xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 9a0c3479e22feda1bdb2db4b97f9deb1b5fa6269)
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/route.h>
63 #include <net/pf_mtag.h>
64 #include <net/pfil.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_icmp.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_carp.h>
75 #include <netinet/pim.h>
76 #include <netinet/tcp_var.h>
77 #include <netinet/udp.h>
78 #include <netinet/udp_var.h>
79 #include <netinet/sctp.h>
80 
81 #include <netinet/ip6.h>
82 #include <netinet/icmp6.h>
83 #ifdef INET6
84 #include <netinet6/in6_pcb.h>
85 #include <netinet6/scope6_var.h>
86 #include <netinet6/ip6_var.h>
87 #endif
88 
89 #include <netpfil/ipfw/ip_fw_private.h>
90 
91 #include <machine/in_cksum.h>	/* XXX for in_cksum */
92 
93 #ifdef MAC
94 #include <security/mac/mac_framework.h>
95 #endif
96 
97 /*
98  * static variables followed by global ones.
99  * All ipfw global variables are here.
100  */
101 
102 /* ipfw_vnet_ready controls when we are open for business */
103 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
104 #define	V_ipfw_vnet_ready	VNET(ipfw_vnet_ready)
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 /* Use 128 tables by default */
123 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
124 
125 /*
126  * Each rule belongs to one of 32 different sets (0..31).
127  * The variable set_disable contains one bit per set.
128  * If the bit is set, all rules in the corresponding set
129  * are disabled. Set RESVD_SET(31) is reserved for the default rule
130  * and rules that are not deleted by the flush command,
131  * and CANNOT be disabled.
132  * Rules in set RESVD_SET can only be deleted individually.
133  */
134 VNET_DEFINE(u_int32_t, set_disable);
135 #define	V_set_disable			VNET(set_disable)
136 
137 VNET_DEFINE(int, fw_verbose);
138 /* counter for ipfw_log(NULL...) */
139 VNET_DEFINE(u_int64_t, norule_counter);
140 VNET_DEFINE(int, verbose_limit);
141 
142 /* layer3_chain contains the list of rules for layer 3 */
143 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
144 
145 ipfw_nat_t *ipfw_nat_ptr = NULL;
146 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
147 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
148 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
149 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
150 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
151 
152 #ifdef SYSCTL_NODE
153 uint32_t dummy_def = IPFW_DEFAULT_RULE;
154 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
155 
156 SYSBEGIN(f3)
157 
158 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
159 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
160     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
161     "Only do a single pass through ipfw when using dummynet(4)");
162 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
163     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
164     "Rule number auto-increment step");
165 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
166     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
167     "Log matches to ipfw rules");
168 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
169     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
170     "Set upper limit of matches of ipfw rules logged");
171 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
172     &dummy_def, 0,
173     "The default/max possible rule number.");
174 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
175     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
176     "Maximum number of tables");
177 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
178     &default_to_accept, 0,
179     "Make the default rule accept all packets.");
180 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
181 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
182 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
183     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
184     "Number of static rules");
185 
186 #ifdef INET6
187 SYSCTL_DECL(_net_inet6_ip6);
188 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
189 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
190     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
191     "Deny packets with unknown IPv6 Extension Headers");
192 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
193     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
194     "Permit single packet IPv6 fragments");
195 #endif /* INET6 */
196 
197 SYSEND
198 
199 #endif /* SYSCTL_NODE */
200 
201 
202 /*
203  * Some macros used in the various matching options.
204  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
205  * Other macros just cast void * into the appropriate type
206  */
207 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
208 #define	TCP(p)		((struct tcphdr *)(p))
209 #define	SCTP(p)		((struct sctphdr *)(p))
210 #define	UDP(p)		((struct udphdr *)(p))
211 #define	ICMP(p)		((struct icmphdr *)(p))
212 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
213 
214 static __inline int
215 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
216 {
217 	int type = icmp->icmp_type;
218 
219 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
220 }
221 
222 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
223     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
224 
225 static int
226 is_icmp_query(struct icmphdr *icmp)
227 {
228 	int type = icmp->icmp_type;
229 
230 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
231 }
232 #undef TT
233 
234 /*
235  * The following checks use two arrays of 8 or 16 bits to store the
236  * bits that we want set or clear, respectively. They are in the
237  * low and high half of cmd->arg1 or cmd->d[0].
238  *
239  * We scan options and store the bits we find set. We succeed if
240  *
241  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
242  *
243  * The code is sometimes optimized not to store additional variables.
244  */
245 
246 static int
247 flags_match(ipfw_insn *cmd, u_int8_t bits)
248 {
249 	u_char want_clear;
250 	bits = ~bits;
251 
252 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
253 		return 0; /* some bits we want set were clear */
254 	want_clear = (cmd->arg1 >> 8) & 0xff;
255 	if ( (want_clear & bits) != want_clear)
256 		return 0; /* some bits we want clear were set */
257 	return 1;
258 }
259 
260 static int
261 ipopts_match(struct ip *ip, ipfw_insn *cmd)
262 {
263 	int optlen, bits = 0;
264 	u_char *cp = (u_char *)(ip + 1);
265 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
266 
267 	for (; x > 0; x -= optlen, cp += optlen) {
268 		int opt = cp[IPOPT_OPTVAL];
269 
270 		if (opt == IPOPT_EOL)
271 			break;
272 		if (opt == IPOPT_NOP)
273 			optlen = 1;
274 		else {
275 			optlen = cp[IPOPT_OLEN];
276 			if (optlen <= 0 || optlen > x)
277 				return 0; /* invalid or truncated */
278 		}
279 		switch (opt) {
280 
281 		default:
282 			break;
283 
284 		case IPOPT_LSRR:
285 			bits |= IP_FW_IPOPT_LSRR;
286 			break;
287 
288 		case IPOPT_SSRR:
289 			bits |= IP_FW_IPOPT_SSRR;
290 			break;
291 
292 		case IPOPT_RR:
293 			bits |= IP_FW_IPOPT_RR;
294 			break;
295 
296 		case IPOPT_TS:
297 			bits |= IP_FW_IPOPT_TS;
298 			break;
299 		}
300 	}
301 	return (flags_match(cmd, bits));
302 }
303 
304 static int
305 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
306 {
307 	int optlen, bits = 0;
308 	u_char *cp = (u_char *)(tcp + 1);
309 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
310 
311 	for (; x > 0; x -= optlen, cp += optlen) {
312 		int opt = cp[0];
313 		if (opt == TCPOPT_EOL)
314 			break;
315 		if (opt == TCPOPT_NOP)
316 			optlen = 1;
317 		else {
318 			optlen = cp[1];
319 			if (optlen <= 0)
320 				break;
321 		}
322 
323 		switch (opt) {
324 
325 		default:
326 			break;
327 
328 		case TCPOPT_MAXSEG:
329 			bits |= IP_FW_TCPOPT_MSS;
330 			break;
331 
332 		case TCPOPT_WINDOW:
333 			bits |= IP_FW_TCPOPT_WINDOW;
334 			break;
335 
336 		case TCPOPT_SACK_PERMITTED:
337 		case TCPOPT_SACK:
338 			bits |= IP_FW_TCPOPT_SACK;
339 			break;
340 
341 		case TCPOPT_TIMESTAMP:
342 			bits |= IP_FW_TCPOPT_TS;
343 			break;
344 
345 		}
346 	}
347 	return (flags_match(cmd, bits));
348 }
349 
350 static int
351 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
352 {
353 	if (ifp == NULL)	/* no iface with this packet, match fails */
354 		return 0;
355 	/* Check by name or by IP address */
356 	if (cmd->name[0] != '\0') { /* match by name */
357 		if (cmd->name[0] == '\1') /* use tablearg to match */
358 			return ipfw_lookup_table_extended(chain, cmd->p.glob,
359 				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
360 		/* Check name */
361 		if (cmd->p.glob) {
362 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
363 				return(1);
364 		} else {
365 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
366 				return(1);
367 		}
368 	} else {
369 #ifdef __FreeBSD__	/* and OSX too ? */
370 		struct ifaddr *ia;
371 
372 		if_addr_rlock(ifp);
373 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
374 			if (ia->ifa_addr->sa_family != AF_INET)
375 				continue;
376 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
377 			    (ia->ifa_addr))->sin_addr.s_addr) {
378 				if_addr_runlock(ifp);
379 				return(1);	/* match */
380 			}
381 		}
382 		if_addr_runlock(ifp);
383 #endif /* __FreeBSD__ */
384 	}
385 	return(0);	/* no match, fail ... */
386 }
387 
388 /*
389  * The verify_path function checks if a route to the src exists and
390  * if it is reachable via ifp (when provided).
391  *
392  * The 'verrevpath' option checks that the interface that an IP packet
393  * arrives on is the same interface that traffic destined for the
394  * packet's source address would be routed out of.
395  * The 'versrcreach' option just checks that the source address is
396  * reachable via any route (except default) in the routing table.
397  * These two are a measure to block forged packets. This is also
398  * commonly known as "anti-spoofing" or Unicast Reverse Path
399  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
400  * is purposely reminiscent of the Cisco IOS command,
401  *
402  *   ip verify unicast reverse-path
403  *   ip verify unicast source reachable-via any
404  *
405  * which implements the same functionality. But note that the syntax
406  * is misleading, and the check may be performed on all IP packets
407  * whether unicast, multicast, or broadcast.
408  */
409 static int
410 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
411 {
412 #ifndef __FreeBSD__
413 	return 0;
414 #else
415 	struct route ro;
416 	struct sockaddr_in *dst;
417 
418 	bzero(&ro, sizeof(ro));
419 
420 	dst = (struct sockaddr_in *)&(ro.ro_dst);
421 	dst->sin_family = AF_INET;
422 	dst->sin_len = sizeof(*dst);
423 	dst->sin_addr = src;
424 	in_rtalloc_ign(&ro, 0, fib);
425 
426 	if (ro.ro_rt == NULL)
427 		return 0;
428 
429 	/*
430 	 * If ifp is provided, check for equality with rtentry.
431 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
432 	 * in order to pass packets injected back by if_simloop():
433 	 * if useloopback == 1 routing entry (via lo0) for our own address
434 	 * may exist, so we need to handle routing assymetry.
435 	 */
436 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
437 		RTFREE(ro.ro_rt);
438 		return 0;
439 	}
440 
441 	/* if no ifp provided, check if rtentry is not default route */
442 	if (ifp == NULL &&
443 	     satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
444 		RTFREE(ro.ro_rt);
445 		return 0;
446 	}
447 
448 	/* or if this is a blackhole/reject route */
449 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
450 		RTFREE(ro.ro_rt);
451 		return 0;
452 	}
453 
454 	/* found valid route */
455 	RTFREE(ro.ro_rt);
456 	return 1;
457 #endif /* __FreeBSD__ */
458 }
459 
460 #ifdef INET6
461 /*
462  * ipv6 specific rules here...
