xref: /netbsd-src/sys/external/bsd/ipf/netinet/ip_fil_netbsd.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: ip_fil_netbsd.c,v 1.29 2018/07/11 05:25:46 maxv 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 1.29 2018/07/11 05:25:46 maxv 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 #if defined(__NetBSD__)
61 #include <netinet/in_offload.h>
62 #endif
63 
64 #include <net/if.h>
65 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/tcp.h>
72 #if __NetBSD_Version__ >= 105190000	/* 1.5T */
73 # include <netinet/tcp_timer.h>
74 # include <netinet/tcp_var.h>
75 #endif
76 #include <netinet/udp.h>
77 #include <netinet/ip_icmp.h>
78 #include "netinet/ip_compat.h"
79 #ifdef USE_INET6
80 # include <netinet/icmp6.h>
81 # if (__NetBSD_Version__ >= 106000000)
82 #  include <netinet6/nd6.h>
83 # endif
84 # if __NetBSD_Version__ >= 499001100
85 #  include <netinet6/scope6_var.h>
86 # endif
87 #endif
88 #include "netinet/ip_fil.h"
89 #include "netinet/ip_nat.h"
90 #include "netinet/ip_frag.h"
91 #include "netinet/ip_state.h"
92 #include "netinet/ip_proxy.h"
93 #include "netinet/ip_auth.h"
94 #include "netinet/ip_sync.h"
95 #include "netinet/ip_lookup.h"
96 #include "netinet/ip_dstlist.h"
97 #ifdef	IPFILTER_SCAN
98 #include "netinet/ip_scan.h"
99 #endif
100 #include <sys/md5.h>
101 #include <sys/kernel.h>
102 #include <sys/conf.h>
103 #ifdef INET
104 extern	int	ip_optcopy (struct ip *, struct ip *);
105 #endif
106 
107 #ifdef IPFILTER_M_IPFILTER
108 MALLOC_DEFINE(M_IPFILTER, "IP Filter", "IP Filter packet filter data structures");
109 #endif
110 
111 #if __NetBSD_Version__ >= 105009999
112 # define	csuminfo	csum_flags
113 #endif
114 
115 #if __NetBSD_Version__ < 200000000
116 extern	struct	protosw	inetsw[];
117 #endif
118 
119 #if (__NetBSD_Version__ >= 599002000)
120 static kauth_listener_t ipf_listener;
121 #endif
122 
123 #if (__NetBSD_Version__ < 399001400)
124 extern int ip6_getpmtu (struct route_in6 *, struct route_in6 *,
125 			    struct ifnet *, struct in6_addr *, u_long *,
126 			    int *);
127 #endif
128 #if (NetBSD >= 199511)
129 static  int     ipfopen(dev_t dev, int flags, int devtype, PROC_T *p);
130 static  int     ipfclose(dev_t dev, int flags, int devtype, PROC_T *p);
131 #else
132 # if (__NetBSD_Version__ >= 399001400)
133 static  int     ipfopen(dev_t dev, int flags, struct lwp *);
134 static  int     ipfclose(dev_t dev, int flags, struct lwp *);
135 # else
136 static  int     ipfopen(dev_t dev, int flags);
137 static  int     ipfclose(dev_t dev, int flags);
138 # endif /* __NetBSD_Version__ >= 399001400 */
139 #endif
140 static  int     ipfread(dev_t, struct uio *, int ioflag);
141 static  int     ipfwrite(dev_t, struct uio *, int ioflag);
142 static  int     ipfpoll(dev_t, int events, PROC_T *);
143 static	void	ipf_timer_func(void *ptr);
144 
145 const struct cdevsw ipl_cdevsw = {
146 	.d_open = ipfopen,
147 	.d_close = ipfclose,
148 	.d_read = ipfread,
149 	.d_write = ipfwrite,
150 	.d_ioctl = ipfioctl,
151 	.d_stop = nostop,
152 	.d_tty = notty,
153 	.d_poll = ipfpoll,
154 	.d_mmap = nommap,
155 #if  (__NetBSD_Version__ >= 200000000)
156 	.d_kqfilter = nokqfilter,
157 #endif
158 	.d_discard = nodiscard,
159 #ifdef D_OTHER
160 	.d_flag = D_OTHER
161 #else
162 	.d_flag = 0
163 #endif
164 };
165 #if (__NetBSD_Version__ >= 799003000)
166 kmutex_t ipf_ref_mutex;
167 int	ipf_active;
168 #endif
169 
170 ipf_main_softc_t ipfmain;
171 
172 static	u_short	ipid = 0;
173 static	int	(*ipf_savep)(void *, ip_t *, int, void *, int, struct mbuf **);
174 static	int	ipf_send_ip(fr_info_t *, mb_t *);
175 #ifdef USE_INET6
176 static int ipf_fastroute6(struct mbuf *, struct mbuf **,
177 			      fr_info_t *, frdest_t *);
178 #endif
179 
180 #if defined(NETBSD_PF)
181 # include <net/pfil.h>
182 /*
183  * We provide the ipf_checkp name just to minimize changes later.
184  */
185 int (*ipf_checkp)(void *, ip_t *ip, int hlen, void *ifp, int out, mb_t **mp);
186 #endif /* NETBSD_PF */
187 
188 #if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)
189 # include <net/pfil.h>
190 
191 static int ipf_check_wrapper(void *, struct mbuf **, struct ifnet *, int );
192 
193 static int
194 ipf_check_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
195 {
196 	struct ip *ip;
197 	int rv, hlen;
198 
199 #if __NetBSD_Version__ >= 200080000
200 	/*
201 	 * ensure that mbufs are writable beforehand
202 	 * as it's assumed by ipf code.
203 	 * XXX inefficient
204 	 */
205 	int error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
206 
207 	if (error) {
208 		m_freem(*mp);
209 		*mp = NULL;
210 		return error;
211 	}
212 #endif
213 	ip = mtod(*mp, struct ip *);
214 	hlen = ip->ip_hl << 2;
215 
216 #ifdef INET
217 #if defined(M_CSUM_TCPv4)
218 	/*
219 	 * If the packet is out-bound, we can't delay checksums
220 	 * here.  For in-bound, the checksum has already been
221 	 * validated.
222 	 */
223 	if (dir == PFIL_OUT) {
224 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
225 			in_undefer_cksum_tcpudp(*mp);
226 			(*mp)->m_pkthdr.csum_flags &=
227 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
228 		}
229 	}
230 #endif /* M_CSUM_TCPv4 */
231 #endif /* INET */
232 
233 	/*
234 	 * Note, we don't need to update the checksum, because
235 	 * it has already been verified.
236 	 */
237 	rv = ipf_check(&ipfmain, ip, hlen, ifp, (dir == PFIL_OUT), mp);
238 
239 	return (rv);
240 }
241 
242 # ifdef USE_INET6
243 #  include <netinet/ip6.h>
244 
245 static int ipf_check_wrapper6(void *, struct mbuf **, struct ifnet *, int );
246 
247 static int
248 ipf_check_wrapper6(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
249 {
250 #if defined(INET6)
251 #  if defined(M_CSUM_TCPv6) && (__NetBSD_Version__ > 200000000)
252 	/*
253 	 * If the packet is out-bound, we can't delay checksums
254 	 * here.  For in-bound, the checksum has already been
255 	 * validated.
256 	 */
257 	if (dir == PFIL_OUT) {
258 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
259 #   if (__NetBSD_Version__ > 399000600)
260 			in6_delayed_cksum(*mp);
261 #   endif
262 			(*mp)->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv6|
263 							M_CSUM_UDPv6);
264 		}
265 	}
266 #  endif
267 #endif /* INET6 */
268 
269 	return (ipf_check(&ipfmain, mtod(*mp, struct ip *), sizeof(struct ip6_hdr),
270 	    ifp, (dir == PFIL_OUT), mp));
271 }
272 # endif
273 
274 
275 # if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
276 
277 #  if (__NetBSD_Version__ >= 799000400)
278 
279 static void ipf_pfilsync(void *, unsigned long, void *);
280 
281 static void
282 ipf_pfilsync(void *hdr, unsigned long cmd, void *arg2)
283 {
284 	/*
285 	 * The interface pointer is useless for create (we have nothing to
286 	 * compare it to) and at detach, the interface name is still in the
287 	 * list of active NICs (albeit, down, but that's not any real
288 	 * indicator) and doing ifunit() on the name will still return the
289 	 * pointer, so it's not much use then, either.
