xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 86aa9539fef591a363b06a0ebd3aa7a07f4c1579)
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 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs);
114 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
115 
116 VNET_DEFINE_STATIC(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_parse(struct tcphdr *tcp, uint16_t *mss)
335 {
336 	u_char *cp = (u_char *)(tcp + 1);
337 	int optlen, bits = 0;
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 		default:
354 			break;
355 
356 		case TCPOPT_MAXSEG:
357 			bits |= IP_FW_TCPOPT_MSS;
358 			if (mss != NULL)
359 				*mss = be16dec(cp + 2);
360 			break;
361 
362 		case TCPOPT_WINDOW:
363 			bits |= IP_FW_TCPOPT_WINDOW;
364 			break;
365 
366 		case TCPOPT_SACK_PERMITTED:
367 		case TCPOPT_SACK:
368 			bits |= IP_FW_TCPOPT_SACK;
369 			break;
370 
371 		case TCPOPT_TIMESTAMP:
372 			bits |= IP_FW_TCPOPT_TS;
373 			break;
374 		}
375 	}
376 	return (bits);
377 }
378 
379 static int
380 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
381 {
382 
383 	return (flags_match(cmd, tcpopts_parse(tcp, NULL)));
384 }
385 
386 static int
387 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
388     uint32_t *tablearg)
389 {
390 
391 	if (ifp == NULL)	/* no iface with this packet, match fails */
392 		return (0);
393 
394 	/* Check by name or by IP address */
395 	if (cmd->name[0] != '\0') { /* match by name */
396 		if (cmd->name[0] == '\1') /* use tablearg to match */
397 			return ipfw_lookup_table(chain, cmd->p.kidx, 0,
398 			    &ifp->if_index, tablearg);
399 		/* Check name */
400 		if (cmd->p.glob) {
401 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
402 				return(1);
403 		} else {
404 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
405 				return(1);
406 		}
407 	} else {
408 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
409 		struct ifaddr *ia;
410 
411 		if_addr_rlock(ifp);
412 		CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
413 			if (ia->ifa_addr->sa_family != AF_INET)
414 				continue;
415 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
416 			    (ia->ifa_addr))->sin_addr.s_addr) {
417 				if_addr_runlock(ifp);
418 				return(1);	/* match */
419 			}
420 		}
421 		if_addr_runlock(ifp);
422 #endif /* __FreeBSD__ */
423 	}
424 	return(0);	/* no match, fail ... */
425 }
426 
427 /*
428  * The verify_path function checks if a route to the src exists and
429  * if it is reachable via ifp (when provided).
430  *
431  * The 'verrevpath' option checks that the interface that an IP packet
432  * arrives on is the same interface that traffic destined for the
433  * packet's source address would be routed out of.
434  * The 'versrcreach' option just checks that the source address is
435  * reachable via any route (except default) in the routing table.
436  * These two are a measure to block forged packets. This is also
437  * commonly known as "anti-spoofing" or Unicast Reverse Path
438  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
439  * is purposely reminiscent of the Cisco IOS command,
440  *
441  *   ip verify unicast reverse-path
442  *   ip verify unicast source reachable-via any
443  *
444  * which implements the same functionality. But note that the syntax
445  * is misleading, and the check may be performed on all IP packets
446  * whether unicast, multicast, or broadcast.
447  */
448 static int
449 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
450 {
451 #if defined(USERSPACE) || !defined(__FreeBSD__)
452 	return 0;
453 #else
454 	struct nhop4_basic nh4;
455 
456 	if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
457 		return (0);
458 
459 	/*
460 	 * If ifp is provided, check for equality with rtentry.
461 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
462 	 * in order to pass packets injected back by if_simloop():
463 	 * routing entry (via lo0) for our own address
464 	 * may exist, so we need to handle routing assymetry.
465 	 */
466 	if (ifp != NULL && ifp != nh4.nh_ifp)
467 		return (0);
468 
469 	/* if no ifp provided, check if rtentry is not default route */
470 	if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
471 		return (0);
472 
473 	/* or if this is a blackhole/reject route */
474 	if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
475 		return (0);
476 
477 	/* found valid route */
478 	return 1;
479 #endif /* __FreeBSD__ */
480 }
481 
482 /*
483  * Generate an SCTP packet containing an ABORT chunk. The verification tag
484  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
485  * reflected is not 0.
486  */
487 
488 static struct mbuf *
489 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
490     int reflected)
491 {
492 	struct mbuf *m;
493 	struct ip *ip;
494 #ifdef INET6
495 	struct ip6_hdr *ip6;
496 #endif
497 	struct sctphdr *sctp;
498 	struct sctp_chunkhdr *chunk;
499 	u_int16_t hlen, plen, tlen;
500 
501 	MGETHDR(m, M_NOWAIT, MT_DATA);
502 	if (m == NULL)
503 		return (NULL);
504 
505 	M_SETFIB(m, id->fib);
506 #ifdef MAC
507 	if (replyto != NULL)
508 		mac_netinet_firewall_reply(replyto, m);
509 	else
510 		mac_netinet_firewall_send(m);
511 #else
512 	(void)replyto;		/* don't warn about unused arg */
513 #endif
514 
515 	switch (id->addr_type) {
516 	case 4:
517 		hlen = sizeof(struct ip);
518 		break;
519 #ifdef INET6
520 	case 6:
521 		hlen = sizeof(struct ip6_hdr);
522 		break;
523 #endif
524 	default:
525 		/* XXX: log me?!? */
526 		FREE_PKT(m);
527 		return (NULL);
528 	}
529 	plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
530 	tlen = hlen + plen;
531 	m->m_data += max_linkhdr;
532 	m->m_flags |= M_SKIP_FIREWALL;
533 	m->m_pkthdr.len = m->m_len = tlen;
534 	m->m_pkthdr.rcvif = NULL;
535 	bzero(m->m_data, tlen);
536 
537 	switch (id->addr_type) {
538 	case 4:
539 		ip = mtod(m, struct ip *);
540 
541 		ip->ip_v = 4;
542 		ip->ip_hl = sizeof(struct ip) >> 2;
543 		ip->ip_tos = IPTOS_LOWDELAY;
544 		ip->ip_len = htons(tlen);
545 		ip->ip_id = htons(0);
546 		ip->ip_off = htons(0);
547 		ip->ip_ttl = V_ip_defttl;
548 		ip->ip_p = IPPROTO_SCTP;
549 		ip->ip_sum = 0;
550 		ip->ip_src.s_addr = htonl(id->dst_ip);
551 		ip->ip_dst.s_addr = htonl(id->src_ip);
552 
553 		sctp = (struct sctphdr *)(ip + 1);
554 		break;
555 #ifdef INET6
556 	case 6:
557 		ip6 = mtod(m, struct ip6_hdr *);
558 
559 		ip6->ip6_vfc = IPV6_VERSION;
560 		ip6->ip6_plen = htons(plen);
561 		ip6->ip6_nxt = IPPROTO_SCTP;
562 		ip6->ip6_hlim = IPV6_DEFHLIM;
563 		ip6->ip6_src = id->dst_ip6;
564 		ip6->ip6_dst = id->src_ip6;
565 
566 		sctp = (struct sctphdr *)(ip6 + 1);
567 		break;
568 #endif
569 	}
570 
571 	sctp->src_port = htons(id->dst_port);
572 	sctp->dest_port = htons(id->src_port);
573 	sctp->v_tag = htonl(vtag);
574 	sctp->checksum = htonl(0);
575 
576 	chunk = (struct sctp_chunkhdr *)(sctp + 1);
577 	chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
578 	chunk->chunk_flags = 0;
579 	if (reflected != 0) {
580 		chunk->chunk_flags |= SCTP_HAD_NO_TCB;
581 	}
582 	chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
583 
584 	sctp->checksum = sctp_calculate_cksum(m, hlen);
585 
586 	return (m);
587 }
588 
589 /*
590  * Generate a TCP packet, containing either a RST or a keepalive.
591  * When flags & TH_RST, we are sending a RST packet, because of a
592  * "reset" action matched the packet.
593  * Otherwise we are sending a keepalive, and flags & TH_
594  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
595  * so that MAC can label the reply appropriately.
596  */
597 struct mbuf *
598 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
599     u_int32_t ack, int flags)
600 {
601 	struct mbuf *m = NULL;		/* stupid compiler */
602 	struct ip *h = NULL;		/* stupid compiler */
603 #ifdef INET6
604 	struct ip6_hdr *h6 = NULL;
605 #endif
606 	struct tcphdr *th = NULL;
607 	int len, dir;
608 
609 	MGETHDR(m, M_NOWAIT, MT_DATA);
610 	if (m == NULL)
611 		return (NULL);
612 
613 	M_SETFIB(m, id->fib);
614 #ifdef MAC
615 	if (replyto != NULL)
616 		mac_netinet_firewall_reply(replyto, m);
617 	else
618 		mac_netinet_firewall_send(m);
619 #else
620 	(void)replyto;		/* don't warn about unused arg */
621 #endif
622 
623 	switch (id->addr_type) {
624 	case 4:
625 		len = sizeof(struct ip) + sizeof(struct tcphdr);
626 		break;
627 #ifdef INET6
628 	case 6:
629 		len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
630 		break;
631 #endif
632 	default:
633 		/* XXX: log me?!? */
634 		FREE_PKT(m);
635 		return (NULL);
636 	}
637 	dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
638 
639 	m->m_data += max_linkhdr;
640 	m->m_flags |= M_SKIP_FIREWALL;
641 	m->m_pkthdr.len = m->m_len = len;
642 	m->m_pkthdr.rcvif = NULL;
643 	bzero(m->m_data, len);
644 
645 	switch (id->addr_type) {
646 	case 4:
647 		h = mtod(m, struct ip *);
648 
649 		/* prepare for checksum */
650 		h->ip_p = IPPROTO_TCP;
651 		h->ip_len = htons(sizeof(struct tcphdr));
652 		if (dir) {
653 			h->ip_src.s_addr = htonl(id->src_ip);
654 			h->ip_dst.s_addr = htonl(id->dst_ip);
655 		} else {
656 			h->ip_src.s_addr = htonl(id->dst_ip);
657 			h->ip_dst.s_addr = htonl(id->src_ip);
658 		}
659 
660 		th = (struct tcphdr *)(h + 1);
661 		break;
662 #ifdef INET6
663 	case 6:
664 		h6 = mtod(m, struct ip6_hdr *);
665 
666 		/* prepare for checksum */
667 		h6->ip6_nxt = IPPROTO_TCP;
668 		h6->ip6_plen = htons(sizeof(struct tcphdr));
669 		if (dir) {
670 			h6->ip6_src = id->src_ip6;
671 			h6->ip6_dst = id->dst_ip6;
672 		} else {
673 			h6->ip6_src = id->dst_ip6;
674 			h6->ip6_dst = id->src_ip6;
675 		}
676 
677 		th = (struct tcphdr *)(h6 + 1);
678 		break;
679 #endif
680 	}
681 
682 	if (dir) {
683 		th->th_sport = htons(id->src_port);
684 		th->th_dport = htons(id->dst_port);
685 	} else {
686 		th->th_sport = htons(id->dst_port);
687 		th->th_dport = htons(id->src_port);
688 	}
689 	th->th_off = sizeof(struct tcphdr) >> 2;
690 
691 	if (flags & TH_RST) {
692 		if (flags & TH_ACK) {
693 			th->th_seq = htonl(ack);
694 			th->th_flags = TH_RST;
695 		} else {
696 			if (flags & TH_SYN)
697 				seq++;
698 			th->th_ack = htonl(seq);
699 			th->th_flags = TH_RST | TH_ACK;
700 		}
701 	} else {
702 		/*
703 		 * Keepalive - use caller provided sequence numbers
704 		 */
705 		th->th_seq = htonl(seq);
706 		th->th_ack = htonl(ack);
707 		th->th_flags = TH_ACK;
708 	}
709 
710 	switch (id->addr_type) {
711 	case 4:
712 		th->th_sum = in_cksum(m, len);
713 
714 		/* finish the ip header */
715 		h->ip_v = 4;
716 		h->ip_hl = sizeof(*h) >> 2;
717 		h->ip_tos = IPTOS_LOWDELAY;
718 		h->ip_off = htons(0);
719 		h->ip_len = htons(len);
720 		h->ip_ttl = V_ip_defttl;
721 		h->ip_sum = 0;
722 		break;
723 #ifdef INET6
724 	case 6:
725 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
726 		    sizeof(struct tcphdr));
727 
728 		/* finish the ip6 header */
729 		h6->ip6_vfc |= IPV6_VERSION;
730 		h6->ip6_hlim = IPV6_DEFHLIM;
731 		break;
732 #endif
733 	}
734 
735 	return (m);
736 }
737 
738 #ifdef INET6
739 /*
740  * ipv6 specific rules here...
