xref: /titanic_41/usr/src/uts/common/inet/ipf/fil.c (revision c4a60461173b2401c7ffb8916e31ecc2cccf65c4)
1 /*
2  * Copyright (C) 1993-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
7  *
8  * Copyright (c) 2014, Joyent, Inc.  All rights reserved.
9  */
10 
11 #if defined(KERNEL) || defined(_KERNEL)
12 # undef KERNEL
13 # undef _KERNEL
14 # define        KERNEL	1
15 # define        _KERNEL	1
16 #endif
17 #include <sys/errno.h>
18 #include <sys/types.h>
19 #include <sys/param.h>
20 #include <sys/time.h>
21 #if defined(__NetBSD__)
22 # if (NetBSD >= 199905) && !defined(IPFILTER_LKM) && defined(_KERNEL)
23 #  include "opt_ipfilter_log.h"
24 # endif
25 #endif
26 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
27     (__FreeBSD_version >= 220000)
28 # if (__FreeBSD_version >= 400000)
29 #  if !defined(IPFILTER_LKM)
30 #   include "opt_inet6.h"
31 #  endif
32 #  if (__FreeBSD_version == 400019)
33 #   define CSUM_DELAY_DATA
34 #  endif
35 # endif
36 # include <sys/filio.h>
37 #else
38 # include <sys/ioctl.h>
39 #endif
40 #if !defined(_AIX51)
41 # include <sys/fcntl.h>
42 #endif
43 #if defined(_KERNEL)
44 # include <sys/systm.h>
45 # include <sys/file.h>
46 #else
47 # include <stdio.h>
48 # include <string.h>
49 # include <stdlib.h>
50 # include <stddef.h>
51 # include <sys/file.h>
52 # define _KERNEL
53 # ifdef __OpenBSD__
54 struct file;
55 # endif
56 # include <sys/uio.h>
57 # undef _KERNEL
58 #endif
59 #if !defined(__SVR4) && !defined(__svr4__) && !defined(__hpux) && \
60     !defined(linux)
61 # include <sys/mbuf.h>
62 #else
63 # if !defined(linux)
64 #  include <sys/byteorder.h>
65 # endif
66 # if (SOLARIS2 < 5) && defined(sun)
67 #  include <sys/dditypes.h>
68 # endif
69 #endif
70 #ifdef __hpux
71 # define _NET_ROUTE_INCLUDED
72 #endif
73 #if !defined(linux)
74 # include <sys/protosw.h>
75 #endif
76 #include <sys/socket.h>
77 #include <net/if.h>
78 #ifdef sun
79 # include <net/af.h>
80 #endif
81 #if !defined(_KERNEL) && defined(__FreeBSD__)
82 # include "radix_ipf.h"
83 #endif
84 #include <net/route.h>
85 #include <netinet/in.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/ip.h>
88 #if !defined(linux)
89 # include <netinet/ip_var.h>
90 #endif
91 #if defined(__sgi) && defined(IFF_DRVRLOCK) /* IRIX 6 */
92 # include <sys/hashing.h>
93 # include <netinet/in_var.h>
94 #endif
95 #include <netinet/tcp.h>
96 #if (!defined(__sgi) && !defined(AIX)) || defined(_KERNEL)
97 # include <netinet/udp.h>
98 # include <netinet/ip_icmp.h>
99 #endif
100 #ifdef __hpux
101 # undef _NET_ROUTE_INCLUDED
102 #endif
103 #include "netinet/ip_compat.h"
104 #ifdef	USE_INET6
105 # include <netinet/icmp6.h>
106 # if !SOLARIS && defined(_KERNEL) && !defined(__osf__) && !defined(__hpux)
107 #  include <netinet6/in6_var.h>
108 # endif
109 #endif
110 #include <netinet/tcpip.h>
111 #include "netinet/ip_fil.h"
112 #include "netinet/ip_nat.h"
113 #include "netinet/ip_frag.h"
114 #include "netinet/ip_state.h"
115 #include "netinet/ip_proxy.h"
116 #include "netinet/ip_auth.h"
117 #include "netinet/ipf_stack.h"
118 #ifdef IPFILTER_SCAN
119 # include "netinet/ip_scan.h"
120 #endif
121 #ifdef IPFILTER_SYNC
122 # include "netinet/ip_sync.h"
123 #endif
124 #include "netinet/ip_pool.h"
125 #include "netinet/ip_htable.h"
126 #ifdef IPFILTER_COMPILED
127 # include "netinet/ip_rules.h"
128 #endif
129 #if defined(IPFILTER_BPF) && defined(_KERNEL)
130 # include <net/bpf.h>
131 #endif
132 #if defined(__FreeBSD_version) && (__FreeBSD_version >= 300000)
133 # include <sys/malloc.h>
134 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
135 #  include "opt_ipfilter.h"
136 # endif
137 #endif
138 #include "netinet/ipl.h"
139 #if defined(_KERNEL)
140 #include <sys/sunddi.h>
141 #endif
142 /* END OF INCLUDES */
143 
144 #if !defined(lint)
145 static const char sccsid[] = "@(#)fil.c	1.36 6/5/96 (C) 1993-2000 Darren Reed";
146 static const char rcsid[] = "@(#)$Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $";
147 #endif
148 
149 #ifndef	_KERNEL
150 # include "ipf.h"
151 # include "ipt.h"
152 # include "bpf-ipf.h"
153 extern	int	opts;
154 
155 # define	FR_VERBOSE(verb_pr)			verbose verb_pr
156 # define	FR_DEBUG(verb_pr)			debug verb_pr
157 #else /* #ifndef _KERNEL */
158 # define	FR_VERBOSE(verb_pr)
159 # define	FR_DEBUG(verb_pr)
160 #endif /* _KERNEL */
161 
162 
163 char	ipfilter_version[] = IPL_VERSION;
164 int	fr_features = 0
165 #ifdef	IPFILTER_LKM
166 		| IPF_FEAT_LKM
167 #endif
168 #ifdef	IPFILTER_LOG
169 		| IPF_FEAT_LOG
170 #endif
171 #ifdef	IPFILTER_LOOKUP
172 		| IPF_FEAT_LOOKUP
173 #endif
174 #ifdef	IPFILTER_BPF
175 		| IPF_FEAT_BPF
176 #endif
177 #ifdef	IPFILTER_COMPILED
178 		| IPF_FEAT_COMPILED
179 #endif
180 #ifdef	IPFILTER_CKSUM
181 		| IPF_FEAT_CKSUM
182 #endif
183 #ifdef	IPFILTER_SYNC
184 		| IPF_FEAT_SYNC
185 #endif
186 #ifdef	IPFILTER_SCAN
187 		| IPF_FEAT_SCAN
188 #endif
189 #ifdef	USE_INET6
190 		| IPF_FEAT_IPV6
191 #endif
192 	;
193 
194 #define	IPF_BUMP(x)	(x)++
195 
196 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
197 static	INLINE int	fr_ipfcheck __P((fr_info_t *, frentry_t *, int));
198 static	int		fr_portcheck __P((frpcmp_t *, u_short *));
199 static	int		frflushlist __P((int, minor_t, int *, frentry_t **,
200 					 ipf_stack_t *));
201 static	ipfunc_t	fr_findfunc __P((ipfunc_t));
202 static	frentry_t	*fr_firewall __P((fr_info_t *, u_32_t *));
203 static	int		fr_funcinit __P((frentry_t *fr, ipf_stack_t *));
204 static	INLINE void	frpr_ah __P((fr_info_t *));
205 static	INLINE void	frpr_esp __P((fr_info_t *));
206 static	INLINE void	frpr_gre __P((fr_info_t *));
207 static	INLINE void	frpr_udp __P((fr_info_t *));
208 static	INLINE void	frpr_tcp __P((fr_info_t *));
209 static	INLINE void	frpr_icmp __P((fr_info_t *));
210 static	INLINE void	frpr_ipv4hdr __P((fr_info_t *));
211 static	INLINE int	frpr_pullup __P((fr_info_t *, int));
212 static	INLINE void	frpr_short __P((fr_info_t *, int));
213 static	INLINE void	frpr_tcpcommon __P((fr_info_t *));
214 static	INLINE void	frpr_udpcommon __P((fr_info_t *));
215 static	INLINE int	fr_updateipid __P((fr_info_t *));
216 #ifdef	IPFILTER_LOOKUP
217 static	int		fr_grpmapinit __P((frentry_t *fr, ipf_stack_t *));
218 static	INLINE void	*fr_resolvelookup __P((u_int, u_int, lookupfunc_t *,
219 					       ipf_stack_t *));
220 #endif
221 static	void		frsynclist __P((int, int, void *, char *, frentry_t *,
222     ipf_stack_t *));
223 static	void		*fr_ifsync __P((int, int, char *, char *,
224 					void *, void *, ipf_stack_t *));
225 static	ipftuneable_t	*fr_findtunebyname __P((const char *, ipf_stack_t *));
226 static	ipftuneable_t	*fr_findtunebycookie __P((void *, void **, ipf_stack_t *));
227 
228 /*
229  * bit values for identifying presence of individual IP options
230  * All of these tables should be ordered by increasing key value on the left
231  * hand side to allow for binary searching of the array and include a trailer
232  * with a 0 for the bitmask for linear searches to easily find the end with.
233  */
234 const	struct	optlist	ipopts[20] = {
235 	{ IPOPT_NOP,	0x000001 },
236 	{ IPOPT_RR,	0x000002 },
237 	{ IPOPT_ZSU,	0x000004 },
238 	{ IPOPT_MTUP,	0x000008 },
239 	{ IPOPT_MTUR,	0x000010 },
240 	{ IPOPT_ENCODE,	0x000020 },
241 	{ IPOPT_TS,	0x000040 },
242 	{ IPOPT_TR,	0x000080 },
243 	{ IPOPT_SECURITY, 0x000100 },
244 	{ IPOPT_LSRR,	0x000200 },
245 	{ IPOPT_E_SEC,	0x000400 },
246 	{ IPOPT_CIPSO,	0x000800 },
247 	{ IPOPT_SATID,	0x001000 },
248 	{ IPOPT_SSRR,	0x002000 },
249 	{ IPOPT_ADDEXT,	0x004000 },
250 	{ IPOPT_VISA,	0x008000 },
251 	{ IPOPT_IMITD,	0x010000 },
252 	{ IPOPT_EIP,	0x020000 },
253 	{ IPOPT_FINN,	0x040000 },
254 	{ 0,		0x000000 }
255 };
256 
257 #ifdef USE_INET6
258 struct optlist ip6exthdr[] = {
259 	{ IPPROTO_HOPOPTS,		0x000001 },
260 	{ IPPROTO_IPV6,			0x000002 },
261 	{ IPPROTO_ROUTING,		0x000004 },
262 	{ IPPROTO_FRAGMENT,		0x000008 },
263 	{ IPPROTO_ESP,			0x000010 },
264 	{ IPPROTO_AH,			0x000020 },
265 	{ IPPROTO_NONE,			0x000040 },
266 	{ IPPROTO_DSTOPTS,		0x000080 },
267 	{ 0,				0 }
268 };
269 #endif
270 
271 struct optlist tcpopts[] = {
272 	{ TCPOPT_NOP,			0x000001 },
273 	{ TCPOPT_MAXSEG,		0x000002 },
274 	{ TCPOPT_WINDOW,		0x000004 },
275 	{ TCPOPT_SACK_PERMITTED,	0x000008 },
276 	{ TCPOPT_SACK,			0x000010 },
277 	{ TCPOPT_TIMESTAMP,		0x000020 },
278 	{ 0,				0x000000 }
279 };
280 
281 /*
282  * bit values for identifying presence of individual IP security options
283  */
284 const	struct	optlist	secopt[8] = {
285 	{ IPSO_CLASS_RES4,	0x01 },
286 	{ IPSO_CLASS_TOPS,	0x02 },
287 	{ IPSO_CLASS_SECR,	0x04 },
288 	{ IPSO_CLASS_RES3,	0x08 },
289 	{ IPSO_CLASS_CONF,	0x10 },
290 	{ IPSO_CLASS_UNCL,	0x20 },
291 	{ IPSO_CLASS_RES2,	0x40 },
292 	{ IPSO_CLASS_RES1,	0x80 }
293 };
294 
295 
296 /*
297  * Table of functions available for use with call rules.
298  */
299 static ipfunc_resolve_t fr_availfuncs[] = {
300 #ifdef	IPFILTER_LOOKUP
301 	{ "fr_srcgrpmap", fr_srcgrpmap, fr_grpmapinit },
302 	{ "fr_dstgrpmap", fr_dstgrpmap, fr_grpmapinit },
303 #endif
304 	{ "", NULL }
305 };
306 
307 
308 /*
309  * Below we declare a list of constants used only by the ipf_extraflush()
310  * routine.  We are placing it here, instead of in ipf_extraflush() itself,
311  * because we want to make it visible to tools such as mdb, nm etc., so the
312  * values can easily be altered during debugging.
313  */
314 static	const	int	idletime_tab[] = {
315 	IPF_TTLVAL(30),		/* 30 seconds */
316 	IPF_TTLVAL(1800),	/* 30 minutes */
317 	IPF_TTLVAL(43200),	/* 12 hours */
318 	IPF_TTLVAL(345600),	/* 4 days */
319 };
320 
321 
322 /*
323  * The next section of code is a a collection of small routines that set
324  * fields in the fr_info_t structure passed based on properties of the
325  * current packet.  There are different routines for the same protocol
326  * for each of IPv4 and IPv6.  Adding a new protocol, for which there
327  * will "special" inspection for setup, is now more easily done by adding
328  * a new routine and expanding the frpr_ipinit*() function rather than by
329  * adding more code to a growing switch statement.
330  */
331 #ifdef USE_INET6
332 static	INLINE int	frpr_ah6 __P((fr_info_t *));
333 static	INLINE void	frpr_esp6 __P((fr_info_t *));
334 static	INLINE void	frpr_gre6 __P((fr_info_t *));
335 static	INLINE void	frpr_udp6 __P((fr_info_t *));
336 static	INLINE void	frpr_tcp6 __P((fr_info_t *));
337 static	INLINE void	frpr_icmp6 __P((fr_info_t *));
338 static	INLINE void	frpr_ipv6hdr __P((fr_info_t *));
339 static	INLINE void	frpr_short6 __P((fr_info_t *, int));
340 static	INLINE int	frpr_hopopts6 __P((fr_info_t *));
341 static	INLINE int	frpr_routing6 __P((fr_info_t *));
342 static	INLINE int	frpr_dstopts6 __P((fr_info_t *));
343 static	INLINE int	frpr_fragment6 __P((fr_info_t *));
344 static	INLINE int	frpr_ipv6exthdr __P((fr_info_t *, int, int));
345 
346 
347 /* ------------------------------------------------------------------------ */
348 /* Function:    frpr_short6                                                 */
349 /* Returns:     void                                                        */
350 /* Parameters:  fin(I) - pointer to packet information                      */
351 /*                                                                          */
352 /* IPv6 Only                                                                */
353 /* This is function enforces the 'is a packet too short to be legit' rule   */
354 /* for IPv6 and marks the packet with FI_SHORT if so.  See function comment */
355 /* for frpr_short() for more details.                                       */
356 /* ------------------------------------------------------------------------ */
frpr_short6(fin,xmin)357 static INLINE void frpr_short6(fin, xmin)
358 fr_info_t *fin;
359 int xmin;
360 {
361 
362 	if (fin->fin_dlen < xmin)
363 		fin->fin_flx |= FI_SHORT;
364 }
365 
366 
367 /* ------------------------------------------------------------------------ */
368 /* Function:    frpr_ipv6hdr                                                */
369 /* Returns:     Nil                                                         */
370 /* Parameters:  fin(I) - pointer to packet information                      */
371 /*                                                                          */
372 /* IPv6 Only                                                                */
373 /* Copy values from the IPv6 header into the fr_info_t struct and call the  */
374 /* per-protocol analyzer if it exists.                                      */
375 /* ------------------------------------------------------------------------ */
frpr_ipv6hdr(fin)376 static INLINE void frpr_ipv6hdr(fin)
377 fr_info_t *fin;
378 {
379 	ip6_t *ip6 = (ip6_t *)fin->fin_ip;
380 	int p, go = 1, i, hdrcount;
381 	fr_ip_t *fi = &fin->fin_fi;
382 
383 	fin->fin_off = 0;
384 
385 	fi->fi_tos = 0;
386 	fi->fi_optmsk = 0;
387 	fi->fi_secmsk = 0;
388 	fi->fi_auth = 0;
389 
390 	p = ip6->ip6_nxt;
391 	fi->fi_ttl = ip6->ip6_hlim;
392 	fi->fi_src.in6 = ip6->ip6_src;
393 	fi->fi_dst.in6 = ip6->ip6_dst;
394 	fin->fin_id = 0;
395 
396 	hdrcount = 0;
397 	while (go && !(fin->fin_flx & (FI_BAD|FI_SHORT))) {
398 		switch (p)
399 		{
400 		case IPPROTO_UDP :
401 			frpr_udp6(fin);
402 			go = 0;
403 			break;
404 
405 		case IPPROTO_TCP :
406 			frpr_tcp6(fin);
407 			go = 0;
408 			break;
409 
410 		case IPPROTO_ICMPV6 :
411 			frpr_icmp6(fin);
412 			go = 0;
413 			break;
414 
415 		case IPPROTO_GRE :
416 			frpr_gre6(fin);
417 			go = 0;
418 			break;
419 
420 		case IPPROTO_HOPOPTS :
421 			/*
422 			 * hop by hop ext header is only allowed
423 			 * right after IPv6 header.
424 			 */
425 			if (hdrcount != 0) {
426 				fin->fin_flx |= FI_BAD;
427 				p = IPPROTO_NONE;
428 			} else {
429 				p = frpr_hopopts6(fin);
430 			}
431 			break;
432 
433 		case IPPROTO_DSTOPTS :
434 			p = frpr_dstopts6(fin);
435 			break;
436 
437 		case IPPROTO_ROUTING :
438 			p = frpr_routing6(fin);
439 			break;
440 
441 		case IPPROTO_AH :
442 			p = frpr_ah6(fin);
443 			break;
444 
445 		case IPPROTO_ESP :
446 			frpr_esp6(fin);
447 			go = 0;
448 			break;
449 
450 		case IPPROTO_IPV6 :
451 			for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
452 				if (ip6exthdr[i].ol_val == p) {
453 					fin->fin_flx |= ip6exthdr[i].ol_bit;
454 					break;
455 				}
456 			go = 0;
457 			break;
458 
459 		case IPPROTO_NONE :
460 			go = 0;
461 			break;
462 
463 		case IPPROTO_FRAGMENT :
464 			p = frpr_fragment6(fin);
465 			if (fin->fin_off != 0)  /* Not the first frag */
466 				go = 0;
467 			break;
468 
469 		default :
470 			go = 0;
471 			break;
472 		}
473 		hdrcount++;
474 
475 		/*
476 		 * It is important to note that at this point, for the
477 		 * extension headers (go != 0), the entire header may not have
478 		 * been pulled up when the code gets to this point.  This is
479 		 * only done for "go != 0" because the other header handlers
480 		 * will all pullup their complete header.  The other indicator
481 		 * of an incomplete packet is that this was just an extension
482 		 * header.
