xref: /illumos-gate/usr/src/uts/common/inet/ipf/fil.c (revision d0dd088c)
1 /*
2  * Copyright (C) 1993-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
7  */
8 
9 #if defined(KERNEL) || defined(_KERNEL)
10 # undef KERNEL
11 # undef _KERNEL
12 # define        KERNEL	1
13 # define        _KERNEL	1
14 #endif
15 #include <sys/errno.h>
16 #include <sys/types.h>
17 #include <sys/param.h>
18 #include <sys/time.h>
19 #if defined(__NetBSD__)
20 # if (NetBSD >= 199905) && !defined(IPFILTER_LKM) && defined(_KERNEL)
21 #  include "opt_ipfilter_log.h"
22 # endif
23 #endif
24 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
25     (__FreeBSD_version >= 220000)
26 # if (__FreeBSD_version >= 400000)
27 #  if !defined(IPFILTER_LKM)
28 #   include "opt_inet6.h"
29 #  endif
30 #  if (__FreeBSD_version == 400019)
31 #   define CSUM_DELAY_DATA
32 #  endif
33 # endif
34 # include <sys/filio.h>
35 #else
36 # include <sys/ioctl.h>
37 #endif
38 #if !defined(_AIX51)
39 # include <sys/fcntl.h>
40 #endif
41 #if defined(_KERNEL)
42 # include <sys/systm.h>
43 # include <sys/file.h>
44 #else
45 # include <stdio.h>
46 # include <string.h>
47 # include <stdlib.h>
48 # include <stddef.h>
49 # include <sys/file.h>
50 # define _KERNEL
51 # ifdef __OpenBSD__
52 struct file;
53 # endif
54 # include <sys/uio.h>
55 # undef _KERNEL
56 #endif
57 #if !defined(__SVR4) && !defined(__svr4__) && !defined(__hpux) && \
58     !defined(linux)
59 # include <sys/mbuf.h>
60 #else
61 # if !defined(linux)
62 #  include <sys/byteorder.h>
63 # endif
64 # if (SOLARIS2 < 5) && defined(sun)
65 #  include <sys/dditypes.h>
66 # endif
67 #endif
68 #ifdef __hpux
69 # define _NET_ROUTE_INCLUDED
70 #endif
71 #if !defined(linux)
72 # include <sys/protosw.h>
73 #endif
74 #include <sys/socket.h>
75 #include <net/if.h>
76 #ifdef sun
77 # include <net/af.h>
78 #endif
79 #if !defined(_KERNEL) && defined(__FreeBSD__)
80 # include "radix_ipf.h"
81 #endif
82 #include <net/route.h>
83 #include <netinet/in.h>
84 #include <netinet/in_systm.h>
85 #include <netinet/ip.h>
86 #if !defined(linux)
87 # include <netinet/ip_var.h>
88 #endif
89 #if defined(__sgi) && defined(IFF_DRVRLOCK) /* IRIX 6 */
90 # include <sys/hashing.h>
91 # include <netinet/in_var.h>
92 #endif
93 #include <netinet/tcp.h>
94 #if (!defined(__sgi) && !defined(AIX)) || defined(_KERNEL)
95 # include <netinet/udp.h>
96 # include <netinet/ip_icmp.h>
97 #endif
98 #ifdef __hpux
99 # undef _NET_ROUTE_INCLUDED
100 #endif
101 #include "netinet/ip_compat.h"
102 #ifdef	USE_INET6
103 # include <netinet/icmp6.h>
104 # if !SOLARIS && defined(_KERNEL) && !defined(__osf__) && !defined(__hpux)
105 #  include <netinet6/in6_var.h>
106 # endif
107 #endif
108 #include <netinet/tcpip.h>
109 #include "netinet/ip_fil.h"
110 #include "netinet/ip_nat.h"
111 #include "netinet/ip_frag.h"
112 #include "netinet/ip_state.h"
113 #include "netinet/ip_proxy.h"
114 #include "netinet/ip_auth.h"
115 #include "netinet/ipf_stack.h"
116 #ifdef IPFILTER_SCAN
117 # include "netinet/ip_scan.h"
118 #endif
119 #ifdef IPFILTER_SYNC
120 # include "netinet/ip_sync.h"
121 #endif
122 #include "netinet/ip_pool.h"
123 #include "netinet/ip_htable.h"
124 #ifdef IPFILTER_COMPILED
125 # include "netinet/ip_rules.h"
126 #endif
127 #if defined(IPFILTER_BPF) && defined(_KERNEL)
128 # include <net/bpf.h>
129 #endif
130 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
131 # include <sys/malloc.h>
132 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
133 #  include "opt_ipfilter.h"
134 # endif
135 #endif
136 #include "netinet/ipl.h"
137 /* END OF INCLUDES */
138 
139 #if !defined(lint)
140 static const char sccsid[] = "@(#)fil.c	1.36 6/5/96 (C) 1993-2000 Darren Reed";
141 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
142 #endif
143 
144 #ifndef	_KERNEL
145 # include "ipf.h"
146 # include "ipt.h"
147 # include "bpf-ipf.h"
148 extern	int	opts;
149 
150 # define	FR_VERBOSE(verb_pr)			verbose verb_pr
151 # define	FR_DEBUG(verb_pr)			debug verb_pr
152 #else /* #ifndef _KERNEL */
153 # define	FR_VERBOSE(verb_pr)
154 # define	FR_DEBUG(verb_pr)
155 #endif /* _KERNEL */
156 
157 
158 char	ipfilter_version[] = IPL_VERSION;
159 int	fr_features = 0
160 #ifdef	IPFILTER_LKM
161 		| IPF_FEAT_LKM
162 #endif
163 #ifdef	IPFILTER_LOG
164 		| IPF_FEAT_LOG
165 #endif
166 #ifdef	IPFILTER_LOOKUP
167 		| IPF_FEAT_LOOKUP
168 #endif
169 #ifdef	IPFILTER_BPF
170 		| IPF_FEAT_BPF
171 #endif
172 #ifdef	IPFILTER_COMPILED
173 		| IPF_FEAT_COMPILED
174 #endif
175 #ifdef	IPFILTER_CKSUM
176 		| IPF_FEAT_CKSUM
177 #endif
178 #ifdef	IPFILTER_SYNC
179 		| IPF_FEAT_SYNC
180 #endif
181 #ifdef	IPFILTER_SCAN
182 		| IPF_FEAT_SCAN
183 #endif
184 #ifdef	USE_INET6
185 		| IPF_FEAT_IPV6
186 #endif
187 	;
188 
189 #define	IPF_BUMP(x)	(x)++
190 
191 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
192 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
193 static	int		fr_portcheck __P((frpcmp_t *, u_short *));
194 static	int		frflushlist __P((int, minor_t, int *, frentry_t **,
195 					 ipf_stack_t *));
196 static	ipfunc_t	fr_findfunc __P((ipfunc_t));
197 static	frentry_t	*fr_firewall __P((fr_info_t *, u_32_t *));
198 static	int		fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
199 static	INLINE void	frpr_ah __P((fr_info_t *));
200 static	INLINE void	frpr_esp __P((fr_info_t *));
201 static	INLINE void	frpr_gre __P((fr_info_t *));
202 static	INLINE void	frpr_udp __P((fr_info_t *));
203 static	INLINE void	frpr_tcp __P((fr_info_t *));
204 static	INLINE void	frpr_icmp __P((fr_info_t *));
205 static	INLINE void	frpr_ipv4hdr __P((fr_info_t *));
206 static	INLINE int	frpr_pullup __P((fr_info_t *, int));
207 static	INLINE void	frpr_short __P((fr_info_t *, int));
208 static	INLINE void	frpr_tcpcommon __P((fr_info_t *));
209 static	INLINE void	frpr_udpcommon __P((fr_info_t *));
210 static	INLINE int	fr_updateipid __P((fr_info_t *));
211 #ifdef	IPFILTER_LOOKUP
212 static	int		fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
213 static	INLINE void	*fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
214 					       ipf_stack_t *));
215 #endif
216 static	void		frsynclist __P((int, int, void *, char *, frentry_t *,
217     ipf_stack_t *));
218 static	void		*fr_ifsync __P((int, int, char *, char *,
219 					void *, void *, ipf_stack_t *));
220 static	ipftuneable_t	*fr_findtunebyname __P((const char *, ipf_stack_t *));
221 static	ipftuneable_t	*fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
222 
223 /*
224  * bit values for identifying presence of individual IP options
225  * All of these tables should be ordered by increasing key value on the left
226  * hand side to allow for binary searching of the array and include a trailer
227  * with a 0 for the bitmask for linear searches to easily find the end with.
228  */
229 const	struct	optlist	ipopts[20] = {
230 	{ IPOPT_NOP,	0x000001 },
231 	{ IPOPT_RR,	0x000002 },
232 	{ IPOPT_ZSU,	0x000004 },
233 	{ IPOPT_MTUP,	0x000008 },
234 	{ IPOPT_MTUR,	0x000010 },
235 	{ IPOPT_ENCODE,	0x000020 },
236 	{ IPOPT_TS,	0x000040 },
237 	{ IPOPT_TR,	0x000080 },
238 	{ IPOPT_SECURITY, 0x000100 },
239 	{ IPOPT_LSRR,	0x000200 },
240 	{ IPOPT_E_SEC,	0x000400 },
241 	{ IPOPT_CIPSO,	0x000800 },
242 	{ IPOPT_SATID,	0x001000 },
243 	{ IPOPT_SSRR,	0x002000 },
244 	{ IPOPT_ADDEXT,	0x004000 },
245 	{ IPOPT_VISA,	0x008000 },
246 	{ IPOPT_IMITD,	0x010000 },
247 	{ IPOPT_EIP,	0x020000 },
248 	{ IPOPT_FINN,	0x040000 },
249 	{ 0,		0x000000 }
250 };
251 
252 #ifdef USE_INET6
253 struct optlist ip6exthdr[] = {
254 	{ IPPROTO_HOPOPTS,		0x000001 },
255 	{ IPPROTO_IPV6,			0x000002 },
256 	{ IPPROTO_ROUTING,		0x000004 },
257 	{ IPPROTO_FRAGMENT,		0x000008 },
258 	{ IPPROTO_ESP,			0x000010 },
259 	{ IPPROTO_AH,			0x000020 },
260 	{ IPPROTO_NONE,			0x000040 },
261 	{ IPPROTO_DSTOPTS,		0x000080 },
262 	{ 0,				0 }
263 };
264 #endif
265 
266 struct optlist tcpopts[] = {
267 	{ TCPOPT_NOP,			0x000001 },
268 	{ TCPOPT_MAXSEG,		0x000002 },
269 	{ TCPOPT_WINDOW,		0x000004 },
270 	{ TCPOPT_SACK_PERMITTED,	0x000008 },
271 	{ TCPOPT_SACK,			0x000010 },
272 	{ TCPOPT_TIMESTAMP,		0x000020 },
273 	{ 0,				0x000000 }
274 };
275 
276 /*
277  * bit values for identifying presence of individual IP security options
278  */
279 const	struct	optlist	secopt[8] = {
280 	{ IPSO_CLASS_RES4,	0x01 },
281 	{ IPSO_CLASS_TOPS,	0x02 },
282 	{ IPSO_CLASS_SECR,	0x04 },
283 	{ IPSO_CLASS_RES3,	0x08 },
284 	{ IPSO_CLASS_CONF,	0x10 },
285 	{ IPSO_CLASS_UNCL,	0x20 },
286 	{ IPSO_CLASS_RES2,	0x40 },
287 	{ IPSO_CLASS_RES1,	0x80 }
288 };
289 
290 
291 /*
292  * Table of functions available for use with call rules.
293  */
294 static ipfunc_resolve_t fr_availfuncs[] = {
295 #ifdef	IPFILTER_LOOKUP
296 	{ "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
297 	{ "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
298 #endif
299 	{ "", NULL }
300 };
301 
302 
303 /*
304  * Below we declare a list of constants used only by the ipf_extraflush()
305  * routine.  We are placing it here, instead of in ipf_extraflush() itself,
306  * because we want to make it visible to tools such as mdb, nm etc., so the
307  * values can easily be altered during debugging.
308  */
309 static	const	int	idletime_tab[] = {
310 	IPF_TTLVAL(30),		/* 30 seconds */
311 	IPF_TTLVAL(1800),	/* 30 minutes */
312 	IPF_TTLVAL(43200),	/* 12 hours */
313 	IPF_TTLVAL(345600),	/* 4 days */
314 };
315 
316 
317 /*
318  * The next section of code is a a collection of small routines that set
319  * fields in the fr_info_t structure passed based on properties of the
320  * current packet.  There are different routines for the same protocol
321  * for each of IPv4 and IPv6.  Adding a new protocol, for which there
322  * will "special" inspection for setup, is now more easily done by adding
323  * a new routine and expanding the frpr_ipinit*() function rather than by
324  * adding more code to a growing switch statement.
325  */
326 #ifdef USE_INET6
327 static	INLINE int	frpr_ah6 __P((fr_info_t *));
328 static	INLINE void	frpr_esp6 __P((fr_info_t *));
329 static	INLINE void	frpr_gre6 __P((fr_info_t *));
330 static	INLINE void	frpr_udp6 __P((fr_info_t *));
331 static	INLINE void	frpr_tcp6 __P((fr_info_t *));
332 static	INLINE void	frpr_icmp6 __P((fr_info_t *));
333 static	INLINE void	frpr_ipv6hdr __P((fr_info_t *));
334 static	INLINE void	frpr_short6 __P((fr_info_t *, int));
335 static	INLINE int	frpr_hopopts6 __P((fr_info_t *));
336 static	INLINE int	frpr_routing6 __P((fr_info_t *));
337 static	INLINE int	frpr_dstopts6 __P((fr_info_t *));
338 static	INLINE int	frpr_fragment6 __P((fr_info_t *));
339 static	INLINE int	frpr_ipv6exthdr __P((fr_info_t *, int, int));
340 
341 
342 /* ------------------------------------------------------------------------ */
343 /* Function:    frpr_short6                                                 */
344 /* Returns:     void                                                        */
345 /* Parameters:  fin(I) - pointer to packet information                      */
346 /*                                                                          */
347 /* IPv6 Only                                                                */
348 /* This is function enforces the 'is a packet too short to be legit' rule   */
349 /* for IPv6 and marks the packet with FI_SHORT if so.  See function comment */
350 /* for frpr_short() for more details.                                       */
351 /* ------------------------------------------------------------------------ */
352 static INLINE void frpr_short6(fin, xmin)
353 fr_info_t *fin;
354 int xmin;
355 {
356 
357 	if (fin->fin_dlen < xmin)
358 		fin->fin_flx |= FI_SHORT;
359 }
360 
361 
362 /* ------------------------------------------------------------------------ */
363 /* Function:    frpr_ipv6hdr                                                */
364 /* Returns:     Nil                                                         */
365 /* Parameters:  fin(I) - pointer to packet information                      */
366 /*                                                                          */
367 /* IPv6 Only                                                                */
368 /* Copy values from the IPv6 header into the fr_info_t struct and call the  */
369 /* per-protocol analyzer if it exists.                                      */
370 /* ------------------------------------------------------------------------ */
371 static INLINE void frpr_ipv6hdr(fin)
372 fr_info_t *fin;
373 {
374 	ip6_t *ip6 = (ip6_t *)fin->fin_ip;
375 	int p, go = 1, i, hdrcount;
376 	fr_ip_t *fi = &fin->fin_fi;
377 
378 	fin->fin_off = 0;
379 
380 	fi->fi_tos = 0;
381 	fi->fi_optmsk = 0;
382 	fi->fi_secmsk = 0;
383 	fi->fi_auth = 0;
384 
385 	p = ip6->ip6_nxt;
386 	fi->fi_ttl = ip6->ip6_hlim;
387 	fi->fi_src.in6 = ip6->ip6_src;
388 	fi->fi_dst.in6 = ip6->ip6_dst;
389 	fin->fin_id = 0;
390 
391 	hdrcount = 0;
392 	while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
393 		switch (p)
394 		{
395 		case IPPROTO_UDP :
396 			frpr_udp6(fin);
397 			go = 0;
398 			break;
399 
400 		case IPPROTO_TCP :
401 			frpr_tcp6(fin);
402 			go = 0;
403 			break;
404 
405 		case IPPROTO_ICMPV6 :
406 			frpr_icmp6(fin);
407 			go = 0;
408 			break;
409 
410 		case IPPROTO_GRE :
411 			frpr_gre6(fin);
412 			go = 0;
413 			break;
414 
415 		case IPPROTO_HOPOPTS :
416 			/*
417 			 * hop by hop ext header is only allowed
418 			 * right after IPv6 header.
419 			 */
420 			if (hdrcount != 0) {
421 				fin->fin_flx |= FI_BAD;
422 				p = IPPROTO_NONE;
423 			} else {
424 				p = frpr_hopopts6(fin);
425 			}
426 			break;
427 
428 		case IPPROTO_DSTOPTS :
429 			p = frpr_dstopts6(fin);
430 			break;
431 
432 		case IPPROTO_ROUTING :
433 			p = frpr_routing6(fin);
434 			break;
435 
436 		case IPPROTO_AH :
437 			p = frpr_ah6(fin);
438 			break;
439 
440 		case IPPROTO_ESP :
441 			frpr_esp6(fin);
442 			go = 0;
443 			break;
444 
445 		case IPPROTO_IPV6 :
446 			for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
447 				if (ip6exthdr[i].ol_val == p) {
448 					fin->fin_flx |= ip6exthdr[i].ol_bit;
449 					break;
450 				}
451 			go = 0;
452 			break;
453 
454 		case IPPROTO_NONE :
455 			go = 0;
456 			break;
457 
458 		case IPPROTO_FRAGMENT :
459 			p = frpr_fragment6(fin);
460 			if (fin->fin_off != 0)  /* Not the first frag */
461 				go = 0;
462 			break;
463 
464 		default :
465 			go = 0;
466 			break;
467 		}
468 		hdrcount++;
469 
470 		/*
471 		 * It is important to note that at this point, for the
472 		 * extension headers (go != 0), the entire header may not have
473 		 * been pulled up when the code gets to this point.  This is
474 		 * only done for "go != 0" because the other header handlers
475 		 * will all pullup their complete header.  The other indicator
476 		 * of an incomplete packet is that this was just an extension
477 		 * header.
478 		 */
479 		if ((go != 0) && (p != IPPROTO_NONE) &&
480 		    (frpr_pullup(fin, 0) == -1)) {
481 			p = IPPROTO_NONE;
482 			go = 0;
483 		}
484 	}
485 	fi->fi_p = p;
486 }
487 
488 
489 /* ------------------------------------------------------------------------ */
490 /* Function:    frpr_ipv6exthdr                                             */
491 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
492 /* Parameters:  fin(I)      - pointer to packet information                 */
493 /*              multiple(I) - flag indicating yes/no if multiple occurances */
494 /*                            of this extension header are allowed.         */
495 /*              proto(I)    - protocol number for this extension header     */
496 /*                                                                          */
497 /* IPv6 Only                                                                */
498 /* This function expects to find an IPv6 extension header at fin_dp.        */
499 /* There must be at least 8 bytes of data at fin_dp for there to be a valid */
500 /* extension header present. If a good one is found, fin_dp is advanced to  */
501 /* point at the first piece of data after the extension header, fin_exthdr  */
502 /* points to the start of the extension header and the "protocol" of the    */
503 /* *NEXT* header is returned.                                               */
504 /* ------------------------------------------------------------------------ */
505 static INLINE int frpr_ipv6exthdr(fin, multiple, proto)
506 fr_info_t *fin;
507 int multiple, proto;
508 {
509 	struct ip6_ext *hdr;
510 	u_short shift;
511 	int i;
512 
513 	fin->fin_flx |= FI_V6EXTHDR;
514 
515 				/* 8 is default length of extension hdr */
516 	if ((fin->fin_dlen - 8) < 0) {
517 		fin->fin_flx |= FI_SHORT;
518 		return IPPROTO_NONE;
519 	}
520 
521 	if (frpr_pullup(fin, 8) == -1)
522 		return IPPROTO_NONE;
523 
524 	hdr = fin->fin_dp;
525 	shift = 8 + (hdr->ip6e_len << 3);
526 	if (shift > fin->fin_dlen) {	/* Nasty extension header length? */
527 		fin->fin_flx |= FI_BAD;
528 		return IPPROTO_NONE;
529 	}
530 
531 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
532 		if (ip6exthdr[i].ol_val == proto) {
533 			/*
534 			 * Most IPv6 extension headers are only allowed once.
535 			 */
536 			if ((multiple == 0) &&
537 			    ((fin->fin_optmsk & ip6exthdr[i].ol_bit) != 0))
538 				fin->fin_flx |= FI_BAD;
539 			else
540 				fin->fin_optmsk |= ip6exthdr[i].ol_bit;
541 			break;
542 		}
543 
544 	fin->fin_dp = (char *)fin->fin_dp + shift;
545 	fin->fin_dlen -= shift;
546 
547 	return hdr->ip6e_nxt;
548 }
549 
550 
551 /* ------------------------------------------------------------------------ */
552 /* Function:    frpr_hopopts6                                               */
553 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
554 /* Parameters:  fin(I) - pointer to packet information                      */
555 /*                                                                          */
556 /* IPv6 Only                                                                */
557 /* This is function checks pending hop by hop options extension header      */
558 /* ------------------------------------------------------------------------ */
559 static INLINE int frpr_hopopts6(fin)
560 fr_info_t *fin;
561 {
562 	return frpr_ipv6exthdr(fin, 0, IPPROTO_HOPOPTS);
563 }
564 
565 
566 /* ------------------------------------------------------------------------ */
567 /* Function:    frpr_routing6                                               */
568 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
569 /* Parameters:  fin(I) - pointer to packet information                      */
570 /*                                                                          */
571 /* IPv6 Only                                                                */
572 /* This is function checks pending routing extension header                 */
573 /* ------------------------------------------------------------------------ */
574 static INLINE int frpr_routing6(fin)
575 fr_info_t *fin;
576 {
577 	struct ip6_ext *hdr;
578 	int shift;
579 
580 	hdr = fin->fin_dp;
581 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_ROUTING) == IPPROTO_NONE)
582 		return IPPROTO_NONE;
583 
584 	shift = 8 + (hdr->ip6e_len << 3);
585 	/*
586 	 * Nasty extension header length?
587 	 */
588 	if ((hdr->ip6e_len << 3) & 15) {
589 		fin->fin_flx |= FI_BAD;
590 		/*
591 		 * Compensate for the changes made in frpr_ipv6exthdr()
592 		 */
593 		fin->fin_dlen += shift;
594 		fin->fin_dp = (char *)fin->fin_dp - shift;
595 		return IPPROTO_NONE;
596 	}
597 
598 	return hdr->ip6e_nxt;
599 }
600 
601 
602 /* ------------------------------------------------------------------------ */
603 /* Function:    frpr_fragment6                                              */
604 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
605 /* Parameters:  fin(I) - pointer to packet information                      */
606 /*                                                                          */
607 /* IPv6 Only                                                                */
608 /* Examine the IPv6 fragment header and extract fragment offset information.*/
609 /*                                                                          */
610 /* We don't know where the transport layer header (or whatever is next is), */
611 /* as it could be behind destination options (amongst others).  Because     */
612 /* there is no fragment cache, there is no knowledge about whether or not an*/
613 /* upper layer header has been seen (or where it ends) and thus we are not  */
614 /* able to continue processing beyond this header with any confidence.      */
615 /* ------------------------------------------------------------------------ */
616 static INLINE int frpr_fragment6(fin)
617 fr_info_t *fin;
618 {
619 	struct ip6_frag *frag;
620 
621 	fin->fin_flx |= FI_FRAG;
622 
623 	/*
624 	 * A fragmented IPv6 packet implies that there must be something
625 	 * else after the fragment.
626 	 */
627 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
628 		return IPPROTO_NONE;
629 
630 	frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
631 
632 	/*
633 	 * If this fragment isn't the last then the packet length must
634 	 * be a multiple of 8.
635 	 */
636 	if ((frag->ip6f_offlg & IP6F_MORE_FRAG) != 0) {
637 		fin->fin_flx |= FI_MOREFRAG;
638 
639 		if ((fin->fin_plen & 0x7) != 0)
640 			fin->fin_flx |= FI_BAD;
641 	}
642 
643 	fin->fin_id = frag->ip6f_ident;
644 	fin->fin_off = ntohs(frag->ip6f_offlg & IP6F_OFF_MASK);
645 	if (fin->fin_off != 0)
646 		fin->fin_flx |= FI_FRAGBODY;
647 
648 	return frag->ip6f_nxt;
649 }
650 
651 
652 /* ------------------------------------------------------------------------ */
653 /* Function:    frpr_dstopts6                                               */
654 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
655 /* Parameters:  fin(I) - pointer to packet information                      */
656 /*              nextheader(I) - stores next header value                    */
657 /*                                                                          */
658 /* IPv6 Only                                                                */
659 /* This is function checks pending destination options extension header     */
660 /* ------------------------------------------------------------------------ */
661 static INLINE int frpr_dstopts6(fin)
662 fr_info_t *fin;
663 {
664 	return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
665 }
666 
667 
668 /* ------------------------------------------------------------------------ */
669 /* Function:    frpr_icmp6                                                  */
670 /* Returns:     void                                                        */
671 /* Parameters:  fin(I) - pointer to packet information                      */
672 /*                                                                          */
673 /* IPv6 Only                                                                */
674 /* This routine is mainly concerned with determining the minimum valid size */
675 /* for an ICMPv6 packet.                                                    */
676 /* ------------------------------------------------------------------------ */
677 static INLINE void frpr_icmp6(fin)
678 fr_info_t *fin;
679 {
680 	int minicmpsz = sizeof(struct icmp6_hdr);
681 	struct icmp6_hdr *icmp6;
682 
683 	if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
684 		return;
685 
686 	if (fin->fin_dlen > 1) {
687 		icmp6 = fin->fin_dp;
688 
689 		fin->fin_data[0] = *(u_short *)icmp6;
690 
691 		if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
692 			fin->fin_flx |= FI_ICMPQUERY;
693 
694 		switch (icmp6->icmp6_type)
695 		{
696 		case ICMP6_ECHO_REPLY :
697 		case ICMP6_ECHO_REQUEST :
698 			if (fin->fin_dlen >= 6)
699 				fin->fin_data[1] = icmp6->icmp6_id;
700 			minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
701 			break;
702 		case ICMP6_DST_UNREACH :
703 		case ICMP6_PACKET_TOO_BIG :
704 		case ICMP6_TIME_EXCEEDED :
705 		case ICMP6_PARAM_PROB :
706 			if ((fin->fin_m != NULL) &&
707 			    (M_LEN(fin->fin_m) < fin->fin_plen)) {
708 				if (fr_coalesce(fin) != 1)
709 					return;
710 			}
711 			fin->fin_flx |= FI_ICMPERR;
712 			minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
713 			break;
714 		default :
715 			break;
716 		}
717 	}
718 
719 	frpr_short6(fin, minicmpsz);
720 }
721 
722 
723 /* ------------------------------------------------------------------------ */
724 /* Function:    frpr_udp6                                                   */
725 /* Returns:     void                                                        */
726 /* Parameters:  fin(I) - pointer to packet information                      */
727 /*                                                                          */
728 /* IPv6 Only                                                                */
729 /* Analyse the packet for IPv6/UDP properties.                              */
730 /* Is not expected to be called for fragmented packets.                     */
731 /* ------------------------------------------------------------------------ */
732 static INLINE void frpr_udp6(fin)
733 fr_info_t *fin;
734 {
735 
736 	fr_checkv6sum(fin);
737 
738 	frpr_short6(fin, sizeof(struct udphdr));
739 	if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
740 		return;
741 
742 	frpr_udpcommon(fin);
743 }
744 
745 
746 /* ------------------------------------------------------------------------ */
747 /* Function:    frpr_tcp6                                                   */
748 /* Returns:     void                                                        */
749 /* Parameters:  fin(I) - pointer to packet information                      */
750 /*                                                                          */
751 /* IPv6 Only                                                                */
752 /* Analyse the packet for IPv6/TCP properties.                              */
753 /* Is not expected to be called for fragmented packets.                     */
754 /* ------------------------------------------------------------------------ */
755 static INLINE void frpr_tcp6(fin)
756 fr_info_t *fin;
757 {
758 
759 	fr_checkv6sum(fin);
760 
761 	frpr_short6(fin, sizeof(struct tcphdr));
762 	if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
763 		return;
764 
765 	frpr_tcpcommon(fin);
766 }
767 
768 
769 /* ------------------------------------------------------------------------ */
770 /* Function:    frpr_esp6                                                   */
771 /* Returns:     void                                                        */
772 /* Parameters:  fin(I) - pointer to packet information                      */
773 /*                                                                          */
774 /* IPv6 Only                                                                */
775 /* Analyse the packet for ESP properties.                                   */
776 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
777 /* even though the newer ESP packets must also have a sequence number that  */
778 /* is 32bits as well, it is not possible(?) to determine the version from a */
779 /* simple packet header.                                                    */
780 /* ------------------------------------------------------------------------ */
781 static INLINE void frpr_esp6(fin)
782 fr_info_t *fin;
783 {
784 	int i;
785 	frpr_short6(fin, sizeof(grehdr_t));
786 
787 	(void) frpr_pullup(fin, 8);
788 
789 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
790 		if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
791 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
792 			break;
793 		}
794 }
795 
796 
797 /* ------------------------------------------------------------------------ */
798 /* Function:    frpr_ah6                                                    */
799 /* Returns:     void                                                        */
800 /* Parameters:  fin(I) - pointer to packet information                      */
801 /*                                                                          */
802 /* IPv6 Only                                                                */
803 /* Analyse the packet for AH properties.                                    */
804 /* The minimum length is taken to be the combination of all fields in the   */
805 /* header being present and no authentication data (null algorithm used.)   */
806 /* ------------------------------------------------------------------------ */
807 static INLINE int frpr_ah6(fin)
808 fr_info_t *fin;
809 {
810 	authhdr_t *ah;
811 	int i, shift;
812 
813 	frpr_short6(fin, 12);
814 
815 	if (frpr_pullup(fin, sizeof(*ah)) == -1)
816 		return IPPROTO_NONE;
817 
818 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
819 		if (ip6exthdr[i].ol_val == IPPROTO_AH) {
820 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
821 			break;
822 		}
823 
824 	ah = (authhdr_t *)fin->fin_dp;
825 
826 	shift = (ah->ah_plen + 2) * 4;
827 	fin->fin_dlen -= shift;
828 	fin->fin_dp = (char*)fin->fin_dp + shift;
829 
830 	return ah->ah_next;
831 }
832 
833 
834 /* ------------------------------------------------------------------------ */
835 /* Function:    frpr_gre6                                                   */
836 /* Returns:     void                                                        */
837 /* Parameters:  fin(I) - pointer to packet information                      */
838 /*                                                                          */
839 /* Analyse the packet for GRE properties.                                   */
840 /* ------------------------------------------------------------------------ */
841 static INLINE void frpr_gre6(fin)
842 fr_info_t *fin;
843 {
844 	grehdr_t *gre;
845 
846 	frpr_short6(fin, sizeof(grehdr_t));
847 
848 	if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
849 		return;
850 
851 	gre = fin->fin_dp;
852 	if (GRE_REV(gre->gr_flags) == 1)
853 		fin->fin_data[0] = gre->gr_call;
854 }
855 #endif	/* USE_INET6 */
856 
857 
858 /* ------------------------------------------------------------------------ */
859 /* Function:    frpr_pullup                                                 */
860 /* Returns:     int     - 0 == pullup succeeded, -1 == failure              */
861 /* Parameters:  fin(I)  - pointer to packet information                     */
862 /*              plen(I) - length (excluding L3 header) to pullup            */
863 /*                                                                          */
864 /* Short inline function to cut down on code duplication to perform a call  */
865 /* to fr_pullup to ensure there is the required amount of data,             */
866 /* consecutively in the packet buffer.                                      */
867 /* ------------------------------------------------------------------------ */
868 static INLINE int frpr_pullup(fin, plen)
869 fr_info_t *fin;
870 int plen;
871 {
872 #if defined(_KERNEL)
873 	if (fin->fin_m != NULL) {
874 		int ipoff;
875 
876 		ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
877 
878 		if (fin->fin_dp != NULL)
879 			plen += (char *)fin->fin_dp -
880 				((char *)fin->fin_ip + fin->fin_hlen);
881 		plen += fin->fin_hlen;
882 		/*
883 		 * We don't do 'plen += ipoff;' here. The fr_pullup() will
884 		 * do it for us.
885 		 */
886 		if (M_LEN(fin->fin_m) < plen + ipoff) {
887 			if (fr_pullup(fin->fin_m, fin, plen) == NULL)
888 				return -1;
889 		}
890 	}
891 #endif
892 	return 0;
893 }
894 
895 
896 /* ------------------------------------------------------------------------ */
897 /* Function:    frpr_short                                                  */
898 /* Returns:     void                                                        */
899 /* Parameters:  fin(I)  - pointer to packet information                     */
900 /*              xmin(I) - minimum header size                               */
901 /*                                                                          */
902 /* Check if a packet is "short" as defined by xmin.  The rule we are        */
903 /* applying here is that the packet must not be fragmented within the layer */
904 /* 4 header.  That is, it must not be a fragment that has its offset set to */
905 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the    */
906 /* entire layer 4 header must be present (min).                             */
907 /* ------------------------------------------------------------------------ */
908 static INLINE void frpr_short(fin, xmin)
909 fr_info_t *fin;
910 int xmin;
911 {
912 
913 	if (fin->fin_off == 0) {
914 		if (fin->fin_dlen < xmin)
915 			fin->fin_flx |= FI_SHORT;
916 	} else if (fin->fin_off < xmin) {
917 		fin->fin_flx |= FI_SHORT;
918 	}
919 }
920 
921 
922 /* ------------------------------------------------------------------------ */
923 /* Function:    frpr_icmp                                                   */
924 /* Returns:     void                                                        */
925 /* Parameters:  fin(I) - pointer to packet information                      */
926 /*                                                                          */
927 /* IPv4 Only                                                                */
928 /* Do a sanity check on the packet for ICMP (v4).  In nearly all cases,     */
929 /* except extrememly bad packets, both type and code will be present.       */
930 /* The expected minimum size of an ICMP packet is very much dependent on    */
931 /* the type of it.                                                          */
932 /*                                                                          */
933 /* XXX - other ICMP sanity checks?                                          */
934 /* ------------------------------------------------------------------------ */
935 static INLINE void frpr_icmp(fin)
936 fr_info_t *fin;
937 {
938 	int minicmpsz = sizeof(struct icmp);
939 	icmphdr_t *icmp;
940 	ip_t *oip;
941 	ipf_stack_t *ifs = fin->fin_ifs;
942 
943 	if (fin->fin_off != 0) {
944 		frpr_short(fin, ICMPERR_ICMPHLEN);
945 		return;
946 	}
947 
948 	if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
949 		return;
950 
951 	fr_checkv4sum(fin);
952 
953 	/*
954 	 * This is a right place to set icmp pointer, since the memory
955 	 * referenced by fin_dp could get reallocated. The code down below can
956 	 * rely on fact icmp variable always points to ICMP header.
