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         int lookupflags;
1084         int match;
1085
1086         id = &args->f_id;
1087         inp = args->inp;
1088
1089         /*
1090          * Check to see if the UDP or TCP stack supplied us with
1091          * the PCB. If so, rather then holding a lock and looking
1092          * up the PCB, we can use the one that was supplied.
1093          */
1094         if (inp && *ugid_lookupp == 0) {
1095                 INP_LOCK_ASSERT(inp);
1096                 if (inp->inp_socket != NULL) {
1097                         *uc = crhold(inp->inp_cred);
1098                         *ugid_lookupp = 1;
1099                 } else
1100                         *ugid_lookupp = -1;
1101         }
1102         /*
1103          * If we have already been here and the packet has no
1104          * PCB entry associated with it, then we can safely
1105          * assume that this is a no match.
1106          */
1107         if (*ugid_lookupp == -1)
1108                 return (0);
1109         if (id->proto == IPPROTO_TCP) {
1110                 lookupflags = 0;
1111                 pi = &V_tcbinfo;
1112         } else if (id->proto == IPPROTO_UDP) {
1113                 lookupflags = INPLOOKUP_WILDCARD;
1114                 pi = &V_udbinfo;
1115         } else if (id->proto == IPPROTO_UDPLITE) {
1116                 lookupflags = INPLOOKUP_WILDCARD;
1117                 pi = &V_ulitecbinfo;
1118         } else
1119                 return 0;
1120         lookupflags |= INPLOOKUP_RLOCKPCB;
1121         match = 0;
1122         if (*ugid_lookupp == 0) {
1123                 if (id->addr_type == 6) {
1124 #ifdef INET6
1125                         if (args->flags & IPFW_ARGS_IN)
1126                                 pcb = in6_pcblookup_mbuf(pi,
1127                                     &id->src_ip6, htons(id->src_port),
1128                                     &id->dst_ip6, htons(id->dst_port),
1129                                     lookupflags, NULL, args->m);
1130                         else
1131                                 pcb = in6_pcblookup_mbuf(pi,
1132                                     &id->dst_ip6, htons(id->dst_port),
1133                                     &id->src_ip6, htons(id->src_port),
1134                                     lookupflags, args->ifp, args->m);
1135 #else
1136                         *ugid_lookupp = -1;
1137                         return (0);
1138 #endif
1139                 } else {
1140                         src_ip.s_addr = htonl(id->src_ip);
1141                         dst_ip.s_addr = htonl(id->dst_ip);
1142                         if (args->flags & IPFW_ARGS_IN)
1143                                 pcb = in_pcblookup_mbuf(pi,
1144                                     src_ip, htons(id->src_port),
1145                                     dst_ip, htons(id->dst_port),
1146                                     lookupflags, NULL, args->m);
1147                         else
1148                                 pcb = in_pcblookup_mbuf(pi,
1149                                     dst_ip, htons(id->dst_port),
1150                                     src_ip, htons(id->src_port),
1151                                     lookupflags, args->ifp, args->m);
1152                 }
1153                 if (pcb != NULL) {
1154                         INP_RLOCK_ASSERT(pcb);
1155                         *uc = crhold(pcb->inp_cred);
1156                         *ugid_lookupp = 1;
1157                         INP_RUNLOCK(pcb);
1158                 }
1159                 if (*ugid_lookupp == 0) {
1160                         /*
1161                          * We tried and failed, set the variable to -1
1162                          * so we will not try again on this packet.
1163                          */
1164                         *ugid_lookupp = -1;
1165                         return (0);
1166                 }
1167         }
1168         if (insn->o.opcode == O_UID)
1169                 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1170         else if (insn->o.opcode == O_GID)
1171                 match = groupmember((gid_t)insn->d[0], *uc);
1172         else if (insn->o.opcode == O_JAIL)
1173                 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1174         return (match);
1175 #endif /* __FreeBSD__ */
1176 #endif /* not supported in userspace */
1177 }
1178
1179 /*
1180  * Helper function to set args with info on the rule after the matching
1181  * one. slot is precise, whereas we guess rule_id as they are
1182  * assigned sequentially.
1183  */
1184 static inline void
1185 set_match(struct ip_fw_args *args, int slot,
1186         struct ip_fw_chain *chain)
1187 {
1188         args->rule.chain_id = chain->id;
1189         args->rule.slot = slot + 1; /* we use 0 as a marker */
1190         args->rule.rule_id = 1 + chain->map[slot]->id;
1191         args->rule.rulenum = chain->map[slot]->rulenum;
1192         args->flags |= IPFW_ARGS_REF;
1193 }
1194
1195 #ifndef LINEAR_SKIPTO
1196 /*
1197  * Helper function to enable cached rule lookups using
1198  * cached_id and cached_pos fields in ipfw rule.
1199  */
1200 static int
1201 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1202     int tablearg, int jump_backwards)
1203 {
1204         int f_pos;
1205
1206         /* If possible use cached f_pos (in f->cached_pos),
1207          * whose version is written in f->cached_id
1208          * (horrible hacks to avoid changing the ABI).
1209          */
1210         if (num != IP_FW_TARG && f->cached_id == chain->id)
1211                 f_pos = f->cached_pos;
1212         else {
1213                 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1214                 /* make sure we do not jump backward */
1215                 if (jump_backwards == 0 && i <= f->rulenum)
1216                         i = f->rulenum + 1;
1217                 if (chain->idxmap != NULL)
1218                         f_pos = chain->idxmap[i];
1219                 else
1220                         f_pos = ipfw_find_rule(chain, i, 0);
1221                 /* update the cache */
1222                 if (num != IP_FW_TARG) {
1223                         f->cached_id = chain->id;
1224                         f->cached_pos = f_pos;
1225                 }
1226         }
1227
1228         return (f_pos);
1229 }
1230 #else
1231 /*
1232  * Helper function to enable real fast rule lookups.
1233  */
1234 static int
1235 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1236     int tablearg, int jump_backwards)
1237 {
1238         int f_pos;
1239
1240         num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1241         /* make sure we do not jump backward */
1242         if (jump_backwards == 0 && num <= f->rulenum)
1243                 num = f->rulenum + 1;
1244         f_pos = chain->idxmap[num];
1245
1246         return (f_pos);
1247 }
1248 #endif
1249
1250 #define TARG(k, f)      IP_FW_ARG_TABLEARG(chain, k, f)
1251 /*
1252  * The main check routine for the firewall.
1253  *
1254  * All arguments are in args so we can modify them and return them
1255  * back to the caller.
1256  *
1257  * Parameters:
1258  *
1259  *      args->m (in/out) The packet; we set to NULL when/if we nuke it.
1260  *              Starts with the IP header.
1261  *      args->L3offset  Number of bytes bypassed if we came from L2.
1262  *                      e.g. often sizeof(eh)  ** NOTYET **
1263  *      args->ifp       Incoming or outgoing interface.
1264  *      args->divert_rule (in/out)
1265  *              Skip up to the first rule past this rule number;
1266  *              upon return, non-zero port number for divert or tee.
1267  *
1268  *      args->rule      Pointer to the last matching rule (in/out)
1269  *      args->next_hop  Socket we are forwarding to (out).
1270  *      args->next_hop6 IPv6 next hop we are forwarding to (out).
1271  *      args->f_id      Addresses grabbed from the packet (out)
1272  *      args->rule.info a cookie depending on rule action
1273  *
1274  * Return value:
1275  *
1276  *      IP_FW_PASS      the packet must be accepted
1277  *      IP_FW_DENY      the packet must be dropped
1278  *      IP_FW_DIVERT    divert packet, port in m_tag
1279  *      IP_FW_TEE       tee packet, port in m_tag
1280  *      IP_FW_DUMMYNET  to dummynet, pipe in args->cookie
1281  *      IP_FW_NETGRAPH  into netgraph, cookie args->cookie
1282  *              args->rule contains the matching rule,
1283  *              args->rule.info has additional information.
1284  *
1285  */
1286 int
1287 ipfw_chk(struct ip_fw_args *args)
1288 {
1289
1290         /*
1291          * Local variables holding state while processing a packet:
1292          *
1293          * IMPORTANT NOTE: to speed up the processing of rules, there
1294          * are some assumption on the values of the variables, which
1295          * are documented here. Should you change them, please check
1296          * the implementation of the various instructions to make sure
1297          * that they still work.
1298          *
1299          * m | args->m  Pointer to the mbuf, as received from the caller.
1300          *      It may change if ipfw_chk() does an m_pullup, or if it
1301          *      consumes the packet because it calls send_reject().
1302          *      XXX This has to change, so that ipfw_chk() never modifies
1303          *      or consumes the buffer.
1304          *      OR
1305          * args->mem    Pointer to contigous memory chunk.
1306          * ip   Is the beginning of the ip(4 or 6) header.
1307          * eh   Ethernet header in case if input is Layer2.
1308          */
1309         struct mbuf *m;
1310         struct ip *ip;
1311         struct ether_header *eh;
1312
1313         /*
1314          * For rules which contain uid/gid or jail constraints, cache
1315          * a copy of the users credentials after the pcb lookup has been
1316          * executed. This will speed up the processing of rules with
1317          * these types of constraints, as well as decrease contention
1318          * on pcb related locks.
1319          */
1320 #ifndef __FreeBSD__
1321         struct bsd_ucred ucred_cache;
1322 #else
1323         struct ucred *ucred_cache = NULL;
1324 #endif
1325         int ucred_lookup = 0;
1326         int f_pos = 0;          /* index of current rule in the array */
1327         int retval = 0;
1328         struct ifnet *oif, *iif;
1329
1330         /*
1331          * hlen The length of the IP header.
1332          */
1333         u_int hlen = 0;         /* hlen >0 means we have an IP pkt */
1334
1335         /*
1336          * offset       The offset of a fragment. offset != 0 means that
1337          *      we have a fragment at this offset of an IPv4 packet.
1338          *      offset == 0 means that (if this is an IPv4 packet)
1339          *      this is the first or only fragment.
1340          *      For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1341          *      or there is a single packet fragment (fragment header added
1342          *      without needed).  We will treat a single packet fragment as if
1343          *      there was no fragment header (or log/block depending on the
1344          *      V_fw_permit_single_frag6 sysctl setting).
1345          */
1346         u_short offset = 0;
1347         u_short ip6f_mf = 0;
1348
1349         /*
1350          * Local copies of addresses. They are only valid if we have
1351          * an IP packet.
1352          *
1353          * proto        The protocol. Set to 0 for non-ip packets,
1354          *      or to the protocol read from the packet otherwise.
1355          *      proto != 0 means that we have an IPv4 packet.
1356          *
1357          * src_port, dst_port   port numbers, in HOST format. Only
1358          *      valid for TCP and UDP packets.
1359          *
1360          * src_ip, dst_ip       ip addresses, in NETWORK format.
1361          *      Only valid for IPv4 packets.
1362          */
1363         uint8_t proto;
1364         uint16_t src_port, dst_port;            /* NOTE: host format    */
1365         struct in_addr src_ip, dst_ip;          /* NOTE: network format */
1366         int iplen = 0;
1367         int pktlen;
1368
1369         struct ipfw_dyn_info dyn_info;
1370         struct ip_fw *q = NULL;
1371         struct ip_fw_chain *chain = &V_layer3_chain;
1372
1373         /*
1374          * We store in ulp a pointer to the upper layer protocol header.
