957 				struct inpcbinfo *pi;
1958 
1959 				if (is_ipv6) /* XXX can we remove this ? */
1960 					break;
1961 
1962 				if (proto == IPPROTO_TCP)
1963 					pi = &V_tcbinfo;
1964 				else if (proto == IPPROTO_UDP)
1965 					pi = &V_udbinfo;
1966 				else
1967 					break;
1968 
1969 				/*
1970 				 * XXXRW: so_user_cookie should almost
1971 				 * certainly be inp_user_cookie?
1972 				 */
1973 
1974 				/* For incomming packet, lookup up the
1975 				inpcb using the src/dest ip/port tuple */
1976 				if (inp == NULL) {
1977 					inp = in_pcblookup(pi,
1978 						src_ip, htons(src_port),
1979 						dst_ip, htons(dst_port),
1980 						INPLOOKUP_RLOCKPCB, NULL);
1981 					if (inp != NULL) {
1982 						tablearg =
1983 						    inp->inp_socket->so_user_cookie;
1984 						if (tablearg)
1985 							match = 1;
1986 						INP_RUNLOCK(inp);
1987 					}
1988 				} else {
1989 					if (inp->inp_socket) {
1990 						tablearg =
1991 						    inp->inp_socket->so_user_cookie;
1992 						if (tablearg)
1993 							match = 1;
1994 					}
1995 				}
1996 #endif /* !USERSPACE */
1997 				break;
1998 			}
1999 
2000 			case O_TAGGED: {
2001 				struct m_tag *mtag;
2002 				uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
2003 
2004 				if (cmdlen == 1) {
2005 					match = m_tag_locate(m, MTAG_IPFW,
2006 					    tag, NULL) != NULL;
2007 					break;
2008 				}
2009 
2010 				/* we have ranges */
2011 				for (mtag = m_tag_first(m);
2012 				    mtag != NULL && !match;
2013 				    mtag = m_tag_next(m, mtag)) {
2014 					uint16_t *p;
2015 					int i;
2016 
2017 					if (mtag->m_tag_cookie != MTAG_IPFW)
2018 						continue;
2019 
2020 					p = ((ipfw_insn_u16 *)cmd)->ports;
2021 					i = cmdlen - 1;
2022 					for(; !match && i > 0; i--, p += 2)
2023 						match =
2024 						    mtag->m_tag_id >= p[0] &&
2025 						    mtag->m_tag_id <= p[1];
2026 				}
2027 				break;
2028 			}
2029 
2030 			/*
2031 			 * The second set of opcodes represents 'actions',
2032 			 * i.e. the terminal part of a rule once the packet
2033 			 * matches all previous patterns.
2034 			 * Typically there is only one action for each rule,
2035 			 * and the opcode is stored at the end of the rule
2036 			 * (but there are exceptions -- see below).
2037 			 *
2038 			 * In general, here we set retval and terminate the
2039 			 * outer loop (would be a 'break 3' in some language,
2040 			 * but we need to set l=0, done=1)
2041 			 *
2042 			 * Exceptions:
2043 			 * O_COUNT and O_SKIPTO actions:
2044 			 *   instead of terminating, we jump to the next rule
2045 			 *   (setting l=0), or to the SKIPTO target (setting
2046 			 *   f/f_len, cmd and l as needed), respectively.
2047 			 *
2048 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2049 			 *   perform some action and set match = 1;
2050 			 *
2051 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2052 			 *   not real 'actions', and are stored right
2053 			 *   before the 'action' part of the rule.
2054 			 *   These opcodes try to install an entry in the
2055 			 *   state tables; if successful, we continue with
2056 			 *   the next opcode (match=1; break;), otherwise
2057 			 *   the packet must be dropped (set retval,
2058 			 *   break loops with l=0, done=1)
2059 			 *
2060 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2061 			 *   cause a lookup of the state table, and a jump
2062 			 *   to the 'action' part of the parent rule
2063 			 *   if an entry is found, or
2064 			 *   (CHECK_STATE only) a jump to the next rule if
2065 			 *   the entry is not found.
2066 			 *   The result of the lookup is cached so that
2067 			 *   further instances of these opcodes become NOPs.
2068 			 *   The jump to the next rule is done by setting
2069 			 *   l=0, cmdlen=0.
