xref: /titanic_52/usr/src/uts/common/inet/ipf/fil.c (revision 37b285d61e57561b538d999834714df1f5c2db53)
1 /*
2  * Copyright (C) 1993-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
7  * Use is subject to license terms.
8  */
9 
10 #if defined(KERNEL) || defined(_KERNEL)
11 # undef KERNEL
12 # undef _KERNEL
13 # define        KERNEL	1
14 # define        _KERNEL	1
15 #endif
16 #include <sys/errno.h>
17 #include <sys/types.h>
18 #include <sys/param.h>
19 #include <sys/time.h>
20 #if defined(__NetBSD__)
21 # if (NetBSD >= 199905) && !defined(IPFILTER_LKM) && defined(_KERNEL)
22 #  include "opt_ipfilter_log.h"
23 # endif
24 #endif
25 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
26     (__FreeBSD_version >= 220000)
27 # if (__FreeBSD_version >= 400000)
28 #  if !defined(IPFILTER_LKM)
29 #   include "opt_inet6.h"
30 #  endif
31 #  if (__FreeBSD_version == 400019)
32 #   define CSUM_DELAY_DATA
33 #  endif
34 # endif
35 # include <sys/filio.h>
36 #else
37 # include <sys/ioctl.h>
38 #endif
39 #if !defined(_AIX51)
40 # include <sys/fcntl.h>
41 #endif
42 #if defined(_KERNEL)
43 # include <sys/systm.h>
44 # include <sys/file.h>
45 #else
46 # include <stdio.h>
47 # include <string.h>
48 # include <stdlib.h>
49 # include <stddef.h>
50 # include <sys/file.h>
51 # define _KERNEL
52 # ifdef __OpenBSD__
53 struct file;
54 # endif
55 # include <sys/uio.h>
56 # undef _KERNEL
57 #endif
58 #if !defined(__SVR4) && !defined(__svr4__) && !defined(__hpux) && \
59     !defined(linux)
60 # include <sys/mbuf.h>
61 #else
62 # if !defined(linux)
63 #  include <sys/byteorder.h>
64 # endif
65 # if (SOLARIS2 < 5) && defined(sun)
66 #  include <sys/dditypes.h>
67 # endif
68 #endif
69 #ifdef __hpux
70 # define _NET_ROUTE_INCLUDED
71 #endif
72 #if !defined(linux)
73 # include <sys/protosw.h>
74 #endif
75 #include <sys/socket.h>
76 #include <net/if.h>
77 #ifdef sun
78 # include <net/af.h>
79 #endif
80 #if !defined(_KERNEL) && defined(__FreeBSD__)
81 # include "radix_ipf.h"
82 #endif
83 #include <net/route.h>
84 #include <netinet/in.h>
85 #include <netinet/in_systm.h>
86 #include <netinet/ip.h>
87 #if !defined(linux)
88 # include <netinet/ip_var.h>
89 #endif
90 #if defined(__sgi) && defined(IFF_DRVRLOCK) /* IRIX 6 */
91 # include <sys/hashing.h>
92 # include <netinet/in_var.h>
93 #endif
94 #include <netinet/tcp.h>
95 #if (!defined(__sgi) && !defined(AIX)) || defined(_KERNEL)
96 # include <netinet/udp.h>
97 # include <netinet/ip_icmp.h>
98 #endif
99 #ifdef __hpux
100 # undef _NET_ROUTE_INCLUDED
101 #endif
102 #include "netinet/ip_compat.h"
103 #ifdef	USE_INET6
104 # include <netinet/icmp6.h>
105 # if !SOLARIS && defined(_KERNEL) && !defined(__osf__) && !defined(__hpux)
106 #  include <netinet6/in6_var.h>
107 # endif
108 #endif
109 #include <netinet/tcpip.h>
110 #include "netinet/ip_fil.h"
111 #include "netinet/ip_nat.h"
112 #include "netinet/ip_frag.h"
113 #include "netinet/ip_state.h"
114 #include "netinet/ip_proxy.h"
115 #include "netinet/ip_auth.h"
116 #include "netinet/ipf_stack.h"
117 #ifdef IPFILTER_SCAN
118 # include "netinet/ip_scan.h"
119 #endif
120 #ifdef IPFILTER_SYNC
121 # include "netinet/ip_sync.h"
122 #endif
123 #include "netinet/ip_pool.h"
124 #include "netinet/ip_htable.h"
125 #ifdef IPFILTER_COMPILED
126 # include "netinet/ip_rules.h"
127 #endif
128 #if defined(IPFILTER_BPF) && defined(_KERNEL)
129 # include <net/bpf.h>
130 #endif
131 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
132 # include <sys/malloc.h>
133 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
134 #  include "opt_ipfilter.h"
135 # endif
136 #endif
137 #include "netinet/ipl.h"
138 /* END OF INCLUDES */
139 
140 #if !defined(lint)
141 static const char sccsid[] = "@(#)fil.c	1.36 6/5/96 (C) 1993-2000 Darren Reed";
142 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
143 #endif
144 
145 #ifndef	_KERNEL
146 # include "ipf.h"
147 # include "ipt.h"
148 # include "bpf-ipf.h"
149 extern	int	opts;
150 
151 # define	FR_VERBOSE(verb_pr)			verbose verb_pr
152 # define	FR_DEBUG(verb_pr)			debug verb_pr
153 #else /* #ifndef _KERNEL */
154 # define	FR_VERBOSE(verb_pr)
155 # define	FR_DEBUG(verb_pr)
156 #endif /* _KERNEL */
157 
158 
159 char	ipfilter_version[] = IPL_VERSION;
160 int	fr_features = 0
161 #ifdef	IPFILTER_LKM
162 		| IPF_FEAT_LKM
163 #endif
164 #ifdef	IPFILTER_LOG
165 		| IPF_FEAT_LOG
166 #endif
167 #ifdef	IPFILTER_LOOKUP
168 		| IPF_FEAT_LOOKUP
169 #endif
170 #ifdef	IPFILTER_BPF
171 		| IPF_FEAT_BPF
172 #endif
173 #ifdef	IPFILTER_COMPILED
174 		| IPF_FEAT_COMPILED
175 #endif
176 #ifdef	IPFILTER_CKSUM
177 		| IPF_FEAT_CKSUM
178 #endif
179 #ifdef	IPFILTER_SYNC
180 		| IPF_FEAT_SYNC
181 #endif
182 #ifdef	IPFILTER_SCAN
183 		| IPF_FEAT_SCAN
184 #endif
185 #ifdef	USE_INET6
186 		| IPF_FEAT_IPV6
187 #endif
188 	;
189 
190 #define	IPF_BUMP(x)	(x)++
191 
192 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
193 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
194 static	int		fr_portcheck __P((frpcmp_t *, u_short *));
195 static	int		frflushlist __P((int, minor_t, int *, frentry_t **,
196 					 ipf_stack_t *));
197 static	ipfunc_t	fr_findfunc __P((ipfunc_t));
198 static	frentry_t	*fr_firewall __P((fr_info_t *, u_32_t *));
199 static	int		fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
200 static	INLINE void	frpr_ah __P((fr_info_t *));
201 static	INLINE void	frpr_esp __P((fr_info_t *));
202 static	INLINE void	frpr_gre __P((fr_info_t *));
203 static	INLINE void	frpr_udp __P((fr_info_t *));
204 static	INLINE void	frpr_tcp __P((fr_info_t *));
205 static	INLINE void	frpr_icmp __P((fr_info_t *));
206 static	INLINE void	frpr_ipv4hdr __P((fr_info_t *));
207 static	INLINE int	frpr_pullup __P((fr_info_t *, int));
208 static	INLINE void	frpr_short __P((fr_info_t *, int));
209 static	INLINE void	frpr_tcpcommon __P((fr_info_t *));
210 static	INLINE void	frpr_udpcommon __P((fr_info_t *));
211 static	INLINE int	fr_updateipid __P((fr_info_t *));
212 #ifdef	IPFILTER_LOOKUP
213 static	int		fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
214 static	INLINE void	*fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
215 					       ipf_stack_t *));
216 #endif
217 static	void		frsynclist __P((int, int, void *, char *, frentry_t *,
218     ipf_stack_t *));
219 static	void		*fr_ifsync __P((int, int, char *, char *,
220 					void *, void *, ipf_stack_t *));
221 static	ipftuneable_t	*fr_findtunebyname __P((const char *, ipf_stack_t *));
222 static	ipftuneable_t	*fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
223 
224 /*
225  * bit values for identifying presence of individual IP options
226  * All of these tables should be ordered by increasing key value on the left
227  * hand side to allow for binary searching of the array and include a trailer
228  * with a 0 for the bitmask for linear searches to easily find the end with.
229  */
230 const	struct	optlist	ipopts[20] = {
231 	{ IPOPT_NOP,	0x000001 },
232 	{ IPOPT_RR,	0x000002 },
233 	{ IPOPT_ZSU,	0x000004 },
234 	{ IPOPT_MTUP,	0x000008 },
235 	{ IPOPT_MTUR,	0x000010 },
236 	{ IPOPT_ENCODE,	0x000020 },
237 	{ IPOPT_TS,	0x000040 },
238 	{ IPOPT_TR,	0x000080 },
239 	{ IPOPT_SECURITY, 0x000100 },
240 	{ IPOPT_LSRR,	0x000200 },
241 	{ IPOPT_E_SEC,	0x000400 },
242 	{ IPOPT_CIPSO,	0x000800 },
243 	{ IPOPT_SATID,	0x001000 },
244 	{ IPOPT_SSRR,	0x002000 },
245 	{ IPOPT_ADDEXT,	0x004000 },
246 	{ IPOPT_VISA,	0x008000 },
247 	{ IPOPT_IMITD,	0x010000 },
248 	{ IPOPT_EIP,	0x020000 },
249 	{ IPOPT_FINN,	0x040000 },
250 	{ 0,		0x000000 }
251 };
252 
253 #ifdef USE_INET6
254 struct optlist ip6exthdr[] = {
255 	{ IPPROTO_HOPOPTS,		0x000001 },
256 	{ IPPROTO_IPV6,			0x000002 },
257 	{ IPPROTO_ROUTING,		0x000004 },
258 	{ IPPROTO_FRAGMENT,		0x000008 },
259 	{ IPPROTO_ESP,			0x000010 },
260 	{ IPPROTO_AH,			0x000020 },
261 	{ IPPROTO_NONE,			0x000040 },
262 	{ IPPROTO_DSTOPTS,		0x000080 },
263 	{ 0,				0 }
264 };
265 #endif
266 
267 struct optlist tcpopts[] = {
268 	{ TCPOPT_NOP,			0x000001 },
269 	{ TCPOPT_MAXSEG,		0x000002 },
270 	{ TCPOPT_WINDOW,		0x000004 },
271 	{ TCPOPT_SACK_PERMITTED,	0x000008 },
272 	{ TCPOPT_SACK,			0x000010 },
273 	{ TCPOPT_TIMESTAMP,		0x000020 },
274 	{ 0,				0x000000 }
275 };
276 
277 /*
278  * bit values for identifying presence of individual IP security options
279  */
280 const	struct	optlist	secopt[8] = {
281 	{ IPSO_CLASS_RES4,	0x01 },
282 	{ IPSO_CLASS_TOPS,	0x02 },
283 	{ IPSO_CLASS_SECR,	0x04 },
284 	{ IPSO_CLASS_RES3,	0x08 },
285 	{ IPSO_CLASS_CONF,	0x10 },
286 	{ IPSO_CLASS_UNCL,	0x20 },
287 	{ IPSO_CLASS_RES2,	0x40 },
288 	{ IPSO_CLASS_RES1,	0x80 }
289 };
290 
291 
292 /*
293  * Table of functions available for use with call rules.
294  */
295 static ipfunc_resolve_t fr_availfuncs[] = {
296 #ifdef	IPFILTER_LOOKUP
297 	{ "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
298 	{ "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
299 #endif
300 	{ "", NULL }
301 };
302 
303 
304 /*
305  * Below we declare a list of constants used only by the ipf_extraflush()
306  * routine.  We are placing it here, instead of in ipf_extraflush() itself,
307  * because we want to make it visible to tools such as mdb, nm etc., so the
308  * values can easily be altered during debugging.
309  */
310 static	const	int	idletime_tab[] = {
311 	IPF_TTLVAL(30),		/* 30 seconds */
312 	IPF_TTLVAL(1800),	/* 30 minutes */
313 	IPF_TTLVAL(43200),	/* 12 hours */
314 	IPF_TTLVAL(345600),	/* 4 days */
315 };
316 
317 
318 /*
319  * The next section of code is a a collection of small routines that set
320  * fields in the fr_info_t structure passed based on properties of the
321  * current packet.  There are different routines for the same protocol
322  * for each of IPv4 and IPv6.  Adding a new protocol, for which there
323  * will "special" inspection for setup, is now more easily done by adding
324  * a new routine and expanding the frpr_ipinit*() function rather than by
325  * adding more code to a growing switch statement.
326  */
327 #ifdef USE_INET6
328 static	INLINE int	frpr_ah6 __P((fr_info_t *));
329 static	INLINE void	frpr_esp6 __P((fr_info_t *));
330 static	INLINE void	frpr_gre6 __P((fr_info_t *));
331 static	INLINE void	frpr_udp6 __P((fr_info_t *));
332 static	INLINE void	frpr_tcp6 __P((fr_info_t *));
333 static	INLINE void	frpr_icmp6 __P((fr_info_t *));
334 static	INLINE void	frpr_ipv6hdr __P((fr_info_t *));
335 static	INLINE void	frpr_short6 __P((fr_info_t *, int));
336 static	INLINE int	frpr_hopopts6 __P((fr_info_t *));
337 static	INLINE int	frpr_routing6 __P((fr_info_t *));
338 static	INLINE int	frpr_dstopts6 __P((fr_info_t *));
339 static	INLINE int	frpr_fragment6 __P((fr_info_t *));
340 static	INLINE int	frpr_ipv6exthdr __P((fr_info_t *, int, int));
341 
342 
343 /* ------------------------------------------------------------------------ */
344 /* Function:    frpr_short6                                                 */
345 /* Returns:     void                                                        */
346 /* Parameters:  fin(I) - pointer to packet information                      */
347 /*                                                                          */
348 /* IPv6 Only                                                                */
349 /* This is function enforces the 'is a packet too short to be legit' rule   */
350 /* for IPv6 and marks the packet with FI_SHORT if so.  See function comment */
351 /* for frpr_short() for more details.                                       */
352 /* ------------------------------------------------------------------------ */
353 static INLINE void frpr_short6(fin, xmin)
354 fr_info_t *fin;
355 int xmin;
356 {
357 
358 	if (fin->fin_dlen < xmin)
359 		fin->fin_flx |= FI_SHORT;
360 }
361 
362 
363 /* ------------------------------------------------------------------------ */
364 /* Function:    frpr_ipv6hdr                                                */
365 /* Returns:     Nil                                                         */
366 /* Parameters:  fin(I) - pointer to packet information                      */
367 /*                                                                          */
368 /* IPv6 Only                                                                */
369 /* Copy values from the IPv6 header into the fr_info_t struct and call the  */
370 /* per-protocol analyzer if it exists.                                      */
371 /* ------------------------------------------------------------------------ */
372 static INLINE void frpr_ipv6hdr(fin)
373 fr_info_t *fin;
374 {
375 	ip6_t *ip6 = (ip6_t *)fin->fin_ip;
376 	int p, go = 1, i, hdrcount;
377 	fr_ip_t *fi = &fin->fin_fi;
378 
379 	fin->fin_off = 0;
380 
381 	fi->fi_tos = 0;
382 	fi->fi_optmsk = 0;
383 	fi->fi_secmsk = 0;
384 	fi->fi_auth = 0;
385 
386 	p = ip6->ip6_nxt;
387 	fi->fi_ttl = ip6->ip6_hlim;
388 	fi->fi_src.in6 = ip6->ip6_src;
389 	fi->fi_dst.in6 = ip6->ip6_dst;
390 	fin->fin_id = 0;
391 
392 	hdrcount = 0;
393 	while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
394 		switch (p)
395 		{
396 		case IPPROTO_UDP :
397 			frpr_udp6(fin);
398 			go = 0;
399 			break;
400 
401 		case IPPROTO_TCP :
402 			frpr_tcp6(fin);
403 			go = 0;
404 			break;
405 
406 		case IPPROTO_ICMPV6 :
407 			frpr_icmp6(fin);
408 			go = 0;
409 			break;
410 
411 		case IPPROTO_GRE :
412 			frpr_gre6(fin);
413 			go = 0;
414 			break;
415 
416 		case IPPROTO_HOPOPTS :
417 			/*
418 			 * hop by hop ext header is only allowed
419 			 * right after IPv6 header.
420 			 */
421 			if (hdrcount != 0) {
422 				fin->fin_flx |= FI_BAD;
423 				p = IPPROTO_NONE;
424 			} else {
425 				p = frpr_hopopts6(fin);
426 			}
427 			break;
428 
429 		case IPPROTO_DSTOPTS :
430 			p = frpr_dstopts6(fin);
431 			break;
432 
433 		case IPPROTO_ROUTING :
434 			p = frpr_routing6(fin);
435 			break;
436 
437 		case IPPROTO_AH :
438 			p = frpr_ah6(fin);
439 			break;
440 
441 		case IPPROTO_ESP :
442 			frpr_esp6(fin);
443 			go = 0;
444 			break;
445 
446 		case IPPROTO_IPV6 :
447 			for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
448 				if (ip6exthdr[i].ol_val == p) {
449 					fin->fin_flx |= ip6exthdr[i].ol_bit;
450 					break;
451 				}
452 			go = 0;
453 			break;
454 
455 		case IPPROTO_NONE :
456 			go = 0;
457 			break;
458 
459 		case IPPROTO_FRAGMENT :
460 			p = frpr_fragment6(fin);
461 			if (fin->fin_off != 0)  /* Not the first frag */
462 				go = 0;
463 			break;
464 
465 		default :
466 			go = 0;
467 			break;
468 		}
469 		hdrcount++;
470 
471 		/*
472 		 * It is important to note that at this point, for the
473 		 * extension headers (go != 0), the entire header may not have
474 		 * been pulled up when the code gets to this point.  This is
475 		 * only done for "go != 0" because the other header handlers
476 		 * will all pullup their complete header.  The other indicator
477 		 * of an incomplete packet is that this was just an extension
478 		 * header.
479 		 */
480 		if ((go != 0) && (p != IPPROTO_NONE) &&
481 		    (frpr_pullup(fin, 0) == -1)) {
482 			p = IPPROTO_NONE;
483 			go = 0;
484 		}
485 	}
486 	fi->fi_p = p;
487 }
488 
489 
490 /* ------------------------------------------------------------------------ */
491 /* Function:    frpr_ipv6exthdr                                             */
492 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
493 /* Parameters:  fin(I)      - pointer to packet information                 */
494 /*              multiple(I) - flag indicating yes/no if multiple occurances */
495 /*                            of this extension header are allowed.         */
496 /*              proto(I)    - protocol number for this extension header     */
497 /*                                                                          */
498 /* IPv6 Only                                                                */
499 /* This function expects to find an IPv6 extension header at fin_dp.        */
500 /* There must be at least 8 bytes of data at fin_dp for there to be a valid */
501 /* extension header present. If a good one is found, fin_dp is advanced to  */
502 /* point at the first piece of data after the extension header, fin_exthdr  */
503 /* points to the start of the extension header and the "protocol" of the    */
504 /* *NEXT* header is returned.                                               */
505 /* ------------------------------------------------------------------------ */
506 static INLINE int frpr_ipv6exthdr(fin, multiple, proto)
507 fr_info_t *fin;
508 int multiple, proto;
509 {
510 	struct ip6_ext *hdr;
511 	u_short shift;
512 	int i;
513 
514 	fin->fin_flx |= FI_V6EXTHDR;
515 
516 				/* 8 is default length of extension hdr */
517 	if ((fin->fin_dlen - 8) < 0) {
518 		fin->fin_flx |= FI_SHORT;
519 		return IPPROTO_NONE;
520 	}
521 
522 	if (frpr_pullup(fin, 8) == -1)
523 		return IPPROTO_NONE;
524 
525 	hdr = fin->fin_dp;
526 	shift = 8 + (hdr->ip6e_len << 3);
527 	if (shift > fin->fin_dlen) {	/* Nasty extension header length? */
528 		fin->fin_flx |= FI_BAD;
529 		return IPPROTO_NONE;
530 	}
531 
532 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
533 		if (ip6exthdr[i].ol_val == proto) {
534 			/*
535 			 * Most IPv6 extension headers are only allowed once.
536 			 */
537 			if ((multiple == 0) &&
538 			    ((fin->fin_optmsk & ip6exthdr[i].ol_bit) != 0))
539 				fin->fin_flx |= FI_BAD;
540 			else
541 				fin->fin_optmsk |= ip6exthdr[i].ol_bit;
542 			break;
543 		}
544 
545 	fin->fin_dp = (char *)fin->fin_dp + shift;
546 	fin->fin_dlen -= shift;
547 
548 	return hdr->ip6e_nxt;
549 }
550 
551 
552 /* ------------------------------------------------------------------------ */
553 /* Function:    frpr_hopopts6                                               */
554 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
555 /* Parameters:  fin(I) - pointer to packet information                      */
556 /*                                                                          */
557 /* IPv6 Only                                                                */
558 /* This is function checks pending hop by hop options extension header      */
559 /* ------------------------------------------------------------------------ */
560 static INLINE int frpr_hopopts6(fin)
561 fr_info_t *fin;
562 {
563 	return frpr_ipv6exthdr(fin, 0, IPPROTO_HOPOPTS);
564 }
565 
566 
567 /* ------------------------------------------------------------------------ */
568 /* Function:    frpr_routing6                                               */
569 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
570 /* Parameters:  fin(I) - pointer to packet information                      */
571 /*                                                                          */
572 /* IPv6 Only                                                                */
573 /* This is function checks pending routing extension header                 */
574 /* ------------------------------------------------------------------------ */
575 static INLINE int frpr_routing6(fin)
576 fr_info_t *fin;
577 {
578 	struct ip6_ext *hdr;
579 	int shift;
580 
581 	hdr = fin->fin_dp;
582 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_ROUTING) == IPPROTO_NONE)
583 		return IPPROTO_NONE;
584 
585 	shift = 8 + (hdr->ip6e_len << 3);
586 	/*
587 	 * Nasty extension header length?
588 	 */
589 	if ((hdr->ip6e_len << 3) & 15) {
590 		fin->fin_flx |= FI_BAD;
591 		/*
592 		 * Compensate for the changes made in frpr_ipv6exthdr()
593 		 */
594 		fin->fin_dlen += shift;
595 		fin->fin_dp = (char *)fin->fin_dp - shift;
596 		return IPPROTO_NONE;
597 	}
598 
599 	return hdr->ip6e_nxt;
600 }
601 
602 
603 /* ------------------------------------------------------------------------ */
604 /* Function:    frpr_fragment6                                              */
605 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
606 /* Parameters:  fin(I) - pointer to packet information                      */
607 /*                                                                          */
608 /* IPv6 Only                                                                */
609 /* Examine the IPv6 fragment header and extract fragment offset information.*/
610 /*                                                                          */
611 /* We don't know where the transport layer header (or whatever is next is), */
612 /* as it could be behind destination options (amongst others).  Because     */
613 /* there is no fragment cache, there is no knowledge about whether or not an*/
614 /* upper layer header has been seen (or where it ends) and thus we are not  */
615 /* able to continue processing beyond this header with any confidence.      */
616 /* ------------------------------------------------------------------------ */
617 static INLINE int frpr_fragment6(fin)
618 fr_info_t *fin;
619 {
620 	struct ip6_frag *frag;
621 
622 	fin->fin_flx |= FI_FRAG;
623 
624 	/*
625 	 * A fragmented IPv6 packet implies that there must be something
626 	 * else after the fragment.
627 	 */
628 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
629 		return IPPROTO_NONE;
630 
631 	frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
632 
633 	/*
634 	 * If this fragment isn't the last then the packet length must
635 	 * be a multiple of 8.
636 	 */
637 	if ((frag->ip6f_offlg & IP6F_MORE_FRAG) != 0) {
638 		fin->fin_flx |= FI_MOREFRAG;
639 
640 		if ((fin->fin_plen & 0x7) != 0)
641 			fin->fin_flx |= FI_BAD;
642 	}
643 
644 	fin->fin_id = frag->ip6f_ident;
645 	fin->fin_off = ntohs(frag->ip6f_offlg & IP6F_OFF_MASK);
646 	if (fin->fin_off != 0)
647 		fin->fin_flx |= FI_FRAGBODY;
648 
649 	return frag->ip6f_nxt;
650 }
651 
652 
653 /* ------------------------------------------------------------------------ */
654 /* Function:    frpr_dstopts6                                               */
655 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
656 /* Parameters:  fin(I) - pointer to packet information                      */
657 /*              nextheader(I) - stores next header value                    */
658 /*                                                                          */
659 /* IPv6 Only                                                                */
660 /* This is function checks pending destination options extension header     */
661 /* ------------------------------------------------------------------------ */
662 static INLINE int frpr_dstopts6(fin)
663 fr_info_t *fin;
664 {
665 	return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
666 }
667 
668 
669 /* ------------------------------------------------------------------------ */
670 /* Function:    frpr_icmp6                                                  */
671 /* Returns:     void                                                        */
672 /* Parameters:  fin(I) - pointer to packet information                      */
673 /*                                                                          */
674 /* IPv6 Only                                                                */
675 /* This routine is mainly concerned with determining the minimum valid size */
676 /* for an ICMPv6 packet.                                                    */
677 /* ------------------------------------------------------------------------ */
678 static INLINE void frpr_icmp6(fin)
679 fr_info_t *fin;
680 {
681 	int minicmpsz = sizeof(struct icmp6_hdr);
682 	struct icmp6_hdr *icmp6;
683 
684 	if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
685 		return;
686 
687 	if (fin->fin_dlen > 1) {
688 		icmp6 = fin->fin_dp;
689 
690 		fin->fin_data[0] = *(u_short *)icmp6;
691 
692 		if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
693 			fin->fin_flx |= FI_ICMPQUERY;
694 
695 		switch (icmp6->icmp6_type)
696 		{
697 		case ICMP6_ECHO_REPLY :
698 		case ICMP6_ECHO_REQUEST :
699 			if (fin->fin_dlen >= 6)
700 				fin->fin_data[1] = icmp6->icmp6_id;
701 			minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
702 			break;
703 		case ICMP6_DST_UNREACH :
704 		case ICMP6_PACKET_TOO_BIG :
705 		case ICMP6_TIME_EXCEEDED :
706 		case ICMP6_PARAM_PROB :
707 			if ((fin->fin_m != NULL) &&
708 			    (M_LEN(fin->fin_m) < fin->fin_plen)) {
709 				if (fr_coalesce(fin) != 1)
710 					return;
711 			}
712 			fin->fin_flx |= FI_ICMPERR;
713 			minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
714 			break;
715 		default :
716 			break;
717 		}
718 	}
719 
720 	frpr_short6(fin, minicmpsz);
721 }
722 
723 
724 /* ------------------------------------------------------------------------ */
725 /* Function:    frpr_udp6                                                   */
726 /* Returns:     void                                                        */
727 /* Parameters:  fin(I) - pointer to packet information                      */
728 /*                                                                          */
729 /* IPv6 Only                                                                */
730 /* Analyse the packet for IPv6/UDP properties.                              */
731 /* Is not expected to be called for fragmented packets.                     */
732 /* ------------------------------------------------------------------------ */
733 static INLINE void frpr_udp6(fin)
734 fr_info_t *fin;
735 {
736 
737 	fr_checkv6sum(fin);
738 
739 	frpr_short6(fin, sizeof(struct udphdr));
740 	if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
741 		return;
742 
743 	frpr_udpcommon(fin);
744 }
745 
746 
747 /* ------------------------------------------------------------------------ */
748 /* Function:    frpr_tcp6                                                   */
749 /* Returns:     void                                                        */
750 /* Parameters:  fin(I) - pointer to packet information                      */
751 /*                                                                          */
752 /* IPv6 Only                                                                */
753 /* Analyse the packet for IPv6/TCP properties.                              */
754 /* Is not expected to be called for fragmented packets.                     */
755 /* ------------------------------------------------------------------------ */
756 static INLINE void frpr_tcp6(fin)
757 fr_info_t *fin;
758 {
759 
760 	fr_checkv6sum(fin);
761 
762 	frpr_short6(fin, sizeof(struct tcphdr));
763 	if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
764 		return;
765 
766 	frpr_tcpcommon(fin);
767 }
768 
769 
770 /* ------------------------------------------------------------------------ */
771 /* Function:    frpr_esp6                                                   */
772 /* Returns:     void                                                        */
773 /* Parameters:  fin(I) - pointer to packet information                      */
774 /*                                                                          */
775 /* IPv6 Only                                                                */
776 /* Analyse the packet for ESP properties.                                   */
777 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
778 /* even though the newer ESP packets must also have a sequence number that  */
779 /* is 32bits as well, it is not possible(?) to determine the version from a */
780 /* simple packet header.                                                    */
781 /* ------------------------------------------------------------------------ */
782 static INLINE void frpr_esp6(fin)
783 fr_info_t *fin;
784 {
785 	int i;
786 	frpr_short6(fin, sizeof(grehdr_t));
787 
788 	(void) frpr_pullup(fin, 8);
789 
790 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
791 		if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
792 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
793 			break;
794 		}
795 }
796 
797 
798 /* ------------------------------------------------------------------------ */
799 /* Function:    frpr_ah6                                                    */
800 /* Returns:     void                                                        */
801 /* Parameters:  fin(I) - pointer to packet information                      */
802 /*                                                                          */
803 /* IPv6 Only                                                                */
804 /* Analyse the packet for AH properties.                                    */
805 /* The minimum length is taken to be the combination of all fields in the   */
806 /* header being present and no authentication data (null algorithm used.)   */
807 /* ------------------------------------------------------------------------ */
808 static INLINE int frpr_ah6(fin)
809 fr_info_t *fin;
810 {
811 	authhdr_t *ah;
812 	int i, shift;
813 
814 	frpr_short6(fin, 12);
815 
816 	if (frpr_pullup(fin, sizeof(*ah)) == -1)
817 		return IPPROTO_NONE;
818 
819 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
820 		if (ip6exthdr[i].ol_val == IPPROTO_AH) {
821 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
822 			break;
823 		}
824 
825 	ah = (authhdr_t *)fin->fin_dp;
826 
827 	shift = (ah->ah_plen + 2) * 4;
828 	fin->fin_dlen -= shift;
829 	fin->fin_dp = (char*)fin->fin_dp + shift;
830 
831 	return ah->ah_next;
832 }
833 
834 
835 /* ------------------------------------------------------------------------ */
836 /* Function:    frpr_gre6                                                   */
837 /* Returns:     void                                                        */
838 /* Parameters:  fin(I) - pointer to packet information                      */
839 /*                                                                          */
840 /* Analyse the packet for GRE properties.                                   */
841 /* ------------------------------------------------------------------------ */
842 static INLINE void frpr_gre6(fin)
843 fr_info_t *fin;
844 {
845 	grehdr_t *gre;
846 
847 	frpr_short6(fin, sizeof(grehdr_t));
848 
849 	if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
850 		return;
851 
852 	gre = fin->fin_dp;
853 	if (GRE_REV(gre->gr_flags) == 1)
854 		fin->fin_data[0] = gre->gr_call;
855 }
856 #endif	/* USE_INET6 */
857 
858 
859 /* ------------------------------------------------------------------------ */
860 /* Function:    frpr_pullup                                                 */
861 /* Returns:     int     - 0 == pullup succeeded, -1 == failure              */
862 /* Parameters:  fin(I)  - pointer to packet information                     */
863 /*              plen(I) - length (excluding L3 header) to pullup            */
864 /*                                                                          */
865 /* Short inline function to cut down on code duplication to perform a call  */
866 /* to fr_pullup to ensure there is the required amount of data,             */
867 /* consecutively in the packet buffer.                                      */
868 /* ------------------------------------------------------------------------ */
869 static INLINE int frpr_pullup(fin, plen)
870 fr_info_t *fin;
871 int plen;
872 {
873 #if defined(_KERNEL)
874 	if (fin->fin_m != NULL) {
875 		int ipoff;
876 
877 		ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
878 
879 		if (fin->fin_dp != NULL)
880 			plen += (char *)fin->fin_dp -
881 				((char *)fin->fin_ip + fin->fin_hlen);
882 		plen += fin->fin_hlen;
883 		/*
884 		 * We don't do 'plen += ipoff;' here. The fr_pullup() will
885 		 * do it for us.
886 		 */
887 		if (M_LEN(fin->fin_m) < plen + ipoff) {
888 			if (fr_pullup(fin->fin_m, fin, plen) == NULL)
889 				return -1;
890 		}
891 	}
892 #endif
893 	return 0;
894 }
895 
896 
897 /* ------------------------------------------------------------------------ */
898 /* Function:    frpr_short                                                  */
899 /* Returns:     void                                                        */
900 /* Parameters:  fin(I)  - pointer to packet information                     */
901 /*              xmin(I) - minimum header size                               */
902 /*                                                                          */
903 /* Check if a packet is "short" as defined by xmin.  The rule we are        */
904 /* applying here is that the packet must not be fragmented within the layer */
905 /* 4 header.  That is, it must not be a fragment that has its offset set to */
906 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the    */
907 /* entire layer 4 header must be present (min).                             */
908 /* ------------------------------------------------------------------------ */
909 static INLINE void frpr_short(fin, xmin)
910 fr_info_t *fin;
911 int xmin;
912 {
913 
914 	if (fin->fin_off == 0) {
915 		if (fin->fin_dlen < xmin)
916 			fin->fin_flx |= FI_SHORT;
917 	} else if (fin->fin_off < xmin) {
918 		fin->fin_flx |= FI_SHORT;
919 	}
920 }
921 
922 
923 /* ------------------------------------------------------------------------ */
924 /* Function:    frpr_icmp                                                   */
925 /* Returns:     void                                                        */
926 /* Parameters:  fin(I) - pointer to packet information                      */
927 /*                                                                          */
928 /* IPv4 Only                                                                */
929 /* Do a sanity check on the packet for ICMP (v4).  In nearly all cases,     */
930 /* except extrememly bad packets, both type and code will be present.       */
931 /* The expected minimum size of an ICMP packet is very much dependent on    */
932 /* the type of it.                                                          */
933 /*                                                                          */
934 /* XXX - other ICMP sanity checks?                                          */
935 /* ------------------------------------------------------------------------ */
936 static INLINE void frpr_icmp(fin)
937 fr_info_t *fin;
938 {
939 	int minicmpsz = sizeof(struct icmp);
940 	icmphdr_t *icmp;
941 	ip_t *oip;
942 	ipf_stack_t *ifs = fin->fin_ifs;
943 
944 	if (fin->fin_off != 0) {
945 		frpr_short(fin, ICMPERR_ICMPHLEN);
946 		return;
947 	}
948 
949 	if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
950 		return;
951 
952 	fr_checkv4sum(fin);
953 
954 	/*
955 	 * This is a right place to set icmp pointer, since the memory
956 	 * referenced by fin_dp could get reallocated. The code down below can
957 	 * rely on fact icmp variable always points to ICMP header.
