xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision d7d962ead0b6e5e8a39202d0590022082bf5bfb6)
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 static int jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num,
148     int tablearg, int jump_backwards);
149 #ifndef LINEAR_SKIPTO
150 static int jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num,
151     int tablearg, int jump_backwards);
152 #define	JUMP(ch, f, num, targ, back)	jump_cached(ch, f, num, targ, back)
153 #else
154 #define	JUMP(ch, f, num, targ, back)	jump_lookup_pos(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 static int
1231 jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1232     int tablearg, int jump_backwards)
1233 {
1234 	int f_pos, i;
1235 
1236 	i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1237 	/* make sure we do not jump backward */
1238 	if (jump_backwards == 0 && i <= f->rulenum)
1239 		i = f->rulenum + 1;
1240 
1241 #ifndef LINEAR_SKIPTO
1242 	if (chain->idxmap != NULL)
1243 		f_pos = chain->idxmap[i];
1244 	else
1245 		f_pos = ipfw_find_rule(chain, i, 0);
1246 #else
1247 	f_pos = chain->idxmap[i];
1248 #endif /* LINEAR_SKIPTO */
1249 
1250 	return (f_pos);
1251 }
1252 
1253 
1254 #ifndef LINEAR_SKIPTO
1255 /*
1256  * Helper function to enable cached rule lookups using
1257  * cache.id and cache.pos fields in ipfw rule.
1258  */
1259 static int
1260 jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1261     int tablearg, int jump_backwards)
1262 {
1263 	int f_pos;
1264 
1265 	/* Can't use cache with IP_FW_TARG */
1266 	if (num == IP_FW_TARG)
1267 		return jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1268 
1269 	/*
1270 	 * If possible use cached f_pos (in f->cache.pos),
1271 	 * whose version is written in f->cache.id (horrible hacks
1272 	 * to avoid changing the ABI).
1273 	 *
1274 	 * Multiple threads can execute the same rule simultaneously,
1275 	 * we need to ensure that cache.pos is updated before cache.id.
1276 	 */
1277 
1278 #ifdef __LP64__
1279 	struct ip_fw_jump_cache cache;
1280 
1281 	cache.raw_value = f->cache.raw_value;
1282 	if (cache.id == chain->id)
1283 		return (cache.pos);
1284 
1285 	f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1286 
1287 	cache.pos = f_pos;
1288 	cache.id = chain->id;
1289 	f->cache.raw_value = cache.raw_value;
1290 #else
1291 	if (f->cache.id == chain->id) {
1292 		/* Load pos after id */
1293 		atomic_thread_fence_acq();
1294 		return (f->cache.pos);
1295 	}
1296 
1297 	f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1298 
1299 	f->cache.pos = f_pos;
1300 	/* Store id after pos */
1301 	atomic_thread_fence_rel();
1302 	f->cache.id = chain->id;
1303 #endif /* !__LP64__ */
1304 	return (f_pos);
1305 }
1306 #endif /* !LINEAR_SKIPTO */
1307 
1308 #define	TARG(k, f)	IP_FW_ARG_TABLEARG(chain, k, f)
1309 /*
1310  * The main check routine for the firewall.
1311  *
1312  * All arguments are in args so we can modify them and return them
1313  * back to the caller.
1314  *
1315  * Parameters:
1316  *
1317  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
1318  *		Starts with the IP header.
1319  *	args->L3offset	Number of bytes bypassed if we came from L2.
1320  *			e.g. often sizeof(eh)  ** NOTYET **
1321  *	args->ifp	Incoming or outgoing interface.
1322  *	args->divert_rule (in/out)
1323  *		Skip up to the first rule past this rule number;
1324  *		upon return, non-zero port number for divert or tee.
1325  *
1326  *	args->rule	Pointer to the last matching rule (in/out)
1327  *	args->next_hop	Socket we are forwarding to (out).
1328  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
1329  *	args->f_id	Addresses grabbed from the packet (out)
1330  * 	args->rule.info	a cookie depending on rule action
1331  *
1332  * Return value:
1333  *
1334  *	IP_FW_PASS	the packet must be accepted
1335  *	IP_FW_DENY	the packet must be dropped
1336  *	IP_FW_DIVERT	divert packet, port in m_tag
1337  *	IP_FW_TEE	tee packet, port in m_tag
1338  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
1339  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
1340  *		args->rule contains the matching rule,
1341  *		args->rule.info has additional information.
1342  *
1343  */
1344 int
1345 ipfw_chk(struct ip_fw_args *args)
1346 {
1347 
1348 	/*
1349 	 * Local variables holding state while processing a packet:
1350 	 *
1351 	 * IMPORTANT NOTE: to speed up the processing of rules, there
1352 	 * are some assumption on the values of the variables, which
1353 	 * are documented here. Should you change them, please check
1354 	 * the implementation of the various instructions to make sure
1355 	 * that they still work.
1356 	 *
1357 	 * m | args->m	Pointer to the mbuf, as received from the caller.
1358 	 *	It may change if ipfw_chk() does an m_pullup, or if it
1359 	 *	consumes the packet because it calls send_reject().
1360 	 *	XXX This has to change, so that ipfw_chk() never modifies
1361 	 *	or consumes the buffer.
1362 	 *	OR
1363 	 * args->mem	Pointer to contigous memory chunk.
1364 	 * ip	Is the beginning of the ip(4 or 6) header.
1365 	 * eh	Ethernet header in case if input is Layer2.
1366 	 */
1367 	struct mbuf *m;
1368 	struct ip *ip;
1369 	struct ether_header *eh;
1370 
1371 	/*
1372 	 * For rules which contain uid/gid or jail constraints, cache
1373 	 * a copy of the users credentials after the pcb lookup has been
1374 	 * executed. This will speed up the processing of rules with
1375 	 * these types of constraints, as well as decrease contention
1376 	 * on pcb related locks.
1377 	 */
1378 #ifndef __FreeBSD__
1379 	struct bsd_ucred ucred_cache;
1380 #else
1381 	struct ucred *ucred_cache = NULL;
1382 #endif
1383 	int ucred_lookup = 0;
1384 	int f_pos = 0;		/* index of current rule in the array */
1385 	int retval = 0;
1386 	struct ifnet *oif, *iif;
1387 
1388 	/*
1389 	 * hlen	The length of the IP header.
1390 	 */
1391 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
1392 
1393 	/*
1394 	 * offset	The offset of a fragment. offset != 0 means that
1395 	 *	we have a fragment at this offset of an IPv4 packet.
1396 	 *	offset == 0 means that (if this is an IPv4 packet)
1397 	 *	this is the first or only fragment.
1398 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1399 	 *	or there is a single packet fragment (fragment header added
1400 	 *	without needed).  We will treat a single packet fragment as if
1401 	 *	there was no fragment header (or log/block depending on the
1402 	 *	V_fw_permit_single_frag6 sysctl setting).
1403 	 */
1404 	u_short offset = 0;
1405 	u_short ip6f_mf = 0;
1406 
1407 	/*
1408 	 * Local copies of addresses. They are only valid if we have
1409 	 * an IP packet.
1410 	 *
1411 	 * proto	The protocol. Set to 0 for non-ip packets,
1412 	 *	or to the protocol read from the packet otherwise.
1413 	 *	proto != 0 means that we have an IPv4 packet.
1414 	 *
1415 	 * src_port, dst_port	port numbers, in HOST format. Only
1416 	 *	valid for TCP and UDP packets.
1417 	 *
1418 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
1419 	 *	Only valid for IPv4 packets.
1420 	 */
1421 	uint8_t proto;
1422 	uint16_t src_port, dst_port;		/* NOTE: host format	*/
1423 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
1424 	int iplen = 0;
1425 	int pktlen;
1426 
1427 	struct ipfw_dyn_info dyn_info;
1428 	struct ip_fw *q = NULL;
1429 	struct ip_fw_chain *chain = &V_layer3_chain;
1430 
1431 	/*
1432 	 * We store in ulp a pointer to the upper layer protocol header.
1433 	 * In the ipv4 case this is easy to determine from the header,
1434 	 * but for ipv6 we might have some additional headers in the middle.
1435 	 * ulp is NULL if not found.
1436 	 */
1437 	void *ulp = NULL;		/* upper layer protocol pointer. */
1438 
1439 	/* XXX ipv6 variables */
1440 	int is_ipv6 = 0;
1441 	uint8_t	icmp6_type = 0;
1442 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
1443 	/* end of ipv6 variables */
1444 
1445 	int is_ipv4 = 0;
1446 
1447 	int done = 0;		/* flag to exit the outer loop */
1448 	IPFW_RLOCK_TRACKER;
1449 	bool mem;
1450 
1451 	if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1452 		if (args->flags & IPFW_ARGS_ETHER) {
1453 			eh = (struct ether_header *)args->mem;
1454 			if (eh->ether_type == htons(ETHERTYPE_VLAN))
1455 				ip = (struct ip *)
1456 				    ((struct ether_vlan_header *)eh + 1);
1457 			else
1458 				ip = (struct ip *)(eh + 1);
1459 		} else {
1460 			eh = NULL;
1461 			ip = (struct ip *)args->mem;
1462 		}
1463 		pktlen = IPFW_ARGS_LENGTH(args->flags);
1464 		args->f_id.fib = args->ifp->if_fib;	/* best guess */
1465 	} else {
1466 		m = args->m;
1467 		if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1468 			return (IP_FW_PASS);	/* accept */
1469 		if (args->flags & IPFW_ARGS_ETHER) {
1470 	                /* We need some amount of data to be contiguous. */
1471 			if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1472 			    (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1473 			    max_protohdr))) == NULL)
1474 				goto pullup_failed;
1475 			eh = mtod(m, struct ether_header *);
1476 			ip = (struct ip *)(eh + 1);
1477 		} else {
1478 			eh = NULL;
1479 			ip = mtod(m, struct ip *);
1480 		}
1481 		pktlen = m->m_pkthdr.len;
1482 		args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1483 	}
1484 
1485 	dst_ip.s_addr = 0;		/* make sure it is initialized */
1486 	src_ip.s_addr = 0;		/* make sure it is initialized */
1487 	src_port = dst_port = 0;
1488 
1489 	DYN_INFO_INIT(&dyn_info);
1490 /*
1491  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1492  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1493  * pointer might become stale after other pullups (but we never use it
1494  * this way).