957 	 */
958 	icmp = fin->fin_dp;
959 	fin->fin_data[0] = *(u_short *)icmp;
960 	fin->fin_data[1] = icmp->icmp_id;
961 
962 	switch (icmp->icmp_type)
963 	{
964 	case ICMP_ECHOREPLY :
965 	case ICMP_ECHO :
966 	/* Router discovery messaes - RFC 1256 */
967 	case ICMP_ROUTERADVERT :
968 	case ICMP_ROUTERSOLICIT :
969 		minicmpsz = ICMP_MINLEN;
970 		break;
971 	/*
972 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
973 	 * 3 * timestamp(3 * 4)
974 	 */
975 	case ICMP_TSTAMP :
976 	case ICMP_TSTAMPREPLY :
977 		minicmpsz = 20;
978 		break;
979 	/*
980 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
981 	 * mask(4)
982 	 */
983 	case ICMP_MASKREQ :
984 	case ICMP_MASKREPLY :
985 			minicmpsz = 12;
986 			break;
987 	/*
988 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
989 	 */
990 	case ICMP_UNREACH :
991 		if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
992 			if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
993 				fin->fin_flx |= FI_BAD;
994 		}
995 		/* FALLTHRU */
996 	case ICMP_SOURCEQUENCH :
997 	case ICMP_REDIRECT :
998 	case ICMP_TIMXCEED :
999 	case ICMP_PARAMPROB :
1000 		fin->fin_flx |= FI_ICMPERR;
1001 		if (fr_coalesce(fin) != 1)
1002 			return;
1003 		/*
1004 		 * ICMP error packets should not be generated for IP
1005 		 * packets that are a fragment that isn't the first
1006 		 * fragment.
1007 		 */
1008 		oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1009 		if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1010 			fin->fin_flx |= FI_BAD;
1011 		break;
1012 	default :
1013 		break;
1014 	}
1015 
1016 	frpr_short(fin, minicmpsz);
1017 }
1018 
1019 
1020 /* ------------------------------------------------------------------------ */
1021 /* Function:    frpr_tcpcommon                                              */
1022 /* Returns:     void                                                        */
1023 /* Parameters:  fin(I) - pointer to packet information                      */
1024 /*                                                                          */
1025 /* TCP header sanity checking.  Look for bad combinations of TCP flags,     */
1026 /* and make some checks with how they interact with other fields.           */
1027 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is     */
1028 /* valid and mark the packet as bad if not.                                 */
1029 /* ------------------------------------------------------------------------ */
1030 static INLINE void frpr_tcpcommon(fin)
1031 fr_info_t *fin;
1032 {
1033 	int flags, tlen;
1034 	tcphdr_t *tcp;
1035 
1036 	fin->fin_flx |= FI_TCPUDP;
1037 	if (fin->fin_off != 0)
1038 		return;
1039 
1040 	if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1041 		return;
1042 	tcp = fin->fin_dp;
1043 
1044 	if (fin->fin_dlen > 3) {
1045 		fin->fin_sport = ntohs(tcp->th_sport);
1046 		fin->fin_dport = ntohs(tcp->th_dport);
1047 	}
1048 
1049 	if ((fin->fin_flx & FI_SHORT) != 0)
1050 		return;
1051 
1052 	/*
1053 	 * Use of the TCP data offset *must* result in a value that is at
1054 	 * least the same size as the TCP header.
1055 	 */
1056 	tlen = TCP_OFF(tcp) << 2;
1057 	if (tlen < sizeof(tcphdr_t)) {
1058 		fin->fin_flx |= FI_BAD;
1059 		return;
1060 	}
1061 
1062 	flags = tcp->th_flags;
1063 	fin->fin_tcpf = tcp->th_flags;
1064 
1065 	/*
1066 	 * If the urgent flag is set, then the urgent pointer must
1067 	 * also be set and vice versa.  Good TCP packets do not have
1068 	 * just one of these set.
1069 	 */
1070 	if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1071 		fin->fin_flx |= FI_BAD;
1072 	} else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1073 		/* Ignore this case, it shows up in "real" traffic with */
1074 		/* bogus values in the urgent pointer field. */
1075 		flags = flags; /* LINT */
1076 	} else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1077 		   ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1078 		/* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1079 		fin->fin_flx |= FI_BAD;
1080 	} else if (!(flags & TH_ACK)) {
1081 		/*
1082 		 * If the ack bit isn't set, then either the SYN or
1083 		 * RST bit must be set.  If the SYN bit is set, then
1084 		 * we expect the ACK field to be 0.  If the ACK is
1085 		 * not set and if URG, PSH or FIN are set, consdier
1086 		 * that to indicate a bad TCP packet.
1087 		 */
1088 		if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1089 			/*
1090 			 * Cisco PIX sets the ACK field to a random value.
1091 			 * In light of this, do not set FI_BAD until a patch
1092 			 * is available from Cisco to ensure that
1093 			 * interoperability between existing systems is
1094 			 * achieved.
1095 			 */
1096 			/*fin->fin_flx |= FI_BAD*/;
1097 			flags = flags; /* LINT */
1098 		} else if (!(flags & (TH_RST|TH_SYN))) {
1099 			fin->fin_flx |= FI_BAD;
1100 		} else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1101 			fin->fin_flx |= FI_BAD;
1102 		}
1103 	}
1104 
1105 	/*
1106 	 * At this point, it's not exactly clear what is to be gained by
1107 	 * marking up which TCP options are and are not present.  The one we
1108 	 * are most interested in is the TCP window scale.  This is only in
1109 	 * a SYN packet [RFC1323] so we don't need this here...?
1110 	 * Now if we were to analyse the header for passive fingerprinting,
1111 	 * then that might add some weight to adding this...
1112 	 */
1113 	if (tlen == sizeof(tcphdr_t))
1114 		return;
1115 
1116 	if (frpr_pullup(fin, tlen) == -1)
1117 		return;
1118 
1119 #if 0
1120 	ip = fin->fin_ip;
1121 	s = (u_char *)(tcp + 1);
1122 	off = IP_HL(ip) << 2;
1123 # ifdef _KERNEL
1124 	if (fin->fin_mp != NULL) {
1125 		mb_t *m = *fin->fin_mp;
1126 
1127 		if (off + tlen > M_LEN(m))
1128 			return;
1129 	}
1130 # endif
1131 	for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1132 		opt = *s;
1133 		if (opt == '\0')
1134 			break;
1135 		else if (opt == TCPOPT_NOP)
1136 			ol = 1;
1137 		else {
1138 			if (tlen < 2)
1139 				break;
1140 			ol = (int)*(s + 1);
1141 			if (ol < 2 || ol > tlen)
1142 				break;
1143 		}
1144 
1145 		for (i = 9, mv = 4; mv >= 0; ) {
1146 			op = ipopts + i;
1147 			if (opt == (u_char)op->ol_val) {
1148 				optmsk |= op->ol_bit;
1149 				break;
1150 			}
1151 		}
1152 		tlen -= ol;
1153 		s += ol;
1154 	}
1155 #endif /* 0 */
1156 }
1157 
1158 
1159 
1160 /* ------------------------------------------------------------------------ */
1161 /* Function:    frpr_udpcommon                                              */
1162 /* Returns:     void                                                        */
1163 /* Parameters:  fin(I) - pointer to packet information                      */
1164 /*                                                                          */
1165 /* Extract the UDP source and destination ports, if present.  If compiled   */
1166 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid.          */
1167 /* ------------------------------------------------------------------------ */
1168 static INLINE void frpr_udpcommon(fin)
1169 fr_info_t *fin;
1170 {
1171 	udphdr_t *udp;
1172 
1173 	fin->fin_flx |= FI_TCPUDP;
1174 
1175 	if (!fin->fin_off && (fin->fin_dlen > 3)) {
1176 		if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1177 			fin->fin_flx |= FI_SHORT;
1178 			return;
1179 		}
1180 
1181 		udp = fin->fin_dp;
1182 
1183 		fin->fin_sport = ntohs(udp->uh_sport);
1184 		fin->fin_dport = ntohs(udp->uh_dport);
1185 	}
1186 }
1187 
1188 
1189 /* ------------------------------------------------------------------------ */
1190 /* Function:    frpr_tcp                                                    */
1191 /* Returns:     void                                                        */
1192 /* Parameters:  fin(I) - pointer to packet information                      */
1193 /*                                                                          */
1194 /* IPv4 Only                                                                */
1195 /* Analyse the packet for IPv4/TCP properties.                              */
1196 /* ------------------------------------------------------------------------ */
1197 static INLINE void frpr_tcp(fin)
1198 fr_info_t *fin;
1199 {
1200 
1201 	fr_checkv4sum(fin);
1202 
1203 	frpr_short(fin, sizeof(tcphdr_t));
1204 
1205 	frpr_tcpcommon(fin);
1206 }
1207 
1208 
1209 /* ------------------------------------------------------------------------ */
1210 /* Function:    frpr_udp                                                    */
1211 /* Returns:     void                                                        */
1212 /* Parameters:  fin(I) - pointer to packet information                      */
1213 /*                                                                          */
1214 /* IPv4 Only                                                                */
1215 /* Analyse the packet for IPv4/UDP properties.                              */
1216 /* ------------------------------------------------------------------------ */
1217 static INLINE void frpr_udp(fin)
1218 fr_info_t *fin;
1219 {
1220 
1221 	fr_checkv4sum(fin);
1222 
1223 	frpr_short(fin, sizeof(udphdr_t));
1224 
1225 	frpr_udpcommon(fin);
1226 }
1227 
1228 
1229 /* ------------------------------------------------------------------------ */
1230 /* Function:    frpr_esp                                                    */
1231 /* Returns:     void                                                        */
1232 /* Parameters:  fin(I) - pointer to packet information                      */
1233 /*                                                                          */
1234 /* Analyse the packet for ESP properties.                                   */
1235 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
1236 /* even though the newer ESP packets must also have a sequence number that  */
1237 /* is 32bits as well, it is not possible(?) to determine the version from a */
1238 /* simple packet header.                                                    */
1239 /* ------------------------------------------------------------------------ */
1240 static INLINE void frpr_esp(fin)
1241 fr_info_t *fin;
1242 {
1243 	if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1244 		return;
1245 
1246 	frpr_short(fin, 8);
1247 }
1248 
1249 
1250 /* ------------------------------------------------------------------------ */
1251 /* Function:    frpr_ah                                                     */
1252 /* Returns:     void                                                        */
1253 /* Parameters:  fin(I) - pointer to packet information                      */
1254 /*                                                                          */
1255 /* Analyse the packet for AH properties.                                    */
1256 /* The minimum length is taken to be the combination of all fields in the   */
1257 /* header being present and no authentication data (null algorithm used.)   */
1258 /* ------------------------------------------------------------------------ */
1259 static INLINE void frpr_ah(fin)
1260 fr_info_t *fin;
1261 {
1262 	authhdr_t *ah;
1263 	int len;
1264 
1265 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1266 		return;
1267 
1268 	ah = (authhdr_t *)fin->fin_dp;
1269 
1270 	len = (ah->ah_plen + 2) << 2;
1271 	frpr_short(fin, len);
1272 }
1273 
1274 
1275 /* ------------------------------------------------------------------------ */
1276 /* Function:    frpr_gre                                                    */
1277 /* Returns:     void                                                        */
1278 /* Parameters:  fin(I) - pointer to packet information                      */
1279 /*                                                                          */
1280 /* Analyse the packet for GRE properties.                                   */
1281 /* ------------------------------------------------------------------------ */
1282 static INLINE void frpr_gre(fin)
1283 fr_info_t *fin;
1284 {
1285 	grehdr_t *gre;
1286 
1287 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1288 		return;
1289 
1290 	frpr_short(fin, sizeof(grehdr_t));
1291 
1292 	if (fin->fin_off == 0) {
1293 		gre = fin->fin_dp;
1294 		if (GRE_REV(gre->gr_flags) == 1)
1295 			fin->fin_data[0] = gre->gr_call;
1296 	}
1297 }
1298 
1299 
1300 /* ------------------------------------------------------------------------ */
1301 /* Function:    frpr_ipv4hdr                                                */
1302 /* Returns:     void                                                        */
1303 /* Parameters:  fin(I) - pointer to packet information                      */
1304 /*                                                                          */
1305 /* IPv4 Only                                                                */
1306 /* Analyze the IPv4 header and set fields in the fr_info_t structure.       */
1307 /* Check all options present and flag their presence if any exist.          */
1308 /* ------------------------------------------------------------------------ */
1309 static INLINE void frpr_ipv4hdr(fin)
1310 fr_info_t *fin;
1311 {
1312 	u_short optmsk = 0, secmsk = 0, auth = 0;
1313 	int hlen, ol, mv, p, i;
1314 	const struct optlist *op;
1315 	u_char *s, opt;
1316 	u_short off;
1317 	fr_ip_t *fi;
1318 	ip_t *ip;
1319 
1320 	fi = &fin->fin_fi;
1321 	hlen = fin->fin_hlen;
1322 
1323 	ip = fin->fin_ip;
1324 	p = ip->ip_p;
1325 	fi->fi_p = p;
1326 	fi->fi_tos = ip->ip_tos;
1327 	fin->fin_id = ip->ip_id;
1328 	off = ip->ip_off;
1329 
1330 	/* Get both TTL and protocol */
1331 	fi->fi_p = ip->ip_p;
1332 	fi->fi_ttl = ip->ip_ttl;
1333 #if 0
1334 	(*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1335 #endif
1336 
1337 	/* Zero out bits not used in IPv6 address */
1338 	fi->fi_src.i6[1] = 0;
1339 	fi->fi_src.i6[2] = 0;
1340 	fi->fi_src.i6[3] = 0;
1341 	fi->fi_dst.i6[1] = 0;
1342 	fi->fi_dst.i6[2] = 0;
1343 	fi->fi_dst.i6[3] = 0;
1344 
1345 	fi->fi_saddr = ip->ip_src.s_addr;
1346 	fi->fi_daddr = ip->ip_dst.s_addr;
1347 
1348 	/*
1349 	 * set packet attribute flags based on the offset and
1350 	 * calculate the byte offset that it represents.
1351 	 */
1352 	off &= IP_MF|IP_OFFMASK;
1353 	if (off != 0) {
1354 		int morefrag = off & IP_MF;
1355 
1356 		fi->fi_flx |= FI_FRAG;
1357 		if (morefrag)
1358 			fi->fi_flx |= FI_MOREFRAG;
1359 		off &= IP_OFFMASK;
1360 		if (off != 0) {
1361 			fin->fin_flx |= FI_FRAGBODY;
1362 			off <<= 3;
1363 			if ((off + fin->fin_dlen > 65535) ||
1364 			    (fin->fin_dlen == 0) ||
1365 			    ((morefrag != 0) && ((fin->fin_dlen & 7) != 0))) {
1366 				/*
1367 				 * The length of the packet, starting at its
1368 				 * offset cannot exceed 65535 (0xffff) as the
1369 				 * length of an IP packet is only 16 bits.
1370 				 *
1371 				 * Any fragment that isn't the last fragment
1372 				 * must have a length greater than 0 and it
1373 				 * must be an even multiple of 8.
1374 				 */
1375 				fi->fi_flx |= FI_BAD;
1376 			}
1377 		}
1378 	}
1379 	fin->fin_off = off;
1380 
1381 	/*
1382 	 * Call per-protocol setup and checking
1383 	 */
1384 	switch (p)
1385 	{
1386 	case IPPROTO_UDP :
1387 		frpr_udp(fin);
1388 		break;
1389 	case IPPROTO_TCP :
1390 		frpr_tcp(fin);
1391 		break;
1392 	case IPPROTO_ICMP :
1393 		frpr_icmp(fin);
1394 		break;
1395 	case IPPROTO_AH :
1396 		frpr_ah(fin);
1397 		break;
1398 	case IPPROTO_ESP :
1399 		frpr_esp(fin);
1400 		break;
1401 	case IPPROTO_GRE :
1402 		frpr_gre(fin);
1403 		break;
1404 	}
1405 
1406 	ip = fin->fin_ip;
1407 	if (ip == NULL)
1408 		return;
1409 
1410 	/*
1411 	 * If it is a standard IP header (no options), set the flag fields
1412 	 * which relate to options to 0.
1413 	 */
1414 	if (hlen == sizeof(*ip)) {
1415 		fi->fi_optmsk = 0;
1416 		fi->fi_secmsk = 0;
1417 		fi->fi_auth = 0;
1418 		return;
1419 	}
1420 
1421 	/*
1422 	 * So the IP header has some IP options attached.  Walk the entire
1423 	 * list of options present with this packet and set flags to indicate
1424 	 * which ones are here and which ones are not.  For the somewhat out
1425 	 * of date and obscure security classification options, set a flag to
1426 	 * represent which classification is present.
1427 	 */
1428 	fi->fi_flx |= FI_OPTIONS;
1429 
1430 	for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1431 		opt = *s;
1432 		if (opt == '\0')
1433 			break;
1434 		else if (opt == IPOPT_NOP)
1435 			ol = 1;
1436 		else {
1437 			if (hlen < 2)
1438 				break;
1439 			ol = (int)*(s + 1);
1440 			if (ol < 2 || ol > hlen)
1441 				break;
1442 		}
1443 		for (i = 9, mv = 4; mv >= 0; ) {
1444 			op = ipopts + i;
1445 			if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1446 				optmsk |= op->ol_bit;
1447 				if (opt == IPOPT_SECURITY) {
1448 					const struct optlist *sp;
1449 					u_char	sec;
1450 					int j, m;
1451 
1452 					sec = *(s + 2);	/* classification */
1453 					for (j = 3, m = 2; m >= 0; ) {
1454 						sp = secopt + j;
1455 						if (sec == sp->ol_val) {
1456 							secmsk |= sp->ol_bit;
1457 							auth = *(s + 3);
1458 							auth *= 256;
1459 							auth += *(s + 4);
1460 							break;
1461 						}
1462 						if (sec < sp->ol_val)
1463 							j -= m;
1464 						else
1465 							j += m;
1466 						m--;
1467 					}
1468 				}
1469 				break;
1470 			}
1471 			if (opt < op->ol_val)
1472 				i -= mv;
1473 			else
1474 				i += mv;
1475 			mv--;
1476 		}
1477 		hlen -= ol;
1478 		s += ol;
1479 	}
1480 
1481 	/*
1482 	 *
1483 	 */
1484 	if (auth && !(auth & 0x0100))
1485 		auth &= 0xff00;
1486 	fi->fi_optmsk = optmsk;
1487 	fi->fi_secmsk = secmsk;
1488 	fi->fi_auth = auth;
1489 }
1490 
1491 
1492 /* ------------------------------------------------------------------------ */
1493 /* Function:    fr_makefrip                                                 */
1494 /* Returns:     int - 1 == hdr checking error, 0 == OK                      */
1495 /* Parameters:  hlen(I) - length of IP packet header                        */
1496 /*              ip(I)   - pointer to the IP header                          */
1497 /*              fin(IO) - pointer to packet information                     */
1498 /*                                                                          */
1499 /* Compact the IP header into a structure which contains just the info.     */
1500 /* which is useful for comparing IP headers with and store this information */
1501 /* in the fr_info_t structure pointer to by fin.  At present, it is assumed */
1502 /* this function will be called with either an IPv4 or IPv6 packet.         */
1503 /* ------------------------------------------------------------------------ */
1504 int	fr_makefrip(hlen, ip, fin)
1505 int hlen;
1506 ip_t *ip;
1507 fr_info_t *fin;
1508 {
1509 	int v;
1510 
1511 	fin->fin_depth = 0;
1512 	fin->fin_hlen = (u_short)hlen;
1513 	fin->fin_ip = ip;
1514 	fin->fin_rule = 0xffffffff;
1515 	fin->fin_group[0] = -1;
1516 	fin->fin_group[1] = '\0';
1517 	fin->fin_dlen = fin->fin_plen - hlen;
1518 	fin->fin_dp = (char *)ip + hlen;
1519 
1520 	v = fin->fin_v;
1521 	if (v == 4)
1522 		frpr_ipv4hdr(fin);
1523 #ifdef	USE_INET6
1524 	else if (v == 6)
1525 		frpr_ipv6hdr(fin);
1526 #endif
1527 	if (fin->fin_ip == NULL)
1528 		return -1;
1529 	return 0;
1530 }
1531 
1532 
1533 /* ------------------------------------------------------------------------ */
1534 /* Function:    fr_portcheck                                                */
1535 /* Returns:     int - 1 == port matched, 0 == port match failed             */
1536 /* Parameters:  frp(I) - pointer to port check `expression'                 */
1537 /*              pop(I) - pointer to port number to evaluate                 */
1538 /*                                                                          */
1539 /* Perform a comparison of a port number against some other(s), using a     */
1540 /* structure with compare information stored in it.                         */
1541 /* ------------------------------------------------------------------------ */
1542 static INLINE int fr_portcheck(frp, pop)
1543 frpcmp_t *frp;
1544 u_short *pop;
1545 {
1546 	u_short tup, po;
1547 	int err = 1;
1548 
1549 	tup = *pop;
1550 	po = frp->frp_port;
1551 
1552 	/*
1553 	 * Do opposite test to that required and continue if that succeeds.
1554 	 */
1555 	switch (frp->frp_cmp)
1556 	{
1557 	case FR_EQUAL :
1558 		if (tup != po) /* EQUAL */
1559 			err = 0;
1560 		break;
1561 	case FR_NEQUAL :
1562 		if (tup == po) /* NOTEQUAL */
1563 			err = 0;
1564 		break;
1565 	case FR_LESST :
1566 		if (tup >= po) /* LESSTHAN */
1567 			err = 0;
1568 		break;
1569 	case FR_GREATERT :
1570 		if (tup <= po) /* GREATERTHAN */
1571 			err = 0;
1572 		break;
1573 	case FR_LESSTE :
1574 		if (tup > po) /* LT or EQ */
1575 			err = 0;
1576 		break;
1577 	case FR_GREATERTE :
1578 		if (tup < po) /* GT or EQ */
1579 			err = 0;
1580 		break;
1581 	case FR_OUTRANGE :
1582 		if (tup >= po && tup <= frp->frp_top) /* Out of range */
1583 			err = 0;
1584 		break;
1585 	case FR_INRANGE :
1586 		if (tup <= po || tup >= frp->frp_top) /* In range */
1587 			err = 0;
1588 		break;
1589 	case FR_INCRANGE :
1590 		if (tup < po || tup > frp->frp_top) /* Inclusive range */
1591 			err = 0;
1592 		break;
1593 	default :
1594 		break;
1595 	}
1596 	return err;
1597 }
1598 
1599 
1600 /* ------------------------------------------------------------------------ */
1601 /* Function:    fr_tcpudpchk                                                */
1602 /* Returns:     int - 1 == protocol matched, 0 == check failed              */
1603 /* Parameters:  fin(I) - pointer to packet information                      */
1604 /*              ft(I)  - pointer to structure with comparison data          */
1605 /*                                                                          */
1606 /* Compares the current pcket (assuming it is TCP/UDP) information with a   */
1607 /* structure containing information that we want to match against.          */
1608 /* ------------------------------------------------------------------------ */
1609 int fr_tcpudpchk(fin, ft)
1610 fr_info_t *fin;
1611 frtuc_t *ft;
1612 {
1613 	int err = 1;
1614 
1615 	/*
1616 	 * Both ports should *always* be in the first fragment.
1617 	 * So far, I cannot find any cases where they can not be.
1618 	 *
1619 	 * compare destination ports
1620 	 */
1621 	if (ft->ftu_dcmp)
1622 		err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1623 
1624 	/*
1625 	 * compare source ports
1626 	 */
1627 	if (err && ft->ftu_scmp)
1628 		err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1629 
1630 	/*
1631 	 * If we don't have all the TCP/UDP header, then how can we
1632 	 * expect to do any sort of match on it ?  If we were looking for
1633 	 * TCP flags, then NO match.  If not, then match (which should
1634 	 * satisfy the "short" class too).
1635 	 */
1636 	if (err && (fin->fin_p == IPPROTO_TCP)) {
1637 		if (fin->fin_flx & FI_SHORT)
1638 			return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1639 		/*
1640 		 * Match the flags ?  If not, abort this match.
1641 		 */
1642 		if (ft->ftu_tcpfm &&
1643 		    ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1644 			FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1645 				 ft->ftu_tcpfm, ft->ftu_tcpf));
1646 			err = 0;
1647 		}
1648 	}
1649 	return err;
1650 }
1651 
1652 
1653 /* ------------------------------------------------------------------------ */
1654 /* Function:    fr_ipfcheck                                                 */
1655 /* Returns:     int - 0 == match, 1 == no match                             */
1656 /* Parameters:  fin(I)     - pointer to packet information                  */
1657 /*              fr(I)      - pointer to filter rule                         */
1658 /*              portcmp(I) - flag indicating whether to attempt matching on */
1659 /*                           TCP/UDP port data.                             */
1660 /*                                                                          */
1661 /* Check to see if a packet matches an IPFilter rule.  Checks of addresses, */
1662 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1663 /* this function.                                                           */
1664 /* ------------------------------------------------------------------------ */
1665 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1666 fr_info_t *fin;
1667 frentry_t *fr;
1668 int portcmp;
1669 {
1670 	u_32_t	*ld, *lm, *lip;
1671 	fripf_t *fri;
1672 	fr_ip_t *fi;
1673 	int i;
1674 	ipf_stack_t *ifs = fin->fin_ifs;
1675 
1676 	fi = &fin->fin_fi;
1677 	fri = fr->fr_ipf;
1678 	lip = (u_32_t *)fi;
1679 	lm = (u_32_t *)&fri->fri_mip;
1680 	ld = (u_32_t *)&fri->fri_ip;
1681 
1682 	/*
1683 	 * first 32 bits to check coversion:
1684 	 * IP version, TOS, TTL, protocol
1685 	 */
1686 	i = ((*lip & *lm) != *ld);
1687 	FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1688 		   *lip, *lm, *ld));
1689 	if (i)
1690 		return 1;
1691 
1692 	/*
1693 	 * Next 32 bits is a constructed bitmask indicating which IP options
1694 	 * are present (if any) in this packet.
1695 	 */
1696 	lip++, lm++, ld++;
1697 	i |= ((*lip & *lm) != *ld);
1698 	FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1699 		   *lip, *lm, *ld));
1700 	if (i)
1701 		return 1;
1702 
1703 	lip++, lm++, ld++;
1704 	/*
1705 	 * Unrolled loops (4 each, for 32 bits) for address checks.
1706 	 */
1707 	/*
1708 	 * Check the source address.
1709 	 */
1710 #ifdef	IPFILTER_LOOKUP
1711 	if (fr->fr_satype == FRI_LOOKUP) {
1712 		fin->fin_flx |= FI_DONTCACHE;
1713 		i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, fin, 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 		fin->fin_flx |= FI_DONTCACHE;
1756 		i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, fin, ifs);
1757 		if (i == -1)
1758 			return 1;
1759 		lip += 3;
1760 		lm += 3;
1761 		ld += 3;
1762 	} else {
1763 #endif
1764 		i = ((*lip & *lm) != *ld);
1765 		FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1766 			   *lip, *lm, *ld));
1767 		if (fi->fi_v == 6) {
1768 			lip++, lm++, ld++;
1769 			i |= ((*lip & *lm) != *ld);
1770 			FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1771 				   *lip, *lm, *ld));
1772 			lip++, lm++, ld++;
1773 			i |= ((*lip & *lm) != *ld);
1774 			FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1775 				   *lip, *lm, *ld));
1776 			lip++, lm++, ld++;
1777 			i |= ((*lip & *lm) != *ld);
1778 			FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1779 				   *lip, *lm, *ld));
1780 		} else {
1781 			lip += 3;
1782 			lm += 3;
1783 			ld += 3;
1784 		}
1785 #ifdef	IPFILTER_LOOKUP
1786 	}
1787 #endif
1788 	i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1789 	if (i)
1790 		return 1;
1791 	/*
1792 	 * IP addresses matched.  The next 32bits contains:
1793 	 * mast of old IP header security & authentication bits.
1794 	 */
1795 	lip++, lm++, ld++;
1796 	i |= ((*lip & *lm) != *ld);
1797 	FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1798 		   *lip, *lm, *ld));
1799 
1800 	/*
1801 	 * Next we have 32 bits of packet flags.
1802 	 */
1803 	lip++, lm++, ld++;
1804 	i |= ((*lip & *lm) != *ld);
1805 	FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1806 		   *lip, *lm, *ld));
1807 
1808 	if (i == 0) {
1809 		/*
1810 		 * If a fragment, then only the first has what we're
1811 		 * looking for here...
1812 		 */
1813 		if (portcmp) {
1814 			if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1815 				i = 1;
1816 		} else {
1817 			if (fr->fr_dcmp || fr->fr_scmp ||
1818 			    fr->fr_tcpf || fr->fr_tcpfm)
1819 				i = 1;
1820 			if (fr->fr_icmpm || fr->fr_icmp) {
1821 				if (((fi->fi_p != IPPROTO_ICMP) &&
1822 				     (fi->fi_p != IPPROTO_ICMPV6)) ||
1823 				    fin->fin_off || (fin->fin_dlen < 2))
1824 					i = 1;
1825 				else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1826 					 fr->fr_icmp) {
1827 					FR_DEBUG(("i. %#x & %#x != %#x\n",
1828 						 fin->fin_data[0],
1829 						 fr->fr_icmpm, fr->fr_icmp));
1830 					i = 1;
1831 				}
1832 			}
1833 		}
1834 	}
1835 	return i;
1836 }
1837 
1838 
1839 /* ------------------------------------------------------------------------ */
1840 /* Function:    fr_scanlist                                                 */
1841 /* Returns:     int - result flags of scanning filter list                  */
1842 /* Parameters:  fin(I) - pointer to packet information                      */
1843 /*              pass(I) - default result to return for filtering            */
1844 /*                                                                          */
1845 /* Check the input/output list of rules for a match to the current packet.  */
1846 /* If a match is found, the value of fr_flags from the rule becomes the     */
1847 /* return value and fin->fin_fr points to the matched rule.                 */
1848 /*                                                                          */
1849 /* This function may be called recusively upto 16 times (limit inbuilt.)    */
1850 /* When unwinding, it should finish up with fin_depth as 0.                 */
1851 /*                                                                          */
1852 /* Could be per interface, but this gets real nasty when you don't have,    */
1853 /* or can't easily change, the kernel source code to .                      */
1854 /* ------------------------------------------------------------------------ */
1855 int fr_scanlist(fin, pass)
1856 fr_info_t *fin;
1857 u_32_t pass;
1858 {
1859 	int rulen, portcmp, off, logged, skip;
1860 	struct frentry *fr, *fnext;
1861 	u_32_t passt, passo;
1862 	ipf_stack_t *ifs = fin->fin_ifs;
1863 
1864 	/*
1865 	 * Do not allow nesting deeper than 16 levels.
1866 	 */
1867 	if (fin->fin_depth >= 16)
1868 		return pass;
1869 
1870 	fr = fin->fin_fr;
1871 
1872 	/*
1873 	 * If there are no rules in this list, return now.
1874 	 */
1875 	if (fr == NULL)
1876 		return pass;
1877 
1878 	skip = 0;
1879 	logged = 0;
1880 	portcmp = 0;
1881 	fin->fin_depth++;
1882 	fin->fin_fr = NULL;
1883 	off = fin->fin_off;
1884 
1885 	if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1886 		portcmp = 1;
1887 
1888 	for (rulen = 0; fr; fr = fnext, rulen++) {
1889 		fnext = fr->fr_next;
1890 		if (skip != 0) {
1891 			FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1892 			skip--;
1893 			continue;
1894 		}
1895 
1896 		/*
1897 		 * In all checks below, a null (zero) value in the
1898 		 * filter struture is taken to mean a wildcard.