958 	 */
959 	icmp = fin->fin_dp;
960 	fin->fin_data[0] = *(u_short *)icmp;
961 	fin->fin_data[1] = icmp->icmp_id;
962 
963 	switch (icmp->icmp_type)
964 	{
965 	case ICMP_ECHOREPLY :
966 	case ICMP_ECHO :
967 	/* Router discovery messaes - RFC 1256 */
968 	case ICMP_ROUTERADVERT :
969 	case ICMP_ROUTERSOLICIT :
970 		minicmpsz = ICMP_MINLEN;
971 		break;
972 	/*
973 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
974 	 * 3 * timestamp(3 * 4)
975 	 */
976 	case ICMP_TSTAMP :
977 	case ICMP_TSTAMPREPLY :
978 		minicmpsz = 20;
979 		break;
980 	/*
981 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
982 	 * mask(4)
983 	 */
984 	case ICMP_MASKREQ :
985 	case ICMP_MASKREPLY :
986 			minicmpsz = 12;
987 			break;
988 	/*
989 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
990 	 */
991 	case ICMP_UNREACH :
992 		if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
993 			if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
994 				fin->fin_flx |= FI_BAD;
995 		}
996 		/* FALLTHRU */
997 	case ICMP_SOURCEQUENCH :
998 	case ICMP_REDIRECT :
999 	case ICMP_TIMXCEED :
1000 	case ICMP_PARAMPROB :
1001 		fin->fin_flx |= FI_ICMPERR;
1002 		if (fr_coalesce(fin) != 1)
1003 			return;
1004 		/*
1005 		 * ICMP error packets should not be generated for IP
1006 		 * packets that are a fragment that isn't the first
1007 		 * fragment.
1008 		 */
1009 		oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1010 		if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1011 			fin->fin_flx |= FI_BAD;
1012 		break;
1013 	default :
1014 		break;
1015 	}
1016 
1017 	frpr_short(fin, minicmpsz);
1018 }
1019 
1020 
1021 /* ------------------------------------------------------------------------ */
1022 /* Function:    frpr_tcpcommon                                              */
1023 /* Returns:     void                                                        */
1024 /* Parameters:  fin(I) - pointer to packet information                      */
1025 /*                                                                          */
1026 /* TCP header sanity checking.  Look for bad combinations of TCP flags,     */
1027 /* and make some checks with how they interact with other fields.           */
1028 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is     */
1029 /* valid and mark the packet as bad if not.                                 */
1030 /* ------------------------------------------------------------------------ */
1031 static INLINE void frpr_tcpcommon(fin)
1032 fr_info_t *fin;
1033 {
1034 	int flags, tlen;
1035 	tcphdr_t *tcp;
1036 
1037 	fin->fin_flx |= FI_TCPUDP;
1038 	if (fin->fin_off != 0)
1039 		return;
1040 
1041 	if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1042 		return;
1043 	tcp = fin->fin_dp;
1044 
1045 	if (fin->fin_dlen > 3) {
1046 		fin->fin_sport = ntohs(tcp->th_sport);
1047 		fin->fin_dport = ntohs(tcp->th_dport);
1048 	}
1049 
1050 	if ((fin->fin_flx & FI_SHORT) != 0)
1051 		return;
1052 
1053 	/*
1054 	 * Use of the TCP data offset *must* result in a value that is at
1055 	 * least the same size as the TCP header.
1056 	 */
1057 	tlen = TCP_OFF(tcp) << 2;
1058 	if (tlen < sizeof(tcphdr_t)) {
1059 		fin->fin_flx |= FI_BAD;
1060 		return;
1061 	}
1062 
1063 	flags = tcp->th_flags;
1064 	fin->fin_tcpf = tcp->th_flags;
1065 
1066 	/*
1067 	 * If the urgent flag is set, then the urgent pointer must
1068 	 * also be set and vice versa.  Good TCP packets do not have
1069 	 * just one of these set.
1070 	 */
1071 	if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1072 		fin->fin_flx |= FI_BAD;
1073 	} else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1074 		/* Ignore this case, it shows up in "real" traffic with */
1075 		/* bogus values in the urgent pointer field. */
1076 		flags = flags; /* LINT */
1077 	} else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1078 		   ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1079 		/* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1080 		fin->fin_flx |= FI_BAD;
1081 	} else if (!(flags & TH_ACK)) {
1082 		/*
1083 		 * If the ack bit isn't set, then either the SYN or
1084 		 * RST bit must be set.  If the SYN bit is set, then
1085 		 * we expect the ACK field to be 0.  If the ACK is
1086 		 * not set and if URG, PSH or FIN are set, consdier
1087 		 * that to indicate a bad TCP packet.
1088 		 */
1089 		if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1090 			/*
1091 			 * Cisco PIX sets the ACK field to a random value.
1092 			 * In light of this, do not set FI_BAD until a patch
1093 			 * is available from Cisco to ensure that
1094 			 * interoperability between existing systems is
1095 			 * achieved.
1096 			 */
1097 			/*fin->fin_flx |= FI_BAD*/;
1098 			flags = flags; /* LINT */
1099 		} else if (!(flags & (TH_RST|TH_SYN))) {
1100 			fin->fin_flx |= FI_BAD;
1101 		} else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1102 			fin->fin_flx |= FI_BAD;
1103 		}
1104 	}
1105 
1106 	/*
1107 	 * At this point, it's not exactly clear what is to be gained by
1108 	 * marking up which TCP options are and are not present.  The one we
1109 	 * are most interested in is the TCP window scale.  This is only in
1110 	 * a SYN packet [RFC1323] so we don't need this here...?
1111 	 * Now if we were to analyse the header for passive fingerprinting,
1112 	 * then that might add some weight to adding this...
1113 	 */
1114 	if (tlen == sizeof(tcphdr_t))
1115 		return;
1116 
1117 	if (frpr_pullup(fin, tlen) == -1)
1118 		return;
1119 
1120 #if 0
1121 	ip = fin->fin_ip;
1122 	s = (u_char *)(tcp + 1);
1123 	off = IP_HL(ip) << 2;
1124 # ifdef _KERNEL
1125 	if (fin->fin_mp != NULL) {
1126 		mb_t *m = *fin->fin_mp;
1127 
1128 		if (off + tlen > M_LEN(m))
1129 			return;
1130 	}
1131 # endif
1132 	for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1133 		opt = *s;
1134 		if (opt == '\0')
1135 			break;
1136 		else if (opt == TCPOPT_NOP)
1137 			ol = 1;
1138 		else {
1139 			if (tlen < 2)
1140 				break;
1141 			ol = (int)*(s + 1);
1142 			if (ol < 2 || ol > tlen)
1143 				break;
1144 		}
1145 
1146 		for (i = 9, mv = 4; mv >= 0; ) {
1147 			op = ipopts + i;
1148 			if (opt == (u_char)op->ol_val) {
1149 				optmsk |= op->ol_bit;
1150 				break;
1151 			}
1152 		}
1153 		tlen -= ol;
1154 		s += ol;
1155 	}
1156 #endif /* 0 */
1157 }
1158 
1159 
1160 
1161 /* ------------------------------------------------------------------------ */
1162 /* Function:    frpr_udpcommon                                              */
1163 /* Returns:     void                                                        */
1164 /* Parameters:  fin(I) - pointer to packet information                      */
1165 /*                                                                          */
1166 /* Extract the UDP source and destination ports, if present.  If compiled   */
1167 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid.          */
1168 /* ------------------------------------------------------------------------ */
1169 static INLINE void frpr_udpcommon(fin)
1170 fr_info_t *fin;
1171 {
1172 	udphdr_t *udp;
1173 
1174 	fin->fin_flx |= FI_TCPUDP;
1175 
1176 	if (!fin->fin_off && (fin->fin_dlen > 3)) {
1177 		if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1178 			fin->fin_flx |= FI_SHORT;
1179 			return;
1180 		}
1181 
1182 		udp = fin->fin_dp;
1183 
1184 		fin->fin_sport = ntohs(udp->uh_sport);
1185 		fin->fin_dport = ntohs(udp->uh_dport);
1186 	}
1187 }
1188 
1189 
1190 /* ------------------------------------------------------------------------ */
1191 /* Function:    frpr_tcp                                                    */
1192 /* Returns:     void                                                        */
1193 /* Parameters:  fin(I) - pointer to packet information                      */
1194 /*                                                                          */
1195 /* IPv4 Only                                                                */
1196 /* Analyse the packet for IPv4/TCP properties.                              */
1197 /* ------------------------------------------------------------------------ */
1198 static INLINE void frpr_tcp(fin)
1199 fr_info_t *fin;
1200 {
1201 
1202 	fr_checkv4sum(fin);
1203 
1204 	frpr_short(fin, sizeof(tcphdr_t));
1205 
1206 	frpr_tcpcommon(fin);
1207 }
1208 
1209 
1210 /* ------------------------------------------------------------------------ */
1211 /* Function:    frpr_udp                                                    */
1212 /* Returns:     void                                                        */
1213 /* Parameters:  fin(I) - pointer to packet information                      */
1214 /*                                                                          */
1215 /* IPv4 Only                                                                */
1216 /* Analyse the packet for IPv4/UDP properties.                              */
1217 /* ------------------------------------------------------------------------ */
1218 static INLINE void frpr_udp(fin)
1219 fr_info_t *fin;
1220 {
1221 
1222 	fr_checkv4sum(fin);
1223 
1224 	frpr_short(fin, sizeof(udphdr_t));
1225 
1226 	frpr_udpcommon(fin);
1227 }
1228 
1229 
1230 /* ------------------------------------------------------------------------ */
1231 /* Function:    frpr_esp                                                    */
1232 /* Returns:     void                                                        */
1233 /* Parameters:  fin(I) - pointer to packet information                      */
1234 /*                                                                          */
1235 /* Analyse the packet for ESP properties.                                   */
1236 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
1237 /* even though the newer ESP packets must also have a sequence number that  */
1238 /* is 32bits as well, it is not possible(?) to determine the version from a */
1239 /* simple packet header.                                                    */
1240 /* ------------------------------------------------------------------------ */
1241 static INLINE void frpr_esp(fin)
1242 fr_info_t *fin;
1243 {
1244 	if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1245 		return;
1246 
1247 	frpr_short(fin, 8);
1248 }
1249 
1250 
1251 /* ------------------------------------------------------------------------ */
1252 /* Function:    frpr_ah                                                     */
1253 /* Returns:     void                                                        */
1254 /* Parameters:  fin(I) - pointer to packet information                      */
1255 /*                                                                          */
1256 /* Analyse the packet for AH properties.                                    */
1257 /* The minimum length is taken to be the combination of all fields in the   */
1258 /* header being present and no authentication data (null algorithm used.)   */
1259 /* ------------------------------------------------------------------------ */
1260 static INLINE void frpr_ah(fin)
1261 fr_info_t *fin;
1262 {
1263 	authhdr_t *ah;
1264 	int len;
1265 
1266 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1267 		return;
1268 
1269 	ah = (authhdr_t *)fin->fin_dp;
1270 
1271 	len = (ah->ah_plen + 2) << 2;
1272 	frpr_short(fin, len);
1273 }
1274 
1275 
1276 /* ------------------------------------------------------------------------ */
1277 /* Function:    frpr_gre                                                    */
1278 /* Returns:     void                                                        */
1279 /* Parameters:  fin(I) - pointer to packet information                      */
1280 /*                                                                          */
1281 /* Analyse the packet for GRE properties.                                   */
1282 /* ------------------------------------------------------------------------ */
1283 static INLINE void frpr_gre(fin)
1284 fr_info_t *fin;
1285 {
1286 	grehdr_t *gre;
1287 
1288 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1289 		return;
1290 
1291 	frpr_short(fin, sizeof(grehdr_t));
1292 
1293 	if (fin->fin_off == 0) {
1294 		gre = fin->fin_dp;
1295 		if (GRE_REV(gre->gr_flags) == 1)
1296 			fin->fin_data[0] = gre->gr_call;
1297 	}
1298 }
1299 
1300 
1301 /* ------------------------------------------------------------------------ */
1302 /* Function:    frpr_ipv4hdr                                                */
1303 /* Returns:     void                                                        */
1304 /* Parameters:  fin(I) - pointer to packet information                      */
1305 /*                                                                          */
1306 /* IPv4 Only                                                                */
1307 /* Analyze the IPv4 header and set fields in the fr_info_t structure.       */
1308 /* Check all options present and flag their presence if any exist.          */
1309 /* ------------------------------------------------------------------------ */
1310 static INLINE void frpr_ipv4hdr(fin)
1311 fr_info_t *fin;
1312 {
1313 	u_short optmsk = 0, secmsk = 0, auth = 0;
1314 	int hlen, ol, mv, p, i;
1315 	const struct optlist *op;
1316 	u_char *s, opt;
1317 	u_short off;
1318 	fr_ip_t *fi;
1319 	ip_t *ip;
1320 
1321 	fi = &fin->fin_fi;
1322 	hlen = fin->fin_hlen;
1323 
1324 	ip = fin->fin_ip;
1325 	p = ip->ip_p;
1326 	fi->fi_p = p;
1327 	fi->fi_tos = ip->ip_tos;
1328 	fin->fin_id = ip->ip_id;
1329 	off = ip->ip_off;
1330 
1331 	/* Get both TTL and protocol */
1332 	fi->fi_p = ip->ip_p;
1333 	fi->fi_ttl = ip->ip_ttl;
1334 #if 0
1335 	(*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1336 #endif
1337 
1338 	/* Zero out bits not used in IPv6 address */
1339 	fi->fi_src.i6[1] = 0;
1340 	fi->fi_src.i6[2] = 0;
1341 	fi->fi_src.i6[3] = 0;
1342 	fi->fi_dst.i6[1] = 0;
1343 	fi->fi_dst.i6[2] = 0;
1344 	fi->fi_dst.i6[3] = 0;
1345 
1346 	fi->fi_saddr = ip->ip_src.s_addr;
1347 	fi->fi_daddr = ip->ip_dst.s_addr;
1348 
1349 	/*
1350 	 * set packet attribute flags based on the offset and
1351 	 * calculate the byte offset that it represents.
1352 	 */
1353 	off &= IP_MF|IP_OFFMASK;
1354 	if (off != 0) {
1355 		int morefrag = off & IP_MF;
1356 
1357 		fi->fi_flx |= FI_FRAG;
1358 		if (morefrag)
1359 			fi->fi_flx |= FI_MOREFRAG;
1360 		off &= IP_OFFMASK;
1361 		if (off != 0) {
1362 			fin->fin_flx |= FI_FRAGBODY;
1363 			off <<= 3;
1364 			if ((off + fin->fin_dlen > 65535) ||
1365 			    (fin->fin_dlen == 0) ||
1366 			    ((morefrag != 0) && ((fin->fin_dlen & 7) != 0))) {
1367 				/*
1368 				 * The length of the packet, starting at its
1369 				 * offset cannot exceed 65535 (0xffff) as the
1370 				 * length of an IP packet is only 16 bits.
1371 				 *
1372 				 * Any fragment that isn't the last fragment
1373 				 * must have a length greater than 0 and it
1374 				 * must be an even multiple of 8.
1375 				 */
1376 				fi->fi_flx |= FI_BAD;
1377 			}
1378 		}
1379 	}
1380 	fin->fin_off = off;
1381 
1382 	/*
1383 	 * Call per-protocol setup and checking
1384 	 */
1385 	switch (p)
1386 	{
1387 	case IPPROTO_UDP :
1388 		frpr_udp(fin);
1389 		break;
1390 	case IPPROTO_TCP :
1391 		frpr_tcp(fin);
1392 		break;
1393 	case IPPROTO_ICMP :
1394 		frpr_icmp(fin);
1395 		break;
1396 	case IPPROTO_AH :
1397 		frpr_ah(fin);
1398 		break;
1399 	case IPPROTO_ESP :
1400 		frpr_esp(fin);
1401 		break;
1402 	case IPPROTO_GRE :
1403 		frpr_gre(fin);
1404 		break;
1405 	}
1406 
1407 	ip = fin->fin_ip;
1408 	if (ip == NULL)
1409 		return;
1410 
1411 	/*
1412 	 * If it is a standard IP header (no options), set the flag fields
1413 	 * which relate to options to 0.
1414 	 */
1415 	if (hlen == sizeof(*ip)) {
1416 		fi->fi_optmsk = 0;
1417 		fi->fi_secmsk = 0;
1418 		fi->fi_auth = 0;
1419 		return;
1420 	}
1421 
1422 	/*
1423 	 * So the IP header has some IP options attached.  Walk the entire
1424 	 * list of options present with this packet and set flags to indicate
1425 	 * which ones are here and which ones are not.  For the somewhat out
1426 	 * of date and obscure security classification options, set a flag to
1427 	 * represent which classification is present.
1428 	 */
1429 	fi->fi_flx |= FI_OPTIONS;
1430 
1431 	for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1432 		opt = *s;
1433 		if (opt == '\0')
1434 			break;
1435 		else if (opt == IPOPT_NOP)
1436 			ol = 1;
1437 		else {
1438 			if (hlen < 2)
1439 				break;
1440 			ol = (int)*(s + 1);
1441 			if (ol < 2 || ol > hlen)
1442 				break;
1443 		}
1444 		for (i = 9, mv = 4; mv >= 0; ) {
1445 			op = ipopts + i;
1446 			if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1447 				optmsk |= op->ol_bit;
1448 				if (opt == IPOPT_SECURITY) {
1449 					const struct optlist *sp;
1450 					u_char	sec;
1451 					int j, m;
1452 
1453 					sec = *(s + 2);	/* classification */
1454 					for (j = 3, m = 2; m >= 0; ) {
1455 						sp = secopt + j;
1456 						if (sec == sp->ol_val) {
1457 							secmsk |= sp->ol_bit;
1458 							auth = *(s + 3);
1459 							auth *= 256;
1460 							auth += *(s + 4);
1461 							break;
1462 						}
1463 						if (sec < sp->ol_val)
1464 							j -= m;
1465 						else
1466 							j += m;
1467 						m--;
1468 					}
1469 				}
1470 				break;
1471 			}
1472 			if (opt < op->ol_val)
1473 				i -= mv;
1474 			else
1475 				i += mv;
1476 			mv--;
1477 		}
1478 		hlen -= ol;
1479 		s += ol;
1480 	}
1481 
1482 	/*
1483 	 *
1484 	 */
1485 	if (auth && !(auth & 0x0100))
1486 		auth &= 0xff00;
1487 	fi->fi_optmsk = optmsk;
1488 	fi->fi_secmsk = secmsk;
1489 	fi->fi_auth = auth;
1490 }
1491 
1492 
1493 /* ------------------------------------------------------------------------ */
1494 /* Function:    fr_makefrip                                                 */
1495 /* Returns:     int - 1 == hdr checking error, 0 == OK                      */
1496 /* Parameters:  hlen(I) - length of IP packet header                        */
1497 /*              ip(I)   - pointer to the IP header                          */
1498 /*              fin(IO) - pointer to packet information                     */
1499 /*                                                                          */
1500 /* Compact the IP header into a structure which contains just the info.     */
1501 /* which is useful for comparing IP headers with and store this information */
1502 /* in the fr_info_t structure pointer to by fin.  At present, it is assumed */
1503 /* this function will be called with either an IPv4 or IPv6 packet.         */
1504 /* ------------------------------------------------------------------------ */
1505 int	fr_makefrip(hlen, ip, fin)
1506 int hlen;
1507 ip_t *ip;
1508 fr_info_t *fin;
1509 {
1510 	int v;
1511 
1512 	fin->fin_depth = 0;
1513 	fin->fin_hlen = (u_short)hlen;
1514 	fin->fin_ip = ip;
1515 	fin->fin_rule = 0xffffffff;
1516 	fin->fin_group[0] = -1;
1517 	fin->fin_group[1] = '\0';
1518 	fin->fin_dlen = fin->fin_plen - hlen;
1519 	fin->fin_dp = (char *)ip + hlen;
1520 
1521 	v = fin->fin_v;
1522 	if (v == 4)
1523 		frpr_ipv4hdr(fin);
1524 #ifdef	USE_INET6
1525 	else if (v == 6)
1526 		frpr_ipv6hdr(fin);
1527 #endif
1528 	if (fin->fin_ip == NULL)
1529 		return -1;
1530 	return 0;
1531 }
1532 
1533 
1534 /* ------------------------------------------------------------------------ */
1535 /* Function:    fr_portcheck                                                */
1536 /* Returns:     int - 1 == port matched, 0 == port match failed             */
1537 /* Parameters:  frp(I) - pointer to port check `expression'                 */
1538 /*              pop(I) - pointer to port number to evaluate                 */
1539 /*                                                                          */
1540 /* Perform a comparison of a port number against some other(s), using a     */
1541 /* structure with compare information stored in it.                         */
1542 /* ------------------------------------------------------------------------ */
1543 static INLINE int fr_portcheck(frp, pop)
1544 frpcmp_t *frp;
1545 u_short *pop;
1546 {
1547 	u_short tup, po;
1548 	int err = 1;
1549 
1550 	tup = *pop;
1551 	po = frp->frp_port;
1552 
1553 	/*
1554 	 * Do opposite test to that required and continue if that succeeds.
1555 	 */
1556 	switch (frp->frp_cmp)
1557 	{
1558 	case FR_EQUAL :
1559 		if (tup != po) /* EQUAL */
1560 			err = 0;
1561 		break;
1562 	case FR_NEQUAL :
1563 		if (tup == po) /* NOTEQUAL */
1564 			err = 0;
1565 		break;
1566 	case FR_LESST :
1567 		if (tup >= po) /* LESSTHAN */
1568 			err = 0;
1569 		break;
1570 	case FR_GREATERT :
1571 		if (tup <= po) /* GREATERTHAN */
1572 			err = 0;
1573 		break;
1574 	case FR_LESSTE :
1575 		if (tup > po) /* LT or EQ */
1576 			err = 0;
1577 		break;
1578 	case FR_GREATERTE :
1579 		if (tup < po) /* GT or EQ */
1580 			err = 0;
1581 		break;
1582 	case FR_OUTRANGE :
1583 		if (tup >= po && tup <= frp->frp_top) /* Out of range */
1584 			err = 0;
1585 		break;
1586 	case FR_INRANGE :
1587 		if (tup <= po || tup >= frp->frp_top) /* In range */
1588 			err = 0;
1589 		break;
1590 	case FR_INCRANGE :
1591 		if (tup < po || tup > frp->frp_top) /* Inclusive range */
1592 			err = 0;
1593 		break;
1594 	default :
1595 		break;
1596 	}
1597 	return err;
1598 }
1599 
1600 
1601 /* ------------------------------------------------------------------------ */
1602 /* Function:    fr_tcpudpchk                                                */
1603 /* Returns:     int - 1 == protocol matched, 0 == check failed              */
1604 /* Parameters:  fin(I) - pointer to packet information                      */
1605 /*              ft(I)  - pointer to structure with comparison data          */
1606 /*                                                                          */
1607 /* Compares the current pcket (assuming it is TCP/UDP) information with a   */
1608 /* structure containing information that we want to match against.          */
1609 /* ------------------------------------------------------------------------ */
1610 int fr_tcpudpchk(fin, ft)
1611 fr_info_t *fin;
1612 frtuc_t *ft;
1613 {
1614 	int err = 1;
1615 
1616 	/*
1617 	 * Both ports should *always* be in the first fragment.
1618 	 * So far, I cannot find any cases where they can not be.
1619 	 *
1620 	 * compare destination ports
1621 	 */
1622 	if (ft->ftu_dcmp)
1623 		err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1624 
1625 	/*
1626 	 * compare source ports
1627 	 */
1628 	if (err && ft->ftu_scmp)
1629 		err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1630 
1631 	/*
1632 	 * If we don't have all the TCP/UDP header, then how can we
1633 	 * expect to do any sort of match on it ?  If we were looking for
1634 	 * TCP flags, then NO match.  If not, then match (which should
1635 	 * satisfy the "short" class too).
1636 	 */
1637 	if (err && (fin->fin_p == IPPROTO_TCP)) {
1638 		if (fin->fin_flx & FI_SHORT)
1639 			return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1640 		/*
1641 		 * Match the flags ?  If not, abort this match.
1642 		 */
1643 		if (ft->ftu_tcpfm &&
1644 		    ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1645 			FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1646 				 ft->ftu_tcpfm, ft->ftu_tcpf));
1647 			err = 0;
1648 		}
1649 	}
1650 	return err;
1651 }
1652 
1653 
1654 /* ------------------------------------------------------------------------ */
1655 /* Function:    fr_ipfcheck                                                 */
1656 /* Returns:     int - 0 == match, 1 == no match                             */
1657 /* Parameters:  fin(I)     - pointer to packet information                  */
1658 /*              fr(I)      - pointer to filter rule                         */
1659 /*              portcmp(I) - flag indicating whether to attempt matching on */
1660 /*                           TCP/UDP port data.                             */
1661 /*                                                                          */
1662 /* Check to see if a packet matches an IPFilter rule.  Checks of addresses, */
1663 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1664 /* this function.                                                           */
1665 /* ------------------------------------------------------------------------ */
1666 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1667 fr_info_t *fin;
1668 frentry_t *fr;
1669 int portcmp;
1670 {
1671 	u_32_t	*ld, *lm, *lip;
1672 	fripf_t *fri;
1673 	fr_ip_t *fi;
1674 	int i;
1675 	ipf_stack_t *ifs = fin->fin_ifs;
1676 
1677 	fi = &fin->fin_fi;
1678 	fri = fr->fr_ipf;
1679 	lip = (u_32_t *)fi;
1680 	lm = (u_32_t *)&fri->fri_mip;
1681 	ld = (u_32_t *)&fri->fri_ip;
1682 
1683 	/*
1684 	 * first 32 bits to check coversion:
1685 	 * IP version, TOS, TTL, protocol
1686 	 */
1687 	i = ((*lip & *lm) != *ld);
1688 	FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1689 		   *lip, *lm, *ld));
1690 	if (i)
1691 		return 1;
1692 
1693 	/*
1694 	 * Next 32 bits is a constructed bitmask indicating which IP options
1695 	 * are present (if any) in this packet.
1696 	 */
1697 	lip++, lm++, ld++;
1698 	i |= ((*lip & *lm) != *ld);
1699 	FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1700 		   *lip, *lm, *ld));
1701 	if (i)
1702 		return 1;
1703 
1704 	lip++, lm++, ld++;
1705 	/*
1706 	 * Unrolled loops (4 each, for 32 bits) for address checks.
1707 	 */
1708 	/*
1709 	 * Check the source address.
1710 	 */
1711 #ifdef	IPFILTER_LOOKUP
1712 	if (fr->fr_satype == FRI_LOOKUP) {
1713 		i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, ifs);
1714 		if (i == -1)
1715 			return 1;
1716 		lip += 3;
1717 		lm += 3;
1718 		ld += 3;
1719 	} else {
1720 #endif
1721 		i = ((*lip & *lm) != *ld);
1722 		FR_DEBUG(("2a. %#08x & %#08x != %#08x\n",
1723 			   *lip, *lm, *ld));
1724 		if (fi->fi_v == 6) {
1725 			lip++, lm++, ld++;
1726 			i |= ((*lip & *lm) != *ld);
1727 			FR_DEBUG(("2b. %#08x & %#08x != %#08x\n",
1728 				   *lip, *lm, *ld));
1729 			lip++, lm++, ld++;
1730 			i |= ((*lip & *lm) != *ld);
1731 			FR_DEBUG(("2c. %#08x & %#08x != %#08x\n",
1732 				   *lip, *lm, *ld));
1733 			lip++, lm++, ld++;
1734 			i |= ((*lip & *lm) != *ld);
1735 			FR_DEBUG(("2d. %#08x & %#08x != %#08x\n",
1736 				   *lip, *lm, *ld));
1737 		} else {
1738 			lip += 3;
1739 			lm += 3;
1740 			ld += 3;
1741 		}
1742 #ifdef	IPFILTER_LOOKUP
1743 	}
1744 #endif
1745 	i ^= (fr->fr_flags & FR_NOTSRCIP) >> 6;
1746 	if (i)
1747 		return 1;
1748 
1749 	/*
1750 	 * Check the destination address.
1751 	 */
1752 	lip++, lm++, ld++;
1753 #ifdef	IPFILTER_LOOKUP
1754 	if (fr->fr_datype == FRI_LOOKUP) {
1755 		i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, ifs);
1756 		if (i == -1)
1757 			return 1;
1758 		lip += 3;
1759 		lm += 3;
1760 		ld += 3;
1761 	} else {
1762 #endif
1763 		i = ((*lip & *lm) != *ld);
1764 		FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1765 			   *lip, *lm, *ld));
1766 		if (fi->fi_v == 6) {
1767 			lip++, lm++, ld++;
1768 			i |= ((*lip & *lm) != *ld);
1769 			FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1770 				   *lip, *lm, *ld));
1771 			lip++, lm++, ld++;
1772 			i |= ((*lip & *lm) != *ld);
1773 			FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1774 				   *lip, *lm, *ld));
1775 			lip++, lm++, ld++;
1776 			i |= ((*lip & *lm) != *ld);
1777 			FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1778 				   *lip, *lm, *ld));
1779 		} else {
1780 			lip += 3;
1781 			lm += 3;
1782 			ld += 3;
1783 		}
1784 #ifdef	IPFILTER_LOOKUP
1785 	}
1786 #endif
1787 	i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1788 	if (i)
1789 		return 1;
1790 	/*
1791 	 * IP addresses matched.  The next 32bits contains:
1792 	 * mast of old IP header security & authentication bits.
1793 	 */
1794 	lip++, lm++, ld++;
1795 	i |= ((*lip & *lm) != *ld);
1796 	FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1797 		   *lip, *lm, *ld));
1798 
1799 	/*
1800 	 * Next we have 32 bits of packet flags.
1801 	 */
1802 	lip++, lm++, ld++;
1803 	i |= ((*lip & *lm) != *ld);
1804 	FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1805 		   *lip, *lm, *ld));
1806 
1807 	if (i == 0) {
1808 		/*
1809 		 * If a fragment, then only the first has what we're
1810 		 * looking for here...
1811 		 */
1812 		if (portcmp) {
1813 			if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1814 				i = 1;
1815 		} else {
1816 			if (fr->fr_dcmp || fr->fr_scmp ||
1817 			    fr->fr_tcpf || fr->fr_tcpfm)
1818 				i = 1;
1819 			if (fr->fr_icmpm || fr->fr_icmp) {
1820 				if (((fi->fi_p != IPPROTO_ICMP) &&
1821 				     (fi->fi_p != IPPROTO_ICMPV6)) ||
1822 				    fin->fin_off || (fin->fin_dlen < 2))
1823 					i = 1;
1824 				else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1825 					 fr->fr_icmp) {
1826 					FR_DEBUG(("i. %#x & %#x != %#x\n",
1827 						 fin->fin_data[0],
1828 						 fr->fr_icmpm, fr->fr_icmp));
1829 					i = 1;
1830 				}
1831 			}
1832 		}
1833 	}
1834 	return i;
1835 }
1836 
1837 
1838 /* ------------------------------------------------------------------------ */
1839 /* Function:    fr_scanlist                                                 */
1840 /* Returns:     int - result flags of scanning filter list                  */
1841 /* Parameters:  fin(I) - pointer to packet information                      */
1842 /*              pass(I) - default result to return for filtering            */
1843 /*                                                                          */
1844 /* Check the input/output list of rules for a match to the current packet.  */
1845 /* If a match is found, the value of fr_flags from the rule becomes the     */
1846 /* return value and fin->fin_fr points to the matched rule.                 */
1847 /*                                                                          */
1848 /* This function may be called recusively upto 16 times (limit inbuilt.)    */
1849 /* When unwinding, it should finish up with fin_depth as 0.                 */
1850 /*                                                                          */
1851 /* Could be per interface, but this gets real nasty when you don't have,    */
1852 /* or can't easily change, the kernel source code to .                      */
1853 /* ------------------------------------------------------------------------ */
1854 int fr_scanlist(fin, pass)
1855 fr_info_t *fin;
1856 u_32_t pass;
1857 {
1858 	int rulen, portcmp, off, logged, skip;
1859 	struct frentry *fr, *fnext;
1860 	u_32_t passt, passo;
1861 	ipf_stack_t *ifs = fin->fin_ifs;
1862 
1863 	/*
1864 	 * Do not allow nesting deeper than 16 levels.
1865 	 */
1866 	if (fin->fin_depth >= 16)
1867 		return pass;
1868 
1869 	fr = fin->fin_fr;
1870 
1871 	/*
1872 	 * If there are no rules in this list, return now.
1873 	 */
1874 	if (fr == NULL)
1875 		return pass;
1876 
1877 	skip = 0;
1878 	logged = 0;
1879 	portcmp = 0;
1880 	fin->fin_depth++;
1881 	fin->fin_fr = NULL;
1882 	off = fin->fin_off;
1883 
1884 	if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1885 		portcmp = 1;
1886 
1887 	for (rulen = 0; fr; fr = fnext, rulen++) {
1888 		fnext = fr->fr_next;
1889 		if (skip != 0) {
1890 			FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1891 			skip--;
1892 			continue;
1893 		}
1894 
1895 		/*
1896 		 * In all checks below, a null (zero) value in the
1897 		 * filter struture is taken to mean a wildcard.
1898 		 *
1899 		 * check that we are working for the right interface
1900 		 */
1901 #ifdef	_KERNEL
1902 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1903 			continue;
1904 #else
1905 		if (opts & (OPT_VERBOSE|OPT_DEBUG))
1906 			printf("\n");
1907 		FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1908 				  FR_ISPASS(pass) ? 'p' :
1909 				  FR_ISACCOUNT(pass) ? 'A' :
1910 				  FR_ISAUTH(pass) ? 'a' :
1911 				  (pass & FR_NOMATCH) ? 'n' :'b'));
1912 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1913 			continue;
1914 		FR_VERBOSE((":i"));
1915 #endif
1916 
1917 		switch (fr->fr_type)
1918 		{
1919 		case FR_T_IPF :
1920 		case FR_T_IPF|FR_T_BUILTIN :
1921 			if (fr_ipfcheck(fin, fr, portcmp))
1922 				continue;
1923 			break;
1924 #if defined(IPFILTER_BPF)
1925 		case FR_T_BPFOPC :
1926 		case FR_T_BPFOPC|FR_T_BUILTIN :
1927 		    {
1928 			u_char *mc;
1929 
1930 			if (*fin->fin_mp == NULL)
1931 				continue;
1932 			if (fin->fin_v != fr->fr_v)
1933 				continue;
1934 			mc = (u_char *)fin->fin_m;
1935 			if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1936 				continue;
1937 			break;
1938 		    }
1939 #endif
1940 		case FR_T_CALLFUNC|FR_T_BUILTIN :
1941 		    {
1942 			frentry_t *f;
1943 
1944 			f = (*fr->fr_func)(fin, &pass);
1945 			if (f != NULL)
1946 				fr = f;
1947 			else
1948 				continue;
1949 			break;
1950 		    }
1951 		default :
1952 			break;
1953 		}
1954 
1955 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1956 			if (fin->fin_nattag == NULL)
1957 				continue;
1958 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1959 				continue;
1960 		}
1961 		FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1962 
1963 		passt = fr->fr_flags;
1964 
1965 		/*
1966 		 * Allowing a rule with the "keep state" flag set to match
1967 		 * packets that have been tagged "out of window" by the TCP
1968 		 * state tracking is foolish as the attempt to add a new
1969 		 * state entry to the table will fail.
