ip_fw_pfil.c revision 227085
1238384Sjkim/*- 2238384Sjkim * Copyright (c) 2004 Andre Oppermann, Internet Business Solutions AG 3238384Sjkim * All rights reserved. 4238384Sjkim * 5238384Sjkim * Redistribution and use in source and binary forms, with or without 6238384Sjkim * modification, are permitted provided that the following conditions 7238384Sjkim * are met: 8238384Sjkim * 1. Redistributions of source code must retain the above copyright 9238384Sjkim * notice, this list of conditions and the following disclaimer. 10238384Sjkim * 2. Redistributions in binary form must reproduce the above copyright 11238384Sjkim * notice, this list of conditions and the following disclaimer in the 12238384Sjkim * documentation and/or other materials provided with the distribution. 13238384Sjkim * 14238384Sjkim * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15238384Sjkim * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16238384Sjkim * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17238384Sjkim * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18238384Sjkim * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19238384Sjkim * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20238384Sjkim * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21238384Sjkim * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22238384Sjkim * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23238384Sjkim * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24238384Sjkim * SUCH DAMAGE. 25238384Sjkim */ 26238384Sjkim 27238384Sjkim#include <sys/cdefs.h> 28238384Sjkim__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw_pfil.c 227085 2011-11-04 16:24:19Z bz $"); 29238384Sjkim 30238384Sjkim#include "opt_ipfw.h" 31238384Sjkim#include "opt_inet.h" 32238384Sjkim#include "opt_inet6.h" 33238384Sjkim#ifndef INET 34238384Sjkim#error IPFIREWALL requires INET. 35238384Sjkim#endif /* INET */ 36238384Sjkim 37238384Sjkim#include <sys/param.h> 38238384Sjkim#include <sys/systm.h> 39238384Sjkim#include <sys/malloc.h> 40238384Sjkim#include <sys/mbuf.h> 41238384Sjkim#include <sys/module.h> 42238384Sjkim#include <sys/kernel.h> 43238384Sjkim#include <sys/lock.h> 44238384Sjkim#include <sys/rwlock.h> 45246772Sjkim#include <sys/socket.h> 46238384Sjkim#include <sys/sysctl.h> 47238384Sjkim 48238384Sjkim#include <net/if.h> 49238384Sjkim#include <net/route.h> 50238384Sjkim#include <net/pfil.h> 51238384Sjkim#include <net/vnet.h> 52238384Sjkim 53238384Sjkim#include <netinet/in.h> 54238384Sjkim#include <netinet/in_systm.h> 55238384Sjkim#include <netinet/ip.h> 56238384Sjkim#include <netinet/ip_var.h> 57238384Sjkim#include <netinet/ip_fw.h> 58238384Sjkim#ifdef INET6 59238384Sjkim#include <netinet/ip6.h> 60238384Sjkim#include <netinet6/ip6_var.h> 61238384Sjkim#endif 62238384Sjkim#include <netinet/ipfw/ip_fw_private.h> 63238384Sjkim#include <netgraph/ng_ipfw.h> 64238384Sjkim 65238384Sjkim#include <machine/in_cksum.h> 66238384Sjkim 67238384Sjkimstatic VNET_DEFINE(int, fw_enable) = 1; 68238384Sjkim#define V_fw_enable