xref: /freebsd/sys/netpfil/ipfw/ip_fw_pfil.c (revision feabce61dcd6e6b2be679e6919a0bda5ab27b19a)
1 /*-
2  * Copyright (c) 2004 Andre Oppermann, Internet Business Solutions AG
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29 
30 #include "opt_ipfw.h"
31 #include "opt_inet.h"
32 #include "opt_inet6.h"
33 #ifndef INET
34 #error IPFIREWALL requires INET.
35 #endif /* INET */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/module.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/rwlock.h>
45 #include <sys/socket.h>
46 #include <sys/sysctl.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 #include <net/ethernet.h>
51 #include <net/pfil.h>
52 #include <net/vnet.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/ip_fw.h>
59 #ifdef INET6
60 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
62 #include <netinet6/scope6_var.h>
63 #endif
64 
65 #include <netgraph/ng_ipfw.h>
66 
67 #include <netpfil/ipfw/ip_fw_private.h>
68 
69 #include <machine/in_cksum.h>
70 
71 static VNET_DEFINE(int, fw_enable) = 1;
72 #define V_fw_enable	VNET(fw_enable)
73 
74 #ifdef INET6
75 static VNET_DEFINE(int, fw6_enable) = 1;
76 #define V_fw6_enable	VNET(fw6_enable)
77 #endif
78 
79 static VNET_DEFINE(int, fwlink_enable) = 0;
80 #define V_fwlink_enable	VNET(fwlink_enable)
81 
82 int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
83 
84 /* Forward declarations. */
85 static int ipfw_divert(struct mbuf **, int, struct ipfw_rule_ref *, int);
86 int ipfw_check_packet(void *, struct mbuf **, struct ifnet *, int,
87 	struct inpcb *);
88 int ipfw_check_frame(void *, struct mbuf **, struct ifnet *, int,
89 	struct inpcb *);
90 
91 #ifdef SYSCTL_NODE
92 
93 SYSBEGIN(f1)
94 
95 SYSCTL_DECL(_net_inet_ip_fw);
96 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
97     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
98     &VNET_NAME(fw_enable), 0, ipfw_chg_hook, "I", "Enable ipfw");
99 #ifdef INET6
100 SYSCTL_DECL(_net_inet6_ip6_fw);
101 SYSCTL_PROC(_net_inet6_ip6_fw, OID_AUTO, enable,
102     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
103     &VNET_NAME(fw6_enable), 0, ipfw_chg_hook, "I", "Enable ipfw+6");
104 #endif /* INET6 */
105 
106 SYSCTL_DECL(_net_link_ether);
107 SYSCTL_PROC(_net_link_ether, OID_AUTO, ipfw,
108     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
109     &VNET_NAME(fwlink_enable), 0, ipfw_chg_hook, "I",
110     "Pass ether pkts through firewall");
111 
112 SYSEND
113 
114 #endif /* SYSCTL_NODE */
115 
116 /*
117  * The pfilter hook to pass packets to ipfw_chk and then to
118  * dummynet, divert, netgraph or other modules.
119  * The packet may be consumed.
120  */
121 int
122 ipfw_check_packet(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir,
123     struct inpcb *inp)
124 {
125 	struct ip_fw_args args;
126 	struct m_tag *tag;
127 	int ipfw;
128 	int ret;
129 
130 	/* convert dir to IPFW values */
131 	dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT;
132 	bzero(&args, sizeof(args));
133 
134 again:
135 	/*
136 	 * extract and remove the tag if present. If we are left
137 	 * with onepass, optimize the outgoing path.
138 	 */
139 	tag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
140 	if (tag != NULL) {
141 		args.rule = *((struct ipfw_rule_ref *)(tag+1));
142 		m_tag_delete(*m0, tag);
143 		if (args.rule.info & IPFW_ONEPASS)
144 			return (0);
145 	}
146 
147 	args.m = *m0;
148 	args.oif = dir == DIR_OUT ? ifp : NULL;
149 	args.inp = inp;
150 
151 	ipfw = ipfw_chk(&args);
152 	*m0 = args.m;
153 
154 	KASSERT(*m0 != NULL || ipfw == IP_FW_DENY, ("%s: m0 is NULL",
155 	    __func__));
156 
157 	/* breaking out of the switch means drop */
158 	ret = 0;	/* default return value for pass */
159 	switch (ipfw) {
160 	case IP_FW_PASS:
161 		/* next_hop may be set by ipfw_chk */
162 		if (args.next_hop == NULL && args.next_hop6 == NULL)
163 			break; /* pass */
164 #if (!defined(INET6) && !defined(INET))
165 		ret = EACCES;
166 #else
167 	    {
168 		struct m_tag *fwd_tag;
169 		size_t len;
170 
171 		KASSERT(args.next_hop == NULL || args.next_hop6 == NULL,
172 		    ("%s: both next_hop=%p and next_hop6=%p not NULL", __func__,
173 		     args.next_hop, args.next_hop6));
174 #ifdef INET6
175 		if (args.next_hop6 != NULL)
176 			len = sizeof(struct sockaddr_in6);
177 #endif
178 #ifdef INET
179 		if (args.next_hop != NULL)
180 			len = sizeof(struct sockaddr_in);
181 #endif
182 
183 		/* Incoming packets should not be tagged so we do not
184 		 * m_tag_find. Outgoing packets may be tagged, so we
185 		 * reuse the tag if present.
