Context Navigation

source: rtems-libbsd/freebsd/sys/netpfil/ipfw/ip_fw2.c @ 3c967ca

55-freebsd-126-freebsd-12

Last change on this file since 3c967ca was 3c967ca, checked in by Sebastian Huber <sebastian.huber@…>, on 06/08/17 at 11:15:12
Use <sys/lock.h> provided by Newlib
Property mode set to `100644`
File size: 78.4 KB

Line
1	#include <machine/rtems-bsd-kernel-space.h>
2
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 <rtems/bsd/local/opt_ipfw.h>
36	#include <rtems/bsd/local/opt_ipdivert.h>
37	#include <rtems/bsd/local/opt_inet.h>
38	#ifndef INET
39	#error "IPFIREWALL requires INET"
40	#endif /* INET */
41	#include <rtems/bsd/local/opt_inet6.h>
42	#include <rtems/bsd/local/opt_ipsec.h>
43
44	#include <sys/param.h>
45	#include <sys/systm.h>
46	#include <sys/condvar.h>
47	#include <sys/counter.h>
48	#include <sys/eventhandler.h>
49	#include <sys/malloc.h>
50	#include <sys/mbuf.h>
51	#include <sys/kernel.h>
52	#include <sys/lock.h>
53	#include <sys/jail.h>
54	#include <sys/module.h>
55	#include <sys/priv.h>
56	#include <sys/proc.h>
57	#include <sys/rwlock.h>
58	#include <sys/rmlock.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/pfil.h>
69	#include <net/vnet.h>
70
71	#include <netpfil/pf/pf_mtag.h>
72
73	#include <netinet/in.h>
74	#include <netinet/in_var.h>
75	#include <netinet/in_pcb.h>
76	#include <netinet/ip.h>
77	#include <netinet/ip_var.h>
78	#include <netinet/ip_icmp.h>
79	#include <netinet/ip_fw.h>
80	#include <netinet/ip_carp.h>
81	#include <netinet/pim.h>
82	#include <netinet/tcp_var.h>
83	#include <netinet/udp.h>
84	#include <netinet/udp_var.h>
85	#include <netinet/sctp.h>
86
87	#include <netinet/ip6.h>
88	#include <netinet/icmp6.h>
89	#include <netinet/in_fib.h>
90	#ifdef INET6
91	#include <netinet6/in6_fib.h>
92	#include <netinet6/in6_pcb.h>
93	#include <netinet6/scope6_var.h>
94	#include <netinet6/ip6_var.h>
95	#endif
96
97	#include <netpfil/ipfw/ip_fw_private.h>
98
99	#include <machine/in_cksum.h> /* XXX for in_cksum */
100
101	#ifdef MAC
102	#include <security/mac/mac_framework.h>
103	#endif
104
105	/*
106	* static variables followed by global ones.
107	* All ipfw global variables are here.
108	*/
109
110	static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
111	#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs)
112
113	static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
114	#define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6)
115
116	#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
117	static int default_to_accept = 1;
118	#else
119	static int default_to_accept;
120	#endif
121
122	VNET_DEFINE(int, autoinc_step);
123	VNET_DEFINE(int, fw_one_pass) = 1;
124
125	VNET_DEFINE(unsigned int, fw_tables_max);
126	VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */
127	/* Use 128 tables by default */
128	static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
129
130	#ifndef LINEAR_SKIPTO
131	static int jump_fast(struct ip_fw_chain chain, struct ip_fw f, int num,
132	int tablearg, int jump_backwards);
133	#define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back)
134	#else
135	static int jump_linear(struct ip_fw_chain chain, struct ip_fw f, int num,
136	int tablearg, int jump_backwards);
137	#define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back)
138	#endif
139
140	/*
141	* Each rule belongs to one of 32 different sets (0..31).
142	* The variable set_disable contains one bit per set.
143	* If the bit is set, all rules in the corresponding set
144	* are disabled. Set RESVD_SET(31) is reserved for the default rule
145	* and rules that are not deleted by the flush command,
146	* and CANNOT be disabled.
147	* Rules in set RESVD_SET can only be deleted individually.
148	*/
149	VNET_DEFINE(u_int32_t, set_disable);
150	#define V_set_disable VNET(set_disable)
151
152	VNET_DEFINE(int, fw_verbose);
153	/* counter for ipfw_log(NULL...) */
154	VNET_DEFINE(u_int64_t, norule_counter);
155	VNET_DEFINE(int, verbose_limit);
156
157	/* layer3_chain contains the list of rules for layer 3 */
158	VNET_DEFINE(struct ip_fw_chain, layer3_chain);
159
160	/* ipfw_vnet_ready controls when we are open for business */
161	VNET_DEFINE(int, ipfw_vnet_ready) = 0;
162
163	VNET_DEFINE(int, ipfw_nat_ready) = 0;
164
165	ipfw_nat_t *ipfw_nat_ptr = NULL;
166	struct cfg_nat (lookup_nat_ptr)(struct nat_list *, int);
167	ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
168	ipfw_nat_cfg_t *ipfw_nat_del_ptr;
169	ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
170	ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
171
172	#ifdef SYSCTL_NODE
173	uint32_t dummy_def = IPFW_DEFAULT_RULE;
174	static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
175	static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
176
177	SYSBEGIN(f3)
178
179	SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
180	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
181	CTLFLAG_VNET \| CTLFLAG_RW \| CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
182	"Only do a single pass through ipfw when using dummynet(4)");
183	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
184	CTLFLAG_VNET \| CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
185	"Rule number auto-increment step");
186	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
187	CTLFLAG_VNET \| CTLFLAG_RW \| CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
188	"Log matches to ipfw rules");
189	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
190	CTLFLAG_VNET \| CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
191	"Set upper limit of matches of ipfw rules logged");
192	SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
193	&dummy_def, 0,
194	"The default/max possible rule number.");
195	SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
196	CTLFLAG_VNET \| CTLTYPE_UINT \| CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
197	"Maximum number of concurrently used tables");
198	SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
199	CTLFLAG_VNET \| CTLTYPE_UINT \| CTLFLAG_RW,
200	0, 0, sysctl_ipfw_tables_sets, "IU",
201	"Use per-set namespace for tables");
202	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
203	&default_to_accept, 0,
204	"Make the default rule accept all packets.");
205	TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
206	SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
207	CTLFLAG_VNET \| CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
208	"Number of static rules");
209
210	#ifdef INET6
211	SYSCTL_DECL(_net_inet6_ip6);
212	SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
213	SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
214	CTLFLAG_VNET \| CTLFLAG_RW \| CTLFLAG_SECURE,
215	&VNET_NAME(fw_deny_unknown_exthdrs), 0,
216	"Deny packets with unknown IPv6 Extension Headers");
217	SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
218	CTLFLAG_VNET \| CTLFLAG_RW \| CTLFLAG_SECURE,
219	&VNET_NAME(fw_permit_single_frag6), 0,
220	"Permit single packet IPv6 fragments");
221	#endif /* INET6 */
222
223	SYSEND
224
225	#endif /* SYSCTL_NODE */
226
227
228	/*
229	* Some macros used in the various matching options.
230	* L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
231	* Other macros just cast void * into the appropriate type
232	*/
233	#define L3HDR(T, ip) ((T )((u_int32_t )(ip) + (ip)->ip_hl))
234	#define TCP(p) ((struct tcphdr *)(p))
235	#define SCTP(p) ((struct sctphdr *)(p))
236	#define UDP(p) ((struct udphdr *)(p))
237	#define ICMP(p) ((struct icmphdr *)(p))
238	#define ICMP6(p) ((struct icmp6_hdr *)(p))
239
240	static __inline int
241	icmptype_match(struct icmphdr icmp, ipfw_insn_u32 cmd)
242	{
243	int type = icmp->icmp_type;
244
245	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
246	}
247
248	#define TT ( (1 << ICMP_ECHO) \| (1 << ICMP_ROUTERSOLICIT) \| \
249	(1 << ICMP_TSTAMP) \| (1 << ICMP_IREQ) \| (1 << ICMP_MASKREQ) )
250
251	static int
252	is_icmp_query(struct icmphdr *icmp)
253	{
254	int type = icmp->icmp_type;
255
256	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
257	}
258	#undef TT
259
260	/*
261	* The following checks use two arrays of 8 or 16 bits to store the
262	* bits that we want set or clear, respectively. They are in the
263	* low and high half of cmd->arg1 or cmd->d[0].
264	*
265	* We scan options and store the bits we find set. We succeed if
266	*
267	* (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
268	*
269	* The code is sometimes optimized not to store additional variables.
270	*/
271
272	static int
273	flags_match(ipfw_insn *cmd, u_int8_t bits)
274	{
275	u_char want_clear;
276	bits = ~bits;
277
278	if ( ((cmd->arg1 & 0xff) & bits) != 0)
279	return 0; /* some bits we want set were clear */
280	want_clear = (cmd->arg1 >> 8) & 0xff;
281	if ( (want_clear & bits) != want_clear)
282	return 0; /* some bits we want clear were set */
283	return 1;
284	}
285
286	static int
287	ipopts_match(struct ip ip, ipfw_insn cmd)
288	{
289	int optlen, bits = 0;
290	u_char cp = (u_char )(ip + 1);
291	int x = (ip->ip_hl << 2) - sizeof (struct ip);
292
293	for (; x > 0; x -= optlen, cp += optlen) {
294	int opt = cp[IPOPT_OPTVAL];
295
296	if (opt == IPOPT_EOL)
297	break;
298	if (opt == IPOPT_NOP)
299	optlen = 1;
300	else {
301	optlen = cp[IPOPT_OLEN];
302	if (optlen <= 0 \|\| optlen > x)
303	return 0; /* invalid or truncated */
304	}
305	switch (opt) {
306
307	default:
308	break;
309
310	case IPOPT_LSRR:
311	bits \|= IP_FW_IPOPT_LSRR;
312	break;
313
314	case IPOPT_SSRR:
315	bits \|= IP_FW_IPOPT_SSRR;
316	break;
317
318	case IPOPT_RR:
319	bits \|= IP_FW_IPOPT_RR;
320	break;
321
322	case IPOPT_TS:
323	bits \|= IP_FW_IPOPT_TS;
324	break;
325	}
326	}
327	return (flags_match(cmd, bits));
328	}
329
330	static int
331	tcpopts_match(struct tcphdr tcp, ipfw_insn cmd)
332	{
333	int optlen, bits = 0;
334	u_char cp = (u_char )(tcp + 1);
335	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
336
337	for (; x > 0; x -= optlen, cp += optlen) {
338	int opt = cp[0];
339	if (opt == TCPOPT_EOL)
340	break;
341	if (opt == TCPOPT_NOP)
342	optlen = 1;
343	else {
344	optlen = cp[1];
345	if (optlen <= 0)
346	break;
347	}
348
349	switch (opt) {
350
351	default:
352	break;
353
354	case TCPOPT_MAXSEG:
355	bits \|= IP_FW_TCPOPT_MSS;
356	break;
357
358	case TCPOPT_WINDOW:
359	bits \|= IP_FW_TCPOPT_WINDOW;
360	break;
361
362	case TCPOPT_SACK_PERMITTED:
363	case TCPOPT_SACK:
364	bits \|= IP_FW_TCPOPT_SACK;
365	break;
366
367	case TCPOPT_TIMESTAMP:
368	bits \|= IP_FW_TCPOPT_TS;
369	break;
370
371	}
372	}
373	return (flags_match(cmd, bits));
374	}
375
376	static int
377	iface_match(struct ifnet ifp, ipfw_insn_if cmd, struct ip_fw_chain *chain,
378	uint32_t *tablearg)
379	{
380
381	if (ifp == NULL) /* no iface with this packet, match fails */
382	return (0);
383
384	/* Check by name or by IP address */
385	if (cmd->name[0] != '\0') { /* match by name */
386	if (cmd->name[0] == '\1') /* use tablearg to match */
387	return ipfw_lookup_table(chain, cmd->p.kidx, 0,
388	&ifp->if_index, tablearg);
389	/* Check name */
390	if (cmd->p.glob) {
391	if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
392	return(1);
393	} else {
394	if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
395	return(1);
396	}
397	} else {
398	#if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
399	struct ifaddr *ia;
400
401	if_addr_rlock(ifp);
402	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
403	if (ia->ifa_addr->sa_family != AF_INET)
404	continue;
405	if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
406	(ia->ifa_addr))->sin_addr.s_addr) {
407	if_addr_runlock(ifp);
408	return(1); /* match */
409	}
410	}
411	if_addr_runlock(ifp);
412	#endif /* __FreeBSD__ */
413	}
414	return(0); /* no match, fail ... */
415	}
416
417	/*
418	* The verify_path function checks if a route to the src exists and
419	* if it is reachable via ifp (when provided).