1899 		 *
1900 		 * check that we are working for the right interface
1901 		 */
1902 #ifdef	_KERNEL
1903 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1904 			continue;
1905 #else
1906 		if (opts & (OPT_VERBOSE|OPT_DEBUG))
1907 			printf("\n");
1908 		FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1909 				  FR_ISPASS(pass) ? 'p' :
1910 				  FR_ISACCOUNT(pass) ? 'A' :
1911 				  FR_ISAUTH(pass) ? 'a' :
1912 				  (pass & FR_NOMATCH) ? 'n' :'b'));
1913 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1914 			continue;
1915 		FR_VERBOSE((":i"));
1916 #endif
1917 
1918 		switch (fr->fr_type)
1919 		{
1920 		case FR_T_IPF :
1921 		case FR_T_IPF|FR_T_BUILTIN :
1922 			if (fr_ipfcheck(fin, fr, portcmp))
1923 				continue;
1924 			break;
1925 #if defined(IPFILTER_BPF)
1926 		case FR_T_BPFOPC :
1927 		case FR_T_BPFOPC|FR_T_BUILTIN :
1928 		    {
1929 			u_char *mc;
1930 
1931 			if (*fin->fin_mp == NULL)
1932 				continue;
1933 			if (fin->fin_v != fr->fr_v)
1934 				continue;
1935 			mc = (u_char *)fin->fin_m;
1936 			if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1937 				continue;
1938 			break;
1939 		    }
1940 #endif
1941 		case FR_T_CALLFUNC|FR_T_BUILTIN :
1942 		    {
1943 			frentry_t *f;
1944 
1945 			f = (*fr->fr_func)(fin, &pass);
1946 			if (f != NULL)
1947 				fr = f;
1948 			else
1949 				continue;
1950 			break;
1951 		    }
1952 		default :
1953 			break;
1954 		}
1955 
1956 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1957 			if (fin->fin_nattag == NULL)
1958 				continue;
1959 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1960 				continue;
1961 		}
1962 		FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1963 
1964 		passt = fr->fr_flags;
1965 
1966 		/*
1967 		 * Allowing a rule with the "keep state" flag set to match
1968 		 * packets that have been tagged "out of window" by the TCP
1969 		 * state tracking is foolish as the attempt to add a new
1970 		 * state entry to the table will fail.
1971 		 */
1972 		if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1973 			continue;
1974 
1975 		/*
1976 		 * If the rule is a "call now" rule, then call the function
1977 		 * in the rule, if it exists and use the results from that.
1978 		 * If the function pointer is bad, just make like we ignore
1979 		 * it, except for increasing the hit counter.
1980 		 */
1981 		IPF_BUMP(fr->fr_hits);
1982 		fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
1983 		if ((passt & FR_CALLNOW) != 0) {
1984 			if ((fr->fr_func != NULL) &&
1985 			    (fr->fr_func != (ipfunc_t)-1)) {
1986 				frentry_t *frs;
1987 
1988 				frs = fin->fin_fr;
1989 				fin->fin_fr = fr;
1990 				fr = (*fr->fr_func)(fin, &passt);
1991 				if (fr == NULL) {
1992 					fin->fin_fr = frs;
1993 					continue;
1994 				}
1995 				passt = fr->fr_flags;
1996 				fin->fin_fr = fr;
1997 			}
1998 		} else {
1999 			fin->fin_fr = fr;
2000 		}
2001 
2002 #ifdef  IPFILTER_LOG
2003 		/*
2004 		 * Just log this packet...
2005 		 */
2006 		if ((passt & FR_LOGMASK) == FR_LOG) {
2007 			if (ipflog(fin, passt) == -1) {
2008 				if (passt & FR_LOGORBLOCK) {
2009 					passt &= ~FR_CMDMASK;
2010 					passt |= FR_BLOCK|FR_QUICK;
2011 				}
2012 				IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2013 			}
2014 			IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2015 			logged = 1;
2016 		}
2017 #endif /* IPFILTER_LOG */
2018 		passo = pass;
2019 		if (FR_ISSKIP(passt))
2020 			skip = fr->fr_arg;
2021 		else if ((passt & FR_LOGMASK) != FR_LOG)
2022 			pass = passt;
2023 		if (passt & (FR_RETICMP|FR_FAKEICMP))
2024 			fin->fin_icode = fr->fr_icode;
2025 		FR_DEBUG(("pass %#x\n", pass));
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 			fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2175 			pass = fr->fr_flags;
2176 		}
2177 	} else {
2178 		RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2179 
2180 		pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2181 
2182 		if (((pass & FR_KEEPSTATE) == 0) &&
2183 		    ((fin->fin_flx & FI_DONTCACHE) == 0)) {
2184 			WRITE_ENTER(&ifs->ifs_ipf_frcache);
2185 			bcopy((char *)fin, (char *)fc, FI_COPYSIZE);
2186 			RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2187 		}
2188 
2189 		fr = fin->fin_fr;
2190 	}
2191 
2192 	if ((pass & FR_NOMATCH)) {
2193 		IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2194 	}
2195 
2196 	/*
2197 	 * Apply packets per second rate-limiting to a rule as required.
2198 	 */
2199 	if ((fr != NULL) && (fr->fr_pps != 0) &&
2200 	    !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2201 		pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2202 		pass |= FR_BLOCK;
2203 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2204 	}
2205 
2206 	/*
2207 	 * If we fail to add a packet to the authorization queue, then we
2208 	 * drop the packet later.  However, if it was added then pretend
2209 	 * we've dropped it already.
2210 	 */
2211 	if (FR_ISAUTH(pass)) {
2212 		if (fr_newauth(fin->fin_m, fin) != 0) {
2213 #ifdef	_KERNEL
2214 			fin->fin_m = *fin->fin_mp = NULL;
2215 #else
2216 			;
2217 #endif
2218 			fin->fin_error = 0;
2219 		} else
2220 			fin->fin_error = ENOSPC;
2221 	}
2222 
2223 	if ((fr != NULL) && (fr->fr_func != NULL) &&
2224 	    (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2225 		(void) (*fr->fr_func)(fin, &pass);
2226 
2227 	/*
2228 	 * If a rule is a pre-auth rule, check again in the list of rules
2229 	 * loaded for authenticated use.  It does not particulary matter
2230 	 * if this search fails because a "preauth" result, from a rule,
2231 	 * is treated as "not a pass", hence the packet is blocked.
2232 	 */
2233 	if (FR_ISPREAUTH(pass)) {
2234 		if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2235 			pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2236 	}
2237 
2238 	/*
2239 	 * If the rule has "keep frag" and the packet is actually a fragment,
2240 	 * then create a fragment state entry.
2241 	 */
2242 	if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2243 		if (fin->fin_flx & FI_FRAG) {
2244 			if (fr_newfrag(fin, pass) == -1) {
2245 				IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2246 			} else {
2247 				IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2248 			}
2249 		} else {
2250 			IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2251 		}
2252 	}
2253 
2254 	/*
2255 	 * Finally, if we've asked to track state for this packet, set it up.
2256 	 */
2257 	if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2258 		if (fr_addstate(fin, NULL, 0) != NULL) {
2259 			IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2260 		} else {
2261 			IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2262 			if (FR_ISPASS(pass)) {
2263 				pass &= ~FR_CMDMASK;
2264 				pass |= FR_BLOCK;
2265 			}
2266 		}
2267 	}
2268 
2269 	fr = fin->fin_fr;
2270 
2271 	if (passp != NULL)
2272 		*passp = pass;
2273 
2274 	return fr;
2275 }
2276 
2277 /* ------------------------------------------------------------------------ */
2278 /* Function:    fr_check                                                    */
2279 /* Returns:     int -  0 == packet allowed through,                         */
2280 /*              User space:                                                 */
2281 /*                    -1 == packet blocked                                  */
2282 /*                     1 == packet not matched                              */
2283 /*                    -2 == requires authentication                         */
2284 /*              Kernel:                                                     */
2285 /*                   > 0 == filter error # for packet                       */
2286 /* Parameters: ip(I)   - pointer to start of IPv4/6 packet                  */
2287 /*             hlen(I) - length of header                                   */
2288 /*             ifp(I)  - pointer to interface this packet is on             */
2289 /*             out(I)  - 0 == packet going in, 1 == packet going out        */
2290 /*             mp(IO)  - pointer to caller's buffer pointer that holds this */
2291 /*                       IP packet.                                         */
2292 /* Solaris & HP-UX ONLY :                                                   */
2293 /*             qpi(I)  - pointer to STREAMS queue information for this      */
2294 /*                       interface & direction.                             */
2295 /*                                                                          */
2296 /* fr_check() is the master function for all IPFilter packet processing.    */
2297 /* It orchestrates: Network Address Translation (NAT), checking for packet  */
2298 /* authorisation (or pre-authorisation), presence of related state info.,   */
2299 /* generating log entries, IP packet accounting, routing of packets as      */
2300 /* directed by firewall rules and of course whether or not to allow the     */
2301 /* packet to be further processed by the kernel.                            */
2302 /*                                                                          */
2303 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer  */
2304 /* freed.  Packets passed may be returned with the pointer pointed to by    */
2305 /* by "mp" changed to a new buffer.                                         */
2306 /* ------------------------------------------------------------------------ */
2307 int fr_check(ip, hlen, ifp, out
2308 #if defined(_KERNEL) && defined(MENTAT)
2309 , qif, mp, ifs)
2310 void *qif;
2311 #else
2312 , mp, ifs)
2313 #endif
2314 mb_t **mp;
2315 ip_t *ip;
2316 int hlen;
2317 void *ifp;
2318 int out;
2319 ipf_stack_t *ifs;
2320 {
2321 	/*
2322 	 * The above really sucks, but short of writing a diff
2323 	 */
2324 	fr_info_t frinfo;
2325 	fr_info_t *fin = &frinfo;
2326 	u_32_t pass;
2327 	frentry_t *fr = NULL;
2328 	int v = IP_V(ip);
2329 	mb_t *mc = NULL;
2330 	mb_t *m;
2331 #ifdef USE_INET6
2332 	ip6_t *ip6;
2333 #endif
2334 #ifdef	_KERNEL
2335 # ifdef MENTAT
2336 	qpktinfo_t *qpi = qif;
2337 #endif
2338 #endif
2339 
2340 	SPL_INT(s);
2341 	pass = ifs->ifs_fr_pass;
2342 
2343 	/*
2344 	 * The first part of fr_check() deals with making sure that what goes
2345 	 * into the filtering engine makes some sense.  Information about the
2346 	 * the packet is distilled, collected into a fr_info_t structure and
2347 	 * the an attempt to ensure the buffer the packet is in is big enough
2348 	 * to hold all the required packet headers.
2349 	 */
2350 #ifdef	_KERNEL
2351 # ifdef MENTAT
2352 	if (!OK_32PTR(ip))
2353 		return 2;
2354 # endif
2355 
2356 
2357 	if (ifs->ifs_fr_running <= 0) {
2358 		return 0;
2359 	}
2360 
2361 	bzero((char *)fin, sizeof(*fin));
2362 
2363 # ifdef MENTAT
2364 	fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2365 					 FI_BROADCAST);
2366 	m = qpi->qpi_m;
2367 	fin->fin_qfm = m;
2368 	fin->fin_qpi = qpi;
2369 # else /* MENTAT */
2370 
2371 	m = *mp;
2372 
2373 #  if defined(M_MCAST)
2374 	if ((m->m_flags & M_MCAST) != 0)
2375 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2376 #  endif
2377 #  if defined(M_MLOOP)
2378 	if ((m->m_flags & M_MLOOP) != 0)
2379 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2380 #  endif
2381 #  if defined(M_BCAST)
2382 	if ((m->m_flags & M_BCAST) != 0)
2383 		fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2384 #  endif
2385 #  ifdef M_CANFASTFWD
2386 	/*
2387 	 * XXX For now, IP Filter and fast-forwarding of cached flows
2388 	 * XXX are mutually exclusive.  Eventually, IP Filter should
2389 	 * XXX get a "can-fast-forward" filter rule.
2390 	 */
2391 	m->m_flags &= ~M_CANFASTFWD;
2392 #  endif /* M_CANFASTFWD */
2393 #  ifdef CSUM_DELAY_DATA
2394 	/*
2395 	 * disable delayed checksums.
2396 	 */
2397 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2398 		in_delayed_cksum(m);
2399 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2400 	}
2401 #  endif /* CSUM_DELAY_DATA */
2402 # endif /* MENTAT */
2403 #else
2404 
2405 	bzero((char *)fin, sizeof(*fin));
2406 	m = *mp;
2407 #endif /* _KERNEL */
2408 
2409 	fin->fin_v = v;
2410 	fin->fin_m = m;
2411 	fin->fin_ip = ip;
2412 	fin->fin_mp = mp;
2413 	fin->fin_out = out;
2414 	fin->fin_ifp = ifp;
2415 	fin->fin_error = ENETUNREACH;
2416 	fin->fin_hlen = (u_short)hlen;
2417 	fin->fin_dp = (char *)ip + hlen;
2418 	fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2419 	fin->fin_ifs = ifs;
2420 
2421 	SPL_NET(s);
2422 
2423 #ifdef	USE_INET6
2424 	if (v == 6) {
2425 		IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2426 		/*
2427 		 * Jumbo grams are quite likely too big for internal buffer
2428 		 * structures to handle comfortably, for now, so just drop
2429 		 * them.
2430 		 */
2431 		ip6 = (ip6_t *)ip;
2432 		fin->fin_plen = ntohs(ip6->ip6_plen);
2433 		if (fin->fin_plen == 0) {
2434 			READ_ENTER(&ifs->ifs_ipf_mutex);
2435 			pass = FR_BLOCK|FR_NOMATCH;
2436 			goto filtered;
2437 		}
2438 		fin->fin_plen += sizeof(ip6_t);
2439 	} else
2440 #endif
2441 	{
2442 #if (OpenBSD >= 200311) && defined(_KERNEL)
2443 		ip->ip_len = ntohs(ip->ip_len);
2444 		ip->ip_off = ntohs(ip->ip_off);
2445 #endif
2446 		fin->fin_plen = ip->ip_len;
2447 	}
2448 
2449 	if (fr_makefrip(hlen, ip, fin) == -1) {
2450 		READ_ENTER(&ifs->ifs_ipf_mutex);
2451 		pass = FR_BLOCK;
2452 		goto filtered;
2453 	}
2454 
2455 	/*
2456 	 * For at least IPv6 packets, if a m_pullup() fails then this pointer
2457 	 * becomes NULL and so we have no packet to free.
2458 	 */
2459 	if (*fin->fin_mp == NULL)
2460 		goto finished;
2461 
2462 	if (!out) {
2463 		if (v == 4) {
2464 #ifdef _KERNEL
2465 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2466 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2467 				fin->fin_flx |= FI_BADSRC;
2468 			}
2469 #endif
2470 			if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2471 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2472 				fin->fin_flx |= FI_LOWTTL;
2473 			}
2474 		}
2475 #ifdef USE_INET6
2476 		else  if (v == 6) {
2477 			ip6 = (ip6_t *)ip;
2478 #ifdef _KERNEL
2479 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2480 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2481 				fin->fin_flx |= FI_BADSRC;
2482 			}
2483 #endif
2484 			if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2485 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2486 				fin->fin_flx |= FI_LOWTTL;
2487 			}
2488 		}
2489 #endif
2490 	}
2491 
2492 	if (fin->fin_flx & FI_SHORT) {
2493 		IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2494 	}
2495 
2496 	READ_ENTER(&ifs->ifs_ipf_mutex);
2497 
2498 	/*
2499 	 * Check auth now.  This, combined with the check below to see if apass
2500 	 * is 0 is to ensure that we don't count the packet twice, which can
2501 	 * otherwise occur when we reprocess it.  As it is, we only count it
2502 	 * after it has no auth. table matchup.  This also stops NAT from
2503 	 * occuring until after the packet has been auth'd.
2504 	 */
2505 	fr = fr_checkauth(fin, &pass);
2506 	if (!out) {
2507 		switch (fin->fin_v)
2508 		{
2509 		case 4 :
2510 			if (fr_checknatin(fin, &pass) == -1) {
2511 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2512 				goto finished;
2513 			}
2514 			break;
2515 #ifdef	USE_INET6
2516 		case 6 :
2517 			if (fr_checknat6in(fin, &pass) == -1) {
2518 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2519 				goto finished;
2520 			}
2521 			break;
2522 #endif
2523 		default :
2524 			break;
2525 		}
2526 	}
2527 	if (!out)
2528 		(void) fr_acctpkt(fin, NULL);
2529 
2530 	if (fr == NULL)
2531 		if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2532 			fr = fr_knownfrag(fin, &pass);
2533 	if (fr == NULL)
2534 		fr = fr_checkstate(fin, &pass);
2535 
2536 	if ((pass & FR_NOMATCH) || (fr == NULL))
2537 		fr = fr_firewall(fin, &pass);
2538 
2539 	fin->fin_fr = fr;
2540 
2541 	/*
2542 	 * Only count/translate packets which will be passed on, out the
2543 	 * interface.
2544 	 */
2545 	if (out && FR_ISPASS(pass)) {
2546 		(void) fr_acctpkt(fin, NULL);
2547 
2548 		switch (fin->fin_v)
2549 		{
2550 		case 4 :
2551 			if (fr_checknatout(fin, &pass) == -1) {
2552 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2553 				goto finished;
2554 			}
2555 			break;
2556 #ifdef	USE_INET6
2557 		case 6 :
2558 			if (fr_checknat6out(fin, &pass) == -1) {
2559 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2560 				goto finished;
2561 			}
2562 			break;
2563 #endif
2564 		default :
2565 			break;
2566 		}
2567 
2568 		if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2569 			if (fr_updateipid(fin) == -1) {
2570 				IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2571 				pass &= ~FR_CMDMASK;
2572 				pass |= FR_BLOCK;
2573 			} else {
2574 				IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2575 			}
2576 		}
2577 	}
2578 
2579 #ifdef	IPFILTER_LOG
2580 	if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2581 		(void) fr_dolog(fin, &pass);
2582 	}
2583 #endif
2584 
2585 	/*
2586 	 * The FI_STATE flag is cleared here so that calling fr_checkstate
2587 	 * will work when called from inside of fr_fastroute.  Although
2588 	 * there is a similar flag, FI_NATED, for NAT, it does have the same
2589 	 * impact on code execution.
2590 	 */
2591 	fin->fin_flx &= ~FI_STATE;
2592 
2593 	/*
2594 	 * Only allow FR_DUP to work if a rule matched - it makes no sense to
2595 	 * set FR_DUP as a "default" as there are no instructions about where
2596 	 * to send the packet.  Use fin_m here because it may have changed
2597 	 * (without an update of 'm') in prior processing.
2598 	 */
2599 	if ((fr != NULL) && (pass & FR_DUP)) {
2600 		mc = M_DUPLICATE(fin->fin_m);
2601 #ifdef _KERNEL
2602 		mc->b_rptr += fin->fin_ipoff;
2603 #endif
2604 	}
2605 
2606 	/*
2607 	 * We don't want to send RST for packets, which are going to be
2608 	 * dropped, just because they don't fit into TCP window. Those packets
2609 	 * will be dropped silently. In other words, we want to drop packet,
2610 	 * while keeping session alive.
2611 	 */
2612 	if ((pass & (FR_RETRST|FR_RETICMP)) && ((fin->fin_flx & FI_OOW) == 0)) {
2613 		/*
2614 		 * Should we return an ICMP packet to indicate error
2615 		 * status passing through the packet filter ?
2616 		 * WARNING: ICMP error packets AND TCP RST packets should
2617 		 * ONLY be sent in repsonse to incoming packets.  Sending them
2618 		 * in response to outbound packets can result in a panic on
2619 		 * some operating systems.
2620 		 */
2621 		if (!out) {
2622 			if (pass & FR_RETICMP) {
2623 				int dst;
2624 
2625 				if ((pass & FR_RETMASK) == FR_FAKEICMP)
2626 					dst = 1;
2627 				else
2628 					dst = 0;
2629 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2630 				/*
2631 				 * Assume it's possible to enter insane rule:
2632 				 * 	pass return-icmp in proto udp ...
2633 				 * then we have no other option than to forward
2634 				 * packet on loopback and give up any attempt
2635 				 * to create a fake response.
2636 				 */
2637 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2638 				    FR_ISBLOCK(pass)) {
2639 
2640 					if (fr_make_icmp(fin) == 0) {
2641 						IPF_BUMP(
2642 						ifs->ifs_frstats[out].fr_ret);
2643 					}
2644 					/*
2645 					 * we drop packet silently in case we
2646 					 * failed assemble fake response for it
2647 					 */
2648 					else if (*mp != NULL) {
2649 						FREE_MB_T(*mp);
2650 						m = *mp = NULL;
2651 					}
2652 
2653 					IPF_BUMP(
2654 					    ifs->ifs_frstats[out].fr_block);
2655 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2656 
2657 					return (0);
2658 				}
2659 #endif	/* _KERNEL && SOLARIS2 >= 10 */
2660 
2661 				(void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2662 				IPF_BUMP(ifs->ifs_frstats[out].fr_ret);
2663 
2664 			} else if (((pass & FR_RETMASK) == FR_RETRST) &&
2665 				   !(fin->fin_flx & FI_SHORT)) {
2666 
2667 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2668 				/*
2669 				 * Assume it's possible to enter insane rule:
2670 				 * 	pass return-rst in proto tcp ...
2671 				 * then we have no other option than to forward
2672 				 * packet on loopback and give up any attempt
2673 				 * to create a fake response.
2674 				 */
2675 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2676 				    FR_ISBLOCK(pass)) {
2677 					if (fr_make_rst(fin) == 0) {
2678 						IPF_BUMP(
2679 						ifs->ifs_frstats[out].fr_ret);
2680 					}
2681 					else if (mp != NULL) {
2682 					/*
2683 					 * we drop packet silently in case we
2684 					 * failed assemble fake response for it
2685 					 */
2686 						FREE_MB_T(*mp);
2687 						m = *mp = NULL;
2688 					}
2689 
2690 					IPF_BUMP(
2691 					    ifs->ifs_frstats[out].fr_block);
2692 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2693 
2694 					return (0);
2695 				 }
2696 #endif /* _KERNEL && _SOLARIS2 >= 10 */
2697 				if (fr_send_reset(fin) == 0) {
2698 					IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2699 				}
2700 			}
2701 		} else {
2702 			if (pass & FR_RETRST)
2703 				fin->fin_error = ECONNRESET;
2704 		}
2705 	}
2706 
2707 	/*
2708 	 * If we didn't drop off the bottom of the list of rules (and thus
2709 	 * the 'current' rule fr is not NULL), then we may have some extra
2710 	 * instructions about what to do with a packet.
2711 	 * Once we're finished return to our caller, freeing the packet if
2712 	 * we are dropping it (* BSD ONLY *).
2713 	 * Reassign m from fin_m as we may have a new buffer, now.
2714 	 */
2715 filtered:
2716 	m = fin->fin_m;
2717 
2718 	if (fr != NULL) {
2719 		frdest_t *fdp;
2720 
2721 		fdp = &fr->fr_tifs[fin->fin_rev];
2722 
2723 		if (!out && (pass & FR_FASTROUTE)) {
2724 			/*
2725 			 * For fastroute rule, no destioation interface defined
2726 			 * so pass NULL as the frdest_t parameter
2727 			 */
2728 			(void) fr_fastroute(m, mp, fin, NULL);
2729 			m = *mp = NULL;
2730 		} else if ((fdp->fd_ifp != NULL) &&
2731 			   (fdp->fd_ifp != (struct ifnet *)-1)) {
2732 			/* this is for to rules: */
2733 			(void) fr_fastroute(m, mp, fin, fdp);
2734 			m = *mp = NULL;
2735 		}
2736 
2737 		/*
2738 		 * Send a duplicated packet.
2739 		 */
2740 		if (mc != NULL) {
2741 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2742 			/*
2743 			 * We are going to compute chksum for copies of loopback packets
2744 			 * only. IP stack does not compute chksums at all for loopback
2745 			 * packets. We want to get it fixed in their copies, since those
2746 			 * are going to be sent to network.
2747 			 */
2748 			if (IPF_IS_LOOPBACK(qpi->qpi_flags))
2749 				fr_calc_chksum(fin, mc);
2750 #endif
2751 			(void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2752 		}
2753 	}
2754 
2755 	if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT))
2756 		nat_uncreate(fin);
2757 
2758 	/*
2759 	 * This late because the likes of fr_fastroute() use fin_fr.
2760 	 */
2761 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2762 
2763 finished:
2764 	if (!FR_ISPASS(pass)) {
2765 		IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2766 		if (*mp != NULL) {
2767 			FREE_MB_T(*mp);
2768 			m = *mp = NULL;
2769 		}
2770 	} else {
2771 		IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2772 #if defined(_KERNEL) && defined(__sgi)
2773 		if ((fin->fin_hbuf != NULL) &&
2774 		    (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2775 			COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2776 		}
2777 #endif
2778 	}
2779 
2780 	SPL_X(s);
2781 
2782 #ifdef _KERNEL
2783 # if OpenBSD >= 200311
2784 	if (FR_ISPASS(pass) && (v == 4)) {
2785 		ip = fin->fin_ip;
2786 		ip->ip_len = ntohs(ip->ip_len);
2787 		ip->ip_off = ntohs(ip->ip_off);
2788 	}
2789 # endif
2790 	return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2791 #else /* _KERNEL */
2792 	FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2793 	if ((pass & FR_NOMATCH) != 0)
2794 		return 1;
2795 
2796 	if ((pass & FR_RETMASK) != 0)
2797 		switch (pass & FR_RETMASK)
2798 		{
2799 		case FR_RETRST :
2800 			return 3;
2801 		case FR_RETICMP :
2802 			return 4;
2803 		case FR_FAKEICMP :
2804 			return 5;
2805 		}
2806 
2807 	switch (pass & FR_CMDMASK)
2808 	{
2809 	case FR_PASS :
2810 		return 0;
2811 	case FR_BLOCK :
2812 		return -1;
2813 	case FR_AUTH :
2814 		return -2;
2815 	case FR_ACCOUNT :
2816 		return -3;
2817 	case FR_PREAUTH :
2818 		return -4;
2819 	}
2820 	return 2;
2821 #endif /* _KERNEL */
2822 }
2823 
2824 
2825 #ifdef	IPFILTER_LOG
2826 /* ------------------------------------------------------------------------ */
2827 /* Function:    fr_dolog                                                    */
2828 /* Returns:     frentry_t* - returns contents of fin_fr (no change made)    */
2829 /* Parameters:  fin(I) - pointer to packet information                      */
2830 /*              passp(IO) - pointer to current/new filter decision (unused) */
2831 /*                                                                          */
2832 /* Checks flags set to see how a packet should be logged, if it is to be    */
2833 /* logged.  Adjust statistics based on its success or not.                  */
2834 /* ------------------------------------------------------------------------ */
2835 frentry_t *fr_dolog(fin, passp)
2836 fr_info_t *fin;
2837 u_32_t *passp;
2838 {
2839 	u_32_t pass;
2840 	int out;
2841 	ipf_stack_t *ifs = fin->fin_ifs;
2842 
2843 	out = fin->fin_out;
2844 	pass = *passp;
2845 
2846 	if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2847 		pass |= FF_LOGNOMATCH;
2848 		IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2849 		goto logit;
2850 	} else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2851 	    (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2852 		if ((pass & FR_LOGMASK) != FR_LOGP)
2853 			pass |= FF_LOGPASS;
2854 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2855 		goto logit;
2856 	} else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2857 		   (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2858 		if ((pass & FR_LOGMASK) != FR_LOGB)
2859 			pass |= FF_LOGBLOCK;
2860 		IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2861 logit:
2862 		if (ipflog(fin, pass) == -1) {
2863 			IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2864 
2865 			/*
2866 			 * If the "or-block" option has been used then
2867 			 * block the packet if we failed to log it.
2868 			 */
2869 			if ((pass & FR_LOGORBLOCK) &&
2870 			    FR_ISPASS(pass)) {
2871 				pass &= ~FR_CMDMASK;
2872 				pass |= FR_BLOCK;
2873 			}
2874 		}
2875 		*passp = pass;
2876 	}
2877 
2878 	return fin->fin_fr;
2879 }
2880 #endif /* IPFILTER_LOG */
2881 
2882 
2883 /* ------------------------------------------------------------------------ */
2884 /* Function:    ipf_cksum                                                   */
2885 /* Returns:     u_short - IP header checksum                                */
2886 /* Parameters:  addr(I) - pointer to start of buffer to checksum            */
2887 /*              len(I)  - length of buffer in bytes                         */
2888 /*                                                                          */
2889 /* Calculate the two's complement 16 bit checksum of the buffer passed.     */
2890 /*                                                                          */
2891 /* N.B.: addr should be 16bit aligned.                                      */
2892 /* ------------------------------------------------------------------------ */
2893 u_short ipf_cksum(addr, len)
2894 u_short *addr;
2895 int len;
2896 {
2897 	u_32_t sum = 0;
2898 
2899 	for (sum = 0; len > 1; len -= 2)
2900 		sum += *addr++;
2901 
2902 	/* mop up an odd byte, if necessary */
2903 	if (len == 1)
2904 		sum += *(u_char *)addr;
2905 
2906 	/*
2907 	 * add back carry outs from top 16 bits to low 16 bits
2908 	 */
2909 	sum = (sum >> 16) + (sum & 0xffff);	/* add hi 16 to low 16 */
2910 	sum += (sum >> 16);			/* add carry */
2911 	return (u_short)(~sum);
2912 }
2913 
2914 
2915 /* ------------------------------------------------------------------------ */
2916 /* Function:    fr_cksum                                                    */
2917 /* Returns:     u_short - layer 4 checksum                                  */
2918 /* Parameters:  m(I  )     - pointer to buffer holding packet               */
2919 /*              ip(I)      - pointer to IP header                           */
2920 /*              l4proto(I) - protocol to caclulate checksum for             */
2921 /*              l4hdr(I)   - pointer to layer 4 header                      */
2922 /*                                                                          */
2923 /* Calculates the TCP checksum for the packet held in "m", using the data   */
2924 /* in the IP header "ip" to seed it.                                        */
2925 /*                                                                          */
2926 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2927 /* and the TCP header.  We also assume that data blocks aren't allocated in */
2928 /* odd sizes.                                                               */
2929 /*                                                                          */
2930 /* Expects ip_len to be in host byte order when called.                     */
2931 /* ------------------------------------------------------------------------ */
2932 u_short fr_cksum(m, ip, l4proto, l4hdr)
2933 mb_t *m;
2934 ip_t *ip;
2935 int l4proto;
2936 void *l4hdr;
2937 {
2938 	u_short *sp, slen, sumsave, l4hlen, *csump;
2939 	u_int sum, sum2;
2940 	int hlen;
2941 #ifdef	USE_INET6
2942 	ip6_t *ip6;
2943 #endif
2944 
2945 	csump = NULL;
2946 	sumsave = 0;
2947 	l4hlen = 0;
2948 	sp = NULL;
2949 	slen = 0;
2950 	hlen = 0;
2951 	sum = 0;
2952 
2953 	/*
2954 	 * Add up IP Header portion
2955 	 */
2956 #ifdef	USE_INET6
2957 	if (IP_V(ip) == 4) {
2958 #endif
2959 		hlen = IP_HL(ip) << 2;
2960 		slen = ip->ip_len - hlen;
2961 		sum = htons((u_short)l4proto);
2962 		sum += htons(slen);
2963 		sp = (u_short *)&ip->ip_src;
2964 		sum += *sp++;	/* ip_src */
2965 		sum += *sp++;
2966 		sum += *sp++;	/* ip_dst */
2967 		sum += *sp++;
2968 #ifdef	USE_INET6
2969 	} else if (IP_V(ip) == 6) {
2970 		ip6 = (ip6_t *)ip;
2971 		hlen = sizeof(*ip6);
2972 		slen = ntohs(ip6->ip6_plen);
2973 		sum = htons((u_short)l4proto);
2974 		sum += htons(slen);
2975 		sp = (u_short *)&ip6->ip6_src;
2976 		sum += *sp++;	/* ip6_src */
2977 		sum += *sp++;
2978 		sum += *sp++;
2979 		sum += *sp++;
2980 		sum += *sp++;
2981 		sum += *sp++;
2982 		sum += *sp++;
2983 		sum += *sp++;
2984 		sum += *sp++;	/* ip6_dst */
2985 		sum += *sp++;
2986 		sum += *sp++;
2987 		sum += *sp++;
2988 		sum += *sp++;
2989 		sum += *sp++;
2990 		sum += *sp++;
2991 		sum += *sp++;
2992 	}
2993 #endif
2994 
2995 	switch (l4proto)
2996 	{
2997 	case IPPROTO_UDP :
2998 		csump = &((udphdr_t *)l4hdr)->uh_sum;
2999 		l4hlen = sizeof(udphdr_t);
3000 		break;
3001 
3002 	case IPPROTO_TCP :
3003 		csump = &((tcphdr_t *)l4hdr)->th_sum;
3004 		l4hlen = sizeof(tcphdr_t);
3005 		break;
3006 	case IPPROTO_ICMP :
3007 		csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
3008 		l4hlen = 4;
3009 		sum = 0;
3010 		break;
3011 	default :
3012 		break;
3013 	}
3014 
3015 	if (csump != NULL) {
3016 		sumsave = *csump;
3017 		*csump = 0;
3018 	}
3019 
3020 	l4hlen = l4hlen;	/* LINT */
3021 
3022 #ifdef	_KERNEL
3023 # ifdef MENTAT
3024 	{
3025 	void *rp = m->b_rptr;
3026 
3027 	if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
3028 		m->b_rptr = (u_char *)ip;
3029 	sum2 = ip_cksum(m, hlen, sum);	/* hlen == offset */
3030 	m->b_rptr = rp;
3031 	sum2 = (sum2 & 0xffff) + (sum2 >> 16);
3032 	sum2 = ~sum2 & 0xffff;
3033 	}
3034 # else /* MENTAT */
3035 #  if defined(BSD) || defined(sun)
3036 #   if BSD >= 199103
3037 	m->m_data += hlen;
3038 #   else
3039 	m->m_off += hlen;
3040 #   endif
3041 	m->m_len -= hlen;
3042 	sum2 = in_cksum(m, slen);
3043 	m->m_len += hlen;
3044 #   if BSD >= 199103
3045 	m->m_data -= hlen;
3046 #   else
3047 	m->m_off -= hlen;
3048 #   endif
3049 	/*
3050 	 * Both sum and sum2 are partial sums, so combine them together.