1495  */
1496 #define	PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
1497 #define	EHLEN	(eh != NULL ? ((char *)ip - (char *)eh) : 0)
1498 #define	_PULLUP_LOCKED(_len, p, T, unlock)			\
1499 do {								\
1500 	int x = (_len) + T + EHLEN;				\
1501 	if (mem) {						\
1502 		if (__predict_false(pktlen < x)) {		\
1503 			unlock;					\
1504 			goto pullup_failed;			\
1505 		}						\
1506 		p = (char *)args->mem + (_len) + EHLEN;		\
1507 	} else {						\
1508 		if (__predict_false((m)->m_len < x)) {		\
1509 			args->m = m = m_pullup(m, x);		\
1510 			if (m == NULL) {			\
1511 				unlock;				\
1512 				goto pullup_failed;		\
1513 			}					\
1514 		}						\
1515 		p = mtod(m, char *) + (_len) + EHLEN;		\
1516 	}							\
1517 } while (0)
1518 
1519 #define	PULLUP_LEN(_len, p, T)	_PULLUP_LOCKED(_len, p, T, )
1520 #define	PULLUP_LEN_LOCKED(_len, p, T)	\
1521     _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain));	\
1522     UPDATE_POINTERS()
1523 /*
1524  * In case pointers got stale after pullups, update them.
1525  */
1526 #define	UPDATE_POINTERS()					\
1527 do {								\
1528 	if (!mem) {						\
1529 		if (eh != NULL) {				\
1530 			eh = mtod(m, struct ether_header *);	\
1531 			ip = (struct ip *)(eh + 1);		\
1532 		} else						\
1533 			ip = mtod(m, struct ip *);		\
1534 		args->m = m;					\
1535 	}							\
1536 } while (0)
1537 
1538 	/* Identify IP packets and fill up variables. */
1539 	if (pktlen >= sizeof(struct ip6_hdr) &&
1540 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1541 	    ip->ip_v == 6) {
1542 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1543 
1544 		is_ipv6 = 1;
1545 		args->flags |= IPFW_ARGS_IP6;
1546 		hlen = sizeof(struct ip6_hdr);
1547 		proto = ip6->ip6_nxt;
1548 		/* Search extension headers to find upper layer protocols */
1549 		while (ulp == NULL && offset == 0) {
1550 			switch (proto) {
1551 			case IPPROTO_ICMPV6:
1552 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1553 				icmp6_type = ICMP6(ulp)->icmp6_type;
1554 				break;
1555 
1556 			case IPPROTO_TCP:
1557 				PULLUP_TO(hlen, ulp, struct tcphdr);
1558 				dst_port = TCP(ulp)->th_dport;
1559 				src_port = TCP(ulp)->th_sport;
1560 				/* save flags for dynamic rules */
1561 				args->f_id._flags = TCP(ulp)->th_flags;
1562 				break;
1563 
1564 			case IPPROTO_SCTP:
1565 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1566 				    sizeof(struct sctp_chunkhdr) +
1567 				    offsetof(struct sctp_init, a_rwnd))
1568 					PULLUP_LEN(hlen, ulp,
1569 					    sizeof(struct sctphdr) +
1570 					    sizeof(struct sctp_chunkhdr) +
1571 					    offsetof(struct sctp_init, a_rwnd));
1572 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1573 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1574 				else
1575 					PULLUP_LEN(hlen, ulp,
1576 					    sizeof(struct sctphdr));
1577 				src_port = SCTP(ulp)->src_port;
1578 				dst_port = SCTP(ulp)->dest_port;
1579 				break;
1580 
1581 			case IPPROTO_UDP:
1582 			case IPPROTO_UDPLITE:
1583 				PULLUP_TO(hlen, ulp, struct udphdr);
1584 				dst_port = UDP(ulp)->uh_dport;
1585 				src_port = UDP(ulp)->uh_sport;
1586 				break;
1587 
1588 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1589 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1590 				ext_hd |= EXT_HOPOPTS;
1591 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1592 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1593 				ulp = NULL;
1594 				break;
1595 
1596 			case IPPROTO_ROUTING:	/* RFC 2460 */
1597 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1598 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1599 				case 0:
1600 					ext_hd |= EXT_RTHDR0;
1601 					break;
1602 				case 2:
1603 					ext_hd |= EXT_RTHDR2;
1604 					break;
1605 				default:
1606 					if (V_fw_verbose)
1607 						printf("IPFW2: IPV6 - Unknown "
1608 						    "Routing Header type(%d)\n",
1609 						    ((struct ip6_rthdr *)
1610 						    ulp)->ip6r_type);
1611 					if (V_fw_deny_unknown_exthdrs)
1612 					    return (IP_FW_DENY);
1613 					break;
1614 				}
1615 				ext_hd |= EXT_ROUTING;
1616 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1617 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1618 				ulp = NULL;
1619 				break;
1620 
1621 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1622 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1623 				ext_hd |= EXT_FRAGMENT;
1624 				hlen += sizeof (struct ip6_frag);
1625 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1626 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1627 					IP6F_OFF_MASK;
1628 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1629 					IP6F_MORE_FRAG;
1630 				if (V_fw_permit_single_frag6 == 0 &&
1631 				    offset == 0 && ip6f_mf == 0) {
1632 					if (V_fw_verbose)
1633 						printf("IPFW2: IPV6 - Invalid "
1634 						    "Fragment Header\n");
1635 					if (V_fw_deny_unknown_exthdrs)
1636 					    return (IP_FW_DENY);
1637 					break;
1638 				}
1639 				args->f_id.extra =
1640 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1641 				ulp = NULL;
1642 				break;
1643 
1644 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1645 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1646 				ext_hd |= EXT_DSTOPTS;
1647 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1648 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1649 				ulp = NULL;
1650 				break;
1651 
1652 			case IPPROTO_AH:	/* RFC 2402 */
1653 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1654 				ext_hd |= EXT_AH;
1655 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1656 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1657 				ulp = NULL;
1658 				break;
1659 
1660 			case IPPROTO_ESP:	/* RFC 2406 */
1661 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1662 				/* Anything past Seq# is variable length and
1663 				 * data past this ext. header is encrypted. */
1664 				ext_hd |= EXT_ESP;
1665 				break;
1666 
1667 			case IPPROTO_NONE:	/* RFC 2460 */
1668 				/*
1669 				 * Packet ends here, and IPv6 header has
1670 				 * already been pulled up. If ip6e_len!=0
1671 				 * then octets must be ignored.
1672 				 */
1673 				ulp = ip; /* non-NULL to get out of loop. */
1674 				break;
1675 
1676 			case IPPROTO_OSPFIGP:
1677 				/* XXX OSPF header check? */
1678 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1679 				break;
1680 
1681 			case IPPROTO_PIM:
1682 				/* XXX PIM header check? */
1683 				PULLUP_TO(hlen, ulp, struct pim);
1684 				break;
1685 
1686 			case IPPROTO_GRE:	/* RFC 1701 */
1687 				/* XXX GRE header check? */
1688 				PULLUP_TO(hlen, ulp, struct grehdr);
1689 				break;
1690 
1691 			case IPPROTO_CARP:
1692 				PULLUP_TO(hlen, ulp, offsetof(
1693 				    struct carp_header, carp_counter));
1694 				if (CARP_ADVERTISEMENT !=
1695 				    ((struct carp_header *)ulp)->carp_type)
1696 					return (IP_FW_DENY);
1697 				break;
1698 
1699 			case IPPROTO_IPV6:	/* RFC 2893 */
1700 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1701 				break;
1702 
1703 			case IPPROTO_IPV4:	/* RFC 2893 */
1704 				PULLUP_TO(hlen, ulp, struct ip);
1705 				break;
1706 
1707 			default:
1708 				if (V_fw_verbose)
1709 					printf("IPFW2: IPV6 - Unknown "
1710 					    "Extension Header(%d), ext_hd=%x\n",
1711 					     proto, ext_hd);
1712 				if (V_fw_deny_unknown_exthdrs)
1713 				    return (IP_FW_DENY);
1714 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1715 				break;
1716 			} /*switch */
1717 		}
1718 		UPDATE_POINTERS();
1719 		ip6 = (struct ip6_hdr *)ip;
1720 		args->f_id.addr_type = 6;
1721 		args->f_id.src_ip6 = ip6->ip6_src;
1722 		args->f_id.dst_ip6 = ip6->ip6_dst;
1723 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1724 		iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1725 	} else if (pktlen >= sizeof(struct ip) &&
1726 	    (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1727 	    ip->ip_v == 4) {
1728 		is_ipv4 = 1;
1729 		args->flags |= IPFW_ARGS_IP4;
1730 		hlen = ip->ip_hl << 2;
1731 		/*
1732 		 * Collect parameters into local variables for faster
1733 		 * matching.
1734 		 */
1735 		proto = ip->ip_p;
1736 		src_ip = ip->ip_src;
1737 		dst_ip = ip->ip_dst;
1738 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1739 		iplen = ntohs(ip->ip_len);
1740 
1741 		if (offset == 0) {
1742 			switch (proto) {
1743 			case IPPROTO_TCP:
1744 				PULLUP_TO(hlen, ulp, struct tcphdr);
1745 				dst_port = TCP(ulp)->th_dport;
1746 				src_port = TCP(ulp)->th_sport;
1747 				/* save flags for dynamic rules */
1748 				args->f_id._flags = TCP(ulp)->th_flags;
1749 				break;
1750 
1751 			case IPPROTO_SCTP:
1752 				if (pktlen >= hlen + sizeof(struct sctphdr) +
1753 				    sizeof(struct sctp_chunkhdr) +
1754 				    offsetof(struct sctp_init, a_rwnd))
1755 					PULLUP_LEN(hlen, ulp,
1756 					    sizeof(struct sctphdr) +
1757 					    sizeof(struct sctp_chunkhdr) +
1758 					    offsetof(struct sctp_init, a_rwnd));
1759 				else if (pktlen >= hlen + sizeof(struct sctphdr))
1760 					PULLUP_LEN(hlen, ulp, pktlen - hlen);
1761 				else
1762 					PULLUP_LEN(hlen, ulp,
1763 					    sizeof(struct sctphdr));
1764 				src_port = SCTP(ulp)->src_port;
1765 				dst_port = SCTP(ulp)->dest_port;
1766 				break;
1767 
1768 			case IPPROTO_UDP:
1769 			case IPPROTO_UDPLITE:
1770 				PULLUP_TO(hlen, ulp, struct udphdr);
1771 				dst_port = UDP(ulp)->uh_dport;
1772 				src_port = UDP(ulp)->uh_sport;
1773 				break;
1774 
1775 			case IPPROTO_ICMP:
1776 				PULLUP_TO(hlen, ulp, struct icmphdr);
1777 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1778 				break;
1779 
1780 			default:
1781 				break;
1782 			}
1783 		} else {
1784 			if (offset == 1 && proto == IPPROTO_TCP) {
1785 				/* RFC 3128 */
1786 				goto pullup_failed;
1787 			}
1788 		}
1789 
1790 		UPDATE_POINTERS();
1791 		args->f_id.addr_type = 4;
1792 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1793 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1794 	} else {
1795 		proto = 0;
1796 		dst_ip.s_addr = src_ip.s_addr = 0;
1797 
1798 		args->f_id.addr_type = 1; /* XXX */
1799 	}
1800 #undef PULLUP_TO
1801 	pktlen = iplen < pktlen ? iplen: pktlen;
1802 
1803 	/* Properly initialize the rest of f_id */
1804 	args->f_id.proto = proto;
1805 	args->f_id.src_port = src_port = ntohs(src_port);
1806 	args->f_id.dst_port = dst_port = ntohs(dst_port);
1807 
1808 	IPFW_PF_RLOCK(chain);
1809 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1810 		IPFW_PF_RUNLOCK(chain);
1811 		return (IP_FW_PASS);	/* accept */
1812 	}
1813 	if (args->flags & IPFW_ARGS_REF) {
1814 		/*
1815 		 * Packet has already been tagged as a result of a previous
1816 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1817 		 * REASS, NETGRAPH, DIVERT/TEE...)