VNET(fw_enable) 69238384Sjkim 70238384Sjkim#ifdef INET6 71238384Sjkimstatic VNET_DEFINE(int, fw6_enable) = 1; 72238384Sjkim#define V_fw6_enable VNET(fw6_enable) 73238384Sjkim#endif 74238384Sjkim 75238384Sjkimint ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 76238384Sjkim 77238384Sjkim/* Forward declarations. */ 78238384Sjkimstatic int ipfw_divert(struct mbuf **, int, struct ipfw_rule_ref *, int); 79238384Sjkim 80238384Sjkim#ifdef SYSCTL_NODE 81238384Sjkim 82238384SjkimSYSBEGIN(f1) 83238384Sjkim 84238384SjkimSYSCTL_DECL(_net_inet_ip_fw); 85238384SjkimSYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, enable, 86238384Sjkim CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_enable), 0, 87238384Sjkim ipfw_chg_hook, "I", "Enable ipfw"); 88238384Sjkim#ifdef INET6 89238384SjkimSYSCTL_DECL(_net_inet6_ip6_fw); 90238384SjkimSYSCTL_VNET_PROC(_net_inet6_ip6_fw, OID_AUTO, enable, 91238384Sjkim CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw6_enable), 0, 92238384Sjkim ipfw_chg_hook, "I", "Enable ipfw+6"); 93238384Sjkim#endif /* INET6 */ 94238384Sjkim 95238384SjkimSYSEND 96238384Sjkim 97238384Sjkim#endif /* SYSCTL_NODE */ 98238384Sjkim 99238384Sjkim/* 100238384Sjkim * The pfilter hook to pass packets to ipfw_chk and then to 101238384Sjkim * dummynet, divert, netgraph or other modules. 102238384Sjkim * The packet may be consumed. 103238384Sjkim */ 104238384Sjkimint 105238384Sjkimipfw_check_hook(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir, 106238384Sjkim struct inpcb *inp) 107238384Sjkim{ 108238384Sjkim struct ip_fw_args args; 109238384Sjkim struct m_tag *tag; 110238384Sjkim int ipfw; 111238384Sjkim int ret; 112238384Sjkim 113238384Sjkim /* all the processing now uses ip_len in net format */ 114238384Sjkim if (mtod(*m0, struct ip *)->ip_v == 4) 115238384Sjkim SET_NET_IPLEN(mtod(*m0, struct ip *)); 116238384Sjkim 117238384Sjkim /* convert dir to IPFW values */ 118238384Sjkim dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT; 119238384Sjkim bzero(&args, sizeof(args)); 120238384Sjkim 121238384Sjkimagain: 122238384Sjkim /* 123238384Sjkim * extract and remove the tag if present. If we are left 124238384Sjkim * with onepass, optimize the outgoing path. 125238384Sjkim */ 126238384Sjkim tag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL); 127238384Sjkim if (tag != NULL) { 128238384Sjkim args.rule = *((struct ipfw_rule_ref *)(tag+1)); 129238384Sjkim m_tag_delete(*m0, tag); 130238384Sjkim if (args.rule.info & IPFW_ONEPASS) { 131238384Sjkim if (mtod(*m0, struct ip *)->ip_v == 4) 132238384Sjkim SET_HOST_IPLEN(mtod(*m0, struct ip *)); 133238384Sjkim return (0); 134238384Sjkim } 135238384Sjkim } 136238384Sjkim 137238384Sjkim args.m = *m0; 138238384Sjkim args.oif = dir == DIR_OUT ? ifp : NULL; 139238384Sjkim args.inp = inp; 140238384Sjkim 141238384Sjkim ipfw = ipfw_chk(&args); 142238384Sjkim *m0 = args.m; 143238384Sjkim 144238384Sjkim KASSERT(*m0 != NULL || ipfw == IP_FW_DENY, ("%s: m0 is NULL", 145238384Sjkim __func__)); 146238384Sjkim 147238384Sjkim /* breaking out of the switch means drop */ 148238384Sjkim ret = 0; /* default return value for pass */ 149238384Sjkim switch (ipfw) { 150238384Sjkim case IP_FW_PASS: 151238384Sjkim /* next_hop may be set by ipfw_chk */ 152238384Sjkim if (args.next_hop == NULL && args.next_hop6 == NULL) 153238384Sjkim break; /* pass */ 154238384Sjkim#if !defined(IPFIREWALL_FORWARD) || (!defined(INET6) && !defined(INET)) 155238384Sjkim ret = EACCES; 156238384Sjkim#else 157238384Sjkim { 158238384Sjkim struct m_tag *fwd_tag; 159238384Sjkim size_t len; 160238384Sjkim 161238384Sjkim KASSERT(args.next_hop == NULL || args.next_hop6 == NULL, 162238384Sjkim ("%s: both next_hop=%p and next_hop6=%p not NULL", __func__, 163238384Sjkim args.next_hop, args.next_hop6)); 164238384Sjkim#ifdef INET6 165238384Sjkim if (args.next_hop6 != NULL) 166238384Sjkim len = sizeof(struct sockaddr_in6); 167238384Sjkim#endif 168238384Sjkim#ifdef INET 169238384Sjkim if (args.next_hop != NULL) 170238384Sjkim len = sizeof(struct sockaddr_in); 171238384Sjkim#endif 172238384Sjkim 173238384Sjkim /* Incoming packets should not be tagged so we do not 174238384Sjkim * m_tag_find. Outgoing packets may be tagged, so we 175238384Sjkim * reuse the tag if present. 176238384Sjkim */ 177238384Sjkim fwd_tag = (dir == DIR_IN) ? NULL : 178238384Sjkim m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 179238384Sjkim if (fwd_tag != NULL) { 180238384Sjkim m_tag_unlink(*m0, fwd_tag); 181238384Sjkim } else { 182238384Sjkim fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD, len, 183238384Sjkim M_NOWAIT); 184238384Sjkim if (fwd_tag == NULL) { 185238384Sjkim ret = EACCES; 186238384Sjkim break; /* i.e. drop */ 187238384Sjkim } 188238384Sjkim } 189238384Sjkim#ifdef INET6 190238384Sjkim if (args.next_hop6 != NULL) { 191238384Sjkim bcopy(args.next_hop6, (fwd_tag+1), len); 192238384Sjkim if (in6_localip(&args.next_hop6->sin6_addr)) 193238384Sjkim (*m0)->m_flags |= M_FASTFWD_OURS; 194238384Sjkim } 195238384Sjkim#endif 196238384Sjkim#ifdef INET 197279264Sdelphij if (args.next_hop != NULL) { 198279264Sdelphij bcopy(args.next_hop, (fwd_tag+1), len); 199238384Sjkim if (in_localip(args.next_hop->sin_addr)) 200238384Sjkim (*m0)->m_flags |= M_FASTFWD_OURS; 201238384Sjkim } 202238384Sjkim#endif 203238384Sjkim m_tag_prepend(*m0, fwd_tag); 204279264Sdelphij } 205238384Sjkim#endif /* IPFIREWALL_FORWARD */ 206279264Sdelphij break; 207238384Sjkim 208279264Sdelphij case IP_FW_DENY: 209238384Sjkim ret = EACCES; 210279264Sdelphij break; /* i.e. drop */ 211279264Sdelphij 212238384Sjkim case IP_FW_DUMMYNET: 213279264Sdelphij ret = EACCES; 214238384Sjkim if (ip_dn_io_ptr == NULL) 215279264Sdelphij break; /* i.e. drop */ 216238384Sjkim if (mtod(*m0, struct ip *)->ip_v == 4) 217238384Sjkim ret = ip_dn_io_ptr(m0, dir, &args); 218238384Sjkim else if (mtod(*m0, struct ip *)->ip_v == 6) 219238384Sjkim ret = ip_dn_io_ptr(m0, dir | PROTO_IPV6, &args); 220238384Sjkim else 221238384Sjkim break; /* drop it */ 222238384Sjkim /* 223238384Sjkim * XXX should read the return value. 