3051 	 */
3052 	sum += ~sum2 & 0xffff;
3053 	while (sum > 0xffff)
3054 		sum = (sum & 0xffff) + (sum >> 16);
3055 	sum2 = ~sum & 0xffff;
3056 #  else /* defined(BSD) || defined(sun) */
3057 {
3058 	union {
3059 		u_char	c[2];
3060 		u_short	s;
3061 	} bytes;
3062 	u_short len = ip->ip_len;
3063 #   if defined(__sgi)
3064 	int add;
3065 #   endif
3066 
3067 	/*
3068 	 * Add up IP Header portion
3069 	 */
3070 	if (sp != (u_short *)l4hdr)
3071 		sp = (u_short *)l4hdr;
3072 
3073 	switch (l4proto)
3074 	{
3075 	case IPPROTO_UDP :
3076 		sum += *sp++;	/* sport */
3077 		sum += *sp++;	/* dport */
3078 		sum += *sp++;	/* udp length */
3079 		sum += *sp++;	/* checksum */
3080 		break;
3081 
3082 	case IPPROTO_TCP :
3083 		sum += *sp++;	/* sport */
3084 		sum += *sp++;	/* dport */
3085 		sum += *sp++;	/* seq */
3086 		sum += *sp++;
3087 		sum += *sp++;	/* ack */
3088 		sum += *sp++;
3089 		sum += *sp++;	/* off */
3090 		sum += *sp++;	/* win */
3091 		sum += *sp++;	/* checksum */
3092 		sum += *sp++;	/* urp */
3093 		break;
3094 	case IPPROTO_ICMP :
3095 		sum = *sp++;	/* type/code */
3096 		sum += *sp++;	/* checksum */
3097 		break;
3098 	}
3099 
3100 #   ifdef	__sgi
3101 	/*
3102 	 * In case we had to copy the IP & TCP header out of mbufs,
3103 	 * skip over the mbuf bits which are the header
3104 	 */
3105 	if ((caddr_t)ip != mtod(m, caddr_t)) {
3106 		hlen = (caddr_t)sp - (caddr_t)ip;
3107 		while (hlen) {
3108 			add = MIN(hlen, m->m_len);
3109 			sp = (u_short *)(mtod(m, caddr_t) + add);
3110 			hlen -= add;
3111 			if (add == m->m_len) {
3112 				m = m->m_next;
3113 				if (!hlen) {
3114 					if (!m)
3115 						break;
3116 					sp = mtod(m, u_short *);
3117 				}
3118 				PANIC((!m),("fr_cksum(1): not enough data"));
3119 			}
3120 		}
3121 	}
3122 #   endif
3123 
3124 	len -= (l4hlen + hlen);
3125 	if (len <= 0)
3126 		goto nodata;
3127 
3128 	while (len > 1) {
3129 		if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3130 			m = m->m_next;
3131 			PANIC((!m),("fr_cksum(2): not enough data"));
3132 			sp = mtod(m, u_short *);
3133 		}
3134 		if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3135 			bytes.c[0] = *(u_char *)sp;
3136 			m = m->m_next;
3137 			PANIC((!m),("fr_cksum(3): not enough data"));
3138 			sp = mtod(m, u_short *);
3139 			bytes.c[1] = *(u_char *)sp;
3140 			sum += bytes.s;
3141 			sp = (u_short *)((u_char *)sp + 1);
3142 		}
3143 		if ((u_long)sp & 1) {
3144 			bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3145 			sum += bytes.s;
3146 		} else
3147 			sum += *sp++;
3148 		len -= 2;
3149 	}
3150 
3151 	if (len != 0)
3152 		sum += ntohs(*(u_char *)sp << 8);
3153 nodata:
3154 	while (sum > 0xffff)
3155 		sum = (sum & 0xffff) + (sum >> 16);
3156 	sum2 = (u_short)(~sum & 0xffff);
3157 }
3158 #  endif /*  defined(BSD) || defined(sun) */
3159 # endif /* MENTAT */
3160 #else /* _KERNEL */
3161 	for (; slen > 1; slen -= 2)
3162 	        sum += *sp++;
3163 	if (slen)
3164 		sum += ntohs(*(u_char *)sp << 8);
3165 	while (sum > 0xffff)
3166 		sum = (sum & 0xffff) + (sum >> 16);
3167 	sum2 = (u_short)(~sum & 0xffff);
3168 #endif /* _KERNEL */
3169 	if (csump != NULL)
3170 		*csump = sumsave;
3171 	return sum2;
3172 }
3173 
3174 
3175 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3176     defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3177 /*
3178  * Copyright (c) 1982, 1986, 1988, 1991, 1993
3179  *	The Regents of the University of California.  All rights reserved.
3180  *
3181  * Redistribution and use in source and binary forms, with or without
3182  * modification, are permitted provided that the following conditions
3183  * are met:
3184  * 1. Redistributions of source code must retain the above copyright
3185  *    notice, this list of conditions and the following disclaimer.
3186  * 2. Redistributions in binary form must reproduce the above copyright
3187  *    notice, this list of conditions and the following disclaimer in the
3188  *    documentation and/or other materials provided with the distribution.
3189  * 3. Neither the name of the University nor the names of its contributors
3190  *    may be used to endorse or promote products derived from this software
3191  *    without specific prior written permission.
3192  *
3193  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3194  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3195  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3196  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3197  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3198  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3199  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3200  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3201  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3202  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3203  * SUCH DAMAGE.
3204  *
3205  *	@(#)uipc_mbuf.c	8.2 (Berkeley) 1/4/94
3206  * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3207  */
3208 /*
3209  * Copy data from an mbuf chain starting "off" bytes from the beginning,
3210  * continuing for "len" bytes, into the indicated buffer.
3211  */
3212 void
3213 m_copydata(m, off, len, cp)
3214 	mb_t *m;
3215 	int off;
3216 	int len;
3217 	caddr_t cp;
3218 {
3219 	unsigned count;
3220 
3221 	if (off < 0 || len < 0)
3222 		panic("m_copydata");
3223 	while (off > 0) {
3224 		if (m == 0)
3225 			panic("m_copydata");
3226 		if (off < m->m_len)
3227 			break;
3228 		off -= m->m_len;
3229 		m = m->m_next;
3230 	}
3231 	while (len > 0) {
3232 		if (m == 0)
3233 			panic("m_copydata");
3234 		count = MIN(m->m_len - off, len);
3235 		bcopy(mtod(m, caddr_t) + off, cp, count);
3236 		len -= count;
3237 		cp += count;
3238 		off = 0;
3239 		m = m->m_next;
3240 	}
3241 }
3242 
3243 
3244 /*
3245  * Copy data from a buffer back into the indicated mbuf chain,
3246  * starting "off" bytes from the beginning, extending the mbuf
3247  * chain if necessary.
3248  */
3249 void
3250 m_copyback(m0, off, len, cp)
3251 	struct	mbuf *m0;
3252 	int off;
3253 	int len;
3254 	caddr_t cp;
3255 {
3256 	int mlen;
3257 	struct mbuf *m = m0, *n;
3258 	int totlen = 0;
3259 
3260 	if (m0 == 0)
3261 		return;
3262 	while (off > (mlen = m->m_len)) {
3263 		off -= mlen;
3264 		totlen += mlen;
3265 		if (m->m_next == 0) {
3266 			n = m_getclr(M_DONTWAIT, m->m_type);
3267 			if (n == 0)
3268 				goto out;
3269 			n->m_len = min(MLEN, len + off);
3270 			m->m_next = n;
3271 		}
3272 		m = m->m_next;
3273 	}
3274 	while (len > 0) {
3275 		mlen = min(m->m_len - off, len);
3276 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3277 		cp += mlen;
3278 		len -= mlen;
3279 		mlen += off;
3280 		off = 0;
3281 		totlen += mlen;
3282 		if (len == 0)
3283 			break;
3284 		if (m->m_next == 0) {
3285 			n = m_get(M_DONTWAIT, m->m_type);
3286 			if (n == 0)
3287 				break;
3288 			n->m_len = min(MLEN, len);
3289 			m->m_next = n;
3290 		}
3291 		m = m->m_next;
3292 	}
3293 out:
3294 #if 0
3295 	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3296 		m->m_pkthdr.len = totlen;
3297 #endif
3298 	return;
3299 }
3300 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3301 
3302 
3303 /* ------------------------------------------------------------------------ */
3304 /* Function:    fr_findgroup                                                */
3305 /* Returns:     frgroup_t * - NULL = group not found, else pointer to group */
3306 /* Parameters:  group(I) - group name to search for                         */
3307 /*              unit(I)  - device to which this group belongs               */
3308 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3309 /*              fgpp(O)  - pointer to place to store pointer to the pointer */
3310 /*                         to where to add the next (last) group or where   */
3311 /*                         to delete group from.                            */
3312 /*                                                                          */
3313 /* Search amongst the defined groups for a particular group number.         */
3314 /* ------------------------------------------------------------------------ */
3315 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3316 char *group;
3317 minor_t unit;
3318 int set;
3319 frgroup_t ***fgpp;
3320 ipf_stack_t *ifs;
3321 {
3322 	frgroup_t *fg, **fgp;
3323 
3324 	/*
3325 	 * Which list of groups to search in is dependent on which list of
3326 	 * rules are being operated on.
3327 	 */
3328 	fgp = &ifs->ifs_ipfgroups[unit][set];
3329 
3330 	while ((fg = *fgp) != NULL) {
3331 		if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3332 			break;
3333 		else
3334 			fgp = &fg->fg_next;
3335 	}
3336 	if (fgpp != NULL)
3337 		*fgpp = fgp;
3338 	return fg;
3339 }
3340 
3341 
3342 /* ------------------------------------------------------------------------ */
3343 /* Function:    fr_addgroup                                                 */
3344 /* Returns:     frgroup_t * - NULL == did not create group,                 */
3345 /*                            != NULL == pointer to the group               */
3346 /* Parameters:  num(I)   - group number to add                              */
3347 /*              head(I)  - rule pointer that is using this as the head      */
3348 /*              flags(I) - rule flags which describe the type of rule it is */
3349 /*              unit(I)  - device to which this group will belong to        */
3350 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3351 /* Write Locks: ipf_mutex                                                   */
3352 /*                                                                          */
3353 /* Add a new group head, or if it already exists, increase the reference    */
3354 /* count to it.                                                             */
3355 /* ------------------------------------------------------------------------ */
3356 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3357 char *group;
3358 void *head;
3359 u_32_t flags;
3360 minor_t unit;
3361 int set;
3362 ipf_stack_t *ifs;
3363 {
3364 	frgroup_t *fg, **fgp;
3365 	u_32_t gflags;
3366 
3367 	if (group == NULL)
3368 		return NULL;
3369 
3370 	if (unit == IPL_LOGIPF && *group == '\0')
3371 		return NULL;
3372 
3373 	fgp = NULL;
3374 	gflags = flags & FR_INOUT;
3375 
3376 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3377 	if (fg != NULL) {
3378 		if (fg->fg_flags == 0)
3379 			fg->fg_flags = gflags;
3380 		else if (gflags != fg->fg_flags)
3381 			return NULL;
3382 		fg->fg_ref++;
3383 		return fg;
3384 	}
3385 	KMALLOC(fg, frgroup_t *);
3386 	if (fg != NULL) {
3387 		fg->fg_head = head;
3388 		fg->fg_start = NULL;
3389 		fg->fg_next = *fgp;
3390 		bcopy(group, fg->fg_name, FR_GROUPLEN);
3391 		fg->fg_flags = gflags;
3392 		fg->fg_ref = 1;
3393 		*fgp = fg;
3394 	}
3395 	return fg;
3396 }
3397 
3398 
3399 /* ------------------------------------------------------------------------ */
3400 /* Function:    fr_delgroup                                                 */
3401 /* Returns:     Nil                                                         */
3402 /* Parameters:  group(I) - group name to delete                             */
3403 /*              unit(I)  - device to which this group belongs               */
3404 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3405 /* Write Locks: ipf_mutex                                                   */
3406 /*                                                                          */
3407 /* Attempt to delete a group head.                                          */
3408 /* Only do this when its reference count reaches 0.                         */
3409 /* ------------------------------------------------------------------------ */
3410 void fr_delgroup(group, unit, set, ifs)
3411 char *group;
3412 minor_t unit;
3413 int set;
3414 ipf_stack_t *ifs;
3415 {
3416 	frgroup_t *fg, **fgp;
3417 
3418 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3419 	if (fg == NULL)
3420 		return;
3421 
3422 	fg->fg_ref--;
3423 	if (fg->fg_ref == 0) {
3424 		*fgp = fg->fg_next;
3425 		KFREE(fg);
3426 	}
3427 }
3428 
3429 
3430 /* ------------------------------------------------------------------------ */
3431 /* Function:    fr_getrulen                                                 */
3432 /* Returns:     frentry_t * - NULL == not found, else pointer to rule n     */
3433 /* Parameters:  unit(I)  - device for which to count the rule's number      */
3434 /*              flags(I) - which set of rules to find the rule in           */
3435 /*              group(I) - group name                                       */
3436 /*              n(I)     - rule number to find                              */
3437 /*                                                                          */
3438 /* Find rule # n in group # g and return a pointer to it.  Return NULl if   */
3439 /* group # g doesn't exist or there are less than n rules in the group.     */
3440 /* ------------------------------------------------------------------------ */
3441 frentry_t *fr_getrulen(unit, group, n, ifs)
3442 int unit;
3443 char *group;
3444 u_32_t n;
3445 ipf_stack_t *ifs;
3446 {
3447 	frentry_t *fr;
3448 	frgroup_t *fg;
3449 
3450 	fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3451 	if (fg == NULL)
3452 		return NULL;
3453 	for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3454 		;
3455 	if (n != 0)
3456 		return NULL;
3457 	return fr;
3458 }
3459 
3460 
3461 /* ------------------------------------------------------------------------ */
3462 /* Function:    fr_rulen                                                    */
3463 /* Returns:     int - >= 0 - rule number, -1 == search failed               */
3464 /* Parameters:  unit(I) - device for which to count the rule's number       */
3465 /*              fr(I)   - pointer to rule to match                          */
3466 /*                                                                          */
3467 /* Return the number for a rule on a specific filtering device.             */
3468 /* ------------------------------------------------------------------------ */
3469 int fr_rulen(unit, fr, ifs)
3470 int unit;
3471 frentry_t *fr;
3472 ipf_stack_t *ifs;
3473 {
3474 	frentry_t *fh;
3475 	frgroup_t *fg;
3476 	u_32_t n = 0;
3477 
3478 	if (fr == NULL)
3479 		return -1;
3480 	fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3481 	if (fg == NULL)
3482 		return -1;
3483 	for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3484 		if (fh == fr)
3485 			break;
3486 	if (fh == NULL)
3487 		return -1;
3488 	return n;
3489 }
3490 
3491 
3492 /* ------------------------------------------------------------------------ */
3493 /* Function:    frflushlist                                                 */
3494 /* Returns:     int - >= 0 - number of flushed rules                        */
3495 /* Parameters:  set(I)   - which set of rules (inactive/inactive) this is   */
3496 /*              unit(I)  - device for which to flush rules                  */
3497 /*              flags(I) - which set of rules to flush                      */
3498 /*              nfreedp(O) - pointer to int where flush count is stored     */
3499 /*              listp(I)   - pointer to list to flush pointer               */
3500 /* Write Locks: ipf_mutex                                                   */
3501 /*                                                                          */
3502 /* Recursively flush rules from the list, descending groups as they are     */
3503 /* encountered.  if a rule is the head of a group and it has lost all its   */
3504 /* group members, then also delete the group reference.  nfreedp is needed  */
3505 /* to store the accumulating count of rules removed, whereas the returned   */
3506 /* value is just the number removed from the current list.  The latter is   */
3507 /* needed to correctly adjust reference counts on rules that define groups. */
3508 /*                                                                          */
3509 /* NOTE: Rules not loaded from user space cannot be flushed.                */
3510 /* ------------------------------------------------------------------------ */
3511 static int frflushlist(set, unit, nfreedp, listp, ifs)
3512 int set;
3513 minor_t unit;
3514 int *nfreedp;
3515 frentry_t **listp;
3516 ipf_stack_t *ifs;
3517 {
3518 	int freed = 0;
3519 	frentry_t *fp;
3520 
3521 	while ((fp = *listp) != NULL) {
3522 		if ((fp->fr_type & FR_T_BUILTIN) ||
3523 		    !(fp->fr_flags & FR_COPIED)) {
3524 			listp = &fp->fr_next;
3525 			continue;
3526 		}
3527 		*listp = fp->fr_next;
3528 		if (fp->fr_grp != NULL) {
3529 			(void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3530 		}
3531 
3532 		if (fp->fr_grhead != NULL) {
3533 			fr_delgroup(fp->fr_grhead, unit, set, ifs);
3534 			*fp->fr_grhead = '\0';
3535 		}
3536 
3537 		ASSERT(fp->fr_ref > 0);
3538 		fp->fr_next = NULL;
3539 		if (fr_derefrule(&fp, ifs) == 0)
3540 			freed++;
3541 	}
3542 	*nfreedp += freed;
3543 	return freed;
3544 }
3545 
3546 
3547 /* ------------------------------------------------------------------------ */
3548 /* Function:    frflush                                                     */
3549 /* Returns:     int - >= 0 - number of flushed rules                        */
3550 /* Parameters:  unit(I)  - device for which to flush rules                  */
3551 /*              flags(I) - which set of rules to flush                      */
3552 /*                                                                          */
3553 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3554 /* and IPv6) as defined by the value of flags.                              */
3555 /* ------------------------------------------------------------------------ */
3556 int frflush(unit, proto, flags, ifs)
3557 minor_t unit;
3558 int proto, flags;
3559 ipf_stack_t *ifs;
3560 {
3561 	int flushed = 0, set;
3562 
3563 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3564 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
3565 
3566 	set = ifs->ifs_fr_active;
3567 	if ((flags & FR_INACTIVE) == FR_INACTIVE)
3568 		set = 1 - set;
3569 
3570 	if (flags & FR_OUTQUE) {
3571 		if (proto == 0 || proto == 6) {
3572 			(void) frflushlist(set, unit,
3573 			    &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3574 			(void) frflushlist(set, unit,
3575 			    &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3576 		}
3577 		if (proto == 0 || proto == 4) {
3578 			(void) frflushlist(set, unit,
3579 			    &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3580 			(void) frflushlist(set, unit,
3581 			    &flushed, &ifs->ifs_ipacct[1][set], ifs);
3582 		}
3583 	}
3584 	if (flags & FR_INQUE) {
3585 		if (proto == 0 || proto == 6) {
3586 			(void) frflushlist(set, unit,
3587 			    &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3588 			(void) frflushlist(set, unit,
3589 			    &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3590 		}
3591 		if (proto == 0 || proto == 4) {
3592 			(void) frflushlist(set, unit,
3593 			    &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3594 			(void) frflushlist(set, unit,
3595 			    &flushed, &ifs->ifs_ipacct[0][set], ifs);
3596 		}
3597 	}
3598 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3599 
3600 	if (unit == IPL_LOGIPF) {
3601 		int tmp;
3602 
3603 		tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3604 		if (tmp >= 0)
3605 			flushed += tmp;
3606 	}
3607 	return flushed;
3608 }
3609 
3610 
3611 /* ------------------------------------------------------------------------ */
3612 /* Function:    memstr                                                      */
3613 /* Returns:     char *  - NULL if failed, != NULL pointer to matching bytes */
3614 /* Parameters:  src(I)  - pointer to byte sequence to match                 */
3615 /*              dst(I)  - pointer to byte sequence to search                */
3616 /*              slen(I) - match length                                      */
3617 /*              dlen(I) - length available to search in                     */
3618 /*                                                                          */
3619 /* Search dst for a sequence of bytes matching those at src and extend for  */
3620 /* slen bytes.                                                              */
3621 /* ------------------------------------------------------------------------ */
3622 char *memstr(src, dst, slen, dlen)
3623 char *src, *dst;
3624 int slen, dlen;
3625 {
3626 	char *s = NULL;
3627 
3628 	while (dlen >= slen) {
3629 		if (bcmp(src, dst, slen) == 0) {
3630 			s = dst;
3631 			break;
3632 		}
3633 		dst++;
3634 		dlen--;
3635 	}
3636 	return s;
3637 }
3638 /* ------------------------------------------------------------------------ */
3639 /* Function:    fr_fixskip                                                  */
3640 /* Returns:     Nil                                                         */
3641 /* Parameters:  listp(IO)    - pointer to start of list with skip rule      */
3642 /*              rp(I)        - rule added/removed with skip in it.          */
3643 /*              addremove(I) - adjustment (-1/+1) to make to skip count,    */
3644 /*                             depending on whether a rule was just added   */
3645 /*                             or removed.                                  */
3646 /*                                                                          */
3647 /* Adjust all the rules in a list which would have skip'd past the position */
3648 /* where we are inserting to skip to the right place given the change.      */
3649 /* ------------------------------------------------------------------------ */
3650 void fr_fixskip(listp, rp, addremove)
3651 frentry_t **listp, *rp;
3652 int addremove;
3653 {
3654 	int rules, rn;
3655 	frentry_t *fp;
3656 
3657 	rules = 0;
3658 	for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3659 		rules++;
3660 
3661 	if (!fp)
3662 		return;
3663 
3664 	for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3665 		if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3666 			fp->fr_arg += addremove;
3667 }
3668 
3669 
3670 #ifdef	_KERNEL
3671 /* ------------------------------------------------------------------------ */
3672 /* Function:    count4bits                                                  */
3673 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3674 /* Parameters:  ip(I) - 32bit IP address                                    */
3675 /*                                                                          */
3676 /* IPv4 ONLY                                                                */
3677 /* count consecutive 1's in bit mask.  If the mask generated by counting    */
3678 /* consecutive 1's is different to that passed, return -1, else return #    */
3679 /* of bits.                                                                 */
3680 /* ------------------------------------------------------------------------ */
3681 int	count4bits(ip)
3682 u_32_t	ip;
3683 {
3684 	u_32_t	ipn;
3685 	int	cnt = 0, i, j;
3686 
3687 	ip = ipn = ntohl(ip);
3688 	for (i = 32; i; i--, ipn *= 2)
3689 		if (ipn & 0x80000000)
3690 			cnt++;
3691 		else
3692 			break;
3693 	ipn = 0;
3694 	for (i = 32, j = cnt; i; i--, j--) {
3695 		ipn *= 2;
3696 		if (j > 0)
3697 			ipn++;
3698 	}
3699 	if (ipn == ip)
3700 		return cnt;
3701 	return -1;
3702 }
3703 
3704 
3705 #ifdef USE_INET6
3706 /* ------------------------------------------------------------------------ */
3707 /* Function:    count6bits                                                  */
3708 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3709 /* Parameters:  msk(I) - pointer to start of IPv6 bitmask                   */
3710 /*                                                                          */
3711 /* IPv6 ONLY                                                                */
3712 /* count consecutive 1's in bit mask.                                       */
3713 /* ------------------------------------------------------------------------ */
3714 int count6bits(msk)
3715 u_32_t *msk;
3716 {
3717 	int i = 0, k;
3718 	u_32_t j;
3719 
3720 	for (k = 3; k >= 0; k--)
3721 		if (msk[k] == 0xffffffff)
3722 			i += 32;
3723 		else {
3724 			for (j = msk[k]; j; j <<= 1)
3725 				if (j & 0x80000000)
3726 					i++;
3727 		}
3728 	return i;
3729 }
3730 # endif
3731 #endif /* _KERNEL */
3732 
3733 
3734 /* ------------------------------------------------------------------------ */
3735 /* Function:    fr_ifsync                                                   */
3736 /* Returns:     void *    - new interface identifier                        */
3737 /* Parameters:  action(I)  - type of synchronisation to do                  */
3738 /*              v(I)       - IP version being sync'd (v4 or v6)             */
3739 /*              newifp(I)  - interface identifier being introduced/removed  */
3740 /*              oldifp(I)  - interface identifier in a filter rule          */
3741 /*              newname(I) - name associated with newifp interface          */
3742 /*              oldname(I) - name associated with oldifp interface          */
3743 /*		ifs       - pointer to IPF stack instance		    */
3744 /*                                                                          */
3745 /* This function returns what the new value for "oldifp" should be for its  */
3746 /* caller.  In some cases it will not change, in some it will.              */
3747 /* action == IPFSYNC_RESYNC                                                 */
3748 /*   a new value for oldifp will always be looked up, according to oldname, */
3749 /*   the values of newname and newifp are ignored.                          */
3750 /* action == IPFSYNC_NEWIFP                                                 */
3751 /*   if oldname matches newname then we are doing a sync for the matching   */
3752 /*   interface, so we return newifp to be used in place of oldifp.  If the  */
3753 /*   the names don't match, just return oldifp.                             */
3754 /* action == IPFSYNC_OLDIFP                                                 */
3755 /*   if oldifp matches newifp then we are are doing a sync to remove any    */
3756 /*   references to oldifp, so we return "-1".                               */
3757 /* -----								    */
3758 /* NOTE:								    */
3759 /* This function processes NIC event from PF_HOOKS. The action parameter    */
3760 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is	    */
3761 /* one single switch statement() in ipf_nic_event_vx() function, which	    */
3762 /* translates the HOOK event type to action parameter passed to fr_ifsync.  */
3763 /* The translation table looks as follows:				    */
3764 /*	event		| action					    */
3765 /*	----------------+-------------					    */
3766 /*	NE_PLUMB	| IPFSYNC_NEWIFP				    */
3767 /*	NE_UNPLUMB	| IPFSYNC_OLDIFP				    */
3768 /*    NE_ADDRESS_CHANGE	| IPFSYNC_RESYNC				    */
3769 /*									    */
3770 /* The oldname and oldifp parameters are taken from IPF entry (rule, state  */
3771 /* table entry, NAT table entry, fragment ...). The newname and newifp	    */
3772 /* parameters come from hook event data, parameters are taken from event    */
3773 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is	    */
3774 /* notified by hook function.						    */
3775 /*									    */
3776 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3777 /* to plumb the interface, which is already plumbed. In such case we always */
3778 /* get the event from PF_HOOKS as follows:				    */
3779 /*	event:	NE_PLUMB						    */
3780 /*	NIC:	0x0							    */
3781 /* ------------------------------------------------------------------------ */
3782 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3783 int action, v;
3784 char *newname, *oldname;
3785 void *newifp, *oldifp;
3786 ipf_stack_t *ifs;
3787 {
3788 	void *rval = oldifp;
3789 
3790 	switch (action)
3791 	{
3792 	case IPFSYNC_RESYNC :
3793 		if (oldname[0] != '\0') {
3794 			rval = fr_resolvenic(oldname, v, ifs);
3795 		}
3796 		break;
3797 	case IPFSYNC_NEWIFP :
3798 		if (!strncmp(newname, oldname, LIFNAMSIZ))
3799 			rval = newifp;
3800 		break;
3801 	case IPFSYNC_OLDIFP :
3802 		/*
3803 		 * If interface gets unplumbed it must be invalidated, which
3804 		 * means set all existing references to the interface to -1.
3805 		 * We don't want to invalidate references for wildcard
3806 		 * (unbound) rules (entries).
3807 		 */
3808 		if (newifp == oldifp)
3809 			rval = (oldifp) ? (void *)-1 : NULL;
3810 		break;
3811 	}
3812 
3813 	return rval;
3814 }
3815 
3816 
3817 /* ------------------------------------------------------------------------ */
3818 /* Function:    frsynclist                                                  */
3819 /* Returns:     void                                                        */
3820 /* Parameters:  action(I) - type of synchronisation to do                   */
3821 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3822 /*              ifp(I)    - interface identifier associated with action     */
3823 /*              ifname(I) - name associated with ifp parameter              */
3824 /*              fr(I)     - pointer to filter rule                          */
3825 /*		ifs       - pointer to IPF stack instance		    */
3826 /* Write Locks: ipf_mutex                                                   */
3827 /*                                                                          */
3828 /* Walk through a list of filter rules and resolve any interface names into */
3829 /* pointers.  Where dynamic addresses are used, also update the IP address  */
3830 /* used in the rule.  The interface pointer is used to limit the lookups to */
3831 /* a specific set of matching names if it is non-NULL.                      */
3832 /* ------------------------------------------------------------------------ */
3833 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3834 int action, v;
3835 void *ifp;
3836 char *ifname;
3837 frentry_t *fr;
3838 ipf_stack_t *ifs;
3839 {
3840 	frdest_t *fdp;
3841 	int rv, i;
3842 
3843 	for (; fr; fr = fr->fr_next) {
3844 		rv = fr->fr_v;
3845 		if (v != 0 && v != rv)
3846 			continue;
3847 
3848 		/*
3849 		 * Lookup all the interface names that are part of the rule.
3850 		 */
3851 		for (i = 0; i < 4; i++) {
3852 			fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3853 						   fr->fr_ifnames[i],
3854 						   ifp, fr->fr_ifas[i],
3855 						   ifs);
3856 		}
3857 
3858 		fdp = &fr->fr_tifs[0];
3859 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3860 					   ifp, fdp->fd_ifp, ifs);
3861 
3862 		fdp = &fr->fr_tifs[1];
3863 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3864 					   ifp, fdp->fd_ifp, ifs);
3865 
3866 		fdp = &fr->fr_dif;
3867 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3868 					   ifp, fdp->fd_ifp, ifs);
3869 
3870 		if (action != IPFSYNC_RESYNC)
3871 			continue;
3872 
3873 		if (fr->fr_type == FR_T_IPF) {
3874 			if (fr->fr_satype != FRI_NORMAL &&
3875 			    fr->fr_satype != FRI_LOOKUP) {
3876 				(void)fr_ifpaddr(rv, fr->fr_satype,
3877 						 fr->fr_ifas[fr->fr_sifpidx],
3878 						 &fr->fr_src, &fr->fr_smsk,
3879 						 ifs);
3880 			}
3881 			if (fr->fr_datype != FRI_NORMAL &&
3882 			    fr->fr_datype != FRI_LOOKUP) {
3883 				(void)fr_ifpaddr(rv, fr->fr_datype,
3884 						 fr->fr_ifas[fr->fr_difpidx],
3885 						 &fr->fr_dst, &fr->fr_dmsk,
3886 						 ifs);
3887 			}
3888 		}
3889 
3890 #ifdef	IPFILTER_LOOKUP
3891 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3892 		    fr->fr_srcptr == NULL) {
3893 			fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3894 							 fr->fr_srcnum,
3895 							 &fr->fr_srcfunc, ifs);
3896 		}
3897 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3898 		    fr->fr_dstptr == NULL) {
3899 			fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3900 							 fr->fr_dstnum,
3901 							 &fr->fr_dstfunc, ifs);
3902 		}
3903 #endif
3904 	}
3905 }
3906 
3907 
3908 #ifdef	_KERNEL
3909 /* ------------------------------------------------------------------------ */
3910 /* Function:    frsync                                                      */
3911 /* Returns:     void                                                        */
3912 /* Parameters:  action(I) - type of synchronisation to do                   */
3913 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3914 /*              ifp(I)    - interface identifier associated with action     */
3915 /*              name(I)   - name associated with ifp parameter              */
3916 /*                                                                          */
3917 /* frsync() is called when we suspect that the interface list or            */
3918 /* information about interfaces (like IP#) has changed.  Go through all     */
3919 /* filter rules, NAT entries and the state table and check if anything      */
3920 /* needs to be changed/updated.                                             */
3921 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3922 /* in which this was done to support three different types of sync:         */
3923 /* - complete resync of all interface name/identifiers                      */
3924 /* - new interface being announced with its name and identifier             */
3925 /* - interface removal being announced by only its identifier               */
3926 /* ------------------------------------------------------------------------ */
3927 void frsync(action, v, ifp, name, ifs)
3928 int action, v;
3929 void *ifp;
3930 char *name;
3931 ipf_stack_t *ifs;
3932 {
3933 	int i;
3934 
3935 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3936 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3937 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3938 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3939 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3940 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3941 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3942 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3943 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3944 
3945 	for (i = 0; i < IPL_LOGSIZE; i++) {
3946 		frgroup_t *g;
3947 
3948 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3949 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3950 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3951 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3952 	}
3953 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3954 }
3955 
3956 #if SOLARIS2 >= 10
3957 /* ------------------------------------------------------------------------ */
3958 /* Function:    fr_syncindex						    */
3959 /* Returns:     void							    */
3960 /* Parameters:  rules	  - list of rules to be sync'd			    */
3961 /*		ifp	  - interface, which is being sync'd		    */
3962 /*		newifp	  - new ifindex value for interface		    */
3963 /*                                                                          */
3964 /* Function updates all NIC indecis, which match ifp, in every rule. Every  */
3965 /* NIC index matching ifp, will be updated to newifp.			    */
3966 /* ------------------------------------------------------------------------ */
3967 static void fr_syncindex(rules, ifp, newifp)
3968 frentry_t *rules;
3969 void *ifp;
3970 void *newifp;
3971 {
3972 	int i;
3973 	frentry_t *fr;
3974 
3975 	for (fr = rules; fr != NULL; fr = fr->fr_next) {
3976 		/*
3977 		 * Lookup all the interface names that are part of the rule.
