1 /*	$NetBSD: ip_fil_netbsd.c,v 1.18 2016/07/18 21:07:30 pgoyette Exp $	*/
2 
3 /*
4  * Copyright (C) 2012 by Darren Reed.
5  *
6  * See the IPFILTER.LICENCE file for details on licencing.
7  */
8 #if !defined(lint)
9 #if defined(__NetBSD__)
10 #include <sys/cdefs.h>
11 __KERNEL_RCSID(0, "$NetBSD: ip_fil_netbsd.c,v 1.18 2016/07/18 21:07:30 pgoyette Exp $");
12 #else
13 static const char sccsid[] = "@(#)ip_fil.c	2.41 6/5/96 (C) 1993-2000 Darren Reed";
14 static const char rcsid[] = "@(#)Id: ip_fil_netbsd.c,v 1.1.1.2 2012/07/22 13:45:17 darrenr Exp";
15 #endif
16 #endif
17 
18 #if defined(KERNEL) || defined(_KERNEL)
19 # undef KERNEL
20 # undef _KERNEL
21 # define        KERNEL	1
22 # define        _KERNEL	1
23 #endif
24 #include <sys/param.h>
25 #if (NetBSD >= 199905) && !defined(IPFILTER_LKM)
26 # if (__NetBSD_Version__ >= 799003000)
27 #   ifdef _KERNEL_OPT
28 #    include "opt_ipsec.h"
29 #   endif
30 # else
31 #  include "opt_ipsec.h"
32 # endif
33 #endif
34 #include <sys/errno.h>
35 #include <sys/types.h>
36 #include <sys/file.h>
37 #include <sys/ioctl.h>
38 #include <sys/time.h>
39 #include <sys/systm.h>
40 #include <sys/select.h>
41 #if (NetBSD > 199609)
42 # include <sys/dirent.h>
43 #else
44 # include <sys/dir.h>
45 #endif
46 #if (__NetBSD_Version__ >= 599005900)
47 # include <sys/cprng.h>
48 #endif
49 #include <sys/mbuf.h>
50 #include <sys/protosw.h>
51 #include <sys/socket.h>
52 #include <sys/poll.h>
53 #if (__NetBSD_Version__ >= 399002000)
54 # include <sys/kauth.h>
55 #endif
56 #if (__NetBSD_Version__ >= 799003000)
57 #include <sys/module.h>
58 #include <sys/mutex.h>
59 #endif
60 
61 #include <net/if.h>
62 #include <net/route.h>
63 #include <netinet/in.h>
64 #include <netinet/in_var.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/ip.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/tcp.h>
69 #if __NetBSD_Version__ >= 105190000	/* 1.5T */
70 # include <netinet/tcp_timer.h>
71 # include <netinet/tcp_var.h>
72 #endif
73 #include <netinet/udp.h>
74 #include <netinet/tcpip.h>
75 #include <netinet/ip_icmp.h>
76 #include "netinet/ip_compat.h"
77 #ifdef USE_INET6
78 # include <netinet/icmp6.h>
79 # if (__NetBSD_Version__ >= 106000000)
80 #  include <netinet6/nd6.h>
81 # endif
82 #endif
83 #include "netinet/ip_fil.h"
84 #include "netinet/ip_nat.h"
85 #include "netinet/ip_frag.h"
86 #include "netinet/ip_state.h"
87 #include "netinet/ip_proxy.h"
88 #include "netinet/ip_auth.h"
89 #include "netinet/ip_sync.h"
90 #include "netinet/ip_lookup.h"
91 #include "netinet/ip_dstlist.h"
92 #ifdef	IPFILTER_SCAN
93 #include "netinet/ip_scan.h"
94 #endif
95 #include <sys/md5.h>
96 #include <sys/kernel.h>
97 #include <sys/conf.h>
98 #ifdef INET
99 extern	int	ip_optcopy (struct ip *, struct ip *);
100 #endif
101 
102 #ifdef IPFILTER_M_IPFILTER
103 MALLOC_DEFINE(M_IPFILTER, "IP Filter", "IP Filter packet filter data structures");
104 #endif
105 
106 #if __NetBSD_Version__ >= 105009999
107 # define	csuminfo	csum_flags
108 #endif
109 
110 #if __NetBSD_Version__ < 200000000
111 extern	struct	protosw	inetsw[];
112 #endif
113 
114 #if (__NetBSD_Version__ >= 599002000)
115 static kauth_listener_t ipf_listener;
116 #endif
117 
118 #if (__NetBSD_Version__ < 399001400)
119 extern int ip6_getpmtu (struct route_in6 *, struct route_in6 *,
120 			    struct ifnet *, struct in6_addr *, u_long *,
121 			    int *);
122 #endif
123 #if (NetBSD >= 199511)
124 static  int     ipfopen(dev_t dev, int flags, int devtype, PROC_T *p);
125 static  int     ipfclose(dev_t dev, int flags, int devtype, PROC_T *p);
126 #else
127 # if (__NetBSD_Version__ >= 399001400)
128 static  int     ipfopen(dev_t dev, int flags, struct lwp *);
129 static  int     ipfclose(dev_t dev, int flags, struct lwp *);
130 # else
131 static  int     ipfopen(dev_t dev, int flags);
132 static  int     ipfclose(dev_t dev, int flags);
133 # endif /* __NetBSD_Version__ >= 399001400 */
134 #endif
135 static  int     ipfread(dev_t, struct uio *, int ioflag);
136 static  int     ipfwrite(dev_t, struct uio *, int ioflag);
137 static  int     ipfpoll(dev_t, int events, PROC_T *);
138 static	void	ipf_timer_func(void *ptr);
139 
140 const struct cdevsw ipl_cdevsw = {
141 	.d_open = ipfopen,
142 	.d_close = ipfclose,
143 	.d_read = ipfread,
144 	.d_write = ipfwrite,
145 	.d_ioctl = ipfioctl,
146 	.d_stop = nostop,
147 	.d_tty = notty,
148 	.d_poll = ipfpoll,
149 	.d_mmap = nommap,
150 #if  (__NetBSD_Version__ >= 200000000)
151 	.d_kqfilter = nokqfilter,
152 #endif
153 	.d_discard = nodiscard,
154 #ifdef D_OTHER
155 	.d_flag = D_OTHER
156 #else
157 	.d_flag = 0
158 #endif
159 };
160 #if (__NetBSD_Version__ >= 799003000)
161 kmutex_t ipf_ref_mutex;
162 int	ipf_active;
163 #endif
164 
165 ipf_main_softc_t ipfmain;
166 
167 static	u_short	ipid = 0;
168 static	int	(*ipf_savep)(void *, ip_t *, int, void *, int, struct mbuf **);
169 static	int	ipf_send_ip(fr_info_t *, mb_t *);
170 #ifdef USE_INET6
171 static int ipf_fastroute6(struct mbuf *, struct mbuf **,
172 			      fr_info_t *, frdest_t *);
173 #endif
174 
175 #if defined(NETBSD_PF)
176 # include <net/pfil.h>
177 /*
178  * We provide the ipf_checkp name just to minimize changes later.
179  */
180 int (*ipf_checkp)(void *, ip_t *ip, int hlen, void *ifp, int out, mb_t **mp);
181 #endif /* NETBSD_PF */
182 
183 #if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)
184 # include <net/pfil.h>
185 
186 static int ipf_check_wrapper(void *, struct mbuf **, struct ifnet *, int );
187 
188 static int
ipf_check_wrapper(void * arg,struct mbuf ** mp,struct ifnet * ifp,int dir)189 ipf_check_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
190 {
191 	struct ip *ip;
192 	int rv, hlen;
193 
194 #if __NetBSD_Version__ >= 200080000
195 	/*
196 	 * ensure that mbufs are writable beforehand
197 	 * as it's assumed by ipf code.
198 	 * XXX inefficient
199 	 */
200 	int error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
201 
202 	if (error) {
203 		m_freem(*mp);
204 		*mp = NULL;
205 		return error;
206 	}
207 #endif
208 	ip = mtod(*mp, struct ip *);
209 	hlen = ip->ip_hl << 2;
210 
211 #ifdef INET
212 #if defined(M_CSUM_TCPv4)
213 	/*
214 	 * If the packet is out-bound, we can't delay checksums
215 	 * here.  For in-bound, the checksum has already been
216 	 * validated.
217 	 */
218 	if (dir == PFIL_OUT) {
219 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
220 			in_delayed_cksum(*mp);
221 			(*mp)->m_pkthdr.csum_flags &=
222 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
223 		}
224 	}
225 #endif /* M_CSUM_TCPv4 */
226 #endif /* INET */
227 
228 	/*
229 	 * Note, we don't need to update the checksum, because
230 	 * it has already been verified.
231 	 */
232 	rv = ipf_check(&ipfmain, ip, hlen, ifp, (dir == PFIL_OUT), mp);
233 
234 	return (rv);
235 }
236 
237 # ifdef USE_INET6
238 #  include <netinet/ip6.h>
239 
240 static int ipf_check_wrapper6(void *, struct mbuf **, struct ifnet *, int );
241 
242 static int
ipf_check_wrapper6(void * arg,struct mbuf ** mp,struct ifnet * ifp,int dir)243 ipf_check_wrapper6(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
244 {
245 #if defined(INET6)
246 #  if defined(M_CSUM_TCPv6) && (__NetBSD_Version__ > 200000000)
247 	/*
248 	 * If the packet is out-bound, we can't delay checksums
249 	 * here.  For in-bound, the checksum has already been
250 	 * validated.
251 	 */
252 	if (dir == PFIL_OUT) {
253 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
254 #   if (__NetBSD_Version__ > 399000600)
255 			in6_delayed_cksum(*mp);
256 #   endif
257 			(*mp)->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv6|
258 							M_CSUM_UDPv6);
259 		}
260 	}
261 #  endif
262 #endif /* INET6 */
263 
264 	return (ipf_check(&ipfmain, mtod(*mp, struct ip *), sizeof(struct ip6_hdr),
265 	    ifp, (dir == PFIL_OUT), mp));
266 }
267 # endif
268 
269 
270 # if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
271 static int ipf_pfilsync(void *, struct mbuf **, struct ifnet *, int);
272 
273 static int
ipf_pfilsync(void * hdr,struct mbuf ** mp,struct ifnet * ifp,int dir)274 ipf_pfilsync(void *hdr, struct mbuf **mp, struct ifnet *ifp, int dir)
275 {
276 	/*
277 	 * The interface pointer is useless for create (we have nothing to
278 	 * compare it to) and at detach, the interface name is still in the
279 	 * list of active NICs (albeit, down, but that's not any real
280 	 * indicator) and doing ifunit() on the name will still return the
281 	 * pointer, so it's not much use then, either.