741  */
742 static __inline int
743 icmp6type_match (int type, ipfw_insn_u32 *cmd)
744 {
745 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
746 }
747 
748 static int
749 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
750 {
751 	int i;
752 	for (i=0; i <= cmd->o.arg1; ++i )
753 		if (curr_flow == cmd->d[i] )
754 			return 1;
755 	return 0;
756 }
757 
758 /* support for IP6_*_ME opcodes */
759 static const struct in6_addr lla_mask = {{{
760 	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
761 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
762 }}};
763 
764 static int
765 ipfw_localip6(struct in6_addr *in6)
766 {
767 	struct rm_priotracker in6_ifa_tracker;
768 	struct in6_ifaddr *ia;
769 
770 	if (IN6_IS_ADDR_MULTICAST(in6))
771 		return (0);
772 
773 	if (!IN6_IS_ADDR_LINKLOCAL(in6))
774 		return (in6_localip(in6));
775 
776 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
777 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
778 		if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
779 			continue;
780 		if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
781 		    in6, &lla_mask)) {
782 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
783 			return (1);
784 		}
785 	}
786 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
787 	return (0);
788 }
789 
790 static int
791 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
792 {
793 	struct nhop6_basic nh6;
794 
795 	if (IN6_IS_SCOPE_LINKLOCAL(src))
796 		return (1);
797 
798 	if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
799 		return (0);
800 
801 	/* If ifp is provided, check for equality with route table. */
802 	if (ifp != NULL && ifp != nh6.nh_ifp)
803 		return (0);
804 
805 	/* if no ifp provided, check if rtentry is not default route */
806 	if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
807 		return (0);
808 
809 	/* or if this is a blackhole/reject route */
810 	if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
811 		return (0);
812 
813 	/* found valid route */
814 	return 1;
815 }
816 
817 static int
818 is_icmp6_query(int icmp6_type)
819 {
820 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
821 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
822 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
823 	    icmp6_type == ICMP6_WRUREQUEST ||
824 	    icmp6_type == ICMP6_FQDN_QUERY ||
825 	    icmp6_type == ICMP6_NI_QUERY))
826 		return (1);
827 
828 	return (0);
829 }
830 
831 static int
832 map_icmp_unreach(int code)
833 {
834 
835 	/* RFC 7915 p4.2 */
836 	switch (code) {
837 	case ICMP_UNREACH_NET:
838 	case ICMP_UNREACH_HOST:
839 	case ICMP_UNREACH_SRCFAIL:
840 	case ICMP_UNREACH_NET_UNKNOWN:
841 	case ICMP_UNREACH_HOST_UNKNOWN:
842 	case ICMP_UNREACH_TOSNET:
843 	case ICMP_UNREACH_TOSHOST:
844 		return (ICMP6_DST_UNREACH_NOROUTE);
845 	case ICMP_UNREACH_PORT:
846 		return (ICMP6_DST_UNREACH_NOPORT);
847 	default:
848 		/*
849 		 * Map the rest of codes into admit prohibited.
850 		 * XXX: unreach proto should be mapped into ICMPv6
851 		 * parameter problem, but we use only unreach type.
852 		 */
853 		return (ICMP6_DST_UNREACH_ADMIN);
854 	}
855 }
856 
857 static void
858 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
859 {
860 	struct mbuf *m;
861 
862 	m = args->m;
863 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
864 		struct tcphdr *tcp;
865 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
866 
867 		if ((tcp->th_flags & TH_RST) == 0) {
868 			struct mbuf *m0;
869 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
870 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
871 			    tcp->th_flags | TH_RST);
872 			if (m0 != NULL)
873 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
874 				    NULL);
875 		}
876 		FREE_PKT(m);
877 	} else if (code == ICMP6_UNREACH_ABORT &&
878 	    args->f_id.proto == IPPROTO_SCTP) {
879 		struct mbuf *m0;
880 		struct sctphdr *sctp;
881 		u_int32_t v_tag;
882 		int reflected;
883 
884 		sctp = (struct sctphdr *)((char *)ip6 + hlen);
885 		reflected = 1;
886 		v_tag = ntohl(sctp->v_tag);
887 		/* Investigate the first chunk header if available */
888 		if (m->m_len >= hlen + sizeof(struct sctphdr) +
889 		    sizeof(struct sctp_chunkhdr)) {
890 			struct sctp_chunkhdr *chunk;
891 
892 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
893 			switch (chunk->chunk_type) {
894 			case SCTP_INITIATION:
895 				/*
896 				 * Packets containing an INIT chunk MUST have
897 				 * a zero v-tag.
898 				 */
899 				if (v_tag != 0) {
900 					v_tag = 0;
901 					break;
902 				}
903 				/* INIT chunk MUST NOT be bundled */
904 				if (m->m_pkthdr.len >
905 				    hlen + sizeof(struct sctphdr) +
906 				    ntohs(chunk->chunk_length) + 3) {
907 					break;
908 				}
909 				/* Use the initiate tag if available */
910 				if ((m->m_len >= hlen + sizeof(struct sctphdr) +
911 				    sizeof(struct sctp_chunkhdr) +
912 				    offsetof(struct sctp_init, a_rwnd))) {
913 					struct sctp_init *init;
914 
915 					init = (struct sctp_init *)(chunk + 1);
916 					v_tag = ntohl(init->initiate_tag);
917 					reflected = 0;
918 				}
919 				break;
920 			case SCTP_ABORT_ASSOCIATION:
921 				/*
922 				 * If the packet contains an ABORT chunk, don't
923 				 * reply.
924 				 * XXX: We should search through all chunks,
925 				 *      but don't do to avoid attacks.
926 				 */
927 				v_tag = 0;
928 				break;
929 			}
930 		}
931 		if (v_tag == 0) {
932 			m0 = NULL;
933 		} else {
934 			m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
935 			    reflected);
936 		}
937 		if (m0 != NULL)
938 			ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
939 		FREE_PKT(m);
940 	} else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
941 		/* Send an ICMPv6 unreach. */
942 #if 0
943 		/*
944 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
945 		 * as the contents are read. We need to m_adj() the
946 		 * needed amount.
947 		 * The mbuf will however be thrown away so we can adjust it.
948 		 * Remember we did an m_pullup on it already so we
949 		 * can make some assumptions about contiguousness.
950 		 */
951 		if (args->L3offset)
952 			m_adj(m, args->L3offset);
953 #endif
954 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
955 	} else
956 		FREE_PKT(m);
957 
958 	args->m = NULL;
959 }
960 
961 #endif /* INET6 */
962 
963 
964 /*
965  * sends a reject message, consuming the mbuf passed as an argument.
966  */
967 static void
968 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
969 {
970 
971 #if 0
972 	/* XXX When ip is not guaranteed to be at mtod() we will
973 	 * need to account for this */
974 	 * The mbuf will however be thrown away so we can adjust it.
975 	 * Remember we did an m_pullup on it already so we
976 	 * can make some assumptions about contiguousness.
977 	 */
978 	if (args->L3offset)
979 		m_adj(m, args->L3offset);
980 #endif
981 	if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
982 		/* Send an ICMP unreach */
983 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
984 	} else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
985 		struct tcphdr *const tcp =
986 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
987 		if ( (tcp->th_flags & TH_RST) == 0) {
988 			struct mbuf *m;
989 			m = ipfw_send_pkt(args->m, &(args->f_id),
990 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
991 				tcp->th_flags | TH_RST);
992 			if (m != NULL)
993 				ip_output(m, NULL, NULL, 0, NULL, NULL);
994 		}
995 		FREE_PKT(args->m);
996 	} else if (code == ICMP_REJECT_ABORT &&
997 	    args->f_id.proto == IPPROTO_SCTP) {
998 		struct mbuf *m;
999 		struct sctphdr *sctp;
1000 		struct sctp_chunkhdr *chunk;
1001 		struct sctp_init *init;
1002 		u_int32_t v_tag;
1003 		int reflected;
1004 
1005 		sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1006 		reflected = 1;
1007 		v_tag = ntohl(sctp->v_tag);
1008 		if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1009 		    sizeof(struct sctp_chunkhdr)) {
1010 			/* Look at the first chunk header if available */
1011 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
1012 			switch (chunk->chunk_type) {
1013 			case SCTP_INITIATION:
1014 				/*
1015 				 * Packets containing an INIT chunk MUST have
1016 				 * a zero v-tag.
1017 				 */
1018 				if (v_tag != 0) {
1019 					v_tag = 0;
1020 					break;
1021 				}
1022 				/* INIT chunk MUST NOT be bundled */
1023 				if (iplen >
1024 				    (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1025 				    ntohs(chunk->chunk_length) + 3) {
1026 					break;
1027 				}
1028 				/* Use the initiate tag if available */
1029 				if ((iplen >= (ip->ip_hl << 2) +
1030 				    sizeof(struct sctphdr) +
1031 				    sizeof(struct sctp_chunkhdr) +
1032 				    offsetof(struct sctp_init, a_rwnd))) {
1033 					init = (struct sctp_init *)(chunk + 1);
1034 					v_tag = ntohl(init->initiate_tag);
1035 					reflected = 0;
1036 				}
1037 				break;
1038 			case SCTP_ABORT_ASSOCIATION:
1039 				/*
1040 				 * If the packet contains an ABORT chunk, don't
1041 				 * reply.
1042 				 * XXX: We should search through all chunks,
1043 				 * but don't do to avoid attacks.
1044 				 */
1045 				v_tag = 0;
1046 				break;
1047 			}
1048 		}
1049 		if (v_tag == 0) {
1050 			m = NULL;
1051 		} else {
1052 			m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1053 			    reflected);
1054 		}
1055 		if (m != NULL)
1056 			ip_output(m, NULL, NULL, 0, NULL, NULL);
1057 		FREE_PKT(args->m);
1058 	} else
1059 		FREE_PKT(args->m);
1060 	args->m = NULL;
1061 }
1062 
1063 /*
1064  * Support for uid/gid/jail lookup. These tests are expensive
1065  * (because we may need to look into the list of active sockets)
1066  * so we cache the results. ugid_lookupp is 0 if we have not
1067  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1068  * and failed. The function always returns the match value.
1069  * We could actually spare the variable and use *uc, setting
1070  * it to '(void *)check_uidgid if we have no info, NULL if
1071  * we tried and failed, or any other value if successful.
1072  */
1073 static int
1074 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1075     struct ucred **uc)
1076 {
1077 #if defined(USERSPACE)
1078 	return 0;	// not supported in userspace
1079 #else
1080 #ifndef __FreeBSD__
1081 	/* XXX */
1082 	return cred_check(insn, proto, oif,
1083 	    dst_ip, dst_port, src_ip, src_port,
1084 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1085 #else  /* FreeBSD */
1086 	struct in_addr src_ip, dst_ip;
1087 	struct inpcbinfo *pi;
1088 	struct ipfw_flow_id *id;
1089 	struct inpcb *pcb, *inp;
1090 	int lookupflags;
1091 	int match;
1092 
1093 	id = &args->f_id;
1094 	inp = args->inp;
1095 
1096 	/*
1097 	 * Check to see if the UDP or TCP stack supplied us with
1098 	 * the PCB. If so, rather then holding a lock and looking
1099 	 * up the PCB, we can use the one that was supplied.
1100 	 */
1101 	if (inp && *ugid_lookupp == 0) {
1102 		INP_LOCK_ASSERT(inp);
1103 		if (inp->inp_socket != NULL) {
1104 			*uc = crhold(inp->inp_cred);
1105 			*ugid_lookupp = 1;
1106 		} else
1107 			*ugid_lookupp = -1;
1108 	}
1109 	/*
1110 	 * If we have already been here and the packet has no
1111 	 * PCB entry associated with it, then we can safely
1112 	 * assume that this is a no match.
1113 	 */
1114 	if (*ugid_lookupp == -1)
1115 		return (0);
1116 	if (id->proto == IPPROTO_TCP) {
1117 		lookupflags = 0;
1118 		pi = &V_tcbinfo;
1119 	} else if (id->proto == IPPROTO_UDP) {
1120 		lookupflags = INPLOOKUP_WILDCARD;
1121 		pi = &V_udbinfo;
1122 	} else if (id->proto == IPPROTO_UDPLITE) {
1123 		lookupflags = INPLOOKUP_WILDCARD;
1124 		pi = &V_ulitecbinfo;
1125 	} else
1126 		return 0;
1127 	lookupflags |= INPLOOKUP_RLOCKPCB;
1128 	match = 0;
1129 	if (*ugid_lookupp == 0) {
1130 		if (id->addr_type == 6) {
1131 #ifdef INET6
1132 			if (args->flags & IPFW_ARGS_IN)
1133 				pcb = in6_pcblookup_mbuf(pi,
1134 				    &id->src_ip6, htons(id->src_port),
1135 				    &id->dst_ip6, htons(id->dst_port),
1136 				    lookupflags, NULL, args->m);
1137 			else
1138 				pcb = in6_pcblookup_mbuf(pi,
1139 				    &id->dst_ip6, htons(id->dst_port),
1140 				    &id->src_ip6, htons(id->src_port),
1141 				    lookupflags, args->ifp, args->m);
1142 #else
1143 			*ugid_lookupp = -1;
1144 			return (0);
1145 #endif
1146 		} else {
1147 			src_ip.s_addr = htonl(id->src_ip);
1148 			dst_ip.s_addr = htonl(id->dst_ip);
1149 			if (args->flags & IPFW_ARGS_IN)
1150 				pcb = in_pcblookup_mbuf(pi,
1151 				    src_ip, htons(id->src_port),
1152 				    dst_ip, htons(id->dst_port),
1153 				    lookupflags, NULL, args->m);
1154 			else
1155 				pcb = in_pcblookup_mbuf(pi,
1156 				    dst_ip, htons(id->dst_port),
1157 				    src_ip, htons(id->src_port),
1158 				    lookupflags, args->ifp, args->m);
1159 		}
1160 		if (pcb != NULL) {
1161 			INP_RLOCK_ASSERT(pcb);
1162 			*uc = crhold(pcb->inp_cred);
1163 			*ugid_lookupp = 1;
1164 			INP_RUNLOCK(pcb);
1165 		}
1166 		if (*ugid_lookupp == 0) {
1167 			/*
1168 			 * We tried and failed, set the variable to -1
1169 			 * so we will not try again on this packet.
1170 			 */
1171 			*ugid_lookupp = -1;
1172 			return (0);
1173 		}
1174 	}
1175 	if (insn->o.opcode == O_UID)
1176 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1177 	else if (insn->o.opcode == O_GID)
1178 		match = groupmember((gid_t)insn->d[0], *uc);
1179 	else if (insn->o.opcode == O_JAIL)
1180 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1181 	return (match);
1182 #endif /* __FreeBSD__ */
1183 #endif /* not supported in userspace */
1184 }
1185 
1186 /*
1187  * Helper function to set args with info on the rule after the matching
1188  * one. slot is precise, whereas we guess rule_id as they are
1189  * assigned sequentially.
