xref: /freebsd/sys/netpfil/ipfw/ip_fw_pfil.c (revision f5147e312f43a9050468de539aeafa072caa1a60)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2004 Andre Oppermann, Internet Business Solutions AG
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "opt_ipfw.h"
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 #ifndef INET
36 #error IPFIREWALL requires INET.
37 #endif /* INET */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/module.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/rwlock.h>
47 #include <sys/socket.h>
48 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/route.h>
52 #include <net/ethernet.h>
53 #include <net/pfil.h>
54 #include <net/vnet.h>
55 
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/ip_fw.h>
61 #ifdef INET6
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #include <netinet6/scope6_var.h>
65 #endif
66 
67 #include <netgraph/ng_ipfw.h>
68 
69 #include <netpfil/ipfw/ip_fw_private.h>
70 
71 #include <machine/in_cksum.h>
72 
73 static VNET_DEFINE(int, fw_enable) = 1;
74 #define V_fw_enable	VNET(fw_enable)
75 
76 #ifdef INET6
77 static VNET_DEFINE(int, fw6_enable) = 1;
78 #define V_fw6_enable	VNET(fw6_enable)
79 #endif
80 
81 static VNET_DEFINE(int, fwlink_enable) = 0;
82 #define V_fwlink_enable	VNET(fwlink_enable)
83 
84 int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
85 
86 /* Forward declarations. */
87 static int ipfw_divert(struct mbuf **, int, struct ipfw_rule_ref *, int);
88 int ipfw_check_packet(void *, struct mbuf **, struct ifnet *, int,
89 	struct inpcb *);
90 int ipfw_check_frame(void *, struct mbuf **, struct ifnet *, int,
91 	struct inpcb *);
92 
93 #ifdef SYSCTL_NODE
94 
95 SYSBEGIN(f1)
96 
97 SYSCTL_DECL(_net_inet_ip_fw);
98 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
99     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
100     &VNET_NAME(fw_enable), 0, ipfw_chg_hook, "I", "Enable ipfw");
101 #ifdef INET6
102 SYSCTL_DECL(_net_inet6_ip6_fw);
103 SYSCTL_PROC(_net_inet6_ip6_fw, OID_AUTO, enable,
104     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
105     &VNET_NAME(fw6_enable), 0, ipfw_chg_hook, "I", "Enable ipfw+6");
106 #endif /* INET6 */
107 
108 SYSCTL_DECL(_net_link_ether);
109 SYSCTL_PROC(_net_link_ether, OID_AUTO, ipfw,
110     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
111     &VNET_NAME(fwlink_enable), 0, ipfw_chg_hook, "I",
112     "Pass ether pkts through firewall");
113 
114 SYSEND
115 
116 #endif /* SYSCTL_NODE */
117 
118 /*
119  * The pfilter hook to pass packets to ipfw_chk and then to
120  * dummynet, divert, netgraph or other modules.
121  * The packet may be consumed.
122  */
123 int
124 ipfw_check_packet(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir,
125     struct inpcb *inp)
126 {
127 	struct ip_fw_args args;
128 	struct m_tag *tag;
129 	int ipfw;
130 	int ret;
131 
132 	/* convert dir to IPFW values */
133 	dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT;
134 	bzero(&args, sizeof(args));
135 
136 again:
137 	/*
138 	 * extract and remove the tag if present. If we are left
139 	 * with onepass, optimize the outgoing path.
140 	 */
141 	tag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
142 	if (tag != NULL) {
143 		args.rule = *((struct ipfw_rule_ref *)(tag+1));
144 		m_tag_delete(*m0, tag);
145 		if (args.rule.info & IPFW_ONEPASS)
146 			return (0);
147 	}
148 
149 	args.m = *m0;
150 	args.oif = dir == DIR_OUT ? ifp : NULL;
151 	args.inp = inp;
152 
153 	ipfw = ipfw_chk(&args);
154 	*m0 = args.m;
155 
156 	KASSERT(*m0 != NULL || ipfw == IP_FW_DENY, ("%s: m0 is NULL",
157 	    __func__));
158 
159 	/* breaking out of the switch means drop */
160 	ret = 0;	/* default return value for pass */
161 	switch (ipfw) {
162 	case IP_FW_PASS:
163 		/* next_hop may be set by ipfw_chk */
164 		if (args.next_hop == NULL && args.next_hop6 == NULL)
165 			break; /* pass */
166 #if (!defined(INET6) && !defined(INET))
167 		ret = EACCES;
168 #else
169 	    {
170 		struct m_tag *fwd_tag;
171 		size_t len;
172 
173 		KASSERT(args.next_hop == NULL || args.next_hop6 == NULL,
174 		    ("%s: both next_hop=%p and next_hop6=%p not NULL", __func__,
175 		     args.next_hop, args.next_hop6));
176 #ifdef INET6
177 		if (args.next_hop6 != NULL)
178 			len = sizeof(struct sockaddr_in6);
179 #endif
180 #ifdef INET
181 		if (args.next_hop != NULL)
182 			len = sizeof(struct sockaddr_in);
183 #endif
184 
185 		/* Incoming packets should not be tagged so we do not
186 		 * m_tag_find. Outgoing packets may be tagged, so we
187 		 * reuse the tag if present.
