xref: /titanic_50/usr/src/uts/common/inet/ipf/fil.c (revision 6e0414acac7fe81ff262fb9d3d83c0700fe9b695)
1 /*
2  * Copyright (C) 1993-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
7  * Use is subject to license terms.
8  */
9 
10 #if defined(KERNEL) || defined(_KERNEL)
11 # undef KERNEL
12 # undef _KERNEL
13 # define        KERNEL	1
14 # define        _KERNEL	1
15 #endif
16 #include <sys/errno.h>
17 #include <sys/types.h>
18 #include <sys/param.h>
19 #include <sys/time.h>
20 #if defined(__NetBSD__)
21 # if (NetBSD >= 199905) && !defined(IPFILTER_LKM) && defined(_KERNEL)
22 #  include "opt_ipfilter_log.h"
23 # endif
24 #endif
25 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
26     (__FreeBSD_version >= 220000)
27 # if (__FreeBSD_version >= 400000)
28 #  if !defined(IPFILTER_LKM)
29 #   include "opt_inet6.h"
30 #  endif
31 #  if (__FreeBSD_version == 400019)
32 #   define CSUM_DELAY_DATA
33 #  endif
34 # endif
35 # include <sys/filio.h>
36 #else
37 # include <sys/ioctl.h>
38 #endif
39 #if !defined(_AIX51)
40 # include <sys/fcntl.h>
41 #endif
42 #if defined(_KERNEL)
43 # include <sys/systm.h>
44 # include <sys/file.h>
45 #else
46 # include <stdio.h>
47 # include <string.h>
48 # include <stdlib.h>
49 # include <stddef.h>
50 # include <sys/file.h>
51 # define _KERNEL
52 # ifdef __OpenBSD__
53 struct file;
54 # endif
55 # include <sys/uio.h>
56 # undef _KERNEL
57 #endif
58 #if !defined(__SVR4) && !defined(__svr4__) && !defined(__hpux) && \
59     !defined(linux)
60 # include <sys/mbuf.h>
61 #else
62 # if !defined(linux)
63 #  include <sys/byteorder.h>
64 # endif
65 # if (SOLARIS2 < 5) && defined(sun)
66 #  include <sys/dditypes.h>
67 # endif
68 #endif
69 #ifdef __hpux
70 # define _NET_ROUTE_INCLUDED
71 #endif
72 #if !defined(linux)
73 # include <sys/protosw.h>
74 #endif
75 #include <sys/socket.h>
76 #include <net/if.h>
77 #ifdef sun
78 # include <net/af.h>
79 #endif
80 #if !defined(_KERNEL) && defined(__FreeBSD__)
81 # include "radix_ipf.h"
82 #endif
83 #include <net/route.h>
84 #include <netinet/in.h>
85 #include <netinet/in_systm.h>
86 #include <netinet/ip.h>
87 #if !defined(linux)
88 # include <netinet/ip_var.h>
89 #endif
90 #if defined(__sgi) && defined(IFF_DRVRLOCK) /* IRIX 6 */
91 # include <sys/hashing.h>
92 # include <netinet/in_var.h>
93 #endif
94 #include <netinet/tcp.h>
95 #if (!defined(__sgi) && !defined(AIX)) || defined(_KERNEL)
96 # include <netinet/udp.h>
97 # include <netinet/ip_icmp.h>
98 #endif
99 #ifdef __hpux
100 # undef _NET_ROUTE_INCLUDED
101 #endif
102 #include "netinet/ip_compat.h"
103 #ifdef	USE_INET6
104 # include <netinet/icmp6.h>
105 # if !SOLARIS && defined(_KERNEL) && !defined(__osf__) && !defined(__hpux)
106 #  include <netinet6/in6_var.h>
107 # endif
108 #endif
109 #include <netinet/tcpip.h>
110 #include "netinet/ip_fil.h"
111 #include "netinet/ip_nat.h"
112 #include "netinet/ip_frag.h"
113 #include "netinet/ip_state.h"
114 #include "netinet/ip_proxy.h"
115 #include "netinet/ip_auth.h"
116 #include "netinet/ipf_stack.h"
117 #ifdef IPFILTER_SCAN
118 # include "netinet/ip_scan.h"
119 #endif
120 #ifdef IPFILTER_SYNC
121 # include "netinet/ip_sync.h"
122 #endif
123 #include "netinet/ip_pool.h"
124 #include "netinet/ip_htable.h"
125 #ifdef IPFILTER_COMPILED
126 # include "netinet/ip_rules.h"
127 #endif
128 #if defined(IPFILTER_BPF) && defined(_KERNEL)
129 # include <net/bpf.h>
130 #endif
131 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
132 # include <sys/malloc.h>
133 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
134 #  include "opt_ipfilter.h"
135 # endif
136 #endif
137 #include "netinet/ipl.h"
138 /* END OF INCLUDES */
139 
140 #if !defined(lint)
141 static const char sccsid[] = "@(#)fil.c	1.36 6/5/96 (C) 1993-2000 Darren Reed";
142 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
143 #endif
144 
145 #ifndef	_KERNEL
146 # include "ipf.h"
147 # include "ipt.h"
148 # include "bpf-ipf.h"
149 extern	int	opts;
150 
151 # define	FR_VERBOSE(verb_pr)			verbose verb_pr
152 # define	FR_DEBUG(verb_pr)			debug verb_pr
153 #else /* #ifndef _KERNEL */
154 # define	FR_VERBOSE(verb_pr)
155 # define	FR_DEBUG(verb_pr)
156 #endif /* _KERNEL */
157 
158 
159 char	ipfilter_version[] = IPL_VERSION;
160 int	fr_features = 0
161 #ifdef	IPFILTER_LKM
162 		| IPF_FEAT_LKM
163 #endif
164 #ifdef	IPFILTER_LOG
165 		| IPF_FEAT_LOG
166 #endif
167 #ifdef	IPFILTER_LOOKUP
168 		| IPF_FEAT_LOOKUP
169 #endif
170 #ifdef	IPFILTER_BPF
171 		| IPF_FEAT_BPF
172 #endif
173 #ifdef	IPFILTER_COMPILED
174 		| IPF_FEAT_COMPILED
175 #endif
176 #ifdef	IPFILTER_CKSUM
177 		| IPF_FEAT_CKSUM
178 #endif
179 #ifdef	IPFILTER_SYNC
180 		| IPF_FEAT_SYNC
181 #endif
182 #ifdef	IPFILTER_SCAN
183 		| IPF_FEAT_SCAN
184 #endif
185 #ifdef	USE_INET6
186 		| IPF_FEAT_IPV6
187 #endif
188 	;
189 
190 #define	IPF_BUMP(x)	(x)++
191 
192 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
193 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
194 static	int		fr_portcheck __P((frpcmp_t *, u_short *));
195 static	int		frflushlist __P((int, minor_t, int *, frentry_t **,
196 					 ipf_stack_t *));
197 static	ipfunc_t	fr_findfunc __P((ipfunc_t));
198 static	frentry_t	*fr_firewall __P((fr_info_t *, u_32_t *));
199 static	int		fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
200 static	INLINE void	frpr_ah __P((fr_info_t *));
201 static	INLINE void	frpr_esp __P((fr_info_t *));
202 static	INLINE void	frpr_gre __P((fr_info_t *));
203 static	INLINE void	frpr_udp __P((fr_info_t *));
204 static	INLINE void	frpr_tcp __P((fr_info_t *));
205 static	INLINE void	frpr_icmp __P((fr_info_t *));
206 static	INLINE void	frpr_ipv4hdr __P((fr_info_t *));
207 static	INLINE int	frpr_pullup __P((fr_info_t *, int));
208 static	INLINE void	frpr_short __P((fr_info_t *, int));
209 static	INLINE void	frpr_tcpcommon __P((fr_info_t *));
210 static	INLINE void	frpr_udpcommon __P((fr_info_t *));
211 static	INLINE int	fr_updateipid __P((fr_info_t *));
212 #ifdef	IPFILTER_LOOKUP
213 static	int		fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
214 static	INLINE void	*fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
215 					       ipf_stack_t *));
216 #endif
217 static	void		frsynclist __P((int, int, void *, char *, frentry_t *,
218     ipf_stack_t *));
219 static	void		*fr_ifsync __P((int, int, char *, char *,
220 					void *, void *, ipf_stack_t *));
221 static	ipftuneable_t	*fr_findtunebyname __P((const char *, ipf_stack_t *));
222 static	ipftuneable_t	*fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
223 
224 /*
225  * bit values for identifying presence of individual IP options
226  * All of these tables should be ordered by increasing key value on the left
227  * hand side to allow for binary searching of the array and include a trailer
228  * with a 0 for the bitmask for linear searches to easily find the end with.
229  */
230 const	struct	optlist	ipopts[20] = {
231 	{ IPOPT_NOP,	0x000001 },
232 	{ IPOPT_RR,	0x000002 },
233 	{ IPOPT_ZSU,	0x000004 },
234 	{ IPOPT_MTUP,	0x000008 },
235 	{ IPOPT_MTUR,	0x000010 },
236 	{ IPOPT_ENCODE,	0x000020 },
237 	{ IPOPT_TS,	0x000040 },
238 	{ IPOPT_TR,	0x000080 },
239 	{ IPOPT_SECURITY, 0x000100 },
240 	{ IPOPT_LSRR,	0x000200 },
241 	{ IPOPT_E_SEC,	0x000400 },
242 	{ IPOPT_CIPSO,	0x000800 },
243 	{ IPOPT_SATID,	0x001000 },
244 	{ IPOPT_SSRR,	0x002000 },
245 	{ IPOPT_ADDEXT,	0x004000 },
246 	{ IPOPT_VISA,	0x008000 },
247 	{ IPOPT_IMITD,	0x010000 },
248 	{ IPOPT_EIP,	0x020000 },
249 	{ IPOPT_FINN,	0x040000 },
250 	{ 0,		0x000000 }
251 };
252 
253 #ifdef USE_INET6
254 struct optlist ip6exthdr[] = {
255 	{ IPPROTO_HOPOPTS,		0x000001 },
256 	{ IPPROTO_IPV6,			0x000002 },
257 	{ IPPROTO_ROUTING,		0x000004 },
258 	{ IPPROTO_FRAGMENT,		0x000008 },
259 	{ IPPROTO_ESP,			0x000010 },
260 	{ IPPROTO_AH,			0x000020 },
261 	{ IPPROTO_NONE,			0x000040 },
262 	{ IPPROTO_DSTOPTS,		0x000080 },
263 	{ 0,				0 }
264 };
265 #endif
266 
267 struct optlist tcpopts[] = {
268 	{ TCPOPT_NOP,			0x000001 },
269 	{ TCPOPT_MAXSEG,		0x000002 },
270 	{ TCPOPT_WINDOW,		0x000004 },
271 	{ TCPOPT_SACK_PERMITTED,	0x000008 },
272 	{ TCPOPT_SACK,			0x000010 },
273 	{ TCPOPT_TIMESTAMP,		0x000020 },
274 	{ 0,				0x000000 }
275 };
276 
277 /*
278  * bit values for identifying presence of individual IP security options
279  */
280 const	struct	optlist	secopt[8] = {
281 	{ IPSO_CLASS_RES4,	0x01 },
282 	{ IPSO_CLASS_TOPS,	0x02 },
283 	{ IPSO_CLASS_SECR,	0x04 },
284 	{ IPSO_CLASS_RES3,	0x08 },
285 	{ IPSO_CLASS_CONF,	0x10 },
286 	{ IPSO_CLASS_UNCL,	0x20 },
287 	{ IPSO_CLASS_RES2,	0x40 },
288 	{ IPSO_CLASS_RES1,	0x80 }
289 };
290 
291 
292 /*
293  * Table of functions available for use with call rules.
294  */
295 static ipfunc_resolve_t fr_availfuncs[] = {
296 #ifdef	IPFILTER_LOOKUP
297 	{ "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
298 	{ "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
299 #endif
300 	{ "", NULL }
301 };
302 
303 
304 /*
305  * Below we declare a list of constants used only by the ipf_extraflush()
306  * routine.  We are placing it here, instead of in ipf_extraflush() itself,
307  * because we want to make it visible to tools such as mdb, nm etc., so the
308  * values can easily be altered during debugging.
309  */
310 static	const	int	idletime_tab[] = {
311 	IPF_TTLVAL(30),		/* 30 seconds */
312 	IPF_TTLVAL(1800),	/* 30 minutes */
313 	IPF_TTLVAL(43200),	/* 12 hours */
314 	IPF_TTLVAL(345600),	/* 4 days */
315 };
316 
317 
318 /*
319  * The next section of code is a a collection of small routines that set
320  * fields in the fr_info_t structure passed based on properties of the
321  * current packet.  There are different routines for the same protocol
322  * for each of IPv4 and IPv6.  Adding a new protocol, for which there
323  * will "special" inspection for setup, is now more easily done by adding
324  * a new routine and expanding the frpr_ipinit*() function rather than by
325  * adding more code to a growing switch statement.
326  */
327 #ifdef USE_INET6
328 static	INLINE int	frpr_ah6 __P((fr_info_t *));
329 static	INLINE void	frpr_esp6 __P((fr_info_t *));
330 static	INLINE void	frpr_gre6 __P((fr_info_t *));
331 static	INLINE void	frpr_udp6 __P((fr_info_t *));
332 static	INLINE void	frpr_tcp6 __P((fr_info_t *));
333 static	INLINE void	frpr_icmp6 __P((fr_info_t *));
334 static	INLINE int	frpr_ipv6hdr __P((fr_info_t *));
335 static	INLINE void	frpr_short6 __P((fr_info_t *, int));
336 static	INLINE int	frpr_hopopts6 __P((fr_info_t *));
337 static	INLINE int	frpr_routing6 __P((fr_info_t *));
338 static	INLINE int	frpr_dstopts6 __P((fr_info_t *));
339 static	INLINE int	frpr_fragment6 __P((fr_info_t *));
340 static	INLINE int	frpr_ipv6exthdr __P((fr_info_t *, int, int));
341 
342 
343 /* ------------------------------------------------------------------------ */
344 /* Function:    frpr_short6                                                 */
345 /* Returns:     void                                                        */
346 /* Parameters:  fin(I) - pointer to packet information                      */
347 /*                                                                          */
348 /* IPv6 Only                                                                */
349 /* This is function enforces the 'is a packet too short to be legit' rule   */
350 /* for IPv6 and marks the packet with FI_SHORT if so.  See function comment */
351 /* for frpr_short() for more details.                                       */
352 /* ------------------------------------------------------------------------ */
353 static INLINE void frpr_short6(fin, xmin)
354 fr_info_t *fin;
355 int xmin;
356 {
357 
358 	if (fin->fin_dlen < xmin)
359 		fin->fin_flx |= FI_SHORT;
360 }
361 
362 
363 /* ------------------------------------------------------------------------ */
364 /* Function:    frpr_ipv6hdr                                                */
365 /* Returns:     int                                                         */
366 /* Parameters:  fin(I) - pointer to packet information                      */
367 /*                                                                          */
368 /* IPv6 Only                                                                */
369 /* Copy values from the IPv6 header into the fr_info_t struct and call the  */
370 /* per-protocol analyzer if it exists.                                      */
371 /* ------------------------------------------------------------------------ */
372 static INLINE int frpr_ipv6hdr(fin)
373 fr_info_t *fin;
374 {
375 	ip6_t *ip6 = (ip6_t *)fin->fin_ip;
376 	int p, go = 1, i, hdrcount;
377 	fr_ip_t *fi = &fin->fin_fi;
378 
379 	fin->fin_off = 0;
380 
381 	fi->fi_tos = 0;
382 	fi->fi_optmsk = 0;
383 	fi->fi_secmsk = 0;
384 	fi->fi_auth = 0;
385 
386 	p = ip6->ip6_nxt;
387 	fi->fi_ttl = ip6->ip6_hlim;
388 	fi->fi_src.in6 = ip6->ip6_src;
389 	fi->fi_dst.in6 = ip6->ip6_dst;
390 	fin->fin_id = 0;
391 
392 	hdrcount = 0;
393 	while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
394 		switch (p)
395 		{
396 		case IPPROTO_UDP :
397 			frpr_udp6(fin);
398 			go = 0;
399 			break;
400 
401 		case IPPROTO_TCP :
402 			frpr_tcp6(fin);
403 			go = 0;
404 			break;
405 
406 		case IPPROTO_ICMPV6 :
407 			frpr_icmp6(fin);
408 			go = 0;
409 			break;
410 
411 		case IPPROTO_GRE :
412 			frpr_gre6(fin);
413 			go = 0;
414 			break;
415 
416 		case IPPROTO_HOPOPTS :
417 			/*
418 			 * hop by hop ext header is only allowed
419 			 * right after IPv6 header.
420 			 */
421 			if (hdrcount != 0) {
422 				fin->fin_flx |= FI_BAD;
423 				p = IPPROTO_NONE;
424 			} else {
425 				p = frpr_hopopts6(fin);
426 			}
427 			break;
428 
429 		case IPPROTO_DSTOPTS :
430 			p = frpr_dstopts6(fin);
431 			break;
432 
433 		case IPPROTO_ROUTING :
434 			p = frpr_routing6(fin);
435 			break;
436 
437 		case IPPROTO_AH :
438 			p = frpr_ah6(fin);
439 			break;
440 
441 		case IPPROTO_ESP :
442 			frpr_esp6(fin);
443 			go = 0;
444 			break;
445 
446 		case IPPROTO_IPV6 :
447 			for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
448 				if (ip6exthdr[i].ol_val == p) {
449 					fin->fin_flx |= ip6exthdr[i].ol_bit;
450 					break;
451 				}
452 			go = 0;
453 			break;
454 
455 		case IPPROTO_NONE :
456 			go = 0;
457 			break;
458 
459 		case IPPROTO_FRAGMENT :
460 			p = frpr_fragment6(fin);
461 			if (fin->fin_off != 0)  /* Not the first frag */
462 				go = 0;
463 			break;
464 
465 		default :
466 			go = 0;
467 			break;
468 		}
469 		hdrcount++;
470 
471 		/*
472 		 * It is important to note that at this point, for the
473 		 * extension headers (go != 0), the entire header may not have
474 		 * been pulled up when the code gets to this point.  This is
475 		 * only done for "go != 0" because the other header handlers
476 		 * will all pullup their complete header.  The other indicator
477 		 * of an incomplete packet is that this was just an extension
478 		 * header.
479 		 */
480 		if ((go != 0) && (p != IPPROTO_NONE) &&
481 		    (frpr_pullup(fin, 0) == -1)) {
482 			p = IPPROTO_NONE;
483 			go = 0;
484 		}
485 	}
486 	fi->fi_p = p;
487 
488 	if (fin->fin_flx & FI_BAD)
489 		return -1;
490 
491 	return 0;
492 }
493 
494 
495 /* ------------------------------------------------------------------------ */
496 /* Function:    frpr_ipv6exthdr                                             */
497 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
498 /* Parameters:  fin(I)      - pointer to packet information                 */
499 /*              multiple(I) - flag indicating yes/no if multiple occurances */
500 /*                            of this extension header are allowed.         */
501 /*              proto(I)    - protocol number for this extension header     */
502 /*                                                                          */
503 /* IPv6 Only                                                                */
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 	int dlen;
621 
622 	fin->fin_flx |= FI_FRAG;
623 
624 	dlen = fin->fin_dlen;
625 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
626 		return IPPROTO_NONE;
627 
628 	if (frpr_pullup(fin, sizeof(*frag)) == -1)
629 		return IPPROTO_NONE;
630 
631 	frpr_short6(fin, sizeof(*frag));
632 
633 	if ((fin->fin_flx & FI_SHORT) != 0)
634 		return IPPROTO_NONE;
635 
636 	frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
637 	/*
638 	 * Fragment but no fragmentation info set?  Bad packet...
639 	 */
640 	if (frag->ip6f_offlg == 0) {
641 		fin->fin_flx |= FI_BAD;
642 		return IPPROTO_NONE;
643 	}
644 
645 	fin->fin_id = frag->ip6f_ident;
646 	fin->fin_off = frag->ip6f_offlg & IP6F_OFF_MASK;
647 	fin->fin_off = ntohs(fin->fin_off);
648 	if (fin->fin_off != 0)
649 		fin->fin_flx |= FI_FRAGBODY;
650 
651 	fin->fin_dp = (char *)frag + sizeof(*frag);
652 	fin->fin_dlen = dlen - sizeof(*frag);
653 
654 	/* length of hdrs(after frag hdr) + data */
655 	fin->fin_flen = fin->fin_dlen;
656 
657 	/*
658 	 * If the frag is not the last one and the payload length
659 	 * is not multiple of 8, it must be dropped.
660 	 */
661 	if ((frag->ip6f_offlg & IP6F_MORE_FRAG) && (dlen % 8)) {
662 		fin->fin_flx |= FI_BAD;
663 		return IPPROTO_NONE;
664 	}
665 
666 	return frag->ip6f_nxt;
667 }
668 
669 
670 /* ------------------------------------------------------------------------ */
671 /* Function:    frpr_dstopts6                                               */
672 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
673 /* Parameters:  fin(I) - pointer to packet information                      */
674 /*              nextheader(I) - stores next header value                    */
675 /*                                                                          */
676 /* IPv6 Only                                                                */
677 /* This is function checks pending destination options extension header     */
678 /* ------------------------------------------------------------------------ */
679 static INLINE int frpr_dstopts6(fin)
680 fr_info_t *fin;
681 {
682 	return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
683 }
684 
685 
686 /* ------------------------------------------------------------------------ */
687 /* Function:    frpr_icmp6                                                  */
688 /* Returns:     void                                                        */
689 /* Parameters:  fin(I) - pointer to packet information                      */
690 /*                                                                          */
691 /* IPv6 Only                                                                */
692 /* This routine is mainly concerned with determining the minimum valid size */
693 /* for an ICMPv6 packet.                                                    */
694 /* ------------------------------------------------------------------------ */
695 static INLINE void frpr_icmp6(fin)
696 fr_info_t *fin;
697 {
698 	int minicmpsz = sizeof(struct icmp6_hdr);
699 	struct icmp6_hdr *icmp6;
700 
701 	if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
702 		return;
703 
704 	if (fin->fin_dlen > 1) {
705 		icmp6 = fin->fin_dp;
706 
707 		fin->fin_data[0] = *(u_short *)icmp6;
708 
709 		if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
710 			fin->fin_flx |= FI_ICMPQUERY;
711 
712 		switch (icmp6->icmp6_type)
713 		{
714 		case ICMP6_ECHO_REPLY :
715 		case ICMP6_ECHO_REQUEST :
716 			if (fin->fin_dlen >= 6)
717 				fin->fin_data[1] = icmp6->icmp6_id;
718 			minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
719 			break;
720 		case ICMP6_DST_UNREACH :
721 		case ICMP6_PACKET_TOO_BIG :
722 		case ICMP6_TIME_EXCEEDED :
723 		case ICMP6_PARAM_PROB :
724 			if ((fin->fin_m != NULL) &&
725 			    (M_LEN(fin->fin_m) < fin->fin_plen)) {
726 				if (fr_coalesce(fin) != 1)
727 					return;
728 			}
729 			fin->fin_flx |= FI_ICMPERR;
730 			minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
731 			break;
732 		default :
733 			break;
734 		}
735 	}
736 
737 	frpr_short6(fin, minicmpsz);
738 	fin->fin_flen -= fin->fin_dlen - minicmpsz;
739 }
740 
741 
742 /* ------------------------------------------------------------------------ */
743 /* Function:    frpr_udp6                                                   */
744 /* Returns:     void                                                        */
745 /* Parameters:  fin(I) - pointer to packet information                      */
746 /*                                                                          */
747 /* IPv6 Only                                                                */
748 /* Analyse the packet for IPv6/UDP properties.                              */
749 /* Is not expected to be called for fragmented packets.                     */
750 /* ------------------------------------------------------------------------ */
751 static INLINE void frpr_udp6(fin)
752 fr_info_t *fin;
753 {
754 
755 	fr_checkv6sum(fin);
756 
757 	frpr_short6(fin, sizeof(struct udphdr));
758 	if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
759 		return;
760 
761 	fin->fin_flen -= fin->fin_dlen - sizeof(struct udphdr);
762 
763 	frpr_udpcommon(fin);
764 }
765 
766 
767 /* ------------------------------------------------------------------------ */
768 /* Function:    frpr_tcp6                                                   */
769 /* Returns:     void                                                        */
770 /* Parameters:  fin(I) - pointer to packet information                      */
771 /*                                                                          */
772 /* IPv6 Only                                                                */
773 /* Analyse the packet for IPv6/TCP properties.                              */
774 /* Is not expected to be called for fragmented packets.                     */
775 /* ------------------------------------------------------------------------ */
776 static INLINE void frpr_tcp6(fin)
777 fr_info_t *fin;
778 {
779 
780 	fr_checkv6sum(fin);
781 
782 	frpr_short6(fin, sizeof(struct tcphdr));
783 	if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
784 		return;
785 
786 	fin->fin_flen -= fin->fin_dlen - sizeof(struct tcphdr);
787 
788 	frpr_tcpcommon(fin);
789 }
790 
791 
792 /* ------------------------------------------------------------------------ */
793 /* Function:    frpr_esp6                                                   */
794 /* Returns:     void                                                        */
795 /* Parameters:  fin(I) - pointer to packet information                      */
796 /*                                                                          */
797 /* IPv6 Only                                                                */
798 /* Analyse the packet for ESP properties.                                   */
799 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
800 /* even though the newer ESP packets must also have a sequence number that  */
801 /* is 32bits as well, it is not possible(?) to determine the version from a */
802 /* simple packet header.                                                    */
803 /* ------------------------------------------------------------------------ */
804 static INLINE void frpr_esp6(fin)
805 fr_info_t *fin;
806 {
807 	int i;
808 	frpr_short6(fin, sizeof(grehdr_t));
809 
810 	(void) frpr_pullup(fin, 8);
811 
812 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
813 		if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
814 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
815 			break;
816 		}
817 }
818 
819 
820 /* ------------------------------------------------------------------------ */
821 /* Function:    frpr_ah6                                                    */
822 /* Returns:     void                                                        */
823 /* Parameters:  fin(I) - pointer to packet information                      */
824 /*                                                                          */
825 /* IPv6 Only                                                                */
826 /* Analyse the packet for AH properties.                                    */
827 /* The minimum length is taken to be the combination of all fields in the   */
828 /* header being present and no authentication data (null algorithm used.)   */
829 /* ------------------------------------------------------------------------ */
830 static INLINE int frpr_ah6(fin)
831 fr_info_t *fin;
832 {
833 	authhdr_t *ah;
834 	int i, shift;
835 
836 	frpr_short6(fin, 12);
837 
838 	if (frpr_pullup(fin, sizeof(*ah)) == -1)
839 		return IPPROTO_NONE;
840 
841 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
842 		if (ip6exthdr[i].ol_val == IPPROTO_AH) {
843 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
844 			break;
845 		}
846 
847 	ah = (authhdr_t *)fin->fin_dp;
848 
849 	shift = (ah->ah_plen + 2) * 4;
850 	fin->fin_dlen -= shift;
851 	fin->fin_dp = (char*)fin->fin_dp + shift;
852 
853 	return ah->ah_next;
854 }
855 
856 
857 /* ------------------------------------------------------------------------ */
858 /* Function:    frpr_gre6                                                   */
859 /* Returns:     void                                                        */
860 /* Parameters:  fin(I) - pointer to packet information                      */
861 /*                                                                          */
862 /* Analyse the packet for GRE properties.                                   */
863 /* ------------------------------------------------------------------------ */
864 static INLINE void frpr_gre6(fin)
865 fr_info_t *fin;
866 {
867 	grehdr_t *gre;
868 
869 	frpr_short6(fin, sizeof(grehdr_t));
870 
871 	if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
872 		return;
873 
874 	gre = fin->fin_dp;
875 	if (GRE_REV(gre->gr_flags) == 1)
876 		fin->fin_data[0] = gre->gr_call;
877 }
878 #endif	/* USE_INET6 */
879 
880 
881 /* ------------------------------------------------------------------------ */
882 /* Function:    frpr_pullup                                                 */
883 /* Returns:     int     - 0 == pullup succeeded, -1 == failure              */
884 /* Parameters:  fin(I)  - pointer to packet information                     */
885 /*              plen(I) - length (excluding L3 header) to pullup            */
886 /*                                                                          */
887 /* Short inline function to cut down on code duplication to perform a call  */
888 /* to fr_pullup to ensure there is the required amount of data,             */
889 /* consecutively in the packet buffer.                                      */
890 /* ------------------------------------------------------------------------ */
891 static INLINE int frpr_pullup(fin, plen)
892 fr_info_t *fin;
893 int plen;
894 {
895 #if defined(_KERNEL)
896 	if (fin->fin_m != NULL) {
897 		int ipoff;
898 
899 		ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
900 
901 		if (fin->fin_dp != NULL)
902 			plen += (char *)fin->fin_dp -
903 				((char *)fin->fin_ip + fin->fin_hlen);
904 		plen += fin->fin_hlen;
905 		/*
906 		 * We don't do 'plen += ipoff;' here. The fr_pullup() will
907 		 * do it for us.
908 		 */
909 		if (M_LEN(fin->fin_m) < plen + ipoff) {
910 			if (fr_pullup(fin->fin_m, fin, plen) == NULL)
911 				return -1;
912 		}
913 	}
914 #endif
915 	return 0;
916 }
917 
918 
919 /* ------------------------------------------------------------------------ */
920 /* Function:    frpr_short                                                  */
921 /* Returns:     void                                                        */
922 /* Parameters:  fin(I)  - pointer to packet information                     */
923 /*              xmin(I) - minimum header size                               */
924 /*                                                                          */
925 /* Check if a packet is "short" as defined by xmin.  The rule we are        */
926 /* applying here is that the packet must not be fragmented within the layer */
927 /* 4 header.  That is, it must not be a fragment that has its offset set to */
928 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the    */
929 /* entire layer 4 header must be present (min).                             */
930 /* ------------------------------------------------------------------------ */
931 static INLINE void frpr_short(fin, xmin)
932 fr_info_t *fin;
933 int xmin;
934 {
935 
936 	if (fin->fin_off == 0) {
937 		if (fin->fin_dlen < xmin)
938 			fin->fin_flx |= FI_SHORT;
939 	} else if (fin->fin_off < xmin) {
940 		fin->fin_flx |= FI_SHORT;
941 	}
942 }
943 
944 
945 /* ------------------------------------------------------------------------ */
946 /* Function:    frpr_icmp                                                   */
947 /* Returns:     void                                                        */
948 /* Parameters:  fin(I) - pointer to packet information                      */
949 /*                                                                          */
950 /* IPv4 Only                                                                */
951 /* Do a sanity check on the packet for ICMP (v4).  In nearly all cases,     */
952 /* except extrememly bad packets, both type and code will be present.       */
953 /* The expected minimum size of an ICMP packet is very much dependent on    */
954 /* the type of it.                                                          */
955 /*                                                                          */
956 /* XXX - other ICMP sanity checks?                                          */
957 /* ------------------------------------------------------------------------ */
958 static INLINE void frpr_icmp(fin)
959 fr_info_t *fin;
960 {
961 	int minicmpsz = sizeof(struct icmp);
962 	icmphdr_t *icmp;
963 	ip_t *oip;
964 	ipf_stack_t *ifs = fin->fin_ifs;
965 
966 	if (fin->fin_off != 0) {
967 		frpr_short(fin, ICMPERR_ICMPHLEN);
968 		return;
969 	}
970 
971 	if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
972 		return;
973 
974 	fr_checkv4sum(fin);
975 
976 	if (fin->fin_dlen > 1) {
977 		icmp = fin->fin_dp;
978 
979 		fin->fin_data[0] = *(u_short *)icmp;
980 
981 		switch (icmp->icmp_type)
982 		{
983 		case ICMP_ECHOREPLY :
984 		case ICMP_ECHO :
985 		/* Router discovery messaes - RFC 1256 */
986 		case ICMP_ROUTERADVERT :
987 		case ICMP_ROUTERSOLICIT :
988 			minicmpsz = ICMP_MINLEN;
989 			break;
990 		/*
991 		 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
992 		 * 3 * timestamp(3 * 4)
993 		 */
994 		case ICMP_TSTAMP :
995 		case ICMP_TSTAMPREPLY :
996 			minicmpsz = 20;
997 			break;
998 		/*
999 		 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
1000 		 * mask(4)
1001 		 */
1002 		case ICMP_MASKREQ :
1003 		case ICMP_MASKREPLY :
1004 			minicmpsz = 12;
1005 			break;
1006 		/*
1007 		 * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
1008 		 */
1009 		case ICMP_UNREACH :
1010 			if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
1011 				if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
1012 					fin->fin_flx |= FI_BAD;
1013 			}
1014 			/* FALLTHRU */
1015 		case ICMP_SOURCEQUENCH :
1016 		case ICMP_REDIRECT :
1017 		case ICMP_TIMXCEED :
1018 		case ICMP_PARAMPROB :
1019 			fin->fin_flx |= FI_ICMPERR;
1020 			if (fr_coalesce(fin) != 1)
1021 				return;
1022 			/*
1023 			 * ICMP error packets should not be generated for IP
1024 			 * packets that are a fragment that isn't the first
1025 			 * fragment.
