xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision ab2043b81eaba0d7d7769b4a58b2b6d17bc464a3)
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/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_icmp.h>
72 #include <netinet/ip_fw.h>
73 #include <netinet/ip_carp.h>
74 #include <netinet/pim.h>
75 #include <netinet/tcp_var.h>
76 #include <netinet/udp.h>
77 #include <netinet/udp_var.h>
78 #include <netinet/sctp.h>
79 
80 #include <netinet/ip6.h>
81 #include <netinet/icmp6.h>
82 #ifdef INET6
83 #include <netinet6/in6_pcb.h>
84 #include <netinet6/scope6_var.h>
85 #include <netinet6/ip6_var.h>
86 #endif
87 
88 #include <netpfil/ipfw/ip_fw_private.h>
89 
90 #include <machine/in_cksum.h>	/* XXX for in_cksum */
91 
92 #ifdef MAC
93 #include <security/mac/mac_framework.h>
94 #endif
95 
96 /*
97  * static variables followed by global ones.
98  * All ipfw global variables are here.
99  */
100 
101 /* ipfw_vnet_ready controls when we are open for business */
102 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
103 #define	V_ipfw_vnet_ready	VNET(ipfw_vnet_ready)
104 
105 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
106 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
107 
108 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
109 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
110 
111 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
112 static int default_to_accept = 1;
113 #else
114 static int default_to_accept;
115 #endif
116 
117 VNET_DEFINE(int, autoinc_step);
118 VNET_DEFINE(int, fw_one_pass) = 1;
119 
120 VNET_DEFINE(unsigned int, fw_tables_max);
121 /* Use 128 tables by default */
122 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
123 
124 /*
125  * Each rule belongs to one of 32 different sets (0..31).
126  * The variable set_disable contains one bit per set.
127  * If the bit is set, all rules in the corresponding set
128  * are disabled. Set RESVD_SET(31) is reserved for the default rule
129  * and rules that are not deleted by the flush command,
130  * and CANNOT be disabled.
131  * Rules in set RESVD_SET can only be deleted individually.
132  */
133 VNET_DEFINE(u_int32_t, set_disable);
134 #define	V_set_disable			VNET(set_disable)
135 
136 VNET_DEFINE(int, fw_verbose);
137 /* counter for ipfw_log(NULL...) */
138 VNET_DEFINE(u_int64_t, norule_counter);
139 VNET_DEFINE(int, verbose_limit);
140 
141 /* layer3_chain contains the list of rules for layer 3 */
142 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
143 
144 ipfw_nat_t *ipfw_nat_ptr = NULL;
145 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
146 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
147 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
148 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
149 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
150 
151 #ifdef SYSCTL_NODE
152 uint32_t dummy_def = IPFW_DEFAULT_RULE;
153 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
154 
155 SYSBEGIN(f3)
156 
157 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
158 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
159     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
160     "Only do a single pass through ipfw when using dummynet(4)");
161 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
162     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
163     "Rule number auto-increment step");
164 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
165     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
166     "Log matches to ipfw rules");
167 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
168     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
169     "Set upper limit of matches of ipfw rules logged");
170 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
171     &dummy_def, 0,
172     "The default/max possible rule number.");
173 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
174     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
175     "Maximum number of tables");
176 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
177     &default_to_accept, 0,
178     "Make the default rule accept all packets.");
179 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
180 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
181 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
182     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
183     "Number of static rules");
184 
185 #ifdef INET6
186 SYSCTL_DECL(_net_inet6_ip6);
187 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
188 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
189     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
190     "Deny packets with unknown IPv6 Extension Headers");
191 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
192     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
193     "Permit single packet IPv6 fragments");
194 #endif /* INET6 */
195 
196 SYSEND
197 
198 #endif /* SYSCTL_NODE */
199 
200 
201 /*
202  * Some macros used in the various matching options.
203  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
204  * Other macros just cast void * into the appropriate type
205  */
206 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
207 #define	TCP(p)		((struct tcphdr *)(p))
208 #define	SCTP(p)		((struct sctphdr *)(p))
209 #define	UDP(p)		((struct udphdr *)(p))
210 #define	ICMP(p)		((struct icmphdr *)(p))
211 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
212 
213 static __inline int
214 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
215 {
216 	int type = icmp->icmp_type;
217 
218 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
219 }
220 
221 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
222     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
223 
224 static int
225 is_icmp_query(struct icmphdr *icmp)
226 {
227 	int type = icmp->icmp_type;
228 
229 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
230 }
231 #undef TT
232 
233 /*
234  * The following checks use two arrays of 8 or 16 bits to store the
235  * bits that we want set or clear, respectively. They are in the
236  * low and high half of cmd->arg1 or cmd->d[0].
237  *
238  * We scan options and store the bits we find set. We succeed if
239  *
240  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
241  *
242  * The code is sometimes optimized not to store additional variables.
243  */
244 
245 static int
246 flags_match(ipfw_insn *cmd, u_int8_t bits)
247 {
248 	u_char want_clear;
249 	bits = ~bits;
250 
251 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
252 		return 0; /* some bits we want set were clear */
253 	want_clear = (cmd->arg1 >> 8) & 0xff;
254 	if ( (want_clear & bits) != want_clear)
255 		return 0; /* some bits we want clear were set */
256 	return 1;
257 }
258 
259 static int
260 ipopts_match(struct ip *ip, ipfw_insn *cmd)
261 {
262 	int optlen, bits = 0;
263 	u_char *cp = (u_char *)(ip + 1);
264 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
265 
266 	for (; x > 0; x -= optlen, cp += optlen) {
267 		int opt = cp[IPOPT_OPTVAL];
268 
269 		if (opt == IPOPT_EOL)
270 			break;
271 		if (opt == IPOPT_NOP)
272 			optlen = 1;
273 		else {
274 			optlen = cp[IPOPT_OLEN];
275 			if (optlen <= 0 || optlen > x)
276 				return 0; /* invalid or truncated */
277 		}
278 		switch (opt) {
279 
280 		default:
281 			break;
282 
283 		case IPOPT_LSRR:
284 			bits |= IP_FW_IPOPT_LSRR;
285 			break;
286 
287 		case IPOPT_SSRR:
288 			bits |= IP_FW_IPOPT_SSRR;
289 			break;
290 
291 		case IPOPT_RR:
292 			bits |= IP_FW_IPOPT_RR;
293 			break;
294 
295 		case IPOPT_TS:
296 			bits |= IP_FW_IPOPT_TS;
297 			break;
298 		}
299 	}
300 	return (flags_match(cmd, bits));
301 }
302 
303 static int
304 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
305 {
306 	int optlen, bits = 0;
307 	u_char *cp = (u_char *)(tcp + 1);
308 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
309 
310 	for (; x > 0; x -= optlen, cp += optlen) {
311 		int opt = cp[0];
312 		if (opt == TCPOPT_EOL)
313 			break;
314 		if (opt == TCPOPT_NOP)
315 			optlen = 1;
316 		else {
317 			optlen = cp[1];
318 			if (optlen <= 0)
319 				break;
320 		}
321 
322 		switch (opt) {
323 
324 		default:
325 			break;
326 
327 		case TCPOPT_MAXSEG:
328 			bits |= IP_FW_TCPOPT_MSS;
329 			break;
330 
331 		case TCPOPT_WINDOW:
332 			bits |= IP_FW_TCPOPT_WINDOW;
333 			break;
334 
335 		case TCPOPT_SACK_PERMITTED:
336 		case TCPOPT_SACK:
337 			bits |= IP_FW_TCPOPT_SACK;
338 			break;
339 
340 		case TCPOPT_TIMESTAMP:
341 			bits |= IP_FW_TCPOPT_TS;
342 			break;
343 
344 		}
345 	}
346 	return (flags_match(cmd, bits));
347 }
348 
349 static int
350 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
351 {
352 	if (ifp == NULL)	/* no iface with this packet, match fails */
353 		return 0;
354 	/* Check by name or by IP address */
355 	if (cmd->name[0] != '\0') { /* match by name */
356 		if (cmd->name[0] == '\1') /* use tablearg to match */
357 			return ipfw_lookup_table_extended(chain, cmd->p.glob,
358 				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
359 		/* Check name */
360 		if (cmd->p.glob) {
361 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
362 				return(1);
363 		} else {
364 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
365 				return(1);
366 		}
367 	} else {
368 #ifdef __FreeBSD__	/* and OSX too ? */
369 		struct ifaddr *ia;
370 
371 		if_addr_rlock(ifp);
372 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
373 			if (ia->ifa_addr->sa_family != AF_INET)
374 				continue;
375 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
376 			    (ia->ifa_addr))->sin_addr.s_addr) {
377 				if_addr_runlock(ifp);
378 				return(1);	/* match */
379 			}
380 		}
381 		if_addr_runlock(ifp);
382 #endif /* __FreeBSD__ */
383 	}
384 	return(0);	/* no match, fail ... */
385 }
386 
387 /*
388  * The verify_path function checks if a route to the src exists and
389  * if it is reachable via ifp (when provided).
390  *
391  * The 'verrevpath' option checks that the interface that an IP packet
392  * arrives on is the same interface that traffic destined for the
393  * packet's source address would be routed out of.
394  * The 'versrcreach' option just checks that the source address is
395  * reachable via any route (except default) in the routing table.
396  * These two are a measure to block forged packets. This is also
397  * commonly known as "anti-spoofing" or Unicast Reverse Path
398  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
399  * is purposely reminiscent of the Cisco IOS command,
400  *
401  *   ip verify unicast reverse-path
402  *   ip verify unicast source reachable-via any
403  *
404  * which implements the same functionality. But note that the syntax
405  * is misleading, and the check may be performed on all IP packets
406  * whether unicast, multicast, or broadcast.
407  */
408 static int
409 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
410 {
411 #ifndef __FreeBSD__
412 	return 0;
413 #else
414 	struct route ro;
415 	struct sockaddr_in *dst;
416 
417 	bzero(&ro, sizeof(ro));
418 
419 	dst = (struct sockaddr_in *)&(ro.ro_dst);
420 	dst->sin_family = AF_INET;
421 	dst->sin_len = sizeof(*dst);
422 	dst->sin_addr = src;
423 	in_rtalloc_ign(&ro, 0, fib);
424 
425 	if (ro.ro_rt == NULL)
426 		return 0;
427 
428 	/*
429 	 * If ifp is provided, check for equality with rtentry.
430 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
431 	 * in order to pass packets injected back by if_simloop():
432 	 * if useloopback == 1 routing entry (via lo0) for our own address
433 	 * may exist, so we need to handle routing assymetry.
434 	 */
435 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
436 		RTFREE(ro.ro_rt);
437 		return 0;
438 	}
439 
440 	/* if no ifp provided, check if rtentry is not default route */
441 	if (ifp == NULL &&
442 	     satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
443 		RTFREE(ro.ro_rt);
444 		return 0;
445 	}
446 
447 	/* or if this is a blackhole/reject route */
448 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
449 		RTFREE(ro.ro_rt);
450 		return 0;
451 	}
452 
453 	/* found valid route */
454 	RTFREE(ro.ro_rt);
455 	return 1;
456 #endif /* __FreeBSD__ */
457 }
458 
459 #ifdef INET6
460 /*
461  * ipv6 specific rules here...