3978 		 */
3979 		for (i = 0; i < 4; i++)
3980 			if (fr->fr_ifas[i] == ifp)
3981 				fr->fr_ifas[i] = newifp;
3982 
3983 		for (i = 0; i < 2; i++) {
3984 			if (fr->fr_tifs[i].fd_ifp == ifp)
3985 				fr->fr_tifs[i].fd_ifp = newifp;
3986 		}
3987 
3988 		if (fr->fr_dif.fd_ifp == ifp)
3989 			fr->fr_dif.fd_ifp = newifp;
3990 	}
3991 }
3992 
3993 /* ------------------------------------------------------------------------ */
3994 /* Function:    fr_ifindexsync						    */
3995 /* Returns:     void							    */
3996 /* Parameters:	ifp	  - interface, which is being sync'd		    */
3997 /*		newifp	  - new ifindex value for interface		    */
3998 /*              ifs	  - IPF's stack					    */
3999 /*                                                                          */
4000 /* Function assumes ipf_mutex is locked exclusively.			    */
4001 /* 									    */
4002 /* Function updates the NIC references in rules with new interfaces index   */
4003 /* (newifp). Function must process active lists:			    */
4004 /*	with accounting rules (IPv6 and IPv4)				    */
4005 /*	with inbound rules (IPv6 and IPv4)				    */
4006 /*	with outbound rules (IPv6 and IPv4)				    */
4007 /* Function also has to take care of rule groups.			    */
4008 /*                                                                          */
4009 /* NOTE: The ipf_mutex is grabbed exclusively by caller (which is always    */
4010 /* nic_event_hook). The hook function also updates state entries, NAT rules */
4011 /* and NAT entries. We want to do all these update atomically to keep the   */
4012 /* NIC references consistent. The ipf_mutex will synchronize event with	    */
4013 /* fr_check(), which processes packets,	so no packet will enter fr_check(), */
4014 /* while NIC references will be synchronized.				    */
4015 /* ------------------------------------------------------------------------ */
4016 void fr_ifindexsync(ifp, newifp, ifs)
4017 void *ifp;
4018 void *newifp;
4019 ipf_stack_t *ifs;
4020 {
4021 	unsigned int	i;
4022 	frentry_t *rule_lists[8];
4023 	unsigned int	rules = sizeof (rule_lists) / sizeof (frentry_t *);
4024 
4025 	rule_lists[0] = ifs->ifs_ipacct[0][ifs->ifs_fr_active];
4026 	rule_lists[1] =	ifs->ifs_ipacct[1][ifs->ifs_fr_active];
4027 	rule_lists[2] =	ifs->ifs_ipfilter[0][ifs->ifs_fr_active];
4028 	rule_lists[3] =	ifs->ifs_ipfilter[1][ifs->ifs_fr_active];
4029 	rule_lists[4] =	ifs->ifs_ipacct6[0][ifs->ifs_fr_active];
4030 	rule_lists[5] =	ifs->ifs_ipacct6[1][ifs->ifs_fr_active];
4031 	rule_lists[6] =	ifs->ifs_ipfilter6[0][ifs->ifs_fr_active];
4032 	rule_lists[7] =	ifs->ifs_ipfilter6[1][ifs->ifs_fr_active];
4033 
4034 	for (i = 0; i < rules; i++) {
4035 		fr_syncindex(rule_lists[i], ifp, newifp);
4036 	}
4037 
4038 	/*
4039 	 * Update rule groups.
4040 	 */
4041 	for (i = 0; i < IPL_LOGSIZE; i++) {
4042 		frgroup_t *g;
4043 
4044 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
4045 			fr_syncindex(g->fg_start, ifp, newifp);
4046 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
4047 			fr_syncindex(g->fg_start, ifp, newifp);
4048 	}
4049 }
4050 #endif
4051 
4052 /*
4053  * In the functions below, bcopy() is called because the pointer being
4054  * copied _from_ in this instance is a pointer to a char buf (which could
4055  * end up being unaligned) and on the kernel's local stack.
4056  */
4057 /* ------------------------------------------------------------------------ */
4058 /* Function:    copyinptr                                                   */
4059 /* Returns:     int - 0 = success, else failure                             */
4060 /* Parameters:  src(I)  - pointer to the source address                     */
4061 /*              dst(I)  - destination address                               */
4062 /*              size(I) - number of bytes to copy                           */
4063 /*                                                                          */
4064 /* Copy a block of data in from user space, given a pointer to the pointer  */
4065 /* to start copying from (src) and a pointer to where to store it (dst).    */
4066 /* NB: src - pointer to user space pointer, dst - kernel space pointer      */
4067 /* ------------------------------------------------------------------------ */
4068 int copyinptr(src, dst, size)
4069 void *src, *dst;
4070 size_t size;
4071 {
4072 	caddr_t ca;
4073 	int err;
4074 
4075 # if SOLARIS
4076 	err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
4077 	if (err != 0)
4078 		return err;
4079 # else
4080 	bcopy(src, (caddr_t)&ca, sizeof(ca));
4081 # endif
4082 	err = COPYIN(ca, dst, size);
4083 	return err;
4084 }
4085 
4086 
4087 /* ------------------------------------------------------------------------ */
4088 /* Function:    copyoutptr                                                  */
4089 /* Returns:     int - 0 = success, else failure                             */
4090 /* Parameters:  src(I)  - pointer to the source address                     */
4091 /*              dst(I)  - destination address                               */
4092 /*              size(I) - number of bytes to copy                           */
4093 /*                                                                          */
4094 /* Copy a block of data out to user space, given a pointer to the pointer   */
4095 /* to start copying from (src) and a pointer to where to store it (dst).    */
4096 /* NB: src - kernel space pointer, dst - pointer to user space pointer.     */
4097 /* ------------------------------------------------------------------------ */
4098 int copyoutptr(src, dst, size)
4099 void *src, *dst;
4100 size_t size;
4101 {
4102 	caddr_t ca;
4103 	int err;
4104 
4105 # if SOLARIS
4106 	err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
4107 	if (err != 0)
4108 		return err;
4109 # else
4110 	bcopy(dst, (caddr_t)&ca, sizeof(ca));
4111 # endif
4112 	err = COPYOUT(src, ca, size);
4113 	return err;
4114 }
4115 #endif
4116 
4117 
4118 /* ------------------------------------------------------------------------ */
4119 /* Function:    fr_lock                                                     */
4120 /* Returns:	int - 0 = success, else error				    */
4121 /* Parameters:  data(I)  - pointer to lock value to set                     */
4122 /*              lockp(O) - pointer to location to store old lock value      */
4123 /*                                                                          */
4124 /* Get the new value for the lock integer, set it and return the old value  */
4125 /* in *lockp.                                                               */
4126 /* ------------------------------------------------------------------------ */
4127 int fr_lock(data, lockp)
4128 caddr_t data;
4129 int *lockp;
4130 {
4131 	int arg, err;
4132 
4133 	err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
4134 	if (err != 0)
4135 		return (EFAULT);
4136 	err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
4137 	if (err != 0)
4138 		return (EFAULT);
4139 	*lockp = arg;
4140 	return (0);
4141 }
4142 
4143 
4144 /* ------------------------------------------------------------------------ */
4145 /* Function:    fr_getstat                                                  */
4146 /* Returns:     Nil                                                         */
4147 /* Parameters:  fiop(I)  - pointer to ipfilter stats structure              */
4148 /*                                                                          */
4149 /* Stores a copy of current pointers, counters, etc, in the friostat        */
4150 /* structure.                                                               */
4151 /* ------------------------------------------------------------------------ */
4152 void fr_getstat(fiop, ifs)
4153 friostat_t *fiop;
4154 ipf_stack_t *ifs;
4155 {
4156 	int i, j;
4157 
4158 	bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
4159 	    sizeof(filterstats_t) * 2);
4160 	fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
4161 	fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
4162 	fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
4163 	fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
4164 
4165 	for (i = 0; i < 2; i++)
4166 		for (j = 0; j < 2; j++) {
4167 			fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4168 			fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4169 			fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4170 			fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4171 		}
4172 
4173 	fiop->f_ticks = ifs->ifs_fr_ticks;
4174 	fiop->f_active = ifs->ifs_fr_active;
4175 	fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4176 	fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4177 
4178 	fiop->f_running = ifs->ifs_fr_running;
4179 	for (i = 0; i < IPL_LOGSIZE; i++) {
4180 		fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4181 		fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4182 	}
4183 #ifdef  IPFILTER_LOG
4184 	fiop->f_logging = 1;
4185 #else
4186 	fiop->f_logging = 0;
4187 #endif
4188 	fiop->f_defpass = ifs->ifs_fr_pass;
4189 	fiop->f_features = fr_features;
4190 	(void) strncpy(fiop->f_version, ipfilter_version,
4191 		       sizeof(fiop->f_version));
4192 }
4193 
4194 
4195 #ifdef	USE_INET6
4196 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4197 	ICMP6_ECHO_REPLY,	/* 0: ICMP_ECHOREPLY */
4198 	-1,			/* 1: UNUSED */
4199 	-1,			/* 2: UNUSED */
4200 	ICMP6_DST_UNREACH,	/* 3: ICMP_UNREACH */
4201 	-1,			/* 4: ICMP_SOURCEQUENCH */
4202 	ND_REDIRECT,		/* 5: ICMP_REDIRECT */
4203 	-1,			/* 6: UNUSED */
4204 	-1,			/* 7: UNUSED */
4205 	ICMP6_ECHO_REQUEST,	/* 8: ICMP_ECHO */
4206 	-1,			/* 9: UNUSED */
4207 	-1,			/* 10: UNUSED */
4208 	ICMP6_TIME_EXCEEDED,	/* 11: ICMP_TIMXCEED */
4209 	ICMP6_PARAM_PROB,	/* 12: ICMP_PARAMPROB */
4210 	-1,			/* 13: ICMP_TSTAMP */
4211 	-1,			/* 14: ICMP_TSTAMPREPLY */
4212 	-1,			/* 15: ICMP_IREQ */
4213 	-1,			/* 16: ICMP_IREQREPLY */
4214 	-1,			/* 17: ICMP_MASKREQ */
4215 	-1,			/* 18: ICMP_MASKREPLY */
4216 };
4217 
4218 
4219 int	icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4220 	ICMP6_DST_UNREACH_ADDR,		/* 0: ICMP_UNREACH_NET */
4221 	ICMP6_DST_UNREACH_ADDR,		/* 1: ICMP_UNREACH_HOST */
4222 	-1,				/* 2: ICMP_UNREACH_PROTOCOL */
4223 	ICMP6_DST_UNREACH_NOPORT,	/* 3: ICMP_UNREACH_PORT */
4224 	-1,				/* 4: ICMP_UNREACH_NEEDFRAG */
4225 	ICMP6_DST_UNREACH_NOTNEIGHBOR,	/* 5: ICMP_UNREACH_SRCFAIL */
4226 	ICMP6_DST_UNREACH_ADDR,		/* 6: ICMP_UNREACH_NET_UNKNOWN */
4227 	ICMP6_DST_UNREACH_ADDR,		/* 7: ICMP_UNREACH_HOST_UNKNOWN */
4228 	-1,				/* 8: ICMP_UNREACH_ISOLATED */
4229 	ICMP6_DST_UNREACH_ADMIN,	/* 9: ICMP_UNREACH_NET_PROHIB */
4230 	ICMP6_DST_UNREACH_ADMIN,	/* 10: ICMP_UNREACH_HOST_PROHIB */
4231 	-1,				/* 11: ICMP_UNREACH_TOSNET */
4232 	-1,				/* 12: ICMP_UNREACH_TOSHOST */
4233 	ICMP6_DST_UNREACH_ADMIN,	/* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4234 };
4235 int	icmpreplytype6[ICMP6_MAXTYPE + 1];
4236 #endif
4237 
4238 int	icmpreplytype4[ICMP_MAXTYPE + 1];
4239 
4240 
4241 /* ------------------------------------------------------------------------ */
4242 /* Function:    fr_matchicmpqueryreply                                      */
4243 /* Returns:     int - 1 if "icmp" is a valid reply to "ic" else 0.          */
4244 /* Parameters:  v(I)    - IP protocol version (4 or 6)                      */
4245 /*              ic(I)   - ICMP information                                  */
4246 /*              icmp(I) - ICMP packet header                                */
4247 /*              rev(I)  - direction (0 = forward/1 = reverse) of packet     */
4248 /*                                                                          */
4249 /* Check if the ICMP packet defined by the header pointed to by icmp is a   */
4250 /* reply to one as described by what's in ic.  If it is a match, return 1,  */
4251 /* else return 0 for no match.                                              */
4252 /* ------------------------------------------------------------------------ */
4253 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4254 int v;
4255 icmpinfo_t *ic;
4256 icmphdr_t *icmp;
4257 int rev;
4258 {
4259 	int ictype;
4260 
4261 	ictype = ic->ici_type;
4262 
4263 	if (v == 4) {
4264 		/*
4265 		 * If we matched its type on the way in, then when going out
4266 		 * it will still be the same type.
4267 		 */
4268 		if ((!rev && (icmp->icmp_type == ictype)) ||
4269 		    (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4270 			if (icmp->icmp_type != ICMP_ECHOREPLY)
4271 				return 1;
4272 			if (icmp->icmp_id == ic->ici_id)
4273 				return 1;
4274 		}
4275 	}
4276 #ifdef	USE_INET6
4277 	else if (v == 6) {
4278 		if ((!rev && (icmp->icmp_type == ictype)) ||
4279 		    (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4280 			if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4281 				return 1;
4282 			if (icmp->icmp_id == ic->ici_id)
4283 				return 1;
4284 		}
4285 	}
4286 #endif
4287 	return 0;
4288 }
4289 
4290 
4291 #ifdef	IPFILTER_LOOKUP
4292 /* ------------------------------------------------------------------------ */
4293 /* Function:    fr_resolvelookup                                            */
4294 /* Returns:     void * - NULL = failure, else success.                      */
4295 /* Parameters:  type(I)     - type of lookup these parameters are for.      */
4296 /*              number(I)   - table number to use when searching            */
4297 /*              funcptr(IO) - pointer to pointer for storing IP address     */
4298 /*			      searching function.			    */
4299 /*		ifs	    - ipf stack instance			    */
4300 /*                                                                          */
4301 /* Search for the "table" number passed in amongst those configured for     */
4302 /* that particular type.  If the type is recognised then the function to    */
4303 /* call to do the IP address search will be change, regardless of whether   */
4304 /* or not the "table" number exists.                                        */
4305 /* ------------------------------------------------------------------------ */
4306 static void *fr_resolvelookup(type, number, funcptr, ifs)
4307 u_int type, number;
4308 lookupfunc_t *funcptr;
4309 ipf_stack_t *ifs;
4310 {
4311 	char name[FR_GROUPLEN];
4312 	iphtable_t *iph;
4313 	ip_pool_t *ipo;
4314 	void *ptr;
4315 
4316 #if defined(SNPRINTF) && defined(_KERNEL)
4317 	(void) SNPRINTF(name, sizeof(name), "%u", number);
4318 #else
4319 	(void) sprintf(name, "%u", number);
4320 #endif
4321 
4322 	READ_ENTER(&ifs->ifs_ip_poolrw);
4323 
4324 	switch (type)
4325 	{
4326 	case IPLT_POOL :
4327 # if (defined(__osf__) && defined(_KERNEL))
4328 		ptr = NULL;
4329 		*funcptr = NULL;
4330 # else
4331 		ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4332 		ptr = ipo;
4333 		if (ipo != NULL) {
4334 			ATOMIC_INC32(ipo->ipo_ref);
4335 		}
4336 		*funcptr = ip_pool_search;
4337 # endif
4338 		break;
4339 	case IPLT_HASH :
4340 		iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4341 		ptr = iph;
4342 		if (iph != NULL) {
4343 			ATOMIC_INC32(iph->iph_ref);
4344 		}
4345 		*funcptr = fr_iphmfindip;
4346 		break;
4347 	default:
4348 		ptr = NULL;
4349 		*funcptr = NULL;
4350 		break;
4351 	}
4352 	RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4353 
4354 	return ptr;
4355 }
4356 #endif
4357 
4358 
4359 /* ------------------------------------------------------------------------ */
4360 /* Function:    frrequest                                                   */
4361 /* Returns:     int - 0 == success, > 0 == errno value                      */
4362 /* Parameters:  unit(I)     - device for which this is for                  */
4363 /*              req(I)      - ioctl command (SIOC*)                         */
4364 /*              data(I)     - pointr to ioctl data                          */
4365 /*              set(I)      - 1 or 0 (filter set)                           */
4366 /*              makecopy(I) - flag indicating whether data points to a rule */
4367 /*                            in kernel space & hence doesn't need copying. */
4368 /*                                                                          */
4369 /* This function handles all the requests which operate on the list of      */
4370 /* filter rules.  This includes adding, deleting, insertion.  It is also    */
4371 /* responsible for creating groups when a "head" rule is loaded.  Interface */
4372 /* names are resolved here and other sanity checks are made on the content  */
4373 /* of the rule structure being loaded.  If a rule has user defined timeouts */
4374 /* then make sure they are created and initialised before exiting.          */
4375 /* ------------------------------------------------------------------------ */
4376 int frrequest(unit, req, data, set, makecopy, ifs)
4377 int unit;
4378 ioctlcmd_t req;
4379 int set, makecopy;
4380 caddr_t data;
4381 ipf_stack_t *ifs;
4382 {
4383 	frentry_t frd, *fp, *f, **fprev, **ftail;
4384 	int error = 0, in, v;
4385 	void *ptr, *uptr;
4386 	u_int *p, *pp;
4387 	frgroup_t *fg;
4388 	char *group;
4389 
4390 	fg = NULL;
4391 	fp = &frd;
4392 	if (makecopy != 0) {
4393 		error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4394 		if (error)
4395 			return EFAULT;
4396 		if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4397 			return EINVAL;
4398 		fp->fr_ref = 0;
4399 		fp->fr_flags |= FR_COPIED;
4400 	} else {
4401 		fp = (frentry_t *)data;
4402 		if ((fp->fr_type & FR_T_BUILTIN) == 0)
4403 			return EINVAL;
4404 		fp->fr_flags &= ~FR_COPIED;
4405 	}
4406 
4407 	if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4408 	    ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4409 		return EINVAL;
4410 
4411 	v = fp->fr_v;
4412 	uptr = fp->fr_data;
4413 
4414 	/*
4415 	 * Only filter rules for IPv4 or IPv6 are accepted.
4416 	 */
4417 	if (v == 4)
4418 		/*EMPTY*/;
4419 #ifdef	USE_INET6
4420 	else if (v == 6)
4421 		/*EMPTY*/;
4422 #endif
4423 	else {
4424 		return EINVAL;
4425 	}
4426 
4427 	/*
4428 	 * If the rule is being loaded from user space, i.e. we had to copy it
4429 	 * into kernel space, then do not trust the function pointer in the
4430 	 * rule.
4431 	 */
4432 	if ((makecopy == 1) && (fp->fr_func != NULL)) {
4433 		if (fr_findfunc(fp->fr_func) == NULL)
4434 			return ESRCH;
4435 		error = fr_funcinit(fp, ifs);
4436 		if (error != 0)
4437 			return error;
4438 	}
4439 
4440 	ptr = NULL;
4441 	/*
4442 	 * Check that the group number does exist and that its use (in/out)
4443 	 * matches what the rule is.
4444 	 */
4445 	if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4446 		*fp->fr_grhead = '\0';
4447 	group = fp->fr_group;
4448 	if (!strncmp(group, "0", FR_GROUPLEN))
4449 		*group = '\0';
4450 
4451 	if (FR_ISACCOUNT(fp->fr_flags))
4452 		unit = IPL_LOGCOUNT;
4453 
4454 	if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4455 		fg = fr_findgroup(group, unit, set, NULL, ifs);
4456 		if (fg == NULL)
4457 			return ESRCH;
4458 		if (fg->fg_flags == 0)
4459 			fg->fg_flags = fp->fr_flags & FR_INOUT;
4460 		else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4461 			return ESRCH;
4462 	}
4463 
4464 	in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4465 
4466 	/*
4467 	 * Work out which rule list this change is being applied to.
4468 	 */
4469 	ftail = NULL;
4470 	fprev = NULL;
4471 	if (unit == IPL_LOGAUTH)
4472 		fprev = &ifs->ifs_ipauth;
4473 	else if (v == 4) {
4474 		if (FR_ISACCOUNT(fp->fr_flags))
4475 			fprev = &ifs->ifs_ipacct[in][set];
4476 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4477 			fprev = &ifs->ifs_ipfilter[in][set];
4478 	} else if (v == 6) {
4479 		if (FR_ISACCOUNT(fp->fr_flags))
4480 			fprev = &ifs->ifs_ipacct6[in][set];
4481 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4482 			fprev = &ifs->ifs_ipfilter6[in][set];
4483 	}
4484 	if (fprev == NULL)
4485 		return ESRCH;
4486 
4487 	if (*group != '\0') {
4488 	    if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4489 			return ESRCH;
4490 		fprev = &fg->fg_start;
4491 	}
4492 
4493 	ftail = fprev;
4494 	for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4495 		if (fp->fr_collect <= f->fr_collect) {
4496 			ftail = fprev;
4497 			f = NULL;
4498 			break;
4499 		}
4500 		fprev = ftail;
4501 	}
4502 
4503 	/*
4504 	 * Copy in extra data for the rule.
4505 	 */
4506 	if (fp->fr_dsize != 0) {
4507 		if (makecopy != 0) {
4508 			KMALLOCS(ptr, void *, fp->fr_dsize);
4509 			if (!ptr)
4510 				return ENOMEM;
4511 			error = COPYIN(uptr, ptr, fp->fr_dsize);
4512 		} else {
4513 			ptr = uptr;
4514 			error = 0;
4515 		}
4516 		if (error != 0) {
4517 			KFREES(ptr, fp->fr_dsize);
4518 			return EFAULT;
4519 		}
4520 		fp->fr_data = ptr;
4521 	} else
4522 		fp->fr_data = NULL;
4523 
4524 	/*
4525 	 * Perform per-rule type sanity checks of their members.
4526 	 */
4527 	switch (fp->fr_type & ~FR_T_BUILTIN)
4528 	{
4529 #if defined(IPFILTER_BPF)
4530 	case FR_T_BPFOPC :
4531 		if (fp->fr_dsize == 0)
4532 			return EINVAL;
4533 		if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4534 			if (makecopy && fp->fr_data != NULL) {
4535 				KFREES(fp->fr_data, fp->fr_dsize);
4536 			}
4537 			return EINVAL;
4538 		}
4539 		break;
4540 #endif
4541 	case FR_T_IPF :
4542 		if (fp->fr_dsize != sizeof(fripf_t)) {
4543 			if (makecopy && fp->fr_data != NULL) {
4544 				KFREES(fp->fr_data, fp->fr_dsize);
4545 			}
4546 			return EINVAL;
4547 		}
4548 
4549 		/*
4550 		 * Allowing a rule with both "keep state" and "with oow" is
4551 		 * pointless because adding a state entry to the table will
4552 		 * fail with the out of window (oow) flag set.
4553 		 */
4554 		if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW)) {
4555 			if (makecopy && fp->fr_data != NULL) {
4556 				KFREES(fp->fr_data, fp->fr_dsize);
4557 			}
4558 			return EINVAL;
4559 		}
4560 
4561 		switch (fp->fr_satype)
4562 		{
4563 		case FRI_BROADCAST :
4564 		case FRI_DYNAMIC :
4565 		case FRI_NETWORK :
4566 		case FRI_NETMASKED :
4567 		case FRI_PEERADDR :
4568 			if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4569 				if (makecopy && fp->fr_data != NULL) {
4570 					KFREES(fp->fr_data, fp->fr_dsize);
4571 				}
4572 				return EINVAL;
4573 			}
4574 			break;
4575 #ifdef	IPFILTER_LOOKUP
4576 		case FRI_LOOKUP :
4577 			fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4578 							 fp->fr_srcnum,
4579 							 &fp->fr_srcfunc, ifs);
4580 			break;
4581 #endif
4582 		default :
4583 			break;
4584 		}
4585 
4586 		switch (fp->fr_datype)
4587 		{
4588 		case FRI_BROADCAST :
4589 		case FRI_DYNAMIC :
4590 		case FRI_NETWORK :
4591 		case FRI_NETMASKED :
4592 		case FRI_PEERADDR :
4593 			if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4594 				if (makecopy && fp->fr_data != NULL) {
4595 					KFREES(fp->fr_data, fp->fr_dsize);
4596 				}
4597 				return EINVAL;
4598 			}
4599 			break;
4600 #ifdef	IPFILTER_LOOKUP
4601 		case FRI_LOOKUP :
4602 			fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4603 							 fp->fr_dstnum,
4604 							 &fp->fr_dstfunc, ifs);
4605 			break;
4606 #endif
4607 		default :
4608 			break;
4609 		}
4610 		break;
4611 	case FR_T_NONE :
4612 		break;
4613 	case FR_T_CALLFUNC :
4614 		break;
4615 	case FR_T_COMPIPF :
4616 		break;
4617 	default :
4618 		if (makecopy && fp->fr_data != NULL) {
4619 			KFREES(fp->fr_data, fp->fr_dsize);
4620 		}
4621 		return EINVAL;
4622 	}
4623 
4624 	/*
4625 	 * Lookup all the interface names that are part of the rule.
4626 	 */
4627 	frsynclist(0, 0, NULL, NULL, fp, ifs);
4628 	fp->fr_statecnt = 0;
4629 
4630 	/*
4631 	 * Look for an existing matching filter rule, but don't include the
4632 	 * next or interface pointer in the comparison (fr_next, fr_ifa).
4633 	 * This elminates rules which are indentical being loaded.  Checksum
4634 	 * the constant part of the filter rule to make comparisons quicker
4635 	 * (this meaning no pointers are included).
4636 	 */
4637 	for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4638 	     p < pp; p++)
4639 		fp->fr_cksum += *p;
4640 	pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4641 	for (p = (u_int *)fp->fr_data; p < pp; p++)
4642 		fp->fr_cksum += *p;
4643 
4644 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
4645 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
4646 
4647 	for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4648 		if ((fp->fr_cksum != f->fr_cksum) ||
4649 		    (f->fr_dsize != fp->fr_dsize))
4650 			continue;
4651 		if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4652 			continue;
4653 		if ((!ptr && !f->fr_data) ||
4654 		    (ptr && f->fr_data &&
4655 		     !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4656 			break;
4657 	}
4658 
4659 	/*
4660 	 * If zero'ing statistics, copy current to caller and zero.
4661 	 */
4662 	if (req == (ioctlcmd_t)SIOCZRLST) {
4663 		if (f == NULL)
4664 			error = ESRCH;
4665 		else {
4666 			/*
4667 			 * Copy and reduce lock because of impending copyout.
4668 			 * Well we should, but if we do then the atomicity of
4669 			 * this call and the correctness of fr_hits and
4670 			 * fr_bytes cannot be guaranteed.  As it is, this code
4671 			 * only resets them to 0 if they are successfully
4672 			 * copied out into user space.
4673 			 */
4674 			bcopy((char *)f, (char *)fp, sizeof(*f));
4675 
4676 			/*
4677 			 * When we copy this rule back out, set the data
4678 			 * pointer to be what it was in user space.
4679 			 */
4680 			fp->fr_data = uptr;
4681 			error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4682 
4683 			if (error == 0) {
4684 				if ((f->fr_dsize != 0) && (uptr != NULL))
4685 					error = COPYOUT(f->fr_data, uptr,
4686 							f->fr_dsize);
4687 				if (error == 0) {
4688 					f->fr_hits = 0;
4689 					f->fr_bytes = 0;
4690 				}
4691 			}
4692 		}
4693 
4694 		if ((ptr != NULL) && (makecopy != 0)) {
4695 			KFREES(ptr, fp->fr_dsize);
4696 		}
4697 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4698 		return error;
4699 	}
4700 
4701 	if (!f) {
4702 		/*
4703 		 * At the end of this, ftail must point to the place where the
4704 		 * new rule is to be saved/inserted/added.
4705 		 * For SIOCAD*FR, this should be the last rule in the group of
4706 		 * rules that have equal fr_collect fields.
4707 		 * For SIOCIN*FR, ...
4708 		 */
4709 		if (req == (ioctlcmd_t)SIOCADAFR ||
4710 		    req == (ioctlcmd_t)SIOCADIFR) {
4711 
4712 			for (ftail = fprev; (f = *ftail) != NULL; ) {
4713 				if (f->fr_collect > fp->fr_collect)
4714 					break;
4715 				ftail = &f->fr_next;
4716 			}
4717 			f = NULL;
4718 			ptr = NULL;
4719 			error = 0;
4720 		} else if (req == (ioctlcmd_t)SIOCINAFR ||
4721 			   req == (ioctlcmd_t)SIOCINIFR) {
4722 			while ((f = *fprev) != NULL) {
4723 				if (f->fr_collect >= fp->fr_collect)
4724 					break;
4725 				fprev = &f->fr_next;
4726 			}
4727 			ftail = fprev;
4728 			if (fp->fr_hits != 0) {
4729 				while (fp->fr_hits && (f = *ftail)) {
4730 					if (f->fr_collect != fp->fr_collect)
4731 						break;
4732 					fprev = ftail;
4733 					ftail = &f->fr_next;
4734 					fp->fr_hits--;
4735 				}
4736 			}
4737 			f = NULL;
4738 			ptr = NULL;
4739 			error = 0;
4740 		}
4741 	}
4742 
4743 	/*
4744 	 * Request to remove a rule.
4745 	 */
4746 	if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4747 		if (!f)
4748 			error = ESRCH;
4749 		else {
4750 			/*
4751 			 * Do not allow activity from user space to interfere
4752 			 * with rules not loaded that way.
4753 			 */
4754 			if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4755 				error = EPERM;
4756 				goto done;
4757 			}
4758 
4759 			/*
4760 			 * Return EBUSY if the rule is being reference by
4761 			 * something else (eg state information.
4762 			 */
4763 			if (f->fr_ref > 1) {
4764 				error = EBUSY;
4765 				goto done;
4766 			}
4767 #ifdef	IPFILTER_SCAN
4768 			if (f->fr_isctag[0] != '\0' &&
4769 			    (f->fr_isc != (struct ipscan *)-1))
4770 				ipsc_detachfr(f);
4771 #endif
4772 			if (unit == IPL_LOGAUTH) {
4773 				error = fr_preauthcmd(req, f, ftail, ifs);
4774 				goto done;
4775 			}
4776 			if (*f->fr_grhead != '\0')
4777 				fr_delgroup(f->fr_grhead, unit, set, ifs);
4778 			fr_fixskip(ftail, f, -1);
4779 			*ftail = f->fr_next;
4780 			f->fr_next = NULL;
4781 			(void)fr_derefrule(&f, ifs);
4782 		}
4783 	} else {
4784 		/*
4785 		 * Not removing, so we must be adding/inserting a rule.
4786 		 */
4787 		if (f)
4788 			error = EEXIST;
4789 		else {
4790 			if (unit == IPL_LOGAUTH) {
4791 				error = fr_preauthcmd(req, fp, ftail, ifs);
4792 				goto done;
4793 			}
4794 			if (makecopy) {
4795 				KMALLOC(f, frentry_t *);
4796 			} else
4797 				f = fp;
4798 			if (f != NULL) {
4799 				if (fp != f)
4800 					bcopy((char *)fp, (char *)f,
4801 					      sizeof(*f));
4802 				MUTEX_NUKE(&f->fr_lock);
4803 				MUTEX_INIT(&f->fr_lock, "filter rule lock");
4804 #ifdef	IPFILTER_SCAN
4805 				if (f->fr_isctag[0] != '\0' &&
4806 				    ipsc_attachfr(f))
4807 					f->fr_isc = (struct ipscan *)-1;
4808 #endif
4809 				f->fr_hits = 0;
4810 				if (makecopy != 0)
4811 					f->fr_ref = 1;
4812 				f->fr_next = *ftail;
4813 				*ftail = f;
4814 				if (req == (ioctlcmd_t)SIOCINIFR ||
4815 				    req == (ioctlcmd_t)SIOCINAFR)
4816 					fr_fixskip(ftail, f, 1);
4817 				f->fr_grp = NULL;
4818 				group = f->fr_grhead;
4819 				if (*group != '\0') {
4820 					fg = fr_addgroup(group, f, f->fr_flags,
4821 							 unit, set, ifs);
4822 					if (fg != NULL)
4823 						f->fr_grp = &fg->fg_start;
4824 				}
4825 			} else
4826 				error = ENOMEM;
4827 		}
4828 	}
4829 done:
4830 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4831 	if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4832 		KFREES(ptr, fp->fr_dsize);
4833 	}
4834 	return (error);
4835 }
4836 
4837 
4838 /* ------------------------------------------------------------------------ */
4839 /* Function:    fr_funcinit                                                 */
4840 /* Returns:     int - 0 == success, else ESRCH: cannot resolve rule details */
4841 /* Parameters:  fr(I) - pointer to filter rule                              */
4842 /*                                                                          */
4843 /* If a rule is a call rule, then check if the function it points to needs  */
4844 /* an init function to be called now the rule has been loaded.              */
4845 /* ------------------------------------------------------------------------ */
4846 static int fr_funcinit(fr, ifs)
4847 frentry_t *fr;
4848 ipf_stack_t *ifs;
4849 {
4850 	ipfunc_resolve_t *ft;
4851 	int err;
4852 
4853 	err = ESRCH;
4854 
4855 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4856 		if (ft->ipfu_addr == fr->fr_func) {
4857 			err = 0;
4858 			if (ft->ipfu_init != NULL)
4859 				err = (*ft->ipfu_init)(fr, ifs);
4860 			break;
4861 		}
4862 	return err;
4863 }
4864 
4865 
4866 /* ------------------------------------------------------------------------ */
4867 /* Function:    fr_findfunc                                                 */
4868 /* Returns:     ipfunc_t - pointer to function if found, else NULL          */
4869 /* Parameters:  funcptr(I) - function pointer to lookup                     */
4870 /*                                                                          */
4871 /* Look for a function in the table of known functions.                     */
4872 /* ------------------------------------------------------------------------ */
4873 static ipfunc_t fr_findfunc(funcptr)
4874 ipfunc_t funcptr;
4875 {
4876 	ipfunc_resolve_t *ft;
4877 
4878 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4879 		if (ft->ipfu_addr == funcptr)
4880 			return funcptr;
4881 	return NULL;
4882 }
4883 
4884 
4885 /* ------------------------------------------------------------------------ */
4886 /* Function:    fr_resolvefunc                                              */
4887 /* Returns:     int - 0 == success, else error                              */
4888 /* Parameters:  data(IO) - ioctl data pointer to ipfunc_resolve_t struct    */
4889 /*                                                                          */
4890 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4891 /* This will either be the function name (if the pointer is set) or the     */
4892 /* function pointer if the name is set.  When found, fill in the other one  */
4893 /* so that the entire, complete, structure can be copied back to user space.*/
4894 /* ------------------------------------------------------------------------ */
4895 int fr_resolvefunc(data)
4896 void *data;
4897 {
4898 	ipfunc_resolve_t res, *ft;
4899 	int err;
4900 
4901 	err = BCOPYIN(data, &res, sizeof(res));
4902 	if (err != 0)
4903 		return EFAULT;
4904 
4905 	if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4906 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4907 			if (strncmp(res.ipfu_name, ft->ipfu_name,
4908 				    sizeof(res.ipfu_name)) == 0) {
4909 				res.ipfu_addr = ft->ipfu_addr;
4910 				res.ipfu_init = ft->ipfu_init;
4911 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4912 					return EFAULT;
4913 				return 0;
4914 			}
4915 	}
4916 	if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4917 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4918 			if (ft->ipfu_addr == res.ipfu_addr) {
4919 				(void) strncpy(res.ipfu_name, ft->ipfu_name,
4920 					       sizeof(res.ipfu_name));
4921 				res.ipfu_init = ft->ipfu_init;
4922 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4923 					return EFAULT;
4924 				return 0;
4925 			}
4926 	}
4927 	return ESRCH;
4928 }
4929 
4930 
4931 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4932     (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4933     (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4934     (defined(__OpenBSD__) && (OpenBSD < 200006))
4935 /*
4936  * From: NetBSD
4937  * ppsratecheck(): packets (or events) per second limitation.
