1 /*
   2  * Copyright (C) 1993-2003 by Darren Reed.
   3  *
   4  * See the IPFILTER.LICENCE file for details on licencing.
   5  *
   6  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
   7  *
   8  * Copyright (c) 2014, Joyent, Inc.  All rights reserved.
   9  */
  10 
  11 #if defined(KERNEL) || defined(_KERNEL)
  12 # undef KERNEL
  13 # undef _KERNEL
  14 # define        KERNEL  1
  15 # define        _KERNEL 1
  16 #endif
  17 #include <sys/errno.h>
  18 #include <sys/types.h>
  19 #include <sys/param.h>
  20 #include <sys/time.h>
  21 #if defined(__NetBSD__)
  22 # if (NetBSD >= 199905) && !defined(IPFILTER_LKM) && defined(_KERNEL)
  23 #  include "opt_ipfilter_log.h"
  24 # endif
  25 #endif
  26 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
  27     (__FreeBSD_version >= 220000)
  28 # if (__FreeBSD_version >= 400000)
  29 #  if !defined(IPFILTER_LKM)
  30 #   include "opt_inet6.h"
  31 #  endif
  32 #  if (__FreeBSD_version == 400019)
  33 #   define CSUM_DELAY_DATA
  34 #  endif
  35 # endif
  36 # include <sys/filio.h>
  37 #else
  38 # include <sys/ioctl.h>
  39 #endif
  40 #if !defined(_AIX51)
  41 # include <sys/fcntl.h>
  42 #endif
  43 #if defined(_KERNEL)
  44 # include <sys/systm.h>
  45 # include <sys/file.h>
  46 #else
  47 # include <stdio.h>
  48 # include <string.h>
  49 # include <stdlib.h>
  50 # include <stddef.h>
  51 # include <sys/file.h>
  52 # define _KERNEL
  53 # ifdef __OpenBSD__
  54 struct file;
  55 # endif
  56 # include <sys/uio.h>
  57 # undef _KERNEL
  58 #endif
  59 #if !defined(__SVR4) && !defined(__svr4__) && !defined(__hpux) && \
  60     !defined(linux)
  61 # include <sys/mbuf.h>
  62 #else
  63 # if !defined(linux)
  64 #  include <sys/byteorder.h>
  65 # endif
  66 # if (SOLARIS2 < 5) && defined(sun)
  67 #  include <sys/dditypes.h>
  68 # endif
  69 #endif
  70 #ifdef __hpux
  71 # define _NET_ROUTE_INCLUDED
  72 #endif
  73 #if !defined(linux)
  74 # include <sys/protosw.h>
  75 #endif
  76 #include <sys/socket.h>
  77 #include <net/if.h>
  78 #ifdef sun
  79 # include <net/af.h>
  80 #endif
  81 #if !defined(_KERNEL) && defined(__FreeBSD__)
  82 # include "radix_ipf.h"
  83 #endif
  84 #include <net/route.h>
  85 #include <netinet/in.h>
  86 #include <netinet/in_systm.h>
  87 #include <netinet/ip.h>
  88 #if !defined(linux)
  89 # include <netinet/ip_var.h>
  90 #endif
  91 #if defined(__sgi) && defined(IFF_DRVRLOCK) /* IRIX 6 */
  92 # include <sys/hashing.h>
  93 # include <netinet/in_var.h>
  94 #endif
  95 #include <netinet/tcp.h>
  96 #if (!defined(__sgi) && !defined(AIX)) || defined(_KERNEL)
  97 # include <netinet/udp.h>
  98 # include <netinet/ip_icmp.h>
  99 #endif
 100 #ifdef __hpux
 101 # undef _NET_ROUTE_INCLUDED
 102 #endif
 103 #include "netinet/ip_compat.h"
 104 #ifdef  USE_INET6
 105 # include <netinet/icmp6.h>
 106 # if !defined(SOLARIS) && defined(_KERNEL) && !defined(__osf__) && \
 107         !defined(__hpux)
 108 #  include <netinet6/in6_var.h>
 109 # endif
 110 #endif
 111 #include <netinet/tcpip.h>
 112 #include "netinet/ip_fil.h"
 113 #include "netinet/ip_nat.h"
 114 #include "netinet/ip_frag.h"
 115 #include "netinet/ip_state.h"
 116 #include "netinet/ip_proxy.h"
 117 #include "netinet/ip_auth.h"
 118 #include "netinet/ipf_stack.h"
 119 #ifdef IPFILTER_SCAN
 120 # include "netinet/ip_scan.h"
 121 #endif
 122 #ifdef IPFILTER_SYNC
 123 # include "netinet/ip_sync.h"
 124 #endif
 125 #include "netinet/ip_pool.h"
 126 #include "netinet/ip_htable.h"
 127 #ifdef IPFILTER_COMPILED
 128 # include "netinet/ip_rules.h"
 129 #endif
 130 #if defined(IPFILTER_BPF) && defined(_KERNEL)
 131 # include <net/bpf.h>
 132 #endif
 133 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
 134 # include <sys/malloc.h>
 135 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
 136 #  include "opt_ipfilter.h"
 137 # endif
 138 #endif
 139 #include "netinet/ipl.h"
 140 #if defined(_KERNEL)
 141 #include <sys/sunddi.h>
 142 #endif
 143 /* END OF INCLUDES */
 144 
 145 #if !defined(lint)
 146 static const char sccsid[] = "@(#)fil.c 1.36 6/5/96 (C) 1993-2000 Darren Reed";
 147 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
 148 #endif
 149 
 150 #ifndef _KERNEL
 151 # include "ipf.h"
 152 # include "ipt.h"
 153 # include "bpf-ipf.h"
 154 extern  int     opts;
 155 
 156 # define        FR_VERBOSE(verb_pr)                     verbose verb_pr
 157 # define        FR_DEBUG(verb_pr)                       debug verb_pr
 158 #else /* #ifndef _KERNEL */
 159 # define        FR_VERBOSE(verb_pr)
 160 # define        FR_DEBUG(verb_pr)
 161 #endif /* _KERNEL */
 162 
 163 
 164 char    ipfilter_version[] = IPL_VERSION;
 165 int     fr_features = 0
 166 #ifdef  IPFILTER_LKM
 167                 | IPF_FEAT_LKM
 168 #endif
 169 #ifdef  IPFILTER_LOG
 170                 | IPF_FEAT_LOG
 171 #endif
 172 #ifdef  IPFILTER_LOOKUP
 173                 | IPF_FEAT_LOOKUP
 174 #endif
 175 #ifdef  IPFILTER_BPF
 176                 | IPF_FEAT_BPF
 177 #endif
 178 #ifdef  IPFILTER_COMPILED
 179                 | IPF_FEAT_COMPILED
 180 #endif
 181 #ifdef  IPFILTER_CKSUM
 182                 | IPF_FEAT_CKSUM
 183 #endif
 184 #ifdef  IPFILTER_SYNC
 185                 | IPF_FEAT_SYNC
 186 #endif
 187 #ifdef  IPFILTER_SCAN
 188                 | IPF_FEAT_SCAN
 189 #endif
 190 #ifdef  USE_INET6
 191                 | IPF_FEAT_IPV6
 192 #endif
 193         ;
 194 
 195 #define IPF_BUMP(x)     (x)++   
 196 
 197 static  INLINE int      fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
 198 static  INLINE int      fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
 199 static  int             fr_portcheck __P((frpcmp_t *, u_short *));
 200 static  int             frflushlist __P((int, minor_t, int *, frentry_t **,
 201                                          ipf_stack_t *));
 202 static  ipfunc_t        fr_findfunc __P((ipfunc_t));
 203 static  frentry_t       *fr_firewall __P((fr_info_t *, u_32_t *));
 204 static  int             fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
 205 static  INLINE void     frpr_ah __P((fr_info_t *));
 206 static  INLINE void     frpr_esp __P((fr_info_t *));
 207 static  INLINE void     frpr_gre __P((fr_info_t *));
 208 static  INLINE void     frpr_udp __P((fr_info_t *));
 209 static  INLINE void     frpr_tcp __P((fr_info_t *));
 210 static  INLINE void     frpr_icmp __P((fr_info_t *));
 211 static  INLINE void     frpr_ipv4hdr __P((fr_info_t *));
 212 static  INLINE int      frpr_pullup __P((fr_info_t *, int));
 213 static  INLINE void     frpr_short __P((fr_info_t *, int));
 214 static  INLINE void     frpr_tcpcommon __P((fr_info_t *));
 215 static  INLINE void     frpr_udpcommon __P((fr_info_t *));
 216 static  INLINE int      fr_updateipid __P((fr_info_t *));
 217 #ifdef  IPFILTER_LOOKUP
 218 static  int             fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
 219 static  INLINE void     *fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
 220                                                ipf_stack_t *));
 221 #endif
 222 static  void            frsynclist __P((int, int, void *, char *, frentry_t *,
 223     ipf_stack_t *));
 224 static  void            *fr_ifsync __P((int, int, char *, char *,
 225                                         void *, void *, ipf_stack_t *));
 226 static  ipftuneable_t   *fr_findtunebyname __P((const char *, ipf_stack_t *));
 227 static  ipftuneable_t   *fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
 228 
 229 /*
 230  * bit values for identifying presence of individual IP options
 231  * All of these tables should be ordered by increasing key value on the left
 232  * hand side to allow for binary searching of the array and include a trailer
 233  * with a 0 for the bitmask for linear searches to easily find the end with.
 234  */
 235 const   struct  optlist ipopts[20] = {
 236         { IPOPT_NOP,    0x000001 },
 237         { IPOPT_RR,     0x000002 },
 238         { IPOPT_ZSU,    0x000004 },
 239         { IPOPT_MTUP,   0x000008 },
 240         { IPOPT_MTUR,   0x000010 },
 241         { IPOPT_ENCODE, 0x000020 },
 242         { IPOPT_TS,     0x000040 },
 243         { IPOPT_TR,     0x000080 },
 244         { IPOPT_SECURITY, 0x000100 },
 245         { IPOPT_LSRR,   0x000200 },
 246         { IPOPT_E_SEC,  0x000400 },
 247         { IPOPT_CIPSO,  0x000800 },
 248         { IPOPT_SATID,  0x001000 },
 249         { IPOPT_SSRR,   0x002000 },
 250         { IPOPT_ADDEXT, 0x004000 },
 251         { IPOPT_VISA,   0x008000 },
 252         { IPOPT_IMITD,  0x010000 },
 253         { IPOPT_EIP,    0x020000 },
 254         { IPOPT_FINN,   0x040000 },
 255         { 0,            0x000000 }
 256 };
 257 
 258 #ifdef USE_INET6
 259 struct optlist ip6exthdr[] = {
 260         { IPPROTO_HOPOPTS,              0x000001 },
 261         { IPPROTO_IPV6,                 0x000002 },
 262         { IPPROTO_ROUTING,              0x000004 },
 263         { IPPROTO_FRAGMENT,             0x000008 },
 264         { IPPROTO_ESP,                  0x000010 },
 265         { IPPROTO_AH,                   0x000020 },
 266         { IPPROTO_NONE,                 0x000040 },
 267         { IPPROTO_DSTOPTS,              0x000080 },
 268         { 0,                            0 }
 269 };
 270 #endif
 271 
 272 struct optlist tcpopts[] = {
 273         { TCPOPT_NOP,                   0x000001 },
 274         { TCPOPT_MAXSEG,                0x000002 },
 275         { TCPOPT_WINDOW,                0x000004 },
 276         { TCPOPT_SACK_PERMITTED,        0x000008 },
 277         { TCPOPT_SACK,                  0x000010 },
 278         { TCPOPT_TIMESTAMP,             0x000020 },
 279         { 0,                            0x000000 }
 280 };
 281 
 282 /*
 283  * bit values for identifying presence of individual IP security options
 284  */
 285 const   struct  optlist secopt[8] = {
 286         { IPSO_CLASS_RES4,      0x01 },
 287         { IPSO_CLASS_TOPS,      0x02 },
 288         { IPSO_CLASS_SECR,      0x04 },
 289         { IPSO_CLASS_RES3,      0x08 },
 290         { IPSO_CLASS_CONF,      0x10 },
 291         { IPSO_CLASS_UNCL,      0x20 },
 292         { IPSO_CLASS_RES2,      0x40 },
 293         { IPSO_CLASS_RES1,      0x80 }
 294 };
 295 
 296 
 297 /*
 298  * Table of functions available for use with call rules.
 299  */
 300 static ipfunc_resolve_t fr_availfuncs[] = {
 301 #ifdef  IPFILTER_LOOKUP
 302         { "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
 303         { "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
 304 #endif
 305         { "", NULL }
 306 };
 307 
 308 
 309 /*
 310  * Below we declare a list of constants used only by the ipf_extraflush()
 311  * routine.  We are placing it here, instead of in ipf_extraflush() itself,
 312  * because we want to make it visible to tools such as mdb, nm etc., so the
 313  * values can easily be altered during debugging.
 314  */
 315 static  const   int     idletime_tab[] = {
 316         IPF_TTLVAL(30),         /* 30 seconds */
 317         IPF_TTLVAL(1800),       /* 30 minutes */
 318         IPF_TTLVAL(43200),      /* 12 hours */
 319         IPF_TTLVAL(345600),     /* 4 days */
 320 };
 321 
 322 
 323 /*
 324  * The next section of code is a a collection of small routines that set
 325  * fields in the fr_info_t structure passed based on properties of the
 326  * current packet.  There are different routines for the same protocol
 327  * for each of IPv4 and IPv6.  Adding a new protocol, for which there
 328  * will "special" inspection for setup, is now more easily done by adding
 329  * a new routine and expanding the frpr_ipinit*() function rather than by
 330  * adding more code to a growing switch statement.
 331  */
 332 #ifdef USE_INET6
 333 static  INLINE int      frpr_ah6 __P((fr_info_t *));
 334 static  INLINE void     frpr_esp6 __P((fr_info_t *));
 335 static  INLINE void     frpr_gre6 __P((fr_info_t *));
 336 static  INLINE void     frpr_udp6 __P((fr_info_t *));
 337 static  INLINE void     frpr_tcp6 __P((fr_info_t *));
 338 static  INLINE void     frpr_icmp6 __P((fr_info_t *));
 339 static  INLINE void     frpr_ipv6hdr __P((fr_info_t *));
 340 static  INLINE void     frpr_short6 __P((fr_info_t *, int));
 341 static  INLINE int      frpr_hopopts6 __P((fr_info_t *));
 342 static  INLINE int      frpr_routing6 __P((fr_info_t *));
 343 static  INLINE int      frpr_dstopts6 __P((fr_info_t *));
 344 static  INLINE int      frpr_fragment6 __P((fr_info_t *));
 345 static  INLINE int      frpr_ipv6exthdr __P((fr_info_t *, int, int));
 346 
 347 
 348 /* ------------------------------------------------------------------------ */
 349 /* Function:    frpr_short6                                                 */
 350 /* Returns:     void                                                        */
 351 /* Parameters:  fin(I) - pointer to packet information                      */
 352 /*                                                                          */
 353 /* IPv6 Only                                                                */
 354 /* This is function enforces the 'is a packet too short to be legit' rule   */
 355 /* for IPv6 and marks the packet with FI_SHORT if so.  See function comment */
 356 /* for frpr_short() for more details.                                       */
 357 /* ------------------------------------------------------------------------ */
 358 static INLINE void frpr_short6(fin, xmin)
 359 fr_info_t *fin;
 360 int xmin;
 361 {
 362 
 363         if (fin->fin_dlen < xmin)
 364                 fin->fin_flx |= FI_SHORT;
 365 }
 366 
 367 
 368 /* ------------------------------------------------------------------------ */
 369 /* Function:    frpr_ipv6hdr                                                */
 370 /* Returns:     Nil                                                         */
 371 /* Parameters:  fin(I) - pointer to packet information                      */
 372 /*                                                                          */
 373 /* IPv6 Only                                                                */
 374 /* Copy values from the IPv6 header into the fr_info_t struct and call the  */
 375 /* per-protocol analyzer if it exists.                                      */
 376 /* ------------------------------------------------------------------------ */
 377 static INLINE void frpr_ipv6hdr(fin)
 378 fr_info_t *fin;
 379 {
 380         ip6_t *ip6 = (ip6_t *)fin->fin_ip;
 381         int p, go = 1, i, hdrcount;
 382         fr_ip_t *fi = &fin->fin_fi;
 383 
 384         fin->fin_off = 0;
 385 
 386         fi->fi_tos = 0;
 387         fi->fi_optmsk = 0;
 388         fi->fi_secmsk = 0;
 389         fi->fi_auth = 0;
 390 
 391         p = ip6->ip6_nxt;
 392         fi->fi_ttl = ip6->ip6_hlim;
 393         fi->fi_src.in6 = ip6->ip6_src;
 394         fi->fi_dst.in6 = ip6->ip6_dst;
 395         fin->fin_id = 0;
 396 
 397         hdrcount = 0;
 398         while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
 399                 switch (p)
 400                 {
 401                 case IPPROTO_UDP :
 402                         frpr_udp6(fin);
 403                         go = 0;
 404                         break;
 405 
 406                 case IPPROTO_TCP :
 407                         frpr_tcp6(fin);
 408                         go = 0;
 409                         break;
 410 
 411                 case IPPROTO_ICMPV6 :
 412                         frpr_icmp6(fin);
 413                         go = 0;
 414                         break;
 415 
 416                 case IPPROTO_GRE :
 417                         frpr_gre6(fin);
 418                         go = 0;
 419                         break;
 420 
 421                 case IPPROTO_HOPOPTS :
 422                         /*
 423                          * hop by hop ext header is only allowed
 424                          * right after IPv6 header.
 425                          */
 426                         if (hdrcount != 0) {
 427                                 fin->fin_flx |= FI_BAD;
 428                                 p = IPPROTO_NONE;
 429                         } else {
 430                                 p = frpr_hopopts6(fin);
 431                         }
 432                         break;
 433 
 434                 case IPPROTO_DSTOPTS :
 435                         p = frpr_dstopts6(fin);
 436                         break;
 437 
 438                 case IPPROTO_ROUTING :
 439                         p = frpr_routing6(fin);
 440                         break;
 441 
 442                 case IPPROTO_AH :
 443                         p = frpr_ah6(fin);
 444                         break;
 445 
 446                 case IPPROTO_ESP :
 447                         frpr_esp6(fin);
 448                         go = 0;
 449                         break;
 450 
 451                 case IPPROTO_IPV6 :
 452                         for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
 453                                 if (ip6exthdr[i].ol_val == p) {
 454                                         fin->fin_flx |= ip6exthdr[i].ol_bit;
 455                                         break;
 456                                 }
 457                         go = 0;
 458                         break;
 459 
 460                 case IPPROTO_NONE :
 461                         go = 0;
 462                         break;
 463 
 464                 case IPPROTO_FRAGMENT :
 465                         p = frpr_fragment6(fin);
 466                         if (fin->fin_off != 0)  /* Not the first frag */
 467                                 go = 0;
 468                         break;
 469 
 470                 default :
 471                         go = 0;
 472                         break;
 473                 }
 474                 hdrcount++;
 475 
 476                 /*
 477                  * It is important to note that at this point, for the
 478                  * extension headers (go != 0), the entire header may not have
 479                  * been pulled up when the code gets to this point.  This is
 480                  * only done for "go != 0" because the other header handlers
 481                  * will all pullup their complete header.  The other indicator
 482                  * of an incomplete packet is that this was just an extension
 483                  * header.
 484                  */
 485                 if ((go != 0) && (p != IPPROTO_NONE) &&
 486                     (frpr_pullup(fin, 0) == -1)) {
 487                         p = IPPROTO_NONE;
 488                         go = 0;
 489                 }
 490         }
 491         fi->fi_p = p;
 492 }
 493 
 494 
 495 /* ------------------------------------------------------------------------ */
 496 /* Function:    frpr_ipv6exthdr                                             */
 497 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
 498 /* Parameters:  fin(I)      - pointer to packet information                 */
 499 /*              multiple(I) - flag indicating yes/no if multiple occurances */
 500 /*                            of this extension header are allowed.         */
 501 /*              proto(I)    - protocol number for this extension header     */
 502 /*                                                                          */
 503 /* IPv6 Only                                                                */
 504 /* This function expects to find an IPv6 extension header at fin_dp.        */
 505 /* There must be at least 8 bytes of data at fin_dp for there to be a valid */
 506 /* extension header present. If a good one is found, fin_dp is advanced to  */
 507 /* point at the first piece of data after the extension header, fin_exthdr  */
 508 /* points to the start of the extension header and the "protocol" of the    */
 509 /* *NEXT* header is returned.                                               */
 510 /* ------------------------------------------------------------------------ */
 511 static INLINE int frpr_ipv6exthdr(fin, multiple, proto)
 512 fr_info_t *fin;
 513 int multiple, proto;
 514 {
 515         struct ip6_ext *hdr;
 516         u_short shift;
 517         int i;
 518 
 519         fin->fin_flx |= FI_V6EXTHDR;
 520 
 521                                 /* 8 is default length of extension hdr */
 522         if ((fin->fin_dlen - 8) < 0) {
 523                 fin->fin_flx |= FI_SHORT;
 524                 return IPPROTO_NONE;
 525         }
 526 
 527         if (frpr_pullup(fin, 8) == -1)
 528                 return IPPROTO_NONE;
 529 
 530         hdr = fin->fin_dp;
 531         shift = 8 + (hdr->ip6e_len << 3);
 532         if (shift > fin->fin_dlen) {      /* Nasty extension header length? */
 533                 fin->fin_flx |= FI_BAD;
 534                 return IPPROTO_NONE;
 535         }
 536 
 537         for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
 538                 if (ip6exthdr[i].ol_val == proto) {
 539                         /*
 540                          * Most IPv6 extension headers are only allowed once.
 541                          */
 542                         if ((multiple == 0) &&
 543                             ((fin->fin_optmsk & ip6exthdr[i].ol_bit) != 0))
 544                                 fin->fin_flx |= FI_BAD;
 545                         else
 546                                 fin->fin_optmsk |= ip6exthdr[i].ol_bit;
 547                         break;
 548                 }
 549 
 550         fin->fin_dp = (char *)fin->fin_dp + shift;
 551         fin->fin_dlen -= shift;
 552 
 553         return hdr->ip6e_nxt;
 554 }
 555 
 556 
 557 /* ------------------------------------------------------------------------ */
 558 /* Function:    frpr_hopopts6                                               */
 559 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
 560 /* Parameters:  fin(I) - pointer to packet information                      */
 561 /*                                                                          */
 562 /* IPv6 Only                                                                */
 563 /* This is function checks pending hop by hop options extension header      */
 564 /* ------------------------------------------------------------------------ */
 565 static INLINE int frpr_hopopts6(fin)
 566 fr_info_t *fin;
 567 {
 568         return frpr_ipv6exthdr(fin, 0, IPPROTO_HOPOPTS);
 569 }
 570 
 571 
 572 /* ------------------------------------------------------------------------ */
 573 /* Function:    frpr_routing6                                               */
 574 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
 575 /* Parameters:  fin(I) - pointer to packet information                      */
 576 /*                                                                          */
 577 /* IPv6 Only                                                                */
 578 /* This is function checks pending routing extension header                 */
 579 /* ------------------------------------------------------------------------ */
 580 static INLINE int frpr_routing6(fin)
 581 fr_info_t *fin;
 582 {
 583         struct ip6_ext *hdr;
 584         int shift;
 585 
 586         hdr = fin->fin_dp;
 587         if (frpr_ipv6exthdr(fin, 0, IPPROTO_ROUTING) == IPPROTO_NONE)
 588                 return IPPROTO_NONE;
 589 
 590         shift = 8 + (hdr->ip6e_len << 3);
 591         /*
 592          * Nasty extension header length?
 593          */
 594         if ((hdr->ip6e_len << 3) & 15) {
 595                 fin->fin_flx |= FI_BAD;
 596                 /*
 597                  * Compensate for the changes made in frpr_ipv6exthdr()
 598                  */
 599                 fin->fin_dlen += shift;
 600                 fin->fin_dp = (char *)fin->fin_dp - shift;
 601                 return IPPROTO_NONE;
 602         }
 603 
 604         return hdr->ip6e_nxt;
 605 }
 606 
 607 
 608 /* ------------------------------------------------------------------------ */
 609 /* Function:    frpr_fragment6                                              */
 610 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
 611 /* Parameters:  fin(I) - pointer to packet information                      */
 612 /*                                                                          */
 613 /* IPv6 Only                                                                */
 614 /* Examine the IPv6 fragment header and extract fragment offset information.*/
 615 /*                                                                          */
 616 /* We don't know where the transport layer header (or whatever is next is), */
 617 /* as it could be behind destination options (amongst others).  Because     */
 618 /* there is no fragment cache, there is no knowledge about whether or not an*/
 619 /* upper layer header has been seen (or where it ends) and thus we are not  */
 620 /* able to continue processing beyond this header with any confidence.      */
 621 /* ------------------------------------------------------------------------ */
 622 static INLINE int frpr_fragment6(fin)
 623 fr_info_t *fin;
 624 {
 625         struct ip6_frag *frag;
 626 
 627         fin->fin_flx |= FI_FRAG;
 628 
 629         /*
 630          * A fragmented IPv6 packet implies that there must be something
 631          * else after the fragment.
 632          */
 633         if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
 634                 return IPPROTO_NONE;
 635 
 636         frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
 637 
 638         /*
 639          * If this fragment isn't the last then the packet length must
 640          * be a multiple of 8.
 641          */
 642         if ((frag->ip6f_offlg & IP6F_MORE_FRAG) != 0) {
 643                 fin->fin_flx |= FI_MOREFRAG;
 644 
 645                 if ((fin->fin_plen & 0x7) != 0)
 646                         fin->fin_flx |= FI_BAD;
 647         }
 648 
 649         fin->fin_id = frag->ip6f_ident;
 650         fin->fin_off = ntohs(frag->ip6f_offlg & IP6F_OFF_MASK);
 651         if (fin->fin_off != 0)
 652                 fin->fin_flx |= FI_FRAGBODY;
 653 
 654         return frag->ip6f_nxt;
 655 }
 656 
 657 
 658 /* ------------------------------------------------------------------------ */
 659 /* Function:    frpr_dstopts6                                               */
 660 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
 661 /* Parameters:  fin(I) - pointer to packet information                      */
 662 /*              nextheader(I) - stores next header value                    */
 663 /*                                                                          */
 664 /* IPv6 Only                                                                */
 665 /* This is function checks pending destination options extension header     */
 666 /* ------------------------------------------------------------------------ */
 667 static INLINE int frpr_dstopts6(fin)
 668 fr_info_t *fin;
 669 {
 670         return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
 671 }
 672 
 673 
 674 /* ------------------------------------------------------------------------ */
 675 /* Function:    frpr_icmp6                                                  */
 676 /* Returns:     void                                                        */
 677 /* Parameters:  fin(I) - pointer to packet information                      */
 678 /*                                                                          */
 679 /* IPv6 Only                                                                */
 680 /* This routine is mainly concerned with determining the minimum valid size */
 681 /* for an ICMPv6 packet.                                                    */
 682 /* ------------------------------------------------------------------------ */
 683 static INLINE void frpr_icmp6(fin)
 684 fr_info_t *fin;
 685 {
 686         int minicmpsz = sizeof(struct icmp6_hdr);
 687         struct icmp6_hdr *icmp6;
 688 
 689         if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
 690                 return;
 691 
 692         if (fin->fin_dlen > 1) {
 693                 icmp6 = fin->fin_dp;
 694 
 695                 fin->fin_data[0] = *(u_short *)icmp6;
 696 
 697                 if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
 698                         fin->fin_flx |= FI_ICMPQUERY;
 699 
 700                 switch (icmp6->icmp6_type)
 701                 {
 702                 case ICMP6_ECHO_REPLY :
 703                 case ICMP6_ECHO_REQUEST :
 704                         if (fin->fin_dlen >= 6)
 705                                 fin->fin_data[1] = icmp6->icmp6_id;
 706                         minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
 707                         break;
 708                 case ICMP6_DST_UNREACH :
 709                 case ICMP6_PACKET_TOO_BIG :
 710                 case ICMP6_TIME_EXCEEDED :
 711                 case ICMP6_PARAM_PROB :
 712                         if ((fin->fin_m != NULL) &&
 713                             (M_LEN(fin->fin_m) < fin->fin_plen)) {
 714                                 if (fr_coalesce(fin) != 1)
 715                                         return;
 716                         }
 717                         fin->fin_flx |= FI_ICMPERR;
 718                         minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
 719                         break;
 720                 default :
 721                         break;
 722                 }
 723         }
 724 
 725         frpr_short6(fin, minicmpsz);
 726 }
 727 
 728 
 729 /* ------------------------------------------------------------------------ */
 730 /* Function:    frpr_udp6                                                   */
 731 /* Returns:     void                                                        */
 732 /* Parameters:  fin(I) - pointer to packet information                      */
 733 /*                                                                          */
 734 /* IPv6 Only                                                                */
 735 /* Analyse the packet for IPv6/UDP properties.                              */
 736 /* Is not expected to be called for fragmented packets.                     */
 737 /* ------------------------------------------------------------------------ */
 738 static INLINE void frpr_udp6(fin)
 739 fr_info_t *fin;
 740 {
 741 
 742         fr_checkv6sum(fin);
 743 
 744         frpr_short6(fin, sizeof(struct udphdr));
 745         if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
 746                 return;
 747 
 748         frpr_udpcommon(fin);
 749 }
 750 
 751 
 752 /* ------------------------------------------------------------------------ */
 753 /* Function:    frpr_tcp6                                                   */
 754 /* Returns:     void                                                        */
 755 /* Parameters:  fin(I) - pointer to packet information                      */
 756 /*                                                                          */
 757 /* IPv6 Only                                                                */
 758 /* Analyse the packet for IPv6/TCP properties.                              */
 759 /* Is not expected to be called for fragmented packets.                     */
 760 /* ------------------------------------------------------------------------ */
 761 static INLINE void frpr_tcp6(fin)
 762 fr_info_t *fin;
 763 {
 764 
 765         fr_checkv6sum(fin);
 766 
 767         frpr_short6(fin, sizeof(struct tcphdr));
 768         if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
 769                 return;
 770 
 771         frpr_tcpcommon(fin);
 772 }
 773 
 774 
 775 /* ------------------------------------------------------------------------ */
 776 /* Function:    frpr_esp6                                                   */
 777 /* Returns:     void                                                        */
 778 /* Parameters:  fin(I) - pointer to packet information                      */
 779 /*                                                                          */
 780 /* IPv6 Only                                                                */
 781 /* Analyse the packet for ESP properties.                                   */
 782 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
 783 /* even though the newer ESP packets must also have a sequence number that  */
 784 /* is 32bits as well, it is not possible(?) to determine the version from a */
 785 /* simple packet header.                                                    */
 786 /* ------------------------------------------------------------------------ */
 787 static INLINE void frpr_esp6(fin)
 788 fr_info_t *fin;
 789 {
 790         int i;
 791         frpr_short6(fin, sizeof(grehdr_t));
 792 
 793         (void) frpr_pullup(fin, 8);
 794 
 795         for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
 796                 if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
 797                         fin->fin_optmsk |= ip6exthdr[i].ol_bit;
 798                         break;                  
 799                 }
 800 }
 801 
 802 
 803 /* ------------------------------------------------------------------------ */
 804 /* Function:    frpr_ah6                                                    */
 805 /* Returns:     void                                                        */
 806 /* Parameters:  fin(I) - pointer to packet information                      */
 807 /*                                                                          */
 808 /* IPv6 Only                                                                */
 809 /* Analyse the packet for AH properties.                                    */
 810 /* The minimum length is taken to be the combination of all fields in the   */
 811 /* header being present and no authentication data (null algorithm used.)   */
 812 /* ------------------------------------------------------------------------ */
 813 static INLINE int frpr_ah6(fin)
 814 fr_info_t *fin;
 815 {
 816         authhdr_t *ah;
 817         int i, shift;
 818 
 819         frpr_short6(fin, 12);
 820 
 821         if (frpr_pullup(fin, sizeof(*ah)) == -1)
 822                 return IPPROTO_NONE;
 823 
 824         for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
 825                 if (ip6exthdr[i].ol_val == IPPROTO_AH) {
 826                         fin->fin_optmsk |= ip6exthdr[i].ol_bit;
 827                         break;                  
 828                 }
 829 
 830         ah = (authhdr_t *)fin->fin_dp;
 831 
 832         shift = (ah->ah_plen + 2) * 4;
 833         fin->fin_dlen -= shift;
 834         fin->fin_dp = (char*)fin->fin_dp + shift;
 835 
 836         return ah->ah_next;
 837 }
 838 
 839 
 840 /* ------------------------------------------------------------------------ */
 841 /* Function:    frpr_gre6                                                   */
 842 /* Returns:     void                                                        */
 843 /* Parameters:  fin(I) - pointer to packet information                      */
 844 /*                                                                          */
 845 /* Analyse the packet for GRE properties.                                   */
 846 /* ------------------------------------------------------------------------ */
 847 static INLINE void frpr_gre6(fin)
 848 fr_info_t *fin;
 849 {
 850         grehdr_t *gre;
 851 
 852         frpr_short6(fin, sizeof(grehdr_t));
 853 
 854         if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
 855                 return;
 856 
 857         gre = fin->fin_dp;
 858         if (GRE_REV(gre->gr_flags) == 1)
 859                 fin->fin_data[0] = gre->gr_call;
 860 }
 861 #endif  /* USE_INET6 */
 862 
 863 
 864 /* ------------------------------------------------------------------------ */
 865 /* Function:    frpr_pullup                                                 */
 866 /* Returns:     int     - 0 == pullup succeeded, -1 == failure              */
 867 /* Parameters:  fin(I)  - pointer to packet information                     */
 868 /*              plen(I) - length (excluding L3 header) to pullup            */
 869 /*                                                                          */
 870 /* Short inline function to cut down on code duplication to perform a call  */
 871 /* to fr_pullup to ensure there is the required amount of data,             */
 872 /* consecutively in the packet buffer.                                      */
 873 /* ------------------------------------------------------------------------ */
 874 static INLINE int frpr_pullup(fin, plen)
 875 fr_info_t *fin;
 876 int plen;
 877 {
 878 #if defined(_KERNEL)
 879         if (fin->fin_m != NULL) {
 880                 int ipoff;
 881 
 882                 ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
 883 
 884                 if (fin->fin_dp != NULL)
 885                         plen += (char *)fin->fin_dp -
 886                                 ((char *)fin->fin_ip + fin->fin_hlen);
 887                 plen += fin->fin_hlen;
 888                 /*
 889                  * We don't do 'plen += ipoff;' here. The fr_pullup() will
 890                  * do it for us.