186 		 */
187 		fwd_tag = (dir == DIR_IN) ? NULL :
188 			m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
189 		if (fwd_tag != NULL) {
190 			m_tag_unlink(*m0, fwd_tag);
191 		} else {
192 			fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD, len,
193 			    M_NOWAIT);
194 			if (fwd_tag == NULL) {
195 				ret = EACCES;
196 				break; /* i.e. drop */
197 			}
198 		}
199 #ifdef INET6
200 		if (args.next_hop6 != NULL) {
201 			struct sockaddr_in6 *sa6;
202 
203 			sa6 = (struct sockaddr_in6 *)(fwd_tag + 1);
204 			bcopy(args.next_hop6, sa6, len);
205 			/*
206 			 * If nh6 address is link-local we should convert
207 			 * it to kernel internal form before doing any
208 			 * comparisons.
209 			 */
210 			if (sa6_embedscope(sa6, V_ip6_use_defzone) != 0) {
211 				ret = EACCES;
212 				break;
213 			}
214 			if (in6_localip(&sa6->sin6_addr))
215 				(*m0)->m_flags |= M_FASTFWD_OURS;
216 			(*m0)->m_flags |= M_IP6_NEXTHOP;
217 		}
218 #endif
219 #ifdef INET
220 		if (args.next_hop != NULL) {
221 			bcopy(args.next_hop, (fwd_tag+1), len);
222 			if (in_localip(args.next_hop->sin_addr))
223 				(*m0)->m_flags |= M_FASTFWD_OURS;
224 			(*m0)->m_flags |= M_IP_NEXTHOP;
225 		}
226 #endif
227 		m_tag_prepend(*m0, fwd_tag);
228 	    }
229 #endif /* INET || INET6 */
230 		break;
231 
232 	case IP_FW_DENY:
233 		ret = EACCES;
234 		break; /* i.e. drop */
235 
236 	case IP_FW_DUMMYNET:
237 		ret = EACCES;
238 		if (ip_dn_io_ptr == NULL)
239 			break; /* i.e. drop */
240 		if (mtod(*m0, struct ip *)->ip_v == 4)
241 			ret = ip_dn_io_ptr(m0, dir, &args);
242 		else if (mtod(*m0, struct ip *)->ip_v == 6)
243 			ret = ip_dn_io_ptr(m0, dir | PROTO_IPV6, &args);
244 		else
245 			break; /* drop it */
246 		/*
247 		 * XXX should read the return value.
248 		 * dummynet normally eats the packet and sets *m0=NULL
249 		 * unless the packet can be sent immediately. In this
250 		 * case args is updated and we should re-run the
251 		 * check without clearing args.
252 		 */
253 		if (*m0 != NULL)
254 			goto again;
255 		break;
256 
257 	case IP_FW_TEE:
258 	case IP_FW_DIVERT:
259 		if (ip_divert_ptr == NULL) {
260 			ret = EACCES;
261 			break; /* i.e. drop */
262 		}
263 		ret = ipfw_divert(m0, dir, &args.rule,
264 			(ipfw == IP_FW_TEE) ? 1 : 0);
265 		/* continue processing for the original packet (tee). */
266 		if (*m0)
267 			goto again;
268 		break;
269 
270 	case IP_FW_NGTEE:
271 	case IP_FW_NETGRAPH:
272 		if (ng_ipfw_input_p == NULL) {
273 			ret = EACCES;
274 			break; /* i.e. drop */
275 		}
276 		ret = ng_ipfw_input_p(m0, dir, &args,
277 			(ipfw == IP_FW_NGTEE) ? 1 : 0);
278 		if (ipfw == IP_FW_NGTEE) /* ignore errors for NGTEE */
279 			goto again;	/* continue with packet */
280 		break;
281 
282 	case IP_FW_NAT:
283 		/* honor one-pass in case of successful nat */
284 		if (V_fw_one_pass)
285 			break; /* ret is already 0 */
286 		goto again;
287 
288 	case IP_FW_REASS:
289 		goto again;		/* continue with packet */
290 
291 	default:
292 		KASSERT(0, ("%s: unknown retval", __func__));
293 	}
294 
295 	if (ret != 0) {
296 		if (*m0)
297 			FREE_PKT(*m0);
298 		*m0 = NULL;
299 	}
300 
301 	return ret;
302 }
303 
304 /*
305  * ipfw processing for ethernet packets (in and out).