463  */
464 static __inline int
465 icmp6type_match (int type, ipfw_insn_u32 *cmd)
466 {
467 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
468 }
469 
470 static int
471 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
472 {
473 	int i;
474 	for (i=0; i <= cmd->o.arg1; ++i )
475 		if (curr_flow == cmd->d[i] )
476 			return 1;
477 	return 0;
478 }
479 
480 /* support for IP6_*_ME opcodes */
481 static int
482 search_ip6_addr_net (struct in6_addr * ip6_addr)
483 {
484 	struct ifnet *mdc;
485 	struct ifaddr *mdc2;
486 	struct in6_ifaddr *fdm;
487 	struct in6_addr copia;
488 
489 	TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
490 		if_addr_rlock(mdc);
491 		TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
492 			if (mdc2->ifa_addr->sa_family == AF_INET6) {
493 				fdm = (struct in6_ifaddr *)mdc2;
494 				copia = fdm->ia_addr.sin6_addr;
495 				/* need for leaving scope_id in the sock_addr */
496 				in6_clearscope(&copia);
497 				if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
498 					if_addr_runlock(mdc);
499 					return 1;
500 				}
501 			}
502 		}
503 		if_addr_runlock(mdc);
504 	}
505 	return 0;
506 }
507 
508 static int
509 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
510 {
511 	struct route_in6 ro;
512 	struct sockaddr_in6 *dst;
513 
514 	bzero(&ro, sizeof(ro));
515 
516 	dst = (struct sockaddr_in6 * )&(ro.ro_dst);
517 	dst->sin6_family = AF_INET6;
518 	dst->sin6_len = sizeof(*dst);
519 	dst->sin6_addr = *src;
520 
521 	in6_rtalloc_ign(&ro, 0, fib);
522 	if (ro.ro_rt == NULL)
523 		return 0;
524 
525 	/*
526 	 * if ifp is provided, check for equality with rtentry
527 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
528 	 * to support the case of sending packets to an address of our own.
529 	 * (where the former interface is the first argument of if_simloop()
530 	 *  (=ifp), the latter is lo0)
531 	 */
532 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
533 		RTFREE(ro.ro_rt);
534 		return 0;
535 	}
536 
537 	/* if no ifp provided, check if rtentry is not default route */
538 	if (ifp == NULL &&
539 	    IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
540 		RTFREE(ro.ro_rt);
541 		return 0;
542 	}
543 
544 	/* or if this is a blackhole/reject route */
545 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
546 		RTFREE(ro.ro_rt);
547 		return 0;
548 	}
549 
550 	/* found valid route */
551 	RTFREE(ro.ro_rt);
552 	return 1;
553 
554 }
555 
556 static int
557 is_icmp6_query(int icmp6_type)
558 {
559 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
560 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
561 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
562 	    icmp6_type == ICMP6_WRUREQUEST ||
563 	    icmp6_type == ICMP6_FQDN_QUERY ||
564 	    icmp6_type == ICMP6_NI_QUERY))
565 		return (1);
566 
567 	return (0);
568 }
569 
570 static void
571 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
572 {
573 	struct mbuf *m;
574 
575 	m = args->m;
576 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
577 		struct tcphdr *tcp;
578 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
579 
580 		if ((tcp->th_flags & TH_RST) == 0) {
581 			struct mbuf *m0;
582 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
583 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
584 			    tcp->th_flags | TH_RST);
585 			if (m0 != NULL)
586 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
587 				    NULL);
588 		}
589 		FREE_PKT(m);
590 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
591 #if 0
592 		/*
593 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
594 		 * as the contents are read. We need to m_adj() the
595 		 * needed amount.
596 		 * The mbuf will however be thrown away so we can adjust it.
597 		 * Remember we did an m_pullup on it already so we
598 		 * can make some assumptions about contiguousness.
599 		 */
600 		if (args->L3offset)
601 			m_adj(m, args->L3offset);
602 #endif
603 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
604 	} else
605 		FREE_PKT(m);
606 
607 	args->m = NULL;
608 }
609 
610 #endif /* INET6 */
611 
612 
613 /*
614  * sends a reject message, consuming the mbuf passed as an argument.
615  */
616 static void
617 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
618 {
619 
620 #if 0
621 	/* XXX When ip is not guaranteed to be at mtod() we will
622 	 * need to account for this */
623 	 * The mbuf will however be thrown away so we can adjust it.
624 	 * Remember we did an m_pullup on it already so we
625 	 * can make some assumptions about contiguousness.
626 	 */
627 	if (args->L3offset)
628 		m_adj(m, args->L3offset);
629 #endif
630 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
631 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
632 	} else if (args->f_id.proto == IPPROTO_TCP) {
633 		struct tcphdr *const tcp =
634 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
635 		if ( (tcp->th_flags & TH_RST) == 0) {
636 			struct mbuf *m;
637 			m = ipfw_send_pkt(args->m, &(args->f_id),
638 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
639 				tcp->th_flags | TH_RST);
640 			if (m != NULL)
641 				ip_output(m, NULL, NULL, 0, NULL, NULL);
642 		}
643 		FREE_PKT(args->m);
644 	} else
645 		FREE_PKT(args->m);
646 	args->m = NULL;
647 }
648 
649 /*
650  * Support for uid/gid/jail lookup. These tests are expensive
651  * (because we may need to look into the list of active sockets)
652  * so we cache the results. ugid_lookupp is 0 if we have not
653  * yet done a lookup, 1 if we succeeded, and -1 if we tried
654  * and failed. The function always returns the match value.
655  * We could actually spare the variable and use *uc, setting
656  * it to '(void *)check_uidgid if we have no info, NULL if
657  * we tried and failed, or any other value if successful.
658  */
659 static int
660 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
661     struct ucred **uc)
662 {
663 #ifndef __FreeBSD__
664 	/* XXX */
665 	return cred_check(insn, proto, oif,
666 	    dst_ip, dst_port, src_ip, src_port,
667 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
668 #else  /* FreeBSD */
669 	struct in_addr src_ip, dst_ip;
670 	struct inpcbinfo *pi;
671 	struct ipfw_flow_id *id;
672 	struct inpcb *pcb, *inp;
673 	struct ifnet *oif;
674 	int lookupflags;
675 	int match;
676 
677 	id = &args->f_id;
678 	inp = args->inp;
679 	oif = args->oif;
680 
681 	/*
682 	 * Check to see if the UDP or TCP stack supplied us with
683 	 * the PCB. If so, rather then holding a lock and looking
684 	 * up the PCB, we can use the one that was supplied.
685 	 */
686 	if (inp && *ugid_lookupp == 0) {
687 		INP_LOCK_ASSERT(inp);
688 		if (inp->inp_socket != NULL) {
689 			*uc = crhold(inp->inp_cred);
690 			*ugid_lookupp = 1;
691 		} else
692 			*ugid_lookupp = -1;
693 	}
694 	/*
695 	 * If we have already been here and the packet has no
696 	 * PCB entry associated with it, then we can safely
697 	 * assume that this is a no match.
698 	 */
699 	if (*ugid_lookupp == -1)
700 		return (0);
701 	if (id->proto == IPPROTO_TCP) {
702 		lookupflags = 0;
703 		pi = &V_tcbinfo;
704 	} else if (id->proto == IPPROTO_UDP) {
705 		lookupflags = INPLOOKUP_WILDCARD;
706 		pi = &V_udbinfo;
707 	} else
708 		return 0;
709 	lookupflags |= INPLOOKUP_RLOCKPCB;
710 	match = 0;
711 	if (*ugid_lookupp == 0) {
712 		if (id->addr_type == 6) {
713 #ifdef INET6
714 			if (oif == NULL)
715 				pcb = in6_pcblookup_mbuf(pi,
716 				    &id->src_ip6, htons(id->src_port),
717 				    &id->dst_ip6, htons(id->dst_port),
718 				    lookupflags, oif, args->m);
719 			else
720 				pcb = in6_pcblookup_mbuf(pi,
721 				    &id->dst_ip6, htons(id->dst_port),
722 				    &id->src_ip6, htons(id->src_port),
723 				    lookupflags, oif, args->m);
724 #else
725 			*ugid_lookupp = -1;
726 			return (0);
727 #endif
728 		} else {
729 			src_ip.s_addr = htonl(id->src_ip);
730 			dst_ip.s_addr = htonl(id->dst_ip);
731 			if (oif == NULL)
732 				pcb = in_pcblookup_mbuf(pi,
733 				    src_ip, htons(id->src_port),
734 				    dst_ip, htons(id->dst_port),
735 				    lookupflags, oif, args->m);
736 			else
737 				pcb = in_pcblookup_mbuf(pi,
738 				    dst_ip, htons(id->dst_port),
739 				    src_ip, htons(id->src_port),
740 				    lookupflags, oif, args->m);
741 		}
742 		if (pcb != NULL) {
743 			INP_RLOCK_ASSERT(pcb);
744 			*uc = crhold(pcb->inp_cred);
745 			*ugid_lookupp = 1;
746 			INP_RUNLOCK(pcb);
747 		}
748 		if (*ugid_lookupp == 0) {
749 			/*
750 			 * We tried and failed, set the variable to -1
751 			 * so we will not try again on this packet.
752 			 */
753 			*ugid_lookupp = -1;
754 			return (0);
755 		}
756 	}
757 	if (insn->o.opcode == O_UID)
758 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
759 	else if (insn->o.opcode == O_GID)
760 		match = groupmember((gid_t)insn->d[0], *uc);
761 	else if (insn->o.opcode == O_JAIL)
762 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
763 	return (match);
764 #endif /* __FreeBSD__ */
765 }
766 
767 /*
768  * Helper function to set args with info on the rule after the matching
769  * one. slot is precise, whereas we guess rule_id as they are
770  * assigned sequentially.
771  */
772 static inline void
773 set_match(struct ip_fw_args *args, int slot,
774 	struct ip_fw_chain *chain)
775 {
776 	args->rule.chain_id = chain->id;
777 	args->rule.slot = slot + 1; /* we use 0 as a marker */
778 	args->rule.rule_id = 1 + chain->map[slot]->id;
779 	args->rule.rulenum = chain->map[slot]->rulenum;
780 }
781 
782 /*
783  * Helper function to enable cached rule lookups using
784  * x_next and next_rule fields in ipfw rule.
785  */
786 static int
787 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
788     int tablearg, int jump_backwards)
789 {
790 	int f_pos;
791 
792 	/* If possible use cached f_pos (in f->next_rule),
793 	 * whose version is written in f->next_rule
794 	 * (horrible hacks to avoid changing the ABI).
795 	 */
796 	if (num != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id)
797 		f_pos = (uintptr_t)f->next_rule;
798 	else {
799 		int i = IP_FW_ARG_TABLEARG(num);
800 		/* make sure we do not jump backward */
801 		if (jump_backwards == 0 && i <= f->rulenum)
802 			i = f->rulenum + 1;
803 		f_pos = ipfw_find_rule(chain, i, 0);
804 		/* update the cache */
805 		if (num != IP_FW_TABLEARG) {
806 			f->next_rule = (void *)(uintptr_t)f_pos;
807 			f->x_next = (void *)(uintptr_t)chain->id;
808 		}
809 	}
810 
811 	return (f_pos);
812 }
813 
814 /*
815  * The main check routine for the firewall.
816  *
817  * All arguments are in args so we can modify them and return them
818  * back to the caller.
819  *
820  * Parameters:
821  *
822  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
823  *		Starts with the IP header.
824  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
825  *	args->L3offset	Number of bytes bypassed if we came from L2.
826  *			e.g. often sizeof(eh)  ** NOTYET **
827  *	args->oif	Outgoing interface, NULL if packet is incoming.