462  */
463 static __inline int
464 icmp6type_match (int type, ipfw_insn_u32 *cmd)
465 {
466 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
467 }
468 
469 static int
470 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
471 {
472 	int i;
473 	for (i=0; i <= cmd->o.arg1; ++i )
474 		if (curr_flow == cmd->d[i] )
475 			return 1;
476 	return 0;
477 }
478 
479 /* support for IP6_*_ME opcodes */
480 static int
481 search_ip6_addr_net (struct in6_addr * ip6_addr)
482 {
483 	struct ifnet *mdc;
484 	struct ifaddr *mdc2;
485 	struct in6_ifaddr *fdm;
486 	struct in6_addr copia;
487 
488 	TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
489 		if_addr_rlock(mdc);
490 		TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
491 			if (mdc2->ifa_addr->sa_family == AF_INET6) {
492 				fdm = (struct in6_ifaddr *)mdc2;
493 				copia = fdm->ia_addr.sin6_addr;
494 				/* need for leaving scope_id in the sock_addr */
495 				in6_clearscope(&copia);
496 				if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
497 					if_addr_runlock(mdc);
498 					return 1;
499 				}
500 			}
501 		}
502 		if_addr_runlock(mdc);
503 	}
504 	return 0;
505 }
506 
507 static int
508 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
509 {
510 	struct route_in6 ro;
511 	struct sockaddr_in6 *dst;
512 
513 	bzero(&ro, sizeof(ro));
514 
515 	dst = (struct sockaddr_in6 * )&(ro.ro_dst);
516 	dst->sin6_family = AF_INET6;
517 	dst->sin6_len = sizeof(*dst);
518 	dst->sin6_addr = *src;
519 
520 	in6_rtalloc_ign(&ro, 0, fib);
521 	if (ro.ro_rt == NULL)
522 		return 0;
523 
524 	/*
525 	 * if ifp is provided, check for equality with rtentry
526 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
527 	 * to support the case of sending packets to an address of our own.
528 	 * (where the former interface is the first argument of if_simloop()
529 	 *  (=ifp), the latter is lo0)
530 	 */
531 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
532 		RTFREE(ro.ro_rt);
533 		return 0;
534 	}
535 
536 	/* if no ifp provided, check if rtentry is not default route */
537 	if (ifp == NULL &&
538 	    IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
539 		RTFREE(ro.ro_rt);
540 		return 0;
541 	}
542 
543 	/* or if this is a blackhole/reject route */
544 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
545 		RTFREE(ro.ro_rt);
546 		return 0;
547 	}
548 
549 	/* found valid route */
550 	RTFREE(ro.ro_rt);
551 	return 1;
552 
553 }
554 
555 static int
556 is_icmp6_query(int icmp6_type)
557 {
558 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
559 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
560 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
561 	    icmp6_type == ICMP6_WRUREQUEST ||
562 	    icmp6_type == ICMP6_FQDN_QUERY ||
563 	    icmp6_type == ICMP6_NI_QUERY))
564 		return (1);
565 
566 	return (0);
567 }
568 
569 static void
570 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
571 {
572 	struct mbuf *m;
573 
574 	m = args->m;
575 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
576 		struct tcphdr *tcp;
577 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
578 
579 		if ((tcp->th_flags & TH_RST) == 0) {
580 			struct mbuf *m0;
581 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
582 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
583 			    tcp->th_flags | TH_RST);
584 			if (m0 != NULL)
585 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
586 				    NULL);
587 		}
588 		FREE_PKT(m);
589 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
590 #if 0
591 		/*
592 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
593 		 * as the contents are read. We need to m_adj() the
594 		 * needed amount.
595 		 * The mbuf will however be thrown away so we can adjust it.
596 		 * Remember we did an m_pullup on it already so we
597 		 * can make some assumptions about contiguousness.
598 		 */
599 		if (args->L3offset)
600 			m_adj(m, args->L3offset);
601 #endif
602 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
603 	} else
604 		FREE_PKT(m);
605 
606 	args->m = NULL;
607 }
608 
609 #endif /* INET6 */
610 
611 
612 /*
613  * sends a reject message, consuming the mbuf passed as an argument.
614  */
615 static void
616 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
617 {
618 
619 #if 0
620 	/* XXX When ip is not guaranteed to be at mtod() we will
621 	 * need to account for this */
622 	 * The mbuf will however be thrown away so we can adjust it.
623 	 * Remember we did an m_pullup on it already so we
624 	 * can make some assumptions about contiguousness.
625 	 */
626 	if (args->L3offset)
627 		m_adj(m, args->L3offset);
628 #endif
629 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
630 		/* We need the IP header in host order for icmp_error(). */
631 		SET_HOST_IPLEN(ip);
632 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
633 	} else if (args->f_id.proto == IPPROTO_TCP) {
634 		struct tcphdr *const tcp =
635 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
636 		if ( (tcp->th_flags & TH_RST) == 0) {
637 			struct mbuf *m;
638 			m = ipfw_send_pkt(args->m, &(args->f_id),
639 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
640 				tcp->th_flags | TH_RST);
641 			if (m != NULL)
642 				ip_output(m, NULL, NULL, 0, NULL, NULL);
643 		}
644 		FREE_PKT(args->m);
645 	} else
646 		FREE_PKT(args->m);
647 	args->m = NULL;
648 }
649 
650 /*
651  * Support for uid/gid/jail lookup. These tests are expensive
652  * (because we may need to look into the list of active sockets)
653  * so we cache the results. ugid_lookupp is 0 if we have not
654  * yet done a lookup, 1 if we succeeded, and -1 if we tried
655  * and failed. The function always returns the match value.
656  * We could actually spare the variable and use *uc, setting
657  * it to '(void *)check_uidgid if we have no info, NULL if
658  * we tried and failed, or any other value if successful.
659  */
660 static int
661 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
662     struct ucred **uc)
663 {
664 #ifndef __FreeBSD__
665 	/* XXX */
666 	return cred_check(insn, proto, oif,
667 	    dst_ip, dst_port, src_ip, src_port,
668 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
669 #else  /* FreeBSD */
670 	struct in_addr src_ip, dst_ip;
671 	struct inpcbinfo *pi;
672 	struct ipfw_flow_id *id;
673 	struct inpcb *pcb, *inp;
674 	struct ifnet *oif;
675 	int lookupflags;
676 	int match;
677 
678 	id = &args->f_id;
679 	inp = args->inp;
680 	oif = args->oif;
681 
682 	/*
683 	 * Check to see if the UDP or TCP stack supplied us with
684 	 * the PCB. If so, rather then holding a lock and looking
685 	 * up the PCB, we can use the one that was supplied.
686 	 */
687 	if (inp && *ugid_lookupp == 0) {
688 		INP_LOCK_ASSERT(inp);
689 		if (inp->inp_socket != NULL) {
690 			*uc = crhold(inp->inp_cred);
691 			*ugid_lookupp = 1;
692 		} else
693 			*ugid_lookupp = -1;
694 	}
695 	/*
696 	 * If we have already been here and the packet has no
697 	 * PCB entry associated with it, then we can safely
698 	 * assume that this is a no match.
699 	 */
700 	if (*ugid_lookupp == -1)
701 		return (0);
702 	if (id->proto == IPPROTO_TCP) {
703 		lookupflags = 0;
704 		pi = &V_tcbinfo;
705 	} else if (id->proto == IPPROTO_UDP) {
706 		lookupflags = INPLOOKUP_WILDCARD;
707 		pi = &V_udbinfo;
708 	} else
709 		return 0;
710 	lookupflags |= INPLOOKUP_RLOCKPCB;
711 	match = 0;
712 	if (*ugid_lookupp == 0) {
713 		if (id->addr_type == 6) {
714 #ifdef INET6
715 			if (oif == NULL)
716 				pcb = in6_pcblookup_mbuf(pi,
717 				    &id->src_ip6, htons(id->src_port),
718 				    &id->dst_ip6, htons(id->dst_port),
719 				    lookupflags, oif, args->m);
720 			else
721 				pcb = in6_pcblookup_mbuf(pi,
722 				    &id->dst_ip6, htons(id->dst_port),
723 				    &id->src_ip6, htons(id->src_port),
724 				    lookupflags, oif, args->m);
725 #else
726 			*ugid_lookupp = -1;
727 			return (0);
728 #endif
729 		} else {
730 			src_ip.s_addr = htonl(id->src_ip);
731 			dst_ip.s_addr = htonl(id->dst_ip);
732 			if (oif == NULL)
733 				pcb = in_pcblookup_mbuf(pi,
734 				    src_ip, htons(id->src_port),
735 				    dst_ip, htons(id->dst_port),
736 				    lookupflags, oif, args->m);
737 			else
738 				pcb = in_pcblookup_mbuf(pi,
739 				    dst_ip, htons(id->dst_port),
740 				    src_ip, htons(id->src_port),
741 				    lookupflags, oif, args->m);
742 		}
743 		if (pcb != NULL) {
744 			INP_RLOCK_ASSERT(pcb);
745 			*uc = crhold(pcb->inp_cred);
746 			*ugid_lookupp = 1;
747 			INP_RUNLOCK(pcb);
748 		}
749 		if (*ugid_lookupp == 0) {
750 			/*
751 			 * We tried and failed, set the variable to -1
752 			 * so we will not try again on this packet.
753 			 */
754 			*ugid_lookupp = -1;
755 			return (0);
756 		}
757 	}
758 	if (insn->o.opcode == O_UID)
759 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
760 	else if (insn->o.opcode == O_GID)
761 		match = groupmember((gid_t)insn->d[0], *uc);
762 	else if (insn->o.opcode == O_JAIL)
763 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
764 	return (match);
765 #endif /* __FreeBSD__ */
766 }
767 
768 /*
769  * Helper function to set args with info on the rule after the matching
770  * one. slot is precise, whereas we guess rule_id as they are
771  * assigned sequentially.
772  */
773 static inline void
774 set_match(struct ip_fw_args *args, int slot,
775 	struct ip_fw_chain *chain)
776 {
777 	args->rule.chain_id = chain->id;
778 	args->rule.slot = slot + 1; /* we use 0 as a marker */
779 	args->rule.rule_id = 1 + chain->map[slot]->id;
780 	args->rule.rulenum = chain->map[slot]->rulenum;
781 }
782 
783 /*
784  * The main check routine for the firewall.
785  *
786  * All arguments are in args so we can modify them and return them
787  * back to the caller.
788  *
789  * Parameters:
790  *
791  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
792  *		Starts with the IP header.
793  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
794  *	args->L3offset	Number of bytes bypassed if we came from L2.
795  *			e.g. often sizeof(eh)  ** NOTYET **
796  *	args->oif	Outgoing interface, NULL if packet is incoming.
797  *		The incoming interface is in the mbuf. (in)
798  *	args->divert_rule (in/out)
799  *		Skip up to the first rule past this rule number;
800  *		upon return, non-zero port number for divert or tee.
801  *
802  *	args->rule	Pointer to the last matching rule (in/out)
803  *	args->next_hop	Socket we are forwarding to (out).
804  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
805  *	args->f_id	Addresses grabbed from the packet (out)
806  * 	args->rule.info	a cookie depending on rule action
807  *
808  * Return value:
809  *
810  *	IP_FW_PASS	the packet must be accepted
811  *	IP_FW_DENY	the packet must be dropped
812  *	IP_FW_DIVERT	divert packet, port in m_tag
813  *	IP_FW_TEE	tee packet, port in m_tag
814  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
815  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
816  *		args->rule contains the matching rule,
817  *		args->rule.info has additional information.
