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_match(struct tcphdr *tcp, ipfw_insn *cmd)
 335 {
 336         int optlen, bits = 0;
 337         u_char *cp = (u_char *)(tcp + 1);
 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
 354                 default:
 355                         break;
 356
 357                 case TCPOPT_MAXSEG:
 358                         bits |= IP_FW_TCPOPT_MSS;
 359                         break;
 360
 361                 case TCPOPT_WINDOW:
 362                         bits |= IP_FW_TCPOPT_WINDOW;
 363                         break;
 364
 365                 case TCPOPT_SACK_PERMITTED:
 366                 case TCPOPT_SACK:
 367                         bits |= IP_FW_TCPOPT_SACK;
 368                         break;
 369
 370                 case TCPOPT_TIMESTAMP:
 371                         bits |= IP_FW_TCPOPT_TS;
 372                         break;
 373
 374                 }
 375         }
 376         return (flags_match(cmd, bits));
 377 }
 378
 379 static int
 380 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
 381     uint32_t *tablearg)
 382 {
 383
 384         if (ifp == NULL)        /* no iface with this packet, match fails */
 385                 return (0);
 386
 387         /* Check by name or by IP address */
 388         if (cmd->name[0] != '\0') { /* match by name */
 389                 if (cmd->name[0] == '\1') /* use tablearg to match */
 390                         return ipfw_lookup_table(chain, cmd->p.kidx, 0,
 391                             &ifp->if_index, tablearg);
 392                 /* Check name */
 393                 if (cmd->p.glob) {
 394                         if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
 395                                 return(1);
 396                 } else {
 397                         if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
 398                                 return(1);
 399                 }
 400         } else {
 401 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
 402                 struct ifaddr *ia;
 403
 404                 if_addr_rlock(ifp);
 405                 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
 406                         if (ia->ifa_addr->sa_family != AF_INET)
 407                                 continue;
 408                         if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
 409                             (ia->ifa_addr))->sin_addr.s_addr) {
 410                                 if_addr_runlock(ifp);
 411                                 return(1);      /* match */
 412                         }
 413                 }
 414                 if_addr_runlock(ifp);
 415 #endif /* __FreeBSD__ */
 416         }
 417         return(0);      /* no match, fail ... */
 418 }
 419
 420 /*
 421  * The verify_path function checks if a route to the src exists and
 422  * if it is reachable via ifp (when provided).
 423  *
 424  * The 'verrevpath' option checks that the interface that an IP packet
 425  * arrives on is the same interface that traffic destined for the
 426  * packet's source address would be routed out of.
 427  * The 'versrcreach' option just checks that the source address is
 428  * reachable via any route (except default) in the routing table.
 429  * These two are a measure to block forged packets. This is also
 430  * commonly known as "anti-spoofing" or Unicast Reverse Path
 431  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
 432  * is purposely reminiscent of the Cisco IOS command,
 433  *
 434  *   ip verify unicast reverse-path
 435  *   ip verify unicast source reachable-via any
 436  *
 437  * which implements the same functionality. But note that the syntax
 438  * is misleading, and the check may be performed on all IP packets
 439  * whether unicast, multicast, or broadcast.
 440  */
 441 static int
 442 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
 443 {
 444 #if defined(USERSPACE) || !defined(__FreeBSD__)
 445         return 0;
 446 #else
 447         struct nhop4_basic nh4;
 448
 449         if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
 450                 return (0);
 451
 452         /*
 453          * If ifp is provided, check for equality with rtentry.
 454          * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
 455          * in order to pass packets injected back by if_simloop():
 456          * routing entry (via lo0) for our own address
 457          * may exist, so we need to handle routing assymetry.
 458          */
 459         if (ifp != NULL && ifp != nh4.nh_ifp)
 460                 return (0);
 461
 462         /* if no ifp provided, check if rtentry is not default route */
 463         if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
 464                 return (0);
 465
 466         /* or if this is a blackhole/reject route */
 467         if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 468                 return (0);
 469
 470         /* found valid route */
 471         return 1;
 472 #endif /* __FreeBSD__ */
 473 }
 474
 475 /*
 476  * Generate an SCTP packet containing an ABORT chunk. The verification tag
 477  * is given by vtag. The T-bit is set in the ABORT chunk if and only if
 478  * reflected is not 0.
 479  */
 480
 481 static struct mbuf *
 482 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
 483     int reflected)
 484 {
 485         struct mbuf *m;
 486         struct ip *ip;
 487 #ifdef INET6
 488         struct ip6_hdr *ip6;
 489 #endif
 490         struct sctphdr *sctp;
 491         struct sctp_chunkhdr *chunk;
 492         u_int16_t hlen, plen, tlen;
 493
 494         MGETHDR(m, M_NOWAIT, MT_DATA);
 495         if (m == NULL)
 496                 return (NULL);
 497
 498         M_SETFIB(m, id->fib);
 499 #ifdef MAC
 500         if (replyto != NULL)
 501                 mac_netinet_firewall_reply(replyto, m);
 502         else
 503                 mac_netinet_firewall_send(m);
 504 #else
 505         (void)replyto;          /* don't warn about unused arg */
 506 #endif
 507
 508         switch (id->addr_type) {
 509         case 4:
 510                 hlen = sizeof(struct ip);
 511                 break;
 512 #ifdef INET6
 513         case 6:
 514                 hlen = sizeof(struct ip6_hdr);
 515                 break;
 516 #endif
 517         default:
 518                 /* XXX: log me?!? */
 519                 FREE_PKT(m);
 520                 return (NULL);
 521         }
 522         plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
 523         tlen = hlen + plen;
 524         m->m_data += max_linkhdr;
 525         m->m_flags |= M_SKIP_FIREWALL;
 526         m->m_pkthdr.len = m->m_len = tlen;
 527         m->m_pkthdr.rcvif = NULL;
 528         bzero(m->m_data, tlen);
 529
 530         switch (id->addr_type) {
 531         case 4:
 532                 ip = mtod(m, struct ip *);
 533
 534                 ip->ip_v = 4;
 535                 ip->ip_hl = sizeof(struct ip) >> 2;
 536                 ip->ip_tos = IPTOS_LOWDELAY;
 537                 ip->ip_len = htons(tlen);
 538                 ip->ip_id = htons(0);
 539                 ip->ip_off = htons(0);
 540                 ip->ip_ttl = V_ip_defttl;
 541                 ip->ip_p = IPPROTO_SCTP;
 542                 ip->ip_sum = 0;
 543                 ip->ip_src.s_addr = htonl(id->dst_ip);
 544                 ip->ip_dst.s_addr = htonl(id->src_ip);
 545
 546                 sctp = (struct sctphdr *)(ip + 1);
 547                 break;
 548 #ifdef INET6
 549         case 6:
 550                 ip6 = mtod(m, struct ip6_hdr *);
 551
 552                 ip6->ip6_vfc = IPV6_VERSION;
 553                 ip6->ip6_plen = htons(plen);
 554                 ip6->ip6_nxt = IPPROTO_SCTP;
 555                 ip6->ip6_hlim = IPV6_DEFHLIM;
 556                 ip6->ip6_src = id->dst_ip6;
 557                 ip6->ip6_dst = id->src_ip6;
 558
 559                 sctp = (struct sctphdr *)(ip6 + 1);
 560                 break;
 561 #endif
 562         }
 563
 564         sctp->src_port = htons(id->dst_port);
 565         sctp->dest_port = htons(id->src_port);
 566         sctp->v_tag = htonl(vtag);
 567         sctp->checksum = htonl(0);
 568
 569         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 570         chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
 571         chunk->chunk_flags = 0;
 572         if (reflected != 0) {
 573                 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
 574         }
 575         chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
 576
 577         sctp->checksum = sctp_calculate_cksum(m, hlen);
 578
 579         return (m);
 580 }
 581
 582 /*
 583  * Generate a TCP packet, containing either a RST or a keepalive.
 584  * When flags & TH_RST, we are sending a RST packet, because of a
 585  * "reset" action matched the packet.
 586  * Otherwise we are sending a keepalive, and flags & TH_
 587  * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
 588  * so that MAC can label the reply appropriately.
 589  */
 590 struct mbuf *
 591 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
 592     u_int32_t ack, int flags)
 593 {
 594         struct mbuf *m = NULL;          /* stupid compiler */
 595         struct ip *h = NULL;            /* stupid compiler */
 596 #ifdef INET6
 597         struct ip6_hdr *h6 = NULL;
 598 #endif
 599         struct tcphdr *th = NULL;
 600         int len, dir;
 601
 602         MGETHDR(m, M_NOWAIT, MT_DATA);
 603         if (m == NULL)
 604                 return (NULL);
 605
 606         M_SETFIB(m, id->fib);
 607 #ifdef MAC
 608         if (replyto != NULL)
 609                 mac_netinet_firewall_reply(replyto, m);
 610         else
 611                 mac_netinet_firewall_send(m);
 612 #else
 613         (void)replyto;          /* don't warn about unused arg */
 614 #endif
 615
 616         switch (id->addr_type) {
 617         case 4:
 618                 len = sizeof(struct ip) + sizeof(struct tcphdr);
 619                 break;
 620 #ifdef INET6
 621         case 6:
 622                 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
 623                 break;
 624 #endif
 625         default:
 626                 /* XXX: log me?!? */
 627                 FREE_PKT(m);
 628                 return (NULL);
 629         }
 630         dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
 631
 632         m->m_data += max_linkhdr;
 633         m->m_flags |= M_SKIP_FIREWALL;
 634         m->m_pkthdr.len = m->m_len = len;
 635         m->m_pkthdr.rcvif = NULL;
 636         bzero(m->m_data, len);
 637
 638         switch (id->addr_type) {
 639         case 4:
 640                 h = mtod(m, struct ip *);
 641
 642                 /* prepare for checksum */
 643                 h->ip_p = IPPROTO_TCP;
 644                 h->ip_len = htons(sizeof(struct tcphdr));
 645                 if (dir) {
 646                         h->ip_src.s_addr = htonl(id->src_ip);
 647                         h->ip_dst.s_addr = htonl(id->dst_ip);
 648                 } else {
 649                         h->ip_src.s_addr = htonl(id->dst_ip);
 650                         h->ip_dst.s_addr = htonl(id->src_ip);
 651                 }
 652
 653                 th = (struct tcphdr *)(h + 1);
 654                 break;
 655 #ifdef INET6
 656         case 6:
 657                 h6 = mtod(m, struct ip6_hdr *);
 658
 659                 /* prepare for checksum */
 660                 h6->ip6_nxt = IPPROTO_TCP;
 661                 h6->ip6_plen = htons(sizeof(struct tcphdr));
 662                 if (dir) {
 663                         h6->ip6_src = id->src_ip6;
 664                         h6->ip6_dst = id->dst_ip6;
 665                 } else {
 666                         h6->ip6_src = id->dst_ip6;
 667                         h6->ip6_dst = id->src_ip6;
 668                 }
 669
 670                 th = (struct tcphdr *)(h6 + 1);
 671                 break;
 672 #endif
 673         }
 674
 675         if (dir) {
 676                 th->th_sport = htons(id->src_port);
 677                 th->th_dport = htons(id->dst_port);
 678         } else {
 679                 th->th_sport = htons(id->dst_port);
 680                 th->th_dport = htons(id->src_port);
 681         }
 682         th->th_off = sizeof(struct tcphdr) >> 2;
 683
 684         if (flags & TH_RST) {
 685                 if (flags & TH_ACK) {
 686                         th->th_seq = htonl(ack);
 687                         th->th_flags = TH_RST;
 688                 } else {
 689                         if (flags & TH_SYN)
 690                                 seq++;
 691                         th->th_ack = htonl(seq);
 692                         th->th_flags = TH_RST | TH_ACK;
 693                 }
 694         } else {
 695                 /*
 696                  * Keepalive - use caller provided sequence numbers
 697                  */
 698                 th->th_seq = htonl(seq);
 699                 th->th_ack = htonl(ack);
 700                 th->th_flags = TH_ACK;
 701         }
 702
 703         switch (id->addr_type) {
 704         case 4:
 705                 th->th_sum = in_cksum(m, len);
 706
 707                 /* finish the ip header */
 708                 h->ip_v = 4;
 709                 h->ip_hl = sizeof(*h) >> 2;
 710                 h->ip_tos = IPTOS_LOWDELAY;
 711                 h->ip_off = htons(0);
 712                 h->ip_len = htons(len);
 713                 h->ip_ttl = V_ip_defttl;
 714                 h->ip_sum = 0;
 715                 break;
 716 #ifdef INET6
 717         case 6:
 718                 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
 719                     sizeof(struct tcphdr));
 720
 721                 /* finish the ip6 header */
 722                 h6->ip6_vfc |= IPV6_VERSION;
 723                 h6->ip6_hlim = IPV6_DEFHLIM;
 724                 break;
 725 #endif
 726         }
 727
 728         return (m);
 729 }
 730
 731 #ifdef INET6
 732 /*
 733  * ipv6 specific rules here...