828  *		The incoming interface is in the mbuf. (in)
829  *	args->divert_rule (in/out)
830  *		Skip up to the first rule past this rule number;
831  *		upon return, non-zero port number for divert or tee.
832  *
833  *	args->rule	Pointer to the last matching rule (in/out)
834  *	args->next_hop	Socket we are forwarding to (out).
835  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
836  *	args->f_id	Addresses grabbed from the packet (out)
837  * 	args->rule.info	a cookie depending on rule action
838  *
839  * Return value:
840  *
841  *	IP_FW_PASS	the packet must be accepted
842  *	IP_FW_DENY	the packet must be dropped
843  *	IP_FW_DIVERT	divert packet, port in m_tag
844  *	IP_FW_TEE	tee packet, port in m_tag
845  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
846  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
847  *		args->rule contains the matching rule,
848  *		args->rule.info has additional information.
849  *
850  */
851 int
852 ipfw_chk(struct ip_fw_args *args)
853 {
854 
855 	/*
856 	 * Local variables holding state while processing a packet:
857 	 *
858 	 * IMPORTANT NOTE: to speed up the processing of rules, there
859 	 * are some assumption on the values of the variables, which
860 	 * are documented here. Should you change them, please check
861 	 * the implementation of the various instructions to make sure
862 	 * that they still work.
863 	 *
864 	 * args->eh	The MAC header. It is non-null for a layer2
865 	 *	packet, it is NULL for a layer-3 packet.
866 	 * **notyet**
867 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
868 	 *
869 	 * m | args->m	Pointer to the mbuf, as received from the caller.
870 	 *	It may change if ipfw_chk() does an m_pullup, or if it
871 	 *	consumes the packet because it calls send_reject().
872 	 *	XXX This has to change, so that ipfw_chk() never modifies
873 	 *	or consumes the buffer.
874 	 * ip	is the beginning of the ip(4 or 6) header.
875 	 *	Calculated by adding the L3offset to the start of data.
876 	 *	(Until we start using L3offset, the packet is
877 	 *	supposed to start with the ip header).
878 	 */
879 	struct mbuf *m = args->m;
880 	struct ip *ip = mtod(m, struct ip *);
881 
882 	/*
883 	 * For rules which contain uid/gid or jail constraints, cache
884 	 * a copy of the users credentials after the pcb lookup has been
885 	 * executed. This will speed up the processing of rules with
886 	 * these types of constraints, as well as decrease contention
887 	 * on pcb related locks.
888 	 */
889 #ifndef __FreeBSD__
890 	struct bsd_ucred ucred_cache;
891 #else
892 	struct ucred *ucred_cache = NULL;
893 #endif
894 	int ucred_lookup = 0;
895 
896 	/*
897 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
898 	 *	inbound path (ether_input, ip_input).
899 	 *	If non-NULL, ipfw_chk has been called on the outbound path
900 	 *	(ether_output, ip_output).
901 	 */
902 	struct ifnet *oif = args->oif;
903 
904 	int f_pos = 0;		/* index of current rule in the array */
905 	int retval = 0;
906 
907 	/*
908 	 * hlen	The length of the IP header.
909 	 */
910 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
911 
912 	/*
913 	 * offset	The offset of a fragment. offset != 0 means that
914 	 *	we have a fragment at this offset of an IPv4 packet.
915 	 *	offset == 0 means that (if this is an IPv4 packet)
916 	 *	this is the first or only fragment.
917 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
918 	 *	or there is a single packet fragement (fragement header added
919 	 *	without needed).  We will treat a single packet fragment as if
920 	 *	there was no fragment header (or log/block depending on the
921 	 *	V_fw_permit_single_frag6 sysctl setting).
922 	 */
923 	u_short offset = 0;
924 	u_short ip6f_mf = 0;
925 
926 	/*
927 	 * Local copies of addresses. They are only valid if we have
928 	 * an IP packet.
929 	 *
930 	 * proto	The protocol. Set to 0 for non-ip packets,
931 	 *	or to the protocol read from the packet otherwise.
932 	 *	proto != 0 means that we have an IPv4 packet.
933 	 *
934 	 * src_port, dst_port	port numbers, in HOST format. Only
935 	 *	valid for TCP and UDP packets.
936 	 *
937 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
938 	 *	Only valid for IPv4 packets.
939 	 */
940 	uint8_t proto;
941 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
942 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
943 	uint16_t iplen=0;
944 	int pktlen;
945 	uint16_t	etype = 0;	/* Host order stored ether type */
946 
947 	/*
948 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
949 	 * 	MATCH_NONE when checked and not matched (q = NULL),
950 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
951 	 */
952 	int dyn_dir = MATCH_UNKNOWN;
953 	ipfw_dyn_rule *q = NULL;
954 	struct ip_fw_chain *chain = &V_layer3_chain;
955 
956 	/*
957 	 * We store in ulp a pointer to the upper layer protocol header.
958 	 * In the ipv4 case this is easy to determine from the header,
959 	 * but for ipv6 we might have some additional headers in the middle.
960 	 * ulp is NULL if not found.
961 	 */
962 	void *ulp = NULL;		/* upper layer protocol pointer. */
963 
964 	/* XXX ipv6 variables */
965 	int is_ipv6 = 0;
966 	uint8_t	icmp6_type = 0;
967 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
968 	/* end of ipv6 variables */
969 
970 	int is_ipv4 = 0;
971 
972 	int done = 0;		/* flag to exit the outer loop */
973 
974 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
975 		return (IP_FW_PASS);	/* accept */
976 
977 	dst_ip.s_addr = 0;		/* make sure it is initialized */
978 	src_ip.s_addr = 0;		/* make sure it is initialized */
979 	pktlen = m->m_pkthdr.len;
980 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
981 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
982 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
983 
984 /*
985  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
986  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
987  * pointer might become stale after other pullups (but we never use it
988  * this way).
989  */
990 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
991 #define PULLUP_LEN(_len, p, T)					\
992 do {								\
993 	int x = (_len) + T;					\
994 	if ((m)->m_len < x) {					\
995 		args->m = m = m_pullup(m, x);			\
996 		if (m == NULL)					\
997 			goto pullup_failed;			\
998 	}							\
999 	p = (mtod(m, char *) + (_len));				\
1000 } while (0)
1001 
1002 	/*
1003 	 * if we have an ether header,
1004 	 */
1005 	if (args->eh)
1006 		etype = ntohs(args->eh->ether_type);
1007 
1008 	/* Identify IP packets and fill up variables. */
1009 	if (pktlen >= sizeof(struct ip6_hdr) &&
1010 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1011 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1012 		is_ipv6 = 1;
1013 		args->f_id.addr_type = 6;
1014 		hlen = sizeof(struct ip6_hdr);
1015 		proto = ip6->ip6_nxt;
1016 
1017 		/* Search extension headers to find upper layer protocols */
1018 		while (ulp == NULL && offset == 0) {
1019 			switch (proto) {
1020 			case IPPROTO_ICMPV6:
1021 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1022 				icmp6_type = ICMP6(ulp)->icmp6_type;
1023 				break;
1024 
1025 			case IPPROTO_TCP:
1026 				PULLUP_TO(hlen, ulp, struct tcphdr);
1027 				dst_port = TCP(ulp)->th_dport;
1028 				src_port = TCP(ulp)->th_sport;
1029 				/* save flags for dynamic rules */
1030 				args->f_id._flags = TCP(ulp)->th_flags;
1031 				break;
1032 
1033 			case IPPROTO_SCTP:
1034 				PULLUP_TO(hlen, ulp, struct sctphdr);
1035 				src_port = SCTP(ulp)->src_port;
1036 				dst_port = SCTP(ulp)->dest_port;
1037 				break;
1038 
1039 			case IPPROTO_UDP:
1040 				PULLUP_TO(hlen, ulp, struct udphdr);
1041 				dst_port = UDP(ulp)->uh_dport;
1042 				src_port = UDP(ulp)->uh_sport;
1043 				break;
1044 
1045 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1046 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1047 				ext_hd |= EXT_HOPOPTS;
1048 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1049 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1050 				ulp = NULL;
1051 				break;
1052 
1053 			case IPPROTO_ROUTING:	/* RFC 2460 */
1054 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1055 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1056 				case 0:
1057 					ext_hd |= EXT_RTHDR0;
1058 					break;
1059 				case 2:
1060 					ext_hd |= EXT_RTHDR2;
1061 					break;
1062 				default:
1063 					if (V_fw_verbose)
1064 						printf("IPFW2: IPV6 - Unknown "
1065 						    "Routing Header type(%d)\n",
1066 						    ((struct ip6_rthdr *)
1067 						    ulp)->ip6r_type);
1068 					if (V_fw_deny_unknown_exthdrs)
1069 					    return (IP_FW_DENY);
1070 					break;
1071 				}
1072 				ext_hd |= EXT_ROUTING;
1073 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1074 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1075 				ulp = NULL;
1076 				break;
1077 
1078 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1079 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1080 				ext_hd |= EXT_FRAGMENT;
1081 				hlen += sizeof (struct ip6_frag);
1082 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1083 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1084 					IP6F_OFF_MASK;
1085 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1086 					IP6F_MORE_FRAG;
1087 				if (V_fw_permit_single_frag6 == 0 &&
1088 				    offset == 0 && ip6f_mf == 0) {
1089 					if (V_fw_verbose)
1090 						printf("IPFW2: IPV6 - Invalid "
1091 						    "Fragment Header\n");
1092 					if (V_fw_deny_unknown_exthdrs)
1093 					    return (IP_FW_DENY);
1094 					break;
1095 				}
1096 				args->f_id.extra =
1097 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1098 				ulp = NULL;
1099 				break;
1100 
1101 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1102 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1103 				ext_hd |= EXT_DSTOPTS;
1104 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1105 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1106 				ulp = NULL;
1107 				break;
1108 
1109 			case IPPROTO_AH:	/* RFC 2402 */
1110 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1111 				ext_hd |= EXT_AH;
1112 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1113 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1114 				ulp = NULL;
1115 				break;
1116 
1117 			case IPPROTO_ESP:	/* RFC 2406 */
1118 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1119 				/* Anything past Seq# is variable length and
1120 				 * data past this ext. header is encrypted. */
1121 				ext_hd |= EXT_ESP;
1122 				break;
1123 
1124 			case IPPROTO_NONE:	/* RFC 2460 */
1125 				/*
1126 				 * Packet ends here, and IPv6 header has
1127 				 * already been pulled up. If ip6e_len!=0
1128 				 * then octets must be ignored.