1190  */
1191 static inline void
1192 set_match(struct ip_fw_args *args, int slot,
1193 	struct ip_fw_chain *chain)
1194 {
1195 	args->rule.chain_id = chain->id;
1196 	args->rule.slot = slot + 1; /* we use 0 as a marker */
1197 	args->rule.rule_id = 1 + chain->map[slot]->id;
1198 	args->rule.rulenum = chain->map[slot]->rulenum;
1199 	args->flags |= IPFW_ARGS_REF;
1200 }
1201 
1202 #ifndef LINEAR_SKIPTO
1203 /*
1204  * Helper function to enable cached rule lookups using
1205  * cached_id and cached_pos fields in ipfw rule.
1206  */
1207 static int
1208 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1209     int tablearg, int jump_backwards)
1210 {
1211 	int f_pos;
1212 
1213 	/* If possible use cached f_pos (in f->cached_pos),
1214 	 * whose version is written in f->cached_id
1215 	 * (horrible hacks to avoid changing the ABI).
1216 	 */
1217 	if (num != IP_FW_TARG && f->cached_id == chain->id)
1218 		f_pos = f->cached_pos;
1219 	else {
1220 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1221 		/* make sure we do not jump backward */
1222 		if (jump_backwards == 0 && i <= f->rulenum)
1223 			i = f->rulenum + 1;
1224 		if (chain->idxmap != NULL)
1225 			f_pos = chain->idxmap[i];
1226 		else
1227 			f_pos = ipfw_find_rule(chain, i, 0);
1228 		/* update the cache */
1229 		if (num != IP_FW_TARG) {
1230 			f->cached_id = chain->id;
1231 			f->cached_pos = f_pos;
1232 		}
1233 	}
1234 
1235 	return (f_pos);
1236 }
1237 #else
1238 /*
1239  * Helper function to enable real fast rule lookups.
1240  */
1241 static int
1242 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1243     int tablearg, int jump_backwards)
1244 {
1245 	int f_pos;
1246 
1247 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1248 	/* make sure we do not jump backward */
1249 	if (jump_backwards == 0 && num <= f->rulenum)
1250 		num = f->rulenum + 1;
1251 	f_pos = chain->idxmap[num];
1252 
1253 	return (f_pos);
1254 }
1255 #endif
1256 
1257 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
1258 /*
1259  * The main check routine for the firewall.
1260  *
1261  * All arguments are in args so we can modify them and return them
1262  * back to the caller.
1263  *
1264  * Parameters:
1265  *
1266  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
1267  *		Starts with the IP header.
1268  *	args->L3offset	Number of bytes bypassed if we came from L2.
1269  *			e.g. often sizeof(eh)  ** NOTYET **
1270  *	args->ifp	Incoming or outgoing interface.
1271  *	args->divert_rule (in/out)
1272  *		Skip up to the first rule past this rule number;
1273  *		upon return, non-zero port number for divert or tee.
1274  *
1275  *	args->rule	Pointer to the last matching rule (in/out)
1276  *	args->next_hop	Socket we are forwarding to (out).
1277  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
1278  *	args->f_id	Addresses grabbed from the packet (out)
1279  * 	args->rule.info	a cookie depending on rule action
1280  *
1281  * Return value:
1282  *
1283  *	IP_FW_PASS	the packet must be accepted
1284  *	IP_FW_DENY	the packet must be dropped
1285  *	IP_FW_DIVERT	divert packet, port in m_tag
1286  *	IP_FW_TEE	tee packet, port in m_tag
1287  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
1288  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
1289  *		args->rule contains the matching rule,
1290  *		args->rule.info has additional information.
1291  *
1292  */
1293 int
1294 ipfw_chk(struct ip_fw_args *args)
1295 {
1296 
1297 	/*
1298 	 * Local variables holding state while processing a packet:
1299 	 *
1300 	 * IMPORTANT NOTE: to speed up the processing of rules, there
1301 	 * are some assumption on the values of the variables, which
1302 	 * are documented here. Should you change them, please check
1303 	 * the implementation of the various instructions to make sure
1304 	 * that they still work.
1305 	 *
1306 	 * m | args->m	Pointer to the mbuf, as received from the caller.
1307 	 *	It may change if ipfw_chk() does an m_pullup, or if it
1308 	 *	consumes the packet because it calls send_reject().
1309 	 *	XXX This has to change, so that ipfw_chk() never modifies
1310 	 *	or consumes the buffer.
1311 	 *	OR
1312 	 * args->mem	Pointer to contigous memory chunk.
1313 	 * ip	Is the beginning of the ip(4 or 6) header.
1314 	 * eh	Ethernet header in case if input is Layer2.
1315 	 */
1316 	struct mbuf *m;
1317 	struct ip *ip;
1318 	struct ether_header *eh;
1319 
1320 	/*
1321 	 * For rules which contain uid/gid or jail constraints, cache
1322 	 * a copy of the users credentials after the pcb lookup has been
1323 	 * executed. This will speed up the processing of rules with
1324 	 * these types of constraints, as well as decrease contention
1325 	 * on pcb related locks.
1326 	 */
1327 #ifndef __FreeBSD__
1328 	struct bsd_ucred ucred_cache;
1329 #else
1330 	struct ucred *ucred_cache = NULL;
1331 #endif
1332 	int ucred_lookup = 0;
1333 	int f_pos = 0;		/* index of current rule in the array */
1334 	int retval = 0;
1335 	struct ifnet *oif, *iif;
1336 
1337 	/*
1338 	 * hlen	The length of the IP header.
1339 	 */
1340 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
1341 
1342 	/*
1343 	 * offset	The offset of a fragment. offset != 0 means that
1344 	 *	we have a fragment at this offset of an IPv4 packet.
1345 	 *	offset == 0 means that (if this is an IPv4 packet)
1346 	 *	this is the first or only fragment.
1347 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1348 	 *	or there is a single packet fragment (fragment header added
1349 	 *	without needed).  We will treat a single packet fragment as if
1350 	 *	there was no fragment header (or log/block depending on the
1351 	 *	V_fw_permit_single_frag6 sysctl setting).
1352 	 */
1353 	u_short offset = 0;
1354 	u_short ip6f_mf = 0;
1355 
1356 	/*
1357 	 * Local copies of addresses. They are only valid if we have
1358 	 * an IP packet.
1359 	 *
1360 	 * proto	The protocol. Set to 0 for non-ip packets,
1361 	 *	or to the protocol read from the packet otherwise.
1362 	 *	proto != 0 means that we have an IPv4 packet.
1363 	 *
1364 	 * src_port, dst_port	port numbers, in HOST format. Only
1365 	 *	valid for TCP and UDP packets.
1366 	 *
1367 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
1368 	 *	Only valid for IPv4 packets.
1369 	 */
1370 	uint8_t proto;
1371 	uint16_t src_port, dst_port;		/* NOTE: host format	*/
1372 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
1373 	int iplen = 0;
1374 	int pktlen;
1375 
1376 	struct ipfw_dyn_info dyn_info;
1377 	struct ip_fw *q = NULL;
1378 	struct ip_fw_chain *chain = &V_layer3_chain;
1379 
1380 	/*
1381 	 * We store in ulp a pointer to the upper layer protocol header.
1382 	 * In the ipv4 case this is easy to determine from the header,
1383 	 * but for ipv6 we might have some additional headers in the middle.
1384 	 * ulp is NULL if not found.
1385 	 */
1386 	void *ulp = NULL;		/* upper layer protocol pointer. */
1387 
1388 	/* XXX ipv6 variables */
1389 	int is_ipv6 = 0;
1390 	uint8_t	icmp6_type = 0;
1391 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1392 	/* end of ipv6 variables */
1393 
1394 	int is_ipv4 = 0;
1395 
1396 	int done = 0;		/* flag to exit the outer loop */
1397 	IPFW_RLOCK_TRACKER;
1398 	bool mem;
1399 
1400 	if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1401 		if (args->flags & IPFW_ARGS_ETHER) {
1402 			eh = (struct ether_header *)args->mem;
1403 			if (eh->ether_type == htons(ETHERTYPE_VLAN))
1404 				ip = (struct ip *)
1405 				    ((struct ether_vlan_header *)eh + 1);
1406 			else
1407 				ip = (struct ip *)(eh + 1);
1408 		} else {
1409 			eh = NULL;
1410 			ip = (struct ip *)args->mem;
1411 		}
1412 		pktlen = IPFW_ARGS_LENGTH(args->flags);
1413 		args->f_id.fib = args->ifp->if_fib;	/* best guess */
1414 	} else {
1415 		m = args->m;
1416 		if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1417 			return (IP_FW_PASS);	/* accept */
1418 		if (args->flags & IPFW_ARGS_ETHER) {
1419 	                /* We need some amount of data to be contiguous. */
1420 			if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1421 			    (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1422 			    max_protohdr))) == NULL)
1423 				goto pullup_failed;
1424 			eh = mtod(m, struct ether_header *);
1425 			ip = (struct ip *)(eh + 1);
1426 		} else {
1427 			eh = NULL;
1428 			ip = mtod(m, struct ip *);
1429 		}
1430 		pktlen = m->m_pkthdr.len;
1431 		args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1432 	}
1433 
1434 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1435 	src_ip.s_addr = 0;		/* make sure it is initialized */
1436 	src_port = dst_port = 0;
1437 
1438 	DYN_INFO_INIT(&dyn_info);
1439 /*
1440  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1441  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1442  * pointer might become stale after other pullups (but we never use it
1443  * this way).
1444  */
1445 #define	PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1446 #define	EHLEN	(eh != NULL ? ((char *)ip - (char *)eh) : 0)
1447 #define	_PULLUP_LOCKED(_len, p, T, unlock)			\
1448 do {								\
1449 	int x = (_len) + T + EHLEN;				\
1450 	if (mem) {						\
1451 		MPASS(pktlen >= x);				\
1452 		p = (char *)args->mem + (_len) + EHLEN;		\
1453 	} else {						\
1454 		if (__predict_false((m)->m_len < x)) {		\
1455 			args->m = m = m_pullup(m, x);		\
1456 			if (m == NULL) {			\
1457 				unlock;				\
1458 				goto pullup_failed;		\
1459 			}					\
1460 		}						\
1461 		p = mtod(m, char *) + (_len) + EHLEN;		\
1462 	}							\
1463 } while (0)
1464 
1465 #define	PULLUP_LEN(_len, p, T)	_PULLUP_LOCKED(_len, p, T, )
1466 #define	PULLUP_LEN_LOCKED(_len, p, T)	\
1467     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain))
1468 /*
1469  * In case pointers got stale after pullups, update them.
1470  */
1471 #define	UPDATE_POINTERS()					\
1472 do {								\
1473 	if (!mem) {						\
1474 		if (eh != NULL) {				\
1475 			eh = mtod(m, struct ether_header *);	\
1476 			ip = (struct ip *)(eh + 1);		\
1477 		} else						\
1478 			ip = mtod(m, struct ip *);		\
1479 		args->m = m;					\
1480 	}							\
1481 } while (0)
1482 
1483 	/* Identify IP packets and fill up variables. */
1484 	if (pktlen >= sizeof(struct ip6_hdr) &&
1485 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1486 	    ip->ip_v == 6) {
1487 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1488 
1489 		is_ipv6 = 1;
1490 		args->flags |= IPFW_ARGS_IP6;
1491 		hlen = sizeof(struct ip6_hdr);
1492 		proto = ip6->ip6_nxt;
1493 		/* Search extension headers to find upper layer protocols */
1494 		while (ulp == NULL && offset == 0) {
1495 			switch (proto) {
1496 			case IPPROTO_ICMPV6:
1497 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1498 				icmp6_type = ICMP6(ulp)->icmp6_type;
1499 				break;
1500 
1501 			case IPPROTO_TCP:
1502 				PULLUP_TO(hlen, ulp, struct tcphdr);
1503 				dst_port = TCP(ulp)->th_dport;
1504 				src_port = TCP(ulp)->th_sport;
1505 				/* save flags for dynamic rules */
1506 				args->f_id._flags = TCP(ulp)->th_flags;
1507 				break;
1508 
1509 			case IPPROTO_SCTP:
1510 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1511 				    sizeof(struct sctp_chunkhdr) +
1512 				    offsetof(struct sctp_init, a_rwnd))
1513 					PULLUP_LEN(hlen, ulp,
1514 					    sizeof(struct sctphdr) +
1515 					    sizeof(struct sctp_chunkhdr) +
1516 					    offsetof(struct sctp_init, a_rwnd));
1517 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1518 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1519 				else
1520 					PULLUP_LEN(hlen, ulp,
1521 					    sizeof(struct sctphdr));
1522 				src_port = SCTP(ulp)->src_port;
1523 				dst_port = SCTP(ulp)->dest_port;
1524 				break;
1525 
1526 			case IPPROTO_UDP:
1527 			case IPPROTO_UDPLITE:
1528 				PULLUP_TO(hlen, ulp, struct udphdr);
1529 				dst_port = UDP(ulp)->uh_dport;
1530 				src_port = UDP(ulp)->uh_sport;
1531 				break;
1532 
1533 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1534 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1535 				ext_hd |= EXT_HOPOPTS;
1536 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1537 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1538 				ulp = NULL;
1539 				break;
1540 
1541 			case IPPROTO_ROUTING:	/* RFC 2460 */
1542 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1543 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1544 				case 0:
1545 					ext_hd |= EXT_RTHDR0;
1546 					break;
1547 				case 2:
1548 					ext_hd |= EXT_RTHDR2;
1549 					break;
1550 				default:
1551 					if (V_fw_verbose)
1552 						printf("IPFW2: IPV6 - Unknown "
1553 						    "Routing Header type(%d)\n",
1554 						    ((struct ip6_rthdr *)
1555 						    ulp)->ip6r_type);
1556 					if (V_fw_deny_unknown_exthdrs)
1557 					    return (IP_FW_DENY);
1558 					break;
1559 				}
1560 				ext_hd |= EXT_ROUTING;
1561 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1562 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1563 				ulp = NULL;
1564 				break;
1565 
1566 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1567 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1568 				ext_hd |= EXT_FRAGMENT;
1569 				hlen += sizeof (struct ip6_frag);
1570 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1571 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1572 					IP6F_OFF_MASK;
1573 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1574 					IP6F_MORE_FRAG;
1575 				if (V_fw_permit_single_frag6 == 0 &&
1576 				    offset == 0 && ip6f_mf == 0) {
1577 					if (V_fw_verbose)
1578 						printf("IPFW2: IPV6 - Invalid "
1579 						    "Fragment Header\n");
1580 					if (V_fw_deny_unknown_exthdrs)
1581 					    return (IP_FW_DENY);
1582 					break;
1583 				}
1584 				args->f_id.extra =
1585 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1586 				ulp = NULL;
1587 				break;
1588 
1589 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1590 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1591 				ext_hd |= EXT_DSTOPTS;
1592 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1593 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1594 				ulp = NULL;
1595 				break;
1596 
1597 			case IPPROTO_AH:	/* RFC 2402 */
1598 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1599 				ext_hd |= EXT_AH;
1600 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1601 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1602 				ulp = NULL;
1603 				break;
1604 
1605 			case IPPROTO_ESP:	/* RFC 2406 */
1606 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1607 				/* Anything past Seq# is variable length and
1608 				 * data past this ext. header is encrypted. */
1609 				ext_hd |= EXT_ESP;
1610 				break;
1611 
1612 			case IPPROTO_NONE:	/* RFC 2460 */
1613 				/*
1614 				 * Packet ends here, and IPv6 header has
1615 				 * already been pulled up. If ip6e_len!=0
1616 				 * then octets must be ignored.