1818 		 * Validate the slot and continue from the next one
1819 		 * if still present, otherwise do a lookup.
1820 		 */
1821 		f_pos = (args->rule.chain_id == chain->id) ?
1822 		    args->rule.slot :
1823 		    ipfw_find_rule(chain, args->rule.rulenum,
1824 			args->rule.rule_id);
1825 	} else {
1826 		f_pos = 0;
1827 	}
1828 
1829 	if (args->flags & IPFW_ARGS_IN) {
1830 		iif = args->ifp;
1831 		oif = NULL;
1832 	} else {
1833 		MPASS(args->flags & IPFW_ARGS_OUT);
1834 		iif = mem ? NULL : m_rcvif(m);
1835 		oif = args->ifp;
1836 	}
1837 
1838 	/*
1839 	 * Now scan the rules, and parse microinstructions for each rule.
1840 	 * We have two nested loops and an inner switch. Sometimes we
1841 	 * need to break out of one or both loops, or re-enter one of
1842 	 * the loops with updated variables. Loop variables are:
1843 	 *
1844 	 *	f_pos (outer loop) points to the current rule.
1845 	 *		On output it points to the matching rule.
1846 	 *	done (outer loop) is used as a flag to break the loop.
1847 	 *	l (inner loop)	residual length of current rule.
1848 	 *		cmd points to the current microinstruction.
1849 	 *
1850 	 * We break the inner loop by setting l=0 and possibly
1851 	 * cmdlen=0 if we don't want to advance cmd.
1852 	 * We break the outer loop by setting done=1
1853 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1854 	 * as needed.
1855 	 */
1856 	for (; f_pos < chain->n_rules; f_pos++) {
1857 		ipfw_insn *cmd;
1858 		uint32_t tablearg = 0;
1859 		int l, cmdlen, skip_or; /* skip rest of OR block */
1860 		struct ip_fw *f;
1861 
1862 		f = chain->map[f_pos];
1863 		if (V_set_disable & (1 << f->set) )
1864 			continue;
1865 
1866 		skip_or = 0;
1867 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1868 		    l -= cmdlen, cmd += cmdlen) {
1869 			int match;
1870 
1871 			/*
1872 			 * check_body is a jump target used when we find a
1873 			 * CHECK_STATE, and need to jump to the body of
1874 			 * the target rule.
1875 			 */
1876 
1877 /* check_body: */
1878 			cmdlen = F_LEN(cmd);
1879 			/*
1880 			 * An OR block (insn_1 || .. || insn_n) has the
1881 			 * F_OR bit set in all but the last instruction.
1882 			 * The first match will set "skip_or", and cause
1883 			 * the following instructions to be skipped until
1884 			 * past the one with the F_OR bit clear.
1885 			 */
1886 			if (skip_or) {		/* skip this instruction */
1887 				if ((cmd->len & F_OR) == 0)
1888 					skip_or = 0;	/* next one is good */
1889 				continue;
1890 			}
1891 			match = 0; /* set to 1 if we succeed */
1892 
1893 			switch (cmd->opcode) {
1894 			/*
1895 			 * The first set of opcodes compares the packet's
1896 			 * fields with some pattern, setting 'match' if a
1897 			 * match is found. At the end of the loop there is
1898 			 * logic to deal with F_NOT and F_OR flags associated
1899 			 * with the opcode.
1900 			 */
1901 			case O_NOP:
1902 				match = 1;
1903 				break;
1904 
1905 			case O_FORWARD_MAC:
1906 				printf("ipfw: opcode %d unimplemented\n",
1907 				    cmd->opcode);
1908 				break;
1909 
1910 			case O_GID:
1911 			case O_UID:
1912 			case O_JAIL:
1913 				/*
1914 				 * We only check offset == 0 && proto != 0,
1915 				 * as this ensures that we have a
1916 				 * packet with the ports info.
1917 				 */
1918 				if (offset != 0)
1919 					break;
1920 				if (proto == IPPROTO_TCP ||
1921 				    proto == IPPROTO_UDP ||
1922 				    proto == IPPROTO_UDPLITE)
1923 					match = check_uidgid(
1924 						    (ipfw_insn_u32 *)cmd,
1925 						    args, &ucred_lookup,
1926 #ifdef __FreeBSD__
1927 						    &ucred_cache);
1928 #else
1929 						    (void *)&ucred_cache);
1930 #endif
1931 				break;
1932 
1933 			case O_RECV:
1934 				match = iface_match(iif, (ipfw_insn_if *)cmd,
1935 				    chain, &tablearg);
1936 				break;
1937 
1938 			case O_XMIT:
1939 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1940 				    chain, &tablearg);
1941 				break;
1942 
1943 			case O_VIA:
1944 				match = iface_match(args->ifp,
1945 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1946 				break;
1947 
1948 			case O_MACADDR2:
1949 				if (args->flags & IPFW_ARGS_ETHER) {
1950 					u_int32_t *want = (u_int32_t *)
1951 						((ipfw_insn_mac *)cmd)->addr;
1952 					u_int32_t *mask = (u_int32_t *)
1953 						((ipfw_insn_mac *)cmd)->mask;
1954 					u_int32_t *hdr = (u_int32_t *)eh;
1955 
1956 					match =
1957 					    ( want[0] == (hdr[0] & mask[0]) &&
1958 					      want[1] == (hdr[1] & mask[1]) &&
1959 					      want[2] == (hdr[2] & mask[2]) );
1960 				}
1961 				break;
1962 
1963 			case O_MAC_TYPE:
1964 				if (args->flags & IPFW_ARGS_ETHER) {
1965 					u_int16_t *p =
1966 					    ((ipfw_insn_u16 *)cmd)->ports;
1967 					int i;
1968 
1969 					for (i = cmdlen - 1; !match && i>0;
1970 					    i--, p += 2)
1971 						match =
1972 						    (ntohs(eh->ether_type) >=
1973 						    p[0] &&
1974 						    ntohs(eh->ether_type) <=
1975 						    p[1]);
1976 				}
1977 				break;
1978 
1979 			case O_FRAG:
1980 				if (is_ipv4) {
1981 					/*
1982 					 * Since flags_match() works with
1983 					 * uint8_t we pack ip_off into 8 bits.
1984 					 * For this match offset is a boolean.
1985 					 */
1986 					match = flags_match(cmd,
1987 					    ((ntohs(ip->ip_off) & ~IP_OFFMASK)
1988 					    >> 8) | (offset != 0));
1989 				} else {
1990 					/*
1991 					 * Compatiblity: historically bare
1992 					 * "frag" would match IPv6 fragments.
1993 					 */
1994 					match = (cmd->arg1 == 0x1 &&
1995 					    (offset != 0));
1996 				}
1997 				break;
1998 
1999 			case O_IN:	/* "out" is "not in" */
2000 				match = (oif == NULL);
2001 				break;
2002 
2003 			case O_LAYER2:
2004 				match = (args->flags & IPFW_ARGS_ETHER);
2005 				break;
2006 
2007 			case O_DIVERTED:
2008 				if ((args->flags & IPFW_ARGS_REF) == 0)
2009 					break;
2010 				/*
2011 				 * For diverted packets, args->rule.info
2012 				 * contains the divert port (in host format)
2013 				 * reason and direction.
2014 				 */
2015 				match = ((args->rule.info & IPFW_IS_MASK) ==
2016 				    IPFW_IS_DIVERT) && (
2017 				    ((args->rule.info & IPFW_INFO_IN) ?
2018 					1: 2) & cmd->arg1);
2019 				break;
2020 
2021 			case O_PROTO:
2022 				/*
2023 				 * We do not allow an arg of 0 so the
2024 				 * check of "proto" only suffices.
2025 				 */
2026 				match = (proto == cmd->arg1);
2027 				break;
2028 
2029 			case O_IP_SRC:
2030 				match = is_ipv4 &&
2031 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2032 				    src_ip.s_addr);
2033 				break;
2034 
2035 			case O_IP_DST_LOOKUP:
2036 			{
2037 				void *pkey;
2038 				uint32_t vidx, key;
2039 				uint16_t keylen;
2040 
2041 				if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
2042 					/* Determine lookup key type */
2043 					vidx = ((ipfw_insn_u32 *)cmd)->d[1];
2044 					if (vidx != 4 /* uid */ &&
2045 					    vidx != 5 /* jail */ &&
2046 					    is_ipv6 == 0 && is_ipv4 == 0)
2047 						break;
2048 					/* Determine key length */
2049 					if (vidx == 0 /* dst-ip */ ||
2050 					    vidx == 1 /* src-ip */)
2051 						keylen = is_ipv6 ?