1129 				 */
1130 				ulp = ip; /* non-NULL to get out of loop. */
1131 				break;
1132 
1133 			case IPPROTO_OSPFIGP:
1134 				/* XXX OSPF header check? */
1135 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1136 				break;
1137 
1138 			case IPPROTO_PIM:
1139 				/* XXX PIM header check? */
1140 				PULLUP_TO(hlen, ulp, struct pim);
1141 				break;
1142 
1143 			case IPPROTO_CARP:
1144 				PULLUP_TO(hlen, ulp, struct carp_header);
1145 				if (((struct carp_header *)ulp)->carp_version !=
1146 				    CARP_VERSION)
1147 					return (IP_FW_DENY);
1148 				if (((struct carp_header *)ulp)->carp_type !=
1149 				    CARP_ADVERTISEMENT)
1150 					return (IP_FW_DENY);
1151 				break;
1152 
1153 			case IPPROTO_IPV6:	/* RFC 2893 */
1154 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1155 				break;
1156 
1157 			case IPPROTO_IPV4:	/* RFC 2893 */
1158 				PULLUP_TO(hlen, ulp, struct ip);
1159 				break;
1160 
1161 			default:
1162 				if (V_fw_verbose)
1163 					printf("IPFW2: IPV6 - Unknown "
1164 					    "Extension Header(%d), ext_hd=%x\n",
1165 					     proto, ext_hd);
1166 				if (V_fw_deny_unknown_exthdrs)
1167 				    return (IP_FW_DENY);
1168 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1169 				break;
1170 			} /*switch */
1171 		}
1172 		ip = mtod(m, struct ip *);
1173 		ip6 = (struct ip6_hdr *)ip;
1174 		args->f_id.src_ip6 = ip6->ip6_src;
1175 		args->f_id.dst_ip6 = ip6->ip6_dst;
1176 		args->f_id.src_ip = 0;
1177 		args->f_id.dst_ip = 0;
1178 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1179 	} else if (pktlen >= sizeof(struct ip) &&
1180 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1181 	    	is_ipv4 = 1;
1182 		hlen = ip->ip_hl << 2;
1183 		args->f_id.addr_type = 4;
1184 
1185 		/*
1186 		 * Collect parameters into local variables for faster matching.
1187 		 */
1188 		proto = ip->ip_p;
1189 		src_ip = ip->ip_src;
1190 		dst_ip = ip->ip_dst;
1191 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1192 		iplen = ntohs(ip->ip_len);
1193 		pktlen = iplen < pktlen ? iplen : pktlen;
1194 
1195 		if (offset == 0) {
1196 			switch (proto) {
1197 			case IPPROTO_TCP:
1198 				PULLUP_TO(hlen, ulp, struct tcphdr);
1199 				dst_port = TCP(ulp)->th_dport;
1200 				src_port = TCP(ulp)->th_sport;
1201 				/* save flags for dynamic rules */
1202 				args->f_id._flags = TCP(ulp)->th_flags;
1203 				break;
1204 
1205 			case IPPROTO_SCTP:
1206 				PULLUP_TO(hlen, ulp, struct sctphdr);
1207 				src_port = SCTP(ulp)->src_port;
1208 				dst_port = SCTP(ulp)->dest_port;
1209 				break;
1210 
1211 			case IPPROTO_UDP:
1212 				PULLUP_TO(hlen, ulp, struct udphdr);
1213 				dst_port = UDP(ulp)->uh_dport;
1214 				src_port = UDP(ulp)->uh_sport;
1215 				break;
1216 
1217 			case IPPROTO_ICMP:
1218 				PULLUP_TO(hlen, ulp, struct icmphdr);
1219 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1220 				break;
1221 
1222 			default:
1223 				break;
1224 			}
1225 		}
1226 
1227 		ip = mtod(m, struct ip *);
1228 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1229 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1230 	}
1231 #undef PULLUP_TO
1232 	if (proto) { /* we may have port numbers, store them */
1233 		args->f_id.proto = proto;
1234 		args->f_id.src_port = src_port = ntohs(src_port);
1235 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1236 	}
1237 
1238 	IPFW_PF_RLOCK(chain);
1239 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1240 		IPFW_PF_RUNLOCK(chain);
1241 		return (IP_FW_PASS);	/* accept */
1242 	}
1243 	if (args->rule.slot) {
1244 		/*
1245 		 * Packet has already been tagged as a result of a previous
1246 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1247 		 * REASS, NETGRAPH, DIVERT/TEE...)
1248 		 * Validate the slot and continue from the next one
1249 		 * if still present, otherwise do a lookup.
1250 		 */
1251 		f_pos = (args->rule.chain_id == chain->id) ?
1252 		    args->rule.slot :
1253 		    ipfw_find_rule(chain, args->rule.rulenum,
1254 			args->rule.rule_id);
1255 	} else {
1256 		f_pos = 0;
1257 	}
1258 
1259 	/*
1260 	 * Now scan the rules, and parse microinstructions for each rule.
1261 	 * We have two nested loops and an inner switch. Sometimes we
1262 	 * need to break out of one or both loops, or re-enter one of
1263 	 * the loops with updated variables. Loop variables are:
1264 	 *
1265 	 *	f_pos (outer loop) points to the current rule.
1266 	 *		On output it points to the matching rule.
1267 	 *	done (outer loop) is used as a flag to break the loop.
1268 	 *	l (inner loop)	residual length of current rule.
1269 	 *		cmd points to the current microinstruction.
1270 	 *
1271 	 * We break the inner loop by setting l=0 and possibly
1272 	 * cmdlen=0 if we don't want to advance cmd.
1273 	 * We break the outer loop by setting done=1
1274 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1275 	 * as needed.
1276 	 */
1277 	for (; f_pos < chain->n_rules; f_pos++) {
1278 		ipfw_insn *cmd;
1279 		uint32_t tablearg = 0;
1280 		int l, cmdlen, skip_or; /* skip rest of OR block */
1281 		struct ip_fw *f;
1282 
1283 		f = chain->map[f_pos];
1284 		if (V_set_disable & (1 << f->set) )
1285 			continue;
1286 
1287 		skip_or = 0;
1288 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1289 		    l -= cmdlen, cmd += cmdlen) {
1290 			int match;
1291 
1292 			/*
1293 			 * check_body is a jump target used when we find a
1294 			 * CHECK_STATE, and need to jump to the body of
1295 			 * the target rule.
1296 			 */
1297 
1298 /* check_body: */
1299 			cmdlen = F_LEN(cmd);
1300 			/*
1301 			 * An OR block (insn_1 || .. || insn_n) has the
1302 			 * F_OR bit set in all but the last instruction.
1303 			 * The first match will set "skip_or", and cause
1304 			 * the following instructions to be skipped until
1305 			 * past the one with the F_OR bit clear.
1306 			 */
1307 			if (skip_or) {		/* skip this instruction */
1308 				if ((cmd->len & F_OR) == 0)
1309 					skip_or = 0;	/* next one is good */
1310 				continue;
1311 			}
1312 			match = 0; /* set to 1 if we succeed */
1313 
1314 			switch (cmd->opcode) {
1315 			/*
1316 			 * The first set of opcodes compares the packet's
1317 			 * fields with some pattern, setting 'match' if a
1318 			 * match is found. At the end of the loop there is
1319 			 * logic to deal with F_NOT and F_OR flags associated
1320 			 * with the opcode.
1321 			 */
1322 			case O_NOP:
1323 				match = 1;
1324 				break;
1325 
1326 			case O_FORWARD_MAC:
1327 				printf("ipfw: opcode %d unimplemented\n",
1328 				    cmd->opcode);
1329 				break;
1330 
1331 			case O_GID:
1332 			case O_UID:
1333 			case O_JAIL:
1334 				/*
1335 				 * We only check offset == 0 && proto != 0,
1336 				 * as this ensures that we have a
1337 				 * packet with the ports info.
1338 				 */
1339 				if (offset != 0)
1340 					break;
1341 				if (proto == IPPROTO_TCP ||
1342 				    proto == IPPROTO_UDP)
1343 					match = check_uidgid(
1344 						    (ipfw_insn_u32 *)cmd,
1345 						    args, &ucred_lookup,
1346 #ifdef __FreeBSD__
1347 						    &ucred_cache);
1348 #else
1349 						    (void *)&ucred_cache);
1350 #endif
1351 				break;
1352 
1353 			case O_RECV:
1354 				match = iface_match(m->m_pkthdr.rcvif,
1355 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1356 				break;
1357 
1358 			case O_XMIT:
1359 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1360 				    chain, &tablearg);
1361 				break;
1362 
1363 			case O_VIA:
1364 				match = iface_match(oif ? oif :
1365 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1366 				    chain, &tablearg);
1367 				break;
1368 
1369 			case O_MACADDR2:
1370 				if (args->eh != NULL) {	/* have MAC header */
1371 					u_int32_t *want = (u_int32_t *)
1372 						((ipfw_insn_mac *)cmd)->addr;
1373 					u_int32_t *mask = (u_int32_t *)
1374 						((ipfw_insn_mac *)cmd)->mask;
1375 					u_int32_t *hdr = (u_int32_t *)args->eh;
1376 
1377 					match =
1378 					    ( want[0] == (hdr[0] & mask[0]) &&
1379 					      want[1] == (hdr[1] & mask[1]) &&
1380 					      want[2] == (hdr[2] & mask[2]) );
1381 				}
1382 				break;
1383 
1384 			case O_MAC_TYPE:
1385 				if (args->eh != NULL) {
1386 					u_int16_t *p =
1387 					    ((ipfw_insn_u16 *)cmd)->ports;
1388 					int i;
1389 
1390 					for (i = cmdlen - 1; !match && i>0;
1391 					    i--, p += 2)
1392 						match = (etype >= p[0] &&
1393 						    etype <= p[1]);
1394 				}
1395 				break;
1396 
1397 			case O_FRAG:
1398 				match = (offset != 0);
1399 				break;
1400 
1401 			case O_IN:	/* "out" is "not in" */
1402 				match = (oif == NULL);
1403 				break;
1404 
1405 			case O_LAYER2:
1406 				match = (args->eh != NULL);
1407 				break;
1408 
1409 			case O_DIVERTED:
1410 			    {
1411 				/* For diverted packets, args->rule.info
1412 				 * contains the divert port (in host format)
1413 				 * reason and direction.
1414 				 */
1415 				uint32_t i = args->rule.info;
1416 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1417 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1418 			    }
1419 				break;
1420 
1421 			case O_PROTO:
1422 				/*
1423 				 * We do not allow an arg of 0 so the
1424 				 * check of "proto" only suffices.
1425 				 */
1426 				match = (proto == cmd->arg1);
1427 				break;
1428 
1429 			case O_IP_SRC:
1430 				match = is_ipv4 &&
1431 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1432 				    src_ip.s_addr);
1433 				break;
1434 
1435 			case O_IP_SRC_LOOKUP:
1436 			case O_IP_DST_LOOKUP:
1437 				if (is_ipv4) {
1438 				    uint32_t key =
1439 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1440 					    dst_ip.s_addr : src_ip.s_addr;
1441 				    uint32_t v = 0;
1442 
1443 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1444 					/* generic lookup. The key must be
1445 					 * in 32bit big-endian format.
