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, const ipfw_insn *cmd, int iplen,
997 struct ip *ip)
998 {
999 int code, mtu;
1000
1001 code = cmd->arg1;
1002 if (code == ICMP_UNREACH_NEEDFRAG &&
1003 cmd->len == F_INSN_SIZE(ipfw_insn_u16))
1004 mtu = ((const ipfw_insn_u16 *)cmd)->ports[0];
1005 else
1006 mtu = 0;
1007
1008 #if 0
1009 /* XXX When ip is not guaranteed to be at mtod() we will
1010 * need to account for this */
1011 * The mbuf will however be thrown away so we can adjust it.
1012 * Remember we did an m_pullup on it already so we
1013 * can make some assumptions about contiguousness.
1014 */
1015 if (args->L3offset)
1016 m_adj(m, args->L3offset);
1017 #endif
1018 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
1019 /* Send an ICMP unreach */
1020 icmp_error(args->m, ICMP_UNREACH, code, 0L, mtu);
1021 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
1022 struct tcphdr *const tcp =
1023 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1024 if ( (tcp->th_flags & TH_RST) == 0) {
1025 struct mbuf *m;
1026 m = ipfw_send_pkt(args->m, &(args->f_id),
1027 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1028 tcp->th_flags | TH_RST);
1029 if (m != NULL)
1030 ip_output(m, NULL, NULL, 0, NULL, NULL);
1031 }
1032 FREE_PKT(args->m);
1033 } else if (code == ICMP_REJECT_ABORT &&
1034 args->f_id.proto == IPPROTO_SCTP) {
1035 struct mbuf *m;
1036 struct sctphdr *sctp;
1037 struct sctp_chunkhdr *chunk;
1038 struct sctp_init *init;
1039 u_int32_t v_tag;
1040 int reflected;
1041
1042 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1043 reflected = 1;
1044 v_tag = ntohl(sctp->v_tag);
1045 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1046 sizeof(struct sctp_chunkhdr)) {
1047 /* Look at the first chunk header if available */
1048 chunk = (struct sctp_chunkhdr *)(sctp + 1);
1049 switch (chunk->chunk_type) {
1050 case SCTP_INITIATION:
1051 /*
1052 * Packets containing an INIT chunk MUST have
1053 * a zero v-tag.
1054 */
1055 if (v_tag != 0) {
1056 v_tag = 0;
1057 break;
1058 }
1059 /* INIT chunk MUST NOT be bundled */
1060 if (iplen >
1061 (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1062 ntohs(chunk->chunk_length) + 3) {
1063 break;
1064 }
1065 /* Use the initiate tag if available */
1066 if ((iplen >= (ip->ip_hl << 2) +
1067 sizeof(struct sctphdr) +
1068 sizeof(struct sctp_chunkhdr) +
1069 offsetof(struct sctp_init, a_rwnd))) {
1070 init = (struct sctp_init *)(chunk + 1);
1071 v_tag = ntohl(init->initiate_tag);
1072 reflected = 0;
1073 }
1074 break;
1075 case SCTP_ABORT_ASSOCIATION:
1076 /*
1077 * If the packet contains an ABORT chunk, don't
1078 * reply.
1079 * XXX: We should search through all chunks,
1080 * but do not do that to avoid attacks.
1081 */
1082 v_tag = 0;
1083 break;
1084 }
1085 }
1086 if (v_tag == 0) {
1087 m = NULL;
1088 } else {
1089 m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1090 reflected);
1091 }
1092 if (m != NULL)
1093 ip_output(m, NULL, NULL, 0, NULL, NULL);
1094 FREE_PKT(args->m);
1095 } else
1096 FREE_PKT(args->m);
1097 args->m = NULL;
1098 }
1099
1100 /*
1101 * Support for uid/gid/jail lookup. These tests are expensive
1102 * (because we may need to look into the list of active sockets)
1103 * so we cache the results. ugid_lookupp is 0 if we have not
1104 * yet done a lookup, 1 if we succeeded, and -1 if we tried
1105 * and failed. The function always returns the match value.
1106 * We could actually spare the variable and use *uc, setting
1107 * it to '(void *)check_uidgid if we have no info, NULL if
1108 * we tried and failed, or any other value if successful.
1109 */
1110 static int
1111 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1112 struct ucred **uc)
1113 {
1114 #if defined(USERSPACE)
1115 return 0; // not supported in userspace
1116 #else
1117 #ifndef __FreeBSD__
1118 /* XXX */
1119 return cred_check(insn, proto, oif,
1120 dst_ip, dst_port, src_ip, src_port,
1121 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1122 #else /* FreeBSD */
1123 struct in_addr src_ip, dst_ip;
1124 struct inpcbinfo *pi;
1125 struct ipfw_flow_id *id;
1126 struct inpcb *pcb, *inp;
1127 int lookupflags;
1128 int match;
1129
1130 id = &args->f_id;
1131 inp = args->inp;
1132
1133 /*
1134 * Check to see if the UDP or TCP stack supplied us with
1135 * the PCB. If so, rather then holding a lock and looking
1136 * up the PCB, we can use the one that was supplied.
1137 */
1138 if (inp && *ugid_lookupp == 0) {
1139 INP_LOCK_ASSERT(inp);
1140 if (inp->inp_socket != NULL) {
1141 *uc = crhold(inp->inp_cred);
1142 *ugid_lookupp = 1;
1143 } else
1144 *ugid_lookupp = -1;
1145 }
1146 /*
1147 * If we have already been here and the packet has no
1148 * PCB entry associated with it, then we can safely
1149 * assume that this is a no match.
1150 */
1151 if (*ugid_lookupp == -1)
1152 return (0);
1153 if (id->proto == IPPROTO_TCP) {
1154 lookupflags = 0;
1155 pi = &V_tcbinfo;
1156 } else if (id->proto == IPPROTO_UDP) {
1157 lookupflags = INPLOOKUP_WILDCARD;
1158 pi = &V_udbinfo;
1159 } else if (id->proto == IPPROTO_UDPLITE) {
1160 lookupflags = INPLOOKUP_WILDCARD;
1161 pi = &V_ulitecbinfo;
1162 } else
1163 return 0;
1164 lookupflags |= INPLOOKUP_RLOCKPCB;
1165 match = 0;
1166 if (*ugid_lookupp == 0) {
1167 if (id->addr_type == 6) {
1168 #ifdef INET6
1169 if (args->flags & IPFW_ARGS_IN)
1170 pcb = in6_pcblookup_mbuf(pi,
1171 &id->src_ip6, htons(id->src_port),
1172 &id->dst_ip6, htons(id->dst_port),
1173 lookupflags, NULL, args->m);
1174 else
1175 pcb = in6_pcblookup_mbuf(pi,
1176 &id->dst_ip6, htons(id->dst_port),
1177 &id->src_ip6, htons(id->src_port),
1178 lookupflags, args->ifp, args->m);
1179 #else
1180 *ugid_lookupp = -1;
1181 return (0);
1182 #endif
1183 } else {
1184 src_ip.s_addr = htonl(id->src_ip);
1185 dst_ip.s_addr = htonl(id->dst_ip);
1186 if (args->flags & IPFW_ARGS_IN)
1187 pcb = in_pcblookup_mbuf(pi,
1188 src_ip, htons(id->src_port),
1189 dst_ip, htons(id->dst_port),
1190 lookupflags, NULL, args->m);
1191 else
1192 pcb = in_pcblookup_mbuf(pi,
1193 dst_ip, htons(id->dst_port),
1194 src_ip, htons(id->src_port),
1195 lookupflags, args->ifp, args->m);
1196 }
1197 if (pcb != NULL) {
1198 INP_RLOCK_ASSERT(pcb);
1199 *uc = crhold(pcb->inp_cred);
1200 *ugid_lookupp = 1;
1201 INP_RUNLOCK(pcb);
1202 }
1203 if (*ugid_lookupp == 0) {
1204 /*
1205 * We tried and failed, set the variable to -1
1206 * so we will not try again on this packet.
1207 */
1208 *ugid_lookupp = -1;
1209 return (0);
1210 }
1211 }
1212 if (insn->o.opcode == O_UID)
1213 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1214 else if (insn->o.opcode == O_GID)
1215 match = groupmember((gid_t)insn->d[0], *uc);
1216 else if (insn->o.opcode == O_JAIL)
1217 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1218 return (match);
1219 #endif /* __FreeBSD__ */
1220 #endif /* not supported in userspace */
1221 }
1222
1223 /*
1224 * Helper function to set args with info on the rule after the matching
1225 * one. slot is precise, whereas we guess rule_id as they are
1226 * assigned sequentially.
1227 */
1228 static inline void
1229 set_match(struct ip_fw_args *args, int slot,
1230 struct ip_fw_chain *chain)
1231 {
1232 args->rule.chain_id = chain->id;
1233 args->rule.slot = slot + 1; /* we use 0 as a marker */
1234 args->rule.rule_id = 1 + chain->map[slot]->id;
1235 args->rule.rulenum = chain->map[slot]->rulenum;
1236 args->flags |= IPFW_ARGS_REF;
1237 }
1238
1239 static int
1240 jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1241 int tablearg, int jump_backwards)
1242 {
1243 int f_pos, i;
1244
1245 i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1246 /* make sure we do not jump backward */
1247 if (jump_backwards == 0 && i <= f->rulenum)
1248 i = f->rulenum + 1;
1249
1250 #ifndef LINEAR_SKIPTO
1251 if (chain->idxmap != NULL)
1252 f_pos = chain->idxmap[i];
1253 else
1254 f_pos = ipfw_find_rule(chain, i, 0);
1255 #else
1256 f_pos = chain->idxmap[i];
1257 #endif /* LINEAR_SKIPTO */
1258
1259 return (f_pos);
1260 }
1261
1262
1263 #ifndef LINEAR_SKIPTO
1264 /*
1265 * Helper function to enable cached rule lookups using
1266 * cache.id and cache.pos fields in ipfw rule.
1267 */
1268 static int
1269 jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1270 int tablearg, int jump_backwards)
1271 {
1272 int f_pos;
1273
1274 /* Can't use cache with IP_FW_TARG */
1275 if (num == IP_FW_TARG)
1276 return jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1277
1278 /*
1279 * If possible use cached f_pos (in f->cache.pos),
1280 * whose version is written in f->cache.id (horrible hacks
1281 * to avoid changing the ABI).
1282 *
1283 * Multiple threads can execute the same rule simultaneously,
1284 * we need to ensure that cache.pos is updated before cache.id.
1285 */
1286
1287 #ifdef __LP64__
1288 struct ip_fw_jump_cache cache;
1289
1290 cache.raw_value = f->cache.raw_value;
1291 if (cache.id == chain->id)
1292 return (cache.pos);
1293
1294 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1295
1296 cache.pos = f_pos;
1297 cache.id = chain->id;
1298 f->cache.raw_value = cache.raw_value;
1299 #else
1300 if (f->cache.id == chain->id) {
1301 /* Load pos after id */
1302 atomic_thread_fence_acq();
1303 return (f->cache.pos);
1304 }
1305
1306 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards);
1307
1308 f->cache.pos = f_pos;
1309 /* Store id after pos */
1310 atomic_thread_fence_rel();
1311 f->cache.id = chain->id;
1312 #endif /* !__LP64__ */
1313 return (f_pos);
1314 }
1315 #endif /* !LINEAR_SKIPTO */
1316
1317 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f)
1318 /*
1319 * The main check routine for the firewall.
