sys/netpfil/ipfw/ip_fw2.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
   3  *
   4  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  *
  15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  25  * SUCH DAMAGE.
  26  */
  27
  28 #include <sys/cdefs.h>
  29 __FBSDID("$FreeBSD$");
  30
  31 /*
  32  * The FreeBSD IP packet firewall, main file
  33  */
  34
  35 #include "opt_ipfw.h"
  36 #include "opt_ipdivert.h"
  37 #include "opt_inet.h"
  38 #ifndef INET
  39 #error "IPFIREWALL requires INET"
  40 #endif /* INET */
  41 #include "opt_inet6.h"
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/condvar.h>
  46 #include <sys/counter.h>
  47 #include <sys/eventhandler.h>
  48 #include <sys/malloc.h>
  49 #include <sys/mbuf.h>
  50 #include <sys/kernel.h>
  51 #include <sys/lock.h>
  52 #include <sys/jail.h>
  53 #include <sys/module.h>
  54 #include <sys/priv.h>
  55 #include <sys/proc.h>
  56 #include <sys/rwlock.h>
  57 #include <sys/rmlock.h>
  58 #include <sys/socket.h>
  59 #include <sys/socketvar.h>
  60 #include <sys/sysctl.h>
  61 #include <sys/syslog.h>
  62 #include <sys/ucred.h>
  63 #include <net/ethernet.h> /* for ETHERTYPE_IP */
  64 #include <net/if.h>
  65 #include <net/if_var.h>
  66 #include <net/route.h>
  67 #include <net/pfil.h>
  68 #include <net/vnet.h>
  69
  70 #include <netpfil/pf/pf_mtag.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/in_pcb.h>
  75 #include <netinet/ip.h>
  76 #include <netinet/ip_var.h>
  77 #include <netinet/ip_icmp.h>
  78 #include <netinet/ip_fw.h>
  79 #include <netinet/ip_carp.h>
  80 #include <netinet/pim.h>
  81 #include <netinet/tcp_var.h>
  82 #include <netinet/udp.h>
  83 #include <netinet/udp_var.h>
  84 #include <netinet/sctp.h>
  85 #include <netinet/sctp_crc32.h>
  86 #include <netinet/sctp_header.h>
  87
  88 #include <netinet/ip6.h>
  89 #include <netinet/icmp6.h>
  90 #include <netinet/in_fib.h>
  91 #ifdef INET6
  92 #include <netinet6/in6_fib.h>
  93 #include <netinet6/in6_pcb.h>
  94 #include <netinet6/scope6_var.h>
  95 #include <netinet6/ip6_var.h>
  96 #endif
  97
  98 #include <net/if_gre.h> /* for struct grehdr */
  99
 100 #include <netpfil/ipfw/ip_fw_private.h>
 101
 102 #include <machine/in_cksum.h>   /* XXX for in_cksum */
 103
 104 #ifdef MAC
 105 #include <security/mac/mac_framework.h>
 106 #endif
 107
 108 /*
 109  * static variables followed by global ones.
 110  * All ipfw global variables are here.
 111  */
 112
 113 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs);
 114 #define V_fw_deny_unknown_exthdrs       VNET(fw_deny_unknown_exthdrs)
 115
 116 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1;
 117 #define V_fw_permit_single_frag6        VNET(fw_permit_single_frag6)
 118
 119 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
 120 static int default_to_accept = 1;
 121 #else
 122 static int default_to_accept;
 123 #endif
 124
 125 VNET_DEFINE(int, autoinc_step);
 126 VNET_DEFINE(int, fw_one_pass) = 1;
 127
 128 VNET_DEFINE(unsigned int, fw_tables_max);
 129 VNET_DEFINE(unsigned int, fw_tables_sets) = 0;  /* Don't use set-aware tables */
 130 /* Use 128 tables by default */
 131 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
 132
 133 #ifndef LINEAR_SKIPTO
 134 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 135     int tablearg, int jump_backwards);
 136 #define JUMP(ch, f, num, targ, back)    jump_fast(ch, f, num, targ, back)
 137 #else
 138 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
 139     int tablearg, int jump_backwards);
 140 #define JUMP(ch, f, num, targ, back)    jump_linear(ch, f, num, targ, back)
 141 #endif
 142
 143 /*
 144  * Each rule belongs to one of 32 different sets (0..31).
 145  * The variable set_disable contains one bit per set.
 146  * If the bit is set, all rules in the corresponding set
 147  * are disabled. Set RESVD_SET(31) is reserved for the default rule
 148  * and rules that are not deleted by the flush command,
 149  * and CANNOT be disabled.
 150  * Rules in set RESVD_SET can only be deleted individually.
 151  */
 152 VNET_DEFINE(u_int32_t, set_disable);
 153 #define V_set_disable                   VNET(set_disable)
 154
 155 VNET_DEFINE(int, fw_verbose);
 156 /* counter for ipfw_log(NULL...) */
 157 VNET_DEFINE(u_int64_t, norule_counter);
 158 VNET_DEFINE(int, verbose_limit);
 159
 160 /* layer3_chain contains the list of rules for layer 3 */
 161 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
 162
 163 /* ipfw_vnet_ready controls when we are open for business */
 164 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
 165
 166 VNET_DEFINE(int, ipfw_nat_ready) = 0;
 167
 168 ipfw_nat_t *ipfw_nat_ptr = NULL;
 169 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
 170 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
 171 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
 172 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
 173 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
 174
 175 #ifdef SYSCTL_NODE
 176 uint32_t dummy_def = IPFW_DEFAULT_RULE;
 177 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
 178 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
 179
 180 SYSBEGIN(f3)
 181
 182 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
 184     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
 185     "Only do a single pass through ipfw when using dummynet(4)");
 186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
 187     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
 188     "Rule number auto-increment step");
 189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
 190     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
 191     "Log matches to ipfw rules");
 192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
 193     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
 194     "Set upper limit of matches of ipfw rules logged");
 195 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
 196     &dummy_def, 0,
 197     "The default/max possible rule number.");
 198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
 199     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
 200     "Maximum number of concurrently used tables");
 201 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
 202     CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
 203     0, 0, sysctl_ipfw_tables_sets, "IU",
 204     "Use per-set namespace for tables");
 205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
 206     &default_to_accept, 0,
 207     "Make the default rule accept all packets.");
 208 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
 209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
 210     CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
 211     "Number of static rules");
 212
 213 #ifdef INET6
 214 SYSCTL_DECL(_net_inet6_ip6);
 215 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
 216 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
 217     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 218     &VNET_NAME(fw_deny_unknown_exthdrs), 0,
 219     "Deny packets with unknown IPv6 Extension Headers");
 220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
 221     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
 222     &VNET_NAME(fw_permit_single_frag6), 0,
 223     "Permit single packet IPv6 fragments");
 224 #endif /* INET6 */
 225
 226 SYSEND
 227
 228 #endif /* SYSCTL_NODE */
 229
 230
 231 /*
 232  * Some macros used in the various matching options.
 233  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
 234  * Other macros just cast void * into the appropriate type
 235  */
 236 #define L3HDR(T, ip)    ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
 237 #define TCP(p)          ((struct tcphdr *)(p))
 238 #define SCTP(p)         ((struct sctphdr *)(p))
 239 #define UDP(p)          ((struct udphdr *)(p))
 240 #define ICMP(p)         ((struct icmphdr *)(p))
 241 #define ICMP6(p)        ((struct icmp6_hdr *)(p))
 242
 243 static __inline int
 244 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
 245 {
 246         int type = icmp->icmp_type;
 247
 248         return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
 249 }
 250
 251 #define TT      ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
 252     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
 253
 254 static int
 255 is_icmp_query(struct icmphdr *icmp)
 256 {
 257         int type = icmp->icmp_type;
 258
 259         return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
 260 }
 261 #undef TT
 262
 263 /*
 264  * The following checks use two arrays of 8 or 16 bits to store the
 265  * bits that we want set or clear, respectively. They are in the
 266  * low and high half of cmd->arg1 or cmd->d[0].
 267  *
 268  * We scan options and store the bits we find set. We succeed if
 269  *
 270  *      (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
 271  *
 272  * The code is sometimes optimized not to store additional variables.
 273  */
 274
 275 static int
 276 flags_match(ipfw_insn *cmd, u_int8_t bits)
 277 {
 278         u_char want_clear;
 279         bits = ~bits;
 280
 281         if ( ((cmd->arg1 & 0xff) & bits) != 0)
 282                 return 0; /* some bits we want set were clear */
 283         want_clear = (cmd->arg1 >> 8) & 0xff;
 284         if ( (want_clear & bits) != want_clear)
 285                 return 0; /* some bits we want clear were set */
 286         return 1;
 287 }
 288
 289 static int
 290 ipopts_match(struct ip *ip, ipfw_insn *cmd)
 291 {
 292         int optlen, bits = 0;
 293         u_char *cp = (u_char *)(ip + 1);
 294         int x = (ip->ip_hl << 2) - sizeof (struct ip);
 295
 296         for (; x > 0; x -= optlen, cp += optlen) {
 297                 int opt = cp[IPOPT_OPTVAL];
 298
 299                 if (opt == IPOPT_EOL)
 300                         break;
 301                 if (opt == IPOPT_NOP)
 302                         optlen = 1;
 303                 else {
 304                         optlen = cp[IPOPT_OLEN];
 305                         if (optlen <= 0 || optlen > x)
 306                                 return 0; /* invalid or truncated */
 307                 }
 308                 switch (opt) {
 309
 310                 default:
 311                         break;
 312
 313                 case IPOPT_LSRR:
 314                         bits |= IP_FW_IPOPT_LSRR;
 315                         break;
 316
 317                 case IPOPT_SSRR:
 318                         bits |= IP_FW_IPOPT_SSRR;
 319                         break;
 320
 321                 case IPOPT_RR:
 322                         bits |= IP_FW_IPOPT_RR;
 323                         break;
 324
 325                 case IPOPT_TS:
 326                         bits |= IP_FW_IPOPT_TS;
 327                         break;
 328                 }
 329         }
 330         return (flags_match(cmd, bits));
 331 }
 332
 333 static int
 334 tcpopts_parse(struct tcphdr *tcp, uint16_t *mss)
 335 {
 336         u_char *cp = (u_char *)(tcp + 1);
 337         int optlen, bits = 0;
 338         int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
 339
 340         for (; x > 0; x -= optlen, cp += optlen) {
 341                 int opt = cp[0];
 342                 if (opt == TCPOPT_EOL)
 343                         break;
 344                 if (opt == TCPOPT_NOP)
 345                         optlen = 1;
 346                 else {
 347                         optlen = cp[1];
 348                         if (optlen <= 0)
 349                                 break;
 350                 }
 351
 352                 switch (opt) {
 353                 default:
 354                         break;
 355
 356                 case TCPOPT_MAXSEG:
 357                         bits |= IP_FW_TCPOPT_MSS;
 358                         if (mss != NULL)
 359                                 *mss = be16dec(cp + 2);
 360                         break;
 361
 362                 case TCPOPT_WINDOW:
 363                         bits |= IP_FW_TCPOPT_WINDOW;
 364                         break;
 365
 366                 case TCPOPT_SACK_PERMITTED:
 367                 case TCPOPT_SACK:
 368                         bits |= IP_FW_TCPOPT_SACK;
 369                         break;
 370
 371                 case TCPOPT_TIMESTAMP:
 372                         bits |= IP_FW_TCPOPT_TS;
 373                         break;
 374                 }
 375         }
 376         return (bits);
 377 }
 378
 379 static int
 380 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
 381 {
 382
 383         return (flags_match(cmd, tcpopts_parse(tcp, NULL)));
 384 }
 385
 386 static int
 387 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 388     uint32_t *tablearg)
 389 {
 390
 391         if (ifp == NULL)        /* no iface with this packet, match fails */
 392                 return (0);
 393
 394         /* Check by name or by IP address */
 395         if (cmd->name[0] != '\0') { /* match by name */
 396                 if (cmd->name[0] == '\1') /* use tablearg to match */
 397                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 398                             &ifp->if_index, tablearg);
 399                 /* Check name */
 400                 if (cmd->p.glob) {
 401                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 402                                 return(1);
 403                 } else {
 404                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 405                                 return(1);
 406                 }
 407         } else {
 408 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 409                 struct ifaddr *ia;
 410
 411                 if_addr_rlock(ifp);
 412                 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 413                         if (ia->ifa_addr->sa_family != AF_INET)
 414                                 continue;
 415                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 416                             (ia->ifa_addr))->sin_addr.s_addr) {
 417                                 if_addr_runlock(ifp);
 418                                 return(1);      /* match */
 419                         }
 420                 }
 421                 if_addr_runlock(ifp);
 422 #endif /* __FreeBSD__ */
 423         }
 424         return(0);      /* no match, fail ... */
 425 }
 426
 427 /*
 428  * The verify_path function checks if a route to the src exists and
 429  * if it is reachable via ifp (when provided).
 430  *
 431  * The 'verrevpath' option checks that the interface that an IP packet
 432  * arrives on is the same interface that traffic destined for the
 433  * packet's source address would be routed out of.
 434  * The 'versrcreach' option just checks that the source address is
 435  * reachable via any route (except default) in the routing table.
 436  * These two are a measure to block forged packets. This is also
 437  * commonly known as "anti-spoofing" or Unicast Reverse Path
 438  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 439  * is purposely reminiscent of the Cisco IOS command,
 440  *
 441  *   ip verify unicast reverse-path
 442  *   ip verify unicast source reachable-via any
 443  *
 444  * which implements the same functionality. But note that the syntax
 445  * is misleading, and the check may be performed on all IP packets
 446  * whether unicast, multicast, or broadcast.
 447  */
 448 static int
 449 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 450 {
 451 #if defined(USERSPACE) || !defined(__FreeBSD__)
 452         return 0;
 453 #else
 454         struct nhop4_basic nh4;
 455
 456         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 457                 return (0);
 458
 459         /*
 460          * If ifp is provided, check for equality with rtentry.
 461          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 462          * in order to pass packets injected back by if_simloop():
 463          * routing entry (via lo0) for our own address
 464          * may exist, so we need to handle routing assymetry.
 465          */
 466         if (ifp != NULL && ifp != nh4.nh_ifp)
 467                 return (0);
 468
 469         /* if no ifp provided, check if rtentry is not default route */
 470         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 471                 return (0);
 472
 473         /* or if this is a blackhole/reject route */
 474         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 475                 return (0);
 476
 477         /* found valid route */
 478         return 1;
 479 #endif /* __FreeBSD__ */
 480 }
 481
 482 /*
 483  * Generate an SCTP packet containing an ABORT chunk. The verification tag
 484  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
 485  * reflected is not 0.
