xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision d06955f9bdb1416d9196043ed781f9b36dae9adc)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
31 /*
32  * The FreeBSD IP packet firewall, main file
33  */
34 
35 #include "opt_ipfw.h"
36 #include "opt_ipdivert.h"
37 #include "opt_inet.h"
38 #ifndef INET
39 #error "IPFIREWALL requires INET"
40 #endif /* INET */
41 #include "opt_inet6.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/condvar.h>
46 #include <sys/counter.h>
47 #include <sys/eventhandler.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/jail.h>
53 #include <sys/module.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/rwlock.h>
57 #include <sys/rmlock.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/syslog.h>
62 #include <sys/ucred.h>
63 #include <net/ethernet.h> /* for ETHERTYPE_IP */
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/route.h>
67 #include <net/pfil.h>
68 #include <net/vnet.h>
69 
70 #include <netpfil/pf/pf_mtag.h>
71 
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_icmp.h>
78 #include <netinet/ip_fw.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet/pim.h>
81 #include <netinet/tcp_var.h>
82 #include <netinet/udp.h>
83 #include <netinet/udp_var.h>
84 #include <netinet/sctp.h>
85 #include <netinet/sctp_crc32.h>
86 #include <netinet/sctp_header.h>
87 
88 #include <netinet/ip6.h>
89 #include <netinet/icmp6.h>
90 #include <netinet/in_fib.h>
91 #ifdef INET6
92 #include <netinet6/in6_fib.h>
93 #include <netinet6/in6_pcb.h>
94 #include <netinet6/scope6_var.h>
95 #include <netinet6/ip6_var.h>
96 #endif
97 
98 #include <net/if_gre.h> /* for struct grehdr */
99 
100 #include <netpfil/ipfw/ip_fw_private.h>
101 
102 #include <machine/in_cksum.h>	/* XXX for in_cksum */
103 
104 #ifdef MAC
105 #include <security/mac/mac_framework.h>
106 #endif
107 
108 /*
109  * static variables followed by global ones.
110  * All ipfw global variables are here.
111  */
112 
113 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
114 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
115 
116 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
117 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
118 
119 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
120 static int default_to_accept = 1;
121 #else
122 static int default_to_accept;
123 #endif
124 
125 VNET_DEFINE(int, autoinc_step);
126 VNET_DEFINE(int, fw_one_pass) = 1;
127 
128 VNET_DEFINE(unsigned int, fw_tables_max);
129 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;	/* Don't use set-aware tables */
130 /* Use 128 tables by default */
131 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
132 
133 #ifndef LINEAR_SKIPTO
134 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
135     int tablearg, int jump_backwards);
136 #define	JUMP(ch, f, num, targ, back)	jump_fast(ch, f, num, targ, back)
137 #else
138 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
139     int tablearg, int jump_backwards);
140 #define	JUMP(ch, f, num, targ, back)	jump_linear(ch, f, num, targ, back)
141 #endif
142 
143 /*
144  * Each rule belongs to one of 32 different sets (0..31).
145  * The variable set_disable contains one bit per set.
146  * If the bit is set, all rules in the corresponding set
147  * are disabled. Set RESVD_SET(31) is reserved for the default rule
148  * and rules that are not deleted by the flush command,
149  * and CANNOT be disabled.
150  * Rules in set RESVD_SET can only be deleted individually.
151  */
152 VNET_DEFINE(u_int32_t, set_disable);
153 #define	V_set_disable			VNET(set_disable)
154 
155 VNET_DEFINE(int, fw_verbose);
156 /* counter for ipfw_log(NULL...) */
157 VNET_DEFINE(u_int64_t, norule_counter);
158 VNET_DEFINE(int, verbose_limit);
159 
160 /* layer3_chain contains the list of rules for layer 3 */
161 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
162 
163 /* ipfw_vnet_ready controls when we are open for business */
164 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
165 
166 VNET_DEFINE(int, ipfw_nat_ready) = 0;
167 
168 ipfw_nat_t *ipfw_nat_ptr = NULL;
169 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
170 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
171 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
172 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
173 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
174 
175 #ifdef SYSCTL_NODE
176 uint32_t dummy_def = IPFW_DEFAULT_RULE;
177 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
178 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
179 
180 SYSBEGIN(f3)
181 
182 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
184     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
185     "Only do a single pass through ipfw when using dummynet(4)");
186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
187     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
188     "Rule number auto-increment step");
189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
190     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
191     "Log matches to ipfw rules");
192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
193     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
194     "Set upper limit of matches of ipfw rules logged");
195 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
196     &dummy_def, 0,
197     "The default/max possible rule number.");
198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
199     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
200     "Maximum number of concurrently used tables");
201 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
202     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
203     0, 0, sysctl_ipfw_tables_sets, "IU",
204     "Use per-set namespace for tables");
205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
206     &default_to_accept, 0,
207     "Make the default rule accept all packets.");
208 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
210     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
211     "Number of static rules");
212 
213 #ifdef INET6
214 SYSCTL_DECL(_net_inet6_ip6);
215 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
216 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
217     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
218     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
219     "Deny packets with unknown IPv6 Extension Headers");
220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
221     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
222     &VNET_NAME(fw_permit_single_frag6), 0,
223     "Permit single packet IPv6 fragments");
224 #endif /* INET6 */
225 
226 SYSEND
227 
228 #endif /* SYSCTL_NODE */
229 
230 
231 /*
232  * Some macros used in the various matching options.
233  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
234  * Other macros just cast void * into the appropriate type
235  */
236 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
237 #define	TCP(p)		((struct tcphdr *)(p))
238 #define	SCTP(p)		((struct sctphdr *)(p))
239 #define	UDP(p)		((struct udphdr *)(p))
240 #define	ICMP(p)		((struct icmphdr *)(p))
241 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
242 
243 static __inline int
244 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
245 {
246 	int type = icmp->icmp_type;
247 
248 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
249 }
250 
251 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
252     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
253 
254 static int
255 is_icmp_query(struct icmphdr *icmp)
256 {
257 	int type = icmp->icmp_type;
258 
259 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
260 }
261 #undef TT
262 
263 /*
264  * The following checks use two arrays of 8 or 16 bits to store the
265  * bits that we want set or clear, respectively. They are in the
266  * low and high half of cmd->arg1 or cmd->d[0].
267  *
268  * We scan options and store the bits we find set. We succeed if
269  *
270  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
271  *
272  * The code is sometimes optimized not to store additional variables.
273  */
274 
275 static int
276 flags_match(ipfw_insn *cmd, u_int8_t bits)
277 {
278 	u_char want_clear;
279 	bits = ~bits;
280 
281 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
282 		return 0; /* some bits we want set were clear */
283 	want_clear = (cmd->arg1 >> 8) & 0xff;
284 	if ( (want_clear & bits) != want_clear)
285 		return 0; /* some bits we want clear were set */
286 	return 1;
287 }
288 
289 static int
290 ipopts_match(struct ip *ip, ipfw_insn *cmd)
291 {
292 	int optlen, bits = 0;
293 	u_char *cp = (u_char *)(ip + 1);
294 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
295 
296 	for (; x > 0; x -= optlen, cp += optlen) {
297 		int opt = cp[IPOPT_OPTVAL];
298 
299 		if (opt == IPOPT_EOL)
300 			break;
301 		if (opt == IPOPT_NOP)
302 			optlen = 1;
303 		else {
304 			optlen = cp[IPOPT_OLEN];
305 			if (optlen <= 0 || optlen > x)
306 				return 0; /* invalid or truncated */
307 		}
308 		switch (opt) {
309 
310 		default:
311 			break;
312 
313 		case IPOPT_LSRR:
314 			bits |= IP_FW_IPOPT_LSRR;
315 			break;
316 
317 		case IPOPT_SSRR:
318 			bits |= IP_FW_IPOPT_SSRR;
319 			break;
320 
321 		case IPOPT_RR:
322 			bits |= IP_FW_IPOPT_RR;
323 			break;
324 
325 		case IPOPT_TS:
326 			bits |= IP_FW_IPOPT_TS;
327 			break;
328 		}
329 	}
330 	return (flags_match(cmd, bits));
331 }
332 
333 static int
334 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
335 {
336 	int optlen, bits = 0;
337 	u_char *cp = (u_char *)(tcp + 1);
338 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
339 
340 	for (; x > 0; x -= optlen, cp += optlen) {
341 		int opt = cp[0];
342 		if (opt == TCPOPT_EOL)
343 			break;
344 		if (opt == TCPOPT_NOP)
345 			optlen = 1;
346 		else {
347 			optlen = cp[1];
348 			if (optlen <= 0)
349 				break;
350 		}
351 
352 		switch (opt) {
353 
354 		default:
355 			break;
356 
357 		case TCPOPT_MAXSEG:
358 			bits |= IP_FW_TCPOPT_MSS;
359 			break;
360 
361 		case TCPOPT_WINDOW:
362 			bits |= IP_FW_TCPOPT_WINDOW;
363 			break;
364 
365 		case TCPOPT_SACK_PERMITTED:
366 		case TCPOPT_SACK:
367 			bits |= IP_FW_TCPOPT_SACK;
368 			break;
369 
370 		case TCPOPT_TIMESTAMP:
371 			bits |= IP_FW_TCPOPT_TS;
372 			break;
373 
374 		}
375 	}
376 	return (flags_match(cmd, bits));
377 }
378 
379 static int
380 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
381     uint32_t *tablearg)
382 {
383 
384 	if (ifp == NULL)	/* no iface with this packet, match fails */
385 		return (0);
386 
387 	/* Check by name or by IP address */
388 	if (cmd->name[0] != '\0') { /* match by name */
389 		if (cmd->name[0] == '\1') /* use tablearg to match */
390 			return ipfw_lookup_table(chain, cmd->p.kidx, 0,
391 			    &ifp->if_index, tablearg);
392 		/* Check name */
393 		if (cmd->p.glob) {
394 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
395 				return(1);
396 		} else {
397 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
398 				return(1);
399 		}
400 	} else {
401 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
402 		struct ifaddr *ia;
403 
404 		if_addr_rlock(ifp);
405 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
406 			if (ia->ifa_addr->sa_family != AF_INET)
407 				continue;
408 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
409 			    (ia->ifa_addr))->sin_addr.s_addr) {
410 				if_addr_runlock(ifp);
411 				return(1);	/* match */
412 			}
413 		}
414 		if_addr_runlock(ifp);
415 #endif /* __FreeBSD__ */
416 	}
417 	return(0);	/* no match, fail ... */
418 }
419 
420 /*
421  * The verify_path function checks if a route to the src exists and
422  * if it is reachable via ifp (when provided).
423  *
424  * The 'verrevpath' option checks that the interface that an IP packet
425  * arrives on is the same interface that traffic destined for the
426  * packet's source address would be routed out of.
427  * The 'versrcreach' option just checks that the source address is
428  * reachable via any route (except default) in the routing table.
429  * These two are a measure to block forged packets. This is also
430  * commonly known as "anti-spoofing" or Unicast Reverse Path
431  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
432  * is purposely reminiscent of the Cisco IOS command,
433  *
434  *   ip verify unicast reverse-path
435  *   ip verify unicast source reachable-via any
436  *
437  * which implements the same functionality. But note that the syntax
438  * is misleading, and the check may be performed on all IP packets
439  * whether unicast, multicast, or broadcast.
440  */
441 static int
442 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
443 {
444 #if defined(USERSPACE) || !defined(__FreeBSD__)
445 	return 0;
446 #else
447 	struct nhop4_basic nh4;
448 
449 	if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
450 		return (0);
451 
452 	/*
453 	 * If ifp is provided, check for equality with rtentry.
454 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
455 	 * in order to pass packets injected back by if_simloop():
456 	 * routing entry (via lo0) for our own address
457 	 * may exist, so we need to handle routing assymetry.
458 	 */
459 	if (ifp != NULL && ifp != nh4.nh_ifp)
460 		return (0);
461 
462 	/* if no ifp provided, check if rtentry is not default route */
463 	if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
464 		return (0);
465 
466 	/* or if this is a blackhole/reject route */
467 	if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
468 		return (0);
469 
470 	/* found valid route */
471 	return 1;
472 #endif /* __FreeBSD__ */
473 }
474 
475 /*
476  * Generate an SCTP packet containing an ABORT chunk. The verification tag
477  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
478  * reflected is not 0.
479  */
480 
481 static struct mbuf *
482 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
483     int reflected)
484 {
485 	struct mbuf *m;
486 	struct ip *ip;
487 #ifdef INET6
488 	struct ip6_hdr *ip6;
489 #endif
490 	struct sctphdr *sctp;
491 	struct sctp_chunkhdr *chunk;
492 	u_int16_t hlen, plen, tlen;
493 
494 	MGETHDR(m, M_NOWAIT, MT_DATA);
495 	if (m == NULL)
496 		return (NULL);
497 
498 	M_SETFIB(m, id->fib);
499 #ifdef MAC
500 	if (replyto != NULL)
501 		mac_netinet_firewall_reply(replyto, m);
502 	else
503 		mac_netinet_firewall_send(m);
504 #else
505 	(void)replyto;		/* don't warn about unused arg */
506 #endif
507 
508 	switch (id->addr_type) {
509 	case 4:
510 		hlen = sizeof(struct ip);
511 		break;
512 #ifdef INET6
513 	case 6:
514 		hlen = sizeof(struct ip6_hdr);
515 		break;
516 #endif
517 	default:
518 		/* XXX: log me?!? */
519 		FREE_PKT(m);
520 		return (NULL);
521 	}
522 	plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
523 	tlen = hlen + plen;
524 	m->m_data += max_linkhdr;
525 	m->m_flags |= M_SKIP_FIREWALL;
526 	m->m_pkthdr.len = m->m_len = tlen;
527 	m->m_pkthdr.rcvif = NULL;
528 	bzero(m->m_data, tlen);
529 
530 	switch (id->addr_type) {
531 	case 4:
532 		ip = mtod(m, struct ip *);
533 
534 		ip->ip_v = 4;
535 		ip->ip_hl = sizeof(struct ip) >> 2;
536 		ip->ip_tos = IPTOS_LOWDELAY;
537 		ip->ip_len = htons(tlen);
538 		ip->ip_id = htons(0);
539 		ip->ip_off = htons(0);
540 		ip->ip_ttl = V_ip_defttl;
541 		ip->ip_p = IPPROTO_SCTP;
542 		ip->ip_sum = 0;
543 		ip->ip_src.s_addr = htonl(id->dst_ip);
544 		ip->ip_dst.s_addr = htonl(id->src_ip);
545 
546 		sctp = (struct sctphdr *)(ip + 1);
547 		break;
548 #ifdef INET6
549 	case 6:
550 		ip6 = mtod(m, struct ip6_hdr *);
551 
552 		ip6->ip6_vfc = IPV6_VERSION;
553 		ip6->ip6_plen = htons(plen);
554 		ip6->ip6_nxt = IPPROTO_SCTP;
555 		ip6->ip6_hlim = IPV6_DEFHLIM;
556 		ip6->ip6_src = id->dst_ip6;
557 		ip6->ip6_dst = id->src_ip6;
558 
559 		sctp = (struct sctphdr *)(ip6 + 1);
560 		break;
561 #endif
562 	}
563 
564 	sctp->src_port = htons(id->dst_port);
565 	sctp->dest_port = htons(id->src_port);
566 	sctp->v_tag = htonl(vtag);
567 	sctp->checksum = htonl(0);
568 
569 	chunk = (struct sctp_chunkhdr *)(sctp + 1);
570 	chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
571 	chunk->chunk_flags = 0;
572 	if (reflected != 0) {
573 		chunk->chunk_flags |= SCTP_HAD_NO_TCB;
574 	}
575 	chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
576 
577 	sctp->checksum = sctp_calculate_cksum(m, hlen);
578 
579 	return (m);
580 }
581 
582 /*
583  * Generate a TCP packet, containing either a RST or a keepalive.
