sys/netpfil/ipfw/ip_fw2.c

   1 /*-
   2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
   3  *
   4  * Redistribution and use in source and binary forms, with or without
   5  * modification, are permitted provided that the following conditions
   6  * are met:
   7  * 1. Redistributions of source code must retain the above copyright
   8  *    notice, this list of conditions and the following disclaimer.
   9  * 2. Redistributions in binary form must reproduce the above copyright
  10  *    notice, this list of conditions and the following disclaimer in the
  11  *    documentation and/or other materials provided with the distribution.
  12  *
  13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  23  * SUCH DAMAGE.
  24  */
  25
  26 #include <sys/cdefs.h>
  27 __FBSDID("$FreeBSD$");
  28
  29 /*
  30  * The FreeBSD IP packet firewall, main file
  31  */
  32
  33 #include "opt_ipfw.h"
  34 #include "opt_ipdivert.h"
  35 #include "opt_inet.h"
  36 #ifndef INET
  37 #error IPFIREWALL requires INET.
  38 #endif /* INET */
  39 #include "opt_inet6.h"
  40 #include "opt_ipsec.h"
  41
  42 #include <sys/param.h>
  43 #include <sys/systm.h>
  44 #include <sys/condvar.h>
  45 #include <sys/eventhandler.h>
  46 #include <sys/malloc.h>
  47 #include <sys/mbuf.h>
  48 #include <sys/kernel.h>
  49 #include <sys/lock.h>
  50 #include <sys/jail.h>
  51 #include <sys/module.h>
  52 #include <sys/priv.h>
  53 #include <sys/proc.h>
  54 #include <sys/rwlock.h>
  55 #include <sys/socket.h>
  56 #include <sys/socketvar.h>
  57 #include <sys/sysctl.h>
  58 #include <sys/syslog.h>
  59 #include <sys/ucred.h>
  60 #include <net/ethernet.h> /* for ETHERTYPE_IP */
  61 #include <net/if.h>
  62 #include <net/route.h>
  63 #include <net/pf_mtag.h>
  64 #include <net/pfil.h>
  65 #include <net/vnet.h>
  66
  67 #include <netinet/in.h>
  68 #include <netinet/in_var.h>
  69 #include <netinet/in_pcb.h>
  70 #include <netinet/ip.h>
  71 #include <netinet/ip_var.h>
  72 #include <netinet/ip_icmp.h>
  73 #include <netinet/ip_fw.h>
  74 #include <netinet/ip_carp.h>
  75 #include <netinet/pim.h>
  76 #include <netinet/tcp_var.h>
  77 #include <netinet/udp.h>
  78 #include <netinet/udp_var.h>
  79 #include <netinet/sctp.h>
  80
  81 #include <netinet/ip6.h>
  82 #include <netinet/icmp6.h>
  83 #ifdef INET6
  84 #include <netinet6/in6_pcb.h>
  85 #include <netinet6/scope6_var.h>
  86 #include <netinet6/ip6_var.h>
  87 #endif
  88
  89 #include <net/if_gre.h> /* for struct gre_h */
  90
  91 #include <netpfil/ipfw/ip_fw_private.h>
  92
  93 #include <machine/in_cksum.h>   /* XXX for in_cksum */
  94
  95 #ifdef MAC
  96 #include <security/mac/mac_framework.h>
  97 #endif
  98
  99 /*
 100  * static variables followed by global ones.
 101  * All ipfw global variables are here.
 102  */
 103
 104 /* ipfw_vnet_ready controls when we are open for business */
 105 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 106 #define V_ipfw_vnet_ready       VNET(ipfw_vnet_ready)
 107
 108 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
 109 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 110
 111 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
 112 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 113
 114 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 115 static int default_to_accept = 1;
 116 #else
 117 static int default_to_accept;
 118 #endif
 119
 120 VNET_DEFINE(int, autoinc_step);
 121 VNET_DEFINE(int, fw_one_pass) = 1;
 122
 123 VNET_DEFINE(unsigned int, fw_tables_max);
 124 /* Use 128 tables by default */
 125 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 126
 127 /*
 128  * Each rule belongs to one of 32 different sets (0..31).
 129  * The variable set_disable contains one bit per set.
 130  * If the bit is set, all rules in the corresponding set
 131  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 132  * and rules that are not deleted by the flush command,
 133  * and CANNOT be disabled.
 134  * Rules in set RESVD_SET can only be deleted individually.
 135  */
 136 VNET_DEFINE(u_int32_t, set_disable);
 137 #define V_set_disable                   VNET(set_disable)
 138
 139 VNET_DEFINE(int, fw_verbose);
 140 /* counter for ipfw_log(NULL...) */
 141 VNET_DEFINE(u_int64_t, norule_counter);
 142 VNET_DEFINE(int, verbose_limit);
 143
 144 /* layer3_chain contains the list of rules for layer 3 */
 145 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 146
 147 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 148
 149 ipfw_nat_t *ipfw_nat_ptr = NULL;
 150 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 151 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 152 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 153 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 154 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 155
 156 #ifdef SYSCTL_NODE
 157 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 158 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 159
 160 SYSBEGIN(f3)
 161
 162 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 163 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 164     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 165     "Only do a single pass through ipfw when using dummynet(4)");
 166 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 167     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 168     "Rule number auto-increment step");
 169 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 170     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 171     "Log matches to ipfw rules");
 172 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 173     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 174     "Set upper limit of matches of ipfw rules logged");
 175 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 176     &dummy_def, 0,
 177     "The default/max possible rule number.");
 178 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 179     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 180     "Maximum number of tables");
 181 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 182     &default_to_accept, 0,
 183     "Make the default rule accept all packets.");
 184 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
 185 TUNABLE_INT("net.inet.ip.fw.tables_max", &default_fw_tables);
 186 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 187     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 188     "Number of static rules");
 189
 190 #ifdef INET6
 191 SYSCTL_DECL(_net_inet6_ip6);
 192 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 193 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 194     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 195     "Deny packets with unknown IPv6 Extension Headers");
 196 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 197     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
 198     "Permit single packet IPv6 fragments");
 199 #endif /* INET6 */
 200
 201 SYSEND
 202
 203 #endif /* SYSCTL_NODE */
 204
 205
 206 /*
 207  * Some macros used in the various matching options.
 208  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 209  * Other macros just cast void * into the appropriate type
 210  */
 211 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 212 #define TCP(p)          ((struct tcphdr *)(p))
 213 #define SCTP(p)         ((struct sctphdr *)(p))
 214 #define UDP(p)          ((struct udphdr *)(p))
 215 #define ICMP(p)         ((struct icmphdr *)(p))
 216 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 217
 218 static __inline int
 219 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 220 {
 221         int type = icmp->icmp_type;
 222
 223         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 224 }
 225
 226 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 227     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 228
 229 static int
 230 is_icmp_query(struct icmphdr *icmp)
 231 {
 232         int type = icmp->icmp_type;
 233
 234         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 235 }
 236 #undef TT
 237
 238 /*
 239  * The following checks use two arrays of 8 or 16 bits to store the
 240  * bits that we want set or clear, respectively. They are in the
 241  * low and high half of cmd->arg1 or cmd->d[0].
 242  *
 243  * We scan options and store the bits we find set. We succeed if
 244  *
 245  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 246  *
 247  * The code is sometimes optimized not to store additional variables.
 248  */
 249
 250 static int
 251 flags_match(ipfw_insn *cmd, u_int8_t bits)
 252 {
 253         u_char want_clear;
 254         bits = ~bits;
 255
 256         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 257                 return 0; /* some bits we want set were clear */
 258         want_clear = (cmd->arg1 >> 8) & 0xff;
 259         if ( (want_clear & bits) != want_clear)
 260                 return 0; /* some bits we want clear were set */
 261         return 1;
 262 }
 263
 264 static int
 265 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 266 {
 267         int optlen, bits = 0;
 268         u_char *cp = (u_char *)(ip + 1);
 269         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 270
 271         for (; x > 0; x -= optlen, cp += optlen) {
 272                 int opt = cp[IPOPT_OPTVAL];
 273
 274                 if (opt == IPOPT_EOL)
 275                         break;
 276                 if (opt == IPOPT_NOP)
 277                         optlen = 1;
 278                 else {
 279                         optlen = cp[IPOPT_OLEN];
 280                         if (optlen <= 0 || optlen > x)
 281                                 return 0; /* invalid or truncated */
 282                 }
 283                 switch (opt) {
 284
 285                 default:
 286                         break;
 287
 288                 case IPOPT_LSRR:
 289                         bits |= IP_FW_IPOPT_LSRR;
 290                         break;
 291
 292                 case IPOPT_SSRR:
 293                         bits |= IP_FW_IPOPT_SSRR;
 294                         break;
 295
 296                 case IPOPT_RR:
 297                         bits |= IP_FW_IPOPT_RR;
 298                         break;
 299
 300                 case IPOPT_TS:
 301                         bits |= IP_FW_IPOPT_TS;
 302                         break;
 303                 }
 304         }
 305         return (flags_match(cmd, bits));
 306 }
 307
 308 static int
 309 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 310 {
 311         int optlen, bits = 0;
 312         u_char *cp = (u_char *)(tcp + 1);
 313         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 314
 315         for (; x > 0; x -= optlen, cp += optlen) {
 316                 int opt = cp[0];
 317                 if (opt == TCPOPT_EOL)
 318                         break;
 319                 if (opt == TCPOPT_NOP)
 320                         optlen = 1;
 321                 else {
 322                         optlen = cp[1];
 323                         if (optlen <= 0)
 324                                 break;
 325                 }
 326
 327                 switch (opt) {
 328
 329                 default:
 330                         break;
 331
 332                 case TCPOPT_MAXSEG:
 333                         bits |= IP_FW_TCPOPT_MSS;
 334                         break;
 335
 336                 case TCPOPT_WINDOW:
 337                         bits |= IP_FW_TCPOPT_WINDOW;
 338                         break;
 339
 340                 case TCPOPT_SACK_PERMITTED:
 341                 case TCPOPT_SACK:
 342                         bits |= IP_FW_TCPOPT_SACK;
 343                         break;
 344
 345                 case TCPOPT_TIMESTAMP:
 346                         bits |= IP_FW_TCPOPT_TS;
 347                         break;
 348
 349                 }
 350         }
 351         return (flags_match(cmd, bits));
 352 }
 353
 354 static int
 355 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
 356 {
 357         if (ifp == NULL)        /* no iface with this packet, match fails */
 358                 return 0;
 359         /* Check by name or by IP address */
 360         if (cmd->name[0] != '\0') { /* match by name */
 361                 if (cmd->name[0] == '\1') /* use tablearg to match */
 362                         return ipfw_lookup_table_extended(chain, cmd->p.glob,
 363                                 ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
 364                 /* Check name */
 365                 if (cmd->p.glob) {
 366                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 367                                 return(1);
 368                 } else {
 369                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 370                                 return(1);
 371                 }
 372         } else {
 373 #ifdef __FreeBSD__      /* and OSX too ? */
 374                 struct ifaddr *ia;
 375
 376                 if_addr_rlock(ifp);
 377                 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 378                         if (ia->ifa_addr->sa_family != AF_INET)
 379                                 continue;
 380                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 381                             (ia->ifa_addr))->sin_addr.s_addr) {
 382                                 if_addr_runlock(ifp);
 383                                 return(1);      /* match */
 384                         }
 385                 }
 386                 if_addr_runlock(ifp);
 387 #endif /* __FreeBSD__ */
 388         }
 389         return(0);      /* no match, fail ... */
 390 }
 391
 392 /*
 393  * The verify_path function checks if a route to the src exists and
 394  * if it is reachable via ifp (when provided).
 395  *
 396  * The 'verrevpath' option checks that the interface that an IP packet
 397  * arrives on is the same interface that traffic destined for the
 398  * packet's source address would be routed out of.
 399  * The 'versrcreach' option just checks that the source address is
 400  * reachable via any route (except default) in the routing table.