483 		 */
484 		if ((go != 0) && (p != IPPROTO_NONE) &&
485 		    (frpr_pullup(fin, 0) == -1)) {
486 			p = IPPROTO_NONE;
487 			go = 0;
488 		}
489 	}
490 	fi->fi_p = p;
491 }
492 
493 
494 /* ------------------------------------------------------------------------ */
495 /* Function:    frpr_ipv6exthdr                                             */
496 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
497 /* Parameters:  fin(I)      - pointer to packet information                 */
498 /*              multiple(I) - flag indicating yes/no if multiple occurances */
499 /*                            of this extension header are allowed.         */
500 /*              proto(I)    - protocol number for this extension header     */
501 /*                                                                          */
502 /* IPv6 Only                                                                */
503 /* This function expects to find an IPv6 extension header at fin_dp.        */
504 /* There must be at least 8 bytes of data at fin_dp for there to be a valid */
505 /* extension header present. If a good one is found, fin_dp is advanced to  */
506 /* point at the first piece of data after the extension header, fin_exthdr  */
507 /* points to the start of the extension header and the "protocol" of the    */
508 /* *NEXT* header is returned.                                               */
509 /* ------------------------------------------------------------------------ */
frpr_ipv6exthdr(fin,multiple,proto)510 static INLINE int frpr_ipv6exthdr(fin, multiple, proto)
511 fr_info_t *fin;
512 int multiple, proto;
513 {
514 	struct ip6_ext *hdr;
515 	u_short shift;
516 	int i;
517 
518 	fin->fin_flx |= FI_V6EXTHDR;
519 
520 				/* 8 is default length of extension hdr */
521 	if ((fin->fin_dlen - 8) < 0) {
522 		fin->fin_flx |= FI_SHORT;
523 		return IPPROTO_NONE;
524 	}
525 
526 	if (frpr_pullup(fin, 8) == -1)
527 		return IPPROTO_NONE;
528 
529 	hdr = fin->fin_dp;
530 	shift = 8 + (hdr->ip6e_len << 3);
531 	if (shift > fin->fin_dlen) {	/* Nasty extension header length? */
532 		fin->fin_flx |= FI_BAD;
533 		return IPPROTO_NONE;
534 	}
535 
536 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
537 		if (ip6exthdr[i].ol_val == proto) {
538 			/*
539 			 * Most IPv6 extension headers are only allowed once.
540 			 */
541 			if ((multiple == 0) &&
542 			    ((fin->fin_optmsk & ip6exthdr[i].ol_bit) != 0))
543 				fin->fin_flx |= FI_BAD;
544 			else
545 				fin->fin_optmsk |= ip6exthdr[i].ol_bit;
546 			break;
547 		}
548 
549 	fin->fin_dp = (char *)fin->fin_dp + shift;
550 	fin->fin_dlen -= shift;
551 
552 	return hdr->ip6e_nxt;
553 }
554 
555 
556 /* ------------------------------------------------------------------------ */
557 /* Function:    frpr_hopopts6                                               */
558 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
559 /* Parameters:  fin(I) - pointer to packet information                      */
560 /*                                                                          */
561 /* IPv6 Only                                                                */
562 /* This is function checks pending hop by hop options extension header      */
563 /* ------------------------------------------------------------------------ */
frpr_hopopts6(fin)564 static INLINE int frpr_hopopts6(fin)
565 fr_info_t *fin;
566 {
567 	return frpr_ipv6exthdr(fin, 0, IPPROTO_HOPOPTS);
568 }
569 
570 
571 /* ------------------------------------------------------------------------ */
572 /* Function:    frpr_routing6                                               */
573 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
574 /* Parameters:  fin(I) - pointer to packet information                      */
575 /*                                                                          */
576 /* IPv6 Only                                                                */
577 /* This is function checks pending routing extension header                 */
578 /* ------------------------------------------------------------------------ */
frpr_routing6(fin)579 static INLINE int frpr_routing6(fin)
580 fr_info_t *fin;
581 {
582 	struct ip6_ext *hdr;
583 	int shift;
584 
585 	hdr = fin->fin_dp;
586 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_ROUTING) == IPPROTO_NONE)
587 		return IPPROTO_NONE;
588 
589 	shift = 8 + (hdr->ip6e_len << 3);
590 	/*
591 	 * Nasty extension header length?
592 	 */
593 	if ((hdr->ip6e_len << 3) & 15) {
594 		fin->fin_flx |= FI_BAD;
595 		/*
596 		 * Compensate for the changes made in frpr_ipv6exthdr()
597 		 */
598 		fin->fin_dlen += shift;
599 		fin->fin_dp = (char *)fin->fin_dp - shift;
600 		return IPPROTO_NONE;
601 	}
602 
603 	return hdr->ip6e_nxt;
604 }
605 
606 
607 /* ------------------------------------------------------------------------ */
608 /* Function:    frpr_fragment6                                              */
609 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
610 /* Parameters:  fin(I) - pointer to packet information                      */
611 /*                                                                          */
612 /* IPv6 Only                                                                */
613 /* Examine the IPv6 fragment header and extract fragment offset information.*/
614 /*                                                                          */
615 /* We don't know where the transport layer header (or whatever is next is), */
616 /* as it could be behind destination options (amongst others).  Because     */
617 /* there is no fragment cache, there is no knowledge about whether or not an*/
618 /* upper layer header has been seen (or where it ends) and thus we are not  */
619 /* able to continue processing beyond this header with any confidence.      */
620 /* ------------------------------------------------------------------------ */
frpr_fragment6(fin)621 static INLINE int frpr_fragment6(fin)
622 fr_info_t *fin;
623 {
624 	struct ip6_frag *frag;
625 
626 	fin->fin_flx |= FI_FRAG;
627 
628 	/*
629 	 * A fragmented IPv6 packet implies that there must be something
630 	 * else after the fragment.
631 	 */
632 	if (frpr_ipv6exthdr(fin, 0, IPPROTO_FRAGMENT) == IPPROTO_NONE)
633 		return IPPROTO_NONE;
634 
635 	frag = (struct ip6_frag *)((char *)fin->fin_dp - sizeof(*frag));
636 
637 	/*
638 	 * If this fragment isn't the last then the packet length must
639 	 * be a multiple of 8.
640 	 */
641 	if ((frag->ip6f_offlg & IP6F_MORE_FRAG) != 0) {
642 		fin->fin_flx |= FI_MOREFRAG;
643 
644 		if ((fin->fin_plen & 0x7) != 0)
645 			fin->fin_flx |= FI_BAD;
646 	}
647 
648 	fin->fin_id = frag->ip6f_ident;
649 	fin->fin_off = ntohs(frag->ip6f_offlg & IP6F_OFF_MASK);
650 	if (fin->fin_off != 0)
651 		fin->fin_flx |= FI_FRAGBODY;
652 
653 	return frag->ip6f_nxt;
654 }
655 
656 
657 /* ------------------------------------------------------------------------ */
658 /* Function:    frpr_dstopts6                                               */
659 /* Returns:     int    - value of the next header or IPPROTO_NONE if error  */
660 /* Parameters:  fin(I) - pointer to packet information                      */
661 /*              nextheader(I) - stores next header value                    */
662 /*                                                                          */
663 /* IPv6 Only                                                                */
664 /* This is function checks pending destination options extension header     */
665 /* ------------------------------------------------------------------------ */
frpr_dstopts6(fin)666 static INLINE int frpr_dstopts6(fin)
667 fr_info_t *fin;
668 {
669 	return frpr_ipv6exthdr(fin, 1, IPPROTO_DSTOPTS);
670 }
671 
672 
673 /* ------------------------------------------------------------------------ */
674 /* Function:    frpr_icmp6                                                  */
675 /* Returns:     void                                                        */
676 /* Parameters:  fin(I) - pointer to packet information                      */
677 /*                                                                          */
678 /* IPv6 Only                                                                */
679 /* This routine is mainly concerned with determining the minimum valid size */
680 /* for an ICMPv6 packet.                                                    */
681 /* ------------------------------------------------------------------------ */
frpr_icmp6(fin)682 static INLINE void frpr_icmp6(fin)
683 fr_info_t *fin;
684 {
685 	int minicmpsz = sizeof(struct icmp6_hdr);
686 	struct icmp6_hdr *icmp6;
687 
688 	if (frpr_pullup(fin, ICMP6ERR_MINPKTLEN - sizeof(ip6_t)) == -1)
689 		return;
690 
691 	if (fin->fin_dlen > 1) {
692 		icmp6 = fin->fin_dp;
693 
694 		fin->fin_data[0] = *(u_short *)icmp6;
695 
696 		if ((icmp6->icmp6_type & ICMP6_INFOMSG_MASK) != 0)
697 			fin->fin_flx |= FI_ICMPQUERY;
698 
699 		switch (icmp6->icmp6_type)
700 		{
701 		case ICMP6_ECHO_REPLY :
702 		case ICMP6_ECHO_REQUEST :
703 			if (fin->fin_dlen >= 6)
704 				fin->fin_data[1] = icmp6->icmp6_id;
705 			minicmpsz = ICMP6ERR_MINPKTLEN - sizeof(ip6_t);
706 			break;
707 		case ICMP6_DST_UNREACH :
708 		case ICMP6_PACKET_TOO_BIG :
709 		case ICMP6_TIME_EXCEEDED :
710 		case ICMP6_PARAM_PROB :
711 			if ((fin->fin_m != NULL) &&
712 			    (M_LEN(fin->fin_m) < fin->fin_plen)) {
713 				if (fr_coalesce(fin) != 1)
714 					return;
715 			}
716 			fin->fin_flx |= FI_ICMPERR;
717 			minicmpsz = ICMP6ERR_IPICMPHLEN - sizeof(ip6_t);
718 			break;
719 		default :
720 			break;
721 		}
722 	}
723 
724 	frpr_short6(fin, minicmpsz);
725 }
726 
727 
728 /* ------------------------------------------------------------------------ */
729 /* Function:    frpr_udp6                                                   */
730 /* Returns:     void                                                        */
731 /* Parameters:  fin(I) - pointer to packet information                      */
732 /*                                                                          */
733 /* IPv6 Only                                                                */
734 /* Analyse the packet for IPv6/UDP properties.                              */
735 /* Is not expected to be called for fragmented packets.                     */
736 /* ------------------------------------------------------------------------ */
frpr_udp6(fin)737 static INLINE void frpr_udp6(fin)
738 fr_info_t *fin;
739 {
740 
741 	fr_checkv6sum(fin);
742 
743 	frpr_short6(fin, sizeof(struct udphdr));
744 	if (frpr_pullup(fin, sizeof(struct udphdr)) == -1)
745 		return;
746 
747 	frpr_udpcommon(fin);
748 }
749 
750 
751 /* ------------------------------------------------------------------------ */
752 /* Function:    frpr_tcp6                                                   */
753 /* Returns:     void                                                        */
754 /* Parameters:  fin(I) - pointer to packet information                      */
755 /*                                                                          */
756 /* IPv6 Only                                                                */
757 /* Analyse the packet for IPv6/TCP properties.                              */
758 /* Is not expected to be called for fragmented packets.                     */
759 /* ------------------------------------------------------------------------ */
frpr_tcp6(fin)760 static INLINE void frpr_tcp6(fin)
761 fr_info_t *fin;
762 {
763 
764 	fr_checkv6sum(fin);
765 
766 	frpr_short6(fin, sizeof(struct tcphdr));
767 	if (frpr_pullup(fin, sizeof(struct tcphdr)) == -1)
768 		return;
769 
770 	frpr_tcpcommon(fin);
771 }
772 
773 
774 /* ------------------------------------------------------------------------ */
775 /* Function:    frpr_esp6                                                   */
776 /* Returns:     void                                                        */
777 /* Parameters:  fin(I) - pointer to packet information                      */
778 /*                                                                          */
779 /* IPv6 Only                                                                */
780 /* Analyse the packet for ESP properties.                                   */
781 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
782 /* even though the newer ESP packets must also have a sequence number that  */
783 /* is 32bits as well, it is not possible(?) to determine the version from a */
784 /* simple packet header.                                                    */
785 /* ------------------------------------------------------------------------ */
frpr_esp6(fin)786 static INLINE void frpr_esp6(fin)
787 fr_info_t *fin;
788 {
789 	int i;
790 	frpr_short6(fin, sizeof(grehdr_t));
791 
792 	(void) frpr_pullup(fin, 8);
793 
794 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
795 		if (ip6exthdr[i].ol_val == IPPROTO_ESP) {
796 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
797 			break;
798 		}
799 }
800 
801 
802 /* ------------------------------------------------------------------------ */
803 /* Function:    frpr_ah6                                                    */
804 /* Returns:     void                                                        */
805 /* Parameters:  fin(I) - pointer to packet information                      */
806 /*                                                                          */
807 /* IPv6 Only                                                                */
808 /* Analyse the packet for AH properties.                                    */
809 /* The minimum length is taken to be the combination of all fields in the   */
810 /* header being present and no authentication data (null algorithm used.)   */
811 /* ------------------------------------------------------------------------ */
frpr_ah6(fin)812 static INLINE int frpr_ah6(fin)
813 fr_info_t *fin;
814 {
815 	authhdr_t *ah;
816 	int i, shift;
817 
818 	frpr_short6(fin, 12);
819 
820 	if (frpr_pullup(fin, sizeof(*ah)) == -1)
821 		return IPPROTO_NONE;
822 
823 	for (i = 0; ip6exthdr[i].ol_bit != 0; i++)
824 		if (ip6exthdr[i].ol_val == IPPROTO_AH) {
825 			fin->fin_optmsk |= ip6exthdr[i].ol_bit;
826 			break;
827 		}
828 
829 	ah = (authhdr_t *)fin->fin_dp;
830 
831 	shift = (ah->ah_plen + 2) * 4;
832 	fin->fin_dlen -= shift;
833 	fin->fin_dp = (char*)fin->fin_dp + shift;
834 
835 	return ah->ah_next;
836 }
837 
838 
839 /* ------------------------------------------------------------------------ */
840 /* Function:    frpr_gre6                                                   */
841 /* Returns:     void                                                        */
842 /* Parameters:  fin(I) - pointer to packet information                      */
843 /*                                                                          */
844 /* Analyse the packet for GRE properties.                                   */
845 /* ------------------------------------------------------------------------ */
frpr_gre6(fin)846 static INLINE void frpr_gre6(fin)
847 fr_info_t *fin;
848 {
849 	grehdr_t *gre;
850 
851 	frpr_short6(fin, sizeof(grehdr_t));
852 
853 	if (frpr_pullup(fin, sizeof(grehdr_t)) == -1)
854 		return;
855 
856 	gre = fin->fin_dp;
857 	if (GRE_REV(gre->gr_flags) == 1)
858 		fin->fin_data[0] = gre->gr_call;
859 }
860 #endif	/* USE_INET6 */
861 
862 
863 /* ------------------------------------------------------------------------ */
864 /* Function:    frpr_pullup                                                 */
865 /* Returns:     int     - 0 == pullup succeeded, -1 == failure              */
866 /* Parameters:  fin(I)  - pointer to packet information                     */
867 /*              plen(I) - length (excluding L3 header) to pullup            */
868 /*                                                                          */
869 /* Short inline function to cut down on code duplication to perform a call  */
870 /* to fr_pullup to ensure there is the required amount of data,             */
871 /* consecutively in the packet buffer.                                      */
872 /* ------------------------------------------------------------------------ */
frpr_pullup(fin,plen)873 static INLINE int frpr_pullup(fin, plen)
874 fr_info_t *fin;
875 int plen;
876 {
877 #if defined(_KERNEL)
878 	if (fin->fin_m != NULL) {
879 		int ipoff;
880 
881 		ipoff = (char *)fin->fin_ip - MTOD(fin->fin_m, char *);
882 
883 		if (fin->fin_dp != NULL)
884 			plen += (char *)fin->fin_dp -
885 				((char *)fin->fin_ip + fin->fin_hlen);
886 		plen += fin->fin_hlen;
887 		/*
888 		 * We don't do 'plen += ipoff;' here. The fr_pullup() will
889 		 * do it for us.
890 		 */
891 		if (M_LEN(fin->fin_m) < plen + ipoff) {
892 			if (fr_pullup(fin->fin_m, fin, plen) == NULL)
893 				return -1;
894 		}
895 	}
896 #endif
897 	return 0;
898 }
899 
900 
901 /* ------------------------------------------------------------------------ */
902 /* Function:    frpr_short                                                  */
903 /* Returns:     void                                                        */
904 /* Parameters:  fin(I)  - pointer to packet information                     */
905 /*              xmin(I) - minimum header size                               */
906 /*                                                                          */
907 /* Check if a packet is "short" as defined by xmin.  The rule we are        */
908 /* applying here is that the packet must not be fragmented within the layer */
909 /* 4 header.  That is, it must not be a fragment that has its offset set to */
910 /* start within the layer 4 header (hdrmin) or if it is at offset 0, the    */
911 /* entire layer 4 header must be present (min).                             */
912 /* ------------------------------------------------------------------------ */
frpr_short(fin,xmin)913 static INLINE void frpr_short(fin, xmin)
914 fr_info_t *fin;
915 int xmin;
916 {
917 
918 	if (fin->fin_off == 0) {
919 		if (fin->fin_dlen < xmin)
920 			fin->fin_flx |= FI_SHORT;
921 	} else if (fin->fin_off < xmin) {
922 		fin->fin_flx |= FI_SHORT;
923 	}
924 }
925 
926 
927 /* ------------------------------------------------------------------------ */
928 /* Function:    frpr_icmp                                                   */
929 /* Returns:     void                                                        */
930 /* Parameters:  fin(I) - pointer to packet information                      */
931 /*                                                                          */
932 /* IPv4 Only                                                                */
933 /* Do a sanity check on the packet for ICMP (v4).  In nearly all cases,     */
934 /* except extrememly bad packets, both type and code will be present.       */
935 /* The expected minimum size of an ICMP packet is very much dependent on    */
936 /* the type of it.                                                          */
937 /*                                                                          */
938 /* XXX - other ICMP sanity checks?                                          */
939 /* ------------------------------------------------------------------------ */
frpr_icmp(fin)940 static INLINE void frpr_icmp(fin)
941 fr_info_t *fin;
942 {
943 	int minicmpsz = sizeof(struct icmp);
944 	icmphdr_t *icmp;
945 	ip_t *oip;
946 	ipf_stack_t *ifs = fin->fin_ifs;
947 
948 	if (fin->fin_off != 0) {
949 		frpr_short(fin, ICMPERR_ICMPHLEN);
950 		return;
951 	}
952 
953 	if (frpr_pullup(fin, ICMPERR_ICMPHLEN) == -1)
954 		return;
955 
956 	fr_checkv4sum(fin);
957 
958 	/*
959 	 * This is a right place to set icmp pointer, since the memory
960 	 * referenced by fin_dp could get reallocated. The code down below can
961 	 * rely on fact icmp variable always points to ICMP header.