1320 *
1321 * All arguments are in args so we can modify them and return them
1322 * back to the caller.
1323 *
1324 * Parameters:
1325 *
1326 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1327 * Starts with the IP header.
1328 * args->L3offset Number of bytes bypassed if we came from L2.
1329 * e.g. often sizeof(eh) ** NOTYET **
1330 * args->ifp Incoming or outgoing interface.
1331 * args->divert_rule (in/out)
1332 * Skip up to the first rule past this rule number;
1333 * upon return, non-zero port number for divert or tee.
1334 *
1335 * args->rule Pointer to the last matching rule (in/out)
1336 * args->next_hop Socket we are forwarding to (out).
1337 * args->next_hop6 IPv6 next hop we are forwarding to (out).
1338 * args->f_id Addresses grabbed from the packet (out)
1339 * args->rule.info a cookie depending on rule action
1340 *
1341 * Return value:
1342 *
1343 * IP_FW_PASS the packet must be accepted
1344 * IP_FW_DENY the packet must be dropped
1345 * IP_FW_DIVERT divert packet, port in m_tag
1346 * IP_FW_TEE tee packet, port in m_tag
1347 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
1348 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
1349 * args->rule contains the matching rule,
1350 * args->rule.info has additional information.
1351 *
1352 */
1353 int
1354 ipfw_chk(struct ip_fw_args *args)
1355 {
1356
1357 /*
1358 * Local variables holding state while processing a packet:
1359 *
1360 * IMPORTANT NOTE: to speed up the processing of rules, there
1361 * are some assumption on the values of the variables, which
1362 * are documented here. Should you change them, please check
1363 * the implementation of the various instructions to make sure
1364 * that they still work.
1365 *
1366 * m | args->m Pointer to the mbuf, as received from the caller.
1367 * It may change if ipfw_chk() does an m_pullup, or if it
1368 * consumes the packet because it calls send_reject().
1369 * XXX This has to change, so that ipfw_chk() never modifies
1370 * or consumes the buffer.
1371 * OR
1372 * args->mem Pointer to contigous memory chunk.
1373 * ip Is the beginning of the ip(4 or 6) header.
1374 * eh Ethernet header in case if input is Layer2.
1375 */
1376 struct mbuf *m;
1377 struct ip *ip;
1378 struct ether_header *eh;
1379
1380 /*
1381 * For rules which contain uid/gid or jail constraints, cache
1382 * a copy of the users credentials after the pcb lookup has been
1383 * executed. This will speed up the processing of rules with
1384 * these types of constraints, as well as decrease contention
1385 * on pcb related locks.
1386 */
1387 #ifndef __FreeBSD__
1388 struct bsd_ucred ucred_cache;
1389 #else
1390 struct ucred *ucred_cache = NULL;
1391 #endif
1392 int ucred_lookup = 0;
1393 int f_pos = 0; /* index of current rule in the array */
1394 int retval = 0;
1395 struct ifnet *oif, *iif;
1396
1397 /*
1398 * hlen The length of the IP header.
1399 */
1400 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1401
1402 /*
1403 * offset The offset of a fragment. offset != 0 means that
1404 * we have a fragment at this offset of an IPv4 packet.
1405 * offset == 0 means that (if this is an IPv4 packet)
1406 * this is the first or only fragment.
1407 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1408 * or there is a single packet fragment (fragment header added
1409 * without needed). We will treat a single packet fragment as if
1410 * there was no fragment header (or log/block depending on the
1411 * V_fw_permit_single_frag6 sysctl setting).
1412 */
1413 u_short offset = 0;
1414 u_short ip6f_mf = 0;
1415
1416 /*
1417 * Local copies of addresses. They are only valid if we have
1418 * an IP packet.
1419 *
1420 * proto The protocol. Set to 0 for non-ip packets,
1421 * or to the protocol read from the packet otherwise.
1422 * proto != 0 means that we have an IPv4 packet.
1423 *
1424 * src_port, dst_port port numbers, in HOST format. Only
1425 * valid for TCP and UDP packets.
1426 *
1427 * src_ip, dst_ip ip addresses, in NETWORK format.
1428 * Only valid for IPv4 packets.
1429 */
1430 uint8_t proto;
1431 uint16_t src_port, dst_port; /* NOTE: host format */
1432 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1433 int iplen = 0;
1434 int pktlen;
1435
1436 struct ipfw_dyn_info dyn_info;
1437 struct ip_fw *q = NULL;
1438 struct ip_fw_chain *chain = &V_layer3_chain;
1439
1440 /*
1441 * We store in ulp a pointer to the upper layer protocol header.
1442 * In the ipv4 case this is easy to determine from the header,
1443 * but for ipv6 we might have some additional headers in the middle.
1444 * ulp is NULL if not found.
1445 */
1446 void *ulp = NULL; /* upper layer protocol pointer. */
1447
1448 /* XXX ipv6 variables */
1449 int is_ipv6 = 0;
1450 #ifdef INET6
1451 uint8_t icmp6_type = 0;
1452 #endif
1453 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
1454 /* end of ipv6 variables */
1455
1456 int is_ipv4 = 0;
1457
1458 int done = 0; /* flag to exit the outer loop */
1459 IPFW_RLOCK_TRACKER;
1460 bool mem;
1461
1462 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1463 if (args->flags & IPFW_ARGS_ETHER) {
1464 eh = (struct ether_header *)args->mem;
1465 if (eh->ether_type == htons(ETHERTYPE_VLAN))
1466 ip = (struct ip *)
1467 ((struct ether_vlan_header *)eh + 1);
1468 else
1469 ip = (struct ip *)(eh + 1);
1470 } else {
1471 eh = NULL;
1472 ip = (struct ip *)args->mem;
1473 }
1474 pktlen = IPFW_ARGS_LENGTH(args->flags);
1475 args->f_id.fib = args->ifp->if_fib; /* best guess */
1476 } else {
1477 m = args->m;
1478 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1479 return (IP_FW_PASS); /* accept */
1480 if (args->flags & IPFW_ARGS_ETHER) {
1481 /* We need some amount of data to be contiguous. */
1482 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1483 (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1484 max_protohdr))) == NULL)
1485 goto pullup_failed;
1486 eh = mtod(m, struct ether_header *);
1487 ip = (struct ip *)(eh + 1);
1488 } else {
1489 eh = NULL;
1490 ip = mtod(m, struct ip *);
1491 }
1492 pktlen = m->m_pkthdr.len;
1493 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1494 }
1495
1496 dst_ip.s_addr = 0; /* make sure it is initialized */
1497 src_ip.s_addr = 0; /* make sure it is initialized */
1498 src_port = dst_port = 0;
1499
1500 DYN_INFO_INIT(&dyn_info);
1501 /*
1502 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1503 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1504 * pointer might become stale after other pullups (but we never use it
1505 * this way).
1506 */
1507 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
1508 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0)
1509 #define _PULLUP_LOCKED(_len, p, T, unlock) \
1510 do { \
1511 int x = (_len) + T + EHLEN; \
1512 if (mem) { \
1513 if (__predict_false(pktlen < x)) { \
1514 unlock; \
1515 goto pullup_failed; \
1516 } \
1517 p = (char *)args->mem + (_len) + EHLEN; \
1518 } else { \
1519 if (__predict_false((m)->m_len < x)) { \
1520 args->m = m = m_pullup(m, x); \
1521 if (m == NULL) { \
1522 unlock; \
1523 goto pullup_failed; \
1524 } \
1525 } \
1526 p = mtod(m, char *) + (_len) + EHLEN; \
1527 } \
1528 } while (0)
1529
1530 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, )
1531 #define PULLUP_LEN_LOCKED(_len, p, T) \
1532 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \
1533 UPDATE_POINTERS()
1534 /*
1535 * In case pointers got stale after pullups, update them.
1536 */
1537 #define UPDATE_POINTERS() \
1538 do { \
1539 if (!mem) { \
1540 if (eh != NULL) { \
1541 eh = mtod(m, struct ether_header *); \
1542 ip = (struct ip *)(eh + 1); \
1543 } else \
1544 ip = mtod(m, struct ip *); \
1545 args->m = m; \
1546 } \
1547 } while (0)
1548
1549 /* Identify IP packets and fill up variables. */
1550 if (pktlen >= sizeof(struct ip6_hdr) &&
1551 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1552 ip->ip_v == 6) {
1553 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1554
1555 is_ipv6 = 1;
1556 args->flags |= IPFW_ARGS_IP6;
1557 hlen = sizeof(struct ip6_hdr);
1558 proto = ip6->ip6_nxt;
1559 /* Search extension headers to find upper layer protocols */
1560 while (ulp == NULL && offset == 0) {
1561 switch (proto) {
1562 case IPPROTO_ICMPV6:
1563 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1564 #ifdef INET6
1565 icmp6_type = ICMP6(ulp)->icmp6_type;
1566 #endif
1567 break;
1568
1569 case IPPROTO_TCP:
1570 PULLUP_TO(hlen, ulp, struct tcphdr);
1571 dst_port = TCP(ulp)->th_dport;
1572 src_port = TCP(ulp)->th_sport;
1573 /* save flags for dynamic rules */
1574 args->f_id._flags = TCP(ulp)->th_flags;
1575 break;
1576
1577 case IPPROTO_SCTP:
1578 if (pktlen >= hlen + sizeof(struct sctphdr) +
1579 sizeof(struct sctp_chunkhdr) +
1580 offsetof(struct sctp_init, a_rwnd))
1581 PULLUP_LEN(hlen, ulp,
1582 sizeof(struct sctphdr) +
1583 sizeof(struct sctp_chunkhdr) +
1584 offsetof(struct sctp_init, a_rwnd));
1585 else if (pktlen >= hlen + sizeof(struct sctphdr))
1586 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1587 else
1588 PULLUP_LEN(hlen, ulp,
1589 sizeof(struct sctphdr));
1590 src_port = SCTP(ulp)->src_port;
1591 dst_port = SCTP(ulp)->dest_port;
1592 break;
1593
1594 case IPPROTO_UDP:
1595 case IPPROTO_UDPLITE:
1596 PULLUP_TO(hlen, ulp, struct udphdr);
1597 dst_port = UDP(ulp)->uh_dport;
1598 src_port = UDP(ulp)->uh_sport;
1599 break;
1600
1601 case IPPROTO_HOPOPTS: /* RFC 2460 */
1602 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1603 ext_hd |= EXT_HOPOPTS;
1604 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1605 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1606 ulp = NULL;
1607 break;
1608
1609 case IPPROTO_ROUTING: /* RFC 2460 */
1610 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1611 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1612 case 0:
1613 ext_hd |= EXT_RTHDR0;
1614 break;
1615 case 2:
1616 ext_hd |= EXT_RTHDR2;
1617 break;
1618 default:
1619 if (V_fw_verbose)
1620 printf("IPFW2: IPV6 - Unknown "
1621 "Routing Header type(%d)\n",
1622 ((struct ip6_rthdr *)
1623 ulp)->ip6r_type);
1624 if (V_fw_deny_unknown_exthdrs)
1625 return (IP_FW_DENY);
1626 break;
1627 }
1628 ext_hd |= EXT_ROUTING;
1629 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1630 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1631 ulp = NULL;
1632 break;
1633
1634 case IPPROTO_FRAGMENT: /* RFC 2460 */
1635 PULLUP_TO(hlen, ulp, struct ip6_frag);
1636 ext_hd |= EXT_FRAGMENT;
1637 hlen += sizeof (struct ip6_frag);
1638 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1639 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1640 IP6F_OFF_MASK;
1641 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1642 IP6F_MORE_FRAG;
1643 if (V_fw_permit_single_frag6 == 0 &&
1644 offset == 0 && ip6f_mf == 0) {
1645 if (V_fw_verbose)
1646 printf("IPFW2: IPV6 - Invalid "
1647 "Fragment Header\n");
1648 if (V_fw_deny_unknown_exthdrs)
1649 return (IP_FW_DENY);
1650 break;
1651 }
1652 args->f_id.extra =
1653 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1654 ulp = NULL;
1655 break;
1656
1657 case IPPROTO_DSTOPTS: /* RFC 2460 */
1658 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1659 ext_hd |= EXT_DSTOPTS;
1660 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1661 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1662 ulp = NULL;
1663 break;
1664
1665 case IPPROTO_AH: /* RFC 2402 */
1666 PULLUP_TO(hlen, ulp, struct ip6_ext);
1667 ext_hd |= EXT_AH;
1668 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1669 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1670 ulp = NULL;
1671 break;
1672
1673 case IPPROTO_ESP: /* RFC 2406 */
1674 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1675 /* Anything past Seq# is variable length and
1676 * data past this ext. header is encrypted. */
1677 ext_hd |= EXT_ESP;
1678 break;
1679
1680 case IPPROTO_NONE: /* RFC 2460 */
1681 /*
1682 * Packet ends here, and IPv6 header has
1683 * already been pulled up. If ip6e_len!=0
1684 * then octets must be ignored.