 401  * These two are a measure to block forged packets. This is also
 402  * commonly known as "anti-spoofing" or Unicast Reverse Path
 403  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 404  * is purposely reminiscent of the Cisco IOS command,
 405  *
 406  *   ip verify unicast reverse-path
 407  *   ip verify unicast source reachable-via any
 408  *
 409  * which implements the same functionality. But note that the syntax
 410  * is misleading, and the check may be performed on all IP packets
 411  * whether unicast, multicast, or broadcast.
 412  */
 413 static int
 414 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 415 {
 416 #ifndef __FreeBSD__
 417         return 0;
 418 #else
 419         struct route ro;
 420         struct sockaddr_in *dst;
 421
 422         bzero(&ro, sizeof(ro));
 423
 424         dst = (struct sockaddr_in *)&(ro.ro_dst);
 425         dst->sin_family = AF_INET;
 426         dst->sin_len = sizeof(*dst);
 427         dst->sin_addr = src;
 428         in_rtalloc_ign(&ro, 0, fib);
 429
 430         if (ro.ro_rt == NULL)
 431                 return 0;
 432
 433         /*
 434          * If ifp is provided, check for equality with rtentry.
 435          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 436          * in order to pass packets injected back by if_simloop():
 437          * if useloopback == 1 routing entry (via lo0) for our own address
 438          * may exist, so we need to handle routing assymetry.
 439          */
 440         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 441                 RTFREE(ro.ro_rt);
 442                 return 0;
 443         }
 444
 445         /* if no ifp provided, check if rtentry is not default route */
 446         if (ifp == NULL &&
 447              satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
 448                 RTFREE(ro.ro_rt);
 449                 return 0;
 450         }
 451
 452         /* or if this is a blackhole/reject route */
 453         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 454                 RTFREE(ro.ro_rt);
 455                 return 0;
 456         }
 457
 458         /* found valid route */
 459         RTFREE(ro.ro_rt);
 460         return 1;
 461 #endif /* __FreeBSD__ */
 462 }
 463
 464 #ifdef INET6
 465 /*
 466  * ipv6 specific rules here...
 467  */
 468 static __inline int
 469 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 470 {
 471         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 472 }
 473
 474 static int
 475 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 476 {
 477         int i;
 478         for (i=0; i <= cmd->o.arg1; ++i )
 479                 if (curr_flow == cmd->d[i] )
 480                         return 1;
 481         return 0;
 482 }
 483
 484 /* support for IP6_*_ME opcodes */
 485 static int
 486 search_ip6_addr_net (struct in6_addr * ip6_addr)
 487 {
 488         struct ifnet *mdc;
 489         struct ifaddr *mdc2;
 490         struct in6_ifaddr *fdm;
 491         struct in6_addr copia;
 492
 493         TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
 494                 if_addr_rlock(mdc);
 495                 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
 496                         if (mdc2->ifa_addr->sa_family == AF_INET6) {
 497                                 fdm = (struct in6_ifaddr *)mdc2;
 498                                 copia = fdm->ia_addr.sin6_addr;
 499                                 /* need for leaving scope_id in the sock_addr */
 500                                 in6_clearscope(&copia);
 501                                 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
 502                                         if_addr_runlock(mdc);
 503                                         return 1;
 504                                 }
 505                         }
 506                 }
 507                 if_addr_runlock(mdc);
 508         }
 509         return 0;
 510 }
 511
 512 static int
 513 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 514 {
 515         struct route_in6 ro;
 516         struct sockaddr_in6 *dst;
 517
 518         bzero(&ro, sizeof(ro));
 519
 520         dst = (struct sockaddr_in6 * )&(ro.ro_dst);
 521         dst->sin6_family = AF_INET6;
 522         dst->sin6_len = sizeof(*dst);
 523         dst->sin6_addr = *src;
 524
 525         in6_rtalloc_ign(&ro, 0, fib);
 526         if (ro.ro_rt == NULL)
 527                 return 0;
 528
 529         /*
 530          * if ifp is provided, check for equality with rtentry
 531          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 532          * to support the case of sending packets to an address of our own.
 533          * (where the former interface is the first argument of if_simloop()
 534          *  (=ifp), the latter is lo0)
 535          */
 536         if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
 537                 RTFREE(ro.ro_rt);
 538                 return 0;
 539         }
 540
 541         /* if no ifp provided, check if rtentry is not default route */
 542         if (ifp == NULL &&
 543             IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
 544                 RTFREE(ro.ro_rt);
 545                 return 0;
 546         }
 547
 548         /* or if this is a blackhole/reject route */
 549         if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
 550                 RTFREE(ro.ro_rt);
 551                 return 0;
 552         }
 553
 554         /* found valid route */
 555         RTFREE(ro.ro_rt);
 556         return 1;
 557
 558 }
 559
 560 static int
 561 is_icmp6_query(int icmp6_type)
 562 {
 563         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 564             (icmp6_type == ICMP6_ECHO_REQUEST ||
 565             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 566             icmp6_type == ICMP6_WRUREQUEST ||
 567             icmp6_type == ICMP6_FQDN_QUERY ||
 568             icmp6_type == ICMP6_NI_QUERY))
 569                 return (1);
 570
 571         return (0);
 572 }
 573
 574 static void
 575 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 576 {
 577         struct mbuf *m;
 578
 579         m = args->m;
 580         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 581                 struct tcphdr *tcp;
 582                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 583
 584                 if ((tcp->th_flags & TH_RST) == 0) {
 585                         struct mbuf *m0;
 586                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 587                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 588                             tcp->th_flags | TH_RST);
 589                         if (m0 != NULL)
 590                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 591                                     NULL);
 592                 }
 593                 FREE_PKT(m);
 594         } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
 595 #if 0
 596                 /*
 597                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 598                  * as the contents are read. We need to m_adj() the
 599                  * needed amount.
 600                  * The mbuf will however be thrown away so we can adjust it.
 601                  * Remember we did an m_pullup on it already so we
 602                  * can make some assumptions about contiguousness.
 603                  */
 604                 if (args->L3offset)
 605                         m_adj(m, args->L3offset);
 606 #endif
 607                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 608         } else
 609                 FREE_PKT(m);
 610
 611         args->m = NULL;
 612 }
 613
 614 #endif /* INET6 */
 615
 616
 617 /*
 618  * sends a reject message, consuming the mbuf passed as an argument.
 619  */
 620 static void
 621 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 622 {
 623
 624 #if 0
 625         /* XXX When ip is not guaranteed to be at mtod() we will
 626          * need to account for this */
 627          * The mbuf will however be thrown away so we can adjust it.
 628          * Remember we did an m_pullup on it already so we
 629          * can make some assumptions about contiguousness.
 630          */
 631         if (args->L3offset)
 632                 m_adj(m, args->L3offset);
 633 #endif
 634         if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
 635                 /* We need the IP header in host order for icmp_error(). */
 636                 SET_HOST_IPLEN(ip);
 637                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 638         } else if (args->f_id.proto == IPPROTO_TCP) {
 639                 struct tcphdr *const tcp =
 640                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 641                 if ( (tcp->th_flags & TH_RST) == 0) {
 642                         struct mbuf *m;
 643                         m = ipfw_send_pkt(args->m, &(args->f_id),
 644                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 645                                 tcp->th_flags | TH_RST);
 646                         if (m != NULL)
 647                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 648                 }
 649                 FREE_PKT(args->m);
 650         } else
 651                 FREE_PKT(args->m);
 652         args->m = NULL;
 653 }
 654
 655 /*
 656  * Support for uid/gid/jail lookup. These tests are expensive
 657  * (because we may need to look into the list of active sockets)
 658  * so we cache the results. ugid_lookupp is 0 if we have not
 659  * yet done a lookup, 1 if we succeeded, and -1 if we tried
 660  * and failed. The function always returns the match value.
 661  * We could actually spare the variable and use *uc, setting
 662  * it to '(void *)check_uidgid if we have no info, NULL if
 663  * we tried and failed, or any other value if successful.
 664  */
 665 static int
 666 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
 667     struct ucred **uc)
 668 {
 669 #ifndef __FreeBSD__
 670         /* XXX */
 671         return cred_check(insn, proto, oif,
 672             dst_ip, dst_port, src_ip, src_port,
 673             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
 674 #else  /* FreeBSD */
 675         struct in_addr src_ip, dst_ip;
 676         struct inpcbinfo *pi;
 677         struct ipfw_flow_id *id;
 678         struct inpcb *pcb, *inp;
 679         struct ifnet *oif;
 680         int lookupflags;
 681         int match;
 682
 683         id = &args->f_id;
 684         inp = args->inp;
 685         oif = args->oif;
 686
 687         /*
 688          * Check to see if the UDP or TCP stack supplied us with
 689          * the PCB. If so, rather then holding a lock and looking
 690          * up the PCB, we can use the one that was supplied.
 691          */
 692         if (inp && *ugid_lookupp == 0) {
 693                 INP_LOCK_ASSERT(inp);
 694                 if (inp->inp_socket != NULL) {
 695                         *uc = crhold(inp->inp_cred);
 696                         *ugid_lookupp = 1;
 697                 } else
 698                         *ugid_lookupp = -1;
 699         }
 700         /*
 701          * If we have already been here and the packet has no
 702          * PCB entry associated with it, then we can safely
 703          * assume that this is a no match.
 704          */
 705         if (*ugid_lookupp == -1)
 706                 return (0);
 707         if (id->proto == IPPROTO_TCP) {
 708                 lookupflags = 0;
 709                 pi = &V_tcbinfo;
 710         } else if (id->proto == IPPROTO_UDP) {
 711                 lookupflags = INPLOOKUP_WILDCARD;
 712                 pi = &V_udbinfo;
 713         } else
 714                 return 0;
 715         lookupflags |= INPLOOKUP_RLOCKPCB;
 716         match = 0;
 717         if (*ugid_lookupp == 0) {
 718                 if (id->addr_type == 6) {
 719 #ifdef INET6
 720                         if (oif == NULL)
 721                                 pcb = in6_pcblookup_mbuf(pi,
 722                                     &id->src_ip6, htons(id->src_port),
 723                                     &id->dst_ip6, htons(id->dst_port),
 724                                     lookupflags, oif, args->m);
 725                         else
 726                                 pcb = in6_pcblookup_mbuf(pi,
 727                                     &id->dst_ip6, htons(id->dst_port),
 728                                     &id->src_ip6, htons(id->src_port),
 729                                     lookupflags, oif, args->m);
 730 #else
 731                         *ugid_lookupp = -1;
 732                         return (0);
 733 #endif
 734                 } else {
 735                         src_ip.s_addr = htonl(id->src_ip);
 736                         dst_ip.s_addr = htonl(id->dst_ip);
 737                         if (oif == NULL)
 738                                 pcb = in_pcblookup_mbuf(pi,
 739                                     src_ip, htons(id->src_port),
 740                                     dst_ip, htons(id->dst_port),
 741                                     lookupflags, oif, args->m);
 742                         else
 743                                 pcb = in_pcblookup_mbuf(pi,
 744                                     dst_ip, htons(id->dst_port),
 745                                     src_ip, htons(id->src_port),
 746                                     lookupflags, oif, args->m);
 747                 }
 748                 if (pcb != NULL) {
 749                         INP_RLOCK_ASSERT(pcb);
 750                         *uc = crhold(pcb->inp_cred);
 751                         *ugid_lookupp = 1;
 752                         INP_RUNLOCK(pcb);
 753                 }
 754                 if (*ugid_lookupp == 0) {
 755                         /*
 756                          * We tried and failed, set the variable to -1
 757                          * so we will not try again on this packet.