4938  */
4939 int
4940 ppsratecheck(lasttime, curpps, maxpps)
4941 	struct timeval *lasttime;
4942 	int *curpps;
4943 	int maxpps;	/* maximum pps allowed */
4944 {
4945 	struct timeval tv, delta;
4946 	int rv;
4947 
4948 	GETKTIME(&tv);
4949 
4950 	delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4951 	delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4952 	if (delta.tv_usec < 0) {
4953 		delta.tv_sec--;
4954 		delta.tv_usec += 1000000;
4955 	}
4956 
4957 	/*
4958 	 * check for 0,0 is so that the message will be seen at least once.
4959 	 * if more than one second have passed since the last update of
4960 	 * lasttime, reset the counter.
4961 	 *
4962 	 * we do increment *curpps even in *curpps < maxpps case, as some may
4963 	 * try to use *curpps for stat purposes as well.
4964 	 */
4965 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4966 	    delta.tv_sec >= 1) {
4967 		*lasttime = tv;
4968 		*curpps = 0;
4969 		rv = 1;
4970 	} else if (maxpps < 0)
4971 		rv = 1;
4972 	else if (*curpps < maxpps)
4973 		rv = 1;
4974 	else
4975 		rv = 0;
4976 	*curpps = *curpps + 1;
4977 
4978 	return (rv);
4979 }
4980 #endif
4981 
4982 
4983 /* ------------------------------------------------------------------------ */
4984 /* Function:    fr_derefrule                                                */
4985 /* Returns:     int   - 0 == rule freed up, else rule not freed             */
4986 /* Parameters:  fr(I) - pointer to filter rule                              */
4987 /*                                                                          */
4988 /* Decrement the reference counter to a rule by one.  If it reaches zero,   */
4989 /* free it and any associated storage space being used by it.               */
4990 /* ------------------------------------------------------------------------ */
4991 int fr_derefrule(frp, ifs)
4992 frentry_t **frp;
4993 ipf_stack_t *ifs;
4994 {
4995 	frentry_t *fr;
4996 
4997 	fr = *frp;
4998 
4999 	MUTEX_ENTER(&fr->fr_lock);
5000 	fr->fr_ref--;
5001 	if (fr->fr_ref == 0) {
5002 		MUTEX_EXIT(&fr->fr_lock);
5003 		MUTEX_DESTROY(&fr->fr_lock);
5004 
5005 #ifdef IPFILTER_LOOKUP
5006 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
5007 		    ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
5008 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
5009 		    ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
5010 #endif
5011 
5012 		if (fr->fr_dsize) {
5013 			KFREES(fr->fr_data, fr->fr_dsize);
5014 		}
5015 		if ((fr->fr_flags & FR_COPIED) != 0) {
5016 			KFREE(fr);
5017 			return 0;
5018 		}
5019 		return 1;
5020 	} else {
5021 		MUTEX_EXIT(&fr->fr_lock);
5022 	}
5023 	*frp = NULL;
5024 	return -1;
5025 }
5026 
5027 
5028 #ifdef	IPFILTER_LOOKUP
5029 /* ------------------------------------------------------------------------ */
5030 /* Function:    fr_grpmapinit                                               */
5031 /* Returns:     int - 0 == success, else ESRCH because table entry not found*/
5032 /* Parameters:  fr(I) - pointer to rule to find hash table for              */
5033 /*                                                                          */
5034 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr.  */
5035 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap.                   */
5036 /* ------------------------------------------------------------------------ */
5037 static int fr_grpmapinit(fr, ifs)
5038 frentry_t *fr;
5039 ipf_stack_t *ifs;
5040 {
5041 	char name[FR_GROUPLEN];
5042 	iphtable_t *iph;
5043 
5044 #if defined(SNPRINTF) && defined(_KERNEL)
5045 	(void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
5046 #else
5047 	(void) sprintf(name, "%d", fr->fr_arg);
5048 #endif
5049 	iph = fr_findhtable(IPL_LOGIPF, name, ifs);
5050 	if (iph == NULL)
5051 		return ESRCH;
5052 	if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
5053 		return ESRCH;
5054 	fr->fr_ptr = iph;
5055 	return 0;
5056 }
5057 
5058 
5059 /* ------------------------------------------------------------------------ */
5060 /* Function:    fr_srcgrpmap                                                */
5061 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5062 /* Parameters:  fin(I)    - pointer to packet information                   */
5063 /*              passp(IO) - pointer to current/new filter decision (unused) */
5064 /*                                                                          */
5065 /* Look for a rule group head in a hash table, using the source address as  */
5066 /* the key, and descend into that group and continue matching rules against */
5067 /* the packet.                                                              */
5068 /* ------------------------------------------------------------------------ */
5069 frentry_t *fr_srcgrpmap(fin, passp)
5070 fr_info_t *fin;
5071 u_32_t *passp;
5072 {
5073 	frgroup_t *fg;
5074 	void *rval;
5075 	ipf_stack_t *ifs = fin->fin_ifs;
5076 
5077 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
5078 	if (rval == NULL)
5079 		return NULL;
5080 
5081 	fg = rval;
5082 	fin->fin_fr = fg->fg_start;
5083 	(void) fr_scanlist(fin, *passp);
5084 	return fin->fin_fr;
5085 }
5086 
5087 
5088 /* ------------------------------------------------------------------------ */
5089 /* Function:    fr_dstgrpmap                                                */
5090 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5091 /* Parameters:  fin(I)    - pointer to packet information                   */
5092 /*              passp(IO) - pointer to current/new filter decision (unused) */
5093 /*                                                                          */
5094 /* Look for a rule group head in a hash table, using the destination        */
5095 /* address as the key, and descend into that group and continue matching    */
5096 /* rules against  the packet.                                               */
5097 /* ------------------------------------------------------------------------ */
5098 frentry_t *fr_dstgrpmap(fin, passp)
5099 fr_info_t *fin;
5100 u_32_t *passp;
5101 {
5102 	frgroup_t *fg;
5103 	void *rval;
5104 	ipf_stack_t *ifs = fin->fin_ifs;
5105 
5106 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
5107 	if (rval == NULL)
5108 		return NULL;
5109 
5110 	fg = rval;
5111 	fin->fin_fr = fg->fg_start;
5112 	(void) fr_scanlist(fin, *passp);
5113 	return fin->fin_fr;
5114 }
5115 #endif /* IPFILTER_LOOKUP */
5116 
5117 /*
5118  * Queue functions
5119  * ===============
5120  * These functions manage objects on queues for efficient timeouts.  There are
5121  * a number of system defined queues as well as user defined timeouts.  It is
5122  * expected that a lock is held in the domain in which the queue belongs
5123  * (i.e. either state or NAT) when calling any of these functions that prevents
5124  * fr_freetimeoutqueue() from being called at the same time as any other.
5125  */
5126 
5127 
5128 /* ------------------------------------------------------------------------ */
5129 /* Function:    fr_addtimeoutqueue                                          */
5130 /* Returns:     struct ifqtq * - NULL if malloc fails, else pointer to      */
5131 /*                               timeout queue with given interval.         */
5132 /* Parameters:  parent(I)  - pointer to pointer to parent node of this list */
5133 /*                           of interface queues.                           */
5134 /*              seconds(I) - timeout value in seconds for this queue.       */
5135 /*                                                                          */
5136 /* This routine first looks for a timeout queue that matches the interval   */
5137 /* being requested.  If it finds one, increments the reference counter and  */
5138 /* returns a pointer to it.  If none are found, it allocates a new one and  */
5139 /* inserts it at the top of the list.                                       */
5140 /*                                                                          */
5141 /* Locking.                                                                 */
5142 /* It is assumed that the caller of this function has an appropriate lock   */
5143 /* held (exclusively) in the domain that encompases 'parent'.               */
5144 /* ------------------------------------------------------------------------ */
5145 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
5146 ipftq_t **parent;
5147 u_int seconds;
5148 ipf_stack_t *ifs;
5149 {
5150 	ipftq_t *ifq;
5151 	u_int period;
5152 
5153 	period = seconds * IPF_HZ_DIVIDE;
5154 
5155 	MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
5156 	for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
5157 		if (ifq->ifq_ttl == period) {
5158 			/*
5159 			 * Reset the delete flag, if set, so the structure
5160 			 * gets reused rather than freed and reallocated.
5161 			 */
5162 			MUTEX_ENTER(&ifq->ifq_lock);
5163 			ifq->ifq_flags &= ~IFQF_DELETE;
5164 			ifq->ifq_ref++;
5165 			MUTEX_EXIT(&ifq->ifq_lock);
5166 			MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5167 
5168 			return ifq;
5169 		}
5170 	}
5171 
5172 	KMALLOC(ifq, ipftq_t *);
5173 	if (ifq != NULL) {
5174 		ifq->ifq_ttl = period;
5175 		ifq->ifq_head = NULL;
5176 		ifq->ifq_tail = &ifq->ifq_head;
5177 		ifq->ifq_next = *parent;
5178 		ifq->ifq_pnext = parent;
5179 		ifq->ifq_ref = 1;
5180 		ifq->ifq_flags = IFQF_USER;
5181 		*parent = ifq;
5182 		ifs->ifs_fr_userifqs++;
5183 		MUTEX_NUKE(&ifq->ifq_lock);
5184 		MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5185 	}
5186 	MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5187 	return ifq;
5188 }
5189 
5190 
5191 /* ------------------------------------------------------------------------ */
5192 /* Function:    fr_deletetimeoutqueue                                       */
5193 /* Returns:     int    - new reference count value of the timeout queue     */
5194 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5195 /* Locks:       ifq->ifq_lock                                               */
5196 /*                                                                          */
5197 /* This routine must be called when we're discarding a pointer to a timeout */
5198 /* queue object, taking care of the reference counter.                      */
5199 /*                                                                          */
5200 /* Now that this just sets a DELETE flag, it requires the expire code to    */
5201 /* check the list of user defined timeout queues and call the free function */
5202 /* below (currently commented out) to stop memory leaking.  It is done this */
5203 /* way because the locking may not be sufficient to safely do a free when   */
5204 /* this function is called.                                                 */
5205 /* ------------------------------------------------------------------------ */
5206 int fr_deletetimeoutqueue(ifq)
5207 ipftq_t *ifq;
5208 {
5209 
5210 	ifq->ifq_ref--;
5211 	if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5212 		ifq->ifq_flags |= IFQF_DELETE;
5213 	}
5214 
5215 	return ifq->ifq_ref;
5216 }
5217 
5218 
5219 /* ------------------------------------------------------------------------ */
5220 /* Function:    fr_freetimeoutqueue                                         */
5221 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5222 /* Returns:     Nil                                                         */
5223 /*                                                                          */
5224 /* Locking:                                                                 */
5225 /* It is assumed that the caller of this function has an appropriate lock   */
5226 /* held (exclusively) in the domain that encompases the callers "domain".   */
5227 /* The ifq_lock for this structure should not be held.                      */
5228 /*                                                                          */
5229 /* Remove a user definde timeout queue from the list of queues it is in and */
5230 /* tidy up after this is done.                                              */
5231 /* ------------------------------------------------------------------------ */
5232 void fr_freetimeoutqueue(ifq, ifs)
5233 ipftq_t *ifq;
5234 ipf_stack_t *ifs;
5235 {
5236 
5237 
5238 	if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5239 	    ((ifq->ifq_flags & IFQF_USER) == 0)) {
5240 		printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5241 		       (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5242 		       ifq->ifq_ref);
5243 		return;
5244 	}
5245 
5246 	/*
5247 	 * Remove from its position in the list.
5248 	 */
5249 	*ifq->ifq_pnext = ifq->ifq_next;
5250 	if (ifq->ifq_next != NULL)
5251 		ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5252 
5253 	MUTEX_DESTROY(&ifq->ifq_lock);
5254 	ifs->ifs_fr_userifqs--;
5255 	KFREE(ifq);
5256 }
5257 
5258 
5259 /* ------------------------------------------------------------------------ */
5260 /* Function:    fr_deletequeueentry                                         */
5261 /* Returns:     Nil                                                         */
5262 /* Parameters:  tqe(I) - timeout queue entry to delete                      */
5263 /*              ifq(I) - timeout queue to remove entry from                 */
5264 /*                                                                          */
5265 /* Remove a tail queue entry from its queue and make it an orphan.          */
5266 /* fr_deletetimeoutqueue is called to make sure the reference count on the  */
5267 /* queue is correct.  We can't, however, call fr_freetimeoutqueue because   */
5268 /* the correct lock(s) may not be held that would make it safe to do so.    */
5269 /* ------------------------------------------------------------------------ */
5270 void fr_deletequeueentry(tqe)
5271 ipftqent_t *tqe;
5272 {
5273 	ipftq_t *ifq;
5274 
5275 	ifq = tqe->tqe_ifq;
5276 	if (ifq == NULL)
5277 		return;
5278 
5279 	MUTEX_ENTER(&ifq->ifq_lock);
5280 
5281 	if (tqe->tqe_pnext != NULL) {
5282 		*tqe->tqe_pnext = tqe->tqe_next;
5283 		if (tqe->tqe_next != NULL)
5284 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5285 		else    /* we must be the tail anyway */
5286 			ifq->ifq_tail = tqe->tqe_pnext;
5287 
5288 		tqe->tqe_pnext = NULL;
5289 		tqe->tqe_ifq = NULL;
5290 	}
5291 
5292 	(void) fr_deletetimeoutqueue(ifq);
5293 
5294 	MUTEX_EXIT(&ifq->ifq_lock);
5295 }
5296 
5297 
5298 /* ------------------------------------------------------------------------ */
5299 /* Function:    fr_queuefront                                               */
5300 /* Returns:     Nil                                                         */
5301 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5302 /*                                                                          */
5303 /* Move a queue entry to the front of the queue, if it isn't already there. */
5304 /* ------------------------------------------------------------------------ */
5305 void fr_queuefront(tqe)
5306 ipftqent_t *tqe;
5307 {
5308 	ipftq_t *ifq;
5309 
5310 	ifq = tqe->tqe_ifq;
5311 	if (ifq == NULL)
5312 		return;
5313 
5314 	MUTEX_ENTER(&ifq->ifq_lock);
5315 	if (ifq->ifq_head != tqe) {
5316 		*tqe->tqe_pnext = tqe->tqe_next;
5317 		if (tqe->tqe_next)
5318 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5319 		else
5320 			ifq->ifq_tail = tqe->tqe_pnext;
5321 
5322 		tqe->tqe_next = ifq->ifq_head;
5323 		ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5324 		ifq->ifq_head = tqe;
5325 		tqe->tqe_pnext = &ifq->ifq_head;
5326 	}
5327 	MUTEX_EXIT(&ifq->ifq_lock);
5328 }
5329 
5330 
5331 /* ------------------------------------------------------------------------ */
5332 /* Function:    fr_queueback                                                */
5333 /* Returns:     Nil                                                         */
5334 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5335 /*                                                                          */
5336 /* Move a queue entry to the back of the queue, if it isn't already there.  */
5337 /* ------------------------------------------------------------------------ */
5338 void fr_queueback(tqe, ifs)
5339 ipftqent_t *tqe;
5340 ipf_stack_t *ifs;
5341 {
5342 	ipftq_t *ifq;
5343 
5344 	ifq = tqe->tqe_ifq;
5345 	if (ifq == NULL)
5346 		return;
5347 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5348 
5349 	MUTEX_ENTER(&ifq->ifq_lock);
5350 	if (tqe->tqe_next == NULL) {		/* at the end already ? */
5351 		MUTEX_EXIT(&ifq->ifq_lock);
5352 		return;
5353 	}
5354 
5355 	/*
5356 	 * Remove from list
5357 	 */
5358 	*tqe->tqe_pnext = tqe->tqe_next;
5359 	tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5360 
5361 	/*
5362 	 * Make it the last entry.
5363 	 */
5364 	tqe->tqe_next = NULL;
5365 	tqe->tqe_pnext = ifq->ifq_tail;
5366 	*ifq->ifq_tail = tqe;
5367 	ifq->ifq_tail = &tqe->tqe_next;
5368 	MUTEX_EXIT(&ifq->ifq_lock);
5369 }
5370 
5371 
5372 /* ------------------------------------------------------------------------ */
5373 /* Function:    fr_queueappend                                              */
5374 /* Returns:     Nil                                                         */
5375 /* Parameters:  tqe(I)    - pointer to timeout queue entry                  */
5376 /*              ifq(I)    - pointer to timeout queue                        */
5377 /*              parent(I) - owing object pointer                            */
5378 /*                                                                          */
5379 /* Add a new item to this queue and put it on the very end.                 */
5380 /* ------------------------------------------------------------------------ */
5381 void fr_queueappend(tqe, ifq, parent, ifs)
5382 ipftqent_t *tqe;
5383 ipftq_t *ifq;
5384 void *parent;
5385 ipf_stack_t *ifs;
5386 {
5387 
5388 	MUTEX_ENTER(&ifq->ifq_lock);
5389 	tqe->tqe_parent = parent;
5390 	tqe->tqe_pnext = ifq->ifq_tail;
5391 	*ifq->ifq_tail = tqe;
5392 	ifq->ifq_tail = &tqe->tqe_next;
5393 	tqe->tqe_next = NULL;
5394 	tqe->tqe_ifq = ifq;
5395 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5396 	ifq->ifq_ref++;
5397 	MUTEX_EXIT(&ifq->ifq_lock);
5398 }
5399 
5400 
5401 /* ------------------------------------------------------------------------ */
5402 /* Function:    fr_movequeue                                                */
5403 /* Returns:     Nil                                                         */
5404 /* Parameters:  tq(I)   - pointer to timeout queue information              */
5405 /*              oifp(I) - old timeout queue entry was on                    */
5406 /*              nifp(I) - new timeout queue to put entry on                 */
5407 /*		ifs	- ipf stack instance				    */
5408 /*                                                                          */
5409 /* Move a queue entry from one timeout queue to another timeout queue.      */
5410 /* If it notices that the current entry is already last and does not need   */
5411 /* to move queue, the return.                                               */
5412 /* ------------------------------------------------------------------------ */
5413 void fr_movequeue(tqe, oifq, nifq, ifs)
5414 ipftqent_t *tqe;
5415 ipftq_t *oifq, *nifq;
5416 ipf_stack_t *ifs;
5417 {
5418 	/*
5419 	 * If the queue isn't changing, and the clock hasn't ticked
5420 	 * since the last update, the operation will be a no-op.
5421 	 */
5422 	if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5423 		return;
5424 
5425 	/*
5426 	 * Grab the lock and update the timers.
5427 	 */
5428 	MUTEX_ENTER(&oifq->ifq_lock);
5429 	tqe->tqe_touched = ifs->ifs_fr_ticks;
5430 	tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5431 
5432 	/*
5433 	 * The remainder of the operation can still be a no-op.
5434 	 *
5435 	 * If the queue isn't changing, check to see if
5436 	 * an update would be meaningless.
5437 	 */
5438 	if (oifq == nifq) {
5439 		if ((tqe->tqe_next == NULL) ||
5440 		    (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5441 			MUTEX_EXIT(&oifq->ifq_lock);
5442 			return;
5443 		}
5444 	}
5445 
5446 	/*
5447 	 * Remove from the old queue
5448 	 */
5449 	*tqe->tqe_pnext = tqe->tqe_next;
5450 	if (tqe->tqe_next)
5451 		tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5452 	else
5453 		oifq->ifq_tail = tqe->tqe_pnext;
5454 	tqe->tqe_next = NULL;
5455 
5456 	/*
5457 	 * If we're moving from one queue to another, release the lock on the
5458 	 * old queue and get a lock on the new queue.  For user defined queues,
5459 	 * if we're moving off it, call delete in case it can now be freed.
5460 	 */
5461 	if (oifq != nifq) {
5462 		tqe->tqe_ifq = NULL;
5463 
5464 		(void) fr_deletetimeoutqueue(oifq);
5465 
5466 		MUTEX_EXIT(&oifq->ifq_lock);
5467 
5468 		MUTEX_ENTER(&nifq->ifq_lock);
5469 
5470 		tqe->tqe_ifq = nifq;
5471 		nifq->ifq_ref++;
5472 	}
5473 
5474 	/*
5475 	 * Add to the bottom of the new queue
5476 	 */
5477 	tqe->tqe_pnext = nifq->ifq_tail;
5478 	*nifq->ifq_tail = tqe;
5479 	nifq->ifq_tail = &tqe->tqe_next;
5480 	MUTEX_EXIT(&nifq->ifq_lock);
5481 }
5482 
5483 
5484 /* ------------------------------------------------------------------------ */
5485 /* Function:    fr_updateipid                                               */
5486 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
5487 /* Parameters:  fin(I) - pointer to packet information                      */
5488 /*                                                                          */
5489 /* When we are doing NAT, change the IP of every packet to represent a      */
5490 /* single sequence of packets coming from the host, hiding any host         */
5491 /* specific sequencing that might otherwise be revealed.  If the packet is  */
5492 /* a fragment, then store the 'new' IPid in the fragment cache and look up  */
5493 /* the fragment cache for non-leading fragments.  If a non-leading fragment */
5494 /* has no match in the cache, return an error.                              */
5495 /* ------------------------------------------------------------------------ */
5496 static INLINE int fr_updateipid(fin)
5497 fr_info_t *fin;
5498 {
5499 	u_short id, ido, sums;
5500 	u_32_t sumd, sum;
5501 	ip_t *ip;
5502 
5503 	if (fin->fin_off != 0) {
5504 		sum = fr_ipid_knownfrag(fin);
5505 		if (sum == 0xffffffff)
5506 			return -1;
5507 		sum &= 0xffff;
5508 		id = (u_short)sum;
5509 	} else {
5510 		id = fr_nextipid(fin);
5511 		if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5512 			(void) fr_ipid_newfrag(fin, (u_32_t)id);
5513 	}
5514 
5515 	ip = fin->fin_ip;
5516 	ido = ntohs(ip->ip_id);
5517 	if (id == ido)
5518 		return 0;
5519 	ip->ip_id = htons(id);
5520 	CALC_SUMD(ido, id, sumd);	/* DESTRUCTIVE MACRO! id,ido change */
5521 	sum = (~ntohs(ip->ip_sum)) & 0xffff;
5522 	sum += sumd;
5523 	sum = (sum >> 16) + (sum & 0xffff);
5524 	sum = (sum >> 16) + (sum & 0xffff);
5525 	sums = ~(u_short)sum;
5526 	ip->ip_sum = htons(sums);
5527 	return 0;
5528 }
5529 
5530 
5531 #ifdef	NEED_FRGETIFNAME
5532 /* ------------------------------------------------------------------------ */
5533 /* Function:    fr_getifname                                                */
5534 /* Returns:     char *    - pointer to interface name                       */
5535 /* Parameters:  ifp(I)    - pointer to network interface                    */
5536 /*              buffer(O) - pointer to where to store interface name        */
5537 /*                                                                          */
5538 /* Constructs an interface name in the buffer passed.  The buffer passed is */
5539 /* expected to be at least LIFNAMSIZ in bytes big.  If buffer is passed in  */
5540 /* as a NULL pointer then return a pointer to a static array.               */
5541 /* ------------------------------------------------------------------------ */
5542 char *fr_getifname(ifp, buffer)
5543 struct ifnet *ifp;
5544 char *buffer;
5545 {
5546 	static char namebuf[LIFNAMSIZ];
5547 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5548      defined(__sgi) || defined(linux) || defined(_AIX51) || \
5549      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5550 	int unit, space;
5551 	char temp[20];
5552 	char *s;
5553 # endif
5554 
5555 	ASSERT(buffer != NULL);
5556 #ifdef notdef
5557 	if (buffer == NULL)
5558 		buffer = namebuf;
5559 #endif
5560 	(void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5561 	buffer[LIFNAMSIZ - 1] = '\0';
5562 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5563      defined(__sgi) || defined(_AIX51) || \
5564      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5565 	for (s = buffer; *s; s++)
5566 		;
5567 	unit = ifp->if_unit;
5568 	space = LIFNAMSIZ - (s - buffer);
5569 	if (space > 0) {
5570 #  if defined(SNPRINTF) && defined(_KERNEL)
5571 		(void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5572 #  else
5573 		(void) sprintf(temp, "%d", unit);
5574 #  endif
5575 		(void) strncpy(s, temp, space);
5576 	}
5577 # endif
5578 	return buffer;
5579 }
5580 #endif
5581 
5582 
5583 /* ------------------------------------------------------------------------ */
5584 /* Function:    fr_ioctlswitch                                              */
5585 /* Returns:     int     - -1 continue processing, else ioctl return value   */
5586 /* Parameters:  unit(I) - device unit opened                                */
5587 /*              data(I) - pointer to ioctl data                             */
5588 /*              cmd(I)  - ioctl command                                     */
5589 /*              mode(I) - mode value                                        */
5590 /*                                                                          */
5591 /* Based on the value of unit, call the appropriate ioctl handler or return */
5592 /* EIO if ipfilter is not running.   Also checks if write perms are req'd   */
5593 /* for the device in order to execute the ioctl.                            */
5594 /* ------------------------------------------------------------------------ */
5595 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5596 int unit, mode, uid;
5597 ioctlcmd_t cmd;
5598 void *data, *ctx;
5599 ipf_stack_t *ifs;
5600 {
5601 	int error = 0;
5602 
5603 	switch (unit)
5604 	{
5605 	case IPL_LOGIPF :
5606 		error = -1;
5607 		break;
5608 	case IPL_LOGNAT :
5609 		if (ifs->ifs_fr_running > 0)
5610 			error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5611 		else
5612 			error = EIO;
5613 		break;
5614 	case IPL_LOGSTATE :
5615 		if (ifs->ifs_fr_running > 0)
5616 			error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5617 		else
5618 			error = EIO;
5619 		break;
5620 	case IPL_LOGAUTH :
5621 		if (ifs->ifs_fr_running > 0) {
5622 			if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5623 			    (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5624 				if (!(mode & FWRITE)) {
5625 					error = EPERM;
5626 				} else {
5627 					error = frrequest(unit, cmd, data,
5628 						  ifs->ifs_fr_active, 1, ifs);
5629 				}
5630 			} else {
5631 				error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5632 			}
5633 		} else
5634 			error = EIO;
5635 		break;
5636 	case IPL_LOGSYNC :
5637 #ifdef IPFILTER_SYNC
5638 		if (ifs->ifs_fr_running > 0)
5639 			error = fr_sync_ioctl(data, cmd, mode, ifs);
5640 		else
5641 #endif
5642 			error = EIO;
5643 		break;
5644 	case IPL_LOGSCAN :
5645 #ifdef IPFILTER_SCAN
5646 		if (ifs->ifs_fr_running > 0)
5647 			error = fr_scan_ioctl(data, cmd, mode, ifs);
5648 		else
5649 #endif
5650 			error = EIO;
5651 		break;
5652 	case IPL_LOGLOOKUP :
5653 #ifdef IPFILTER_LOOKUP
5654 		if (ifs->ifs_fr_running > 0)
5655 			error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5656 		else
5657 #endif
5658 			error = EIO;
5659 		break;
5660 	default :
5661 		error = EIO;
5662 		break;
5663 	}
5664 
5665 	return error;
5666 }
5667 
5668 
5669 /*
5670  * This array defines the expected size of objects coming into the kernel
5671  * for the various recognised object types.
5672  */
5673 #define	NUM_OBJ_TYPES	19
5674 
5675 static	int	fr_objbytes[NUM_OBJ_TYPES][2] = {
5676 	{ 1,	sizeof(struct frentry) },		/* frentry */
5677 	{ 0,	sizeof(struct friostat) },
5678 	{ 0,	sizeof(struct fr_info) },
5679 	{ 0,	sizeof(struct fr_authstat) },
5680 	{ 0,	sizeof(struct ipfrstat) },
5681 	{ 0,	sizeof(struct ipnat) },
5682 	{ 0,	sizeof(struct natstat) },
5683 	{ 0,	sizeof(struct ipstate_save) },
5684 	{ 1,	sizeof(struct nat_save) },		/* nat_save */
5685 	{ 0,	sizeof(struct natlookup) },
5686 	{ 1,	sizeof(struct ipstate) },		/* ipstate */
5687 	{ 0,	sizeof(struct ips_stat) },
5688 	{ 0,	sizeof(struct frauth) },
5689 	{ 0,	sizeof(struct ipftune) },
5690 	{ 0,	sizeof(struct nat) },                   /* nat_t */
5691 	{ 0,	sizeof(struct ipfruleiter) },
5692 	{ 0,	sizeof(struct ipfgeniter) },
5693 	{ 0,	sizeof(struct ipftable) },
5694 	{ 0,	sizeof(struct ipflookupiter) }
5695 };
5696 
5697 
5698 /* ------------------------------------------------------------------------ */
5699 /* Function:    fr_inobj                                                    */
5700 /* Returns:     int     - 0 = success, else failure                         */
5701 /* Parameters:  data(I) - pointer to ioctl data                             */
5702 /*              ptr(I)  - pointer to store real data in                     */
5703 /*              type(I) - type of structure being moved                     */
5704 /*                                                                          */
5705 /* Copy in the contents of what the ipfobj_t points to.  In future, we      */
5706 /* add things to check for version numbers, sizes, etc, to make it backward */
5707 /* compatible at the ABI for user land.                                     */
5708 /* ------------------------------------------------------------------------ */
5709 int fr_inobj(data, ptr, type)
5710 void *data;
5711 void *ptr;
5712 int type;
5713 {
5714 	ipfobj_t obj;
5715 	int error = 0;
5716 
5717 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5718 		return EINVAL;
5719 
5720 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5721 	if (error != 0)
5722 		return EFAULT;
5723 
5724 	if (obj.ipfo_type != type)
5725 		return EINVAL;
5726 
5727 #ifndef	IPFILTER_COMPAT
5728 	if ((fr_objbytes[type][0] & 1) != 0) {
5729 		if (obj.ipfo_size < fr_objbytes[type][1])
5730 			return EINVAL;
5731 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5732 		return EINVAL;
5733 #else
5734 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5735 		error = fr_incomptrans(&obj, ptr);
5736 		return error;
5737 	}
5738 
5739 	if ((fr_objbytes[type][0] & 1) != 0 &&
5740 	    obj.ipfo_size < fr_objbytes[type][1] ||
5741 	    obj.ipfo_size != fr_objbytes[type][1])
5742 		return EINVAL;
5743 #endif
5744 
5745 	if ((fr_objbytes[type][0] & 1) != 0) {
5746 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5747 				fr_objbytes[type][1]);
5748 	} else {
5749 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5750 				obj.ipfo_size);
5751 	}
5752 	return error;
5753 }
5754 
5755 
5756 /* ------------------------------------------------------------------------ */
5757 /* Function:    fr_inobjsz                                                  */
5758 /* Returns:     int     - 0 = success, else failure                         */
5759 /* Parameters:  data(I) - pointer to ioctl data                             */
5760 /*              ptr(I)  - pointer to store real data in                     */
5761 /*              type(I) - type of structure being moved                     */
5762 /*              sz(I)   - size of data to copy                              */
5763 /*                                                                          */
5764 /* As per fr_inobj, except the size of the object to copy in is passed in   */
5765 /* but it must not be smaller than the size defined for the type and the    */
5766 /* type must allow for varied sized objects.  The extra requirement here is */
5767 /* that sz must match the size of the object being passed in - this is not  */
5768 /* not possible nor required in fr_inobj().                                 */
5769 /* ------------------------------------------------------------------------ */
5770 int fr_inobjsz(data, ptr, type, sz)
5771 void *data;
5772 void *ptr;
5773 int type, sz;
5774 {
5775 	ipfobj_t obj;
5776 	int error;
5777 
5778 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5779 		return EINVAL;
5780 	if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5781 		return EINVAL;
5782 
5783 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5784 	if (error != 0)
5785 		return EFAULT;
5786 
5787 	if (obj.ipfo_type != type)
5788 		return EINVAL;
5789 
5790 #ifndef	IPFILTER_COMPAT
5791 	if (obj.ipfo_size != sz)
5792 		return EINVAL;
5793 #else
5794 	if (obj.ipfo_rev != IPFILTER_VERSION)
5795 		/*XXX compatibility hook here */
5796 		/*EMPTY*/;
5797 	if (obj.ipfo_size != sz)
5798 		/* XXX compatibility hook here */
5799 		return EINVAL;
5800 #endif
5801 
5802 	error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5803 	return error;
5804 }
5805 
5806 
5807 /* ------------------------------------------------------------------------ */
5808 /* Function:    fr_outobjsz                                                 */
5809 /* Returns:     int     - 0 = success, else failure                         */
5810 /* Parameters:  data(I) - pointer to ioctl data                             */
5811 /*              ptr(I)  - pointer to store real data in                     */
5812 /*              type(I) - type of structure being moved                     */
5813 /*              sz(I)   - size of data to copy                              */
5814 /*                                                                          */
5815 /* As per fr_outobj, except the size of the object to copy out is passed in */
5816 /* but it must not be smaller than the size defined for the type and the    */
5817 /* type must allow for varied sized objects.  The extra requirement here is */
5818 /* that sz must match the size of the object being passed in - this is not  */
5819 /* not possible nor required in fr_outobj().                                */
5820 /* ------------------------------------------------------------------------ */
5821 int fr_outobjsz(data, ptr, type, sz)
5822 void *data;
5823 void *ptr;
5824 int type, sz;
5825 {
5826 	ipfobj_t obj;
5827 	int error;
5828 
5829 	if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5830 	    ((fr_objbytes[type][0] & 1) == 0) ||
5831 	    (sz < fr_objbytes[type][1]))
5832 		return EINVAL;
5833 
5834 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5835 	if (error != 0)
5836 		return EFAULT;
5837 
5838 	if (obj.ipfo_type != type)
5839 		return EINVAL;
5840 
5841 #ifndef	IPFILTER_COMPAT
5842 	if (obj.ipfo_size != sz)
5843 		return EINVAL;
5844 #else
5845 	if (obj.ipfo_rev != IPFILTER_VERSION)
5846 		/* XXX compatibility hook here */
5847 		/*EMPTY*/;
5848 	if (obj.ipfo_size != sz)
5849 		/* XXX compatibility hook here */
5850 		return EINVAL;
5851 #endif
5852 
5853 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5854 	return error;
5855 }
5856 
5857 
5858 /* ------------------------------------------------------------------------ */
5859 /* Function:    fr_outobj                                                   */
5860 /* Returns:     int     - 0 = success, else failure                         */
5861 /* Parameters:  data(I) - pointer to ioctl data                             */
5862 /*              ptr(I)  - pointer to store real data in                     */
5863 /*              type(I) - type of structure being moved                     */
5864 /*                                                                          */
5865 /* Copy out the contents of what ptr is to where ipfobj points to.  In      */
5866 /* future, we add things to check for version numbers, sizes, etc, to make  */
5867 /* it backward  compatible at the ABI for user land.                        */
5868 /* ------------------------------------------------------------------------ */
5869 int fr_outobj(data, ptr, type)
5870 void *data;
5871 void *ptr;
5872 int type;
5873 {
5874 	ipfobj_t obj;
5875 	int error;
5876 
5877 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5878 		return EINVAL;
5879 
5880 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5881 	if (error != 0)
5882 		return EFAULT;
5883 
5884 	if (obj.ipfo_type != type)
5885 		return EINVAL;
5886 
5887 #ifndef	IPFILTER_COMPAT
5888 	if ((fr_objbytes[type][0] & 1) != 0) {
5889 		if (obj.ipfo_size < fr_objbytes[type][1])
5890 			return EINVAL;
5891 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5892 		return EINVAL;
5893 #else
5894 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5895 		error = fr_outcomptrans(&obj, ptr);
5896 		return error;
5897 	}
5898 
5899 	if ((fr_objbytes[type][0] & 1) != 0 &&
5900 	    obj.ipfo_size < fr_objbytes[type][1] ||
5901 	    obj.ipfo_size != fr_objbytes[type][1])
5902 		return EINVAL;
5903 #endif
5904 
5905 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5906 	return error;
5907 }
5908 
5909 
5910 /* ------------------------------------------------------------------------ */
5911 /* Function:    fr_checkl4sum                                               */
5912 /* Returns:     int     - 0 = good, -1 = bad, 1 = cannot check              */
5913 /* Parameters:  fin(I) - pointer to packet information                      */
5914 /*                                                                          */
5915 /* If possible, calculate the layer 4 checksum for the packet.  If this is  */
5916 /* not possible, return without indicating a failure or success but in a    */
5917 /* way that is ditinguishable.                                              */
5918 /* ------------------------------------------------------------------------ */
5919 int fr_checkl4sum(fin)
5920 fr_info_t *fin;
5921 {
5922 	u_short sum, hdrsum, *csump;
5923 	udphdr_t *udp;
5924 	int dosum;
5925 	ipf_stack_t *ifs = fin->fin_ifs;
5926 
5927 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5928 	net_handle_t net_data_p;
5929 	if (fin->fin_v == 4)
5930 		net_data_p = ifs->ifs_ipf_ipv4;
5931 	else
5932 		net_data_p = ifs->ifs_ipf_ipv6;
5933 #endif
5934 
5935 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
5936 		return 0;
5937 
5938 	/*
5939 	 * If the TCP packet isn't a fragment, isn't too short and otherwise
5940 	 * isn't already considered "bad", then validate the checksum.  If
5941 	 * this check fails then considered the packet to be "bad".