1026 			 */
1027 			oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1028 			if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1029 				fin->fin_flx |= FI_BAD;
1030 			break;
1031 		default :
1032 			break;
1033 		}
1034 
1035 		if (fin->fin_dlen >= 6)				/* ID field */
1036 			fin->fin_data[1] = icmp->icmp_id;
1037 	}
1038 
1039 	frpr_short(fin, minicmpsz);
1040 }
1041 
1042 
1043 /* ------------------------------------------------------------------------ */
1044 /* Function:    frpr_tcpcommon                                              */
1045 /* Returns:     void                                                        */
1046 /* Parameters:  fin(I) - pointer to packet information                      */
1047 /*                                                                          */
1048 /* TCP header sanity checking.  Look for bad combinations of TCP flags,     */
1049 /* and make some checks with how they interact with other fields.           */
1050 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is     */
1051 /* valid and mark the packet as bad if not.                                 */
1052 /* ------------------------------------------------------------------------ */
1053 static INLINE void frpr_tcpcommon(fin)
1054 fr_info_t *fin;
1055 {
1056 	int flags, tlen;
1057 	tcphdr_t *tcp;
1058 
1059 	fin->fin_flx |= FI_TCPUDP;
1060 	if (fin->fin_off != 0)
1061 		return;
1062 
1063 	if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1064 		return;
1065 	tcp = fin->fin_dp;
1066 
1067 	if (fin->fin_dlen > 3) {
1068 		fin->fin_sport = ntohs(tcp->th_sport);
1069 		fin->fin_dport = ntohs(tcp->th_dport);
1070 	}
1071 
1072 	if ((fin->fin_flx & FI_SHORT) != 0)
1073 		return;
1074 
1075 	/*
1076 	 * Use of the TCP data offset *must* result in a value that is at
1077 	 * least the same size as the TCP header.
1078 	 */
1079 	tlen = TCP_OFF(tcp) << 2;
1080 	if (tlen < sizeof(tcphdr_t)) {
1081 		fin->fin_flx |= FI_BAD;
1082 		return;
1083 	}
1084 
1085 	flags = tcp->th_flags;
1086 	fin->fin_tcpf = tcp->th_flags;
1087 
1088 	/*
1089 	 * If the urgent flag is set, then the urgent pointer must
1090 	 * also be set and vice versa.  Good TCP packets do not have
1091 	 * just one of these set.
1092 	 */
1093 	if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1094 		fin->fin_flx |= FI_BAD;
1095 	} else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1096 		/* Ignore this case, it shows up in "real" traffic with */
1097 		/* bogus values in the urgent pointer field. */
1098 		flags = flags; /* LINT */
1099 	} else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1100 		   ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1101 		/* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1102 		fin->fin_flx |= FI_BAD;
1103 	} else if (!(flags & TH_ACK)) {
1104 		/*
1105 		 * If the ack bit isn't set, then either the SYN or
1106 		 * RST bit must be set.  If the SYN bit is set, then
1107 		 * we expect the ACK field to be 0.  If the ACK is
1108 		 * not set and if URG, PSH or FIN are set, consdier
1109 		 * that to indicate a bad TCP packet.
1110 		 */
1111 		if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1112 			/*
1113 			 * Cisco PIX sets the ACK field to a random value.
1114 			 * In light of this, do not set FI_BAD until a patch
1115 			 * is available from Cisco to ensure that
1116 			 * interoperability between existing systems is
1117 			 * achieved.
1118 			 */
1119 			/*fin->fin_flx |= FI_BAD*/;
1120 			flags = flags; /* LINT */
1121 		} else if (!(flags & (TH_RST|TH_SYN))) {
1122 			fin->fin_flx |= FI_BAD;
1123 		} else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1124 			fin->fin_flx |= FI_BAD;
1125 		}
1126 	}
1127 
1128 	/*
1129 	 * At this point, it's not exactly clear what is to be gained by
1130 	 * marking up which TCP options are and are not present.  The one we
1131 	 * are most interested in is the TCP window scale.  This is only in
1132 	 * a SYN packet [RFC1323] so we don't need this here...?
1133 	 * Now if we were to analyse the header for passive fingerprinting,
1134 	 * then that might add some weight to adding this...
1135 	 */
1136 	if (tlen == sizeof(tcphdr_t))
1137 		return;
1138 
1139 	if (frpr_pullup(fin, tlen) == -1)
1140 		return;
1141 
1142 #if 0
1143 	ip = fin->fin_ip;
1144 	s = (u_char *)(tcp + 1);
1145 	off = IP_HL(ip) << 2;
1146 # ifdef _KERNEL
1147 	if (fin->fin_mp != NULL) {
1148 		mb_t *m = *fin->fin_mp;
1149 
1150 		if (off + tlen > M_LEN(m))
1151 			return;
1152 	}
1153 # endif
1154 	for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1155 		opt = *s;
1156 		if (opt == '\0')
1157 			break;
1158 		else if (opt == TCPOPT_NOP)
1159 			ol = 1;
1160 		else {
1161 			if (tlen < 2)
1162 				break;
1163 			ol = (int)*(s + 1);
1164 			if (ol < 2 || ol > tlen)
1165 				break;
1166 		}
1167 
1168 		for (i = 9, mv = 4; mv >= 0; ) {
1169 			op = ipopts + i;
1170 			if (opt == (u_char)op->ol_val) {
1171 				optmsk |= op->ol_bit;
1172 				break;
1173 			}
1174 		}
1175 		tlen -= ol;
1176 		s += ol;
1177 	}
1178 #endif /* 0 */
1179 }
1180 
1181 
1182 
1183 /* ------------------------------------------------------------------------ */
1184 /* Function:    frpr_udpcommon                                              */
1185 /* Returns:     void                                                        */
1186 /* Parameters:  fin(I) - pointer to packet information                      */
1187 /*                                                                          */
1188 /* Extract the UDP source and destination ports, if present.  If compiled   */
1189 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid.          */
1190 /* ------------------------------------------------------------------------ */
1191 static INLINE void frpr_udpcommon(fin)
1192 fr_info_t *fin;
1193 {
1194 	udphdr_t *udp;
1195 
1196 	fin->fin_flx |= FI_TCPUDP;
1197 
1198 	if (!fin->fin_off && (fin->fin_dlen > 3)) {
1199 		if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1200 			fin->fin_flx |= FI_SHORT;
1201 			return;
1202 		}
1203 
1204 		udp = fin->fin_dp;
1205 
1206 		fin->fin_sport = ntohs(udp->uh_sport);
1207 		fin->fin_dport = ntohs(udp->uh_dport);
1208 	}
1209 }
1210 
1211 
1212 /* ------------------------------------------------------------------------ */
1213 /* Function:    frpr_tcp                                                    */
1214 /* Returns:     void                                                        */
1215 /* Parameters:  fin(I) - pointer to packet information                      */
1216 /*                                                                          */
1217 /* IPv4 Only                                                                */
1218 /* Analyse the packet for IPv4/TCP properties.                              */
1219 /* ------------------------------------------------------------------------ */
1220 static INLINE void frpr_tcp(fin)
1221 fr_info_t *fin;
1222 {
1223 
1224 	fr_checkv4sum(fin);
1225 
1226 	frpr_short(fin, sizeof(tcphdr_t));
1227 
1228 	frpr_tcpcommon(fin);
1229 }
1230 
1231 
1232 /* ------------------------------------------------------------------------ */
1233 /* Function:    frpr_udp                                                    */
1234 /* Returns:     void                                                        */
1235 /* Parameters:  fin(I) - pointer to packet information                      */
1236 /*                                                                          */
1237 /* IPv4 Only                                                                */
1238 /* Analyse the packet for IPv4/UDP properties.                              */
1239 /* ------------------------------------------------------------------------ */
1240 static INLINE void frpr_udp(fin)
1241 fr_info_t *fin;
1242 {
1243 
1244 	fr_checkv4sum(fin);
1245 
1246 	frpr_short(fin, sizeof(udphdr_t));
1247 
1248 	frpr_udpcommon(fin);
1249 }
1250 
1251 
1252 /* ------------------------------------------------------------------------ */
1253 /* Function:    frpr_esp                                                    */
1254 /* Returns:     void                                                        */
1255 /* Parameters:  fin(I) - pointer to packet information                      */
1256 /*                                                                          */
1257 /* Analyse the packet for ESP properties.                                   */
1258 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
1259 /* even though the newer ESP packets must also have a sequence number that  */
1260 /* is 32bits as well, it is not possible(?) to determine the version from a */
1261 /* simple packet header.                                                    */
1262 /* ------------------------------------------------------------------------ */
1263 static INLINE void frpr_esp(fin)
1264 fr_info_t *fin;
1265 {
1266 	if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1267 		return;
1268 
1269 	frpr_short(fin, 8);
1270 }
1271 
1272 
1273 /* ------------------------------------------------------------------------ */
1274 /* Function:    frpr_ah                                                     */
1275 /* Returns:     void                                                        */
1276 /* Parameters:  fin(I) - pointer to packet information                      */
1277 /*                                                                          */
1278 /* Analyse the packet for AH properties.                                    */
1279 /* The minimum length is taken to be the combination of all fields in the   */
1280 /* header being present and no authentication data (null algorithm used.)   */
1281 /* ------------------------------------------------------------------------ */
1282 static INLINE void frpr_ah(fin)
1283 fr_info_t *fin;
1284 {
1285 	authhdr_t *ah;
1286 	int len;
1287 
1288 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1289 		return;
1290 
1291 	ah = (authhdr_t *)fin->fin_dp;
1292 
1293 	len = (ah->ah_plen + 2) << 2;
1294 	frpr_short(fin, len);
1295 }
1296 
1297 
1298 /* ------------------------------------------------------------------------ */
1299 /* Function:    frpr_gre                                                    */
1300 /* Returns:     void                                                        */
1301 /* Parameters:  fin(I) - pointer to packet information                      */
1302 /*                                                                          */
1303 /* Analyse the packet for GRE properties.                                   */
1304 /* ------------------------------------------------------------------------ */
1305 static INLINE void frpr_gre(fin)
1306 fr_info_t *fin;
1307 {
1308 	grehdr_t *gre;
1309 
1310 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1311 		return;
1312 
1313 	frpr_short(fin, sizeof(grehdr_t));
1314 
1315 	if (fin->fin_off == 0) {
1316 		gre = fin->fin_dp;
1317 		if (GRE_REV(gre->gr_flags) == 1)
1318 			fin->fin_data[0] = gre->gr_call;
1319 	}
1320 }
1321 
1322 
1323 /* ------------------------------------------------------------------------ */
1324 /* Function:    frpr_ipv4hdr                                                */
1325 /* Returns:     void                                                        */
1326 /* Parameters:  fin(I) - pointer to packet information                      */
1327 /*                                                                          */
1328 /* IPv4 Only                                                                */
1329 /* Analyze the IPv4 header and set fields in the fr_info_t structure.       */
1330 /* Check all options present and flag their presence if any exist.          */
1331 /* ------------------------------------------------------------------------ */
1332 static INLINE void frpr_ipv4hdr(fin)
1333 fr_info_t *fin;
1334 {
1335 	u_short optmsk = 0, secmsk = 0, auth = 0;
1336 	int hlen, ol, mv, p, i;
1337 	const struct optlist *op;
1338 	u_char *s, opt;
1339 	u_short off;
1340 	fr_ip_t *fi;
1341 	ip_t *ip;
1342 
1343 	fi = &fin->fin_fi;
1344 	hlen = fin->fin_hlen;
1345 
1346 	ip = fin->fin_ip;
1347 	p = ip->ip_p;
1348 	fi->fi_p = p;
1349 	fi->fi_tos = ip->ip_tos;
1350 	fin->fin_id = ip->ip_id;
1351 	off = ip->ip_off;
1352 
1353 	/* Get both TTL and protocol */
1354 	fi->fi_p = ip->ip_p;
1355 	fi->fi_ttl = ip->ip_ttl;
1356 #if 0
1357 	(*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1358 #endif
1359 
1360 	/* Zero out bits not used in IPv6 address */
1361 	fi->fi_src.i6[1] = 0;
1362 	fi->fi_src.i6[2] = 0;
1363 	fi->fi_src.i6[3] = 0;
1364 	fi->fi_dst.i6[1] = 0;
1365 	fi->fi_dst.i6[2] = 0;
1366 	fi->fi_dst.i6[3] = 0;
1367 
1368 	fi->fi_saddr = ip->ip_src.s_addr;
1369 	fi->fi_daddr = ip->ip_dst.s_addr;
1370 
1371 	/*
1372 	 * set packet attribute flags based on the offset and
1373 	 * calculate the byte offset that it represents.
1374 	 */
1375 	off &= IP_MF|IP_OFFMASK;
1376 	if (off != 0) {
1377 		fi->fi_flx |= FI_FRAG;
1378 		off &= IP_OFFMASK;
1379 		if (off != 0) {
1380 			fin->fin_flx |= FI_FRAGBODY;
1381 			off <<= 3;
1382 			if ((off + fin->fin_dlen > 65535) ||
1383 			    (fin->fin_dlen == 0) ||
1384 			    ((ip->ip_off & IP_MF) && (fin->fin_dlen & 7))) {
1385 				/*
1386 				 * The length of the packet, starting at its
1387 				 * offset cannot exceed 65535 (0xffff) as the
1388 				 * length of an IP packet is only 16 bits.
1389 				 *
1390 				 * Any fragment that isn't the last fragment
1391 				 * must have a length greater than 0 and it
1392 				 * must be an even multiple of 8.
1393 				 */
1394 				fi->fi_flx |= FI_BAD;
1395 			}
1396 		}
1397 	}
1398 	fin->fin_off = off;
1399 
1400 	/*
1401 	 * Call per-protocol setup and checking
1402 	 */
1403 	switch (p)
1404 	{
1405 	case IPPROTO_UDP :
1406 		frpr_udp(fin);
1407 		break;
1408 	case IPPROTO_TCP :
1409 		frpr_tcp(fin);
1410 		break;
1411 	case IPPROTO_ICMP :
1412 		frpr_icmp(fin);
1413 		break;
1414 	case IPPROTO_AH :
1415 		frpr_ah(fin);
1416 		break;
1417 	case IPPROTO_ESP :
1418 		frpr_esp(fin);
1419 		break;
1420 	case IPPROTO_GRE :
1421 		frpr_gre(fin);
1422 		break;
1423 	}
1424 
1425 	ip = fin->fin_ip;
1426 	if (ip == NULL)
1427 		return;
1428 
1429 	/*
1430 	 * If it is a standard IP header (no options), set the flag fields
1431 	 * which relate to options to 0.
1432 	 */
1433 	if (hlen == sizeof(*ip)) {
1434 		fi->fi_optmsk = 0;
1435 		fi->fi_secmsk = 0;
1436 		fi->fi_auth = 0;
1437 		return;
1438 	}
1439 
1440 	/*
1441 	 * So the IP header has some IP options attached.  Walk the entire
1442 	 * list of options present with this packet and set flags to indicate
1443 	 * which ones are here and which ones are not.  For the somewhat out
1444 	 * of date and obscure security classification options, set a flag to
1445 	 * represent which classification is present.
1446 	 */
1447 	fi->fi_flx |= FI_OPTIONS;
1448 
1449 	for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1450 		opt = *s;
1451 		if (opt == '\0')
1452 			break;
1453 		else if (opt == IPOPT_NOP)
1454 			ol = 1;
1455 		else {
1456 			if (hlen < 2)
1457 				break;
1458 			ol = (int)*(s + 1);
1459 			if (ol < 2 || ol > hlen)
1460 				break;
1461 		}
1462 		for (i = 9, mv = 4; mv >= 0; ) {
1463 			op = ipopts + i;
1464 			if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1465 				optmsk |= op->ol_bit;
1466 				if (opt == IPOPT_SECURITY) {
1467 					const struct optlist *sp;
1468 					u_char	sec;
1469 					int j, m;
1470 
1471 					sec = *(s + 2);	/* classification */
1472 					for (j = 3, m = 2; m >= 0; ) {
1473 						sp = secopt + j;
1474 						if (sec == sp->ol_val) {
1475 							secmsk |= sp->ol_bit;
1476 							auth = *(s + 3);
1477 							auth *= 256;
1478 							auth += *(s + 4);
1479 							break;
1480 						}
1481 						if (sec < sp->ol_val)
1482 							j -= m;
1483 						else
1484 							j += m;
1485 						m--;
1486 					}
1487 				}
1488 				break;
1489 			}
1490 			if (opt < op->ol_val)
1491 				i -= mv;
1492 			else
1493 				i += mv;
1494 			mv--;
1495 		}
1496 		hlen -= ol;
1497 		s += ol;
1498 	}
1499 
1500 	/*
1501 	 *
1502 	 */
1503 	if (auth && !(auth & 0x0100))
1504 		auth &= 0xff00;
1505 	fi->fi_optmsk = optmsk;
1506 	fi->fi_secmsk = secmsk;
1507 	fi->fi_auth = auth;
1508 }
1509 
1510 
1511 /* ------------------------------------------------------------------------ */
1512 /* Function:    fr_makefrip                                                 */
1513 /* Returns:     int - 1 == hdr checking error, 0 == OK                      */
1514 /* Parameters:  hlen(I) - length of IP packet header                        */
1515 /*              ip(I)   - pointer to the IP header                          */
1516 /*              fin(IO) - pointer to packet information                     */
1517 /*                                                                          */
1518 /* Compact the IP header into a structure which contains just the info.     */
1519 /* which is useful for comparing IP headers with and store this information */
1520 /* in the fr_info_t structure pointer to by fin.  At present, it is assumed */
1521 /* this function will be called with either an IPv4 or IPv6 packet.         */
1522 /* ------------------------------------------------------------------------ */
1523 int	fr_makefrip(hlen, ip, fin)
1524 int hlen;
1525 ip_t *ip;
1526 fr_info_t *fin;
1527 {
1528 	int v;
1529 
1530 	fin->fin_nat = NULL;
1531 	fin->fin_state = NULL;
1532 	fin->fin_depth = 0;
1533 	fin->fin_hlen = (u_short)hlen;
1534 	fin->fin_ip = ip;
1535 	fin->fin_rule = 0xffffffff;
1536 	fin->fin_group[0] = -1;
1537 	fin->fin_group[1] = '\0';
1538 	fin->fin_dlen = fin->fin_plen - hlen;
1539 	fin->fin_dp = (char *)ip + hlen;
1540 
1541 	v = fin->fin_v;
1542 	if (v == 4)
1543 		frpr_ipv4hdr(fin);
1544 #ifdef	USE_INET6
1545 	else if (v == 6) {
1546 		if (frpr_ipv6hdr(fin) == -1)
1547 			return -1;
1548 	}
1549 #endif
1550 	if (fin->fin_ip == NULL)
1551 		return -1;
1552 	return 0;
1553 }
1554 
1555 
1556 /* ------------------------------------------------------------------------ */
1557 /* Function:    fr_portcheck                                                */
1558 /* Returns:     int - 1 == port matched, 0 == port match failed             */
1559 /* Parameters:  frp(I) - pointer to port check `expression'                 */
1560 /*              pop(I) - pointer to port number to evaluate                 */
1561 /*                                                                          */
1562 /* Perform a comparison of a port number against some other(s), using a     */
1563 /* structure with compare information stored in it.                         */
1564 /* ------------------------------------------------------------------------ */
1565 static INLINE int fr_portcheck(frp, pop)
1566 frpcmp_t *frp;
1567 u_short *pop;
1568 {
1569 	u_short tup, po;
1570 	int err = 1;
1571 
1572 	tup = *pop;
1573 	po = frp->frp_port;
1574 
1575 	/*
1576 	 * Do opposite test to that required and continue if that succeeds.
1577 	 */
1578 	switch (frp->frp_cmp)
1579 	{
1580 	case FR_EQUAL :
1581 		if (tup != po) /* EQUAL */
1582 			err = 0;
1583 		break;
1584 	case FR_NEQUAL :
1585 		if (tup == po) /* NOTEQUAL */
1586 			err = 0;
1587 		break;
1588 	case FR_LESST :
1589 		if (tup >= po) /* LESSTHAN */
1590 			err = 0;
1591 		break;
1592 	case FR_GREATERT :
1593 		if (tup <= po) /* GREATERTHAN */
1594 			err = 0;
1595 		break;
1596 	case FR_LESSTE :
1597 		if (tup > po) /* LT or EQ */
1598 			err = 0;
1599 		break;
1600 	case FR_GREATERTE :
1601 		if (tup < po) /* GT or EQ */
1602 			err = 0;
1603 		break;
1604 	case FR_OUTRANGE :
1605 		if (tup >= po && tup <= frp->frp_top) /* Out of range */
1606 			err = 0;
1607 		break;
1608 	case FR_INRANGE :
1609 		if (tup <= po || tup >= frp->frp_top) /* In range */
1610 			err = 0;
1611 		break;
1612 	case FR_INCRANGE :
1613 		if (tup < po || tup > frp->frp_top) /* Inclusive range */
1614 			err = 0;
1615 		break;
1616 	default :
1617 		break;
1618 	}
1619 	return err;
1620 }
1621 
1622 
1623 /* ------------------------------------------------------------------------ */
1624 /* Function:    fr_tcpudpchk                                                */
1625 /* Returns:     int - 1 == protocol matched, 0 == check failed              */
1626 /* Parameters:  fin(I) - pointer to packet information                      */
1627 /*              ft(I)  - pointer to structure with comparison data          */
1628 /*                                                                          */
1629 /* Compares the current pcket (assuming it is TCP/UDP) information with a   */
1630 /* structure containing information that we want to match against.          */
1631 /* ------------------------------------------------------------------------ */
1632 int fr_tcpudpchk(fin, ft)
1633 fr_info_t *fin;
1634 frtuc_t *ft;
1635 {
1636 	int err = 1;
1637 
1638 	/*
1639 	 * Both ports should *always* be in the first fragment.
1640 	 * So far, I cannot find any cases where they can not be.
1641 	 *
1642 	 * compare destination ports
1643 	 */
1644 	if (ft->ftu_dcmp)
1645 		err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1646 
1647 	/*
1648 	 * compare source ports
1649 	 */
1650 	if (err && ft->ftu_scmp)
1651 		err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1652 
1653 	/*
1654 	 * If we don't have all the TCP/UDP header, then how can we
1655 	 * expect to do any sort of match on it ?  If we were looking for
1656 	 * TCP flags, then NO match.  If not, then match (which should
1657 	 * satisfy the "short" class too).
1658 	 */
1659 	if (err && (fin->fin_p == IPPROTO_TCP)) {
1660 		if (fin->fin_flx & FI_SHORT)
1661 			return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1662 		/*
1663 		 * Match the flags ?  If not, abort this match.
1664 		 */
1665 		if (ft->ftu_tcpfm &&
1666 		    ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1667 			FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1668 				 ft->ftu_tcpfm, ft->ftu_tcpf));
1669 			err = 0;
1670 		}
1671 	}
1672 	return err;
1673 }
1674 
1675 
1676 /* ------------------------------------------------------------------------ */
1677 /* Function:    fr_ipfcheck                                                 */
1678 /* Returns:     int - 0 == match, 1 == no match                             */
1679 /* Parameters:  fin(I)     - pointer to packet information                  */
1680 /*              fr(I)      - pointer to filter rule                         */
1681 /*              portcmp(I) - flag indicating whether to attempt matching on */
1682 /*                           TCP/UDP port data.                             */
1683 /*                                                                          */
1684 /* Check to see if a packet matches an IPFilter rule.  Checks of addresses, */
1685 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1686 /* this function.                                                           */
1687 /* ------------------------------------------------------------------------ */
1688 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1689 fr_info_t *fin;
1690 frentry_t *fr;
1691 int portcmp;
1692 {
1693 	u_32_t	*ld, *lm, *lip;
1694 	fripf_t *fri;
1695 	fr_ip_t *fi;
1696 	int i;
1697 	ipf_stack_t *ifs = fin->fin_ifs;
1698 
1699 	fi = &fin->fin_fi;
1700 	fri = fr->fr_ipf;
1701 	lip = (u_32_t *)fi;
1702 	lm = (u_32_t *)&fri->fri_mip;
1703 	ld = (u_32_t *)&fri->fri_ip;
1704 
1705 	/*
1706 	 * first 32 bits to check coversion:
1707 	 * IP version, TOS, TTL, protocol
1708 	 */
1709 	i = ((*lip & *lm) != *ld);
1710 	FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1711 		   *lip, *lm, *ld));
1712 	if (i)
1713 		return 1;
1714 
1715 	/*
1716 	 * Next 32 bits is a constructed bitmask indicating which IP options
1717 	 * are present (if any) in this packet.
1718 	 */
1719 	lip++, lm++, ld++;
1720 	i |= ((*lip & *lm) != *ld);
1721 	FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1722 		   *lip, *lm, *ld));
1723 	if (i)
1724 		return 1;
1725 
1726 	lip++, lm++, ld++;
1727 	/*
1728 	 * Unrolled loops (4 each, for 32 bits) for address checks.
1729 	 */
1730 	/*
1731 	 * Check the source address.
1732 	 */
1733 #ifdef	IPFILTER_LOOKUP
1734 	if (fr->fr_satype == FRI_LOOKUP) {
1735 		i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, ifs);
1736 		if (i == -1)
1737 			return 1;
1738 		lip += 3;
1739 		lm += 3;
1740 		ld += 3;
1741 	} else {
1742 #endif
1743 		i = ((*lip & *lm) != *ld);
1744 		FR_DEBUG(("2a. %#08x & %#08x != %#08x\n",
1745 			   *lip, *lm, *ld));
1746 		if (fi->fi_v == 6) {
1747 			lip++, lm++, ld++;
1748 			i |= ((*lip & *lm) != *ld);
1749 			FR_DEBUG(("2b. %#08x & %#08x != %#08x\n",
1750 				   *lip, *lm, *ld));
1751 			lip++, lm++, ld++;
1752 			i |= ((*lip & *lm) != *ld);
1753 			FR_DEBUG(("2c. %#08x & %#08x != %#08x\n",
1754 				   *lip, *lm, *ld));
1755 			lip++, lm++, ld++;
1756 			i |= ((*lip & *lm) != *ld);
1757 			FR_DEBUG(("2d. %#08x & %#08x != %#08x\n",
1758 				   *lip, *lm, *ld));
1759 		} else {
1760 			lip += 3;
1761 			lm += 3;
1762 			ld += 3;
1763 		}
1764 #ifdef	IPFILTER_LOOKUP
1765 	}
1766 #endif
1767 	i ^= (fr->fr_flags & FR_NOTSRCIP) >> 6;
1768 	if (i)
1769 		return 1;
1770 
1771 	/*
1772 	 * Check the destination address.
1773 	 */
1774 	lip++, lm++, ld++;
1775 #ifdef	IPFILTER_LOOKUP
1776 	if (fr->fr_datype == FRI_LOOKUP) {
1777 		i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, ifs);
1778 		if (i == -1)
1779 			return 1;
1780 		lip += 3;
1781 		lm += 3;
1782 		ld += 3;
1783 	} else {
1784 #endif
1785 		i = ((*lip & *lm) != *ld);
1786 		FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1787 			   *lip, *lm, *ld));
1788 		if (fi->fi_v == 6) {
1789 			lip++, lm++, ld++;
1790 			i |= ((*lip & *lm) != *ld);
1791 			FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1792 				   *lip, *lm, *ld));
1793 			lip++, lm++, ld++;
1794 			i |= ((*lip & *lm) != *ld);
1795 			FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1796 				   *lip, *lm, *ld));
1797 			lip++, lm++, ld++;
1798 			i |= ((*lip & *lm) != *ld);
1799 			FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1800 				   *lip, *lm, *ld));
1801 		} else {
1802 			lip += 3;
1803 			lm += 3;
1804 			ld += 3;
1805 		}
1806 #ifdef	IPFILTER_LOOKUP
1807 	}
1808 #endif
1809 	i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1810 	if (i)
1811 		return 1;
1812 	/*
1813 	 * IP addresses matched.  The next 32bits contains:
1814 	 * mast of old IP header security & authentication bits.
1815 	 */
1816 	lip++, lm++, ld++;
1817 	i |= ((*lip & *lm) != *ld);
1818 	FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1819 		   *lip, *lm, *ld));
1820 
1821 	/*
1822 	 * Next we have 32 bits of packet flags.
1823 	 */
1824 	lip++, lm++, ld++;
1825 	i |= ((*lip & *lm) != *ld);
1826 	FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1827 		   *lip, *lm, *ld));
1828 
1829 	if (i == 0) {
1830 		/*
1831 		 * If a fragment, then only the first has what we're
1832 		 * looking for here...
1833 		 */
1834 		if (portcmp) {
1835 			if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1836 				i = 1;
1837 		} else {
1838 			if (fr->fr_dcmp || fr->fr_scmp ||
1839 			    fr->fr_tcpf || fr->fr_tcpfm)
1840 				i = 1;
1841 			if (fr->fr_icmpm || fr->fr_icmp) {
1842 				if (((fi->fi_p != IPPROTO_ICMP) &&
1843 				     (fi->fi_p != IPPROTO_ICMPV6)) ||
1844 				    fin->fin_off || (fin->fin_dlen < 2))
1845 					i = 1;
1846 				else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1847 					 fr->fr_icmp) {
1848 					FR_DEBUG(("i. %#x & %#x != %#x\n",
1849 						 fin->fin_data[0],
1850 						 fr->fr_icmpm, fr->fr_icmp));
1851 					i = 1;
1852 				}
1853 			}
1854 		}
1855 	}
1856 	return i;
1857 }
1858 
1859 
1860 /* ------------------------------------------------------------------------ */
1861 /* Function:    fr_scanlist                                                 */
1862 /* Returns:     int - result flags of scanning filter list                  */
1863 /* Parameters:  fin(I) - pointer to packet information                      */
1864 /*              pass(I) - default result to return for filtering            */
1865 /*                                                                          */
1866 /* Check the input/output list of rules for a match to the current packet.  */
1867 /* If a match is found, the value of fr_flags from the rule becomes the     */
1868 /* return value and fin->fin_fr points to the matched rule.                 */
1869 /*                                                                          */
1870 /* This function may be called recusively upto 16 times (limit inbuilt.)    */
1871 /* When unwinding, it should finish up with fin_depth as 0.                 */
1872 /*                                                                          */
1873 /* Could be per interface, but this gets real nasty when you don't have,    */
1874 /* or can't easily change, the kernel source code to .                      */
1875 /* ------------------------------------------------------------------------ */
1876 int fr_scanlist(fin, pass)
1877 fr_info_t *fin;
1878 u_32_t pass;
1879 {
1880 	int rulen, portcmp, off, logged, skip;
1881 	struct frentry *fr, *fnext;
1882 	u_32_t passt, passo;
1883 	ipf_stack_t *ifs = fin->fin_ifs;
1884 
1885 	/*
1886 	 * Do not allow nesting deeper than 16 levels.
1887 	 */
1888 	if (fin->fin_depth >= 16)
1889 		return pass;
1890 
1891 	fr = fin->fin_fr;
1892 
1893 	/*
1894 	 * If there are no rules in this list, return now.
1895 	 */
1896 	if (fr == NULL)
1897 		return pass;
1898 
1899 	skip = 0;
1900 	logged = 0;
1901 	portcmp = 0;
1902 	fin->fin_depth++;
1903 	fin->fin_fr = NULL;
1904 	off = fin->fin_off;
1905 
1906 	if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1907 		portcmp = 1;
1908 
1909 	for (rulen = 0; fr; fr = fnext, rulen++) {
1910 		fnext = fr->fr_next;
1911 		if (skip != 0) {
1912 			FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1913 			skip--;
1914 			continue;
1915 		}
1916 
1917 		/*
1918 		 * In all checks below, a null (zero) value in the
1919 		 * filter struture is taken to mean a wildcard.
1920 		 *
1921 		 * check that we are working for the right interface
1922 		 */
1923 #ifdef	_KERNEL
1924 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1925 			continue;
1926 #else
1927 		if (opts & (OPT_VERBOSE|OPT_DEBUG))
1928 			printf("\n");
1929 		FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1930 				  FR_ISPASS(pass) ? 'p' :
1931 				  FR_ISACCOUNT(pass) ? 'A' :
1932 				  FR_ISAUTH(pass) ? 'a' :
1933 				  (pass & FR_NOMATCH) ? 'n' :'b'));
1934 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1935 			continue;
1936 		FR_VERBOSE((":i"));
1937 #endif
1938 
1939 		switch (fr->fr_type)
1940 		{
1941 		case FR_T_IPF :
1942 		case FR_T_IPF|FR_T_BUILTIN :
1943 			if (fr_ipfcheck(fin, fr, portcmp))
1944 				continue;
1945 			break;
1946 #if defined(IPFILTER_BPF)
1947 		case FR_T_BPFOPC :
1948 		case FR_T_BPFOPC|FR_T_BUILTIN :
1949 		    {
1950 			u_char *mc;
1951 
1952 			if (*fin->fin_mp == NULL)
1953 				continue;
1954 			if (fin->fin_v != fr->fr_v)
1955 				continue;
1956 			mc = (u_char *)fin->fin_m;
1957 			if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1958 				continue;
1959 			break;
1960 		    }
1961 #endif
1962 		case FR_T_CALLFUNC|FR_T_BUILTIN :
1963 		    {
1964 			frentry_t *f;
1965 
1966 			f = (*fr->fr_func)(fin, &pass);
1967 			if (f != NULL)
1968 				fr = f;
1969 			else
1970 				continue;
1971 			break;
1972 		    }
1973 		default :
1974 			break;
1975 		}
1976 
1977 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1978 			if (fin->fin_nattag == NULL)
1979 				continue;
1980 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1981 				continue;
1982 		}
1983 		FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1984 
1985 		passt = fr->fr_flags;
1986 
1987 		/*
1988 		 * Allowing a rule with the "keep state" flag set to match
1989 		 * packets that have been tagged "out of window" by the TCP
1990 		 * state tracking is foolish as the attempt to add a new
1991 		 * state entry to the table will fail.