290 	 */
291 	ipf_sync(&ipfmain, NULL);
292 }
293 
294 #  else
295 
296 static int ipf_pfilsync(void *, struct mbuf **, struct ifnet *, int);
297 
298 static int
299 ipf_pfilsync(void *hdr, struct mbuf **mp, struct ifnet *ifp, int dir)
300 {
301 	ipf_sync(&ipfmain, NULL);
302 	return 0;
303 }
304 
305 #  endif
306 # endif
307 
308 #endif /* __NetBSD_Version__ >= 105110000 */
309 
310 
311 #if defined(IPFILTER_LKM)
312 int
313 ipf_identify(s)
314 	char *s;
315 {
316 	if (strcmp(s, "ipl") == 0)
317 		return 1;
318 	return 0;
319 }
320 #endif /* IPFILTER_LKM */
321 
322 #if (__NetBSD_Version__ >= 599002000)
323 static int
324 ipf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
325     void *arg0, void *arg1, void *arg2, void *arg3)
326 {
327 	int result;
328 	enum kauth_network_req req;
329 
330 	result = KAUTH_RESULT_DEFER;
331 	req = (enum kauth_network_req)arg0;
332 
333 	if (action != KAUTH_NETWORK_FIREWALL)
334 		return result;
335 
336 	/* These must have came from device context. */
337 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
338 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
339 		result = KAUTH_RESULT_ALLOW;
340 
341 	return result;
342 }
343 #endif
344 
345 /*
346  * Try to detect the case when compiling for NetBSD with pseudo-device
347  */
348 void
349 ipfilterattach(int count)
350 {
351 
352 #if (__NetBSD_Version__ >= 799003000)
353 	return;
354 #else
355 #if (__NetBSD_Version__ >= 599002000)
356 	ipf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
357 	    ipf_listener_cb, NULL);
358 #endif
359 
360 	if (ipf_load_all() == 0)
361 		(void) ipf_create_all(&ipfmain);
362 #endif
363 }
364 
365 
366 int
367 ipfattach(ipf_main_softc_t *softc)
368 {
369 	SPL_INT(s);
370 #if (__NetBSD_Version__ >= 499005500)
371 	int i;
372 #endif
373 #if defined(NETBSD_PF) && (__NetBSD_Version__ >= 104200000)
374 	int error = 0;
375 # if defined(__NetBSD_Version__) && (__NetBSD_Version__ >= 105110000)
376         pfil_head_t *ph_inet;
377 #  ifdef USE_INET6
378         pfil_head_t *ph_inet6;
379 #  endif
380 #  if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
381         pfil_head_t *ph_ifsync;
382 #  endif
383 # endif
384 #endif
385 
386 	SPL_NET(s);
387 	if ((softc->ipf_running > 0) || (ipf_checkp == ipf_check)) {
388 		printf("IP Filter: already initialized\n");
389 		SPL_X(s);
390 		IPFERROR(130017);
391 		return EBUSY;
392 	}
393 
394 	if (ipf_init_all(softc) < 0) {
395 		SPL_X(s);
396 		IPFERROR(130015);
397 		return EIO;
398 	}
399 
400 #ifdef NETBSD_PF
401 # if (__NetBSD_Version__ >= 104200000)
402 #  if __NetBSD_Version__ >= 105110000
403 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
404 #   ifdef USE_INET6
405 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
406 #   endif
407 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
408 	ph_ifsync = pfil_head_get(PFIL_TYPE_IFNET, 0);
409 #   endif
410 
411 	if (ph_inet == NULL
412 #   ifdef USE_INET6
413 	    && ph_inet6 == NULL
414 #   endif
415 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
416 	    && ph_ifsync == NULL
417 #   endif
418 	   ) {
419 		SPL_X(s);
420 		IPFERROR(130016);
421 		return ENODEV;
422 	}
423 
424 	if (ph_inet != NULL)
425 		error = pfil_add_hook((void *)ipf_check_wrapper, NULL,
426 				      PFIL_IN|PFIL_OUT, ph_inet);
427 	else
428 		error = 0;
429 #  else
430 	error = pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
431 			      &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
432 #  endif
433 	if (error) {
434 		IPFERROR(130013);
435 		goto pfil_error;
436 	}
437 # else
438 	pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT);
439 # endif
440 
441 # ifdef USE_INET6
442 #  if __NetBSD_Version__ >= 105110000
443 	if (ph_inet6 != NULL)
444 		error = pfil_add_hook((void *)ipf_check_wrapper6, NULL,
445 				      PFIL_IN|PFIL_OUT, ph_inet6);
446 	else
447 		error = 0;
448 	if (error) {
449 		pfil_remove_hook((void *)ipf_check_wrapper6, NULL,
450 				 PFIL_IN|PFIL_OUT, ph_inet6);
451 		ipfmain.ipf_interror = 130014;
452 		goto pfil_error;
453 	}
454 #  else
455 	error = pfil_add_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
456 			      &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
457 	if (error) {
458 		pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
459 				 &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
460 		IPFERROR(130014);
461 		goto pfil_error;
462 	}
463 #  endif
464 # endif
465 
466 # if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
467 	if (ph_ifsync != NULL)
468 #if (__NetBSD_Version__ >= 799000400)
469 		(void) pfil_add_ihook((void *)ipf_pfilsync, NULL,
470 				      PFIL_IFNET, ph_ifsync);
471 #else
472 		(void) pfil_add_hook((void *)ipf_pfilsync, NULL,
473 				     PFIL_IFNET, ph_ifsync);
474 #endif
475 # endif
476 #endif
477 
478 #if (__NetBSD_Version__ >= 499005500)
479 	for (i = 0; i < IPL_LOGSIZE; i++)
480 		selinit(&ipfmain.ipf_selwait[i]);
481 #else
482 	bzero((char *)ipfmain.ipf_selwait, sizeof(ipfmain.ipf_selwait));
483 #endif
484 	ipf_savep = ipf_checkp;
485 	ipf_checkp = ipf_check;
486 
487 #ifdef INET
488 	if (softc->ipf_control_forwarding & 1)
489 		ipforwarding = 1;
490 #endif
491 
492 	ipid = 0;
493 
494 	SPL_X(s);
495 
496 #if (__NetBSD_Version__ >= 104010000)
497 # if (__NetBSD_Version__ >= 499002000)
498 	callout_init(&softc->ipf_slow_ch, 0);
499 # else
500 	callout_init(&softc->ipf_slow_ch);
501 # endif
502 	callout_reset(&softc->ipf_slow_ch, (hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT,
503 		     ipf_timer_func, softc);
504 #else
505 	timeout(ipf_timer_func, softc, (hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT);
506 #endif
507 
508 	return 0;
509 
510 #if __NetBSD_Version__ >= 105110000
511 pfil_error:
512 	SPL_X(s);
513 	ipf_fini_all(softc);
514 	return error;
515 #endif
516 }
517 
518 static void
519 ipf_timer_func(void *ptr)
520 {
521 	ipf_main_softc_t *softc = ptr;
522 	SPL_INT(s);
523 
524 	SPL_NET(s);
525 	READ_ENTER(&softc->ipf_global);
526 
527 	if (softc->ipf_running > 0)
528 		ipf_slowtimer(softc);
529 
530 	if (softc->ipf_running == -1 || softc->ipf_running == 1) {
531 #if NETBSD_GE_REV(104240000)
532 		callout_reset(&softc->ipf_slow_ch, hz / 2,
533 			      ipf_timer_func, softc);
534 #else
535 		timeout(ipf_timer_func, softc,
536 			(hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT);
537 #endif
538 	}
539 	RWLOCK_EXIT(&softc->ipf_global);
540 	SPL_X(s);
541 }
542 
543 
544 /*
545  * Disable the filter by removing the hooks from the IP input/output
546  * stream.
547  */
548 int
549 ipfdetach(ipf_main_softc_t *softc)
550 {
551 	SPL_INT(s);
552 #if (__NetBSD_Version__ >= 499005500)
553 	int i;
554 #endif
555 #if defined(NETBSD_PF) && (__NetBSD_Version__ >= 104200000)
556 	int error = 0;
557 # if __NetBSD_Version__ >= 105150000
558 	pfil_head_t *ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
559 #  ifdef USE_INET6
560 	pfil_head_t *ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
561 #  endif
562 #  if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
563 	struct pfil_head *ph_ifsync = pfil_head_get(PFIL_TYPE_IFNET, 0);
564 #  endif
565 # endif
566 #endif
567 
568 	SPL_NET(s);
569 
570 #if (__NetBSD_Version__ >= 104010000)
571 	if (softc->ipf_running > 0)
572 		callout_stop(&softc->ipf_slow_ch);
573 #else
574 	untimeout(ipf_slowtimer, NULL);
575 #endif /* NetBSD */
576 
577 	ipf_checkp = ipf_savep;
578 	(void) ipf_flush(softc, IPL_LOGIPF, FR_INQUE|FR_OUTQUE|FR_INACTIVE);
579 	(void) ipf_flush(softc, IPL_LOGIPF, FR_INQUE|FR_OUTQUE);
580 
581 #ifdef INET
582 	if (softc->ipf_control_forwarding & 2)
583 		ipforwarding = 0;
584 #endif
585 
586 #ifdef NETBSD_PF
587 # if (__NetBSD_Version__ >= 104200000)
588 #  if __NetBSD_Version__ >= 105110000
589 #   if defined(PFIL_TYPE_IFNET) && defined(PFIL_IFNET)
590 #    if __NetBSD_Version__ >= 799000400
591 	(void) pfil_remove_ihook((void *)ipf_pfilsync, NULL,
592 				PFIL_IFNET, ph_ifsync);
593 #    else
594 	(void) pfil_remove_hook((void *)ipf_pfilsync, NULL,
595 				PFIL_IFNET, ph_ifsync);
596 #    endif
597 #   endif
598 
599 	if (ph_inet != NULL)
600 		error = pfil_remove_hook((void *)ipf_check_wrapper, NULL,
601 					 PFIL_IN|PFIL_OUT, ph_inet);
602 	else
603 		error = 0;
604 #  else
605 	error = pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
606 				 &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
607 #  endif
608 	if (error) {
609 		SPL_X(s);
610 		IPFERROR(130011);
611 		return error;
612 	}
613 # else
614 	pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT);
615 # endif
616 # ifdef USE_INET6
617 #  if __NetBSD_Version__ >= 105110000
618 	if (ph_inet6 != NULL)
619 		error = pfil_remove_hook((void *)ipf_check_wrapper6, NULL,
620 					 PFIL_IN|PFIL_OUT, ph_inet6);
621 	else
622 		error = 0;
623 #  else
624 	error = pfil_remove_hook((void *)ipf_check, PFIL_IN|PFIL_OUT,
625 				 &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
626 #  endif
627 	if (error) {
628 		SPL_X(s);
629 		IPFERROR(130012);
630 		return error;
631 	}
632 # endif
633 #endif
634 	SPL_X(s);
635 
636 #if (__NetBSD_Version__ >= 499005500)
637 	for (i = 0; i < IPL_LOGSIZE; i++)
638 		seldestroy(&ipfmain.ipf_selwait[i]);
639 #endif
640 
641 	ipf_fini_all(softc);
642 
643 	return 0;
644 }
645 
646 
647 /*
648  * Filter ioctl interface.