584  * When flags & TH_RST, we are sending a RST packet, because of a
585  * "reset" action matched the packet.
586  * Otherwise we are sending a keepalive, and flags & TH_
587  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
588  * so that MAC can label the reply appropriately.
589  */
590 struct mbuf *
591 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
592     u_int32_t ack, int flags)
593 {
594 	struct mbuf *m = NULL;		/* stupid compiler */
595 	struct ip *h = NULL;		/* stupid compiler */
596 #ifdef INET6
597 	struct ip6_hdr *h6 = NULL;
598 #endif
599 	struct tcphdr *th = NULL;
600 	int len, dir;
601 
602 	MGETHDR(m, M_NOWAIT, MT_DATA);
603 	if (m == NULL)
604 		return (NULL);
605 
606 	M_SETFIB(m, id->fib);
607 #ifdef MAC
608 	if (replyto != NULL)
609 		mac_netinet_firewall_reply(replyto, m);
610 	else
611 		mac_netinet_firewall_send(m);
612 #else
613 	(void)replyto;		/* don't warn about unused arg */
614 #endif
615 
616 	switch (id->addr_type) {
617 	case 4:
618 		len = sizeof(struct ip) + sizeof(struct tcphdr);
619 		break;
620 #ifdef INET6
621 	case 6:
622 		len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
623 		break;
624 #endif
625 	default:
626 		/* XXX: log me?!? */
627 		FREE_PKT(m);
628 		return (NULL);
629 	}
630 	dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
631 
632 	m->m_data += max_linkhdr;
633 	m->m_flags |= M_SKIP_FIREWALL;
634 	m->m_pkthdr.len = m->m_len = len;
635 	m->m_pkthdr.rcvif = NULL;
636 	bzero(m->m_data, len);
637 
638 	switch (id->addr_type) {
639 	case 4:
640 		h = mtod(m, struct ip *);
641 
642 		/* prepare for checksum */
643 		h->ip_p = IPPROTO_TCP;
644 		h->ip_len = htons(sizeof(struct tcphdr));
645 		if (dir) {
646 			h->ip_src.s_addr = htonl(id->src_ip);
647 			h->ip_dst.s_addr = htonl(id->dst_ip);
648 		} else {
649 			h->ip_src.s_addr = htonl(id->dst_ip);
650 			h->ip_dst.s_addr = htonl(id->src_ip);
651 		}
652 
653 		th = (struct tcphdr *)(h + 1);
654 		break;
655 #ifdef INET6
656 	case 6:
657 		h6 = mtod(m, struct ip6_hdr *);
658 
659 		/* prepare for checksum */
660 		h6->ip6_nxt = IPPROTO_TCP;
661 		h6->ip6_plen = htons(sizeof(struct tcphdr));
662 		if (dir) {
663 			h6->ip6_src = id->src_ip6;
664 			h6->ip6_dst = id->dst_ip6;
665 		} else {
666 			h6->ip6_src = id->dst_ip6;
667 			h6->ip6_dst = id->src_ip6;
668 		}
669 
670 		th = (struct tcphdr *)(h6 + 1);
671 		break;
672 #endif
673 	}
674 
675 	if (dir) {
676 		th->th_sport = htons(id->src_port);
677 		th->th_dport = htons(id->dst_port);
678 	} else {
679 		th->th_sport = htons(id->dst_port);
680 		th->th_dport = htons(id->src_port);
681 	}
682 	th->th_off = sizeof(struct tcphdr) >> 2;
683 
684 	if (flags & TH_RST) {
685 		if (flags & TH_ACK) {
686 			th->th_seq = htonl(ack);
687 			th->th_flags = TH_RST;
688 		} else {
689 			if (flags & TH_SYN)
690 				seq++;
691 			th->th_ack = htonl(seq);
692 			th->th_flags = TH_RST | TH_ACK;
693 		}
694 	} else {
695 		/*
696 		 * Keepalive - use caller provided sequence numbers
697 		 */
698 		th->th_seq = htonl(seq);
699 		th->th_ack = htonl(ack);
700 		th->th_flags = TH_ACK;
701 	}
702 
703 	switch (id->addr_type) {
704 	case 4:
705 		th->th_sum = in_cksum(m, len);
706 
707 		/* finish the ip header */
708 		h->ip_v = 4;
709 		h->ip_hl = sizeof(*h) >> 2;
710 		h->ip_tos = IPTOS_LOWDELAY;
711 		h->ip_off = htons(0);
712 		h->ip_len = htons(len);
713 		h->ip_ttl = V_ip_defttl;
714 		h->ip_sum = 0;
715 		break;
716 #ifdef INET6
717 	case 6:
718 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
719 		    sizeof(struct tcphdr));
720 
721 		/* finish the ip6 header */
722 		h6->ip6_vfc |= IPV6_VERSION;
723 		h6->ip6_hlim = IPV6_DEFHLIM;
724 		break;
725 #endif
726 	}
727 
728 	return (m);
729 }
730 
731 #ifdef INET6
732 /*
733  * ipv6 specific rules here...
734  */
735 static __inline int
736 icmp6type_match (int type, ipfw_insn_u32 *cmd)
737 {
738 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
739 }
740 
741 static int
742 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
743 {
744 	int i;
745 	for (i=0; i <= cmd->o.arg1; ++i )
746 		if (curr_flow == cmd->d[i] )
747 			return 1;
748 	return 0;
749 }
750 
751 /* support for IP6_*_ME opcodes */
752 static const struct in6_addr lla_mask = {{{
753 	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
754 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
755 }}};
756 
757 static int
758 ipfw_localip6(struct in6_addr *in6)
759 {
760 	struct rm_priotracker in6_ifa_tracker;
761 	struct in6_ifaddr *ia;
762 
763 	if (IN6_IS_ADDR_MULTICAST(in6))
764 		return (0);
765 
766 	if (!IN6_IS_ADDR_LINKLOCAL(in6))
767 		return (in6_localip(in6));
768 
769 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
770 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
771 		if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
772 			continue;
773 		if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
774 		    in6, &lla_mask)) {
775 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
776 			return (1);
777 		}
778 	}
779 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
780 	return (0);
781 }
782 
783 static int
784 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
785 {
786 	struct nhop6_basic nh6;
787 
788 	if (IN6_IS_SCOPE_LINKLOCAL(src))
789 		return (1);
790 
791 	if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
792 		return (0);
793 
794 	/* If ifp is provided, check for equality with route table. */
795 	if (ifp != NULL && ifp != nh6.nh_ifp)
796 		return (0);
797 
798 	/* if no ifp provided, check if rtentry is not default route */
799 	if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
800 		return (0);
801 
802 	/* or if this is a blackhole/reject route */
803 	if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
804 		return (0);
805 
806 	/* found valid route */
807 	return 1;
808 }
809 
810 static int
811 is_icmp6_query(int icmp6_type)
812 {
813 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
814 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
815 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
816 	    icmp6_type == ICMP6_WRUREQUEST ||
817 	    icmp6_type == ICMP6_FQDN_QUERY ||
818 	    icmp6_type == ICMP6_NI_QUERY))
819 		return (1);
820 
821 	return (0);
822 }
823 
824 static void
825 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
826 {
827 	struct mbuf *m;
828 
829 	m = args->m;
830 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
831 		struct tcphdr *tcp;
832 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
833 
834 		if ((tcp->th_flags & TH_RST) == 0) {
835 			struct mbuf *m0;
836 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
837 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
838 			    tcp->th_flags | TH_RST);
839 			if (m0 != NULL)
840 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
841 				    NULL);
842 		}
843 		FREE_PKT(m);
844 	} else if (code == ICMP6_UNREACH_ABORT &&
845 	    args->f_id.proto == IPPROTO_SCTP) {
846 		struct mbuf *m0;
847 		struct sctphdr *sctp;
848 		u_int32_t v_tag;
849 		int reflected;
850 
851 		sctp = (struct sctphdr *)((char *)ip6 + hlen);
852 		reflected = 1;
853 		v_tag = ntohl(sctp->v_tag);
854 		/* Investigate the first chunk header if available */
855 		if (m->m_len >= hlen + sizeof(struct sctphdr) +
856 		    sizeof(struct sctp_chunkhdr)) {
857 			struct sctp_chunkhdr *chunk;
858 
859 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
860 			switch (chunk->chunk_type) {
861 			case SCTP_INITIATION:
862 				/*
863 				 * Packets containing an INIT chunk MUST have
864 				 * a zero v-tag.
865 				 */
866 				if (v_tag != 0) {
867 					v_tag = 0;
868 					break;
869 				}
870 				/* INIT chunk MUST NOT be bundled */
871 				if (m->m_pkthdr.len >
872 				    hlen + sizeof(struct sctphdr) +
873 				    ntohs(chunk->chunk_length) + 3) {
874 					break;
875 				}
876 				/* Use the initiate tag if available */
877 				if ((m->m_len >= hlen + sizeof(struct sctphdr) +
878 				    sizeof(struct sctp_chunkhdr) +
879 				    offsetof(struct sctp_init, a_rwnd))) {
880 					struct sctp_init *init;
881 
882 					init = (struct sctp_init *)(chunk + 1);
883 					v_tag = ntohl(init->initiate_tag);
884 					reflected = 0;
885 				}
886 				break;
887 			case SCTP_ABORT_ASSOCIATION:
888 				/*
889 				 * If the packet contains an ABORT chunk, don't
890 				 * reply.
891 				 * XXX: We should search through all chunks,
892 				 *      but don't do to avoid attacks.
893 				 */
894 				v_tag = 0;
895 				break;
896 			}
897 		}
898 		if (v_tag == 0) {
899 			m0 = NULL;
900 		} else {
901 			m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
902 			    reflected);
903 		}
904 		if (m0 != NULL)
905 			ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
906 		FREE_PKT(m);
907 	} else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
908 		/* Send an ICMPv6 unreach. */
909 #if 0
910 		/*
911 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
912 		 * as the contents are read. We need to m_adj() the
913 		 * needed amount.
914 		 * The mbuf will however be thrown away so we can adjust it.
915 		 * Remember we did an m_pullup on it already so we
916 		 * can make some assumptions about contiguousness.
917 		 */
918 		if (args->L3offset)
919 			m_adj(m, args->L3offset);
920 #endif
921 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
922 	} else
923 		FREE_PKT(m);
924 
925 	args->m = NULL;
926 }
927 
928 #endif /* INET6 */
929 
930 
931 /*
932  * sends a reject message, consuming the mbuf passed as an argument.
933  */
934 static void
935 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
936 {
937 
938 #if 0
939 	/* XXX When ip is not guaranteed to be at mtod() we will
940 	 * need to account for this */
941 	 * The mbuf will however be thrown away so we can adjust it.
942 	 * Remember we did an m_pullup on it already so we
943 	 * can make some assumptions about contiguousness.
944 	 */
945 	if (args->L3offset)
946 		m_adj(m, args->L3offset);
947 #endif
948 	if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
949 		/* Send an ICMP unreach */
950 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
951 	} else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
952 		struct tcphdr *const tcp =
953 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
954 		if ( (tcp->th_flags & TH_RST) == 0) {
955 			struct mbuf *m;
956 			m = ipfw_send_pkt(args->m, &(args->f_id),
957 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
958 				tcp->th_flags | TH_RST);
959 			if (m != NULL)
960 				ip_output(m, NULL, NULL, 0, NULL, NULL);
961 		}
962 		FREE_PKT(args->m);
963 	} else if (code == ICMP_REJECT_ABORT &&
964 	    args->f_id.proto == IPPROTO_SCTP) {
965 		struct mbuf *m;
966 		struct sctphdr *sctp;
967 		struct sctp_chunkhdr *chunk;
968 		struct sctp_init *init;
969 		u_int32_t v_tag;
970 		int reflected;
971 
972 		sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
973 		reflected = 1;
974 		v_tag = ntohl(sctp->v_tag);
975 		if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
976 		    sizeof(struct sctp_chunkhdr)) {
977 			/* Look at the first chunk header if available */
978 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
979 			switch (chunk->chunk_type) {
980 			case SCTP_INITIATION:
981 				/*
982 				 * Packets containing an INIT chunk MUST have
983 				 * a zero v-tag.