 891                  */
 892                 if (M_LEN(fin->fin_m) < plen + ipoff) {
 893                         if (fr_pullup(fin->fin_m, fin, plen) == NULL)
 894                                 return -1;
 895                 }
 896         }
 897 #endif
 898         return 0;
 899 }
 900 
 901 
 902 /* ------------------------------------------------------------------------ */
 903 /* Function:    frpr_short                                                  */
 904 /* Returns:     void                                                        */
 905 /* Parameters:  fin(I)  - pointer to packet information                     */
 906 /*              xmin(I) - minimum header size                               */
 907 /*                                                                          */
 908 /* Check if a packet is "short" as defined by xmin.  The rule we are        */
 909 /* applying here is that the packet must not be fragmented within the layer */
 910 /* 4 header.  That is, it must not be a fragment that has its offset set to */
 911 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the    */
 912 /* entire layer 4 header must be present (min).                             */
 913 /* ------------------------------------------------------------------------ */
 914 static INLINE void frpr_short(fin, xmin)
 915 fr_info_t *fin;
 916 int xmin;
 917 {
 918 
 919         if (fin->fin_off == 0) {
 920                 if (fin->fin_dlen < xmin)
 921                         fin->fin_flx |= FI_SHORT;
 922         } else if (fin->fin_off < xmin) {
 923                 fin->fin_flx |= FI_SHORT;
 924         }
 925 }
 926 
 927 
 928 /* ------------------------------------------------------------------------ */
 929 /* Function:    frpr_icmp                                                   */
 930 /* Returns:     void                                                        */
 931 /* Parameters:  fin(I) - pointer to packet information                      */
 932 /*                                                                          */
 933 /* IPv4 Only                                                                */
 934 /* Do a sanity check on the packet for ICMP (v4).  In nearly all cases,     */
 935 /* except extrememly bad packets, both type and code will be present.       */
 936 /* The expected minimum size of an ICMP packet is very much dependent on    */
 937 /* the type of it.                                                          */
 938 /*                                                                          */
 939 /* XXX - other ICMP sanity checks?                                          */
 940 /* ------------------------------------------------------------------------ */
 941 static INLINE void frpr_icmp(fin)
 942 fr_info_t *fin;
 943 {
 944         int minicmpsz = sizeof(struct icmp);
 945         icmphdr_t *icmp;
 946         ip_t *oip;
 947         ipf_stack_t *ifs = fin->fin_ifs;
 948 
 949         if (fin->fin_off != 0) {
 950                 frpr_short(fin, ICMPERR_ICMPHLEN);
 951                 return;
 952         }
 953 
 954         if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
 955                 return;
 956 
 957         fr_checkv4sum(fin);
 958 
 959         /*
 960          * This is a right place to set icmp pointer, since the memory
 961          * referenced by fin_dp could get reallocated. The code down below can
 962          * rely on fact icmp variable always points to ICMP header.
 963          */
 964         icmp = fin->fin_dp;
 965         fin->fin_data[0] = *(u_short *)icmp;
 966         fin->fin_data[1] = icmp->icmp_id;
 967 
 968         switch (icmp->icmp_type)
 969         {
 970         case ICMP_ECHOREPLY :
 971         case ICMP_ECHO :
 972         /* Router discovery messaes - RFC 1256 */
 973         case ICMP_ROUTERADVERT :
 974         case ICMP_ROUTERSOLICIT :
 975                 minicmpsz = ICMP_MINLEN;
 976                 break;
 977         /*
 978          * type(1) + code(1) + cksum(2) + id(2) seq(2) +
 979          * 3 * timestamp(3 * 4)
 980          */
 981         case ICMP_TSTAMP :
 982         case ICMP_TSTAMPREPLY :
 983                 minicmpsz = 20;
 984                 break;
 985         /*
 986          * type(1) + code(1) + cksum(2) + id(2) seq(2) +
 987          * mask(4)
 988          */
 989         case ICMP_MASKREQ :
 990         case ICMP_MASKREPLY :
 991                         minicmpsz = 12;
 992                         break;
 993         /*
 994          * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
 995          */
 996         case ICMP_UNREACH :
 997                 if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
 998                         if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
 999                                 fin->fin_flx |= FI_BAD;
1000                 }
1001                 /* FALLTHRU */
1002         case ICMP_SOURCEQUENCH :
1003         case ICMP_REDIRECT :
1004         case ICMP_TIMXCEED :
1005         case ICMP_PARAMPROB :
1006                 fin->fin_flx |= FI_ICMPERR;
1007                 if (fr_coalesce(fin) != 1)
1008                         return;
1009                 /*
1010                  * ICMP error packets should not be generated for IP
1011                  * packets that are a fragment that isn't the first
1012                  * fragment.
1013                  */
1014                 oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1015                 if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1016                         fin->fin_flx |= FI_BAD;
1017                 break;
1018         default :
1019                 break;
1020         }
1021 
1022         frpr_short(fin, minicmpsz);
1023 }
1024 
1025 
1026 /* ------------------------------------------------------------------------ */
1027 /* Function:    frpr_tcpcommon                                              */
1028 /* Returns:     void                                                        */
1029 /* Parameters:  fin(I) - pointer to packet information                      */
1030 /*                                                                          */
1031 /* TCP header sanity checking.  Look for bad combinations of TCP flags,     */
1032 /* and make some checks with how they interact with other fields.           */
1033 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is     */
1034 /* valid and mark the packet as bad if not.                                 */
1035 /* ------------------------------------------------------------------------ */
1036 static INLINE void frpr_tcpcommon(fin)
1037 fr_info_t *fin;
1038 {
1039         int flags, tlen;
1040         tcphdr_t *tcp;
1041 
1042         fin->fin_flx |= FI_TCPUDP;
1043         if (fin->fin_off != 0)
1044                 return;
1045 
1046         if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1047                 return;
1048         tcp = fin->fin_dp;
1049 
1050         if (fin->fin_dlen > 3) {
1051                 fin->fin_sport = ntohs(tcp->th_sport);
1052                 fin->fin_dport = ntohs(tcp->th_dport);
1053         }
1054 
1055         if ((fin->fin_flx & FI_SHORT) != 0)
1056                 return;
1057 
1058         /*
1059          * Use of the TCP data offset *must* result in a value that is at
1060          * least the same size as the TCP header.
1061          */
1062         tlen = TCP_OFF(tcp) << 2;
1063         if (tlen < sizeof(tcphdr_t)) {
1064                 fin->fin_flx |= FI_BAD;
1065                 return;
1066         }
1067 
1068         flags = tcp->th_flags;
1069         fin->fin_tcpf = tcp->th_flags;
1070 
1071         /*
1072          * If the urgent flag is set, then the urgent pointer must
1073          * also be set and vice versa.  Good TCP packets do not have
1074          * just one of these set.
1075          */
1076         if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1077                 fin->fin_flx |= FI_BAD;
1078         } else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1079                 /* Ignore this case, it shows up in "real" traffic with */
1080                 /* bogus values in the urgent pointer field. */
1081                 flags = flags; /* LINT */
1082         } else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1083                    ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1084                 /* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1085                 fin->fin_flx |= FI_BAD;
1086         } else if (!(flags & TH_ACK)) {
1087                 /*
1088                  * If the ack bit isn't set, then either the SYN or
1089                  * RST bit must be set.  If the SYN bit is set, then
1090                  * we expect the ACK field to be 0.  If the ACK is
1091                  * not set and if URG, PSH or FIN are set, consdier
1092                  * that to indicate a bad TCP packet.
1093                  */
1094                 if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1095                         /*
1096                          * Cisco PIX sets the ACK field to a random value.
1097                          * In light of this, do not set FI_BAD until a patch
1098                          * is available from Cisco to ensure that
1099                          * interoperability between existing systems is
1100                          * achieved.
1101                          */
1102                         /*fin->fin_flx |= FI_BAD*/;
1103                         flags = flags; /* LINT */
1104                 } else if (!(flags & (TH_RST|TH_SYN))) {
1105                         fin->fin_flx |= FI_BAD;
1106                 } else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1107                         fin->fin_flx |= FI_BAD;
1108                 }
1109         }
1110 
1111         /*
1112          * At this point, it's not exactly clear what is to be gained by
1113          * marking up which TCP options are and are not present.  The one we
1114          * are most interested in is the TCP window scale.  This is only in
1115          * a SYN packet [RFC1323] so we don't need this here...?
1116          * Now if we were to analyse the header for passive fingerprinting,
1117          * then that might add some weight to adding this...
1118          */
1119         if (tlen == sizeof(tcphdr_t))
1120                 return;
1121 
1122         if (frpr_pullup(fin, tlen) == -1)
1123                 return;
1124 
1125 #if 0
1126         ip = fin->fin_ip;
1127         s = (u_char *)(tcp + 1);
1128         off = IP_HL(ip) << 2;
1129 # ifdef _KERNEL
1130         if (fin->fin_mp != NULL) {
1131                 mb_t *m = *fin->fin_mp;
1132 
1133                 if (off + tlen > M_LEN(m))
1134                         return;
1135         }
1136 # endif
1137         for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1138                 opt = *s;
1139                 if (opt == '\0')
1140                         break;
1141                 else if (opt == TCPOPT_NOP)
1142                         ol = 1;
1143                 else {
1144                         if (tlen < 2)
1145                                 break;
1146                         ol = (int)*(s + 1);
1147                         if (ol < 2 || ol > tlen)
1148                                 break;
1149                 }
1150 
1151                 for (i = 9, mv = 4; mv >= 0; ) {
1152                         op = ipopts + i;
1153                         if (opt == (u_char)op->ol_val) {
1154                                 optmsk |= op->ol_bit;
1155                                 break;
1156                         }
1157                 }
1158                 tlen -= ol;
1159                 s += ol;
1160         }
1161 #endif /* 0 */
1162 }
1163 
1164 
1165 
1166 /* ------------------------------------------------------------------------ */
1167 /* Function:    frpr_udpcommon                                              */
1168 /* Returns:     void                                                        */
1169 /* Parameters:  fin(I) - pointer to packet information                      */
1170 /*                                                                          */
1171 /* Extract the UDP source and destination ports, if present.  If compiled   */
1172 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid.          */
1173 /* ------------------------------------------------------------------------ */
1174 static INLINE void frpr_udpcommon(fin)
1175 fr_info_t *fin;
1176 {
1177         udphdr_t *udp;
1178 
1179         fin->fin_flx |= FI_TCPUDP;
1180 
1181         if (!fin->fin_off && (fin->fin_dlen > 3)) {
1182                 if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1183                         fin->fin_flx |= FI_SHORT;
1184                         return;
1185                 }
1186 
1187                 udp = fin->fin_dp;
1188 
1189                 fin->fin_sport = ntohs(udp->uh_sport);
1190                 fin->fin_dport = ntohs(udp->uh_dport);
1191         }
1192 }
1193 
1194 
1195 /* ------------------------------------------------------------------------ */
1196 /* Function:    frpr_tcp                                                    */
1197 /* Returns:     void                                                        */
1198 /* Parameters:  fin(I) - pointer to packet information                      */
1199 /*                                                                          */
1200 /* IPv4 Only                                                                */
1201 /* Analyse the packet for IPv4/TCP properties.                              */
1202 /* ------------------------------------------------------------------------ */
1203 static INLINE void frpr_tcp(fin)
1204 fr_info_t *fin;
1205 {
1206 
1207         fr_checkv4sum(fin);
1208 
1209         frpr_short(fin, sizeof(tcphdr_t));
1210 
1211         frpr_tcpcommon(fin);
1212 }
1213 
1214 
1215 /* ------------------------------------------------------------------------ */
1216 /* Function:    frpr_udp                                                    */
1217 /* Returns:     void                                                        */
1218 /* Parameters:  fin(I) - pointer to packet information                      */
1219 /*                                                                          */
1220 /* IPv4 Only                                                                */
1221 /* Analyse the packet for IPv4/UDP properties.                              */
1222 /* ------------------------------------------------------------------------ */
1223 static INLINE void frpr_udp(fin)
1224 fr_info_t *fin;
1225 {
1226 
1227         fr_checkv4sum(fin);
1228 
1229         frpr_short(fin, sizeof(udphdr_t));
1230 
1231         frpr_udpcommon(fin);
1232 }
1233 
1234 
1235 /* ------------------------------------------------------------------------ */
1236 /* Function:    frpr_esp                                                    */
1237 /* Returns:     void                                                        */
1238 /* Parameters:  fin(I) - pointer to packet information                      */
1239 /*                                                                          */
1240 /* Analyse the packet for ESP properties.                                   */
1241 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
1242 /* even though the newer ESP packets must also have a sequence number that  */
1243 /* is 32bits as well, it is not possible(?) to determine the version from a */
1244 /* simple packet header.                                                    */
1245 /* ------------------------------------------------------------------------ */
1246 static INLINE void frpr_esp(fin)
1247 fr_info_t *fin;
1248 {
1249         if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1250                 return;
1251 
1252         frpr_short(fin, 8);
1253 }
1254 
1255 
1256 /* ------------------------------------------------------------------------ */
1257 /* Function:    frpr_ah                                                     */
1258 /* Returns:     void                                                        */
1259 /* Parameters:  fin(I) - pointer to packet information                      */
1260 /*                                                                          */
1261 /* Analyse the packet for AH properties.                                    */
1262 /* The minimum length is taken to be the combination of all fields in the   */
1263 /* header being present and no authentication data (null algorithm used.)   */
1264 /* ------------------------------------------------------------------------ */
1265 static INLINE void frpr_ah(fin)
1266 fr_info_t *fin;
1267 {
1268         authhdr_t *ah;
1269         int len;
1270 
1271         if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1272                 return;
1273 
1274         ah = (authhdr_t *)fin->fin_dp;
1275 
1276         len = (ah->ah_plen + 2) << 2;
1277         frpr_short(fin, len);
1278 }
1279 
1280 
1281 /* ------------------------------------------------------------------------ */
1282 /* Function:    frpr_gre                                                    */
1283 /* Returns:     void                                                        */
1284 /* Parameters:  fin(I) - pointer to packet information                      */
1285 /*                                                                          */
1286 /* Analyse the packet for GRE properties.                                   */
1287 /* ------------------------------------------------------------------------ */
1288 static INLINE void frpr_gre(fin)
1289 fr_info_t *fin;
1290 {
1291         grehdr_t *gre;
1292 
1293         if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1294                 return;
1295 
1296         frpr_short(fin, sizeof(grehdr_t));
1297 
1298         if (fin->fin_off == 0) {
1299                 gre = fin->fin_dp;
1300                 if (GRE_REV(gre->gr_flags) == 1)
1301                         fin->fin_data[0] = gre->gr_call;
1302         }
1303 }
1304 
1305 
1306 /* ------------------------------------------------------------------------ */
1307 /* Function:    frpr_ipv4hdr                                                */
1308 /* Returns:     void                                                        */
1309 /* Parameters:  fin(I) - pointer to packet information                      */
1310 /*                                                                          */
1311 /* IPv4 Only                                                                */
1312 /* Analyze the IPv4 header and set fields in the fr_info_t structure.       */
1313 /* Check all options present and flag their presence if any exist.          */
1314 /* ------------------------------------------------------------------------ */
1315 static INLINE void frpr_ipv4hdr(fin)
1316 fr_info_t *fin;
1317 {
1318         u_short optmsk = 0, secmsk = 0, auth = 0;
1319         int hlen, ol, mv, p, i;
1320         const struct optlist *op;
1321         u_char *s, opt;
1322         u_short off;
1323         fr_ip_t *fi;
1324         ip_t *ip;
1325 
1326         fi = &fin->fin_fi;
1327         hlen = fin->fin_hlen;
1328 
1329         ip = fin->fin_ip;
1330         p = ip->ip_p;
1331         fi->fi_p = p;
1332         fi->fi_tos = ip->ip_tos;
1333         fin->fin_id = ip->ip_id;
1334         off = ip->ip_off;
1335 
1336         /* Get both TTL and protocol */
1337         fi->fi_p = ip->ip_p;
1338         fi->fi_ttl = ip->ip_ttl;
1339 #if 0
1340         (*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1341 #endif
1342 
1343         /* Zero out bits not used in IPv6 address */
1344         fi->fi_src.i6[1] = 0;
1345         fi->fi_src.i6[2] = 0;
1346         fi->fi_src.i6[3] = 0;
1347         fi->fi_dst.i6[1] = 0;
1348         fi->fi_dst.i6[2] = 0;
1349         fi->fi_dst.i6[3] = 0;
1350 
1351         fi->fi_saddr = ip->ip_src.s_addr;
1352         fi->fi_daddr = ip->ip_dst.s_addr;
1353 
1354         /*
1355          * set packet attribute flags based on the offset and
1356          * calculate the byte offset that it represents.
1357          */
1358         off &= IP_MF|IP_OFFMASK;
1359         if (off != 0) {
1360                 int morefrag = off & IP_MF;
1361 
1362                 fi->fi_flx |= FI_FRAG;
1363                 if (morefrag)
1364                         fi->fi_flx |= FI_MOREFRAG;
1365                 off &= IP_OFFMASK;
1366                 if (off != 0) {
1367                         fin->fin_flx |= FI_FRAGBODY;
1368                         off <<= 3;
1369                         if ((off + fin->fin_dlen > 65535) || 
1370                             (fin->fin_dlen == 0) ||
1371                             ((morefrag != 0) && ((fin->fin_dlen & 7) != 0))) {
1372                                 /* 
1373                                  * The length of the packet, starting at its
1374                                  * offset cannot exceed 65535 (0xffff) as the 
1375                                  * length of an IP packet is only 16 bits.
1376                                  *
1377                                  * Any fragment that isn't the last fragment
1378                                  * must have a length greater than 0 and it
1379                                  * must be an even multiple of 8.
1380                                  */
1381                                 fi->fi_flx |= FI_BAD;
1382                         }
1383                 }
1384         }
1385         fin->fin_off = off;
1386 
1387         /*
1388          * Call per-protocol setup and checking
1389          */
1390         switch (p)
1391         {
1392         case IPPROTO_UDP :
1393                 frpr_udp(fin);
1394                 break;
1395         case IPPROTO_TCP :
1396                 frpr_tcp(fin);
1397                 break;
1398         case IPPROTO_ICMP :
1399                 frpr_icmp(fin);
1400                 break;
1401         case IPPROTO_AH :
1402                 frpr_ah(fin);
1403                 break;
1404         case IPPROTO_ESP :
1405                 frpr_esp(fin);
1406                 break;
1407         case IPPROTO_GRE :
1408                 frpr_gre(fin);
1409                 break;
1410         }
1411 
1412         ip = fin->fin_ip;
1413         if (ip == NULL)
1414                 return;
1415 
1416         /*
1417          * If it is a standard IP header (no options), set the flag fields
1418          * which relate to options to 0.
1419          */
1420         if (hlen == sizeof(*ip)) {
1421                 fi->fi_optmsk = 0;
1422                 fi->fi_secmsk = 0;
1423                 fi->fi_auth = 0;
1424                 return;
1425         }
1426 
1427         /*
1428          * So the IP header has some IP options attached.  Walk the entire
1429          * list of options present with this packet and set flags to indicate
1430          * which ones are here and which ones are not.  For the somewhat out
1431          * of date and obscure security classification options, set a flag to
1432          * represent which classification is present.
1433          */
1434         fi->fi_flx |= FI_OPTIONS;
1435 
1436         for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1437                 opt = *s;
1438                 if (opt == '\0')
1439                         break;
1440                 else if (opt == IPOPT_NOP)
1441                         ol = 1;
1442                 else {
1443                         if (hlen < 2)
1444                                 break;
1445                         ol = (int)*(s + 1);
1446                         if (ol < 2 || ol > hlen)
1447                                 break;
1448                 }
1449                 for (i = 9, mv = 4; mv >= 0; ) {
1450                         op = ipopts + i;
1451                         if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1452                                 optmsk |= op->ol_bit;
1453                                 if (opt == IPOPT_SECURITY) {
1454                                         const struct optlist *sp;
1455                                         u_char  sec;
1456                                         int j, m;
1457 
1458                                         sec = *(s + 2); /* classification */
1459                                         for (j = 3, m = 2; m >= 0; ) {
1460                                                 sp = secopt + j;
1461                                                 if (sec == sp->ol_val) {
1462                                                         secmsk |= sp->ol_bit;
1463                                                         auth = *(s + 3);
1464                                                         auth *= 256;
1465                                                         auth += *(s + 4);
1466                                                         break;
1467                                                 }
1468                                                 if (sec < sp->ol_val)
1469                                                         j -= m;
1470                                                 else
1471                                                         j += m;
1472                                                 m--;
1473                                         }
1474                                 }
1475                                 break;
1476                         }
1477                         if (opt < op->ol_val)
1478                                 i -= mv;
1479                         else
1480                                 i += mv;
1481                         mv--;
1482                 }
1483                 hlen -= ol;
1484                 s += ol;
1485         }
1486 
1487         /*
1488          *
1489          */
1490         if (auth && !(auth & 0x0100))
1491                 auth &= 0xff00;
1492         fi->fi_optmsk = optmsk;
1493         fi->fi_secmsk = secmsk;
1494         fi->fi_auth = auth;
1495 }
1496 
1497 
1498 /* ------------------------------------------------------------------------ */
1499 /* Function:    fr_makefrip                                                 */
1500 /* Returns:     int - 1 == hdr checking error, 0 == OK                      */
1501 /* Parameters:  hlen(I) - length of IP packet header                        */
1502 /*              ip(I)   - pointer to the IP header                          */
1503 /*              fin(IO) - pointer to packet information                     */
1504 /*                                                                          */
1505 /* Compact the IP header into a structure which contains just the info.     */
1506 /* which is useful for comparing IP headers with and store this information */
1507 /* in the fr_info_t structure pointer to by fin.  At present, it is assumed */
1508 /* this function will be called with either an IPv4 or IPv6 packet.         */
1509 /* ------------------------------------------------------------------------ */
1510 int     fr_makefrip(hlen, ip, fin)
1511 int hlen;
1512 ip_t *ip;
1513 fr_info_t *fin;
1514 {
1515         int v;
1516 
1517         fin->fin_depth = 0;
1518         fin->fin_hlen = (u_short)hlen;
1519         fin->fin_ip = ip;
1520         fin->fin_rule = 0xffffffff;
1521         fin->fin_group[0] = -1;
1522         fin->fin_group[1] = '\0';
1523         fin->fin_dlen = fin->fin_plen - hlen;
1524         fin->fin_dp = (char *)ip + hlen;
1525 
1526         v = fin->fin_v;
1527         if (v == 4)
1528                 frpr_ipv4hdr(fin);
1529 #ifdef  USE_INET6
1530         else if (v == 6)
1531                 frpr_ipv6hdr(fin);
1532 #endif
1533         if (fin->fin_ip == NULL)
1534                 return -1;
1535         return 0;
1536 }
1537 
1538 
1539 /* ------------------------------------------------------------------------ */
1540 /* Function:    fr_portcheck                                                */
1541 /* Returns:     int - 1 == port matched, 0 == port match failed             */
1542 /* Parameters:  frp(I) - pointer to port check `expression'                 */
1543 /*              pop(I) - pointer to port number to evaluate                 */
1544 /*                                                                          */
1545 /* Perform a comparison of a port number against some other(s), using a     */
1546 /* structure with compare information stored in it.                         */
1547 /* ------------------------------------------------------------------------ */
1548 static INLINE int fr_portcheck(frp, pop)
1549 frpcmp_t *frp;
1550 u_short *pop;
1551 {
1552         u_short tup, po;
1553         int err = 1;
1554 
1555         tup = *pop;
1556         po = frp->frp_port;
1557 
1558         /*
1559          * Do opposite test to that required and continue if that succeeds.
1560          */
1561         switch (frp->frp_cmp)
1562         {
1563         case FR_EQUAL :
1564                 if (tup != po) /* EQUAL */
1565                         err = 0;
1566                 break;
1567         case FR_NEQUAL :
1568                 if (tup == po) /* NOTEQUAL */
1569                         err = 0;
1570                 break;
1571         case FR_LESST :
1572                 if (tup >= po) /* LESSTHAN */
1573                         err = 0;
1574                 break;
1575         case FR_GREATERT :
1576                 if (tup <= po) /* GREATERTHAN */
1577                         err = 0;
1578                 break;
1579         case FR_LESSTE :
1580                 if (tup > po) /* LT or EQ */
1581                         err = 0;
1582                 break;
1583         case FR_GREATERTE :
1584                 if (tup < po) /* GT or EQ */
1585                         err = 0;
1586                 break;
1587         case FR_OUTRANGE :
1588                 if (tup >= po && tup <= frp->frp_top) /* Out of range */
1589                         err = 0;
1590                 break;
1591         case FR_INRANGE :
1592                 if (tup <= po || tup >= frp->frp_top) /* In range */
1593                         err = 0;
1594                 break;
1595         case FR_INCRANGE :
1596                 if (tup < po || tup > frp->frp_top) /* Inclusive range */
1597                         err = 0;
1598                 break;
1599         default :
1600                 break;
1601         }
1602         return err;
1603 }
1604 
1605 
1606 /* ------------------------------------------------------------------------ */
1607 /* Function:    fr_tcpudpchk                                                */
1608 /* Returns:     int - 1 == protocol matched, 0 == check failed              */
1609 /* Parameters:  fin(I) - pointer to packet information                      */
1610 /*              ft(I)  - pointer to structure with comparison data          */
1611 /*                                                                          */
1612 /* Compares the current pcket (assuming it is TCP/UDP) information with a   */
1613 /* structure containing information that we want to match against.          */
1614 /* ------------------------------------------------------------------------ */
1615 int fr_tcpudpchk(fin, ft)
1616 fr_info_t *fin;
1617 frtuc_t *ft;
1618 {
1619         int err = 1;
1620 
1621         /*
1622          * Both ports should *always* be in the first fragment.
1623          * So far, I cannot find any cases where they can not be.
1624          *
1625          * compare destination ports
1626          */
1627         if (ft->ftu_dcmp)
1628                 err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1629 
1630         /*
1631          * compare source ports
1632          */
1633         if (err && ft->ftu_scmp)
1634                 err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1635 
1636         /*
1637          * If we don't have all the TCP/UDP header, then how can we
1638          * expect to do any sort of match on it ?  If we were looking for
1639          * TCP flags, then NO match.  If not, then match (which should
1640          * satisfy the "short" class too).
1641          */
1642         if (err && (fin->fin_p == IPPROTO_TCP)) {
1643                 if (fin->fin_flx & FI_SHORT)
1644                         return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1645                 /*
1646                  * Match the flags ?  If not, abort this match.
1647                  */
1648                 if (ft->ftu_tcpfm &&
1649                     ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1650                         FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1651                                  ft->ftu_tcpfm, ft->ftu_tcpf));
1652                         err = 0;
1653                 }
1654         }
1655         return err;
1656 }
1657 
1658 
1659 /* ------------------------------------------------------------------------ */
1660 /* Function:    fr_ipfcheck                                                 */
1661 /* Returns:     int - 0 == match, 1 == no match                             */
1662 /* Parameters:  fin(I)     - pointer to packet information                  */
1663 /*              fr(I)      - pointer to filter rule                         */
1664 /*              portcmp(I) - flag indicating whether to attempt matching on */
1665 /*                           TCP/UDP port data.                             */
1666 /*                                                                          */
1667 /* Check to see if a packet matches an IPFilter rule.  Checks of addresses, */
1668 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1669 /* this function.                                                           */
1670 /* ------------------------------------------------------------------------ */
1671 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1672 fr_info_t *fin;
1673 frentry_t *fr;
1674 int portcmp;
1675 {
1676         u_32_t  *ld, *lm, *lip;
1677         fripf_t *fri;
1678         fr_ip_t *fi;
1679         int i;
1680         ipf_stack_t *ifs = fin->fin_ifs;
1681 
1682         fi = &fin->fin_fi;
1683         fri = fr->fr_ipf;
1684         lip = (u_32_t *)fi;
1685         lm = (u_32_t *)&fri->fri_mip;
1686         ld = (u_32_t *)&fri->fri_ip;
1687 
1688         /*
1689          * first 32 bits to check coversion:
1690          * IP version, TOS, TTL, protocol
1691          */
1692         i = ((*lip & *lm) != *ld);
1693         FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1694                    *lip, *lm, *ld));
1695         if (i)
1696                 return 1;
1697 
1698         /*
1699          * Next 32 bits is a constructed bitmask indicating which IP options
1700          * are present (if any) in this packet.
1701          */
1702         lip++, lm++, ld++;
1703         i |= ((*lip & *lm) != *ld);
1704         FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1705                    *lip, *lm, *ld));
1706         if (i)
1707                 return 1;
1708 
1709         lip++, lm++, ld++;
1710         /*
1711          * Unrolled loops (4 each, for 32 bits) for address checks.
1712          */
1713         /*
1714          * Check the source address.
1715          */
1716 #ifdef  IPFILTER_LOOKUP
1717         if (fr->fr_satype == FRI_LOOKUP) {
1718                 fin->fin_flx |= FI_DONTCACHE;
1719                 i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, fin, ifs);
1720                 if (i == -1)
1721                         return 1;
1722                 lip += 3;
1723                 lm += 3;
1724                 ld += 3;
1725         } else {
1726 #endif
1727                 i = ((*lip & *lm) != *ld);
1728                 FR_DEBUG(("2a. %#08x & %#08x != %#08x\n",
1729                            *lip, *lm, *ld));
1730                 if (fi->fi_v == 6) {
1731                         lip++, lm++, ld++;
1732                         i |= ((*lip & *lm) != *ld);
1733                         FR_DEBUG(("2b. %#08x & %#08x != %#08x\n",
1734                                    *lip, *lm, *ld));
1735                         lip++, lm++, ld++;
1736                         i |= ((*lip & *lm) != *ld);
1737                         FR_DEBUG(("2c. %#08x & %#08x != %#08x\n",
1738                                    *lip, *lm, *ld));
1739                         lip++, lm++, ld++;
1740                         i |= ((*lip & *lm) != *ld);
1741                         FR_DEBUG(("2d. %#08x & %#08x != %#08x\n",
1742                                    *lip, *lm, *ld));
1743                 } else {
1744                         lip += 3;
1745                         lm += 3;
1746                         ld += 3;
1747                 }
1748 #ifdef  IPFILTER_LOOKUP
1749         }
1750 #endif
1751         i ^= (fr->fr_flags & FR_NOTSRCIP) >> 6;
1752         if (i)
1753                 return 1;
1754 
1755         /*
1756          * Check the destination address.
1757          */
1758         lip++, lm++, ld++;
1759 #ifdef  IPFILTER_LOOKUP
1760         if (fr->fr_datype == FRI_LOOKUP) {
1761                 fin->fin_flx |= FI_DONTCACHE;
1762                 i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, fin, ifs);
1763                 if (i == -1)
1764                         return 1;
1765                 lip += 3;
1766                 lm += 3;
1767                 ld += 3;
1768         } else {
1769 #endif
1770                 i = ((*lip & *lm) != *ld);
1771                 FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1772                            *lip, *lm, *ld));
1773                 if (fi->fi_v == 6) {
1774                         lip++, lm++, ld++;
1775                         i |= ((*lip & *lm) != *ld);
1776                         FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1777                                    *lip, *lm, *ld));
1778                         lip++, lm++, ld++;
1779                         i |= ((*lip & *lm) != *ld);
1780                         FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1781                                    *lip, *lm, *ld));
1782                         lip++, lm++, ld++;
1783                         i |= ((*lip & *lm) != *ld);
1784                         FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1785                                    *lip, *lm, *ld));
1786                 } else {
1787                         lip += 3;
1788                         lm += 3;
1789                         ld += 3;
1790                 }
1791 #ifdef  IPFILTER_LOOKUP
1792         }
1793 #endif
1794         i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1795         if (i)
1796                 return 1;
1797         /*
1798          * IP addresses matched.  The next 32bits contains:
1799          * mast of old IP header security & authentication bits.
1800          */
1801         lip++, lm++, ld++;
1802         i |= ((*lip & *lm) != *ld);
1803         FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1804                    *lip, *lm, *ld));
1805 
1806         /*
1807          * Next we have 32 bits of packet flags.
1808          */
1809         lip++, lm++, ld++;
1810         i |= ((*lip & *lm) != *ld);
1811         FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1812                    *lip, *lm, *ld));
1813 
1814         if (i == 0) {
1815                 /*
1816                  * If a fragment, then only the first has what we're
1817                  * looking for here...
1818                  */
1819                 if (portcmp) {
1820                         if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1821                                 i = 1;
1822                 } else {
1823                         if (fr->fr_dcmp || fr->fr_scmp ||
1824                             fr->fr_tcpf || fr->fr_tcpfm)
1825                                 i = 1;
1826                         if (fr->fr_icmpm || fr->fr_icmp) {
1827                                 if (((fi->fi_p != IPPROTO_ICMP) &&
1828                                      (fi->fi_p != IPPROTO_ICMPV6)) ||
1829                                     fin->fin_off || (fin->fin_dlen < 2))
1830                                         i = 1;
1831                                 else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1832                                          fr->fr_icmp) {
1833                                         FR_DEBUG(("i. %#x & %#x != %#x\n",
1834                                                  fin->fin_data[0],
1835                                                  fr->fr_icmpm, fr->fr_icmp));
1836                                         i = 1;
1837                                 }
1838                         }
1839                 }
1840         }
1841         return i;
1842 }
1843 
1844 
1845 /* ------------------------------------------------------------------------ */
1846 /* Function:    fr_scanlist                                                 */
1847 /* Returns:     int - result flags of scanning filter list                  */
1848 /* Parameters:  fin(I) - pointer to packet information                      */
1849 /*              pass(I) - default result to return for filtering            */
1850 /*                                                                          */
1851 /* Check the input/output list of rules for a match to the current packet.  */
1852 /* If a match is found, the value of fr_flags from the rule becomes the     */
1853 /* return value and fin->fin_fr points to the matched rule.                 */
1854 /*                                                                          */
1855 /* This function may be called recusively upto 16 times (limit inbuilt.)    */
1856 /* When unwinding, it should finish up with fin_depth as 0.                 */
1857 /*                                                                          */
1858 /* Could be per interface, but this gets real nasty when you don't have,    */
1859 /* or can't easily change, the kernel source code to .                      */
1860 /* ------------------------------------------------------------------------ */
1861 int fr_scanlist(fin, pass)
1862 fr_info_t *fin;
1863 u_32_t pass;
1864 {
1865         int rulen, portcmp, off, logged, skip;
1866         struct frentry *fr, *fnext;
1867         u_32_t passt, passo;
1868         ipf_stack_t *ifs = fin->fin_ifs;
1869 
1870         /*
1871          * Do not allow nesting deeper than 16 levels.
1872          */
1873         if (fin->fin_depth >= 16)
1874                 return pass;
1875 
1876         fr = fin->fin_fr;
1877 
1878         /*
1879          * If there are no rules in this list, return now.
1880          */
1881         if (fr == NULL)
1882                 return pass;
1883 
1884         skip = 0;
1885         logged = 0;
1886         portcmp = 0;
1887         fin->fin_depth++;
1888         fin->fin_fr = NULL;
1889         off = fin->fin_off;
1890 
1891         if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1892                 portcmp = 1;
1893 
1894         for (rulen = 0; fr; fr = fnext, rulen++) {
1895                 fnext = fr->fr_next;
1896                 if (skip != 0) {
1897                         FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1898                         skip--;
1899                         continue;
1900                 }
1901 
1902                 /*
1903                  * In all checks below, a null (zero) value in the
1904                  * filter struture is taken to mean a wildcard.
