xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision c66ec88fed842fbaad62c30d510644ceb7bd2d71)
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/sdt.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/syslog.h>
63 #include <sys/ucred.h>
64 #include <net/ethernet.h> /* for ETHERTYPE_IP */
65 #include <net/if.h>
66 #include <net/if_var.h>
67 #include <net/route.h>
68 #include <net/route/nhop.h>
69 #include <net/pfil.h>
70 #include <net/vnet.h>
71 
72 #include <netpfil/pf/pf_mtag.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_var.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 #include <netinet/ip_fw.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/pim.h>
83 #include <netinet/tcp_var.h>
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
86 #include <netinet/sctp.h>
87 #include <netinet/sctp_crc32.h>
88 #include <netinet/sctp_header.h>
89 
90 #include <netinet/ip6.h>
91 #include <netinet/icmp6.h>
92 #include <netinet/in_fib.h>
93 #ifdef INET6
94 #include <netinet6/in6_fib.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/scope6_var.h>
97 #include <netinet6/ip6_var.h>
98 #endif
99 
100 #include <net/if_gre.h> /* for struct grehdr */
101 
102 #include <netpfil/ipfw/ip_fw_private.h>
103 
104 #include <machine/in_cksum.h>	/* XXX for in_cksum */
105 
106 #ifdef MAC
107 #include <security/mac/mac_framework.h>
108 #endif
109 
110 #define	IPFW_PROBE(probe, arg0, arg1, arg2, arg3, arg4, arg5)		\
111     SDT_PROBE6(ipfw, , , probe, arg0, arg1, arg2, arg3, arg4, arg5)
112 
113 SDT_PROVIDER_DEFINE(ipfw);
114 SDT_PROBE_DEFINE6(ipfw, , , rule__matched,
115     "int",			/* retval */
116     "int",			/* af */
117     "void *",			/* src addr */
118     "void *",			/* dst addr */
119     "struct ip_fw_args *",	/* args */
120     "struct ip_fw *"		/* rule */);
121 
122 /*
123  * static variables followed by global ones.
124  * All ipfw global variables are here.
125  */
126 
127 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs);
128 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
129 
130 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1;
131 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
132 
133 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
134 static int default_to_accept = 1;
135 #else
136 static int default_to_accept;
137 #endif
138 
139 VNET_DEFINE(int, autoinc_step);
140 VNET_DEFINE(int, fw_one_pass) = 1;
141 
142 VNET_DEFINE(unsigned int, fw_tables_max);
143 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;	/* Don't use set-aware tables */
144 /* Use 128 tables by default */
145 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
146 
147 #ifndef LINEAR_SKIPTO
148 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
149     int tablearg, int jump_backwards);
150 #define	JUMP(ch, f, num, targ, back)	jump_fast(ch, f, num, targ, back)
151 #else
152 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
153     int tablearg, int jump_backwards);
154 #define	JUMP(ch, f, num, targ, back)	jump_linear(ch, f, num, targ, back)
155 #endif
156 
157 /*
158  * Each rule belongs to one of 32 different sets (0..31).
159  * The variable set_disable contains one bit per set.
160  * If the bit is set, all rules in the corresponding set
161  * are disabled. Set RESVD_SET(31) is reserved for the default rule
162  * and rules that are not deleted by the flush command,
163  * and CANNOT be disabled.
164  * Rules in set RESVD_SET can only be deleted individually.
165  */
166 VNET_DEFINE(u_int32_t, set_disable);
167 #define	V_set_disable			VNET(set_disable)
168 
169 VNET_DEFINE(int, fw_verbose);
170 /* counter for ipfw_log(NULL...) */
171 VNET_DEFINE(u_int64_t, norule_counter);
172 VNET_DEFINE(int, verbose_limit);
173 
174 /* layer3_chain contains the list of rules for layer 3 */
175 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
176 
177 /* ipfw_vnet_ready controls when we are open for business */
178 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
179 
180 VNET_DEFINE(int, ipfw_nat_ready) = 0;
181 
182 ipfw_nat_t *ipfw_nat_ptr = NULL;
183 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
184 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
185 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
186 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
187 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
188 
189 #ifdef SYSCTL_NODE
190 uint32_t dummy_def = IPFW_DEFAULT_RULE;
191 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
192 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
193 
194 SYSBEGIN(f3)
195 
196 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
197     "Firewall");
198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
199     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
200     "Only do a single pass through ipfw when using dummynet(4)");
201 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
202     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
203     "Rule number auto-increment step");
204 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
205     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
206     "Log matches to ipfw rules");
207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
208     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
209     "Set upper limit of matches of ipfw rules logged");
210 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
211     &dummy_def, 0,
212     "The default/max possible rule number.");
213 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
214     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE,
215     0, 0, sysctl_ipfw_table_num, "IU",
216     "Maximum number of concurrently used tables");
217 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
218     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE,
219     0, 0, sysctl_ipfw_tables_sets, "IU",
220     "Use per-set namespace for tables");
221 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
222     &default_to_accept, 0,
223     "Make the default rule accept all packets.");
224 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
225 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
226     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
227     "Number of static rules");
228 
229 #ifdef INET6
230 SYSCTL_DECL(_net_inet6_ip6);
231 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
232     "Firewall");
233 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
234     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
235     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
236     "Deny packets with unknown IPv6 Extension Headers");
237 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
238     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
239     &VNET_NAME(fw_permit_single_frag6), 0,
240     "Permit single packet IPv6 fragments");
241 #endif /* INET6 */
242 
243 SYSEND
244 
245 #endif /* SYSCTL_NODE */
246 
247 /*
248  * Some macros used in the various matching options.
249  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
250  * Other macros just cast void * into the appropriate type
251  */
252 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
253 #define	TCP(p)		((struct tcphdr *)(p))
254 #define	SCTP(p)		((struct sctphdr *)(p))
255 #define	UDP(p)		((struct udphdr *)(p))
256 #define	ICMP(p)		((struct icmphdr *)(p))
257 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
258 
259 static __inline int
260 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
261 {
262 	int type = icmp->icmp_type;
263 
264 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
265 }
266 
267 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
268     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
269 
270 static int
271 is_icmp_query(struct icmphdr *icmp)
272 {
273 	int type = icmp->icmp_type;
274 
275 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
276 }
277 #undef TT
278 
279 /*
280  * The following checks use two arrays of 8 or 16 bits to store the
281  * bits that we want set or clear, respectively. They are in the
282  * low and high half of cmd->arg1 or cmd->d[0].
283  *
284  * We scan options and store the bits we find set. We succeed if
285  *
286  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
287  *
288  * The code is sometimes optimized not to store additional variables.
289  */
290 
291 static int
292 flags_match(ipfw_insn *cmd, u_int8_t bits)
293 {
294 	u_char want_clear;
295 	bits = ~bits;
296 
297 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
298 		return 0; /* some bits we want set were clear */
299 	want_clear = (cmd->arg1 >> 8) & 0xff;
300 	if ( (want_clear & bits) != want_clear)
301 		return 0; /* some bits we want clear were set */
302 	return 1;
303 }
304 
305 static int
306 ipopts_match(struct ip *ip, ipfw_insn *cmd)
307 {
308 	int optlen, bits = 0;
309 	u_char *cp = (u_char *)(ip + 1);
310 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
311 
312 	for (; x > 0; x -= optlen, cp += optlen) {
313 		int opt = cp[IPOPT_OPTVAL];
314 
315 		if (opt == IPOPT_EOL)
316 			break;
317 		if (opt == IPOPT_NOP)
318 			optlen = 1;
319 		else {
320 			optlen = cp[IPOPT_OLEN];
321 			if (optlen <= 0 || optlen > x)
322 				return 0; /* invalid or truncated */
323 		}
324 		switch (opt) {
325 		default:
326 			break;
327 
328 		case IPOPT_LSRR:
329 			bits |= IP_FW_IPOPT_LSRR;
330 			break;
331 
332 		case IPOPT_SSRR:
333 			bits |= IP_FW_IPOPT_SSRR;
334 			break;
335 
336 		case IPOPT_RR:
337 			bits |= IP_FW_IPOPT_RR;
338 			break;
339 
340 		case IPOPT_TS:
341 			bits |= IP_FW_IPOPT_TS;
342 			break;
343 		}
344 	}
345 	return (flags_match(cmd, bits));
346 }
347 
348 /*
349  * Parse TCP options. The logic copied from tcp_dooptions().
350  */
351 static int
352 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss)
353 {
354 	const u_char *cp = (const u_char *)(tcp + 1);
355 	int optlen, bits = 0;
356 	int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr);
357 
358 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
359 		int opt = cp[0];
360 		if (opt == TCPOPT_EOL)
361 			break;
362 		if (opt == TCPOPT_NOP)
363 			optlen = 1;
364 		else {
365 			if (cnt < 2)
366 				break;
367 			optlen = cp[1];
368 			if (optlen < 2 || optlen > cnt)
369 				break;
370 		}
371 
372 		switch (opt) {
373 		default:
374 			break;
375 
376 		case TCPOPT_MAXSEG:
377 			if (optlen != TCPOLEN_MAXSEG)
378 				break;
379 			bits |= IP_FW_TCPOPT_MSS;
380 			if (mss != NULL)
381 				*mss = be16dec(cp + 2);
382 			break;
383 
384 		case TCPOPT_WINDOW:
385 			if (optlen == TCPOLEN_WINDOW)
386 				bits |= IP_FW_TCPOPT_WINDOW;
387 			break;
388 
389 		case TCPOPT_SACK_PERMITTED:
390 			if (optlen == TCPOLEN_SACK_PERMITTED)
391 				bits |= IP_FW_TCPOPT_SACK;
392 			break;
393 
394 		case TCPOPT_SACK:
395 			if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0)
396 				bits |= IP_FW_TCPOPT_SACK;
397 			break;
398 
399 		case TCPOPT_TIMESTAMP:
400 			if (optlen == TCPOLEN_TIMESTAMP)
401 				bits |= IP_FW_TCPOPT_TS;
402 			break;
403 		}
404 	}
405 	return (bits);
406 }
407 
408 static int
409 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
410 {
411 
412 	return (flags_match(cmd, tcpopts_parse(tcp, NULL)));
413 }
414 
415 static int
416 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
417     uint32_t *tablearg)
418 {
419 
420 	if (ifp == NULL)	/* no iface with this packet, match fails */
421 		return (0);
422 
423 	/* Check by name or by IP address */
424 	if (cmd->name[0] != '\0') { /* match by name */
425 		if (cmd->name[0] == '\1') /* use tablearg to match */
426 			return ipfw_lookup_table(chain, cmd->p.kidx, 0,
427 			    &ifp->if_index, tablearg);
428 		/* Check name */
429 		if (cmd->p.glob) {
430 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
431 				return(1);
432 		} else {
433 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
434 				return(1);
435 		}
436 	} else {
437 #if !defined(USERSPACE) && defined(__FreeBSD__)	/* and OSX too ? */
438 		struct ifaddr *ia;
439 
440 		NET_EPOCH_ASSERT();
441 
442 		CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
443 			if (ia->ifa_addr->sa_family != AF_INET)
444 				continue;
445 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
446 			    (ia->ifa_addr))->sin_addr.s_addr)
447 				return (1);	/* match */
448 		}
449 #endif /* __FreeBSD__ */
450 	}
451 	return(0);	/* no match, fail ... */
452 }
453 
454 /*
455  * The verify_path function checks if a route to the src exists and
456  * if it is reachable via ifp (when provided).
457  *
458  * The 'verrevpath' option checks that the interface that an IP packet
459  * arrives on is the same interface that traffic destined for the
460  * packet's source address would be routed out of.
461  * The 'versrcreach' option just checks that the source address is
462  * reachable via any route (except default) in the routing table.
463  * These two are a measure to block forged packets. This is also
464  * commonly known as "anti-spoofing" or Unicast Reverse Path
465  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
466  * is purposely reminiscent of the Cisco IOS command,
467  *
468  *   ip verify unicast reverse-path
469  *   ip verify unicast source reachable-via any
470  *
471  * which implements the same functionality. But note that the syntax
472  * is misleading, and the check may be performed on all IP packets
473  * whether unicast, multicast, or broadcast.
474  */
475 static int
476 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
477 {
478 #if defined(USERSPACE) || !defined(__FreeBSD__)
479 	return 0;
480 #else
481 	struct nhop_object *nh;
482 
483 	nh = fib4_lookup(fib, src, 0, NHR_NONE, 0);
484 	if (nh == NULL)
485 		return (0);
486 
487 	/*
488 	 * If ifp is provided, check for equality with rtentry.
489 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
490 	 * in order to pass packets injected back by if_simloop():
491 	 * routing entry (via lo0) for our own address
492 	 * may exist, so we need to handle routing assymetry.
493 	 */
494 	if (ifp != NULL && ifp != nh->nh_aifp)
495 		return (0);
496 
497 	/* if no ifp provided, check if rtentry is not default route */
498 	if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0)
499 		return (0);
500 
501 	/* or if this is a blackhole/reject route */
502 	if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
503 		return (0);
504 
505 	/* found valid route */
506 	return 1;
507 #endif /* __FreeBSD__ */
508 }
509 
510 /*
511  * Generate an SCTP packet containing an ABORT chunk. The verification tag
512  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
513  * reflected is not 0.
514  */
515 
516 static struct mbuf *
517 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
518     int reflected)
519 {
520 	struct mbuf *m;
521 	struct ip *ip;
522 #ifdef INET6
523 	struct ip6_hdr *ip6;
524 #endif
525 	struct sctphdr *sctp;
526 	struct sctp_chunkhdr *chunk;
527 	u_int16_t hlen, plen, tlen;
528 
529 	MGETHDR(m, M_NOWAIT, MT_DATA);
530 	if (m == NULL)
531 		return (NULL);
532 
533 	M_SETFIB(m, id->fib);
534 #ifdef MAC
535 	if (replyto != NULL)
536 		mac_netinet_firewall_reply(replyto, m);
537 	else
538 		mac_netinet_firewall_send(m);
539 #else
540 	(void)replyto;		/* don't warn about unused arg */
541 #endif
542 
543 	switch (id->addr_type) {
544 	case 4:
545 		hlen = sizeof(struct ip);
546 		break;
547 #ifdef INET6
548 	case 6:
549 		hlen = sizeof(struct ip6_hdr);
550 		break;
551 #endif
552 	default:
553 		/* XXX: log me?!? */
554 		FREE_PKT(m);
555 		return (NULL);
556 	}
557 	plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
558 	tlen = hlen + plen;
559 	m->m_data += max_linkhdr;
560 	m->m_flags |= M_SKIP_FIREWALL;
561 	m->m_pkthdr.len = m->m_len = tlen;
562 	m->m_pkthdr.rcvif = NULL;
563 	bzero(m->m_data, tlen);
564 
565 	switch (id->addr_type) {
566 	case 4:
567 		ip = mtod(m, struct ip *);
568 
569 		ip->ip_v = 4;
570 		ip->ip_hl = sizeof(struct ip) >> 2;
571 		ip->ip_tos = IPTOS_LOWDELAY;
572 		ip->ip_len = htons(tlen);
573 		ip->ip_id = htons(0);
574 		ip->ip_off = htons(0);
575 		ip->ip_ttl = V_ip_defttl;
576 		ip->ip_p = IPPROTO_SCTP;
577 		ip->ip_sum = 0;
578 		ip->ip_src.s_addr = htonl(id->dst_ip);
579 		ip->ip_dst.s_addr = htonl(id->src_ip);
580 
581 		sctp = (struct sctphdr *)(ip + 1);
582 		break;
583 #ifdef INET6
584 	case 6:
585 		ip6 = mtod(m, struct ip6_hdr *);
586 
587 		ip6->ip6_vfc = IPV6_VERSION;
588 		ip6->ip6_plen = htons(plen);
589 		ip6->ip6_nxt = IPPROTO_SCTP;
590 		ip6->ip6_hlim = IPV6_DEFHLIM;
591 		ip6->ip6_src = id->dst_ip6;
592 		ip6->ip6_dst = id->src_ip6;
593 
594 		sctp = (struct sctphdr *)(ip6 + 1);
595 		break;
596 #endif
597 	}
598 
599 	sctp->src_port = htons(id->dst_port);
600 	sctp->dest_port = htons(id->src_port);
601 	sctp->v_tag = htonl(vtag);
602 	sctp->checksum = htonl(0);
603 
604 	chunk = (struct sctp_chunkhdr *)(sctp + 1);
605 	chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
606 	chunk->chunk_flags = 0;
607 	if (reflected != 0) {
608 		chunk->chunk_flags |= SCTP_HAD_NO_TCB;
609 	}
610 	chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
611 
612 	sctp->checksum = sctp_calculate_cksum(m, hlen);
613 
614 	return (m);
615 }
616 
617 /*
618  * Generate a TCP packet, containing either a RST or a keepalive.