306  */
307 int
308 ipfw_check_frame(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir,
309     struct inpcb *inp)
310 {
311 	struct ether_header *eh;
312 	struct ether_header save_eh;
313 	struct mbuf *m;
314 	int i, ret;
315 	struct ip_fw_args args;
316 	struct m_tag *mtag;
317 
318 	/* fetch start point from rule, if any.  remove the tag if present. */
319 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
320 	if (mtag == NULL) {
321 		args.rule.slot = 0;
322 	} else {
323 		args.rule = *((struct ipfw_rule_ref *)(mtag+1));
324 		m_tag_delete(*m0, mtag);
325 		if (args.rule.info & IPFW_ONEPASS)
326 			return (0);
327 	}
328 
329 	/* I need some amt of data to be contiguous */
330 	m = *m0;
331 	i = min(m->m_pkthdr.len, max_protohdr);
332 	if (m->m_len < i) {
333 		m = m_pullup(m, i);
334 		if (m == NULL) {
335 			*m0 = m;
336 			return (0);
337 		}
338 	}
339 	eh = mtod(m, struct ether_header *);
340 	save_eh = *eh;			/* save copy for restore below */
341 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
342 
343 	args.m = m;		/* the packet we are looking at		*/
344 	args.oif = dir == PFIL_OUT ? ifp: NULL;	/* destination, if any	*/
345 	args.next_hop = NULL;	/* we do not support forward yet	*/
346 	args.next_hop6 = NULL;	/* we do not support forward yet	*/
347 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
348 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
349 	i = ipfw_chk(&args);
350 	m = args.m;
351 	if (m != NULL) {
352 		/*
353 		 * Restore Ethernet header, as needed, in case the
354 		 * mbuf chain was replaced by ipfw.
355 		 */
356 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
357 		if (m == NULL) {
358 			*m0 = NULL;
359 			return (0);
360 		}
361 		if (eh != mtod(m, struct ether_header *))
362 			bcopy(&save_eh, mtod(m, struct ether_header *),
363 				ETHER_HDR_LEN);
364 	}
365 	*m0 = m;
366 
367 	ret = 0;
368 	/* Check result of ipfw_chk() */
369 	switch (i) {
370 	case IP_FW_PASS:
371 		break;
372 
373 	case IP_FW_DENY:
374 		ret = EACCES;
375 		break; /* i.e. drop */
376 
377 	case IP_FW_DUMMYNET:
378 		ret = EACCES;
379 
380 		if (ip_dn_io_ptr == NULL)
381 			break; /* i.e. drop */
382 
383 		*m0 = NULL;
384 		dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT;
385 		ip_dn_io_ptr(&m, dir | PROTO_LAYER2, &args);
386 		return 0;
387 
388 	default:
389 		KASSERT(0, ("%s: unknown retval", __func__));
390 	}
391 
392 	if (ret != 0) {
393 		if (*m0)
394 			FREE_PKT(*m0);
395 		*m0 = NULL;
396 	}
397 
398 	return ret;
399 }
400 
401 /* do the divert, return 1 on error 0 on success */
402 static int
403 ipfw_divert(struct mbuf **m0, int incoming, struct ipfw_rule_ref *rule,
404 	int tee)
405 {
406 	/*
407 	 * ipfw_chk() has already tagged the packet with the divert tag.
408 	 * If tee is set, copy packet and return original.
409 	 * If not tee, consume packet and send it to divert socket.
410 	 */
411 	struct mbuf *clone;
412 	struct ip *ip = mtod(*m0, struct ip *);
413 	struct m_tag *tag;
414 
415 	/* Cloning needed for tee? */
416 	if (tee == 0) {
417 		clone = *m0;	/* use the original mbuf */
418 		*m0 = NULL;
419 	} else {
420 		clone = m_dup(*m0, M_NOWAIT);
421 		/* If we cannot duplicate the mbuf, we sacrifice the divert
422 		 * chain and continue with the tee-ed packet.
423 		 */
424 		if (clone == NULL)
425 			return 1;
426 	}
427 
428 	/*
429 	 * Divert listeners can normally handle non-fragmented packets,
430 	 * but we can only reass in the non-tee case.
431 	 * This means that listeners on a tee rule may get fragments,
432 	 * and have to live with that.
433 	 * Note that we now have the 'reass' ipfw option so if we care
434 	 * we can do it before a 'tee'.