282 	 */
283 	ipf_sync(&ipfmain, NULL);
284 	return 0;
285 }
286 # endif
287 
288 #endif /* __NetBSD_Version__ >= 105110000 */
289 
290 
291 #if defined(IPFILTER_LKM)
292 int
ipf_identify(s)293 ipf_identify(s)
294 	char *s;
295 {
296 	if (strcmp(s, "ipl") == 0)
297 		return 1;
298 	return 0;
299 }
300 #endif /* IPFILTER_LKM */
301 
302 #if (__NetBSD_Version__ >= 599002000)
303 static int
ipf_listener_cb(kauth_cred_t cred,kauth_action_t action,void * cookie,void * arg0,void * arg1,void * arg2,void * arg3)304 ipf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
305     void *arg0, void *arg1, void *arg2, void *arg3)
306 {
307 	int result;
308 	enum kauth_network_req req;
309 
310 	result = KAUTH_RESULT_DEFER;
311 	req = (enum kauth_network_req)arg0;
312 
313 	if (action != KAUTH_NETWORK_FIREWALL)
314 		return result;
315 
316 	/* These must have came from device context. */
317 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
318 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
319 		result = KAUTH_RESULT_ALLOW;
320 
321 	return result;
322 }
323 #endif
324 
325 /*
326  * Try to detect the case when compiling for NetBSD with pseudo-device
327  */
328 void
ipfilterattach(int count)329 ipfilterattach(int count)
330 {
331 
332 #if (__NetBSD_Version__ >= 799003000)
333 	return;
334 #else
335 #if (__NetBSD_Version__ >= 599002000)
336 	ipf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
337 	    ipf_listener_cb, NULL);
338 #endif
339 
340 	if (ipf_load_all() == 0)
341 		(void) ipf_create_all(&ipfmain);
342 #endif
343 }
344 
345 
346 int
ipfattach(ipf_main_softc_t * softc)347 ipfattach(ipf_main_softc_t *softc)
348 {
349 	SPL_INT(s);
350 #if (__NetBSD_Version__ >= 499005500)
351 	int i;
352 #endif
353 #if defined(NETBSD_PF) && (__NetBSD_Version__ >= 104200000)
354 	int error = 0;
355 # if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)
356         pfil_head_t *ph_inet;
357 #  ifdef USE_INET6
358         pfil_head_t *ph_inet6;
359 #  endif
360 #  if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
361         pfil_head_t *ph_ifsync;
362 #  endif
363 # endif
364 #endif
365 
366 	SPL_NET(s);
367 	if ((softc->ipf_running > 0) || (ipf_checkp == ipf_check)) {
368 		printf("IP Filter: already initialized\n");
369 		SPL_X(s);
370 		IPFERROR(130017);
371 		return EBUSY;
372 	}
373 
374 	if (ipf_init_all(softc) < 0) {
375 		SPL_X(s);
376 		IPFERROR(130015);
377 		return EIO;
378 	}
379 
380 #ifdef NETBSD_PF
381 # if (__NetBSD_Version__ >= 104200000)
382 #  if __NetBSD_Version__ >= 105110000
383 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
384 #   ifdef USE_INET6
385 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
386 #   endif
387 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
388 	ph_ifsync = pfil_head_get(PFIL_TYPE_IFNET, 0);
389 #   endif
390 
391 	if (ph_inet == NULL
392 #   ifdef USE_INET6
393 	    && ph_inet6 == NULL
394 #   endif
395 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
396 	    && ph_ifsync == NULL
397 #   endif
398 	   ) {
399 		SPL_X(s);
400 		IPFERROR(130016);
401 		return ENODEV;
402 	}
403 
404 	if (ph_inet != NULL)
405 		error = pfil_add_hook((void *)ipf_check_wrapper, NULL,
406 				      PFIL_IN|PFIL_OUT, ph_inet);
407 	else
408 		error = 0;
409 #  else
410 	error = pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
411 			      &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
412 #  endif
413 	if (error) {
414 		IPFERROR(130013);
415 		goto pfil_error;
416 	}
417 # else
418 	pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT);
419 # endif
420 
421 # ifdef USE_INET6
422 #  if __NetBSD_Version__ >= 105110000
423 	if (ph_inet6 != NULL)
424 		error = pfil_add_hook((void *)ipf_check_wrapper6, NULL,
425 				      PFIL_IN|PFIL_OUT, ph_inet6);
426 	else
427 		error = 0;
428 	if (error) {
429 		pfil_remove_hook((void *)ipf_check_wrapper6, NULL,
430 				 PFIL_IN|PFIL_OUT, ph_inet6);
431 		ipfmain.ipf_interror = 130014;
432 		goto pfil_error;
433 	}
434 #  else
435 	error = pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
436 			      &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
437 	if (error) {
438 		pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
439 				 &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
440 		IPFERROR(130014);
441 		goto pfil_error;
442 	}
443 #  endif
444 # endif
445 
446 # if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
447 	if (ph_ifsync != NULL)
448 		(void) pfil_add_hook((void *)ipf_pfilsync, NULL,
449 				     PFIL_IFNET, ph_ifsync);
450 # endif
451 #endif
452 
453 #if (__NetBSD_Version__ >= 499005500)
454 	for (i = 0; i < IPL_LOGSIZE; i++)
455 		selinit(&ipfmain.ipf_selwait[i]);
456 #else
457 	bzero((char *)ipfmain.ipf_selwait, sizeof(ipfmain.ipf_selwait));
458 #endif
459 	ipf_savep = ipf_checkp;
460 	ipf_checkp = ipf_check;
461 
462 #ifdef INET
463 	if (softc->ipf_control_forwarding & 1)
464 		ipforwarding = 1;
465 #endif
466 
467 	ipid = 0;
468 
469 	SPL_X(s);
470 
471 #if (__NetBSD_Version__ >= 104010000)
472 # if (__NetBSD_Version__ >= 499002000)
473 	callout_init(&softc->ipf_slow_ch, 0);
474 # else
475 	callout_init(&softc->ipf_slow_ch);
476 # endif
477 	callout_reset(&softc->ipf_slow_ch, (hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT,
478 		     ipf_timer_func, softc);
479 #else
480 	timeout(ipf_timer_func, softc, (hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT);
481 #endif
482 
483 	return 0;
484 
485 #if __NetBSD_Version__ >= 105110000
486 pfil_error:
487 	SPL_X(s);
488 	ipf_fini_all(softc);
489 	return error;
490 #endif
491 }
492 
493 static void
ipf_timer_func(void * ptr)494 ipf_timer_func(void *ptr)
495 {
496 	ipf_main_softc_t *softc = ptr;
497 	SPL_INT(s);
498 
499 	SPL_NET(s);
500 	READ_ENTER(&softc->ipf_global);
501 
502 	if (softc->ipf_running > 0)
503 		ipf_slowtimer(softc);
504 
505 	if (softc->ipf_running == -1 || softc->ipf_running == 1) {
506 #if NETBSD_GE_REV(104240000)
507 		callout_reset(&softc->ipf_slow_ch, hz / 2,
508 			      ipf_timer_func, softc);
509 #else
510 		timeout(ipf_timer_func, softc,
511 			(hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT);
512 #endif
513 	}
514 	RWLOCK_EXIT(&softc->ipf_global);
515 	SPL_X(s);
516 }
517 
518 
519 /*
520  * Disable the filter by removing the hooks from the IP input/output
521  * stream.
522  */
523 int
ipfdetach(ipf_main_softc_t * softc)524 ipfdetach(ipf_main_softc_t *softc)
525 {
526 	SPL_INT(s);
527 #if (__NetBSD_Version__ >= 499005500)
528 	int i;
529 #endif
530 #if defined(NETBSD_PF) && (__NetBSD_Version__ >= 104200000)
531 	int error = 0;
532 # if __NetBSD_Version__ >= 105150000
533 	pfil_head_t *ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
534 #  ifdef USE_INET6
535 	pfil_head_t *ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
536 #  endif
537 #  if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
538 	struct pfil_head *ph_ifsync = pfil_head_get(PFIL_TYPE_IFNET, 0);
539 #  endif
540 # endif
541 #endif
542 
543 	SPL_NET(s);
544 
545 #if (__NetBSD_Version__ >= 104010000)
546 	if (softc->ipf_running > 0)
547 		callout_stop(&softc->ipf_slow_ch);
548 #else
549 	untimeout(ipf_slowtimer, NULL);
550 #endif /* NetBSD */
551 
552 	ipf_checkp = ipf_savep;
553 	(void) ipf_flush(softc, IPL_LOGIPF, FR_INQUE|FR_OUTQUE|FR_INACTIVE);
554 	(void) ipf_flush(softc, IPL_LOGIPF, FR_INQUE|FR_OUTQUE);
555 
556 #ifdef INET
557 	if (softc->ipf_control_forwarding & 2)
558 		ipforwarding = 0;
559 #endif
560 
561 #ifdef NETBSD_PF
562 # if (__NetBSD_Version__ >= 104200000)
563 #  if __NetBSD_Version__ >= 105110000
564 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
565 	(void) pfil_remove_hook((void *)ipf_pfilsync, NULL,
566 				PFIL_IFNET, ph_ifsync);
567 #   endif
568 
569 	if (ph_inet != NULL)
570 		error = pfil_remove_hook((void *)ipf_check_wrapper, NULL,
571 					 PFIL_IN|PFIL_OUT, ph_inet);
572 	else
573 		error = 0;
574 #  else
575 	error = pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
576 				 &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
577 #  endif
578 	if (error) {
579 		SPL_X(s);
580 		IPFERROR(130011);
581 		return error;
582 	}
583 # else
584 	pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT);
585 # endif
586 # ifdef USE_INET6
587 #  if __NetBSD_Version__ >= 105110000
588 	if (ph_inet6 != NULL)
589 		error = pfil_remove_hook((void *)ipf_check_wrapper6, NULL,
590 					 PFIL_IN|PFIL_OUT, ph_inet6);
591 	else
592 		error = 0;
593 #  else
594 	error = pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
595 				 &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
596 #  endif
597 	if (error) {
598 		SPL_X(s);
599 		IPFERROR(130012);
600 		return error;
601 	}
602 # endif
603 #endif
604 	SPL_X(s);
605 
606 #if (__NetBSD_Version__ >= 499005500)
607 	for (i = 0; i < IPL_LOGSIZE; i++)
608 		seldestroy(&ipfmain.ipf_selwait[i]);
609 #endif
610 
611 	ipf_fini_all(softc);
612 
613 	return 0;
614 }
615 
616 
617 /*
618  * Filter ioctl interface.