 758                          */
 759                         *ugid_lookupp = -1;
 760                         return (0);
 761                 }
 762         }
 763         if (insn->o.opcode == O_UID)
 764                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
 765         else if (insn->o.opcode == O_GID)
 766                 match = groupmember((gid_t)insn->d[0], *uc);
 767         else if (insn->o.opcode == O_JAIL)
 768                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
 769         return (match);
 770 #endif /* __FreeBSD__ */
 771 }
 772
 773 /*
 774  * Helper function to set args with info on the rule after the matching
 775  * one. slot is precise, whereas we guess rule_id as they are
 776  * assigned sequentially.
 777  */
 778 static inline void
 779 set_match(struct ip_fw_args *args, int slot,
 780         struct ip_fw_chain *chain)
 781 {
 782         args->rule.chain_id = chain->id;
 783         args->rule.slot = slot + 1; /* we use 0 as a marker */
 784         args->rule.rule_id = 1 + chain->map[slot]->id;
 785         args->rule.rulenum = chain->map[slot]->rulenum;
 786 }
 787
 788 /*
 789  * Helper function to enable cached rule lookups using
 790  * x_next and next_rule fields in ipfw rule.
 791  */
 792 static int
 793 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 794     int tablearg, int jump_backwards)
 795 {
 796         int f_pos;
 797
 798         /* If possible use cached f_pos (in f->next_rule),
 799          * whose version is written in f->next_rule
 800          * (horrible hacks to avoid changing the ABI).
 801          */
 802         if (num != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id)
 803                 f_pos = (uintptr_t)f->next_rule;
 804         else {
 805                 int i = IP_FW_ARG_TABLEARG(num);
 806                 /* make sure we do not jump backward */
 807                 if (jump_backwards == 0 && i <= f->rulenum)
 808                         i = f->rulenum + 1;
 809                 f_pos = ipfw_find_rule(chain, i, 0);
 810                 /* update the cache */
 811                 if (num != IP_FW_TABLEARG) {
 812                         f->next_rule = (void *)(uintptr_t)f_pos;
 813                         f->x_next = (void *)(uintptr_t)chain->id;
 814                 }
 815         }
 816
 817         return (f_pos);
 818 }
 819
 820 /*
 821  * The main check routine for the firewall.
 822  *
 823  * All arguments are in args so we can modify them and return them
 824  * back to the caller.
 825  *
 826  * Parameters:
 827  *
 828  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
 829  *              Starts with the IP header.
 830  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
 831  *      args->L3offset  Number of bytes bypassed if we came from L2.
 832  *                      e.g. often sizeof(eh)  ** NOTYET **
 833  *      args->oif       Outgoing interface, NULL if packet is incoming.
 834  *              The incoming interface is in the mbuf. (in)
 835  *      args->divert_rule (in/out)
 836  *              Skip up to the first rule past this rule number;
 837  *              upon return, non-zero port number for divert or tee.
 838  *
 839  *      args->rule      Pointer to the last matching rule (in/out)
 840  *      args->next_hop  Socket we are forwarding to (out).
 841  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
 842  *      args->f_id      Addresses grabbed from the packet (out)
 843  *      args->rule.info a cookie depending on rule action
 844  *
 845  * Return value:
 846  *
 847  *      IP_FW_PASS      the packet must be accepted
 848  *      IP_FW_DENY      the packet must be dropped
 849  *      IP_FW_DIVERT    divert packet, port in m_tag
 850  *      IP_FW_TEE       tee packet, port in m_tag
 851  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
 852  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
 853  *              args->rule contains the matching rule,
 854  *              args->rule.info has additional information.
 855  *
 856  */
 857 int
 858 ipfw_chk(struct ip_fw_args *args)
 859 {
 860
 861         /*
 862          * Local variables holding state while processing a packet:
 863          *
 864          * IMPORTANT NOTE: to speed up the processing of rules, there
 865          * are some assumption on the values of the variables, which
 866          * are documented here. Should you change them, please check
 867          * the implementation of the various instructions to make sure
 868          * that they still work.
 869          *
 870          * args->eh     The MAC header. It is non-null for a layer2
 871          *      packet, it is NULL for a layer-3 packet.
 872          * **notyet**
 873          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
 874          *
 875          * m | args->m  Pointer to the mbuf, as received from the caller.
 876          *      It may change if ipfw_chk() does an m_pullup, or if it
 877          *      consumes the packet because it calls send_reject().
 878          *      XXX This has to change, so that ipfw_chk() never modifies
 879          *      or consumes the buffer.
 880          * ip   is the beginning of the ip(4 or 6) header.
 881          *      Calculated by adding the L3offset to the start of data.
 882          *      (Until we start using L3offset, the packet is
 883          *      supposed to start with the ip header).
 884          */
 885         struct mbuf *m = args->m;
 886         struct ip *ip = mtod(m, struct ip *);
 887
 888         /*
 889          * For rules which contain uid/gid or jail constraints, cache
 890          * a copy of the users credentials after the pcb lookup has been
 891          * executed. This will speed up the processing of rules with
 892          * these types of constraints, as well as decrease contention
 893          * on pcb related locks.
 894          */
 895 #ifndef __FreeBSD__
 896         struct bsd_ucred ucred_cache;
 897 #else
 898         struct ucred *ucred_cache = NULL;
 899 #endif
 900         int ucred_lookup = 0;
 901
 902         /*
 903          * oif | args->oif      If NULL, ipfw_chk has been called on the
 904          *      inbound path (ether_input, ip_input).
 905          *      If non-NULL, ipfw_chk has been called on the outbound path
 906          *      (ether_output, ip_output).
 907          */
 908         struct ifnet *oif = args->oif;
 909
 910         int f_pos = 0;          /* index of current rule in the array */
 911         int retval = 0;
 912
 913         /*
 914          * hlen The length of the IP header.
 915          */
 916         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
 917
 918         /*
 919          * offset       The offset of a fragment. offset != 0 means that
 920          *      we have a fragment at this offset of an IPv4 packet.
 921          *      offset == 0 means that (if this is an IPv4 packet)
 922          *      this is the first or only fragment.
 923          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
 924          *      or there is a single packet fragement (fragement header added
 925          *      without needed).  We will treat a single packet fragment as if
 926          *      there was no fragment header (or log/block depending on the
 927          *      V_fw_permit_single_frag6 sysctl setting).
 928          */
 929         u_short offset = 0;
 930         u_short ip6f_mf = 0;
 931
 932         /*
 933          * Local copies of addresses. They are only valid if we have
 934          * an IP packet.
 935          *
 936          * proto        The protocol. Set to 0 for non-ip packets,
 937          *      or to the protocol read from the packet otherwise.
 938          *      proto != 0 means that we have an IPv4 packet.
 939          *
 940          * src_port, dst_port   port numbers, in HOST format. Only
 941          *      valid for TCP and UDP packets.
 942          *
 943          * src_ip, dst_ip       ip addresses, in NETWORK format.
 944          *      Only valid for IPv4 packets.
 945          */
 946         uint8_t proto;
 947         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
 948         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
 949         uint16_t iplen=0;
 950         int pktlen;
 951         uint16_t        etype = 0;      /* Host order stored ether type */
 952
 953         /*
 954          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
 955          *      MATCH_NONE when checked and not matched (q = NULL),
 956          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
 957          */
 958         int dyn_dir = MATCH_UNKNOWN;
 959         ipfw_dyn_rule *q = NULL;
 960         struct ip_fw_chain *chain = &V_layer3_chain;
 961
 962         /*
 963          * We store in ulp a pointer to the upper layer protocol header.
 964          * In the ipv4 case this is easy to determine from the header,
 965          * but for ipv6 we might have some additional headers in the middle.
 966          * ulp is NULL if not found.
 967          */
 968         void *ulp = NULL;               /* upper layer protocol pointer. */
 969
 970         /* XXX ipv6 variables */
 971         int is_ipv6 = 0;
 972         uint8_t icmp6_type = 0;
 973         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
 974         /* end of ipv6 variables */
 975
 976         int is_ipv4 = 0;
 977
 978         int done = 0;           /* flag to exit the outer loop */
 979
 980         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
 981                 return (IP_FW_PASS);    /* accept */
 982
 983         dst_ip.s_addr = 0;              /* make sure it is initialized */
 984         src_ip.s_addr = 0;              /* make sure it is initialized */
 985         pktlen = m->m_pkthdr.len;
 986         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
 987         proto = args->f_id.proto = 0;   /* mark f_id invalid */
 988                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
 989
 990 /*
 991  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
 992  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
 993  * pointer might become stale after other pullups (but we never use it
 994  * this way).
 995  */
 996 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
 997 #define PULLUP_LEN(_len, p, T)                                  \
 998 do {                                                            \
 999         int x = (_len) + T;                                     \
1000         if ((m)->m_len < x) {                                   \
1001                 args->m = m = m_pullup(m, x);                   \
1002                 if (m == NULL)                                  \
1003                         goto pullup_failed;                     \
1004         }                                                       \
1005         p = (mtod(m, char *) + (_len));                         \
1006 } while (0)
1007
1008         /*
1009          * if we have an ether header,
1010          */
1011         if (args->eh)
1012                 etype = ntohs(args->eh->ether_type);
1013
1014         /* Identify IP packets and fill up variables. */
1015         if (pktlen >= sizeof(struct ip6_hdr) &&
1016             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1017                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1018                 is_ipv6 = 1;
1019                 args->f_id.addr_type = 6;
1020                 hlen = sizeof(struct ip6_hdr);
1021                 proto = ip6->ip6_nxt;
1022
1023                 /* Search extension headers to find upper layer protocols */
1024                 while (ulp == NULL && offset == 0) {
1025                         switch (proto) {
1026                         case IPPROTO_ICMPV6:
1027                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1028                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1029                                 break;
1030
1031                         case IPPROTO_TCP:
1032                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1033                                 dst_port = TCP(ulp)->th_dport;
1034                                 src_port = TCP(ulp)->th_sport;
1035                                 /* save flags for dynamic rules */
1036                                 args->f_id._flags = TCP(ulp)->th_flags;
1037                                 break;
1038
1039                         case IPPROTO_SCTP:
1040                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1041                                 src_port = SCTP(ulp)->src_port;
1042                                 dst_port = SCTP(ulp)->dest_port;
1043                                 break;
1044
1045                         case IPPROTO_UDP:
1046                                 PULLUP_TO(hlen, ulp, struct udphdr);
1047                                 dst_port = UDP(ulp)->uh_dport;
1048                                 src_port = UDP(ulp)->uh_sport;
1049                                 break;
1050
1051                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1052                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1053                                 ext_hd |= EXT_HOPOPTS;
1054                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1055                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1056                                 ulp = NULL;
1057                                 break;
1058
1059                         case IPPROTO_ROUTING:   /* RFC 2460 */
1060                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1061                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1062                                 case 0:
1063                                         ext_hd |= EXT_RTHDR0;
1064                                         break;
1065                                 case 2:
1066                                         ext_hd |= EXT_RTHDR2;
1067                                         break;
1068                                 default:
1069                                         if (V_fw_verbose)
1070                                                 printf("IPFW2: IPV6 - Unknown "
1071                                                     "Routing Header type(%d)\n",
1072                                                     ((struct ip6_rthdr *)
1073                                                     ulp)->ip6r_type);
1074                                         if (V_fw_deny_unknown_exthdrs)
1075                                             return (IP_FW_DENY);
1076                                         break;
1077                                 }
1078                                 ext_hd |= EXT_ROUTING;
1079                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1080                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1081                                 ulp = NULL;
1082                                 break;
1083
1084                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1085                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1086                                 ext_hd |= EXT_FRAGMENT;
1087                                 hlen += sizeof (struct ip6_frag);
1088                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1089                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1090                                         IP6F_OFF_MASK;
1091                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1092                                         IP6F_MORE_FRAG;
1093                                 if (V_fw_permit_single_frag6 == 0 &&
1094                                     offset == 0 && ip6f_mf == 0) {
1095                                         if (V_fw_verbose)
1096                                                 printf("IPFW2: IPV6 - Invalid "
1097                                                     "Fragment Header\n");
1098                                         if (V_fw_deny_unknown_exthdrs)
1099                                             return (IP_FW_DENY);
1100                                         break;
1101                                 }
1102                                 args->f_id.extra =
1103                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1104                                 ulp = NULL;
1105                                 break;
1106
1107                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1108                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1109                                 ext_hd |= EXT_DSTOPTS;
1110                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1111                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1112                                 ulp = NULL;
1113                                 break;
1114
1115                         case IPPROTO_AH:        /* RFC 2402 */
1116                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1117                                 ext_hd |= EXT_AH;
1118                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1119                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1120                                 ulp = NULL;
1121                                 break;
1122
1123                         case IPPROTO_ESP:       /* RFC 2406 */
1124                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1125                                 /* Anything past Seq# is variable length and
1126                                  * data past this ext. header is encrypted. */
1127                                 ext_hd |= EXT_ESP;
1128                                 break;
1129
1130                         case IPPROTO_NONE:      /* RFC 2460 */
1131                                 /*
1132                                  * Packet ends here, and IPv6 header has
1133                                  * already been pulled up. If ip6e_len!=0
1134                                  * then octets must be ignored.