1905                  *
1906                  * check that we are working for the right interface
1907                  */
1908 #ifdef  _KERNEL
1909                 if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1910                         continue;
1911 #else
1912                 if (opts & (OPT_VERBOSE|OPT_DEBUG))
1913                         printf("\n");
1914                 FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1915                                   FR_ISPASS(pass) ? 'p' :
1916                                   FR_ISACCOUNT(pass) ? 'A' :
1917                                   FR_ISAUTH(pass) ? 'a' :
1918                                   (pass & FR_NOMATCH) ? 'n' :'b'));
1919                 if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1920                         continue;
1921                 FR_VERBOSE((":i"));
1922 #endif
1923 
1924                 switch (fr->fr_type)
1925                 {
1926                 case FR_T_IPF :
1927                 case FR_T_IPF|FR_T_BUILTIN :
1928                         if (fr_ipfcheck(fin, fr, portcmp))
1929                                 continue;
1930                         break;
1931 #if defined(IPFILTER_BPF)
1932                 case FR_T_BPFOPC :
1933                 case FR_T_BPFOPC|FR_T_BUILTIN :
1934                     {
1935                         u_char *mc;
1936 
1937                         if (*fin->fin_mp == NULL)
1938                                 continue;
1939                         if (fin->fin_v != fr->fr_v)
1940                                 continue;
1941                         mc = (u_char *)fin->fin_m;
1942                         if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1943                                 continue;
1944                         break;
1945                     }
1946 #endif
1947                 case FR_T_CALLFUNC|FR_T_BUILTIN :
1948                     {
1949                         frentry_t *f;
1950 
1951                         f = (*fr->fr_func)(fin, &pass);
1952                         if (f != NULL)
1953                                 fr = f;
1954                         else
1955                                 continue;
1956                         break;
1957                     }
1958                 default :
1959                         break;
1960                 }
1961 
1962                 if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1963                         if (fin->fin_nattag == NULL)
1964                                 continue;
1965                         if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1966                                 continue;
1967                 }
1968                 FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1969 
1970                 passt = fr->fr_flags;
1971 
1972                 /*
1973                  * Allowing a rule with the "keep state" flag set to match
1974                  * packets that have been tagged "out of window" by the TCP
1975                  * state tracking is foolish as the attempt to add a new
1976                  * state entry to the table will fail.
1977                  */
1978                 if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1979                         continue;
1980 
1981                 /*
1982                  * If the rule is a "call now" rule, then call the function
1983                  * in the rule, if it exists and use the results from that.
1984                  * If the function pointer is bad, just make like we ignore
1985                  * it, except for increasing the hit counter.
1986                  */
1987                 IPF_BUMP(fr->fr_hits);
1988                 fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
1989                 if ((passt & FR_CALLNOW) != 0) {
1990                         if ((fr->fr_func != NULL) &&
1991                             (fr->fr_func != (ipfunc_t)-1)) {
1992                                 frentry_t *frs;
1993 
1994                                 frs = fin->fin_fr;
1995                                 fin->fin_fr = fr;
1996                                 fr = (*fr->fr_func)(fin, &passt);
1997                                 if (fr == NULL) {
1998                                         fin->fin_fr = frs;
1999                                         continue;
2000                                 }
2001                                 passt = fr->fr_flags;
2002                                 fin->fin_fr = fr;
2003                         }
2004                 } else {
2005                         fin->fin_fr = fr;
2006                 }
2007 
2008 #ifdef  IPFILTER_LOG
2009                 /*
2010                  * Just log this packet...
2011                  */
2012                 if ((passt & FR_LOGMASK) == FR_LOG) {
2013                         if (ipflog(fin, passt) == -1) {
2014                                 if (passt & FR_LOGORBLOCK) {
2015                                         passt &= ~FR_CMDMASK;
2016                                         passt |= FR_BLOCK|FR_QUICK;
2017                                 }
2018                                 IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2019                         }
2020                         IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2021                         logged = 1;
2022                 }
2023 #endif /* IPFILTER_LOG */
2024                 passo = pass;
2025                 if (FR_ISSKIP(passt))
2026                         skip = fr->fr_arg;
2027                 else if ((passt & FR_LOGMASK) != FR_LOG)
2028                         pass = passt;
2029                 if (passt & (FR_RETICMP|FR_FAKEICMP))
2030                         fin->fin_icode = fr->fr_icode;
2031                 FR_DEBUG(("pass %#x\n", pass));
2032                 fin->fin_rule = rulen;
2033                 (void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2034                 if (fr->fr_grp != NULL) {
2035                         fin->fin_fr = *fr->fr_grp;
2036                         pass = fr_scanlist(fin, pass);
2037                         if (fin->fin_fr == NULL) {
2038                                 fin->fin_rule = rulen;
2039                                 (void) strncpy(fin->fin_group, fr->fr_group,
2040                                                FR_GROUPLEN);
2041                                 fin->fin_fr = fr;
2042                         }
2043                         if (fin->fin_flx & FI_DONTCACHE)
2044                                 logged = 1;
2045                 }
2046 
2047                 if (pass & FR_QUICK) {
2048                         /*
2049                          * Finally, if we've asked to track state for this
2050                          * packet, set it up.  Add state for "quick" rules
2051                          * here so that if the action fails we can consider
2052                          * the rule to "not match" and keep on processing
2053                          * filter rules.
2054                          */
2055                         if ((pass & FR_KEEPSTATE) &&
2056                             !(fin->fin_flx & FI_STATE)) {
2057                                 int out = fin->fin_out;
2058 
2059                                 if (fr_addstate(fin, NULL, 0) != NULL) {
2060                                         IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2061                                 } else {
2062                                         IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2063                                         pass = passo;
2064                                         continue;
2065                                 }
2066                         }
2067                         break;
2068                 }
2069         }
2070         if (logged)
2071                 fin->fin_flx |= FI_DONTCACHE;
2072         fin->fin_depth--;
2073         return pass;
2074 }
2075 
2076 
2077 /* ------------------------------------------------------------------------ */
2078 /* Function:    fr_acctpkt                                                  */
2079 /* Returns:     frentry_t* - always returns NULL                            */
2080 /* Parameters:  fin(I) - pointer to packet information                      */
2081 /*              passp(IO) - pointer to current/new filter decision (unused) */
2082 /*                                                                          */
2083 /* Checks a packet against accounting rules, if there are any for the given */
2084 /* IP protocol version.                                                     */
2085 /*                                                                          */
2086 /* N.B.: this function returns NULL to match the prototype used by other    */
2087 /* functions called from the IPFilter "mainline" in fr_check().             */
2088 /* ------------------------------------------------------------------------ */
2089 frentry_t *fr_acctpkt(fin, passp)
2090 fr_info_t *fin;
2091 u_32_t *passp;
2092 {
2093         char group[FR_GROUPLEN];
2094         frentry_t *fr, *frsave;
2095         u_32_t pass, rulen;
2096         ipf_stack_t *ifs = fin->fin_ifs;
2097 
2098         passp = passp;
2099 #ifdef  USE_INET6
2100         if (fin->fin_v == 6)
2101                 fr = ifs->ifs_ipacct6[fin->fin_out][ifs->ifs_fr_active];
2102         else
2103 #endif
2104                 fr = ifs->ifs_ipacct[fin->fin_out][ifs->ifs_fr_active];
2105 
2106         if (fr != NULL) {
2107                 frsave = fin->fin_fr;
2108                 bcopy(fin->fin_group, group, FR_GROUPLEN);
2109                 rulen = fin->fin_rule;
2110                 fin->fin_fr = fr;
2111                 pass = fr_scanlist(fin, FR_NOMATCH);
2112                 if (FR_ISACCOUNT(pass)) {
2113                         IPF_BUMP(ifs->ifs_frstats[0].fr_acct);
2114                 }
2115                 fin->fin_fr = frsave;
2116                 bcopy(group, fin->fin_group, FR_GROUPLEN);
2117                 fin->fin_rule = rulen;
2118         }
2119         return NULL;
2120 }
2121 
2122 
2123 /* ------------------------------------------------------------------------ */
2124 /* Function:    fr_firewall                                                 */
2125 /* Returns:     frentry_t* - returns pointer to matched rule, if no matches */
2126 /*                           were found, returns NULL.                      */
2127 /* Parameters:  fin(I) - pointer to packet information                      */
2128 /*              passp(IO) - pointer to current/new filter decision (unused) */
2129 /*                                                                          */
2130 /* Applies an appropriate set of firewall rules to the packet, to see if    */
2131 /* there are any matches.  The first check is to see if a match can be seen */
2132 /* in the cache.  If not, then search an appropriate list of rules.  Once a */
2133 /* matching rule is found, take any appropriate actions as defined by the   */
2134 /* rule - except logging.                                                   */
2135 /* ------------------------------------------------------------------------ */
2136 static frentry_t *fr_firewall(fin, passp)
2137 fr_info_t *fin;
2138 u_32_t *passp;
2139 {
2140         frentry_t *fr;
2141         fr_info_t *fc;
2142         u_32_t pass;
2143         int out;
2144         ipf_stack_t *ifs = fin->fin_ifs;
2145 
2146         out = fin->fin_out;
2147         pass = *passp;
2148 
2149 #ifdef  USE_INET6
2150         if (fin->fin_v == 6)
2151                 fin->fin_fr = ifs->ifs_ipfilter6[out][ifs->ifs_fr_active];
2152         else
2153 #endif
2154                 fin->fin_fr = ifs->ifs_ipfilter[out][ifs->ifs_fr_active];
2155 
2156         /*
2157          * If there are no rules loaded skip all checks and return.
2158          */
2159         if (fin->fin_fr == NULL) {
2160 
2161                 if ((pass & FR_NOMATCH)) {
2162                         IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2163                 }
2164 
2165                 return (NULL);
2166         }
2167 
2168         fc = &ifs->ifs_frcache[out][CACHE_HASH(fin)];
2169         READ_ENTER(&ifs->ifs_ipf_frcache);
2170         if (!bcmp((char *)fin, (char *)fc, FI_CSIZE)) {
2171                 /*
2172                  * copy cached data so we can unlock the mutexes earlier.
2173                  */
2174                 bcopy((char *)fc, (char *)fin, FI_COPYSIZE);
2175                 RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2176                 IPF_BUMP(ifs->ifs_frstats[out].fr_chit);
2177 
2178                 if ((fr = fin->fin_fr) != NULL) {
2179                         IPF_BUMP(fr->fr_hits);
2180                         fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2181                         pass = fr->fr_flags;
2182                 }
2183         } else {
2184                 RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2185 
2186                 pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2187 
2188                 if (((pass & FR_KEEPSTATE) == 0) &&
2189                     ((fin->fin_flx & FI_DONTCACHE) == 0)) {
2190                         WRITE_ENTER(&ifs->ifs_ipf_frcache);
2191                         bcopy((char *)fin, (char *)fc, FI_COPYSIZE);
2192                         RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2193                 }
2194 
2195                 fr = fin->fin_fr;
2196         }
2197 
2198         if ((pass & FR_NOMATCH)) {
2199                 IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2200         }
2201 
2202         /*
2203          * Apply packets per second rate-limiting to a rule as required.
2204          */
2205         if ((fr != NULL) && (fr->fr_pps != 0) &&
2206             !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2207                 pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2208                 pass |= FR_BLOCK;
2209                 IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2210         }
2211 
2212         /*
2213          * If we fail to add a packet to the authorization queue, then we
2214          * drop the packet later.  However, if it was added then pretend
2215          * we've dropped it already.
2216          */
2217         if (FR_ISAUTH(pass)) {
2218                 if (fr_newauth(fin->fin_m, fin) != 0) {
2219 #ifdef  _KERNEL
2220                         fin->fin_m = *fin->fin_mp = NULL;
2221 #else
2222                         ;
2223 #endif
2224                         fin->fin_error = 0;
2225                 } else
2226                         fin->fin_error = ENOSPC;
2227         }
2228 
2229         if ((fr != NULL) && (fr->fr_func != NULL) &&
2230             (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2231                 (void) (*fr->fr_func)(fin, &pass);
2232 
2233         /*
2234          * If a rule is a pre-auth rule, check again in the list of rules
2235          * loaded for authenticated use.  It does not particulary matter
2236          * if this search fails because a "preauth" result, from a rule,
2237          * is treated as "not a pass", hence the packet is blocked.
2238          */
2239         if (FR_ISPREAUTH(pass)) {
2240                 if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2241                         pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2242         }
2243 
2244         /*
2245          * If the rule has "keep frag" and the packet is actually a fragment,
2246          * then create a fragment state entry.
2247          */
2248         if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2249                 if (fin->fin_flx & FI_FRAG) {
2250                         if (fr_newfrag(fin, pass) == -1) {
2251                                 IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2252                         } else {
2253                                 IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2254                         }
2255                 } else {
2256                         IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2257                 }
2258         }
2259 
2260         /*
2261          * Finally, if we've asked to track state for this packet, set it up.
2262          */
2263         if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2264                 if (fr_addstate(fin, NULL, 0) != NULL) {
2265                         IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2266                 } else {
2267                         IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2268                         if (FR_ISPASS(pass)) {
2269                                 pass &= ~FR_CMDMASK;
2270                                 pass |= FR_BLOCK;
2271                         }
2272                 }
2273         }
2274 
2275         fr = fin->fin_fr;
2276 
2277         if (passp != NULL)
2278                 *passp = pass;
2279 
2280         return fr;
2281 }
2282 
2283 /* ------------------------------------------------------------------------ */
2284 /* Function:    fr_check                                                    */
2285 /* Returns:     int -  0 == packet allowed through,                         */
2286 /*              User space:                                                 */
2287 /*                    -1 == packet blocked                                  */
2288 /*                     1 == packet not matched                              */
2289 /*                    -2 == requires authentication                         */
2290 /*              Kernel:                                                     */
2291 /*                   > 0 == filter error # for packet                       */
2292 /* Parameters: ip(I)   - pointer to start of IPv4/6 packet                  */
2293 /*             hlen(I) - length of header                                   */
2294 /*             ifp(I)  - pointer to interface this packet is on             */
2295 /*             out(I)  - 0 == packet going in, 1 == packet going out        */
2296 /*             mp(IO)  - pointer to caller's buffer pointer that holds this */
2297 /*                       IP packet.                                         */
2298 /* Solaris & HP-UX ONLY :                                                   */
2299 /*             qpi(I)  - pointer to STREAMS queue information for this      */
2300 /*                       interface & direction.                             */
2301 /*                                                                          */
2302 /* fr_check() is the master function for all IPFilter packet processing.    */
2303 /* It orchestrates: Network Address Translation (NAT), checking for packet  */
2304 /* authorisation (or pre-authorisation), presence of related state info.,   */
2305 /* generating log entries, IP packet accounting, routing of packets as      */
2306 /* directed by firewall rules and of course whether or not to allow the     */
2307 /* packet to be further processed by the kernel.                            */
2308 /*                                                                          */
2309 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer  */
2310 /* freed.  Packets passed may be returned with the pointer pointed to by    */
2311 /* by "mp" changed to a new buffer.                                         */
2312 /* ------------------------------------------------------------------------ */
2313 int fr_check(ip, hlen, ifp, out
2314 #if defined(_KERNEL) && defined(MENTAT)
2315 , qif, mp, ifs)
2316 void *qif;
2317 #else
2318 , mp, ifs)
2319 #endif
2320 mb_t **mp;
2321 ip_t *ip;
2322 int hlen;
2323 void *ifp;
2324 int out;
2325 ipf_stack_t *ifs;
2326 {
2327         /*
2328          * The above really sucks, but short of writing a diff
2329          */
2330         fr_info_t frinfo;
2331         fr_info_t *fin = &frinfo;
2332         u_32_t pass;
2333         frentry_t *fr = NULL;
2334         int v = IP_V(ip);
2335         mb_t *mc = NULL;
2336         mb_t *m;
2337 #ifdef USE_INET6
2338         ip6_t *ip6;
2339 #endif
2340 #ifdef  _KERNEL
2341 # ifdef MENTAT
2342         qpktinfo_t *qpi = qif;
2343 #endif
2344 #endif
2345 
2346         SPL_INT(s);
2347         pass = ifs->ifs_fr_pass;
2348 
2349         /*
2350          * The first part of fr_check() deals with making sure that what goes
2351          * into the filtering engine makes some sense.  Information about the
2352          * the packet is distilled, collected into a fr_info_t structure and
2353          * the an attempt to ensure the buffer the packet is in is big enough
2354          * to hold all the required packet headers.
2355          */
2356 #ifdef  _KERNEL
2357 # ifdef MENTAT
2358         if (!OK_32PTR(ip))
2359                 return 2;
2360 # endif
2361 
2362 
2363         if (ifs->ifs_fr_running <= 0) {
2364                 return 0;
2365         }
2366 
2367         bzero((char *)fin, sizeof(*fin));
2368 
2369 # ifdef MENTAT
2370         fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2371                                          FI_BROADCAST);
2372         m = qpi->qpi_m;
2373         fin->fin_qfm = m;
2374         fin->fin_qpi = qpi;
2375 # else /* MENTAT */
2376 
2377         m = *mp;
2378 
2379 #  if defined(M_MCAST)
2380         if ((m->m_flags & M_MCAST) != 0)
2381                 fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2382 #  endif
2383 #  if defined(M_MLOOP)
2384         if ((m->m_flags & M_MLOOP) != 0)
2385                 fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2386 #  endif
2387 #  if defined(M_BCAST)
2388         if ((m->m_flags & M_BCAST) != 0)
2389                 fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2390 #  endif
2391 #  ifdef M_CANFASTFWD
2392         /*
2393          * XXX For now, IP Filter and fast-forwarding of cached flows
2394          * XXX are mutually exclusive.  Eventually, IP Filter should
2395          * XXX get a "can-fast-forward" filter rule.
2396          */
2397         m->m_flags &= ~M_CANFASTFWD;
2398 #  endif /* M_CANFASTFWD */
2399 #  ifdef CSUM_DELAY_DATA
2400         /*
2401          * disable delayed checksums.
2402          */
2403         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2404                 in_delayed_cksum(m);
2405                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2406         }
2407 #  endif /* CSUM_DELAY_DATA */
2408 # endif /* MENTAT */
2409 #else
2410 
2411         bzero((char *)fin, sizeof(*fin));
2412         m = *mp;
2413 #endif /* _KERNEL */
2414 
2415         fin->fin_v = v;
2416         fin->fin_m = m;
2417         fin->fin_ip = ip;
2418         fin->fin_mp = mp;
2419         fin->fin_out = out;
2420         fin->fin_ifp = ifp;
2421         fin->fin_error = ENETUNREACH;
2422         fin->fin_hlen = (u_short)hlen;
2423         fin->fin_dp = (char *)ip + hlen;
2424         fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2425         fin->fin_ifs = ifs;
2426 
2427         SPL_NET(s);
2428 
2429 #ifdef  USE_INET6
2430         if (v == 6) {
2431                 IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2432                 /*
2433                  * Jumbo grams are quite likely too big for internal buffer
2434                  * structures to handle comfortably, for now, so just drop
2435                  * them.
2436                  */
2437                 ip6 = (ip6_t *)ip;
2438                 fin->fin_plen = ntohs(ip6->ip6_plen);
2439                 if (fin->fin_plen == 0) {
2440                         READ_ENTER(&ifs->ifs_ipf_mutex);
2441                         pass = FR_BLOCK|FR_NOMATCH;
2442                         goto filtered;
2443                 }
2444                 fin->fin_plen += sizeof(ip6_t);
2445         } else
2446 #endif
2447         {
2448 #if (OpenBSD >= 200311) && defined(_KERNEL)
2449                 ip->ip_len = ntohs(ip->ip_len);
2450                 ip->ip_off = ntohs(ip->ip_off);
2451 #endif
2452                 fin->fin_plen = ip->ip_len;
2453         }
2454 
2455         if (fr_makefrip(hlen, ip, fin) == -1) {
2456                 READ_ENTER(&ifs->ifs_ipf_mutex);
2457                 pass = FR_BLOCK;
2458                 goto filtered;
2459         }
2460 
2461         /*
2462          * For at least IPv6 packets, if a m_pullup() fails then this pointer
2463          * becomes NULL and so we have no packet to free.
2464          */
2465         if (*fin->fin_mp == NULL)
2466                 goto finished;
2467 
2468         if (!out) {
2469                 if (v == 4) {
2470 #ifdef _KERNEL
2471                         if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2472                                 IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2473                                 fin->fin_flx |= FI_BADSRC;
2474                         }
2475 #endif
2476                         if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2477                                 IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2478                                 fin->fin_flx |= FI_LOWTTL;
2479                         }
2480                 }
2481 #ifdef USE_INET6
2482                 else  if (v == 6) {
2483                         ip6 = (ip6_t *)ip;
2484 #ifdef _KERNEL
2485                         if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2486                                 IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2487                                 fin->fin_flx |= FI_BADSRC;
2488                         }
2489 #endif
2490                         if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2491                                 IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2492                                 fin->fin_flx |= FI_LOWTTL;
2493                         }
2494                 }
2495 #endif
2496         }
2497 
2498         if (fin->fin_flx & FI_SHORT) {
2499                 IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2500         }
2501 
2502         READ_ENTER(&ifs->ifs_ipf_mutex);
2503 
2504         /*
2505          * Check auth now.  This, combined with the check below to see if apass
2506          * is 0 is to ensure that we don't count the packet twice, which can
2507          * otherwise occur when we reprocess it.  As it is, we only count it
2508          * after it has no auth. table matchup.  This also stops NAT from
2509          * occuring until after the packet has been auth'd.
2510          */
2511         fr = fr_checkauth(fin, &pass);
2512         if (!out) {
2513                 switch (fin->fin_v)
2514                 {
2515                 case 4 :
2516                         if (fr_checknatin(fin, &pass) == -1) {
2517                                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2518                                 goto finished;
2519                         }
2520                         break;
2521 #ifdef  USE_INET6
2522                 case 6 :
2523                         if (fr_checknat6in(fin, &pass) == -1) {
2524                                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2525                                 goto finished;
2526                         }
2527                         break;
2528 #endif
2529                 default :
2530                         break;
2531                 }
2532         }
2533         if (!out)
2534                 (void) fr_acctpkt(fin, NULL);
2535 
2536         if (fr == NULL)
2537                 if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2538                         fr = fr_knownfrag(fin, &pass);
2539         if (fr == NULL)
2540                 fr = fr_checkstate(fin, &pass);
2541 
2542         if ((pass & FR_NOMATCH) || (fr == NULL))
2543                 fr = fr_firewall(fin, &pass);
2544 
2545         fin->fin_fr = fr;
2546 
2547         /*
2548          * Only count/translate packets which will be passed on, out the
2549          * interface.
2550          */
2551         if (out && FR_ISPASS(pass)) {
2552                 (void) fr_acctpkt(fin, NULL);
2553 
2554                 switch (fin->fin_v)
2555                 {
2556                 case 4 :
2557                         if (fr_checknatout(fin, &pass) == -1) {
2558                                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2559                                 goto finished;
2560                         }
2561                         break;
2562 #ifdef  USE_INET6
2563                 case 6 :
2564                         if (fr_checknat6out(fin, &pass) == -1) {
2565                                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2566                                 goto finished;
2567                         }
2568                         break;
2569 #endif
2570                 default :
2571                         break;
2572                 }
2573 
2574                 if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2575                         if (fr_updateipid(fin) == -1) {
2576                                 IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2577                                 pass &= ~FR_CMDMASK;
2578                                 pass |= FR_BLOCK;
2579                         } else {
2580                                 IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2581                         }
2582                 }
2583         }
2584 
2585 #ifdef  IPFILTER_LOG
2586         if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2587                 (void) fr_dolog(fin, &pass);
2588         }
2589 #endif
2590 
2591         if (IFS_CFWLOG(ifs) && FR_ISBLOCK(pass))
2592                 ipf_block_cfwlog(fr, fin, ifs);
2593 
2594         /*
2595          * The FI_STATE flag is cleared here so that calling fr_checkstate
2596          * will work when called from inside of fr_fastroute.  Although
2597          * there is a similar flag, FI_NATED, for NAT, it does have the same
2598          * impact on code execution.
2599          */
2600         fin->fin_flx &= ~FI_STATE;
2601 
2602         /*
2603          * Only allow FR_DUP to work if a rule matched - it makes no sense to
2604          * set FR_DUP as a "default" as there are no instructions about where
2605          * to send the packet.  Use fin_m here because it may have changed
2606          * (without an update of 'm') in prior processing.
2607          */
2608         if ((fr != NULL) && (pass & FR_DUP)) {
2609                 mc = M_DUPLICATE(fin->fin_m);
2610 #ifdef _KERNEL
2611                 mc->b_rptr += fin->fin_ipoff;
2612 #endif
2613         }
2614 
2615         /*
2616          * We don't want to send RST for packets, which are going to be
2617          * dropped, just because they don't fit into TCP window. Those packets
2618          * will be dropped silently. In other words, we want to drop packet,
2619          * while keeping session alive.
2620          */
2621         if ((pass & (FR_RETRST|FR_RETICMP)) && ((fin->fin_flx & FI_OOW) == 0)) {
2622                 /*
2623                  * Should we return an ICMP packet to indicate error
2624                  * status passing through the packet filter ?
2625                  * WARNING: ICMP error packets AND TCP RST packets should
2626                  * ONLY be sent in repsonse to incoming packets.  Sending them
2627                  * in response to outbound packets can result in a panic on
2628                  * some operating systems.
2629                  */
2630                 if (!out) {
2631                         if (pass & FR_RETICMP) {
2632                                 int dst;
2633 
2634                                 if ((pass & FR_RETMASK) == FR_FAKEICMP)
2635                                         dst = 1;
2636                                 else
2637                                         dst = 0;
2638 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2639                                 /*
2640                                  * Assume it's possible to enter insane rule:
2641                                  *      pass return-icmp in proto udp ...
2642                                  * then we have no other option than to forward
2643                                  * packet on loopback and give up any attempt
2644                                  * to create a fake response.
2645                                  */
2646                                 if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2647                                     FR_ISBLOCK(pass)) {
2648 
2649                                         if (fr_make_icmp(fin) == 0) {
2650                                                 IPF_BUMP(
2651                                                 ifs->ifs_frstats[out].fr_ret);
2652                                         }
2653                                         /* 
2654                                          * we drop packet silently in case we
2655                                          * failed assemble fake response for it
2656                                          */
2657                                         else if (*mp != NULL) {
2658                                                 FREE_MB_T(*mp);
2659                                                 m = *mp = NULL;
2660                                         }
2661 
2662                                         IPF_BUMP(
2663                                             ifs->ifs_frstats[out].fr_block);
2664                                         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2665                                         
2666                                         return (0);
2667                                 }
2668 #endif  /* _KERNEL && SOLARIS2 >= 10 */
2669 
2670                                 (void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2671                                 IPF_BUMP(ifs->ifs_frstats[out].fr_ret);
2672 
2673                         } else if (((pass & FR_RETMASK) == FR_RETRST) &&
2674                                    !(fin->fin_flx & FI_SHORT)) {
2675 
2676 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2677                                 /*
2678                                  * Assume it's possible to enter insane rule:
2679                                  *      pass return-rst in proto tcp ...
2680                                  * then we have no other option than to forward
2681                                  * packet on loopback and give up any attempt
2682                                  * to create a fake response.
2683                                  */
2684                                 if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2685                                     FR_ISBLOCK(pass)) {
2686                                         if (fr_make_rst(fin) == 0) {
2687                                                 IPF_BUMP(
2688                                                 ifs->ifs_frstats[out].fr_ret);
2689                                         }
2690                                         else if (mp != NULL) {
2691                                         /* 
2692                                          * we drop packet silently in case we
2693                                          * failed assemble fake response for it
2694                                          */
2695                                                 FREE_MB_T(*mp);
2696                                                 m = *mp = NULL;
2697                                         }
2698 
2699                                         IPF_BUMP(
2700                                             ifs->ifs_frstats[out].fr_block);
2701                                         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2702                                         
2703                                         return (0);
2704                                  }
2705 #endif /* _KERNEL && _SOLARIS2 >= 10 */
2706                                 if (fr_send_reset(fin) == 0) {
2707                                         IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2708                                 }
2709                         }
2710                 } else {
2711                         if (pass & FR_RETRST)
2712                                 fin->fin_error = ECONNRESET;
2713                 }
2714         }
2715 
2716         /*
2717          * If we didn't drop off the bottom of the list of rules (and thus
2718          * the 'current' rule fr is not NULL), then we may have some extra
2719          * instructions about what to do with a packet.
2720          * Once we're finished return to our caller, freeing the packet if
2721          * we are dropping it (* BSD ONLY *).
2722          * Reassign m from fin_m as we may have a new buffer, now.
2723          */
2724 filtered:
2725         m = fin->fin_m;
2726 
2727         if (fr != NULL) {
2728                 frdest_t *fdp;
2729 
2730                 fdp = &fr->fr_tifs[fin->fin_rev];
2731 
2732                 if (!out && (pass & FR_FASTROUTE)) {
2733                         /*
2734                          * For fastroute rule, no destioation interface defined
2735                          * so pass NULL as the frdest_t parameter
2736                          */
2737                         (void) fr_fastroute(m, mp, fin, NULL);
2738                         m = *mp = NULL;
2739                 } else if ((fdp->fd_ifp != NULL) &&
2740                            (fdp->fd_ifp != (struct ifnet *)-1)) {
2741                         /* this is for to rules: */
2742                         (void) fr_fastroute(m, mp, fin, fdp);
2743                         m = *mp = NULL;
2744                 }
2745 
2746                 /*
2747                  * Send a duplicated packet.
2748                  */
2749                 if (mc != NULL) {
2750 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2751                         /*
2752                          * We are going to compute chksum for copies of loopback packets
2753                          * only. IP stack does not compute chksums at all for loopback
2754                          * packets. We want to get it fixed in their copies, since those
2755                          * are going to be sent to network.
2756                          */
2757                         if (IPF_IS_LOOPBACK(qpi->qpi_flags))
2758                                 fr_calc_chksum(fin, mc);
2759 #endif
2760                         (void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2761                 }
2762         }
2763 
2764         if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT))
2765                 nat_uncreate(fin);
2766 
2767         /*
2768          * This late because the likes of fr_fastroute() use fin_fr.
2769          */
2770         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2771 
2772 finished:
2773         if (!FR_ISPASS(pass)) {
2774                 IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2775                 if (*mp != NULL) {
2776                         FREE_MB_T(*mp);
2777                         m = *mp = NULL;
2778                 }
2779         } else {
2780                 IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2781 #if defined(_KERNEL) && defined(__sgi)
2782                 if ((fin->fin_hbuf != NULL) &&
2783                     (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2784                         COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2785                 }
2786 #endif
2787         }
2788 
2789         SPL_X(s);
2790 
2791 #ifdef _KERNEL
2792 # if OpenBSD >= 200311
2793         if (FR_ISPASS(pass) && (v == 4)) {
2794                 ip = fin->fin_ip;
2795                 ip->ip_len = ntohs(ip->ip_len);
2796                 ip->ip_off = ntohs(ip->ip_off);
2797         }
2798 # endif
2799         return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2800 #else /* _KERNEL */
2801         FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2802         if ((pass & FR_NOMATCH) != 0)
2803                 return 1;
2804 
2805         if ((pass & FR_RETMASK) != 0)
2806                 switch (pass & FR_RETMASK)
2807                 {
2808                 case FR_RETRST :
2809                         return 3;
2810                 case FR_RETICMP :
2811                         return 4;
2812                 case FR_FAKEICMP :
2813                         return 5;
2814                 }
2815 
2816         switch (pass & FR_CMDMASK)
2817         {
2818         case FR_PASS :
2819                 return 0;
2820         case FR_BLOCK :
2821                 return -1;
2822         case FR_AUTH :
2823                 return -2;
2824         case FR_ACCOUNT :
2825                 return -3;
2826         case FR_PREAUTH :
2827                 return -4;
2828         }
2829         return 2;
2830 #endif /* _KERNEL */
2831 }
2832 
2833 
2834 #ifdef  IPFILTER_LOG
2835 /* ------------------------------------------------------------------------ */
2836 /* Function:    fr_dolog                                                    */
2837 /* Returns:     frentry_t* - returns contents of fin_fr (no change made)    */
2838 /* Parameters:  fin(I) - pointer to packet information                      */
2839 /*              passp(IO) - pointer to current/new filter decision (unused) */
2840 /*                                                                          */
2841 /* Checks flags set to see how a packet should be logged, if it is to be    */
2842 /* logged.  Adjust statistics based on its success or not.                  */
2843 /* ------------------------------------------------------------------------ */
2844 frentry_t *fr_dolog(fin, passp)
2845 fr_info_t *fin;
2846 u_32_t *passp;
2847 {
2848         u_32_t pass;
2849         int out;
2850         ipf_stack_t *ifs = fin->fin_ifs;
2851 
2852         out = fin->fin_out;
2853         pass = *passp;
2854 
2855         if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2856                 pass |= FF_LOGNOMATCH;
2857                 IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2858                 goto logit;
2859         } else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2860             (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2861                 if ((pass & FR_LOGMASK) != FR_LOGP)
2862                         pass |= FF_LOGPASS;
2863                 IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2864                 goto logit;
2865         } else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2866                    (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2867                 if ((pass & FR_LOGMASK) != FR_LOGB)
2868                         pass |= FF_LOGBLOCK;
2869                 IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2870 logit:
2871                 if (ipflog(fin, pass) == -1) {
2872                         IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2873 
2874                         /*
2875                          * If the "or-block" option has been used then
2876                          * block the packet if we failed to log it.