1992 		 */
1993 		if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1994 			continue;
1995 
1996 		/*
1997 		 * If the rule is a "call now" rule, then call the function
1998 		 * in the rule, if it exists and use the results from that.
1999 		 * If the function pointer is bad, just make like we ignore
2000 		 * it, except for increasing the hit counter.
2001 		 */
2002 		if ((passt & FR_CALLNOW) != 0) {
2003 			IPF_BUMP(fr->fr_hits);
2004 			if ((fr->fr_func != NULL) &&
2005 			    (fr->fr_func != (ipfunc_t)-1)) {
2006 				frentry_t *frs;
2007 
2008 				frs = fin->fin_fr;
2009 				fin->fin_fr = fr;
2010 				fr = (*fr->fr_func)(fin, &passt);
2011 				if (fr == NULL) {
2012 					fin->fin_fr = frs;
2013 					continue;
2014 				}
2015 				passt = fr->fr_flags;
2016 				fin->fin_fr = fr;
2017 			}
2018 		} else {
2019 			fin->fin_fr = fr;
2020 		}
2021 
2022 #ifdef  IPFILTER_LOG
2023 		/*
2024 		 * Just log this packet...
2025 		 */
2026 		if ((passt & FR_LOGMASK) == FR_LOG) {
2027 			if (ipflog(fin, passt) == -1) {
2028 				if (passt & FR_LOGORBLOCK) {
2029 					passt &= ~FR_CMDMASK;
2030 					passt |= FR_BLOCK|FR_QUICK;
2031 				}
2032 				IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2033 			}
2034 			IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2035 			logged = 1;
2036 		}
2037 #endif /* IPFILTER_LOG */
2038 		fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2039 		passo = pass;
2040 		if (FR_ISSKIP(passt))
2041 			skip = fr->fr_arg;
2042 		else if ((passt & FR_LOGMASK) != FR_LOG)
2043 			pass = passt;
2044 		if (passt & (FR_RETICMP|FR_FAKEICMP))
2045 			fin->fin_icode = fr->fr_icode;
2046 		FR_DEBUG(("pass %#x\n", pass));
2047 		IPF_BUMP(fr->fr_hits);
2048 		fin->fin_rule = rulen;
2049 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2050 		if (fr->fr_grp != NULL) {
2051 			fin->fin_fr = *fr->fr_grp;
2052 			pass = fr_scanlist(fin, pass);
2053 			if (fin->fin_fr == NULL) {
2054 				fin->fin_rule = rulen;
2055 				(void) strncpy(fin->fin_group, fr->fr_group,
2056 					       FR_GROUPLEN);
2057 				fin->fin_fr = fr;
2058 			}
2059 			if (fin->fin_flx & FI_DONTCACHE)
2060 				logged = 1;
2061 		}
2062 
2063 		if (pass & FR_QUICK) {
2064 			/*
2065 			 * Finally, if we've asked to track state for this
2066 			 * packet, set it up.  Add state for "quick" rules
2067 			 * here so that if the action fails we can consider
2068 			 * the rule to "not match" and keep on processing
2069 			 * filter rules.
2070 			 */
2071 			if ((pass & FR_KEEPSTATE) &&
2072 			    !(fin->fin_flx & FI_STATE)) {
2073 				int out = fin->fin_out;
2074 
2075 				if (fr_addstate(fin, NULL, 0) != NULL) {
2076 					IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2077 				} else {
2078 					IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2079 					pass = passo;
2080 					continue;
2081 				}
2082 			}
2083 			break;
2084 		}
2085 	}
2086 	if (logged)
2087 		fin->fin_flx |= FI_DONTCACHE;
2088 	fin->fin_depth--;
2089 	return pass;
2090 }
2091 
2092 
2093 /* ------------------------------------------------------------------------ */
2094 /* Function:    fr_acctpkt                                                  */
2095 /* Returns:     frentry_t* - always returns NULL                            */
2096 /* Parameters:  fin(I) - pointer to packet information                      */
2097 /*              passp(IO) - pointer to current/new filter decision (unused) */
2098 /*                                                                          */
2099 /* Checks a packet against accounting rules, if there are any for the given */
2100 /* IP protocol version.                                                     */
2101 /*                                                                          */
2102 /* N.B.: this function returns NULL to match the prototype used by other    */
2103 /* functions called from the IPFilter "mainline" in fr_check().             */
2104 /* ------------------------------------------------------------------------ */
2105 frentry_t *fr_acctpkt(fin, passp)
2106 fr_info_t *fin;
2107 u_32_t *passp;
2108 {
2109 	char group[FR_GROUPLEN];
2110 	frentry_t *fr, *frsave;
2111 	u_32_t pass, rulen;
2112 	ipf_stack_t *ifs = fin->fin_ifs;
2113 
2114 	passp = passp;
2115 #ifdef	USE_INET6
2116 	if (fin->fin_v == 6)
2117 		fr = ifs->ifs_ipacct6[fin->fin_out][ifs->ifs_fr_active];
2118 	else
2119 #endif
2120 		fr = ifs->ifs_ipacct[fin->fin_out][ifs->ifs_fr_active];
2121 
2122 	if (fr != NULL) {
2123 		frsave = fin->fin_fr;
2124 		bcopy(fin->fin_group, group, FR_GROUPLEN);
2125 		rulen = fin->fin_rule;
2126 		fin->fin_fr = fr;
2127 		pass = fr_scanlist(fin, FR_NOMATCH);
2128 		if (FR_ISACCOUNT(pass)) {
2129 			IPF_BUMP(ifs->ifs_frstats[0].fr_acct);
2130 		}
2131 		fin->fin_fr = frsave;
2132 		bcopy(group, fin->fin_group, FR_GROUPLEN);
2133 		fin->fin_rule = rulen;
2134 	}
2135 	return NULL;
2136 }
2137 
2138 
2139 /* ------------------------------------------------------------------------ */
2140 /* Function:    fr_firewall                                                 */
2141 /* Returns:     frentry_t* - returns pointer to matched rule, if no matches */
2142 /*                           were found, returns NULL.                      */
2143 /* Parameters:  fin(I) - pointer to packet information                      */
2144 /*              passp(IO) - pointer to current/new filter decision (unused) */
2145 /*                                                                          */
2146 /* Applies an appropriate set of firewall rules to the packet, to see if    */
2147 /* there are any matches.  The first check is to see if a match can be seen */
2148 /* in the cache.  If not, then search an appropriate list of rules.  Once a */
2149 /* matching rule is found, take any appropriate actions as defined by the   */
2150 /* rule - except logging.                                                   */
2151 /* ------------------------------------------------------------------------ */
2152 static frentry_t *fr_firewall(fin, passp)
2153 fr_info_t *fin;
2154 u_32_t *passp;
2155 {
2156 	frentry_t *fr;
2157 	u_32_t pass;
2158 	int out;
2159 	ipf_stack_t *ifs = fin->fin_ifs;
2160 
2161 	out = fin->fin_out;
2162 	pass = *passp;
2163 
2164 #ifdef	USE_INET6
2165 	if (fin->fin_v == 6)
2166 		fin->fin_fr = ifs->ifs_ipfilter6[out][ifs->ifs_fr_active];
2167 	else
2168 #endif
2169 		fin->fin_fr = ifs->ifs_ipfilter[out][ifs->ifs_fr_active];
2170 	if (fin->fin_fr != NULL)
2171 		pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2172 
2173 	if ((pass & FR_NOMATCH)) {
2174 		IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2175 	}
2176 	fr = fin->fin_fr;
2177 
2178 	/*
2179 	 * Apply packets per second rate-limiting to a rule as required.
2180 	 */
2181 	if ((fr != NULL) && (fr->fr_pps != 0) &&
2182 	    !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2183 		pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2184 		pass |= FR_BLOCK;
2185 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2186 	}
2187 
2188 	/*
2189 	 * If we fail to add a packet to the authorization queue, then we
2190 	 * drop the packet later.  However, if it was added then pretend
2191 	 * we've dropped it already.
2192 	 */
2193 	if (FR_ISAUTH(pass)) {
2194 		if (fr_newauth(fin->fin_m, fin) != 0) {
2195 #ifdef	_KERNEL
2196 			fin->fin_m = *fin->fin_mp = NULL;
2197 #else
2198 			;
2199 #endif
2200 			fin->fin_error = 0;
2201 		} else
2202 			fin->fin_error = ENOSPC;
2203 	}
2204 
2205 	if ((fr != NULL) && (fr->fr_func != NULL) &&
2206 	    (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2207 		(void) (*fr->fr_func)(fin, &pass);
2208 
2209 	/*
2210 	 * If a rule is a pre-auth rule, check again in the list of rules
2211 	 * loaded for authenticated use.  It does not particulary matter
2212 	 * if this search fails because a "preauth" result, from a rule,
2213 	 * is treated as "not a pass", hence the packet is blocked.
2214 	 */
2215 	if (FR_ISPREAUTH(pass)) {
2216 		if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2217 			pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2218 	}
2219 
2220 	/*
2221 	 * If the rule has "keep frag" and the packet is actually a fragment,
2222 	 * then create a fragment state entry.
2223 	 */
2224 	if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2225 		if (fin->fin_flx & FI_FRAG) {
2226 			if (fr_newfrag(fin, pass) == -1) {
2227 				IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2228 			} else {
2229 				IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2230 			}
2231 		} else {
2232 			IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2233 		}
2234 	}
2235 
2236 	/*
2237 	 * Finally, if we've asked to track state for this packet, set it up.
2238 	 */
2239 	if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2240 		if (fr_addstate(fin, NULL, 0) != NULL) {
2241 			IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2242 		} else {
2243 			IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2244 			if (FR_ISPASS(pass)) {
2245 				pass &= ~FR_CMDMASK;
2246 				pass |= FR_BLOCK;
2247 			}
2248 		}
2249 	}
2250 
2251 	fr = fin->fin_fr;
2252 
2253 	if (passp != NULL)
2254 		*passp = pass;
2255 
2256 	return fr;
2257 }
2258 
2259 
2260 /* ------------------------------------------------------------------------ */
2261 /* Function:    fr_check                                                    */
2262 /* Returns:     int -  0 == packet allowed through,                         */
2263 /*              User space:                                                 */
2264 /*                    -1 == packet blocked                                  */
2265 /*                     1 == packet not matched                              */
2266 /*                    -2 == requires authentication                         */
2267 /*              Kernel:                                                     */
2268 /*                   > 0 == filter error # for packet                       */
2269 /* Parameters: ip(I)   - pointer to start of IPv4/6 packet                  */
2270 /*             hlen(I) - length of header                                   */
2271 /*             ifp(I)  - pointer to interface this packet is on             */
2272 /*             out(I)  - 0 == packet going in, 1 == packet going out        */
2273 /*             mp(IO)  - pointer to caller's buffer pointer that holds this */
2274 /*                       IP packet.                                         */
2275 /* Solaris & HP-UX ONLY :                                                   */
2276 /*             qpi(I)  - pointer to STREAMS queue information for this      */
2277 /*                       interface & direction.                             */
2278 /*                                                                          */
2279 /* fr_check() is the master function for all IPFilter packet processing.    */
2280 /* It orchestrates: Network Address Translation (NAT), checking for packet  */
2281 /* authorisation (or pre-authorisation), presence of related state info.,   */
2282 /* generating log entries, IP packet accounting, routing of packets as      */
2283 /* directed by firewall rules and of course whether or not to allow the     */
2284 /* packet to be further processed by the kernel.                            */
2285 /*                                                                          */
2286 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer  */
2287 /* freed.  Packets passed may be returned with the pointer pointed to by    */
2288 /* by "mp" changed to a new buffer.                                         */
2289 /* ------------------------------------------------------------------------ */
2290 int fr_check(ip, hlen, ifp, out
2291 #if defined(_KERNEL) && defined(MENTAT)
2292 , qif, mp, ifs)
2293 void *qif;
2294 #else
2295 , mp, ifs)
2296 #endif
2297 mb_t **mp;
2298 ip_t *ip;
2299 int hlen;
2300 void *ifp;
2301 int out;
2302 ipf_stack_t *ifs;
2303 {
2304 	/*
2305 	 * The above really sucks, but short of writing a diff
2306 	 */
2307 	fr_info_t frinfo;
2308 	fr_info_t *fin = &frinfo;
2309 	u_32_t pass;
2310 	frentry_t *fr = NULL;
2311 	int v = IP_V(ip);
2312 	mb_t *mc = NULL;
2313 	mb_t *m;
2314 #ifdef USE_INET6
2315 	ip6_t *ip6;
2316 #endif
2317 #ifdef	_KERNEL
2318 # ifdef MENTAT
2319 	qpktinfo_t *qpi = qif;
2320 #endif
2321 #endif
2322 
2323 	SPL_INT(s);
2324 	pass = ifs->ifs_fr_pass;
2325 
2326 	/*
2327 	 * The first part of fr_check() deals with making sure that what goes
2328 	 * into the filtering engine makes some sense.  Information about the
2329 	 * the packet is distilled, collected into a fr_info_t structure and
2330 	 * the an attempt to ensure the buffer the packet is in is big enough
2331 	 * to hold all the required packet headers.
2332 	 */
2333 #ifdef	_KERNEL
2334 # ifdef MENTAT
2335 	if (!OK_32PTR(ip))
2336 		return 2;
2337 # endif
2338 
2339 
2340 	if (ifs->ifs_fr_running <= 0) {
2341 		return 0;
2342 	}
2343 
2344 	bzero((char *)fin, sizeof(*fin));
2345 
2346 # ifdef MENTAT
2347 	fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2348 					 FI_BROADCAST);
2349 	m = qpi->qpi_m;
2350 	fin->fin_qfm = m;
2351 	fin->fin_qpi = qpi;
2352 # else /* MENTAT */
2353 
2354 	m = *mp;
2355 
2356 #  if defined(M_MCAST)
2357 	if ((m->m_flags & M_MCAST) != 0)
2358 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2359 #  endif
2360 #  if defined(M_MLOOP)
2361 	if ((m->m_flags & M_MLOOP) != 0)
2362 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2363 #  endif
2364 #  if defined(M_BCAST)
2365 	if ((m->m_flags & M_BCAST) != 0)
2366 		fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2367 #  endif
2368 #  ifdef M_CANFASTFWD
2369 	/*
2370 	 * XXX For now, IP Filter and fast-forwarding of cached flows
2371 	 * XXX are mutually exclusive.  Eventually, IP Filter should
2372 	 * XXX get a "can-fast-forward" filter rule.
2373 	 */
2374 	m->m_flags &= ~M_CANFASTFWD;
2375 #  endif /* M_CANFASTFWD */
2376 #  ifdef CSUM_DELAY_DATA
2377 	/*
2378 	 * disable delayed checksums.
2379 	 */
2380 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2381 		in_delayed_cksum(m);
2382 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2383 	}
2384 #  endif /* CSUM_DELAY_DATA */
2385 # endif /* MENTAT */
2386 #else
2387 
2388 	bzero((char *)fin, sizeof(*fin));
2389 	m = *mp;
2390 #endif /* _KERNEL */
2391 
2392 	fin->fin_v = v;
2393 	fin->fin_m = m;
2394 	fin->fin_ip = ip;
2395 	fin->fin_mp = mp;
2396 	fin->fin_out = out;
2397 	fin->fin_ifp = ifp;
2398 	fin->fin_error = ENETUNREACH;
2399 	fin->fin_hlen = (u_short)hlen;
2400 	fin->fin_dp = (char *)ip + hlen;
2401 	fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2402 	fin->fin_ifs = ifs;
2403 
2404 	SPL_NET(s);
2405 
2406 #ifdef	USE_INET6
2407 	if (v == 6) {
2408 		IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2409 		/*
2410 		 * Jumbo grams are quite likely too big for internal buffer
2411 		 * structures to handle comfortably, for now, so just drop
2412 		 * them.
2413 		 */
2414 		ip6 = (ip6_t *)ip;
2415 		fin->fin_plen = ntohs(ip6->ip6_plen);
2416 		if (fin->fin_plen == 0) {
2417 			READ_ENTER(&ifs->ifs_ipf_mutex);
2418 			pass = FR_BLOCK|FR_NOMATCH;
2419 			goto filtered;
2420 		}
2421 		fin->fin_plen += sizeof(ip6_t);
2422 	} else
2423 #endif
2424 	{
2425 #if (OpenBSD >= 200311) && defined(_KERNEL)
2426 		ip->ip_len = ntohs(ip->ip_len);
2427 		ip->ip_off = ntohs(ip->ip_off);
2428 #endif
2429 		fin->fin_plen = ip->ip_len;
2430 	}
2431 
2432 	if (fr_makefrip(hlen, ip, fin) == -1) {
2433 		READ_ENTER(&ifs->ifs_ipf_mutex);
2434 		pass = FR_BLOCK;
2435 		goto filtered;
2436 	}
2437 
2438 	/*
2439 	 * For at least IPv6 packets, if a m_pullup() fails then this pointer
2440 	 * becomes NULL and so we have no packet to free.
2441 	 */
2442 	if (*fin->fin_mp == NULL)
2443 		goto finished;
2444 
2445 	if (!out) {
2446 		if (v == 4) {
2447 #ifdef _KERNEL
2448 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2449 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2450 				fin->fin_flx |= FI_BADSRC;
2451 			}
2452 #endif
2453 			if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2454 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2455 				fin->fin_flx |= FI_LOWTTL;
2456 			}
2457 		}
2458 #ifdef USE_INET6
2459 		else  if (v == 6) {
2460 			ip6 = (ip6_t *)ip;
2461 #ifdef _KERNEL
2462 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2463 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2464 				fin->fin_flx |= FI_BADSRC;
2465 			}
2466 #endif
2467 			if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2468 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2469 				fin->fin_flx |= FI_LOWTTL;
2470 			}
2471 		}
2472 #endif
2473 	}
2474 
2475 	if (fin->fin_flx & FI_SHORT) {
2476 		IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2477 	}
2478 
2479 	READ_ENTER(&ifs->ifs_ipf_mutex);
2480 
2481 	/*
2482 	 * Check auth now.  This, combined with the check below to see if apass
2483 	 * is 0 is to ensure that we don't count the packet twice, which can
2484 	 * otherwise occur when we reprocess it.  As it is, we only count it
2485 	 * after it has no auth. table matchup.  This also stops NAT from
2486 	 * occuring until after the packet has been auth'd.
2487 	 */
2488 	fr = fr_checkauth(fin, &pass);
2489 	if (!out) {
2490 		switch (fin->fin_v)
2491 		{
2492 		case 4 :
2493 			if (fr_checknatin(fin, &pass) == -1) {
2494 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2495 				goto finished;
2496 			}
2497 			break;
2498 #ifdef	USE_INET6
2499 		case 6 :
2500 			if (fr_checknat6in(fin, &pass) == -1) {
2501 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2502 				goto finished;
2503 			}
2504 			break;
2505 #endif
2506 		default :
2507 			break;
2508 		}
2509 	}
2510 	if (!out)
2511 		(void) fr_acctpkt(fin, NULL);
2512 
2513 	if (fr == NULL)
2514 		if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2515 			fr = fr_knownfrag(fin, &pass);
2516 	if (fr == NULL)
2517 		fr = fr_checkstate(fin, &pass);
2518 
2519 	if ((pass & FR_NOMATCH) || (fr == NULL))
2520 		fr = fr_firewall(fin, &pass);
2521 
2522 	fin->fin_fr = fr;
2523 
2524 	/*
2525 	 * Only count/translate packets which will be passed on, out the
2526 	 * interface.
2527 	 */
2528 	if (out && FR_ISPASS(pass)) {
2529 		(void) fr_acctpkt(fin, NULL);
2530 
2531 		switch (fin->fin_v)
2532 		{
2533 		case 4 :
2534 			if (fr_checknatout(fin, &pass) == -1) {
2535 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2536 				goto finished;
2537 			}
2538 			break;
2539 #ifdef	USE_INET6
2540 		case 6 :
2541 			if (fr_checknat6out(fin, &pass) == -1) {
2542 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2543 				goto finished;
2544 			}
2545 			break;
2546 #endif
2547 		default :
2548 			break;
2549 		}
2550 
2551 		if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2552 			if (fr_updateipid(fin) == -1) {
2553 				IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2554 				pass &= ~FR_CMDMASK;
2555 				pass |= FR_BLOCK;
2556 			} else {
2557 				IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2558 			}
2559 		}
2560 	}
2561 
2562 #ifdef	IPFILTER_LOG
2563 	if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2564 		(void) fr_dolog(fin, &pass);
2565 	}
2566 #endif
2567 
2568 	if (fin->fin_state != NULL)
2569 		fr_statederef((ipstate_t **)&fin->fin_state, ifs);
2570 
2571 	/*
2572 	 * Only allow FR_DUP to work if a rule matched - it makes no sense to
2573 	 * set FR_DUP as a "default" as there are no instructions about where
2574 	 * to send the packet.  Use fin_m here because it may have changed
2575 	 * (without an update of 'm') in prior processing.
2576 	 */
2577 	if ((fr != NULL) && (pass & FR_DUP)) {
2578 		mc = M_DUPLICATE(fin->fin_m);
2579 	}
2580 
2581 	if (pass & (FR_RETRST|FR_RETICMP)) {
2582 		/*
2583 		 * Should we return an ICMP packet to indicate error
2584 		 * status passing through the packet filter ?
2585 		 * WARNING: ICMP error packets AND TCP RST packets should
2586 		 * ONLY be sent in repsonse to incoming packets.  Sending them
2587 		 * in response to outbound packets can result in a panic on
2588 		 * some operating systems.
2589 		 */
2590 		if (!out) {
2591 			nat_t *savenat;
2592 
2593 			/*
2594 			 * In case this packet is associated with a NAT entry,
2595 			 * we need to save a pointer to it, set it to NULL,
2596 			 * and restore the pointer after any return packets
2597 			 * have been sent.
2598 			 *
2599 			 * We must do this to keep calls to fr_send_icmp_err()
2600 			 * and fr_send_reset() from trying to dereference the
2601 			 * entry; as the dereference happens (for all packets)
2602  			 * later on.
2603 			 */
2604 			savenat = fin->fin_nat;
2605 			fin->fin_nat = NULL;
2606 
2607 			if (pass & FR_RETICMP) {
2608 				int dst;
2609 
2610 				if ((pass & FR_RETMASK) == FR_FAKEICMP)
2611 					dst = 1;
2612 				else
2613 					dst = 0;
2614 				(void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2615 				IPF_BUMP(ifs->ifs_frstats[0].fr_ret);
2616 			} else if (((pass & FR_RETMASK) == FR_RETRST) &&
2617 				   !(fin->fin_flx & FI_SHORT)) {
2618 				if (fr_send_reset(fin) == 0) {
2619 					IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2620 				}
2621 			}
2622 
2623 			fin->fin_nat = savenat;
2624 		} else {
2625 			if (pass & FR_RETRST)
2626 				fin->fin_error = ECONNRESET;
2627 		}
2628 	}
2629 
2630 	/*
2631 	 * fr_send_icmp_err() and fr_send_reset() are the last functions that
2632 	 * could require use of a NAT entry.  If this packet is associated with
2633 	 * one, it's time to clean up our reference to it.
2634 	 *
2635 	 * Note:
2636 	 * If a new NAT entry was created as result of this packet, and the
2637 	 * packet was blocked, then the entry should be forcibly removed
2638 	 * from the table.  Otherwise, it's sufficient to simply give up the
2639 	 * reference to it.
2640 	 */
2641 	if (fin->fin_nat != NULL) {
2642 		if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT)) {
2643 			WRITE_ENTER(&ifs->ifs_ipf_nat);
2644 			(void) nat_delete((nat_t *)fin->fin_nat, NL_DESTROY, ifs);
2645 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
2646 			fin->fin_nat = NULL;
2647 		} else {
2648 			fr_natderef((nat_t **)&fin->fin_nat, ifs);
2649 		}
2650 	}
2651 
2652 	/*
2653 	 * If we didn't drop off the bottom of the list of rules (and thus
2654 	 * the 'current' rule fr is not NULL), then we may have some extra
2655 	 * instructions about what to do with a packet.
2656 	 * Once we're finished return to our caller, freeing the packet if
2657 	 * we are dropping it (* BSD ONLY *).
2658 	 * Reassign m from fin_m as we may have a new buffer, now.
2659 	 */
2660 filtered:
2661 	m = fin->fin_m;
2662 
2663 	if (fr != NULL) {
2664 		frdest_t *fdp;
2665 
2666 		fdp = &fr->fr_tifs[fin->fin_rev];
2667 
2668 		if (!out && (pass & FR_FASTROUTE)) {
2669 			/*
2670 			 * For fastroute rule, no destioation interface defined
2671 			 * so pass NULL as the frdest_t parameter
2672 			 */
2673 			(void) fr_fastroute(m, mp, fin, NULL);
2674 			m = *mp = NULL;
2675 		} else if ((fdp->fd_ifp != NULL) &&
2676 			   (fdp->fd_ifp != (struct ifnet *)-1)) {
2677 			/* this is for to rules: */
2678 			(void) fr_fastroute(m, mp, fin, fdp);
2679 			m = *mp = NULL;
2680 		}
2681 
2682 		/*
2683 		 * Generate a duplicated packet.
2684 		 */
2685 		if (mc != NULL)
2686 			(void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2687 	}
2688 
2689 	/*
2690 	 * This late because the likes of fr_fastroute() use fin_fr.
2691 	 */
2692 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2693 
2694 finished:
2695 	if (!FR_ISPASS(pass)) {
2696 		IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2697 		if (*mp != NULL) {
2698 			FREE_MB_T(*mp);
2699 			m = *mp = NULL;
2700 		}
2701 	} else {
2702 		IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2703 #if defined(_KERNEL) && defined(__sgi)
2704 		if ((fin->fin_hbuf != NULL) &&
2705 		    (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2706 			COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2707 		}
2708 #endif
2709 	}
2710 
2711 	SPL_X(s);
2712 
2713 #ifdef _KERNEL
2714 # if OpenBSD >= 200311
2715 	if (FR_ISPASS(pass) && (v == 4)) {
2716 		ip = fin->fin_ip;
2717 		ip->ip_len = ntohs(ip->ip_len);
2718 		ip->ip_off = ntohs(ip->ip_off);
2719 	}
2720 # endif
2721 	return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2722 #else /* _KERNEL */
2723 	FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2724 	if ((pass & FR_NOMATCH) != 0)
2725 		return 1;
2726 
2727 	if ((pass & FR_RETMASK) != 0)
2728 		switch (pass & FR_RETMASK)
2729 		{
2730 		case FR_RETRST :
2731 			return 3;
2732 		case FR_RETICMP :
2733 			return 4;
2734 		case FR_FAKEICMP :
2735 			return 5;
2736 		}
2737 
2738 	switch (pass & FR_CMDMASK)
2739 	{
2740 	case FR_PASS :
2741 		return 0;
2742 	case FR_BLOCK :
2743 		return -1;
2744 	case FR_AUTH :
2745 		return -2;
2746 	case FR_ACCOUNT :
2747 		return -3;
2748 	case FR_PREAUTH :
2749 		return -4;
2750 	}
2751 	return 2;
2752 #endif /* _KERNEL */
2753 }
2754 
2755 
2756 #ifdef	IPFILTER_LOG
2757 /* ------------------------------------------------------------------------ */
2758 /* Function:    fr_dolog                                                    */
2759 /* Returns:     frentry_t* - returns contents of fin_fr (no change made)    */
2760 /* Parameters:  fin(I) - pointer to packet information                      */
2761 /*              passp(IO) - pointer to current/new filter decision (unused) */
2762 /*                                                                          */
2763 /* Checks flags set to see how a packet should be logged, if it is to be    */
2764 /* logged.  Adjust statistics based on its success or not.                  */
2765 /* ------------------------------------------------------------------------ */
2766 frentry_t *fr_dolog(fin, passp)
2767 fr_info_t *fin;
2768 u_32_t *passp;
2769 {
2770 	u_32_t pass;
2771 	int out;
2772 	ipf_stack_t *ifs = fin->fin_ifs;
2773 
2774 	out = fin->fin_out;
2775 	pass = *passp;
2776 
2777 	if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2778 		pass |= FF_LOGNOMATCH;
2779 		IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2780 		goto logit;
2781 	} else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2782 	    (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2783 		if ((pass & FR_LOGMASK) != FR_LOGP)
2784 			pass |= FF_LOGPASS;
2785 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2786 		goto logit;
2787 	} else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2788 		   (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2789 		if ((pass & FR_LOGMASK) != FR_LOGB)
2790 			pass |= FF_LOGBLOCK;
2791 		IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2792 logit:
2793 		if (ipflog(fin, pass) == -1) {
2794 			IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2795 
2796 			/*
2797 			 * If the "or-block" option has been used then
2798 			 * block the packet if we failed to log it.
2799 			 */
2800 			if ((pass & FR_LOGORBLOCK) &&
2801 			    FR_ISPASS(pass)) {
2802 				pass &= ~FR_CMDMASK;
2803 				pass |= FR_BLOCK;
2804 			}
2805 		}
2806 		*passp = pass;
2807 	}
2808 
2809 	return fin->fin_fr;
2810 }
2811 #endif /* IPFILTER_LOG */
2812 
2813 
2814 /* ------------------------------------------------------------------------ */
2815 /* Function:    ipf_cksum                                                   */
2816 /* Returns:     u_short - IP header checksum                                */
2817 /* Parameters:  addr(I) - pointer to start of buffer to checksum            */
2818 /*              len(I)  - length of buffer in bytes                         */
2819 /*                                                                          */
2820 /* Calculate the two's complement 16 bit checksum of the buffer passed.     */
2821 /*                                                                          */
2822 /* N.B.: addr should be 16bit aligned.                                      */
2823 /* ------------------------------------------------------------------------ */
2824 u_short ipf_cksum(addr, len)
2825 u_short *addr;
2826 int len;
2827 {
2828 	u_32_t sum = 0;
2829 
2830 	for (sum = 0; len > 1; len -= 2)
2831 		sum += *addr++;
2832 
2833 	/* mop up an odd byte, if necessary */
2834 	if (len == 1)
2835 		sum += *(u_char *)addr;
2836 
2837 	/*
2838 	 * add back carry outs from top 16 bits to low 16 bits
2839 	 */
2840 	sum = (sum >> 16) + (sum & 0xffff);	/* add hi 16 to low 16 */
2841 	sum += (sum >> 16);			/* add carry */
2842 	return (u_short)(~sum);
2843 }
2844 
2845 
2846 /* ------------------------------------------------------------------------ */
2847 /* Function:    fr_cksum                                                    */
2848 /* Returns:     u_short - layer 4 checksum                                  */
2849 /* Parameters:  m(I  )     - pointer to buffer holding packet               */
2850 /*              ip(I)      - pointer to IP header                           */
2851 /*              l4proto(I) - protocol to caclulate checksum for             */
2852 /*              l4hdr(I)   - pointer to layer 4 header                      */
2853 /*                                                                          */
2854 /* Calculates the TCP checksum for the packet held in "m", using the data   */
2855 /* in the IP header "ip" to seed it.                                        */
2856 /*                                                                          */
2857 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2858 /* and the TCP header.  We also assume that data blocks aren't allocated in */
2859 /* odd sizes.                                                               */
2860 /*                                                                          */
2861 /* Expects ip_len to be in host byte order when called.                     */
2862 /* ------------------------------------------------------------------------ */
2863 u_short fr_cksum(m, ip, l4proto, l4hdr)
2864 mb_t *m;
2865 ip_t *ip;
2866 int l4proto;
2867 void *l4hdr;
2868 {
2869 	u_short *sp, slen, sumsave, l4hlen, *csump;
2870 	u_int sum, sum2;
2871 	int hlen;
2872 #ifdef	USE_INET6
2873 	ip6_t *ip6;
2874 #endif
2875 
2876 	csump = NULL;
2877 	sumsave = 0;
2878 	l4hlen = 0;
2879 	sp = NULL;
2880 	slen = 0;
2881 	hlen = 0;
2882 	sum = 0;
2883 
2884 	/*
2885 	 * Add up IP Header portion
2886 	 */
2887 #ifdef	USE_INET6
2888 	if (IP_V(ip) == 4) {
2889 #endif
2890 		hlen = IP_HL(ip) << 2;
2891 		slen = ip->ip_len - hlen;
2892 		sum = htons((u_short)l4proto);
2893 		sum += htons(slen);
2894 		sp = (u_short *)&ip->ip_src;
2895 		sum += *sp++;	/* ip_src */
2896 		sum += *sp++;
2897 		sum += *sp++;	/* ip_dst */
2898 		sum += *sp++;
2899 #ifdef	USE_INET6
2900 	} else if (IP_V(ip) == 6) {
2901 		ip6 = (ip6_t *)ip;
2902 		hlen = sizeof(*ip6);
2903 		slen = ntohs(ip6->ip6_plen);
2904 		sum = htons((u_short)l4proto);
2905 		sum += htons(slen);
2906 		sp = (u_short *)&ip6->ip6_src;
2907 		sum += *sp++;	/* ip6_src */
2908 		sum += *sp++;
2909 		sum += *sp++;
2910 		sum += *sp++;
2911 		sum += *sp++;
2912 		sum += *sp++;
2913 		sum += *sp++;
2914 		sum += *sp++;
2915 		sum += *sp++;	/* ip6_dst */
2916 		sum += *sp++;
2917 		sum += *sp++;
2918 		sum += *sp++;
2919 		sum += *sp++;
2920 		sum += *sp++;
2921 		sum += *sp++;
2922 		sum += *sp++;
2923 	}
2924 #endif
2925 
2926 	switch (l4proto)
2927 	{
2928 	case IPPROTO_UDP :
2929 		csump = &((udphdr_t *)l4hdr)->uh_sum;
2930 		l4hlen = sizeof(udphdr_t);
2931 		break;
2932 
2933 	case IPPROTO_TCP :
2934 		csump = &((tcphdr_t *)l4hdr)->th_sum;
2935 		l4hlen = sizeof(tcphdr_t);
2936 		break;
2937 	case IPPROTO_ICMP :
2938 		csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
2939 		l4hlen = 4;
2940 		sum = 0;
2941 		break;
2942 	default :
2943 		break;
2944 	}
2945 
2946 	if (csump != NULL) {
2947 		sumsave = *csump;
2948 		*csump = 0;
2949 	}
2950 
2951 	l4hlen = l4hlen;	/* LINT */
2952 
2953 #ifdef	_KERNEL
2954 # ifdef MENTAT
2955 	{
2956 	void *rp = m->b_rptr;
2957 
2958 	if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
2959 		m->b_rptr = (u_char *)ip;
2960 	sum2 = ip_cksum(m, hlen, sum);	/* hlen == offset */
2961 	m->b_rptr = rp;
2962 	sum2 = (sum2 & 0xffff) + (sum2 >> 16);
2963 	sum2 = ~sum2 & 0xffff;
2964 	}
2965 # else /* MENTAT */
2966 #  if defined(BSD) || defined(sun)
2967 #   if BSD >= 199103
2968 	m->m_data += hlen;
2969 #   else
2970 	m->m_off += hlen;
2971 #   endif
2972 	m->m_len -= hlen;
2973 	sum2 = in_cksum(m, slen);
2974 	m->m_len += hlen;
2975 #   if BSD >= 199103
2976 	m->m_data -= hlen;
2977 #   else
2978 	m->m_off -= hlen;
2979 #   endif
2980 	/*
2981 	 * Both sum and sum2 are partial sums, so combine them together.