1970 		 */
1971 		if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1972 			continue;
1973 
1974 		/*
1975 		 * If the rule is a "call now" rule, then call the function
1976 		 * in the rule, if it exists and use the results from that.
1977 		 * If the function pointer is bad, just make like we ignore
1978 		 * it, except for increasing the hit counter.
1979 		 */
1980 		if ((passt & FR_CALLNOW) != 0) {
1981 			IPF_BUMP(fr->fr_hits);
1982 			if ((fr->fr_func != NULL) &&
1983 			    (fr->fr_func != (ipfunc_t)-1)) {
1984 				frentry_t *frs;
1985 
1986 				frs = fin->fin_fr;
1987 				fin->fin_fr = fr;
1988 				fr = (*fr->fr_func)(fin, &passt);
1989 				if (fr == NULL) {
1990 					fin->fin_fr = frs;
1991 					continue;
1992 				}
1993 				passt = fr->fr_flags;
1994 				fin->fin_fr = fr;
1995 			}
1996 		} else {
1997 			fin->fin_fr = fr;
1998 		}
1999 
2000 #ifdef  IPFILTER_LOG
2001 		/*
2002 		 * Just log this packet...
2003 		 */
2004 		if ((passt & FR_LOGMASK) == FR_LOG) {
2005 			if (ipflog(fin, passt) == -1) {
2006 				if (passt & FR_LOGORBLOCK) {
2007 					passt &= ~FR_CMDMASK;
2008 					passt |= FR_BLOCK|FR_QUICK;
2009 				}
2010 				IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2011 			}
2012 			IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2013 			logged = 1;
2014 		}
2015 #endif /* IPFILTER_LOG */
2016 		fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2017 		passo = pass;
2018 		if (FR_ISSKIP(passt))
2019 			skip = fr->fr_arg;
2020 		else if ((passt & FR_LOGMASK) != FR_LOG)
2021 			pass = passt;
2022 		if (passt & (FR_RETICMP|FR_FAKEICMP))
2023 			fin->fin_icode = fr->fr_icode;
2024 		FR_DEBUG(("pass %#x\n", pass));
2025 		IPF_BUMP(fr->fr_hits);
2026 		fin->fin_rule = rulen;
2027 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2028 		if (fr->fr_grp != NULL) {
2029 			fin->fin_fr = *fr->fr_grp;
2030 			pass = fr_scanlist(fin, pass);
2031 			if (fin->fin_fr == NULL) {
2032 				fin->fin_rule = rulen;
2033 				(void) strncpy(fin->fin_group, fr->fr_group,
2034 					       FR_GROUPLEN);
2035 				fin->fin_fr = fr;
2036 			}
2037 			if (fin->fin_flx & FI_DONTCACHE)
2038 				logged = 1;
2039 		}
2040 
2041 		if (pass & FR_QUICK) {
2042 			/*
2043 			 * Finally, if we've asked to track state for this
2044 			 * packet, set it up.  Add state for "quick" rules
2045 			 * here so that if the action fails we can consider
2046 			 * the rule to "not match" and keep on processing
2047 			 * filter rules.
2048 			 */
2049 			if ((pass & FR_KEEPSTATE) &&
2050 			    !(fin->fin_flx & FI_STATE)) {
2051 				int out = fin->fin_out;
2052 
2053 				if (fr_addstate(fin, NULL, 0) != NULL) {
2054 					IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2055 				} else {
2056 					IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2057 					pass = passo;
2058 					continue;
2059 				}
2060 			}
2061 			break;
2062 		}
2063 	}
2064 	if (logged)
2065 		fin->fin_flx |= FI_DONTCACHE;
2066 	fin->fin_depth--;
2067 	return pass;
2068 }
2069 
2070 
2071 /* ------------------------------------------------------------------------ */
2072 /* Function:    fr_acctpkt                                                  */
2073 /* Returns:     frentry_t* - always returns NULL                            */
2074 /* Parameters:  fin(I) - pointer to packet information                      */
2075 /*              passp(IO) - pointer to current/new filter decision (unused) */
2076 /*                                                                          */
2077 /* Checks a packet against accounting rules, if there are any for the given */
2078 /* IP protocol version.                                                     */
2079 /*                                                                          */
2080 /* N.B.: this function returns NULL to match the prototype used by other    */
2081 /* functions called from the IPFilter "mainline" in fr_check().             */
2082 /* ------------------------------------------------------------------------ */
2083 frentry_t *fr_acctpkt(fin, passp)
2084 fr_info_t *fin;
2085 u_32_t *passp;
2086 {
2087 	char group[FR_GROUPLEN];
2088 	frentry_t *fr, *frsave;
2089 	u_32_t pass, rulen;
2090 	ipf_stack_t *ifs = fin->fin_ifs;
2091 
2092 	passp = passp;
2093 #ifdef	USE_INET6
2094 	if (fin->fin_v == 6)
2095 		fr = ifs->ifs_ipacct6[fin->fin_out][ifs->ifs_fr_active];
2096 	else
2097 #endif
2098 		fr = ifs->ifs_ipacct[fin->fin_out][ifs->ifs_fr_active];
2099 
2100 	if (fr != NULL) {
2101 		frsave = fin->fin_fr;
2102 		bcopy(fin->fin_group, group, FR_GROUPLEN);
2103 		rulen = fin->fin_rule;
2104 		fin->fin_fr = fr;
2105 		pass = fr_scanlist(fin, FR_NOMATCH);
2106 		if (FR_ISACCOUNT(pass)) {
2107 			IPF_BUMP(ifs->ifs_frstats[0].fr_acct);
2108 		}
2109 		fin->fin_fr = frsave;
2110 		bcopy(group, fin->fin_group, FR_GROUPLEN);
2111 		fin->fin_rule = rulen;
2112 	}
2113 	return NULL;
2114 }
2115 
2116 
2117 /* ------------------------------------------------------------------------ */
2118 /* Function:    fr_firewall                                                 */
2119 /* Returns:     frentry_t* - returns pointer to matched rule, if no matches */
2120 /*                           were found, returns NULL.                      */
2121 /* Parameters:  fin(I) - pointer to packet information                      */
2122 /*              passp(IO) - pointer to current/new filter decision (unused) */
2123 /*                                                                          */
2124 /* Applies an appropriate set of firewall rules to the packet, to see if    */
2125 /* there are any matches.  The first check is to see if a match can be seen */
2126 /* in the cache.  If not, then search an appropriate list of rules.  Once a */
2127 /* matching rule is found, take any appropriate actions as defined by the   */
2128 /* rule - except logging.                                                   */
2129 /* ------------------------------------------------------------------------ */
2130 static frentry_t *fr_firewall(fin, passp)
2131 fr_info_t *fin;
2132 u_32_t *passp;
2133 {
2134 	frentry_t *fr;
2135 	fr_info_t *fc;
2136 	u_32_t pass;
2137 	int out;
2138 	ipf_stack_t *ifs = fin->fin_ifs;
2139 
2140 	out = fin->fin_out;
2141 	pass = *passp;
2142 
2143 #ifdef	USE_INET6
2144 	if (fin->fin_v == 6)
2145 		fin->fin_fr = ifs->ifs_ipfilter6[out][ifs->ifs_fr_active];
2146 	else
2147 #endif
2148 		fin->fin_fr = ifs->ifs_ipfilter[out][ifs->ifs_fr_active];
2149 
2150 	/*
2151 	 * If there are no rules loaded skip all checks and return.
2152 	 */
2153 	if (fin->fin_fr == NULL) {
2154 
2155 		if ((pass & FR_NOMATCH)) {
2156 			IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2157 		}
2158 
2159 		return (NULL);
2160 	}
2161 
2162 	fc = &ifs->ifs_frcache[out][CACHE_HASH(fin)];
2163 	READ_ENTER(&ifs->ifs_ipf_frcache);
2164 	if (!bcmp((char *)fin, (char *)fc, FI_CSIZE)) {
2165 		/*
2166 		 * copy cached data so we can unlock the mutexes earlier.
2167 		 */
2168 		bcopy((char *)fc, (char *)fin, FI_COPYSIZE);
2169 		RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2170 		IPF_BUMP(ifs->ifs_frstats[out].fr_chit);
2171 
2172 		if ((fr = fin->fin_fr) != NULL) {
2173 			IPF_BUMP(fr->fr_hits);
2174 			pass = fr->fr_flags;
2175 		}
2176 	} else {
2177 		RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2178 
2179 		pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2180 
2181 		if (((pass & FR_KEEPSTATE) == 0) &&
2182 		    ((fin->fin_flx & FI_DONTCACHE) == 0)) {
2183 			WRITE_ENTER(&ifs->ifs_ipf_frcache);
2184 			bcopy((char *)fin, (char *)fc, FI_COPYSIZE);
2185 			RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2186 		}
2187 
2188 		fr = fin->fin_fr;
2189 	}
2190 
2191 	if ((pass & FR_NOMATCH)) {
2192 		IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2193 	}
2194 
2195 	/*
2196 	 * Apply packets per second rate-limiting to a rule as required.
2197 	 */
2198 	if ((fr != NULL) && (fr->fr_pps != 0) &&
2199 	    !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2200 		pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2201 		pass |= FR_BLOCK;
2202 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2203 	}
2204 
2205 	/*
2206 	 * If we fail to add a packet to the authorization queue, then we
2207 	 * drop the packet later.  However, if it was added then pretend
2208 	 * we've dropped it already.
2209 	 */
2210 	if (FR_ISAUTH(pass)) {
2211 		if (fr_newauth(fin->fin_m, fin) != 0) {
2212 #ifdef	_KERNEL
2213 			fin->fin_m = *fin->fin_mp = NULL;
2214 #else
2215 			;
2216 #endif
2217 			fin->fin_error = 0;
2218 		} else
2219 			fin->fin_error = ENOSPC;
2220 	}
2221 
2222 	if ((fr != NULL) && (fr->fr_func != NULL) &&
2223 	    (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2224 		(void) (*fr->fr_func)(fin, &pass);
2225 
2226 	/*
2227 	 * If a rule is a pre-auth rule, check again in the list of rules
2228 	 * loaded for authenticated use.  It does not particulary matter
2229 	 * if this search fails because a "preauth" result, from a rule,
2230 	 * is treated as "not a pass", hence the packet is blocked.
2231 	 */
2232 	if (FR_ISPREAUTH(pass)) {
2233 		if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2234 			pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2235 	}
2236 
2237 	/*
2238 	 * If the rule has "keep frag" and the packet is actually a fragment,
2239 	 * then create a fragment state entry.
2240 	 */
2241 	if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2242 		if (fin->fin_flx & FI_FRAG) {
2243 			if (fr_newfrag(fin, pass) == -1) {
2244 				IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2245 			} else {
2246 				IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2247 			}
2248 		} else {
2249 			IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2250 		}
2251 	}
2252 
2253 	/*
2254 	 * Finally, if we've asked to track state for this packet, set it up.
2255 	 */
2256 	if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2257 		if (fr_addstate(fin, NULL, 0) != NULL) {
2258 			IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2259 		} else {
2260 			IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2261 			if (FR_ISPASS(pass)) {
2262 				pass &= ~FR_CMDMASK;
2263 				pass |= FR_BLOCK;
2264 			}
2265 		}
2266 	}
2267 
2268 	fr = fin->fin_fr;
2269 
2270 	if (passp != NULL)
2271 		*passp = pass;
2272 
2273 	return fr;
2274 }
2275 
2276 /* ------------------------------------------------------------------------ */
2277 /* Function:    fr_check                                                    */
2278 /* Returns:     int -  0 == packet allowed through,                         */
2279 /*              User space:                                                 */
2280 /*                    -1 == packet blocked                                  */
2281 /*                     1 == packet not matched                              */
2282 /*                    -2 == requires authentication                         */
2283 /*              Kernel:                                                     */
2284 /*                   > 0 == filter error # for packet                       */
2285 /* Parameters: ip(I)   - pointer to start of IPv4/6 packet                  */
2286 /*             hlen(I) - length of header                                   */
2287 /*             ifp(I)  - pointer to interface this packet is on             */
2288 /*             out(I)  - 0 == packet going in, 1 == packet going out        */
2289 /*             mp(IO)  - pointer to caller's buffer pointer that holds this */
2290 /*                       IP packet.                                         */
2291 /* Solaris & HP-UX ONLY :                                                   */
2292 /*             qpi(I)  - pointer to STREAMS queue information for this      */
2293 /*                       interface & direction.                             */
2294 /*                                                                          */
2295 /* fr_check() is the master function for all IPFilter packet processing.    */
2296 /* It orchestrates: Network Address Translation (NAT), checking for packet  */
2297 /* authorisation (or pre-authorisation), presence of related state info.,   */
2298 /* generating log entries, IP packet accounting, routing of packets as      */
2299 /* directed by firewall rules and of course whether or not to allow the     */
2300 /* packet to be further processed by the kernel.                            */
2301 /*                                                                          */
2302 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer  */
2303 /* freed.  Packets passed may be returned with the pointer pointed to by    */
2304 /* by "mp" changed to a new buffer.                                         */
2305 /* ------------------------------------------------------------------------ */
2306 int fr_check(ip, hlen, ifp, out
2307 #if defined(_KERNEL) && defined(MENTAT)
2308 , qif, mp, ifs)
2309 void *qif;
2310 #else
2311 , mp, ifs)
2312 #endif
2313 mb_t **mp;
2314 ip_t *ip;
2315 int hlen;
2316 void *ifp;
2317 int out;
2318 ipf_stack_t *ifs;
2319 {
2320 	/*
2321 	 * The above really sucks, but short of writing a diff
2322 	 */
2323 	fr_info_t frinfo;
2324 	fr_info_t *fin = &frinfo;
2325 	u_32_t pass;
2326 	frentry_t *fr = NULL;
2327 	int v = IP_V(ip);
2328 	mb_t *mc = NULL;
2329 	mb_t *m;
2330 #ifdef USE_INET6
2331 	ip6_t *ip6;
2332 #endif
2333 #ifdef	_KERNEL
2334 # ifdef MENTAT
2335 	qpktinfo_t *qpi = qif;
2336 #endif
2337 #endif
2338 
2339 	SPL_INT(s);
2340 	pass = ifs->ifs_fr_pass;
2341 
2342 	/*
2343 	 * The first part of fr_check() deals with making sure that what goes
2344 	 * into the filtering engine makes some sense.  Information about the
2345 	 * the packet is distilled, collected into a fr_info_t structure and
2346 	 * the an attempt to ensure the buffer the packet is in is big enough
2347 	 * to hold all the required packet headers.
2348 	 */
2349 #ifdef	_KERNEL
2350 # ifdef MENTAT
2351 	if (!OK_32PTR(ip))
2352 		return 2;
2353 # endif
2354 
2355 
2356 	if (ifs->ifs_fr_running <= 0) {
2357 		return 0;
2358 	}
2359 
2360 	bzero((char *)fin, sizeof(*fin));
2361 
2362 # ifdef MENTAT
2363 	fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2364 					 FI_BROADCAST);
2365 	m = qpi->qpi_m;
2366 	fin->fin_qfm = m;
2367 	fin->fin_qpi = qpi;
2368 # else /* MENTAT */
2369 
2370 	m = *mp;
2371 
2372 #  if defined(M_MCAST)
2373 	if ((m->m_flags & M_MCAST) != 0)
2374 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2375 #  endif
2376 #  if defined(M_MLOOP)
2377 	if ((m->m_flags & M_MLOOP) != 0)
2378 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2379 #  endif
2380 #  if defined(M_BCAST)
2381 	if ((m->m_flags & M_BCAST) != 0)
2382 		fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2383 #  endif
2384 #  ifdef M_CANFASTFWD
2385 	/*
2386 	 * XXX For now, IP Filter and fast-forwarding of cached flows
2387 	 * XXX are mutually exclusive.  Eventually, IP Filter should
2388 	 * XXX get a "can-fast-forward" filter rule.
2389 	 */
2390 	m->m_flags &= ~M_CANFASTFWD;
2391 #  endif /* M_CANFASTFWD */
2392 #  ifdef CSUM_DELAY_DATA
2393 	/*
2394 	 * disable delayed checksums.
2395 	 */
2396 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2397 		in_delayed_cksum(m);
2398 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2399 	}
2400 #  endif /* CSUM_DELAY_DATA */
2401 # endif /* MENTAT */
2402 #else
2403 
2404 	bzero((char *)fin, sizeof(*fin));
2405 	m = *mp;
2406 #endif /* _KERNEL */
2407 
2408 	fin->fin_v = v;
2409 	fin->fin_m = m;
2410 	fin->fin_ip = ip;
2411 	fin->fin_mp = mp;
2412 	fin->fin_out = out;
2413 	fin->fin_ifp = ifp;
2414 	fin->fin_error = ENETUNREACH;
2415 	fin->fin_hlen = (u_short)hlen;
2416 	fin->fin_dp = (char *)ip + hlen;
2417 	fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2418 	fin->fin_ifs = ifs;
2419 
2420 	SPL_NET(s);
2421 
2422 #ifdef	USE_INET6
2423 	if (v == 6) {
2424 		IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2425 		/*
2426 		 * Jumbo grams are quite likely too big for internal buffer
2427 		 * structures to handle comfortably, for now, so just drop
2428 		 * them.
2429 		 */
2430 		ip6 = (ip6_t *)ip;
2431 		fin->fin_plen = ntohs(ip6->ip6_plen);
2432 		if (fin->fin_plen == 0) {
2433 			READ_ENTER(&ifs->ifs_ipf_mutex);
2434 			pass = FR_BLOCK|FR_NOMATCH;
2435 			goto filtered;
2436 		}
2437 		fin->fin_plen += sizeof(ip6_t);
2438 	} else
2439 #endif
2440 	{
2441 #if (OpenBSD >= 200311) && defined(_KERNEL)
2442 		ip->ip_len = ntohs(ip->ip_len);
2443 		ip->ip_off = ntohs(ip->ip_off);
2444 #endif
2445 		fin->fin_plen = ip->ip_len;
2446 	}
2447 
2448 	if (fr_makefrip(hlen, ip, fin) == -1) {
2449 		READ_ENTER(&ifs->ifs_ipf_mutex);
2450 		pass = FR_BLOCK;
2451 		goto filtered;
2452 	}
2453 
2454 	/*
2455 	 * For at least IPv6 packets, if a m_pullup() fails then this pointer
2456 	 * becomes NULL and so we have no packet to free.
2457 	 */
2458 	if (*fin->fin_mp == NULL)
2459 		goto finished;
2460 
2461 	if (!out) {
2462 		if (v == 4) {
2463 #ifdef _KERNEL
2464 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2465 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2466 				fin->fin_flx |= FI_BADSRC;
2467 			}
2468 #endif
2469 			if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2470 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2471 				fin->fin_flx |= FI_LOWTTL;
2472 			}
2473 		}
2474 #ifdef USE_INET6
2475 		else  if (v == 6) {
2476 			ip6 = (ip6_t *)ip;
2477 #ifdef _KERNEL
2478 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2479 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2480 				fin->fin_flx |= FI_BADSRC;
2481 			}
2482 #endif
2483 			if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2484 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2485 				fin->fin_flx |= FI_LOWTTL;
2486 			}
2487 		}
2488 #endif
2489 	}
2490 
2491 	if (fin->fin_flx & FI_SHORT) {
2492 		IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2493 	}
2494 
2495 	READ_ENTER(&ifs->ifs_ipf_mutex);
2496 
2497 	/*
2498 	 * Check auth now.  This, combined with the check below to see if apass
2499 	 * is 0 is to ensure that we don't count the packet twice, which can
2500 	 * otherwise occur when we reprocess it.  As it is, we only count it
2501 	 * after it has no auth. table matchup.  This also stops NAT from
2502 	 * occuring until after the packet has been auth'd.
2503 	 */
2504 	fr = fr_checkauth(fin, &pass);
2505 	if (!out) {
2506 		switch (fin->fin_v)
2507 		{
2508 		case 4 :
2509 			if (fr_checknatin(fin, &pass) == -1) {
2510 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2511 				goto finished;
2512 			}
2513 			break;
2514 #ifdef	USE_INET6
2515 		case 6 :
2516 			if (fr_checknat6in(fin, &pass) == -1) {
2517 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2518 				goto finished;
2519 			}
2520 			break;
2521 #endif
2522 		default :
2523 			break;
2524 		}
2525 	}
2526 	if (!out)
2527 		(void) fr_acctpkt(fin, NULL);
2528 
2529 	if (fr == NULL)
2530 		if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2531 			fr = fr_knownfrag(fin, &pass);
2532 	if (fr == NULL)
2533 		fr = fr_checkstate(fin, &pass);
2534 
2535 	if ((pass & FR_NOMATCH) || (fr == NULL))
2536 		fr = fr_firewall(fin, &pass);
2537 
2538 	fin->fin_fr = fr;
2539 
2540 	/*
2541 	 * Only count/translate packets which will be passed on, out the
2542 	 * interface.
2543 	 */
2544 	if (out && FR_ISPASS(pass)) {
2545 		(void) fr_acctpkt(fin, NULL);
2546 
2547 		switch (fin->fin_v)
2548 		{
2549 		case 4 :
2550 			if (fr_checknatout(fin, &pass) == -1) {
2551 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2552 				goto finished;
2553 			}
2554 			break;
2555 #ifdef	USE_INET6
2556 		case 6 :
2557 			if (fr_checknat6out(fin, &pass) == -1) {
2558 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2559 				goto finished;
2560 			}
2561 			break;
2562 #endif
2563 		default :
2564 			break;
2565 		}
2566 
2567 		if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2568 			if (fr_updateipid(fin) == -1) {
2569 				IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2570 				pass &= ~FR_CMDMASK;
2571 				pass |= FR_BLOCK;
2572 			} else {
2573 				IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2574 			}
2575 		}
2576 	}
2577 
2578 #ifdef	IPFILTER_LOG
2579 	if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2580 		(void) fr_dolog(fin, &pass);
2581 	}
2582 #endif
2583 
2584 	/*
2585 	 * The FI_STATE flag is cleared here so that calling fr_checkstate
2586 	 * will work when called from inside of fr_fastroute.  Although
2587 	 * there is a similar flag, FI_NATED, for NAT, it does have the same
2588 	 * impact on code execution.
2589 	 */
2590 	fin->fin_flx &= ~FI_STATE;
2591 
2592 	/*
2593 	 * Only allow FR_DUP to work if a rule matched - it makes no sense to
2594 	 * set FR_DUP as a "default" as there are no instructions about where
2595 	 * to send the packet.  Use fin_m here because it may have changed
2596 	 * (without an update of 'm') in prior processing.
2597 	 */
2598 	if ((fr != NULL) && (pass & FR_DUP)) {
2599 		mc = M_DUPLICATE(fin->fin_m);
2600 #ifdef _KERNEL
2601 		mc->b_rptr += fin->fin_ipoff;
2602 #endif
2603 	}
2604 
2605 	if (pass & (FR_RETRST|FR_RETICMP)) {
2606 		/*
2607 		 * Should we return an ICMP packet to indicate error
2608 		 * status passing through the packet filter ?
2609 		 * WARNING: ICMP error packets AND TCP RST packets should
2610 		 * ONLY be sent in repsonse to incoming packets.  Sending them
2611 		 * in response to outbound packets can result in a panic on
2612 		 * some operating systems.
2613 		 */
2614 		if (!out) {
2615 			if (pass & FR_RETICMP) {
2616 				int dst;
2617 
2618 				if ((pass & FR_RETMASK) == FR_FAKEICMP)
2619 					dst = 1;
2620 				else
2621 					dst = 0;
2622 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2623 				/*
2624 				 * Assume it's possible to enter insane rule:
2625 				 * 	pass return-icmp in proto udp ...
2626 				 * then we have no other option than to forward
2627 				 * packet on loopback and give up any attempt
2628 				 * to create a fake response.
2629 				 */
2630 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2631 				    FR_ISBLOCK(pass)) {
2632 
2633 					if (fr_make_icmp(fin) == 0) {
2634 						IPF_BUMP(
2635 						ifs->ifs_frstats[out].fr_ret);
2636 					}
2637 					/*
2638 					 * we drop packet silently in case we
2639 					 * failed assemble fake response for it
2640 					 */
2641 					else if (*mp != NULL) {
2642 						FREE_MB_T(*mp);
2643 						m = *mp = NULL;
2644 					}
2645 
2646 					IPF_BUMP(
2647 					    ifs->ifs_frstats[out].fr_block);
2648 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2649 
2650 					return (0);
2651 				}
2652 #endif	/* _KERNEL && SOLARIS2 >= 10 */
2653 
2654 				(void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2655 				IPF_BUMP(ifs->ifs_frstats[out].fr_ret);
2656 
2657 			} else if (((pass & FR_RETMASK) == FR_RETRST) &&
2658 				   !(fin->fin_flx & FI_SHORT)) {
2659 
2660 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2661 				/*
2662 				 * Assume it's possible to enter insane rule:
2663 				 * 	pass return-rst in proto tcp ...
2664 				 * then we have no other option than to forward
2665 				 * packet on loopback and give up any attempt
2666 				 * to create a fake response.
2667 				 */
2668 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2669 				    FR_ISBLOCK(pass)) {
2670 					if (fr_make_rst(fin) == 0) {
2671 						IPF_BUMP(
2672 						ifs->ifs_frstats[out].fr_ret);
2673 					}
2674 					else if (mp != NULL) {
2675 					/*
2676 					 * we drop packet silently in case we
2677 					 * failed assemble fake response for it
2678 					 */
2679 						FREE_MB_T(*mp);
2680 						m = *mp = NULL;
2681 					}
2682 
2683 					IPF_BUMP(
2684 					    ifs->ifs_frstats[out].fr_block);
2685 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2686 
2687 					return (0);
2688 				 }
2689 #endif /* _KERNEL && _SOLARIS2 >= 10 */
2690 				if (fr_send_reset(fin) == 0) {
2691 					IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2692 				}
2693 			}
2694 		} else {
2695 			if (pass & FR_RETRST)
2696 				fin->fin_error = ECONNRESET;
2697 		}
2698 	}
2699 
2700 	/*
2701 	 * If we didn't drop off the bottom of the list of rules (and thus
2702 	 * the 'current' rule fr is not NULL), then we may have some extra
2703 	 * instructions about what to do with a packet.
2704 	 * Once we're finished return to our caller, freeing the packet if
2705 	 * we are dropping it (* BSD ONLY *).
2706 	 * Reassign m from fin_m as we may have a new buffer, now.
2707 	 */
2708 filtered:
2709 	m = fin->fin_m;
2710 
2711 	if (fr != NULL) {
2712 		frdest_t *fdp;
2713 
2714 		fdp = &fr->fr_tifs[fin->fin_rev];
2715 
2716 		if (!out && (pass & FR_FASTROUTE)) {
2717 			/*
2718 			 * For fastroute rule, no destioation interface defined
2719 			 * so pass NULL as the frdest_t parameter
2720 			 */
2721 			(void) fr_fastroute(m, mp, fin, NULL);
2722 			m = *mp = NULL;
2723 		} else if ((fdp->fd_ifp != NULL) &&
2724 			   (fdp->fd_ifp != (struct ifnet *)-1)) {
2725 			/* this is for to rules: */
2726 			(void) fr_fastroute(m, mp, fin, fdp);
2727 			m = *mp = NULL;
2728 		}
2729 
2730 		/*
2731 		 * Generate a duplicated packet.
2732 		 */
2733 		if (mc != NULL)
2734 			(void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2735 	}
2736 
2737 	if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT))
2738 		nat_uncreate(fin);
2739 
2740 	/*
2741 	 * This late because the likes of fr_fastroute() use fin_fr.
2742 	 */
2743 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2744 
2745 finished:
2746 	if (!FR_ISPASS(pass)) {
2747 		IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2748 		if (*mp != NULL) {
2749 			FREE_MB_T(*mp);
2750 			m = *mp = NULL;
2751 		}
2752 	} else {
2753 		IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2754 #if defined(_KERNEL) && defined(__sgi)
2755 		if ((fin->fin_hbuf != NULL) &&
2756 		    (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2757 			COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2758 		}
2759 #endif
2760 	}
2761 
2762 	SPL_X(s);
2763 
2764 #ifdef _KERNEL
2765 # if OpenBSD >= 200311
2766 	if (FR_ISPASS(pass) && (v == 4)) {
2767 		ip = fin->fin_ip;
2768 		ip->ip_len = ntohs(ip->ip_len);
2769 		ip->ip_off = ntohs(ip->ip_off);
2770 	}
2771 # endif
2772 	return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2773 #else /* _KERNEL */
2774 	FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2775 	if ((pass & FR_NOMATCH) != 0)
2776 		return 1;
2777 
2778 	if ((pass & FR_RETMASK) != 0)
2779 		switch (pass & FR_RETMASK)
2780 		{
2781 		case FR_RETRST :
2782 			return 3;
2783 		case FR_RETICMP :
2784 			return 4;
2785 		case FR_FAKEICMP :
2786 			return 5;
2787 		}
2788 
2789 	switch (pass & FR_CMDMASK)
2790 	{
2791 	case FR_PASS :
2792 		return 0;
2793 	case FR_BLOCK :
2794 		return -1;
2795 	case FR_AUTH :
2796 		return -2;
2797 	case FR_ACCOUNT :
2798 		return -3;
2799 	case FR_PREAUTH :
2800 		return -4;
2801 	}
2802 	return 2;
2803 #endif /* _KERNEL */
2804 }
2805 
2806 
2807 #ifdef	IPFILTER_LOG
2808 /* ------------------------------------------------------------------------ */
2809 /* Function:    fr_dolog                                                    */
2810 /* Returns:     frentry_t* - returns contents of fin_fr (no change made)    */
2811 /* Parameters:  fin(I) - pointer to packet information                      */
2812 /*              passp(IO) - pointer to current/new filter decision (unused) */
2813 /*                                                                          */
2814 /* Checks flags set to see how a packet should be logged, if it is to be    */
2815 /* logged.  Adjust statistics based on its success or not.                  */
2816 /* ------------------------------------------------------------------------ */
2817 frentry_t *fr_dolog(fin, passp)
2818 fr_info_t *fin;
2819 u_32_t *passp;
2820 {
2821 	u_32_t pass;
2822 	int out;
2823 	ipf_stack_t *ifs = fin->fin_ifs;
2824 
2825 	out = fin->fin_out;
2826 	pass = *passp;
2827 
2828 	if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2829 		pass |= FF_LOGNOMATCH;
2830 		IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2831 		goto logit;
2832 	} else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2833 	    (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2834 		if ((pass & FR_LOGMASK) != FR_LOGP)
2835 			pass |= FF_LOGPASS;
2836 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2837 		goto logit;
2838 	} else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2839 		   (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2840 		if ((pass & FR_LOGMASK) != FR_LOGB)
2841 			pass |= FF_LOGBLOCK;
2842 		IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2843 logit:
2844 		if (ipflog(fin, pass) == -1) {
2845 			IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2846 
2847 			/*
2848 			 * If the "or-block" option has been used then
2849 			 * block the packet if we failed to log it.