435 	 */
436 	if (!tee) switch (ip->ip_v) {
437 	case IPVERSION:
438 	    if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
439 		int hlen;
440 		struct mbuf *reass;
441 
442 		reass = ip_reass(clone); /* Reassemble packet. */
443 		if (reass == NULL)
444 			return 0; /* not an error */
445 		/* if reass = NULL then it was consumed by ip_reass */
446 		/*
447 		 * IP header checksum fixup after reassembly and leave header
448 		 * in network byte order.
449 		 */
450 		ip = mtod(reass, struct ip *);
451 		hlen = ip->ip_hl << 2;
452 		ip->ip_sum = 0;
453 		if (hlen == sizeof(struct ip))
454 			ip->ip_sum = in_cksum_hdr(ip);
455 		else
456 			ip->ip_sum = in_cksum(reass, hlen);
457 		clone = reass;
458 	    }
459 	    break;
460 #ifdef INET6
461 	case IPV6_VERSION >> 4:
462 	    {
463 	    struct ip6_hdr *const ip6 = mtod(clone, struct ip6_hdr *);
464 
465 		if (ip6->ip6_nxt == IPPROTO_FRAGMENT) {
466 			int nxt, off;
467 
468 			off = sizeof(struct ip6_hdr);
469 			nxt = frag6_input(&clone, &off, 0);
470 			if (nxt == IPPROTO_DONE)
471 				return (0);
472 		}
473 		break;
474 	    }
475 #endif
476 	}
477 
478 	/* attach a tag to the packet with the reinject info */
479 	tag = m_tag_alloc(MTAG_IPFW_RULE, 0,
480 		    sizeof(struct ipfw_rule_ref), M_NOWAIT);
481 	if (tag == NULL) {
482 		FREE_PKT(clone);
483 		return 1;
484 	}
485 	*((struct ipfw_rule_ref *)(tag+1)) = *rule;
486 	m_tag_prepend(clone, tag);
487 
488 	/* Do the dirty job... */
489 	ip_divert_ptr(clone, incoming);
490 	return 0;
491 }
492 
493 /*
494  * attach or detach hooks for a given protocol family
495  */
496 static int
497 ipfw_hook(int onoff, int pf)
498 {
499 	struct pfil_head *pfh;
500 	pfil_func_t hook_func;
501 
502 	pfh = pfil_head_get(PFIL_TYPE_AF, pf);
503 	if (pfh == NULL)
504 		return ENOENT;
505 
506 	hook_func = (pf == AF_LINK) ? ipfw_check_frame : ipfw_check_packet;
507 
508 	(void) (onoff ? pfil_add_hook : pfil_remove_hook)
509 	    (hook_func, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh);
510 
511 	return 0;
512 }
513 
514 int
515 ipfw_attach_hooks(int arg)
516 {
517 	int error = 0;
518 
519 	if (arg == 0) /* detach */
520 		ipfw_hook(0, AF_INET);
521 	else if (V_fw_enable && ipfw_hook(1, AF_INET) != 0) {
522                 error = ENOENT; /* see ip_fw_pfil.c::ipfw_hook() */
523                 printf("ipfw_hook() error\n");
524         }
525 #ifdef INET6
526 	if (arg == 0) /* detach */
527 		ipfw_hook(0, AF_INET6);
528 	else if (V_fw6_enable && ipfw_hook(1, AF_INET6) != 0) {
529                 error = ENOENT;
530                 printf("ipfw6_hook() error\n");
531         }
532 #endif
533 	if (arg == 0) /* detach */
534 		ipfw_hook(0, AF_LINK);
535 	else if (V_fwlink_enable && ipfw_hook(1, AF_LINK) != 0) {
536                 error = ENOENT;
537                 printf("ipfw_link_hook() error\n");
538         }
539 	return error;
540 }
541 
542 int
543 ipfw_chg_hook(SYSCTL_HANDLER_ARGS)
544 {
545 	int newval;
546 	int error;
547 	int af;
548 
549 	if (arg1 == &V_fw_enable)
550 		af = AF_INET;
551 #ifdef INET6
552 	else if (arg1 == &V_fw6_enable)
553 		af = AF_INET6;
554 #endif
555 	else if (arg1 == &V_fwlink_enable)
556 		af = AF_LINK;
557 	else
558 		return (EINVAL);
559 
560 	newval = *(int *)arg1;
561 	/* Handle sysctl change */
562 	error = sysctl_handle_int(oidp, &newval, 0, req);
563 
564 	if (error)
565 		return (error);
566 
567 	/* Formalize new value */
568 	newval = (newval) ? 1 : 0;
569 
570 	if (*(int *)arg1 == newval)
571 		return (0);
572 
573 	error = ipfw_hook(newval, af);
574 	if (error)
575 		return (error);
576 	*(int *)arg1 = newval;
577 
578 	return (0);
579 }
580 /* end of file */
581