984 				 */
985 				if (v_tag != 0) {
986 					v_tag = 0;
987 					break;
988 				}
989 				/* INIT chunk MUST NOT be bundled */
990 				if (iplen >
991 				    (ip->ip_hl << 2) + sizeof(struct sctphdr) +
992 				    ntohs(chunk->chunk_length) + 3) {
993 					break;
994 				}
995 				/* Use the initiate tag if available */
996 				if ((iplen >= (ip->ip_hl << 2) +
997 				    sizeof(struct sctphdr) +
998 				    sizeof(struct sctp_chunkhdr) +
999 				    offsetof(struct sctp_init, a_rwnd))) {
1000 					init = (struct sctp_init *)(chunk + 1);
1001 					v_tag = ntohl(init->initiate_tag);
1002 					reflected = 0;
1003 				}
1004 				break;
1005 			case SCTP_ABORT_ASSOCIATION:
1006 				/*
1007 				 * If the packet contains an ABORT chunk, don't
1008 				 * reply.
1009 				 * XXX: We should search through all chunks,
1010 				 * but don't do to avoid attacks.
1011 				 */
1012 				v_tag = 0;
1013 				break;
1014 			}
1015 		}
1016 		if (v_tag == 0) {
1017 			m = NULL;
1018 		} else {
1019 			m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1020 			    reflected);
1021 		}
1022 		if (m != NULL)
1023 			ip_output(m, NULL, NULL, 0, NULL, NULL);
1024 		FREE_PKT(args->m);
1025 	} else
1026 		FREE_PKT(args->m);
1027 	args->m = NULL;
1028 }
1029 
1030 /*
1031  * Support for uid/gid/jail lookup. These tests are expensive
1032  * (because we may need to look into the list of active sockets)
1033  * so we cache the results. ugid_lookupp is 0 if we have not
1034  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1035  * and failed. The function always returns the match value.
1036  * We could actually spare the variable and use *uc, setting
1037  * it to '(void *)check_uidgid if we have no info, NULL if
1038  * we tried and failed, or any other value if successful.
1039  */
1040 static int
1041 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1042     struct ucred **uc)
1043 {
1044 #if defined(USERSPACE)
1045 	return 0;	// not supported in userspace
1046 #else
1047 #ifndef __FreeBSD__
1048 	/* XXX */
1049 	return cred_check(insn, proto, oif,
1050 	    dst_ip, dst_port, src_ip, src_port,
1051 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1052 #else  /* FreeBSD */
1053 	struct in_addr src_ip, dst_ip;
1054 	struct inpcbinfo *pi;
1055 	struct ipfw_flow_id *id;
1056 	struct inpcb *pcb, *inp;
1057 	struct ifnet *oif;
1058 	int lookupflags;
1059 	int match;
1060 
1061 	id = &args->f_id;
1062 	inp = args->inp;
1063 	oif = args->oif;
1064 
1065 	/*
1066 	 * Check to see if the UDP or TCP stack supplied us with
1067 	 * the PCB. If so, rather then holding a lock and looking
1068 	 * up the PCB, we can use the one that was supplied.
1069 	 */
1070 	if (inp && *ugid_lookupp == 0) {
1071 		INP_LOCK_ASSERT(inp);
1072 		if (inp->inp_socket != NULL) {
1073 			*uc = crhold(inp->inp_cred);
1074 			*ugid_lookupp = 1;
1075 		} else
1076 			*ugid_lookupp = -1;
1077 	}
1078 	/*
1079 	 * If we have already been here and the packet has no
1080 	 * PCB entry associated with it, then we can safely
1081 	 * assume that this is a no match.
1082 	 */
1083 	if (*ugid_lookupp == -1)
1084 		return (0);
1085 	if (id->proto == IPPROTO_TCP) {
1086 		lookupflags = 0;
1087 		pi = &V_tcbinfo;
1088 	} else if (id->proto == IPPROTO_UDP) {
1089 		lookupflags = INPLOOKUP_WILDCARD;
1090 		pi = &V_udbinfo;
1091 	} else
1092 		return 0;
1093 	lookupflags |= INPLOOKUP_RLOCKPCB;
1094 	match = 0;
1095 	if (*ugid_lookupp == 0) {
1096 		if (id->addr_type == 6) {
1097 #ifdef INET6
1098 			if (oif == NULL)
1099 				pcb = in6_pcblookup_mbuf(pi,
1100 				    &id->src_ip6, htons(id->src_port),
1101 				    &id->dst_ip6, htons(id->dst_port),
1102 				    lookupflags, oif, args->m);
1103 			else
1104 				pcb = in6_pcblookup_mbuf(pi,
1105 				    &id->dst_ip6, htons(id->dst_port),
1106 				    &id->src_ip6, htons(id->src_port),
1107 				    lookupflags, oif, args->m);
1108 #else
1109 			*ugid_lookupp = -1;
1110 			return (0);
1111 #endif
1112 		} else {
1113 			src_ip.s_addr = htonl(id->src_ip);
1114 			dst_ip.s_addr = htonl(id->dst_ip);
1115 			if (oif == NULL)
1116 				pcb = in_pcblookup_mbuf(pi,
1117 				    src_ip, htons(id->src_port),
1118 				    dst_ip, htons(id->dst_port),
1119 				    lookupflags, oif, args->m);
1120 			else
1121 				pcb = in_pcblookup_mbuf(pi,
1122 				    dst_ip, htons(id->dst_port),
1123 				    src_ip, htons(id->src_port),
1124 				    lookupflags, oif, args->m);
1125 		}
1126 		if (pcb != NULL) {
1127 			INP_RLOCK_ASSERT(pcb);
1128 			*uc = crhold(pcb->inp_cred);
1129 			*ugid_lookupp = 1;
1130 			INP_RUNLOCK(pcb);
1131 		}
1132 		if (*ugid_lookupp == 0) {
1133 			/*
1134 			 * We tried and failed, set the variable to -1
1135 			 * so we will not try again on this packet.
1136 			 */
1137 			*ugid_lookupp = -1;
1138 			return (0);
1139 		}
1140 	}
1141 	if (insn->o.opcode == O_UID)
1142 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1143 	else if (insn->o.opcode == O_GID)
1144 		match = groupmember((gid_t)insn->d[0], *uc);
1145 	else if (insn->o.opcode == O_JAIL)
1146 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1147 	return (match);
1148 #endif /* __FreeBSD__ */
1149 #endif /* not supported in userspace */
1150 }
1151 
1152 /*
1153  * Helper function to set args with info on the rule after the matching
1154  * one. slot is precise, whereas we guess rule_id as they are
1155  * assigned sequentially.
1156  */
1157 static inline void
1158 set_match(struct ip_fw_args *args, int slot,
1159 	struct ip_fw_chain *chain)
1160 {
1161 	args->rule.chain_id = chain->id;
1162 	args->rule.slot = slot + 1; /* we use 0 as a marker */
1163 	args->rule.rule_id = 1 + chain->map[slot]->id;
1164 	args->rule.rulenum = chain->map[slot]->rulenum;
1165 }
1166 
1167 #ifndef LINEAR_SKIPTO
1168 /*
1169  * Helper function to enable cached rule lookups using
1170  * cached_id and cached_pos fields in ipfw rule.
1171  */
1172 static int
1173 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1174     int tablearg, int jump_backwards)
1175 {
1176 	int f_pos;
1177 
1178 	/* If possible use cached f_pos (in f->cached_pos),
1179 	 * whose version is written in f->cached_id
1180 	 * (horrible hacks to avoid changing the ABI).
1181 	 */
1182 	if (num != IP_FW_TARG && f->cached_id == chain->id)
1183 		f_pos = f->cached_pos;
1184 	else {
1185 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1186 		/* make sure we do not jump backward */
1187 		if (jump_backwards == 0 && i <= f->rulenum)
1188 			i = f->rulenum + 1;
1189 		if (chain->idxmap != NULL)
1190 			f_pos = chain->idxmap[i];
1191 		else
1192 			f_pos = ipfw_find_rule(chain, i, 0);
1193 		/* update the cache */
1194 		if (num != IP_FW_TARG) {
1195 			f->cached_id = chain->id;
1196 			f->cached_pos = f_pos;
1197 		}
1198 	}
1199 
1200 	return (f_pos);
1201 }
1202 #else
1203 /*
1204  * Helper function to enable real fast rule lookups.
1205  */
1206 static int
1207 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1208     int tablearg, int jump_backwards)
1209 {
1210 	int f_pos;
1211 
1212 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1213 	/* make sure we do not jump backward */
1214 	if (jump_backwards == 0 && num <= f->rulenum)
1215 		num = f->rulenum + 1;
1216 	f_pos = chain->idxmap[num];
1217 
1218 	return (f_pos);
1219 }
1220 #endif
1221 
1222 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
1223 /*
1224  * The main check routine for the firewall.
1225  *
1226  * All arguments are in args so we can modify them and return them
1227  * back to the caller.
1228  *
1229  * Parameters:
1230  *
1231  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
1232  *		Starts with the IP header.
1233  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
1234  *	args->L3offset	Number of bytes bypassed if we came from L2.
1235  *			e.g. often sizeof(eh)  ** NOTYET **
1236  *	args->oif	Outgoing interface, NULL if packet is incoming.
1237  *		The incoming interface is in the mbuf. (in)
1238  *	args->divert_rule (in/out)
1239  *		Skip up to the first rule past this rule number;
1240  *		upon return, non-zero port number for divert or tee.
1241  *
1242  *	args->rule	Pointer to the last matching rule (in/out)
1243  *	args->next_hop	Socket we are forwarding to (out).
1244  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
1245  *	args->f_id	Addresses grabbed from the packet (out)
1246  * 	args->rule.info	a cookie depending on rule action
1247  *
1248  * Return value:
1249  *
1250  *	IP_FW_PASS	the packet must be accepted
1251  *	IP_FW_DENY	the packet must be dropped
1252  *	IP_FW_DIVERT	divert packet, port in m_tag
1253  *	IP_FW_TEE	tee packet, port in m_tag
1254  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
1255  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
1256  *		args->rule contains the matching rule,
1257  *		args->rule.info has additional information.
1258  *
1259  */
1260 int
1261 ipfw_chk(struct ip_fw_args *args)
1262 {
1263 
1264 	/*
1265 	 * Local variables holding state while processing a packet:
1266 	 *
1267 	 * IMPORTANT NOTE: to speed up the processing of rules, there
1268 	 * are some assumption on the values of the variables, which
1269 	 * are documented here. Should you change them, please check
1270 	 * the implementation of the various instructions to make sure
1271 	 * that they still work.
1272 	 *
1273 	 * args->eh	The MAC header. It is non-null for a layer2
1274 	 *	packet, it is NULL for a layer-3 packet.
1275 	 * **notyet**
1276 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1277 	 *
1278 	 * m | args->m	Pointer to the mbuf, as received from the caller.
1279 	 *	It may change if ipfw_chk() does an m_pullup, or if it
1280 	 *	consumes the packet because it calls send_reject().
1281 	 *	XXX This has to change, so that ipfw_chk() never modifies
1282 	 *	or consumes the buffer.
1283 	 * ip	is the beginning of the ip(4 or 6) header.
1284 	 *	Calculated by adding the L3offset to the start of data.
1285 	 *	(Until we start using L3offset, the packet is
1286 	 *	supposed to start with the ip header).
1287 	 */
1288 	struct mbuf *m = args->m;
1289 	struct ip *ip = mtod(m, struct ip *);
1290 
1291 	/*
1292 	 * For rules which contain uid/gid or jail constraints, cache
1293 	 * a copy of the users credentials after the pcb lookup has been
1294 	 * executed. This will speed up the processing of rules with
1295 	 * these types of constraints, as well as decrease contention
1296 	 * on pcb related locks.
1297 	 */
1298 #ifndef __FreeBSD__
1299 	struct bsd_ucred ucred_cache;
1300 #else
1301 	struct ucred *ucred_cache = NULL;
1302 #endif
1303 	int ucred_lookup = 0;
1304 
1305 	/*
1306 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
1307 	 *	inbound path (ether_input, ip_input).
1308 	 *	If non-NULL, ipfw_chk has been called on the outbound path
1309 	 *	(ether_output, ip_output).
1310 	 */
1311 	struct ifnet *oif = args->oif;
1312 
1313 	int f_pos = 0;		/* index of current rule in the array */
1314 	int retval = 0;
1315 
1316 	/*
1317 	 * hlen	The length of the IP header.
1318 	 */
1319 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
1320 
1321 	/*
1322 	 * offset	The offset of a fragment. offset != 0 means that
1323 	 *	we have a fragment at this offset of an IPv4 packet.
1324 	 *	offset == 0 means that (if this is an IPv4 packet)
1325 	 *	this is the first or only fragment.
1326 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1327 	 *	or there is a single packet fragment (fragment header added
1328 	 *	without needed).  We will treat a single packet fragment as if
1329 	 *	there was no fragment header (or log/block depending on the
1330 	 *	V_fw_permit_single_frag6 sysctl setting).
1331 	 */
1332 	u_short offset = 0;
1333 	u_short ip6f_mf = 0;
1334 
1335 	/*
1336 	 * Local copies of addresses. They are only valid if we have
1337 	 * an IP packet.
1338 	 *
1339 	 * proto	The protocol. Set to 0 for non-ip packets,
1340 	 *	or to the protocol read from the packet otherwise.
1341 	 *	proto != 0 means that we have an IPv4 packet.
1342 	 *
1343 	 * src_port, dst_port	port numbers, in HOST format. Only
1344 	 *	valid for TCP and UDP packets.
1345 	 *
1346 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
1347 	 *	Only valid for IPv4 packets.
1348 	 */
1349 	uint8_t proto;
1350 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
1351 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
1352 	int iplen = 0;
1353 	int pktlen;
1354 	uint16_t	etype = 0;	/* Host order stored ether type */
1355 
1356 	/*
1357 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1358 	 * 	MATCH_NONE when checked and not matched (q = NULL),
1359 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1360 	 */
1361 	int dyn_dir = MATCH_UNKNOWN;
1362 	uint16_t dyn_name = 0;
1363 	struct ip_fw *q = NULL;
1364 	struct ip_fw_chain *chain = &V_layer3_chain;
1365 
1366 	/*
1367 	 * We store in ulp a pointer to the upper layer protocol header.
1368 	 * In the ipv4 case this is easy to determine from the header,
1369 	 * but for ipv6 we might have some additional headers in the middle.
1370 	 * ulp is NULL if not found.