1617 				 */
1618 				ulp = ip; /* non-NULL to get out of loop. */
1619 				break;
1620 
1621 			case IPPROTO_OSPFIGP:
1622 				/* XXX OSPF header check? */
1623 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1624 				break;
1625 
1626 			case IPPROTO_PIM:
1627 				/* XXX PIM header check? */
1628 				PULLUP_TO(hlen, ulp, struct pim);
1629 				break;
1630 
1631 			case IPPROTO_GRE:	/* RFC 1701 */
1632 				/* XXX GRE header check? */
1633 				PULLUP_TO(hlen, ulp, struct grehdr);
1634 				break;
1635 
1636 			case IPPROTO_CARP:
1637 				PULLUP_TO(hlen, ulp, offsetof(
1638 				    struct carp_header, carp_counter));
1639 				if (CARP_ADVERTISEMENT !=
1640 				    ((struct carp_header *)ulp)->carp_type)
1641 					return (IP_FW_DENY);
1642 				break;
1643 
1644 			case IPPROTO_IPV6:	/* RFC 2893 */
1645 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1646 				break;
1647 
1648 			case IPPROTO_IPV4:	/* RFC 2893 */
1649 				PULLUP_TO(hlen, ulp, struct ip);
1650 				break;
1651 
1652 			default:
1653 				if (V_fw_verbose)
1654 					printf("IPFW2: IPV6 - Unknown "
1655 					    "Extension Header(%d), ext_hd=%x\n",
1656 					     proto, ext_hd);
1657 				if (V_fw_deny_unknown_exthdrs)
1658 				    return (IP_FW_DENY);
1659 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1660 				break;
1661 			} /*switch */
1662 		}
1663 		UPDATE_POINTERS();
1664 		ip6 = (struct ip6_hdr *)ip;
1665 		args->f_id.addr_type = 6;
1666 		args->f_id.src_ip6 = ip6->ip6_src;
1667 		args->f_id.dst_ip6 = ip6->ip6_dst;
1668 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1669 		iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1670 	} else if (pktlen >= sizeof(struct ip) &&
1671 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1672 	    ip->ip_v == 4) {
1673 		is_ipv4 = 1;
1674 		args->flags |= IPFW_ARGS_IP4;
1675 		hlen = ip->ip_hl << 2;
1676 		/*
1677 		 * Collect parameters into local variables for faster
1678 		 * matching.
1679 		 */
1680 		proto = ip->ip_p;
1681 		src_ip = ip->ip_src;
1682 		dst_ip = ip->ip_dst;
1683 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1684 		iplen = ntohs(ip->ip_len);
1685 
1686 		if (offset == 0) {
1687 			switch (proto) {
1688 			case IPPROTO_TCP:
1689 				PULLUP_TO(hlen, ulp, struct tcphdr);
1690 				dst_port = TCP(ulp)->th_dport;
1691 				src_port = TCP(ulp)->th_sport;
1692 				/* save flags for dynamic rules */
1693 				args->f_id._flags = TCP(ulp)->th_flags;
1694 				break;
1695 
1696 			case IPPROTO_SCTP:
1697 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1698 				    sizeof(struct sctp_chunkhdr) +
1699 				    offsetof(struct sctp_init, a_rwnd))
1700 					PULLUP_LEN(hlen, ulp,
1701 					    sizeof(struct sctphdr) +
1702 					    sizeof(struct sctp_chunkhdr) +
1703 					    offsetof(struct sctp_init, a_rwnd));
1704 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1705 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1706 				else
1707 					PULLUP_LEN(hlen, ulp,
1708 					    sizeof(struct sctphdr));
1709 				src_port = SCTP(ulp)->src_port;
1710 				dst_port = SCTP(ulp)->dest_port;
1711 				break;
1712 
1713 			case IPPROTO_UDP:
1714 			case IPPROTO_UDPLITE:
1715 				PULLUP_TO(hlen, ulp, struct udphdr);
1716 				dst_port = UDP(ulp)->uh_dport;
1717 				src_port = UDP(ulp)->uh_sport;
1718 				break;
1719 
1720 			case IPPROTO_ICMP:
1721 				PULLUP_TO(hlen, ulp, struct icmphdr);
1722 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1723 				break;
1724 
1725 			default:
1726 				break;
1727 			}
1728 		} else {
1729 			if (offset == 1 && proto == IPPROTO_TCP) {
1730 				/* RFC 3128 */
1731 				goto pullup_failed;
1732 			}
1733 		}
1734 
1735 		UPDATE_POINTERS();
1736 		args->f_id.addr_type = 4;
1737 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1738 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1739 	} else {
1740 		proto = 0;
1741 		dst_ip.s_addr = src_ip.s_addr = 0;
1742 
1743 		args->f_id.addr_type = 1; /* XXX */
1744 	}
1745 #undef PULLUP_TO
1746 	pktlen = iplen < pktlen ? iplen: pktlen;
1747 
1748 	/* Properly initialize the rest of f_id */
1749 	args->f_id.proto = proto;
1750 	args->f_id.src_port = src_port = ntohs(src_port);
1751 	args->f_id.dst_port = dst_port = ntohs(dst_port);
1752 
1753 	IPFW_PF_RLOCK(chain);
1754 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1755 		IPFW_PF_RUNLOCK(chain);
1756 		return (IP_FW_PASS);	/* accept */
1757 	}
1758 	if (args->flags & IPFW_ARGS_REF) {
1759 		/*
1760 		 * Packet has already been tagged as a result of a previous
1761 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1762 		 * REASS, NETGRAPH, DIVERT/TEE...)
1763 		 * Validate the slot and continue from the next one
1764 		 * if still present, otherwise do a lookup.
1765 		 */
1766 		f_pos = (args->rule.chain_id == chain->id) ?
1767 		    args->rule.slot :
1768 		    ipfw_find_rule(chain, args->rule.rulenum,
1769 			args->rule.rule_id);
1770 	} else {
1771 		f_pos = 0;
1772 	}
1773 
1774 	if (args->flags & IPFW_ARGS_IN) {
1775 		iif = args->ifp;
1776 		oif = NULL;
1777 	} else {
1778 		MPASS(args->flags & IPFW_ARGS_OUT);
1779 		iif = mem ? NULL : m_rcvif(m);
1780 		oif = args->ifp;
1781 	}
1782 
1783 	/*
1784 	 * Now scan the rules, and parse microinstructions for each rule.
1785 	 * We have two nested loops and an inner switch. Sometimes we
1786 	 * need to break out of one or both loops, or re-enter one of
1787 	 * the loops with updated variables. Loop variables are:
1788 	 *
1789 	 *	f_pos (outer loop) points to the current rule.
1790 	 *		On output it points to the matching rule.
1791 	 *	done (outer loop) is used as a flag to break the loop.
1792 	 *	l (inner loop)	residual length of current rule.
1793 	 *		cmd points to the current microinstruction.
1794 	 *
1795 	 * We break the inner loop by setting l=0 and possibly
1796 	 * cmdlen=0 if we don't want to advance cmd.
1797 	 * We break the outer loop by setting done=1
1798 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1799 	 * as needed.
1800 	 */
1801 	for (; f_pos < chain->n_rules; f_pos++) {
1802 		ipfw_insn *cmd;
1803 		uint32_t tablearg = 0;
1804 		int l, cmdlen, skip_or; /* skip rest of OR block */
1805 		struct ip_fw *f;
1806 
1807 		f = chain->map[f_pos];
1808 		if (V_set_disable & (1 << f->set) )
1809 			continue;
1810 
1811 		skip_or = 0;
1812 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1813 		    l -= cmdlen, cmd += cmdlen) {
1814 			int match;
1815 
1816 			/*
1817 			 * check_body is a jump target used when we find a
1818 			 * CHECK_STATE, and need to jump to the body of
1819 			 * the target rule.
1820 			 */
1821 
1822 /* check_body: */
1823 			cmdlen = F_LEN(cmd);
1824 			/*
1825 			 * An OR block (insn_1 || .. || insn_n) has the
1826 			 * F_OR bit set in all but the last instruction.
1827 			 * The first match will set "skip_or", and cause
1828 			 * the following instructions to be skipped until
1829 			 * past the one with the F_OR bit clear.
1830 			 */
1831 			if (skip_or) {		/* skip this instruction */
1832 				if ((cmd->len & F_OR) == 0)
1833 					skip_or = 0;	/* next one is good */
1834 				continue;
1835 			}
1836 			match = 0; /* set to 1 if we succeed */
1837 
1838 			switch (cmd->opcode) {
1839 			/*
1840 			 * The first set of opcodes compares the packet's
1841 			 * fields with some pattern, setting 'match' if a
1842 			 * match is found. At the end of the loop there is
1843 			 * logic to deal with F_NOT and F_OR flags associated
1844 			 * with the opcode.
1845 			 */
1846 			case O_NOP:
1847 				match = 1;
1848 				break;
1849 
1850 			case O_FORWARD_MAC:
1851 				printf("ipfw: opcode %d unimplemented\n",
1852 				    cmd->opcode);
1853 				break;
1854 
1855 			case O_GID:
1856 			case O_UID:
1857 			case O_JAIL:
1858 				/*
1859 				 * We only check offset == 0 && proto != 0,
1860 				 * as this ensures that we have a
1861 				 * packet with the ports info.
1862 				 */
1863 				if (offset != 0)
1864 					break;
1865 				if (proto == IPPROTO_TCP ||
1866 				    proto == IPPROTO_UDP ||
1867 				    proto == IPPROTO_UDPLITE)
1868 					match = check_uidgid(
1869 						    (ipfw_insn_u32 *)cmd,
1870 						    args, &ucred_lookup,
1871 #ifdef __FreeBSD__
1872 						    &ucred_cache);
1873 #else
1874 						    (void *)&ucred_cache);
1875 #endif
1876 				break;
1877 
1878 			case O_RECV:
1879 				match = iface_match(iif, (ipfw_insn_if *)cmd,
1880 				    chain, &tablearg);
1881 				break;
1882 
1883 			case O_XMIT:
1884 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1885 				    chain, &tablearg);
1886 				break;
1887 
1888 			case O_VIA:
1889 				match = iface_match(args->ifp,
1890 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1891 				break;
1892 
1893 			case O_MACADDR2:
1894 				if (args->flags & IPFW_ARGS_ETHER) {
1895 					u_int32_t *want = (u_int32_t *)
1896 						((ipfw_insn_mac *)cmd)->addr;
1897 					u_int32_t *mask = (u_int32_t *)
1898 						((ipfw_insn_mac *)cmd)->mask;
1899 					u_int32_t *hdr = (u_int32_t *)eh;
1900 
1901 					match =
1902 					    ( want[0] == (hdr[0] & mask[0]) &&
1903 					      want[1] == (hdr[1] & mask[1]) &&
1904 					      want[2] == (hdr[2] & mask[2]) );
1905 				}
1906 				break;
1907 
1908 			case O_MAC_TYPE:
1909 				if (args->flags & IPFW_ARGS_ETHER) {
1910 					u_int16_t *p =
1911 					    ((ipfw_insn_u16 *)cmd)->ports;
1912 					int i;
1913 
1914 					for (i = cmdlen - 1; !match && i>0;
1915 					    i--, p += 2)
1916 						match =
1917 						    (ntohs(eh->ether_type) >=
1918 						    p[0] &&
1919 						    ntohs(eh->ether_type) <=
1920 						    p[1]);
1921 				}
1922 				break;
1923 
1924 			case O_FRAG:
1925 				match = (offset != 0);
1926 				break;
1927 
1928 			case O_IN:	/* "out" is "not in" */
1929 				match = (oif == NULL);
1930 				break;
1931 
1932 			case O_LAYER2:
1933 				match = (args->flags & IPFW_ARGS_ETHER);
1934 				break;
1935 
1936 			case O_DIVERTED:
1937 				if ((args->flags & IPFW_ARGS_REF) == 0)
1938 					break;
1939 				/*
1940 				 * For diverted packets, args->rule.info
1941 				 * contains the divert port (in host format)
1942 				 * reason and direction.
1943 				 */
1944 				match = ((args->rule.info & IPFW_IS_MASK) ==
1945 				    IPFW_IS_DIVERT) && (
1946 				    ((args->rule.info & IPFW_INFO_IN) ?
1947 					1: 2) & cmd->arg1);
1948 				break;
1949 
1950 			case O_PROTO:
1951 				/*
1952 				 * We do not allow an arg of 0 so the
1953 				 * check of "proto" only suffices.