420	*
421	* The 'verrevpath' option checks that the interface that an IP packet
422	* arrives on is the same interface that traffic destined for the
423	* packet's source address would be routed out of.
424	* The 'versrcreach' option just checks that the source address is
425	* reachable via any route (except default) in the routing table.
426	* These two are a measure to block forged packets. This is also
427	* commonly known as "anti-spoofing" or Unicast Reverse Path
428	* Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
429	* is purposely reminiscent of the Cisco IOS command,
430	*
431	* ip verify unicast reverse-path
432	* ip verify unicast source reachable-via any
433	*
434	* which implements the same functionality. But note that the syntax
435	* is misleading, and the check may be performed on all IP packets
436	* whether unicast, multicast, or broadcast.
437	*/
438	static int
439	verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
440	{
441	#if defined(USERSPACE) \|\| !defined(__FreeBSD__)
442	return 0;
443	#else
444	struct nhop4_basic nh4;
445
446	if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
447	return (0);
448
449	/*
450	* If ifp is provided, check for equality with rtentry.
451	* We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
452	* in order to pass packets injected back by if_simloop():
453	* routing entry (via lo0) for our own address
454	* may exist, so we need to handle routing assymetry.
455	*/
456	if (ifp != NULL && ifp != nh4.nh_ifp)
457	return (0);
458
459	/* if no ifp provided, check if rtentry is not default route */
460	if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
461	return (0);
462
463	/* or if this is a blackhole/reject route */
464	if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT\|NHF_BLACKHOLE)) != 0)
465	return (0);
466
467	/* found valid route */
468	return 1;
469	#endif /* __FreeBSD__ */
470	}
471
472	#ifdef INET6
473	/*
474	* ipv6 specific rules here...
475	*/
476	static __inline int
477	icmp6type_match (int type, ipfw_insn_u32 *cmd)
478	{
479	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
480	}
481
482	static int
483	flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
484	{
485	int i;
486	for (i=0; i <= cmd->o.arg1; ++i )
487	if (curr_flow == cmd->d[i] )
488	return 1;
489	return 0;
490	}
491
492	/* support for IP6__ME opcodes /
493	static const struct in6_addr lla_mask = {{{
494	0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
495	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
496	}}};
497
498	static int
499	ipfw_localip6(struct in6_addr *in6)
500	{
501	struct rm_priotracker in6_ifa_tracker;
502	struct in6_ifaddr *ia;
503
504	if (IN6_IS_ADDR_MULTICAST(in6))
505	return (0);
506
507	if (!IN6_IS_ADDR_LINKLOCAL(in6))
508	return (in6_localip(in6));
509
510	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
511	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
512	if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
513	continue;
514	if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
515	in6, &lla_mask)) {
516	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
517	return (1);
518	}
519	}
520	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
521	return (0);
522	}
523
524	static int
525	verify_path6(struct in6_addr src, struct ifnet ifp, u_int fib)
526	{
527	struct nhop6_basic nh6;
528
529	if (IN6_IS_SCOPE_LINKLOCAL(src))
530	return (1);
531
532	if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
533	return (0);
534
535	/* If ifp is provided, check for equality with route table. */
536	if (ifp != NULL && ifp != nh6.nh_ifp)
537	return (0);
538
539	/* if no ifp provided, check if rtentry is not default route */
540	if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
541	return (0);
542
543	/* or if this is a blackhole/reject route */
544	if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT\|NHF_BLACKHOLE)) != 0)
545	return (0);
546
547	/* found valid route */
548	return 1;
549	}
550
551	static int
552	is_icmp6_query(int icmp6_type)
553	{
554	if ((icmp6_type <= ICMP6_MAXTYPE) &&
555	(icmp6_type == ICMP6_ECHO_REQUEST \|\|
556	icmp6_type == ICMP6_MEMBERSHIP_QUERY \|\|
557	icmp6_type == ICMP6_WRUREQUEST \|\|
558	icmp6_type == ICMP6_FQDN_QUERY \|\|
559	icmp6_type == ICMP6_NI_QUERY))
560	return (1);
561
562	return (0);
563	}
564
565	static void
566	send_reject6(struct ip_fw_args args, int code, u_int hlen, struct ip6_hdr ip6)
567	{
568	struct mbuf *m;
569
570	m = args->m;
571	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
572	struct tcphdr *tcp;
573	tcp = (struct tcphdr )((char )ip6 + hlen);
574
575	if ((tcp->th_flags & TH_RST) == 0) {
576	struct mbuf *m0;
577	m0 = ipfw_send_pkt(args->m, &(args->f_id),
578	ntohl(tcp->th_seq), ntohl(tcp->th_ack),
579	tcp->th_flags \| TH_RST);
580	if (m0 != NULL)
581	ip6_output(m0, NULL, NULL, 0, NULL, NULL,
582	NULL);
583	}
584	FREE_PKT(m);
585	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
586	#if 0
587	/*
588	* Unlike above, the mbufs need to line up with the ip6 hdr,
589	* as the contents are read. We need to m_adj() the
590	* needed amount.
591	* The mbuf will however be thrown away so we can adjust it.
592	* Remember we did an m_pullup on it already so we
593	* can make some assumptions about contiguousness.
594	*/
595	if (args->L3offset)
596	m_adj(m, args->L3offset);
597	#endif
598	icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
599	} else
600	FREE_PKT(m);
601
602	args->m = NULL;
603	}
604
605	#endif /* INET6 */
606
607
608	/*
609	* sends a reject message, consuming the mbuf passed as an argument.
610	*/
611	static void
612	send_reject(struct ip_fw_args args, int code, int iplen, struct ip ip)
613	{
614
615	#if 0
616	/* XXX When ip is not guaranteed to be at mtod() we will
617	* need to account for this */
618	* The mbuf will however be thrown away so we can adjust it.
619	* Remember we did an m_pullup on it already so we
620	* can make some assumptions about contiguousness.
621	*/
622	if (args->L3offset)
623	m_adj(m, args->L3offset);
624	#endif
625	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
626	icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
627	} else if (args->f_id.proto == IPPROTO_TCP) {
628	struct tcphdr *const tcp =
629	L3HDR(struct tcphdr, mtod(args->m, struct ip *));
630	if ( (tcp->th_flags & TH_RST) == 0) {
631	struct mbuf *m;
632	m = ipfw_send_pkt(args->m, &(args->f_id),
633	ntohl(tcp->th_seq), ntohl(tcp->th_ack),
634	tcp->th_flags \| TH_RST);
635	if (m != NULL)
636	ip_output(m, NULL, NULL, 0, NULL, NULL);
637	}
638	FREE_PKT(args->m);
639	} else
640	FREE_PKT(args->m);
641	args->m = NULL;
642	}
643
644	/*
645	* Support for uid/gid/jail lookup. These tests are expensive
646	* (because we may need to look into the list of active sockets)
647	* so we cache the results. ugid_lookupp is 0 if we have not
648	* yet done a lookup, 1 if we succeeded, and -1 if we tried
649	* and failed. The function always returns the match value.
650	* We could actually spare the variable and use *uc, setting
651	* it to '(void *)check_uidgid if we have no info, NULL if
652	* we tried and failed, or any other value if successful.
653	*/
654	static int
655	check_uidgid(ipfw_insn_u32 insn, struct ip_fw_args args, int *ugid_lookupp,
656	struct ucred **uc)
657	{
658	#if defined(USERSPACE)
659	return 0; // not supported in userspace
660	#else
661	#ifndef __FreeBSD__
662	/* XXX */
663	return cred_check(insn, proto, oif,
664	dst_ip, dst_port, src_ip, src_port,
665	(struct bsd_ucred )uc, ugid_lookupp, ((struct mbuf )inp)->m_skb);
666	#else /* FreeBSD */
667	struct in_addr src_ip, dst_ip;
668	struct inpcbinfo *pi;
669	struct ipfw_flow_id *id;
670	struct inpcb pcb, inp;
671	struct ifnet *oif;
672	int lookupflags;
673	int match;
674
675	id = &args->f_id;
676	inp = args->inp;
677	oif = args->oif;
678
679	/*
680	* Check to see if the UDP or TCP stack supplied us with
681	* the PCB. If so, rather then holding a lock and looking
682	* up the PCB, we can use the one that was supplied.
683	*/
684	if (inp && *ugid_lookupp == 0) {
685	INP_LOCK_ASSERT(inp);
686	if (inp->inp_socket != NULL) {
687	*uc = crhold(inp->inp_cred);
688	*ugid_lookupp = 1;
689	} else
690	*ugid_lookupp = -1;
691	}
692	/*
693	* If we have already been here and the packet has no
694	* PCB entry associated with it, then we can safely
695	* assume that this is a no match.
696	*/
697	if (*ugid_lookupp == -1)
698	return (0);
699	if (id->proto == IPPROTO_TCP) {
700	lookupflags = 0;
701	pi = &V_tcbinfo;
702	} else if (id->proto == IPPROTO_UDP) {
703	lookupflags = INPLOOKUP_WILDCARD;
704	pi = &V_udbinfo;
705	} else
706	return 0;
707	lookupflags \|= INPLOOKUP_RLOCKPCB;
708	match = 0;
709	if (*ugid_lookupp == 0) {
710	if (id->addr_type == 6) {
711	#ifdef INET6
712	if (oif == NULL)
713	pcb = in6_pcblookup_mbuf(pi,
714	&id->src_ip6, htons(id->src_port),
715	&id->dst_ip6, htons(id->dst_port),
716	lookupflags, oif, args->m);
717	else
718	pcb = in6_pcblookup_mbuf(pi,
719	&id->dst_ip6, htons(id->dst_port),
720	&id->src_ip6, htons(id->src_port),
721	lookupflags, oif, args->m);
722	#else
723	*ugid_lookupp = -1;
724	return (0);
725	#endif
726	} else {
727	src_ip.s_addr = htonl(id->src_ip);
728	dst_ip.s_addr = htonl(id->dst_ip);
729	if (oif == NULL)
730	pcb = in_pcblookup_mbuf(pi,
731	src_ip, htons(id->src_port),
732	dst_ip, htons(id->dst_port),
733	lookupflags, oif, args->m);
734	else
735	pcb = in_pcblookup_mbuf(pi,
736	dst_ip, htons(id->dst_port),
737	src_ip, htons(id->src_port),
738	lookupflags, oif, args->m);
739	}
740	if (pcb != NULL) {
741	INP_RLOCK_ASSERT(pcb);
742	*uc = crhold(pcb->inp_cred);
743	*ugid_lookupp = 1;
744	INP_RUNLOCK(pcb);
745	}
746	if (*ugid_lookupp == 0) {
747	/*
748	* We tried and failed, set the variable to -1
749	* so we will not try again on this packet.
750	*/
751	*ugid_lookupp = -1;
752	return (0);
753	}
754	}
755	if (insn->o.opcode == O_UID)
756	#ifndef __rtems__
757	match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
758	#else /* __rtems__ */
759	match = (BSD_DEFAULT_UID == (uid_t)insn->d[0]);
760	#endif /* __rtems__ */
761	else if (insn->o.opcode == O_GID)
762	match = groupmember((gid_t)insn->d[0], *uc);
763	else if (insn->o.opcode == O_JAIL)
764	#ifndef __rtems__
765	match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
766	#else /* __rtems__ */
767	match = (BSD_DEFAULT_PRISON->pr_id == (int)insn->d[0]);
768	#endif /* __rtems__ */
769	return (match);
770	#endif /* __FreeBSD__ */
771	#endif /* not supported in userspace */
772	}
773
774	/*
775	* Helper function to set args with info on the rule after the matching
776	* one. slot is precise, whereas we guess rule_id as they are
777	* assigned sequentially.
778	*/
779	static inline void
780	set_match(struct ip_fw_args *args, int slot,
781	struct ip_fw_chain *chain)
782	{
783	args->rule.chain_id = chain->id;
784	args->rule.slot = slot + 1; /* we use 0 as a marker */
785	args->rule.rule_id = 1 + chain->map[slot]->id;
786	args->rule.rulenum = chain->map[slot]->rulenum;
787	}
788
789	#ifndef LINEAR_SKIPTO
790	/*
791	* Helper function to enable cached rule lookups using
792	* cached_id and cached_pos fields in ipfw rule.
793	*/
794	static int
795	jump_fast(struct ip_fw_chain chain, struct ip_fw f, int num,
796	int tablearg, int jump_backwards)
797	{
798	int f_pos;
799
800	/* If possible use cached f_pos (in f->cached_pos),
801	* whose version is written in f->cached_id
802	* (horrible hacks to avoid changing the ABI).