 734  */
 735 static __inline int
 736 icmp6type_match (int type, ipfw_insn_u32 *cmd)
 737 {
 738         return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
 739 }
 740
 741 static int
 742 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
 743 {
 744         int i;
 745         for (i=0; i <= cmd->o.arg1; ++i )
 746                 if (curr_flow == cmd->d[i] )
 747                         return 1;
 748         return 0;
 749 }
 750
 751 /* support for IP6_*_ME opcodes */
 752 static const struct in6_addr lla_mask = {{{
 753         0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
 754         0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 755 }}};
 756
 757 static int
 758 ipfw_localip6(struct in6_addr *in6)
 759 {
 760         struct rm_priotracker in6_ifa_tracker;
 761         struct in6_ifaddr *ia;
 762
 763         if (IN6_IS_ADDR_MULTICAST(in6))
 764                 return (0);
 765
 766         if (!IN6_IS_ADDR_LINKLOCAL(in6))
 767                 return (in6_localip(in6));
 768
 769         IN6_IFADDR_RLOCK(&in6_ifa_tracker);
 770         CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 771                 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
 772                         continue;
 773                 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 774                     in6, &lla_mask)) {
 775                         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 776                         return (1);
 777                 }
 778         }
 779         IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
 780         return (0);
 781 }
 782
 783 static int
 784 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
 785 {
 786         struct nhop6_basic nh6;
 787
 788         if (IN6_IS_SCOPE_LINKLOCAL(src))
 789                 return (1);
 790
 791         if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
 792                 return (0);
 793
 794         /* If ifp is provided, check for equality with route table. */
 795         if (ifp != NULL && ifp != nh6.nh_ifp)
 796                 return (0);
 797
 798         /* if no ifp provided, check if rtentry is not default route */
 799         if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
 800                 return (0);
 801
 802         /* or if this is a blackhole/reject route */
 803         if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
 804                 return (0);
 805
 806         /* found valid route */
 807         return 1;
 808 }
 809
 810 static int
 811 is_icmp6_query(int icmp6_type)
 812 {
 813         if ((icmp6_type <= ICMP6_MAXTYPE) &&
 814             (icmp6_type == ICMP6_ECHO_REQUEST ||
 815             icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
 816             icmp6_type == ICMP6_WRUREQUEST ||
 817             icmp6_type == ICMP6_FQDN_QUERY ||
 818             icmp6_type == ICMP6_NI_QUERY))
 819                 return (1);
 820
 821         return (0);
 822 }
 823
 824 static int
 825 map_icmp_unreach(int code)
 826 {
 827
 828         /* RFC 7915 p4.2 */
 829         switch (code) {
 830         case ICMP_UNREACH_NET:
 831         case ICMP_UNREACH_HOST:
 832         case ICMP_UNREACH_SRCFAIL:
 833         case ICMP_UNREACH_NET_UNKNOWN:
 834         case ICMP_UNREACH_HOST_UNKNOWN:
 835         case ICMP_UNREACH_TOSNET:
 836         case ICMP_UNREACH_TOSHOST:
 837                 return (ICMP6_DST_UNREACH_NOROUTE);
 838         case ICMP_UNREACH_PORT:
 839                 return (ICMP6_DST_UNREACH_NOPORT);
 840         default:
 841                 /*
 842                  * Map the rest of codes into admit prohibited.
 843                  * XXX: unreach proto should be mapped into ICMPv6
 844                  * parameter problem, but we use only unreach type.
 845                  */
 846                 return (ICMP6_DST_UNREACH_ADMIN);
 847         }
 848 }
 849
 850 static void
 851 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
 852 {
 853         struct mbuf *m;
 854
 855         m = args->m;
 856         if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
 857                 struct tcphdr *tcp;
 858                 tcp = (struct tcphdr *)((char *)ip6 + hlen);
 859
 860                 if ((tcp->th_flags & TH_RST) == 0) {
 861                         struct mbuf *m0;
 862                         m0 = ipfw_send_pkt(args->m, &(args->f_id),
 863                             ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 864                             tcp->th_flags | TH_RST);
 865                         if (m0 != NULL)
 866                                 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
 867                                     NULL);
 868                 }
 869                 FREE_PKT(m);
 870         } else if (code == ICMP6_UNREACH_ABORT &&
 871             args->f_id.proto == IPPROTO_SCTP) {
 872                 struct mbuf *m0;
 873                 struct sctphdr *sctp;
 874                 u_int32_t v_tag;
 875                 int reflected;
 876
 877                 sctp = (struct sctphdr *)((char *)ip6 + hlen);
 878                 reflected = 1;
 879                 v_tag = ntohl(sctp->v_tag);
 880                 /* Investigate the first chunk header if available */
 881                 if (m->m_len >= hlen + sizeof(struct sctphdr) +
 882                     sizeof(struct sctp_chunkhdr)) {
 883                         struct sctp_chunkhdr *chunk;
 884
 885                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
 886                         switch (chunk->chunk_type) {
 887                         case SCTP_INITIATION:
 888                                 /*
 889                                  * Packets containing an INIT chunk MUST have
 890                                  * a zero v-tag.
 891                                  */
 892                                 if (v_tag != 0) {
 893                                         v_tag = 0;
 894                                         break;
 895                                 }
 896                                 /* INIT chunk MUST NOT be bundled */
 897                                 if (m->m_pkthdr.len >
 898                                     hlen + sizeof(struct sctphdr) +
 899                                     ntohs(chunk->chunk_length) + 3) {
 900                                         break;
 901                                 }
 902                                 /* Use the initiate tag if available */
 903                                 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
 904                                     sizeof(struct sctp_chunkhdr) +
 905                                     offsetof(struct sctp_init, a_rwnd))) {
 906                                         struct sctp_init *init;
 907
 908                                         init = (struct sctp_init *)(chunk + 1);
 909                                         v_tag = ntohl(init->initiate_tag);
 910                                         reflected = 0;
 911                                 }
 912                                 break;
 913                         case SCTP_ABORT_ASSOCIATION:
 914                                 /*
 915                                  * If the packet contains an ABORT chunk, don't
 916                                  * reply.
 917                                  * XXX: We should search through all chunks,
 918                                  *      but don't do to avoid attacks.
 919                                  */
 920                                 v_tag = 0;
 921                                 break;
 922                         }
 923                 }
 924                 if (v_tag == 0) {
 925                         m0 = NULL;
 926                 } else {
 927                         m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
 928                             reflected);
 929                 }
 930                 if (m0 != NULL)
 931                         ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
 932                 FREE_PKT(m);
 933         } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
 934                 /* Send an ICMPv6 unreach. */
 935 #if 0
 936                 /*
 937                  * Unlike above, the mbufs need to line up with the ip6 hdr,
 938                  * as the contents are read. We need to m_adj() the
 939                  * needed amount.
 940                  * The mbuf will however be thrown away so we can adjust it.
 941                  * Remember we did an m_pullup on it already so we
 942                  * can make some assumptions about contiguousness.
 943                  */
 944                 if (args->L3offset)
 945                         m_adj(m, args->L3offset);
 946 #endif
 947                 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
 948         } else
 949                 FREE_PKT(m);
 950
 951         args->m = NULL;
 952 }
 953
 954 #endif /* INET6 */
 955
 956
 957 /*
 958  * sends a reject message, consuming the mbuf passed as an argument.
 959  */
 960 static void
 961 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
 962 {
 963
 964 #if 0
 965         /* XXX When ip is not guaranteed to be at mtod() we will
 966          * need to account for this */
 967          * The mbuf will however be thrown away so we can adjust it.
 968          * Remember we did an m_pullup on it already so we
 969          * can make some assumptions about contiguousness.
 970          */
 971         if (args->L3offset)
 972                 m_adj(m, args->L3offset);
 973 #endif
 974         if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
 975                 /* Send an ICMP unreach */
 976                 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
 977         } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
 978                 struct tcphdr *const tcp =
 979                     L3HDR(struct tcphdr, mtod(args->m, struct ip *));
 980                 if ( (tcp->th_flags & TH_RST) == 0) {
 981                         struct mbuf *m;
 982                         m = ipfw_send_pkt(args->m, &(args->f_id),
 983                                 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
 984                                 tcp->th_flags | TH_RST);
 985                         if (m != NULL)
 986                                 ip_output(m, NULL, NULL, 0, NULL, NULL);
 987                 }
 988                 FREE_PKT(args->m);
 989         } else if (code == ICMP_REJECT_ABORT &&
 990             args->f_id.proto == IPPROTO_SCTP) {
 991                 struct mbuf *m;
 992                 struct sctphdr *sctp;
 993                 struct sctp_chunkhdr *chunk;
 994                 struct sctp_init *init;
 995                 u_int32_t v_tag;
 996                 int reflected;
 997
 998                 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
 999                 reflected = 1;
1000                 v_tag = ntohl(sctp->v_tag);
1001                 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1002                     sizeof(struct sctp_chunkhdr)) {
1003                         /* Look at the first chunk header if available */
1004                         chunk = (struct sctp_chunkhdr *)(sctp + 1);
1005                         switch (chunk->chunk_type) {
1006                         case SCTP_INITIATION:
1007                                 /*
1008                                  * Packets containing an INIT chunk MUST have
1009                                  * a zero v-tag.
1010                                  */
1011                                 if (v_tag != 0) {
1012                                         v_tag = 0;
1013                                         break;
1014                                 }
1015                                 /* INIT chunk MUST NOT be bundled */
1016                                 if (iplen >
1017                                     (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1018                                     ntohs(chunk->chunk_length) + 3) {
1019                                         break;
1020                                 }
1021                                 /* Use the initiate tag if available */
1022                                 if ((iplen >= (ip->ip_hl << 2) +
1023                                     sizeof(struct sctphdr) +
1024                                     sizeof(struct sctp_chunkhdr) +
1025                                     offsetof(struct sctp_init, a_rwnd))) {
1026                                         init = (struct sctp_init *)(chunk + 1);
1027                                         v_tag = ntohl(init->initiate_tag);
1028                                         reflected = 0;
1029                                 }
1030                                 break;
1031                         case SCTP_ABORT_ASSOCIATION:
1032                                 /*
1033                                  * If the packet contains an ABORT chunk, don't
1034                                  * reply.
1035                                  * XXX: We should search through all chunks,
1036                                  * but don't do to avoid attacks.
1037                                  */
1038                                 v_tag = 0;
1039                                 break;
1040                         }
1041                 }
1042                 if (v_tag == 0) {
1043                         m = NULL;
1044                 } else {
1045                         m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1046                             reflected);
1047                 }
1048                 if (m != NULL)
1049                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1050                 FREE_PKT(args->m);
1051         } else
1052                 FREE_PKT(args->m);
1053         args->m = NULL;
1054 }
1055
1056 /*
1057  * Support for uid/gid/jail lookup. These tests are expensive
1058  * (because we may need to look into the list of active sockets)
1059  * so we cache the results. ugid_lookupp is 0 if we have not
1060  * yet done a lookup, 1 if we succeeded, and -1 if we tried
1061  * and failed. The function always returns the match value.
1062  * We could actually spare the variable and use *uc, setting
1063  * it to '(void *)check_uidgid if we have no info, NULL if
1064  * we tried and failed, or any other value if successful.
1065  */
1066 static int
1067 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1068     struct ucred **uc)
1069 {
1070 #if defined(USERSPACE)
1071         return 0;       // not supported in userspace
1072 #else
1073 #ifndef __FreeBSD__
1074         /* XXX */
1075         return cred_check(insn, proto, oif,
1076             dst_ip, dst_port, src_ip, src_port,
1077             (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1078 #else  /* FreeBSD */
1079         struct in_addr src_ip, dst_ip;
1080         struct inpcbinfo *pi;
1081         struct ipfw_flow_id *id;
1082         struct inpcb *pcb, *inp;
1083         struct ifnet *oif;
1084         int lookupflags;
1085         int match;
1086
1087         id = &args->f_id;
1088         inp = args->inp;
1089         oif = args->oif;
1090
1091         /*
1092          * Check to see if the UDP or TCP stack supplied us with
1093          * the PCB. If so, rather then holding a lock and looking
1094          * up the PCB, we can use the one that was supplied.
1095          */
1096         if (inp && *ugid_lookupp == 0) {
1097                 INP_LOCK_ASSERT(inp);
1098                 if (inp->inp_socket != NULL) {
1099                         *uc = crhold(inp->inp_cred);
1100                         *ugid_lookupp = 1;
1101                 } else
1102                         *ugid_lookupp = -1;
1103         }
1104         /*
1105          * If we have already been here and the packet has no
1106          * PCB entry associated with it, then we can safely
1107          * assume that this is a no match.