1446 					 */
1447 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1448 					if (v == 0)
1449 					    key = dst_ip.s_addr;
1450 					else if (v == 1)
1451 					    key = src_ip.s_addr;
1452 					else if (v == 6) /* dscp */
1453 					    key = (ip->ip_tos >> 2) & 0x3f;
1454 					else if (offset != 0)
1455 					    break;
1456 					else if (proto != IPPROTO_TCP &&
1457 						proto != IPPROTO_UDP)
1458 					    break;
1459 					else if (v == 2)
1460 					    key = htonl(dst_port);
1461 					else if (v == 3)
1462 					    key = htonl(src_port);
1463 					else if (v == 4 || v == 5) {
1464 					    check_uidgid(
1465 						(ipfw_insn_u32 *)cmd,
1466 						args, &ucred_lookup,
1467 #ifdef __FreeBSD__
1468 						&ucred_cache);
1469 					    if (v == 4 /* O_UID */)
1470 						key = ucred_cache->cr_uid;
1471 					    else if (v == 5 /* O_JAIL */)
1472 						key = ucred_cache->cr_prison->pr_id;
1473 #else /* !__FreeBSD__ */
1474 						(void *)&ucred_cache);
1475 					    if (v ==4 /* O_UID */)
1476 						key = ucred_cache.uid;
1477 					    else if (v == 5 /* O_JAIL */)
1478 						key = ucred_cache.xid;
1479 #endif /* !__FreeBSD__ */
1480 					    key = htonl(key);
1481 					} else
1482 					    break;
1483 				    }
1484 				    match = ipfw_lookup_table(chain,
1485 					cmd->arg1, key, &v);
1486 				    if (!match)
1487 					break;
1488 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1489 					match =
1490 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1491 				    else
1492 					tablearg = v;
1493 				} else if (is_ipv6) {
1494 					uint32_t v = 0;
1495 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1496 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1497 					match = ipfw_lookup_table_extended(chain,
1498 							cmd->arg1, pkey, &v,
1499 							IPFW_TABLE_CIDR);
1500 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1501 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1502 					if (match)
1503 						tablearg = v;
1504 				}
1505 				break;
1506 
1507 			case O_IP_SRC_MASK:
1508 			case O_IP_DST_MASK:
1509 				if (is_ipv4) {
1510 				    uint32_t a =
1511 					(cmd->opcode == O_IP_DST_MASK) ?
1512 					    dst_ip.s_addr : src_ip.s_addr;
1513 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1514 				    int i = cmdlen-1;
1515 
1516 				    for (; !match && i>0; i-= 2, p+= 2)
1517 					match = (p[0] == (a & p[1]));
1518 				}
1519 				break;
1520 
1521 			case O_IP_SRC_ME:
1522 				if (is_ipv4) {
1523 					struct ifnet *tif;
1524 
1525 					INADDR_TO_IFP(src_ip, tif);
1526 					match = (tif != NULL);
1527 					break;
1528 				}
1529 #ifdef INET6
1530 				/* FALLTHROUGH */
1531 			case O_IP6_SRC_ME:
1532 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1533 #endif
1534 				break;
1535 
1536 			case O_IP_DST_SET:
1537 			case O_IP_SRC_SET:
1538 				if (is_ipv4) {
1539 					u_int32_t *d = (u_int32_t *)(cmd+1);
1540 					u_int32_t addr =
1541 					    cmd->opcode == O_IP_DST_SET ?
1542 						args->f_id.dst_ip :
1543 						args->f_id.src_ip;
1544 
1545 					    if (addr < d[0])
1546 						    break;
1547 					    addr -= d[0]; /* subtract base */
1548 					    match = (addr < cmd->arg1) &&
1549 						( d[ 1 + (addr>>5)] &
1550 						  (1<<(addr & 0x1f)) );
1551 				}
1552 				break;
1553 
1554 			case O_IP_DST:
1555 				match = is_ipv4 &&
1556 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1557 				    dst_ip.s_addr);
1558 				break;
1559 
1560 			case O_IP_DST_ME:
1561 				if (is_ipv4) {
1562 					struct ifnet *tif;
1563 
1564 					INADDR_TO_IFP(dst_ip, tif);
1565 					match = (tif != NULL);
1566 					break;
1567 				}
1568 #ifdef INET6
1569 				/* FALLTHROUGH */
1570 			case O_IP6_DST_ME:
1571 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1572 #endif
1573 				break;
1574 
1575 
1576 			case O_IP_SRCPORT:
1577 			case O_IP_DSTPORT:
1578 				/*
1579 				 * offset == 0 && proto != 0 is enough
1580 				 * to guarantee that we have a
1581 				 * packet with port info.
1582 				 */
1583 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1584 				    && offset == 0) {
1585 					u_int16_t x =
1586 					    (cmd->opcode == O_IP_SRCPORT) ?
1587 						src_port : dst_port ;
1588 					u_int16_t *p =
1589 					    ((ipfw_insn_u16 *)cmd)->ports;
1590 					int i;
1591 
1592 					for (i = cmdlen - 1; !match && i>0;
1593 					    i--, p += 2)
1594 						match = (x>=p[0] && x<=p[1]);
1595 				}
1596 				break;
1597 
1598 			case O_ICMPTYPE:
1599 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1600 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1601 				break;
1602 
1603 #ifdef INET6
1604 			case O_ICMP6TYPE:
1605 				match = is_ipv6 && offset == 0 &&
1606 				    proto==IPPROTO_ICMPV6 &&
1607 				    icmp6type_match(
1608 					ICMP6(ulp)->icmp6_type,
1609 					(ipfw_insn_u32 *)cmd);
1610 				break;
1611 #endif /* INET6 */
1612 
1613 			case O_IPOPT:
1614 				match = (is_ipv4 &&
1615 				    ipopts_match(ip, cmd) );
1616 				break;
1617 
1618 			case O_IPVER:
1619 				match = (is_ipv4 &&
1620 				    cmd->arg1 == ip->ip_v);
1621 				break;
1622 
1623 			case O_IPID:
1624 			case O_IPLEN:
1625 			case O_IPTTL:
1626 				if (is_ipv4) {	/* only for IP packets */
1627 				    uint16_t x;
1628 				    uint16_t *p;
1629 				    int i;
1630 
1631 				    if (cmd->opcode == O_IPLEN)
1632 					x = iplen;
1633 				    else if (cmd->opcode == O_IPTTL)
1634 					x = ip->ip_ttl;
1635 				    else /* must be IPID */
1636 					x = ntohs(ip->ip_id);
1637 				    if (cmdlen == 1) {
1638 					match = (cmd->arg1 == x);
1639 					break;
1640 				    }
1641 				    /* otherwise we have ranges */
1642 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1643 				    i = cmdlen - 1;
1644 				    for (; !match && i>0; i--, p += 2)
1645 					match = (x >= p[0] && x <= p[1]);
1646 				}
1647 				break;
1648 
1649 			case O_IPPRECEDENCE:
1650 				match = (is_ipv4 &&
1651 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1652 				break;
1653 
1654 			case O_IPTOS:
1655 				match = (is_ipv4 &&
1656 				    flags_match(cmd, ip->ip_tos));
1657 				break;
1658 
1659 			case O_DSCP:
1660 			    {
1661 				uint32_t *p;
1662 				uint16_t x;
1663 
1664 				p = ((ipfw_insn_u32 *)cmd)->d;
1665 
1666 				if (is_ipv4)
1667 					x = ip->ip_tos >> 2;
1668 				else if (is_ipv6) {
1669 					uint8_t *v;
1670 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
1671 					x = (*v & 0x0F) << 2;
1672 					v++;
1673 					x |= *v >> 6;
1674 				} else
1675 					break;
1676 
1677 				/* DSCP bitmask is stored as low_u32 high_u32 */
1678 				if (x > 32)
1679 					match = *(p + 1) & (1 << (x - 32));
1680 				else
1681 					match = *p & (1 << x);
1682 			    }
1683 				break;
1684 
1685 			case O_TCPDATALEN:
1686 				if (proto == IPPROTO_TCP && offset == 0) {
1687 				    struct tcphdr *tcp;
1688 				    uint16_t x;
1689 				    uint16_t *p;
1690 				    int i;
1691 
1692 				    tcp = TCP(ulp);
1693 				    x = iplen -
1694 					((ip->ip_hl + tcp->th_off) << 2);
1695 				    if (cmdlen == 1) {
1696 					match = (cmd->arg1 == x);
1697 					break;
1698 				    }
1699 				    /* otherwise we have ranges */
1700 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1701 				    i = cmdlen - 1;
1702 				    for (; !match && i>0; i--, p += 2)
1703 					match = (x >= p[0] && x <= p[1]);
1704 				}
1705 				break;
1706 
1707 			case O_TCPFLAGS:
1708 				match = (proto == IPPROTO_TCP && offset == 0 &&
1709 				    flags_match(cmd, TCP(ulp)->th_flags));
1710 				break;
1711 
1712 			case O_TCPOPTS:
1713 				PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1714 				match = (proto == IPPROTO_TCP && offset == 0 &&
1715 				    tcpopts_match(TCP(ulp), cmd));
1716 				break;
1717 
1718 			case O_TCPSEQ:
1719 				match = (proto == IPPROTO_TCP && offset == 0 &&
1720 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1721 					TCP(ulp)->th_seq);
1722 				break;
1723 
1724 			case O_TCPACK:
1725 				match = (proto == IPPROTO_TCP && offset == 0 &&
1726 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1727 					TCP(ulp)->th_ack);
1728 				break;
1729 
1730 			case O_TCPWIN:
1731 				if (proto == IPPROTO_TCP && offset == 0) {
1732 				    uint16_t x;
1733 				    uint16_t *p;
1734 				    int i;
1735 
1736 				    x = ntohs(TCP(ulp)->th_win);
1737 				    if (cmdlen == 1) {
1738 					match = (cmd->arg1 == x);
1739 					break;
1740 				    }
1741 				    /* Otherwise we have ranges. */
1742 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1743 				    i = cmdlen - 1;
1744 				    for (; !match && i > 0; i--, p += 2)
1745 					match = (x >= p[0] && x <= p[1]);
1746 				}
1747 				break;
1748 
1749 			case O_ESTAB:
1750 				/* reject packets which have SYN only */
1751 				/* XXX should i also check for TH_ACK ? */
1752 				match = (proto == IPPROTO_TCP && offset == 0 &&
1753 				    (TCP(ulp)->th_flags &
1754 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1755 				break;
1756 
1757 			case O_ALTQ: {
1758 				struct pf_mtag *at;
1759 				struct m_tag *mtag;
1760 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1761 
1762 				/*
1763 				 * ALTQ uses mbuf tags from another
1764 				 * packet filtering system - pf(4).
1765 				 * We allocate a tag in its format
1766 				 * and fill it in, pretending to be pf(4).
1767 				 */
1768 				match = 1;
1769 				at = pf_find_mtag(m);
1770 				if (at != NULL && at->qid != 0)
1771 					break;
1772 				mtag = m_tag_get(PACKET_TAG_PF,
1773 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1774 				if (mtag == NULL) {
1775 					/*
1776 					 * Let the packet fall back to the
1777 					 * default ALTQ.
1778 					 */
1779 					break;
1780 				}
1781 				m_tag_prepend(m, mtag);
1782 				at = (struct pf_mtag *)(mtag + 1);
1783 				at->qid = altq->qid;
1784 				at->hdr = ip;
1785 				break;
1786 			}
1787 
1788 			case O_LOG:
1789 				ipfw_log(f, hlen, args, m,
1790 				    oif, offset | ip6f_mf, tablearg, ip);
1791 				match = 1;
1792 				break;
1793 
1794 			case O_PROB:
1795 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1796 				break;
1797 
1798 			case O_VERREVPATH:
1799 				/* Outgoing packets automatically pass/match */
1800 				match = ((oif != NULL) ||
1801 				    (m->m_pkthdr.rcvif == NULL) ||
1802 				    (
1803 #ifdef INET6
1804 				    is_ipv6 ?
1805 					verify_path6(&(args->f_id.src_ip6),
1806 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1807 #endif
1808 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1809 				        args->f_id.fib)));
1810 				break;
1811 
1812 			case O_VERSRCREACH:
1813 				/* Outgoing packets automatically pass/match */
1814 				match = (hlen > 0 && ((oif != NULL) ||
1815 #ifdef INET6
1816 				    is_ipv6 ?