619  * When flags & TH_RST, we are sending a RST packet, because of a
620  * "reset" action matched the packet.
621  * Otherwise we are sending a keepalive, and flags & TH_
622  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
623  * so that MAC can label the reply appropriately.
624  */
625 struct mbuf *
626 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
627     u_int32_t ack, int flags)
628 {
629 	struct mbuf *m = NULL;		/* stupid compiler */
630 	struct ip *h = NULL;		/* stupid compiler */
631 #ifdef INET6
632 	struct ip6_hdr *h6 = NULL;
633 #endif
634 	struct tcphdr *th = NULL;
635 	int len, dir;
636 
637 	MGETHDR(m, M_NOWAIT, MT_DATA);
638 	if (m == NULL)
639 		return (NULL);
640 
641 	M_SETFIB(m, id->fib);
642 #ifdef MAC
643 	if (replyto != NULL)
644 		mac_netinet_firewall_reply(replyto, m);
645 	else
646 		mac_netinet_firewall_send(m);
647 #else
648 	(void)replyto;		/* don't warn about unused arg */
649 #endif
650 
651 	switch (id->addr_type) {
652 	case 4:
653 		len = sizeof(struct ip) + sizeof(struct tcphdr);
654 		break;
655 #ifdef INET6
656 	case 6:
657 		len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
658 		break;
659 #endif
660 	default:
661 		/* XXX: log me?!? */
662 		FREE_PKT(m);
663 		return (NULL);
664 	}
665 	dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
666 
667 	m->m_data += max_linkhdr;
668 	m->m_flags |= M_SKIP_FIREWALL;
669 	m->m_pkthdr.len = m->m_len = len;
670 	m->m_pkthdr.rcvif = NULL;
671 	bzero(m->m_data, len);
672 
673 	switch (id->addr_type) {
674 	case 4:
675 		h = mtod(m, struct ip *);
676 
677 		/* prepare for checksum */
678 		h->ip_p = IPPROTO_TCP;
679 		h->ip_len = htons(sizeof(struct tcphdr));
680 		if (dir) {
681 			h->ip_src.s_addr = htonl(id->src_ip);
682 			h->ip_dst.s_addr = htonl(id->dst_ip);
683 		} else {
684 			h->ip_src.s_addr = htonl(id->dst_ip);
685 			h->ip_dst.s_addr = htonl(id->src_ip);
686 		}
687 
688 		th = (struct tcphdr *)(h + 1);
689 		break;
690 #ifdef INET6
691 	case 6:
692 		h6 = mtod(m, struct ip6_hdr *);
693 
694 		/* prepare for checksum */
695 		h6->ip6_nxt = IPPROTO_TCP;
696 		h6->ip6_plen = htons(sizeof(struct tcphdr));
697 		if (dir) {
698 			h6->ip6_src = id->src_ip6;
699 			h6->ip6_dst = id->dst_ip6;
700 		} else {
701 			h6->ip6_src = id->dst_ip6;
702 			h6->ip6_dst = id->src_ip6;
703 		}
704 
705 		th = (struct tcphdr *)(h6 + 1);
706 		break;
707 #endif
708 	}
709 
710 	if (dir) {
711 		th->th_sport = htons(id->src_port);
712 		th->th_dport = htons(id->dst_port);
713 	} else {
714 		th->th_sport = htons(id->dst_port);
715 		th->th_dport = htons(id->src_port);
716 	}
717 	th->th_off = sizeof(struct tcphdr) >> 2;
718 
719 	if (flags & TH_RST) {
720 		if (flags & TH_ACK) {
721 			th->th_seq = htonl(ack);
722 			th->th_flags = TH_RST;
723 		} else {
724 			if (flags & TH_SYN)
725 				seq++;
726 			th->th_ack = htonl(seq);
727 			th->th_flags = TH_RST | TH_ACK;
728 		}
729 	} else {
730 		/*
731 		 * Keepalive - use caller provided sequence numbers
732 		 */
733 		th->th_seq = htonl(seq);
734 		th->th_ack = htonl(ack);
735 		th->th_flags = TH_ACK;
736 	}
737 
738 	switch (id->addr_type) {
739 	case 4:
740 		th->th_sum = in_cksum(m, len);
741 
742 		/* finish the ip header */
743 		h->ip_v = 4;
744 		h->ip_hl = sizeof(*h) >> 2;
745 		h->ip_tos = IPTOS_LOWDELAY;
746 		h->ip_off = htons(0);
747 		h->ip_len = htons(len);
748 		h->ip_ttl = V_ip_defttl;
749 		h->ip_sum = 0;
750 		break;
751 #ifdef INET6
752 	case 6:
753 		th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
754 		    sizeof(struct tcphdr));
755 
756 		/* finish the ip6 header */
757 		h6->ip6_vfc |= IPV6_VERSION;
758 		h6->ip6_hlim = IPV6_DEFHLIM;
759 		break;
760 #endif
761 	}
762 
763 	return (m);
764 }
765 
766 #ifdef INET6
767 /*
768  * ipv6 specific rules here...
769  */
770 static __inline int
771 icmp6type_match(int type, ipfw_insn_u32 *cmd)
772 {
773 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
774 }
775 
776 static int
777 flow6id_match(int curr_flow, ipfw_insn_u32 *cmd)
778 {
779 	int i;
780 	for (i=0; i <= cmd->o.arg1; ++i)
781 		if (curr_flow == cmd->d[i])
782 			return 1;
783 	return 0;
784 }
785 
786 /* support for IP6_*_ME opcodes */
787 static const struct in6_addr lla_mask = {{{
788 	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
789 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
790 }}};
791 
792 static int
793 ipfw_localip6(struct in6_addr *in6)
794 {
795 	struct rm_priotracker in6_ifa_tracker;
796 	struct in6_ifaddr *ia;
797 
798 	if (IN6_IS_ADDR_MULTICAST(in6))
799 		return (0);
800 
801 	if (!IN6_IS_ADDR_LINKLOCAL(in6))
802 		return (in6_localip(in6));
803 
804 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
805 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
806 		if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
807 			continue;
808 		if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
809 		    in6, &lla_mask)) {
810 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
811 			return (1);
812 		}
813 	}
814 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
815 	return (0);
816 }
817 
818 static int
819 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
820 {
821 	struct nhop_object *nh;
822 
823 	if (IN6_IS_SCOPE_LINKLOCAL(src))
824 		return (1);
825 
826 	nh = fib6_lookup(fib, src, 0, NHR_NONE, 0);
827 	if (nh == NULL)
828 		return (0);
829 
830 	/* If ifp is provided, check for equality with route table. */
831 	if (ifp != NULL && ifp != nh->nh_aifp)
832 		return (0);
833 
834 	/* if no ifp provided, check if rtentry is not default route */
835 	if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0)
836 		return (0);
837 
838 	/* or if this is a blackhole/reject route */
839 	if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
840 		return (0);
841 
842 	/* found valid route */
843 	return 1;
844 }
845 
846 static int
847 is_icmp6_query(int icmp6_type)
848 {
849 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
850 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
851 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
852 	    icmp6_type == ICMP6_WRUREQUEST ||
853 	    icmp6_type == ICMP6_FQDN_QUERY ||
854 	    icmp6_type == ICMP6_NI_QUERY))
855 		return (1);
856 
857 	return (0);
858 }
859 
860 static int
861 map_icmp_unreach(int code)
862 {
863 
864 	/* RFC 7915 p4.2 */
865 	switch (code) {
866 	case ICMP_UNREACH_NET:
867 	case ICMP_UNREACH_HOST:
868 	case ICMP_UNREACH_SRCFAIL:
869 	case ICMP_UNREACH_NET_UNKNOWN:
870 	case ICMP_UNREACH_HOST_UNKNOWN:
871 	case ICMP_UNREACH_TOSNET:
872 	case ICMP_UNREACH_TOSHOST:
873 		return (ICMP6_DST_UNREACH_NOROUTE);
874 	case ICMP_UNREACH_PORT:
875 		return (ICMP6_DST_UNREACH_NOPORT);
876 	default:
877 		/*
878 		 * Map the rest of codes into admit prohibited.
879 		 * XXX: unreach proto should be mapped into ICMPv6
880 		 * parameter problem, but we use only unreach type.
881 		 */
882 		return (ICMP6_DST_UNREACH_ADMIN);
883 	}
884 }
885 
886 static void
887 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
888 {
889 	struct mbuf *m;
890 
891 	m = args->m;
892 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
893 		struct tcphdr *tcp;
894 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
895 
896 		if ((tcp->th_flags & TH_RST) == 0) {
897 			struct mbuf *m0;
898 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
899 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
900 			    tcp->th_flags | TH_RST);
901 			if (m0 != NULL)
902 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
903 				    NULL);
904 		}
905 		FREE_PKT(m);
906 	} else if (code == ICMP6_UNREACH_ABORT &&
907 	    args->f_id.proto == IPPROTO_SCTP) {
908 		struct mbuf *m0;
909 		struct sctphdr *sctp;
910 		u_int32_t v_tag;
911 		int reflected;
912 
913 		sctp = (struct sctphdr *)((char *)ip6 + hlen);
914 		reflected = 1;
915 		v_tag = ntohl(sctp->v_tag);
916 		/* Investigate the first chunk header if available */
917 		if (m->m_len >= hlen + sizeof(struct sctphdr) +
918 		    sizeof(struct sctp_chunkhdr)) {
919 			struct sctp_chunkhdr *chunk;
920 
921 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
922 			switch (chunk->chunk_type) {
923 			case SCTP_INITIATION:
924 				/*
925 				 * Packets containing an INIT chunk MUST have
926 				 * a zero v-tag.
927 				 */
928 				if (v_tag != 0) {
929 					v_tag = 0;
930 					break;
931 				}
932 				/* INIT chunk MUST NOT be bundled */
933 				if (m->m_pkthdr.len >
934 				    hlen + sizeof(struct sctphdr) +
935 				    ntohs(chunk->chunk_length) + 3) {
936 					break;
937 				}
938 				/* Use the initiate tag if available */
939 				if ((m->m_len >= hlen + sizeof(struct sctphdr) +
940 				    sizeof(struct sctp_chunkhdr) +
941 				    offsetof(struct sctp_init, a_rwnd))) {
942 					struct sctp_init *init;
943 
944 					init = (struct sctp_init *)(chunk + 1);
945 					v_tag = ntohl(init->initiate_tag);
946 					reflected = 0;
947 				}
948 				break;
949 			case SCTP_ABORT_ASSOCIATION:
950 				/*
951 				 * If the packet contains an ABORT chunk, don't
952 				 * reply.
953 				 * XXX: We should search through all chunks,
954 				 * but do not do that to avoid attacks.
955 				 */
956 				v_tag = 0;
957 				break;
958 			}
959 		}
960 		if (v_tag == 0) {
961 			m0 = NULL;
962 		} else {
963 			m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
964 			    reflected);
965 		}
966 		if (m0 != NULL)
967 			ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
968 		FREE_PKT(m);
969 	} else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
970 		/* Send an ICMPv6 unreach. */
971 #if 0
972 		/*
973 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
974 		 * as the contents are read. We need to m_adj() the
975 		 * needed amount.
976 		 * The mbuf will however be thrown away so we can adjust it.
977 		 * Remember we did an m_pullup on it already so we
978 		 * can make some assumptions about contiguousness.
979 		 */
980 		if (args->L3offset)
981 			m_adj(m, args->L3offset);
982 #endif
983 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
984 	} else
985 		FREE_PKT(m);
986 
987 	args->m = NULL;
988 }
989 
990 #endif /* INET6 */
991 
992 /*
993  * sends a reject message, consuming the mbuf passed as an argument.
994  */
995 static void
996 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
997 {
998 
999 #if 0
1000 	/* XXX When ip is not guaranteed to be at mtod() we will
1001 	 * need to account for this */
1002 	 * The mbuf will however be thrown away so we can adjust it.
1003 	 * Remember we did an m_pullup on it already so we
1004 	 * can make some assumptions about contiguousness.
1005 	 */
1006 	if (args->L3offset)
1007 		m_adj(m, args->L3offset);
1008 #endif
1009 	if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
1010 		/* Send an ICMP unreach */
1011 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1012 	} else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
1013 		struct tcphdr *const tcp =
1014 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1015 		if ( (tcp->th_flags & TH_RST) == 0) {
1016 			struct mbuf *m;
1017 			m = ipfw_send_pkt(args->m, &(args->f_id),
1018 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1019 				tcp->th_flags | TH_RST);
1020 			if (m != NULL)
1021 				ip_output(m, NULL, NULL, 0, NULL, NULL);
1022 		}
1023 		FREE_PKT(args->m);
1024 	} else if (code == ICMP_REJECT_ABORT &&
1025 	    args->f_id.proto == IPPROTO_SCTP) {
1026 		struct mbuf *m;
1027 		struct sctphdr *sctp;
1028 		struct sctp_chunkhdr *chunk;
1029 		struct sctp_init *init;
1030 		u_int32_t v_tag;
1031 		int reflected;
1032 
1033 		sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1034 		reflected = 1;
1035 		v_tag = ntohl(sctp->v_tag);
1036 		if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1037 		    sizeof(struct sctp_chunkhdr)) {
1038 			/* Look at the first chunk header if available */
1039 			chunk = (struct sctp_chunkhdr *)(sctp + 1);
1040 			switch (chunk->chunk_type) {
1041 			case SCTP_INITIATION:
1042 				/*
1043 				 * Packets containing an INIT chunk MUST have
1044 				 * a zero v-tag.