1135                                  */
1136                                 ulp = ip; /* non-NULL to get out of loop. */
1137                                 break;
1138
1139                         case IPPROTO_OSPFIGP:
1140                                 /* XXX OSPF header check? */
1141                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1142                                 break;
1143
1144                         case IPPROTO_PIM:
1145                                 /* XXX PIM header check? */
1146                                 PULLUP_TO(hlen, ulp, struct pim);
1147                                 break;
1148
1149                         case IPPROTO_GRE:       /* RFC 1701 */
1150                                 /* XXX GRE header check? */
1151                                 PULLUP_TO(hlen, ulp, struct gre_h);
1152                                 break;
1153
1154                         case IPPROTO_CARP:
1155                                 PULLUP_TO(hlen, ulp, struct carp_header);
1156                                 if (((struct carp_header *)ulp)->carp_version !=
1157                                     CARP_VERSION)
1158                                         return (IP_FW_DENY);
1159                                 if (((struct carp_header *)ulp)->carp_type !=
1160                                     CARP_ADVERTISEMENT)
1161                                         return (IP_FW_DENY);
1162                                 break;
1163
1164                         case IPPROTO_IPV6:      /* RFC 2893 */
1165                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1166                                 break;
1167
1168                         case IPPROTO_IPV4:      /* RFC 2893 */
1169                                 PULLUP_TO(hlen, ulp, struct ip);
1170                                 break;
1171
1172                         default:
1173                                 if (V_fw_verbose)
1174                                         printf("IPFW2: IPV6 - Unknown "
1175                                             "Extension Header(%d), ext_hd=%x\n",
1176                                              proto, ext_hd);
1177                                 if (V_fw_deny_unknown_exthdrs)
1178                                     return (IP_FW_DENY);
1179                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1180                                 break;
1181                         } /*switch */
1182                 }
1183                 ip = mtod(m, struct ip *);
1184                 ip6 = (struct ip6_hdr *)ip;
1185                 args->f_id.src_ip6 = ip6->ip6_src;
1186                 args->f_id.dst_ip6 = ip6->ip6_dst;
1187                 args->f_id.src_ip = 0;
1188                 args->f_id.dst_ip = 0;
1189                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1190         } else if (pktlen >= sizeof(struct ip) &&
1191             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1192                 is_ipv4 = 1;
1193                 hlen = ip->ip_hl << 2;
1194                 args->f_id.addr_type = 4;
1195
1196                 /*
1197                  * Collect parameters into local variables for faster matching.
1198                  */
1199                 proto = ip->ip_p;
1200                 src_ip = ip->ip_src;
1201                 dst_ip = ip->ip_dst;
1202                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1203                 iplen = ntohs(ip->ip_len);
1204                 pktlen = iplen < pktlen ? iplen : pktlen;
1205
1206                 if (offset == 0) {
1207                         switch (proto) {
1208                         case IPPROTO_TCP:
1209                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1210                                 dst_port = TCP(ulp)->th_dport;
1211                                 src_port = TCP(ulp)->th_sport;
1212                                 /* save flags for dynamic rules */
1213                                 args->f_id._flags = TCP(ulp)->th_flags;
1214                                 break;
1215
1216                         case IPPROTO_SCTP:
1217                                 PULLUP_TO(hlen, ulp, struct sctphdr);
1218                                 src_port = SCTP(ulp)->src_port;
1219                                 dst_port = SCTP(ulp)->dest_port;
1220                                 break;
1221
1222                         case IPPROTO_UDP:
1223                                 PULLUP_TO(hlen, ulp, struct udphdr);
1224                                 dst_port = UDP(ulp)->uh_dport;
1225                                 src_port = UDP(ulp)->uh_sport;
1226                                 break;
1227
1228                         case IPPROTO_ICMP:
1229                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1230                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1231                                 break;
1232
1233                         default:
1234                                 break;
1235                         }
1236                 }
1237
1238                 ip = mtod(m, struct ip *);
1239                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1240                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1241         }
1242 #undef PULLUP_TO
1243         if (proto) { /* we may have port numbers, store them */
1244                 args->f_id.proto = proto;
1245                 args->f_id.src_port = src_port = ntohs(src_port);
1246                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1247         }
1248
1249         IPFW_RLOCK(chain);
1250         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1251                 IPFW_RUNLOCK(chain);
1252                 return (IP_FW_PASS);    /* accept */
1253         }
1254         if (args->rule.slot) {
1255                 /*
1256                  * Packet has already been tagged as a result of a previous
1257                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1258                  * REASS, NETGRAPH, DIVERT/TEE...)
1259                  * Validate the slot and continue from the next one
1260                  * if still present, otherwise do a lookup.
1261                  */
1262                 f_pos = (args->rule.chain_id == chain->id) ?
1263                     args->rule.slot :
1264                     ipfw_find_rule(chain, args->rule.rulenum,
1265                         args->rule.rule_id);
1266         } else {
1267                 f_pos = 0;
1268         }
1269
1270         /*
1271          * Now scan the rules, and parse microinstructions for each rule.
1272          * We have two nested loops and an inner switch. Sometimes we
1273          * need to break out of one or both loops, or re-enter one of
1274          * the loops with updated variables. Loop variables are:
1275          *
1276          *      f_pos (outer loop) points to the current rule.
1277          *              On output it points to the matching rule.
1278          *      done (outer loop) is used as a flag to break the loop.
1279          *      l (inner loop)  residual length of current rule.
1280          *              cmd points to the current microinstruction.
1281          *
1282          * We break the inner loop by setting l=0 and possibly
1283          * cmdlen=0 if we don't want to advance cmd.
1284          * We break the outer loop by setting done=1
1285          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1286          * as needed.
1287          */
1288         for (; f_pos < chain->n_rules; f_pos++) {
1289                 ipfw_insn *cmd;
1290                 uint32_t tablearg = 0;
1291                 int l, cmdlen, skip_or; /* skip rest of OR block */
1292                 struct ip_fw *f;
1293
1294                 f = chain->map[f_pos];
1295                 if (V_set_disable & (1 << f->set) )
1296                         continue;
1297
1298                 skip_or = 0;
1299                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1300                     l -= cmdlen, cmd += cmdlen) {
1301                         int match;
1302
1303                         /*
1304                          * check_body is a jump target used when we find a
1305                          * CHECK_STATE, and need to jump to the body of
1306                          * the target rule.
1307                          */
1308
1309 /* check_body: */
1310                         cmdlen = F_LEN(cmd);
1311                         /*
1312                          * An OR block (insn_1 || .. || insn_n) has the
1313                          * F_OR bit set in all but the last instruction.
1314                          * The first match will set "skip_or", and cause
1315                          * the following instructions to be skipped until
1316                          * past the one with the F_OR bit clear.
1317                          */
1318                         if (skip_or) {          /* skip this instruction */
1319                                 if ((cmd->len & F_OR) == 0)
1320                                         skip_or = 0;    /* next one is good */
1321                                 continue;
1322                         }
1323                         match = 0; /* set to 1 if we succeed */
1324
1325                         switch (cmd->opcode) {
1326                         /*
1327                          * The first set of opcodes compares the packet's
1328                          * fields with some pattern, setting 'match' if a
1329                          * match is found. At the end of the loop there is
1330                          * logic to deal with F_NOT and F_OR flags associated
1331                          * with the opcode.
1332                          */
1333                         case O_NOP:
1334                                 match = 1;
1335                                 break;
1336
1337                         case O_FORWARD_MAC:
1338                                 printf("ipfw: opcode %d unimplemented\n",
1339                                     cmd->opcode);
1340                                 break;
1341
1342                         case O_GID:
1343                         case O_UID:
1344                         case O_JAIL:
1345                                 /*
1346                                  * We only check offset == 0 && proto != 0,
1347                                  * as this ensures that we have a
1348                                  * packet with the ports info.
1349                                  */
1350                                 if (offset != 0)
1351                                         break;
1352                                 if (proto == IPPROTO_TCP ||
1353                                     proto == IPPROTO_UDP)
1354                                         match = check_uidgid(
1355                                                     (ipfw_insn_u32 *)cmd,
1356                                                     args, &ucred_lookup,
1357 #ifdef __FreeBSD__
1358                                                     &ucred_cache);
1359 #else
1360                                                     (void *)&ucred_cache);
1361 #endif
1362                                 break;
1363
1364                         case O_RECV:
1365                                 match = iface_match(m->m_pkthdr.rcvif,
1366                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1367                                 break;
1368
1369                         case O_XMIT:
1370                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1371                                     chain, &tablearg);
1372                                 break;
1373
1374                         case O_VIA:
1375                                 match = iface_match(oif ? oif :
1376                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1377                                     chain, &tablearg);
1378                                 break;
1379
1380                         case O_MACADDR2:
1381                                 if (args->eh != NULL) { /* have MAC header */
1382                                         u_int32_t *want = (u_int32_t *)
1383                                                 ((ipfw_insn_mac *)cmd)->addr;
1384                                         u_int32_t *mask = (u_int32_t *)
1385                                                 ((ipfw_insn_mac *)cmd)->mask;
1386                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1387
1388                                         match =
1389                                             ( want[0] == (hdr[0] & mask[0]) &&
1390                                               want[1] == (hdr[1] & mask[1]) &&
1391                                               want[2] == (hdr[2] & mask[2]) );
1392                                 }
1393                                 break;
1394
1395                         case O_MAC_TYPE:
1396                                 if (args->eh != NULL) {
1397                                         u_int16_t *p =
1398                                             ((ipfw_insn_u16 *)cmd)->ports;
1399                                         int i;
1400
1401                                         for (i = cmdlen - 1; !match && i>0;
1402                                             i--, p += 2)
1403                                                 match = (etype >= p[0] &&
1404                                                     etype <= p[1]);
1405                                 }
1406                                 break;
1407
1408                         case O_FRAG:
1409                                 match = (offset != 0);
1410                                 break;
1411
1412                         case O_IN:      /* "out" is "not in" */
1413                                 match = (oif == NULL);
1414                                 break;
1415
1416                         case O_LAYER2:
1417                                 match = (args->eh != NULL);
1418                                 break;
1419
1420                         case O_DIVERTED:
1421                             {
1422                                 /* For diverted packets, args->rule.info
1423                                  * contains the divert port (in host format)
1424                                  * reason and direction.
1425                                  */
1426                                 uint32_t i = args->rule.info;
1427                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1428                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1429                             }
1430                                 break;
1431
1432                         case O_PROTO:
1433                                 /*
1434                                  * We do not allow an arg of 0 so the
1435                                  * check of "proto" only suffices.