619  */
620 int
ipfioctl(dev_t dev,u_long cmd,void * data,int mode,struct lwp * p)621 ipfioctl(dev_t dev, u_long cmd,
622 #if  (__NetBSD_Version__ >= 499001000)
623 	void *data,
624 #else
625 	caddr_t data,
626 #endif
627 	int mode
628 #if (NetBSD >= 199511)
629 # if  (__NetBSD_Version__ >= 399001400)
630 	, struct lwp *p
631 #   if (__NetBSD_Version__ >= 399002000)
632 #     define	UID(l)	kauth_cred_getuid((l)->l_cred)
633 #  else
634 #     define	UID(l)	((l)->l_proc->p_cred->p_ruid)
635 #  endif
636 # else
637 	, struct proc *p
638 #  define	UID(p)	((p)->p_cred->p_ruid)
639 # endif
640 #endif
641 )
642 {
643 	int error = 0, unit = 0;
644 	SPL_INT(s);
645 
646 #if (__NetBSD_Version__ >= 399002000)
647 	if ((mode & FWRITE) &&
648 	    kauth_authorize_network(p->l_cred, KAUTH_NETWORK_FIREWALL,
649 				    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL,
650 				    NULL, NULL)) {
651 		ipfmain.ipf_interror = 130005;
652 		return EPERM;
653 	}
654 #else
655 	if ((securelevel >= 2) && (mode & FWRITE)) {
656 		ipfmain.ipf_interror = 130001;
657 		return EPERM;
658 	}
659 #endif
660 
661   	unit = GET_MINOR(dev);
662   	if ((IPL_LOGMAX < unit) || (unit < 0)) {
663 		ipfmain.ipf_interror = 130002;
664 		return ENXIO;
665 	}
666 
667 	if (ipfmain.ipf_running <= 0) {
668 		if (unit != IPL_LOGIPF && cmd != SIOCIPFINTERROR) {
669 			ipfmain.ipf_interror = 130003;
670 			return EIO;
671 		}
672 		if (cmd != SIOCIPFGETNEXT && cmd != SIOCIPFGET &&
673 		    cmd != SIOCIPFSET && cmd != SIOCFRENB &&
674 		    cmd != SIOCGETFS && cmd != SIOCGETFF &&
675 		    cmd != SIOCIPFINTERROR) {
676 			ipfmain.ipf_interror = 130004;
677 			return EIO;
678 		}
679 	}
680 
681 	SPL_NET(s);
682 
683 	error = ipf_ioctlswitch(&ipfmain, unit, data, cmd, mode, UID(p), p);
684 	if (error != -1) {
685 		SPL_X(s);
686 		return error;
687 	}
688 
689 	SPL_X(s);
690 	return error;
691 }
692 
693 
694 /*
695  * ipf_send_reset - this could conceivably be a call to tcp_respond(), but that
696  * requires a large amount of setting up and isn't any more efficient.
697  */
698 int
ipf_send_reset(fr_info_t * fin)699 ipf_send_reset(fr_info_t *fin)
700 {
701 	struct tcphdr *tcp, *tcp2;
702 	int tlen = 0, hlen;
703 	struct mbuf *m;
704 #ifdef USE_INET6
705 	ip6_t *ip6;
706 #endif
707 	ip_t *ip;
708 
709 	tcp = fin->fin_dp;
710 	if (tcp->th_flags & TH_RST)
711 		return -1;		/* feedback loop */
712 
713 	if (ipf_checkl4sum(fin) == -1)
714 		return -1;
715 
716 	tlen = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
717 			((tcp->th_flags & TH_SYN) ? 1 : 0) +
718 			((tcp->th_flags & TH_FIN) ? 1 : 0);
719 
720 #ifdef USE_INET6
721 	hlen = (fin->fin_v == 6) ? sizeof(ip6_t) : sizeof(ip_t);
722 #else
723 	hlen = sizeof(ip_t);
724 #endif
725 #ifdef MGETHDR
726 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
727 #else
728 	MGET(m, M_DONTWAIT, MT_HEADER);
729 #endif
730 	if (m == NULL)
731 		return -1;
732 	if (sizeof(*tcp2) + hlen > MHLEN) {
733 		MCLGET(m, M_DONTWAIT);
734 		if (m == NULL)
735 			return -1;
736 		if ((m->m_flags & M_EXT) == 0) {
737 			FREE_MB_T(m);
738 			return -1;
739 		}
740 	}
741 
742 	m->m_len = sizeof(*tcp2) + hlen;
743 	m->m_data += max_linkhdr;
744 	m->m_pkthdr.len = m->m_len;
745 	m_reset_rcvif(m);
746 	ip = mtod(m, struct ip *);
747 	bzero((char *)ip, hlen);
748 #ifdef USE_INET6
749 	ip6 = (ip6_t *)ip;
750 #endif
751 	bzero((char *)ip, sizeof(*tcp2) + hlen);
752 	tcp2 = (struct tcphdr *)((char *)ip + hlen);
753 	tcp2->th_sport = tcp->th_dport;
754 	tcp2->th_dport = tcp->th_sport;
755 
756 	if (tcp->th_flags & TH_ACK) {
757 		tcp2->th_seq = tcp->th_ack;
758 		tcp2->th_flags = TH_RST;
759 		tcp2->th_ack = 0;
760 	} else {
761 		tcp2->th_seq = 0;
762 		tcp2->th_ack = ntohl(tcp->th_seq);
763 		tcp2->th_ack += tlen;
764 		tcp2->th_ack = htonl(tcp2->th_ack);
765 		tcp2->th_flags = TH_RST|TH_ACK;
766 	}
767 	tcp2->th_x2 = 0;
768 	TCP_OFF_A(tcp2, sizeof(*tcp2) >> 2);
769 	tcp2->th_win = tcp->th_win;
770 	tcp2->th_sum = 0;
771 	tcp2->th_urp = 0;
772 
773 #ifdef USE_INET6
774 	if (fin->fin_v == 6) {
775 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
776 		ip6->ip6_plen = htons(sizeof(struct tcphdr));
777 		ip6->ip6_nxt = IPPROTO_TCP;
778 		ip6->ip6_hlim = 0;
779 		ip6->ip6_src = fin->fin_dst6.in6;
780 		ip6->ip6_dst = fin->fin_src6.in6;
781 		tcp2->th_sum = in6_cksum(m, IPPROTO_TCP,
782 					 sizeof(*ip6), sizeof(*tcp2));
783 		return ipf_send_ip(fin, m);
784 	}
785 #endif
786 #ifdef INET
787 	ip->ip_p = IPPROTO_TCP;
788 	ip->ip_len = htons(sizeof(struct tcphdr));
789 	ip->ip_src.s_addr = fin->fin_daddr;
790 	ip->ip_dst.s_addr = fin->fin_saddr;
791 	tcp2->th_sum = in_cksum(m, hlen + sizeof(*tcp2));
792 	ip->ip_len = hlen + sizeof(*tcp2);
793 	return ipf_send_ip(fin, m);
794 #else
795 	return 0;
796 #endif
797 }
798 
799 
800 /*
801  * Expects ip_len to be in host byte order when called.