1045 				 */
1046 				if (v_tag != 0) {
1047 					v_tag = 0;
1048 					break;
1049 				}
1050 				/* INIT chunk MUST NOT be bundled */
1051 				if (iplen >
1052 				    (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1053 				    ntohs(chunk->chunk_length) + 3) {
1054 					break;
1055 				}
1056 				/* Use the initiate tag if available */
1057 				if ((iplen >= (ip->ip_hl << 2) +
1058 				    sizeof(struct sctphdr) +
1059 				    sizeof(struct sctp_chunkhdr) +
1060 				    offsetof(struct sctp_init, a_rwnd))) {
1061 					init = (struct sctp_init *)(chunk + 1);
1062 					v_tag = ntohl(init->initiate_tag);
1063 					reflected = 0;
1064 				}
1065 				break;
1066 			case SCTP_ABORT_ASSOCIATION:
1067 				/*
1068 				 * If the packet contains an ABORT chunk, don't
1069 				 * reply.
1070 				 * XXX: We should search through all chunks,
1071 				 * but do not do that to avoid attacks.
1072 				 */
1073 				v_tag = 0;
1074 				break;
1075 			}
1076 		}
1077 		if (v_tag == 0) {
1078 			m = NULL;
1079 		} else {
1080 			m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1081 			    reflected);
1082 		}
1083 		if (m != NULL)
1084 			ip_output(m, NULL, NULL, 0, NULL, NULL);
1085 		FREE_PKT(args->m);
1086 	} else
1087 		FREE_PKT(args->m);
1088 	args->m = NULL;
1089 }
1090 
1091 /*
1092  * Support for uid/gid/jail lookup. These tests are expensive
1093  * (because we may need to look into the list of active sockets)
1094  * so we cache the results. ugid_lookupp is 0 if we have not
1095  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1096  * and failed. The function always returns the match value.
1097  * We could actually spare the variable and use *uc, setting
1098  * it to '(void *)check_uidgid if we have no info, NULL if
1099  * we tried and failed, or any other value if successful.
1100  */
1101 static int
1102 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1103     struct ucred **uc)
1104 {
1105 #if defined(USERSPACE)
1106 	return 0;	// not supported in userspace
1107 #else
1108 #ifndef __FreeBSD__
1109 	/* XXX */
1110 	return cred_check(insn, proto, oif,
1111 	    dst_ip, dst_port, src_ip, src_port,
1112 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1113 #else  /* FreeBSD */
1114 	struct in_addr src_ip, dst_ip;
1115 	struct inpcbinfo *pi;
1116 	struct ipfw_flow_id *id;
1117 	struct inpcb *pcb, *inp;
1118 	int lookupflags;
1119 	int match;
1120 
1121 	id = &args->f_id;
1122 	inp = args->inp;
1123 
1124 	/*
1125 	 * Check to see if the UDP or TCP stack supplied us with
1126 	 * the PCB. If so, rather then holding a lock and looking
1127 	 * up the PCB, we can use the one that was supplied.
1128 	 */
1129 	if (inp && *ugid_lookupp == 0) {
1130 		INP_LOCK_ASSERT(inp);
1131 		if (inp->inp_socket != NULL) {
1132 			*uc = crhold(inp->inp_cred);
1133 			*ugid_lookupp = 1;
1134 		} else
1135 			*ugid_lookupp = -1;
1136 	}
1137 	/*
1138 	 * If we have already been here and the packet has no
1139 	 * PCB entry associated with it, then we can safely
1140 	 * assume that this is a no match.
1141 	 */
1142 	if (*ugid_lookupp == -1)
1143 		return (0);
1144 	if (id->proto == IPPROTO_TCP) {
1145 		lookupflags = 0;
1146 		pi = &V_tcbinfo;
1147 	} else if (id->proto == IPPROTO_UDP) {
1148 		lookupflags = INPLOOKUP_WILDCARD;
1149 		pi = &V_udbinfo;
1150 	} else if (id->proto == IPPROTO_UDPLITE) {
1151 		lookupflags = INPLOOKUP_WILDCARD;
1152 		pi = &V_ulitecbinfo;
1153 	} else
1154 		return 0;
1155 	lookupflags |= INPLOOKUP_RLOCKPCB;
1156 	match = 0;
1157 	if (*ugid_lookupp == 0) {
1158 		if (id->addr_type == 6) {
1159 #ifdef INET6
1160 			if (args->flags & IPFW_ARGS_IN)
1161 				pcb = in6_pcblookup_mbuf(pi,
1162 				    &id->src_ip6, htons(id->src_port),
1163 				    &id->dst_ip6, htons(id->dst_port),
1164 				    lookupflags, NULL, args->m);
1165 			else
1166 				pcb = in6_pcblookup_mbuf(pi,
1167 				    &id->dst_ip6, htons(id->dst_port),
1168 				    &id->src_ip6, htons(id->src_port),
1169 				    lookupflags, args->ifp, args->m);
1170 #else
1171 			*ugid_lookupp = -1;
1172 			return (0);
1173 #endif
1174 		} else {
1175 			src_ip.s_addr = htonl(id->src_ip);
1176 			dst_ip.s_addr = htonl(id->dst_ip);
1177 			if (args->flags & IPFW_ARGS_IN)
1178 				pcb = in_pcblookup_mbuf(pi,
1179 				    src_ip, htons(id->src_port),
1180 				    dst_ip, htons(id->dst_port),
1181 				    lookupflags, NULL, args->m);
1182 			else
1183 				pcb = in_pcblookup_mbuf(pi,
1184 				    dst_ip, htons(id->dst_port),
1185 				    src_ip, htons(id->src_port),
1186 				    lookupflags, args->ifp, args->m);
1187 		}
1188 		if (pcb != NULL) {
1189 			INP_RLOCK_ASSERT(pcb);
1190 			*uc = crhold(pcb->inp_cred);
1191 			*ugid_lookupp = 1;
1192 			INP_RUNLOCK(pcb);
1193 		}
1194 		if (*ugid_lookupp == 0) {
1195 			/*
1196 			 * We tried and failed, set the variable to -1
1197 			 * so we will not try again on this packet.
1198 			 */
1199 			*ugid_lookupp = -1;
1200 			return (0);
1201 		}
1202 	}
1203 	if (insn->o.opcode == O_UID)
1204 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1205 	else if (insn->o.opcode == O_GID)
1206 		match = groupmember((gid_t)insn->d[0], *uc);
1207 	else if (insn->o.opcode == O_JAIL)
1208 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1209 	return (match);
1210 #endif /* __FreeBSD__ */
1211 #endif /* not supported in userspace */
1212 }
1213 
1214 /*
1215  * Helper function to set args with info on the rule after the matching
1216  * one. slot is precise, whereas we guess rule_id as they are
1217  * assigned sequentially.
1218  */
1219 static inline void
1220 set_match(struct ip_fw_args *args, int slot,
1221 	struct ip_fw_chain *chain)
1222 {
1223 	args->rule.chain_id = chain->id;
1224 	args->rule.slot = slot + 1; /* we use 0 as a marker */
1225 	args->rule.rule_id = 1 + chain->map[slot]->id;
1226 	args->rule.rulenum = chain->map[slot]->rulenum;
1227 	args->flags |= IPFW_ARGS_REF;
1228 }
1229 
1230 #ifndef LINEAR_SKIPTO
1231 /*
1232  * Helper function to enable cached rule lookups using
1233  * cached_id and cached_pos fields in ipfw rule.
1234  */
1235 static int
1236 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1237     int tablearg, int jump_backwards)
1238 {
1239 	int f_pos;
1240 
1241 	/* If possible use cached f_pos (in f->cached_pos),
1242 	 * whose version is written in f->cached_id
1243 	 * (horrible hacks to avoid changing the ABI).
1244 	 */
1245 	if (num != IP_FW_TARG && f->cached_id == chain->id)
1246 		f_pos = f->cached_pos;
1247 	else {
1248 		int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1249 		/* make sure we do not jump backward */
1250 		if (jump_backwards == 0 && i <= f->rulenum)
1251 			i = f->rulenum + 1;
1252 		if (chain->idxmap != NULL)
1253 			f_pos = chain->idxmap[i];
1254 		else
1255 			f_pos = ipfw_find_rule(chain, i, 0);
1256 		/* update the cache */
1257 		if (num != IP_FW_TARG) {
1258 			f->cached_id = chain->id;
1259 			f->cached_pos = f_pos;
1260 		}
1261 	}
1262 
1263 	return (f_pos);
1264 }
1265 #else
1266 /*
1267  * Helper function to enable real fast rule lookups.
1268  */
1269 static int
1270 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1271     int tablearg, int jump_backwards)
1272 {
1273 	int f_pos;
1274 
1275 	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1276 	/* make sure we do not jump backward */
1277 	if (jump_backwards == 0 && num <= f->rulenum)
1278 		num = f->rulenum + 1;
1279 	f_pos = chain->idxmap[num];
1280 
1281 	return (f_pos);
1282 }
1283 #endif
1284 
1285 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
1286 /*
1287  * The main check routine for the firewall.
1288  *
1289  * All arguments are in args so we can modify them and return them
1290  * back to the caller.
1291  *
1292  * Parameters:
1293  *
1294  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
1295  *		Starts with the IP header.
1296  *	args->L3offset	Number of bytes bypassed if we came from L2.
1297  *			e.g. often sizeof(eh)  ** NOTYET **
1298  *	args->ifp	Incoming or outgoing interface.
1299  *	args->divert_rule (in/out)
1300  *		Skip up to the first rule past this rule number;
1301  *		upon return, non-zero port number for divert or tee.
1302  *
1303  *	args->rule	Pointer to the last matching rule (in/out)
1304  *	args->next_hop	Socket we are forwarding to (out).
1305  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
1306  *	args->f_id	Addresses grabbed from the packet (out)
1307  * 	args->rule.info	a cookie depending on rule action
1308  *
1309  * Return value:
1310  *
1311  *	IP_FW_PASS	the packet must be accepted
1312  *	IP_FW_DENY	the packet must be dropped
1313  *	IP_FW_DIVERT	divert packet, port in m_tag
1314  *	IP_FW_TEE	tee packet, port in m_tag
1315  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
1316  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
1317  *		args->rule contains the matching rule,
1318  *		args->rule.info has additional information.
1319  *
1320  */
1321 int
1322 ipfw_chk(struct ip_fw_args *args)
1323 {
1324 
1325 	/*
1326 	 * Local variables holding state while processing a packet:
1327 	 *
1328 	 * IMPORTANT NOTE: to speed up the processing of rules, there
1329 	 * are some assumption on the values of the variables, which
1330 	 * are documented here. Should you change them, please check
1331 	 * the implementation of the various instructions to make sure
1332 	 * that they still work.
1333 	 *
1334 	 * m | args->m	Pointer to the mbuf, as received from the caller.
1335 	 *	It may change if ipfw_chk() does an m_pullup, or if it
1336 	 *	consumes the packet because it calls send_reject().
1337 	 *	XXX This has to change, so that ipfw_chk() never modifies
1338 	 *	or consumes the buffer.
1339 	 *	OR
1340 	 * args->mem	Pointer to contigous memory chunk.
1341 	 * ip	Is the beginning of the ip(4 or 6) header.
1342 	 * eh	Ethernet header in case if input is Layer2.
1343 	 */
1344 	struct mbuf *m;
1345 	struct ip *ip;
1346 	struct ether_header *eh;
1347 
1348 	/*
1349 	 * For rules which contain uid/gid or jail constraints, cache
1350 	 * a copy of the users credentials after the pcb lookup has been
1351 	 * executed. This will speed up the processing of rules with
1352 	 * these types of constraints, as well as decrease contention
1353 	 * on pcb related locks.
1354 	 */
1355 #ifndef __FreeBSD__
1356 	struct bsd_ucred ucred_cache;
1357 #else
1358 	struct ucred *ucred_cache = NULL;
1359 #endif
1360 	int ucred_lookup = 0;
1361 	int f_pos = 0;		/* index of current rule in the array */
1362 	int retval = 0;
1363 	struct ifnet *oif, *iif;
1364 
1365 	/*
1366 	 * hlen	The length of the IP header.
1367 	 */
1368 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
1369 
1370 	/*
1371 	 * offset	The offset of a fragment. offset != 0 means that
1372 	 *	we have a fragment at this offset of an IPv4 packet.
1373 	 *	offset == 0 means that (if this is an IPv4 packet)
1374 	 *	this is the first or only fragment.
1375 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1376 	 *	or there is a single packet fragment (fragment header added
1377 	 *	without needed).  We will treat a single packet fragment as if
1378 	 *	there was no fragment header (or log/block depending on the
1379 	 *	V_fw_permit_single_frag6 sysctl setting).
1380 	 */
1381 	u_short offset = 0;
1382 	u_short ip6f_mf = 0;
1383 
1384 	/*
1385 	 * Local copies of addresses. They are only valid if we have
1386 	 * an IP packet.
1387 	 *
1388 	 * proto	The protocol. Set to 0 for non-ip packets,
1389 	 *	or to the protocol read from the packet otherwise.
1390 	 *	proto != 0 means that we have an IPv4 packet.
1391 	 *
1392 	 * src_port, dst_port	port numbers, in HOST format. Only
1393 	 *	valid for TCP and UDP packets.
1394 	 *
1395 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
1396 	 *	Only valid for IPv4 packets.
1397 	 */
1398 	uint8_t proto;
1399 	uint16_t src_port, dst_port;		/* NOTE: host format	*/
1400 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
1401 	int iplen = 0;
1402 	int pktlen;
1403 
1404 	struct ipfw_dyn_info dyn_info;
1405 	struct ip_fw *q = NULL;
1406 	struct ip_fw_chain *chain = &V_layer3_chain;
1407 
1408 	/*
1409 	 * We store in ulp a pointer to the upper layer protocol header.
1410 	 * In the ipv4 case this is easy to determine from the header,
1411 	 * but for ipv6 we might have some additional headers in the middle.
1412 	 * ulp is NULL if not found.
1413 	 */
1414 	void *ulp = NULL;		/* upper layer protocol pointer. */
1415 
1416 	/* XXX ipv6 variables */
1417 	int is_ipv6 = 0;
1418 	uint8_t	icmp6_type = 0;
1419 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1420 	/* end of ipv6 variables */
1421 
1422 	int is_ipv4 = 0;
1423 
1424 	int done = 0;		/* flag to exit the outer loop */
1425 	IPFW_RLOCK_TRACKER;
1426 	bool mem;
1427 
1428 	if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1429 		if (args->flags & IPFW_ARGS_ETHER) {
1430 			eh = (struct ether_header *)args->mem;
1431 			if (eh->ether_type == htons(ETHERTYPE_VLAN))
1432 				ip = (struct ip *)
1433 				    ((struct ether_vlan_header *)eh + 1);
1434 			else
1435 				ip = (struct ip *)(eh + 1);
1436 		} else {
1437 			eh = NULL;
1438 			ip = (struct ip *)args->mem;
1439 		}
1440 		pktlen = IPFW_ARGS_LENGTH(args->flags);
1441 		args->f_id.fib = args->ifp->if_fib;	/* best guess */
1442 	} else {
1443 		m = args->m;
1444 		if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1445 			return (IP_FW_PASS);	/* accept */
1446 		if (args->flags & IPFW_ARGS_ETHER) {
1447 	                /* We need some amount of data to be contiguous. */
1448 			if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1449 			    (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1450 			    max_protohdr))) == NULL)
1451 				goto pullup_failed;
1452 			eh = mtod(m, struct ether_header *);
1453 			ip = (struct ip *)(eh + 1);
1454 		} else {
1455 			eh = NULL;
1456 			ip = mtod(m, struct ip *);
1457 		}
1458 		pktlen = m->m_pkthdr.len;
1459 		args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1460 	}
1461 
1462 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1463 	src_ip.s_addr = 0;		/* make sure it is initialized */
1464 	src_port = dst_port = 0;
1465 
1466 	DYN_INFO_INIT(&dyn_info);
1467 /*
1468  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1469  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1470  * pointer might become stale after other pullups (but we never use it
1471  * this way).