818  *
819  */
820 int
821 ipfw_chk(struct ip_fw_args *args)
822 {
823 
824 	/*
825 	 * Local variables holding state while processing a packet:
826 	 *
827 	 * IMPORTANT NOTE: to speed up the processing of rules, there
828 	 * are some assumption on the values of the variables, which
829 	 * are documented here. Should you change them, please check
830 	 * the implementation of the various instructions to make sure
831 	 * that they still work.
832 	 *
833 	 * args->eh	The MAC header. It is non-null for a layer2
834 	 *	packet, it is NULL for a layer-3 packet.
835 	 * **notyet**
836 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
837 	 *
838 	 * m | args->m	Pointer to the mbuf, as received from the caller.
839 	 *	It may change if ipfw_chk() does an m_pullup, or if it
840 	 *	consumes the packet because it calls send_reject().
841 	 *	XXX This has to change, so that ipfw_chk() never modifies
842 	 *	or consumes the buffer.
843 	 * ip	is the beginning of the ip(4 or 6) header.
844 	 *	Calculated by adding the L3offset to the start of data.
845 	 *	(Until we start using L3offset, the packet is
846 	 *	supposed to start with the ip header).
847 	 */
848 	struct mbuf *m = args->m;
849 	struct ip *ip = mtod(m, struct ip *);
850 
851 	/*
852 	 * For rules which contain uid/gid or jail constraints, cache
853 	 * a copy of the users credentials after the pcb lookup has been
854 	 * executed. This will speed up the processing of rules with
855 	 * these types of constraints, as well as decrease contention
856 	 * on pcb related locks.
857 	 */
858 #ifndef __FreeBSD__
859 	struct bsd_ucred ucred_cache;
860 #else
861 	struct ucred *ucred_cache = NULL;
862 #endif
863 	int ucred_lookup = 0;
864 
865 	/*
866 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
867 	 *	inbound path (ether_input, ip_input).
868 	 *	If non-NULL, ipfw_chk has been called on the outbound path
869 	 *	(ether_output, ip_output).
870 	 */
871 	struct ifnet *oif = args->oif;
872 
873 	int f_pos = 0;		/* index of current rule in the array */
874 	int retval = 0;
875 
876 	/*
877 	 * hlen	The length of the IP header.
878 	 */
879 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
880 
881 	/*
882 	 * offset	The offset of a fragment. offset != 0 means that
883 	 *	we have a fragment at this offset of an IPv4 packet.
884 	 *	offset == 0 means that (if this is an IPv4 packet)
885 	 *	this is the first or only fragment.
886 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
887 	 *	or there is a single packet fragement (fragement header added
888 	 *	without needed).  We will treat a single packet fragment as if
889 	 *	there was no fragment header (or log/block depending on the
890 	 *	V_fw_permit_single_frag6 sysctl setting).
891 	 */
892 	u_short offset = 0;
893 	u_short ip6f_mf = 0;
894 
895 	/*
896 	 * Local copies of addresses. They are only valid if we have
897 	 * an IP packet.
898 	 *
899 	 * proto	The protocol. Set to 0 for non-ip packets,
900 	 *	or to the protocol read from the packet otherwise.
901 	 *	proto != 0 means that we have an IPv4 packet.
902 	 *
903 	 * src_port, dst_port	port numbers, in HOST format. Only
904 	 *	valid for TCP and UDP packets.
905 	 *
906 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
907 	 *	Only valid for IPv4 packets.
908 	 */
909 	uint8_t proto;
910 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
911 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
912 	uint16_t iplen=0;
913 	int pktlen;
914 	uint16_t	etype = 0;	/* Host order stored ether type */
915 
916 	/*
917 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
918 	 * 	MATCH_NONE when checked and not matched (q = NULL),
919 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
920 	 */
921 	int dyn_dir = MATCH_UNKNOWN;
922 	ipfw_dyn_rule *q = NULL;
923 	struct ip_fw_chain *chain = &V_layer3_chain;
924 
925 	/*
926 	 * We store in ulp a pointer to the upper layer protocol header.
927 	 * In the ipv4 case this is easy to determine from the header,
928 	 * but for ipv6 we might have some additional headers in the middle.
929 	 * ulp is NULL if not found.
930 	 */
931 	void *ulp = NULL;		/* upper layer protocol pointer. */
932 
933 	/* XXX ipv6 variables */
934 	int is_ipv6 = 0;
935 	uint8_t	icmp6_type = 0;
936 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
937 	/* end of ipv6 variables */
938 
939 	int is_ipv4 = 0;
940 
941 	int done = 0;		/* flag to exit the outer loop */
942 
943 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
944 		return (IP_FW_PASS);	/* accept */
945 
946 	dst_ip.s_addr = 0;		/* make sure it is initialized */
947 	src_ip.s_addr = 0;		/* make sure it is initialized */
948 	pktlen = m->m_pkthdr.len;
949 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
950 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
951 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
952 
953 /*
954  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
955  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
956  * pointer might become stale after other pullups (but we never use it
957  * this way).
958  */
959 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
960 #define PULLUP_LEN(_len, p, T)					\
961 do {								\
962 	int x = (_len) + T;					\
963 	if ((m)->m_len < x) {					\
964 		args->m = m = m_pullup(m, x);			\
965 		if (m == NULL)					\
966 			goto pullup_failed;			\
967 	}							\
968 	p = (mtod(m, char *) + (_len));				\
969 } while (0)
970 
971 	/*
972 	 * if we have an ether header,
973 	 */
974 	if (args->eh)
975 		etype = ntohs(args->eh->ether_type);
976 
977 	/* Identify IP packets and fill up variables. */
978 	if (pktlen >= sizeof(struct ip6_hdr) &&
979 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
980 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
981 		is_ipv6 = 1;
982 		args->f_id.addr_type = 6;
983 		hlen = sizeof(struct ip6_hdr);
984 		proto = ip6->ip6_nxt;
985 
986 		/* Search extension headers to find upper layer protocols */
987 		while (ulp == NULL && offset == 0) {
988 			switch (proto) {
989 			case IPPROTO_ICMPV6:
990 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
991 				icmp6_type = ICMP6(ulp)->icmp6_type;
992 				break;
993 
994 			case IPPROTO_TCP:
995 				PULLUP_TO(hlen, ulp, struct tcphdr);
996 				dst_port = TCP(ulp)->th_dport;
997 				src_port = TCP(ulp)->th_sport;
998 				/* save flags for dynamic rules */
999 				args->f_id._flags = TCP(ulp)->th_flags;
1000 				break;
1001 
1002 			case IPPROTO_SCTP:
1003 				PULLUP_TO(hlen, ulp, struct sctphdr);
1004 				src_port = SCTP(ulp)->src_port;
1005 				dst_port = SCTP(ulp)->dest_port;
1006 				break;
1007 
1008 			case IPPROTO_UDP:
1009 				PULLUP_TO(hlen, ulp, struct udphdr);
1010 				dst_port = UDP(ulp)->uh_dport;
1011 				src_port = UDP(ulp)->uh_sport;
1012 				break;
1013 
1014 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1015 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1016 				ext_hd |= EXT_HOPOPTS;
1017 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1018 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1019 				ulp = NULL;
1020 				break;
1021 
1022 			case IPPROTO_ROUTING:	/* RFC 2460 */
1023 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1024 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1025 				case 0:
1026 					ext_hd |= EXT_RTHDR0;
1027 					break;
1028 				case 2:
1029 					ext_hd |= EXT_RTHDR2;
1030 					break;
1031 				default:
1032 					if (V_fw_verbose)
1033 						printf("IPFW2: IPV6 - Unknown "
1034 						    "Routing Header type(%d)\n",
1035 						    ((struct ip6_rthdr *)
1036 						    ulp)->ip6r_type);
1037 					if (V_fw_deny_unknown_exthdrs)
1038 					    return (IP_FW_DENY);
1039 					break;
1040 				}
1041 				ext_hd |= EXT_ROUTING;
1042 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1043 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1044 				ulp = NULL;
1045 				break;
1046 
1047 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1048 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1049 				ext_hd |= EXT_FRAGMENT;
1050 				hlen += sizeof (struct ip6_frag);
1051 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1052 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1053 					IP6F_OFF_MASK;
1054 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1055 					IP6F_MORE_FRAG;
1056 				if (V_fw_permit_single_frag6 == 0 &&
1057 				    offset == 0 && ip6f_mf == 0) {
1058 					if (V_fw_verbose)
1059 						printf("IPFW2: IPV6 - Invalid "
1060 						    "Fragment Header\n");
1061 					if (V_fw_deny_unknown_exthdrs)
1062 					    return (IP_FW_DENY);
1063 					break;
1064 				}
1065 				args->f_id.extra =
1066 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1067 				ulp = NULL;
1068 				break;
1069 
1070 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1071 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1072 				ext_hd |= EXT_DSTOPTS;
1073 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1074 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1075 				ulp = NULL;
1076 				break;
1077 
1078 			case IPPROTO_AH:	/* RFC 2402 */
1079 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1080 				ext_hd |= EXT_AH;
1081 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1082 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1083 				ulp = NULL;
1084 				break;
1085 
1086 			case IPPROTO_ESP:	/* RFC 2406 */
1087 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1088 				/* Anything past Seq# is variable length and
1089 				 * data past this ext. header is encrypted. */
1090 				ext_hd |= EXT_ESP;
1091 				break;
1092 
1093 			case IPPROTO_NONE:	/* RFC 2460 */
1094 				/*
1095 				 * Packet ends here, and IPv6 header has
1096 				 * already been pulled up. If ip6e_len!=0
1097 				 * then octets must be ignored.
1098 				 */
1099 				ulp = ip; /* non-NULL to get out of loop. */
1100 				break;
1101 
1102 			case IPPROTO_OSPFIGP:
1103 				/* XXX OSPF header check? */
1104 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1105 				break;
1106 
1107 			case IPPROTO_PIM:
1108 				/* XXX PIM header check? */
1109 				PULLUP_TO(hlen, ulp, struct pim);
1110 				break;
1111 
1112 			case IPPROTO_CARP:
1113 				PULLUP_TO(hlen, ulp, struct carp_header);
1114 				if (((struct carp_header *)ulp)->carp_version !=
1115 				    CARP_VERSION)
1116 					return (IP_FW_DENY);
1117 				if (((struct carp_header *)ulp)->carp_type !=
1118 				    CARP_ADVERTISEMENT)
1119 					return (IP_FW_DENY);
1120 				break;
1121 
1122 			case IPPROTO_IPV6:	/* RFC 2893 */
1123 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1124 				break;
1125 
1126 			case IPPROTO_IPV4:	/* RFC 2893 */
1127 				PULLUP_TO(hlen, ulp, struct ip);
1128 				break;
1129 
1130 			default:
1131 				if (V_fw_verbose)
1132 					printf("IPFW2: IPV6 - Unknown "
1133 					    "Extension Header(%d), ext_hd=%x\n",
1134 					     proto, ext_hd);
1135 				if (V_fw_deny_unknown_exthdrs)
1136 				    return (IP_FW_DENY);
1137 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1138 				break;
1139 			} /*switch */
1140 		}
1141 		ip = mtod(m, struct ip *);
1142 		ip6 = (struct ip6_hdr *)ip;
1143 		args->f_id.src_ip6 = ip6->ip6_src;
1144 		args->f_id.dst_ip6 = ip6->ip6_dst;
1145 		args->f_id.src_ip = 0;
1146 		args->f_id.dst_ip = 0;
1147 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1148 	} else if (pktlen >= sizeof(struct ip) &&
1149 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1150 	    	is_ipv4 = 1;
1151 		hlen = ip->ip_hl << 2;
1152 		args->f_id.addr_type = 4;
1153 
1154 		/*
1155 		 * Collect parameters into local variables for faster matching.