1375          * In the ipv4 case this is easy to determine from the header,
1376          * but for ipv6 we might have some additional headers in the middle.
1377          * ulp is NULL if not found.
1378          */
1379         void *ulp = NULL;               /* upper layer protocol pointer. */
1380
1381         /* XXX ipv6 variables */
1382         int is_ipv6 = 0;
1383         uint8_t icmp6_type = 0;
1384         uint16_t ext_hd = 0;    /* bits vector for extension header filtering */
1385         /* end of ipv6 variables */
1386
1387         int is_ipv4 = 0;
1388
1389         int done = 0;           /* flag to exit the outer loop */
1390         IPFW_RLOCK_TRACKER;
1391         bool mem;
1392
1393         if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1394                 if (args->flags & IPFW_ARGS_ETHER) {
1395                         eh = (struct ether_header *)args->mem;
1396                         if (eh->ether_type == htons(ETHERTYPE_VLAN))
1397                                 ip = (struct ip *)
1398                                     ((struct ether_vlan_header *)eh + 1);
1399                         else
1400                                 ip = (struct ip *)(eh + 1);
1401                 } else {
1402                         eh = NULL;
1403                         ip = (struct ip *)args->mem;
1404                 }
1405                 pktlen = IPFW_ARGS_LENGTH(args->flags);
1406                 args->f_id.fib = args->ifp->if_fib;     /* best guess */
1407         } else {
1408                 m = args->m;
1409                 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1410                         return (IP_FW_PASS);    /* accept */
1411                 if (args->flags & IPFW_ARGS_ETHER) {
1412                         /* We need some amount of data to be contiguous. */
1413                         if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1414                             (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1415                             max_protohdr))) == NULL)
1416                                 goto pullup_failed;
1417                         eh = mtod(m, struct ether_header *);
1418                         ip = (struct ip *)(eh + 1);
1419                 } else {
1420                         eh = NULL;
1421                         ip = mtod(m, struct ip *);
1422                 }
1423                 pktlen = m->m_pkthdr.len;
1424                 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1425         }
1426
1427         dst_ip.s_addr = 0;              /* make sure it is initialized */
1428         src_ip.s_addr = 0;              /* make sure it is initialized */
1429         src_port = dst_port = 0;
1430
1431         DYN_INFO_INIT(&dyn_info);
1432 /*
1433  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1434  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1435  * pointer might become stale after other pullups (but we never use it
1436  * this way).
1437  */
1438 #define PULLUP_TO(_len, p, T)   PULLUP_LEN(_len, p, sizeof(T))
1439 #define EHLEN   (eh != NULL ? ((char *)ip - (char *)eh) : 0)
1440 #define PULLUP_LEN(_len, p, T)                                  \
1441 do {                                                            \
1442         int x = (_len) + T + EHLEN;                             \
1443         if (mem) {                                              \
1444                 MPASS(pktlen >= x);                             \
1445                 p = (char *)args->mem + (_len) + EHLEN;         \
1446         } else {                                                \
1447                 if (__predict_false((m)->m_len < x)) {          \
1448                         args->m = m = m_pullup(m, x);           \
1449                         if (m == NULL)                          \
1450                                 goto pullup_failed;             \
1451                 }                                               \
1452                 p = mtod(m, char *) + (_len) + EHLEN;           \
1453         }                                                       \
1454 } while (0)
1455 /*
1456  * In case pointers got stale after pullups, update them.
1457  */
1458 #define UPDATE_POINTERS()                                       \
1459 do {                                                            \
1460         if (!mem) {                                             \
1461                 if (eh != NULL) {                               \
1462                         eh = mtod(m, struct ether_header *);    \
1463                         ip = (struct ip *)(eh + 1);             \
1464                 } else                                          \
1465                         ip = mtod(m, struct ip *);              \
1466                 args->m = m;                                    \
1467         }                                                       \
1468 } while (0)
1469
1470         /* Identify IP packets and fill up variables. */
1471         if (pktlen >= sizeof(struct ip6_hdr) &&
1472             (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1473             ip->ip_v == 6) {
1474                 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1475
1476                 is_ipv6 = 1;
1477                 args->flags |= IPFW_ARGS_IP6;
1478                 hlen = sizeof(struct ip6_hdr);
1479                 proto = ip6->ip6_nxt;
1480                 /* Search extension headers to find upper layer protocols */
1481                 while (ulp == NULL && offset == 0) {
1482                         switch (proto) {
1483                         case IPPROTO_ICMPV6:
1484                                 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1485                                 icmp6_type = ICMP6(ulp)->icmp6_type;
1486                                 break;
1487
1488                         case IPPROTO_TCP:
1489                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1490                                 dst_port = TCP(ulp)->th_dport;
1491                                 src_port = TCP(ulp)->th_sport;
1492                                 /* save flags for dynamic rules */
1493                                 args->f_id._flags = TCP(ulp)->th_flags;
1494                                 break;
1495
1496                         case IPPROTO_SCTP:
1497                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1498                                     sizeof(struct sctp_chunkhdr) +
1499                                     offsetof(struct sctp_init, a_rwnd))
1500                                         PULLUP_LEN(hlen, ulp,
1501                                             sizeof(struct sctphdr) +
1502                                             sizeof(struct sctp_chunkhdr) +
1503                                             offsetof(struct sctp_init, a_rwnd));
1504                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1505                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1506                                 else
1507                                         PULLUP_LEN(hlen, ulp,
1508                                             sizeof(struct sctphdr));
1509                                 src_port = SCTP(ulp)->src_port;
1510                                 dst_port = SCTP(ulp)->dest_port;
1511                                 break;
1512
1513                         case IPPROTO_UDP:
1514                         case IPPROTO_UDPLITE:
1515                                 PULLUP_TO(hlen, ulp, struct udphdr);
1516                                 dst_port = UDP(ulp)->uh_dport;
1517                                 src_port = UDP(ulp)->uh_sport;
1518                                 break;
1519
1520                         case IPPROTO_HOPOPTS:   /* RFC 2460 */
1521                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1522                                 ext_hd |= EXT_HOPOPTS;
1523                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1524                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1525                                 ulp = NULL;
1526                                 break;
1527
1528                         case IPPROTO_ROUTING:   /* RFC 2460 */
1529                                 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1530                                 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1531                                 case 0:
1532                                         ext_hd |= EXT_RTHDR0;
1533                                         break;
1534                                 case 2:
1535                                         ext_hd |= EXT_RTHDR2;
1536                                         break;
1537                                 default:
1538                                         if (V_fw_verbose)
1539                                                 printf("IPFW2: IPV6 - Unknown "
1540                                                     "Routing Header type(%d)\n",
1541                                                     ((struct ip6_rthdr *)
1542                                                     ulp)->ip6r_type);
1543                                         if (V_fw_deny_unknown_exthdrs)
1544                                             return (IP_FW_DENY);
1545                                         break;
1546                                 }
1547                                 ext_hd |= EXT_ROUTING;
1548                                 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1549                                 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1550                                 ulp = NULL;
1551                                 break;
1552
1553                         case IPPROTO_FRAGMENT:  /* RFC 2460 */
1554                                 PULLUP_TO(hlen, ulp, struct ip6_frag);
1555                                 ext_hd |= EXT_FRAGMENT;
1556                                 hlen += sizeof (struct ip6_frag);
1557                                 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1558                                 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1559                                         IP6F_OFF_MASK;
1560                                 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1561                                         IP6F_MORE_FRAG;
1562                                 if (V_fw_permit_single_frag6 == 0 &&
1563                                     offset == 0 && ip6f_mf == 0) {
1564                                         if (V_fw_verbose)
1565                                                 printf("IPFW2: IPV6 - Invalid "
1566                                                     "Fragment Header\n");
1567                                         if (V_fw_deny_unknown_exthdrs)
1568                                             return (IP_FW_DENY);
1569                                         break;
1570                                 }
1571                                 args->f_id.extra =
1572                                     ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1573                                 ulp = NULL;
1574                                 break;
1575
1576                         case IPPROTO_DSTOPTS:   /* RFC 2460 */
1577                                 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1578                                 ext_hd |= EXT_DSTOPTS;
1579                                 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1580                                 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1581                                 ulp = NULL;
1582                                 break;
1583
1584                         case IPPROTO_AH:        /* RFC 2402 */
1585                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1586                                 ext_hd |= EXT_AH;
1587                                 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1588                                 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1589                                 ulp = NULL;
1590                                 break;
1591
1592                         case IPPROTO_ESP:       /* RFC 2406 */
1593                                 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1594                                 /* Anything past Seq# is variable length and
1595                                  * data past this ext. header is encrypted. */
1596                                 ext_hd |= EXT_ESP;
1597                                 break;
1598
1599                         case IPPROTO_NONE:      /* RFC 2460 */
1600                                 /*
1601                                  * Packet ends here, and IPv6 header has
1602                                  * already been pulled up. If ip6e_len!=0
1603                                  * then octets must be ignored.
1604                                  */
1605                                 ulp = ip; /* non-NULL to get out of loop. */
1606                                 break;
1607
1608                         case IPPROTO_OSPFIGP:
1609                                 /* XXX OSPF header check? */
1610                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1611                                 break;
1612
1613                         case IPPROTO_PIM:
1614                                 /* XXX PIM header check? */
1615                                 PULLUP_TO(hlen, ulp, struct pim);
1616                                 break;
1617
1618                         case IPPROTO_GRE:       /* RFC 1701 */
1619                                 /* XXX GRE header check? */
1620                                 PULLUP_TO(hlen, ulp, struct grehdr);
1621                                 break;
1622
1623                         case IPPROTO_CARP:
1624                                 PULLUP_TO(hlen, ulp, offsetof(
1625                                     struct carp_header, carp_counter));
1626                                 if (CARP_ADVERTISEMENT !=
1627                                     ((struct carp_header *)ulp)->carp_type)
1628                                         return (IP_FW_DENY);
1629                                 break;
1630
1631                         case IPPROTO_IPV6:      /* RFC 2893 */
1632                                 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1633                                 break;
1634
1635                         case IPPROTO_IPV4:      /* RFC 2893 */
1636                                 PULLUP_TO(hlen, ulp, struct ip);
1637                                 break;
1638
1639                         default:
1640                                 if (V_fw_verbose)
1641                                         printf("IPFW2: IPV6 - Unknown "
1642                                             "Extension Header(%d), ext_hd=%x\n",
1643                                              proto, ext_hd);
1644                                 if (V_fw_deny_unknown_exthdrs)
1645                                     return (IP_FW_DENY);
1646                                 PULLUP_TO(hlen, ulp, struct ip6_ext);
1647                                 break;
1648                         } /*switch */
1649                 }
1650                 UPDATE_POINTERS();
1651                 ip6 = (struct ip6_hdr *)ip;
1652                 args->f_id.addr_type = 6;
1653                 args->f_id.src_ip6 = ip6->ip6_src;
1654                 args->f_id.dst_ip6 = ip6->ip6_dst;
1655                 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1656                 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1657         } else if (pktlen >= sizeof(struct ip) &&
1658             (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1659             ip->ip_v == 4) {
1660                 is_ipv4 = 1;
1661                 args->flags |= IPFW_ARGS_IP4;
1662                 hlen = ip->ip_hl << 2;
1663                 /*
1664                  * Collect parameters into local variables for faster
1665                  * matching.