2877                          */
2878                         if ((pass & FR_LOGORBLOCK) &&
2879                             FR_ISPASS(pass)) {
2880                                 pass &= ~FR_CMDMASK;
2881                                 pass |= FR_BLOCK;
2882                         }
2883                 }
2884                 *passp = pass;
2885         }
2886 
2887         return fin->fin_fr;
2888 }
2889 #endif /* IPFILTER_LOG */
2890 
2891 
2892 /* ------------------------------------------------------------------------ */
2893 /* Function:    ipf_cksum                                                   */
2894 /* Returns:     u_short - IP header checksum                                */
2895 /* Parameters:  addr(I) - pointer to start of buffer to checksum            */
2896 /*              len(I)  - length of buffer in bytes                         */
2897 /*                                                                          */
2898 /* Calculate the two's complement 16 bit checksum of the buffer passed.     */
2899 /*                                                                          */
2900 /* N.B.: addr should be 16bit aligned.                                      */
2901 /* ------------------------------------------------------------------------ */
2902 u_short ipf_cksum(addr, len)
2903 u_short *addr;
2904 int len;
2905 {
2906         u_32_t sum = 0;
2907 
2908         for (sum = 0; len > 1; len -= 2)
2909                 sum += *addr++;
2910 
2911         /* mop up an odd byte, if necessary */
2912         if (len == 1)
2913                 sum += *(u_char *)addr;
2914 
2915         /*
2916          * add back carry outs from top 16 bits to low 16 bits
2917          */
2918         sum = (sum >> 16) + (sum & 0xffff);   /* add hi 16 to low 16 */
2919         sum += (sum >> 16);                       /* add carry */
2920         return (u_short)(~sum);
2921 }
2922 
2923 
2924 /* ------------------------------------------------------------------------ */
2925 /* Function:    fr_cksum                                                    */
2926 /* Returns:     u_short - layer 4 checksum                                  */
2927 /* Parameters:  m(I  )     - pointer to buffer holding packet               */
2928 /*              ip(I)      - pointer to IP header                           */
2929 /*              l4proto(I) - protocol to caclulate checksum for             */
2930 /*              l4hdr(I)   - pointer to layer 4 header                      */
2931 /*                                                                          */
2932 /* Calculates the TCP checksum for the packet held in "m", using the data   */
2933 /* in the IP header "ip" to seed it.                                        */
2934 /*                                                                          */
2935 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2936 /* and the TCP header.  We also assume that data blocks aren't allocated in */
2937 /* odd sizes.                                                               */
2938 /*                                                                          */
2939 /* Expects ip_len to be in host byte order when called.                     */
2940 /* ------------------------------------------------------------------------ */
2941 u_short fr_cksum(m, ip, l4proto, l4hdr)
2942 mb_t *m;
2943 ip_t *ip;
2944 int l4proto;
2945 void *l4hdr;
2946 {
2947         u_short *sp, slen, sumsave, l4hlen, *csump;
2948         u_int sum, sum2;
2949         int hlen;
2950 #ifdef  USE_INET6
2951         ip6_t *ip6;
2952 #endif
2953 
2954         csump = NULL;
2955         sumsave = 0;
2956         l4hlen = 0;
2957         sp = NULL;
2958         slen = 0;
2959         hlen = 0;
2960         sum = 0;
2961 
2962         /*
2963          * Add up IP Header portion
2964          */
2965 #ifdef  USE_INET6
2966         if (IP_V(ip) == 4) {
2967 #endif
2968                 hlen = IP_HL(ip) << 2;
2969                 slen = ip->ip_len - hlen;
2970                 sum = htons((u_short)l4proto);
2971                 sum += htons(slen);
2972                 sp = (u_short *)&ip->ip_src;
2973                 sum += *sp++;   /* ip_src */
2974                 sum += *sp++;
2975                 sum += *sp++;   /* ip_dst */
2976                 sum += *sp++;
2977 #ifdef  USE_INET6
2978         } else if (IP_V(ip) == 6) {
2979                 ip6 = (ip6_t *)ip;
2980                 hlen = sizeof(*ip6);
2981                 slen = ntohs(ip6->ip6_plen);
2982                 sum = htons((u_short)l4proto);
2983                 sum += htons(slen);
2984                 sp = (u_short *)&ip6->ip6_src;
2985                 sum += *sp++;   /* ip6_src */
2986                 sum += *sp++;
2987                 sum += *sp++;
2988                 sum += *sp++;
2989                 sum += *sp++;
2990                 sum += *sp++;
2991                 sum += *sp++;
2992                 sum += *sp++;
2993                 sum += *sp++;   /* ip6_dst */
2994                 sum += *sp++;
2995                 sum += *sp++;
2996                 sum += *sp++;
2997                 sum += *sp++;
2998                 sum += *sp++;
2999                 sum += *sp++;
3000                 sum += *sp++;
3001         }
3002 #endif
3003 
3004         switch (l4proto)
3005         {
3006         case IPPROTO_UDP :
3007                 csump = &((udphdr_t *)l4hdr)->uh_sum;
3008                 l4hlen = sizeof(udphdr_t);
3009                 break;
3010 
3011         case IPPROTO_TCP :
3012                 csump = &((tcphdr_t *)l4hdr)->th_sum;
3013                 l4hlen = sizeof(tcphdr_t);
3014                 break;
3015         case IPPROTO_ICMP :
3016                 csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
3017                 l4hlen = 4;
3018                 sum = 0;
3019                 break;
3020         default :
3021                 break;
3022         }
3023 
3024         if (csump != NULL) {
3025                 sumsave = *csump;
3026                 *csump = 0;
3027         }
3028 
3029         l4hlen = l4hlen;        /* LINT */
3030 
3031 #ifdef  _KERNEL
3032 # ifdef MENTAT
3033         {
3034         void *rp = m->b_rptr;
3035 
3036         if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
3037                 m->b_rptr = (u_char *)ip;
3038         sum2 = ip_cksum(m, hlen, sum);  /* hlen == offset */
3039         m->b_rptr = rp;
3040         sum2 = (sum2 & 0xffff) + (sum2 >> 16);
3041         sum2 = ~sum2 & 0xffff;
3042         }
3043 # else /* MENTAT */
3044 #  if defined(BSD) || defined(sun)
3045 #   if BSD >= 199103
3046         m->m_data += hlen;
3047 #   else
3048         m->m_off += hlen;
3049 #   endif
3050         m->m_len -= hlen;
3051         sum2 = in_cksum(m, slen);
3052         m->m_len += hlen;
3053 #   if BSD >= 199103
3054         m->m_data -= hlen;
3055 #   else
3056         m->m_off -= hlen;
3057 #   endif
3058         /*
3059          * Both sum and sum2 are partial sums, so combine them together.
3060          */
3061         sum += ~sum2 & 0xffff;
3062         while (sum > 0xffff)
3063                 sum = (sum & 0xffff) + (sum >> 16);
3064         sum2 = ~sum & 0xffff;
3065 #  else /* defined(BSD) || defined(sun) */
3066 {
3067         union {
3068                 u_char  c[2];
3069                 u_short s;
3070         } bytes;
3071         u_short len = ip->ip_len;
3072 #   if defined(__sgi)
3073         int add;
3074 #   endif
3075 
3076         /*
3077          * Add up IP Header portion
3078          */
3079         if (sp != (u_short *)l4hdr)
3080                 sp = (u_short *)l4hdr;
3081 
3082         switch (l4proto)
3083         {
3084         case IPPROTO_UDP :
3085                 sum += *sp++;   /* sport */
3086                 sum += *sp++;   /* dport */
3087                 sum += *sp++;   /* udp length */
3088                 sum += *sp++;   /* checksum */
3089                 break;
3090 
3091         case IPPROTO_TCP :
3092                 sum += *sp++;   /* sport */
3093                 sum += *sp++;   /* dport */
3094                 sum += *sp++;   /* seq */
3095                 sum += *sp++;
3096                 sum += *sp++;   /* ack */
3097                 sum += *sp++;
3098                 sum += *sp++;   /* off */
3099                 sum += *sp++;   /* win */
3100                 sum += *sp++;   /* checksum */
3101                 sum += *sp++;   /* urp */
3102                 break;
3103         case IPPROTO_ICMP :
3104                 sum = *sp++;    /* type/code */
3105                 sum += *sp++;   /* checksum */
3106                 break;
3107         }
3108 
3109 #   ifdef       __sgi
3110         /*
3111          * In case we had to copy the IP & TCP header out of mbufs,
3112          * skip over the mbuf bits which are the header
3113          */
3114         if ((caddr_t)ip != mtod(m, caddr_t)) {
3115                 hlen = (caddr_t)sp - (caddr_t)ip;
3116                 while (hlen) {
3117                         add = MIN(hlen, m->m_len);
3118                         sp = (u_short *)(mtod(m, caddr_t) + add);
3119                         hlen -= add;
3120                         if (add == m->m_len) {
3121                                 m = m->m_next;
3122                                 if (!hlen) {
3123                                         if (!m)
3124                                                 break;
3125                                         sp = mtod(m, u_short *);
3126                                 }
3127                                 PANIC((!m),("fr_cksum(1): not enough data"));
3128                         }
3129                 }
3130         }
3131 #   endif
3132 
3133         len -= (l4hlen + hlen);
3134         if (len <= 0)
3135                 goto nodata;
3136 
3137         while (len > 1) {
3138                 if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3139                         m = m->m_next;
3140                         PANIC((!m),("fr_cksum(2): not enough data"));
3141                         sp = mtod(m, u_short *);
3142                 }
3143                 if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3144                         bytes.c[0] = *(u_char *)sp;
3145                         m = m->m_next;
3146                         PANIC((!m),("fr_cksum(3): not enough data"));
3147                         sp = mtod(m, u_short *);
3148                         bytes.c[1] = *(u_char *)sp;
3149                         sum += bytes.s;
3150                         sp = (u_short *)((u_char *)sp + 1);
3151                 }
3152                 if ((u_long)sp & 1) {
3153                         bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3154                         sum += bytes.s;
3155                 } else
3156                         sum += *sp++;
3157                 len -= 2;
3158         }
3159 
3160         if (len != 0)
3161                 sum += ntohs(*(u_char *)sp << 8);
3162 nodata:
3163         while (sum > 0xffff)
3164                 sum = (sum & 0xffff) + (sum >> 16);
3165         sum2 = (u_short)(~sum & 0xffff);
3166 }
3167 #  endif /*  defined(BSD) || defined(sun) */
3168 # endif /* MENTAT */
3169 #else /* _KERNEL */
3170         for (; slen > 1; slen -= 2)
3171                 sum += *sp++;
3172         if (slen)
3173                 sum += ntohs(*(u_char *)sp << 8);
3174         while (sum > 0xffff)
3175                 sum = (sum & 0xffff) + (sum >> 16);
3176         sum2 = (u_short)(~sum & 0xffff);
3177 #endif /* _KERNEL */
3178         if (csump != NULL)
3179                 *csump = sumsave;
3180         return sum2;
3181 }
3182 
3183 
3184 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3185     defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3186 /*
3187  * Copyright (c) 1982, 1986, 1988, 1991, 1993
3188  *      The Regents of the University of California.  All rights reserved.
3189  *
3190  * Redistribution and use in source and binary forms, with or without
3191  * modification, are permitted provided that the following conditions
3192  * are met:
3193  * 1. Redistributions of source code must retain the above copyright
3194  *    notice, this list of conditions and the following disclaimer.
3195  * 2. Redistributions in binary form must reproduce the above copyright
3196  *    notice, this list of conditions and the following disclaimer in the
3197  *    documentation and/or other materials provided with the distribution.
3198  * 3. Neither the name of the University nor the names of its contributors
3199  *    may be used to endorse or promote products derived from this software
3200  *    without specific prior written permission.
3201  *
3202  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3203  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3204  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3205  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3206  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3207  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3208  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3209  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3210  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3211  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3212  * SUCH DAMAGE.
3213  *
3214  *      @(#)uipc_mbuf.c 8.2 (Berkeley) 1/4/94
3215  * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3216  */
3217 /*
3218  * Copy data from an mbuf chain starting "off" bytes from the beginning,
3219  * continuing for "len" bytes, into the indicated buffer.
3220  */
3221 void
3222 m_copydata(m, off, len, cp)
3223         mb_t *m;
3224         int off;
3225         int len;
3226         caddr_t cp;
3227 {
3228         unsigned count;
3229 
3230         if (off < 0 || len < 0)
3231                 panic("m_copydata");
3232         while (off > 0) {
3233                 if (m == 0)
3234                         panic("m_copydata");
3235                 if (off < m->m_len)
3236                         break;
3237                 off -= m->m_len;
3238                 m = m->m_next;
3239         }
3240         while (len > 0) {
3241                 if (m == 0)
3242                         panic("m_copydata");
3243                 count = MIN(m->m_len - off, len);
3244                 bcopy(mtod(m, caddr_t) + off, cp, count);
3245                 len -= count;
3246                 cp += count;
3247                 off = 0;
3248                 m = m->m_next;
3249         }
3250 }
3251 
3252 
3253 /*
3254  * Copy data from a buffer back into the indicated mbuf chain,
3255  * starting "off" bytes from the beginning, extending the mbuf
3256  * chain if necessary.
3257  */
3258 void
3259 m_copyback(m0, off, len, cp)
3260         struct  mbuf *m0;
3261         int off;
3262         int len;
3263         caddr_t cp;
3264 {
3265         int mlen;
3266         struct mbuf *m = m0, *n;
3267         int totlen = 0;
3268 
3269         if (m0 == 0)
3270                 return;
3271         while (off > (mlen = m->m_len)) {
3272                 off -= mlen;
3273                 totlen += mlen;
3274                 if (m->m_next == 0) {
3275                         n = m_getclr(M_DONTWAIT, m->m_type);
3276                         if (n == 0)
3277                                 goto out;
3278                         n->m_len = min(MLEN, len + off);
3279                         m->m_next = n;
3280                 }
3281                 m = m->m_next;
3282         }
3283         while (len > 0) {
3284                 mlen = min(m->m_len - off, len);
3285                 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3286                 cp += mlen;
3287                 len -= mlen;
3288                 mlen += off;
3289                 off = 0;
3290                 totlen += mlen;
3291                 if (len == 0)
3292                         break;
3293                 if (m->m_next == 0) {
3294                         n = m_get(M_DONTWAIT, m->m_type);
3295                         if (n == 0)
3296                                 break;
3297                         n->m_len = min(MLEN, len);
3298                         m->m_next = n;
3299                 }
3300                 m = m->m_next;
3301         }
3302 out:
3303 #if 0
3304         if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3305                 m->m_pkthdr.len = totlen;
3306 #endif
3307         return;
3308 }
3309 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3310 
3311 
3312 /* ------------------------------------------------------------------------ */
3313 /* Function:    fr_findgroup                                                */
3314 /* Returns:     frgroup_t * - NULL = group not found, else pointer to group */
3315 /* Parameters:  group(I) - group name to search for                         */
3316 /*              unit(I)  - device to which this group belongs               */
3317 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3318 /*              fgpp(O)  - pointer to place to store pointer to the pointer */
3319 /*                         to where to add the next (last) group or where   */
3320 /*                         to delete group from.                            */
3321 /*                                                                          */
3322 /* Search amongst the defined groups for a particular group number.         */
3323 /* ------------------------------------------------------------------------ */
3324 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3325 char *group;
3326 minor_t unit;
3327 int set;
3328 frgroup_t ***fgpp;
3329 ipf_stack_t *ifs;
3330 {
3331         frgroup_t *fg, **fgp;
3332 
3333         /*
3334          * Which list of groups to search in is dependent on which list of
3335          * rules are being operated on.
3336          */
3337         fgp = &ifs->ifs_ipfgroups[unit][set];
3338 
3339         while ((fg = *fgp) != NULL) {
3340                 if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3341                         break;
3342                 else
3343                         fgp = &fg->fg_next;
3344         }
3345         if (fgpp != NULL)
3346                 *fgpp = fgp;
3347         return fg;
3348 }
3349 
3350 
3351 /* ------------------------------------------------------------------------ */
3352 /* Function:    fr_addgroup                                                 */
3353 /* Returns:     frgroup_t * - NULL == did not create group,                 */
3354 /*                            != NULL == pointer to the group               */
3355 /* Parameters:  num(I)   - group number to add                              */
3356 /*              head(I)  - rule pointer that is using this as the head      */
3357 /*              flags(I) - rule flags which describe the type of rule it is */
3358 /*              unit(I)  - device to which this group will belong to        */
3359 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3360 /* Write Locks: ipf_mutex                                                   */
3361 /*                                                                          */
3362 /* Add a new group head, or if it already exists, increase the reference    */
3363 /* count to it.                                                             */
3364 /* ------------------------------------------------------------------------ */
3365 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3366 char *group;
3367 void *head;
3368 u_32_t flags;
3369 minor_t unit;
3370 int set;
3371 ipf_stack_t *ifs;
3372 {
3373         frgroup_t *fg, **fgp;
3374         u_32_t gflags;
3375 
3376         if (group == NULL)
3377                 return NULL;
3378 
3379         if (unit == IPL_LOGIPF && *group == '\0')
3380                 return NULL;
3381 
3382         fgp = NULL;
3383         gflags = flags & FR_INOUT;
3384 
3385         fg = fr_findgroup(group, unit, set, &fgp, ifs);
3386         if (fg != NULL) {
3387                 if (fg->fg_flags == 0)
3388                         fg->fg_flags = gflags;
3389                 else if (gflags != fg->fg_flags)
3390                         return NULL;
3391                 fg->fg_ref++;
3392                 return fg;
3393         }
3394         KMALLOC(fg, frgroup_t *);
3395         if (fg != NULL) {
3396                 fg->fg_head = head;
3397                 fg->fg_start = NULL;
3398                 fg->fg_next = *fgp;
3399                 bcopy(group, fg->fg_name, FR_GROUPLEN);
3400                 fg->fg_flags = gflags;
3401                 fg->fg_ref = 1;
3402                 *fgp = fg;
3403         }
3404         return fg;
3405 }
3406 
3407 
3408 /* ------------------------------------------------------------------------ */
3409 /* Function:    fr_delgroup                                                 */
3410 /* Returns:     Nil                                                         */
3411 /* Parameters:  group(I) - group name to delete                             */
3412 /*              unit(I)  - device to which this group belongs               */
3413 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3414 /* Write Locks: ipf_mutex                                                   */
3415 /*                                                                          */
3416 /* Attempt to delete a group head.                                          */
3417 /* Only do this when its reference count reaches 0.                         */
3418 /* ------------------------------------------------------------------------ */
3419 void fr_delgroup(group, unit, set, ifs)
3420 char *group;
3421 minor_t unit;
3422 int set;
3423 ipf_stack_t *ifs;
3424 {
3425         frgroup_t *fg, **fgp;
3426 
3427         fg = fr_findgroup(group, unit, set, &fgp, ifs);
3428         if (fg == NULL)
3429                 return;
3430 
3431         fg->fg_ref--;
3432         if (fg->fg_ref == 0) {
3433                 *fgp = fg->fg_next;
3434                 KFREE(fg);
3435         }
3436 }
3437 
3438 
3439 /* ------------------------------------------------------------------------ */
3440 /* Function:    fr_getrulen                                                 */
3441 /* Returns:     frentry_t * - NULL == not found, else pointer to rule n     */
3442 /* Parameters:  unit(I)  - device for which to count the rule's number      */
3443 /*              flags(I) - which set of rules to find the rule in           */
3444 /*              group(I) - group name                                       */
3445 /*              n(I)     - rule number to find                              */
3446 /*                                                                          */
3447 /* Find rule # n in group # g and return a pointer to it.  Return NULl if   */
3448 /* group # g doesn't exist or there are less than n rules in the group.     */
3449 /* ------------------------------------------------------------------------ */
3450 frentry_t *fr_getrulen(unit, group, n, ifs)
3451 int unit;
3452 char *group;
3453 u_32_t n;
3454 ipf_stack_t *ifs;
3455 {
3456         frentry_t *fr;
3457         frgroup_t *fg;
3458 
3459         fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3460         if (fg == NULL)
3461                 return NULL;
3462         for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3463                 ;
3464         if (n != 0)
3465                 return NULL;
3466         return fr;
3467 }
3468 
3469 
3470 /* ------------------------------------------------------------------------ */
3471 /* Function:    fr_rulen                                                    */
3472 /* Returns:     int - >= 0 - rule number, -1 == search failed               */
3473 /* Parameters:  unit(I) - device for which to count the rule's number       */
3474 /*              fr(I)   - pointer to rule to match                          */
3475 /*                                                                          */
3476 /* Return the number for a rule on a specific filtering device.             */
3477 /* ------------------------------------------------------------------------ */
3478 int fr_rulen(unit, fr, ifs)
3479 int unit;
3480 frentry_t *fr;
3481 ipf_stack_t *ifs;
3482 {
3483         frentry_t *fh;
3484         frgroup_t *fg;
3485         u_32_t n = 0;
3486 
3487         if (fr == NULL)
3488                 return -1;
3489         fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3490         if (fg == NULL)
3491                 return -1;
3492         for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3493                 if (fh == fr)
3494                         break;
3495         if (fh == NULL)
3496                 return -1;
3497         return n;
3498 }
3499 
3500 
3501 /* ------------------------------------------------------------------------ */
3502 /* Function:    frflushlist                                                 */
3503 /* Returns:     int - >= 0 - number of flushed rules                        */
3504 /* Parameters:  set(I)   - which set of rules (inactive/inactive) this is   */
3505 /*              unit(I)  - device for which to flush rules                  */
3506 /*              flags(I) - which set of rules to flush                      */
3507 /*              nfreedp(O) - pointer to int where flush count is stored     */
3508 /*              listp(I)   - pointer to list to flush pointer               */
3509 /* Write Locks: ipf_mutex                                                   */
3510 /*                                                                          */
3511 /* Recursively flush rules from the list, descending groups as they are     */
3512 /* encountered.  if a rule is the head of a group and it has lost all its   */
3513 /* group members, then also delete the group reference.  nfreedp is needed  */
3514 /* to store the accumulating count of rules removed, whereas the returned   */
3515 /* value is just the number removed from the current list.  The latter is   */
3516 /* needed to correctly adjust reference counts on rules that define groups. */
3517 /*                                                                          */
3518 /* NOTE: Rules not loaded from user space cannot be flushed.                */
3519 /* ------------------------------------------------------------------------ */
3520 static int frflushlist(set, unit, nfreedp, listp, ifs)
3521 int set;
3522 minor_t unit;
3523 int *nfreedp;
3524 frentry_t **listp;
3525 ipf_stack_t *ifs;
3526 {
3527         int freed = 0;
3528         frentry_t *fp;
3529 
3530         while ((fp = *listp) != NULL) {
3531                 if ((fp->fr_type & FR_T_BUILTIN) ||
3532                     !(fp->fr_flags & FR_COPIED)) {
3533                         listp = &fp->fr_next;
3534                         continue;
3535                 }
3536                 *listp = fp->fr_next;
3537                 if (fp->fr_grp != NULL) {
3538                         (void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3539                 }
3540 
3541                 if (fp->fr_grhead != NULL) {
3542                         fr_delgroup(fp->fr_grhead, unit, set, ifs);
3543                         *fp->fr_grhead = '\0';
3544                 }
3545 
3546                 ASSERT(fp->fr_ref > 0);
3547                 fp->fr_next = NULL;
3548                 if (fr_derefrule(&fp, ifs) == 0)
3549                         freed++;
3550         }
3551         *nfreedp += freed;
3552         return freed;
3553 }
3554 
3555 
3556 /* ------------------------------------------------------------------------ */
3557 /* Function:    frflush                                                     */
3558 /* Returns:     int - >= 0 - number of flushed rules                        */
3559 /* Parameters:  unit(I)  - device for which to flush rules                  */
3560 /*              flags(I) - which set of rules to flush                      */
3561 /*                                                                          */
3562 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3563 /* and IPv6) as defined by the value of flags.                              */
3564 /* ------------------------------------------------------------------------ */
3565 int frflush(unit, proto, flags, ifs)
3566 minor_t unit;
3567 int proto, flags;
3568 ipf_stack_t *ifs;
3569 {
3570         int flushed = 0, set;
3571 
3572         WRITE_ENTER(&ifs->ifs_ipf_mutex);
3573         bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
3574 
3575         set = ifs->ifs_fr_active;
3576         if ((flags & FR_INACTIVE) == FR_INACTIVE)
3577                 set = 1 - set;
3578 
3579         if (flags & FR_OUTQUE) {
3580                 if (proto == 0 || proto == 6) {
3581                         (void) frflushlist(set, unit,
3582                             &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3583                         (void) frflushlist(set, unit,
3584                             &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3585                 }
3586                 if (proto == 0 || proto == 4) {
3587                         (void) frflushlist(set, unit,
3588                             &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3589                         (void) frflushlist(set, unit,
3590                             &flushed, &ifs->ifs_ipacct[1][set], ifs);
3591                 }
3592         }
3593         if (flags & FR_INQUE) {
3594                 if (proto == 0 || proto == 6) {
3595                         (void) frflushlist(set, unit,
3596                             &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3597                         (void) frflushlist(set, unit,
3598                             &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3599                 }
3600                 if (proto == 0 || proto == 4) {
3601                         (void) frflushlist(set, unit,
3602                             &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3603                         (void) frflushlist(set, unit,
3604                             &flushed, &ifs->ifs_ipacct[0][set], ifs);
3605                 }
3606         }
3607         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3608 
3609         if (unit == IPL_LOGIPF) {
3610                 int tmp;
3611 
3612                 tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3613                 if (tmp >= 0)
3614                         flushed += tmp;
3615         }
3616         return flushed;
3617 }
3618 
3619 
3620 /* ------------------------------------------------------------------------ */
3621 /* Function:    memstr                                                      */
3622 /* Returns:     char *  - NULL if failed, != NULL pointer to matching bytes */
3623 /* Parameters:  src(I)  - pointer to byte sequence to match                 */
3624 /*              dst(I)  - pointer to byte sequence to search                */
3625 /*              slen(I) - match length                                      */
3626 /*              dlen(I) - length available to search in                     */
3627 /*                                                                          */
3628 /* Search dst for a sequence of bytes matching those at src and extend for  */
3629 /* slen bytes.                                                              */
3630 /* ------------------------------------------------------------------------ */
3631 char *memstr(src, dst, slen, dlen)
3632 char *src, *dst;
3633 int slen, dlen;
3634 {
3635         char *s = NULL;
3636 
3637         while (dlen >= slen) {
3638                 if (bcmp(src, dst, slen) == 0) {
3639                         s = dst;
3640                         break;
3641                 }
3642                 dst++;
3643                 dlen--;
3644         }
3645         return s;
3646 }
3647 /* ------------------------------------------------------------------------ */
3648 /* Function:    fr_fixskip                                                  */
3649 /* Returns:     Nil                                                         */
3650 /* Parameters:  listp(IO)    - pointer to start of list with skip rule      */
3651 /*              rp(I)        - rule added/removed with skip in it.          */
3652 /*              addremove(I) - adjustment (-1/+1) to make to skip count,    */
3653 /*                             depending on whether a rule was just added   */
3654 /*                             or removed.                                  */
3655 /*                                                                          */
3656 /* Adjust all the rules in a list which would have skip'd past the position */
3657 /* where we are inserting to skip to the right place given the change.      */
3658 /* ------------------------------------------------------------------------ */
3659 void fr_fixskip(listp, rp, addremove)
3660 frentry_t **listp, *rp;
3661 int addremove;
3662 {
3663         int rules, rn;
3664         frentry_t *fp;
3665 
3666         rules = 0;
3667         for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3668                 rules++;
3669 
3670         if (!fp)
3671                 return;
3672 
3673         for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3674                 if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3675                         fp->fr_arg += addremove;
3676 }
3677 
3678 
3679 #ifdef  _KERNEL
3680 /* ------------------------------------------------------------------------ */
3681 /* Function:    count4bits                                                  */
3682 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3683 /* Parameters:  ip(I) - 32bit IP address                                    */
3684 /*                                                                          */
3685 /* IPv4 ONLY                                                                */
3686 /* count consecutive 1's in bit mask.  If the mask generated by counting    */
3687 /* consecutive 1's is different to that passed, return -1, else return #    */
3688 /* of bits.                                                                 */
3689 /* ------------------------------------------------------------------------ */
3690 int     count4bits(ip)
3691 u_32_t  ip;
3692 {
3693         u_32_t  ipn;
3694         int     cnt = 0, i, j;
3695 
3696         ip = ipn = ntohl(ip);
3697         for (i = 32; i; i--, ipn *= 2)
3698                 if (ipn & 0x80000000)
3699                         cnt++;
3700                 else
3701                         break;
3702         ipn = 0;
3703         for (i = 32, j = cnt; i; i--, j--) {
3704                 ipn *= 2;
3705                 if (j > 0)
3706                         ipn++;
3707         }
3708         if (ipn == ip)
3709                 return cnt;
3710         return -1;
3711 }
3712 
3713 
3714 #ifdef USE_INET6
3715 /* ------------------------------------------------------------------------ */
3716 /* Function:    count6bits                                                  */
3717 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3718 /* Parameters:  msk(I) - pointer to start of IPv6 bitmask                   */
3719 /*                                                                          */
3720 /* IPv6 ONLY                                                                */
3721 /* count consecutive 1's in bit mask.                                       */
3722 /* ------------------------------------------------------------------------ */
3723 int count6bits(msk)
3724 u_32_t *msk;
3725 {
3726         int i = 0, k;
3727         u_32_t j;
3728 
3729         for (k = 3; k >= 0; k--)
3730                 if (msk[k] == 0xffffffff)
3731                         i += 32;
3732                 else {
3733                         for (j = msk[k]; j; j <<= 1)
3734                                 if (j & 0x80000000)
3735                                         i++;
3736                 }
3737         return i;
3738 }
3739 # endif
3740 #endif /* _KERNEL */
3741 
3742 
3743 /* ------------------------------------------------------------------------ */
3744 /* Function:    fr_ifsync                                                   */
3745 /* Returns:     void *    - new interface identifier                        */
3746 /* Parameters:  action(I)  - type of synchronisation to do                  */
3747 /*              v(I)       - IP version being sync'd (v4 or v6)             */
3748 /*              newifp(I)  - interface identifier being introduced/removed  */
3749 /*              oldifp(I)  - interface identifier in a filter rule          */
3750 /*              newname(I) - name associated with newifp interface          */
3751 /*              oldname(I) - name associated with oldifp interface          */
3752 /*              ifs       - pointer to IPF stack instance                   */
3753 /*                                                                          */
3754 /* This function returns what the new value for "oldifp" should be for its  */
3755 /* caller.  In some cases it will not change, in some it will.              */
3756 /* action == IPFSYNC_RESYNC                                                 */
3757 /*   a new value for oldifp will always be looked up, according to oldname, */
3758 /*   the values of newname and newifp are ignored.                          */
3759 /* action == IPFSYNC_NEWIFP                                                 */
3760 /*   if oldname matches newname then we are doing a sync for the matching   */
3761 /*   interface, so we return newifp to be used in place of oldifp.  If the  */
3762 /*   the names don't match, just return oldifp.                             */
3763 /* action == IPFSYNC_OLDIFP                                                 */
3764 /*   if oldifp matches newifp then we are are doing a sync to remove any    */
3765 /*   references to oldifp, so we return "-1".                               */
3766 /* -----                                                                    */
3767 /* NOTE:                                                                    */
3768 /* This function processes NIC event from PF_HOOKS. The action parameter    */
3769 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is       */
3770 /* one single switch statement() in ipf_nic_event_vx() function, which      */
3771 /* translates the HOOK event type to action parameter passed to fr_ifsync.  */
3772 /* The translation table looks as follows:                                  */
3773 /*      event           | action                                            */
3774 /*      ----------------+-------------                                      */
3775 /*      NE_PLUMB        | IPFSYNC_NEWIFP                                    */
3776 /*      NE_UNPLUMB      | IPFSYNC_OLDIFP                                    */
3777 /*    NE_ADDRESS_CHANGE | IPFSYNC_RESYNC                                    */
3778 /*                                                                          */
3779 /* The oldname and oldifp parameters are taken from IPF entry (rule, state  */
3780 /* table entry, NAT table entry, fragment ...). The newname and newifp      */
3781 /* parameters come from hook event data, parameters are taken from event    */
3782 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is        */
3783 /* notified by hook function.                                               */
3784 /*                                                                          */
3785 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3786 /* to plumb the interface, which is already plumbed. In such case we always */
3787 /* get the event from PF_HOOKS as follows:                                  */
3788 /*      event:  NE_PLUMB                                                    */
3789 /*      NIC:    0x0                                                         */
3790 /* ------------------------------------------------------------------------ */
3791 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3792 int action, v;
3793 char *newname, *oldname;
3794 void *newifp, *oldifp;
3795 ipf_stack_t *ifs;
3796 {
3797         void *rval = oldifp;
3798 
3799         switch (action)
3800         {
3801         case IPFSYNC_RESYNC :
3802                 if (oldname[0] != '\0') {
3803                         rval = fr_resolvenic(oldname, v, ifs);
3804                 }
3805                 break;
3806         case IPFSYNC_NEWIFP :
3807                 if (!strncmp(newname, oldname, LIFNAMSIZ))
3808                         rval = newifp;
3809                 break;
3810         case IPFSYNC_OLDIFP :
3811                 /* 
3812                  * If interface gets unplumbed it must be invalidated, which
3813                  * means set all existing references to the interface to -1.
3814                  * We don't want to invalidate references for wildcard
3815                  * (unbound) rules (entries).
3816                  */
3817                 if (newifp == oldifp)
3818                         rval = (oldifp) ? (void *)-1 : NULL;
3819                 break;
3820         }
3821 
3822         return rval;
3823 }
3824 
3825 
3826 /* ------------------------------------------------------------------------ */
3827 /* Function:    frsynclist                                                  */
3828 /* Returns:     void                                                        */
3829 /* Parameters:  action(I) - type of synchronisation to do                   */
3830 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3831 /*              ifp(I)    - interface identifier associated with action     */
3832 /*              ifname(I) - name associated with ifp parameter              */
3833 /*              fr(I)     - pointer to filter rule                          */
3834 /*              ifs       - pointer to IPF stack instance                   */
3835 /* Write Locks: ipf_mutex                                                   */
3836 /*                                                                          */
3837 /* Walk through a list of filter rules and resolve any interface names into */
3838 /* pointers.  Where dynamic addresses are used, also update the IP address  */
3839 /* used in the rule.  The interface pointer is used to limit the lookups to */
3840 /* a specific set of matching names if it is non-NULL.                      */
3841 /* ------------------------------------------------------------------------ */
3842 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3843 int action, v;
3844 void *ifp;
3845 char *ifname;
3846 frentry_t *fr;
3847 ipf_stack_t *ifs;
3848 {
3849         frdest_t *fdp;
3850         int rv, i;
3851 
3852         for (; fr; fr = fr->fr_next) {
3853                 rv = fr->fr_v;
3854                 if (v != 0 && v != rv)
3855                         continue;
3856 
3857                 /*
3858                  * Lookup all the interface names that are part of the rule.