1108          */
1109         if (*ugid_lookupp == -1)
1110                 return (0);
1111         if (id->proto == IPPROTO_TCP) {
1112                 lookupflags = 0;
1113                 pi = &V_tcbinfo;
1114         } else if (id->proto == IPPROTO_UDP) {
1115                 lookupflags = INPLOOKUP_WILDCARD;
1116                 pi = &V_udbinfo;
1117         } else if (id->proto == IPPROTO_UDPLITE) {
1118                 lookupflags = INPLOOKUP_WILDCARD;
1119                 pi = &V_ulitecbinfo;
1120         } else
1121                 return 0;
1122         lookupflags |= INPLOOKUP_RLOCKPCB;
1123         match = 0;
1124         if (*ugid_lookupp == 0) {
1125                 if (id->addr_type == 6) {
1126 #ifdef INET6
1127                         if (oif == NULL)
1128                                 pcb = in6_pcblookup_mbuf(pi,
1129                                     &id->src_ip6, htons(id->src_port),
1130                                     &id->dst_ip6, htons(id->dst_port),
1131                                     lookupflags, oif, args->m);
1132                         else
1133                                 pcb = in6_pcblookup_mbuf(pi,
1134                                     &id->dst_ip6, htons(id->dst_port),
1135                                     &id->src_ip6, htons(id->src_port),
1136                                     lookupflags, oif, args->m);
1137 #else
1138                         *ugid_lookupp = -1;
1139                         return (0);
1140 #endif
1141                 } else {
1142                         src_ip.s_addr = htonl(id->src_ip);
1143                         dst_ip.s_addr = htonl(id->dst_ip);
1144                         if (oif == NULL)
1145                                 pcb = in_pcblookup_mbuf(pi,
1146                                     src_ip, htons(id->src_port),
1147                                     dst_ip, htons(id->dst_port),
1148                                     lookupflags, oif, args->m);
1149                         else
1150                                 pcb = in_pcblookup_mbuf(pi,
1151                                     dst_ip, htons(id->dst_port),
1152                                     src_ip, htons(id->src_port),
1153                                     lookupflags, oif, args->m);
1154                 }
1155                 if (pcb != NULL) {
1156                         INP_RLOCK_ASSERT(pcb);
1157                         *uc = crhold(pcb->inp_cred);
1158                         *ugid_lookupp = 1;
1159                         INP_RUNLOCK(pcb);
1160                 }
1161                 if (*ugid_lookupp == 0) {
1162                         /*
1163                          * We tried and failed, set the variable to -1
1164                          * so we will not try again on this packet.
1165                          */
1166                         *ugid_lookupp = -1;
1167                         return (0);
1168                 }
1169         }
1170         if (insn->o.opcode == O_UID)
1171                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1172         else if (insn->o.opcode == O_GID)
1173                 match = groupmember((gid_t)insn->d[0], *uc);
1174         else if (insn->o.opcode == O_JAIL)
1175                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1176         return (match);
1177 #endif /* __FreeBSD__ */
1178 #endif /* not supported in userspace */
1179 }
1180
1181 /*
1182  * Helper function to set args with info on the rule after the matching
1183  * one. slot is precise, whereas we guess rule_id as they are
1184  * assigned sequentially.
1185  */
1186 static inline void
1187 set_match(struct ip_fw_args *args, int slot,
1188         struct ip_fw_chain *chain)
1189 {
1190         args->rule.chain_id = chain->id;
1191         args->rule.slot = slot + 1; /* we use 0 as a marker */
1192         args->rule.rule_id = 1 + chain->map[slot]->id;
1193         args->rule.rulenum = chain->map[slot]->rulenum;
1194 }
1195
1196 #ifndef LINEAR_SKIPTO
1197 /*
1198  * Helper function to enable cached rule lookups using
1199  * cached_id and cached_pos fields in ipfw rule.
1200  */
1201 static int
1202 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1203     int tablearg, int jump_backwards)
1204 {
1205         int f_pos;
1206
1207         /* If possible use cached f_pos (in f->cached_pos),
1208          * whose version is written in f->cached_id
1209          * (horrible hacks to avoid changing the ABI).
1210          */
1211         if (num != IP_FW_TARG && f->cached_id == chain->id)
1212                 f_pos = f->cached_pos;
1213         else {
1214                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1215                 /* make sure we do not jump backward */
1216                 if (jump_backwards == 0 && i <= f->rulenum)
1217                         i = f->rulenum + 1;
1218                 if (chain->idxmap != NULL)
1219                         f_pos = chain->idxmap[i];
1220                 else
1221                         f_pos = ipfw_find_rule(chain, i, 0);
1222                 /* update the cache */
1223                 if (num != IP_FW_TARG) {
1224                         f->cached_id = chain->id;
1225                         f->cached_pos = f_pos;
1226                 }
1227         }
1228
1229         return (f_pos);
1230 }
1231 #else
1232 /*
1233  * Helper function to enable real fast rule lookups.
1234  */
1235 static int
1236 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1237     int tablearg, int jump_backwards)
1238 {
1239         int f_pos;
1240
1241         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1242         /* make sure we do not jump backward */
1243         if (jump_backwards == 0 && num <= f->rulenum)
1244                 num = f->rulenum + 1;
1245         f_pos = chain->idxmap[num];
1246
1247         return (f_pos);
1248 }
1249 #endif
1250
1251 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1252 /*
1253  * The main check routine for the firewall.
1254  *
1255  * All arguments are in args so we can modify them and return them
1256  * back to the caller.
1257  *
1258  * Parameters:
1259  *
1260  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1261  *              Starts with the IP header.
1262  *      args->eh (in)   Mac header if present, NULL for layer3 packet.
1263  *      args->L3offset  Number of bytes bypassed if we came from L2.
1264  *                      e.g. often sizeof(eh)  ** NOTYET **
1265  *      args->oif       Outgoing interface, NULL if packet is incoming.
1266  *              The incoming interface is in the mbuf. (in)
1267  *      args->divert_rule (in/out)
1268  *              Skip up to the first rule past this rule number;
1269  *              upon return, non-zero port number for divert or tee.
1270  *
1271  *      args->rule      Pointer to the last matching rule (in/out)
1272  *      args->next_hop  Socket we are forwarding to (out).
1273  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1274  *      args->f_id      Addresses grabbed from the packet (out)
1275  *      args->rule.info a cookie depending on rule action
1276  *
1277  * Return value:
1278  *
1279  *      IP_FW_PASS      the packet must be accepted
1280  *      IP_FW_DENY      the packet must be dropped
1281  *      IP_FW_DIVERT    divert packet, port in m_tag
1282  *      IP_FW_TEE       tee packet, port in m_tag
1283  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1284  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1285  *              args->rule contains the matching rule,
1286  *              args->rule.info has additional information.
1287  *
1288  */
1289 int
1290 ipfw_chk(struct ip_fw_args *args)
1291 {
1292
1293         /*
1294          * Local variables holding state while processing a packet:
1295          *
1296          * IMPORTANT NOTE: to speed up the processing of rules, there
1297          * are some assumption on the values of the variables, which
1298          * are documented here. Should you change them, please check
1299          * the implementation of the various instructions to make sure
1300          * that they still work.
1301          *
1302          * args->eh     The MAC header. It is non-null for a layer2
1303          *      packet, it is NULL for a layer-3 packet.
1304          * **notyet**
1305          * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1306          *
1307          * m | args->m  Pointer to the mbuf, as received from the caller.
1308          *      It may change if ipfw_chk() does an m_pullup, or if it
1309          *      consumes the packet because it calls send_reject().
1310          *      XXX This has to change, so that ipfw_chk() never modifies
1311          *      or consumes the buffer.
1312          * ip   is the beginning of the ip(4 or 6) header.
1313          *      Calculated by adding the L3offset to the start of data.
1314          *      (Until we start using L3offset, the packet is
1315          *      supposed to start with the ip header).
1316          */
1317         struct mbuf *m = args->m;
1318         struct ip *ip = mtod(m, struct ip *);
1319
1320         /*
1321          * For rules which contain uid/gid or jail constraints, cache
1322          * a copy of the users credentials after the pcb lookup has been
1323          * executed. This will speed up the processing of rules with
1324          * these types of constraints, as well as decrease contention
1325          * on pcb related locks.
1326          */
1327 #ifndef __FreeBSD__
1328         struct bsd_ucred ucred_cache;
1329 #else
1330         struct ucred *ucred_cache = NULL;
1331 #endif
1332         int ucred_lookup = 0;
1333
1334         /*
1335          * oif | args->oif      If NULL, ipfw_chk has been called on the
1336          *      inbound path (ether_input, ip_input).
1337          *      If non-NULL, ipfw_chk has been called on the outbound path
1338          *      (ether_output, ip_output).
1339          */
1340         struct ifnet *oif = args->oif;
1341
1342         int f_pos = 0;          /* index of current rule in the array */
1343         int retval = 0;
1344
1345         /*
1346          * hlen The length of the IP header.
1347          */
1348         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1349
1350         /*
1351          * offset       The offset of a fragment. offset != 0 means that
1352          *      we have a fragment at this offset of an IPv4 packet.
1353          *      offset == 0 means that (if this is an IPv4 packet)
1354          *      this is the first or only fragment.
1355          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1356          *      or there is a single packet fragment (fragment header added
1357          *      without needed).  We will treat a single packet fragment as if
1358          *      there was no fragment header (or log/block depending on the
1359          *      V_fw_permit_single_frag6 sysctl setting).
1360          */
1361         u_short offset = 0;
1362         u_short ip6f_mf = 0;
1363
1364         /*
1365          * Local copies of addresses. They are only valid if we have
1366          * an IP packet.
1367          *
1368          * proto        The protocol. Set to 0 for non-ip packets,
1369          *      or to the protocol read from the packet otherwise.
1370          *      proto != 0 means that we have an IPv4 packet.
1371          *
1372          * src_port, dst_port   port numbers, in HOST format. Only
1373          *      valid for TCP and UDP packets.
1374          *
1375          * src_ip, dst_ip       ip addresses, in NETWORK format.
1376          *      Only valid for IPv4 packets.
1377          */
1378         uint8_t proto;
1379         uint16_t src_port = 0, dst_port = 0;    /* NOTE: host format    */
1380         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1381         int iplen = 0;
1382         int pktlen;
1383         uint16_t        etype = 0;      /* Host order stored ether type */
1384
1385         /*
1386          * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1387          *      MATCH_NONE when checked and not matched (q = NULL),
1388          *      MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1389          */
1390         struct ipfw_dyn_info dyn_info;
1391         struct ip_fw *q = NULL;
1392         struct ip_fw_chain *chain = &V_layer3_chain;
1393
1394         /*
1395          * We store in ulp a pointer to the upper layer protocol header.
1396          * In the ipv4 case this is easy to determine from the header,
1397          * but for ipv6 we might have some additional headers in the middle.
1398          * ulp is NULL if not found.
1399          */
1400         void *ulp = NULL;               /* upper layer protocol pointer. */
1401
1402         /* XXX ipv6 variables */
1403         int is_ipv6 = 0;
1404         uint8_t icmp6_type = 0;
1405         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1406         /* end of ipv6 variables */
1407
1408         int is_ipv4 = 0;
1409
1410         int done = 0;           /* flag to exit the outer loop */
1411
1412         if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1413                 return (IP_FW_PASS);    /* accept */
1414
1415         dst_ip.s_addr = 0;              /* make sure it is initialized */
1416         src_ip.s_addr = 0;              /* make sure it is initialized */
1417         pktlen = m->m_pkthdr.len;
1418         args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1419         proto = args->f_id.proto = 0;   /* mark f_id invalid */
1420                 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1421
1422         DYN_INFO_INIT(&dyn_info);
1423 /*
1424  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1425  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1426  * pointer might become stale after other pullups (but we never use it
1427  * this way).
1428  */
1429 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1430 #define PULLUP_LEN(_len, p, T)                                  \
1431 do {                                                            \
1432         int x = (_len) + T;                                     \
1433         if ((m)->m_len < x) {                                   \
1434                 args->m = m = m_pullup(m, x);                   \
1435                 if (m == NULL)                                  \
1436                         goto pullup_failed;                     \
1437         }                                                       \
1438         p = (mtod(m, char *) + (_len));                         \
1439 } while (0)
1440
1441         /*
1442          * if we have an ether header,
1443          */
1444         if (args->eh)
1445                 etype = ntohs(args->eh->ether_type);
1446
1447         /* Identify IP packets and fill up variables. */
1448         if (pktlen >= sizeof(struct ip6_hdr) &&
1449             (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1450                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1451                 is_ipv6 = 1;
1452                 args->f_id.addr_type = 6;
1453                 hlen = sizeof(struct ip6_hdr);
1454                 proto = ip6->ip6_nxt;
1455
1456                 /* Search extension headers to find upper layer protocols */
1457                 while (ulp == NULL && offset == 0) {
1458                         switch (proto) {
1459                         case IPPROTO_ICMPV6:
1460                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1461                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1462                                 break;
1463
1464                         case IPPROTO_TCP:
1465                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1466                                 dst_port = TCP(ulp)->th_dport;
1467                                 src_port = TCP(ulp)->th_sport;
1468                                 /* save flags for dynamic rules */
1469                                 args->f_id._flags = TCP(ulp)->th_flags;
1470                                 break;
1471
1472                         case IPPROTO_SCTP:
1473                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1474                                     sizeof(struct sctp_chunkhdr) +
1475                                     offsetof(struct sctp_init, a_rwnd))
1476                                         PULLUP_LEN(hlen, ulp,
1477                                             sizeof(struct sctphdr) +
1478                                             sizeof(struct sctp_chunkhdr) +
1479                                             offsetof(struct sctp_init, a_rwnd));
1480                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1481                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1482                                 else
1483                                         PULLUP_LEN(hlen, ulp,
1484                                             sizeof(struct sctphdr));
1485                                 src_port = SCTP(ulp)->src_port;
1486                                 dst_port = SCTP(ulp)->dest_port;
1487                                 break;
1488
1489                         case IPPROTO_UDP:
1490                         case IPPROTO_UDPLITE:
1491                                 PULLUP_TO(hlen, ulp, struct udphdr);
1492                                 dst_port = UDP(ulp)->uh_dport;
1493                                 src_port = UDP(ulp)->uh_sport;
1494                                 break;
1495
1496                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1497                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1498                                 ext_hd |= EXT_HOPOPTS;
1499                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1500                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1501                                 ulp = NULL;
1502                                 break;
1503
1504                         case IPPROTO_ROUTING:   /* RFC 2460 */
1505                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1506                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1507                                 case 0:
1508                                         ext_hd |= EXT_RTHDR0;
1509                                         break;
1510                                 case 2:
1511                                         ext_hd |= EXT_RTHDR2;
1512                                         break;
1513                                 default:
1514                                         if (V_fw_verbose)
1515                                                 printf("IPFW2: IPV6 - Unknown "
1516                                                     "Routing Header type(%d)\n",
1517                                                     ((struct ip6_rthdr *)
1518                                                     ulp)->ip6r_type);
1519                                         if (V_fw_deny_unknown_exthdrs)
1520                                             return (IP_FW_DENY);
1521                                         break;
1522                                 }
1523                                 ext_hd |= EXT_ROUTING;
1524                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1525                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1526                                 ulp = NULL;
1527                                 break;
1528
1529                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1530                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1531                                 ext_hd |= EXT_FRAGMENT;
1532                                 hlen += sizeof (struct ip6_frag);
1533                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1534                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1535                                         IP6F_OFF_MASK;
1536                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1537                                         IP6F_MORE_FRAG;
1538                                 if (V_fw_permit_single_frag6 == 0 &&
1539                                     offset == 0 && ip6f_mf == 0) {
1540                                         if (V_fw_verbose)
1541                                                 printf("IPFW2: IPV6 - Invalid "
1542                                                     "Fragment Header\n");
1543                                         if (V_fw_deny_unknown_exthdrs)
1544                                             return (IP_FW_DENY);
1545                                         break;
1546                                 }
1547                                 args->f_id.extra =
1548                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1549                                 ulp = NULL;
1550                                 break;
1551
1552                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1553                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1554                                 ext_hd |= EXT_DSTOPTS;
1555                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1556                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1557                                 ulp = NULL;
1558                                 break;
1559
1560                         case IPPROTO_AH:        /* RFC 2402 */
1561                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1562                                 ext_hd |= EXT_AH;
1563                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1564                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1565                                 ulp = NULL;
1566                                 break;
1567
1568                         case IPPROTO_ESP:       /* RFC 2406 */
1569                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1570                                 /* Anything past Seq# is variable length and
1571                                  * data past this ext. header is encrypted. */
1572                                 ext_hd |= EXT_ESP;
1573                                 break;
1574
1575                         case IPPROTO_NONE:      /* RFC 2460 */
1576                                 /*
1577                                  * Packet ends here, and IPv6 header has
1578                                  * already been pulled up. If ip6e_len!=0
1579                                  * then octets must be ignored.
