xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 15c433351f54e7cd5bec8d36c8e89e6a7fa55b26)
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 fragement (fragement 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 	ipfw_dyn_rule *q = NULL;
975 	struct ip_fw_chain *chain = &V_layer3_chain;
976 
977 	/*
978 	 * We store in ulp a pointer to the upper layer protocol header.
979 	 * In the ipv4 case this is easy to determine from the header,
980 	 * but for ipv6 we might have some additional headers in the middle.
981 	 * ulp is NULL if not found.
982 	 */
983 	void *ulp = NULL;		/* upper layer protocol pointer. */
984 
985 	/* XXX ipv6 variables */
986 	int is_ipv6 = 0;
987 	uint8_t	icmp6_type = 0;
988 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
989 	/* end of ipv6 variables */
990 
991 	int is_ipv4 = 0;
992 
993 	int done = 0;		/* flag to exit the outer loop */
994 	IPFW_RLOCK_TRACKER;
995 
996 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
997 		return (IP_FW_PASS);	/* accept */
998 
999 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1000 	src_ip.s_addr = 0;		/* make sure it is initialized */
1001 	pktlen = m->m_pkthdr.len;
1002 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1003 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
1004 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1005 
1006 /*
1007  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1008  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1009  * pointer might become stale after other pullups (but we never use it
1010  * this way).
1011  */
1012 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1013 #define PULLUP_LEN(_len, p, T)					\
1014 do {								\
1015 	int x = (_len) + T;					\
1016 	if ((m)->m_len < x) {					\
1017 		args->m = m = m_pullup(m, x);			\
1018 		if (m == NULL)					\
1019 			goto pullup_failed;			\
1020 	}							\
1021 	p = (mtod(m, char *) + (_len));				\
1022 } while (0)
1023 
1024 	/*
1025 	 * if we have an ether header,
1026 	 */
1027 	if (args->eh)
1028 		etype = ntohs(args->eh->ether_type);
1029 
1030 	/* Identify IP packets and fill up variables. */
1031 	if (pktlen >= sizeof(struct ip6_hdr) &&
1032 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1033 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1034 		is_ipv6 = 1;
1035 		args->f_id.addr_type = 6;
1036 		hlen = sizeof(struct ip6_hdr);
1037 		proto = ip6->ip6_nxt;
1038 
1039 		/* Search extension headers to find upper layer protocols */
1040 		while (ulp == NULL && offset == 0) {
1041 			switch (proto) {
1042 			case IPPROTO_ICMPV6:
1043 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1044 				icmp6_type = ICMP6(ulp)->icmp6_type;
1045 				break;
1046 
1047 			case IPPROTO_TCP:
1048 				PULLUP_TO(hlen, ulp, struct tcphdr);
1049 				dst_port = TCP(ulp)->th_dport;
1050 				src_port = TCP(ulp)->th_sport;
1051 				/* save flags for dynamic rules */
1052 				args->f_id._flags = TCP(ulp)->th_flags;
1053 				break;
1054 
1055 			case IPPROTO_SCTP:
1056 				PULLUP_TO(hlen, ulp, struct sctphdr);
1057 				src_port = SCTP(ulp)->src_port;
1058 				dst_port = SCTP(ulp)->dest_port;
1059 				break;
1060 
1061 			case IPPROTO_UDP:
1062 				PULLUP_TO(hlen, ulp, struct udphdr);
1063 				dst_port = UDP(ulp)->uh_dport;
1064 				src_port = UDP(ulp)->uh_sport;
1065 				break;
1066 
1067 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1068 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1069 				ext_hd |= EXT_HOPOPTS;
1070 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1071 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1072 				ulp = NULL;
1073 				break;
1074 
1075 			case IPPROTO_ROUTING:	/* RFC 2460 */
1076 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1077 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1078 				case 0:
1079 					ext_hd |= EXT_RTHDR0;
1080 					break;
1081 				case 2:
1082 					ext_hd |= EXT_RTHDR2;
1083 					break;
1084 				default:
1085 					if (V_fw_verbose)
1086 						printf("IPFW2: IPV6 - Unknown "
1087 						    "Routing Header type(%d)\n",
1088 						    ((struct ip6_rthdr *)
1089 						    ulp)->ip6r_type);
1090 					if (V_fw_deny_unknown_exthdrs)
1091 					    return (IP_FW_DENY);
1092 					break;
1093 				}
1094 				ext_hd |= EXT_ROUTING;
1095 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1096 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1097 				ulp = NULL;
1098 				break;
1099 
1100 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1101 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1102 				ext_hd |= EXT_FRAGMENT;
1103 				hlen += sizeof (struct ip6_frag);
1104 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1105 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1106 					IP6F_OFF_MASK;
1107 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1108 					IP6F_MORE_FRAG;
1109 				if (V_fw_permit_single_frag6 == 0 &&
1110 				    offset == 0 && ip6f_mf == 0) {
1111 					if (V_fw_verbose)
1112 						printf("IPFW2: IPV6 - Invalid "
1113 						    "Fragment Header\n");
1114 					if (V_fw_deny_unknown_exthdrs)
1115 					    return (IP_FW_DENY);
1116 					break;
1117 				}
1118 				args->f_id.extra =
1119 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1120 				ulp = NULL;
1121 				break;
1122 
1123 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1124 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1125 				ext_hd |= EXT_DSTOPTS;
1126 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1127 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1128 				ulp = NULL;
1129 				break;
1130 
1131 			case IPPROTO_AH:	/* RFC 2402 */
1132 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1133 				ext_hd |= EXT_AH;
1134 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1135 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1136 				ulp = NULL;
1137 				break;
1138 
1139 			case IPPROTO_ESP:	/* RFC 2406 */
1140 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1141 				/* Anything past Seq# is variable length and
1142 				 * data past this ext. header is encrypted. */
1143 				ext_hd |= EXT_ESP;
1144 				break;
1145 
1146 			case IPPROTO_NONE:	/* RFC 2460 */
1147 				/*
1148 				 * Packet ends here, and IPv6 header has
1149 				 * already been pulled up. If ip6e_len!=0
1150 				 * then octets must be ignored.
1151 				 */
1152 				ulp = ip; /* non-NULL to get out of loop. */
1153 				break;
1154 
1155 			case IPPROTO_OSPFIGP:
1156 				/* XXX OSPF header check? */
1157 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1158 				break;
1159 
1160 			case IPPROTO_PIM:
1161 				/* XXX PIM header check? */
1162 				PULLUP_TO(hlen, ulp, struct pim);
1163 				break;
1164 
1165 			case IPPROTO_CARP:
1166 				PULLUP_TO(hlen, ulp, struct carp_header);
1167 				if (((struct carp_header *)ulp)->carp_version !=
1168 				    CARP_VERSION)
1169 					return (IP_FW_DENY);
1170 				if (((struct carp_header *)ulp)->carp_type !=
1171 				    CARP_ADVERTISEMENT)
1172 					return (IP_FW_DENY);
1173 				break;
1174 
1175 			case IPPROTO_IPV6:	/* RFC 2893 */
1176 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1177 				break;
1178 
1179 			case IPPROTO_IPV4:	/* RFC 2893 */
1180 				PULLUP_TO(hlen, ulp, struct ip);
1181 				break;
1182 
1183 			default:
1184 				if (V_fw_verbose)
1185 					printf("IPFW2: IPV6 - Unknown "
1186 					    "Extension Header(%d), ext_hd=%x\n",
1187 					     proto, ext_hd);
1188 				if (V_fw_deny_unknown_exthdrs)
1189 				    return (IP_FW_DENY);
1190 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1191 				break;
1192 			} /*switch */
1193 		}
1194 		ip = mtod(m, struct ip *);
1195 		ip6 = (struct ip6_hdr *)ip;
1196 		args->f_id.src_ip6 = ip6->ip6_src;
1197 		args->f_id.dst_ip6 = ip6->ip6_dst;
1198 		args->f_id.src_ip = 0;
1199 		args->f_id.dst_ip = 0;
1200 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1201 	} else if (pktlen >= sizeof(struct ip) &&
1202 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1203 	    	is_ipv4 = 1;
1204 		hlen = ip->ip_hl << 2;
1205 		args->f_id.addr_type = 4;
1206 
1207 		/*
1208 		 * Collect parameters into local variables for faster matching.