1954 				 */
1955 				match = (proto == cmd->arg1);
1956 				break;
1957 
1958 			case O_IP_SRC:
1959 				match = is_ipv4 &&
1960 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1961 				    src_ip.s_addr);
1962 				break;
1963 
1964 			case O_IP_DST_LOOKUP:
1965 			{
1966 				void *pkey;
1967 				uint32_t vidx, key;
1968 				uint16_t keylen;
1969 
1970 				if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1971 					/* Determine lookup key type */
1972 					vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1973 					if (vidx != 4 /* uid */ &&
1974 					    vidx != 5 /* jail */ &&
1975 					    is_ipv6 == 0 && is_ipv4 == 0)
1976 						break;
1977 					/* Determine key length */
1978 					if (vidx == 0 /* dst-ip */ ||
1979 					    vidx == 1 /* src-ip */)
1980 						keylen = is_ipv6 ?
1981 						    sizeof(struct in6_addr):
1982 						    sizeof(in_addr_t);
1983 					else {
1984 						keylen = sizeof(key);
1985 						pkey = &key;
1986 					}
1987 					if (vidx == 0 /* dst-ip */)
1988 						pkey = is_ipv4 ? (void *)&dst_ip:
1989 						    (void *)&args->f_id.dst_ip6;
1990 					else if (vidx == 1 /* src-ip */)
1991 						pkey = is_ipv4 ? (void *)&src_ip:
1992 						    (void *)&args->f_id.src_ip6;
1993 					else if (vidx == 6 /* dscp */) {
1994 						if (is_ipv4)
1995 							key = ip->ip_tos >> 2;
1996 						else {
1997 							key = args->f_id.flow_id6;
1998 							key = (key & 0x0f) << 2 |
1999 							    (key & 0xf000) >> 14;
2000 						}
2001 						key &= 0x3f;
2002 					} else if (vidx == 2 /* dst-port */ ||
2003 					    vidx == 3 /* src-port */) {
2004 						/* Skip fragments */
2005 						if (offset != 0)
2006 							break;
2007 						/* Skip proto without ports */
2008 						if (proto != IPPROTO_TCP &&
2009 						    proto != IPPROTO_UDP &&
2010 						    proto != IPPROTO_UDPLITE &&
2011 						    proto != IPPROTO_SCTP)
2012 							break;
2013 						if (vidx == 2 /* dst-port */)
2014 							key = dst_port;
2015 						else
2016 							key = src_port;
2017 					}
2018 #ifndef USERSPACE
2019 					else if (vidx == 4 /* uid */ ||
2020 					    vidx == 5 /* jail */) {
2021 						check_uidgid(
2022 						    (ipfw_insn_u32 *)cmd,
2023 						    args, &ucred_lookup,
2024 #ifdef __FreeBSD__
2025 						    &ucred_cache);
2026 						if (vidx == 4 /* uid */)
2027 							key = ucred_cache->cr_uid;
2028 						else if (vidx == 5 /* jail */)
2029 							key = ucred_cache->cr_prison->pr_id;
2030 #else /* !__FreeBSD__ */
2031 						    (void *)&ucred_cache);
2032 						if (vidx == 4 /* uid */)
2033 							key = ucred_cache.uid;
2034 						else if (vidx == 5 /* jail */)
2035 							key = ucred_cache.xid;
2036 #endif /* !__FreeBSD__ */
2037 					}
2038 #endif /* !USERSPACE */
2039 					else
2040 						break;
2041 					match = ipfw_lookup_table(chain,
2042 					    cmd->arg1, keylen, pkey, &vidx);
2043 					if (!match)
2044 						break;
2045 					tablearg = vidx;
2046 					break;
2047 				}
2048 				/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2049 				/* FALLTHROUGH */
2050 			}
2051 			case O_IP_SRC_LOOKUP:
2052 			{
2053 				void *pkey;
2054 				uint32_t vidx;
2055 				uint16_t keylen;
2056 
2057 				if (is_ipv4) {
2058 					keylen = sizeof(in_addr_t);
2059 					if (cmd->opcode == O_IP_DST_LOOKUP)
2060 						pkey = &dst_ip;
2061 					else
2062 						pkey = &src_ip;
2063 				} else if (is_ipv6) {
2064 					keylen = sizeof(struct in6_addr);
2065 					if (cmd->opcode == O_IP_DST_LOOKUP)
2066 						pkey = &args->f_id.dst_ip6;
2067 					else
2068 						pkey = &args->f_id.src_ip6;
2069 				} else
2070 					break;
2071 				match = ipfw_lookup_table(chain, cmd->arg1,
2072 				    keylen, pkey, &vidx);
2073 				if (!match)
2074 					break;
2075 				if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2076 					match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2077 					    TARG_VAL(chain, vidx, tag);
2078 					if (!match)
2079 						break;
2080 				}
2081 				tablearg = vidx;
2082 				break;
2083 			}
2084 
2085 			case O_IP_FLOW_LOOKUP:
2086 				{
2087 					uint32_t v = 0;
2088 					match = ipfw_lookup_table(chain,
2089 					    cmd->arg1, 0, &args->f_id, &v);
2090 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2091 						match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2092 						    TARG_VAL(chain, v, tag);
2093 					if (match)
2094 						tablearg = v;
2095 				}
2096 				break;
2097 			case O_IP_SRC_MASK:
2098 			case O_IP_DST_MASK:
2099 				if (is_ipv4) {
2100 				    uint32_t a =
2101 					(cmd->opcode == O_IP_DST_MASK) ?
2102 					    dst_ip.s_addr : src_ip.s_addr;
2103 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2104 				    int i = cmdlen-1;
2105 
2106 				    for (; !match && i>0; i-= 2, p+= 2)
2107 					match = (p[0] == (a & p[1]));
2108 				}
2109 				break;
2110 
2111 			case O_IP_SRC_ME:
2112 				if (is_ipv4) {
2113 					match = in_localip(src_ip);
2114 					break;
2115 				}
2116 #ifdef INET6
2117 				/* FALLTHROUGH */
2118 			case O_IP6_SRC_ME:
2119 				match = is_ipv6 &&
2120 				    ipfw_localip6(&args->f_id.src_ip6);
2121 #endif
2122 				break;
2123 
2124 			case O_IP_DST_SET:
2125 			case O_IP_SRC_SET:
2126 				if (is_ipv4) {
2127 					u_int32_t *d = (u_int32_t *)(cmd+1);
2128 					u_int32_t addr =
2129 					    cmd->opcode == O_IP_DST_SET ?
2130 						args->f_id.dst_ip :
2131 						args->f_id.src_ip;
2132 
2133 					    if (addr < d[0])
2134 						    break;
2135 					    addr -= d[0]; /* subtract base */
2136 					    match = (addr < cmd->arg1) &&
2137 						( d[ 1 + (addr>>5)] &
2138 						  (1<<(addr & 0x1f)) );
2139 				}
2140 				break;
2141 
2142 			case O_IP_DST:
2143 				match = is_ipv4 &&
2144 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2145 				    dst_ip.s_addr);
2146 				break;
2147 
2148 			case O_IP_DST_ME:
2149 				if (is_ipv4) {
2150 					match = in_localip(dst_ip);
2151 					break;
2152 				}
2153 #ifdef INET6
2154 				/* FALLTHROUGH */
2155 			case O_IP6_DST_ME:
2156 				match = is_ipv6 &&
2157 				    ipfw_localip6(&args->f_id.dst_ip6);
2158 #endif
2159 				break;
2160 
2161 
2162 			case O_IP_SRCPORT:
2163 			case O_IP_DSTPORT:
2164 				/*
2165 				 * offset == 0 && proto != 0 is enough
2166 				 * to guarantee that we have a
2167 				 * packet with port info.
2168 				 */
2169 				if ((proto == IPPROTO_UDP ||
2170 				    proto == IPPROTO_UDPLITE ||
2171 				    proto == IPPROTO_TCP ||
2172 				    proto == IPPROTO_SCTP) && offset == 0) {
2173 					u_int16_t x =
2174 					    (cmd->opcode == O_IP_SRCPORT) ?
2175 						src_port : dst_port ;
2176 					u_int16_t *p =
2177 					    ((ipfw_insn_u16 *)cmd)->ports;
2178 					int i;
2179 
2180 					for (i = cmdlen - 1; !match && i>0;
2181 					    i--, p += 2)
2182 						match = (x>=p[0] && x<=p[1]);
2183 				}
2184 				break;
2185 
2186 			case O_ICMPTYPE:
2187 				match = (offset == 0 && proto==IPPROTO_ICMP &&
2188 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2189 				break;
2190 
2191 #ifdef INET6
2192 			case O_ICMP6TYPE:
2193 				match = is_ipv6 && offset == 0 &&
2194 				    proto==IPPROTO_ICMPV6 &&
2195 				    icmp6type_match(
2196 					ICMP6(ulp)->icmp6_type,
2197 					(ipfw_insn_u32 *)cmd);
2198 				break;
2199 #endif /* INET6 */
2200 
2201 			case O_IPOPT:
2202 				match = (is_ipv4 &&
2203 				    ipopts_match(ip, cmd) );
2204 				break;
2205 
2206 			case O_IPVER:
2207 				match = (is_ipv4 &&
2208 				    cmd->arg1 == ip->ip_v);
2209 				break;
2210 
2211 			case O_IPID:
2212 			case O_IPTTL:
2213 				if (!is_ipv4)
2214 					break;
2215 			case O_IPLEN:
2216 				{	/* only for IP packets */
2217 				    uint16_t x;
2218 				    uint16_t *p;
2219 				    int i;
2220 
2221 				    if (cmd->opcode == O_IPLEN)
2222 					x = iplen;
2223 				    else if (cmd->opcode == O_IPTTL)
2224 					x = ip->ip_ttl;
2225 				    else /* must be IPID */
2226 					x = ntohs(ip->ip_id);
2227 				    if (cmdlen == 1) {
2228 					match = (cmd->arg1 == x);
2229 					break;
2230 				    }
2231 				    /* otherwise we have ranges */
2232 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2233 				    i = cmdlen - 1;
2234 				    for (; !match && i>0; i--, p += 2)
2235 					match = (x >= p[0] && x <= p[1]);
2236 				}
2237 				break;
2238 
2239 			case O_IPPRECEDENCE:
2240 				match = (is_ipv4 &&
2241 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2242 				break;
2243 
2244 			case O_IPTOS:
2245 				match = (is_ipv4 &&
2246 				    flags_match(cmd, ip->ip_tos));
2247 				break;
2248 
2249 			case O_DSCP:
2250 			    {
2251 				uint32_t *p;
2252 				uint16_t x;
2253 
2254 				p = ((ipfw_insn_u32 *)cmd)->d;
2255 
2256 				if (is_ipv4)
2257 					x = ip->ip_tos >> 2;
2258 				else if (is_ipv6) {
2259 					uint8_t *v;
2260 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2261 					x = (*v & 0x0F) << 2;
2262 					v++;
2263 					x |= *v >> 6;
2264 				} else
2265 					break;
2266 
2267 				/* DSCP bitmask is stored as low_u32 high_u32 */
2268 				if (x >= 32)
2269 					match = *(p + 1) & (1 << (x - 32));
2270 				else
2271 					match = *p & (1 << x);
2272 			    }
2273 				break;
2274 
2275 			case O_TCPDATALEN:
2276 				if (proto == IPPROTO_TCP && offset == 0) {
2277 				    struct tcphdr *tcp;
2278 				    uint16_t x;
2279 				    uint16_t *p;
2280 				    int i;
2281 #ifdef INET6
2282 				    if (is_ipv6) {
2283 					    struct ip6_hdr *ip6;
2284 
2285 					    ip6 = (struct ip6_hdr *)ip;
2286 					    if (ip6->ip6_plen == 0) {
2287 						    /*
2288 						     * Jumbo payload is not
2289 						     * supported by this
2290 						     * opcode.
2291 						     */
2292 						    break;
2293 					    }
2294 					    x = iplen - hlen;
2295 				    } else
2296 #endif /* INET6 */
2297 					    x = iplen - (ip->ip_hl << 2);
2298 				    tcp = TCP(ulp);
2299 				    x -= tcp->th_off << 2;
2300 				    if (cmdlen == 1) {
2301 					match = (cmd->arg1 == x);
2302 					break;
2303 				    }
2304 				    /* otherwise we have ranges */
2305 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2306 				    i = cmdlen - 1;
2307 				    for (; !match && i>0; i--, p += 2)
2308 					match = (x >= p[0] && x <= p[1]);
2309 				}
2310 				break;
2311 
2312 			case O_TCPFLAGS:
2313 				match = (proto == IPPROTO_TCP && offset == 0 &&
2314 				    flags_match(cmd, TCP(ulp)->th_flags));
2315 				break;
2316 
2317 			case O_TCPOPTS:
2318 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
2319 					PULLUP_LEN_LOCKED(hlen, ulp,
2320 					    (TCP(ulp)->th_off << 2));
2321 					match = tcpopts_match(TCP(ulp), cmd);
2322 				}
2323 				break;
2324 
2325 			case O_TCPSEQ:
2326 				match = (proto == IPPROTO_TCP && offset == 0 &&
2327 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2328 					TCP(ulp)->th_seq);
2329 				break;
2330 
2331 			case O_TCPACK:
2332 				match = (proto == IPPROTO_TCP && offset == 0 &&
2333 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2334 					TCP(ulp)->th_ack);
2335 				break;
2336 
2337 			case O_TCPMSS:
2338 				if (proto == IPPROTO_TCP &&
2339 				    (args->f_id._flags & TH_SYN) != 0 &&
2340 				    ulp != NULL) {
2341 					uint16_t mss, *p;
2342 					int i;
2343 
2344 					PULLUP_LEN_LOCKED(hlen, ulp,
2345 					    (TCP(ulp)->th_off << 2));
2346 					if ((tcpopts_parse(TCP(ulp), &mss) &
2347 					    IP_FW_TCPOPT_MSS) == 0)
2348 						break;
2349 					if (cmdlen == 1) {
2350 						match = (cmd->arg1 == mss);
2351 						break;
2352 					}
2353 					/* Otherwise we have ranges. */
2354 					p = ((ipfw_insn_u16 *)cmd)->ports;
2355 					i = cmdlen - 1;
2356 					for (; !match && i > 0; i--, p += 2)
2357 						match = (mss >= p[0] &&
2358 						    mss <= p[1]);
2359 				}
2360 				break;
2361 
2362 			case O_TCPWIN:
2363 				if (proto == IPPROTO_TCP && offset == 0) {
2364 				    uint16_t x;
2365 				    uint16_t *p;
2366 				    int i;
2367 
2368 				    x = ntohs(TCP(ulp)->th_win);
2369 				    if (cmdlen == 1) {
2370 					match = (cmd->arg1 == x);
2371 					break;
2372 				    }
2373 				    /* Otherwise we have ranges. */
2374 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2375 				    i = cmdlen - 1;
2376 				    for (; !match && i > 0; i--, p += 2)
2377 					match = (x >= p[0] && x <= p[1]);
2378 				}
2379 				break;
2380 
2381 			case O_ESTAB:
2382 				/* reject packets which have SYN only */
2383 				/* XXX should i also check for TH_ACK ? */
2384 				match = (proto == IPPROTO_TCP && offset == 0 &&
2385 				    (TCP(ulp)->th_flags &
2386 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2387 				break;
2388 
2389 			case O_ALTQ: {
2390 				struct pf_mtag *at;
2391 				struct m_tag *mtag;
2392 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2393 
2394 				/*
2395 				 * ALTQ uses mbuf tags from another
2396 				 * packet filtering system - pf(4).
