xref: /freebsd/sys/netpfil/ipfw/ip_fw_pfil.c (revision 7431dfd4580e850375fe5478d92ec770344db098)
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 #endif
63 
64 #include <netgraph/ng_ipfw.h>
65 
66 #include <netpfil/ipfw/ip_fw_private.h>
67 
68 #include <machine/in_cksum.h>
69 
70 static VNET_DEFINE(int, fw_enable) = 1;
71 #define V_fw_enable	VNET(fw_enable)
72 
73 #ifdef INET6
74 static VNET_DEFINE(int, fw6_enable) = 1;
75 #define V_fw6_enable	VNET(fw6_enable)
76 #endif
77 
78 static VNET_DEFINE(int, fwlink_enable) = 0;
79 #define V_fwlink_enable	VNET(fwlink_enable)
80 
81 int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
82 
83 /* Forward declarations. */
84 static int ipfw_divert(struct mbuf **, int, struct ipfw_rule_ref *, int);
85 int ipfw_check_packet(void *, struct mbuf **, struct ifnet *, int,
86 	struct inpcb *);
87 int ipfw_check_frame(void *, struct mbuf **, struct ifnet *, int,
88 	struct inpcb *);
89 
90 #ifdef SYSCTL_NODE
91 
92 SYSBEGIN(f1)
93 
94 SYSCTL_DECL(_net_inet_ip_fw);
95 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, enable,
96     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_enable), 0,
97     ipfw_chg_hook, "I", "Enable ipfw");
98 #ifdef INET6
99 SYSCTL_DECL(_net_inet6_ip6_fw);
100 SYSCTL_VNET_PROC(_net_inet6_ip6_fw, OID_AUTO, enable,
101     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw6_enable), 0,
102     ipfw_chg_hook, "I", "Enable ipfw+6");
103 #endif /* INET6 */
104 
105 SYSCTL_DECL(_net_link_ether);
106 SYSCTL_VNET_PROC(_net_link_ether, OID_AUTO, ipfw,
107     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fwlink_enable), 0,
108     ipfw_chg_hook, "I", "Pass ether pkts through firewall");
109 
110 SYSEND
111 
112 #endif /* SYSCTL_NODE */
113 
114 /*
115  * The pfilter hook to pass packets to ipfw_chk and then to
116  * dummynet, divert, netgraph or other modules.
117  * The packet may be consumed.
118  */
119 int
120 ipfw_check_packet(void *arg, struct mbuf **m0, struct ifnet *ifp, int dir,
121     struct inpcb *inp)
122 {
123 	struct ip_fw_args args;
124 	struct m_tag *tag;
125 	int ipfw;
126 	int ret;
127 
128 	/* convert dir to IPFW values */
129 	dir = (dir == PFIL_IN) ? DIR_IN : DIR_OUT;
130 	bzero(&args, sizeof(args));
131 
132 again:
133 	/*
134 	 * extract and remove the tag if present. If we are left
135 	 * with onepass, optimize the outgoing path.
136 	 */
137 	tag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
138 	if (tag != NULL) {
139 		args.rule = *((struct ipfw_rule_ref *)(tag+1));
140 		m_tag_delete(*m0, tag);
141 		if (args.rule.info & IPFW_ONEPASS)
142 			return (0);
143 	}
144 
145 	args.m = *m0;
146 	args.oif = dir == DIR_OUT ? ifp : NULL;
147 	args.inp = inp;
148 
149 	ipfw = ipfw_chk(&args);
150 	*m0 = args.m;
151 
152 	KASSERT(*m0 != NULL || ipfw == IP_FW_DENY, ("%s: m0 is NULL",
153 	    __func__));
154 
155 	/* breaking out of the switch means drop */
156 	ret = 0;	/* default return value for pass */
157 	switch (ipfw) {
158 	case IP_FW_PASS:
159 		/* next_hop may be set by ipfw_chk */
160 		if (args.next_hop == NULL && args.next_hop6 == NULL)
161 			break; /* pass */
162 #if (!defined(INET6) && !defined(INET))
163 		ret = EACCES;
164 #else
165 	    {
166 		struct m_tag *fwd_tag;
167 		size_t len;
168 
169 		KASSERT(args.next_hop == NULL || args.next_hop6 == NULL,
170 		    ("%s: both next_hop=%p and next_hop6=%p not NULL", __func__,
171 		     args.next_hop, args.next_hop6));
172 #ifdef INET6
173 		if (args.next_hop6 != NULL)
174 			len = sizeof(struct sockaddr_in6);
175 #endif
176 #ifdef INET
177 		if (args.next_hop != NULL)
178 			len = sizeof(struct sockaddr_in);
179 #endif
180 
181 		/* Incoming packets should not be tagged so we do not
182 		 * m_tag_find. Outgoing packets may be tagged, so we
183 		 * reuse the tag if present.