3859                  */
3860                 for (i = 0; i < 4; i++) {
3861                         fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3862                                                    fr->fr_ifnames[i],
3863                                                    ifp, fr->fr_ifas[i],
3864                                                    ifs);
3865                 }
3866 
3867                 fdp = &fr->fr_tifs[0];
3868                 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3869                                            ifp, fdp->fd_ifp, ifs);
3870 
3871                 fdp = &fr->fr_tifs[1];
3872                 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3873                                            ifp, fdp->fd_ifp, ifs);
3874 
3875                 fdp = &fr->fr_dif;
3876                 fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3877                                            ifp, fdp->fd_ifp, ifs);
3878 
3879                 if (action != IPFSYNC_RESYNC)
3880                         continue;
3881 
3882                 if (fr->fr_type == FR_T_IPF) {
3883                         if (fr->fr_satype != FRI_NORMAL &&
3884                             fr->fr_satype != FRI_LOOKUP) {
3885                                 (void)fr_ifpaddr(rv, fr->fr_satype,
3886                                                  fr->fr_ifas[fr->fr_sifpidx],
3887                                                  &fr->fr_src, &fr->fr_smsk,
3888                                                  ifs);
3889                         }
3890                         if (fr->fr_datype != FRI_NORMAL &&
3891                             fr->fr_datype != FRI_LOOKUP) {
3892                                 (void)fr_ifpaddr(rv, fr->fr_datype,
3893                                                  fr->fr_ifas[fr->fr_difpidx],
3894                                                  &fr->fr_dst, &fr->fr_dmsk,
3895                                                  ifs);
3896                         }
3897                 }
3898 
3899 #ifdef  IPFILTER_LOOKUP
3900                 if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3901                     fr->fr_srcptr == NULL) {
3902                         fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3903                                                          fr->fr_srcnum,
3904                                                          &fr->fr_srcfunc, ifs);
3905                 }
3906                 if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3907                     fr->fr_dstptr == NULL) {
3908                         fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3909                                                          fr->fr_dstnum,
3910                                                          &fr->fr_dstfunc, ifs);
3911                 }
3912 #endif
3913         }
3914 }
3915 
3916 
3917 #ifdef  _KERNEL
3918 /* ------------------------------------------------------------------------ */
3919 /* Function:    frsync                                                      */
3920 /* Returns:     void                                                        */
3921 /* Parameters:  action(I) - type of synchronisation to do                   */
3922 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3923 /*              ifp(I)    - interface identifier associated with action     */
3924 /*              name(I)   - name associated with ifp parameter              */
3925 /*                                                                          */
3926 /* frsync() is called when we suspect that the interface list or            */
3927 /* information about interfaces (like IP#) has changed.  Go through all     */
3928 /* filter rules, NAT entries and the state table and check if anything      */
3929 /* needs to be changed/updated.                                             */
3930 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3931 /* in which this was done to support three different types of sync:         */
3932 /* - complete resync of all interface name/identifiers                      */
3933 /* - new interface being announced with its name and identifier             */
3934 /* - interface removal being announced by only its identifier               */
3935 /* ------------------------------------------------------------------------ */
3936 void frsync(action, v, ifp, name, ifs)
3937 int action, v;
3938 void *ifp;
3939 char *name;
3940 ipf_stack_t *ifs;
3941 {
3942         int i;
3943 
3944         WRITE_ENTER(&ifs->ifs_ipf_mutex);
3945         frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3946         frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3947         frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3948         frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3949         frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3950         frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3951         frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3952         frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3953 
3954         for (i = 0; i < IPL_LOGSIZE; i++) {
3955                 frgroup_t *g;
3956 
3957                 for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3958                         frsynclist(action, v, ifp, name, g->fg_start, ifs);
3959                 for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3960                         frsynclist(action, v, ifp, name, g->fg_start, ifs);
3961         }
3962         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3963 }
3964 
3965 #if SOLARIS2 >= 10
3966 /* ------------------------------------------------------------------------ */
3967 /* Function:    fr_syncindex                                                */
3968 /* Returns:     void                                                        */
3969 /* Parameters:  rules     - list of rules to be sync'd                      */
3970 /*              ifp       - interface, which is being sync'd                */
3971 /*              newifp    - new ifindex value for interface                 */
3972 /*                                                                          */
3973 /* Function updates all NIC indecis, which match ifp, in every rule. Every  */
3974 /* NIC index matching ifp, will be updated to newifp.                       */
3975 /* ------------------------------------------------------------------------ */
3976 static void fr_syncindex(rules, ifp, newifp)
3977 frentry_t *rules;
3978 void *ifp;
3979 void *newifp;
3980 {
3981         int i;
3982         frentry_t *fr;
3983 
3984         for (fr = rules; fr != NULL; fr = fr->fr_next) {
3985                 /*
3986                  * Lookup all the interface names that are part of the rule.
3987                  */
3988                 for (i = 0; i < 4; i++)
3989                         if (fr->fr_ifas[i] == ifp)
3990                                 fr->fr_ifas[i] = newifp;
3991 
3992                 for (i = 0; i < 2; i++) {
3993                         if (fr->fr_tifs[i].fd_ifp == ifp)
3994                                 fr->fr_tifs[i].fd_ifp = newifp;
3995                 }
3996 
3997                 if (fr->fr_dif.fd_ifp == ifp)
3998                         fr->fr_dif.fd_ifp = newifp;
3999         }
4000 }
4001 
4002 /* ------------------------------------------------------------------------ */
4003 /* Function:    fr_ifindexsync                                              */
4004 /* Returns:     void                                                        */
4005 /* Parameters:  ifp       - interface, which is being sync'd                */
4006 /*              newifp    - new ifindex value for interface                 */
4007 /*              ifs       - IPF's stack                                     */
4008 /*                                                                          */
4009 /* Function assumes ipf_mutex is locked exclusively.                        */
4010 /*                                                                          */
4011 /* Function updates the NIC references in rules with new interfaces index   */
4012 /* (newifp). Function must process active lists:                            */
4013 /*      with accounting rules (IPv6 and IPv4)                               */
4014 /*      with inbound rules (IPv6 and IPv4)                                  */
4015 /*      with outbound rules (IPv6 and IPv4)                                 */
4016 /* Function also has to take care of rule groups.                           */
4017 /*                                                                          */
4018 /* NOTE: The ipf_mutex is grabbed exclusively by caller (which is always    */
4019 /* nic_event_hook). The hook function also updates state entries, NAT rules */
4020 /* and NAT entries. We want to do all these update atomically to keep the   */
4021 /* NIC references consistent. The ipf_mutex will synchronize event with     */
4022 /* fr_check(), which processes packets, so no packet will enter fr_check(), */
4023 /* while NIC references will be synchronized.                               */
4024 /* ------------------------------------------------------------------------ */
4025 void fr_ifindexsync(ifp, newifp, ifs)
4026 void *ifp;
4027 void *newifp;
4028 ipf_stack_t *ifs;
4029 {
4030         unsigned int    i;
4031         frentry_t *rule_lists[8];
4032         unsigned int    rules = sizeof (rule_lists) / sizeof (frentry_t *);
4033 
4034         rule_lists[0] = ifs->ifs_ipacct[0][ifs->ifs_fr_active];
4035         rule_lists[1] = ifs->ifs_ipacct[1][ifs->ifs_fr_active];
4036         rule_lists[2] = ifs->ifs_ipfilter[0][ifs->ifs_fr_active];
4037         rule_lists[3] = ifs->ifs_ipfilter[1][ifs->ifs_fr_active];
4038         rule_lists[4] = ifs->ifs_ipacct6[0][ifs->ifs_fr_active];
4039         rule_lists[5] = ifs->ifs_ipacct6[1][ifs->ifs_fr_active];
4040         rule_lists[6] = ifs->ifs_ipfilter6[0][ifs->ifs_fr_active];
4041         rule_lists[7] = ifs->ifs_ipfilter6[1][ifs->ifs_fr_active];
4042 
4043         for (i = 0; i < rules; i++) {
4044                 fr_syncindex(rule_lists[i], ifp, newifp);
4045         } 
4046 
4047         /*
4048          * Update rule groups.
4049          */
4050         for (i = 0; i < IPL_LOGSIZE; i++) {
4051                 frgroup_t *g;
4052 
4053                 for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
4054                         fr_syncindex(g->fg_start, ifp, newifp);
4055                 for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
4056                         fr_syncindex(g->fg_start, ifp, newifp);
4057         }
4058 }
4059 #endif
4060 
4061 /*
4062  * In the functions below, bcopy() is called because the pointer being
4063  * copied _from_ in this instance is a pointer to a char buf (which could
4064  * end up being unaligned) and on the kernel's local stack.
4065  */
4066 /* ------------------------------------------------------------------------ */
4067 /* Function:    copyinptr                                                   */
4068 /* Returns:     int - 0 = success, else failure                             */
4069 /* Parameters:  src(I)  - pointer to the source address                     */
4070 /*              dst(I)  - destination address                               */
4071 /*              size(I) - number of bytes to copy                           */
4072 /*                                                                          */
4073 /* Copy a block of data in from user space, given a pointer to the pointer  */
4074 /* to start copying from (src) and a pointer to where to store it (dst).    */
4075 /* NB: src - pointer to user space pointer, dst - kernel space pointer      */
4076 /* ------------------------------------------------------------------------ */
4077 int copyinptr(src, dst, size)
4078 void *src, *dst;
4079 size_t size;
4080 {
4081         caddr_t ca;
4082         int err;
4083 
4084 # ifdef SOLARIS
4085         err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
4086         if (err != 0)
4087                 return err;
4088 # else
4089         bcopy(src, (caddr_t)&ca, sizeof(ca));
4090 # endif
4091         err = COPYIN(ca, dst, size);
4092         return err;
4093 }
4094 
4095 
4096 /* ------------------------------------------------------------------------ */
4097 /* Function:    copyoutptr                                                  */
4098 /* Returns:     int - 0 = success, else failure                             */
4099 /* Parameters:  src(I)  - pointer to the source address                     */
4100 /*              dst(I)  - destination address                               */
4101 /*              size(I) - number of bytes to copy                           */
4102 /*                                                                          */
4103 /* Copy a block of data out to user space, given a pointer to the pointer   */
4104 /* to start copying from (src) and a pointer to where to store it (dst).    */
4105 /* NB: src - kernel space pointer, dst - pointer to user space pointer.     */
4106 /* ------------------------------------------------------------------------ */
4107 int copyoutptr(src, dst, size)
4108 void *src, *dst;
4109 size_t size;
4110 {
4111         caddr_t ca;
4112         int err;
4113 
4114 # ifdef SOLARIS
4115         err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
4116         if (err != 0)
4117                 return err;
4118 # else
4119         bcopy(dst, (caddr_t)&ca, sizeof(ca));
4120 # endif
4121         err = COPYOUT(src, ca, size);
4122         return err;
4123 }
4124 #endif
4125 
4126 
4127 /* ------------------------------------------------------------------------ */
4128 /* Function:    fr_lock                                                     */
4129 /* Returns:     int - 0 = success, else error                               */
4130 /* Parameters:  data(I)  - pointer to lock value to set                     */
4131 /*              lockp(O) - pointer to location to store old lock value      */
4132 /*                                                                          */
4133 /* Get the new value for the lock integer, set it and return the old value  */
4134 /* in *lockp.                                                               */
4135 /* ------------------------------------------------------------------------ */
4136 int fr_lock(data, lockp)
4137 caddr_t data;
4138 int *lockp;
4139 {
4140         int arg, err;
4141 
4142         err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
4143         if (err != 0)
4144                 return (EFAULT);
4145         err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
4146         if (err != 0)
4147                 return (EFAULT);
4148         *lockp = arg;
4149         return (0);
4150 }
4151 
4152 
4153 /* ------------------------------------------------------------------------ */
4154 /* Function:    fr_getstat                                                  */
4155 /* Returns:     Nil                                                         */
4156 /* Parameters:  fiop(I)  - pointer to ipfilter stats structure              */
4157 /*                                                                          */
4158 /* Stores a copy of current pointers, counters, etc, in the friostat        */
4159 /* structure.                                                               */
4160 /* ------------------------------------------------------------------------ */
4161 void fr_getstat(fiop, ifs)
4162 friostat_t *fiop;
4163 ipf_stack_t *ifs;
4164 {
4165         int i, j;
4166 
4167         bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
4168             sizeof(filterstats_t) * 2);
4169         fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
4170         fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
4171         fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
4172         fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
4173 
4174         for (i = 0; i < 2; i++)
4175                 for (j = 0; j < 2; j++) {
4176                         fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4177                         fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4178                         fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4179                         fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4180                 }
4181 
4182         fiop->f_ticks = ifs->ifs_fr_ticks;
4183         fiop->f_active = ifs->ifs_fr_active;
4184         fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4185         fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4186 
4187         fiop->f_running = ifs->ifs_fr_running;
4188         for (i = 0; i < IPL_LOGSIZE; i++) {
4189                 fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4190                 fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4191         }
4192 #ifdef  IPFILTER_LOG
4193         fiop->f_logging = 1;
4194 #else
4195         fiop->f_logging = 0;
4196 #endif
4197         fiop->f_defpass = ifs->ifs_fr_pass;
4198         fiop->f_features = fr_features;
4199         (void) strncpy(fiop->f_version, ipfilter_version,
4200                        sizeof(fiop->f_version));
4201 }
4202 
4203 
4204 #ifdef  USE_INET6
4205 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4206         ICMP6_ECHO_REPLY,       /* 0: ICMP_ECHOREPLY */
4207         -1,                     /* 1: UNUSED */
4208         -1,                     /* 2: UNUSED */
4209         ICMP6_DST_UNREACH,      /* 3: ICMP_UNREACH */
4210         -1,                     /* 4: ICMP_SOURCEQUENCH */
4211         ND_REDIRECT,            /* 5: ICMP_REDIRECT */
4212         -1,                     /* 6: UNUSED */
4213         -1,                     /* 7: UNUSED */
4214         ICMP6_ECHO_REQUEST,     /* 8: ICMP_ECHO */
4215         -1,                     /* 9: UNUSED */
4216         -1,                     /* 10: UNUSED */
4217         ICMP6_TIME_EXCEEDED,    /* 11: ICMP_TIMXCEED */
4218         ICMP6_PARAM_PROB,       /* 12: ICMP_PARAMPROB */
4219         -1,                     /* 13: ICMP_TSTAMP */
4220         -1,                     /* 14: ICMP_TSTAMPREPLY */
4221         -1,                     /* 15: ICMP_IREQ */
4222         -1,                     /* 16: ICMP_IREQREPLY */
4223         -1,                     /* 17: ICMP_MASKREQ */
4224         -1,                     /* 18: ICMP_MASKREPLY */
4225 };
4226 
4227 
4228 int     icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4229         ICMP6_DST_UNREACH_ADDR,         /* 0: ICMP_UNREACH_NET */
4230         ICMP6_DST_UNREACH_ADDR,         /* 1: ICMP_UNREACH_HOST */
4231         -1,                             /* 2: ICMP_UNREACH_PROTOCOL */
4232         ICMP6_DST_UNREACH_NOPORT,       /* 3: ICMP_UNREACH_PORT */
4233         -1,                             /* 4: ICMP_UNREACH_NEEDFRAG */
4234         ICMP6_DST_UNREACH_NOTNEIGHBOR,  /* 5: ICMP_UNREACH_SRCFAIL */
4235         ICMP6_DST_UNREACH_ADDR,         /* 6: ICMP_UNREACH_NET_UNKNOWN */
4236         ICMP6_DST_UNREACH_ADDR,         /* 7: ICMP_UNREACH_HOST_UNKNOWN */
4237         -1,                             /* 8: ICMP_UNREACH_ISOLATED */
4238         ICMP6_DST_UNREACH_ADMIN,        /* 9: ICMP_UNREACH_NET_PROHIB */
4239         ICMP6_DST_UNREACH_ADMIN,        /* 10: ICMP_UNREACH_HOST_PROHIB */
4240         -1,                             /* 11: ICMP_UNREACH_TOSNET */
4241         -1,                             /* 12: ICMP_UNREACH_TOSHOST */
4242         ICMP6_DST_UNREACH_ADMIN,        /* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4243 };
4244 int     icmpreplytype6[ICMP6_MAXTYPE + 1];
4245 #endif
4246 
4247 int     icmpreplytype4[ICMP_MAXTYPE + 1];
4248 
4249 
4250 /* ------------------------------------------------------------------------ */
4251 /* Function:    fr_matchicmpqueryreply                                      */
4252 /* Returns:     int - 1 if "icmp" is a valid reply to "ic" else 0.          */
4253 /* Parameters:  v(I)    - IP protocol version (4 or 6)                      */
4254 /*              ic(I)   - ICMP information                                  */
4255 /*              icmp(I) - ICMP packet header                                */
4256 /*              rev(I)  - direction (0 = forward/1 = reverse) of packet     */
4257 /*                                                                          */
4258 /* Check if the ICMP packet defined by the header pointed to by icmp is a   */
4259 /* reply to one as described by what's in ic.  If it is a match, return 1,  */
4260 /* else return 0 for no match.                                              */
4261 /* ------------------------------------------------------------------------ */
4262 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4263 int v;
4264 icmpinfo_t *ic;
4265 icmphdr_t *icmp;
4266 int rev;
4267 {
4268         int ictype;
4269 
4270         ictype = ic->ici_type;
4271 
4272         if (v == 4) {
4273                 /*
4274                  * If we matched its type on the way in, then when going out
4275                  * it will still be the same type.
4276                  */
4277                 if ((!rev && (icmp->icmp_type == ictype)) ||
4278                     (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4279                         if (icmp->icmp_type != ICMP_ECHOREPLY)
4280                                 return 1;
4281                         if (icmp->icmp_id == ic->ici_id)
4282                                 return 1;
4283                 }
4284         }
4285 #ifdef  USE_INET6
4286         else if (v == 6) {
4287                 if ((!rev && (icmp->icmp_type == ictype)) ||
4288                     (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4289                         if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4290                                 return 1;
4291                         if (icmp->icmp_id == ic->ici_id)
4292                                 return 1;
4293                 }
4294         }
4295 #endif
4296         return 0;
4297 }
4298 
4299 
4300 #ifdef  IPFILTER_LOOKUP
4301 /* ------------------------------------------------------------------------ */
4302 /* Function:    fr_resolvelookup                                            */
4303 /* Returns:     void * - NULL = failure, else success.                      */
4304 /* Parameters:  type(I)     - type of lookup these parameters are for.      */
4305 /*              number(I)   - table number to use when searching            */
4306 /*              funcptr(IO) - pointer to pointer for storing IP address     */
4307 /*                            searching function.                           */
4308 /*              ifs         - ipf stack instance                            */
4309 /*                                                                          */
4310 /* Search for the "table" number passed in amongst those configured for     */
4311 /* that particular type.  If the type is recognised then the function to    */
4312 /* call to do the IP address search will be change, regardless of whether   */
4313 /* or not the "table" number exists.                                        */
4314 /* ------------------------------------------------------------------------ */
4315 static void *fr_resolvelookup(type, number, funcptr, ifs)
4316 u_int type, number;
4317 lookupfunc_t *funcptr;
4318 ipf_stack_t *ifs;
4319 {
4320         char name[FR_GROUPLEN];
4321         iphtable_t *iph;
4322         ip_pool_t *ipo;
4323         void *ptr;
4324 
4325 #if defined(SNPRINTF) && defined(_KERNEL)
4326         (void) SNPRINTF(name, sizeof(name), "%u", number);
4327 #else
4328         (void) sprintf(name, "%u", number);
4329 #endif
4330 
4331         READ_ENTER(&ifs->ifs_ip_poolrw);
4332 
4333         switch (type)
4334         {
4335         case IPLT_POOL :
4336 # if (defined(__osf__) && defined(_KERNEL))
4337                 ptr = NULL;
4338                 *funcptr = NULL;
4339 # else
4340                 ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4341                 ptr = ipo;
4342                 if (ipo != NULL) {
4343                         ATOMIC_INC32(ipo->ipo_ref);
4344                 }
4345                 *funcptr = ip_pool_search;
4346 # endif
4347                 break;
4348         case IPLT_HASH :
4349                 iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4350                 ptr = iph;
4351                 if (iph != NULL) {
4352                         ATOMIC_INC32(iph->iph_ref);
4353                 }
4354                 *funcptr = fr_iphmfindip;
4355                 break;
4356         default:
4357                 ptr = NULL;
4358                 *funcptr = NULL;
4359                 break;
4360         }
4361         RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4362 
4363         return ptr;
4364 }
4365 #endif
4366 
4367 
4368 /* ------------------------------------------------------------------------ */
4369 /* Function:    frrequest                                                   */
4370 /* Returns:     int - 0 == success, > 0 == errno value                      */
4371 /* Parameters:  unit(I)     - device for which this is for                  */
4372 /*              req(I)      - ioctl command (SIOC*)                         */
4373 /*              data(I)     - pointr to ioctl data                          */
4374 /*              set(I)      - 1 or 0 (filter set)                           */
4375 /*              makecopy(I) - flag indicating whether data points to a rule */
4376 /*                            in kernel space & hence doesn't need copying. */
4377 /*                                                                          */
4378 /* This function handles all the requests which operate on the list of      */
4379 /* filter rules.  This includes adding, deleting, insertion.  It is also    */
4380 /* responsible for creating groups when a "head" rule is loaded.  Interface */
4381 /* names are resolved here and other sanity checks are made on the content  */
4382 /* of the rule structure being loaded.  If a rule has user defined timeouts */
4383 /* then make sure they are created and initialised before exiting.          */
4384 /* ------------------------------------------------------------------------ */
4385 int frrequest(unit, req, data, set, makecopy, ifs)
4386 int unit;
4387 ioctlcmd_t req;
4388 int set, makecopy;
4389 caddr_t data;
4390 ipf_stack_t *ifs;
4391 {
4392         frentry_t frd, *fp, *f, **fprev, **ftail;
4393         int error = 0, in, v;
4394         void *ptr, *uptr;
4395         u_int *p, *pp;
4396         frgroup_t *fg;
4397         char *group;
4398 
4399         fg = NULL;
4400         fp = &frd;
4401         if (makecopy != 0) {
4402                 error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4403                 if (error)
4404                         return EFAULT;
4405                 if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4406                         return EINVAL;
4407                 fp->fr_ref = 0;
4408                 fp->fr_flags |= FR_COPIED;
4409         } else {
4410                 fp = (frentry_t *)data;
4411                 if ((fp->fr_type & FR_T_BUILTIN) == 0)
4412                         return EINVAL;
4413                 fp->fr_flags &= ~FR_COPIED;
4414         }
4415 
4416         if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4417             ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4418                 return EINVAL;
4419 
4420         v = fp->fr_v;
4421         uptr = fp->fr_data;
4422 
4423         /*
4424          * Only filter rules for IPv4 or IPv6 are accepted.
4425          */
4426         if (v == 4)
4427                 /*EMPTY*/;
4428 #ifdef  USE_INET6
4429         else if (v == 6)
4430                 /*EMPTY*/;
4431 #endif
4432         else {
4433                 return EINVAL;
4434         }
4435 
4436         /*
4437          * If the rule is being loaded from user space, i.e. we had to copy it
4438          * into kernel space, then do not trust the function pointer in the
4439          * rule.
4440          */
4441         if ((makecopy == 1) && (fp->fr_func != NULL)) {
4442                 if (fr_findfunc(fp->fr_func) == NULL)
4443                         return ESRCH;
4444                 error = fr_funcinit(fp, ifs);
4445                 if (error != 0)
4446                         return error;
4447         }
4448 
4449         ptr = NULL;
4450         /*
4451          * Check that the group number does exist and that its use (in/out)
4452          * matches what the rule is.
4453          */
4454         if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4455                 *fp->fr_grhead = '\0';
4456         group = fp->fr_group;
4457         if (!strncmp(group, "0", FR_GROUPLEN))
4458                 *group = '\0';
4459 
4460         if (FR_ISACCOUNT(fp->fr_flags))
4461                 unit = IPL_LOGCOUNT;
4462 
4463         if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4464                 fg = fr_findgroup(group, unit, set, NULL, ifs);
4465                 if (fg == NULL)
4466                         return ESRCH;
4467                 if (fg->fg_flags == 0)
4468                         fg->fg_flags = fp->fr_flags & FR_INOUT;
4469                 else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4470                         return ESRCH;
4471         }
4472 
4473         in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4474 
4475         /*
4476          * Work out which rule list this change is being applied to.
4477          */
4478         ftail = NULL;
4479         fprev = NULL;
4480         if (unit == IPL_LOGAUTH)
4481                 fprev = &ifs->ifs_ipauth;
4482         else if (v == 4) {
4483                 if (FR_ISACCOUNT(fp->fr_flags))
4484                         fprev = &ifs->ifs_ipacct[in][set];
4485                 else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4486                         fprev = &ifs->ifs_ipfilter[in][set];
4487         } else if (v == 6) {
4488                 if (FR_ISACCOUNT(fp->fr_flags))
4489                         fprev = &ifs->ifs_ipacct6[in][set];
4490                 else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4491                         fprev = &ifs->ifs_ipfilter6[in][set];
4492         }
4493         if (fprev == NULL)
4494                 return ESRCH;
4495 
4496         if (*group != '\0') {
4497             if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4498                         return ESRCH;
4499                 fprev = &fg->fg_start;
4500         }
4501 
4502         ftail = fprev;
4503         for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4504                 if (fp->fr_collect <= f->fr_collect) {
4505                         ftail = fprev;
4506                         f = NULL;
4507                         break;
4508                 }
4509                 fprev = ftail;
4510         }
4511 
4512         /*
4513          * Copy in extra data for the rule.
4514          */
4515         if (fp->fr_dsize != 0) {
4516                 if (makecopy != 0) {
4517                         KMALLOCS(ptr, void *, fp->fr_dsize);
4518                         if (!ptr)
4519                                 return ENOMEM;
4520                         error = COPYIN(uptr, ptr, fp->fr_dsize);
4521                 } else {
4522                         ptr = uptr;
4523                         error = 0;
4524                 }
4525                 if (error != 0) {
4526                         KFREES(ptr, fp->fr_dsize);
4527                         return EFAULT;
4528                 }
4529                 fp->fr_data = ptr;
4530         } else
4531                 fp->fr_data = NULL;
4532 
4533         /*
4534          * Perform per-rule type sanity checks of their members.
4535          */
4536         switch (fp->fr_type & ~FR_T_BUILTIN)
4537         {
4538 #if defined(IPFILTER_BPF)
4539         case FR_T_BPFOPC :
4540                 if (fp->fr_dsize == 0)
4541                         return EINVAL;
4542                 if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4543                         if (makecopy && fp->fr_data != NULL) {
4544                                 KFREES(fp->fr_data, fp->fr_dsize);
4545                         }
4546                         return EINVAL;
4547                 }
4548                 break;
4549 #endif
4550         case FR_T_IPF :
4551                 if (fp->fr_dsize != sizeof(fripf_t)) {
4552                         if (makecopy && fp->fr_data != NULL) {
4553                                 KFREES(fp->fr_data, fp->fr_dsize);
4554                         }
4555                         return EINVAL;
4556                 }
4557 
4558                 /*
4559                  * Allowing a rule with both "keep state" and "with oow" is
4560                  * pointless because adding a state entry to the table will
4561                  * fail with the out of window (oow) flag set.
4562                  */
4563                 if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW)) {
4564                         if (makecopy && fp->fr_data != NULL) {
4565                                 KFREES(fp->fr_data, fp->fr_dsize);
4566                         }
4567                         return EINVAL;
4568                 }
4569 
4570                 switch (fp->fr_satype)
4571                 {
4572                 case FRI_BROADCAST :
4573                 case FRI_DYNAMIC :
4574                 case FRI_NETWORK :
4575                 case FRI_NETMASKED :
4576                 case FRI_PEERADDR :
4577                         if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4578                                 if (makecopy && fp->fr_data != NULL) {
4579                                         KFREES(fp->fr_data, fp->fr_dsize);
4580                                 }
4581                                 return EINVAL;
4582                         }
4583                         break;
4584 #ifdef  IPFILTER_LOOKUP
4585                 case FRI_LOOKUP :
4586                         fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4587                                                          fp->fr_srcnum,
4588                                                          &fp->fr_srcfunc, ifs);
4589                         break;
4590 #endif
4591                 default :
4592                         break;
4593                 }
4594 
4595                 switch (fp->fr_datype)
4596                 {
4597                 case FRI_BROADCAST :
4598                 case FRI_DYNAMIC :
4599                 case FRI_NETWORK :
4600                 case FRI_NETMASKED :
4601                 case FRI_PEERADDR :
4602                         if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4603                                 if (makecopy && fp->fr_data != NULL) {
4604                                         KFREES(fp->fr_data, fp->fr_dsize);
4605                                 }
4606                                 return EINVAL;
4607                         }
4608                         break;
4609 #ifdef  IPFILTER_LOOKUP
4610                 case FRI_LOOKUP :
4611                         fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4612                                                          fp->fr_dstnum,
4613                                                          &fp->fr_dstfunc, ifs);
4614                         break;
4615 #endif
4616                 default :
4617                         break;
4618                 }
4619                 break;
4620         case FR_T_NONE :
4621                 break;
4622         case FR_T_CALLFUNC :
4623                 break;
4624         case FR_T_COMPIPF :
4625                 break;
4626         default :
4627                 if (makecopy && fp->fr_data != NULL) {
4628                         KFREES(fp->fr_data, fp->fr_dsize);
4629                 }
4630                 return EINVAL;
4631         }
4632 
4633         /*
4634          * Lookup all the interface names that are part of the rule.
4635          */
4636         frsynclist(0, 0, NULL, NULL, fp, ifs);
4637         fp->fr_statecnt = 0;
4638 
4639         /*
4640          * Look for an existing matching filter rule, but don't include the
4641          * next or interface pointer in the comparison (fr_next, fr_ifa).
4642          * This elminates rules which are indentical being loaded.  Checksum
4643          * the constant part of the filter rule to make comparisons quicker
4644          * (this meaning no pointers are included).
4645          */
4646         for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4647              p < pp; p++)
4648                 fp->fr_cksum += *p;
4649         pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4650         for (p = (u_int *)fp->fr_data; p < pp; p++)
4651                 fp->fr_cksum += *p;
4652 
4653         WRITE_ENTER(&ifs->ifs_ipf_mutex);
4654         bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
4655 
4656         for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4657                 if ((fp->fr_cksum != f->fr_cksum) ||
4658                     (f->fr_dsize != fp->fr_dsize))
4659                         continue;
4660                 if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4661                         continue;
4662                 if ((!ptr && !f->fr_data) ||
4663                     (ptr && f->fr_data &&
4664                      !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4665                         break;
4666         }
4667 
4668         /*
4669          * If zero'ing statistics, copy current to caller and zero.
4670          */
4671         if (req == (ioctlcmd_t)SIOCZRLST) {
4672                 if (f == NULL)
4673                         error = ESRCH;
4674                 else {
4675                         /*
4676                          * Copy and reduce lock because of impending copyout.
4677                          * Well we should, but if we do then the atomicity of
4678                          * this call and the correctness of fr_hits and
4679                          * fr_bytes cannot be guaranteed.  As it is, this code
4680                          * only resets them to 0 if they are successfully
4681                          * copied out into user space.
4682                          */
4683                         bcopy((char *)f, (char *)fp, sizeof(*f));
4684 
4685                         /*
4686                          * When we copy this rule back out, set the data
4687                          * pointer to be what it was in user space.
4688                          */
4689                         fp->fr_data = uptr;
4690                         error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4691 
4692                         if (error == 0) {
4693                                 if ((f->fr_dsize != 0) && (uptr != NULL))
4694                                         error = COPYOUT(f->fr_data, uptr,
4695                                                         f->fr_dsize);
4696                                 if (error == 0) {
4697                                         f->fr_hits = 0;
4698                                         f->fr_bytes = 0;
4699                                 }
4700                         }
4701                 }
4702 
4703                 if ((ptr != NULL) && (makecopy != 0)) {
4704                         KFREES(ptr, fp->fr_dsize);
4705                 }
4706                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4707                 return error;
4708         }
4709 
4710         if (!f) {
4711                 /*
4712                  * At the end of this, ftail must point to the place where the
4713                  * new rule is to be saved/inserted/added.
4714                  * For SIOCAD*FR, this should be the last rule in the group of
4715                  * rules that have equal fr_collect fields.
4716                  * For SIOCIN*FR, ...
4717                  */
4718                 if (req == (ioctlcmd_t)SIOCADAFR ||
4719                     req == (ioctlcmd_t)SIOCADIFR) {
4720 
4721                         for (ftail = fprev; (f = *ftail) != NULL; ) {
4722                                 if (f->fr_collect > fp->fr_collect)
4723                                         break;
4724                                 ftail = &f->fr_next;
4725                         }
4726                         f = NULL;
4727                         ptr = NULL;
4728                         error = 0;
4729                 } else if (req == (ioctlcmd_t)SIOCINAFR ||
4730                            req == (ioctlcmd_t)SIOCINIFR) {
4731                         while ((f = *fprev) != NULL) {
4732                                 if (f->fr_collect >= fp->fr_collect)
4733                                         break;
4734                                 fprev = &f->fr_next;
4735                         }
4736                         ftail = fprev;
4737                         if (fp->fr_hits != 0) {
4738                                 while (fp->fr_hits && (f = *ftail)) {
4739                                         if (f->fr_collect != fp->fr_collect)
4740                                                 break;
4741                                         fprev = ftail;
4742                                         ftail = &f->fr_next;
4743                                         fp->fr_hits--;
4744                                 }
4745                         }
4746                         f = NULL;
4747                         ptr = NULL;
4748                         error = 0;
4749                 }
4750         }
4751 
4752         /*
4753          * Request to remove a rule.
4754          */
4755         if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4756                 if (!f)
4757                         error = ESRCH;
4758                 else {
4759                         /*
4760                          * Do not allow activity from user space to interfere
4761                          * with rules not loaded that way.
4762                          */
4763                         if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4764                                 error = EPERM;
4765                                 goto done;
4766                         }
4767 
4768                         /*
4769                          * Return EBUSY if the rule is being reference by
4770                          * something else (eg state information.
4771                          */
4772                         if (f->fr_ref > 1) {
4773                                 error = EBUSY;
4774                                 goto done;
4775                         }
4776 #ifdef  IPFILTER_SCAN
4777                         if (f->fr_isctag[0] != '\0' &&
4778                             (f->fr_isc != (struct ipscan *)-1))
4779                                 ipsc_detachfr(f);
4780 #endif
4781                         if (unit == IPL_LOGAUTH) {
4782                                 error = fr_preauthcmd(req, f, ftail, ifs);
4783                                 goto done;
4784                         }
4785                         if (*f->fr_grhead != '\0')
4786                                 fr_delgroup(f->fr_grhead, unit, set, ifs);
4787                         fr_fixskip(ftail, f, -1);
4788                         *ftail = f->fr_next;
4789                         f->fr_next = NULL;
4790                         (void)fr_derefrule(&f, ifs);
4791                 }
4792         } else {
4793                 /*
4794                  * Not removing, so we must be adding/inserting a rule.