1472  */
1473 #define	PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1474 #define	EHLEN	(eh != NULL ? ((char *)ip - (char *)eh) : 0)
1475 #define	_PULLUP_LOCKED(_len, p, T, unlock)			\
1476 do {								\
1477 	int x = (_len) + T + EHLEN;				\
1478 	if (mem) {						\
1479 		if (__predict_false(pktlen < x)) {		\
1480 			unlock;					\
1481 			goto pullup_failed;			\
1482 		}						\
1483 		p = (char *)args->mem + (_len) + EHLEN;		\
1484 	} else {						\
1485 		if (__predict_false((m)->m_len < x)) {		\
1486 			args->m = m = m_pullup(m, x);		\
1487 			if (m == NULL) {			\
1488 				unlock;				\
1489 				goto pullup_failed;		\
1490 			}					\
1491 		}						\
1492 		p = mtod(m, char *) + (_len) + EHLEN;		\
1493 	}							\
1494 } while (0)
1495 
1496 #define	PULLUP_LEN(_len, p, T)	_PULLUP_LOCKED(_len, p, T, )
1497 #define	PULLUP_LEN_LOCKED(_len, p, T)	\
1498     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain));	\
1499     UPDATE_POINTERS()
1500 /*
1501  * In case pointers got stale after pullups, update them.
1502  */
1503 #define	UPDATE_POINTERS()					\
1504 do {								\
1505 	if (!mem) {						\
1506 		if (eh != NULL) {				\
1507 			eh = mtod(m, struct ether_header *);	\
1508 			ip = (struct ip *)(eh + 1);		\
1509 		} else						\
1510 			ip = mtod(m, struct ip *);		\
1511 		args->m = m;					\
1512 	}							\
1513 } while (0)
1514 
1515 	/* Identify IP packets and fill up variables. */
1516 	if (pktlen >= sizeof(struct ip6_hdr) &&
1517 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1518 	    ip->ip_v == 6) {
1519 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1520 
1521 		is_ipv6 = 1;
1522 		args->flags |= IPFW_ARGS_IP6;
1523 		hlen = sizeof(struct ip6_hdr);
1524 		proto = ip6->ip6_nxt;
1525 		/* Search extension headers to find upper layer protocols */
1526 		while (ulp == NULL && offset == 0) {
1527 			switch (proto) {
1528 			case IPPROTO_ICMPV6:
1529 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1530 				icmp6_type = ICMP6(ulp)->icmp6_type;
1531 				break;
1532 
1533 			case IPPROTO_TCP:
1534 				PULLUP_TO(hlen, ulp, struct tcphdr);
1535 				dst_port = TCP(ulp)->th_dport;
1536 				src_port = TCP(ulp)->th_sport;
1537 				/* save flags for dynamic rules */
1538 				args->f_id._flags = TCP(ulp)->th_flags;
1539 				break;
1540 
1541 			case IPPROTO_SCTP:
1542 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1543 				    sizeof(struct sctp_chunkhdr) +
1544 				    offsetof(struct sctp_init, a_rwnd))
1545 					PULLUP_LEN(hlen, ulp,
1546 					    sizeof(struct sctphdr) +
1547 					    sizeof(struct sctp_chunkhdr) +
1548 					    offsetof(struct sctp_init, a_rwnd));
1549 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1550 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1551 				else
1552 					PULLUP_LEN(hlen, ulp,
1553 					    sizeof(struct sctphdr));
1554 				src_port = SCTP(ulp)->src_port;
1555 				dst_port = SCTP(ulp)->dest_port;
1556 				break;
1557 
1558 			case IPPROTO_UDP:
1559 			case IPPROTO_UDPLITE:
1560 				PULLUP_TO(hlen, ulp, struct udphdr);
1561 				dst_port = UDP(ulp)->uh_dport;
1562 				src_port = UDP(ulp)->uh_sport;
1563 				break;
1564 
1565 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1566 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1567 				ext_hd |= EXT_HOPOPTS;
1568 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1569 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1570 				ulp = NULL;
1571 				break;
1572 
1573 			case IPPROTO_ROUTING:	/* RFC 2460 */
1574 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1575 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1576 				case 0:
1577 					ext_hd |= EXT_RTHDR0;
1578 					break;
1579 				case 2:
1580 					ext_hd |= EXT_RTHDR2;
1581 					break;
1582 				default:
1583 					if (V_fw_verbose)
1584 						printf("IPFW2: IPV6 - Unknown "
1585 						    "Routing Header type(%d)\n",
1586 						    ((struct ip6_rthdr *)
1587 						    ulp)->ip6r_type);
1588 					if (V_fw_deny_unknown_exthdrs)
1589 					    return (IP_FW_DENY);
1590 					break;
1591 				}
1592 				ext_hd |= EXT_ROUTING;
1593 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1594 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1595 				ulp = NULL;
1596 				break;
1597 
1598 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1599 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1600 				ext_hd |= EXT_FRAGMENT;
1601 				hlen += sizeof (struct ip6_frag);
1602 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1603 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1604 					IP6F_OFF_MASK;
1605 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1606 					IP6F_MORE_FRAG;
1607 				if (V_fw_permit_single_frag6 == 0 &&
1608 				    offset == 0 && ip6f_mf == 0) {
1609 					if (V_fw_verbose)
1610 						printf("IPFW2: IPV6 - Invalid "
1611 						    "Fragment Header\n");
1612 					if (V_fw_deny_unknown_exthdrs)
1613 					    return (IP_FW_DENY);
1614 					break;
1615 				}
1616 				args->f_id.extra =
1617 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1618 				ulp = NULL;
1619 				break;
1620 
1621 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1622 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1623 				ext_hd |= EXT_DSTOPTS;
1624 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1625 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1626 				ulp = NULL;
1627 				break;
1628 
1629 			case IPPROTO_AH:	/* RFC 2402 */
1630 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1631 				ext_hd |= EXT_AH;
1632 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1633 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1634 				ulp = NULL;
1635 				break;
1636 
1637 			case IPPROTO_ESP:	/* RFC 2406 */
1638 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1639 				/* Anything past Seq# is variable length and
1640 				 * data past this ext. header is encrypted. */
1641 				ext_hd |= EXT_ESP;
1642 				break;
1643 
1644 			case IPPROTO_NONE:	/* RFC 2460 */
1645 				/*
1646 				 * Packet ends here, and IPv6 header has
1647 				 * already been pulled up. If ip6e_len!=0
1648 				 * then octets must be ignored.
1649 				 */
1650 				ulp = ip; /* non-NULL to get out of loop. */
1651 				break;
1652 
1653 			case IPPROTO_OSPFIGP:
1654 				/* XXX OSPF header check? */
1655 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1656 				break;
1657 
1658 			case IPPROTO_PIM:
1659 				/* XXX PIM header check? */
1660 				PULLUP_TO(hlen, ulp, struct pim);
1661 				break;
1662 
1663 			case IPPROTO_GRE:	/* RFC 1701 */
1664 				/* XXX GRE header check? */
1665 				PULLUP_TO(hlen, ulp, struct grehdr);
1666 				break;
1667 
1668 			case IPPROTO_CARP:
1669 				PULLUP_TO(hlen, ulp, offsetof(
1670 				    struct carp_header, carp_counter));
1671 				if (CARP_ADVERTISEMENT !=
1672 				    ((struct carp_header *)ulp)->carp_type)
1673 					return (IP_FW_DENY);
1674 				break;
1675 
1676 			case IPPROTO_IPV6:	/* RFC 2893 */
1677 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1678 				break;
1679 
1680 			case IPPROTO_IPV4:	/* RFC 2893 */
1681 				PULLUP_TO(hlen, ulp, struct ip);
1682 				break;
1683 
1684 			default:
1685 				if (V_fw_verbose)
1686 					printf("IPFW2: IPV6 - Unknown "
1687 					    "Extension Header(%d), ext_hd=%x\n",
1688 					     proto, ext_hd);
1689 				if (V_fw_deny_unknown_exthdrs)
1690 				    return (IP_FW_DENY);
1691 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1692 				break;
1693 			} /*switch */
1694 		}
1695 		UPDATE_POINTERS();
1696 		ip6 = (struct ip6_hdr *)ip;
1697 		args->f_id.addr_type = 6;
1698 		args->f_id.src_ip6 = ip6->ip6_src;
1699 		args->f_id.dst_ip6 = ip6->ip6_dst;
1700 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1701 		iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1702 	} else if (pktlen >= sizeof(struct ip) &&
1703 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1704 	    ip->ip_v == 4) {
1705 		is_ipv4 = 1;
1706 		args->flags |= IPFW_ARGS_IP4;
1707 		hlen = ip->ip_hl << 2;
1708 		/*
1709 		 * Collect parameters into local variables for faster
1710 		 * matching.
1711 		 */
1712 		proto = ip->ip_p;
1713 		src_ip = ip->ip_src;
1714 		dst_ip = ip->ip_dst;
1715 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1716 		iplen = ntohs(ip->ip_len);
1717 
1718 		if (offset == 0) {
1719 			switch (proto) {
1720 			case IPPROTO_TCP:
1721 				PULLUP_TO(hlen, ulp, struct tcphdr);
1722 				dst_port = TCP(ulp)->th_dport;
1723 				src_port = TCP(ulp)->th_sport;
1724 				/* save flags for dynamic rules */
1725 				args->f_id._flags = TCP(ulp)->th_flags;
1726 				break;
1727 
1728 			case IPPROTO_SCTP:
1729 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1730 				    sizeof(struct sctp_chunkhdr) +
1731 				    offsetof(struct sctp_init, a_rwnd))
1732 					PULLUP_LEN(hlen, ulp,
1733 					    sizeof(struct sctphdr) +
1734 					    sizeof(struct sctp_chunkhdr) +
1735 					    offsetof(struct sctp_init, a_rwnd));
1736 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1737 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1738 				else
1739 					PULLUP_LEN(hlen, ulp,
1740 					    sizeof(struct sctphdr));
1741 				src_port = SCTP(ulp)->src_port;
1742 				dst_port = SCTP(ulp)->dest_port;
1743 				break;
1744 
1745 			case IPPROTO_UDP:
1746 			case IPPROTO_UDPLITE:
1747 				PULLUP_TO(hlen, ulp, struct udphdr);
1748 				dst_port = UDP(ulp)->uh_dport;
1749 				src_port = UDP(ulp)->uh_sport;
1750 				break;
1751 
1752 			case IPPROTO_ICMP:
1753 				PULLUP_TO(hlen, ulp, struct icmphdr);
1754 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1755 				break;
1756 
1757 			default:
1758 				break;
1759 			}
1760 		} else {
1761 			if (offset == 1 && proto == IPPROTO_TCP) {
1762 				/* RFC 3128 */
1763 				goto pullup_failed;
1764 			}
1765 		}
1766 
1767 		UPDATE_POINTERS();
1768 		args->f_id.addr_type = 4;
1769 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1770 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1771 	} else {
1772 		proto = 0;
1773 		dst_ip.s_addr = src_ip.s_addr = 0;
1774 
1775 		args->f_id.addr_type = 1; /* XXX */
1776 	}
1777 #undef PULLUP_TO
1778 	pktlen = iplen < pktlen ? iplen: pktlen;
1779 
1780 	/* Properly initialize the rest of f_id */
1781 	args->f_id.proto = proto;
1782 	args->f_id.src_port = src_port = ntohs(src_port);
1783 	args->f_id.dst_port = dst_port = ntohs(dst_port);
1784 
1785 	IPFW_PF_RLOCK(chain);
1786 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1787 		IPFW_PF_RUNLOCK(chain);
1788 		return (IP_FW_PASS);	/* accept */
1789 	}
1790 	if (args->flags & IPFW_ARGS_REF) {
1791 		/*
1792 		 * Packet has already been tagged as a result of a previous
1793 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1794 		 * REASS, NETGRAPH, DIVERT/TEE...)
1795 		 * Validate the slot and continue from the next one
1796 		 * if still present, otherwise do a lookup.
1797 		 */
1798 		f_pos = (args->rule.chain_id == chain->id) ?
1799 		    args->rule.slot :
1800 		    ipfw_find_rule(chain, args->rule.rulenum,
1801 			args->rule.rule_id);
1802 	} else {
1803 		f_pos = 0;
1804 	}
1805 
1806 	if (args->flags & IPFW_ARGS_IN) {
1807 		iif = args->ifp;
1808 		oif = NULL;
1809 	} else {
1810 		MPASS(args->flags & IPFW_ARGS_OUT);
1811 		iif = mem ? NULL : m_rcvif(m);
1812 		oif = args->ifp;
1813 	}
1814 
1815 	/*
1816 	 * Now scan the rules, and parse microinstructions for each rule.
1817 	 * We have two nested loops and an inner switch. Sometimes we
1818 	 * need to break out of one or both loops, or re-enter one of
1819 	 * the loops with updated variables. Loop variables are:
1820 	 *
1821 	 *	f_pos (outer loop) points to the current rule.
1822 	 *		On output it points to the matching rule.
1823 	 *	done (outer loop) is used as a flag to break the loop.
1824 	 *	l (inner loop)	residual length of current rule.
1825 	 *		cmd points to the current microinstruction.
1826 	 *
1827 	 * We break the inner loop by setting l=0 and possibly
1828 	 * cmdlen=0 if we don't want to advance cmd.
1829 	 * We break the outer loop by setting done=1
1830 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1831 	 * as needed.
1832 	 */
1833 	for (; f_pos < chain->n_rules; f_pos++) {
1834 		ipfw_insn *cmd;
1835 		uint32_t tablearg = 0;
1836 		int l, cmdlen, skip_or; /* skip rest of OR block */
1837 		struct ip_fw *f;
1838 
1839 		f = chain->map[f_pos];
1840 		if (V_set_disable & (1 << f->set) )
1841 			continue;
1842 
1843 		skip_or = 0;
1844 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1845 		    l -= cmdlen, cmd += cmdlen) {
1846 			int match;
1847 
1848 			/*
1849 			 * check_body is a jump target used when we find a
1850 			 * CHECK_STATE, and need to jump to the body of
1851 			 * the target rule.
1852 			 */
1853 
1854 /* check_body: */
1855 			cmdlen = F_LEN(cmd);
1856 			/*
1857 			 * An OR block (insn_1 || .. || insn_n) has the
1858 			 * F_OR bit set in all but the last instruction.
1859 			 * The first match will set "skip_or", and cause
1860 			 * the following instructions to be skipped until
1861 			 * past the one with the F_OR bit clear.