962 	 */
963 	icmp = fin->fin_dp;
964 	fin->fin_data[0] = *(u_short *)icmp;
965 	fin->fin_data[1] = icmp->icmp_id;
966 
967 	switch (icmp->icmp_type)
968 	{
969 	case ICMP_ECHOREPLY :
970 	case ICMP_ECHO :
971 	/* Router discovery messaes - RFC 1256 */
972 	case ICMP_ROUTERADVERT :
973 	case ICMP_ROUTERSOLICIT :
974 		minicmpsz = ICMP_MINLEN;
975 		break;
976 	/*
977 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
978 	 * 3 * timestamp(3 * 4)
979 	 */
980 	case ICMP_TSTAMP :
981 	case ICMP_TSTAMPREPLY :
982 		minicmpsz = 20;
983 		break;
984 	/*
985 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) +
986 	 * mask(4)
987 	 */
988 	case ICMP_MASKREQ :
989 	case ICMP_MASKREPLY :
990 			minicmpsz = 12;
991 			break;
992 	/*
993 	 * type(1) + code(1) + cksum(2) + id(2) seq(2) + ip(20+)
994 	 */
995 	case ICMP_UNREACH :
996 		if (icmp->icmp_code == ICMP_UNREACH_NEEDFRAG) {
997 			if (icmp->icmp_nextmtu < ifs->ifs_fr_icmpminfragmtu)
998 				fin->fin_flx |= FI_BAD;
999 		}
1000 		/* FALLTHRU */
1001 	case ICMP_SOURCEQUENCH :
1002 	case ICMP_REDIRECT :
1003 	case ICMP_TIMXCEED :
1004 	case ICMP_PARAMPROB :
1005 		fin->fin_flx |= FI_ICMPERR;
1006 		if (fr_coalesce(fin) != 1)
1007 			return;
1008 		/*
1009 		 * ICMP error packets should not be generated for IP
1010 		 * packets that are a fragment that isn't the first
1011 		 * fragment.
1012 		 */
1013 		oip = (ip_t *)((char *)fin->fin_dp + ICMPERR_ICMPHLEN);
1014 		if ((ntohs(oip->ip_off) & IP_OFFMASK) != 0)
1015 			fin->fin_flx |= FI_BAD;
1016 		break;
1017 	default :
1018 		break;
1019 	}
1020 
1021 	frpr_short(fin, minicmpsz);
1022 }
1023 
1024 
1025 /* ------------------------------------------------------------------------ */
1026 /* Function:    frpr_tcpcommon                                              */
1027 /* Returns:     void                                                        */
1028 /* Parameters:  fin(I) - pointer to packet information                      */
1029 /*                                                                          */
1030 /* TCP header sanity checking.  Look for bad combinations of TCP flags,     */
1031 /* and make some checks with how they interact with other fields.           */
1032 /* If compiled with IPFILTER_CKSUM, check to see if the TCP checksum is     */
1033 /* valid and mark the packet as bad if not.                                 */
1034 /* ------------------------------------------------------------------------ */
frpr_tcpcommon(fin)1035 static INLINE void frpr_tcpcommon(fin)
1036 fr_info_t *fin;
1037 {
1038 	int flags, tlen;
1039 	tcphdr_t *tcp;
1040 
1041 	fin->fin_flx |= FI_TCPUDP;
1042 	if (fin->fin_off != 0)
1043 		return;
1044 
1045 	if (frpr_pullup(fin, sizeof(*tcp)) == -1)
1046 		return;
1047 	tcp = fin->fin_dp;
1048 
1049 	if (fin->fin_dlen > 3) {
1050 		fin->fin_sport = ntohs(tcp->th_sport);
1051 		fin->fin_dport = ntohs(tcp->th_dport);
1052 	}
1053 
1054 	if ((fin->fin_flx & FI_SHORT) != 0)
1055 		return;
1056 
1057 	/*
1058 	 * Use of the TCP data offset *must* result in a value that is at
1059 	 * least the same size as the TCP header.
1060 	 */
1061 	tlen = TCP_OFF(tcp) << 2;
1062 	if (tlen < sizeof(tcphdr_t)) {
1063 		fin->fin_flx |= FI_BAD;
1064 		return;
1065 	}
1066 
1067 	flags = tcp->th_flags;
1068 	fin->fin_tcpf = tcp->th_flags;
1069 
1070 	/*
1071 	 * If the urgent flag is set, then the urgent pointer must
1072 	 * also be set and vice versa.  Good TCP packets do not have
1073 	 * just one of these set.
1074 	 */
1075 	if ((flags & TH_URG) != 0 && (tcp->th_urp == 0)) {
1076 		fin->fin_flx |= FI_BAD;
1077 	} else if ((flags & TH_URG) == 0 && (tcp->th_urp != 0)) {
1078 		/* Ignore this case, it shows up in "real" traffic with */
1079 		/* bogus values in the urgent pointer field. */
1080 		flags = flags; /* LINT */
1081 	} else if (((flags & (TH_SYN|TH_FIN)) != 0) &&
1082 		   ((flags & (TH_RST|TH_ACK)) == TH_RST)) {
1083 		/* TH_FIN|TH_RST|TH_ACK seems to appear "naturally" */
1084 		fin->fin_flx |= FI_BAD;
1085 	} else if (!(flags & TH_ACK)) {
1086 		/*
1087 		 * If the ack bit isn't set, then either the SYN or
1088 		 * RST bit must be set.  If the SYN bit is set, then
1089 		 * we expect the ACK field to be 0.  If the ACK is
1090 		 * not set and if URG, PSH or FIN are set, consdier
1091 		 * that to indicate a bad TCP packet.
1092 		 */
1093 		if ((flags == TH_SYN) && (tcp->th_ack != 0)) {
1094 			/*
1095 			 * Cisco PIX sets the ACK field to a random value.
1096 			 * In light of this, do not set FI_BAD until a patch
1097 			 * is available from Cisco to ensure that
1098 			 * interoperability between existing systems is
1099 			 * achieved.
1100 			 */
1101 			/*fin->fin_flx |= FI_BAD*/;
1102 			flags = flags; /* LINT */
1103 		} else if (!(flags & (TH_RST|TH_SYN))) {
1104 			fin->fin_flx |= FI_BAD;
1105 		} else if ((flags & (TH_URG|TH_PUSH|TH_FIN)) != 0) {
1106 			fin->fin_flx |= FI_BAD;
1107 		}
1108 	}
1109 
1110 	/*
1111 	 * At this point, it's not exactly clear what is to be gained by
1112 	 * marking up which TCP options are and are not present.  The one we
1113 	 * are most interested in is the TCP window scale.  This is only in
1114 	 * a SYN packet [RFC1323] so we don't need this here...?
1115 	 * Now if we were to analyse the header for passive fingerprinting,
1116 	 * then that might add some weight to adding this...
1117 	 */
1118 	if (tlen == sizeof(tcphdr_t))
1119 		return;
1120 
1121 	if (frpr_pullup(fin, tlen) == -1)
1122 		return;
1123 
1124 #if 0
1125 	ip = fin->fin_ip;
1126 	s = (u_char *)(tcp + 1);
1127 	off = IP_HL(ip) << 2;
1128 # ifdef _KERNEL
1129 	if (fin->fin_mp != NULL) {
1130 		mb_t *m = *fin->fin_mp;
1131 
1132 		if (off + tlen > M_LEN(m))
1133 			return;
1134 	}
1135 # endif
1136 	for (tlen -= (int)sizeof(*tcp); tlen > 0; ) {
1137 		opt = *s;
1138 		if (opt == '\0')
1139 			break;
1140 		else if (opt == TCPOPT_NOP)
1141 			ol = 1;
1142 		else {
1143 			if (tlen < 2)
1144 				break;
1145 			ol = (int)*(s + 1);
1146 			if (ol < 2 || ol > tlen)
1147 				break;
1148 		}
1149 
1150 		for (i = 9, mv = 4; mv >= 0; ) {
1151 			op = ipopts + i;
1152 			if (opt == (u_char)op->ol_val) {
1153 				optmsk |= op->ol_bit;
1154 				break;
1155 			}
1156 		}
1157 		tlen -= ol;
1158 		s += ol;
1159 	}
1160 #endif /* 0 */
1161 }
1162 
1163 
1164 
1165 /* ------------------------------------------------------------------------ */
1166 /* Function:    frpr_udpcommon                                              */
1167 /* Returns:     void                                                        */
1168 /* Parameters:  fin(I) - pointer to packet information                      */
1169 /*                                                                          */
1170 /* Extract the UDP source and destination ports, if present.  If compiled   */
1171 /* with IPFILTER_CKSUM, check to see if the UDP checksum is valid.          */
1172 /* ------------------------------------------------------------------------ */
frpr_udpcommon(fin)1173 static INLINE void frpr_udpcommon(fin)
1174 fr_info_t *fin;
1175 {
1176 	udphdr_t *udp;
1177 
1178 	fin->fin_flx |= FI_TCPUDP;
1179 
1180 	if (!fin->fin_off && (fin->fin_dlen > 3)) {
1181 		if (frpr_pullup(fin, sizeof(*udp)) == -1) {
1182 			fin->fin_flx |= FI_SHORT;
1183 			return;
1184 		}
1185 
1186 		udp = fin->fin_dp;
1187 
1188 		fin->fin_sport = ntohs(udp->uh_sport);
1189 		fin->fin_dport = ntohs(udp->uh_dport);
1190 	}
1191 }
1192 
1193 
1194 /* ------------------------------------------------------------------------ */
1195 /* Function:    frpr_tcp                                                    */
1196 /* Returns:     void                                                        */
1197 /* Parameters:  fin(I) - pointer to packet information                      */
1198 /*                                                                          */
1199 /* IPv4 Only                                                                */
1200 /* Analyse the packet for IPv4/TCP properties.                              */
1201 /* ------------------------------------------------------------------------ */
frpr_tcp(fin)1202 static INLINE void frpr_tcp(fin)
1203 fr_info_t *fin;
1204 {
1205 
1206 	fr_checkv4sum(fin);
1207 
1208 	frpr_short(fin, sizeof(tcphdr_t));
1209 
1210 	frpr_tcpcommon(fin);
1211 }
1212 
1213 
1214 /* ------------------------------------------------------------------------ */
1215 /* Function:    frpr_udp                                                    */
1216 /* Returns:     void                                                        */
1217 /* Parameters:  fin(I) - pointer to packet information                      */
1218 /*                                                                          */
1219 /* IPv4 Only                                                                */
1220 /* Analyse the packet for IPv4/UDP properties.                              */
1221 /* ------------------------------------------------------------------------ */
frpr_udp(fin)1222 static INLINE void frpr_udp(fin)
1223 fr_info_t *fin;
1224 {
1225 
1226 	fr_checkv4sum(fin);
1227 
1228 	frpr_short(fin, sizeof(udphdr_t));
1229 
1230 	frpr_udpcommon(fin);
1231 }
1232 
1233 
1234 /* ------------------------------------------------------------------------ */
1235 /* Function:    frpr_esp                                                    */
1236 /* Returns:     void                                                        */
1237 /* Parameters:  fin(I) - pointer to packet information                      */
1238 /*                                                                          */
1239 /* Analyse the packet for ESP properties.                                   */
1240 /* The minimum length is taken to be the SPI (32bits) plus a tail (32bits)  */
1241 /* even though the newer ESP packets must also have a sequence number that  */
1242 /* is 32bits as well, it is not possible(?) to determine the version from a */
1243 /* simple packet header.                                                    */
1244 /* ------------------------------------------------------------------------ */
frpr_esp(fin)1245 static INLINE void frpr_esp(fin)
1246 fr_info_t *fin;
1247 {
1248 	if ((fin->fin_off == 0) && (frpr_pullup(fin, 8) == -1))
1249 		return;
1250 
1251 	frpr_short(fin, 8);
1252 }
1253 
1254 
1255 /* ------------------------------------------------------------------------ */
1256 /* Function:    frpr_ah                                                     */
1257 /* Returns:     void                                                        */
1258 /* Parameters:  fin(I) - pointer to packet information                      */
1259 /*                                                                          */
1260 /* Analyse the packet for AH properties.                                    */
1261 /* The minimum length is taken to be the combination of all fields in the   */
1262 /* header being present and no authentication data (null algorithm used.)   */
1263 /* ------------------------------------------------------------------------ */
frpr_ah(fin)1264 static INLINE void frpr_ah(fin)
1265 fr_info_t *fin;
1266 {
1267 	authhdr_t *ah;
1268 	int len;
1269 
1270 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(*ah)) == -1))
1271 		return;
1272 
1273 	ah = (authhdr_t *)fin->fin_dp;
1274 
1275 	len = (ah->ah_plen + 2) << 2;
1276 	frpr_short(fin, len);
1277 }
1278 
1279 
1280 /* ------------------------------------------------------------------------ */
1281 /* Function:    frpr_gre                                                    */
1282 /* Returns:     void                                                        */
1283 /* Parameters:  fin(I) - pointer to packet information                      */
1284 /*                                                                          */
1285 /* Analyse the packet for GRE properties.                                   */
1286 /* ------------------------------------------------------------------------ */
frpr_gre(fin)1287 static INLINE void frpr_gre(fin)
1288 fr_info_t *fin;
1289 {
1290 	grehdr_t *gre;
1291 
1292 	if ((fin->fin_off == 0) && (frpr_pullup(fin, sizeof(grehdr_t)) == -1))
1293 		return;
1294 
1295 	frpr_short(fin, sizeof(grehdr_t));
1296 
1297 	if (fin->fin_off == 0) {
1298 		gre = fin->fin_dp;
1299 		if (GRE_REV(gre->gr_flags) == 1)
1300 			fin->fin_data[0] = gre->gr_call;
1301 	}
1302 }
1303 
1304 
1305 /* ------------------------------------------------------------------------ */
1306 /* Function:    frpr_ipv4hdr                                                */
1307 /* Returns:     void                                                        */
1308 /* Parameters:  fin(I) - pointer to packet information                      */
1309 /*                                                                          */
1310 /* IPv4 Only                                                                */
1311 /* Analyze the IPv4 header and set fields in the fr_info_t structure.       */
1312 /* Check all options present and flag their presence if any exist.          */
1313 /* ------------------------------------------------------------------------ */
frpr_ipv4hdr(fin)1314 static INLINE void frpr_ipv4hdr(fin)
1315 fr_info_t *fin;
1316 {
1317 	u_short optmsk = 0, secmsk = 0, auth = 0;
1318 	int hlen, ol, mv, p, i;
1319 	const struct optlist *op;
1320 	u_char *s, opt;
1321 	u_short off;
1322 	fr_ip_t *fi;
1323 	ip_t *ip;
1324 
1325 	fi = &fin->fin_fi;
1326 	hlen = fin->fin_hlen;
1327 
1328 	ip = fin->fin_ip;
1329 	p = ip->ip_p;
1330 	fi->fi_p = p;
1331 	fi->fi_tos = ip->ip_tos;
1332 	fin->fin_id = ip->ip_id;
1333 	off = ip->ip_off;
1334 
1335 	/* Get both TTL and protocol */
1336 	fi->fi_p = ip->ip_p;
1337 	fi->fi_ttl = ip->ip_ttl;
1338 #if 0
1339 	(*(((u_short *)fi) + 1)) = (*(((u_short *)ip) + 4));
1340 #endif
1341 
1342 	/* Zero out bits not used in IPv6 address */
1343 	fi->fi_src.i6[1] = 0;
1344 	fi->fi_src.i6[2] = 0;
1345 	fi->fi_src.i6[3] = 0;
1346 	fi->fi_dst.i6[1] = 0;
1347 	fi->fi_dst.i6[2] = 0;
1348 	fi->fi_dst.i6[3] = 0;
1349 
1350 	fi->fi_saddr = ip->ip_src.s_addr;
1351 	fi->fi_daddr = ip->ip_dst.s_addr;
1352 
1353 	/*
1354 	 * set packet attribute flags based on the offset and
1355 	 * calculate the byte offset that it represents.
1356 	 */
1357 	off &= IP_MF|IP_OFFMASK;
1358 	if (off != 0) {
1359 		int morefrag = off & IP_MF;
1360 
1361 		fi->fi_flx |= FI_FRAG;
1362 		if (morefrag)
1363 			fi->fi_flx |= FI_MOREFRAG;
1364 		off &= IP_OFFMASK;
1365 		if (off != 0) {
1366 			fin->fin_flx |= FI_FRAGBODY;
1367 			off <<= 3;
1368 			if ((off + fin->fin_dlen > 65535) ||
1369 			    (fin->fin_dlen == 0) ||
1370 			    ((morefrag != 0) && ((fin->fin_dlen & 7) != 0))) {
1371 				/*
1372 				 * The length of the packet, starting at its
1373 				 * offset cannot exceed 65535 (0xffff) as the
1374 				 * length of an IP packet is only 16 bits.
1375 				 *
1376 				 * Any fragment that isn't the last fragment
1377 				 * must have a length greater than 0 and it
1378 				 * must be an even multiple of 8.
1379 				 */
1380 				fi->fi_flx |= FI_BAD;
1381 			}
1382 		}
1383 	}
1384 	fin->fin_off = off;
1385 
1386 	/*
1387 	 * Call per-protocol setup and checking
1388 	 */
1389 	switch (p)
1390 	{
1391 	case IPPROTO_UDP :
1392 		frpr_udp(fin);
1393 		break;
1394 	case IPPROTO_TCP :
1395 		frpr_tcp(fin);
1396 		break;
1397 	case IPPROTO_ICMP :
1398 		frpr_icmp(fin);
1399 		break;
1400 	case IPPROTO_AH :
1401 		frpr_ah(fin);
1402 		break;
1403 	case IPPROTO_ESP :
1404 		frpr_esp(fin);
1405 		break;
1406 	case IPPROTO_GRE :
1407 		frpr_gre(fin);
1408 		break;
1409 	}
1410 
1411 	ip = fin->fin_ip;
1412 	if (ip == NULL)
1413 		return;
1414 
1415 	/*
1416 	 * If it is a standard IP header (no options), set the flag fields
1417 	 * which relate to options to 0.
1418 	 */
1419 	if (hlen == sizeof(*ip)) {
1420 		fi->fi_optmsk = 0;
1421 		fi->fi_secmsk = 0;
1422 		fi->fi_auth = 0;
1423 		return;
1424 	}
1425 
1426 	/*
1427 	 * So the IP header has some IP options attached.  Walk the entire
1428 	 * list of options present with this packet and set flags to indicate
1429 	 * which ones are here and which ones are not.  For the somewhat out
1430 	 * of date and obscure security classification options, set a flag to
1431 	 * represent which classification is present.