1156 		 */
1157 		proto = ip->ip_p;
1158 		src_ip = ip->ip_src;
1159 		dst_ip = ip->ip_dst;
1160 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1161 		iplen = ntohs(ip->ip_len);
1162 		pktlen = iplen < pktlen ? iplen : pktlen;
1163 
1164 		if (offset == 0) {
1165 			switch (proto) {
1166 			case IPPROTO_TCP:
1167 				PULLUP_TO(hlen, ulp, struct tcphdr);
1168 				dst_port = TCP(ulp)->th_dport;
1169 				src_port = TCP(ulp)->th_sport;
1170 				/* save flags for dynamic rules */
1171 				args->f_id._flags = TCP(ulp)->th_flags;
1172 				break;
1173 
1174 			case IPPROTO_SCTP:
1175 				PULLUP_TO(hlen, ulp, struct sctphdr);
1176 				src_port = SCTP(ulp)->src_port;
1177 				dst_port = SCTP(ulp)->dest_port;
1178 				break;
1179 
1180 			case IPPROTO_UDP:
1181 				PULLUP_TO(hlen, ulp, struct udphdr);
1182 				dst_port = UDP(ulp)->uh_dport;
1183 				src_port = UDP(ulp)->uh_sport;
1184 				break;
1185 
1186 			case IPPROTO_ICMP:
1187 				PULLUP_TO(hlen, ulp, struct icmphdr);
1188 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1189 				break;
1190 
1191 			default:
1192 				break;
1193 			}
1194 		}
1195 
1196 		ip = mtod(m, struct ip *);
1197 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1198 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1199 	}
1200 #undef PULLUP_TO
1201 	if (proto) { /* we may have port numbers, store them */
1202 		args->f_id.proto = proto;
1203 		args->f_id.src_port = src_port = ntohs(src_port);
1204 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1205 	}
1206 
1207 	IPFW_RLOCK(chain);
1208 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1209 		IPFW_RUNLOCK(chain);
1210 		return (IP_FW_PASS);	/* accept */
1211 	}
1212 	if (args->rule.slot) {
1213 		/*
1214 		 * Packet has already been tagged as a result of a previous
1215 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1216 		 * REASS, NETGRAPH, DIVERT/TEE...)
1217 		 * Validate the slot and continue from the next one
1218 		 * if still present, otherwise do a lookup.
1219 		 */
1220 		f_pos = (args->rule.chain_id == chain->id) ?
1221 		    args->rule.slot :
1222 		    ipfw_find_rule(chain, args->rule.rulenum,
1223 			args->rule.rule_id);
1224 	} else {
1225 		f_pos = 0;
1226 	}
1227 
1228 	/*
1229 	 * Now scan the rules, and parse microinstructions for each rule.
1230 	 * We have two nested loops and an inner switch. Sometimes we
1231 	 * need to break out of one or both loops, or re-enter one of
1232 	 * the loops with updated variables. Loop variables are:
1233 	 *
1234 	 *	f_pos (outer loop) points to the current rule.
1235 	 *		On output it points to the matching rule.
1236 	 *	done (outer loop) is used as a flag to break the loop.
1237 	 *	l (inner loop)	residual length of current rule.
1238 	 *		cmd points to the current microinstruction.
1239 	 *
1240 	 * We break the inner loop by setting l=0 and possibly
1241 	 * cmdlen=0 if we don't want to advance cmd.
1242 	 * We break the outer loop by setting done=1
1243 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1244 	 * as needed.
1245 	 */
1246 	for (; f_pos < chain->n_rules; f_pos++) {
1247 		ipfw_insn *cmd;
1248 		uint32_t tablearg = 0;
1249 		int l, cmdlen, skip_or; /* skip rest of OR block */
1250 		struct ip_fw *f;
1251 
1252 		f = chain->map[f_pos];
1253 		if (V_set_disable & (1 << f->set) )
1254 			continue;
1255 
1256 		skip_or = 0;
1257 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1258 		    l -= cmdlen, cmd += cmdlen) {
1259 			int match;
1260 
1261 			/*
1262 			 * check_body is a jump target used when we find a
1263 			 * CHECK_STATE, and need to jump to the body of
1264 			 * the target rule.
1265 			 */
1266 
1267 /* check_body: */
1268 			cmdlen = F_LEN(cmd);
1269 			/*
1270 			 * An OR block (insn_1 || .. || insn_n) has the
1271 			 * F_OR bit set in all but the last instruction.
1272 			 * The first match will set "skip_or", and cause
1273 			 * the following instructions to be skipped until
1274 			 * past the one with the F_OR bit clear.
1275 			 */
1276 			if (skip_or) {		/* skip this instruction */
1277 				if ((cmd->len & F_OR) == 0)
1278 					skip_or = 0;	/* next one is good */
1279 				continue;
1280 			}
1281 			match = 0; /* set to 1 if we succeed */
1282 
1283 			switch (cmd->opcode) {
1284 			/*
1285 			 * The first set of opcodes compares the packet's
1286 			 * fields with some pattern, setting 'match' if a
1287 			 * match is found. At the end of the loop there is
1288 			 * logic to deal with F_NOT and F_OR flags associated
1289 			 * with the opcode.
1290 			 */
1291 			case O_NOP:
1292 				match = 1;
1293 				break;
1294 
1295 			case O_FORWARD_MAC:
1296 				printf("ipfw: opcode %d unimplemented\n",
1297 				    cmd->opcode);
1298 				break;
1299 
1300 			case O_GID:
1301 			case O_UID:
1302 			case O_JAIL:
1303 				/*
1304 				 * We only check offset == 0 && proto != 0,
1305 				 * as this ensures that we have a
1306 				 * packet with the ports info.
1307 				 */
1308 				if (offset != 0)
1309 					break;
1310 				if (proto == IPPROTO_TCP ||
1311 				    proto == IPPROTO_UDP)
1312 					match = check_uidgid(
1313 						    (ipfw_insn_u32 *)cmd,
1314 						    args, &ucred_lookup,
1315 #ifdef __FreeBSD__
1316 						    &ucred_cache);
1317 #else
1318 						    (void *)&ucred_cache);
1319 #endif
1320 				break;
1321 
1322 			case O_RECV:
1323 				match = iface_match(m->m_pkthdr.rcvif,
1324 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1325 				break;
1326 
1327 			case O_XMIT:
1328 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1329 				    chain, &tablearg);
1330 				break;
1331 
1332 			case O_VIA:
1333 				match = iface_match(oif ? oif :
1334 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1335 				    chain, &tablearg);
1336 				break;
1337 
1338 			case O_MACADDR2:
1339 				if (args->eh != NULL) {	/* have MAC header */
1340 					u_int32_t *want = (u_int32_t *)
1341 						((ipfw_insn_mac *)cmd)->addr;
1342 					u_int32_t *mask = (u_int32_t *)
1343 						((ipfw_insn_mac *)cmd)->mask;
1344 					u_int32_t *hdr = (u_int32_t *)args->eh;
1345 
1346 					match =
1347 					    ( want[0] == (hdr[0] & mask[0]) &&
1348 					      want[1] == (hdr[1] & mask[1]) &&
1349 					      want[2] == (hdr[2] & mask[2]) );
1350 				}
1351 				break;
1352 
1353 			case O_MAC_TYPE:
1354 				if (args->eh != NULL) {
1355 					u_int16_t *p =
1356 					    ((ipfw_insn_u16 *)cmd)->ports;
1357 					int i;
1358 
1359 					for (i = cmdlen - 1; !match && i>0;
1360 					    i--, p += 2)
1361 						match = (etype >= p[0] &&
1362 						    etype <= p[1]);
1363 				}
1364 				break;
1365 
1366 			case O_FRAG:
1367 				match = (offset != 0);
1368 				break;
1369 
1370 			case O_IN:	/* "out" is "not in" */
1371 				match = (oif == NULL);
1372 				break;
1373 
1374 			case O_LAYER2:
1375 				match = (args->eh != NULL);
1376 				break;
1377 
1378 			case O_DIVERTED:
1379 			    {
1380 				/* For diverted packets, args->rule.info
1381 				 * contains the divert port (in host format)
1382 				 * reason and direction.
1383 				 */
1384 				uint32_t i = args->rule.info;
1385 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1386 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1387 			    }
1388 				break;
1389 
1390 			case O_PROTO:
1391 				/*
1392 				 * We do not allow an arg of 0 so the
1393 				 * check of "proto" only suffices.
1394 				 */
1395 				match = (proto == cmd->arg1);
1396 				break;
1397 
1398 			case O_IP_SRC:
1399 				match = is_ipv4 &&
1400 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1401 				    src_ip.s_addr);
1402 				break;
1403 
1404 			case O_IP_SRC_LOOKUP:
1405 			case O_IP_DST_LOOKUP:
1406 				if (is_ipv4) {
1407 				    uint32_t key =
1408 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1409 					    dst_ip.s_addr : src_ip.s_addr;
1410 				    uint32_t v = 0;
1411 
1412 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1413 					/* generic lookup. The key must be
1414 					 * in 32bit big-endian format.
1415 					 */
1416 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1417 					if (v == 0)
1418 					    key = dst_ip.s_addr;
1419 					else if (v == 1)
1420 					    key = src_ip.s_addr;
1421 					else if (v == 6) /* dscp */
1422 					    key = (ip->ip_tos >> 2) & 0x3f;
1423 					else if (offset != 0)
1424 					    break;
1425 					else if (proto != IPPROTO_TCP &&
1426 						proto != IPPROTO_UDP)
1427 					    break;
1428 					else if (v == 2)
1429 					    key = htonl(dst_port);
1430 					else if (v == 3)
1431 					    key = htonl(src_port);
1432 					else if (v == 4 || v == 5) {
1433 					    check_uidgid(
1434 						(ipfw_insn_u32 *)cmd,
1435 						args, &ucred_lookup,
1436 #ifdef __FreeBSD__
1437 						&ucred_cache);
1438 					    if (v == 4 /* O_UID */)
1439 						key = ucred_cache->cr_uid;
1440 					    else if (v == 5 /* O_JAIL */)
1441 						key = ucred_cache->cr_prison->pr_id;
1442 #else /* !__FreeBSD__ */
1443 						(void *)&ucred_cache);
1444 					    if (v ==4 /* O_UID */)
1445 						key = ucred_cache.uid;
1446 					    else if (v == 5 /* O_JAIL */)
1447 						key = ucred_cache.xid;
1448 #endif /* !__FreeBSD__ */
1449 					    key = htonl(key);
1450 					} else
1451 					    break;
1452 				    }
1453 				    match = ipfw_lookup_table(chain,
1454 					cmd->arg1, key, &v);
1455 				    if (!match)
1456 					break;
1457 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1458 					match =
1459 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1460 				    else
1461 					tablearg = v;
1462 				} else if (is_ipv6) {
1463 					uint32_t v = 0;
1464 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1465 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1466 					match = ipfw_lookup_table_extended(chain,
1467 							cmd->arg1, pkey, &v,
1468 							IPFW_TABLE_CIDR);
1469 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1470 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1471 					if (match)
1472 						tablearg = v;
1473 				}
1474 				break;
1475 
1476 			case O_IP_SRC_MASK:
1477 			case O_IP_DST_MASK:
1478 				if (is_ipv4) {
1479 				    uint32_t a =
1480 					(cmd->opcode == O_IP_DST_MASK) ?