188 		 */
189 		fwd_tag = (dir == DIR_IN) ? NULL :
190 			m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
191 		if (fwd_tag != NULL) {
192 			m_tag_unlink(*m0, fwd_tag);
193 		} else {
194 			fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD, len,
195 			    M_NOWAIT);
196 			if (fwd_tag == NULL) {
197 				ret = EACCES;
198 				break; /* i.e. drop */
199 			}
200 		}
201 #ifdef INET6
202 		if (args.next_hop6 != NULL) {
203 			struct sockaddr_in6 *sa6;
204 
205 			sa6 = (struct sockaddr_in6 *)(fwd_tag + 1);
206 			bcopy(args.next_hop6, sa6, len);
207 			/*
208 			 * If nh6 address is link-local we should convert
209 			 * it to kernel internal form before doing any
210 			 * comparisons.
211 			 */
212 			if (sa6_embedscope(sa6, V_ip6_use_defzone) != 0) {
213 				ret = EACCES;
214 				break;
215 			}
216 			if (in6_localip(&sa6->sin6_addr))
217 				(*m0)->m_flags |= M_FASTFWD_OURS;
218 			(*m0)->m_flags |= M_IP6_NEXTHOP;
219 		}
220 #endif
221 #ifdef INET
222 		if (args.next_hop != NULL) {
223 			bcopy(args.next_hop, (fwd_tag+1), len);
224 			if (in_localip(args.next_hop->sin_addr))
225 				(*m0)->m_flags |= M_FASTFWD_OURS;
226 			(*m0)->m_flags |= M_IP_NEXTHOP;
227 		}
228 #endif
229 		m_tag_prepend(*m0, fwd_tag);
230 	    }
231 #endif /* INET || INET6 */
232 		break;
233 
234 	case IP_FW_DENY:
235 		ret = EACCES;
236 		break; /* i.e. drop */
237 
238 	case IP_FW_DUMMYNET:
239 		ret = EACCES;
240 		if (ip_dn_io_ptr == NULL)
241 			break; /* i.e. drop */
242 		if (mtod(*m0, struct ip *)->ip_v == 4)
243 			ret = ip_dn_io_ptr(m0, dir, &args);
244 		else if (mtod(*m0, struct ip *)->ip_v == 6)
245 			ret = ip_dn_io_ptr(m0, dir | PROTO_IPV6, &args);
246 		else
247 			break; /* drop it */
248 		/*
249 		 * XXX should read the return value.
250 		 * dummynet normally eats the packet and sets *m0=NULL
251 		 * unless the packet can be sent immediately. In this
252 		 * case args is updated and we should re-run the
253 		 * check without clearing args.
254 		 */
255 		if (*m0 != NULL)
256 			goto again;
257 		break;
258 
259 	case IP_FW_TEE:
260 	case IP_FW_DIVERT:
261 		if (ip_divert_ptr == NULL) {
262 			ret = EACCES;
263 			break; /* i.e. drop */
264 		}
265 		ret = ipfw_divert(m0, dir, &args.rule,
266 			(ipfw == IP_FW_TEE) ? 1 : 0);
267 		/* continue processing for the original packet (tee). */
268 		if (*m0)
269 			goto again;
270 		break;
271 
272 	case IP_FW_NGTEE:
273 	case IP_FW_NETGRAPH:
274 		if (ng_ipfw_input_p == NULL) {
275 			ret = EACCES;
276 			break; /* i.e. drop */
277 		}
278 		ret = ng_ipfw_input_p(m0, dir, &args,
279 			(ipfw == IP_FW_NGTEE) ? 1 : 0);
280 		if (ipfw == IP_FW_NGTEE) /* ignore errors for NGTEE */
281 			goto again;	/* continue with packet */
282 		break;
283 
284 	case IP_FW_NAT:
285 		/* honor one-pass in case of successful nat */
286 		if (V_fw_one_pass)
287 			break; /* ret is already 0 */
288 		goto again;
289 
290 	case IP_FW_REASS:
291 		goto again;		/* continue with packet */
292 
293 	default:
294 		KASSERT(0, ("%s: unknown retval", __func__));
295 	}
296 
297 	if (ret != 0) {
298 		if (*m0)
299 			FREE_PKT(*m0);
300 		*m0 = NULL;
301 	}
302 
303 	return ret;
304 }
305 
306 /*
307  * ipfw processing for ethernet packets (in and out).