803	*/
804	if (num != IP_FW_TARG && f->cached_id == chain->id)
805	f_pos = f->cached_pos;
806	else {
807	int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
808	/* make sure we do not jump backward */
809	if (jump_backwards == 0 && i <= f->rulenum)
810	i = f->rulenum + 1;
811	if (chain->idxmap != NULL)
812	f_pos = chain->idxmap[i];
813	else
814	f_pos = ipfw_find_rule(chain, i, 0);
815	/* update the cache */
816	if (num != IP_FW_TARG) {
817	f->cached_id = chain->id;
818	f->cached_pos = f_pos;
819	}
820	}
821
822	return (f_pos);
823	}
824	#else
825	/*
826	* Helper function to enable real fast rule lookups.
827	*/
828	static int
829	jump_linear(struct ip_fw_chain chain, struct ip_fw f, int num,
830	int tablearg, int jump_backwards)
831	{
832	int f_pos;
833
834	num = IP_FW_ARG_TABLEARG(chain, num, skipto);
835	/* make sure we do not jump backward */
836	if (jump_backwards == 0 && num <= f->rulenum)
837	num = f->rulenum + 1;
838	f_pos = chain->idxmap[num];
839
840	return (f_pos);
841	}
842	#endif
843
844	#define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f)
845	/*
846	* The main check routine for the firewall.
847	*
848	* All arguments are in args so we can modify them and return them
849	* back to the caller.
850	*
851	* Parameters:
852	*
853	* args->m (in/out) The packet; we set to NULL when/if we nuke it.
854	* Starts with the IP header.
855	* args->eh (in) Mac header if present, NULL for layer3 packet.
856	* args->L3offset Number of bytes bypassed if we came from L2.
857	* e.g. often sizeof(eh) NOTYET
858	* args->oif Outgoing interface, NULL if packet is incoming.
859	* The incoming interface is in the mbuf. (in)
860	* args->divert_rule (in/out)
861	* Skip up to the first rule past this rule number;
862	* upon return, non-zero port number for divert or tee.
863	*
864	* args->rule Pointer to the last matching rule (in/out)
865	* args->next_hop Socket we are forwarding to (out).
866	* args->next_hop6 IPv6 next hop we are forwarding to (out).
867	* args->f_id Addresses grabbed from the packet (out)
868	* args->rule.info a cookie depending on rule action
869	*
870	* Return value:
871	*
872	* IP_FW_PASS the packet must be accepted
873	* IP_FW_DENY the packet must be dropped
874	* IP_FW_DIVERT divert packet, port in m_tag
875	* IP_FW_TEE tee packet, port in m_tag
876	* IP_FW_DUMMYNET to dummynet, pipe in args->cookie
877	* IP_FW_NETGRAPH into netgraph, cookie args->cookie
878	* args->rule contains the matching rule,
879	* args->rule.info has additional information.
880	*
881	*/
882	int
883	ipfw_chk(struct ip_fw_args *args)
884	{
885
886	/*
887	* Local variables holding state while processing a packet:
888	*
889	* IMPORTANT NOTE: to speed up the processing of rules, there
890	* are some assumption on the values of the variables, which
891	* are documented here. Should you change them, please check
892	* the implementation of the various instructions to make sure
893	* that they still work.
894	*
895	* args->eh The MAC header. It is non-null for a layer2
896	* packet, it is NULL for a layer-3 packet.
897	* notyet
898	* args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
899	*
900	* m \| args->m Pointer to the mbuf, as received from the caller.
901	* It may change if ipfw_chk() does an m_pullup, or if it
902	* consumes the packet because it calls send_reject().
903	* XXX This has to change, so that ipfw_chk() never modifies
904	* or consumes the buffer.
905	* ip is the beginning of the ip(4 or 6) header.
906	* Calculated by adding the L3offset to the start of data.
907	* (Until we start using L3offset, the packet is
908	* supposed to start with the ip header).
909	*/
910	struct mbuf *m = args->m;
911	struct ip ip = mtod(m, struct ip );
912
913	/*
914	* For rules which contain uid/gid or jail constraints, cache
915	* a copy of the users credentials after the pcb lookup has been
916	* executed. This will speed up the processing of rules with
917	* these types of constraints, as well as decrease contention
918	* on pcb related locks.
919	*/
920	#ifndef __FreeBSD__
921	struct bsd_ucred ucred_cache;
922	#else
923	struct ucred *ucred_cache = NULL;
924	#endif
925	int ucred_lookup = 0;
926
927	/*
928	* oif \| args->oif If NULL, ipfw_chk has been called on the
929	* inbound path (ether_input, ip_input).
930	* If non-NULL, ipfw_chk has been called on the outbound path
931	* (ether_output, ip_output).
932	*/
933	struct ifnet *oif = args->oif;
934
935	int f_pos = 0; /* index of current rule in the array */
936	int retval = 0;
937
938	/*
939	* hlen The length of the IP header.
940	*/
941	u_int hlen = 0; /* hlen >0 means we have an IP pkt */
942
943	/*
944	* offset The offset of a fragment. offset != 0 means that
945	* we have a fragment at this offset of an IPv4 packet.
946	* offset == 0 means that (if this is an IPv4 packet)
947	* this is the first or only fragment.
948	* For IPv6 offset\|ip6f_mf == 0 means there is no Fragment Header
949	* or there is a single packet fragment (fragment header added
950	* without needed). We will treat a single packet fragment as if
951	* there was no fragment header (or log/block depending on the
952	* V_fw_permit_single_frag6 sysctl setting).
953	*/
954	u_short offset = 0;
955	u_short ip6f_mf = 0;
956
957	/*
958	* Local copies of addresses. They are only valid if we have
959	* an IP packet.
960	*
961	* proto The protocol. Set to 0 for non-ip packets,
962	* or to the protocol read from the packet otherwise.
963	* proto != 0 means that we have an IPv4 packet.
964	*
965	* src_port, dst_port port numbers, in HOST format. Only
966	* valid for TCP and UDP packets.
967	*
968	* src_ip, dst_ip ip addresses, in NETWORK format.
969	* Only valid for IPv4 packets.
970	*/
971	uint8_t proto;
972	uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
973	struct in_addr src_ip, dst_ip; /* NOTE: network format */
974	uint16_t iplen=0;
975	int pktlen;
976	uint16_t etype = 0; /* Host order stored ether type */
977
978	/*
979	* dyn_dir = MATCH_UNKNOWN when rules unchecked,
980	* MATCH_NONE when checked and not matched (q = NULL),
981	* MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
982	*/
983	int dyn_dir = MATCH_UNKNOWN;
984	uint16_t dyn_name = 0;
985	ipfw_dyn_rule *q = NULL;
986	struct ip_fw_chain *chain = &V_layer3_chain;
987
988	/*
989	* We store in ulp a pointer to the upper layer protocol header.
990	* In the ipv4 case this is easy to determine from the header,
991	* but for ipv6 we might have some additional headers in the middle.
992	* ulp is NULL if not found.
993	*/
994	void ulp = NULL; / upper layer protocol pointer. */
995
996	/* XXX ipv6 variables */
997	int is_ipv6 = 0;
998	uint8_t icmp6_type = 0;
999	uint16_t ext_hd = 0; /* bits vector for extension header filtering */
1000	/* end of ipv6 variables */
1001
1002	int is_ipv4 = 0;
1003
1004	int done = 0; /* flag to exit the outer loop */
1005
1006	if (m->m_flags & M_SKIP_FIREWALL \|\| (! V_ipfw_vnet_ready))
1007	return (IP_FW_PASS); /* accept */
1008
1009	dst_ip.s_addr = 0; /* make sure it is initialized */
1010	src_ip.s_addr = 0; /* make sure it is initialized */
1011	pktlen = m->m_pkthdr.len;
1012	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1013	proto = args->f_id.proto = 0; /* mark f_id invalid */
1014	/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1015
1016	/*
1017	* PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1018	* then it sets p to point at the offset "len" in the mbuf. WARNING: the
1019	* pointer might become stale after other pullups (but we never use it
1020	* this way).
1021	*/
1022	#define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
1023	#define PULLUP_LEN(_len, p, T) \
1024	do { \
1025	int x = (_len) + T; \
1026	if ((m)->m_len < x) { \
1027	args->m = m = m_pullup(m, x); \
1028	if (m == NULL) \
1029	goto pullup_failed; \
1030	} \
1031	p = (mtod(m, char *) + (_len)); \
1032	} while (0)
1033
1034	/*
1035	* if we have an ether header,
1036	*/
1037	if (args->eh)
1038	etype = ntohs(args->eh->ether_type);
1039
1040	/* Identify IP packets and fill up variables. */
1041	if (pktlen >= sizeof(struct ip6_hdr) &&
1042	(args->eh == NULL \|\| etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1043	struct ip6_hdr ip6 = (struct ip6_hdr )ip;
1044	is_ipv6 = 1;
1045	args->f_id.addr_type = 6;
1046	hlen = sizeof(struct ip6_hdr);
1047	proto = ip6->ip6_nxt;
1048
1049	/* Search extension headers to find upper layer protocols */
1050	while (ulp == NULL && offset == 0) {
1051	switch (proto) {
1052	case IPPROTO_ICMPV6:
1053	PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1054	icmp6_type = ICMP6(ulp)->icmp6_type;
1055	break;
1056
1057	case IPPROTO_TCP:
1058	PULLUP_TO(hlen, ulp, struct tcphdr);
1059	dst_port = TCP(ulp)->th_dport;
1060	src_port = TCP(ulp)->th_sport;
1061	/* save flags for dynamic rules */
1062	args->f_id._flags = TCP(ulp)->th_flags;
1063	break;
1064
1065	case IPPROTO_SCTP:
1066	PULLUP_TO(hlen, ulp, struct sctphdr);
1067	src_port = SCTP(ulp)->src_port;
1068	dst_port = SCTP(ulp)->dest_port;
1069	break;
1070
1071	case IPPROTO_UDP:
1072	PULLUP_TO(hlen, ulp, struct udphdr);
1073	dst_port = UDP(ulp)->uh_dport;
1074	src_port = UDP(ulp)->uh_sport;
1075	break;
1076
1077	case IPPROTO_HOPOPTS: /* RFC 2460 */
1078	PULLUP_TO(hlen, ulp, struct ip6_hbh);
1079	ext_hd \|= EXT_HOPOPTS;
1080	hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1081	proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1082	ulp = NULL;
1083	break;
1084
1085	case IPPROTO_ROUTING: /* RFC 2460 */
1086	PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1087	switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1088	case 0:
1089	ext_hd \|= EXT_RTHDR0;
1090	break;
1091	case 2:
1092	ext_hd \|= EXT_RTHDR2;
1093	break;
1094	default:
1095	if (V_fw_verbose)
1096	printf("IPFW2: IPV6 - Unknown "
1097	"Routing Header type(%d)\n",
1098	((struct ip6_rthdr *)
1099	ulp)->ip6r_type);
1100	if (V_fw_deny_unknown_exthdrs)
1101	return (IP_FW_DENY);
1102	break;
1103	}
1104	ext_hd \|= EXT_ROUTING;
1105	hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1106	proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1107	ulp = NULL;
1108	break;
1109
1110	case IPPROTO_FRAGMENT: /* RFC 2460 */
1111	PULLUP_TO(hlen, ulp, struct ip6_frag);
1112	ext_hd \|= EXT_FRAGMENT;
1113	hlen += sizeof (struct ip6_frag);
1114	proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1115	offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1116	IP6F_OFF_MASK;
1117	ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1118	IP6F_MORE_FRAG;
1119	if (V_fw_permit_single_frag6 == 0 &&
1120	offset == 0 && ip6f_mf == 0) {
1121	if (V_fw_verbose)
1122	printf("IPFW2: IPV6 - Invalid "
1123	"Fragment Header\n");
1124	if (V_fw_deny_unknown_exthdrs)
1125	return (IP_FW_DENY);
1126	break;
1127	}
1128	args->f_id.extra =
1129	ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1130	ulp = NULL;
1131	break;
1132
1133	case IPPROTO_DSTOPTS: /* RFC 2460 */
1134	PULLUP_TO(hlen, ulp, struct ip6_hbh);
1135	ext_hd \|= EXT_DSTOPTS;
1136	hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1137	proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1138	ulp = NULL;
1139	break;
1140
1141	case IPPROTO_AH: /* RFC 2402 */
1142	PULLUP_TO(hlen, ulp, struct ip6_ext);
1143	ext_hd \|= EXT_AH;
1144	hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1145	proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1146	ulp = NULL;
1147	break;
1148
1149	case IPPROTO_ESP: /* RFC 2406 */
1150	PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1151	/* Anything past Seq# is variable length and
1152	* data past this ext. header is encrypted. */
1153	ext_hd \|= EXT_ESP;
1154	break;
1155
1156	case IPPROTO_NONE: /* RFC 2460 */
1157	/*
1158	* Packet ends here, and IPv6 header has
1159	* already been pulled up. If ip6e_len!=0
1160	* then octets must be ignored.