1685 */
1686 ulp = ip; /* non-NULL to get out of loop. */
1687 break;
1688
1689 case IPPROTO_OSPFIGP:
1690 /* XXX OSPF header check? */
1691 PULLUP_TO(hlen, ulp, struct ip6_ext);
1692 break;
1693
1694 case IPPROTO_PIM:
1695 /* XXX PIM header check? */
1696 PULLUP_TO(hlen, ulp, struct pim);
1697 break;
1698
1699 case IPPROTO_GRE: /* RFC 1701 */
1700 /* XXX GRE header check? */
1701 PULLUP_TO(hlen, ulp, struct grehdr);
1702 break;
1703
1704 case IPPROTO_CARP:
1705 PULLUP_TO(hlen, ulp, offsetof(
1706 struct carp_header, carp_counter));
1707 if (CARP_ADVERTISEMENT !=
1708 ((struct carp_header *)ulp)->carp_type)
1709 return (IP_FW_DENY);
1710 break;
1711
1712 case IPPROTO_IPV6: /* RFC 2893 */
1713 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1714 break;
1715
1716 case IPPROTO_IPV4: /* RFC 2893 */
1717 PULLUP_TO(hlen, ulp, struct ip);
1718 break;
1719
1720 default:
1721 if (V_fw_verbose)
1722 printf("IPFW2: IPV6 - Unknown "
1723 "Extension Header(%d), ext_hd=%x\n",
1724 proto, ext_hd);
1725 if (V_fw_deny_unknown_exthdrs)
1726 return (IP_FW_DENY);
1727 PULLUP_TO(hlen, ulp, struct ip6_ext);
1728 break;
1729 } /*switch */
1730 }
1731 UPDATE_POINTERS();
1732 ip6 = (struct ip6_hdr *)ip;
1733 args->f_id.addr_type = 6;
1734 args->f_id.src_ip6 = ip6->ip6_src;
1735 args->f_id.dst_ip6 = ip6->ip6_dst;
1736 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1737 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1738 } else if (pktlen >= sizeof(struct ip) &&
1739 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1740 ip->ip_v == 4) {
1741 is_ipv4 = 1;
1742 args->flags |= IPFW_ARGS_IP4;
1743 hlen = ip->ip_hl << 2;
1744 /*
1745 * Collect parameters into local variables for faster
1746 * matching.
1747 */
1748 proto = ip->ip_p;
1749 src_ip = ip->ip_src;
1750 dst_ip = ip->ip_dst;
1751 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1752 iplen = ntohs(ip->ip_len);
1753
1754 if (offset == 0) {
1755 switch (proto) {
1756 case IPPROTO_TCP:
1757 PULLUP_TO(hlen, ulp, struct tcphdr);
1758 dst_port = TCP(ulp)->th_dport;
1759 src_port = TCP(ulp)->th_sport;
1760 /* save flags for dynamic rules */
1761 args->f_id._flags = TCP(ulp)->th_flags;
1762 break;
1763
1764 case IPPROTO_SCTP:
1765 if (pktlen >= hlen + sizeof(struct sctphdr) +
1766 sizeof(struct sctp_chunkhdr) +
1767 offsetof(struct sctp_init, a_rwnd))
1768 PULLUP_LEN(hlen, ulp,
1769 sizeof(struct sctphdr) +
1770 sizeof(struct sctp_chunkhdr) +
1771 offsetof(struct sctp_init, a_rwnd));
1772 else if (pktlen >= hlen + sizeof(struct sctphdr))
1773 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1774 else
1775 PULLUP_LEN(hlen, ulp,
1776 sizeof(struct sctphdr));
1777 src_port = SCTP(ulp)->src_port;
1778 dst_port = SCTP(ulp)->dest_port;
1779 break;
1780
1781 case IPPROTO_UDP:
1782 case IPPROTO_UDPLITE:
1783 PULLUP_TO(hlen, ulp, struct udphdr);
1784 dst_port = UDP(ulp)->uh_dport;
1785 src_port = UDP(ulp)->uh_sport;
1786 break;
1787
1788 case IPPROTO_ICMP:
1789 PULLUP_TO(hlen, ulp, struct icmphdr);
1790 //args->f_id.flags = ICMP(ulp)->icmp_type;
1791 break;
1792
1793 default:
1794 break;
1795 }
1796 } else {
1797 if (offset == 1 && proto == IPPROTO_TCP) {
1798 /* RFC 3128 */
1799 goto pullup_failed;
1800 }
1801 }
1802
1803 UPDATE_POINTERS();
1804 args->f_id.addr_type = 4;
1805 args->f_id.src_ip = ntohl(src_ip.s_addr);
1806 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1807 } else {
1808 proto = 0;
1809 dst_ip.s_addr = src_ip.s_addr = 0;
1810
1811 args->f_id.addr_type = 1; /* XXX */
1812 }
1813 #undef PULLUP_TO
1814 pktlen = iplen < pktlen ? iplen: pktlen;
1815
1816 /* Properly initialize the rest of f_id */
1817 args->f_id.proto = proto;
1818 args->f_id.src_port = src_port = ntohs(src_port);
1819 args->f_id.dst_port = dst_port = ntohs(dst_port);
1820
1821 IPFW_PF_RLOCK(chain);
1822 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1823 IPFW_PF_RUNLOCK(chain);
1824 return (IP_FW_PASS); /* accept */
1825 }
1826 if (args->flags & IPFW_ARGS_REF) {
1827 /*
1828 * Packet has already been tagged as a result of a previous
1829 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1830 * REASS, NETGRAPH, DIVERT/TEE...)
1831 * Validate the slot and continue from the next one
1832 * if still present, otherwise do a lookup.
1833 */
1834 f_pos = (args->rule.chain_id == chain->id) ?
1835 args->rule.slot :
1836 ipfw_find_rule(chain, args->rule.rulenum,
1837 args->rule.rule_id);
1838 } else {
1839 f_pos = 0;
1840 }
1841
1842 if (args->flags & IPFW_ARGS_IN) {
1843 iif = args->ifp;
1844 oif = NULL;
1845 } else {
1846 MPASS(args->flags & IPFW_ARGS_OUT);
1847 iif = mem ? NULL : m_rcvif(m);
1848 oif = args->ifp;
1849 }
1850
1851 /*
1852 * Now scan the rules, and parse microinstructions for each rule.
1853 * We have two nested loops and an inner switch. Sometimes we
1854 * need to break out of one or both loops, or re-enter one of
1855 * the loops with updated variables. Loop variables are:
1856 *
1857 * f_pos (outer loop) points to the current rule.
1858 * On output it points to the matching rule.
1859 * done (outer loop) is used as a flag to break the loop.
1860 * l (inner loop) residual length of current rule.
1861 * cmd points to the current microinstruction.
1862 *
1863 * We break the inner loop by setting l=0 and possibly
1864 * cmdlen=0 if we don't want to advance cmd.
1865 * We break the outer loop by setting done=1
1866 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1867 * as needed.
1868 */
1869 for (; f_pos < chain->n_rules; f_pos++) {
1870 ipfw_insn *cmd;
1871 uint32_t tablearg = 0;
1872 int l, cmdlen, skip_or; /* skip rest of OR block */
1873 struct ip_fw *f;
1874
1875 f = chain->map[f_pos];
1876 if (V_set_disable & (1 << f->set) )
1877 continue;
1878
1879 skip_or = 0;
1880 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1881 l -= cmdlen, cmd += cmdlen) {
1882 int match;
1883
1884 /*
1885 * check_body is a jump target used when we find a
1886 * CHECK_STATE, and need to jump to the body of
1887 * the target rule.
1888 */
1889
1890 /* check_body: */
1891 cmdlen = F_LEN(cmd);
1892 /*
1893 * An OR block (insn_1 || .. || insn_n) has the
1894 * F_OR bit set in all but the last instruction.
1895 * The first match will set "skip_or", and cause
1896 * the following instructions to be skipped until
1897 * past the one with the F_OR bit clear.
1898 */
1899 if (skip_or) { /* skip this instruction */
1900 if ((cmd->len & F_OR) == 0)
1901 skip_or = 0; /* next one is good */
1902 continue;
1903 }
1904 match = 0; /* set to 1 if we succeed */
1905
1906 switch (cmd->opcode) {
1907 /*
1908 * The first set of opcodes compares the packet's
1909 * fields with some pattern, setting 'match' if a
1910 * match is found. At the end of the loop there is
1911 * logic to deal with F_NOT and F_OR flags associated
1912 * with the opcode.
1913 */
1914 case O_NOP:
1915 match = 1;
1916 break;
1917
1918 case O_FORWARD_MAC:
1919 printf("ipfw: opcode %d unimplemented\n",
1920 cmd->opcode);
1921 break;
1922
1923 case O_GID:
1924 case O_UID:
1925 case O_JAIL:
1926 /*
1927 * We only check offset == 0 && proto != 0,
1928 * as this ensures that we have a
1929 * packet with the ports info.