2397 				 * We allocate a tag in its format
2398 				 * and fill it in, pretending to be pf(4).
2399 				 */
2400 				match = 1;
2401 				at = pf_find_mtag(m);
2402 				if (at != NULL && at->qid != 0)
2403 					break;
2404 				mtag = m_tag_get(PACKET_TAG_PF,
2405 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2406 				if (mtag == NULL) {
2407 					/*
2408 					 * Let the packet fall back to the
2409 					 * default ALTQ.
2410 					 */
2411 					break;
2412 				}
2413 				m_tag_prepend(m, mtag);
2414 				at = (struct pf_mtag *)(mtag + 1);
2415 				at->qid = altq->qid;
2416 				at->hdr = ip;
2417 				break;
2418 			}
2419 
2420 			case O_LOG:
2421 				ipfw_log(chain, f, hlen, args,
2422 				    offset | ip6f_mf, tablearg, ip);
2423 				match = 1;
2424 				break;
2425 
2426 			case O_PROB:
2427 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2428 				break;
2429 
2430 			case O_VERREVPATH:
2431 				/* Outgoing packets automatically pass/match */
2432 				match = (args->flags & IPFW_ARGS_OUT ||
2433 				    (
2434 #ifdef INET6
2435 				    is_ipv6 ?
2436 					verify_path6(&(args->f_id.src_ip6),
2437 					    iif, args->f_id.fib) :
2438 #endif
2439 				    verify_path(src_ip, iif, args->f_id.fib)));
2440 				break;
2441 
2442 			case O_VERSRCREACH:
2443 				/* Outgoing packets automatically pass/match */
2444 				match = (hlen > 0 && ((oif != NULL) || (
2445 #ifdef INET6
2446 				    is_ipv6 ?
2447 				        verify_path6(&(args->f_id.src_ip6),
2448 				            NULL, args->f_id.fib) :
2449 #endif
2450 				    verify_path(src_ip, NULL, args->f_id.fib))));
2451 				break;
2452 
2453 			case O_ANTISPOOF:
2454 				/* Outgoing packets automatically pass/match */
2455 				if (oif == NULL && hlen > 0 &&
2456 				    (  (is_ipv4 && in_localaddr(src_ip))
2457 #ifdef INET6
2458 				    || (is_ipv6 &&
2459 				        in6_localaddr(&(args->f_id.src_ip6)))
2460 #endif
2461 				    ))
2462 					match =
2463 #ifdef INET6
2464 					    is_ipv6 ? verify_path6(
2465 					        &(args->f_id.src_ip6), iif,
2466 						args->f_id.fib) :
2467 #endif
2468 					    verify_path(src_ip, iif,
2469 					        args->f_id.fib);
2470 				else
2471 					match = 1;
2472 				break;
2473 
2474 			case O_IPSEC:
2475 				match = (m_tag_find(m,
2476 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2477 				/* otherwise no match */
2478 				break;
2479 
2480 #ifdef INET6
2481 			case O_IP6_SRC:
2482 				match = is_ipv6 &&
2483 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2484 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2485 				break;
2486 
2487 			case O_IP6_DST:
2488 				match = is_ipv6 &&
2489 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2490 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2491 				break;
2492 			case O_IP6_SRC_MASK:
2493 			case O_IP6_DST_MASK:
2494 				if (is_ipv6) {
2495 					int i = cmdlen - 1;
2496 					struct in6_addr p;
2497 					struct in6_addr *d =
2498 					    &((ipfw_insn_ip6 *)cmd)->addr6;
2499 
2500 					for (; !match && i > 0; d += 2,
2501 					    i -= F_INSN_SIZE(struct in6_addr)
2502 					    * 2) {
2503 						p = (cmd->opcode ==
2504 						    O_IP6_SRC_MASK) ?
2505 						    args->f_id.src_ip6:
2506 						    args->f_id.dst_ip6;
2507 						APPLY_MASK(&p, &d[1]);
2508 						match =
2509 						    IN6_ARE_ADDR_EQUAL(&d[0],
2510 						    &p);
2511 					}
2512 				}
2513 				break;
2514 
2515 			case O_FLOW6ID:
2516 				match = is_ipv6 &&
2517 				    flow6id_match(args->f_id.flow_id6,
2518 				    (ipfw_insn_u32 *) cmd);
2519 				break;
2520 
2521 			case O_EXT_HDR:
2522 				match = is_ipv6 &&
2523 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
2524 				break;
2525 
2526 			case O_IP6:
2527 				match = is_ipv6;
2528 				break;
2529 #endif
2530 
2531 			case O_IP4:
2532 				match = is_ipv4;
2533 				break;
2534 
2535 			case O_TAG: {
2536 				struct m_tag *mtag;
2537 				uint32_t tag = TARG(cmd->arg1, tag);
2538 
2539 				/* Packet is already tagged with this tag? */
2540 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2541 
2542 				/* We have `untag' action when F_NOT flag is
2543 				 * present. And we must remove this mtag from
2544 				 * mbuf and reset `match' to zero (`match' will
2545 				 * be inversed later).
2546 				 * Otherwise we should allocate new mtag and
2547 				 * push it into mbuf.
2548 				 */
2549 				if (cmd->len & F_NOT) { /* `untag' action */
2550 					if (mtag != NULL)
2551 						m_tag_delete(m, mtag);
2552 					match = 0;
2553 				} else {
2554 					if (mtag == NULL) {
2555 						mtag = m_tag_alloc( MTAG_IPFW,
2556 						    tag, 0, M_NOWAIT);
2557 						if (mtag != NULL)
2558 							m_tag_prepend(m, mtag);
2559 					}
2560 					match = 1;
2561 				}
2562 				break;
2563 			}
2564 
2565 			case O_FIB: /* try match the specified fib */
2566 				if (args->f_id.fib == cmd->arg1)
2567 					match = 1;
2568 				break;
2569 
2570 			case O_SOCKARG:	{
2571 #ifndef USERSPACE	/* not supported in userspace */
2572 				struct inpcb *inp = args->inp;
2573 				struct inpcbinfo *pi;
2574 
2575 				if (is_ipv6) /* XXX can we remove this ? */
2576 					break;
2577 
2578 				if (proto == IPPROTO_TCP)
2579 					pi = &V_tcbinfo;
2580 				else if (proto == IPPROTO_UDP)
2581 					pi = &V_udbinfo;
2582 				else if (proto == IPPROTO_UDPLITE)
2583 					pi = &V_ulitecbinfo;
2584 				else
2585 					break;
2586 
2587 				/*
2588 				 * XXXRW: so_user_cookie should almost
2589 				 * certainly be inp_user_cookie?
2590 				 */
2591 
2592 				/* For incoming packet, lookup up the
2593 				inpcb using the src/dest ip/port tuple */
2594 				if (inp == NULL) {
2595 					inp = in_pcblookup(pi,
2596 						src_ip, htons(src_port),
2597 						dst_ip, htons(dst_port),
2598 						INPLOOKUP_RLOCKPCB, NULL);
2599 					if (inp != NULL) {
2600 						tablearg =
2601 						    inp->inp_socket->so_user_cookie;
2602 						if (tablearg)
2603 							match = 1;
2604 						INP_RUNLOCK(inp);
2605 					}
2606 				} else {
2607 					if (inp->inp_socket) {
2608 						tablearg =
2609 						    inp->inp_socket->so_user_cookie;
2610 						if (tablearg)
2611 							match = 1;
2612 					}
2613 				}
2614 #endif /* !USERSPACE */
2615 				break;
2616 			}
2617 
2618 			case O_TAGGED: {
2619 				struct m_tag *mtag;
2620 				uint32_t tag = TARG(cmd->arg1, tag);
2621 
2622 				if (cmdlen == 1) {
2623 					match = m_tag_locate(m, MTAG_IPFW,
2624 					    tag, NULL) != NULL;
2625 					break;
2626 				}
2627 
2628 				/* we have ranges */
2629 				for (mtag = m_tag_first(m);
2630 				    mtag != NULL && !match;
2631 				    mtag = m_tag_next(m, mtag)) {
2632 					uint16_t *p;
2633 					int i;
2634 
2635 					if (mtag->m_tag_cookie != MTAG_IPFW)
2636 						continue;
2637 
2638 					p = ((ipfw_insn_u16 *)cmd)->ports;
2639 					i = cmdlen - 1;
2640 					for(; !match && i > 0; i--, p += 2)
2641 						match =
2642 						    mtag->m_tag_id >= p[0] &&
2643 						    mtag->m_tag_id <= p[1];
2644 				}
2645 				break;
2646 			}
2647 
2648 			/*
2649 			 * The second set of opcodes represents 'actions',
2650 			 * i.e. the terminal part of a rule once the packet
2651 			 * matches all previous patterns.
2652 			 * Typically there is only one action for each rule,
2653 			 * and the opcode is stored at the end of the rule
2654 			 * (but there are exceptions -- see below).
2655 			 *
2656 			 * In general, here we set retval and terminate the
2657 			 * outer loop (would be a 'break 3' in some language,
2658 			 * but we need to set l=0, done=1)
2659 			 *
2660 			 * Exceptions:
2661 			 * O_COUNT and O_SKIPTO actions:
2662 			 *   instead of terminating, we jump to the next rule
2663 			 *   (setting l=0), or to the SKIPTO target (setting
2664 			 *   f/f_len, cmd and l as needed), respectively.
2665 			 *
2666 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2667 			 *   perform some action and set match = 1;
2668 			 *
2669 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2670 			 *   not real 'actions', and are stored right
2671 			 *   before the 'action' part of the rule (one
2672 			 *   exception is O_SKIP_ACTION which could be
2673 			 *   between these opcodes and 'action' one).
2674 			 *   These opcodes try to install an entry in the
2675 			 *   state tables; if successful, we continue with
2676 			 *   the next opcode (match=1; break;), otherwise
2677 			 *   the packet must be dropped (set retval,
2678 			 *   break loops with l=0, done=1)
2679 			 *
2680 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2681 			 *   cause a lookup of the state table, and a jump
2682 			 *   to the 'action' part of the parent rule
2683 			 *   if an entry is found, or
2684 			 *   (CHECK_STATE only) a jump to the next rule if
2685 			 *   the entry is not found.
2686 			 *   The result of the lookup is cached so that
2687 			 *   further instances of these opcodes become NOPs.
2688 			 *   The jump to the next rule is done by setting
2689 			 *   l=0, cmdlen=0.
2690 			 *
2691 			 * O_SKIP_ACTION: this opcode is not a real 'action'
2692 			 *  either, and is stored right before the 'action'
2693 			 *  part of the rule, right after the O_KEEP_STATE
2694 			 *  opcode. It causes match failure so the real
2695 			 *  'action' could be executed only if the rule
2696 			 *  is checked via dynamic rule from the state
2697 			 *  table, as in such case execution starts
2698 			 *  from the true 'action' opcode directly.
2699 			 *
2700 			 */
2701 			case O_LIMIT:
2702 			case O_KEEP_STATE:
2703 				if (ipfw_dyn_install_state(chain, f,
2704 				    (ipfw_insn_limit *)cmd, args, ulp,
2705 				    pktlen, &dyn_info, tablearg)) {
2706 					/* error or limit violation */
2707 					retval = IP_FW_DENY;
2708 					l = 0;	/* exit inner loop */
2709 					done = 1; /* exit outer loop */
2710 				}
2711 				match = 1;
2712 				break;
2713 
2714 			case O_PROBE_STATE:
2715 			case O_CHECK_STATE:
2716 				/*
2717 				 * dynamic rules are checked at the first
2718 				 * keep-state or check-state occurrence,
2719 				 * with the result being stored in dyn_info.
2720 				 * The compiler introduces a PROBE_STATE
2721 				 * instruction for us when we have a
2722 				 * KEEP_STATE (because PROBE_STATE needs
2723 				 * to be run first).
2724 				 */
2725 				if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2726 				    (q = ipfw_dyn_lookup_state(args, ulp,
2727 				    pktlen, cmd, &dyn_info)) != NULL) {
2728 					/*
2729 					 * Found dynamic entry, jump to the
2730 					 * 'action' part of the parent rule
2731 					 * by setting f, cmd, l and clearing
2732 					 * cmdlen.
