xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 8b238f4126d32df3e70056bc32536b7248ebffa0)
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 		NET_EPOCH_ASSERT();
412 
413 		CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
414 			if (ia->ifa_addr->sa_family != AF_INET)
415 				continue;
416 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
417 			    (ia->ifa_addr))->sin_addr.s_addr)
418 				return (1);	/* match */
419 		}
420 #endif /* __FreeBSD__ */
421 	}
422 	return(0);	/* no match, fail ... */
423 }
424 
425 /*
426  * The verify_path function checks if a route to the src exists and
427  * if it is reachable via ifp (when provided).
428  *
429  * The 'verrevpath' option checks that the interface that an IP packet
430  * arrives on is the same interface that traffic destined for the
431  * packet's source address would be routed out of.
432  * The 'versrcreach' option just checks that the source address is
433  * reachable via any route (except default) in the routing table.
434  * These two are a measure to block forged packets. This is also
435  * commonly known as "anti-spoofing" or Unicast Reverse Path
436  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
437  * is purposely reminiscent of the Cisco IOS command,
438  *
439  *   ip verify unicast reverse-path
440  *   ip verify unicast source reachable-via any
441  *
442  * which implements the same functionality. But note that the syntax
443  * is misleading, and the check may be performed on all IP packets
444  * whether unicast, multicast, or broadcast.
445  */
446 static int
447 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
448 {
449 #if defined(USERSPACE) || !defined(__FreeBSD__)
450 	return 0;
451 #else
452 	struct nhop4_basic nh4;
453 
454 	if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
455 		return (0);
456 
457 	/*
458 	 * If ifp is provided, check for equality with rtentry.
459 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
460 	 * in order to pass packets injected back by if_simloop():
461 	 * routing entry (via lo0) for our own address
462 	 * may exist, so we need to handle routing assymetry.
463 	 */
464 	if (ifp != NULL && ifp != nh4.nh_ifp)
465 		return (0);
466 
467 	/* if no ifp provided, check if rtentry is not default route */
468 	if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
469 		return (0);
470 
471 	/* or if this is a blackhole/reject route */
472 	if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
473 		return (0);
474 
475 	/* found valid route */
476 	return 1;
477 #endif /* __FreeBSD__ */
478 }
479 
480 /*
481  * Generate an SCTP packet containing an ABORT chunk. The verification tag
482  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
483  * reflected is not 0.
484  */
485 
486 static struct mbuf *
487 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
488     int reflected)
489 {
490 	struct mbuf *m;
491 	struct ip *ip;
492 #ifdef INET6
493 	struct ip6_hdr *ip6;
494 #endif
495 	struct sctphdr *sctp;
496 	struct sctp_chunkhdr *chunk;
497 	u_int16_t hlen, plen, tlen;
498 
499 	MGETHDR(m, M_NOWAIT, MT_DATA);
500 	if (m == NULL)
501 		return (NULL);
502 
503 	M_SETFIB(m, id->fib);
504 #ifdef MAC
505 	if (replyto != NULL)
506 		mac_netinet_firewall_reply(replyto, m);
507 	else
508 		mac_netinet_firewall_send(m);
509 #else
510 	(void)replyto;		/* don't warn about unused arg */
511 #endif
512 
513 	switch (id->addr_type) {
514 	case 4:
515 		hlen = sizeof(struct ip);
516 		break;
517 #ifdef INET6
518 	case 6:
519 		hlen = sizeof(struct ip6_hdr);
520 		break;
521 #endif
522 	default:
523 		/* XXX: log me?!? */
524 		FREE_PKT(m);
525 		return (NULL);
526 	}
527 	plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
528 	tlen = hlen + plen;
529 	m->m_data += max_linkhdr;
530 	m->m_flags |= M_SKIP_FIREWALL;
531 	m->m_pkthdr.len = m->m_len = tlen;
532 	m->m_pkthdr.rcvif = NULL;
533 	bzero(m->m_data, tlen);
534 
535 	switch (id->addr_type) {
536 	case 4:
537 		ip = mtod(m, struct ip *);
538 
539 		ip->ip_v = 4;
540 		ip->ip_hl = sizeof(struct ip) >> 2;
541 		ip->ip_tos = IPTOS_LOWDELAY;
542 		ip->ip_len = htons(tlen);
543 		ip->ip_id = htons(0);
544 		ip->ip_off = htons(0);
545 		ip->ip_ttl = V_ip_defttl;
546 		ip->ip_p = IPPROTO_SCTP;
547 		ip->ip_sum = 0;
548 		ip->ip_src.s_addr = htonl(id->dst_ip);
549 		ip->ip_dst.s_addr = htonl(id->src_ip);
550 
551 		sctp = (struct sctphdr *)(ip + 1);
552 		break;
553 #ifdef INET6
554 	case 6:
555 		ip6 = mtod(m, struct ip6_hdr *);
556 
557 		ip6->ip6_vfc = IPV6_VERSION;
558 		ip6->ip6_plen = htons(plen);
559 		ip6->ip6_nxt = IPPROTO_SCTP;
560 		ip6->ip6_hlim = IPV6_DEFHLIM;
561 		ip6->ip6_src = id->dst_ip6;
562 		ip6->ip6_dst = id->src_ip6;
563 
564 		sctp = (struct sctphdr *)(ip6 + 1);
565 		break;
566 #endif
567 	}
568 
569 	sctp->src_port = htons(id->dst_port);
570 	sctp->dest_port = htons(id->src_port);
571 	sctp->v_tag = htonl(vtag);
572 	sctp->checksum = htonl(0);
573 
574 	chunk = (struct sctp_chunkhdr *)(sctp + 1);
575 	chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
576 	chunk->chunk_flags = 0;
577 	if (reflected != 0) {
578 		chunk->chunk_flags |= SCTP_HAD_NO_TCB;
579 	}
580 	chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
581 
582 	sctp->checksum = sctp_calculate_cksum(m, hlen);
583 
584 	return (m);
585 }
586 
587 /*
588  * Generate a TCP packet, containing either a RST or a keepalive.
589  * When flags & TH_RST, we are sending a RST packet, because of a
590  * "reset" action matched the packet.
591  * Otherwise we are sending a keepalive, and flags & TH_
592  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
593  * so that MAC can label the reply appropriately.
594  */
595 struct mbuf *
596 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
597     u_int32_t ack, int flags)
598 {
599 	struct mbuf *m = NULL;		/* stupid compiler */
600 	struct ip *h = NULL;		/* stupid compiler */
601 #ifdef INET6
602 	struct ip6_hdr *h6 = NULL;
603 #endif
604 	struct tcphdr *th = NULL;
605 	int len, dir;
606 
607 	MGETHDR(m, M_NOWAIT, MT_DATA);
608 	if (m == NULL)
609 		return (NULL);
610 
611 	M_SETFIB(m, id->fib);
612 #ifdef MAC
613 	if (replyto != NULL)
614 		mac_netinet_firewall_reply(replyto, m);
615 	else
616 		mac_netinet_firewall_send(m);
617 #else
618 	(void)replyto;		/* don't warn about unused arg */
619 #endif
620 
621 	switch (id->addr_type) {
622 	case 4:
623 		len = sizeof(struct ip) + sizeof(struct tcphdr);
624 		break;
625 #ifdef INET6
626 	case 6:
627 		len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
628 		break;
629 #endif
630 	default:
631 		/* XXX: log me?!? */
632 		FREE_PKT(m);
633 		return (NULL);
634 	}
635 	dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
636 
637 	m->m_data += max_linkhdr;
638 	m->m_flags |= M_SKIP_FIREWALL;
639 	m->m_pkthdr.len = m->m_len = len;
640 	m->m_pkthdr.rcvif = NULL;
641 	bzero(m->m_data, len);
642 
643 	switch (id->addr_type) {
644 	case 4:
645 		h = mtod(m, struct ip *);
646 
647 		/* prepare for checksum */
648 		h->ip_p = IPPROTO_TCP;
649 		h->ip_len = htons(sizeof(struct tcphdr));
650 		if (dir) {
651 			h->ip_src.s_addr = htonl(id->src_ip);
652 			h->ip_dst.s_addr = htonl(id->dst_ip);
653 		} else {
654 			h->ip_src.s_addr = htonl(id->dst_ip);
655 			h->ip_dst.s_addr = htonl(id->src_ip);
656 		}
657 
658 		th = (struct tcphdr *)(h + 1);
659 		break;
660 #ifdef INET6
661 	case 6:
662 		h6 = mtod(m, struct ip6_hdr *);
663 
664 		/* prepare for checksum */
665 		h6->ip6_nxt = IPPROTO_TCP;
666 		h6->ip6_plen = htons(sizeof(struct tcphdr));
667 		if (dir) {
668 			h6->ip6_src = id->src_ip6;
669 			h6->ip6_dst = id->dst_ip6;
670 		} else {
671 			h6->ip6_src = id->dst_ip6;
672 			h6->ip6_dst = id->src_ip6;
673 		}
674 
675 		th = (struct tcphdr *)(h6 + 1);
676 		break;
677 #endif
678 	}
679 
680 	if (dir) {
681 		th->th_sport = htons(id->src_port);
682 		th->th_dport = htons(id->dst_port);
683 	} else {
684 		th->th_sport = htons(id->dst_port);
685 		th->th_dport = htons(id->src_port);
686 	}
687 	th->th_off = sizeof(struct tcphdr) >> 2;
688 
689 	if (flags & TH_RST) {
690 		if (flags & TH_ACK) {
691 			th->th_seq = htonl(ack);
692 			th->th_flags = TH_RST;
693 		} else {
694 			if (flags & TH_SYN)
695 				seq++;
696 			th->th_ack = htonl(seq);
697 			th->th_flags = TH_RST | TH_ACK;
698 		}
699 	} else {
700 		/*
701 		 * Keepalive - use caller provided sequence numbers
702 		 */
703 		th->th_seq = htonl(seq);
704 		th->th_ack = htonl(ack);
705 		th->th_flags = TH_ACK;
706 	}
707 
708 	switch (id->addr_type) {
709 	case 4:
710 		th->th_sum = in_cksum(m, len);
711 
712 		/* finish the ip header */
713 		h->ip_v = 4;
714 		h->ip_hl = sizeof(*h) >> 2;
715 		h->ip_tos = IPTOS_LOWDELAY;
716 		h->ip_off = htons(0);
717 		h->ip_len = htons(len);
718 		h->ip_ttl = V_ip_defttl;
719 		h->ip_sum = 0;
720 		break;
721 #ifdef INET6
722 	case 6:
723 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
724 		    sizeof(struct tcphdr));
725 
726 		/* finish the ip6 header */
727 		h6->ip6_vfc |= IPV6_VERSION;
728 		h6->ip6_hlim = IPV6_DEFHLIM;
729 		break;
730 #endif
731 	}
732 
733 	return (m);
734 }
735 
736 #ifdef INET6
737 /*
738  * ipv6 specific rules here...
739  */
740 static __inline int
741 icmp6type_match (int type, ipfw_insn_u32 *cmd)
742 {
743 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
744 }
745 
746 static int
747 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
748 {
749 	int i;
750 	for (i=0; i <= cmd->o.arg1; ++i )
751 		if (curr_flow == cmd->d[i] )
752 			return 1;
753 	return 0;
754 }
755 
756 /* support for IP6_*_ME opcodes */
757 static const struct in6_addr lla_mask = {{{
758 	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
759 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
760 }}};
761 
762 static int
763 ipfw_localip6(struct in6_addr *in6)
764 {
765 	struct rm_priotracker in6_ifa_tracker;
766 	struct in6_ifaddr *ia;
767 
768 	if (IN6_IS_ADDR_MULTICAST(in6))
769 		return (0);
770 
771 	if (!IN6_IS_ADDR_LINKLOCAL(in6))
772 		return (in6_localip(in6));
773 
774 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
775 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
776 		if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
777 			continue;
778 		if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
779 		    in6, &lla_mask)) {
780 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
781 			return (1);
782 		}
783 	}
784 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
785 	return (0);
786 }
787 
788 static int
789 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
790 {
791 	struct nhop6_basic nh6;
792 
793 	if (IN6_IS_SCOPE_LINKLOCAL(src))
794 		return (1);
795 
796 	if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
797 		return (0);
798 
799 	/* If ifp is provided, check for equality with route table. */
800 	if (ifp != NULL && ifp != nh6.nh_ifp)
801 		return (0);
802 
803 	/* if no ifp provided, check if rtentry is not default route */
804 	if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
805 		return (0);
806 
807 	/* or if this is a blackhole/reject route */
808 	if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
809 		return (0);
810 
811 	/* found valid route */
812 	return 1;
813 }
814 
815 static int
816 is_icmp6_query(int icmp6_type)
817 {
818 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
819 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
820 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
821 	    icmp6_type == ICMP6_WRUREQUEST ||
822 	    icmp6_type == ICMP6_FQDN_QUERY ||
823 	    icmp6_type == ICMP6_NI_QUERY))
824 		return (1);
825 
826 	return (0);
827 }
828 
829 static int
830 map_icmp_unreach(int code)
831 {
832 
833 	/* RFC 7915 p4.2 */
834 	switch (code) {
835 	case ICMP_UNREACH_NET:
836 	case ICMP_UNREACH_HOST:
837 	case ICMP_UNREACH_SRCFAIL:
838 	case ICMP_UNREACH_NET_UNKNOWN:
839 	case ICMP_UNREACH_HOST_UNKNOWN:
840 	case ICMP_UNREACH_TOSNET:
841 	case ICMP_UNREACH_TOSHOST:
842 		return (ICMP6_DST_UNREACH_NOROUTE);
843 	case ICMP_UNREACH_PORT:
844 		return (ICMP6_DST_UNREACH_NOPORT);
845 	default:
846 		/*
847 		 * Map the rest of codes into admit prohibited.