2982 	 */
2983 	sum += ~sum2 & 0xffff;
2984 	while (sum > 0xffff)
2985 		sum = (sum & 0xffff) + (sum >> 16);
2986 	sum2 = ~sum & 0xffff;
2987 #  else /* defined(BSD) || defined(sun) */
2988 {
2989 	union {
2990 		u_char	c[2];
2991 		u_short	s;
2992 	} bytes;
2993 	u_short len = ip->ip_len;
2994 #   if defined(__sgi)
2995 	int add;
2996 #   endif
2997 
2998 	/*
2999 	 * Add up IP Header portion
3000 	 */
3001 	if (sp != (u_short *)l4hdr)
3002 		sp = (u_short *)l4hdr;
3003 
3004 	switch (l4proto)
3005 	{
3006 	case IPPROTO_UDP :
3007 		sum += *sp++;	/* sport */
3008 		sum += *sp++;	/* dport */
3009 		sum += *sp++;	/* udp length */
3010 		sum += *sp++;	/* checksum */
3011 		break;
3012 
3013 	case IPPROTO_TCP :
3014 		sum += *sp++;	/* sport */
3015 		sum += *sp++;	/* dport */
3016 		sum += *sp++;	/* seq */
3017 		sum += *sp++;
3018 		sum += *sp++;	/* ack */
3019 		sum += *sp++;
3020 		sum += *sp++;	/* off */
3021 		sum += *sp++;	/* win */
3022 		sum += *sp++;	/* checksum */
3023 		sum += *sp++;	/* urp */
3024 		break;
3025 	case IPPROTO_ICMP :
3026 		sum = *sp++;	/* type/code */
3027 		sum += *sp++;	/* checksum */
3028 		break;
3029 	}
3030 
3031 #   ifdef	__sgi
3032 	/*
3033 	 * In case we had to copy the IP & TCP header out of mbufs,
3034 	 * skip over the mbuf bits which are the header
3035 	 */
3036 	if ((caddr_t)ip != mtod(m, caddr_t)) {
3037 		hlen = (caddr_t)sp - (caddr_t)ip;
3038 		while (hlen) {
3039 			add = MIN(hlen, m->m_len);
3040 			sp = (u_short *)(mtod(m, caddr_t) + add);
3041 			hlen -= add;
3042 			if (add == m->m_len) {
3043 				m = m->m_next;
3044 				if (!hlen) {
3045 					if (!m)
3046 						break;
3047 					sp = mtod(m, u_short *);
3048 				}
3049 				PANIC((!m),("fr_cksum(1): not enough data"));
3050 			}
3051 		}
3052 	}
3053 #   endif
3054 
3055 	len -= (l4hlen + hlen);
3056 	if (len <= 0)
3057 		goto nodata;
3058 
3059 	while (len > 1) {
3060 		if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3061 			m = m->m_next;
3062 			PANIC((!m),("fr_cksum(2): not enough data"));
3063 			sp = mtod(m, u_short *);
3064 		}
3065 		if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3066 			bytes.c[0] = *(u_char *)sp;
3067 			m = m->m_next;
3068 			PANIC((!m),("fr_cksum(3): not enough data"));
3069 			sp = mtod(m, u_short *);
3070 			bytes.c[1] = *(u_char *)sp;
3071 			sum += bytes.s;
3072 			sp = (u_short *)((u_char *)sp + 1);
3073 		}
3074 		if ((u_long)sp & 1) {
3075 			bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3076 			sum += bytes.s;
3077 		} else
3078 			sum += *sp++;
3079 		len -= 2;
3080 	}
3081 
3082 	if (len != 0)
3083 		sum += ntohs(*(u_char *)sp << 8);
3084 nodata:
3085 	while (sum > 0xffff)
3086 		sum = (sum & 0xffff) + (sum >> 16);
3087 	sum2 = (u_short)(~sum & 0xffff);
3088 }
3089 #  endif /*  defined(BSD) || defined(sun) */
3090 # endif /* MENTAT */
3091 #else /* _KERNEL */
3092 	for (; slen > 1; slen -= 2)
3093 	        sum += *sp++;
3094 	if (slen)
3095 		sum += ntohs(*(u_char *)sp << 8);
3096 	while (sum > 0xffff)
3097 		sum = (sum & 0xffff) + (sum >> 16);
3098 	sum2 = (u_short)(~sum & 0xffff);
3099 #endif /* _KERNEL */
3100 	if (csump != NULL)
3101 		*csump = sumsave;
3102 	return sum2;
3103 }
3104 
3105 
3106 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3107     defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3108 /*
3109  * Copyright (c) 1982, 1986, 1988, 1991, 1993
3110  *	The Regents of the University of California.  All rights reserved.
3111  *
3112  * Redistribution and use in source and binary forms, with or without
3113  * modification, are permitted provided that the following conditions
3114  * are met:
3115  * 1. Redistributions of source code must retain the above copyright
3116  *    notice, this list of conditions and the following disclaimer.
3117  * 2. Redistributions in binary form must reproduce the above copyright
3118  *    notice, this list of conditions and the following disclaimer in the
3119  *    documentation and/or other materials provided with the distribution.
3120  * 3. Neither the name of the University nor the names of its contributors
3121  *    may be used to endorse or promote products derived from this software
3122  *    without specific prior written permission.
3123  *
3124  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3125  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3126  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3127  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3128  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3129  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3130  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3131  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3132  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3133  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3134  * SUCH DAMAGE.
3135  *
3136  *	@(#)uipc_mbuf.c	8.2 (Berkeley) 1/4/94
3137  * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3138  */
3139 /*
3140  * Copy data from an mbuf chain starting "off" bytes from the beginning,
3141  * continuing for "len" bytes, into the indicated buffer.
3142  */
3143 void
3144 m_copydata(m, off, len, cp)
3145 	mb_t *m;
3146 	int off;
3147 	int len;
3148 	caddr_t cp;
3149 {
3150 	unsigned count;
3151 
3152 	if (off < 0 || len < 0)
3153 		panic("m_copydata");
3154 	while (off > 0) {
3155 		if (m == 0)
3156 			panic("m_copydata");
3157 		if (off < m->m_len)
3158 			break;
3159 		off -= m->m_len;
3160 		m = m->m_next;
3161 	}
3162 	while (len > 0) {
3163 		if (m == 0)
3164 			panic("m_copydata");
3165 		count = MIN(m->m_len - off, len);
3166 		bcopy(mtod(m, caddr_t) + off, cp, count);
3167 		len -= count;
3168 		cp += count;
3169 		off = 0;
3170 		m = m->m_next;
3171 	}
3172 }
3173 
3174 
3175 /*
3176  * Copy data from a buffer back into the indicated mbuf chain,
3177  * starting "off" bytes from the beginning, extending the mbuf
3178  * chain if necessary.
3179  */
3180 void
3181 m_copyback(m0, off, len, cp)
3182 	struct	mbuf *m0;
3183 	int off;
3184 	int len;
3185 	caddr_t cp;
3186 {
3187 	int mlen;
3188 	struct mbuf *m = m0, *n;
3189 	int totlen = 0;
3190 
3191 	if (m0 == 0)
3192 		return;
3193 	while (off > (mlen = m->m_len)) {
3194 		off -= mlen;
3195 		totlen += mlen;
3196 		if (m->m_next == 0) {
3197 			n = m_getclr(M_DONTWAIT, m->m_type);
3198 			if (n == 0)
3199 				goto out;
3200 			n->m_len = min(MLEN, len + off);
3201 			m->m_next = n;
3202 		}
3203 		m = m->m_next;
3204 	}
3205 	while (len > 0) {
3206 		mlen = min(m->m_len - off, len);
3207 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3208 		cp += mlen;
3209 		len -= mlen;
3210 		mlen += off;
3211 		off = 0;
3212 		totlen += mlen;
3213 		if (len == 0)
3214 			break;
3215 		if (m->m_next == 0) {
3216 			n = m_get(M_DONTWAIT, m->m_type);
3217 			if (n == 0)
3218 				break;
3219 			n->m_len = min(MLEN, len);
3220 			m->m_next = n;
3221 		}
3222 		m = m->m_next;
3223 	}
3224 out:
3225 #if 0
3226 	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3227 		m->m_pkthdr.len = totlen;
3228 #endif
3229 	return;
3230 }
3231 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3232 
3233 
3234 /* ------------------------------------------------------------------------ */
3235 /* Function:    fr_findgroup                                                */
3236 /* Returns:     frgroup_t * - NULL = group not found, else pointer to group */
3237 /* Parameters:  group(I) - group name to search for                         */
3238 /*              unit(I)  - device to which this group belongs               */
3239 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3240 /*              fgpp(O)  - pointer to place to store pointer to the pointer */
3241 /*                         to where to add the next (last) group or where   */
3242 /*                         to delete group from.                            */
3243 /*                                                                          */
3244 /* Search amongst the defined groups for a particular group number.         */
3245 /* ------------------------------------------------------------------------ */
3246 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3247 char *group;
3248 minor_t unit;
3249 int set;
3250 frgroup_t ***fgpp;
3251 ipf_stack_t *ifs;
3252 {
3253 	frgroup_t *fg, **fgp;
3254 
3255 	/*
3256 	 * Which list of groups to search in is dependent on which list of
3257 	 * rules are being operated on.
3258 	 */
3259 	fgp = &ifs->ifs_ipfgroups[unit][set];
3260 
3261 	while ((fg = *fgp) != NULL) {
3262 		if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3263 			break;
3264 		else
3265 			fgp = &fg->fg_next;
3266 	}
3267 	if (fgpp != NULL)
3268 		*fgpp = fgp;
3269 	return fg;
3270 }
3271 
3272 
3273 /* ------------------------------------------------------------------------ */
3274 /* Function:    fr_addgroup                                                 */
3275 /* Returns:     frgroup_t * - NULL == did not create group,                 */
3276 /*                            != NULL == pointer to the group               */
3277 /* Parameters:  num(I)   - group number to add                              */
3278 /*              head(I)  - rule pointer that is using this as the head      */
3279 /*              flags(I) - rule flags which describe the type of rule it is */
3280 /*              unit(I)  - device to which this group will belong to        */
3281 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3282 /* Write Locks: ipf_mutex                                                   */
3283 /*                                                                          */
3284 /* Add a new group head, or if it already exists, increase the reference    */
3285 /* count to it.                                                             */
3286 /* ------------------------------------------------------------------------ */
3287 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3288 char *group;
3289 void *head;
3290 u_32_t flags;
3291 minor_t unit;
3292 int set;
3293 ipf_stack_t *ifs;
3294 {
3295 	frgroup_t *fg, **fgp;
3296 	u_32_t gflags;
3297 
3298 	if (group == NULL)
3299 		return NULL;
3300 
3301 	if (unit == IPL_LOGIPF && *group == '\0')
3302 		return NULL;
3303 
3304 	fgp = NULL;
3305 	gflags = flags & FR_INOUT;
3306 
3307 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3308 	if (fg != NULL) {
3309 		if (fg->fg_flags == 0)
3310 			fg->fg_flags = gflags;
3311 		else if (gflags != fg->fg_flags)
3312 			return NULL;
3313 		fg->fg_ref++;
3314 		return fg;
3315 	}
3316 	KMALLOC(fg, frgroup_t *);
3317 	if (fg != NULL) {
3318 		fg->fg_head = head;
3319 		fg->fg_start = NULL;
3320 		fg->fg_next = *fgp;
3321 		bcopy(group, fg->fg_name, FR_GROUPLEN);
3322 		fg->fg_flags = gflags;
3323 		fg->fg_ref = 1;
3324 		*fgp = fg;
3325 	}
3326 	return fg;
3327 }
3328 
3329 
3330 /* ------------------------------------------------------------------------ */
3331 /* Function:    fr_delgroup                                                 */
3332 /* Returns:     Nil                                                         */
3333 /* Parameters:  group(I) - group name to delete                             */
3334 /*              unit(I)  - device to which this group belongs               */
3335 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3336 /* Write Locks: ipf_mutex                                                   */
3337 /*                                                                          */
3338 /* Attempt to delete a group head.                                          */
3339 /* Only do this when its reference count reaches 0.                         */
3340 /* ------------------------------------------------------------------------ */
3341 void fr_delgroup(group, unit, set, ifs)
3342 char *group;
3343 minor_t unit;
3344 int set;
3345 ipf_stack_t *ifs;
3346 {
3347 	frgroup_t *fg, **fgp;
3348 
3349 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3350 	if (fg == NULL)
3351 		return;
3352 
3353 	fg->fg_ref--;
3354 	if (fg->fg_ref == 0) {
3355 		*fgp = fg->fg_next;
3356 		KFREE(fg);
3357 	}
3358 }
3359 
3360 
3361 /* ------------------------------------------------------------------------ */
3362 /* Function:    fr_getrulen                                                 */
3363 /* Returns:     frentry_t * - NULL == not found, else pointer to rule n     */
3364 /* Parameters:  unit(I)  - device for which to count the rule's number      */
3365 /*              flags(I) - which set of rules to find the rule in           */
3366 /*              group(I) - group name                                       */
3367 /*              n(I)     - rule number to find                              */
3368 /*                                                                          */
3369 /* Find rule # n in group # g and return a pointer to it.  Return NULl if   */
3370 /* group # g doesn't exist or there are less than n rules in the group.     */
3371 /* ------------------------------------------------------------------------ */
3372 frentry_t *fr_getrulen(unit, group, n, ifs)
3373 int unit;
3374 char *group;
3375 u_32_t n;
3376 ipf_stack_t *ifs;
3377 {
3378 	frentry_t *fr;
3379 	frgroup_t *fg;
3380 
3381 	fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3382 	if (fg == NULL)
3383 		return NULL;
3384 	for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3385 		;
3386 	if (n != 0)
3387 		return NULL;
3388 	return fr;
3389 }
3390 
3391 
3392 /* ------------------------------------------------------------------------ */
3393 /* Function:    fr_rulen                                                    */
3394 /* Returns:     int - >= 0 - rule number, -1 == search failed               */
3395 /* Parameters:  unit(I) - device for which to count the rule's number       */
3396 /*              fr(I)   - pointer to rule to match                          */
3397 /*                                                                          */
3398 /* Return the number for a rule on a specific filtering device.             */
3399 /* ------------------------------------------------------------------------ */
3400 int fr_rulen(unit, fr, ifs)
3401 int unit;
3402 frentry_t *fr;
3403 ipf_stack_t *ifs;
3404 {
3405 	frentry_t *fh;
3406 	frgroup_t *fg;
3407 	u_32_t n = 0;
3408 
3409 	if (fr == NULL)
3410 		return -1;
3411 	fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3412 	if (fg == NULL)
3413 		return -1;
3414 	for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3415 		if (fh == fr)
3416 			break;
3417 	if (fh == NULL)
3418 		return -1;
3419 	return n;
3420 }
3421 
3422 
3423 /* ------------------------------------------------------------------------ */
3424 /* Function:    frflushlist                                                 */
3425 /* Returns:     int - >= 0 - number of flushed rules                        */
3426 /* Parameters:  set(I)   - which set of rules (inactive/inactive) this is   */
3427 /*              unit(I)  - device for which to flush rules                  */
3428 /*              flags(I) - which set of rules to flush                      */
3429 /*              nfreedp(O) - pointer to int where flush count is stored     */
3430 /*              listp(I)   - pointer to list to flush pointer               */
3431 /* Write Locks: ipf_mutex                                                   */
3432 /*                                                                          */
3433 /* Recursively flush rules from the list, descending groups as they are     */
3434 /* encountered.  if a rule is the head of a group and it has lost all its   */
3435 /* group members, then also delete the group reference.  nfreedp is needed  */
3436 /* to store the accumulating count of rules removed, whereas the returned   */
3437 /* value is just the number removed from the current list.  The latter is   */
3438 /* needed to correctly adjust reference counts on rules that define groups. */
3439 /*                                                                          */
3440 /* NOTE: Rules not loaded from user space cannot be flushed.                */
3441 /* ------------------------------------------------------------------------ */
3442 static int frflushlist(set, unit, nfreedp, listp, ifs)
3443 int set;
3444 minor_t unit;
3445 int *nfreedp;
3446 frentry_t **listp;
3447 ipf_stack_t *ifs;
3448 {
3449 	int freed = 0;
3450 	frentry_t *fp;
3451 
3452 	while ((fp = *listp) != NULL) {
3453 		if ((fp->fr_type & FR_T_BUILTIN) ||
3454 		    !(fp->fr_flags & FR_COPIED)) {
3455 			listp = &fp->fr_next;
3456 			continue;
3457 		}
3458 		*listp = fp->fr_next;
3459 		if (fp->fr_grp != NULL) {
3460 			(void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3461 		}
3462 
3463 		if (fp->fr_grhead != NULL) {
3464 			fr_delgroup(fp->fr_grhead, unit, set, ifs);
3465 			*fp->fr_grhead = '\0';
3466 		}
3467 
3468 		ASSERT(fp->fr_ref > 0);
3469 		fp->fr_next = NULL;
3470 		if (fr_derefrule(&fp, ifs) == 0)
3471 			freed++;
3472 	}
3473 	*nfreedp += freed;
3474 	return freed;
3475 }
3476 
3477 
3478 /* ------------------------------------------------------------------------ */
3479 /* Function:    frflush                                                     */
3480 /* Returns:     int - >= 0 - number of flushed rules                        */
3481 /* Parameters:  unit(I)  - device for which to flush rules                  */
3482 /*              flags(I) - which set of rules to flush                      */
3483 /*                                                                          */
3484 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3485 /* and IPv6) as defined by the value of flags.                              */
3486 /* ------------------------------------------------------------------------ */
3487 int frflush(unit, proto, flags, ifs)
3488 minor_t unit;
3489 int proto, flags;
3490 ipf_stack_t *ifs;
3491 {
3492 	int flushed = 0, set;
3493 
3494 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3495 
3496 	set = ifs->ifs_fr_active;
3497 	if ((flags & FR_INACTIVE) == FR_INACTIVE)
3498 		set = 1 - set;
3499 
3500 	if (flags & FR_OUTQUE) {
3501 		if (proto == 0 || proto == 6) {
3502 			(void) frflushlist(set, unit,
3503 			    &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3504 			(void) frflushlist(set, unit,
3505 			    &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3506 		}
3507 		if (proto == 0 || proto == 4) {
3508 			(void) frflushlist(set, unit,
3509 			    &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3510 			(void) frflushlist(set, unit,
3511 			    &flushed, &ifs->ifs_ipacct[1][set], ifs);
3512 		}
3513 	}
3514 	if (flags & FR_INQUE) {
3515 		if (proto == 0 || proto == 6) {
3516 			(void) frflushlist(set, unit,
3517 			    &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3518 			(void) frflushlist(set, unit,
3519 			    &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3520 		}
3521 		if (proto == 0 || proto == 4) {
3522 			(void) frflushlist(set, unit,
3523 			    &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3524 			(void) frflushlist(set, unit,
3525 			    &flushed, &ifs->ifs_ipacct[0][set], ifs);
3526 		}
3527 	}
3528 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3529 
3530 	if (unit == IPL_LOGIPF) {
3531 		int tmp;
3532 
3533 		tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3534 		if (tmp >= 0)
3535 			flushed += tmp;
3536 	}
3537 	return flushed;
3538 }
3539 
3540 
3541 /* ------------------------------------------------------------------------ */
3542 /* Function:    memstr                                                      */
3543 /* Returns:     char *  - NULL if failed, != NULL pointer to matching bytes */
3544 /* Parameters:  src(I)  - pointer to byte sequence to match                 */
3545 /*              dst(I)  - pointer to byte sequence to search                */
3546 /*              slen(I) - match length                                      */
3547 /*              dlen(I) - length available to search in                     */
3548 /*                                                                          */
3549 /* Search dst for a sequence of bytes matching those at src and extend for  */
3550 /* slen bytes.                                                              */
3551 /* ------------------------------------------------------------------------ */
3552 char *memstr(src, dst, slen, dlen)
3553 char *src, *dst;
3554 int slen, dlen;
3555 {
3556 	char *s = NULL;
3557 
3558 	while (dlen >= slen) {
3559 		if (bcmp(src, dst, slen) == 0) {
3560 			s = dst;
3561 			break;
3562 		}
3563 		dst++;
3564 		dlen--;
3565 	}
3566 	return s;
3567 }
3568 /* ------------------------------------------------------------------------ */
3569 /* Function:    fr_fixskip                                                  */
3570 /* Returns:     Nil                                                         */
3571 /* Parameters:  listp(IO)    - pointer to start of list with skip rule      */
3572 /*              rp(I)        - rule added/removed with skip in it.          */
3573 /*              addremove(I) - adjustment (-1/+1) to make to skip count,    */
3574 /*                             depending on whether a rule was just added   */
3575 /*                             or removed.                                  */
3576 /*                                                                          */
3577 /* Adjust all the rules in a list which would have skip'd past the position */
3578 /* where we are inserting to skip to the right place given the change.      */
3579 /* ------------------------------------------------------------------------ */
3580 void fr_fixskip(listp, rp, addremove)
3581 frentry_t **listp, *rp;
3582 int addremove;
3583 {
3584 	int rules, rn;
3585 	frentry_t *fp;
3586 
3587 	rules = 0;
3588 	for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3589 		rules++;
3590 
3591 	if (!fp)
3592 		return;
3593 
3594 	for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3595 		if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3596 			fp->fr_arg += addremove;
3597 }
3598 
3599 
3600 #ifdef	_KERNEL
3601 /* ------------------------------------------------------------------------ */
3602 /* Function:    count4bits                                                  */
3603 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3604 /* Parameters:  ip(I) - 32bit IP address                                    */
3605 /*                                                                          */
3606 /* IPv4 ONLY                                                                */
3607 /* count consecutive 1's in bit mask.  If the mask generated by counting    */
3608 /* consecutive 1's is different to that passed, return -1, else return #    */
3609 /* of bits.                                                                 */
3610 /* ------------------------------------------------------------------------ */
3611 int	count4bits(ip)
3612 u_32_t	ip;
3613 {
3614 	u_32_t	ipn;
3615 	int	cnt = 0, i, j;
3616 
3617 	ip = ipn = ntohl(ip);
3618 	for (i = 32; i; i--, ipn *= 2)
3619 		if (ipn & 0x80000000)
3620 			cnt++;
3621 		else
3622 			break;
3623 	ipn = 0;
3624 	for (i = 32, j = cnt; i; i--, j--) {
3625 		ipn *= 2;
3626 		if (j > 0)
3627 			ipn++;
3628 	}
3629 	if (ipn == ip)
3630 		return cnt;
3631 	return -1;
3632 }
3633 
3634 
3635 #ifdef USE_INET6
3636 /* ------------------------------------------------------------------------ */
3637 /* Function:    count6bits                                                  */
3638 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3639 /* Parameters:  msk(I) - pointer to start of IPv6 bitmask                   */
3640 /*                                                                          */
3641 /* IPv6 ONLY                                                                */
3642 /* count consecutive 1's in bit mask.                                       */
3643 /* ------------------------------------------------------------------------ */
3644 int count6bits(msk)
3645 u_32_t *msk;
3646 {
3647 	int i = 0, k;
3648 	u_32_t j;
3649 
3650 	for (k = 3; k >= 0; k--)
3651 		if (msk[k] == 0xffffffff)
3652 			i += 32;
3653 		else {
3654 			for (j = msk[k]; j; j <<= 1)
3655 				if (j & 0x80000000)
3656 					i++;
3657 		}
3658 	return i;
3659 }
3660 # endif
3661 #endif /* _KERNEL */
3662 
3663 
3664 /* ------------------------------------------------------------------------ */
3665 /* Function:    fr_ifsync                                                   */
3666 /* Returns:     void *    - new interface identifier                        */
3667 /* Parameters:  action(I)  - type of synchronisation to do                  */
3668 /*              v(I)       - IP version being sync'd (v4 or v6)             */
3669 /*              newifp(I)  - interface identifier being introduced/removed  */
3670 /*              oldifp(I)  - interface identifier in a filter rule          */
3671 /*              newname(I) - name associated with newifp interface          */
3672 /*              oldname(I) - name associated with oldifp interface          */
3673 /*		ifs       - pointer to IPF stack instance		    */
3674 /*                                                                          */
3675 /* This function returns what the new value for "oldifp" should be for its  */
3676 /* caller.  In some cases it will not change, in some it will.              */
3677 /* action == IPFSYNC_RESYNC                                                 */
3678 /*   a new value for oldifp will always be looked up, according to oldname, */
3679 /*   the values of newname and newifp are ignored.                          */
3680 /* action == IPFSYNC_NEWIFP                                                 */
3681 /*   if oldname matches newname then we are doing a sync for the matching   */
3682 /*   interface, so we return newifp to be used in place of oldifp.  If the  */
3683 /*   the names don't match, just return oldifp.                             */
3684 /* action == IPFSYNC_OLDIFP                                                 */
3685 /*   if oldifp matches newifp then we are are doing a sync to remove any    */
3686 /*   references to oldifp, so we return "-1".                               */
3687 /* -----								    */
3688 /* NOTE:								    */
3689 /* This function processes NIC event from PF_HOOKS. The action parameter    */
3690 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is	    */
3691 /* one single switch statement() in ipf_nic_event_vx() function, which	    */
3692 /* translates the HOOK event type to action parameter passed to fr_ifsync.  */
3693 /* The translation table looks as follows:				    */
3694 /*	event		| action					    */
3695 /*	----------------+-------------					    */
3696 /*	NE_PLUMB	| IPFSYNC_NEWIFP				    */
3697 /*	NE_UNPLUMB	| IPFSYNC_OLDIFP				    */
3698 /*    NE_ADDRESS_CHANGE	| IPFSYNC_RESYNC				    */
3699 /*									    */
3700 /* The oldname and oldifp parameters are taken from IPF entry (rule, state  */
3701 /* table entry, NAT table entry, fragment ...). The newname and newifp	    */
3702 /* parameters come from hook event data, parameters are taken from event    */
3703 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is	    */
3704 /* notified by hook function.						    */
3705 /*									    */
3706 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3707 /* to plumb the interface, which is already plumbed. In such case we always */
3708 /* get the event from PF_HOOKS as follows:				    */
3709 /*	event:	NE_PLUMB						    */
3710 /*	NIC:	0x0							    */
3711 /* ------------------------------------------------------------------------ */
3712 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3713 int action, v;
3714 char *newname, *oldname;
3715 void *newifp, *oldifp;
3716 ipf_stack_t *ifs;
3717 {
3718 	void *rval = oldifp;
3719 
3720 	switch (action)
3721 	{
3722 	case IPFSYNC_RESYNC :
3723 		if (oldname[0] != '\0') {
3724 			rval = fr_resolvenic(oldname, v, ifs);
3725 		}
3726 		break;
3727 	case IPFSYNC_NEWIFP :
3728 		if (!strncmp(newname, oldname, LIFNAMSIZ))
3729 			rval = newifp;
3730 		break;
3731 	case IPFSYNC_OLDIFP :
3732 		/*
3733 		 * If interface gets unplumbed it must be invalidated, which
3734 		 * means set all existing references to the interface to -1.
3735 		 * We don't want to invalidate references for wildcard
3736 		 * (unbound) rules (entries).
3737 		 */
3738 		if (newifp == oldifp)
3739 			rval = (oldifp) ? (void *)-1 : NULL;
3740 		break;
3741 	}
3742 
3743 	return rval;
3744 }
3745 
3746 
3747 /* ------------------------------------------------------------------------ */
3748 /* Function:    frsynclist                                                  */
3749 /* Returns:     void                                                        */
3750 /* Parameters:  action(I) - type of synchronisation to do                   */
3751 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3752 /*              ifp(I)    - interface identifier associated with action     */
3753 /*              ifname(I) - name associated with ifp parameter              */
3754 /*              fr(I)     - pointer to filter rule                          */
3755 /*		ifs       - pointer to IPF stack instance		    */
3756 /* Write Locks: ipf_mutex                                                   */
3757 /*                                                                          */
3758 /* Walk through a list of filter rules and resolve any interface names into */
3759 /* pointers.  Where dynamic addresses are used, also update the IP address  */
3760 /* used in the rule.  The interface pointer is used to limit the lookups to */
3761 /* a specific set of matching names if it is non-NULL.                      */
3762 /* ------------------------------------------------------------------------ */
3763 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3764 int action, v;
3765 void *ifp;
3766 char *ifname;
3767 frentry_t *fr;
3768 ipf_stack_t *ifs;
3769 {
3770 	frdest_t *fdp;
3771 	int rv, i;
3772 
3773 	for (; fr; fr = fr->fr_next) {
3774 		rv = fr->fr_v;
3775 		if (v != 0 && v != rv)
3776 			continue;
3777 
3778 		/*
3779 		 * Lookup all the interface names that are part of the rule.
3780 		 */
3781 		for (i = 0; i < 4; i++) {
3782 			fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3783 						   fr->fr_ifnames[i],
3784 						   ifp, fr->fr_ifas[i],
3785 						   ifs);
3786 		}
3787 
3788 		fdp = &fr->fr_tifs[0];
3789 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3790 					   ifp, fdp->fd_ifp, ifs);
3791 
3792 		fdp = &fr->fr_tifs[1];
3793 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3794 					   ifp, fdp->fd_ifp, ifs);
3795 
3796 		fdp = &fr->fr_dif;
3797 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3798 					   ifp, fdp->fd_ifp, ifs);
3799 
3800 		if (action != IPFSYNC_RESYNC)
3801 			continue;
3802 
3803 		if (fr->fr_type == FR_T_IPF) {
3804 			if (fr->fr_satype != FRI_NORMAL &&
3805 			    fr->fr_satype != FRI_LOOKUP) {
3806 				(void)fr_ifpaddr(rv, fr->fr_satype,
3807 						 fr->fr_ifas[fr->fr_sifpidx],
3808 						 &fr->fr_src, &fr->fr_smsk,
3809 						 ifs);
3810 			}
3811 			if (fr->fr_datype != FRI_NORMAL &&
3812 			    fr->fr_datype != FRI_LOOKUP) {
3813 				(void)fr_ifpaddr(rv, fr->fr_datype,
3814 						 fr->fr_ifas[fr->fr_difpidx],
3815 						 &fr->fr_dst, &fr->fr_dmsk,
3816 						 ifs);
3817 			}
3818 		}
3819 
3820 #ifdef	IPFILTER_LOOKUP
3821 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3822 		    fr->fr_srcptr == NULL) {
3823 			fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3824 							 fr->fr_srcnum,
3825 							 &fr->fr_srcfunc, ifs);
3826 		}
3827 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3828 		    fr->fr_dstptr == NULL) {
3829 			fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3830 							 fr->fr_dstnum,
3831 							 &fr->fr_dstfunc, ifs);
3832 		}
3833 #endif
3834 	}
3835 }
3836 
3837 
3838 #ifdef	_KERNEL
3839 /* ------------------------------------------------------------------------ */
3840 /* Function:    frsync                                                      */
3841 /* Returns:     void                                                        */
3842 /* Parameters:  action(I) - type of synchronisation to do                   */
3843 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3844 /*              ifp(I)    - interface identifier associated with action     */
3845 /*              name(I)   - name associated with ifp parameter              */
3846 /*                                                                          */
3847 /* frsync() is called when we suspect that the interface list or            */
3848 /* information about interfaces (like IP#) has changed.  Go through all     */
3849 /* filter rules, NAT entries and the state table and check if anything      */
3850 /* needs to be changed/updated.                                             */
3851 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3852 /* in which this was done to support three different types of sync:         */
3853 /* - complete resync of all interface name/identifiers                      */
3854 /* - new interface being announced with its name and identifier             */
3855 /* - interface removal being announced by only its identifier               */
3856 /* ------------------------------------------------------------------------ */
3857 void frsync(action, v, ifp, name, ifs)
3858 int action, v;
3859 void *ifp;
3860 char *name;
3861 ipf_stack_t *ifs;
3862 {
3863 	int i;
3864 
3865 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3866 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3867 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3868 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3869 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3870 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3871 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3872 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3873 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3874 
3875 	for (i = 0; i < IPL_LOGSIZE; i++) {
3876 		frgroup_t *g;
3877 
3878 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3879 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3880 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3881 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3882 	}
3883 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3884 }
3885 
3886 
3887 /*
3888  * In the functions below, bcopy() is called because the pointer being
3889  * copied _from_ in this instance is a pointer to a char buf (which could
3890  * end up being unaligned) and on the kernel's local stack.