2850 			 */
2851 			if ((pass & FR_LOGORBLOCK) &&
2852 			    FR_ISPASS(pass)) {
2853 				pass &= ~FR_CMDMASK;
2854 				pass |= FR_BLOCK;
2855 			}
2856 		}
2857 		*passp = pass;
2858 	}
2859 
2860 	return fin->fin_fr;
2861 }
2862 #endif /* IPFILTER_LOG */
2863 
2864 
2865 /* ------------------------------------------------------------------------ */
2866 /* Function:    ipf_cksum                                                   */
2867 /* Returns:     u_short - IP header checksum                                */
2868 /* Parameters:  addr(I) - pointer to start of buffer to checksum            */
2869 /*              len(I)  - length of buffer in bytes                         */
2870 /*                                                                          */
2871 /* Calculate the two's complement 16 bit checksum of the buffer passed.     */
2872 /*                                                                          */
2873 /* N.B.: addr should be 16bit aligned.                                      */
2874 /* ------------------------------------------------------------------------ */
2875 u_short ipf_cksum(addr, len)
2876 u_short *addr;
2877 int len;
2878 {
2879 	u_32_t sum = 0;
2880 
2881 	for (sum = 0; len > 1; len -= 2)
2882 		sum += *addr++;
2883 
2884 	/* mop up an odd byte, if necessary */
2885 	if (len == 1)
2886 		sum += *(u_char *)addr;
2887 
2888 	/*
2889 	 * add back carry outs from top 16 bits to low 16 bits
2890 	 */
2891 	sum = (sum >> 16) + (sum & 0xffff);	/* add hi 16 to low 16 */
2892 	sum += (sum >> 16);			/* add carry */
2893 	return (u_short)(~sum);
2894 }
2895 
2896 
2897 /* ------------------------------------------------------------------------ */
2898 /* Function:    fr_cksum                                                    */
2899 /* Returns:     u_short - layer 4 checksum                                  */
2900 /* Parameters:  m(I  )     - pointer to buffer holding packet               */
2901 /*              ip(I)      - pointer to IP header                           */
2902 /*              l4proto(I) - protocol to caclulate checksum for             */
2903 /*              l4hdr(I)   - pointer to layer 4 header                      */
2904 /*                                                                          */
2905 /* Calculates the TCP checksum for the packet held in "m", using the data   */
2906 /* in the IP header "ip" to seed it.                                        */
2907 /*                                                                          */
2908 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2909 /* and the TCP header.  We also assume that data blocks aren't allocated in */
2910 /* odd sizes.                                                               */
2911 /*                                                                          */
2912 /* Expects ip_len to be in host byte order when called.                     */
2913 /* ------------------------------------------------------------------------ */
2914 u_short fr_cksum(m, ip, l4proto, l4hdr)
2915 mb_t *m;
2916 ip_t *ip;
2917 int l4proto;
2918 void *l4hdr;
2919 {
2920 	u_short *sp, slen, sumsave, l4hlen, *csump;
2921 	u_int sum, sum2;
2922 	int hlen;
2923 #ifdef	USE_INET6
2924 	ip6_t *ip6;
2925 #endif
2926 
2927 	csump = NULL;
2928 	sumsave = 0;
2929 	l4hlen = 0;
2930 	sp = NULL;
2931 	slen = 0;
2932 	hlen = 0;
2933 	sum = 0;
2934 
2935 	/*
2936 	 * Add up IP Header portion
2937 	 */
2938 #ifdef	USE_INET6
2939 	if (IP_V(ip) == 4) {
2940 #endif
2941 		hlen = IP_HL(ip) << 2;
2942 		slen = ip->ip_len - hlen;
2943 		sum = htons((u_short)l4proto);
2944 		sum += htons(slen);
2945 		sp = (u_short *)&ip->ip_src;
2946 		sum += *sp++;	/* ip_src */
2947 		sum += *sp++;
2948 		sum += *sp++;	/* ip_dst */
2949 		sum += *sp++;
2950 #ifdef	USE_INET6
2951 	} else if (IP_V(ip) == 6) {
2952 		ip6 = (ip6_t *)ip;
2953 		hlen = sizeof(*ip6);
2954 		slen = ntohs(ip6->ip6_plen);
2955 		sum = htons((u_short)l4proto);
2956 		sum += htons(slen);
2957 		sp = (u_short *)&ip6->ip6_src;
2958 		sum += *sp++;	/* ip6_src */
2959 		sum += *sp++;
2960 		sum += *sp++;
2961 		sum += *sp++;
2962 		sum += *sp++;
2963 		sum += *sp++;
2964 		sum += *sp++;
2965 		sum += *sp++;
2966 		sum += *sp++;	/* ip6_dst */
2967 		sum += *sp++;
2968 		sum += *sp++;
2969 		sum += *sp++;
2970 		sum += *sp++;
2971 		sum += *sp++;
2972 		sum += *sp++;
2973 		sum += *sp++;
2974 	}
2975 #endif
2976 
2977 	switch (l4proto)
2978 	{
2979 	case IPPROTO_UDP :
2980 		csump = &((udphdr_t *)l4hdr)->uh_sum;
2981 		l4hlen = sizeof(udphdr_t);
2982 		break;
2983 
2984 	case IPPROTO_TCP :
2985 		csump = &((tcphdr_t *)l4hdr)->th_sum;
2986 		l4hlen = sizeof(tcphdr_t);
2987 		break;
2988 	case IPPROTO_ICMP :
2989 		csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
2990 		l4hlen = 4;
2991 		sum = 0;
2992 		break;
2993 	default :
2994 		break;
2995 	}
2996 
2997 	if (csump != NULL) {
2998 		sumsave = *csump;
2999 		*csump = 0;
3000 	}
3001 
3002 	l4hlen = l4hlen;	/* LINT */
3003 
3004 #ifdef	_KERNEL
3005 # ifdef MENTAT
3006 	{
3007 	void *rp = m->b_rptr;
3008 
3009 	if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
3010 		m->b_rptr = (u_char *)ip;
3011 	sum2 = ip_cksum(m, hlen, sum);	/* hlen == offset */
3012 	m->b_rptr = rp;
3013 	sum2 = (sum2 & 0xffff) + (sum2 >> 16);
3014 	sum2 = ~sum2 & 0xffff;
3015 	}
3016 # else /* MENTAT */
3017 #  if defined(BSD) || defined(sun)
3018 #   if BSD >= 199103
3019 	m->m_data += hlen;
3020 #   else
3021 	m->m_off += hlen;
3022 #   endif
3023 	m->m_len -= hlen;
3024 	sum2 = in_cksum(m, slen);
3025 	m->m_len += hlen;
3026 #   if BSD >= 199103
3027 	m->m_data -= hlen;
3028 #   else
3029 	m->m_off -= hlen;
3030 #   endif
3031 	/*
3032 	 * Both sum and sum2 are partial sums, so combine them together.
3033 	 */
3034 	sum += ~sum2 & 0xffff;
3035 	while (sum > 0xffff)
3036 		sum = (sum & 0xffff) + (sum >> 16);
3037 	sum2 = ~sum & 0xffff;
3038 #  else /* defined(BSD) || defined(sun) */
3039 {
3040 	union {
3041 		u_char	c[2];
3042 		u_short	s;
3043 	} bytes;
3044 	u_short len = ip->ip_len;
3045 #   if defined(__sgi)
3046 	int add;
3047 #   endif
3048 
3049 	/*
3050 	 * Add up IP Header portion
3051 	 */
3052 	if (sp != (u_short *)l4hdr)
3053 		sp = (u_short *)l4hdr;
3054 
3055 	switch (l4proto)
3056 	{
3057 	case IPPROTO_UDP :
3058 		sum += *sp++;	/* sport */
3059 		sum += *sp++;	/* dport */
3060 		sum += *sp++;	/* udp length */
3061 		sum += *sp++;	/* checksum */
3062 		break;
3063 
3064 	case IPPROTO_TCP :
3065 		sum += *sp++;	/* sport */
3066 		sum += *sp++;	/* dport */
3067 		sum += *sp++;	/* seq */
3068 		sum += *sp++;
3069 		sum += *sp++;	/* ack */
3070 		sum += *sp++;
3071 		sum += *sp++;	/* off */
3072 		sum += *sp++;	/* win */
3073 		sum += *sp++;	/* checksum */
3074 		sum += *sp++;	/* urp */
3075 		break;
3076 	case IPPROTO_ICMP :
3077 		sum = *sp++;	/* type/code */
3078 		sum += *sp++;	/* checksum */
3079 		break;
3080 	}
3081 
3082 #   ifdef	__sgi
3083 	/*
3084 	 * In case we had to copy the IP & TCP header out of mbufs,
3085 	 * skip over the mbuf bits which are the header
3086 	 */
3087 	if ((caddr_t)ip != mtod(m, caddr_t)) {
3088 		hlen = (caddr_t)sp - (caddr_t)ip;
3089 		while (hlen) {
3090 			add = MIN(hlen, m->m_len);
3091 			sp = (u_short *)(mtod(m, caddr_t) + add);
3092 			hlen -= add;
3093 			if (add == m->m_len) {
3094 				m = m->m_next;
3095 				if (!hlen) {
3096 					if (!m)
3097 						break;
3098 					sp = mtod(m, u_short *);
3099 				}
3100 				PANIC((!m),("fr_cksum(1): not enough data"));
3101 			}
3102 		}
3103 	}
3104 #   endif
3105 
3106 	len -= (l4hlen + hlen);
3107 	if (len <= 0)
3108 		goto nodata;
3109 
3110 	while (len > 1) {
3111 		if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3112 			m = m->m_next;
3113 			PANIC((!m),("fr_cksum(2): not enough data"));
3114 			sp = mtod(m, u_short *);
3115 		}
3116 		if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3117 			bytes.c[0] = *(u_char *)sp;
3118 			m = m->m_next;
3119 			PANIC((!m),("fr_cksum(3): not enough data"));
3120 			sp = mtod(m, u_short *);
3121 			bytes.c[1] = *(u_char *)sp;
3122 			sum += bytes.s;
3123 			sp = (u_short *)((u_char *)sp + 1);
3124 		}
3125 		if ((u_long)sp & 1) {
3126 			bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3127 			sum += bytes.s;
3128 		} else
3129 			sum += *sp++;
3130 		len -= 2;
3131 	}
3132 
3133 	if (len != 0)
3134 		sum += ntohs(*(u_char *)sp << 8);
3135 nodata:
3136 	while (sum > 0xffff)
3137 		sum = (sum & 0xffff) + (sum >> 16);
3138 	sum2 = (u_short)(~sum & 0xffff);
3139 }
3140 #  endif /*  defined(BSD) || defined(sun) */
3141 # endif /* MENTAT */
3142 #else /* _KERNEL */
3143 	for (; slen > 1; slen -= 2)
3144 	        sum += *sp++;
3145 	if (slen)
3146 		sum += ntohs(*(u_char *)sp << 8);
3147 	while (sum > 0xffff)
3148 		sum = (sum & 0xffff) + (sum >> 16);
3149 	sum2 = (u_short)(~sum & 0xffff);
3150 #endif /* _KERNEL */
3151 	if (csump != NULL)
3152 		*csump = sumsave;
3153 	return sum2;
3154 }
3155 
3156 
3157 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3158     defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3159 /*
3160  * Copyright (c) 1982, 1986, 1988, 1991, 1993
3161  *	The Regents of the University of California.  All rights reserved.
3162  *
3163  * Redistribution and use in source and binary forms, with or without
3164  * modification, are permitted provided that the following conditions
3165  * are met:
3166  * 1. Redistributions of source code must retain the above copyright
3167  *    notice, this list of conditions and the following disclaimer.
3168  * 2. Redistributions in binary form must reproduce the above copyright
3169  *    notice, this list of conditions and the following disclaimer in the
3170  *    documentation and/or other materials provided with the distribution.
3171  * 3. Neither the name of the University nor the names of its contributors
3172  *    may be used to endorse or promote products derived from this software
3173  *    without specific prior written permission.
3174  *
3175  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3176  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3177  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3178  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3179  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3180  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3181  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3182  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3183  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3184  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3185  * SUCH DAMAGE.
3186  *
3187  *	@(#)uipc_mbuf.c	8.2 (Berkeley) 1/4/94
3188  * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3189  */
3190 /*
3191  * Copy data from an mbuf chain starting "off" bytes from the beginning,
3192  * continuing for "len" bytes, into the indicated buffer.
3193  */
3194 void
3195 m_copydata(m, off, len, cp)
3196 	mb_t *m;
3197 	int off;
3198 	int len;
3199 	caddr_t cp;
3200 {
3201 	unsigned count;
3202 
3203 	if (off < 0 || len < 0)
3204 		panic("m_copydata");
3205 	while (off > 0) {
3206 		if (m == 0)
3207 			panic("m_copydata");
3208 		if (off < m->m_len)
3209 			break;
3210 		off -= m->m_len;
3211 		m = m->m_next;
3212 	}
3213 	while (len > 0) {
3214 		if (m == 0)
3215 			panic("m_copydata");
3216 		count = MIN(m->m_len - off, len);
3217 		bcopy(mtod(m, caddr_t) + off, cp, count);
3218 		len -= count;
3219 		cp += count;
3220 		off = 0;
3221 		m = m->m_next;
3222 	}
3223 }
3224 
3225 
3226 /*
3227  * Copy data from a buffer back into the indicated mbuf chain,
3228  * starting "off" bytes from the beginning, extending the mbuf
3229  * chain if necessary.
3230  */
3231 void
3232 m_copyback(m0, off, len, cp)
3233 	struct	mbuf *m0;
3234 	int off;
3235 	int len;
3236 	caddr_t cp;
3237 {
3238 	int mlen;
3239 	struct mbuf *m = m0, *n;
3240 	int totlen = 0;
3241 
3242 	if (m0 == 0)
3243 		return;
3244 	while (off > (mlen = m->m_len)) {
3245 		off -= mlen;
3246 		totlen += mlen;
3247 		if (m->m_next == 0) {
3248 			n = m_getclr(M_DONTWAIT, m->m_type);
3249 			if (n == 0)
3250 				goto out;
3251 			n->m_len = min(MLEN, len + off);
3252 			m->m_next = n;
3253 		}
3254 		m = m->m_next;
3255 	}
3256 	while (len > 0) {
3257 		mlen = min(m->m_len - off, len);
3258 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3259 		cp += mlen;
3260 		len -= mlen;
3261 		mlen += off;
3262 		off = 0;
3263 		totlen += mlen;
3264 		if (len == 0)
3265 			break;
3266 		if (m->m_next == 0) {
3267 			n = m_get(M_DONTWAIT, m->m_type);
3268 			if (n == 0)
3269 				break;
3270 			n->m_len = min(MLEN, len);
3271 			m->m_next = n;
3272 		}
3273 		m = m->m_next;
3274 	}
3275 out:
3276 #if 0
3277 	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3278 		m->m_pkthdr.len = totlen;
3279 #endif
3280 	return;
3281 }
3282 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3283 
3284 
3285 /* ------------------------------------------------------------------------ */
3286 /* Function:    fr_findgroup                                                */
3287 /* Returns:     frgroup_t * - NULL = group not found, else pointer to group */
3288 /* Parameters:  group(I) - group name to search for                         */
3289 /*              unit(I)  - device to which this group belongs               */
3290 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3291 /*              fgpp(O)  - pointer to place to store pointer to the pointer */
3292 /*                         to where to add the next (last) group or where   */
3293 /*                         to delete group from.                            */
3294 /*                                                                          */
3295 /* Search amongst the defined groups for a particular group number.         */
3296 /* ------------------------------------------------------------------------ */
3297 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3298 char *group;
3299 minor_t unit;
3300 int set;
3301 frgroup_t ***fgpp;
3302 ipf_stack_t *ifs;
3303 {
3304 	frgroup_t *fg, **fgp;
3305 
3306 	/*
3307 	 * Which list of groups to search in is dependent on which list of
3308 	 * rules are being operated on.
3309 	 */
3310 	fgp = &ifs->ifs_ipfgroups[unit][set];
3311 
3312 	while ((fg = *fgp) != NULL) {
3313 		if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3314 			break;
3315 		else
3316 			fgp = &fg->fg_next;
3317 	}
3318 	if (fgpp != NULL)
3319 		*fgpp = fgp;
3320 	return fg;
3321 }
3322 
3323 
3324 /* ------------------------------------------------------------------------ */
3325 /* Function:    fr_addgroup                                                 */
3326 /* Returns:     frgroup_t * - NULL == did not create group,                 */
3327 /*                            != NULL == pointer to the group               */
3328 /* Parameters:  num(I)   - group number to add                              */
3329 /*              head(I)  - rule pointer that is using this as the head      */
3330 /*              flags(I) - rule flags which describe the type of rule it is */
3331 /*              unit(I)  - device to which this group will belong to        */
3332 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3333 /* Write Locks: ipf_mutex                                                   */
3334 /*                                                                          */
3335 /* Add a new group head, or if it already exists, increase the reference    */
3336 /* count to it.                                                             */
3337 /* ------------------------------------------------------------------------ */
3338 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3339 char *group;
3340 void *head;
3341 u_32_t flags;
3342 minor_t unit;
3343 int set;
3344 ipf_stack_t *ifs;
3345 {
3346 	frgroup_t *fg, **fgp;
3347 	u_32_t gflags;
3348 
3349 	if (group == NULL)
3350 		return NULL;
3351 
3352 	if (unit == IPL_LOGIPF && *group == '\0')
3353 		return NULL;
3354 
3355 	fgp = NULL;
3356 	gflags = flags & FR_INOUT;
3357 
3358 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3359 	if (fg != NULL) {
3360 		if (fg->fg_flags == 0)
3361 			fg->fg_flags = gflags;
3362 		else if (gflags != fg->fg_flags)
3363 			return NULL;
3364 		fg->fg_ref++;
3365 		return fg;
3366 	}
3367 	KMALLOC(fg, frgroup_t *);
3368 	if (fg != NULL) {
3369 		fg->fg_head = head;
3370 		fg->fg_start = NULL;
3371 		fg->fg_next = *fgp;
3372 		bcopy(group, fg->fg_name, FR_GROUPLEN);
3373 		fg->fg_flags = gflags;
3374 		fg->fg_ref = 1;
3375 		*fgp = fg;
3376 	}
3377 	return fg;
3378 }
3379 
3380 
3381 /* ------------------------------------------------------------------------ */
3382 /* Function:    fr_delgroup                                                 */
3383 /* Returns:     Nil                                                         */
3384 /* Parameters:  group(I) - group name to delete                             */
3385 /*              unit(I)  - device to which this group belongs               */
3386 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3387 /* Write Locks: ipf_mutex                                                   */
3388 /*                                                                          */
3389 /* Attempt to delete a group head.                                          */
3390 /* Only do this when its reference count reaches 0.                         */
3391 /* ------------------------------------------------------------------------ */
3392 void fr_delgroup(group, unit, set, ifs)
3393 char *group;
3394 minor_t unit;
3395 int set;
3396 ipf_stack_t *ifs;
3397 {
3398 	frgroup_t *fg, **fgp;
3399 
3400 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3401 	if (fg == NULL)
3402 		return;
3403 
3404 	fg->fg_ref--;
3405 	if (fg->fg_ref == 0) {
3406 		*fgp = fg->fg_next;
3407 		KFREE(fg);
3408 	}
3409 }
3410 
3411 
3412 /* ------------------------------------------------------------------------ */
3413 /* Function:    fr_getrulen                                                 */
3414 /* Returns:     frentry_t * - NULL == not found, else pointer to rule n     */
3415 /* Parameters:  unit(I)  - device for which to count the rule's number      */
3416 /*              flags(I) - which set of rules to find the rule in           */
3417 /*              group(I) - group name                                       */
3418 /*              n(I)     - rule number to find                              */
3419 /*                                                                          */
3420 /* Find rule # n in group # g and return a pointer to it.  Return NULl if   */
3421 /* group # g doesn't exist or there are less than n rules in the group.     */
3422 /* ------------------------------------------------------------------------ */
3423 frentry_t *fr_getrulen(unit, group, n, ifs)
3424 int unit;
3425 char *group;
3426 u_32_t n;
3427 ipf_stack_t *ifs;
3428 {
3429 	frentry_t *fr;
3430 	frgroup_t *fg;
3431 
3432 	fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3433 	if (fg == NULL)
3434 		return NULL;
3435 	for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3436 		;
3437 	if (n != 0)
3438 		return NULL;
3439 	return fr;
3440 }
3441 
3442 
3443 /* ------------------------------------------------------------------------ */
3444 /* Function:    fr_rulen                                                    */
3445 /* Returns:     int - >= 0 - rule number, -1 == search failed               */
3446 /* Parameters:  unit(I) - device for which to count the rule's number       */
3447 /*              fr(I)   - pointer to rule to match                          */
3448 /*                                                                          */
3449 /* Return the number for a rule on a specific filtering device.             */
3450 /* ------------------------------------------------------------------------ */
3451 int fr_rulen(unit, fr, ifs)
3452 int unit;
3453 frentry_t *fr;
3454 ipf_stack_t *ifs;
3455 {
3456 	frentry_t *fh;
3457 	frgroup_t *fg;
3458 	u_32_t n = 0;
3459 
3460 	if (fr == NULL)
3461 		return -1;
3462 	fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3463 	if (fg == NULL)
3464 		return -1;
3465 	for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3466 		if (fh == fr)
3467 			break;
3468 	if (fh == NULL)
3469 		return -1;
3470 	return n;
3471 }
3472 
3473 
3474 /* ------------------------------------------------------------------------ */
3475 /* Function:    frflushlist                                                 */
3476 /* Returns:     int - >= 0 - number of flushed rules                        */
3477 /* Parameters:  set(I)   - which set of rules (inactive/inactive) this is   */
3478 /*              unit(I)  - device for which to flush rules                  */
3479 /*              flags(I) - which set of rules to flush                      */
3480 /*              nfreedp(O) - pointer to int where flush count is stored     */
3481 /*              listp(I)   - pointer to list to flush pointer               */
3482 /* Write Locks: ipf_mutex                                                   */
3483 /*                                                                          */
3484 /* Recursively flush rules from the list, descending groups as they are     */
3485 /* encountered.  if a rule is the head of a group and it has lost all its   */
3486 /* group members, then also delete the group reference.  nfreedp is needed  */
3487 /* to store the accumulating count of rules removed, whereas the returned   */
3488 /* value is just the number removed from the current list.  The latter is   */
3489 /* needed to correctly adjust reference counts on rules that define groups. */
3490 /*                                                                          */
3491 /* NOTE: Rules not loaded from user space cannot be flushed.                */
3492 /* ------------------------------------------------------------------------ */
3493 static int frflushlist(set, unit, nfreedp, listp, ifs)
3494 int set;
3495 minor_t unit;
3496 int *nfreedp;
3497 frentry_t **listp;
3498 ipf_stack_t *ifs;
3499 {
3500 	int freed = 0;
3501 	frentry_t *fp;
3502 
3503 	while ((fp = *listp) != NULL) {
3504 		if ((fp->fr_type & FR_T_BUILTIN) ||
3505 		    !(fp->fr_flags & FR_COPIED)) {
3506 			listp = &fp->fr_next;
3507 			continue;
3508 		}
3509 		*listp = fp->fr_next;
3510 		if (fp->fr_grp != NULL) {
3511 			(void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3512 		}
3513 
3514 		if (fp->fr_grhead != NULL) {
3515 			fr_delgroup(fp->fr_grhead, unit, set, ifs);
3516 			*fp->fr_grhead = '\0';
3517 		}
3518 
3519 		ASSERT(fp->fr_ref > 0);
3520 		fp->fr_next = NULL;
3521 		if (fr_derefrule(&fp, ifs) == 0)
3522 			freed++;
3523 	}
3524 	*nfreedp += freed;
3525 	return freed;
3526 }
3527 
3528 
3529 /* ------------------------------------------------------------------------ */
3530 /* Function:    frflush                                                     */
3531 /* Returns:     int - >= 0 - number of flushed rules                        */
3532 /* Parameters:  unit(I)  - device for which to flush rules                  */
3533 /*              flags(I) - which set of rules to flush                      */
3534 /*                                                                          */
3535 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3536 /* and IPv6) as defined by the value of flags.                              */
3537 /* ------------------------------------------------------------------------ */
3538 int frflush(unit, proto, flags, ifs)
3539 minor_t unit;
3540 int proto, flags;
3541 ipf_stack_t *ifs;
3542 {
3543 	int flushed = 0, set;
3544 
3545 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3546 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
3547 
3548 	set = ifs->ifs_fr_active;
3549 	if ((flags & FR_INACTIVE) == FR_INACTIVE)
3550 		set = 1 - set;
3551 
3552 	if (flags & FR_OUTQUE) {
3553 		if (proto == 0 || proto == 6) {
3554 			(void) frflushlist(set, unit,
3555 			    &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3556 			(void) frflushlist(set, unit,
3557 			    &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3558 		}
3559 		if (proto == 0 || proto == 4) {
3560 			(void) frflushlist(set, unit,
3561 			    &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3562 			(void) frflushlist(set, unit,
3563 			    &flushed, &ifs->ifs_ipacct[1][set], ifs);
3564 		}
3565 	}
3566 	if (flags & FR_INQUE) {
3567 		if (proto == 0 || proto == 6) {
3568 			(void) frflushlist(set, unit,
3569 			    &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3570 			(void) frflushlist(set, unit,
3571 			    &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3572 		}
3573 		if (proto == 0 || proto == 4) {
3574 			(void) frflushlist(set, unit,
3575 			    &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3576 			(void) frflushlist(set, unit,
3577 			    &flushed, &ifs->ifs_ipacct[0][set], ifs);
3578 		}
3579 	}
3580 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3581 
3582 	if (unit == IPL_LOGIPF) {
3583 		int tmp;
3584 
3585 		tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3586 		if (tmp >= 0)
3587 			flushed += tmp;
3588 	}
3589 	return flushed;
3590 }
3591 
3592 
3593 /* ------------------------------------------------------------------------ */
3594 /* Function:    memstr                                                      */
3595 /* Returns:     char *  - NULL if failed, != NULL pointer to matching bytes */
3596 /* Parameters:  src(I)  - pointer to byte sequence to match                 */
3597 /*              dst(I)  - pointer to byte sequence to search                */
3598 /*              slen(I) - match length                                      */
3599 /*              dlen(I) - length available to search in                     */
3600 /*                                                                          */
3601 /* Search dst for a sequence of bytes matching those at src and extend for  */
3602 /* slen bytes.                                                              */
3603 /* ------------------------------------------------------------------------ */
3604 char *memstr(src, dst, slen, dlen)
3605 char *src, *dst;
3606 int slen, dlen;
3607 {
3608 	char *s = NULL;
3609 
3610 	while (dlen >= slen) {
3611 		if (bcmp(src, dst, slen) == 0) {
3612 			s = dst;
3613 			break;
3614 		}
3615 		dst++;
3616 		dlen--;
3617 	}
3618 	return s;
3619 }
3620 /* ------------------------------------------------------------------------ */
3621 /* Function:    fr_fixskip                                                  */
3622 /* Returns:     Nil                                                         */
3623 /* Parameters:  listp(IO)    - pointer to start of list with skip rule      */
3624 /*              rp(I)        - rule added/removed with skip in it.          */
3625 /*              addremove(I) - adjustment (-1/+1) to make to skip count,    */
3626 /*                             depending on whether a rule was just added   */
3627 /*                             or removed.                                  */
3628 /*                                                                          */
3629 /* Adjust all the rules in a list which would have skip'd past the position */
3630 /* where we are inserting to skip to the right place given the change.      */
3631 /* ------------------------------------------------------------------------ */
3632 void fr_fixskip(listp, rp, addremove)
3633 frentry_t **listp, *rp;
3634 int addremove;
3635 {
3636 	int rules, rn;
3637 	frentry_t *fp;
3638 
3639 	rules = 0;
3640 	for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3641 		rules++;
3642 
3643 	if (!fp)
3644 		return;
3645 
3646 	for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3647 		if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3648 			fp->fr_arg += addremove;
3649 }
3650 
3651 
3652 #ifdef	_KERNEL
3653 /* ------------------------------------------------------------------------ */
3654 /* Function:    count4bits                                                  */
3655 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3656 /* Parameters:  ip(I) - 32bit IP address                                    */
3657 /*                                                                          */
3658 /* IPv4 ONLY                                                                */
3659 /* count consecutive 1's in bit mask.  If the mask generated by counting    */
3660 /* consecutive 1's is different to that passed, return -1, else return #    */
3661 /* of bits.                                                                 */
3662 /* ------------------------------------------------------------------------ */
3663 int	count4bits(ip)
3664 u_32_t	ip;
3665 {
3666 	u_32_t	ipn;
3667 	int	cnt = 0, i, j;
3668 
3669 	ip = ipn = ntohl(ip);
3670 	for (i = 32; i; i--, ipn *= 2)
3671 		if (ipn & 0x80000000)
3672 			cnt++;
3673 		else
3674 			break;
3675 	ipn = 0;
3676 	for (i = 32, j = cnt; i; i--, j--) {
3677 		ipn *= 2;
3678 		if (j > 0)
3679 			ipn++;
3680 	}
3681 	if (ipn == ip)
3682 		return cnt;
3683 	return -1;
3684 }
3685 
3686 
3687 #ifdef USE_INET6
3688 /* ------------------------------------------------------------------------ */
3689 /* Function:    count6bits                                                  */
3690 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3691 /* Parameters:  msk(I) - pointer to start of IPv6 bitmask                   */
3692 /*                                                                          */
3693 /* IPv6 ONLY                                                                */
3694 /* count consecutive 1's in bit mask.                                       */
3695 /* ------------------------------------------------------------------------ */
3696 int count6bits(msk)
3697 u_32_t *msk;
3698 {
3699 	int i = 0, k;
3700 	u_32_t j;
3701 
3702 	for (k = 3; k >= 0; k--)
3703 		if (msk[k] == 0xffffffff)
3704 			i += 32;
3705 		else {
3706 			for (j = msk[k]; j; j <<= 1)
3707 				if (j & 0x80000000)
3708 					i++;
3709 		}
3710 	return i;
3711 }
3712 # endif
3713 #endif /* _KERNEL */
3714 
3715 
3716 /* ------------------------------------------------------------------------ */
3717 /* Function:    fr_ifsync                                                   */
3718 /* Returns:     void *    - new interface identifier                        */
3719 /* Parameters:  action(I)  - type of synchronisation to do                  */
3720 /*              v(I)       - IP version being sync'd (v4 or v6)             */
3721 /*              newifp(I)  - interface identifier being introduced/removed  */
3722 /*              oldifp(I)  - interface identifier in a filter rule          */
3723 /*              newname(I) - name associated with newifp interface          */
3724 /*              oldname(I) - name associated with oldifp interface          */
3725 /*		ifs       - pointer to IPF stack instance		    */
3726 /*                                                                          */
3727 /* This function returns what the new value for "oldifp" should be for its  */
3728 /* caller.  In some cases it will not change, in some it will.              */
3729 /* action == IPFSYNC_RESYNC                                                 */
3730 /*   a new value for oldifp will always be looked up, according to oldname, */
3731 /*   the values of newname and newifp are ignored.                          */
3732 /* action == IPFSYNC_NEWIFP                                                 */
3733 /*   if oldname matches newname then we are doing a sync for the matching   */
3734 /*   interface, so we return newifp to be used in place of oldifp.  If the  */
3735 /*   the names don't match, just return oldifp.                             */
3736 /* action == IPFSYNC_OLDIFP                                                 */
3737 /*   if oldifp matches newifp then we are are doing a sync to remove any    */
3738 /*   references to oldifp, so we return "-1".                               */
3739 /* -----								    */
3740 /* NOTE:								    */
3741 /* This function processes NIC event from PF_HOOKS. The action parameter    */
3742 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is	    */
3743 /* one single switch statement() in ipf_nic_event_vx() function, which	    */
3744 /* translates the HOOK event type to action parameter passed to fr_ifsync.  */
3745 /* The translation table looks as follows:				    */
3746 /*	event		| action					    */
3747 /*	----------------+-------------					    */
3748 /*	NE_PLUMB	| IPFSYNC_NEWIFP				    */
3749 /*	NE_UNPLUMB	| IPFSYNC_OLDIFP				    */
3750 /*    NE_ADDRESS_CHANGE	| IPFSYNC_RESYNC				    */
3751 /*									    */
3752 /* The oldname and oldifp parameters are taken from IPF entry (rule, state  */
3753 /* table entry, NAT table entry, fragment ...). The newname and newifp	    */
3754 /* parameters come from hook event data, parameters are taken from event    */
3755 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is	    */
3756 /* notified by hook function.						    */
3757 /*									    */
3758 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3759 /* to plumb the interface, which is already plumbed. In such case we always */
3760 /* get the event from PF_HOOKS as follows:				    */
3761 /*	event:	NE_PLUMB						    */
3762 /*	NIC:	0x0							    */
3763 /* ------------------------------------------------------------------------ */
3764 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3765 int action, v;
3766 char *newname, *oldname;
3767 void *newifp, *oldifp;
3768 ipf_stack_t *ifs;
3769 {
3770 	void *rval = oldifp;
3771 
3772 	switch (action)
3773 	{
3774 	case IPFSYNC_RESYNC :
3775 		if (oldname[0] != '\0') {
3776 			rval = fr_resolvenic(oldname, v, ifs);
3777 		}
3778 		break;
3779 	case IPFSYNC_NEWIFP :
3780 		if (!strncmp(newname, oldname, LIFNAMSIZ))
3781 			rval = newifp;
3782 		break;
3783 	case IPFSYNC_OLDIFP :
3784 		/*
3785 		 * If interface gets unplumbed it must be invalidated, which
3786 		 * means set all existing references to the interface to -1.
3787 		 * We don't want to invalidate references for wildcard
3788 		 * (unbound) rules (entries).
3789 		 */
3790 		if (newifp == oldifp)
3791 			rval = (oldifp) ? (void *)-1 : NULL;
3792 		break;
3793 	}
3794 
3795 	return rval;
3796 }
3797 
3798 
3799 /* ------------------------------------------------------------------------ */
3800 /* Function:    frsynclist                                                  */
3801 /* Returns:     void                                                        */
3802 /* Parameters:  action(I) - type of synchronisation to do                   */
3803 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3804 /*              ifp(I)    - interface identifier associated with action     */
3805 /*              ifname(I) - name associated with ifp parameter              */
3806 /*              fr(I)     - pointer to filter rule                          */
3807 /*		ifs       - pointer to IPF stack instance		    */
3808 /* Write Locks: ipf_mutex                                                   */
3809 /*                                                                          */
3810 /* Walk through a list of filter rules and resolve any interface names into */
3811 /* pointers.  Where dynamic addresses are used, also update the IP address  */
3812 /* used in the rule.  The interface pointer is used to limit the lookups to */
3813 /* a specific set of matching names if it is non-NULL.                      */
3814 /* ------------------------------------------------------------------------ */
3815 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3816 int action, v;
3817 void *ifp;
3818 char *ifname;
3819 frentry_t *fr;
3820 ipf_stack_t *ifs;
3821 {
3822 	frdest_t *fdp;
3823 	int rv, i;
3824 
3825 	for (; fr; fr = fr->fr_next) {
3826 		rv = fr->fr_v;
3827 		if (v != 0 && v != rv)
3828 			continue;
3829 
3830 		/*
3831 		 * Lookup all the interface names that are part of the rule.
3832 		 */
3833 		for (i = 0; i < 4; i++) {
3834 			fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3835 						   fr->fr_ifnames[i],
3836 						   ifp, fr->fr_ifas[i],
3837 						   ifs);
3838 		}
3839 
3840 		fdp = &fr->fr_tifs[0];
3841 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3842 					   ifp, fdp->fd_ifp, ifs);
3843 
3844 		fdp = &fr->fr_tifs[1];
3845 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3846 					   ifp, fdp->fd_ifp, ifs);
3847 
3848 		fdp = &fr->fr_dif;
3849 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3850 					   ifp, fdp->fd_ifp, ifs);
3851 
3852 		if (action != IPFSYNC_RESYNC)
3853 			continue;
3854 
3855 		if (fr->fr_type == FR_T_IPF) {
3856 			if (fr->fr_satype != FRI_NORMAL &&
3857 			    fr->fr_satype != FRI_LOOKUP) {
3858 				(void)fr_ifpaddr(rv, fr->fr_satype,
3859 						 fr->fr_ifas[fr->fr_sifpidx],
3860 						 &fr->fr_src, &fr->fr_smsk,
3861 						 ifs);
3862 			}
3863 			if (fr->fr_datype != FRI_NORMAL &&
3864 			    fr->fr_datype != FRI_LOOKUP) {
3865 				(void)fr_ifpaddr(rv, fr->fr_datype,
3866 						 fr->fr_ifas[fr->fr_difpidx],
3867 						 &fr->fr_dst, &fr->fr_dmsk,
3868 						 ifs);
3869 			}
3870 		}
3871 
3872 #ifdef	IPFILTER_LOOKUP
3873 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3874 		    fr->fr_srcptr == NULL) {
3875 			fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3876 							 fr->fr_srcnum,
3877 							 &fr->fr_srcfunc, ifs);
3878 		}
3879 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3880 		    fr->fr_dstptr == NULL) {
3881 			fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3882 							 fr->fr_dstnum,
3883 							 &fr->fr_dstfunc, ifs);
3884 		}
3885 #endif
3886 	}
3887 }
3888 
3889 
3890 #ifdef	_KERNEL
3891 /* ------------------------------------------------------------------------ */
3892 /* Function:    frsync                                                      */
3893 /* Returns:     void                                                        */
3894 /* Parameters:  action(I) - type of synchronisation to do                   */
3895 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3896 /*              ifp(I)    - interface identifier associated with action     */
3897 /*              name(I)   - name associated with ifp parameter              */
3898 /*                                                                          */
3899 /* frsync() is called when we suspect that the interface list or            */
3900 /* information about interfaces (like IP#) has changed.  Go through all     */
3901 /* filter rules, NAT entries and the state table and check if anything      */
3902 /* needs to be changed/updated.                                             */
3903 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3904 /* in which this was done to support three different types of sync:         */
3905 /* - complete resync of all interface name/identifiers                      */
3906 /* - new interface being announced with its name and identifier             */
3907 /* - interface removal being announced by only its identifier               */
3908 /* ------------------------------------------------------------------------ */
3909 void frsync(action, v, ifp, name, ifs)
3910 int action, v;
3911 void *ifp;
3912 char *name;
3913 ipf_stack_t *ifs;
3914 {
3915 	int i;
3916 
3917 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3918 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3919 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3920 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3921 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3922 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3923 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3924 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3925 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3926 
3927 	for (i = 0; i < IPL_LOGSIZE; i++) {
3928 		frgroup_t *g;
3929 
3930 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3931 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3932 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3933 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3934 	}
3935 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3936 }
3937 
3938 #if SOLARIS2 >= 10
3939 /* ------------------------------------------------------------------------ */
3940 /* Function:    fr_syncindex						    */
3941 /* Returns:     void							    */
3942 /* Parameters:  rules	  - list of rules to be sync'd			    */
3943 /*		ifp	  - interface, which is being sync'd		    */
3944 /*		newifp	  - new ifindex value for interface		    */
3945 /*                                                                          */
3946 /* Function updates all NIC indecis, which match ifp, in every rule. Every  */
3947 /* NIC index matching ifp, will be updated to newifp.			    */
3948 /* ------------------------------------------------------------------------ */
3949 static void fr_syncindex(rules, ifp, newifp)
3950 frentry_t *rules;
3951 void *ifp;
3952 void *newifp;
3953 {
3954 	int i;
3955 	frentry_t *fr;
3956 
3957 	for (fr = rules; fr != NULL; fr = fr->fr_next) {
3958 		/*
3959 		 * Lookup all the interface names that are part of the rule.