649  */
650 int
651 ipfioctl(dev_t dev, u_long cmd,
652 #if  (__NetBSD_Version__ >= 499001000)
653 	void *data,
654 #else
655 	caddr_t data,
656 #endif
657 	int mode
658 #if (NetBSD >= 199511)
659 # if  (__NetBSD_Version__ >= 399001400)
660 	, struct lwp *p
661 #   if (__NetBSD_Version__ >= 399002000)
662 #     define	UID(l)	kauth_cred_getuid((l)->l_cred)
663 #  else
664 #     define	UID(l)	((l)->l_proc->p_cred->p_ruid)
665 #  endif
666 # else
667 	, struct proc *p
668 #  define	UID(p)	((p)->p_cred->p_ruid)
669 # endif
670 #endif
671 )
672 {
673 	int error = 0, unit = 0;
674 	SPL_INT(s);
675 
676 #if (__NetBSD_Version__ >= 399002000)
677 	if ((mode & FWRITE) &&
678 	    kauth_authorize_network(p->l_cred, KAUTH_NETWORK_FIREWALL,
679 				    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL,
680 				    NULL, NULL)) {
681 		ipfmain.ipf_interror = 130005;
682 		return EPERM;
683 	}
684 #else
685 	if ((securelevel >= 2) && (mode & FWRITE)) {
686 		ipfmain.ipf_interror = 130001;
687 		return EPERM;
688 	}
689 #endif
690 
691   	unit = GET_MINOR(dev);
692   	if ((IPL_LOGMAX < unit) || (unit < 0)) {
693 		ipfmain.ipf_interror = 130002;
694 		return ENXIO;
695 	}
696 
697 	if (ipfmain.ipf_running <= 0) {
698 		if (unit != IPL_LOGIPF && cmd != SIOCIPFINTERROR) {
699 			ipfmain.ipf_interror = 130003;
700 			return EIO;
701 		}
702 		if (cmd != SIOCIPFGETNEXT && cmd != SIOCIPFGET &&
703 		    cmd != SIOCIPFSET && cmd != SIOCFRENB &&
704 		    cmd != SIOCGETFS && cmd != SIOCGETFF &&
705 		    cmd != SIOCIPFINTERROR) {
706 			ipfmain.ipf_interror = 130004;
707 			return EIO;
708 		}
709 	}
710 
711 	SPL_NET(s);
712 
713 	error = ipf_ioctlswitch(&ipfmain, unit, data, cmd, mode, UID(p), p);
714 	if (error != -1) {
715 		SPL_X(s);
716 		return error;
717 	}
718 
719 	SPL_X(s);
720 	return error;
721 }
722 
723 
724 /*
725  * ipf_send_reset - this could conceivably be a call to tcp_respond(), but that
726  * requires a large amount of setting up and isn't any more efficient.
727  */
728 int
729 ipf_send_reset(fr_info_t *fin)
730 {
731 	struct tcphdr *tcp, *tcp2;
732 	int tlen = 0, hlen;
733 	struct mbuf *m;
734 #ifdef USE_INET6
735 	ip6_t *ip6;
736 #endif
737 	ip_t *ip;
738 
739 	tcp = fin->fin_dp;
740 	if (tcp->th_flags & TH_RST)
741 		return -1;		/* feedback loop */
742 
743 	if (ipf_checkl4sum(fin) == -1)
744 		return -1;
745 
746 	tlen = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
747 			((tcp->th_flags & TH_SYN) ? 1 : 0) +
748 			((tcp->th_flags & TH_FIN) ? 1 : 0);
749 
750 #ifdef USE_INET6
751 	hlen = (fin->fin_v == 6) ? sizeof(ip6_t) : sizeof(ip_t);
752 #else
753 	hlen = sizeof(ip_t);
754 #endif
755 #ifdef MGETHDR
756 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
757 #else
758 	MGET(m, M_DONTWAIT, MT_HEADER);
759 #endif
760 	if (m == NULL)
761 		return -1;
762 	if (sizeof(*tcp2) + hlen > MHLEN) {
763 		MCLGET(m, M_DONTWAIT);
764 		if (m == NULL)
765 			return -1;
766 		if ((m->m_flags & M_EXT) == 0) {
767 			FREE_MB_T(m);
768 			return -1;
769 		}
770 	}
771 
772 	m->m_len = sizeof(*tcp2) + hlen;
773 	m->m_data += max_linkhdr;
774 	m->m_pkthdr.len = m->m_len;
775 	m_reset_rcvif(m);
776 	ip = mtod(m, struct ip *);
777 	bzero((char *)ip, hlen);
778 #ifdef USE_INET6
779 	ip6 = (ip6_t *)ip;
780 #endif
781 	bzero((char *)ip, sizeof(*tcp2) + hlen);
782 	tcp2 = (struct tcphdr *)((char *)ip + hlen);
783 	tcp2->th_sport = tcp->th_dport;
784 	tcp2->th_dport = tcp->th_sport;
785 
786 	if (tcp->th_flags & TH_ACK) {
787 		tcp2->th_seq = tcp->th_ack;
788 		tcp2->th_flags = TH_RST;
789 		tcp2->th_ack = 0;
790 	} else {
791 		tcp2->th_seq = 0;
792 		tcp2->th_ack = ntohl(tcp->th_seq);
793 		tcp2->th_ack += tlen;
794 		tcp2->th_ack = htonl(tcp2->th_ack);
795 		tcp2->th_flags = TH_RST|TH_ACK;
796 	}
797 	tcp2->th_x2 = 0;
798 	TCP_OFF_A(tcp2, sizeof(*tcp2) >> 2);
799 	tcp2->th_win = tcp->th_win;
800 	tcp2->th_sum = 0;
801 	tcp2->th_urp = 0;
802 
803 #ifdef USE_INET6
804 	if (fin->fin_v == 6) {
805 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
806 		ip6->ip6_plen = htons(sizeof(struct tcphdr));
807 		ip6->ip6_nxt = IPPROTO_TCP;
808 		ip6->ip6_hlim = 0;
809 		ip6->ip6_src = fin->fin_dst6.in6;
810 		ip6->ip6_dst = fin->fin_src6.in6;
811 		tcp2->th_sum = in6_cksum(m, IPPROTO_TCP,
812 					 sizeof(*ip6), sizeof(*tcp2));
813 		return ipf_send_ip(fin, m);
814 	}
815 #endif
816 #ifdef INET
817 	ip->ip_p = IPPROTO_TCP;
818 	ip->ip_len = htons(sizeof(struct tcphdr));
819 	ip->ip_src.s_addr = fin->fin_daddr;
820 	ip->ip_dst.s_addr = fin->fin_saddr;
821 	tcp2->th_sum = in_cksum(m, hlen + sizeof(*tcp2));
822 	ip->ip_len = hlen + sizeof(*tcp2);
823 	return ipf_send_ip(fin, m);
824 #else
825 	return 0;
826 #endif
827 }
828 
829 
830 /*
831  * Expects ip_len to be in host byte order when called.
832  */
833 static int
834 ipf_send_ip(fr_info_t *fin, mb_t *m)
835 {
836 	fr_info_t fnew;
837 #ifdef INET
838 	ip_t *oip;
839 #endif
840 	ip_t *ip;
841 	int hlen;
842 
843 	ip = mtod(m, ip_t *);
844 	bzero((char *)&fnew, sizeof(fnew));
845 	fnew.fin_main_soft = fin->fin_main_soft;
846 
847 	IP_V_A(ip, fin->fin_v);
848 	switch (fin->fin_v)
849 	{
850 #ifdef INET
851 	case 4 :
852 		oip = fin->fin_ip;
853 		hlen = sizeof(*oip);
854 		fnew.fin_v = 4;
855 		fnew.fin_p = ip->ip_p;
856 		fnew.fin_plen = ntohs(ip->ip_len);
857 		HTONS(ip->ip_len);
858 		IP_HL_A(ip, sizeof(*oip) >> 2);
859 		ip->ip_tos = oip->ip_tos;
860 		ip->ip_id = ipf_nextipid(fin);
861 		ip->ip_off = htons(ip_mtudisc ? IP_DF : 0);
862 		ip->ip_ttl = ip_defttl;
863 		ip->ip_sum = 0;
864 		break;
865 #endif
866 #ifdef USE_INET6
867 	case 6 :
868 	{
869 		ip6_t *ip6 = (ip6_t *)ip;
870 
871 		ip6->ip6_vfc = 0x60;
872 		ip6->ip6_hlim = IPDEFTTL;
873 
874 		hlen = sizeof(*ip6);
875 		fnew.fin_p = ip6->ip6_nxt;
876 		fnew.fin_v = 6;
877 		fnew.fin_plen = ntohs(ip6->ip6_plen) + hlen;
878 		break;
879 	}
880 #endif
881 	default :
882 		return EINVAL;
883 	}
884 #ifdef KAME_IPSEC
885 	m_reset_rcvif(m);
886 #endif
887 
888 	fnew.fin_ifp = fin->fin_ifp;
889 	fnew.fin_flx = FI_NOCKSUM;
890 	fnew.fin_m = m;
891 	fnew.fin_ip = ip;
892 	fnew.fin_mp = &m;
893 	fnew.fin_hlen = hlen;
894 	fnew.fin_dp = (char *)ip + hlen;
895 	(void) ipf_makefrip(hlen, ip, &fnew);
896 
897 	return ipf_fastroute(m, &m, &fnew, NULL);
898 }
899 
900 
901 int
902 ipf_send_icmp_err(int type, fr_info_t *fin, int dst)
903 {
904 	int err, hlen, xtra, iclen, ohlen, avail;
905 	struct in_addr dst4;
906 	struct icmp *icmp;
907 	struct mbuf *m;
908 	i6addr_t dst6;
909 	void *ifp;
910 #ifdef USE_INET6
911 	int code;
912 	ip6_t *ip6;
913 #endif
914 	ip_t *ip, *ip2;
915 
916 	if ((type < 0) || (type > ICMP_MAXTYPE))
917 		return -1;
918 
919 #ifdef USE_INET6
920 	code = fin->fin_icode;
921 	if ((code < 0) || (code >= sizeof(icmptoicmp6unreach)/sizeof(int)))
922 		return -1;
923 #endif
924 
925 	if (ipf_checkl4sum(fin) == -1)
926 		return -1;
927 #ifdef MGETHDR
928 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
929 #else
930 	MGET(m, M_DONTWAIT, MT_HEADER);
931 #endif
932 	if (m == NULL)
933 		return -1;
934 	avail = MHLEN;
935 
936 	xtra = 0;
937 	hlen = 0;
938 	ohlen = 0;
939 	dst4.s_addr = 0;
940 	ifp = fin->fin_ifp;
941 	if (fin->fin_v == 4) {
942 		if ((fin->fin_p == IPPROTO_ICMP) && !(fin->fin_flx & FI_SHORT))
943 			switch (ntohs(fin->fin_data[0]) >> 8)