802  */
803 static int
ipf_send_ip(fr_info_t * fin,mb_t * m)804 ipf_send_ip(fr_info_t *fin, mb_t *m)
805 {
806 	fr_info_t fnew;
807 #ifdef INET
808 	ip_t *oip;
809 #endif
810 	ip_t *ip;
811 	int hlen;
812 
813 	ip = mtod(m, ip_t *);
814 	bzero((char *)&fnew, sizeof(fnew));
815 	fnew.fin_main_soft = fin->fin_main_soft;
816 
817 	IP_V_A(ip, fin->fin_v);
818 	switch (fin->fin_v)
819 	{
820 #ifdef INET
821 	case 4 :
822 		oip = fin->fin_ip;
823 		hlen = sizeof(*oip);
824 		fnew.fin_v = 4;
825 		fnew.fin_p = ip->ip_p;
826 		fnew.fin_plen = ntohs(ip->ip_len);
827 		HTONS(ip->ip_len);
828 		IP_HL_A(ip, sizeof(*oip) >> 2);
829 		ip->ip_tos = oip->ip_tos;
830 		ip->ip_id = ipf_nextipid(fin);
831 		ip->ip_off = htons(ip_mtudisc ? IP_DF : 0);
832 		ip->ip_ttl = ip_defttl;
833 		ip->ip_sum = 0;
834 		break;
835 #endif
836 #ifdef USE_INET6
837 	case 6 :
838 	{
839 		ip6_t *ip6 = (ip6_t *)ip;
840 
841 		ip6->ip6_vfc = 0x60;
842 		ip6->ip6_hlim = IPDEFTTL;
843 
844 		hlen = sizeof(*ip6);
845 		fnew.fin_p = ip6->ip6_nxt;
846 		fnew.fin_v = 6;
847 		fnew.fin_plen = ntohs(ip6->ip6_plen) + hlen;
848 		break;
849 	}
850 #endif
851 	default :
852 		return EINVAL;
853 	}
854 #ifdef KAME_IPSEC
855 	m_reset_rcvif(m);
856 #endif
857 
858 	fnew.fin_ifp = fin->fin_ifp;
859 	fnew.fin_flx = FI_NOCKSUM;
860 	fnew.fin_m = m;
861 	fnew.fin_ip = ip;
862 	fnew.fin_mp = &m;
863 	fnew.fin_hlen = hlen;
864 	fnew.fin_dp = (char *)ip + hlen;
865 	(void) ipf_makefrip(hlen, ip, &fnew);
866 
867 	return ipf_fastroute(m, &m, &fnew, NULL);
868 }
869 
870 
871 int
ipf_send_icmp_err(int type,fr_info_t * fin,int dst)872 ipf_send_icmp_err(int type, fr_info_t *fin, int dst)
873 {
874 	int err, hlen, xtra, iclen, ohlen, avail;
875 	struct in_addr dst4;
876 	struct icmp *icmp;
877 	struct mbuf *m;
878 	i6addr_t dst6;
879 	void *ifp;
880 #ifdef USE_INET6
881 	int code;
882 	ip6_t *ip6;
883 #endif
884 	ip_t *ip, *ip2;
885 
886 	if ((type < 0) || (type > ICMP_MAXTYPE))
887 		return -1;
888 
889 #ifdef USE_INET6
890 	code = fin->fin_icode;
891 	if ((code < 0) || (code >= sizeof(icmptoicmp6unreach)/sizeof(int)))
892 		return -1;
893 #endif
894 
895 	if (ipf_checkl4sum(fin) == -1)
896 		return -1;
897 #ifdef MGETHDR
898 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
899 #else
900 	MGET(m, M_DONTWAIT, MT_HEADER);
901 #endif
902 	if (m == NULL)
903 		return -1;
904 	avail = MHLEN;
905 
906 	xtra = 0;
907 	hlen = 0;
908 	ohlen = 0;
909 	dst4.s_addr = 0;
910 	ifp = fin->fin_ifp;
911 	if (fin->fin_v == 4) {
912 		if ((fin->fin_p == IPPROTO_ICMP) && !(fin->fin_flx & FI_SHORT))
913 			switch (ntohs(fin->fin_data[0]) >> 8)
914 			{
915 			case ICMP_ECHO :
916 			case ICMP_TSTAMP :
917 			case ICMP_IREQ :
918 			case ICMP_MASKREQ :
919 				break;
920 			default :
921 				FREE_MB_T(m);
922 				return 0;
923 			}
924 
925 		if (dst == 0) {
926 			if (ipf_ifpaddr(&ipfmain, 4, FRI_NORMAL, ifp,
927 				       &dst6, NULL) == -1) {
928 				FREE_MB_T(m);
929 				return -1;
930 			}
931 			dst4 = dst6.in4;
932 		} else
933 			dst4.s_addr = fin->fin_daddr;
934 
935 		hlen = sizeof(ip_t);
936 		ohlen = fin->fin_hlen;
937 		iclen = hlen + offsetof(struct icmp, icmp_ip) + ohlen;
938 		if (fin->fin_hlen < fin->fin_plen)
939 			xtra = MIN(fin->fin_dlen, 8);
940 		else
941 			xtra = 0;
942 	}
943 
944 #ifdef USE_INET6
945 	else if (fin->fin_v == 6) {
946 		hlen = sizeof(ip6_t);
947 		ohlen = sizeof(ip6_t);
948 		iclen = hlen + offsetof(struct icmp, icmp_ip) + ohlen;
949 		type = icmptoicmp6types[type];
950 		if (type == ICMP6_DST_UNREACH)
951 			code = icmptoicmp6unreach[code];
952 
953 		if (iclen + max_linkhdr + fin->fin_plen > avail) {
954 			MCLGET(m, M_DONTWAIT);
955 			if (m == NULL)
956 				return -1;
957 			if ((m->m_flags & M_EXT) == 0) {
958 				FREE_MB_T(m);
959 				return -1;
960 			}
961 			avail = MCLBYTES;
962 		}
963 		xtra = MIN(fin->fin_plen, avail - iclen - max_linkhdr);
964 		xtra = MIN(xtra, IPV6_MMTU - iclen);
965 		if (dst == 0) {
966 			if (ipf_ifpaddr(&ipfmain, 6, FRI_NORMAL, ifp,
967 				       &dst6, NULL) == -1) {
968 				FREE_MB_T(m);
969 				return -1;
970 			}
971 		} else
972 			dst6 = fin->fin_dst6;
973 	}
974 #endif
975 	else {
976 		FREE_MB_T(m);
977 		return -1;
978 	}
979 
980 	avail -= (max_linkhdr + iclen);
981 	if (avail < 0) {
982 		FREE_MB_T(m);
983 		return -1;
984 	}
985 	if (xtra > avail)
986 		xtra = avail;
987 	iclen += xtra;
988 	m->m_data += max_linkhdr;
989 	m_reset_rcvif(m);
990 	m->m_pkthdr.len = iclen;
991 	m->m_len = iclen;
992 	ip = mtod(m, ip_t *);
993 	icmp = (struct icmp *)((char *)ip + hlen);
994 	ip2 = (ip_t *)&icmp->icmp_ip;
995 
996 	icmp->icmp_type = type;
997 	icmp->icmp_code = fin->fin_icode;
998 	icmp->icmp_cksum = 0;
999 #ifdef icmp_nextmtu
1000 	if (type == ICMP_UNREACH && fin->fin_icode == ICMP_UNREACH_NEEDFRAG) {
1001 		if (fin->fin_mtu != 0) {
1002 			icmp->icmp_nextmtu = htons(fin->fin_mtu);
1003 
1004 		} else if (ifp != NULL) {
1005 			icmp->icmp_nextmtu = htons(GETIFMTU_4(ifp));
1006 
1007 		} else {	/* make up a number... */
1008 			icmp->icmp_nextmtu = htons(fin->fin_plen - 20);
1009 		}
1010 	}
1011 #endif
1012 
1013 	bcopy((char *)fin->fin_ip, (char *)ip2, ohlen);
1014 
1015 #if defined(M_CSUM_IPv4)
1016 	/*
1017 	 * Clear any in-bound checksum flags for this packet.
1018 	 */
1019 	m->m_pkthdr.csuminfo = 0;
1020 #endif /* __NetBSD__ && M_CSUM_IPv4 */
1021 
1022 #ifdef USE_INET6
1023 	ip6 = (ip6_t *)ip;
1024 	if (fin->fin_v == 6) {
1025 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
1026 		ip6->ip6_plen = htons(iclen - hlen);
1027 		ip6->ip6_nxt = IPPROTO_ICMPV6;
1028 		ip6->ip6_hlim = 0;
1029 		ip6->ip6_src = dst6.in6;
1030 		ip6->ip6_dst = fin->fin_src6.in6;
1031 		if (xtra > 0)
1032 			bcopy((char *)fin->fin_ip + ohlen,
1033 			      (char *)&icmp->icmp_ip + ohlen, xtra);
1034 		icmp->icmp_cksum = in6_cksum(m, IPPROTO_ICMPV6,
1035 					     sizeof(*ip6), iclen - hlen);
1036 	} else
1037 #endif
1038 	{
1039 		ip->ip_p = IPPROTO_ICMP;
1040 		ip->ip_src.s_addr = dst4.s_addr;
1041 		ip->ip_dst.s_addr = fin->fin_saddr;
1042 
1043 		if (xtra > 0)
1044 			bcopy((char *)fin->fin_ip + ohlen,
1045 			      (char *)&icmp->icmp_ip + ohlen, xtra);
1046 		icmp->icmp_cksum = ipf_cksum((u_short *)icmp,
1047 					     sizeof(*icmp) + 8);
1048 		ip->ip_len = iclen;
1049 		ip->ip_p = IPPROTO_ICMP;
1050 	}
1051 	err = ipf_send_ip(fin, m);
1052 	return err;
1053 }
1054 
1055 
1056 /*
1057  * m0 - pointer to mbuf where the IP packet starts
1058  * mpp - pointer to the mbuf pointer that is the start of the mbuf chain
1059  */
1060 int
ipf_fastroute(mb_t * m0,mb_t ** mpp,fr_info_t * fin,frdest_t * fdp)1061 ipf_fastroute(mb_t *m0, mb_t **mpp, fr_info_t *fin, frdest_t *fdp)
1062 {
1063 	register struct ip *ip, *mhip;
1064 	register struct mbuf *m = *mpp;
1065 	register struct route *ro;
1066 	int len, off, error = 0, hlen, code;
1067 	struct ifnet *ifp, *sifp;
1068 	ipf_main_softc_t *softc;
1069 #if __NetBSD_Version__ >= 499001100
1070 	union {
1071 		struct sockaddr         dst;
1072 		struct sockaddr_in      dst4;
1073 	} u;
1074 #else
1075 	struct sockaddr_in *dst4;
1076 #endif
1077 	struct sockaddr *dst;
1078 	u_short ip_off, ip_len;
1079 	struct route iproute;
1080 	struct rtentry *rt;
1081 	frdest_t node;
1082 	frentry_t *fr;
1083 
1084 	if (fin->fin_v == 6) {
1085 #ifdef USE_INET6
1086 		error = ipf_fastroute6(m0, mpp, fin, fdp);
1087 #else
1088 		error = EPROTONOSUPPORT;
1089 #endif
1090 		if ((error != 0) && (*mpp != NULL))
1091 			FREE_MB_T(*mpp);
1092 		return error;
1093 	}
1094 #ifndef INET
1095 	FREE_MB_T(*mpp);
1096 	return EPROTONOSUPPORT;
1097 #else
1098 
1099 	hlen = fin->fin_hlen;
1100 	ip = mtod(m0, struct ip *);
1101 	softc = fin->fin_main_soft;
1102 	rt = NULL;
1103 	ifp = NULL;
1104 
1105 # if defined(M_CSUM_IPv4)
1106 	/*
1107 	 * Clear any in-bound checksum flags for this packet.
1108 	 */
1109 	m0->m_pkthdr.csuminfo = 0;
1110 # endif /* __NetBSD__ && M_CSUM_IPv4 */
1111 
1112 	/*
1113 	 * Route packet.