 486  */
 487
 488 static struct mbuf *
 489 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
 490     int reflected)
 491 {
 492         struct mbuf *m;
 493         struct ip *ip;
 494 #ifdef INET6
 495         struct ip6_hdr *ip6;
 496 #endif
 497         struct sctphdr *sctp;
 498         struct sctp_chunkhdr *chunk;
 499         u_int16_t hlen, plen, tlen;
 500
 501         MGETHDR(m, M_NOWAIT, MT_DATA);
 502         if (m == NULL)
 503                 return (NULL);
 504
 505         M_SETFIB(m, id->fib);
 506 #ifdef MAC
 507         if (replyto != NULL)
 508                 mac_netinet_firewall_reply(replyto, m);
 509         else
 510                 mac_netinet_firewall_send(m);
 511 #else
 512         (void)replyto;          /* don't warn about unused arg */
 513 #endif
 514
 515         switch (id->addr_type) {
 516         case 4:
 517                 hlen = sizeof(struct ip);
 518                 break;
 519 #ifdef INET6
 520         case 6:
 521                 hlen = sizeof(struct ip6_hdr);
 522                 break;
 523 #endif
 524         default:
 525                 /* XXX: log me?!? */
 526                 FREE_PKT(m);
 527                 return (NULL);
 528         }
 529         plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
 530         tlen = hlen + plen;
 531         m->m_data += max_linkhdr;
 532         m->m_flags |= M_SKIP_FIREWALL;
 533         m->m_pkthdr.len = m->m_len = tlen;
 534         m->m_pkthdr.rcvif = NULL;
 535         bzero(m->m_data, tlen);
 536
 537         switch (id->addr_type) {
 538         case 4:
 539                 ip = mtod(m, struct ip *);
 540
 541                 ip->ip_v = 4;
 542                 ip->ip_hl = sizeof(struct ip) >> 2;
 543                 ip->ip_tos = IPTOS_LOWDELAY;
 544                 ip->ip_len = htons(tlen);
 545                 ip->ip_id = htons(0);
 546                 ip->ip_off = htons(0);
 547                 ip->ip_ttl = V_ip_defttl;
 548                 ip->ip_p = IPPROTO_SCTP;
 549                 ip->ip_sum = 0;
 550                 ip->ip_src.s_addr = htonl(id->dst_ip);
 551                 ip->ip_dst.s_addr = htonl(id->src_ip);
 552
 553                 sctp = (struct sctphdr *)(ip + 1);
 554                 break;
 555 #ifdef INET6
 556         case 6:
 557                 ip6 = mtod(m, struct ip6_hdr *);
 558
 559                 ip6->ip6_vfc = IPV6_VERSION;
 560                 ip6->ip6_plen = htons(plen);
 561                 ip6->ip6_nxt = IPPROTO_SCTP;
 562                 ip6->ip6_hlim = IPV6_DEFHLIM;
 563                 ip6->ip6_src = id->dst_ip6;
 564                 ip6->ip6_dst = id->src_ip6;
 565
 566                 sctp = (struct sctphdr *)(ip6 + 1);
 567                 break;
 568 #endif
 569         }
 570
 571         sctp->src_port = htons(id->dst_port);
 572         sctp->dest_port = htons(id->src_port);
 573         sctp->v_tag = htonl(vtag);
 574         sctp->checksum = htonl(0);
 575
 576         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 577         chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
 578         chunk->chunk_flags = 0;
 579         if (reflected != 0) {
 580                 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
 581         }
 582         chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
 583
 584         sctp->checksum = sctp_calculate_cksum(m, hlen);
 585
 586         return (m);
 587 }
 588
 589 /*
 590  * Generate a TCP packet, containing either a RST or a keepalive.
 591  * When flags & TH_RST, we are sending a RST packet, because of a
 592  * "reset" action matched the packet.
 593  * Otherwise we are sending a keepalive, and flags & TH_
 594  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
 595  * so that MAC can label the reply appropriately.
 596  */
 597 struct mbuf *
 598 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
 599     u_int32_t ack, int flags)
 600 {
 601         struct mbuf *m = NULL;          /* stupid compiler */
 602         struct ip *h = NULL;            /* stupid compiler */
 603 #ifdef INET6
 604         struct ip6_hdr *h6 = NULL;
 605 #endif
 606         struct tcphdr *th = NULL;
 607         int len, dir;
 608
 609         MGETHDR(m, M_NOWAIT, MT_DATA);
 610         if (m == NULL)
 611                 return (NULL);
 612
 613         M_SETFIB(m, id->fib);
 614 #ifdef MAC
 615         if (replyto != NULL)
 616                 mac_netinet_firewall_reply(replyto, m);
 617         else
 618                 mac_netinet_firewall_send(m);
 619 #else
 620         (void)replyto;          /* don't warn about unused arg */
 621 #endif
 622
 623         switch (id->addr_type) {
 624         case 4:
 625                 len = sizeof(struct ip) + sizeof(struct tcphdr);
 626                 break;
 627 #ifdef INET6
 628         case 6:
 629                 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
 630                 break;
 631 #endif
 632         default:
 633                 /* XXX: log me?!? */
 634                 FREE_PKT(m);
 635                 return (NULL);
 636         }
 637         dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
 638
 639         m->m_data += max_linkhdr;
 640         m->m_flags |= M_SKIP_FIREWALL;
 641         m->m_pkthdr.len = m->m_len = len;
 642         m->m_pkthdr.rcvif = NULL;
 643         bzero(m->m_data, len);
 644
 645         switch (id->addr_type) {
 646         case 4:
 647                 h = mtod(m, struct ip *);
 648
 649                 /* prepare for checksum */
 650                 h->ip_p = IPPROTO_TCP;
 651                 h->ip_len = htons(sizeof(struct tcphdr));
 652                 if (dir) {
 653                         h->ip_src.s_addr = htonl(id->src_ip);
 654                         h->ip_dst.s_addr = htonl(id->dst_ip);
 655                 } else {
 656                         h->ip_src.s_addr = htonl(id->dst_ip);
 657                         h->ip_dst.s_addr = htonl(id->src_ip);
 658                 }
 659
 660                 th = (struct tcphdr *)(h + 1);
 661                 break;
 662 #ifdef INET6
 663         case 6:
 664                 h6 = mtod(m, struct ip6_hdr *);
 665
 666                 /* prepare for checksum */
 667                 h6->ip6_nxt = IPPROTO_TCP;
 668                 h6->ip6_plen = htons(sizeof(struct tcphdr));
 669                 if (dir) {
 670                         h6->ip6_src = id->src_ip6;
 671                         h6->ip6_dst = id->dst_ip6;
 672                 } else {
 673                         h6->ip6_src = id->dst_ip6;
 674                         h6->ip6_dst = id->src_ip6;
 675                 }
 676
 677                 th = (struct tcphdr *)(h6 + 1);
 678                 break;
 679 #endif
 680         }
 681
 682         if (dir) {
 683                 th->th_sport = htons(id->src_port);
 684                 th->th_dport = htons(id->dst_port);
 685         } else {
 686                 th->th_sport = htons(id->dst_port);
 687                 th->th_dport = htons(id->src_port);
 688         }
 689         th->th_off = sizeof(struct tcphdr) >> 2;
 690
 691         if (flags & TH_RST) {
 692                 if (flags & TH_ACK) {
 693                         th->th_seq = htonl(ack);
 694                         th->th_flags = TH_RST;
 695                 } else {
 696                         if (flags & TH_SYN)
 697                                 seq++;
 698                         th->th_ack = htonl(seq);
 699                         th->th_flags = TH_RST | TH_ACK;
 700                 }
 701         } else {
 702                 /*
 703                  * Keepalive - use caller provided sequence numbers
 704                  */
 705                 th->th_seq = htonl(seq);
 706                 th->th_ack = htonl(ack);
 707                 th->th_flags = TH_ACK;
 708         }
 709
 710         switch (id->addr_type) {
 711         case 4:
 712                 th->th_sum = in_cksum(m, len);
 713
 714                 /* finish the ip header */
 715                 h->ip_v = 4;
 716                 h->ip_hl = sizeof(*h) >> 2;
 717                 h->ip_tos = IPTOS_LOWDELAY;
 718                 h->ip_off = htons(0);
 719                 h->ip_len = htons(len);
 720                 h->ip_ttl = V_ip_defttl;
 721                 h->ip_sum = 0;
 722                 break;
 723 #ifdef INET6
 724         case 6:
 725                 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
 726                     sizeof(struct tcphdr));
 727
 728                 /* finish the ip6 header */
 729                 h6->ip6_vfc |= IPV6_VERSION;
 730                 h6->ip6_hlim = IPV6_DEFHLIM;
 731                 break;
 732 #endif
 733         }
 734
 735         return (m);
 736 }
 737
 738 #ifdef INET6
 739 /*
 740  * ipv6 specific rules here...
 741  */
 742 static __inline int
 743 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 744 {
 745         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 746 }
 747
 748 static int
 749 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 750 {
 751         int i;
 752         for (i=0; i <= cmd->o.arg1; ++i )
 753                 if (curr_flow == cmd->d[i] )
 754                         return 1;
 755         return 0;
 756 }
 757
 758 /* support for IP6_*_ME opcodes */
 759 static const struct in6_addr lla_mask = {{{
 760         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 761         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 762 }}};
 763
 764 static int
 765 ipfw_localip6(struct in6_addr *in6)
 766 {
 767         struct rm_priotracker in6_ifa_tracker;
 768         struct in6_ifaddr *ia;
 769
 770         if (IN6_IS_ADDR_MULTICAST(in6))
 771                 return (0);
 772
 773         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 774                 return (in6_localip(in6));
 775
 776         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 777         CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 778                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 779                         continue;
 780                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 781                     in6, &lla_mask)) {
 782                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 783                         return (1);
 784                 }
 785         }
 786         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 787         return (0);
 788 }
 789
 790 static int
 791 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 792 {
 793         struct nhop6_basic nh6;
 794
 795         if (IN6_IS_SCOPE_LINKLOCAL(src))
 796                 return (1);
 797
 798         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 799                 return (0);
 800
 801         /* If ifp is provided, check for equality with route table. */
 802         if (ifp != NULL && ifp != nh6.nh_ifp)
 803                 return (0);
 804
 805         /* if no ifp provided, check if rtentry is not default route */
 806         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 807                 return (0);
 808
 809         /* or if this is a blackhole/reject route */
 810         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 811                 return (0);
 812
 813         /* found valid route */
 814         return 1;
 815 }
 816
 817 static int
 818 is_icmp6_query(int icmp6_type)
 819 {
 820         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 821             (icmp6_type == ICMP6_ECHO_REQUEST ||
 822             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 823             icmp6_type == ICMP6_WRUREQUEST ||
 824             icmp6_type == ICMP6_FQDN_QUERY ||
 825             icmp6_type == ICMP6_NI_QUERY))
 826                 return (1);
 827
 828         return (0);
 829 }
 830
 831 static int
 832 map_icmp_unreach(int code)
 833 {
 834
 835         /* RFC 7915 p4.2 */
 836         switch (code) {
 837         case ICMP_UNREACH_NET:
 838         case ICMP_UNREACH_HOST:
 839         case ICMP_UNREACH_SRCFAIL:
 840         case ICMP_UNREACH_NET_UNKNOWN:
 841         case ICMP_UNREACH_HOST_UNKNOWN:
 842         case ICMP_UNREACH_TOSNET:
 843         case ICMP_UNREACH_TOSHOST:
 844                 return (ICMP6_DST_UNREACH_NOROUTE);
 845         case ICMP_UNREACH_PORT:
 846                 return (ICMP6_DST_UNREACH_NOPORT);
 847         default:
 848                 /*
 849                  * Map the rest of codes into admit prohibited.
 850                  * XXX: unreach proto should be mapped into ICMPv6
 851                  * parameter problem, but we use only unreach type.
 852                  */
 853                 return (ICMP6_DST_UNREACH_ADMIN);
 854         }
 855 }
 856
 857 static void
 858 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 859 {
 860         struct mbuf *m;
 861
 862         m = args->m;
 863         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 864                 struct tcphdr *tcp;
 865                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 866
 867                 if ((tcp->th_flags & TH_RST) == 0) {
 868                         struct mbuf *m0;
 869                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 870                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 871                             tcp->th_flags | TH_RST);
 872                         if (m0 != NULL)
 873                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 874                                     NULL);
 875                 }
 876                 FREE_PKT(m);
 877         } else if (code == ICMP6_UNREACH_ABORT &&
 878             args->f_id.proto == IPPROTO_SCTP) {
 879                 struct mbuf *m0;
 880                 struct sctphdr *sctp;
 881                 u_int32_t v_tag;
 882                 int reflected;
 883
 884                 sctp = (struct sctphdr *)((char *)ip6 + hlen);
 885                 reflected = 1;
 886                 v_tag = ntohl(sctp->v_tag);
 887                 /* Investigate the first chunk header if available */
 888                 if (m->m_len >= hlen + sizeof(struct sctphdr) +
 889                     sizeof(struct sctp_chunkhdr)) {
 890                         struct sctp_chunkhdr *chunk;
 891
 892                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 893                         switch (chunk->chunk_type) {
 894                         case SCTP_INITIATION:
 895                                 /*
 896                                  * Packets containing an INIT chunk MUST have
 897                                  * a zero v-tag.
 898                                  */
 899                                 if (v_tag != 0) {
 900                                         v_tag = 0;
 901                                         break;
 902                                 }
 903                                 /* INIT chunk MUST NOT be bundled */
 904                                 if (m->m_pkthdr.len >
 905                                     hlen + sizeof(struct sctphdr) +
 906                                     ntohs(chunk->chunk_length) + 3) {
 907                                         break;
 908                                 }
 909                                 /* Use the initiate tag if available */
 910                                 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
 911                                     sizeof(struct sctp_chunkhdr) +
 912                                     offsetof(struct sctp_init, a_rwnd))) {
 913                                         struct sctp_init *init;
 914
 915                                         init = (struct sctp_init *)(chunk + 1);
 916                                         v_tag = ntohl(init->initiate_tag);
 917                                         reflected = 0;
 918                                 }
 919                                 break;
 920                         case SCTP_ABORT_ASSOCIATION:
 921                                 /*
 922                                  * If the packet contains an ABORT chunk, don't
 923                                  * reply.
 924                                  * XXX: We should search through all chunks,
 925                                  *      but don't do to avoid attacks.
 926                                  */
 927                                 v_tag = 0;
 928                                 break;
 929                         }
 930                 }
 931                 if (v_tag == 0) {
 932                         m0 = NULL;
 933                 } else {
 934                         m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
 935                             reflected);
 936                 }
 937                 if (m0 != NULL)
 938                         ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
 939                 FREE_PKT(m);
 940         } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
 941                 /* Send an ICMPv6 unreach. */
 942 #if 0
 943                 /*
 944                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 945                  * as the contents are read. We need to m_adj() the
 946                  * needed amount.
 947                  * The mbuf will however be thrown away so we can adjust it.