2733 					 */
2734 					f = q;
2735 					f_pos = dyn_info.f_pos;
2736 					cmd = ACTION_PTR(f);
2737 					l = f->cmd_len - f->act_ofs;
2738 					cmdlen = 0;
2739 					match = 1;
2740 					break;
2741 				}
2742 				/*
2743 				 * Dynamic entry not found. If CHECK_STATE,
2744 				 * skip to next rule, if PROBE_STATE just
2745 				 * ignore and continue with next opcode.
2746 				 */
2747 				if (cmd->opcode == O_CHECK_STATE)
2748 					l = 0;	/* exit inner loop */
2749 				match = 1;
2750 				break;
2751 
2752 			case O_SKIP_ACTION:
2753 				match = 0;	/* skip to the next rule */
2754 				l = 0;		/* exit inner loop */
2755 				break;
2756 
2757 			case O_ACCEPT:
2758 				retval = 0;	/* accept */
2759 				l = 0;		/* exit inner loop */
2760 				done = 1;	/* exit outer loop */
2761 				break;
2762 
2763 			case O_PIPE:
2764 			case O_QUEUE:
2765 				set_match(args, f_pos, chain);
2766 				args->rule.info = TARG(cmd->arg1, pipe);
2767 				if (cmd->opcode == O_PIPE)
2768 					args->rule.info |= IPFW_IS_PIPE;
2769 				if (V_fw_one_pass)
2770 					args->rule.info |= IPFW_ONEPASS;
2771 				retval = IP_FW_DUMMYNET;
2772 				l = 0;          /* exit inner loop */
2773 				done = 1;       /* exit outer loop */
2774 				break;
2775 
2776 			case O_DIVERT:
2777 			case O_TEE:
2778 				if (args->flags & IPFW_ARGS_ETHER)
2779 					break;	/* not on layer 2 */
2780 				/* otherwise this is terminal */
2781 				l = 0;		/* exit inner loop */
2782 				done = 1;	/* exit outer loop */
2783 				retval = (cmd->opcode == O_DIVERT) ?
2784 					IP_FW_DIVERT : IP_FW_TEE;
2785 				set_match(args, f_pos, chain);
2786 				args->rule.info = TARG(cmd->arg1, divert);
2787 				break;
2788 
2789 			case O_COUNT:
2790 				IPFW_INC_RULE_COUNTER(f, pktlen);
2791 				l = 0;		/* exit inner loop */
2792 				break;
2793 
2794 			case O_SKIPTO:
2795 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2796 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2797 			    /*
2798 			     * Skip disabled rules, and re-enter
2799 			     * the inner loop with the correct
2800 			     * f_pos, f, l and cmd.
2801 			     * Also clear cmdlen and skip_or
2802 			     */
2803 			    for (; f_pos < chain->n_rules - 1 &&
2804 				    (V_set_disable &
2805 				     (1 << chain->map[f_pos]->set));
2806 				    f_pos++)
2807 				;
2808 			    /* Re-enter the inner loop at the skipto rule. */
2809 			    f = chain->map[f_pos];
2810 			    l = f->cmd_len;
2811 			    cmd = f->cmd;
2812 			    match = 1;
2813 			    cmdlen = 0;
2814 			    skip_or = 0;
2815 			    continue;
2816 			    break;	/* not reached */
2817 
2818 			case O_CALLRETURN: {
2819 				/*
2820 				 * Implementation of `subroutine' call/return,
2821 				 * in the stack carried in an mbuf tag. This
2822 				 * is different from `skipto' in that any call
2823 				 * address is possible (`skipto' must prevent
2824 				 * backward jumps to avoid endless loops).
2825 				 * We have `return' action when F_NOT flag is
2826 				 * present. The `m_tag_id' field is used as
2827 				 * stack pointer.
2828 				 */
2829 				struct m_tag *mtag;
2830 				uint16_t jmpto, *stack;
2831 
2832 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2833 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2834 				/*
2835 				 * Hand-rolled version of m_tag_locate() with
2836 				 * wildcard `type'.
2837 				 * If not already tagged, allocate new tag.
2838 				 */
2839 				mtag = m_tag_first(m);
2840 				while (mtag != NULL) {
2841 					if (mtag->m_tag_cookie ==
2842 					    MTAG_IPFW_CALL)
2843 						break;
2844 					mtag = m_tag_next(m, mtag);
2845 				}
2846 				if (mtag == NULL && IS_CALL) {
2847 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2848 					    IPFW_CALLSTACK_SIZE *
2849 					    sizeof(uint16_t), M_NOWAIT);
2850 					if (mtag != NULL)
2851 						m_tag_prepend(m, mtag);
2852 				}
2853 
2854 				/*
2855 				 * On error both `call' and `return' just
2856 				 * continue with next rule.
2857 				 */
2858 				if (IS_RETURN && (mtag == NULL ||
2859 				    mtag->m_tag_id == 0)) {
2860 					l = 0;		/* exit inner loop */
2861 					break;
2862 				}
2863 				if (IS_CALL && (mtag == NULL ||
2864 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2865 					printf("ipfw: call stack error, "
2866 					    "go to next rule\n");
2867 					l = 0;		/* exit inner loop */
2868 					break;
2869 				}
2870 
2871 				IPFW_INC_RULE_COUNTER(f, pktlen);
2872 				stack = (uint16_t *)(mtag + 1);
2873 
2874 				/*
2875 				 * The `call' action may use cached f_pos
2876 				 * (in f->next_rule), whose version is written
2877 				 * in f->next_rule.
2878 				 * The `return' action, however, doesn't have
2879 				 * fixed jump address in cmd->arg1 and can't use
2880 				 * cache.
2881 				 */
2882 				if (IS_CALL) {
2883 					stack[mtag->m_tag_id] = f->rulenum;
2884 					mtag->m_tag_id++;
2885 			    		f_pos = JUMP(chain, f, cmd->arg1,
2886 					    tablearg, 1);
2887 				} else {	/* `return' action */
2888 					mtag->m_tag_id--;
2889 					jmpto = stack[mtag->m_tag_id] + 1;
2890 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2891 				}
2892 
2893 				/*
2894 				 * Skip disabled rules, and re-enter
2895 				 * the inner loop with the correct
2896 				 * f_pos, f, l and cmd.
2897 				 * Also clear cmdlen and skip_or
2898 				 */
2899 				for (; f_pos < chain->n_rules - 1 &&
2900 				    (V_set_disable &
2901 				    (1 << chain->map[f_pos]->set)); f_pos++)
2902 					;
2903 				/* Re-enter the inner loop at the dest rule. */
2904 				f = chain->map[f_pos];
2905 				l = f->cmd_len;
2906 				cmd = f->cmd;
2907 				cmdlen = 0;
2908 				skip_or = 0;
2909 				continue;
2910 				break;	/* NOTREACHED */
2911 			}
2912 #undef IS_CALL
2913 #undef IS_RETURN
2914 
2915 			case O_REJECT:
2916 				/*
2917 				 * Drop the packet and send a reject notice
2918 				 * if the packet is not ICMP (or is an ICMP
2919 				 * query), and it is not multicast/broadcast.
2920 				 */
2921 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2922 				    (proto != IPPROTO_ICMP ||
2923 				     is_icmp_query(ICMP(ulp))) &&
2924 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2925 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2926 					send_reject(args, cmd->arg1, iplen, ip);
2927 					m = args->m;
2928 				}
2929 				/* FALLTHROUGH */
2930 #ifdef INET6
2931 			case O_UNREACH6:
2932 				if (hlen > 0 && is_ipv6 &&
2933 				    ((offset & IP6F_OFF_MASK) == 0) &&
2934 				    (proto != IPPROTO_ICMPV6 ||
2935 				     (is_icmp6_query(icmp6_type) == 1)) &&
2936 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2937 				    !IN6_IS_ADDR_MULTICAST(
2938 					&args->f_id.dst_ip6)) {
2939 					send_reject6(args,
2940 					    cmd->opcode == O_REJECT ?
2941 					    map_icmp_unreach(cmd->arg1):
2942 					    cmd->arg1, hlen,
2943 					    (struct ip6_hdr *)ip);
2944 					m = args->m;
2945 				}
2946 				/* FALLTHROUGH */
2947 #endif
2948 			case O_DENY:
2949 				retval = IP_FW_DENY;
2950 				l = 0;		/* exit inner loop */
2951 				done = 1;	/* exit outer loop */
2952 				break;
2953 
2954 			case O_FORWARD_IP:
2955 				if (args->flags & IPFW_ARGS_ETHER)
2956 					break;	/* not valid on layer2 pkts */
2957 				if (q != f ||
2958 				    dyn_info.direction == MATCH_FORWARD) {
2959 				    struct sockaddr_in *sa;
2960 
2961 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2962 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2963 #ifdef INET6
2964 					/*
2965 					 * We use O_FORWARD_IP opcode for
2966 					 * fwd rule with tablearg, but tables
2967 					 * now support IPv6 addresses. And
2968 					 * when we are inspecting IPv6 packet,
2969 					 * we can use nh6 field from
2970 					 * table_value as next_hop6 address.
2971 					 */
2972 					if (is_ipv6) {
2973 						struct ip_fw_nh6 *nh6;
2974 
2975 						args->flags |= IPFW_ARGS_NH6;
2976 						nh6 = &args->hopstore6;
2977 						nh6->sin6_addr = TARG_VAL(
2978 						    chain, tablearg, nh6);
2979 						nh6->sin6_port = sa->sin_port;
2980 						nh6->sin6_scope_id = TARG_VAL(
2981 						    chain, tablearg, zoneid);
2982 					} else
2983 #endif
2984 					{
2985 						args->flags |= IPFW_ARGS_NH4;
2986 						args->hopstore.sin_port =
2987 						    sa->sin_port;
2988 						sa = &args->hopstore;
2989 						sa->sin_family = AF_INET;
2990 						sa->sin_len = sizeof(*sa);
2991 						sa->sin_addr.s_addr = htonl(
2992 						    TARG_VAL(chain, tablearg,
2993 						    nh4));
2994 					}
2995 				    } else {
2996 					    args->flags |= IPFW_ARGS_NH4PTR;
2997 					    args->next_hop = sa;
2998 				    }
2999 				}
3000 				retval = IP_FW_PASS;
3001 				l = 0;          /* exit inner loop */
3002 				done = 1;       /* exit outer loop */
3003 				break;
3004 
3005 #ifdef INET6
3006 			case O_FORWARD_IP6:
3007 				if (args->flags & IPFW_ARGS_ETHER)
3008 					break;	/* not valid on layer2 pkts */
3009 				if (q != f ||
3010 				    dyn_info.direction == MATCH_FORWARD) {
3011 					struct sockaddr_in6 *sin6;
3012 
3013 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3014 					args->flags |= IPFW_ARGS_NH6PTR;
3015 					args->next_hop6 = sin6;
3016 				}
3017 				retval = IP_FW_PASS;
3018 				l = 0;		/* exit inner loop */
3019 				done = 1;	/* exit outer loop */
3020 				break;
3021 #endif
3022 
3023 			case O_NETGRAPH:
3024 			case O_NGTEE:
3025 				set_match(args, f_pos, chain);
3026 				args->rule.info = TARG(cmd->arg1, netgraph);
3027 				if (V_fw_one_pass)
3028 					args->rule.info |= IPFW_ONEPASS;
3029 				retval = (cmd->opcode == O_NETGRAPH) ?
3030 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
3031 				l = 0;          /* exit inner loop */
3032 				done = 1;       /* exit outer loop */
3033 				break;
3034 
3035 			case O_SETFIB: {
3036 				uint32_t fib;
3037 
3038 				IPFW_INC_RULE_COUNTER(f, pktlen);
3039 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
3040 				if (fib >= rt_numfibs)
3041 					fib = 0;
3042 				M_SETFIB(m, fib);
3043 				args->f_id.fib = fib; /* XXX */
3044 				l = 0;		/* exit inner loop */
3045 				break;
3046 		        }
3047 
3048 			case O_SETDSCP: {
3049 				uint16_t code;
3050 
3051 				code = TARG(cmd->arg1, dscp) & 0x3F;
3052 				l = 0;		/* exit inner loop */
3053 				if (is_ipv4) {
3054 					uint16_t old;
3055 
3056 					old = *(uint16_t *)ip;
3057 					ip->ip_tos = (code << 2) |
3058 					    (ip->ip_tos & 0x03);
3059 					ip->ip_sum = cksum_adjust(ip->ip_sum,
3060 					    old, *(uint16_t *)ip);
3061 				} else if (is_ipv6) {
3062 					uint8_t *v;
3063 
3064 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
3065 					*v = (*v & 0xF0) | (code >> 2);
3066 					v++;
3067 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
3068 				} else
3069 					break;
3070 
3071 				IPFW_INC_RULE_COUNTER(f, pktlen);
3072 				break;
3073 			}
3074 
3075 			case O_NAT:
3076 				l = 0;          /* exit inner loop */
3077 				done = 1;       /* exit outer loop */
3078 				/*
3079 				 * Ensure that we do not invoke NAT handler for
3080 				 * non IPv4 packets. Libalias expects only IPv4.
3081 				 */
3082 				if (!is_ipv4 || !IPFW_NAT_LOADED) {
3083 				    retval = IP_FW_DENY;
3084 				    break;
3085 				}
3086 
3087 				struct cfg_nat *t;
3088 				int nat_id;
3089 
3090 				args->rule.info = 0;
3091 				set_match(args, f_pos, chain);
3092 				/* Check if this is 'global' nat rule */
3093 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3094 					retval = ipfw_nat_ptr(args, NULL, m);
3095 					break;
3096 				}
3097 				t = ((ipfw_insn_nat *)cmd)->nat;
3098 				if (t == NULL) {
3099 					nat_id = TARG(cmd->arg1, nat);
3100 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3101 
3102 					if (t == NULL) {
3103 					    retval = IP_FW_DENY;
3104 					    break;
3105 					}
3106 					if (cmd->arg1 != IP_FW_TARG)
3107 					    ((ipfw_insn_nat *)cmd)->nat = t;
3108 				}
3109 				retval = ipfw_nat_ptr(args, t, m);
3110 				break;
3111 
3112 			case O_REASS: {
3113 				int ip_off;
3114 
3115 				l = 0;	/* in any case exit inner loop */
3116 				if (is_ipv6) /* IPv6 is not supported yet */
3117 					break;
3118 				IPFW_INC_RULE_COUNTER(f, pktlen);
3119 				ip_off = ntohs(ip->ip_off);
3120 
3121 				/* if not fragmented, go to next rule */
3122 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3123 				    break;
3124 
3125 				args->m = m = ip_reass(m);
3126 
3127 				/*
3128 				 * do IP header checksum fixup.