1114 	 */
1115 	ro = &iproute;
1116 	memset(ro, 0, sizeof(*ro));
1117 	fr = fin->fin_fr;
1118 
1119 	if ((fr != NULL) && !(fr->fr_flags & FR_KEEPSTATE) && (fdp != NULL) &&
1120 	    (fdp->fd_type == FRD_DSTLIST)) {
1121 		if (ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL, &node) == 0)
1122 			fdp = &node;
1123 	}
1124 	if (fdp != NULL)
1125 		ifp = fdp->fd_ptr;
1126 	else
1127 		ifp = fin->fin_ifp;
1128 
1129 	if ((ifp == NULL) && ((fr == NULL) || !(fr->fr_flags & FR_FASTROUTE))) {
1130 		error = -2;
1131 		goto bad;
1132 	}
1133 
1134 # if __NetBSD_Version__ >= 499001100
1135 	if ((fdp != NULL) && (fdp->fd_ip.s_addr != 0))
1136 		sockaddr_in_init(&u.dst4, &fdp->fd_ip, 0);
1137 	else
1138 		sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1139 	dst = &u.dst;
1140 	rtcache_setdst(ro, dst);
1141 	rt = rtcache_init(ro);
1142 # else
1143 	dst4 = (struct sockaddr_in *)&ro->ro_dst;
1144 	dst = (struct sockaddr *)dst4;
1145 	dst4->sin_family = AF_INET;
1146 	dst4->sin_addr = ip->ip_dst;
1147 
1148 	if ((fdp != NULL) && (fdp->fd_ip.s_addr != 0))
1149 		dst4->sin_addr = fdp->fd_ip;
1150 
1151 	dst4->sin_len = sizeof(*dst);
1152 	rtalloc(ro);
1153 	rt = ro->ro_rt;
1154 # endif
1155 	if ((ifp == NULL) && (rt != NULL))
1156 		ifp = rt->rt_ifp;
1157 	if ((rt == NULL) || (ifp == NULL)) {
1158 #ifdef INET
1159 		if (in_localaddr(ip->ip_dst))
1160 			error = EHOSTUNREACH;
1161 		else
1162 #endif
1163 			error = ENETUNREACH;
1164 		goto bad;
1165 	}
1166 
1167 
1168 	if (rt->rt_flags & RTF_GATEWAY)
1169 		dst = rt->rt_gateway;
1170 
1171 	rt->rt_use++;
1172 
1173 	/*
1174 	 * For input packets which are being "fastrouted", they won't
1175 	 * go back through output filtering and miss their chance to get
1176 	 * NAT'd and counted.  Duplicated packets aren't considered to be
1177 	 * part of the normal packet stream, so do not NAT them or pass
1178 	 * them through stateful checking, etc.
1179 	 */
1180 	if ((fdp != &fr->fr_dif) && (fin->fin_out == 0)) {
1181 		sifp = fin->fin_ifp;
1182 		fin->fin_ifp = ifp;
1183 		fin->fin_out = 1;
1184 		(void) ipf_acctpkt(fin, NULL);
1185 		fin->fin_fr = NULL;
1186 		if (!fr || !(fr->fr_flags & FR_RETMASK)) {
1187 			u_32_t pass;
1188 
1189 			(void) ipf_state_check(fin, &pass);
1190 		}
1191 
1192 		switch (ipf_nat_checkout(fin, NULL))
1193 		{
1194 		case 0 :
1195 			break;
1196 		case 1 :
1197 			ip->ip_sum = 0;
1198 			break;
1199 		case -1 :
1200 			error = -1;
1201 			goto bad;
1202 			break;
1203 		}
1204 
1205 		fin->fin_ifp = sifp;
1206 		fin->fin_out = 0;
1207 	} else
1208 		ip->ip_sum = 0;
1209 	/*
1210 	 * If small enough for interface, can just send directly.
1211 	 */
1212 	m_set_rcvif(m, ifp);
1213 
1214 	ip_len = ntohs(ip->ip_len);
1215 	if (ip_len <= ifp->if_mtu) {
1216 # if defined(M_CSUM_IPv4)
1217 #  if (__NetBSD_Version__ >= 105009999)
1218 		if (ifp->if_csum_flags_tx & M_CSUM_IPv4)
1219 			m->m_pkthdr.csuminfo |= M_CSUM_IPv4;
1220 #  else
1221 		if (ifp->if_capabilities & IFCAP_CSUM_IPv4)
1222 			m->m_pkthdr.csuminfo |= M_CSUM_IPv4;
1223 #  endif /* (__NetBSD_Version__ >= 105009999) */
1224 		else if (ip->ip_sum == 0)
1225 			ip->ip_sum = in_cksum(m, hlen);
1226 # else
1227 		if (!ip->ip_sum)
1228 			ip->ip_sum = in_cksum(m, hlen);
1229 # endif /* M_CSUM_IPv4 */
1230 
1231 		error = if_output_lock(ifp, ifp, m, dst, rt);
1232 		goto done;
1233 	}
1234 
1235 	/*
1236 	 * Too large for interface; fragment if possible.
1237 	 * Must be able to put at least 8 bytes per fragment.
1238 	 */
1239 	ip_off = ntohs(ip->ip_off);
1240 	if (ip_off & IP_DF) {
1241 		error = EMSGSIZE;
1242 		goto bad;
1243 	}
1244 	len = (ifp->if_mtu - hlen) &~ 7;
1245 	if (len < 8) {
1246 		error = EMSGSIZE;
1247 		goto bad;
1248 	}
1249 
1250     {
1251 	int mhlen, firstlen = len;
1252 	struct mbuf **mnext = &m->m_act;
1253 
1254 	/*
1255 	 * Loop through length of segment after first fragment,
1256 	 * make new header and copy data of each part and link onto chain.
1257 	 */
1258 	m0 = m;
1259 	mhlen = sizeof (struct ip);
1260 	for (off = hlen + len; off < ip_len; off += len) {
1261 # ifdef MGETHDR
1262 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
1263 # else
1264 		MGET(m, M_DONTWAIT, MT_HEADER);
1265 # endif
1266 		if (m == 0) {
1267 			m = m0;
1268 			error = ENOBUFS;
1269 			goto bad;
1270 		}
1271 		m->m_data += max_linkhdr;
1272 		mhip = mtod(m, struct ip *);
1273 		bcopy((char *)ip, (char *)mhip, sizeof(*ip));
1274 #ifdef INET
1275 		if (hlen > sizeof (struct ip)) {
1276 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1277 			IP_HL_A(mhip, mhlen >> 2);
1278 		}
1279 #endif
1280 		m->m_len = mhlen;
1281 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1282 		if (off + len >= ip_len)
1283 			len = ip_len - off;
1284 		else
1285 			mhip->ip_off |= IP_MF;
1286 		mhip->ip_len = htons((u_short)(len + mhlen));
1287 		m->m_next = m_copy(m0, off, len);
1288 		if (m->m_next == 0) {
1289 			error = ENOBUFS;	/* ??? */
1290 			goto sendorfree;
1291 		}
1292 		m->m_pkthdr.len = mhlen + len;
1293 		m_reset_rcvif(m);
1294 		mhip->ip_off = htons((u_short)mhip->ip_off);
1295 		mhip->ip_sum = 0;
1296 #ifdef INET
1297 		mhip->ip_sum = in_cksum(m, mhlen);
1298 #endif
1299 		*mnext = m;
1300 		mnext = &m->m_act;
1301 	}
1302 	/*
1303 	 * Update first fragment by trimming what's been copied out
1304 	 * and updating header, then send each fragment (in order).
1305 	 */
1306 	m_adj(m0, hlen + firstlen - ip_len);
1307 	ip->ip_len = htons((u_short)(hlen + firstlen));
1308 	ip->ip_off = htons((u_short)IP_MF);
1309 	ip->ip_sum = 0;
1310 #ifdef INET
1311 	ip->ip_sum = in_cksum(m0, hlen);
1312 #endif
1313 sendorfree:
1314 	for (m = m0; m; m = m0) {
1315 		m0 = m->m_act;
1316 		m->m_act = 0;
1317 		if (error == 0) {
1318 			KERNEL_LOCK(1, NULL);
1319 			error = (*ifp->if_output)(ifp, m, dst, rt);
1320 			KERNEL_UNLOCK_ONE(NULL);
1321 		} else {
1322 			FREE_MB_T(m);
1323 		}
1324 	}
1325     }
1326 done:
1327 	if (!error)
1328 		softc->ipf_frouteok[0]++;
1329 	else
1330 		softc->ipf_frouteok[1]++;
1331 
1332 # if __NetBSD_Version__ >= 499001100
1333 	rtcache_free(ro);
1334 # else
1335 	if (rt) {
1336 		RTFREE(rt);
1337 	}
1338 # endif
1339 	return error;
1340 bad:
1341 	if (error == EMSGSIZE) {
1342 		sifp = fin->fin_ifp;
1343 		code = fin->fin_icode;
1344 		fin->fin_icode = ICMP_UNREACH_NEEDFRAG;
1345 		fin->fin_ifp = ifp;
1346 		(void) ipf_send_icmp_err(ICMP_UNREACH, fin, 1);
1347 		fin->fin_ifp = sifp;
1348 		fin->fin_icode = code;
1349 	}
1350 	FREE_MB_T(m);
1351 	goto done;
1352 #endif /* INET */
1353 }
1354 
1355 
1356 #if defined(USE_INET6)
1357 /*
1358  * This is the IPv6 specific fastroute code.  It doesn't clean up the mbuf's
1359  * or ensure that it is an IPv6 packet that is being forwarded, those are
1360  * expected to be done by the called (ipf_fastroute).