3960 		 */
3961 		for (i = 0; i < 4; i++)
3962 			if (fr->fr_ifas[i] == ifp)
3963 				fr->fr_ifas[i] = newifp;
3964 
3965 		for (i = 0; i < 2; i++) {
3966 			if (fr->fr_tifs[i].fd_ifp == ifp)
3967 				fr->fr_tifs[i].fd_ifp = newifp;
3968 		}
3969 
3970 		if (fr->fr_dif.fd_ifp == ifp)
3971 			fr->fr_dif.fd_ifp = newifp;
3972 	}
3973 }
3974 
3975 /* ------------------------------------------------------------------------ */
3976 /* Function:    fr_ifindexsync						    */
3977 /* Returns:     void							    */
3978 /* Parameters:	ifp	  - interface, which is being sync'd		    */
3979 /*		newifp	  - new ifindex value for interface		    */
3980 /*              ifs	  - IPF's stack					    */
3981 /*                                                                          */
3982 /* Function assumes ipf_mutex is locked exclusively.			    */
3983 /* 									    */
3984 /* Function updates the NIC references in rules with new interfaces index   */
3985 /* (newifp). Function must process active lists:			    */
3986 /*	with accounting rules (IPv6 and IPv4)				    */
3987 /*	with inbound rules (IPv6 and IPv4)				    */
3988 /*	with outbound rules (IPv6 and IPv4)				    */
3989 /* Function also has to take care of rule groups.			    */
3990 /*                                                                          */
3991 /* NOTE: The ipf_mutex is grabbed exclusively by caller (which is always    */
3992 /* nic_event_hook). The hook function also updates state entries, NAT rules */
3993 /* and NAT entries. We want to do all these update atomically to keep the   */
3994 /* NIC references consistent. The ipf_mutex will synchronize event with	    */
3995 /* fr_check(), which processes packets,	so no packet will enter fr_check(), */
3996 /* while NIC references will be synchronized.				    */
3997 /* ------------------------------------------------------------------------ */
3998 void fr_ifindexsync(ifp, newifp, ifs)
3999 void *ifp;
4000 void *newifp;
4001 ipf_stack_t *ifs;
4002 {
4003 	unsigned int	i;
4004 	frentry_t *rule_lists[8];
4005 	unsigned int	rules = sizeof (rule_lists) / sizeof (frentry_t *);
4006 
4007 	rule_lists[0] = ifs->ifs_ipacct[0][ifs->ifs_fr_active];
4008 	rule_lists[1] =	ifs->ifs_ipacct[1][ifs->ifs_fr_active];
4009 	rule_lists[2] =	ifs->ifs_ipfilter[0][ifs->ifs_fr_active];
4010 	rule_lists[3] =	ifs->ifs_ipfilter[1][ifs->ifs_fr_active];
4011 	rule_lists[4] =	ifs->ifs_ipacct6[0][ifs->ifs_fr_active];
4012 	rule_lists[5] =	ifs->ifs_ipacct6[1][ifs->ifs_fr_active];
4013 	rule_lists[6] =	ifs->ifs_ipfilter6[0][ifs->ifs_fr_active];
4014 	rule_lists[7] =	ifs->ifs_ipfilter6[1][ifs->ifs_fr_active];
4015 
4016 	for (i = 0; i < rules; i++) {
4017 		fr_syncindex(rule_lists[i], ifp, newifp);
4018 	}
4019 
4020 	/*
4021 	 * Update rule groups.
4022 	 */
4023 	for (i = 0; i < IPL_LOGSIZE; i++) {
4024 		frgroup_t *g;
4025 
4026 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
4027 			fr_syncindex(g->fg_start, ifp, newifp);
4028 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
4029 			fr_syncindex(g->fg_start, ifp, newifp);
4030 	}
4031 }
4032 #endif
4033 
4034 /*
4035  * In the functions below, bcopy() is called because the pointer being
4036  * copied _from_ in this instance is a pointer to a char buf (which could
4037  * end up being unaligned) and on the kernel's local stack.
4038  */
4039 /* ------------------------------------------------------------------------ */
4040 /* Function:    copyinptr                                                   */
4041 /* Returns:     int - 0 = success, else failure                             */
4042 /* Parameters:  src(I)  - pointer to the source address                     */
4043 /*              dst(I)  - destination address                               */
4044 /*              size(I) - number of bytes to copy                           */
4045 /*                                                                          */
4046 /* Copy a block of data in from user space, given a pointer to the pointer  */
4047 /* to start copying from (src) and a pointer to where to store it (dst).    */
4048 /* NB: src - pointer to user space pointer, dst - kernel space pointer      */
4049 /* ------------------------------------------------------------------------ */
4050 int copyinptr(src, dst, size)
4051 void *src, *dst;
4052 size_t size;
4053 {
4054 	caddr_t ca;
4055 	int err;
4056 
4057 # if SOLARIS
4058 	err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
4059 	if (err != 0)
4060 		return err;
4061 # else
4062 	bcopy(src, (caddr_t)&ca, sizeof(ca));
4063 # endif
4064 	err = COPYIN(ca, dst, size);
4065 	return err;
4066 }
4067 
4068 
4069 /* ------------------------------------------------------------------------ */
4070 /* Function:    copyoutptr                                                  */
4071 /* Returns:     int - 0 = success, else failure                             */
4072 /* Parameters:  src(I)  - pointer to the source address                     */
4073 /*              dst(I)  - destination address                               */
4074 /*              size(I) - number of bytes to copy                           */
4075 /*                                                                          */
4076 /* Copy a block of data out to user space, given a pointer to the pointer   */
4077 /* to start copying from (src) and a pointer to where to store it (dst).    */
4078 /* NB: src - kernel space pointer, dst - pointer to user space pointer.     */
4079 /* ------------------------------------------------------------------------ */
4080 int copyoutptr(src, dst, size)
4081 void *src, *dst;
4082 size_t size;
4083 {
4084 	caddr_t ca;
4085 	int err;
4086 
4087 # if SOLARIS
4088 	err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
4089 	if (err != 0)
4090 		return err;
4091 # else
4092 	bcopy(dst, (caddr_t)&ca, sizeof(ca));
4093 # endif
4094 	err = COPYOUT(src, ca, size);
4095 	return err;
4096 }
4097 #endif
4098 
4099 
4100 /* ------------------------------------------------------------------------ */
4101 /* Function:    fr_lock                                                     */
4102 /* Returns:	int - 0 = success, else error				    */
4103 /* Parameters:  data(I)  - pointer to lock value to set                     */
4104 /*              lockp(O) - pointer to location to store old lock value      */
4105 /*                                                                          */
4106 /* Get the new value for the lock integer, set it and return the old value  */
4107 /* in *lockp.                                                               */
4108 /* ------------------------------------------------------------------------ */
4109 int fr_lock(data, lockp)
4110 caddr_t data;
4111 int *lockp;
4112 {
4113 	int arg, err;
4114 
4115 	err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
4116 	if (err != 0)
4117 		return (EFAULT);
4118 	err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
4119 	if (err != 0)
4120 		return (EFAULT);
4121 	*lockp = arg;
4122 	return (0);
4123 }
4124 
4125 
4126 /* ------------------------------------------------------------------------ */
4127 /* Function:    fr_getstat                                                  */
4128 /* Returns:     Nil                                                         */
4129 /* Parameters:  fiop(I)  - pointer to ipfilter stats structure              */
4130 /*                                                                          */
4131 /* Stores a copy of current pointers, counters, etc, in the friostat        */
4132 /* structure.                                                               */
4133 /* ------------------------------------------------------------------------ */
4134 void fr_getstat(fiop, ifs)
4135 friostat_t *fiop;
4136 ipf_stack_t *ifs;
4137 {
4138 	int i, j;
4139 
4140 	bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
4141 	    sizeof(filterstats_t) * 2);
4142 	fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
4143 	fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
4144 	fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
4145 	fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
4146 
4147 	for (i = 0; i < 2; i++)
4148 		for (j = 0; j < 2; j++) {
4149 			fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4150 			fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4151 			fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4152 			fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4153 		}
4154 
4155 	fiop->f_ticks = ifs->ifs_fr_ticks;
4156 	fiop->f_active = ifs->ifs_fr_active;
4157 	fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4158 	fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4159 
4160 	fiop->f_running = ifs->ifs_fr_running;
4161 	for (i = 0; i < IPL_LOGSIZE; i++) {
4162 		fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4163 		fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4164 	}
4165 #ifdef  IPFILTER_LOG
4166 	fiop->f_logging = 1;
4167 #else
4168 	fiop->f_logging = 0;
4169 #endif
4170 	fiop->f_defpass = ifs->ifs_fr_pass;
4171 	fiop->f_features = fr_features;
4172 	(void) strncpy(fiop->f_version, ipfilter_version,
4173 		       sizeof(fiop->f_version));
4174 }
4175 
4176 
4177 #ifdef	USE_INET6
4178 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4179 	ICMP6_ECHO_REPLY,	/* 0: ICMP_ECHOREPLY */
4180 	-1,			/* 1: UNUSED */
4181 	-1,			/* 2: UNUSED */
4182 	ICMP6_DST_UNREACH,	/* 3: ICMP_UNREACH */
4183 	-1,			/* 4: ICMP_SOURCEQUENCH */
4184 	ND_REDIRECT,		/* 5: ICMP_REDIRECT */
4185 	-1,			/* 6: UNUSED */
4186 	-1,			/* 7: UNUSED */
4187 	ICMP6_ECHO_REQUEST,	/* 8: ICMP_ECHO */
4188 	-1,			/* 9: UNUSED */
4189 	-1,			/* 10: UNUSED */
4190 	ICMP6_TIME_EXCEEDED,	/* 11: ICMP_TIMXCEED */
4191 	ICMP6_PARAM_PROB,	/* 12: ICMP_PARAMPROB */
4192 	-1,			/* 13: ICMP_TSTAMP */
4193 	-1,			/* 14: ICMP_TSTAMPREPLY */
4194 	-1,			/* 15: ICMP_IREQ */
4195 	-1,			/* 16: ICMP_IREQREPLY */
4196 	-1,			/* 17: ICMP_MASKREQ */
4197 	-1,			/* 18: ICMP_MASKREPLY */
4198 };
4199 
4200 
4201 int	icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4202 	ICMP6_DST_UNREACH_ADDR,		/* 0: ICMP_UNREACH_NET */
4203 	ICMP6_DST_UNREACH_ADDR,		/* 1: ICMP_UNREACH_HOST */
4204 	-1,				/* 2: ICMP_UNREACH_PROTOCOL */
4205 	ICMP6_DST_UNREACH_NOPORT,	/* 3: ICMP_UNREACH_PORT */
4206 	-1,				/* 4: ICMP_UNREACH_NEEDFRAG */
4207 	ICMP6_DST_UNREACH_NOTNEIGHBOR,	/* 5: ICMP_UNREACH_SRCFAIL */
4208 	ICMP6_DST_UNREACH_ADDR,		/* 6: ICMP_UNREACH_NET_UNKNOWN */
4209 	ICMP6_DST_UNREACH_ADDR,		/* 7: ICMP_UNREACH_HOST_UNKNOWN */
4210 	-1,				/* 8: ICMP_UNREACH_ISOLATED */
4211 	ICMP6_DST_UNREACH_ADMIN,	/* 9: ICMP_UNREACH_NET_PROHIB */
4212 	ICMP6_DST_UNREACH_ADMIN,	/* 10: ICMP_UNREACH_HOST_PROHIB */
4213 	-1,				/* 11: ICMP_UNREACH_TOSNET */
4214 	-1,				/* 12: ICMP_UNREACH_TOSHOST */
4215 	ICMP6_DST_UNREACH_ADMIN,	/* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4216 };
4217 int	icmpreplytype6[ICMP6_MAXTYPE + 1];
4218 #endif
4219 
4220 int	icmpreplytype4[ICMP_MAXTYPE + 1];
4221 
4222 
4223 /* ------------------------------------------------------------------------ */
4224 /* Function:    fr_matchicmpqueryreply                                      */
4225 /* Returns:     int - 1 if "icmp" is a valid reply to "ic" else 0.          */
4226 /* Parameters:  v(I)    - IP protocol version (4 or 6)                      */
4227 /*              ic(I)   - ICMP information                                  */
4228 /*              icmp(I) - ICMP packet header                                */
4229 /*              rev(I)  - direction (0 = forward/1 = reverse) of packet     */
4230 /*                                                                          */
4231 /* Check if the ICMP packet defined by the header pointed to by icmp is a   */
4232 /* reply to one as described by what's in ic.  If it is a match, return 1,  */
4233 /* else return 0 for no match.                                              */
4234 /* ------------------------------------------------------------------------ */
4235 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4236 int v;
4237 icmpinfo_t *ic;
4238 icmphdr_t *icmp;
4239 int rev;
4240 {
4241 	int ictype;
4242 
4243 	ictype = ic->ici_type;
4244 
4245 	if (v == 4) {
4246 		/*
4247 		 * If we matched its type on the way in, then when going out
4248 		 * it will still be the same type.
4249 		 */
4250 		if ((!rev && (icmp->icmp_type == ictype)) ||
4251 		    (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4252 			if (icmp->icmp_type != ICMP_ECHOREPLY)
4253 				return 1;
4254 			if (icmp->icmp_id == ic->ici_id)
4255 				return 1;
4256 		}
4257 	}
4258 #ifdef	USE_INET6
4259 	else if (v == 6) {
4260 		if ((!rev && (icmp->icmp_type == ictype)) ||
4261 		    (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4262 			if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4263 				return 1;
4264 			if (icmp->icmp_id == ic->ici_id)
4265 				return 1;
4266 		}
4267 	}
4268 #endif
4269 	return 0;
4270 }
4271 
4272 
4273 #ifdef	IPFILTER_LOOKUP
4274 /* ------------------------------------------------------------------------ */
4275 /* Function:    fr_resolvelookup                                            */
4276 /* Returns:     void * - NULL = failure, else success.                      */
4277 /* Parameters:  type(I)     - type of lookup these parameters are for.      */
4278 /*              number(I)   - table number to use when searching            */
4279 /*              funcptr(IO) - pointer to pointer for storing IP address     */
4280 /*                           searching function.                            */
4281 /*                                                                          */
4282 /* Search for the "table" number passed in amongst those configured for     */
4283 /* that particular type.  If the type is recognised then the function to    */
4284 /* call to do the IP address search will be change, regardless of whether   */
4285 /* or not the "table" number exists.                                        */
4286 /* ------------------------------------------------------------------------ */
4287 static void *fr_resolvelookup(type, number, funcptr, ifs)
4288 u_int type, number;
4289 lookupfunc_t *funcptr;
4290 ipf_stack_t *ifs;
4291 {
4292 	char name[FR_GROUPLEN];
4293 	iphtable_t *iph;
4294 	ip_pool_t *ipo;
4295 	void *ptr;
4296 
4297 #if defined(SNPRINTF) && defined(_KERNEL)
4298 	(void) SNPRINTF(name, sizeof(name), "%u", number);
4299 #else
4300 	(void) sprintf(name, "%u", number);
4301 #endif
4302 
4303 	READ_ENTER(&ifs->ifs_ip_poolrw);
4304 
4305 	switch (type)
4306 	{
4307 	case IPLT_POOL :
4308 # if (defined(__osf__) && defined(_KERNEL))
4309 		ptr = NULL;
4310 		*funcptr = NULL;
4311 # else
4312 		ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4313 		ptr = ipo;
4314 		if (ipo != NULL) {
4315 			ATOMIC_INC32(ipo->ipo_ref);
4316 		}
4317 		*funcptr = ip_pool_search;
4318 # endif
4319 		break;
4320 	case IPLT_HASH :
4321 		iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4322 		ptr = iph;
4323 		if (iph != NULL) {
4324 			ATOMIC_INC32(iph->iph_ref);
4325 		}
4326 		*funcptr = fr_iphmfindip;
4327 		break;
4328 	default:
4329 		ptr = NULL;
4330 		*funcptr = NULL;
4331 		break;
4332 	}
4333 	RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4334 
4335 	return ptr;
4336 }
4337 #endif
4338 
4339 
4340 /* ------------------------------------------------------------------------ */
4341 /* Function:    frrequest                                                   */
4342 /* Returns:     int - 0 == success, > 0 == errno value                      */
4343 /* Parameters:  unit(I)     - device for which this is for                  */
4344 /*              req(I)      - ioctl command (SIOC*)                         */
4345 /*              data(I)     - pointr to ioctl data                          */
4346 /*              set(I)      - 1 or 0 (filter set)                           */
4347 /*              makecopy(I) - flag indicating whether data points to a rule */
4348 /*                            in kernel space & hence doesn't need copying. */
4349 /*                                                                          */
4350 /* This function handles all the requests which operate on the list of      */
4351 /* filter rules.  This includes adding, deleting, insertion.  It is also    */
4352 /* responsible for creating groups when a "head" rule is loaded.  Interface */
4353 /* names are resolved here and other sanity checks are made on the content  */
4354 /* of the rule structure being loaded.  If a rule has user defined timeouts */
4355 /* then make sure they are created and initialised before exiting.          */
4356 /* ------------------------------------------------------------------------ */
4357 int frrequest(unit, req, data, set, makecopy, ifs)
4358 int unit;
4359 ioctlcmd_t req;
4360 int set, makecopy;
4361 caddr_t data;
4362 ipf_stack_t *ifs;
4363 {
4364 	frentry_t frd, *fp, *f, **fprev, **ftail;
4365 	int error = 0, in, v;
4366 	void *ptr, *uptr;
4367 	u_int *p, *pp;
4368 	frgroup_t *fg;
4369 	char *group;
4370 
4371 	fg = NULL;
4372 	fp = &frd;
4373 	if (makecopy != 0) {
4374 		error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4375 		if (error)
4376 			return EFAULT;
4377 		if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4378 			return EINVAL;
4379 		fp->fr_ref = 0;
4380 		fp->fr_flags |= FR_COPIED;
4381 	} else {
4382 		fp = (frentry_t *)data;
4383 		if ((fp->fr_type & FR_T_BUILTIN) == 0)
4384 			return EINVAL;
4385 		fp->fr_flags &= ~FR_COPIED;
4386 	}
4387 
4388 	if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4389 	    ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4390 		return EINVAL;
4391 
4392 	v = fp->fr_v;
4393 	uptr = fp->fr_data;
4394 
4395 	/*
4396 	 * Only filter rules for IPv4 or IPv6 are accepted.
4397 	 */
4398 	if (v == 4)
4399 		/*EMPTY*/;
4400 #ifdef	USE_INET6
4401 	else if (v == 6)
4402 		/*EMPTY*/;
4403 #endif
4404 	else {
4405 		return EINVAL;
4406 	}
4407 
4408 	/*
4409 	 * If the rule is being loaded from user space, i.e. we had to copy it
4410 	 * into kernel space, then do not trust the function pointer in the
4411 	 * rule.
4412 	 */
4413 	if ((makecopy == 1) && (fp->fr_func != NULL)) {
4414 		if (fr_findfunc(fp->fr_func) == NULL)
4415 			return ESRCH;
4416 		error = fr_funcinit(fp, ifs);
4417 		if (error != 0)
4418 			return error;
4419 	}
4420 
4421 	ptr = NULL;
4422 	/*
4423 	 * Check that the group number does exist and that its use (in/out)
4424 	 * matches what the rule is.
4425 	 */
4426 	if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4427 		*fp->fr_grhead = '\0';
4428 	group = fp->fr_group;
4429 	if (!strncmp(group, "0", FR_GROUPLEN))
4430 		*group = '\0';
4431 
4432 	if (FR_ISACCOUNT(fp->fr_flags))
4433 		unit = IPL_LOGCOUNT;
4434 
4435 	if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4436 		fg = fr_findgroup(group, unit, set, NULL, ifs);
4437 		if (fg == NULL)
4438 			return ESRCH;
4439 		if (fg->fg_flags == 0)
4440 			fg->fg_flags = fp->fr_flags & FR_INOUT;
4441 		else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4442 			return ESRCH;
4443 	}
4444 
4445 	in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4446 
4447 	/*
4448 	 * Work out which rule list this change is being applied to.
4449 	 */
4450 	ftail = NULL;
4451 	fprev = NULL;
4452 	if (unit == IPL_LOGAUTH)
4453 		fprev = &ifs->ifs_ipauth;
4454 	else if (v == 4) {
4455 		if (FR_ISACCOUNT(fp->fr_flags))
4456 			fprev = &ifs->ifs_ipacct[in][set];
4457 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4458 			fprev = &ifs->ifs_ipfilter[in][set];
4459 	} else if (v == 6) {
4460 		if (FR_ISACCOUNT(fp->fr_flags))
4461 			fprev = &ifs->ifs_ipacct6[in][set];
4462 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4463 			fprev = &ifs->ifs_ipfilter6[in][set];
4464 	}
4465 	if (fprev == NULL)
4466 		return ESRCH;
4467 
4468 	if (*group != '\0') {
4469 	    if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4470 			return ESRCH;
4471 		fprev = &fg->fg_start;
4472 	}
4473 
4474 	ftail = fprev;
4475 	for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4476 		if (fp->fr_collect <= f->fr_collect) {
4477 			ftail = fprev;
4478 			f = NULL;
4479 			break;
4480 		}
4481 		fprev = ftail;
4482 	}
4483 
4484 	/*
4485 	 * Copy in extra data for the rule.
4486 	 */
4487 	if (fp->fr_dsize != 0) {
4488 		if (makecopy != 0) {
4489 			KMALLOCS(ptr, void *, fp->fr_dsize);
4490 			if (!ptr)
4491 				return ENOMEM;
4492 			error = COPYIN(uptr, ptr, fp->fr_dsize);
4493 		} else {
4494 			ptr = uptr;
4495 			error = 0;
4496 		}
4497 		if (error != 0) {
4498 			KFREES(ptr, fp->fr_dsize);
4499 			return EFAULT;
4500 		}
4501 		fp->fr_data = ptr;
4502 	} else
4503 		fp->fr_data = NULL;
4504 
4505 	/*
4506 	 * Perform per-rule type sanity checks of their members.
4507 	 */
4508 	switch (fp->fr_type & ~FR_T_BUILTIN)
4509 	{
4510 #if defined(IPFILTER_BPF)
4511 	case FR_T_BPFOPC :
4512 		if (fp->fr_dsize == 0)
4513 			return EINVAL;
4514 		if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4515 			if (makecopy && fp->fr_data != NULL) {
4516 				KFREES(fp->fr_data, fp->fr_dsize);
4517 			}
4518 			return EINVAL;
4519 		}
4520 		break;
4521 #endif
4522 	case FR_T_IPF :
4523 		if (fp->fr_dsize != sizeof(fripf_t)) {
4524 			if (makecopy && fp->fr_data != NULL) {
4525 				KFREES(fp->fr_data, fp->fr_dsize);
4526 			}
4527 			return EINVAL;
4528 		}
4529 
4530 		/*
4531 		 * Allowing a rule with both "keep state" and "with oow" is
4532 		 * pointless because adding a state entry to the table will
4533 		 * fail with the out of window (oow) flag set.
4534 		 */
4535 		if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW)) {
4536 			if (makecopy && fp->fr_data != NULL) {
4537 				KFREES(fp->fr_data, fp->fr_dsize);
4538 			}
4539 			return EINVAL;
4540 		}
4541 
4542 		switch (fp->fr_satype)
4543 		{
4544 		case FRI_BROADCAST :
4545 		case FRI_DYNAMIC :
4546 		case FRI_NETWORK :
4547 		case FRI_NETMASKED :
4548 		case FRI_PEERADDR :
4549 			if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4550 				if (makecopy && fp->fr_data != NULL) {
4551 					KFREES(fp->fr_data, fp->fr_dsize);
4552 				}
4553 				return EINVAL;
4554 			}
4555 			break;
4556 #ifdef	IPFILTER_LOOKUP
4557 		case FRI_LOOKUP :
4558 			fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4559 							 fp->fr_srcnum,
4560 							 &fp->fr_srcfunc, ifs);
4561 			break;
4562 #endif
4563 		default :
4564 			break;
4565 		}
4566 
4567 		switch (fp->fr_datype)
4568 		{
4569 		case FRI_BROADCAST :
4570 		case FRI_DYNAMIC :
4571 		case FRI_NETWORK :
4572 		case FRI_NETMASKED :
4573 		case FRI_PEERADDR :
4574 			if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4575 				if (makecopy && fp->fr_data != NULL) {
4576 					KFREES(fp->fr_data, fp->fr_dsize);
4577 				}
4578 				return EINVAL;
4579 			}
4580 			break;
4581 #ifdef	IPFILTER_LOOKUP
4582 		case FRI_LOOKUP :
4583 			fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4584 							 fp->fr_dstnum,
4585 							 &fp->fr_dstfunc, ifs);
4586 			break;
4587 #endif
4588 		default :
4589 			break;
4590 		}
4591 		break;
4592 	case FR_T_NONE :
4593 		break;
4594 	case FR_T_CALLFUNC :
4595 		break;
4596 	case FR_T_COMPIPF :
4597 		break;
4598 	default :
4599 		if (makecopy && fp->fr_data != NULL) {
4600 			KFREES(fp->fr_data, fp->fr_dsize);
4601 		}
4602 		return EINVAL;
4603 	}
4604 
4605 	/*
4606 	 * Lookup all the interface names that are part of the rule.
4607 	 */
4608 	frsynclist(0, 0, NULL, NULL, fp, ifs);
4609 	fp->fr_statecnt = 0;
4610 
4611 	/*
4612 	 * Look for an existing matching filter rule, but don't include the
4613 	 * next or interface pointer in the comparison (fr_next, fr_ifa).
4614 	 * This elminates rules which are indentical being loaded.  Checksum
4615 	 * the constant part of the filter rule to make comparisons quicker
4616 	 * (this meaning no pointers are included).
4617 	 */
4618 	for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4619 	     p < pp; p++)
4620 		fp->fr_cksum += *p;
4621 	pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4622 	for (p = (u_int *)fp->fr_data; p < pp; p++)
4623 		fp->fr_cksum += *p;
4624 
4625 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
4626 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
4627 
4628 	for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4629 		if ((fp->fr_cksum != f->fr_cksum) ||
4630 		    (f->fr_dsize != fp->fr_dsize))
4631 			continue;
4632 		if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4633 			continue;
4634 		if ((!ptr && !f->fr_data) ||
4635 		    (ptr && f->fr_data &&
4636 		     !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4637 			break;
4638 	}
4639 
4640 	/*
4641 	 * If zero'ing statistics, copy current to caller and zero.
4642 	 */
4643 	if (req == (ioctlcmd_t)SIOCZRLST) {
4644 		if (f == NULL)
4645 			error = ESRCH;
4646 		else {
4647 			/*
4648 			 * Copy and reduce lock because of impending copyout.
4649 			 * Well we should, but if we do then the atomicity of
4650 			 * this call and the correctness of fr_hits and
4651 			 * fr_bytes cannot be guaranteed.  As it is, this code
4652 			 * only resets them to 0 if they are successfully
4653 			 * copied out into user space.
4654 			 */
4655 			bcopy((char *)f, (char *)fp, sizeof(*f));
4656 
4657 			/*
4658 			 * When we copy this rule back out, set the data
4659 			 * pointer to be what it was in user space.
4660 			 */
4661 			fp->fr_data = uptr;
4662 			error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4663 
4664 			if (error == 0) {
4665 				if ((f->fr_dsize != 0) && (uptr != NULL))
4666 					error = COPYOUT(f->fr_data, uptr,
4667 							f->fr_dsize);
4668 				if (error == 0) {
4669 					f->fr_hits = 0;
4670 					f->fr_bytes = 0;
4671 				}
4672 			}
4673 		}
4674 
4675 		if ((ptr != NULL) && (makecopy != 0)) {
4676 			KFREES(ptr, fp->fr_dsize);
4677 		}
4678 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4679 		return error;
4680 	}
4681 
4682 	if (!f) {
4683 		/*
4684 		 * At the end of this, ftail must point to the place where the
4685 		 * new rule is to be saved/inserted/added.
4686 		 * For SIOCAD*FR, this should be the last rule in the group of
4687 		 * rules that have equal fr_collect fields.
4688 		 * For SIOCIN*FR, ...
4689 		 */
4690 		if (req == (ioctlcmd_t)SIOCADAFR ||
4691 		    req == (ioctlcmd_t)SIOCADIFR) {
4692 
4693 			for (ftail = fprev; (f = *ftail) != NULL; ) {
4694 				if (f->fr_collect > fp->fr_collect)
4695 					break;
4696 				ftail = &f->fr_next;
4697 			}
4698 			f = NULL;
4699 			ptr = NULL;
4700 			error = 0;
4701 		} else if (req == (ioctlcmd_t)SIOCINAFR ||
4702 			   req == (ioctlcmd_t)SIOCINIFR) {
4703 			while ((f = *fprev) != NULL) {
4704 				if (f->fr_collect >= fp->fr_collect)
4705 					break;
4706 				fprev = &f->fr_next;
4707 			}
4708 			ftail = fprev;
4709 			if (fp->fr_hits != 0) {
4710 				while (fp->fr_hits && (f = *ftail)) {
4711 					if (f->fr_collect != fp->fr_collect)
4712 						break;
4713 					fprev = ftail;
4714 					ftail = &f->fr_next;
4715 					fp->fr_hits--;
4716 				}
4717 			}
4718 			f = NULL;
4719 			ptr = NULL;
4720 			error = 0;
4721 		}
4722 	}
4723 
4724 	/*
4725 	 * Request to remove a rule.
4726 	 */
4727 	if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4728 		if (!f)
4729 			error = ESRCH;
4730 		else {
4731 			/*
4732 			 * Do not allow activity from user space to interfere
4733 			 * with rules not loaded that way.
4734 			 */
4735 			if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4736 				error = EPERM;
4737 				goto done;
4738 			}
4739 
4740 			/*
4741 			 * Return EBUSY if the rule is being reference by
4742 			 * something else (eg state information.
4743 			 */
4744 			if (f->fr_ref > 1) {
4745 				error = EBUSY;
4746 				goto done;
4747 			}
4748 #ifdef	IPFILTER_SCAN
4749 			if (f->fr_isctag[0] != '\0' &&
4750 			    (f->fr_isc != (struct ipscan *)-1))
4751 				ipsc_detachfr(f);
4752 #endif
4753 			if (unit == IPL_LOGAUTH) {
4754 				error = fr_preauthcmd(req, f, ftail, ifs);
4755 				goto done;
4756 			}
4757 			if (*f->fr_grhead != '\0')
4758 				fr_delgroup(f->fr_grhead, unit, set, ifs);
4759 			fr_fixskip(ftail, f, -1);
4760 			*ftail = f->fr_next;
4761 			f->fr_next = NULL;
4762 			(void)fr_derefrule(&f, ifs);
4763 		}
4764 	} else {
4765 		/*
4766 		 * Not removing, so we must be adding/inserting a rule.