1436                                  */
1437                                 match = (proto == cmd->arg1);
1438                                 break;
1439
1440                         case O_IP_SRC:
1441                                 match = is_ipv4 &&
1442                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1443                                     src_ip.s_addr);
1444                                 break;
1445
1446                         case O_IP_SRC_LOOKUP:
1447                         case O_IP_DST_LOOKUP:
1448                                 if (is_ipv4) {
1449                                     uint32_t key =
1450                                         (cmd->opcode == O_IP_DST_LOOKUP) ?
1451                                             dst_ip.s_addr : src_ip.s_addr;
1452                                     uint32_t v = 0;
1453
1454                                     if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1455                                         /* generic lookup. The key must be
1456                                          * in 32bit big-endian format.
1457                                          */
1458                                         v = ((ipfw_insn_u32 *)cmd)->d[1];
1459                                         if (v == 0)
1460                                             key = dst_ip.s_addr;
1461                                         else if (v == 1)
1462                                             key = src_ip.s_addr;
1463                                         else if (v == 6) /* dscp */
1464                                             key = (ip->ip_tos >> 2) & 0x3f;
1465                                         else if (offset != 0)
1466                                             break;
1467                                         else if (proto != IPPROTO_TCP &&
1468                                                 proto != IPPROTO_UDP)
1469                                             break;
1470                                         else if (v == 2)
1471                                             key = htonl(dst_port);
1472                                         else if (v == 3)
1473                                             key = htonl(src_port);
1474                                         else if (v == 4 || v == 5) {
1475                                             check_uidgid(
1476                                                 (ipfw_insn_u32 *)cmd,
1477                                                 args, &ucred_lookup,
1478 #ifdef __FreeBSD__
1479                                                 &ucred_cache);
1480                                             if (v == 4 /* O_UID */)
1481                                                 key = ucred_cache->cr_uid;
1482                                             else if (v == 5 /* O_JAIL */)
1483                                                 key = ucred_cache->cr_prison->pr_id;
1484 #else /* !__FreeBSD__ */
1485                                                 (void *)&ucred_cache);
1486                                             if (v ==4 /* O_UID */)
1487                                                 key = ucred_cache.uid;
1488                                             else if (v == 5 /* O_JAIL */)
1489                                                 key = ucred_cache.xid;
1490 #endif /* !__FreeBSD__ */
1491                                             key = htonl(key);
1492                                         } else
1493                                             break;
1494                                     }
1495                                     match = ipfw_lookup_table(chain,
1496                                         cmd->arg1, key, &v);
1497                                     if (!match)
1498                                         break;
1499                                     if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1500                                         match =
1501                                             ((ipfw_insn_u32 *)cmd)->d[0] == v;
1502                                     else
1503                                         tablearg = v;
1504                                 } else if (is_ipv6) {
1505                                         uint32_t v = 0;
1506                                         void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1507                                                 &args->f_id.dst_ip6: &args->f_id.src_ip6;
1508                                         match = ipfw_lookup_table_extended(chain,
1509                                                         cmd->arg1, pkey, &v,
1510                                                         IPFW_TABLE_CIDR);
1511                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1512                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1513                                         if (match)
1514                                                 tablearg = v;
1515                                 }
1516                                 break;
1517
1518                         case O_IP_SRC_MASK:
1519                         case O_IP_DST_MASK:
1520                                 if (is_ipv4) {
1521                                     uint32_t a =
1522                                         (cmd->opcode == O_IP_DST_MASK) ?
1523                                             dst_ip.s_addr : src_ip.s_addr;
1524                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1525                                     int i = cmdlen-1;
1526
1527                                     for (; !match && i>0; i-= 2, p+= 2)
1528                                         match = (p[0] == (a & p[1]));
1529                                 }
1530                                 break;
1531
1532                         case O_IP_SRC_ME:
1533                                 if (is_ipv4) {
1534                                         struct ifnet *tif;
1535
1536                                         INADDR_TO_IFP(src_ip, tif);
1537                                         match = (tif != NULL);
1538                                         break;
1539                                 }
1540 #ifdef INET6
1541                                 /* FALLTHROUGH */
1542                         case O_IP6_SRC_ME:
1543                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1544 #endif
1545                                 break;
1546
1547                         case O_IP_DST_SET:
1548                         case O_IP_SRC_SET:
1549                                 if (is_ipv4) {
1550                                         u_int32_t *d = (u_int32_t *)(cmd+1);
1551                                         u_int32_t addr =
1552                                             cmd->opcode == O_IP_DST_SET ?
1553                                                 args->f_id.dst_ip :
1554                                                 args->f_id.src_ip;
1555
1556                                             if (addr < d[0])
1557                                                     break;
1558                                             addr -= d[0]; /* subtract base */
1559                                             match = (addr < cmd->arg1) &&
1560                                                 ( d[ 1 + (addr>>5)] &
1561                                                   (1<<(addr & 0x1f)) );
1562                                 }
1563                                 break;
1564
1565                         case O_IP_DST:
1566                                 match = is_ipv4 &&
1567                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1568                                     dst_ip.s_addr);
1569                                 break;
1570
1571                         case O_IP_DST_ME:
1572                                 if (is_ipv4) {
1573                                         struct ifnet *tif;
1574
1575                                         INADDR_TO_IFP(dst_ip, tif);
1576                                         match = (tif != NULL);
1577                                         break;
1578                                 }
1579 #ifdef INET6
1580                                 /* FALLTHROUGH */
1581                         case O_IP6_DST_ME:
1582                                 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1583 #endif
1584                                 break;
1585
1586
1587                         case O_IP_SRCPORT:
1588                         case O_IP_DSTPORT:
1589                                 /*
1590                                  * offset == 0 && proto != 0 is enough
1591                                  * to guarantee that we have a
1592                                  * packet with port info.
1593                                  */
1594                                 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1595                                     && offset == 0) {
1596                                         u_int16_t x =
1597                                             (cmd->opcode == O_IP_SRCPORT) ?
1598                                                 src_port : dst_port ;
1599                                         u_int16_t *p =
1600                                             ((ipfw_insn_u16 *)cmd)->ports;
1601                                         int i;
1602
1603                                         for (i = cmdlen - 1; !match && i>0;
1604                                             i--, p += 2)
1605                                                 match = (x>=p[0] && x<=p[1]);
1606                                 }
1607                                 break;
1608
1609                         case O_ICMPTYPE:
1610                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
1611                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1612                                 break;
1613
1614 #ifdef INET6
1615                         case O_ICMP6TYPE:
1616                                 match = is_ipv6 && offset == 0 &&
1617                                     proto==IPPROTO_ICMPV6 &&
1618                                     icmp6type_match(
1619                                         ICMP6(ulp)->icmp6_type,
1620                                         (ipfw_insn_u32 *)cmd);
1621                                 break;
1622 #endif /* INET6 */
1623
1624                         case O_IPOPT:
1625                                 match = (is_ipv4 &&
1626                                     ipopts_match(ip, cmd) );
1627                                 break;
1628
1629                         case O_IPVER:
1630                                 match = (is_ipv4 &&
1631                                     cmd->arg1 == ip->ip_v);
1632                                 break;
1633
1634                         case O_IPID:
1635                         case O_IPLEN:
1636                         case O_IPTTL:
1637                                 if (is_ipv4) {  /* only for IP packets */
1638                                     uint16_t x;
1639                                     uint16_t *p;
1640                                     int i;
1641
1642                                     if (cmd->opcode == O_IPLEN)
1643                                         x = iplen;
1644                                     else if (cmd->opcode == O_IPTTL)
1645                                         x = ip->ip_ttl;
1646                                     else /* must be IPID */
1647                                         x = ntohs(ip->ip_id);
1648                                     if (cmdlen == 1) {
1649                                         match = (cmd->arg1 == x);
1650                                         break;
1651                                     }
1652                                     /* otherwise we have ranges */
1653                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1654                                     i = cmdlen - 1;
1655                                     for (; !match && i>0; i--, p += 2)
1656                                         match = (x >= p[0] && x <= p[1]);
1657                                 }
1658                                 break;
1659
1660                         case O_IPPRECEDENCE:
1661                                 match = (is_ipv4 &&
1662                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1663                                 break;
1664
1665                         case O_IPTOS:
1666                                 match = (is_ipv4 &&
1667                                     flags_match(cmd, ip->ip_tos));
1668                                 break;
1669
1670                         case O_DSCP:
1671                             {
1672                                 uint32_t *p;
1673                                 uint16_t x;
1674
1675                                 p = ((ipfw_insn_u32 *)cmd)->d;
1676
1677                                 if (is_ipv4)
1678                                         x = ip->ip_tos >> 2;
1679                                 else if (is_ipv6) {
1680                                         uint8_t *v;
1681                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
1682                                         x = (*v & 0x0F) << 2;
1683                                         v++;
1684                                         x |= *v >> 6;
1685                                 } else
1686                                         break;
1687
1688                                 /* DSCP bitmask is stored as low_u32 high_u32 */
1689                                 if (x >= 32)
1690                                         match = *(p + 1) & (1 << (x - 32));
1691                                 else
1692                                         match = *p & (1 << x);
1693                             }
1694                                 break;
1695
1696                         case O_TCPDATALEN:
1697                                 if (proto == IPPROTO_TCP && offset == 0) {
1698                                     struct tcphdr *tcp;
1699                                     uint16_t x;
1700                                     uint16_t *p;
1701                                     int i;
1702
1703                                     tcp = TCP(ulp);
1704                                     x = iplen -
1705                                         ((ip->ip_hl + tcp->th_off) << 2);
1706                                     if (cmdlen == 1) {
1707                                         match = (cmd->arg1 == x);
1708                                         break;
1709                                     }
1710                                     /* otherwise we have ranges */
1711                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1712                                     i = cmdlen - 1;
1713                                     for (; !match && i>0; i--, p += 2)
1714                                         match = (x >= p[0] && x <= p[1]);
1715                                 }
1716                                 break;
1717
1718                         case O_TCPFLAGS:
1719                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1720                                     flags_match(cmd, TCP(ulp)->th_flags));
1721                                 break;
1722
1723                         case O_TCPOPTS:
1724                                 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1725                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1726                                     tcpopts_match(TCP(ulp), cmd));
1727                                 break;
1728
1729                         case O_TCPSEQ:
1730                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1731                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1732                                         TCP(ulp)->th_seq);
1733                                 break;
1734
1735                         case O_TCPACK:
1736                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1737                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
1738                                         TCP(ulp)->th_ack);
1739                                 break;
1740
1741                         case O_TCPWIN:
1742                                 if (proto == IPPROTO_TCP && offset == 0) {
1743                                     uint16_t x;
1744                                     uint16_t *p;
1745                                     int i;
1746
1747                                     x = ntohs(TCP(ulp)->th_win);
1748                                     if (cmdlen == 1) {
1749                                         match = (cmd->arg1 == x);
1750                                         break;
1751                                     }
1752                                     /* Otherwise we have ranges. */
1753                                     p = ((ipfw_insn_u16 *)cmd)->ports;
1754                                     i = cmdlen - 1;
1755                                     for (; !match && i > 0; i--, p += 2)
1756                                         match = (x >= p[0] && x <= p[1]);
1757                                 }
1758                                 break;
1759
1760                         case O_ESTAB:
1761                                 /* reject packets which have SYN only */
1762                                 /* XXX should i also check for TH_ACK ? */
1763                                 match = (proto == IPPROTO_TCP && offset == 0 &&
1764                                     (TCP(ulp)->th_flags &
1765                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1766                                 break;
1767
1768                         case O_ALTQ: {
1769                                 struct pf_mtag *at;
1770                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1771
1772                                 match = 1;
1773                                 at = pf_find_mtag(m);
1774                                 if (at != NULL && at->qid != 0)
1775                                         break;
1776                                 at = pf_get_mtag(m);
1777                                 if (at == NULL) {
1778                                         /*
1779                                          * Let the packet fall back to the
1780                                          * default ALTQ.