5942 	 */
5943 	if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
5944 		return 1;
5945 
5946 	csump = NULL;
5947 	hdrsum = 0;
5948 	dosum = 0;
5949 	sum = 0;
5950 
5951 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5952 	ASSERT(fin->fin_m != NULL);
5953 	if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
5954 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
5955 			hdrsum = 0;
5956 			sum = 0;
5957 	} else {
5958 #endif
5959 		switch (fin->fin_p)
5960 		{
5961 		case IPPROTO_TCP :
5962 			csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
5963 			dosum = 1;
5964 			break;
5965 
5966 		case IPPROTO_UDP :
5967 			udp = fin->fin_dp;
5968 			if (udp->uh_sum != 0) {
5969 				csump = &udp->uh_sum;
5970 				dosum = 1;
5971 			}
5972 			break;
5973 
5974 		case IPPROTO_ICMP :
5975 			csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
5976 			dosum = 1;
5977 			break;
5978 
5979 		default :
5980 			return 1;
5981 			/*NOTREACHED*/
5982 		}
5983 
5984 		if (csump != NULL)
5985 			hdrsum = *csump;
5986 
5987 		if (dosum)
5988 			sum = fr_cksum(fin->fin_m, fin->fin_ip,
5989 				       fin->fin_p, fin->fin_dp);
5990 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5991 	}
5992 #endif
5993 #if !defined(_KERNEL)
5994 	if (sum == hdrsum) {
5995 		FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
5996 	} else {
5997 		FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
5998 	}
5999 #endif
6000 	if (hdrsum == sum)
6001 		return 0;
6002 	return -1;
6003 }
6004 
6005 
6006 /* ------------------------------------------------------------------------ */
6007 /* Function:    fr_ifpfillv4addr                                            */
6008 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6009 /* Parameters:  atype(I)   - type of network address update to perform      */
6010 /*              sin(I)     - pointer to source of address information       */
6011 /*              mask(I)    - pointer to source of netmask information       */
6012 /*              inp(I)     - pointer to destination address store           */
6013 /*              inpmask(I) - pointer to destination netmask store           */
6014 /*                                                                          */
6015 /* Given a type of network address update (atype) to perform, copy          */
6016 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6017 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6018 /* which case the operation fails.  For all values of atype other than      */
6019 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6020 /* value.                                                                   */
6021 /* ------------------------------------------------------------------------ */
6022 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
6023 int atype;
6024 struct sockaddr_in *sin, *mask;
6025 struct in_addr *inp, *inpmask;
6026 {
6027 	if (inpmask != NULL && atype != FRI_NETMASKED)
6028 		inpmask->s_addr = 0xffffffff;
6029 
6030 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6031 		if (atype == FRI_NETMASKED) {
6032 			if (inpmask == NULL)
6033 				return -1;
6034 			inpmask->s_addr = mask->sin_addr.s_addr;
6035 		}
6036 		inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
6037 	} else {
6038 		inp->s_addr = sin->sin_addr.s_addr;
6039 	}
6040 	return 0;
6041 }
6042 
6043 
6044 #ifdef	USE_INET6
6045 /* ------------------------------------------------------------------------ */
6046 /* Function:    fr_ifpfillv6addr                                            */
6047 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6048 /* Parameters:  atype(I)   - type of network address update to perform      */
6049 /*              sin(I)     - pointer to source of address information       */
6050 /*              mask(I)    - pointer to source of netmask information       */
6051 /*              inp(I)     - pointer to destination address store           */
6052 /*              inpmask(I) - pointer to destination netmask store           */
6053 /*                                                                          */
6054 /* Given a type of network address update (atype) to perform, copy          */
6055 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6056 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6057 /* which case the operation fails.  For all values of atype other than      */
6058 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6059 /* value.                                                                   */
6060 /* ------------------------------------------------------------------------ */
6061 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
6062 int atype;
6063 struct sockaddr_in6 *sin, *mask;
6064 struct in_addr *inp, *inpmask;
6065 {
6066 	i6addr_t *src, *dst, *and, *dmask;
6067 
6068 	src = (i6addr_t *)&sin->sin6_addr;
6069 	and = (i6addr_t *)&mask->sin6_addr;
6070 	dst = (i6addr_t *)inp;
6071 	dmask = (i6addr_t *)inpmask;
6072 
6073 	if (inpmask != NULL && atype != FRI_NETMASKED) {
6074 		dmask->i6[0] = 0xffffffff;
6075 		dmask->i6[1] = 0xffffffff;
6076 		dmask->i6[2] = 0xffffffff;
6077 		dmask->i6[3] = 0xffffffff;
6078 	}
6079 
6080 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6081 		if (atype == FRI_NETMASKED) {
6082 			if (inpmask == NULL)
6083 				return -1;
6084 			dmask->i6[0] = and->i6[0];
6085 			dmask->i6[1] = and->i6[1];
6086 			dmask->i6[2] = and->i6[2];
6087 			dmask->i6[3] = and->i6[3];
6088 		}
6089 
6090 		dst->i6[0] = src->i6[0] & and->i6[0];
6091 		dst->i6[1] = src->i6[1] & and->i6[1];
6092 		dst->i6[2] = src->i6[2] & and->i6[2];
6093 		dst->i6[3] = src->i6[3] & and->i6[3];
6094 	} else {
6095 		dst->i6[0] = src->i6[0];
6096 		dst->i6[1] = src->i6[1];
6097 		dst->i6[2] = src->i6[2];
6098 		dst->i6[3] = src->i6[3];
6099 	}
6100 	return 0;
6101 }
6102 #endif
6103 
6104 
6105 /* ------------------------------------------------------------------------ */
6106 /* Function:    fr_matchtag                                                 */
6107 /* Returns:     0 == mismatch, 1 == match.                                  */
6108 /* Parameters:  tag1(I) - pointer to first tag to compare                   */
6109 /*              tag2(I) - pointer to second tag to compare                  */
6110 /*                                                                          */
6111 /* Returns true (non-zero) or false(0) if the two tag structures can be     */
6112 /* considered to be a match or not match, respectively.  The tag is 16      */
6113 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so    */
6114 /* compare the ints instead, for speed. tag1 is the master of the           */
6115 /* comparison.  This function should only be called with both tag1 and tag2 */
6116 /* as non-NULL pointers.                                                    */
6117 /* ------------------------------------------------------------------------ */
6118 int fr_matchtag(tag1, tag2)
6119 ipftag_t *tag1, *tag2;
6120 {
6121 	if (tag1 == tag2)
6122 		return 1;
6123 
6124 	if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
6125 		return 1;
6126 
6127 	if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
6128 	    (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
6129 	    (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
6130 	    (tag1->ipt_num[3] == tag2->ipt_num[3]))
6131 		return 1;
6132 	return 0;
6133 }
6134 
6135 
6136 /* ------------------------------------------------------------------------ */
6137 /* Function:    fr_coalesce                                                 */
6138 /* Returns:     1 == success, -1 == failure, 0 == no change                 */
6139 /* Parameters:  fin(I) - pointer to packet information                      */
6140 /*                                                                          */
6141 /* Attempt to get all of the packet data into a single, contiguous buffer.  */
6142 /* If this call returns a failure then the buffers have also been freed.    */
6143 /* ------------------------------------------------------------------------ */
6144 int fr_coalesce(fin)
6145 fr_info_t *fin;
6146 {
6147 	ipf_stack_t *ifs = fin->fin_ifs;
6148 	if ((fin->fin_flx & FI_COALESCE) != 0)
6149 		return 1;
6150 
6151 	/*
6152 	 * If the mbuf pointers indicate that there is no mbuf to work with,
6153 	 * return but do not indicate success or failure.
6154 	 */
6155 	if (fin->fin_m == NULL || fin->fin_mp == NULL)
6156 		return 0;
6157 
6158 #if defined(_KERNEL)
6159 	if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
6160 		IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
6161 # ifdef MENTAT
6162 		FREE_MB_T(*fin->fin_mp);
6163 # endif
6164 		*fin->fin_mp = NULL;
6165 		fin->fin_m = NULL;
6166 		return -1;
6167 	}
6168 #else
6169 	fin = fin;	/* LINT */
6170 #endif
6171 	return 1;
6172 }
6173 
6174 
6175 /*
6176  * The following table lists all of the tunable variables that can be
6177  * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT.  The format of each row
6178  * in the table below is as follows:
6179  *
6180  * pointer to value, name of value, minimum, maximum, size of the value's
6181  *     container, value attribute flags
6182  *
6183  * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6184  * means the value can only be written to when IPFilter is loaded but disabled.
6185  * The obvious implication is if neither of these are set then the value can be
6186  * changed at any time without harm.
6187  */
6188 ipftuneable_t lcl_ipf_tuneables[] = {
6189 	/* filtering */
6190 	{ { NULL },	"fr_flags",		0,	0xffffffff,
6191 			0,		0 },
6192 	{ { NULL },	"fr_active",		0,	0,
6193 			0,		IPFT_RDONLY },
6194 	{ { NULL },	"fr_control_forwarding",	0, 1,
6195 			0,		0 },
6196 	{ { NULL },	"fr_update_ipid",	0,	1,
6197 			0,		0 },
6198 	{ { NULL },	"fr_chksrc",		0,	1,
6199 			0,		0 },
6200 	{ { NULL },	"fr_minttl",		0,	1,
6201 			0,		0 },
6202 	{ { NULL }, 	"fr_icmpminfragmtu",	0,	1,
6203 			0,		0 },
6204 	{ { NULL },		"fr_pass",		0,	0xffffffff,
6205 			0,		0 },
6206 #if SOLARIS2 >= 10
6207 	{ { NULL },	"ipf_loopback",		0,	1,
6208 			0,		IPFT_WRDISABLED },
6209 #endif
6210 	/* state */
6211 	{ { NULL }, "fr_tcpidletimeout",	1,	0x7fffffff,
6212 			0,	IPFT_WRDISABLED },
6213 	{ { NULL },	"fr_tcpclosewait",	1,	0x7fffffff,
6214 			0,	IPFT_WRDISABLED },
6215 	{ { NULL },	"fr_tcplastack",	1,	0x7fffffff,
6216 			0,		IPFT_WRDISABLED },
6217 	{ { NULL },	"fr_tcptimeout",	1,	0x7fffffff,
6218 			0,		IPFT_WRDISABLED },
6219 	{ { NULL },	"fr_tcpclosed",		1,	0x7fffffff,
6220 			0,		IPFT_WRDISABLED },
6221 	{ { NULL },	"fr_tcphalfclosed",	1,	0x7fffffff,
6222 			0,	IPFT_WRDISABLED },
6223 	{ { NULL },	"fr_udptimeout",	1,	0x7fffffff,
6224 			0,		IPFT_WRDISABLED },
6225 	{ { NULL }, "fr_udpacktimeout",	1,	0x7fffffff,
6226 			0,	IPFT_WRDISABLED },
6227 	{ { NULL },	"fr_icmptimeout",	1,	0x7fffffff,
6228 			0,		IPFT_WRDISABLED },
6229 	{ { NULL }, "fr_icmpacktimeout",	1,	0x7fffffff,
6230 			0,	IPFT_WRDISABLED },
6231 	{ { NULL }, "fr_iptimeout",		1,	0x7fffffff,
6232 			0,		IPFT_WRDISABLED },
6233 	{ { NULL },	"fr_statemax",		1,	0x7fffffff,
6234 			0,		0 },
6235 	{ { NULL },	"fr_statesize",		1,	0x7fffffff,
6236 			0,		IPFT_WRDISABLED },
6237 	{ { NULL },	"fr_state_lock",	0,	1,
6238 			0,		IPFT_RDONLY },
6239 	{ { NULL }, "fr_state_maxbucket", 1,	0x7fffffff,
6240 			0,	IPFT_WRDISABLED },
6241 	{ { NULL }, "fr_state_maxbucket_reset",	0, 1,
6242 			0, IPFT_WRDISABLED },
6243 	{ { NULL },	"ipstate_logging",	0,	1,
6244 			0,	0 },
6245 	{ { NULL },	"state_flush_level_hi",	1,	100,
6246 			0,		0 },
6247 	{ { NULL },	"state_flush_level_lo",	1,	100,
6248 			0,		0 },
6249 	/* nat */
6250 	{ { NULL },		"fr_nat_lock",		0,	1,
6251 			0,		IPFT_RDONLY },
6252 	{ { NULL },	"ipf_nattable_sz",	1,	0x7fffffff,
6253 			0,	IPFT_WRDISABLED },
6254 	{ { NULL }, "ipf_nattable_max",	1,	0x7fffffff,
6255 			0,	0 },
6256 	{ { NULL },	"ipf_natrules_sz",	1,	0x7fffffff,
6257 			0,	IPFT_WRDISABLED },
6258 	{ { NULL },	"ipf_rdrrules_sz",	1,	0x7fffffff,
6259 			0,	IPFT_WRDISABLED },
6260 	{ { NULL },	"ipf_hostmap_sz",	1,	0x7fffffff,
6261 			0,		IPFT_WRDISABLED },
6262 	{ { NULL }, "fr_nat_maxbucket",	1,	0x7fffffff,
6263 			0,	IPFT_WRDISABLED },
6264 	{ { NULL },	"fr_nat_maxbucket_reset",	0, 1,
6265 			0,	IPFT_WRDISABLED },
6266 	{ { NULL },		"nat_logging",		0,	1,
6267 			0,		0 },
6268 	{ { NULL },	"fr_defnatage",		1,	0x7fffffff,
6269 			0,		IPFT_WRDISABLED },
6270 	{ { NULL },	"fr_defnatipage",	1,	0x7fffffff,
6271 			0,		IPFT_WRDISABLED },
6272 	{ { NULL }, "fr_defnaticmpage",	1,	0x7fffffff,
6273 			0,	IPFT_WRDISABLED },
6274 	{ { NULL },	"nat_flush_level_hi",	1,	100,
6275 			0,		0 },
6276 	{ { NULL },	"nat_flush_level_lo",	1,	100,
6277 			0,		0 },
6278 	/* frag */
6279 	{ { NULL },	"ipfr_size",		1,	0x7fffffff,
6280 			0,		IPFT_WRDISABLED },
6281 	{ { NULL },	"fr_ipfrttl",		1,	0x7fffffff,
6282 			0,		IPFT_WRDISABLED },
6283 #ifdef IPFILTER_LOG
6284 	/* log */
6285 	{ { NULL },	"ipl_suppress",		0,	1,
6286 			0,		0 },
6287 	{ { NULL },	"ipl_buffer_sz",	0,	0,
6288 			0,		IPFT_RDONLY },
6289 	{ { NULL },	"ipl_logmax",		0,	0x7fffffff,
6290 			0,		IPFT_WRDISABLED },
6291 	{ { NULL },	"ipl_logall",		0,	1,
6292 			0,		0 },
6293 	{ { NULL },	"ipl_logsize",		0,	0x80000,
6294 			0,		0 },
6295 #endif
6296 	{ { NULL },		NULL,			0,	0 }
6297 };
6298 
6299 static ipftuneable_t *
6300 tune_lookup(ipf_stack_t *ifs, char *name)
6301 {
6302     int i;
6303 
6304     for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6305 	if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6306 	    return (&ifs->ifs_ipf_tuneables[i]);
6307     }
6308     return (NULL);
6309 }
6310 
6311 #ifdef _KERNEL
6312 extern dev_info_t *ipf_dev_info;
6313 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6314 #endif
6315 
6316 /* -------------------------------------------------------------------- */
6317 /* Function:	ipftuneable_setdefs()					*/
6318 /* Returns:		void						*/
6319 /* Parameters:	ifs - pointer to newly allocated IPF instance		*/
6320 /*				assigned to	IP instance		*/
6321 /*									*/
6322 /* Function initializes IPF instance variables. Function is invoked	*/
6323 /* from	ipftuneable_alloc(). ipftuneable_alloc() is called only one	*/
6324 /* time during IP instance lifetime - at the time of IP instance	*/
6325 /* creation. Anytime IP	instance is being created new private IPF	*/
6326 /* instance is allocated and assigned to it. The moment of IP 		*/
6327 /* instance creation is the right time to initialize those IPF 		*/
6328 /* variables.								*/
6329 /*									*/
6330 /* -------------------------------------------------------------------- */
6331 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6332 {
6333 	ifs->ifs_ipfr_size = IPFT_SIZE;
6334 	ifs->ifs_fr_ipfrttl = 120;	/* 60 seconds */
6335 
6336 	/* it comes from fr_authinit() in IPF auth */
6337 	ifs->ifs_fr_authsize = FR_NUMAUTH;
6338 	ifs->ifs_fr_defaultauthage = 600;
6339 
6340 	/* it comes from fr_stateinit() in IPF state */
6341 	ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5);	/* five days */
6342 	ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6343 	ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6344 	ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6345 	ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6346 	ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600);	/* 2 hours */
6347 	ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6348 	ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6349 	ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6350 	ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6351 	ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6352 	ifs->ifs_fr_statemax = IPSTATE_MAX;
6353 	ifs->ifs_fr_statesize = IPSTATE_SIZE;
6354 	ifs->ifs_fr_state_maxbucket_reset = 1;
6355 	ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6356 	ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6357 
6358 	/* it comes from fr_natinit() in ipnat */
6359 	ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6360 	ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6361 	ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6362 	ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6363 	ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6364 	ifs->ifs_fr_nat_maxbucket_reset = 1;
6365 	ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6366 	ifs->ifs_fr_defnatipage = 120;		/* 60 seconds */
6367 	ifs->ifs_fr_defnaticmpage = 6;		/* 3 seconds */
6368 	ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6369 	ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6370 
6371 #ifdef IPFILTER_LOG
6372 	/* it comes from fr_loginit() in IPF log */
6373 	ifs->ifs_ipl_suppress = 1;
6374 	ifs->ifs_ipl_logmax = IPL_LOGMAX;
6375 	ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6376 
6377 	/* from fr_natinit() */
6378 	ifs->ifs_nat_logging = 1;
6379 
6380 	/* from fr_stateinit() */
6381 	ifs->ifs_ipstate_logging = 1;
6382 #else
6383 	/* from fr_natinit() */
6384 	ifs->ifs_nat_logging = 0;
6385 
6386 	/* from fr_stateinit() */
6387 	ifs->ifs_ipstate_logging = 0;
6388 #endif
6389 	ifs->ifs_ipf_loopback = 0;
6390 
6391 }
6392 /*
6393  * Allocate a per-stack tuneable and copy in the names. Then
6394  * set it to point to each of the per-stack tunables.
6395  */
6396 void
6397 ipftuneable_alloc(ipf_stack_t *ifs)
6398 {
6399     ipftuneable_t *item;
6400 
6401     KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6402 	sizeof (lcl_ipf_tuneables));
6403     bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6404 	sizeof (lcl_ipf_tuneables));
6405 
6406 #define TUNE_SET(_ifs, _name, _field)			\
6407     item = tune_lookup((_ifs), (_name));		\
6408     if (item != NULL) {					\
6409 	item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field);	\
6410 	item->ipft_sz = sizeof ((_ifs)->_field);	\
6411     }
6412 
6413     TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6414     TUNE_SET(ifs, "fr_active", ifs_fr_active);
6415     TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6416     TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6417     TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6418     TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6419     TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6420     TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6421     TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6422     TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6423     TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6424     TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6425     TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6426     TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6427     TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6428     TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6429     TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6430     TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6431     TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6432     TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6433     TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6434     TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6435     TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6436     TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6437     TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6438     TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6439     TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6440     TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6441     TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6442     TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6443     TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6444     TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6445     TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6446     TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6447     TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6448     TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6449     TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6450     TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6451     TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6452     TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6453     TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6454     TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6455     TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6456     TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6457 #ifdef IPFILTER_LOG
6458     TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6459     TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6460     TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6461     TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6462     TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6463 #endif
6464 #undef TUNE_SET
6465 
6466 	ipftuneable_setdefs(ifs);
6467 
6468 #ifdef _KERNEL
6469     (void) ipf_property_update(ipf_dev_info, ifs);
6470 #endif
6471 }
6472 
6473 void
6474 ipftuneable_free(ipf_stack_t *ifs)
6475 {
6476 	KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6477 	ifs->ifs_ipf_tuneables = NULL;
6478 }
6479 
6480 /* ------------------------------------------------------------------------ */
6481 /* Function:    fr_findtunebycookie                                         */
6482 /* Returns:     NULL = search failed, else pointer to tune struct           */
6483 /* Parameters:  cookie(I) - cookie value to search for amongst tuneables    */
6484 /*              next(O)   - pointer to place to store the cookie for the    */
6485 /*                          "next" tuneable, if it is desired.              */
6486 /*                                                                          */
6487 /* This function is used to walk through all of the existing tunables with  */
6488 /* successive calls.  It searches the known tunables for the one which has  */
6489 /* a matching value for "cookie" - ie its address.  When returning a match, */
6490 /* the next one to be found may be returned inside next.                    */
6491 /* ------------------------------------------------------------------------ */
6492 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6493 void *cookie, **next;
6494 ipf_stack_t * ifs;
6495 {
6496 	ipftuneable_t *ta, **tap;
6497 
6498 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6499 		if (ta == cookie) {
6500 			if (next != NULL) {
6501 				/*
6502 				 * If the next entry in the array has a name
6503 				 * present, then return a pointer to it for
6504 				 * where to go next, else return a pointer to
6505 				 * the dynaminc list as a key to search there
6506 				 * next.  This facilitates a weak linking of
6507 				 * the two "lists" together.
6508 				 */
6509 				if ((ta + 1)->ipft_name != NULL)
6510 					*next = ta + 1;
6511 				else
6512 					*next = &ifs->ifs_ipf_tunelist;
6513 			}
6514 			return ta;
6515 		}
6516 
6517 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6518 		if (tap == cookie) {
6519 			if (next != NULL)
6520 				*next = &ta->ipft_next;
6521 			return ta;
6522 		}
6523 
6524 	if (next != NULL)
6525 		*next = NULL;
6526 	return NULL;
6527 }
6528 
6529 
6530 /* ------------------------------------------------------------------------ */
6531 /* Function:    fr_findtunebyname                                           */
6532 /* Returns:     NULL = search failed, else pointer to tune struct           */
6533 /* Parameters:  name(I) - name of the tuneable entry to find.               */
6534 /*                                                                          */
6535 /* Search the static array of tuneables and the list of dynamic tuneables   */
6536 /* for an entry with a matching name.  If we can find one, return a pointer */
6537 /* to the matching structure.                                               */
6538 /* ------------------------------------------------------------------------ */
6539 static ipftuneable_t *fr_findtunebyname(name, ifs)
6540 const char *name;
6541 ipf_stack_t *ifs;
6542 {
6543 	ipftuneable_t *ta;
6544 
6545 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6546 		if (!strcmp(ta->ipft_name, name)) {
6547 			return ta;
6548 		}
6549 
6550 	for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6551 		if (!strcmp(ta->ipft_name, name)) {
6552 			return ta;
6553 		}
6554 
6555 	return NULL;
6556 }
6557 
6558 
6559 /* ------------------------------------------------------------------------ */
6560 /* Function:    fr_addipftune                                               */
6561 /* Returns:     int - 0 == success, else failure                            */
6562 /* Parameters:  newtune - pointer to new tune struct to add to tuneables    */
6563 /*                                                                          */
6564 /* Appends the tune structure pointer to by "newtune" to the end of the     */
6565 /* current list of "dynamic" tuneable parameters.  Once added, the owner    */
6566 /* of the object is not expected to ever change "ipft_next".                */
6567 /* ------------------------------------------------------------------------ */
6568 int fr_addipftune(newtune, ifs)
6569 ipftuneable_t *newtune;
6570 ipf_stack_t *ifs;
6571 {
6572 	ipftuneable_t *ta, **tap;
6573 
6574 	ta = fr_findtunebyname(newtune->ipft_name, ifs);
6575 	if (ta != NULL)
6576 		return EEXIST;
6577 
6578 	for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6579 		;
6580 
6581 	newtune->ipft_next = NULL;
6582 	*tap = newtune;
6583 	return 0;
6584 }
6585 
6586 
6587 /* ------------------------------------------------------------------------ */
6588 /* Function:    fr_delipftune                                               */
6589 /* Returns:     int - 0 == success, else failure                            */
6590 /* Parameters:  oldtune - pointer to tune struct to remove from the list of */
6591 /*                        current dynamic tuneables                         */
6592 /*                                                                          */
6593 /* Search for the tune structure, by pointer, in the list of those that are */
6594 /* dynamically added at run time.  If found, adjust the list so that this   */
6595 /* structure is no longer part of it.                                       */
6596 /* ------------------------------------------------------------------------ */
6597 int fr_delipftune(oldtune, ifs)
6598 ipftuneable_t *oldtune;
6599 ipf_stack_t *ifs;
6600 {
6601 	ipftuneable_t *ta, **tap;
6602 
6603 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6604 		if (ta == oldtune) {
6605 			*tap = oldtune->ipft_next;
6606 			oldtune->ipft_next = NULL;
6607 			return 0;
6608 		}
6609 
6610 	return ESRCH;
6611 }
6612 
6613 
6614 /* ------------------------------------------------------------------------ */
6615 /* Function:    fr_ipftune                                                  */
6616 /* Returns:     int - 0 == success, else failure                            */
6617 /* Parameters:  cmd(I)  - ioctl command number                              */
6618 /*              data(I) - pointer to ioctl data structure                   */
6619 /*                                                                          */
6620 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET.  These  */
6621 /* three ioctls provide the means to access and control global variables    */
6622 /* within IPFilter, allowing (for example) timeouts and table sizes to be   */
6623 /* changed without rebooting, reloading or recompiling.  The initialisation */
6624 /* and 'destruction' routines of the various components of ipfilter are all */
6625 /* each responsible for handling their own values being too big.            */
6626 /* ------------------------------------------------------------------------ */
6627 int fr_ipftune(cmd, data, ifs)
6628 ioctlcmd_t cmd;
6629 void *data;
6630 ipf_stack_t *ifs;
6631 {
6632 	ipftuneable_t *ta;
6633 	ipftune_t tu;
6634 	void *cookie;
6635 	int error;
6636 
6637 	error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6638 	if (error != 0)
6639 		return error;
6640 
6641 	tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6642 	cookie = tu.ipft_cookie;
6643 	ta = NULL;
6644 
6645 	switch (cmd)
6646 	{
6647 	case SIOCIPFGETNEXT :
6648 		/*
6649 		 * If cookie is non-NULL, assume it to be a pointer to the last
6650 		 * entry we looked at, so find it (if possible) and return a
6651 		 * pointer to the next one after it.  The last entry in the
6652 		 * the table is a NULL entry, so when we get to it, set cookie
6653 		 * to NULL and return that, indicating end of list, erstwhile
6654 		 * if we come in with cookie set to NULL, we are starting anew
6655 		 * at the front of the list.
6656 		 */
6657 		if (cookie != NULL) {
6658 			ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6659 		} else {
6660 			ta = ifs->ifs_ipf_tuneables;
6661 			tu.ipft_cookie = ta + 1;
6662 		}
6663 		if (ta != NULL) {
6664 			/*
6665 			 * Entry found, but does the data pointed to by that
6666 			 * row fit in what we can return?
6667 			 */
6668 			if (ta->ipft_sz > sizeof(tu.ipft_un))
6669 				return EINVAL;
6670 
6671 			tu.ipft_vlong = 0;
6672 			if (ta->ipft_sz == sizeof(u_long))
6673 				tu.ipft_vlong = *ta->ipft_plong;
6674 			else if (ta->ipft_sz == sizeof(u_int))
6675 				tu.ipft_vint = *ta->ipft_pint;
6676 			else if (ta->ipft_sz == sizeof(u_short))
6677 				tu.ipft_vshort = *ta->ipft_pshort;
6678 			else if (ta->ipft_sz == sizeof(u_char))
6679 				tu.ipft_vchar = *ta->ipft_pchar;
6680 
6681 			tu.ipft_sz = ta->ipft_sz;
6682 			tu.ipft_min = ta->ipft_min;
6683 			tu.ipft_max = ta->ipft_max;
6684 			tu.ipft_flags = ta->ipft_flags;
6685 			bcopy(ta->ipft_name, tu.ipft_name,
6686 			      MIN(sizeof(tu.ipft_name),
6687 				  strlen(ta->ipft_name) + 1));
6688 		}
6689 		error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6690 		break;
6691 
6692 	case SIOCIPFGET :
6693 	case SIOCIPFSET :
6694 		/*
6695 		 * Search by name or by cookie value for a particular entry
6696 		 * in the tuning paramter table.
6697 		 */
6698 		error = ESRCH;
6699 		if (cookie != NULL) {
6700 			ta = fr_findtunebycookie(cookie, NULL, ifs);
6701 			if (ta != NULL)
6702 				error = 0;
6703 		} else if (tu.ipft_name[0] != '\0') {
6704 			ta = fr_findtunebyname(tu.ipft_name, ifs);
6705 			if (ta != NULL)
6706 				error = 0;
6707 		}
6708 		if (error != 0)
6709 			break;
6710 
6711 		if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6712 			/*
6713 			 * Fetch the tuning parameters for a particular value
6714 			 */
6715 			tu.ipft_vlong = 0;
6716 			if (ta->ipft_sz == sizeof(u_long))
6717 				tu.ipft_vlong = *ta->ipft_plong;
6718 			else if (ta->ipft_sz == sizeof(u_int))
6719 				tu.ipft_vint = *ta->ipft_pint;
6720 			else if (ta->ipft_sz == sizeof(u_short))
6721 				tu.ipft_vshort = *ta->ipft_pshort;
6722 			else if (ta->ipft_sz == sizeof(u_char))
6723 				tu.ipft_vchar = *ta->ipft_pchar;
6724 			tu.ipft_cookie = ta;
6725 			tu.ipft_sz = ta->ipft_sz;
6726 			tu.ipft_min = ta->ipft_min;
6727 			tu.ipft_max = ta->ipft_max;
6728 			tu.ipft_flags = ta->ipft_flags;
6729 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6730 
6731 		} else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6732 			/*
6733 			 * Set an internal parameter.  The hard part here is
6734 			 * getting the new value safely and correctly out of
6735 			 * the kernel (given we only know its size, not type.)