2052 						    sizeof(struct in6_addr):
2053 						    sizeof(in_addr_t);
2054 					else {
2055 						keylen = sizeof(key);
2056 						pkey = &key;
2057 					}
2058 					if (vidx == 0 /* dst-ip */)
2059 						pkey = is_ipv4 ? (void *)&dst_ip:
2060 						    (void *)&args->f_id.dst_ip6;
2061 					else if (vidx == 1 /* src-ip */)
2062 						pkey = is_ipv4 ? (void *)&src_ip:
2063 						    (void *)&args->f_id.src_ip6;
2064 					else if (vidx == 6 /* dscp */) {
2065 						if (is_ipv4)
2066 							key = ip->ip_tos >> 2;
2067 						else {
2068 							key = args->f_id.flow_id6;
2069 							key = (key & 0x0f) << 2 |
2070 							    (key & 0xf000) >> 14;
2071 						}
2072 						key &= 0x3f;
2073 					} else if (vidx == 2 /* dst-port */ ||
2074 					    vidx == 3 /* src-port */) {
2075 						/* Skip fragments */
2076 						if (offset != 0)
2077 							break;
2078 						/* Skip proto without ports */
2079 						if (proto != IPPROTO_TCP &&
2080 						    proto != IPPROTO_UDP &&
2081 						    proto != IPPROTO_UDPLITE &&
2082 						    proto != IPPROTO_SCTP)
2083 							break;
2084 						if (vidx == 2 /* dst-port */)
2085 							key = dst_port;
2086 						else
2087 							key = src_port;
2088 					}
2089 #ifndef USERSPACE
2090 					else if (vidx == 4 /* uid */ ||
2091 					    vidx == 5 /* jail */) {
2092 						check_uidgid(
2093 						    (ipfw_insn_u32 *)cmd,
2094 						    args, &ucred_lookup,
2095 #ifdef __FreeBSD__
2096 						    &ucred_cache);
2097 						if (vidx == 4 /* uid */)
2098 							key = ucred_cache->cr_uid;
2099 						else if (vidx == 5 /* jail */)
2100 							key = ucred_cache->cr_prison->pr_id;
2101 #else /* !__FreeBSD__ */
2102 						    (void *)&ucred_cache);
2103 						if (vidx == 4 /* uid */)
2104 							key = ucred_cache.uid;
2105 						else if (vidx == 5 /* jail */)
2106 							key = ucred_cache.xid;
2107 #endif /* !__FreeBSD__ */
2108 					}
2109 #endif /* !USERSPACE */
2110 					else
2111 						break;
2112 					match = ipfw_lookup_table(chain,
2113 					    cmd->arg1, keylen, pkey, &vidx);
2114 					if (!match)
2115 						break;
2116 					tablearg = vidx;
2117 					break;
2118 				}
2119 				/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2120 				/* FALLTHROUGH */
2121 			}
2122 			case O_IP_SRC_LOOKUP:
2123 			{
2124 				void *pkey;
2125 				uint32_t vidx;
2126 				uint16_t keylen;
2127 
2128 				if (is_ipv4) {
2129 					keylen = sizeof(in_addr_t);
2130 					if (cmd->opcode == O_IP_DST_LOOKUP)
2131 						pkey = &dst_ip;
2132 					else
2133 						pkey = &src_ip;
2134 				} else if (is_ipv6) {
2135 					keylen = sizeof(struct in6_addr);
2136 					if (cmd->opcode == O_IP_DST_LOOKUP)
2137 						pkey = &args->f_id.dst_ip6;
2138 					else
2139 						pkey = &args->f_id.src_ip6;
2140 				} else
2141 					break;
2142 				match = ipfw_lookup_table(chain, cmd->arg1,
2143 				    keylen, pkey, &vidx);
2144 				if (!match)
2145 					break;
2146 				if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2147 					match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2148 					    TARG_VAL(chain, vidx, tag);
2149 					if (!match)
2150 						break;
2151 				}
2152 				tablearg = vidx;
2153 				break;
2154 			}
2155 
2156 			case O_IP_FLOW_LOOKUP:
2157 				{
2158 					uint32_t v = 0;
2159 					match = ipfw_lookup_table(chain,
2160 					    cmd->arg1, 0, &args->f_id, &v);
2161 					if (!match)
2162 						break;
2163 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2164 						match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2165 						    TARG_VAL(chain, v, tag);
2166 					if (match)
2167 						tablearg = v;
2168 				}
2169 				break;
2170 			case O_IP_SRC_MASK:
2171 			case O_IP_DST_MASK:
2172 				if (is_ipv4) {
2173 				    uint32_t a =
2174 					(cmd->opcode == O_IP_DST_MASK) ?
2175 					    dst_ip.s_addr : src_ip.s_addr;
2176 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2177 				    int i = cmdlen-1;
2178 
2179 				    for (; !match && i>0; i-= 2, p+= 2)
2180 					match = (p[0] == (a & p[1]));
2181 				}
2182 				break;
2183 
2184 			case O_IP_SRC_ME:
2185 				if (is_ipv4) {
2186 					match = in_localip(src_ip);
2187 					break;
2188 				}
2189 #ifdef INET6
2190 				/* FALLTHROUGH */
2191 			case O_IP6_SRC_ME:
2192 				match = is_ipv6 &&
2193 				    ipfw_localip6(&args->f_id.src_ip6);
2194 #endif
2195 				break;
2196 
2197 			case O_IP_DST_SET:
2198 			case O_IP_SRC_SET:
2199 				if (is_ipv4) {
2200 					u_int32_t *d = (u_int32_t *)(cmd+1);
2201 					u_int32_t addr =
2202 					    cmd->opcode == O_IP_DST_SET ?
2203 						args->f_id.dst_ip :
2204 						args->f_id.src_ip;
2205 
2206 					    if (addr < d[0])
2207 						    break;
2208 					    addr -= d[0]; /* subtract base */
2209 					    match = (addr < cmd->arg1) &&
2210 						( d[ 1 + (addr>>5)] &
2211 						  (1<<(addr & 0x1f)) );
2212 				}
2213 				break;
2214 
2215 			case O_IP_DST:
2216 				match = is_ipv4 &&
2217 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2218 				    dst_ip.s_addr);
2219 				break;
2220 
2221 			case O_IP_DST_ME:
2222 				if (is_ipv4) {
2223 					match = in_localip(dst_ip);
2224 					break;
2225 				}
2226 #ifdef INET6
2227 				/* FALLTHROUGH */
2228 			case O_IP6_DST_ME:
2229 				match = is_ipv6 &&
2230 				    ipfw_localip6(&args->f_id.dst_ip6);
2231 #endif
2232 				break;
2233 
2234 			case O_IP_SRCPORT:
2235 			case O_IP_DSTPORT:
2236 				/*
2237 				 * offset == 0 && proto != 0 is enough
2238 				 * to guarantee that we have a
2239 				 * packet with port info.
2240 				 */
2241 				if ((proto == IPPROTO_UDP ||
2242 				    proto == IPPROTO_UDPLITE ||
2243 				    proto == IPPROTO_TCP ||
2244 				    proto == IPPROTO_SCTP) && offset == 0) {
2245 					u_int16_t x =
2246 					    (cmd->opcode == O_IP_SRCPORT) ?
2247 						src_port : dst_port ;
2248 					u_int16_t *p =
2249 					    ((ipfw_insn_u16 *)cmd)->ports;
2250 					int i;
2251 
2252 					for (i = cmdlen - 1; !match && i>0;
2253 					    i--, p += 2)
2254 						match = (x>=p[0] && x<=p[1]);
2255 				}
2256 				break;
2257 
2258 			case O_ICMPTYPE:
2259 				match = (offset == 0 && proto==IPPROTO_ICMP &&
2260 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2261 				break;
2262 
2263 #ifdef INET6
2264 			case O_ICMP6TYPE:
2265 				match = is_ipv6 && offset == 0 &&
2266 				    proto==IPPROTO_ICMPV6 &&
2267 				    icmp6type_match(
2268 					ICMP6(ulp)->icmp6_type,
2269 					(ipfw_insn_u32 *)cmd);
2270 				break;
2271 #endif /* INET6 */
2272 
2273 			case O_IPOPT:
2274 				match = (is_ipv4 &&
2275 				    ipopts_match(ip, cmd) );
2276 				break;
2277 
2278 			case O_IPVER:
2279 				match = ((is_ipv4 || is_ipv6) &&
2280 				    cmd->arg1 == ip->ip_v);
2281 				break;
2282 
2283 			case O_IPID:
2284 			case O_IPTTL:
2285 				if (!is_ipv4)
2286 					break;
2287 			case O_IPLEN:
2288 				{	/* only for IP packets */
2289 				    uint16_t x;
2290 				    uint16_t *p;
2291 				    int i;
2292 
2293 				    if (cmd->opcode == O_IPLEN)
2294 					x = iplen;
2295 				    else if (cmd->opcode == O_IPTTL)
2296 					x = ip->ip_ttl;
2297 				    else /* must be IPID */
2298 					x = ntohs(ip->ip_id);
2299 				    if (cmdlen == 1) {
2300 					match = (cmd->arg1 == x);
2301 					break;
2302 				    }
2303 				    /* otherwise we have ranges */
2304 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2305 				    i = cmdlen - 1;
2306 				    for (; !match && i>0; i--, p += 2)
2307 					match = (x >= p[0] && x <= p[1]);
2308 				}
2309 				break;
2310 
2311 			case O_IPPRECEDENCE:
2312 				match = (is_ipv4 &&
2313 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2314 				break;
2315 
2316 			case O_IPTOS:
2317 				match = (is_ipv4 &&
2318 				    flags_match(cmd, ip->ip_tos));
2319 				break;
2320 
2321 			case O_DSCP:
2322 			    {
2323 				uint32_t *p;
2324 				uint16_t x;
2325 
2326 				p = ((ipfw_insn_u32 *)cmd)->d;
2327 
2328 				if (is_ipv4)
2329 					x = ip->ip_tos >> 2;
2330 				else if (is_ipv6) {
2331 					uint8_t *v;
2332 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
2333 					x = (*v & 0x0F) << 2;
2334 					v++;
2335 					x |= *v >> 6;
2336 				} else
2337 					break;
2338 
2339 				/* DSCP bitmask is stored as low_u32 high_u32 */
2340 				if (x >= 32)
2341 					match = *(p + 1) & (1 << (x - 32));
2342 				else
2343 					match = *p & (1 << x);
2344 			    }
2345 				break;
2346 
2347 			case O_TCPDATALEN:
2348 				if (proto == IPPROTO_TCP && offset == 0) {
2349 				    struct tcphdr *tcp;
2350 				    uint16_t x;
2351 				    uint16_t *p;
2352 				    int i;
2353 #ifdef INET6
2354 				    if (is_ipv6) {
2355 					    struct ip6_hdr *ip6;
2356 
2357 					    ip6 = (struct ip6_hdr *)ip;
2358 					    if (ip6->ip6_plen == 0) {
2359 						    /*
2360 						     * Jumbo payload is not
2361 						     * supported by this
2362 						     * opcode.