3129 				 */
3130 				if (m == NULL) { /* fragment got swallowed */
3131 				    retval = IP_FW_DENY;
3132 				} else { /* good, packet complete */
3133 				    int hlen;
3134 
3135 				    ip = mtod(m, struct ip *);
3136 				    hlen = ip->ip_hl << 2;
3137 				    ip->ip_sum = 0;
3138 				    if (hlen == sizeof(struct ip))
3139 					ip->ip_sum = in_cksum_hdr(ip);
3140 				    else
3141 					ip->ip_sum = in_cksum(m, hlen);
3142 				    retval = IP_FW_REASS;
3143 				    args->rule.info = 0;
3144 				    set_match(args, f_pos, chain);
3145 				}
3146 				done = 1;	/* exit outer loop */
3147 				break;
3148 			}
3149 			case O_EXTERNAL_ACTION:
3150 				l = 0; /* in any case exit inner loop */
3151 				retval = ipfw_run_eaction(chain, args,
3152 				    cmd, &done);
3153 				/*
3154 				 * If both @retval and @done are zero,
3155 				 * consider this as rule matching and
3156 				 * update counters.
3157 				 */
3158 				if (retval == 0 && done == 0) {
3159 					IPFW_INC_RULE_COUNTER(f, pktlen);
3160 					/*
3161 					 * Reset the result of the last
3162 					 * dynamic state lookup.
3163 					 * External action can change
3164 					 * @args content, and it may be
3165 					 * used for new state lookup later.
3166 					 */
3167 					DYN_INFO_INIT(&dyn_info);
3168 				}
3169 				break;
3170 
3171 			default:
3172 				panic("-- unknown opcode %d\n", cmd->opcode);
3173 			} /* end of switch() on opcodes */
3174 			/*
3175 			 * if we get here with l=0, then match is irrelevant.
3176 			 */
3177 
3178 			if (cmd->len & F_NOT)
3179 				match = !match;
3180 
3181 			if (match) {
3182 				if (cmd->len & F_OR)
3183 					skip_or = 1;
3184 			} else {
3185 				if (!(cmd->len & F_OR)) /* not an OR block, */
3186 					break;		/* try next rule    */
3187 			}
3188 
3189 		}	/* end of inner loop, scan opcodes */
3190 #undef PULLUP_LEN
3191 #undef PULLUP_LEN_LOCKED
3192 
3193 		if (done)
3194 			break;
3195 
3196 /* next_rule:; */	/* try next rule		*/
3197 
3198 	}		/* end of outer for, scan rules */
3199 
3200 	if (done) {
3201 		struct ip_fw *rule = chain->map[f_pos];
3202 		/* Update statistics */
3203 		IPFW_INC_RULE_COUNTER(rule, pktlen);
3204 	} else {
3205 		retval = IP_FW_DENY;
3206 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3207 	}
3208 	IPFW_PF_RUNLOCK(chain);
3209 #ifdef __FreeBSD__
3210 	if (ucred_cache != NULL)
3211 		crfree(ucred_cache);
3212 #endif
3213 	return (retval);
3214 
3215 pullup_failed:
3216 	if (V_fw_verbose)
3217 		printf("ipfw: pullup failed\n");
3218 	return (IP_FW_DENY);
3219 }
3220 
3221 /*
3222  * Set maximum number of tables that can be used in given VNET ipfw instance.
3223  */
3224 #ifdef SYSCTL_NODE
3225 static int
3226 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3227 {
3228 	int error;
3229 	unsigned int ntables;
3230 
3231 	ntables = V_fw_tables_max;
3232 
3233 	error = sysctl_handle_int(oidp, &ntables, 0, req);
3234 	/* Read operation or some error */
3235 	if ((error != 0) || (req->newptr == NULL))
3236 		return (error);
3237 
3238 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
3239 }
3240 
3241 /*
3242  * Switches table namespace between global and per-set.
3243  */
3244 static int
3245 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3246 {
3247 	int error;
3248 	unsigned int sets;
3249 
3250 	sets = V_fw_tables_sets;
3251 
3252 	error = sysctl_handle_int(oidp, &sets, 0, req);
3253 	/* Read operation or some error */
3254 	if ((error != 0) || (req->newptr == NULL))
3255 		return (error);
3256 
3257 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3258 }
3259 #endif
3260 
3261 /*
3262  * Module and VNET glue
3263  */
3264 
3265 /*
3266  * Stuff that must be initialised only on boot or module load
3267  */
3268 static int
3269 ipfw_init(void)
3270 {
3271 	int error = 0;
3272 
3273 	/*
3274  	 * Only print out this stuff the first time around,
3275 	 * when called from the sysinit code.
3276 	 */
3277 	printf("ipfw2 "
3278 #ifdef INET6
3279 		"(+ipv6) "
3280 #endif
3281 		"initialized, divert %s, nat %s, "
3282 		"default to %s, logging ",
3283 #ifdef IPDIVERT
3284 		"enabled",
3285 #else
3286 		"loadable",
3287 #endif
3288 #ifdef IPFIREWALL_NAT
3289 		"enabled",
3290 #else
3291 		"loadable",
3292 #endif
3293 		default_to_accept ? "accept" : "deny");
3294 
3295 	/*
3296 	 * Note: V_xxx variables can be accessed here but the vnet specific
3297 	 * initializer may not have been called yet for the VIMAGE case.
3298 	 * Tuneables will have been processed. We will print out values for
3299 	 * the default vnet.
3300 	 * XXX This should all be rationalized AFTER 8.0
3301 	 */
3302 	if (V_fw_verbose == 0)
3303 		printf("disabled\n");
3304 	else if (V_verbose_limit == 0)
3305 		printf("unlimited\n");
3306 	else
3307 		printf("limited to %d packets/entry by default\n",
3308 		    V_verbose_limit);
3309 
3310 	/* Check user-supplied table count for validness */
3311 	if (default_fw_tables > IPFW_TABLES_MAX)
3312 	  default_fw_tables = IPFW_TABLES_MAX;
3313 
3314 	ipfw_init_sopt_handler();
3315 	ipfw_init_obj_rewriter();
3316 	ipfw_iface_init();
3317 	return (error);
3318 }
3319 
3320 /*
3321  * Called for the removal of the last instance only on module unload.
3322  */
3323 static void
3324 ipfw_destroy(void)
3325 {
3326 
3327 	ipfw_iface_destroy();
3328 	ipfw_destroy_sopt_handler();
3329 	ipfw_destroy_obj_rewriter();
3330 	printf("IP firewall unloaded\n");
3331 }
3332 
3333 /*
3334  * Stuff that must be initialized for every instance
3335  * (including the first of course).
3336  */
3337 static int
3338 vnet_ipfw_init(const void *unused)
3339 {
3340 	int error, first;
3341 	struct ip_fw *rule = NULL;
3342 	struct ip_fw_chain *chain;
3343 
3344 	chain = &V_layer3_chain;
3345 
3346 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3347 
3348 	/* First set up some values that are compile time options */
3349 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
3350 	V_fw_deny_unknown_exthdrs = 1;
3351 #ifdef IPFIREWALL_VERBOSE
3352 	V_fw_verbose = 1;
3353 #endif
3354 #ifdef IPFIREWALL_VERBOSE_LIMIT
3355 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3356 #endif
3357 #ifdef IPFIREWALL_NAT
3358 	LIST_INIT(&chain->nat);
3359 #endif
3360 
3361 	/* Init shared services hash table */
3362 	ipfw_init_srv(chain);
3363 
3364 	ipfw_init_counters();
3365 	/* Set initial number of tables */
3366 	V_fw_tables_max = default_fw_tables;
3367 	error = ipfw_init_tables(chain, first);
3368 	if (error) {
3369 		printf("ipfw2: setting up tables failed\n");
3370 		free(chain->map, M_IPFW);
3371 		free(rule, M_IPFW);
3372 		return (ENOSPC);
3373 	}
3374 
3375 	IPFW_LOCK_INIT(chain);
3376 
3377 	/* fill and insert the default rule */
3378 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3379 	rule->flags |= IPFW_RULE_NOOPT;
3380 	rule->cmd_len = 1;
3381 	rule->cmd[0].len = 1;
3382 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3383 	chain->default_rule = rule;
3384 	ipfw_add_protected_rule(chain, rule, 0);
3385 
3386 	ipfw_dyn_init(chain);
3387 	ipfw_eaction_init(chain, first);
3388 #ifdef LINEAR_SKIPTO
3389 	ipfw_init_skipto_cache(chain);
3390 #endif
3391 	ipfw_bpf_init(first);
3392 
3393 	/* First set up some values that are compile time options */
3394 	V_ipfw_vnet_ready = 1;		/* Open for business */
3395 
3396 	/*
3397 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3398 	 * Even if the latter two fail we still keep the module alive
3399 	 * because the sockopt and layer2 paths are still useful.
3400 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3401 	 * so we can ignore the exact return value and just set a flag.
3402 	 *
3403 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3404 	 * changes in the underlying (per-vnet) variables trigger
3405 	 * immediate hook()/unhook() calls.
3406 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
3407 	 * is checked on each packet because there are no pfil hooks.
3408 	 */
3409 	V_ip_fw_ctl_ptr = ipfw_ctl3;
3410 	error = ipfw_attach_hooks();
3411 	return (error);
3412 }
3413 
3414 /*
3415  * Called for the removal of each instance.
3416  */
3417 static int
3418 vnet_ipfw_uninit(const void *unused)
3419 {
3420 	struct ip_fw *reap;
3421 	struct ip_fw_chain *chain = &V_layer3_chain;
3422 	int i, last;
3423 
3424 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3425 	/*
3426 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
3427 	 * Then grab, release and grab again the WLOCK so we make
3428 	 * sure the update is propagated and nobody will be in.
3429 	 */
3430 	ipfw_detach_hooks();
3431 	V_ip_fw_ctl_ptr = NULL;
3432 
3433 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3434 
3435 	IPFW_UH_WLOCK(chain);
3436 	IPFW_UH_WUNLOCK(chain);
3437 
3438 	ipfw_dyn_uninit(0);	/* run the callout_drain */
3439 
3440 	IPFW_UH_WLOCK(chain);
3441 
3442 	reap = NULL;
3443 	IPFW_WLOCK(chain);
3444 	for (i = 0; i < chain->n_rules; i++)
3445 		ipfw_reap_add(chain, &reap, chain->map[i]);
3446 	free(chain->map, M_IPFW);
3447 #ifdef LINEAR_SKIPTO
3448 	ipfw_destroy_skipto_cache(chain);
3449 #endif
3450 	IPFW_WUNLOCK(chain);
3451 	IPFW_UH_WUNLOCK(chain);
3452 	ipfw_destroy_tables(chain, last);
3453 	ipfw_eaction_uninit(chain, last);
3454 	if (reap != NULL)
3455 		ipfw_reap_rules(reap);
3456 	vnet_ipfw_iface_destroy(chain);
3457 	ipfw_destroy_srv(chain);
3458 	IPFW_LOCK_DESTROY(chain);
3459 	ipfw_dyn_uninit(1);	/* free the remaining parts */
3460 	ipfw_destroy_counters();
3461 	ipfw_bpf_uninit(last);
3462 	return (0);
3463 }
3464 
3465 /*
3466  * Module event handler.
3467  * In general we have the choice of handling most of these events by the
3468  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3469  * use the SYSINIT handlers as they are more capable of expressing the
3470  * flow of control during module and vnet operations, so this is just
3471  * a skeleton. Note there is no SYSINIT equivalent of the module
3472  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3473  */
3474 static int
3475 ipfw_modevent(module_t mod, int type, void *unused)
3476 {
3477 	int err = 0;
3478 
3479 	switch (type) {
3480 	case MOD_LOAD:
3481 		/* Called once at module load or
3482 	 	 * system boot if compiled in. */
3483 		break;
3484 	case MOD_QUIESCE:
3485 		/* Called before unload. May veto unloading. */
3486 		break;
3487 	case MOD_UNLOAD:
3488 		/* Called during unload. */
3489 		break;
3490 	case MOD_SHUTDOWN:
3491 		/* Called during system shutdown. */
3492 		break;
3493 	default:
3494 		err = EOPNOTSUPP;
3495 		break;
3496 	}
3497 	return err;
3498 }
3499 
3500 static moduledata_t ipfwmod = {
3501 	"ipfw",
3502 	ipfw_modevent,
3503 	0
3504 };
3505 
3506 /* Define startup order. */
3507 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
3508 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
3509 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
3510 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
3511 
3512 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3513 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3514 MODULE_VERSION(ipfw, 3);
3515 /* should declare some dependencies here */
3516 
3517 /*
3518  * Starting up. Done in order after ipfwmod() has been called.
3519  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3520  */
3521 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3522 	    ipfw_init, NULL);
3523 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3524 	    vnet_ipfw_init, NULL);
3525 
3526 /*
3527  * Closing up shop. These are done in REVERSE ORDER, but still
3528  * after ipfwmod() has been called. Not called on reboot.
3529  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3530  * or when the module is unloaded.
3531  */
3532 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3533 	    ipfw_destroy, NULL);
3534 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3535 	    vnet_ipfw_uninit, NULL);
3536 /* end of file */
3537