3891  */
3892 /* ------------------------------------------------------------------------ */
3893 /* Function:    copyinptr                                                   */
3894 /* Returns:     int - 0 = success, else failure                             */
3895 /* Parameters:  src(I)  - pointer to the source address                     */
3896 /*              dst(I)  - destination address                               */
3897 /*              size(I) - number of bytes to copy                           */
3898 /*                                                                          */
3899 /* Copy a block of data in from user space, given a pointer to the pointer  */
3900 /* to start copying from (src) and a pointer to where to store it (dst).    */
3901 /* NB: src - pointer to user space pointer, dst - kernel space pointer      */
3902 /* ------------------------------------------------------------------------ */
3903 int copyinptr(src, dst, size)
3904 void *src, *dst;
3905 size_t size;
3906 {
3907 	caddr_t ca;
3908 	int err;
3909 
3910 # if SOLARIS
3911 	err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
3912 	if (err != 0)
3913 		return err;
3914 # else
3915 	bcopy(src, (caddr_t)&ca, sizeof(ca));
3916 # endif
3917 	err = COPYIN(ca, dst, size);
3918 	return err;
3919 }
3920 
3921 
3922 /* ------------------------------------------------------------------------ */
3923 /* Function:    copyoutptr                                                  */
3924 /* Returns:     int - 0 = success, else failure                             */
3925 /* Parameters:  src(I)  - pointer to the source address                     */
3926 /*              dst(I)  - destination address                               */
3927 /*              size(I) - number of bytes to copy                           */
3928 /*                                                                          */
3929 /* Copy a block of data out to user space, given a pointer to the pointer   */
3930 /* to start copying from (src) and a pointer to where to store it (dst).    */
3931 /* NB: src - kernel space pointer, dst - pointer to user space pointer.     */
3932 /* ------------------------------------------------------------------------ */
3933 int copyoutptr(src, dst, size)
3934 void *src, *dst;
3935 size_t size;
3936 {
3937 	caddr_t ca;
3938 	int err;
3939 
3940 # if SOLARIS
3941 	err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
3942 	if (err != 0)
3943 		return err;
3944 # else
3945 	bcopy(dst, (caddr_t)&ca, sizeof(ca));
3946 # endif
3947 	err = COPYOUT(src, ca, size);
3948 	return err;
3949 }
3950 #endif
3951 
3952 
3953 /* ------------------------------------------------------------------------ */
3954 /* Function:    fr_lock                                                     */
3955 /* Returns:	int - 0 = success, else error				    */
3956 /* Parameters:  data(I)  - pointer to lock value to set                     */
3957 /*              lockp(O) - pointer to location to store old lock value      */
3958 /*                                                                          */
3959 /* Get the new value for the lock integer, set it and return the old value  */
3960 /* in *lockp.                                                               */
3961 /* ------------------------------------------------------------------------ */
3962 int fr_lock(data, lockp)
3963 caddr_t data;
3964 int *lockp;
3965 {
3966 	int arg, err;
3967 
3968 	err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
3969 	if (err != 0)
3970 		return (EFAULT);
3971 	err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
3972 	if (err != 0)
3973 		return (EFAULT);
3974 	*lockp = arg;
3975 	return (0);
3976 }
3977 
3978 
3979 /* ------------------------------------------------------------------------ */
3980 /* Function:    fr_getstat                                                  */
3981 /* Returns:     Nil                                                         */
3982 /* Parameters:  fiop(I)  - pointer to ipfilter stats structure              */
3983 /*                                                                          */
3984 /* Stores a copy of current pointers, counters, etc, in the friostat        */
3985 /* structure.                                                               */
3986 /* ------------------------------------------------------------------------ */
3987 void fr_getstat(fiop, ifs)
3988 friostat_t *fiop;
3989 ipf_stack_t *ifs;
3990 {
3991 	int i, j;
3992 
3993 	bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
3994 	    sizeof(filterstats_t) * 2);
3995 	fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
3996 	fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
3997 	fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
3998 	fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
3999 
4000 	for (i = 0; i < 2; i++)
4001 		for (j = 0; j < 2; j++) {
4002 			fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4003 			fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4004 			fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4005 			fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4006 		}
4007 
4008 	fiop->f_ticks = ifs->ifs_fr_ticks;
4009 	fiop->f_active = ifs->ifs_fr_active;
4010 	fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4011 	fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4012 
4013 	fiop->f_running = ifs->ifs_fr_running;
4014 	for (i = 0; i < IPL_LOGSIZE; i++) {
4015 		fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4016 		fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4017 	}
4018 #ifdef  IPFILTER_LOG
4019 	fiop->f_logging = 1;
4020 #else
4021 	fiop->f_logging = 0;
4022 #endif
4023 	fiop->f_defpass = ifs->ifs_fr_pass;
4024 	fiop->f_features = fr_features;
4025 	(void) strncpy(fiop->f_version, ipfilter_version,
4026 		       sizeof(fiop->f_version));
4027 }
4028 
4029 
4030 #ifdef	USE_INET6
4031 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4032 	ICMP6_ECHO_REPLY,	/* 0: ICMP_ECHOREPLY */
4033 	-1,			/* 1: UNUSED */
4034 	-1,			/* 2: UNUSED */
4035 	ICMP6_DST_UNREACH,	/* 3: ICMP_UNREACH */
4036 	-1,			/* 4: ICMP_SOURCEQUENCH */
4037 	ND_REDIRECT,		/* 5: ICMP_REDIRECT */
4038 	-1,			/* 6: UNUSED */
4039 	-1,			/* 7: UNUSED */
4040 	ICMP6_ECHO_REQUEST,	/* 8: ICMP_ECHO */
4041 	-1,			/* 9: UNUSED */
4042 	-1,			/* 10: UNUSED */
4043 	ICMP6_TIME_EXCEEDED,	/* 11: ICMP_TIMXCEED */
4044 	ICMP6_PARAM_PROB,	/* 12: ICMP_PARAMPROB */
4045 	-1,			/* 13: ICMP_TSTAMP */
4046 	-1,			/* 14: ICMP_TSTAMPREPLY */
4047 	-1,			/* 15: ICMP_IREQ */
4048 	-1,			/* 16: ICMP_IREQREPLY */
4049 	-1,			/* 17: ICMP_MASKREQ */
4050 	-1,			/* 18: ICMP_MASKREPLY */
4051 };
4052 
4053 
4054 int	icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4055 	ICMP6_DST_UNREACH_ADDR,		/* 0: ICMP_UNREACH_NET */
4056 	ICMP6_DST_UNREACH_ADDR,		/* 1: ICMP_UNREACH_HOST */
4057 	-1,				/* 2: ICMP_UNREACH_PROTOCOL */
4058 	ICMP6_DST_UNREACH_NOPORT,	/* 3: ICMP_UNREACH_PORT */
4059 	-1,				/* 4: ICMP_UNREACH_NEEDFRAG */
4060 	ICMP6_DST_UNREACH_NOTNEIGHBOR,	/* 5: ICMP_UNREACH_SRCFAIL */
4061 	ICMP6_DST_UNREACH_ADDR,		/* 6: ICMP_UNREACH_NET_UNKNOWN */
4062 	ICMP6_DST_UNREACH_ADDR,		/* 7: ICMP_UNREACH_HOST_UNKNOWN */
4063 	-1,				/* 8: ICMP_UNREACH_ISOLATED */
4064 	ICMP6_DST_UNREACH_ADMIN,	/* 9: ICMP_UNREACH_NET_PROHIB */
4065 	ICMP6_DST_UNREACH_ADMIN,	/* 10: ICMP_UNREACH_HOST_PROHIB */
4066 	-1,				/* 11: ICMP_UNREACH_TOSNET */
4067 	-1,				/* 12: ICMP_UNREACH_TOSHOST */
4068 	ICMP6_DST_UNREACH_ADMIN,	/* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4069 };
4070 int	icmpreplytype6[ICMP6_MAXTYPE + 1];
4071 #endif
4072 
4073 int	icmpreplytype4[ICMP_MAXTYPE + 1];
4074 
4075 
4076 /* ------------------------------------------------------------------------ */
4077 /* Function:    fr_matchicmpqueryreply                                      */
4078 /* Returns:     int - 1 if "icmp" is a valid reply to "ic" else 0.          */
4079 /* Parameters:  v(I)    - IP protocol version (4 or 6)                      */
4080 /*              ic(I)   - ICMP information                                  */
4081 /*              icmp(I) - ICMP packet header                                */
4082 /*              rev(I)  - direction (0 = forward/1 = reverse) of packet     */
4083 /*                                                                          */
4084 /* Check if the ICMP packet defined by the header pointed to by icmp is a   */
4085 /* reply to one as described by what's in ic.  If it is a match, return 1,  */
4086 /* else return 0 for no match.                                              */
4087 /* ------------------------------------------------------------------------ */
4088 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4089 int v;
4090 icmpinfo_t *ic;
4091 icmphdr_t *icmp;
4092 int rev;
4093 {
4094 	int ictype;
4095 
4096 	ictype = ic->ici_type;
4097 
4098 	if (v == 4) {
4099 		/*
4100 		 * If we matched its type on the way in, then when going out
4101 		 * it will still be the same type.
4102 		 */
4103 		if ((!rev && (icmp->icmp_type == ictype)) ||
4104 		    (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4105 			if (icmp->icmp_type != ICMP_ECHOREPLY)
4106 				return 1;
4107 			if (icmp->icmp_id == ic->ici_id)
4108 				return 1;
4109 		}
4110 	}
4111 #ifdef	USE_INET6
4112 	else if (v == 6) {
4113 		if ((!rev && (icmp->icmp_type == ictype)) ||
4114 		    (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4115 			if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4116 				return 1;
4117 			if (icmp->icmp_id == ic->ici_id)
4118 				return 1;
4119 		}
4120 	}
4121 #endif
4122 	return 0;
4123 }
4124 
4125 
4126 #ifdef	IPFILTER_LOOKUP
4127 /* ------------------------------------------------------------------------ */
4128 /* Function:    fr_resolvelookup                                            */
4129 /* Returns:     void * - NULL = failure, else success.                      */
4130 /* Parameters:  type(I)     - type of lookup these parameters are for.      */
4131 /*              number(I)   - table number to use when searching            */
4132 /*              funcptr(IO) - pointer to pointer for storing IP address     */
4133 /*                           searching function.                            */
4134 /*                                                                          */
4135 /* Search for the "table" number passed in amongst those configured for     */
4136 /* that particular type.  If the type is recognised then the function to    */
4137 /* call to do the IP address search will be change, regardless of whether   */
4138 /* or not the "table" number exists.                                        */
4139 /* ------------------------------------------------------------------------ */
4140 static void *fr_resolvelookup(type, number, funcptr, ifs)
4141 u_int type, number;
4142 lookupfunc_t *funcptr;
4143 ipf_stack_t *ifs;
4144 {
4145 	char name[FR_GROUPLEN];
4146 	iphtable_t *iph;
4147 	ip_pool_t *ipo;
4148 	void *ptr;
4149 
4150 #if defined(SNPRINTF) && defined(_KERNEL)
4151 	(void) SNPRINTF(name, sizeof(name), "%u", number);
4152 #else
4153 	(void) sprintf(name, "%u", number);
4154 #endif
4155 
4156 	READ_ENTER(&ifs->ifs_ip_poolrw);
4157 
4158 	switch (type)
4159 	{
4160 	case IPLT_POOL :
4161 # if (defined(__osf__) && defined(_KERNEL))
4162 		ptr = NULL;
4163 		*funcptr = NULL;
4164 # else
4165 		ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4166 		ptr = ipo;
4167 		if (ipo != NULL) {
4168 			ATOMIC_INC32(ipo->ipo_ref);
4169 		}
4170 		*funcptr = ip_pool_search;
4171 # endif
4172 		break;
4173 	case IPLT_HASH :
4174 		iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4175 		ptr = iph;
4176 		if (iph != NULL) {
4177 			ATOMIC_INC32(iph->iph_ref);
4178 		}
4179 		*funcptr = fr_iphmfindip;
4180 		break;
4181 	default:
4182 		ptr = NULL;
4183 		*funcptr = NULL;
4184 		break;
4185 	}
4186 	RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4187 
4188 	return ptr;
4189 }
4190 #endif
4191 
4192 
4193 /* ------------------------------------------------------------------------ */
4194 /* Function:    frrequest                                                   */
4195 /* Returns:     int - 0 == success, > 0 == errno value                      */
4196 /* Parameters:  unit(I)     - device for which this is for                  */
4197 /*              req(I)      - ioctl command (SIOC*)                         */
4198 /*              data(I)     - pointr to ioctl data                          */
4199 /*              set(I)      - 1 or 0 (filter set)                           */
4200 /*              makecopy(I) - flag indicating whether data points to a rule */
4201 /*                            in kernel space & hence doesn't need copying. */
4202 /*                                                                          */
4203 /* This function handles all the requests which operate on the list of      */
4204 /* filter rules.  This includes adding, deleting, insertion.  It is also    */
4205 /* responsible for creating groups when a "head" rule is loaded.  Interface */
4206 /* names are resolved here and other sanity checks are made on the content  */
4207 /* of the rule structure being loaded.  If a rule has user defined timeouts */
4208 /* then make sure they are created and initialised before exiting.          */
4209 /* ------------------------------------------------------------------------ */
4210 int frrequest(unit, req, data, set, makecopy, ifs)
4211 int unit;
4212 ioctlcmd_t req;
4213 int set, makecopy;
4214 caddr_t data;
4215 ipf_stack_t *ifs;
4216 {
4217 	frentry_t frd, *fp, *f, **fprev, **ftail;
4218 	int error = 0, in, v;
4219 	void *ptr, *uptr;
4220 	u_int *p, *pp;
4221 	frgroup_t *fg;
4222 	char *group;
4223 
4224 	fg = NULL;
4225 	fp = &frd;
4226 	if (makecopy != 0) {
4227 		error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4228 		if (error)
4229 			return EFAULT;
4230 		if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4231 			return EINVAL;
4232 		fp->fr_ref = 0;
4233 		fp->fr_flags |= FR_COPIED;
4234 	} else {
4235 		fp = (frentry_t *)data;
4236 		if ((fp->fr_type & FR_T_BUILTIN) == 0)
4237 			return EINVAL;
4238 		fp->fr_flags &= ~FR_COPIED;
4239 	}
4240 
4241 	if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4242 	    ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4243 		return EINVAL;
4244 
4245 	v = fp->fr_v;
4246 	uptr = fp->fr_data;
4247 
4248 	/*
4249 	 * Only filter rules for IPv4 or IPv6 are accepted.
4250 	 */
4251 	if (v == 4)
4252 		/*EMPTY*/;
4253 #ifdef	USE_INET6
4254 	else if (v == 6)
4255 		/*EMPTY*/;
4256 #endif
4257 	else {
4258 		return EINVAL;
4259 	}
4260 
4261 	/*
4262 	 * If the rule is being loaded from user space, i.e. we had to copy it
4263 	 * into kernel space, then do not trust the function pointer in the
4264 	 * rule.
4265 	 */
4266 	if ((makecopy == 1) && (fp->fr_func != NULL)) {
4267 		if (fr_findfunc(fp->fr_func) == NULL)
4268 			return ESRCH;
4269 		error = fr_funcinit(fp, ifs);
4270 		if (error != 0)
4271 			return error;
4272 	}
4273 
4274 	ptr = NULL;
4275 	/*
4276 	 * Check that the group number does exist and that its use (in/out)
4277 	 * matches what the rule is.
4278 	 */
4279 	if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4280 		*fp->fr_grhead = '\0';
4281 	group = fp->fr_group;
4282 	if (!strncmp(group, "0", FR_GROUPLEN))
4283 		*group = '\0';
4284 
4285 	if (FR_ISACCOUNT(fp->fr_flags))
4286 		unit = IPL_LOGCOUNT;
4287 
4288 	if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4289 		fg = fr_findgroup(group, unit, set, NULL, ifs);
4290 		if (fg == NULL)
4291 			return ESRCH;
4292 		if (fg->fg_flags == 0)
4293 			fg->fg_flags = fp->fr_flags & FR_INOUT;
4294 		else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4295 			return ESRCH;
4296 	}
4297 
4298 	in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4299 
4300 	/*
4301 	 * Work out which rule list this change is being applied to.
4302 	 */
4303 	ftail = NULL;
4304 	fprev = NULL;
4305 	if (unit == IPL_LOGAUTH)
4306 		fprev = &ifs->ifs_ipauth;
4307 	else if (v == 4) {
4308 		if (FR_ISACCOUNT(fp->fr_flags))
4309 			fprev = &ifs->ifs_ipacct[in][set];
4310 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4311 			fprev = &ifs->ifs_ipfilter[in][set];
4312 	} else if (v == 6) {
4313 		if (FR_ISACCOUNT(fp->fr_flags))
4314 			fprev = &ifs->ifs_ipacct6[in][set];
4315 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4316 			fprev = &ifs->ifs_ipfilter6[in][set];
4317 	}
4318 	if (fprev == NULL)
4319 		return ESRCH;
4320 
4321 	if (*group != '\0') {
4322 	    if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4323 			return ESRCH;
4324 		fprev = &fg->fg_start;
4325 	}
4326 
4327 	ftail = fprev;
4328 	for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4329 		if (fp->fr_collect <= f->fr_collect) {
4330 			ftail = fprev;
4331 			f = NULL;
4332 			break;
4333 		}
4334 		fprev = ftail;
4335 	}
4336 
4337 	/*
4338 	 * Copy in extra data for the rule.
4339 	 */
4340 	if (fp->fr_dsize != 0) {
4341 		if (makecopy != 0) {
4342 			KMALLOCS(ptr, void *, fp->fr_dsize);
4343 			if (!ptr)
4344 				return ENOMEM;
4345 			error = COPYIN(uptr, ptr, fp->fr_dsize);
4346 		} else {
4347 			ptr = uptr;
4348 			error = 0;
4349 		}
4350 		if (error != 0) {
4351 			KFREES(ptr, fp->fr_dsize);
4352 			return ENOMEM;
4353 		}
4354 		fp->fr_data = ptr;
4355 	} else
4356 		fp->fr_data = NULL;
4357 
4358 	/*
4359 	 * Perform per-rule type sanity checks of their members.
4360 	 */
4361 	switch (fp->fr_type & ~FR_T_BUILTIN)
4362 	{
4363 #if defined(IPFILTER_BPF)
4364 	case FR_T_BPFOPC :
4365 		if (fp->fr_dsize == 0)
4366 			return EINVAL;
4367 		if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4368 			if (makecopy && fp->fr_data != NULL) {
4369 				KFREES(fp->fr_data, fp->fr_dsize);
4370 			}
4371 			return EINVAL;
4372 		}
4373 		break;
4374 #endif
4375 	case FR_T_IPF :
4376 		if (fp->fr_dsize != sizeof(fripf_t))
4377 			return EINVAL;
4378 
4379 		/*
4380 		 * Allowing a rule with both "keep state" and "with oow" is
4381 		 * pointless because adding a state entry to the table will
4382 		 * fail with the out of window (oow) flag set.
4383 		 */
4384 		if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW))
4385 			return EINVAL;
4386 
4387 		switch (fp->fr_satype)
4388 		{
4389 		case FRI_BROADCAST :
4390 		case FRI_DYNAMIC :
4391 		case FRI_NETWORK :
4392 		case FRI_NETMASKED :
4393 		case FRI_PEERADDR :
4394 			if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4395 				if (makecopy && fp->fr_data != NULL) {
4396 					KFREES(fp->fr_data, fp->fr_dsize);
4397 				}
4398 				return EINVAL;
4399 			}
4400 			break;
4401 #ifdef	IPFILTER_LOOKUP
4402 		case FRI_LOOKUP :
4403 			fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4404 							 fp->fr_srcnum,
4405 							 &fp->fr_srcfunc, ifs);
4406 			break;
4407 #endif
4408 		default :
4409 			break;
4410 		}
4411 
4412 		switch (fp->fr_datype)
4413 		{
4414 		case FRI_BROADCAST :
4415 		case FRI_DYNAMIC :
4416 		case FRI_NETWORK :
4417 		case FRI_NETMASKED :
4418 		case FRI_PEERADDR :
4419 			if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4420 				if (makecopy && fp->fr_data != NULL) {
4421 					KFREES(fp->fr_data, fp->fr_dsize);
4422 				}
4423 				return EINVAL;
4424 			}
4425 			break;
4426 #ifdef	IPFILTER_LOOKUP
4427 		case FRI_LOOKUP :
4428 			fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4429 							 fp->fr_dstnum,
4430 							 &fp->fr_dstfunc, ifs);
4431 			break;
4432 #endif
4433 		default :
4434 			break;
4435 		}
4436 		break;
4437 	case FR_T_NONE :
4438 		break;
4439 	case FR_T_CALLFUNC :
4440 		break;
4441 	case FR_T_COMPIPF :
4442 		break;
4443 	default :
4444 		if (makecopy && fp->fr_data != NULL) {
4445 			KFREES(fp->fr_data, fp->fr_dsize);
4446 		}
4447 		return EINVAL;
4448 	}
4449 
4450 	/*
4451 	 * Lookup all the interface names that are part of the rule.
4452 	 */
4453 	frsynclist(0, 0, NULL, NULL, fp, ifs);
4454 	fp->fr_statecnt = 0;
4455 
4456 	/*
4457 	 * Look for an existing matching filter rule, but don't include the
4458 	 * next or interface pointer in the comparison (fr_next, fr_ifa).
4459 	 * This elminates rules which are indentical being loaded.  Checksum
4460 	 * the constant part of the filter rule to make comparisons quicker
4461 	 * (this meaning no pointers are included).
4462 	 */
4463 	for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4464 	     p < pp; p++)
4465 		fp->fr_cksum += *p;
4466 	pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4467 	for (p = (u_int *)fp->fr_data; p < pp; p++)
4468 		fp->fr_cksum += *p;
4469 
4470 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
4471 
4472 	for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4473 		if ((fp->fr_cksum != f->fr_cksum) ||
4474 		    (f->fr_dsize != fp->fr_dsize))
4475 			continue;
4476 		if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4477 			continue;
4478 		if ((!ptr && !f->fr_data) ||
4479 		    (ptr && f->fr_data &&
4480 		     !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4481 			break;
4482 	}
4483 
4484 	/*
4485 	 * If zero'ing statistics, copy current to caller and zero.
4486 	 */
4487 	if (req == (ioctlcmd_t)SIOCZRLST) {
4488 		if (f == NULL)
4489 			error = ESRCH;
4490 		else {
4491 			/*
4492 			 * Copy and reduce lock because of impending copyout.
4493 			 * Well we should, but if we do then the atomicity of
4494 			 * this call and the correctness of fr_hits and
4495 			 * fr_bytes cannot be guaranteed.  As it is, this code
4496 			 * only resets them to 0 if they are successfully
4497 			 * copied out into user space.
4498 			 */
4499 			bcopy((char *)f, (char *)fp, sizeof(*f));
4500 
4501 			/*
4502 			 * When we copy this rule back out, set the data
4503 			 * pointer to be what it was in user space.
4504 			 */
4505 			fp->fr_data = uptr;
4506 			error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4507 
4508 			if (error == 0) {
4509 				if ((f->fr_dsize != 0) && (uptr != NULL))
4510 					error = COPYOUT(f->fr_data, uptr,
4511 							f->fr_dsize);
4512 				if (error == 0) {
4513 					f->fr_hits = 0;
4514 					f->fr_bytes = 0;
4515 				}
4516 			}
4517 		}
4518 
4519 		if ((ptr != NULL) && (makecopy != 0)) {
4520 			KFREES(ptr, fp->fr_dsize);
4521 		}
4522 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4523 		return error;
4524 	}
4525 
4526 	if (!f) {
4527 		/*
4528 		 * At the end of this, ftail must point to the place where the
4529 		 * new rule is to be saved/inserted/added.
4530 		 * For SIOCAD*FR, this should be the last rule in the group of
4531 		 * rules that have equal fr_collect fields.
4532 		 * For SIOCIN*FR, ...
4533 		 */
4534 		if (req == (ioctlcmd_t)SIOCADAFR ||
4535 		    req == (ioctlcmd_t)SIOCADIFR) {
4536 
4537 			for (ftail = fprev; (f = *ftail) != NULL; ) {
4538 				if (f->fr_collect > fp->fr_collect)
4539 					break;
4540 				ftail = &f->fr_next;
4541 			}
4542 			f = NULL;
4543 			ptr = NULL;
4544 			error = 0;
4545 		} else if (req == (ioctlcmd_t)SIOCINAFR ||
4546 			   req == (ioctlcmd_t)SIOCINIFR) {
4547 			while ((f = *fprev) != NULL) {
4548 				if (f->fr_collect >= fp->fr_collect)
4549 					break;
4550 				fprev = &f->fr_next;
4551 			}
4552 			ftail = fprev;
4553 			if (fp->fr_hits != 0) {
4554 				while (fp->fr_hits && (f = *ftail)) {
4555 					if (f->fr_collect != fp->fr_collect)
4556 						break;
4557 					fprev = ftail;
4558 					ftail = &f->fr_next;
4559 					fp->fr_hits--;
4560 				}
4561 			}
4562 			f = NULL;
4563 			ptr = NULL;
4564 			error = 0;
4565 		}
4566 	}
4567 
4568 	/*
4569 	 * Request to remove a rule.
4570 	 */
4571 	if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4572 		if (!f)
4573 			error = ESRCH;
4574 		else {
4575 			/*
4576 			 * Do not allow activity from user space to interfere
4577 			 * with rules not loaded that way.
4578 			 */
4579 			if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4580 				error = EPERM;
4581 				goto done;
4582 			}
4583 
4584 			/*
4585 			 * Return EBUSY if the rule is being reference by
4586 			 * something else (eg state information.
4587 			 */
4588 			if (f->fr_ref > 1) {
4589 				error = EBUSY;
4590 				goto done;
4591 			}
4592 #ifdef	IPFILTER_SCAN
4593 			if (f->fr_isctag[0] != '\0' &&
4594 			    (f->fr_isc != (struct ipscan *)-1))
4595 				ipsc_detachfr(f);
4596 #endif
4597 			if (unit == IPL_LOGAUTH) {
4598 				error = fr_preauthcmd(req, f, ftail, ifs);
4599 				goto done;
4600 			}
4601 			if (*f->fr_grhead != '\0')
4602 				fr_delgroup(f->fr_grhead, unit, set, ifs);
4603 			fr_fixskip(ftail, f, -1);
4604 			*ftail = f->fr_next;
4605 			f->fr_next = NULL;
4606 			(void)fr_derefrule(&f, ifs);
4607 		}
4608 	} else {
4609 		/*
4610 		 * Not removing, so we must be adding/inserting a rule.
4611 		 */
4612 		if (f)
4613 			error = EEXIST;
4614 		else {
4615 			if (unit == IPL_LOGAUTH) {
4616 				error = fr_preauthcmd(req, fp, ftail, ifs);
4617 				goto done;
4618 			}
4619 			if (makecopy) {
4620 				KMALLOC(f, frentry_t *);
4621 			} else
4622 				f = fp;
4623 			if (f != NULL) {
4624 				if (fp != f)
4625 					bcopy((char *)fp, (char *)f,
4626 					      sizeof(*f));
4627 				MUTEX_NUKE(&f->fr_lock);
4628 				MUTEX_INIT(&f->fr_lock, "filter rule lock");
4629 #ifdef	IPFILTER_SCAN
4630 				if (f->fr_isctag[0] != '\0' &&
4631 				    ipsc_attachfr(f))
4632 					f->fr_isc = (struct ipscan *)-1;
4633 #endif
4634 				f->fr_hits = 0;
4635 				if (makecopy != 0)
4636 					f->fr_ref = 1;
4637 				f->fr_next = *ftail;
4638 				*ftail = f;
4639 				if (req == (ioctlcmd_t)SIOCINIFR ||
4640 				    req == (ioctlcmd_t)SIOCINAFR)
4641 					fr_fixskip(ftail, f, 1);
4642 				f->fr_grp = NULL;
4643 				group = f->fr_grhead;
4644 				if (*group != '\0') {
4645 					fg = fr_addgroup(group, f, f->fr_flags,
4646 							 unit, set, ifs);
4647 					if (fg != NULL)
4648 						f->fr_grp = &fg->fg_start;
4649 				}
4650 			} else
4651 				error = ENOMEM;
4652 		}
4653 	}
4654 done:
4655 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4656 	if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4657 		KFREES(ptr, fp->fr_dsize);
4658 	}
4659 	return (error);
4660 }
4661 
4662 
4663 /* ------------------------------------------------------------------------ */
4664 /* Function:    fr_funcinit                                                 */
4665 /* Returns:     int - 0 == success, else ESRCH: cannot resolve rule details */
4666 /* Parameters:  fr(I) - pointer to filter rule                              */
4667 /*                                                                          */
4668 /* If a rule is a call rule, then check if the function it points to needs  */
4669 /* an init function to be called now the rule has been loaded.              */
4670 /* ------------------------------------------------------------------------ */
4671 static int fr_funcinit(fr, ifs)
4672 frentry_t *fr;
4673 ipf_stack_t *ifs;
4674 {
4675 	ipfunc_resolve_t *ft;
4676 	int err;
4677 
4678 	err = ESRCH;
4679 
4680 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4681 		if (ft->ipfu_addr == fr->fr_func) {
4682 			err = 0;
4683 			if (ft->ipfu_init != NULL)
4684 				err = (*ft->ipfu_init)(fr, ifs);
4685 			break;
4686 		}
4687 	return err;
4688 }
4689 
4690 
4691 /* ------------------------------------------------------------------------ */
4692 /* Function:    fr_findfunc                                                 */
4693 /* Returns:     ipfunc_t - pointer to function if found, else NULL          */
4694 /* Parameters:  funcptr(I) - function pointer to lookup                     */
4695 /*                                                                          */
4696 /* Look for a function in the table of known functions.                     */
4697 /* ------------------------------------------------------------------------ */
4698 static ipfunc_t fr_findfunc(funcptr)
4699 ipfunc_t funcptr;
4700 {
4701 	ipfunc_resolve_t *ft;
4702 
4703 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4704 		if (ft->ipfu_addr == funcptr)
4705 			return funcptr;
4706 	return NULL;
4707 }
4708 
4709 
4710 /* ------------------------------------------------------------------------ */
4711 /* Function:    fr_resolvefunc                                              */
4712 /* Returns:     int - 0 == success, else error                              */
4713 /* Parameters:  data(IO) - ioctl data pointer to ipfunc_resolve_t struct    */
4714 /*                                                                          */
4715 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4716 /* This will either be the function name (if the pointer is set) or the     */
4717 /* function pointer if the name is set.  When found, fill in the other one  */
4718 /* so that the entire, complete, structure can be copied back to user space.*/
4719 /* ------------------------------------------------------------------------ */
4720 int fr_resolvefunc(data)
4721 void *data;
4722 {
4723 	ipfunc_resolve_t res, *ft;
4724 	int err;
4725 
4726 	err = BCOPYIN(data, &res, sizeof(res));
4727 	if (err != 0)
4728 		return EFAULT;
4729 
4730 	if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4731 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4732 			if (strncmp(res.ipfu_name, ft->ipfu_name,
4733 				    sizeof(res.ipfu_name)) == 0) {
4734 				res.ipfu_addr = ft->ipfu_addr;
4735 				res.ipfu_init = ft->ipfu_init;
4736 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4737 					return EFAULT;
4738 				return 0;
4739 			}
4740 	}
4741 	if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4742 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4743 			if (ft->ipfu_addr == res.ipfu_addr) {
4744 				(void) strncpy(res.ipfu_name, ft->ipfu_name,
4745 					       sizeof(res.ipfu_name));
4746 				res.ipfu_init = ft->ipfu_init;
4747 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4748 					return EFAULT;
4749 				return 0;
4750 			}
4751 	}
4752 	return ESRCH;
4753 }
4754 
4755 
4756 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4757     (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4758     (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4759     (defined(__OpenBSD__) && (OpenBSD < 200006))
4760 /*
4761  * From: NetBSD
4762  * ppsratecheck(): packets (or events) per second limitation.
4763  */
4764 int
4765 ppsratecheck(lasttime, curpps, maxpps)
4766 	struct timeval *lasttime;
4767 	int *curpps;
4768 	int maxpps;	/* maximum pps allowed */
4769 {
4770 	struct timeval tv, delta;
4771 	int rv;
4772 
4773 	GETKTIME(&tv);
4774 
4775 	delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4776 	delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4777 	if (delta.tv_usec < 0) {
4778 		delta.tv_sec--;
4779 		delta.tv_usec += 1000000;
4780 	}
4781 
4782 	/*
4783 	 * check for 0,0 is so that the message will be seen at least once.
4784 	 * if more than one second have passed since the last update of
4785 	 * lasttime, reset the counter.