224238384Sjkim * dummynet normally eats the packet and sets *m0=NULL 225238384Sjkim * unless the packet can be sent immediately. In this 226238384Sjkim * case args is updated and we should re-run the 227238384Sjkim * check without clearing args. 228238384Sjkim */ 229238384Sjkim if (*m0 != NULL) 230238384Sjkim goto again; 231238384Sjkim break; 232238384Sjkim 233238384Sjkim case IP_FW_TEE: 234238384Sjkim case IP_FW_DIVERT: 235238384Sjkim if (ip_divert_ptr == NULL) { 236238384Sjkim ret = EACCES; 237238384Sjkim break; /* i.e. drop */ 238238384Sjkim } 239238384Sjkim ret = ipfw_divert(m0, dir, &args.rule, 240238384Sjkim (ipfw == IP_FW_TEE) ? 1 : 0); 241238384Sjkim /* continue processing for the original packet (tee). */ 242238384Sjkim if (*m0) 243238384Sjkim goto again; 244238384Sjkim break; 245238384Sjkim 246238384Sjkim case IP_FW_NGTEE: 247238384Sjkim case IP_FW_NETGRAPH: 248238384Sjkim if (ng_ipfw_input_p == NULL) { 249238384Sjkim ret = EACCES; 250238384Sjkim break; /* i.e. drop */ 251238384Sjkim } 252238384Sjkim ret = ng_ipfw_input_p(m0, dir, &args, 253238384Sjkim (ipfw == IP_FW_NGTEE) ? 1 : 0); 254238384Sjkim if (ipfw == IP_FW_NGTEE) /* ignore errors for NGTEE */ 255 goto again; /* continue with packet */ 256 break; 257 258 case IP_FW_NAT: 259 /* honor one-pass in case of successful nat */ 260 if (V_fw_one_pass) 261 break; /* ret is already 0 */ 262 goto again; 263 264 case IP_FW_REASS: 265 goto again; /* continue with packet */ 266 267 default: 268 KASSERT(0, ("%s: unknown retval", __func__)); 269 } 270 271 if (ret != 0) { 272 if (*m0) 273 FREE_PKT(*m0); 274 *m0 = NULL; 275 } 276 if (*m0 && mtod(*m0, struct ip *)->ip_v == 4) 277 SET_HOST_IPLEN(mtod(*m0, struct ip *)); 278 return ret; 279} 280 281/* do the divert, return 1 on error 0 on success */ 282static int 283ipfw_divert(struct mbuf **m0, int incoming, struct ipfw_rule_ref *rule, 284 int tee) 285{ 286 /* 287 * ipfw_chk() has already tagged the packet with the divert tag. 288 * If tee is set, copy packet and return original. 289 * If not tee, consume packet and send it to divert socket. 290 */ 291 struct mbuf *clone; 292 struct ip *ip = mtod(*m0, struct ip *); 293 struct m_tag *tag; 294 295 /* Cloning needed for tee? */ 296 if (tee == 0) { 297 clone = *m0; /* use the original mbuf */ 298 *m0 = NULL; 299 } else { 300 clone = m_dup(*m0, M_DONTWAIT); 301 /* If we cannot duplicate the mbuf, we sacrifice the divert 302 * chain and continue with the tee-ed packet. 303 */ 304 if (clone == NULL) 305 return 1; 306 } 307 308 /* 309 * Divert listeners can normally handle non-fragmented packets, 310 * but we can only reass in the non-tee case. 311 * This means that listeners on a tee rule may get fragments, 312 * and have to live with that. 313 * Note that we now have the 'reass' ipfw option so if we care 314 * we can do it before a 'tee'. 