1371 	 */
1372 	void *ulp = NULL;		/* upper layer protocol pointer. */
1373 
1374 	/* XXX ipv6 variables */
1375 	int is_ipv6 = 0;
1376 	uint8_t	icmp6_type = 0;
1377 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1378 	/* end of ipv6 variables */
1379 
1380 	int is_ipv4 = 0;
1381 
1382 	int done = 0;		/* flag to exit the outer loop */
1383 
1384 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1385 		return (IP_FW_PASS);	/* accept */
1386 
1387 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1388 	src_ip.s_addr = 0;		/* make sure it is initialized */
1389 	pktlen = m->m_pkthdr.len;
1390 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1391 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
1392 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1393 
1394 /*
1395  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1396  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1397  * pointer might become stale after other pullups (but we never use it
1398  * this way).
1399  */
1400 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1401 #define PULLUP_LEN(_len, p, T)					\
1402 do {								\
1403 	int x = (_len) + T;					\
1404 	if ((m)->m_len < x) {					\
1405 		args->m = m = m_pullup(m, x);			\
1406 		if (m == NULL)					\
1407 			goto pullup_failed;			\
1408 	}							\
1409 	p = (mtod(m, char *) + (_len));				\
1410 } while (0)
1411 
1412 	/*
1413 	 * if we have an ether header,
1414 	 */
1415 	if (args->eh)
1416 		etype = ntohs(args->eh->ether_type);
1417 
1418 	/* Identify IP packets and fill up variables. */
1419 	if (pktlen >= sizeof(struct ip6_hdr) &&
1420 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1421 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1422 		is_ipv6 = 1;
1423 		args->f_id.addr_type = 6;
1424 		hlen = sizeof(struct ip6_hdr);
1425 		proto = ip6->ip6_nxt;
1426 
1427 		/* Search extension headers to find upper layer protocols */
1428 		while (ulp == NULL && offset == 0) {
1429 			switch (proto) {
1430 			case IPPROTO_ICMPV6:
1431 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1432 				icmp6_type = ICMP6(ulp)->icmp6_type;
1433 				break;
1434 
1435 			case IPPROTO_TCP:
1436 				PULLUP_TO(hlen, ulp, struct tcphdr);
1437 				dst_port = TCP(ulp)->th_dport;
1438 				src_port = TCP(ulp)->th_sport;
1439 				/* save flags for dynamic rules */
1440 				args->f_id._flags = TCP(ulp)->th_flags;
1441 				break;
1442 
1443 			case IPPROTO_SCTP:
1444 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1445 				    sizeof(struct sctp_chunkhdr) +
1446 				    offsetof(struct sctp_init, a_rwnd))
1447 					PULLUP_LEN(hlen, ulp,
1448 					    sizeof(struct sctphdr) +
1449 					    sizeof(struct sctp_chunkhdr) +
1450 					    offsetof(struct sctp_init, a_rwnd));
1451 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1452 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1453 				else
1454 					PULLUP_LEN(hlen, ulp,
1455 					    sizeof(struct sctphdr));
1456 				src_port = SCTP(ulp)->src_port;
1457 				dst_port = SCTP(ulp)->dest_port;
1458 				break;
1459 
1460 			case IPPROTO_UDP:
1461 				PULLUP_TO(hlen, ulp, struct udphdr);
1462 				dst_port = UDP(ulp)->uh_dport;
1463 				src_port = UDP(ulp)->uh_sport;
1464 				break;
1465 
1466 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1467 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1468 				ext_hd |= EXT_HOPOPTS;
1469 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1470 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1471 				ulp = NULL;
1472 				break;
1473 
1474 			case IPPROTO_ROUTING:	/* RFC 2460 */
1475 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1476 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1477 				case 0:
1478 					ext_hd |= EXT_RTHDR0;
1479 					break;
1480 				case 2:
1481 					ext_hd |= EXT_RTHDR2;
1482 					break;
1483 				default:
1484 					if (V_fw_verbose)
1485 						printf("IPFW2: IPV6 - Unknown "
1486 						    "Routing Header type(%d)\n",
1487 						    ((struct ip6_rthdr *)
1488 						    ulp)->ip6r_type);
1489 					if (V_fw_deny_unknown_exthdrs)
1490 					    return (IP_FW_DENY);
1491 					break;
1492 				}
1493 				ext_hd |= EXT_ROUTING;
1494 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1495 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1496 				ulp = NULL;
1497 				break;
1498 
1499 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1500 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1501 				ext_hd |= EXT_FRAGMENT;
1502 				hlen += sizeof (struct ip6_frag);
1503 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1504 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1505 					IP6F_OFF_MASK;
1506 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1507 					IP6F_MORE_FRAG;
1508 				if (V_fw_permit_single_frag6 == 0 &&
1509 				    offset == 0 && ip6f_mf == 0) {
1510 					if (V_fw_verbose)
1511 						printf("IPFW2: IPV6 - Invalid "
1512 						    "Fragment Header\n");
1513 					if (V_fw_deny_unknown_exthdrs)
1514 					    return (IP_FW_DENY);
1515 					break;
1516 				}
1517 				args->f_id.extra =
1518 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1519 				ulp = NULL;
1520 				break;
1521 
1522 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1523 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1524 				ext_hd |= EXT_DSTOPTS;
1525 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1526 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1527 				ulp = NULL;
1528 				break;
1529 
1530 			case IPPROTO_AH:	/* RFC 2402 */
1531 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1532 				ext_hd |= EXT_AH;
1533 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1534 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1535 				ulp = NULL;
1536 				break;
1537 
1538 			case IPPROTO_ESP:	/* RFC 2406 */
1539 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1540 				/* Anything past Seq# is variable length and
1541 				 * data past this ext. header is encrypted. */
1542 				ext_hd |= EXT_ESP;
1543 				break;
1544 
1545 			case IPPROTO_NONE:	/* RFC 2460 */
1546 				/*
1547 				 * Packet ends here, and IPv6 header has
1548 				 * already been pulled up. If ip6e_len!=0
1549 				 * then octets must be ignored.
1550 				 */
1551 				ulp = ip; /* non-NULL to get out of loop. */
1552 				break;
1553 
1554 			case IPPROTO_OSPFIGP:
1555 				/* XXX OSPF header check? */
1556 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1557 				break;
1558 
1559 			case IPPROTO_PIM:
1560 				/* XXX PIM header check? */
1561 				PULLUP_TO(hlen, ulp, struct pim);
1562 				break;
1563 
1564 			case IPPROTO_GRE:	/* RFC 1701 */
1565 				/* XXX GRE header check? */
1566 				PULLUP_TO(hlen, ulp, struct grehdr);
1567 				break;
1568 
1569 			case IPPROTO_CARP:
1570 				PULLUP_TO(hlen, ulp, struct carp_header);
1571 				if (((struct carp_header *)ulp)->carp_version !=
1572 				    CARP_VERSION)
1573 					return (IP_FW_DENY);
1574 				if (((struct carp_header *)ulp)->carp_type !=
1575 				    CARP_ADVERTISEMENT)
1576 					return (IP_FW_DENY);
1577 				break;
1578 
1579 			case IPPROTO_IPV6:	/* RFC 2893 */
1580 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1581 				break;
1582 
1583 			case IPPROTO_IPV4:	/* RFC 2893 */
1584 				PULLUP_TO(hlen, ulp, struct ip);
1585 				break;
1586 
1587 			default:
1588 				if (V_fw_verbose)
1589 					printf("IPFW2: IPV6 - Unknown "
1590 					    "Extension Header(%d), ext_hd=%x\n",
1591 					     proto, ext_hd);
1592 				if (V_fw_deny_unknown_exthdrs)
1593 				    return (IP_FW_DENY);
1594 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1595 				break;
1596 			} /*switch */
1597 		}
1598 		ip = mtod(m, struct ip *);
1599 		ip6 = (struct ip6_hdr *)ip;
1600 		args->f_id.src_ip6 = ip6->ip6_src;
1601 		args->f_id.dst_ip6 = ip6->ip6_dst;
1602 		args->f_id.src_ip = 0;
1603 		args->f_id.dst_ip = 0;
1604 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1605 		iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1606 	} else if (pktlen >= sizeof(struct ip) &&
1607 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1608 	    	is_ipv4 = 1;
1609 		hlen = ip->ip_hl << 2;
1610 		args->f_id.addr_type = 4;
1611 
1612 		/*
1613 		 * Collect parameters into local variables for faster matching.
1614 		 */
1615 		proto = ip->ip_p;
1616 		src_ip = ip->ip_src;
1617 		dst_ip = ip->ip_dst;
1618 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1619 		iplen = ntohs(ip->ip_len);
1620 
1621 		if (offset == 0) {
1622 			switch (proto) {
1623 			case IPPROTO_TCP:
1624 				PULLUP_TO(hlen, ulp, struct tcphdr);
1625 				dst_port = TCP(ulp)->th_dport;
1626 				src_port = TCP(ulp)->th_sport;
1627 				/* save flags for dynamic rules */
1628 				args->f_id._flags = TCP(ulp)->th_flags;
1629 				break;
1630 
1631 			case IPPROTO_SCTP:
1632 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1633 				    sizeof(struct sctp_chunkhdr) +
1634 				    offsetof(struct sctp_init, a_rwnd))
1635 					PULLUP_LEN(hlen, ulp,
1636 					    sizeof(struct sctphdr) +
1637 					    sizeof(struct sctp_chunkhdr) +
1638 					    offsetof(struct sctp_init, a_rwnd));
1639 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1640 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1641 				else
1642 					PULLUP_LEN(hlen, ulp,
1643 					    sizeof(struct sctphdr));
1644 				src_port = SCTP(ulp)->src_port;
1645 				dst_port = SCTP(ulp)->dest_port;
1646 				break;
1647 
1648 			case IPPROTO_UDP:
1649 				PULLUP_TO(hlen, ulp, struct udphdr);
1650 				dst_port = UDP(ulp)->uh_dport;
1651 				src_port = UDP(ulp)->uh_sport;
1652 				break;
1653 
1654 			case IPPROTO_ICMP:
1655 				PULLUP_TO(hlen, ulp, struct icmphdr);
1656 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1657 				break;
1658 
1659 			default:
1660 				break;
1661 			}
1662 		}
1663 
1664 		ip = mtod(m, struct ip *);
1665 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1666 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1667 	}
1668 #undef PULLUP_TO
1669 	pktlen = iplen < pktlen ? iplen: pktlen;
1670 	if (proto) { /* we may have port numbers, store them */
1671 		args->f_id.proto = proto;
1672 		args->f_id.src_port = src_port = ntohs(src_port);
1673 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1674 	}
1675 
1676 	IPFW_PF_RLOCK(chain);
1677 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1678 		IPFW_PF_RUNLOCK(chain);
1679 		return (IP_FW_PASS);	/* accept */
1680 	}
1681 	if (args->rule.slot) {
1682 		/*
1683 		 * Packet has already been tagged as a result of a previous
1684 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1685 		 * REASS, NETGRAPH, DIVERT/TEE...)
1686 		 * Validate the slot and continue from the next one
1687 		 * if still present, otherwise do a lookup.
1688 		 */
1689 		f_pos = (args->rule.chain_id == chain->id) ?
1690 		    args->rule.slot :
1691 		    ipfw_find_rule(chain, args->rule.rulenum,
1692 			args->rule.rule_id);
1693 	} else {
1694 		f_pos = 0;
1695 	}
1696 
1697 	/*
1698 	 * Now scan the rules, and parse microinstructions for each rule.
1699 	 * We have two nested loops and an inner switch. Sometimes we
1700 	 * need to break out of one or both loops, or re-enter one of
1701 	 * the loops with updated variables. Loop variables are:
1702 	 *
1703 	 *	f_pos (outer loop) points to the current rule.
1704 	 *		On output it points to the matching rule.
1705 	 *	done (outer loop) is used as a flag to break the loop.
1706 	 *	l (inner loop)	residual length of current rule.
1707 	 *		cmd points to the current microinstruction.
1708 	 *
1709 	 * We break the inner loop by setting l=0 and possibly
1710 	 * cmdlen=0 if we don't want to advance cmd.
1711 	 * We break the outer loop by setting done=1
1712 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1713 	 * as needed.
1714 	 */
1715 	for (; f_pos < chain->n_rules; f_pos++) {
1716 		ipfw_insn *cmd;
1717 		uint32_t tablearg = 0;
1718 		int l, cmdlen, skip_or; /* skip rest of OR block */
1719 		struct ip_fw *f;
1720 
1721 		f = chain->map[f_pos];
1722 		if (V_set_disable & (1 << f->set) )
1723 			continue;
1724 
1725 		skip_or = 0;
1726 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1727 		    l -= cmdlen, cmd += cmdlen) {
1728 			int match;
1729 
1730 			/*
1731 			 * check_body is a jump target used when we find a
1732 			 * CHECK_STATE, and need to jump to the body of
1733 			 * the target rule.
1734 			 */
1735 
1736 /* check_body: */
1737 			cmdlen = F_LEN(cmd);
1738 			/*
1739 			 * An OR block (insn_1 || .. || insn_n) has the
1740 			 * F_OR bit set in all but the last instruction.
1741 			 * The first match will set "skip_or", and cause
1742 			 * the following instructions to be skipped until
1743 			 * past the one with the F_OR bit clear.
1744 			 */
1745 			if (skip_or) {		/* skip this instruction */
1746 				if ((cmd->len & F_OR) == 0)
1747 					skip_or = 0;	/* next one is good */
1748 				continue;
1749 			}
1750 			match = 0; /* set to 1 if we succeed */
1751 
1752 			switch (cmd->opcode) {
1753 			/*
1754 			 * The first set of opcodes compares the packet's
1755 			 * fields with some pattern, setting 'match' if a
1756 			 * match is found. At the end of the loop there is
1757 			 * logic to deal with F_NOT and F_OR flags associated
1758 			 * with the opcode.
1759 			 */
1760 			case O_NOP:
1761 				match = 1;
1762 				break;
1763 
1764 			case O_FORWARD_MAC:
1765 				printf("ipfw: opcode %d unimplemented\n",
1766 				    cmd->opcode);
1767 				break;
1768 
1769 			case O_GID:
1770 			case O_UID:
1771 			case O_JAIL:
1772 				/*
1773 				 * We only check offset == 0 && proto != 0,
1774 				 * as this ensures that we have a
1775 				 * packet with the ports info.
