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