1580                                  */
1581                                 ulp = ip; /* non-NULL to get out of loop. */
1582                                 break;
1583
1584                         case IPPROTO_OSPFIGP:
1585                                 /* XXX OSPF header check? */
1586                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1587                                 break;
1588
1589                         case IPPROTO_PIM:
1590                                 /* XXX PIM header check? */
1591                                 PULLUP_TO(hlen, ulp, struct pim);
1592                                 break;
1593
1594                         case IPPROTO_GRE:       /* RFC 1701 */
1595                                 /* XXX GRE header check? */
1596                                 PULLUP_TO(hlen, ulp, struct grehdr);
1597                                 break;
1598
1599                         case IPPROTO_CARP:
1600                                 PULLUP_TO(hlen, ulp, struct carp_header);
1601                                 if (((struct carp_header *)ulp)->carp_version !=
1602                                     CARP_VERSION)
1603                                         return (IP_FW_DENY);
1604                                 if (((struct carp_header *)ulp)->carp_type !=
1605                                     CARP_ADVERTISEMENT)
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.src_ip6 = ip6->ip6_src;
1631                 args->f_id.dst_ip6 = ip6->ip6_dst;
1632                 args->f_id.src_ip = 0;
1633                 args->f_id.dst_ip = 0;
1634                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1635                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1636         } else if (pktlen >= sizeof(struct ip) &&
1637             (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1638                 is_ipv4 = 1;
1639                 hlen = ip->ip_hl << 2;
1640                 args->f_id.addr_type = 4;
1641
1642                 /*
1643                  * Collect parameters into local variables for faster matching.
1644                  */
1645                 proto = ip->ip_p;
1646                 src_ip = ip->ip_src;
1647                 dst_ip = ip->ip_dst;
1648                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1649                 iplen = ntohs(ip->ip_len);
1650
1651                 if (offset == 0) {
1652                         switch (proto) {
1653                         case IPPROTO_TCP:
1654                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1655                                 dst_port = TCP(ulp)->th_dport;
1656                                 src_port = TCP(ulp)->th_sport;
1657                                 /* save flags for dynamic rules */
1658                                 args->f_id._flags = TCP(ulp)->th_flags;
1659                                 break;
1660
1661                         case IPPROTO_SCTP:
1662                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1663                                     sizeof(struct sctp_chunkhdr) +
1664                                     offsetof(struct sctp_init, a_rwnd))
1665                                         PULLUP_LEN(hlen, ulp,
1666                                             sizeof(struct sctphdr) +
1667                                             sizeof(struct sctp_chunkhdr) +
1668                                             offsetof(struct sctp_init, a_rwnd));
1669                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1670                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1671                                 else
1672                                         PULLUP_LEN(hlen, ulp,
1673                                             sizeof(struct sctphdr));
1674                                 src_port = SCTP(ulp)->src_port;
1675                                 dst_port = SCTP(ulp)->dest_port;
1676                                 break;
1677
1678                         case IPPROTO_UDP:
1679                         case IPPROTO_UDPLITE:
1680                                 PULLUP_TO(hlen, ulp, struct udphdr);
1681                                 dst_port = UDP(ulp)->uh_dport;
1682                                 src_port = UDP(ulp)->uh_sport;
1683                                 break;
1684
1685                         case IPPROTO_ICMP:
1686                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1687                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1688                                 break;
1689
1690                         default:
1691                                 break;
1692                         }
1693                 }
1694
1695                 ip = mtod(m, struct ip *);
1696                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1697                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1698         }
1699 #undef PULLUP_TO
1700         pktlen = iplen < pktlen ? iplen: pktlen;
1701         if (proto) { /* we may have port numbers, store them */
1702                 args->f_id.proto = proto;
1703                 args->f_id.src_port = src_port = ntohs(src_port);
1704                 args->f_id.dst_port = dst_port = ntohs(dst_port);
1705         }
1706
1707         IPFW_PF_RLOCK(chain);
1708         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1709                 IPFW_PF_RUNLOCK(chain);
1710                 return (IP_FW_PASS);    /* accept */
1711         }
1712         if (args->rule.slot) {
1713                 /*
1714                  * Packet has already been tagged as a result of a previous
1715                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1716                  * REASS, NETGRAPH, DIVERT/TEE...)
1717                  * Validate the slot and continue from the next one
1718                  * if still present, otherwise do a lookup.
1719                  */
1720                 f_pos = (args->rule.chain_id == chain->id) ?
1721                     args->rule.slot :
1722                     ipfw_find_rule(chain, args->rule.rulenum,
1723                         args->rule.rule_id);
1724         } else {
1725                 f_pos = 0;
1726         }
1727
1728         /*
1729          * Now scan the rules, and parse microinstructions for each rule.
1730          * We have two nested loops and an inner switch. Sometimes we
1731          * need to break out of one or both loops, or re-enter one of
1732          * the loops with updated variables. Loop variables are:
1733          *
1734          *      f_pos (outer loop) points to the current rule.
1735          *              On output it points to the matching rule.
1736          *      done (outer loop) is used as a flag to break the loop.
1737          *      l (inner loop)  residual length of current rule.
1738          *              cmd points to the current microinstruction.
1739          *
1740          * We break the inner loop by setting l=0 and possibly
1741          * cmdlen=0 if we don't want to advance cmd.
1742          * We break the outer loop by setting done=1
1743          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1744          * as needed.
1745          */
1746         for (; f_pos < chain->n_rules; f_pos++) {
1747                 ipfw_insn *cmd;
1748                 uint32_t tablearg = 0;
1749                 int l, cmdlen, skip_or; /* skip rest of OR block */
1750                 struct ip_fw *f;
1751
1752                 f = chain->map[f_pos];
1753                 if (V_set_disable & (1 << f->set) )
1754                         continue;
1755
1756                 skip_or = 0;
1757                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1758                     l -= cmdlen, cmd += cmdlen) {
1759                         int match;
1760
1761                         /*
1762                          * check_body is a jump target used when we find a
1763                          * CHECK_STATE, and need to jump to the body of
1764                          * the target rule.
1765                          */
1766
1767 /* check_body: */
1768                         cmdlen = F_LEN(cmd);
1769                         /*
1770                          * An OR block (insn_1 || .. || insn_n) has the
1771                          * F_OR bit set in all but the last instruction.
1772                          * The first match will set "skip_or", and cause
1773                          * the following instructions to be skipped until
1774                          * past the one with the F_OR bit clear.
1775                          */
1776                         if (skip_or) {          /* skip this instruction */
1777                                 if ((cmd->len & F_OR) == 0)
1778                                         skip_or = 0;    /* next one is good */
1779                                 continue;
1780                         }
1781                         match = 0; /* set to 1 if we succeed */
1782
1783                         switch (cmd->opcode) {
1784                         /*
1785                          * The first set of opcodes compares the packet's
1786                          * fields with some pattern, setting 'match' if a
1787                          * match is found. At the end of the loop there is
1788                          * logic to deal with F_NOT and F_OR flags associated
1789                          * with the opcode.
1790                          */
1791                         case O_NOP:
1792                                 match = 1;
1793                                 break;
1794
1795                         case O_FORWARD_MAC:
1796                                 printf("ipfw: opcode %d unimplemented\n",
1797                                     cmd->opcode);
1798                                 break;
1799
1800                         case O_GID:
1801                         case O_UID:
1802                         case O_JAIL:
1803                                 /*
1804                                  * We only check offset == 0 && proto != 0,
1805                                  * as this ensures that we have a
1806                                  * packet with the ports info.
1807                                  */
1808                                 if (offset != 0)
1809                                         break;
1810                                 if (proto == IPPROTO_TCP ||
1811                                     proto == IPPROTO_UDP ||
1812                                     proto == IPPROTO_UDPLITE)
1813                                         match = check_uidgid(
1814                                                     (ipfw_insn_u32 *)cmd,
1815                                                     args, &ucred_lookup,
1816 #ifdef __FreeBSD__
1817                                                     &ucred_cache);
1818 #else
1819                                                     (void *)&ucred_cache);
1820 #endif
1821                                 break;
1822
1823                         case O_RECV:
1824                                 match = iface_match(m->m_pkthdr.rcvif,
1825                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1826                                 break;
1827
1828                         case O_XMIT:
1829                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1830                                     chain, &tablearg);
1831                                 break;
1832
1833                         case O_VIA:
1834                                 match = iface_match(oif ? oif :
1835                                     m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1836                                     chain, &tablearg);
1837                                 break;
1838
1839                         case O_MACADDR2:
1840                                 if (args->eh != NULL) { /* have MAC header */
1841                                         u_int32_t *want = (u_int32_t *)
1842                                                 ((ipfw_insn_mac *)cmd)->addr;
1843                                         u_int32_t *mask = (u_int32_t *)
1844                                                 ((ipfw_insn_mac *)cmd)->mask;
1845                                         u_int32_t *hdr = (u_int32_t *)args->eh;
1846
1847                                         match =
1848                                             ( want[0] == (hdr[0] & mask[0]) &&
1849                                               want[1] == (hdr[1] & mask[1]) &&
1850                                               want[2] == (hdr[2] & mask[2]) );
1851                                 }
1852                                 break;
1853
1854                         case O_MAC_TYPE:
1855                                 if (args->eh != NULL) {
1856                                         u_int16_t *p =
1857                                             ((ipfw_insn_u16 *)cmd)->ports;
1858                                         int i;
1859
1860                                         for (i = cmdlen - 1; !match && i>0;
1861                                             i--, p += 2)
1862                                                 match = (etype >= p[0] &&
1863                                                     etype <= p[1]);
1864                                 }
1865                                 break;
1866
1867                         case O_FRAG:
1868                                 match = (offset != 0);
1869                                 break;
1870
1871                         case O_IN:      /* "out" is "not in" */
1872                                 match = (oif == NULL);
1873                                 break;
1874
1875                         case O_LAYER2:
1876                                 match = (args->eh != NULL);
1877                                 break;
1878
1879                         case O_DIVERTED:
1880                             {
1881                                 /* For diverted packets, args->rule.info
1882                                  * contains the divert port (in host format)
1883                                  * reason and direction.
1884                                  */
1885                                 uint32_t i = args->rule.info;
1886                                 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1887                                     cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1888                             }
1889                                 break;
1890
1891                         case O_PROTO:
1892                                 /*
1893                                  * We do not allow an arg of 0 so the
1894                                  * check of "proto" only suffices.
1895                                  */
1896                                 match = (proto == cmd->arg1);
1897                                 break;
1898
1899                         case O_IP_SRC:
1900                                 match = is_ipv4 &&
1901                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1902                                     src_ip.s_addr);
1903                                 break;
1904
1905                         case O_IP_DST_LOOKUP:
1906                         {
1907                                 void *pkey;
1908                                 uint32_t vidx, key;
1909                                 uint16_t keylen;
1910
1911                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1912                                         /* Determine lookup key type */
1913                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1914                                         if (vidx != 4 /* uid */ &&
1915                                             vidx != 5 /* jail */ &&
1916                                             is_ipv6 == 0 && is_ipv4 == 0)
1917                                                 break;
1918                                         /* Determine key length */
1919                                         if (vidx == 0 /* dst-ip */ ||
1920                                             vidx == 1 /* src-ip */)
1921                                                 keylen = is_ipv6 ?