4767 		 */
4768 		if (f)
4769 			error = EEXIST;
4770 		else {
4771 			if (unit == IPL_LOGAUTH) {
4772 				error = fr_preauthcmd(req, fp, ftail, ifs);
4773 				goto done;
4774 			}
4775 			if (makecopy) {
4776 				KMALLOC(f, frentry_t *);
4777 			} else
4778 				f = fp;
4779 			if (f != NULL) {
4780 				if (fp != f)
4781 					bcopy((char *)fp, (char *)f,
4782 					      sizeof(*f));
4783 				MUTEX_NUKE(&f->fr_lock);
4784 				MUTEX_INIT(&f->fr_lock, "filter rule lock");
4785 #ifdef	IPFILTER_SCAN
4786 				if (f->fr_isctag[0] != '\0' &&
4787 				    ipsc_attachfr(f))
4788 					f->fr_isc = (struct ipscan *)-1;
4789 #endif
4790 				f->fr_hits = 0;
4791 				if (makecopy != 0)
4792 					f->fr_ref = 1;
4793 				f->fr_next = *ftail;
4794 				*ftail = f;
4795 				if (req == (ioctlcmd_t)SIOCINIFR ||
4796 				    req == (ioctlcmd_t)SIOCINAFR)
4797 					fr_fixskip(ftail, f, 1);
4798 				f->fr_grp = NULL;
4799 				group = f->fr_grhead;
4800 				if (*group != '\0') {
4801 					fg = fr_addgroup(group, f, f->fr_flags,
4802 							 unit, set, ifs);
4803 					if (fg != NULL)
4804 						f->fr_grp = &fg->fg_start;
4805 				}
4806 			} else
4807 				error = ENOMEM;
4808 		}
4809 	}
4810 done:
4811 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4812 	if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4813 		KFREES(ptr, fp->fr_dsize);
4814 	}
4815 	return (error);
4816 }
4817 
4818 
4819 /* ------------------------------------------------------------------------ */
4820 /* Function:    fr_funcinit                                                 */
4821 /* Returns:     int - 0 == success, else ESRCH: cannot resolve rule details */
4822 /* Parameters:  fr(I) - pointer to filter rule                              */
4823 /*                                                                          */
4824 /* If a rule is a call rule, then check if the function it points to needs  */
4825 /* an init function to be called now the rule has been loaded.              */
4826 /* ------------------------------------------------------------------------ */
4827 static int fr_funcinit(fr, ifs)
4828 frentry_t *fr;
4829 ipf_stack_t *ifs;
4830 {
4831 	ipfunc_resolve_t *ft;
4832 	int err;
4833 
4834 	err = ESRCH;
4835 
4836 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4837 		if (ft->ipfu_addr == fr->fr_func) {
4838 			err = 0;
4839 			if (ft->ipfu_init != NULL)
4840 				err = (*ft->ipfu_init)(fr, ifs);
4841 			break;
4842 		}
4843 	return err;
4844 }
4845 
4846 
4847 /* ------------------------------------------------------------------------ */
4848 /* Function:    fr_findfunc                                                 */
4849 /* Returns:     ipfunc_t - pointer to function if found, else NULL          */
4850 /* Parameters:  funcptr(I) - function pointer to lookup                     */
4851 /*                                                                          */
4852 /* Look for a function in the table of known functions.                     */
4853 /* ------------------------------------------------------------------------ */
4854 static ipfunc_t fr_findfunc(funcptr)
4855 ipfunc_t funcptr;
4856 {
4857 	ipfunc_resolve_t *ft;
4858 
4859 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4860 		if (ft->ipfu_addr == funcptr)
4861 			return funcptr;
4862 	return NULL;
4863 }
4864 
4865 
4866 /* ------------------------------------------------------------------------ */
4867 /* Function:    fr_resolvefunc                                              */
4868 /* Returns:     int - 0 == success, else error                              */
4869 /* Parameters:  data(IO) - ioctl data pointer to ipfunc_resolve_t struct    */
4870 /*                                                                          */
4871 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4872 /* This will either be the function name (if the pointer is set) or the     */
4873 /* function pointer if the name is set.  When found, fill in the other one  */
4874 /* so that the entire, complete, structure can be copied back to user space.*/
4875 /* ------------------------------------------------------------------------ */
4876 int fr_resolvefunc(data)
4877 void *data;
4878 {
4879 	ipfunc_resolve_t res, *ft;
4880 	int err;
4881 
4882 	err = BCOPYIN(data, &res, sizeof(res));
4883 	if (err != 0)
4884 		return EFAULT;
4885 
4886 	if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4887 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4888 			if (strncmp(res.ipfu_name, ft->ipfu_name,
4889 				    sizeof(res.ipfu_name)) == 0) {
4890 				res.ipfu_addr = ft->ipfu_addr;
4891 				res.ipfu_init = ft->ipfu_init;
4892 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4893 					return EFAULT;
4894 				return 0;
4895 			}
4896 	}
4897 	if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4898 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4899 			if (ft->ipfu_addr == res.ipfu_addr) {
4900 				(void) strncpy(res.ipfu_name, ft->ipfu_name,
4901 					       sizeof(res.ipfu_name));
4902 				res.ipfu_init = ft->ipfu_init;
4903 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4904 					return EFAULT;
4905 				return 0;
4906 			}
4907 	}
4908 	return ESRCH;
4909 }
4910 
4911 
4912 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4913     (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4914     (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4915     (defined(__OpenBSD__) && (OpenBSD < 200006))
4916 /*
4917  * From: NetBSD
4918  * ppsratecheck(): packets (or events) per second limitation.
4919  */
4920 int
4921 ppsratecheck(lasttime, curpps, maxpps)
4922 	struct timeval *lasttime;
4923 	int *curpps;
4924 	int maxpps;	/* maximum pps allowed */
4925 {
4926 	struct timeval tv, delta;
4927 	int rv;
4928 
4929 	GETKTIME(&tv);
4930 
4931 	delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4932 	delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4933 	if (delta.tv_usec < 0) {
4934 		delta.tv_sec--;
4935 		delta.tv_usec += 1000000;
4936 	}
4937 
4938 	/*
4939 	 * check for 0,0 is so that the message will be seen at least once.
4940 	 * if more than one second have passed since the last update of
4941 	 * lasttime, reset the counter.
4942 	 *
4943 	 * we do increment *curpps even in *curpps < maxpps case, as some may
4944 	 * try to use *curpps for stat purposes as well.
4945 	 */
4946 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4947 	    delta.tv_sec >= 1) {
4948 		*lasttime = tv;
4949 		*curpps = 0;
4950 		rv = 1;
4951 	} else if (maxpps < 0)
4952 		rv = 1;
4953 	else if (*curpps < maxpps)
4954 		rv = 1;
4955 	else
4956 		rv = 0;
4957 	*curpps = *curpps + 1;
4958 
4959 	return (rv);
4960 }
4961 #endif
4962 
4963 
4964 /* ------------------------------------------------------------------------ */
4965 /* Function:    fr_derefrule                                                */
4966 /* Returns:     int   - 0 == rule freed up, else rule not freed             */
4967 /* Parameters:  fr(I) - pointer to filter rule                              */
4968 /*                                                                          */
4969 /* Decrement the reference counter to a rule by one.  If it reaches zero,   */
4970 /* free it and any associated storage space being used by it.               */
4971 /* ------------------------------------------------------------------------ */
4972 int fr_derefrule(frp, ifs)
4973 frentry_t **frp;
4974 ipf_stack_t *ifs;
4975 {
4976 	frentry_t *fr;
4977 
4978 	fr = *frp;
4979 
4980 	MUTEX_ENTER(&fr->fr_lock);
4981 	fr->fr_ref--;
4982 	if (fr->fr_ref == 0) {
4983 		MUTEX_EXIT(&fr->fr_lock);
4984 		MUTEX_DESTROY(&fr->fr_lock);
4985 
4986 #ifdef IPFILTER_LOOKUP
4987 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
4988 		    ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
4989 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
4990 		    ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
4991 #endif
4992 
4993 		if (fr->fr_dsize) {
4994 			KFREES(fr->fr_data, fr->fr_dsize);
4995 		}
4996 		if ((fr->fr_flags & FR_COPIED) != 0) {
4997 			KFREE(fr);
4998 			return 0;
4999 		}
5000 		return 1;
5001 	} else {
5002 		MUTEX_EXIT(&fr->fr_lock);
5003 	}
5004 	*frp = NULL;
5005 	return -1;
5006 }
5007 
5008 
5009 #ifdef	IPFILTER_LOOKUP
5010 /* ------------------------------------------------------------------------ */
5011 /* Function:    fr_grpmapinit                                               */
5012 /* Returns:     int - 0 == success, else ESRCH because table entry not found*/
5013 /* Parameters:  fr(I) - pointer to rule to find hash table for              */
5014 /*                                                                          */
5015 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr.  */
5016 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap.                   */
5017 /* ------------------------------------------------------------------------ */
5018 static int fr_grpmapinit(fr, ifs)
5019 frentry_t *fr;
5020 ipf_stack_t *ifs;
5021 {
5022 	char name[FR_GROUPLEN];
5023 	iphtable_t *iph;
5024 
5025 #if defined(SNPRINTF) && defined(_KERNEL)
5026 	(void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
5027 #else
5028 	(void) sprintf(name, "%d", fr->fr_arg);
5029 #endif
5030 	iph = fr_findhtable(IPL_LOGIPF, name, ifs);
5031 	if (iph == NULL)
5032 		return ESRCH;
5033 	if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
5034 		return ESRCH;
5035 	fr->fr_ptr = iph;
5036 	return 0;
5037 }
5038 
5039 
5040 /* ------------------------------------------------------------------------ */
5041 /* Function:    fr_srcgrpmap                                                */
5042 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5043 /* Parameters:  fin(I)    - pointer to packet information                   */
5044 /*              passp(IO) - pointer to current/new filter decision (unused) */
5045 /*                                                                          */
5046 /* Look for a rule group head in a hash table, using the source address as  */
5047 /* the key, and descend into that group and continue matching rules against */
5048 /* the packet.                                                              */
5049 /* ------------------------------------------------------------------------ */
5050 frentry_t *fr_srcgrpmap(fin, passp)
5051 fr_info_t *fin;
5052 u_32_t *passp;
5053 {
5054 	frgroup_t *fg;
5055 	void *rval;
5056 	ipf_stack_t *ifs = fin->fin_ifs;
5057 
5058 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
5059 	if (rval == NULL)
5060 		return NULL;
5061 
5062 	fg = rval;
5063 	fin->fin_fr = fg->fg_start;
5064 	(void) fr_scanlist(fin, *passp);
5065 	return fin->fin_fr;
5066 }
5067 
5068 
5069 /* ------------------------------------------------------------------------ */
5070 /* Function:    fr_dstgrpmap                                                */
5071 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5072 /* Parameters:  fin(I)    - pointer to packet information                   */
5073 /*              passp(IO) - pointer to current/new filter decision (unused) */
5074 /*                                                                          */
5075 /* Look for a rule group head in a hash table, using the destination        */
5076 /* address as the key, and descend into that group and continue matching    */
5077 /* rules against  the packet.                                               */
5078 /* ------------------------------------------------------------------------ */
5079 frentry_t *fr_dstgrpmap(fin, passp)
5080 fr_info_t *fin;
5081 u_32_t *passp;
5082 {
5083 	frgroup_t *fg;
5084 	void *rval;
5085 	ipf_stack_t *ifs = fin->fin_ifs;
5086 
5087 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
5088 	if (rval == NULL)
5089 		return NULL;
5090 
5091 	fg = rval;
5092 	fin->fin_fr = fg->fg_start;
5093 	(void) fr_scanlist(fin, *passp);
5094 	return fin->fin_fr;
5095 }
5096 #endif /* IPFILTER_LOOKUP */
5097 
5098 /*
5099  * Queue functions
5100  * ===============
5101  * These functions manage objects on queues for efficient timeouts.  There are
5102  * a number of system defined queues as well as user defined timeouts.  It is
5103  * expected that a lock is held in the domain in which the queue belongs
5104  * (i.e. either state or NAT) when calling any of these functions that prevents
5105  * fr_freetimeoutqueue() from being called at the same time as any other.
5106  */
5107 
5108 
5109 /* ------------------------------------------------------------------------ */
5110 /* Function:    fr_addtimeoutqueue                                          */
5111 /* Returns:     struct ifqtq * - NULL if malloc fails, else pointer to      */
5112 /*                               timeout queue with given interval.         */
5113 /* Parameters:  parent(I)  - pointer to pointer to parent node of this list */
5114 /*                           of interface queues.                           */
5115 /*              seconds(I) - timeout value in seconds for this queue.       */
5116 /*                                                                          */
5117 /* This routine first looks for a timeout queue that matches the interval   */
5118 /* being requested.  If it finds one, increments the reference counter and  */
5119 /* returns a pointer to it.  If none are found, it allocates a new one and  */
5120 /* inserts it at the top of the list.                                       */
5121 /*                                                                          */
5122 /* Locking.                                                                 */
5123 /* It is assumed that the caller of this function has an appropriate lock   */
5124 /* held (exclusively) in the domain that encompases 'parent'.               */
5125 /* ------------------------------------------------------------------------ */
5126 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
5127 ipftq_t **parent;
5128 u_int seconds;
5129 ipf_stack_t *ifs;
5130 {
5131 	ipftq_t *ifq;
5132 	u_int period;
5133 
5134 	period = seconds * IPF_HZ_DIVIDE;
5135 
5136 	MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
5137 	for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
5138 		if (ifq->ifq_ttl == period) {
5139 			/*
5140 			 * Reset the delete flag, if set, so the structure
5141 			 * gets reused rather than freed and reallocated.
5142 			 */
5143 			MUTEX_ENTER(&ifq->ifq_lock);
5144 			ifq->ifq_flags &= ~IFQF_DELETE;
5145 			ifq->ifq_ref++;
5146 			MUTEX_EXIT(&ifq->ifq_lock);
5147 			MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5148 
5149 			return ifq;
5150 		}
5151 	}
5152 
5153 	KMALLOC(ifq, ipftq_t *);
5154 	if (ifq != NULL) {
5155 		ifq->ifq_ttl = period;
5156 		ifq->ifq_head = NULL;
5157 		ifq->ifq_tail = &ifq->ifq_head;
5158 		ifq->ifq_next = *parent;
5159 		ifq->ifq_pnext = parent;
5160 		ifq->ifq_ref = 1;
5161 		ifq->ifq_flags = IFQF_USER;
5162 		*parent = ifq;
5163 		ifs->ifs_fr_userifqs++;
5164 		MUTEX_NUKE(&ifq->ifq_lock);
5165 		MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5166 	}
5167 	MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5168 	return ifq;
5169 }
5170 
5171 
5172 /* ------------------------------------------------------------------------ */
5173 /* Function:    fr_deletetimeoutqueue                                       */
5174 /* Returns:     int    - new reference count value of the timeout queue     */
5175 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5176 /* Locks:       ifq->ifq_lock                                               */
5177 /*                                                                          */
5178 /* This routine must be called when we're discarding a pointer to a timeout */
5179 /* queue object, taking care of the reference counter.                      */
5180 /*                                                                          */
5181 /* Now that this just sets a DELETE flag, it requires the expire code to    */
5182 /* check the list of user defined timeout queues and call the free function */
5183 /* below (currently commented out) to stop memory leaking.  It is done this */
5184 /* way because the locking may not be sufficient to safely do a free when   */
5185 /* this function is called.                                                 */
5186 /* ------------------------------------------------------------------------ */
5187 int fr_deletetimeoutqueue(ifq)
5188 ipftq_t *ifq;
5189 {
5190 
5191 	ifq->ifq_ref--;
5192 	if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5193 		ifq->ifq_flags |= IFQF_DELETE;
5194 	}
5195 
5196 	return ifq->ifq_ref;
5197 }
5198 
5199 
5200 /* ------------------------------------------------------------------------ */
5201 /* Function:    fr_freetimeoutqueue                                         */
5202 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5203 /* Returns:     Nil                                                         */
5204 /*                                                                          */
5205 /* Locking:                                                                 */
5206 /* It is assumed that the caller of this function has an appropriate lock   */
5207 /* held (exclusively) in the domain that encompases the callers "domain".   */
5208 /* The ifq_lock for this structure should not be held.                      */
5209 /*                                                                          */
5210 /* Remove a user definde timeout queue from the list of queues it is in and */
5211 /* tidy up after this is done.                                              */
5212 /* ------------------------------------------------------------------------ */
5213 void fr_freetimeoutqueue(ifq, ifs)
5214 ipftq_t *ifq;
5215 ipf_stack_t *ifs;
5216 {
5217 
5218 
5219 	if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5220 	    ((ifq->ifq_flags & IFQF_USER) == 0)) {
5221 		printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5222 		       (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5223 		       ifq->ifq_ref);
5224 		return;
5225 	}
5226 
5227 	/*
5228 	 * Remove from its position in the list.
5229 	 */
5230 	*ifq->ifq_pnext = ifq->ifq_next;
5231 	if (ifq->ifq_next != NULL)
5232 		ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5233 
5234 	MUTEX_DESTROY(&ifq->ifq_lock);
5235 	ifs->ifs_fr_userifqs--;
5236 	KFREE(ifq);
5237 }
5238 
5239 
5240 /* ------------------------------------------------------------------------ */
5241 /* Function:    fr_deletequeueentry                                         */
5242 /* Returns:     Nil                                                         */
5243 /* Parameters:  tqe(I) - timeout queue entry to delete                      */
5244 /*              ifq(I) - timeout queue to remove entry from                 */
5245 /*                                                                          */
5246 /* Remove a tail queue entry from its queue and make it an orphan.          */
5247 /* fr_deletetimeoutqueue is called to make sure the reference count on the  */
5248 /* queue is correct.  We can't, however, call fr_freetimeoutqueue because   */
5249 /* the correct lock(s) may not be held that would make it safe to do so.    */
5250 /* ------------------------------------------------------------------------ */
5251 void fr_deletequeueentry(tqe)
5252 ipftqent_t *tqe;
5253 {
5254 	ipftq_t *ifq;
5255 
5256 	ifq = tqe->tqe_ifq;
5257 	if (ifq == NULL)
5258 		return;
5259 
5260 	MUTEX_ENTER(&ifq->ifq_lock);
5261 
5262 	if (tqe->tqe_pnext != NULL) {
5263 		*tqe->tqe_pnext = tqe->tqe_next;
5264 		if (tqe->tqe_next != NULL)
5265 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5266 		else    /* we must be the tail anyway */
5267 			ifq->ifq_tail = tqe->tqe_pnext;
5268 
5269 		tqe->tqe_pnext = NULL;
5270 		tqe->tqe_ifq = NULL;
5271 	}
5272 
5273 	(void) fr_deletetimeoutqueue(ifq);
5274 
5275 	MUTEX_EXIT(&ifq->ifq_lock);
5276 }
5277 
5278 
5279 /* ------------------------------------------------------------------------ */
5280 /* Function:    fr_queuefront                                               */
5281 /* Returns:     Nil                                                         */
5282 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5283 /*                                                                          */
5284 /* Move a queue entry to the front of the queue, if it isn't already there. */
5285 /* ------------------------------------------------------------------------ */
5286 void fr_queuefront(tqe)
5287 ipftqent_t *tqe;
5288 {
5289 	ipftq_t *ifq;
5290 
5291 	ifq = tqe->tqe_ifq;
5292 	if (ifq == NULL)
5293 		return;
5294 
5295 	MUTEX_ENTER(&ifq->ifq_lock);
5296 	if (ifq->ifq_head != tqe) {
5297 		*tqe->tqe_pnext = tqe->tqe_next;
5298 		if (tqe->tqe_next)
5299 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5300 		else
5301 			ifq->ifq_tail = tqe->tqe_pnext;
5302 
5303 		tqe->tqe_next = ifq->ifq_head;
5304 		ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5305 		ifq->ifq_head = tqe;
5306 		tqe->tqe_pnext = &ifq->ifq_head;
5307 	}
5308 	MUTEX_EXIT(&ifq->ifq_lock);
5309 }
5310 
5311 
5312 /* ------------------------------------------------------------------------ */
5313 /* Function:    fr_queueback                                                */
5314 /* Returns:     Nil                                                         */
5315 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5316 /*                                                                          */
5317 /* Move a queue entry to the back of the queue, if it isn't already there.  */
5318 /* ------------------------------------------------------------------------ */
5319 void fr_queueback(tqe, ifs)
5320 ipftqent_t *tqe;
5321 ipf_stack_t *ifs;
5322 {
5323 	ipftq_t *ifq;
5324 
5325 	ifq = tqe->tqe_ifq;
5326 	if (ifq == NULL)
5327 		return;
5328 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5329 
5330 	MUTEX_ENTER(&ifq->ifq_lock);
5331 	if (tqe->tqe_next == NULL) {		/* at the end already ? */
5332 		MUTEX_EXIT(&ifq->ifq_lock);
5333 		return;
5334 	}
5335 
5336 	/*
5337 	 * Remove from list
5338 	 */
5339 	*tqe->tqe_pnext = tqe->tqe_next;
5340 	tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5341 
5342 	/*
5343 	 * Make it the last entry.
5344 	 */
5345 	tqe->tqe_next = NULL;
5346 	tqe->tqe_pnext = ifq->ifq_tail;
5347 	*ifq->ifq_tail = tqe;
5348 	ifq->ifq_tail = &tqe->tqe_next;
5349 	MUTEX_EXIT(&ifq->ifq_lock);
5350 }
5351 
5352 
5353 /* ------------------------------------------------------------------------ */
5354 /* Function:    fr_queueappend                                              */
5355 /* Returns:     Nil                                                         */
5356 /* Parameters:  tqe(I)    - pointer to timeout queue entry                  */
5357 /*              ifq(I)    - pointer to timeout queue                        */
5358 /*              parent(I) - owing object pointer                            */
5359 /*                                                                          */
5360 /* Add a new item to this queue and put it on the very end.                 */
5361 /* ------------------------------------------------------------------------ */
5362 void fr_queueappend(tqe, ifq, parent, ifs)
5363 ipftqent_t *tqe;
5364 ipftq_t *ifq;
5365 void *parent;
5366 ipf_stack_t *ifs;
5367 {
5368 
5369 	MUTEX_ENTER(&ifq->ifq_lock);
5370 	tqe->tqe_parent = parent;
5371 	tqe->tqe_pnext = ifq->ifq_tail;
5372 	*ifq->ifq_tail = tqe;
5373 	ifq->ifq_tail = &tqe->tqe_next;
5374 	tqe->tqe_next = NULL;
5375 	tqe->tqe_ifq = ifq;
5376 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5377 	ifq->ifq_ref++;
5378 	MUTEX_EXIT(&ifq->ifq_lock);
5379 }
5380 
5381 
5382 /* ------------------------------------------------------------------------ */
5383 /* Function:    fr_movequeue                                                */
5384 /* Returns:     Nil                                                         */
5385 /* Parameters:  tq(I)   - pointer to timeout queue information              */
5386 /*              oifp(I) - old timeout queue entry was on                    */
5387 /*              nifp(I) - new timeout queue to put entry on                 */
5388 /*		ifs	- ipf stack instance				    */
5389 /*                                                                          */
5390 /* Move a queue entry from one timeout queue to another timeout queue.      */
5391 /* If it notices that the current entry is already last and does not need   */
5392 /* to move queue, the return.                                               */
5393 /* ------------------------------------------------------------------------ */
5394 void fr_movequeue(tqe, oifq, nifq, ifs)
5395 ipftqent_t *tqe;
5396 ipftq_t *oifq, *nifq;
5397 ipf_stack_t *ifs;
5398 {
5399 	/*
5400 	 * If the queue isn't changing, and the clock hasn't ticked
5401 	 * since the last update, the operation will be a no-op.
5402 	 */
5403 	if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5404 		return;
5405 
5406 	/*
5407 	 * Grab the lock and update the timers.
5408 	 */
5409 	MUTEX_ENTER(&oifq->ifq_lock);
5410 	tqe->tqe_touched = ifs->ifs_fr_ticks;
5411 	tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5412 
5413 	/*
5414 	 * The remainder of the operation can still be a no-op.
5415 	 *
5416 	 * If the queue isn't changing, check to see if
5417 	 * an update would be meaningless.
5418 	 */
5419 	if (oifq == nifq) {
5420 		if ((tqe->tqe_next == NULL) ||
5421 		    (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5422 			MUTEX_EXIT(&oifq->ifq_lock);
5423 			return;
5424 		}
5425 	}
5426 
5427 	/*
5428 	 * Remove from the old queue
5429 	 */
5430 	*tqe->tqe_pnext = tqe->tqe_next;
5431 	if (tqe->tqe_next)
5432 		tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5433 	else
5434 		oifq->ifq_tail = tqe->tqe_pnext;
5435 	tqe->tqe_next = NULL;
5436 
5437 	/*
5438 	 * If we're moving from one queue to another, release the lock on the
5439 	 * old queue and get a lock on the new queue.  For user defined queues,
5440 	 * if we're moving off it, call delete in case it can now be freed.
5441 	 */
5442 	if (oifq != nifq) {
5443 		tqe->tqe_ifq = NULL;
5444 
5445 		(void) fr_deletetimeoutqueue(oifq);
5446 
5447 		MUTEX_EXIT(&oifq->ifq_lock);
5448 
5449 		MUTEX_ENTER(&nifq->ifq_lock);
5450 
5451 		tqe->tqe_ifq = nifq;
5452 		nifq->ifq_ref++;
5453 	}
5454 
5455 	/*
5456 	 * Add to the bottom of the new queue
5457 	 */
5458 	tqe->tqe_pnext = nifq->ifq_tail;
5459 	*nifq->ifq_tail = tqe;
5460 	nifq->ifq_tail = &tqe->tqe_next;
5461 	MUTEX_EXIT(&nifq->ifq_lock);
5462 }
5463 
5464 
5465 /* ------------------------------------------------------------------------ */
5466 /* Function:    fr_updateipid                                               */
5467 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
5468 /* Parameters:  fin(I) - pointer to packet information                      */
5469 /*                                                                          */
5470 /* When we are doing NAT, change the IP of every packet to represent a      */
5471 /* single sequence of packets coming from the host, hiding any host         */
5472 /* specific sequencing that might otherwise be revealed.  If the packet is  */
5473 /* a fragment, then store the 'new' IPid in the fragment cache and look up  */
5474 /* the fragment cache for non-leading fragments.  If a non-leading fragment */
5475 /* has no match in the cache, return an error.                              */
5476 /* ------------------------------------------------------------------------ */
5477 static INLINE int fr_updateipid(fin)
5478 fr_info_t *fin;
5479 {
5480 	u_short id, ido, sums;
5481 	u_32_t sumd, sum;
5482 	ip_t *ip;
5483 
5484 	if (fin->fin_off != 0) {
5485 		sum = fr_ipid_knownfrag(fin);
5486 		if (sum == 0xffffffff)
5487 			return -1;
5488 		sum &= 0xffff;
5489 		id = (u_short)sum;
5490 	} else {
5491 		id = fr_nextipid(fin);
5492 		if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5493 			(void) fr_ipid_newfrag(fin, (u_32_t)id);
5494 	}
5495 
5496 	ip = fin->fin_ip;
5497 	ido = ntohs(ip->ip_id);
5498 	if (id == ido)
5499 		return 0;
5500 	ip->ip_id = htons(id);
5501 	CALC_SUMD(ido, id, sumd);	/* DESTRUCTIVE MACRO! id,ido change */
5502 	sum = (~ntohs(ip->ip_sum)) & 0xffff;
5503 	sum += sumd;
5504 	sum = (sum >> 16) + (sum & 0xffff);
5505 	sum = (sum >> 16) + (sum & 0xffff);
5506 	sums = ~(u_short)sum;
5507 	ip->ip_sum = htons(sums);
5508 	return 0;
5509 }
5510 
5511 
5512 #ifdef	NEED_FRGETIFNAME
5513 /* ------------------------------------------------------------------------ */
5514 /* Function:    fr_getifname                                                */
5515 /* Returns:     char *    - pointer to interface name                       */
5516 /* Parameters:  ifp(I)    - pointer to network interface                    */
5517 /*              buffer(O) - pointer to where to store interface name        */
5518 /*                                                                          */
5519 /* Constructs an interface name in the buffer passed.  The buffer passed is */
5520 /* expected to be at least LIFNAMSIZ in bytes big.  If buffer is passed in  */
5521 /* as a NULL pointer then return a pointer to a static array.               */
5522 /* ------------------------------------------------------------------------ */
5523 char *fr_getifname(ifp, buffer)
5524 struct ifnet *ifp;
5525 char *buffer;
5526 {
5527 	static char namebuf[LIFNAMSIZ];
5528 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5529      defined(__sgi) || defined(linux) || defined(_AIX51) || \
5530      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5531 	int unit, space;
5532 	char temp[20];
5533 	char *s;
5534 # endif
5535 
5536 	ASSERT(buffer != NULL);
5537 #ifdef notdef
5538 	if (buffer == NULL)
5539 		buffer = namebuf;
5540 #endif
5541 	(void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5542 	buffer[LIFNAMSIZ - 1] = '\0';
5543 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5544      defined(__sgi) || defined(_AIX51) || \
5545      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5546 	for (s = buffer; *s; s++)
5547 		;
5548 	unit = ifp->if_unit;
5549 	space = LIFNAMSIZ - (s - buffer);
5550 	if (space > 0) {
5551 #  if defined(SNPRINTF) && defined(_KERNEL)
5552 		(void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5553 #  else
5554 		(void) sprintf(temp, "%d", unit);
5555 #  endif
5556 		(void) strncpy(s, temp, space);
5557 	}
5558 # endif
5559 	return buffer;
5560 }
5561 #endif
5562 
5563 
5564 /* ------------------------------------------------------------------------ */
5565 /* Function:    fr_ioctlswitch                                              */
5566 /* Returns:     int     - -1 continue processing, else ioctl return value   */
5567 /* Parameters:  unit(I) - device unit opened                                */
5568 /*              data(I) - pointer to ioctl data                             */
5569 /*              cmd(I)  - ioctl command                                     */
5570 /*              mode(I) - mode value                                        */
5571 /*                                                                          */
5572 /* Based on the value of unit, call the appropriate ioctl handler or return */
5573 /* EIO if ipfilter is not running.   Also checks if write perms are req'd   */
5574 /* for the device in order to execute the ioctl.                            */
5575 /* ------------------------------------------------------------------------ */
5576 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5577 int unit, mode, uid;
5578 ioctlcmd_t cmd;
5579 void *data, *ctx;
5580 ipf_stack_t *ifs;
5581 {
5582 	int error = 0;
5583 
5584 	switch (unit)
5585 	{
5586 	case IPL_LOGIPF :
5587 		error = -1;
5588 		break;
5589 	case IPL_LOGNAT :
5590 		if (ifs->ifs_fr_running > 0)
5591 			error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5592 		else
5593 			error = EIO;
5594 		break;
5595 	case IPL_LOGSTATE :
5596 		if (ifs->ifs_fr_running > 0)
5597 			error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5598 		else
5599 			error = EIO;
5600 		break;
5601 	case IPL_LOGAUTH :
5602 		if (ifs->ifs_fr_running > 0) {
5603 			if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5604 			    (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5605 				if (!(mode & FWRITE)) {
5606 					error = EPERM;
5607 				} else {
5608 					error = frrequest(unit, cmd, data,
5609 						  ifs->ifs_fr_active, 1, ifs);
5610 				}
5611 			} else {
5612 				error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5613 			}
5614 		} else
5615 			error = EIO;
5616 		break;
5617 	case IPL_LOGSYNC :
5618 #ifdef IPFILTER_SYNC
5619 		if (ifs->ifs_fr_running > 0)
5620 			error = fr_sync_ioctl(data, cmd, mode, ifs);
5621 		else
5622 #endif
5623 			error = EIO;
5624 		break;
5625 	case IPL_LOGSCAN :
5626 #ifdef IPFILTER_SCAN
5627 		if (ifs->ifs_fr_running > 0)
5628 			error = fr_scan_ioctl(data, cmd, mode, ifs);
5629 		else
5630 #endif
5631 			error = EIO;
5632 		break;
5633 	case IPL_LOGLOOKUP :
5634 #ifdef IPFILTER_LOOKUP
5635 		if (ifs->ifs_fr_running > 0)
5636 			error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5637 		else
5638 #endif
5639 			error = EIO;
5640 		break;
5641 	default :
5642 		error = EIO;
5643 		break;
5644 	}
5645 
5646 	return error;
5647 }
5648 
5649 
5650 /*
5651  * This array defines the expected size of objects coming into the kernel
5652  * for the various recognised object types.