4795                  */
4796                 if (f)
4797                         error = EEXIST;
4798                 else {
4799                         if (unit == IPL_LOGAUTH) {
4800                                 error = fr_preauthcmd(req, fp, ftail, ifs);
4801                                 goto done;
4802                         }
4803                         if (makecopy) {
4804                                 KMALLOC(f, frentry_t *);
4805                         } else
4806                                 f = fp;
4807                         if (f != NULL) {
4808                                 if (fp != f)
4809                                         bcopy((char *)fp, (char *)f,
4810                                               sizeof(*f));
4811                                 MUTEX_NUKE(&f->fr_lock);
4812                                 MUTEX_INIT(&f->fr_lock, "filter rule lock");
4813 #ifdef  IPFILTER_SCAN
4814                                 if (f->fr_isctag[0] != '\0' &&
4815                                     ipsc_attachfr(f))
4816                                         f->fr_isc = (struct ipscan *)-1;
4817 #endif
4818                                 f->fr_hits = 0;
4819                                 if (makecopy != 0)
4820                                         f->fr_ref = 1;
4821                                 f->fr_next = *ftail;
4822                                 *ftail = f;
4823                                 if (req == (ioctlcmd_t)SIOCINIFR ||
4824                                     req == (ioctlcmd_t)SIOCINAFR)
4825                                         fr_fixskip(ftail, f, 1);
4826                                 f->fr_grp = NULL;
4827                                 group = f->fr_grhead;
4828                                 if (*group != '\0') {
4829                                         fg = fr_addgroup(group, f, f->fr_flags,
4830                                                          unit, set, ifs);
4831                                         if (fg != NULL)
4832                                                 f->fr_grp = &fg->fg_start;
4833                                 }
4834                         } else
4835                                 error = ENOMEM;
4836                 }
4837         }
4838 done:
4839         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4840         if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4841                 KFREES(ptr, fp->fr_dsize);
4842         }
4843         return (error);
4844 }
4845 
4846 
4847 /* ------------------------------------------------------------------------ */
4848 /* Function:    fr_funcinit                                                 */
4849 /* Returns:     int - 0 == success, else ESRCH: cannot resolve rule details */
4850 /* Parameters:  fr(I) - pointer to filter rule                              */
4851 /*                                                                          */
4852 /* If a rule is a call rule, then check if the function it points to needs  */
4853 /* an init function to be called now the rule has been loaded.              */
4854 /* ------------------------------------------------------------------------ */
4855 static int fr_funcinit(fr, ifs)
4856 frentry_t *fr;
4857 ipf_stack_t *ifs;
4858 {
4859         ipfunc_resolve_t *ft;
4860         int err;
4861 
4862         err = ESRCH;
4863 
4864         for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4865                 if (ft->ipfu_addr == fr->fr_func) {
4866                         err = 0;
4867                         if (ft->ipfu_init != NULL)
4868                                 err = (*ft->ipfu_init)(fr, ifs);
4869                         break;
4870                 }
4871         return err;
4872 }
4873 
4874 
4875 /* ------------------------------------------------------------------------ */
4876 /* Function:    fr_findfunc                                                 */
4877 /* Returns:     ipfunc_t - pointer to function if found, else NULL          */
4878 /* Parameters:  funcptr(I) - function pointer to lookup                     */
4879 /*                                                                          */
4880 /* Look for a function in the table of known functions.                     */
4881 /* ------------------------------------------------------------------------ */
4882 static ipfunc_t fr_findfunc(funcptr)
4883 ipfunc_t funcptr;
4884 {
4885         ipfunc_resolve_t *ft;
4886 
4887         for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4888                 if (ft->ipfu_addr == funcptr)
4889                         return funcptr;
4890         return NULL;
4891 }
4892 
4893 
4894 /* ------------------------------------------------------------------------ */
4895 /* Function:    fr_resolvefunc                                              */
4896 /* Returns:     int - 0 == success, else error                              */
4897 /* Parameters:  data(IO) - ioctl data pointer to ipfunc_resolve_t struct    */
4898 /*                                                                          */
4899 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4900 /* This will either be the function name (if the pointer is set) or the     */
4901 /* function pointer if the name is set.  When found, fill in the other one  */
4902 /* so that the entire, complete, structure can be copied back to user space.*/
4903 /* ------------------------------------------------------------------------ */
4904 int fr_resolvefunc(data)
4905 void *data;
4906 {
4907         ipfunc_resolve_t res, *ft;
4908         int err;
4909 
4910         err = BCOPYIN(data, &res, sizeof(res));
4911         if (err != 0)
4912                 return EFAULT;
4913 
4914         if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4915                 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4916                         if (strncmp(res.ipfu_name, ft->ipfu_name,
4917                                     sizeof(res.ipfu_name)) == 0) {
4918                                 res.ipfu_addr = ft->ipfu_addr;
4919                                 res.ipfu_init = ft->ipfu_init;
4920                                 if (COPYOUT(&res, data, sizeof(res)) != 0)
4921                                         return EFAULT;
4922                                 return 0;
4923                         }
4924         }
4925         if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4926                 for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4927                         if (ft->ipfu_addr == res.ipfu_addr) {
4928                                 (void) strncpy(res.ipfu_name, ft->ipfu_name,
4929                                                sizeof(res.ipfu_name));
4930                                 res.ipfu_init = ft->ipfu_init;
4931                                 if (COPYOUT(&res, data, sizeof(res)) != 0)
4932                                         return EFAULT;
4933                                 return 0;
4934                         }
4935         }
4936         return ESRCH;
4937 }
4938 
4939 
4940 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4941     (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4942     (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4943     (defined(__OpenBSD__) && (OpenBSD < 200006))
4944 /*
4945  * From: NetBSD
4946  * ppsratecheck(): packets (or events) per second limitation.
4947  */
4948 int
4949 ppsratecheck(lasttime, curpps, maxpps)
4950         struct timeval *lasttime;
4951         int *curpps;
4952         int maxpps;     /* maximum pps allowed */
4953 {
4954         struct timeval tv, delta;
4955         int rv;
4956 
4957         GETKTIME(&tv);
4958 
4959         delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4960         delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4961         if (delta.tv_usec < 0) {
4962                 delta.tv_sec--;
4963                 delta.tv_usec += 1000000;
4964         }
4965 
4966         /*
4967          * check for 0,0 is so that the message will be seen at least once.
4968          * if more than one second have passed since the last update of
4969          * lasttime, reset the counter.
4970          *
4971          * we do increment *curpps even in *curpps < maxpps case, as some may
4972          * try to use *curpps for stat purposes as well.
4973          */
4974         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4975             delta.tv_sec >= 1) {
4976                 *lasttime = tv;
4977                 *curpps = 0;
4978                 rv = 1;
4979         } else if (maxpps < 0)
4980                 rv = 1;
4981         else if (*curpps < maxpps)
4982                 rv = 1;
4983         else
4984                 rv = 0;
4985         *curpps = *curpps + 1;
4986 
4987         return (rv);
4988 }
4989 #endif
4990 
4991 
4992 /* ------------------------------------------------------------------------ */
4993 /* Function:    fr_derefrule                                                */
4994 /* Returns:     int   - 0 == rule freed up, else rule not freed             */
4995 /* Parameters:  fr(I) - pointer to filter rule                              */
4996 /*                                                                          */
4997 /* Decrement the reference counter to a rule by one.  If it reaches zero,   */
4998 /* free it and any associated storage space being used by it.               */
4999 /* ------------------------------------------------------------------------ */
5000 int fr_derefrule(frp, ifs)
5001 frentry_t **frp;
5002 ipf_stack_t *ifs;
5003 {
5004         frentry_t *fr;
5005 
5006         fr = *frp;
5007 
5008         MUTEX_ENTER(&fr->fr_lock);
5009         fr->fr_ref--;
5010         if (fr->fr_ref == 0) {
5011                 MUTEX_EXIT(&fr->fr_lock);
5012                 MUTEX_DESTROY(&fr->fr_lock);
5013 
5014 #ifdef IPFILTER_LOOKUP
5015                 if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
5016                     ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
5017                 if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
5018                     ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
5019 #endif
5020 
5021                 if (fr->fr_dsize) {
5022                         KFREES(fr->fr_data, fr->fr_dsize);
5023                 }
5024                 if ((fr->fr_flags & FR_COPIED) != 0) {
5025                         KFREE(fr);
5026                         return 0;
5027                 }
5028                 return 1;
5029         } else {
5030                 MUTEX_EXIT(&fr->fr_lock);
5031         }
5032         *frp = NULL;
5033         return -1;
5034 }
5035 
5036 
5037 #ifdef  IPFILTER_LOOKUP
5038 /* ------------------------------------------------------------------------ */
5039 /* Function:    fr_grpmapinit                                               */
5040 /* Returns:     int - 0 == success, else ESRCH because table entry not found*/
5041 /* Parameters:  fr(I) - pointer to rule to find hash table for              */
5042 /*                                                                          */
5043 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr.  */
5044 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap.                   */
5045 /* ------------------------------------------------------------------------ */
5046 static int fr_grpmapinit(fr, ifs)
5047 frentry_t *fr;
5048 ipf_stack_t *ifs;
5049 {
5050         char name[FR_GROUPLEN];
5051         iphtable_t *iph;
5052 
5053 #if defined(SNPRINTF) && defined(_KERNEL)
5054         (void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
5055 #else
5056         (void) sprintf(name, "%d", fr->fr_arg);
5057 #endif
5058         iph = fr_findhtable(IPL_LOGIPF, name, ifs);
5059         if (iph == NULL)
5060                 return ESRCH;
5061         if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
5062                 return ESRCH;
5063         fr->fr_ptr = iph;
5064         return 0;
5065 }
5066 
5067 
5068 /* ------------------------------------------------------------------------ */
5069 /* Function:    fr_srcgrpmap                                                */
5070 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5071 /* Parameters:  fin(I)    - pointer to packet information                   */
5072 /*              passp(IO) - pointer to current/new filter decision (unused) */
5073 /*                                                                          */
5074 /* Look for a rule group head in a hash table, using the source address as  */
5075 /* the key, and descend into that group and continue matching rules against */
5076 /* the packet.                                                              */
5077 /* ------------------------------------------------------------------------ */
5078 frentry_t *fr_srcgrpmap(fin, passp)
5079 fr_info_t *fin;
5080 u_32_t *passp;
5081 {
5082         frgroup_t *fg;
5083         void *rval;
5084         ipf_stack_t *ifs = fin->fin_ifs;
5085 
5086         rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
5087         if (rval == NULL)
5088                 return NULL;
5089 
5090         fg = rval;
5091         fin->fin_fr = fg->fg_start;
5092         (void) fr_scanlist(fin, *passp);
5093         return fin->fin_fr;
5094 }
5095 
5096 
5097 /* ------------------------------------------------------------------------ */
5098 /* Function:    fr_dstgrpmap                                                */
5099 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5100 /* Parameters:  fin(I)    - pointer to packet information                   */
5101 /*              passp(IO) - pointer to current/new filter decision (unused) */
5102 /*                                                                          */
5103 /* Look for a rule group head in a hash table, using the destination        */
5104 /* address as the key, and descend into that group and continue matching    */
5105 /* rules against  the packet.                                               */
5106 /* ------------------------------------------------------------------------ */
5107 frentry_t *fr_dstgrpmap(fin, passp)
5108 fr_info_t *fin;
5109 u_32_t *passp;
5110 {
5111         frgroup_t *fg;
5112         void *rval;
5113         ipf_stack_t *ifs = fin->fin_ifs;
5114 
5115         rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
5116         if (rval == NULL)
5117                 return NULL;
5118 
5119         fg = rval;
5120         fin->fin_fr = fg->fg_start;
5121         (void) fr_scanlist(fin, *passp);
5122         return fin->fin_fr;
5123 }
5124 #endif /* IPFILTER_LOOKUP */
5125 
5126 /*
5127  * Queue functions
5128  * ===============
5129  * These functions manage objects on queues for efficient timeouts.  There are
5130  * a number of system defined queues as well as user defined timeouts.  It is
5131  * expected that a lock is held in the domain in which the queue belongs
5132  * (i.e. either state or NAT) when calling any of these functions that prevents
5133  * fr_freetimeoutqueue() from being called at the same time as any other.
5134  */
5135 
5136 
5137 /* ------------------------------------------------------------------------ */
5138 /* Function:    fr_addtimeoutqueue                                          */
5139 /* Returns:     struct ifqtq * - NULL if malloc fails, else pointer to      */
5140 /*                               timeout queue with given interval.         */
5141 /* Parameters:  parent(I)  - pointer to pointer to parent node of this list */
5142 /*                           of interface queues.                           */
5143 /*              seconds(I) - timeout value in seconds for this queue.       */
5144 /*                                                                          */
5145 /* This routine first looks for a timeout queue that matches the interval   */
5146 /* being requested.  If it finds one, increments the reference counter and  */
5147 /* returns a pointer to it.  If none are found, it allocates a new one and  */
5148 /* inserts it at the top of the list.                                       */
5149 /*                                                                          */
5150 /* Locking.                                                                 */
5151 /* It is assumed that the caller of this function has an appropriate lock   */
5152 /* held (exclusively) in the domain that encompases 'parent'.               */
5153 /* ------------------------------------------------------------------------ */
5154 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
5155 ipftq_t **parent;
5156 u_int seconds;
5157 ipf_stack_t *ifs;
5158 {
5159         ipftq_t *ifq;
5160         u_int period;
5161 
5162         period = seconds * IPF_HZ_DIVIDE;
5163 
5164         MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
5165         for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
5166                 if (ifq->ifq_ttl == period) {
5167                         /*
5168                          * Reset the delete flag, if set, so the structure
5169                          * gets reused rather than freed and reallocated.
5170                          */
5171                         MUTEX_ENTER(&ifq->ifq_lock);
5172                         ifq->ifq_flags &= ~IFQF_DELETE;
5173                         ifq->ifq_ref++;
5174                         MUTEX_EXIT(&ifq->ifq_lock);
5175                         MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5176 
5177                         return ifq;
5178                 }
5179         }
5180 
5181         KMALLOC(ifq, ipftq_t *);
5182         if (ifq != NULL) {
5183                 ifq->ifq_ttl = period;
5184                 ifq->ifq_head = NULL;
5185                 ifq->ifq_tail = &ifq->ifq_head;
5186                 ifq->ifq_next = *parent;
5187                 ifq->ifq_pnext = parent;
5188                 ifq->ifq_ref = 1;
5189                 ifq->ifq_flags = IFQF_USER;
5190                 *parent = ifq;
5191                 ifs->ifs_fr_userifqs++;
5192                 MUTEX_NUKE(&ifq->ifq_lock);
5193                 MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5194         }
5195         MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5196         return ifq;
5197 }
5198 
5199 
5200 /* ------------------------------------------------------------------------ */
5201 /* Function:    fr_deletetimeoutqueue                                       */
5202 /* Returns:     int    - new reference count value of the timeout queue     */
5203 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5204 /* Locks:       ifq->ifq_lock                                               */
5205 /*                                                                          */
5206 /* This routine must be called when we're discarding a pointer to a timeout */
5207 /* queue object, taking care of the reference counter.                      */
5208 /*                                                                          */
5209 /* Now that this just sets a DELETE flag, it requires the expire code to    */
5210 /* check the list of user defined timeout queues and call the free function */
5211 /* below (currently commented out) to stop memory leaking.  It is done this */
5212 /* way because the locking may not be sufficient to safely do a free when   */
5213 /* this function is called.                                                 */
5214 /* ------------------------------------------------------------------------ */
5215 int fr_deletetimeoutqueue(ifq)
5216 ipftq_t *ifq;
5217 {
5218 
5219         ifq->ifq_ref--;
5220         if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5221                 ifq->ifq_flags |= IFQF_DELETE;
5222         }
5223 
5224         return ifq->ifq_ref;
5225 }
5226 
5227 
5228 /* ------------------------------------------------------------------------ */
5229 /* Function:    fr_freetimeoutqueue                                         */
5230 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5231 /* Returns:     Nil                                                         */
5232 /*                                                                          */
5233 /* Locking:                                                                 */
5234 /* It is assumed that the caller of this function has an appropriate lock   */
5235 /* held (exclusively) in the domain that encompases the callers "domain".   */
5236 /* The ifq_lock for this structure should not be held.                      */
5237 /*                                                                          */
5238 /* Remove a user definde timeout queue from the list of queues it is in and */
5239 /* tidy up after this is done.                                              */
5240 /* ------------------------------------------------------------------------ */
5241 void fr_freetimeoutqueue(ifq, ifs)
5242 ipftq_t *ifq;
5243 ipf_stack_t *ifs;
5244 {
5245 
5246 
5247         if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5248             ((ifq->ifq_flags & IFQF_USER) == 0)) {
5249                 printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5250                        (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5251                        ifq->ifq_ref);
5252                 return;
5253         }
5254 
5255         /*
5256          * Remove from its position in the list.
5257          */
5258         *ifq->ifq_pnext = ifq->ifq_next;
5259         if (ifq->ifq_next != NULL)
5260                 ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5261 
5262         MUTEX_DESTROY(&ifq->ifq_lock);
5263         ifs->ifs_fr_userifqs--;
5264         KFREE(ifq);
5265 }
5266 
5267 
5268 /* ------------------------------------------------------------------------ */
5269 /* Function:    fr_deletequeueentry                                         */
5270 /* Returns:     Nil                                                         */
5271 /* Parameters:  tqe(I) - timeout queue entry to delete                      */
5272 /*              ifq(I) - timeout queue to remove entry from                 */
5273 /*                                                                          */
5274 /* Remove a tail queue entry from its queue and make it an orphan.          */
5275 /* fr_deletetimeoutqueue is called to make sure the reference count on the  */
5276 /* queue is correct.  We can't, however, call fr_freetimeoutqueue because   */
5277 /* the correct lock(s) may not be held that would make it safe to do so.    */
5278 /* ------------------------------------------------------------------------ */
5279 void fr_deletequeueentry(tqe)
5280 ipftqent_t *tqe;
5281 {
5282         ipftq_t *ifq;
5283 
5284         ifq = tqe->tqe_ifq;
5285         if (ifq == NULL)
5286                 return;
5287 
5288         MUTEX_ENTER(&ifq->ifq_lock);
5289 
5290         if (tqe->tqe_pnext != NULL) {
5291                 *tqe->tqe_pnext = tqe->tqe_next;
5292                 if (tqe->tqe_next != NULL)
5293                         tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5294                 else    /* we must be the tail anyway */
5295                         ifq->ifq_tail = tqe->tqe_pnext;
5296 
5297                 tqe->tqe_pnext = NULL;
5298                 tqe->tqe_ifq = NULL;
5299         }
5300 
5301         (void) fr_deletetimeoutqueue(ifq);
5302 
5303         MUTEX_EXIT(&ifq->ifq_lock);
5304 }
5305 
5306 
5307 /* ------------------------------------------------------------------------ */
5308 /* Function:    fr_queuefront                                               */
5309 /* Returns:     Nil                                                         */
5310 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5311 /*                                                                          */
5312 /* Move a queue entry to the front of the queue, if it isn't already there. */
5313 /* ------------------------------------------------------------------------ */
5314 void fr_queuefront(tqe)
5315 ipftqent_t *tqe;
5316 {
5317         ipftq_t *ifq;
5318 
5319         ifq = tqe->tqe_ifq;
5320         if (ifq == NULL)
5321                 return;
5322 
5323         MUTEX_ENTER(&ifq->ifq_lock);
5324         if (ifq->ifq_head != tqe) {
5325                 *tqe->tqe_pnext = tqe->tqe_next;
5326                 if (tqe->tqe_next)
5327                         tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5328                 else
5329                         ifq->ifq_tail = tqe->tqe_pnext;
5330 
5331                 tqe->tqe_next = ifq->ifq_head;
5332                 ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5333                 ifq->ifq_head = tqe;
5334                 tqe->tqe_pnext = &ifq->ifq_head;
5335         }
5336         MUTEX_EXIT(&ifq->ifq_lock);
5337 }
5338 
5339 
5340 /* ------------------------------------------------------------------------ */
5341 /* Function:    fr_queueback                                                */
5342 /* Returns:     Nil                                                         */
5343 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5344 /*                                                                          */
5345 /* Move a queue entry to the back of the queue, if it isn't already there.  */
5346 /* ------------------------------------------------------------------------ */
5347 void fr_queueback(tqe, ifs)
5348 ipftqent_t *tqe;
5349 ipf_stack_t *ifs;
5350 {
5351         ipftq_t *ifq;
5352 
5353         ifq = tqe->tqe_ifq;
5354         if (ifq == NULL)
5355                 return;
5356         tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5357 
5358         MUTEX_ENTER(&ifq->ifq_lock);
5359         if (tqe->tqe_next == NULL) {         /* at the end already ? */
5360                 MUTEX_EXIT(&ifq->ifq_lock);
5361                 return;
5362         }
5363 
5364         /*
5365          * Remove from list
5366          */
5367         *tqe->tqe_pnext = tqe->tqe_next;
5368         tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5369 
5370         /*
5371          * Make it the last entry.
5372          */
5373         tqe->tqe_next = NULL;
5374         tqe->tqe_pnext = ifq->ifq_tail;
5375         *ifq->ifq_tail = tqe;
5376         ifq->ifq_tail = &tqe->tqe_next;
5377         MUTEX_EXIT(&ifq->ifq_lock);
5378 }
5379 
5380 
5381 /* ------------------------------------------------------------------------ */
5382 /* Function:    fr_queueappend                                              */
5383 /* Returns:     Nil                                                         */
5384 /* Parameters:  tqe(I)    - pointer to timeout queue entry                  */
5385 /*              ifq(I)    - pointer to timeout queue                        */
5386 /*              parent(I) - owing object pointer                            */
5387 /*                                                                          */
5388 /* Add a new item to this queue and put it on the very end.                 */
5389 /* ------------------------------------------------------------------------ */
5390 void fr_queueappend(tqe, ifq, parent, ifs)
5391 ipftqent_t *tqe;
5392 ipftq_t *ifq;
5393 void *parent;
5394 ipf_stack_t *ifs;
5395 {
5396 
5397         MUTEX_ENTER(&ifq->ifq_lock);
5398         tqe->tqe_parent = parent;
5399         tqe->tqe_pnext = ifq->ifq_tail;
5400         *ifq->ifq_tail = tqe;
5401         ifq->ifq_tail = &tqe->tqe_next;
5402         tqe->tqe_next = NULL;
5403         tqe->tqe_ifq = ifq;
5404         tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5405         ifq->ifq_ref++;
5406         MUTEX_EXIT(&ifq->ifq_lock);
5407 }
5408 
5409 
5410 /* ------------------------------------------------------------------------ */
5411 /* Function:    fr_movequeue                                                */
5412 /* Returns:     Nil                                                         */
5413 /* Parameters:  tq(I)   - pointer to timeout queue information              */
5414 /*              oifp(I) - old timeout queue entry was on                    */
5415 /*              nifp(I) - new timeout queue to put entry on                 */
5416 /*              ifs     - ipf stack instance                                */
5417 /*                                                                          */
5418 /* Move a queue entry from one timeout queue to another timeout queue.      */
5419 /* If it notices that the current entry is already last and does not need   */
5420 /* to move queue, the return.                                               */
5421 /* ------------------------------------------------------------------------ */
5422 void fr_movequeue(tqe, oifq, nifq, ifs)
5423 ipftqent_t *tqe;
5424 ipftq_t *oifq, *nifq;
5425 ipf_stack_t *ifs;
5426 {
5427         /*
5428          * If the queue isn't changing, and the clock hasn't ticked
5429          * since the last update, the operation will be a no-op.
5430          */
5431         if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5432                 return;
5433 
5434         /*
5435          * Grab the lock and update the timers.
5436          */
5437         MUTEX_ENTER(&oifq->ifq_lock);
5438         tqe->tqe_touched = ifs->ifs_fr_ticks;
5439         tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5440 
5441         /*
5442          * The remainder of the operation can still be a no-op.
5443          *
5444          * If the queue isn't changing, check to see if
5445          * an update would be meaningless.
5446          */
5447         if (oifq == nifq) {
5448                 if ((tqe->tqe_next == NULL) ||
5449                     (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5450                         MUTEX_EXIT(&oifq->ifq_lock);
5451                         return;
5452                 }
5453         }
5454 
5455         /*
5456          * Remove from the old queue
5457          */
5458         *tqe->tqe_pnext = tqe->tqe_next;
5459         if (tqe->tqe_next)
5460                 tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5461         else
5462                 oifq->ifq_tail = tqe->tqe_pnext;
5463         tqe->tqe_next = NULL;
5464 
5465         /*
5466          * If we're moving from one queue to another, release the lock on the
5467          * old queue and get a lock on the new queue.  For user defined queues,
5468          * if we're moving off it, call delete in case it can now be freed.
5469          */
5470         if (oifq != nifq) {
5471                 tqe->tqe_ifq = NULL;
5472 
5473                 (void) fr_deletetimeoutqueue(oifq);
5474 
5475                 MUTEX_EXIT(&oifq->ifq_lock);
5476 
5477                 MUTEX_ENTER(&nifq->ifq_lock);
5478 
5479                 tqe->tqe_ifq = nifq;
5480                 nifq->ifq_ref++;
5481         }
5482 
5483         /*
5484          * Add to the bottom of the new queue
5485          */
5486         tqe->tqe_pnext = nifq->ifq_tail;
5487         *nifq->ifq_tail = tqe;
5488         nifq->ifq_tail = &tqe->tqe_next;
5489         MUTEX_EXIT(&nifq->ifq_lock);
5490 }
5491 
5492 
5493 /* ------------------------------------------------------------------------ */
5494 /* Function:    fr_updateipid                                               */
5495 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
5496 /* Parameters:  fin(I) - pointer to packet information                      */
5497 /*                                                                          */
5498 /* When we are doing NAT, change the IP of every packet to represent a      */
5499 /* single sequence of packets coming from the host, hiding any host         */
5500 /* specific sequencing that might otherwise be revealed.  If the packet is  */
5501 /* a fragment, then store the 'new' IPid in the fragment cache and look up  */
5502 /* the fragment cache for non-leading fragments.  If a non-leading fragment */
5503 /* has no match in the cache, return an error.                              */
5504 /* ------------------------------------------------------------------------ */
5505 static INLINE int fr_updateipid(fin)
5506 fr_info_t *fin;
5507 {
5508         u_short id, ido, sums;
5509         u_32_t sumd, sum;
5510         ip_t *ip;
5511 
5512         if (fin->fin_off != 0) {
5513                 sum = fr_ipid_knownfrag(fin);
5514                 if (sum == 0xffffffff)
5515                         return -1;
5516                 sum &= 0xffff;
5517                 id = (u_short)sum;
5518         } else {
5519                 id = fr_nextipid(fin);
5520                 if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5521                         (void) fr_ipid_newfrag(fin, (u_32_t)id);
5522         }
5523 
5524         ip = fin->fin_ip;
5525         ido = ntohs(ip->ip_id);
5526         if (id == ido)
5527                 return 0;
5528         ip->ip_id = htons(id);
5529         CALC_SUMD(ido, id, sumd);       /* DESTRUCTIVE MACRO! id,ido change */
5530         sum = (~ntohs(ip->ip_sum)) & 0xffff;
5531         sum += sumd;
5532         sum = (sum >> 16) + (sum & 0xffff);
5533         sum = (sum >> 16) + (sum & 0xffff);
5534         sums = ~(u_short)sum;
5535         ip->ip_sum = htons(sums);
5536         return 0;
5537 }
5538 
5539 
5540 #ifdef  NEED_FRGETIFNAME
5541 /* ------------------------------------------------------------------------ */
5542 /* Function:    fr_getifname                                                */
5543 /* Returns:     char *    - pointer to interface name                       */
5544 /* Parameters:  ifp(I)    - pointer to network interface                    */
5545 /*              buffer(O) - pointer to where to store interface name        */
5546 /*                                                                          */
5547 /* Constructs an interface name in the buffer passed.  The buffer passed is */
5548 /* expected to be at least LIFNAMSIZ in bytes big.  If buffer is passed in  */
5549 /* as a NULL pointer then return a pointer to a static array.               */
5550 /* ------------------------------------------------------------------------ */
5551 char *fr_getifname(ifp, buffer)
5552 struct ifnet *ifp;
5553 char *buffer;
5554 {
5555         static char namebuf[LIFNAMSIZ];
5556 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5557      defined(__sgi) || defined(linux) || defined(_AIX51) || \
5558      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5559         int unit, space;
5560         char temp[20];
5561         char *s;
5562 # endif
5563 
5564         ASSERT(buffer != NULL);
5565 #ifdef notdef
5566         if (buffer == NULL)
5567                 buffer = namebuf;
5568 #endif
5569         (void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5570         buffer[LIFNAMSIZ - 1] = '\0';
5571 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5572      defined(__sgi) || defined(_AIX51) || \
5573      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5574         for (s = buffer; *s; s++)
5575                 ;
5576         unit = ifp->if_unit;
5577         space = LIFNAMSIZ - (s - buffer);
5578         if (space > 0) {
5579 #  if defined(SNPRINTF) && defined(_KERNEL)
5580                 (void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5581 #  else
5582                 (void) sprintf(temp, "%d", unit);
5583 #  endif
5584                 (void) strncpy(s, temp, space);
5585         }
5586 # endif
5587         return buffer;
5588 }
5589 #endif
5590 
5591 
5592 /* ------------------------------------------------------------------------ */
5593 /* Function:    fr_ioctlswitch                                              */
5594 /* Returns:     int     - -1 continue processing, else ioctl return value   */
5595 /* Parameters:  unit(I) - device unit opened                                */
5596 /*              data(I) - pointer to ioctl data                             */
5597 /*              cmd(I)  - ioctl command                                     */
5598 /*              mode(I) - mode value                                        */
5599 /*                                                                          */
5600 /* Based on the value of unit, call the appropriate ioctl handler or return */
5601 /* EIO if ipfilter is not running.   Also checks if write perms are req'd   */
5602 /* for the device in order to execute the ioctl.                            */
5603 /* ------------------------------------------------------------------------ */
5604 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5605 int unit, mode, uid;
5606 ioctlcmd_t cmd;
5607 void *data, *ctx;
5608 ipf_stack_t *ifs;
5609 {
5610         int error = 0;
5611 
5612         switch (unit)
5613         {
5614         case IPL_LOGIPF :
5615                 error = -1;
5616                 break;
5617         case IPL_LOGNAT :
5618                 if (ifs->ifs_fr_running > 0)
5619                         error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5620                 else
5621                         error = EIO;
5622                 break;
5623         case IPL_LOGSTATE :
5624                 if (ifs->ifs_fr_running > 0)
5625                         error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5626                 else
5627                         error = EIO;
5628                 break;
5629         case IPL_LOGAUTH :
5630                 if (ifs->ifs_fr_running > 0) {
5631                         if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5632                             (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5633                                 if (!(mode & FWRITE)) {
5634                                         error = EPERM;
5635                                 } else {
5636                                         error = frrequest(unit, cmd, data,
5637                                                   ifs->ifs_fr_active, 1, ifs);
5638                                 }
5639                         } else {
5640                                 error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5641                         }
5642                 } else
5643                         error = EIO;
5644                 break;
5645         case IPL_LOGSYNC :
5646 #ifdef IPFILTER_SYNC
5647                 if (ifs->ifs_fr_running > 0)
5648                         error = fr_sync_ioctl(data, cmd, mode, ifs);
5649                 else
5650 #endif
5651                         error = EIO;
5652                 break;
5653         case IPL_LOGSCAN :
5654 #ifdef IPFILTER_SCAN
5655                 if (ifs->ifs_fr_running > 0)
5656                         error = fr_scan_ioctl(data, cmd, mode, ifs);
5657                 else
5658 #endif
5659                         error = EIO;
5660                 break;
5661         case IPL_LOGLOOKUP :
5662 #ifdef IPFILTER_LOOKUP
5663                 if (ifs->ifs_fr_running > 0)
5664                         error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5665                 else
5666 #endif
5667                         error = EIO;
5668                 break;
5669         default :
5670                 error = EIO;
5671                 break;
5672         }
5673 
5674         return error;
5675 }
5676 
5677 
5678 /*
5679  * This array defines the expected size of objects coming into the kernel
5680  * for the various recognised object types.