184 		 */
185 		fwd_tag = (dir == DIR_IN) ? NULL :
186 			m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
187 		if (fwd_tag != NULL) {
188 			m_tag_unlink(*m0, fwd_tag);
189 		} else {
190 			fwd_tag = m_tag_get(PACKET_TAG_IPFORWARD, len,
191 			    M_NOWAIT);
192 			if (fwd_tag == NULL) {
193 				ret = EACCES;
194 				break; /* i.e. drop */
195 			}
196 		}
197 #ifdef INET6
198 		if (args.next_hop6 != NULL) {
199 			bcopy(args.next_hop6, (fwd_tag+1), len);
200 			if (in6_localip(&args.next_hop6->sin6_addr))
201 				(*m0)->m_flags |= M_FASTFWD_OURS;
202 			(*m0)->m_flags |= M_IP6_NEXTHOP;
203 		}
204 #endif
205 #ifdef INET
206 		if (args.next_hop != NULL) {
207 			bcopy(args.next_hop, (fwd_tag+1), len);
208 			if (in_localip(args.next_hop->sin_addr))
209 				(*m0)->m_flags |= M_FASTFWD_OURS;
210 			(*m0)->m_flags |= M_IP_NEXTHOP;
211 		}
212 #endif
213 		m_tag_prepend(*m0, fwd_tag);
214 	    }
215 #endif /* INET || INET6 */
216 		break;
217 
218 	case IP_FW_DENY:
219 		ret = EACCES;
220 		break; /* i.e. drop */
221 
222 	case IP_FW_DUMMYNET:
223 		ret = EACCES;
224 		if (ip_dn_io_ptr == NULL)
225 			break; /* i.e. drop */
226 		if (mtod(*m0, struct ip *)->ip_v == 4)
227 			ret = ip_dn_io_ptr(m0, dir, &args);
228 		else if (mtod(*m0, struct ip *)->ip_v == 6)
229 			ret = ip_dn_io_ptr(m0, dir | PROTO_IPV6, &args);
230 		else
231 			break; /* drop it */
232 		/*
233 		 * XXX should read the return value.
234 		 * dummynet normally eats the packet and sets *m0=NULL
235 		 * unless the packet can be sent immediately. In this
236 		 * case args is updated and we should re-run the
237 		 * check without clearing args.
238 		 */
239 		if (*m0 != NULL)
240 			goto again;
241 		break;
242 
243 	case IP_FW_TEE:
244 	case IP_FW_DIVERT:
245 		if (ip_divert_ptr == NULL) {
246 			ret = EACCES;
247 			break; /* i.e. drop */
248 		}
249 		ret = ipfw_divert(m0, dir, &args.rule,
250 			(ipfw == IP_FW_TEE) ? 1 : 0);
251 		/* continue processing for the original packet (tee). */
252 		if (*m0)
253 			goto again;
254 		break;
255 
256 	case IP_FW_NGTEE:
257 	case IP_FW_NETGRAPH:
258 		if (ng_ipfw_input_p == NULL) {
259 			ret = EACCES;
260 			break; /* i.e. drop */
261 		}
262 		ret = ng_ipfw_input_p(m0, dir, &args,
263 			(ipfw == IP_FW_NGTEE) ? 1 : 0);
264 		if (ipfw == IP_FW_NGTEE) /* ignore errors for NGTEE */
265 			goto again;	/* continue with packet */
266 		break;
267 
268 	case IP_FW_NAT:
269 		/* honor one-pass in case of successful nat */
270 		if (V_fw_one_pass)
271 			break; /* ret is already 0 */
272 		goto again;
273 
274 	case IP_FW_REASS:
275 		goto again;		/* continue with packet */
276 
277 	default:
278 		KASSERT(0, ("%s: unknown retval", __func__));
279 	}
280 
281 	if (ret != 0) {
282 		if (*m0)
283 			FREE_PKT(*m0);
284 		*m0 = NULL;
285 	}
286 
287 	return ret;
288 }
289 
290 /*
291  * ipfw processing for ethernet packets (in and out).
292  * Inteface is NULL from ether_demux, and ifp from
293  * ether_output_frame.