2070 			 */
2071 			case O_LIMIT:
2072 			case O_KEEP_STATE:
2073 				if (ipfw_install_state(f,
2074 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2075 					/* error or limit violation */
2076 					retval = IP_FW_DENY;
2077 					l = 0;	/* exit inner loop */
2078 					done = 1; /* exit outer loop */
2079 				}
2080 				match = 1;
2081 				break;
2082 
2083 			case O_PROBE_STATE:
2084 			case O_CHECK_STATE:
2085 				/*
2086 				 * dynamic rules are checked at the first
2087 				 * keep-state or check-state occurrence,
2088 				 * with the result being stored in dyn_dir.
2089 				 * The compiler introduces a PROBE_STATE
2090 				 * instruction for us when we have a
2091 				 * KEEP_STATE (because PROBE_STATE needs
2092 				 * to be run first).
2093 				 */
2094 				if (dyn_dir == MATCH_UNKNOWN &&
2095 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2096 				     &dyn_dir, proto == IPPROTO_TCP ?
2097 					TCP(ulp) : NULL))
2098 					!= NULL) {
2099 					/*
2100 					 * Found dynamic entry, update stats
2101 					 * and jump to the 'action' part of
2102 					 * the parent rule by setting
2103 					 * f, cmd, l and clearing cmdlen.
2104 					 */
2105 					IPFW_INC_DYN_COUNTER(q, pktlen);
2106 					/* XXX we would like to have f_pos
2107 					 * readily accessible in the dynamic
2108 				         * rule, instead of having to
2109 					 * lookup q->rule.
2110 					 */
2111 					f = q->rule;
2112 					f_pos = ipfw_find_rule(chain,
2113 						f->rulenum, f->id);
2114 					cmd = ACTION_PTR(f);
2115 					l = f->cmd_len - f->act_ofs;
2116 					ipfw_dyn_unlock(q);
2117 					cmdlen = 0;
2118 					match = 1;
2119 					break;
2120 				}
2121 				/*
2122 				 * Dynamic entry not found. If CHECK_STATE,
2123 				 * skip to next rule, if PROBE_STATE just
2124 				 * ignore and continue with next opcode.
2125 				 */
2126 				if (cmd->opcode == O_CHECK_STATE)
2127 					l = 0;	/* exit inner loop */
2128 				match = 1;
2129 				break;
2130 
2131 			case O_ACCEPT:
2132 				retval = 0;	/* accept */
2133 				l = 0;		/* exit inner loop */
2134 				done = 1;	/* exit outer loop */
2135 				break;
2136 
2137 			case O_PIPE:
2138 			case O_QUEUE:
2139 				set_match(args, f_pos, chain);
2140 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2141 				if (cmd->opcode == O_PIPE)
2142 					args->rule.info |= IPFW_IS_PIPE;
2143 				if (V_fw_one_pass)
2144 					args->rule.info |= IPFW_ONEPASS;
2145 				retval = IP_FW_DUMMYNET;
2146 				l = 0;          /* exit inner loop */
2147 				done = 1;       /* exit outer loop */
2148 				break;
2149 
2150 			case O_DIVERT:
2151 			case O_TEE:
2152 				if (args->eh) /* not on layer 2 */
2153 				    break;
2154 				/* otherwise this is terminal */
2155 				l = 0;		/* exit inner loop */
2156 				done = 1;	/* exit outer loop */
2157 				retval = (cmd->opcode == O_DIVERT) ?
2158 					IP_FW_DIVERT : IP_FW_TEE;
2159 				set_match(args, f_pos, chain);
2160 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2161 				break;
2162 
2163 			case O_COUNT:
2164 				IPFW_INC_RULE_COUNTER(f, pktlen);
2165 				l = 0;		/* exit inner loop */
2166 				break;
2167 
2168 			case O_SKIPTO:
2169 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2170 			    f_pos = jump_fast(chain, f, cmd->arg1, tablearg, 0);
2171 			    /*
2172 			     * Skip disabled rules, and re-enter
2173 			     * the inner loop with the correct
2174 			     * f_pos, f, l and cmd.