1930 */
1931 if (offset != 0)
1932 break;
1933 if (proto == IPPROTO_TCP ||
1934 proto == IPPROTO_UDP ||
1935 proto == IPPROTO_UDPLITE)
1936 match = check_uidgid(
1937 (ipfw_insn_u32 *)cmd,
1938 args, &ucred_lookup,
1939 #ifdef __FreeBSD__
1940 &ucred_cache);
1941 #else
1942 (void *)&ucred_cache);
1943 #endif
1944 break;
1945
1946 case O_RECV:
1947 match = iface_match(iif, (ipfw_insn_if *)cmd,
1948 chain, &tablearg);
1949 break;
1950
1951 case O_XMIT:
1952 match = iface_match(oif, (ipfw_insn_if *)cmd,
1953 chain, &tablearg);
1954 break;
1955
1956 case O_VIA:
1957 match = iface_match(args->ifp,
1958 (ipfw_insn_if *)cmd, chain, &tablearg);
1959 break;
1960
1961 case O_MACADDR2:
1962 if (args->flags & IPFW_ARGS_ETHER) {
1963 u_int32_t *want = (u_int32_t *)
1964 ((ipfw_insn_mac *)cmd)->addr;
1965 u_int32_t *mask = (u_int32_t *)
1966 ((ipfw_insn_mac *)cmd)->mask;
1967 u_int32_t *hdr = (u_int32_t *)eh;
1968
1969 match =
1970 ( want[0] == (hdr[0] & mask[0]) &&
1971 want[1] == (hdr[1] & mask[1]) &&
1972 want[2] == (hdr[2] & mask[2]) );
1973 }
1974 break;
1975
1976 case O_MAC_TYPE:
1977 if (args->flags & IPFW_ARGS_ETHER) {
1978 u_int16_t *p =
1979 ((ipfw_insn_u16 *)cmd)->ports;
1980 int i;
1981
1982 for (i = cmdlen - 1; !match && i>0;
1983 i--, p += 2)
1984 match =
1985 (ntohs(eh->ether_type) >=
1986 p[0] &&
1987 ntohs(eh->ether_type) <=
1988 p[1]);
1989 }
1990 break;
1991
1992 case O_FRAG:
1993 if (is_ipv4) {
1994 /*
1995 * Since flags_match() works with
1996 * uint8_t we pack ip_off into 8 bits.
1997 * For this match offset is a boolean.
1998 */
1999 match = flags_match(cmd,
2000 ((ntohs(ip->ip_off) & ~IP_OFFMASK)
2001 >> 8) | (offset != 0));
2002 } else {
2003 /*
2004 * Compatiblity: historically bare
2005 * "frag" would match IPv6 fragments.
2006 */
2007 match = (cmd->arg1 == 0x1 &&
2008 (offset != 0));
2009 }
2010 break;
2011
2012 case O_IN: /* "out" is "not in" */
2013 match = (oif == NULL);
2014 break;
2015
2016 case O_LAYER2:
2017 match = (args->flags & IPFW_ARGS_ETHER);
2018 break;
2019
2020 case O_DIVERTED:
2021 if ((args->flags & IPFW_ARGS_REF) == 0)
2022 break;
2023 /*
2024 * For diverted packets, args->rule.info
2025 * contains the divert port (in host format)
2026 * reason and direction.
2027 */
2028 match = ((args->rule.info & IPFW_IS_MASK) ==
2029 IPFW_IS_DIVERT) && (
2030 ((args->rule.info & IPFW_INFO_IN) ?
2031 1: 2) & cmd->arg1);
2032 break;
2033
2034 case O_PROTO:
2035 /*
2036 * We do not allow an arg of 0 so the
2037 * check of "proto" only suffices.
2038 */
2039 match = (proto == cmd->arg1);
2040 break;
2041
2042 case O_IP_SRC:
2043 match = is_ipv4 &&
2044 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2045 src_ip.s_addr);
2046 break;
2047
2048 case O_IP_DST_LOOKUP:
2049 {
2050 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
2051 void *pkey;
2052 uint32_t vidx, key;
2053 uint16_t keylen = 0; /* zero if can't match the packet */
2054
2055 /* Determine lookup key type */
2056 vidx = ((ipfw_insn_u32 *)cmd)->d[1];
2057 switch (vidx) {
2058 case LOOKUP_DST_IP:
2059 case LOOKUP_SRC_IP:
2060 /* Need IP frame */
2061 if (is_ipv6 == 0 && is_ipv4 == 0)
2062 break;
2063 if (vidx == LOOKUP_DST_IP)
2064 pkey = is_ipv6 ?
2065 (void *)&args->f_id.dst_ip6:
2066 (void *)&dst_ip;
2067 else
2068 pkey = is_ipv6 ?
2069 (void *)&args->f_id.src_ip6:
2070 (void *)&src_ip;
2071 keylen = is_ipv6 ?
2072 sizeof(struct in6_addr):
2073 sizeof(in_addr_t);
2074 break;
2075 case LOOKUP_DST_PORT:
2076 case LOOKUP_SRC_PORT:
2077 /* Need IP frame */
2078 if (is_ipv6 == 0 && is_ipv4 == 0)
2079 break;
2080 /* Skip fragments */
2081 if (offset != 0)
2082 break;
2083 /* Skip proto without ports */
2084 if (proto != IPPROTO_TCP &&
2085 proto != IPPROTO_UDP &&
2086 proto != IPPROTO_UDPLITE &&
2087 proto != IPPROTO_SCTP)
2088 break;
2089 key = vidx == LOOKUP_DST_PORT ?
2090 dst_port:
2091 src_port;
2092 pkey = &key;
2093 keylen = sizeof(key);
2094 break;
2095 case LOOKUP_UID:
2096 case LOOKUP_JAIL:
2097 check_uidgid(
2098 (ipfw_insn_u32 *)cmd,
2099 args, &ucred_lookup,
2100 &ucred_cache);
2101 key = vidx == LOOKUP_UID ?
2102 ucred_cache->cr_uid:
2103 ucred_cache->cr_prison->pr_id;
2104 pkey = &key;
2105 keylen = sizeof(key);
2106 break;
2107 case LOOKUP_DSCP:
2108 /* Need IP frame */
2109 if (is_ipv6 == 0 && is_ipv4 == 0)
2110 break;
2111 if (is_ipv6)
2112 key = IPV6_DSCP(
2113 (struct ip6_hdr *)ip) >> 2;
2114 else
2115 key = ip->ip_tos >> 2;
2116 pkey = &key;
2117 keylen = sizeof(key);
2118 break;
2119 case LOOKUP_DST_MAC:
2120 case LOOKUP_SRC_MAC:
2121 /* Need ether frame */
2122 if ((args->flags & IPFW_ARGS_ETHER) == 0)
2123 break;
2124 pkey = vidx == LOOKUP_DST_MAC ?
2125 eh->ether_dhost:
2126 eh->ether_shost;
2127 keylen = ETHER_ADDR_LEN;
2128 break;
2129 }
2130 if (keylen == 0)
2131 break;
2132 match = ipfw_lookup_table(chain,
2133 cmd->arg1, keylen, pkey, &vidx);
2134 if (!match)
2135 break;
2136 tablearg = vidx;
2137 break;
2138 }
2139 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2140 /* FALLTHROUGH */
2141 }
2142 case O_IP_SRC_LOOKUP:
2143 {
2144 void *pkey;
2145 uint32_t vidx;
2146 uint16_t keylen;
2147
2148 if (is_ipv4) {
2149 keylen = sizeof(in_addr_t);
2150 if (cmd->opcode == O_IP_DST_LOOKUP)
2151 pkey = &dst_ip;
2152 else
2153 pkey = &src_ip;
2154 } else if (is_ipv6) {
2155 keylen = sizeof(struct in6_addr);
2156 if (cmd->opcode == O_IP_DST_LOOKUP)
2157 pkey = &args->f_id.dst_ip6;
2158 else
2159 pkey = &args->f_id.src_ip6;
2160 } else
2161 break;
2162 match = ipfw_lookup_table(chain, cmd->arg1,
2163 keylen, pkey, &vidx);
2164 if (!match)
2165 break;
2166 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2167 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2168 TARG_VAL(chain, vidx, tag);
2169 if (!match)
2170 break;
2171 }
2172 tablearg = vidx;
2173 break;
2174 }
2175
2176 case O_MAC_SRC_LOOKUP:
2177 case O_MAC_DST_LOOKUP:
2178 {
2179 void *pkey;
2180 uint32_t vidx;
2181 uint16_t keylen = ETHER_ADDR_LEN;
2182
2183 /* Need ether frame */
2184 if ((args->flags & IPFW_ARGS_ETHER) == 0)
2185 break;
2186
2187 if (cmd->opcode == O_MAC_DST_LOOKUP)
2188 pkey = eh->ether_dhost;
2189 else
2190 pkey = eh->ether_shost;
2191
2192 match = ipfw_lookup_table(chain, cmd->arg1,
2193 keylen, pkey, &vidx);
2194 if (!match)
2195 break;
2196 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2197 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2198 TARG_VAL(chain, vidx, tag);
2199 if (!match)
2200 break;
2201 }
2202 tablearg = vidx;
2203 break;
2204 }
2205
2206 case O_IP_FLOW_LOOKUP:
2207 {
2208 uint32_t v = 0;
2209 match = ipfw_lookup_table(chain,
2210 cmd->arg1, 0, &args->f_id, &v);
2211 if (!match)
2212 break;
2213 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2214 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2215 TARG_VAL(chain, v, tag);
2216 if (match)
2217 tablearg = v;
2218 }
2219 break;
2220 case O_IP_SRC_MASK:
2221 case O_IP_DST_MASK:
2222 if (is_ipv4) {
2223 uint32_t a =
2224 (cmd->opcode == O_IP_DST_MASK) ?
2225 dst_ip.s_addr : src_ip.s_addr;
2226 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2227 int i = cmdlen-1;
2228
2229 for (; !match && i>0; i-= 2, p+= 2)
2230 match = (p[0] == (a & p[1]));
2231 }
2232 break;
2233
2234 case O_IP_SRC_ME:
2235 if (is_ipv4) {
2236 match = in_localip(src_ip);
2237 break;
2238 }
2239 #ifdef INET6
2240 /* FALLTHROUGH */
2241 case O_IP6_SRC_ME:
2242 match = is_ipv6 &&
2243 ipfw_localip6(&args->f_id.src_ip6);
2244 #endif
2245 break;
2246
2247 case O_IP_DST_SET:
2248 case O_IP_SRC_SET:
2249 if (is_ipv4) {
2250 u_int32_t *d = (u_int32_t *)(cmd+1);
2251 u_int32_t addr =
2252 cmd->opcode == O_IP_DST_SET ?
2253 args->f_id.dst_ip :
2254 args->f_id.src_ip;
2255
2256 if (addr < d[0])
2257 break;
2258 addr -= d[0]; /* subtract base */
2259 match = (addr < cmd->arg1) &&
2260 ( d[ 1 + (addr>>5)] &
2261 (1<<(addr & 0x1f)) );
2262 }
2263 break;
2264
2265 case O_IP_DST:
2266 match = is_ipv4 &&
2267 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2268 dst_ip.s_addr);
2269 break;
2270
2271 case O_IP_DST_ME:
2272 if (is_ipv4) {
2273 match = in_localip(dst_ip);
2274 break;
2275 }
2276 #ifdef INET6
2277 /* FALLTHROUGH */
2278 case O_IP6_DST_ME:
2279 match = is_ipv6 &&
2280 ipfw_localip6(&args->f_id.dst_ip6);
2281 #endif
2282 break;
2283
2284 case O_IP_SRCPORT:
2285 case O_IP_DSTPORT:
2286 /*
2287 * offset == 0 && proto != 0 is enough
2288 * to guarantee that we have a
2289 * packet with port info.
2290 */
2291 if ((proto == IPPROTO_UDP ||
2292 proto == IPPROTO_UDPLITE ||
2293 proto == IPPROTO_TCP ||
2294 proto == IPPROTO_SCTP) && offset == 0) {
2295 u_int16_t x =
2296 (cmd->opcode == O_IP_SRCPORT) ?