1432 	 */
1433 	fi->fi_flx |= FI_OPTIONS;
1434 
1435 	for (s = (u_char *)(ip + 1), hlen -= (int)sizeof(*ip); hlen > 0; ) {
1436 		opt = *s;
1437 		if (opt == '\0')
1438 			break;
1439 		else if (opt == IPOPT_NOP)
1440 			ol = 1;
1441 		else {
1442 			if (hlen < 2)
1443 				break;
1444 			ol = (int)*(s + 1);
1445 			if (ol < 2 || ol > hlen)
1446 				break;
1447 		}
1448 		for (i = 9, mv = 4; mv >= 0; ) {
1449 			op = ipopts + i;
1450 			if ((opt == (u_char)op->ol_val) && (ol > 4)) {
1451 				optmsk |= op->ol_bit;
1452 				if (opt == IPOPT_SECURITY) {
1453 					const struct optlist *sp;
1454 					u_char	sec;
1455 					int j, m;
1456 
1457 					sec = *(s + 2);	/* classification */
1458 					for (j = 3, m = 2; m >= 0; ) {
1459 						sp = secopt + j;
1460 						if (sec == sp->ol_val) {
1461 							secmsk |= sp->ol_bit;
1462 							auth = *(s + 3);
1463 							auth *= 256;
1464 							auth += *(s + 4);
1465 							break;
1466 						}
1467 						if (sec < sp->ol_val)
1468 							j -= m;
1469 						else
1470 							j += m;
1471 						m--;
1472 					}
1473 				}
1474 				break;
1475 			}
1476 			if (opt < op->ol_val)
1477 				i -= mv;
1478 			else
1479 				i += mv;
1480 			mv--;
1481 		}
1482 		hlen -= ol;
1483 		s += ol;
1484 	}
1485 
1486 	/*
1487 	 *
1488 	 */
1489 	if (auth && !(auth & 0x0100))
1490 		auth &= 0xff00;
1491 	fi->fi_optmsk = optmsk;
1492 	fi->fi_secmsk = secmsk;
1493 	fi->fi_auth = auth;
1494 }
1495 
1496 
1497 /* ------------------------------------------------------------------------ */
1498 /* Function:    fr_makefrip                                                 */
1499 /* Returns:     int - 1 == hdr checking error, 0 == OK                      */
1500 /* Parameters:  hlen(I) - length of IP packet header                        */
1501 /*              ip(I)   - pointer to the IP header                          */
1502 /*              fin(IO) - pointer to packet information                     */
1503 /*                                                                          */
1504 /* Compact the IP header into a structure which contains just the info.     */
1505 /* which is useful for comparing IP headers with and store this information */
1506 /* in the fr_info_t structure pointer to by fin.  At present, it is assumed */
1507 /* this function will be called with either an IPv4 or IPv6 packet.         */
1508 /* ------------------------------------------------------------------------ */
fr_makefrip(hlen,ip,fin)1509 int	fr_makefrip(hlen, ip, fin)
1510 int hlen;
1511 ip_t *ip;
1512 fr_info_t *fin;
1513 {
1514 	int v;
1515 
1516 	fin->fin_depth = 0;
1517 	fin->fin_hlen = (u_short)hlen;
1518 	fin->fin_ip = ip;
1519 	fin->fin_rule = 0xffffffff;
1520 	fin->fin_group[0] = -1;
1521 	fin->fin_group[1] = '\0';
1522 	fin->fin_dlen = fin->fin_plen - hlen;
1523 	fin->fin_dp = (char *)ip + hlen;
1524 
1525 	v = fin->fin_v;
1526 	if (v == 4)
1527 		frpr_ipv4hdr(fin);
1528 #ifdef	USE_INET6
1529 	else if (v == 6)
1530 		frpr_ipv6hdr(fin);
1531 #endif
1532 	if (fin->fin_ip == NULL)
1533 		return -1;
1534 	return 0;
1535 }
1536 
1537 
1538 /* ------------------------------------------------------------------------ */
1539 /* Function:    fr_portcheck                                                */
1540 /* Returns:     int - 1 == port matched, 0 == port match failed             */
1541 /* Parameters:  frp(I) - pointer to port check `expression'                 */
1542 /*              pop(I) - pointer to port number to evaluate                 */
1543 /*                                                                          */
1544 /* Perform a comparison of a port number against some other(s), using a     */
1545 /* structure with compare information stored in it.                         */
1546 /* ------------------------------------------------------------------------ */
fr_portcheck(frp,pop)1547 static INLINE int fr_portcheck(frp, pop)
1548 frpcmp_t *frp;
1549 u_short *pop;
1550 {
1551 	u_short tup, po;
1552 	int err = 1;
1553 
1554 	tup = *pop;
1555 	po = frp->frp_port;
1556 
1557 	/*
1558 	 * Do opposite test to that required and continue if that succeeds.
1559 	 */
1560 	switch (frp->frp_cmp)
1561 	{
1562 	case FR_EQUAL :
1563 		if (tup != po) /* EQUAL */
1564 			err = 0;
1565 		break;
1566 	case FR_NEQUAL :
1567 		if (tup == po) /* NOTEQUAL */
1568 			err = 0;
1569 		break;
1570 	case FR_LESST :
1571 		if (tup >= po) /* LESSTHAN */
1572 			err = 0;
1573 		break;
1574 	case FR_GREATERT :
1575 		if (tup <= po) /* GREATERTHAN */
1576 			err = 0;
1577 		break;
1578 	case FR_LESSTE :
1579 		if (tup > po) /* LT or EQ */
1580 			err = 0;
1581 		break;
1582 	case FR_GREATERTE :
1583 		if (tup < po) /* GT or EQ */
1584 			err = 0;
1585 		break;
1586 	case FR_OUTRANGE :
1587 		if (tup >= po && tup <= frp->frp_top) /* Out of range */
1588 			err = 0;
1589 		break;
1590 	case FR_INRANGE :
1591 		if (tup <= po || tup >= frp->frp_top) /* In range */
1592 			err = 0;
1593 		break;
1594 	case FR_INCRANGE :
1595 		if (tup < po || tup > frp->frp_top) /* Inclusive range */
1596 			err = 0;
1597 		break;
1598 	default :
1599 		break;
1600 	}
1601 	return err;
1602 }
1603 
1604 
1605 /* ------------------------------------------------------------------------ */
1606 /* Function:    fr_tcpudpchk                                                */
1607 /* Returns:     int - 1 == protocol matched, 0 == check failed              */
1608 /* Parameters:  fin(I) - pointer to packet information                      */
1609 /*              ft(I)  - pointer to structure with comparison data          */
1610 /*                                                                          */
1611 /* Compares the current pcket (assuming it is TCP/UDP) information with a   */
1612 /* structure containing information that we want to match against.          */
1613 /* ------------------------------------------------------------------------ */
fr_tcpudpchk(fin,ft)1614 int fr_tcpudpchk(fin, ft)
1615 fr_info_t *fin;
1616 frtuc_t *ft;
1617 {
1618 	int err = 1;
1619 
1620 	/*
1621 	 * Both ports should *always* be in the first fragment.
1622 	 * So far, I cannot find any cases where they can not be.
1623 	 *
1624 	 * compare destination ports
1625 	 */
1626 	if (ft->ftu_dcmp)
1627 		err = fr_portcheck(&ft->ftu_dst, &fin->fin_dport);
1628 
1629 	/*
1630 	 * compare source ports
1631 	 */
1632 	if (err && ft->ftu_scmp)
1633 		err = fr_portcheck(&ft->ftu_src, &fin->fin_sport);
1634 
1635 	/*
1636 	 * If we don't have all the TCP/UDP header, then how can we
1637 	 * expect to do any sort of match on it ?  If we were looking for
1638 	 * TCP flags, then NO match.  If not, then match (which should
1639 	 * satisfy the "short" class too).
1640 	 */
1641 	if (err && (fin->fin_p == IPPROTO_TCP)) {
1642 		if (fin->fin_flx & FI_SHORT)
1643 			return !(ft->ftu_tcpf | ft->ftu_tcpfm);
1644 		/*
1645 		 * Match the flags ?  If not, abort this match.
1646 		 */
1647 		if (ft->ftu_tcpfm &&
1648 		    ft->ftu_tcpf != (fin->fin_tcpf & ft->ftu_tcpfm)) {
1649 			FR_DEBUG(("f. %#x & %#x != %#x\n", fin->fin_tcpf,
1650 				 ft->ftu_tcpfm, ft->ftu_tcpf));
1651 			err = 0;
1652 		}
1653 	}
1654 	return err;
1655 }
1656 
1657 
1658 /* ------------------------------------------------------------------------ */
1659 /* Function:    fr_ipfcheck                                                 */
1660 /* Returns:     int - 0 == match, 1 == no match                             */
1661 /* Parameters:  fin(I)     - pointer to packet information                  */
1662 /*              fr(I)      - pointer to filter rule                         */
1663 /*              portcmp(I) - flag indicating whether to attempt matching on */
1664 /*                           TCP/UDP port data.                             */
1665 /*                                                                          */
1666 /* Check to see if a packet matches an IPFilter rule.  Checks of addresses, */
1667 /* port numbers, etc, for "standard" IPFilter rules are all orchestrated in */
1668 /* this function.                                                           */
1669 /* ------------------------------------------------------------------------ */
fr_ipfcheck(fin,fr,portcmp)1670 static INLINE int fr_ipfcheck(fin, fr, portcmp)
1671 fr_info_t *fin;
1672 frentry_t *fr;
1673 int portcmp;
1674 {
1675 	u_32_t	*ld, *lm, *lip;
1676 	fripf_t *fri;
1677 	fr_ip_t *fi;
1678 	int i;
1679 	ipf_stack_t *ifs = fin->fin_ifs;
1680 
1681 	fi = &fin->fin_fi;
1682 	fri = fr->fr_ipf;
1683 	lip = (u_32_t *)fi;
1684 	lm = (u_32_t *)&fri->fri_mip;
1685 	ld = (u_32_t *)&fri->fri_ip;
1686 
1687 	/*
1688 	 * first 32 bits to check coversion:
1689 	 * IP version, TOS, TTL, protocol
1690 	 */
1691 	i = ((*lip & *lm) != *ld);
1692 	FR_DEBUG(("0. %#08x & %#08x != %#08x\n",
1693 		   *lip, *lm, *ld));
1694 	if (i)
1695 		return 1;
1696 
1697 	/*
1698 	 * Next 32 bits is a constructed bitmask indicating which IP options
1699 	 * are present (if any) in this packet.
1700 	 */
1701 	lip++, lm++, ld++;
1702 	i |= ((*lip & *lm) != *ld);
1703 	FR_DEBUG(("1. %#08x & %#08x != %#08x\n",
1704 		   *lip, *lm, *ld));
1705 	if (i)
1706 		return 1;
1707 
1708 	lip++, lm++, ld++;
1709 	/*
1710 	 * Unrolled loops (4 each, for 32 bits) for address checks.
1711 	 */
1712 	/*
1713 	 * Check the source address.
1714 	 */
1715 #ifdef	IPFILTER_LOOKUP
1716 	if (fr->fr_satype == FRI_LOOKUP) {
1717 		fin->fin_flx |= FI_DONTCACHE;
1718 		i = (*fr->fr_srcfunc)(fr->fr_srcptr, fi->fi_v, lip, fin, ifs);
1719 		if (i == -1)
1720 			return 1;
1721 		lip += 3;
1722 		lm += 3;
1723 		ld += 3;
1724 	} else {
1725 #endif
1726 		i = ((*lip & *lm) != *ld);
1727 		FR_DEBUG(("2a. %#08x & %#08x != %#08x\n",
1728 			   *lip, *lm, *ld));
1729 		if (fi->fi_v == 6) {
1730 			lip++, lm++, ld++;
1731 			i |= ((*lip & *lm) != *ld);
1732 			FR_DEBUG(("2b. %#08x & %#08x != %#08x\n",
1733 				   *lip, *lm, *ld));
1734 			lip++, lm++, ld++;
1735 			i |= ((*lip & *lm) != *ld);
1736 			FR_DEBUG(("2c. %#08x & %#08x != %#08x\n",
1737 				   *lip, *lm, *ld));
1738 			lip++, lm++, ld++;
1739 			i |= ((*lip & *lm) != *ld);
1740 			FR_DEBUG(("2d. %#08x & %#08x != %#08x\n",
1741 				   *lip, *lm, *ld));
1742 		} else {
1743 			lip += 3;
1744 			lm += 3;
1745 			ld += 3;
1746 		}
1747 #ifdef	IPFILTER_LOOKUP
1748 	}
1749 #endif
1750 	i ^= (fr->fr_flags & FR_NOTSRCIP) >> 6;
1751 	if (i)
1752 		return 1;
1753 
1754 	/*
1755 	 * Check the destination address.
1756 	 */
1757 	lip++, lm++, ld++;
1758 #ifdef	IPFILTER_LOOKUP
1759 	if (fr->fr_datype == FRI_LOOKUP) {
1760 		fin->fin_flx |= FI_DONTCACHE;
1761 		i = (*fr->fr_dstfunc)(fr->fr_dstptr, fi->fi_v, lip, fin, ifs);
1762 		if (i == -1)
1763 			return 1;
1764 		lip += 3;
1765 		lm += 3;
1766 		ld += 3;
1767 	} else {
1768 #endif
1769 		i = ((*lip & *lm) != *ld);
1770 		FR_DEBUG(("3a. %#08x & %#08x != %#08x\n",
1771 			   *lip, *lm, *ld));
1772 		if (fi->fi_v == 6) {
1773 			lip++, lm++, ld++;
1774 			i |= ((*lip & *lm) != *ld);
1775 			FR_DEBUG(("3b. %#08x & %#08x != %#08x\n",
1776 				   *lip, *lm, *ld));
1777 			lip++, lm++, ld++;
1778 			i |= ((*lip & *lm) != *ld);
1779 			FR_DEBUG(("3c. %#08x & %#08x != %#08x\n",
1780 				   *lip, *lm, *ld));
1781 			lip++, lm++, ld++;
1782 			i |= ((*lip & *lm) != *ld);
1783 			FR_DEBUG(("3d. %#08x & %#08x != %#08x\n",
1784 				   *lip, *lm, *ld));
1785 		} else {
1786 			lip += 3;
1787 			lm += 3;
1788 			ld += 3;
1789 		}
1790 #ifdef	IPFILTER_LOOKUP
1791 	}
1792 #endif
1793 	i ^= (fr->fr_flags & FR_NOTDSTIP) >> 7;
1794 	if (i)
1795 		return 1;
1796 	/*
1797 	 * IP addresses matched.  The next 32bits contains:
1798 	 * mast of old IP header security & authentication bits.
1799 	 */
1800 	lip++, lm++, ld++;
1801 	i |= ((*lip & *lm) != *ld);
1802 	FR_DEBUG(("4. %#08x & %#08x != %#08x\n",
1803 		   *lip, *lm, *ld));
1804 
1805 	/*
1806 	 * Next we have 32 bits of packet flags.
1807 	 */
1808 	lip++, lm++, ld++;
1809 	i |= ((*lip & *lm) != *ld);
1810 	FR_DEBUG(("5. %#08x & %#08x != %#08x\n",
1811 		   *lip, *lm, *ld));
1812 
1813 	if (i == 0) {
1814 		/*
1815 		 * If a fragment, then only the first has what we're
1816 		 * looking for here...
1817 		 */
1818 		if (portcmp) {
1819 			if (!fr_tcpudpchk(fin, &fr->fr_tuc))
1820 				i = 1;
1821 		} else {
1822 			if (fr->fr_dcmp || fr->fr_scmp ||
1823 			    fr->fr_tcpf || fr->fr_tcpfm)
1824 				i = 1;
1825 			if (fr->fr_icmpm || fr->fr_icmp) {
1826 				if (((fi->fi_p != IPPROTO_ICMP) &&
1827 				     (fi->fi_p != IPPROTO_ICMPV6)) ||
1828 				    fin->fin_off || (fin->fin_dlen < 2))
1829 					i = 1;
1830 				else if ((fin->fin_data[0] & fr->fr_icmpm) !=
1831 					 fr->fr_icmp) {
1832 					FR_DEBUG(("i. %#x & %#x != %#x\n",
1833 						 fin->fin_data[0],
1834 						 fr->fr_icmpm, fr->fr_icmp));
1835 					i = 1;
1836 				}
1837 			}
1838 		}
1839 	}
1840 	return i;
1841 }
1842 
1843 
1844 /* ------------------------------------------------------------------------ */
1845 /* Function:    fr_scanlist                                                 */
1846 /* Returns:     int - result flags of scanning filter list                  */
1847 /* Parameters:  fin(I) - pointer to packet information                      */
1848 /*              pass(I) - default result to return for filtering            */
1849 /*                                                                          */
1850 /* Check the input/output list of rules for a match to the current packet.  */
1851 /* If a match is found, the value of fr_flags from the rule becomes the     */
1852 /* return value and fin->fin_fr points to the matched rule.                 */
1853 /*                                                                          */
1854 /* This function may be called recusively upto 16 times (limit inbuilt.)    */
1855 /* When unwinding, it should finish up with fin_depth as 0.                 */
1856 /*                                                                          */
1857 /* Could be per interface, but this gets real nasty when you don't have,    */
1858 /* or can't easily change, the kernel source code to .                      */
1859 /* ------------------------------------------------------------------------ */
fr_scanlist(fin,pass)1860 int fr_scanlist(fin, pass)
1861 fr_info_t *fin;
1862 u_32_t pass;
1863 {
1864 	int rulen, portcmp, off, logged, skip;
1865 	struct frentry *fr, *fnext;
1866 	u_32_t passt, passo;
1867 	ipf_stack_t *ifs = fin->fin_ifs;
1868 
1869 	/*
1870 	 * Do not allow nesting deeper than 16 levels.
1871 	 */
1872 	if (fin->fin_depth >= 16)
1873 		return pass;
1874 
1875 	fr = fin->fin_fr;
1876 
1877 	/*
1878 	 * If there are no rules in this list, return now.
1879 	 */
1880 	if (fr == NULL)
1881 		return pass;
1882 
1883 	skip = 0;
1884 	logged = 0;
1885 	portcmp = 0;
1886 	fin->fin_depth++;
1887 	fin->fin_fr = NULL;
1888 	off = fin->fin_off;
1889 
1890 	if ((fin->fin_flx & FI_TCPUDP) && (fin->fin_dlen > 3) && !off)
1891 		portcmp = 1;
1892 
1893 	for (rulen = 0; fr; fr = fnext, rulen++) {
1894 		fnext = fr->fr_next;
1895 		if (skip != 0) {
1896 			FR_VERBOSE(("%d (%#x)\n", skip, fr->fr_flags));
1897 			skip--;
1898 			continue;
1899 		}
1900 
1901 		/*
1902 		 * In all checks below, a null (zero) value in the
1903 		 * filter struture is taken to mean a wildcard.