1666                  */
1667                 proto = ip->ip_p;
1668                 src_ip = ip->ip_src;
1669                 dst_ip = ip->ip_dst;
1670                 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1671                 iplen = ntohs(ip->ip_len);
1672
1673                 if (offset == 0) {
1674                         switch (proto) {
1675                         case IPPROTO_TCP:
1676                                 PULLUP_TO(hlen, ulp, struct tcphdr);
1677                                 dst_port = TCP(ulp)->th_dport;
1678                                 src_port = TCP(ulp)->th_sport;
1679                                 /* save flags for dynamic rules */
1680                                 args->f_id._flags = TCP(ulp)->th_flags;
1681                                 break;
1682
1683                         case IPPROTO_SCTP:
1684                                 if (pktlen >= hlen + sizeof(struct sctphdr) +
1685                                     sizeof(struct sctp_chunkhdr) +
1686                                     offsetof(struct sctp_init, a_rwnd))
1687                                         PULLUP_LEN(hlen, ulp,
1688                                             sizeof(struct sctphdr) +
1689                                             sizeof(struct sctp_chunkhdr) +
1690                                             offsetof(struct sctp_init, a_rwnd));
1691                                 else if (pktlen >= hlen + sizeof(struct sctphdr))
1692                                         PULLUP_LEN(hlen, ulp, pktlen - hlen);
1693                                 else
1694                                         PULLUP_LEN(hlen, ulp,
1695                                             sizeof(struct sctphdr));
1696                                 src_port = SCTP(ulp)->src_port;
1697                                 dst_port = SCTP(ulp)->dest_port;
1698                                 break;
1699
1700                         case IPPROTO_UDP:
1701                         case IPPROTO_UDPLITE:
1702                                 PULLUP_TO(hlen, ulp, struct udphdr);
1703                                 dst_port = UDP(ulp)->uh_dport;
1704                                 src_port = UDP(ulp)->uh_sport;
1705                                 break;
1706
1707                         case IPPROTO_ICMP:
1708                                 PULLUP_TO(hlen, ulp, struct icmphdr);
1709                                 //args->f_id.flags = ICMP(ulp)->icmp_type;
1710                                 break;
1711
1712                         default:
1713                                 break;
1714                         }
1715                 }
1716
1717                 UPDATE_POINTERS();
1718                 args->f_id.addr_type = 4;
1719                 args->f_id.src_ip = ntohl(src_ip.s_addr);
1720                 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1721         } else {
1722                 proto = 0;
1723                 dst_ip.s_addr = src_ip.s_addr = 0;
1724
1725                 args->f_id.addr_type = 1; /* XXX */
1726         }
1727 #undef PULLUP_TO
1728         pktlen = iplen < pktlen ? iplen: pktlen;
1729
1730         /* Properly initialize the rest of f_id */
1731         args->f_id.proto = proto;
1732         args->f_id.src_port = src_port = ntohs(src_port);
1733         args->f_id.dst_port = dst_port = ntohs(dst_port);
1734
1735         IPFW_PF_RLOCK(chain);
1736         if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1737                 IPFW_PF_RUNLOCK(chain);
1738                 return (IP_FW_PASS);    /* accept */
1739         }
1740         if (args->flags & IPFW_ARGS_REF) {
1741                 /*
1742                  * Packet has already been tagged as a result of a previous
1743                  * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1744                  * REASS, NETGRAPH, DIVERT/TEE...)
1745                  * Validate the slot and continue from the next one
1746                  * if still present, otherwise do a lookup.
1747                  */
1748                 f_pos = (args->rule.chain_id == chain->id) ?
1749                     args->rule.slot :
1750                     ipfw_find_rule(chain, args->rule.rulenum,
1751                         args->rule.rule_id);
1752         } else {
1753                 f_pos = 0;
1754         }
1755
1756         if (args->flags & IPFW_ARGS_IN) {
1757                 iif = args->ifp;
1758                 oif = NULL;
1759         } else {
1760                 MPASS(args->flags & IPFW_ARGS_OUT);
1761                 iif = mem ? NULL : m->m_pkthdr.rcvif;
1762                 oif = args->ifp;
1763         }
1764
1765         /*
1766          * Now scan the rules, and parse microinstructions for each rule.
1767          * We have two nested loops and an inner switch. Sometimes we
1768          * need to break out of one or both loops, or re-enter one of
1769          * the loops with updated variables. Loop variables are:
1770          *
1771          *      f_pos (outer loop) points to the current rule.
1772          *              On output it points to the matching rule.
1773          *      done (outer loop) is used as a flag to break the loop.
1774          *      l (inner loop)  residual length of current rule.
1775          *              cmd points to the current microinstruction.
1776          *
1777          * We break the inner loop by setting l=0 and possibly
1778          * cmdlen=0 if we don't want to advance cmd.
1779          * We break the outer loop by setting done=1
1780          * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1781          * as needed.
1782          */
1783         for (; f_pos < chain->n_rules; f_pos++) {
1784                 ipfw_insn *cmd;
1785                 uint32_t tablearg = 0;
1786                 int l, cmdlen, skip_or; /* skip rest of OR block */
1787                 struct ip_fw *f;
1788
1789                 f = chain->map[f_pos];
1790                 if (V_set_disable & (1 << f->set) )
1791                         continue;
1792
1793                 skip_or = 0;
1794                 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1795                     l -= cmdlen, cmd += cmdlen) {
1796                         int match;
1797
1798                         /*
1799                          * check_body is a jump target used when we find a
1800                          * CHECK_STATE, and need to jump to the body of
1801                          * the target rule.
1802                          */
1803
1804 /* check_body: */
1805                         cmdlen = F_LEN(cmd);
1806                         /*
1807                          * An OR block (insn_1 || .. || insn_n) has the
1808                          * F_OR bit set in all but the last instruction.
1809                          * The first match will set "skip_or", and cause
1810                          * the following instructions to be skipped until
1811                          * past the one with the F_OR bit clear.
1812                          */
1813                         if (skip_or) {          /* skip this instruction */
1814                                 if ((cmd->len & F_OR) == 0)
1815                                         skip_or = 0;    /* next one is good */
1816                                 continue;
1817                         }
1818                         match = 0; /* set to 1 if we succeed */
1819
1820                         switch (cmd->opcode) {
1821                         /*
1822                          * The first set of opcodes compares the packet's
1823                          * fields with some pattern, setting 'match' if a
1824                          * match is found. At the end of the loop there is
1825                          * logic to deal with F_NOT and F_OR flags associated
1826                          * with the opcode.
1827                          */
1828                         case O_NOP:
1829                                 match = 1;
1830                                 break;
1831
1832                         case O_FORWARD_MAC:
1833                                 printf("ipfw: opcode %d unimplemented\n",
1834                                     cmd->opcode);
1835                                 break;
1836
1837                         case O_GID:
1838                         case O_UID:
1839                         case O_JAIL:
1840                                 /*
1841                                  * We only check offset == 0 && proto != 0,
1842                                  * as this ensures that we have a
1843                                  * packet with the ports info.
1844                                  */
1845                                 if (offset != 0)
1846                                         break;
1847                                 if (proto == IPPROTO_TCP ||
1848                                     proto == IPPROTO_UDP ||
1849                                     proto == IPPROTO_UDPLITE)
1850                                         match = check_uidgid(
1851                                                     (ipfw_insn_u32 *)cmd,
1852                                                     args, &ucred_lookup,
1853 #ifdef __FreeBSD__
1854                                                     &ucred_cache);
1855 #else
1856                                                     (void *)&ucred_cache);
1857 #endif
1858                                 break;
1859
1860                         case O_RECV:
1861                                 match = iface_match(iif, (ipfw_insn_if *)cmd,
1862                                     chain, &tablearg);
1863                                 break;
1864
1865                         case O_XMIT:
1866                                 match = iface_match(oif, (ipfw_insn_if *)cmd,
1867                                     chain, &tablearg);
1868                                 break;
1869
1870                         case O_VIA:
1871                                 match = iface_match(args->ifp,
1872                                     (ipfw_insn_if *)cmd, chain, &tablearg);
1873                                 break;
1874
1875                         case O_MACADDR2:
1876                                 if (args->flags & IPFW_ARGS_ETHER) {
1877                                         u_int32_t *want = (u_int32_t *)
1878                                                 ((ipfw_insn_mac *)cmd)->addr;
1879                                         u_int32_t *mask = (u_int32_t *)
1880                                                 ((ipfw_insn_mac *)cmd)->mask;
1881                                         u_int32_t *hdr = (u_int32_t *)eh;
1882
1883                                         match =
1884                                             ( want[0] == (hdr[0] & mask[0]) &&
1885                                               want[1] == (hdr[1] & mask[1]) &&
1886                                               want[2] == (hdr[2] & mask[2]) );
1887                                 }
1888                                 break;
1889
1890                         case O_MAC_TYPE:
1891                                 if (args->flags & IPFW_ARGS_ETHER) {
1892                                         u_int16_t *p =
1893                                             ((ipfw_insn_u16 *)cmd)->ports;
1894                                         int i;
1895
1896                                         for (i = cmdlen - 1; !match && i>0;
1897                                             i--, p += 2)
1898                                                 match =
1899                                                     (ntohs(eh->ether_type) >=
1900                                                     p[0] &&
1901                                                     ntohs(eh->ether_type) <=
1902                                                     p[1]);
1903                                 }
1904                                 break;
1905
1906                         case O_FRAG:
1907                                 match = (offset != 0);
1908                                 break;
1909
1910                         case O_IN:      /* "out" is "not in" */
1911                                 match = (oif == NULL);
1912                                 break;
1913
1914                         case O_LAYER2:
1915                                 match = (args->flags & IPFW_ARGS_ETHER);
1916                                 break;
1917
1918                         case O_DIVERTED:
1919                                 if ((args->flags & IPFW_ARGS_REF) == 0)
1920                                         break;
1921                                 /*
1922                                  * For diverted packets, args->rule.info
1923                                  * contains the divert port (in host format)
1924                                  * reason and direction.
1925                                  */
1926                                 match = ((args->rule.info & IPFW_IS_MASK) ==
1927                                     IPFW_IS_DIVERT) && (
1928                                     ((args->rule.info & IPFW_INFO_IN) ?
1929                                         1: 2) & cmd->arg1);
1930                                 break;
1931
1932                         case O_PROTO:
1933                                 /*
1934                                  * We do not allow an arg of 0 so the
1935                                  * check of "proto" only suffices.
1936                                  */
1937                                 match = (proto == cmd->arg1);
1938                                 break;
1939
1940                         case O_IP_SRC:
1941                                 match = is_ipv4 &&
1942                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1943                                     src_ip.s_addr);
1944                                 break;
1945
1946                         case O_IP_DST_LOOKUP:
1947                         {
1948                                 void *pkey;
1949                                 uint32_t vidx, key;
1950                                 uint16_t keylen;
1951
1952                                 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1953                                         /* Determine lookup key type */
1954                                         vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1955                                         if (vidx != 4 /* uid */ &&
1956                                             vidx != 5 /* jail */ &&
1957                                             is_ipv6 == 0 && is_ipv4 == 0)
1958                                                 break;
1959                                         /* Determine key length */
1960                                         if (vidx == 0 /* dst-ip */ ||
1961                                             vidx == 1 /* src-ip */)
1962                                                 keylen = is_ipv6 ?