944 			{
945 			case ICMP_ECHO :
946 			case ICMP_TSTAMP :
947 			case ICMP_IREQ :
948 			case ICMP_MASKREQ :
949 				break;
950 			default :
951 				FREE_MB_T(m);
952 				return 0;
953 			}
954 
955 		if (dst == 0) {
956 			if (ipf_ifpaddr(&ipfmain, 4, FRI_NORMAL, ifp,
957 				       &dst6, NULL) == -1) {
958 				FREE_MB_T(m);
959 				return -1;
960 			}
961 			dst4 = dst6.in4;
962 		} else
963 			dst4.s_addr = fin->fin_daddr;
964 
965 		hlen = sizeof(ip_t);
966 		ohlen = fin->fin_hlen;
967 		iclen = hlen + offsetof(struct icmp, icmp_ip) + ohlen;
968 		if (fin->fin_hlen < fin->fin_plen)
969 			xtra = MIN(fin->fin_dlen, 8);
970 		else
971 			xtra = 0;
972 	}
973 
974 #ifdef USE_INET6
975 	else if (fin->fin_v == 6) {
976 		hlen = sizeof(ip6_t);
977 		ohlen = sizeof(ip6_t);
978 		iclen = hlen + offsetof(struct icmp, icmp_ip) + ohlen;
979 		type = icmptoicmp6types[type];
980 		if (type == ICMP6_DST_UNREACH)
981 			code = icmptoicmp6unreach[code];
982 
983 		if (iclen + max_linkhdr + fin->fin_plen > avail) {
984 			MCLGET(m, M_DONTWAIT);
985 			if (m == NULL)
986 				return -1;
987 			if ((m->m_flags & M_EXT) == 0) {
988 				FREE_MB_T(m);
989 				return -1;
990 			}
991 			avail = MCLBYTES;
992 		}
993 		xtra = MIN(fin->fin_plen, avail - iclen - max_linkhdr);
994 		xtra = MIN(xtra, IPV6_MMTU - iclen);
995 		if (dst == 0 && !IN6_IS_ADDR_LINKLOCAL(&fin->fin_dst6.in6)) {
996 			if (ipf_ifpaddr(&ipfmain, 6, FRI_NORMAL, ifp,
997 				       &dst6, NULL) == -1) {
998 				FREE_MB_T(m);
999 				return -1;
1000 			}
1001 		} else
1002 			dst6 = fin->fin_dst6;
1003 	}
1004 #endif
1005 	else {
1006 		FREE_MB_T(m);
1007 		return -1;
1008 	}
1009 
1010 	avail -= (max_linkhdr + iclen);
1011 	if (avail < 0) {
1012 		FREE_MB_T(m);
1013 		return -1;
1014 	}
1015 	if (xtra > avail)
1016 		xtra = avail;
1017 	iclen += xtra;
1018 	m->m_data += max_linkhdr;
1019 	m_reset_rcvif(m);
1020 	m->m_pkthdr.len = iclen;
1021 	m->m_len = iclen;
1022 	ip = mtod(m, ip_t *);
1023 	icmp = (struct icmp *)((char *)ip + hlen);
1024 	ip2 = (ip_t *)&icmp->icmp_ip;
1025 
1026 	icmp->icmp_type = type;
1027 	icmp->icmp_code = fin->fin_icode;
1028 	icmp->icmp_cksum = 0;
1029 #ifdef icmp_nextmtu
1030 	if (type == ICMP_UNREACH && fin->fin_icode == ICMP_UNREACH_NEEDFRAG) {
1031 		if (fin->fin_mtu != 0) {
1032 			icmp->icmp_nextmtu = htons(fin->fin_mtu);
1033 
1034 		} else if (ifp != NULL) {
1035 			icmp->icmp_nextmtu = htons(GETIFMTU_4(ifp));
1036 
1037 		} else {	/* make up a number... */
1038 			icmp->icmp_nextmtu = htons(fin->fin_plen - 20);
1039 		}
1040 	}
1041 #endif
1042 
1043 	bcopy((char *)fin->fin_ip, (char *)ip2, ohlen);
1044 
1045 #if defined(M_CSUM_IPv4)
1046 	/*
1047 	 * Clear any in-bound checksum flags for this packet.
1048 	 */
1049 	m->m_pkthdr.csuminfo = 0;
1050 #endif /* __NetBSD__ && M_CSUM_IPv4 */
1051 
1052 #ifdef USE_INET6
1053 	ip6 = (ip6_t *)ip;
1054 	if (fin->fin_v == 6) {
1055 		ip6->ip6_flow = ((ip6_t *)fin->fin_ip)->ip6_flow;
1056 		ip6->ip6_plen = htons(iclen - hlen);
1057 		ip6->ip6_nxt = IPPROTO_ICMPV6;
1058 		ip6->ip6_hlim = 0;
1059 		ip6->ip6_src = dst6.in6;
1060 		ip6->ip6_dst = fin->fin_src6.in6;
1061 		if (xtra > 0)
1062 			bcopy((char *)fin->fin_ip + ohlen,
1063 			      (char *)&icmp->icmp_ip + ohlen, xtra);
1064 		icmp->icmp_cksum = in6_cksum(m, IPPROTO_ICMPV6,
1065 					     sizeof(*ip6), iclen - hlen);
1066 	} else
1067 #endif
1068 	{
1069 		ip->ip_p = IPPROTO_ICMP;
1070 		ip->ip_src.s_addr = dst4.s_addr;
1071 		ip->ip_dst.s_addr = fin->fin_saddr;
1072 
1073 		if (xtra > 0)
1074 			bcopy((char *)fin->fin_ip + ohlen,
1075 			      (char *)&icmp->icmp_ip + ohlen, xtra);
1076 		icmp->icmp_cksum = ipf_cksum((u_short *)icmp,
1077 					     sizeof(*icmp) + 8);
1078 		ip->ip_len = iclen;
1079 		ip->ip_p = IPPROTO_ICMP;
1080 	}
1081 	err = ipf_send_ip(fin, m);
1082 	return err;
1083 }
1084 
1085 
1086 /*
1087  * m0 - pointer to mbuf where the IP packet starts
1088  * mpp - pointer to the mbuf pointer that is the start of the mbuf chain
1089  */
1090 int
1091 ipf_fastroute(mb_t *m0, mb_t **mpp, fr_info_t *fin, frdest_t *fdp)
1092 {
1093 	register struct ip *ip, *mhip;
1094 	register struct mbuf *m = *mpp;
1095 	register struct route *ro;
1096 	int len, off, error = 0, hlen, code;
1097 	struct ifnet *ifp, *sifp;
1098 	ipf_main_softc_t *softc;
1099 #if __NetBSD_Version__ >= 499001100
1100 	union {
1101 		struct sockaddr         dst;
1102 		struct sockaddr_in      dst4;
1103 	} u;
1104 #else
1105 	struct sockaddr_in *dst4;
1106 #endif
1107 	struct sockaddr *dst;
1108 	u_short ip_off, ip_len;
1109 	struct route iproute;
1110 	struct rtentry *rt;
1111 	frdest_t node;
1112 	frentry_t *fr;
1113 
1114 	if (fin->fin_v == 6) {
1115 #ifdef USE_INET6
1116 		error = ipf_fastroute6(m0, mpp, fin, fdp);
1117 #else
1118 		error = EPROTONOSUPPORT;
1119 #endif
1120 		if ((error != 0) && (*mpp != NULL))
1121 			FREE_MB_T(*mpp);
1122 		return error;
1123 	}
1124 #ifndef INET
1125 	FREE_MB_T(*mpp);
1126 	return EPROTONOSUPPORT;
1127 #else
1128 
1129 	hlen = fin->fin_hlen;
1130 	ip = mtod(m0, struct ip *);
1131 	softc = fin->fin_main_soft;
1132 	rt = NULL;
1133 	ifp = NULL;
1134 
1135 # if defined(M_CSUM_IPv4)
1136 	/*
1137 	 * Clear any in-bound checksum flags for this packet.
1138 	 */
1139 	m0->m_pkthdr.csuminfo = 0;
1140 # endif /* __NetBSD__ && M_CSUM_IPv4 */
1141 
1142 	/*
1143 	 * Route packet.
1144 	 */
1145 	ro = &iproute;
1146 	memset(ro, 0, sizeof(*ro));
1147 	fr = fin->fin_fr;
1148 
1149 	if ((fr != NULL) && !(fr->fr_flags & FR_KEEPSTATE) && (fdp != NULL) &&
1150 	    (fdp->fd_type == FRD_DSTLIST)) {
1151 		if (ipf_dstlist_select_node(fin, fdp->fd_ptr, NULL, &node) == 0)
1152 			fdp = &node;
1153 	}
1154 	if (fdp != NULL)
1155 		ifp = fdp->fd_ptr;
1156 	else
1157 		ifp = fin->fin_ifp;
1158 
1159 	if ((ifp == NULL) && ((fr == NULL) || !(fr->fr_flags & FR_FASTROUTE))) {
1160 		error = -2;
1161 		goto bad;
1162 	}
1163 
1164 # if __NetBSD_Version__ >= 499001100
1165 	if ((fdp != NULL) && (fdp->fd_ip.s_addr != 0))
1166 		sockaddr_in_init(&u.dst4, &fdp->fd_ip, 0);
1167 	else
1168 		sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1169 	dst = &u.dst;
1170 	rtcache_setdst(ro, dst);
1171 	rt = rtcache_init(ro);
1172 # else
1173 	dst4 = (struct sockaddr_in *)&ro->ro_dst;
1174 	dst = (struct sockaddr *)dst4;
1175 	dst4->sin_family = AF_INET;
1176 	dst4->sin_addr = ip->ip_dst;
1177 
1178 	if ((fdp != NULL) && (fdp->fd_ip.s_addr != 0))
1179 		dst4->sin_addr = fdp->fd_ip;
1180 
1181 	dst4->sin_len = sizeof(*dst);
1182 	rtalloc(ro);
1183 	rt = ro->ro_rt;
1184 # endif
1185 	if ((ifp == NULL) && (rt != NULL))
1186 		ifp = rt->rt_ifp;
1187 	if ((rt == NULL) || (ifp == NULL)) {
1188 #ifdef INET
1189 		if (in_localaddr(ip->ip_dst))
1190 			error = EHOSTUNREACH;
1191 		else
1192 #endif
1193 			error = ENETUNREACH;
1194 		goto bad;
1195 	}
1196 
1197 
1198 	if (rt->rt_flags & RTF_GATEWAY)
1199 		dst = rt->rt_gateway;
1200 
1201 	rt->rt_use++;
1202 
1203 	/*
1204 	 * For input packets which are being "fastrouted", they won't
1205 	 * go back through output filtering and miss their chance to get
1206 	 * NAT'd and counted.  Duplicated packets aren't considered to be
1207 	 * part of the normal packet stream, so do not NAT them or pass
1208 	 * them through stateful checking, etc.
1209 	 */
1210 	if ((fdp != &fr->fr_dif) && (fin->fin_out == 0)) {
1211 		sifp = fin->fin_ifp;
1212 		fin->fin_ifp = ifp;
1213 		fin->fin_out = 1;
1214 		(void) ipf_acctpkt(fin, NULL);
1215 		fin->fin_fr = NULL;
1216 		if (!fr || !(fr->fr_flags & FR_RETMASK)) {
1217 			u_32_t pass;
1218 
1219 			(void) ipf_state_check(fin, &pass);
1220 		}
1221 
1222 		switch (ipf_nat_checkout(fin, NULL))
1223 		{
1224 		case 0 :
1225 			break;
1226 		case 1 :
1227 			ip->ip_sum = 0;
1228 			break;
1229 		case -1 :
1230 			error = -1;
1231 			goto bad;
1232 			break;
1233 		}
1234 
1235 		fin->fin_ifp = sifp;
1236 		fin->fin_out = 0;
1237 	} else
1238 		ip->ip_sum = 0;
1239 	/*
1240 	 * If small enough for interface, can just send directly.
1241 	 */
1242 	m_set_rcvif(m, ifp);
1243 
1244 	ip_len = ntohs(ip->ip_len);
1245 	if (ip_len <= ifp->if_mtu) {
1246 # if defined(M_CSUM_IPv4)
1247 #  if (__NetBSD_Version__ >= 105009999)
1248 		if (ifp->if_csum_flags_tx & M_CSUM_IPv4)
1249 			m->m_pkthdr.csuminfo |= M_CSUM_IPv4;
1250 #  else
1251 		if (ifp->if_capabilities & IFCAP_CSUM_IPv4)
1252 			m->m_pkthdr.csuminfo |= M_CSUM_IPv4;
1253 #  endif /* (__NetBSD_Version__ >= 105009999) */
1254 		else if (ip->ip_sum == 0)
1255 			ip->ip_sum = in_cksum(m, hlen);
1256 # else
1257 		if (!ip->ip_sum)
1258 			ip->ip_sum = in_cksum(m, hlen);
1259 # endif /* M_CSUM_IPv4 */
1260 
1261 		error = if_output_lock(ifp, ifp, m, dst, rt);
1262 		goto done;
1263 	}
1264 
1265 	/*
1266 	 * Too large for interface; fragment if possible.
1267 	 * Must be able to put at least 8 bytes per fragment.
1268 	 */
1269 	ip_off = ntohs(ip->ip_off);
1270 	if (ip_off & IP_DF) {
1271 		error = EMSGSIZE;
1272 		goto bad;
1273 	}
1274 	len = (ifp->if_mtu - hlen) &~ 7;
1275 	if (len < 8) {
1276 		error = EMSGSIZE;
1277 		goto bad;
1278 	}
1279 
1280     {
1281 	int mhlen, firstlen = len;
1282 	struct mbuf **mnext = &m->m_act;
1283 
1284 	/*
1285 	 * Loop through length of segment after first fragment,
1286 	 * make new header and copy data of each part and link onto chain.
1287 	 */
1288 	m0 = m;
1289 	mhlen = sizeof (struct ip);
1290 	for (off = hlen + len; off < ip_len; off += len) {
1291 # ifdef MGETHDR
1292 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
1293 # else
1294 		MGET(m, M_DONTWAIT, MT_HEADER);
1295 # endif
1296 		if (m == 0) {
1297 			m = m0;
1298 			error = ENOBUFS;
1299 			goto bad;
1300 		}
1301 		m->m_data += max_linkhdr;
1302 		mhip = mtod(m, struct ip *);
1303 		bcopy((char *)ip, (char *)mhip, sizeof(*ip));
1304 #ifdef INET
1305 		if (hlen > sizeof (struct ip)) {
1306 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1307 			IP_HL_A(mhip, mhlen >> 2);
1308 		}
1309 #endif
1310 		m->m_len = mhlen;
1311 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1312 		if (off + len >= ip_len)
1313 			len = ip_len - off;
1314 		else
1315 			mhip->ip_off |= IP_MF;
1316 		mhip->ip_len = htons((u_short)(len + mhlen));
1317 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
1318 		if (m->m_next == 0) {
1319 			error = ENOBUFS;	/* ??? */
1320 			goto sendorfree;
1321 		}
1322 		m->m_pkthdr.len = mhlen + len;
1323 		m_reset_rcvif(m);
1324 		mhip->ip_off = htons((u_short)mhip->ip_off);
1325 		mhip->ip_sum = 0;
1326 #ifdef INET
1327 		mhip->ip_sum = in_cksum(m, mhlen);
1328 #endif
1329 		*mnext = m;
1330 		mnext = &m->m_act;
1331 	}
1332 	/*
1333 	 * Update first fragment by trimming what's been copied out
1334 	 * and updating header, then send each fragment (in order).
1335 	 */
1336 	m_adj(m0, hlen + firstlen - ip_len);
1337 	ip->ip_len = htons((u_short)(hlen + firstlen));
1338 	ip->ip_off = htons((u_short)IP_MF);
1339 	ip->ip_sum = 0;
1340 #ifdef INET
1341 	ip->ip_sum = in_cksum(m0, hlen);
1342 #endif
1343 sendorfree:
1344 	for (m = m0; m; m = m0) {
1345 		m0 = m->m_act;
1346 		m->m_act = 0;
1347 		if (error == 0) {
1348 			KERNEL_LOCK(1, NULL);
1349 			error = (*ifp->if_output)(ifp, m, dst, rt);
1350 			KERNEL_UNLOCK_ONE(NULL);
1351 		} else {
1352 			FREE_MB_T(m);
1353 		}
1354 	}
1355     }
1356 done:
1357 	if (!error)
1358 		softc->ipf_frouteok[0]++;
1359 	else
1360 		softc->ipf_frouteok[1]++;
1361 
1362 # if __NetBSD_Version__ >= 499001100
1363 	rtcache_unref(rt, ro);
1364 	rtcache_free(ro);
1365 # else
1366 	if (rt) {
1367 		RTFREE(rt);
1368 	}
1369 # endif
1370 	return error;
1371 bad:
1372 	if (error == EMSGSIZE) {
1373 		sifp = fin->fin_ifp;
1374 		code = fin->fin_icode;
1375 		fin->fin_icode = ICMP_UNREACH_NEEDFRAG;
1376 		fin->fin_ifp = ifp;
1377 		(void) ipf_send_icmp_err(ICMP_UNREACH, fin, 1);
1378 		fin->fin_ifp = sifp;
1379 		fin->fin_icode = code;
1380 	}
1381 	FREE_MB_T(m);
1382 	goto done;
1383 #endif /* INET */
1384 }
1385 
1386 
1387 #if defined(USE_INET6)
1388 /*
1389  * This is the IPv6 specific fastroute code.  It doesn't clean up the mbuf's
1390  * or ensure that it is an IPv6 packet that is being forwarded, those are
1391  * expected to be done by the called (ipf_fastroute).
1392  */
1393 static int
1394 ipf_fastroute6(struct mbuf *m0, struct mbuf **mpp, fr_info_t *fin,
1395     frdest_t *fdp)
1396 {
1397 # if __NetBSD_Version__ >= 499001100
1398 	struct route ip6route;
1399 	const struct sockaddr *dst;
1400 	union {
1401 		struct sockaddr         dst;
1402 		struct sockaddr_in6     dst6;
1403 	} u;
1404 	struct route *ro;
1405 # else
1406 	struct route_in6 ip6route;
1407 	struct sockaddr_in6 *dst6;
1408 	struct route_in6 *ro;
1409 # endif
1410 	struct rtentry *rt;
1411 	struct ifnet *ifp;
1412 	u_long mtu;
1413 	int error;
1414 
1415 	error = 0;
1416 	ro = &ip6route;
1417 
1418 	if (fdp != NULL)
1419 		ifp = fdp->fd_ptr;
1420 	else
1421 		ifp = fin->fin_ifp;
1422 	memset(ro, 0, sizeof(*ro));
1423 # if __NetBSD_Version__ >= 499001100
1424 	if (fdp != NULL && IP6_NOTZERO(&fdp->fd_ip6))
1425 		sockaddr_in6_init(&u.dst6, &fdp->fd_ip6.in6, 0, 0, 0);
1426 	else
1427 		sockaddr_in6_init(&u.dst6, &fin->fin_fi.fi_dst.in6, 0, 0, 0);
1428 	if ((error = in6_setscope(&u.dst6.sin6_addr, ifp,
1429 	    &u.dst6.sin6_scope_id)) != 0)
1430 		return error;
1431 	if ((error = sa6_embedscope(&u.dst6, 0)) != 0)
1432 		return error;
1433 
1434 	dst = &u.dst;
1435 	rtcache_setdst(ro, dst);
1436 
1437 	rt = rtcache_init(ro);
1438 	if ((ifp == NULL) && (rt != NULL))
1439 		ifp = rt->rt_ifp;
1440 # else
1441 	dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
1442 	dst6->sin6_family = AF_INET6;
1443 	dst6->sin6_len = sizeof(struct sockaddr_in6);
1444 	dst6->sin6_addr = fin->fin_fi.fi_dst.in6;
1445 	/* KAME */
1446 	if (IN6_IS_ADDR_LINKLOCAL(&dst6->sin6_addr))
1447 		dst6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1448 
1449 	if (fdp != NULL) {
1450 		if (IP6_NOTZERO(&fdp->fd_ip6))
1451 			dst6->sin6_addr = fdp->fd_ip6.in6;
1452 	}
1453 
1454 	rtalloc((struct route *)ro);
1455 
1456 	if ((ifp == NULL) && (ro->ro_rt != NULL))
1457 		ifp = ro->ro_rt->rt_ifp;
1458 	rt = ro->ro_rt;
1459 # endif
1460 	if ((rt == NULL) || (ifp == NULL)) {
1461 
1462 		error = EHOSTUNREACH;
1463 		goto bad;
1464 	}
1465 
1466 	{
1467 # if (__NetBSD_Version__ >= 106010000) && !defined(IN6_LINKMTU)
1468 		struct in6_ifextra *ife;
1469 # endif
1470 		if (rt->rt_flags & RTF_GATEWAY)
1471 # if __NetBSD_Version__ >= 499001100
1472 			dst = rt->rt_gateway;
1473 # else
1474 			dst6 = (struct sockaddr_in6 *)rt->rt_gateway;
1475 # endif
1476 		rt->rt_use++;
1477 
1478 		/* Determine path MTU. */
1479 # if (__NetBSD_Version__ <= 106009999)
1480 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
1481 # else
1482 #  ifdef IN6_LINKMTU
1483 		mtu = IN6_LINKMTU(ifp);
1484 #  else
1485 		ife = (struct in6_ifextra *)(ifp)->if_afdata[AF_INET6];
1486 		mtu = ife->nd_ifinfo[ifp->if_index].linkmtu;
1487 #  endif
1488 # endif
1489 		if ((error == 0) && (m0->m_pkthdr.len <= mtu)) {
1490 # if __NetBSD_Version__ >= 499001100
1491 			error = ip6_if_output(ifp, ifp, m0, satocsin6(dst), rt);
1492 # else
1493 			error = nd6_output(ifp, ifp, m0, dst6, rt);
1494 # endif
1495 		} else {
1496 			error = EMSGSIZE;
1497 		}
1498 	}
1499 bad:
1500 # if __NetBSD_Version__ >= 499001100
1501 	rtcache_unref(rt, ro);
1502 	rtcache_free(ro);
1503 # else
1504 	if (ro->ro_rt != NULL) {
1505 		RTFREE(((struct route *)ro)->ro_rt);
1506 	}
1507 # endif
1508 	return error;
1509 }
1510 #endif	/* INET6 */
1511 
1512 
1513 int
1514 ipf_verifysrc(fr_info_t *fin)
1515 {
1516 #if __NetBSD_Version__ >= 499001100
1517 	union {
1518 		struct sockaddr         dst;
1519 		struct sockaddr_in      dst4;
1520 	} u;
1521 	struct rtentry *rt;
1522 #else
1523 	struct sockaddr_in *dst;
1524 #endif
1525 	struct route iproute;
1526 	int rc;
1527 
1528 #if __NetBSD_Version__ >= 499001100
1529 	sockaddr_in_init(&u.dst4, &fin->fin_src, 0);
1530 	rtcache_setdst(&iproute, &u.dst);
1531 	rt = rtcache_init(&iproute);
1532 	if (rt == NULL)
1533 		rc = 0;
1534 	else
1535 		rc = (fin->fin_ifp == rt->rt_ifp);
1536 	rtcache_unref(rt, &iproute);
1537 	rtcache_free(&iproute);
1538 #else
1539 	dst = (struct sockaddr_in *)&iproute.ro_dst;
1540 	dst->sin_len = sizeof(*dst);
1541 	dst->sin_family = AF_INET;
1542 	dst->sin_addr = fin->fin_src;
1543 	rtalloc(&iproute);
1544 	if (iproute.ro_rt == NULL)
1545 		return 0;
1546 	rc = (fin->fin_ifp == iproute.ro_rt->rt_ifp);
1547 	RTFREE(iproute.ro_rt);
1548 #endif
1549 	return rc;
1550 }
1551 
1552 
1553 /*
1554  * return the first IP Address associated with an interface
1555  */
1556 int
1557 ipf_ifpaddr(ipf_main_softc_t *softc, int v, int atype, void *ifptr,
1558     i6addr_t *inp, i6addr_t *inpmask)
1559 {
1560 #ifdef USE_INET6
1561 	struct in6_addr *inp6 = NULL;
1562 #endif
1563 	struct sockaddr *sock, *mask;
1564 	struct sockaddr_in *sin;
1565 	struct ifaddr *ifa;
1566 	struct ifnet *ifp;
1567 
1568 	if ((ifptr == NULL) || (ifptr == (void *)-1))
1569 		return -1;
1570 
1571 	ifp = ifptr;
1572 	mask = NULL;
1573 
1574 	if (v == 4)
1575 		inp->in4.s_addr = 0;
1576 #ifdef USE_INET6
1577 	else if (v == 6)
1578 		bzero((char *)inp, sizeof(*inp));
1579 #endif
1580 
1581 	ifa = IFADDR_READER_FIRST(ifp);
1582 	sock = ifa ? ifa->ifa_addr : NULL;
1583 	while (sock != NULL && ifa != NULL) {
1584 		sin = (struct sockaddr_in *)sock;
1585 		if ((v == 4) && (sin->sin_family == AF_INET))
1586 			break;
1587 #ifdef USE_INET6
1588 		if ((v == 6) && (sin->sin_family == AF_INET6)) {
1589 			inp6 = &((struct sockaddr_in6 *)sin)->sin6_addr;
1590 			if (!IN6_IS_ADDR_LINKLOCAL(inp6) &&
1591 			    !IN6_IS_ADDR_LOOPBACK(inp6))
1592 				break;
1593 		}
1594 #endif
1595 		ifa = IFADDR_READER_NEXT(ifa);
1596 		if (ifa != NULL)
1597 			sock = ifa->ifa_addr;
1598 	}
1599 	if (ifa == NULL || sock == NULL)
1600 		return -1;
1601 
1602 	mask = ifa->ifa_netmask;
1603 	if (atype == FRI_BROADCAST)
1604 		sock = ifa->ifa_broadaddr;
1605 	else if (atype == FRI_PEERADDR)
1606 		sock = ifa->ifa_dstaddr;
1607 
1608 #ifdef USE_INET6
1609 	if (v == 6)
1610 		return ipf_ifpfillv6addr(atype, (struct sockaddr_in6 *)sock,
1611 					(struct sockaddr_in6 *)mask,
1612 					inp, inpmask);
1613 #endif
1614 	return ipf_ifpfillv4addr(atype, (struct sockaddr_in *)sock,
1615 				(struct sockaddr_in *)mask,
1616 				&inp->in4, &inpmask->in4);
1617 }
1618 
1619 
1620 u_32_t
1621 ipf_newisn(fr_info_t *fin)
1622 {
1623 #if __NetBSD_Version__ >= 105190000	/* 1.5T */
1624 	size_t asz;
1625 
1626 	if (fin->fin_v == 4)
1627 		asz = sizeof(struct in_addr);
1628 	else if (fin->fin_v == 6)
1629 		asz = sizeof(fin->fin_src);
1630 	else	/* XXX: no way to return error */
1631 		return 0;
1632 #ifdef INET
1633 	return tcp_new_iss1((void *)&fin->fin_src, (void *)&fin->fin_dst,
1634 			    fin->fin_sport, fin->fin_dport, asz, 0);
1635 #else
1636 	return ENOSYS;
1637 #endif
1638 #else
1639 	static int iss_seq_off = 0;
1640 	u_char hash[16];
1641 	u_32_t newiss;
1642 	MD5_CTX ctx;
1643 
1644 	/*
1645 	 * Compute the base value of the ISS.  It is a hash
1646 	 * of (saddr, sport, daddr, dport, secret).
1647 	 */
1648 	MD5Init(&ctx);
1649 
1650 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_src,
1651 		  sizeof(fin->fin_fi.fi_src));
1652 	MD5Update(&ctx, (u_char *) &fin->fin_fi.fi_dst,
1653 		  sizeof(fin->fin_fi.fi_dst));
1654 	MD5Update(&ctx, (u_char *) &fin->fin_dat, sizeof(fin->fin_dat));
1655 
1656 	MD5Update(&ctx, ipf_iss_secret, sizeof(ipf_iss_secret));
1657 
1658 	MD5Final(hash, &ctx);
1659 
1660 	memcpy(&newiss, hash, sizeof(newiss));
1661 
1662 	/*
1663 	 * Now increment our "timer", and add it in to
1664 	 * the computed value.
1665 	 *
1666 	 * XXX Use `addin'?
1667 	 * XXX TCP_ISSINCR too large to use?