2175 			     * Also clear cmdlen and skip_or
2176 			     */
2177 			    for (; f_pos < chain->n_rules - 1 &&
2178 				    (V_set_disable &
2179 				     (1 << chain->map[f_pos]->set));
2180 				    f_pos++)
2181 				;
2182 			    /* Re-enter the inner loop at the skipto rule. */
2183 			    f = chain->map[f_pos];
2184 			    l = f->cmd_len;
2185 			    cmd = f->cmd;
2186 			    match = 1;
2187 			    cmdlen = 0;
2188 			    skip_or = 0;
2189 			    continue;
2190 			    break;	/* not reached */
2191 
2192 			case O_CALLRETURN: {
2193 				/*
2194 				 * Implementation of `subroutine' call/return,
2195 				 * in the stack carried in an mbuf tag. This
2196 				 * is different from `skipto' in that any call
2197 				 * address is possible (`skipto' must prevent
2198 				 * backward jumps to avoid endless loops).
2199 				 * We have `return' action when F_NOT flag is
2200 				 * present. The `m_tag_id' field is used as
2201 				 * stack pointer.
2202 				 */
2203 				struct m_tag *mtag;
2204 				uint16_t jmpto, *stack;
2205 
2206 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2207 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2208 				/*
2209 				 * Hand-rolled version of m_tag_locate() with
2210 				 * wildcard `type'.
2211 				 * If not already tagged, allocate new tag.
2212 				 */
2213 				mtag = m_tag_first(m);
2214 				while (mtag != NULL) {
2215 					if (mtag->m_tag_cookie ==
2216 					    MTAG_IPFW_CALL)
2217 						break;
2218 					mtag = m_tag_next(m, mtag);
2219 				}
2220 				if (mtag == NULL && IS_CALL) {
2221 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2222 					    IPFW_CALLSTACK_SIZE *
2223 					    sizeof(uint16_t), M_NOWAIT);
2224 					if (mtag != NULL)
2225 						m_tag_prepend(m, mtag);
2226 				}
2227 
2228 				/*
2229 				 * On error both `call' and `return' just
2230 				 * continue with next rule.
2231 				 */
2232 				if (IS_RETURN && (mtag == NULL ||
2233 				    mtag->m_tag_id == 0)) {
2234 					l = 0;		/* exit inner loop */
2235 					break;
2236 				}
2237 				if (IS_CALL && (mtag == NULL ||
2238 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2239 					printf("ipfw: call stack error, "
2240 					    "go to next rule\n");
2241 					l = 0;		/* exit inner loop */
2242 					break;
2243 				}
2244 
2245 				IPFW_INC_RULE_COUNTER(f, pktlen);
2246 				stack = (uint16_t *)(mtag + 1);
2247 
2248 				/*
2249 				 * The `call' action may use cached f_pos
2250 				 * (in f->next_rule), whose version is written
2251 				 * in f->next_rule.
2252 				 * The `return' action, however, doesn't have
2253 				 * fixed jump address in cmd->arg1 and can't use
2254 				 * cache.
2255 				 */
2256 				if (IS_CALL) {
2257 					stack[mtag->m_tag_id] = f->rulenum;
2258 					mtag->m_tag_id++;
2259 			    		f_pos = jump_fast(chain, f, cmd->arg1,
2260 					    tablearg, 1);
2261 				} else {	/* `return' action */
2262 					mtag->m_tag_id--;
2263 					jmpto = stack[mtag->m_tag_id] + 1;
2264 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2265 				}
2266 
2267 				/*
2268 				 * Skip disabled rules, and re-enter
2269 				 * the inner loop with the correct
2270 				 * f_pos, f, l and cmd.
2271 				 * Also clear cmdlen and skip_or
2272 				 */
2273 				for (; f_pos < chain->n_rules - 1 &&
2274 				    (V_set_disable &
2275 				    (1 << chain->map[f_pos]->set)); f_pos++)
2276 					;
2277 				/* Re-enter the inner loop at the dest rule. */
2278 				f = chain->map[f_pos];
2279 				l = f->cmd_len;
2280 				cmd = f->cmd;
2281 				cmdlen = 0;
2282 				skip_or = 0;
2283 				continue;
2284 				break;	/* NOTREACHED */
2285 			}
2286 #undef IS_CALL
2287 #undef IS_RETURN
2288 
2289 			case O_REJECT:
2290 				/*
2291 				 * Drop the packet and send a reject notice
2292 				 * if the packet is not ICMP (or is an ICMP
2293 				 * query), and it is not multicast/broadcast.