1862 			 */
1863 			if (skip_or) {		/* skip this instruction */
1864 				if ((cmd->len & F_OR) == 0)
1865 					skip_or = 0;	/* next one is good */
1866 				continue;
1867 			}
1868 			match = 0; /* set to 1 if we succeed */
1869 
1870 			switch (cmd->opcode) {
1871 			/*
1872 			 * The first set of opcodes compares the packet's
1873 			 * fields with some pattern, setting 'match' if a
1874 			 * match is found. At the end of the loop there is
1875 			 * logic to deal with F_NOT and F_OR flags associated
1876 			 * with the opcode.
1877 			 */
1878 			case O_NOP:
1879 				match = 1;
1880 				break;
1881 
1882 			case O_FORWARD_MAC:
1883 				printf("ipfw: opcode %d unimplemented\n",
1884 				    cmd->opcode);
1885 				break;
1886 
1887 			case O_GID:
1888 			case O_UID:
1889 			case O_JAIL:
1890 				/*
1891 				 * We only check offset == 0 && proto != 0,
1892 				 * as this ensures that we have a
1893 				 * packet with the ports info.
1894 				 */
1895 				if (offset != 0)
1896 					break;
1897 				if (proto == IPPROTO_TCP ||
1898 				    proto == IPPROTO_UDP ||
1899 				    proto == IPPROTO_UDPLITE)
1900 					match = check_uidgid(
1901 						    (ipfw_insn_u32 *)cmd,
1902 						    args, &ucred_lookup,
1903 #ifdef __FreeBSD__
1904 						    &ucred_cache);
1905 #else
1906 						    (void *)&ucred_cache);
1907 #endif
1908 				break;
1909 
1910 			case O_RECV:
1911 				match = iface_match(iif, (ipfw_insn_if *)cmd,
1912 				    chain, &tablearg);
1913 				break;
1914 
1915 			case O_XMIT:
1916 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1917 				    chain, &tablearg);
1918 				break;
1919 
1920 			case O_VIA:
1921 				match = iface_match(args->ifp,
1922 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1923 				break;
1924 
1925 			case O_MACADDR2:
1926 				if (args->flags & IPFW_ARGS_ETHER) {
1927 					u_int32_t *want = (u_int32_t *)
1928 						((ipfw_insn_mac *)cmd)->addr;
1929 					u_int32_t *mask = (u_int32_t *)
1930 						((ipfw_insn_mac *)cmd)->mask;
1931 					u_int32_t *hdr = (u_int32_t *)eh;
1932 
1933 					match =
1934 					    ( want[0] == (hdr[0] & mask[0]) &&
1935 					      want[1] == (hdr[1] & mask[1]) &&
1936 					      want[2] == (hdr[2] & mask[2]) );
1937 				}
1938 				break;
1939 
1940 			case O_MAC_TYPE:
1941 				if (args->flags & IPFW_ARGS_ETHER) {
1942 					u_int16_t *p =
1943 					    ((ipfw_insn_u16 *)cmd)->ports;
1944 					int i;
1945 
1946 					for (i = cmdlen - 1; !match && i>0;
1947 					    i--, p += 2)
1948 						match =
1949 						    (ntohs(eh->ether_type) >=
1950 						    p[0] &&
1951 						    ntohs(eh->ether_type) <=
1952 						    p[1]);
1953 				}
1954 				break;
1955 
1956 			case O_FRAG:
1957 				if (is_ipv4) {
1958 					/*
1959 					 * Since flags_match() works with
1960 					 * uint8_t we pack ip_off into 8 bits.
1961 					 * For this match offset is a boolean.
1962 					 */
1963 					match = flags_match(cmd,
1964 					    ((ntohs(ip->ip_off) & ~IP_OFFMASK)
1965 					    >> 8) | (offset != 0));
1966 				} else {
1967 					/*
1968 					 * Compatiblity: historically bare
1969 					 * "frag" would match IPv6 fragments.
1970 					 */
1971 					match = (cmd->arg1 == 0x1 &&
1972 					    (offset != 0));
1973 				}
1974 				break;
1975 
1976 			case O_IN:	/* "out" is "not in" */
1977 				match = (oif == NULL);
1978 				break;
1979 
1980 			case O_LAYER2:
1981 				match = (args->flags & IPFW_ARGS_ETHER);
1982 				break;
1983 
1984 			case O_DIVERTED:
1985 				if ((args->flags & IPFW_ARGS_REF) == 0)
1986 					break;
1987 				/*
1988 				 * For diverted packets, args->rule.info
1989 				 * contains the divert port (in host format)
1990 				 * reason and direction.
1991 				 */
1992 				match = ((args->rule.info & IPFW_IS_MASK) ==
1993 				    IPFW_IS_DIVERT) && (
1994 				    ((args->rule.info & IPFW_INFO_IN) ?
1995 					1: 2) & cmd->arg1);
1996 				break;
1997 
1998 			case O_PROTO:
1999 				/*
2000 				 * We do not allow an arg of 0 so the
2001 				 * check of "proto" only suffices.
2002 				 */
2003 				match = (proto == cmd->arg1);
2004 				break;
2005 
2006 			case O_IP_SRC:
2007 				match = is_ipv4 &&
2008 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2009 				    src_ip.s_addr);
2010 				break;
2011 
2012 			case O_IP_DST_LOOKUP:
2013 			{
2014 				void *pkey;
2015 				uint32_t vidx, key;
2016 				uint16_t keylen;
2017 
2018 				if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
2019 					/* Determine lookup key type */
2020 					vidx = ((ipfw_insn_u32 *)cmd)->d[1];
2021 					if (vidx != 4 /* uid */ &&
2022 					    vidx != 5 /* jail */ &&
2023 					    is_ipv6 == 0 && is_ipv4 == 0)
2024 						break;
2025 					/* Determine key length */
2026 					if (vidx == 0 /* dst-ip */ ||
2027 					    vidx == 1 /* src-ip */)
2028 						keylen = is_ipv6 ?
2029 						    sizeof(struct in6_addr):
2030 						    sizeof(in_addr_t);
2031 					else {
2032 						keylen = sizeof(key);
2033 						pkey = &key;
2034 					}
2035 					if (vidx == 0 /* dst-ip */)
2036 						pkey = is_ipv4 ? (void *)&dst_ip:
2037 						    (void *)&args->f_id.dst_ip6;
2038 					else if (vidx == 1 /* src-ip */)
2039 						pkey = is_ipv4 ? (void *)&src_ip:
2040 						    (void *)&args->f_id.src_ip6;
2041 					else if (vidx == 6 /* dscp */) {
2042 						if (is_ipv4)
2043 							key = ip->ip_tos >> 2;
2044 						else {
2045 							key = args->f_id.flow_id6;
2046 							key = (key & 0x0f) << 2 |
2047 							    (key & 0xf000) >> 14;
2048 						}
2049 						key &= 0x3f;
2050 					} else if (vidx == 2 /* dst-port */ ||
2051 					    vidx == 3 /* src-port */) {
2052 						/* Skip fragments */
2053 						if (offset != 0)
2054 							break;
2055 						/* Skip proto without ports */
2056 						if (proto != IPPROTO_TCP &&
2057 						    proto != IPPROTO_UDP &&
2058 						    proto != IPPROTO_UDPLITE &&
2059 						    proto != IPPROTO_SCTP)
2060 							break;
2061 						if (vidx == 2 /* dst-port */)
2062 							key = dst_port;
2063 						else
2064 							key = src_port;
2065 					}
2066 #ifndef USERSPACE
2067 					else if (vidx == 4 /* uid */ ||
2068 					    vidx == 5 /* jail */) {
2069 						check_uidgid(
2070 						    (ipfw_insn_u32 *)cmd,
2071 						    args, &ucred_lookup,
2072 #ifdef __FreeBSD__
2073 						    &ucred_cache);
2074 						if (vidx == 4 /* uid */)
2075 							key = ucred_cache->cr_uid;
2076 						else if (vidx == 5 /* jail */)
2077 							key = ucred_cache->cr_prison->pr_id;
2078 #else /* !__FreeBSD__ */
2079 						    (void *)&ucred_cache);
2080 						if (vidx == 4 /* uid */)
2081 							key = ucred_cache.uid;
2082 						else if (vidx == 5 /* jail */)
2083 							key = ucred_cache.xid;
2084 #endif /* !__FreeBSD__ */
2085 					}
2086 #endif /* !USERSPACE */
2087 					else
2088 						break;
2089 					match = ipfw_lookup_table(chain,
2090 					    cmd->arg1, keylen, pkey, &vidx);
2091 					if (!match)
2092 						break;
2093 					tablearg = vidx;
2094 					break;
2095 				}
2096 				/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2097 				/* FALLTHROUGH */
2098 			}
2099 			case O_IP_SRC_LOOKUP:
2100 			{
2101 				void *pkey;
2102 				uint32_t vidx;
2103 				uint16_t keylen;
2104 
2105 				if (is_ipv4) {
2106 					keylen = sizeof(in_addr_t);
2107 					if (cmd->opcode == O_IP_DST_LOOKUP)
2108 						pkey = &dst_ip;
2109 					else
2110 						pkey = &src_ip;
2111 				} else if (is_ipv6) {
2112 					keylen = sizeof(struct in6_addr);
2113 					if (cmd->opcode == O_IP_DST_LOOKUP)
2114 						pkey = &args->f_id.dst_ip6;
2115 					else
2116 						pkey = &args->f_id.src_ip6;
2117 				} else
2118 					break;
2119 				match = ipfw_lookup_table(chain, cmd->arg1,
2120 				    keylen, pkey, &vidx);
2121 				if (!match)
2122 					break;
2123 				if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2124 					match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2125 					    TARG_VAL(chain, vidx, tag);
2126 					if (!match)
2127 						break;
2128 				}
2129 				tablearg = vidx;
2130 				break;
2131 			}
2132 
2133 			case O_IP_FLOW_LOOKUP:
2134 				{
2135 					uint32_t v = 0;
2136 					match = ipfw_lookup_table(chain,
2137 					    cmd->arg1, 0, &args->f_id, &v);
2138 					if (!match)
2139 						break;
2140 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2141 						match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2142 						    TARG_VAL(chain, v, tag);
2143 					if (match)
2144 						tablearg = v;
2145 				}
2146 				break;
2147 			case O_IP_SRC_MASK:
2148 			case O_IP_DST_MASK:
2149 				if (is_ipv4) {
2150 				    uint32_t a =
2151 					(cmd->opcode == O_IP_DST_MASK) ?
2152 					    dst_ip.s_addr : src_ip.s_addr;
2153 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2154 				    int i = cmdlen-1;
2155 
2156 				    for (; !match && i>0; i-= 2, p+= 2)
2157 					match = (p[0] == (a & p[1]));
2158 				}
2159 				break;
2160 
2161 			case O_IP_SRC_ME:
2162 				if (is_ipv4) {
2163 					match = in_localip(src_ip);
2164 					break;
2165 				}
2166 #ifdef INET6
2167 				/* FALLTHROUGH */
2168 			case O_IP6_SRC_ME:
2169 				match = is_ipv6 &&
2170 				    ipfw_localip6(&args->f_id.src_ip6);
2171 #endif
2172 				break;
2173 
2174 			case O_IP_DST_SET:
2175 			case O_IP_SRC_SET:
2176 				if (is_ipv4) {
2177 					u_int32_t *d = (u_int32_t *)(cmd+1);
2178 					u_int32_t addr =
2179 					    cmd->opcode == O_IP_DST_SET ?
2180 						args->f_id.dst_ip :
2181 						args->f_id.src_ip;
2182 
2183 					    if (addr < d[0])
2184 						    break;
2185 					    addr -= d[0]; /* subtract base */
2186 					    match = (addr < cmd->arg1) &&
2187 						( d[ 1 + (addr>>5)] &
2188 						  (1<<(addr & 0x1f)) );
2189 				}
2190 				break;
2191 
2192 			case O_IP_DST:
2193 				match = is_ipv4 &&
2194 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2195 				    dst_ip.s_addr);
2196 				break;
2197 
2198 			case O_IP_DST_ME:
2199 				if (is_ipv4) {
2200 					match = in_localip(dst_ip);
2201 					break;
2202 				}
2203 #ifdef INET6
2204 				/* FALLTHROUGH */
2205 			case O_IP6_DST_ME:
2206 				match = is_ipv6 &&
2207 				    ipfw_localip6(&args->f_id.dst_ip6);
2208 #endif
2209 				break;
2210 
2211 			case O_IP_SRCPORT:
2212 			case O_IP_DSTPORT:
2213 				/*
2214 				 * offset == 0 && proto != 0 is enough
2215 				 * to guarantee that we have a
2216 				 * packet with port info.
2217 				 */
2218 				if ((proto == IPPROTO_UDP ||
2219 				    proto == IPPROTO_UDPLITE ||
2220 				    proto == IPPROTO_TCP ||
2221 				    proto == IPPROTO_SCTP) && offset == 0) {
2222 					u_int16_t x =
2223 					    (cmd->opcode == O_IP_SRCPORT) ?
2224 						src_port : dst_port ;
2225 					u_int16_t *p =
2226 					    ((ipfw_insn_u16 *)cmd)->ports;
2227 					int i;
2228 
2229 					for (i = cmdlen - 1; !match && i>0;
2230 					    i--, p += 2)
2231 						match = (x>=p[0] && x<=p[1]);
2232 				}
2233 				break;
2234 
2235 			case O_ICMPTYPE:
2236 				match = (offset == 0 && proto==IPPROTO_ICMP &&
2237 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2238 				break;
2239 
2240 #ifdef INET6
2241 			case O_ICMP6TYPE:
2242 				match = is_ipv6 && offset == 0 &&
2243 				    proto==IPPROTO_ICMPV6 &&
2244 				    icmp6type_match(
2245 					ICMP6(ulp)->icmp6_type,
2246 					(ipfw_insn_u32 *)cmd);
2247 				break;
2248 #endif /* INET6 */
2249 
2250 			case O_IPOPT:
2251 				match = (is_ipv4 &&
2252 				    ipopts_match(ip, cmd) );
2253 				break;
2254 
2255 			case O_IPVER:
2256 				match = ((is_ipv4 || is_ipv6) &&
2257 				    cmd->arg1 == ip->ip_v);
2258 				break;
2259 
2260 			case O_IPID:
2261 			case O_IPTTL:
2262 				if (!is_ipv4)
2263 					break;
2264 			case O_IPLEN:
2265 				{	/* only for IP packets */
2266 				    uint16_t x;
2267 				    uint16_t *p;
2268 				    int i;
2269 
2270 				    if (cmd->opcode == O_IPLEN)
2271 					x = iplen;
2272 				    else if (cmd->opcode == O_IPTTL)
2273 					x = ip->ip_ttl;
2274 				    else /* must be IPID */
2275 					x = ntohs(ip->ip_id);
2276 				    if (cmdlen == 1) {
2277 					match = (cmd->arg1 == x);
2278 					break;
2279 				    }
2280 				    /* otherwise we have ranges */
2281 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2282 				    i = cmdlen - 1;
2283 				    for (; !match && i>0; i--, p += 2)
2284 					match = (x >= p[0] && x <= p[1]);
2285 				}
2286 				break;
2287 
2288 			case O_IPPRECEDENCE:
2289 				match = (is_ipv4 &&
2290 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2291 				break;
2292 
2293 			case O_IPTOS:
2294 				match = (is_ipv4 &&
2295 				    flags_match(cmd, ip->ip_tos));
2296 				break;
2297 
2298 			case O_DSCP:
2299 			    {
2300 				uint32_t *p;
2301 				uint16_t x;
2302 
2303 				p = ((ipfw_insn_u32 *)cmd)->d;
2304 
2305 				if (is_ipv4)
2306 					x = ip->ip_tos >> 2;
2307 				else if (is_ipv6) {
2308 					uint8_t *v;
2309 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2310 					x = (*v & 0x0F) << 2;
2311 					v++;
2312 					x |= *v >> 6;
2313 				} else
2314 					break;
2315 
2316 				/* DSCP bitmask is stored as low_u32 high_u32 */
2317 				if (x >= 32)
2318 					match = *(p + 1) & (1 << (x - 32));
2319 				else
2320 					match = *p & (1 << x);
2321 			    }
2322 				break;
2323 
2324 			case O_TCPDATALEN:
2325 				if (proto == IPPROTO_TCP && offset == 0) {
2326 				    struct tcphdr *tcp;
2327 				    uint16_t x;
2328 				    uint16_t *p;
2329 				    int i;
2330 #ifdef INET6
2331 				    if (is_ipv6) {
2332 					    struct ip6_hdr *ip6;
2333 
2334 					    ip6 = (struct ip6_hdr *)ip;
2335 					    if (ip6->ip6_plen == 0) {
2336 						    /*
2337 						     * Jumbo payload is not
2338 						     * supported by this
2339 						     * opcode.