848 		 * XXX: unreach proto should be mapped into ICMPv6
849 		 * parameter problem, but we use only unreach type.
850 		 */
851 		return (ICMP6_DST_UNREACH_ADMIN);
852 	}
853 }
854 
855 static void
856 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
857 {
858 	struct mbuf *m;
859 
860 	m = args->m;
861 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
862 		struct tcphdr *tcp;
863 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
864 
865 		if ((tcp->th_flags & TH_RST) == 0) {
866 			struct mbuf *m0;
867 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
868 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
869 			    tcp->th_flags | TH_RST);
870 			if (m0 != NULL)
871 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
872 				    NULL);
873 		}
874 		FREE_PKT(m);
875 	} else if (code == ICMP6_UNREACH_ABORT &&
876 	    args->f_id.proto == IPPROTO_SCTP) {
877 		struct mbuf *m0;
878 		struct sctphdr *sctp;
879 		u_int32_t v_tag;
880 		int reflected;
881 
882 		sctp = (struct sctphdr *)((char *)ip6 + hlen);
883 		reflected = 1;
884 		v_tag = ntohl(sctp->v_tag);
885 		/* Investigate the first chunk header if available */
886 		if (m->m_len >= hlen + sizeof(struct sctphdr) +
887 		    sizeof(struct sctp_chunkhdr)) {
888 			struct sctp_chunkhdr *chunk;
889 
890 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
891 			switch (chunk->chunk_type) {
892 			case SCTP_INITIATION:
893 				/*
894 				 * Packets containing an INIT chunk MUST have
895 				 * a zero v-tag.
896 				 */
897 				if (v_tag != 0) {
898 					v_tag = 0;
899 					break;
900 				}
901 				/* INIT chunk MUST NOT be bundled */
902 				if (m->m_pkthdr.len >
903 				    hlen + sizeof(struct sctphdr) +
904 				    ntohs(chunk->chunk_length) + 3) {
905 					break;
906 				}
907 				/* Use the initiate tag if available */
908 				if ((m->m_len >= hlen + sizeof(struct sctphdr) +
909 				    sizeof(struct sctp_chunkhdr) +
910 				    offsetof(struct sctp_init, a_rwnd))) {
911 					struct sctp_init *init;
912 
913 					init = (struct sctp_init *)(chunk + 1);
914 					v_tag = ntohl(init->initiate_tag);
915 					reflected = 0;
916 				}
917 				break;
918 			case SCTP_ABORT_ASSOCIATION:
919 				/*
920 				 * If the packet contains an ABORT chunk, don't
921 				 * reply.
922 				 * XXX: We should search through all chunks,
923 				 *      but don't do to avoid attacks.
924 				 */
925 				v_tag = 0;
926 				break;
927 			}
928 		}
929 		if (v_tag == 0) {
930 			m0 = NULL;
931 		} else {
932 			m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
933 			    reflected);
934 		}
935 		if (m0 != NULL)
936 			ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
937 		FREE_PKT(m);
938 	} else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
939 		/* Send an ICMPv6 unreach. */
940 #if 0
941 		/*
942 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
943 		 * as the contents are read. We need to m_adj() the
944 		 * needed amount.
945 		 * The mbuf will however be thrown away so we can adjust it.
946 		 * Remember we did an m_pullup on it already so we
947 		 * can make some assumptions about contiguousness.
948 		 */
949 		if (args->L3offset)
950 			m_adj(m, args->L3offset);
951 #endif
952 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
953 	} else
954 		FREE_PKT(m);
955 
956 	args->m = NULL;
957 }
958 
959 #endif /* INET6 */
960 
961 
962 /*
963  * sends a reject message, consuming the mbuf passed as an argument.
964  */
965 static void
966 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
967 {
968 
969 #if 0
970 	/* XXX When ip is not guaranteed to be at mtod() we will
971 	 * need to account for this */
972 	 * The mbuf will however be thrown away so we can adjust it.
973 	 * Remember we did an m_pullup on it already so we
974 	 * can make some assumptions about contiguousness.
975 	 */
976 	if (args->L3offset)
977 		m_adj(m, args->L3offset);
978 #endif
979 	if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
980 		/* Send an ICMP unreach */
981 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
982 	} else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
983 		struct tcphdr *const tcp =
984 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
985 		if ( (tcp->th_flags & TH_RST) == 0) {
986 			struct mbuf *m;
987 			m = ipfw_send_pkt(args->m, &(args->f_id),
988 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
989 				tcp->th_flags | TH_RST);
990 			if (m != NULL)
991 				ip_output(m, NULL, NULL, 0, NULL, NULL);
992 		}
993 		FREE_PKT(args->m);
994 	} else if (code == ICMP_REJECT_ABORT &&
995 	    args->f_id.proto == IPPROTO_SCTP) {
996 		struct mbuf *m;
997 		struct sctphdr *sctp;
998 		struct sctp_chunkhdr *chunk;
999 		struct sctp_init *init;
1000 		u_int32_t v_tag;
1001 		int reflected;
1002 
1003 		sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1004 		reflected = 1;
1005 		v_tag = ntohl(sctp->v_tag);
1006 		if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1007 		    sizeof(struct sctp_chunkhdr)) {
1008 			/* Look at the first chunk header if available */
1009 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
1010 			switch (chunk->chunk_type) {
1011 			case SCTP_INITIATION:
1012 				/*
1013 				 * Packets containing an INIT chunk MUST have
1014 				 * a zero v-tag.
1015 				 */
1016 				if (v_tag != 0) {
1017 					v_tag = 0;
1018 					break;
1019 				}
1020 				/* INIT chunk MUST NOT be bundled */
1021 				if (iplen >
1022 				    (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1023 				    ntohs(chunk->chunk_length) + 3) {
1024 					break;
1025 				}
1026 				/* Use the initiate tag if available */
1027 				if ((iplen >= (ip->ip_hl << 2) +
1028 				    sizeof(struct sctphdr) +
1029 				    sizeof(struct sctp_chunkhdr) +
1030 				    offsetof(struct sctp_init, a_rwnd))) {
1031 					init = (struct sctp_init *)(chunk + 1);
1032 					v_tag = ntohl(init->initiate_tag);
1033 					reflected = 0;
1034 				}
1035 				break;
1036 			case SCTP_ABORT_ASSOCIATION:
1037 				/*
1038 				 * If the packet contains an ABORT chunk, don't
1039 				 * reply.
1040 				 * XXX: We should search through all chunks,
1041 				 * but don't do to avoid attacks.
1042 				 */
1043 				v_tag = 0;
1044 				break;
1045 			}
1046 		}
1047 		if (v_tag == 0) {
1048 			m = NULL;
1049 		} else {
1050 			m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1051 			    reflected);
1052 		}
1053 		if (m != NULL)
1054 			ip_output(m, NULL, NULL, 0, NULL, NULL);
1055 		FREE_PKT(args->m);
1056 	} else
1057 		FREE_PKT(args->m);
1058 	args->m = NULL;
1059 }
1060 
1061 /*
1062  * Support for uid/gid/jail lookup. These tests are expensive
1063  * (because we may need to look into the list of active sockets)
1064  * so we cache the results. ugid_lookupp is 0 if we have not
1065  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1066  * and failed. The function always returns the match value.
1067  * We could actually spare the variable and use *uc, setting
1068  * it to '(void *)check_uidgid if we have no info, NULL if
1069  * we tried and failed, or any other value if successful.
1070  */
1071 static int
1072 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1073     struct ucred **uc)
1074 {
1075 #if defined(USERSPACE)
1076 	return 0;	// not supported in userspace
1077 #else
1078 #ifndef __FreeBSD__
1079 	/* XXX */
1080 	return cred_check(insn, proto, oif,
1081 	    dst_ip, dst_port, src_ip, src_port,
1082 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1083 #else  /* FreeBSD */
1084 	struct in_addr src_ip, dst_ip;
1085 	struct inpcbinfo *pi;
1086 	struct ipfw_flow_id *id;
1087 	struct inpcb *pcb, *inp;
1088 	int lookupflags;
1089 	int match;
1090 
1091 	id = &args->f_id;
1092 	inp = args->inp;
1093 
1094 	/*
1095 	 * Check to see if the UDP or TCP stack supplied us with
1096 	 * the PCB. If so, rather then holding a lock and looking
1097 	 * up the PCB, we can use the one that was supplied.
1098 	 */
1099 	if (inp && *ugid_lookupp == 0) {
1100 		INP_LOCK_ASSERT(inp);
1101 		if (inp->inp_socket != NULL) {
1102 			*uc = crhold(inp->inp_cred);
1103 			*ugid_lookupp = 1;
1104 		} else
1105 			*ugid_lookupp = -1;
1106 	}
1107 	/*
1108 	 * If we have already been here and the packet has no
1109 	 * PCB entry associated with it, then we can safely
1110 	 * assume that this is a no match.
1111 	 */
1112 	if (*ugid_lookupp == -1)
1113 		return (0);
1114 	if (id->proto == IPPROTO_TCP) {
1115 		lookupflags = 0;
1116 		pi = &V_tcbinfo;
1117 	} else if (id->proto == IPPROTO_UDP) {
1118 		lookupflags = INPLOOKUP_WILDCARD;
1119 		pi = &V_udbinfo;
1120 	} else if (id->proto == IPPROTO_UDPLITE) {
1121 		lookupflags = INPLOOKUP_WILDCARD;
1122 		pi = &V_ulitecbinfo;
1123 	} else
1124 		return 0;
1125 	lookupflags |= INPLOOKUP_RLOCKPCB;
1126 	match = 0;
1127 	if (*ugid_lookupp == 0) {
1128 		if (id->addr_type == 6) {
1129 #ifdef INET6
1130 			if (args->flags & IPFW_ARGS_IN)
1131 				pcb = in6_pcblookup_mbuf(pi,
1132 				    &id->src_ip6, htons(id->src_port),
1133 				    &id->dst_ip6, htons(id->dst_port),
1134 				    lookupflags, NULL, args->m);
1135 			else
1136 				pcb = in6_pcblookup_mbuf(pi,
1137 				    &id->dst_ip6, htons(id->dst_port),
1138 				    &id->src_ip6, htons(id->src_port),
1139 				    lookupflags, args->ifp, args->m);
1140 #else
1141 			*ugid_lookupp = -1;
1142 			return (0);
1143 #endif
1144 		} else {
1145 			src_ip.s_addr = htonl(id->src_ip);
1146 			dst_ip.s_addr = htonl(id->dst_ip);
1147 			if (args->flags & IPFW_ARGS_IN)
1148 				pcb = in_pcblookup_mbuf(pi,
1149 				    src_ip, htons(id->src_port),
1150 				    dst_ip, htons(id->dst_port),
1151 				    lookupflags, NULL, args->m);
1152 			else
1153 				pcb = in_pcblookup_mbuf(pi,
1154 				    dst_ip, htons(id->dst_port),
1155 				    src_ip, htons(id->src_port),
1156 				    lookupflags, args->ifp, args->m);
1157 		}
1158 		if (pcb != NULL) {
1159 			INP_RLOCK_ASSERT(pcb);
1160 			*uc = crhold(pcb->inp_cred);
1161 			*ugid_lookupp = 1;
1162 			INP_RUNLOCK(pcb);
1163 		}
1164 		if (*ugid_lookupp == 0) {
1165 			/*
1166 			 * We tried and failed, set the variable to -1
1167 			 * so we will not try again on this packet.
1168 			 */
1169 			*ugid_lookupp = -1;
1170 			return (0);
1171 		}
1172 	}
1173 	if (insn->o.opcode == O_UID)
1174 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1175 	else if (insn->o.opcode == O_GID)
1176 		match = groupmember((gid_t)insn->d[0], *uc);
1177 	else if (insn->o.opcode == O_JAIL)
1178 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1179 	return (match);
1180 #endif /* __FreeBSD__ */
1181 #endif /* not supported in userspace */
1182 }
1183 
1184 /*
1185  * Helper function to set args with info on the rule after the matching
1186  * one. slot is precise, whereas we guess rule_id as they are
1187  * assigned sequentially.
1188  */
1189 static inline void
1190 set_match(struct ip_fw_args *args, int slot,
1191 	struct ip_fw_chain *chain)
1192 {
1193 	args->rule.chain_id = chain->id;
1194 	args->rule.slot = slot + 1; /* we use 0 as a marker */
1195 	args->rule.rule_id = 1 + chain->map[slot]->id;
1196 	args->rule.rulenum = chain->map[slot]->rulenum;
1197 	args->flags |= IPFW_ARGS_REF;
1198 }
1199 
1200 #ifndef LINEAR_SKIPTO
1201 /*
1202  * Helper function to enable cached rule lookups using
1203  * cached_id and cached_pos fields in ipfw rule.
1204  */
1205 static int
1206 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1207     int tablearg, int jump_backwards)
1208 {
1209 	int f_pos;
1210 
1211 	/* If possible use cached f_pos (in f->cached_pos),
1212 	 * whose version is written in f->cached_id
1213 	 * (horrible hacks to avoid changing the ABI).
1214 	 */
1215 	if (num != IP_FW_TARG && f->cached_id == chain->id)
1216 		f_pos = f->cached_pos;
1217 	else {
1218 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1219 		/* make sure we do not jump backward */
1220 		if (jump_backwards == 0 && i <= f->rulenum)
1221 			i = f->rulenum + 1;
1222 		if (chain->idxmap != NULL)
1223 			f_pos = chain->idxmap[i];
1224 		else
1225 			f_pos = ipfw_find_rule(chain, i, 0);
1226 		/* update the cache */
1227 		if (num != IP_FW_TARG) {
1228 			f->cached_id = chain->id;
1229 			f->cached_pos = f_pos;
1230 		}
1231 	}
1232 
1233 	return (f_pos);
1234 }
1235 #else
1236 /*
1237  * Helper function to enable real fast rule lookups.