1817 				        verify_path6(&(args->f_id.src_ip6),
1818 				            NULL, args->f_id.fib) :
1819 #endif
1820 				    verify_path(src_ip, NULL, args->f_id.fib)));
1821 				break;
1822 
1823 			case O_ANTISPOOF:
1824 				/* Outgoing packets automatically pass/match */
1825 				if (oif == NULL && hlen > 0 &&
1826 				    (  (is_ipv4 && in_localaddr(src_ip))
1827 #ifdef INET6
1828 				    || (is_ipv6 &&
1829 				        in6_localaddr(&(args->f_id.src_ip6)))
1830 #endif
1831 				    ))
1832 					match =
1833 #ifdef INET6
1834 					    is_ipv6 ? verify_path6(
1835 					        &(args->f_id.src_ip6),
1836 					        m->m_pkthdr.rcvif,
1837 						args->f_id.fib) :
1838 #endif
1839 					    verify_path(src_ip,
1840 					    	m->m_pkthdr.rcvif,
1841 					        args->f_id.fib);
1842 				else
1843 					match = 1;
1844 				break;
1845 
1846 			case O_IPSEC:
1847 #ifdef IPSEC
1848 				match = (m_tag_find(m,
1849 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1850 #endif
1851 				/* otherwise no match */
1852 				break;
1853 
1854 #ifdef INET6
1855 			case O_IP6_SRC:
1856 				match = is_ipv6 &&
1857 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1858 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1859 				break;
1860 
1861 			case O_IP6_DST:
1862 				match = is_ipv6 &&
1863 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1864 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1865 				break;
1866 			case O_IP6_SRC_MASK:
1867 			case O_IP6_DST_MASK:
1868 				if (is_ipv6) {
1869 					int i = cmdlen - 1;
1870 					struct in6_addr p;
1871 					struct in6_addr *d =
1872 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1873 
1874 					for (; !match && i > 0; d += 2,
1875 					    i -= F_INSN_SIZE(struct in6_addr)
1876 					    * 2) {
1877 						p = (cmd->opcode ==
1878 						    O_IP6_SRC_MASK) ?
1879 						    args->f_id.src_ip6:
1880 						    args->f_id.dst_ip6;
1881 						APPLY_MASK(&p, &d[1]);
1882 						match =
1883 						    IN6_ARE_ADDR_EQUAL(&d[0],
1884 						    &p);
1885 					}
1886 				}
1887 				break;
1888 
1889 			case O_FLOW6ID:
1890 				match = is_ipv6 &&
1891 				    flow6id_match(args->f_id.flow_id6,
1892 				    (ipfw_insn_u32 *) cmd);
1893 				break;
1894 
1895 			case O_EXT_HDR:
1896 				match = is_ipv6 &&
1897 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1898 				break;
1899 
1900 			case O_IP6:
1901 				match = is_ipv6;
1902 				break;
1903 #endif
1904 
1905 			case O_IP4:
1906 				match = is_ipv4;
1907 				break;
1908 
1909 			case O_TAG: {
1910 				struct m_tag *mtag;
1911 				uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1912 
1913 				/* Packet is already tagged with this tag? */
1914 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1915 
1916 				/* We have `untag' action when F_NOT flag is
1917 				 * present. And we must remove this mtag from
1918 				 * mbuf and reset `match' to zero (`match' will
1919 				 * be inversed later).
1920 				 * Otherwise we should allocate new mtag and
1921 				 * push it into mbuf.
1922 				 */
1923 				if (cmd->len & F_NOT) { /* `untag' action */
1924 					if (mtag != NULL)
1925 						m_tag_delete(m, mtag);
1926 					match = 0;
1927 				} else {
1928 					if (mtag == NULL) {
1929 						mtag = m_tag_alloc( MTAG_IPFW,
1930 						    tag, 0, M_NOWAIT);
1931 						if (mtag != NULL)
1932 							m_tag_prepend(m, mtag);
1933 					}
1934 					match = 1;
1935 				}
1936 				break;
1937 			}
1938 
1939 			case O_FIB: /* try match the specified fib */
1940 				if (args->f_id.fib == cmd->arg1)
1941 					match = 1;
1942 				break;
1943 
1944 			case O_SOCKARG:	{
1945 				struct inpcb *inp = args->inp;
1946 				struct inpcbinfo *pi;
1947 
1948 				if (is_ipv6) /* XXX can we remove this ? */
1949 					break;
1950 
1951 				if (proto == IPPROTO_TCP)
1952 					pi = &V_tcbinfo;
1953 				else if (proto == IPPROTO_UDP)
1954 					pi = &V_udbinfo;
1955 				else
1956 					break;
1957 
1958 				/*
1959 				 * XXXRW: so_user_cookie should almost
1960 				 * certainly be inp_user_cookie?
1961 				 */
1962 
1963 				/* For incomming packet, lookup up the
1964 				inpcb using the src/dest ip/port tuple */
1965 				if (inp == NULL) {
1966 					inp = in_pcblookup(pi,
1967 						src_ip, htons(src_port),
1968 						dst_ip, htons(dst_port),
1969 						INPLOOKUP_RLOCKPCB, NULL);
1970 					if (inp != NULL) {
1971 						tablearg =
1972 						    inp->inp_socket->so_user_cookie;
1973 						if (tablearg)
1974 							match = 1;
1975 						INP_RUNLOCK(inp);
1976 					}
1977 				} else {
1978 					if (inp->inp_socket) {
1979 						tablearg =
1980 						    inp->inp_socket->so_user_cookie;
1981 						if (tablearg)
1982 							match = 1;
1983 					}
1984 				}
1985 				break;
1986 			}
1987 
1988 			case O_TAGGED: {
1989 				struct m_tag *mtag;
1990 				uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1991 
1992 				if (cmdlen == 1) {
1993 					match = m_tag_locate(m, MTAG_IPFW,
1994 					    tag, NULL) != NULL;
1995 					break;
1996 				}
1997 
1998 				/* we have ranges */
1999 				for (mtag = m_tag_first(m);
2000 				    mtag != NULL && !match;
2001 				    mtag = m_tag_next(m, mtag)) {
2002 					uint16_t *p;
2003 					int i;
2004 
2005 					if (mtag->m_tag_cookie != MTAG_IPFW)
2006 						continue;
2007 
2008 					p = ((ipfw_insn_u16 *)cmd)->ports;
2009 					i = cmdlen - 1;
2010 					for(; !match && i > 0; i--, p += 2)
2011 						match =
2012 						    mtag->m_tag_id >= p[0] &&
2013 						    mtag->m_tag_id <= p[1];
2014 				}
2015 				break;
2016 			}
2017 
2018 			/*
2019 			 * The second set of opcodes represents 'actions',
2020 			 * i.e. the terminal part of a rule once the packet
2021 			 * matches all previous patterns.
2022 			 * Typically there is only one action for each rule,
2023 			 * and the opcode is stored at the end of the rule
2024 			 * (but there are exceptions -- see below).
2025 			 *
2026 			 * In general, here we set retval and terminate the
2027 			 * outer loop (would be a 'break 3' in some language,
2028 			 * but we need to set l=0, done=1)
2029 			 *
2030 			 * Exceptions:
2031 			 * O_COUNT and O_SKIPTO actions:
2032 			 *   instead of terminating, we jump to the next rule
2033 			 *   (setting l=0), or to the SKIPTO target (setting
2034 			 *   f/f_len, cmd and l as needed), respectively.
2035 			 *
2036 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2037 			 *   perform some action and set match = 1;
2038 			 *
2039 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2040 			 *   not real 'actions', and are stored right
2041 			 *   before the 'action' part of the rule.
2042 			 *   These opcodes try to install an entry in the
2043 			 *   state tables; if successful, we continue with
2044 			 *   the next opcode (match=1; break;), otherwise
2045 			 *   the packet must be dropped (set retval,
2046 			 *   break loops with l=0, done=1)
2047 			 *
2048 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2049 			 *   cause a lookup of the state table, and a jump
2050 			 *   to the 'action' part of the parent rule
2051 			 *   if an entry is found, or
2052 			 *   (CHECK_STATE only) a jump to the next rule if
2053 			 *   the entry is not found.
2054 			 *   The result of the lookup is cached so that
2055 			 *   further instances of these opcodes become NOPs.
2056 			 *   The jump to the next rule is done by setting
2057 			 *   l=0, cmdlen=0.
2058 			 */
2059 			case O_LIMIT:
2060 			case O_KEEP_STATE:
2061 				if (ipfw_install_state(f,
2062 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2063 					/* error or limit violation */
2064 					retval = IP_FW_DENY;
2065 					l = 0;	/* exit inner loop */
2066 					done = 1; /* exit outer loop */
2067 				}
2068 				match = 1;
2069 				break;
2070 
2071 			case O_PROBE_STATE:
2072 			case O_CHECK_STATE:
2073 				/*
2074 				 * dynamic rules are checked at the first
2075 				 * keep-state or check-state occurrence,
2076 				 * with the result being stored in dyn_dir.
2077 				 * The compiler introduces a PROBE_STATE
2078 				 * instruction for us when we have a
2079 				 * KEEP_STATE (because PROBE_STATE needs
2080 				 * to be run first).
2081 				 */
2082 				if (dyn_dir == MATCH_UNKNOWN &&
2083 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2084 				     &dyn_dir, proto == IPPROTO_TCP ?
2085 					TCP(ulp) : NULL))
2086 					!= NULL) {
2087 					/*
2088 					 * Found dynamic entry, update stats
2089 					 * and jump to the 'action' part of
2090 					 * the parent rule by setting
2091 					 * f, cmd, l and clearing cmdlen.
2092 					 */
2093 					IPFW_INC_DYN_COUNTER(q, pktlen);
2094 					/* XXX we would like to have f_pos
2095 					 * readily accessible in the dynamic
2096 				         * rule, instead of having to
2097 					 * lookup q->rule.
2098 					 */
2099 					f = q->rule;
2100 					f_pos = ipfw_find_rule(chain,
2101 						f->rulenum, f->id);
2102 					cmd = ACTION_PTR(f);
2103 					l = f->cmd_len - f->act_ofs;
2104 					ipfw_dyn_unlock(q);
2105 					cmdlen = 0;
2106 					match = 1;
2107 					break;
2108 				}
2109 				/*
2110 				 * Dynamic entry not found. If CHECK_STATE,
2111 				 * skip to next rule, if PROBE_STATE just
2112 				 * ignore and continue with next opcode.
2113 				 */
2114 				if (cmd->opcode == O_CHECK_STATE)
2115 					l = 0;	/* exit inner loop */
2116 				match = 1;
2117 				break;
2118 
2119 			case O_ACCEPT:
2120 				retval = 0;	/* accept */
2121 				l = 0;		/* exit inner loop */
2122 				done = 1;	/* exit outer loop */
2123 				break;
2124 
2125 			case O_PIPE:
2126 			case O_QUEUE:
2127 				set_match(args, f_pos, chain);
2128 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2129 				if (cmd->opcode == O_PIPE)
2130 					args->rule.info |= IPFW_IS_PIPE;
2131 				if (V_fw_one_pass)
2132 					args->rule.info |= IPFW_ONEPASS;
2133 				retval = IP_FW_DUMMYNET;
2134 				l = 0;          /* exit inner loop */
2135 				done = 1;       /* exit outer loop */
2136 				break;
2137 
2138 			case O_DIVERT:
2139 			case O_TEE:
2140 				if (args->eh) /* not on layer 2 */
2141 				    break;
2142 				/* otherwise this is terminal */
2143 				l = 0;		/* exit inner loop */
2144 				done = 1;	/* exit outer loop */
2145 				retval = (cmd->opcode == O_DIVERT) ?