 948                  * Remember we did an m_pullup on it already so we
 949                  * can make some assumptions about contiguousness.
 950                  */
 951                 if (args->L3offset)
 952                         m_adj(m, args->L3offset);
 953 #endif
 954                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 955         } else
 956                 FREE_PKT(m);
 957
 958         args->m = NULL;
 959 }
 960
 961 #endif /* INET6 */
 962
 963
 964 /*
 965  * sends a reject message, consuming the mbuf passed as an argument.
 966  */
 967 static void
 968 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 969 {
 970
 971 #if 0
 972         /* XXX When ip is not guaranteed to be at mtod() we will
 973          * need to account for this */
 974          * The mbuf will however be thrown away so we can adjust it.
 975          * Remember we did an m_pullup on it already so we
 976          * can make some assumptions about contiguousness.
 977          */
 978         if (args->L3offset)
 979                 m_adj(m, args->L3offset);
 980 #endif
 981         if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
 982                 /* Send an ICMP unreach */
 983                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 984         } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
 985                 struct tcphdr *const tcp =
 986                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 987                 if ( (tcp->th_flags & TH_RST) == 0) {
 988                         struct mbuf *m;
 989                         m = ipfw_send_pkt(args->m, &(args->f_id),
 990                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 991                                 tcp->th_flags | TH_RST);
 992                         if (m != NULL)
 993                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 994                 }
 995                 FREE_PKT(args->m);
 996         } else if (code == ICMP_REJECT_ABORT &&
 997             args->f_id.proto == IPPROTO_SCTP) {
 998                 struct mbuf *m;
 999                 struct sctphdr *sctp;
1000                 struct sctp_chunkhdr *chunk;
1001                 struct sctp_init *init;
1002                 u_int32_t v_tag;
1003                 int reflected;
1004
1005                 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1006                 reflected = 1;
1007                 v_tag = ntohl(sctp->v_tag);
1008                 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1009                     sizeof(struct sctp_chunkhdr)) {
1010                         /* Look at the first chunk header if available */
1011                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
1012                         switch (chunk->chunk_type) {
1013                         case SCTP_INITIATION:
1014                                 /*
1015                                  * Packets containing an INIT chunk MUST have
1016                                  * a zero v-tag.
1017                                  */
1018                                 if (v_tag != 0) {
1019                                         v_tag = 0;
1020                                         break;
1021                                 }
1022                                 /* INIT chunk MUST NOT be bundled */
1023                                 if (iplen >
1024                                     (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1025                                     ntohs(chunk->chunk_length) + 3) {
1026                                         break;
1027                                 }
1028                                 /* Use the initiate tag if available */
1029                                 if ((iplen >= (ip->ip_hl << 2) +
1030                                     sizeof(struct sctphdr) +
1031                                     sizeof(struct sctp_chunkhdr) +
1032                                     offsetof(struct sctp_init, a_rwnd))) {
1033                                         init = (struct sctp_init *)(chunk + 1);
1034                                         v_tag = ntohl(init->initiate_tag);
1035                                         reflected = 0;
1036                                 }
1037                                 break;
1038                         case SCTP_ABORT_ASSOCIATION:
1039                                 /*
1040                                  * If the packet contains an ABORT chunk, don't
1041                                  * reply.
1042                                  * XXX: We should search through all chunks,
1043                                  * but don't do to avoid attacks.
1044                                  */
1045                                 v_tag = 0;
1046                                 break;
1047                         }
1048                 }
1049                 if (v_tag == 0) {
1050                         m = NULL;
1051                 } else {
1052                         m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1053                             reflected);
1054                 }
1055                 if (m != NULL)
1056                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1057                 FREE_PKT(args->m);
1058         } else
1059                 FREE_PKT(args->m);
1060         args->m = NULL;
1061 }
1062
1063 /*
1064  * Support for uid/gid/jail lookup. These tests are expensive
1065  * (because we may need to look into the list of active sockets)
1066  * so we cache the results. ugid_lookupp is 0 if we have not
1067  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1068  * and failed. The function always returns the match value.
1069  * We could actually spare the variable and use *uc, setting
1070  * it to '(void *)check_uidgid if we have no info, NULL if
1071  * we tried and failed, or any other value if successful.
1072  */
1073 static int
1074 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1075     struct ucred **uc)
1076 {
1077 #if defined(USERSPACE)
1078         return 0;       // not supported in userspace
1079 #else
1080 #ifndef __FreeBSD__
1081         /* XXX */
1082         return cred_check(insn, proto, oif,
1083             dst_ip, dst_port, src_ip, src_port,
1084             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1085 #else  /* FreeBSD */
1086         struct in_addr src_ip, dst_ip;
1087         struct inpcbinfo *pi;
1088         struct ipfw_flow_id *id;
1089         struct inpcb *pcb, *inp;
1090         struct ifnet *oif;
1091         int lookupflags;
1092         int match;
1093
1094         id = &args->f_id;
1095         inp = args->inp;
1096         oif = args->oif;
1097
1098         /*
1099          * Check to see if the UDP or TCP stack supplied us with
1100          * the PCB. If so, rather then holding a lock and looking
1101          * up the PCB, we can use the one that was supplied.
1102          */
1103         if (inp && *ugid_lookupp == 0) {
1104                 INP_LOCK_ASSERT(inp);
1105                 if (inp->inp_socket != NULL) {
1106                         *uc = crhold(inp->inp_cred);
1107                         *ugid_lookupp = 1;
1108                 } else
1109                         *ugid_lookupp = -1;
1110         }
1111         /*
1112          * If we have already been here and the packet has no
1113          * PCB entry associated with it, then we can safely
1114          * assume that this is a no match.
1115          */
1116         if (*ugid_lookupp == -1)
1117                 return (0);
1118         if (id->proto == IPPROTO_TCP) {
1119                 lookupflags = 0;
1120                 pi = &V_tcbinfo;
1121         } else if (id->proto == IPPROTO_UDP) {
1122                 lookupflags = INPLOOKUP_WILDCARD;
1123                 pi = &V_udbinfo;
1124         } else if (id->proto == IPPROTO_UDPLITE) {
1125                 lookupflags = INPLOOKUP_WILDCARD;
1126                 pi = &V_ulitecbinfo;
1127         } else
1128                 return 0;
1129         lookupflags |= INPLOOKUP_RLOCKPCB;
1130         match = 0;
1131         if (*ugid_lookupp == 0) {
1132                 if (id->addr_type == 6) {
1133 #ifdef INET6
1134                         if (oif == NULL)
1135                                 pcb = in6_pcblookup_mbuf(pi,
1136                                     &id->src_ip6, htons(id->src_port),
1137                                     &id->dst_ip6, htons(id->dst_port),
1138                                     lookupflags, oif, args->m);
1139                         else
1140                                 pcb = in6_pcblookup_mbuf(pi,
1141                                     &id->dst_ip6, htons(id->dst_port),
1142                                     &id->src_ip6, htons(id->src_port),
1143                                     lookupflags, oif, args->m);
1144 #else
1145                         *ugid_lookupp = -1;
1146                         return (0);
1147 #endif
1148                 } else {
1149                         src_ip.s_addr = htonl(id->src_ip);
1150                         dst_ip.s_addr = htonl(id->dst_ip);
1151                         if (oif == NULL)
1152                                 pcb = in_pcblookup_mbuf(pi,
1153                                     src_ip, htons(id->src_port),
1154                                     dst_ip, htons(id->dst_port),
1155                                     lookupflags, oif, args->m);
1156                         else
1157                                 pcb = in_pcblookup_mbuf(pi,
1158                                     dst_ip, htons(id->dst_port),
1159                                     src_ip, htons(id->src_port),
1160                                     lookupflags, oif, args->m);
1161                 }
1162                 if (pcb != NULL) {
1163                         INP_RLOCK_ASSERT(pcb);
1164                         *uc = crhold(pcb->inp_cred);
1165                         *ugid_lookupp = 1;
1166                         INP_RUNLOCK(pcb);
1167                 }
1168                 if (*ugid_lookupp == 0) {
1169                         /*
1170                          * We tried and failed, set the variable to -1
1171                          * so we will not try again on this packet.
1172                          */
1173                         *ugid_lookupp = -1;
1174                         return (0);
1175                 }
1176         }
1177         if (insn->o.opcode == O_UID)
1178                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1179         else if (insn->o.opcode == O_GID)
1180                 match = groupmember((gid_t)insn->d[0], *uc);
1181         else if (insn->o.opcode == O_JAIL)
1182                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1183         return (match);
1184 #endif /* __FreeBSD__ */
1185 #endif /* not supported in userspace */
1186 }
1187
1188 /*
1189  * Helper function to set args with info on the rule after the matching
1190  * one. slot is precise, whereas we guess rule_id as they are
1191  * assigned sequentially.
1192  */
1193 static inline void
1194 set_match(struct ip_fw_args *args, int slot,
1195         struct ip_fw_chain *chain)
1196 {
1197         args->rule.chain_id = chain->id;
1198         args->rule.slot = slot + 1; /* we use 0 as a marker */
1199         args->rule.rule_id = 1 + chain->map[slot]->id;
1200         args->rule.rulenum = chain->map[slot]->rulenum;
1201         args->flags |= IPFW_ARGS_REF;
1202 }
1203
1204 #ifndef LINEAR_SKIPTO
1205 /*
1206  * Helper function to enable cached rule lookups using
1207  * cached_id and cached_pos fields in ipfw rule.
1208  */
1209 static int
1210 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1211     int tablearg, int jump_backwards)
1212 {
1213         int f_pos;
1214
1215         /* If possible use cached f_pos (in f->cached_pos),
1216          * whose version is written in f->cached_id
1217          * (horrible hacks to avoid changing the ABI).
1218          */
1219         if (num != IP_FW_TARG && f->cached_id == chain->id)
1220                 f_pos = f->cached_pos;
1221         else {
1222                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1223                 /* make sure we do not jump backward */
1224                 if (jump_backwards == 0 && i <= f->rulenum)
1225                         i = f->rulenum + 1;
1226                 if (chain->idxmap != NULL)
1227                         f_pos = chain->idxmap[i];
1228                 else
1229                         f_pos = ipfw_find_rule(chain, i, 0);
1230                 /* update the cache */
1231                 if (num != IP_FW_TARG) {
1232                         f->cached_id = chain->id;
1233                         f->cached_pos = f_pos;
1234                 }
1235         }
1236
1237         return (f_pos);
1238 }
1239 #else
1240 /*
1241  * Helper function to enable real fast rule lookups.
1242  */
1243 static int
1244 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1245     int tablearg, int jump_backwards)
1246 {
1247         int f_pos;
1248
1249         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1250         /* make sure we do not jump backward */
1251         if (jump_backwards == 0 && num <= f->rulenum)
1252                 num = f->rulenum + 1;
1253         f_pos = chain->idxmap[num];
1254
1255         return (f_pos);
1256 }
1257 #endif
1258
1259 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1260 /*
1261  * The main check routine for the firewall.
1262  *
1263  * All arguments are in args so we can modify them and return them
1264  * back to the caller.
1265  *
1266  * Parameters:
1267  *
1268  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1269  *              Starts with the IP header.
1270  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
1271  *      args->L3offset  Number of bytes bypassed if we came from L2.
1272  *                      e.g. often sizeof(eh)  ** NOTYET **
1273  *      args->oif       Outgoing interface, NULL if packet is incoming.
1274  *              The incoming interface is in the mbuf. (in)
1275  *      args->divert_rule (in/out)
1276  *              Skip up to the first rule past this rule number;
1277  *              upon return, non-zero port number for divert or tee.
1278  *
1279  *      args->rule      Pointer to the last matching rule (in/out)
1280  *      args->next_hop  Socket we are forwarding to (out).
1281  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1282  *      args->f_id      Addresses grabbed from the packet (out)
1283  *      args->rule.info a cookie depending on rule action
1284  *
1285  * Return value:
1286  *
1287  *      IP_FW_PASS      the packet must be accepted
1288  *      IP_FW_DENY      the packet must be dropped
1289  *      IP_FW_DIVERT    divert packet, port in m_tag
1290  *      IP_FW_TEE       tee packet, port in m_tag
1291  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1292  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1293  *              args->rule contains the matching rule,
1294  *              args->rule.info has additional information.
1295  *
1296  */
1297 int
1298 ipfw_chk(struct ip_fw_args *args)
1299 {
1300
1301         /*
1302          * Local variables holding state while processing a packet:
1303          *
1304          * IMPORTANT NOTE: to speed up the processing of rules, there
1305          * are some assumption on the values of the variables, which
1306          * are documented here. Should you change them, please check
1307          * the implementation of the various instructions to make sure
1308          * that they still work.
1309          *
1310          * args->eh     The MAC header. It is non-null for a layer2
1311          *      packet, it is NULL for a layer-3 packet.
1312          * **notyet**
1313          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1314          *
1315          * m | args->m  Pointer to the mbuf, as received from the caller.
1316          *      It may change if ipfw_chk() does an m_pullup, or if it
1317          *      consumes the packet because it calls send_reject().
1318          *      XXX This has to change, so that ipfw_chk() never modifies
1319          *      or consumes the buffer.
1320          * ip   is the beginning of the ip(4 or 6) header.
1321          *      Calculated by adding the L3offset to the start of data.
1322          *      (Until we start using L3offset, the packet is
1323          *      supposed to start with the ip header).
1324          */
1325         struct mbuf *m = args->m;
1326         struct ip *ip = mtod(m, struct ip *);
1327
1328         /*
1329          * For rules which contain uid/gid or jail constraints, cache
1330          * a copy of the users credentials after the pcb lookup has been
1331          * executed. This will speed up the processing of rules with
1332          * these types of constraints, as well as decrease contention
1333          * on pcb related locks.
1334          */
1335 #ifndef __FreeBSD__
1336         struct bsd_ucred ucred_cache;
1337 #else
1338         struct ucred *ucred_cache = NULL;
1339 #endif
1340         int ucred_lookup = 0;
1341
1342         /*
1343          * oif | args->oif      If NULL, ipfw_chk has been called on the
1344          *      inbound path (ether_input, ip_input).
1345          *      If non-NULL, ipfw_chk has been called on the outbound path
1346          *      (ether_output, ip_output).
1347          */
1348         struct ifnet *oif = args->oif;
1349
1350         int f_pos = 0;          /* index of current rule in the array */
1351         int retval = 0;
1352
1353         /*
1354          * hlen The length of the IP header.
1355          */
1356         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1357
1358         /*
1359          * offset       The offset of a fragment. offset != 0 means that
1360          *      we have a fragment at this offset of an IPv4 packet.
1361          *      offset == 0 means that (if this is an IPv4 packet)
1362          *      this is the first or only fragment.
1363          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1364          *      or there is a single packet fragment (fragment header added
1365          *      without needed).  We will treat a single packet fragment as if
1366          *      there was no fragment header (or log/block depending on the
1367          *      V_fw_permit_single_frag6 sysctl setting).