4786 	 *
4787 	 * we do increment *curpps even in *curpps < maxpps case, as some may
4788 	 * try to use *curpps for stat purposes as well.
4789 	 */
4790 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4791 	    delta.tv_sec >= 1) {
4792 		*lasttime = tv;
4793 		*curpps = 0;
4794 		rv = 1;
4795 	} else if (maxpps < 0)
4796 		rv = 1;
4797 	else if (*curpps < maxpps)
4798 		rv = 1;
4799 	else
4800 		rv = 0;
4801 	*curpps = *curpps + 1;
4802 
4803 	return (rv);
4804 }
4805 #endif
4806 
4807 
4808 /* ------------------------------------------------------------------------ */
4809 /* Function:    fr_derefrule                                                */
4810 /* Returns:     int   - 0 == rule freed up, else rule not freed             */
4811 /* Parameters:  fr(I) - pointer to filter rule                              */
4812 /*                                                                          */
4813 /* Decrement the reference counter to a rule by one.  If it reaches zero,   */
4814 /* free it and any associated storage space being used by it.               */
4815 /* ------------------------------------------------------------------------ */
4816 int fr_derefrule(frp, ifs)
4817 frentry_t **frp;
4818 ipf_stack_t *ifs;
4819 {
4820 	frentry_t *fr;
4821 
4822 	fr = *frp;
4823 
4824 	MUTEX_ENTER(&fr->fr_lock);
4825 	fr->fr_ref--;
4826 	if (fr->fr_ref == 0) {
4827 		MUTEX_EXIT(&fr->fr_lock);
4828 		MUTEX_DESTROY(&fr->fr_lock);
4829 
4830 #ifdef IPFILTER_LOOKUP
4831 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
4832 		    ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
4833 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
4834 		    ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
4835 #endif
4836 
4837 		if (fr->fr_dsize) {
4838 			KFREES(fr->fr_data, fr->fr_dsize);
4839 		}
4840 		if ((fr->fr_flags & FR_COPIED) != 0) {
4841 			KFREE(fr);
4842 			return 0;
4843 		}
4844 		return 1;
4845 	} else {
4846 		MUTEX_EXIT(&fr->fr_lock);
4847 	}
4848 	*frp = NULL;
4849 	return -1;
4850 }
4851 
4852 
4853 #ifdef	IPFILTER_LOOKUP
4854 /* ------------------------------------------------------------------------ */
4855 /* Function:    fr_grpmapinit                                               */
4856 /* Returns:     int - 0 == success, else ESRCH because table entry not found*/
4857 /* Parameters:  fr(I) - pointer to rule to find hash table for              */
4858 /*                                                                          */
4859 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr.  */
4860 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap.                   */
4861 /* ------------------------------------------------------------------------ */
4862 static int fr_grpmapinit(fr, ifs)
4863 frentry_t *fr;
4864 ipf_stack_t *ifs;
4865 {
4866 	char name[FR_GROUPLEN];
4867 	iphtable_t *iph;
4868 
4869 #if defined(SNPRINTF) && defined(_KERNEL)
4870 	(void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
4871 #else
4872 	(void) sprintf(name, "%d", fr->fr_arg);
4873 #endif
4874 	iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4875 	if (iph == NULL)
4876 		return ESRCH;
4877 	if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
4878 		return ESRCH;
4879 	fr->fr_ptr = iph;
4880 	return 0;
4881 }
4882 
4883 
4884 /* ------------------------------------------------------------------------ */
4885 /* Function:    fr_srcgrpmap                                                */
4886 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
4887 /* Parameters:  fin(I)    - pointer to packet information                   */
4888 /*              passp(IO) - pointer to current/new filter decision (unused) */
4889 /*                                                                          */
4890 /* Look for a rule group head in a hash table, using the source address as  */
4891 /* the key, and descend into that group and continue matching rules against */
4892 /* the packet.                                                              */
4893 /* ------------------------------------------------------------------------ */
4894 frentry_t *fr_srcgrpmap(fin, passp)
4895 fr_info_t *fin;
4896 u_32_t *passp;
4897 {
4898 	frgroup_t *fg;
4899 	void *rval;
4900 	ipf_stack_t *ifs = fin->fin_ifs;
4901 
4902 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
4903 	if (rval == NULL)
4904 		return NULL;
4905 
4906 	fg = rval;
4907 	fin->fin_fr = fg->fg_start;
4908 	(void) fr_scanlist(fin, *passp);
4909 	return fin->fin_fr;
4910 }
4911 
4912 
4913 /* ------------------------------------------------------------------------ */
4914 /* Function:    fr_dstgrpmap                                                */
4915 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
4916 /* Parameters:  fin(I)    - pointer to packet information                   */
4917 /*              passp(IO) - pointer to current/new filter decision (unused) */
4918 /*                                                                          */
4919 /* Look for a rule group head in a hash table, using the destination        */
4920 /* address as the key, and descend into that group and continue matching    */
4921 /* rules against  the packet.                                               */
4922 /* ------------------------------------------------------------------------ */
4923 frentry_t *fr_dstgrpmap(fin, passp)
4924 fr_info_t *fin;
4925 u_32_t *passp;
4926 {
4927 	frgroup_t *fg;
4928 	void *rval;
4929 	ipf_stack_t *ifs = fin->fin_ifs;
4930 
4931 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
4932 	if (rval == NULL)
4933 		return NULL;
4934 
4935 	fg = rval;
4936 	fin->fin_fr = fg->fg_start;
4937 	(void) fr_scanlist(fin, *passp);
4938 	return fin->fin_fr;
4939 }
4940 #endif /* IPFILTER_LOOKUP */
4941 
4942 /*
4943  * Queue functions
4944  * ===============
4945  * These functions manage objects on queues for efficient timeouts.  There are
4946  * a number of system defined queues as well as user defined timeouts.  It is
4947  * expected that a lock is held in the domain in which the queue belongs
4948  * (i.e. either state or NAT) when calling any of these functions that prevents
4949  * fr_freetimeoutqueue() from being called at the same time as any other.
4950  */
4951 
4952 
4953 /* ------------------------------------------------------------------------ */
4954 /* Function:    fr_addtimeoutqueue                                          */
4955 /* Returns:     struct ifqtq * - NULL if malloc fails, else pointer to      */
4956 /*                               timeout queue with given interval.         */
4957 /* Parameters:  parent(I)  - pointer to pointer to parent node of this list */
4958 /*                           of interface queues.                           */
4959 /*              seconds(I) - timeout value in seconds for this queue.       */
4960 /*                                                                          */
4961 /* This routine first looks for a timeout queue that matches the interval   */
4962 /* being requested.  If it finds one, increments the reference counter and  */
4963 /* returns a pointer to it.  If none are found, it allocates a new one and  */
4964 /* inserts it at the top of the list.                                       */
4965 /*                                                                          */
4966 /* Locking.                                                                 */
4967 /* It is assumed that the caller of this function has an appropriate lock   */
4968 /* held (exclusively) in the domain that encompases 'parent'.               */
4969 /* ------------------------------------------------------------------------ */
4970 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
4971 ipftq_t **parent;
4972 u_int seconds;
4973 ipf_stack_t *ifs;
4974 {
4975 	ipftq_t *ifq;
4976 	u_int period;
4977 
4978 	period = seconds * IPF_HZ_DIVIDE;
4979 
4980 	MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
4981 	for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
4982 		if (ifq->ifq_ttl == period) {
4983 			/*
4984 			 * Reset the delete flag, if set, so the structure
4985 			 * gets reused rather than freed and reallocated.
4986 			 */
4987 			MUTEX_ENTER(&ifq->ifq_lock);
4988 			ifq->ifq_flags &= ~IFQF_DELETE;
4989 			ifq->ifq_ref++;
4990 			MUTEX_EXIT(&ifq->ifq_lock);
4991 			MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
4992 
4993 			return ifq;
4994 		}
4995 	}
4996 
4997 	KMALLOC(ifq, ipftq_t *);
4998 	if (ifq != NULL) {
4999 		ifq->ifq_ttl = period;
5000 		ifq->ifq_head = NULL;
5001 		ifq->ifq_tail = &ifq->ifq_head;
5002 		ifq->ifq_next = *parent;
5003 		ifq->ifq_pnext = parent;
5004 		ifq->ifq_ref = 1;
5005 		ifq->ifq_flags = IFQF_USER;
5006 		*parent = ifq;
5007 		ifs->ifs_fr_userifqs++;
5008 		MUTEX_NUKE(&ifq->ifq_lock);
5009 		MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5010 	}
5011 	MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5012 	return ifq;
5013 }
5014 
5015 
5016 /* ------------------------------------------------------------------------ */
5017 /* Function:    fr_deletetimeoutqueue                                       */
5018 /* Returns:     int    - new reference count value of the timeout queue     */
5019 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5020 /* Locks:       ifq->ifq_lock                                               */
5021 /*                                                                          */
5022 /* This routine must be called when we're discarding a pointer to a timeout */
5023 /* queue object, taking care of the reference counter.                      */
5024 /*                                                                          */
5025 /* Now that this just sets a DELETE flag, it requires the expire code to    */
5026 /* check the list of user defined timeout queues and call the free function */
5027 /* below (currently commented out) to stop memory leaking.  It is done this */
5028 /* way because the locking may not be sufficient to safely do a free when   */
5029 /* this function is called.                                                 */
5030 /* ------------------------------------------------------------------------ */
5031 int fr_deletetimeoutqueue(ifq)
5032 ipftq_t *ifq;
5033 {
5034 
5035 	ifq->ifq_ref--;
5036 	if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5037 		ifq->ifq_flags |= IFQF_DELETE;
5038 	}
5039 
5040 	return ifq->ifq_ref;
5041 }
5042 
5043 
5044 /* ------------------------------------------------------------------------ */
5045 /* Function:    fr_freetimeoutqueue                                         */
5046 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5047 /* Returns:     Nil                                                         */
5048 /*                                                                          */
5049 /* Locking:                                                                 */
5050 /* It is assumed that the caller of this function has an appropriate lock   */
5051 /* held (exclusively) in the domain that encompases the callers "domain".   */
5052 /* The ifq_lock for this structure should not be held.                      */
5053 /*                                                                          */
5054 /* Remove a user definde timeout queue from the list of queues it is in and */
5055 /* tidy up after this is done.                                              */
5056 /* ------------------------------------------------------------------------ */
5057 void fr_freetimeoutqueue(ifq, ifs)
5058 ipftq_t *ifq;
5059 ipf_stack_t *ifs;
5060 {
5061 
5062 
5063 	if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5064 	    ((ifq->ifq_flags & IFQF_USER) == 0)) {
5065 		printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5066 		       (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5067 		       ifq->ifq_ref);
5068 		return;
5069 	}
5070 
5071 	/*
5072 	 * Remove from its position in the list.
5073 	 */
5074 	*ifq->ifq_pnext = ifq->ifq_next;
5075 	if (ifq->ifq_next != NULL)
5076 		ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5077 
5078 	MUTEX_DESTROY(&ifq->ifq_lock);
5079 	ifs->ifs_fr_userifqs--;
5080 	KFREE(ifq);
5081 }
5082 
5083 
5084 /* ------------------------------------------------------------------------ */
5085 /* Function:    fr_deletequeueentry                                         */
5086 /* Returns:     Nil                                                         */
5087 /* Parameters:  tqe(I) - timeout queue entry to delete                      */
5088 /*              ifq(I) - timeout queue to remove entry from                 */
5089 /*                                                                          */
5090 /* Remove a tail queue entry from its queue and make it an orphan.          */
5091 /* fr_deletetimeoutqueue is called to make sure the reference count on the  */
5092 /* queue is correct.  We can't, however, call fr_freetimeoutqueue because   */
5093 /* the correct lock(s) may not be held that would make it safe to do so.    */
5094 /* ------------------------------------------------------------------------ */
5095 void fr_deletequeueentry(tqe)
5096 ipftqent_t *tqe;
5097 {
5098 	ipftq_t *ifq;
5099 
5100 	ifq = tqe->tqe_ifq;
5101 	if (ifq == NULL)
5102 		return;
5103 
5104 	MUTEX_ENTER(&ifq->ifq_lock);
5105 
5106 	if (tqe->tqe_pnext != NULL) {
5107 		*tqe->tqe_pnext = tqe->tqe_next;
5108 		if (tqe->tqe_next != NULL)
5109 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5110 		else    /* we must be the tail anyway */
5111 			ifq->ifq_tail = tqe->tqe_pnext;
5112 
5113 		tqe->tqe_pnext = NULL;
5114 		tqe->tqe_ifq = NULL;
5115 	}
5116 
5117 	(void) fr_deletetimeoutqueue(ifq);
5118 
5119 	MUTEX_EXIT(&ifq->ifq_lock);
5120 }
5121 
5122 
5123 /* ------------------------------------------------------------------------ */
5124 /* Function:    fr_queuefront                                               */
5125 /* Returns:     Nil                                                         */
5126 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5127 /*                                                                          */
5128 /* Move a queue entry to the front of the queue, if it isn't already there. */
5129 /* ------------------------------------------------------------------------ */
5130 void fr_queuefront(tqe)
5131 ipftqent_t *tqe;
5132 {
5133 	ipftq_t *ifq;
5134 
5135 	ifq = tqe->tqe_ifq;
5136 	if (ifq == NULL)
5137 		return;
5138 
5139 	MUTEX_ENTER(&ifq->ifq_lock);
5140 	if (ifq->ifq_head != tqe) {
5141 		*tqe->tqe_pnext = tqe->tqe_next;
5142 		if (tqe->tqe_next)
5143 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5144 		else
5145 			ifq->ifq_tail = tqe->tqe_pnext;
5146 
5147 		tqe->tqe_next = ifq->ifq_head;
5148 		ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5149 		ifq->ifq_head = tqe;
5150 		tqe->tqe_pnext = &ifq->ifq_head;
5151 	}
5152 	MUTEX_EXIT(&ifq->ifq_lock);
5153 }
5154 
5155 
5156 /* ------------------------------------------------------------------------ */
5157 /* Function:    fr_queueback                                                */
5158 /* Returns:     Nil                                                         */
5159 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5160 /*                                                                          */
5161 /* Move a queue entry to the back of the queue, if it isn't already there.  */
5162 /* ------------------------------------------------------------------------ */
5163 void fr_queueback(tqe, ifs)
5164 ipftqent_t *tqe;
5165 ipf_stack_t *ifs;
5166 {
5167 	ipftq_t *ifq;
5168 
5169 	ifq = tqe->tqe_ifq;
5170 	if (ifq == NULL)
5171 		return;
5172 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5173 
5174 	MUTEX_ENTER(&ifq->ifq_lock);
5175 	if (tqe->tqe_next == NULL) {		/* at the end already ? */
5176 		MUTEX_EXIT(&ifq->ifq_lock);
5177 		return;
5178 	}
5179 
5180 	/*
5181 	 * Remove from list
5182 	 */
5183 	*tqe->tqe_pnext = tqe->tqe_next;
5184 	tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5185 
5186 	/*
5187 	 * Make it the last entry.
5188 	 */
5189 	tqe->tqe_next = NULL;
5190 	tqe->tqe_pnext = ifq->ifq_tail;
5191 	*ifq->ifq_tail = tqe;
5192 	ifq->ifq_tail = &tqe->tqe_next;
5193 	MUTEX_EXIT(&ifq->ifq_lock);
5194 }
5195 
5196 
5197 /* ------------------------------------------------------------------------ */
5198 /* Function:    fr_queueappend                                              */
5199 /* Returns:     Nil                                                         */
5200 /* Parameters:  tqe(I)    - pointer to timeout queue entry                  */
5201 /*              ifq(I)    - pointer to timeout queue                        */
5202 /*              parent(I) - owing object pointer                            */
5203 /*                                                                          */
5204 /* Add a new item to this queue and put it on the very end.                 */
5205 /* ------------------------------------------------------------------------ */
5206 void fr_queueappend(tqe, ifq, parent, ifs)
5207 ipftqent_t *tqe;
5208 ipftq_t *ifq;
5209 void *parent;
5210 ipf_stack_t *ifs;
5211 {
5212 
5213 	MUTEX_ENTER(&ifq->ifq_lock);
5214 	tqe->tqe_parent = parent;
5215 	tqe->tqe_pnext = ifq->ifq_tail;
5216 	*ifq->ifq_tail = tqe;
5217 	ifq->ifq_tail = &tqe->tqe_next;
5218 	tqe->tqe_next = NULL;
5219 	tqe->tqe_ifq = ifq;
5220 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5221 	ifq->ifq_ref++;
5222 	MUTEX_EXIT(&ifq->ifq_lock);
5223 }
5224 
5225 
5226 /* ------------------------------------------------------------------------ */
5227 /* Function:    fr_movequeue                                                */
5228 /* Returns:     Nil                                                         */
5229 /* Parameters:  tq(I)   - pointer to timeout queue information              */
5230 /*              oifp(I) - old timeout queue entry was on                    */
5231 /*              nifp(I) - new timeout queue to put entry on                 */
5232 /*		ifs	- ipf stack instance				    */
5233 /*                                                                          */
5234 /* Move a queue entry from one timeout queue to another timeout queue.      */
5235 /* If it notices that the current entry is already last and does not need   */
5236 /* to move queue, the return.                                               */
5237 /* ------------------------------------------------------------------------ */
5238 void fr_movequeue(tqe, oifq, nifq, ifs)
5239 ipftqent_t *tqe;
5240 ipftq_t *oifq, *nifq;
5241 ipf_stack_t *ifs;
5242 {
5243 	/*
5244 	 * If the queue isn't changing, and the clock hasn't ticked
5245 	 * since the last update, the operation will be a no-op.
5246 	 */
5247 	if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5248 		return;
5249 
5250 	/*
5251 	 * Grab the lock and update the timers.
5252 	 */
5253 	MUTEX_ENTER(&oifq->ifq_lock);
5254 	tqe->tqe_touched = ifs->ifs_fr_ticks;
5255 	tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5256 
5257 	/*
5258 	 * The remainder of the operation can still be a no-op.
5259 	 *
5260 	 * If the queue isn't changing, check to see if
5261 	 * an update would be meaningless.
5262 	 */
5263 	if (oifq == nifq) {
5264 		if ((tqe->tqe_next == NULL) ||
5265 		    (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5266 			MUTEX_EXIT(&oifq->ifq_lock);
5267 			return;
5268 		}
5269 	}
5270 
5271 	/*
5272 	 * Remove from the old queue
5273 	 */
5274 	*tqe->tqe_pnext = tqe->tqe_next;
5275 	if (tqe->tqe_next)
5276 		tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5277 	else
5278 		oifq->ifq_tail = tqe->tqe_pnext;
5279 	tqe->tqe_next = NULL;
5280 
5281 	/*
5282 	 * If we're moving from one queue to another, release the lock on the
5283 	 * old queue and get a lock on the new queue.  For user defined queues,
5284 	 * if we're moving off it, call delete in case it can now be freed.
5285 	 */
5286 	if (oifq != nifq) {
5287 		tqe->tqe_ifq = NULL;
5288 
5289 		(void) fr_deletetimeoutqueue(oifq);
5290 
5291 		MUTEX_EXIT(&oifq->ifq_lock);
5292 
5293 		MUTEX_ENTER(&nifq->ifq_lock);
5294 
5295 		tqe->tqe_ifq = nifq;
5296 		nifq->ifq_ref++;
5297 	}
5298 
5299 	/*
5300 	 * Add to the bottom of the new queue
5301 	 */
5302 	tqe->tqe_pnext = nifq->ifq_tail;
5303 	*nifq->ifq_tail = tqe;
5304 	nifq->ifq_tail = &tqe->tqe_next;
5305 	MUTEX_EXIT(&nifq->ifq_lock);
5306 }
5307 
5308 
5309 /* ------------------------------------------------------------------------ */
5310 /* Function:    fr_updateipid                                               */
5311 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
5312 /* Parameters:  fin(I) - pointer to packet information                      */
5313 /*                                                                          */
5314 /* When we are doing NAT, change the IP of every packet to represent a      */
5315 /* single sequence of packets coming from the host, hiding any host         */
5316 /* specific sequencing that might otherwise be revealed.  If the packet is  */
5317 /* a fragment, then store the 'new' IPid in the fragment cache and look up  */
5318 /* the fragment cache for non-leading fragments.  If a non-leading fragment */
5319 /* has no match in the cache, return an error.                              */
5320 /* ------------------------------------------------------------------------ */
5321 static INLINE int fr_updateipid(fin)
5322 fr_info_t *fin;
5323 {
5324 	u_short id, ido, sums;
5325 	u_32_t sumd, sum;
5326 	ip_t *ip;
5327 
5328 	if (fin->fin_off != 0) {
5329 		sum = fr_ipid_knownfrag(fin);
5330 		if (sum == 0xffffffff)
5331 			return -1;
5332 		sum &= 0xffff;
5333 		id = (u_short)sum;
5334 	} else {
5335 		id = fr_nextipid(fin);
5336 		if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5337 			(void) fr_ipid_newfrag(fin, (u_32_t)id);
5338 	}
5339 
5340 	ip = fin->fin_ip;
5341 	ido = ntohs(ip->ip_id);
5342 	if (id == ido)
5343 		return 0;
5344 	ip->ip_id = htons(id);
5345 	CALC_SUMD(ido, id, sumd);	/* DESTRUCTIVE MACRO! id,ido change */
5346 	sum = (~ntohs(ip->ip_sum)) & 0xffff;
5347 	sum += sumd;
5348 	sum = (sum >> 16) + (sum & 0xffff);
5349 	sum = (sum >> 16) + (sum & 0xffff);
5350 	sums = ~(u_short)sum;
5351 	ip->ip_sum = htons(sums);
5352 	return 0;
5353 }
5354 
5355 
5356 #ifdef	NEED_FRGETIFNAME
5357 /* ------------------------------------------------------------------------ */
5358 /* Function:    fr_getifname                                                */
5359 /* Returns:     char *    - pointer to interface name                       */
5360 /* Parameters:  ifp(I)    - pointer to network interface                    */
5361 /*              buffer(O) - pointer to where to store interface name        */
5362 /*                                                                          */
5363 /* Constructs an interface name in the buffer passed.  The buffer passed is */
5364 /* expected to be at least LIFNAMSIZ in bytes big.  If buffer is passed in  */
5365 /* as a NULL pointer then return a pointer to a static array.               */
5366 /* ------------------------------------------------------------------------ */
5367 char *fr_getifname(ifp, buffer)
5368 struct ifnet *ifp;
5369 char *buffer;
5370 {
5371 	static char namebuf[LIFNAMSIZ];
5372 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5373      defined(__sgi) || defined(linux) || defined(_AIX51) || \
5374      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5375 	int unit, space;
5376 	char temp[20];
5377 	char *s;
5378 # endif
5379 
5380 	ASSERT(buffer != NULL);
5381 #ifdef notdef
5382 	if (buffer == NULL)
5383 		buffer = namebuf;
5384 #endif
5385 	(void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5386 	buffer[LIFNAMSIZ - 1] = '\0';
5387 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5388      defined(__sgi) || defined(_AIX51) || \
5389      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5390 	for (s = buffer; *s; s++)
5391 		;
5392 	unit = ifp->if_unit;
5393 	space = LIFNAMSIZ - (s - buffer);
5394 	if (space > 0) {
5395 #  if defined(SNPRINTF) && defined(_KERNEL)
5396 		(void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5397 #  else
5398 		(void) sprintf(temp, "%d", unit);
5399 #  endif
5400 		(void) strncpy(s, temp, space);
5401 	}
5402 # endif
5403 	return buffer;
5404 }
5405 #endif
5406 
5407 
5408 /* ------------------------------------------------------------------------ */
5409 /* Function:    fr_ioctlswitch                                              */
5410 /* Returns:     int     - -1 continue processing, else ioctl return value   */
5411 /* Parameters:  unit(I) - device unit opened                                */
5412 /*              data(I) - pointer to ioctl data                             */
5413 /*              cmd(I)  - ioctl command                                     */
5414 /*              mode(I) - mode value                                        */
5415 /*                                                                          */
5416 /* Based on the value of unit, call the appropriate ioctl handler or return */
5417 /* EIO if ipfilter is not running.   Also checks if write perms are req'd   */
5418 /* for the device in order to execute the ioctl.                            */
5419 /* ------------------------------------------------------------------------ */
5420 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5421 int unit, mode, uid;
5422 ioctlcmd_t cmd;
5423 void *data, *ctx;
5424 ipf_stack_t *ifs;
5425 {
5426 	int error = 0;
5427 
5428 	switch (unit)
5429 	{
5430 	case IPL_LOGIPF :
5431 		error = -1;
5432 		break;
5433 	case IPL_LOGNAT :
5434 		if (ifs->ifs_fr_running > 0)
5435 			error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5436 		else
5437 			error = EIO;
5438 		break;
5439 	case IPL_LOGSTATE :
5440 		if (ifs->ifs_fr_running > 0)
5441 			error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5442 		else
5443 			error = EIO;
5444 		break;
5445 	case IPL_LOGAUTH :
5446 		if (ifs->ifs_fr_running > 0) {
5447 			if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5448 			    (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5449 				if (!(mode & FWRITE)) {
5450 					error = EPERM;
5451 				} else {
5452 					error = frrequest(unit, cmd, data,
5453 						  ifs->ifs_fr_active, 1, ifs);
5454 				}
5455 			} else {
5456 				error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5457 			}
5458 		} else
5459 			error = EIO;
5460 		break;
5461 	case IPL_LOGSYNC :
5462 #ifdef IPFILTER_SYNC
5463 		if (ifs->ifs_fr_running > 0)
5464 			error = fr_sync_ioctl(data, cmd, mode, ifs);
5465 		else
5466 #endif
5467 			error = EIO;
5468 		break;
5469 	case IPL_LOGSCAN :
5470 #ifdef IPFILTER_SCAN
5471 		if (ifs->ifs_fr_running > 0)
5472 			error = fr_scan_ioctl(data, cmd, mode, ifs);
5473 		else
5474 #endif
5475 			error = EIO;
5476 		break;
5477 	case IPL_LOGLOOKUP :
5478 #ifdef IPFILTER_LOOKUP
5479 		if (ifs->ifs_fr_running > 0)
5480 			error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5481 		else
5482 #endif
5483 			error = EIO;
5484 		break;
5485 	default :
5486 		error = EIO;
5487 		break;
5488 	}
5489 
5490 	return error;
5491 }
5492 
5493 
5494 /*
5495  * This array defines the expected size of objects coming into the kernel
5496  * for the various recognised object types.
5497  */
5498 #define	NUM_OBJ_TYPES	19
5499 
5500 static	int	fr_objbytes[NUM_OBJ_TYPES][2] = {
5501 	{ 1,	sizeof(struct frentry) },		/* frentry */
5502 	{ 0,	sizeof(struct friostat) },
5503 	{ 0,	sizeof(struct fr_info) },
5504 	{ 0,	sizeof(struct fr_authstat) },
5505 	{ 0,	sizeof(struct ipfrstat) },
5506 	{ 0,	sizeof(struct ipnat) },
5507 	{ 0,	sizeof(struct natstat) },
5508 	{ 0,	sizeof(struct ipstate_save) },
5509 	{ 1,	sizeof(struct nat_save) },		/* nat_save */
5510 	{ 0,	sizeof(struct natlookup) },
5511 	{ 1,	sizeof(struct ipstate) },		/* ipstate */
5512 	{ 0,	sizeof(struct ips_stat) },
5513 	{ 0,	sizeof(struct frauth) },
5514 	{ 0,	sizeof(struct ipftune) },
5515 	{ 0,	sizeof(struct nat) },                   /* nat_t */
5516 	{ 0,	sizeof(struct ipfruleiter) },
5517 	{ 0,	sizeof(struct ipfgeniter) },
5518 	{ 0,	sizeof(struct ipftable) },
5519 	{ 0,	sizeof(struct ipflookupiter) }
5520 };
5521 
5522 
5523 /* ------------------------------------------------------------------------ */
5524 /* Function:    fr_inobj                                                    */
5525 /* Returns:     int     - 0 = success, else failure                         */
5526 /* Parameters:  data(I) - pointer to ioctl data                             */
5527 /*              ptr(I)  - pointer to store real data in                     */
5528 /*              type(I) - type of structure being moved                     */
5529 /*                                                                          */
5530 /* Copy in the contents of what the ipfobj_t points to.  In future, we      */
5531 /* add things to check for version numbers, sizes, etc, to make it backward */
5532 /* compatible at the ABI for user land.                                     */
5533 /* ------------------------------------------------------------------------ */
5534 int fr_inobj(data, ptr, type)
5535 void *data;
5536 void *ptr;
5537 int type;
5538 {
5539 	ipfobj_t obj;
5540 	int error = 0;
5541 
5542 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5543 		return EINVAL;
5544 
5545 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5546 	if (error != 0)
5547 		return EFAULT;
5548 
5549 	if (obj.ipfo_type != type)
5550 		return EINVAL;
5551 
5552 #ifndef	IPFILTER_COMPAT
5553 	if ((fr_objbytes[type][0] & 1) != 0) {
5554 		if (obj.ipfo_size < fr_objbytes[type][1])
5555 			return EINVAL;
5556 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5557 		return EINVAL;
5558 #else
5559 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5560 		error = fr_incomptrans(&obj, ptr);
5561 		return error;
5562 	}
5563 
5564 	if ((fr_objbytes[type][0] & 1) != 0 &&
5565 	    obj.ipfo_size < fr_objbytes[type][1] ||
5566 	    obj.ipfo_size != fr_objbytes[type][1])
5567 		return EINVAL;
5568 #endif
5569 
5570 	if ((fr_objbytes[type][0] & 1) != 0) {
5571 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5572 				fr_objbytes[type][1]);
5573 	} else {
5574 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5575 				obj.ipfo_size);
5576 	}
5577 	return error;
5578 }
5579 
5580 
5581 /* ------------------------------------------------------------------------ */
5582 /* Function:    fr_inobjsz                                                  */
5583 /* Returns:     int     - 0 = success, else failure                         */
5584 /* Parameters:  data(I) - pointer to ioctl data                             */
5585 /*              ptr(I)  - pointer to store real data in                     */
5586 /*              type(I) - type of structure being moved                     */
5587 /*              sz(I)   - size of data to copy                              */
5588 /*                                                                          */
5589 /* As per fr_inobj, except the size of the object to copy in is passed in   */
5590 /* but it must not be smaller than the size defined for the type and the    */
5591 /* type must allow for varied sized objects.  The extra requirement here is */
5592 /* that sz must match the size of the object being passed in - this is not  */
5593 /* not possible nor required in fr_inobj().                                 */
5594 /* ------------------------------------------------------------------------ */
5595 int fr_inobjsz(data, ptr, type, sz)
5596 void *data;
5597 void *ptr;
5598 int type, sz;
5599 {
5600 	ipfobj_t obj;
5601 	int error;
5602 
5603 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5604 		return EINVAL;
5605 	if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5606 		return EINVAL;
5607 
5608 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5609 	if (error != 0)
5610 		return EFAULT;
5611 
5612 	if (obj.ipfo_type != type)
5613 		return EINVAL;
5614 
5615 #ifndef	IPFILTER_COMPAT
5616 	if (obj.ipfo_size != sz)
5617 		return EINVAL;
5618 #else
5619 	if (obj.ipfo_rev != IPFILTER_VERSION)
5620 		/*XXX compatibility hook here */
5621 		/*EMPTY*/;
5622 	if (obj.ipfo_size != sz)
5623 		/* XXX compatibility hook here */
5624 		return EINVAL;
5625 #endif
5626 
5627 	error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5628 	return error;
5629 }
5630 
5631 
5632 /* ------------------------------------------------------------------------ */
5633 /* Function:    fr_outobjsz                                                 */
5634 /* Returns:     int     - 0 = success, else failure                         */
5635 /* Parameters:  data(I) - pointer to ioctl data                             */
5636 /*              ptr(I)  - pointer to store real data in                     */
5637 /*              type(I) - type of structure being moved                     */
5638 /*              sz(I)   - size of data to copy                              */
5639 /*                                                                          */
5640 /* As per fr_outobj, except the size of the object to copy out is passed in */
5641 /* but it must not be smaller than the size defined for the type and the    */
5642 /* type must allow for varied sized objects.  The extra requirement here is */
5643 /* that sz must match the size of the object being passed in - this is not  */
5644 /* not possible nor required in fr_outobj().                                */
5645 /* ------------------------------------------------------------------------ */
5646 int fr_outobjsz(data, ptr, type, sz)
5647 void *data;
5648 void *ptr;
5649 int type, sz;
5650 {
5651 	ipfobj_t obj;
5652 	int error;
5653 
5654 	if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5655 	    ((fr_objbytes[type][0] & 1) == 0) ||
5656 	    (sz < fr_objbytes[type][1]))
5657 		return EINVAL;
5658 
5659 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5660 	if (error != 0)
5661 		return EFAULT;
5662 
5663 	if (obj.ipfo_type != type)
5664 		return EINVAL;
5665 
5666 #ifndef	IPFILTER_COMPAT
5667 	if (obj.ipfo_size != sz)
5668 		return EINVAL;
5669 #else
5670 	if (obj.ipfo_rev != IPFILTER_VERSION)
5671 		/* XXX compatibility hook here */
5672 		/*EMPTY*/;
5673 	if (obj.ipfo_size != sz)
5674 		/* XXX compatibility hook here */
5675 		return EINVAL;
5676 #endif
5677 
5678 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5679 	return error;
5680 }
5681 
5682 
5683 /* ------------------------------------------------------------------------ */
5684 /* Function:    fr_outobj                                                   */
5685 /* Returns:     int     - 0 = success, else failure                         */
5686 /* Parameters:  data(I) - pointer to ioctl data                             */
5687 /*              ptr(I)  - pointer to store real data in                     */
5688 /*              type(I) - type of structure being moved                     */
5689 /*                                                                          */
5690 /* Copy out the contents of what ptr is to where ipfobj points to.  In      */
5691 /* future, we add things to check for version numbers, sizes, etc, to make  */
5692 /* it backward  compatible at the ABI for user land.                        */
5693 /* ------------------------------------------------------------------------ */
5694 int fr_outobj(data, ptr, type)
5695 void *data;
5696 void *ptr;
5697 int type;
5698 {
5699 	ipfobj_t obj;
5700 	int error;
5701 
5702 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5703 		return EINVAL;
5704 
5705 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5706 	if (error != 0)
5707 		return EFAULT;
5708 
5709 	if (obj.ipfo_type != type)
5710 		return EINVAL;
5711 
5712 #ifndef	IPFILTER_COMPAT
5713 	if ((fr_objbytes[type][0] & 1) != 0) {
5714 		if (obj.ipfo_size < fr_objbytes[type][1])
5715 			return EINVAL;
5716 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5717 		return EINVAL;
5718 #else
5719 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5720 		error = fr_outcomptrans(&obj, ptr);
5721 		return error;
5722 	}
5723 
5724 	if ((fr_objbytes[type][0] & 1) != 0 &&
5725 	    obj.ipfo_size < fr_objbytes[type][1] ||
5726 	    obj.ipfo_size != fr_objbytes[type][1])
5727 		return EINVAL;
5728 #endif
5729 
5730 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5731 	return error;
5732 }
5733 
5734 
5735 /* ------------------------------------------------------------------------ */
5736 /* Function:    fr_checkl4sum                                               */
5737 /* Returns:     int     - 0 = good, -1 = bad, 1 = cannot check              */
5738 /* Parameters:  fin(I) - pointer to packet information                      */
5739 /*                                                                          */
5740 /* If possible, calculate the layer 4 checksum for the packet.  If this is  */
5741 /* not possible, return without indicating a failure or success but in a    */
5742 /* way that is ditinguishable.                                              */
5743 /* ------------------------------------------------------------------------ */
5744 int fr_checkl4sum(fin)
5745 fr_info_t *fin;
5746 {
5747 	u_short sum, hdrsum, *csump;
5748 	udphdr_t *udp;
5749 	int dosum;
5750 	ipf_stack_t *ifs = fin->fin_ifs;
5751 
5752 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5753 	net_handle_t net_data_p;
5754 	if (fin->fin_v == 4)
5755 		net_data_p = ifs->ifs_ipf_ipv4;
5756 	else
5757 		net_data_p = ifs->ifs_ipf_ipv6;
5758 #endif
5759 
5760 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
5761 		return 0;
5762 
5763 	/*
5764 	 * If the TCP packet isn't a fragment, isn't too short and otherwise
5765 	 * isn't already considered "bad", then validate the checksum.  If
5766 	 * this check fails then considered the packet to be "bad".
