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