2340 						     */
2341 						    break;
2342 					    }
2343 					    x = iplen - hlen;
2344 				    } else
2345 #endif /* INET6 */
2346 					    x = iplen - (ip->ip_hl << 2);
2347 				    tcp = TCP(ulp);
2348 				    x -= tcp->th_off << 2;
2349 				    if (cmdlen == 1) {
2350 					match = (cmd->arg1 == x);
2351 					break;
2352 				    }
2353 				    /* otherwise we have ranges */
2354 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2355 				    i = cmdlen - 1;
2356 				    for (; !match && i>0; i--, p += 2)
2357 					match = (x >= p[0] && x <= p[1]);
2358 				}
2359 				break;
2360 
2361 			case O_TCPFLAGS:
2362 				match = (proto == IPPROTO_TCP && offset == 0 &&
2363 				    flags_match(cmd, TCP(ulp)->th_flags));
2364 				break;
2365 
2366 			case O_TCPOPTS:
2367 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
2368 					PULLUP_LEN_LOCKED(hlen, ulp,
2369 					    (TCP(ulp)->th_off << 2));
2370 					match = tcpopts_match(TCP(ulp), cmd);
2371 				}
2372 				break;
2373 
2374 			case O_TCPSEQ:
2375 				match = (proto == IPPROTO_TCP && offset == 0 &&
2376 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2377 					TCP(ulp)->th_seq);
2378 				break;
2379 
2380 			case O_TCPACK:
2381 				match = (proto == IPPROTO_TCP && offset == 0 &&
2382 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2383 					TCP(ulp)->th_ack);
2384 				break;
2385 
2386 			case O_TCPMSS:
2387 				if (proto == IPPROTO_TCP &&
2388 				    (args->f_id._flags & TH_SYN) != 0 &&
2389 				    ulp != NULL) {
2390 					uint16_t mss, *p;
2391 					int i;
2392 
2393 					PULLUP_LEN_LOCKED(hlen, ulp,
2394 					    (TCP(ulp)->th_off << 2));
2395 					if ((tcpopts_parse(TCP(ulp), &mss) &
2396 					    IP_FW_TCPOPT_MSS) == 0)
2397 						break;
2398 					if (cmdlen == 1) {
2399 						match = (cmd->arg1 == mss);
2400 						break;
2401 					}
2402 					/* Otherwise we have ranges. */
2403 					p = ((ipfw_insn_u16 *)cmd)->ports;
2404 					i = cmdlen - 1;
2405 					for (; !match && i > 0; i--, p += 2)
2406 						match = (mss >= p[0] &&
2407 						    mss <= p[1]);
2408 				}
2409 				break;
2410 
2411 			case O_TCPWIN:
2412 				if (proto == IPPROTO_TCP && offset == 0) {
2413 				    uint16_t x;
2414 				    uint16_t *p;
2415 				    int i;
2416 
2417 				    x = ntohs(TCP(ulp)->th_win);
2418 				    if (cmdlen == 1) {
2419 					match = (cmd->arg1 == x);
2420 					break;
2421 				    }
2422 				    /* Otherwise we have ranges. */
2423 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2424 				    i = cmdlen - 1;
2425 				    for (; !match && i > 0; i--, p += 2)
2426 					match = (x >= p[0] && x <= p[1]);
2427 				}
2428 				break;
2429 
2430 			case O_ESTAB:
2431 				/* reject packets which have SYN only */
2432 				/* XXX should i also check for TH_ACK ? */
2433 				match = (proto == IPPROTO_TCP && offset == 0 &&
2434 				    (TCP(ulp)->th_flags &
2435 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2436 				break;
2437 
2438 			case O_ALTQ: {
2439 				struct pf_mtag *at;
2440 				struct m_tag *mtag;
2441 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2442 
2443 				/*
2444 				 * ALTQ uses mbuf tags from another
2445 				 * packet filtering system - pf(4).
2446 				 * We allocate a tag in its format
2447 				 * and fill it in, pretending to be pf(4).
2448 				 */
2449 				match = 1;
2450 				at = pf_find_mtag(m);
2451 				if (at != NULL && at->qid != 0)
2452 					break;
2453 				mtag = m_tag_get(PACKET_TAG_PF,
2454 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2455 				if (mtag == NULL) {
2456 					/*
2457 					 * Let the packet fall back to the
2458 					 * default ALTQ.
2459 					 */
2460 					break;
2461 				}
2462 				m_tag_prepend(m, mtag);
2463 				at = (struct pf_mtag *)(mtag + 1);
2464 				at->qid = altq->qid;
2465 				at->hdr = ip;
2466 				break;
2467 			}
2468 
2469 			case O_LOG:
2470 				ipfw_log(chain, f, hlen, args,
2471 				    offset | ip6f_mf, tablearg, ip);
2472 				match = 1;
2473 				break;
2474 
2475 			case O_PROB:
2476 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2477 				break;
2478 
2479 			case O_VERREVPATH:
2480 				/* Outgoing packets automatically pass/match */
2481 				match = (args->flags & IPFW_ARGS_OUT ||
2482 				    (
2483 #ifdef INET6
2484 				    is_ipv6 ?
2485 					verify_path6(&(args->f_id.src_ip6),
2486 					    iif, args->f_id.fib) :
2487 #endif
2488 				    verify_path(src_ip, iif, args->f_id.fib)));
2489 				break;
2490 
2491 			case O_VERSRCREACH:
2492 				/* Outgoing packets automatically pass/match */
2493 				match = (hlen > 0 && ((oif != NULL) || (
2494 #ifdef INET6
2495 				    is_ipv6 ?
2496 				        verify_path6(&(args->f_id.src_ip6),
2497 				            NULL, args->f_id.fib) :
2498 #endif
2499 				    verify_path(src_ip, NULL, args->f_id.fib))));
2500 				break;
2501 
2502 			case O_ANTISPOOF:
2503 				/* Outgoing packets automatically pass/match */
2504 				if (oif == NULL && hlen > 0 &&
2505 				    (  (is_ipv4 && in_localaddr(src_ip))
2506 #ifdef INET6
2507 				    || (is_ipv6 &&
2508 				        in6_localaddr(&(args->f_id.src_ip6)))
2509 #endif
2510 				    ))
2511 					match =
2512 #ifdef INET6
2513 					    is_ipv6 ? verify_path6(
2514 					        &(args->f_id.src_ip6), iif,
2515 						args->f_id.fib) :
2516 #endif
2517 					    verify_path(src_ip, iif,
2518 					        args->f_id.fib);
2519 				else
2520 					match = 1;
2521 				break;
2522 
2523 			case O_IPSEC:
2524 				match = (m_tag_find(m,
2525 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2526 				/* otherwise no match */
2527 				break;
2528 
2529 #ifdef INET6
2530 			case O_IP6_SRC:
2531 				match = is_ipv6 &&
2532 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2533 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2534 				break;
2535 
2536 			case O_IP6_DST:
2537 				match = is_ipv6 &&
2538 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2539 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2540 				break;
2541 			case O_IP6_SRC_MASK:
2542 			case O_IP6_DST_MASK:
2543 				if (is_ipv6) {
2544 					int i = cmdlen - 1;
2545 					struct in6_addr p;
2546 					struct in6_addr *d =
2547 					    &((ipfw_insn_ip6 *)cmd)->addr6;
2548 
2549 					for (; !match && i > 0; d += 2,
2550 					    i -= F_INSN_SIZE(struct in6_addr)
2551 					    * 2) {
2552 						p = (cmd->opcode ==
2553 						    O_IP6_SRC_MASK) ?
2554 						    args->f_id.src_ip6:
2555 						    args->f_id.dst_ip6;
2556 						APPLY_MASK(&p, &d[1]);
2557 						match =
2558 						    IN6_ARE_ADDR_EQUAL(&d[0],
2559 						    &p);
2560 					}
2561 				}
2562 				break;
2563 
2564 			case O_FLOW6ID:
2565 				match = is_ipv6 &&
2566 				    flow6id_match(args->f_id.flow_id6,
2567 				    (ipfw_insn_u32 *) cmd);
2568 				break;
2569 
2570 			case O_EXT_HDR:
2571 				match = is_ipv6 &&
2572 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
2573 				break;
2574 
2575 			case O_IP6:
2576 				match = is_ipv6;
2577 				break;
2578 #endif
2579 
2580 			case O_IP4:
2581 				match = is_ipv4;
2582 				break;
2583 
2584 			case O_TAG: {
2585 				struct m_tag *mtag;
2586 				uint32_t tag = TARG(cmd->arg1, tag);
2587 
2588 				/* Packet is already tagged with this tag? */
2589 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2590 
2591 				/* We have `untag' action when F_NOT flag is
2592 				 * present. And we must remove this mtag from
2593 				 * mbuf and reset `match' to zero (`match' will
2594 				 * be inversed later).
2595 				 * Otherwise we should allocate new mtag and
2596 				 * push it into mbuf.
2597 				 */
2598 				if (cmd->len & F_NOT) { /* `untag' action */
2599 					if (mtag != NULL)
2600 						m_tag_delete(m, mtag);
2601 					match = 0;
2602 				} else {
2603 					if (mtag == NULL) {
2604 						mtag = m_tag_alloc( MTAG_IPFW,
2605 						    tag, 0, M_NOWAIT);
2606 						if (mtag != NULL)
2607 							m_tag_prepend(m, mtag);
2608 					}
2609 					match = 1;
2610 				}
2611 				break;
2612 			}
2613 
2614 			case O_FIB: /* try match the specified fib */
2615 				if (args->f_id.fib == cmd->arg1)
2616 					match = 1;
2617 				break;
2618 
2619 			case O_SOCKARG:	{
2620 #ifndef USERSPACE	/* not supported in userspace */
2621 				struct inpcb *inp = args->inp;
2622 				struct inpcbinfo *pi;
2623 
2624 				if (is_ipv6) /* XXX can we remove this ? */
2625 					break;
2626 
2627 				if (proto == IPPROTO_TCP)
2628 					pi = &V_tcbinfo;
2629 				else if (proto == IPPROTO_UDP)
2630 					pi = &V_udbinfo;
2631 				else if (proto == IPPROTO_UDPLITE)
2632 					pi = &V_ulitecbinfo;
2633 				else
2634 					break;
2635 
2636 				/*
2637 				 * XXXRW: so_user_cookie should almost
2638 				 * certainly be inp_user_cookie?
2639 				 */
2640 
2641 				/* For incoming packet, lookup up the
2642 				inpcb using the src/dest ip/port tuple */
2643 				if (inp == NULL) {
2644 					inp = in_pcblookup(pi,
2645 						src_ip, htons(src_port),
2646 						dst_ip, htons(dst_port),
2647 						INPLOOKUP_RLOCKPCB, NULL);
2648 					if (inp != NULL) {
2649 						tablearg =
2650 						    inp->inp_socket->so_user_cookie;
2651 						if (tablearg)
2652 							match = 1;
2653 						INP_RUNLOCK(inp);
2654 					}
2655 				} else {
2656 					if (inp->inp_socket) {
2657 						tablearg =
2658 						    inp->inp_socket->so_user_cookie;
2659 						if (tablearg)
2660 							match = 1;
2661 					}
2662 				}
2663 #endif /* !USERSPACE */
2664 				break;
2665 			}
2666 
2667 			case O_TAGGED: {
2668 				struct m_tag *mtag;
2669 				uint32_t tag = TARG(cmd->arg1, tag);
2670 
2671 				if (cmdlen == 1) {
2672 					match = m_tag_locate(m, MTAG_IPFW,
2673 					    tag, NULL) != NULL;
2674 					break;
2675 				}
2676 
2677 				/* we have ranges */
2678 				for (mtag = m_tag_first(m);
2679 				    mtag != NULL && !match;
2680 				    mtag = m_tag_next(m, mtag)) {
2681 					uint16_t *p;
2682 					int i;
2683 
2684 					if (mtag->m_tag_cookie != MTAG_IPFW)
2685 						continue;
2686 
2687 					p = ((ipfw_insn_u16 *)cmd)->ports;
2688 					i = cmdlen - 1;
2689 					for(; !match && i > 0; i--, p += 2)
2690 						match =
2691 						    mtag->m_tag_id >= p[0] &&
2692 						    mtag->m_tag_id <= p[1];
2693 				}
2694 				break;
2695 			}
2696 
2697 			/*
2698 			 * The second set of opcodes represents 'actions',
2699 			 * i.e. the terminal part of a rule once the packet
2700 			 * matches all previous patterns.
2701 			 * Typically there is only one action for each rule,
2702 			 * and the opcode is stored at the end of the rule
2703 			 * (but there are exceptions -- see below).
2704 			 *
2705 			 * In general, here we set retval and terminate the
2706 			 * outer loop (would be a 'break 3' in some language,
2707 			 * but we need to set l=0, done=1)
2708 			 *
2709 			 * Exceptions:
2710 			 * O_COUNT and O_SKIPTO actions:
2711 			 *   instead of terminating, we jump to the next rule
2712 			 *   (setting l=0), or to the SKIPTO target (setting
2713 			 *   f/f_len, cmd and l as needed), respectively.
2714 			 *
2715 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2716 			 *   perform some action and set match = 1;
2717 			 *
2718 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2719 			 *   not real 'actions', and are stored right
2720 			 *   before the 'action' part of the rule (one
2721 			 *   exception is O_SKIP_ACTION which could be
2722 			 *   between these opcodes and 'action' one).
2723 			 *   These opcodes try to install an entry in the
2724 			 *   state tables; if successful, we continue with
2725 			 *   the next opcode (match=1; break;), otherwise
2726 			 *   the packet must be dropped (set retval,
2727 			 *   break loops with l=0, done=1)
2728 			 *
2729 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2730 			 *   cause a lookup of the state table, and a jump
2731 			 *   to the 'action' part of the parent rule
2732 			 *   if an entry is found, or
2733 			 *   (CHECK_STATE only) a jump to the next rule if
2734 			 *   the entry is not found.