2363 						     */
2364 						    break;
2365 					    }
2366 					    x = iplen - hlen;
2367 				    } else
2368 #endif /* INET6 */
2369 					    x = iplen - (ip->ip_hl << 2);
2370 				    tcp = TCP(ulp);
2371 				    x -= tcp->th_off << 2;
2372 				    if (cmdlen == 1) {
2373 					match = (cmd->arg1 == x);
2374 					break;
2375 				    }
2376 				    /* otherwise we have ranges */
2377 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2378 				    i = cmdlen - 1;
2379 				    for (; !match && i>0; i--, p += 2)
2380 					match = (x >= p[0] && x <= p[1]);
2381 				}
2382 				break;
2383 
2384 			case O_TCPFLAGS:
2385 				match = (proto == IPPROTO_TCP && offset == 0 &&
2386 				    flags_match(cmd, TCP(ulp)->th_flags));
2387 				break;
2388 
2389 			case O_TCPOPTS:
2390 				if (proto == IPPROTO_TCP && offset == 0 && ulp){
2391 					PULLUP_LEN_LOCKED(hlen, ulp,
2392 					    (TCP(ulp)->th_off << 2));
2393 					match = tcpopts_match(TCP(ulp), cmd);
2394 				}
2395 				break;
2396 
2397 			case O_TCPSEQ:
2398 				match = (proto == IPPROTO_TCP && offset == 0 &&
2399 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2400 					TCP(ulp)->th_seq);
2401 				break;
2402 
2403 			case O_TCPACK:
2404 				match = (proto == IPPROTO_TCP && offset == 0 &&
2405 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
2406 					TCP(ulp)->th_ack);
2407 				break;
2408 
2409 			case O_TCPMSS:
2410 				if (proto == IPPROTO_TCP &&
2411 				    (args->f_id._flags & TH_SYN) != 0 &&
2412 				    ulp != NULL) {
2413 					uint16_t mss, *p;
2414 					int i;
2415 
2416 					PULLUP_LEN_LOCKED(hlen, ulp,
2417 					    (TCP(ulp)->th_off << 2));
2418 					if ((tcpopts_parse(TCP(ulp), &mss) &
2419 					    IP_FW_TCPOPT_MSS) == 0)
2420 						break;
2421 					if (cmdlen == 1) {
2422 						match = (cmd->arg1 == mss);
2423 						break;
2424 					}
2425 					/* Otherwise we have ranges. */
2426 					p = ((ipfw_insn_u16 *)cmd)->ports;
2427 					i = cmdlen - 1;
2428 					for (; !match && i > 0; i--, p += 2)
2429 						match = (mss >= p[0] &&
2430 						    mss <= p[1]);
2431 				}
2432 				break;
2433 
2434 			case O_TCPWIN:
2435 				if (proto == IPPROTO_TCP && offset == 0) {
2436 				    uint16_t x;
2437 				    uint16_t *p;
2438 				    int i;
2439 
2440 				    x = ntohs(TCP(ulp)->th_win);
2441 				    if (cmdlen == 1) {
2442 					match = (cmd->arg1 == x);
2443 					break;
2444 				    }
2445 				    /* Otherwise we have ranges. */
2446 				    p = ((ipfw_insn_u16 *)cmd)->ports;
2447 				    i = cmdlen - 1;
2448 				    for (; !match && i > 0; i--, p += 2)
2449 					match = (x >= p[0] && x <= p[1]);
2450 				}
2451 				break;
2452 
2453 			case O_ESTAB:
2454 				/* reject packets which have SYN only */
2455 				/* XXX should i also check for TH_ACK ? */
2456 				match = (proto == IPPROTO_TCP && offset == 0 &&
2457 				    (TCP(ulp)->th_flags &
2458 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2459 				break;
2460 
2461 			case O_ALTQ: {
2462 				struct pf_mtag *at;
2463 				struct m_tag *mtag;
2464 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2465 
2466 				/*
2467 				 * ALTQ uses mbuf tags from another
2468 				 * packet filtering system - pf(4).
2469 				 * We allocate a tag in its format
2470 				 * and fill it in, pretending to be pf(4).
2471 				 */
2472 				match = 1;
2473 				at = pf_find_mtag(m);
2474 				if (at != NULL && at->qid != 0)
2475 					break;
2476 				mtag = m_tag_get(PACKET_TAG_PF,
2477 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2478 				if (mtag == NULL) {
2479 					/*
2480 					 * Let the packet fall back to the
2481 					 * default ALTQ.
2482 					 */
2483 					break;
2484 				}
2485 				m_tag_prepend(m, mtag);
2486 				at = (struct pf_mtag *)(mtag + 1);
2487 				at->qid = altq->qid;
2488 				at->hdr = ip;
2489 				break;
2490 			}
2491 
2492 			case O_LOG:
2493 				ipfw_log(chain, f, hlen, args,
2494 				    offset | ip6f_mf, tablearg, ip);
2495 				match = 1;
2496 				break;
2497 
2498 			case O_PROB:
2499 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2500 				break;
2501 
2502 			case O_VERREVPATH:
2503 				/* Outgoing packets automatically pass/match */
2504 				match = (args->flags & IPFW_ARGS_OUT ||
2505 				    (
2506 #ifdef INET6
2507 				    is_ipv6 ?
2508 					verify_path6(&(args->f_id.src_ip6),
2509 					    iif, args->f_id.fib) :
2510 #endif
2511 				    verify_path(src_ip, iif, args->f_id.fib)));
2512 				break;
2513 
2514 			case O_VERSRCREACH:
2515 				/* Outgoing packets automatically pass/match */
2516 				match = (hlen > 0 && ((oif != NULL) || (
2517 #ifdef INET6
2518 				    is_ipv6 ?
2519 				        verify_path6(&(args->f_id.src_ip6),
2520 				            NULL, args->f_id.fib) :
2521 #endif
2522 				    verify_path(src_ip, NULL, args->f_id.fib))));
2523 				break;
2524 
2525 			case O_ANTISPOOF:
2526 				/* Outgoing packets automatically pass/match */
2527 				if (oif == NULL && hlen > 0 &&
2528 				    (  (is_ipv4 && in_localaddr(src_ip))
2529 #ifdef INET6
2530 				    || (is_ipv6 &&
2531 				        in6_localaddr(&(args->f_id.src_ip6)))
2532 #endif
2533 				    ))
2534 					match =
2535 #ifdef INET6
2536 					    is_ipv6 ? verify_path6(
2537 					        &(args->f_id.src_ip6), iif,
2538 						args->f_id.fib) :
2539 #endif
2540 					    verify_path(src_ip, iif,
2541 					        args->f_id.fib);
2542 				else
2543 					match = 1;
2544 				break;
2545 
2546 			case O_IPSEC:
2547 				match = (m_tag_find(m,
2548 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2549 				/* otherwise no match */
2550 				break;
2551 
2552 #ifdef INET6
2553 			case O_IP6_SRC:
2554 				match = is_ipv6 &&
2555 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2556 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2557 				break;
2558 
2559 			case O_IP6_DST:
2560 				match = is_ipv6 &&
2561 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2562 				    &((ipfw_insn_ip6 *)cmd)->addr6);
2563 				break;
2564 			case O_IP6_SRC_MASK:
2565 			case O_IP6_DST_MASK:
2566 				if (is_ipv6) {
2567 					int i = cmdlen - 1;
2568 					struct in6_addr p;
2569 					struct in6_addr *d =
2570 					    &((ipfw_insn_ip6 *)cmd)->addr6;
2571 
2572 					for (; !match && i > 0; d += 2,
2573 					    i -= F_INSN_SIZE(struct in6_addr)
2574 					    * 2) {
2575 						p = (cmd->opcode ==
2576 						    O_IP6_SRC_MASK) ?
2577 						    args->f_id.src_ip6:
2578 						    args->f_id.dst_ip6;
2579 						APPLY_MASK(&p, &d[1]);
2580 						match =
2581 						    IN6_ARE_ADDR_EQUAL(&d[0],
2582 						    &p);
2583 					}
2584 				}
2585 				break;
2586 
2587 			case O_FLOW6ID:
2588 				match = is_ipv6 &&
2589 				    flow6id_match(args->f_id.flow_id6,
2590 				    (ipfw_insn_u32 *) cmd);
2591 				break;
2592 
2593 			case O_EXT_HDR:
2594 				match = is_ipv6 &&
2595 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
2596 				break;
2597 
2598 			case O_IP6:
2599 				match = is_ipv6;
2600 				break;
2601 #endif
2602 
2603 			case O_IP4:
2604 				match = is_ipv4;
2605 				break;
2606 
2607 			case O_TAG: {
2608 				struct m_tag *mtag;
2609 				uint32_t tag = TARG(cmd->arg1, tag);
2610 
2611 				/* Packet is already tagged with this tag? */
2612 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2613 
2614 				/* We have `untag' action when F_NOT flag is
2615 				 * present. And we must remove this mtag from
2616 				 * mbuf and reset `match' to zero (`match' will
2617 				 * be inversed later).
2618 				 * Otherwise we should allocate new mtag and
2619 				 * push it into mbuf.
2620 				 */
2621 				if (cmd->len & F_NOT) { /* `untag' action */
2622 					if (mtag != NULL)
2623 						m_tag_delete(m, mtag);
2624 					match = 0;
2625 				} else {
2626 					if (mtag == NULL) {
2627 						mtag = m_tag_alloc( MTAG_IPFW,
2628 						    tag, 0, M_NOWAIT);
2629 						if (mtag != NULL)
2630 							m_tag_prepend(m, mtag);
2631 					}
2632 					match = 1;
2633 				}
2634 				break;
2635 			}
2636 
2637 			case O_FIB: /* try match the specified fib */
2638 				if (args->f_id.fib == cmd->arg1)
2639 					match = 1;
2640 				break;
2641 
2642 			case O_SOCKARG:	{
2643 #ifndef USERSPACE	/* not supported in userspace */
2644 				struct inpcb *inp = args->inp;
2645 				struct inpcbinfo *pi;
2646 				bool inp_locked = false;
2647 
2648 				if (proto == IPPROTO_TCP)
2649 					pi = &V_tcbinfo;
2650 				else if (proto == IPPROTO_UDP)
2651 					pi = &V_udbinfo;
2652 				else if (proto == IPPROTO_UDPLITE)
2653 					pi = &V_ulitecbinfo;
2654 				else
2655 					break;
2656 
2657 				/*
2658 				 * XXXRW: so_user_cookie should almost
2659 				 * certainly be inp_user_cookie?
2660 				 */
2661 
2662 				/*
2663 				 * For incoming packet lookup the inpcb
2664 				 * using the src/dest ip/port tuple.
2665 				 */
2666 				if (is_ipv4 && inp == NULL) {
2667 					inp = in_pcblookup(pi,
2668 					    src_ip, htons(src_port),
2669 					    dst_ip, htons(dst_port),
2670 					    INPLOOKUP_RLOCKPCB, NULL);
2671 					inp_locked = true;
2672 				}
2673 #ifdef INET6
2674 				if (is_ipv6 && inp == NULL) {
2675 					inp = in6_pcblookup(pi,
2676 					    &args->f_id.src_ip6,
2677 					    htons(src_port),
2678 					    &args->f_id.dst_ip6,
2679 					    htons(dst_port),
2680 					    INPLOOKUP_RLOCKPCB, NULL);
2681 					inp_locked = true;
2682 				}
2683 #endif /* INET6 */
2684 				if (inp != NULL) {
2685 					if (inp->inp_socket) {
2686 						tablearg =
2687 						    inp->inp_socket->so_user_cookie;
2688 						if (tablearg)
2689 							match = 1;
2690 					}
2691 					if (inp_locked)
2692 						INP_RUNLOCK(inp);
2693 				}
2694 #endif /* !USERSPACE */
2695 				break;
2696 			}
2697 
2698 			case O_TAGGED: {
2699 				struct m_tag *mtag;
2700 				uint32_t tag = TARG(cmd->arg1, tag);
2701 
2702 				if (cmdlen == 1) {
2703 					match = m_tag_locate(m, MTAG_IPFW,
2704 					    tag, NULL) != NULL;
2705 					break;
2706 				}
2707 
2708 				/* we have ranges */
2709 				for (mtag = m_tag_first(m);
2710 				    mtag != NULL && !match;
2711 				    mtag = m_tag_next(m, mtag)) {
2712 					uint16_t *p;
2713 					int i;
2714 
2715 					if (mtag->m_tag_cookie != MTAG_IPFW)
2716 						continue;
2717 
2718 					p = ((ipfw_insn_u16 *)cmd)->ports;
2719 					i = cmdlen - 1;
2720 					for(; !match && i > 0; i--, p += 2)
2721 						match =
2722 						    mtag->m_tag_id >= p[0] &&
2723 						    mtag->m_tag_id <= p[1];
2724 				}
2725 				break;
2726 			}
2727 
2728 			/*
2729 			 * The second set of opcodes represents 'actions',
2730 			 * i.e. the terminal part of a rule once the packet
2731 			 * matches all previous patterns.