308  */
309 int
310 ipfw_check_frame(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir,
311     struct inpcb *inp)
312 {
313 	struct ether_header *eh;
314 	struct ether_header save_eh;
315 	struct mbuf *m;
316 	int i, ret;
317 	struct ip_fw_args args;
318 	struct m_tag *mtag;
319 
320 	/* fetch start point from rule, if any.  remove the tag if present. */
321 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
322 	if (mtag == NULL) {
323 		args.rule.slot = 0;
324 	} else {
325 		args.rule = *((struct ipfw_rule_ref *)(mtag+1));
326 		m_tag_delete(*m0, mtag);
327 		if (args.rule.info & IPFW_ONEPASS)
328 			return (0);
329 	}
330 
331 	/* I need some amt of data to be contiguous */
332 	m = *m0;
333 	i = min(m->m_pkthdr.len, max_protohdr);
334 	if (m->m_len < i) {
335 		m = m_pullup(m, i);
336 		if (m == NULL) {
337 			*m0 = m;
338 			return (0);
339 		}
340 	}
341 	eh = mtod(m, struct ether_header *);
342 	save_eh = *eh;			/* save copy for restore below */
343 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
344 
345 	args.m = m;		/* the packet we are looking at		*/
346 	args.oif = dir == PFIL_OUT ? ifp: NULL;	/* destination, if any	*/
347 	args.next_hop = NULL;	/* we do not support forward yet	*/
348 	args.next_hop6 = NULL;	/* we do not support forward yet	*/
349 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
350 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
351 	i = ipfw_chk(&args);
352 	m = args.m;
353 	if (m != NULL) {
354 		/*
355 		 * Restore Ethernet header, as needed, in case the
356 		 * mbuf chain was replaced by ipfw.
357 		 */
358 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
359 		if (m == NULL) {
360 			*m0 = NULL;
361 			return (0);
362 		}
363 		if (eh != mtod(m, struct ether_header *))
364 			bcopy(&save_eh, mtod(m, struct ether_header *),
365 				ETHER_HDR_LEN);
366 	}
367 	*m0 = m;
368 
369 	ret = 0;
370 	/* Check result of ipfw_chk() */
371 	switch (i) {
372 	case IP_FW_PASS:
373 		break;
374 
375 	case IP_FW_DENY:
376 		ret = EACCES;
377 		break; /* i.e. drop */
378 
379 	case IP_FW_DUMMYNET:
380 		ret = EACCES;
381 
382 		if (ip_dn_io_ptr == NULL)
383 			break; /* i.e. drop */
384 
385 		*m0 = NULL;
386 		dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT;
387 		ip_dn_io_ptr(&m, dir | PROTO_LAYER2, &args);
388 		return 0;
389 
390 	default:
391 		KASSERT(0, ("%s: unknown retval", __func__));
392 	}
393 
394 	if (ret != 0) {
395 		if (*m0)
396 			FREE_PKT(*m0);
397 		*m0 = NULL;
398 	}
399 
400 	return ret;
401 }
402 
403 /* do the divert, return 1 on error 0 on success */
404 static int
405 ipfw_divert(struct mbuf **m0, int incoming, struct ipfw_rule_ref *rule,
406 	int tee)
407 {
408 	/*
409 	 * ipfw_chk() has already tagged the packet with the divert tag.
410 	 * If tee is set, copy packet and return original.
411 	 * If not tee, consume packet and send it to divert socket.
412 	 */
413 	struct mbuf *clone;
414 	struct ip *ip = mtod(*m0, struct ip *);
415 	struct m_tag *tag;
416 
417 	/* Cloning needed for tee? */
418 	if (tee == 0) {
419 		clone = *m0;	/* use the original mbuf */
420 		*m0 = NULL;
421 	} else {
422 		clone = m_dup(*m0, M_NOWAIT);
423 		/* If we cannot duplicate the mbuf, we sacrifice the divert
424 		 * chain and continue with the tee-ed packet.
425 		 */
426 		if (clone == NULL)
427 			return 1;
428 	}
429 
430 	/*
431 	 * Divert listeners can normally handle non-fragmented packets,
432 	 * but we can only reass in the non-tee case.
433 	 * This means that listeners on a tee rule may get fragments,
434 	 * and have to live with that.
435 	 * Note that we now have the 'reass' ipfw option so if we care
436 	 * we can do it before a 'tee'.