1368          */
1369         u_short offset = 0;
1370         u_short ip6f_mf = 0;
1371
1372         /*
1373          * Local copies of addresses. They are only valid if we have
1374          * an IP packet.
1375          *
1376          * proto        The protocol. Set to 0 for non-ip packets,
1377          *      or to the protocol read from the packet otherwise.
1378          *      proto != 0 means that we have an IPv4 packet.
1379          *
1380          * src_port, dst_port   port numbers, in HOST format. Only
1381          *      valid for TCP and UDP packets.
1382          *
1383          * src_ip, dst_ip       ip addresses, in NETWORK format.
1384          *      Only valid for IPv4 packets.
1385          */
1386         uint8_t proto;
1387         uint16_t src_port, dst_port;            /* NOTE: host format    */
1388         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1389         int iplen = 0;
1390         int pktlen;
1391         uint16_t etype;                 /* Host order stored ether type */
1392
1393         struct ipfw_dyn_info dyn_info;
1394         struct ip_fw *q = NULL;
1395         struct ip_fw_chain *chain = &V_layer3_chain;
1396
1397         /*
1398          * We store in ulp a pointer to the upper layer protocol header.
1399          * In the ipv4 case this is easy to determine from the header,
1400          * but for ipv6 we might have some additional headers in the middle.
1401          * ulp is NULL if not found.
1402          */
1403         void *ulp = NULL;               /* upper layer protocol pointer. */
1404
1405         /* XXX ipv6 variables */
1406         int is_ipv6 = 0;
1407         uint8_t icmp6_type = 0;
1408         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1409         /* end of ipv6 variables */
1410
1411         int is_ipv4 = 0;
1412
1413         int done = 0;           /* flag to exit the outer loop */
1414
1415         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1416                 return (IP_FW_PASS);    /* accept */
1417
1418         dst_ip.s_addr = 0;              /* make sure it is initialized */
1419         src_ip.s_addr = 0;              /* make sure it is initialized */
1420         src_port = dst_port = 0;
1421         pktlen = m->m_pkthdr.len;
1422
1423         DYN_INFO_INIT(&dyn_info);
1424 /*
1425  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1426  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1427  * pointer might become stale after other pullups (but we never use it
1428  * this way).
1429  */
1430 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1431 #define PULLUP_LEN(_len, p, T)                                  \
1432 do {                                                            \
1433         int x = (_len) + T;                                     \
1434         if ((m)->m_len < x) {                                   \
1435                 args->m = m = m_pullup(m, x);                   \
1436                 if (m == NULL)                                  \
1437                         goto pullup_failed;                     \
1438         }                                                       \
1439         p = (mtod(m, char *) + (_len));                         \
1440 } while (0)
1441
1442         /*
1443          * if we have an ether header,
1444          */
1445         if (args->flags & IPFW_ARGS_ETHER)
1446                 etype = ntohs(args->eh->ether_type);
1447         else
1448                 etype = 0;
1449
1450         /* Identify IP packets and fill up variables. */
1451         if (pktlen >= sizeof(struct ip6_hdr) &&
1452             (etype == 0 || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1453                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1454
1455                 is_ipv6 = 1;
1456                 hlen = sizeof(struct ip6_hdr);
1457                 proto = ip6->ip6_nxt;
1458                 /* Search extension headers to find upper layer protocols */
1459                 while (ulp == NULL && offset == 0) {
1460                         switch (proto) {
1461                         case IPPROTO_ICMPV6:
1462                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1463                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1464                                 break;
1465
1466                         case IPPROTO_TCP:
1467                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1468                                 dst_port = TCP(ulp)->th_dport;
1469                                 src_port = TCP(ulp)->th_sport;
1470                                 /* save flags for dynamic rules */
1471                                 args->f_id._flags = TCP(ulp)->th_flags;
1472                                 break;
1473
1474                         case IPPROTO_SCTP:
1475                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1476                                     sizeof(struct sctp_chunkhdr) +
1477                                     offsetof(struct sctp_init, a_rwnd))
1478                                         PULLUP_LEN(hlen, ulp,
1479                                             sizeof(struct sctphdr) +
1480                                             sizeof(struct sctp_chunkhdr) +
1481                                             offsetof(struct sctp_init, a_rwnd));
1482                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1483                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1484                                 else
1485                                         PULLUP_LEN(hlen, ulp,
1486                                             sizeof(struct sctphdr));
1487                                 src_port = SCTP(ulp)->src_port;
1488                                 dst_port = SCTP(ulp)->dest_port;
1489                                 break;
1490
1491                         case IPPROTO_UDP:
1492                         case IPPROTO_UDPLITE:
1493                                 PULLUP_TO(hlen, ulp, struct udphdr);
1494                                 dst_port = UDP(ulp)->uh_dport;
1495                                 src_port = UDP(ulp)->uh_sport;
1496                                 break;
1497
1498                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1499                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1500                                 ext_hd |= EXT_HOPOPTS;
1501                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1502                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1503                                 ulp = NULL;
1504                                 break;
1505
1506                         case IPPROTO_ROUTING:   /* RFC 2460 */
1507                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1508                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1509                                 case 0:
1510                                         ext_hd |= EXT_RTHDR0;
1511                                         break;
1512                                 case 2:
1513                                         ext_hd |= EXT_RTHDR2;
1514                                         break;
1515                                 default:
1516                                         if (V_fw_verbose)
1517                                                 printf("IPFW2: IPV6 - Unknown "
1518                                                     "Routing Header type(%d)\n",
1519                                                     ((struct ip6_rthdr *)
1520                                                     ulp)->ip6r_type);
1521                                         if (V_fw_deny_unknown_exthdrs)
1522                                             return (IP_FW_DENY);
1523                                         break;
1524                                 }
1525                                 ext_hd |= EXT_ROUTING;
1526                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1527                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1528                                 ulp = NULL;
1529                                 break;
1530
1531                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1532                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1533                                 ext_hd |= EXT_FRAGMENT;
1534                                 hlen += sizeof (struct ip6_frag);
1535                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1536                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1537                                         IP6F_OFF_MASK;
1538                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1539                                         IP6F_MORE_FRAG;
1540                                 if (V_fw_permit_single_frag6 == 0 &&
1541                                     offset == 0 && ip6f_mf == 0) {
1542                                         if (V_fw_verbose)
1543                                                 printf("IPFW2: IPV6 - Invalid "
1544                                                     "Fragment Header\n");
1545                                         if (V_fw_deny_unknown_exthdrs)
1546                                             return (IP_FW_DENY);
1547                                         break;
1548                                 }
1549                                 args->f_id.extra =
1550                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1551                                 ulp = NULL;
1552                                 break;
1553
1554                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1555                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1556                                 ext_hd |= EXT_DSTOPTS;
1557                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1558                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1559                                 ulp = NULL;
1560                                 break;
1561
1562                         case IPPROTO_AH:        /* RFC 2402 */
1563                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1564                                 ext_hd |= EXT_AH;
1565                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1566                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1567                                 ulp = NULL;
1568                                 break;
1569
1570                         case IPPROTO_ESP:       /* RFC 2406 */
1571                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1572                                 /* Anything past Seq# is variable length and
1573                                  * data past this ext. header is encrypted. */
1574                                 ext_hd |= EXT_ESP;
1575                                 break;
1576
1577                         case IPPROTO_NONE:      /* RFC 2460 */
1578                                 /*
1579                                  * Packet ends here, and IPv6 header has
1580                                  * already been pulled up. If ip6e_len!=0
1581                                  * then octets must be ignored.
1582                                  */
1583                                 ulp = ip; /* non-NULL to get out of loop. */
1584                                 break;
1585
1586                         case IPPROTO_OSPFIGP:
1587                                 /* XXX OSPF header check? */
1588                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1589                                 break;
1590
1591                         case IPPROTO_PIM:
1592                                 /* XXX PIM header check? */
1593                                 PULLUP_TO(hlen, ulp, struct pim);
1594                                 break;
1595
1596                         case IPPROTO_GRE:       /* RFC 1701 */
1597                                 /* XXX GRE header check? */
1598                                 PULLUP_TO(hlen, ulp, struct grehdr);
1599                                 break;
1600
1601                         case IPPROTO_CARP:
1602                                 PULLUP_TO(hlen, ulp, offsetof(
1603                                     struct carp_header, carp_counter));
1604                                 if (CARP_ADVERTISEMENT !=
1605                                     ((struct carp_header *)ulp)->carp_type)
1606                                         return (IP_FW_DENY);
1607                                 break;
1608
1609                         case IPPROTO_IPV6:      /* RFC 2893 */
1610                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1611                                 break;
1612
1613                         case IPPROTO_IPV4:      /* RFC 2893 */
1614                                 PULLUP_TO(hlen, ulp, struct ip);
1615                                 break;
1616
1617                         default:
1618                                 if (V_fw_verbose)
1619                                         printf("IPFW2: IPV6 - Unknown "
1620                                             "Extension Header(%d), ext_hd=%x\n",
1621                                              proto, ext_hd);
1622                                 if (V_fw_deny_unknown_exthdrs)
1623                                     return (IP_FW_DENY);
1624                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1625                                 break;
1626                         } /*switch */
1627                 }
1628                 ip = mtod(m, struct ip *);
1629                 ip6 = (struct ip6_hdr *)ip;
1630                 args->f_id.addr_type = 6;
1631                 args->f_id.src_ip6 = ip6->ip6_src;
1632                 args->f_id.dst_ip6 = ip6->ip6_dst;
1633                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1634                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1635         } else if (pktlen >= sizeof(struct ip) &&
1636             (etype == 0 || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1637                 is_ipv4 = 1;
1638                 hlen = ip->ip_hl << 2;
1639                 /*
1640                  * Collect parameters into local variables for faster
1641                  * matching.
1642                  */
1643                 proto = ip->ip_p;
1644                 src_ip = ip->ip_src;
1645                 dst_ip = ip->ip_dst;
1646                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1647                 iplen = ntohs(ip->ip_len);
1648
1649                 if (offset == 0) {
1650                         switch (proto) {
1651                         case IPPROTO_TCP:
1652                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1653                                 dst_port = TCP(ulp)->th_dport;
1654                                 src_port = TCP(ulp)->th_sport;
1655                                 /* save flags for dynamic rules */
1656                                 args->f_id._flags = TCP(ulp)->th_flags;
1657                                 break;
1658
1659                         case IPPROTO_SCTP:
1660                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1661                                     sizeof(struct sctp_chunkhdr) +
1662                                     offsetof(struct sctp_init, a_rwnd))
1663                                         PULLUP_LEN(hlen, ulp,
1664                                             sizeof(struct sctphdr) +
1665                                             sizeof(struct sctp_chunkhdr) +
1666                                             offsetof(struct sctp_init, a_rwnd));
1667                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1668                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1669                                 else
1670                                         PULLUP_LEN(hlen, ulp,
1671                                             sizeof(struct sctphdr));
1672                                 src_port = SCTP(ulp)->src_port;
1673                                 dst_port = SCTP(ulp)->dest_port;
1674                                 break;
1675
1676                         case IPPROTO_UDP:
1677                         case IPPROTO_UDPLITE:
1678                                 PULLUP_TO(hlen, ulp, struct udphdr);
1679                                 dst_port = UDP(ulp)->uh_dport;
1680                                 src_port = UDP(ulp)->uh_sport;
1681                                 break;
1682
1683                         case IPPROTO_ICMP:
1684                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1685                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1686                                 break;
1687
1688                         default:
1689                                 break;
1690                         }
1691                 }
1692
1693                 ip = mtod(m, struct ip *);
1694                 args->f_id.addr_type = 4;
1695                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1696                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1697         } else {
1698                 proto = 0;
1699                 dst_ip.s_addr = src_ip.s_addr = 0;
1700
1701                 args->f_id.addr_type = 1; /* XXX */
1702         }
1703 #undef PULLUP_TO
1704         pktlen = iplen < pktlen ? iplen: pktlen;
1705
1706         /* Properly initialize the rest of f_id */
1707         args->f_id.proto = proto;
1708         args->f_id.src_port = src_port = ntohs(src_port);
1709         args->f_id.dst_port = dst_port = ntohs(dst_port);
1710         args->f_id.fib = M_GETFIB(m);
1711
1712         IPFW_PF_RLOCK(chain);
1713         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1714                 IPFW_PF_RUNLOCK(chain);
1715                 return (IP_FW_PASS);    /* accept */
1716         }
1717         if (args->flags & IPFW_ARGS_REF) {
1718                 /*
1719                  * Packet has already been tagged as a result of a previous
1720                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1721                  * REASS, NETGRAPH, DIVERT/TEE...)
1722                  * Validate the slot and continue from the next one
1723                  * if still present, otherwise do a lookup.
1724                  */
1725                 f_pos = (args->rule.chain_id == chain->id) ?
1726                     args->rule.slot :
1727                     ipfw_find_rule(chain, args->rule.rulenum,
1728                         args->rule.rule_id);
1729         } else {
1730                 f_pos = 0;
1731         }
1732
1733         /*
1734          * Now scan the rules, and parse microinstructions for each rule.
1735          * We have two nested loops and an inner switch. Sometimes we
1736          * need to break out of one or both loops, or re-enter one of
1737          * the loops with updated variables. Loop variables are:
1738          *
1739          *      f_pos (outer loop) points to the current rule.
1740          *              On output it points to the matching rule.
1741          *      done (outer loop) is used as a flag to break the loop.
1742          *      l (inner loop)  residual length of current rule.
1743          *              cmd points to the current microinstruction.
1744          *
1745          * We break the inner loop by setting l=0 and possibly
1746          * cmdlen=0 if we don't want to advance cmd.
1747          * We break the outer loop by setting done=1
1748          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1749          * as needed.
1750          */
1751         for (; f_pos < chain->n_rules; f_pos++) {
1752                 ipfw_insn *cmd;
1753                 uint32_t tablearg = 0;
1754                 int l, cmdlen, skip_or; /* skip rest of OR block */
1755                 struct ip_fw *f;
1756
1757                 f = chain->map[f_pos];
1758                 if (V_set_disable & (1 << f->set) )
1759                         continue;
1760
1761                 skip_or = 0;
1762                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1763                     l -= cmdlen, cmd += cmdlen) {
1764                         int match;
1765
1766                         /*
1767                          * check_body is a jump target used when we find a
1768                          * CHECK_STATE, and need to jump to the body of
1769                          * the target rule.
1770                          */
1771
1772 /* check_body: */
1773                         cmdlen = F_LEN(cmd);
1774                         /*
1775                          * An OR block (insn_1 || .. || insn_n) has the
1776                          * F_OR bit set in all but the last instruction.
1777                          * The first match will set "skip_or", and cause
1778                          * the following instructions to be skipped until
1779                          * past the one with the F_OR bit clear.