2297 src_port : dst_port ;
2298 u_int16_t *p =
2299 ((ipfw_insn_u16 *)cmd)->ports;
2300 int i;
2301
2302 for (i = cmdlen - 1; !match && i>0;
2303 i--, p += 2)
2304 match = (x>=p[0] && x<=p[1]);
2305 }
2306 break;
2307
2308 case O_ICMPTYPE:
2309 match = (offset == 0 && proto==IPPROTO_ICMP &&
2310 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2311 break;
2312
2313 #ifdef INET6
2314 case O_ICMP6TYPE:
2315 match = is_ipv6 && offset == 0 &&
2316 proto==IPPROTO_ICMPV6 &&
2317 icmp6type_match(
2318 ICMP6(ulp)->icmp6_type,
2319 (ipfw_insn_u32 *)cmd);
2320 break;
2321 #endif /* INET6 */
2322
2323 case O_IPOPT:
2324 match = (is_ipv4 &&
2325 ipopts_match(ip, cmd) );
2326 break;
2327
2328 case O_IPVER:
2329 match = ((is_ipv4 || is_ipv6) &&
2330 cmd->arg1 == ip->ip_v);
2331 break;
2332
2333 case O_IPID:
2334 case O_IPTTL:
2335 if (!is_ipv4)
2336 break;
2337 case O_IPLEN:
2338 { /* only for IP packets */
2339 uint16_t x;
2340 uint16_t *p;
2341 int i;
2342
2343 if (cmd->opcode == O_IPLEN)
2344 x = iplen;
2345 else if (cmd->opcode == O_IPTTL)
2346 x = ip->ip_ttl;
2347 else /* must be IPID */
2348 x = ntohs(ip->ip_id);
2349 if (cmdlen == 1) {
2350 match = (cmd->arg1 == x);
2351 break;
2352 }
2353 /* otherwise we have ranges */
2354 p = ((ipfw_insn_u16 *)cmd)->ports;
2355 i = cmdlen - 1;
2356 for (; !match && i>0; i--, p += 2)
2357 match = (x >= p[0] && x <= p[1]);
2358 }
2359 break;
2360
2361 case O_IPPRECEDENCE:
2362 match = (is_ipv4 &&
2363 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2364 break;
2365
2366 case O_IPTOS:
2367 match = (is_ipv4 &&
2368 flags_match(cmd, ip->ip_tos));
2369 break;
2370
2371 case O_DSCP:
2372 {
2373 uint32_t *p;
2374 uint16_t x;
2375
2376 p = ((ipfw_insn_u32 *)cmd)->d;
2377
2378 if (is_ipv4)
2379 x = ip->ip_tos >> 2;
2380 else if (is_ipv6) {
2381 x = IPV6_DSCP(
2382 (struct ip6_hdr *)ip) >> 2;
2383 x &= 0x3f;
2384 } else
2385 break;
2386
2387 /* DSCP bitmask is stored as low_u32 high_u32 */
2388 if (x >= 32)
2389 match = *(p + 1) & (1 << (x - 32));
2390 else
2391 match = *p & (1 << x);
2392 }
2393 break;
2394
2395 case O_TCPDATALEN:
2396 if (proto == IPPROTO_TCP && offset == 0) {
2397 struct tcphdr *tcp;
2398 uint16_t x;
2399 uint16_t *p;
2400 int i;
2401 #ifdef INET6
2402 if (is_ipv6) {
2403 struct ip6_hdr *ip6;
2404
2405 ip6 = (struct ip6_hdr *)ip;
2406 if (ip6->ip6_plen == 0) {
2407 /*
2408 * Jumbo payload is not
2409 * supported by this
2410 * opcode.
2411 */
2412 break;
2413 }
2414 x = iplen - hlen;
2415 } else
2416 #endif /* INET6 */
2417 x = iplen - (ip->ip_hl << 2);
2418 tcp = TCP(ulp);
2419 x -= tcp->th_off << 2;
2420 if (cmdlen == 1) {
2421 match = (cmd->arg1 == x);
2422 break;
2423 }
2424 /* otherwise we have ranges */
2425 p = ((ipfw_insn_u16 *)cmd)->ports;
2426 i = cmdlen - 1;
2427 for (; !match && i>0; i--, p += 2)
2428 match = (x >= p[0] && x <= p[1]);
2429 }
2430 break;
2431
2432 case O_TCPFLAGS:
2433 match = (proto == IPPROTO_TCP && offset == 0 &&
2434 flags_match(cmd, TCP(ulp)->th_flags));
2435 break;
2436
2437 case O_TCPOPTS:
2438 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2439 PULLUP_LEN_LOCKED(hlen, ulp,
2440 (TCP(ulp)->th_off << 2));
2441 match = tcpopts_match(TCP(ulp), cmd);
2442 }
2443 break;
2444
2445 case O_TCPSEQ:
2446 match = (proto == IPPROTO_TCP && offset == 0 &&
2447 ((ipfw_insn_u32 *)cmd)->d[0] ==
2448 TCP(ulp)->th_seq);
2449 break;
2450
2451 case O_TCPACK:
2452 match = (proto == IPPROTO_TCP && offset == 0 &&
2453 ((ipfw_insn_u32 *)cmd)->d[0] ==
2454 TCP(ulp)->th_ack);
2455 break;
2456
2457 case O_TCPMSS:
2458 if (proto == IPPROTO_TCP &&
2459 (args->f_id._flags & TH_SYN) != 0 &&
2460 ulp != NULL) {
2461 uint16_t mss, *p;
2462 int i;
2463
2464 PULLUP_LEN_LOCKED(hlen, ulp,
2465 (TCP(ulp)->th_off << 2));
2466 if ((tcpopts_parse(TCP(ulp), &mss) &
2467 IP_FW_TCPOPT_MSS) == 0)
2468 break;
2469 if (cmdlen == 1) {
2470 match = (cmd->arg1 == mss);
2471 break;
2472 }
2473 /* Otherwise we have ranges. */
2474 p = ((ipfw_insn_u16 *)cmd)->ports;
2475 i = cmdlen - 1;
2476 for (; !match && i > 0; i--, p += 2)
2477 match = (mss >= p[0] &&
2478 mss <= p[1]);
2479 }
2480 break;
2481
2482 case O_TCPWIN:
2483 if (proto == IPPROTO_TCP && offset == 0) {
2484 uint16_t x;
2485 uint16_t *p;
2486 int i;
2487
2488 x = ntohs(TCP(ulp)->th_win);
2489 if (cmdlen == 1) {
2490 match = (cmd->arg1 == x);
2491 break;
2492 }
2493 /* Otherwise we have ranges. */
2494 p = ((ipfw_insn_u16 *)cmd)->ports;
2495 i = cmdlen - 1;
2496 for (; !match && i > 0; i--, p += 2)
2497 match = (x >= p[0] && x <= p[1]);
2498 }
2499 break;
2500
2501 case O_ESTAB:
2502 /* reject packets which have SYN only */
2503 /* XXX should i also check for TH_ACK ? */
2504 match = (proto == IPPROTO_TCP && offset == 0 &&
2505 (TCP(ulp)->th_flags &
2506 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2507 break;
2508
2509 case O_ALTQ: {
2510 struct pf_mtag *at;
2511 struct m_tag *mtag;
2512 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2513
2514 /*
2515 * ALTQ uses mbuf tags from another
2516 * packet filtering system - pf(4).
2517 * We allocate a tag in its format
2518 * and fill it in, pretending to be pf(4).
2519 */
2520 match = 1;
2521 at = pf_find_mtag(m);
2522 if (at != NULL && at->qid != 0)
2523 break;
2524 mtag = m_tag_get(PACKET_TAG_PF,
2525 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2526 if (mtag == NULL) {
2527 /*
2528 * Let the packet fall back to the
2529 * default ALTQ.
2530 */
2531 break;
2532 }
2533 m_tag_prepend(m, mtag);
2534 at = (struct pf_mtag *)(mtag + 1);
2535 at->qid = altq->qid;
2536 at->hdr = ip;
2537 break;
2538 }
2539
2540 case O_LOG:
2541 ipfw_log(chain, f, hlen, args,
2542 offset | ip6f_mf, tablearg, ip);
2543 match = 1;
2544 break;
2545
2546 case O_PROB:
2547 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2548 break;
2549
2550 case O_VERREVPATH:
2551 /* Outgoing packets automatically pass/match */
2552 match = (args->flags & IPFW_ARGS_OUT ||
2553 (
2554 #ifdef INET6
2555 is_ipv6 ?
2556 verify_path6(&(args->f_id.src_ip6),
2557 iif, args->f_id.fib) :
2558 #endif
2559 verify_path(src_ip, iif, args->f_id.fib)));
2560 break;
2561
2562 case O_VERSRCREACH:
2563 /* Outgoing packets automatically pass/match */
2564 match = (hlen > 0 && ((oif != NULL) || (
2565 #ifdef INET6
2566 is_ipv6 ?
2567 verify_path6(&(args->f_id.src_ip6),
2568 NULL, args->f_id.fib) :
2569 #endif
2570 verify_path(src_ip, NULL, args->f_id.fib))));
2571 break;
2572
2573 case O_ANTISPOOF:
2574 /* Outgoing packets automatically pass/match */
2575 if (oif == NULL && hlen > 0 &&
2576 ( (is_ipv4 && in_localaddr(src_ip))
2577 #ifdef INET6
2578 || (is_ipv6 &&
2579 in6_localaddr(&(args->f_id.src_ip6)))
2580 #endif
2581 ))
2582 match =
2583 #ifdef INET6
2584 is_ipv6 ? verify_path6(
2585 &(args->f_id.src_ip6), iif,
2586 args->f_id.fib) :
2587 #endif
2588 verify_path(src_ip, iif,
2589 args->f_id.fib);
2590 else
2591 match = 1;
2592 break;
2593
2594 case O_IPSEC:
2595 match = (m_tag_find(m,
2596 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2597 /* otherwise no match */
2598 break;
2599
2600 #ifdef INET6
2601 case O_IP6_SRC:
2602 match = is_ipv6 &&
2603 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2604 &((ipfw_insn_ip6 *)cmd)->addr6);
2605 break;
2606
2607 case O_IP6_DST:
2608 match = is_ipv6 &&
2609 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2610 &((ipfw_insn_ip6 *)cmd)->addr6);
2611 break;
2612 case O_IP6_SRC_MASK:
2613 case O_IP6_DST_MASK:
2614 if (is_ipv6) {
2615 int i = cmdlen - 1;
2616 struct in6_addr p;
2617 struct in6_addr *d =
2618 &((ipfw_insn_ip6 *)cmd)->addr6;
2619
2620 for (; !match && i > 0; d += 2,
2621 i -= F_INSN_SIZE(struct in6_addr)
2622 * 2) {
2623 p = (cmd->opcode ==
2624 O_IP6_SRC_MASK) ?
2625 args->f_id.src_ip6:
2626 args->f_id.dst_ip6;
2627 APPLY_MASK(&p, &d[1]);
2628 match =
2629 IN6_ARE_ADDR_EQUAL(&d[0],
2630 &p);
2631 }
2632 }
2633 break;
2634
2635 case O_FLOW6ID:
2636 match = is_ipv6 &&
2637 flow6id_match(args->f_id.flow_id6,
2638 (ipfw_insn_u32 *) cmd);
2639 break;
2640
2641 case O_EXT_HDR:
2642 match = is_ipv6 &&
2643 (ext_hd & ((ipfw_insn *) cmd)->arg1);
2644 break;
2645
2646 case O_IP6:
2647 match = is_ipv6;
2648 break;
2649 #endif
2650
2651 case O_IP4:
2652 match = is_ipv4;
2653 break;
2654
2655 case O_TAG: {
2656 struct m_tag *mtag;
2657 uint32_t tag = TARG(cmd->arg1, tag);
2658
2659 /* Packet is already tagged with this tag? */
2660 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2661
2662 /* We have `untag' action when F_NOT flag is
2663 * present. And we must remove this mtag from
2664 * mbuf and reset `match' to zero (`match' will
2665 * be inversed later).