1161	*/
1162	ulp = ip; /* non-NULL to get out of loop. */
1163	break;
1164
1165	case IPPROTO_OSPFIGP:
1166	/* XXX OSPF header check? */
1167	PULLUP_TO(hlen, ulp, struct ip6_ext);
1168	break;
1169
1170	case IPPROTO_PIM:
1171	/* XXX PIM header check? */
1172	PULLUP_TO(hlen, ulp, struct pim);
1173	break;
1174
1175	case IPPROTO_CARP:
1176	PULLUP_TO(hlen, ulp, struct carp_header);
1177	if (((struct carp_header *)ulp)->carp_version !=
1178	CARP_VERSION)
1179	return (IP_FW_DENY);
1180	if (((struct carp_header *)ulp)->carp_type !=
1181	CARP_ADVERTISEMENT)
1182	return (IP_FW_DENY);
1183	break;
1184
1185	case IPPROTO_IPV6: /* RFC 2893 */
1186	PULLUP_TO(hlen, ulp, struct ip6_hdr);
1187	break;
1188
1189	case IPPROTO_IPV4: /* RFC 2893 */
1190	PULLUP_TO(hlen, ulp, struct ip);
1191	break;
1192
1193	default:
1194	if (V_fw_verbose)
1195	printf("IPFW2: IPV6 - Unknown "
1196	"Extension Header(%d), ext_hd=%x\n",
1197	proto, ext_hd);
1198	if (V_fw_deny_unknown_exthdrs)
1199	return (IP_FW_DENY);
1200	PULLUP_TO(hlen, ulp, struct ip6_ext);
1201	break;
1202	} /switch /
1203	}
1204	ip = mtod(m, struct ip *);
1205	ip6 = (struct ip6_hdr *)ip;
1206	args->f_id.src_ip6 = ip6->ip6_src;
1207	args->f_id.dst_ip6 = ip6->ip6_dst;
1208	args->f_id.src_ip = 0;
1209	args->f_id.dst_ip = 0;
1210	args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1211	} else if (pktlen >= sizeof(struct ip) &&
1212	(args->eh == NULL \|\| etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1213	is_ipv4 = 1;
1214	hlen = ip->ip_hl << 2;
1215	args->f_id.addr_type = 4;
1216
1217	/*
1218	* Collect parameters into local variables for faster matching.
1219	*/
1220	proto = ip->ip_p;
1221	src_ip = ip->ip_src;
1222	dst_ip = ip->ip_dst;
1223	offset = ntohs(ip->ip_off) & IP_OFFMASK;
1224	iplen = ntohs(ip->ip_len);
1225	pktlen = iplen < pktlen ? iplen : pktlen;
1226
1227	if (offset == 0) {
1228	switch (proto) {
1229	case IPPROTO_TCP:
1230	PULLUP_TO(hlen, ulp, struct tcphdr);
1231	dst_port = TCP(ulp)->th_dport;
1232	src_port = TCP(ulp)->th_sport;
1233	/* save flags for dynamic rules */
1234	args->f_id._flags = TCP(ulp)->th_flags;
1235	break;
1236
1237	case IPPROTO_SCTP:
1238	PULLUP_TO(hlen, ulp, struct sctphdr);
1239	src_port = SCTP(ulp)->src_port;
1240	dst_port = SCTP(ulp)->dest_port;
1241	break;
1242
1243	case IPPROTO_UDP:
1244	PULLUP_TO(hlen, ulp, struct udphdr);
1245	dst_port = UDP(ulp)->uh_dport;
1246	src_port = UDP(ulp)->uh_sport;
1247	break;
1248
1249	case IPPROTO_ICMP:
1250	PULLUP_TO(hlen, ulp, struct icmphdr);
1251	//args->f_id.flags = ICMP(ulp)->icmp_type;
1252	break;
1253
1254	default:
1255	break;
1256	}
1257	}
1258
1259	ip = mtod(m, struct ip *);
1260	args->f_id.src_ip = ntohl(src_ip.s_addr);
1261	args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1262	}
1263	#undef PULLUP_TO
1264	if (proto) { /* we may have port numbers, store them */
1265	args->f_id.proto = proto;
1266	args->f_id.src_port = src_port = ntohs(src_port);
1267	args->f_id.dst_port = dst_port = ntohs(dst_port);
1268	}
1269
1270	IPFW_PF_RLOCK(chain);
1271	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1272	IPFW_PF_RUNLOCK(chain);
1273	return (IP_FW_PASS); /* accept */
1274	}
1275	if (args->rule.slot) {
1276	/*
1277	* Packet has already been tagged as a result of a previous
1278	* match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1279	* REASS, NETGRAPH, DIVERT/TEE...)
1280	* Validate the slot and continue from the next one
1281	* if still present, otherwise do a lookup.
1282	*/
1283	f_pos = (args->rule.chain_id == chain->id) ?
1284	args->rule.slot :
1285	ipfw_find_rule(chain, args->rule.rulenum,
1286	args->rule.rule_id);
1287	} else {
1288	f_pos = 0;
1289	}
1290
1291	/*
1292	* Now scan the rules, and parse microinstructions for each rule.
1293	* We have two nested loops and an inner switch. Sometimes we
1294	* need to break out of one or both loops, or re-enter one of
1295	* the loops with updated variables. Loop variables are:
1296	*
1297	* f_pos (outer loop) points to the current rule.
1298	* On output it points to the matching rule.
1299	* done (outer loop) is used as a flag to break the loop.
1300	* l (inner loop) residual length of current rule.
1301	* cmd points to the current microinstruction.
1302	*
1303	* We break the inner loop by setting l=0 and possibly
1304	* cmdlen=0 if we don't want to advance cmd.
1305	* We break the outer loop by setting done=1
1306	* We can restart the inner loop by setting l>0 and f_pos, f, cmd
1307	* as needed.
1308	*/
1309	for (; f_pos < chain->n_rules; f_pos++) {
1310	ipfw_insn *cmd;
1311	uint32_t tablearg = 0;
1312	int l, cmdlen, skip_or; /* skip rest of OR block */
1313	struct ip_fw *f;
1314
1315	f = chain->map[f_pos];
1316	if (V_set_disable & (1 << f->set) )
1317	continue;
1318
1319	skip_or = 0;
1320	for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1321	l -= cmdlen, cmd += cmdlen) {
1322	int match;
1323
1324	/*
1325	* check_body is a jump target used when we find a
1326	* CHECK_STATE, and need to jump to the body of
1327	* the target rule.
1328	*/
1329
1330	/* check_body: */
1331	cmdlen = F_LEN(cmd);
1332	/*
1333	* An OR block (insn_1 \|\| .. \|\| insn_n) has the
1334	* F_OR bit set in all but the last instruction.
1335	* The first match will set "skip_or", and cause
1336	* the following instructions to be skipped until
1337	* past the one with the F_OR bit clear.
1338	*/
1339	if (skip_or) { /* skip this instruction */
1340	if ((cmd->len & F_OR) == 0)
1341	skip_or = 0; /* next one is good */
1342	continue;
1343	}
1344	match = 0; /* set to 1 if we succeed */
1345
1346	switch (cmd->opcode) {
1347	/*
1348	* The first set of opcodes compares the packet's
1349	* fields with some pattern, setting 'match' if a
1350	* match is found. At the end of the loop there is
1351	* logic to deal with F_NOT and F_OR flags associated
1352	* with the opcode.
1353	*/
1354	case O_NOP:
1355	match = 1;
1356	break;
1357
1358	case O_FORWARD_MAC:
1359	printf("ipfw: opcode %d unimplemented\n",
1360	cmd->opcode);
1361	break;
1362
1363	case O_GID:
1364	case O_UID:
1365	case O_JAIL:
1366	/*
1367	* We only check offset == 0 && proto != 0,
1368	* as this ensures that we have a
1369	* packet with the ports info.
1370	*/
1371	if (offset != 0)
1372	break;
1373	if (proto == IPPROTO_TCP \|\|
1374	proto == IPPROTO_UDP)
1375	match = check_uidgid(
1376	(ipfw_insn_u32 *)cmd,
1377	args, &ucred_lookup,
1378	#ifdef __FreeBSD__
1379	&ucred_cache);
1380	#else
1381	(void *)&ucred_cache);
1382	#endif
1383	break;
1384
1385	case O_RECV:
1386	match = iface_match(m->m_pkthdr.rcvif,
1387	(ipfw_insn_if *)cmd, chain, &tablearg);
1388	break;
1389
1390	case O_XMIT:
1391	match = iface_match(oif, (ipfw_insn_if *)cmd,
1392	chain, &tablearg);
1393	break;
1394
1395	case O_VIA:
1396	match = iface_match(oif ? oif :
1397	m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1398	chain, &tablearg);
1399	break;
1400
1401	case O_MACADDR2:
1402	if (args->eh != NULL) { /* have MAC header */
1403	u_int32_t want = (u_int32_t )
1404	((ipfw_insn_mac *)cmd)->addr;
1405	u_int32_t mask = (u_int32_t )
1406	((ipfw_insn_mac *)cmd)->mask;
1407	u_int32_t hdr = (u_int32_t )args->eh;
1408
1409	match =
1410	( want[0] == (hdr[0] & mask[0]) &&
1411	want[1] == (hdr[1] & mask[1]) &&
1412	want[2] == (hdr[2] & mask[2]) );
1413	}
1414	break;
1415
1416	case O_MAC_TYPE:
1417	if (args->eh != NULL) {
1418	u_int16_t *p =
1419	((ipfw_insn_u16 *)cmd)->ports;
1420	int i;
1421
1422	for (i = cmdlen - 1; !match && i>0;
1423	i--, p += 2)
1424	match = (etype >= p[0] &&
1425	etype <= p[1]);
1426	}
1427	break;
1428
1429	case O_FRAG:
1430	match = (offset != 0);
1431	break;
1432
1433	case O_IN: /* "out" is "not in" */
1434	match = (oif == NULL);
1435	break;
1436
1437	case O_LAYER2:
1438	match = (args->eh != NULL);
1439	break;
1440
1441	case O_DIVERTED:
1442	{
1443	/* For diverted packets, args->rule.info
1444	* contains the divert port (in host format)
1445	* reason and direction.
1446	*/
1447	uint32_t i = args->rule.info;
1448	match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1449	cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1450	}
1451	break;
1452
1453	case O_PROTO:
1454	/*
1455	* We do not allow an arg of 0 so the
1456	* check of "proto" only suffices.
1457	*/
1458	match = (proto == cmd->arg1);
1459	break;
1460
1461	case O_IP_SRC:
1462	match = is_ipv4 &&
1463	(((ipfw_insn_ip *)cmd)->addr.s_addr ==
1464	src_ip.s_addr);
1465	break;
1466
1467	case O_IP_DST_LOOKUP:
1468	{
1469	void *pkey;
1470	uint32_t vidx, key;
1471	uint16_t keylen;
1472
1473	if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1474	/* Determine lookup key type */
1475	vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1476	if (vidx != 4 /* uid */ &&
1477	vidx != 5 /* jail */ &&
1478	is_ipv6 == 0 && is_ipv4 == 0)
1479	break;
1480	/* Determine key length */
1481	if (vidx == 0 /* dst-ip */ \|\|
1482	vidx == 1 /* src-ip */)
1483	keylen = is_ipv6 ?