1781                                          */
1782                                         break;
1783                                 }
1784                                 at->qid = altq->qid;
1785                                 at->hdr = ip;
1786                                 break;
1787                         }
1788
1789                         case O_LOG:
1790                                 ipfw_log(f, hlen, args, m,
1791                                     oif, offset | ip6f_mf, tablearg, ip);
1792                                 match = 1;
1793                                 break;
1794
1795                         case O_PROB:
1796                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1797                                 break;
1798
1799                         case O_VERREVPATH:
1800                                 /* Outgoing packets automatically pass/match */
1801                                 match = ((oif != NULL) ||
1802                                     (m->m_pkthdr.rcvif == NULL) ||
1803                                     (
1804 #ifdef INET6
1805                                     is_ipv6 ?
1806                                         verify_path6(&(args->f_id.src_ip6),
1807                                             m->m_pkthdr.rcvif, args->f_id.fib) :
1808 #endif
1809                                     verify_path(src_ip, m->m_pkthdr.rcvif,
1810                                         args->f_id.fib)));
1811                                 break;
1812
1813                         case O_VERSRCREACH:
1814                                 /* Outgoing packets automatically pass/match */
1815                                 match = (hlen > 0 && ((oif != NULL) ||
1816 #ifdef INET6
1817                                     is_ipv6 ?
1818                                         verify_path6(&(args->f_id.src_ip6),
1819                                             NULL, args->f_id.fib) :
1820 #endif
1821                                     verify_path(src_ip, NULL, args->f_id.fib)));
1822                                 break;
1823
1824                         case O_ANTISPOOF:
1825                                 /* Outgoing packets automatically pass/match */
1826                                 if (oif == NULL && hlen > 0 &&
1827                                     (  (is_ipv4 && in_localaddr(src_ip))
1828 #ifdef INET6
1829                                     || (is_ipv6 &&
1830                                         in6_localaddr(&(args->f_id.src_ip6)))
1831 #endif
1832                                     ))
1833                                         match =
1834 #ifdef INET6
1835                                             is_ipv6 ? verify_path6(
1836                                                 &(args->f_id.src_ip6),
1837                                                 m->m_pkthdr.rcvif,
1838                                                 args->f_id.fib) :
1839 #endif
1840                                             verify_path(src_ip,
1841                                                 m->m_pkthdr.rcvif,
1842                                                 args->f_id.fib);
1843                                 else
1844                                         match = 1;
1845                                 break;
1846
1847                         case O_IPSEC:
1848 #ifdef IPSEC
1849                                 match = (m_tag_find(m,
1850                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1851 #endif
1852                                 /* otherwise no match */
1853                                 break;
1854
1855 #ifdef INET6
1856                         case O_IP6_SRC:
1857                                 match = is_ipv6 &&
1858                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1859                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1860                                 break;
1861
1862                         case O_IP6_DST:
1863                                 match = is_ipv6 &&
1864                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1865                                     &((ipfw_insn_ip6 *)cmd)->addr6);
1866                                 break;
1867                         case O_IP6_SRC_MASK:
1868                         case O_IP6_DST_MASK:
1869                                 if (is_ipv6) {
1870                                         int i = cmdlen - 1;
1871                                         struct in6_addr p;
1872                                         struct in6_addr *d =
1873                                             &((ipfw_insn_ip6 *)cmd)->addr6;
1874
1875                                         for (; !match && i > 0; d += 2,
1876                                             i -= F_INSN_SIZE(struct in6_addr)
1877                                             * 2) {
1878                                                 p = (cmd->opcode ==
1879                                                     O_IP6_SRC_MASK) ?
1880                                                     args->f_id.src_ip6:
1881                                                     args->f_id.dst_ip6;
1882                                                 APPLY_MASK(&p, &d[1]);
1883                                                 match =
1884                                                     IN6_ARE_ADDR_EQUAL(&d[0],
1885                                                     &p);
1886                                         }
1887                                 }
1888                                 break;
1889
1890                         case O_FLOW6ID:
1891                                 match = is_ipv6 &&
1892                                     flow6id_match(args->f_id.flow_id6,
1893                                     (ipfw_insn_u32 *) cmd);
1894                                 break;
1895
1896                         case O_EXT_HDR:
1897                                 match = is_ipv6 &&
1898                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
1899                                 break;
1900
1901                         case O_IP6:
1902                                 match = is_ipv6;
1903                                 break;
1904 #endif
1905
1906                         case O_IP4:
1907                                 match = is_ipv4;
1908                                 break;
1909
1910                         case O_TAG: {
1911                                 struct m_tag *mtag;
1912                                 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1913
1914                                 /* Packet is already tagged with this tag? */
1915                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1916
1917                                 /* We have `untag' action when F_NOT flag is
1918                                  * present. And we must remove this mtag from
1919                                  * mbuf and reset `match' to zero (`match' will
1920                                  * be inversed later).
1921                                  * Otherwise we should allocate new mtag and
1922                                  * push it into mbuf.
1923                                  */
1924                                 if (cmd->len & F_NOT) { /* `untag' action */
1925                                         if (mtag != NULL)
1926                                                 m_tag_delete(m, mtag);
1927                                         match = 0;
1928                                 } else {
1929                                         if (mtag == NULL) {
1930                                                 mtag = m_tag_alloc( MTAG_IPFW,
1931                                                     tag, 0, M_NOWAIT);
1932                                                 if (mtag != NULL)
1933                                                         m_tag_prepend(m, mtag);
1934                                         }
1935                                         match = 1;
1936                                 }
1937                                 break;
1938                         }
1939
1940                         case O_FIB: /* try match the specified fib */
1941                                 if (args->f_id.fib == cmd->arg1)
1942                                         match = 1;
1943                                 break;
1944
1945                         case O_SOCKARG: {
1946                                 struct inpcb *inp = args->inp;
1947                                 struct inpcbinfo *pi;
1948
1949                                 if (is_ipv6) /* XXX can we remove this ? */
1950                                         break;
1951
1952                                 if (proto == IPPROTO_TCP)
1953                                         pi = &V_tcbinfo;
1954                                 else if (proto == IPPROTO_UDP)
1955                                         pi = &V_udbinfo;
1956                                 else
1957                                         break;
1958
1959                                 /*
1960                                  * XXXRW: so_user_cookie should almost
1961                                  * certainly be inp_user_cookie?
1962                                  */
1963
1964                                 /* For incomming packet, lookup up the
1965                                 inpcb using the src/dest ip/port tuple */
1966                                 if (inp == NULL) {
1967                                         inp = in_pcblookup(pi,
1968                                                 src_ip, htons(src_port),
1969                                                 dst_ip, htons(dst_port),
1970                                                 INPLOOKUP_RLOCKPCB, NULL);
1971                                         if (inp != NULL) {
1972                                                 tablearg =
1973                                                     inp->inp_socket->so_user_cookie;
1974                                                 if (tablearg)
1975                                                         match = 1;
1976                                                 INP_RUNLOCK(inp);
1977                                         }
1978                                 } else {
1979                                         if (inp->inp_socket) {
1980                                                 tablearg =
1981                                                     inp->inp_socket->so_user_cookie;
1982                                                 if (tablearg)
1983                                                         match = 1;
1984                                         }
1985                                 }
1986                                 break;
1987                         }
1988
1989                         case O_TAGGED: {
1990                                 struct m_tag *mtag;
1991                                 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1992
1993                                 if (cmdlen == 1) {
1994                                         match = m_tag_locate(m, MTAG_IPFW,
1995                                             tag, NULL) != NULL;
1996                                         break;
1997                                 }
1998
1999                                 /* we have ranges */
2000                                 for (mtag = m_tag_first(m);
2001                                     mtag != NULL && !match;
2002                                     mtag = m_tag_next(m, mtag)) {
2003                                         uint16_t *p;
2004                                         int i;
2005
2006                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2007                                                 continue;
2008
2009                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2010                                         i = cmdlen - 1;
2011                                         for(; !match && i > 0; i--, p += 2)
2012                                                 match =
2013                                                     mtag->m_tag_id >= p[0] &&
2014                                                     mtag->m_tag_id <= p[1];
2015                                 }
2016                                 break;
2017                         }
2018
2019                         /*
2020                          * The second set of opcodes represents 'actions',
2021                          * i.e. the terminal part of a rule once the packet
2022                          * matches all previous patterns.
2023                          * Typically there is only one action for each rule,
2024                          * and the opcode is stored at the end of the rule
2025                          * (but there are exceptions -- see below).
2026                          *
2027                          * In general, here we set retval and terminate the
2028                          * outer loop (would be a 'break 3' in some language,
2029                          * but we need to set l=0, done=1)
2030                          *
2031                          * Exceptions:
2032                          * O_COUNT and O_SKIPTO actions:
2033                          *   instead of terminating, we jump to the next rule
2034                          *   (setting l=0), or to the SKIPTO target (setting
2035                          *   f/f_len, cmd and l as needed), respectively.
2036                          *
2037                          * O_TAG, O_LOG and O_ALTQ action parameters:
2038                          *   perform some action and set match = 1;
2039                          *
2040                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2041                          *   not real 'actions', and are stored right
2042                          *   before the 'action' part of the rule.
2043                          *   These opcodes try to install an entry in the
2044                          *   state tables; if successful, we continue with
2045                          *   the next opcode (match=1; break;), otherwise
2046                          *   the packet must be dropped (set retval,
2047                          *   break loops with l=0, done=1)
2048                          *
2049                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2050                          *   cause a lookup of the state table, and a jump
2051                          *   to the 'action' part of the parent rule
2052                          *   if an entry is found, or
2053                          *   (CHECK_STATE only) a jump to the next rule if
2054                          *   the entry is not found.
2055                          *   The result of the lookup is cached so that
2056                          *   further instances of these opcodes become NOPs.
2057                          *   The jump to the next rule is done by setting
2058                          *   l=0, cmdlen=0.
2059                          */
2060                         case O_LIMIT:
2061                         case O_KEEP_STATE:
2062                                 if (ipfw_install_state(f,
2063                                     (ipfw_insn_limit *)cmd, args, tablearg)) {
2064                                         /* error or limit violation */
2065                                         retval = IP_FW_DENY;
2066                                         l = 0;  /* exit inner loop */
2067                                         done = 1; /* exit outer loop */
2068                                 }
2069                                 match = 1;
2070                                 break;
2071
2072                         case O_PROBE_STATE:
2073                         case O_CHECK_STATE:
2074                                 /*
2075                                  * dynamic rules are checked at the first
2076                                  * keep-state or check-state occurrence,
2077                                  * with the result being stored in dyn_dir.
2078                                  * The compiler introduces a PROBE_STATE
2079                                  * instruction for us when we have a
2080                                  * KEEP_STATE (because PROBE_STATE needs
2081                                  * to be run first).
2082                                  */
2083                                 if (dyn_dir == MATCH_UNKNOWN &&
2084                                     (q = ipfw_lookup_dyn_rule(&args->f_id,
2085                                      &dyn_dir, proto == IPPROTO_TCP ?
2086                                         TCP(ulp) : NULL))
2087                                         != NULL) {
2088                                         /*
2089                                          * Found dynamic entry, update stats
2090                                          * and jump to the 'action' part of
2091                                          * the parent rule by setting
2092                                          * f, cmd, l and clearing cmdlen.
2093                                          */
2094                                         IPFW_INC_DYN_COUNTER(q, pktlen);
2095                                         /* XXX we would like to have f_pos
2096                                          * readily accessible in the dynamic
2097                                          * rule, instead of having to
2098                                          * lookup q->rule.
2099                                          */
2100                                         f = q->rule;
2101                                         f_pos = ipfw_find_rule(chain,
2102                                                 f->rulenum, f->id);
2103                                         cmd = ACTION_PTR(f);
2104                                         l = f->cmd_len - f->act_ofs;
2105                                         ipfw_dyn_unlock(q);
2106                                         cmdlen = 0;
2107                                         match = 1;
2108                                         break;
2109                                 }
2110                                 /*
2111                                  * Dynamic entry not found. If CHECK_STATE,
2112                                  * skip to next rule, if PROBE_STATE just
2113                                  * ignore and continue with next opcode.