1963                                                     sizeof(struct in6_addr):
1964                                                     sizeof(in_addr_t);
1965                                         else {
1966                                                 keylen = sizeof(key);
1967                                                 pkey = &key;
1968                                         }
1969                                         if (vidx == 0 /* dst-ip */)
1970                                                 pkey = is_ipv4 ? (void *)&dst_ip:
1971                                                     (void *)&args->f_id.dst_ip6;
1972                                         else if (vidx == 1 /* src-ip */)
1973                                                 pkey = is_ipv4 ? (void *)&src_ip:
1974                                                     (void *)&args->f_id.src_ip6;
1975                                         else if (vidx == 6 /* dscp */) {
1976                                                 if (is_ipv4)
1977                                                         key = ip->ip_tos >> 2;
1978                                                 else {
1979                                                         key = args->f_id.flow_id6;
1980                                                         key = (key & 0x0f) << 2 |
1981                                                             (key & 0xf000) >> 14;
1982                                                 }
1983                                                 key &= 0x3f;
1984                                         } else if (vidx == 2 /* dst-port */ ||
1985                                             vidx == 3 /* src-port */) {
1986                                                 /* Skip fragments */
1987                                                 if (offset != 0)
1988                                                         break;
1989                                                 /* Skip proto without ports */
1990                                                 if (proto != IPPROTO_TCP &&
1991                                                     proto != IPPROTO_UDP &&
1992                                                     proto != IPPROTO_UDPLITE &&
1993                                                     proto != IPPROTO_SCTP)
1994                                                         break;
1995                                                 if (vidx == 2 /* dst-port */)
1996                                                         key = dst_port;
1997                                                 else
1998                                                         key = src_port;
1999                                         }
2000 #ifndef USERSPACE
2001                                         else if (vidx == 4 /* uid */ ||
2002                                             vidx == 5 /* jail */) {
2003                                                 check_uidgid(
2004                                                     (ipfw_insn_u32 *)cmd,
2005                                                     args, &ucred_lookup,
2006 #ifdef __FreeBSD__
2007                                                     &ucred_cache);
2008                                                 if (vidx == 4 /* uid */)
2009                                                         key = ucred_cache->cr_uid;
2010                                                 else if (vidx == 5 /* jail */)
2011                                                         key = ucred_cache->cr_prison->pr_id;
2012 #else /* !__FreeBSD__ */
2013                                                     (void *)&ucred_cache);
2014                                                 if (vidx == 4 /* uid */)
2015                                                         key = ucred_cache.uid;
2016                                                 else if (vidx == 5 /* jail */)
2017                                                         key = ucred_cache.xid;
2018 #endif /* !__FreeBSD__ */
2019                                         }
2020 #endif /* !USERSPACE */
2021                                         else
2022                                                 break;
2023                                         match = ipfw_lookup_table(chain,
2024                                             cmd->arg1, keylen, pkey, &vidx);
2025                                         if (!match)
2026                                                 break;
2027                                         tablearg = vidx;
2028                                         break;
2029                                 }
2030                                 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2031                                 /* FALLTHROUGH */
2032                         }
2033                         case O_IP_SRC_LOOKUP:
2034                         {
2035                                 void *pkey;
2036                                 uint32_t vidx;
2037                                 uint16_t keylen;
2038
2039                                 if (is_ipv4) {
2040                                         keylen = sizeof(in_addr_t);
2041                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2042                                                 pkey = &dst_ip;
2043                                         else
2044                                                 pkey = &src_ip;
2045                                 } else if (is_ipv6) {
2046                                         keylen = sizeof(struct in6_addr);
2047                                         if (cmd->opcode == O_IP_DST_LOOKUP)
2048                                                 pkey = &args->f_id.dst_ip6;
2049                                         else
2050                                                 pkey = &args->f_id.src_ip6;
2051                                 } else
2052                                         break;
2053                                 match = ipfw_lookup_table(chain, cmd->arg1,
2054                                     keylen, pkey, &vidx);
2055                                 if (!match)
2056                                         break;
2057                                 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2058                                         match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2059                                             TARG_VAL(chain, vidx, tag);
2060                                         if (!match)
2061                                                 break;
2062                                 }
2063                                 tablearg = vidx;
2064                                 break;
2065                         }
2066
2067                         case O_IP_FLOW_LOOKUP:
2068                                 {
2069                                         uint32_t v = 0;
2070                                         match = ipfw_lookup_table(chain,
2071                                             cmd->arg1, 0, &args->f_id, &v);
2072                                         if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2073                                                 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2074                                                     TARG_VAL(chain, v, tag);
2075                                         if (match)
2076                                                 tablearg = v;
2077                                 }
2078                                 break;
2079                         case O_IP_SRC_MASK:
2080                         case O_IP_DST_MASK:
2081                                 if (is_ipv4) {
2082                                     uint32_t a =
2083                                         (cmd->opcode == O_IP_DST_MASK) ?
2084                                             dst_ip.s_addr : src_ip.s_addr;
2085                                     uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2086                                     int i = cmdlen-1;
2087
2088                                     for (; !match && i>0; i-= 2, p+= 2)
2089                                         match = (p[0] == (a & p[1]));
2090                                 }
2091                                 break;
2092
2093                         case O_IP_SRC_ME:
2094                                 if (is_ipv4) {
2095                                         match = in_localip(src_ip);
2096                                         break;
2097                                 }
2098 #ifdef INET6
2099                                 /* FALLTHROUGH */
2100                         case O_IP6_SRC_ME:
2101                                 match = is_ipv6 &&
2102                                     ipfw_localip6(&args->f_id.src_ip6);
2103 #endif
2104                                 break;
2105
2106                         case O_IP_DST_SET:
2107                         case O_IP_SRC_SET:
2108                                 if (is_ipv4) {
2109                                         u_int32_t *d = (u_int32_t *)(cmd+1);
2110                                         u_int32_t addr =
2111                                             cmd->opcode == O_IP_DST_SET ?
2112                                                 args->f_id.dst_ip :
2113                                                 args->f_id.src_ip;
2114
2115                                             if (addr < d[0])
2116                                                     break;
2117                                             addr -= d[0]; /* subtract base */
2118                                             match = (addr < cmd->arg1) &&
2119                                                 ( d[ 1 + (addr>>5)] &
2120                                                   (1<<(addr & 0x1f)) );
2121                                 }
2122                                 break;
2123
2124                         case O_IP_DST:
2125                                 match = is_ipv4 &&
2126                                     (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2127                                     dst_ip.s_addr);
2128                                 break;
2129
2130                         case O_IP_DST_ME:
2131                                 if (is_ipv4) {
2132                                         match = in_localip(dst_ip);
2133                                         break;
2134                                 }
2135 #ifdef INET6
2136                                 /* FALLTHROUGH */
2137                         case O_IP6_DST_ME:
2138                                 match = is_ipv6 &&
2139                                     ipfw_localip6(&args->f_id.dst_ip6);
2140 #endif
2141                                 break;
2142
2143
2144                         case O_IP_SRCPORT:
2145                         case O_IP_DSTPORT:
2146                                 /*
2147                                  * offset == 0 && proto != 0 is enough
2148                                  * to guarantee that we have a
2149                                  * packet with port info.
2150                                  */
2151                                 if ((proto == IPPROTO_UDP ||
2152                                     proto == IPPROTO_UDPLITE ||
2153                                     proto == IPPROTO_TCP ||
2154                                     proto == IPPROTO_SCTP) && offset == 0) {
2155                                         u_int16_t x =
2156                                             (cmd->opcode == O_IP_SRCPORT) ?
2157                                                 src_port : dst_port ;
2158                                         u_int16_t *p =
2159                                             ((ipfw_insn_u16 *)cmd)->ports;
2160                                         int i;
2161
2162                                         for (i = cmdlen - 1; !match && i>0;
2163                                             i--, p += 2)
2164                                                 match = (x>=p[0] && x<=p[1]);
2165                                 }
2166                                 break;
2167
2168                         case O_ICMPTYPE:
2169                                 match = (offset == 0 && proto==IPPROTO_ICMP &&
2170                                     icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2171                                 break;
2172
2173 #ifdef INET6
2174                         case O_ICMP6TYPE:
2175                                 match = is_ipv6 && offset == 0 &&
2176                                     proto==IPPROTO_ICMPV6 &&
2177                                     icmp6type_match(
2178                                         ICMP6(ulp)->icmp6_type,
2179                                         (ipfw_insn_u32 *)cmd);
2180                                 break;
2181 #endif /* INET6 */
2182
2183                         case O_IPOPT:
2184                                 match = (is_ipv4 &&
2185                                     ipopts_match(ip, cmd) );
2186                                 break;
2187
2188                         case O_IPVER:
2189                                 match = (is_ipv4 &&
2190                                     cmd->arg1 == ip->ip_v);
2191                                 break;
2192
2193                         case O_IPID:
2194                         case O_IPLEN:
2195                         case O_IPTTL:
2196                                 if (is_ipv4) {  /* only for IP packets */
2197                                     uint16_t x;
2198                                     uint16_t *p;
2199                                     int i;
2200
2201                                     if (cmd->opcode == O_IPLEN)
2202                                         x = iplen;
2203                                     else if (cmd->opcode == O_IPTTL)
2204                                         x = ip->ip_ttl;
2205                                     else /* must be IPID */
2206                                         x = ntohs(ip->ip_id);
2207                                     if (cmdlen == 1) {
2208                                         match = (cmd->arg1 == x);
2209                                         break;
2210                                     }
2211                                     /* otherwise we have ranges */
2212                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2213                                     i = cmdlen - 1;
2214                                     for (; !match && i>0; i--, p += 2)
2215                                         match = (x >= p[0] && x <= p[1]);
2216                                 }
2217                                 break;
2218
2219                         case O_IPPRECEDENCE:
2220                                 match = (is_ipv4 &&
2221                                     (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2222                                 break;
2223
2224                         case O_IPTOS:
2225                                 match = (is_ipv4 &&
2226                                     flags_match(cmd, ip->ip_tos));
2227                                 break;
2228
2229                         case O_DSCP:
2230                             {
2231                                 uint32_t *p;
2232                                 uint16_t x;
2233
2234                                 p = ((ipfw_insn_u32 *)cmd)->d;
2235
2236                                 if (is_ipv4)
2237                                         x = ip->ip_tos >> 2;
2238                                 else if (is_ipv6) {
2239                                         uint8_t *v;
2240                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
2241                                         x = (*v & 0x0F) << 2;
2242                                         v++;
2243                                         x |= *v >> 6;
2244                                 } else
2245                                         break;
2246
2247                                 /* DSCP bitmask is stored as low_u32 high_u32 */
2248                                 if (x >= 32)
2249                                         match = *(p + 1) & (1 << (x - 32));
2250                                 else
2251                                         match = *p & (1 << x);
2252                             }
2253                                 break;
2254
2255                         case O_TCPDATALEN:
2256                                 if (proto == IPPROTO_TCP && offset == 0) {
2257                                     struct tcphdr *tcp;
2258                                     uint16_t x;
2259                                     uint16_t *p;
2260                                     int i;
2261 #ifdef INET6
2262                                     if (is_ipv6) {
2263                                             struct ip6_hdr *ip6;
2264
2265                                             ip6 = (struct ip6_hdr *)ip;
2266                                             if (ip6->ip6_plen == 0) {
2267                                                     /*
2268                                                      * Jumbo payload is not
2269                                                      * supported by this
2270                                                      * opcode.