1668 	 */
1669 	iss_seq_off += 0x00010000;
1670 	newiss += iss_seq_off;
1671 	return newiss;
1672 #endif
1673 }
1674 
1675 
1676 /* ------------------------------------------------------------------------ */
1677 /* Function:    ipf_nextipid                                                 */
1678 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
1679 /* Parameters:  fin(I) - pointer to packet information                      */
1680 /*                                                                          */
1681 /* Returns the next IPv4 ID to use for this packet.                         */
1682 /* ------------------------------------------------------------------------ */
1683 u_short
1684 ipf_nextipid(fr_info_t *fin)
1685 {
1686 #ifdef USE_MUTEXES
1687 	ipf_main_softc_t *softc = fin->fin_main_soft;
1688 #endif
1689 	u_short id;
1690 
1691 	MUTEX_ENTER(&softc->ipf_rw);
1692 	id = ipid++;
1693 	MUTEX_EXIT(&softc->ipf_rw);
1694 
1695 	return id;
1696 }
1697 
1698 
1699 EXTERN_INLINE int
1700 ipf_checkv4sum(fr_info_t *fin)
1701 {
1702 #ifdef M_CSUM_TCP_UDP_BAD
1703 	int manual, pflag, cflags, active;
1704 	mb_t *m;
1705 
1706 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
1707 		return 0;
1708 
1709 	if ((fin->fin_flx & FI_SHORT) != 0)
1710 		return 1;
1711 
1712 	if (fin->fin_cksum != FI_CK_NEEDED)
1713 		return (fin->fin_cksum > FI_CK_NEEDED) ? 0 : -1;
1714 
1715 	manual = 0;
1716 	m = fin->fin_m;
1717 	if (m == NULL) {
1718 		manual = 1;
1719 		goto skipauto;
1720 	}
1721 
1722 	switch (fin->fin_p)
1723 	{
1724 	case IPPROTO_UDP :
1725 		pflag = M_CSUM_UDPv4;
1726 		break;
1727 	case IPPROTO_TCP :
1728 		pflag = M_CSUM_TCPv4;
1729 		break;
1730 	default :
1731 		pflag = 0;
1732 		manual = 1;
1733 		break;
1734 	}
1735 
1736 	active = ((struct ifnet *)fin->fin_ifp)->if_csum_flags_rx & pflag;
1737 	active |= M_CSUM_TCP_UDP_BAD | M_CSUM_DATA;
1738 	cflags = m->m_pkthdr.csum_flags & active;
1739 
1740 	if (pflag != 0) {
1741 		if (cflags == (pflag | M_CSUM_TCP_UDP_BAD)) {
1742 			fin->fin_flx |= FI_BAD;
1743 			fin->fin_cksum = FI_CK_BAD;
1744 		} else if (cflags == (pflag | M_CSUM_DATA)) {
1745 			if ((m->m_pkthdr.csum_data ^ 0xffff) != 0) {
1746 				fin->fin_flx |= FI_BAD;
1747 				fin->fin_cksum = FI_CK_BAD;
1748 			} else {
1749 				fin->fin_cksum = FI_CK_SUMOK;
1750 			}
1751 		} else if (cflags == pflag) {
1752 			fin->fin_cksum = FI_CK_SUMOK;
1753 		} else {
1754 			manual = 1;
1755 		}
1756 	}
1757 skipauto:
1758 	if (manual != 0) {
1759 		if (ipf_checkl4sum(fin) == -1) {
1760 			fin->fin_flx |= FI_BAD;
1761 			return -1;
1762 		}
1763 	}
1764 #else
1765 	if (ipf_checkl4sum(fin) == -1) {
1766 		fin->fin_flx |= FI_BAD;
1767 		return -1;
1768 	}
1769 #endif
1770 	return 0;
1771 }
1772 
1773 
1774 #ifdef USE_INET6
1775 EXTERN_INLINE int
1776 ipf_checkv6sum(fr_info_t *fin)
1777 {
1778 # ifdef M_CSUM_TCP_UDP_BAD
1779 	int manual, pflag, cflags, active;
1780 	mb_t *m;
1781 
1782 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
1783 		return 0;
1784 
1785 	if ((fin->fin_flx & FI_SHORT) != 0)
1786 		return 1;
1787 
1788 	if (fin->fin_cksum != FI_CK_SUMOK)
1789 		return (fin->fin_cksum > FI_CK_NEEDED) ? 0 : -1;
1790 
1791 
1792 	manual = 0;
1793 	m = fin->fin_m;
1794 
1795 	switch (fin->fin_p)
1796 	{
1797 	case IPPROTO_UDP :
1798 		pflag = M_CSUM_UDPv6;
1799 		break;
1800 	case IPPROTO_TCP :
1801 		pflag = M_CSUM_TCPv6;
1802 		break;
1803 	default :
1804 		pflag = 0;
1805 		manual = 1;
1806 		break;
1807 	}
1808 
1809 	active = ((struct ifnet *)fin->fin_ifp)->if_csum_flags_rx & pflag;
1810 	active |= M_CSUM_TCP_UDP_BAD | M_CSUM_DATA;
1811 	cflags = m->m_pkthdr.csum_flags & active;
1812 
1813 	if (pflag != 0) {
1814 		if (cflags == (pflag | M_CSUM_TCP_UDP_BAD)) {
1815 			fin->fin_flx |= FI_BAD;
1816 		} else if (cflags == (pflag | M_CSUM_DATA)) {
1817 			if ((m->m_pkthdr.csum_data ^ 0xffff) != 0)
1818 				fin->fin_flx |= FI_BAD;
1819 		} else if (cflags == pflag) {
1820 			;
1821 		} else {
1822 			manual = 1;
1823 		}
1824 	}
1825 	if (manual != 0) {
1826 		if (ipf_checkl4sum(fin) == -1) {
1827 			fin->fin_flx |= FI_BAD;
1828 			return -1;
1829 		}
1830 	}
1831 # else
1832 	if (ipf_checkl4sum(fin) == -1) {
1833 		fin->fin_flx |= FI_BAD;
1834 		return -1;
1835 	}
1836 # endif
1837 	return 0;
1838 }
1839 #endif /* USE_INET6 */
1840 
1841 
1842 size_t
1843 mbufchainlen(struct mbuf *m0)
1844 {
1845 	size_t len;
1846 
1847 	if ((m0->m_flags & M_PKTHDR) != 0) {
1848 		len = m0->m_pkthdr.len;
1849 	} else {
1850 		struct mbuf *m;
1851 
1852 		for (m = m0, len = 0; m != NULL; m = m->m_next)
1853 			len += m->m_len;
1854 	}
1855 	return len;
1856 }
1857 
1858 
1859 /* ------------------------------------------------------------------------ */
1860 /* Function:    ipf_pullup                                                  */
1861 /* Returns:     NULL == pullup failed, else pointer to protocol header      */
1862 /* Parameters:  xmin(I)- pointer to buffer where data packet starts         */
1863 /*              fin(I) - pointer to packet information                      */
1864 /*              len(I) - number of bytes to pullup                          */
1865 /*                                                                          */
1866 /* Attempt to move at least len bytes (from the start of the buffer) into a */
1867 /* single buffer for ease of access.  Operating system native functions are */
1868 /* used to manage buffers - if necessary.  If the entire packet ends up in  */
1869 /* a single buffer, set the FI_COALESCE flag even though ipf_coalesce() has */
1870 /* not been called.  Both fin_ip and fin_dp are updated before exiting _IF_ */
1871 /* and ONLY if the pullup succeeds.                                         */
1872 /*                                                                          */
1873 /* We assume that 'xmin' is a pointer to a buffer that is part of the chain */
1874 /* of buffers that starts at *fin->fin_mp.                                  */
1875 /* ------------------------------------------------------------------------ */
1876 void *
1877 ipf_pullup(mb_t *xmin, fr_info_t *fin, int len)
1878 {
1879 	int dpoff, ipoff;
1880 	mb_t *m = xmin;
1881 	char *ip;
1882 
1883 	if (m == NULL)
1884 		return NULL;
1885 
1886 	ip = (char *)fin->fin_ip;
1887 	if ((fin->fin_flx & FI_COALESCE) != 0)
1888 		return ip;
1889 
1890 	ipoff = fin->fin_ipoff;
1891 	if (fin->fin_dp != NULL)
1892 		dpoff = (char *)fin->fin_dp - (char *)ip;
1893 	else
1894 		dpoff = 0;
1895 
1896 	if (M_LEN(m) < len) {
1897 		mb_t *n = *fin->fin_mp;
1898 		/*
1899 		 * Assume that M_PKTHDR is set and just work with what is left
1900 		 * rather than check..
1901 		 * Should not make any real difference, anyway.
1902 		 */
1903 		if (m != n) {
1904 			/*
1905 			 * Record the mbuf that points to the mbuf that we're
1906 			 * about to go to work on so that we can update the
1907 			 * m_next appropriately later.
1908 			 */
1909 			for (; n->m_next != m; n = n->m_next)
1910 				;
1911 		} else {
1912 			n = NULL;
1913 		}
1914 
1915 #ifdef MHLEN
1916 		if (len > MHLEN)
1917 #else
1918 		if (len > MLEN)
1919 #endif
1920 		{
1921 #ifdef HAVE_M_PULLDOWN
1922 			if (m_pulldown(m, 0, len, NULL) == NULL)
1923 				m = NULL;
1924 #else
1925 			FREE_MB_T(*fin->fin_mp);
1926 			m = NULL;
1927 			n = NULL;
1928 #endif
1929 		} else
1930 		{
1931 			m = m_pullup(m, len);
1932 		}
1933 		if (n != NULL)
1934 			n->m_next = m;
1935 		if (m == NULL) {
1936 			/*
1937 			 * When n is non-NULL, it indicates that m pointed to
1938 			 * a sub-chain (tail) of the mbuf and that the head
1939 			 * of this chain has not yet been free'd.