2294 				 */
2295 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2296 				    (proto != IPPROTO_ICMP ||
2297 				     is_icmp_query(ICMP(ulp))) &&
2298 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2299 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2300 					send_reject(args, cmd->arg1, iplen, ip);
2301 					m = args->m;
2302 				}
2303 				/* FALLTHROUGH */
2304 #ifdef INET6
2305 			case O_UNREACH6:
2306 				if (hlen > 0 && is_ipv6 &&
2307 				    ((offset & IP6F_OFF_MASK) == 0) &&
2308 				    (proto != IPPROTO_ICMPV6 ||
2309 				     (is_icmp6_query(icmp6_type) == 1)) &&
2310 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2311 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2312 					send_reject6(
2313 					    args, cmd->arg1, hlen,
2314 					    (struct ip6_hdr *)ip);
2315 					m = args->m;
2316 				}
2317 				/* FALLTHROUGH */
2318 #endif
2319 			case O_DENY:
2320 				retval = IP_FW_DENY;
2321 				l = 0;		/* exit inner loop */
2322 				done = 1;	/* exit outer loop */
2323 				break;
2324 
2325 			case O_FORWARD_IP:
2326 				if (args->eh)	/* not valid on layer2 pkts */
2327 					break;
2328 				if (q == NULL || q->rule != f ||
2329 				    dyn_dir == MATCH_FORWARD) {
2330 				    struct sockaddr_in *sa;
2331 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2332 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2333 					bcopy(sa, &args->hopstore,
2334 							sizeof(*sa));
2335 					args->hopstore.sin_addr.s_addr =
2336 						    htonl(tablearg);
2337 					args->next_hop = &args->hopstore;
2338 				    } else {
2339 					args->next_hop = sa;
2340 				    }
2341 				}
2342 				retval = IP_FW_PASS;
2343 				l = 0;          /* exit inner loop */
2344 				done = 1;       /* exit outer loop */
2345 				break;
2346 
2347 #ifdef INET6
2348 			case O_FORWARD_IP6:
2349 				if (args->eh)	/* not valid on layer2 pkts */
2350 					break;
2351 				if (q == NULL || q->rule != f ||
2352 				    dyn_dir == MATCH_FORWARD) {
2353 					struct sockaddr_in6 *sin6;
2354 
2355 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2356 					args->next_hop6 = sin6;
2357 				}
2358 				retval = IP_FW_PASS;
2359 				l = 0;		/* exit inner loop */
2360 				done = 1;	/* exit outer loop */
2361 				break;
2362 #endif
2363 
2364 			case O_NETGRAPH:
2365 			case O_NGTEE:
2366 				set_match(args, f_pos, chain);
2367 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2368 				if (V_fw_one_pass)
2369 					args->rule.info |= IPFW_ONEPASS;
2370 				retval = (cmd->opcode == O_NETGRAPH) ?
2371 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2372 				l = 0;          /* exit inner loop */
2373 				done = 1;       /* exit outer loop */
2374 				break;
2375 
2376 			case O_SETFIB: {
2377 				uint32_t fib;
2378 
2379 				IPFW_INC_RULE_COUNTER(f, pktlen);
2380 				fib = IP_FW_ARG_TABLEARG(cmd->arg1);
2381 				if (fib >= rt_numfibs)
2382 					fib = 0;
2383 				M_SETFIB(m, fib);
2384 				args->f_id.fib = fib;
2385 				l = 0;		/* exit inner loop */
2386 				break;
2387 		        }
2388 
2389 			case O_SETDSCP: {
2390 				uint16_t code;
2391 
2392 				code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
2393 				l = 0;		/* exit inner loop */
2394 				if (is_ipv4) {
2395 					uint16_t a;
2396 
2397 					a = ip->ip_tos;
2398 					ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2399 					a += ntohs(ip->ip_sum) - ip->ip_tos;
2400 					ip->ip_sum = htons(a);
2401 				} else if (is_ipv6) {
2402 					uint8_t *v;
2403 