1209 		 */
1210 		proto = ip->ip_p;
1211 		src_ip = ip->ip_src;
1212 		dst_ip = ip->ip_dst;
1213 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1214 		iplen = ntohs(ip->ip_len);
1215 		pktlen = iplen < pktlen ? iplen : pktlen;
1216 
1217 		if (offset == 0) {
1218 			switch (proto) {
1219 			case IPPROTO_TCP:
1220 				PULLUP_TO(hlen, ulp, struct tcphdr);
1221 				dst_port = TCP(ulp)->th_dport;
1222 				src_port = TCP(ulp)->th_sport;
1223 				/* save flags for dynamic rules */
1224 				args->f_id._flags = TCP(ulp)->th_flags;
1225 				break;
1226 
1227 			case IPPROTO_SCTP:
1228 				PULLUP_TO(hlen, ulp, struct sctphdr);
1229 				src_port = SCTP(ulp)->src_port;
1230 				dst_port = SCTP(ulp)->dest_port;
1231 				break;
1232 
1233 			case IPPROTO_UDP:
1234 				PULLUP_TO(hlen, ulp, struct udphdr);
1235 				dst_port = UDP(ulp)->uh_dport;
1236 				src_port = UDP(ulp)->uh_sport;
1237 				break;
1238 
1239 			case IPPROTO_ICMP:
1240 				PULLUP_TO(hlen, ulp, struct icmphdr);
1241 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1242 				break;
1243 
1244 			default:
1245 				break;
1246 			}
1247 		}
1248 
1249 		ip = mtod(m, struct ip *);
1250 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1251 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1252 	}
1253 #undef PULLUP_TO
1254 	if (proto) { /* we may have port numbers, store them */
1255 		args->f_id.proto = proto;
1256 		args->f_id.src_port = src_port = ntohs(src_port);
1257 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1258 	}
1259 
1260 	IPFW_PF_RLOCK(chain);
1261 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1262 		IPFW_PF_RUNLOCK(chain);
1263 		return (IP_FW_PASS);	/* accept */
1264 	}
1265 	if (args->rule.slot) {
1266 		/*
1267 		 * Packet has already been tagged as a result of a previous
1268 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1269 		 * REASS, NETGRAPH, DIVERT/TEE...)
1270 		 * Validate the slot and continue from the next one
1271 		 * if still present, otherwise do a lookup.
1272 		 */
1273 		f_pos = (args->rule.chain_id == chain->id) ?
1274 		    args->rule.slot :
1275 		    ipfw_find_rule(chain, args->rule.rulenum,
1276 			args->rule.rule_id);
1277 	} else {
1278 		f_pos = 0;
1279 	}
1280 
1281 	/*
1282 	 * Now scan the rules, and parse microinstructions for each rule.
1283 	 * We have two nested loops and an inner switch. Sometimes we
1284 	 * need to break out of one or both loops, or re-enter one of
1285 	 * the loops with updated variables. Loop variables are:
1286 	 *
1287 	 *	f_pos (outer loop) points to the current rule.
1288 	 *		On output it points to the matching rule.
1289 	 *	done (outer loop) is used as a flag to break the loop.
1290 	 *	l (inner loop)	residual length of current rule.
1291 	 *		cmd points to the current microinstruction.
1292 	 *
1293 	 * We break the inner loop by setting l=0 and possibly
1294 	 * cmdlen=0 if we don't want to advance cmd.
1295 	 * We break the outer loop by setting done=1
1296 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1297 	 * as needed.
1298 	 */
1299 	for (; f_pos < chain->n_rules; f_pos++) {
1300 		ipfw_insn *cmd;
1301 		uint32_t tablearg = 0;
1302 		int l, cmdlen, skip_or; /* skip rest of OR block */
1303 		struct ip_fw *f;
1304 
1305 		f = chain->map[f_pos];
1306 		if (V_set_disable & (1 << f->set) )
1307 			continue;
1308 
1309 		skip_or = 0;
1310 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1311 		    l -= cmdlen, cmd += cmdlen) {
1312 			int match;
1313 
1314 			/*
1315 			 * check_body is a jump target used when we find a
1316 			 * CHECK_STATE, and need to jump to the body of
1317 			 * the target rule.
1318 			 */
1319 
1320 /* check_body: */
1321 			cmdlen = F_LEN(cmd);
1322 			/*
1323 			 * An OR block (insn_1 || .. || insn_n) has the
1324 			 * F_OR bit set in all but the last instruction.
1325 			 * The first match will set "skip_or", and cause
1326 			 * the following instructions to be skipped until
1327 			 * past the one with the F_OR bit clear.
1328 			 */
1329 			if (skip_or) {		/* skip this instruction */
1330 				if ((cmd->len & F_OR) == 0)
1331 					skip_or = 0;	/* next one is good */
1332 				continue;
1333 			}
1334 			match = 0; /* set to 1 if we succeed */
1335 
1336 			switch (cmd->opcode) {
1337 			/*
1338 			 * The first set of opcodes compares the packet's
1339 			 * fields with some pattern, setting 'match' if a
1340 			 * match is found. At the end of the loop there is
1341 			 * logic to deal with F_NOT and F_OR flags associated
1342 			 * with the opcode.
1343 			 */
1344 			case O_NOP:
1345 				match = 1;
1346 				break;
1347 
1348 			case O_FORWARD_MAC:
1349 				printf("ipfw: opcode %d unimplemented\n",
1350 				    cmd->opcode);
1351 				break;
1352 
1353 			case O_GID:
1354 			case O_UID:
1355 			case O_JAIL:
1356 				/*
1357 				 * We only check offset == 0 && proto != 0,
1358 				 * as this ensures that we have a
1359 				 * packet with the ports info.
1360 				 */
1361 				if (offset != 0)
1362 					break;
1363 				if (proto == IPPROTO_TCP ||
1364 				    proto == IPPROTO_UDP)
1365 					match = check_uidgid(
1366 						    (ipfw_insn_u32 *)cmd,
1367 						    args, &ucred_lookup,
1368 #ifdef __FreeBSD__
1369 						    &ucred_cache);
1370 #else
1371 						    (void *)&ucred_cache);
1372 #endif
1373 				break;
1374 
1375 			case O_RECV:
1376 				match = iface_match(m->m_pkthdr.rcvif,
1377 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1378 				break;
1379 
1380 			case O_XMIT:
1381 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1382 				    chain, &tablearg);
1383 				break;
1384 
1385 			case O_VIA:
1386 				match = iface_match(oif ? oif :
1387 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1388 				    chain, &tablearg);
1389 				break;
1390 
1391 			case O_MACADDR2:
1392 				if (args->eh != NULL) {	/* have MAC header */
1393 					u_int32_t *want = (u_int32_t *)
1394 						((ipfw_insn_mac *)cmd)->addr;
1395 					u_int32_t *mask = (u_int32_t *)
1396 						((ipfw_insn_mac *)cmd)->mask;
1397 					u_int32_t *hdr = (u_int32_t *)args->eh;
1398 
1399 					match =
1400 					    ( want[0] == (hdr[0] & mask[0]) &&
1401 					      want[1] == (hdr[1] & mask[1]) &&
1402 					      want[2] == (hdr[2] & mask[2]) );
1403 				}
1404 				break;
1405 
1406 			case O_MAC_TYPE:
1407 				if (args->eh != NULL) {
1408 					u_int16_t *p =
1409 					    ((ipfw_insn_u16 *)cmd)->ports;
1410 					int i;
1411 
1412 					for (i = cmdlen - 1; !match && i>0;
1413 					    i--, p += 2)
1414 						match = (etype >= p[0] &&
1415 						    etype <= p[1]);
1416 				}
1417 				break;
1418 
1419 			case O_FRAG:
1420 				match = (offset != 0);
1421 				break;
1422 
1423 			case O_IN:	/* "out" is "not in" */
1424 				match = (oif == NULL);
1425 				break;
1426 
1427 			case O_LAYER2:
1428 				match = (args->eh != NULL);
1429 				break;
1430 
1431 			case O_DIVERTED:
1432 			    {
1433 				/* For diverted packets, args->rule.info
1434 				 * contains the divert port (in host format)
1435 				 * reason and direction.
1436 				 */
1437 				uint32_t i = args->rule.info;
1438 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1439 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1440 			    }
1441 				break;
1442 
1443 			case O_PROTO:
1444 				/*
1445 				 * We do not allow an arg of 0 so the
1446 				 * check of "proto" only suffices.
1447 				 */
1448 				match = (proto == cmd->arg1);
1449 				break;
1450 
1451 			case O_IP_SRC:
1452 				match = is_ipv4 &&
1453 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1454 				    src_ip.s_addr);
1455 				break;
1456 
1457 			case O_IP_SRC_LOOKUP:
1458 			case O_IP_DST_LOOKUP:
1459 				if (is_ipv4) {
1460 				    uint32_t key =
1461 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1462 					    dst_ip.s_addr : src_ip.s_addr;
1463 				    uint32_t v = 0;
1464 
1465 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1466 					/* generic lookup. The key must be
1467 					 * in 32bit big-endian format.
1468 					 */
1469 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1470 					if (v == 0)
1471 					    key = dst_ip.s_addr;
1472 					else if (v == 1)
1473 					    key = src_ip.s_addr;
1474 					else if (v == 6) /* dscp */
1475 					    key = (ip->ip_tos >> 2) & 0x3f;
1476 					else if (offset != 0)
1477 					    break;
1478 					else if (proto != IPPROTO_TCP &&
1479 						proto != IPPROTO_UDP)
1480 					    break;
1481 					else if (v == 2)
1482 					    key = dst_port;
1483 					else if (v == 3)
1484 					    key = src_port;
1485 #ifndef USERSPACE
1486 					else if (v == 4 || v == 5) {
1487 					    check_uidgid(
1488 						(ipfw_insn_u32 *)cmd,
1489 						args, &ucred_lookup,
1490 #ifdef __FreeBSD__
1491 						&ucred_cache);
1492 					    if (v == 4 /* O_UID */)
1493 						key = ucred_cache->cr_uid;
1494 					    else if (v == 5 /* O_JAIL */)
1495 						key = ucred_cache->cr_prison->pr_id;
1496 #else /* !__FreeBSD__ */
1497 						(void *)&ucred_cache);
1498 					    if (v ==4 /* O_UID */)
1499 						key = ucred_cache.uid;
1500 					    else if (v == 5 /* O_JAIL */)
1501 						key = ucred_cache.xid;
1502 #endif /* !__FreeBSD__ */
1503 					}
1504 #endif /* !USERSPACE */
1505 					else
1506 					    break;
1507 				    }
1508 				    match = ipfw_lookup_table(chain,
1509 					cmd->arg1, key, &v);
1510 				    if (!match)
1511 					break;
1512 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1513 					match =
1514 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1515 				    else
1516 					tablearg = v;
1517 				} else if (is_ipv6) {
1518 					uint32_t v = 0;
1519 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1520 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1521 					match = ipfw_lookup_table_extended(chain,
1522 							cmd->arg1,
1523 							sizeof(struct in6_addr),
1524 							pkey, &v);
1525 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1526 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1527 					if (match)
1528 						tablearg = v;
1529 				}
1530 				break;
1531 
1532 			case O_IP_FLOW_LOOKUP:
1533 				{
1534 					uint32_t v = 0;
1535 					match = ipfw_lookup_table_extended(chain,
1536 					    cmd->arg1, 0, &args->f_id, &v);
1537 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1538 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1539 					if (match)
1540 						tablearg = v;
1541 				}
1542 				break;
1543 			case O_IP_SRC_MASK:
1544 			case O_IP_DST_MASK:
1545 				if (is_ipv4) {
1546 				    uint32_t a =
1547 					(cmd->opcode == O_IP_DST_MASK) ?
1548 					    dst_ip.s_addr : src_ip.s_addr;
1549 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1550 				    int i = cmdlen-1;
1551 
1552 				    for (; !match && i>0; i-= 2, p+= 2)
1553 					match = (p[0] == (a & p[1]));
1554 				}
1555 				break;
1556 
1557 			case O_IP_SRC_ME:
1558 				if (is_ipv4) {
1559 					struct ifnet *tif;
1560 
1561 					INADDR_TO_IFP(src_ip, tif);
1562 					match = (tif != NULL);
1563 					break;
1564 				}
1565 #ifdef INET6
1566 				/* FALLTHROUGH */
1567 			case O_IP6_SRC_ME:
1568 				match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
1569 #endif
1570 				break;
1571 
1572 			case O_IP_DST_SET:
1573 			case O_IP_SRC_SET:
1574 				if (is_ipv4) {
1575 					u_int32_t *d = (u_int32_t *)(cmd+1);
1576 					u_int32_t addr =
1577 					    cmd->opcode == O_IP_DST_SET ?
1578 						args->f_id.dst_ip :
1579 						args->f_id.src_ip;
1580 
1581 					    if (addr < d[0])
1582 						    break;
1583 					    addr -= d[0]; /* subtract base */
1584 					    match = (addr < cmd->arg1) &&
1585 						( d[ 1 + (addr>>5)] &
1586 						  (1<<(addr & 0x1f)) );
1587 				}
1588 				break;
1589 
1590 			case O_IP_DST:
1591 				match = is_ipv4 &&
1592 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1593 				    dst_ip.s_addr);
1594 				break;
1595 
1596 			case O_IP_DST_ME:
1597 				if (is_ipv4) {
1598 					struct ifnet *tif;
1599 
1600 					INADDR_TO_IFP(dst_ip, tif);
1601 					match = (tif != NULL);
1602 					break;
1603 				}
1604 #ifdef INET6
1605 				/* FALLTHROUGH */
1606 			case O_IP6_DST_ME:
1607 				match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
1608 #endif
1609 				break;
1610 
1611 
1612 			case O_IP_SRCPORT:
1613 			case O_IP_DSTPORT:
1614 				/*
1615 				 * offset == 0 && proto != 0 is enough
1616 				 * to guarantee that we have a
1617 				 * packet with port info.
1618 				 */
1619 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1620 				    && offset == 0) {
1621 					u_int16_t x =
1622 					    (cmd->opcode == O_IP_SRCPORT) ?
1623 						src_port : dst_port ;
1624 					u_int16_t *p =
1625 					    ((ipfw_insn_u16 *)cmd)->ports;
1626 					int i;
1627 
1628 					for (i = cmdlen - 1; !match && i>0;
1629 					    i--, p += 2)
1630 						match = (x>=p[0] && x<=p[1]);
1631 				}
1632 				break;
1633 
1634 			case O_ICMPTYPE:
1635 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1636 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1637 				break;
1638 
1639 #ifdef INET6
1640 			case O_ICMP6TYPE:
1641 				match = is_ipv6 && offset == 0 &&
1642 				    proto==IPPROTO_ICMPV6 &&
1643 				    icmp6type_match(
1644 					ICMP6(ulp)->icmp6_type,
1645 					(ipfw_insn_u32 *)cmd);
1646 				break;
1647 #endif /* INET6 */
1648 
1649 			case O_IPOPT:
1650 				match = (is_ipv4 &&
1651 				    ipopts_match(ip, cmd) );
1652 				break;
1653 
1654 			case O_IPVER:
1655 				match = (is_ipv4 &&
1656 				    cmd->arg1 == ip->ip_v);
1657 				break;
1658 
1659 			case O_IPID:
1660 			case O_IPLEN:
1661 			case O_IPTTL:
1662 				if (is_ipv4) {	/* only for IP packets */
1663 				    uint16_t x;
1664 				    uint16_t *p;
1665 				    int i;
1666 
1667 				    if (cmd->opcode == O_IPLEN)
1668 					x = iplen;
1669 				    else if (cmd->opcode == O_IPTTL)
1670 					x = ip->ip_ttl;
1671 				    else /* must be IPID */
1672 					x = ntohs(ip->ip_id);
1673 				    if (cmdlen == 1) {
1674 					match = (cmd->arg1 == x);
1675 					break;
1676 				    }
1677 				    /* otherwise we have ranges */
1678 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1679 				    i = cmdlen - 1;
1680 				    for (; !match && i>0; i--, p += 2)
1681 					match = (x >= p[0] && x <= p[1]);
1682 				}
1683 				break;
1684 
1685 			case O_IPPRECEDENCE:
1686 				match = (is_ipv4 &&
1687 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1688 				break;
1689 
1690 			case O_IPTOS:
1691 				match = (is_ipv4 &&
1692 				    flags_match(cmd, ip->ip_tos));
1693 				break;
1694 
1695 			case O_DSCP:
1696 			    {
1697 				uint32_t *p;
1698 				uint16_t x;
1699 
1700 				p = ((ipfw_insn_u32 *)cmd)->d;
1701 
1702 				if (is_ipv4)
1703 					x = ip->ip_tos >> 2;
1704 				else if (is_ipv6) {
1705 					uint8_t *v;
1706 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
1707 					x = (*v & 0x0F) << 2;
1708 					v++;
1709 					x |= *v >> 6;
1710 				} else
1711 					break;
1712 
1713 				/* DSCP bitmask is stored as low_u32 high_u32 */
1714 				if (x >= 32)
1715 					match = *(p + 1) & (1 << (x - 32));
1716 				else
1717 					match = *p & (1 << x);
1718 			    }
1719 				break;
1720 
1721 			case O_TCPDATALEN:
1722 				if (proto == IPPROTO_TCP && offset == 0) {
1723 				    struct tcphdr *tcp;
1724 				    uint16_t x;
1725 				    uint16_t *p;
1726 				    int i;
1727 
1728 				    tcp = TCP(ulp);
1729 				    x = iplen -
1730 					((ip->ip_hl + tcp->th_off) << 2);
1731 				    if (cmdlen == 1) {
1732 					match = (cmd->arg1 == x);
1733 					break;
1734 				    }
1735 				    /* otherwise we have ranges */
1736 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1737 				    i = cmdlen - 1;
1738 				    for (; !match && i>0; i--, p += 2)
1739 					match = (x >= p[0] && x <= p[1]);
1740 				}
1741 				break;
1742 
1743 			case O_TCPFLAGS:
1744 				match = (proto == IPPROTO_TCP && offset == 0 &&
1745 				    flags_match(cmd, TCP(ulp)->th_flags));
1746 				break;
1747 
1748 			case O_TCPOPTS:
1749 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
1750 					PULLUP_LEN(hlen, ulp,
1751 					    (TCP(ulp)->th_off << 2));
1752 					match = tcpopts_match(TCP(ulp), cmd);
1753 				}
1754 				break;
1755 
1756 			case O_TCPSEQ:
1757 				match = (proto == IPPROTO_TCP && offset == 0 &&
1758 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1759 					TCP(ulp)->th_seq);
1760 				break;
1761 
1762 			case O_TCPACK:
1763 				match = (proto == IPPROTO_TCP && offset == 0 &&
1764 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1765 					TCP(ulp)->th_ack);
1766 				break;
1767 
1768 			case O_TCPWIN:
1769 				if (proto == IPPROTO_TCP && offset == 0) {
1770 				    uint16_t x;
1771 				    uint16_t *p;
1772 				    int i;
1773 
1774 				    x = ntohs(TCP(ulp)->th_win);
1775 				    if (cmdlen == 1) {
1776 					match = (cmd->arg1 == x);
1777 					break;
1778 				    }
1779 				    /* Otherwise we have ranges. */
1780 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1781 				    i = cmdlen - 1;
1782 				    for (; !match && i > 0; i--, p += 2)
1783 					match = (x >= p[0] && x <= p[1]);
1784 				}
1785 				break;
1786 
1787 			case O_ESTAB:
1788 				/* reject packets which have SYN only */
1789 				/* XXX should i also check for TH_ACK ? */
1790 				match = (proto == IPPROTO_TCP && offset == 0 &&
1791 				    (TCP(ulp)->th_flags &
1792 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1793 				break;
1794 
1795 			case O_ALTQ: {
1796 				struct pf_mtag *at;
1797 				struct m_tag *mtag;
1798 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1799 
1800 				/*
1801 				 * ALTQ uses mbuf tags from another
1802 				 * packet filtering system - pf(4).
1803 				 * We allocate a tag in its format
1804 				 * and fill it in, pretending to be pf(4).
1805 				 */
1806 				match = 1;
1807 				at = pf_find_mtag(m);
1808 				if (at != NULL && at->qid != 0)
1809 					break;
1810 				mtag = m_tag_get(PACKET_TAG_PF,
1811 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1812 				if (mtag == NULL) {
1813 					/*
1814 					 * Let the packet fall back to the
1815 					 * default ALTQ.
1816 					 */
1817 					break;
1818 				}
1819 				m_tag_prepend(m, mtag);
1820 				at = (struct pf_mtag *)(mtag + 1);
1821 				at->qid = altq->qid;
1822 				at->hdr = ip;
1823 				break;
1824 			}
1825 
1826 			case O_LOG:
1827 				ipfw_log(chain, f, hlen, args, m,
1828 				    oif, offset | ip6f_mf, tablearg, ip);
1829 				match = 1;
1830 				break;
1831 
1832 			case O_PROB:
1833 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1834 				break;
1835 
1836 			case O_VERREVPATH:
1837 				/* Outgoing packets automatically pass/match */
1838 				match = ((oif != NULL) ||
1839 				    (m->m_pkthdr.rcvif == NULL) ||
1840 				    (
1841 #ifdef INET6
1842 				    is_ipv6 ?
1843 					verify_path6(&(args->f_id.src_ip6),
1844 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1845 #endif
1846 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1847 				        args->f_id.fib)));
1848 				break;
1849 
1850 			case O_VERSRCREACH:
1851 				/* Outgoing packets automatically pass/match */
1852 				match = (hlen > 0 && ((oif != NULL) ||
1853 #ifdef INET6
1854 				    is_ipv6 ?
1855 				        verify_path6(&(args->f_id.src_ip6),
1856 				            NULL, args->f_id.fib) :
1857 #endif
1858 				    verify_path(src_ip, NULL, args->f_id.fib)));
1859 				break;
1860 
1861 			case O_ANTISPOOF:
1862 				/* Outgoing packets automatically pass/match */
1863 				if (oif == NULL && hlen > 0 &&
1864 				    (  (is_ipv4 && in_localaddr(src_ip))
1865 #ifdef INET6
1866 				    || (is_ipv6 &&
1867 				        in6_localaddr(&(args->f_id.src_ip6)))
1868 #endif
1869 				    ))
1870 					match =
1871 #ifdef INET6
1872 					    is_ipv6 ? verify_path6(
1873 					        &(args->f_id.src_ip6),
1874 					        m->m_pkthdr.rcvif,
1875 						args->f_id.fib) :
1876 #endif
1877 					    verify_path(src_ip,
1878 					    	m->m_pkthdr.rcvif,
1879 					        args->f_id.fib);
1880 				else
1881 					match = 1;
1882 				break;
1883 
1884 			case O_IPSEC:
1885 #ifdef IPSEC
1886 				match = (m_tag_find(m,
1887 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1888 #endif
1889 				/* otherwise no match */
1890 				break;
1891 
1892 #ifdef INET6
1893 			case O_IP6_SRC:
1894 				match = is_ipv6 &&
1895 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1896 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1897 				break;
1898 
1899 			case O_IP6_DST:
1900 				match = is_ipv6 &&
1901 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1902 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1903 				break;
1904 			case O_IP6_SRC_MASK:
1905 			case O_IP6_DST_MASK:
1906 				if (is_ipv6) {
1907 					int i = cmdlen - 1;
1908 					struct in6_addr p;
1909 					struct in6_addr *d =
1910 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1911 
1912 					for (; !match && i > 0; d += 2,
1913 					    i -= F_INSN_SIZE(struct in6_addr)
1914 					    * 2) {
1915 						p = (cmd->opcode ==
1916 						    O_IP6_SRC_MASK) ?
1917 						    args->f_id.src_ip6:
1918 						    args->f_id.dst_ip6;
1919 						APPLY_MASK(&p, &d[1]);
1920 						match =
1921 						    IN6_ARE_ADDR_EQUAL(&d[0],
1922 						    &p);
1923 					}
1924 				}
1925 				break;
1926 
1927 			case O_FLOW6ID:
1928 				match = is_ipv6 &&
1929 				    flow6id_match(args->f_id.flow_id6,
1930 				    (ipfw_insn_u32 *) cmd);
1931 				break;
1932 
1933 			case O_EXT_HDR:
1934 				match = is_ipv6 &&
1935 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1936 				break;
1937 
1938 			case O_IP6:
1939 				match = is_ipv6;
1940 				break;
1941 #endif
1942 
1943 			case O_IP4:
1944 				match = is_ipv4;
1945 				break;
1946 
1947 			case O_TAG: {
1948 				struct m_tag *mtag;
1949 				uint32_t tag = TARG(cmd->arg1, tag);
1950 
1951 				/* Packet is already tagged with this tag? */
1952 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1953 
1954 				/* We have `untag' action when F_NOT flag is
1955 				 * present. And we must remove this mtag from
1956 				 * mbuf and reset `match' to zero (`match' will
1957 				 * be inversed later).