5767 	 */
5768 	if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
5769 		return 1;
5770 
5771 	csump = NULL;
5772 	hdrsum = 0;
5773 	dosum = 0;
5774 	sum = 0;
5775 
5776 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5777 	ASSERT(fin->fin_m != NULL);
5778 	if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
5779 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
5780 			hdrsum = 0;
5781 			sum = 0;
5782 	} else {
5783 #endif
5784 		switch (fin->fin_p)
5785 		{
5786 		case IPPROTO_TCP :
5787 			csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
5788 			dosum = 1;
5789 			break;
5790 
5791 		case IPPROTO_UDP :
5792 			udp = fin->fin_dp;
5793 			if (udp->uh_sum != 0) {
5794 				csump = &udp->uh_sum;
5795 				dosum = 1;
5796 			}
5797 			break;
5798 
5799 		case IPPROTO_ICMP :
5800 			csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
5801 			dosum = 1;
5802 			break;
5803 
5804 		default :
5805 			return 1;
5806 			/*NOTREACHED*/
5807 		}
5808 
5809 		if (csump != NULL)
5810 			hdrsum = *csump;
5811 
5812 		if (dosum)
5813 			sum = fr_cksum(fin->fin_m, fin->fin_ip,
5814 				       fin->fin_p, fin->fin_dp);
5815 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5816 	}
5817 #endif
5818 #if !defined(_KERNEL)
5819 	if (sum == hdrsum) {
5820 		FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
5821 	} else {
5822 		FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
5823 	}
5824 #endif
5825 	if (hdrsum == sum)
5826 		return 0;
5827 	return -1;
5828 }
5829 
5830 
5831 /* ------------------------------------------------------------------------ */
5832 /* Function:    fr_ifpfillv4addr                                            */
5833 /* Returns:     int     - 0 = address update, -1 = address not updated      */
5834 /* Parameters:  atype(I)   - type of network address update to perform      */
5835 /*              sin(I)     - pointer to source of address information       */
5836 /*              mask(I)    - pointer to source of netmask information       */
5837 /*              inp(I)     - pointer to destination address store           */
5838 /*              inpmask(I) - pointer to destination netmask store           */
5839 /*                                                                          */
5840 /* Given a type of network address update (atype) to perform, copy          */
5841 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
5842 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
5843 /* which case the operation fails.  For all values of atype other than      */
5844 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
5845 /* value.                                                                   */
5846 /* ------------------------------------------------------------------------ */
5847 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
5848 int atype;
5849 struct sockaddr_in *sin, *mask;
5850 struct in_addr *inp, *inpmask;
5851 {
5852 	if (inpmask != NULL && atype != FRI_NETMASKED)
5853 		inpmask->s_addr = 0xffffffff;
5854 
5855 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
5856 		if (atype == FRI_NETMASKED) {
5857 			if (inpmask == NULL)
5858 				return -1;
5859 			inpmask->s_addr = mask->sin_addr.s_addr;
5860 		}
5861 		inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
5862 	} else {
5863 		inp->s_addr = sin->sin_addr.s_addr;
5864 	}
5865 	return 0;
5866 }
5867 
5868 
5869 #ifdef	USE_INET6
5870 /* ------------------------------------------------------------------------ */
5871 /* Function:    fr_ifpfillv6addr                                            */
5872 /* Returns:     int     - 0 = address update, -1 = address not updated      */
5873 /* Parameters:  atype(I)   - type of network address update to perform      */
5874 /*              sin(I)     - pointer to source of address information       */
5875 /*              mask(I)    - pointer to source of netmask information       */
5876 /*              inp(I)     - pointer to destination address store           */
5877 /*              inpmask(I) - pointer to destination netmask store           */
5878 /*                                                                          */
5879 /* Given a type of network address update (atype) to perform, copy          */
5880 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
5881 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
5882 /* which case the operation fails.  For all values of atype other than      */
5883 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
5884 /* value.                                                                   */
5885 /* ------------------------------------------------------------------------ */
5886 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
5887 int atype;
5888 struct sockaddr_in6 *sin, *mask;
5889 struct in_addr *inp, *inpmask;
5890 {
5891 	i6addr_t *src, *dst, *and, *dmask;
5892 
5893 	src = (i6addr_t *)&sin->sin6_addr;
5894 	and = (i6addr_t *)&mask->sin6_addr;
5895 	dst = (i6addr_t *)inp;
5896 	dmask = (i6addr_t *)inpmask;
5897 
5898 	if (inpmask != NULL && atype != FRI_NETMASKED) {
5899 		dmask->i6[0] = 0xffffffff;
5900 		dmask->i6[1] = 0xffffffff;
5901 		dmask->i6[2] = 0xffffffff;
5902 		dmask->i6[3] = 0xffffffff;
5903 	}
5904 
5905 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
5906 		if (atype == FRI_NETMASKED) {
5907 			if (inpmask == NULL)
5908 				return -1;
5909 			dmask->i6[0] = and->i6[0];
5910 			dmask->i6[1] = and->i6[1];
5911 			dmask->i6[2] = and->i6[2];
5912 			dmask->i6[3] = and->i6[3];
5913 		}
5914 
5915 		dst->i6[0] = src->i6[0] & and->i6[0];
5916 		dst->i6[1] = src->i6[1] & and->i6[1];
5917 		dst->i6[2] = src->i6[2] & and->i6[2];
5918 		dst->i6[3] = src->i6[3] & and->i6[3];
5919 	} else {
5920 		dst->i6[0] = src->i6[0];
5921 		dst->i6[1] = src->i6[1];
5922 		dst->i6[2] = src->i6[2];
5923 		dst->i6[3] = src->i6[3];
5924 	}
5925 	return 0;
5926 }
5927 #endif
5928 
5929 
5930 /* ------------------------------------------------------------------------ */
5931 /* Function:    fr_matchtag                                                 */
5932 /* Returns:     0 == mismatch, 1 == match.                                  */
5933 /* Parameters:  tag1(I) - pointer to first tag to compare                   */
5934 /*              tag2(I) - pointer to second tag to compare                  */
5935 /*                                                                          */
5936 /* Returns true (non-zero) or false(0) if the two tag structures can be     */
5937 /* considered to be a match or not match, respectively.  The tag is 16      */
5938 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so    */
5939 /* compare the ints instead, for speed. tag1 is the master of the           */
5940 /* comparison.  This function should only be called with both tag1 and tag2 */
5941 /* as non-NULL pointers.                                                    */
5942 /* ------------------------------------------------------------------------ */
5943 int fr_matchtag(tag1, tag2)
5944 ipftag_t *tag1, *tag2;
5945 {
5946 	if (tag1 == tag2)
5947 		return 1;
5948 
5949 	if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
5950 		return 1;
5951 
5952 	if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
5953 	    (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
5954 	    (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
5955 	    (tag1->ipt_num[3] == tag2->ipt_num[3]))
5956 		return 1;
5957 	return 0;
5958 }
5959 
5960 
5961 /* ------------------------------------------------------------------------ */
5962 /* Function:    fr_coalesce                                                 */
5963 /* Returns:     1 == success, -1 == failure, 0 == no change                 */
5964 /* Parameters:  fin(I) - pointer to packet information                      */
5965 /*                                                                          */
5966 /* Attempt to get all of the packet data into a single, contiguous buffer.  */
5967 /* If this call returns a failure then the buffers have also been freed.    */
5968 /* ------------------------------------------------------------------------ */
5969 int fr_coalesce(fin)
5970 fr_info_t *fin;
5971 {
5972 	ipf_stack_t *ifs = fin->fin_ifs;
5973 	if ((fin->fin_flx & FI_COALESCE) != 0)
5974 		return 1;
5975 
5976 	/*
5977 	 * If the mbuf pointers indicate that there is no mbuf to work with,
5978 	 * return but do not indicate success or failure.
5979 	 */
5980 	if (fin->fin_m == NULL || fin->fin_mp == NULL)
5981 		return 0;
5982 
5983 #if defined(_KERNEL)
5984 	if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
5985 		IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
5986 # ifdef MENTAT
5987 		FREE_MB_T(*fin->fin_mp);
5988 # endif
5989 		*fin->fin_mp = NULL;
5990 		fin->fin_m = NULL;
5991 		return -1;
5992 	}
5993 #else
5994 	fin = fin;	/* LINT */
5995 #endif
5996 	return 1;
5997 }
5998 
5999 
6000 /*
6001  * The following table lists all of the tunable variables that can be
6002  * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT.  The format of each row
6003  * in the table below is as follows:
6004  *
6005  * pointer to value, name of value, minimum, maximum, size of the value's
6006  *     container, value attribute flags
6007  *
6008  * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6009  * means the value can only be written to when IPFilter is loaded but disabled.
6010  * The obvious implication is if neither of these are set then the value can be
6011  * changed at any time without harm.
6012  */
6013 ipftuneable_t lcl_ipf_tuneables[] = {
6014 	/* filtering */
6015 	{ { NULL },	"fr_flags",		0,	0xffffffff,
6016 			0,		0 },
6017 	{ { NULL },	"fr_active",		0,	0,
6018 			0,		IPFT_RDONLY },
6019 	{ { NULL },	"fr_control_forwarding",	0, 1,
6020 			0,		0 },
6021 	{ { NULL },	"fr_update_ipid",	0,	1,
6022 			0,		0 },
6023 	{ { NULL },	"fr_chksrc",		0,	1,
6024 			0,		0 },
6025 	{ { NULL },	"fr_minttl",		0,	1,
6026 			0,		0 },
6027 	{ { NULL }, 	"fr_icmpminfragmtu",	0,	1,
6028 			0,		0 },
6029 	{ { NULL },		"fr_pass",		0,	0xffffffff,
6030 			0,		0 },
6031 #if SOLARIS2 >= 10
6032 	{ { NULL },	"ipf_loopback",		0,	1,
6033 			0,		IPFT_WRDISABLED },
6034 #endif
6035 	/* state */
6036 	{ { NULL }, "fr_tcpidletimeout",	1,	0x7fffffff,
6037 			0,	IPFT_WRDISABLED },
6038 	{ { NULL },	"fr_tcpclosewait",	1,	0x7fffffff,
6039 			0,	IPFT_WRDISABLED },
6040 	{ { NULL },	"fr_tcplastack",	1,	0x7fffffff,
6041 			0,		IPFT_WRDISABLED },
6042 	{ { NULL },	"fr_tcptimeout",	1,	0x7fffffff,
6043 			0,		IPFT_WRDISABLED },
6044 	{ { NULL },	"fr_tcpclosed",		1,	0x7fffffff,
6045 			0,		IPFT_WRDISABLED },
6046 	{ { NULL },	"fr_tcphalfclosed",	1,	0x7fffffff,
6047 			0,	IPFT_WRDISABLED },
6048 	{ { NULL },	"fr_udptimeout",	1,	0x7fffffff,
6049 			0,		IPFT_WRDISABLED },
6050 	{ { NULL }, "fr_udpacktimeout",	1,	0x7fffffff,
6051 			0,	IPFT_WRDISABLED },
6052 	{ { NULL },	"fr_icmptimeout",	1,	0x7fffffff,
6053 			0,		IPFT_WRDISABLED },
6054 	{ { NULL }, "fr_icmpacktimeout",	1,	0x7fffffff,
6055 			0,	IPFT_WRDISABLED },
6056 	{ { NULL }, "fr_iptimeout",		1,	0x7fffffff,
6057 			0,		IPFT_WRDISABLED },
6058 	{ { NULL },	"fr_statemax",		1,	0x7fffffff,
6059 			0,		0 },
6060 	{ { NULL },	"fr_statesize",		1,	0x7fffffff,
6061 			0,		IPFT_WRDISABLED },
6062 	{ { NULL },	"fr_state_lock",	0,	1,
6063 			0,		IPFT_RDONLY },
6064 	{ { NULL }, "fr_state_maxbucket", 1,	0x7fffffff,
6065 			0,	IPFT_WRDISABLED },
6066 	{ { NULL }, "fr_state_maxbucket_reset",	0, 1,
6067 			0, IPFT_WRDISABLED },
6068 	{ { NULL },	"ipstate_logging",	0,	1,
6069 			0,	0 },
6070 	{ { NULL },	"state_flush_level_hi",	1,	100,
6071 			0,		0 },
6072 	{ { NULL },	"state_flush_level_lo",	1,	100,
6073 			0,		0 },
6074 	/* nat */
6075 	{ { NULL },		"fr_nat_lock",		0,	1,
6076 			0,		IPFT_RDONLY },
6077 	{ { NULL },	"ipf_nattable_sz",	1,	0x7fffffff,
6078 			0,	IPFT_WRDISABLED },
6079 	{ { NULL }, "ipf_nattable_max",	1,	0x7fffffff,
6080 			0,	0 },
6081 	{ { NULL },	"ipf_natrules_sz",	1,	0x7fffffff,
6082 			0,	IPFT_WRDISABLED },
6083 	{ { NULL },	"ipf_rdrrules_sz",	1,	0x7fffffff,
6084 			0,	IPFT_WRDISABLED },
6085 	{ { NULL },	"ipf_hostmap_sz",	1,	0x7fffffff,
6086 			0,		IPFT_WRDISABLED },
6087 	{ { NULL }, "fr_nat_maxbucket",	1,	0x7fffffff,
6088 			0,	IPFT_WRDISABLED },
6089 	{ { NULL },	"fr_nat_maxbucket_reset",	0, 1,
6090 			0,	IPFT_WRDISABLED },
6091 	{ { NULL },		"nat_logging",		0,	1,
6092 			0,		0 },
6093 	{ { NULL },	"fr_defnatage",		1,	0x7fffffff,
6094 			0,		IPFT_WRDISABLED },
6095 	{ { NULL },	"fr_defnatipage",	1,	0x7fffffff,
6096 			0,		IPFT_WRDISABLED },
6097 	{ { NULL }, "fr_defnaticmpage",	1,	0x7fffffff,
6098 			0,	IPFT_WRDISABLED },
6099 	{ { NULL },	"nat_flush_level_hi",	1,	100,
6100 			0,		0 },
6101 	{ { NULL },	"nat_flush_level_lo",	1,	100,
6102 			0,		0 },
6103 	/* frag */
6104 	{ { NULL },	"ipfr_size",		1,	0x7fffffff,
6105 			0,		IPFT_WRDISABLED },
6106 	{ { NULL },	"fr_ipfrttl",		1,	0x7fffffff,
6107 			0,		IPFT_WRDISABLED },
6108 #ifdef IPFILTER_LOG
6109 	/* log */
6110 	{ { NULL },	"ipl_suppress",		0,	1,
6111 			0,		0 },
6112 	{ { NULL },	"ipl_buffer_sz",	0,	0,
6113 			0,		IPFT_RDONLY },
6114 	{ { NULL },	"ipl_logmax",		0,	0x7fffffff,
6115 			0,		IPFT_WRDISABLED },
6116 	{ { NULL },	"ipl_logall",		0,	1,
6117 			0,		0 },
6118 	{ { NULL },	"ipl_logsize",		0,	0x80000,
6119 			0,		0 },
6120 #endif
6121 	{ { NULL },		NULL,			0,	0 }
6122 };
6123 
6124 static ipftuneable_t *
6125 tune_lookup(ipf_stack_t *ifs, char *name)
6126 {
6127     int i;
6128 
6129     for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6130 	if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6131 	    return (&ifs->ifs_ipf_tuneables[i]);
6132     }
6133     return (NULL);
6134 }
6135 
6136 #ifdef _KERNEL
6137 extern dev_info_t *ipf_dev_info;
6138 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6139 #endif
6140 
6141 /* -------------------------------------------------------------------- */
6142 /* Function:	ipftuneable_setdefs()					*/
6143 /* Returns:		void						*/
6144 /* Parameters:	ifs - pointer to newly allocated IPF instance		*/
6145 /*				assigned to	IP instance		*/
6146 /*									*/
6147 /* Function initializes IPF instance variables. Function is invoked	*/
6148 /* from	ipftuneable_alloc(). ipftuneable_alloc() is called only one	*/
6149 /* time during IP instance lifetime - at the time of IP instance	*/
6150 /* creation. Anytime IP	instance is being created new private IPF	*/
6151 /* instance is allocated and assigned to it. The moment of IP 		*/
6152 /* instance creation is the right time to initialize those IPF 		*/
6153 /* variables.								*/
6154 /*									*/
6155 /* -------------------------------------------------------------------- */
6156 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6157 {
6158 	ifs->ifs_ipfr_size = IPFT_SIZE;
6159 	ifs->ifs_fr_ipfrttl = 120;	/* 60 seconds */
6160 
6161 	/* it comes from fr_authinit() in IPF auth */
6162 	ifs->ifs_fr_authsize = FR_NUMAUTH;
6163 	ifs->ifs_fr_defaultauthage = 600;
6164 
6165 	/* it comes from fr_stateinit() in IPF state */
6166 	ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5);	/* five days */
6167 	ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6168 	ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6169 	ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6170 	ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6171 	ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600);	/* 2 hours */
6172 	ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6173 	ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6174 	ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6175 	ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6176 	ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6177 	ifs->ifs_fr_statemax = IPSTATE_MAX;
6178 	ifs->ifs_fr_statesize = IPSTATE_SIZE;
6179 	ifs->ifs_fr_state_maxbucket_reset = 1;
6180 	ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6181 	ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6182 
6183 	/* it comes from fr_natinit() in ipnat */
6184 	ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6185 	ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6186 	ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6187 	ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6188 	ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6189 	ifs->ifs_fr_nat_maxbucket_reset = 1;
6190 	ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6191 	ifs->ifs_fr_defnatipage = 120;		/* 60 seconds */
6192 	ifs->ifs_fr_defnaticmpage = 6;		/* 3 seconds */
6193 	ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6194 	ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6195 
6196 #ifdef IPFILTER_LOG
6197 	/* it comes from fr_loginit() in IPF log */
6198 	ifs->ifs_ipl_suppress = 1;
6199 	ifs->ifs_ipl_logmax = IPL_LOGMAX;
6200 	ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6201 
6202 	/* from fr_natinit() */
6203 	ifs->ifs_nat_logging = 1;
6204 
6205 	/* from fr_stateinit() */
6206 	ifs->ifs_ipstate_logging = 1;
6207 #else
6208 	/* from fr_natinit() */
6209 	ifs->ifs_nat_logging = 0;
6210 
6211 	/* from fr_stateinit() */
6212 	ifs->ifs_ipstate_logging = 0;
6213 #endif
6214 	ifs->ifs_ipf_loopback = 0;
6215 
6216 }
6217 /*
6218  * Allocate a per-stack tuneable and copy in the names. Then
6219  * set it to point to each of the per-stack tunables.
6220  */
6221 void
6222 ipftuneable_alloc(ipf_stack_t *ifs)
6223 {
6224     ipftuneable_t *item;
6225 
6226     KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6227 	sizeof (lcl_ipf_tuneables));
6228     bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6229 	sizeof (lcl_ipf_tuneables));
6230 
6231 #define TUNE_SET(_ifs, _name, _field)			\
6232     item = tune_lookup((_ifs), (_name));		\
6233     if (item != NULL) {					\
6234 	item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field);	\
6235 	item->ipft_sz = sizeof ((_ifs)->_field);	\
6236     }
6237 
6238     TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6239     TUNE_SET(ifs, "fr_active", ifs_fr_active);
6240     TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6241     TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6242     TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6243     TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6244     TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6245     TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6246     TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6247     TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6248     TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6249     TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6250     TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6251     TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6252     TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6253     TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6254     TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6255     TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6256     TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6257     TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6258     TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6259     TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6260     TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6261     TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6262     TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6263     TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6264     TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6265     TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6266     TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6267     TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6268     TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6269     TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6270     TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6271     TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6272     TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6273     TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6274     TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6275     TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6276     TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6277     TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6278     TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6279     TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6280     TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6281     TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6282 #ifdef IPFILTER_LOG
6283     TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6284     TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6285     TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6286     TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6287     TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6288 #endif
6289 #undef TUNE_SET
6290 
6291 	ipftuneable_setdefs(ifs);
6292 
6293 #ifdef _KERNEL
6294     (void) ipf_property_update(ipf_dev_info, ifs);
6295 #endif
6296 }
6297 
6298 void
6299 ipftuneable_free(ipf_stack_t *ifs)
6300 {
6301 	KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6302 	ifs->ifs_ipf_tuneables = NULL;
6303 }
6304 
6305 /* ------------------------------------------------------------------------ */
6306 /* Function:    fr_findtunebycookie                                         */
6307 /* Returns:     NULL = search failed, else pointer to tune struct           */
6308 /* Parameters:  cookie(I) - cookie value to search for amongst tuneables    */
6309 /*              next(O)   - pointer to place to store the cookie for the    */
6310 /*                          "next" tuneable, if it is desired.              */
6311 /*                                                                          */
6312 /* This function is used to walk through all of the existing tunables with  */
6313 /* successive calls.  It searches the known tunables for the one which has  */
6314 /* a matching value for "cookie" - ie its address.  When returning a match, */
6315 /* the next one to be found may be returned inside next.                    */
6316 /* ------------------------------------------------------------------------ */
6317 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6318 void *cookie, **next;
6319 ipf_stack_t * ifs;
6320 {
6321 	ipftuneable_t *ta, **tap;
6322 
6323 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6324 		if (ta == cookie) {
6325 			if (next != NULL) {
6326 				/*
6327 				 * If the next entry in the array has a name
6328 				 * present, then return a pointer to it for
6329 				 * where to go next, else return a pointer to
6330 				 * the dynaminc list as a key to search there
6331 				 * next.  This facilitates a weak linking of
6332 				 * the two "lists" together.
6333 				 */
6334 				if ((ta + 1)->ipft_name != NULL)
6335 					*next = ta + 1;
6336 				else
6337 					*next = &ifs->ifs_ipf_tunelist;
6338 			}
6339 			return ta;
6340 		}
6341 
6342 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6343 		if (tap == cookie) {
6344 			if (next != NULL)
6345 				*next = &ta->ipft_next;
6346 			return ta;
6347 		}
6348 
6349 	if (next != NULL)
6350 		*next = NULL;
6351 	return NULL;
6352 }
6353 
6354 
6355 /* ------------------------------------------------------------------------ */
6356 /* Function:    fr_findtunebyname                                           */
6357 /* Returns:     NULL = search failed, else pointer to tune struct           */
6358 /* Parameters:  name(I) - name of the tuneable entry to find.               */
6359 /*                                                                          */
6360 /* Search the static array of tuneables and the list of dynamic tuneables   */
6361 /* for an entry with a matching name.  If we can find one, return a pointer */
6362 /* to the matching structure.                                               */
6363 /* ------------------------------------------------------------------------ */
6364 static ipftuneable_t *fr_findtunebyname(name, ifs)
6365 const char *name;
6366 ipf_stack_t *ifs;
6367 {
6368 	ipftuneable_t *ta;
6369 
6370 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6371 		if (!strcmp(ta->ipft_name, name)) {
6372 			return ta;
6373 		}
6374 
6375 	for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6376 		if (!strcmp(ta->ipft_name, name)) {
6377 			return ta;
6378 		}
6379 
6380 	return NULL;
6381 }
6382 
6383 
6384 /* ------------------------------------------------------------------------ */
6385 /* Function:    fr_addipftune                                               */
6386 /* Returns:     int - 0 == success, else failure                            */
6387 /* Parameters:  newtune - pointer to new tune struct to add to tuneables    */
6388 /*                                                                          */
6389 /* Appends the tune structure pointer to by "newtune" to the end of the     */
6390 /* current list of "dynamic" tuneable parameters.  Once added, the owner    */
6391 /* of the object is not expected to ever change "ipft_next".                */
6392 /* ------------------------------------------------------------------------ */
6393 int fr_addipftune(newtune, ifs)
6394 ipftuneable_t *newtune;
6395 ipf_stack_t *ifs;
6396 {
6397 	ipftuneable_t *ta, **tap;
6398 
6399 	ta = fr_findtunebyname(newtune->ipft_name, ifs);
6400 	if (ta != NULL)
6401 		return EEXIST;
6402 
6403 	for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6404 		;
6405 
6406 	newtune->ipft_next = NULL;
6407 	*tap = newtune;
6408 	return 0;
6409 }
6410 
6411 
6412 /* ------------------------------------------------------------------------ */
6413 /* Function:    fr_delipftune                                               */
6414 /* Returns:     int - 0 == success, else failure                            */
6415 /* Parameters:  oldtune - pointer to tune struct to remove from the list of */
6416 /*                        current dynamic tuneables                         */
6417 /*                                                                          */
6418 /* Search for the tune structure, by pointer, in the list of those that are */
6419 /* dynamically added at run time.  If found, adjust the list so that this   */
6420 /* structure is no longer part of it.                                       */
6421 /* ------------------------------------------------------------------------ */
6422 int fr_delipftune(oldtune, ifs)
6423 ipftuneable_t *oldtune;
6424 ipf_stack_t *ifs;
6425 {
6426 	ipftuneable_t *ta, **tap;
6427 
6428 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6429 		if (ta == oldtune) {
6430 			*tap = oldtune->ipft_next;
6431 			oldtune->ipft_next = NULL;
6432 			return 0;
6433 		}
6434 
6435 	return ESRCH;
6436 }
6437 
6438 
6439 /* ------------------------------------------------------------------------ */
6440 /* Function:    fr_ipftune                                                  */
6441 /* Returns:     int - 0 == success, else failure                            */
6442 /* Parameters:  cmd(I)  - ioctl command number                              */
6443 /*              data(I) - pointer to ioctl data structure                   */
6444 /*                                                                          */
6445 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET.  These  */
6446 /* three ioctls provide the means to access and control global variables    */
6447 /* within IPFilter, allowing (for example) timeouts and table sizes to be   */
6448 /* changed without rebooting, reloading or recompiling.  The initialisation */
6449 /* and 'destruction' routines of the various components of ipfilter are all */
6450 /* each responsible for handling their own values being too big.            */
6451 /* ------------------------------------------------------------------------ */
6452 int fr_ipftune(cmd, data, ifs)
6453 ioctlcmd_t cmd;
6454 void *data;
6455 ipf_stack_t *ifs;
6456 {
6457 	ipftuneable_t *ta;
6458 	ipftune_t tu;
6459 	void *cookie;
6460 	int error;
6461 
6462 	error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6463 	if (error != 0)
6464 		return error;
6465 
6466 	tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6467 	cookie = tu.ipft_cookie;
6468 	ta = NULL;
6469 
6470 	switch (cmd)
6471 	{
6472 	case SIOCIPFGETNEXT :
6473 		/*
6474 		 * If cookie is non-NULL, assume it to be a pointer to the last
6475 		 * entry we looked at, so find it (if possible) and return a
6476 		 * pointer to the next one after it.  The last entry in the
6477 		 * the table is a NULL entry, so when we get to it, set cookie
6478 		 * to NULL and return that, indicating end of list, erstwhile
6479 		 * if we come in with cookie set to NULL, we are starting anew
6480 		 * at the front of the list.
6481 		 */
6482 		if (cookie != NULL) {
6483 			ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6484 		} else {
6485 			ta = ifs->ifs_ipf_tuneables;
6486 			tu.ipft_cookie = ta + 1;
6487 		}
6488 		if (ta != NULL) {
6489 			/*
6490 			 * Entry found, but does the data pointed to by that
6491 			 * row fit in what we can return?
6492 			 */
6493 			if (ta->ipft_sz > sizeof(tu.ipft_un))
6494 				return EINVAL;
6495 
6496 			tu.ipft_vlong = 0;
6497 			if (ta->ipft_sz == sizeof(u_long))
6498 				tu.ipft_vlong = *ta->ipft_plong;
6499 			else if (ta->ipft_sz == sizeof(u_int))
6500 				tu.ipft_vint = *ta->ipft_pint;
6501 			else if (ta->ipft_sz == sizeof(u_short))
6502 				tu.ipft_vshort = *ta->ipft_pshort;
6503 			else if (ta->ipft_sz == sizeof(u_char))
6504 				tu.ipft_vchar = *ta->ipft_pchar;
6505 
6506 			tu.ipft_sz = ta->ipft_sz;
6507 			tu.ipft_min = ta->ipft_min;
6508 			tu.ipft_max = ta->ipft_max;
6509 			tu.ipft_flags = ta->ipft_flags;
6510 			bcopy(ta->ipft_name, tu.ipft_name,
6511 			      MIN(sizeof(tu.ipft_name),
6512 				  strlen(ta->ipft_name) + 1));
6513 		}
6514 		error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6515 		break;
6516 
6517 	case SIOCIPFGET :
6518 	case SIOCIPFSET :
6519 		/*
6520 		 * Search by name or by cookie value for a particular entry
6521 		 * in the tuning paramter table.
6522 		 */
6523 		error = ESRCH;
6524 		if (cookie != NULL) {
6525 			ta = fr_findtunebycookie(cookie, NULL, ifs);
6526 			if (ta != NULL)
6527 				error = 0;
6528 		} else if (tu.ipft_name[0] != '\0') {
6529 			ta = fr_findtunebyname(tu.ipft_name, ifs);
6530 			if (ta != NULL)
6531 				error = 0;
6532 		}
6533 		if (error != 0)
6534 			break;
6535 
6536 		if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6537 			/*
6538 			 * Fetch the tuning parameters for a particular value
6539 			 */
6540 			tu.ipft_vlong = 0;
6541 			if (ta->ipft_sz == sizeof(u_long))
6542 				tu.ipft_vlong = *ta->ipft_plong;
6543 			else if (ta->ipft_sz == sizeof(u_int))
6544 				tu.ipft_vint = *ta->ipft_pint;
6545 			else if (ta->ipft_sz == sizeof(u_short))
6546 				tu.ipft_vshort = *ta->ipft_pshort;
6547 			else if (ta->ipft_sz == sizeof(u_char))
6548 				tu.ipft_vchar = *ta->ipft_pchar;
6549 			tu.ipft_cookie = ta;
6550 			tu.ipft_sz = ta->ipft_sz;
6551 			tu.ipft_min = ta->ipft_min;
6552 			tu.ipft_max = ta->ipft_max;
6553 			tu.ipft_flags = ta->ipft_flags;
6554 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6555 
6556 		} else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6557 			/*
6558 			 * Set an internal parameter.  The hard part here is
6559 			 * getting the new value safely and correctly out of
6560 			 * the kernel (given we only know its size, not type.)