1922                                                     sizeof(struct in6_addr):
1923                                                     sizeof(in_addr_t);
1924                                         else {
1925                                                 keylen = sizeof(key);
1926                                                 pkey = &key;
1927                                         }
1928                                         if (vidx == 0 /* dst-ip */)
1929                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1930                                                     (void *)&args->f_id.dst_ip6;
1931                                         else if (vidx == 1 /* src-ip */)
1932                                                 pkey = is_ipv4 ? (void *)&src_ip:
1933                                                     (void *)&args->f_id.src_ip6;
1934                                         else if (vidx == 6 /* dscp */) {
1935                                                 if (is_ipv4)
1936                                                         key = ip->ip_tos >> 2;
1937                                                 else {
1938                                                         key = args->f_id.flow_id6;
1939                                                         key = (key & 0x0f) << 2 |
1940                                                             (key & 0xf000) >> 14;
1941                                                 }
1942                                                 key &= 0x3f;
1943                                         } else if (vidx == 2 /* dst-port */ ||
1944                                             vidx == 3 /* src-port */) {
1945                                                 /* Skip fragments */
1946                                                 if (offset != 0)
1947                                                         break;
1948                                                 /* Skip proto without ports */
1949                                                 if (proto != IPPROTO_TCP &&
1950                                                     proto != IPPROTO_UDP &&
1951                                                     proto != IPPROTO_UDPLITE &&
1952                                                     proto != IPPROTO_SCTP)
1953                                                         break;
1954                                                 if (vidx == 2 /* dst-port */)
1955                                                         key = dst_port;
1956                                                 else
1957                                                         key = src_port;
1958                                         }
1959 #ifndef USERSPACE
1960                                         else if (vidx == 4 /* uid */ ||
1961                                             vidx == 5 /* jail */) {
1962                                                 check_uidgid(
1963                                                     (ipfw_insn_u32 *)cmd,
1964                                                     args, &ucred_lookup,
1965 #ifdef __FreeBSD__
1966                                                     &ucred_cache);
1967                                                 if (vidx == 4 /* uid */)
1968                                                         key = ucred_cache->cr_uid;
1969                                                 else if (vidx == 5 /* jail */)
1970                                                         key = ucred_cache->cr_prison->pr_id;
1971 #else /* !__FreeBSD__ */
1972                                                     (void *)&ucred_cache);
1973                                                 if (vidx == 4 /* uid */)
1974                                                         key = ucred_cache.uid;
1975                                                 else if (vidx == 5 /* jail */)
1976                                                         key = ucred_cache.xid;
1977 #endif /* !__FreeBSD__ */
1978                                         }
1979 #endif /* !USERSPACE */
1980                                         else
1981                                                 break;
1982                                         match = ipfw_lookup_table(chain,
1983                                             cmd->arg1, keylen, pkey, &vidx);
1984                                         if (!match)
1985                                                 break;
1986                                         tablearg = vidx;
1987                                         break;
1988                                 }
1989                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1990                                 /* FALLTHROUGH */
1991                         }
1992                         case O_IP_SRC_LOOKUP:
1993                         {
1994                                 void *pkey;
1995                                 uint32_t vidx;
1996                                 uint16_t keylen;
1997
1998                                 if (is_ipv4) {
1999                                         keylen = sizeof(in_addr_t);
2000                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2001                                                 pkey = &dst_ip;
2002                                         else
2003                                                 pkey = &src_ip;
2004                                 } else if (is_ipv6) {
2005                                         keylen = sizeof(struct in6_addr);
2006                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2007                                                 pkey = &args->f_id.dst_ip6;
2008                                         else
2009                                                 pkey = &args->f_id.src_ip6;
2010                                 } else
2011                                         break;
2012                                 match = ipfw_lookup_table(chain, cmd->arg1,
2013                                     keylen, pkey, &vidx);
2014                                 if (!match)
2015                                         break;
2016                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2017                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2018                                             TARG_VAL(chain, vidx, tag);
2019                                         if (!match)
2020                                                 break;
2021                                 }
2022                                 tablearg = vidx;
2023                                 break;
2024                         }
2025
2026                         case O_IP_FLOW_LOOKUP:
2027                                 {
2028                                         uint32_t v = 0;
2029                                         match = ipfw_lookup_table(chain,
2030                                             cmd->arg1, 0, &args->f_id, &v);
2031                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2032                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2033                                                     TARG_VAL(chain, v, tag);
2034                                         if (match)
2035                                                 tablearg = v;
2036                                 }
2037                                 break;
2038                         case O_IP_SRC_MASK:
2039                         case O_IP_DST_MASK:
2040                                 if (is_ipv4) {
2041                                     uint32_t a =
2042                                         (cmd->opcode == O_IP_DST_MASK) ?
2043                                             dst_ip.s_addr : src_ip.s_addr;
2044                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2045                                     int i = cmdlen-1;
2046
2047                                     for (; !match && i>0; i-= 2, p+= 2)
2048                                         match = (p[0] == (a & p[1]));
2049                                 }
2050                                 break;
2051
2052                         case O_IP_SRC_ME:
2053                                 if (is_ipv4) {
2054                                         match = in_localip(src_ip);
2055                                         break;
2056                                 }
2057 #ifdef INET6
2058                                 /* FALLTHROUGH */
2059                         case O_IP6_SRC_ME:
2060                                 match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
2061 #endif
2062                                 break;
2063
2064                         case O_IP_DST_SET:
2065                         case O_IP_SRC_SET:
2066                                 if (is_ipv4) {
2067                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2068                                         u_int32_t addr =
2069                                             cmd->opcode == O_IP_DST_SET ?
2070                                                 args->f_id.dst_ip :
2071                                                 args->f_id.src_ip;
2072
2073                                             if (addr < d[0])
2074                                                     break;
2075                                             addr -= d[0]; /* subtract base */
2076                                             match = (addr < cmd->arg1) &&
2077                                                 ( d[ 1 + (addr>>5)] &
2078                                                   (1<<(addr & 0x1f)) );
2079                                 }
2080                                 break;
2081
2082                         case O_IP_DST:
2083                                 match = is_ipv4 &&
2084                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2085                                     dst_ip.s_addr);
2086                                 break;
2087
2088                         case O_IP_DST_ME:
2089                                 if (is_ipv4) {
2090                                         match = in_localip(dst_ip);
2091                                         break;
2092                                 }
2093 #ifdef INET6
2094                                 /* FALLTHROUGH */
2095                         case O_IP6_DST_ME:
2096                                 match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
2097 #endif
2098                                 break;
2099
2100
2101                         case O_IP_SRCPORT:
2102                         case O_IP_DSTPORT:
2103                                 /*
2104                                  * offset == 0 && proto != 0 is enough
2105                                  * to guarantee that we have a
2106                                  * packet with port info.
2107                                  */
2108                                 if ((proto == IPPROTO_UDP ||
2109                                     proto == IPPROTO_UDPLITE ||
2110                                     proto == IPPROTO_TCP ||
2111                                     proto == IPPROTO_SCTP) && offset == 0) {
2112                                         u_int16_t x =
2113                                             (cmd->opcode == O_IP_SRCPORT) ?
2114                                                 src_port : dst_port ;
2115                                         u_int16_t *p =
2116                                             ((ipfw_insn_u16 *)cmd)->ports;
2117                                         int i;
2118
2119                                         for (i = cmdlen - 1; !match && i>0;
2120                                             i--, p += 2)
2121                                                 match = (x>=p[0] && x<=p[1]);
2122                                 }
2123                                 break;
2124
2125                         case O_ICMPTYPE:
2126                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2127                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2128                                 break;
2129
2130 #ifdef INET6
2131                         case O_ICMP6TYPE:
2132                                 match = is_ipv6 && offset == 0 &&
2133                                     proto==IPPROTO_ICMPV6 &&
2134                                     icmp6type_match(
2135                                         ICMP6(ulp)->icmp6_type,
2136                                         (ipfw_insn_u32 *)cmd);
2137                                 break;
2138 #endif /* INET6 */
2139
2140                         case O_IPOPT:
2141                                 match = (is_ipv4 &&
2142                                     ipopts_match(ip, cmd) );
2143                                 break;
2144
2145                         case O_IPVER:
2146                                 match = (is_ipv4 &&
2147                                     cmd->arg1 == ip->ip_v);
2148                                 break;
2149
2150                         case O_IPID:
2151                         case O_IPLEN:
2152                         case O_IPTTL:
2153                                 if (is_ipv4) {  /* only for IP packets */
2154                                     uint16_t x;
2155                                     uint16_t *p;
2156                                     int i;
2157
2158                                     if (cmd->opcode == O_IPLEN)
2159                                         x = iplen;
2160                                     else if (cmd->opcode == O_IPTTL)
2161                                         x = ip->ip_ttl;
2162                                     else /* must be IPID */
2163                                         x = ntohs(ip->ip_id);
2164                                     if (cmdlen == 1) {
2165                                         match = (cmd->arg1 == x);
2166                                         break;
2167                                     }
2168                                     /* otherwise we have ranges */
2169                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2170                                     i = cmdlen - 1;
2171                                     for (; !match && i>0; i--, p += 2)
2172                                         match = (x >= p[0] && x <= p[1]);
2173                                 }
2174                                 break;
2175
2176                         case O_IPPRECEDENCE:
2177                                 match = (is_ipv4 &&
2178                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2179                                 break;
2180
2181                         case O_IPTOS:
2182                                 match = (is_ipv4 &&
2183                                     flags_match(cmd, ip->ip_tos));
2184                                 break;
2185
2186                         case O_DSCP:
2187                             {
2188                                 uint32_t *p;
2189                                 uint16_t x;
2190
2191                                 p = ((ipfw_insn_u32 *)cmd)->d;
2192
2193                                 if (is_ipv4)
2194                                         x = ip->ip_tos >> 2;
2195                                 else if (is_ipv6) {
2196                                         uint8_t *v;
2197                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2198                                         x = (*v & 0x0F) << 2;
2199                                         v++;
2200                                         x |= *v >> 6;
2201                                 } else
2202                                         break;
2203
2204                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2205                                 if (x >= 32)
2206                                         match = *(p + 1) & (1 << (x - 32));
2207                                 else
2208                                         match = *p & (1 << x);
2209                             }
2210                                 break;
2211
2212                         case O_TCPDATALEN:
2213                                 if (proto == IPPROTO_TCP && offset == 0) {
2214                                     struct tcphdr *tcp;
2215                                     uint16_t x;
2216                                     uint16_t *p;
2217                                     int i;
2218 #ifdef INET6
2219                                     if (is_ipv6) {
2220                                             struct ip6_hdr *ip6;
2221
2222                                             ip6 = (struct ip6_hdr *)ip;
2223                                             if (ip6->ip6_plen == 0) {
2224                                                     /*
2225                                                      * Jumbo payload is not
2226                                                      * supported by this
2227                                                      * opcode.
2228                                                      */
2229                                                     break;
2230                                             }
2231                                             x = iplen - hlen;
2232                                     } else
2233 #endif /* INET6 */
2234                                             x = iplen - (ip->ip_hl << 2);
2235                                     tcp = TCP(ulp);
2236                                     x -= tcp->th_off << 2;
2237                                     if (cmdlen == 1) {
2238                                         match = (cmd->arg1 == x);
2239                                         break;
2240                                     }
2241                                     /* otherwise we have ranges */
2242                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2243                                     i = cmdlen - 1;
2244                                     for (; !match && i>0; i--, p += 2)
2245                                         match = (x >= p[0] && x <= p[1]);
2246                                 }
2247                                 break;
2248
2249                         case O_TCPFLAGS:
2250                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2251                                     flags_match(cmd, TCP(ulp)->th_flags));
2252                                 break;
2253
2254                         case O_TCPOPTS:
2255                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2256                                         PULLUP_LEN(hlen, ulp,
2257                                             (TCP(ulp)->th_off << 2));
2258                                         match = tcpopts_match(TCP(ulp), cmd);
2259                                 }
2260                                 break;
2261
2262                         case O_TCPSEQ:
2263                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2264                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2265                                         TCP(ulp)->th_seq);
2266                                 break;
2267
2268                         case O_TCPACK:
2269                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2270                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2271                                         TCP(ulp)->th_ack);
2272                                 break;
2273
2274                         case O_TCPWIN:
2275                                 if (proto == IPPROTO_TCP && offset == 0) {
2276                                     uint16_t x;
2277                                     uint16_t *p;
2278                                     int i;
2279
2280                                     x = ntohs(TCP(ulp)->th_win);
2281                                     if (cmdlen == 1) {
2282                                         match = (cmd->arg1 == x);
2283                                         break;
2284                                     }
2285                                     /* Otherwise we have ranges. */
2286                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2287                                     i = cmdlen - 1;
2288                                     for (; !match && i > 0; i--, p += 2)
2289                                         match = (x >= p[0] && x <= p[1]);
2290                                 }
2291                                 break;
2292
2293                         case O_ESTAB:
2294                                 /* reject packets which have SYN only */
2295                                 /* XXX should i also check for TH_ACK ? */
2296                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2297                                     (TCP(ulp)->th_flags &
2298                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2299                                 break;
2300
2301                         case O_ALTQ: {
2302                                 struct pf_mtag *at;
2303                                 struct m_tag *mtag;
2304                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2305
2306                                 /*
2307                                  * ALTQ uses mbuf tags from another
2308                                  * packet filtering system - pf(4).