1484	sizeof(struct in6_addr):
1485	sizeof(in_addr_t);
1486	else {
1487	keylen = sizeof(key);
1488	pkey = &key;
1489	}
1490	if (vidx == 0 /* dst-ip */)
1491	pkey = is_ipv4 ? (void *)&dst_ip:
1492	(void *)&args->f_id.dst_ip6;
1493	else if (vidx == 1 /* src-ip */)
1494	pkey = is_ipv4 ? (void *)&src_ip:
1495	(void *)&args->f_id.src_ip6;
1496	else if (vidx == 6 /* dscp */) {
1497	if (is_ipv4)
1498	key = ip->ip_tos >> 2;
1499	else {
1500	key = args->f_id.flow_id6;
1501	key = (key & 0x0f) << 2 \|
1502	(key & 0xf000) >> 14;
1503	}
1504	key &= 0x3f;
1505	} else if (vidx == 2 /* dst-port */ \|\|
1506	vidx == 3 /* src-port */) {
1507	/* Skip fragments */
1508	if (offset != 0)
1509	break;
1510	/* Skip proto without ports */
1511	if (proto != IPPROTO_TCP &&
1512	proto != IPPROTO_UDP &&
1513	proto != IPPROTO_SCTP)
1514	break;
1515	if (vidx == 2 /* dst-port */)
1516	key = dst_port;
1517	else
1518	key = src_port;
1519	}
1520	#ifndef USERSPACE
1521	else if (vidx == 4 /* uid */ \|\|
1522	vidx == 5 /* jail */) {
1523	check_uidgid(
1524	(ipfw_insn_u32 *)cmd,
1525	args, &ucred_lookup,
1526	#ifdef __FreeBSD__
1527	&ucred_cache);
1528	if (vidx == 4 /* uid */)
1529	#ifndef __rtems__
1530	key = ucred_cache->cr_uid;
1531	#else /* __rtems__ */
1532	key = BSD_DEFAULT_UID;
1533	#endif /* __rtems__ */
1534	else if (vidx == 5 /* jail */)
1535	#ifndef __rtems__
1536	key = ucred_cache->cr_prison->pr_id;
1537	#else /* __rtems__ */
1538	key = BSD_DEFAULT_PRISON->pr_id;
1539	#endif /* __rtems__ */
1540	#else /* !__FreeBSD__ */
1541	(void *)&ucred_cache);
1542	if (vidx == 4 /* uid */)
1543	key = ucred_cache.uid;
1544	else if (vidx == 5 /* jail */)
1545	key = ucred_cache.xid;
1546	#endif /* !__FreeBSD__ */
1547	}
1548	#endif /* !USERSPACE */
1549	else
1550	break;
1551	match = ipfw_lookup_table(chain,
1552	cmd->arg1, keylen, pkey, &vidx);
1553	if (!match)
1554	break;
1555	tablearg = vidx;
1556	break;
1557	}
1558	/* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1559	/* FALLTHROUGH */
1560	}
1561	case O_IP_SRC_LOOKUP:
1562	{
1563	void *pkey;
1564	uint32_t vidx;
1565	uint16_t keylen;
1566
1567	if (is_ipv4) {
1568	keylen = sizeof(in_addr_t);
1569	if (cmd->opcode == O_IP_DST_LOOKUP)
1570	pkey = &dst_ip;
1571	else
1572	pkey = &src_ip;
1573	} else if (is_ipv6) {
1574	keylen = sizeof(struct in6_addr);
1575	if (cmd->opcode == O_IP_DST_LOOKUP)
1576	pkey = &args->f_id.dst_ip6;
1577	else
1578	pkey = &args->f_id.src_ip6;
1579	} else
1580	break;
1581	match = ipfw_lookup_table(chain, cmd->arg1,
1582	keylen, pkey, &vidx);
1583	if (!match)
1584	break;
1585	if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
1586	match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1587	TARG_VAL(chain, vidx, tag);
1588	if (!match)
1589	break;
1590	}
1591	tablearg = vidx;
1592	break;
1593	}
1594
1595	case O_IP_FLOW_LOOKUP:
1596	{
1597	uint32_t v = 0;
1598	match = ipfw_lookup_table(chain,
1599	cmd->arg1, 0, &args->f_id, &v);
1600	if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1601	match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1602	TARG_VAL(chain, v, tag);
1603	if (match)
1604	tablearg = v;
1605	}
1606	break;
1607	case O_IP_SRC_MASK:
1608	case O_IP_DST_MASK:
1609	if (is_ipv4) {
1610	uint32_t a =
1611	(cmd->opcode == O_IP_DST_MASK) ?
1612	dst_ip.s_addr : src_ip.s_addr;
1613	uint32_t p = ((ipfw_insn_u32 )cmd)->d;
1614	int i = cmdlen-1;
1615
1616	for (; !match && i>0; i-= 2, p+= 2)
1617	match = (p[0] == (a & p[1]));
1618	}
1619	break;
1620
1621	case O_IP_SRC_ME:
1622	if (is_ipv4) {
1623	struct ifnet *tif;
1624
1625	INADDR_TO_IFP(src_ip, tif);
1626	match = (tif != NULL);
1627	break;
1628	}
1629	#ifdef INET6
1630	/* FALLTHROUGH */
1631	case O_IP6_SRC_ME:
1632	match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
1633	#endif
1634	break;
1635
1636	case O_IP_DST_SET:
1637	case O_IP_SRC_SET:
1638	if (is_ipv4) {
1639	u_int32_t d = (u_int32_t )(cmd+1);
1640	u_int32_t addr =
1641	cmd->opcode == O_IP_DST_SET ?
1642	args->f_id.dst_ip :
1643	args->f_id.src_ip;
1644
1645	if (addr < d[0])
1646	break;
1647	addr -= d[0]; /* subtract base */
1648	match = (addr < cmd->arg1) &&
1649	( d[ 1 + (addr>>5)] &
1650	(1<<(addr & 0x1f)) );
1651	}
1652	break;
1653
1654	case O_IP_DST:
1655	match = is_ipv4 &&
1656	(((ipfw_insn_ip *)cmd)->addr.s_addr ==
1657	dst_ip.s_addr);
1658	break;
1659
1660	case O_IP_DST_ME:
1661	if (is_ipv4) {
1662	struct ifnet *tif;
1663
1664	INADDR_TO_IFP(dst_ip, tif);
1665	match = (tif != NULL);
1666	break;
1667	}
1668	#ifdef INET6
1669	/* FALLTHROUGH */
1670	case O_IP6_DST_ME:
1671	match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
1672	#endif
1673	break;
1674
1675
1676	case O_IP_SRCPORT:
1677	case O_IP_DSTPORT:
1678	/*
1679	* offset == 0 && proto != 0 is enough
1680	* to guarantee that we have a
1681	* packet with port info.
1682	*/
1683	if ((proto==IPPROTO_UDP \|\| proto==IPPROTO_TCP)
1684	&& offset == 0) {
1685	u_int16_t x =
1686	(cmd->opcode == O_IP_SRCPORT) ?
1687	src_port : dst_port ;
1688	u_int16_t *p =
1689	((ipfw_insn_u16 *)cmd)->ports;
1690	int i;
1691
1692	for (i = cmdlen - 1; !match && i>0;
1693	i--, p += 2)
1694	match = (x>=p[0] && x<=p[1]);
1695	}
1696	break;
1697
1698	case O_ICMPTYPE:
1699	match = (offset == 0 && proto==IPPROTO_ICMP &&
1700	icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1701	break;
1702
1703	#ifdef INET6
1704	case O_ICMP6TYPE:
1705	match = is_ipv6 && offset == 0 &&
1706	proto==IPPROTO_ICMPV6 &&
1707	icmp6type_match(
1708	ICMP6(ulp)->icmp6_type,
1709	(ipfw_insn_u32 *)cmd);
1710	break;
1711	#endif /* INET6 */
1712
1713	case O_IPOPT:
1714	match = (is_ipv4 &&
1715	ipopts_match(ip, cmd) );
1716	break;
1717
1718	case O_IPVER:
1719	match = (is_ipv4 &&
1720	cmd->arg1 == ip->ip_v);
1721	break;
1722
1723	case O_IPID:
1724	case O_IPLEN:
1725	case O_IPTTL:
1726	if (is_ipv4) { /* only for IP packets */
1727	uint16_t x;
1728	uint16_t *p;
1729	int i;
1730
1731	if (cmd->opcode == O_IPLEN)
1732	x = iplen;
1733	else if (cmd->opcode == O_IPTTL)
1734	x = ip->ip_ttl;
1735	else /* must be IPID */
1736	x = ntohs(ip->ip_id);
1737	if (cmdlen == 1) {
1738	match = (cmd->arg1 == x);
1739	break;
1740	}
1741	/* otherwise we have ranges */
1742	p = ((ipfw_insn_u16 *)cmd)->ports;
1743	i = cmdlen - 1;
1744	for (; !match && i>0; i--, p += 2)
1745	match = (x >= p[0] && x <= p[1]);
1746	}
1747	break;
1748
1749	case O_IPPRECEDENCE:
1750	match = (is_ipv4 &&
1751	(cmd->arg1 == (ip->ip_tos & 0xe0)) );
1752	break;
1753
1754	case O_IPTOS:
1755	match = (is_ipv4 &&
1756	flags_match(cmd, ip->ip_tos));
1757	break;
1758
1759	case O_DSCP:
1760	{
1761	uint32_t *p;
1762	uint16_t x;
1763
1764	p = ((ipfw_insn_u32 *)cmd)->d;
1765
1766	if (is_ipv4)
1767	x = ip->ip_tos >> 2;
1768	else if (is_ipv6) {
1769	uint8_t *v;
1770	v = &((struct ip6_hdr *)ip)->ip6_vfc;
1771	x = (*v & 0x0F) << 2;
1772	v++;
1773	x \|= *v >> 6;
1774	} else
1775	break;
1776
1777	/* DSCP bitmask is stored as low_u32 high_u32 */
1778	if (x >= 32)
1779	match = *(p + 1) & (1 << (x - 32));
1780	else
1781	match = *p & (1 << x);
1782	}
1783	break;
1784
1785	case O_TCPDATALEN:
1786	if (proto == IPPROTO_TCP && offset == 0) {
1787	struct tcphdr *tcp;
1788	uint16_t x;
1789	uint16_t *p;
1790	int i;
1791
1792	tcp = TCP(ulp);
1793	x = iplen -
1794	((ip->ip_hl + tcp->th_off) << 2);
1795	if (cmdlen == 1) {
1796	match = (cmd->arg1 == x);
1797	break;
1798	}
1799	/* otherwise we have ranges */
1800	p = ((ipfw_insn_u16 *)cmd)->ports;
1801	i = cmdlen - 1;
1802	for (; !match && i>0; i--, p += 2)
1803	match = (x >= p[0] && x <= p[1]);
1804	}
1805	break;
1806
1807	case O_TCPFLAGS:
1808	match = (proto == IPPROTO_TCP && offset == 0 &&
1809	flags_match(cmd, TCP(ulp)->th_flags));
1810	break;
1811
1812	case O_TCPOPTS:
1813	if (proto == IPPROTO_TCP && offset == 0 && ulp){
1814	PULLUP_LEN(hlen, ulp,
1815	(TCP(ulp)->th_off << 2));
1816	match = tcpopts_match(TCP(ulp), cmd);
1817	}
1818	break;
1819
1820	case O_TCPSEQ:
1821	match = (proto == IPPROTO_TCP && offset == 0 &&
1822	((ipfw_insn_u32 *)cmd)->d[0] ==
1823	TCP(ulp)->th_seq);
1824	break;
1825
1826	case O_TCPACK:
1827	match = (proto == IPPROTO_TCP && offset == 0 &&
1828	((ipfw_insn_u32 *)cmd)->d[0] ==
1829	TCP(ulp)->th_ack);
1830	break;
1831
1832	case O_TCPWIN:
1833	if (proto == IPPROTO_TCP && offset == 0) {
1834	uint16_t x;
1835	uint16_t *p;
1836	int i;
1837
1838	x = ntohs(TCP(ulp)->th_win);
1839	if (cmdlen == 1) {
1840	match = (cmd->arg1 == x);
1841	break;
1842	}
1843	/* Otherwise we have ranges. */
1844	p = ((ipfw_insn_u16 *)cmd)->ports;
1845	i = cmdlen - 1;
1846	for (; !match && i > 0; i--, p += 2)
1847	match = (x >= p[0] && x <= p[1]);
1848	}
1849	break;
1850
1851	case O_ESTAB:
1852	/* reject packets which have SYN only */
1853	/* XXX should i also check for TH_ACK ? */
1854	match = (proto == IPPROTO_TCP && offset == 0 &&
1855	(TCP(ulp)->th_flags &
1856	(TH_RST \| TH_ACK \| TH_SYN)) != TH_SYN);
1857	break;
1858
1859	case O_ALTQ: {
1860	struct pf_mtag *at;
1861	struct m_tag *mtag;
1862	ipfw_insn_altq altq = (ipfw_insn_altq )cmd;
1863
1864	/*
1865	* ALTQ uses mbuf tags from another
1866	* packet filtering system - pf(4).
1867	* We allocate a tag in its format
1868	* and fill it in, pretending to be pf(4).
1869	*/
1870	match = 1;
1871	at = pf_find_mtag(m);
1872	if (at != NULL && at->qid != 0)
1873	break;
1874	mtag = m_tag_get(PACKET_TAG_PF,
1875	sizeof(struct pf_mtag), M_NOWAIT \| M_ZERO);
1876	if (mtag == NULL) {
1877	/*
1878	* Let the packet fall back to the
1879	* default ALTQ.
1880	*/
1881	break;
1882	}
1883	m_tag_prepend(m, mtag);
1884	at = (struct pf_mtag *)(mtag + 1);
1885	at->qid = altq->qid;
1886	at->hdr = ip;
1887	break;
1888	}
1889
1890	case O_LOG:
1891	ipfw_log(chain, f, hlen, args, m,
1892	oif, offset \| ip6f_mf, tablearg, ip);
1893	match = 1;
1894	break;
1895
1896	case O_PROB:
1897	match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1898	break;
1899
1900	case O_VERREVPATH:
1901	/* Outgoing packets automatically pass/match */
1902	match = ((oif != NULL) \|\|
1903	(m->m_pkthdr.rcvif == NULL) \|\|
1904	(
1905	#ifdef INET6
1906	is_ipv6 ?