2732 			 * Typically there is only one action for each rule,
2733 			 * and the opcode is stored at the end of the rule
2734 			 * (but there are exceptions -- see below).
2735 			 *
2736 			 * In general, here we set retval and terminate the
2737 			 * outer loop (would be a 'break 3' in some language,
2738 			 * but we need to set l=0, done=1)
2739 			 *
2740 			 * Exceptions:
2741 			 * O_COUNT and O_SKIPTO actions:
2742 			 *   instead of terminating, we jump to the next rule
2743 			 *   (setting l=0), or to the SKIPTO target (setting
2744 			 *   f/f_len, cmd and l as needed), respectively.
2745 			 *
2746 			 * O_TAG, O_LOG and O_ALTQ action parameters:
2747 			 *   perform some action and set match = 1;
2748 			 *
2749 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
2750 			 *   not real 'actions', and are stored right
2751 			 *   before the 'action' part of the rule (one
2752 			 *   exception is O_SKIP_ACTION which could be
2753 			 *   between these opcodes and 'action' one).
2754 			 *   These opcodes try to install an entry in the
2755 			 *   state tables; if successful, we continue with
2756 			 *   the next opcode (match=1; break;), otherwise
2757 			 *   the packet must be dropped (set retval,
2758 			 *   break loops with l=0, done=1)
2759 			 *
2760 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2761 			 *   cause a lookup of the state table, and a jump
2762 			 *   to the 'action' part of the parent rule
2763 			 *   if an entry is found, or
2764 			 *   (CHECK_STATE only) a jump to the next rule if
2765 			 *   the entry is not found.
2766 			 *   The result of the lookup is cached so that
2767 			 *   further instances of these opcodes become NOPs.
2768 			 *   The jump to the next rule is done by setting
2769 			 *   l=0, cmdlen=0.
2770 			 *
2771 			 * O_SKIP_ACTION: this opcode is not a real 'action'
2772 			 *  either, and is stored right before the 'action'
2773 			 *  part of the rule, right after the O_KEEP_STATE
2774 			 *  opcode. It causes match failure so the real
2775 			 *  'action' could be executed only if the rule
2776 			 *  is checked via dynamic rule from the state
2777 			 *  table, as in such case execution starts
2778 			 *  from the true 'action' opcode directly.
2779 			 *
2780 			 */
2781 			case O_LIMIT:
2782 			case O_KEEP_STATE:
2783 				if (ipfw_dyn_install_state(chain, f,
2784 				    (ipfw_insn_limit *)cmd, args, ulp,
2785 				    pktlen, &dyn_info, tablearg)) {
2786 					/* error or limit violation */
2787 					retval = IP_FW_DENY;
2788 					l = 0;	/* exit inner loop */
2789 					done = 1; /* exit outer loop */
2790 				}
2791 				match = 1;
2792 				break;
2793 
2794 			case O_PROBE_STATE:
2795 			case O_CHECK_STATE:
2796 				/*
2797 				 * dynamic rules are checked at the first
2798 				 * keep-state or check-state occurrence,
2799 				 * with the result being stored in dyn_info.
2800 				 * The compiler introduces a PROBE_STATE
2801 				 * instruction for us when we have a
2802 				 * KEEP_STATE (because PROBE_STATE needs
2803 				 * to be run first).
2804 				 */
2805 				if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2806 				    (q = ipfw_dyn_lookup_state(args, ulp,
2807 				    pktlen, cmd, &dyn_info)) != NULL) {
2808 					/*
2809 					 * Found dynamic entry, jump to the
2810 					 * 'action' part of the parent rule
2811 					 * by setting f, cmd, l and clearing
2812 					 * cmdlen.
2813 					 */
2814 					f = q;
2815 					f_pos = dyn_info.f_pos;
2816 					cmd = ACTION_PTR(f);
2817 					l = f->cmd_len - f->act_ofs;
2818 					cmdlen = 0;
2819 					match = 1;
2820 					break;
2821 				}
2822 				/*
2823 				 * Dynamic entry not found. If CHECK_STATE,
2824 				 * skip to next rule, if PROBE_STATE just
2825 				 * ignore and continue with next opcode.
2826 				 */
2827 				if (cmd->opcode == O_CHECK_STATE)
2828 					l = 0;	/* exit inner loop */
2829 				match = 1;
2830 				break;
2831 
2832 			case O_SKIP_ACTION:
2833 				match = 0;	/* skip to the next rule */
2834 				l = 0;		/* exit inner loop */
2835 				break;
2836 
2837 			case O_ACCEPT:
2838 				retval = 0;	/* accept */
2839 				l = 0;		/* exit inner loop */
2840 				done = 1;	/* exit outer loop */
2841 				break;
2842 
2843 			case O_PIPE:
2844 			case O_QUEUE:
2845 				set_match(args, f_pos, chain);
2846 				args->rule.info = TARG(cmd->arg1, pipe);
2847 				if (cmd->opcode == O_PIPE)
2848 					args->rule.info |= IPFW_IS_PIPE;
2849 				if (V_fw_one_pass)
2850 					args->rule.info |= IPFW_ONEPASS;
2851 				retval = IP_FW_DUMMYNET;
2852 				l = 0;          /* exit inner loop */
2853 				done = 1;       /* exit outer loop */
2854 				break;
2855 
2856 			case O_DIVERT:
2857 			case O_TEE:
2858 				if (args->flags & IPFW_ARGS_ETHER)
2859 					break;	/* not on layer 2 */
2860 				/* otherwise this is terminal */
2861 				l = 0;		/* exit inner loop */
2862 				done = 1;	/* exit outer loop */
2863 				retval = (cmd->opcode == O_DIVERT) ?
2864 					IP_FW_DIVERT : IP_FW_TEE;
2865 				set_match(args, f_pos, chain);
2866 				args->rule.info = TARG(cmd->arg1, divert);
2867 				break;
2868 
2869 			case O_COUNT:
2870 				IPFW_INC_RULE_COUNTER(f, pktlen);
2871 				l = 0;		/* exit inner loop */
2872 				break;
2873 
2874 			case O_SKIPTO:
2875 			    IPFW_INC_RULE_COUNTER(f, pktlen);
2876 			    f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2877 			    /*
2878 			     * Skip disabled rules, and re-enter
2879 			     * the inner loop with the correct
2880 			     * f_pos, f, l and cmd.
2881 			     * Also clear cmdlen and skip_or
2882 			     */
2883 			    for (; f_pos < chain->n_rules - 1 &&
2884 				    (V_set_disable &
2885 				     (1 << chain->map[f_pos]->set));
2886 				    f_pos++)
2887 				;
2888 			    /* Re-enter the inner loop at the skipto rule. */
2889 			    f = chain->map[f_pos];
2890 			    l = f->cmd_len;
2891 			    cmd = f->cmd;
2892 			    match = 1;
2893 			    cmdlen = 0;
2894 			    skip_or = 0;
2895 			    continue;
2896 			    break;	/* not reached */
2897 
2898 			case O_CALLRETURN: {
2899 				/*
2900 				 * Implementation of `subroutine' call/return,
2901 				 * in the stack carried in an mbuf tag. This
2902 				 * is different from `skipto' in that any call
2903 				 * address is possible (`skipto' must prevent
2904 				 * backward jumps to avoid endless loops).
2905 				 * We have `return' action when F_NOT flag is
2906 				 * present. The `m_tag_id' field is used as
2907 				 * stack pointer.
2908 				 */
2909 				struct m_tag *mtag;
2910 				uint16_t jmpto, *stack;
2911 
2912 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2913 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2914 				/*
2915 				 * Hand-rolled version of m_tag_locate() with
2916 				 * wildcard `type'.
2917 				 * If not already tagged, allocate new tag.
2918 				 */
2919 				mtag = m_tag_first(m);
2920 				while (mtag != NULL) {
2921 					if (mtag->m_tag_cookie ==
2922 					    MTAG_IPFW_CALL)
2923 						break;
2924 					mtag = m_tag_next(m, mtag);
2925 				}
2926 				if (mtag == NULL && IS_CALL) {
2927 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2928 					    IPFW_CALLSTACK_SIZE *
2929 					    sizeof(uint16_t), M_NOWAIT);
2930 					if (mtag != NULL)
2931 						m_tag_prepend(m, mtag);
2932 				}
2933 
2934 				/*
2935 				 * On error both `call' and `return' just
2936 				 * continue with next rule.
2937 				 */
2938 				if (IS_RETURN && (mtag == NULL ||
2939 				    mtag->m_tag_id == 0)) {
2940 					l = 0;		/* exit inner loop */
2941 					break;
2942 				}
2943 				if (IS_CALL && (mtag == NULL ||
2944 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2945 					printf("ipfw: call stack error, "
2946 					    "go to next rule\n");
2947 					l = 0;		/* exit inner loop */
2948 					break;
2949 				}
2950 
2951 				IPFW_INC_RULE_COUNTER(f, pktlen);
2952 				stack = (uint16_t *)(mtag + 1);
2953 
2954 				/*
2955 				 * The `call' action may use cached f_pos
2956 				 * (in f->next_rule), whose version is written
2957 				 * in f->next_rule.
2958 				 * The `return' action, however, doesn't have
2959 				 * fixed jump address in cmd->arg1 and can't use
2960 				 * cache.
2961 				 */
2962 				if (IS_CALL) {
2963 					stack[mtag->m_tag_id] = f->rulenum;
2964 					mtag->m_tag_id++;
2965 			    		f_pos = JUMP(chain, f, cmd->arg1,
2966 					    tablearg, 1);
2967 				} else {	/* `return' action */
2968 					mtag->m_tag_id--;
2969 					jmpto = stack[mtag->m_tag_id] + 1;
2970 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2971 				}
2972 
2973 				/*
2974 				 * Skip disabled rules, and re-enter
2975 				 * the inner loop with the correct
2976 				 * f_pos, f, l and cmd.
2977 				 * Also clear cmdlen and skip_or
2978 				 */
2979 				for (; f_pos < chain->n_rules - 1 &&
2980 				    (V_set_disable &
2981 				    (1 << chain->map[f_pos]->set)); f_pos++)
2982 					;
2983 				/* Re-enter the inner loop at the dest rule. */
2984 				f = chain->map[f_pos];
2985 				l = f->cmd_len;
2986 				cmd = f->cmd;
2987 				cmdlen = 0;
2988 				skip_or = 0;
2989 				continue;
2990 				break;	/* NOTREACHED */
2991 			}
2992 #undef IS_CALL
2993 #undef IS_RETURN
2994 
2995 			case O_REJECT:
2996 				/*
2997 				 * Drop the packet and send a reject notice
2998 				 * if the packet is not ICMP (or is an ICMP
2999 				 * query), and it is not multicast/broadcast.