2146 					IP_FW_DIVERT : IP_FW_TEE;
2147 				set_match(args, f_pos, chain);
2148 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2149 				break;
2150 
2151 			case O_COUNT:
2152 				IPFW_INC_RULE_COUNTER(f, pktlen);
2153 				l = 0;		/* exit inner loop */
2154 				break;
2155 
2156 			case O_SKIPTO:
2157 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2158 			    f_pos = jump_fast(chain, f, cmd->arg1, tablearg, 0);
2159 			    /*
2160 			     * Skip disabled rules, and re-enter
2161 			     * the inner loop with the correct
2162 			     * f_pos, f, l and cmd.
2163 			     * Also clear cmdlen and skip_or
2164 			     */
2165 			    for (; f_pos < chain->n_rules - 1 &&
2166 				    (V_set_disable &
2167 				     (1 << chain->map[f_pos]->set));
2168 				    f_pos++)
2169 				;
2170 			    /* Re-enter the inner loop at the skipto rule. */
2171 			    f = chain->map[f_pos];
2172 			    l = f->cmd_len;
2173 			    cmd = f->cmd;
2174 			    match = 1;
2175 			    cmdlen = 0;
2176 			    skip_or = 0;
2177 			    continue;
2178 			    break;	/* not reached */
2179 
2180 			case O_CALLRETURN: {
2181 				/*
2182 				 * Implementation of `subroutine' call/return,
2183 				 * in the stack carried in an mbuf tag. This
2184 				 * is different from `skipto' in that any call
2185 				 * address is possible (`skipto' must prevent
2186 				 * backward jumps to avoid endless loops).
2187 				 * We have `return' action when F_NOT flag is
2188 				 * present. The `m_tag_id' field is used as
2189 				 * stack pointer.
2190 				 */
2191 				struct m_tag *mtag;
2192 				uint16_t jmpto, *stack;
2193 
2194 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2195 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2196 				/*
2197 				 * Hand-rolled version of m_tag_locate() with
2198 				 * wildcard `type'.
2199 				 * If not already tagged, allocate new tag.
2200 				 */
2201 				mtag = m_tag_first(m);
2202 				while (mtag != NULL) {
2203 					if (mtag->m_tag_cookie ==
2204 					    MTAG_IPFW_CALL)
2205 						break;
2206 					mtag = m_tag_next(m, mtag);
2207 				}
2208 				if (mtag == NULL && IS_CALL) {
2209 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2210 					    IPFW_CALLSTACK_SIZE *
2211 					    sizeof(uint16_t), M_NOWAIT);
2212 					if (mtag != NULL)
2213 						m_tag_prepend(m, mtag);
2214 				}
2215 
2216 				/*
2217 				 * On error both `call' and `return' just
2218 				 * continue with next rule.
2219 				 */
2220 				if (IS_RETURN && (mtag == NULL ||
2221 				    mtag->m_tag_id == 0)) {
2222 					l = 0;		/* exit inner loop */
2223 					break;
2224 				}
2225 				if (IS_CALL && (mtag == NULL ||
2226 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2227 					printf("ipfw: call stack error, "
2228 					    "go to next rule\n");
2229 					l = 0;		/* exit inner loop */
2230 					break;
2231 				}
2232 
2233 				IPFW_INC_RULE_COUNTER(f, pktlen);
2234 				stack = (uint16_t *)(mtag + 1);
2235 
2236 				/*
2237 				 * The `call' action may use cached f_pos
2238 				 * (in f->next_rule), whose version is written
2239 				 * in f->next_rule.
2240 				 * The `return' action, however, doesn't have
2241 				 * fixed jump address in cmd->arg1 and can't use
2242 				 * cache.
2243 				 */
2244 				if (IS_CALL) {
2245 					stack[mtag->m_tag_id] = f->rulenum;
2246 					mtag->m_tag_id++;
2247 			    		f_pos = jump_fast(chain, f, cmd->arg1,
2248 					    tablearg, 1);
2249 				} else {	/* `return' action */
2250 					mtag->m_tag_id--;
2251 					jmpto = stack[mtag->m_tag_id] + 1;
2252 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2253 				}
2254 
2255 				/*
2256 				 * Skip disabled rules, and re-enter
2257 				 * the inner loop with the correct
2258 				 * f_pos, f, l and cmd.
2259 				 * Also clear cmdlen and skip_or
2260 				 */
2261 				for (; f_pos < chain->n_rules - 1 &&
2262 				    (V_set_disable &
2263 				    (1 << chain->map[f_pos]->set)); f_pos++)
2264 					;
2265 				/* Re-enter the inner loop at the dest rule. */
2266 				f = chain->map[f_pos];
2267 				l = f->cmd_len;
2268 				cmd = f->cmd;
2269 				cmdlen = 0;
2270 				skip_or = 0;
2271 				continue;
2272 				break;	/* NOTREACHED */
2273 			}
2274 #undef IS_CALL
2275 #undef IS_RETURN
2276 
2277 			case O_REJECT:
2278 				/*
2279 				 * Drop the packet and send a reject notice
2280 				 * if the packet is not ICMP (or is an ICMP
2281 				 * query), and it is not multicast/broadcast.
2282 				 */
2283 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2284 				    (proto != IPPROTO_ICMP ||
2285 				     is_icmp_query(ICMP(ulp))) &&
2286 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2287 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2288 					send_reject(args, cmd->arg1, iplen, ip);
2289 					m = args->m;
2290 				}
2291 				/* FALLTHROUGH */
2292 #ifdef INET6
2293 			case O_UNREACH6:
2294 				if (hlen > 0 && is_ipv6 &&
2295 				    ((offset & IP6F_OFF_MASK) == 0) &&
2296 				    (proto != IPPROTO_ICMPV6 ||
2297 				     (is_icmp6_query(icmp6_type) == 1)) &&
2298 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2299 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2300 					send_reject6(
2301 					    args, cmd->arg1, hlen,
2302 					    (struct ip6_hdr *)ip);
2303 					m = args->m;
2304 				}
2305 				/* FALLTHROUGH */
2306 #endif
2307 			case O_DENY:
2308 				retval = IP_FW_DENY;
2309 				l = 0;		/* exit inner loop */
2310 				done = 1;	/* exit outer loop */
2311 				break;
2312 
2313 			case O_FORWARD_IP:
2314 				if (args->eh)	/* not valid on layer2 pkts */
2315 					break;
2316 				if (q == NULL || q->rule != f ||
2317 				    dyn_dir == MATCH_FORWARD) {
2318 				    struct sockaddr_in *sa;
2319 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2320 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2321 					bcopy(sa, &args->hopstore,
2322 							sizeof(*sa));
2323 					args->hopstore.sin_addr.s_addr =
2324 						    htonl(tablearg);
2325 					args->next_hop = &args->hopstore;
2326 				    } else {
2327 					args->next_hop = sa;
2328 				    }
2329 				}
2330 				retval = IP_FW_PASS;
2331 				l = 0;          /* exit inner loop */
2332 				done = 1;       /* exit outer loop */
2333 				break;
2334 
2335 #ifdef INET6
2336 			case O_FORWARD_IP6:
2337 				if (args->eh)	/* not valid on layer2 pkts */
2338 					break;
2339 				if (q == NULL || q->rule != f ||
2340 				    dyn_dir == MATCH_FORWARD) {
2341 					struct sockaddr_in6 *sin6;
2342 
2343 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2344 					args->next_hop6 = sin6;
2345 				}
2346 				retval = IP_FW_PASS;
2347 				l = 0;		/* exit inner loop */
2348 				done = 1;	/* exit outer loop */
2349 				break;
2350 #endif
2351 
2352 			case O_NETGRAPH:
2353 			case O_NGTEE:
2354 				set_match(args, f_pos, chain);
2355 				args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2356 				if (V_fw_one_pass)
2357 					args->rule.info |= IPFW_ONEPASS;
2358 				retval = (cmd->opcode == O_NETGRAPH) ?
2359 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2360 				l = 0;          /* exit inner loop */
2361 				done = 1;       /* exit outer loop */
2362 				break;
2363 
2364 			case O_SETFIB: {
2365 				uint32_t fib;
2366 
2367 				IPFW_INC_RULE_COUNTER(f, pktlen);
2368 				fib = IP_FW_ARG_TABLEARG(cmd->arg1);
2369 				if (fib >= rt_numfibs)
2370 					fib = 0;
2371 				M_SETFIB(m, fib);
2372 				args->f_id.fib = fib;
2373 				l = 0;		/* exit inner loop */
2374 				break;
2375 		        }
2376 
2377 			case O_SETDSCP: {
2378 				uint16_t code;
2379 
2380 				code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
2381 				l = 0;		/* exit inner loop */
2382 				if (is_ipv4) {
2383 					uint16_t a;
2384 
2385 					a = ip->ip_tos;
2386 					ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2387 					a += ntohs(ip->ip_sum) - ip->ip_tos;
2388 					ip->ip_sum = htons(a);
2389 				} else if (is_ipv6) {
2390 					uint8_t *v;
2391 
2392 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2393 					*v = (*v & 0xF0) | (code >> 2);
2394 					v++;
2395 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2396 				} else
2397 					break;
2398 
2399 				IPFW_INC_RULE_COUNTER(f, pktlen);
2400 				break;
2401 			}
2402 
2403 			case O_NAT:
2404  				if (!IPFW_NAT_LOADED) {
2405 				    retval = IP_FW_DENY;
2406 				} else {
2407 				    struct cfg_nat *t;
2408 				    int nat_id;
2409 
2410 				    set_match(args, f_pos, chain);
2411 				    /* Check if this is 'global' nat rule */
2412 				    if (cmd->arg1 == 0) {
2413 					    retval = ipfw_nat_ptr(args, NULL, m);
2414 					    l = 0;
2415 					    done = 1;
2416 					    break;
2417 				    }
2418 				    t = ((ipfw_insn_nat *)cmd)->nat;
2419 				    if (t == NULL) {
2420 					nat_id = IP_FW_ARG_TABLEARG(cmd->arg1);
2421 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2422 
2423 					if (t == NULL) {
2424 					    retval = IP_FW_DENY;
2425 					    l = 0;	/* exit inner loop */
2426 					    done = 1;	/* exit outer loop */
2427 					    break;
2428 					}
2429 					if (cmd->arg1 != IP_FW_TABLEARG)
2430 					    ((ipfw_insn_nat *)cmd)->nat = t;
2431 				    }
2432 				    retval = ipfw_nat_ptr(args, t, m);
2433 				}
2434 				l = 0;          /* exit inner loop */
2435 				done = 1;       /* exit outer loop */
2436 				break;
2437 
2438 			case O_REASS: {
2439 				int ip_off;
2440 
2441 				IPFW_INC_RULE_COUNTER(f, pktlen);
2442 				l = 0;	/* in any case exit inner loop */
2443 				ip_off = ntohs(ip->ip_off);
2444 
2445 				/* if not fragmented, go to next rule */
2446 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2447 				    break;
2448 
2449 				args->m = m = ip_reass(m);
2450 
2451 				/*
2452 				 * do IP header checksum fixup.
