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