1238  */
1239 static int
1240 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1241     int tablearg, int jump_backwards)
1242 {
1243 	int f_pos;
1244 
1245 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1246 	/* make sure we do not jump backward */
1247 	if (jump_backwards == 0 && num <= f->rulenum)
1248 		num = f->rulenum + 1;
1249 	f_pos = chain->idxmap[num];
1250 
1251 	return (f_pos);
1252 }
1253 #endif
1254 
1255 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
1256 /*
1257  * The main check routine for the firewall.
1258  *
1259  * All arguments are in args so we can modify them and return them
1260  * back to the caller.
1261  *
1262  * Parameters:
1263  *
1264  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
1265  *		Starts with the IP header.
1266  *	args->L3offset	Number of bytes bypassed if we came from L2.
1267  *			e.g. often sizeof(eh)  ** NOTYET **
1268  *	args->ifp	Incoming or outgoing interface.
1269  *	args->divert_rule (in/out)
1270  *		Skip up to the first rule past this rule number;
1271  *		upon return, non-zero port number for divert or tee.
1272  *
1273  *	args->rule	Pointer to the last matching rule (in/out)
1274  *	args->next_hop	Socket we are forwarding to (out).
1275  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
1276  *	args->f_id	Addresses grabbed from the packet (out)
1277  * 	args->rule.info	a cookie depending on rule action
1278  *
1279  * Return value:
1280  *
1281  *	IP_FW_PASS	the packet must be accepted
1282  *	IP_FW_DENY	the packet must be dropped
1283  *	IP_FW_DIVERT	divert packet, port in m_tag
1284  *	IP_FW_TEE	tee packet, port in m_tag
1285  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
1286  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
1287  *		args->rule contains the matching rule,
1288  *		args->rule.info has additional information.
1289  *
1290  */
1291 int
1292 ipfw_chk(struct ip_fw_args *args)
1293 {
1294 
1295 	/*
1296 	 * Local variables holding state while processing a packet:
1297 	 *
1298 	 * IMPORTANT NOTE: to speed up the processing of rules, there
1299 	 * are some assumption on the values of the variables, which
1300 	 * are documented here. Should you change them, please check
1301 	 * the implementation of the various instructions to make sure
1302 	 * that they still work.
1303 	 *
1304 	 * m | args->m	Pointer to the mbuf, as received from the caller.
1305 	 *	It may change if ipfw_chk() does an m_pullup, or if it
1306 	 *	consumes the packet because it calls send_reject().
1307 	 *	XXX This has to change, so that ipfw_chk() never modifies
1308 	 *	or consumes the buffer.
1309 	 *	OR
1310 	 * args->mem	Pointer to contigous memory chunk.
1311 	 * ip	Is the beginning of the ip(4 or 6) header.
1312 	 * eh	Ethernet header in case if input is Layer2.
1313 	 */
1314 	struct mbuf *m;
1315 	struct ip *ip;
1316 	struct ether_header *eh;
1317 
1318 	/*
1319 	 * For rules which contain uid/gid or jail constraints, cache
1320 	 * a copy of the users credentials after the pcb lookup has been
1321 	 * executed. This will speed up the processing of rules with
1322 	 * these types of constraints, as well as decrease contention
1323 	 * on pcb related locks.
1324 	 */
1325 #ifndef __FreeBSD__
1326 	struct bsd_ucred ucred_cache;
1327 #else
1328 	struct ucred *ucred_cache = NULL;
1329 #endif
1330 	int ucred_lookup = 0;
1331 	int f_pos = 0;		/* index of current rule in the array */
1332 	int retval = 0;
1333 	struct ifnet *oif, *iif;
1334 
1335 	/*
1336 	 * hlen	The length of the IP header.
1337 	 */
1338 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
1339 
1340 	/*
1341 	 * offset	The offset of a fragment. offset != 0 means that
1342 	 *	we have a fragment at this offset of an IPv4 packet.
1343 	 *	offset == 0 means that (if this is an IPv4 packet)
1344 	 *	this is the first or only fragment.
1345 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1346 	 *	or there is a single packet fragment (fragment header added
1347 	 *	without needed).  We will treat a single packet fragment as if
1348 	 *	there was no fragment header (or log/block depending on the
1349 	 *	V_fw_permit_single_frag6 sysctl setting).
1350 	 */
1351 	u_short offset = 0;
1352 	u_short ip6f_mf = 0;
1353 
1354 	/*
1355 	 * Local copies of addresses. They are only valid if we have
1356 	 * an IP packet.
1357 	 *
1358 	 * proto	The protocol. Set to 0 for non-ip packets,
1359 	 *	or to the protocol read from the packet otherwise.
1360 	 *	proto != 0 means that we have an IPv4 packet.
1361 	 *
1362 	 * src_port, dst_port	port numbers, in HOST format. Only
1363 	 *	valid for TCP and UDP packets.
1364 	 *
1365 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
1366 	 *	Only valid for IPv4 packets.
1367 	 */
1368 	uint8_t proto;
1369 	uint16_t src_port, dst_port;		/* NOTE: host format	*/
1370 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
1371 	int iplen = 0;
1372 	int pktlen;
1373 
1374 	struct ipfw_dyn_info dyn_info;
1375 	struct ip_fw *q = NULL;
1376 	struct ip_fw_chain *chain = &V_layer3_chain;
1377 
1378 	/*
1379 	 * We store in ulp a pointer to the upper layer protocol header.
1380 	 * In the ipv4 case this is easy to determine from the header,
1381 	 * but for ipv6 we might have some additional headers in the middle.
1382 	 * ulp is NULL if not found.
1383 	 */
1384 	void *ulp = NULL;		/* upper layer protocol pointer. */
1385 
1386 	/* XXX ipv6 variables */
1387 	int is_ipv6 = 0;
1388 	uint8_t	icmp6_type = 0;
1389 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1390 	/* end of ipv6 variables */
1391 
1392 	int is_ipv4 = 0;
1393 
1394 	int done = 0;		/* flag to exit the outer loop */
1395 	IPFW_RLOCK_TRACKER;
1396 	bool mem;
1397 
1398 	if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1399 		if (args->flags & IPFW_ARGS_ETHER) {
1400 			eh = (struct ether_header *)args->mem;
1401 			if (eh->ether_type == htons(ETHERTYPE_VLAN))
1402 				ip = (struct ip *)
1403 				    ((struct ether_vlan_header *)eh + 1);
1404 			else
1405 				ip = (struct ip *)(eh + 1);
1406 		} else {
1407 			eh = NULL;
1408 			ip = (struct ip *)args->mem;
1409 		}
1410 		pktlen = IPFW_ARGS_LENGTH(args->flags);
1411 		args->f_id.fib = args->ifp->if_fib;	/* best guess */
1412 	} else {
1413 		m = args->m;
1414 		if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1415 			return (IP_FW_PASS);	/* accept */
1416 		if (args->flags & IPFW_ARGS_ETHER) {
1417 	                /* We need some amount of data to be contiguous. */
1418 			if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1419 			    (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1420 			    max_protohdr))) == NULL)
1421 				goto pullup_failed;
1422 			eh = mtod(m, struct ether_header *);
1423 			ip = (struct ip *)(eh + 1);
1424 		} else {
1425 			eh = NULL;
1426 			ip = mtod(m, struct ip *);
1427 		}
1428 		pktlen = m->m_pkthdr.len;
1429 		args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1430 	}
1431 
1432 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1433 	src_ip.s_addr = 0;		/* make sure it is initialized */
1434 	src_port = dst_port = 0;
1435 
1436 	DYN_INFO_INIT(&dyn_info);
1437 /*
1438  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1439  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1440  * pointer might become stale after other pullups (but we never use it
1441  * this way).
1442  */
1443 #define	PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1444 #define	EHLEN	(eh != NULL ? ((char *)ip - (char *)eh) : 0)
1445 #define	_PULLUP_LOCKED(_len, p, T, unlock)			\
1446 do {								\
1447 	int x = (_len) + T + EHLEN;				\
1448 	if (mem) {						\
1449 		if (__predict_false(pktlen < x)) {		\
1450 			unlock;					\
1451 			goto pullup_failed;			\
1452 		}						\
1453 		p = (char *)args->mem + (_len) + EHLEN;		\
1454 	} else {						\
1455 		if (__predict_false((m)->m_len < x)) {		\
1456 			args->m = m = m_pullup(m, x);		\
1457 			if (m == NULL) {			\
1458 				unlock;				\
1459 				goto pullup_failed;		\
1460 			}					\
1461 		}						\
1462 		p = mtod(m, char *) + (_len) + EHLEN;		\
1463 	}							\
1464 } while (0)
1465 
1466 #define	PULLUP_LEN(_len, p, T)	_PULLUP_LOCKED(_len, p, T, )
1467 #define	PULLUP_LEN_LOCKED(_len, p, T)	\
1468     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain));	\
1469     UPDATE_POINTERS()
1470 /*
1471  * In case pointers got stale after pullups, update them.
1472  */
1473 #define	UPDATE_POINTERS()					\
1474 do {								\
1475 	if (!mem) {						\
1476 		if (eh != NULL) {				\
1477 			eh = mtod(m, struct ether_header *);	\
1478 			ip = (struct ip *)(eh + 1);		\
1479 		} else						\
1480 			ip = mtod(m, struct ip *);		\
1481 		args->m = m;					\
1482 	}							\
1483 } while (0)
1484 
1485 	/* Identify IP packets and fill up variables. */
1486 	if (pktlen >= sizeof(struct ip6_hdr) &&
1487 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1488 	    ip->ip_v == 6) {
1489 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1490 
1491 		is_ipv6 = 1;
1492 		args->flags |= IPFW_ARGS_IP6;
1493 		hlen = sizeof(struct ip6_hdr);
1494 		proto = ip6->ip6_nxt;
1495 		/* Search extension headers to find upper layer protocols */
1496 		while (ulp == NULL && offset == 0) {
1497 			switch (proto) {
1498 			case IPPROTO_ICMPV6:
1499 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1500 				icmp6_type = ICMP6(ulp)->icmp6_type;
1501 				break;
1502 
1503 			case IPPROTO_TCP:
1504 				PULLUP_TO(hlen, ulp, struct tcphdr);
1505 				dst_port = TCP(ulp)->th_dport;
1506 				src_port = TCP(ulp)->th_sport;
1507 				/* save flags for dynamic rules */
1508 				args->f_id._flags = TCP(ulp)->th_flags;
1509 				break;
1510 
1511 			case IPPROTO_SCTP:
1512 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1513 				    sizeof(struct sctp_chunkhdr) +
1514 				    offsetof(struct sctp_init, a_rwnd))
1515 					PULLUP_LEN(hlen, ulp,
1516 					    sizeof(struct sctphdr) +
1517 					    sizeof(struct sctp_chunkhdr) +
1518 					    offsetof(struct sctp_init, a_rwnd));
1519 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1520 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1521 				else
1522 					PULLUP_LEN(hlen, ulp,
1523 					    sizeof(struct sctphdr));
1524 				src_port = SCTP(ulp)->src_port;
1525 				dst_port = SCTP(ulp)->dest_port;
1526 				break;
1527 
1528 			case IPPROTO_UDP:
1529 			case IPPROTO_UDPLITE:
1530 				PULLUP_TO(hlen, ulp, struct udphdr);
1531 				dst_port = UDP(ulp)->uh_dport;
1532 				src_port = UDP(ulp)->uh_sport;
1533 				break;
1534 
1535 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1536 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1537 				ext_hd |= EXT_HOPOPTS;
1538 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1539 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1540 				ulp = NULL;
1541 				break;
1542 
1543 			case IPPROTO_ROUTING:	/* RFC 2460 */
1544 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1545 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1546 				case 0:
1547 					ext_hd |= EXT_RTHDR0;
1548 					break;
1549 				case 2:
1550 					ext_hd |= EXT_RTHDR2;
1551 					break;
1552 				default:
1553 					if (V_fw_verbose)
1554 						printf("IPFW2: IPV6 - Unknown "
1555 						    "Routing Header type(%d)\n",
1556 						    ((struct ip6_rthdr *)
1557 						    ulp)->ip6r_type);
1558 					if (V_fw_deny_unknown_exthdrs)
1559 					    return (IP_FW_DENY);
1560 					break;
1561 				}
1562 				ext_hd |= EXT_ROUTING;
1563 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1564 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1565 				ulp = NULL;
1566 				break;
1567 
1568 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1569 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1570 				ext_hd |= EXT_FRAGMENT;
1571 				hlen += sizeof (struct ip6_frag);
1572 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1573 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1574 					IP6F_OFF_MASK;
1575 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1576 					IP6F_MORE_FRAG;
1577 				if (V_fw_permit_single_frag6 == 0 &&
1578 				    offset == 0 && ip6f_mf == 0) {
1579 					if (V_fw_verbose)
1580 						printf("IPFW2: IPV6 - Invalid "
1581 						    "Fragment Header\n");
1582 					if (V_fw_deny_unknown_exthdrs)
1583 					    return (IP_FW_DENY);
1584 					break;
1585 				}
1586 				args->f_id.extra =
1587 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1588 				ulp = NULL;
1589 				break;
1590 
1591 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1592 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1593 				ext_hd |= EXT_DSTOPTS;
1594 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1595 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1596 				ulp = NULL;
1597 				break;
1598 
1599 			case IPPROTO_AH:	/* RFC 2402 */
1600 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1601 				ext_hd |= EXT_AH;
1602 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1603 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1604 				ulp = NULL;
1605 				break;
1606 
1607 			case IPPROTO_ESP:	/* RFC 2406 */
1608 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1609 				/* Anything past Seq# is variable length and
1610 				 * data past this ext. header is encrypted. */
1611 				ext_hd |= EXT_ESP;
1612 				break;
1613 
1614 			case IPPROTO_NONE:	/* RFC 2460 */
1615 				/*
1616 				 * Packet ends here, and IPv6 header has
1617 				 * already been pulled up. If ip6e_len!=0
1618 				 * then octets must be ignored.