2271                                                      */
2272                                                     break;
2273                                             }
2274                                             x = iplen - hlen;
2275                                     } else
2276 #endif /* INET6 */
2277                                             x = iplen - (ip->ip_hl << 2);
2278                                     tcp = TCP(ulp);
2279                                     x -= tcp->th_off << 2;
2280                                     if (cmdlen == 1) {
2281                                         match = (cmd->arg1 == x);
2282                                         break;
2283                                     }
2284                                     /* otherwise we have ranges */
2285                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2286                                     i = cmdlen - 1;
2287                                     for (; !match && i>0; i--, p += 2)
2288                                         match = (x >= p[0] && x <= p[1]);
2289                                 }
2290                                 break;
2291
2292                         case O_TCPFLAGS:
2293                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2294                                     flags_match(cmd, TCP(ulp)->th_flags));
2295                                 break;
2296
2297                         case O_TCPOPTS:
2298                                 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2299                                         PULLUP_LEN(hlen, ulp,
2300                                             (TCP(ulp)->th_off << 2));
2301                                         match = tcpopts_match(TCP(ulp), cmd);
2302                                 }
2303                                 break;
2304
2305                         case O_TCPSEQ:
2306                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2307                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2308                                         TCP(ulp)->th_seq);
2309                                 break;
2310
2311                         case O_TCPACK:
2312                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2313                                     ((ipfw_insn_u32 *)cmd)->d[0] ==
2314                                         TCP(ulp)->th_ack);
2315                                 break;
2316
2317                         case O_TCPWIN:
2318                                 if (proto == IPPROTO_TCP && offset == 0) {
2319                                     uint16_t x;
2320                                     uint16_t *p;
2321                                     int i;
2322
2323                                     x = ntohs(TCP(ulp)->th_win);
2324                                     if (cmdlen == 1) {
2325                                         match = (cmd->arg1 == x);
2326                                         break;
2327                                     }
2328                                     /* Otherwise we have ranges. */
2329                                     p = ((ipfw_insn_u16 *)cmd)->ports;
2330                                     i = cmdlen - 1;
2331                                     for (; !match && i > 0; i--, p += 2)
2332                                         match = (x >= p[0] && x <= p[1]);
2333                                 }
2334                                 break;
2335
2336                         case O_ESTAB:
2337                                 /* reject packets which have SYN only */
2338                                 /* XXX should i also check for TH_ACK ? */
2339                                 match = (proto == IPPROTO_TCP && offset == 0 &&
2340                                     (TCP(ulp)->th_flags &
2341                                      (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2342                                 break;
2343
2344                         case O_ALTQ: {
2345                                 struct pf_mtag *at;
2346                                 struct m_tag *mtag;
2347                                 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2348
2349                                 /*
2350                                  * ALTQ uses mbuf tags from another
2351                                  * packet filtering system - pf(4).
2352                                  * We allocate a tag in its format
2353                                  * and fill it in, pretending to be pf(4).
2354                                  */
2355                                 match = 1;
2356                                 at = pf_find_mtag(m);
2357                                 if (at != NULL && at->qid != 0)
2358                                         break;
2359                                 mtag = m_tag_get(PACKET_TAG_PF,
2360                                     sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2361                                 if (mtag == NULL) {
2362                                         /*
2363                                          * Let the packet fall back to the
2364                                          * default ALTQ.
2365                                          */
2366                                         break;
2367                                 }
2368                                 m_tag_prepend(m, mtag);
2369                                 at = (struct pf_mtag *)(mtag + 1);
2370                                 at->qid = altq->qid;
2371                                 at->hdr = ip;
2372                                 break;
2373                         }
2374
2375                         case O_LOG:
2376                                 ipfw_log(chain, f, hlen, args,
2377                                     offset | ip6f_mf, tablearg, ip);
2378                                 match = 1;
2379                                 break;
2380
2381                         case O_PROB:
2382                                 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2383                                 break;
2384
2385                         case O_VERREVPATH:
2386                                 /* Outgoing packets automatically pass/match */
2387                                 match = (args->flags & IPFW_ARGS_OUT ||
2388                                     (
2389 #ifdef INET6
2390                                     is_ipv6 ?
2391                                         verify_path6(&(args->f_id.src_ip6),
2392                                             iif, args->f_id.fib) :
2393 #endif
2394                                     verify_path(src_ip, iif, args->f_id.fib)));
2395                                 break;
2396
2397                         case O_VERSRCREACH:
2398                                 /* Outgoing packets automatically pass/match */
2399                                 match = (hlen > 0 && ((oif != NULL) || (
2400 #ifdef INET6
2401                                     is_ipv6 ?
2402                                         verify_path6(&(args->f_id.src_ip6),
2403                                             NULL, args->f_id.fib) :
2404 #endif
2405                                     verify_path(src_ip, NULL, args->f_id.fib))));
2406                                 break;
2407
2408                         case O_ANTISPOOF:
2409                                 /* Outgoing packets automatically pass/match */
2410                                 if (oif == NULL && hlen > 0 &&
2411                                     (  (is_ipv4 && in_localaddr(src_ip))
2412 #ifdef INET6
2413                                     || (is_ipv6 &&
2414                                         in6_localaddr(&(args->f_id.src_ip6)))
2415 #endif
2416                                     ))
2417                                         match =
2418 #ifdef INET6
2419                                             is_ipv6 ? verify_path6(
2420                                                 &(args->f_id.src_ip6), iif,
2421                                                 args->f_id.fib) :
2422 #endif
2423                                             verify_path(src_ip, iif,
2424                                                 args->f_id.fib);
2425                                 else
2426                                         match = 1;
2427                                 break;
2428
2429                         case O_IPSEC:
2430                                 match = (m_tag_find(m,
2431                                     PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2432                                 /* otherwise no match */
2433                                 break;
2434
2435 #ifdef INET6
2436                         case O_IP6_SRC:
2437                                 match = is_ipv6 &&
2438                                     IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2439                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2440                                 break;
2441
2442                         case O_IP6_DST:
2443                                 match = is_ipv6 &&
2444                                 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2445                                     &((ipfw_insn_ip6 *)cmd)->addr6);
2446                                 break;
2447                         case O_IP6_SRC_MASK:
2448                         case O_IP6_DST_MASK:
2449                                 if (is_ipv6) {
2450                                         int i = cmdlen - 1;
2451                                         struct in6_addr p;
2452                                         struct in6_addr *d =
2453                                             &((ipfw_insn_ip6 *)cmd)->addr6;
2454
2455                                         for (; !match && i > 0; d += 2,
2456                                             i -= F_INSN_SIZE(struct in6_addr)
2457                                             * 2) {
2458                                                 p = (cmd->opcode ==
2459                                                     O_IP6_SRC_MASK) ?
2460                                                     args->f_id.src_ip6:
2461                                                     args->f_id.dst_ip6;
2462                                                 APPLY_MASK(&p, &d[1]);
2463                                                 match =
2464                                                     IN6_ARE_ADDR_EQUAL(&d[0],
2465                                                     &p);
2466                                         }
2467                                 }
2468                                 break;
2469
2470                         case O_FLOW6ID:
2471                                 match = is_ipv6 &&
2472                                     flow6id_match(args->f_id.flow_id6,
2473                                     (ipfw_insn_u32 *) cmd);
2474                                 break;
2475
2476                         case O_EXT_HDR:
2477                                 match = is_ipv6 &&
2478                                     (ext_hd & ((ipfw_insn *) cmd)->arg1);
2479                                 break;
2480
2481                         case O_IP6:
2482                                 match = is_ipv6;
2483                                 break;
2484 #endif
2485
2486                         case O_IP4:
2487                                 match = is_ipv4;
2488                                 break;
2489
2490                         case O_TAG: {
2491                                 struct m_tag *mtag;
2492                                 uint32_t tag = TARG(cmd->arg1, tag);
2493
2494                                 /* Packet is already tagged with this tag? */
2495                                 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2496
2497                                 /* We have `untag' action when F_NOT flag is
2498                                  * present. And we must remove this mtag from
2499                                  * mbuf and reset `match' to zero (`match' will
2500                                  * be inversed later).
2501                                  * Otherwise we should allocate new mtag and
2502                                  * push it into mbuf.
2503                                  */
2504                                 if (cmd->len & F_NOT) { /* `untag' action */
2505                                         if (mtag != NULL)
2506                                                 m_tag_delete(m, mtag);
2507                                         match = 0;
2508                                 } else {
2509                                         if (mtag == NULL) {
2510                                                 mtag = m_tag_alloc( MTAG_IPFW,
2511                                                     tag, 0, M_NOWAIT);
2512                                                 if (mtag != NULL)
2513                                                         m_tag_prepend(m, mtag);
2514                                         }
2515                                         match = 1;
2516                                 }
2517                                 break;
2518                         }
2519
2520                         case O_FIB: /* try match the specified fib */
2521                                 if (args->f_id.fib == cmd->arg1)
2522                                         match = 1;
2523                                 break;
2524
2525                         case O_SOCKARG: {
2526 #ifndef USERSPACE       /* not supported in userspace */
2527                                 struct inpcb *inp = args->inp;
2528                                 struct inpcbinfo *pi;
2529
2530                                 if (is_ipv6) /* XXX can we remove this ? */
2531                                         break;
2532
2533                                 if (proto == IPPROTO_TCP)
2534                                         pi = &V_tcbinfo;
2535                                 else if (proto == IPPROTO_UDP)
2536                                         pi = &V_udbinfo;
2537                                 else if (proto == IPPROTO_UDPLITE)
2538                                         pi = &V_ulitecbinfo;
2539                                 else
2540                                         break;
2541
2542                                 /*
2543                                  * XXXRW: so_user_cookie should almost
2544                                  * certainly be inp_user_cookie?