5681  */
5682 #define NUM_OBJ_TYPES   19
5683 
5684 static  int     fr_objbytes[NUM_OBJ_TYPES][2] = {
5685         { 1,    sizeof(struct frentry) },               /* frentry */
5686         { 0,    sizeof(struct friostat) },
5687         { 0,    sizeof(struct fr_info) },
5688         { 0,    sizeof(struct fr_authstat) },
5689         { 0,    sizeof(struct ipfrstat) },
5690         { 0,    sizeof(struct ipnat) },
5691         { 0,    sizeof(struct natstat) },
5692         { 0,    sizeof(struct ipstate_save) },
5693         { 1,    sizeof(struct nat_save) },              /* nat_save */
5694         { 0,    sizeof(struct natlookup) },
5695         { 1,    sizeof(struct ipstate) },               /* ipstate */
5696         { 0,    sizeof(struct ips_stat) },
5697         { 0,    sizeof(struct frauth) },
5698         { 0,    sizeof(struct ipftune) },
5699         { 0,    sizeof(struct nat) },                   /* nat_t */
5700         { 0,    sizeof(struct ipfruleiter) },
5701         { 0,    sizeof(struct ipfgeniter) },
5702         { 0,    sizeof(struct ipftable) },
5703         { 0,    sizeof(struct ipflookupiter) }
5704 };
5705 
5706 
5707 /* ------------------------------------------------------------------------ */
5708 /* Function:    fr_getzoneid                                                */
5709 /* Returns:     int     - 0 = success, else failure                         */
5710 /* Parameters:  idsp(I) - pointer to ipf_devstate_t                         */
5711 /*              data(I) - pointer to ioctl data                             */
5712 /*                                                                          */
5713 /* Set the zone ID in idsp based on the zone name in ipfzoneobj.  Further   */
5714 /* ioctls will act on the IPF stack for that zone ID.                       */
5715 /* ------------------------------------------------------------------------ */
5716 #if defined(_KERNEL)
5717 int fr_setzoneid(idsp, data)
5718 ipf_devstate_t *idsp;
5719 void *data;
5720 {
5721         int error = 0;
5722         ipfzoneobj_t ipfzo;
5723         zone_t *zone;
5724 
5725         error = BCOPYIN(data, &ipfzo, sizeof(ipfzo));
5726         if (error != 0)
5727                 return EFAULT;
5728 
5729         if (memchr(ipfzo.ipfz_zonename, '\0', ZONENAME_MAX) == NULL)
5730                 return EFAULT;
5731 
5732         /*
5733          * The global zone doesn't have a GZ-controlled stack, so no
5734          * sense in going any further
5735          */
5736         if (strcmp(ipfzo.ipfz_zonename, "global") == 0)
5737                 return ENODEV;
5738 
5739         if ((zone = zone_find_by_name(ipfzo.ipfz_zonename)) == NULL)
5740                 return ENODEV;
5741 
5742         /*
5743          * Store the zone ID that to control, and whether it's the
5744          * GZ-controlled stack that's wanted
5745          */
5746         idsp->ipfs_zoneid = zone->zone_id;
5747         idsp->ipfs_gz = (ipfzo.ipfz_gz == 1) ? B_TRUE : B_FALSE;
5748         zone_rele(zone);
5749 
5750         return error;
5751 }
5752 #endif
5753 
5754 
5755 /* ------------------------------------------------------------------------ */
5756 /* Function:    fr_inobj                                                    */
5757 /* Returns:     int     - 0 = success, else failure                         */
5758 /* Parameters:  data(I) - pointer to ioctl data                             */
5759 /*              ptr(I)  - pointer to store real data in                     */
5760 /*              type(I) - type of structure being moved                     */
5761 /*                                                                          */
5762 /* Copy in the contents of what the ipfobj_t points to.  In future, we      */
5763 /* add things to check for version numbers, sizes, etc, to make it backward */
5764 /* compatible at the ABI for user land.                                     */
5765 /* ------------------------------------------------------------------------ */
5766 int fr_inobj(data, ptr, type)
5767 void *data;
5768 void *ptr;
5769 int type;
5770 {
5771         ipfobj_t obj;
5772         int error = 0;
5773 
5774         if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5775                 return EINVAL;
5776 
5777         error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5778         if (error != 0)
5779                 return EFAULT;
5780 
5781         if (obj.ipfo_type != type)
5782                 return EINVAL;
5783 
5784 #ifndef IPFILTER_COMPAT
5785         if ((fr_objbytes[type][0] & 1) != 0) {
5786                 if (obj.ipfo_size < fr_objbytes[type][1])
5787                         return EINVAL;
5788         } else if (obj.ipfo_size != fr_objbytes[type][1])
5789                 return EINVAL;
5790 #else
5791         if (obj.ipfo_rev != IPFILTER_VERSION) {
5792                 error = fr_incomptrans(&obj, ptr);
5793                 return error;
5794         }
5795 
5796         if ((fr_objbytes[type][0] & 1) != 0 &&
5797             obj.ipfo_size < fr_objbytes[type][1] ||
5798             obj.ipfo_size != fr_objbytes[type][1])
5799                 return EINVAL;
5800 #endif
5801 
5802         if ((fr_objbytes[type][0] & 1) != 0) {
5803                 error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5804                                 fr_objbytes[type][1]);
5805         } else {
5806                 error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5807                                 obj.ipfo_size);
5808         }
5809         return error;
5810 }
5811 
5812 
5813 /* ------------------------------------------------------------------------ */
5814 /* Function:    fr_inobjsz                                                  */
5815 /* Returns:     int     - 0 = success, else failure                         */
5816 /* Parameters:  data(I) - pointer to ioctl data                             */
5817 /*              ptr(I)  - pointer to store real data in                     */
5818 /*              type(I) - type of structure being moved                     */
5819 /*              sz(I)   - size of data to copy                              */
5820 /*                                                                          */
5821 /* As per fr_inobj, except the size of the object to copy in is passed in   */
5822 /* but it must not be smaller than the size defined for the type and the    */
5823 /* type must allow for varied sized objects.  The extra requirement here is */
5824 /* that sz must match the size of the object being passed in - this is not  */
5825 /* not possible nor required in fr_inobj().                                 */
5826 /* ------------------------------------------------------------------------ */
5827 int fr_inobjsz(data, ptr, type, sz)
5828 void *data;
5829 void *ptr;
5830 int type, sz;
5831 {
5832         ipfobj_t obj;
5833         int error;
5834 
5835         if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5836                 return EINVAL;
5837         if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5838                 return EINVAL;
5839 
5840         error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5841         if (error != 0)
5842                 return EFAULT;
5843 
5844         if (obj.ipfo_type != type)
5845                 return EINVAL;
5846 
5847 #ifndef IPFILTER_COMPAT
5848         if (obj.ipfo_size != sz)
5849                 return EINVAL;
5850 #else
5851         if (obj.ipfo_rev != IPFILTER_VERSION)
5852                 /*XXX compatibility hook here */
5853                 /*EMPTY*/;
5854         if (obj.ipfo_size != sz)
5855                 /* XXX compatibility hook here */
5856                 return EINVAL;
5857 #endif
5858 
5859         error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5860         return error;
5861 }
5862 
5863 
5864 /* ------------------------------------------------------------------------ */
5865 /* Function:    fr_outobjsz                                                 */
5866 /* Returns:     int     - 0 = success, else failure                         */
5867 /* Parameters:  data(I) - pointer to ioctl data                             */
5868 /*              ptr(I)  - pointer to store real data in                     */
5869 /*              type(I) - type of structure being moved                     */
5870 /*              sz(I)   - size of data to copy                              */
5871 /*                                                                          */
5872 /* As per fr_outobj, except the size of the object to copy out is passed in */
5873 /* but it must not be smaller than the size defined for the type and the    */
5874 /* type must allow for varied sized objects.  The extra requirement here is */
5875 /* that sz must match the size of the object being passed in - this is not  */
5876 /* not possible nor required in fr_outobj().                                */
5877 /* ------------------------------------------------------------------------ */
5878 int fr_outobjsz(data, ptr, type, sz)
5879 void *data;
5880 void *ptr;
5881 int type, sz;
5882 {
5883         ipfobj_t obj;
5884         int error;
5885 
5886         if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5887             ((fr_objbytes[type][0] & 1) == 0) ||
5888             (sz < fr_objbytes[type][1]))
5889                 return EINVAL;
5890 
5891         error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5892         if (error != 0)
5893                 return EFAULT;
5894 
5895         if (obj.ipfo_type != type)
5896                 return EINVAL;
5897 
5898 #ifndef IPFILTER_COMPAT
5899         if (obj.ipfo_size != sz)
5900                 return EINVAL;
5901 #else
5902         if (obj.ipfo_rev != IPFILTER_VERSION)
5903                 /* XXX compatibility hook here */
5904                 /*EMPTY*/;
5905         if (obj.ipfo_size != sz)
5906                 /* XXX compatibility hook here */
5907                 return EINVAL;
5908 #endif
5909 
5910         error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5911         return error;
5912 }
5913 
5914 
5915 /* ------------------------------------------------------------------------ */
5916 /* Function:    fr_outobj                                                   */
5917 /* Returns:     int     - 0 = success, else failure                         */
5918 /* Parameters:  data(I) - pointer to ioctl data                             */
5919 /*              ptr(I)  - pointer to store real data in                     */
5920 /*              type(I) - type of structure being moved                     */
5921 /*                                                                          */
5922 /* Copy out the contents of what ptr is to where ipfobj points to.  In      */
5923 /* future, we add things to check for version numbers, sizes, etc, to make  */
5924 /* it backward  compatible at the ABI for user land.                        */
5925 /* ------------------------------------------------------------------------ */
5926 int fr_outobj(data, ptr, type)
5927 void *data;
5928 void *ptr;
5929 int type;
5930 {
5931         ipfobj_t obj;
5932         int error;
5933 
5934         if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5935                 return EINVAL;
5936 
5937         error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5938         if (error != 0)
5939                 return EFAULT;
5940 
5941         if (obj.ipfo_type != type)
5942                 return EINVAL;
5943 
5944 #ifndef IPFILTER_COMPAT
5945         if ((fr_objbytes[type][0] & 1) != 0) {
5946                 if (obj.ipfo_size < fr_objbytes[type][1])
5947                         return EINVAL;
5948         } else if (obj.ipfo_size != fr_objbytes[type][1])
5949                 return EINVAL;
5950 #else
5951         if (obj.ipfo_rev != IPFILTER_VERSION) {
5952                 error = fr_outcomptrans(&obj, ptr);
5953                 return error;
5954         }
5955 
5956         if ((fr_objbytes[type][0] & 1) != 0 &&
5957             obj.ipfo_size < fr_objbytes[type][1] ||
5958             obj.ipfo_size != fr_objbytes[type][1])
5959                 return EINVAL;
5960 #endif
5961 
5962         error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5963         return error;
5964 }
5965 
5966 
5967 /* ------------------------------------------------------------------------ */
5968 /* Function:    fr_checkl4sum                                               */
5969 /* Returns:     int     - 0 = good, -1 = bad, 1 = cannot check              */
5970 /* Parameters:  fin(I) - pointer to packet information                      */
5971 /*                                                                          */
5972 /* If possible, calculate the layer 4 checksum for the packet.  If this is  */
5973 /* not possible, return without indicating a failure or success but in a    */
5974 /* way that is ditinguishable.                                              */
5975 /* ------------------------------------------------------------------------ */
5976 int fr_checkl4sum(fin)
5977 fr_info_t *fin;
5978 {
5979         u_short sum, hdrsum, *csump;
5980         udphdr_t *udp;
5981         int dosum;
5982         ipf_stack_t *ifs = fin->fin_ifs;
5983 
5984 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
5985         net_handle_t net_data_p;
5986         if (fin->fin_v == 4)
5987                 net_data_p = ifs->ifs_ipf_ipv4;
5988         else
5989                 net_data_p = ifs->ifs_ipf_ipv6;
5990 #endif
5991 
5992         if ((fin->fin_flx & FI_NOCKSUM) != 0)
5993                 return 0;
5994 
5995         /*
5996          * If the TCP packet isn't a fragment, isn't too short and otherwise
5997          * isn't already considered "bad", then validate the checksum.  If
5998          * this check fails then considered the packet to be "bad".
5999          */
6000         if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
6001                 return 1;
6002 
6003         csump = NULL;
6004         hdrsum = 0;
6005         dosum = 0;
6006         sum = 0;
6007 
6008 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
6009         ASSERT(fin->fin_m != NULL);
6010         if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
6011             NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
6012                         hdrsum = 0;
6013                         sum = 0;
6014         } else {
6015 #endif
6016                 switch (fin->fin_p)
6017                 {
6018                 case IPPROTO_TCP :
6019                         csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
6020                         dosum = 1;
6021                         break;
6022 
6023                 case IPPROTO_UDP :
6024                         udp = fin->fin_dp;
6025                         if (udp->uh_sum != 0) {
6026                                 csump = &udp->uh_sum;
6027                                 dosum = 1;
6028                         }
6029                         break;
6030 
6031                 case IPPROTO_ICMP :
6032                         csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
6033                         dosum = 1;
6034                         break;
6035 
6036                 default :
6037                         return 1;
6038                         /*NOTREACHED*/
6039                 }
6040 
6041                 if (csump != NULL)
6042                         hdrsum = *csump;
6043 
6044                 if (dosum)
6045                         sum = fr_cksum(fin->fin_m, fin->fin_ip,
6046                                        fin->fin_p, fin->fin_dp);
6047 #if defined(SOLARIS) && defined(_KERNEL) && (SOLARIS2 >= 6)
6048         }
6049 #endif
6050 #if !defined(_KERNEL)
6051         if (sum == hdrsum) {
6052                 FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
6053         } else {
6054                 FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
6055         }
6056 #endif
6057         if (hdrsum == sum)
6058                 return 0;
6059         return -1;
6060 }
6061 
6062 
6063 /* ------------------------------------------------------------------------ */
6064 /* Function:    fr_ifpfillv4addr                                            */
6065 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6066 /* Parameters:  atype(I)   - type of network address update to perform      */
6067 /*              sin(I)     - pointer to source of address information       */
6068 /*              mask(I)    - pointer to source of netmask information       */
6069 /*              inp(I)     - pointer to destination address store           */
6070 /*              inpmask(I) - pointer to destination netmask store           */
6071 /*                                                                          */
6072 /* Given a type of network address update (atype) to perform, copy          */
6073 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6074 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6075 /* which case the operation fails.  For all values of atype other than      */
6076 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6077 /* value.                                                                   */
6078 /* ------------------------------------------------------------------------ */
6079 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
6080 int atype;
6081 struct sockaddr_in *sin, *mask;
6082 struct in_addr *inp, *inpmask;
6083 {
6084         if (inpmask != NULL && atype != FRI_NETMASKED)
6085                 inpmask->s_addr = 0xffffffff;
6086 
6087         if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6088                 if (atype == FRI_NETMASKED) {
6089                         if (inpmask == NULL)
6090                                 return -1;
6091                         inpmask->s_addr = mask->sin_addr.s_addr;
6092                 }
6093                 inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
6094         } else {
6095                 inp->s_addr = sin->sin_addr.s_addr;
6096         }
6097         return 0;
6098 }
6099 
6100 
6101 #ifdef  USE_INET6
6102 /* ------------------------------------------------------------------------ */
6103 /* Function:    fr_ifpfillv6addr                                            */
6104 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6105 /* Parameters:  atype(I)   - type of network address update to perform      */
6106 /*              sin(I)     - pointer to source of address information       */
6107 /*              mask(I)    - pointer to source of netmask information       */
6108 /*              inp(I)     - pointer to destination address store           */
6109 /*              inpmask(I) - pointer to destination netmask store           */
6110 /*                                                                          */
6111 /* Given a type of network address update (atype) to perform, copy          */
6112 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6113 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6114 /* which case the operation fails.  For all values of atype other than      */
6115 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6116 /* value.                                                                   */
6117 /* ------------------------------------------------------------------------ */
6118 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
6119 int atype;
6120 struct sockaddr_in6 *sin, *mask;
6121 struct in_addr *inp, *inpmask;
6122 {
6123         i6addr_t *src, *dst, *and, *dmask;
6124 
6125         src = (i6addr_t *)&sin->sin6_addr;
6126         and = (i6addr_t *)&mask->sin6_addr;
6127         dst = (i6addr_t *)inp;
6128         dmask = (i6addr_t *)inpmask;
6129 
6130         if (inpmask != NULL && atype != FRI_NETMASKED) {
6131                 dmask->i6[0] = 0xffffffff;
6132                 dmask->i6[1] = 0xffffffff;
6133                 dmask->i6[2] = 0xffffffff;
6134                 dmask->i6[3] = 0xffffffff;
6135         }
6136 
6137         if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6138                 if (atype == FRI_NETMASKED) {
6139                         if (inpmask == NULL)
6140                                 return -1;
6141                         dmask->i6[0] = and->i6[0];
6142                         dmask->i6[1] = and->i6[1];
6143                         dmask->i6[2] = and->i6[2];
6144                         dmask->i6[3] = and->i6[3];
6145                 }
6146 
6147                 dst->i6[0] = src->i6[0] & and->i6[0];
6148                 dst->i6[1] = src->i6[1] & and->i6[1];
6149                 dst->i6[2] = src->i6[2] & and->i6[2];
6150                 dst->i6[3] = src->i6[3] & and->i6[3];
6151         } else {
6152                 dst->i6[0] = src->i6[0];
6153                 dst->i6[1] = src->i6[1];
6154                 dst->i6[2] = src->i6[2];
6155                 dst->i6[3] = src->i6[3];
6156         }
6157         return 0;
6158 }
6159 #endif
6160 
6161 
6162 /* ------------------------------------------------------------------------ */
6163 /* Function:    fr_matchtag                                                 */
6164 /* Returns:     0 == mismatch, 1 == match.                                  */
6165 /* Parameters:  tag1(I) - pointer to first tag to compare                   */
6166 /*              tag2(I) - pointer to second tag to compare                  */
6167 /*                                                                          */
6168 /* Returns true (non-zero) or false(0) if the two tag structures can be     */
6169 /* considered to be a match or not match, respectively.  The tag is 16      */
6170 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so    */
6171 /* compare the ints instead, for speed. tag1 is the master of the           */
6172 /* comparison.  This function should only be called with both tag1 and tag2 */
6173 /* as non-NULL pointers.                                                    */
6174 /* ------------------------------------------------------------------------ */
6175 int fr_matchtag(tag1, tag2)
6176 ipftag_t *tag1, *tag2;
6177 {
6178         if (tag1 == tag2)
6179                 return 1;
6180 
6181         if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
6182                 return 1;
6183 
6184         if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
6185             (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
6186             (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
6187             (tag1->ipt_num[3] == tag2->ipt_num[3]))
6188                 return 1;
6189         return 0;
6190 }
6191 
6192 
6193 /* ------------------------------------------------------------------------ */
6194 /* Function:    fr_coalesce                                                 */
6195 /* Returns:     1 == success, -1 == failure, 0 == no change                 */
6196 /* Parameters:  fin(I) - pointer to packet information                      */
6197 /*                                                                          */
6198 /* Attempt to get all of the packet data into a single, contiguous buffer.  */
6199 /* If this call returns a failure then the buffers have also been freed.    */
6200 /* ------------------------------------------------------------------------ */
6201 int fr_coalesce(fin)
6202 fr_info_t *fin;
6203 {
6204         ipf_stack_t *ifs = fin->fin_ifs;
6205         if ((fin->fin_flx & FI_COALESCE) != 0)
6206                 return 1;
6207 
6208         /*
6209          * If the mbuf pointers indicate that there is no mbuf to work with,
6210          * return but do not indicate success or failure.
6211          */
6212         if (fin->fin_m == NULL || fin->fin_mp == NULL)
6213                 return 0;
6214 
6215 #if defined(_KERNEL)
6216         if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
6217                 IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
6218 # ifdef MENTAT
6219                 FREE_MB_T(*fin->fin_mp);
6220 # endif
6221                 *fin->fin_mp = NULL;
6222                 fin->fin_m = NULL;
6223                 return -1;
6224         }
6225 #else
6226         fin = fin;      /* LINT */
6227 #endif
6228         return 1;
6229 }
6230 
6231 
6232 /*
6233  * The following table lists all of the tunable variables that can be
6234  * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT.  The format of each row
6235  * in the table below is as follows:
6236  *
6237  * pointer to value, name of value, minimum, maximum, size of the value's
6238  *     container, value attribute flags
6239  *
6240  * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6241  * means the value can only be written to when IPFilter is loaded but disabled.
6242  * The obvious implication is if neither of these are set then the value can be
6243  * changed at any time without harm.
6244  */
6245 ipftuneable_t lcl_ipf_tuneables[] = {
6246         /* filtering */
6247         { { NULL },     "fr_flags",             0,      0xffffffff,
6248                         0,              0 },
6249         { { NULL },     "fr_active",            0,      0,
6250                         0,              IPFT_RDONLY },
6251         { { NULL },     "fr_control_forwarding",        0, 1,
6252                         0,              0 },
6253         { { NULL },     "fr_update_ipid",       0,      1,
6254                         0,              0 },
6255         { { NULL },     "fr_chksrc",            0,      1,
6256                         0,              0 },
6257         { { NULL },     "fr_minttl",            0,      1,
6258                         0,              0 },
6259         { { NULL },     "fr_icmpminfragmtu",    0,      1,
6260                         0,              0 },
6261         { { NULL },             "fr_pass",              0,      0xffffffff,
6262                         0,              0 },
6263 #if SOLARIS2 >= 10
6264         { { NULL },     "ipf_loopback",         0,      1,
6265                         0,              IPFT_WRDISABLED },
6266 #endif
6267         /* state */
6268         { { NULL }, "fr_tcpidletimeout",        1,      0x7fffffff,
6269                         0,      IPFT_WRDISABLED },
6270         { { NULL },     "fr_tcpclosewait",      1,      0x7fffffff,
6271                         0,      IPFT_WRDISABLED },
6272         { { NULL },     "fr_tcplastack",        1,      0x7fffffff,
6273                         0,              IPFT_WRDISABLED },
6274         { { NULL },     "fr_tcptimeout",        1,      0x7fffffff,
6275                         0,              IPFT_WRDISABLED },
6276         { { NULL },     "fr_tcpclosed",         1,      0x7fffffff,
6277                         0,              IPFT_WRDISABLED },
6278         { { NULL },     "fr_tcphalfclosed",     1,      0x7fffffff,
6279                         0,      IPFT_WRDISABLED },
6280         { { NULL },     "fr_udptimeout",        1,      0x7fffffff,
6281                         0,              IPFT_WRDISABLED },
6282         { { NULL }, "fr_udpacktimeout", 1,      0x7fffffff,
6283                         0,      IPFT_WRDISABLED },
6284         { { NULL },     "fr_icmptimeout",       1,      0x7fffffff,
6285                         0,              IPFT_WRDISABLED },
6286         { { NULL }, "fr_icmpacktimeout",        1,      0x7fffffff,
6287                         0,      IPFT_WRDISABLED },
6288         { { NULL }, "fr_iptimeout",             1,      0x7fffffff,
6289                         0,              IPFT_WRDISABLED },
6290         { { NULL },     "fr_statemax",          1,      0x7fffffff,
6291                         0,              0 },
6292         { { NULL },     "fr_statesize",         1,      0x7fffffff,
6293                         0,              IPFT_WRDISABLED },
6294         { { NULL },     "fr_state_lock",        0,      1,
6295                         0,              IPFT_RDONLY },
6296         { { NULL }, "fr_state_maxbucket", 1,    0x7fffffff,
6297                         0,      IPFT_WRDISABLED },
6298         { { NULL }, "fr_state_maxbucket_reset", 0, 1,
6299                         0, IPFT_WRDISABLED },
6300         { { NULL },     "ipstate_logging",      0,      1,
6301                         0,      0 },
6302         { { NULL },     "state_flush_level_hi", 1,      100,
6303                         0,              0 },
6304         { { NULL },     "state_flush_level_lo", 1,      100,
6305                         0,              0 },
6306         /* nat */
6307         { { NULL },             "fr_nat_lock",          0,      1,
6308                         0,              IPFT_RDONLY },
6309         { { NULL },     "ipf_nattable_sz",      1,      0x7fffffff,
6310                         0,      IPFT_WRDISABLED },
6311         { { NULL }, "ipf_nattable_max", 1,      0x7fffffff,
6312                         0,      0 },
6313         { { NULL },     "ipf_natrules_sz",      1,      0x7fffffff,
6314                         0,      IPFT_WRDISABLED },
6315         { { NULL },     "ipf_rdrrules_sz",      1,      0x7fffffff,
6316                         0,      IPFT_WRDISABLED },
6317         { { NULL },     "ipf_hostmap_sz",       1,      0x7fffffff,
6318                         0,              IPFT_WRDISABLED },
6319         { { NULL }, "fr_nat_maxbucket", 1,      0x7fffffff,
6320                         0,      IPFT_WRDISABLED },
6321         { { NULL },     "fr_nat_maxbucket_reset",       0, 1,
6322                         0,      IPFT_WRDISABLED },
6323         { { NULL },             "nat_logging",          0,      1,
6324                         0,              0 },
6325         { { NULL },     "fr_defnatage",         1,      0x7fffffff,
6326                         0,              IPFT_WRDISABLED },
6327         { { NULL },     "fr_defnatipage",       1,      0x7fffffff,
6328                         0,              IPFT_WRDISABLED },
6329         { { NULL }, "fr_defnaticmpage", 1,      0x7fffffff,
6330                         0,      IPFT_WRDISABLED },
6331         { { NULL },     "nat_flush_level_hi",   1,      100,
6332                         0,              0 },
6333         { { NULL },     "nat_flush_level_lo",   1,      100,
6334                         0,              0 },
6335         /* frag */
6336         { { NULL },     "ipfr_size",            1,      0x7fffffff,
6337                         0,              IPFT_WRDISABLED },
6338         { { NULL },     "fr_ipfrttl",           1,      0x7fffffff,
6339                         0,              IPFT_WRDISABLED },
6340 #ifdef IPFILTER_LOG
6341         /* log */
6342         { { NULL },     "ipl_suppress",         0,      1,
6343                         0,              0 },
6344         { { NULL },     "ipl_buffer_sz",        0,      0,
6345                         0,              IPFT_RDONLY },
6346         { { NULL },     "ipl_logmax",           0,      0x7fffffff,
6347                         0,              IPFT_WRDISABLED },
6348         { { NULL },     "ipl_logall",           0,      1,
6349                         0,              0 },
6350         { { NULL },     "ipl_logsize",          0,      0x80000,
6351                         0,              0 },
6352 #endif
6353         { { NULL },             NULL,                   0,      0 }
6354 };
6355 
6356 static ipftuneable_t *
6357 tune_lookup(ipf_stack_t *ifs, char *name)
6358 {
6359     int i;
6360 
6361     for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6362         if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6363             return (&ifs->ifs_ipf_tuneables[i]);
6364     }
6365     return (NULL);
6366 }
6367 
6368 #ifdef _KERNEL
6369 extern dev_info_t *ipf_dev_info;
6370 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6371 #endif
6372 
6373 /* -------------------------------------------------------------------- */
6374 /* Function:    ipftuneable_setdefs()                                   */
6375 /* Returns:             void                                            */
6376 /* Parameters:  ifs - pointer to newly allocated IPF instance           */
6377 /*                              assigned to     IP instance             */
6378 /*                                                                      */
6379 /* Function initializes IPF instance variables. Function is invoked     */
6380 /* from ipftuneable_alloc(). ipftuneable_alloc() is called only one     */
6381 /* time during IP instance lifetime - at the time of IP instance        */
6382 /* creation. Anytime IP instance is being created new private IPF       */
6383 /* instance is allocated and assigned to it. The moment of IP           */
6384 /* instance creation is the right time to initialize those IPF          */
6385 /* variables.                                                           */
6386 /*                                                                      */
6387 /* -------------------------------------------------------------------- */
6388 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6389 {
6390         ifs->ifs_ipfr_size = IPFT_SIZE;
6391         ifs->ifs_fr_ipfrttl = 120;   /* 60 seconds */
6392 
6393         /* it comes from fr_authinit() in IPF auth */
6394         ifs->ifs_fr_authsize = FR_NUMAUTH;
6395         ifs->ifs_fr_defaultauthage = 600;
6396 
6397         /* it comes from fr_stateinit() in IPF state */
6398         ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5);      /* five days */
6399         ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6400         ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6401         ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6402         ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6403         ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600);    /* 2 hours */
6404         ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6405         ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6406         ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6407         ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6408         ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6409         ifs->ifs_fr_statemax = IPSTATE_MAX;
6410         ifs->ifs_fr_statesize = IPSTATE_SIZE;
6411         ifs->ifs_fr_state_maxbucket_reset = 1;
6412         ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6413         ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6414 
6415         /* it comes from fr_natinit() in ipnat */
6416         ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6417         ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6418         ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6419         ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6420         ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6421         ifs->ifs_fr_nat_maxbucket_reset = 1;
6422         ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6423         ifs->ifs_fr_defnatipage = 120;               /* 60 seconds */
6424         ifs->ifs_fr_defnaticmpage = 6;               /* 3 seconds */
6425         ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6426         ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6427 
6428 #ifdef IPFILTER_LOG
6429         /* it comes from fr_loginit() in IPF log */
6430         ifs->ifs_ipl_suppress = 1;
6431         ifs->ifs_ipl_logmax = IPL_LOGMAX;
6432         ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6433 
6434         /* from fr_natinit() */
6435         ifs->ifs_nat_logging = 1;
6436 
6437         /* from fr_stateinit() */
6438         ifs->ifs_ipstate_logging = 1;
6439 #else
6440         /* from fr_natinit() */
6441         ifs->ifs_nat_logging = 0;
6442 
6443         /* from fr_stateinit() */
6444         ifs->ifs_ipstate_logging = 0;
6445 #endif
6446         ifs->ifs_ipf_loopback = 0;
6447 
6448 }
6449 /*
6450  * Allocate a per-stack tuneable and copy in the names. Then
6451  * set it to point to each of the per-stack tunables.
6452  */
6453 void
6454 ipftuneable_alloc(ipf_stack_t *ifs)
6455 {
6456     ipftuneable_t *item;
6457 
6458     KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6459         sizeof (lcl_ipf_tuneables));
6460     bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6461         sizeof (lcl_ipf_tuneables));
6462 
6463 #define TUNE_SET(_ifs, _name, _field)                   \
6464     item = tune_lookup((_ifs), (_name));                \
6465     if (item != NULL) {                                 \
6466         item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field); \
6467         item->ipft_sz = sizeof ((_ifs)->_field);  \
6468     }
6469 
6470     TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6471     TUNE_SET(ifs, "fr_active", ifs_fr_active);
6472     TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6473     TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6474     TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6475     TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6476     TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6477     TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6478     TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6479     TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6480     TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6481     TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6482     TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6483     TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6484     TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6485     TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6486     TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6487     TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6488     TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6489     TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6490     TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6491     TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6492     TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6493     TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6494     TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6495     TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6496     TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6497     TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6498     TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6499     TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6500     TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6501     TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6502     TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6503     TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6504     TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6505     TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6506     TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6507     TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6508     TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6509     TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6510     TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6511     TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6512     TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6513     TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6514 #ifdef IPFILTER_LOG
6515     TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6516     TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6517     TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6518     TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6519     TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6520 #endif
6521 #undef TUNE_SET
6522 
6523         ipftuneable_setdefs(ifs);
6524 
6525 #ifdef _KERNEL
6526     (void) ipf_property_update(ipf_dev_info, ifs);
6527 #endif
6528 }
6529 
6530 void
6531 ipftuneable_free(ipf_stack_t *ifs)
6532 {
6533         KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6534         ifs->ifs_ipf_tuneables = NULL;
6535 }
6536 
6537 /* ------------------------------------------------------------------------ */
6538 /* Function:    fr_findtunebycookie                                         */
6539 /* Returns:     NULL = search failed, else pointer to tune struct           */
6540 /* Parameters:  cookie(I) - cookie value to search for amongst tuneables    */
6541 /*              next(O)   - pointer to place to store the cookie for the    */
6542 /*                          "next" tuneable, if it is desired.              */
6543 /*                                                                          */
6544 /* This function is used to walk through all of the existing tunables with  */
6545 /* successive calls.  It searches the known tunables for the one which has  */
6546 /* a matching value for "cookie" - ie its address.  When returning a match, */
6547 /* the next one to be found may be returned inside next.                    */
6548 /* ------------------------------------------------------------------------ */
6549 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6550 void *cookie, **next;
6551 ipf_stack_t * ifs;
6552 {
6553         ipftuneable_t *ta, **tap;
6554 
6555         for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6556                 if (ta == cookie) {
6557                         if (next != NULL) {
6558                                 /*
6559                                  * If the next entry in the array has a name
6560                                  * present, then return a pointer to it for
6561                                  * where to go next, else return a pointer to
6562                                  * the dynaminc list as a key to search there
6563                                  * next.  This facilitates a weak linking of
6564                                  * the two "lists" together.
6565                                  */
6566                                 if ((ta + 1)->ipft_name != NULL)
6567                                         *next = ta + 1;
6568                                 else
6569                                         *next = &ifs->ifs_ipf_tunelist;
6570                         }
6571                         return ta;
6572                 }
6573 
6574         for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6575                 if (tap == cookie) {
6576                         if (next != NULL)
6577                                 *next = &ta->ipft_next;
6578                         return ta;
6579                 }
6580 
6581         if (next != NULL)
6582                 *next = NULL;
6583         return NULL;
6584 }
6585 
6586 
6587 /* ------------------------------------------------------------------------ */
6588 /* Function:    fr_findtunebyname                                           */
6589 /* Returns:     NULL = search failed, else pointer to tune struct           */
6590 /* Parameters:  name(I) - name of the tuneable entry to find.               */
6591 /*                                                                          */
6592 /* Search the static array of tuneables and the list of dynamic tuneables   */
6593 /* for an entry with a matching name.  If we can find one, return a pointer */
6594 /* to the matching structure.                                               */
6595 /* ------------------------------------------------------------------------ */
6596 static ipftuneable_t *fr_findtunebyname(name, ifs)
6597 const char *name;
6598 ipf_stack_t *ifs;
6599 {
6600         ipftuneable_t *ta;
6601 
6602         for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6603                 if (!strcmp(ta->ipft_name, name)) {
6604                         return ta;
6605                 }
6606 
6607         for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6608                 if (!strcmp(ta->ipft_name, name)) {
6609                         return ta;
6610                 }
6611 
6612         return NULL;
6613 }
6614 
6615 
6616 /* ------------------------------------------------------------------------ */
6617 /* Function:    fr_addipftune                                               */
6618 /* Returns:     int - 0 == success, else failure                            */
6619 /* Parameters:  newtune - pointer to new tune struct to add to tuneables    */
6620 /*                                                                          */
6621 /* Appends the tune structure pointer to by "newtune" to the end of the     */
6622 /* current list of "dynamic" tuneable parameters.  Once added, the owner    */
6623 /* of the object is not expected to ever change "ipft_next".                */
6624 /* ------------------------------------------------------------------------ */
6625 int fr_addipftune(newtune, ifs)
6626 ipftuneable_t *newtune;
6627 ipf_stack_t *ifs;
6628 {
6629         ipftuneable_t *ta, **tap;
6630 
6631         ta = fr_findtunebyname(newtune->ipft_name, ifs);
6632         if (ta != NULL)
6633                 return EEXIST;
6634 
6635         for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6636                 ;
6637 
6638         newtune->ipft_next = NULL;
6639         *tap = newtune;
6640         return 0;
6641 }
6642 
6643 
6644 /* ------------------------------------------------------------------------ */
6645 /* Function:    fr_delipftune                                               */
6646 /* Returns:     int - 0 == success, else failure                            */
6647 /* Parameters:  oldtune - pointer to tune struct to remove from the list of */
6648 /*                        current dynamic tuneables                         */
6649 /*                                                                          */
6650 /* Search for the tune structure, by pointer, in the list of those that are */
6651 /* dynamically added at run time.  If found, adjust the list so that this   */
6652 /* structure is no longer part of it.                                       */
6653 /* ------------------------------------------------------------------------ */
6654 int fr_delipftune(oldtune, ifs)
6655 ipftuneable_t *oldtune;
6656 ipf_stack_t *ifs;
6657 {
6658         ipftuneable_t *ta, **tap;
6659 
6660         for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6661                 if (ta == oldtune) {
6662                         *tap = oldtune->ipft_next;
6663                         oldtune->ipft_next = NULL;
6664                         return 0;
6665                 }
6666 
6667         return ESRCH;
6668 }
6669 
6670 
6671 /* ------------------------------------------------------------------------ */
6672 /* Function:    fr_ipftune                                                  */
6673 /* Returns:     int - 0 == success, else failure                            */
6674 /* Parameters:  cmd(I)  - ioctl command number                              */
6675 /*              data(I) - pointer to ioctl data structure                   */
6676 /*                                                                          */
6677 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET.  These  */
6678 /* three ioctls provide the means to access and control global variables    */
6679 /* within IPFilter, allowing (for example) timeouts and table sizes to be   */
6680 /* changed without rebooting, reloading or recompiling.  The initialisation */
6681 /* and 'destruction' routines of the various components of ipfilter are all */
6682 /* each responsible for handling their own values being too big.            */
6683 /* ------------------------------------------------------------------------ */
6684 int fr_ipftune(cmd, data, ifs)
6685 ioctlcmd_t cmd;
6686 void *data;
6687 ipf_stack_t *ifs;
6688 {
6689         ipftuneable_t *ta;
6690         ipftune_t tu;
6691         void *cookie;
6692         int error;
6693 
6694         error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6695         if (error != 0)
6696                 return error;
6697 
6698         tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6699         cookie = tu.ipft_cookie;
6700         ta = NULL;
6701 
6702         switch (cmd)
6703         {
6704         case SIOCIPFGETNEXT :
6705                 /*
6706                  * If cookie is non-NULL, assume it to be a pointer to the last
6707                  * entry we looked at, so find it (if possible) and return a
6708                  * pointer to the next one after it.  The last entry in the
6709                  * the table is a NULL entry, so when we get to it, set cookie
6710                  * to NULL and return that, indicating end of list, erstwhile
6711                  * if we come in with cookie set to NULL, we are starting anew
6712                  * at the front of the list.
6713                  */
6714                 if (cookie != NULL) {
6715                         ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6716                 } else {
6717                         ta = ifs->ifs_ipf_tuneables;
6718                         tu.ipft_cookie = ta + 1;
6719                 }
6720                 if (ta != NULL) {
6721                         /*
6722                          * Entry found, but does the data pointed to by that
6723                          * row fit in what we can return?
6724                          */
6725                         if (ta->ipft_sz > sizeof(tu.ipft_un))
6726                                 return EINVAL;
6727 
6728                         tu.ipft_vlong = 0;
6729                         if (ta->ipft_sz == sizeof(u_long))
6730                                 tu.ipft_vlong = *ta->ipft_plong;
6731                         else if (ta->ipft_sz == sizeof(u_int))
6732                                 tu.ipft_vint = *ta->ipft_pint;
6733                         else if (ta->ipft_sz == sizeof(u_short))
6734                                 tu.ipft_vshort = *ta->ipft_pshort;
6735                         else if (ta->ipft_sz == sizeof(u_char))
6736                                 tu.ipft_vchar = *ta->ipft_pchar;
6737 
6738                         tu.ipft_sz = ta->ipft_sz;
6739                         tu.ipft_min = ta->ipft_min;
6740                         tu.ipft_max = ta->ipft_max;
6741                         tu.ipft_flags = ta->ipft_flags;
6742                         bcopy(ta->ipft_name, tu.ipft_name,
6743                               MIN(sizeof(tu.ipft_name),
6744                                   strlen(ta->ipft_name) + 1));
6745                 }
6746                 error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6747                 break;
6748 
6749         case SIOCIPFGET :
6750         case SIOCIPFSET :
6751                 /*
6752                  * Search by name or by cookie value for a particular entry
6753                  * in the tuning paramter table.