2309                                  * We allocate a tag in its format
2310                                  * and fill it in, pretending to be pf(4).
2311                                  */
2312                                 match = 1;
2313                                 at = pf_find_mtag(m);
2314                                 if (at != NULL && at->qid != 0)
2315                                         break;
2316                                 mtag = m_tag_get(PACKET_TAG_PF,
2317                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2318                                 if (mtag == NULL) {
2319                                         /*
2320                                          * Let the packet fall back to the
2321                                          * default ALTQ.
2322                                          */
2323                                         break;
2324                                 }
2325                                 m_tag_prepend(m, mtag);
2326                                 at = (struct pf_mtag *)(mtag + 1);
2327                                 at->qid = altq->qid;
2328                                 at->hdr = ip;
2329                                 break;
2330                         }
2331
2332                         case O_LOG:
2333                                 ipfw_log(chain, f, hlen, args, m,
2334                                     oif, offset | ip6f_mf, tablearg, ip);
2335                                 match = 1;
2336                                 break;
2337
2338                         case O_PROB:
2339                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2340                                 break;
2341
2342                         case O_VERREVPATH:
2343                                 /* Outgoing packets automatically pass/match */
2344                                 match = ((oif != NULL) ||
2345                                     (m->m_pkthdr.rcvif == NULL) ||
2346                                     (
2347 #ifdef INET6
2348                                     is_ipv6 ?
2349                                         verify_path6(&(args->f_id.src_ip6),
2350                                             m->m_pkthdr.rcvif, args->f_id.fib) :
2351 #endif
2352                                     verify_path(src_ip, m->m_pkthdr.rcvif,
2353                                         args->f_id.fib)));
2354                                 break;
2355
2356                         case O_VERSRCREACH:
2357                                 /* Outgoing packets automatically pass/match */
2358                                 match = (hlen > 0 && ((oif != NULL) ||
2359 #ifdef INET6
2360                                     is_ipv6 ?
2361                                         verify_path6(&(args->f_id.src_ip6),
2362                                             NULL, args->f_id.fib) :
2363 #endif
2364                                     verify_path(src_ip, NULL, args->f_id.fib)));
2365                                 break;
2366
2367                         case O_ANTISPOOF:
2368                                 /* Outgoing packets automatically pass/match */
2369                                 if (oif == NULL && hlen > 0 &&
2370                                     (  (is_ipv4 && in_localaddr(src_ip))
2371 #ifdef INET6
2372                                     || (is_ipv6 &&
2373                                         in6_localaddr(&(args->f_id.src_ip6)))
2374 #endif
2375                                     ))
2376                                         match =
2377 #ifdef INET6
2378                                             is_ipv6 ? verify_path6(
2379                                                 &(args->f_id.src_ip6),
2380                                                 m->m_pkthdr.rcvif,
2381                                                 args->f_id.fib) :
2382 #endif
2383                                             verify_path(src_ip,
2384                                                 m->m_pkthdr.rcvif,
2385                                                 args->f_id.fib);
2386                                 else
2387                                         match = 1;
2388                                 break;
2389
2390                         case O_IPSEC:
2391                                 match = (m_tag_find(m,
2392                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2393                                 /* otherwise no match */
2394                                 break;
2395
2396 #ifdef INET6
2397                         case O_IP6_SRC:
2398                                 match = is_ipv6 &&
2399                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2400                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2401                                 break;
2402
2403                         case O_IP6_DST:
2404                                 match = is_ipv6 &&
2405                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2406                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2407                                 break;
2408                         case O_IP6_SRC_MASK:
2409                         case O_IP6_DST_MASK:
2410                                 if (is_ipv6) {
2411                                         int i = cmdlen - 1;
2412                                         struct in6_addr p;
2413                                         struct in6_addr *d =
2414                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2415
2416                                         for (; !match && i > 0; d += 2,
2417                                             i -= F_INSN_SIZE(struct in6_addr)
2418                                             * 2) {
2419                                                 p = (cmd->opcode ==
2420                                                     O_IP6_SRC_MASK) ?
2421                                                     args->f_id.src_ip6:
2422                                                     args->f_id.dst_ip6;
2423                                                 APPLY_MASK(&p, &d[1]);
2424                                                 match =
2425                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2426                                                     &p);
2427                                         }
2428                                 }
2429                                 break;
2430
2431                         case O_FLOW6ID:
2432                                 match = is_ipv6 &&
2433                                     flow6id_match(args->f_id.flow_id6,
2434                                     (ipfw_insn_u32 *) cmd);
2435                                 break;
2436
2437                         case O_EXT_HDR:
2438                                 match = is_ipv6 &&
2439                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2440                                 break;
2441
2442                         case O_IP6:
2443                                 match = is_ipv6;
2444                                 break;
2445 #endif
2446
2447                         case O_IP4:
2448                                 match = is_ipv4;
2449                                 break;
2450
2451                         case O_TAG: {
2452                                 struct m_tag *mtag;
2453                                 uint32_t tag = TARG(cmd->arg1, tag);
2454
2455                                 /* Packet is already tagged with this tag? */
2456                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2457
2458                                 /* We have `untag' action when F_NOT flag is
2459                                  * present. And we must remove this mtag from
2460                                  * mbuf and reset `match' to zero (`match' will
2461                                  * be inversed later).
2462                                  * Otherwise we should allocate new mtag and
2463                                  * push it into mbuf.
2464                                  */
2465                                 if (cmd->len & F_NOT) { /* `untag' action */
2466                                         if (mtag != NULL)
2467                                                 m_tag_delete(m, mtag);
2468                                         match = 0;
2469                                 } else {
2470                                         if (mtag == NULL) {
2471                                                 mtag = m_tag_alloc( MTAG_IPFW,
2472                                                     tag, 0, M_NOWAIT);
2473                                                 if (mtag != NULL)
2474                                                         m_tag_prepend(m, mtag);
2475                                         }
2476                                         match = 1;
2477                                 }
2478                                 break;
2479                         }
2480
2481                         case O_FIB: /* try match the specified fib */
2482                                 if (args->f_id.fib == cmd->arg1)
2483                                         match = 1;
2484                                 break;
2485
2486                         case O_SOCKARG: {
2487 #ifndef USERSPACE       /* not supported in userspace */
2488                                 struct inpcb *inp = args->inp;
2489                                 struct inpcbinfo *pi;
2490
2491                                 if (is_ipv6) /* XXX can we remove this ? */
2492                                         break;
2493
2494                                 if (proto == IPPROTO_TCP)
2495                                         pi = &V_tcbinfo;
2496                                 else if (proto == IPPROTO_UDP)
2497                                         pi = &V_udbinfo;
2498                                 else if (proto == IPPROTO_UDPLITE)
2499                                         pi = &V_ulitecbinfo;
2500                                 else
2501                                         break;
2502
2503                                 /*
2504                                  * XXXRW: so_user_cookie should almost
2505                                  * certainly be inp_user_cookie?
2506                                  */
2507
2508                                 /* For incoming packet, lookup up the
2509                                 inpcb using the src/dest ip/port tuple */
2510                                 if (inp == NULL) {
2511                                         inp = in_pcblookup(pi,
2512                                                 src_ip, htons(src_port),
2513                                                 dst_ip, htons(dst_port),
2514                                                 INPLOOKUP_RLOCKPCB, NULL);
2515                                         if (inp != NULL) {
2516                                                 tablearg =
2517                                                     inp->inp_socket->so_user_cookie;
2518                                                 if (tablearg)
2519                                                         match = 1;
2520                                                 INP_RUNLOCK(inp);
2521                                         }
2522                                 } else {
2523                                         if (inp->inp_socket) {
2524                                                 tablearg =
2525                                                     inp->inp_socket->so_user_cookie;
2526                                                 if (tablearg)
2527                                                         match = 1;
2528                                         }
2529                                 }
2530 #endif /* !USERSPACE */
2531                                 break;
2532                         }
2533
2534                         case O_TAGGED: {
2535                                 struct m_tag *mtag;
2536                                 uint32_t tag = TARG(cmd->arg1, tag);
2537
2538                                 if (cmdlen == 1) {
2539                                         match = m_tag_locate(m, MTAG_IPFW,
2540                                             tag, NULL) != NULL;
2541                                         break;
2542                                 }
2543
2544                                 /* we have ranges */
2545                                 for (mtag = m_tag_first(m);
2546                                     mtag != NULL && !match;
2547                                     mtag = m_tag_next(m, mtag)) {
2548                                         uint16_t *p;
2549                                         int i;
2550
2551                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2552                                                 continue;
2553
2554                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2555                                         i = cmdlen - 1;
2556                                         for(; !match && i > 0; i--, p += 2)
2557                                                 match =
2558                                                     mtag->m_tag_id >= p[0] &&
2559                                                     mtag->m_tag_id <= p[1];
2560                                 }
2561                                 break;
2562                         }
2563
2564                         /*
2565                          * The second set of opcodes represents 'actions',
2566                          * i.e. the terminal part of a rule once the packet
2567                          * matches all previous patterns.
2568                          * Typically there is only one action for each rule,
2569                          * and the opcode is stored at the end of the rule
2570                          * (but there are exceptions -- see below).
2571                          *
2572                          * In general, here we set retval and terminate the
2573                          * outer loop (would be a 'break 3' in some language,
2574                          * but we need to set l=0, done=1)
2575                          *
2576                          * Exceptions:
2577                          * O_COUNT and O_SKIPTO actions:
2578                          *   instead of terminating, we jump to the next rule
2579                          *   (setting l=0), or to the SKIPTO target (setting
2580                          *   f/f_len, cmd and l as needed), respectively.
2581                          *
2582                          * O_TAG, O_LOG and O_ALTQ action parameters:
2583                          *   perform some action and set match = 1;
2584                          *
2585                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2586                          *   not real 'actions', and are stored right
2587                          *   before the 'action' part of the rule (one
2588                          *   exception is O_SKIP_ACTION which could be
2589                          *   between these opcodes and 'action' one).
2590                          *   These opcodes try to install an entry in the
2591                          *   state tables; if successful, we continue with
2592                          *   the next opcode (match=1; break;), otherwise
2593                          *   the packet must be dropped (set retval,
2594                          *   break loops with l=0, done=1)
2595                          *
2596                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2597                          *   cause a lookup of the state table, and a jump
2598                          *   to the 'action' part of the parent rule
2599                          *   if an entry is found, or
2600                          *   (CHECK_STATE only) a jump to the next rule if
2601                          *   the entry is not found.
2602                          *   The result of the lookup is cached so that
2603                          *   further instances of these opcodes become NOPs.
2604                          *   The jump to the next rule is done by setting
2605                          *   l=0, cmdlen=0.
2606                          *
2607                          * O_SKIP_ACTION: this opcode is not a real 'action'
2608                          *  either, and is stored right before the 'action'
2609                          *  part of the rule, right after the O_KEEP_STATE
2610                          *  opcode. It causes match failure so the real
2611                          *  'action' could be executed only if the rule
2612                          *  is checked via dynamic rule from the state
2613                          *  table, as in such case execution starts
2614                          *  from the true 'action' opcode directly.
2615                          *
2616                          */
2617                         case O_LIMIT:
2618                         case O_KEEP_STATE:
2619                                 if (ipfw_dyn_install_state(chain, f,
2620                                     (ipfw_insn_limit *)cmd, args, ulp,
2621                                     pktlen, &dyn_info, tablearg)) {
2622                                         /* error or limit violation */
2623                                         retval = IP_FW_DENY;
2624                                         l = 0;  /* exit inner loop */
2625                                         done = 1; /* exit outer loop */
2626                                 }
2627                                 match = 1;
2628                                 break;
2629
2630                         case O_PROBE_STATE:
2631                         case O_CHECK_STATE:
2632                                 /*
2633                                  * dynamic rules are checked at the first
2634                                  * keep-state or check-state occurrence,
2635                                  * with the result being stored in dyn_info.
2636                                  * The compiler introduces a PROBE_STATE
2637                                  * instruction for us when we have a
2638                                  * KEEP_STATE (because PROBE_STATE needs
2639                                  * to be run first).
2640                                  */
2641                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2642                                     (q = ipfw_dyn_lookup_state(args, ulp,
2643                                     pktlen, cmd, &dyn_info)) != NULL) {
2644                                         /*
2645                                          * Found dynamic entry, jump to the
2646                                          * 'action' part of the parent rule
2647                                          * by setting f, cmd, l and clearing
2648                                          * cmdlen.
2649                                          */
2650                                         f = q;
2651                                         f_pos = dyn_info.f_pos;
2652                                         cmd = ACTION_PTR(f);
2653                                         l = f->cmd_len - f->act_ofs;
2654                                         cmdlen = 0;
2655                                         match = 1;
2656                                         break;
2657                                 }
2658                                 /*
2659                                  * Dynamic entry not found. If CHECK_STATE,
2660                                  * skip to next rule, if PROBE_STATE just
2661                                  * ignore and continue with next opcode.