1940 			 */
1941 			if (n != NULL) {
1942 				FREE_MB_T(*fin->fin_mp);
1943 			}
1944 
1945 			*fin->fin_mp = NULL;
1946 			fin->fin_m = NULL;
1947 			return NULL;
1948 		}
1949 
1950 		if (n == NULL)
1951 			*fin->fin_mp = m;
1952 
1953 		while (M_LEN(m) == 0) {
1954 			m = m->m_next;
1955 		}
1956 		fin->fin_m = m;
1957 		ip = MTOD(m, char *) + ipoff;
1958 
1959 		fin->fin_ip = (ip_t *)ip;
1960 		if (fin->fin_dp != NULL)
1961 			fin->fin_dp = (char *)fin->fin_ip + dpoff;
1962 		if (fin->fin_fraghdr != NULL)
1963 			fin->fin_fraghdr = (char *)ip +
1964 					   ((char *)fin->fin_fraghdr -
1965 					    (char *)fin->fin_ip);
1966 	}
1967 
1968 	if (len == fin->fin_plen)
1969 		fin->fin_flx |= FI_COALESCE;
1970 	return ip;
1971 }
1972 
1973 
1974 int
1975 ipf_inject(fr_info_t *fin, mb_t *m)
1976 {
1977 	int error;
1978 
1979 	if (fin->fin_out == 0) {
1980 		if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1981 			FREE_MB_T(m);
1982 			error = ENOBUFS;
1983 		} else {
1984 			error = 0;
1985 		}
1986 	} else {
1987 		error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1988 	}
1989 	return error;
1990 }
1991 
1992 
1993 u_32_t
1994 ipf_random(void)
1995 {
1996 	int number;
1997 
1998 #ifdef _CPRNG_H
1999 	number = cprng_fast32();
2000 #else
2001 	number = arc4random();
2002 #endif
2003 	return number;
2004 }
2005 
2006 
2007 /*
2008  * routines below for saving IP headers to buffer
2009  */
2010 static int ipfopen(dev_t dev, int flags
2011 #if (NetBSD >= 199511)
2012     , int devtype, PROC_T *p
2013 #endif
2014 )
2015 {
2016 	u_int unit = GET_MINOR(dev);
2017 	int error;
2018 
2019 	if (IPL_LOGMAX < unit) {
2020 		error = ENXIO;
2021 	} else {
2022 		switch (unit)
2023 		{
2024 		case IPL_LOGIPF :
2025 		case IPL_LOGNAT :
2026 		case IPL_LOGSTATE :
2027 		case IPL_LOGAUTH :
2028 		case IPL_LOGLOOKUP :
2029 		case IPL_LOGSYNC :
2030 #ifdef IPFILTER_SCAN
2031 		case IPL_LOGSCAN :
2032 #endif
2033 			error = 0;
2034 			break;
2035 		default :
2036 			error = ENXIO;
2037 			break;
2038 		}
2039 	}
2040 #if (__NetBSD_Version__ >= 799003000)
2041 	if (error == 0) {
2042 		mutex_enter(&ipf_ref_mutex);
2043 		ipf_active = 1;
2044 		mutex_exit(&ipf_ref_mutex);
2045 	}
2046 #endif
2047 	return error;
2048 }
2049 
2050 
2051 static int ipfclose(dev_t dev, int flags
2052 #if (NetBSD >= 199511)
2053 	, int devtype, PROC_T *p
2054 #endif
2055 )
2056 {
2057 	u_int	unit = GET_MINOR(dev);
2058 
2059 	if (IPL_LOGMAX < unit)
2060 		return ENXIO;
2061 	else {
2062 #if (__NetBSD_Version__ >= 799003000)
2063 		mutex_enter(&ipf_ref_mutex);
2064 		ipf_active = 0;
2065 		mutex_exit(&ipf_ref_mutex);
2066 #endif
2067 		return 0;
2068 	}
2069 }
2070 
2071 /*
2072  * ipfread/ipflog
2073  * both of these must operate with at least splnet() lest they be
2074  * called during packet processing and cause an inconsistancy to appear in
2075  * the filter lists.
2076  */
2077 static int ipfread(dev_t dev, struct uio *uio, int ioflag)
2078 {
2079 
2080 	if (ipfmain.ipf_running < 1) {
2081 		ipfmain.ipf_interror = 130006;
2082 		return EIO;
2083 	}
2084 
2085 	if (GET_MINOR(dev) == IPL_LOGSYNC)
2086 		return ipf_sync_read(&ipfmain, uio);
2087 
2088 #ifdef IPFILTER_LOG
2089 	return ipf_log_read(&ipfmain, GET_MINOR(dev), uio);
2090 #else
2091 	ipfmain.ipf_interror = 130007;
2092 	return ENXIO;
2093 #endif
2094 }
2095 
2096 
2097 /*
2098  * ipfwrite
2099  * both of these must operate with at least splnet() lest they be
2100  * called during packet processing and cause an inconsistancy to appear in
2101  * the filter lists.
2102  */
2103 static int ipfwrite(dev_t dev, struct uio *uio, int ioflag)
2104 {
2105 
2106 	if (ipfmain.ipf_running < 1) {
2107 		ipfmain.ipf_interror = 130008;
2108 		return EIO;
2109 	}
2110 
2111 	if (GET_MINOR(dev) == IPL_LOGSYNC)
2112 		return ipf_sync_write(&ipfmain, uio);
2113 	ipfmain.ipf_interror = 130009;
2114 	return ENXIO;
2115 }
2116 
2117 
2118 static int ipfpoll(dev_t dev, int events, PROC_T *p)
2119 {
2120 	u_int unit = GET_MINOR(dev);
2121 	int revents = 0;
2122 
2123 	if (IPL_LOGMAX < unit) {
2124 		ipfmain.ipf_interror = 130010;
2125 		return ENXIO;
2126 	}
2127 
2128 	switch (unit)
2129 	{
2130 	case IPL_LOGIPF :
2131 	case IPL_LOGNAT :
2132 	case IPL_LOGSTATE :
2133 #ifdef IPFILTER_LOG
2134 		if ((events & (POLLIN | POLLRDNORM)) &&
2135 		    ipf_log_canread(&ipfmain, unit))
2136 			revents |= events & (POLLIN | POLLRDNORM);
2137 #endif
2138 		break;
2139 	case IPL_LOGAUTH :
2140 		if ((events & (POLLIN | POLLRDNORM)) &&
2141 		    ipf_auth_waiting(&ipfmain))
2142 			revents |= events & (POLLIN | POLLRDNORM);
2143 		break;
2144 	case IPL_LOGSYNC :
2145 		if ((events & (POLLIN | POLLRDNORM)) &&
2146 		    ipf_sync_canread(&ipfmain))
2147 			revents |= events & (POLLIN | POLLRDNORM);
2148 		if ((events & (POLLOUT | POLLWRNORM)) &&
2149 		    ipf_sync_canwrite(&ipfmain))
2150 			revents |= events & (POLLOUT | POLLWRNORM);
2151 		break;
2152 	case IPL_LOGSCAN :
2153 	case IPL_LOGLOOKUP :
2154 	default :
2155 		break;
2156 	}
2157 
2158 	if ((revents == 0) && (((events & (POLLIN|POLLRDNORM)) != 0)))
2159 		selrecord(p, &ipfmain.ipf_selwait[unit]);
2160 	return revents;
2161 }
2162 
2163 u_int
2164 ipf_pcksum(fr_info_t *fin, int hlen, u_int sum)
2165 {
2166 	struct mbuf *m;
2167 	u_int sum2;
2168 	int off;
2169 
2170 	m = fin->fin_m;
2171 	off = (char *)fin->fin_dp - (char *)fin->fin_ip;
2172 	m->m_data += hlen;
2173 	m->m_len -= hlen;
2174 	sum2 = in_cksum(fin->fin_m, fin->fin_plen - off);
2175 	m->m_len += hlen;
2176 	m->m_data -= hlen;
2177 
2178 	/*
2179 	 * Both sum and sum2 are partial sums, so combine them together.
2180 	 */
2181 	sum += ~sum2 & 0xffff;
2182 	while (sum > 0xffff)
2183 		sum = (sum & 0xffff) + (sum >> 16);
2184 	sum2 = ~sum & 0xffff;
2185 	return sum2;
2186 }
2187 
2188 #if (__NetBSD_Version__ >= 799003000)
2189 
2190 /* NetBSD module interface */
2191 
2192 MODULE(MODULE_CLASS_DRIVER, ipl, "bpf_filter");
2193 
2194 static int ipl_init(void *);
2195 static int ipl_fini(void *);
2196 static int ipl_modcmd(modcmd_t, void *);
2197 
2198 #ifdef _MODULE
2199 static devmajor_t ipl_cmaj = -1, ipl_bmaj = -1;
2200 #endif
2201 
2202 static int
2203 ipl_modcmd(modcmd_t cmd, void *opaque)
2204 {
2205 
2206 	switch (cmd) {
2207 	case MODULE_CMD_INIT:
2208 		return ipl_init(opaque);
2209 	case MODULE_CMD_FINI:
2210 		return ipl_fini(opaque);
2211 	default:
2212 		return ENOTTY;
2213 	}
2214 }
2215 
2216 static int
2217 ipl_init(void *opaque)
2218 {
2219 	int error;
2220 
2221 	ipf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
2222 	    ipf_listener_cb, NULL);
2223 
2224 	if ((error = ipf_load_all()) != 0)
2225 		return error;
2226 
2227 	if (ipf_create_all(&ipfmain) == NULL) {
2228 		ipf_unload_all();
2229 		return ENODEV;
2230 	}
2231 
2232 	/* Initialize our mutex and reference count */
2233 	mutex_init(&ipf_ref_mutex, MUTEX_DEFAULT, IPL_NONE);
2234 	ipf_active = 0;
2235 
2236 #ifdef _MODULE
2237 	/*
2238 	 * Insert ourself into the cdevsw list.
2239 	 */
2240 	error = devsw_attach("ipl", NULL, &ipl_bmaj, &ipl_cdevsw, &ipl_cmaj);
2241 	if (error)
2242 		ipl_fini(opaque);
2243 #endif
2244 
2245 	return error;
2246 }
2247 
2248 static int
2249 ipl_fini(void *opaque)
2250 {
2251 
2252 #ifdef _MODULE
2253 	(void)devsw_detach(NULL, &ipl_cdevsw);
2254 #endif
2255 
2256 	/*
2257 	 * Grab the mutex, verify that there are no references
2258 	 * and that there are no running filters.  If either
2259 	 * of these exists, reinsert our cdevsw entry and return
2260 	 * an error.
2261 	 */
2262 	mutex_enter(&ipf_ref_mutex);
2263 	if (ipf_active != 0 || ipfmain.ipf_running > 0) {
2264 #ifdef _MODULE
2265 		(void)devsw_attach("ipl", NULL, &ipl_bmaj,
2266 		    &ipl_cdevsw, &ipl_cmaj);
2267 #endif
2268 		mutex_exit(&ipf_ref_mutex);
2269 		return EBUSY;
2270 	}
2271 
2272 	/* Clean up the rest of our state before being unloaded */
2273 
2274 	mutex_exit(&ipf_ref_mutex);
2275 	mutex_destroy(&ipf_ref_mutex);
2276 	ipf_destroy_all(&ipfmain);
2277 	ipf_unload_all();
2278 	kauth_unlisten_scope(ipf_listener);
2279 
2280 	return 0;
2281 }
2282 #endif /* (__NetBSD_Version__ >= 799003000) */
2283