1904 		 *
1905 		 * check that we are working for the right interface
1906 		 */
1907 #ifdef	_KERNEL
1908 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1909 			continue;
1910 #else
1911 		if (opts & (OPT_VERBOSE|OPT_DEBUG))
1912 			printf("\n");
1913 		FR_VERBOSE(("%c", FR_ISSKIP(pass) ? 's' :
1914 				  FR_ISPASS(pass) ? 'p' :
1915 				  FR_ISACCOUNT(pass) ? 'A' :
1916 				  FR_ISAUTH(pass) ? 'a' :
1917 				  (pass & FR_NOMATCH) ? 'n' :'b'));
1918 		if (fr->fr_ifa && fr->fr_ifa != fin->fin_ifp)
1919 			continue;
1920 		FR_VERBOSE((":i"));
1921 #endif
1922 
1923 		switch (fr->fr_type)
1924 		{
1925 		case FR_T_IPF :
1926 		case FR_T_IPF|FR_T_BUILTIN :
1927 			if (fr_ipfcheck(fin, fr, portcmp))
1928 				continue;
1929 			break;
1930 #if defined(IPFILTER_BPF)
1931 		case FR_T_BPFOPC :
1932 		case FR_T_BPFOPC|FR_T_BUILTIN :
1933 		    {
1934 			u_char *mc;
1935 
1936 			if (*fin->fin_mp == NULL)
1937 				continue;
1938 			if (fin->fin_v != fr->fr_v)
1939 				continue;
1940 			mc = (u_char *)fin->fin_m;
1941 			if (!bpf_filter(fr->fr_data, mc, fin->fin_plen, 0))
1942 				continue;
1943 			break;
1944 		    }
1945 #endif
1946 		case FR_T_CALLFUNC|FR_T_BUILTIN :
1947 		    {
1948 			frentry_t *f;
1949 
1950 			f = (*fr->fr_func)(fin, &pass);
1951 			if (f != NULL)
1952 				fr = f;
1953 			else
1954 				continue;
1955 			break;
1956 		    }
1957 		default :
1958 			break;
1959 		}
1960 
1961 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
1962 			if (fin->fin_nattag == NULL)
1963 				continue;
1964 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) == 0)
1965 				continue;
1966 		}
1967 		FR_VERBOSE(("=%s.%d *", fr->fr_group, rulen));
1968 
1969 		passt = fr->fr_flags;
1970 
1971 		/*
1972 		 * Allowing a rule with the "keep state" flag set to match
1973 		 * packets that have been tagged "out of window" by the TCP
1974 		 * state tracking is foolish as the attempt to add a new
1975 		 * state entry to the table will fail.
1976 		 */
1977 		if ((passt & FR_KEEPSTATE) && (fin->fin_flx & FI_OOW))
1978 			continue;
1979 
1980 		/*
1981 		 * If the rule is a "call now" rule, then call the function
1982 		 * in the rule, if it exists and use the results from that.
1983 		 * If the function pointer is bad, just make like we ignore
1984 		 * it, except for increasing the hit counter.
1985 		 */
1986 		IPF_BUMP(fr->fr_hits);
1987 		fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
1988 		if ((passt & FR_CALLNOW) != 0) {
1989 			if ((fr->fr_func != NULL) &&
1990 			    (fr->fr_func != (ipfunc_t)-1)) {
1991 				frentry_t *frs;
1992 
1993 				frs = fin->fin_fr;
1994 				fin->fin_fr = fr;
1995 				fr = (*fr->fr_func)(fin, &passt);
1996 				if (fr == NULL) {
1997 					fin->fin_fr = frs;
1998 					continue;
1999 				}
2000 				passt = fr->fr_flags;
2001 				fin->fin_fr = fr;
2002 			}
2003 		} else {
2004 			fin->fin_fr = fr;
2005 		}
2006 
2007 #ifdef  IPFILTER_LOG
2008 		/*
2009 		 * Just log this packet...
2010 		 */
2011 		if ((passt & FR_LOGMASK) == FR_LOG) {
2012 			if (ipflog(fin, passt) == -1) {
2013 				if (passt & FR_LOGORBLOCK) {
2014 					passt &= ~FR_CMDMASK;
2015 					passt |= FR_BLOCK|FR_QUICK;
2016 				}
2017 				IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_skip);
2018 			}
2019 			IPF_BUMP(ifs->ifs_frstats[fin->fin_out].fr_pkl);
2020 			logged = 1;
2021 		}
2022 #endif /* IPFILTER_LOG */
2023 		passo = pass;
2024 		if (FR_ISSKIP(passt))
2025 			skip = fr->fr_arg;
2026 		else if ((passt & FR_LOGMASK) != FR_LOG)
2027 			pass = passt;
2028 		if (passt & (FR_RETICMP|FR_FAKEICMP))
2029 			fin->fin_icode = fr->fr_icode;
2030 		FR_DEBUG(("pass %#x\n", pass));
2031 		fin->fin_rule = rulen;
2032 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2033 		if (fr->fr_grp != NULL) {
2034 			fin->fin_fr = *fr->fr_grp;
2035 			pass = fr_scanlist(fin, pass);
2036 			if (fin->fin_fr == NULL) {
2037 				fin->fin_rule = rulen;
2038 				(void) strncpy(fin->fin_group, fr->fr_group,
2039 					       FR_GROUPLEN);
2040 				fin->fin_fr = fr;
2041 			}
2042 			if (fin->fin_flx & FI_DONTCACHE)
2043 				logged = 1;
2044 		}
2045 
2046 		if (pass & FR_QUICK) {
2047 			/*
2048 			 * Finally, if we've asked to track state for this
2049 			 * packet, set it up.  Add state for "quick" rules
2050 			 * here so that if the action fails we can consider
2051 			 * the rule to "not match" and keep on processing
2052 			 * filter rules.
2053 			 */
2054 			if ((pass & FR_KEEPSTATE) &&
2055 			    !(fin->fin_flx & FI_STATE)) {
2056 				int out = fin->fin_out;
2057 
2058 				if (fr_addstate(fin, NULL, 0) != NULL) {
2059 					IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2060 				} else {
2061 					IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2062 					pass = passo;
2063 					continue;
2064 				}
2065 			}
2066 			break;
2067 		}
2068 	}
2069 	if (logged)
2070 		fin->fin_flx |= FI_DONTCACHE;
2071 	fin->fin_depth--;
2072 	return pass;
2073 }
2074 
2075 
2076 /* ------------------------------------------------------------------------ */
2077 /* Function:    fr_acctpkt                                                  */
2078 /* Returns:     frentry_t* - always returns NULL                            */
2079 /* Parameters:  fin(I) - pointer to packet information                      */
2080 /*              passp(IO) - pointer to current/new filter decision (unused) */
2081 /*                                                                          */
2082 /* Checks a packet against accounting rules, if there are any for the given */
2083 /* IP protocol version.                                                     */
2084 /*                                                                          */
2085 /* N.B.: this function returns NULL to match the prototype used by other    */
2086 /* functions called from the IPFilter "mainline" in fr_check().             */
2087 /* ------------------------------------------------------------------------ */
fr_acctpkt(fin,passp)2088 frentry_t *fr_acctpkt(fin, passp)
2089 fr_info_t *fin;
2090 u_32_t *passp;
2091 {
2092 	char group[FR_GROUPLEN];
2093 	frentry_t *fr, *frsave;
2094 	u_32_t pass, rulen;
2095 	ipf_stack_t *ifs = fin->fin_ifs;
2096 
2097 	passp = passp;
2098 #ifdef	USE_INET6
2099 	if (fin->fin_v == 6)
2100 		fr = ifs->ifs_ipacct6[fin->fin_out][ifs->ifs_fr_active];
2101 	else
2102 #endif
2103 		fr = ifs->ifs_ipacct[fin->fin_out][ifs->ifs_fr_active];
2104 
2105 	if (fr != NULL) {
2106 		frsave = fin->fin_fr;
2107 		bcopy(fin->fin_group, group, FR_GROUPLEN);
2108 		rulen = fin->fin_rule;
2109 		fin->fin_fr = fr;
2110 		pass = fr_scanlist(fin, FR_NOMATCH);
2111 		if (FR_ISACCOUNT(pass)) {
2112 			IPF_BUMP(ifs->ifs_frstats[0].fr_acct);
2113 		}
2114 		fin->fin_fr = frsave;
2115 		bcopy(group, fin->fin_group, FR_GROUPLEN);
2116 		fin->fin_rule = rulen;
2117 	}
2118 	return NULL;
2119 }
2120 
2121 
2122 /* ------------------------------------------------------------------------ */
2123 /* Function:    fr_firewall                                                 */
2124 /* Returns:     frentry_t* - returns pointer to matched rule, if no matches */
2125 /*                           were found, returns NULL.                      */
2126 /* Parameters:  fin(I) - pointer to packet information                      */
2127 /*              passp(IO) - pointer to current/new filter decision (unused) */
2128 /*                                                                          */
2129 /* Applies an appropriate set of firewall rules to the packet, to see if    */
2130 /* there are any matches.  The first check is to see if a match can be seen */
2131 /* in the cache.  If not, then search an appropriate list of rules.  Once a */
2132 /* matching rule is found, take any appropriate actions as defined by the   */
2133 /* rule - except logging.                                                   */
2134 /* ------------------------------------------------------------------------ */
fr_firewall(fin,passp)2135 static frentry_t *fr_firewall(fin, passp)
2136 fr_info_t *fin;
2137 u_32_t *passp;
2138 {
2139 	frentry_t *fr;
2140 	fr_info_t *fc;
2141 	u_32_t pass;
2142 	int out;
2143 	ipf_stack_t *ifs = fin->fin_ifs;
2144 
2145 	out = fin->fin_out;
2146 	pass = *passp;
2147 
2148 #ifdef	USE_INET6
2149 	if (fin->fin_v == 6)
2150 		fin->fin_fr = ifs->ifs_ipfilter6[out][ifs->ifs_fr_active];
2151 	else
2152 #endif
2153 		fin->fin_fr = ifs->ifs_ipfilter[out][ifs->ifs_fr_active];
2154 
2155 	/*
2156 	 * If there are no rules loaded skip all checks and return.
2157 	 */
2158 	if (fin->fin_fr == NULL) {
2159 
2160 		if ((pass & FR_NOMATCH)) {
2161 			IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2162 		}
2163 
2164 		return (NULL);
2165 	}
2166 
2167 	fc = &ifs->ifs_frcache[out][CACHE_HASH(fin)];
2168 	READ_ENTER(&ifs->ifs_ipf_frcache);
2169 	if (!bcmp((char *)fin, (char *)fc, FI_CSIZE)) {
2170 		/*
2171 		 * copy cached data so we can unlock the mutexes earlier.
2172 		 */
2173 		bcopy((char *)fc, (char *)fin, FI_COPYSIZE);
2174 		RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2175 		IPF_BUMP(ifs->ifs_frstats[out].fr_chit);
2176 
2177 		if ((fr = fin->fin_fr) != NULL) {
2178 			IPF_BUMP(fr->fr_hits);
2179 			fr->fr_bytes += (U_QUAD_T)fin->fin_plen;
2180 			pass = fr->fr_flags;
2181 		}
2182 	} else {
2183 		RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2184 
2185 		pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2186 
2187 		if (((pass & FR_KEEPSTATE) == 0) &&
2188 		    ((fin->fin_flx & FI_DONTCACHE) == 0)) {
2189 			WRITE_ENTER(&ifs->ifs_ipf_frcache);
2190 			bcopy((char *)fin, (char *)fc, FI_COPYSIZE);
2191 			RWLOCK_EXIT(&ifs->ifs_ipf_frcache);
2192 		}
2193 
2194 		fr = fin->fin_fr;
2195 	}
2196 
2197 	if ((pass & FR_NOMATCH)) {
2198 		IPF_BUMP(ifs->ifs_frstats[out].fr_nom);
2199 	}
2200 
2201 	/*
2202 	 * Apply packets per second rate-limiting to a rule as required.
2203 	 */
2204 	if ((fr != NULL) && (fr->fr_pps != 0) &&
2205 	    !ppsratecheck(&fr->fr_lastpkt, &fr->fr_curpps, fr->fr_pps)) {
2206 		pass &= ~(FR_CMDMASK|FR_DUP|FR_RETICMP|FR_RETRST);
2207 		pass |= FR_BLOCK;
2208 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppshit);
2209 	}
2210 
2211 	/*
2212 	 * If we fail to add a packet to the authorization queue, then we
2213 	 * drop the packet later.  However, if it was added then pretend
2214 	 * we've dropped it already.
2215 	 */
2216 	if (FR_ISAUTH(pass)) {
2217 		if (fr_newauth(fin->fin_m, fin) != 0) {
2218 #ifdef	_KERNEL
2219 			fin->fin_m = *fin->fin_mp = NULL;
2220 #else
2221 			;
2222 #endif
2223 			fin->fin_error = 0;
2224 		} else
2225 			fin->fin_error = ENOSPC;
2226 	}
2227 
2228 	if ((fr != NULL) && (fr->fr_func != NULL) &&
2229 	    (fr->fr_func != (ipfunc_t)-1) && !(pass & FR_CALLNOW))
2230 		(void) (*fr->fr_func)(fin, &pass);
2231 
2232 	/*
2233 	 * If a rule is a pre-auth rule, check again in the list of rules
2234 	 * loaded for authenticated use.  It does not particulary matter
2235 	 * if this search fails because a "preauth" result, from a rule,
2236 	 * is treated as "not a pass", hence the packet is blocked.
2237 	 */
2238 	if (FR_ISPREAUTH(pass)) {
2239 		if ((fin->fin_fr = ifs->ifs_ipauth) != NULL)
2240 			pass = fr_scanlist(fin, ifs->ifs_fr_pass);
2241 	}
2242 
2243 	/*
2244 	 * If the rule has "keep frag" and the packet is actually a fragment,
2245 	 * then create a fragment state entry.
2246 	 */
2247 	if ((pass & (FR_KEEPFRAG|FR_KEEPSTATE)) == FR_KEEPFRAG) {
2248 		if (fin->fin_flx & FI_FRAG) {
2249 			if (fr_newfrag(fin, pass) == -1) {
2250 				IPF_BUMP(ifs->ifs_frstats[out].fr_bnfr);
2251 			} else {
2252 				IPF_BUMP(ifs->ifs_frstats[out].fr_nfr);
2253 			}
2254 		} else {
2255 			IPF_BUMP(ifs->ifs_frstats[out].fr_cfr);
2256 		}
2257 	}
2258 
2259 	/*
2260 	 * Finally, if we've asked to track state for this packet, set it up.
2261 	 */
2262 	if ((pass & FR_KEEPSTATE) && !(fin->fin_flx & FI_STATE)) {
2263 		if (fr_addstate(fin, NULL, 0) != NULL) {
2264 			IPF_BUMP(ifs->ifs_frstats[out].fr_ads);
2265 		} else {
2266 			IPF_BUMP(ifs->ifs_frstats[out].fr_bads);
2267 			if (FR_ISPASS(pass)) {
2268 				pass &= ~FR_CMDMASK;
2269 				pass |= FR_BLOCK;
2270 			}
2271 		}
2272 	}
2273 
2274 	fr = fin->fin_fr;
2275 
2276 	if (passp != NULL)
2277 		*passp = pass;
2278 
2279 	return fr;
2280 }
2281 
2282 /* ------------------------------------------------------------------------ */
2283 /* Function:    fr_check                                                    */
2284 /* Returns:     int -  0 == packet allowed through,                         */
2285 /*              User space:                                                 */
2286 /*                    -1 == packet blocked                                  */
2287 /*                     1 == packet not matched                              */
2288 /*                    -2 == requires authentication                         */
2289 /*              Kernel:                                                     */
2290 /*                   > 0 == filter error # for packet                       */
2291 /* Parameters: ip(I)   - pointer to start of IPv4/6 packet                  */
2292 /*             hlen(I) - length of header                                   */
2293 /*             ifp(I)  - pointer to interface this packet is on             */
2294 /*             out(I)  - 0 == packet going in, 1 == packet going out        */
2295 /*             mp(IO)  - pointer to caller's buffer pointer that holds this */
2296 /*                       IP packet.                                         */
2297 /* Solaris & HP-UX ONLY :                                                   */
2298 /*             qpi(I)  - pointer to STREAMS queue information for this      */
2299 /*                       interface & direction.                             */
2300 /*                                                                          */
2301 /* fr_check() is the master function for all IPFilter packet processing.    */
2302 /* It orchestrates: Network Address Translation (NAT), checking for packet  */
2303 /* authorisation (or pre-authorisation), presence of related state info.,   */
2304 /* generating log entries, IP packet accounting, routing of packets as      */
2305 /* directed by firewall rules and of course whether or not to allow the     */
2306 /* packet to be further processed by the kernel.                            */
2307 /*                                                                          */
2308 /* For packets blocked, the contents of "mp" will be NULL'd and the buffer  */
2309 /* freed.  Packets passed may be returned with the pointer pointed to by    */
2310 /* by "mp" changed to a new buffer.                                         */
2311 /* ------------------------------------------------------------------------ */
fr_check(ip,hlen,ifp,out,qif,mp,ifs)2312 int fr_check(ip, hlen, ifp, out
2313 #if defined(_KERNEL) && defined(MENTAT)
2314 , qif, mp, ifs)
2315 void *qif;
2316 #else
2317 , mp, ifs)
2318 #endif
2319 mb_t **mp;
2320 ip_t *ip;
2321 int hlen;
2322 void *ifp;
2323 int out;
2324 ipf_stack_t *ifs;
2325 {
2326 	/*
2327 	 * The above really sucks, but short of writing a diff
2328 	 */
2329 	fr_info_t frinfo;
2330 	fr_info_t *fin = &frinfo;
2331 	u_32_t pass;
2332 	frentry_t *fr = NULL;
2333 	int v = IP_V(ip);
2334 	mb_t *mc = NULL;
2335 	mb_t *m;
2336 #ifdef USE_INET6
2337 	ip6_t *ip6;
2338 #endif
2339 #ifdef	_KERNEL
2340 # ifdef MENTAT
2341 	qpktinfo_t *qpi = qif;
2342 #endif
2343 #endif
2344 
2345 	SPL_INT(s);
2346 	pass = ifs->ifs_fr_pass;
2347 
2348 	/*
2349 	 * The first part of fr_check() deals with making sure that what goes
2350 	 * into the filtering engine makes some sense.  Information about the
2351 	 * the packet is distilled, collected into a fr_info_t structure and
2352 	 * the an attempt to ensure the buffer the packet is in is big enough
2353 	 * to hold all the required packet headers.