2666 * Otherwise we should allocate new mtag and
2667 * push it into mbuf.
2668 */
2669 if (cmd->len & F_NOT) { /* `untag' action */
2670 if (mtag != NULL)
2671 m_tag_delete(m, mtag);
2672 match = 0;
2673 } else {
2674 if (mtag == NULL) {
2675 mtag = m_tag_alloc( MTAG_IPFW,
2676 tag, 0, M_NOWAIT);
2677 if (mtag != NULL)
2678 m_tag_prepend(m, mtag);
2679 }
2680 match = 1;
2681 }
2682 break;
2683 }
2684
2685 case O_FIB: /* try match the specified fib */
2686 if (args->f_id.fib == cmd->arg1)
2687 match = 1;
2688 break;
2689
2690 case O_SOCKARG: {
2691 #ifndef USERSPACE /* not supported in userspace */
2692 struct inpcb *inp = args->inp;
2693 struct inpcbinfo *pi;
2694 bool inp_locked = false;
2695
2696 if (proto == IPPROTO_TCP)
2697 pi = &V_tcbinfo;
2698 else if (proto == IPPROTO_UDP)
2699 pi = &V_udbinfo;
2700 else if (proto == IPPROTO_UDPLITE)
2701 pi = &V_ulitecbinfo;
2702 else
2703 break;
2704
2705 /*
2706 * XXXRW: so_user_cookie should almost
2707 * certainly be inp_user_cookie?
2708 */
2709
2710 /*
2711 * For incoming packet lookup the inpcb
2712 * using the src/dest ip/port tuple.
2713 */
2714 if (is_ipv4 && inp == NULL) {
2715 inp = in_pcblookup(pi,
2716 src_ip, htons(src_port),
2717 dst_ip, htons(dst_port),
2718 INPLOOKUP_RLOCKPCB, NULL);
2719 inp_locked = true;
2720 }
2721 #ifdef INET6
2722 if (is_ipv6 && inp == NULL) {
2723 inp = in6_pcblookup(pi,
2724 &args->f_id.src_ip6,
2725 htons(src_port),
2726 &args->f_id.dst_ip6,
2727 htons(dst_port),
2728 INPLOOKUP_RLOCKPCB, NULL);
2729 inp_locked = true;
2730 }
2731 #endif /* INET6 */
2732 if (inp != NULL) {
2733 if (inp->inp_socket) {
2734 tablearg =
2735 inp->inp_socket->so_user_cookie;
2736 if (tablearg)
2737 match = 1;
2738 }
2739 if (inp_locked)
2740 INP_RUNLOCK(inp);
2741 }
2742 #endif /* !USERSPACE */
2743 break;
2744 }
2745
2746 case O_TAGGED: {
2747 struct m_tag *mtag;
2748 uint32_t tag = TARG(cmd->arg1, tag);
2749
2750 if (cmdlen == 1) {
2751 match = m_tag_locate(m, MTAG_IPFW,
2752 tag, NULL) != NULL;
2753 break;
2754 }
2755
2756 /* we have ranges */
2757 for (mtag = m_tag_first(m);
2758 mtag != NULL && !match;
2759 mtag = m_tag_next(m, mtag)) {
2760 uint16_t *p;
2761 int i;
2762
2763 if (mtag->m_tag_cookie != MTAG_IPFW)
2764 continue;
2765
2766 p = ((ipfw_insn_u16 *)cmd)->ports;
2767 i = cmdlen - 1;
2768 for(; !match && i > 0; i--, p += 2)
2769 match =
2770 mtag->m_tag_id >= p[0] &&
2771 mtag->m_tag_id <= p[1];
2772 }
2773 break;
2774 }
2775
2776 /*
2777 * The second set of opcodes represents 'actions',
2778 * i.e. the terminal part of a rule once the packet
2779 * matches all previous patterns.
2780 * Typically there is only one action for each rule,
2781 * and the opcode is stored at the end of the rule
2782 * (but there are exceptions -- see below).
2783 *
2784 * In general, here we set retval and terminate the
2785 * outer loop (would be a 'break 3' in some language,
2786 * but we need to set l=0, done=1)
2787 *
2788 * Exceptions:
2789 * O_COUNT and O_SKIPTO actions:
2790 * instead of terminating, we jump to the next rule
2791 * (setting l=0), or to the SKIPTO target (setting
2792 * f/f_len, cmd and l as needed), respectively.
2793 *
2794 * O_TAG, O_LOG and O_ALTQ action parameters:
2795 * perform some action and set match = 1;
2796 *
2797 * O_LIMIT and O_KEEP_STATE: these opcodes are
2798 * not real 'actions', and are stored right
2799 * before the 'action' part of the rule (one
2800 * exception is O_SKIP_ACTION which could be
2801 * between these opcodes and 'action' one).
2802 * These opcodes try to install an entry in the
2803 * state tables; if successful, we continue with
2804 * the next opcode (match=1; break;), otherwise
2805 * the packet must be dropped (set retval,
2806 * break loops with l=0, done=1)
2807 *
2808 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2809 * cause a lookup of the state table, and a jump
2810 * to the 'action' part of the parent rule
2811 * if an entry is found, or
2812 * (CHECK_STATE only) a jump to the next rule if
2813 * the entry is not found.
2814 * The result of the lookup is cached so that
2815 * further instances of these opcodes become NOPs.
2816 * The jump to the next rule is done by setting
2817 * l=0, cmdlen=0.
2818 *
2819 * O_SKIP_ACTION: this opcode is not a real 'action'
2820 * either, and is stored right before the 'action'
2821 * part of the rule, right after the O_KEEP_STATE
2822 * opcode. It causes match failure so the real
2823 * 'action' could be executed only if the rule
2824 * is checked via dynamic rule from the state
2825 * table, as in such case execution starts
2826 * from the true 'action' opcode directly.
2827 *
2828 */
2829 case O_LIMIT:
2830 case O_KEEP_STATE:
2831 if (ipfw_dyn_install_state(chain, f,
2832 (ipfw_insn_limit *)cmd, args, ulp,
2833 pktlen, &dyn_info, tablearg)) {
2834 /* error or limit violation */
2835 retval = IP_FW_DENY;
2836 l = 0; /* exit inner loop */
2837 done = 1; /* exit outer loop */
2838 }
2839 match = 1;
2840 break;
2841
2842 case O_PROBE_STATE:
2843 case O_CHECK_STATE:
2844 /*
2845 * dynamic rules are checked at the first
2846 * keep-state or check-state occurrence,
2847 * with the result being stored in dyn_info.
2848 * The compiler introduces a PROBE_STATE
2849 * instruction for us when we have a
2850 * KEEP_STATE (because PROBE_STATE needs
2851 * to be run first).
2852 */
2853 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2854 (q = ipfw_dyn_lookup_state(args, ulp,
2855 pktlen, cmd, &dyn_info)) != NULL) {
2856 /*
2857 * Found dynamic entry, jump to the
2858 * 'action' part of the parent rule
2859 * by setting f, cmd, l and clearing
2860 * cmdlen.
2861 */
2862 f = q;
2863 f_pos = dyn_info.f_pos;
2864 cmd = ACTION_PTR(f);
2865 l = f->cmd_len - f->act_ofs;
2866 cmdlen = 0;
2867 match = 1;
2868 break;
2869 }
2870 /*
2871 * Dynamic entry not found. If CHECK_STATE,
2872 * skip to next rule, if PROBE_STATE just
2873 * ignore and continue with next opcode.
2874 */
2875 if (cmd->opcode == O_CHECK_STATE)
2876 l = 0; /* exit inner loop */
2877 match = 1;
2878 break;
2879
2880 case O_SKIP_ACTION:
2881 match = 0; /* skip to the next rule */
2882 l = 0; /* exit inner loop */
2883 break;
2884
2885 case O_ACCEPT:
2886 retval = 0; /* accept */
2887 l = 0; /* exit inner loop */
2888 done = 1; /* exit outer loop */
2889 break;
2890
2891 case O_PIPE:
2892 case O_QUEUE:
2893 set_match(args, f_pos, chain);
2894 args->rule.info = TARG(cmd->arg1, pipe);
2895 if (cmd->opcode == O_PIPE)
2896 args->rule.info |= IPFW_IS_PIPE;
2897 if (V_fw_one_pass)
2898 args->rule.info |= IPFW_ONEPASS;
2899 retval = IP_FW_DUMMYNET;
2900 l = 0; /* exit inner loop */
2901 done = 1; /* exit outer loop */
2902 break;
2903
2904 case O_DIVERT:
2905 case O_TEE:
2906 if (args->flags & IPFW_ARGS_ETHER)
2907 break; /* not on layer 2 */
2908 /* otherwise this is terminal */
2909 l = 0; /* exit inner loop */
2910 done = 1; /* exit outer loop */
2911 retval = (cmd->opcode == O_DIVERT) ?
2912 IP_FW_DIVERT : IP_FW_TEE;
2913 set_match(args, f_pos, chain);
2914 args->rule.info = TARG(cmd->arg1, divert);
2915 break;
2916
2917 case O_COUNT:
2918 IPFW_INC_RULE_COUNTER(f, pktlen);
2919 l = 0; /* exit inner loop */
2920 break;
2921
2922 case O_SKIPTO:
2923 IPFW_INC_RULE_COUNTER(f, pktlen);
2924 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2925 /*
2926 * Skip disabled rules, and re-enter
2927 * the inner loop with the correct
2928 * f_pos, f, l and cmd.
2929 * Also clear cmdlen and skip_or
2930 */
2931 for (; f_pos < chain->n_rules - 1 &&
2932 (V_set_disable &
2933 (1 << chain->map[f_pos]->set));
2934 f_pos++)
2935 ;
2936 /* Re-enter the inner loop at the skipto rule. */
2937 f = chain->map[f_pos];
2938 l = f->cmd_len;
2939 cmd = f->cmd;
2940 match = 1;
2941 cmdlen = 0;
2942 skip_or = 0;
2943 continue;
2944 break; /* not reached */
2945
2946 case O_CALLRETURN: {
2947 /*
2948 * Implementation of `subroutine' call/return,
2949 * in the stack carried in an mbuf tag. This
2950 * is different from `skipto' in that any call
2951 * address is possible (`skipto' must prevent
2952 * backward jumps to avoid endless loops).
2953 * We have `return' action when F_NOT flag is
2954 * present. The `m_tag_id' field is used as
2955 * stack pointer.
2956 */
2957 struct m_tag *mtag;
2958 uint16_t jmpto, *stack;
2959
2960 #define IS_CALL ((cmd->len & F_NOT) == 0)
2961 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2962 /*
2963 * Hand-rolled version of m_tag_locate() with
2964 * wildcard `type'.
2965 * If not already tagged, allocate new tag.
2966 */
2967 mtag = m_tag_first(m);
2968 while (mtag != NULL) {
2969 if (mtag->m_tag_cookie ==
2970 MTAG_IPFW_CALL)
2971 break;
2972 mtag = m_tag_next(m, mtag);
2973 }
2974 if (mtag == NULL && IS_CALL) {
2975 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2976 IPFW_CALLSTACK_SIZE *
2977 sizeof(uint16_t), M_NOWAIT);
2978 if (mtag != NULL)
2979 m_tag_prepend(m, mtag);
2980 }
2981
2982 /*
2983 * On error both `call' and `return' just
2984 * continue with next rule.