1619 				 */
1620 				ulp = ip; /* non-NULL to get out of loop. */
1621 				break;
1622 
1623 			case IPPROTO_OSPFIGP:
1624 				/* XXX OSPF header check? */
1625 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1626 				break;
1627 
1628 			case IPPROTO_PIM:
1629 				/* XXX PIM header check? */
1630 				PULLUP_TO(hlen, ulp, struct pim);
1631 				break;
1632 
1633 			case IPPROTO_GRE:	/* RFC 1701 */
1634 				/* XXX GRE header check? */
1635 				PULLUP_TO(hlen, ulp, struct grehdr);
1636 				break;
1637 
1638 			case IPPROTO_CARP:
1639 				PULLUP_TO(hlen, ulp, offsetof(
1640 				    struct carp_header, carp_counter));
1641 				if (CARP_ADVERTISEMENT !=
1642 				    ((struct carp_header *)ulp)->carp_type)
1643 					return (IP_FW_DENY);
1644 				break;
1645 
1646 			case IPPROTO_IPV6:	/* RFC 2893 */
1647 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1648 				break;
1649 
1650 			case IPPROTO_IPV4:	/* RFC 2893 */
1651 				PULLUP_TO(hlen, ulp, struct ip);
1652 				break;
1653 
1654 			default:
1655 				if (V_fw_verbose)
1656 					printf("IPFW2: IPV6 - Unknown "
1657 					    "Extension Header(%d), ext_hd=%x\n",
1658 					     proto, ext_hd);
1659 				if (V_fw_deny_unknown_exthdrs)
1660 				    return (IP_FW_DENY);
1661 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1662 				break;
1663 			} /*switch */
1664 		}
1665 		UPDATE_POINTERS();
1666 		ip6 = (struct ip6_hdr *)ip;
1667 		args->f_id.addr_type = 6;
1668 		args->f_id.src_ip6 = ip6->ip6_src;
1669 		args->f_id.dst_ip6 = ip6->ip6_dst;
1670 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1671 		iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1672 	} else if (pktlen >= sizeof(struct ip) &&
1673 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1674 	    ip->ip_v == 4) {
1675 		is_ipv4 = 1;
1676 		args->flags |= IPFW_ARGS_IP4;
1677 		hlen = ip->ip_hl << 2;
1678 		/*
1679 		 * Collect parameters into local variables for faster
1680 		 * matching.
1681 		 */
1682 		proto = ip->ip_p;
1683 		src_ip = ip->ip_src;
1684 		dst_ip = ip->ip_dst;
1685 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1686 		iplen = ntohs(ip->ip_len);
1687 
1688 		if (offset == 0) {
1689 			switch (proto) {
1690 			case IPPROTO_TCP:
1691 				PULLUP_TO(hlen, ulp, struct tcphdr);
1692 				dst_port = TCP(ulp)->th_dport;
1693 				src_port = TCP(ulp)->th_sport;
1694 				/* save flags for dynamic rules */
1695 				args->f_id._flags = TCP(ulp)->th_flags;
1696 				break;
1697 
1698 			case IPPROTO_SCTP:
1699 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1700 				    sizeof(struct sctp_chunkhdr) +
1701 				    offsetof(struct sctp_init, a_rwnd))
1702 					PULLUP_LEN(hlen, ulp,
1703 					    sizeof(struct sctphdr) +
1704 					    sizeof(struct sctp_chunkhdr) +
1705 					    offsetof(struct sctp_init, a_rwnd));
1706 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1707 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1708 				else
1709 					PULLUP_LEN(hlen, ulp,
1710 					    sizeof(struct sctphdr));
1711 				src_port = SCTP(ulp)->src_port;
1712 				dst_port = SCTP(ulp)->dest_port;
1713 				break;
1714 
1715 			case IPPROTO_UDP:
1716 			case IPPROTO_UDPLITE:
1717 				PULLUP_TO(hlen, ulp, struct udphdr);
1718 				dst_port = UDP(ulp)->uh_dport;
1719 				src_port = UDP(ulp)->uh_sport;
1720 				break;
1721 
1722 			case IPPROTO_ICMP:
1723 				PULLUP_TO(hlen, ulp, struct icmphdr);
1724 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1725 				break;
1726 
1727 			default:
1728 				break;
1729 			}
1730 		} else {
1731 			if (offset == 1 && proto == IPPROTO_TCP) {
1732 				/* RFC 3128 */
1733 				goto pullup_failed;
1734 			}
1735 		}
1736 
1737 		UPDATE_POINTERS();
1738 		args->f_id.addr_type = 4;
1739 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1740 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1741 	} else {
1742 		proto = 0;
1743 		dst_ip.s_addr = src_ip.s_addr = 0;
1744 
1745 		args->f_id.addr_type = 1; /* XXX */
1746 	}
1747 #undef PULLUP_TO
1748 	pktlen = iplen < pktlen ? iplen: pktlen;
1749 
1750 	/* Properly initialize the rest of f_id */
1751 	args->f_id.proto = proto;
1752 	args->f_id.src_port = src_port = ntohs(src_port);
1753 	args->f_id.dst_port = dst_port = ntohs(dst_port);
1754 
1755 	IPFW_PF_RLOCK(chain);
1756 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1757 		IPFW_PF_RUNLOCK(chain);
1758 		return (IP_FW_PASS);	/* accept */
1759 	}
1760 	if (args->flags & IPFW_ARGS_REF) {
1761 		/*
1762 		 * Packet has already been tagged as a result of a previous
1763 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1764 		 * REASS, NETGRAPH, DIVERT/TEE...)
1765 		 * Validate the slot and continue from the next one
1766 		 * if still present, otherwise do a lookup.
1767 		 */
1768 		f_pos = (args->rule.chain_id == chain->id) ?
1769 		    args->rule.slot :
1770 		    ipfw_find_rule(chain, args->rule.rulenum,
1771 			args->rule.rule_id);
1772 	} else {
1773 		f_pos = 0;
1774 	}
1775 
1776 	if (args->flags & IPFW_ARGS_IN) {
1777 		iif = args->ifp;
1778 		oif = NULL;
1779 	} else {
1780 		MPASS(args->flags & IPFW_ARGS_OUT);
1781 		iif = mem ? NULL : m_rcvif(m);
1782 		oif = args->ifp;
1783 	}
1784 
1785 	/*
1786 	 * Now scan the rules, and parse microinstructions for each rule.
1787 	 * We have two nested loops and an inner switch. Sometimes we
1788 	 * need to break out of one or both loops, or re-enter one of
1789 	 * the loops with updated variables. Loop variables are:
1790 	 *
1791 	 *	f_pos (outer loop) points to the current rule.
1792 	 *		On output it points to the matching rule.
1793 	 *	done (outer loop) is used as a flag to break the loop.
1794 	 *	l (inner loop)	residual length of current rule.
1795 	 *		cmd points to the current microinstruction.
1796 	 *
1797 	 * We break the inner loop by setting l=0 and possibly
1798 	 * cmdlen=0 if we don't want to advance cmd.
1799 	 * We break the outer loop by setting done=1
1800 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1801 	 * as needed.
1802 	 */
1803 	for (; f_pos < chain->n_rules; f_pos++) {
1804 		ipfw_insn *cmd;
1805 		uint32_t tablearg = 0;
1806 		int l, cmdlen, skip_or; /* skip rest of OR block */
1807 		struct ip_fw *f;
1808 
1809 		f = chain->map[f_pos];
1810 		if (V_set_disable & (1 << f->set) )
1811 			continue;
1812 
1813 		skip_or = 0;
1814 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1815 		    l -= cmdlen, cmd += cmdlen) {
1816 			int match;
1817 
1818 			/*
1819 			 * check_body is a jump target used when we find a
1820 			 * CHECK_STATE, and need to jump to the body of
1821 			 * the target rule.
1822 			 */
1823 
1824 /* check_body: */
1825 			cmdlen = F_LEN(cmd);
1826 			/*
1827 			 * An OR block (insn_1 || .. || insn_n) has the
1828 			 * F_OR bit set in all but the last instruction.
1829 			 * The first match will set "skip_or", and cause
1830 			 * the following instructions to be skipped until
1831 			 * past the one with the F_OR bit clear.
1832 			 */
1833 			if (skip_or) {		/* skip this instruction */
1834 				if ((cmd->len & F_OR) == 0)
1835 					skip_or = 0;	/* next one is good */
1836 				continue;
1837 			}
1838 			match = 0; /* set to 1 if we succeed */
1839 
1840 			switch (cmd->opcode) {
1841 			/*
1842 			 * The first set of opcodes compares the packet's
1843 			 * fields with some pattern, setting 'match' if a
1844 			 * match is found. At the end of the loop there is
1845 			 * logic to deal with F_NOT and F_OR flags associated
1846 			 * with the opcode.
1847 			 */
1848 			case O_NOP:
1849 				match = 1;
1850 				break;
1851 
1852 			case O_FORWARD_MAC:
1853 				printf("ipfw: opcode %d unimplemented\n",
1854 				    cmd->opcode);
1855 				break;
1856 
1857 			case O_GID:
1858 			case O_UID:
1859 			case O_JAIL:
1860 				/*
1861 				 * We only check offset == 0 && proto != 0,
1862 				 * as this ensures that we have a
1863 				 * packet with the ports info.
1864 				 */
1865 				if (offset != 0)
1866 					break;
1867 				if (proto == IPPROTO_TCP ||
1868 				    proto == IPPROTO_UDP ||
1869 				    proto == IPPROTO_UDPLITE)
1870 					match = check_uidgid(
1871 						    (ipfw_insn_u32 *)cmd,
1872 						    args, &ucred_lookup,
1873 #ifdef __FreeBSD__
1874 						    &ucred_cache);
1875 #else
1876 						    (void *)&ucred_cache);
1877 #endif
1878 				break;
1879 
1880 			case O_RECV:
1881 				match = iface_match(iif, (ipfw_insn_if *)cmd,
1882 				    chain, &tablearg);
1883 				break;
1884 
1885 			case O_XMIT:
1886 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1887 				    chain, &tablearg);
1888 				break;
1889 
1890 			case O_VIA:
1891 				match = iface_match(args->ifp,
1892 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1893 				break;
1894 
1895 			case O_MACADDR2:
1896 				if (args->flags & IPFW_ARGS_ETHER) {
1897 					u_int32_t *want = (u_int32_t *)
1898 						((ipfw_insn_mac *)cmd)->addr;
1899 					u_int32_t *mask = (u_int32_t *)
1900 						((ipfw_insn_mac *)cmd)->mask;
1901 					u_int32_t *hdr = (u_int32_t *)eh;
1902 
1903 					match =
1904 					    ( want[0] == (hdr[0] & mask[0]) &&
1905 					      want[1] == (hdr[1] & mask[1]) &&
1906 					      want[2] == (hdr[2] & mask[2]) );
1907 				}
1908 				break;
1909 
1910 			case O_MAC_TYPE:
1911 				if (args->flags & IPFW_ARGS_ETHER) {
1912 					u_int16_t *p =
1913 					    ((ipfw_insn_u16 *)cmd)->ports;
1914 					int i;
1915 
1916 					for (i = cmdlen - 1; !match && i>0;
1917 					    i--, p += 2)
1918 						match =
1919 						    (ntohs(eh->ether_type) >=
1920 						    p[0] &&
1921 						    ntohs(eh->ether_type) <=
1922 						    p[1]);
1923 				}
1924 				break;
1925 
1926 			case O_FRAG:
1927 				match = (offset != 0);
1928 				break;
1929 
1930 			case O_IN:	/* "out" is "not in" */
1931 				match = (oif == NULL);
1932 				break;
1933 
1934 			case O_LAYER2:
1935 				match = (args->flags & IPFW_ARGS_ETHER);
1936 				break;
1937 
1938 			case O_DIVERTED:
1939 				if ((args->flags & IPFW_ARGS_REF) == 0)
1940 					break;
1941 				/*
1942 				 * For diverted packets, args->rule.info
1943 				 * contains the divert port (in host format)
1944 				 * reason and direction.
1945 				 */
1946 				match = ((args->rule.info & IPFW_IS_MASK) ==
1947 				    IPFW_IS_DIVERT) && (
1948 				    ((args->rule.info & IPFW_INFO_IN) ?
1949 					1: 2) & cmd->arg1);
1950 				break;
1951 
1952 			case O_PROTO:
1953 				/*
1954 				 * We do not allow an arg of 0 so the
1955 				 * check of "proto" only suffices.
1956 				 */
1957 				match = (proto == cmd->arg1);
1958 				break;
1959 
1960 			case O_IP_SRC:
1961 				match = is_ipv4 &&
1962 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1963 				    src_ip.s_addr);
1964 				break;
1965 
1966 			case O_IP_DST_LOOKUP:
1967 			{
1968 				void *pkey;
1969 				uint32_t vidx, key;
1970 				uint16_t keylen;
1971 
1972 				if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1973 					/* Determine lookup key type */
1974 					vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1975 					if (vidx != 4 /* uid */ &&
1976 					    vidx != 5 /* jail */ &&
1977 					    is_ipv6 == 0 && is_ipv4 == 0)
1978 						break;
1979 					/* Determine key length */
1980 					if (vidx == 0 /* dst-ip */ ||
1981 					    vidx == 1 /* src-ip */)
1982 						keylen = is_ipv6 ?