2354 	 */
2355 #ifdef	_KERNEL
2356 # ifdef MENTAT
2357 	if (!OK_32PTR(ip))
2358 		return 2;
2359 # endif
2360 
2361 
2362 	if (ifs->ifs_fr_running <= 0) {
2363 		return 0;
2364 	}
2365 
2366 	bzero((char *)fin, sizeof(*fin));
2367 
2368 # ifdef MENTAT
2369 	fin->fin_flx = qpi->qpi_flags & (FI_NOCKSUM|FI_MBCAST|FI_MULTICAST|
2370 					 FI_BROADCAST);
2371 	m = qpi->qpi_m;
2372 	fin->fin_qfm = m;
2373 	fin->fin_qpi = qpi;
2374 # else /* MENTAT */
2375 
2376 	m = *mp;
2377 
2378 #  if defined(M_MCAST)
2379 	if ((m->m_flags & M_MCAST) != 0)
2380 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2381 #  endif
2382 #  if defined(M_MLOOP)
2383 	if ((m->m_flags & M_MLOOP) != 0)
2384 		fin->fin_flx |= FI_MBCAST|FI_MULTICAST;
2385 #  endif
2386 #  if defined(M_BCAST)
2387 	if ((m->m_flags & M_BCAST) != 0)
2388 		fin->fin_flx |= FI_MBCAST|FI_BROADCAST;
2389 #  endif
2390 #  ifdef M_CANFASTFWD
2391 	/*
2392 	 * XXX For now, IP Filter and fast-forwarding of cached flows
2393 	 * XXX are mutually exclusive.  Eventually, IP Filter should
2394 	 * XXX get a "can-fast-forward" filter rule.
2395 	 */
2396 	m->m_flags &= ~M_CANFASTFWD;
2397 #  endif /* M_CANFASTFWD */
2398 #  ifdef CSUM_DELAY_DATA
2399 	/*
2400 	 * disable delayed checksums.
2401 	 */
2402 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2403 		in_delayed_cksum(m);
2404 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2405 	}
2406 #  endif /* CSUM_DELAY_DATA */
2407 # endif /* MENTAT */
2408 #else
2409 
2410 	bzero((char *)fin, sizeof(*fin));
2411 	m = *mp;
2412 #endif /* _KERNEL */
2413 
2414 	fin->fin_v = v;
2415 	fin->fin_m = m;
2416 	fin->fin_ip = ip;
2417 	fin->fin_mp = mp;
2418 	fin->fin_out = out;
2419 	fin->fin_ifp = ifp;
2420 	fin->fin_error = ENETUNREACH;
2421 	fin->fin_hlen = (u_short)hlen;
2422 	fin->fin_dp = (char *)ip + hlen;
2423 	fin->fin_ipoff = (char *)ip - MTOD(m, char *);
2424 	fin->fin_ifs = ifs;
2425 
2426 	SPL_NET(s);
2427 
2428 #ifdef	USE_INET6
2429 	if (v == 6) {
2430 		IPF_BUMP(ifs->ifs_frstats[out].fr_ipv6);
2431 		/*
2432 		 * Jumbo grams are quite likely too big for internal buffer
2433 		 * structures to handle comfortably, for now, so just drop
2434 		 * them.
2435 		 */
2436 		ip6 = (ip6_t *)ip;
2437 		fin->fin_plen = ntohs(ip6->ip6_plen);
2438 		if (fin->fin_plen == 0) {
2439 			READ_ENTER(&ifs->ifs_ipf_mutex);
2440 			pass = FR_BLOCK|FR_NOMATCH;
2441 			goto filtered;
2442 		}
2443 		fin->fin_plen += sizeof(ip6_t);
2444 	} else
2445 #endif
2446 	{
2447 #if (OpenBSD >= 200311) && defined(_KERNEL)
2448 		ip->ip_len = ntohs(ip->ip_len);
2449 		ip->ip_off = ntohs(ip->ip_off);
2450 #endif
2451 		fin->fin_plen = ip->ip_len;
2452 	}
2453 
2454 	if (fr_makefrip(hlen, ip, fin) == -1) {
2455 		READ_ENTER(&ifs->ifs_ipf_mutex);
2456 		pass = FR_BLOCK;
2457 		goto filtered;
2458 	}
2459 
2460 	/*
2461 	 * For at least IPv6 packets, if a m_pullup() fails then this pointer
2462 	 * becomes NULL and so we have no packet to free.
2463 	 */
2464 	if (*fin->fin_mp == NULL)
2465 		goto finished;
2466 
2467 	if (!out) {
2468 		if (v == 4) {
2469 #ifdef _KERNEL
2470 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2471 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2472 				fin->fin_flx |= FI_BADSRC;
2473 			}
2474 #endif
2475 			if (fin->fin_ip->ip_ttl < ifs->ifs_fr_minttl) {
2476 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2477 				fin->fin_flx |= FI_LOWTTL;
2478 			}
2479 		}
2480 #ifdef USE_INET6
2481 		else  if (v == 6) {
2482 			ip6 = (ip6_t *)ip;
2483 #ifdef _KERNEL
2484 			if (ifs->ifs_fr_chksrc && !fr_verifysrc(fin)) {
2485 				IPF_BUMP(ifs->ifs_frstats[0].fr_badsrc);
2486 				fin->fin_flx |= FI_BADSRC;
2487 			}
2488 #endif
2489 			if (ip6->ip6_hlim < ifs->ifs_fr_minttl) {
2490 				IPF_BUMP(ifs->ifs_frstats[0].fr_badttl);
2491 				fin->fin_flx |= FI_LOWTTL;
2492 			}
2493 		}
2494 #endif
2495 	}
2496 
2497 	if (fin->fin_flx & FI_SHORT) {
2498 		IPF_BUMP(ifs->ifs_frstats[out].fr_short);
2499 	}
2500 
2501 	READ_ENTER(&ifs->ifs_ipf_mutex);
2502 
2503 	/*
2504 	 * Check auth now.  This, combined with the check below to see if apass
2505 	 * is 0 is to ensure that we don't count the packet twice, which can
2506 	 * otherwise occur when we reprocess it.  As it is, we only count it
2507 	 * after it has no auth. table matchup.  This also stops NAT from
2508 	 * occuring until after the packet has been auth'd.
2509 	 */
2510 	fr = fr_checkauth(fin, &pass);
2511 	if (!out) {
2512 		switch (fin->fin_v)
2513 		{
2514 		case 4 :
2515 			if (fr_checknatin(fin, &pass) == -1) {
2516 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2517 				goto finished;
2518 			}
2519 			break;
2520 #ifdef	USE_INET6
2521 		case 6 :
2522 			if (fr_checknat6in(fin, &pass) == -1) {
2523 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2524 				goto finished;
2525 			}
2526 			break;
2527 #endif
2528 		default :
2529 			break;
2530 		}
2531 	}
2532 	if (!out)
2533 		(void) fr_acctpkt(fin, NULL);
2534 
2535 	if (fr == NULL)
2536 		if ((fin->fin_flx & (FI_FRAG|FI_BAD)) == FI_FRAG)
2537 			fr = fr_knownfrag(fin, &pass);
2538 	if (fr == NULL)
2539 		fr = fr_checkstate(fin, &pass);
2540 
2541 	if ((pass & FR_NOMATCH) || (fr == NULL))
2542 		fr = fr_firewall(fin, &pass);
2543 
2544 	fin->fin_fr = fr;
2545 
2546 	/*
2547 	 * Only count/translate packets which will be passed on, out the
2548 	 * interface.
2549 	 */
2550 	if (out && FR_ISPASS(pass)) {
2551 		(void) fr_acctpkt(fin, NULL);
2552 
2553 		switch (fin->fin_v)
2554 		{
2555 		case 4 :
2556 			if (fr_checknatout(fin, &pass) == -1) {
2557 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2558 				goto finished;
2559 			}
2560 			break;
2561 #ifdef	USE_INET6
2562 		case 6 :
2563 			if (fr_checknat6out(fin, &pass) == -1) {
2564 				RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2565 				goto finished;
2566 			}
2567 			break;
2568 #endif
2569 		default :
2570 			break;
2571 		}
2572 
2573 		if ((ifs->ifs_fr_update_ipid != 0) && (v == 4)) {
2574 			if (fr_updateipid(fin) == -1) {
2575 				IPF_BUMP(ifs->ifs_frstats[1].fr_ipud);
2576 				pass &= ~FR_CMDMASK;
2577 				pass |= FR_BLOCK;
2578 			} else {
2579 				IPF_BUMP(ifs->ifs_frstats[0].fr_ipud);
2580 			}
2581 		}
2582 	}
2583 
2584 #ifdef	IPFILTER_LOG
2585 	if ((ifs->ifs_fr_flags & FF_LOGGING) || (pass & FR_LOGMASK)) {
2586 		(void) fr_dolog(fin, &pass);
2587 	}
2588 #endif
2589 
2590 	/*
2591 	 * The FI_STATE flag is cleared here so that calling fr_checkstate
2592 	 * will work when called from inside of fr_fastroute.  Although
2593 	 * there is a similar flag, FI_NATED, for NAT, it does have the same
2594 	 * impact on code execution.
2595 	 */
2596 	fin->fin_flx &= ~FI_STATE;
2597 
2598 	/*
2599 	 * Only allow FR_DUP to work if a rule matched - it makes no sense to
2600 	 * set FR_DUP as a "default" as there are no instructions about where
2601 	 * to send the packet.  Use fin_m here because it may have changed
2602 	 * (without an update of 'm') in prior processing.
2603 	 */
2604 	if ((fr != NULL) && (pass & FR_DUP)) {
2605 		mc = M_DUPLICATE(fin->fin_m);
2606 #ifdef _KERNEL
2607 		mc->b_rptr += fin->fin_ipoff;
2608 #endif
2609 	}
2610 
2611 	/*
2612 	 * We don't want to send RST for packets, which are going to be
2613 	 * dropped, just because they don't fit into TCP window. Those packets
2614 	 * will be dropped silently. In other words, we want to drop packet,
2615 	 * while keeping session alive.
2616 	 */
2617 	if ((pass & (FR_RETRST|FR_RETICMP)) && ((fin->fin_flx & FI_OOW) == 0)) {
2618 		/*
2619 		 * Should we return an ICMP packet to indicate error
2620 		 * status passing through the packet filter ?
2621 		 * WARNING: ICMP error packets AND TCP RST packets should
2622 		 * ONLY be sent in repsonse to incoming packets.  Sending them
2623 		 * in response to outbound packets can result in a panic on
2624 		 * some operating systems.
2625 		 */
2626 		if (!out) {
2627 			if (pass & FR_RETICMP) {
2628 				int dst;
2629 
2630 				if ((pass & FR_RETMASK) == FR_FAKEICMP)
2631 					dst = 1;
2632 				else
2633 					dst = 0;
2634 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2635 				/*
2636 				 * Assume it's possible to enter insane rule:
2637 				 * 	pass return-icmp in proto udp ...
2638 				 * then we have no other option than to forward
2639 				 * packet on loopback and give up any attempt
2640 				 * to create a fake response.
2641 				 */
2642 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2643 				    FR_ISBLOCK(pass)) {
2644 
2645 					if (fr_make_icmp(fin) == 0) {
2646 						IPF_BUMP(
2647 						ifs->ifs_frstats[out].fr_ret);
2648 					}
2649 					/*
2650 					 * we drop packet silently in case we
2651 					 * failed assemble fake response for it
2652 					 */
2653 					else if (*mp != NULL) {
2654 						FREE_MB_T(*mp);
2655 						m = *mp = NULL;
2656 					}
2657 
2658 					IPF_BUMP(
2659 					    ifs->ifs_frstats[out].fr_block);
2660 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2661 
2662 					return (0);
2663 				}
2664 #endif	/* _KERNEL && SOLARIS2 >= 10 */
2665 
2666 				(void) fr_send_icmp_err(ICMP_UNREACH, fin, dst);
2667 				IPF_BUMP(ifs->ifs_frstats[out].fr_ret);
2668 
2669 			} else if (((pass & FR_RETMASK) == FR_RETRST) &&
2670 				   !(fin->fin_flx & FI_SHORT)) {
2671 
2672 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2673 				/*
2674 				 * Assume it's possible to enter insane rule:
2675 				 * 	pass return-rst in proto tcp ...
2676 				 * then we have no other option than to forward
2677 				 * packet on loopback and give up any attempt
2678 				 * to create a fake response.
2679 				 */
2680 				if (IPF_IS_LOOPBACK(qpi->qpi_flags) &&
2681 				    FR_ISBLOCK(pass)) {
2682 					if (fr_make_rst(fin) == 0) {
2683 						IPF_BUMP(
2684 						ifs->ifs_frstats[out].fr_ret);
2685 					}
2686 					else if (mp != NULL) {
2687 					/*
2688 					 * we drop packet silently in case we
2689 					 * failed assemble fake response for it
2690 					 */
2691 						FREE_MB_T(*mp);
2692 						m = *mp = NULL;
2693 					}
2694 
2695 					IPF_BUMP(
2696 					    ifs->ifs_frstats[out].fr_block);
2697 					RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2698 
2699 					return (0);
2700 				 }
2701 #endif /* _KERNEL && _SOLARIS2 >= 10 */
2702 				if (fr_send_reset(fin) == 0) {
2703 					IPF_BUMP(ifs->ifs_frstats[1].fr_ret);
2704 				}
2705 			}
2706 		} else {
2707 			if (pass & FR_RETRST)
2708 				fin->fin_error = ECONNRESET;
2709 		}
2710 	}
2711 
2712 	/*
2713 	 * If we didn't drop off the bottom of the list of rules (and thus
2714 	 * the 'current' rule fr is not NULL), then we may have some extra
2715 	 * instructions about what to do with a packet.
2716 	 * Once we're finished return to our caller, freeing the packet if
2717 	 * we are dropping it (* BSD ONLY *).
2718 	 * Reassign m from fin_m as we may have a new buffer, now.
2719 	 */
2720 filtered:
2721 	m = fin->fin_m;
2722 
2723 	if (fr != NULL) {
2724 		frdest_t *fdp;
2725 
2726 		fdp = &fr->fr_tifs[fin->fin_rev];
2727 
2728 		if (!out && (pass & FR_FASTROUTE)) {
2729 			/*
2730 			 * For fastroute rule, no destioation interface defined
2731 			 * so pass NULL as the frdest_t parameter
2732 			 */
2733 			(void) fr_fastroute(m, mp, fin, NULL);
2734 			m = *mp = NULL;
2735 		} else if ((fdp->fd_ifp != NULL) &&
2736 			   (fdp->fd_ifp != (struct ifnet *)-1)) {
2737 			/* this is for to rules: */
2738 			(void) fr_fastroute(m, mp, fin, fdp);
2739 			m = *mp = NULL;
2740 		}
2741 
2742 		/*
2743 		 * Send a duplicated packet.
2744 		 */
2745 		if (mc != NULL) {
2746 #if defined(_KERNEL) && (SOLARIS2 >= 10)
2747 			/*
2748 			 * We are going to compute chksum for copies of loopback packets
2749 			 * only. IP stack does not compute chksums at all for loopback
2750 			 * packets. We want to get it fixed in their copies, since those
2751 			 * are going to be sent to network.
2752 			 */
2753 			if (IPF_IS_LOOPBACK(qpi->qpi_flags))
2754 				fr_calc_chksum(fin, mc);
2755 #endif
2756 			(void) fr_fastroute(mc, &mc, fin, &fr->fr_dif);
2757 		}
2758 	}
2759 
2760 	if (FR_ISBLOCK(pass) && (fin->fin_flx & FI_NEWNAT))
2761 		nat_uncreate(fin);
2762 
2763 	/*
2764 	 * This late because the likes of fr_fastroute() use fin_fr.
2765 	 */
2766 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
2767 
2768 finished:
2769 	if (!FR_ISPASS(pass)) {
2770 		IPF_BUMP(ifs->ifs_frstats[out].fr_block);
2771 		if (*mp != NULL) {
2772 			FREE_MB_T(*mp);
2773 			m = *mp = NULL;
2774 		}
2775 	} else {
2776 		IPF_BUMP(ifs->ifs_frstats[out].fr_pass);
2777 #if defined(_KERNEL) && defined(__sgi)
2778 		if ((fin->fin_hbuf != NULL) &&
2779 		    (mtod(fin->fin_m, struct ip *) != fin->fin_ip)) {
2780 			COPYBACK(m, 0, fin->fin_plen, fin->fin_hbuf);
2781 		}
2782 #endif
2783 	}
2784 
2785 	SPL_X(s);
2786 
2787 #ifdef _KERNEL
2788 # if OpenBSD >= 200311
2789 	if (FR_ISPASS(pass) && (v == 4)) {
2790 		ip = fin->fin_ip;
2791 		ip->ip_len = ntohs(ip->ip_len);
2792 		ip->ip_off = ntohs(ip->ip_off);
2793 	}
2794 # endif
2795 	return (FR_ISPASS(pass)) ? 0 : fin->fin_error;
2796 #else /* _KERNEL */
2797 	FR_VERBOSE(("fin_flx %#x pass %#x ", fin->fin_flx, pass));
2798 	if ((pass & FR_NOMATCH) != 0)
2799 		return 1;
2800 
2801 	if ((pass & FR_RETMASK) != 0)
2802 		switch (pass & FR_RETMASK)
2803 		{
2804 		case FR_RETRST :
2805 			return 3;
2806 		case FR_RETICMP :
2807 			return 4;
2808 		case FR_FAKEICMP :
2809 			return 5;
2810 		}
2811 
2812 	switch (pass & FR_CMDMASK)
2813 	{
2814 	case FR_PASS :
2815 		return 0;
2816 	case FR_BLOCK :
2817 		return -1;
2818 	case FR_AUTH :
2819 		return -2;
2820 	case FR_ACCOUNT :
2821 		return -3;
2822 	case FR_PREAUTH :
2823 		return -4;
2824 	}
2825 	return 2;
2826 #endif /* _KERNEL */
2827 }
2828 
2829 
2830 #ifdef	IPFILTER_LOG
2831 /* ------------------------------------------------------------------------ */
2832 /* Function:    fr_dolog                                                    */
2833 /* Returns:     frentry_t* - returns contents of fin_fr (no change made)    */
2834 /* Parameters:  fin(I) - pointer to packet information                      */
2835 /*              passp(IO) - pointer to current/new filter decision (unused) */
2836 /*                                                                          */
2837 /* Checks flags set to see how a packet should be logged, if it is to be    */
2838 /* logged.  Adjust statistics based on its success or not.                  */
2839 /* ------------------------------------------------------------------------ */
fr_dolog(fin,passp)2840 frentry_t *fr_dolog(fin, passp)
2841 fr_info_t *fin;
2842 u_32_t *passp;
2843 {
2844 	u_32_t pass;
2845 	int out;
2846 	ipf_stack_t *ifs = fin->fin_ifs;
2847 
2848 	out = fin->fin_out;
2849 	pass = *passp;
2850 
2851 	if ((ifs->ifs_fr_flags & FF_LOGNOMATCH) && (pass & FR_NOMATCH)) {
2852 		pass |= FF_LOGNOMATCH;
2853 		IPF_BUMP(ifs->ifs_frstats[out].fr_npkl);
2854 		goto logit;
2855 	} else if (((pass & FR_LOGMASK) == FR_LOGP) ||
2856 	    (FR_ISPASS(pass) && (ifs->ifs_fr_flags & FF_LOGPASS))) {
2857 		if ((pass & FR_LOGMASK) != FR_LOGP)
2858 			pass |= FF_LOGPASS;
2859 		IPF_BUMP(ifs->ifs_frstats[out].fr_ppkl);
2860 		goto logit;
2861 	} else if (((pass & FR_LOGMASK) == FR_LOGB) ||
2862 		   (FR_ISBLOCK(pass) && (ifs->ifs_fr_flags & FF_LOGBLOCK))) {
2863 		if ((pass & FR_LOGMASK) != FR_LOGB)
2864 			pass |= FF_LOGBLOCK;
2865 		IPF_BUMP(ifs->ifs_frstats[out].fr_bpkl);
2866 logit:
2867 		if (ipflog(fin, pass) == -1) {
2868 			IPF_BUMP(ifs->ifs_frstats[out].fr_skip);
2869 
2870 			/*
2871 			 * If the "or-block" option has been used then
2872 			 * block the packet if we failed to log it.