2404 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2405 					*v = (*v & 0xF0) | (code >> 2);
2406 					v++;
2407 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2408 				} else
2409 					break;
2410 
2411 				IPFW_INC_RULE_COUNTER(f, pktlen);
2412 				break;
2413 			}
2414 
2415 			case O_NAT:
2416 				l = 0;          /* exit inner loop */
2417 				done = 1;       /* exit outer loop */
2418  				if (!IPFW_NAT_LOADED) {
2419 				    retval = IP_FW_DENY;
2420 				    break;
2421 				}
2422 
2423 				struct cfg_nat *t;
2424 				int nat_id;
2425 
2426 				set_match(args, f_pos, chain);
2427 				/* Check if this is 'global' nat rule */
2428 				if (cmd->arg1 == 0) {
2429 					retval = ipfw_nat_ptr(args, NULL, m);
2430 					break;
2431 				}
2432 				t = ((ipfw_insn_nat *)cmd)->nat;
2433 				if (t == NULL) {
2434 					nat_id = IP_FW_ARG_TABLEARG(cmd->arg1);
2435 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2436 
2437 					if (t == NULL) {
2438 					    retval = IP_FW_DENY;
2439 					    break;
2440 					}
2441 					if (cmd->arg1 != IP_FW_TABLEARG)
2442 					    ((ipfw_insn_nat *)cmd)->nat = t;
2443 				}
2444 				retval = ipfw_nat_ptr(args, t, m);
2445 				break;
2446 
2447 			case O_REASS: {
2448 				int ip_off;
2449 
2450 				IPFW_INC_RULE_COUNTER(f, pktlen);
2451 				l = 0;	/* in any case exit inner loop */
2452 				ip_off = ntohs(ip->ip_off);
2453 
2454 				/* if not fragmented, go to next rule */
2455 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2456 				    break;
2457 
2458 				args->m = m = ip_reass(m);
2459 
2460 				/*
2461 				 * do IP header checksum fixup.
2462 				 */
2463 				if (m == NULL) { /* fragment got swallowed */
2464 				    retval = IP_FW_DENY;
2465 				} else { /* good, packet complete */
2466 				    int hlen;
2467 
2468 				    ip = mtod(m, struct ip *);
2469 				    hlen = ip->ip_hl << 2;
2470 				    ip->ip_sum = 0;
2471 				    if (hlen == sizeof(struct ip))
2472 					ip->ip_sum = in_cksum_hdr(ip);
2473 				    else
2474 					ip->ip_sum = in_cksum(m, hlen);
2475 				    retval = IP_FW_REASS;
2476 				    set_match(args, f_pos, chain);
2477 				}
2478 				done = 1;	/* exit outer loop */
2479 				break;
2480 			}
2481 
2482 			default:
2483 				panic("-- unknown opcode %d\n", cmd->opcode);
2484 			} /* end of switch() on opcodes */
2485 			/*
2486 			 * if we get here with l=0, then match is irrelevant.
2487 			 */
2488 
2489 			if (cmd->len & F_NOT)
2490 				match = !match;
2491 
2492 			if (match) {
2493 				if (cmd->len & F_OR)
2494 					skip_or = 1;
2495 			} else {
2496 				if (!(cmd->len & F_OR)) /* not an OR block, */
2497 					break;		/* try next rule    */
2498 			}
2499 
2500 		}	/* end of inner loop, scan opcodes */
2501 #undef PULLUP_LEN
2502 
2503 		if (done)
2504 			break;
2505 
2506 /* next_rule:; */	/* try next rule		*/
2507 
2508 	}		/* end of outer for, scan rules */
2509 
2510 	if (done) {
2511 		struct ip_fw *rule = chain->map[f_pos];
2512 		/* Update statistics */
2513 		IPFW_INC_RULE_COUNTER(rule, pktlen);
2514 	} else {
2515 		retval = IP_FW_DENY;
2516 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2517 	}
2518 	IPFW_PF_RUNLOCK(chain);
2519 #ifdef __FreeBSD__
2520 	if (ucred_cache != NULL)
2521 		crfree(ucred_cache);
2522 #endif
2523 	return (retval);
2524 
2525 pullup_failed:
2526 	if (V_fw_verbose)
2527 		printf("ipfw: pullup failed\n");
2528 	return (IP_FW_DENY);
2529 }
2530 
2531 /*
2532  * Set maximum number of tables that can be used in given VNET ipfw instance.