2545                                  */
2546
2547                                 /* For incoming packet, lookup up the
2548                                 inpcb using the src/dest ip/port tuple */
2549                                 if (inp == NULL) {
2550                                         inp = in_pcblookup(pi,
2551                                                 src_ip, htons(src_port),
2552                                                 dst_ip, htons(dst_port),
2553                                                 INPLOOKUP_RLOCKPCB, NULL);
2554                                         if (inp != NULL) {
2555                                                 tablearg =
2556                                                     inp->inp_socket->so_user_cookie;
2557                                                 if (tablearg)
2558                                                         match = 1;
2559                                                 INP_RUNLOCK(inp);
2560                                         }
2561                                 } else {
2562                                         if (inp->inp_socket) {
2563                                                 tablearg =
2564                                                     inp->inp_socket->so_user_cookie;
2565                                                 if (tablearg)
2566                                                         match = 1;
2567                                         }
2568                                 }
2569 #endif /* !USERSPACE */
2570                                 break;
2571                         }
2572
2573                         case O_TAGGED: {
2574                                 struct m_tag *mtag;
2575                                 uint32_t tag = TARG(cmd->arg1, tag);
2576
2577                                 if (cmdlen == 1) {
2578                                         match = m_tag_locate(m, MTAG_IPFW,
2579                                             tag, NULL) != NULL;
2580                                         break;
2581                                 }
2582
2583                                 /* we have ranges */
2584                                 for (mtag = m_tag_first(m);
2585                                     mtag != NULL && !match;
2586                                     mtag = m_tag_next(m, mtag)) {
2587                                         uint16_t *p;
2588                                         int i;
2589
2590                                         if (mtag->m_tag_cookie != MTAG_IPFW)
2591                                                 continue;
2592
2593                                         p = ((ipfw_insn_u16 *)cmd)->ports;
2594                                         i = cmdlen - 1;
2595                                         for(; !match && i > 0; i--, p += 2)
2596                                                 match =
2597                                                     mtag->m_tag_id >= p[0] &&
2598                                                     mtag->m_tag_id <= p[1];
2599                                 }
2600                                 break;
2601                         }
2602
2603                         /*
2604                          * The second set of opcodes represents 'actions',
2605                          * i.e. the terminal part of a rule once the packet
2606                          * matches all previous patterns.
2607                          * Typically there is only one action for each rule,
2608                          * and the opcode is stored at the end of the rule
2609                          * (but there are exceptions -- see below).
2610                          *
2611                          * In general, here we set retval and terminate the
2612                          * outer loop (would be a 'break 3' in some language,
2613                          * but we need to set l=0, done=1)
2614                          *
2615                          * Exceptions:
2616                          * O_COUNT and O_SKIPTO actions:
2617                          *   instead of terminating, we jump to the next rule
2618                          *   (setting l=0), or to the SKIPTO target (setting
2619                          *   f/f_len, cmd and l as needed), respectively.
2620                          *
2621                          * O_TAG, O_LOG and O_ALTQ action parameters:
2622                          *   perform some action and set match = 1;
2623                          *
2624                          * O_LIMIT and O_KEEP_STATE: these opcodes are
2625                          *   not real 'actions', and are stored right
2626                          *   before the 'action' part of the rule (one
2627                          *   exception is O_SKIP_ACTION which could be
2628                          *   between these opcodes and 'action' one).
2629                          *   These opcodes try to install an entry in the
2630                          *   state tables; if successful, we continue with
2631                          *   the next opcode (match=1; break;), otherwise
2632                          *   the packet must be dropped (set retval,
2633                          *   break loops with l=0, done=1)
2634                          *
2635                          * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2636                          *   cause a lookup of the state table, and a jump
2637                          *   to the 'action' part of the parent rule
2638                          *   if an entry is found, or
2639                          *   (CHECK_STATE only) a jump to the next rule if
2640                          *   the entry is not found.
2641                          *   The result of the lookup is cached so that
2642                          *   further instances of these opcodes become NOPs.
2643                          *   The jump to the next rule is done by setting
2644                          *   l=0, cmdlen=0.
2645                          *
2646                          * O_SKIP_ACTION: this opcode is not a real 'action'
2647                          *  either, and is stored right before the 'action'
2648                          *  part of the rule, right after the O_KEEP_STATE
2649                          *  opcode. It causes match failure so the real
2650                          *  'action' could be executed only if the rule
2651                          *  is checked via dynamic rule from the state
2652                          *  table, as in such case execution starts
2653                          *  from the true 'action' opcode directly.
2654                          *
2655                          */
2656                         case O_LIMIT:
2657                         case O_KEEP_STATE:
2658                                 if (ipfw_dyn_install_state(chain, f,
2659                                     (ipfw_insn_limit *)cmd, args, ulp,
2660                                     pktlen, &dyn_info, tablearg)) {
2661                                         /* error or limit violation */
2662                                         retval = IP_FW_DENY;
2663                                         l = 0;  /* exit inner loop */
2664                                         done = 1; /* exit outer loop */
2665                                 }
2666                                 match = 1;
2667                                 break;
2668
2669                         case O_PROBE_STATE:
2670                         case O_CHECK_STATE:
2671                                 /*
2672                                  * dynamic rules are checked at the first
2673                                  * keep-state or check-state occurrence,
2674                                  * with the result being stored in dyn_info.
2675                                  * The compiler introduces a PROBE_STATE
2676                                  * instruction for us when we have a
2677                                  * KEEP_STATE (because PROBE_STATE needs
2678                                  * to be run first).
2679                                  */
2680                                 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2681                                     (q = ipfw_dyn_lookup_state(args, ulp,
2682                                     pktlen, cmd, &dyn_info)) != NULL) {
2683                                         /*
2684                                          * Found dynamic entry, jump to the
2685                                          * 'action' part of the parent rule
2686                                          * by setting f, cmd, l and clearing
2687                                          * cmdlen.
2688                                          */
2689                                         f = q;
2690                                         f_pos = dyn_info.f_pos;
2691                                         cmd = ACTION_PTR(f);
2692                                         l = f->cmd_len - f->act_ofs;
2693                                         cmdlen = 0;
2694                                         match = 1;
2695                                         break;
2696                                 }
2697                                 /*
2698                                  * Dynamic entry not found. If CHECK_STATE,
2699                                  * skip to next rule, if PROBE_STATE just
2700                                  * ignore and continue with next opcode.
2701                                  */
2702                                 if (cmd->opcode == O_CHECK_STATE)
2703                                         l = 0;  /* exit inner loop */
2704                                 match = 1;
2705                                 break;
2706
2707                         case O_SKIP_ACTION:
2708                                 match = 0;      /* skip to the next rule */
2709                                 l = 0;          /* exit inner loop */
2710                                 break;
2711
2712                         case O_ACCEPT:
2713                                 retval = 0;     /* accept */
2714                                 l = 0;          /* exit inner loop */
2715                                 done = 1;       /* exit outer loop */
2716                                 break;
2717
2718                         case O_PIPE:
2719                         case O_QUEUE:
2720                                 set_match(args, f_pos, chain);
2721                                 args->rule.info = TARG(cmd->arg1, pipe);
2722                                 if (cmd->opcode == O_PIPE)
2723                                         args->rule.info |= IPFW_IS_PIPE;
2724                                 if (V_fw_one_pass)
2725                                         args->rule.info |= IPFW_ONEPASS;
2726                                 retval = IP_FW_DUMMYNET;
2727                                 l = 0;          /* exit inner loop */
2728                                 done = 1;       /* exit outer loop */
2729                                 break;
2730
2731                         case O_DIVERT:
2732                         case O_TEE:
2733                                 if (args->flags & IPFW_ARGS_ETHER)
2734                                         break;  /* not on layer 2 */
2735                                 /* otherwise this is terminal */
2736                                 l = 0;          /* exit inner loop */
2737                                 done = 1;       /* exit outer loop */
2738                                 retval = (cmd->opcode == O_DIVERT) ?
2739                                         IP_FW_DIVERT : IP_FW_TEE;
2740                                 set_match(args, f_pos, chain);
2741                                 args->rule.info = TARG(cmd->arg1, divert);
2742                                 break;
2743
2744                         case O_COUNT:
2745                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2746                                 l = 0;          /* exit inner loop */
2747                                 break;
2748
2749                         case O_SKIPTO:
2750                             IPFW_INC_RULE_COUNTER(f, pktlen);
2751                             f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2752                             /*
2753                              * Skip disabled rules, and re-enter
2754                              * the inner loop with the correct
2755                              * f_pos, f, l and cmd.
2756                              * Also clear cmdlen and skip_or
2757                              */
2758                             for (; f_pos < chain->n_rules - 1 &&
2759                                     (V_set_disable &
2760                                      (1 << chain->map[f_pos]->set));
2761                                     f_pos++)
2762                                 ;
2763                             /* Re-enter the inner loop at the skipto rule. */
2764                             f = chain->map[f_pos];
2765                             l = f->cmd_len;
2766                             cmd = f->cmd;
2767                             match = 1;
2768                             cmdlen = 0;
2769                             skip_or = 0;
2770                             continue;
2771                             break;      /* not reached */
2772
2773                         case O_CALLRETURN: {
2774                                 /*
2775                                  * Implementation of `subroutine' call/return,
2776                                  * in the stack carried in an mbuf tag. This
2777                                  * is different from `skipto' in that any call
2778                                  * address is possible (`skipto' must prevent
2779                                  * backward jumps to avoid endless loops).
2780                                  * We have `return' action when F_NOT flag is
2781                                  * present. The `m_tag_id' field is used as
2782                                  * stack pointer.
2783                                  */
2784                                 struct m_tag *mtag;
2785                                 uint16_t jmpto, *stack;
2786
2787 #define IS_CALL         ((cmd->len & F_NOT) == 0)
2788 #define IS_RETURN       ((cmd->len & F_NOT) != 0)
2789                                 /*
2790                                  * Hand-rolled version of m_tag_locate() with
2791                                  * wildcard `type'.
2792                                  * If not already tagged, allocate new tag.
2793                                  */
2794                                 mtag = m_tag_first(m);
2795                                 while (mtag != NULL) {
2796                                         if (mtag->m_tag_cookie ==
2797                                             MTAG_IPFW_CALL)
2798                                                 break;
2799                                         mtag = m_tag_next(m, mtag);
2800                                 }
2801                                 if (mtag == NULL && IS_CALL) {
2802                                         mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2803                                             IPFW_CALLSTACK_SIZE *
2804                                             sizeof(uint16_t), M_NOWAIT);
2805                                         if (mtag != NULL)
2806                                                 m_tag_prepend(m, mtag);
2807                                 }
2808
2809                                 /*
2810                                  * On error both `call' and `return' just
2811                                  * continue with next rule.
2812                                  */
2813                                 if (IS_RETURN && (mtag == NULL ||
2814                                     mtag->m_tag_id == 0)) {
2815                                         l = 0;          /* exit inner loop */
2816                                         break;
2817                                 }
2818                                 if (IS_CALL && (mtag == NULL ||
2819                                     mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2820                                         printf("ipfw: call stack error, "
2821                                             "go to next rule\n");
2822                                         l = 0;          /* exit inner loop */
2823                                         break;
2824                                 }
2825
2826                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2827                                 stack = (uint16_t *)(mtag + 1);
2828
2829                                 /*
2830                                  * The `call' action may use cached f_pos
2831                                  * (in f->next_rule), whose version is written
2832                                  * in f->next_rule.
2833                                  * The `return' action, however, doesn't have
2834                                  * fixed jump address in cmd->arg1 and can't use
2835                                  * cache.
2836                                  */
2837                                 if (IS_CALL) {
2838                                         stack[mtag->m_tag_id] = f->rulenum;
2839                                         mtag->m_tag_id++;
2840                                         f_pos = JUMP(chain, f, cmd->arg1,
2841                                             tablearg, 1);
2842                                 } else {        /* `return' action */
2843                                         mtag->m_tag_id--;
2844                                         jmpto = stack[mtag->m_tag_id] + 1;
2845                                         f_pos = ipfw_find_rule(chain, jmpto, 0);
2846                                 }
2847
2848                                 /*
2849                                  * Skip disabled rules, and re-enter
2850                                  * the inner loop with the correct
2851                                  * f_pos, f, l and cmd.