2114                                  */
2115                                 if (cmd->opcode == O_CHECK_STATE)
2116                                         l = 0;  /* exit inner loop */
2117                                 match = 1;
2118                                 break;
2119
2120                         case O_ACCEPT:
2121                                 retval = 0;     /* accept */
2122                                 l = 0;          /* exit inner loop */
2123                                 done = 1;       /* exit outer loop */
2124                                 break;
2125
2126                         case O_PIPE:
2127                         case O_QUEUE:
2128                                 set_match(args, f_pos, chain);
2129                                 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2130                                 if (cmd->opcode == O_PIPE)
2131                                         args->rule.info |= IPFW_IS_PIPE;
2132                                 if (V_fw_one_pass)
2133                                         args->rule.info |= IPFW_ONEPASS;
2134                                 retval = IP_FW_DUMMYNET;
2135                                 l = 0;          /* exit inner loop */
2136                                 done = 1;       /* exit outer loop */
2137                                 break;
2138
2139                         case O_DIVERT:
2140                         case O_TEE:
2141                                 if (args->eh) /* not on layer 2 */
2142                                     break;
2143                                 /* otherwise this is terminal */
2144                                 l = 0;          /* exit inner loop */
2145                                 done = 1;       /* exit outer loop */
2146                                 retval = (cmd->opcode == O_DIVERT) ?
2147                                         IP_FW_DIVERT : IP_FW_TEE;
2148                                 set_match(args, f_pos, chain);
2149                                 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2150                                 break;
2151
2152                         case O_COUNT:
2153                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2154                                 l = 0;          /* exit inner loop */
2155                                 break;
2156
2157                         case O_SKIPTO:
2158                             IPFW_INC_RULE_COUNTER(f, pktlen);
2159                             f_pos = jump_fast(chain, f, cmd->arg1, tablearg, 0);
2160                             /*
2161                              * Skip disabled rules, and re-enter
2162                              * the inner loop with the correct
2163                              * f_pos, f, l and cmd.
2164                              * Also clear cmdlen and skip_or
2165                              */
2166                             for (; f_pos < chain->n_rules - 1 &&
2167                                     (V_set_disable &
2168                                      (1 << chain->map[f_pos]->set));
2169                                     f_pos++)
2170                                 ;
2171                             /* Re-enter the inner loop at the skipto rule. */
2172                             f = chain->map[f_pos];
2173                             l = f->cmd_len;
2174                             cmd = f->cmd;
2175                             match = 1;
2176                             cmdlen = 0;
2177                             skip_or = 0;
2178                             continue;
2179                             break;      /* not reached */
2180
2181                         case O_CALLRETURN: {
2182                                 /*
2183                                  * Implementation of `subroutine' call/return,
2184                                  * in the stack carried in an mbuf tag. This
2185                                  * is different from `skipto' in that any call
2186                                  * address is possible (`skipto' must prevent
2187                                  * backward jumps to avoid endless loops).
2188                                  * We have `return' action when F_NOT flag is
2189                                  * present. The `m_tag_id' field is used as
2190                                  * stack pointer.
2191                                  */
2192                                 struct m_tag *mtag;
2193                                 uint16_t jmpto, *stack;
2194
2195 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2196 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2197                                 /*
2198                                  * Hand-rolled version of m_tag_locate() with
2199                                  * wildcard `type'.
2200                                  * If not already tagged, allocate new tag.
2201                                  */
2202                                 mtag = m_tag_first(m);
2203                                 while (mtag != NULL) {
2204                                         if (mtag->m_tag_cookie ==
2205                                             MTAG_IPFW_CALL)
2206                                                 break;
2207                                         mtag = m_tag_next(m, mtag);
2208                                 }
2209                                 if (mtag == NULL && IS_CALL) {
2210                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2211                                             IPFW_CALLSTACK_SIZE *
2212                                             sizeof(uint16_t), M_NOWAIT);
2213                                         if (mtag != NULL)
2214                                                 m_tag_prepend(m, mtag);
2215                                 }
2216
2217                                 /*
2218                                  * On error both `call' and `return' just
2219                                  * continue with next rule.
2220                                  */
2221                                 if (IS_RETURN && (mtag == NULL ||
2222                                     mtag->m_tag_id == 0)) {
2223                                         l = 0;          /* exit inner loop */
2224                                         break;
2225                                 }
2226                                 if (IS_CALL && (mtag == NULL ||
2227                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2228                                         printf("ipfw: call stack error, "
2229                                             "go to next rule\n");
2230                                         l = 0;          /* exit inner loop */
2231                                         break;
2232                                 }
2233
2234                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2235                                 stack = (uint16_t *)(mtag + 1);
2236
2237                                 /*
2238                                  * The `call' action may use cached f_pos
2239                                  * (in f->next_rule), whose version is written
2240                                  * in f->next_rule.
2241                                  * The `return' action, however, doesn't have
2242                                  * fixed jump address in cmd->arg1 and can't use
2243                                  * cache.
2244                                  */
2245                                 if (IS_CALL) {
2246                                         stack[mtag->m_tag_id] = f->rulenum;
2247                                         mtag->m_tag_id++;
2248                                         f_pos = jump_fast(chain, f, cmd->arg1,
2249                                             tablearg, 1);
2250                                 } else {        /* `return' action */
2251                                         mtag->m_tag_id--;
2252                                         jmpto = stack[mtag->m_tag_id] + 1;
2253                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2254                                 }
2255
2256                                 /*
2257                                  * Skip disabled rules, and re-enter
2258                                  * the inner loop with the correct
2259                                  * f_pos, f, l and cmd.
2260                                  * Also clear cmdlen and skip_or
2261                                  */
2262                                 for (; f_pos < chain->n_rules - 1 &&
2263                                     (V_set_disable &
2264                                     (1 << chain->map[f_pos]->set)); f_pos++)
2265                                         ;
2266                                 /* Re-enter the inner loop at the dest rule. */
2267                                 f = chain->map[f_pos];
2268                                 l = f->cmd_len;
2269                                 cmd = f->cmd;
2270                                 cmdlen = 0;
2271                                 skip_or = 0;
2272                                 continue;
2273                                 break;  /* NOTREACHED */
2274                         }
2275 #undef IS_CALL
2276 #undef IS_RETURN
2277
2278                         case O_REJECT:
2279                                 /*
2280                                  * Drop the packet and send a reject notice
2281                                  * if the packet is not ICMP (or is an ICMP
2282                                  * query), and it is not multicast/broadcast.
2283                                  */
2284                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2285                                     (proto != IPPROTO_ICMP ||
2286                                      is_icmp_query(ICMP(ulp))) &&
2287                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2288                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2289                                         send_reject(args, cmd->arg1, iplen, ip);
2290                                         m = args->m;
2291                                 }
2292                                 /* FALLTHROUGH */
2293 #ifdef INET6
2294                         case O_UNREACH6:
2295                                 if (hlen > 0 && is_ipv6 &&
2296                                     ((offset & IP6F_OFF_MASK) == 0) &&
2297                                     (proto != IPPROTO_ICMPV6 ||
2298                                      (is_icmp6_query(icmp6_type) == 1)) &&
2299                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2300                                     !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2301                                         send_reject6(
2302                                             args, cmd->arg1, hlen,
2303                                             (struct ip6_hdr *)ip);
2304                                         m = args->m;
2305                                 }
2306                                 /* FALLTHROUGH */
2307 #endif
2308                         case O_DENY:
2309                                 retval = IP_FW_DENY;
2310                                 l = 0;          /* exit inner loop */
2311                                 done = 1;       /* exit outer loop */
2312                                 break;
2313
2314                         case O_FORWARD_IP:
2315                                 if (args->eh)   /* not valid on layer2 pkts */
2316                                         break;
2317                                 if (q == NULL || q->rule != f ||
2318                                     dyn_dir == MATCH_FORWARD) {
2319                                     struct sockaddr_in *sa;
2320                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2321                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2322                                         bcopy(sa, &args->hopstore,
2323                                                         sizeof(*sa));
2324                                         args->hopstore.sin_addr.s_addr =
2325                                                     htonl(tablearg);
2326                                         args->next_hop = &args->hopstore;
2327                                     } else {
2328                                         args->next_hop = sa;
2329                                     }
2330                                 }
2331                                 retval = IP_FW_PASS;
2332                                 l = 0;          /* exit inner loop */
2333                                 done = 1;       /* exit outer loop */
2334                                 break;
2335
2336 #ifdef INET6
2337                         case O_FORWARD_IP6:
2338                                 if (args->eh)   /* not valid on layer2 pkts */
2339                                         break;
2340                                 if (q == NULL || q->rule != f ||
2341                                     dyn_dir == MATCH_FORWARD) {
2342                                         struct sockaddr_in6 *sin6;
2343
2344                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2345                                         args->next_hop6 = sin6;
2346                                 }
2347                                 retval = IP_FW_PASS;
2348                                 l = 0;          /* exit inner loop */
2349                                 done = 1;       /* exit outer loop */
2350                                 break;
2351 #endif
2352
2353                         case O_NETGRAPH:
2354                         case O_NGTEE:
2355                                 set_match(args, f_pos, chain);
2356                                 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2357                                 if (V_fw_one_pass)
2358                                         args->rule.info |= IPFW_ONEPASS;
2359                                 retval = (cmd->opcode == O_NETGRAPH) ?
2360                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2361                                 l = 0;          /* exit inner loop */
2362                                 done = 1;       /* exit outer loop */
2363                                 break;
2364
2365                         case O_SETFIB: {
2366                                 uint32_t fib;
2367
2368                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2369                                 fib = IP_FW_ARG_TABLEARG(cmd->arg1);
2370                                 if (fib >= rt_numfibs)
2371                                         fib = 0;
2372                                 M_SETFIB(m, fib);
2373                                 args->f_id.fib = fib;
2374                                 l = 0;          /* exit inner loop */
2375                                 break;
2376                         }
2377
2378                         case O_SETDSCP: {
2379                                 uint16_t code;
2380
2381                                 code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
2382                                 l = 0;          /* exit inner loop */
2383                                 if (is_ipv4) {
2384                                         uint16_t a;
2385
2386                                         a = ip->ip_tos;
2387                                         ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2388                                         a += ntohs(ip->ip_sum) - ip->ip_tos;
2389                                         ip->ip_sum = htons(a);
2390                                 } else if (is_ipv6) {
2391                                         uint8_t *v;
2392
2393                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2394                                         *v = (*v & 0xF0) | (code >> 2);
2395                                         v++;
2396                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2397                                 } else
2398                                         break;
2399
2400                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2401                                 break;
2402                         }
2403
2404                         case O_NAT:
2405                                 l = 0;          /* exit inner loop */
2406                                 done = 1;       /* exit outer loop */
2407                                 if (!IPFW_NAT_LOADED) {
2408                                     retval = IP_FW_DENY;
2409                                     break;
2410                                 }
2411
2412                                 struct cfg_nat *t;
2413                                 int nat_id;
2414
2415                                 set_match(args, f_pos, chain);
2416                                 /* Check if this is 'global' nat rule */
2417                                 if (cmd->arg1 == 0) {
2418                                         retval = ipfw_nat_ptr(args, NULL, m);
2419                                         break;
2420                                 }
2421                                 t = ((ipfw_insn_nat *)cmd)->nat;
2422                                 if (t == NULL) {
2423                                         nat_id = IP_FW_ARG_TABLEARG(cmd->arg1);
2424                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2425
2426                                         if (t == NULL) {
2427                                             retval = IP_FW_DENY;
2428                                             break;
2429                                         }
2430                                         if (cmd->arg1 != IP_FW_TABLEARG)
2431                                             ((ipfw_insn_nat *)cmd)->nat = t;
2432                                 }
2433                                 retval = ipfw_nat_ptr(args, t, m);
2434                                 break;
2435
2436                         case O_REASS: {
2437                                 int ip_off;
2438
2439                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2440                                 l = 0;  /* in any case exit inner loop */
2441                                 ip_off = ntohs(ip->ip_off);
2442
2443                                 /* if not fragmented, go to next rule */
2444                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2445                                     break;
2446                                 /*
2447                                  * ip_reass() expects len & off in host
2448                                  * byte order.