1983 						    sizeof(struct in6_addr):
1984 						    sizeof(in_addr_t);
1985 					else {
1986 						keylen = sizeof(key);
1987 						pkey = &key;
1988 					}
1989 					if (vidx == 0 /* dst-ip */)
1990 						pkey = is_ipv4 ? (void *)&dst_ip:
1991 						    (void *)&args->f_id.dst_ip6;
1992 					else if (vidx == 1 /* src-ip */)
1993 						pkey = is_ipv4 ? (void *)&src_ip:
1994 						    (void *)&args->f_id.src_ip6;
1995 					else if (vidx == 6 /* dscp */) {
1996 						if (is_ipv4)
1997 							key = ip->ip_tos >> 2;
1998 						else {
1999 							key = args->f_id.flow_id6;
2000 							key = (key & 0x0f) << 2 |
2001 							    (key & 0xf000) >> 14;
2002 						}
2003 						key &= 0x3f;
2004 					} else if (vidx == 2 /* dst-port */ ||
2005 					    vidx == 3 /* src-port */) {
2006 						/* Skip fragments */
2007 						if (offset != 0)
2008 							break;
2009 						/* Skip proto without ports */
2010 						if (proto != IPPROTO_TCP &&
2011 						    proto != IPPROTO_UDP &&
2012 						    proto != IPPROTO_UDPLITE &&
2013 						    proto != IPPROTO_SCTP)
2014 							break;
2015 						if (vidx == 2 /* dst-port */)
2016 							key = dst_port;
2017 						else
2018 							key = src_port;
2019 					}
2020 #ifndef USERSPACE
2021 					else if (vidx == 4 /* uid */ ||
2022 					    vidx == 5 /* jail */) {
2023 						check_uidgid(
2024 						    (ipfw_insn_u32 *)cmd,
2025 						    args, &ucred_lookup,
2026 #ifdef __FreeBSD__
2027 						    &ucred_cache);
2028 						if (vidx == 4 /* uid */)
2029 							key = ucred_cache->cr_uid;
2030 						else if (vidx == 5 /* jail */)
2031 							key = ucred_cache->cr_prison->pr_id;
2032 #else /* !__FreeBSD__ */
2033 						    (void *)&ucred_cache);
2034 						if (vidx == 4 /* uid */)
2035 							key = ucred_cache.uid;
2036 						else if (vidx == 5 /* jail */)
2037 							key = ucred_cache.xid;
2038 #endif /* !__FreeBSD__ */
2039 					}
2040 #endif /* !USERSPACE */
2041 					else
2042 						break;
2043 					match = ipfw_lookup_table(chain,
2044 					    cmd->arg1, keylen, pkey, &vidx);
2045 					if (!match)
2046 						break;
2047 					tablearg = vidx;
2048 					break;
2049 				}
2050 				/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2051 				/* FALLTHROUGH */
2052 			}
2053 			case O_IP_SRC_LOOKUP:
2054 			{
2055 				void *pkey;
2056 				uint32_t vidx;
2057 				uint16_t keylen;
2058 
2059 				if (is_ipv4) {
2060 					keylen = sizeof(in_addr_t);
2061 					if (cmd->opcode == O_IP_DST_LOOKUP)
2062 						pkey = &dst_ip;
2063 					else
2064 						pkey = &src_ip;
2065 				} else if (is_ipv6) {
2066 					keylen = sizeof(struct in6_addr);
2067 					if (cmd->opcode == O_IP_DST_LOOKUP)
2068 						pkey = &args->f_id.dst_ip6;
2069 					else
2070 						pkey = &args->f_id.src_ip6;
2071 				} else
2072 					break;
2073 				match = ipfw_lookup_table(chain, cmd->arg1,
2074 				    keylen, pkey, &vidx);
2075 				if (!match)
2076 					break;
2077 				if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2078 					match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2079 					    TARG_VAL(chain, vidx, tag);
2080 					if (!match)
2081 						break;
2082 				}
2083 				tablearg = vidx;
2084 				break;
2085 			}
2086 
2087 			case O_IP_FLOW_LOOKUP:
2088 				{
2089 					uint32_t v = 0;
2090 					match = ipfw_lookup_table(chain,
2091 					    cmd->arg1, 0, &args->f_id, &v);
2092 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2093 						match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2094 						    TARG_VAL(chain, v, tag);
2095 					if (match)
2096 						tablearg = v;
2097 				}
2098 				break;
2099 			case O_IP_SRC_MASK:
2100 			case O_IP_DST_MASK:
2101 				if (is_ipv4) {
2102 				    uint32_t a =
2103 					(cmd->opcode == O_IP_DST_MASK) ?
2104 					    dst_ip.s_addr : src_ip.s_addr;
2105 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2106 				    int i = cmdlen-1;
2107 
2108 				    for (; !match && i>0; i-= 2, p+= 2)
2109 					match = (p[0] == (a & p[1]));
2110 				}
2111 				break;
2112 
2113 			case O_IP_SRC_ME:
2114 				if (is_ipv4) {
2115 					match = in_localip(src_ip);
2116 					break;
2117 				}
2118 #ifdef INET6
2119 				/* FALLTHROUGH */
2120 			case O_IP6_SRC_ME:
2121 				match = is_ipv6 &&
2122 				    ipfw_localip6(&args->f_id.src_ip6);
2123 #endif
2124 				break;
2125 
2126 			case O_IP_DST_SET:
2127 			case O_IP_SRC_SET:
2128 				if (is_ipv4) {
2129 					u_int32_t *d = (u_int32_t *)(cmd+1);
2130 					u_int32_t addr =
2131 					    cmd->opcode == O_IP_DST_SET ?
2132 						args->f_id.dst_ip :
2133 						args->f_id.src_ip;
2134 
2135 					    if (addr < d[0])
2136 						    break;
2137 					    addr -= d[0]; /* subtract base */
2138 					    match = (addr < cmd->arg1) &&
2139 						( d[ 1 + (addr>>5)] &
2140 						  (1<<(addr & 0x1f)) );
2141 				}
2142 				break;
2143 
2144 			case O_IP_DST:
2145 				match = is_ipv4 &&
2146 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2147 				    dst_ip.s_addr);
2148 				break;
2149 
2150 			case O_IP_DST_ME:
2151 				if (is_ipv4) {
2152 					match = in_localip(dst_ip);
2153 					break;
2154 				}
2155 #ifdef INET6
2156 				/* FALLTHROUGH */
2157 			case O_IP6_DST_ME:
2158 				match = is_ipv6 &&
2159 				    ipfw_localip6(&args->f_id.dst_ip6);
2160 #endif
2161 				break;
2162 
2163 
2164 			case O_IP_SRCPORT:
2165 			case O_IP_DSTPORT:
2166 				/*
2167 				 * offset == 0 && proto != 0 is enough
2168 				 * to guarantee that we have a
2169 				 * packet with port info.
2170 				 */
2171 				if ((proto == IPPROTO_UDP ||
2172 				    proto == IPPROTO_UDPLITE ||
2173 				    proto == IPPROTO_TCP ||
2174 				    proto == IPPROTO_SCTP) && offset == 0) {
2175 					u_int16_t x =
2176 					    (cmd->opcode == O_IP_SRCPORT) ?
2177 						src_port : dst_port ;
2178 					u_int16_t *p =
2179 					    ((ipfw_insn_u16 *)cmd)->ports;
2180 					int i;
2181 
2182 					for (i = cmdlen - 1; !match && i>0;
2183 					    i--, p += 2)
2184 						match = (x>=p[0] && x<=p[1]);
2185 				}
2186 				break;
2187 
2188 			case O_ICMPTYPE:
2189 				match = (offset == 0 && proto==IPPROTO_ICMP &&
2190 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2191 				break;
2192 
2193 #ifdef INET6
2194 			case O_ICMP6TYPE:
2195 				match = is_ipv6 && offset == 0 &&
2196 				    proto==IPPROTO_ICMPV6 &&
2197 				    icmp6type_match(
2198 					ICMP6(ulp)->icmp6_type,
2199 					(ipfw_insn_u32 *)cmd);
2200 				break;
2201 #endif /* INET6 */
2202 
2203 			case O_IPOPT:
2204 				match = (is_ipv4 &&
2205 				    ipopts_match(ip, cmd) );
2206 				break;
2207 
2208 			case O_IPVER:
2209 				match = (is_ipv4 &&
2210 				    cmd->arg1 == ip->ip_v);
2211 				break;
2212 
2213 			case O_IPID:
2214 			case O_IPTTL:
2215 				if (!is_ipv4)
2216 					break;
2217 			case O_IPLEN:
2218 				{	/* only for IP packets */
2219 				    uint16_t x;
2220 				    uint16_t *p;
2221 				    int i;
2222 
2223 				    if (cmd->opcode == O_IPLEN)
2224 					x = iplen;
2225 				    else if (cmd->opcode == O_IPTTL)
2226 					x = ip->ip_ttl;
2227 				    else /* must be IPID */
2228 					x = ntohs(ip->ip_id);
2229 				    if (cmdlen == 1) {
2230 					match = (cmd->arg1 == x);
2231 					break;
2232 				    }
2233 				    /* otherwise we have ranges */
2234 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2235 				    i = cmdlen - 1;
2236 				    for (; !match && i>0; i--, p += 2)
2237 					match = (x >= p[0] && x <= p[1]);
2238 				}
2239 				break;
2240 
2241 			case O_IPPRECEDENCE:
2242 				match = (is_ipv4 &&
2243 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2244 				break;
2245 
2246 			case O_IPTOS:
2247 				match = (is_ipv4 &&
2248 				    flags_match(cmd, ip->ip_tos));
2249 				break;
2250 
2251 			case O_DSCP:
2252 			    {
2253 				uint32_t *p;
2254 				uint16_t x;
2255 
2256 				p = ((ipfw_insn_u32 *)cmd)->d;
2257 
2258 				if (is_ipv4)
2259 					x = ip->ip_tos >> 2;
2260 				else if (is_ipv6) {
2261 					uint8_t *v;
2262 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2263 					x = (*v & 0x0F) << 2;
2264 					v++;
2265 					x |= *v >> 6;
2266 				} else
2267 					break;
2268 
2269 				/* DSCP bitmask is stored as low_u32 high_u32 */
2270 				if (x >= 32)
2271 					match = *(p + 1) & (1 << (x - 32));
2272 				else
2273 					match = *p & (1 << x);
2274 			    }
2275 				break;
2276 
2277 			case O_TCPDATALEN:
2278 				if (proto == IPPROTO_TCP && offset == 0) {
2279 				    struct tcphdr *tcp;
2280 				    uint16_t x;
2281 				    uint16_t *p;
2282 				    int i;
2283 #ifdef INET6
2284 				    if (is_ipv6) {
2285 					    struct ip6_hdr *ip6;
2286 
2287 					    ip6 = (struct ip6_hdr *)ip;
2288 					    if (ip6->ip6_plen == 0) {
2289 						    /*
2290 						     * Jumbo payload is not
2291 						     * supported by this
2292 						     * opcode.
2293 						     */
2294 						    break;
2295 					    }
2296 					    x = iplen - hlen;
2297 				    } else
2298 #endif /* INET6 */
2299 					    x = iplen - (ip->ip_hl << 2);
2300 				    tcp = TCP(ulp);
2301 				    x -= tcp->th_off << 2;
2302 				    if (cmdlen == 1) {
2303 					match = (cmd->arg1 == x);
2304 					break;
2305 				    }
2306 				    /* otherwise we have ranges */
2307 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2308 				    i = cmdlen - 1;
2309 				    for (; !match && i>0; i--, p += 2)
2310 					match = (x >= p[0] && x <= p[1]);
2311 				}
2312 				break;
2313 
2314 			case O_TCPFLAGS:
2315 				match = (proto == IPPROTO_TCP && offset == 0 &&
2316 				    flags_match(cmd, TCP(ulp)->th_flags));
2317 				break;
2318 
2319 			case O_TCPOPTS:
2320 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
2321 					PULLUP_LEN_LOCKED(hlen, ulp,
2322 					    (TCP(ulp)->th_off << 2));
2323 					match = tcpopts_match(TCP(ulp), cmd);
2324 				}
2325 				break;
2326 
2327 			case O_TCPSEQ:
2328 				match = (proto == IPPROTO_TCP && offset == 0 &&
2329 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2330 					TCP(ulp)->th_seq);
2331 				break;
2332 
2333 			case O_TCPACK:
2334 				match = (proto == IPPROTO_TCP && offset == 0 &&
2335 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2336 					TCP(ulp)->th_ack);
2337 				break;
2338 
2339 			case O_TCPMSS:
2340 				if (proto == IPPROTO_TCP &&
2341 				    (args->f_id._flags & TH_SYN) != 0 &&
2342 				    ulp != NULL) {
2343 					uint16_t mss, *p;
2344 					int i;
2345 
2346 					PULLUP_LEN_LOCKED(hlen, ulp,
2347 					    (TCP(ulp)->th_off << 2));
2348 					if ((tcpopts_parse(TCP(ulp), &mss) &
2349 					    IP_FW_TCPOPT_MSS) == 0)
2350 						break;
2351 					if (cmdlen == 1) {
2352 						match = (cmd->arg1 == mss);
2353 						break;
2354 					}
2355 					/* Otherwise we have ranges. */
2356 					p = ((ipfw_insn_u16 *)cmd)->ports;
2357 					i = cmdlen - 1;
2358 					for (; !match && i > 0; i--, p += 2)
2359 						match = (mss >= p[0] &&
2360 						    mss <= p[1]);
2361 				}
2362 				break;
2363 
2364 			case O_TCPWIN:
2365 				if (proto == IPPROTO_TCP && offset == 0) {
2366 				    uint16_t x;
2367 				    uint16_t *p;
2368 				    int i;
2369 
2370 				    x = ntohs(TCP(ulp)->th_win);
2371 				    if (cmdlen == 1) {
2372 					match = (cmd->arg1 == x);
2373 					break;
2374 				    }
2375 				    /* Otherwise we have ranges. */
2376 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2377 				    i = cmdlen - 1;
2378 				    for (; !match && i > 0; i--, p += 2)
2379 					match = (x >= p[0] && x <= p[1]);
2380 				}
2381 				break;
2382 
2383 			case O_ESTAB:
2384 				/* reject packets which have SYN only */
2385 				/* XXX should i also check for TH_ACK ? */
2386 				match = (proto == IPPROTO_TCP && offset == 0 &&
2387 				    (TCP(ulp)->th_flags &
2388 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2389 				break;
2390 
2391 			case O_ALTQ: {
2392 				struct pf_mtag *at;
2393 				struct m_tag *mtag;
2394 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2395 
2396 				/*
2397 				 * ALTQ uses mbuf tags from another
2398 				 * packet filtering system - pf(4).