1907	verify_path6(&(args->f_id.src_ip6),
1908	m->m_pkthdr.rcvif, args->f_id.fib) :
1909	#endif
1910	verify_path(src_ip, m->m_pkthdr.rcvif,
1911	args->f_id.fib)));
1912	break;
1913
1914	case O_VERSRCREACH:
1915	/* Outgoing packets automatically pass/match */
1916	match = (hlen > 0 && ((oif != NULL) \|\|
1917	#ifdef INET6
1918	is_ipv6 ?
1919	verify_path6(&(args->f_id.src_ip6),
1920	NULL, args->f_id.fib) :
1921	#endif
1922	verify_path(src_ip, NULL, args->f_id.fib)));
1923	break;
1924
1925	case O_ANTISPOOF:
1926	/* Outgoing packets automatically pass/match */
1927	if (oif == NULL && hlen > 0 &&
1928	( (is_ipv4 && in_localaddr(src_ip))
1929	#ifdef INET6
1930	\|\| (is_ipv6 &&
1931	in6_localaddr(&(args->f_id.src_ip6)))
1932	#endif
1933	))
1934	match =
1935	#ifdef INET6
1936	is_ipv6 ? verify_path6(
1937	&(args->f_id.src_ip6),
1938	m->m_pkthdr.rcvif,
1939	args->f_id.fib) :
1940	#endif
1941	verify_path(src_ip,
1942	m->m_pkthdr.rcvif,
1943	args->f_id.fib);
1944	else
1945	match = 1;
1946	break;
1947
1948	case O_IPSEC:
1949	#ifdef IPSEC
1950	match = (m_tag_find(m,
1951	PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1952	#endif
1953	/* otherwise no match */
1954	break;
1955
1956	#ifdef INET6
1957	case O_IP6_SRC:
1958	match = is_ipv6 &&
1959	IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1960	&((ipfw_insn_ip6 *)cmd)->addr6);
1961	break;
1962
1963	case O_IP6_DST:
1964	match = is_ipv6 &&
1965	IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1966	&((ipfw_insn_ip6 *)cmd)->addr6);
1967	break;
1968	case O_IP6_SRC_MASK:
1969	case O_IP6_DST_MASK:
1970	if (is_ipv6) {
1971	int i = cmdlen - 1;
1972	struct in6_addr p;
1973	struct in6_addr *d =
1974	&((ipfw_insn_ip6 *)cmd)->addr6;
1975
1976	for (; !match && i > 0; d += 2,
1977	i -= F_INSN_SIZE(struct in6_addr)
1978	* 2) {
1979	p = (cmd->opcode ==
1980	O_IP6_SRC_MASK) ?
1981	args->f_id.src_ip6:
1982	args->f_id.dst_ip6;
1983	APPLY_MASK(&p, &d[1]);
1984	match =
1985	IN6_ARE_ADDR_EQUAL(&d[0],
1986	&p);
1987	}
1988	}
1989	break;
1990
1991	case O_FLOW6ID:
1992	match = is_ipv6 &&
1993	flow6id_match(args->f_id.flow_id6,
1994	(ipfw_insn_u32 *) cmd);
1995	break;
1996
1997	case O_EXT_HDR:
1998	match = is_ipv6 &&
1999	(ext_hd & ((ipfw_insn *) cmd)->arg1);
2000	break;
2001
2002	case O_IP6:
2003	match = is_ipv6;
2004	break;
2005	#endif
2006
2007	case O_IP4:
2008	match = is_ipv4;
2009	break;
2010
2011	case O_TAG: {
2012	struct m_tag *mtag;
2013	uint32_t tag = TARG(cmd->arg1, tag);
2014
2015	/* Packet is already tagged with this tag? */
2016	mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2017
2018	/* We have `untag' action when F_NOT flag is
2019	* present. And we must remove this mtag from
2020	* mbuf and reset `match' to zero (`match' will
2021	* be inversed later).
2022	* Otherwise we should allocate new mtag and
2023	* push it into mbuf.
2024	*/
2025	if (cmd->len & F_NOT) { /* `untag' action */
2026	if (mtag != NULL)
2027	m_tag_delete(m, mtag);
2028	match = 0;
2029	} else {
2030	if (mtag == NULL) {
2031	mtag = m_tag_alloc( MTAG_IPFW,
2032	tag, 0, M_NOWAIT);
2033	if (mtag != NULL)
2034	m_tag_prepend(m, mtag);
2035	}
2036	match = 1;
2037	}
2038	break;
2039	}
2040
2041	case O_FIB: /* try match the specified fib */
2042	if (args->f_id.fib == cmd->arg1)
2043	match = 1;
2044	break;
2045
2046	case O_SOCKARG: {
2047	#ifndef USERSPACE /* not supported in userspace */
2048	struct inpcb *inp = args->inp;
2049	struct inpcbinfo *pi;
2050
2051	if (is_ipv6) /* XXX can we remove this ? */
2052	break;
2053
2054	if (proto == IPPROTO_TCP)
2055	pi = &V_tcbinfo;
2056	else if (proto == IPPROTO_UDP)
2057	pi = &V_udbinfo;
2058	else
2059	break;
2060
2061	/*
2062	* XXXRW: so_user_cookie should almost
2063	* certainly be inp_user_cookie?
2064	*/
2065
2066	/* For incoming packet, lookup up the
2067	inpcb using the src/dest ip/port tuple */
2068	if (inp == NULL) {
2069	inp = in_pcblookup(pi,
2070	src_ip, htons(src_port),
2071	dst_ip, htons(dst_port),
2072	INPLOOKUP_RLOCKPCB, NULL);
2073	if (inp != NULL) {
2074	tablearg =
2075	inp->inp_socket->so_user_cookie;
2076	if (tablearg)
2077	match = 1;
2078	INP_RUNLOCK(inp);
2079	}
2080	} else {
2081	if (inp->inp_socket) {
2082	tablearg =
2083	inp->inp_socket->so_user_cookie;
2084	if (tablearg)
2085	match = 1;
2086	}
2087	}
2088	#endif /* !USERSPACE */
2089	break;
2090	}
2091
2092	case O_TAGGED: {
2093	struct m_tag *mtag;
2094	uint32_t tag = TARG(cmd->arg1, tag);
2095
2096	if (cmdlen == 1) {
2097	match = m_tag_locate(m, MTAG_IPFW,
2098	tag, NULL) != NULL;
2099	break;
2100	}
2101
2102	/* we have ranges */
2103	for (mtag = m_tag_first(m);
2104	mtag != NULL && !match;
2105	mtag = m_tag_next(m, mtag)) {
2106	uint16_t *p;
2107	int i;
2108
2109	if (mtag->m_tag_cookie != MTAG_IPFW)
2110	continue;
2111
2112	p = ((ipfw_insn_u16 *)cmd)->ports;
2113	i = cmdlen - 1;
2114	for(; !match && i > 0; i--, p += 2)
2115	match =
2116	mtag->m_tag_id >= p[0] &&
2117	mtag->m_tag_id <= p[1];
2118	}
2119	break;
2120	}
2121
2122	/*
2123	* The second set of opcodes represents 'actions',
2124	* i.e. the terminal part of a rule once the packet
2125	* matches all previous patterns.
2126	* Typically there is only one action for each rule,
2127	* and the opcode is stored at the end of the rule
2128	* (but there are exceptions -- see below).
2129	*
2130	* In general, here we set retval and terminate the
2131	* outer loop (would be a 'break 3' in some language,
2132	* but we need to set l=0, done=1)
2133	*
2134	* Exceptions:
2135	* O_COUNT and O_SKIPTO actions:
2136	* instead of terminating, we jump to the next rule
2137	* (setting l=0), or to the SKIPTO target (setting
2138	* f/f_len, cmd and l as needed), respectively.
2139	*
2140	* O_TAG, O_LOG and O_ALTQ action parameters:
2141	* perform some action and set match = 1;
2142	*
2143	* O_LIMIT and O_KEEP_STATE: these opcodes are
2144	* not real 'actions', and are stored right
2145	* before the 'action' part of the rule.
2146	* These opcodes try to install an entry in the
2147	* state tables; if successful, we continue with
2148	* the next opcode (match=1; break;), otherwise
2149	* the packet must be dropped (set retval,
2150	* break loops with l=0, done=1)
2151	*
2152	* O_PROBE_STATE and O_CHECK_STATE: these opcodes
2153	* cause a lookup of the state table, and a jump
2154	* to the 'action' part of the parent rule
2155	* if an entry is found, or
2156	* (CHECK_STATE only) a jump to the next rule if
2157	* the entry is not found.
2158	* The result of the lookup is cached so that
2159	* further instances of these opcodes become NOPs.
2160	* The jump to the next rule is done by setting
2161	* l=0, cmdlen=0.
2162	*/
2163	case O_LIMIT:
2164	case O_KEEP_STATE:
2165	if (ipfw_install_state(chain, f,
2166	(ipfw_insn_limit *)cmd, args, tablearg)) {
2167	/* error or limit violation */
2168	retval = IP_FW_DENY;
2169	l = 0; /* exit inner loop */
2170	done = 1; /* exit outer loop */
2171	}
2172	match = 1;
2173	break;
2174
2175	case O_PROBE_STATE:
2176	case O_CHECK_STATE:
2177	/*
2178	* dynamic rules are checked at the first
2179	* keep-state or check-state occurrence,
2180	* with the result being stored in dyn_dir
2181	* and dyn_name.
2182	* The compiler introduces a PROBE_STATE
2183	* instruction for us when we have a
2184	* KEEP_STATE (because PROBE_STATE needs
2185	* to be run first).
2186	*
2187	* (dyn_dir == MATCH_UNKNOWN) means this is
2188	* first lookup for such f_id. Do lookup.
2189	*
2190	* (dyn_dir != MATCH_UNKNOWN &&
2191	* dyn_name != 0 && dyn_name != cmd->arg1)
2192	* means previous lookup didn't find dynamic
2193	* rule for specific state name and current
2194	* lookup will search rule with another state
2195	* name. Redo lookup.
2196	*
2197	* (dyn_dir != MATCH_UNKNOWN && dyn_name == 0)
2198	* means previous lookup was for `any' name
2199	* and it didn't find rule. No need to do
2200	* lookup again.
2201	*/
2202	if ((dyn_dir == MATCH_UNKNOWN \|\|
2203	(dyn_name != 0 &&
2204	dyn_name != cmd->arg1)) &&
2205	(q = ipfw_lookup_dyn_rule(&args->f_id,
2206	&dyn_dir, proto == IPPROTO_TCP ?
2207	TCP(ulp): NULL,
2208	(dyn_name = cmd->arg1))) != NULL) {
2209	/*
2210	* Found dynamic entry, update stats
2211	* and jump to the 'action' part of
2212	* the parent rule by setting
2213	* f, cmd, l and clearing cmdlen.
2214	*/
2215	IPFW_INC_DYN_COUNTER(q, pktlen);
2216	/* XXX we would like to have f_pos
2217	* readily accessible in the dynamic
2218	* rule, instead of having to
2219	* lookup q->rule.
2220	*/
2221	f = q->rule;
2222	f_pos = ipfw_find_rule(chain,
2223	f->rulenum, f->id);
2224	cmd = ACTION_PTR(f);
2225	l = f->cmd_len - f->act_ofs;
2226	ipfw_dyn_unlock(q);
2227	cmdlen = 0;
2228	match = 1;
2229	break;
2230	}
2231	/*
2232	* Dynamic entry not found. If CHECK_STATE,
2233	* skip to next rule, if PROBE_STATE just
2234	* ignore and continue with next opcode.
2235	*/
2236	if (cmd->opcode == O_CHECK_STATE)
2237	l = 0; /* exit inner loop */
2238	match = 1;
2239	break;
2240
2241	case O_ACCEPT:
2242	retval = 0; /* accept */
2243	l = 0; /* exit inner loop */
2244	done = 1; /* exit outer loop */
2245	break;
2246
2247	case O_PIPE:
2248	case O_QUEUE:
2249	set_match(args, f_pos, chain);
2250	args->rule.info = TARG(cmd->arg1, pipe);
2251	if (cmd->opcode == O_PIPE)
2252	args->rule.info \|= IPFW_IS_PIPE;
2253	if (V_fw_one_pass)
2254	args->rule.info \|= IPFW_ONEPASS;
2255	retval = IP_FW_DUMMYNET;
2256	l = 0; /* exit inner loop */
2257	done = 1; /* exit outer loop */
2258	break;
2259
2260	case O_DIVERT:
2261	case O_TEE:
2262	if (args->eh) /* not on layer 2 */
2263	break;
2264	/* otherwise this is terminal */
2265	l = 0; /* exit inner loop */
2266	done = 1; /* exit outer loop */
2267	retval = (cmd->opcode == O_DIVERT) ?