2985 */
2986 if (IS_RETURN && (mtag == NULL ||
2987 mtag->m_tag_id == 0)) {
2988 l = 0; /* exit inner loop */
2989 break;
2990 }
2991 if (IS_CALL && (mtag == NULL ||
2992 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2993 printf("ipfw: call stack error, "
2994 "go to next rule\n");
2995 l = 0; /* exit inner loop */
2996 break;
2997 }
2998
2999 IPFW_INC_RULE_COUNTER(f, pktlen);
3000 stack = (uint16_t *)(mtag + 1);
3001
3002 /*
3003 * The `call' action may use cached f_pos
3004 * (in f->next_rule), whose version is written
3005 * in f->next_rule.
3006 * The `return' action, however, doesn't have
3007 * fixed jump address in cmd->arg1 and can't use
3008 * cache.
3009 */
3010 if (IS_CALL) {
3011 stack[mtag->m_tag_id] = f->rulenum;
3012 mtag->m_tag_id++;
3013 f_pos = JUMP(chain, f, cmd->arg1,
3014 tablearg, 1);
3015 } else { /* `return' action */
3016 mtag->m_tag_id--;
3017 jmpto = stack[mtag->m_tag_id] + 1;
3018 f_pos = ipfw_find_rule(chain, jmpto, 0);
3019 }
3020
3021 /*
3022 * Skip disabled rules, and re-enter
3023 * the inner loop with the correct
3024 * f_pos, f, l and cmd.
3025 * Also clear cmdlen and skip_or
3026 */
3027 for (; f_pos < chain->n_rules - 1 &&
3028 (V_set_disable &
3029 (1 << chain->map[f_pos]->set)); f_pos++)
3030 ;
3031 /* Re-enter the inner loop at the dest rule. */
3032 f = chain->map[f_pos];
3033 l = f->cmd_len;
3034 cmd = f->cmd;
3035 cmdlen = 0;
3036 skip_or = 0;
3037 continue;
3038 break; /* NOTREACHED */
3039 }
3040 #undef IS_CALL
3041 #undef IS_RETURN
3042
3043 case O_REJECT:
3044 /*
3045 * Drop the packet and send a reject notice
3046 * if the packet is not ICMP (or is an ICMP
3047 * query), and it is not multicast/broadcast.
3048 */
3049 if (hlen > 0 && is_ipv4 && offset == 0 &&
3050 (proto != IPPROTO_ICMP ||
3051 is_icmp_query(ICMP(ulp))) &&
3052 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3053 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3054 send_reject(args, cmd, iplen, ip);
3055 m = args->m;
3056 }
3057 /* FALLTHROUGH */
3058 #ifdef INET6
3059 case O_UNREACH6:
3060 if (hlen > 0 && is_ipv6 &&
3061 ((offset & IP6F_OFF_MASK) == 0) &&
3062 (proto != IPPROTO_ICMPV6 ||
3063 (is_icmp6_query(icmp6_type) == 1)) &&
3064 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3065 !IN6_IS_ADDR_MULTICAST(
3066 &args->f_id.dst_ip6)) {
3067 send_reject6(args,
3068 cmd->opcode == O_REJECT ?
3069 map_icmp_unreach(cmd->arg1):
3070 cmd->arg1, hlen,
3071 (struct ip6_hdr *)ip);
3072 m = args->m;
3073 }
3074 /* FALLTHROUGH */
3075 #endif
3076 case O_DENY:
3077 retval = IP_FW_DENY;
3078 l = 0; /* exit inner loop */
3079 done = 1; /* exit outer loop */
3080 break;
3081
3082 case O_FORWARD_IP:
3083 if (args->flags & IPFW_ARGS_ETHER)
3084 break; /* not valid on layer2 pkts */
3085 if (q != f ||
3086 dyn_info.direction == MATCH_FORWARD) {
3087 struct sockaddr_in *sa;
3088
3089 sa = &(((ipfw_insn_sa *)cmd)->sa);
3090 if (sa->sin_addr.s_addr == INADDR_ANY) {
3091 #ifdef INET6
3092 /*
3093 * We use O_FORWARD_IP opcode for
3094 * fwd rule with tablearg, but tables
3095 * now support IPv6 addresses. And
3096 * when we are inspecting IPv6 packet,
3097 * we can use nh6 field from
3098 * table_value as next_hop6 address.
3099 */
3100 if (is_ipv6) {
3101 struct ip_fw_nh6 *nh6;
3102
3103 args->flags |= IPFW_ARGS_NH6;
3104 nh6 = &args->hopstore6;
3105 nh6->sin6_addr = TARG_VAL(
3106 chain, tablearg, nh6);
3107 nh6->sin6_port = sa->sin_port;
3108 nh6->sin6_scope_id = TARG_VAL(
3109 chain, tablearg, zoneid);
3110 } else
3111 #endif
3112 {
3113 args->flags |= IPFW_ARGS_NH4;
3114 args->hopstore.sin_port =
3115 sa->sin_port;
3116 sa = &args->hopstore;
3117 sa->sin_family = AF_INET;
3118 sa->sin_len = sizeof(*sa);
3119 sa->sin_addr.s_addr = htonl(
3120 TARG_VAL(chain, tablearg,
3121 nh4));
3122 }
3123 } else {
3124 args->flags |= IPFW_ARGS_NH4PTR;
3125 args->next_hop = sa;
3126 }
3127 }
3128 retval = IP_FW_PASS;
3129 l = 0; /* exit inner loop */
3130 done = 1; /* exit outer loop */
3131 break;
3132
3133 #ifdef INET6
3134 case O_FORWARD_IP6:
3135 if (args->flags & IPFW_ARGS_ETHER)
3136 break; /* not valid on layer2 pkts */
3137 if (q != f ||
3138 dyn_info.direction == MATCH_FORWARD) {
3139 struct sockaddr_in6 *sin6;
3140
3141 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
3142 args->flags |= IPFW_ARGS_NH6PTR;
3143 args->next_hop6 = sin6;
3144 }
3145 retval = IP_FW_PASS;
3146 l = 0; /* exit inner loop */
3147 done = 1; /* exit outer loop */
3148 break;
3149 #endif
3150
3151 case O_NETGRAPH:
3152 case O_NGTEE:
3153 set_match(args, f_pos, chain);
3154 args->rule.info = TARG(cmd->arg1, netgraph);
3155 if (V_fw_one_pass)
3156 args->rule.info |= IPFW_ONEPASS;
3157 retval = (cmd->opcode == O_NETGRAPH) ?
3158 IP_FW_NETGRAPH : IP_FW_NGTEE;
3159 l = 0; /* exit inner loop */
3160 done = 1; /* exit outer loop */
3161 break;
3162
3163 case O_SETFIB: {
3164 uint32_t fib;
3165
3166 IPFW_INC_RULE_COUNTER(f, pktlen);
3167 fib = TARG(cmd->arg1, fib) & 0x7FFF;
3168 if (fib >= rt_numfibs)
3169 fib = 0;
3170 M_SETFIB(m, fib);
3171 args->f_id.fib = fib; /* XXX */
3172 l = 0; /* exit inner loop */
3173 break;
3174 }
3175
3176 case O_SETDSCP: {
3177 uint16_t code;
3178
3179 code = TARG(cmd->arg1, dscp) & 0x3F;
3180 l = 0; /* exit inner loop */
3181 if (is_ipv4) {
3182 uint16_t old;
3183
3184 old = *(uint16_t *)ip;
3185 ip->ip_tos = (code << 2) |
3186 (ip->ip_tos & 0x03);
3187 ip->ip_sum = cksum_adjust(ip->ip_sum,
3188 old, *(uint16_t *)ip);
3189 } else if (is_ipv6) {
3190 /* update cached value */
3191 args->f_id.flow_id6 =
3192 ntohl(*(uint32_t *)ip) & ~0x0FC00000;
3193 args->f_id.flow_id6 |= code << 22;
3194
3195 *((uint32_t *)ip) =
3196 htonl(args->f_id.flow_id6);
3197 } else
3198 break;
3199
3200 IPFW_INC_RULE_COUNTER(f, pktlen);
3201 break;
3202 }
3203
3204 case O_NAT:
3205 l = 0; /* exit inner loop */
3206 done = 1; /* exit outer loop */
3207 /*
3208 * Ensure that we do not invoke NAT handler for
3209 * non IPv4 packets. Libalias expects only IPv4.
3210 */
3211 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3212 retval = IP_FW_DENY;
3213 break;
3214 }
3215
3216 struct cfg_nat *t;
3217 int nat_id;
3218
3219 args->rule.info = 0;
3220 set_match(args, f_pos, chain);
3221 /* Check if this is 'global' nat rule */
3222 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3223 retval = ipfw_nat_ptr(args, NULL, m);
3224 break;
3225 }
3226 t = ((ipfw_insn_nat *)cmd)->nat;
3227 if (t == NULL) {
3228 nat_id = TARG(cmd->arg1, nat);
3229 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3230
3231 if (t == NULL) {
3232 retval = IP_FW_DENY;
3233 break;
3234 }
3235 if (cmd->arg1 != IP_FW_TARG)
3236 ((ipfw_insn_nat *)cmd)->nat = t;
3237 }
3238 retval = ipfw_nat_ptr(args, t, m);
3239 break;
3240
3241 case O_REASS: {
3242 int ip_off;
3243
3244 l = 0; /* in any case exit inner loop */
3245 if (is_ipv6) /* IPv6 is not supported yet */
3246 break;
3247 IPFW_INC_RULE_COUNTER(f, pktlen);
3248 ip_off = ntohs(ip->ip_off);
3249
3250 /* if not fragmented, go to next rule */
3251 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3252 break;
3253
3254 args->m = m = ip_reass(m);
3255
3256 /*
3257 * do IP header checksum fixup.
3258 */
3259 if (m == NULL) { /* fragment got swallowed */
3260 retval = IP_FW_DENY;
3261 } else { /* good, packet complete */
3262 int hlen;
3263
3264 ip = mtod(m, struct ip *);
3265 hlen = ip->ip_hl << 2;
3266 ip->ip_sum = 0;
3267 if (hlen == sizeof(struct ip))
3268 ip->ip_sum = in_cksum_hdr(ip);
3269 else
3270 ip->ip_sum = in_cksum(m, hlen);
3271 retval = IP_FW_REASS;
3272 args->rule.info = 0;
3273 set_match(args, f_pos, chain);
3274 }
3275 done = 1; /* exit outer loop */
3276 break;
3277 }
3278 case O_EXTERNAL_ACTION:
3279 l = 0; /* in any case exit inner loop */
3280 retval = ipfw_run_eaction(chain, args,
3281 cmd, &done);
3282 /*
3283 * If both @retval and @done are zero,
3284 * consider this as rule matching and
3285 * update counters.
3286 */
3287 if (retval == 0 && done == 0) {
3288 IPFW_INC_RULE_COUNTER(f, pktlen);
3289 /*
3290 * Reset the result of the last
3291 * dynamic state lookup.
3292 * External action can change
3293 * @args content, and it may be
3294 * used for new state lookup later.
3295 */
3296 DYN_INFO_INIT(&dyn_info);
3297 }
3298 break;
3299
3300 default:
3301 panic("-- unknown opcode %d\n", cmd->opcode);
3302 } /* end of switch() on opcodes */
3303 /*
3304 * if we get here with l=0, then match is irrelevant.