1780                          */
1781                         if (skip_or) {          /* skip this instruction */
1782                                 if ((cmd->len & F_OR) == 0)
1783                                         skip_or = 0;    /* next one is good */
1784                                 continue;
1785                         }
1786                         match = 0; /* set to 1 if we succeed */
1787
1788                         switch (cmd->opcode) {
1789                         /*
1790                          * The first set of opcodes compares the packet's
1791                          * fields with some pattern, setting 'match' if a
1792                          * match is found. At the end of the loop there is
1793                          * logic to deal with F_NOT and F_OR flags associated
1794                          * with the opcode.
1795                          */
1796                         case O_NOP:
1797                                 match = 1;
1798                                 break;
1799
1800                         case O_FORWARD_MAC:
1801                                 printf("ipfw: opcode %d unimplemented\n",
1802                                     cmd->opcode);
1803                                 break;
1804
1805                         case O_GID:
1806                         case O_UID:
1807                         case O_JAIL:
1808                                 /*
1809                                  * We only check offset == 0 && proto != 0,
1810                                  * as this ensures that we have a
1811                                  * packet with the ports info.
1812                                  */
1813                                 if (offset != 0)
1814                                         break;
1815                                 if (proto == IPPROTO_TCP ||
1816                                     proto == IPPROTO_UDP ||
1817                                     proto == IPPROTO_UDPLITE)
1818                                         match = check_uidgid(
1819                                                     (ipfw_insn_u32 *)cmd,
1820                                                     args, &ucred_lookup,
1821 #ifdef __FreeBSD__
1822                                                     &ucred_cache);
1823 #else
1824                                                     (void *)&ucred_cache);
1825 #endif
1826                                 break;
1827
1828                         case O_RECV:
1829                                 match = iface_match(m->m_pkthdr.rcvif,
1830                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1831                                 break;
1832
1833                         case O_XMIT:
1834                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1835                                     chain, &tablearg);
1836                                 break;
1837
1838                         case O_VIA:
1839                                 match = iface_match(oif ? oif :
1840                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1841                                     chain, &tablearg);
1842                                 break;
1843
1844                         case O_MACADDR2:
1845                                 if (args->flags & IPFW_ARGS_ETHER) {
1846                                         u_int32_t *want = (u_int32_t *)
1847                                                 ((ipfw_insn_mac *)cmd)->addr;
1848                                         u_int32_t *mask = (u_int32_t *)
1849                                                 ((ipfw_insn_mac *)cmd)->mask;
1850                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1851
1852                                         match =
1853                                             ( want[0] == (hdr[0] & mask[0]) &&
1854                                               want[1] == (hdr[1] & mask[1]) &&
1855                                               want[2] == (hdr[2] & mask[2]) );
1856                                 }
1857                                 break;
1858
1859                         case O_MAC_TYPE:
1860                                 if (args->flags & IPFW_ARGS_ETHER) {
1861                                         u_int16_t *p =
1862                                             ((ipfw_insn_u16 *)cmd)->ports;
1863                                         int i;
1864
1865                                         for (i = cmdlen - 1; !match && i>0;
1866                                             i--, p += 2)
1867                                                 match = (etype >= p[0] &&
1868                                                     etype <= p[1]);
1869                                 }
1870                                 break;
1871
1872                         case O_FRAG:
1873                                 match = (offset != 0);
1874                                 break;
1875
1876                         case O_IN:      /* "out" is "not in" */
1877                                 match = (oif == NULL);
1878                                 break;
1879
1880                         case O_LAYER2:
1881                                 match = (args->flags & IPFW_ARGS_ETHER);
1882                                 break;
1883
1884                         case O_DIVERTED:
1885                                 if ((args->flags & IPFW_ARGS_REF) == 0)
1886                                         break;
1887                                 /*
1888                                  * For diverted packets, args->rule.info
1889                                  * contains the divert port (in host format)
1890                                  * reason and direction.
1891                                  */
1892                                 match = ((args->rule.info & IPFW_IS_MASK) ==
1893                                     IPFW_IS_DIVERT) && (
1894                                     ((args->rule.info & IPFW_INFO_IN) ?
1895                                         1: 2) & cmd->arg1);
1896                                 break;
1897
1898                         case O_PROTO:
1899                                 /*
1900                                  * We do not allow an arg of 0 so the
1901                                  * check of "proto" only suffices.
1902                                  */
1903                                 match = (proto == cmd->arg1);
1904                                 break;
1905
1906                         case O_IP_SRC:
1907                                 match = is_ipv4 &&
1908                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1909                                     src_ip.s_addr);
1910                                 break;
1911
1912                         case O_IP_DST_LOOKUP:
1913                         {
1914                                 void *pkey;
1915                                 uint32_t vidx, key;
1916                                 uint16_t keylen;
1917
1918                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1919                                         /* Determine lookup key type */
1920                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1921                                         if (vidx != 4 /* uid */ &&
1922                                             vidx != 5 /* jail */ &&
1923                                             is_ipv6 == 0 && is_ipv4 == 0)
1924                                                 break;
1925                                         /* Determine key length */
1926                                         if (vidx == 0 /* dst-ip */ ||
1927                                             vidx == 1 /* src-ip */)
1928                                                 keylen = is_ipv6 ?
1929                                                     sizeof(struct in6_addr):
1930                                                     sizeof(in_addr_t);
1931                                         else {
1932                                                 keylen = sizeof(key);
1933                                                 pkey = &key;
1934                                         }
1935                                         if (vidx == 0 /* dst-ip */)
1936                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1937                                                     (void *)&args->f_id.dst_ip6;
1938                                         else if (vidx == 1 /* src-ip */)
1939                                                 pkey = is_ipv4 ? (void *)&src_ip:
1940                                                     (void *)&args->f_id.src_ip6;
1941                                         else if (vidx == 6 /* dscp */) {
1942                                                 if (is_ipv4)
1943                                                         key = ip->ip_tos >> 2;
1944                                                 else {
1945                                                         key = args->f_id.flow_id6;
1946                                                         key = (key & 0x0f) << 2 |
1947                                                             (key & 0xf000) >> 14;
1948                                                 }
1949                                                 key &= 0x3f;
1950                                         } else if (vidx == 2 /* dst-port */ ||
1951                                             vidx == 3 /* src-port */) {
1952                                                 /* Skip fragments */
1953                                                 if (offset != 0)
1954                                                         break;
1955                                                 /* Skip proto without ports */
1956                                                 if (proto != IPPROTO_TCP &&
1957                                                     proto != IPPROTO_UDP &&
1958                                                     proto != IPPROTO_UDPLITE &&
1959                                                     proto != IPPROTO_SCTP)
1960                                                         break;
1961                                                 if (vidx == 2 /* dst-port */)
1962                                                         key = dst_port;
1963                                                 else
1964                                                         key = src_port;
1965                                         }
1966 #ifndef USERSPACE
1967                                         else if (vidx == 4 /* uid */ ||
1968                                             vidx == 5 /* jail */) {
1969                                                 check_uidgid(
1970                                                     (ipfw_insn_u32 *)cmd,
1971                                                     args, &ucred_lookup,
1972 #ifdef __FreeBSD__
1973                                                     &ucred_cache);
1974                                                 if (vidx == 4 /* uid */)
1975                                                         key = ucred_cache->cr_uid;
1976                                                 else if (vidx == 5 /* jail */)
1977                                                         key = ucred_cache->cr_prison->pr_id;
1978 #else /* !__FreeBSD__ */
1979                                                     (void *)&ucred_cache);
1980                                                 if (vidx == 4 /* uid */)
1981                                                         key = ucred_cache.uid;
1982                                                 else if (vidx == 5 /* jail */)
1983                                                         key = ucred_cache.xid;
1984 #endif /* !__FreeBSD__ */
1985                                         }
1986 #endif /* !USERSPACE */
1987                                         else
1988                                                 break;
1989                                         match = ipfw_lookup_table(chain,
1990                                             cmd->arg1, keylen, pkey, &vidx);
1991                                         if (!match)
1992                                                 break;
1993                                         tablearg = vidx;
1994                                         break;
1995                                 }
1996                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1997                                 /* FALLTHROUGH */
1998                         }
1999                         case O_IP_SRC_LOOKUP:
2000                         {
2001                                 void *pkey;
2002                                 uint32_t vidx;
2003                                 uint16_t keylen;
2004
2005                                 if (is_ipv4) {
2006                                         keylen = sizeof(in_addr_t);
2007                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2008                                                 pkey = &dst_ip;
2009                                         else
2010                                                 pkey = &src_ip;
2011                                 } else if (is_ipv6) {
2012                                         keylen = sizeof(struct in6_addr);
2013                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2014                                                 pkey = &args->f_id.dst_ip6;
2015                                         else
2016                                                 pkey = &args->f_id.src_ip6;
2017                                 } else
2018                                         break;
2019                                 match = ipfw_lookup_table(chain, cmd->arg1,
2020                                     keylen, pkey, &vidx);
2021                                 if (!match)
2022                                         break;
2023                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2024                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2025                                             TARG_VAL(chain, vidx, tag);
2026                                         if (!match)
2027                                                 break;
2028                                 }
2029                                 tablearg = vidx;
2030                                 break;
2031                         }
2032
2033                         case O_IP_FLOW_LOOKUP:
2034                                 {
2035                                         uint32_t v = 0;
2036                                         match = ipfw_lookup_table(chain,
2037                                             cmd->arg1, 0, &args->f_id, &v);
2038                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2039                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2040                                                     TARG_VAL(chain, v, tag);
2041                                         if (match)
2042                                                 tablearg = v;
2043                                 }
2044                                 break;
2045                         case O_IP_SRC_MASK:
2046                         case O_IP_DST_MASK:
2047                                 if (is_ipv4) {
2048                                     uint32_t a =
2049                                         (cmd->opcode == O_IP_DST_MASK) ?
2050                                             dst_ip.s_addr : src_ip.s_addr;
2051                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2052                                     int i = cmdlen-1;
2053
2054                                     for (; !match && i>0; i-= 2, p+= 2)
2055                                         match = (p[0] == (a & p[1]));
2056                                 }
2057                                 break;
2058
2059                         case O_IP_SRC_ME:
2060                                 if (is_ipv4) {
2061                                         match = in_localip(src_ip);
2062                                         break;
2063                                 }
2064 #ifdef INET6
2065                                 /* FALLTHROUGH */
2066                         case O_IP6_SRC_ME:
2067                                 match = is_ipv6 &&
2068                                     ipfw_localip6(&args->f_id.src_ip6);
2069 #endif
2070                                 break;
2071
2072                         case O_IP_DST_SET:
2073                         case O_IP_SRC_SET:
2074                                 if (is_ipv4) {
2075                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2076                                         u_int32_t addr =
2077                                             cmd->opcode == O_IP_DST_SET ?
2078                                                 args->f_id.dst_ip :
2079                                                 args->f_id.src_ip;
2080
2081                                             if (addr < d[0])
2082                                                     break;
2083                                             addr -= d[0]; /* subtract base */
2084                                             match = (addr < cmd->arg1) &&
2085                                                 ( d[ 1 + (addr>>5)] &
2086                                                   (1<<(addr & 0x1f)) );
2087                                 }
2088                                 break;
2089
2090                         case O_IP_DST:
2091                                 match = is_ipv4 &&
2092                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2093                                     dst_ip.s_addr);
2094                                 break;
2095
2096                         case O_IP_DST_ME:
2097                                 if (is_ipv4) {
2098                                         match = in_localip(dst_ip);
2099                                         break;
2100                                 }
2101 #ifdef INET6
2102                                 /* FALLTHROUGH */
2103                         case O_IP6_DST_ME:
2104                                 match = is_ipv6 &&
2105                                     ipfw_localip6(&args->f_id.dst_ip6);
2106 #endif
2107                                 break;
2108
2109
2110                         case O_IP_SRCPORT:
2111                         case O_IP_DSTPORT:
2112                                 /*
2113                                  * offset == 0 && proto != 0 is enough
2114                                  * to guarantee that we have a
2115                                  * packet with port info.
2116                                  */
2117                                 if ((proto == IPPROTO_UDP ||
2118                                     proto == IPPROTO_UDPLITE ||
2119                                     proto == IPPROTO_TCP ||
2120                                     proto == IPPROTO_SCTP) && offset == 0) {
2121                                         u_int16_t x =
2122                                             (cmd->opcode == O_IP_SRCPORT) ?
2123                                                 src_port : dst_port ;
2124                                         u_int16_t *p =
2125                                             ((ipfw_insn_u16 *)cmd)->ports;
2126                                         int i;
2127
2128                                         for (i = cmdlen - 1; !match && i>0;
2129                                             i--, p += 2)
2130                                                 match = (x>=p[0] && x<=p[1]);
2131                                 }
2132                                 break;
2133
2134                         case O_ICMPTYPE:
2135                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2136                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2137                                 break;
2138
2139 #ifdef INET6
2140                         case O_ICMP6TYPE:
2141                                 match = is_ipv6 && offset == 0 &&
2142                                     proto==IPPROTO_ICMPV6 &&
2143                                     icmp6type_match(
2144                                         ICMP6(ulp)->icmp6_type,
2145                                         (ipfw_insn_u32 *)cmd);
2146                                 break;
2147 #endif /* INET6 */
2148
2149                         case O_IPOPT:
2150                                 match = (is_ipv4 &&
2151                                     ipopts_match(ip, cmd) );
2152                                 break;
2153
2154                         case O_IPVER:
2155                                 match = (is_ipv4 &&
2156                                     cmd->arg1 == ip->ip_v);
2157                                 break;
2158
2159                         case O_IPID:
2160                         case O_IPTTL:
2161                                 if (!is_ipv4)
2162                                         break;
2163                         case O_IPLEN:
2164                                 {       /* only for IP packets */
2165                                     uint16_t x;
2166                                     uint16_t *p;
2167                                     int i;
2168
2169                                     if (cmd->opcode == O_IPLEN)
2170                                         x = iplen;
2171                                     else if (cmd->opcode == O_IPTTL)
2172                                         x = ip->ip_ttl;
2173                                     else /* must be IPID */
2174                                         x = ntohs(ip->ip_id);
2175                                     if (cmdlen == 1) {
2176                                         match = (cmd->arg1 == x);
2177                                         break;
2178                                     }
2179                                     /* otherwise we have ranges */
2180                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2181                                     i = cmdlen - 1;
2182                                     for (; !match && i>0; i--, p += 2)
2183                                         match = (x >= p[0] && x <= p[1]);
2184                                 }
2185                                 break;
2186
2187                         case O_IPPRECEDENCE:
2188                                 match = (is_ipv4 &&
2189                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2190                                 break;
2191
2192                         case O_IPTOS:
2193                                 match = (is_ipv4 &&
2194                                     flags_match(cmd, ip->ip_tos));
2195                                 break;
2196
2197                         case O_DSCP:
2198                             {
2199                                 uint32_t *p;
2200                                 uint16_t x;
2201
2202                                 p = ((ipfw_insn_u32 *)cmd)->d;
2203
2204                                 if (is_ipv4)
2205                                         x = ip->ip_tos >> 2;
2206                                 else if (is_ipv6) {
2207                                         uint8_t *v;
2208                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2209                                         x = (*v & 0x0F) << 2;
2210                                         v++;
2211                                         x |= *v >> 6;
2212                                 } else
2213                                         break;
2214
2215                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2216                                 if (x >= 32)
2217                                         match = *(p + 1) & (1 << (x - 32));
2218                                 else
2219                                         match = *p & (1 << x);
2220                             }
2221                                 break;
2222
2223                         case O_TCPDATALEN:
2224                                 if (proto == IPPROTO_TCP && offset == 0) {
2225                                     struct tcphdr *tcp;
2226                                     uint16_t x;
2227                                     uint16_t *p;
2228                                     int i;
2229 #ifdef INET6
2230                                     if (is_ipv6) {
2231                                             struct ip6_hdr *ip6;
2232
2233                                             ip6 = (struct ip6_hdr *)ip;
2234                                             if (ip6->ip6_plen == 0) {
2235                                                     /*
2236                                                      * Jumbo payload is not
2237                                                      * supported by this
2238                                                      * opcode.