1776 				 */
1777 				if (offset != 0)
1778 					break;
1779 				if (proto == IPPROTO_TCP ||
1780 				    proto == IPPROTO_UDP)
1781 					match = check_uidgid(
1782 						    (ipfw_insn_u32 *)cmd,
1783 						    args, &ucred_lookup,
1784 #ifdef __FreeBSD__
1785 						    &ucred_cache);
1786 #else
1787 						    (void *)&ucred_cache);
1788 #endif
1789 				break;
1790 
1791 			case O_RECV:
1792 				match = iface_match(m->m_pkthdr.rcvif,
1793 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1794 				break;
1795 
1796 			case O_XMIT:
1797 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1798 				    chain, &tablearg);
1799 				break;
1800 
1801 			case O_VIA:
1802 				match = iface_match(oif ? oif :
1803 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1804 				    chain, &tablearg);
1805 				break;
1806 
1807 			case O_MACADDR2:
1808 				if (args->eh != NULL) {	/* have MAC header */
1809 					u_int32_t *want = (u_int32_t *)
1810 						((ipfw_insn_mac *)cmd)->addr;
1811 					u_int32_t *mask = (u_int32_t *)
1812 						((ipfw_insn_mac *)cmd)->mask;
1813 					u_int32_t *hdr = (u_int32_t *)args->eh;
1814 
1815 					match =
1816 					    ( want[0] == (hdr[0] & mask[0]) &&
1817 					      want[1] == (hdr[1] & mask[1]) &&
1818 					      want[2] == (hdr[2] & mask[2]) );
1819 				}
1820 				break;
1821 
1822 			case O_MAC_TYPE:
1823 				if (args->eh != NULL) {
1824 					u_int16_t *p =
1825 					    ((ipfw_insn_u16 *)cmd)->ports;
1826 					int i;
1827 
1828 					for (i = cmdlen - 1; !match && i>0;
1829 					    i--, p += 2)
1830 						match = (etype >= p[0] &&
1831 						    etype <= p[1]);
1832 				}
1833 				break;
1834 
1835 			case O_FRAG:
1836 				match = (offset != 0);
1837 				break;
1838 
1839 			case O_IN:	/* "out" is "not in" */
1840 				match = (oif == NULL);
1841 				break;
1842 
1843 			case O_LAYER2:
1844 				match = (args->eh != NULL);
1845 				break;
1846 
1847 			case O_DIVERTED:
1848 			    {
1849 				/* For diverted packets, args->rule.info
1850 				 * contains the divert port (in host format)
1851 				 * reason and direction.
1852 				 */
1853 				uint32_t i = args->rule.info;
1854 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1855 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1856 			    }
1857 				break;
1858 
1859 			case O_PROTO:
1860 				/*
1861 				 * We do not allow an arg of 0 so the
1862 				 * check of "proto" only suffices.
1863 				 */
1864 				match = (proto == cmd->arg1);
1865 				break;
1866 
1867 			case O_IP_SRC:
1868 				match = is_ipv4 &&
1869 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1870 				    src_ip.s_addr);
1871 				break;
1872 
1873 			case O_IP_DST_LOOKUP:
1874 			{
1875 				void *pkey;
1876 				uint32_t vidx, key;
1877 				uint16_t keylen;
1878 
1879 				if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1880 					/* Determine lookup key type */
1881 					vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1882 					if (vidx != 4 /* uid */ &&
1883 					    vidx != 5 /* jail */ &&
1884 					    is_ipv6 == 0 && is_ipv4 == 0)
1885 						break;
1886 					/* Determine key length */
1887 					if (vidx == 0 /* dst-ip */ ||
1888 					    vidx == 1 /* src-ip */)
1889 						keylen = is_ipv6 ?
1890 						    sizeof(struct in6_addr):
1891 						    sizeof(in_addr_t);
1892 					else {
1893 						keylen = sizeof(key);
1894 						pkey = &key;
1895 					}
1896 					if (vidx == 0 /* dst-ip */)
1897 						pkey = is_ipv4 ? (void *)&dst_ip:
1898 						    (void *)&args->f_id.dst_ip6;
1899 					else if (vidx == 1 /* src-ip */)
1900 						pkey = is_ipv4 ? (void *)&src_ip:
1901 						    (void *)&args->f_id.src_ip6;
1902 					else if (vidx == 6 /* dscp */) {
1903 						if (is_ipv4)
1904 							key = ip->ip_tos >> 2;
1905 						else {
1906 							key = args->f_id.flow_id6;
1907 							key = (key & 0x0f) << 2 |
1908 							    (key & 0xf000) >> 14;
1909 						}
1910 						key &= 0x3f;
1911 					} else if (vidx == 2 /* dst-port */ ||
1912 					    vidx == 3 /* src-port */) {
1913 						/* Skip fragments */
1914 						if (offset != 0)
1915 							break;
1916 						/* Skip proto without ports */
1917 						if (proto != IPPROTO_TCP &&
1918 						    proto != IPPROTO_UDP &&
1919 						    proto != IPPROTO_SCTP)
1920 							break;
1921 						if (vidx == 2 /* dst-port */)
1922 							key = dst_port;
1923 						else
1924 							key = src_port;
1925 					}
1926 #ifndef USERSPACE
1927 					else if (vidx == 4 /* uid */ ||
1928 					    vidx == 5 /* jail */) {
1929 						check_uidgid(
1930 						    (ipfw_insn_u32 *)cmd,
1931 						    args, &ucred_lookup,
1932 #ifdef __FreeBSD__
1933 						    &ucred_cache);
1934 						if (vidx == 4 /* uid */)
1935 							key = ucred_cache->cr_uid;
1936 						else if (vidx == 5 /* jail */)
1937 							key = ucred_cache->cr_prison->pr_id;
1938 #else /* !__FreeBSD__ */
1939 						    (void *)&ucred_cache);
1940 						if (vidx == 4 /* uid */)
1941 							key = ucred_cache.uid;
1942 						else if (vidx == 5 /* jail */)
1943 							key = ucred_cache.xid;
1944 #endif /* !__FreeBSD__ */
1945 					}
1946 #endif /* !USERSPACE */
1947 					else
1948 						break;
1949 					match = ipfw_lookup_table(chain,
1950 					    cmd->arg1, keylen, pkey, &vidx);
1951 					if (!match)
1952 						break;
1953 					tablearg = vidx;
1954 					break;
1955 				}
1956 				/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1957 				/* FALLTHROUGH */
1958 			}
1959 			case O_IP_SRC_LOOKUP:
1960 			{
1961 				void *pkey;
1962 				uint32_t vidx;
1963 				uint16_t keylen;
1964 
1965 				if (is_ipv4) {
1966 					keylen = sizeof(in_addr_t);
1967 					if (cmd->opcode == O_IP_DST_LOOKUP)
1968 						pkey = &dst_ip;
1969 					else
1970 						pkey = &src_ip;
1971 				} else if (is_ipv6) {
1972 					keylen = sizeof(struct in6_addr);
1973 					if (cmd->opcode == O_IP_DST_LOOKUP)
1974 						pkey = &args->f_id.dst_ip6;
1975 					else
1976 						pkey = &args->f_id.src_ip6;
1977 				} else
1978 					break;
1979 				match = ipfw_lookup_table(chain, cmd->arg1,
1980 				    keylen, pkey, &vidx);
1981 				if (!match)
1982 					break;
1983 				if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
1984 					match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1985 					    TARG_VAL(chain, vidx, tag);
1986 					if (!match)
1987 						break;
1988 				}
1989 				tablearg = vidx;
1990 				break;
1991 			}
1992 
1993 			case O_IP_FLOW_LOOKUP:
1994 				{
1995 					uint32_t v = 0;
1996 					match = ipfw_lookup_table(chain,
1997 					    cmd->arg1, 0, &args->f_id, &v);
1998 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1999 						match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2000 						    TARG_VAL(chain, v, tag);
2001 					if (match)
2002 						tablearg = v;
2003 				}
2004 				break;
2005 			case O_IP_SRC_MASK:
2006 			case O_IP_DST_MASK:
2007 				if (is_ipv4) {
2008 				    uint32_t a =
2009 					(cmd->opcode == O_IP_DST_MASK) ?
2010 					    dst_ip.s_addr : src_ip.s_addr;
2011 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2012 				    int i = cmdlen-1;
2013 
2014 				    for (; !match && i>0; i-= 2, p+= 2)
2015 					match = (p[0] == (a & p[1]));
2016 				}
2017 				break;
2018 
2019 			case O_IP_SRC_ME:
2020 				if (is_ipv4) {
2021 					match = in_localip(src_ip);
2022 					break;
2023 				}
2024 #ifdef INET6
2025 				/* FALLTHROUGH */
2026 			case O_IP6_SRC_ME:
2027 				match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
2028 #endif
2029 				break;
2030 
2031 			case O_IP_DST_SET:
2032 			case O_IP_SRC_SET:
2033 				if (is_ipv4) {
2034 					u_int32_t *d = (u_int32_t *)(cmd+1);
2035 					u_int32_t addr =
2036 					    cmd->opcode == O_IP_DST_SET ?
2037 						args->f_id.dst_ip :
2038 						args->f_id.src_ip;
2039 
2040 					    if (addr < d[0])
2041 						    break;
2042 					    addr -= d[0]; /* subtract base */
2043 					    match = (addr < cmd->arg1) &&
2044 						( d[ 1 + (addr>>5)] &
2045 						  (1<<(addr & 0x1f)) );
2046 				}
2047 				break;
2048 
2049 			case O_IP_DST:
2050 				match = is_ipv4 &&
2051 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2052 				    dst_ip.s_addr);
2053 				break;
2054 
2055 			case O_IP_DST_ME:
2056 				if (is_ipv4) {
2057 					match = in_localip(dst_ip);
2058 					break;
2059 				}
2060 #ifdef INET6
2061 				/* FALLTHROUGH */
2062 			case O_IP6_DST_ME:
2063 				match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
2064 #endif
2065 				break;
2066 
2067 
2068 			case O_IP_SRCPORT:
2069 			case O_IP_DSTPORT:
2070 				/*
2071 				 * offset == 0 && proto != 0 is enough
2072 				 * to guarantee that we have a
2073 				 * packet with port info.
2074 				 */
2075 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP ||
2076 				    proto==IPPROTO_SCTP) && offset == 0) {
2077 					u_int16_t x =
2078 					    (cmd->opcode == O_IP_SRCPORT) ?
2079 						src_port : dst_port ;
2080 					u_int16_t *p =
2081 					    ((ipfw_insn_u16 *)cmd)->ports;
2082 					int i;
2083 
2084 					for (i = cmdlen - 1; !match && i>0;
2085 					    i--, p += 2)
2086 						match = (x>=p[0] && x<=p[1]);
2087 				}
2088 				break;
2089 
2090 			case O_ICMPTYPE:
2091 				match = (offset == 0 && proto==IPPROTO_ICMP &&
2092 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2093 				break;
2094 
2095 #ifdef INET6
2096 			case O_ICMP6TYPE:
2097 				match = is_ipv6 && offset == 0 &&
2098 				    proto==IPPROTO_ICMPV6 &&
2099 				    icmp6type_match(
2100 					ICMP6(ulp)->icmp6_type,
2101 					(ipfw_insn_u32 *)cmd);
2102 				break;
2103 #endif /* INET6 */
2104 
2105 			case O_IPOPT:
2106 				match = (is_ipv4 &&
2107 				    ipopts_match(ip, cmd) );
2108 				break;
2109 
2110 			case O_IPVER:
2111 				match = (is_ipv4 &&
2112 				    cmd->arg1 == ip->ip_v);
2113 				break;
2114 
2115 			case O_IPID:
2116 			case O_IPLEN:
2117 			case O_IPTTL:
2118 				if (is_ipv4) {	/* only for IP packets */
2119 				    uint16_t x;
2120 				    uint16_t *p;
2121 				    int i;
2122 
2123 				    if (cmd->opcode == O_IPLEN)
2124 					x = iplen;
2125 				    else if (cmd->opcode == O_IPTTL)
2126 					x = ip->ip_ttl;
2127 				    else /* must be IPID */
2128 					x = ntohs(ip->ip_id);
2129 				    if (cmdlen == 1) {
2130 					match = (cmd->arg1 == x);
2131 					break;
2132 				    }
2133 				    /* otherwise we have ranges */
2134 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2135 				    i = cmdlen - 1;
2136 				    for (; !match && i>0; i--, p += 2)
2137 					match = (x >= p[0] && x <= p[1]);
2138 				}
2139 				break;
2140 
2141 			case O_IPPRECEDENCE:
2142 				match = (is_ipv4 &&
2143 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2144 				break;
2145 
2146 			case O_IPTOS:
2147 				match = (is_ipv4 &&
2148 				    flags_match(cmd, ip->ip_tos));
2149 				break;
2150 
2151 			case O_DSCP:
2152 			    {
2153 				uint32_t *p;
2154 				uint16_t x;
2155 
2156 				p = ((ipfw_insn_u32 *)cmd)->d;
2157 
2158 				if (is_ipv4)
2159 					x = ip->ip_tos >> 2;
2160 				else if (is_ipv6) {
2161 					uint8_t *v;
2162 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2163 					x = (*v & 0x0F) << 2;
2164 					v++;
2165 					x |= *v >> 6;
2166 				} else
2167 					break;
2168 
2169 				/* DSCP bitmask is stored as low_u32 high_u32 */
2170 				if (x >= 32)
2171 					match = *(p + 1) & (1 << (x - 32));
2172 				else
2173 					match = *p & (1 << x);
2174 			    }
2175 				break;
2176 
2177 			case O_TCPDATALEN:
2178 				if (proto == IPPROTO_TCP && offset == 0) {
2179 				    struct tcphdr *tcp;
2180 				    uint16_t x;
2181 				    uint16_t *p;
2182 				    int i;
2183 #ifdef INET6
2184 				    if (is_ipv6) {
2185 					    struct ip6_hdr *ip6;
2186 
2187 					    ip6 = (struct ip6_hdr *)ip;
2188 					    if (ip6->ip6_plen == 0) {
2189 						    /*
2190 						     * Jumbo payload is not
2191 						     * supported by this
2192 						     * opcode.