2852                                  * Also clear cmdlen and skip_or
2853                                  */
2854                                 for (; f_pos < chain->n_rules - 1 &&
2855                                     (V_set_disable &
2856                                     (1 << chain->map[f_pos]->set)); f_pos++)
2857                                         ;
2858                                 /* Re-enter the inner loop at the dest rule. */
2859                                 f = chain->map[f_pos];
2860                                 l = f->cmd_len;
2861                                 cmd = f->cmd;
2862                                 cmdlen = 0;
2863                                 skip_or = 0;
2864                                 continue;
2865                                 break;  /* NOTREACHED */
2866                         }
2867 #undef IS_CALL
2868 #undef IS_RETURN
2869
2870                         case O_REJECT:
2871                                 /*
2872                                  * Drop the packet and send a reject notice
2873                                  * if the packet is not ICMP (or is an ICMP
2874                                  * query), and it is not multicast/broadcast.
2875                                  */
2876                                 if (hlen > 0 && is_ipv4 && offset == 0 &&
2877                                     (proto != IPPROTO_ICMP ||
2878                                      is_icmp_query(ICMP(ulp))) &&
2879                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2880                                     !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2881                                         send_reject(args, cmd->arg1, iplen, ip);
2882                                         m = args->m;
2883                                 }
2884                                 /* FALLTHROUGH */
2885 #ifdef INET6
2886                         case O_UNREACH6:
2887                                 if (hlen > 0 && is_ipv6 &&
2888                                     ((offset & IP6F_OFF_MASK) == 0) &&
2889                                     (proto != IPPROTO_ICMPV6 ||
2890                                      (is_icmp6_query(icmp6_type) == 1)) &&
2891                                     !(m->m_flags & (M_BCAST|M_MCAST)) &&
2892                                     !IN6_IS_ADDR_MULTICAST(
2893                                         &args->f_id.dst_ip6)) {
2894                                         send_reject6(args,
2895                                             cmd->opcode == O_REJECT ?
2896                                             map_icmp_unreach(cmd->arg1):
2897                                             cmd->arg1, hlen,
2898                                             (struct ip6_hdr *)ip);
2899                                         m = args->m;
2900                                 }
2901                                 /* FALLTHROUGH */
2902 #endif
2903                         case O_DENY:
2904                                 retval = IP_FW_DENY;
2905                                 l = 0;          /* exit inner loop */
2906                                 done = 1;       /* exit outer loop */
2907                                 break;
2908
2909                         case O_FORWARD_IP:
2910                                 if (args->flags & IPFW_ARGS_ETHER)
2911                                         break;  /* not valid on layer2 pkts */
2912                                 if (q != f ||
2913                                     dyn_info.direction == MATCH_FORWARD) {
2914                                     struct sockaddr_in *sa;
2915
2916                                     sa = &(((ipfw_insn_sa *)cmd)->sa);
2917                                     if (sa->sin_addr.s_addr == INADDR_ANY) {
2918 #ifdef INET6
2919                                         /*
2920                                          * We use O_FORWARD_IP opcode for
2921                                          * fwd rule with tablearg, but tables
2922                                          * now support IPv6 addresses. And
2923                                          * when we are inspecting IPv6 packet,
2924                                          * we can use nh6 field from
2925                                          * table_value as next_hop6 address.
2926                                          */
2927                                         if (is_ipv6) {
2928                                                 struct ip_fw_nh6 *nh6;
2929
2930                                                 args->flags |= IPFW_ARGS_NH6;
2931                                                 nh6 = &args->hopstore6;
2932                                                 nh6->sin6_addr = TARG_VAL(
2933                                                     chain, tablearg, nh6);
2934                                                 nh6->sin6_port = sa->sin_port;
2935                                                 nh6->sin6_scope_id = TARG_VAL(
2936                                                     chain, tablearg, zoneid);
2937                                         } else
2938 #endif
2939                                         {
2940                                                 args->flags |= IPFW_ARGS_NH4;
2941                                                 args->hopstore.sin_port =
2942                                                     sa->sin_port;
2943                                                 sa = &args->hopstore;
2944                                                 sa->sin_family = AF_INET;
2945                                                 sa->sin_len = sizeof(*sa);
2946                                                 sa->sin_addr.s_addr = htonl(
2947                                                     TARG_VAL(chain, tablearg,
2948                                                     nh4));
2949                                         }
2950                                     } else {
2951                                             args->flags |= IPFW_ARGS_NH4PTR;
2952                                             args->next_hop = sa;
2953                                     }
2954                                 }
2955                                 retval = IP_FW_PASS;
2956                                 l = 0;          /* exit inner loop */
2957                                 done = 1;       /* exit outer loop */
2958                                 break;
2959
2960 #ifdef INET6
2961                         case O_FORWARD_IP6:
2962                                 if (args->flags & IPFW_ARGS_ETHER)
2963                                         break;  /* not valid on layer2 pkts */
2964                                 if (q != f ||
2965                                     dyn_info.direction == MATCH_FORWARD) {
2966                                         struct sockaddr_in6 *sin6;
2967
2968                                         sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2969                                         args->flags |= IPFW_ARGS_NH6PTR;
2970                                         args->next_hop6 = sin6;
2971                                 }
2972                                 retval = IP_FW_PASS;
2973                                 l = 0;          /* exit inner loop */
2974                                 done = 1;       /* exit outer loop */
2975                                 break;
2976 #endif
2977
2978                         case O_NETGRAPH:
2979                         case O_NGTEE:
2980                                 set_match(args, f_pos, chain);
2981                                 args->rule.info = TARG(cmd->arg1, netgraph);
2982                                 if (V_fw_one_pass)
2983                                         args->rule.info |= IPFW_ONEPASS;
2984                                 retval = (cmd->opcode == O_NETGRAPH) ?
2985                                     IP_FW_NETGRAPH : IP_FW_NGTEE;
2986                                 l = 0;          /* exit inner loop */
2987                                 done = 1;       /* exit outer loop */
2988                                 break;
2989
2990                         case O_SETFIB: {
2991                                 uint32_t fib;
2992
2993                                 IPFW_INC_RULE_COUNTER(f, pktlen);
2994                                 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2995                                 if (fib >= rt_numfibs)
2996                                         fib = 0;
2997                                 M_SETFIB(m, fib);
2998                                 args->f_id.fib = fib; /* XXX */
2999                                 l = 0;          /* exit inner loop */
3000                                 break;
3001                         }
3002
3003                         case O_SETDSCP: {
3004                                 uint16_t code;
3005
3006                                 code = TARG(cmd->arg1, dscp) & 0x3F;
3007                                 l = 0;          /* exit inner loop */
3008                                 if (is_ipv4) {
3009                                         uint16_t old;
3010
3011                                         old = *(uint16_t *)ip;
3012                                         ip->ip_tos = (code << 2) |
3013                                             (ip->ip_tos & 0x03);
3014                                         ip->ip_sum = cksum_adjust(ip->ip_sum,
3015                                             old, *(uint16_t *)ip);
3016                                 } else if (is_ipv6) {
3017                                         uint8_t *v;
3018
3019                                         v = &((struct ip6_hdr *)ip)->ip6_vfc;
3020                                         *v = (*v & 0xF0) | (code >> 2);
3021                                         v++;
3022                                         *v = (*v & 0x3F) | ((code & 0x03) << 6);
3023                                 } else
3024                                         break;
3025
3026                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3027                                 break;
3028                         }
3029
3030                         case O_NAT:
3031                                 l = 0;          /* exit inner loop */
3032                                 done = 1;       /* exit outer loop */
3033                                 /*
3034                                  * Ensure that we do not invoke NAT handler for
3035                                  * non IPv4 packets. Libalias expects only IPv4.
3036                                  */
3037                                 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3038                                     retval = IP_FW_DENY;
3039                                     break;
3040                                 }
3041
3042                                 struct cfg_nat *t;
3043                                 int nat_id;
3044
3045                                 args->rule.info = 0;
3046                                 set_match(args, f_pos, chain);
3047                                 /* Check if this is 'global' nat rule */
3048                                 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3049                                         retval = ipfw_nat_ptr(args, NULL, m);
3050                                         break;
3051                                 }
3052                                 t = ((ipfw_insn_nat *)cmd)->nat;
3053                                 if (t == NULL) {
3054                                         nat_id = TARG(cmd->arg1, nat);
3055                                         t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3056
3057                                         if (t == NULL) {
3058                                             retval = IP_FW_DENY;
3059                                             break;
3060                                         }
3061                                         if (cmd->arg1 != IP_FW_TARG)
3062                                             ((ipfw_insn_nat *)cmd)->nat = t;
3063                                 }
3064                                 retval = ipfw_nat_ptr(args, t, m);
3065                                 break;
3066
3067                         case O_REASS: {
3068                                 int ip_off;
3069
3070                                 l = 0;  /* in any case exit inner loop */
3071                                 if (is_ipv6) /* IPv6 is not supported yet */
3072                                         break;
3073                                 IPFW_INC_RULE_COUNTER(f, pktlen);
3074                                 ip_off = ntohs(ip->ip_off);
3075
3076                                 /* if not fragmented, go to next rule */
3077                                 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3078                                     break;
3079
3080                                 args->m = m = ip_reass(m);
3081
3082                                 /*
3083                                  * do IP header checksum fixup.
3084                                  */
3085                                 if (m == NULL) { /* fragment got swallowed */
3086                                     retval = IP_FW_DENY;
3087                                 } else { /* good, packet complete */
3088                                     int hlen;
3089
3090                                     ip = mtod(m, struct ip *);
3091                                     hlen = ip->ip_hl << 2;
3092                                     ip->ip_sum = 0;
3093                                     if (hlen == sizeof(struct ip))
3094                                         ip->ip_sum = in_cksum_hdr(ip);
3095                                     else
3096                                         ip->ip_sum = in_cksum(m, hlen);
3097                                     retval = IP_FW_REASS;
3098                                     args->rule.info = 0;
3099                                     set_match(args, f_pos, chain);
3100                                 }
3101                                 done = 1;       /* exit outer loop */
3102                                 break;
3103                         }
3104                         case O_EXTERNAL_ACTION:
3105                                 l = 0; /* in any case exit inner loop */
3106                                 retval = ipfw_run_eaction(chain, args,
3107                                     cmd, &done);
3108                                 /*
3109                                  * If both @retval and @done are zero,
3110                                  * consider this as rule matching and
3111                                  * update counters.
3112                                  */
3113                                 if (retval == 0 && done == 0) {
3114                                         IPFW_INC_RULE_COUNTER(f, pktlen);
3115                                         /*
3116                                          * Reset the result of the last
3117                                          * dynamic state lookup.
3118                                          * External action can change
3119                                          * @args content, and it may be
3120                                          * used for new state lookup later.
3121                                          */
3122                                         DYN_INFO_INIT(&dyn_info);
3123                                 }
3124                                 break;
3125
3126                         default:
3127                                 panic("-- unknown opcode %d\n", cmd->opcode);
3128                         } /* end of switch() on opcodes */
3129                         /*
3130                          * if we get here with l=0, then match is irrelevant.