6561 			 */
6562 			u_long in;
6563 
6564 			if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6565 			    (ifs->ifs_fr_running > 0)) {
6566 				error = EBUSY;
6567 				break;
6568 			}
6569 
6570 			in = tu.ipft_vlong;
6571 			if (in < ta->ipft_min || in > ta->ipft_max) {
6572 				error = EINVAL;
6573 				break;
6574 			}
6575 
6576 			if (ta->ipft_sz == sizeof(u_long)) {
6577 				tu.ipft_vlong = *ta->ipft_plong;
6578 				*ta->ipft_plong = in;
6579 			} else if (ta->ipft_sz == sizeof(u_int)) {
6580 				tu.ipft_vint = *ta->ipft_pint;
6581 				*ta->ipft_pint = (u_int)(in & 0xffffffff);
6582 			} else if (ta->ipft_sz == sizeof(u_short)) {
6583 				tu.ipft_vshort = *ta->ipft_pshort;
6584 				*ta->ipft_pshort = (u_short)(in & 0xffff);
6585 			} else if (ta->ipft_sz == sizeof(u_char)) {
6586 				tu.ipft_vchar = *ta->ipft_pchar;
6587 				*ta->ipft_pchar = (u_char)(in & 0xff);
6588 			}
6589 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6590 		}
6591 		break;
6592 
6593 	default :
6594 		error = EINVAL;
6595 		break;
6596 	}
6597 
6598 	return error;
6599 }
6600 
6601 
6602 /* ------------------------------------------------------------------------ */
6603 /* Function:    fr_initialise                                               */
6604 /* Returns:     int - 0 == success,  < 0 == failure                         */
6605 /* Parameters:  None.                                                       */
6606 /*                                                                          */
6607 /* Call of the initialise functions for all the various subsystems inside   */
6608 /* of IPFilter.  If any of them should fail, return immeadiately a failure  */
6609 /* BUT do not try to recover from the error here.                           */
6610 /* ------------------------------------------------------------------------ */
6611 int fr_initialise(ifs)
6612 ipf_stack_t *ifs;
6613 {
6614 	int i;
6615 
6616 #ifdef IPFILTER_LOG
6617 	i = fr_loginit(ifs);
6618 	if (i < 0)
6619 		return -10 + i;
6620 #endif
6621 	i = fr_natinit(ifs);
6622 	if (i < 0)
6623 		return -20 + i;
6624 
6625 	i = fr_stateinit(ifs);
6626 	if (i < 0)
6627 		return -30 + i;
6628 
6629 	i = fr_authinit(ifs);
6630 	if (i < 0)
6631 		return -40 + i;
6632 
6633 	i = fr_fraginit(ifs);
6634 	if (i < 0)
6635 		return -50 + i;
6636 
6637 	i = appr_init(ifs);
6638 	if (i < 0)
6639 		return -60 + i;
6640 
6641 #ifdef IPFILTER_SYNC
6642 	i = ipfsync_init(ifs);
6643 	if (i < 0)
6644 		return -70 + i;
6645 #endif
6646 #ifdef IPFILTER_SCAN
6647 	i = ipsc_init(ifs);
6648 	if (i < 0)
6649 		return -80 + i;
6650 #endif
6651 #ifdef IPFILTER_LOOKUP
6652 	i = ip_lookup_init(ifs);
6653 	if (i < 0)
6654 		return -90 + i;
6655 #endif
6656 #ifdef IPFILTER_COMPILED
6657 	ipfrule_add(ifs);
6658 #endif
6659 	return 0;
6660 }
6661 
6662 
6663 /* ------------------------------------------------------------------------ */
6664 /* Function:    fr_deinitialise                                             */
6665 /* Returns:     None.                                                       */
6666 /* Parameters:  None.                                                       */
6667 /*                                                                          */
6668 /* Call all the various subsystem cleanup routines to deallocate memory or  */
6669 /* destroy locks or whatever they've done that they need to now undo.       */
6670 /* The order here IS important as there are some cross references of        */
6671 /* internal data structures.                                                */
6672 /* ------------------------------------------------------------------------ */
6673 void fr_deinitialise(ifs)
6674 ipf_stack_t *ifs;
6675 {
6676 	fr_fragunload(ifs);
6677 	fr_authunload(ifs);
6678 	fr_natunload(ifs);
6679 	fr_stateunload(ifs);
6680 #ifdef IPFILTER_SCAN
6681 	fr_scanunload(ifs);
6682 #endif
6683 	appr_unload(ifs);
6684 
6685 #ifdef IPFILTER_COMPILED
6686 	ipfrule_remove(ifs);
6687 #endif
6688 
6689 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6690 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6691 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6692 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6693 
6694 #ifdef IPFILTER_LOOKUP
6695 	ip_lookup_unload(ifs);
6696 #endif
6697 
6698 #ifdef IPFILTER_LOG
6699 	fr_logunload(ifs);
6700 #endif
6701 }
6702 
6703 
6704 /* ------------------------------------------------------------------------ */
6705 /* Function:    fr_zerostats                                                */
6706 /* Returns:     int - 0 = success, else failure                             */
6707 /* Parameters:  data(O) - pointer to pointer for copying data back to       */
6708 /*                                                                          */
6709 /* Copies the current statistics out to userspace and then zero's the       */
6710 /* current ones in the kernel. The lock is only held across the bzero() as  */
6711 /* the copyout may result in paging (ie network activity.)                  */
6712 /* ------------------------------------------------------------------------ */
6713 int	fr_zerostats(data, ifs)
6714 caddr_t	data;
6715 ipf_stack_t *ifs;
6716 {
6717 	friostat_t fio;
6718 	int error;
6719 
6720 	fr_getstat(&fio, ifs);
6721 	error = copyoutptr(&fio, data, sizeof(fio));
6722 	if (error)
6723 		return EFAULT;
6724 
6725 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
6726 	bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6727 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6728 
6729 	return 0;
6730 }
6731 
6732 
6733 #ifdef _KERNEL
6734 /* ------------------------------------------------------------------------ */
6735 /* Function:    fr_resolvedest                                              */
6736 /* Returns:     Nil                                                         */
6737 /* Parameters:  fdp(IO) - pointer to destination information to resolve     */
6738 /*              v(I)    - IP protocol version to match                      */
6739 /*                                                                          */
6740 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6741 /* if a matching name can be found for the particular IP protocol version   */
6742 /* then store the interface pointer in the frdest struct.  If no match is   */
6743 /* found, then set the interface pointer to be -1 as NULL is considered to  */
6744 /* indicate there is no information at all in the structure.                */
6745 /* ------------------------------------------------------------------------ */
6746 void fr_resolvedest(fdp, v, ifs)
6747 frdest_t *fdp;
6748 int v;
6749 ipf_stack_t *ifs;
6750 {
6751 	fdp->fd_ifp = NULL;
6752 
6753   	if (*fdp->fd_ifname != '\0') {
6754  		fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6755 		if (fdp->fd_ifp == NULL)
6756 			fdp->fd_ifp = (void *)-1;
6757 	}
6758 }
6759 #endif /* _KERNEL */
6760 
6761 
6762 /* ------------------------------------------------------------------------ */
6763 /* Function:    fr_resolvenic                                               */
6764 /* Returns:     void* - NULL = wildcard name, -1 = failed to find NIC, else */
6765 /*                      pointer to interface structure for NIC              */
6766 /* Parameters:  name(I) - complete interface name                           */
6767 /*              v(I)    - IP protocol version                               */
6768 /*                                                                          */
6769 /* Look for a network interface structure that firstly has a matching name  */
6770 /* to that passed in and that is also being used for that IP protocol       */
6771 /* version (necessary on some platforms where there are separate listings   */
6772 /* for both IPv4 and IPv6 on the same physical NIC.                         */
6773 /*                                                                          */
6774 /* One might wonder why name gets terminated with a \0 byte in here.  The   */
6775 /* reason is an interface name could get into the kernel structures of ipf  */
6776 /* in any number of ways and so long as they all use the same sized array   */
6777 /* to put the name in, it makes sense to ensure it gets null terminated     */
6778 /* before it is used for its intended purpose - finding its match in the    */
6779 /* kernel's list of configured interfaces.                                  */
6780 /*                                                                          */
6781 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an     */
6782 /*       array for the name that is LIFNAMSIZ bytes (at least) in length.   */
6783 /* ------------------------------------------------------------------------ */
6784 void *fr_resolvenic(name, v, ifs)
6785 char *name;
6786 int v;
6787 ipf_stack_t *ifs;
6788 {
6789 	void *nic;
6790 
6791 	if (name[0] == '\0')
6792 		return NULL;
6793 
6794 	if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
6795 		return NULL;
6796 	}
6797 
6798 	name[LIFNAMSIZ - 1] = '\0';
6799 
6800 	nic = GETIFP(name, v, ifs);
6801 	if (nic == NULL)
6802 		nic = (void *)-1;
6803 	return nic;
6804 }
6805 
6806 
6807 /* ------------------------------------------------------------------------ */
6808 /* Function:    ipf_expiretokens                                            */
6809 /* Returns:     None.                                                       */
6810 /* Parameters:  ifs - ipf stack instance                                    */
6811 /*                                                                          */
6812 /* This function is run every ipf tick to see if there are any tokens that  */
6813 /* have been held for too long and need to be freed up.                     */
6814 /* ------------------------------------------------------------------------ */
6815 void ipf_expiretokens(ifs)
6816 ipf_stack_t *ifs;
6817 {
6818 	ipftoken_t *it;
6819 
6820 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
6821 	while ((it = ifs->ifs_ipftokenhead) != NULL) {
6822 		if (it->ipt_die > ifs->ifs_fr_ticks)
6823 			break;
6824 
6825 		ipf_freetoken(it, ifs);
6826 	}
6827 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
6828 }
6829 
6830 
6831 /* ------------------------------------------------------------------------ */
6832 /* Function:    ipf_deltoken                                                */
6833 /* Returns:     int - 0 = success, else error                               */
6834 /* Parameters:  type(I) - the token type to match                           */
6835 /*              uid(I)  - uid owning the token                              */
6836 /*              ptr(I)  - context pointer for the token                     */
6837 /*              ifs - ipf stack instance                                    */
6838 /*                                                                          */
6839 /* This function looks for a a token in the current list that matches up    */
6840 /* the fields (type, uid, ptr).  If none is found, ESRCH is returned, else  */
6841 /* call ipf_freetoken() to remove it from the list.                         */
6842 /* ------------------------------------------------------------------------ */
6843 int ipf_deltoken(type, uid, ptr, ifs)
6844 int type, uid;
6845 void *ptr;
6846 ipf_stack_t *ifs;
6847 {
6848 	ipftoken_t *it;
6849 	int error = ESRCH;
6850 
6851 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
6852 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
6853 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
6854 		    uid == it->ipt_uid) {
6855 			ipf_freetoken(it, ifs);
6856 			error = 0;
6857 			break;
6858 	}
6859 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
6860 
6861 	return error;
6862 }
6863 
6864 
6865 /* ------------------------------------------------------------------------ */
6866 /* Function:    ipf_unlinktoken                                             */
6867 /* Returns:     None.                                                       */
6868 /* Parameters:  token(I) - pointer to token structure                       */
6869 /*              ifs - ipf stack instance                                    */
6870 /*                                                                          */
6871 /* This function unlinks a token structure from the linked list of tokens   */
6872 /* that it belongs to.  The head pointer never needs to be explicitly       */
6873 /* adjusted, but the tail does due to the linked list implementation.       */
6874 /* ------------------------------------------------------------------------ */
6875 static void ipf_unlinktoken(token, ifs)
6876 ipftoken_t *token;
6877 ipf_stack_t *ifs;
6878 {
6879 
6880 	if (ifs->ifs_ipftokentail == &token->ipt_next)
6881 		ifs->ifs_ipftokentail = token->ipt_pnext;
6882 
6883 	*token->ipt_pnext = token->ipt_next;
6884 	if (token->ipt_next != NULL)
6885 		token->ipt_next->ipt_pnext = token->ipt_pnext;
6886 }
6887 
6888 
6889 /* ------------------------------------------------------------------------ */
6890 /* Function:    ipf_findtoken                                               */
6891 /* Returns:     ipftoken_t * - NULL if no memory, else pointer to token     */
6892 /* Parameters:  type(I) - the token type to match                           */
6893 /*              uid(I) - uid owning the token                               */
6894 /*              ptr(I) - context pointer for the token                      */
6895 /*              ifs - ipf stack instance                                    */
6896 /*                                                                          */
6897 /* This function looks for a live token in the list of current tokens that  */
6898 /* matches the tuple (type, uid, ptr).  If one cannot be found then one is  */
6899 /* allocated.  If one is found then it is moved to the top of the list of   */
6900 /* currently active tokens.                                                 */
6901 /*                                                                          */
6902 /* NOTE: It is by design that this function returns holding a read lock on  */
6903 /*       ipf_tokens.  Callers must make sure they release it!               */
6904 /* ------------------------------------------------------------------------ */
6905 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
6906 int type, uid;
6907 void *ptr;
6908 ipf_stack_t *ifs;
6909 {
6910 	ipftoken_t *it, *new;
6911 
6912 	KMALLOC(new, ipftoken_t *);
6913 
6914 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
6915 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
6916 		if (it->ipt_alive == 0)
6917 			continue;
6918 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
6919 		    uid == it->ipt_uid)
6920 			break;
6921 	}
6922 
6923 	if (it == NULL) {
6924 		it = new;
6925 		new = NULL;
6926 		if (it == NULL)
6927 			return NULL;
6928 		it->ipt_data = NULL;
6929 		it->ipt_ctx = ptr;
6930 		it->ipt_uid = uid;
6931 		it->ipt_type = type;
6932 		it->ipt_next = NULL;
6933 		it->ipt_alive = 1;
6934 	} else {
6935 		if (new != NULL) {
6936 			KFREE(new);
6937 			new = NULL;
6938 		}
6939 
6940 		ipf_unlinktoken(it, ifs);
6941 	}
6942 	it->ipt_pnext = ifs->ifs_ipftokentail;
6943 	*ifs->ifs_ipftokentail = it;
6944 	ifs->ifs_ipftokentail = &it->ipt_next;
6945 	it->ipt_next = NULL;
6946 
6947 	it->ipt_die = ifs->ifs_fr_ticks + 2;
6948 
6949 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
6950 
6951 	return it;
6952 }
6953 
6954 
6955 /* ------------------------------------------------------------------------ */
6956 /* Function:    ipf_freetoken                                               */
6957 /* Returns:     None.                                                       */
6958 /* Parameters:  token(I) - pointer to token structure                       */
6959 /*              ifs - ipf stack instance                                    */
6960 /*                                                                          */
6961 /* This function unlinks a token from the linked list and on the path to    */
6962 /* free'ing the data, it calls the dereference function that is associated  */
6963 /* with the type of data pointed to by the token as it is considered to     */
6964 /* hold a reference to it.                                                  */
6965 /* ------------------------------------------------------------------------ */
6966 void ipf_freetoken(token, ifs)
6967 ipftoken_t *token;
6968 ipf_stack_t *ifs;
6969 {
6970 	void *data, **datap;
6971 
6972 	ipf_unlinktoken(token, ifs);
6973 
6974 	data = token->ipt_data;
6975 	datap = &data;
6976 
6977 	if ((data != NULL) && (data != (void *)-1)) {
6978 		switch (token->ipt_type)
6979 		{
6980 		case IPFGENITER_IPF :
6981 			(void)fr_derefrule((frentry_t **)datap, ifs);
6982 			break;
6983 		case IPFGENITER_IPNAT :
6984 			WRITE_ENTER(&ifs->ifs_ipf_nat);
6985 			fr_ipnatderef((ipnat_t **)datap, ifs);
6986 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
6987 			break;
6988 		case IPFGENITER_NAT :
6989 			fr_natderef((nat_t **)datap, ifs);
6990 			break;
6991 		case IPFGENITER_STATE :
6992 			fr_statederef((ipstate_t **)datap, ifs);
6993 			break;
6994 		case IPFGENITER_FRAG :
6995 			fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
6996 			break;
6997 		case IPFGENITER_NATFRAG :
6998  			fr_fragderef((ipfr_t **)datap,
6999 				     &ifs->ifs_ipf_natfrag, ifs);
7000 			break;
7001 		case IPFGENITER_HOSTMAP :
7002 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7003 			fr_hostmapdel((hostmap_t **)datap);
7004 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7005 			break;
7006 		default :
7007 			(void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7008 			break;
7009 		}
7010 	}
7011 
7012 	KFREE(token);
7013 }
7014 
7015 
7016 /* ------------------------------------------------------------------------ */
7017 /* Function:    ipf_getnextrule                                             */
7018 /* Returns:     int - 0 = success, else error                               */
7019 /* Parameters:  t(I)   - pointer to destination information to resolve      */
7020 /*              ptr(I) - pointer to ipfobj_t to copyin from user space      */
7021 /*              ifs - ipf stack instance                                    */
7022 /*                                                                          */
7023 /* This function's first job is to bring in the ipfruleiter_t structure via */
7024 /* the ipfobj_t structure to determine what should be the next rule to      */
7025 /* return. Once the ipfruleiter_t has been brought in, it then tries to     */
7026 /* find the 'next rule'.  This may include searching rule group lists or    */
7027 /* just be as simple as looking at the 'next' field in the rule structure.  */
7028 /* When we have found the rule to return, increase its reference count and  */
7029 /* if we used an existing rule to get here, decrease its reference count.   */
7030 /* ------------------------------------------------------------------------ */
7031 int ipf_getnextrule(t, ptr, ifs)
7032 ipftoken_t *t;
7033 void *ptr;
7034 ipf_stack_t *ifs;
7035 {
7036 	frentry_t *fr, *next, zero;
7037 	int error, out, count;
7038 	ipfruleiter_t it;
7039 	frgroup_t *fg;
7040 	char *dst;
7041 
7042 	if (t == NULL || ptr == NULL)
7043 		return EFAULT;
7044 	error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7045 	if (error != 0)
7046 		return error;
7047 	if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7048 		return EINVAL;
7049 	if ((it.iri_inout < 0) || (it.iri_inout > 3))
7050 		return EINVAL;
7051 	if (it.iri_nrules == 0)
7052 		return EINVAL;
7053 	if ((it.iri_active != 0) && (it.iri_active != 1))
7054 		return EINVAL;
7055 	if (it.iri_rule == NULL)
7056 		return EFAULT;
7057 
7058 	/*
7059 	 * Use bitmask on it.iri_inout to determine direction.
7060 	 * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7061 	 * F_IN (0) and F_ACIN (2) mask to out = 0.
7062 	 */
7063 	out = it.iri_inout & F_OUT;
7064 	READ_ENTER(&ifs->ifs_ipf_mutex);
7065 
7066 	/*
7067 	 * Retrieve "previous" entry from token and find the next entry.
7068 	 */
7069 	fr = t->ipt_data;
7070 	if (fr == NULL) {
7071 		if (*it.iri_group == '\0') {
7072 			/*
7073 			 * Use bitmask again to determine accounting or not.
7074 			 * F_ACIN will mask to accounting cases F_ACIN (2)
7075 			 * or F_ACOUT (3), but not F_IN or F_OUT.
7076 			 */
7077 			if ((it.iri_inout & F_ACIN) != 0) {
7078 				if (it.iri_ver == AF_INET)
7079 					next = ifs->ifs_ipacct
7080 					    [out][it.iri_active];
7081 				else
7082 					next = ifs->ifs_ipacct6
7083 					    [out][it.iri_active];
7084 			} else {
7085 				if (it.iri_ver == AF_INET)
7086 					next = ifs->ifs_ipfilter
7087 					    [out][it.iri_active];
7088 				else
7089 					next = ifs->ifs_ipfilter6
7090 					    [out][it.iri_active];
7091 			}
7092 		} else {
7093 			fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7094 					  it.iri_active, NULL, ifs);
7095 			if (fg != NULL)
7096 				next = fg->fg_start;
7097 			else
7098 				next = NULL;
7099 		}
7100 	} else {
7101 		next = fr->fr_next;
7102 	}
7103 
7104 	dst = (char *)it.iri_rule;
7105 	/*
7106 	 * The ipfruleiter may ask for more than 1 rule at a time to be
7107 	 * copied out, so long as that many exist in the list to start with!
7108 	 */
7109 	for (count = it.iri_nrules; count > 0; count--) {
7110 		/*
7111 		 * If we found an entry, add reference to it and update token.
7112 		 * Otherwise, zero out data to be returned and NULL out token.
7113 		 */
7114 		if (next != NULL) {
7115 			MUTEX_ENTER(&next->fr_lock);
7116 			next->fr_ref++;
7117 			MUTEX_EXIT(&next->fr_lock);
7118 			t->ipt_data = next;
7119 		} else {
7120 			bzero(&zero, sizeof(zero));
7121 			next = &zero;
7122 			t->ipt_data = NULL;
7123 		}
7124 
7125 		/*
7126 		 * Now that we have ref, it's save to give up lock.
7127 		 */
7128 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7129 
7130 		/*
7131 		 * Copy out data and clean up references and token as needed.
7132 		 */
7133 		error = COPYOUT(next, dst, sizeof(*next));
7134 		if (error != 0)
7135 			error = EFAULT;
7136 		if (t->ipt_data == NULL) {
7137 			ipf_freetoken(t, ifs);
7138 			break;
7139 		} else {
7140 			if (fr != NULL)
7141 				(void) fr_derefrule(&fr, ifs);
7142 			if (next->fr_data != NULL) {
7143 				dst += sizeof(*next);
7144 				error = COPYOUT(next->fr_data, dst,
7145 						next->fr_dsize);
7146 				if (error != 0)
7147 					error = EFAULT;
7148 				else
7149 					dst += next->fr_dsize;
7150 			}
7151 			if (next->fr_next == NULL) {
7152 				ipf_freetoken(t, ifs);
7153 				break;
7154 			}
7155 		}
7156 
7157 		if ((count == 1) || (error != 0))
7158 			break;
7159 
7160 		READ_ENTER(&ifs->ifs_ipf_mutex);
7161 		fr = next;
7162 		next = fr->fr_next;
7163 	}
7164 
7165 	return error;
7166 }
7167 
7168 
7169 /* ------------------------------------------------------------------------ */
7170 /* Function:    fr_frruleiter                                               */
7171 /* Returns:     int - 0 = success, else error                               */
7172 /* Parameters:  data(I) - the token type to match                           */
7173 /*              uid(I) - uid owning the token                               */
7174 /*              ptr(I) - context pointer for the token                      */
7175 /*              ifs - ipf stack instance                                    */
7176 /*                                                                          */
7177 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7178 /* ipf_getnextrule.  It's role is to find the right token in the kernel for */
7179 /* the process doing the ioctl and use that to ask for the next rule.       */
7180 /* ------------------------------------------------------------------------ */
7181 int ipf_frruleiter(data, uid, ctx, ifs)
7182 void *data, *ctx;
7183 int uid;
7184 ipf_stack_t *ifs;
7185 {
7186 	ipftoken_t *token;
7187 	int error;
7188 
7189 	token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7190 	if (token != NULL)
7191 		error = ipf_getnextrule(token, data, ifs);
7192 	else
7193 		error = EFAULT;
7194 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7195 
7196 	return error;
7197 }
7198 
7199 
7200 /* ------------------------------------------------------------------------ */
7201 /* Function:    ipf_geniter                                                 */
7202 /* Returns:     int - 0 = success, else error                               */
7203 /* Parameters:  token(I) - pointer to ipftoken structure                    */
7204 /*              itp(I) - pointer to ipfgeniter structure                    */
7205 /*              ifs - ipf stack instance                                    */
7206 /*                                                                          */
7207 /* Generic iterator called from ipf_genericiter.  Currently only used for   */
7208 /* walking through list of fragments.                                       */
7209 /* ------------------------------------------------------------------------ */
7210 int ipf_geniter(token, itp, ifs)
7211 ipftoken_t *token;
7212 ipfgeniter_t *itp;
7213 ipf_stack_t *ifs;
7214 {
7215 	int error;
7216 
7217 	switch (itp->igi_type)
7218 	{
7219 	case IPFGENITER_FRAG :
7220 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7221 				    &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7222 				    ifs);
7223 		break;
7224 	default :
7225 		error = EINVAL;
7226 		break;
7227 	}
7228 
7229 	return error;
7230 }
7231 
7232 
7233 /* ------------------------------------------------------------------------ */
7234 /* Function:    ipf_genericiter                                             */
7235 /* Returns:     int - 0 = success, else error                               */
7236 /* Parameters:  data(I) - the token type to match                           */
7237 /*              uid(I) - uid owning the token                               */
7238 /*              ptr(I) - context pointer for the token                      */
7239 /*              ifs - ipf stack instance                                    */
7240 /*                                                                          */
7241 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7242 /* ipf_geniter when handling SIOCGENITER.  It's role is to find the right   */
7243 /* token in the kernel for the process using the ioctl, and to use that     */
7244 /* token when calling ipf_geniter.                                          */
7245 /* ------------------------------------------------------------------------ */
7246 int ipf_genericiter(data, uid, ctx, ifs)
7247 void *data, *ctx;
7248 int uid;
7249 ipf_stack_t *ifs;
7250 {
7251 	ipftoken_t *token;
7252 	ipfgeniter_t iter;
7253 	int error;
7254 
7255 	error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7256 	if (error != 0)
7257 		return error;
7258 
7259 	token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7260 	if (token != NULL) {
7261 		token->ipt_subtype = iter.igi_type;
7262 		error = ipf_geniter(token, &iter, ifs);
7263 	} else
7264 		error = EFAULT;
7265 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7266 
7267 	return error;
7268 }
7269 
7270 
7271 /* --------------------------------------------------------------------- */
7272 /* Function:    ipf_earlydrop                                            */
7273 /* Returns:     number of dropped/removed entries from the queue         */
7274 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7275 /*              ifq	- pointer to queue with entries to be deleted    */
7276 /*              idletime - entry must be idle this long to be deleted    */
7277 /*              ifs     - ipf stack instance                             */
7278 /*                                                                       */
7279 /* Function is invoked from state/NAT flush routines to remove entries   */
7280 /* from specified timeout queue, based on how long they've sat idle,     */
7281 /* without waiting for it to happen on its own.                          */
7282 /* --------------------------------------------------------------------- */
7283 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7284 int flushtype;
7285 ipftq_t *ifq;
7286 int idletime;
7287 ipf_stack_t *ifs;
7288 {
7289         ipftqent_t *tqe, *tqn;
7290         unsigned int dropped;
7291         int droptick;
7292 	void *ent;
7293 
7294         if (ifq == NULL)
7295                 return (0);
7296 
7297         dropped = 0;
7298 
7299         /*
7300          * Determine the tick representing the idle time we're interested
7301          * in.  If an entry exists in the queue, and it was touched before
7302          * that tick, then it's been idle longer than idletime, so it should
7303 	 * be deleted.
7304          */
7305         droptick = ifs->ifs_fr_ticks - idletime;
7306         tqn = ifq->ifq_head;
7307         while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7308                 tqn = tqe->tqe_next;
7309 		ent = tqe->tqe_parent;
7310 		switch (flushtype)
7311 		{
7312 		case NAT_FLUSH:
7313 			if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7314 				dropped++;
7315 			break;
7316 		case STATE_FLUSH:
7317 			if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7318 				dropped++;
7319 			break;
7320 		default:
7321 			return (0);
7322 		}
7323         }
7324         return (dropped);
7325 }
7326 
7327 
7328 /* --------------------------------------------------------------------- */
7329 /* Function:    ipf_flushclosing                                         */
7330 /* Returns:     int - number of entries deleted                          */
7331 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7332 /*              stateval - TCP state at which to start removing entries  */
7333 /*              ipfqs - pointer to timeout queues                        */
7334 /*              userqs - pointer to user defined queues                  */
7335 /*              ifs  - ipf stack instance                                */
7336 /*                                                                       */
7337 /* Remove state/NAT table entries for TCP connections which are in the   */
7338 /* process of closing, and have at least reached the state specified by  */
7339 /* the 'stateval' parameter.                                             */
7340 /* --------------------------------------------------------------------- */
7341 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7342 int flushtype, stateval;
7343 ipftq_t *ipfqs, *userqs;
7344 ipf_stack_t *ifs;
7345 {
7346 	ipftq_t *ifq, *ifqn;
7347         ipftqent_t *tqe, *tqn;
7348         int dropped;
7349 	void *ent;
7350 	nat_t *nat;
7351 	ipstate_t *is;
7352 
7353         dropped = 0;
7354 
7355         /*
7356          * Start by deleting any entries in specific timeout queues.
7357          */
7358 	ifqn = &ipfqs[stateval];
7359         while ((ifq = ifqn) != NULL) {
7360                 ifqn = ifq->ifq_next;
7361                 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7362         }
7363 
7364         /*
7365          * Next, look through user defined queues for closing entries.
7366          */
7367 	ifqn = userqs;
7368         while ((ifq = ifqn) != NULL) {
7369                 ifqn = ifq->ifq_next;
7370                 tqn = ifq->ifq_head;
7371                 while ((tqe = tqn) != NULL) {
7372                         tqn = tqe->tqe_next;
7373 			ent = tqe->tqe_parent;
7374 			switch (flushtype)
7375 			{
7376 			case NAT_FLUSH:
7377 				nat = (nat_t *)ent;
7378 				if ((nat->nat_p == IPPROTO_TCP) &&
7379 				    (nat->nat_tcpstate[0] >= stateval) &&
7380 				    (nat->nat_tcpstate[1] >= stateval) &&
7381 				    (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7382 					dropped++;
7383 				break;
7384 			case STATE_FLUSH:
7385 				is = (ipstate_t *)ent;
7386 				if ((is->is_p == IPPROTO_TCP) &&
7387 				    (is->is_state[0] >= stateval) &&
7388 				    (is->is_state[1] >= stateval) &&
7389 				    (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7390 					dropped++;
7391 				break;
7392 			default:
7393 				return (0);
7394 			}
7395                 }
7396         }
7397         return (dropped);
7398 }
7399 
7400 
7401 /* --------------------------------------------------------------------- */
7402 /* Function:    ipf_extraflush                                           */
7403 /* Returns:     int - number of entries flushed (0 = none)               */
7404 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7405 /*              ipfqs - pointer to 'established' timeout queue           */
7406 /*              userqs - pointer to user defined queues                  */
7407 /*              ifs  - ipf stack instance                                */
7408 /*                                                                       */
7409 /* This function gets called when either NAT or state tables fill up.    */
7410 /* We need to try a bit harder to free up some space.  The function will */
7411 /* flush entries for TCP connections which have been idle a long time.   */
7412 /*                                                                       */
7413 /* Currently, the idle time is checked using values from ideltime_tab[]	 */
7414 /* --------------------------------------------------------------------- */
7415 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7416 int flushtype;
7417 ipftq_t *ipfqs, *userqs;
7418 ipf_stack_t *ifs;
7419 {
7420 	ipftq_t *ifq, *ifqn;
7421 	int idletime, removed, idle_idx;
7422 
7423 	removed = 0;
7424 
7425 	/*
7426   	 * Determine initial threshold for minimum idle time based on
7427 	 * how long ipfilter has been running.  Ipfilter needs to have
7428 	 * been up as long as the smallest interval to continue on.
7429 	 *
7430 	 * Minimum idle times stored in idletime_tab and indexed by
7431 	 * idle_idx.  Start at upper end of array and work backwards.
7432 	 *
7433 	 * Once the index is found, set the initial idle time to the
7434 	 * first interval before the current ipfilter run time.
7435 	 */
7436 	if (ifs->ifs_fr_ticks < idletime_tab[0])
7437 		return (0);
7438 	idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7439 	if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7440 		idletime = idletime_tab[idle_idx];
7441 	} else {
7442 		while ((idle_idx > 0) &&
7443 		    (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7444 			idle_idx--;
7445 
7446 		idletime = (ifs->ifs_fr_ticks /
7447 			    idletime_tab[idle_idx]) *
7448 			    idletime_tab[idle_idx];
7449 	}
7450 
7451 	while (idle_idx >= 0) {
7452 		/*
7453 		 * Check to see if we need to delete more entries.
7454 		 * If we do, start with appropriate timeout queue.
7455 		 */
7456 		if (flushtype == NAT_FLUSH) {
7457 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7458 			    ifs->ifs_nat_flush_level_lo)
7459 				break;
7460 		} else if (flushtype == STATE_FLUSH) {
7461 			if (ST_TAB_WATER_LEVEL(ifs) <=
7462 			    ifs->ifs_state_flush_level_lo)
7463 				break;
7464 		} else {
7465 			break;
7466 		}
7467 
7468 		removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7469 
7470 		/*
7471 		 * Next, check the user defined queues.  But first, make
7472 		 * certain that timeout queue deletions didn't do enough.
7473 		 */
7474 		if (flushtype == NAT_FLUSH) {
7475 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7476 			    ifs->ifs_nat_flush_level_lo)
7477 				break;
7478 		} else {
7479 			if (ST_TAB_WATER_LEVEL(ifs) <=
7480 			    ifs->ifs_state_flush_level_lo)
7481 				break;
7482 		}
7483 		ifqn = userqs;
7484 		while ((ifq = ifqn) != NULL) {
7485 			ifqn = ifq->ifq_next;
7486 			removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7487 		}
7488 
7489 		/*
7490 		 * Adjust the granularity of idle time.
7491 		 *
7492 		 * If we reach an interval boundary, we need to
7493 		 * either adjust the idle time accordingly or exit
7494 		 * the loop altogether (if this is very last check).
7495 		 */
7496 		idletime -= idletime_tab[idle_idx];
7497 		if (idletime < idletime_tab[idle_idx]) {
7498 			if (idle_idx != 0) {
7499 				idletime = idletime_tab[idle_idx] -
7500 				idletime_tab[idle_idx - 1];
7501 				idle_idx--;
7502 			} else {
7503 				break;
7504 			}
7505 		}
7506 	}
7507 
7508 	return (removed);
7509 }
7510