2533  */
2534 #ifdef SYSCTL_NODE
2535 static int
2536 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2537 {
2538 	int error;
2539 	unsigned int ntables;
2540 
2541 	ntables = V_fw_tables_max;
2542 
2543 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2544 	/* Read operation or some error */
2545 	if ((error != 0) || (req->newptr == NULL))
2546 		return (error);
2547 
2548 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2549 }
2550 #endif
2551 /*
2552  * Module and VNET glue
2553  */
2554 
2555 /*
2556  * Stuff that must be initialised only on boot or module load
2557  */
2558 static int
2559 ipfw_init(void)
2560 {
2561 	int error = 0;
2562 
2563 	/*
2564  	 * Only print out this stuff the first time around,
2565 	 * when called from the sysinit code.
2566 	 */
2567 	printf("ipfw2 "
2568 #ifdef INET6
2569 		"(+ipv6) "
2570 #endif
2571 		"initialized, divert %s, nat %s, "
2572 		"default to %s, logging ",
2573 #ifdef IPDIVERT
2574 		"enabled",
2575 #else
2576 		"loadable",
2577 #endif
2578 #ifdef IPFIREWALL_NAT
2579 		"enabled",
2580 #else
2581 		"loadable",
2582 #endif
2583 		default_to_accept ? "accept" : "deny");
2584 
2585 	/*
2586 	 * Note: V_xxx variables can be accessed here but the vnet specific
2587 	 * initializer may not have been called yet for the VIMAGE case.
2588 	 * Tuneables will have been processed. We will print out values for
2589 	 * the default vnet.
2590 	 * XXX This should all be rationalized AFTER 8.0
2591 	 */
2592 	if (V_fw_verbose == 0)
2593 		printf("disabled\n");
2594 	else if (V_verbose_limit == 0)
2595 		printf("unlimited\n");
2596 	else
2597 		printf("limited to %d packets/entry by default\n",
2598 		    V_verbose_limit);
2599 
2600 	/* Check user-supplied table count for validness */
2601 	if (default_fw_tables > IPFW_TABLES_MAX)
2602 	  default_fw_tables = IPFW_TABLES_MAX;
2603 
2604 	ipfw_log_bpf(1); /* init */
2605 	return (error);
2606 }
2607 
2608 /*
2609  * Called for the removal of the last instance only on module unload.
2610  */
2611 static void
2612 ipfw_destroy(void)
2613 {
2614 
2615 	ipfw_log_bpf(0); /* uninit */
2616 	printf("IP firewall unloaded\n");
2617 }
2618 
2619 /*
2620  * Stuff that must be initialized for every instance
2621  * (including the first of course).
2622  */
2623 static int
2624 vnet_ipfw_init(const void *unused)
2625 {
2626 	int error;
2627 	struct ip_fw *rule = NULL;
2628 	struct ip_fw_chain *chain;
2629 
2630 	chain = &V_layer3_chain;
2631 
2632 	/* First set up some values that are compile time options */
2633 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2634 	V_fw_deny_unknown_exthdrs = 1;
2635 #ifdef IPFIREWALL_VERBOSE
2636 	V_fw_verbose = 1;
2637 #endif
2638 #ifdef IPFIREWALL_VERBOSE_LIMIT
2639 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2640 #endif
2641 #ifdef IPFIREWALL_NAT
2642 	LIST_INIT(&chain->nat);
2643 #endif
2644 
2645 	/* insert the default rule and create the initial map */
2646 	chain->n_rules = 1;
2647 	chain->static_len = sizeof(struct ip_fw);
2648 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2649 	if (chain->map)
2650 		rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2651 
2652 	/* Set initial number of tables */
2653 	V_fw_tables_max = default_fw_tables;
2654 	error = ipfw_init_tables(chain);
2655 	if (error) {
2656 		printf("ipfw2: setting up tables failed\n");
2657 		free(chain->map, M_IPFW);
2658 		free(rule, M_IPFW);
2659 		return (ENOSPC);
2660 	}
2661 
2662 	/* fill and insert the default rule */
2663 	rule->act_ofs = 0;
2664 	rule->rulenum = IPFW_DEFAULT_RULE;
2665 	rule->cmd_len = 1;
2666 	rule->set = RESVD_SET;
2667 	rule->cmd[0].len = 1;
2668 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2669 	chain->default_rule = chain->map[0] = rule;
2670 	chain->id = rule->id = 1;
2671 
2672 	IPFW_LOCK_INIT(chain);
2673 	ipfw_dyn_init(chain);
2674 
2675 	/* First set up some values that are compile time options */
2676 	V_ipfw_vnet_ready = 1;		/* Open for business */
2677 
2678 	/*
2679 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2680 	 * Even if the latter two fail we still keep the module alive
2681 	 * because the sockopt and layer2 paths are still useful.