2735 			 *   The result of the lookup is cached so that
2736 			 *   further instances of these opcodes become NOPs.
2737 			 *   The jump to the next rule is done by setting
2738 			 *   l=0, cmdlen=0.
2739 			 *
2740 			 * O_SKIP_ACTION: this opcode is not a real 'action'
2741 			 *  either, and is stored right before the 'action'
2742 			 *  part of the rule, right after the O_KEEP_STATE
2743 			 *  opcode. It causes match failure so the real
2744 			 *  'action' could be executed only if the rule
2745 			 *  is checked via dynamic rule from the state
2746 			 *  table, as in such case execution starts
2747 			 *  from the true 'action' opcode directly.
2748 			 *
2749 			 */
2750 			case O_LIMIT:
2751 			case O_KEEP_STATE:
2752 				if (ipfw_dyn_install_state(chain, f,
2753 				    (ipfw_insn_limit *)cmd, args, ulp,
2754 				    pktlen, &dyn_info, tablearg)) {
2755 					/* error or limit violation */
2756 					retval = IP_FW_DENY;
2757 					l = 0;	/* exit inner loop */
2758 					done = 1; /* exit outer loop */
2759 				}
2760 				match = 1;
2761 				break;
2762 
2763 			case O_PROBE_STATE:
2764 			case O_CHECK_STATE:
2765 				/*
2766 				 * dynamic rules are checked at the first
2767 				 * keep-state or check-state occurrence,
2768 				 * with the result being stored in dyn_info.
2769 				 * The compiler introduces a PROBE_STATE
2770 				 * instruction for us when we have a
2771 				 * KEEP_STATE (because PROBE_STATE needs
2772 				 * to be run first).
2773 				 */
2774 				if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2775 				    (q = ipfw_dyn_lookup_state(args, ulp,
2776 				    pktlen, cmd, &dyn_info)) != NULL) {
2777 					/*
2778 					 * Found dynamic entry, jump to the
2779 					 * 'action' part of the parent rule
2780 					 * by setting f, cmd, l and clearing
2781 					 * cmdlen.
2782 					 */
2783 					f = q;
2784 					f_pos = dyn_info.f_pos;
2785 					cmd = ACTION_PTR(f);
2786 					l = f->cmd_len - f->act_ofs;
2787 					cmdlen = 0;
2788 					match = 1;
2789 					break;
2790 				}
2791 				/*
2792 				 * Dynamic entry not found. If CHECK_STATE,
2793 				 * skip to next rule, if PROBE_STATE just
2794 				 * ignore and continue with next opcode.
2795 				 */
2796 				if (cmd->opcode == O_CHECK_STATE)
2797 					l = 0;	/* exit inner loop */
2798 				match = 1;
2799 				break;
2800 
2801 			case O_SKIP_ACTION:
2802 				match = 0;	/* skip to the next rule */
2803 				l = 0;		/* exit inner loop */
2804 				break;
2805 
2806 			case O_ACCEPT:
2807 				retval = 0;	/* accept */
2808 				l = 0;		/* exit inner loop */
2809 				done = 1;	/* exit outer loop */
2810 				break;
2811 
2812 			case O_PIPE:
2813 			case O_QUEUE:
2814 				set_match(args, f_pos, chain);
2815 				args->rule.info = TARG(cmd->arg1, pipe);
2816 				if (cmd->opcode == O_PIPE)
2817 					args->rule.info |= IPFW_IS_PIPE;
2818 				if (V_fw_one_pass)
2819 					args->rule.info |= IPFW_ONEPASS;
2820 				retval = IP_FW_DUMMYNET;
2821 				l = 0;          /* exit inner loop */
2822 				done = 1;       /* exit outer loop */
2823 				break;
2824 
2825 			case O_DIVERT:
2826 			case O_TEE:
2827 				if (args->flags & IPFW_ARGS_ETHER)
2828 					break;	/* not on layer 2 */
2829 				/* otherwise this is terminal */
2830 				l = 0;		/* exit inner loop */
2831 				done = 1;	/* exit outer loop */
2832 				retval = (cmd->opcode == O_DIVERT) ?
2833 					IP_FW_DIVERT : IP_FW_TEE;
2834 				set_match(args, f_pos, chain);
2835 				args->rule.info = TARG(cmd->arg1, divert);
2836 				break;
2837 
2838 			case O_COUNT:
2839 				IPFW_INC_RULE_COUNTER(f, pktlen);
2840 				l = 0;		/* exit inner loop */
2841 				break;
2842 
2843 			case O_SKIPTO:
2844 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2845 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2846 			    /*
2847 			     * Skip disabled rules, and re-enter
2848 			     * the inner loop with the correct
2849 			     * f_pos, f, l and cmd.
2850 			     * Also clear cmdlen and skip_or
2851 			     */
2852 			    for (; f_pos < chain->n_rules - 1 &&
2853 				    (V_set_disable &
2854 				     (1 << chain->map[f_pos]->set));
2855 				    f_pos++)
2856 				;
2857 			    /* Re-enter the inner loop at the skipto rule. */
2858 			    f = chain->map[f_pos];
2859 			    l = f->cmd_len;
2860 			    cmd = f->cmd;
2861 			    match = 1;
2862 			    cmdlen = 0;
2863 			    skip_or = 0;
2864 			    continue;
2865 			    break;	/* not reached */
2866 
2867 			case O_CALLRETURN: {
2868 				/*
2869 				 * Implementation of `subroutine' call/return,
2870 				 * in the stack carried in an mbuf tag. This
2871 				 * is different from `skipto' in that any call
2872 				 * address is possible (`skipto' must prevent
2873 				 * backward jumps to avoid endless loops).
2874 				 * We have `return' action when F_NOT flag is
2875 				 * present. The `m_tag_id' field is used as
2876 				 * stack pointer.
2877 				 */
2878 				struct m_tag *mtag;
2879 				uint16_t jmpto, *stack;
2880 
2881 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2882 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2883 				/*
2884 				 * Hand-rolled version of m_tag_locate() with
2885 				 * wildcard `type'.
2886 				 * If not already tagged, allocate new tag.
2887 				 */
2888 				mtag = m_tag_first(m);
2889 				while (mtag != NULL) {
2890 					if (mtag->m_tag_cookie ==
2891 					    MTAG_IPFW_CALL)
2892 						break;
2893 					mtag = m_tag_next(m, mtag);
2894 				}
2895 				if (mtag == NULL && IS_CALL) {
2896 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2897 					    IPFW_CALLSTACK_SIZE *
2898 					    sizeof(uint16_t), M_NOWAIT);
2899 					if (mtag != NULL)
2900 						m_tag_prepend(m, mtag);
2901 				}
2902 
2903 				/*
2904 				 * On error both `call' and `return' just
2905 				 * continue with next rule.
2906 				 */
2907 				if (IS_RETURN && (mtag == NULL ||
2908 				    mtag->m_tag_id == 0)) {
2909 					l = 0;		/* exit inner loop */
2910 					break;
2911 				}
2912 				if (IS_CALL && (mtag == NULL ||
2913 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2914 					printf("ipfw: call stack error, "
2915 					    "go to next rule\n");
2916 					l = 0;		/* exit inner loop */
2917 					break;
2918 				}
2919 
2920 				IPFW_INC_RULE_COUNTER(f, pktlen);
2921 				stack = (uint16_t *)(mtag + 1);
2922 
2923 				/*
2924 				 * The `call' action may use cached f_pos
2925 				 * (in f->next_rule), whose version is written
2926 				 * in f->next_rule.
2927 				 * The `return' action, however, doesn't have
2928 				 * fixed jump address in cmd->arg1 and can't use
2929 				 * cache.
2930 				 */
2931 				if (IS_CALL) {
2932 					stack[mtag->m_tag_id] = f->rulenum;
2933 					mtag->m_tag_id++;
2934 			    		f_pos = JUMP(chain, f, cmd->arg1,
2935 					    tablearg, 1);
2936 				} else {	/* `return' action */
2937 					mtag->m_tag_id--;
2938 					jmpto = stack[mtag->m_tag_id] + 1;
2939 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2940 				}
2941 
2942 				/*
2943 				 * Skip disabled rules, and re-enter
2944 				 * the inner loop with the correct
2945 				 * f_pos, f, l and cmd.
2946 				 * Also clear cmdlen and skip_or
2947 				 */
2948 				for (; f_pos < chain->n_rules - 1 &&
2949 				    (V_set_disable &
2950 				    (1 << chain->map[f_pos]->set)); f_pos++)
2951 					;
2952 				/* Re-enter the inner loop at the dest rule. */
2953 				f = chain->map[f_pos];
2954 				l = f->cmd_len;
2955 				cmd = f->cmd;
2956 				cmdlen = 0;
2957 				skip_or = 0;
2958 				continue;
2959 				break;	/* NOTREACHED */
2960 			}
2961 #undef IS_CALL
2962 #undef IS_RETURN
2963 
2964 			case O_REJECT:
2965 				/*
2966 				 * Drop the packet and send a reject notice
2967 				 * if the packet is not ICMP (or is an ICMP
2968 				 * query), and it is not multicast/broadcast.
2969 				 */
2970 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2971 				    (proto != IPPROTO_ICMP ||
2972 				     is_icmp_query(ICMP(ulp))) &&
2973 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2974 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2975 					send_reject(args, cmd->arg1, iplen, ip);
2976 					m = args->m;
2977 				}
2978 				/* FALLTHROUGH */
2979 #ifdef INET6
2980 			case O_UNREACH6:
2981 				if (hlen > 0 && is_ipv6 &&
2982 				    ((offset & IP6F_OFF_MASK) == 0) &&
2983 				    (proto != IPPROTO_ICMPV6 ||
2984 				     (is_icmp6_query(icmp6_type) == 1)) &&
2985 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2986 				    !IN6_IS_ADDR_MULTICAST(
2987 					&args->f_id.dst_ip6)) {
2988 					send_reject6(args,
2989 					    cmd->opcode == O_REJECT ?
2990 					    map_icmp_unreach(cmd->arg1):
2991 					    cmd->arg1, hlen,
2992 					    (struct ip6_hdr *)ip);
2993 					m = args->m;
2994 				}
2995 				/* FALLTHROUGH */
2996 #endif
2997 			case O_DENY:
2998 				retval = IP_FW_DENY;
2999 				l = 0;		/* exit inner loop */
3000 				done = 1;	/* exit outer loop */
3001 				break;
3002 
3003 			case O_FORWARD_IP:
3004 				if (args->flags & IPFW_ARGS_ETHER)
3005 					break;	/* not valid on layer2 pkts */
3006 				if (q != f ||
3007 				    dyn_info.direction == MATCH_FORWARD) {
3008 				    struct sockaddr_in *sa;
3009 
3010 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
3011 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
3012 #ifdef INET6
3013 					/*
3014 					 * We use O_FORWARD_IP opcode for
3015 					 * fwd rule with tablearg, but tables
3016 					 * now support IPv6 addresses. And
3017 					 * when we are inspecting IPv6 packet,
3018 					 * we can use nh6 field from
3019 					 * table_value as next_hop6 address.
3020 					 */
3021 					if (is_ipv6) {
3022 						struct ip_fw_nh6 *nh6;
3023 
3024 						args->flags |= IPFW_ARGS_NH6;
3025 						nh6 = &args->hopstore6;
3026 						nh6->sin6_addr = TARG_VAL(
3027 						    chain, tablearg, nh6);
3028 						nh6->sin6_port = sa->sin_port;
3029 						nh6->sin6_scope_id = TARG_VAL(
3030 						    chain, tablearg, zoneid);
3031 					} else
3032 #endif
3033 					{
3034 						args->flags |= IPFW_ARGS_NH4;
3035 						args->hopstore.sin_port =
3036 						    sa->sin_port;
3037 						sa = &args->hopstore;
3038 						sa->sin_family = AF_INET;
3039 						sa->sin_len = sizeof(*sa);
3040 						sa->sin_addr.s_addr = htonl(
3041 						    TARG_VAL(chain, tablearg,
3042 						    nh4));
3043 					}
3044 				    } else {
3045 					    args->flags |= IPFW_ARGS_NH4PTR;
3046 					    args->next_hop = sa;
3047 				    }
3048 				}
3049 				retval = IP_FW_PASS;
3050 				l = 0;          /* exit inner loop */
3051 				done = 1;       /* exit outer loop */
3052 				break;
3053 
3054 #ifdef INET6
3055 			case O_FORWARD_IP6:
3056 				if (args->flags & IPFW_ARGS_ETHER)
3057 					break;	/* not valid on layer2 pkts */
3058 				if (q != f ||
3059 				    dyn_info.direction == MATCH_FORWARD) {
3060 					struct sockaddr_in6 *sin6;
3061 
3062 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3063 					args->flags |= IPFW_ARGS_NH6PTR;
3064 					args->next_hop6 = sin6;
3065 				}
3066 				retval = IP_FW_PASS;
3067 				l = 0;		/* exit inner loop */
3068 				done = 1;	/* exit outer loop */
3069 				break;
3070 #endif
3071 
3072 			case O_NETGRAPH:
3073 			case O_NGTEE:
3074 				set_match(args, f_pos, chain);
3075 				args->rule.info = TARG(cmd->arg1, netgraph);
3076 				if (V_fw_one_pass)
3077 					args->rule.info |= IPFW_ONEPASS;
3078 				retval = (cmd->opcode == O_NETGRAPH) ?
3079 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
3080 				l = 0;          /* exit inner loop */
3081 				done = 1;       /* exit outer loop */
3082 				break;
3083 
3084 			case O_SETFIB: {
3085 				uint32_t fib;
3086 
3087 				IPFW_INC_RULE_COUNTER(f, pktlen);
3088 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
3089 				if (fib >= rt_numfibs)
3090 					fib = 0;
3091 				M_SETFIB(m, fib);
3092 				args->f_id.fib = fib; /* XXX */
3093 				l = 0;		/* exit inner loop */
3094 				break;
3095 		        }
3096 
3097 			case O_SETDSCP: {
3098 				uint16_t code;
3099 
3100 				code = TARG(cmd->arg1, dscp) & 0x3F;
3101 				l = 0;		/* exit inner loop */
3102 				if (is_ipv4) {
3103 					uint16_t old;
3104 
3105 					old = *(uint16_t *)ip;
3106 					ip->ip_tos = (code << 2) |
3107 					    (ip->ip_tos & 0x03);
3108 					ip->ip_sum = cksum_adjust(ip->ip_sum,
3109 					    old, *(uint16_t *)ip);
3110 				} else if (is_ipv6) {
3111 					uint8_t *v;
3112 
3113 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
3114 					*v = (*v & 0xF0) | (code >> 2);
3115 					v++;
3116 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
3117 				} else
3118 					break;
3119 
3120 				IPFW_INC_RULE_COUNTER(f, pktlen);
3121 				break;
3122 			}
3123 
3124 			case O_NAT:
3125 				l = 0;          /* exit inner loop */
3126 				done = 1;       /* exit outer loop */
3127 				/*
3128 				 * Ensure that we do not invoke NAT handler for
3129 				 * non IPv4 packets. Libalias expects only IPv4.