6736 			 */
6737 			u_long in;
6738 
6739 			if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6740 			    (ifs->ifs_fr_running > 0)) {
6741 				error = EBUSY;
6742 				break;
6743 			}
6744 
6745 			in = tu.ipft_vlong;
6746 			if (in < ta->ipft_min || in > ta->ipft_max) {
6747 				error = EINVAL;
6748 				break;
6749 			}
6750 
6751 			if (ta->ipft_sz == sizeof(u_long)) {
6752 				tu.ipft_vlong = *ta->ipft_plong;
6753 				*ta->ipft_plong = in;
6754 			} else if (ta->ipft_sz == sizeof(u_int)) {
6755 				tu.ipft_vint = *ta->ipft_pint;
6756 				*ta->ipft_pint = (u_int)(in & 0xffffffff);
6757 			} else if (ta->ipft_sz == sizeof(u_short)) {
6758 				tu.ipft_vshort = *ta->ipft_pshort;
6759 				*ta->ipft_pshort = (u_short)(in & 0xffff);
6760 			} else if (ta->ipft_sz == sizeof(u_char)) {
6761 				tu.ipft_vchar = *ta->ipft_pchar;
6762 				*ta->ipft_pchar = (u_char)(in & 0xff);
6763 			}
6764 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6765 		}
6766 		break;
6767 
6768 	default :
6769 		error = EINVAL;
6770 		break;
6771 	}
6772 
6773 	return error;
6774 }
6775 
6776 
6777 /* ------------------------------------------------------------------------ */
6778 /* Function:    fr_initialise                                               */
6779 /* Returns:     int - 0 == success,  < 0 == failure                         */
6780 /* Parameters:  None.                                                       */
6781 /*                                                                          */
6782 /* Call of the initialise functions for all the various subsystems inside   */
6783 /* of IPFilter.  If any of them should fail, return immeadiately a failure  */
6784 /* BUT do not try to recover from the error here.                           */
6785 /* ------------------------------------------------------------------------ */
6786 int fr_initialise(ifs)
6787 ipf_stack_t *ifs;
6788 {
6789 	int i;
6790 
6791 #ifdef IPFILTER_LOG
6792 	i = fr_loginit(ifs);
6793 	if (i < 0)
6794 		return -10 + i;
6795 #endif
6796 	i = fr_natinit(ifs);
6797 	if (i < 0)
6798 		return -20 + i;
6799 
6800 	i = fr_stateinit(ifs);
6801 	if (i < 0)
6802 		return -30 + i;
6803 
6804 	i = fr_authinit(ifs);
6805 	if (i < 0)
6806 		return -40 + i;
6807 
6808 	i = fr_fraginit(ifs);
6809 	if (i < 0)
6810 		return -50 + i;
6811 
6812 	i = appr_init(ifs);
6813 	if (i < 0)
6814 		return -60 + i;
6815 
6816 #ifdef IPFILTER_SYNC
6817 	i = ipfsync_init(ifs);
6818 	if (i < 0)
6819 		return -70 + i;
6820 #endif
6821 #ifdef IPFILTER_SCAN
6822 	i = ipsc_init(ifs);
6823 	if (i < 0)
6824 		return -80 + i;
6825 #endif
6826 #ifdef IPFILTER_LOOKUP
6827 	i = ip_lookup_init(ifs);
6828 	if (i < 0)
6829 		return -90 + i;
6830 #endif
6831 #ifdef IPFILTER_COMPILED
6832 	ipfrule_add(ifs);
6833 #endif
6834 	return 0;
6835 }
6836 
6837 
6838 /* ------------------------------------------------------------------------ */
6839 /* Function:    fr_deinitialise                                             */
6840 /* Returns:     None.                                                       */
6841 /* Parameters:  None.                                                       */
6842 /*                                                                          */
6843 /* Call all the various subsystem cleanup routines to deallocate memory or  */
6844 /* destroy locks or whatever they've done that they need to now undo.       */
6845 /* The order here IS important as there are some cross references of        */
6846 /* internal data structures.                                                */
6847 /* ------------------------------------------------------------------------ */
6848 void fr_deinitialise(ifs)
6849 ipf_stack_t *ifs;
6850 {
6851 	fr_fragunload(ifs);
6852 	fr_authunload(ifs);
6853 	fr_natunload(ifs);
6854 	fr_stateunload(ifs);
6855 #ifdef IPFILTER_SCAN
6856 	fr_scanunload(ifs);
6857 #endif
6858 	appr_unload(ifs);
6859 
6860 #ifdef IPFILTER_COMPILED
6861 	ipfrule_remove(ifs);
6862 #endif
6863 
6864 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6865 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6866 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6867 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6868 
6869 #ifdef IPFILTER_LOOKUP
6870 	ip_lookup_unload(ifs);
6871 #endif
6872 
6873 #ifdef IPFILTER_LOG
6874 	fr_logunload(ifs);
6875 #endif
6876 }
6877 
6878 
6879 /* ------------------------------------------------------------------------ */
6880 /* Function:    fr_zerostats                                                */
6881 /* Returns:     int - 0 = success, else failure                             */
6882 /* Parameters:  data(O) - pointer to pointer for copying data back to       */
6883 /*                                                                          */
6884 /* Copies the current statistics out to userspace and then zero's the       */
6885 /* current ones in the kernel. The lock is only held across the bzero() as  */
6886 /* the copyout may result in paging (ie network activity.)                  */
6887 /* ------------------------------------------------------------------------ */
6888 int	fr_zerostats(data, ifs)
6889 caddr_t	data;
6890 ipf_stack_t *ifs;
6891 {
6892 	friostat_t fio;
6893 	int error;
6894 
6895 	fr_getstat(&fio, ifs);
6896 	error = copyoutptr(&fio, data, sizeof(fio));
6897 	if (error)
6898 		return EFAULT;
6899 
6900 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
6901 	bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6902 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6903 
6904 	return 0;
6905 }
6906 
6907 
6908 #ifdef _KERNEL
6909 /* ------------------------------------------------------------------------ */
6910 /* Function:    fr_resolvedest                                              */
6911 /* Returns:     Nil                                                         */
6912 /* Parameters:  fdp(IO) - pointer to destination information to resolve     */
6913 /*              v(I)    - IP protocol version to match                      */
6914 /*                                                                          */
6915 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6916 /* if a matching name can be found for the particular IP protocol version   */
6917 /* then store the interface pointer in the frdest struct.  If no match is   */
6918 /* found, then set the interface pointer to be -1 as NULL is considered to  */
6919 /* indicate there is no information at all in the structure.                */
6920 /* ------------------------------------------------------------------------ */
6921 void fr_resolvedest(fdp, v, ifs)
6922 frdest_t *fdp;
6923 int v;
6924 ipf_stack_t *ifs;
6925 {
6926 	fdp->fd_ifp = NULL;
6927 
6928   	if (*fdp->fd_ifname != '\0') {
6929  		fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6930 		if (fdp->fd_ifp == NULL)
6931 			fdp->fd_ifp = (void *)-1;
6932 	}
6933 }
6934 #endif /* _KERNEL */
6935 
6936 
6937 /* ------------------------------------------------------------------------ */
6938 /* Function:    fr_resolvenic                                               */
6939 /* Returns:     void* - NULL = wildcard name, -1 = failed to find NIC, else */
6940 /*                      pointer to interface structure for NIC              */
6941 /* Parameters:  name(I) - complete interface name                           */
6942 /*              v(I)    - IP protocol version                               */
6943 /*                                                                          */
6944 /* Look for a network interface structure that firstly has a matching name  */
6945 /* to that passed in and that is also being used for that IP protocol       */
6946 /* version (necessary on some platforms where there are separate listings   */
6947 /* for both IPv4 and IPv6 on the same physical NIC.                         */
6948 /*                                                                          */
6949 /* One might wonder why name gets terminated with a \0 byte in here.  The   */
6950 /* reason is an interface name could get into the kernel structures of ipf  */
6951 /* in any number of ways and so long as they all use the same sized array   */
6952 /* to put the name in, it makes sense to ensure it gets null terminated     */
6953 /* before it is used for its intended purpose - finding its match in the    */
6954 /* kernel's list of configured interfaces.                                  */
6955 /*                                                                          */
6956 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an     */
6957 /*       array for the name that is LIFNAMSIZ bytes (at least) in length.   */
6958 /* ------------------------------------------------------------------------ */
6959 void *fr_resolvenic(name, v, ifs)
6960 char *name;
6961 int v;
6962 ipf_stack_t *ifs;
6963 {
6964 	void *nic;
6965 
6966 	if (name[0] == '\0')
6967 		return NULL;
6968 
6969 	if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
6970 		return NULL;
6971 	}
6972 
6973 	name[LIFNAMSIZ - 1] = '\0';
6974 
6975 	nic = GETIFP(name, v, ifs);
6976 	if (nic == NULL)
6977 		nic = (void *)-1;
6978 	return nic;
6979 }
6980 
6981 
6982 /* ------------------------------------------------------------------------ */
6983 /* Function:    ipf_expiretokens                                            */
6984 /* Returns:     None.                                                       */
6985 /* Parameters:  ifs - ipf stack instance                                    */
6986 /*                                                                          */
6987 /* This function is run every ipf tick to see if there are any tokens that  */
6988 /* have been held for too long and need to be freed up.                     */
6989 /* ------------------------------------------------------------------------ */
6990 void ipf_expiretokens(ifs)
6991 ipf_stack_t *ifs;
6992 {
6993 	ipftoken_t *it;
6994 
6995 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
6996 	while ((it = ifs->ifs_ipftokenhead) != NULL) {
6997 		if (it->ipt_die > ifs->ifs_fr_ticks)
6998 			break;
6999 
7000 		ipf_freetoken(it, ifs);
7001 	}
7002 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7003 }
7004 
7005 
7006 /* ------------------------------------------------------------------------ */
7007 /* Function:    ipf_deltoken                                                */
7008 /* Returns:     int - 0 = success, else error                               */
7009 /* Parameters:  type(I) - the token type to match                           */
7010 /*              uid(I)  - uid owning the token                              */
7011 /*              ptr(I)  - context pointer for the token                     */
7012 /*              ifs - ipf stack instance                                    */
7013 /*                                                                          */
7014 /* This function looks for a a token in the current list that matches up    */
7015 /* the fields (type, uid, ptr).  If none is found, ESRCH is returned, else  */
7016 /* call ipf_freetoken() to remove it from the list.                         */
7017 /* ------------------------------------------------------------------------ */
7018 int ipf_deltoken(type, uid, ptr, ifs)
7019 int type, uid;
7020 void *ptr;
7021 ipf_stack_t *ifs;
7022 {
7023 	ipftoken_t *it;
7024 	int error = ESRCH;
7025 
7026 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7027 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
7028 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7029 		    uid == it->ipt_uid) {
7030 			ipf_freetoken(it, ifs);
7031 			error = 0;
7032 			break;
7033 	}
7034 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7035 
7036 	return error;
7037 }
7038 
7039 
7040 /* ------------------------------------------------------------------------ */
7041 /* Function:    ipf_unlinktoken                                             */
7042 /* Returns:     None.                                                       */
7043 /* Parameters:  token(I) - pointer to token structure                       */
7044 /*              ifs - ipf stack instance                                    */
7045 /*                                                                          */
7046 /* This function unlinks a token structure from the linked list of tokens   */
7047 /* that it belongs to.  The head pointer never needs to be explicitly       */
7048 /* adjusted, but the tail does due to the linked list implementation.       */
7049 /* ------------------------------------------------------------------------ */
7050 static void ipf_unlinktoken(token, ifs)
7051 ipftoken_t *token;
7052 ipf_stack_t *ifs;
7053 {
7054 
7055 	if (ifs->ifs_ipftokentail == &token->ipt_next)
7056 		ifs->ifs_ipftokentail = token->ipt_pnext;
7057 
7058 	*token->ipt_pnext = token->ipt_next;
7059 	if (token->ipt_next != NULL)
7060 		token->ipt_next->ipt_pnext = token->ipt_pnext;
7061 }
7062 
7063 
7064 /* ------------------------------------------------------------------------ */
7065 /* Function:    ipf_findtoken                                               */
7066 /* Returns:     ipftoken_t * - NULL if no memory, else pointer to token     */
7067 /* Parameters:  type(I) - the token type to match                           */
7068 /*              uid(I) - uid owning the token                               */
7069 /*              ptr(I) - context pointer for the token                      */
7070 /*              ifs - ipf stack instance                                    */
7071 /*                                                                          */
7072 /* This function looks for a live token in the list of current tokens that  */
7073 /* matches the tuple (type, uid, ptr).  If one cannot be found then one is  */
7074 /* allocated.  If one is found then it is moved to the top of the list of   */
7075 /* currently active tokens.                                                 */
7076 /*                                                                          */
7077 /* NOTE: It is by design that this function returns holding a read lock on  */
7078 /*       ipf_tokens.  Callers must make sure they release it!               */
7079 /* ------------------------------------------------------------------------ */
7080 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
7081 int type, uid;
7082 void *ptr;
7083 ipf_stack_t *ifs;
7084 {
7085 	ipftoken_t *it, *new;
7086 
7087 	KMALLOC(new, ipftoken_t *);
7088 
7089 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7090 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
7091 		if (it->ipt_alive == 0)
7092 			continue;
7093 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7094 		    uid == it->ipt_uid)
7095 			break;
7096 	}
7097 
7098 	if (it == NULL) {
7099 		it = new;
7100 		new = NULL;
7101 		if (it == NULL)
7102 			return NULL;
7103 		it->ipt_data = NULL;
7104 		it->ipt_ctx = ptr;
7105 		it->ipt_uid = uid;
7106 		it->ipt_type = type;
7107 		it->ipt_next = NULL;
7108 		it->ipt_alive = 1;
7109 	} else {
7110 		if (new != NULL) {
7111 			KFREE(new);
7112 			new = NULL;
7113 		}
7114 
7115 		ipf_unlinktoken(it, ifs);
7116 	}
7117 	it->ipt_pnext = ifs->ifs_ipftokentail;
7118 	*ifs->ifs_ipftokentail = it;
7119 	ifs->ifs_ipftokentail = &it->ipt_next;
7120 	it->ipt_next = NULL;
7121 
7122 	it->ipt_die = ifs->ifs_fr_ticks + 2;
7123 
7124 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
7125 
7126 	return it;
7127 }
7128 
7129 
7130 /* ------------------------------------------------------------------------ */
7131 /* Function:    ipf_freetoken                                               */
7132 /* Returns:     None.                                                       */
7133 /* Parameters:  token(I) - pointer to token structure                       */
7134 /*              ifs - ipf stack instance                                    */
7135 /*                                                                          */
7136 /* This function unlinks a token from the linked list and on the path to    */
7137 /* free'ing the data, it calls the dereference function that is associated  */
7138 /* with the type of data pointed to by the token as it is considered to     */
7139 /* hold a reference to it.                                                  */
7140 /* ------------------------------------------------------------------------ */
7141 void ipf_freetoken(token, ifs)
7142 ipftoken_t *token;
7143 ipf_stack_t *ifs;
7144 {
7145 	void *data, **datap;
7146 
7147 	ipf_unlinktoken(token, ifs);
7148 
7149 	data = token->ipt_data;
7150 	datap = &data;
7151 
7152 	if ((data != NULL) && (data != (void *)-1)) {
7153 		switch (token->ipt_type)
7154 		{
7155 		case IPFGENITER_IPF :
7156 			(void)fr_derefrule((frentry_t **)datap, ifs);
7157 			break;
7158 		case IPFGENITER_IPNAT :
7159 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7160 			fr_ipnatderef((ipnat_t **)datap, ifs);
7161 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7162 			break;
7163 		case IPFGENITER_NAT :
7164 			fr_natderef((nat_t **)datap, ifs);
7165 			break;
7166 		case IPFGENITER_STATE :
7167 			fr_statederef((ipstate_t **)datap, ifs);
7168 			break;
7169 		case IPFGENITER_FRAG :
7170 			fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
7171 			break;
7172 		case IPFGENITER_NATFRAG :
7173  			fr_fragderef((ipfr_t **)datap,
7174 				     &ifs->ifs_ipf_natfrag, ifs);
7175 			break;
7176 		case IPFGENITER_HOSTMAP :
7177 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7178 			fr_hostmapdel((hostmap_t **)datap);
7179 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7180 			break;
7181 		default :
7182 			(void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7183 			break;
7184 		}
7185 	}
7186 
7187 	KFREE(token);
7188 }
7189 
7190 
7191 /* ------------------------------------------------------------------------ */
7192 /* Function:    ipf_getnextrule                                             */
7193 /* Returns:     int - 0 = success, else error                               */
7194 /* Parameters:  t(I)   - pointer to destination information to resolve      */
7195 /*              ptr(I) - pointer to ipfobj_t to copyin from user space      */
7196 /*              ifs - ipf stack instance                                    */
7197 /*                                                                          */
7198 /* This function's first job is to bring in the ipfruleiter_t structure via */
7199 /* the ipfobj_t structure to determine what should be the next rule to      */
7200 /* return. Once the ipfruleiter_t has been brought in, it then tries to     */
7201 /* find the 'next rule'.  This may include searching rule group lists or    */
7202 /* just be as simple as looking at the 'next' field in the rule structure.  */
7203 /* When we have found the rule to return, increase its reference count and  */
7204 /* if we used an existing rule to get here, decrease its reference count.   */
7205 /* ------------------------------------------------------------------------ */
7206 int ipf_getnextrule(t, ptr, ifs)
7207 ipftoken_t *t;
7208 void *ptr;
7209 ipf_stack_t *ifs;
7210 {
7211 	frentry_t *fr, *next, zero;
7212 	int error, out, count;
7213 	ipfruleiter_t it;
7214 	frgroup_t *fg;
7215 	char *dst;
7216 
7217 	if (t == NULL || ptr == NULL)
7218 		return EFAULT;
7219 	error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7220 	if (error != 0)
7221 		return error;
7222 	if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7223 		return EINVAL;
7224 	if ((it.iri_inout < 0) || (it.iri_inout > 3))
7225 		return EINVAL;
7226 	if (it.iri_nrules == 0)
7227 		return EINVAL;
7228 	if ((it.iri_active != 0) && (it.iri_active != 1))
7229 		return EINVAL;
7230 	if (it.iri_rule == NULL)
7231 		return EFAULT;
7232 
7233 	/*
7234 	 * Use bitmask on it.iri_inout to determine direction.
7235 	 * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7236 	 * F_IN (0) and F_ACIN (2) mask to out = 0.
7237 	 */
7238 	out = it.iri_inout & F_OUT;
7239 	READ_ENTER(&ifs->ifs_ipf_mutex);
7240 
7241 	/*
7242 	 * Retrieve "previous" entry from token and find the next entry.
7243 	 */
7244 	fr = t->ipt_data;
7245 	if (fr == NULL) {
7246 		if (*it.iri_group == '\0') {
7247 			/*
7248 			 * Use bitmask again to determine accounting or not.
7249 			 * F_ACIN will mask to accounting cases F_ACIN (2)
7250 			 * or F_ACOUT (3), but not F_IN or F_OUT.
7251 			 */
7252 			if ((it.iri_inout & F_ACIN) != 0) {
7253 				if (it.iri_ver == AF_INET)
7254 					next = ifs->ifs_ipacct
7255 					    [out][it.iri_active];
7256 				else
7257 					next = ifs->ifs_ipacct6
7258 					    [out][it.iri_active];
7259 			} else {
7260 				if (it.iri_ver == AF_INET)
7261 					next = ifs->ifs_ipfilter
7262 					    [out][it.iri_active];
7263 				else
7264 					next = ifs->ifs_ipfilter6
7265 					    [out][it.iri_active];
7266 			}
7267 		} else {
7268 			fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7269 					  it.iri_active, NULL, ifs);
7270 			if (fg != NULL)
7271 				next = fg->fg_start;
7272 			else
7273 				next = NULL;
7274 		}
7275 	} else {
7276 		next = fr->fr_next;
7277 	}
7278 
7279 	dst = (char *)it.iri_rule;
7280 	/*
7281 	 * The ipfruleiter may ask for more than 1 rule at a time to be
7282 	 * copied out, so long as that many exist in the list to start with!
7283 	 */
7284 	for (count = it.iri_nrules; count > 0; count--) {
7285 		/*
7286 		 * If we found an entry, add reference to it and update token.
7287 		 * Otherwise, zero out data to be returned and NULL out token.
7288 		 */
7289 		if (next != NULL) {
7290 			MUTEX_ENTER(&next->fr_lock);
7291 			next->fr_ref++;
7292 			MUTEX_EXIT(&next->fr_lock);
7293 			t->ipt_data = next;
7294 		} else {
7295 			bzero(&zero, sizeof(zero));
7296 			next = &zero;
7297 			t->ipt_data = NULL;
7298 		}
7299 
7300 		/*
7301 		 * Now that we have ref, it's save to give up lock.
7302 		 */
7303 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7304 
7305 		/*
7306 		 * Copy out data and clean up references and token as needed.
7307 		 */
7308 		error = COPYOUT(next, dst, sizeof(*next));
7309 		if (error != 0)
7310 			error = EFAULT;
7311 		if (t->ipt_data == NULL) {
7312 			ipf_freetoken(t, ifs);
7313 			break;
7314 		} else {
7315 			if (fr != NULL)
7316 				(void) fr_derefrule(&fr, ifs);
7317 			if (next->fr_data != NULL) {
7318 				dst += sizeof(*next);
7319 				error = COPYOUT(next->fr_data, dst,
7320 						next->fr_dsize);
7321 				if (error != 0)
7322 					error = EFAULT;
7323 				else
7324 					dst += next->fr_dsize;
7325 			}
7326 			if (next->fr_next == NULL) {
7327 				ipf_freetoken(t, ifs);
7328 				break;
7329 			}
7330 		}
7331 
7332 		if ((count == 1) || (error != 0))
7333 			break;
7334 
7335 		READ_ENTER(&ifs->ifs_ipf_mutex);
7336 		fr = next;
7337 		next = fr->fr_next;
7338 	}
7339 
7340 	return error;
7341 }
7342 
7343 
7344 /* ------------------------------------------------------------------------ */
7345 /* Function:    fr_frruleiter                                               */
7346 /* Returns:     int - 0 = success, else error                               */
7347 /* Parameters:  data(I) - the token type to match                           */
7348 /*              uid(I) - uid owning the token                               */
7349 /*              ptr(I) - context pointer for the token                      */
7350 /*              ifs - ipf stack instance                                    */
7351 /*                                                                          */
7352 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7353 /* ipf_getnextrule.  It's role is to find the right token in the kernel for */
7354 /* the process doing the ioctl and use that to ask for the next rule.       */
7355 /* ------------------------------------------------------------------------ */
7356 int ipf_frruleiter(data, uid, ctx, ifs)
7357 void *data, *ctx;
7358 int uid;
7359 ipf_stack_t *ifs;
7360 {
7361 	ipftoken_t *token;
7362 	int error;
7363 
7364 	token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7365 	if (token != NULL)
7366 		error = ipf_getnextrule(token, data, ifs);
7367 	else
7368 		error = EFAULT;
7369 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7370 
7371 	return error;
7372 }
7373 
7374 
7375 /* ------------------------------------------------------------------------ */
7376 /* Function:    ipf_geniter                                                 */
7377 /* Returns:     int - 0 = success, else error                               */
7378 /* Parameters:  token(I) - pointer to ipftoken structure                    */
7379 /*              itp(I) - pointer to ipfgeniter structure                    */
7380 /*              ifs - ipf stack instance                                    */
7381 /*                                                                          */
7382 /* Generic iterator called from ipf_genericiter.  Currently only used for   */
7383 /* walking through list of fragments.                                       */
7384 /* ------------------------------------------------------------------------ */
7385 int ipf_geniter(token, itp, ifs)
7386 ipftoken_t *token;
7387 ipfgeniter_t *itp;
7388 ipf_stack_t *ifs;
7389 {
7390 	int error;
7391 
7392 	switch (itp->igi_type)
7393 	{
7394 	case IPFGENITER_FRAG :
7395 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7396 				    &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7397 				    ifs);
7398 		break;
7399 	default :
7400 		error = EINVAL;
7401 		break;
7402 	}
7403 
7404 	return error;
7405 }
7406 
7407 
7408 /* ------------------------------------------------------------------------ */
7409 /* Function:    ipf_genericiter                                             */
7410 /* Returns:     int - 0 = success, else error                               */
7411 /* Parameters:  data(I) - the token type to match                           */
7412 /*              uid(I) - uid owning the token                               */
7413 /*              ptr(I) - context pointer for the token                      */
7414 /*              ifs - ipf stack instance                                    */
7415 /*                                                                          */
7416 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7417 /* ipf_geniter when handling SIOCGENITER.  It's role is to find the right   */
7418 /* token in the kernel for the process using the ioctl, and to use that     */
7419 /* token when calling ipf_geniter.                                          */
7420 /* ------------------------------------------------------------------------ */
7421 int ipf_genericiter(data, uid, ctx, ifs)
7422 void *data, *ctx;
7423 int uid;
7424 ipf_stack_t *ifs;
7425 {
7426 	ipftoken_t *token;
7427 	ipfgeniter_t iter;
7428 	int error;
7429 
7430 	error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7431 	if (error != 0)
7432 		return error;
7433 
7434 	token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7435 	if (token != NULL) {
7436 		token->ipt_subtype = iter.igi_type;
7437 		error = ipf_geniter(token, &iter, ifs);
7438 	} else
7439 		error = EFAULT;
7440 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7441 
7442 	return error;
7443 }
7444 
7445 
7446 /* --------------------------------------------------------------------- */
7447 /* Function:    ipf_earlydrop                                            */
7448 /* Returns:     number of dropped/removed entries from the queue         */
7449 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7450 /*              ifq	- pointer to queue with entries to be deleted    */
7451 /*              idletime - entry must be idle this long to be deleted    */
7452 /*              ifs     - ipf stack instance                             */
7453 /*                                                                       */
7454 /* Function is invoked from state/NAT flush routines to remove entries   */
7455 /* from specified timeout queue, based on how long they've sat idle,     */
7456 /* without waiting for it to happen on its own.                          */
7457 /* --------------------------------------------------------------------- */
7458 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7459 int flushtype;
7460 ipftq_t *ifq;
7461 int idletime;
7462 ipf_stack_t *ifs;
7463 {
7464         ipftqent_t *tqe, *tqn;
7465         unsigned int dropped;
7466         int droptick;
7467 	void *ent;
7468 
7469         if (ifq == NULL)
7470                 return (0);
7471 
7472         dropped = 0;
7473 
7474         /*
7475          * Determine the tick representing the idle time we're interested
7476          * in.  If an entry exists in the queue, and it was touched before
7477          * that tick, then it's been idle longer than idletime, so it should
7478 	 * be deleted.
7479          */
7480         droptick = ifs->ifs_fr_ticks - idletime;
7481         tqn = ifq->ifq_head;
7482         while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7483                 tqn = tqe->tqe_next;
7484 		ent = tqe->tqe_parent;
7485 		switch (flushtype)
7486 		{
7487 		case NAT_FLUSH:
7488 			if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7489 				dropped++;
7490 			break;
7491 		case STATE_FLUSH:
7492 			if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7493 				dropped++;
7494 			break;
7495 		default:
7496 			return (0);
7497 		}
7498         }
7499         return (dropped);
7500 }
7501 
7502 
7503 /* --------------------------------------------------------------------- */
7504 /* Function:    ipf_flushclosing                                         */
7505 /* Returns:     int - number of entries deleted                          */
7506 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7507 /*              stateval - TCP state at which to start removing entries  */
7508 /*              ipfqs - pointer to timeout queues                        */
7509 /*              userqs - pointer to user defined queues                  */
7510 /*              ifs  - ipf stack instance                                */
7511 /*                                                                       */
7512 /* Remove state/NAT table entries for TCP connections which are in the   */
7513 /* process of closing, and have at least reached the state specified by  */
7514 /* the 'stateval' parameter.                                             */
7515 /* --------------------------------------------------------------------- */
7516 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7517 int flushtype, stateval;
7518 ipftq_t *ipfqs, *userqs;
7519 ipf_stack_t *ifs;
7520 {
7521 	ipftq_t *ifq, *ifqn;
7522         ipftqent_t *tqe, *tqn;
7523         int dropped;
7524 	void *ent;
7525 	nat_t *nat;
7526 	ipstate_t *is;
7527 
7528         dropped = 0;
7529 
7530         /*
7531          * Start by deleting any entries in specific timeout queues.
7532          */
7533 	ifqn = &ipfqs[stateval];
7534         while ((ifq = ifqn) != NULL) {
7535                 ifqn = ifq->ifq_next;
7536                 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7537         }
7538 
7539         /*
7540          * Next, look through user defined queues for closing entries.
7541          */
7542 	ifqn = userqs;
7543         while ((ifq = ifqn) != NULL) {
7544                 ifqn = ifq->ifq_next;
7545                 tqn = ifq->ifq_head;
7546                 while ((tqe = tqn) != NULL) {
7547                         tqn = tqe->tqe_next;
7548 			ent = tqe->tqe_parent;
7549 			switch (flushtype)
7550 			{
7551 			case NAT_FLUSH:
7552 				nat = (nat_t *)ent;
7553 				if ((nat->nat_p == IPPROTO_TCP) &&
7554 				    (nat->nat_tcpstate[0] >= stateval) &&
7555 				    (nat->nat_tcpstate[1] >= stateval) &&
7556 				    (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7557 					dropped++;
7558 				break;
7559 			case STATE_FLUSH:
7560 				is = (ipstate_t *)ent;
7561 				if ((is->is_p == IPPROTO_TCP) &&
7562 				    (is->is_state[0] >= stateval) &&
7563 				    (is->is_state[1] >= stateval) &&
7564 				    (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7565 					dropped++;
7566 				break;
7567 			default:
7568 				return (0);
7569 			}
7570                 }
7571         }
7572         return (dropped);
7573 }
7574 
7575 
7576 /* --------------------------------------------------------------------- */
7577 /* Function:    ipf_extraflush                                           */
7578 /* Returns:     int - number of entries flushed (0 = none)               */
7579 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7580 /*              ipfqs - pointer to 'established' timeout queue           */
7581 /*              userqs - pointer to user defined queues                  */
7582 /*              ifs  - ipf stack instance                                */
7583 /*                                                                       */
7584 /* This function gets called when either NAT or state tables fill up.    */
7585 /* We need to try a bit harder to free up some space.  The function will */
7586 /* flush entries for TCP connections which have been idle a long time.   */
7587 /*                                                                       */
7588 /* Currently, the idle time is checked using values from ideltime_tab[]	 */
7589 /* --------------------------------------------------------------------- */
7590 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7591 int flushtype;
7592 ipftq_t *ipfqs, *userqs;
7593 ipf_stack_t *ifs;
7594 {
7595 	ipftq_t *ifq, *ifqn;
7596 	int idletime, removed, idle_idx;
7597 
7598 	removed = 0;
7599 
7600 	/*
7601   	 * Determine initial threshold for minimum idle time based on
7602 	 * how long ipfilter has been running.  Ipfilter needs to have
7603 	 * been up as long as the smallest interval to continue on.
7604 	 *
7605 	 * Minimum idle times stored in idletime_tab and indexed by
7606 	 * idle_idx.  Start at upper end of array and work backwards.
7607 	 *
7608 	 * Once the index is found, set the initial idle time to the
7609 	 * first interval before the current ipfilter run time.
7610 	 */
7611 	if (ifs->ifs_fr_ticks < idletime_tab[0])
7612 		return (0);
7613 	idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7614 	if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7615 		idletime = idletime_tab[idle_idx];
7616 	} else {
7617 		while ((idle_idx > 0) &&
7618 		    (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7619 			idle_idx--;
7620 
7621 		idletime = (ifs->ifs_fr_ticks /
7622 			    idletime_tab[idle_idx]) *
7623 			    idletime_tab[idle_idx];
7624 	}
7625 
7626 	while (idle_idx >= 0) {
7627 		/*
7628 		 * Check to see if we need to delete more entries.
7629 		 * If we do, start with appropriate timeout queue.
7630 		 */
7631 		if (flushtype == NAT_FLUSH) {
7632 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7633 			    ifs->ifs_nat_flush_level_lo)
7634 				break;
7635 		} else if (flushtype == STATE_FLUSH) {
7636 			if (ST_TAB_WATER_LEVEL(ifs) <=
7637 			    ifs->ifs_state_flush_level_lo)
7638 				break;
7639 		} else {
7640 			break;
7641 		}
7642 
7643 		removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7644 
7645 		/*
7646 		 * Next, check the user defined queues.  But first, make
7647 		 * certain that timeout queue deletions didn't do enough.
7648 		 */
7649 		if (flushtype == NAT_FLUSH) {
7650 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7651 			    ifs->ifs_nat_flush_level_lo)
7652 				break;
7653 		} else {
7654 			if (ST_TAB_WATER_LEVEL(ifs) <=
7655 			    ifs->ifs_state_flush_level_lo)
7656 				break;
7657 		}
7658 		ifqn = userqs;
7659 		while ((ifq = ifqn) != NULL) {
7660 			ifqn = ifq->ifq_next;
7661 			removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7662 		}
7663 
7664 		/*
7665 		 * Adjust the granularity of idle time.
7666 		 *
7667 		 * If we reach an interval boundary, we need to
7668 		 * either adjust the idle time accordingly or exit
7669 		 * the loop altogether (if this is very last check).
7670 		 */
7671 		idletime -= idletime_tab[idle_idx];
7672 		if (idletime < idletime_tab[idle_idx]) {
7673 			if (idle_idx != 0) {
7674 				idletime = idletime_tab[idle_idx] -
7675 				idletime_tab[idle_idx - 1];
7676 				idle_idx--;
7677 			} else {
7678 				break;
7679 			}
7680 		}
7681 	}
7682 
7683 	return (removed);
7684 }
7685