3000 				 */
3001 				if (hlen > 0 && is_ipv4 && offset == 0 &&
3002 				    (proto != IPPROTO_ICMP ||
3003 				     is_icmp_query(ICMP(ulp))) &&
3004 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
3005 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3006 					send_reject(args, cmd->arg1, iplen, ip);
3007 					m = args->m;
3008 				}
3009 				/* FALLTHROUGH */
3010 #ifdef INET6
3011 			case O_UNREACH6:
3012 				if (hlen > 0 && is_ipv6 &&
3013 				    ((offset & IP6F_OFF_MASK) == 0) &&
3014 				    (proto != IPPROTO_ICMPV6 ||
3015 				     (is_icmp6_query(icmp6_type) == 1)) &&
3016 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
3017 				    !IN6_IS_ADDR_MULTICAST(
3018 					&args->f_id.dst_ip6)) {
3019 					send_reject6(args,
3020 					    cmd->opcode == O_REJECT ?
3021 					    map_icmp_unreach(cmd->arg1):
3022 					    cmd->arg1, hlen,
3023 					    (struct ip6_hdr *)ip);
3024 					m = args->m;
3025 				}
3026 				/* FALLTHROUGH */
3027 #endif
3028 			case O_DENY:
3029 				retval = IP_FW_DENY;
3030 				l = 0;		/* exit inner loop */
3031 				done = 1;	/* exit outer loop */
3032 				break;
3033 
3034 			case O_FORWARD_IP:
3035 				if (args->flags & IPFW_ARGS_ETHER)
3036 					break;	/* not valid on layer2 pkts */
3037 				if (q != f ||
3038 				    dyn_info.direction == MATCH_FORWARD) {
3039 				    struct sockaddr_in *sa;
3040 
3041 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
3042 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
3043 #ifdef INET6
3044 					/*
3045 					 * We use O_FORWARD_IP opcode for
3046 					 * fwd rule with tablearg, but tables
3047 					 * now support IPv6 addresses. And
3048 					 * when we are inspecting IPv6 packet,
3049 					 * we can use nh6 field from
3050 					 * table_value as next_hop6 address.
3051 					 */
3052 					if (is_ipv6) {
3053 						struct ip_fw_nh6 *nh6;
3054 
3055 						args->flags |= IPFW_ARGS_NH6;
3056 						nh6 = &args->hopstore6;
3057 						nh6->sin6_addr = TARG_VAL(
3058 						    chain, tablearg, nh6);
3059 						nh6->sin6_port = sa->sin_port;
3060 						nh6->sin6_scope_id = TARG_VAL(
3061 						    chain, tablearg, zoneid);
3062 					} else
3063 #endif
3064 					{
3065 						args->flags |= IPFW_ARGS_NH4;
3066 						args->hopstore.sin_port =
3067 						    sa->sin_port;
3068 						sa = &args->hopstore;
3069 						sa->sin_family = AF_INET;
3070 						sa->sin_len = sizeof(*sa);
3071 						sa->sin_addr.s_addr = htonl(
3072 						    TARG_VAL(chain, tablearg,
3073 						    nh4));
3074 					}
3075 				    } else {
3076 					    args->flags |= IPFW_ARGS_NH4PTR;
3077 					    args->next_hop = sa;
3078 				    }
3079 				}
3080 				retval = IP_FW_PASS;
3081 				l = 0;          /* exit inner loop */
3082 				done = 1;       /* exit outer loop */
3083 				break;
3084 
3085 #ifdef INET6
3086 			case O_FORWARD_IP6:
3087 				if (args->flags & IPFW_ARGS_ETHER)
3088 					break;	/* not valid on layer2 pkts */
3089 				if (q != f ||
3090 				    dyn_info.direction == MATCH_FORWARD) {
3091 					struct sockaddr_in6 *sin6;
3092 
3093 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3094 					args->flags |= IPFW_ARGS_NH6PTR;
3095 					args->next_hop6 = sin6;
3096 				}
3097 				retval = IP_FW_PASS;
3098 				l = 0;		/* exit inner loop */
3099 				done = 1;	/* exit outer loop */
3100 				break;
3101 #endif
3102 
3103 			case O_NETGRAPH:
3104 			case O_NGTEE:
3105 				set_match(args, f_pos, chain);
3106 				args->rule.info = TARG(cmd->arg1, netgraph);
3107 				if (V_fw_one_pass)
3108 					args->rule.info |= IPFW_ONEPASS;
3109 				retval = (cmd->opcode == O_NETGRAPH) ?
3110 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
3111 				l = 0;          /* exit inner loop */
3112 				done = 1;       /* exit outer loop */
3113 				break;
3114 
3115 			case O_SETFIB: {
3116 				uint32_t fib;
3117 
3118 				IPFW_INC_RULE_COUNTER(f, pktlen);
3119 				fib = TARG(cmd->arg1, fib) & 0x7FFF;
3120 				if (fib >= rt_numfibs)
3121 					fib = 0;
3122 				M_SETFIB(m, fib);
3123 				args->f_id.fib = fib; /* XXX */
3124 				l = 0;		/* exit inner loop */
3125 				break;
3126 		        }
3127 
3128 			case O_SETDSCP: {
3129 				uint16_t code;
3130 
3131 				code = TARG(cmd->arg1, dscp) & 0x3F;
3132 				l = 0;		/* exit inner loop */
3133 				if (is_ipv4) {
3134 					uint16_t old;
3135 
3136 					old = *(uint16_t *)ip;
3137 					ip->ip_tos = (code << 2) |
3138 					    (ip->ip_tos & 0x03);
3139 					ip->ip_sum = cksum_adjust(ip->ip_sum,
3140 					    old, *(uint16_t *)ip);
3141 				} else if (is_ipv6) {
3142 					uint8_t *v;
3143 
3144 					v = &((struct ip6_hdr *)ip)->ip6_vfc;
3145 					*v = (*v & 0xF0) | (code >> 2);
3146 					v++;
3147 					*v = (*v & 0x3F) | ((code & 0x03) << 6);
3148 				} else
3149 					break;
3150 
3151 				IPFW_INC_RULE_COUNTER(f, pktlen);
3152 				break;
3153 			}
3154 
3155 			case O_NAT:
3156 				l = 0;          /* exit inner loop */
3157 				done = 1;       /* exit outer loop */
3158 				/*
3159 				 * Ensure that we do not invoke NAT handler for
3160 				 * non IPv4 packets. Libalias expects only IPv4.
3161 				 */
3162 				if (!is_ipv4 || !IPFW_NAT_LOADED) {
3163 				    retval = IP_FW_DENY;
3164 				    break;
3165 				}
3166 
3167 				struct cfg_nat *t;
3168 				int nat_id;
3169 
3170 				args->rule.info = 0;
3171 				set_match(args, f_pos, chain);
3172 				/* Check if this is 'global' nat rule */
3173 				if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3174 					retval = ipfw_nat_ptr(args, NULL, m);
3175 					break;
3176 				}
3177 				t = ((ipfw_insn_nat *)cmd)->nat;
3178 				if (t == NULL) {
3179 					nat_id = TARG(cmd->arg1, nat);
3180 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3181 
3182 					if (t == NULL) {
3183 					    retval = IP_FW_DENY;
3184 					    break;
3185 					}
3186 					if (cmd->arg1 != IP_FW_TARG)
3187 					    ((ipfw_insn_nat *)cmd)->nat = t;
3188 				}
3189 				retval = ipfw_nat_ptr(args, t, m);
3190 				break;
3191 
3192 			case O_REASS: {
3193 				int ip_off;
3194 
3195 				l = 0;	/* in any case exit inner loop */
3196 				if (is_ipv6) /* IPv6 is not supported yet */
3197 					break;
3198 				IPFW_INC_RULE_COUNTER(f, pktlen);
3199 				ip_off = ntohs(ip->ip_off);
3200 
3201 				/* if not fragmented, go to next rule */
3202 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3203 				    break;
3204 
3205 				args->m = m = ip_reass(m);
3206 
3207 				/*
3208 				 * do IP header checksum fixup.
3209 				 */
3210 				if (m == NULL) { /* fragment got swallowed */
3211 				    retval = IP_FW_DENY;
3212 				} else { /* good, packet complete */
3213 				    int hlen;
3214 
3215 				    ip = mtod(m, struct ip *);
3216 				    hlen = ip->ip_hl << 2;
3217 				    ip->ip_sum = 0;
3218 				    if (hlen == sizeof(struct ip))
3219 					ip->ip_sum = in_cksum_hdr(ip);
3220 				    else
3221 					ip->ip_sum = in_cksum(m, hlen);
3222 				    retval = IP_FW_REASS;
3223 				    args->rule.info = 0;
3224 				    set_match(args, f_pos, chain);
3225 				}
3226 				done = 1;	/* exit outer loop */
3227 				break;
3228 			}
3229 			case O_EXTERNAL_ACTION:
3230 				l = 0; /* in any case exit inner loop */
3231 				retval = ipfw_run_eaction(chain, args,
3232 				    cmd, &done);
3233 				/*
3234 				 * If both @retval and @done are zero,
3235 				 * consider this as rule matching and
3236 				 * update counters.
3237 				 */
3238 				if (retval == 0 && done == 0) {
3239 					IPFW_INC_RULE_COUNTER(f, pktlen);
3240 					/*
3241 					 * Reset the result of the last
3242 					 * dynamic state lookup.
3243 					 * External action can change
3244 					 * @args content, and it may be
3245 					 * used for new state lookup later.
3246 					 */
3247 					DYN_INFO_INIT(&dyn_info);
3248 				}
3249 				break;
3250 
3251 			default:
3252 				panic("-- unknown opcode %d\n", cmd->opcode);
3253 			} /* end of switch() on opcodes */
3254 			/*
3255 			 * if we get here with l=0, then match is irrelevant.