1361  */
1362 static int
ipf_fastroute6(struct mbuf * m0,struct mbuf ** mpp,fr_info_t * fin,frdest_t * fdp)1363 ipf_fastroute6(struct mbuf *m0, struct mbuf **mpp, fr_info_t *fin,
1364     frdest_t *fdp)
1365 {
1366 # if __NetBSD_Version__ >= 499001100
1367 	struct route ip6route;
1368 	const struct sockaddr *dst;
1369 	union {
1370 		struct sockaddr         dst;
1371 		struct sockaddr_in6     dst6;
1372 	} u;
1373 	struct route *ro;
1374 # else
1375 	struct route_in6 ip6route;
1376 	struct sockaddr_in6 *dst6;
1377 	struct route_in6 *ro;
1378 # endif
1379 	struct rtentry *rt;
1380 	struct ifnet *ifp;
1381 	u_long mtu;
1382 	int error;
1383 
1384 	error = 0;
1385 	ro = &ip6route;
1386 
1387 	if (fdp != NULL)
1388 		ifp = fdp->fd_ptr;
1389 	else
1390 		ifp = fin->fin_ifp;
1391 	memset(ro, 0, sizeof(*ro));
1392 # if __NetBSD_Version__ >= 499001100
1393 	if (fdp != NULL && IP6_NOTZERO(&fdp->fd_ip6))
1394 		sockaddr_in6_init(&u.dst6, &fdp->fd_ip6.in6, 0, 0, 0);
1395 	else
1396 		sockaddr_in6_init(&u.dst6, &fin->fin_fi.fi_dst.in6, 0, 0, 0);
1397 	dst = &u.dst;
1398 	rtcache_setdst(ro, dst);
1399 
1400 	rt = rtcache_init(ro);
1401 	if ((ifp == NULL) && (rt != NULL))
1402 		ifp = rt->rt_ifp;
1403 # else
1404 	dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
1405 	dst6->sin6_family = AF_INET6;
1406 	dst6->sin6_len = sizeof(struct sockaddr_in6);
1407 	dst6->sin6_addr = fin->fin_fi.fi_dst.in6;
1408 
1409 	if (fdp != NULL) {
1410 		if (IP6_NOTZERO(&fdp->fd_ip6))
1411 			dst6->sin6_addr = fdp->fd_ip6.in6;
1412 	}
1413 
1414 	rtalloc((struct route *)ro);
1415 
1416 	if ((ifp == NULL) && (ro->ro_rt != NULL))
1417 		ifp = ro->ro_rt->rt_ifp;
1418 	rt = ro->ro_rt;
1419 # endif
1420 	if ((rt == NULL) || (ifp == NULL)) {
1421 
1422 		error = EHOSTUNREACH;
1423 		goto bad;
1424 	}
1425 
1426 	/* KAME */
1427 # if __NetBSD_Version__ >= 499001100
1428 	if (IN6_IS_ADDR_LINKLOCAL(&u.dst6.sin6_addr))
1429 		u.dst6.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1430 # else
1431 	if (IN6_IS_ADDR_LINKLOCAL(&dst6->sin6_addr))
1432 		dst6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1433 # endif
1434 
1435 	{
1436 # if (__NetBSD_Version__ >= 106010000) && !defined(IN6_LINKMTU)
1437 		struct in6_ifextra *ife;
1438 # endif
1439 		if (rt->rt_flags & RTF_GATEWAY)
1440 # if __NetBSD_Version__ >= 499001100
1441 			dst = rt->rt_gateway;
1442 # else
1443 			dst6 = (struct sockaddr_in6 *)rt->rt_gateway;
1444 # endif
1445 		rt->rt_use++;
1446 
1447 		/* Determine path MTU. */
1448 # if (__NetBSD_Version__ <= 106009999)
1449 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
1450 # else
1451 #  ifdef IN6_LINKMTU
1452 		mtu = IN6_LINKMTU(ifp);
1453 #  else
1454 		ife = (struct in6_ifextra *)(ifp)->if_afdata[AF_INET6];
1455 		mtu = ife->nd_ifinfo[ifp->if_index].linkmtu;
1456 #  endif
1457 # endif
1458 		if ((error == 0) && (m0->m_pkthdr.len <= mtu)) {
1459 # if __NetBSD_Version__ >= 499001100
1460 			error = nd6_output(ifp, ifp, m0, satocsin6(dst), rt);
1461 # else
1462 			error = nd6_output(ifp, ifp, m0, dst6, rt);
1463 # endif
1464 		} else {
1465 			error = EMSGSIZE;
1466 		}
1467 	}
1468 bad:
1469 # if __NetBSD_Version__ >= 499001100
1470 	rtcache_free(ro);
1471 # else
1472 	if (ro->ro_rt != NULL) {
1473 		RTFREE(((struct route *)ro)->ro_rt);
1474 	}
1475 # endif
1476 	return error;
1477 }
1478 #endif	/* INET6 */
1479 
1480 
1481 int
ipf_verifysrc(fr_info_t * fin)1482 ipf_verifysrc(fr_info_t *fin)
1483 {
1484 #if __NetBSD_Version__ >= 499001100
1485 	union {
1486 		struct sockaddr         dst;
1487 		struct sockaddr_in      dst4;
1488 	} u;
1489 	struct rtentry *rt;
1490 #else
1491 	struct sockaddr_in *dst;
1492 #endif
1493 	struct route iproute;
1494 	int rc;
1495 
1496 #if __NetBSD_Version__ >= 499001100
1497 	sockaddr_in_init(&u.dst4, &fin->fin_src, 0);
1498 	rtcache_setdst(&iproute, &u.dst);
1499 	rt = rtcache_init(&iproute);
1500 	if (rt == NULL)
1501 		rc = 0;
1502 	else
1503 		rc = (fin->fin_ifp == rt->rt_ifp);
1504 	rtcache_free(&iproute);
1505 #else
1506 	dst = (struct sockaddr_in *)&iproute.ro_dst;
1507 	dst->sin_len = sizeof(*dst);
1508 	dst->sin_family = AF_INET;
1509 	dst->sin_addr = fin->fin_src;
1510 	rtalloc(&iproute);
1511 	if (iproute.ro_rt == NULL)
1512 		return 0;
1513 	rc = (fin->fin_ifp == iproute.ro_rt->rt_ifp);
1514 	RTFREE(iproute.ro_rt);
1515 #endif
1516 	return rc;
1517 }
1518 
1519 
1520 /*
1521  * return the first IP Address associated with an interface
1522  */
1523 int
ipf_ifpaddr(ipf_main_softc_t * softc,int v,int atype,void * ifptr,i6addr_t * inp,i6addr_t * inpmask)1524 ipf_ifpaddr(ipf_main_softc_t *softc, int v, int atype, void *ifptr,
1525     i6addr_t *inp, i6addr_t *inpmask)
1526 {
1527 #ifdef USE_INET6
1528 	struct in6_addr *inp6 = NULL;
1529 #endif
1530 	struct sockaddr *sock, *mask;
1531 	struct sockaddr_in *sin;
1532 	struct ifaddr *ifa;
1533 	struct ifnet *ifp;
1534 
1535 	if ((ifptr == NULL) || (ifptr == (void *)-1))
1536 		return -1;
1537 
1538 	ifp = ifptr;
1539 	mask = NULL;
1540 
1541 	if (v == 4)
1542 		inp->in4.s_addr = 0;
1543 #ifdef USE_INET6
1544 	else if (v == 6)
1545 		bzero((char *)inp, sizeof(*inp));
1546 #endif
1547 
1548 	ifa = IFADDR_READER_FIRST(ifp);
1549 	sock = ifa ? ifa->ifa_addr : NULL;
1550 	while (sock != NULL && ifa != NULL) {
1551 		sin = (struct sockaddr_in *)sock;
1552 		if ((v == 4) && (sin->sin_family == AF_INET))
1553 			break;
1554 #ifdef USE_INET6
1555 		if ((v == 6) && (sin->sin_family == AF_INET6)) {
1556 			inp6 = &((struct sockaddr_in6 *)sin)->sin6_addr;
1557 			if (!IN6_IS_ADDR_LINKLOCAL(inp6) &&
1558 			    !IN6_IS_ADDR_LOOPBACK(inp6))
1559 				break;
1560 		}
1561 #endif
1562 		ifa = IFADDR_READER_NEXT(ifa);
1563 		if (ifa != NULL)
1564 			sock = ifa->ifa_addr;
1565 	}
1566 	if (ifa == NULL || sock == NULL)
1567 		return -1;
1568 
1569 	mask = ifa->ifa_netmask;
1570 	if (atype == FRI_BROADCAST)
1571 		sock = ifa->ifa_broadaddr;
1572 	else if (atype == FRI_PEERADDR)
1573 		sock = ifa->ifa_dstaddr;
1574 
1575 #ifdef USE_INET6
1576 	if (v == 6)
1577 		return ipf_ifpfillv6addr(atype, (struct sockaddr_in6 *)sock,
1578 					(struct sockaddr_in6 *)mask,
1579 					inp, inpmask);
1580 #endif
1581 	return ipf_ifpfillv4addr(atype, (struct sockaddr_in *)sock,
1582 				(struct sockaddr_in *)mask,
1583 				&inp->in4, &inpmask->in4);
1584 }
1585 
1586 
1587 u_32_t
ipf_newisn(fr_info_t * fin)1588 ipf_newisn(fr_info_t *fin)
1589 {
1590 #if __NetBSD_Version__ >= 105190000	/* 1.5T */
1591 	size_t asz;
1592 
1593 	if (fin->fin_v == 4)
1594 		asz = sizeof(struct in_addr);
1595 	else if (fin->fin_v == 6)
1596 		asz = sizeof(fin->fin_src);
1597 	else	/* XXX: no way to return error */
1598 		return 0;
1599 #ifdef INET
1600 	return tcp_new_iss1((void *)&fin->fin_src, (void *)&fin->fin_dst,
1601 			    fin->fin_sport, fin->fin_dport, asz, 0);
1602 #else
1603 	return ENOSYS;
1604 #endif
1605 #else
1606 	static int iss_seq_off = 0;
1607 	u_char hash[16];
1608 	u_32_t newiss;
1609 	MD5_CTX ctx;
1610 
1611 	/*
1612 	 * Compute the base value of the ISS.  It is a hash
1613 	 * of (saddr, sport, daddr, dport, secret).
1614 	 */
1615 	MD5Init(&ctx);
1616 
1617 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_src,
1618 		  sizeof(fin->fin_fi.fi_src));
1619 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_dst,
1620 		  sizeof(fin->fin_fi.fi_dst));
1621 	MD5Update(&ctx, (u_char *) &fin->fin_dat, sizeof(fin->fin_dat));
1622 
1623 	MD5Update(&ctx, ipf_iss_secret, sizeof(ipf_iss_secret));
1624 
1625 	MD5Final(hash, &ctx);
1626 
1627 	memcpy(&newiss, hash, sizeof(newiss));
1628 
1629 	/*
1630 	 * Now increment our "timer", and add it in to
1631 	 * the computed value.
1632 	 *
1633 	 * XXX Use `addin'?
1634 	 * XXX TCP_ISSINCR too large to use?