2239                                                      */
2240                                                     break;
2241                                             }
2242                                             x = iplen - hlen;
2243                                     } else
2244 #endif /* INET6 */
2245                                             x = iplen - (ip->ip_hl << 2);
2246                                     tcp = TCP(ulp);
2247                                     x -= tcp->th_off << 2;
2248                                     if (cmdlen == 1) {
2249                                         match = (cmd->arg1 == x);
2250                                         break;
2251                                     }
2252                                     /* otherwise we have ranges */
2253                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2254                                     i = cmdlen - 1;
2255                                     for (; !match && i>0; i--, p += 2)
2256                                         match = (x >= p[0] && x <= p[1]);
2257                                 }
2258                                 break;
2259
2260                         case O_TCPFLAGS:
2261                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2262                                     flags_match(cmd, TCP(ulp)->th_flags));
2263                                 break;
2264
2265                         case O_TCPOPTS:
2266                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2267                                         PULLUP_LEN(hlen, ulp,
2268                                             (TCP(ulp)->th_off << 2));
2269                                         match = tcpopts_match(TCP(ulp), cmd);
2270                                 }
2271                                 break;
2272
2273                         case O_TCPSEQ:
2274                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2275                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2276                                         TCP(ulp)->th_seq);
2277                                 break;
2278
2279                         case O_TCPACK:
2280                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2281                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2282                                         TCP(ulp)->th_ack);
2283                                 break;
2284
2285                         case O_TCPMSS:
2286                                 if (proto == IPPROTO_TCP &&
2287                                     (args->f_id._flags & TH_SYN) != 0 &&
2288                                     ulp != NULL) {
2289                                         uint16_t mss, *p;
2290                                         int i;
2291
2292                                         PULLUP_LEN(hlen, ulp,
2293                                             (TCP(ulp)->th_off << 2));
2294                                         if ((tcpopts_parse(TCP(ulp), &mss) &
2295                                             IP_FW_TCPOPT_MSS) == 0)
2296                                                 break;
2297                                         if (cmdlen == 1) {
2298                                                 match = (cmd->arg1 == mss);
2299                                                 break;
2300                                         }
2301                                         /* Otherwise we have ranges. */
2302                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2303                                         i = cmdlen - 1;
2304                                         for (; !match && i > 0; i--, p += 2)
2305                                                 match = (mss >= p[0] &&
2306                                                     mss <= p[1]);
2307                                 }
2308                                 break;
2309
2310                         case O_TCPWIN:
2311                                 if (proto == IPPROTO_TCP && offset == 0) {
2312                                     uint16_t x;
2313                                     uint16_t *p;
2314                                     int i;
2315
2316                                     x = ntohs(TCP(ulp)->th_win);
2317                                     if (cmdlen == 1) {
2318                                         match = (cmd->arg1 == x);
2319                                         break;
2320                                     }
2321                                     /* Otherwise we have ranges. */
2322                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2323                                     i = cmdlen - 1;
2324                                     for (; !match && i > 0; i--, p += 2)
2325                                         match = (x >= p[0] && x <= p[1]);
2326                                 }
2327                                 break;
2328
2329                         case O_ESTAB:
2330                                 /* reject packets which have SYN only */
2331                                 /* XXX should i also check for TH_ACK ? */
2332                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2333                                     (TCP(ulp)->th_flags &
2334                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2335                                 break;
2336
2337                         case O_ALTQ: {
2338                                 struct pf_mtag *at;
2339                                 struct m_tag *mtag;
2340                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2341
2342                                 /*
2343                                  * ALTQ uses mbuf tags from another
2344                                  * packet filtering system - pf(4).
2345                                  * We allocate a tag in its format
2346                                  * and fill it in, pretending to be pf(4).
2347                                  */
2348                                 match = 1;
2349                                 at = pf_find_mtag(m);
2350                                 if (at != NULL && at->qid != 0)
2351                                         break;
2352                                 mtag = m_tag_get(PACKET_TAG_PF,
2353                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2354                                 if (mtag == NULL) {
2355                                         /*
2356                                          * Let the packet fall back to the
2357                                          * default ALTQ.
2358                                          */
2359                                         break;
2360                                 }
2361                                 m_tag_prepend(m, mtag);
2362                                 at = (struct pf_mtag *)(mtag + 1);
2363                                 at->qid = altq->qid;
2364                                 at->hdr = ip;
2365                                 break;
2366                         }
2367
2368                         case O_LOG:
2369                                 ipfw_log(chain, f, hlen, args, m,
2370                                     oif, offset | ip6f_mf, tablearg, ip);
2371                                 match = 1;
2372                                 break;
2373
2374                         case O_PROB:
2375                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2376                                 break;
2377
2378                         case O_VERREVPATH:
2379                                 /* Outgoing packets automatically pass/match */
2380                                 match = ((oif != NULL) ||
2381                                     (m->m_pkthdr.rcvif == NULL) ||
2382                                     (
2383 #ifdef INET6
2384                                     is_ipv6 ?
2385                                         verify_path6(&(args->f_id.src_ip6),
2386                                             m->m_pkthdr.rcvif, args->f_id.fib) :
2387 #endif
2388                                     verify_path(src_ip, m->m_pkthdr.rcvif,
2389                                         args->f_id.fib)));
2390                                 break;
2391
2392                         case O_VERSRCREACH:
2393                                 /* Outgoing packets automatically pass/match */
2394                                 match = (hlen > 0 && ((oif != NULL) || (
2395 #ifdef INET6
2396                                     is_ipv6 ?
2397                                         verify_path6(&(args->f_id.src_ip6),
2398                                             NULL, args->f_id.fib) :
2399 #endif
2400                                     verify_path(src_ip, NULL, args->f_id.fib))));
2401                                 break;
2402
2403                         case O_ANTISPOOF:
2404                                 /* Outgoing packets automatically pass/match */
2405                                 if (oif == NULL && hlen > 0 &&
2406                                     (  (is_ipv4 && in_localaddr(src_ip))
2407 #ifdef INET6
2408                                     || (is_ipv6 &&
2409                                         in6_localaddr(&(args->f_id.src_ip6)))
2410 #endif
2411                                     ))
2412                                         match =
2413 #ifdef INET6
2414                                             is_ipv6 ? verify_path6(
2415                                                 &(args->f_id.src_ip6),
2416                                                 m->m_pkthdr.rcvif,
2417                                                 args->f_id.fib) :
2418 #endif
2419                                             verify_path(src_ip,
2420                                                 m->m_pkthdr.rcvif,
2421                                                 args->f_id.fib);
2422                                 else
2423                                         match = 1;
2424                                 break;
2425
2426                         case O_IPSEC:
2427                                 match = (m_tag_find(m,
2428                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2429                                 /* otherwise no match */
2430                                 break;
2431
2432 #ifdef INET6
2433                         case O_IP6_SRC:
2434                                 match = is_ipv6 &&
2435                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2436                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2437                                 break;
2438
2439                         case O_IP6_DST:
2440                                 match = is_ipv6 &&
2441                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2442                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2443                                 break;
2444                         case O_IP6_SRC_MASK:
2445                         case O_IP6_DST_MASK:
2446                                 if (is_ipv6) {
2447                                         int i = cmdlen - 1;
2448                                         struct in6_addr p;
2449                                         struct in6_addr *d =
2450                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2451
2452                                         for (; !match && i > 0; d += 2,
2453                                             i -= F_INSN_SIZE(struct in6_addr)
2454                                             * 2) {
2455                                                 p = (cmd->opcode ==
2456                                                     O_IP6_SRC_MASK) ?
2457                                                     args->f_id.src_ip6:
2458                                                     args->f_id.dst_ip6;
2459                                                 APPLY_MASK(&p, &d[1]);
2460                                                 match =
2461                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2462                                                     &p);
2463                                         }
2464                                 }
2465                                 break;
2466
2467                         case O_FLOW6ID:
2468                                 match = is_ipv6 &&
2469                                     flow6id_match(args->f_id.flow_id6,
2470                                     (ipfw_insn_u32 *) cmd);
2471                                 break;
2472
2473                         case O_EXT_HDR:
2474                                 match = is_ipv6 &&
2475                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2476                                 break;
2477
2478                         case O_IP6:
2479                                 match = is_ipv6;
2480                                 break;
2481 #endif
2482
2483                         case O_IP4:
2484                                 match = is_ipv4;
2485                                 break;
2486
2487                         case O_TAG: {
2488                                 struct m_tag *mtag;
2489                                 uint32_t tag = TARG(cmd->arg1, tag);
2490
2491                                 /* Packet is already tagged with this tag? */
2492                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2493
2494                                 /* We have `untag' action when F_NOT flag is
2495                                  * present. And we must remove this mtag from
2496                                  * mbuf and reset `match' to zero (`match' will
2497                                  * be inversed later).
2498                                  * Otherwise we should allocate new mtag and
2499                                  * push it into mbuf.
2500                                  */
2501                                 if (cmd->len & F_NOT) { /* `untag' action */
2502                                         if (mtag != NULL)
2503                                                 m_tag_delete(m, mtag);
2504                                         match = 0;
2505                                 } else {
2506                                         if (mtag == NULL) {
2507                                                 mtag = m_tag_alloc( MTAG_IPFW,
2508                                                     tag, 0, M_NOWAIT);
2509                                                 if (mtag != NULL)
2510                                                         m_tag_prepend(m, mtag);
2511                                         }
2512                                         match = 1;
2513                                 }
2514                                 break;
2515                         }
2516
2517                         case O_FIB: /* try match the specified fib */
2518                                 if (args->f_id.fib == cmd->arg1)
2519                                         match = 1;
2520                                 break;
2521
2522                         case O_SOCKARG: {
2523 #ifndef USERSPACE       /* not supported in userspace */
2524                                 struct inpcb *inp = args->inp;
2525                                 struct inpcbinfo *pi;
2526
2527                                 if (is_ipv6) /* XXX can we remove this ? */
2528                                         break;
2529
2530                                 if (proto == IPPROTO_TCP)
2531                                         pi = &V_tcbinfo;
2532                                 else if (proto == IPPROTO_UDP)
2533                                         pi = &V_udbinfo;
2534                                 else if (proto == IPPROTO_UDPLITE)
2535                                         pi = &V_ulitecbinfo;
2536                                 else
2537                                         break;
2538
2539                                 /*
2540                                  * XXXRW: so_user_cookie should almost
2541                                  * certainly be inp_user_cookie?
2542                                  */
2543
2544                                 /* For incoming packet, lookup up the
2545                                 inpcb using the src/dest ip/port tuple */
2546                                 if (inp == NULL) {
2547                                         inp = in_pcblookup(pi,
2548                                                 src_ip, htons(src_port),
2549                                                 dst_ip, htons(dst_port),
2550                                                 INPLOOKUP_RLOCKPCB, NULL);
2551                                         if (inp != NULL) {
2552                                                 tablearg =
2553                                                     inp->inp_socket->so_user_cookie;
2554                                                 if (tablearg)
2555                                                         match = 1;
2556                                                 INP_RUNLOCK(inp);
2557                                         }
2558                                 } else {
2559                                         if (inp->inp_socket) {
2560                                                 tablearg =
2561                                                     inp->inp_socket->so_user_cookie;
2562                                                 if (tablearg)
2563                                                         match = 1;
2564                                         }
2565                                 }
2566 #endif /* !USERSPACE */
2567                                 break;
2568                         }
2569
2570                         case O_TAGGED: {
2571                                 struct m_tag *mtag;
2572                                 uint32_t tag = TARG(cmd->arg1, tag);
2573
2574                                 if (cmdlen == 1) {
2575                                         match = m_tag_locate(m, MTAG_IPFW,
2576                                             tag, NULL) != NULL;
2577                                         break;
2578                                 }
2579
2580                                 /* we have ranges */
2581                                 for (mtag = m_tag_first(m);
2582                                     mtag != NULL && !match;
2583                                     mtag = m_tag_next(m, mtag)) {
2584                                         uint16_t *p;
2585                                         int i;
2586
2587                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2588                                                 continue;
2589
2590                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2591                                         i = cmdlen - 1;
2592                                         for(; !match && i > 0; i--, p += 2)
2593                                                 match =
2594                                                     mtag->m_tag_id >= p[0] &&
2595                                                     mtag->m_tag_id <= p[1];
2596                                 }
2597                                 break;
2598                         }
2599
2600                         /*
2601                          * The second set of opcodes represents 'actions',
2602                          * i.e. the terminal part of a rule once the packet
2603                          * matches all previous patterns.
2604                          * Typically there is only one action for each rule,
2605                          * and the opcode is stored at the end of the rule
2606                          * (but there are exceptions -- see below).
2607                          *
2608                          * In general, here we set retval and terminate the
2609                          * outer loop (would be a 'break 3' in some language,
2610                          * but we need to set l=0, done=1)
2611                          *
2612                          * Exceptions:
2613                          * O_COUNT and O_SKIPTO actions:
2614                          *   instead of terminating, we jump to the next rule
2615                          *   (setting l=0), or to the SKIPTO target (setting
2616                          *   f/f_len, cmd and l as needed), respectively.
2617                          *
2618                          * O_TAG, O_LOG and O_ALTQ action parameters:
2619                          *   perform some action and set match = 1;
2620                          *
2621                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2622                          *   not real 'actions', and are stored right
2623                          *   before the 'action' part of the rule (one
2624                          *   exception is O_SKIP_ACTION which could be
2625                          *   between these opcodes and 'action' one).