2399 				 * We allocate a tag in its format
2400 				 * and fill it in, pretending to be pf(4).
2401 				 */
2402 				match = 1;
2403 				at = pf_find_mtag(m);
2404 				if (at != NULL && at->qid != 0)
2405 					break;
2406 				mtag = m_tag_get(PACKET_TAG_PF,
2407 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2408 				if (mtag == NULL) {
2409 					/*
2410 					 * Let the packet fall back to the
2411 					 * default ALTQ.
2412 					 */
2413 					break;
2414 				}
2415 				m_tag_prepend(m, mtag);
2416 				at = (struct pf_mtag *)(mtag + 1);
2417 				at->qid = altq->qid;
2418 				at->hdr = ip;
2419 				break;
2420 			}
2421 
2422 			case O_LOG:
2423 				ipfw_log(chain, f, hlen, args,
2424 				    offset | ip6f_mf, tablearg, ip);
2425 				match = 1;
2426 				break;
2427 
2428 			case O_PROB:
2429 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2430 				break;
2431 
2432 			case O_VERREVPATH:
2433 				/* Outgoing packets automatically pass/match */
2434 				match = (args->flags & IPFW_ARGS_OUT ||
2435 				    (
2436 #ifdef INET6
2437 				    is_ipv6 ?
2438 					verify_path6(&(args->f_id.src_ip6),
2439 					    iif, args->f_id.fib) :
2440 #endif
2441 				    verify_path(src_ip, iif, args->f_id.fib)));
2442 				break;
2443 
2444 			case O_VERSRCREACH:
2445 				/* Outgoing packets automatically pass/match */
2446 				match = (hlen > 0 && ((oif != NULL) || (
2447 #ifdef INET6
2448 				    is_ipv6 ?
2449 				        verify_path6(&(args->f_id.src_ip6),
2450 				            NULL, args->f_id.fib) :
2451 #endif
2452 				    verify_path(src_ip, NULL, args->f_id.fib))));
2453 				break;
2454 
2455 			case O_ANTISPOOF:
2456 				/* Outgoing packets automatically pass/match */
2457 				if (oif == NULL && hlen > 0 &&
2458 				    (  (is_ipv4 && in_localaddr(src_ip))
2459 #ifdef INET6
2460 				    || (is_ipv6 &&
2461 				        in6_localaddr(&(args->f_id.src_ip6)))
2462 #endif
2463 				    ))
2464 					match =
2465 #ifdef INET6
2466 					    is_ipv6 ? verify_path6(
2467 					        &(args->f_id.src_ip6), iif,
2468 						args->f_id.fib) :
2469 #endif
2470 					    verify_path(src_ip, iif,
2471 					        args->f_id.fib);
2472 				else
2473 					match = 1;
2474 				break;
2475 
2476 			case O_IPSEC:
2477 				match = (m_tag_find(m,
2478 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2479 				/* otherwise no match */
2480 				break;
2481 
2482 #ifdef INET6
2483 			case O_IP6_SRC:
2484 				match = is_ipv6 &&
2485 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2486 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2487 				break;
2488 
2489 			case O_IP6_DST:
2490 				match = is_ipv6 &&
2491 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2492 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2493 				break;
2494 			case O_IP6_SRC_MASK:
2495 			case O_IP6_DST_MASK:
2496 				if (is_ipv6) {
2497 					int i = cmdlen - 1;
2498 					struct in6_addr p;
2499 					struct in6_addr *d =
2500 					    &((ipfw_insn_ip6 *)cmd)->addr6;
2501 
2502 					for (; !match && i > 0; d += 2,
2503 					    i -= F_INSN_SIZE(struct in6_addr)
2504 					    * 2) {
2505 						p = (cmd->opcode ==
2506 						    O_IP6_SRC_MASK) ?
2507 						    args->f_id.src_ip6:
2508 						    args->f_id.dst_ip6;
2509 						APPLY_MASK(&p, &d[1]);
2510 						match =
2511 						    IN6_ARE_ADDR_EQUAL(&d[0],
2512 						    &p);
2513 					}
2514 				}
2515 				break;
2516 
2517 			case O_FLOW6ID:
2518 				match = is_ipv6 &&
2519 				    flow6id_match(args->f_id.flow_id6,
2520 				    (ipfw_insn_u32 *) cmd);
2521 				break;
2522 
2523 			case O_EXT_HDR:
2524 				match = is_ipv6 &&
2525 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
2526 				break;
2527 
2528 			case O_IP6:
2529 				match = is_ipv6;
2530 				break;
2531 #endif
2532 
2533 			case O_IP4:
2534 				match = is_ipv4;
2535 				break;
2536 
2537 			case O_TAG: {
2538 				struct m_tag *mtag;
2539 				uint32_t tag = TARG(cmd->arg1, tag);
2540 
2541 				/* Packet is already tagged with this tag? */
2542 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2543 
2544 				/* We have `untag' action when F_NOT flag is
2545 				 * present. And we must remove this mtag from
2546 				 * mbuf and reset `match' to zero (`match' will
2547 				 * be inversed later).
2548 				 * Otherwise we should allocate new mtag and
2549 				 * push it into mbuf.
2550 				 */
2551 				if (cmd->len & F_NOT) { /* `untag' action */
2552 					if (mtag != NULL)
2553 						m_tag_delete(m, mtag);
2554 					match = 0;
2555 				} else {
2556 					if (mtag == NULL) {
2557 						mtag = m_tag_alloc( MTAG_IPFW,
2558 						    tag, 0, M_NOWAIT);
2559 						if (mtag != NULL)
2560 							m_tag_prepend(m, mtag);
2561 					}
2562 					match = 1;
2563 				}
2564 				break;
2565 			}
2566 
2567 			case O_FIB: /* try match the specified fib */
2568 				if (args->f_id.fib == cmd->arg1)
2569 					match = 1;
2570 				break;
2571 
2572 			case O_SOCKARG:	{
2573 #ifndef USERSPACE	/* not supported in userspace */
2574 				struct inpcb *inp = args->inp;
2575 				struct inpcbinfo *pi;
2576 
2577 				if (is_ipv6) /* XXX can we remove this ? */
2578 					break;
2579 
2580 				if (proto == IPPROTO_TCP)
2581 					pi = &V_tcbinfo;
2582 				else if (proto == IPPROTO_UDP)
2583 					pi = &V_udbinfo;
2584 				else if (proto == IPPROTO_UDPLITE)
2585 					pi = &V_ulitecbinfo;
2586 				else
2587 					break;
2588 
2589 				/*
2590 				 * XXXRW: so_user_cookie should almost
2591 				 * certainly be inp_user_cookie?
2592 				 */
2593 
2594 				/* For incoming packet, lookup up the
2595 				inpcb using the src/dest ip/port tuple */
2596 				if (inp == NULL) {
2597 					inp = in_pcblookup(pi,
2598 						src_ip, htons(src_port),
2599 						dst_ip, htons(dst_port),
2600 						INPLOOKUP_RLOCKPCB, NULL);
2601 					if (inp != NULL) {
2602 						tablearg =
2603 						    inp->inp_socket->so_user_cookie;
2604 						if (tablearg)
2605 							match = 1;
2606 						INP_RUNLOCK(inp);
2607 					}
2608 				} else {
2609 					if (inp->inp_socket) {
2610 						tablearg =
2611 						    inp->inp_socket->so_user_cookie;
2612 						if (tablearg)
2613 							match = 1;
2614 					}
2615 				}
2616 #endif /* !USERSPACE */
2617 				break;
2618 			}
2619 
2620 			case O_TAGGED: {
2621 				struct m_tag *mtag;
2622 				uint32_t tag = TARG(cmd->arg1, tag);
2623 
2624 				if (cmdlen == 1) {
2625 					match = m_tag_locate(m, MTAG_IPFW,
2626 					    tag, NULL) != NULL;
2627 					break;
2628 				}
2629 
2630 				/* we have ranges */
2631 				for (mtag = m_tag_first(m);
2632 				    mtag != NULL && !match;
2633 				    mtag = m_tag_next(m, mtag)) {
2634 					uint16_t *p;
2635 					int i;
2636 
2637 					if (mtag->m_tag_cookie != MTAG_IPFW)
2638 						continue;
2639 
2640 					p = ((ipfw_insn_u16 *)cmd)->ports;
2641 					i = cmdlen - 1;
2642 					for(; !match && i > 0; i--, p += 2)
2643 						match =
2644 						    mtag->m_tag_id >= p[0] &&
2645 						    mtag->m_tag_id <= p[1];
2646 				}
2647 				break;
2648 			}
2649 
2650 			/*
2651 			 * The second set of opcodes represents 'actions',
2652 			 * i.e. the terminal part of a rule once the packet
2653 			 * matches all previous patterns.
2654 			 * Typically there is only one action for each rule,
2655 			 * and the opcode is stored at the end of the rule
2656 			 * (but there are exceptions -- see below).
2657 			 *
2658 			 * In general, here we set retval and terminate the
2659 			 * outer loop (would be a 'break 3' in some language,
2660 			 * but we need to set l=0, done=1)
2661 			 *
2662 			 * Exceptions:
2663 			 * O_COUNT and O_SKIPTO actions:
2664 			 *   instead of terminating, we jump to the next rule
2665 			 *   (setting l=0), or to the SKIPTO target (setting
2666 			 *   f/f_len, cmd and l as needed), respectively.
2667 			 *
2668 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2669 			 *   perform some action and set match = 1;
2670 			 *
2671 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2672 			 *   not real 'actions', and are stored right
2673 			 *   before the 'action' part of the rule (one
2674 			 *   exception is O_SKIP_ACTION which could be
2675 			 *   between these opcodes and 'action' one).
2676 			 *   These opcodes try to install an entry in the
2677 			 *   state tables; if successful, we continue with
2678 			 *   the next opcode (match=1; break;), otherwise
2679 			 *   the packet must be dropped (set retval,
2680 			 *   break loops with l=0, done=1)
2681 			 *
2682 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2683 			 *   cause a lookup of the state table, and a jump
2684 			 *   to the 'action' part of the parent rule
2685 			 *   if an entry is found, or
2686 			 *   (CHECK_STATE only) a jump to the next rule if
2687 			 *   the entry is not found.
2688 			 *   The result of the lookup is cached so that
2689 			 *   further instances of these opcodes become NOPs.
2690 			 *   The jump to the next rule is done by setting
2691 			 *   l=0, cmdlen=0.
2692 			 *
2693 			 * O_SKIP_ACTION: this opcode is not a real 'action'
2694 			 *  either, and is stored right before the 'action'
2695 			 *  part of the rule, right after the O_KEEP_STATE
2696 			 *  opcode. It causes match failure so the real
2697 			 *  'action' could be executed only if the rule
2698 			 *  is checked via dynamic rule from the state
2699 			 *  table, as in such case execution starts
2700 			 *  from the true 'action' opcode directly.
2701 			 *
2702 			 */
2703 			case O_LIMIT:
2704 			case O_KEEP_STATE:
2705 				if (ipfw_dyn_install_state(chain, f,
2706 				    (ipfw_insn_limit *)cmd, args, ulp,
2707 				    pktlen, &dyn_info, tablearg)) {
2708 					/* error or limit violation */
2709 					retval = IP_FW_DENY;
2710 					l = 0;	/* exit inner loop */
2711 					done = 1; /* exit outer loop */
2712 				}
2713 				match = 1;
2714 				break;
2715 
2716 			case O_PROBE_STATE:
2717 			case O_CHECK_STATE:
2718 				/*
2719 				 * dynamic rules are checked at the first
2720 				 * keep-state or check-state occurrence,
2721 				 * with the result being stored in dyn_info.
2722 				 * The compiler introduces a PROBE_STATE
2723 				 * instruction for us when we have a
2724 				 * KEEP_STATE (because PROBE_STATE needs
2725 				 * to be run first).
2726 				 */
2727 				if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2728 				    (q = ipfw_dyn_lookup_state(args, ulp,
2729 				    pktlen, cmd, &dyn_info)) != NULL) {
2730 					/*
2731 					 * Found dynamic entry, jump to the
2732 					 * 'action' part of the parent rule
2733 					 * by setting f, cmd, l and clearing
2734 					 * cmdlen.
2735 					 */
2736 					f = q;
2737 					f_pos = dyn_info.f_pos;
2738 					cmd = ACTION_PTR(f);
2739 					l = f->cmd_len - f->act_ofs;
2740 					cmdlen = 0;
2741 					match = 1;
2742 					break;
2743 				}
2744 				/*
2745 				 * Dynamic entry not found. If CHECK_STATE,
2746 				 * skip to next rule, if PROBE_STATE just
2747 				 * ignore and continue with next opcode.