2453 				 */
2454 				if (m == NULL) { /* fragment got swallowed */
2455 				    retval = IP_FW_DENY;
2456 				} else { /* good, packet complete */
2457 				    int hlen;
2458 
2459 				    ip = mtod(m, struct ip *);
2460 				    hlen = ip->ip_hl << 2;
2461 				    ip->ip_sum = 0;
2462 				    if (hlen == sizeof(struct ip))
2463 					ip->ip_sum = in_cksum_hdr(ip);
2464 				    else
2465 					ip->ip_sum = in_cksum(m, hlen);
2466 				    retval = IP_FW_REASS;
2467 				    set_match(args, f_pos, chain);
2468 				}
2469 				done = 1;	/* exit outer loop */
2470 				break;
2471 			}
2472 
2473 			default:
2474 				panic("-- unknown opcode %d\n", cmd->opcode);
2475 			} /* end of switch() on opcodes */
2476 			/*
2477 			 * if we get here with l=0, then match is irrelevant.
2478 			 */
2479 
2480 			if (cmd->len & F_NOT)
2481 				match = !match;
2482 
2483 			if (match) {
2484 				if (cmd->len & F_OR)
2485 					skip_or = 1;
2486 			} else {
2487 				if (!(cmd->len & F_OR)) /* not an OR block, */
2488 					break;		/* try next rule    */
2489 			}
2490 
2491 		}	/* end of inner loop, scan opcodes */
2492 #undef PULLUP_LEN
2493 
2494 		if (done)
2495 			break;
2496 
2497 /* next_rule:; */	/* try next rule		*/
2498 
2499 	}		/* end of outer for, scan rules */
2500 
2501 	if (done) {
2502 		struct ip_fw *rule = chain->map[f_pos];
2503 		/* Update statistics */
2504 		IPFW_INC_RULE_COUNTER(rule, pktlen);
2505 	} else {
2506 		retval = IP_FW_DENY;
2507 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2508 	}
2509 	IPFW_PF_RUNLOCK(chain);
2510 #ifdef __FreeBSD__
2511 	if (ucred_cache != NULL)
2512 		crfree(ucred_cache);
2513 #endif
2514 	return (retval);
2515 
2516 pullup_failed:
2517 	if (V_fw_verbose)
2518 		printf("ipfw: pullup failed\n");
2519 	return (IP_FW_DENY);
2520 }
2521 
2522 /*
2523  * Set maximum number of tables that can be used in given VNET ipfw instance.
2524  */
2525 #ifdef SYSCTL_NODE
2526 static int
2527 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2528 {
2529 	int error;
2530 	unsigned int ntables;
2531 
2532 	ntables = V_fw_tables_max;
2533 
2534 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2535 	/* Read operation or some error */
2536 	if ((error != 0) || (req->newptr == NULL))
2537 		return (error);
2538 
2539 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2540 }
2541 #endif
2542 /*
2543  * Module and VNET glue
2544  */
2545 
2546 /*
2547  * Stuff that must be initialised only on boot or module load
2548  */
2549 static int
2550 ipfw_init(void)
2551 {
2552 	int error = 0;
2553 
2554 	/*
2555  	 * Only print out this stuff the first time around,
2556 	 * when called from the sysinit code.
2557 	 */
2558 	printf("ipfw2 "
2559 #ifdef INET6
2560 		"(+ipv6) "
2561 #endif
2562 		"initialized, divert %s, nat %s, "
2563 		"default to %s, logging ",
2564 #ifdef IPDIVERT
2565 		"enabled",
2566 #else
2567 		"loadable",
2568 #endif
2569 #ifdef IPFIREWALL_NAT
2570 		"enabled",
2571 #else
2572 		"loadable",
2573 #endif
2574 		default_to_accept ? "accept" : "deny");
2575 
2576 	/*
2577 	 * Note: V_xxx variables can be accessed here but the vnet specific
2578 	 * initializer may not have been called yet for the VIMAGE case.
2579 	 * Tuneables will have been processed. We will print out values for
2580 	 * the default vnet.
2581 	 * XXX This should all be rationalized AFTER 8.0
2582 	 */
2583 	if (V_fw_verbose == 0)
2584 		printf("disabled\n");
2585 	else if (V_verbose_limit == 0)
2586 		printf("unlimited\n");
2587 	else
2588 		printf("limited to %d packets/entry by default\n",
2589 		    V_verbose_limit);
2590 
2591 	/* Check user-supplied table count for validness */
2592 	if (default_fw_tables > IPFW_TABLES_MAX)
2593 	  default_fw_tables = IPFW_TABLES_MAX;
2594 
2595 	ipfw_log_bpf(1); /* init */
2596 	return (error);
2597 }
2598 
2599 /*
2600  * Called for the removal of the last instance only on module unload.
2601  */
2602 static void
2603 ipfw_destroy(void)
2604 {
2605 
2606 	ipfw_log_bpf(0); /* uninit */
2607 	printf("IP firewall unloaded\n");
2608 }
2609 
2610 /*
2611  * Stuff that must be initialized for every instance
2612  * (including the first of course).
2613  */
2614 static int
2615 vnet_ipfw_init(const void *unused)
2616 {
2617 	int error;
2618 	struct ip_fw *rule = NULL;
2619 	struct ip_fw_chain *chain;
2620 
2621 	chain = &V_layer3_chain;
2622 
2623 	/* First set up some values that are compile time options */
2624 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2625 	V_fw_deny_unknown_exthdrs = 1;
2626 #ifdef IPFIREWALL_VERBOSE
2627 	V_fw_verbose = 1;
2628 #endif
2629 #ifdef IPFIREWALL_VERBOSE_LIMIT
2630 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2631 #endif
2632 #ifdef IPFIREWALL_NAT
2633 	LIST_INIT(&chain->nat);
2634 #endif
2635 
2636 	/* insert the default rule and create the initial map */
2637 	chain->n_rules = 1;
2638 	chain->static_len = sizeof(struct ip_fw);
2639 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2640 	if (chain->map)
2641 		rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2642 
2643 	/* Set initial number of tables */
2644 	V_fw_tables_max = default_fw_tables;
2645 	error = ipfw_init_tables(chain);
2646 	if (error) {
2647 		printf("ipfw2: setting up tables failed\n");
2648 		free(chain->map, M_IPFW);
2649 		free(rule, M_IPFW);
2650 		return (ENOSPC);
2651 	}
2652 
2653 	/* fill and insert the default rule */
2654 	rule->act_ofs = 0;
2655 	rule->rulenum = IPFW_DEFAULT_RULE;
2656 	rule->cmd_len = 1;
2657 	rule->set = RESVD_SET;
2658 	rule->cmd[0].len = 1;
2659 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2660 	chain->rules = chain->default_rule = chain->map[0] = rule;
2661 	chain->id = rule->id = 1;
2662 
2663 	IPFW_LOCK_INIT(chain);
2664 	ipfw_dyn_init(chain);
2665 
2666 	/* First set up some values that are compile time options */
2667 	V_ipfw_vnet_ready = 1;		/* Open for business */
2668 
2669 	/*
2670 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2671 	 * Even if the latter two fail we still keep the module alive
2672 	 * because the sockopt and layer2 paths are still useful.
2673 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2674 	 * so we can ignore the exact return value and just set a flag.
2675 	 *
2676 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2677 	 * changes in the underlying (per-vnet) variables trigger
2678 	 * immediate hook()/unhook() calls.
2679 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2680 	 * is checked on each packet because there are no pfil hooks.
2681 	 */
2682 	V_ip_fw_ctl_ptr = ipfw_ctl;
2683 	error = ipfw_attach_hooks(1);
2684 	return (error);
2685 }
2686 
2687 /*
2688  * Called for the removal of each instance.
2689  */
2690 static int
2691 vnet_ipfw_uninit(const void *unused)
2692 {
2693 	struct ip_fw *reap, *rule;
2694 	struct ip_fw_chain *chain = &V_layer3_chain;
2695 	int i;
2696 
2697 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2698 	/*
2699 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2700 	 * Then grab, release and grab again the WLOCK so we make
2701 	 * sure the update is propagated and nobody will be in.
2702 	 */
2703 	(void)ipfw_attach_hooks(0 /* detach */);
2704 	V_ip_fw_ctl_ptr = NULL;
2705 	IPFW_UH_WLOCK(chain);
2706 	IPFW_UH_WUNLOCK(chain);
2707 	IPFW_UH_WLOCK(chain);
2708 
2709 	IPFW_WLOCK(chain);
2710 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2711 	IPFW_WUNLOCK(chain);
2712 
2713 	ipfw_destroy_tables(chain);
2714 	reap = NULL;
2715 	IPFW_WLOCK(chain);
2716 	for (i = 0; i < chain->n_rules; i++) {
2717 		rule = chain->map[i];
2718 		rule->x_next = reap;
2719 		reap = rule;
2720 	}
2721 	if (chain->map)
2722 		free(chain->map, M_IPFW);
2723 	IPFW_WUNLOCK(chain);
2724 	IPFW_UH_WUNLOCK(chain);
2725 	if (reap != NULL)
2726 		ipfw_reap_rules(reap);
2727 	IPFW_LOCK_DESTROY(chain);
2728 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2729 	return 0;
2730 }
2731 
2732 /*
2733  * Module event handler.
2734  * In general we have the choice of handling most of these events by the
2735  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2736  * use the SYSINIT handlers as they are more capable of expressing the
2737  * flow of control during module and vnet operations, so this is just
2738  * a skeleton. Note there is no SYSINIT equivalent of the module
2739  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2740  */
2741 static int
2742 ipfw_modevent(module_t mod, int type, void *unused)
2743 {
2744 	int err = 0;
2745 
2746 	switch (type) {
2747 	case MOD_LOAD:
2748 		/* Called once at module load or
2749 	 	 * system boot if compiled in. */
2750 		break;
2751 	case MOD_QUIESCE:
2752 		/* Called before unload. May veto unloading. */
2753 		break;
2754 	case MOD_UNLOAD:
2755 		/* Called during unload. */
2756 		break;
2757 	case MOD_SHUTDOWN:
2758 		/* Called during system shutdown. */
2759 		break;
2760 	default:
2761 		err = EOPNOTSUPP;
2762 		break;
2763 	}
2764 	return err;
2765 }
2766 
2767 static moduledata_t ipfwmod = {
2768 	"ipfw",
2769 	ipfw_modevent,
2770 	0
2771 };
2772 
2773 /* Define startup order. */
2774 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2775 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2776 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2777 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2778 
2779 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2780 MODULE_VERSION(ipfw, 2);
2781 /* should declare some dependencies here */
2782 
2783 /*
2784  * Starting up. Done in order after ipfwmod() has been called.
2785  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2786  */
2787 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2788 	    ipfw_init, NULL);
2789 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2790 	    vnet_ipfw_init, NULL);
2791 
2792 /*
2793  * Closing up shop. These are done in REVERSE ORDER, but still
2794  * after ipfwmod() has been called. Not called on reboot.
2795  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2796  * or when the module is unloaded.
2797  */
2798 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2799 	    ipfw_destroy, NULL);
2800 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2801 	    vnet_ipfw_uninit, NULL);
2802 /* end of file */
2803