2193 						     */
2194 						    break;
2195 					    }
2196 					    x = iplen - hlen;
2197 				    } else
2198 #endif /* INET6 */
2199 					    x = iplen - (ip->ip_hl << 2);
2200 				    tcp = TCP(ulp);
2201 				    x -= tcp->th_off << 2;
2202 				    if (cmdlen == 1) {
2203 					match = (cmd->arg1 == x);
2204 					break;
2205 				    }
2206 				    /* otherwise we have ranges */
2207 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2208 				    i = cmdlen - 1;
2209 				    for (; !match && i>0; i--, p += 2)
2210 					match = (x >= p[0] && x <= p[1]);
2211 				}
2212 				break;
2213 
2214 			case O_TCPFLAGS:
2215 				match = (proto == IPPROTO_TCP && offset == 0 &&
2216 				    flags_match(cmd, TCP(ulp)->th_flags));
2217 				break;
2218 
2219 			case O_TCPOPTS:
2220 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
2221 					PULLUP_LEN(hlen, ulp,
2222 					    (TCP(ulp)->th_off << 2));
2223 					match = tcpopts_match(TCP(ulp), cmd);
2224 				}
2225 				break;
2226 
2227 			case O_TCPSEQ:
2228 				match = (proto == IPPROTO_TCP && offset == 0 &&
2229 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2230 					TCP(ulp)->th_seq);
2231 				break;
2232 
2233 			case O_TCPACK:
2234 				match = (proto == IPPROTO_TCP && offset == 0 &&
2235 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2236 					TCP(ulp)->th_ack);
2237 				break;
2238 
2239 			case O_TCPWIN:
2240 				if (proto == IPPROTO_TCP && offset == 0) {
2241 				    uint16_t x;
2242 				    uint16_t *p;
2243 				    int i;
2244 
2245 				    x = ntohs(TCP(ulp)->th_win);
2246 				    if (cmdlen == 1) {
2247 					match = (cmd->arg1 == x);
2248 					break;
2249 				    }
2250 				    /* Otherwise we have ranges. */
2251 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2252 				    i = cmdlen - 1;
2253 				    for (; !match && i > 0; i--, p += 2)
2254 					match = (x >= p[0] && x <= p[1]);
2255 				}
2256 				break;
2257 
2258 			case O_ESTAB:
2259 				/* reject packets which have SYN only */
2260 				/* XXX should i also check for TH_ACK ? */
2261 				match = (proto == IPPROTO_TCP && offset == 0 &&
2262 				    (TCP(ulp)->th_flags &
2263 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2264 				break;
2265 
2266 			case O_ALTQ: {
2267 				struct pf_mtag *at;
2268 				struct m_tag *mtag;
2269 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2270 
2271 				/*
2272 				 * ALTQ uses mbuf tags from another
2273 				 * packet filtering system - pf(4).
2274 				 * We allocate a tag in its format
2275 				 * and fill it in, pretending to be pf(4).
2276 				 */
2277 				match = 1;
2278 				at = pf_find_mtag(m);
2279 				if (at != NULL && at->qid != 0)
2280 					break;
2281 				mtag = m_tag_get(PACKET_TAG_PF,
2282 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2283 				if (mtag == NULL) {
2284 					/*
2285 					 * Let the packet fall back to the
2286 					 * default ALTQ.
2287 					 */
2288 					break;
2289 				}
2290 				m_tag_prepend(m, mtag);
2291 				at = (struct pf_mtag *)(mtag + 1);
2292 				at->qid = altq->qid;
2293 				at->hdr = ip;
2294 				break;
2295 			}
2296 
2297 			case O_LOG:
2298 				ipfw_log(chain, f, hlen, args, m,
2299 				    oif, offset | ip6f_mf, tablearg, ip);
2300 				match = 1;
2301 				break;
2302 
2303 			case O_PROB:
2304 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2305 				break;
2306 
2307 			case O_VERREVPATH:
2308 				/* Outgoing packets automatically pass/match */
2309 				match = ((oif != NULL) ||
2310 				    (m->m_pkthdr.rcvif == NULL) ||
2311 				    (
2312 #ifdef INET6
2313 				    is_ipv6 ?
2314 					verify_path6(&(args->f_id.src_ip6),
2315 					    m->m_pkthdr.rcvif, args->f_id.fib) :
2316 #endif
2317 				    verify_path(src_ip, m->m_pkthdr.rcvif,
2318 				        args->f_id.fib)));
2319 				break;
2320 
2321 			case O_VERSRCREACH:
2322 				/* Outgoing packets automatically pass/match */
2323 				match = (hlen > 0 && ((oif != NULL) ||
2324 #ifdef INET6
2325 				    is_ipv6 ?
2326 				        verify_path6(&(args->f_id.src_ip6),
2327 				            NULL, args->f_id.fib) :
2328 #endif
2329 				    verify_path(src_ip, NULL, args->f_id.fib)));
2330 				break;
2331 
2332 			case O_ANTISPOOF:
2333 				/* Outgoing packets automatically pass/match */
2334 				if (oif == NULL && hlen > 0 &&
2335 				    (  (is_ipv4 && in_localaddr(src_ip))
2336 #ifdef INET6
2337 				    || (is_ipv6 &&
2338 				        in6_localaddr(&(args->f_id.src_ip6)))
2339 #endif
2340 				    ))
2341 					match =
2342 #ifdef INET6
2343 					    is_ipv6 ? verify_path6(
2344 					        &(args->f_id.src_ip6),
2345 					        m->m_pkthdr.rcvif,
2346 						args->f_id.fib) :
2347 #endif
2348 					    verify_path(src_ip,
2349 					    	m->m_pkthdr.rcvif,
2350 					        args->f_id.fib);
2351 				else
2352 					match = 1;
2353 				break;
2354 
2355 			case O_IPSEC:
2356 				match = (m_tag_find(m,
2357 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2358 				/* otherwise no match */
2359 				break;
2360 
2361 #ifdef INET6
2362 			case O_IP6_SRC:
2363 				match = is_ipv6 &&
2364 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2365 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2366 				break;
2367 
2368 			case O_IP6_DST:
2369 				match = is_ipv6 &&
2370 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2371 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2372 				break;
2373 			case O_IP6_SRC_MASK:
2374 			case O_IP6_DST_MASK:
2375 				if (is_ipv6) {
2376 					int i = cmdlen - 1;
2377 					struct in6_addr p;
2378 					struct in6_addr *d =
2379 					    &((ipfw_insn_ip6 *)cmd)->addr6;
2380 
2381 					for (; !match && i > 0; d += 2,
2382 					    i -= F_INSN_SIZE(struct in6_addr)
2383 					    * 2) {
2384 						p = (cmd->opcode ==
2385 						    O_IP6_SRC_MASK) ?
2386 						    args->f_id.src_ip6:
2387 						    args->f_id.dst_ip6;
2388 						APPLY_MASK(&p, &d[1]);
2389 						match =
2390 						    IN6_ARE_ADDR_EQUAL(&d[0],
2391 						    &p);
2392 					}
2393 				}
2394 				break;
2395 
2396 			case O_FLOW6ID:
2397 				match = is_ipv6 &&
2398 				    flow6id_match(args->f_id.flow_id6,
2399 				    (ipfw_insn_u32 *) cmd);
2400 				break;
2401 
2402 			case O_EXT_HDR:
2403 				match = is_ipv6 &&
2404 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
2405 				break;
2406 
2407 			case O_IP6:
2408 				match = is_ipv6;
2409 				break;
2410 #endif
2411 
2412 			case O_IP4:
2413 				match = is_ipv4;
2414 				break;
2415 
2416 			case O_TAG: {
2417 				struct m_tag *mtag;
2418 				uint32_t tag = TARG(cmd->arg1, tag);
2419 
2420 				/* Packet is already tagged with this tag? */
2421 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2422 
2423 				/* We have `untag' action when F_NOT flag is
2424 				 * present. And we must remove this mtag from
2425 				 * mbuf and reset `match' to zero (`match' will
2426 				 * be inversed later).
2427 				 * Otherwise we should allocate new mtag and
2428 				 * push it into mbuf.
2429 				 */
2430 				if (cmd->len & F_NOT) { /* `untag' action */
2431 					if (mtag != NULL)
2432 						m_tag_delete(m, mtag);
2433 					match = 0;
2434 				} else {
2435 					if (mtag == NULL) {
2436 						mtag = m_tag_alloc( MTAG_IPFW,
2437 						    tag, 0, M_NOWAIT);
2438 						if (mtag != NULL)
2439 							m_tag_prepend(m, mtag);
2440 					}
2441 					match = 1;
2442 				}
2443 				break;
2444 			}
2445 
2446 			case O_FIB: /* try match the specified fib */
2447 				if (args->f_id.fib == cmd->arg1)
2448 					match = 1;
2449 				break;
2450 
2451 			case O_SOCKARG:	{
2452 #ifndef USERSPACE	/* not supported in userspace */
2453 				struct inpcb *inp = args->inp;
2454 				struct inpcbinfo *pi;
2455 
2456 				if (is_ipv6) /* XXX can we remove this ? */
2457 					break;
2458 
2459 				if (proto == IPPROTO_TCP)
2460 					pi = &V_tcbinfo;
2461 				else if (proto == IPPROTO_UDP)
2462 					pi = &V_udbinfo;
2463 				else
2464 					break;
2465 
2466 				/*
2467 				 * XXXRW: so_user_cookie should almost
2468 				 * certainly be inp_user_cookie?
2469 				 */
2470 
2471 				/* For incoming packet, lookup up the
2472 				inpcb using the src/dest ip/port tuple */
2473 				if (inp == NULL) {
2474 					inp = in_pcblookup(pi,
2475 						src_ip, htons(src_port),
2476 						dst_ip, htons(dst_port),
2477 						INPLOOKUP_RLOCKPCB, NULL);
2478 					if (inp != NULL) {
2479 						tablearg =
2480 						    inp->inp_socket->so_user_cookie;
2481 						if (tablearg)
2482 							match = 1;
2483 						INP_RUNLOCK(inp);
2484 					}
2485 				} else {
2486 					if (inp->inp_socket) {
2487 						tablearg =
2488 						    inp->inp_socket->so_user_cookie;
2489 						if (tablearg)
2490 							match = 1;
2491 					}
2492 				}
2493 #endif /* !USERSPACE */
2494 				break;
2495 			}
2496 
2497 			case O_TAGGED: {
2498 				struct m_tag *mtag;
2499 				uint32_t tag = TARG(cmd->arg1, tag);
2500 
2501 				if (cmdlen == 1) {
2502 					match = m_tag_locate(m, MTAG_IPFW,
2503 					    tag, NULL) != NULL;
2504 					break;
2505 				}
2506 
2507 				/* we have ranges */
2508 				for (mtag = m_tag_first(m);
2509 				    mtag != NULL && !match;
2510 				    mtag = m_tag_next(m, mtag)) {
2511 					uint16_t *p;
2512 					int i;
2513 
2514 					if (mtag->m_tag_cookie != MTAG_IPFW)
2515 						continue;
2516 
2517 					p = ((ipfw_insn_u16 *)cmd)->ports;
2518 					i = cmdlen - 1;
2519 					for(; !match && i > 0; i--, p += 2)
2520 						match =
2521 						    mtag->m_tag_id >= p[0] &&
2522 						    mtag->m_tag_id <= p[1];
2523 				}
2524 				break;
2525 			}
2526 
2527 			/*
2528 			 * The second set of opcodes represents 'actions',
2529 			 * i.e. the terminal part of a rule once the packet
2530 			 * matches all previous patterns.
2531 			 * Typically there is only one action for each rule,
2532 			 * and the opcode is stored at the end of the rule
2533 			 * (but there are exceptions -- see below).
2534 			 *
2535 			 * In general, here we set retval and terminate the
2536 			 * outer loop (would be a 'break 3' in some language,
2537 			 * but we need to set l=0, done=1)
2538 			 *
2539 			 * Exceptions:
2540 			 * O_COUNT and O_SKIPTO actions:
2541 			 *   instead of terminating, we jump to the next rule
2542 			 *   (setting l=0), or to the SKIPTO target (setting
2543 			 *   f/f_len, cmd and l as needed), respectively.
2544 			 *
2545 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2546 			 *   perform some action and set match = 1;
2547 			 *
2548 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2549 			 *   not real 'actions', and are stored right
2550 			 *   before the 'action' part of the rule.
2551 			 *   These opcodes try to install an entry in the
2552 			 *   state tables; if successful, we continue with
2553 			 *   the next opcode (match=1; break;), otherwise
2554 			 *   the packet must be dropped (set retval,
2555 			 *   break loops with l=0, done=1)
2556 			 *
2557 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2558 			 *   cause a lookup of the state table, and a jump
2559 			 *   to the 'action' part of the parent rule
2560 			 *   if an entry is found, or
2561 			 *   (CHECK_STATE only) a jump to the next rule if
2562 			 *   the entry is not found.
2563 			 *   The result of the lookup is cached so that
2564 			 *   further instances of these opcodes become NOPs.
2565 			 *   The jump to the next rule is done by setting
2566 			 *   l=0, cmdlen=0.
2567 			 */
2568 			case O_LIMIT:
2569 			case O_KEEP_STATE:
2570 				if (ipfw_dyn_install_state(chain, f,
2571 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2572 					/* error or limit violation */
2573 					retval = IP_FW_DENY;
2574 					l = 0;	/* exit inner loop */
2575 					done = 1; /* exit outer loop */
2576 				}
2577 				match = 1;
2578 				break;
2579 
2580 			case O_PROBE_STATE:
2581 			case O_CHECK_STATE:
2582 				/*
2583 				 * dynamic rules are checked at the first
2584 				 * keep-state or check-state occurrence,
2585 				 * with the result being stored in dyn_dir
2586 				 * and dyn_name.
2587 				 * The compiler introduces a PROBE_STATE
2588 				 * instruction for us when we have a
2589 				 * KEEP_STATE (because PROBE_STATE needs
2590 				 * to be run first).
2591 				 *
2592 				 * (dyn_dir == MATCH_UNKNOWN) means this is
2593 				 * first lookup for such f_id. Do lookup.