1481 					    dst_ip.s_addr : src_ip.s_addr;
1482 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1483 				    int i = cmdlen-1;
1484 
1485 				    for (; !match && i>0; i-= 2, p+= 2)
1486 					match = (p[0] == (a & p[1]));
1487 				}
1488 				break;
1489 
1490 			case O_IP_SRC_ME:
1491 				if (is_ipv4) {
1492 					struct ifnet *tif;
1493 
1494 					INADDR_TO_IFP(src_ip, tif);
1495 					match = (tif != NULL);
1496 					break;
1497 				}
1498 #ifdef INET6
1499 				/* FALLTHROUGH */
1500 			case O_IP6_SRC_ME:
1501 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1502 #endif
1503 				break;
1504 
1505 			case O_IP_DST_SET:
1506 			case O_IP_SRC_SET:
1507 				if (is_ipv4) {
1508 					u_int32_t *d = (u_int32_t *)(cmd+1);
1509 					u_int32_t addr =
1510 					    cmd->opcode == O_IP_DST_SET ?
1511 						args->f_id.dst_ip :
1512 						args->f_id.src_ip;
1513 
1514 					    if (addr < d[0])
1515 						    break;
1516 					    addr -= d[0]; /* subtract base */
1517 					    match = (addr < cmd->arg1) &&
1518 						( d[ 1 + (addr>>5)] &
1519 						  (1<<(addr & 0x1f)) );
1520 				}
1521 				break;
1522 
1523 			case O_IP_DST:
1524 				match = is_ipv4 &&
1525 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1526 				    dst_ip.s_addr);
1527 				break;
1528 
1529 			case O_IP_DST_ME:
1530 				if (is_ipv4) {
1531 					struct ifnet *tif;
1532 
1533 					INADDR_TO_IFP(dst_ip, tif);
1534 					match = (tif != NULL);
1535 					break;
1536 				}
1537 #ifdef INET6
1538 				/* FALLTHROUGH */
1539 			case O_IP6_DST_ME:
1540 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1541 #endif
1542 				break;
1543 
1544 
1545 			case O_IP_SRCPORT:
1546 			case O_IP_DSTPORT:
1547 				/*
1548 				 * offset == 0 && proto != 0 is enough
1549 				 * to guarantee that we have a
1550 				 * packet with port info.
1551 				 */
1552 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1553 				    && offset == 0) {
1554 					u_int16_t x =
1555 					    (cmd->opcode == O_IP_SRCPORT) ?
1556 						src_port : dst_port ;
1557 					u_int16_t *p =
1558 					    ((ipfw_insn_u16 *)cmd)->ports;
1559 					int i;
1560 
1561 					for (i = cmdlen - 1; !match && i>0;
1562 					    i--, p += 2)
1563 						match = (x>=p[0] && x<=p[1]);
1564 				}
1565 				break;
1566 
1567 			case O_ICMPTYPE:
1568 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1569 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1570 				break;
1571 
1572 #ifdef INET6
1573 			case O_ICMP6TYPE:
1574 				match = is_ipv6 && offset == 0 &&
1575 				    proto==IPPROTO_ICMPV6 &&
1576 				    icmp6type_match(
1577 					ICMP6(ulp)->icmp6_type,
1578 					(ipfw_insn_u32 *)cmd);
1579 				break;
1580 #endif /* INET6 */
1581 
1582 			case O_IPOPT:
1583 				match = (is_ipv4 &&
1584 				    ipopts_match(ip, cmd) );
1585 				break;
1586 
1587 			case O_IPVER:
1588 				match = (is_ipv4 &&
1589 				    cmd->arg1 == ip->ip_v);
1590 				break;
1591 
1592 			case O_IPID:
1593 			case O_IPLEN:
1594 			case O_IPTTL:
1595 				if (is_ipv4) {	/* only for IP packets */
1596 				    uint16_t x;
1597 				    uint16_t *p;
1598 				    int i;
1599 
1600 				    if (cmd->opcode == O_IPLEN)
1601 					x = iplen;
1602 				    else if (cmd->opcode == O_IPTTL)
1603 					x = ip->ip_ttl;
1604 				    else /* must be IPID */
1605 					x = ntohs(ip->ip_id);
1606 				    if (cmdlen == 1) {
1607 					match = (cmd->arg1 == x);
1608 					break;
1609 				    }
1610 				    /* otherwise we have ranges */
1611 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1612 				    i = cmdlen - 1;
1613 				    for (; !match && i>0; i--, p += 2)
1614 					match = (x >= p[0] && x <= p[1]);
1615 				}
1616 				break;
1617 
1618 			case O_IPPRECEDENCE:
1619 				match = (is_ipv4 &&
1620 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1621 				break;
1622 
1623 			case O_IPTOS:
1624 				match = (is_ipv4 &&
1625 				    flags_match(cmd, ip->ip_tos));
1626 				break;
1627 
1628 			case O_TCPDATALEN:
1629 				if (proto == IPPROTO_TCP && offset == 0) {
1630 				    struct tcphdr *tcp;
1631 				    uint16_t x;
1632 				    uint16_t *p;
1633 				    int i;
1634 
1635 				    tcp = TCP(ulp);
1636 				    x = iplen -
1637 					((ip->ip_hl + tcp->th_off) << 2);
1638 				    if (cmdlen == 1) {
1639 					match = (cmd->arg1 == x);
1640 					break;
1641 				    }
1642 				    /* otherwise we have ranges */
1643 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1644 				    i = cmdlen - 1;
1645 				    for (; !match && i>0; i--, p += 2)
1646 					match = (x >= p[0] && x <= p[1]);
1647 				}
1648 				break;
1649 
1650 			case O_TCPFLAGS:
1651 				match = (proto == IPPROTO_TCP && offset == 0 &&
1652 				    flags_match(cmd, TCP(ulp)->th_flags));
1653 				break;
1654 
1655 			case O_TCPOPTS:
1656 				PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1657 				match = (proto == IPPROTO_TCP && offset == 0 &&
1658 				    tcpopts_match(TCP(ulp), cmd));
1659 				break;
1660 
1661 			case O_TCPSEQ:
1662 				match = (proto == IPPROTO_TCP && offset == 0 &&
1663 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1664 					TCP(ulp)->th_seq);
1665 				break;
1666 
1667 			case O_TCPACK:
1668 				match = (proto == IPPROTO_TCP && offset == 0 &&
1669 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1670 					TCP(ulp)->th_ack);
1671 				break;
1672 
1673 			case O_TCPWIN:
1674 				if (proto == IPPROTO_TCP && offset == 0) {
1675 				    uint16_t x;
1676 				    uint16_t *p;
1677 				    int i;
1678 
1679 				    x = ntohs(TCP(ulp)->th_win);
1680 				    if (cmdlen == 1) {
1681 					match = (cmd->arg1 == x);
1682 					break;
1683 				    }
1684 				    /* Otherwise we have ranges. */
1685 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1686 				    i = cmdlen - 1;
1687 				    for (; !match && i > 0; i--, p += 2)
1688 					match = (x >= p[0] && x <= p[1]);
1689 				}
1690 				break;
1691 
1692 			case O_ESTAB:
1693 				/* reject packets which have SYN only */
1694 				/* XXX should i also check for TH_ACK ? */
1695 				match = (proto == IPPROTO_TCP && offset == 0 &&
1696 				    (TCP(ulp)->th_flags &
1697 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1698 				break;
1699 
1700 			case O_ALTQ: {
1701 				struct pf_mtag *at;
1702 				struct m_tag *mtag;
1703 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1704 
1705 				/*
1706 				 * ALTQ uses mbuf tags from another
1707 				 * packet filtering system - pf(4).
1708 				 * We allocate a tag in its format
1709 				 * and fill it in, pretending to be pf(4).
1710 				 */
1711 				match = 1;
1712 				at = pf_find_mtag(m);
1713 				if (at != NULL && at->qid != 0)
1714 					break;
1715 				mtag = m_tag_get(PACKET_TAG_PF,
1716 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1717 				if (mtag == NULL) {
1718 					/*
1719 					 * Let the packet fall back to the
1720 					 * default ALTQ.
1721 					 */
1722 					break;
1723 				}
1724 				m_tag_prepend(m, mtag);
1725 				at = (struct pf_mtag *)(mtag + 1);
1726 				at->qid = altq->qid;
1727 				at->hdr = ip;
1728 				break;
1729 			}
1730 
1731 			case O_LOG:
1732 				ipfw_log(f, hlen, args, m,
1733 				    oif, offset | ip6f_mf, tablearg, ip);
1734 				match = 1;
1735 				break;
1736 
1737 			case O_PROB:
1738 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1739 				break;
1740 
1741 			case O_VERREVPATH:
1742 				/* Outgoing packets automatically pass/match */
1743 				match = ((oif != NULL) ||
1744 				    (m->m_pkthdr.rcvif == NULL) ||
1745 				    (
1746 #ifdef INET6
1747 				    is_ipv6 ?
1748 					verify_path6(&(args->f_id.src_ip6),
1749 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1750 #endif
1751 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1752 				        args->f_id.fib)));
1753 				break;
1754 
1755 			case O_VERSRCREACH:
1756 				/* Outgoing packets automatically pass/match */
1757 				match = (hlen > 0 && ((oif != NULL) ||
1758 #ifdef INET6
1759 				    is_ipv6 ?
1760 				        verify_path6(&(args->f_id.src_ip6),
1761 				            NULL, args->f_id.fib) :
1762 #endif
1763 				    verify_path(src_ip, NULL, args->f_id.fib)));
1764 				break;
1765 
1766 			case O_ANTISPOOF:
1767 				/* Outgoing packets automatically pass/match */
1768 				if (oif == NULL && hlen > 0 &&
1769 				    (  (is_ipv4 && in_localaddr(src_ip))
1770 #ifdef INET6
1771 				    || (is_ipv6 &&
1772 				        in6_localaddr(&(args->f_id.src_ip6)))
1773 #endif
1774 				    ))
1775 					match =
1776 #ifdef INET6
1777 					    is_ipv6 ? verify_path6(
1778 					        &(args->f_id.src_ip6),
1779 					        m->m_pkthdr.rcvif,
1780 						args->f_id.fib) :
1781 #endif
1782 					    verify_path(src_ip,
1783 					    	m->m_pkthdr.rcvif,
1784 					        args->f_id.fib);
1785 				else
1786 					match = 1;
1787 				break;
1788 
1789 			case O_IPSEC:
1790 #ifdef IPSEC
1791 				match = (m_tag_find(m,
1792 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1793 #endif
1794 				/* otherwise no match */
1795 				break;
1796 
1797 #ifdef INET6
1798 			case O_IP6_SRC:
1799 				match = is_ipv6 &&
1800 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1801 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1802 				break;
1803 
1804 			case O_IP6_DST:
1805 				match = is_ipv6 &&
1806 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1807 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1808 				break;
1809 			case O_IP6_SRC_MASK:
1810 			case O_IP6_DST_MASK:
1811 				if (is_ipv6) {
1812 					int i = cmdlen - 1;
1813 					struct in6_addr p;
1814 					struct in6_addr *d =
1815 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1816 
1817 					for (; !match && i > 0; d += 2,
1818 					    i -= F_INSN_SIZE(struct in6_addr)
1819 					    * 2) {
1820 						p = (cmd->opcode ==
1821 						    O_IP6_SRC_MASK) ?
1822 						    args->f_id.src_ip6:
1823 						    args->f_id.dst_ip6;
1824 						APPLY_MASK(&p, &d[1]);
1825 						match =
1826 						    IN6_ARE_ADDR_EQUAL(&d[0],
1827 						    &p);
1828 					}
1829 				}
1830 				break;
1831 
1832 			case O_FLOW6ID:
1833 				match = is_ipv6 &&
1834 				    flow6id_match(args->f_id.flow_id6,
1835 				    (ipfw_insn_u32 *) cmd);
1836 				break;
1837 
1838 			case O_EXT_HDR:
1839 				match = is_ipv6 &&
1840 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1841 				break;
1842 
1843 			case O_IP6:
1844 				match = is_ipv6;
1845 				break;
1846 #endif
1847 
1848 			case O_IP4:
1849 				match = is_ipv4;
1850 				break;
1851 
1852 			case O_TAG: {
1853 				struct m_tag *mtag;
1854 				uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1855 				    tablearg : cmd->arg1;
1856 
1857 				/* Packet is already tagged with this tag? */
1858 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1859 
1860 				/* We have `untag' action when F_NOT flag is
1861 				 * present. And we must remove this mtag from
1862 				 * mbuf and reset `match' to zero (`match' will
1863 				 * be inversed later).