3305 */
3306
3307 if (cmd->len & F_NOT)
3308 match = !match;
3309
3310 if (match) {
3311 if (cmd->len & F_OR)
3312 skip_or = 1;
3313 } else {
3314 if (!(cmd->len & F_OR)) /* not an OR block, */
3315 break; /* try next rule */
3316 }
3317
3318 } /* end of inner loop, scan opcodes */
3319 #undef PULLUP_LEN
3320 #undef PULLUP_LEN_LOCKED
3321
3322 if (done)
3323 break;
3324
3325 /* next_rule:; */ /* try next rule */
3326
3327 } /* end of outer for, scan rules */
3328
3329 if (done) {
3330 struct ip_fw *rule = chain->map[f_pos];
3331 /* Update statistics */
3332 IPFW_INC_RULE_COUNTER(rule, pktlen);
3333 IPFW_PROBE(rule__matched, retval,
3334 is_ipv4 ? AF_INET : AF_INET6,
3335 is_ipv4 ? (uintptr_t)&src_ip :
3336 (uintptr_t)&args->f_id.src_ip6,
3337 is_ipv4 ? (uintptr_t)&dst_ip :
3338 (uintptr_t)&args->f_id.dst_ip6,
3339 args, rule);
3340 } else {
3341 retval = IP_FW_DENY;
3342 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3343 }
3344 IPFW_PF_RUNLOCK(chain);
3345 #ifdef __FreeBSD__
3346 if (ucred_cache != NULL)
3347 crfree(ucred_cache);
3348 #endif
3349 return (retval);
3350
3351 pullup_failed:
3352 if (V_fw_verbose)
3353 printf("ipfw: pullup failed\n");
3354 return (IP_FW_DENY);
3355 }
3356
3357 /*
3358 * Set maximum number of tables that can be used in given VNET ipfw instance.
3359 */
3360 #ifdef SYSCTL_NODE
3361 static int
3362 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3363 {
3364 int error;
3365 unsigned int ntables;
3366
3367 ntables = V_fw_tables_max;
3368
3369 error = sysctl_handle_int(oidp, &ntables, 0, req);
3370 /* Read operation or some error */
3371 if ((error != 0) || (req->newptr == NULL))
3372 return (error);
3373
3374 return (ipfw_resize_tables(&V_layer3_chain, ntables));
3375 }
3376
3377 /*
3378 * Switches table namespace between global and per-set.
3379 */
3380 static int
3381 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3382 {
3383 int error;
3384 unsigned int sets;
3385
3386 sets = V_fw_tables_sets;
3387
3388 error = sysctl_handle_int(oidp, &sets, 0, req);
3389 /* Read operation or some error */
3390 if ((error != 0) || (req->newptr == NULL))
3391 return (error);
3392
3393 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3394 }
3395 #endif
3396
3397 /*
3398 * Module and VNET glue
3399 */
3400
3401 /*
3402 * Stuff that must be initialised only on boot or module load
3403 */
3404 static int
3405 ipfw_init(void)
3406 {
3407 int error = 0;
3408
3409 /*
3410 * Only print out this stuff the first time around,
3411 * when called from the sysinit code.
3412 */
3413 printf("ipfw2 "
3414 #ifdef INET6
3415 "(+ipv6) "
3416 #endif
3417 "initialized, divert %s, nat %s, "
3418 "default to %s, logging ",
3419 #ifdef IPDIVERT
3420 "enabled",
3421 #else
3422 "loadable",
3423 #endif
3424 #ifdef IPFIREWALL_NAT
3425 "enabled",
3426 #else
3427 "loadable",
3428 #endif
3429 default_to_accept ? "accept" : "deny");
3430
3431 /*
3432 * Note: V_xxx variables can be accessed here but the vnet specific
3433 * initializer may not have been called yet for the VIMAGE case.
3434 * Tuneables will have been processed. We will print out values for
3435 * the default vnet.
3436 * XXX This should all be rationalized AFTER 8.0
3437 */
3438 if (V_fw_verbose == 0)
3439 printf("disabled\n");
3440 else if (V_verbose_limit == 0)
3441 printf("unlimited\n");
3442 else
3443 printf("limited to %d packets/entry by default\n",
3444 V_verbose_limit);
3445
3446 /* Check user-supplied table count for validness */
3447 if (default_fw_tables > IPFW_TABLES_MAX)
3448 default_fw_tables = IPFW_TABLES_MAX;
3449
3450 ipfw_init_sopt_handler();
3451 ipfw_init_obj_rewriter();
3452 ipfw_iface_init();
3453 return (error);
3454 }
3455
3456 /*
3457 * Called for the removal of the last instance only on module unload.
3458 */
3459 static void
3460 ipfw_destroy(void)
3461 {
3462
3463 ipfw_iface_destroy();
3464 ipfw_destroy_sopt_handler();
3465 ipfw_destroy_obj_rewriter();
3466 printf("IP firewall unloaded\n");
3467 }
3468
3469 /*
3470 * Stuff that must be initialized for every instance
3471 * (including the first of course).
3472 */
3473 static int
3474 vnet_ipfw_init(const void *unused)
3475 {
3476 int error, first;
3477 struct ip_fw *rule = NULL;
3478 struct ip_fw_chain *chain;
3479
3480 chain = &V_layer3_chain;
3481
3482 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3483
3484 /* First set up some values that are compile time options */
3485 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
3486 V_fw_deny_unknown_exthdrs = 1;
3487 #ifdef IPFIREWALL_VERBOSE
3488 V_fw_verbose = 1;
3489 #endif
3490 #ifdef IPFIREWALL_VERBOSE_LIMIT
3491 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3492 #endif
3493 #ifdef IPFIREWALL_NAT
3494 LIST_INIT(&chain->nat);
3495 #endif
3496
3497 /* Init shared services hash table */
3498 ipfw_init_srv(chain);
3499
3500 ipfw_init_counters();
3501 /* Set initial number of tables */
3502 V_fw_tables_max = default_fw_tables;
3503 error = ipfw_init_tables(chain, first);
3504 if (error) {
3505 printf("ipfw2: setting up tables failed\n");
3506 free(chain->map, M_IPFW);
3507 free(rule, M_IPFW);
3508 return (ENOSPC);
3509 }
3510
3511 IPFW_LOCK_INIT(chain);
3512
3513 /* fill and insert the default rule */
3514 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3515 rule->flags |= IPFW_RULE_NOOPT;
3516 rule->cmd_len = 1;
3517 rule->cmd[0].len = 1;
3518 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3519 chain->default_rule = rule;
3520 ipfw_add_protected_rule(chain, rule, 0);
3521
3522 ipfw_dyn_init(chain);
3523 ipfw_eaction_init(chain, first);
3524 #ifdef LINEAR_SKIPTO
3525 ipfw_init_skipto_cache(chain);
3526 #endif
3527 ipfw_bpf_init(first);
3528
3529 /* First set up some values that are compile time options */
3530 V_ipfw_vnet_ready = 1; /* Open for business */
3531
3532 /*
3533 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3534 * Even if the latter two fail we still keep the module alive
3535 * because the sockopt and layer2 paths are still useful.
3536 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3537 * so we can ignore the exact return value and just set a flag.
3538 *
3539 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3540 * changes in the underlying (per-vnet) variables trigger
3541 * immediate hook()/unhook() calls.
3542 * In layer2 we have the same behaviour, except that V_ether_ipfw
3543 * is checked on each packet because there are no pfil hooks.
3544 */
3545 V_ip_fw_ctl_ptr = ipfw_ctl3;
3546 error = ipfw_attach_hooks();
3547 return (error);
3548 }
3549
3550 /*
3551 * Called for the removal of each instance.
3552 */
3553 static int
3554 vnet_ipfw_uninit(const void *unused)
3555 {
3556 struct ip_fw *reap;
3557 struct ip_fw_chain *chain = &V_layer3_chain;
3558 int i, last;
3559
3560 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3561 /*
3562 * disconnect from ipv4, ipv6, layer2 and sockopt.
3563 * Then grab, release and grab again the WLOCK so we make
3564 * sure the update is propagated and nobody will be in.
3565 */
3566 ipfw_detach_hooks();
3567 V_ip_fw_ctl_ptr = NULL;
3568
3569 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3570
3571 IPFW_UH_WLOCK(chain);
3572 IPFW_UH_WUNLOCK(chain);
3573
3574 ipfw_dyn_uninit(0); /* run the callout_drain */
3575
3576 IPFW_UH_WLOCK(chain);
3577
3578 reap = NULL;
3579 IPFW_WLOCK(chain);
3580 for (i = 0; i < chain->n_rules; i++)
3581 ipfw_reap_add(chain, &reap, chain->map[i]);
3582 free(chain->map, M_IPFW);
3583 #ifdef LINEAR_SKIPTO
3584 ipfw_destroy_skipto_cache(chain);
3585 #endif
3586 IPFW_WUNLOCK(chain);
3587 IPFW_UH_WUNLOCK(chain);
3588 ipfw_destroy_tables(chain, last);
3589 ipfw_eaction_uninit(chain, last);
3590 if (reap != NULL)
3591 ipfw_reap_rules(reap);
3592 vnet_ipfw_iface_destroy(chain);
3593 ipfw_destroy_srv(chain);
3594 IPFW_LOCK_DESTROY(chain);
3595 ipfw_dyn_uninit(1); /* free the remaining parts */
3596 ipfw_destroy_counters();
3597 ipfw_bpf_uninit(last);
3598 return (0);
3599 }
3600
3601 /*
3602 * Module event handler.
3603 * In general we have the choice of handling most of these events by the
3604 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3605 * use the SYSINIT handlers as they are more capable of expressing the
3606 * flow of control during module and vnet operations, so this is just
3607 * a skeleton. Note there is no SYSINIT equivalent of the module
3608 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3609 */
3610 static int
3611 ipfw_modevent(module_t mod, int type, void *unused)
3612 {
3613 int err = 0;
3614
3615 switch (type) {
3616 case MOD_LOAD:
3617 /* Called once at module load or
3618 * system boot if compiled in. */
3619 break;
3620 case MOD_QUIESCE:
3621 /* Called before unload. May veto unloading. */
3622 break;
3623 case MOD_UNLOAD:
3624 /* Called during unload. */
3625 break;
3626 case MOD_SHUTDOWN:
3627 /* Called during system shutdown. */
3628 break;
3629 default:
3630 err = EOPNOTSUPP;
3631 break;
3632 }
3633 return err;
3634 }
3635
3636 static moduledata_t ipfwmod = {
3637 "ipfw",
3638 ipfw_modevent,
3639 0
3640 };
3641
3642 /* Define startup order. */
3643 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL
3644 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
3645 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3646 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3647
3648 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3649 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3650 MODULE_VERSION(ipfw, 3);
3651 /* should declare some dependencies here */
3652
3653 /*
3654 * Starting up. Done in order after ipfwmod() has been called.
3655 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3656 */
3657 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3658 ipfw_init, NULL);
3659 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3660 vnet_ipfw_init, NULL);
3661
3662 /*
3663 * Closing up shop. These are done in REVERSE ORDER, but still
3664 * after ipfwmod() has been called. Not called on reboot.
3665 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3666 * or when the module is unloaded.
3667 */
3668 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3669 ipfw_destroy, NULL);
3670 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3671 vnet_ipfw_uninit, NULL);
3672 /* end of file */
Cache object: 55fa0353e68d0f2eccfd699943630859
|