2594 				 *
2595 				 * (dyn_dir != MATCH_UNKNOWN &&
2596 				 *  dyn_name != 0 && dyn_name != cmd->arg1)
2597 				 * means previous lookup didn't find dynamic
2598 				 * rule for specific state name and current
2599 				 * lookup will search rule with another state
2600 				 * name. Redo lookup.
2601 				 *
2602 				 * (dyn_dir != MATCH_UNKNOWN && dyn_name == 0)
2603 				 * means previous lookup was for `any' name
2604 				 * and it didn't find rule. No need to do
2605 				 * lookup again.
2606 				 */
2607 				if ((dyn_dir == MATCH_UNKNOWN ||
2608 				    (dyn_name != 0 &&
2609 				    dyn_name != cmd->arg1)) &&
2610 				    (q = ipfw_dyn_lookup_state(&args->f_id,
2611 				     ulp, pktlen, &dyn_dir,
2612 				     (dyn_name = cmd->arg1))) != NULL) {
2613 					/*
2614 					 * Found dynamic entry, jump to the
2615 					 * 'action' part of the parent rule
2616 					 * by setting f, cmd, l and clearing
2617 					 * cmdlen.
2618 					 */
2619 					f = q;
2620 					/* XXX we would like to have f_pos
2621 					 * readily accessible in the dynamic
2622 				         * rule, instead of having to
2623 					 * lookup q->rule.
2624 					 */
2625 					f_pos = ipfw_find_rule(chain,
2626 					    f->rulenum, f->id);
2627 					cmd = ACTION_PTR(f);
2628 					l = f->cmd_len - f->act_ofs;
2629 					cmdlen = 0;
2630 					match = 1;
2631 					break;
2632 				}
2633 				/*
2634 				 * Dynamic entry not found. If CHECK_STATE,
2635 				 * skip to next rule, if PROBE_STATE just
2636 				 * ignore and continue with next opcode.
2637 				 */
2638 				if (cmd->opcode == O_CHECK_STATE)
2639 					l = 0;	/* exit inner loop */
2640 				match = 1;
2641 				break;
2642 
2643 			case O_ACCEPT:
2644 				retval = 0;	/* accept */
2645 				l = 0;		/* exit inner loop */
2646 				done = 1;	/* exit outer loop */
2647 				break;
2648 
2649 			case O_PIPE:
2650 			case O_QUEUE:
2651 				set_match(args, f_pos, chain);
2652 				args->rule.info = TARG(cmd->arg1, pipe);
2653 				if (cmd->opcode == O_PIPE)
2654 					args->rule.info |= IPFW_IS_PIPE;
2655 				if (V_fw_one_pass)
2656 					args->rule.info |= IPFW_ONEPASS;
2657 				retval = IP_FW_DUMMYNET;
2658 				l = 0;          /* exit inner loop */
2659 				done = 1;       /* exit outer loop */
2660 				break;
2661 
2662 			case O_DIVERT:
2663 			case O_TEE:
2664 				if (args->eh) /* not on layer 2 */
2665 				    break;
2666 				/* otherwise this is terminal */
2667 				l = 0;		/* exit inner loop */
2668 				done = 1;	/* exit outer loop */
2669 				retval = (cmd->opcode == O_DIVERT) ?
2670 					IP_FW_DIVERT : IP_FW_TEE;
2671 				set_match(args, f_pos, chain);
2672 				args->rule.info = TARG(cmd->arg1, divert);
2673 				break;
2674 
2675 			case O_COUNT:
2676 				IPFW_INC_RULE_COUNTER(f, pktlen);
2677 				l = 0;		/* exit inner loop */
2678 				break;
2679 
2680 			case O_SKIPTO:
2681 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2682 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2683 			    /*
2684 			     * Skip disabled rules, and re-enter
2685 			     * the inner loop with the correct
2686 			     * f_pos, f, l and cmd.
2687 			     * Also clear cmdlen and skip_or
2688 			     */
2689 			    for (; f_pos < chain->n_rules - 1 &&
2690 				    (V_set_disable &
2691 				     (1 << chain->map[f_pos]->set));
2692 				    f_pos++)
2693 				;
2694 			    /* Re-enter the inner loop at the skipto rule. */
2695 			    f = chain->map[f_pos];
2696 			    l = f->cmd_len;
2697 			    cmd = f->cmd;
2698 			    match = 1;
2699 			    cmdlen = 0;
2700 			    skip_or = 0;
2701 			    continue;
2702 			    break;	/* not reached */
2703 
2704 			case O_CALLRETURN: {
2705 				/*
2706 				 * Implementation of `subroutine' call/return,
2707 				 * in the stack carried in an mbuf tag. This
2708 				 * is different from `skipto' in that any call
2709 				 * address is possible (`skipto' must prevent
2710 				 * backward jumps to avoid endless loops).
2711 				 * We have `return' action when F_NOT flag is
2712 				 * present. The `m_tag_id' field is used as
2713 				 * stack pointer.
2714 				 */
2715 				struct m_tag *mtag;
2716 				uint16_t jmpto, *stack;
2717 
2718 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2719 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2720 				/*
2721 				 * Hand-rolled version of m_tag_locate() with
2722 				 * wildcard `type'.
2723 				 * If not already tagged, allocate new tag.
2724 				 */
2725 				mtag = m_tag_first(m);
2726 				while (mtag != NULL) {
2727 					if (mtag->m_tag_cookie ==
2728 					    MTAG_IPFW_CALL)
2729 						break;
2730 					mtag = m_tag_next(m, mtag);
2731 				}
2732 				if (mtag == NULL && IS_CALL) {
2733 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2734 					    IPFW_CALLSTACK_SIZE *
2735 					    sizeof(uint16_t), M_NOWAIT);
2736 					if (mtag != NULL)
2737 						m_tag_prepend(m, mtag);
2738 				}
2739 
2740 				/*
2741 				 * On error both `call' and `return' just
2742 				 * continue with next rule.
2743 				 */
2744 				if (IS_RETURN && (mtag == NULL ||
2745 				    mtag->m_tag_id == 0)) {
2746 					l = 0;		/* exit inner loop */
2747 					break;
2748 				}
2749 				if (IS_CALL && (mtag == NULL ||
2750 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2751 					printf("ipfw: call stack error, "
2752 					    "go to next rule\n");
2753 					l = 0;		/* exit inner loop */
2754 					break;
2755 				}
2756 
2757 				IPFW_INC_RULE_COUNTER(f, pktlen);
2758 				stack = (uint16_t *)(mtag + 1);
2759 
2760 				/*
2761 				 * The `call' action may use cached f_pos
2762 				 * (in f->next_rule), whose version is written
2763 				 * in f->next_rule.
2764 				 * The `return' action, however, doesn't have
2765 				 * fixed jump address in cmd->arg1 and can't use
2766 				 * cache.
2767 				 */
2768 				if (IS_CALL) {
2769 					stack[mtag->m_tag_id] = f->rulenum;
2770 					mtag->m_tag_id++;
2771 			    		f_pos = JUMP(chain, f, cmd->arg1,
2772 					    tablearg, 1);
2773 				} else {	/* `return' action */
2774 					mtag->m_tag_id--;
2775 					jmpto = stack[mtag->m_tag_id] + 1;
2776 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2777 				}
2778 
2779 				/*
2780 				 * Skip disabled rules, and re-enter
2781 				 * the inner loop with the correct
2782 				 * f_pos, f, l and cmd.
2783 				 * Also clear cmdlen and skip_or
2784 				 */
2785 				for (; f_pos < chain->n_rules - 1 &&
2786 				    (V_set_disable &
2787 				    (1 << chain->map[f_pos]->set)); f_pos++)
2788 					;
2789 				/* Re-enter the inner loop at the dest rule. */
2790 				f = chain->map[f_pos];
2791 				l = f->cmd_len;
2792 				cmd = f->cmd;
2793 				cmdlen = 0;
2794 				skip_or = 0;
2795 				continue;
2796 				break;	/* NOTREACHED */
2797 			}
2798 #undef IS_CALL
2799 #undef IS_RETURN
2800 
2801 			case O_REJECT:
2802 				/*
2803 				 * Drop the packet and send a reject notice
2804 				 * if the packet is not ICMP (or is an ICMP
2805 				 * query), and it is not multicast/broadcast.
2806 				 */
2807 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2808 				    (proto != IPPROTO_ICMP ||
2809 				     is_icmp_query(ICMP(ulp))) &&
2810 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2811 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2812 					send_reject(args, cmd->arg1, iplen, ip);
2813 					m = args->m;
2814 				}
2815 				/* FALLTHROUGH */
2816 #ifdef INET6
2817 			case O_UNREACH6:
2818 				if (hlen > 0 && is_ipv6 &&
2819 				    ((offset & IP6F_OFF_MASK) == 0) &&
2820 				    (proto != IPPROTO_ICMPV6 ||
2821 				     (is_icmp6_query(icmp6_type) == 1)) &&
2822 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2823 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2824 					send_reject6(
2825 					    args, cmd->arg1, hlen,
2826 					    (struct ip6_hdr *)ip);
2827 					m = args->m;
2828 				}
2829 				/* FALLTHROUGH */
2830 #endif
2831 			case O_DENY:
2832 				retval = IP_FW_DENY;
2833 				l = 0;		/* exit inner loop */
2834 				done = 1;	/* exit outer loop */
2835 				break;
2836 
2837 			case O_FORWARD_IP:
2838 				if (args->eh)	/* not valid on layer2 pkts */
2839 					break;
2840 				if (q != f || dyn_dir == MATCH_FORWARD) {
2841 				    struct sockaddr_in *sa;
2842 
2843 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2844 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2845 #ifdef INET6
2846 					/*
2847 					 * We use O_FORWARD_IP opcode for
2848 					 * fwd rule with tablearg, but tables
2849 					 * now support IPv6 addresses. And
2850 					 * when we are inspecting IPv6 packet,
2851 					 * we can use nh6 field from
2852 					 * table_value as next_hop6 address.
2853 					 */
2854 					if (is_ipv6) {
2855 						struct sockaddr_in6 *sa6;
2856 
2857 						sa6 = args->next_hop6 =
2858 						    &args->hopstore6;
2859 						sa6->sin6_family = AF_INET6;
2860 						sa6->sin6_len = sizeof(*sa6);
2861 						sa6->sin6_addr = TARG_VAL(
2862 						    chain, tablearg, nh6);
2863 						sa6->sin6_port = sa->sin_port;
2864 						/*
2865 						 * Set sin6_scope_id only for
2866 						 * link-local unicast addresses.
2867 						 */
2868 						if (IN6_IS_ADDR_LINKLOCAL(
2869 						    &sa6->sin6_addr))
2870 							sa6->sin6_scope_id =
2871 							    TARG_VAL(chain,
2872 								tablearg,
2873 								zoneid);
2874 					} else
2875 #endif
2876 					{
2877 						args->hopstore.sin_port =
2878 						    sa->sin_port;
2879 						sa = args->next_hop =
2880 						    &args->hopstore;
2881 						sa->sin_family = AF_INET;
2882 						sa->sin_len = sizeof(*sa);
2883 						sa->sin_addr.s_addr = htonl(
2884 						    TARG_VAL(chain, tablearg,
2885 						    nh4));
2886 					}
2887 				    } else {
2888 					args->next_hop = sa;
2889 				    }
2890 				}
2891 				retval = IP_FW_PASS;
2892 				l = 0;          /* exit inner loop */
2893 				done = 1;       /* exit outer loop */
2894 				break;
2895 
2896 #ifdef INET6
2897 			case O_FORWARD_IP6:
2898 				if (args->eh)	/* not valid on layer2 pkts */
2899 					break;
2900 				if (q != f || dyn_dir == MATCH_FORWARD) {
2901 					struct sockaddr_in6 *sin6;
2902 
2903 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2904 					args->next_hop6 = sin6;
2905 				}
2906 				retval = IP_FW_PASS;
2907 				l = 0;		/* exit inner loop */
2908 				done = 1;	/* exit outer loop */
2909 				break;
2910 #endif
2911 
2912 			case O_NETGRAPH:
2913 			case O_NGTEE:
2914 				set_match(args, f_pos, chain);
2915 				args->rule.info = TARG(cmd->arg1, netgraph);
2916 				if (V_fw_one_pass)
2917 					args->rule.info |= IPFW_ONEPASS;
2918 				retval = (cmd->opcode == O_NETGRAPH) ?
2919 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2920 				l = 0;          /* exit inner loop */
2921 				done = 1;       /* exit outer loop */
2922 				break;
2923 
2924 			case O_SETFIB: {
2925 				uint32_t fib;
2926 
2927 				IPFW_INC_RULE_COUNTER(f, pktlen);
2928 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
2929 				if (fib >= rt_numfibs)
2930 					fib = 0;
2931 				M_SETFIB(m, fib);
2932 				args->f_id.fib = fib;
2933 				l = 0;		/* exit inner loop */
2934 				break;
2935 		        }
2936 
2937 			case O_SETDSCP: {
2938 				uint16_t code;
2939 
2940 				code = TARG(cmd->arg1, dscp) & 0x3F;
2941 				l = 0;		/* exit inner loop */
2942 				if (is_ipv4) {
2943 					uint16_t old;
2944 
2945 					old = *(uint16_t *)ip;
2946 					ip->ip_tos = (code << 2) |
2947 					    (ip->ip_tos & 0x03);
2948 					ip->ip_sum = cksum_adjust(ip->ip_sum,
2949 					    old, *(uint16_t *)ip);
2950 				} else if (is_ipv6) {
2951 					uint8_t *v;
2952 
2953 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2954 					*v = (*v & 0xF0) | (code >> 2);
2955 					v++;
2956 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
2957 				} else
2958 					break;
2959 
2960 				IPFW_INC_RULE_COUNTER(f, pktlen);
2961 				break;
2962 			}
2963 
2964 			case O_NAT:
2965 				l = 0;          /* exit inner loop */
2966 				done = 1;       /* exit outer loop */
2967 				/*
2968 				 * Ensure that we do not invoke NAT handler for
2969 				 * non IPv4 packets. Libalias expects only IPv4.