1864 				 * Otherwise we should allocate new mtag and
1865 				 * push it into mbuf.
1866 				 */
1867 				if (cmd->len & F_NOT) { /* `untag' action */
1868 					if (mtag != NULL)
1869 						m_tag_delete(m, mtag);
1870 					match = 0;
1871 				} else {
1872 					if (mtag == NULL) {
1873 						mtag = m_tag_alloc( MTAG_IPFW,
1874 						    tag, 0, M_NOWAIT);
1875 						if (mtag != NULL)
1876 							m_tag_prepend(m, mtag);
1877 					}
1878 					match = 1;
1879 				}
1880 				break;
1881 			}
1882 
1883 			case O_FIB: /* try match the specified fib */
1884 				if (args->f_id.fib == cmd->arg1)
1885 					match = 1;
1886 				break;
1887 
1888 			case O_SOCKARG:	{
1889 				struct inpcb *inp = args->inp;
1890 				struct inpcbinfo *pi;
1891 
1892 				if (is_ipv6) /* XXX can we remove this ? */
1893 					break;
1894 
1895 				if (proto == IPPROTO_TCP)
1896 					pi = &V_tcbinfo;
1897 				else if (proto == IPPROTO_UDP)
1898 					pi = &V_udbinfo;
1899 				else
1900 					break;
1901 
1902 				/*
1903 				 * XXXRW: so_user_cookie should almost
1904 				 * certainly be inp_user_cookie?
1905 				 */
1906 
1907 				/* For incomming packet, lookup up the
1908 				inpcb using the src/dest ip/port tuple */
1909 				if (inp == NULL) {
1910 					inp = in_pcblookup(pi,
1911 						src_ip, htons(src_port),
1912 						dst_ip, htons(dst_port),
1913 						INPLOOKUP_RLOCKPCB, NULL);
1914 					if (inp != NULL) {
1915 						tablearg =
1916 						    inp->inp_socket->so_user_cookie;
1917 						if (tablearg)
1918 							match = 1;
1919 						INP_RUNLOCK(inp);
1920 					}
1921 				} else {
1922 					if (inp->inp_socket) {
1923 						tablearg =
1924 						    inp->inp_socket->so_user_cookie;
1925 						if (tablearg)
1926 							match = 1;
1927 					}
1928 				}
1929 				break;
1930 			}
1931 
1932 			case O_TAGGED: {
1933 				struct m_tag *mtag;
1934 				uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1935 				    tablearg : cmd->arg1;
1936 
1937 				if (cmdlen == 1) {
1938 					match = m_tag_locate(m, MTAG_IPFW,
1939 					    tag, NULL) != NULL;
1940 					break;
1941 				}
1942 
1943 				/* we have ranges */
1944 				for (mtag = m_tag_first(m);
1945 				    mtag != NULL && !match;
1946 				    mtag = m_tag_next(m, mtag)) {
1947 					uint16_t *p;
1948 					int i;
1949 
1950 					if (mtag->m_tag_cookie != MTAG_IPFW)
1951 						continue;
1952 
1953 					p = ((ipfw_insn_u16 *)cmd)->ports;
1954 					i = cmdlen - 1;
1955 					for(; !match && i > 0; i--, p += 2)
1956 						match =
1957 						    mtag->m_tag_id >= p[0] &&
1958 						    mtag->m_tag_id <= p[1];
1959 				}
1960 				break;
1961 			}
1962 
1963 			/*
1964 			 * The second set of opcodes represents 'actions',
1965 			 * i.e. the terminal part of a rule once the packet
1966 			 * matches all previous patterns.
1967 			 * Typically there is only one action for each rule,
1968 			 * and the opcode is stored at the end of the rule
1969 			 * (but there are exceptions -- see below).
1970 			 *
1971 			 * In general, here we set retval and terminate the
1972 			 * outer loop (would be a 'break 3' in some language,
1973 			 * but we need to set l=0, done=1)
1974 			 *
1975 			 * Exceptions:
1976 			 * O_COUNT and O_SKIPTO actions:
1977 			 *   instead of terminating, we jump to the next rule
1978 			 *   (setting l=0), or to the SKIPTO target (setting
1979 			 *   f/f_len, cmd and l as needed), respectively.
1980 			 *
1981 			 * O_TAG, O_LOG and O_ALTQ action parameters:
1982 			 *   perform some action and set match = 1;
1983 			 *
1984 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
1985 			 *   not real 'actions', and are stored right
1986 			 *   before the 'action' part of the rule.
1987 			 *   These opcodes try to install an entry in the
1988 			 *   state tables; if successful, we continue with
1989 			 *   the next opcode (match=1; break;), otherwise
1990 			 *   the packet must be dropped (set retval,
1991 			 *   break loops with l=0, done=1)
1992 			 *
1993 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1994 			 *   cause a lookup of the state table, and a jump
1995 			 *   to the 'action' part of the parent rule
1996 			 *   if an entry is found, or
1997 			 *   (CHECK_STATE only) a jump to the next rule if
1998 			 *   the entry is not found.
1999 			 *   The result of the lookup is cached so that
2000 			 *   further instances of these opcodes become NOPs.
2001 			 *   The jump to the next rule is done by setting
2002 			 *   l=0, cmdlen=0.
2003 			 */
2004 			case O_LIMIT:
2005 			case O_KEEP_STATE:
2006 				if (ipfw_install_state(f,
2007 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2008 					/* error or limit violation */
2009 					retval = IP_FW_DENY;
2010 					l = 0;	/* exit inner loop */
2011 					done = 1; /* exit outer loop */
2012 				}
2013 				match = 1;
2014 				break;
2015 
2016 			case O_PROBE_STATE:
2017 			case O_CHECK_STATE:
2018 				/*
2019 				 * dynamic rules are checked at the first
2020 				 * keep-state or check-state occurrence,
2021 				 * with the result being stored in dyn_dir.
2022 				 * The compiler introduces a PROBE_STATE
2023 				 * instruction for us when we have a
2024 				 * KEEP_STATE (because PROBE_STATE needs
2025 				 * to be run first).
2026 				 */
2027 				if (dyn_dir == MATCH_UNKNOWN &&
2028 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2029 				     &dyn_dir, proto == IPPROTO_TCP ?
2030 					TCP(ulp) : NULL))
2031 					!= NULL) {
2032 					/*
2033 					 * Found dynamic entry, update stats
2034 					 * and jump to the 'action' part of
2035 					 * the parent rule by setting
2036 					 * f, cmd, l and clearing cmdlen.
2037 					 */
2038 					q->pcnt++;
2039 					q->bcnt += pktlen;
2040 					/* XXX we would like to have f_pos
2041 					 * readily accessible in the dynamic
2042 				         * rule, instead of having to
2043 					 * lookup q->rule.
2044 					 */
2045 					f = q->rule;
2046 					f_pos = ipfw_find_rule(chain,
2047 						f->rulenum, f->id);
2048 					cmd = ACTION_PTR(f);
2049 					l = f->cmd_len - f->act_ofs;
2050 					ipfw_dyn_unlock();
2051 					cmdlen = 0;
2052 					match = 1;
2053 					break;
2054 				}
2055 				/*
2056 				 * Dynamic entry not found. If CHECK_STATE,
2057 				 * skip to next rule, if PROBE_STATE just
2058 				 * ignore and continue with next opcode.
2059 				 */
2060 				if (cmd->opcode == O_CHECK_STATE)
2061 					l = 0;	/* exit inner loop */
2062 				match = 1;
2063 				break;
2064 
2065 			case O_ACCEPT:
2066 				retval = 0;	/* accept */
2067 				l = 0;		/* exit inner loop */
2068 				done = 1;	/* exit outer loop */
2069 				break;
2070 
2071 			case O_PIPE:
2072 			case O_QUEUE:
2073 				set_match(args, f_pos, chain);
2074 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2075 					tablearg : cmd->arg1;
2076 				if (cmd->opcode == O_PIPE)
2077 					args->rule.info |= IPFW_IS_PIPE;
2078 				if (V_fw_one_pass)
2079 					args->rule.info |= IPFW_ONEPASS;
2080 				retval = IP_FW_DUMMYNET;
2081 				l = 0;          /* exit inner loop */
2082 				done = 1;       /* exit outer loop */
2083 				break;
2084 
2085 			case O_DIVERT:
2086 			case O_TEE:
2087 				if (args->eh) /* not on layer 2 */
2088 				    break;
2089 				/* otherwise this is terminal */
2090 				l = 0;		/* exit inner loop */
2091 				done = 1;	/* exit outer loop */
2092 				retval = (cmd->opcode == O_DIVERT) ?
2093 					IP_FW_DIVERT : IP_FW_TEE;
2094 				set_match(args, f_pos, chain);
2095 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2096 				    tablearg : cmd->arg1;
2097 				break;
2098 
2099 			case O_COUNT:
2100 				f->pcnt++;	/* update stats */
2101 				f->bcnt += pktlen;
2102 				f->timestamp = time_uptime;
2103 				l = 0;		/* exit inner loop */
2104 				break;
2105 
2106 			case O_SKIPTO:
2107 			    f->pcnt++;	/* update stats */
2108 			    f->bcnt += pktlen;
2109 			    f->timestamp = time_uptime;
2110 			    /* If possible use cached f_pos (in f->next_rule),
2111 			     * whose version is written in f->next_rule
2112 			     * (horrible hacks to avoid changing the ABI).
2113 			     */
2114 			    if (cmd->arg1 != IP_FW_TABLEARG &&
2115 				    (uintptr_t)f->x_next == chain->id) {
2116 				f_pos = (uintptr_t)f->next_rule;
2117 			    } else {
2118 				int i = (cmd->arg1 == IP_FW_TABLEARG) ?
2119 					tablearg : cmd->arg1;
2120 				/* make sure we do not jump backward */
2121 				if (i <= f->rulenum)
2122 				    i = f->rulenum + 1;
2123 				f_pos = ipfw_find_rule(chain, i, 0);
2124 				/* update the cache */
2125 				if (cmd->arg1 != IP_FW_TABLEARG) {
2126 				    f->next_rule =
2127 					(void *)(uintptr_t)f_pos;
2128 				    f->x_next =
2129 					(void *)(uintptr_t)chain->id;
2130 				}
2131 			    }
2132 			    /*
2133 			     * Skip disabled rules, and re-enter
2134 			     * the inner loop with the correct
2135 			     * f_pos, f, l and cmd.
2136 			     * Also clear cmdlen and skip_or
2137 			     */
2138 			    for (; f_pos < chain->n_rules - 1 &&
2139 				    (V_set_disable &
2140 				     (1 << chain->map[f_pos]->set));
2141 				    f_pos++)
2142 				;
2143 			    /* Re-enter the inner loop at the skipto rule. */
2144 			    f = chain->map[f_pos];
2145 			    l = f->cmd_len;
2146 			    cmd = f->cmd;
2147 			    match = 1;
2148 			    cmdlen = 0;
2149 			    skip_or = 0;
2150 			    continue;
2151 			    break;	/* not reached */
2152 
2153 			case O_CALLRETURN: {
2154 				/*
2155 				 * Implementation of `subroutine' call/return,
2156 				 * in the stack carried in an mbuf tag. This
2157 				 * is different from `skipto' in that any call
2158 				 * address is possible (`skipto' must prevent
2159 				 * backward jumps to avoid endless loops).
2160 				 * We have `return' action when F_NOT flag is
2161 				 * present. The `m_tag_id' field is used as
2162 				 * stack pointer.
2163 				 */
2164 				struct m_tag *mtag;
2165 				uint16_t jmpto, *stack;
2166 
2167 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2168 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2169 				/*
2170 				 * Hand-rolled version of m_tag_locate() with
2171 				 * wildcard `type'.
2172 				 * If not already tagged, allocate new tag.
2173 				 */
2174 				mtag = m_tag_first(m);
2175 				while (mtag != NULL) {
2176 					if (mtag->m_tag_cookie ==
2177 					    MTAG_IPFW_CALL)
2178 						break;
2179 					mtag = m_tag_next(m, mtag);
2180 				}
2181 				if (mtag == NULL && IS_CALL) {
2182 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2183 					    IPFW_CALLSTACK_SIZE *
2184 					    sizeof(uint16_t), M_NOWAIT);
2185 					if (mtag != NULL)
2186 						m_tag_prepend(m, mtag);
2187 				}
2188 
2189 				/*
2190 				 * On error both `call' and `return' just
2191 				 * continue with next rule.
2192 				 */
2193 				if (IS_RETURN && (mtag == NULL ||
2194 				    mtag->m_tag_id == 0)) {
2195 					l = 0;		/* exit inner loop */
2196 					break;
2197 				}
2198 				if (IS_CALL && (mtag == NULL ||
2199 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2200 					printf("ipfw: call stack error, "
2201 					    "go to next rule\n");
2202 					l = 0;		/* exit inner loop */
2203 					break;
2204 				}
2205 
2206 				f->pcnt++;	/* update stats */
2207 				f->bcnt += pktlen;
2208 				f->timestamp = time_uptime;
2209 				stack = (uint16_t *)(mtag + 1);
2210 
2211 				/*
2212 				 * The `call' action may use cached f_pos
2213 				 * (in f->next_rule), whose version is written
2214 				 * in f->next_rule.
2215 				 * The `return' action, however, doesn't have
2216 				 * fixed jump address in cmd->arg1 and can't use
2217 				 * cache.
2218 				 */
2219 				if (IS_CALL) {
2220 					stack[mtag->m_tag_id] = f->rulenum;
2221 					mtag->m_tag_id++;
2222 					if (cmd->arg1 != IP_FW_TABLEARG &&
2223 					    (uintptr_t)f->x_next == chain->id) {
2224 						f_pos = (uintptr_t)f->next_rule;
2225 					} else {
2226 						jmpto = (cmd->arg1 ==
2227 						    IP_FW_TABLEARG) ? tablearg:
2228 						    cmd->arg1;
2229 						f_pos = ipfw_find_rule(chain,
2230 						    jmpto, 0);
2231 						/* update the cache */
2232 						if (cmd->arg1 !=
2233 						    IP_FW_TABLEARG) {
2234 							f->next_rule =
2235 							    (void *)(uintptr_t)
2236 							    f_pos;
2237 							f->x_next =
2238 							    (void *)(uintptr_t)
2239 							    chain->id;
2240 						}
2241 					}
2242 				} else {	/* `return' action */
2243 					mtag->m_tag_id--;
2244 					jmpto = stack[mtag->m_tag_id] + 1;
2245 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2246 				}
2247 
2248 				/*
2249 				 * Skip disabled rules, and re-enter
2250 				 * the inner loop with the correct
2251 				 * f_pos, f, l and cmd.
2252 				 * Also clear cmdlen and skip_or
2253 				 */
2254 				for (; f_pos < chain->n_rules - 1 &&
2255 				    (V_set_disable &
2256 				    (1 << chain->map[f_pos]->set)); f_pos++)
2257 					;
2258 				/* Re-enter the inner loop at the dest rule. */
2259 				f = chain->map[f_pos];
2260 				l = f->cmd_len;
2261 				cmd = f->cmd;
2262 				cmdlen = 0;
2263 				skip_or = 0;
2264 				continue;
2265 				break;	/* NOTREACHED */
2266 			}
2267 #undef IS_CALL
2268 #undef IS_RETURN
2269 
2270 			case O_REJECT:
2271 				/*
2272 				 * Drop the packet and send a reject notice
2273 				 * if the packet is not ICMP (or is an ICMP
2274 				 * query), and it is not multicast/broadcast.
2275 				 */
2276 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2277 				    (proto != IPPROTO_ICMP ||
2278 				     is_icmp_query(ICMP(ulp))) &&
2279 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2280 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2281 					send_reject(args, cmd->arg1, iplen, ip);
2282 					m = args->m;
2283 				}
2284 				/* FALLTHROUGH */
2285 #ifdef INET6
2286 			case O_UNREACH6:
2287 				if (hlen > 0 && is_ipv6 &&
2288 				    ((offset & IP6F_OFF_MASK) == 0) &&
2289 				    (proto != IPPROTO_ICMPV6 ||
2290 				     (is_icmp6_query(icmp6_type) == 1)) &&
2291 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2292 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2293 					send_reject6(
2294 					    args, cmd->arg1, hlen,
2295 					    (struct ip6_hdr *)ip);
2296 					m = args->m;
2297 				}
2298 				/* FALLTHROUGH */
2299 #endif
2300 			case O_DENY:
2301 				retval = IP_FW_DENY;
2302 				l = 0;		/* exit inner loop */
2303 				done = 1;	/* exit outer loop */
2304 				break;
2305 
2306 			case O_FORWARD_IP:
2307 				if (args->eh)	/* not valid on layer2 pkts */
2308 					break;
2309 				if (q == NULL || q->rule != f ||
2310 				    dyn_dir == MATCH_FORWARD) {
2311 				    struct sockaddr_in *sa;
2312 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2313 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2314 					bcopy(sa, &args->hopstore,
2315 							sizeof(*sa));
2316 					args->hopstore.sin_addr.s_addr =
2317 						    htonl(tablearg);
2318 					args->next_hop = &args->hopstore;
2319 				    } else {
2320 					args->next_hop = sa;
2321 				    }
2322 				}
2323 				retval = IP_FW_PASS;
2324 				l = 0;          /* exit inner loop */
2325 				done = 1;       /* exit outer loop */
2326 				break;
2327 
2328 #ifdef INET6
2329 			case O_FORWARD_IP6:
2330 				if (args->eh)	/* not valid on layer2 pkts */
2331 					break;
2332 				if (q == NULL || q->rule != f ||
2333 				    dyn_dir == MATCH_FORWARD) {
2334 					struct sockaddr_in6 *sin6;
2335 
2336 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2337 					args->next_hop6 = sin6;
2338 				}
2339 				retval = IP_FW_PASS;
2340 				l = 0;		/* exit inner loop */
2341 				done = 1;	/* exit outer loop */
2342 				break;
2343 #endif
2344 
2345 			case O_NETGRAPH:
2346 			case O_NGTEE:
2347 				set_match(args, f_pos, chain);
2348 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2349 					tablearg : cmd->arg1;
2350 				if (V_fw_one_pass)
2351 					args->rule.info |= IPFW_ONEPASS;
2352 				retval = (cmd->opcode == O_NETGRAPH) ?
2353 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2354 				l = 0;          /* exit inner loop */
2355 				done = 1;       /* exit outer loop */
2356 				break;
2357 
2358 			case O_SETFIB: {
2359 				uint32_t fib;
2360 
2361 				f->pcnt++;	/* update stats */
2362 				f->bcnt += pktlen;
2363 				f->timestamp = time_uptime;
2364 				fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg:
2365 				    cmd->arg1;
2366 				if (fib >= rt_numfibs)
2367 					fib = 0;
2368 				M_SETFIB(m, fib);
2369 				args->f_id.fib = fib;
2370 				l = 0;		/* exit inner loop */
2371 				break;
2372 		        }
2373 
2374 			case O_NAT:
2375  				if (!IPFW_NAT_LOADED) {
2376 				    retval = IP_FW_DENY;
2377 				} else {
2378 				    struct cfg_nat *t;
2379 				    int nat_id;
2380 
2381 				    set_match(args, f_pos, chain);
2382 				    /* Check if this is 'global' nat rule */
2383 				    if (cmd->arg1 == 0) {
2384 					    retval = ipfw_nat_ptr(args, NULL, m);
2385 					    l = 0;
2386 					    done = 1;
2387 					    break;
2388 				    }
2389 				    t = ((ipfw_insn_nat *)cmd)->nat;
2390 				    if (t == NULL) {
2391 					nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
2392 						tablearg : cmd->arg1;
2393 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2394 
2395 					if (t == NULL) {
2396 					    retval = IP_FW_DENY;
2397 					    l = 0;	/* exit inner loop */
2398 					    done = 1;	/* exit outer loop */
2399 					    break;
2400 					}
2401 					if (cmd->arg1 != IP_FW_TABLEARG)
2402 					    ((ipfw_insn_nat *)cmd)->nat = t;
2403 				    }
2404 				    retval = ipfw_nat_ptr(args, t, m);
2405 				}
2406 				l = 0;          /* exit inner loop */
2407 				done = 1;       /* exit outer loop */
2408 				break;
2409 
2410 			case O_REASS: {
2411 				int ip_off;
2412 
2413 				f->pcnt++;
2414 				f->bcnt += pktlen;
2415 				l = 0;	/* in any case exit inner loop */
2416 				ip_off = ntohs(ip->ip_off);
2417 
2418 				/* if not fragmented, go to next rule */
2419 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2420 				    break;
2421 				/*
2422 				 * ip_reass() expects len & off in host
2423 				 * byte order.
2424 				 */
2425 				SET_HOST_IPLEN(ip);
2426 
2427 				args->m = m = ip_reass(m);
2428 
2429 				/*
2430 				 * do IP header checksum fixup.
2431 				 */
2432 				if (m == NULL) { /* fragment got swallowed */
2433 				    retval = IP_FW_DENY;
2434 				} else { /* good, packet complete */
2435 				    int hlen;
2436 
2437 				    ip = mtod(m, struct ip *);
2438 				    hlen = ip->ip_hl << 2;
2439 				    SET_NET_IPLEN(ip);
2440 				    ip->ip_sum = 0;
2441 				    if (hlen == sizeof(struct ip))
2442 					ip->ip_sum = in_cksum_hdr(ip);
2443 				    else
2444 					ip->ip_sum = in_cksum(m, hlen);
2445 				    retval = IP_FW_REASS;
2446 				    set_match(args, f_pos, chain);
2447 				}
2448 				done = 1;	/* exit outer loop */
2449 				break;
2450 			}
2451 
2452 			default:
2453 				panic("-- unknown opcode %d\n", cmd->opcode);
2454 			} /* end of switch() on opcodes */
2455 			/*
2456 			 * if we get here with l=0, then match is irrelevant.
2457 			 */
2458 
2459 			if (cmd->len & F_NOT)
2460 				match = !match;
2461 
2462 			if (match) {
2463 				if (cmd->len & F_OR)
2464 					skip_or = 1;
2465 			} else {
2466 				if (!(cmd->len & F_OR)) /* not an OR block, */
2467 					break;		/* try next rule    */
2468 			}
2469 
2470 		}	/* end of inner loop, scan opcodes */
2471 #undef PULLUP_LEN
2472 
2473 		if (done)
2474 			break;
2475 
2476 /* next_rule:; */	/* try next rule		*/
2477 
2478 	}		/* end of outer for, scan rules */
2479 
2480 	if (done) {
2481 		struct ip_fw *rule = chain->map[f_pos];
2482 		/* Update statistics */
2483 		rule->pcnt++;
2484 		rule->bcnt += pktlen;
2485 		rule->timestamp = time_uptime;
2486 	} else {
2487 		retval = IP_FW_DENY;
2488 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2489 	}
2490 	IPFW_RUNLOCK(chain);
2491 #ifdef __FreeBSD__
2492 	if (ucred_cache != NULL)
2493 		crfree(ucred_cache);
2494 #endif
2495 	return (retval);
2496 
2497 pullup_failed:
2498 	if (V_fw_verbose)
2499 		printf("ipfw: pullup failed\n");
2500 	return (IP_FW_DENY);
2501 }
2502 
2503 /*
2504  * Set maximum number of tables that can be used in given VNET ipfw instance.
2505  */
2506 #ifdef SYSCTL_NODE
2507 static int
2508 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2509 {
2510 	int error;
2511 	unsigned int ntables;
2512 
2513 	ntables = V_fw_tables_max;
2514 
2515 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2516 	/* Read operation or some error */
2517 	if ((error != 0) || (req->newptr == NULL))
2518 		return (error);
2519 
2520 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2521 }
2522 #endif
2523 /*
2524  * Module and VNET glue
2525  */
2526 
2527 /*
2528  * Stuff that must be initialised only on boot or module load
2529  */
2530 static int
2531 ipfw_init(void)
2532 {
2533 	int error = 0;
2534 
2535 	ipfw_dyn_attach();
2536 	/*
2537  	 * Only print out this stuff the first time around,
2538 	 * when called from the sysinit code.
2539 	 */
2540 	printf("ipfw2 "
2541 #ifdef INET6
2542 		"(+ipv6) "
2543 #endif
2544 		"initialized, divert %s, nat %s, "
2545 		"rule-based forwarding "
2546 #ifdef IPFIREWALL_FORWARD
2547 		"enabled, "
2548 #else
2549 		"disabled, "
2550 #endif
2551 		"default to %s, logging ",
2552 #ifdef IPDIVERT
2553 		"enabled",
2554 #else
2555 		"loadable",
2556 #endif
2557 #ifdef IPFIREWALL_NAT
2558 		"enabled",
2559 #else
2560 		"loadable",
2561 #endif
2562 		default_to_accept ? "accept" : "deny");
2563 
2564 	/*
2565 	 * Note: V_xxx variables can be accessed here but the vnet specific
2566 	 * initializer may not have been called yet for the VIMAGE case.
2567 	 * Tuneables will have been processed. We will print out values for
2568 	 * the default vnet.
2569 	 * XXX This should all be rationalized AFTER 8.0
2570 	 */
2571 	if (V_fw_verbose == 0)
2572 		printf("disabled\n");
2573 	else if (V_verbose_limit == 0)
2574 		printf("unlimited\n");
2575 	else
2576 		printf("limited to %d packets/entry by default\n",
2577 		    V_verbose_limit);
2578 
2579 	/* Check user-supplied table count for validness */
2580 	if (default_fw_tables > IPFW_TABLES_MAX)
2581 	  default_fw_tables = IPFW_TABLES_MAX;
2582 
2583 	ipfw_log_bpf(1); /* init */
2584 	return (error);
2585 }
2586 
2587 /*
2588  * Called for the removal of the last instance only on module unload.
2589  */
2590 static void
2591 ipfw_destroy(void)
2592 {
2593 
2594 	ipfw_log_bpf(0); /* uninit */
2595 	ipfw_dyn_detach();
2596 	printf("IP firewall unloaded\n");
2597 }
2598 
2599 /*
2600  * Stuff that must be initialized for every instance
2601  * (including the first of course).
2602  */
2603 static int
2604 vnet_ipfw_init(const void *unused)
2605 {
2606 	int error;
2607 	struct ip_fw *rule = NULL;
2608 	struct ip_fw_chain *chain;
2609 
2610 	chain = &V_layer3_chain;
2611 
2612 	/* First set up some values that are compile time options */
2613 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2614 	V_fw_deny_unknown_exthdrs = 1;
2615 #ifdef IPFIREWALL_VERBOSE
2616 	V_fw_verbose = 1;
2617 #endif
2618 #ifdef IPFIREWALL_VERBOSE_LIMIT
2619 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2620 #endif
2621 #ifdef IPFIREWALL_NAT
2622 	LIST_INIT(&chain->nat);
2623 #endif
2624 
2625 	/* insert the default rule and create the initial map */
2626 	chain->n_rules = 1;
2627 	chain->static_len = sizeof(struct ip_fw);
2628 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2629 	if (chain->map)
2630 		rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2631 
2632 	/* Set initial number of tables */
2633 	V_fw_tables_max = default_fw_tables;
2634 	error = ipfw_init_tables(chain);
2635 	if (error) {
2636 		printf("ipfw2: setting up tables failed\n");
2637 		free(chain->map, M_IPFW);
2638 		free(rule, M_IPFW);
2639 		return (ENOSPC);
2640 	}
2641 
2642 	/* fill and insert the default rule */
2643 	rule->act_ofs = 0;
2644 	rule->rulenum = IPFW_DEFAULT_RULE;
2645 	rule->cmd_len = 1;
2646 	rule->set = RESVD_SET;
2647 	rule->cmd[0].len = 1;
2648 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2649 	chain->rules = chain->default_rule = chain->map[0] = rule;
2650 	chain->id = rule->id = 1;
2651 
2652 	IPFW_LOCK_INIT(chain);
2653 	ipfw_dyn_init();
2654 
2655 	/* First set up some values that are compile time options */
2656 	V_ipfw_vnet_ready = 1;		/* Open for business */
2657 
2658 	/*
2659 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2660 	 * Even if the latter two fail we still keep the module alive
2661 	 * because the sockopt and layer2 paths are still useful.
2662 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2663 	 * so we can ignore the exact return value and just set a flag.
2664 	 *
2665 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2666 	 * changes in the underlying (per-vnet) variables trigger
2667 	 * immediate hook()/unhook() calls.
2668 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2669 	 * is checked on each packet because there are no pfil hooks.
2670 	 */
2671 	V_ip_fw_ctl_ptr = ipfw_ctl;
2672 	error = ipfw_attach_hooks(1);
2673 	return (error);
2674 }
2675 
2676 /*
2677  * Called for the removal of each instance.
2678  */
2679 static int
2680 vnet_ipfw_uninit(const void *unused)
2681 {
2682 	struct ip_fw *reap, *rule;
2683 	struct ip_fw_chain *chain = &V_layer3_chain;
2684 	int i;
2685 
2686 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2687 	/*
2688 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2689 	 * Then grab, release and grab again the WLOCK so we make
2690 	 * sure the update is propagated and nobody will be in.
2691 	 */
2692 	(void)ipfw_attach_hooks(0 /* detach */);
2693 	V_ip_fw_ctl_ptr = NULL;
2694 	IPFW_UH_WLOCK(chain);
2695 	IPFW_UH_WUNLOCK(chain);
2696 	IPFW_UH_WLOCK(chain);
2697 
2698 	IPFW_WLOCK(chain);
2699 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2700 	IPFW_WUNLOCK(chain);
2701 
2702 	ipfw_destroy_tables(chain);
2703 	reap = NULL;
2704 	IPFW_WLOCK(chain);
2705 	for (i = 0; i < chain->n_rules; i++) {
2706 		rule = chain->map[i];
2707 		rule->x_next = reap;
2708 		reap = rule;
2709 	}
2710 	if (chain->map)
2711 		free(chain->map, M_IPFW);
2712 	IPFW_WUNLOCK(chain);
2713 	IPFW_UH_WUNLOCK(chain);
2714 	if (reap != NULL)
2715 		ipfw_reap_rules(reap);
2716 	IPFW_LOCK_DESTROY(chain);
2717 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2718 	return 0;
2719 }
2720 
2721 /*
2722  * Module event handler.
2723  * In general we have the choice of handling most of these events by the
2724  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2725  * use the SYSINIT handlers as they are more capable of expressing the
2726  * flow of control during module and vnet operations, so this is just
2727  * a skeleton. Note there is no SYSINIT equivalent of the module
2728  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2729  */
2730 static int
2731 ipfw_modevent(module_t mod, int type, void *unused)
2732 {
2733 	int err = 0;
2734 
2735 	switch (type) {
2736 	case MOD_LOAD:
2737 		/* Called once at module load or
2738 	 	 * system boot if compiled in. */
2739 		break;
2740 	case MOD_QUIESCE:
2741 		/* Called before unload. May veto unloading. */
2742 		break;
2743 	case MOD_UNLOAD:
2744 		/* Called during unload. */
2745 		break;
2746 	case MOD_SHUTDOWN:
2747 		/* Called during system shutdown. */
2748 		break;
2749 	default:
2750 		err = EOPNOTSUPP;
2751 		break;
2752 	}
2753 	return err;
2754 }
2755 
2756 static moduledata_t ipfwmod = {
2757 	"ipfw",
2758 	ipfw_modevent,
2759 	0
2760 };
2761 
2762 /* Define startup order. */
2763 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2764 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2765 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2766 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2767 
2768 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2769 MODULE_VERSION(ipfw, 2);
2770 /* should declare some dependencies here */
2771 
2772 /*
2773  * Starting up. Done in order after ipfwmod() has been called.
2774  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2775  */
2776 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2777 	    ipfw_init, NULL);
2778 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2779 	    vnet_ipfw_init, NULL);
2780 
2781 /*
2782  * Closing up shop. These are done in REVERSE ORDER, but still
2783  * after ipfwmod() has been called. Not called on reboot.
2784  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2785  * or when the module is unloaded.
2786  */
2787 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2788 	    ipfw_destroy, NULL);
2789 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2790 	    vnet_ipfw_uninit, NULL);
2791 /* end of file */
2792