315 */ 316 if (!tee) switch (ip->ip_v) { 317 case IPVERSION: 318 if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) { 319 int hlen; 320 struct mbuf *reass; 321 322 SET_HOST_IPLEN(ip); /* ip_reass wants host order */ 323 reass = ip_reass(clone); /* Reassemble packet. */ 324 if (reass == NULL) 325 return 0; /* not an error */ 326 /* if reass = NULL then it was consumed by ip_reass */ 327 /* 328 * IP header checksum fixup after reassembly and leave header 329 * in network byte order. 330 */ 331 ip = mtod(reass, struct ip *); 332 hlen = ip->ip_hl << 2; 333 SET_NET_IPLEN(ip); 334 ip->ip_sum = 0; 335 if (hlen == sizeof(struct ip)) 336 ip->ip_sum = in_cksum_hdr(ip); 337 else 338 ip->ip_sum = in_cksum(reass, hlen); 339 clone = reass; 340 } 341 break; 342#ifdef INET6 343 case IPV6_VERSION >> 4: 344 { 345 struct ip6_hdr *const ip6 = mtod(clone, struct ip6_hdr *); 346 347 if (ip6->ip6_nxt == IPPROTO_FRAGMENT) { 348 int nxt, off; 349 350 off = sizeof(struct ip6_hdr); 351 nxt = frag6_input(&clone, &off, 0); 352 if (nxt == IPPROTO_DONE) 353 return (0); 354 } 355 break; 356 } 357#endif 358 } 359 360 /* attach a tag to the packet with the reinject info */ 361 tag = m_tag_alloc(MTAG_IPFW_RULE, 0, 362 sizeof(struct ipfw_rule_ref), M_NOWAIT); 363 if (tag == NULL) { 364 FREE_PKT(clone); 365 return 1; 366 } 367 *((struct ipfw_rule_ref *)(tag+1)) = *rule; 368 m_tag_prepend(clone, tag); 369 370 /* Do the dirty job... */ 371 ip_divert_ptr(clone, incoming); 372 return 0; 373} 374 375/* 376 * attach or detach hooks for a given protocol family 377 */ 378static int 379ipfw_hook(int onoff, int pf) 380{ 381 struct pfil_head *pfh; 382 383 pfh = pfil_head_get(PFIL_TYPE_AF, pf); 384 if (pfh == NULL) 385 return ENOENT; 386 387 (void) (onoff ? pfil_add_hook : pfil_remove_hook) 388 (ipfw_check_hook, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh); 389 390 return 0; 391} 392 393int 394ipfw_attach_hooks(int arg) 395{ 396 int error = 0; 397 398 if (arg == 0) /* detach */ 399 ipfw_hook(0, AF_INET); 400 else if (V_fw_enable && ipfw_hook(1, AF_INET) != 0) { 401 error = ENOENT; /* see ip_fw_pfil.c::ipfw_hook() */ 402 printf("ipfw_hook() error\n"); 403 } 404#ifdef INET6 405 if (arg == 0) /* detach */ 406 ipfw_hook(0, AF_INET6); 407 else if (V_fw6_enable && ipfw_hook(1, AF_INET6) != 0) { 408 error = ENOENT; 409 printf("ipfw6_hook() error\n"); 410 } 411#endif 412 return error; 413} 414 415int 416ipfw_chg_hook(SYSCTL_HANDLER_ARGS) 417{ 418 int enable; 419 int oldenable; 420 int error; 421 int af; 422 423 if (arg1 == &VNET_NAME(fw_enable)) { 424 enable = V_fw_enable; 425 af = AF_INET; 426 } 427#ifdef INET6 428 else if (arg1 == &VNET_NAME(fw6_enable)) { 429 enable = V_fw6_enable; 430 af = AF_INET6; 431 } 432#endif 433 else 434 return (EINVAL); 435 436 oldenable = enable; 437 438 error = sysctl_handle_int(oidp, &enable, 0, req); 439 440 if (error) 441 return (error); 442 443 enable = (enable) ? 1 : 0; 444 445 if (enable == oldenable) 446 return (0); 447 448 error = ipfw_hook(enable, af); 449 if (error) 450 return (error); 451 if (af == AF_INET) 452 V_fw_enable = enable; 453#ifdef INET6 454 else if (af == AF_INET6) 455 V_fw6_enable = enable; 456#endif 457 458 return (0); 459} 460/* end of file */ 461