3130 				 */
3131 				if (!is_ipv4 || !IPFW_NAT_LOADED) {
3132 				    retval = IP_FW_DENY;
3133 				    break;
3134 				}
3135 
3136 				struct cfg_nat *t;
3137 				int nat_id;
3138 
3139 				args->rule.info = 0;
3140 				set_match(args, f_pos, chain);
3141 				/* Check if this is 'global' nat rule */
3142 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3143 					retval = ipfw_nat_ptr(args, NULL, m);
3144 					break;
3145 				}
3146 				t = ((ipfw_insn_nat *)cmd)->nat;
3147 				if (t == NULL) {
3148 					nat_id = TARG(cmd->arg1, nat);
3149 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3150 
3151 					if (t == NULL) {
3152 					    retval = IP_FW_DENY;
3153 					    break;
3154 					}
3155 					if (cmd->arg1 != IP_FW_TARG)
3156 					    ((ipfw_insn_nat *)cmd)->nat = t;
3157 				}
3158 				retval = ipfw_nat_ptr(args, t, m);
3159 				break;
3160 
3161 			case O_REASS: {
3162 				int ip_off;
3163 
3164 				l = 0;	/* in any case exit inner loop */
3165 				if (is_ipv6) /* IPv6 is not supported yet */
3166 					break;
3167 				IPFW_INC_RULE_COUNTER(f, pktlen);
3168 				ip_off = ntohs(ip->ip_off);
3169 
3170 				/* if not fragmented, go to next rule */
3171 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3172 				    break;
3173 
3174 				args->m = m = ip_reass(m);
3175 
3176 				/*
3177 				 * do IP header checksum fixup.
3178 				 */
3179 				if (m == NULL) { /* fragment got swallowed */
3180 				    retval = IP_FW_DENY;
3181 				} else { /* good, packet complete */
3182 				    int hlen;
3183 
3184 				    ip = mtod(m, struct ip *);
3185 				    hlen = ip->ip_hl << 2;
3186 				    ip->ip_sum = 0;
3187 				    if (hlen == sizeof(struct ip))
3188 					ip->ip_sum = in_cksum_hdr(ip);
3189 				    else
3190 					ip->ip_sum = in_cksum(m, hlen);
3191 				    retval = IP_FW_REASS;
3192 				    args->rule.info = 0;
3193 				    set_match(args, f_pos, chain);
3194 				}
3195 				done = 1;	/* exit outer loop */
3196 				break;
3197 			}
3198 			case O_EXTERNAL_ACTION:
3199 				l = 0; /* in any case exit inner loop */
3200 				retval = ipfw_run_eaction(chain, args,
3201 				    cmd, &done);
3202 				/*
3203 				 * If both @retval and @done are zero,
3204 				 * consider this as rule matching and
3205 				 * update counters.
3206 				 */
3207 				if (retval == 0 && done == 0) {
3208 					IPFW_INC_RULE_COUNTER(f, pktlen);
3209 					/*
3210 					 * Reset the result of the last
3211 					 * dynamic state lookup.
3212 					 * External action can change
3213 					 * @args content, and it may be
3214 					 * used for new state lookup later.
3215 					 */
3216 					DYN_INFO_INIT(&dyn_info);
3217 				}
3218 				break;
3219 
3220 			default:
3221 				panic("-- unknown opcode %d\n", cmd->opcode);
3222 			} /* end of switch() on opcodes */
3223 			/*
3224 			 * if we get here with l=0, then match is irrelevant.
3225 			 */
3226 
3227 			if (cmd->len & F_NOT)
3228 				match = !match;
3229 
3230 			if (match) {
3231 				if (cmd->len & F_OR)
3232 					skip_or = 1;
3233 			} else {
3234 				if (!(cmd->len & F_OR)) /* not an OR block, */
3235 					break;		/* try next rule    */
3236 			}
3237 
3238 		}	/* end of inner loop, scan opcodes */
3239 #undef PULLUP_LEN
3240 #undef PULLUP_LEN_LOCKED
3241 
3242 		if (done)
3243 			break;
3244 
3245 /* next_rule:; */	/* try next rule		*/
3246 
3247 	}		/* end of outer for, scan rules */
3248 
3249 	if (done) {
3250 		struct ip_fw *rule = chain->map[f_pos];
3251 		/* Update statistics */
3252 		IPFW_INC_RULE_COUNTER(rule, pktlen);
3253 		IPFW_PROBE(rule__matched, retval,
3254 		    is_ipv4 ? AF_INET : AF_INET6,
3255 		    is_ipv4 ? (uintptr_t)&src_ip :
3256 		        (uintptr_t)&args->f_id.src_ip6,
3257 		    is_ipv4 ? (uintptr_t)&dst_ip :
3258 		        (uintptr_t)&args->f_id.dst_ip6,
3259 		    args, rule);
3260 	} else {
3261 		retval = IP_FW_DENY;
3262 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3263 	}
3264 	IPFW_PF_RUNLOCK(chain);
3265 #ifdef __FreeBSD__
3266 	if (ucred_cache != NULL)
3267 		crfree(ucred_cache);
3268 #endif
3269 	return (retval);
3270 
3271 pullup_failed:
3272 	if (V_fw_verbose)
3273 		printf("ipfw: pullup failed\n");
3274 	return (IP_FW_DENY);
3275 }
3276 
3277 /*
3278  * Set maximum number of tables that can be used in given VNET ipfw instance.
3279  */
3280 #ifdef SYSCTL_NODE
3281 static int
3282 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3283 {
3284 	int error;
3285 	unsigned int ntables;
3286 
3287 	ntables = V_fw_tables_max;
3288 
3289 	error = sysctl_handle_int(oidp, &ntables, 0, req);
3290 	/* Read operation or some error */
3291 	if ((error != 0) || (req->newptr == NULL))
3292 		return (error);
3293 
3294 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
3295 }
3296 
3297 /*
3298  * Switches table namespace between global and per-set.
3299  */
3300 static int
3301 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3302 {
3303 	int error;
3304 	unsigned int sets;
3305 
3306 	sets = V_fw_tables_sets;
3307 
3308 	error = sysctl_handle_int(oidp, &sets, 0, req);
3309 	/* Read operation or some error */
3310 	if ((error != 0) || (req->newptr == NULL))
3311 		return (error);
3312 
3313 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3314 }
3315 #endif
3316 
3317 /*
3318  * Module and VNET glue
3319  */
3320 
3321 /*
3322  * Stuff that must be initialised only on boot or module load
3323  */
3324 static int
3325 ipfw_init(void)
3326 {
3327 	int error = 0;
3328 
3329 	/*
3330  	 * Only print out this stuff the first time around,
3331 	 * when called from the sysinit code.
3332 	 */
3333 	printf("ipfw2 "
3334 #ifdef INET6
3335 		"(+ipv6) "
3336 #endif
3337 		"initialized, divert %s, nat %s, "
3338 		"default to %s, logging ",
3339 #ifdef IPDIVERT
3340 		"enabled",
3341 #else
3342 		"loadable",
3343 #endif
3344 #ifdef IPFIREWALL_NAT
3345 		"enabled",
3346 #else
3347 		"loadable",
3348 #endif
3349 		default_to_accept ? "accept" : "deny");
3350 
3351 	/*
3352 	 * Note: V_xxx variables can be accessed here but the vnet specific
3353 	 * initializer may not have been called yet for the VIMAGE case.
3354 	 * Tuneables will have been processed. We will print out values for
3355 	 * the default vnet.
3356 	 * XXX This should all be rationalized AFTER 8.0
3357 	 */
3358 	if (V_fw_verbose == 0)
3359 		printf("disabled\n");
3360 	else if (V_verbose_limit == 0)
3361 		printf("unlimited\n");
3362 	else
3363 		printf("limited to %d packets/entry by default\n",
3364 		    V_verbose_limit);
3365 
3366 	/* Check user-supplied table count for validness */
3367 	if (default_fw_tables > IPFW_TABLES_MAX)
3368 	  default_fw_tables = IPFW_TABLES_MAX;
3369 
3370 	ipfw_init_sopt_handler();
3371 	ipfw_init_obj_rewriter();
3372 	ipfw_iface_init();
3373 	return (error);
3374 }
3375 
3376 /*
3377  * Called for the removal of the last instance only on module unload.
3378  */
3379 static void
3380 ipfw_destroy(void)
3381 {
3382 
3383 	ipfw_iface_destroy();
3384 	ipfw_destroy_sopt_handler();
3385 	ipfw_destroy_obj_rewriter();
3386 	printf("IP firewall unloaded\n");
3387 }
3388 
3389 /*
3390  * Stuff that must be initialized for every instance
3391  * (including the first of course).
3392  */
3393 static int
3394 vnet_ipfw_init(const void *unused)
3395 {
3396 	int error, first;
3397 	struct ip_fw *rule = NULL;
3398 	struct ip_fw_chain *chain;
3399 
3400 	chain = &V_layer3_chain;
3401 
3402 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3403 
3404 	/* First set up some values that are compile time options */
3405 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
3406 	V_fw_deny_unknown_exthdrs = 1;
3407 #ifdef IPFIREWALL_VERBOSE
3408 	V_fw_verbose = 1;
3409 #endif
3410 #ifdef IPFIREWALL_VERBOSE_LIMIT
3411 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3412 #endif
3413 #ifdef IPFIREWALL_NAT
3414 	LIST_INIT(&chain->nat);
3415 #endif
3416 
3417 	/* Init shared services hash table */
3418 	ipfw_init_srv(chain);
3419 
3420 	ipfw_init_counters();
3421 	/* Set initial number of tables */
3422 	V_fw_tables_max = default_fw_tables;
3423 	error = ipfw_init_tables(chain, first);
3424 	if (error) {
3425 		printf("ipfw2: setting up tables failed\n");
3426 		free(chain->map, M_IPFW);
3427 		free(rule, M_IPFW);
3428 		return (ENOSPC);
3429 	}
3430 
3431 	IPFW_LOCK_INIT(chain);
3432 
3433 	/* fill and insert the default rule */
3434 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3435 	rule->flags |= IPFW_RULE_NOOPT;
3436 	rule->cmd_len = 1;
3437 	rule->cmd[0].len = 1;
3438 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3439 	chain->default_rule = rule;
3440 	ipfw_add_protected_rule(chain, rule, 0);
3441 
3442 	ipfw_dyn_init(chain);
3443 	ipfw_eaction_init(chain, first);
3444 #ifdef LINEAR_SKIPTO
3445 	ipfw_init_skipto_cache(chain);
3446 #endif
3447 	ipfw_bpf_init(first);
3448 
3449 	/* First set up some values that are compile time options */
3450 	V_ipfw_vnet_ready = 1;		/* Open for business */
3451 
3452 	/*
3453 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3454 	 * Even if the latter two fail we still keep the module alive
3455 	 * because the sockopt and layer2 paths are still useful.
3456 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3457 	 * so we can ignore the exact return value and just set a flag.
3458 	 *
3459 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3460 	 * changes in the underlying (per-vnet) variables trigger
3461 	 * immediate hook()/unhook() calls.
3462 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
3463 	 * is checked on each packet because there are no pfil hooks.
3464 	 */
3465 	V_ip_fw_ctl_ptr = ipfw_ctl3;
3466 	error = ipfw_attach_hooks();
3467 	return (error);
3468 }
3469 
3470 /*
3471  * Called for the removal of each instance.
3472  */
3473 static int
3474 vnet_ipfw_uninit(const void *unused)
3475 {
3476 	struct ip_fw *reap;
3477 	struct ip_fw_chain *chain = &V_layer3_chain;
3478 	int i, last;
3479 
3480 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3481 	/*
3482 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
3483 	 * Then grab, release and grab again the WLOCK so we make
3484 	 * sure the update is propagated and nobody will be in.
3485 	 */
3486 	ipfw_detach_hooks();
3487 	V_ip_fw_ctl_ptr = NULL;
3488 
3489 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3490 
3491 	IPFW_UH_WLOCK(chain);
3492 	IPFW_UH_WUNLOCK(chain);
3493 
3494 	ipfw_dyn_uninit(0);	/* run the callout_drain */
3495 
3496 	IPFW_UH_WLOCK(chain);
3497 
3498 	reap = NULL;
3499 	IPFW_WLOCK(chain);
3500 	for (i = 0; i < chain->n_rules; i++)
3501 		ipfw_reap_add(chain, &reap, chain->map[i]);
3502 	free(chain->map, M_IPFW);
3503 #ifdef LINEAR_SKIPTO
3504 	ipfw_destroy_skipto_cache(chain);
3505 #endif
3506 	IPFW_WUNLOCK(chain);
3507 	IPFW_UH_WUNLOCK(chain);
3508 	ipfw_destroy_tables(chain, last);
3509 	ipfw_eaction_uninit(chain, last);
3510 	if (reap != NULL)
3511 		ipfw_reap_rules(reap);
3512 	vnet_ipfw_iface_destroy(chain);
3513 	ipfw_destroy_srv(chain);
3514 	IPFW_LOCK_DESTROY(chain);
3515 	ipfw_dyn_uninit(1);	/* free the remaining parts */
3516 	ipfw_destroy_counters();
3517 	ipfw_bpf_uninit(last);
3518 	return (0);
3519 }
3520 
3521 /*
3522  * Module event handler.
3523  * In general we have the choice of handling most of these events by the
3524  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3525  * use the SYSINIT handlers as they are more capable of expressing the
3526  * flow of control during module and vnet operations, so this is just
3527  * a skeleton. Note there is no SYSINIT equivalent of the module
3528  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3529  */
3530 static int
3531 ipfw_modevent(module_t mod, int type, void *unused)
3532 {
3533 	int err = 0;
3534 
3535 	switch (type) {
3536 	case MOD_LOAD:
3537 		/* Called once at module load or
3538 	 	 * system boot if compiled in. */
3539 		break;
3540 	case MOD_QUIESCE:
3541 		/* Called before unload. May veto unloading. */
3542 		break;
3543 	case MOD_UNLOAD:
3544 		/* Called during unload. */
3545 		break;
3546 	case MOD_SHUTDOWN:
3547 		/* Called during system shutdown. */
3548 		break;
3549 	default:
3550 		err = EOPNOTSUPP;
3551 		break;
3552 	}
3553 	return err;
3554 }
3555 
3556 static moduledata_t ipfwmod = {
3557 	"ipfw",
3558 	ipfw_modevent,
3559 	0
3560 };
3561 
3562 /* Define startup order. */
3563 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
3564 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
3565 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
3566 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
3567 
3568 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3569 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3570 MODULE_VERSION(ipfw, 3);
3571 /* should declare some dependencies here */
3572 
3573 /*
3574  * Starting up. Done in order after ipfwmod() has been called.
3575  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3576  */
3577 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3578 	    ipfw_init, NULL);
3579 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3580 	    vnet_ipfw_init, NULL);
3581 
3582 /*
3583  * Closing up shop. These are done in REVERSE ORDER, but still
3584  * after ipfwmod() has been called. Not called on reboot.
3585  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3586  * or when the module is unloaded.
3587  */
3588 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3589 	    ipfw_destroy, NULL);
3590 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3591 	    vnet_ipfw_uninit, NULL);
3592 /* end of file */
3593