5653  */
5654 #define	NUM_OBJ_TYPES	19
5655 
5656 static	int	fr_objbytes[NUM_OBJ_TYPES][2] = {
5657 	{ 1,	sizeof(struct frentry) },		/* frentry */
5658 	{ 0,	sizeof(struct friostat) },
5659 	{ 0,	sizeof(struct fr_info) },
5660 	{ 0,	sizeof(struct fr_authstat) },
5661 	{ 0,	sizeof(struct ipfrstat) },
5662 	{ 0,	sizeof(struct ipnat) },
5663 	{ 0,	sizeof(struct natstat) },
5664 	{ 0,	sizeof(struct ipstate_save) },
5665 	{ 1,	sizeof(struct nat_save) },		/* nat_save */
5666 	{ 0,	sizeof(struct natlookup) },
5667 	{ 1,	sizeof(struct ipstate) },		/* ipstate */
5668 	{ 0,	sizeof(struct ips_stat) },
5669 	{ 0,	sizeof(struct frauth) },
5670 	{ 0,	sizeof(struct ipftune) },
5671 	{ 0,	sizeof(struct nat) },                   /* nat_t */
5672 	{ 0,	sizeof(struct ipfruleiter) },
5673 	{ 0,	sizeof(struct ipfgeniter) },
5674 	{ 0,	sizeof(struct ipftable) },
5675 	{ 0,	sizeof(struct ipflookupiter) }
5676 };
5677 
5678 
5679 /* ------------------------------------------------------------------------ */
5680 /* Function:    fr_inobj                                                    */
5681 /* Returns:     int     - 0 = success, else failure                         */
5682 /* Parameters:  data(I) - pointer to ioctl data                             */
5683 /*              ptr(I)  - pointer to store real data in                     */
5684 /*              type(I) - type of structure being moved                     */
5685 /*                                                                          */
5686 /* Copy in the contents of what the ipfobj_t points to.  In future, we      */
5687 /* add things to check for version numbers, sizes, etc, to make it backward */
5688 /* compatible at the ABI for user land.                                     */
5689 /* ------------------------------------------------------------------------ */
5690 int fr_inobj(data, ptr, type)
5691 void *data;
5692 void *ptr;
5693 int type;
5694 {
5695 	ipfobj_t obj;
5696 	int error = 0;
5697 
5698 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5699 		return EINVAL;
5700 
5701 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5702 	if (error != 0)
5703 		return EFAULT;
5704 
5705 	if (obj.ipfo_type != type)
5706 		return EINVAL;
5707 
5708 #ifndef	IPFILTER_COMPAT
5709 	if ((fr_objbytes[type][0] & 1) != 0) {
5710 		if (obj.ipfo_size < fr_objbytes[type][1])
5711 			return EINVAL;
5712 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5713 		return EINVAL;
5714 #else
5715 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5716 		error = fr_incomptrans(&obj, ptr);
5717 		return error;
5718 	}
5719 
5720 	if ((fr_objbytes[type][0] & 1) != 0 &&
5721 	    obj.ipfo_size < fr_objbytes[type][1] ||
5722 	    obj.ipfo_size != fr_objbytes[type][1])
5723 		return EINVAL;
5724 #endif
5725 
5726 	if ((fr_objbytes[type][0] & 1) != 0) {
5727 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5728 				fr_objbytes[type][1]);
5729 	} else {
5730 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5731 				obj.ipfo_size);
5732 	}
5733 	return error;
5734 }
5735 
5736 
5737 /* ------------------------------------------------------------------------ */
5738 /* Function:    fr_inobjsz                                                  */
5739 /* Returns:     int     - 0 = success, else failure                         */
5740 /* Parameters:  data(I) - pointer to ioctl data                             */
5741 /*              ptr(I)  - pointer to store real data in                     */
5742 /*              type(I) - type of structure being moved                     */
5743 /*              sz(I)   - size of data to copy                              */
5744 /*                                                                          */
5745 /* As per fr_inobj, except the size of the object to copy in is passed in   */
5746 /* but it must not be smaller than the size defined for the type and the    */
5747 /* type must allow for varied sized objects.  The extra requirement here is */
5748 /* that sz must match the size of the object being passed in - this is not  */
5749 /* not possible nor required in fr_inobj().                                 */
5750 /* ------------------------------------------------------------------------ */
5751 int fr_inobjsz(data, ptr, type, sz)
5752 void *data;
5753 void *ptr;
5754 int type, sz;
5755 {
5756 	ipfobj_t obj;
5757 	int error;
5758 
5759 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5760 		return EINVAL;
5761 	if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5762 		return EINVAL;
5763 
5764 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5765 	if (error != 0)
5766 		return EFAULT;
5767 
5768 	if (obj.ipfo_type != type)
5769 		return EINVAL;
5770 
5771 #ifndef	IPFILTER_COMPAT
5772 	if (obj.ipfo_size != sz)
5773 		return EINVAL;
5774 #else
5775 	if (obj.ipfo_rev != IPFILTER_VERSION)
5776 		/*XXX compatibility hook here */
5777 		/*EMPTY*/;
5778 	if (obj.ipfo_size != sz)
5779 		/* XXX compatibility hook here */
5780 		return EINVAL;
5781 #endif
5782 
5783 	error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5784 	return error;
5785 }
5786 
5787 
5788 /* ------------------------------------------------------------------------ */
5789 /* Function:    fr_outobjsz                                                 */
5790 /* Returns:     int     - 0 = success, else failure                         */
5791 /* Parameters:  data(I) - pointer to ioctl data                             */
5792 /*              ptr(I)  - pointer to store real data in                     */
5793 /*              type(I) - type of structure being moved                     */
5794 /*              sz(I)   - size of data to copy                              */
5795 /*                                                                          */
5796 /* As per fr_outobj, except the size of the object to copy out is passed in */
5797 /* but it must not be smaller than the size defined for the type and the    */
5798 /* type must allow for varied sized objects.  The extra requirement here is */
5799 /* that sz must match the size of the object being passed in - this is not  */
5800 /* not possible nor required in fr_outobj().                                */
5801 /* ------------------------------------------------------------------------ */
5802 int fr_outobjsz(data, ptr, type, sz)
5803 void *data;
5804 void *ptr;
5805 int type, sz;
5806 {
5807 	ipfobj_t obj;
5808 	int error;
5809 
5810 	if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5811 	    ((fr_objbytes[type][0] & 1) == 0) ||
5812 	    (sz < fr_objbytes[type][1]))
5813 		return EINVAL;
5814 
5815 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5816 	if (error != 0)
5817 		return EFAULT;
5818 
5819 	if (obj.ipfo_type != type)
5820 		return EINVAL;
5821 
5822 #ifndef	IPFILTER_COMPAT
5823 	if (obj.ipfo_size != sz)
5824 		return EINVAL;
5825 #else
5826 	if (obj.ipfo_rev != IPFILTER_VERSION)
5827 		/* XXX compatibility hook here */
5828 		/*EMPTY*/;
5829 	if (obj.ipfo_size != sz)
5830 		/* XXX compatibility hook here */
5831 		return EINVAL;
5832 #endif
5833 
5834 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5835 	return error;
5836 }
5837 
5838 
5839 /* ------------------------------------------------------------------------ */
5840 /* Function:    fr_outobj                                                   */
5841 /* Returns:     int     - 0 = success, else failure                         */
5842 /* Parameters:  data(I) - pointer to ioctl data                             */
5843 /*              ptr(I)  - pointer to store real data in                     */
5844 /*              type(I) - type of structure being moved                     */
5845 /*                                                                          */
5846 /* Copy out the contents of what ptr is to where ipfobj points to.  In      */
5847 /* future, we add things to check for version numbers, sizes, etc, to make  */
5848 /* it backward  compatible at the ABI for user land.                        */
5849 /* ------------------------------------------------------------------------ */
5850 int fr_outobj(data, ptr, type)
5851 void *data;
5852 void *ptr;
5853 int type;
5854 {
5855 	ipfobj_t obj;
5856 	int error;
5857 
5858 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5859 		return EINVAL;
5860 
5861 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5862 	if (error != 0)
5863 		return EFAULT;
5864 
5865 	if (obj.ipfo_type != type)
5866 		return EINVAL;
5867 
5868 #ifndef	IPFILTER_COMPAT
5869 	if ((fr_objbytes[type][0] & 1) != 0) {
5870 		if (obj.ipfo_size < fr_objbytes[type][1])
5871 			return EINVAL;
5872 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5873 		return EINVAL;
5874 #else
5875 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5876 		error = fr_outcomptrans(&obj, ptr);
5877 		return error;
5878 	}
5879 
5880 	if ((fr_objbytes[type][0] & 1) != 0 &&
5881 	    obj.ipfo_size < fr_objbytes[type][1] ||
5882 	    obj.ipfo_size != fr_objbytes[type][1])
5883 		return EINVAL;
5884 #endif
5885 
5886 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5887 	return error;
5888 }
5889 
5890 
5891 /* ------------------------------------------------------------------------ */
5892 /* Function:    fr_checkl4sum                                               */
5893 /* Returns:     int     - 0 = good, -1 = bad, 1 = cannot check              */
5894 /* Parameters:  fin(I) - pointer to packet information                      */
5895 /*                                                                          */
5896 /* If possible, calculate the layer 4 checksum for the packet.  If this is  */
5897 /* not possible, return without indicating a failure or success but in a    */
5898 /* way that is ditinguishable.                                              */
5899 /* ------------------------------------------------------------------------ */
5900 int fr_checkl4sum(fin)
5901 fr_info_t *fin;
5902 {
5903 	u_short sum, hdrsum, *csump;
5904 	udphdr_t *udp;
5905 	int dosum;
5906 	ipf_stack_t *ifs = fin->fin_ifs;
5907 
5908 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5909 	net_handle_t net_data_p;
5910 	if (fin->fin_v == 4)
5911 		net_data_p = ifs->ifs_ipf_ipv4;
5912 	else
5913 		net_data_p = ifs->ifs_ipf_ipv6;
5914 #endif
5915 
5916 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
5917 		return 0;
5918 
5919 	/*
5920 	 * If the TCP packet isn't a fragment, isn't too short and otherwise
5921 	 * isn't already considered "bad", then validate the checksum.  If
5922 	 * this check fails then considered the packet to be "bad".
5923 	 */
5924 	if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
5925 		return 1;
5926 
5927 	csump = NULL;
5928 	hdrsum = 0;
5929 	dosum = 0;
5930 	sum = 0;
5931 
5932 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5933 	ASSERT(fin->fin_m != NULL);
5934 	if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
5935 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
5936 			hdrsum = 0;
5937 			sum = 0;
5938 	} else {
5939 #endif
5940 		switch (fin->fin_p)
5941 		{
5942 		case IPPROTO_TCP :
5943 			csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
5944 			dosum = 1;
5945 			break;
5946 
5947 		case IPPROTO_UDP :
5948 			udp = fin->fin_dp;
5949 			if (udp->uh_sum != 0) {
5950 				csump = &udp->uh_sum;
5951 				dosum = 1;
5952 			}
5953 			break;
5954 
5955 		case IPPROTO_ICMP :
5956 			csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
5957 			dosum = 1;
5958 			break;
5959 
5960 		default :
5961 			return 1;
5962 			/*NOTREACHED*/
5963 		}
5964 
5965 		if (csump != NULL)
5966 			hdrsum = *csump;
5967 
5968 		if (dosum)
5969 			sum = fr_cksum(fin->fin_m, fin->fin_ip,
5970 				       fin->fin_p, fin->fin_dp);
5971 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5972 	}
5973 #endif
5974 #if !defined(_KERNEL)
5975 	if (sum == hdrsum) {
5976 		FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
5977 	} else {
5978 		FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
5979 	}
5980 #endif
5981 	if (hdrsum == sum)
5982 		return 0;
5983 	return -1;
5984 }
5985 
5986 
5987 /* ------------------------------------------------------------------------ */
5988 /* Function:    fr_ifpfillv4addr                                            */
5989 /* Returns:     int     - 0 = address update, -1 = address not updated      */
5990 /* Parameters:  atype(I)   - type of network address update to perform      */
5991 /*              sin(I)     - pointer to source of address information       */
5992 /*              mask(I)    - pointer to source of netmask information       */
5993 /*              inp(I)     - pointer to destination address store           */
5994 /*              inpmask(I) - pointer to destination netmask store           */
5995 /*                                                                          */
5996 /* Given a type of network address update (atype) to perform, copy          */
5997 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
5998 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
5999 /* which case the operation fails.  For all values of atype other than      */
6000 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6001 /* value.                                                                   */
6002 /* ------------------------------------------------------------------------ */
6003 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
6004 int atype;
6005 struct sockaddr_in *sin, *mask;
6006 struct in_addr *inp, *inpmask;
6007 {
6008 	if (inpmask != NULL && atype != FRI_NETMASKED)
6009 		inpmask->s_addr = 0xffffffff;
6010 
6011 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6012 		if (atype == FRI_NETMASKED) {
6013 			if (inpmask == NULL)
6014 				return -1;
6015 			inpmask->s_addr = mask->sin_addr.s_addr;
6016 		}
6017 		inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
6018 	} else {
6019 		inp->s_addr = sin->sin_addr.s_addr;
6020 	}
6021 	return 0;
6022 }
6023 
6024 
6025 #ifdef	USE_INET6
6026 /* ------------------------------------------------------------------------ */
6027 /* Function:    fr_ifpfillv6addr                                            */
6028 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6029 /* Parameters:  atype(I)   - type of network address update to perform      */
6030 /*              sin(I)     - pointer to source of address information       */
6031 /*              mask(I)    - pointer to source of netmask information       */
6032 /*              inp(I)     - pointer to destination address store           */
6033 /*              inpmask(I) - pointer to destination netmask store           */
6034 /*                                                                          */
6035 /* Given a type of network address update (atype) to perform, copy          */
6036 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6037 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6038 /* which case the operation fails.  For all values of atype other than      */
6039 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6040 /* value.                                                                   */
6041 /* ------------------------------------------------------------------------ */
6042 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
6043 int atype;
6044 struct sockaddr_in6 *sin, *mask;
6045 struct in_addr *inp, *inpmask;
6046 {
6047 	i6addr_t *src, *dst, *and, *dmask;
6048 
6049 	src = (i6addr_t *)&sin->sin6_addr;
6050 	and = (i6addr_t *)&mask->sin6_addr;
6051 	dst = (i6addr_t *)inp;
6052 	dmask = (i6addr_t *)inpmask;
6053 
6054 	if (inpmask != NULL && atype != FRI_NETMASKED) {
6055 		dmask->i6[0] = 0xffffffff;
6056 		dmask->i6[1] = 0xffffffff;
6057 		dmask->i6[2] = 0xffffffff;
6058 		dmask->i6[3] = 0xffffffff;
6059 	}
6060 
6061 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6062 		if (atype == FRI_NETMASKED) {
6063 			if (inpmask == NULL)
6064 				return -1;
6065 			dmask->i6[0] = and->i6[0];
6066 			dmask->i6[1] = and->i6[1];
6067 			dmask->i6[2] = and->i6[2];
6068 			dmask->i6[3] = and->i6[3];
6069 		}
6070 
6071 		dst->i6[0] = src->i6[0] & and->i6[0];
6072 		dst->i6[1] = src->i6[1] & and->i6[1];
6073 		dst->i6[2] = src->i6[2] & and->i6[2];
6074 		dst->i6[3] = src->i6[3] & and->i6[3];
6075 	} else {
6076 		dst->i6[0] = src->i6[0];
6077 		dst->i6[1] = src->i6[1];
6078 		dst->i6[2] = src->i6[2];
6079 		dst->i6[3] = src->i6[3];
6080 	}
6081 	return 0;
6082 }
6083 #endif
6084 
6085 
6086 /* ------------------------------------------------------------------------ */
6087 /* Function:    fr_matchtag                                                 */
6088 /* Returns:     0 == mismatch, 1 == match.                                  */
6089 /* Parameters:  tag1(I) - pointer to first tag to compare                   */
6090 /*              tag2(I) - pointer to second tag to compare                  */
6091 /*                                                                          */
6092 /* Returns true (non-zero) or false(0) if the two tag structures can be     */
6093 /* considered to be a match or not match, respectively.  The tag is 16      */
6094 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so    */
6095 /* compare the ints instead, for speed. tag1 is the master of the           */
6096 /* comparison.  This function should only be called with both tag1 and tag2 */
6097 /* as non-NULL pointers.                                                    */
6098 /* ------------------------------------------------------------------------ */
6099 int fr_matchtag(tag1, tag2)
6100 ipftag_t *tag1, *tag2;
6101 {
6102 	if (tag1 == tag2)
6103 		return 1;
6104 
6105 	if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
6106 		return 1;
6107 
6108 	if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
6109 	    (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
6110 	    (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
6111 	    (tag1->ipt_num[3] == tag2->ipt_num[3]))
6112 		return 1;
6113 	return 0;
6114 }
6115 
6116 
6117 /* ------------------------------------------------------------------------ */
6118 /* Function:    fr_coalesce                                                 */
6119 /* Returns:     1 == success, -1 == failure, 0 == no change                 */
6120 /* Parameters:  fin(I) - pointer to packet information                      */
6121 /*                                                                          */
6122 /* Attempt to get all of the packet data into a single, contiguous buffer.  */
6123 /* If this call returns a failure then the buffers have also been freed.    */
6124 /* ------------------------------------------------------------------------ */
6125 int fr_coalesce(fin)
6126 fr_info_t *fin;
6127 {
6128 	ipf_stack_t *ifs = fin->fin_ifs;
6129 	if ((fin->fin_flx & FI_COALESCE) != 0)
6130 		return 1;
6131 
6132 	/*
6133 	 * If the mbuf pointers indicate that there is no mbuf to work with,
6134 	 * return but do not indicate success or failure.
6135 	 */
6136 	if (fin->fin_m == NULL || fin->fin_mp == NULL)
6137 		return 0;
6138 
6139 #if defined(_KERNEL)
6140 	if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
6141 		IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
6142 # ifdef MENTAT
6143 		FREE_MB_T(*fin->fin_mp);
6144 # endif
6145 		*fin->fin_mp = NULL;
6146 		fin->fin_m = NULL;
6147 		return -1;
6148 	}
6149 #else
6150 	fin = fin;	/* LINT */
6151 #endif
6152 	return 1;
6153 }
6154 
6155 
6156 /*
6157  * The following table lists all of the tunable variables that can be
6158  * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT.  The format of each row
6159  * in the table below is as follows:
6160  *
6161  * pointer to value, name of value, minimum, maximum, size of the value's
6162  *     container, value attribute flags
6163  *
6164  * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6165  * means the value can only be written to when IPFilter is loaded but disabled.
6166  * The obvious implication is if neither of these are set then the value can be
6167  * changed at any time without harm.
6168  */
6169 ipftuneable_t lcl_ipf_tuneables[] = {
6170 	/* filtering */
6171 	{ { NULL },	"fr_flags",		0,	0xffffffff,
6172 			0,		0 },
6173 	{ { NULL },	"fr_active",		0,	0,
6174 			0,		IPFT_RDONLY },
6175 	{ { NULL },	"fr_control_forwarding",	0, 1,
6176 			0,		0 },
6177 	{ { NULL },	"fr_update_ipid",	0,	1,
6178 			0,		0 },
6179 	{ { NULL },	"fr_chksrc",		0,	1,
6180 			0,		0 },
6181 	{ { NULL },	"fr_minttl",		0,	1,
6182 			0,		0 },
6183 	{ { NULL }, 	"fr_icmpminfragmtu",	0,	1,
6184 			0,		0 },
6185 	{ { NULL },		"fr_pass",		0,	0xffffffff,
6186 			0,		0 },
6187 #if SOLARIS2 >= 10
6188 	{ { NULL },	"ipf_loopback",		0,	1,
6189 			0,		IPFT_WRDISABLED },
6190 #endif
6191 	/* state */
6192 	{ { NULL }, "fr_tcpidletimeout",	1,	0x7fffffff,
6193 			0,	IPFT_WRDISABLED },
6194 	{ { NULL },	"fr_tcpclosewait",	1,	0x7fffffff,
6195 			0,	IPFT_WRDISABLED },
6196 	{ { NULL },	"fr_tcplastack",	1,	0x7fffffff,
6197 			0,		IPFT_WRDISABLED },
6198 	{ { NULL },	"fr_tcptimeout",	1,	0x7fffffff,
6199 			0,		IPFT_WRDISABLED },
6200 	{ { NULL },	"fr_tcpclosed",		1,	0x7fffffff,
6201 			0,		IPFT_WRDISABLED },
6202 	{ { NULL },	"fr_tcphalfclosed",	1,	0x7fffffff,
6203 			0,	IPFT_WRDISABLED },
6204 	{ { NULL },	"fr_udptimeout",	1,	0x7fffffff,
6205 			0,		IPFT_WRDISABLED },
6206 	{ { NULL }, "fr_udpacktimeout",	1,	0x7fffffff,
6207 			0,	IPFT_WRDISABLED },
6208 	{ { NULL },	"fr_icmptimeout",	1,	0x7fffffff,
6209 			0,		IPFT_WRDISABLED },
6210 	{ { NULL }, "fr_icmpacktimeout",	1,	0x7fffffff,
6211 			0,	IPFT_WRDISABLED },
6212 	{ { NULL }, "fr_iptimeout",		1,	0x7fffffff,
6213 			0,		IPFT_WRDISABLED },
6214 	{ { NULL },	"fr_statemax",		1,	0x7fffffff,
6215 			0,		0 },
6216 	{ { NULL },	"fr_statesize",		1,	0x7fffffff,
6217 			0,		IPFT_WRDISABLED },
6218 	{ { NULL },	"fr_state_lock",	0,	1,
6219 			0,		IPFT_RDONLY },
6220 	{ { NULL }, "fr_state_maxbucket", 1,	0x7fffffff,
6221 			0,	IPFT_WRDISABLED },
6222 	{ { NULL }, "fr_state_maxbucket_reset",	0, 1,
6223 			0, IPFT_WRDISABLED },
6224 	{ { NULL },	"ipstate_logging",	0,	1,
6225 			0,	0 },
6226 	{ { NULL },	"state_flush_level_hi",	1,	100,
6227 			0,		0 },
6228 	{ { NULL },	"state_flush_level_lo",	1,	100,
6229 			0,		0 },
6230 	/* nat */
6231 	{ { NULL },		"fr_nat_lock",		0,	1,
6232 			0,		IPFT_RDONLY },
6233 	{ { NULL },	"ipf_nattable_sz",	1,	0x7fffffff,
6234 			0,	IPFT_WRDISABLED },
6235 	{ { NULL }, "ipf_nattable_max",	1,	0x7fffffff,
6236 			0,	0 },
6237 	{ { NULL },	"ipf_natrules_sz",	1,	0x7fffffff,
6238 			0,	IPFT_WRDISABLED },
6239 	{ { NULL },	"ipf_rdrrules_sz",	1,	0x7fffffff,
6240 			0,	IPFT_WRDISABLED },
6241 	{ { NULL },	"ipf_hostmap_sz",	1,	0x7fffffff,
6242 			0,		IPFT_WRDISABLED },
6243 	{ { NULL }, "fr_nat_maxbucket",	1,	0x7fffffff,
6244 			0,	IPFT_WRDISABLED },
6245 	{ { NULL },	"fr_nat_maxbucket_reset",	0, 1,
6246 			0,	IPFT_WRDISABLED },
6247 	{ { NULL },		"nat_logging",		0,	1,
6248 			0,		0 },
6249 	{ { NULL },	"fr_defnatage",		1,	0x7fffffff,
6250 			0,		IPFT_WRDISABLED },
6251 	{ { NULL },	"fr_defnatipage",	1,	0x7fffffff,
6252 			0,		IPFT_WRDISABLED },
6253 	{ { NULL }, "fr_defnaticmpage",	1,	0x7fffffff,
6254 			0,	IPFT_WRDISABLED },
6255 	{ { NULL },	"nat_flush_level_hi",	1,	100,
6256 			0,		0 },
6257 	{ { NULL },	"nat_flush_level_lo",	1,	100,
6258 			0,		0 },
6259 	/* frag */
6260 	{ { NULL },	"ipfr_size",		1,	0x7fffffff,
6261 			0,		IPFT_WRDISABLED },
6262 	{ { NULL },	"fr_ipfrttl",		1,	0x7fffffff,
6263 			0,		IPFT_WRDISABLED },
6264 #ifdef IPFILTER_LOG
6265 	/* log */
6266 	{ { NULL },	"ipl_suppress",		0,	1,
6267 			0,		0 },
6268 	{ { NULL },	"ipl_buffer_sz",	0,	0,
6269 			0,		IPFT_RDONLY },
6270 	{ { NULL },	"ipl_logmax",		0,	0x7fffffff,
6271 			0,		IPFT_WRDISABLED },
6272 	{ { NULL },	"ipl_logall",		0,	1,
6273 			0,		0 },
6274 	{ { NULL },	"ipl_logsize",		0,	0x80000,
6275 			0,		0 },
6276 #endif
6277 	{ { NULL },		NULL,			0,	0 }
6278 };
6279 
6280 static ipftuneable_t *
6281 tune_lookup(ipf_stack_t *ifs, char *name)
6282 {
6283     int i;
6284 
6285     for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6286 	if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6287 	    return (&ifs->ifs_ipf_tuneables[i]);
6288     }
6289     return (NULL);
6290 }
6291 
6292 #ifdef _KERNEL
6293 extern dev_info_t *ipf_dev_info;
6294 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6295 #endif
6296 
6297 /* -------------------------------------------------------------------- */
6298 /* Function:	ipftuneable_setdefs()					*/
6299 /* Returns:		void						*/
6300 /* Parameters:	ifs - pointer to newly allocated IPF instance		*/
6301 /*				assigned to	IP instance		*/
6302 /*									*/
6303 /* Function initializes IPF instance variables. Function is invoked	*/
6304 /* from	ipftuneable_alloc(). ipftuneable_alloc() is called only one	*/
6305 /* time during IP instance lifetime - at the time of IP instance	*/
6306 /* creation. Anytime IP	instance is being created new private IPF	*/
6307 /* instance is allocated and assigned to it. The moment of IP 		*/
6308 /* instance creation is the right time to initialize those IPF 		*/
6309 /* variables.								*/
6310 /*									*/
6311 /* -------------------------------------------------------------------- */
6312 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6313 {
6314 	ifs->ifs_ipfr_size = IPFT_SIZE;
6315 	ifs->ifs_fr_ipfrttl = 120;	/* 60 seconds */
6316 
6317 	/* it comes from fr_authinit() in IPF auth */
6318 	ifs->ifs_fr_authsize = FR_NUMAUTH;
6319 	ifs->ifs_fr_defaultauthage = 600;
6320 
6321 	/* it comes from fr_stateinit() in IPF state */
6322 	ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5);	/* five days */
6323 	ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6324 	ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6325 	ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6326 	ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6327 	ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600);	/* 2 hours */
6328 	ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6329 	ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6330 	ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6331 	ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6332 	ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6333 	ifs->ifs_fr_statemax = IPSTATE_MAX;
6334 	ifs->ifs_fr_statesize = IPSTATE_SIZE;
6335 	ifs->ifs_fr_state_maxbucket_reset = 1;
6336 	ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6337 	ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6338 
6339 	/* it comes from fr_natinit() in ipnat */
6340 	ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6341 	ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6342 	ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6343 	ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6344 	ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6345 	ifs->ifs_fr_nat_maxbucket_reset = 1;
6346 	ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6347 	ifs->ifs_fr_defnatipage = 120;		/* 60 seconds */
6348 	ifs->ifs_fr_defnaticmpage = 6;		/* 3 seconds */
6349 	ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6350 	ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6351 
6352 #ifdef IPFILTER_LOG
6353 	/* it comes from fr_loginit() in IPF log */
6354 	ifs->ifs_ipl_suppress = 1;
6355 	ifs->ifs_ipl_logmax = IPL_LOGMAX;
6356 	ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6357 
6358 	/* from fr_natinit() */
6359 	ifs->ifs_nat_logging = 1;
6360 
6361 	/* from fr_stateinit() */
6362 	ifs->ifs_ipstate_logging = 1;
6363 #else
6364 	/* from fr_natinit() */
6365 	ifs->ifs_nat_logging = 0;
6366 
6367 	/* from fr_stateinit() */
6368 	ifs->ifs_ipstate_logging = 0;
6369 #endif
6370 	ifs->ifs_ipf_loopback = 0;
6371 
6372 }
6373 /*
6374  * Allocate a per-stack tuneable and copy in the names. Then
6375  * set it to point to each of the per-stack tunables.
6376  */
6377 void
6378 ipftuneable_alloc(ipf_stack_t *ifs)
6379 {
6380     ipftuneable_t *item;
6381 
6382     KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6383 	sizeof (lcl_ipf_tuneables));
6384     bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6385 	sizeof (lcl_ipf_tuneables));
6386 
6387 #define TUNE_SET(_ifs, _name, _field)			\
6388     item = tune_lookup((_ifs), (_name));		\
6389     if (item != NULL) {					\
6390 	item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field);	\
6391 	item->ipft_sz = sizeof ((_ifs)->_field);	\
6392     }
6393 
6394     TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6395     TUNE_SET(ifs, "fr_active", ifs_fr_active);
6396     TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6397     TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6398     TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6399     TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6400     TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6401     TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6402     TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6403     TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6404     TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6405     TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6406     TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6407     TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6408     TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6409     TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6410     TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6411     TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6412     TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6413     TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6414     TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6415     TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6416     TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6417     TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6418     TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6419     TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6420     TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6421     TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6422     TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6423     TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6424     TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6425     TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6426     TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6427     TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6428     TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6429     TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6430     TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6431     TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6432     TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6433     TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6434     TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6435     TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6436     TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6437     TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6438 #ifdef IPFILTER_LOG
6439     TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6440     TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6441     TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6442     TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6443     TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6444 #endif
6445 #undef TUNE_SET
6446 
6447 	ipftuneable_setdefs(ifs);
6448 
6449 #ifdef _KERNEL
6450     (void) ipf_property_update(ipf_dev_info, ifs);
6451 #endif
6452 }
6453 
6454 void
6455 ipftuneable_free(ipf_stack_t *ifs)
6456 {
6457 	KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6458 	ifs->ifs_ipf_tuneables = NULL;
6459 }
6460 
6461 /* ------------------------------------------------------------------------ */
6462 /* Function:    fr_findtunebycookie                                         */
6463 /* Returns:     NULL = search failed, else pointer to tune struct           */
6464 /* Parameters:  cookie(I) - cookie value to search for amongst tuneables    */
6465 /*              next(O)   - pointer to place to store the cookie for the    */
6466 /*                          "next" tuneable, if it is desired.              */
6467 /*                                                                          */
6468 /* This function is used to walk through all of the existing tunables with  */
6469 /* successive calls.  It searches the known tunables for the one which has  */
6470 /* a matching value for "cookie" - ie its address.  When returning a match, */
6471 /* the next one to be found may be returned inside next.                    */
6472 /* ------------------------------------------------------------------------ */
6473 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6474 void *cookie, **next;
6475 ipf_stack_t * ifs;
6476 {
6477 	ipftuneable_t *ta, **tap;
6478 
6479 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6480 		if (ta == cookie) {
6481 			if (next != NULL) {
6482 				/*
6483 				 * If the next entry in the array has a name
6484 				 * present, then return a pointer to it for
6485 				 * where to go next, else return a pointer to
6486 				 * the dynaminc list as a key to search there
6487 				 * next.  This facilitates a weak linking of
6488 				 * the two "lists" together.
6489 				 */
6490 				if ((ta + 1)->ipft_name != NULL)
6491 					*next = ta + 1;
6492 				else
6493 					*next = &ifs->ifs_ipf_tunelist;
6494 			}
6495 			return ta;
6496 		}
6497 
6498 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6499 		if (tap == cookie) {
6500 			if (next != NULL)
6501 				*next = &ta->ipft_next;
6502 			return ta;
6503 		}
6504 
6505 	if (next != NULL)
6506 		*next = NULL;
6507 	return NULL;
6508 }
6509 
6510 
6511 /* ------------------------------------------------------------------------ */
6512 /* Function:    fr_findtunebyname                                           */
6513 /* Returns:     NULL = search failed, else pointer to tune struct           */
6514 /* Parameters:  name(I) - name of the tuneable entry to find.               */
6515 /*                                                                          */
6516 /* Search the static array of tuneables and the list of dynamic tuneables   */
6517 /* for an entry with a matching name.  If we can find one, return a pointer */
6518 /* to the matching structure.                                               */
6519 /* ------------------------------------------------------------------------ */
6520 static ipftuneable_t *fr_findtunebyname(name, ifs)
6521 const char *name;
6522 ipf_stack_t *ifs;
6523 {
6524 	ipftuneable_t *ta;
6525 
6526 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6527 		if (!strcmp(ta->ipft_name, name)) {
6528 			return ta;
6529 		}
6530 
6531 	for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6532 		if (!strcmp(ta->ipft_name, name)) {
6533 			return ta;
6534 		}
6535 
6536 	return NULL;
6537 }
6538 
6539 
6540 /* ------------------------------------------------------------------------ */
6541 /* Function:    fr_addipftune                                               */
6542 /* Returns:     int - 0 == success, else failure                            */
6543 /* Parameters:  newtune - pointer to new tune struct to add to tuneables    */
6544 /*                                                                          */
6545 /* Appends the tune structure pointer to by "newtune" to the end of the     */
6546 /* current list of "dynamic" tuneable parameters.  Once added, the owner    */
6547 /* of the object is not expected to ever change "ipft_next".                */
6548 /* ------------------------------------------------------------------------ */
6549 int fr_addipftune(newtune, ifs)
6550 ipftuneable_t *newtune;
6551 ipf_stack_t *ifs;
6552 {
6553 	ipftuneable_t *ta, **tap;
6554 
6555 	ta = fr_findtunebyname(newtune->ipft_name, ifs);
6556 	if (ta != NULL)
6557 		return EEXIST;
6558 
6559 	for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6560 		;
6561 
6562 	newtune->ipft_next = NULL;
6563 	*tap = newtune;
6564 	return 0;
6565 }
6566 
6567 
6568 /* ------------------------------------------------------------------------ */
6569 /* Function:    fr_delipftune                                               */
6570 /* Returns:     int - 0 == success, else failure                            */
6571 /* Parameters:  oldtune - pointer to tune struct to remove from the list of */
6572 /*                        current dynamic tuneables                         */
6573 /*                                                                          */
6574 /* Search for the tune structure, by pointer, in the list of those that are */
6575 /* dynamically added at run time.  If found, adjust the list so that this   */
6576 /* structure is no longer part of it.                                       */
6577 /* ------------------------------------------------------------------------ */
6578 int fr_delipftune(oldtune, ifs)
6579 ipftuneable_t *oldtune;
6580 ipf_stack_t *ifs;
6581 {
6582 	ipftuneable_t *ta, **tap;
6583 
6584 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6585 		if (ta == oldtune) {
6586 			*tap = oldtune->ipft_next;
6587 			oldtune->ipft_next = NULL;
6588 			return 0;
6589 		}
6590 
6591 	return ESRCH;
6592 }
6593 
6594 
6595 /* ------------------------------------------------------------------------ */
6596 /* Function:    fr_ipftune                                                  */
6597 /* Returns:     int - 0 == success, else failure                            */
6598 /* Parameters:  cmd(I)  - ioctl command number                              */
6599 /*              data(I) - pointer to ioctl data structure                   */
6600 /*                                                                          */
6601 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET.  These  */
6602 /* three ioctls provide the means to access and control global variables    */
6603 /* within IPFilter, allowing (for example) timeouts and table sizes to be   */
6604 /* changed without rebooting, reloading or recompiling.  The initialisation */
6605 /* and 'destruction' routines of the various components of ipfilter are all */
6606 /* each responsible for handling their own values being too big.            */
6607 /* ------------------------------------------------------------------------ */
6608 int fr_ipftune(cmd, data, ifs)
6609 ioctlcmd_t cmd;
6610 void *data;
6611 ipf_stack_t *ifs;
6612 {
6613 	ipftuneable_t *ta;
6614 	ipftune_t tu;
6615 	void *cookie;
6616 	int error;
6617 
6618 	error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6619 	if (error != 0)
6620 		return error;
6621 
6622 	tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6623 	cookie = tu.ipft_cookie;
6624 	ta = NULL;
6625 
6626 	switch (cmd)
6627 	{
6628 	case SIOCIPFGETNEXT :
6629 		/*
6630 		 * If cookie is non-NULL, assume it to be a pointer to the last
6631 		 * entry we looked at, so find it (if possible) and return a
6632 		 * pointer to the next one after it.  The last entry in the
6633 		 * the table is a NULL entry, so when we get to it, set cookie
6634 		 * to NULL and return that, indicating end of list, erstwhile
6635 		 * if we come in with cookie set to NULL, we are starting anew
6636 		 * at the front of the list.
6637 		 */
6638 		if (cookie != NULL) {
6639 			ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6640 		} else {
6641 			ta = ifs->ifs_ipf_tuneables;
6642 			tu.ipft_cookie = ta + 1;
6643 		}
6644 		if (ta != NULL) {
6645 			/*
6646 			 * Entry found, but does the data pointed to by that
6647 			 * row fit in what we can return?
6648 			 */
6649 			if (ta->ipft_sz > sizeof(tu.ipft_un))
6650 				return EINVAL;
6651 
6652 			tu.ipft_vlong = 0;
6653 			if (ta->ipft_sz == sizeof(u_long))
6654 				tu.ipft_vlong = *ta->ipft_plong;
6655 			else if (ta->ipft_sz == sizeof(u_int))
6656 				tu.ipft_vint = *ta->ipft_pint;
6657 			else if (ta->ipft_sz == sizeof(u_short))
6658 				tu.ipft_vshort = *ta->ipft_pshort;
6659 			else if (ta->ipft_sz == sizeof(u_char))
6660 				tu.ipft_vchar = *ta->ipft_pchar;
6661 
6662 			tu.ipft_sz = ta->ipft_sz;
6663 			tu.ipft_min = ta->ipft_min;
6664 			tu.ipft_max = ta->ipft_max;
6665 			tu.ipft_flags = ta->ipft_flags;
6666 			bcopy(ta->ipft_name, tu.ipft_name,
6667 			      MIN(sizeof(tu.ipft_name),
6668 				  strlen(ta->ipft_name) + 1));
6669 		}
6670 		error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6671 		break;
6672 
6673 	case SIOCIPFGET :
6674 	case SIOCIPFSET :
6675 		/*
6676 		 * Search by name or by cookie value for a particular entry
6677 		 * in the tuning paramter table.