294  */
295 int
296 ipfw_check_frame(void *arg, struct mbuf **m0, struct ifnet *dst, int dir,
297     struct inpcb *inp)
298 {
299 	struct ether_header *eh;
300 	struct ether_header save_eh;
301 	struct mbuf *m;
302 	int i, ret;
303 	struct ip_fw_args args;
304 	struct m_tag *mtag;
305 
306 	/* fetch start point from rule, if any */
307 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
308 	if (mtag == NULL) {
309 		args.rule.slot = 0;
310 	} else {
311 		/* dummynet packet, already partially processed */
312 		struct ipfw_rule_ref *r;
313 
314 		/* XXX can we free it after use ? */
315 		mtag->m_tag_id = PACKET_TAG_NONE;
316 		r = (struct ipfw_rule_ref *)(mtag + 1);
317 		if (r->info & IPFW_ONEPASS)
318 			return (0);
319 		args.rule = *r;
320 	}
321 
322 	/* I need some amt of data to be contiguous */
323 	m = *m0;
324 	i = min(m->m_pkthdr.len, max_protohdr);
325 	if (m->m_len < i) {
326 		m = m_pullup(m, i);
327 		if (m == NULL) {
328 			*m0 = m;
329 			return (0);
330 		}
331 	}
332 	eh = mtod(m, struct ether_header *);
333 	save_eh = *eh;			/* save copy for restore below */
334 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
335 
336 	args.m = m;		/* the packet we are looking at		*/
337 	args.oif = dir == PFIL_OUT ? dst: NULL;	/* destination, if any	*/
338 	args.next_hop = NULL;	/* we do not support forward yet	*/
339 	args.next_hop6 = NULL;	/* we do not support forward yet	*/
340 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
341 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
342 	i = ipfw_chk(&args);
343 	m = args.m;
344 	if (m != NULL) {
345 		/*
346 		 * Restore Ethernet header, as needed, in case the
347 		 * mbuf chain was replaced by ipfw.
348 		 */
349 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
350 		if (m == NULL) {
351 			*m0 = NULL;
352 			return (0);
353 		}
354 		if (eh != mtod(m, struct ether_header *))
355 			bcopy(&save_eh, mtod(m, struct ether_header *),
356 				ETHER_HDR_LEN);
357 	}
358 	*m0 = m;
359 
360 	ret = 0;
361 	/* Check result of ipfw_chk() */
362 	switch (i) {
363 	case IP_FW_PASS:
364 		break;
365 
366 	case IP_FW_DENY:
367 		ret = EACCES;
368 		break; /* i.e. drop */
369 
370 	case IP_FW_DUMMYNET:
371 		ret = EACCES;
372 		int dir;
373 
374 		if (ip_dn_io_ptr == NULL)
375 			break; /* i.e. drop */
376 
377 		*m0 = NULL;
378 		dir = PROTO_LAYER2 | (dst ? DIR_OUT : DIR_IN);
379 		ip_dn_io_ptr(&m, dir, &args);
380 		return 0;
381 
382 	default:
383 		KASSERT(0, ("%s: unknown retval", __func__));
384 	}
385 
386 	if (ret != 0) {
387 		if (*m0)
388 			FREE_PKT(*m0);
389 		*m0 = NULL;
390 	}
391 
392 	return ret;
393 }
394 
395 /* do the divert, return 1 on error 0 on success */
396 static int
397 ipfw_divert(struct mbuf **m0, int incoming, struct ipfw_rule_ref *rule,
398 	int tee)
399 {
400 	/*
401 	 * ipfw_chk() has already tagged the packet with the divert tag.
402 	 * If tee is set, copy packet and return original.
403 	 * If not tee, consume packet and send it to divert socket.
404 	 */
405 	struct mbuf *clone;
406 	struct ip *ip = mtod(*m0, struct ip *);
407 	struct m_tag *tag;
408 
409 	/* Cloning needed for tee? */
410 	if (tee == 0) {
411 		clone = *m0;	/* use the original mbuf */
412 		*m0 = NULL;
413 	} else {
414 		clone = m_dup(*m0, M_NOWAIT);
415 		/* If we cannot duplicate the mbuf, we sacrifice the divert
416 		 * chain and continue with the tee-ed packet.
417 		 */
418 		if (clone == NULL)
419 			return 1;
420 	}
421 
422 	/*
423 	 * Divert listeners can normally handle non-fragmented packets,
424 	 * but we can only reass in the non-tee case.
425 	 * This means that listeners on a tee rule may get fragments,
426 	 * and have to live with that.
427 	 * Note that we now have the 'reass' ipfw option so if we care
428 	 * we can do it before a 'tee'.
429 	 */
430 	if (!tee) switch (ip->ip_v) {
431 	case IPVERSION:
432 	    if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
433 		int hlen;
434 		struct mbuf *reass;
435 
436 		reass = ip_reass(clone); /* Reassemble packet. */
437 		if (reass == NULL)
438 			return 0; /* not an error */
439 		/* if reass = NULL then it was consumed by ip_reass */
440 		/*
441 		 * IP header checksum fixup after reassembly and leave header
442 		 * in network byte order.