2449                                  */
2450                                 SET_HOST_IPLEN(ip);
2451
2452                                 args->m = m = ip_reass(m);
2453
2454                                 /*
2455                                  * do IP header checksum fixup.
2456                                  */
2457                                 if (m == NULL) { /* fragment got swallowed */
2458                                     retval = IP_FW_DENY;
2459                                 } else { /* good, packet complete */
2460                                     int hlen;
2461
2462                                     ip = mtod(m, struct ip *);
2463                                     hlen = ip->ip_hl << 2;
2464                                     SET_NET_IPLEN(ip);
2465                                     ip->ip_sum = 0;
2466                                     if (hlen == sizeof(struct ip))
2467                                         ip->ip_sum = in_cksum_hdr(ip);
2468                                     else
2469                                         ip->ip_sum = in_cksum(m, hlen);
2470                                     retval = IP_FW_REASS;
2471                                     set_match(args, f_pos, chain);
2472                                 }
2473                                 done = 1;       /* exit outer loop */
2474                                 break;
2475                         }
2476
2477                         default:
2478                                 panic("-- unknown opcode %d\n", cmd->opcode);
2479                         } /* end of switch() on opcodes */
2480                         /*
2481                          * if we get here with l=0, then match is irrelevant.
2482                          */
2483
2484                         if (cmd->len & F_NOT)
2485                                 match = !match;
2486
2487                         if (match) {
2488                                 if (cmd->len & F_OR)
2489                                         skip_or = 1;
2490                         } else {
2491                                 if (!(cmd->len & F_OR)) /* not an OR block, */
2492                                         break;          /* try next rule    */
2493                         }
2494
2495                 }       /* end of inner loop, scan opcodes */
2496 #undef PULLUP_LEN
2497
2498                 if (done)
2499                         break;
2500
2501 /* next_rule:; */       /* try next rule                */
2502
2503         }               /* end of outer for, scan rules */
2504
2505         if (done) {
2506                 struct ip_fw *rule = chain->map[f_pos];
2507                 /* Update statistics */
2508                 IPFW_INC_RULE_COUNTER(rule, pktlen);
2509         } else {
2510                 retval = IP_FW_DENY;
2511                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2512         }
2513         IPFW_RUNLOCK(chain);
2514 #ifdef __FreeBSD__
2515         if (ucred_cache != NULL)
2516                 crfree(ucred_cache);
2517 #endif
2518         return (retval);
2519
2520 pullup_failed:
2521         if (V_fw_verbose)
2522                 printf("ipfw: pullup failed\n");
2523         return (IP_FW_DENY);
2524 }
2525
2526 /*
2527  * Set maximum number of tables that can be used in given VNET ipfw instance.
2528  */
2529 #ifdef SYSCTL_NODE
2530 static int
2531 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2532 {
2533         int error;
2534         unsigned int ntables;
2535
2536         ntables = V_fw_tables_max;
2537
2538         error = sysctl_handle_int(oidp, &ntables, 0, req);
2539         /* Read operation or some error */
2540         if ((error != 0) || (req->newptr == NULL))
2541                 return (error);
2542
2543         return (ipfw_resize_tables(&V_layer3_chain, ntables));
2544 }
2545 #endif
2546 /*
2547  * Module and VNET glue
2548  */
2549
2550 /*
2551  * Stuff that must be initialised only on boot or module load
2552  */
2553 static int
2554 ipfw_init(void)
2555 {
2556         int error = 0;
2557
2558         /*
2559          * Only print out this stuff the first time around,
2560          * when called from the sysinit code.
2561          */
2562         printf("ipfw2 "
2563 #ifdef INET6
2564                 "(+ipv6) "
2565 #endif
2566                 "initialized, divert %s, nat %s, "
2567                 "default to %s, logging ",
2568 #ifdef IPDIVERT
2569                 "enabled",
2570 #else
2571                 "loadable",
2572 #endif
2573 #ifdef IPFIREWALL_NAT
2574                 "enabled",
2575 #else
2576                 "loadable",
2577 #endif
2578                 default_to_accept ? "accept" : "deny");
2579
2580         /*
2581          * Note: V_xxx variables can be accessed here but the vnet specific
2582          * initializer may not have been called yet for the VIMAGE case.
2583          * Tuneables will have been processed. We will print out values for
2584          * the default vnet.
2585          * XXX This should all be rationalized AFTER 8.0
2586          */
2587         if (V_fw_verbose == 0)
2588                 printf("disabled\n");
2589         else if (V_verbose_limit == 0)
2590                 printf("unlimited\n");
2591         else
2592                 printf("limited to %d packets/entry by default\n",
2593                     V_verbose_limit);
2594
2595         /* Check user-supplied table count for validness */
2596         if (default_fw_tables > IPFW_TABLES_MAX)
2597           default_fw_tables = IPFW_TABLES_MAX;
2598
2599         ipfw_log_bpf(1); /* init */
2600         return (error);
2601 }
2602
2603 /*
2604  * Called for the removal of the last instance only on module unload.
2605  */
2606 static void
2607 ipfw_destroy(void)
2608 {
2609
2610         ipfw_log_bpf(0); /* uninit */
2611         printf("IP firewall unloaded\n");
2612 }
2613
2614 /*
2615  * Stuff that must be initialized for every instance
2616  * (including the first of course).
2617  */
2618 static int
2619 vnet_ipfw_init(const void *unused)
2620 {
2621         int error;
2622         struct ip_fw *rule = NULL;
2623         struct ip_fw_chain *chain;
2624
2625         chain = &V_layer3_chain;
2626
2627         /* First set up some values that are compile time options */
2628         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
2629         V_fw_deny_unknown_exthdrs = 1;
2630 #ifdef IPFIREWALL_VERBOSE
2631         V_fw_verbose = 1;
2632 #endif
2633 #ifdef IPFIREWALL_VERBOSE_LIMIT
2634         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2635 #endif
2636 #ifdef IPFIREWALL_NAT
2637         LIST_INIT(&chain->nat);
2638 #endif
2639
2640         /* insert the default rule and create the initial map */
2641         chain->n_rules = 1;
2642         chain->static_len = sizeof(struct ip_fw);
2643         chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2644         if (chain->map)
2645                 rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2646
2647         /* Set initial number of tables */
2648         V_fw_tables_max = default_fw_tables;
2649         error = ipfw_init_tables(chain);
2650         if (error) {
2651                 printf("ipfw2: setting up tables failed\n");
2652                 free(chain->map, M_IPFW);
2653                 free(rule, M_IPFW);
2654                 return (ENOSPC);
2655         }
2656
2657         /* fill and insert the default rule */
2658         rule->act_ofs = 0;
2659         rule->rulenum = IPFW_DEFAULT_RULE;
2660         rule->cmd_len = 1;
2661         rule->set = RESVD_SET;
2662         rule->cmd[0].len = 1;
2663         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2664         chain->default_rule = chain->map[0] = rule;
2665         chain->id = rule->id = 1;
2666
2667         IPFW_LOCK_INIT(chain);
2668         ipfw_dyn_init(chain);
2669
2670         /* First set up some values that are compile time options */
2671         V_ipfw_vnet_ready = 1;          /* Open for business */
2672
2673         /*
2674          * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2675          * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2676          * we still keep the module alive because the sockopt and
2677          * layer2 paths are still useful.
2678          * ipfw[6]_hook return 0 on success, ENOENT on failure,
2679          * so we can ignore the exact return value and just set a flag.
2680          *
2681          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2682          * changes in the underlying (per-vnet) variables trigger
2683          * immediate hook()/unhook() calls.
2684          * In layer2 we have the same behaviour, except that V_ether_ipfw
2685          * is checked on each packet because there are no pfil hooks.
2686          */
2687         V_ip_fw_ctl_ptr = ipfw_ctl;
2688         V_ip_fw_chk_ptr = ipfw_chk;
2689         error = ipfw_attach_hooks(1);
2690         return (error);
2691 }
2692
2693 /*
2694  * Called for the removal of each instance.
2695  */
2696 static int
2697 vnet_ipfw_uninit(const void *unused)
2698 {
2699         struct ip_fw *reap, *rule;
2700         struct ip_fw_chain *chain = &V_layer3_chain;
2701         int i;
2702
2703         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2704         /*
2705          * disconnect from ipv4, ipv6, layer2 and sockopt.
2706          * Then grab, release and grab again the WLOCK so we make
2707          * sure the update is propagated and nobody will be in.
2708          */
2709         (void)ipfw_attach_hooks(0 /* detach */);
2710         V_ip_fw_chk_ptr = NULL;
2711         V_ip_fw_ctl_ptr = NULL;
2712         IPFW_UH_WLOCK(chain);
2713         IPFW_UH_WUNLOCK(chain);
2714         IPFW_UH_WLOCK(chain);
2715
2716         IPFW_WLOCK(chain);
2717         ipfw_dyn_uninit(0);     /* run the callout_drain */
2718         IPFW_WUNLOCK(chain);
2719
2720         ipfw_destroy_tables(chain);
2721         reap = NULL;
2722         IPFW_WLOCK(chain);
2723         for (i = 0; i < chain->n_rules; i++) {
2724                 rule = chain->map[i];
2725                 rule->x_next = reap;
2726                 reap = rule;
2727         }
2728         if (chain->map)
2729                 free(chain->map, M_IPFW);
2730         IPFW_WUNLOCK(chain);
2731         IPFW_UH_WUNLOCK(chain);
2732         if (reap != NULL)
2733                 ipfw_reap_rules(reap);
2734         IPFW_LOCK_DESTROY(chain);
2735         ipfw_dyn_uninit(1);     /* free the remaining parts */
2736         return 0;
2737 }
2738
2739 /*
2740  * Module event handler.
2741  * In general we have the choice of handling most of these events by the
2742  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2743  * use the SYSINIT handlers as they are more capable of expressing the
2744  * flow of control during module and vnet operations, so this is just
2745  * a skeleton. Note there is no SYSINIT equivalent of the module
2746  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2747  */
2748 static int
2749 ipfw_modevent(module_t mod, int type, void *unused)
2750 {
2751         int err = 0;
2752
2753         switch (type) {
2754         case MOD_LOAD:
2755                 /* Called once at module load or
2756                  * system boot if compiled in. */
2757                 break;
2758         case MOD_QUIESCE:
2759                 /* Called before unload. May veto unloading. */
2760                 break;
2761         case MOD_UNLOAD:
2762                 /* Called during unload. */
2763                 break;
2764         case MOD_SHUTDOWN:
2765                 /* Called during system shutdown. */
2766                 break;
2767         default:
2768                 err = EOPNOTSUPP;
2769                 break;
2770         }
2771         return err;
2772 }
2773
2774 static moduledata_t ipfwmod = {
2775         "ipfw",
2776         ipfw_modevent,
2777         0
2778 };
2779
2780 /* Define startup order. */
2781 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_IFATTACHDOMAIN
2782 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
2783 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2784 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2785
2786 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2787 MODULE_VERSION(ipfw, 2);
2788 /* should declare some dependencies here */
2789
2790 /*
2791  * Starting up. Done in order after ipfwmod() has been called.
2792  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2793  */
2794 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2795             ipfw_init, NULL);
2796 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2797             vnet_ipfw_init, NULL);
2798
2799 /*
2800  * Closing up shop. These are done in REVERSE ORDER, but still
2801  * after ipfwmod() has been called. Not called on reboot.
2802  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2803  * or when the module is unloaded.
2804  */
2805 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2806             ipfw_destroy, NULL);
2807 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2808             vnet_ipfw_uninit, NULL);
2809 /* end of file */