2970 				 */
2971 				if (!is_ipv4 || !IPFW_NAT_LOADED) {
2972 				    retval = IP_FW_DENY;
2973 				    break;
2974 				}
2975 
2976 				struct cfg_nat *t;
2977 				int nat_id;
2978 
2979 				set_match(args, f_pos, chain);
2980 				/* Check if this is 'global' nat rule */
2981 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
2982 					retval = ipfw_nat_ptr(args, NULL, m);
2983 					break;
2984 				}
2985 				t = ((ipfw_insn_nat *)cmd)->nat;
2986 				if (t == NULL) {
2987 					nat_id = TARG(cmd->arg1, nat);
2988 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2989 
2990 					if (t == NULL) {
2991 					    retval = IP_FW_DENY;
2992 					    break;
2993 					}
2994 					if (cmd->arg1 != IP_FW_TARG)
2995 					    ((ipfw_insn_nat *)cmd)->nat = t;
2996 				}
2997 				retval = ipfw_nat_ptr(args, t, m);
2998 				break;
2999 
3000 			case O_REASS: {
3001 				int ip_off;
3002 
3003 				IPFW_INC_RULE_COUNTER(f, pktlen);
3004 				l = 0;	/* in any case exit inner loop */
3005 				ip_off = ntohs(ip->ip_off);
3006 
3007 				/* if not fragmented, go to next rule */
3008 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3009 				    break;
3010 
3011 				args->m = m = ip_reass(m);
3012 
3013 				/*
3014 				 * do IP header checksum fixup.
3015 				 */
3016 				if (m == NULL) { /* fragment got swallowed */
3017 				    retval = IP_FW_DENY;
3018 				} else { /* good, packet complete */
3019 				    int hlen;
3020 
3021 				    ip = mtod(m, struct ip *);
3022 				    hlen = ip->ip_hl << 2;
3023 				    ip->ip_sum = 0;
3024 				    if (hlen == sizeof(struct ip))
3025 					ip->ip_sum = in_cksum_hdr(ip);
3026 				    else
3027 					ip->ip_sum = in_cksum(m, hlen);
3028 				    retval = IP_FW_REASS;
3029 				    set_match(args, f_pos, chain);
3030 				}
3031 				done = 1;	/* exit outer loop */
3032 				break;
3033 			}
3034 			case O_EXTERNAL_ACTION:
3035 				l = 0; /* in any case exit inner loop */
3036 				retval = ipfw_run_eaction(chain, args,
3037 				    cmd, &done);
3038 				/*
3039 				 * If both @retval and @done are zero,
3040 				 * consider this as rule matching and
3041 				 * update counters.
3042 				 */
3043 				if (retval == 0 && done == 0) {
3044 					IPFW_INC_RULE_COUNTER(f, pktlen);
3045 					/*
3046 					 * Reset the result of the last
3047 					 * dynamic state lookup.
3048 					 * External action can change
3049 					 * @args content, and it may be
3050 					 * used for new state lookup later.
3051 					 */
3052 					dyn_dir = MATCH_UNKNOWN;
3053 				}
3054 				break;
3055 
3056 			default:
3057 				panic("-- unknown opcode %d\n", cmd->opcode);
3058 			} /* end of switch() on opcodes */
3059 			/*
3060 			 * if we get here with l=0, then match is irrelevant.
3061 			 */
3062 
3063 			if (cmd->len & F_NOT)
3064 				match = !match;
3065 
3066 			if (match) {
3067 				if (cmd->len & F_OR)
3068 					skip_or = 1;
3069 			} else {
3070 				if (!(cmd->len & F_OR)) /* not an OR block, */
3071 					break;		/* try next rule    */
3072 			}
3073 
3074 		}	/* end of inner loop, scan opcodes */
3075 #undef PULLUP_LEN
3076 
3077 		if (done)
3078 			break;
3079 
3080 /* next_rule:; */	/* try next rule		*/
3081 
3082 	}		/* end of outer for, scan rules */
3083 
3084 	if (done) {
3085 		struct ip_fw *rule = chain->map[f_pos];
3086 		/* Update statistics */
3087 		IPFW_INC_RULE_COUNTER(rule, pktlen);
3088 	} else {
3089 		retval = IP_FW_DENY;
3090 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3091 	}
3092 	IPFW_PF_RUNLOCK(chain);
3093 #ifdef __FreeBSD__
3094 	if (ucred_cache != NULL)
3095 		crfree(ucred_cache);
3096 #endif
3097 	return (retval);
3098 
3099 pullup_failed:
3100 	if (V_fw_verbose)
3101 		printf("ipfw: pullup failed\n");
3102 	return (IP_FW_DENY);
3103 }
3104 
3105 /*
3106  * Set maximum number of tables that can be used in given VNET ipfw instance.
3107  */
3108 #ifdef SYSCTL_NODE
3109 static int
3110 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3111 {
3112 	int error;
3113 	unsigned int ntables;
3114 
3115 	ntables = V_fw_tables_max;
3116 
3117 	error = sysctl_handle_int(oidp, &ntables, 0, req);
3118 	/* Read operation or some error */
3119 	if ((error != 0) || (req->newptr == NULL))
3120 		return (error);
3121 
3122 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
3123 }
3124 
3125 /*
3126  * Switches table namespace between global and per-set.
3127  */
3128 static int
3129 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3130 {
3131 	int error;
3132 	unsigned int sets;
3133 
3134 	sets = V_fw_tables_sets;
3135 
3136 	error = sysctl_handle_int(oidp, &sets, 0, req);
3137 	/* Read operation or some error */
3138 	if ((error != 0) || (req->newptr == NULL))
3139 		return (error);
3140 
3141 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3142 }
3143 #endif
3144 
3145 /*
3146  * Module and VNET glue
3147  */
3148 
3149 /*
3150  * Stuff that must be initialised only on boot or module load
3151  */
3152 static int
3153 ipfw_init(void)
3154 {
3155 	int error = 0;
3156 
3157 	/*
3158  	 * Only print out this stuff the first time around,
3159 	 * when called from the sysinit code.
3160 	 */
3161 	printf("ipfw2 "
3162 #ifdef INET6
3163 		"(+ipv6) "
3164 #endif
3165 		"initialized, divert %s, nat %s, "
3166 		"default to %s, logging ",
3167 #ifdef IPDIVERT
3168 		"enabled",
3169 #else
3170 		"loadable",
3171 #endif
3172 #ifdef IPFIREWALL_NAT
3173 		"enabled",
3174 #else
3175 		"loadable",
3176 #endif
3177 		default_to_accept ? "accept" : "deny");
3178 
3179 	/*
3180 	 * Note: V_xxx variables can be accessed here but the vnet specific
3181 	 * initializer may not have been called yet for the VIMAGE case.
3182 	 * Tuneables will have been processed. We will print out values for
3183 	 * the default vnet.
3184 	 * XXX This should all be rationalized AFTER 8.0
3185 	 */
3186 	if (V_fw_verbose == 0)
3187 		printf("disabled\n");
3188 	else if (V_verbose_limit == 0)
3189 		printf("unlimited\n");
3190 	else
3191 		printf("limited to %d packets/entry by default\n",
3192 		    V_verbose_limit);
3193 
3194 	/* Check user-supplied table count for validness */
3195 	if (default_fw_tables > IPFW_TABLES_MAX)
3196 	  default_fw_tables = IPFW_TABLES_MAX;
3197 
3198 	ipfw_init_sopt_handler();
3199 	ipfw_init_obj_rewriter();
3200 	ipfw_iface_init();
3201 	return (error);
3202 }
3203 
3204 /*
3205  * Called for the removal of the last instance only on module unload.
3206  */
3207 static void
3208 ipfw_destroy(void)
3209 {
3210 
3211 	ipfw_iface_destroy();
3212 	ipfw_destroy_sopt_handler();
3213 	ipfw_destroy_obj_rewriter();
3214 	printf("IP firewall unloaded\n");
3215 }
3216 
3217 /*
3218  * Stuff that must be initialized for every instance
3219  * (including the first of course).
3220  */
3221 static int
3222 vnet_ipfw_init(const void *unused)
3223 {
3224 	int error, first;
3225 	struct ip_fw *rule = NULL;
3226 	struct ip_fw_chain *chain;
3227 
3228 	chain = &V_layer3_chain;
3229 
3230 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3231 
3232 	/* First set up some values that are compile time options */
3233 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
3234 	V_fw_deny_unknown_exthdrs = 1;
3235 #ifdef IPFIREWALL_VERBOSE
3236 	V_fw_verbose = 1;
3237 #endif
3238 #ifdef IPFIREWALL_VERBOSE_LIMIT
3239 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3240 #endif
3241 #ifdef IPFIREWALL_NAT
3242 	LIST_INIT(&chain->nat);
3243 #endif
3244 
3245 	/* Init shared services hash table */
3246 	ipfw_init_srv(chain);
3247 
3248 	ipfw_init_counters();
3249 	/* Set initial number of tables */
3250 	V_fw_tables_max = default_fw_tables;
3251 	error = ipfw_init_tables(chain, first);
3252 	if (error) {
3253 		printf("ipfw2: setting up tables failed\n");
3254 		free(chain->map, M_IPFW);
3255 		free(rule, M_IPFW);
3256 		return (ENOSPC);
3257 	}
3258 
3259 	IPFW_LOCK_INIT(chain);
3260 
3261 	/* fill and insert the default rule */
3262 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3263 	rule->cmd_len = 1;
3264 	rule->cmd[0].len = 1;
3265 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3266 	chain->default_rule = rule;
3267 	ipfw_add_protected_rule(chain, rule, 0);
3268 
3269 	ipfw_dyn_init(chain);
3270 	ipfw_eaction_init(chain, first);
3271 #ifdef LINEAR_SKIPTO
3272 	ipfw_init_skipto_cache(chain);
3273 #endif
3274 	ipfw_bpf_init(first);
3275 
3276 	/* First set up some values that are compile time options */
3277 	V_ipfw_vnet_ready = 1;		/* Open for business */
3278 
3279 	/*
3280 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3281 	 * Even if the latter two fail we still keep the module alive
3282 	 * because the sockopt and layer2 paths are still useful.
3283 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3284 	 * so we can ignore the exact return value and just set a flag.
3285 	 *
3286 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3287 	 * changes in the underlying (per-vnet) variables trigger
3288 	 * immediate hook()/unhook() calls.
3289 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
3290 	 * is checked on each packet because there are no pfil hooks.
3291 	 */
3292 	V_ip_fw_ctl_ptr = ipfw_ctl3;
3293 	error = ipfw_attach_hooks(1);
3294 	return (error);
3295 }
3296 
3297 /*
3298  * Called for the removal of each instance.
3299  */
3300 static int
3301 vnet_ipfw_uninit(const void *unused)
3302 {
3303 	struct ip_fw *reap;
3304 	struct ip_fw_chain *chain = &V_layer3_chain;
3305 	int i, last;
3306 
3307 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3308 	/*
3309 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
3310 	 * Then grab, release and grab again the WLOCK so we make
3311 	 * sure the update is propagated and nobody will be in.
3312 	 */
3313 	(void)ipfw_attach_hooks(0 /* detach */);
3314 	V_ip_fw_ctl_ptr = NULL;
3315 
3316 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3317 
3318 	IPFW_UH_WLOCK(chain);
3319 	IPFW_UH_WUNLOCK(chain);
3320 
3321 	ipfw_dyn_uninit(0);	/* run the callout_drain */
3322 
3323 	IPFW_UH_WLOCK(chain);
3324 
3325 	reap = NULL;
3326 	IPFW_WLOCK(chain);
3327 	for (i = 0; i < chain->n_rules; i++)
3328 		ipfw_reap_add(chain, &reap, chain->map[i]);
3329 	free(chain->map, M_IPFW);
3330 #ifdef LINEAR_SKIPTO
3331 	ipfw_destroy_skipto_cache(chain);
3332 #endif
3333 	IPFW_WUNLOCK(chain);
3334 	IPFW_UH_WUNLOCK(chain);
3335 	ipfw_destroy_tables(chain, last);
3336 	ipfw_eaction_uninit(chain, last);
3337 	if (reap != NULL)
3338 		ipfw_reap_rules(reap);
3339 	vnet_ipfw_iface_destroy(chain);
3340 	ipfw_destroy_srv(chain);
3341 	IPFW_LOCK_DESTROY(chain);
3342 	ipfw_dyn_uninit(1);	/* free the remaining parts */
3343 	ipfw_destroy_counters();
3344 	ipfw_bpf_uninit(last);
3345 	return (0);
3346 }
3347 
3348 /*
3349  * Module event handler.
3350  * In general we have the choice of handling most of these events by the
3351  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3352  * use the SYSINIT handlers as they are more capable of expressing the
3353  * flow of control during module and vnet operations, so this is just
3354  * a skeleton. Note there is no SYSINIT equivalent of the module
3355  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3356  */
3357 static int
3358 ipfw_modevent(module_t mod, int type, void *unused)
3359 {
3360 	int err = 0;
3361 
3362 	switch (type) {
3363 	case MOD_LOAD:
3364 		/* Called once at module load or
3365 	 	 * system boot if compiled in. */
3366 		break;
3367 	case MOD_QUIESCE:
3368 		/* Called before unload. May veto unloading. */
3369 		break;
3370 	case MOD_UNLOAD:
3371 		/* Called during unload. */
3372 		break;
3373 	case MOD_SHUTDOWN:
3374 		/* Called during system shutdown. */
3375 		break;
3376 	default:
3377 		err = EOPNOTSUPP;
3378 		break;
3379 	}
3380 	return err;
3381 }
3382 
3383 static moduledata_t ipfwmod = {
3384 	"ipfw",
3385 	ipfw_modevent,
3386 	0
3387 };
3388 
3389 /* Define startup order. */
3390 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
3391 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
3392 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
3393 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
3394 
3395 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3396 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3397 MODULE_VERSION(ipfw, 3);
3398 /* should declare some dependencies here */
3399 
3400 /*
3401  * Starting up. Done in order after ipfwmod() has been called.
3402  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3403  */
3404 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3405 	    ipfw_init, NULL);
3406 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3407 	    vnet_ipfw_init, NULL);
3408 
3409 /*
3410  * Closing up shop. These are done in REVERSE ORDER, but still
3411  * after ipfwmod() has been called. Not called on reboot.
3412  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3413  * or when the module is unloaded.
3414  */
3415 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3416 	    ipfw_destroy, NULL);
3417 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3418 	    vnet_ipfw_uninit, NULL);
3419 /* end of file */
3420