2873 			 */
2874 			if ((pass & FR_LOGORBLOCK) &&
2875 			    FR_ISPASS(pass)) {
2876 				pass &= ~FR_CMDMASK;
2877 				pass |= FR_BLOCK;
2878 			}
2879 		}
2880 		*passp = pass;
2881 	}
2882 
2883 	return fin->fin_fr;
2884 }
2885 #endif /* IPFILTER_LOG */
2886 
2887 
2888 /* ------------------------------------------------------------------------ */
2889 /* Function:    ipf_cksum                                                   */
2890 /* Returns:     u_short - IP header checksum                                */
2891 /* Parameters:  addr(I) - pointer to start of buffer to checksum            */
2892 /*              len(I)  - length of buffer in bytes                         */
2893 /*                                                                          */
2894 /* Calculate the two's complement 16 bit checksum of the buffer passed.     */
2895 /*                                                                          */
2896 /* N.B.: addr should be 16bit aligned.                                      */
2897 /* ------------------------------------------------------------------------ */
ipf_cksum(addr,len)2898 u_short ipf_cksum(addr, len)
2899 u_short *addr;
2900 int len;
2901 {
2902 	u_32_t sum = 0;
2903 
2904 	for (sum = 0; len > 1; len -= 2)
2905 		sum += *addr++;
2906 
2907 	/* mop up an odd byte, if necessary */
2908 	if (len == 1)
2909 		sum += *(u_char *)addr;
2910 
2911 	/*
2912 	 * add back carry outs from top 16 bits to low 16 bits
2913 	 */
2914 	sum = (sum >> 16) + (sum & 0xffff);	/* add hi 16 to low 16 */
2915 	sum += (sum >> 16);			/* add carry */
2916 	return (u_short)(~sum);
2917 }
2918 
2919 
2920 /* ------------------------------------------------------------------------ */
2921 /* Function:    fr_cksum                                                    */
2922 /* Returns:     u_short - layer 4 checksum                                  */
2923 /* Parameters:  m(I  )     - pointer to buffer holding packet               */
2924 /*              ip(I)      - pointer to IP header                           */
2925 /*              l4proto(I) - protocol to caclulate checksum for             */
2926 /*              l4hdr(I)   - pointer to layer 4 header                      */
2927 /*                                                                          */
2928 /* Calculates the TCP checksum for the packet held in "m", using the data   */
2929 /* in the IP header "ip" to seed it.                                        */
2930 /*                                                                          */
2931 /* NB: This function assumes we've pullup'd enough for all of the IP header */
2932 /* and the TCP header.  We also assume that data blocks aren't allocated in */
2933 /* odd sizes.                                                               */
2934 /*                                                                          */
2935 /* Expects ip_len to be in host byte order when called.                     */
2936 /* ------------------------------------------------------------------------ */
fr_cksum(m,ip,l4proto,l4hdr)2937 u_short fr_cksum(m, ip, l4proto, l4hdr)
2938 mb_t *m;
2939 ip_t *ip;
2940 int l4proto;
2941 void *l4hdr;
2942 {
2943 	u_short *sp, slen, sumsave, l4hlen, *csump;
2944 	u_int sum, sum2;
2945 	int hlen;
2946 #ifdef	USE_INET6
2947 	ip6_t *ip6;
2948 #endif
2949 
2950 	csump = NULL;
2951 	sumsave = 0;
2952 	l4hlen = 0;
2953 	sp = NULL;
2954 	slen = 0;
2955 	hlen = 0;
2956 	sum = 0;
2957 
2958 	/*
2959 	 * Add up IP Header portion
2960 	 */
2961 #ifdef	USE_INET6
2962 	if (IP_V(ip) == 4) {
2963 #endif
2964 		hlen = IP_HL(ip) << 2;
2965 		slen = ip->ip_len - hlen;
2966 		sum = htons((u_short)l4proto);
2967 		sum += htons(slen);
2968 		sp = (u_short *)&ip->ip_src;
2969 		sum += *sp++;	/* ip_src */
2970 		sum += *sp++;
2971 		sum += *sp++;	/* ip_dst */
2972 		sum += *sp++;
2973 #ifdef	USE_INET6
2974 	} else if (IP_V(ip) == 6) {
2975 		ip6 = (ip6_t *)ip;
2976 		hlen = sizeof(*ip6);
2977 		slen = ntohs(ip6->ip6_plen);
2978 		sum = htons((u_short)l4proto);
2979 		sum += htons(slen);
2980 		sp = (u_short *)&ip6->ip6_src;
2981 		sum += *sp++;	/* ip6_src */
2982 		sum += *sp++;
2983 		sum += *sp++;
2984 		sum += *sp++;
2985 		sum += *sp++;
2986 		sum += *sp++;
2987 		sum += *sp++;
2988 		sum += *sp++;
2989 		sum += *sp++;	/* ip6_dst */
2990 		sum += *sp++;
2991 		sum += *sp++;
2992 		sum += *sp++;
2993 		sum += *sp++;
2994 		sum += *sp++;
2995 		sum += *sp++;
2996 		sum += *sp++;
2997 	}
2998 #endif
2999 
3000 	switch (l4proto)
3001 	{
3002 	case IPPROTO_UDP :
3003 		csump = &((udphdr_t *)l4hdr)->uh_sum;
3004 		l4hlen = sizeof(udphdr_t);
3005 		break;
3006 
3007 	case IPPROTO_TCP :
3008 		csump = &((tcphdr_t *)l4hdr)->th_sum;
3009 		l4hlen = sizeof(tcphdr_t);
3010 		break;
3011 	case IPPROTO_ICMP :
3012 		csump = &((icmphdr_t *)l4hdr)->icmp_cksum;
3013 		l4hlen = 4;
3014 		sum = 0;
3015 		break;
3016 	default :
3017 		break;
3018 	}
3019 
3020 	if (csump != NULL) {
3021 		sumsave = *csump;
3022 		*csump = 0;
3023 	}
3024 
3025 	l4hlen = l4hlen;	/* LINT */
3026 
3027 #ifdef	_KERNEL
3028 # ifdef MENTAT
3029 	{
3030 	void *rp = m->b_rptr;
3031 
3032 	if ((unsigned char *)ip > m->b_rptr && (unsigned char *)ip < m->b_wptr)
3033 		m->b_rptr = (u_char *)ip;
3034 	sum2 = ip_cksum(m, hlen, sum);	/* hlen == offset */
3035 	m->b_rptr = rp;
3036 	sum2 = (sum2 & 0xffff) + (sum2 >> 16);
3037 	sum2 = ~sum2 & 0xffff;
3038 	}
3039 # else /* MENTAT */
3040 #  if defined(BSD) || defined(sun)
3041 #   if BSD >= 199103
3042 	m->m_data += hlen;
3043 #   else
3044 	m->m_off += hlen;
3045 #   endif
3046 	m->m_len -= hlen;
3047 	sum2 = in_cksum(m, slen);
3048 	m->m_len += hlen;
3049 #   if BSD >= 199103
3050 	m->m_data -= hlen;
3051 #   else
3052 	m->m_off -= hlen;
3053 #   endif
3054 	/*
3055 	 * Both sum and sum2 are partial sums, so combine them together.
3056 	 */
3057 	sum += ~sum2 & 0xffff;
3058 	while (sum > 0xffff)
3059 		sum = (sum & 0xffff) + (sum >> 16);
3060 	sum2 = ~sum & 0xffff;
3061 #  else /* defined(BSD) || defined(sun) */
3062 {
3063 	union {
3064 		u_char	c[2];
3065 		u_short	s;
3066 	} bytes;
3067 	u_short len = ip->ip_len;
3068 #   if defined(__sgi)
3069 	int add;
3070 #   endif
3071 
3072 	/*
3073 	 * Add up IP Header portion
3074 	 */
3075 	if (sp != (u_short *)l4hdr)
3076 		sp = (u_short *)l4hdr;
3077 
3078 	switch (l4proto)
3079 	{
3080 	case IPPROTO_UDP :
3081 		sum += *sp++;	/* sport */
3082 		sum += *sp++;	/* dport */
3083 		sum += *sp++;	/* udp length */
3084 		sum += *sp++;	/* checksum */
3085 		break;
3086 
3087 	case IPPROTO_TCP :
3088 		sum += *sp++;	/* sport */
3089 		sum += *sp++;	/* dport */
3090 		sum += *sp++;	/* seq */
3091 		sum += *sp++;
3092 		sum += *sp++;	/* ack */
3093 		sum += *sp++;
3094 		sum += *sp++;	/* off */
3095 		sum += *sp++;	/* win */
3096 		sum += *sp++;	/* checksum */
3097 		sum += *sp++;	/* urp */
3098 		break;
3099 	case IPPROTO_ICMP :
3100 		sum = *sp++;	/* type/code */
3101 		sum += *sp++;	/* checksum */
3102 		break;
3103 	}
3104 
3105 #   ifdef	__sgi
3106 	/*
3107 	 * In case we had to copy the IP & TCP header out of mbufs,
3108 	 * skip over the mbuf bits which are the header
3109 	 */
3110 	if ((caddr_t)ip != mtod(m, caddr_t)) {
3111 		hlen = (caddr_t)sp - (caddr_t)ip;
3112 		while (hlen) {
3113 			add = MIN(hlen, m->m_len);
3114 			sp = (u_short *)(mtod(m, caddr_t) + add);
3115 			hlen -= add;
3116 			if (add == m->m_len) {
3117 				m = m->m_next;
3118 				if (!hlen) {
3119 					if (!m)
3120 						break;
3121 					sp = mtod(m, u_short *);
3122 				}
3123 				PANIC((!m),("fr_cksum(1): not enough data"));
3124 			}
3125 		}
3126 	}
3127 #   endif
3128 
3129 	len -= (l4hlen + hlen);
3130 	if (len <= 0)
3131 		goto nodata;
3132 
3133 	while (len > 1) {
3134 		if (((caddr_t)sp - mtod(m, caddr_t)) >= m->m_len) {
3135 			m = m->m_next;
3136 			PANIC((!m),("fr_cksum(2): not enough data"));
3137 			sp = mtod(m, u_short *);
3138 		}
3139 		if (((caddr_t)(sp + 1) - mtod(m, caddr_t)) > m->m_len) {
3140 			bytes.c[0] = *(u_char *)sp;
3141 			m = m->m_next;
3142 			PANIC((!m),("fr_cksum(3): not enough data"));
3143 			sp = mtod(m, u_short *);
3144 			bytes.c[1] = *(u_char *)sp;
3145 			sum += bytes.s;
3146 			sp = (u_short *)((u_char *)sp + 1);
3147 		}
3148 		if ((u_long)sp & 1) {
3149 			bcopy((char *)sp++, (char *)&bytes.s, sizeof(bytes.s));
3150 			sum += bytes.s;
3151 		} else
3152 			sum += *sp++;
3153 		len -= 2;
3154 	}
3155 
3156 	if (len != 0)
3157 		sum += ntohs(*(u_char *)sp << 8);
3158 nodata:
3159 	while (sum > 0xffff)
3160 		sum = (sum & 0xffff) + (sum >> 16);
3161 	sum2 = (u_short)(~sum & 0xffff);
3162 }
3163 #  endif /*  defined(BSD) || defined(sun) */
3164 # endif /* MENTAT */
3165 #else /* _KERNEL */
3166 	for (; slen > 1; slen -= 2)
3167 	        sum += *sp++;
3168 	if (slen)
3169 		sum += ntohs(*(u_char *)sp << 8);
3170 	while (sum > 0xffff)
3171 		sum = (sum & 0xffff) + (sum >> 16);
3172 	sum2 = (u_short)(~sum & 0xffff);
3173 #endif /* _KERNEL */
3174 	if (csump != NULL)
3175 		*csump = sumsave;
3176 	return sum2;
3177 }
3178 
3179 
3180 #if defined(_KERNEL) && ( ((BSD < 199103) && !defined(MENTAT)) || \
3181     defined(__sgi) ) && !defined(linux) && !defined(_AIX51)
3182 /*
3183  * Copyright (c) 1982, 1986, 1988, 1991, 1993
3184  *	The Regents of the University of California.  All rights reserved.
3185  *
3186  * Redistribution and use in source and binary forms, with or without
3187  * modification, are permitted provided that the following conditions
3188  * are met:
3189  * 1. Redistributions of source code must retain the above copyright
3190  *    notice, this list of conditions and the following disclaimer.
3191  * 2. Redistributions in binary form must reproduce the above copyright
3192  *    notice, this list of conditions and the following disclaimer in the
3193  *    documentation and/or other materials provided with the distribution.
3194  * 3. Neither the name of the University nor the names of its contributors
3195  *    may be used to endorse or promote products derived from this software
3196  *    without specific prior written permission.
3197  *
3198  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
3199  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
3200  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
3201  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
3202  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
3203  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3204  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3205  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3206  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3207  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3208  * SUCH DAMAGE.
3209  *
3210  *	@(#)uipc_mbuf.c	8.2 (Berkeley) 1/4/94
3211  * $Id: fil.c,v 2.243.2.64 2005/08/13 05:19:59 darrenr Exp $
3212  */
3213 /*
3214  * Copy data from an mbuf chain starting "off" bytes from the beginning,
3215  * continuing for "len" bytes, into the indicated buffer.
3216  */
3217 void
m_copydata(m,off,len,cp)3218 m_copydata(m, off, len, cp)
3219 	mb_t *m;
3220 	int off;
3221 	int len;
3222 	caddr_t cp;
3223 {
3224 	unsigned count;
3225 
3226 	if (off < 0 || len < 0)
3227 		panic("m_copydata");
3228 	while (off > 0) {
3229 		if (m == 0)
3230 			panic("m_copydata");
3231 		if (off < m->m_len)
3232 			break;
3233 		off -= m->m_len;
3234 		m = m->m_next;
3235 	}
3236 	while (len > 0) {
3237 		if (m == 0)
3238 			panic("m_copydata");
3239 		count = MIN(m->m_len - off, len);
3240 		bcopy(mtod(m, caddr_t) + off, cp, count);
3241 		len -= count;
3242 		cp += count;
3243 		off = 0;
3244 		m = m->m_next;
3245 	}
3246 }
3247 
3248 
3249 /*
3250  * Copy data from a buffer back into the indicated mbuf chain,
3251  * starting "off" bytes from the beginning, extending the mbuf
3252  * chain if necessary.
3253  */
3254 void
m_copyback(m0,off,len,cp)3255 m_copyback(m0, off, len, cp)
3256 	struct	mbuf *m0;
3257 	int off;
3258 	int len;
3259 	caddr_t cp;
3260 {
3261 	int mlen;
3262 	struct mbuf *m = m0, *n;
3263 	int totlen = 0;
3264 
3265 	if (m0 == 0)
3266 		return;
3267 	while (off > (mlen = m->m_len)) {
3268 		off -= mlen;
3269 		totlen += mlen;
3270 		if (m->m_next == 0) {
3271 			n = m_getclr(M_DONTWAIT, m->m_type);
3272 			if (n == 0)
3273 				goto out;
3274 			n->m_len = min(MLEN, len + off);
3275 			m->m_next = n;
3276 		}
3277 		m = m->m_next;
3278 	}
3279 	while (len > 0) {
3280 		mlen = min(m->m_len - off, len);
3281 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
3282 		cp += mlen;
3283 		len -= mlen;
3284 		mlen += off;
3285 		off = 0;
3286 		totlen += mlen;
3287 		if (len == 0)
3288 			break;
3289 		if (m->m_next == 0) {
3290 			n = m_get(M_DONTWAIT, m->m_type);
3291 			if (n == 0)
3292 				break;
3293 			n->m_len = min(MLEN, len);
3294 			m->m_next = n;
3295 		}
3296 		m = m->m_next;
3297 	}
3298 out:
3299 #if 0
3300 	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
3301 		m->m_pkthdr.len = totlen;
3302 #endif
3303 	return;
3304 }
3305 #endif /* (_KERNEL) && ( ((BSD < 199103) && !MENTAT) || __sgi) */
3306 
3307 
3308 /* ------------------------------------------------------------------------ */
3309 /* Function:    fr_findgroup                                                */
3310 /* Returns:     frgroup_t * - NULL = group not found, else pointer to group */
3311 /* Parameters:  group(I) - group name to search for                         */
3312 /*              unit(I)  - device to which this group belongs               */
3313 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3314 /*              fgpp(O)  - pointer to place to store pointer to the pointer */
3315 /*                         to where to add the next (last) group or where   */
3316 /*                         to delete group from.                            */
3317 /*                                                                          */
3318 /* Search amongst the defined groups for a particular group number.         */
3319 /* ------------------------------------------------------------------------ */
fr_findgroup(group,unit,set,fgpp,ifs)3320 frgroup_t *fr_findgroup(group, unit, set, fgpp, ifs)
3321 char *group;
3322 minor_t unit;
3323 int set;
3324 frgroup_t ***fgpp;
3325 ipf_stack_t *ifs;
3326 {
3327 	frgroup_t *fg, **fgp;
3328 
3329 	/*
3330 	 * Which list of groups to search in is dependent on which list of
3331 	 * rules are being operated on.