437 	 */
438 	if (!tee) switch (ip->ip_v) {
439 	case IPVERSION:
440 	    if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
441 		int hlen;
442 		struct mbuf *reass;
443 
444 		reass = ip_reass(clone); /* Reassemble packet. */
445 		if (reass == NULL)
446 			return 0; /* not an error */
447 		/* if reass = NULL then it was consumed by ip_reass */
448 		/*
449 		 * IP header checksum fixup after reassembly and leave header
450 		 * in network byte order.
451 		 */
452 		ip = mtod(reass, struct ip *);
453 		hlen = ip->ip_hl << 2;
454 		ip->ip_sum = 0;
455 		if (hlen == sizeof(struct ip))
456 			ip->ip_sum = in_cksum_hdr(ip);
457 		else
458 			ip->ip_sum = in_cksum(reass, hlen);
459 		clone = reass;
460 	    }
461 	    break;
462 #ifdef INET6
463 	case IPV6_VERSION >> 4:
464 	    {
465 	    struct ip6_hdr *const ip6 = mtod(clone, struct ip6_hdr *);
466 
467 		if (ip6->ip6_nxt == IPPROTO_FRAGMENT) {
468 			int nxt, off;
469 
470 			off = sizeof(struct ip6_hdr);
471 			nxt = frag6_input(&clone, &off, 0);
472 			if (nxt == IPPROTO_DONE)
473 				return (0);
474 		}
475 		break;
476 	    }
477 #endif
478 	}
479 
480 	/* attach a tag to the packet with the reinject info */
481 	tag = m_tag_alloc(MTAG_IPFW_RULE, 0,
482 		    sizeof(struct ipfw_rule_ref), M_NOWAIT);
483 	if (tag == NULL) {
484 		FREE_PKT(clone);
485 		return 1;
486 	}
487 	*((struct ipfw_rule_ref *)(tag+1)) = *rule;
488 	m_tag_prepend(clone, tag);
489 
490 	/* Do the dirty job... */
491 	ip_divert_ptr(clone, incoming);
492 	return 0;
493 }
494 
495 /*
496  * attach or detach hooks for a given protocol family
497  */
498 static int
499 ipfw_hook(int onoff, int pf)
500 {
501 	struct pfil_head *pfh;
502 	pfil_func_t hook_func;
503 
504 	pfh = pfil_head_get(PFIL_TYPE_AF, pf);
505 	if (pfh == NULL)
506 		return ENOENT;
507 
508 	hook_func = (pf == AF_LINK) ? ipfw_check_frame : ipfw_check_packet;
509 
510 	(void) (onoff ? pfil_add_hook : pfil_remove_hook)
511 	    (hook_func, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh);
512 
513 	return 0;
514 }
515 
516 int
517 ipfw_attach_hooks(int arg)
518 {
519 	int error = 0;
520 
521 	if (arg == 0) /* detach */
522 		ipfw_hook(0, AF_INET);
523 	else if (V_fw_enable && ipfw_hook(1, AF_INET) != 0) {
524                 error = ENOENT; /* see ip_fw_pfil.c::ipfw_hook() */
525                 printf("ipfw_hook() error\n");
526         }
527 #ifdef INET6
528 	if (arg == 0) /* detach */
529 		ipfw_hook(0, AF_INET6);
530 	else if (V_fw6_enable && ipfw_hook(1, AF_INET6) != 0) {
531                 error = ENOENT;
532                 printf("ipfw6_hook() error\n");
533         }
534 #endif
535 	if (arg == 0) /* detach */
536 		ipfw_hook(0, AF_LINK);
537 	else if (V_fwlink_enable && ipfw_hook(1, AF_LINK) != 0) {
538                 error = ENOENT;
539                 printf("ipfw_link_hook() error\n");
540         }
541 	return error;
542 }
543 
544 int
545 ipfw_chg_hook(SYSCTL_HANDLER_ARGS)
546 {
547 	int newval;
548 	int error;
549 	int af;
550 
551 	if (arg1 == &V_fw_enable)
552 		af = AF_INET;
553 #ifdef INET6
554 	else if (arg1 == &V_fw6_enable)
555 		af = AF_INET6;
556 #endif
557 	else if (arg1 == &V_fwlink_enable)
558 		af = AF_LINK;
559 	else
560 		return (EINVAL);
561 
562 	newval = *(int *)arg1;
563 	/* Handle sysctl change */
564 	error = sysctl_handle_int(oidp, &newval, 0, req);
565 
566 	if (error)
567 		return (error);
568 
569 	/* Formalize new value */
570 	newval = (newval) ? 1 : 0;
571 
572 	if (*(int *)arg1 == newval)
573 		return (0);
574 
575 	error = ipfw_hook(newval, af);
576 	if (error)
577 		return (error);
578 	*(int *)arg1 = newval;
579 
580 	return (0);
581 }
582 /* end of file */
583