6754                  */
6755                 error = ESRCH;
6756                 if (cookie != NULL) {
6757                         ta = fr_findtunebycookie(cookie, NULL, ifs);
6758                         if (ta != NULL)
6759                                 error = 0;
6760                 } else if (tu.ipft_name[0] != '\0') {
6761                         ta = fr_findtunebyname(tu.ipft_name, ifs);
6762                         if (ta != NULL)
6763                                 error = 0;
6764                 }
6765                 if (error != 0)
6766                         break;
6767 
6768                 if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6769                         /*
6770                          * Fetch the tuning parameters for a particular value
6771                          */
6772                         tu.ipft_vlong = 0;
6773                         if (ta->ipft_sz == sizeof(u_long))
6774                                 tu.ipft_vlong = *ta->ipft_plong;
6775                         else if (ta->ipft_sz == sizeof(u_int))
6776                                 tu.ipft_vint = *ta->ipft_pint;
6777                         else if (ta->ipft_sz == sizeof(u_short))
6778                                 tu.ipft_vshort = *ta->ipft_pshort;
6779                         else if (ta->ipft_sz == sizeof(u_char))
6780                                 tu.ipft_vchar = *ta->ipft_pchar;
6781                         tu.ipft_cookie = ta;
6782                         tu.ipft_sz = ta->ipft_sz;
6783                         tu.ipft_min = ta->ipft_min;
6784                         tu.ipft_max = ta->ipft_max;
6785                         tu.ipft_flags = ta->ipft_flags;
6786                         error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6787 
6788                 } else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6789                         /*
6790                          * Set an internal parameter.  The hard part here is
6791                          * getting the new value safely and correctly out of
6792                          * the kernel (given we only know its size, not type.)
6793                          */
6794                         u_long in;
6795 
6796                         if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6797                             (ifs->ifs_fr_running > 0)) {
6798                                 error = EBUSY;
6799                                 break;
6800                         }
6801 
6802                         in = tu.ipft_vlong;
6803                         if (in < ta->ipft_min || in > ta->ipft_max) {
6804                                 error = EINVAL;
6805                                 break;
6806                         }
6807 
6808                         if (ta->ipft_sz == sizeof(u_long)) {
6809                                 tu.ipft_vlong = *ta->ipft_plong;
6810                                 *ta->ipft_plong = in;
6811                         } else if (ta->ipft_sz == sizeof(u_int)) {
6812                                 tu.ipft_vint = *ta->ipft_pint;
6813                                 *ta->ipft_pint = (u_int)(in & 0xffffffff);
6814                         } else if (ta->ipft_sz == sizeof(u_short)) {
6815                                 tu.ipft_vshort = *ta->ipft_pshort;
6816                                 *ta->ipft_pshort = (u_short)(in & 0xffff);
6817                         } else if (ta->ipft_sz == sizeof(u_char)) {
6818                                 tu.ipft_vchar = *ta->ipft_pchar;
6819                                 *ta->ipft_pchar = (u_char)(in & 0xff);
6820                         }
6821                         error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6822                 }
6823                 break;
6824 
6825         default :
6826                 error = EINVAL;
6827                 break;
6828         }
6829 
6830         return error;
6831 }
6832 
6833 
6834 /* ------------------------------------------------------------------------ */
6835 /* Function:    fr_initialise                                               */
6836 /* Returns:     int - 0 == success,  < 0 == failure                         */
6837 /* Parameters:  None.                                                       */
6838 /*                                                                          */
6839 /* Call of the initialise functions for all the various subsystems inside   */
6840 /* of IPFilter.  If any of them should fail, return immeadiately a failure  */
6841 /* BUT do not try to recover from the error here.                           */
6842 /* ------------------------------------------------------------------------ */
6843 int fr_initialise(ifs)
6844 ipf_stack_t *ifs;
6845 {
6846         int i;
6847 
6848 #ifdef IPFILTER_LOG
6849         i = fr_loginit(ifs);
6850         if (i < 0)
6851                 return -10 + i;
6852 #endif
6853         i = fr_natinit(ifs);
6854         if (i < 0)
6855                 return -20 + i;
6856 
6857         i = fr_stateinit(ifs);
6858         if (i < 0)
6859                 return -30 + i;
6860 
6861         i = fr_authinit(ifs);
6862         if (i < 0)
6863                 return -40 + i;
6864 
6865         i = fr_fraginit(ifs);
6866         if (i < 0)
6867                 return -50 + i;
6868 
6869         i = appr_init(ifs);
6870         if (i < 0)
6871                 return -60 + i;
6872 
6873 #ifdef IPFILTER_SYNC
6874         i = ipfsync_init(ifs);
6875         if (i < 0)
6876                 return -70 + i;
6877 #endif
6878 #ifdef IPFILTER_SCAN
6879         i = ipsc_init(ifs);
6880         if (i < 0)
6881                 return -80 + i;
6882 #endif
6883 #ifdef IPFILTER_LOOKUP
6884         i = ip_lookup_init(ifs);
6885         if (i < 0)
6886                 return -90 + i;
6887 #endif
6888 #ifdef IPFILTER_COMPILED
6889         ipfrule_add(ifs);
6890 #endif
6891         return 0;
6892 }
6893 
6894 
6895 /* ------------------------------------------------------------------------ */
6896 /* Function:    fr_deinitialise                                             */
6897 /* Returns:     None.                                                       */
6898 /* Parameters:  None.                                                       */
6899 /*                                                                          */
6900 /* Call all the various subsystem cleanup routines to deallocate memory or  */
6901 /* destroy locks or whatever they've done that they need to now undo.       */
6902 /* The order here IS important as there are some cross references of        */
6903 /* internal data structures.                                                */
6904 /* ------------------------------------------------------------------------ */
6905 void fr_deinitialise(ifs)
6906 ipf_stack_t *ifs;
6907 {
6908         fr_fragunload(ifs);
6909         fr_authunload(ifs);
6910         fr_natunload(ifs);
6911         fr_stateunload(ifs);
6912 #ifdef IPFILTER_SCAN
6913         fr_scanunload(ifs);
6914 #endif
6915         appr_unload(ifs);
6916 
6917 #ifdef IPFILTER_COMPILED
6918         ipfrule_remove(ifs);
6919 #endif
6920 
6921         (void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6922         (void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6923         (void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6924         (void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6925 
6926 #ifdef IPFILTER_LOOKUP
6927         ip_lookup_unload(ifs);
6928 #endif
6929 
6930 #ifdef IPFILTER_LOG
6931         fr_logunload(ifs);
6932 #endif
6933 }
6934 
6935 
6936 /* ------------------------------------------------------------------------ */
6937 /* Function:    fr_zerostats                                                */
6938 /* Returns:     int - 0 = success, else failure                             */
6939 /* Parameters:  data(O) - pointer to pointer for copying data back to       */
6940 /*                                                                          */
6941 /* Copies the current statistics out to userspace and then zero's the       */
6942 /* current ones in the kernel. The lock is only held across the bzero() as  */
6943 /* the copyout may result in paging (ie network activity.)                  */
6944 /* ------------------------------------------------------------------------ */
6945 int     fr_zerostats(data, ifs)
6946 caddr_t data;
6947 ipf_stack_t *ifs;
6948 {
6949         friostat_t fio;
6950         int error;
6951 
6952         fr_getstat(&fio, ifs);
6953         error = copyoutptr(&fio, data, sizeof(fio));
6954         if (error)
6955                 return EFAULT;
6956 
6957         WRITE_ENTER(&ifs->ifs_ipf_mutex);
6958         bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6959         RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6960 
6961         return 0;
6962 }
6963 
6964 
6965 #ifdef _KERNEL
6966 /* ------------------------------------------------------------------------ */
6967 /* Function:    fr_resolvedest                                              */
6968 /* Returns:     Nil                                                         */
6969 /* Parameters:  fdp(IO) - pointer to destination information to resolve     */
6970 /*              v(I)    - IP protocol version to match                      */
6971 /*                                                                          */
6972 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6973 /* if a matching name can be found for the particular IP protocol version   */
6974 /* then store the interface pointer in the frdest struct.  If no match is   */
6975 /* found, then set the interface pointer to be -1 as NULL is considered to  */
6976 /* indicate there is no information at all in the structure.                */
6977 /* ------------------------------------------------------------------------ */
6978 void fr_resolvedest(fdp, v, ifs)
6979 frdest_t *fdp;
6980 int v;
6981 ipf_stack_t *ifs;
6982 {
6983         fdp->fd_ifp = NULL;
6984 
6985         if (*fdp->fd_ifname != '\0') {
6986                 fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6987                 if (fdp->fd_ifp == NULL)
6988                         fdp->fd_ifp = (void *)-1;
6989         }
6990 }
6991 #endif /* _KERNEL */
6992 
6993 
6994 /* ------------------------------------------------------------------------ */
6995 /* Function:    fr_resolvenic                                               */
6996 /* Returns:     void* - NULL = wildcard name, -1 = failed to find NIC, else */
6997 /*                      pointer to interface structure for NIC              */
6998 /* Parameters:  name(I) - complete interface name                           */
6999 /*              v(I)    - IP protocol version                               */
7000 /*                                                                          */
7001 /* Look for a network interface structure that firstly has a matching name  */
7002 /* to that passed in and that is also being used for that IP protocol       */
7003 /* version (necessary on some platforms where there are separate listings   */
7004 /* for both IPv4 and IPv6 on the same physical NIC.                         */
7005 /*                                                                          */
7006 /* One might wonder why name gets terminated with a \0 byte in here.  The   */
7007 /* reason is an interface name could get into the kernel structures of ipf  */
7008 /* in any number of ways and so long as they all use the same sized array   */
7009 /* to put the name in, it makes sense to ensure it gets null terminated     */
7010 /* before it is used for its intended purpose - finding its match in the    */
7011 /* kernel's list of configured interfaces.                                  */
7012 /*                                                                          */
7013 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an     */
7014 /*       array for the name that is LIFNAMSIZ bytes (at least) in length.   */
7015 /* ------------------------------------------------------------------------ */
7016 void *fr_resolvenic(name, v, ifs)
7017 char *name;
7018 int v;
7019 ipf_stack_t *ifs;
7020 {
7021         void *nic;
7022 
7023         if (name[0] == '\0')
7024                 return NULL;
7025 
7026         if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
7027                 return NULL;
7028         }
7029 
7030         name[LIFNAMSIZ - 1] = '\0';
7031 
7032         nic = GETIFP(name, v, ifs);
7033         if (nic == NULL)
7034                 nic = (void *)-1;
7035         return nic;
7036 }
7037 
7038 
7039 /* ------------------------------------------------------------------------ */
7040 /* Function:    ipf_expiretokens                                            */
7041 /* Returns:     None.                                                       */
7042 /* Parameters:  ifs - ipf stack instance                                    */
7043 /*                                                                          */
7044 /* This function is run every ipf tick to see if there are any tokens that  */
7045 /* have been held for too long and need to be freed up.                     */
7046 /* ------------------------------------------------------------------------ */
7047 void ipf_expiretokens(ifs)
7048 ipf_stack_t *ifs;
7049 {
7050         ipftoken_t *it;
7051 
7052         WRITE_ENTER(&ifs->ifs_ipf_tokens);
7053         while ((it = ifs->ifs_ipftokenhead) != NULL) {
7054                 if (it->ipt_die > ifs->ifs_fr_ticks)
7055                         break;
7056 
7057                 ipf_freetoken(it, ifs);
7058         }
7059         RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7060 }
7061 
7062 
7063 /* ------------------------------------------------------------------------ */
7064 /* Function:    ipf_deltoken                                                */
7065 /* Returns:     int - 0 = success, else error                               */
7066 /* Parameters:  type(I) - the token type to match                           */
7067 /*              uid(I)  - uid owning the token                              */
7068 /*              ptr(I)  - context pointer for the token                     */
7069 /*              ifs - ipf stack instance                                    */
7070 /*                                                                          */
7071 /* This function looks for a a token in the current list that matches up    */
7072 /* the fields (type, uid, ptr).  If none is found, ESRCH is returned, else  */
7073 /* call ipf_freetoken() to remove it from the list.                         */
7074 /* ------------------------------------------------------------------------ */
7075 int ipf_deltoken(type, uid, ptr, ifs)
7076 int type, uid;
7077 void *ptr;
7078 ipf_stack_t *ifs;
7079 {
7080         ipftoken_t *it;
7081         int error = ESRCH;
7082 
7083         WRITE_ENTER(&ifs->ifs_ipf_tokens);
7084         for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
7085                 if (ptr == it->ipt_ctx && type == it->ipt_type &&
7086                     uid == it->ipt_uid) {
7087                         ipf_freetoken(it, ifs);
7088                         error = 0;
7089                         break;
7090         }
7091         RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7092 
7093         return error;
7094 }
7095 
7096 
7097 /* ------------------------------------------------------------------------ */
7098 /* Function:    ipf_unlinktoken                                             */
7099 /* Returns:     None.                                                       */
7100 /* Parameters:  token(I) - pointer to token structure                       */
7101 /*              ifs - ipf stack instance                                    */
7102 /*                                                                          */
7103 /* This function unlinks a token structure from the linked list of tokens   */
7104 /* that it belongs to.  The head pointer never needs to be explicitly       */
7105 /* adjusted, but the tail does due to the linked list implementation.       */
7106 /* ------------------------------------------------------------------------ */
7107 static void ipf_unlinktoken(token, ifs)
7108 ipftoken_t *token;
7109 ipf_stack_t *ifs;
7110 {
7111 
7112         if (ifs->ifs_ipftokentail == &token->ipt_next)
7113                 ifs->ifs_ipftokentail = token->ipt_pnext;
7114 
7115         *token->ipt_pnext = token->ipt_next;
7116         if (token->ipt_next != NULL)
7117                 token->ipt_next->ipt_pnext = token->ipt_pnext;
7118 }
7119 
7120 
7121 /* ------------------------------------------------------------------------ */
7122 /* Function:    ipf_findtoken                                               */
7123 /* Returns:     ipftoken_t * - NULL if no memory, else pointer to token     */
7124 /* Parameters:  type(I) - the token type to match                           */
7125 /*              uid(I) - uid owning the token                               */
7126 /*              ptr(I) - context pointer for the token                      */
7127 /*              ifs - ipf stack instance                                    */
7128 /*                                                                          */
7129 /* This function looks for a live token in the list of current tokens that  */
7130 /* matches the tuple (type, uid, ptr).  If one cannot be found then one is  */
7131 /* allocated.  If one is found then it is moved to the top of the list of   */
7132 /* currently active tokens.                                                 */
7133 /*                                                                          */
7134 /* NOTE: It is by design that this function returns holding a read lock on  */
7135 /*       ipf_tokens.  Callers must make sure they release it!               */
7136 /* ------------------------------------------------------------------------ */
7137 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
7138 int type, uid;
7139 void *ptr;
7140 ipf_stack_t *ifs;
7141 {
7142         ipftoken_t *it, *new;
7143 
7144         KMALLOC(new, ipftoken_t *);
7145 
7146         WRITE_ENTER(&ifs->ifs_ipf_tokens);
7147         for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
7148                 if (it->ipt_alive == 0)
7149                         continue;
7150                 if (ptr == it->ipt_ctx && type == it->ipt_type &&
7151                     uid == it->ipt_uid)
7152                         break;
7153         }
7154 
7155         if (it == NULL) {
7156                 it = new;
7157                 new = NULL;
7158                 if (it == NULL)
7159                         return NULL;
7160                 it->ipt_data = NULL;
7161                 it->ipt_ctx = ptr;
7162                 it->ipt_uid = uid;
7163                 it->ipt_type = type;
7164                 it->ipt_next = NULL;
7165                 it->ipt_alive = 1;
7166         } else {
7167                 if (new != NULL) {
7168                         KFREE(new);
7169                         new = NULL;
7170                 }
7171 
7172                 ipf_unlinktoken(it, ifs);
7173         }
7174         it->ipt_pnext = ifs->ifs_ipftokentail;
7175         *ifs->ifs_ipftokentail = it;
7176         ifs->ifs_ipftokentail = &it->ipt_next;
7177         it->ipt_next = NULL;
7178 
7179         it->ipt_die = ifs->ifs_fr_ticks + 2;
7180 
7181         MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
7182 
7183         return it;
7184 }
7185 
7186 
7187 /* ------------------------------------------------------------------------ */
7188 /* Function:    ipf_freetoken                                               */
7189 /* Returns:     None.                                                       */
7190 /* Parameters:  token(I) - pointer to token structure                       */
7191 /*              ifs - ipf stack instance                                    */
7192 /*                                                                          */
7193 /* This function unlinks a token from the linked list and on the path to    */
7194 /* free'ing the data, it calls the dereference function that is associated  */
7195 /* with the type of data pointed to by the token as it is considered to     */
7196 /* hold a reference to it.                                                  */
7197 /* ------------------------------------------------------------------------ */
7198 void ipf_freetoken(token, ifs)
7199 ipftoken_t *token;
7200 ipf_stack_t *ifs;
7201 {
7202         void *data, **datap;
7203 
7204         ipf_unlinktoken(token, ifs);
7205 
7206         data = token->ipt_data;
7207         datap = &data;
7208 
7209         if ((data != NULL) && (data != (void *)-1)) {
7210                 switch (token->ipt_type)
7211                 {
7212                 case IPFGENITER_IPF :
7213                         (void)fr_derefrule((frentry_t **)datap, ifs);
7214                         break;
7215                 case IPFGENITER_IPNAT :
7216                         WRITE_ENTER(&ifs->ifs_ipf_nat);
7217                         fr_ipnatderef((ipnat_t **)datap, ifs);
7218                         RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7219                         break;
7220                 case IPFGENITER_NAT :
7221                         fr_natderef((nat_t **)datap, ifs);
7222                         break;
7223                 case IPFGENITER_STATE :
7224                         fr_statederef((ipstate_t **)datap, ifs);
7225                         break;
7226                 case IPFGENITER_FRAG :
7227                         fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
7228                         break;
7229                 case IPFGENITER_NATFRAG :
7230                         fr_fragderef((ipfr_t **)datap,
7231                                      &ifs->ifs_ipf_natfrag, ifs);
7232                         break;
7233                 case IPFGENITER_HOSTMAP :
7234                         WRITE_ENTER(&ifs->ifs_ipf_nat);
7235                         fr_hostmapdel((hostmap_t **)datap);
7236                         RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7237                         break;
7238                 default :
7239                         (void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7240                         break;
7241                 }
7242         }
7243 
7244         KFREE(token);
7245 }
7246 
7247 
7248 /* ------------------------------------------------------------------------ */
7249 /* Function:    ipf_getnextrule                                             */
7250 /* Returns:     int - 0 = success, else error                               */
7251 /* Parameters:  t(I)   - pointer to destination information to resolve      */
7252 /*              ptr(I) - pointer to ipfobj_t to copyin from user space      */
7253 /*              ifs - ipf stack instance                                    */
7254 /*                                                                          */
7255 /* This function's first job is to bring in the ipfruleiter_t structure via */
7256 /* the ipfobj_t structure to determine what should be the next rule to      */
7257 /* return. Once the ipfruleiter_t has been brought in, it then tries to     */
7258 /* find the 'next rule'.  This may include searching rule group lists or    */
7259 /* just be as simple as looking at the 'next' field in the rule structure.  */
7260 /* When we have found the rule to return, increase its reference count and  */
7261 /* if we used an existing rule to get here, decrease its reference count.   */
7262 /* ------------------------------------------------------------------------ */
7263 int ipf_getnextrule(t, ptr, ifs)
7264 ipftoken_t *t;
7265 void *ptr;
7266 ipf_stack_t *ifs;
7267 {
7268         frentry_t *fr, *next, zero;
7269         int error, out, count;
7270         ipfruleiter_t it;
7271         frgroup_t *fg;
7272         char *dst;
7273 
7274         if (t == NULL || ptr == NULL)
7275                 return EFAULT;
7276         error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7277         if (error != 0)
7278                 return error;
7279         if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7280                 return EINVAL;
7281         if ((it.iri_inout < 0) || (it.iri_inout > 3))
7282                 return EINVAL;
7283         if (it.iri_nrules == 0)
7284                 return EINVAL;
7285         if ((it.iri_active != 0) && (it.iri_active != 1))
7286                 return EINVAL;
7287         if (it.iri_rule == NULL)
7288                 return EFAULT;
7289 
7290         /*
7291          * Use bitmask on it.iri_inout to determine direction.
7292          * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7293          * F_IN (0) and F_ACIN (2) mask to out = 0.
7294          */
7295         out = it.iri_inout & F_OUT;
7296         READ_ENTER(&ifs->ifs_ipf_mutex);
7297 
7298         /*
7299          * Retrieve "previous" entry from token and find the next entry.
7300          */
7301         fr = t->ipt_data;
7302         if (fr == NULL) {
7303                 if (*it.iri_group == '\0') {
7304                         /*
7305                          * Use bitmask again to determine accounting or not.
7306                          * F_ACIN will mask to accounting cases F_ACIN (2)
7307                          * or F_ACOUT (3), but not F_IN or F_OUT.
7308                          */
7309                         if ((it.iri_inout & F_ACIN) != 0) {
7310                                 if (it.iri_ver == AF_INET)
7311                                         next = ifs->ifs_ipacct
7312                                             [out][it.iri_active];
7313                                 else
7314                                         next = ifs->ifs_ipacct6
7315                                             [out][it.iri_active];
7316                         } else {
7317                                 if (it.iri_ver == AF_INET)
7318                                         next = ifs->ifs_ipfilter
7319                                             [out][it.iri_active];
7320                                 else
7321                                         next = ifs->ifs_ipfilter6
7322                                             [out][it.iri_active];
7323                         }
7324                 } else {
7325                         fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7326                                           it.iri_active, NULL, ifs);
7327                         if (fg != NULL)
7328                                 next = fg->fg_start;
7329                         else
7330                                 next = NULL;
7331                 }
7332         } else {
7333                 next = fr->fr_next;
7334         }
7335 
7336         dst = (char *)it.iri_rule;
7337         /*
7338          * The ipfruleiter may ask for more than 1 rule at a time to be
7339          * copied out, so long as that many exist in the list to start with!
7340          */
7341         for (count = it.iri_nrules; count > 0; count--) {
7342                 /*
7343                  * If we found an entry, add reference to it and update token.
7344                  * Otherwise, zero out data to be returned and NULL out token.
7345                  */
7346                 if (next != NULL) {
7347                         MUTEX_ENTER(&next->fr_lock);
7348                         next->fr_ref++;
7349                         MUTEX_EXIT(&next->fr_lock);
7350                         t->ipt_data = next;
7351                 } else {
7352                         bzero(&zero, sizeof(zero));
7353                         next = &zero;
7354                         t->ipt_data = NULL;
7355                 }
7356 
7357                 /*
7358                  * Now that we have ref, it's save to give up lock.
7359                  */
7360                 RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7361  
7362                 /*
7363                  * Copy out data and clean up references and token as needed.
7364                  */
7365                 error = COPYOUT(next, dst, sizeof(*next));
7366                 if (error != 0)
7367                         error = EFAULT;
7368                 if (t->ipt_data == NULL) {
7369                         ipf_freetoken(t, ifs);
7370                         break;
7371                 } else {
7372                         if (fr != NULL)
7373                                 (void) fr_derefrule(&fr, ifs);
7374                         if (next->fr_data != NULL) {
7375                                 dst += sizeof(*next);
7376                                 error = COPYOUT(next->fr_data, dst,
7377                                                 next->fr_dsize);
7378                                 if (error != 0)
7379                                         error = EFAULT;
7380                                 else
7381                                         dst += next->fr_dsize;
7382                         }
7383                         if (next->fr_next == NULL) {
7384                                 ipf_freetoken(t, ifs);
7385                                 break;
7386                         }
7387                 }
7388  
7389                 if ((count == 1) || (error != 0))
7390                         break;
7391 
7392                 READ_ENTER(&ifs->ifs_ipf_mutex);
7393                 fr = next;
7394                 next = fr->fr_next;
7395         }
7396 
7397         return error;
7398 }
7399 
7400 
7401 /* ------------------------------------------------------------------------ */
7402 /* Function:    fr_frruleiter                                               */
7403 /* Returns:     int - 0 = success, else error                               */
7404 /* Parameters:  data(I) - the token type to match                           */
7405 /*              uid(I) - uid owning the token                               */
7406 /*              ptr(I) - context pointer for the token                      */
7407 /*              ifs - ipf stack instance                                    */
7408 /*                                                                          */
7409 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7410 /* ipf_getnextrule.  It's role is to find the right token in the kernel for */
7411 /* the process doing the ioctl and use that to ask for the next rule.       */
7412 /* ------------------------------------------------------------------------ */
7413 int ipf_frruleiter(data, uid, ctx, ifs)
7414 void *data, *ctx;
7415 int uid;
7416 ipf_stack_t *ifs;
7417 {
7418         ipftoken_t *token;
7419         int error;
7420 
7421         token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7422         if (token != NULL)
7423                 error = ipf_getnextrule(token, data, ifs);
7424         else
7425                 error = EFAULT;
7426         RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7427 
7428         return error;
7429 }
7430 
7431 
7432 /* ------------------------------------------------------------------------ */
7433 /* Function:    ipf_geniter                                                 */
7434 /* Returns:     int - 0 = success, else error                               */
7435 /* Parameters:  token(I) - pointer to ipftoken structure                    */
7436 /*              itp(I) - pointer to ipfgeniter structure                    */
7437 /*              ifs - ipf stack instance                                    */
7438 /*                                                                          */
7439 /* Generic iterator called from ipf_genericiter.  Currently only used for   */
7440 /* walking through list of fragments.                                       */
7441 /* ------------------------------------------------------------------------ */
7442 int ipf_geniter(token, itp, ifs)
7443 ipftoken_t *token;
7444 ipfgeniter_t *itp;
7445 ipf_stack_t *ifs;
7446 {
7447         int error;
7448 
7449         switch (itp->igi_type)
7450         {
7451         case IPFGENITER_FRAG :
7452                 error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7453                                     &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7454                                     ifs);
7455                 break;
7456         default :
7457                 error = EINVAL;
7458                 break;
7459         }
7460 
7461         return error;
7462 }
7463 
7464 
7465 /* ------------------------------------------------------------------------ */
7466 /* Function:    ipf_genericiter                                             */
7467 /* Returns:     int - 0 = success, else error                               */
7468 /* Parameters:  data(I) - the token type to match                           */
7469 /*              uid(I) - uid owning the token                               */
7470 /*              ptr(I) - context pointer for the token                      */
7471 /*              ifs - ipf stack instance                                    */
7472 /*                                                                          */
7473 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7474 /* ipf_geniter when handling SIOCGENITER.  It's role is to find the right   */
7475 /* token in the kernel for the process using the ioctl, and to use that     */
7476 /* token when calling ipf_geniter.                                          */
7477 /* ------------------------------------------------------------------------ */
7478 int ipf_genericiter(data, uid, ctx, ifs)
7479 void *data, *ctx;
7480 int uid;
7481 ipf_stack_t *ifs;
7482 {
7483         ipftoken_t *token;
7484         ipfgeniter_t iter;
7485         int error;
7486 
7487         error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7488         if (error != 0)
7489                 return error;
7490 
7491         token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7492         if (token != NULL) {
7493                 token->ipt_subtype = iter.igi_type;
7494                 error = ipf_geniter(token, &iter, ifs);
7495         } else
7496                 error = EFAULT;
7497         RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7498 
7499         return error;
7500 }
7501 
7502 
7503 /* --------------------------------------------------------------------- */
7504 /* Function:    ipf_earlydrop                                            */
7505 /* Returns:     number of dropped/removed entries from the queue         */
7506 /* Parameters:  flushtype - which table we're cleaning (NAT or State)    */
7507 /*              ifq     - pointer to queue with entries to be deleted    */
7508 /*              idletime - entry must be idle this long to be deleted    */
7509 /*              ifs     - ipf stack instance                             */
7510 /*                                                                       */
7511 /* Function is invoked from state/NAT flush routines to remove entries   */
7512 /* from specified timeout queue, based on how long they've sat idle,     */
7513 /* without waiting for it to happen on its own.                          */
7514 /* --------------------------------------------------------------------- */
7515 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7516 int flushtype;
7517 ipftq_t *ifq;
7518 int idletime;
7519 ipf_stack_t *ifs;
7520 {
7521         ipftqent_t *tqe, *tqn;
7522         unsigned int dropped;
7523         int droptick;
7524         void *ent;
7525 
7526         if (ifq == NULL)
7527                 return (0);
7528 
7529         dropped = 0;
7530 
7531         /*
7532          * Determine the tick representing the idle time we're interested
7533          * in.  If an entry exists in the queue, and it was touched before
7534          * that tick, then it's been idle longer than idletime, so it should
7535          * be deleted.
7536          */
7537         droptick = ifs->ifs_fr_ticks - idletime;
7538         tqn = ifq->ifq_head;
7539         while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7540                 tqn = tqe->tqe_next;
7541                 ent = tqe->tqe_parent;
7542                 switch (flushtype)
7543                 {
7544                 case NAT_FLUSH:
7545                         if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7546                                 dropped++;
7547                         break;
7548                 case STATE_FLUSH:
7549                         if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7550                                 dropped++;
7551                         break;
7552                 default:
7553                         return (0);
7554                 }
7555         }
7556         return (dropped);
7557 }
7558 
7559 
7560 /* --------------------------------------------------------------------- */
7561 /* Function:    ipf_flushclosing                                         */
7562 /* Returns:     int - number of entries deleted                          */
7563 /* Parameters:  flushtype - which table we're cleaning (NAT or State)    */
7564 /*              stateval - TCP state at which to start removing entries  */
7565 /*              ipfqs - pointer to timeout queues                        */
7566 /*              userqs - pointer to user defined queues                  */
7567 /*              ifs  - ipf stack instance                                */
7568 /*                                                                       */
7569 /* Remove state/NAT table entries for TCP connections which are in the   */
7570 /* process of closing, and have at least reached the state specified by  */
7571 /* the 'stateval' parameter.                                             */
7572 /* --------------------------------------------------------------------- */
7573 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7574 int flushtype, stateval;
7575 ipftq_t *ipfqs, *userqs;
7576 ipf_stack_t *ifs;
7577 {
7578         ipftq_t *ifq, *ifqn;
7579         ipftqent_t *tqe, *tqn;
7580         int dropped;
7581         void *ent;
7582         nat_t *nat;
7583         ipstate_t *is;
7584 
7585         dropped = 0;
7586 
7587         /*
7588          * Start by deleting any entries in specific timeout queues.
7589          */
7590         ifqn = &ipfqs[stateval];
7591         while ((ifq = ifqn) != NULL) {
7592                 ifqn = ifq->ifq_next;
7593                 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7594         }
7595 
7596         /*
7597          * Next, look through user defined queues for closing entries.
7598          */
7599         ifqn = userqs;
7600         while ((ifq = ifqn) != NULL) {
7601                 ifqn = ifq->ifq_next;
7602                 tqn = ifq->ifq_head;
7603                 while ((tqe = tqn) != NULL) {
7604                         tqn = tqe->tqe_next;
7605                         ent = tqe->tqe_parent;
7606                         switch (flushtype)
7607                         {
7608                         case NAT_FLUSH:
7609                                 nat = (nat_t *)ent;
7610                                 if ((nat->nat_p == IPPROTO_TCP) &&
7611                                     (nat->nat_tcpstate[0] >= stateval) &&
7612                                     (nat->nat_tcpstate[1] >= stateval) &&
7613                                     (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7614                                         dropped++;
7615                                 break;
7616                         case STATE_FLUSH:
7617                                 is = (ipstate_t *)ent;
7618                                 if ((is->is_p == IPPROTO_TCP) &&
7619                                     (is->is_state[0] >= stateval) &&
7620                                     (is->is_state[1] >= stateval) &&
7621                                     (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7622                                         dropped++;
7623                                 break;
7624                         default:
7625                                 return (0);
7626                         }
7627                 }
7628         }
7629         return (dropped);
7630 }
7631 
7632 
7633 /* --------------------------------------------------------------------- */
7634 /* Function:    ipf_extraflush                                           */
7635 /* Returns:     int - number of entries flushed (0 = none)               */
7636 /* Parameters:  flushtype - which table we're cleaning (NAT or State)    */
7637 /*              ipfqs - pointer to 'established' timeout queue           */
7638 /*              userqs - pointer to user defined queues                  */
7639 /*              ifs  - ipf stack instance                                */
7640 /*                                                                       */
7641 /* This function gets called when either NAT or state tables fill up.    */
7642 /* We need to try a bit harder to free up some space.  The function will */
7643 /* flush entries for TCP connections which have been idle a long time.   */
7644 /*                                                                       */
7645 /* Currently, the idle time is checked using values from ideltime_tab[]  */
7646 /* --------------------------------------------------------------------- */
7647 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7648 int flushtype;
7649 ipftq_t *ipfqs, *userqs;
7650 ipf_stack_t *ifs;
7651 {
7652         ipftq_t *ifq, *ifqn;
7653         int idletime, removed, idle_idx;
7654 
7655         removed = 0;
7656 
7657         /*
7658          * Determine initial threshold for minimum idle time based on
7659          * how long ipfilter has been running.  Ipfilter needs to have
7660          * been up as long as the smallest interval to continue on.
7661          *
7662          * Minimum idle times stored in idletime_tab and indexed by
7663          * idle_idx.  Start at upper end of array and work backwards.
7664          *
7665          * Once the index is found, set the initial idle time to the
7666          * first interval before the current ipfilter run time.
7667          */
7668         if (ifs->ifs_fr_ticks < idletime_tab[0])
7669                 return (0);
7670         idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7671         if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7672                 idletime = idletime_tab[idle_idx];
7673         } else {
7674                 while ((idle_idx > 0) &&
7675                     (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7676                         idle_idx--;
7677 
7678                 idletime = (ifs->ifs_fr_ticks /
7679                             idletime_tab[idle_idx]) *
7680                             idletime_tab[idle_idx];
7681         }
7682 
7683         while (idle_idx >= 0) {
7684                 /*
7685                  * Check to see if we need to delete more entries.
7686                  * If we do, start with appropriate timeout queue.
7687                  */
7688                 if (flushtype == NAT_FLUSH) {
7689                         if (NAT_TAB_WATER_LEVEL(ifs) <=
7690                             ifs->ifs_nat_flush_level_lo)
7691                                 break;
7692                 } else if (flushtype == STATE_FLUSH) {
7693                         if (ST_TAB_WATER_LEVEL(ifs) <=
7694                             ifs->ifs_state_flush_level_lo)
7695                                 break;
7696                 } else {
7697                         break;
7698                 }
7699 
7700                 removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7701 
7702                 /*
7703                  * Next, check the user defined queues.  But first, make
7704                  * certain that timeout queue deletions didn't do enough.
7705                  */
7706                 if (flushtype == NAT_FLUSH) {
7707                         if (NAT_TAB_WATER_LEVEL(ifs) <=
7708                             ifs->ifs_nat_flush_level_lo)
7709                                 break;
7710                 } else {
7711                         if (ST_TAB_WATER_LEVEL(ifs) <=
7712                             ifs->ifs_state_flush_level_lo)
7713                                 break;
7714                 }
7715                 ifqn = userqs;
7716                 while ((ifq = ifqn) != NULL) {
7717                         ifqn = ifq->ifq_next;
7718                         removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7719                 }
7720 
7721                 /*
7722                  * Adjust the granularity of idle time.
7723                  *
7724                  * If we reach an interval boundary, we need to
7725                  * either adjust the idle time accordingly or exit
7726                  * the loop altogether (if this is very last check).
7727                  */
7728                 idletime -= idletime_tab[idle_idx];
7729                 if (idletime < idletime_tab[idle_idx]) {
7730                         if (idle_idx != 0) {
7731                                 idletime = idletime_tab[idle_idx] -
7732                                 idletime_tab[idle_idx - 1];
7733                                 idle_idx--;
7734                         } else {
7735                                 break;
7736                         }
7737                 }
7738         }
7739 
7740         return (removed);
7741 }