3131                          */
3132
3133                         if (cmd->len & F_NOT)
3134                                 match = !match;
3135
3136                         if (match) {
3137                                 if (cmd->len & F_OR)
3138                                         skip_or = 1;
3139                         } else {
3140                                 if (!(cmd->len & F_OR)) /* not an OR block, */
3141                                         break;          /* try next rule    */
3142                         }
3143
3144                 }       /* end of inner loop, scan opcodes */
3145 #undef PULLUP_LEN
3146
3147                 if (done)
3148                         break;
3149
3150 /* next_rule:; */       /* try next rule                */
3151
3152         }               /* end of outer for, scan rules */
3153
3154         if (done) {
3155                 struct ip_fw *rule = chain->map[f_pos];
3156                 /* Update statistics */
3157                 IPFW_INC_RULE_COUNTER(rule, pktlen);
3158         } else {
3159                 retval = IP_FW_DENY;
3160                 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3161         }
3162         IPFW_PF_RUNLOCK(chain);
3163 #ifdef __FreeBSD__
3164         if (ucred_cache != NULL)
3165                 crfree(ucred_cache);
3166 #endif
3167         return (retval);
3168
3169 pullup_failed:
3170         if (V_fw_verbose)
3171                 printf("ipfw: pullup failed\n");
3172         return (IP_FW_DENY);
3173 }
3174
3175 /*
3176  * Set maximum number of tables that can be used in given VNET ipfw instance.
3177  */
3178 #ifdef SYSCTL_NODE
3179 static int
3180 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3181 {
3182         int error;
3183         unsigned int ntables;
3184
3185         ntables = V_fw_tables_max;
3186
3187         error = sysctl_handle_int(oidp, &ntables, 0, req);
3188         /* Read operation or some error */
3189         if ((error != 0) || (req->newptr == NULL))
3190                 return (error);
3191
3192         return (ipfw_resize_tables(&V_layer3_chain, ntables));
3193 }
3194
3195 /*
3196  * Switches table namespace between global and per-set.
3197  */
3198 static int
3199 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3200 {
3201         int error;
3202         unsigned int sets;
3203
3204         sets = V_fw_tables_sets;
3205
3206         error = sysctl_handle_int(oidp, &sets, 0, req);
3207         /* Read operation or some error */
3208         if ((error != 0) || (req->newptr == NULL))
3209                 return (error);
3210
3211         return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3212 }
3213 #endif
3214
3215 /*
3216  * Module and VNET glue
3217  */
3218
3219 /*
3220  * Stuff that must be initialised only on boot or module load
3221  */
3222 static int
3223 ipfw_init(void)
3224 {
3225         int error = 0;
3226
3227         /*
3228          * Only print out this stuff the first time around,
3229          * when called from the sysinit code.
3230          */
3231         printf("ipfw2 "
3232 #ifdef INET6
3233                 "(+ipv6) "
3234 #endif
3235                 "initialized, divert %s, nat %s, "
3236                 "default to %s, logging ",
3237 #ifdef IPDIVERT
3238                 "enabled",
3239 #else
3240                 "loadable",
3241 #endif
3242 #ifdef IPFIREWALL_NAT
3243                 "enabled",
3244 #else
3245                 "loadable",
3246 #endif
3247                 default_to_accept ? "accept" : "deny");
3248
3249         /*
3250          * Note: V_xxx variables can be accessed here but the vnet specific
3251          * initializer may not have been called yet for the VIMAGE case.
3252          * Tuneables will have been processed. We will print out values for
3253          * the default vnet.
3254          * XXX This should all be rationalized AFTER 8.0
3255          */
3256         if (V_fw_verbose == 0)
3257                 printf("disabled\n");
3258         else if (V_verbose_limit == 0)
3259                 printf("unlimited\n");
3260         else
3261                 printf("limited to %d packets/entry by default\n",
3262                     V_verbose_limit);
3263
3264         /* Check user-supplied table count for validness */
3265         if (default_fw_tables > IPFW_TABLES_MAX)
3266           default_fw_tables = IPFW_TABLES_MAX;
3267
3268         ipfw_init_sopt_handler();
3269         ipfw_init_obj_rewriter();
3270         ipfw_iface_init();
3271         return (error);
3272 }
3273
3274 /*
3275  * Called for the removal of the last instance only on module unload.
3276  */
3277 static void
3278 ipfw_destroy(void)
3279 {
3280
3281         ipfw_iface_destroy();
3282         ipfw_destroy_sopt_handler();
3283         ipfw_destroy_obj_rewriter();
3284         printf("IP firewall unloaded\n");
3285 }
3286
3287 /*
3288  * Stuff that must be initialized for every instance
3289  * (including the first of course).
3290  */
3291 static int
3292 vnet_ipfw_init(const void *unused)
3293 {
3294         int error, first;
3295         struct ip_fw *rule = NULL;
3296         struct ip_fw_chain *chain;
3297
3298         chain = &V_layer3_chain;
3299
3300         first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3301
3302         /* First set up some values that are compile time options */
3303         V_autoinc_step = 100;   /* bounded to 1..1000 in add_rule() */
3304         V_fw_deny_unknown_exthdrs = 1;
3305 #ifdef IPFIREWALL_VERBOSE
3306         V_fw_verbose = 1;
3307 #endif
3308 #ifdef IPFIREWALL_VERBOSE_LIMIT
3309         V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3310 #endif
3311 #ifdef IPFIREWALL_NAT
3312         LIST_INIT(&chain->nat);
3313 #endif
3314
3315         /* Init shared services hash table */
3316         ipfw_init_srv(chain);
3317
3318         ipfw_init_counters();
3319         /* Set initial number of tables */
3320         V_fw_tables_max = default_fw_tables;
3321         error = ipfw_init_tables(chain, first);
3322         if (error) {
3323                 printf("ipfw2: setting up tables failed\n");
3324                 free(chain->map, M_IPFW);
3325                 free(rule, M_IPFW);
3326                 return (ENOSPC);
3327         }
3328
3329         IPFW_LOCK_INIT(chain);
3330
3331         /* fill and insert the default rule */
3332         rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3333         rule->cmd_len = 1;
3334         rule->cmd[0].len = 1;
3335         rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3336         chain->default_rule = rule;
3337         ipfw_add_protected_rule(chain, rule, 0);
3338
3339         ipfw_dyn_init(chain);
3340         ipfw_eaction_init(chain, first);
3341 #ifdef LINEAR_SKIPTO
3342         ipfw_init_skipto_cache(chain);
3343 #endif
3344         ipfw_bpf_init(first);
3345
3346         /* First set up some values that are compile time options */
3347         V_ipfw_vnet_ready = 1;          /* Open for business */
3348
3349         /*
3350          * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3351          * Even if the latter two fail we still keep the module alive
3352          * because the sockopt and layer2 paths are still useful.
3353          * ipfw[6]_hook return 0 on success, ENOENT on failure,
3354          * so we can ignore the exact return value and just set a flag.
3355          *
3356          * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3357          * changes in the underlying (per-vnet) variables trigger
3358          * immediate hook()/unhook() calls.
3359          * In layer2 we have the same behaviour, except that V_ether_ipfw
3360          * is checked on each packet because there are no pfil hooks.
3361          */
3362         V_ip_fw_ctl_ptr = ipfw_ctl3;
3363         error = ipfw_attach_hooks(1);
3364         return (error);
3365 }
3366
3367 /*
3368  * Called for the removal of each instance.
3369  */
3370 static int
3371 vnet_ipfw_uninit(const void *unused)
3372 {
3373         struct ip_fw *reap;
3374         struct ip_fw_chain *chain = &V_layer3_chain;
3375         int i, last;
3376
3377         V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3378         /*
3379          * disconnect from ipv4, ipv6, layer2 and sockopt.
3380          * Then grab, release and grab again the WLOCK so we make
3381          * sure the update is propagated and nobody will be in.
3382          */
3383         (void)ipfw_attach_hooks(0 /* detach */);
3384         V_ip_fw_ctl_ptr = NULL;
3385
3386         last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3387
3388         IPFW_UH_WLOCK(chain);
3389         IPFW_UH_WUNLOCK(chain);
3390
3391         ipfw_dyn_uninit(0);     /* run the callout_drain */
3392
3393         IPFW_UH_WLOCK(chain);
3394
3395         reap = NULL;
3396         IPFW_WLOCK(chain);
3397         for (i = 0; i < chain->n_rules; i++)
3398                 ipfw_reap_add(chain, &reap, chain->map[i]);
3399         free(chain->map, M_IPFW);
3400 #ifdef LINEAR_SKIPTO
3401         ipfw_destroy_skipto_cache(chain);
3402 #endif
3403         IPFW_WUNLOCK(chain);
3404         IPFW_UH_WUNLOCK(chain);
3405         ipfw_destroy_tables(chain, last);
3406         ipfw_eaction_uninit(chain, last);
3407         if (reap != NULL)
3408                 ipfw_reap_rules(reap);
3409         vnet_ipfw_iface_destroy(chain);
3410         ipfw_destroy_srv(chain);
3411         IPFW_LOCK_DESTROY(chain);
3412         ipfw_dyn_uninit(1);     /* free the remaining parts */
3413         ipfw_destroy_counters();
3414         ipfw_bpf_uninit(last);
3415         return (0);
3416 }
3417
3418 /*
3419  * Module event handler.
3420  * In general we have the choice of handling most of these events by the
3421  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3422  * use the SYSINIT handlers as they are more capable of expressing the
3423  * flow of control during module and vnet operations, so this is just
3424  * a skeleton. Note there is no SYSINIT equivalent of the module
3425  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3426  */
3427 static int
3428 ipfw_modevent(module_t mod, int type, void *unused)
3429 {
3430         int err = 0;
3431
3432         switch (type) {
3433         case MOD_LOAD:
3434                 /* Called once at module load or
3435                  * system boot if compiled in. */
3436                 break;
3437         case MOD_QUIESCE:
3438                 /* Called before unload. May veto unloading. */
3439                 break;
3440         case MOD_UNLOAD:
3441                 /* Called during unload. */
3442                 break;
3443         case MOD_SHUTDOWN:
3444                 /* Called during system shutdown. */
3445                 break;
3446         default:
3447                 err = EOPNOTSUPP;
3448                 break;
3449         }
3450         return err;
3451 }
3452
3453 static moduledata_t ipfwmod = {
3454         "ipfw",
3455         ipfw_modevent,
3456         0
3457 };
3458
3459 /* Define startup order. */
3460 #define IPFW_SI_SUB_FIREWALL    SI_SUB_PROTO_FIREWALL
3461 #define IPFW_MODEVENT_ORDER     (SI_ORDER_ANY - 255) /* On boot slot in here. */
3462 #define IPFW_MODULE_ORDER       (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3463 #define IPFW_VNET_ORDER         (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3464
3465 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3466 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3467 MODULE_VERSION(ipfw, 3);
3468 /* should declare some dependencies here */
3469
3470 /*
3471  * Starting up. Done in order after ipfwmod() has been called.
3472  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3473  */
3474 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3475             ipfw_init, NULL);
3476 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3477             vnet_ipfw_init, NULL);
3478
3479 /*
3480  * Closing up shop. These are done in REVERSE ORDER, but still
3481  * after ipfwmod() has been called. Not called on reboot.
3482  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3483  * or when the module is unloaded.
3484  */
3485 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3486             ipfw_destroy, NULL);
3487 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3488             vnet_ipfw_uninit, NULL);
3489 /* end of file */