2268	IP_FW_DIVERT : IP_FW_TEE;
2269	set_match(args, f_pos, chain);
2270	args->rule.info = TARG(cmd->arg1, divert);
2271	break;
2272
2273	case O_COUNT:
2274	IPFW_INC_RULE_COUNTER(f, pktlen);
2275	l = 0; /* exit inner loop */
2276	break;
2277
2278	case O_SKIPTO:
2279	IPFW_INC_RULE_COUNTER(f, pktlen);
2280	f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2281	/*
2282	* Skip disabled rules, and re-enter
2283	* the inner loop with the correct
2284	* f_pos, f, l and cmd.
2285	* Also clear cmdlen and skip_or
2286	*/
2287	for (; f_pos < chain->n_rules - 1 &&
2288	(V_set_disable &
2289	(1 << chain->map[f_pos]->set));
2290	f_pos++)
2291	;
2292	/* Re-enter the inner loop at the skipto rule. */
2293	f = chain->map[f_pos];
2294	l = f->cmd_len;
2295	cmd = f->cmd;
2296	match = 1;
2297	cmdlen = 0;
2298	skip_or = 0;
2299	continue;
2300	break; /* not reached */
2301
2302	case O_CALLRETURN: {
2303	/*
2304	* Implementation of `subroutine' call/return,
2305	* in the stack carried in an mbuf tag. This
2306	* is different from `skipto' in that any call
2307	* address is possible (`skipto' must prevent
2308	* backward jumps to avoid endless loops).
2309	* We have `return' action when F_NOT flag is
2310	* present. The `m_tag_id' field is used as
2311	* stack pointer.
2312	*/
2313	struct m_tag *mtag;
2314	uint16_t jmpto, *stack;
2315
2316	#define IS_CALL ((cmd->len & F_NOT) == 0)
2317	#define IS_RETURN ((cmd->len & F_NOT) != 0)
2318	/*
2319	* Hand-rolled version of m_tag_locate() with
2320	* wildcard `type'.
2321	* If not already tagged, allocate new tag.
2322	*/
2323	mtag = m_tag_first(m);
2324	while (mtag != NULL) {
2325	if (mtag->m_tag_cookie ==
2326	MTAG_IPFW_CALL)
2327	break;
2328	mtag = m_tag_next(m, mtag);
2329	}
2330	if (mtag == NULL && IS_CALL) {
2331	mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2332	IPFW_CALLSTACK_SIZE *
2333	sizeof(uint16_t), M_NOWAIT);
2334	if (mtag != NULL)
2335	m_tag_prepend(m, mtag);
2336	}
2337
2338	/*
2339	* On error both `call' and `return' just
2340	* continue with next rule.
2341	*/
2342	if (IS_RETURN && (mtag == NULL \|\|
2343	mtag->m_tag_id == 0)) {
2344	l = 0; /* exit inner loop */
2345	break;
2346	}
2347	if (IS_CALL && (mtag == NULL \|\|
2348	mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2349	printf("ipfw: call stack error, "
2350	"go to next rule\n");
2351	l = 0; /* exit inner loop */
2352	break;
2353	}
2354
2355	IPFW_INC_RULE_COUNTER(f, pktlen);
2356	stack = (uint16_t *)(mtag + 1);
2357
2358	/*
2359	* The `call' action may use cached f_pos
2360	* (in f->next_rule), whose version is written
2361	* in f->next_rule.
2362	* The `return' action, however, doesn't have
2363	* fixed jump address in cmd->arg1 and can't use
2364	* cache.
2365	*/
2366	if (IS_CALL) {
2367	stack[mtag->m_tag_id] = f->rulenum;
2368	mtag->m_tag_id++;
2369	f_pos = JUMP(chain, f, cmd->arg1,
2370	tablearg, 1);
2371	} else { /* `return' action */
2372	mtag->m_tag_id--;
2373	jmpto = stack[mtag->m_tag_id] + 1;
2374	f_pos = ipfw_find_rule(chain, jmpto, 0);
2375	}
2376
2377	/*
2378	* Skip disabled rules, and re-enter
2379	* the inner loop with the correct
2380	* f_pos, f, l and cmd.
2381	* Also clear cmdlen and skip_or
2382	*/
2383	for (; f_pos < chain->n_rules - 1 &&
2384	(V_set_disable &
2385	(1 << chain->map[f_pos]->set)); f_pos++)
2386	;
2387	/* Re-enter the inner loop at the dest rule. */
2388	f = chain->map[f_pos];
2389	l = f->cmd_len;
2390	cmd = f->cmd;
2391	cmdlen = 0;
2392	skip_or = 0;
2393	continue;
2394	break; /* NOTREACHED */
2395	}
2396	#undef IS_CALL
2397	#undef IS_RETURN
2398
2399	case O_REJECT:
2400	/*
2401	* Drop the packet and send a reject notice
2402	* if the packet is not ICMP (or is an ICMP
2403	* query), and it is not multicast/broadcast.
2404	*/
2405	if (hlen > 0 && is_ipv4 && offset == 0 &&
2406	(proto != IPPROTO_ICMP \|\|
2407	is_icmp_query(ICMP(ulp))) &&
2408	!(m->m_flags & (M_BCAST\|M_MCAST)) &&
2409	!IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2410	send_reject(args, cmd->arg1, iplen, ip);
2411	m = args->m;
2412	}
2413	/* FALLTHROUGH */
2414	#ifdef INET6
2415	case O_UNREACH6:
2416	if (hlen > 0 && is_ipv6 &&
2417	((offset & IP6F_OFF_MASK) == 0) &&
2418	(proto != IPPROTO_ICMPV6 \|\|
2419	(is_icmp6_query(icmp6_type) == 1)) &&
2420	!(m->m_flags & (M_BCAST\|M_MCAST)) &&
2421	!IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2422	send_reject6(
2423	args, cmd->arg1, hlen,
2424	(struct ip6_hdr *)ip);
2425	m = args->m;
2426	}
2427	/* FALLTHROUGH */
2428	#endif
2429	case O_DENY:
2430	retval = IP_FW_DENY;
2431	l = 0; /* exit inner loop */
2432	done = 1; /* exit outer loop */
2433	break;
2434
2435	case O_FORWARD_IP:
2436	if (args->eh) /* not valid on layer2 pkts */
2437	break;
2438	if (q == NULL \|\| q->rule != f \|\|
2439	dyn_dir == MATCH_FORWARD) {
2440	struct sockaddr_in *sa;
2441
2442	sa = &(((ipfw_insn_sa *)cmd)->sa);
2443	if (sa->sin_addr.s_addr == INADDR_ANY) {
2444	#ifdef INET6
2445	/*
2446	* We use O_FORWARD_IP opcode for
2447	* fwd rule with tablearg, but tables
2448	* now support IPv6 addresses. And
2449	* when we are inspecting IPv6 packet,
2450	* we can use nh6 field from
2451	* table_value as next_hop6 address.
2452	*/
2453	if (is_ipv6) {
2454	struct sockaddr_in6 *sa6;
2455
2456	sa6 = args->next_hop6 =
2457	&args->hopstore6;
2458	sa6->sin6_family = AF_INET6;
2459	sa6->sin6_len = sizeof(*sa6);
2460	sa6->sin6_addr = TARG_VAL(
2461	chain, tablearg, nh6);
2462	/*
2463	* Set sin6_scope_id only for
2464	* link-local unicast addresses.
2465	*/
2466	if (IN6_IS_ADDR_LINKLOCAL(
2467	&sa6->sin6_addr))
2468	sa6->sin6_scope_id =
2469	TARG_VAL(chain,
2470	tablearg,
2471	zoneid);
2472	} else
2473	#endif
2474	{
2475	sa = args->next_hop =
2476	&args->hopstore;
2477	sa->sin_family = AF_INET;
2478	sa->sin_len = sizeof(*sa);
2479	sa->sin_addr.s_addr = htonl(
2480	TARG_VAL(chain, tablearg,
2481	nh4));
2482	}
2483	} else {
2484	args->next_hop = sa;
2485	}
2486	}
2487	retval = IP_FW_PASS;
2488	l = 0; /* exit inner loop */
2489	done = 1; /* exit outer loop */
2490	break;
2491
2492	#ifdef INET6
2493	case O_FORWARD_IP6:
2494	if (args->eh) /* not valid on layer2 pkts */
2495	break;
2496	if (q == NULL \|\| q->rule != f \|\|
2497	dyn_dir == MATCH_FORWARD) {
2498	struct sockaddr_in6 *sin6;
2499
2500	sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2501	args->next_hop6 = sin6;
2502	}
2503	retval = IP_FW_PASS;
2504	l = 0; /* exit inner loop */
2505	done = 1; /* exit outer loop */
2506	break;
2507	#endif
2508
2509	case O_NETGRAPH:
2510	case O_NGTEE:
2511	set_match(args, f_pos, chain);
2512	args->rule.info = TARG(cmd->arg1, netgraph);
2513	if (V_fw_one_pass)
2514	args->rule.info \|= IPFW_ONEPASS;
2515	retval = (cmd->opcode == O_NETGRAPH) ?
2516	IP_FW_NETGRAPH : IP_FW_NGTEE;
2517	l = 0; /* exit inner loop */
2518	done = 1; /* exit outer loop */
2519	break;
2520
2521	case O_SETFIB: {
2522	uint32_t fib;
2523
2524	IPFW_INC_RULE_COUNTER(f, pktlen);
2525	fib = TARG(cmd->arg1, fib) & 0x7FFF;
2526	if (fib >= rt_numfibs)
2527	fib = 0;
2528	M_SETFIB(m, fib);
2529	args->f_id.fib = fib;
2530	l = 0; /* exit inner loop */
2531	break;
2532	}
2533
2534	case O_SETDSCP: {
2535	uint16_t code;
2536
2537	code = TARG(cmd->arg1, dscp) & 0x3F;
2538	l = 0; /* exit inner loop */
2539	if (is_ipv4) {
2540	uint16_t old;
2541
2542	old = (uint16_t )ip;
2543	ip->ip_tos = (code << 2) \|
2544	(ip->ip_tos & 0x03);
2545	ip->ip_sum = cksum_adjust(ip->ip_sum,
2546	old, (uint16_t )ip);
2547	} else if (is_ipv6) {
2548	uint8_t *v;
2549
2550	v = &((struct ip6_hdr *)ip)->ip6_vfc;
2551	v = (v & 0xF0) \| (code >> 2);
2552	v++;
2553	v = (v & 0x3F) \| ((code & 0x03) << 6);
2554	} else
2555	break;
2556
2557	IPFW_INC_RULE_COUNTER(f, pktlen);
2558	break;
2559	}
2560
2561	case O_NAT:
2562	l = 0; /* exit inner loop */
2563	done = 1; /* exit outer loop */
2564	if (!IPFW_NAT_LOADED) {
2565	retval = IP_FW_DENY;
2566	break;
2567	}
2568
2569	struct cfg_nat *t;
2570	int nat_id;
2571
2572	set_match(args, f_pos, chain);
2573	/* Check if this is 'global' nat rule */
2574	if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
2575	retval = ipfw_nat_ptr(args, NULL, m);
2576	break;
2577	}
2578	t = ((ipfw_insn_nat *)cmd)->nat;
2579	if (t == NULL) {
2580	nat_id = TARG(cmd->arg1, nat);
2581	t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2582
2583	if (t == NULL) {
2584	retval = IP_FW_DENY;
2585	break;
2586	}
2587	if (cmd->arg1 != IP_FW_TARG)
2588	((ipfw_insn_nat *)cmd)->nat = t;
2589	}
2590	retval = ipfw_nat_ptr(args, t, m);
2591	break;
2592
2593	case O_REASS: {
2594	int ip_off;
2595
2596	IPFW_INC_RULE_COUNTER(f, pktlen);
2597	l = 0; /* in any case exit inner loop */
2598	ip_off = ntohs(ip->ip_off);
2599
2600	/* if not fragmented, go to next rule */
2601	if ((ip_off & (IP_MF \| IP_OFFMASK)) == 0)
2602	break;
2603
2604	args->m = m = ip_reass(m);
2605
2606	/*
2607	* do IP header checksum fixup.