2626                          *   These opcodes try to install an entry in the
2627                          *   state tables; if successful, we continue with
2628                          *   the next opcode (match=1; break;), otherwise
2629                          *   the packet must be dropped (set retval,
2630                          *   break loops with l=0, done=1)
2631                          *
2632                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2633                          *   cause a lookup of the state table, and a jump
2634                          *   to the 'action' part of the parent rule
2635                          *   if an entry is found, or
2636                          *   (CHECK_STATE only) a jump to the next rule if
2637                          *   the entry is not found.
2638                          *   The result of the lookup is cached so that
2639                          *   further instances of these opcodes become NOPs.
2640                          *   The jump to the next rule is done by setting
2641                          *   l=0, cmdlen=0.
2642                          *
2643                          * O_SKIP_ACTION: this opcode is not a real 'action'
2644                          *  either, and is stored right before the 'action'
2645                          *  part of the rule, right after the O_KEEP_STATE
2646                          *  opcode. It causes match failure so the real
2647                          *  'action' could be executed only if the rule
2648                          *  is checked via dynamic rule from the state
2649                          *  table, as in such case execution starts
2650                          *  from the true 'action' opcode directly.
2651                          *
2652                          */
2653                         case O_LIMIT:
2654                         case O_KEEP_STATE:
2655                                 if (ipfw_dyn_install_state(chain, f,
2656                                     (ipfw_insn_limit *)cmd, args, ulp,
2657                                     pktlen, &dyn_info, tablearg)) {
2658                                         /* error or limit violation */
2659                                         retval = IP_FW_DENY;
2660                                         l = 0;  /* exit inner loop */
2661                                         done = 1; /* exit outer loop */
2662                                 }
2663                                 match = 1;
2664                                 break;
2665
2666                         case O_PROBE_STATE:
2667                         case O_CHECK_STATE:
2668                                 /*
2669                                  * dynamic rules are checked at the first
2670                                  * keep-state or check-state occurrence,
2671                                  * with the result being stored in dyn_info.
2672                                  * The compiler introduces a PROBE_STATE
2673                                  * instruction for us when we have a
2674                                  * KEEP_STATE (because PROBE_STATE needs
2675                                  * to be run first).
2676                                  */
2677                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2678                                     (q = ipfw_dyn_lookup_state(args, ulp,
2679                                     pktlen, cmd, &dyn_info)) != NULL) {
2680                                         /*
2681                                          * Found dynamic entry, jump to the
2682                                          * 'action' part of the parent rule
2683                                          * by setting f, cmd, l and clearing
2684                                          * cmdlen.
2685                                          */
2686                                         f = q;
2687                                         f_pos = dyn_info.f_pos;
2688                                         cmd = ACTION_PTR(f);
2689                                         l = f->cmd_len - f->act_ofs;
2690                                         cmdlen = 0;
2691                                         match = 1;
2692                                         break;
2693                                 }
2694                                 /*
2695                                  * Dynamic entry not found. If CHECK_STATE,
2696                                  * skip to next rule, if PROBE_STATE just
2697                                  * ignore and continue with next opcode.
2698                                  */
2699                                 if (cmd->opcode == O_CHECK_STATE)
2700                                         l = 0;  /* exit inner loop */
2701                                 match = 1;
2702                                 break;
2703
2704                         case O_SKIP_ACTION:
2705                                 match = 0;      /* skip to the next rule */
2706                                 l = 0;          /* exit inner loop */
2707                                 break;
2708
2709                         case O_ACCEPT:
2710                                 retval = 0;     /* accept */
2711                                 l = 0;          /* exit inner loop */
2712                                 done = 1;       /* exit outer loop */
2713                                 break;
2714
2715                         case O_PIPE:
2716                         case O_QUEUE:
2717                                 set_match(args, f_pos, chain);
2718                                 args->rule.info = TARG(cmd->arg1, pipe);
2719                                 if (cmd->opcode == O_PIPE)
2720                                         args->rule.info |= IPFW_IS_PIPE;
2721                                 if (V_fw_one_pass)
2722                                         args->rule.info |= IPFW_ONEPASS;
2723                                 retval = IP_FW_DUMMYNET;
2724                                 l = 0;          /* exit inner loop */
2725                                 done = 1;       /* exit outer loop */
2726                                 break;
2727
2728                         case O_DIVERT:
2729                         case O_TEE:
2730                                 if (args->flags & IPFW_ARGS_ETHER)
2731                                         break;  /* not on layer 2 */
2732                                 /* otherwise this is terminal */
2733                                 l = 0;          /* exit inner loop */
2734                                 done = 1;       /* exit outer loop */
2735                                 retval = (cmd->opcode == O_DIVERT) ?
2736                                         IP_FW_DIVERT : IP_FW_TEE;
2737                                 set_match(args, f_pos, chain);
2738                                 args->rule.info = TARG(cmd->arg1, divert);
2739                                 break;
2740
2741                         case O_COUNT:
2742                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2743                                 l = 0;          /* exit inner loop */
2744                                 break;
2745
2746                         case O_SKIPTO:
2747                             IPFW_INC_RULE_COUNTER(f, pktlen);
2748                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2749                             /*
2750                              * Skip disabled rules, and re-enter
2751                              * the inner loop with the correct
2752                              * f_pos, f, l and cmd.
2753                              * Also clear cmdlen and skip_or
2754                              */
2755                             for (; f_pos < chain->n_rules - 1 &&
2756                                     (V_set_disable &
2757                                      (1 << chain->map[f_pos]->set));
2758                                     f_pos++)
2759                                 ;
2760                             /* Re-enter the inner loop at the skipto rule. */
2761                             f = chain->map[f_pos];
2762                             l = f->cmd_len;
2763                             cmd = f->cmd;
2764                             match = 1;
2765                             cmdlen = 0;
2766                             skip_or = 0;
2767                             continue;
2768                             break;      /* not reached */
2769
2770                         case O_CALLRETURN: {
2771                                 /*
2772                                  * Implementation of `subroutine' call/return,
2773                                  * in the stack carried in an mbuf tag. This
2774                                  * is different from `skipto' in that any call
2775                                  * address is possible (`skipto' must prevent
2776                                  * backward jumps to avoid endless loops).
2777                                  * We have `return' action when F_NOT flag is
2778                                  * present. The `m_tag_id' field is used as
2779                                  * stack pointer.
2780                                  */
2781                                 struct m_tag *mtag;
2782                                 uint16_t jmpto, *stack;
2783
2784 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2785 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2786                                 /*
2787                                  * Hand-rolled version of m_tag_locate() with
2788                                  * wildcard `type'.
2789                                  * If not already tagged, allocate new tag.
2790                                  */
2791                                 mtag = m_tag_first(m);
2792                                 while (mtag != NULL) {
2793                                         if (mtag->m_tag_cookie ==
2794                                             MTAG_IPFW_CALL)
2795                                                 break;
2796                                         mtag = m_tag_next(m, mtag);
2797                                 }
2798                                 if (mtag == NULL && IS_CALL) {
2799                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2800                                             IPFW_CALLSTACK_SIZE *
2801                                             sizeof(uint16_t), M_NOWAIT);
2802                                         if (mtag != NULL)
2803                                                 m_tag_prepend(m, mtag);
2804                                 }
2805
2806                                 /*
2807                                  * On error both `call' and `return' just
2808                                  * continue with next rule.
2809                                  */
2810                                 if (IS_RETURN && (mtag == NULL ||
2811                                     mtag->m_tag_id == 0)) {
2812                                         l = 0;          /* exit inner loop */
2813                                         break;
2814                                 }
2815                                 if (IS_CALL && (mtag == NULL ||
2816                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2817                                         printf("ipfw: call stack error, "
2818                                             "go to next rule\n");
2819                                         l = 0;          /* exit inner loop */
2820                                         break;
2821                                 }
2822
2823                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2824                                 stack = (uint16_t *)(mtag + 1);
2825
2826                                 /*
2827                                  * The `call' action may use cached f_pos
2828                                  * (in f->next_rule), whose version is written
2829                                  * in f->next_rule.
2830                                  * The `return' action, however, doesn't have
2831                                  * fixed jump address in cmd->arg1 and can't use
2832                                  * cache.
2833                                  */
2834                                 if (IS_CALL) {
2835                                         stack[mtag->m_tag_id] = f->rulenum;
2836                                         mtag->m_tag_id++;
2837                                         f_pos = JUMP(chain, f, cmd->arg1,
2838                                             tablearg, 1);
2839                                 } else {        /* `return' action */
2840                                         mtag->m_tag_id--;
2841                                         jmpto = stack[mtag->m_tag_id] + 1;
2842                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2843                                 }
2844
2845                                 /*
2846                                  * Skip disabled rules, and re-enter
2847                                  * the inner loop with the correct
2848                                  * f_pos, f, l and cmd.
2849                                  * Also clear cmdlen and skip_or
2850                                  */
2851                                 for (; f_pos < chain->n_rules - 1 &&
2852                                     (V_set_disable &
2853                                     (1 << chain->map[f_pos]->set)); f_pos++)
2854                                         ;
2855                                 /* Re-enter the inner loop at the dest rule. */
2856                                 f = chain->map[f_pos];
2857                                 l = f->cmd_len;
2858                                 cmd = f->cmd;
2859                                 cmdlen = 0;
2860                                 skip_or = 0;
2861                                 continue;
2862                                 break;  /* NOTREACHED */
2863                         }
2864 #undef IS_CALL
2865 #undef IS_RETURN
2866
2867                         case O_REJECT:
2868                                 /*
2869                                  * Drop the packet and send a reject notice
2870                                  * if the packet is not ICMP (or is an ICMP
2871                                  * query), and it is not multicast/broadcast.
2872                                  */
2873                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2874                                     (proto != IPPROTO_ICMP ||
2875                                      is_icmp_query(ICMP(ulp))) &&
2876                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2877                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2878                                         send_reject(args, cmd->arg1, iplen, ip);
2879                                         m = args->m;
2880                                 }
2881                                 /* FALLTHROUGH */
2882 #ifdef INET6
2883                         case O_UNREACH6:
2884                                 if (hlen > 0 && is_ipv6 &&
2885                                     ((offset & IP6F_OFF_MASK) == 0) &&
2886                                     (proto != IPPROTO_ICMPV6 ||
2887                                      (is_icmp6_query(icmp6_type) == 1)) &&
2888                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2889                                     !IN6_IS_ADDR_MULTICAST(
2890                                         &args->f_id.dst_ip6)) {
2891                                         send_reject6(args,
2892                                             cmd->opcode == O_REJECT ?
2893                                             map_icmp_unreach(cmd->arg1):
2894                                             cmd->arg1, hlen,
2895                                             (struct ip6_hdr *)ip);
2896                                         m = args->m;
2897                                 }
2898                                 /* FALLTHROUGH */
2899 #endif
2900                         case O_DENY:
2901                                 retval = IP_FW_DENY;
2902                                 l = 0;          /* exit inner loop */
2903                                 done = 1;       /* exit outer loop */
2904                                 break;
2905
2906                         case O_FORWARD_IP:
2907                                 if (args->flags & IPFW_ARGS_ETHER)
2908                                         break;  /* not valid on layer2 pkts */
2909                                 if (q != f ||
2910                                     dyn_info.direction == MATCH_FORWARD) {
2911                                     struct sockaddr_in *sa;
2912
2913                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2914                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2915 #ifdef INET6
2916                                         /*
2917                                          * We use O_FORWARD_IP opcode for
2918                                          * fwd rule with tablearg, but tables
2919                                          * now support IPv6 addresses. And
2920                                          * when we are inspecting IPv6 packet,
2921                                          * we can use nh6 field from
2922                                          * table_value as next_hop6 address.
2923                                          */
2924                                         if (is_ipv6) {
2925                                                 struct ip_fw_nh6 *nh6;
2926
2927                                                 args->flags |= IPFW_ARGS_NH6;
2928                                                 nh6 = &args->hopstore6;
2929                                                 nh6->sin6_addr = TARG_VAL(
2930                                                     chain, tablearg, nh6);
2931                                                 nh6->sin6_port = sa->sin_port;
2932                                                 nh6->sin6_scope_id = TARG_VAL(
2933                                                     chain, tablearg, zoneid);
2934                                         } else
2935 #endif
2936                                         {
2937                                                 args->flags |= IPFW_ARGS_NH4;
2938                                                 args->hopstore.sin_port =
2939                                                     sa->sin_port;
2940                                                 sa = &args->hopstore;
2941                                                 sa->sin_family = AF_INET;
2942                                                 sa->sin_len = sizeof(*sa);
2943                                                 sa->sin_addr.s_addr = htonl(
2944                                                     TARG_VAL(chain, tablearg,
2945                                                     nh4));
2946                                         }
2947                                     } else {
2948                                             args->flags |= IPFW_ARGS_NH4PTR;
2949                                             args->next_hop = sa;
2950                                     }
2951                                 }
2952                                 retval = IP_FW_PASS;
2953                                 l = 0;          /* exit inner loop */
2954                                 done = 1;       /* exit outer loop */
2955                                 break;
2956
2957 #ifdef INET6
2958                         case O_FORWARD_IP6:
2959                                 if (args->flags & IPFW_ARGS_ETHER)
2960                                         break;  /* not valid on layer2 pkts */
2961                                 if (q != f ||
2962                                     dyn_info.direction == MATCH_FORWARD) {
2963                                         struct sockaddr_in6 *sin6;
2964
2965                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2966                                         args->flags |= IPFW_ARGS_NH6PTR;
2967                                         args->next_hop6 = sin6;
2968                                 }
2969                                 retval = IP_FW_PASS;
2970                                 l = 0;          /* exit inner loop */
2971                                 done = 1;       /* exit outer loop */
2972                                 break;
2973 #endif
2974
2975                         case O_NETGRAPH:
2976                         case O_NGTEE:
2977                                 set_match(args, f_pos, chain);
2978                                 args->rule.info = TARG(cmd->arg1, netgraph);
2979                                 if (V_fw_one_pass)
2980                                         args->rule.info |= IPFW_ONEPASS;
2981                                 retval = (cmd->opcode == O_NETGRAPH) ?
2982                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2983                                 l = 0;          /* exit inner loop */
2984                                 done = 1;       /* exit outer loop */
2985                                 break;
2986
2987                         case O_SETFIB: {
2988                                 uint32_t fib;
2989
2990                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2991                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2992                                 if (fib >= rt_numfibs)
2993                                         fib = 0;
2994                                 M_SETFIB(m, fib);
2995                                 args->f_id.fib = fib; /* XXX */
2996                                 l = 0;          /* exit inner loop */
2997                                 break;
2998                         }
2999
3000                         case O_SETDSCP: {
3001                                 uint16_t code;
3002
3003                                 code = TARG(cmd->arg1, dscp) & 0x3F;
3004                                 l = 0;          /* exit inner loop */
3005                                 if (is_ipv4) {
3006                                         uint16_t old;
3007
3008                                         old = *(uint16_t *)ip;
3009                                         ip->ip_tos = (code << 2) |
3010                                             (ip->ip_tos & 0x03);
3011                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
3012                                             old, *(uint16_t *)ip);
3013                                 } else if (is_ipv6) {
3014                                         uint8_t *v;
3015
3016                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
3017                                         *v = (*v & 0xF0) | (code >> 2);
3018                                         v++;
3019                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
3020                                 } else
3021                                         break;
3022
3023                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3024                                 break;
3025                         }
3026
3027                         case O_NAT:
3028                                 l = 0;          /* exit inner loop */
3029                                 done = 1;       /* exit outer loop */
3030                                 /*
3031                                  * Ensure that we do not invoke NAT handler for
3032                                  * non IPv4 packets. Libalias expects only IPv4.