6678 		 */
6679 		error = ESRCH;
6680 		if (cookie != NULL) {
6681 			ta = fr_findtunebycookie(cookie, NULL, ifs);
6682 			if (ta != NULL)
6683 				error = 0;
6684 		} else if (tu.ipft_name[0] != '\0') {
6685 			ta = fr_findtunebyname(tu.ipft_name, ifs);
6686 			if (ta != NULL)
6687 				error = 0;
6688 		}
6689 		if (error != 0)
6690 			break;
6691 
6692 		if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6693 			/*
6694 			 * Fetch the tuning parameters for a particular value
6695 			 */
6696 			tu.ipft_vlong = 0;
6697 			if (ta->ipft_sz == sizeof(u_long))
6698 				tu.ipft_vlong = *ta->ipft_plong;
6699 			else if (ta->ipft_sz == sizeof(u_int))
6700 				tu.ipft_vint = *ta->ipft_pint;
6701 			else if (ta->ipft_sz == sizeof(u_short))
6702 				tu.ipft_vshort = *ta->ipft_pshort;
6703 			else if (ta->ipft_sz == sizeof(u_char))
6704 				tu.ipft_vchar = *ta->ipft_pchar;
6705 			tu.ipft_cookie = ta;
6706 			tu.ipft_sz = ta->ipft_sz;
6707 			tu.ipft_min = ta->ipft_min;
6708 			tu.ipft_max = ta->ipft_max;
6709 			tu.ipft_flags = ta->ipft_flags;
6710 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6711 
6712 		} else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6713 			/*
6714 			 * Set an internal parameter.  The hard part here is
6715 			 * getting the new value safely and correctly out of
6716 			 * the kernel (given we only know its size, not type.)
6717 			 */
6718 			u_long in;
6719 
6720 			if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6721 			    (ifs->ifs_fr_running > 0)) {
6722 				error = EBUSY;
6723 				break;
6724 			}
6725 
6726 			in = tu.ipft_vlong;
6727 			if (in < ta->ipft_min || in > ta->ipft_max) {
6728 				error = EINVAL;
6729 				break;
6730 			}
6731 
6732 			if (ta->ipft_sz == sizeof(u_long)) {
6733 				tu.ipft_vlong = *ta->ipft_plong;
6734 				*ta->ipft_plong = in;
6735 			} else if (ta->ipft_sz == sizeof(u_int)) {
6736 				tu.ipft_vint = *ta->ipft_pint;
6737 				*ta->ipft_pint = (u_int)(in & 0xffffffff);
6738 			} else if (ta->ipft_sz == sizeof(u_short)) {
6739 				tu.ipft_vshort = *ta->ipft_pshort;
6740 				*ta->ipft_pshort = (u_short)(in & 0xffff);
6741 			} else if (ta->ipft_sz == sizeof(u_char)) {
6742 				tu.ipft_vchar = *ta->ipft_pchar;
6743 				*ta->ipft_pchar = (u_char)(in & 0xff);
6744 			}
6745 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6746 		}
6747 		break;
6748 
6749 	default :
6750 		error = EINVAL;
6751 		break;
6752 	}
6753 
6754 	return error;
6755 }
6756 
6757 
6758 /* ------------------------------------------------------------------------ */
6759 /* Function:    fr_initialise                                               */
6760 /* Returns:     int - 0 == success,  < 0 == failure                         */
6761 /* Parameters:  None.                                                       */
6762 /*                                                                          */
6763 /* Call of the initialise functions for all the various subsystems inside   */
6764 /* of IPFilter.  If any of them should fail, return immeadiately a failure  */
6765 /* BUT do not try to recover from the error here.                           */
6766 /* ------------------------------------------------------------------------ */
6767 int fr_initialise(ifs)
6768 ipf_stack_t *ifs;
6769 {
6770 	int i;
6771 
6772 #ifdef IPFILTER_LOG
6773 	i = fr_loginit(ifs);
6774 	if (i < 0)
6775 		return -10 + i;
6776 #endif
6777 	i = fr_natinit(ifs);
6778 	if (i < 0)
6779 		return -20 + i;
6780 
6781 	i = fr_stateinit(ifs);
6782 	if (i < 0)
6783 		return -30 + i;
6784 
6785 	i = fr_authinit(ifs);
6786 	if (i < 0)
6787 		return -40 + i;
6788 
6789 	i = fr_fraginit(ifs);
6790 	if (i < 0)
6791 		return -50 + i;
6792 
6793 	i = appr_init(ifs);
6794 	if (i < 0)
6795 		return -60 + i;
6796 
6797 #ifdef IPFILTER_SYNC
6798 	i = ipfsync_init(ifs);
6799 	if (i < 0)
6800 		return -70 + i;
6801 #endif
6802 #ifdef IPFILTER_SCAN
6803 	i = ipsc_init(ifs);
6804 	if (i < 0)
6805 		return -80 + i;
6806 #endif
6807 #ifdef IPFILTER_LOOKUP
6808 	i = ip_lookup_init(ifs);
6809 	if (i < 0)
6810 		return -90 + i;
6811 #endif
6812 #ifdef IPFILTER_COMPILED
6813 	ipfrule_add(ifs);
6814 #endif
6815 	return 0;
6816 }
6817 
6818 
6819 /* ------------------------------------------------------------------------ */
6820 /* Function:    fr_deinitialise                                             */
6821 /* Returns:     None.                                                       */
6822 /* Parameters:  None.                                                       */
6823 /*                                                                          */
6824 /* Call all the various subsystem cleanup routines to deallocate memory or  */
6825 /* destroy locks or whatever they've done that they need to now undo.       */
6826 /* The order here IS important as there are some cross references of        */
6827 /* internal data structures.                                                */
6828 /* ------------------------------------------------------------------------ */
6829 void fr_deinitialise(ifs)
6830 ipf_stack_t *ifs;
6831 {
6832 	fr_fragunload(ifs);
6833 	fr_authunload(ifs);
6834 	fr_natunload(ifs);
6835 	fr_stateunload(ifs);
6836 #ifdef IPFILTER_SCAN
6837 	fr_scanunload(ifs);
6838 #endif
6839 	appr_unload(ifs);
6840 
6841 #ifdef IPFILTER_COMPILED
6842 	ipfrule_remove(ifs);
6843 #endif
6844 
6845 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6846 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6847 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6848 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6849 
6850 #ifdef IPFILTER_LOOKUP
6851 	ip_lookup_unload(ifs);
6852 #endif
6853 
6854 #ifdef IPFILTER_LOG
6855 	fr_logunload(ifs);
6856 #endif
6857 }
6858 
6859 
6860 /* ------------------------------------------------------------------------ */
6861 /* Function:    fr_zerostats                                                */
6862 /* Returns:     int - 0 = success, else failure                             */
6863 /* Parameters:  data(O) - pointer to pointer for copying data back to       */
6864 /*                                                                          */
6865 /* Copies the current statistics out to userspace and then zero's the       */
6866 /* current ones in the kernel. The lock is only held across the bzero() as  */
6867 /* the copyout may result in paging (ie network activity.)                  */
6868 /* ------------------------------------------------------------------------ */
6869 int	fr_zerostats(data, ifs)
6870 caddr_t	data;
6871 ipf_stack_t *ifs;
6872 {
6873 	friostat_t fio;
6874 	int error;
6875 
6876 	fr_getstat(&fio, ifs);
6877 	error = copyoutptr(&fio, data, sizeof(fio));
6878 	if (error)
6879 		return EFAULT;
6880 
6881 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
6882 	bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6883 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6884 
6885 	return 0;
6886 }
6887 
6888 
6889 #ifdef _KERNEL
6890 /* ------------------------------------------------------------------------ */
6891 /* Function:    fr_resolvedest                                              */
6892 /* Returns:     Nil                                                         */
6893 /* Parameters:  fdp(IO) - pointer to destination information to resolve     */
6894 /*              v(I)    - IP protocol version to match                      */
6895 /*                                                                          */
6896 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6897 /* if a matching name can be found for the particular IP protocol version   */
6898 /* then store the interface pointer in the frdest struct.  If no match is   */
6899 /* found, then set the interface pointer to be -1 as NULL is considered to  */
6900 /* indicate there is no information at all in the structure.                */
6901 /* ------------------------------------------------------------------------ */
6902 void fr_resolvedest(fdp, v, ifs)
6903 frdest_t *fdp;
6904 int v;
6905 ipf_stack_t *ifs;
6906 {
6907 	fdp->fd_ifp = NULL;
6908 
6909   	if (*fdp->fd_ifname != '\0') {
6910  		fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6911 		if (fdp->fd_ifp == NULL)
6912 			fdp->fd_ifp = (void *)-1;
6913 	}
6914 }
6915 #endif /* _KERNEL */
6916 
6917 
6918 /* ------------------------------------------------------------------------ */
6919 /* Function:    fr_resolvenic                                               */
6920 /* Returns:     void* - NULL = wildcard name, -1 = failed to find NIC, else */
6921 /*                      pointer to interface structure for NIC              */
6922 /* Parameters:  name(I) - complete interface name                           */
6923 /*              v(I)    - IP protocol version                               */
6924 /*                                                                          */
6925 /* Look for a network interface structure that firstly has a matching name  */
6926 /* to that passed in and that is also being used for that IP protocol       */
6927 /* version (necessary on some platforms where there are separate listings   */
6928 /* for both IPv4 and IPv6 on the same physical NIC.                         */
6929 /*                                                                          */
6930 /* One might wonder why name gets terminated with a \0 byte in here.  The   */
6931 /* reason is an interface name could get into the kernel structures of ipf  */
6932 /* in any number of ways and so long as they all use the same sized array   */
6933 /* to put the name in, it makes sense to ensure it gets null terminated     */
6934 /* before it is used for its intended purpose - finding its match in the    */
6935 /* kernel's list of configured interfaces.                                  */
6936 /*                                                                          */
6937 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an     */
6938 /*       array for the name that is LIFNAMSIZ bytes (at least) in length.   */
6939 /* ------------------------------------------------------------------------ */
6940 void *fr_resolvenic(name, v, ifs)
6941 char *name;
6942 int v;
6943 ipf_stack_t *ifs;
6944 {
6945 	void *nic;
6946 
6947 	if (name[0] == '\0')
6948 		return NULL;
6949 
6950 	if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
6951 		return NULL;
6952 	}
6953 
6954 	name[LIFNAMSIZ - 1] = '\0';
6955 
6956 	nic = GETIFP(name, v, ifs);
6957 	if (nic == NULL)
6958 		nic = (void *)-1;
6959 	return nic;
6960 }
6961 
6962 
6963 /* ------------------------------------------------------------------------ */
6964 /* Function:    ipf_expiretokens                                            */
6965 /* Returns:     None.                                                       */
6966 /* Parameters:  ifs - ipf stack instance                                    */
6967 /*                                                                          */
6968 /* This function is run every ipf tick to see if there are any tokens that  */
6969 /* have been held for too long and need to be freed up.                     */
6970 /* ------------------------------------------------------------------------ */
6971 void ipf_expiretokens(ifs)
6972 ipf_stack_t *ifs;
6973 {
6974 	ipftoken_t *it;
6975 
6976 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
6977 	while ((it = ifs->ifs_ipftokenhead) != NULL) {
6978 		if (it->ipt_die > ifs->ifs_fr_ticks)
6979 			break;
6980 
6981 		ipf_freetoken(it, ifs);
6982 	}
6983 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
6984 }
6985 
6986 
6987 /* ------------------------------------------------------------------------ */
6988 /* Function:    ipf_deltoken                                                */
6989 /* Returns:     int - 0 = success, else error                               */
6990 /* Parameters:  type(I) - the token type to match                           */
6991 /*              uid(I)  - uid owning the token                              */
6992 /*              ptr(I)  - context pointer for the token                     */
6993 /*              ifs - ipf stack instance                                    */
6994 /*                                                                          */
6995 /* This function looks for a a token in the current list that matches up    */
6996 /* the fields (type, uid, ptr).  If none is found, ESRCH is returned, else  */
6997 /* call ipf_freetoken() to remove it from the list.                         */
6998 /* ------------------------------------------------------------------------ */
6999 int ipf_deltoken(type, uid, ptr, ifs)
7000 int type, uid;
7001 void *ptr;
7002 ipf_stack_t *ifs;
7003 {
7004 	ipftoken_t *it;
7005 	int error = ESRCH;
7006 
7007 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7008 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
7009 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7010 		    uid == it->ipt_uid) {
7011 			ipf_freetoken(it, ifs);
7012 			error = 0;
7013 			break;
7014 	}
7015 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7016 
7017 	return error;
7018 }
7019 
7020 
7021 /* ------------------------------------------------------------------------ */
7022 /* Function:    ipf_unlinktoken                                             */
7023 /* Returns:     None.                                                       */
7024 /* Parameters:  token(I) - pointer to token structure                       */
7025 /*              ifs - ipf stack instance                                    */
7026 /*                                                                          */
7027 /* This function unlinks a token structure from the linked list of tokens   */
7028 /* that it belongs to.  The head pointer never needs to be explicitly       */
7029 /* adjusted, but the tail does due to the linked list implementation.       */
7030 /* ------------------------------------------------------------------------ */
7031 static void ipf_unlinktoken(token, ifs)
7032 ipftoken_t *token;
7033 ipf_stack_t *ifs;
7034 {
7035 
7036 	if (ifs->ifs_ipftokentail == &token->ipt_next)
7037 		ifs->ifs_ipftokentail = token->ipt_pnext;
7038 
7039 	*token->ipt_pnext = token->ipt_next;
7040 	if (token->ipt_next != NULL)
7041 		token->ipt_next->ipt_pnext = token->ipt_pnext;
7042 }
7043 
7044 
7045 /* ------------------------------------------------------------------------ */
7046 /* Function:    ipf_findtoken                                               */
7047 /* Returns:     ipftoken_t * - NULL if no memory, else pointer to token     */
7048 /* Parameters:  type(I) - the token type to match                           */
7049 /*              uid(I) - uid owning the token                               */
7050 /*              ptr(I) - context pointer for the token                      */
7051 /*              ifs - ipf stack instance                                    */
7052 /*                                                                          */
7053 /* This function looks for a live token in the list of current tokens that  */
7054 /* matches the tuple (type, uid, ptr).  If one cannot be found then one is  */
7055 /* allocated.  If one is found then it is moved to the top of the list of   */
7056 /* currently active tokens.                                                 */
7057 /*                                                                          */
7058 /* NOTE: It is by design that this function returns holding a read lock on  */
7059 /*       ipf_tokens.  Callers must make sure they release it!               */
7060 /* ------------------------------------------------------------------------ */
7061 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
7062 int type, uid;
7063 void *ptr;
7064 ipf_stack_t *ifs;
7065 {
7066 	ipftoken_t *it, *new;
7067 
7068 	KMALLOC(new, ipftoken_t *);
7069 
7070 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7071 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
7072 		if (it->ipt_alive == 0)
7073 			continue;
7074 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7075 		    uid == it->ipt_uid)
7076 			break;
7077 	}
7078 
7079 	if (it == NULL) {
7080 		it = new;
7081 		new = NULL;
7082 		if (it == NULL)
7083 			return NULL;
7084 		it->ipt_data = NULL;
7085 		it->ipt_ctx = ptr;
7086 		it->ipt_uid = uid;
7087 		it->ipt_type = type;
7088 		it->ipt_next = NULL;
7089 		it->ipt_alive = 1;
7090 	} else {
7091 		if (new != NULL) {
7092 			KFREE(new);
7093 			new = NULL;
7094 		}
7095 
7096 		ipf_unlinktoken(it, ifs);
7097 	}
7098 	it->ipt_pnext = ifs->ifs_ipftokentail;
7099 	*ifs->ifs_ipftokentail = it;
7100 	ifs->ifs_ipftokentail = &it->ipt_next;
7101 	it->ipt_next = NULL;
7102 
7103 	it->ipt_die = ifs->ifs_fr_ticks + 2;
7104 
7105 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
7106 
7107 	return it;
7108 }
7109 
7110 
7111 /* ------------------------------------------------------------------------ */
7112 /* Function:    ipf_freetoken                                               */
7113 /* Returns:     None.                                                       */
7114 /* Parameters:  token(I) - pointer to token structure                       */
7115 /*              ifs - ipf stack instance                                    */
7116 /*                                                                          */
7117 /* This function unlinks a token from the linked list and on the path to    */
7118 /* free'ing the data, it calls the dereference function that is associated  */
7119 /* with the type of data pointed to by the token as it is considered to     */
7120 /* hold a reference to it.                                                  */
7121 /* ------------------------------------------------------------------------ */
7122 void ipf_freetoken(token, ifs)
7123 ipftoken_t *token;
7124 ipf_stack_t *ifs;
7125 {
7126 	void *data, **datap;
7127 
7128 	ipf_unlinktoken(token, ifs);
7129 
7130 	data = token->ipt_data;
7131 	datap = &data;
7132 
7133 	if ((data != NULL) && (data != (void *)-1)) {
7134 		switch (token->ipt_type)
7135 		{
7136 		case IPFGENITER_IPF :
7137 			(void)fr_derefrule((frentry_t **)datap, ifs);
7138 			break;
7139 		case IPFGENITER_IPNAT :
7140 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7141 			fr_ipnatderef((ipnat_t **)datap, ifs);
7142 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7143 			break;
7144 		case IPFGENITER_NAT :
7145 			fr_natderef((nat_t **)datap, ifs);
7146 			break;
7147 		case IPFGENITER_STATE :
7148 			fr_statederef((ipstate_t **)datap, ifs);
7149 			break;
7150 		case IPFGENITER_FRAG :
7151 			fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
7152 			break;
7153 		case IPFGENITER_NATFRAG :
7154  			fr_fragderef((ipfr_t **)datap,
7155 				     &ifs->ifs_ipf_natfrag, ifs);
7156 			break;
7157 		case IPFGENITER_HOSTMAP :
7158 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7159 			fr_hostmapdel((hostmap_t **)datap);
7160 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7161 			break;
7162 		default :
7163 			(void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7164 			break;
7165 		}
7166 	}
7167 
7168 	KFREE(token);
7169 }
7170 
7171 
7172 /* ------------------------------------------------------------------------ */
7173 /* Function:    ipf_getnextrule                                             */
7174 /* Returns:     int - 0 = success, else error                               */
7175 /* Parameters:  t(I)   - pointer to destination information to resolve      */
7176 /*              ptr(I) - pointer to ipfobj_t to copyin from user space      */
7177 /*              ifs - ipf stack instance                                    */
7178 /*                                                                          */
7179 /* This function's first job is to bring in the ipfruleiter_t structure via */
7180 /* the ipfobj_t structure to determine what should be the next rule to      */
7181 /* return. Once the ipfruleiter_t has been brought in, it then tries to     */
7182 /* find the 'next rule'.  This may include searching rule group lists or    */
7183 /* just be as simple as looking at the 'next' field in the rule structure.  */
7184 /* When we have found the rule to return, increase its reference count and  */
7185 /* if we used an existing rule to get here, decrease its reference count.   */
7186 /* ------------------------------------------------------------------------ */
7187 int ipf_getnextrule(t, ptr, ifs)
7188 ipftoken_t *t;
7189 void *ptr;
7190 ipf_stack_t *ifs;
7191 {
7192 	frentry_t *fr, *next, zero;
7193 	int error, out, count;
7194 	ipfruleiter_t it;
7195 	frgroup_t *fg;
7196 	char *dst;
7197 
7198 	if (t == NULL || ptr == NULL)
7199 		return EFAULT;
7200 	error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7201 	if (error != 0)
7202 		return error;
7203 	if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7204 		return EINVAL;
7205 	if ((it.iri_inout < 0) || (it.iri_inout > 3))
7206 		return EINVAL;
7207 	if (it.iri_nrules == 0)
7208 		return EINVAL;
7209 	if ((it.iri_active != 0) && (it.iri_active != 1))
7210 		return EINVAL;
7211 	if (it.iri_rule == NULL)
7212 		return EFAULT;
7213 
7214 	/*
7215 	 * Use bitmask on it.iri_inout to determine direction.
7216 	 * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7217 	 * F_IN (0) and F_ACIN (2) mask to out = 0.
7218 	 */
7219 	out = it.iri_inout & F_OUT;
7220 	READ_ENTER(&ifs->ifs_ipf_mutex);
7221 
7222 	/*
7223 	 * Retrieve "previous" entry from token and find the next entry.
7224 	 */
7225 	fr = t->ipt_data;
7226 	if (fr == NULL) {
7227 		if (*it.iri_group == '\0') {
7228 			/*
7229 			 * Use bitmask again to determine accounting or not.
7230 			 * F_ACIN will mask to accounting cases F_ACIN (2)
7231 			 * or F_ACOUT (3), but not F_IN or F_OUT.
7232 			 */
7233 			if ((it.iri_inout & F_ACIN) != 0) {
7234 				if (it.iri_ver == AF_INET)
7235 					next = ifs->ifs_ipacct
7236 					    [out][it.iri_active];
7237 				else
7238 					next = ifs->ifs_ipacct6
7239 					    [out][it.iri_active];
7240 			} else {
7241 				if (it.iri_ver == AF_INET)
7242 					next = ifs->ifs_ipfilter
7243 					    [out][it.iri_active];
7244 				else
7245 					next = ifs->ifs_ipfilter6
7246 					    [out][it.iri_active];
7247 			}
7248 		} else {
7249 			fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7250 					  it.iri_active, NULL, ifs);
7251 			if (fg != NULL)
7252 				next = fg->fg_start;
7253 			else
7254 				next = NULL;
7255 		}
7256 	} else {
7257 		next = fr->fr_next;
7258 	}
7259 
7260 	dst = (char *)it.iri_rule;
7261 	/*
7262 	 * The ipfruleiter may ask for more than 1 rule at a time to be
7263 	 * copied out, so long as that many exist in the list to start with!
7264 	 */
7265 	for (count = it.iri_nrules; count > 0; count--) {
7266 		/*
7267 		 * If we found an entry, add reference to it and update token.
7268 		 * Otherwise, zero out data to be returned and NULL out token.
7269 		 */
7270 		if (next != NULL) {
7271 			MUTEX_ENTER(&next->fr_lock);
7272 			next->fr_ref++;
7273 			MUTEX_EXIT(&next->fr_lock);
7274 			t->ipt_data = next;
7275 		} else {
7276 			bzero(&zero, sizeof(zero));
7277 			next = &zero;
7278 			t->ipt_data = NULL;
7279 		}
7280 
7281 		/*
7282 		 * Now that we have ref, it's save to give up lock.
7283 		 */
7284 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7285 
7286 		/*
7287 		 * Copy out data and clean up references and token as needed.
7288 		 */
7289 		error = COPYOUT(next, dst, sizeof(*next));
7290 		if (error != 0)
7291 			error = EFAULT;
7292 		if (t->ipt_data == NULL) {
7293 			ipf_freetoken(t, ifs);
7294 			break;
7295 		} else {
7296 			if (fr != NULL)
7297 				(void) fr_derefrule(&fr, ifs);
7298 			if (next->fr_data != NULL) {
7299 				dst += sizeof(*next);
7300 				error = COPYOUT(next->fr_data, dst,
7301 						next->fr_dsize);
7302 				if (error != 0)
7303 					error = EFAULT;
7304 				else
7305 					dst += next->fr_dsize;
7306 			}
7307 			if (next->fr_next == NULL) {
7308 				ipf_freetoken(t, ifs);
7309 				break;
7310 			}
7311 		}
7312 
7313 		if ((count == 1) || (error != 0))
7314 			break;
7315 
7316 		READ_ENTER(&ifs->ifs_ipf_mutex);
7317 		fr = next;
7318 		next = fr->fr_next;
7319 	}
7320 
7321 	return error;
7322 }
7323 
7324 
7325 /* ------------------------------------------------------------------------ */
7326 /* Function:    fr_frruleiter                                               */
7327 /* Returns:     int - 0 = success, else error                               */
7328 /* Parameters:  data(I) - the token type to match                           */
7329 /*              uid(I) - uid owning the token                               */
7330 /*              ptr(I) - context pointer for the token                      */
7331 /*              ifs - ipf stack instance                                    */
7332 /*                                                                          */
7333 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7334 /* ipf_getnextrule.  It's role is to find the right token in the kernel for */
7335 /* the process doing the ioctl and use that to ask for the next rule.       */
7336 /* ------------------------------------------------------------------------ */
7337 int ipf_frruleiter(data, uid, ctx, ifs)
7338 void *data, *ctx;
7339 int uid;
7340 ipf_stack_t *ifs;
7341 {
7342 	ipftoken_t *token;
7343 	int error;
7344 
7345 	token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7346 	if (token != NULL)
7347 		error = ipf_getnextrule(token, data, ifs);
7348 	else
7349 		error = EFAULT;
7350 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7351 
7352 	return error;
7353 }
7354 
7355 
7356 /* ------------------------------------------------------------------------ */
7357 /* Function:    ipf_geniter                                                 */
7358 /* Returns:     int - 0 = success, else error                               */
7359 /* Parameters:  token(I) - pointer to ipftoken structure                    */
7360 /*              itp(I) - pointer to ipfgeniter structure                    */
7361 /*              ifs - ipf stack instance                                    */
7362 /*                                                                          */
7363 /* Generic iterator called from ipf_genericiter.  Currently only used for   */
7364 /* walking through list of fragments.                                       */
7365 /* ------------------------------------------------------------------------ */
7366 int ipf_geniter(token, itp, ifs)
7367 ipftoken_t *token;
7368 ipfgeniter_t *itp;
7369 ipf_stack_t *ifs;
7370 {
7371 	int error;
7372 
7373 	switch (itp->igi_type)
7374 	{
7375 	case IPFGENITER_FRAG :
7376 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7377 				    &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7378 				    ifs);
7379 		break;
7380 	default :
7381 		error = EINVAL;
7382 		break;
7383 	}
7384 
7385 	return error;
7386 }
7387 
7388 
7389 /* ------------------------------------------------------------------------ */
7390 /* Function:    ipf_genericiter                                             */
7391 /* Returns:     int - 0 = success, else error                               */
7392 /* Parameters:  data(I) - the token type to match                           */
7393 /*              uid(I) - uid owning the token                               */
7394 /*              ptr(I) - context pointer for the token                      */
7395 /*              ifs - ipf stack instance                                    */
7396 /*                                                                          */
7397 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7398 /* ipf_geniter when handling SIOCGENITER.  It's role is to find the right   */
7399 /* token in the kernel for the process using the ioctl, and to use that     */
7400 /* token when calling ipf_geniter.                                          */
7401 /* ------------------------------------------------------------------------ */
7402 int ipf_genericiter(data, uid, ctx, ifs)
7403 void *data, *ctx;
7404 int uid;
7405 ipf_stack_t *ifs;
7406 {
7407 	ipftoken_t *token;
7408 	ipfgeniter_t iter;
7409 	int error;
7410 
7411 	error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7412 	if (error != 0)
7413 		return error;
7414 
7415 	token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7416 	if (token != NULL) {
7417 		token->ipt_subtype = iter.igi_type;
7418 		error = ipf_geniter(token, &iter, ifs);
7419 	} else
7420 		error = EFAULT;
7421 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7422 
7423 	return error;
7424 }
7425 
7426 
7427 /* --------------------------------------------------------------------- */
7428 /* Function:    ipf_earlydrop                                            */
7429 /* Returns:     number of dropped/removed entries from the queue         */
7430 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7431 /*              ifq	- pointer to queue with entries to be deleted    */
7432 /*              idletime - entry must be idle this long to be deleted    */
7433 /*              ifs     - ipf stack instance                             */
7434 /*                                                                       */
7435 /* Function is invoked from state/NAT flush routines to remove entries   */
7436 /* from specified timeout queue, based on how long they've sat idle,     */
7437 /* without waiting for it to happen on its own.                          */
7438 /* --------------------------------------------------------------------- */
7439 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7440 int flushtype;
7441 ipftq_t *ifq;
7442 int idletime;
7443 ipf_stack_t *ifs;
7444 {
7445         ipftqent_t *tqe, *tqn;
7446         unsigned int dropped;
7447         int droptick;
7448 	void *ent;
7449 
7450         if (ifq == NULL)
7451                 return (0);
7452 
7453         dropped = 0;
7454 
7455         /*
7456          * Determine the tick representing the idle time we're interested
7457          * in.  If an entry exists in the queue, and it was touched before
7458          * that tick, then it's been idle longer than idletime, so it should
7459 	 * be deleted.
7460          */
7461         droptick = ifs->ifs_fr_ticks - idletime;
7462         tqn = ifq->ifq_head;
7463         while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7464                 tqn = tqe->tqe_next;
7465 		ent = tqe->tqe_parent;
7466 		switch (flushtype)
7467 		{
7468 		case NAT_FLUSH:
7469 			if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7470 				dropped++;
7471 			break;
7472 		case STATE_FLUSH:
7473 			if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7474 				dropped++;
7475 			break;
7476 		default:
7477 			return (0);
7478 		}
7479         }
7480         return (dropped);
7481 }
7482 
7483 
7484 /* --------------------------------------------------------------------- */
7485 /* Function:    ipf_flushclosing                                         */
7486 /* Returns:     int - number of entries deleted                          */
7487 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7488 /*              stateval - TCP state at which to start removing entries  */
7489 /*              ipfqs - pointer to timeout queues                        */
7490 /*              userqs - pointer to user defined queues                  */
7491 /*              ifs  - ipf stack instance                                */
7492 /*                                                                       */
7493 /* Remove state/NAT table entries for TCP connections which are in the   */
7494 /* process of closing, and have at least reached the state specified by  */
7495 /* the 'stateval' parameter.                                             */
7496 /* --------------------------------------------------------------------- */
7497 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7498 int flushtype, stateval;
7499 ipftq_t *ipfqs, *userqs;
7500 ipf_stack_t *ifs;
7501 {
7502 	ipftq_t *ifq, *ifqn;
7503         ipftqent_t *tqe, *tqn;
7504         int dropped;
7505 	void *ent;
7506 	nat_t *nat;
7507 	ipstate_t *is;
7508 
7509         dropped = 0;
7510 
7511         /*
7512          * Start by deleting any entries in specific timeout queues.
7513          */
7514 	ifqn = &ipfqs[stateval];
7515         while ((ifq = ifqn) != NULL) {
7516                 ifqn = ifq->ifq_next;
7517                 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7518         }
7519 
7520         /*
7521          * Next, look through user defined queues for closing entries.
7522          */
7523 	ifqn = userqs;
7524         while ((ifq = ifqn) != NULL) {
7525                 ifqn = ifq->ifq_next;
7526                 tqn = ifq->ifq_head;
7527                 while ((tqe = tqn) != NULL) {
7528                         tqn = tqe->tqe_next;
7529 			ent = tqe->tqe_parent;
7530 			switch (flushtype)
7531 			{
7532 			case NAT_FLUSH:
7533 				nat = (nat_t *)ent;
7534 				if ((nat->nat_p == IPPROTO_TCP) &&
7535 				    (nat->nat_tcpstate[0] >= stateval) &&
7536 				    (nat->nat_tcpstate[1] >= stateval) &&
7537 				    (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7538 					dropped++;
7539 				break;
7540 			case STATE_FLUSH:
7541 				is = (ipstate_t *)ent;
7542 				if ((is->is_p == IPPROTO_TCP) &&
7543 				    (is->is_state[0] >= stateval) &&
7544 				    (is->is_state[1] >= stateval) &&
7545 				    (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7546 					dropped++;
7547 				break;
7548 			default:
7549 				return (0);
7550 			}
7551                 }
7552         }
7553         return (dropped);
7554 }
7555 
7556 
7557 /* --------------------------------------------------------------------- */
7558 /* Function:    ipf_extraflush                                           */
7559 /* Returns:     int - number of entries flushed (0 = none)               */
7560 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7561 /*              ipfqs - pointer to 'established' timeout queue           */
7562 /*              userqs - pointer to user defined queues                  */
7563 /*              ifs  - ipf stack instance                                */
7564 /*                                                                       */
7565 /* This function gets called when either NAT or state tables fill up.    */
7566 /* We need to try a bit harder to free up some space.  The function will */
7567 /* flush entries for TCP connections which have been idle a long time.   */
7568 /*                                                                       */
7569 /* Currently, the idle time is checked using values from ideltime_tab[]	 */
7570 /* --------------------------------------------------------------------- */
7571 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7572 int flushtype;
7573 ipftq_t *ipfqs, *userqs;
7574 ipf_stack_t *ifs;
7575 {
7576 	ipftq_t *ifq, *ifqn;
7577 	int idletime, removed, idle_idx;
7578 
7579 	removed = 0;
7580 
7581 	/*
7582   	 * Determine initial threshold for minimum idle time based on
7583 	 * how long ipfilter has been running.  Ipfilter needs to have
7584 	 * been up as long as the smallest interval to continue on.
7585 	 *
7586 	 * Minimum idle times stored in idletime_tab and indexed by
7587 	 * idle_idx.  Start at upper end of array and work backwards.
7588 	 *
7589 	 * Once the index is found, set the initial idle time to the
7590 	 * first interval before the current ipfilter run time.
7591 	 */
7592 	if (ifs->ifs_fr_ticks < idletime_tab[0])
7593 		return (0);
7594 	idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7595 	if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7596 		idletime = idletime_tab[idle_idx];
7597 	} else {
7598 		while ((idle_idx > 0) &&
7599 		    (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7600 			idle_idx--;
7601 
7602 		idletime = (ifs->ifs_fr_ticks /
7603 			    idletime_tab[idle_idx]) *
7604 			    idletime_tab[idle_idx];
7605 	}
7606 
7607 	while (idle_idx >= 0) {
7608 		/*
7609 		 * Check to see if we need to delete more entries.
7610 		 * If we do, start with appropriate timeout queue.
7611 		 */
7612 		if (flushtype == NAT_FLUSH) {
7613 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7614 			    ifs->ifs_nat_flush_level_lo)
7615 				break;
7616 		} else if (flushtype == STATE_FLUSH) {
7617 			if (ST_TAB_WATER_LEVEL(ifs) <=
7618 			    ifs->ifs_state_flush_level_lo)
7619 				break;
7620 		} else {
7621 			break;
7622 		}
7623 
7624 		removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7625 
7626 		/*
7627 		 * Next, check the user defined queues.  But first, make
7628 		 * certain that timeout queue deletions didn't do enough.
7629 		 */
7630 		if (flushtype == NAT_FLUSH) {
7631 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7632 			    ifs->ifs_nat_flush_level_lo)
7633 				break;
7634 		} else {
7635 			if (ST_TAB_WATER_LEVEL(ifs) <=
7636 			    ifs->ifs_state_flush_level_lo)
7637 				break;
7638 		}
7639 		ifqn = userqs;
7640 		while ((ifq = ifqn) != NULL) {
7641 			ifqn = ifq->ifq_next;
7642 			removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7643 		}
7644 
7645 		/*
7646 		 * Adjust the granularity of idle time.
7647 		 *
7648 		 * If we reach an interval boundary, we need to
7649 		 * either adjust the idle time accordingly or exit
7650 		 * the loop altogether (if this is very last check).
7651 		 */
7652 		idletime -= idletime_tab[idle_idx];
7653 		if (idletime < idletime_tab[idle_idx]) {
7654 			if (idle_idx != 0) {
7655 				idletime = idletime_tab[idle_idx] -
7656 				idletime_tab[idle_idx - 1];
7657 				idle_idx--;
7658 			} else {
7659 				break;
7660 			}
7661 		}
7662 	}
7663 
7664 	return (removed);
7665 }
7666