3256 			 */
3257 
3258 			if (cmd->len & F_NOT)
3259 				match = !match;
3260 
3261 			if (match) {
3262 				if (cmd->len & F_OR)
3263 					skip_or = 1;
3264 			} else {
3265 				if (!(cmd->len & F_OR)) /* not an OR block, */
3266 					break;		/* try next rule    */
3267 			}
3268 
3269 		}	/* end of inner loop, scan opcodes */
3270 #undef PULLUP_LEN
3271 #undef PULLUP_LEN_LOCKED
3272 
3273 		if (done)
3274 			break;
3275 
3276 /* next_rule:; */	/* try next rule		*/
3277 
3278 	}		/* end of outer for, scan rules */
3279 
3280 	if (done) {
3281 		struct ip_fw *rule = chain->map[f_pos];
3282 		/* Update statistics */
3283 		IPFW_INC_RULE_COUNTER(rule, pktlen);
3284 		IPFW_PROBE(rule__matched, retval,
3285 		    is_ipv4 ? AF_INET : AF_INET6,
3286 		    is_ipv4 ? (uintptr_t)&src_ip :
3287 		        (uintptr_t)&args->f_id.src_ip6,
3288 		    is_ipv4 ? (uintptr_t)&dst_ip :
3289 		        (uintptr_t)&args->f_id.dst_ip6,
3290 		    args, rule);
3291 	} else {
3292 		retval = IP_FW_DENY;
3293 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3294 	}
3295 	IPFW_PF_RUNLOCK(chain);
3296 #ifdef __FreeBSD__
3297 	if (ucred_cache != NULL)
3298 		crfree(ucred_cache);
3299 #endif
3300 	return (retval);
3301 
3302 pullup_failed:
3303 	if (V_fw_verbose)
3304 		printf("ipfw: pullup failed\n");
3305 	return (IP_FW_DENY);
3306 }
3307 
3308 /*
3309  * Set maximum number of tables that can be used in given VNET ipfw instance.
3310  */
3311 #ifdef SYSCTL_NODE
3312 static int
3313 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3314 {
3315 	int error;
3316 	unsigned int ntables;
3317 
3318 	ntables = V_fw_tables_max;
3319 
3320 	error = sysctl_handle_int(oidp, &ntables, 0, req);
3321 	/* Read operation or some error */
3322 	if ((error != 0) || (req->newptr == NULL))
3323 		return (error);
3324 
3325 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
3326 }
3327 
3328 /*
3329  * Switches table namespace between global and per-set.
3330  */
3331 static int
3332 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3333 {
3334 	int error;
3335 	unsigned int sets;
3336 
3337 	sets = V_fw_tables_sets;
3338 
3339 	error = sysctl_handle_int(oidp, &sets, 0, req);
3340 	/* Read operation or some error */
3341 	if ((error != 0) || (req->newptr == NULL))
3342 		return (error);
3343 
3344 	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3345 }
3346 #endif
3347 
3348 /*
3349  * Module and VNET glue
3350  */
3351 
3352 /*
3353  * Stuff that must be initialised only on boot or module load
3354  */
3355 static int
3356 ipfw_init(void)
3357 {
3358 	int error = 0;
3359 
3360 	/*
3361  	 * Only print out this stuff the first time around,
3362 	 * when called from the sysinit code.
3363 	 */
3364 	printf("ipfw2 "
3365 #ifdef INET6
3366 		"(+ipv6) "
3367 #endif
3368 		"initialized, divert %s, nat %s, "
3369 		"default to %s, logging ",
3370 #ifdef IPDIVERT
3371 		"enabled",
3372 #else
3373 		"loadable",
3374 #endif
3375 #ifdef IPFIREWALL_NAT
3376 		"enabled",
3377 #else
3378 		"loadable",
3379 #endif
3380 		default_to_accept ? "accept" : "deny");
3381 
3382 	/*
3383 	 * Note: V_xxx variables can be accessed here but the vnet specific
3384 	 * initializer may not have been called yet for the VIMAGE case.
3385 	 * Tuneables will have been processed. We will print out values for
3386 	 * the default vnet.
3387 	 * XXX This should all be rationalized AFTER 8.0
3388 	 */
3389 	if (V_fw_verbose == 0)
3390 		printf("disabled\n");
3391 	else if (V_verbose_limit == 0)
3392 		printf("unlimited\n");
3393 	else
3394 		printf("limited to %d packets/entry by default\n",
3395 		    V_verbose_limit);
3396 
3397 	/* Check user-supplied table count for validness */
3398 	if (default_fw_tables > IPFW_TABLES_MAX)
3399 	  default_fw_tables = IPFW_TABLES_MAX;
3400 
3401 	ipfw_init_sopt_handler();
3402 	ipfw_init_obj_rewriter();
3403 	ipfw_iface_init();
3404 	return (error);
3405 }
3406 
3407 /*
3408  * Called for the removal of the last instance only on module unload.
3409  */
3410 static void
3411 ipfw_destroy(void)
3412 {
3413 
3414 	ipfw_iface_destroy();
3415 	ipfw_destroy_sopt_handler();
3416 	ipfw_destroy_obj_rewriter();
3417 	printf("IP firewall unloaded\n");
3418 }
3419 
3420 /*
3421  * Stuff that must be initialized for every instance
3422  * (including the first of course).
3423  */
3424 static int
3425 vnet_ipfw_init(const void *unused)
3426 {
3427 	int error, first;
3428 	struct ip_fw *rule = NULL;
3429 	struct ip_fw_chain *chain;
3430 
3431 	chain = &V_layer3_chain;
3432 
3433 	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3434 
3435 	/* First set up some values that are compile time options */
3436 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
3437 	V_fw_deny_unknown_exthdrs = 1;
3438 #ifdef IPFIREWALL_VERBOSE
3439 	V_fw_verbose = 1;
3440 #endif
3441 #ifdef IPFIREWALL_VERBOSE_LIMIT
3442 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3443 #endif
3444 #ifdef IPFIREWALL_NAT
3445 	LIST_INIT(&chain->nat);
3446 #endif
3447 
3448 	/* Init shared services hash table */
3449 	ipfw_init_srv(chain);
3450 
3451 	ipfw_init_counters();
3452 	/* Set initial number of tables */
3453 	V_fw_tables_max = default_fw_tables;
3454 	error = ipfw_init_tables(chain, first);
3455 	if (error) {
3456 		printf("ipfw2: setting up tables failed\n");
3457 		free(chain->map, M_IPFW);
3458 		free(rule, M_IPFW);
3459 		return (ENOSPC);
3460 	}
3461 
3462 	IPFW_LOCK_INIT(chain);
3463 
3464 	/* fill and insert the default rule */
3465 	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3466 	rule->flags |= IPFW_RULE_NOOPT;
3467 	rule->cmd_len = 1;
3468 	rule->cmd[0].len = 1;
3469 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3470 	chain->default_rule = rule;
3471 	ipfw_add_protected_rule(chain, rule, 0);
3472 
3473 	ipfw_dyn_init(chain);
3474 	ipfw_eaction_init(chain, first);
3475 #ifdef LINEAR_SKIPTO
3476 	ipfw_init_skipto_cache(chain);
3477 #endif
3478 	ipfw_bpf_init(first);
3479 
3480 	/* First set up some values that are compile time options */
3481 	V_ipfw_vnet_ready = 1;		/* Open for business */
3482 
3483 	/*
3484 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3485 	 * Even if the latter two fail we still keep the module alive
3486 	 * because the sockopt and layer2 paths are still useful.
3487 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3488 	 * so we can ignore the exact return value and just set a flag.
3489 	 *
3490 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3491 	 * changes in the underlying (per-vnet) variables trigger
3492 	 * immediate hook()/unhook() calls.
3493 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
3494 	 * is checked on each packet because there are no pfil hooks.
3495 	 */
3496 	V_ip_fw_ctl_ptr = ipfw_ctl3;
3497 	error = ipfw_attach_hooks();
3498 	return (error);
3499 }
3500 
3501 /*
3502  * Called for the removal of each instance.
3503  */
3504 static int
3505 vnet_ipfw_uninit(const void *unused)
3506 {
3507 	struct ip_fw *reap;
3508 	struct ip_fw_chain *chain = &V_layer3_chain;
3509 	int i, last;
3510 
3511 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3512 	/*
3513 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
3514 	 * Then grab, release and grab again the WLOCK so we make
3515 	 * sure the update is propagated and nobody will be in.
3516 	 */
3517 	ipfw_detach_hooks();
3518 	V_ip_fw_ctl_ptr = NULL;
3519 
3520 	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3521 
3522 	IPFW_UH_WLOCK(chain);
3523 	IPFW_UH_WUNLOCK(chain);
3524 
3525 	ipfw_dyn_uninit(0);	/* run the callout_drain */
3526 
3527 	IPFW_UH_WLOCK(chain);
3528 
3529 	reap = NULL;
3530 	IPFW_WLOCK(chain);
3531 	for (i = 0; i < chain->n_rules; i++)
3532 		ipfw_reap_add(chain, &reap, chain->map[i]);
3533 	free(chain->map, M_IPFW);
3534 #ifdef LINEAR_SKIPTO
3535 	ipfw_destroy_skipto_cache(chain);
3536 #endif
3537 	IPFW_WUNLOCK(chain);
3538 	IPFW_UH_WUNLOCK(chain);
3539 	ipfw_destroy_tables(chain, last);
3540 	ipfw_eaction_uninit(chain, last);
3541 	if (reap != NULL)
3542 		ipfw_reap_rules(reap);
3543 	vnet_ipfw_iface_destroy(chain);
3544 	ipfw_destroy_srv(chain);
3545 	IPFW_LOCK_DESTROY(chain);
3546 	ipfw_dyn_uninit(1);	/* free the remaining parts */
3547 	ipfw_destroy_counters();
3548 	ipfw_bpf_uninit(last);
3549 	return (0);
3550 }
3551 
3552 /*
3553  * Module event handler.
3554  * In general we have the choice of handling most of these events by the
3555  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3556  * use the SYSINIT handlers as they are more capable of expressing the
3557  * flow of control during module and vnet operations, so this is just
3558  * a skeleton. Note there is no SYSINIT equivalent of the module
3559  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3560  */
3561 static int
3562 ipfw_modevent(module_t mod, int type, void *unused)
3563 {
3564 	int err = 0;
3565 
3566 	switch (type) {
3567 	case MOD_LOAD:
3568 		/* Called once at module load or
3569 	 	 * system boot if compiled in. */
3570 		break;
3571 	case MOD_QUIESCE:
3572 		/* Called before unload. May veto unloading. */
3573 		break;
3574 	case MOD_UNLOAD:
3575 		/* Called during unload. */
3576 		break;
3577 	case MOD_SHUTDOWN:
3578 		/* Called during system shutdown. */
3579 		break;
3580 	default:
3581 		err = EOPNOTSUPP;
3582 		break;
3583 	}
3584 	return err;
3585 }
3586 
3587 static moduledata_t ipfwmod = {
3588 	"ipfw",
3589 	ipfw_modevent,
3590 	0
3591 };
3592 
3593 /* Define startup order. */
3594 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_FIREWALL
3595 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
3596 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
3597 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
3598 
3599 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3600 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3601 MODULE_VERSION(ipfw, 3);
3602 /* should declare some dependencies here */
3603 
3604 /*
3605  * Starting up. Done in order after ipfwmod() has been called.
3606  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3607  */
3608 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3609 	    ipfw_init, NULL);
3610 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3611 	    vnet_ipfw_init, NULL);
3612 
3613 /*
3614  * Closing up shop. These are done in REVERSE ORDER, but still
3615  * after ipfwmod() has been called. Not called on reboot.
3616  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3617  * or when the module is unloaded.
3618  */
3619 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3620 	    ipfw_destroy, NULL);
3621 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3622 	    vnet_ipfw_uninit, NULL);
3623 /* end of file */
3624