2748 				 */
2749 				if (cmd->opcode == O_CHECK_STATE)
2750 					l = 0;	/* exit inner loop */
2751 				match = 1;
2752 				break;
2753 
2754 			case O_SKIP_ACTION:
2755 				match = 0;	/* skip to the next rule */
2756 				l = 0;		/* exit inner loop */
2757 				break;
2758 
2759 			case O_ACCEPT:
2760 				retval = 0;	/* accept */
2761 				l = 0;		/* exit inner loop */
2762 				done = 1;	/* exit outer loop */
2763 				break;
2764 
2765 			case O_PIPE:
2766 			case O_QUEUE:
2767 				set_match(args, f_pos, chain);
2768 				args->rule.info = TARG(cmd->arg1, pipe);
2769 				if (cmd->opcode == O_PIPE)
2770 					args->rule.info |= IPFW_IS_PIPE;
2771 				if (V_fw_one_pass)
2772 					args->rule.info |= IPFW_ONEPASS;
2773 				retval = IP_FW_DUMMYNET;
2774 				l = 0;          /* exit inner loop */
2775 				done = 1;       /* exit outer loop */
2776 				break;
2777 
2778 			case O_DIVERT:
2779 			case O_TEE:
2780 				if (args->flags & IPFW_ARGS_ETHER)
2781 					break;	/* not on layer 2 */
2782 				/* otherwise this is terminal */
2783 				l = 0;		/* exit inner loop */
2784 				done = 1;	/* exit outer loop */
2785 				retval = (cmd->opcode == O_DIVERT) ?
2786 					IP_FW_DIVERT : IP_FW_TEE;
2787 				set_match(args, f_pos, chain);
2788 				args->rule.info = TARG(cmd->arg1, divert);
2789 				break;
2790 
2791 			case O_COUNT:
2792 				IPFW_INC_RULE_COUNTER(f, pktlen);
2793 				l = 0;		/* exit inner loop */
2794 				break;
2795 
2796 			case O_SKIPTO:
2797 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2798 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2799 			    /*
2800 			     * Skip disabled rules, and re-enter
2801 			     * the inner loop with the correct
2802 			     * f_pos, f, l and cmd.
2803 			     * Also clear cmdlen and skip_or
2804 			     */
2805 			    for (; f_pos < chain->n_rules - 1 &&
2806 				    (V_set_disable &
2807 				     (1 << chain->map[f_pos]->set));
2808 				    f_pos++)
2809 				;
2810 			    /* Re-enter the inner loop at the skipto rule. */
2811 			    f = chain->map[f_pos];
2812 			    l = f->cmd_len;
2813 			    cmd = f->cmd;
2814 			    match = 1;
2815 			    cmdlen = 0;
2816 			    skip_or = 0;
2817 			    continue;
2818 			    break;	/* not reached */
2819 
2820 			case O_CALLRETURN: {
2821 				/*
2822 				 * Implementation of `subroutine' call/return,
2823 				 * in the stack carried in an mbuf tag. This
2824 				 * is different from `skipto' in that any call
2825 				 * address is possible (`skipto' must prevent
2826 				 * backward jumps to avoid endless loops).
2827 				 * We have `return' action when F_NOT flag is
2828 				 * present. The `m_tag_id' field is used as
2829 				 * stack pointer.
2830 				 */
2831 				struct m_tag *mtag;
2832 				uint16_t jmpto, *stack;
2833 
2834 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2835 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2836 				/*
2837 				 * Hand-rolled version of m_tag_locate() with
2838 				 * wildcard `type'.
2839 				 * If not already tagged, allocate new tag.
2840 				 */
2841 				mtag = m_tag_first(m);
2842 				while (mtag != NULL) {
2843 					if (mtag->m_tag_cookie ==
2844 					    MTAG_IPFW_CALL)
2845 						break;
2846 					mtag = m_tag_next(m, mtag);
2847 				}
2848 				if (mtag == NULL && IS_CALL) {
2849 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2850 					    IPFW_CALLSTACK_SIZE *
2851 					    sizeof(uint16_t), M_NOWAIT);
2852 					if (mtag != NULL)
2853 						m_tag_prepend(m, mtag);
2854 				}
2855 
2856 				/*
2857 				 * On error both `call' and `return' just
2858 				 * continue with next rule.
2859 				 */
2860 				if (IS_RETURN && (mtag == NULL ||
2861 				    mtag->m_tag_id == 0)) {
2862 					l = 0;		/* exit inner loop */
2863 					break;
2864 				}
2865 				if (IS_CALL && (mtag == NULL ||
2866 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2867 					printf("ipfw: call stack error, "
2868 					    "go to next rule\n");
2869 					l = 0;		/* exit inner loop */
2870 					break;
2871 				}
2872 
2873 				IPFW_INC_RULE_COUNTER(f, pktlen);
2874 				stack = (uint16_t *)(mtag + 1);
2875 
2876 				/*
2877 				 * The `call' action may use cached f_pos
2878 				 * (in f->next_rule), whose version is written
2879 				 * in f->next_rule.
2880 				 * The `return' action, however, doesn't have
2881 				 * fixed jump address in cmd->arg1 and can't use
2882 				 * cache.
2883 				 */
2884 				if (IS_CALL) {
2885 					stack[mtag->m_tag_id] = f->rulenum;
2886 					mtag->m_tag_id++;
2887 			    		f_pos = JUMP(chain, f, cmd->arg1,
2888 					    tablearg, 1);
2889 				} else {	/* `return' action */
2890 					mtag->m_tag_id--;
2891 					jmpto = stack[mtag->m_tag_id] + 1;
2892 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2893 				}
2894 
2895 				/*
2896 				 * Skip disabled rules, and re-enter
2897 				 * the inner loop with the correct
2898 				 * f_pos, f, l and cmd.
2899 				 * Also clear cmdlen and skip_or
2900 				 */
2901 				for (; f_pos < chain->n_rules - 1 &&
2902 				    (V_set_disable &
2903 				    (1 << chain->map[f_pos]->set)); f_pos++)
2904 					;
2905 				/* Re-enter the inner loop at the dest rule. */
2906 				f = chain->map[f_pos];
2907 				l = f->cmd_len;
2908 				cmd = f->cmd;
2909 				cmdlen = 0;
2910 				skip_or = 0;
2911 				continue;
2912 				break;	/* NOTREACHED */
2913 			}
2914 #undef IS_CALL
2915 #undef IS_RETURN
2916 
2917 			case O_REJECT:
2918 				/*
2919 				 * Drop the packet and send a reject notice
2920 				 * if the packet is not ICMP (or is an ICMP
2921 				 * query), and it is not multicast/broadcast.
2922 				 */
2923 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2924 				    (proto != IPPROTO_ICMP ||
2925 				     is_icmp_query(ICMP(ulp))) &&
2926 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2927 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2928 					send_reject(args, cmd->arg1, iplen, ip);
2929 					m = args->m;
2930 				}
2931 				/* FALLTHROUGH */
2932 #ifdef INET6
2933 			case O_UNREACH6:
2934 				if (hlen > 0 && is_ipv6 &&
2935 				    ((offset & IP6F_OFF_MASK) == 0) &&
2936 				    (proto != IPPROTO_ICMPV6 ||
2937 				     (is_icmp6_query(icmp6_type) == 1)) &&
2938 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2939 				    !IN6_IS_ADDR_MULTICAST(
2940 					&args->f_id.dst_ip6)) {
2941 					send_reject6(args,
2942 					    cmd->opcode == O_REJECT ?
2943 					    map_icmp_unreach(cmd->arg1):
2944 					    cmd->arg1, hlen,
2945 					    (struct ip6_hdr *)ip);
2946 					m = args->m;
2947 				}
2948 				/* FALLTHROUGH */
2949 #endif
2950 			case O_DENY:
2951 				retval = IP_FW_DENY;
2952 				l = 0;		/* exit inner loop */
2953 				done = 1;	/* exit outer loop */
2954 				break;
2955 
2956 			case O_FORWARD_IP:
2957 				if (args->flags & IPFW_ARGS_ETHER)
2958 					break;	/* not valid on layer2 pkts */
2959 				if (q != f ||
2960 				    dyn_info.direction == MATCH_FORWARD) {
2961 				    struct sockaddr_in *sa;
2962 
2963 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2964 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2965 #ifdef INET6
2966 					/*
2967 					 * We use O_FORWARD_IP opcode for
2968 					 * fwd rule with tablearg, but tables
2969 					 * now support IPv6 addresses. And
2970 					 * when we are inspecting IPv6 packet,
2971 					 * we can use nh6 field from
2972 					 * table_value as next_hop6 address.
2973 					 */
2974 					if (is_ipv6) {
2975 						struct ip_fw_nh6 *nh6;
2976 
2977 						args->flags |= IPFW_ARGS_NH6;
2978 						nh6 = &args->hopstore6;
2979 						nh6->sin6_addr = TARG_VAL(
2980 						    chain, tablearg, nh6);
2981 						nh6->sin6_port = sa->sin_port;
2982 						nh6->sin6_scope_id = TARG_VAL(
2983 						    chain, tablearg, zoneid);
2984 					} else
2985 #endif
2986 					{
2987 						args->flags |= IPFW_ARGS_NH4;
2988 						args->hopstore.sin_port =
2989 						    sa->sin_port;
2990 						sa = &args->hopstore;
2991 						sa->sin_family = AF_INET;
2992 						sa->sin_len = sizeof(*sa);
2993 						sa->sin_addr.s_addr = htonl(
2994 						    TARG_VAL(chain, tablearg,
2995 						    nh4));
2996 					}
2997 				    } else {
2998 					    args->flags |= IPFW_ARGS_NH4PTR;
2999 					    args->next_hop = sa;
3000 				    }
3001 				}
3002 				retval = IP_FW_PASS;
3003 				l = 0;          /* exit inner loop */
3004 				done = 1;       /* exit outer loop */
3005 				break;
3006 
3007 #ifdef INET6
3008 			case O_FORWARD_IP6:
3009 				if (args->flags & IPFW_ARGS_ETHER)
3010 					break;	/* not valid on layer2 pkts */
3011 				if (q != f ||
3012 				    dyn_info.direction == MATCH_FORWARD) {
3013 					struct sockaddr_in6 *sin6;
3014 
3015 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3016 					args->flags |= IPFW_ARGS_NH6PTR;
3017 					args->next_hop6 = sin6;
3018 				}
3019 				retval = IP_FW_PASS;
3020 				l = 0;		/* exit inner loop */
3021 				done = 1;	/* exit outer loop */
3022 				break;
3023 #endif
3024 
3025 			case O_NETGRAPH:
3026 			case O_NGTEE:
3027 				set_match(args, f_pos, chain);
3028 				args->rule.info = TARG(cmd->arg1, netgraph);
3029 				if (V_fw_one_pass)
3030 					args->rule.info |= IPFW_ONEPASS;
3031 				retval = (cmd->opcode == O_NETGRAPH) ?
3032 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
3033 				l = 0;          /* exit inner loop */
3034 				done = 1;       /* exit outer loop */
3035 				break;
3036 
3037 			case O_SETFIB: {
3038 				uint32_t fib;
3039 
3040 				IPFW_INC_RULE_COUNTER(f, pktlen);
3041 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
3042 				if (fib >= rt_numfibs)
3043 					fib = 0;
3044 				M_SETFIB(m, fib);
3045 				args->f_id.fib = fib; /* XXX */
3046 				l = 0;		/* exit inner loop */
3047 				break;
3048 		        }
3049 
3050 			case O_SETDSCP: {
3051 				uint16_t code;
3052 
3053 				code = TARG(cmd->arg1, dscp) & 0x3F;
3054 				l = 0;		/* exit inner loop */
3055 				if (is_ipv4) {
3056 					uint16_t old;
3057 
3058 					old = *(uint16_t *)ip;
3059 					ip->ip_tos = (code << 2) |
3060 					    (ip->ip_tos & 0x03);
3061 					ip->ip_sum = cksum_adjust(ip->ip_sum,
3062 					    old, *(uint16_t *)ip);
3063 				} else if (is_ipv6) {
3064 					uint8_t *v;
3065 
3066 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
3067 					*v = (*v & 0xF0) | (code >> 2);
3068 					v++;
3069 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
3070 				} else
3071 					break;
3072 
3073 				IPFW_INC_RULE_COUNTER(f, pktlen);
3074 				break;
3075 			}
3076 
3077 			case O_NAT:
3078 				l = 0;          /* exit inner loop */
3079 				done = 1;       /* exit outer loop */
3080 				/*
3081 				 * Ensure that we do not invoke NAT handler for
3082 				 * non IPv4 packets. Libalias expects only IPv4.
3083 				 */
3084 				if (!is_ipv4 || !IPFW_NAT_LOADED) {
3085 				    retval = IP_FW_DENY;
3086 				    break;
3087 				}
3088 
3089 				struct cfg_nat *t;
3090 				int nat_id;
3091 
3092 				args->rule.info = 0;
3093 				set_match(args, f_pos, chain);
3094 				/* Check if this is 'global' nat rule */
3095 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3096 					retval = ipfw_nat_ptr(args, NULL, m);
3097 					break;
3098 				}
3099 				t = ((ipfw_insn_nat *)cmd)->nat;
3100 				if (t == NULL) {
3101 					nat_id = TARG(cmd->arg1, nat);
3102 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3103 
3104 					if (t == NULL) {
3105 					    retval = IP_FW_DENY;
3106 					    break;
3107 					}
3108 					if (cmd->arg1 != IP_FW_TARG)
3109 					    ((ipfw_insn_nat *)cmd)->nat = t;
3110 				}
3111 				retval = ipfw_nat_ptr(args, t, m);
3112 				break;
3113 
3114 			case O_REASS: {
3115 				int ip_off;
3116 
3117 				l = 0;	/* in any case exit inner loop */
3118 				if (is_ipv6) /* IPv6 is not supported yet */
3119 					break;
3120 				IPFW_INC_RULE_COUNTER(f, pktlen);
3121 				ip_off = ntohs(ip->ip_off);
3122 
3123 				/* if not fragmented, go to next rule */
3124 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3125 				    break;
3126 
3127 				args->m = m = ip_reass(m);
3128 
3129 				/*
3130 				 * do IP header checksum fixup.