2662                                  */
2663                                 if (cmd->opcode == O_CHECK_STATE)
2664                                         l = 0;  /* exit inner loop */
2665                                 match = 1;
2666                                 break;
2667
2668                         case O_SKIP_ACTION:
2669                                 match = 0;      /* skip to the next rule */
2670                                 l = 0;          /* exit inner loop */
2671                                 break;
2672
2673                         case O_ACCEPT:
2674                                 retval = 0;     /* accept */
2675                                 l = 0;          /* exit inner loop */
2676                                 done = 1;       /* exit outer loop */
2677                                 break;
2678
2679                         case O_PIPE:
2680                         case O_QUEUE:
2681                                 set_match(args, f_pos, chain);
2682                                 args->rule.info = TARG(cmd->arg1, pipe);
2683                                 if (cmd->opcode == O_PIPE)
2684                                         args->rule.info |= IPFW_IS_PIPE;
2685                                 if (V_fw_one_pass)
2686                                         args->rule.info |= IPFW_ONEPASS;
2687                                 retval = IP_FW_DUMMYNET;
2688                                 l = 0;          /* exit inner loop */
2689                                 done = 1;       /* exit outer loop */
2690                                 break;
2691
2692                         case O_DIVERT:
2693                         case O_TEE:
2694                                 if (args->eh) /* not on layer 2 */
2695                                     break;
2696                                 /* otherwise this is terminal */
2697                                 l = 0;          /* exit inner loop */
2698                                 done = 1;       /* exit outer loop */
2699                                 retval = (cmd->opcode == O_DIVERT) ?
2700                                         IP_FW_DIVERT : IP_FW_TEE;
2701                                 set_match(args, f_pos, chain);
2702                                 args->rule.info = TARG(cmd->arg1, divert);
2703                                 break;
2704
2705                         case O_COUNT:
2706                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2707                                 l = 0;          /* exit inner loop */
2708                                 break;
2709
2710                         case O_SKIPTO:
2711                             IPFW_INC_RULE_COUNTER(f, pktlen);
2712                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2713                             /*
2714                              * Skip disabled rules, and re-enter
2715                              * the inner loop with the correct
2716                              * f_pos, f, l and cmd.
2717                              * Also clear cmdlen and skip_or
2718                              */
2719                             for (; f_pos < chain->n_rules - 1 &&
2720                                     (V_set_disable &
2721                                      (1 << chain->map[f_pos]->set));
2722                                     f_pos++)
2723                                 ;
2724                             /* Re-enter the inner loop at the skipto rule. */
2725                             f = chain->map[f_pos];
2726                             l = f->cmd_len;
2727                             cmd = f->cmd;
2728                             match = 1;
2729                             cmdlen = 0;
2730                             skip_or = 0;
2731                             continue;
2732                             break;      /* not reached */
2733
2734                         case O_CALLRETURN: {
2735                                 /*
2736                                  * Implementation of `subroutine' call/return,
2737                                  * in the stack carried in an mbuf tag. This
2738                                  * is different from `skipto' in that any call
2739                                  * address is possible (`skipto' must prevent
2740                                  * backward jumps to avoid endless loops).
2741                                  * We have `return' action when F_NOT flag is
2742                                  * present. The `m_tag_id' field is used as
2743                                  * stack pointer.
2744                                  */
2745                                 struct m_tag *mtag;
2746                                 uint16_t jmpto, *stack;
2747
2748 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2749 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2750                                 /*
2751                                  * Hand-rolled version of m_tag_locate() with
2752                                  * wildcard `type'.
2753                                  * If not already tagged, allocate new tag.
2754                                  */
2755                                 mtag = m_tag_first(m);
2756                                 while (mtag != NULL) {
2757                                         if (mtag->m_tag_cookie ==
2758                                             MTAG_IPFW_CALL)
2759                                                 break;
2760                                         mtag = m_tag_next(m, mtag);
2761                                 }
2762                                 if (mtag == NULL && IS_CALL) {
2763                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2764                                             IPFW_CALLSTACK_SIZE *
2765                                             sizeof(uint16_t), M_NOWAIT);
2766                                         if (mtag != NULL)
2767                                                 m_tag_prepend(m, mtag);
2768                                 }
2769
2770                                 /*
2771                                  * On error both `call' and `return' just
2772                                  * continue with next rule.
2773                                  */
2774                                 if (IS_RETURN && (mtag == NULL ||
2775                                     mtag->m_tag_id == 0)) {
2776                                         l = 0;          /* exit inner loop */
2777                                         break;
2778                                 }
2779                                 if (IS_CALL && (mtag == NULL ||
2780                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2781                                         printf("ipfw: call stack error, "
2782                                             "go to next rule\n");
2783                                         l = 0;          /* exit inner loop */
2784                                         break;
2785                                 }
2786
2787                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2788                                 stack = (uint16_t *)(mtag + 1);
2789
2790                                 /*
2791                                  * The `call' action may use cached f_pos
2792                                  * (in f->next_rule), whose version is written
2793                                  * in f->next_rule.
2794                                  * The `return' action, however, doesn't have
2795                                  * fixed jump address in cmd->arg1 and can't use
2796                                  * cache.
2797                                  */
2798                                 if (IS_CALL) {
2799                                         stack[mtag->m_tag_id] = f->rulenum;
2800                                         mtag->m_tag_id++;
2801                                         f_pos = JUMP(chain, f, cmd->arg1,
2802                                             tablearg, 1);
2803                                 } else {        /* `return' action */
2804                                         mtag->m_tag_id--;
2805                                         jmpto = stack[mtag->m_tag_id] + 1;
2806                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2807                                 }
2808
2809                                 /*
2810                                  * Skip disabled rules, and re-enter
2811                                  * the inner loop with the correct
2812                                  * f_pos, f, l and cmd.
2813                                  * Also clear cmdlen and skip_or
2814                                  */
2815                                 for (; f_pos < chain->n_rules - 1 &&
2816                                     (V_set_disable &
2817                                     (1 << chain->map[f_pos]->set)); f_pos++)
2818                                         ;
2819                                 /* Re-enter the inner loop at the dest rule. */
2820                                 f = chain->map[f_pos];
2821                                 l = f->cmd_len;
2822                                 cmd = f->cmd;
2823                                 cmdlen = 0;
2824                                 skip_or = 0;
2825                                 continue;
2826                                 break;  /* NOTREACHED */
2827                         }
2828 #undef IS_CALL
2829 #undef IS_RETURN
2830
2831                         case O_REJECT:
2832                                 /*
2833                                  * Drop the packet and send a reject notice
2834                                  * if the packet is not ICMP (or is an ICMP
2835                                  * query), and it is not multicast/broadcast.
2836                                  */
2837                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2838                                     (proto != IPPROTO_ICMP ||
2839                                      is_icmp_query(ICMP(ulp))) &&
2840                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2841                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2842                                         send_reject(args, cmd->arg1, iplen, ip);
2843                                         m = args->m;
2844                                 }
2845                                 /* FALLTHROUGH */
2846 #ifdef INET6
2847                         case O_UNREACH6:
2848                                 if (hlen > 0 && is_ipv6 &&
2849                                     ((offset & IP6F_OFF_MASK) == 0) &&
2850                                     (proto != IPPROTO_ICMPV6 ||
2851                                      (is_icmp6_query(icmp6_type) == 1)) &&
2852                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2853                                     !IN6_IS_ADDR_MULTICAST(
2854                                         &args->f_id.dst_ip6)) {
2855                                         send_reject6(args,
2856                                             cmd->opcode == O_REJECT ?
2857                                             map_icmp_unreach(cmd->arg1):
2858                                             cmd->arg1, hlen,
2859                                             (struct ip6_hdr *)ip);
2860                                         m = args->m;
2861                                 }
2862                                 /* FALLTHROUGH */
2863 #endif
2864                         case O_DENY:
2865                                 retval = IP_FW_DENY;
2866                                 l = 0;          /* exit inner loop */
2867                                 done = 1;       /* exit outer loop */
2868                                 break;
2869
2870                         case O_FORWARD_IP:
2871                                 if (args->eh)   /* not valid on layer2 pkts */
2872                                         break;
2873                                 if (q != f ||
2874                                     dyn_info.direction == MATCH_FORWARD) {
2875                                     struct sockaddr_in *sa;
2876
2877                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2878                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2879 #ifdef INET6
2880                                         /*
2881                                          * We use O_FORWARD_IP opcode for
2882                                          * fwd rule with tablearg, but tables
2883                                          * now support IPv6 addresses. And
2884                                          * when we are inspecting IPv6 packet,
2885                                          * we can use nh6 field from
2886                                          * table_value as next_hop6 address.
2887                                          */
2888                                         if (is_ipv6) {
2889                                                 struct sockaddr_in6 *sa6;
2890
2891                                                 sa6 = args->next_hop6 =
2892                                                     &args->hopstore6;
2893                                                 sa6->sin6_family = AF_INET6;
2894                                                 sa6->sin6_len = sizeof(*sa6);
2895                                                 sa6->sin6_addr = TARG_VAL(
2896                                                     chain, tablearg, nh6);
2897                                                 sa6->sin6_port = sa->sin_port;
2898                                                 /*
2899                                                  * Set sin6_scope_id only for
2900                                                  * link-local unicast addresses.
2901                                                  */
2902                                                 if (IN6_IS_ADDR_LINKLOCAL(
2903                                                     &sa6->sin6_addr))
2904                                                         sa6->sin6_scope_id =
2905                                                             TARG_VAL(chain,
2906                                                                 tablearg,
2907                                                                 zoneid);
2908                                         } else
2909 #endif
2910                                         {
2911                                                 args->hopstore.sin_port =
2912                                                     sa->sin_port;
2913                                                 sa = args->next_hop =
2914                                                     &args->hopstore;
2915                                                 sa->sin_family = AF_INET;
2916                                                 sa->sin_len = sizeof(*sa);
2917                                                 sa->sin_addr.s_addr = htonl(
2918                                                     TARG_VAL(chain, tablearg,
2919                                                     nh4));
2920                                         }
2921                                     } else {
2922                                         args->next_hop = sa;
2923                                     }
2924                                 }
2925                                 retval = IP_FW_PASS;
2926                                 l = 0;          /* exit inner loop */
2927                                 done = 1;       /* exit outer loop */
2928                                 break;
2929
2930 #ifdef INET6
2931                         case O_FORWARD_IP6:
2932                                 if (args->eh)   /* not valid on layer2 pkts */
2933                                         break;
2934                                 if (q != f ||
2935                                     dyn_info.direction == MATCH_FORWARD) {
2936                                         struct sockaddr_in6 *sin6;
2937
2938                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2939                                         args->next_hop6 = sin6;
2940                                 }
2941                                 retval = IP_FW_PASS;
2942                                 l = 0;          /* exit inner loop */
2943                                 done = 1;       /* exit outer loop */
2944                                 break;
2945 #endif
2946
2947                         case O_NETGRAPH:
2948                         case O_NGTEE:
2949                                 set_match(args, f_pos, chain);
2950                                 args->rule.info = TARG(cmd->arg1, netgraph);
2951                                 if (V_fw_one_pass)
2952                                         args->rule.info |= IPFW_ONEPASS;
2953                                 retval = (cmd->opcode == O_NETGRAPH) ?
2954                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2955                                 l = 0;          /* exit inner loop */
2956                                 done = 1;       /* exit outer loop */
2957                                 break;
2958
2959                         case O_SETFIB: {
2960                                 uint32_t fib;
2961
2962                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2963                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2964                                 if (fib >= rt_numfibs)
2965                                         fib = 0;
2966                                 M_SETFIB(m, fib);
2967                                 args->f_id.fib = fib;
2968                                 l = 0;          /* exit inner loop */
2969                                 break;
2970                         }
2971
2972                         case O_SETDSCP: {
2973                                 uint16_t code;
2974
2975                                 code = TARG(cmd->arg1, dscp) & 0x3F;
2976                                 l = 0;          /* exit inner loop */
2977                                 if (is_ipv4) {
2978                                         uint16_t old;
2979
2980                                         old = *(uint16_t *)ip;
2981                                         ip->ip_tos = (code << 2) |
2982                                             (ip->ip_tos & 0x03);
2983                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
2984                                             old, *(uint16_t *)ip);
2985                                 } else if (is_ipv6) {
2986                                         uint8_t *v;
2987
2988                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2989                                         *v = (*v & 0xF0) | (code >> 2);
2990                                         v++;
2991                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
2992                                 } else
2993                                         break;
2994
2995                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2996                                 break;
2997                         }
2998
2999                         case O_NAT:
3000                                 l = 0;          /* exit inner loop */
3001                                 done = 1;       /* exit outer loop */
3002                                 /*
3003                                  * Ensure that we do not invoke NAT handler for
3004                                  * non IPv4 packets. Libalias expects only IPv4.