1635 	 */
1636 	iss_seq_off += 0x00010000;
1637 	newiss += iss_seq_off;
1638 	return newiss;
1639 #endif
1640 }
1641 
1642 
1643 /* ------------------------------------------------------------------------ */
1644 /* Function:    ipf_nextipid                                                 */
1645 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
1646 /* Parameters:  fin(I) - pointer to packet information                      */
1647 /*                                                                          */
1648 /* Returns the next IPv4 ID to use for this packet.                         */
1649 /* ------------------------------------------------------------------------ */
1650 u_short
ipf_nextipid(fr_info_t * fin)1651 ipf_nextipid(fr_info_t *fin)
1652 {
1653 #ifdef USE_MUTEXES
1654 	ipf_main_softc_t *softc = fin->fin_main_soft;
1655 #endif
1656 	u_short id;
1657 
1658 	MUTEX_ENTER(&softc->ipf_rw);
1659 	id = ipid++;
1660 	MUTEX_EXIT(&softc->ipf_rw);
1661 
1662 	return id;
1663 }
1664 
1665 
1666 EXTERN_INLINE int
ipf_checkv4sum(fr_info_t * fin)1667 ipf_checkv4sum(fr_info_t *fin)
1668 {
1669 #ifdef M_CSUM_TCP_UDP_BAD
1670 	int manual, pflag, cflags, active;
1671 	mb_t *m;
1672 
1673 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
1674 		return 0;
1675 
1676 	if ((fin->fin_flx & FI_SHORT) != 0)
1677 		return 1;
1678 
1679 	if (fin->fin_cksum != FI_CK_NEEDED)
1680 		return (fin->fin_cksum > FI_CK_NEEDED) ? 0 : -1;
1681 
1682 	manual = 0;
1683 	m = fin->fin_m;
1684 	if (m == NULL) {
1685 		manual = 1;
1686 		goto skipauto;
1687 	}
1688 
1689 	switch (fin->fin_p)
1690 	{
1691 	case IPPROTO_UDP :
1692 		pflag = M_CSUM_UDPv4;
1693 		break;
1694 	case IPPROTO_TCP :
1695 		pflag = M_CSUM_TCPv4;
1696 		break;
1697 	default :
1698 		pflag = 0;
1699 		manual = 1;
1700 		break;
1701 	}
1702 
1703 	active = ((struct ifnet *)fin->fin_ifp)->if_csum_flags_rx & pflag;
1704 	active |= M_CSUM_TCP_UDP_BAD | M_CSUM_DATA;
1705 	cflags = m->m_pkthdr.csum_flags & active;
1706 
1707 	if (pflag != 0) {
1708 		if (cflags == (pflag | M_CSUM_TCP_UDP_BAD)) {
1709 			fin->fin_flx |= FI_BAD;
1710 			fin->fin_cksum = FI_CK_BAD;
1711 		} else if (cflags == (pflag | M_CSUM_DATA)) {
1712 			if ((m->m_pkthdr.csum_data ^ 0xffff) != 0) {
1713 				fin->fin_flx |= FI_BAD;
1714 				fin->fin_cksum = FI_CK_BAD;
1715 			} else {
1716 				fin->fin_cksum = FI_CK_SUMOK;
1717 			}
1718 		} else if (cflags == pflag) {
1719 			fin->fin_cksum = FI_CK_SUMOK;
1720 		} else {
1721 			manual = 1;
1722 		}
1723 	}
1724 skipauto:
1725 	if (manual != 0) {
1726 		if (ipf_checkl4sum(fin) == -1) {
1727 			fin->fin_flx |= FI_BAD;
1728 			return -1;
1729 		}
1730 	}
1731 #else
1732 	if (ipf_checkl4sum(fin) == -1) {
1733 		fin->fin_flx |= FI_BAD;
1734 		return -1;
1735 	}
1736 #endif
1737 	return 0;
1738 }
1739 
1740 
1741 #ifdef USE_INET6
1742 EXTERN_INLINE int
ipf_checkv6sum(fr_info_t * fin)1743 ipf_checkv6sum(fr_info_t *fin)
1744 {
1745 # ifdef M_CSUM_TCP_UDP_BAD
1746 	int manual, pflag, cflags, active;
1747 	mb_t *m;
1748 
1749 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
1750 		return 0;
1751 
1752 	if ((fin->fin_flx & FI_SHORT) != 0)
1753 		return 1;
1754 
1755 	if (fin->fin_cksum != FI_CK_SUMOK)
1756 		return (fin->fin_cksum > FI_CK_NEEDED) ? 0 : -1;
1757 
1758 
1759 	manual = 0;
1760 	m = fin->fin_m;
1761 
1762 	switch (fin->fin_p)
1763 	{
1764 	case IPPROTO_UDP :
1765 		pflag = M_CSUM_UDPv6;
1766 		break;
1767 	case IPPROTO_TCP :
1768 		pflag = M_CSUM_TCPv6;
1769 		break;
1770 	default :
1771 		pflag = 0;
1772 		manual = 1;
1773 		break;
1774 	}
1775 
1776 	active = ((struct ifnet *)fin->fin_ifp)->if_csum_flags_rx & pflag;
1777 	active |= M_CSUM_TCP_UDP_BAD | M_CSUM_DATA;
1778 	cflags = m->m_pkthdr.csum_flags & active;
1779 
1780 	if (pflag != 0) {
1781 		if (cflags == (pflag | M_CSUM_TCP_UDP_BAD)) {
1782 			fin->fin_flx |= FI_BAD;
1783 		} else if (cflags == (pflag | M_CSUM_DATA)) {
1784 			if ((m->m_pkthdr.csum_data ^ 0xffff) != 0)
1785 				fin->fin_flx |= FI_BAD;
1786 		} else if (cflags == pflag) {
1787 			;
1788 		} else {
1789 			manual = 1;
1790 		}
1791 	}
1792 	if (manual != 0) {
1793 		if (ipf_checkl4sum(fin) == -1) {
1794 			fin->fin_flx |= FI_BAD;
1795 			return -1;
1796 		}
1797 	}
1798 # else
1799 	if (ipf_checkl4sum(fin) == -1) {
1800 		fin->fin_flx |= FI_BAD;
1801 		return -1;
1802 	}
1803 # endif
1804 	return 0;
1805 }
1806 #endif /* USE_INET6 */
1807 
1808 
1809 size_t
mbufchainlen(struct mbuf * m0)1810 mbufchainlen(struct mbuf *m0)
1811 {
1812 	size_t len;
1813 
1814 	if ((m0->m_flags & M_PKTHDR) != 0) {
1815 		len = m0->m_pkthdr.len;
1816 	} else {
1817 		struct mbuf *m;
1818 
1819 		for (m = m0, len = 0; m != NULL; m = m->m_next)
1820 			len += m->m_len;
1821 	}
1822 	return len;
1823 }
1824 
1825 
1826 /* ------------------------------------------------------------------------ */
1827 /* Function:    ipf_pullup                                                  */
1828 /* Returns:     NULL == pullup failed, else pointer to protocol header      */
1829 /* Parameters:  xmin(I)- pointer to buffer where data packet starts         */
1830 /*              fin(I) - pointer to packet information                      */
1831 /*              len(I) - number of bytes to pullup                          */
1832 /*                                                                          */
1833 /* Attempt to move at least len bytes (from the start of the buffer) into a */
1834 /* single buffer for ease of access.  Operating system native functions are */
1835 /* used to manage buffers - if necessary.  If the entire packet ends up in  */
1836 /* a single buffer, set the FI_COALESCE flag even though ipf_coalesce() has */
1837 /* not been called.  Both fin_ip and fin_dp are updated before exiting _IF_ */
1838 /* and ONLY if the pullup succeeds.                                         */
1839 /*                                                                          */
1840 /* We assume that 'xmin' is a pointer to a buffer that is part of the chain */
1841 /* of buffers that starts at *fin->fin_mp.                                  */
1842 /* ------------------------------------------------------------------------ */
1843 void *
ipf_pullup(mb_t * xmin,fr_info_t * fin,int len)1844 ipf_pullup(mb_t *xmin, fr_info_t *fin, int len)
1845 {
1846 	int dpoff, ipoff;
1847 	mb_t *m = xmin;
1848 	char *ip;
1849 
1850 	if (m == NULL)
1851 		return NULL;
1852 
1853 	ip = (char *)fin->fin_ip;
1854 	if ((fin->fin_flx & FI_COALESCE) != 0)
1855 		return ip;
1856 
1857 	ipoff = fin->fin_ipoff;
1858 	if (fin->fin_dp != NULL)
1859 		dpoff = (char *)fin->fin_dp - (char *)ip;
1860 	else
1861 		dpoff = 0;
1862 
1863 	if (M_LEN(m) < len) {
1864 		mb_t *n = *fin->fin_mp;
1865 		/*
1866 		 * Assume that M_PKTHDR is set and just work with what is left
1867 		 * rather than check..
1868 		 * Should not make any real difference, anyway.
1869 		 */
1870 		if (m != n) {
1871 			/*
1872 			 * Record the mbuf that points to the mbuf that we're
1873 			 * about to go to work on so that we can update the
1874 			 * m_next appropriately later.
1875 			 */
1876 			for (; n->m_next != m; n = n->m_next)
1877 				;
1878 		} else {
1879 			n = NULL;
1880 		}
1881 
1882 #ifdef MHLEN
1883 		if (len > MHLEN)
1884 #else
1885 		if (len > MLEN)
1886 #endif
1887 		{
1888 #ifdef HAVE_M_PULLDOWN
1889 			if (m_pulldown(m, 0, len, NULL) == NULL)
1890 				m = NULL;
1891 #else
1892 			FREE_MB_T(*fin->fin_mp);
1893 			m = NULL;
1894 			n = NULL;
1895 #endif
1896 		} else
1897 		{
1898 			m = m_pullup(m, len);
1899 		}
1900 		if (n != NULL)
1901 			n->m_next = m;
1902 		if (m == NULL) {
1903 			/*
1904 			 * When n is non-NULL, it indicates that m pointed to
1905 			 * a sub-chain (tail) of the mbuf and that the head
1906 			 * of this chain has not yet been free'd.