1958 				 * Otherwise we should allocate new mtag and
1959 				 * push it into mbuf.
1960 				 */
1961 				if (cmd->len & F_NOT) { /* `untag' action */
1962 					if (mtag != NULL)
1963 						m_tag_delete(m, mtag);
1964 					match = 0;
1965 				} else {
1966 					if (mtag == NULL) {
1967 						mtag = m_tag_alloc( MTAG_IPFW,
1968 						    tag, 0, M_NOWAIT);
1969 						if (mtag != NULL)
1970 							m_tag_prepend(m, mtag);
1971 					}
1972 					match = 1;
1973 				}
1974 				break;
1975 			}
1976 
1977 			case O_FIB: /* try match the specified fib */
1978 				if (args->f_id.fib == cmd->arg1)
1979 					match = 1;
1980 				break;
1981 
1982 			case O_SOCKARG:	{
1983 #ifndef USERSPACE	/* not supported in userspace */
1984 				struct inpcb *inp = args->inp;
1985 				struct inpcbinfo *pi;
1986 
1987 				if (is_ipv6) /* XXX can we remove this ? */
1988 					break;
1989 
1990 				if (proto == IPPROTO_TCP)
1991 					pi = &V_tcbinfo;
1992 				else if (proto == IPPROTO_UDP)
1993 					pi = &V_udbinfo;
1994 				else
1995 					break;
1996 
1997 				/*
1998 				 * XXXRW: so_user_cookie should almost
1999 				 * certainly be inp_user_cookie?
2000 				 */
2001 
2002 				/* For incomming packet, lookup up the
2003 				inpcb using the src/dest ip/port tuple */
2004 				if (inp == NULL) {
2005 					inp = in_pcblookup(pi,
2006 						src_ip, htons(src_port),
2007 						dst_ip, htons(dst_port),
2008 						INPLOOKUP_RLOCKPCB, NULL);
2009 					if (inp != NULL) {
2010 						tablearg =
2011 						    inp->inp_socket->so_user_cookie;
2012 						if (tablearg)
2013 							match = 1;
2014 						INP_RUNLOCK(inp);
2015 					}
2016 				} else {
2017 					if (inp->inp_socket) {
2018 						tablearg =
2019 						    inp->inp_socket->so_user_cookie;
2020 						if (tablearg)
2021 							match = 1;
2022 					}
2023 				}
2024 #endif /* !USERSPACE */
2025 				break;
2026 			}
2027 
2028 			case O_TAGGED: {
2029 				struct m_tag *mtag;
2030 				uint32_t tag = TARG(cmd->arg1, tag);
2031 
2032 				if (cmdlen == 1) {
2033 					match = m_tag_locate(m, MTAG_IPFW,
2034 					    tag, NULL) != NULL;
2035 					break;
2036 				}
2037 
2038 				/* we have ranges */
2039 				for (mtag = m_tag_first(m);
2040 				    mtag != NULL && !match;
2041 				    mtag = m_tag_next(m, mtag)) {
2042 					uint16_t *p;
2043 					int i;
2044 
2045 					if (mtag->m_tag_cookie != MTAG_IPFW)
2046 						continue;
2047 
2048 					p = ((ipfw_insn_u16 *)cmd)->ports;
2049 					i = cmdlen - 1;
2050 					for(; !match && i > 0; i--, p += 2)
2051 						match =
2052 						    mtag->m_tag_id >= p[0] &&
2053 						    mtag->m_tag_id <= p[1];
2054 				}
2055 				break;
2056 			}
2057 
2058 			/*
2059 			 * The second set of opcodes represents 'actions',
2060 			 * i.e. the terminal part of a rule once the packet
2061 			 * matches all previous patterns.
2062 			 * Typically there is only one action for each rule,
2063 			 * and the opcode is stored at the end of the rule
2064 			 * (but there are exceptions -- see below).
2065 			 *
2066 			 * In general, here we set retval and terminate the
2067 			 * outer loop (would be a 'break 3' in some language,
2068 			 * but we need to set l=0, done=1)
2069 			 *
2070 			 * Exceptions:
2071 			 * O_COUNT and O_SKIPTO actions:
2072 			 *   instead of terminating, we jump to the next rule
2073 			 *   (setting l=0), or to the SKIPTO target (setting
2074 			 *   f/f_len, cmd and l as needed), respectively.
2075 			 *
2076 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2077 			 *   perform some action and set match = 1;
2078 			 *
2079 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2080 			 *   not real 'actions', and are stored right
2081 			 *   before the 'action' part of the rule.
2082 			 *   These opcodes try to install an entry in the
2083 			 *   state tables; if successful, we continue with
2084 			 *   the next opcode (match=1; break;), otherwise
2085 			 *   the packet must be dropped (set retval,
2086 			 *   break loops with l=0, done=1)
2087 			 *
2088 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2089 			 *   cause a lookup of the state table, and a jump
2090 			 *   to the 'action' part of the parent rule
2091 			 *   if an entry is found, or
2092 			 *   (CHECK_STATE only) a jump to the next rule if
2093 			 *   the entry is not found.
2094 			 *   The result of the lookup is cached so that
2095 			 *   further instances of these opcodes become NOPs.
2096 			 *   The jump to the next rule is done by setting
2097 			 *   l=0, cmdlen=0.
2098 			 */
2099 			case O_LIMIT:
2100 			case O_KEEP_STATE:
2101 				if (ipfw_install_state(chain, f,
2102 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2103 					/* error or limit violation */
2104 					retval = IP_FW_DENY;
2105 					l = 0;	/* exit inner loop */
2106 					done = 1; /* exit outer loop */
2107 				}
2108 				match = 1;
2109 				break;
2110 
2111 			case O_PROBE_STATE:
2112 			case O_CHECK_STATE:
2113 				/*
2114 				 * dynamic rules are checked at the first
2115 				 * keep-state or check-state occurrence,
2116 				 * with the result being stored in dyn_dir.
2117 				 * The compiler introduces a PROBE_STATE
2118 				 * instruction for us when we have a
2119 				 * KEEP_STATE (because PROBE_STATE needs
2120 				 * to be run first).
2121 				 */
2122 				if (dyn_dir == MATCH_UNKNOWN &&
2123 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2124 				     &dyn_dir, proto == IPPROTO_TCP ?
2125 					TCP(ulp) : NULL))
2126 					!= NULL) {
2127 					/*
2128 					 * Found dynamic entry, update stats
2129 					 * and jump to the 'action' part of
2130 					 * the parent rule by setting
2131 					 * f, cmd, l and clearing cmdlen.
2132 					 */
2133 					IPFW_INC_DYN_COUNTER(q, pktlen);
2134 					/* XXX we would like to have f_pos
2135 					 * readily accessible in the dynamic
2136 				         * rule, instead of having to
2137 					 * lookup q->rule.
2138 					 */
2139 					f = q->rule;
2140 					f_pos = ipfw_find_rule(chain,
2141 						f->rulenum, f->id);
2142 					cmd = ACTION_PTR(f);
2143 					l = f->cmd_len - f->act_ofs;
2144 					ipfw_dyn_unlock(q);
2145 					cmdlen = 0;
2146 					match = 1;
2147 					break;
2148 				}
2149 				/*
2150 				 * Dynamic entry not found. If CHECK_STATE,
2151 				 * skip to next rule, if PROBE_STATE just
2152 				 * ignore and continue with next opcode.
2153 				 */
2154 				if (cmd->opcode == O_CHECK_STATE)
2155 					l = 0;	/* exit inner loop */
2156 				match = 1;
2157 				break;
2158 
2159 			case O_ACCEPT:
2160 				retval = 0;	/* accept */
2161 				l = 0;		/* exit inner loop */
2162 				done = 1;	/* exit outer loop */
2163 				break;
2164 
2165 			case O_PIPE:
2166 			case O_QUEUE:
2167 				set_match(args, f_pos, chain);
2168 				args->rule.info = TARG(cmd->arg1, pipe);
2169 				if (cmd->opcode == O_PIPE)
2170 					args->rule.info |= IPFW_IS_PIPE;
2171 				if (V_fw_one_pass)
2172 					args->rule.info |= IPFW_ONEPASS;
2173 				retval = IP_FW_DUMMYNET;
2174 				l = 0;          /* exit inner loop */
2175 				done = 1;       /* exit outer loop */
2176 				break;
2177 
2178 			case O_DIVERT:
2179 			case O_TEE:
2180 				if (args->eh) /* not on layer 2 */
2181 				    break;
2182 				/* otherwise this is terminal */
2183 				l = 0;		/* exit inner loop */
2184 				done = 1;	/* exit outer loop */
2185 				retval = (cmd->opcode == O_DIVERT) ?
2186 					IP_FW_DIVERT : IP_FW_TEE;
2187 				set_match(args, f_pos, chain);
2188 				args->rule.info = TARG(cmd->arg1, divert);
2189 				break;
2190 
2191 			case O_COUNT:
2192 				IPFW_INC_RULE_COUNTER(f, pktlen);
2193 				l = 0;		/* exit inner loop */
2194 				break;
2195 
2196 			case O_SKIPTO:
2197 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2198 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2199 			    /*
2200 			     * Skip disabled rules, and re-enter
2201 			     * the inner loop with the correct
2202 			     * f_pos, f, l and cmd.
2203 			     * Also clear cmdlen and skip_or
2204 			     */
2205 			    for (; f_pos < chain->n_rules - 1 &&
2206 				    (V_set_disable &
2207 				     (1 << chain->map[f_pos]->set));
2208 				    f_pos++)
2209 				;
2210 			    /* Re-enter the inner loop at the skipto rule. */
2211 			    f = chain->map[f_pos];
2212 			    l = f->cmd_len;
2213 			    cmd = f->cmd;
2214 			    match = 1;
2215 			    cmdlen = 0;
2216 			    skip_or = 0;
2217 			    continue;
2218 			    break;	/* not reached */
2219 
2220 			case O_CALLRETURN: {
2221 				/*
2222 				 * Implementation of `subroutine' call/return,
2223 				 * in the stack carried in an mbuf tag. This
2224 				 * is different from `skipto' in that any call
2225 				 * address is possible (`skipto' must prevent
2226 				 * backward jumps to avoid endless loops).
2227 				 * We have `return' action when F_NOT flag is
2228 				 * present. The `m_tag_id' field is used as
2229 				 * stack pointer.
2230 				 */
2231 				struct m_tag *mtag;
2232 				uint16_t jmpto, *stack;
2233 
2234 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2235 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2236 				/*
2237 				 * Hand-rolled version of m_tag_locate() with
2238 				 * wildcard `type'.
2239 				 * If not already tagged, allocate new tag.
2240 				 */
2241 				mtag = m_tag_first(m);
2242 				while (mtag != NULL) {
2243 					if (mtag->m_tag_cookie ==
2244 					    MTAG_IPFW_CALL)
2245 						break;
2246 					mtag = m_tag_next(m, mtag);
2247 				}
2248 				if (mtag == NULL && IS_CALL) {
2249 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2250 					    IPFW_CALLSTACK_SIZE *
2251 					    sizeof(uint16_t), M_NOWAIT);
2252 					if (mtag != NULL)
2253 						m_tag_prepend(m, mtag);
2254 				}
2255 
2256 				/*
2257 				 * On error both `call' and `return' just
2258 				 * continue with next rule.
2259 				 */
2260 				if (IS_RETURN && (mtag == NULL ||
2261 				    mtag->m_tag_id == 0)) {
2262 					l = 0;		/* exit inner loop */
2263 					break;
2264 				}
2265 				if (IS_CALL && (mtag == NULL ||
2266 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2267 					printf("ipfw: call stack error, "
2268 					    "go to next rule\n");
2269 					l = 0;		/* exit inner loop */
2270 					break;
2271 				}
2272 
2273 				IPFW_INC_RULE_COUNTER(f, pktlen);
2274 				stack = (uint16_t *)(mtag + 1);
2275 
2276 				/*
2277 				 * The `call' action may use cached f_pos
2278 				 * (in f->next_rule), whose version is written
2279 				 * in f->next_rule.
2280 				 * The `return' action, however, doesn't have
2281 				 * fixed jump address in cmd->arg1 and can't use
2282 				 * cache.
2283 				 */
2284 				if (IS_CALL) {
2285 					stack[mtag->m_tag_id] = f->rulenum;
2286 					mtag->m_tag_id++;
2287 			    		f_pos = JUMP(chain, f, cmd->arg1,
2288 					    tablearg, 1);
2289 				} else {	/* `return' action */
2290 					mtag->m_tag_id--;
2291 					jmpto = stack[mtag->m_tag_id] + 1;
2292 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2293 				}
2294 
2295 				/*
2296 				 * Skip disabled rules, and re-enter
2297 				 * the inner loop with the correct
2298 				 * f_pos, f, l and cmd.
2299 				 * Also clear cmdlen and skip_or
2300 				 */
2301 				for (; f_pos < chain->n_rules - 1 &&
2302 				    (V_set_disable &
2303 				    (1 << chain->map[f_pos]->set)); f_pos++)
2304 					;
2305 				/* Re-enter the inner loop at the dest rule. */
2306 				f = chain->map[f_pos];
2307 				l = f->cmd_len;
2308 				cmd = f->cmd;
2309 				cmdlen = 0;
2310 				skip_or = 0;
2311 				continue;
2312 				break;	/* NOTREACHED */
2313 			}
2314 #undef IS_CALL
2315 #undef IS_RETURN
2316 
2317 			case O_REJECT:
2318 				/*
2319 				 * Drop the packet and send a reject notice
2320 				 * if the packet is not ICMP (or is an ICMP
2321 				 * query), and it is not multicast/broadcast.
2322 				 */
2323 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2324 				    (proto != IPPROTO_ICMP ||
2325 				     is_icmp_query(ICMP(ulp))) &&
2326 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2327 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2328 					send_reject(args, cmd->arg1, iplen, ip);
2329 					m = args->m;
2330 				}
2331 				/* FALLTHROUGH */
2332 #ifdef INET6
2333 			case O_UNREACH6:
2334 				if (hlen > 0 && is_ipv6 &&
2335 				    ((offset & IP6F_OFF_MASK) == 0) &&
2336 				    (proto != IPPROTO_ICMPV6 ||
2337 				     (is_icmp6_query(icmp6_type) == 1)) &&
2338 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2339 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2340 					send_reject6(
2341 					    args, cmd->arg1, hlen,
2342 					    (struct ip6_hdr *)ip);
2343 					m = args->m;
2344 				}
2345 				/* FALLTHROUGH */
2346 #endif
2347 			case O_DENY:
2348 				retval = IP_FW_DENY;
2349 				l = 0;		/* exit inner loop */
2350 				done = 1;	/* exit outer loop */
2351 				break;
2352 
2353 			case O_FORWARD_IP:
2354 				if (args->eh)	/* not valid on layer2 pkts */
2355 					break;
2356 				if (q == NULL || q->rule != f ||
2357 				    dyn_dir == MATCH_FORWARD) {
2358 				    struct sockaddr_in *sa;
2359 
2360 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2361 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2362 #ifdef INET6
2363 					/*
2364 					 * We use O_FORWARD_IP opcode for
2365 					 * fwd rule with tablearg, but tables
2366 					 * now support IPv6 addresses. And
2367 					 * when we are inspecting IPv6 packet,
2368 					 * we can use nh6 field from
2369 					 * table_value as next_hop6 address.
2370 					 */
2371 					if (is_ipv6) {
2372 						struct sockaddr_in6 *sa6;
2373 
2374 						sa6 = args->next_hop6 =
2375 						    &args->hopstore6;
2376 						sa6->sin6_family = AF_INET6;
2377 						sa6->sin6_len = sizeof(*sa6);
2378 						sa6->sin6_addr = TARG_VAL(
2379 						    chain, tablearg, nh6);
2380 						/*
2381 						 * Set sin6_scope_id only for
2382 						 * link-local unicast addresses.
2383 						 */
2384 						if (IN6_IS_ADDR_LINKLOCAL(
2385 						    &sa6->sin6_addr))
2386 							sa6->sin6_scope_id =
2387 							    TARG_VAL(chain,
2388 								tablearg,
2389 								zoneid);
2390 					} else
2391 #endif
2392 					{
2393 						sa = args->next_hop =
2394 						    &args->hopstore;
2395 						sa->sin_family = AF_INET;
2396 						sa->sin_len = sizeof(*sa);
2397 						sa->sin_addr.s_addr = htonl(
2398 						    TARG_VAL(chain, tablearg,
2399 						    nh4));
2400 					}
2401 				    } else {
2402 					args->next_hop = sa;
2403 				    }
2404 				}
2405 				retval = IP_FW_PASS;
2406 				l = 0;          /* exit inner loop */
2407 				done = 1;       /* exit outer loop */
2408 				break;
2409 
2410 #ifdef INET6
2411 			case O_FORWARD_IP6:
2412 				if (args->eh)	/* not valid on layer2 pkts */
2413 					break;
2414 				if (q == NULL || q->rule != f ||
2415 				    dyn_dir == MATCH_FORWARD) {
2416 					struct sockaddr_in6 *sin6;
2417 
2418 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2419 					args->next_hop6 = sin6;
2420 				}
2421 				retval = IP_FW_PASS;
2422 				l = 0;		/* exit inner loop */
2423 				done = 1;	/* exit outer loop */
2424 				break;
2425 #endif
2426 
2427 			case O_NETGRAPH:
2428 			case O_NGTEE:
2429 				set_match(args, f_pos, chain);
2430 				args->rule.info = TARG(cmd->arg1, netgraph);
2431 				if (V_fw_one_pass)
2432 					args->rule.info |= IPFW_ONEPASS;
2433 				retval = (cmd->opcode == O_NETGRAPH) ?
2434 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2435 				l = 0;          /* exit inner loop */
2436 				done = 1;       /* exit outer loop */
2437 				break;
2438 
2439 			case O_SETFIB: {
2440 				uint32_t fib;
2441 
2442 				IPFW_INC_RULE_COUNTER(f, pktlen);
2443 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
2444 				if (fib >= rt_numfibs)
2445 					fib = 0;
2446 				M_SETFIB(m, fib);
2447 				args->f_id.fib = fib;
2448 				l = 0;		/* exit inner loop */
2449 				break;
2450 		        }
2451 
2452 			case O_SETDSCP: {
2453 				uint16_t code;
2454 
2455 				code = TARG(cmd->arg1, dscp) & 0x3F;
2456 				l = 0;		/* exit inner loop */
2457 				if (is_ipv4) {
2458 					uint16_t old;
2459 
2460 					old = *(uint16_t *)ip;
2461 					ip->ip_tos = (code << 2) |
2462 					    (ip->ip_tos & 0x03);
2463 					ip->ip_sum = cksum_adjust(ip->ip_sum,
2464 					    old, *(uint16_t *)ip);
2465 				} else if (is_ipv6) {
2466 					uint8_t *v;
2467 
2468 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2469 					*v = (*v & 0xF0) | (code >> 2);
2470 					v++;
2471 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2472 				} else
2473 					break;
2474 
2475 				IPFW_INC_RULE_COUNTER(f, pktlen);
2476 				break;
2477 			}
2478 
2479 			case O_NAT:
2480 				l = 0;          /* exit inner loop */
2481 				done = 1;       /* exit outer loop */
2482  				if (!IPFW_NAT_LOADED) {
2483 				    retval = IP_FW_DENY;
2484 				    break;
2485 				}
2486 
2487 				struct cfg_nat *t;
2488 				int nat_id;
2489 
2490 				set_match(args, f_pos, chain);
2491 				/* Check if this is 'global' nat rule */
2492 				if (cmd->arg1 == 0) {
2493 					retval = ipfw_nat_ptr(args, NULL, m);
2494 					break;
2495 				}
2496 				t = ((ipfw_insn_nat *)cmd)->nat;
2497 				if (t == NULL) {
2498 					nat_id = TARG(cmd->arg1, nat);
2499 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2500 
2501 					if (t == NULL) {
2502 					    retval = IP_FW_DENY;
2503 					    break;
2504 					}
2505 					if (cmd->arg1 != IP_FW_TARG)
2506 					    ((ipfw_insn_nat *)cmd)->nat = t;
2507 				}
2508 				retval = ipfw_nat_ptr(args, t, m);
2509 				break;
2510 
2511 			case O_REASS: {
2512 				int ip_off;
2513 
2514 				IPFW_INC_RULE_COUNTER(f, pktlen);
2515 				l = 0;	/* in any case exit inner loop */
2516 				ip_off = ntohs(ip->ip_off);
2517 
2518 				/* if not fragmented, go to next rule */
2519 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2520 				    break;
2521 
2522 				args->m = m = ip_reass(m);
2523 
2524 				/*
2525 				 * do IP header checksum fixup.
2526 				 */
2527 				if (m == NULL) { /* fragment got swallowed */
2528 				    retval = IP_FW_DENY;
2529 				} else { /* good, packet complete */
2530 				    int hlen;
2531 
2532 				    ip = mtod(m, struct ip *);
2533 				    hlen = ip->ip_hl << 2;
2534 				    ip->ip_sum = 0;
2535 				    if (hlen == sizeof(struct ip))
2536 					ip->ip_sum = in_cksum_hdr(ip);
2537 				    else
2538 					ip->ip_sum = in_cksum(m, hlen);
2539 				    retval = IP_FW_REASS;
2540 				    set_match(args, f_pos, chain);
2541 				}
2542 				done = 1;	/* exit outer loop */
2543 				break;
2544 			}
2545 
2546 			default:
2547 				panic("-- unknown opcode %d\n", cmd->opcode);
2548 			} /* end of switch() on opcodes */
2549 			/*
2550 			 * if we get here with l=0, then match is irrelevant.
2551 			 */
2552 
2553 			if (cmd->len & F_NOT)
2554 				match = !match;
2555 
2556 			if (match) {
2557 				if (cmd->len & F_OR)
2558 					skip_or = 1;
2559 			} else {
2560 				if (!(cmd->len & F_OR)) /* not an OR block, */
2561 					break;		/* try next rule    */
2562 			}
2563 
2564 		}	/* end of inner loop, scan opcodes */
2565 #undef PULLUP_LEN
2566 
2567 		if (done)
2568 			break;
2569 
2570 /* next_rule:; */	/* try next rule		*/
2571 
2572 	}		/* end of outer for, scan rules */
2573 
2574 	if (done) {
2575 		struct ip_fw *rule = chain->map[f_pos];
2576 		/* Update statistics */
2577 		IPFW_INC_RULE_COUNTER(rule, pktlen);
2578 	} else {
2579 		retval = IP_FW_DENY;
2580 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2581 	}
2582 	IPFW_PF_RUNLOCK(chain);
2583 #ifdef __FreeBSD__
2584 	if (ucred_cache != NULL)
2585 		crfree(ucred_cache);
2586 #endif
2587 	return (retval);
2588 
2589 pullup_failed:
2590 	if (V_fw_verbose)
2591 		printf("ipfw: pullup failed\n");
2592 	return (IP_FW_DENY);
2593 }
2594 
2595 /*
2596  * Set maximum number of tables that can be used in given VNET ipfw instance.
2597  */
2598 #ifdef SYSCTL_NODE
2599 static int
2600 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2601 {
2602 	int error;
2603 	unsigned int ntables;
2604 
2605 	ntables = V_fw_tables_max;
2606 
2607 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2608 	/* Read operation or some error */
2609 	if ((error != 0) || (req->newptr == NULL))
2610 		return (error);
2611 
2612 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2613 }
2614 
2615 /*
2616  * Switches table namespace between global and per-set.
2617  */
2618 static int
2619 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2620 {
2621 	int error;
2622 	unsigned int sets;
2623 
2624 	sets = V_fw_tables_sets;
2625 
2626 	error = sysctl_handle_int(oidp, &sets, 0, req);
2627 	/* Read operation or some error */
2628 	if ((error != 0) || (req->newptr == NULL))
2629 		return (error);
2630 
2631 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2632 }
2633 #endif
2634 
2635 /*
2636  * Module and VNET glue
2637  */
2638 
2639 /*
2640  * Stuff that must be initialised only on boot or module load
2641  */
2642 static int
2643 ipfw_init(void)
2644 {
2645 	int error = 0;
2646 
2647 	/*
2648  	 * Only print out this stuff the first time around,
2649 	 * when called from the sysinit code.
2650 	 */
2651 	printf("ipfw2 "
2652 #ifdef INET6
2653 		"(+ipv6) "
2654 #endif
2655 		"initialized, divert %s, nat %s, "
2656 		"default to %s, logging ",
2657 #ifdef IPDIVERT
2658 		"enabled",
2659 #else
2660 		"loadable",
2661 #endif
2662 #ifdef IPFIREWALL_NAT
2663 		"enabled",
2664 #else
2665 		"loadable",
2666 #endif
2667 		default_to_accept ? "accept" : "deny");
2668 
2669 	/*
2670 	 * Note: V_xxx variables can be accessed here but the vnet specific
2671 	 * initializer may not have been called yet for the VIMAGE case.
2672 	 * Tuneables will have been processed. We will print out values for
2673 	 * the default vnet.
2674 	 * XXX This should all be rationalized AFTER 8.0
2675 	 */
2676 	if (V_fw_verbose == 0)
2677 		printf("disabled\n");
2678 	else if (V_verbose_limit == 0)
2679 		printf("unlimited\n");
2680 	else
2681 		printf("limited to %d packets/entry by default\n",
2682 		    V_verbose_limit);
2683 
2684 	/* Check user-supplied table count for validness */
2685 	if (default_fw_tables > IPFW_TABLES_MAX)
2686 	  default_fw_tables = IPFW_TABLES_MAX;
2687 
2688 	ipfw_init_sopt_handler();
2689 	ipfw_log_bpf(1); /* init */
2690 	ipfw_iface_init();
2691 	return (error);
2692 }
2693 
2694 /*
2695  * Called for the removal of the last instance only on module unload.
2696  */
2697 static void
2698 ipfw_destroy(void)
2699 {
2700 
2701 	ipfw_iface_destroy();
2702 	ipfw_log_bpf(0); /* uninit */
2703 	ipfw_destroy_sopt_handler();
2704 	printf("IP firewall unloaded\n");
2705 }
2706 
2707 /*
2708  * Stuff that must be initialized for every instance
2709  * (including the first of course).
2710  */
2711 static int
2712 vnet_ipfw_init(const void *unused)
2713 {
2714 	int error, first;
2715 	struct ip_fw *rule = NULL;
2716 	struct ip_fw_chain *chain;
2717 
2718 	chain = &V_layer3_chain;
2719 
2720 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2721 
2722 	/* First set up some values that are compile time options */
2723 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2724 	V_fw_deny_unknown_exthdrs = 1;
2725 #ifdef IPFIREWALL_VERBOSE
2726 	V_fw_verbose = 1;
2727 #endif
2728 #ifdef IPFIREWALL_VERBOSE_LIMIT
2729 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2730 #endif
2731 #ifdef IPFIREWALL_NAT
2732 	LIST_INIT(&chain->nat);
2733 #endif
2734 
2735 	/* Init shared services hash table */
2736 	ipfw_init_srv(chain);
2737 
2738 	ipfw_init_obj_rewriter();
2739 	ipfw_init_counters();
2740 	/* insert the default rule and create the initial map */
2741 	chain->n_rules = 1;
2742 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2743 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2744 
2745 	/* Set initial number of tables */
2746 	V_fw_tables_max = default_fw_tables;
2747 	error = ipfw_init_tables(chain, first);
2748 	if (error) {
2749 		printf("ipfw2: setting up tables failed\n");
2750 		free(chain->map, M_IPFW);
2751 		free(rule, M_IPFW);
2752 		return (ENOSPC);
2753 	}
2754 
2755 	/* fill and insert the default rule */
2756 	rule->act_ofs = 0;
2757 	rule->rulenum = IPFW_DEFAULT_RULE;
2758 	rule->cmd_len = 1;
2759 	rule->set = RESVD_SET;
2760 	rule->cmd[0].len = 1;
2761 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2762 	chain->default_rule = chain->map[0] = rule;
2763 	chain->id = rule->id = 1;
2764 	/* Pre-calculate rules length for legacy dump format */
2765 	chain->static_len = sizeof(struct ip_fw_rule0);
2766 
2767 	IPFW_LOCK_INIT(chain);
2768 	ipfw_dyn_init(chain);
2769 #ifdef LINEAR_SKIPTO
2770 	ipfw_init_skipto_cache(chain);
2771 #endif
2772 
2773 	/* First set up some values that are compile time options */
2774 	V_ipfw_vnet_ready = 1;		/* Open for business */
2775 
2776 	/*
2777 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2778 	 * Even if the latter two fail we still keep the module alive
2779 	 * because the sockopt and layer2 paths are still useful.
2780 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2781 	 * so we can ignore the exact return value and just set a flag.
2782 	 *
2783 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2784 	 * changes in the underlying (per-vnet) variables trigger
2785 	 * immediate hook()/unhook() calls.
2786 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2787 	 * is checked on each packet because there are no pfil hooks.
2788 	 */
2789 	V_ip_fw_ctl_ptr = ipfw_ctl3;
2790 	error = ipfw_attach_hooks(1);
2791 	return (error);
2792 }
2793 
2794 /*
2795  * Called for the removal of each instance.
2796  */
2797 static int
2798 vnet_ipfw_uninit(const void *unused)
2799 {
2800 	struct ip_fw *reap;
2801 	struct ip_fw_chain *chain = &V_layer3_chain;
2802 	int i, last;
2803 
2804 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2805 	/*
2806 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2807 	 * Then grab, release and grab again the WLOCK so we make
2808 	 * sure the update is propagated and nobody will be in.
2809 	 */
2810 	(void)ipfw_attach_hooks(0 /* detach */);
2811 	V_ip_fw_ctl_ptr = NULL;
2812 
2813 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2814 
2815 	IPFW_UH_WLOCK(chain);
2816 	IPFW_UH_WUNLOCK(chain);
2817 
2818 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2819 
2820 	IPFW_UH_WLOCK(chain);
2821 
2822 	reap = NULL;
2823 	IPFW_WLOCK(chain);
2824 	for (i = 0; i < chain->n_rules; i++)
2825 		ipfw_reap_add(chain, &reap, chain->map[i]);
2826 	free(chain->map, M_IPFW);
2827 #ifdef LINEAR_SKIPTO
2828 	ipfw_destroy_skipto_cache(chain);
2829 #endif
2830 	IPFW_WUNLOCK(chain);
2831 	IPFW_UH_WUNLOCK(chain);
2832 	ipfw_destroy_tables(chain, last);
2833 	if (reap != NULL)
2834 		ipfw_reap_rules(reap);
2835 	vnet_ipfw_iface_destroy(chain);
2836 	ipfw_destroy_srv(chain);
2837 	IPFW_LOCK_DESTROY(chain);
2838 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2839 	ipfw_destroy_counters();
2840 	ipfw_destroy_obj_rewriter();
2841 	return (0);
2842 }
2843 
2844 /*
2845  * Module event handler.
2846  * In general we have the choice of handling most of these events by the
2847  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2848  * use the SYSINIT handlers as they are more capable of expressing the
2849  * flow of control during module and vnet operations, so this is just
2850  * a skeleton. Note there is no SYSINIT equivalent of the module
2851  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2852  */
2853 static int
2854 ipfw_modevent(module_t mod, int type, void *unused)
2855 {
2856 	int err = 0;
2857 
2858 	switch (type) {
2859 	case MOD_LOAD:
2860 		/* Called once at module load or
2861 	 	 * system boot if compiled in. */
2862 		break;
2863 	case MOD_QUIESCE:
2864 		/* Called before unload. May veto unloading. */
2865 		break;
2866 	case MOD_UNLOAD:
2867 		/* Called during unload. */
2868 		break;
2869 	case MOD_SHUTDOWN:
2870 		/* Called during system shutdown. */
2871 		break;
2872 	default:
2873 		err = EOPNOTSUPP;
2874 		break;
2875 	}
2876 	return err;
2877 }
2878 
2879 static moduledata_t ipfwmod = {
2880 	"ipfw",
2881 	ipfw_modevent,
2882 	0
2883 };
2884 
2885 /* Define startup order. */
2886 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2887 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2888 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2889 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2890 
2891 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2892 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
2893 MODULE_VERSION(ipfw, 3);
2894 /* should declare some dependencies here */
2895 
2896 /*
2897  * Starting up. Done in order after ipfwmod() has been called.
2898  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2899  */
2900 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2901 	    ipfw_init, NULL);
2902 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2903 	    vnet_ipfw_init, NULL);
2904 
2905 /*
2906  * Closing up shop. These are done in REVERSE ORDER, but still
2907  * after ipfwmod() has been called. Not called on reboot.
2908  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2909  * or when the module is unloaded.
2910  */
2911 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2912 	    ipfw_destroy, NULL);
2913 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2914 	    vnet_ipfw_uninit, NULL);
2915 /* end of file */
2916