3033                                  */
3034                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3035                                     retval = IP_FW_DENY;
3036                                     break;
3037                                 }
3038
3039                                 struct cfg_nat *t;
3040                                 int nat_id;
3041
3042                                 args->rule.info = 0;
3043                                 set_match(args, f_pos, chain);
3044                                 /* Check if this is 'global' nat rule */
3045                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3046                                         retval = ipfw_nat_ptr(args, NULL, m);
3047                                         break;
3048                                 }
3049                                 t = ((ipfw_insn_nat *)cmd)->nat;
3050                                 if (t == NULL) {
3051                                         nat_id = TARG(cmd->arg1, nat);
3052                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3053
3054                                         if (t == NULL) {
3055                                             retval = IP_FW_DENY;
3056                                             break;
3057                                         }
3058                                         if (cmd->arg1 != IP_FW_TARG)
3059                                             ((ipfw_insn_nat *)cmd)->nat = t;
3060                                 }
3061                                 retval = ipfw_nat_ptr(args, t, m);
3062                                 break;
3063
3064                         case O_REASS: {
3065                                 int ip_off;
3066
3067                                 l = 0;  /* in any case exit inner loop */
3068                                 if (is_ipv6) /* IPv6 is not supported yet */
3069                                         break;
3070                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3071                                 ip_off = ntohs(ip->ip_off);
3072
3073                                 /* if not fragmented, go to next rule */
3074                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3075                                     break;
3076
3077                                 args->m = m = ip_reass(m);
3078
3079                                 /*
3080                                  * do IP header checksum fixup.
3081                                  */
3082                                 if (m == NULL) { /* fragment got swallowed */
3083                                     retval = IP_FW_DENY;
3084                                 } else { /* good, packet complete */
3085                                     int hlen;
3086
3087                                     ip = mtod(m, struct ip *);
3088                                     hlen = ip->ip_hl << 2;
3089                                     ip->ip_sum = 0;
3090                                     if (hlen == sizeof(struct ip))
3091                                         ip->ip_sum = in_cksum_hdr(ip);
3092                                     else
3093                                         ip->ip_sum = in_cksum(m, hlen);
3094                                     retval = IP_FW_REASS;
3095                                     args->rule.info = 0;
3096                                     set_match(args, f_pos, chain);
3097                                 }
3098                                 done = 1;       /* exit outer loop */
3099                                 break;
3100                         }
3101                         case O_EXTERNAL_ACTION:
3102                                 l = 0; /* in any case exit inner loop */
3103                                 retval = ipfw_run_eaction(chain, args,
3104                                     cmd, &done);
3105                                 /*
3106                                  * If both @retval and @done are zero,
3107                                  * consider this as rule matching and
3108                                  * update counters.
3109                                  */
3110                                 if (retval == 0 && done == 0) {
3111                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3112                                         /*
3113                                          * Reset the result of the last
3114                                          * dynamic state lookup.
3115                                          * External action can change
3116                                          * @args content, and it may be
3117                                          * used for new state lookup later.
3118                                          */
3119                                         DYN_INFO_INIT(&dyn_info);
3120                                 }
3121                                 break;
3122
3123                         default:
3124                                 panic("-- unknown opcode %d\n", cmd->opcode);
3125                         } /* end of switch() on opcodes */
3126                         /*
3127                          * if we get here with l=0, then match is irrelevant.
3128                          */
3129
3130                         if (cmd->len & F_NOT)
3131                                 match = !match;
3132
3133                         if (match) {
3134                                 if (cmd->len & F_OR)
3135                                         skip_or = 1;
3136                         } else {
3137                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3138                                         break;          /* try next rule    */
3139                         }
3140
3141                 }       /* end of inner loop, scan opcodes */
3142 #undef PULLUP_LEN
3143
3144                 if (done)
3145                         break;
3146
3147 /* next_rule:; */       /* try next rule                */
3148
3149         }               /* end of outer for, scan rules */
3150
3151         if (done) {
3152                 struct ip_fw *rule = chain->map[f_pos];
3153                 /* Update statistics */
3154                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3155         } else {
3156                 retval = IP_FW_DENY;
3157                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3158         }
3159         IPFW_PF_RUNLOCK(chain);
3160 #ifdef __FreeBSD__
3161         if (ucred_cache != NULL)
3162                 crfree(ucred_cache);
3163 #endif
3164         return (retval);
3165
3166 pullup_failed:
3167         if (V_fw_verbose)
3168                 printf("ipfw: pullup failed\n");
3169         return (IP_FW_DENY);
3170 }
3171
3172 /*
3173  * Set maximum number of tables that can be used in given VNET ipfw instance.
3174  */
3175 #ifdef SYSCTL_NODE
3176 static int
3177 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3178 {
3179         int error;
3180         unsigned int ntables;
3181
3182         ntables = V_fw_tables_max;
3183
3184         error = sysctl_handle_int(oidp, &ntables, 0, req);
3185         /* Read operation or some error */
3186         if ((error != 0) || (req->newptr == NULL))
3187                 return (error);
3188
3189         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3190 }
3191
3192 /*
3193  * Switches table namespace between global and per-set.
3194  */
3195 static int
3196 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3197 {
3198         int error;
3199         unsigned int sets;
3200
3201         sets = V_fw_tables_sets;
3202
3203         error = sysctl_handle_int(oidp, &sets, 0, req);
3204         /* Read operation or some error */
3205         if ((error != 0) || (req->newptr == NULL))
3206                 return (error);
3207
3208         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3209 }
3210 #endif
3211
3212 /*
3213  * Module and VNET glue
3214  */
3215
3216 /*
3217  * Stuff that must be initialised only on boot or module load
3218  */
3219 static int
3220 ipfw_init(void)
3221 {
3222         int error = 0;
3223
3224         /*
3225          * Only print out this stuff the first time around,
3226          * when called from the sysinit code.
3227          */
3228         printf("ipfw2 "
3229 #ifdef INET6
3230                 "(+ipv6) "
3231 #endif
3232                 "initialized, divert %s, nat %s, "
3233                 "default to %s, logging ",
3234 #ifdef IPDIVERT
3235                 "enabled",
3236 #else
3237                 "loadable",
3238 #endif
3239 #ifdef IPFIREWALL_NAT
3240                 "enabled",
3241 #else
3242                 "loadable",
3243 #endif
3244                 default_to_accept ? "accept" : "deny");
3245
3246         /*
3247          * Note: V_xxx variables can be accessed here but the vnet specific
3248          * initializer may not have been called yet for the VIMAGE case.
3249          * Tuneables will have been processed. We will print out values for
3250          * the default vnet.
3251          * XXX This should all be rationalized AFTER 8.0
3252          */
3253         if (V_fw_verbose == 0)
3254                 printf("disabled\n");
3255         else if (V_verbose_limit == 0)
3256                 printf("unlimited\n");
3257         else
3258                 printf("limited to %d packets/entry by default\n",
3259                     V_verbose_limit);
3260
3261         /* Check user-supplied table count for validness */
3262         if (default_fw_tables > IPFW_TABLES_MAX)
3263           default_fw_tables = IPFW_TABLES_MAX;
3264
3265         ipfw_init_sopt_handler();
3266         ipfw_init_obj_rewriter();
3267         ipfw_iface_init();
3268         return (error);
3269 }
3270
3271 /*
3272  * Called for the removal of the last instance only on module unload.
3273  */
3274 static void
3275 ipfw_destroy(void)
3276 {
3277
3278         ipfw_iface_destroy();
3279         ipfw_destroy_sopt_handler();
3280         ipfw_destroy_obj_rewriter();
3281         printf("IP firewall unloaded\n");
3282 }
3283
3284 /*
3285  * Stuff that must be initialized for every instance
3286  * (including the first of course).
3287  */
3288 static int
3289 vnet_ipfw_init(const void *unused)
3290 {
3291         int error, first;
3292         struct ip_fw *rule = NULL;
3293         struct ip_fw_chain *chain;
3294
3295         chain = &V_layer3_chain;
3296
3297         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3298
3299         /* First set up some values that are compile time options */
3300         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3301         V_fw_deny_unknown_exthdrs = 1;
3302 #ifdef IPFIREWALL_VERBOSE
3303         V_fw_verbose = 1;
3304 #endif
3305 #ifdef IPFIREWALL_VERBOSE_LIMIT
3306         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3307 #endif
3308 #ifdef IPFIREWALL_NAT
3309         LIST_INIT(&chain->nat);
3310 #endif
3311
3312         /* Init shared services hash table */
3313         ipfw_init_srv(chain);
3314
3315         ipfw_init_counters();
3316         /* Set initial number of tables */
3317         V_fw_tables_max = default_fw_tables;
3318         error = ipfw_init_tables(chain, first);
3319         if (error) {
3320                 printf("ipfw2: setting up tables failed\n");
3321                 free(chain->map, M_IPFW);
3322                 free(rule, M_IPFW);
3323                 return (ENOSPC);
3324         }
3325
3326         IPFW_LOCK_INIT(chain);
3327
3328         /* fill and insert the default rule */
3329         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3330         rule->flags |= IPFW_RULE_NOOPT;
3331         rule->cmd_len = 1;
3332         rule->cmd[0].len = 1;
3333         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3334         chain->default_rule = rule;
3335         ipfw_add_protected_rule(chain, rule, 0);
3336
3337         ipfw_dyn_init(chain);
3338         ipfw_eaction_init(chain, first);
3339 #ifdef LINEAR_SKIPTO
3340         ipfw_init_skipto_cache(chain);
3341 #endif
3342         ipfw_bpf_init(first);
3343
3344         /* First set up some values that are compile time options */
3345         V_ipfw_vnet_ready = 1;          /* Open for business */
3346
3347         /*
3348          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3349          * Even if the latter two fail we still keep the module alive
3350          * because the sockopt and layer2 paths are still useful.
3351          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3352          * so we can ignore the exact return value and just set a flag.
3353          *
3354          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3355          * changes in the underlying (per-vnet) variables trigger
3356          * immediate hook()/unhook() calls.
3357          * In layer2 we have the same behaviour, except that V_ether_ipfw
3358          * is checked on each packet because there are no pfil hooks.
3359          */
3360         V_ip_fw_ctl_ptr = ipfw_ctl3;
3361         error = ipfw_attach_hooks(1);
3362         return (error);
3363 }
3364
3365 /*
3366  * Called for the removal of each instance.
3367  */
3368 static int
3369 vnet_ipfw_uninit(const void *unused)
3370 {
3371         struct ip_fw *reap;
3372         struct ip_fw_chain *chain = &V_layer3_chain;
3373         int i, last;
3374
3375         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3376         /*
3377          * disconnect from ipv4, ipv6, layer2 and sockopt.
3378          * Then grab, release and grab again the WLOCK so we make
3379          * sure the update is propagated and nobody will be in.
3380          */
3381         (void)ipfw_attach_hooks(0 /* detach */);
3382         V_ip_fw_ctl_ptr = NULL;
3383
3384         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3385
3386         IPFW_UH_WLOCK(chain);
3387         IPFW_UH_WUNLOCK(chain);
3388
3389         ipfw_dyn_uninit(0);     /* run the callout_drain */
3390
3391         IPFW_UH_WLOCK(chain);
3392
3393         reap = NULL;
3394         IPFW_WLOCK(chain);
3395         for (i = 0; i < chain->n_rules; i++)
3396                 ipfw_reap_add(chain, &reap, chain->map[i]);
3397         free(chain->map, M_IPFW);
3398 #ifdef LINEAR_SKIPTO
3399         ipfw_destroy_skipto_cache(chain);
3400 #endif
3401         IPFW_WUNLOCK(chain);
3402         IPFW_UH_WUNLOCK(chain);
3403         ipfw_destroy_tables(chain, last);
3404         ipfw_eaction_uninit(chain, last);
3405         if (reap != NULL)
3406                 ipfw_reap_rules(reap);
3407         vnet_ipfw_iface_destroy(chain);
3408         ipfw_destroy_srv(chain);
3409         IPFW_LOCK_DESTROY(chain);
3410         ipfw_dyn_uninit(1);     /* free the remaining parts */
3411         ipfw_destroy_counters();
3412         ipfw_bpf_uninit(last);
3413         return (0);
3414 }
3415
3416 /*
3417  * Module event handler.
3418  * In general we have the choice of handling most of these events by the
3419  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3420  * use the SYSINIT handlers as they are more capable of expressing the
3421  * flow of control during module and vnet operations, so this is just
3422  * a skeleton. Note there is no SYSINIT equivalent of the module
3423  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3424  */
3425 static int
3426 ipfw_modevent(module_t mod, int type, void *unused)
3427 {
3428         int err = 0;
3429
3430         switch (type) {
3431         case MOD_LOAD:
3432                 /* Called once at module load or
3433                  * system boot if compiled in. */
3434                 break;
3435         case MOD_QUIESCE:
3436                 /* Called before unload. May veto unloading. */
3437                 break;
3438         case MOD_UNLOAD:
3439                 /* Called during unload. */
3440                 break;
3441         case MOD_SHUTDOWN:
3442                 /* Called during system shutdown. */
3443                 break;
3444         default:
3445                 err = EOPNOTSUPP;
3446                 break;
3447         }
3448         return err;
3449 }
3450
3451 static moduledata_t ipfwmod = {
3452         "ipfw",
3453         ipfw_modevent,
3454         0
3455 };
3456
3457 /* Define startup order. */
3458 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3459 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3460 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3461 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3462
3463 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3464 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3465 MODULE_VERSION(ipfw, 3);
3466 /* should declare some dependencies here */
3467
3468 /*
3469  * Starting up. Done in order after ipfwmod() has been called.
3470  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3471  */
3472 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3473             ipfw_init, NULL);
3474 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3475             vnet_ipfw_init, NULL);
3476
3477 /*
3478  * Closing up shop. These are done in REVERSE ORDER, but still
3479  * after ipfwmod() has been called. Not called on reboot.
3480  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3481  * or when the module is unloaded.
3482  */
3483 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3484             ipfw_destroy, NULL);
3485 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3486             vnet_ipfw_uninit, NULL);
3487 /* end of file */