1907 			 */
1908 			if (n != NULL) {
1909 				FREE_MB_T(*fin->fin_mp);
1910 			}
1911 
1912 			*fin->fin_mp = NULL;
1913 			fin->fin_m = NULL;
1914 			return NULL;
1915 		}
1916 
1917 		if (n == NULL)
1918 			*fin->fin_mp = m;
1919 
1920 		while (M_LEN(m) == 0) {
1921 			m = m->m_next;
1922 		}
1923 		fin->fin_m = m;
1924 		ip = MTOD(m, char *) + ipoff;
1925 
1926 		fin->fin_ip = (ip_t *)ip;
1927 		if (fin->fin_dp != NULL)
1928 			fin->fin_dp = (char *)fin->fin_ip + dpoff;
1929 		if (fin->fin_fraghdr != NULL)
1930 			fin->fin_fraghdr = (char *)ip +
1931 					   ((char *)fin->fin_fraghdr -
1932 					    (char *)fin->fin_ip);
1933 	}
1934 
1935 	if (len == fin->fin_plen)
1936 		fin->fin_flx |= FI_COALESCE;
1937 	return ip;
1938 }
1939 
1940 
1941 int
ipf_inject(fr_info_t * fin,mb_t * m)1942 ipf_inject(fr_info_t *fin, mb_t *m)
1943 {
1944 	int error;
1945 
1946 	if (fin->fin_out == 0) {
1947 		if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1948 			FREE_MB_T(m);
1949 			error = ENOBUFS;
1950 		} else {
1951 			error = 0;
1952 		}
1953 	} else {
1954 		error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1955 	}
1956 	return error;
1957 }
1958 
1959 
1960 u_32_t
ipf_random(void)1961 ipf_random(void)
1962 {
1963 	int number;
1964 
1965 #ifdef _CPRNG_H
1966 	number = cprng_fast32();
1967 #else
1968 	number = arc4random();
1969 #endif
1970 	return number;
1971 }
1972 
1973 
1974 /*
1975  * routines below for saving IP headers to buffer
1976  */
ipfopen(dev_t dev,int flags,int devtype,PROC_T * p)1977 static int ipfopen(dev_t dev, int flags
1978 #if (NetBSD >= 199511)
1979     , int devtype, PROC_T *p
1980 #endif
1981 )
1982 {
1983 	u_int unit = GET_MINOR(dev);
1984 	int error;
1985 
1986 	if (IPL_LOGMAX < unit) {
1987 		error = ENXIO;
1988 	} else {
1989 		switch (unit)
1990 		{
1991 		case IPL_LOGIPF :
1992 		case IPL_LOGNAT :
1993 		case IPL_LOGSTATE :
1994 		case IPL_LOGAUTH :
1995 		case IPL_LOGLOOKUP :
1996 		case IPL_LOGSYNC :
1997 #ifdef IPFILTER_SCAN
1998 		case IPL_LOGSCAN :
1999 #endif
2000 			error = 0;
2001 			break;
2002 		default :
2003 			error = ENXIO;
2004 			break;
2005 		}
2006 	}
2007 #if (__NetBSD_Version__ >= 799003000)
2008 	if (error == 0) {
2009 		mutex_enter(&ipf_ref_mutex);
2010 		ipf_active = 1;
2011 		mutex_exit(&ipf_ref_mutex);
2012 	}
2013 #endif
2014 	return error;
2015 }
2016 
2017 
ipfclose(dev_t dev,int flags,int devtype,PROC_T * p)2018 static int ipfclose(dev_t dev, int flags
2019 #if (NetBSD >= 199511)
2020 	, int devtype, PROC_T *p
2021 #endif
2022 )
2023 {
2024 	u_int	unit = GET_MINOR(dev);
2025 
2026 	if (IPL_LOGMAX < unit)
2027 		return ENXIO;
2028 	else {
2029 #if (__NetBSD_Version__ >= 799003000)
2030 		mutex_enter(&ipf_ref_mutex);
2031 		ipf_active = 0;
2032 		mutex_exit(&ipf_ref_mutex);
2033 #endif
2034 		return 0;
2035 	}
2036 }
2037 
2038 /*
2039  * ipfread/ipflog
2040  * both of these must operate with at least splnet() lest they be
2041  * called during packet processing and cause an inconsistancy to appear in
2042  * the filter lists.
2043  */
ipfread(dev_t dev,struct uio * uio,int ioflag)2044 static int ipfread(dev_t dev, struct uio *uio, int ioflag)
2045 {
2046 
2047 	if (ipfmain.ipf_running < 1) {
2048 		ipfmain.ipf_interror = 130006;
2049 		return EIO;
2050 	}
2051 
2052 	if (GET_MINOR(dev) == IPL_LOGSYNC)
2053 		return ipf_sync_read(&ipfmain, uio);
2054 
2055 #ifdef IPFILTER_LOG
2056 	return ipf_log_read(&ipfmain, GET_MINOR(dev), uio);
2057 #else
2058 	ipfmain.ipf_interror = 130007;
2059 	return ENXIO;
2060 #endif
2061 }
2062 
2063 
2064 /*
2065  * ipfwrite
2066  * both of these must operate with at least splnet() lest they be
2067  * called during packet processing and cause an inconsistancy to appear in
2068  * the filter lists.
2069  */
ipfwrite(dev_t dev,struct uio * uio,int ioflag)2070 static int ipfwrite(dev_t dev, struct uio *uio, int ioflag)
2071 {
2072 
2073 	if (ipfmain.ipf_running < 1) {
2074 		ipfmain.ipf_interror = 130008;
2075 		return EIO;
2076 	}
2077 
2078 	if (GET_MINOR(dev) == IPL_LOGSYNC)
2079 		return ipf_sync_write(&ipfmain, uio);
2080 	ipfmain.ipf_interror = 130009;
2081 	return ENXIO;
2082 }
2083 
2084 
ipfpoll(dev_t dev,int events,PROC_T * p)2085 static int ipfpoll(dev_t dev, int events, PROC_T *p)
2086 {
2087 	u_int unit = GET_MINOR(dev);
2088 	int revents = 0;
2089 
2090 	if (IPL_LOGMAX < unit) {
2091 		ipfmain.ipf_interror = 130010;
2092 		return ENXIO;
2093 	}
2094 
2095 	switch (unit)
2096 	{
2097 	case IPL_LOGIPF :
2098 	case IPL_LOGNAT :
2099 	case IPL_LOGSTATE :
2100 #ifdef IPFILTER_LOG
2101 		if ((events & (POLLIN | POLLRDNORM)) &&
2102 		    ipf_log_canread(&ipfmain, unit))
2103 			revents |= events & (POLLIN | POLLRDNORM);
2104 #endif
2105 		break;
2106 	case IPL_LOGAUTH :
2107 		if ((events & (POLLIN | POLLRDNORM)) &&
2108 		    ipf_auth_waiting(&ipfmain))
2109 			revents |= events & (POLLIN | POLLRDNORM);
2110 		break;
2111 	case IPL_LOGSYNC :
2112 		if ((events & (POLLIN | POLLRDNORM)) &&
2113 		    ipf_sync_canread(&ipfmain))
2114 			revents |= events & (POLLIN | POLLRDNORM);
2115 		if ((events & (POLLOUT | POLLWRNORM)) &&
2116 		    ipf_sync_canwrite(&ipfmain))
2117 			revents |= events & (POLLOUT | POLLWRNORM);
2118 		break;
2119 	case IPL_LOGSCAN :
2120 	case IPL_LOGLOOKUP :
2121 	default :
2122 		break;
2123 	}
2124 
2125 	if ((revents == 0) && (((events & (POLLIN|POLLRDNORM)) != 0)))
2126 		selrecord(p, &ipfmain.ipf_selwait[unit]);
2127 	return revents;
2128 }
2129 
2130 u_int
ipf_pcksum(fr_info_t * fin,int hlen,u_int sum)2131 ipf_pcksum(fr_info_t *fin, int hlen, u_int sum)
2132 {
2133 	struct mbuf *m;
2134 	u_int sum2;
2135 	int off;
2136 
2137 	m = fin->fin_m;
2138 	off = (char *)fin->fin_dp - (char *)fin->fin_ip;
2139 	m->m_data += hlen;
2140 	m->m_len -= hlen;
2141 	sum2 = in_cksum(fin->fin_m, fin->fin_plen - off);
2142 	m->m_len += hlen;
2143 	m->m_data -= hlen;
2144 
2145 	/*
2146 	 * Both sum and sum2 are partial sums, so combine them together.
2147 	 */
2148 	sum += ~sum2 & 0xffff;
2149 	while (sum > 0xffff)
2150 		sum = (sum & 0xffff) + (sum >> 16);
2151 	sum2 = ~sum & 0xffff;
2152 	return sum2;
2153 }
2154 
2155 #if (__NetBSD_Version__ >= 799003000)
2156 
2157 /* NetBSD module interface */
2158 
2159 MODULE(MODULE_CLASS_DRIVER, ipl, "bpf_filter");
2160 
2161 static int ipl_init(void *);
2162 static int ipl_fini(void *);
2163 static int ipl_modcmd(modcmd_t, void *);
2164 
2165 #ifdef _MODULE
2166 static devmajor_t ipl_cmaj = -1, ipl_bmaj = -1;
2167 #endif
2168 
2169 static int
ipl_modcmd(modcmd_t cmd,void * opaque)2170 ipl_modcmd(modcmd_t cmd, void *opaque)
2171 {
2172 
2173 	switch (cmd) {
2174 	case MODULE_CMD_INIT:
2175 		return ipl_init(opaque);
2176 	case MODULE_CMD_FINI:
2177 		return ipl_fini(opaque);
2178 	default:
2179 		return ENOTTY;
2180 	}
2181 }
2182 
2183 static int
ipl_init(void * opaque)2184 ipl_init(void *opaque)
2185 {
2186 	int error;
2187 
2188 	ipf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
2189 	    ipf_listener_cb, NULL);
2190 
2191 	if ((error = ipf_load_all()) != 0)
2192 		return error;
2193 
2194 	if (ipf_create_all(&ipfmain) == NULL) {
2195 		ipf_unload_all();
2196 		return ENODEV;
2197 	}
2198 
2199 	/* Initialize our mutex and reference count */
2200 	mutex_init(&ipf_ref_mutex, MUTEX_DEFAULT, IPL_NONE);
2201 	ipf_active = 0;
2202 
2203 #ifdef _MODULE
2204 	/*
2205 	 * Insert ourself into the cdevsw list.
2206 	 */
2207 	error = devsw_attach("ipl", NULL, &ipl_bmaj, &ipl_cdevsw, &ipl_cmaj);
2208 	if (error)
2209 		ipl_fini(opaque);
2210 #endif
2211 
2212 	return error;
2213 }
2214 
2215 static int
ipl_fini(void * opaque)2216 ipl_fini(void *opaque)
2217 {
2218 
2219 #ifdef _MODULE
2220 	(void)devsw_detach(NULL, &ipl_cdevsw);
2221 #endif
2222 
2223 	/*
2224 	 * Grab the mutex, verify that there are no references
2225 	 * and that there are no running filters.  If either
2226 	 * of these exists, reinsert our cdevsw entry and return
2227 	 * an error.
2228 	 */
2229 	mutex_enter(&ipf_ref_mutex);
2230 	if (ipf_active != 0 || ipfmain.ipf_running > 0) {
2231 #ifdef _MODULE
2232 		(void)devsw_attach("ipl", NULL, &ipl_bmaj,
2233 		    &ipl_cdevsw, &ipl_cmaj);
2234 #endif
2235 		mutex_exit(&ipf_ref_mutex);
2236 		return EBUSY;
2237 	}
2238 
2239 	/* Clean up the rest of our state before being unloaded */
2240 
2241 	mutex_exit(&ipf_ref_mutex);
2242 	mutex_destroy(&ipf_ref_mutex);
2243 	ipf_destroy_all(&ipfmain);
2244 	ipf_unload_all();
2245 	kauth_unlisten_scope(ipf_listener);
2246 
2247 	return 0;
2248 }
2249 #endif /* (__NetBSD_Version__ >= 799003000) */
2250