2682 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2683 	 * so we can ignore the exact return value and just set a flag.
2684 	 *
2685 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2686 	 * changes in the underlying (per-vnet) variables trigger
2687 	 * immediate hook()/unhook() calls.
2688 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2689 	 * is checked on each packet because there are no pfil hooks.
2690 	 */
2691 	V_ip_fw_ctl_ptr = ipfw_ctl;
2692 	error = ipfw_attach_hooks(1);
2693 	return (error);
2694 }
2695 
2696 /*
2697  * Called for the removal of each instance.
2698  */
2699 static int
2700 vnet_ipfw_uninit(const void *unused)
2701 {
2702 	struct ip_fw *reap, *rule;
2703 	struct ip_fw_chain *chain = &V_layer3_chain;
2704 	int i;
2705 
2706 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2707 	/*
2708 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2709 	 * Then grab, release and grab again the WLOCK so we make
2710 	 * sure the update is propagated and nobody will be in.
2711 	 */
2712 	(void)ipfw_attach_hooks(0 /* detach */);
2713 	V_ip_fw_ctl_ptr = NULL;
2714 	IPFW_UH_WLOCK(chain);
2715 	IPFW_UH_WUNLOCK(chain);
2716 	IPFW_UH_WLOCK(chain);
2717 
2718 	IPFW_WLOCK(chain);
2719 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2720 	IPFW_WUNLOCK(chain);
2721 
2722 	ipfw_destroy_tables(chain);
2723 	reap = NULL;
2724 	IPFW_WLOCK(chain);
2725 	for (i = 0; i < chain->n_rules; i++) {
2726 		rule = chain->map[i];
2727 		rule->x_next = reap;
2728 		reap = rule;
2729 	}
2730 	if (chain->map)
2731 		free(chain->map, M_IPFW);
2732 	IPFW_WUNLOCK(chain);
2733 	IPFW_UH_WUNLOCK(chain);
2734 	if (reap != NULL)
2735 		ipfw_reap_rules(reap);
2736 	IPFW_LOCK_DESTROY(chain);
2737 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2738 	return 0;
2739 }
2740 
2741 /*
2742  * Module event handler.
2743  * In general we have the choice of handling most of these events by the
2744  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2745  * use the SYSINIT handlers as they are more capable of expressing the
2746  * flow of control during module and vnet operations, so this is just
2747  * a skeleton. Note there is no SYSINIT equivalent of the module
2748  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2749  */
2750 static int
2751 ipfw_modevent(module_t mod, int type, void *unused)
2752 {
2753 	int err = 0;
2754 
2755 	switch (type) {
2756 	case MOD_LOAD:
2757 		/* Called once at module load or
2758 	 	 * system boot if compiled in. */
2759 		break;
2760 	case MOD_QUIESCE:
2761 		/* Called before unload. May veto unloading. */
2762 		break;
2763 	case MOD_UNLOAD:
2764 		/* Called during unload. */
2765 		break;
2766 	case MOD_SHUTDOWN:
2767 		/* Called during system shutdown. */
2768 		break;
2769 	default:
2770 		err = EOPNOTSUPP;
2771 		break;
2772 	}
2773 	return err;
2774 }
2775 
2776 static moduledata_t ipfwmod = {
2777 	"ipfw",
2778 	ipfw_modevent,
2779 	0
2780 };
2781 
2782 /* Define startup order. */
2783 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2784 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2785 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2786 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2787 
2788 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2789 MODULE_VERSION(ipfw, 2);
2790 /* should declare some dependencies here */
2791 
2792 /*
2793  * Starting up. Done in order after ipfwmod() has been called.
2794  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2795  */
2796 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2797 	    ipfw_init, NULL);
2798 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2799 	    vnet_ipfw_init, NULL);
2800 
2801 /*
2802  * Closing up shop. These are done in REVERSE ORDER, but still
2803  * after ipfwmod() has been called. Not called on reboot.
2804  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2805  * or when the module is unloaded.
2806  */
2807 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2808 	    ipfw_destroy, NULL);
2809 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2810 	    vnet_ipfw_uninit, NULL);
2811 /* end of file */
2812