3005                                  */
3006                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3007                                     retval = IP_FW_DENY;
3008                                     break;
3009                                 }
3010
3011                                 struct cfg_nat *t;
3012                                 int nat_id;
3013
3014                                 set_match(args, f_pos, chain);
3015                                 /* Check if this is 'global' nat rule */
3016                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3017                                         retval = ipfw_nat_ptr(args, NULL, m);
3018                                         break;
3019                                 }
3020                                 t = ((ipfw_insn_nat *)cmd)->nat;
3021                                 if (t == NULL) {
3022                                         nat_id = TARG(cmd->arg1, nat);
3023                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3024
3025                                         if (t == NULL) {
3026                                             retval = IP_FW_DENY;
3027                                             break;
3028                                         }
3029                                         if (cmd->arg1 != IP_FW_TARG)
3030                                             ((ipfw_insn_nat *)cmd)->nat = t;
3031                                 }
3032                                 retval = ipfw_nat_ptr(args, t, m);
3033                                 break;
3034
3035                         case O_REASS: {
3036                                 int ip_off;
3037
3038                                 l = 0;  /* in any case exit inner loop */
3039                                 if (is_ipv6) /* IPv6 is not supported yet */
3040                                         break;
3041                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3042                                 ip_off = ntohs(ip->ip_off);
3043
3044                                 /* if not fragmented, go to next rule */
3045                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3046                                     break;
3047
3048                                 args->m = m = ip_reass(m);
3049
3050                                 /*
3051                                  * do IP header checksum fixup.
3052                                  */
3053                                 if (m == NULL) { /* fragment got swallowed */
3054                                     retval = IP_FW_DENY;
3055                                 } else { /* good, packet complete */
3056                                     int hlen;
3057
3058                                     ip = mtod(m, struct ip *);
3059                                     hlen = ip->ip_hl << 2;
3060                                     ip->ip_sum = 0;
3061                                     if (hlen == sizeof(struct ip))
3062                                         ip->ip_sum = in_cksum_hdr(ip);
3063                                     else
3064                                         ip->ip_sum = in_cksum(m, hlen);
3065                                     retval = IP_FW_REASS;
3066                                     set_match(args, f_pos, chain);
3067                                 }
3068                                 done = 1;       /* exit outer loop */
3069                                 break;
3070                         }
3071                         case O_EXTERNAL_ACTION:
3072                                 l = 0; /* in any case exit inner loop */
3073                                 retval = ipfw_run_eaction(chain, args,
3074                                     cmd, &done);
3075                                 /*
3076                                  * If both @retval and @done are zero,
3077                                  * consider this as rule matching and
3078                                  * update counters.
3079                                  */
3080                                 if (retval == 0 && done == 0) {
3081                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3082                                         /*
3083                                          * Reset the result of the last
3084                                          * dynamic state lookup.
3085                                          * External action can change
3086                                          * @args content, and it may be
3087                                          * used for new state lookup later.
3088                                          */
3089                                         DYN_INFO_INIT(&dyn_info);
3090                                 }
3091                                 break;
3092
3093                         default:
3094                                 panic("-- unknown opcode %d\n", cmd->opcode);
3095                         } /* end of switch() on opcodes */
3096                         /*
3097                          * if we get here with l=0, then match is irrelevant.
3098                          */
3099
3100                         if (cmd->len & F_NOT)
3101                                 match = !match;
3102
3103                         if (match) {
3104                                 if (cmd->len & F_OR)
3105                                         skip_or = 1;
3106                         } else {
3107                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3108                                         break;          /* try next rule    */
3109                         }
3110
3111                 }       /* end of inner loop, scan opcodes */
3112 #undef PULLUP_LEN
3113
3114                 if (done)
3115                         break;
3116
3117 /* next_rule:; */       /* try next rule                */
3118
3119         }               /* end of outer for, scan rules */
3120
3121         if (done) {
3122                 struct ip_fw *rule = chain->map[f_pos];
3123                 /* Update statistics */
3124                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3125         } else {
3126                 retval = IP_FW_DENY;
3127                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3128         }
3129         IPFW_PF_RUNLOCK(chain);
3130 #ifdef __FreeBSD__
3131         if (ucred_cache != NULL)
3132                 crfree(ucred_cache);
3133 #endif
3134         return (retval);
3135
3136 pullup_failed:
3137         if (V_fw_verbose)
3138                 printf("ipfw: pullup failed\n");
3139         return (IP_FW_DENY);
3140 }
3141
3142 /*
3143  * Set maximum number of tables that can be used in given VNET ipfw instance.
3144  */
3145 #ifdef SYSCTL_NODE
3146 static int
3147 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3148 {
3149         int error;
3150         unsigned int ntables;
3151
3152         ntables = V_fw_tables_max;
3153
3154         error = sysctl_handle_int(oidp, &ntables, 0, req);
3155         /* Read operation or some error */
3156         if ((error != 0) || (req->newptr == NULL))
3157                 return (error);
3158
3159         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3160 }
3161
3162 /*
3163  * Switches table namespace between global and per-set.
3164  */
3165 static int
3166 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3167 {
3168         int error;
3169         unsigned int sets;
3170
3171         sets = V_fw_tables_sets;
3172
3173         error = sysctl_handle_int(oidp, &sets, 0, req);
3174         /* Read operation or some error */
3175         if ((error != 0) || (req->newptr == NULL))
3176                 return (error);
3177
3178         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3179 }
3180 #endif
3181
3182 /*
3183  * Module and VNET glue
3184  */
3185
3186 /*
3187  * Stuff that must be initialised only on boot or module load
3188  */
3189 static int
3190 ipfw_init(void)
3191 {
3192         int error = 0;
3193
3194         /*
3195          * Only print out this stuff the first time around,
3196          * when called from the sysinit code.
3197          */
3198         printf("ipfw2 "
3199 #ifdef INET6
3200                 "(+ipv6) "
3201 #endif
3202                 "initialized, divert %s, nat %s, "
3203                 "default to %s, logging ",
3204 #ifdef IPDIVERT
3205                 "enabled",
3206 #else
3207                 "loadable",
3208 #endif
3209 #ifdef IPFIREWALL_NAT
3210                 "enabled",
3211 #else
3212                 "loadable",
3213 #endif
3214                 default_to_accept ? "accept" : "deny");
3215
3216         /*
3217          * Note: V_xxx variables can be accessed here but the vnet specific
3218          * initializer may not have been called yet for the VIMAGE case.
3219          * Tuneables will have been processed. We will print out values for
3220          * the default vnet.
3221          * XXX This should all be rationalized AFTER 8.0
3222          */
3223         if (V_fw_verbose == 0)
3224                 printf("disabled\n");
3225         else if (V_verbose_limit == 0)
3226                 printf("unlimited\n");
3227         else
3228                 printf("limited to %d packets/entry by default\n",
3229                     V_verbose_limit);
3230
3231         /* Check user-supplied table count for validness */
3232         if (default_fw_tables > IPFW_TABLES_MAX)
3233           default_fw_tables = IPFW_TABLES_MAX;
3234
3235         ipfw_init_sopt_handler();
3236         ipfw_init_obj_rewriter();
3237         ipfw_iface_init();
3238         return (error);
3239 }
3240
3241 /*
3242  * Called for the removal of the last instance only on module unload.
3243  */
3244 static void
3245 ipfw_destroy(void)
3246 {
3247
3248         ipfw_iface_destroy();
3249         ipfw_destroy_sopt_handler();
3250         ipfw_destroy_obj_rewriter();
3251         printf("IP firewall unloaded\n");
3252 }
3253
3254 /*
3255  * Stuff that must be initialized for every instance
3256  * (including the first of course).
3257  */
3258 static int
3259 vnet_ipfw_init(const void *unused)
3260 {
3261         int error, first;
3262         struct ip_fw *rule = NULL;
3263         struct ip_fw_chain *chain;
3264
3265         chain = &V_layer3_chain;
3266
3267         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3268
3269         /* First set up some values that are compile time options */
3270         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3271         V_fw_deny_unknown_exthdrs = 1;
3272 #ifdef IPFIREWALL_VERBOSE
3273         V_fw_verbose = 1;
3274 #endif
3275 #ifdef IPFIREWALL_VERBOSE_LIMIT
3276         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3277 #endif
3278 #ifdef IPFIREWALL_NAT
3279         LIST_INIT(&chain->nat);
3280 #endif
3281
3282         /* Init shared services hash table */
3283         ipfw_init_srv(chain);
3284
3285         ipfw_init_counters();
3286         /* Set initial number of tables */
3287         V_fw_tables_max = default_fw_tables;
3288         error = ipfw_init_tables(chain, first);
3289         if (error) {
3290                 printf("ipfw2: setting up tables failed\n");
3291                 free(chain->map, M_IPFW);
3292                 free(rule, M_IPFW);
3293                 return (ENOSPC);
3294         }
3295
3296         IPFW_LOCK_INIT(chain);
3297
3298         /* fill and insert the default rule */
3299         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3300         rule->cmd_len = 1;
3301         rule->cmd[0].len = 1;
3302         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3303         chain->default_rule = rule;
3304         ipfw_add_protected_rule(chain, rule, 0);
3305
3306         ipfw_dyn_init(chain);
3307         ipfw_eaction_init(chain, first);
3308 #ifdef LINEAR_SKIPTO
3309         ipfw_init_skipto_cache(chain);
3310 #endif
3311         ipfw_bpf_init(first);
3312
3313         /* First set up some values that are compile time options */
3314         V_ipfw_vnet_ready = 1;          /* Open for business */
3315
3316         /*
3317          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3318          * Even if the latter two fail we still keep the module alive
3319          * because the sockopt and layer2 paths are still useful.
3320          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3321          * so we can ignore the exact return value and just set a flag.
3322          *
3323          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3324          * changes in the underlying (per-vnet) variables trigger
3325          * immediate hook()/unhook() calls.
3326          * In layer2 we have the same behaviour, except that V_ether_ipfw
3327          * is checked on each packet because there are no pfil hooks.
3328          */
3329         V_ip_fw_ctl_ptr = ipfw_ctl3;
3330         error = ipfw_attach_hooks(1);
3331         return (error);
3332 }
3333
3334 /*
3335  * Called for the removal of each instance.
3336  */
3337 static int
3338 vnet_ipfw_uninit(const void *unused)
3339 {
3340         struct ip_fw *reap;
3341         struct ip_fw_chain *chain = &V_layer3_chain;
3342         int i, last;
3343
3344         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3345         /*
3346          * disconnect from ipv4, ipv6, layer2 and sockopt.
3347          * Then grab, release and grab again the WLOCK so we make
3348          * sure the update is propagated and nobody will be in.
3349          */
3350         (void)ipfw_attach_hooks(0 /* detach */);
3351         V_ip_fw_ctl_ptr = NULL;
3352
3353         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3354
3355         IPFW_UH_WLOCK(chain);
3356         IPFW_UH_WUNLOCK(chain);
3357
3358         ipfw_dyn_uninit(0);     /* run the callout_drain */
3359
3360         IPFW_UH_WLOCK(chain);
3361
3362         reap = NULL;
3363         IPFW_WLOCK(chain);
3364         for (i = 0; i < chain->n_rules; i++)
3365                 ipfw_reap_add(chain, &reap, chain->map[i]);
3366         free(chain->map, M_IPFW);
3367 #ifdef LINEAR_SKIPTO
3368         ipfw_destroy_skipto_cache(chain);
3369 #endif
3370         IPFW_WUNLOCK(chain);
3371         IPFW_UH_WUNLOCK(chain);
3372         ipfw_destroy_tables(chain, last);
3373         ipfw_eaction_uninit(chain, last);
3374         if (reap != NULL)
3375                 ipfw_reap_rules(reap);
3376         vnet_ipfw_iface_destroy(chain);
3377         ipfw_destroy_srv(chain);
3378         IPFW_LOCK_DESTROY(chain);
3379         ipfw_dyn_uninit(1);     /* free the remaining parts */
3380         ipfw_destroy_counters();
3381         ipfw_bpf_uninit(last);
3382         return (0);
3383 }
3384
3385 /*
3386  * Module event handler.
3387  * In general we have the choice of handling most of these events by the
3388  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3389  * use the SYSINIT handlers as they are more capable of expressing the
3390  * flow of control during module and vnet operations, so this is just
3391  * a skeleton. Note there is no SYSINIT equivalent of the module
3392  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3393  */
3394 static int
3395 ipfw_modevent(module_t mod, int type, void *unused)
3396 {
3397         int err = 0;
3398
3399         switch (type) {
3400         case MOD_LOAD:
3401                 /* Called once at module load or
3402                  * system boot if compiled in. */
3403                 break;
3404         case MOD_QUIESCE:
3405                 /* Called before unload. May veto unloading. */
3406                 break;
3407         case MOD_UNLOAD:
3408                 /* Called during unload. */
3409                 break;
3410         case MOD_SHUTDOWN:
3411                 /* Called during system shutdown. */
3412                 break;
3413         default:
3414                 err = EOPNOTSUPP;
3415                 break;
3416         }
3417         return err;
3418 }
3419
3420 static moduledata_t ipfwmod = {
3421         "ipfw",
3422         ipfw_modevent,
3423         0
3424 };
3425
3426 /* Define startup order. */
3427 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3428 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3429 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3430 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3431
3432 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3433 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3434 MODULE_VERSION(ipfw, 3);
3435 /* should declare some dependencies here */
3436
3437 /*
3438  * Starting up. Done in order after ipfwmod() has been called.
3439  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3440  */
3441 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3442             ipfw_init, NULL);
3443 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3444             vnet_ipfw_init, NULL);
3445
3446 /*
3447  * Closing up shop. These are done in REVERSE ORDER, but still
3448  * after ipfwmod() has been called. Not called on reboot.
3449  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3450  * or when the module is unloaded.
3451  */
3452 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3453             ipfw_destroy, NULL);
3454 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3455             vnet_ipfw_uninit, NULL);
3456 /* end of file */