2608	*/
2609	if (m == NULL) { /* fragment got swallowed */
2610	retval = IP_FW_DENY;
2611	} else { /* good, packet complete */
2612	int hlen;
2613
2614	ip = mtod(m, struct ip *);
2615	hlen = ip->ip_hl << 2;
2616	ip->ip_sum = 0;
2617	if (hlen == sizeof(struct ip))
2618	ip->ip_sum = in_cksum_hdr(ip);
2619	else
2620	ip->ip_sum = in_cksum(m, hlen);
2621	retval = IP_FW_REASS;
2622	set_match(args, f_pos, chain);
2623	}
2624	done = 1; /* exit outer loop */
2625	break;
2626	}
2627	case O_EXTERNAL_ACTION:
2628	l = 0; /* in any case exit inner loop */
2629	retval = ipfw_run_eaction(chain, args,
2630	cmd, &done);
2631	/*
2632	* If both @retval and @done are zero,
2633	* consider this as rule matching and
2634	* update counters.
2635	*/
2636	if (retval == 0 && done == 0) {
2637	IPFW_INC_RULE_COUNTER(f, pktlen);
2638	/*
2639	* Reset the result of the last
2640	* dynamic state lookup.
2641	* External action can change
2642	* @args content, and it may be
2643	* used for new state lookup later.
2644	*/
2645	dyn_dir = MATCH_UNKNOWN;
2646	}
2647	break;
2648
2649	default:
2650	panic("-- unknown opcode %d\n", cmd->opcode);
2651	} /* end of switch() on opcodes */
2652	/*
2653	* if we get here with l=0, then match is irrelevant.
2654	*/
2655
2656	if (cmd->len & F_NOT)
2657	match = !match;
2658
2659	if (match) {
2660	if (cmd->len & F_OR)
2661	skip_or = 1;
2662	} else {
2663	if (!(cmd->len & F_OR)) /* not an OR block, */
2664	break; /* try next rule */
2665	}
2666
2667	} /* end of inner loop, scan opcodes */
2668	#undef PULLUP_LEN
2669
2670	if (done)
2671	break;
2672
2673	/* next_rule:; / / try next rule */
2674
2675	} /* end of outer for, scan rules */
2676
2677	if (done) {
2678	struct ip_fw *rule = chain->map[f_pos];
2679	/* Update statistics */
2680	IPFW_INC_RULE_COUNTER(rule, pktlen);
2681	} else {
2682	retval = IP_FW_DENY;
2683	printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2684	}
2685	IPFW_PF_RUNLOCK(chain);
2686	#ifdef __FreeBSD__
2687	if (ucred_cache != NULL)
2688	crfree(ucred_cache);
2689	#endif
2690	return (retval);
2691
2692	pullup_failed:
2693	if (V_fw_verbose)
2694	printf("ipfw: pullup failed\n");
2695	return (IP_FW_DENY);
2696	}
2697
2698	/*
2699	* Set maximum number of tables that can be used in given VNET ipfw instance.
2700	*/
2701	#ifdef SYSCTL_NODE
2702	static int
2703	sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2704	{
2705	int error;
2706	unsigned int ntables;
2707
2708	ntables = V_fw_tables_max;
2709
2710	error = sysctl_handle_int(oidp, &ntables, 0, req);
2711	/* Read operation or some error */
2712	if ((error != 0) \|\| (req->newptr == NULL))
2713	return (error);
2714
2715	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2716	}
2717
2718	/*
2719	* Switches table namespace between global and per-set.
2720	*/
2721	static int
2722	sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
2723	{
2724	int error;
2725	unsigned int sets;
2726
2727	sets = V_fw_tables_sets;
2728
2729	error = sysctl_handle_int(oidp, &sets, 0, req);
2730	/* Read operation or some error */
2731	if ((error != 0) \|\| (req->newptr == NULL))
2732	return (error);
2733
2734	return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
2735	}
2736	#endif
2737
2738	/*
2739	* Module and VNET glue
2740	*/
2741
2742	/*
2743	* Stuff that must be initialised only on boot or module load
2744	*/
2745	static int
2746	ipfw_init(void)
2747	{
2748	int error = 0;
2749
2750	/*
2751	* Only print out this stuff the first time around,
2752	* when called from the sysinit code.
2753	*/
2754	printf("ipfw2 "
2755	#ifdef INET6
2756	"(+ipv6) "
2757	#endif
2758	"initialized, divert %s, nat %s, "
2759	"default to %s, logging ",
2760	#ifdef IPDIVERT
2761	"enabled",
2762	#else
2763	"loadable",
2764	#endif
2765	#ifdef IPFIREWALL_NAT
2766	"enabled",
2767	#else
2768	"loadable",
2769	#endif
2770	default_to_accept ? "accept" : "deny");
2771
2772	/*
2773	* Note: V_xxx variables can be accessed here but the vnet specific
2774	* initializer may not have been called yet for the VIMAGE case.
2775	* Tuneables will have been processed. We will print out values for
2776	* the default vnet.
2777	* XXX This should all be rationalized AFTER 8.0
2778	*/
2779	if (V_fw_verbose == 0)
2780	printf("disabled\n");
2781	else if (V_verbose_limit == 0)
2782	printf("unlimited\n");
2783	else
2784	printf("limited to %d packets/entry by default\n",
2785	V_verbose_limit);
2786
2787	/* Check user-supplied table count for validness */
2788	if (default_fw_tables > IPFW_TABLES_MAX)
2789	default_fw_tables = IPFW_TABLES_MAX;
2790
2791	ipfw_init_sopt_handler();
2792	ipfw_init_obj_rewriter();
2793	ipfw_iface_init();
2794	return (error);
2795	}
2796
2797	#ifndef __rtems__
2798	/*
2799	* Called for the removal of the last instance only on module unload.
2800	*/
2801	static void
2802	ipfw_destroy(void)
2803	{
2804
2805	ipfw_iface_destroy();
2806	ipfw_destroy_sopt_handler();
2807	ipfw_destroy_obj_rewriter();
2808	printf("IP firewall unloaded\n");
2809	}
2810	#endif /* __rtems__ */
2811
2812	/*
2813	* Stuff that must be initialized for every instance
2814	* (including the first of course).
2815	*/
2816	static int
2817	vnet_ipfw_init(const void *unused)
2818	{
2819	int error, first;
2820	struct ip_fw *rule = NULL;
2821	struct ip_fw_chain *chain;
2822
2823	chain = &V_layer3_chain;
2824
2825	first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2826
2827	/* First set up some values that are compile time options */
2828	V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
2829	V_fw_deny_unknown_exthdrs = 1;
2830	#ifdef IPFIREWALL_VERBOSE
2831	V_fw_verbose = 1;
2832	#endif
2833	#ifdef IPFIREWALL_VERBOSE_LIMIT
2834	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2835	#endif
2836	#ifdef IPFIREWALL_NAT
2837	LIST_INIT(&chain->nat);
2838	#endif
2839
2840	/* Init shared services hash table */
2841	ipfw_init_srv(chain);
2842
2843	ipfw_init_counters();
2844	/* insert the default rule and create the initial map */
2845	chain->n_rules = 1;
2846	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK \| M_ZERO);
2847	rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
2848
2849	/* Set initial number of tables */
2850	V_fw_tables_max = default_fw_tables;
2851	error = ipfw_init_tables(chain, first);
2852	if (error) {
2853	printf("ipfw2: setting up tables failed\n");
2854	free(chain->map, M_IPFW);
2855	free(rule, M_IPFW);
2856	return (ENOSPC);
2857	}
2858
2859	/* fill and insert the default rule */
2860	rule->act_ofs = 0;
2861	rule->rulenum = IPFW_DEFAULT_RULE;
2862	rule->cmd_len = 1;
2863	rule->set = RESVD_SET;
2864	rule->cmd[0].len = 1;
2865	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2866	chain->default_rule = chain->map[0] = rule;
2867	chain->id = rule->id = 1;
2868	/* Pre-calculate rules length for legacy dump format */
2869	chain->static_len = sizeof(struct ip_fw_rule0);
2870
2871	IPFW_LOCK_INIT(chain);
2872	ipfw_dyn_init(chain);
2873	ipfw_eaction_init(chain, first);
2874	#ifdef LINEAR_SKIPTO
2875	ipfw_init_skipto_cache(chain);
2876	#endif
2877	ipfw_bpf_init(first);
2878
2879	/* First set up some values that are compile time options */
2880	V_ipfw_vnet_ready = 1; /* Open for business */
2881
2882	/*
2883	* Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2884	* Even if the latter two fail we still keep the module alive
2885	* because the sockopt and layer2 paths are still useful.
2886	* ipfw[6]_hook return 0 on success, ENOENT on failure,
2887	* so we can ignore the exact return value and just set a flag.
2888	*
2889	* Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2890	* changes in the underlying (per-vnet) variables trigger
2891	* immediate hook()/unhook() calls.
2892	* In layer2 we have the same behaviour, except that V_ether_ipfw
2893	* is checked on each packet because there are no pfil hooks.
2894	*/
2895	V_ip_fw_ctl_ptr = ipfw_ctl3;
2896	error = ipfw_attach_hooks(1);
2897	return (error);
2898	}
2899
2900	#ifndef __rtems__
2901	/*
2902	* Called for the removal of each instance.
2903	*/
2904	static int
2905	vnet_ipfw_uninit(const void *unused)
2906	{
2907	struct ip_fw *reap;
2908	struct ip_fw_chain *chain = &V_layer3_chain;
2909	int i, last;
2910
2911	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2912	/*
2913	* disconnect from ipv4, ipv6, layer2 and sockopt.
2914	* Then grab, release and grab again the WLOCK so we make
2915	* sure the update is propagated and nobody will be in.
2916	*/
2917	(void)ipfw_attach_hooks(0 /* detach */);
2918	V_ip_fw_ctl_ptr = NULL;
2919
2920	last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
2921
2922	IPFW_UH_WLOCK(chain);
2923	IPFW_UH_WUNLOCK(chain);
2924
2925	ipfw_dyn_uninit(0); /* run the callout_drain */
2926
2927	IPFW_UH_WLOCK(chain);
2928
2929	reap = NULL;
2930	IPFW_WLOCK(chain);
2931	for (i = 0; i < chain->n_rules; i++)
2932	ipfw_reap_add(chain, &reap, chain->map[i]);
2933	free(chain->map, M_IPFW);
2934	#ifdef LINEAR_SKIPTO
2935	ipfw_destroy_skipto_cache(chain);
2936	#endif
2937	IPFW_WUNLOCK(chain);
2938	IPFW_UH_WUNLOCK(chain);
2939	ipfw_destroy_tables(chain, last);
2940	ipfw_eaction_uninit(chain, last);
2941	if (reap != NULL)
2942	ipfw_reap_rules(reap);
2943	vnet_ipfw_iface_destroy(chain);
2944	ipfw_destroy_srv(chain);
2945	IPFW_LOCK_DESTROY(chain);
2946	ipfw_dyn_uninit(1); /* free the remaining parts */
2947	ipfw_destroy_counters();
2948	ipfw_bpf_uninit(last);
2949	return (0);
2950	}
2951	#endif /* __rtems__ */
2952
2953	/*
2954	* Module event handler.
2955	* In general we have the choice of handling most of these events by the
2956	* event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2957	* use the SYSINIT handlers as they are more capable of expressing the
2958	* flow of control during module and vnet operations, so this is just
2959	* a skeleton. Note there is no SYSINIT equivalent of the module
2960	* SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2961	*/
2962	static int
2963	ipfw_modevent(module_t mod, int type, void *unused)
2964	{
2965	int err = 0;
2966
2967	switch (type) {
2968	case MOD_LOAD:
2969	/* Called once at module load or
2970	* system boot if compiled in. */
2971	break;
2972	case MOD_QUIESCE:
2973	/* Called before unload. May veto unloading. */
2974	break;
2975	case MOD_UNLOAD:
2976	/* Called during unload. */
2977	break;
2978	case MOD_SHUTDOWN:
2979	/* Called during system shutdown. */
2980	break;
2981	default:
2982	err = EOPNOTSUPP;
2983	break;
2984	}
2985	return err;
2986	}
2987
2988	static moduledata_t ipfwmod = {
2989	"ipfw",
2990	ipfw_modevent,
2991	0
2992	};
2993
2994	/* Define startup order. */
2995	#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL
2996	#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
2997	#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2998	#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2999
3000	DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3001	FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3002	MODULE_VERSION(ipfw, 3);
3003	/* should declare some dependencies here */
3004
3005	/*
3006	* Starting up. Done in order after ipfwmod() has been called.
3007	* VNET_SYSINIT is also called for each existing vnet and each new vnet.
3008	*/
3009	SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3010	ipfw_init, NULL);
3011	VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3012	vnet_ipfw_init, NULL);
3013
3014	/*
3015	* Closing up shop. These are done in REVERSE ORDER, but still
3016	* after ipfwmod() has been called. Not called on reboot.
3017	* VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3018	* or when the module is unloaded.
3019	*/
3020	SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3021	ipfw_destroy, NULL);
3022	VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3023	vnet_ipfw_uninit, NULL);
3024	/* end of file */

Note: See TracBrowser for help on using the repository browser.

Download in other formats: