xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 273c26a3c3bea87a241d6879abd4f991db180bf0)
1 /*-
2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
28 
29 /*
30  * The FreeBSD IP packet firewall, main file
31  */
32 
33 #include "opt_ipfw.h"
34 #include "opt_ipdivert.h"
35 #include "opt_inet.h"
36 #ifndef INET
37 #error "IPFIREWALL requires INET"
38 #endif /* INET */
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/condvar.h>
45 #include <sys/counter.h>
46 #include <sys/eventhandler.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/jail.h>
52 #include <sys/module.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/rwlock.h>
56 #include <sys/rmlock.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/syslog.h>
61 #include <sys/ucred.h>
62 #include <net/ethernet.h> /* for ETHERTYPE_IP */
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #include <net/pfil.h>
67 #include <net/vnet.h>
68 
69 #include <netpfil/pf/pf_mtag.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet/pim.h>
80 #include <netinet/tcp_var.h>
81 #include <netinet/udp.h>
82 #include <netinet/udp_var.h>
83 #include <netinet/sctp.h>
84 
85 #include <netinet/ip6.h>
86 #include <netinet/icmp6.h>
87 #include <netinet/in_fib.h>
88 #ifdef INET6
89 #include <netinet6/in6_fib.h>
90 #include <netinet6/in6_pcb.h>
91 #include <netinet6/scope6_var.h>
92 #include <netinet6/ip6_var.h>
93 #endif
94 
95 #include <netpfil/ipfw/ip_fw_private.h>
96 
97 #include <machine/in_cksum.h>	/* XXX for in_cksum */
98 
99 #ifdef MAC
100 #include <security/mac/mac_framework.h>
101 #endif
102 
103 /*
104  * static variables followed by global ones.
105  * All ipfw global variables are here.
106  */
107 
108 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
109 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
110 
111 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
112 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
113 
114 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
115 static int default_to_accept = 1;
116 #else
117 static int default_to_accept;
118 #endif
119 
120 VNET_DEFINE(int, autoinc_step);
121 VNET_DEFINE(int, fw_one_pass) = 1;
122 
123 VNET_DEFINE(unsigned int, fw_tables_max);
124 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;	/* Don't use set-aware tables */
125 /* Use 128 tables by default */
126 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
127 
128 #ifndef LINEAR_SKIPTO
129 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
130     int tablearg, int jump_backwards);
131 #define	JUMP(ch, f, num, targ, back)	jump_fast(ch, f, num, targ, back)
132 #else
133 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
134     int tablearg, int jump_backwards);
135 #define	JUMP(ch, f, num, targ, back)	jump_linear(ch, f, num, targ, back)
136 #endif
137 
138 /*
139  * Each rule belongs to one of 32 different sets (0..31).
140  * The variable set_disable contains one bit per set.
141  * If the bit is set, all rules in the corresponding set
142  * are disabled. Set RESVD_SET(31) is reserved for the default rule
143  * and rules that are not deleted by the flush command,
144  * and CANNOT be disabled.
145  * Rules in set RESVD_SET can only be deleted individually.
146  */
147 VNET_DEFINE(u_int32_t, set_disable);
148 #define	V_set_disable			VNET(set_disable)
149 
150 VNET_DEFINE(int, fw_verbose);
151 /* counter for ipfw_log(NULL...) */
152 VNET_DEFINE(u_int64_t, norule_counter);
153 VNET_DEFINE(int, verbose_limit);
154 
155 /* layer3_chain contains the list of rules for layer 3 */
156 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
157 
158 /* ipfw_vnet_ready controls when we are open for business */
159 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
160 
161 VNET_DEFINE(int, ipfw_nat_ready) = 0;
162 
163 ipfw_nat_t *ipfw_nat_ptr = NULL;
164 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
165 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
166 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
167 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
168 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
169 
170 #ifdef SYSCTL_NODE
171 uint32_t dummy_def = IPFW_DEFAULT_RULE;
172 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
173 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
174 
175 SYSBEGIN(f3)
176 
177 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
178 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
179     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
180     "Only do a single pass through ipfw when using dummynet(4)");
181 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
182     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
183     "Rule number auto-increment step");
184 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
185     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
186     "Log matches to ipfw rules");
187 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
188     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
189     "Set upper limit of matches of ipfw rules logged");
190 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
191     &dummy_def, 0,
192     "The default/max possible rule number.");
193 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
194     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
195     "Maximum number of concurrently used tables");
196 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
197     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
198     0, 0, sysctl_ipfw_tables_sets, "IU",
199     "Use per-set namespace for tables");
200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
201     &default_to_accept, 0,
202     "Make the default rule accept all packets.");
203 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
204 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
205     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
206     "Number of static rules");
207 
208 #ifdef INET6
209 SYSCTL_DECL(_net_inet6_ip6);
210 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
211 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
212     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
213     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
214     "Deny packets with unknown IPv6 Extension Headers");
215 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
216     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
217     &VNET_NAME(fw_permit_single_frag6), 0,
218     "Permit single packet IPv6 fragments");
219 #endif /* INET6 */
220 
221 SYSEND
222 
223 #endif /* SYSCTL_NODE */
224 
225 
226 /*
227  * Some macros used in the various matching options.
228  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
229  * Other macros just cast void * into the appropriate type
230  */
231 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
232 #define	TCP(p)		((struct tcphdr *)(p))
233 #define	SCTP(p)		((struct sctphdr *)(p))
234 #define	UDP(p)		((struct udphdr *)(p))
235 #define	ICMP(p)		((struct icmphdr *)(p))
236 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
237 
238 static __inline int
239 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
240 {
241 	int type = icmp->icmp_type;
242 
243 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
244 }
245 
246 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
247     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
248 
249 static int
250 is_icmp_query(struct icmphdr *icmp)
251 {
252 	int type = icmp->icmp_type;
253 
254 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
255 }
256 #undef TT
257 
258 /*
259  * The following checks use two arrays of 8 or 16 bits to store the
260  * bits that we want set or clear, respectively. They are in the
261  * low and high half of cmd->arg1 or cmd->d[0].
262  *
263  * We scan options and store the bits we find set. We succeed if
264  *
265  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
266  *
267  * The code is sometimes optimized not to store additional variables.
268  */
269 
270 static int
271 flags_match(ipfw_insn *cmd, u_int8_t bits)
272 {
273 	u_char want_clear;
274 	bits = ~bits;
275 
276 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
277 		return 0; /* some bits we want set were clear */
278 	want_clear = (cmd->arg1 >> 8) & 0xff;
279 	if ( (want_clear & bits) != want_clear)
280 		return 0; /* some bits we want clear were set */
281 	return 1;
282 }
283 
284 static int
285 ipopts_match(struct ip *ip, ipfw_insn *cmd)
286 {
287 	int optlen, bits = 0;
288 	u_char *cp = (u_char *)(ip + 1);
289 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
290 
291 	for (; x > 0; x -= optlen, cp += optlen) {
292 		int opt = cp[IPOPT_OPTVAL];
293 
294 		if (opt == IPOPT_EOL)
295 			break;
296 		if (opt == IPOPT_NOP)
297 			optlen = 1;
298 		else {
299 			optlen = cp[IPOPT_OLEN];
300 			if (optlen <= 0 || optlen > x)
301 				return 0; /* invalid or truncated */
302 		}
303 		switch (opt) {
304 
305 		default:
306 			break;
307 
308 		case IPOPT_LSRR:
309 			bits |= IP_FW_IPOPT_LSRR;
310 			break;
311 
312 		case IPOPT_SSRR:
313 			bits |= IP_FW_IPOPT_SSRR;
314 			break;
315 
316 		case IPOPT_RR:
317 			bits |= IP_FW_IPOPT_RR;
318 			break;
319 
320 		case IPOPT_TS:
321 			bits |= IP_FW_IPOPT_TS;
322 			break;
323 		}
324 	}
325 	return (flags_match(cmd, bits));
326 }
327 
328 static int
329 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
330 {
331 	int optlen, bits = 0;
332 	u_char *cp = (u_char *)(tcp + 1);
333 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
334 
335 	for (; x > 0; x -= optlen, cp += optlen) {
336 		int opt = cp[0];
337 		if (opt == TCPOPT_EOL)
338 			break;
339 		if (opt == TCPOPT_NOP)
340 			optlen = 1;
341 		else {
342 			optlen = cp[1];
343 			if (optlen <= 0)
344 				break;
345 		}
346 
347 		switch (opt) {
348 
349 		default:
350 			break;
351 
352 		case TCPOPT_MAXSEG:
353 			bits |= IP_FW_TCPOPT_MSS;
354 			break;
355 
356 		case TCPOPT_WINDOW:
357 			bits |= IP_FW_TCPOPT_WINDOW;
358 			break;
359 
360 		case TCPOPT_SACK_PERMITTED:
361 		case TCPOPT_SACK:
362 			bits |= IP_FW_TCPOPT_SACK;
363 			break;
364 
365 		case TCPOPT_TIMESTAMP:
366 			bits |= IP_FW_TCPOPT_TS;
367 			break;
368 
369 		}
370 	}
371 	return (flags_match(cmd, bits));
372 }
373 
374 static int
375 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
376     uint32_t *tablearg)
377 {
378 
379 	if (ifp == NULL)	/* no iface with this packet, match fails */
380 		return (0);
381 
382 	/* Check by name or by IP address */
383 	if (cmd->name[0] != '\0') { /* match by name */
384 		if (cmd->name[0] == '\1') /* use tablearg to match */
385 			return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0,
386 				&ifp->if_index, tablearg);
387 		/* Check name */
388 		if (cmd->p.glob) {
389 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
390 				return(1);
391 		} else {
392 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
393 				return(1);
394 		}
395 	} else {
396 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
397 		struct ifaddr *ia;
398 
399 		if_addr_rlock(ifp);
400 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
401 			if (ia->ifa_addr->sa_family != AF_INET)
402 				continue;
403 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
404 			    (ia->ifa_addr))->sin_addr.s_addr) {
405 				if_addr_runlock(ifp);
406 				return(1);	/* match */
407 			}
408 		}
409 		if_addr_runlock(ifp);
410 #endif /* __FreeBSD__ */
411 	}
412 	return(0);	/* no match, fail ... */
413 }
414 
415 /*
416  * The verify_path function checks if a route to the src exists and
417  * if it is reachable via ifp (when provided).
418  *
419  * The 'verrevpath' option checks that the interface that an IP packet
420  * arrives on is the same interface that traffic destined for the
421  * packet's source address would be routed out of.
422  * The 'versrcreach' option just checks that the source address is
423  * reachable via any route (except default) in the routing table.
424  * These two are a measure to block forged packets. This is also
425  * commonly known as "anti-spoofing" or Unicast Reverse Path
426  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
427  * is purposely reminiscent of the Cisco IOS command,
428  *
429  *   ip verify unicast reverse-path
430  *   ip verify unicast source reachable-via any
431  *
432  * which implements the same functionality. But note that the syntax
433  * is misleading, and the check may be performed on all IP packets
434  * whether unicast, multicast, or broadcast.
435  */
436 static int
437 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
438 {
439 #if defined(USERSPACE) || !defined(__FreeBSD__)
440 	return 0;
441 #else
442 	struct nhop4_basic nh4;
443 
444 	if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
445 		return (0);
446 
447 	/*
448 	 * If ifp is provided, check for equality with rtentry.
449 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
450 	 * in order to pass packets injected back by if_simloop():
451 	 * routing entry (via lo0) for our own address
452 	 * may exist, so we need to handle routing assymetry.
453 	 */
454 	if (ifp != NULL && ifp != nh4.nh_ifp)
455 		return (0);
456 
457 	/* if no ifp provided, check if rtentry is not default route */
458 	if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
459 		return (0);
460 
461 	/* or if this is a blackhole/reject route */
462 	if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
463 		return (0);
464 
465 	/* found valid route */
466 	return 1;
467 #endif /* __FreeBSD__ */
468 }
469 
470 #ifdef INET6
471 /*
472  * ipv6 specific rules here...
473  */
474 static __inline int
475 icmp6type_match (int type, ipfw_insn_u32 *cmd)
476 {
477 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
478 }
479 
480 static int
481 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
482 {
483 	int i;
484 	for (i=0; i <= cmd->o.arg1; ++i )
485 		if (curr_flow == cmd->d[i] )
486 			return 1;
487 	return 0;
488 }
489 
490 /* support for IP6_*_ME opcodes */
491 static const struct in6_addr lla_mask = {{{
492 	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
493 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
494 }}};
495 
496 static int
497 ipfw_localip6(struct in6_addr *in6)
498 {
499 	struct rm_priotracker in6_ifa_tracker;
500 	struct in6_ifaddr *ia;
501 
502 	if (IN6_IS_ADDR_MULTICAST(in6))
503 		return (0);
504 
505 	if (!IN6_IS_ADDR_LINKLOCAL(in6))
506 		return (in6_localip(in6));
507 
508 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
509 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
510 		if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
511 			continue;
512 		if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
513 		    in6, &lla_mask)) {
514 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
515 			return (1);
516 		}
517 	}
518 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
519 	return (0);
520 }
521 
522 static int
523 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
524 {
525 	struct nhop6_basic nh6;
526 
527 	if (IN6_IS_SCOPE_LINKLOCAL(src))
528 		return (1);
529 
530 	if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
531 		return (0);
532 
533 	/* If ifp is provided, check for equality with route table. */
534 	if (ifp != NULL && ifp != nh6.nh_ifp)
535 		return (0);
536 
537 	/* if no ifp provided, check if rtentry is not default route */
538 	if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
539 		return (0);
540 
541 	/* or if this is a blackhole/reject route */
542 	if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
543 		return (0);
544 
545 	/* found valid route */
546 	return 1;
547 }
548 
549 static int
550 is_icmp6_query(int icmp6_type)
551 {
552 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
553 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
554 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
555 	    icmp6_type == ICMP6_WRUREQUEST ||
556 	    icmp6_type == ICMP6_FQDN_QUERY ||
557 	    icmp6_type == ICMP6_NI_QUERY))
558 		return (1);
559 
560 	return (0);
561 }
562 
563 static void
564 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
565 {
566 	struct mbuf *m;
567 
568 	m = args->m;
569 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
570 		struct tcphdr *tcp;
571 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
572 
573 		if ((tcp->th_flags & TH_RST) == 0) {
574 			struct mbuf *m0;
575 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
576 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
577 			    tcp->th_flags | TH_RST);
578 			if (m0 != NULL)
579 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
580 				    NULL);
581 		}
582 		FREE_PKT(m);
583 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
584 #if 0
585 		/*
586 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
587 		 * as the contents are read. We need to m_adj() the
588 		 * needed amount.
589 		 * The mbuf will however be thrown away so we can adjust it.
590 		 * Remember we did an m_pullup on it already so we
591 		 * can make some assumptions about contiguousness.
592 		 */
593 		if (args->L3offset)
594 			m_adj(m, args->L3offset);
595 #endif
596 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
597 	} else
598 		FREE_PKT(m);
599 
600 	args->m = NULL;
601 }
602 
603 #endif /* INET6 */
604 
605 
606 /*
607  * sends a reject message, consuming the mbuf passed as an argument.
608  */
609 static void
610 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
611 {
612 
613 #if 0
614 	/* XXX When ip is not guaranteed to be at mtod() we will
615 	 * need to account for this */
616 	 * The mbuf will however be thrown away so we can adjust it.
617 	 * Remember we did an m_pullup on it already so we
618 	 * can make some assumptions about contiguousness.
619 	 */
620 	if (args->L3offset)
621 		m_adj(m, args->L3offset);
622 #endif
623 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
624 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
625 	} else if (args->f_id.proto == IPPROTO_TCP) {
626 		struct tcphdr *const tcp =
627 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
628 		if ( (tcp->th_flags & TH_RST) == 0) {
629 			struct mbuf *m;
630 			m = ipfw_send_pkt(args->m, &(args->f_id),
631 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
632 				tcp->th_flags | TH_RST);
633 			if (m != NULL)
634 				ip_output(m, NULL, NULL, 0, NULL, NULL);
635 		}
636 		FREE_PKT(args->m);
637 	} else
638 		FREE_PKT(args->m);
639 	args->m = NULL;
640 }
641 
642 /*
643  * Support for uid/gid/jail lookup. These tests are expensive
644  * (because we may need to look into the list of active sockets)
645  * so we cache the results. ugid_lookupp is 0 if we have not
646  * yet done a lookup, 1 if we succeeded, and -1 if we tried
647  * and failed. The function always returns the match value.
648  * We could actually spare the variable and use *uc, setting
649  * it to '(void *)check_uidgid if we have no info, NULL if
650  * we tried and failed, or any other value if successful.
651  */
652 static int
653 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
654     struct ucred **uc)
655 {
656 #if defined(USERSPACE)
657 	return 0;	// not supported in userspace
658 #else
659 #ifndef __FreeBSD__
660 	/* XXX */
661 	return cred_check(insn, proto, oif,
662 	    dst_ip, dst_port, src_ip, src_port,
663 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
664 #else  /* FreeBSD */
665 	struct in_addr src_ip, dst_ip;
666 	struct inpcbinfo *pi;
667 	struct ipfw_flow_id *id;
668 	struct inpcb *pcb, *inp;
669 	struct ifnet *oif;
670 	int lookupflags;
671 	int match;
672 
673 	id = &args->f_id;
674 	inp = args->inp;
675 	oif = args->oif;
676 
677 	/*
678 	 * Check to see if the UDP or TCP stack supplied us with
679 	 * the PCB. If so, rather then holding a lock and looking
680 	 * up the PCB, we can use the one that was supplied.
681 	 */
682 	if (inp && *ugid_lookupp == 0) {
683 		INP_LOCK_ASSERT(inp);
684 		if (inp->inp_socket != NULL) {
685 			*uc = crhold(inp->inp_cred);
686 			*ugid_lookupp = 1;
687 		} else
688 			*ugid_lookupp = -1;
689 	}
690 	/*
691 	 * If we have already been here and the packet has no
692 	 * PCB entry associated with it, then we can safely
693 	 * assume that this is a no match.
694 	 */
695 	if (*ugid_lookupp == -1)
696 		return (0);
697 	if (id->proto == IPPROTO_TCP) {
698 		lookupflags = 0;
699 		pi = &V_tcbinfo;
700 	} else if (id->proto == IPPROTO_UDP) {
701 		lookupflags = INPLOOKUP_WILDCARD;
702 		pi = &V_udbinfo;
703 	} else
704 		return 0;
705 	lookupflags |= INPLOOKUP_RLOCKPCB;
706 	match = 0;
707 	if (*ugid_lookupp == 0) {
708 		if (id->addr_type == 6) {
709 #ifdef INET6
710 			if (oif == NULL)
711 				pcb = in6_pcblookup_mbuf(pi,
712 				    &id->src_ip6, htons(id->src_port),
713 				    &id->dst_ip6, htons(id->dst_port),
714 				    lookupflags, oif, args->m);
715 			else
716 				pcb = in6_pcblookup_mbuf(pi,
717 				    &id->dst_ip6, htons(id->dst_port),
718 				    &id->src_ip6, htons(id->src_port),
719 				    lookupflags, oif, args->m);
720 #else
721 			*ugid_lookupp = -1;
722 			return (0);
723 #endif
724 		} else {
725 			src_ip.s_addr = htonl(id->src_ip);
726 			dst_ip.s_addr = htonl(id->dst_ip);
727 			if (oif == NULL)
728 				pcb = in_pcblookup_mbuf(pi,
729 				    src_ip, htons(id->src_port),
730 				    dst_ip, htons(id->dst_port),
731 				    lookupflags, oif, args->m);
732 			else
733 				pcb = in_pcblookup_mbuf(pi,
734 				    dst_ip, htons(id->dst_port),
735 				    src_ip, htons(id->src_port),
736 				    lookupflags, oif, args->m);
737 		}
738 		if (pcb != NULL) {
739 			INP_RLOCK_ASSERT(pcb);
740 			*uc = crhold(pcb->inp_cred);
741 			*ugid_lookupp = 1;
742 			INP_RUNLOCK(pcb);
743 		}
744 		if (*ugid_lookupp == 0) {
745 			/*
746 			 * We tried and failed, set the variable to -1
747 			 * so we will not try again on this packet.
748 			 */
749 			*ugid_lookupp = -1;
750 			return (0);
751 		}
752 	}
753 	if (insn->o.opcode == O_UID)
754 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
755 	else if (insn->o.opcode == O_GID)
756 		match = groupmember((gid_t)insn->d[0], *uc);
757 	else if (insn->o.opcode == O_JAIL)
758 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
759 	return (match);
760 #endif /* __FreeBSD__ */
761 #endif /* not supported in userspace */
762 }
763 
764 /*
765  * Helper function to set args with info on the rule after the matching
766  * one. slot is precise, whereas we guess rule_id as they are
767  * assigned sequentially.
768  */
769 static inline void
770 set_match(struct ip_fw_args *args, int slot,
771 	struct ip_fw_chain *chain)
772 {
773 	args->rule.chain_id = chain->id;
774 	args->rule.slot = slot + 1; /* we use 0 as a marker */
775 	args->rule.rule_id = 1 + chain->map[slot]->id;
776 	args->rule.rulenum = chain->map[slot]->rulenum;
777 }
778 
779 #ifndef LINEAR_SKIPTO
780 /*
781  * Helper function to enable cached rule lookups using
782  * cached_id and cached_pos fields in ipfw rule.
783  */
784 static int
785 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
786     int tablearg, int jump_backwards)
787 {
788 	int f_pos;
789 
790 	/* If possible use cached f_pos (in f->cached_pos),
791 	 * whose version is written in f->cached_id
792 	 * (horrible hacks to avoid changing the ABI).
793 	 */
794 	if (num != IP_FW_TARG && f->cached_id == chain->id)
795 		f_pos = f->cached_pos;
796 	else {
797 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
798 		/* make sure we do not jump backward */
799 		if (jump_backwards == 0 && i <= f->rulenum)
800 			i = f->rulenum + 1;
801 		if (chain->idxmap != NULL)
802 			f_pos = chain->idxmap[i];
803 		else
804 			f_pos = ipfw_find_rule(chain, i, 0);
805 		/* update the cache */
806 		if (num != IP_FW_TARG) {
807 			f->cached_id = chain->id;
808 			f->cached_pos = f_pos;
809 		}
810 	}
811 
812 	return (f_pos);
813 }
814 #else
815 /*
816  * Helper function to enable real fast rule lookups.
817  */
818 static int
819 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
820     int tablearg, int jump_backwards)
821 {
822 	int f_pos;
823 
824 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
825 	/* make sure we do not jump backward */
826 	if (jump_backwards == 0 && num <= f->rulenum)
827 		num = f->rulenum + 1;
828 	f_pos = chain->idxmap[num];
829 
830 	return (f_pos);
831 }
832 #endif
833 
834 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
835 /*
836  * The main check routine for the firewall.
837  *
838  * All arguments are in args so we can modify them and return them
839  * back to the caller.
840  *
841  * Parameters:
842  *
843  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
844  *		Starts with the IP header.
845  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
846  *	args->L3offset	Number of bytes bypassed if we came from L2.
847  *			e.g. often sizeof(eh)  ** NOTYET **
848  *	args->oif	Outgoing interface, NULL if packet is incoming.
849  *		The incoming interface is in the mbuf. (in)
850  *	args->divert_rule (in/out)
851  *		Skip up to the first rule past this rule number;
852  *		upon return, non-zero port number for divert or tee.
853  *
854  *	args->rule	Pointer to the last matching rule (in/out)
855  *	args->next_hop	Socket we are forwarding to (out).
856  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
857  *	args->f_id	Addresses grabbed from the packet (out)
858  * 	args->rule.info	a cookie depending on rule action
859  *
860  * Return value:
861  *
862  *	IP_FW_PASS	the packet must be accepted
863  *	IP_FW_DENY	the packet must be dropped
864  *	IP_FW_DIVERT	divert packet, port in m_tag
865  *	IP_FW_TEE	tee packet, port in m_tag
866  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
867  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
868  *		args->rule contains the matching rule,
869  *		args->rule.info has additional information.
870  *
871  */
872 int
873 ipfw_chk(struct ip_fw_args *args)
874 {
875 
876 	/*
877 	 * Local variables holding state while processing a packet:
878 	 *
879 	 * IMPORTANT NOTE: to speed up the processing of rules, there
880 	 * are some assumption on the values of the variables, which
881 	 * are documented here. Should you change them, please check
882 	 * the implementation of the various instructions to make sure
883 	 * that they still work.
884 	 *
885 	 * args->eh	The MAC header. It is non-null for a layer2
886 	 *	packet, it is NULL for a layer-3 packet.
887 	 * **notyet**
888 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
889 	 *
890 	 * m | args->m	Pointer to the mbuf, as received from the caller.
891 	 *	It may change if ipfw_chk() does an m_pullup, or if it
892 	 *	consumes the packet because it calls send_reject().
893 	 *	XXX This has to change, so that ipfw_chk() never modifies
894 	 *	or consumes the buffer.
895 	 * ip	is the beginning of the ip(4 or 6) header.
896 	 *	Calculated by adding the L3offset to the start of data.
897 	 *	(Until we start using L3offset, the packet is
898 	 *	supposed to start with the ip header).
899 	 */
900 	struct mbuf *m = args->m;
901 	struct ip *ip = mtod(m, struct ip *);
902 
903 	/*
904 	 * For rules which contain uid/gid or jail constraints, cache
905 	 * a copy of the users credentials after the pcb lookup has been
906 	 * executed. This will speed up the processing of rules with
907 	 * these types of constraints, as well as decrease contention
908 	 * on pcb related locks.
909 	 */
910 #ifndef __FreeBSD__
911 	struct bsd_ucred ucred_cache;
912 #else
913 	struct ucred *ucred_cache = NULL;
914 #endif
915 	int ucred_lookup = 0;
916 
917 	/*
918 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
919 	 *	inbound path (ether_input, ip_input).
920 	 *	If non-NULL, ipfw_chk has been called on the outbound path
921 	 *	(ether_output, ip_output).
922 	 */
923 	struct ifnet *oif = args->oif;
924 
925 	int f_pos = 0;		/* index of current rule in the array */
926 	int retval = 0;
927 
928 	/*
929 	 * hlen	The length of the IP header.
930 	 */
931 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
932 
933 	/*
934 	 * offset	The offset of a fragment. offset != 0 means that
935 	 *	we have a fragment at this offset of an IPv4 packet.
936 	 *	offset == 0 means that (if this is an IPv4 packet)
937 	 *	this is the first or only fragment.
938 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
939 	 *	or there is a single packet fragment (fragment header added
940 	 *	without needed).  We will treat a single packet fragment as if
941 	 *	there was no fragment header (or log/block depending on the
942 	 *	V_fw_permit_single_frag6 sysctl setting).
943 	 */
944 	u_short offset = 0;
945 	u_short ip6f_mf = 0;
946 
947 	/*
948 	 * Local copies of addresses. They are only valid if we have
949 	 * an IP packet.
950 	 *
951 	 * proto	The protocol. Set to 0 for non-ip packets,
952 	 *	or to the protocol read from the packet otherwise.
953 	 *	proto != 0 means that we have an IPv4 packet.
954 	 *
955 	 * src_port, dst_port	port numbers, in HOST format. Only
956 	 *	valid for TCP and UDP packets.
957 	 *
958 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
959 	 *	Only valid for IPv4 packets.
960 	 */
961 	uint8_t proto;
962 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
963 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
964 	uint16_t iplen=0;
965 	int pktlen;
966 	uint16_t	etype = 0;	/* Host order stored ether type */
967 
968 	/*
969 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
970 	 * 	MATCH_NONE when checked and not matched (q = NULL),
971 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
972 	 */
973 	int dyn_dir = MATCH_UNKNOWN;
974 	uint16_t dyn_name = 0;
975 	ipfw_dyn_rule *q = NULL;
976 	struct ip_fw_chain *chain = &V_layer3_chain;
977 
978 	/*
979 	 * We store in ulp a pointer to the upper layer protocol header.
980 	 * In the ipv4 case this is easy to determine from the header,
981 	 * but for ipv6 we might have some additional headers in the middle.
982 	 * ulp is NULL if not found.
983 	 */
984 	void *ulp = NULL;		/* upper layer protocol pointer. */
985 
986 	/* XXX ipv6 variables */
987 	int is_ipv6 = 0;
988 	uint8_t	icmp6_type = 0;
989 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
990 	/* end of ipv6 variables */
991 
992 	int is_ipv4 = 0;
993 
994 	int done = 0;		/* flag to exit the outer loop */
995 	IPFW_RLOCK_TRACKER;
996 
997 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
998 		return (IP_FW_PASS);	/* accept */
999 
1000 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1001 	src_ip.s_addr = 0;		/* make sure it is initialized */
1002 	pktlen = m->m_pkthdr.len;
1003 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1004 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
1005 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1006 
1007 /*
1008  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1009  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1010  * pointer might become stale after other pullups (but we never use it
1011  * this way).
1012  */
1013 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1014 #define PULLUP_LEN(_len, p, T)					\
1015 do {								\
1016 	int x = (_len) + T;					\
1017 	if ((m)->m_len < x) {					\
1018 		args->m = m = m_pullup(m, x);			\
1019 		if (m == NULL)					\
1020 			goto pullup_failed;			\
1021 	}							\
1022 	p = (mtod(m, char *) + (_len));				\
1023 } while (0)
1024 
1025 	/*
1026 	 * if we have an ether header,
1027 	 */
1028 	if (args->eh)
1029 		etype = ntohs(args->eh->ether_type);
1030 
1031 	/* Identify IP packets and fill up variables. */
1032 	if (pktlen >= sizeof(struct ip6_hdr) &&
1033 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1034 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1035 		is_ipv6 = 1;
1036 		args->f_id.addr_type = 6;
1037 		hlen = sizeof(struct ip6_hdr);
1038 		proto = ip6->ip6_nxt;
1039 
1040 		/* Search extension headers to find upper layer protocols */
1041 		while (ulp == NULL && offset == 0) {
1042 			switch (proto) {
1043 			case IPPROTO_ICMPV6:
1044 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1045 				icmp6_type = ICMP6(ulp)->icmp6_type;
1046 				break;
1047 
1048 			case IPPROTO_TCP:
1049 				PULLUP_TO(hlen, ulp, struct tcphdr);
1050 				dst_port = TCP(ulp)->th_dport;
1051 				src_port = TCP(ulp)->th_sport;
1052 				/* save flags for dynamic rules */
1053 				args->f_id._flags = TCP(ulp)->th_flags;
1054 				break;
1055 
1056 			case IPPROTO_SCTP:
1057 				PULLUP_TO(hlen, ulp, struct sctphdr);
1058 				src_port = SCTP(ulp)->src_port;
1059 				dst_port = SCTP(ulp)->dest_port;
1060 				break;
1061 
1062 			case IPPROTO_UDP:
1063 				PULLUP_TO(hlen, ulp, struct udphdr);
1064 				dst_port = UDP(ulp)->uh_dport;
1065 				src_port = UDP(ulp)->uh_sport;
1066 				break;
1067 
1068 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1069 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1070 				ext_hd |= EXT_HOPOPTS;
1071 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1072 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1073 				ulp = NULL;
1074 				break;
1075 
1076 			case IPPROTO_ROUTING:	/* RFC 2460 */
1077 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1078 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1079 				case 0:
1080 					ext_hd |= EXT_RTHDR0;
1081 					break;
1082 				case 2:
1083 					ext_hd |= EXT_RTHDR2;
1084 					break;
1085 				default:
1086 					if (V_fw_verbose)
1087 						printf("IPFW2: IPV6 - Unknown "
1088 						    "Routing Header type(%d)\n",
1089 						    ((struct ip6_rthdr *)
1090 						    ulp)->ip6r_type);
1091 					if (V_fw_deny_unknown_exthdrs)
1092 					    return (IP_FW_DENY);
1093 					break;
1094 				}
1095 				ext_hd |= EXT_ROUTING;
1096 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1097 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1098 				ulp = NULL;
1099 				break;
1100 
1101 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1102 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1103 				ext_hd |= EXT_FRAGMENT;
1104 				hlen += sizeof (struct ip6_frag);
1105 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1106 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1107 					IP6F_OFF_MASK;
1108 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1109 					IP6F_MORE_FRAG;
1110 				if (V_fw_permit_single_frag6 == 0 &&
1111 				    offset == 0 && ip6f_mf == 0) {
1112 					if (V_fw_verbose)
1113 						printf("IPFW2: IPV6 - Invalid "
1114 						    "Fragment Header\n");
1115 					if (V_fw_deny_unknown_exthdrs)
1116 					    return (IP_FW_DENY);
1117 					break;
1118 				}
1119 				args->f_id.extra =
1120 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1121 				ulp = NULL;
1122 				break;
1123 
1124 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1125 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1126 				ext_hd |= EXT_DSTOPTS;
1127 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1128 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1129 				ulp = NULL;
1130 				break;
1131 
1132 			case IPPROTO_AH:	/* RFC 2402 */
1133 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1134 				ext_hd |= EXT_AH;
1135 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1136 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1137 				ulp = NULL;
1138 				break;
1139 
1140 			case IPPROTO_ESP:	/* RFC 2406 */
1141 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1142 				/* Anything past Seq# is variable length and
1143 				 * data past this ext. header is encrypted. */
1144 				ext_hd |= EXT_ESP;
1145 				break;
1146 
1147 			case IPPROTO_NONE:	/* RFC 2460 */
1148 				/*
1149 				 * Packet ends here, and IPv6 header has
1150 				 * already been pulled up. If ip6e_len!=0
1151 				 * then octets must be ignored.
1152 				 */
1153 				ulp = ip; /* non-NULL to get out of loop. */
1154 				break;
1155 
1156 			case IPPROTO_OSPFIGP:
1157 				/* XXX OSPF header check? */
1158 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1159 				break;
1160 
1161 			case IPPROTO_PIM:
1162 				/* XXX PIM header check? */
1163 				PULLUP_TO(hlen, ulp, struct pim);
1164 				break;
1165 
1166 			case IPPROTO_CARP:
1167 				PULLUP_TO(hlen, ulp, struct carp_header);
1168 				if (((struct carp_header *)ulp)->carp_version !=
1169 				    CARP_VERSION)
1170 					return (IP_FW_DENY);
1171 				if (((struct carp_header *)ulp)->carp_type !=
1172 				    CARP_ADVERTISEMENT)
1173 					return (IP_FW_DENY);
1174 				break;
1175 
1176 			case IPPROTO_IPV6:	/* RFC 2893 */
1177 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1178 				break;
1179 
1180 			case IPPROTO_IPV4:	/* RFC 2893 */
1181 				PULLUP_TO(hlen, ulp, struct ip);
1182 				break;
1183 
1184 			default:
1185 				if (V_fw_verbose)
1186 					printf("IPFW2: IPV6 - Unknown "
1187 					    "Extension Header(%d), ext_hd=%x\n",
1188 					     proto, ext_hd);
1189 				if (V_fw_deny_unknown_exthdrs)
1190 				    return (IP_FW_DENY);
1191 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1192 				break;
1193 			} /*switch */
1194 		}
1195 		ip = mtod(m, struct ip *);
1196 		ip6 = (struct ip6_hdr *)ip;
1197 		args->f_id.src_ip6 = ip6->ip6_src;
1198 		args->f_id.dst_ip6 = ip6->ip6_dst;
1199 		args->f_id.src_ip = 0;
1200 		args->f_id.dst_ip = 0;
1201 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1202 	} else if (pktlen >= sizeof(struct ip) &&
1203 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1204 	    	is_ipv4 = 1;
1205 		hlen = ip->ip_hl << 2;
1206 		args->f_id.addr_type = 4;
1207 
1208 		/*
1209 		 * Collect parameters into local variables for faster matching.
1210 		 */
1211 		proto = ip->ip_p;
1212 		src_ip = ip->ip_src;
1213 		dst_ip = ip->ip_dst;
1214 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1215 		iplen = ntohs(ip->ip_len);
1216 		pktlen = iplen < pktlen ? iplen : pktlen;
1217 
1218 		if (offset == 0) {
1219 			switch (proto) {
1220 			case IPPROTO_TCP:
1221 				PULLUP_TO(hlen, ulp, struct tcphdr);
1222 				dst_port = TCP(ulp)->th_dport;
1223 				src_port = TCP(ulp)->th_sport;
1224 				/* save flags for dynamic rules */
1225 				args->f_id._flags = TCP(ulp)->th_flags;
1226 				break;
1227 
1228 			case IPPROTO_SCTP:
1229 				PULLUP_TO(hlen, ulp, struct sctphdr);
1230 				src_port = SCTP(ulp)->src_port;
1231 				dst_port = SCTP(ulp)->dest_port;
1232 				break;
1233 
1234 			case IPPROTO_UDP:
1235 				PULLUP_TO(hlen, ulp, struct udphdr);
1236 				dst_port = UDP(ulp)->uh_dport;
1237 				src_port = UDP(ulp)->uh_sport;
1238 				break;
1239 
1240 			case IPPROTO_ICMP:
1241 				PULLUP_TO(hlen, ulp, struct icmphdr);
1242 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1243 				break;
1244 
1245 			default:
1246 				break;
1247 			}
1248 		}
1249 
1250 		ip = mtod(m, struct ip *);
1251 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1252 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1253 	}
1254 #undef PULLUP_TO
1255 	if (proto) { /* we may have port numbers, store them */
1256 		args->f_id.proto = proto;
1257 		args->f_id.src_port = src_port = ntohs(src_port);
1258 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1259 	}
1260 
1261 	IPFW_PF_RLOCK(chain);
1262 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1263 		IPFW_PF_RUNLOCK(chain);
1264 		return (IP_FW_PASS);	/* accept */
1265 	}
1266 	if (args->rule.slot) {
1267 		/*
1268 		 * Packet has already been tagged as a result of a previous
1269 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1270 		 * REASS, NETGRAPH, DIVERT/TEE...)
1271 		 * Validate the slot and continue from the next one
1272 		 * if still present, otherwise do a lookup.
1273 		 */
1274 		f_pos = (args->rule.chain_id == chain->id) ?
1275 		    args->rule.slot :
1276 		    ipfw_find_rule(chain, args->rule.rulenum,
1277 			args->rule.rule_id);
1278 	} else {
1279 		f_pos = 0;
1280 	}
1281 
1282 	/*
1283 	 * Now scan the rules, and parse microinstructions for each rule.
1284 	 * We have two nested loops and an inner switch. Sometimes we
1285 	 * need to break out of one or both loops, or re-enter one of
1286 	 * the loops with updated variables. Loop variables are:
1287 	 *
1288 	 *	f_pos (outer loop) points to the current rule.
1289 	 *		On output it points to the matching rule.
1290 	 *	done (outer loop) is used as a flag to break the loop.
1291 	 *	l (inner loop)	residual length of current rule.
1292 	 *		cmd points to the current microinstruction.
1293 	 *
1294 	 * We break the inner loop by setting l=0 and possibly
1295 	 * cmdlen=0 if we don't want to advance cmd.
1296 	 * We break the outer loop by setting done=1
1297 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1298 	 * as needed.
1299 	 */
1300 	for (; f_pos < chain->n_rules; f_pos++) {
1301 		ipfw_insn *cmd;
1302 		uint32_t tablearg = 0;
1303 		int l, cmdlen, skip_or; /* skip rest of OR block */
1304 		struct ip_fw *f;
1305 
1306 		f = chain->map[f_pos];
1307 		if (V_set_disable & (1 << f->set) )
1308 			continue;
1309 
1310 		skip_or = 0;
1311 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1312 		    l -= cmdlen, cmd += cmdlen) {
1313 			int match;
1314 
1315 			/*
1316 			 * check_body is a jump target used when we find a
1317 			 * CHECK_STATE, and need to jump to the body of
1318 			 * the target rule.
1319 			 */
1320 
1321 /* check_body: */
1322 			cmdlen = F_LEN(cmd);
1323 			/*
1324 			 * An OR block (insn_1 || .. || insn_n) has the
1325 			 * F_OR bit set in all but the last instruction.
1326 			 * The first match will set "skip_or", and cause
1327 			 * the following instructions to be skipped until
1328 			 * past the one with the F_OR bit clear.
1329 			 */
1330 			if (skip_or) {		/* skip this instruction */
1331 				if ((cmd->len & F_OR) == 0)
1332 					skip_or = 0;	/* next one is good */
1333 				continue;
1334 			}
1335 			match = 0; /* set to 1 if we succeed */
1336 
1337 			switch (cmd->opcode) {
1338 			/*
1339 			 * The first set of opcodes compares the packet's
1340 			 * fields with some pattern, setting 'match' if a
1341 			 * match is found. At the end of the loop there is
1342 			 * logic to deal with F_NOT and F_OR flags associated
1343 			 * with the opcode.
1344 			 */
1345 			case O_NOP:
1346 				match = 1;
1347 				break;
1348 
1349 			case O_FORWARD_MAC:
1350 				printf("ipfw: opcode %d unimplemented\n",
1351 				    cmd->opcode);
1352 				break;
1353 
1354 			case O_GID:
1355 			case O_UID:
1356 			case O_JAIL:
1357 				/*
1358 				 * We only check offset == 0 && proto != 0,
1359 				 * as this ensures that we have a
1360 				 * packet with the ports info.
1361 				 */
1362 				if (offset != 0)
1363 					break;
1364 				if (proto == IPPROTO_TCP ||
1365 				    proto == IPPROTO_UDP)
1366 					match = check_uidgid(
1367 						    (ipfw_insn_u32 *)cmd,
1368 						    args, &ucred_lookup,
1369 #ifdef __FreeBSD__
1370 						    &ucred_cache);
1371 #else
1372 						    (void *)&ucred_cache);
1373 #endif
1374 				break;
1375 
1376 			case O_RECV:
1377 				match = iface_match(m->m_pkthdr.rcvif,
1378 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1379 				break;
1380 
1381 			case O_XMIT:
1382 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1383 				    chain, &tablearg);
1384 				break;
1385 
1386 			case O_VIA:
1387 				match = iface_match(oif ? oif :
1388 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1389 				    chain, &tablearg);
1390 				break;
1391 
1392 			case O_MACADDR2:
1393 				if (args->eh != NULL) {	/* have MAC header */
1394 					u_int32_t *want = (u_int32_t *)
1395 						((ipfw_insn_mac *)cmd)->addr;
1396 					u_int32_t *mask = (u_int32_t *)
1397 						((ipfw_insn_mac *)cmd)->mask;
1398 					u_int32_t *hdr = (u_int32_t *)args->eh;
1399 
1400 					match =
1401 					    ( want[0] == (hdr[0] & mask[0]) &&
1402 					      want[1] == (hdr[1] & mask[1]) &&
1403 					      want[2] == (hdr[2] & mask[2]) );
1404 				}
1405 				break;
1406 
1407 			case O_MAC_TYPE:
1408 				if (args->eh != NULL) {
1409 					u_int16_t *p =
1410 					    ((ipfw_insn_u16 *)cmd)->ports;
1411 					int i;
1412 
1413 					for (i = cmdlen - 1; !match && i>0;
1414 					    i--, p += 2)
1415 						match = (etype >= p[0] &&
1416 						    etype <= p[1]);
1417 				}
1418 				break;
1419 
1420 			case O_FRAG:
1421 				match = (offset != 0);
1422 				break;
1423 
1424 			case O_IN:	/* "out" is "not in" */
1425 				match = (oif == NULL);
1426 				break;
1427 
1428 			case O_LAYER2:
1429 				match = (args->eh != NULL);
1430 				break;
1431 
1432 			case O_DIVERTED:
1433 			    {
1434 				/* For diverted packets, args->rule.info
1435 				 * contains the divert port (in host format)
1436 				 * reason and direction.
1437 				 */
1438 				uint32_t i = args->rule.info;
1439 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1440 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1441 			    }
1442 				break;
1443 
1444 			case O_PROTO:
1445 				/*
1446 				 * We do not allow an arg of 0 so the
1447 				 * check of "proto" only suffices.
1448 				 */
1449 				match = (proto == cmd->arg1);
1450 				break;
1451 
1452 			case O_IP_SRC:
1453 				match = is_ipv4 &&
1454 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1455 				    src_ip.s_addr);
1456 				break;
1457 
1458 			case O_IP_SRC_LOOKUP:
1459 			case O_IP_DST_LOOKUP:
1460 				if (is_ipv4) {
1461 				    uint32_t key =
1462 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1463 					    dst_ip.s_addr : src_ip.s_addr;
1464 				    uint32_t v = 0;
1465 
1466 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1467 					/* generic lookup. The key must be
1468 					 * in 32bit big-endian format.
1469 					 */
1470 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1471 					if (v == 0)
1472 					    key = dst_ip.s_addr;
1473 					else if (v == 1)
1474 					    key = src_ip.s_addr;
1475 					else if (v == 6) /* dscp */
1476 					    key = (ip->ip_tos >> 2) & 0x3f;
1477 					else if (offset != 0)
1478 					    break;
1479 					else if (proto != IPPROTO_TCP &&
1480 						proto != IPPROTO_UDP)
1481 					    break;
1482 					else if (v == 2)
1483 					    key = dst_port;
1484 					else if (v == 3)
1485 					    key = src_port;
1486 #ifndef USERSPACE
1487 					else if (v == 4 || v == 5) {
1488 					    check_uidgid(
1489 						(ipfw_insn_u32 *)cmd,
1490 						args, &ucred_lookup,
1491 #ifdef __FreeBSD__
1492 						&ucred_cache);
1493 					    if (v == 4 /* O_UID */)
1494 						key = ucred_cache->cr_uid;
1495 					    else if (v == 5 /* O_JAIL */)
1496 						key = ucred_cache->cr_prison->pr_id;
1497 #else /* !__FreeBSD__ */
1498 						(void *)&ucred_cache);
1499 					    if (v ==4 /* O_UID */)
1500 						key = ucred_cache.uid;
1501 					    else if (v == 5 /* O_JAIL */)
1502 						key = ucred_cache.xid;
1503 #endif /* !__FreeBSD__ */
1504 					}
1505 #endif /* !USERSPACE */
1506 					else
1507 					    break;
1508 				    }
1509 				    match = ipfw_lookup_table(chain,
1510 					cmd->arg1, key, &v);
1511 				    if (!match)
1512 					break;
1513 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1514 					match =
1515 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1516 				    else
1517 					tablearg = v;
1518 				} else if (is_ipv6) {
1519 					uint32_t v = 0;
1520 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1521 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1522 					match = ipfw_lookup_table_extended(chain,
1523 							cmd->arg1,
1524 							sizeof(struct in6_addr),
1525 							pkey, &v);
1526 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1527 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1528 					if (match)
1529 						tablearg = v;
1530 				}
1531 				break;
1532 
1533 			case O_IP_FLOW_LOOKUP:
1534 				{
1535 					uint32_t v = 0;
1536 					match = ipfw_lookup_table_extended(chain,
1537 					    cmd->arg1, 0, &args->f_id, &v);
1538 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1539 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1540 					if (match)
1541 						tablearg = v;
1542 				}
1543 				break;
1544 			case O_IP_SRC_MASK:
1545 			case O_IP_DST_MASK:
1546 				if (is_ipv4) {
1547 				    uint32_t a =
1548 					(cmd->opcode == O_IP_DST_MASK) ?
1549 					    dst_ip.s_addr : src_ip.s_addr;
1550 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1551 				    int i = cmdlen-1;
1552 
1553 				    for (; !match && i>0; i-= 2, p+= 2)
1554 					match = (p[0] == (a & p[1]));
1555 				}
1556 				break;
1557 
1558 			case O_IP_SRC_ME:
1559 				if (is_ipv4) {
1560 					struct ifnet *tif;
1561 
1562 					INADDR_TO_IFP(src_ip, tif);
1563 					match = (tif != NULL);
1564 					break;
1565 				}
1566 #ifdef INET6
1567 				/* FALLTHROUGH */
1568 			case O_IP6_SRC_ME:
1569 				match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
1570 #endif
1571 				break;
1572 
1573 			case O_IP_DST_SET:
1574 			case O_IP_SRC_SET:
1575 				if (is_ipv4) {
1576 					u_int32_t *d = (u_int32_t *)(cmd+1);
1577 					u_int32_t addr =
1578 					    cmd->opcode == O_IP_DST_SET ?
1579 						args->f_id.dst_ip :
1580 						args->f_id.src_ip;
1581 
1582 					    if (addr < d[0])
1583 						    break;
1584 					    addr -= d[0]; /* subtract base */
1585 					    match = (addr < cmd->arg1) &&
1586 						( d[ 1 + (addr>>5)] &
1587 						  (1<<(addr & 0x1f)) );
1588 				}
1589 				break;
1590 
1591 			case O_IP_DST:
1592 				match = is_ipv4 &&
1593 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1594 				    dst_ip.s_addr);
1595 				break;
1596 
1597 			case O_IP_DST_ME:
1598 				if (is_ipv4) {
1599 					struct ifnet *tif;
1600 
1601 					INADDR_TO_IFP(dst_ip, tif);
1602 					match = (tif != NULL);
1603 					break;
1604 				}
1605 #ifdef INET6
1606 				/* FALLTHROUGH */
1607 			case O_IP6_DST_ME:
1608 				match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
1609 #endif
1610 				break;
1611 
1612 
1613 			case O_IP_SRCPORT:
1614 			case O_IP_DSTPORT:
1615 				/*
1616 				 * offset == 0 && proto != 0 is enough
1617 				 * to guarantee that we have a
1618 				 * packet with port info.
1619 				 */
1620 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1621 				    && offset == 0) {
1622 					u_int16_t x =
1623 					    (cmd->opcode == O_IP_SRCPORT) ?
1624 						src_port : dst_port ;
1625 					u_int16_t *p =
1626 					    ((ipfw_insn_u16 *)cmd)->ports;
1627 					int i;
1628 
1629 					for (i = cmdlen - 1; !match && i>0;
1630 					    i--, p += 2)
1631 						match = (x>=p[0] && x<=p[1]);
1632 				}
1633 				break;
1634 
1635 			case O_ICMPTYPE:
1636 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1637 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1638 				break;
1639 
1640 #ifdef INET6
1641 			case O_ICMP6TYPE:
1642 				match = is_ipv6 && offset == 0 &&
1643 				    proto==IPPROTO_ICMPV6 &&
1644 				    icmp6type_match(
1645 					ICMP6(ulp)->icmp6_type,
1646 					(ipfw_insn_u32 *)cmd);
1647 				break;
1648 #endif /* INET6 */
1649 
1650 			case O_IPOPT:
1651 				match = (is_ipv4 &&
1652 				    ipopts_match(ip, cmd) );
1653 				break;
1654 
1655 			case O_IPVER:
1656 				match = (is_ipv4 &&
1657 				    cmd->arg1 == ip->ip_v);
1658 				break;
1659 
1660 			case O_IPID:
1661 			case O_IPLEN:
1662 			case O_IPTTL:
1663 				if (is_ipv4) {	/* only for IP packets */
1664 				    uint16_t x;
1665 				    uint16_t *p;
1666 				    int i;
1667 
1668 				    if (cmd->opcode == O_IPLEN)
1669 					x = iplen;
1670 				    else if (cmd->opcode == O_IPTTL)
1671 					x = ip->ip_ttl;
1672 				    else /* must be IPID */
1673 					x = ntohs(ip->ip_id);
1674 				    if (cmdlen == 1) {
1675 					match = (cmd->arg1 == x);
1676 					break;
1677 				    }
1678 				    /* otherwise we have ranges */
1679 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1680 				    i = cmdlen - 1;
1681 				    for (; !match && i>0; i--, p += 2)
1682 					match = (x >= p[0] && x <= p[1]);
1683 				}
1684 				break;
1685 
1686 			case O_IPPRECEDENCE:
1687 				match = (is_ipv4 &&
1688 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1689 				break;
1690 
1691 			case O_IPTOS:
1692 				match = (is_ipv4 &&
1693 				    flags_match(cmd, ip->ip_tos));
1694 				break;
1695 
1696 			case O_DSCP:
1697 			    {
1698 				uint32_t *p;
1699 				uint16_t x;
1700 
1701 				p = ((ipfw_insn_u32 *)cmd)->d;
1702 
1703 				if (is_ipv4)
1704 					x = ip->ip_tos >> 2;
1705 				else if (is_ipv6) {
1706 					uint8_t *v;
1707 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
1708 					x = (*v & 0x0F) << 2;
1709 					v++;
1710 					x |= *v >> 6;
1711 				} else
1712 					break;
1713 
1714 				/* DSCP bitmask is stored as low_u32 high_u32 */
1715 				if (x >= 32)
1716 					match = *(p + 1) & (1 << (x - 32));
1717 				else
1718 					match = *p & (1 << x);
1719 			    }
1720 				break;
1721 
1722 			case O_TCPDATALEN:
1723 				if (proto == IPPROTO_TCP && offset == 0) {
1724 				    struct tcphdr *tcp;
1725 				    uint16_t x;
1726 				    uint16_t *p;
1727 				    int i;
1728 
1729 				    tcp = TCP(ulp);
1730 				    x = iplen -
1731 					((ip->ip_hl + tcp->th_off) << 2);
1732 				    if (cmdlen == 1) {
1733 					match = (cmd->arg1 == x);
1734 					break;
1735 				    }
1736 				    /* otherwise we have ranges */
1737 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1738 				    i = cmdlen - 1;
1739 				    for (; !match && i>0; i--, p += 2)
1740 					match = (x >= p[0] && x <= p[1]);
1741 				}
1742 				break;
1743 
1744 			case O_TCPFLAGS:
1745 				match = (proto == IPPROTO_TCP && offset == 0 &&
1746 				    flags_match(cmd, TCP(ulp)->th_flags));
1747 				break;
1748 
1749 			case O_TCPOPTS:
1750 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
1751 					PULLUP_LEN(hlen, ulp,
1752 					    (TCP(ulp)->th_off << 2));
1753 					match = tcpopts_match(TCP(ulp), cmd);
1754 				}
1755 				break;
1756 
1757 			case O_TCPSEQ:
1758 				match = (proto == IPPROTO_TCP && offset == 0 &&
1759 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1760 					TCP(ulp)->th_seq);
1761 				break;
1762 
1763 			case O_TCPACK:
1764 				match = (proto == IPPROTO_TCP && offset == 0 &&
1765 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1766 					TCP(ulp)->th_ack);
1767 				break;
1768 
1769 			case O_TCPWIN:
1770 				if (proto == IPPROTO_TCP && offset == 0) {
1771 				    uint16_t x;
1772 				    uint16_t *p;
1773 				    int i;
1774 
1775 				    x = ntohs(TCP(ulp)->th_win);
1776 				    if (cmdlen == 1) {
1777 					match = (cmd->arg1 == x);
1778 					break;
1779 				    }
1780 				    /* Otherwise we have ranges. */
1781 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1782 				    i = cmdlen - 1;
1783 				    for (; !match && i > 0; i--, p += 2)
1784 					match = (x >= p[0] && x <= p[1]);
1785 				}
1786 				break;
1787 
1788 			case O_ESTAB:
1789 				/* reject packets which have SYN only */
1790 				/* XXX should i also check for TH_ACK ? */
1791 				match = (proto == IPPROTO_TCP && offset == 0 &&
1792 				    (TCP(ulp)->th_flags &
1793 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1794 				break;
1795 
1796 			case O_ALTQ: {
1797 				struct pf_mtag *at;
1798 				struct m_tag *mtag;
1799 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1800 
1801 				/*
1802 				 * ALTQ uses mbuf tags from another
1803 				 * packet filtering system - pf(4).
1804 				 * We allocate a tag in its format
1805 				 * and fill it in, pretending to be pf(4).
1806 				 */
1807 				match = 1;
1808 				at = pf_find_mtag(m);
1809 				if (at != NULL && at->qid != 0)
1810 					break;
1811 				mtag = m_tag_get(PACKET_TAG_PF,
1812 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1813 				if (mtag == NULL) {
1814 					/*
1815 					 * Let the packet fall back to the
1816 					 * default ALTQ.
1817 					 */
1818 					break;
1819 				}
1820 				m_tag_prepend(m, mtag);
1821 				at = (struct pf_mtag *)(mtag + 1);
1822 				at->qid = altq->qid;
1823 				at->hdr = ip;
1824 				break;
1825 			}
1826 
1827 			case O_LOG:
1828 				ipfw_log(chain, f, hlen, args, m,
1829 				    oif, offset | ip6f_mf, tablearg, ip);
1830 				match = 1;
1831 				break;
1832 
1833 			case O_PROB:
1834 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1835 				break;
1836 
1837 			case O_VERREVPATH:
1838 				/* Outgoing packets automatically pass/match */
1839 				match = ((oif != NULL) ||
1840 				    (m->m_pkthdr.rcvif == NULL) ||
1841 				    (
1842 #ifdef INET6
1843 				    is_ipv6 ?
1844 					verify_path6(&(args->f_id.src_ip6),
1845 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1846 #endif
1847 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1848 				        args->f_id.fib)));
1849 				break;
1850 
1851 			case O_VERSRCREACH:
1852 				/* Outgoing packets automatically pass/match */
1853 				match = (hlen > 0 && ((oif != NULL) ||
1854 #ifdef INET6
1855 				    is_ipv6 ?
1856 				        verify_path6(&(args->f_id.src_ip6),
1857 				            NULL, args->f_id.fib) :
1858 #endif
1859 				    verify_path(src_ip, NULL, args->f_id.fib)));
1860 				break;
1861 
1862 			case O_ANTISPOOF:
1863 				/* Outgoing packets automatically pass/match */
1864 				if (oif == NULL && hlen > 0 &&
1865 				    (  (is_ipv4 && in_localaddr(src_ip))
1866 #ifdef INET6
1867 				    || (is_ipv6 &&
1868 				        in6_localaddr(&(args->f_id.src_ip6)))
1869 #endif
1870 				    ))
1871 					match =
1872 #ifdef INET6
1873 					    is_ipv6 ? verify_path6(
1874 					        &(args->f_id.src_ip6),
1875 					        m->m_pkthdr.rcvif,
1876 						args->f_id.fib) :
1877 #endif
1878 					    verify_path(src_ip,
1879 					    	m->m_pkthdr.rcvif,
1880 					        args->f_id.fib);
1881 				else
1882 					match = 1;
1883 				break;
1884 
1885 			case O_IPSEC:
1886 #ifdef IPSEC
1887 				match = (m_tag_find(m,
1888 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1889 #endif
1890 				/* otherwise no match */
1891 				break;
1892 
1893 #ifdef INET6
1894 			case O_IP6_SRC:
1895 				match = is_ipv6 &&
1896 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1897 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1898 				break;
1899 
1900 			case O_IP6_DST:
1901 				match = is_ipv6 &&
1902 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1903 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1904 				break;
1905 			case O_IP6_SRC_MASK:
1906 			case O_IP6_DST_MASK:
1907 				if (is_ipv6) {
1908 					int i = cmdlen - 1;
1909 					struct in6_addr p;
1910 					struct in6_addr *d =
1911 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1912 
1913 					for (; !match && i > 0; d += 2,
1914 					    i -= F_INSN_SIZE(struct in6_addr)
1915 					    * 2) {
1916 						p = (cmd->opcode ==
1917 						    O_IP6_SRC_MASK) ?
1918 						    args->f_id.src_ip6:
1919 						    args->f_id.dst_ip6;
1920 						APPLY_MASK(&p, &d[1]);
1921 						match =
1922 						    IN6_ARE_ADDR_EQUAL(&d[0],
1923 						    &p);
1924 					}
1925 				}
1926 				break;
1927 
1928 			case O_FLOW6ID:
1929 				match = is_ipv6 &&
1930 				    flow6id_match(args->f_id.flow_id6,
1931 				    (ipfw_insn_u32 *) cmd);
1932 				break;
1933 
1934 			case O_EXT_HDR:
1935 				match = is_ipv6 &&
1936 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1937 				break;
1938 
1939 			case O_IP6:
1940 				match = is_ipv6;
1941 				break;
1942 #endif
1943 
1944 			case O_IP4:
1945 				match = is_ipv4;
1946 				break;
1947 
1948 			case O_TAG: {
1949 				struct m_tag *mtag;
1950 				uint32_t tag = TARG(cmd->arg1, tag);
1951 
1952 				/* Packet is already tagged with this tag? */
1953 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1954 
1955 				/* We have `untag' action when F_NOT flag is
1956 				 * present. And we must remove this mtag from
1957 				 * mbuf and reset `match' to zero (`match' will
1958 				 * be inversed later).
1959 				 * Otherwise we should allocate new mtag and
1960 				 * push it into mbuf.
1961 				 */
1962 				if (cmd->len & F_NOT) { /* `untag' action */
1963 					if (mtag != NULL)
1964 						m_tag_delete(m, mtag);
1965 					match = 0;
1966 				} else {
1967 					if (mtag == NULL) {
1968 						mtag = m_tag_alloc( MTAG_IPFW,
1969 						    tag, 0, M_NOWAIT);
1970 						if (mtag != NULL)
1971 							m_tag_prepend(m, mtag);
1972 					}
1973 					match = 1;
1974 				}
1975 				break;
1976 			}
1977 
1978 			case O_FIB: /* try match the specified fib */
1979 				if (args->f_id.fib == cmd->arg1)
1980 					match = 1;
1981 				break;
1982 
1983 			case O_SOCKARG:	{
1984 #ifndef USERSPACE	/* not supported in userspace */
1985 				struct inpcb *inp = args->inp;
1986 				struct inpcbinfo *pi;
1987 
1988 				if (is_ipv6) /* XXX can we remove this ? */
1989 					break;
1990 
1991 				if (proto == IPPROTO_TCP)
1992 					pi = &V_tcbinfo;
1993 				else if (proto == IPPROTO_UDP)
1994 					pi = &V_udbinfo;
1995 				else
1996 					break;
1997 
1998 				/*
1999 				 * XXXRW: so_user_cookie should almost
2000 				 * certainly be inp_user_cookie?
2001 				 */
2002 
2003 				/* For incoming packet, lookup up the
2004 				inpcb using the src/dest ip/port tuple */
2005 				if (inp == NULL) {
2006 					inp = in_pcblookup(pi,
2007 						src_ip, htons(src_port),
2008 						dst_ip, htons(dst_port),
2009 						INPLOOKUP_RLOCKPCB, NULL);
2010 					if (inp != NULL) {
2011 						tablearg =
2012 						    inp->inp_socket->so_user_cookie;
2013 						if (tablearg)
2014 							match = 1;
2015 						INP_RUNLOCK(inp);
2016 					}
2017 				} else {
2018 					if (inp->inp_socket) {
2019 						tablearg =
2020 						    inp->inp_socket->so_user_cookie;
2021 						if (tablearg)
2022 							match = 1;
2023 					}
2024 				}
2025 #endif /* !USERSPACE */
2026 				break;
2027 			}
2028 
2029 			case O_TAGGED: {
2030 				struct m_tag *mtag;
2031 				uint32_t tag = TARG(cmd->arg1, tag);
2032 
2033 				if (cmdlen == 1) {
2034 					match = m_tag_locate(m, MTAG_IPFW,
2035 					    tag, NULL) != NULL;
2036 					break;
2037 				}
2038 
2039 				/* we have ranges */
2040 				for (mtag = m_tag_first(m);
2041 				    mtag != NULL && !match;
2042 				    mtag = m_tag_next(m, mtag)) {
2043 					uint16_t *p;
2044 					int i;
2045 
2046 					if (mtag->m_tag_cookie != MTAG_IPFW)
2047 						continue;
2048 
2049 					p = ((ipfw_insn_u16 *)cmd)->ports;
2050 					i = cmdlen - 1;
2051 					for(; !match && i > 0; i--, p += 2)
2052 						match =
2053 						    mtag->m_tag_id >= p[0] &&
2054 						    mtag->m_tag_id <= p[1];
2055 				}
2056 				break;
2057 			}
2058 
2059 			/*
2060 			 * The second set of opcodes represents 'actions',
2061 			 * i.e. the terminal part of a rule once the packet
2062 			 * matches all previous patterns.
2063 			 * Typically there is only one action for each rule,
2064 			 * and the opcode is stored at the end of the rule
2065 			 * (but there are exceptions -- see below).
2066 			 *
2067 			 * In general, here we set retval and terminate the
2068 			 * outer loop (would be a 'break 3' in some language,
2069 			 * but we need to set l=0, done=1)
2070 			 *
2071 			 * Exceptions:
2072 			 * O_COUNT and O_SKIPTO actions:
2073 			 *   instead of terminating, we jump to the next rule
2074 			 *   (setting l=0), or to the SKIPTO target (setting
2075 			 *   f/f_len, cmd and l as needed), respectively.
2076 			 *
2077 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2078 			 *   perform some action and set match = 1;
2079 			 *
2080 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2081 			 *   not real 'actions', and are stored right
2082 			 *   before the 'action' part of the rule.
2083 			 *   These opcodes try to install an entry in the
2084 			 *   state tables; if successful, we continue with
2085 			 *   the next opcode (match=1; break;), otherwise
2086 			 *   the packet must be dropped (set retval,
2087 			 *   break loops with l=0, done=1)
2088 			 *
2089 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2090 			 *   cause a lookup of the state table, and a jump
2091 			 *   to the 'action' part of the parent rule
2092 			 *   if an entry is found, or
2093 			 *   (CHECK_STATE only) a jump to the next rule if
2094 			 *   the entry is not found.
2095 			 *   The result of the lookup is cached so that
2096 			 *   further instances of these opcodes become NOPs.
2097 			 *   The jump to the next rule is done by setting
2098 			 *   l=0, cmdlen=0.
2099 			 */
2100 			case O_LIMIT:
2101 			case O_KEEP_STATE:
2102 				if (ipfw_install_state(chain, f,
2103 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2104 					/* error or limit violation */
2105 					retval = IP_FW_DENY;
2106 					l = 0;	/* exit inner loop */
2107 					done = 1; /* exit outer loop */
2108 				}
2109 				match = 1;
2110 				break;
2111 
2112 			case O_PROBE_STATE:
2113 			case O_CHECK_STATE:
2114 				/*
2115 				 * dynamic rules are checked at the first
2116 				 * keep-state or check-state occurrence,
2117 				 * with the result being stored in dyn_dir
2118 				 * and dyn_name.
2119 				 * The compiler introduces a PROBE_STATE
2120 				 * instruction for us when we have a
2121 				 * KEEP_STATE (because PROBE_STATE needs
2122 				 * to be run first).
2123 				 *
2124 				 * (dyn_dir == MATCH_UNKNOWN) means this is
2125 				 * first lookup for such f_id. Do lookup.
2126 				 *
2127 				 * (dyn_dir != MATCH_UNKNOWN &&
2128 				 *  dyn_name != 0 && dyn_name != cmd->arg1)
2129 				 * means previous lookup didn't find dynamic
2130 				 * rule for specific state name and current
2131 				 * lookup will search rule with another state
2132 				 * name. Redo lookup.
2133 				 *
2134 				 * (dyn_dir != MATCH_UNKNOWN && dyn_name == 0)
2135 				 * means previous lookup was for `any' name
2136 				 * and it didn't find rule. No need to do
2137 				 * lookup again.
2138 				 */
2139 				if ((dyn_dir == MATCH_UNKNOWN ||
2140 				    (dyn_name != 0 &&
2141 				    dyn_name != cmd->arg1)) &&
2142 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2143 				     &dyn_dir, proto == IPPROTO_TCP ?
2144 				     TCP(ulp): NULL,
2145 				     (dyn_name = cmd->arg1))) != NULL) {
2146 					/*
2147 					 * Found dynamic entry, update stats
2148 					 * and jump to the 'action' part of
2149 					 * the parent rule by setting
2150 					 * f, cmd, l and clearing cmdlen.
2151 					 */
2152 					IPFW_INC_DYN_COUNTER(q, pktlen);
2153 					/* XXX we would like to have f_pos
2154 					 * readily accessible in the dynamic
2155 				         * rule, instead of having to
2156 					 * lookup q->rule.
2157 					 */
2158 					f = q->rule;
2159 					f_pos = ipfw_find_rule(chain,
2160 						f->rulenum, f->id);
2161 					cmd = ACTION_PTR(f);
2162 					l = f->cmd_len - f->act_ofs;
2163 					ipfw_dyn_unlock(q);
2164 					cmdlen = 0;
2165 					match = 1;
2166 					break;
2167 				}
2168 				/*
2169 				 * Dynamic entry not found. If CHECK_STATE,
2170 				 * skip to next rule, if PROBE_STATE just
2171 				 * ignore and continue with next opcode.
2172 				 */
2173 				if (cmd->opcode == O_CHECK_STATE)
2174 					l = 0;	/* exit inner loop */
2175 				match = 1;
2176 				break;
2177 
2178 			case O_ACCEPT:
2179 				retval = 0;	/* accept */
2180 				l = 0;		/* exit inner loop */
2181 				done = 1;	/* exit outer loop */
2182 				break;
2183 
2184 			case O_PIPE:
2185 			case O_QUEUE:
2186 				set_match(args, f_pos, chain);
2187 				args->rule.info = TARG(cmd->arg1, pipe);
2188 				if (cmd->opcode == O_PIPE)
2189 					args->rule.info |= IPFW_IS_PIPE;
2190 				if (V_fw_one_pass)
2191 					args->rule.info |= IPFW_ONEPASS;
2192 				retval = IP_FW_DUMMYNET;
2193 				l = 0;          /* exit inner loop */
2194 				done = 1;       /* exit outer loop */
2195 				break;
2196 
2197 			case O_DIVERT:
2198 			case O_TEE:
2199 				if (args->eh) /* not on layer 2 */
2200 				    break;
2201 				/* otherwise this is terminal */
2202 				l = 0;		/* exit inner loop */
2203 				done = 1;	/* exit outer loop */
2204 				retval = (cmd->opcode == O_DIVERT) ?
2205 					IP_FW_DIVERT : IP_FW_TEE;
2206 				set_match(args, f_pos, chain);
2207 				args->rule.info = TARG(cmd->arg1, divert);
2208 				break;
2209 
2210 			case O_COUNT:
2211 				IPFW_INC_RULE_COUNTER(f, pktlen);
2212 				l = 0;		/* exit inner loop */
2213 				break;
2214 
2215 			case O_SKIPTO:
2216 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2217 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2218 			    /*
2219 			     * Skip disabled rules, and re-enter
2220 			     * the inner loop with the correct
2221 			     * f_pos, f, l and cmd.
2222 			     * Also clear cmdlen and skip_or
2223 			     */
2224 			    for (; f_pos < chain->n_rules - 1 &&
2225 				    (V_set_disable &
2226 				     (1 << chain->map[f_pos]->set));
2227 				    f_pos++)
2228 				;
2229 			    /* Re-enter the inner loop at the skipto rule. */
2230 			    f = chain->map[f_pos];
2231 			    l = f->cmd_len;
2232 			    cmd = f->cmd;
2233 			    match = 1;
2234 			    cmdlen = 0;
2235 			    skip_or = 0;
2236 			    continue;
2237 			    break;	/* not reached */
2238 
2239 			case O_CALLRETURN: {
2240 				/*
2241 				 * Implementation of `subroutine' call/return,
2242 				 * in the stack carried in an mbuf tag. This
2243 				 * is different from `skipto' in that any call
2244 				 * address is possible (`skipto' must prevent
2245 				 * backward jumps to avoid endless loops).
2246 				 * We have `return' action when F_NOT flag is
2247 				 * present. The `m_tag_id' field is used as
2248 				 * stack pointer.
2249 				 */
2250 				struct m_tag *mtag;
2251 				uint16_t jmpto, *stack;
2252 
2253 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2254 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2255 				/*
2256 				 * Hand-rolled version of m_tag_locate() with
2257 				 * wildcard `type'.
2258 				 * If not already tagged, allocate new tag.
2259 				 */
2260 				mtag = m_tag_first(m);
2261 				while (mtag != NULL) {
2262 					if (mtag->m_tag_cookie ==
2263 					    MTAG_IPFW_CALL)
2264 						break;
2265 					mtag = m_tag_next(m, mtag);
2266 				}
2267 				if (mtag == NULL && IS_CALL) {
2268 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2269 					    IPFW_CALLSTACK_SIZE *
2270 					    sizeof(uint16_t), M_NOWAIT);
2271 					if (mtag != NULL)
2272 						m_tag_prepend(m, mtag);
2273 				}
2274 
2275 				/*
2276 				 * On error both `call' and `return' just
2277 				 * continue with next rule.
2278 				 */
2279 				if (IS_RETURN && (mtag == NULL ||
2280 				    mtag->m_tag_id == 0)) {
2281 					l = 0;		/* exit inner loop */
2282 					break;
2283 				}
2284 				if (IS_CALL && (mtag == NULL ||
2285 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2286 					printf("ipfw: call stack error, "
2287 					    "go to next rule\n");
2288 					l = 0;		/* exit inner loop */
2289 					break;
2290 				}
2291 
2292 				IPFW_INC_RULE_COUNTER(f, pktlen);
2293 				stack = (uint16_t *)(mtag + 1);
2294 
2295 				/*
2296 				 * The `call' action may use cached f_pos
2297 				 * (in f->next_rule), whose version is written
2298 				 * in f->next_rule.
2299 				 * The `return' action, however, doesn't have
2300 				 * fixed jump address in cmd->arg1 and can't use
2301 				 * cache.
2302 				 */
2303 				if (IS_CALL) {
2304 					stack[mtag->m_tag_id] = f->rulenum;
2305 					mtag->m_tag_id++;
2306 			    		f_pos = JUMP(chain, f, cmd->arg1,
2307 					    tablearg, 1);
2308 				} else {	/* `return' action */
2309 					mtag->m_tag_id--;
2310 					jmpto = stack[mtag->m_tag_id] + 1;
2311 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2312 				}
2313 
2314 				/*
2315 				 * Skip disabled rules, and re-enter
2316 				 * the inner loop with the correct
2317 				 * f_pos, f, l and cmd.
2318 				 * Also clear cmdlen and skip_or
2319 				 */
2320 				for (; f_pos < chain->n_rules - 1 &&
2321 				    (V_set_disable &
2322 				    (1 << chain->map[f_pos]->set)); f_pos++)
2323 					;
2324 				/* Re-enter the inner loop at the dest rule. */
2325 				f = chain->map[f_pos];
2326 				l = f->cmd_len;
2327 				cmd = f->cmd;
2328 				cmdlen = 0;
2329 				skip_or = 0;
2330 				continue;
2331 				break;	/* NOTREACHED */
2332 			}
2333 #undef IS_CALL
2334 #undef IS_RETURN
2335 
2336 			case O_REJECT:
2337 				/*
2338 				 * Drop the packet and send a reject notice
2339 				 * if the packet is not ICMP (or is an ICMP
2340 				 * query), and it is not multicast/broadcast.
2341 				 */
2342 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2343 				    (proto != IPPROTO_ICMP ||
2344 				     is_icmp_query(ICMP(ulp))) &&
2345 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2346 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2347 					send_reject(args, cmd->arg1, iplen, ip);
2348 					m = args->m;
2349 				}
2350 				/* FALLTHROUGH */
2351 #ifdef INET6
2352 			case O_UNREACH6:
2353 				if (hlen > 0 && is_ipv6 &&
2354 				    ((offset & IP6F_OFF_MASK) == 0) &&
2355 				    (proto != IPPROTO_ICMPV6 ||
2356 				     (is_icmp6_query(icmp6_type) == 1)) &&
2357 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2358 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2359 					send_reject6(
2360 					    args, cmd->arg1, hlen,
2361 					    (struct ip6_hdr *)ip);
2362 					m = args->m;
2363 				}
2364 				/* FALLTHROUGH */
2365 #endif
2366 			case O_DENY:
2367 				retval = IP_FW_DENY;
2368 				l = 0;		/* exit inner loop */
2369 				done = 1;	/* exit outer loop */
2370 				break;
2371 
2372 			case O_FORWARD_IP:
2373 				if (args->eh)	/* not valid on layer2 pkts */
2374 					break;
2375 				if (q == NULL || q->rule != f ||
2376 				    dyn_dir == MATCH_FORWARD) {
2377 				    struct sockaddr_in *sa;
2378 
2379 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2380 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2381 #ifdef INET6
2382 					/*
2383 					 * We use O_FORWARD_IP opcode for
2384 					 * fwd rule with tablearg, but tables
2385 					 * now support IPv6 addresses. And
2386 					 * when we are inspecting IPv6 packet,
2387 					 * we can use nh6 field from
2388 					 * table_value as next_hop6 address.
2389 					 */
2390 					if (is_ipv6) {
2391 						struct sockaddr_in6 *sa6;
2392 
2393 						sa6 = args->next_hop6 =
2394 						    &args->hopstore6;
2395 						sa6->sin6_family = AF_INET6;
2396 						sa6->sin6_len = sizeof(*sa6);
2397 						sa6->sin6_addr = TARG_VAL(
2398 						    chain, tablearg, nh6);
2399 						/*
2400 						 * Set sin6_scope_id only for
2401 						 * link-local unicast addresses.
2402 						 */
2403 						if (IN6_IS_ADDR_LINKLOCAL(
2404 						    &sa6->sin6_addr))
2405 							sa6->sin6_scope_id =
2406 							    TARG_VAL(chain,
2407 								tablearg,
2408 								zoneid);
2409 					} else
2410 #endif
2411 					{
2412 						sa = args->next_hop =
2413 						    &args->hopstore;
2414 						sa->sin_family = AF_INET;
2415 						sa->sin_len = sizeof(*sa);
2416 						sa->sin_addr.s_addr = htonl(
2417 						    TARG_VAL(chain, tablearg,
2418 						    nh4));
2419 					}
2420 				    } else {
2421 					args->next_hop = sa;
2422 				    }
2423 				}
2424 				retval = IP_FW_PASS;
2425 				l = 0;          /* exit inner loop */
2426 				done = 1;       /* exit outer loop */
2427 				break;
2428 
2429 #ifdef INET6
2430 			case O_FORWARD_IP6:
2431 				if (args->eh)	/* not valid on layer2 pkts */
2432 					break;
2433 				if (q == NULL || q->rule != f ||
2434 				    dyn_dir == MATCH_FORWARD) {
2435 					struct sockaddr_in6 *sin6;
2436 
2437 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2438 					args->next_hop6 = sin6;
2439 				}
2440 				retval = IP_FW_PASS;
2441 				l = 0;		/* exit inner loop */
2442 				done = 1;	/* exit outer loop */
2443 				break;
2444 #endif
2445 
2446 			case O_NETGRAPH:
2447 			case O_NGTEE:
2448 				set_match(args, f_pos, chain);
2449 				args->rule.info = TARG(cmd->arg1, netgraph);
2450 				if (V_fw_one_pass)
2451 					args->rule.info |= IPFW_ONEPASS;
2452 				retval = (cmd->opcode == O_NETGRAPH) ?
2453 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2454 				l = 0;          /* exit inner loop */
2455 				done = 1;       /* exit outer loop */
2456 				break;
2457 
2458 			case O_SETFIB: {
2459 				uint32_t fib;
2460 
2461 				IPFW_INC_RULE_COUNTER(f, pktlen);
2462 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
2463 				if (fib >= rt_numfibs)
2464 					fib = 0;
2465 				M_SETFIB(m, fib);
2466 				args->f_id.fib = fib;
2467 				l = 0;		/* exit inner loop */
2468 				break;
2469 		        }
2470 
2471 			case O_SETDSCP: {
2472 				uint16_t code;
2473 
2474 				code = TARG(cmd->arg1, dscp) & 0x3F;
2475 				l = 0;		/* exit inner loop */
2476 				if (is_ipv4) {
2477 					uint16_t old;
2478 
2479 					old = *(uint16_t *)ip;
2480 					ip->ip_tos = (code << 2) |
2481 					    (ip->ip_tos & 0x03);
2482 					ip->ip_sum = cksum_adjust(ip->ip_sum,
2483 					    old, *(uint16_t *)ip);
2484 				} else if (is_ipv6) {
2485 					uint8_t *v;
2486 
2487 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2488 					*v = (*v & 0xF0) | (code >> 2);
2489 					v++;
2490 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2491 				} else
2492 					break;
2493 
2494 				IPFW_INC_RULE_COUNTER(f, pktlen);
2495 				break;
2496 			}
2497 
2498 			case O_NAT:
2499 				l = 0;          /* exit inner loop */
2500 				done = 1;       /* exit outer loop */
2501  				if (!IPFW_NAT_LOADED) {
2502 				    retval = IP_FW_DENY;
2503 				    break;
2504 				}
2505 
2506 				struct cfg_nat *t;
2507 				int nat_id;
2508 
2509 				set_match(args, f_pos, chain);
2510 				/* Check if this is 'global' nat rule */
2511 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
2512 					retval = ipfw_nat_ptr(args, NULL, m);
2513 					break;
2514 				}
2515 				t = ((ipfw_insn_nat *)cmd)->nat;
2516 				if (t == NULL) {
2517 					nat_id = TARG(cmd->arg1, nat);
2518 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2519 
2520 					if (t == NULL) {
2521 					    retval = IP_FW_DENY;
2522 					    break;
2523 					}
2524 					if (cmd->arg1 != IP_FW_TARG)
2525 					    ((ipfw_insn_nat *)cmd)->nat = t;
2526 				}
2527 				retval = ipfw_nat_ptr(args, t, m);
2528 				break;
2529 
2530 			case O_REASS: {
2531 				int ip_off;
2532 
2533 				IPFW_INC_RULE_COUNTER(f, pktlen);
2534 				l = 0;	/* in any case exit inner loop */
2535 				ip_off = ntohs(ip->ip_off);
2536 
2537 				/* if not fragmented, go to next rule */
2538 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2539 				    break;
2540 
2541 				args->m = m = ip_reass(m);
2542 
2543 				/*
2544 				 * do IP header checksum fixup.
2545 				 */
2546 				if (m == NULL) { /* fragment got swallowed */
2547 				    retval = IP_FW_DENY;
2548 				} else { /* good, packet complete */
2549 				    int hlen;
2550 
2551 				    ip = mtod(m, struct ip *);
2552 				    hlen = ip->ip_hl << 2;
2553 				    ip->ip_sum = 0;
2554 				    if (hlen == sizeof(struct ip))
2555 					ip->ip_sum = in_cksum_hdr(ip);
2556 				    else
2557 					ip->ip_sum = in_cksum(m, hlen);
2558 				    retval = IP_FW_REASS;
2559 				    set_match(args, f_pos, chain);
2560 				}
2561 				done = 1;	/* exit outer loop */
2562 				break;
2563 			}
2564 			case O_EXTERNAL_ACTION:
2565 				l = 0; /* in any case exit inner loop */
2566 				retval = ipfw_run_eaction(chain, args,
2567 				    cmd, &done);
2568 				break;
2569 
2570 			default:
2571 				panic("-- unknown opcode %d\n", cmd->opcode);
2572 			} /* end of switch() on opcodes */
2573 			/*
2574 			 * if we get here with l=0, then match is irrelevant.
2575 			 */
2576 
2577 			if (cmd->len & F_NOT)
2578 				match = !match;
2579 
2580 			if (match) {
2581 				if (cmd->len & F_OR)
2582 					skip_or = 1;
2583 			} else {
2584 				if (!(cmd->len & F_OR)) /* not an OR block, */
2585 					break;		/* try next rule    */
2586 			}
2587 
2588 		}	/* end of inner loop, scan opcodes */
2589 #undef PULLUP_LEN
2590 
2591 		if (done)
2592 			break;
2593 
2594 /* next_rule:; */	/* try next rule		*/
2595 
2596 	}		/* end of outer for, scan rules */
2597 
2598 	if (done) {
2599 		struct ip_fw *rule = chain->map[f_pos];
2600 		/* Update statistics */
2601 		IPFW_INC_RULE_COUNTER(rule, pktlen);
2602 	} else {
2603 		retval = IP_FW_DENY;
2604 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2605 	}
2606 	IPFW_PF_RUNLOCK(chain);
2607 #ifdef __FreeBSD__
2608 	if (ucred_cache != NULL)
2609 		crfree(ucred_cache);
2610 #endif
2611 	return (retval);
2612 
2613 pullup_failed:
2614 	if (V_fw_verbose)
2615 		printf("ipfw: pullup failed\n");
2616 	return (IP_FW_DENY);
2617 }
2618 
2619 /*
2620  * Set maximum number of tables that can be used in given VNET ipfw instance.
2621  */
2622 #ifdef SYSCTL_NODE
2623 static int
2624 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2625 {
2626 	int error;
2627 	unsigned int ntables;
2628 
2629 	ntables = V_fw_tables_max;
2630 
2631 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2632 	/* Read operation or some error */
2633 	if ((error != 0) || (req->newptr == NULL))
2634 		return (error);
2635 
2636 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2637 }
2638 
2639 /*
2640  * Switches table namespace between global and per-set.
2641  */
2642 static int
2643 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2644 {
2645 	int error;
2646 	unsigned int sets;
2647 
2648 	sets = V_fw_tables_sets;
2649 
2650 	error = sysctl_handle_int(oidp, &sets, 0, req);
2651 	/* Read operation or some error */
2652 	if ((error != 0) || (req->newptr == NULL))
2653 		return (error);
2654 
2655 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2656 }
2657 #endif
2658 
2659 /*
2660  * Module and VNET glue
2661  */
2662 
2663 /*
2664  * Stuff that must be initialised only on boot or module load
2665  */
2666 static int
2667 ipfw_init(void)
2668 {
2669 	int error = 0;
2670 
2671 	/*
2672  	 * Only print out this stuff the first time around,
2673 	 * when called from the sysinit code.
2674 	 */
2675 	printf("ipfw2 "
2676 #ifdef INET6
2677 		"(+ipv6) "
2678 #endif
2679 		"initialized, divert %s, nat %s, "
2680 		"default to %s, logging ",
2681 #ifdef IPDIVERT
2682 		"enabled",
2683 #else
2684 		"loadable",
2685 #endif
2686 #ifdef IPFIREWALL_NAT
2687 		"enabled",
2688 #else
2689 		"loadable",
2690 #endif
2691 		default_to_accept ? "accept" : "deny");
2692 
2693 	/*
2694 	 * Note: V_xxx variables can be accessed here but the vnet specific
2695 	 * initializer may not have been called yet for the VIMAGE case.
2696 	 * Tuneables will have been processed. We will print out values for
2697 	 * the default vnet.
2698 	 * XXX This should all be rationalized AFTER 8.0
2699 	 */
2700 	if (V_fw_verbose == 0)
2701 		printf("disabled\n");
2702 	else if (V_verbose_limit == 0)
2703 		printf("unlimited\n");
2704 	else
2705 		printf("limited to %d packets/entry by default\n",
2706 		    V_verbose_limit);
2707 
2708 	/* Check user-supplied table count for validness */
2709 	if (default_fw_tables > IPFW_TABLES_MAX)
2710 	  default_fw_tables = IPFW_TABLES_MAX;
2711 
2712 	ipfw_init_sopt_handler();
2713 	ipfw_init_obj_rewriter();
2714 	ipfw_iface_init();
2715 	return (error);
2716 }
2717 
2718 /*
2719  * Called for the removal of the last instance only on module unload.
2720  */
2721 static void
2722 ipfw_destroy(void)
2723 {
2724 
2725 	ipfw_iface_destroy();
2726 	ipfw_destroy_sopt_handler();
2727 	ipfw_destroy_obj_rewriter();
2728 	printf("IP firewall unloaded\n");
2729 }
2730 
2731 /*
2732  * Stuff that must be initialized for every instance
2733  * (including the first of course).
2734  */
2735 static int
2736 vnet_ipfw_init(const void *unused)
2737 {
2738 	int error, first;
2739 	struct ip_fw *rule = NULL;
2740 	struct ip_fw_chain *chain;
2741 
2742 	chain = &V_layer3_chain;
2743 
2744 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2745 
2746 	/* First set up some values that are compile time options */
2747 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2748 	V_fw_deny_unknown_exthdrs = 1;
2749 #ifdef IPFIREWALL_VERBOSE
2750 	V_fw_verbose = 1;
2751 #endif
2752 #ifdef IPFIREWALL_VERBOSE_LIMIT
2753 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2754 #endif
2755 #ifdef IPFIREWALL_NAT
2756 	LIST_INIT(&chain->nat);
2757 #endif
2758 
2759 	/* Init shared services hash table */
2760 	ipfw_init_srv(chain);
2761 
2762 	ipfw_init_counters();
2763 	/* insert the default rule and create the initial map */
2764 	chain->n_rules = 1;
2765 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2766 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2767 
2768 	/* Set initial number of tables */
2769 	V_fw_tables_max = default_fw_tables;
2770 	error = ipfw_init_tables(chain, first);
2771 	if (error) {
2772 		printf("ipfw2: setting up tables failed\n");
2773 		free(chain->map, M_IPFW);
2774 		free(rule, M_IPFW);
2775 		return (ENOSPC);
2776 	}
2777 
2778 	/* fill and insert the default rule */
2779 	rule->act_ofs = 0;
2780 	rule->rulenum = IPFW_DEFAULT_RULE;
2781 	rule->cmd_len = 1;
2782 	rule->set = RESVD_SET;
2783 	rule->cmd[0].len = 1;
2784 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2785 	chain->default_rule = chain->map[0] = rule;
2786 	chain->id = rule->id = 1;
2787 	/* Pre-calculate rules length for legacy dump format */
2788 	chain->static_len = sizeof(struct ip_fw_rule0);
2789 
2790 	IPFW_LOCK_INIT(chain);
2791 	ipfw_dyn_init(chain);
2792 	ipfw_eaction_init(chain, first);
2793 #ifdef LINEAR_SKIPTO
2794 	ipfw_init_skipto_cache(chain);
2795 #endif
2796 	ipfw_bpf_init(first);
2797 
2798 	/* First set up some values that are compile time options */
2799 	V_ipfw_vnet_ready = 1;		/* Open for business */
2800 
2801 	/*
2802 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2803 	 * Even if the latter two fail we still keep the module alive
2804 	 * because the sockopt and layer2 paths are still useful.
2805 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2806 	 * so we can ignore the exact return value and just set a flag.
2807 	 *
2808 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2809 	 * changes in the underlying (per-vnet) variables trigger
2810 	 * immediate hook()/unhook() calls.
2811 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2812 	 * is checked on each packet because there are no pfil hooks.
2813 	 */
2814 	V_ip_fw_ctl_ptr = ipfw_ctl3;
2815 	error = ipfw_attach_hooks(1);
2816 	return (error);
2817 }
2818 
2819 /*
2820  * Called for the removal of each instance.
2821  */
2822 static int
2823 vnet_ipfw_uninit(const void *unused)
2824 {
2825 	struct ip_fw *reap;
2826 	struct ip_fw_chain *chain = &V_layer3_chain;
2827 	int i, last;
2828 
2829 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2830 	/*
2831 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2832 	 * Then grab, release and grab again the WLOCK so we make
2833 	 * sure the update is propagated and nobody will be in.
2834 	 */
2835 	(void)ipfw_attach_hooks(0 /* detach */);
2836 	V_ip_fw_ctl_ptr = NULL;
2837 
2838 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2839 
2840 	IPFW_UH_WLOCK(chain);
2841 	IPFW_UH_WUNLOCK(chain);
2842 
2843 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2844 
2845 	IPFW_UH_WLOCK(chain);
2846 
2847 	reap = NULL;
2848 	IPFW_WLOCK(chain);
2849 	for (i = 0; i < chain->n_rules; i++)
2850 		ipfw_reap_add(chain, &reap, chain->map[i]);
2851 	free(chain->map, M_IPFW);
2852 #ifdef LINEAR_SKIPTO
2853 	ipfw_destroy_skipto_cache(chain);
2854 #endif
2855 	IPFW_WUNLOCK(chain);
2856 	IPFW_UH_WUNLOCK(chain);
2857 	ipfw_destroy_tables(chain, last);
2858 	ipfw_eaction_uninit(chain, last);
2859 	if (reap != NULL)
2860 		ipfw_reap_rules(reap);
2861 	vnet_ipfw_iface_destroy(chain);
2862 	ipfw_destroy_srv(chain);
2863 	IPFW_LOCK_DESTROY(chain);
2864 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2865 	ipfw_destroy_counters();
2866 	ipfw_bpf_uninit(last);
2867 	return (0);
2868 }
2869 
2870 /*
2871  * Module event handler.
2872  * In general we have the choice of handling most of these events by the
2873  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2874  * use the SYSINIT handlers as they are more capable of expressing the
2875  * flow of control during module and vnet operations, so this is just
2876  * a skeleton. Note there is no SYSINIT equivalent of the module
2877  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2878  */
2879 static int
2880 ipfw_modevent(module_t mod, int type, void *unused)
2881 {
2882 	int err = 0;
2883 
2884 	switch (type) {
2885 	case MOD_LOAD:
2886 		/* Called once at module load or
2887 	 	 * system boot if compiled in. */
2888 		break;
2889 	case MOD_QUIESCE:
2890 		/* Called before unload. May veto unloading. */
2891 		break;
2892 	case MOD_UNLOAD:
2893 		/* Called during unload. */
2894 		break;
2895 	case MOD_SHUTDOWN:
2896 		/* Called during system shutdown. */
2897 		break;
2898 	default:
2899 		err = EOPNOTSUPP;
2900 		break;
2901 	}
2902 	return err;
2903 }
2904 
2905 static moduledata_t ipfwmod = {
2906 	"ipfw",
2907 	ipfw_modevent,
2908 	0
2909 };
2910 
2911 /* Define startup order. */
2912 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
2913 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2914 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2915 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2916 
2917 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2918 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2919 MODULE_VERSION(ipfw, 3);
2920 /* should declare some dependencies here */
2921 
2922 /*
2923  * Starting up. Done in order after ipfwmod() has been called.
2924  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2925  */
2926 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2927 	    ipfw_init, NULL);
2928 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2929 	    vnet_ipfw_init, NULL);
2930 
2931 /*
2932  * Closing up shop. These are done in REVERSE ORDER, but still
2933  * after ipfwmod() has been called. Not called on reboot.
2934  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2935  * or when the module is unloaded.
2936  */
2937 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2938 	    ipfw_destroy, NULL);
2939 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2940 	    vnet_ipfw_uninit, NULL);
2941 /* end of file */
2942