3131 				 */
3132 				if (m == NULL) { /* fragment got swallowed */
3133 				    retval = IP_FW_DENY;
3134 				} else { /* good, packet complete */
3135 				    int hlen;
3136 
3137 				    ip = mtod(m, struct ip *);
3138 				    hlen = ip->ip_hl << 2;
3139 				    ip->ip_sum = 0;
3140 				    if (hlen == sizeof(struct ip))
3141 					ip->ip_sum = in_cksum_hdr(ip);
3142 				    else
3143 					ip->ip_sum = in_cksum(m, hlen);
3144 				    retval = IP_FW_REASS;
3145 				    args->rule.info = 0;
3146 				    set_match(args, f_pos, chain);
3147 				}
3148 				done = 1;	/* exit outer loop */
3149 				break;
3150 			}
3151 			case O_EXTERNAL_ACTION:
3152 				l = 0; /* in any case exit inner loop */
3153 				retval = ipfw_run_eaction(chain, args,
3154 				    cmd, &done);
3155 				/*
3156 				 * If both @retval and @done are zero,
3157 				 * consider this as rule matching and
3158 				 * update counters.
3159 				 */
3160 				if (retval == 0 && done == 0) {
3161 					IPFW_INC_RULE_COUNTER(f, pktlen);
3162 					/*
3163 					 * Reset the result of the last
3164 					 * dynamic state lookup.
3165 					 * External action can change
3166 					 * @args content, and it may be
3167 					 * used for new state lookup later.
3168 					 */
3169 					DYN_INFO_INIT(&dyn_info);
3170 				}
3171 				break;
3172 
3173 			default:
3174 				panic("-- unknown opcode %d\n", cmd->opcode);
3175 			} /* end of switch() on opcodes */
3176 			/*
3177 			 * if we get here with l=0, then match is irrelevant.
3178 			 */
3179 
3180 			if (cmd->len & F_NOT)
3181 				match = !match;
3182 
3183 			if (match) {
3184 				if (cmd->len & F_OR)
3185 					skip_or = 1;
3186 			} else {
3187 				if (!(cmd->len & F_OR)) /* not an OR block, */
3188 					break;		/* try next rule    */
3189 			}
3190 
3191 		}	/* end of inner loop, scan opcodes */
3192 #undef PULLUP_LEN
3193 #undef PULLUP_LEN_LOCKED
3194 
3195 		if (done)
3196 			break;
3197 
3198 /* next_rule:; */	/* try next rule		*/
3199 
3200 	}		/* end of outer for, scan rules */
3201 
3202 	if (done) {
3203 		struct ip_fw *rule = chain->map[f_pos];
3204 		/* Update statistics */
3205 		IPFW_INC_RULE_COUNTER(rule, pktlen);
3206 	} else {
3207 		retval = IP_FW_DENY;
3208 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3209 	}
3210 	IPFW_PF_RUNLOCK(chain);
3211 #ifdef __FreeBSD__
3212 	if (ucred_cache != NULL)
3213 		crfree(ucred_cache);
3214 #endif
3215 	return (retval);
3216 
3217 pullup_failed:
3218 	if (V_fw_verbose)
3219 		printf("ipfw: pullup failed\n");
3220 	return (IP_FW_DENY);
3221 }
3222 
3223 /*
3224  * Set maximum number of tables that can be used in given VNET ipfw instance.
3225  */
3226 #ifdef SYSCTL_NODE
3227 static int
3228 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3229 {
3230 	int error;
3231 	unsigned int ntables;
3232 
3233 	ntables = V_fw_tables_max;
3234 
3235 	error = sysctl_handle_int(oidp, &ntables, 0, req);
3236 	/* Read operation or some error */
3237 	if ((error != 0) || (req->newptr == NULL))
3238 		return (error);
3239 
3240 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
3241 }
3242 
3243 /*
3244  * Switches table namespace between global and per-set.
3245  */
3246 static int
3247 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3248 {
3249 	int error;
3250 	unsigned int sets;
3251 
3252 	sets = V_fw_tables_sets;
3253 
3254 	error = sysctl_handle_int(oidp, &sets, 0, req);
3255 	/* Read operation or some error */
3256 	if ((error != 0) || (req->newptr == NULL))
3257 		return (error);
3258 
3259 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3260 }
3261 #endif
3262 
3263 /*
3264  * Module and VNET glue
3265  */
3266 
3267 /*
3268  * Stuff that must be initialised only on boot or module load
3269  */
3270 static int
3271 ipfw_init(void)
3272 {
3273 	int error = 0;
3274 
3275 	/*
3276  	 * Only print out this stuff the first time around,
3277 	 * when called from the sysinit code.
3278 	 */
3279 	printf("ipfw2 "
3280 #ifdef INET6
3281 		"(+ipv6) "
3282 #endif
3283 		"initialized, divert %s, nat %s, "
3284 		"default to %s, logging ",
3285 #ifdef IPDIVERT
3286 		"enabled",
3287 #else
3288 		"loadable",
3289 #endif
3290 #ifdef IPFIREWALL_NAT
3291 		"enabled",
3292 #else
3293 		"loadable",
3294 #endif
3295 		default_to_accept ? "accept" : "deny");
3296 
3297 	/*
3298 	 * Note: V_xxx variables can be accessed here but the vnet specific
3299 	 * initializer may not have been called yet for the VIMAGE case.
3300 	 * Tuneables will have been processed. We will print out values for
3301 	 * the default vnet.
3302 	 * XXX This should all be rationalized AFTER 8.0
3303 	 */
3304 	if (V_fw_verbose == 0)
3305 		printf("disabled\n");
3306 	else if (V_verbose_limit == 0)
3307 		printf("unlimited\n");
3308 	else
3309 		printf("limited to %d packets/entry by default\n",
3310 		    V_verbose_limit);
3311 
3312 	/* Check user-supplied table count for validness */
3313 	if (default_fw_tables > IPFW_TABLES_MAX)
3314 	  default_fw_tables = IPFW_TABLES_MAX;
3315 
3316 	ipfw_init_sopt_handler();
3317 	ipfw_init_obj_rewriter();
3318 	ipfw_iface_init();
3319 	return (error);
3320 }
3321 
3322 /*
3323  * Called for the removal of the last instance only on module unload.
3324  */
3325 static void
3326 ipfw_destroy(void)
3327 {
3328 
3329 	ipfw_iface_destroy();
3330 	ipfw_destroy_sopt_handler();
3331 	ipfw_destroy_obj_rewriter();
3332 	printf("IP firewall unloaded\n");
3333 }
3334 
3335 /*
3336  * Stuff that must be initialized for every instance
3337  * (including the first of course).
3338  */
3339 static int
3340 vnet_ipfw_init(const void *unused)
3341 {
3342 	int error, first;
3343 	struct ip_fw *rule = NULL;
3344 	struct ip_fw_chain *chain;
3345 
3346 	chain = &V_layer3_chain;
3347 
3348 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3349 
3350 	/* First set up some values that are compile time options */
3351 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
3352 	V_fw_deny_unknown_exthdrs = 1;
3353 #ifdef IPFIREWALL_VERBOSE
3354 	V_fw_verbose = 1;
3355 #endif
3356 #ifdef IPFIREWALL_VERBOSE_LIMIT
3357 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3358 #endif
3359 #ifdef IPFIREWALL_NAT
3360 	LIST_INIT(&chain->nat);
3361 #endif
3362 
3363 	/* Init shared services hash table */
3364 	ipfw_init_srv(chain);
3365 
3366 	ipfw_init_counters();
3367 	/* Set initial number of tables */
3368 	V_fw_tables_max = default_fw_tables;
3369 	error = ipfw_init_tables(chain, first);
3370 	if (error) {
3371 		printf("ipfw2: setting up tables failed\n");
3372 		free(chain->map, M_IPFW);
3373 		free(rule, M_IPFW);
3374 		return (ENOSPC);
3375 	}
3376 
3377 	IPFW_LOCK_INIT(chain);
3378 
3379 	/* fill and insert the default rule */
3380 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3381 	rule->flags |= IPFW_RULE_NOOPT;
3382 	rule->cmd_len = 1;
3383 	rule->cmd[0].len = 1;
3384 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3385 	chain->default_rule = rule;
3386 	ipfw_add_protected_rule(chain, rule, 0);
3387 
3388 	ipfw_dyn_init(chain);
3389 	ipfw_eaction_init(chain, first);
3390 #ifdef LINEAR_SKIPTO
3391 	ipfw_init_skipto_cache(chain);
3392 #endif
3393 	ipfw_bpf_init(first);
3394 
3395 	/* First set up some values that are compile time options */
3396 	V_ipfw_vnet_ready = 1;		/* Open for business */
3397 
3398 	/*
3399 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3400 	 * Even if the latter two fail we still keep the module alive
3401 	 * because the sockopt and layer2 paths are still useful.
3402 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3403 	 * so we can ignore the exact return value and just set a flag.
3404 	 *
3405 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3406 	 * changes in the underlying (per-vnet) variables trigger
3407 	 * immediate hook()/unhook() calls.
3408 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
3409 	 * is checked on each packet because there are no pfil hooks.
3410 	 */
3411 	V_ip_fw_ctl_ptr = ipfw_ctl3;
3412 	error = ipfw_attach_hooks();
3413 	return (error);
3414 }
3415 
3416 /*
3417  * Called for the removal of each instance.
3418  */
3419 static int
3420 vnet_ipfw_uninit(const void *unused)
3421 {
3422 	struct ip_fw *reap;
3423 	struct ip_fw_chain *chain = &V_layer3_chain;
3424 	int i, last;
3425 
3426 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3427 	/*
3428 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
3429 	 * Then grab, release and grab again the WLOCK so we make
3430 	 * sure the update is propagated and nobody will be in.
3431 	 */
3432 	ipfw_detach_hooks();
3433 	V_ip_fw_ctl_ptr = NULL;
3434 
3435 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3436 
3437 	IPFW_UH_WLOCK(chain);
3438 	IPFW_UH_WUNLOCK(chain);
3439 
3440 	ipfw_dyn_uninit(0);	/* run the callout_drain */
3441 
3442 	IPFW_UH_WLOCK(chain);
3443 
3444 	reap = NULL;
3445 	IPFW_WLOCK(chain);
3446 	for (i = 0; i < chain->n_rules; i++)
3447 		ipfw_reap_add(chain, &reap, chain->map[i]);
3448 	free(chain->map, M_IPFW);
3449 #ifdef LINEAR_SKIPTO
3450 	ipfw_destroy_skipto_cache(chain);
3451 #endif
3452 	IPFW_WUNLOCK(chain);
3453 	IPFW_UH_WUNLOCK(chain);
3454 	ipfw_destroy_tables(chain, last);
3455 	ipfw_eaction_uninit(chain, last);
3456 	if (reap != NULL)
3457 		ipfw_reap_rules(reap);
3458 	vnet_ipfw_iface_destroy(chain);
3459 	ipfw_destroy_srv(chain);
3460 	IPFW_LOCK_DESTROY(chain);
3461 	ipfw_dyn_uninit(1);	/* free the remaining parts */
3462 	ipfw_destroy_counters();
3463 	ipfw_bpf_uninit(last);
3464 	return (0);
3465 }
3466 
3467 /*
3468  * Module event handler.
3469  * In general we have the choice of handling most of these events by the
3470  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3471  * use the SYSINIT handlers as they are more capable of expressing the
3472  * flow of control during module and vnet operations, so this is just
3473  * a skeleton. Note there is no SYSINIT equivalent of the module
3474  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3475  */
3476 static int
3477 ipfw_modevent(module_t mod, int type, void *unused)
3478 {
3479 	int err = 0;
3480 
3481 	switch (type) {
3482 	case MOD_LOAD:
3483 		/* Called once at module load or
3484 	 	 * system boot if compiled in. */
3485 		break;
3486 	case MOD_QUIESCE:
3487 		/* Called before unload. May veto unloading. */
3488 		break;
3489 	case MOD_UNLOAD:
3490 		/* Called during unload. */
3491 		break;
3492 	case MOD_SHUTDOWN:
3493 		/* Called during system shutdown. */
3494 		break;
3495 	default:
3496 		err = EOPNOTSUPP;
3497 		break;
3498 	}
3499 	return err;
3500 }
3501 
3502 static moduledata_t ipfwmod = {
3503 	"ipfw",
3504 	ipfw_modevent,
3505 	0
3506 };
3507 
3508 /* Define startup order. */
3509 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
3510 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
3511 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
3512 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
3513 
3514 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3515 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3516 MODULE_VERSION(ipfw, 3);
3517 /* should declare some dependencies here */
3518 
3519 /*
3520  * Starting up. Done in order after ipfwmod() has been called.
3521  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3522  */
3523 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3524 	    ipfw_init, NULL);
3525 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3526 	    vnet_ipfw_init, NULL);
3527 
3528 /*
3529  * Closing up shop. These are done in REVERSE ORDER, but still
3530  * after ipfwmod() has been called. Not called on reboot.
3531  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3532  * or when the module is unloaded.
3533  */
3534 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3535 	    ipfw_destroy, NULL);
3536 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3537 	    vnet_ipfw_uninit, NULL);
3538 /* end of file */
3539