443 		 */
444 		ip = mtod(reass, struct ip *);
445 		hlen = ip->ip_hl << 2;
446 		ip->ip_sum = 0;
447 		if (hlen == sizeof(struct ip))
448 			ip->ip_sum = in_cksum_hdr(ip);
449 		else
450 			ip->ip_sum = in_cksum(reass, hlen);
451 		clone = reass;
452 	    }
453 	    break;
454 #ifdef INET6
455 	case IPV6_VERSION >> 4:
456 	    {
457 	    struct ip6_hdr *const ip6 = mtod(clone, struct ip6_hdr *);
458 
459 		if (ip6->ip6_nxt == IPPROTO_FRAGMENT) {
460 			int nxt, off;
461 
462 			off = sizeof(struct ip6_hdr);
463 			nxt = frag6_input(&clone, &off, 0);
464 			if (nxt == IPPROTO_DONE)
465 				return (0);
466 		}
467 		break;
468 	    }
469 #endif
470 	}
471 
472 	/* attach a tag to the packet with the reinject info */
473 	tag = m_tag_alloc(MTAG_IPFW_RULE, 0,
474 		    sizeof(struct ipfw_rule_ref), M_NOWAIT);
475 	if (tag == NULL) {
476 		FREE_PKT(clone);
477 		return 1;
478 	}
479 	*((struct ipfw_rule_ref *)(tag+1)) = *rule;
480 	m_tag_prepend(clone, tag);
481 
482 	/* Do the dirty job... */
483 	ip_divert_ptr(clone, incoming);
484 	return 0;
485 }
486 
487 /*
488  * attach or detach hooks for a given protocol family
489  */
490 static int
491 ipfw_hook(int onoff, int pf)
492 {
493 	struct pfil_head *pfh;
494 	void *hook_func;
495 
496 	pfh = pfil_head_get(PFIL_TYPE_AF, pf);
497 	if (pfh == NULL)
498 		return ENOENT;
499 
500 	hook_func = (pf == AF_LINK) ? ipfw_check_frame : ipfw_check_packet;
501 
502 	(void) (onoff ? pfil_add_hook : pfil_remove_hook)
503 	    (hook_func, NULL, PFIL_IN | PFIL_OUT | PFIL_WAITOK, pfh);
504 
505 	return 0;
506 }
507 
508 int
509 ipfw_attach_hooks(int arg)
510 {
511 	int error = 0;
512 
513 	if (arg == 0) /* detach */
514 		ipfw_hook(0, AF_INET);
515 	else if (V_fw_enable && ipfw_hook(1, AF_INET) != 0) {
516                 error = ENOENT; /* see ip_fw_pfil.c::ipfw_hook() */
517                 printf("ipfw_hook() error\n");
518         }
519 #ifdef INET6
520 	if (arg == 0) /* detach */
521 		ipfw_hook(0, AF_INET6);
522 	else if (V_fw6_enable && ipfw_hook(1, AF_INET6) != 0) {
523                 error = ENOENT;
524                 printf("ipfw6_hook() error\n");
525         }
526 #endif
527 	if (arg == 0) /* detach */
528 		ipfw_hook(0, AF_LINK);
529 	else if (V_fwlink_enable && ipfw_hook(1, AF_LINK) != 0) {
530                 error = ENOENT;
531                 printf("ipfw_link_hook() error\n");
532         }
533 	return error;
534 }
535 
536 int
537 ipfw_chg_hook(SYSCTL_HANDLER_ARGS)
538 {
539 	int newval;
540 	int error;
541 	int af;
542 
543 	if (arg1 == &V_fw_enable)
544 		af = AF_INET;
545 #ifdef INET6
546 	else if (arg1 == &V_fw6_enable)
547 		af = AF_INET6;
548 #endif
549 	else if (arg1 == &V_fwlink_enable)
550 		af = AF_LINK;
551 	else
552 		return (EINVAL);
553 
554 	newval = *(int *)arg1;
555 	/* Handle sysctl change */
556 	error = sysctl_handle_int(oidp, &newval, 0, req);
557 
558 	if (error)
559 		return (error);
560 
561 	/* Formalize new value */
562 	newval = (newval) ? 1 : 0;
563 
564 	if (*(int *)arg1 == newval)
565 		return (0);
566 
567 	error = ipfw_hook(newval, af);
568 	if (error)
569 		return (error);
570 	*(int *)arg1 = newval;
571 
572 	return (0);
573 }
574 /* end of file */
575