3332 	 */
3333 	fgp = &ifs->ifs_ipfgroups[unit][set];
3334 
3335 	while ((fg = *fgp) != NULL) {
3336 		if (strncmp(group, fg->fg_name, FR_GROUPLEN) == 0)
3337 			break;
3338 		else
3339 			fgp = &fg->fg_next;
3340 	}
3341 	if (fgpp != NULL)
3342 		*fgpp = fgp;
3343 	return fg;
3344 }
3345 
3346 
3347 /* ------------------------------------------------------------------------ */
3348 /* Function:    fr_addgroup                                                 */
3349 /* Returns:     frgroup_t * - NULL == did not create group,                 */
3350 /*                            != NULL == pointer to the group               */
3351 /* Parameters:  num(I)   - group number to add                              */
3352 /*              head(I)  - rule pointer that is using this as the head      */
3353 /*              flags(I) - rule flags which describe the type of rule it is */
3354 /*              unit(I)  - device to which this group will belong to        */
3355 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3356 /* Write Locks: ipf_mutex                                                   */
3357 /*                                                                          */
3358 /* Add a new group head, or if it already exists, increase the reference    */
3359 /* count to it.                                                             */
3360 /* ------------------------------------------------------------------------ */
fr_addgroup(group,head,flags,unit,set,ifs)3361 frgroup_t *fr_addgroup(group, head, flags, unit, set, ifs)
3362 char *group;
3363 void *head;
3364 u_32_t flags;
3365 minor_t unit;
3366 int set;
3367 ipf_stack_t *ifs;
3368 {
3369 	frgroup_t *fg, **fgp;
3370 	u_32_t gflags;
3371 
3372 	if (group == NULL)
3373 		return NULL;
3374 
3375 	if (unit == IPL_LOGIPF && *group == '\0')
3376 		return NULL;
3377 
3378 	fgp = NULL;
3379 	gflags = flags & FR_INOUT;
3380 
3381 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3382 	if (fg != NULL) {
3383 		if (fg->fg_flags == 0)
3384 			fg->fg_flags = gflags;
3385 		else if (gflags != fg->fg_flags)
3386 			return NULL;
3387 		fg->fg_ref++;
3388 		return fg;
3389 	}
3390 	KMALLOC(fg, frgroup_t *);
3391 	if (fg != NULL) {
3392 		fg->fg_head = head;
3393 		fg->fg_start = NULL;
3394 		fg->fg_next = *fgp;
3395 		bcopy(group, fg->fg_name, FR_GROUPLEN);
3396 		fg->fg_flags = gflags;
3397 		fg->fg_ref = 1;
3398 		*fgp = fg;
3399 	}
3400 	return fg;
3401 }
3402 
3403 
3404 /* ------------------------------------------------------------------------ */
3405 /* Function:    fr_delgroup                                                 */
3406 /* Returns:     Nil                                                         */
3407 /* Parameters:  group(I) - group name to delete                             */
3408 /*              unit(I)  - device to which this group belongs               */
3409 /*              set(I)   - which set of rules (inactive/inactive) this is   */
3410 /* Write Locks: ipf_mutex                                                   */
3411 /*                                                                          */
3412 /* Attempt to delete a group head.                                          */
3413 /* Only do this when its reference count reaches 0.                         */
3414 /* ------------------------------------------------------------------------ */
fr_delgroup(group,unit,set,ifs)3415 void fr_delgroup(group, unit, set, ifs)
3416 char *group;
3417 minor_t unit;
3418 int set;
3419 ipf_stack_t *ifs;
3420 {
3421 	frgroup_t *fg, **fgp;
3422 
3423 	fg = fr_findgroup(group, unit, set, &fgp, ifs);
3424 	if (fg == NULL)
3425 		return;
3426 
3427 	fg->fg_ref--;
3428 	if (fg->fg_ref == 0) {
3429 		*fgp = fg->fg_next;
3430 		KFREE(fg);
3431 	}
3432 }
3433 
3434 
3435 /* ------------------------------------------------------------------------ */
3436 /* Function:    fr_getrulen                                                 */
3437 /* Returns:     frentry_t * - NULL == not found, else pointer to rule n     */
3438 /* Parameters:  unit(I)  - device for which to count the rule's number      */
3439 /*              flags(I) - which set of rules to find the rule in           */
3440 /*              group(I) - group name                                       */
3441 /*              n(I)     - rule number to find                              */
3442 /*                                                                          */
3443 /* Find rule # n in group # g and return a pointer to it.  Return NULl if   */
3444 /* group # g doesn't exist or there are less than n rules in the group.     */
3445 /* ------------------------------------------------------------------------ */
fr_getrulen(unit,group,n,ifs)3446 frentry_t *fr_getrulen(unit, group, n, ifs)
3447 int unit;
3448 char *group;
3449 u_32_t n;
3450 ipf_stack_t *ifs;
3451 {
3452 	frentry_t *fr;
3453 	frgroup_t *fg;
3454 
3455 	fg = fr_findgroup(group, unit, ifs->ifs_fr_active, NULL, ifs);
3456 	if (fg == NULL)
3457 		return NULL;
3458 	for (fr = fg->fg_head; fr && n; fr = fr->fr_next, n--)
3459 		;
3460 	if (n != 0)
3461 		return NULL;
3462 	return fr;
3463 }
3464 
3465 
3466 /* ------------------------------------------------------------------------ */
3467 /* Function:    fr_rulen                                                    */
3468 /* Returns:     int - >= 0 - rule number, -1 == search failed               */
3469 /* Parameters:  unit(I) - device for which to count the rule's number       */
3470 /*              fr(I)   - pointer to rule to match                          */
3471 /*                                                                          */
3472 /* Return the number for a rule on a specific filtering device.             */
3473 /* ------------------------------------------------------------------------ */
fr_rulen(unit,fr,ifs)3474 int fr_rulen(unit, fr, ifs)
3475 int unit;
3476 frentry_t *fr;
3477 ipf_stack_t *ifs;
3478 {
3479 	frentry_t *fh;
3480 	frgroup_t *fg;
3481 	u_32_t n = 0;
3482 
3483 	if (fr == NULL)
3484 		return -1;
3485 	fg = fr_findgroup(fr->fr_group, unit, ifs->ifs_fr_active, NULL, ifs);
3486 	if (fg == NULL)
3487 		return -1;
3488 	for (fh = fg->fg_head; fh; n++, fh = fh->fr_next)
3489 		if (fh == fr)
3490 			break;
3491 	if (fh == NULL)
3492 		return -1;
3493 	return n;
3494 }
3495 
3496 
3497 /* ------------------------------------------------------------------------ */
3498 /* Function:    frflushlist                                                 */
3499 /* Returns:     int - >= 0 - number of flushed rules                        */
3500 /* Parameters:  set(I)   - which set of rules (inactive/inactive) this is   */
3501 /*              unit(I)  - device for which to flush rules                  */
3502 /*              flags(I) - which set of rules to flush                      */
3503 /*              nfreedp(O) - pointer to int where flush count is stored     */
3504 /*              listp(I)   - pointer to list to flush pointer               */
3505 /* Write Locks: ipf_mutex                                                   */
3506 /*                                                                          */
3507 /* Recursively flush rules from the list, descending groups as they are     */
3508 /* encountered.  if a rule is the head of a group and it has lost all its   */
3509 /* group members, then also delete the group reference.  nfreedp is needed  */
3510 /* to store the accumulating count of rules removed, whereas the returned   */
3511 /* value is just the number removed from the current list.  The latter is   */
3512 /* needed to correctly adjust reference counts on rules that define groups. */
3513 /*                                                                          */
3514 /* NOTE: Rules not loaded from user space cannot be flushed.                */
3515 /* ------------------------------------------------------------------------ */
frflushlist(set,unit,nfreedp,listp,ifs)3516 static int frflushlist(set, unit, nfreedp, listp, ifs)
3517 int set;
3518 minor_t unit;
3519 int *nfreedp;
3520 frentry_t **listp;
3521 ipf_stack_t *ifs;
3522 {
3523 	int freed = 0;
3524 	frentry_t *fp;
3525 
3526 	while ((fp = *listp) != NULL) {
3527 		if ((fp->fr_type & FR_T_BUILTIN) ||
3528 		    !(fp->fr_flags & FR_COPIED)) {
3529 			listp = &fp->fr_next;
3530 			continue;
3531 		}
3532 		*listp = fp->fr_next;
3533 		if (fp->fr_grp != NULL) {
3534 			(void) frflushlist(set, unit, nfreedp, fp->fr_grp, ifs);
3535 		}
3536 
3537 		if (fp->fr_grhead != NULL) {
3538 			fr_delgroup(fp->fr_grhead, unit, set, ifs);
3539 			*fp->fr_grhead = '\0';
3540 		}
3541 
3542 		ASSERT(fp->fr_ref > 0);
3543 		fp->fr_next = NULL;
3544 		if (fr_derefrule(&fp, ifs) == 0)
3545 			freed++;
3546 	}
3547 	*nfreedp += freed;
3548 	return freed;
3549 }
3550 
3551 
3552 /* ------------------------------------------------------------------------ */
3553 /* Function:    frflush                                                     */
3554 /* Returns:     int - >= 0 - number of flushed rules                        */
3555 /* Parameters:  unit(I)  - device for which to flush rules                  */
3556 /*              flags(I) - which set of rules to flush                      */
3557 /*                                                                          */
3558 /* Calls flushlist() for all filter rules (accounting, firewall - both IPv4 */
3559 /* and IPv6) as defined by the value of flags.                              */
3560 /* ------------------------------------------------------------------------ */
frflush(unit,proto,flags,ifs)3561 int frflush(unit, proto, flags, ifs)
3562 minor_t unit;
3563 int proto, flags;
3564 ipf_stack_t *ifs;
3565 {
3566 	int flushed = 0, set;
3567 
3568 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3569 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
3570 
3571 	set = ifs->ifs_fr_active;
3572 	if ((flags & FR_INACTIVE) == FR_INACTIVE)
3573 		set = 1 - set;
3574 
3575 	if (flags & FR_OUTQUE) {
3576 		if (proto == 0 || proto == 6) {
3577 			(void) frflushlist(set, unit,
3578 			    &flushed, &ifs->ifs_ipfilter6[1][set], ifs);
3579 			(void) frflushlist(set, unit,
3580 			    &flushed, &ifs->ifs_ipacct6[1][set], ifs);
3581 		}
3582 		if (proto == 0 || proto == 4) {
3583 			(void) frflushlist(set, unit,
3584 			    &flushed, &ifs->ifs_ipfilter[1][set], ifs);
3585 			(void) frflushlist(set, unit,
3586 			    &flushed, &ifs->ifs_ipacct[1][set], ifs);
3587 		}
3588 	}
3589 	if (flags & FR_INQUE) {
3590 		if (proto == 0 || proto == 6) {
3591 			(void) frflushlist(set, unit,
3592 			    &flushed, &ifs->ifs_ipfilter6[0][set], ifs);
3593 			(void) frflushlist(set, unit,
3594 			    &flushed, &ifs->ifs_ipacct6[0][set], ifs);
3595 		}
3596 		if (proto == 0 || proto == 4) {
3597 			(void) frflushlist(set, unit,
3598 			    &flushed, &ifs->ifs_ipfilter[0][set], ifs);
3599 			(void) frflushlist(set, unit,
3600 			    &flushed, &ifs->ifs_ipacct[0][set], ifs);
3601 		}
3602 	}
3603 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3604 
3605 	if (unit == IPL_LOGIPF) {
3606 		int tmp;
3607 
3608 		tmp = frflush(IPL_LOGCOUNT, proto, flags, ifs);
3609 		if (tmp >= 0)
3610 			flushed += tmp;
3611 	}
3612 	return flushed;
3613 }
3614 
3615 
3616 /* ------------------------------------------------------------------------ */
3617 /* Function:    memstr                                                      */
3618 /* Returns:     char *  - NULL if failed, != NULL pointer to matching bytes */
3619 /* Parameters:  src(I)  - pointer to byte sequence to match                 */
3620 /*              dst(I)  - pointer to byte sequence to search                */
3621 /*              slen(I) - match length                                      */
3622 /*              dlen(I) - length available to search in                     */
3623 /*                                                                          */
3624 /* Search dst for a sequence of bytes matching those at src and extend for  */
3625 /* slen bytes.                                                              */
3626 /* ------------------------------------------------------------------------ */
memstr(src,dst,slen,dlen)3627 char *memstr(src, dst, slen, dlen)
3628 char *src, *dst;
3629 int slen, dlen;
3630 {
3631 	char *s = NULL;
3632 
3633 	while (dlen >= slen) {
3634 		if (bcmp(src, dst, slen) == 0) {
3635 			s = dst;
3636 			break;
3637 		}
3638 		dst++;
3639 		dlen--;
3640 	}
3641 	return s;
3642 }
3643 /* ------------------------------------------------------------------------ */
3644 /* Function:    fr_fixskip                                                  */
3645 /* Returns:     Nil                                                         */
3646 /* Parameters:  listp(IO)    - pointer to start of list with skip rule      */
3647 /*              rp(I)        - rule added/removed with skip in it.          */
3648 /*              addremove(I) - adjustment (-1/+1) to make to skip count,    */
3649 /*                             depending on whether a rule was just added   */
3650 /*                             or removed.                                  */
3651 /*                                                                          */
3652 /* Adjust all the rules in a list which would have skip'd past the position */
3653 /* where we are inserting to skip to the right place given the change.      */
3654 /* ------------------------------------------------------------------------ */
fr_fixskip(listp,rp,addremove)3655 void fr_fixskip(listp, rp, addremove)
3656 frentry_t **listp, *rp;
3657 int addremove;
3658 {
3659 	int rules, rn;
3660 	frentry_t *fp;
3661 
3662 	rules = 0;
3663 	for (fp = *listp; (fp != NULL) && (fp != rp); fp = fp->fr_next)
3664 		rules++;
3665 
3666 	if (!fp)
3667 		return;
3668 
3669 	for (rn = 0, fp = *listp; fp && (fp != rp); fp = fp->fr_next, rn++)
3670 		if (FR_ISSKIP(fp->fr_flags) && (rn + fp->fr_arg >= rules))
3671 			fp->fr_arg += addremove;
3672 }
3673 
3674 
3675 #ifdef	_KERNEL
3676 /* ------------------------------------------------------------------------ */
3677 /* Function:    count4bits                                                  */
3678 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3679 /* Parameters:  ip(I) - 32bit IP address                                    */
3680 /*                                                                          */
3681 /* IPv4 ONLY                                                                */
3682 /* count consecutive 1's in bit mask.  If the mask generated by counting    */
3683 /* consecutive 1's is different to that passed, return -1, else return #    */
3684 /* of bits.                                                                 */
3685 /* ------------------------------------------------------------------------ */
count4bits(ip)3686 int	count4bits(ip)
3687 u_32_t	ip;
3688 {
3689 	u_32_t	ipn;
3690 	int	cnt = 0, i, j;
3691 
3692 	ip = ipn = ntohl(ip);
3693 	for (i = 32; i; i--, ipn *= 2)
3694 		if (ipn & 0x80000000)
3695 			cnt++;
3696 		else
3697 			break;
3698 	ipn = 0;
3699 	for (i = 32, j = cnt; i; i--, j--) {
3700 		ipn *= 2;
3701 		if (j > 0)
3702 			ipn++;
3703 	}
3704 	if (ipn == ip)
3705 		return cnt;
3706 	return -1;
3707 }
3708 
3709 
3710 #ifdef USE_INET6
3711 /* ------------------------------------------------------------------------ */
3712 /* Function:    count6bits                                                  */
3713 /* Returns:     int - >= 0 - number of consecutive bits in input            */
3714 /* Parameters:  msk(I) - pointer to start of IPv6 bitmask                   */
3715 /*                                                                          */
3716 /* IPv6 ONLY                                                                */
3717 /* count consecutive 1's in bit mask.                                       */
3718 /* ------------------------------------------------------------------------ */
count6bits(msk)3719 int count6bits(msk)
3720 u_32_t *msk;
3721 {
3722 	int i = 0, k;
3723 	u_32_t j;
3724 
3725 	for (k = 3; k >= 0; k--)
3726 		if (msk[k] == 0xffffffff)
3727 			i += 32;
3728 		else {
3729 			for (j = msk[k]; j; j <<= 1)
3730 				if (j & 0x80000000)
3731 					i++;
3732 		}
3733 	return i;
3734 }
3735 # endif
3736 #endif /* _KERNEL */
3737 
3738 
3739 /* ------------------------------------------------------------------------ */
3740 /* Function:    fr_ifsync                                                   */
3741 /* Returns:     void *    - new interface identifier                        */
3742 /* Parameters:  action(I)  - type of synchronisation to do                  */
3743 /*              v(I)       - IP version being sync'd (v4 or v6)             */
3744 /*              newifp(I)  - interface identifier being introduced/removed  */
3745 /*              oldifp(I)  - interface identifier in a filter rule          */
3746 /*              newname(I) - name associated with newifp interface          */
3747 /*              oldname(I) - name associated with oldifp interface          */
3748 /*		ifs       - pointer to IPF stack instance		    */
3749 /*                                                                          */
3750 /* This function returns what the new value for "oldifp" should be for its  */
3751 /* caller.  In some cases it will not change, in some it will.              */
3752 /* action == IPFSYNC_RESYNC                                                 */
3753 /*   a new value for oldifp will always be looked up, according to oldname, */
3754 /*   the values of newname and newifp are ignored.                          */
3755 /* action == IPFSYNC_NEWIFP                                                 */
3756 /*   if oldname matches newname then we are doing a sync for the matching   */
3757 /*   interface, so we return newifp to be used in place of oldifp.  If the  */
3758 /*   the names don't match, just return oldifp.                             */
3759 /* action == IPFSYNC_OLDIFP                                                 */
3760 /*   if oldifp matches newifp then we are are doing a sync to remove any    */
3761 /*   references to oldifp, so we return "-1".                               */
3762 /* -----								    */
3763 /* NOTE:								    */
3764 /* This function processes NIC event from PF_HOOKS. The action parameter    */
3765 /* is set in ipf_nic_event_v4()/ipf_nic_event_v6() function. There is	    */
3766 /* one single switch statement() in ipf_nic_event_vx() function, which	    */
3767 /* translates the HOOK event type to action parameter passed to fr_ifsync.  */
3768 /* The translation table looks as follows:				    */
3769 /*	event		| action					    */
3770 /*	----------------+-------------					    */
3771 /*	NE_PLUMB	| IPFSYNC_NEWIFP				    */
3772 /*	NE_UNPLUMB	| IPFSYNC_OLDIFP				    */
3773 /*    NE_ADDRESS_CHANGE	| IPFSYNC_RESYNC				    */
3774 /*									    */
3775 /* The oldname and oldifp parameters are taken from IPF entry (rule, state  */
3776 /* table entry, NAT table entry, fragment ...). The newname and newifp	    */
3777 /* parameters come from hook event data, parameters are taken from event    */
3778 /* in ipf_nic_event_vx() functions. Any time NIC changes, the IPF is	    */
3779 /* notified by hook function.						    */
3780 /*									    */
3781 /* We get NE_UNPLUMB event from PF_HOOKS even if someone coincidently tries */
3782 /* to plumb the interface, which is already plumbed. In such case we always */
3783 /* get the event from PF_HOOKS as follows:				    */
3784 /*	event:	NE_PLUMB						    */
3785 /*	NIC:	0x0							    */
3786 /* ------------------------------------------------------------------------ */
fr_ifsync(action,v,newname,oldname,newifp,oldifp,ifs)3787 static void *fr_ifsync(action, v, newname, oldname, newifp, oldifp, ifs)
3788 int action, v;
3789 char *newname, *oldname;
3790 void *newifp, *oldifp;
3791 ipf_stack_t *ifs;
3792 {
3793 	void *rval = oldifp;
3794 
3795 	switch (action)
3796 	{
3797 	case IPFSYNC_RESYNC :
3798 		if (oldname[0] != '\0') {
3799 			rval = fr_resolvenic(oldname, v, ifs);
3800 		}
3801 		break;
3802 	case IPFSYNC_NEWIFP :
3803 		if (!strncmp(newname, oldname, LIFNAMSIZ))
3804 			rval = newifp;
3805 		break;
3806 	case IPFSYNC_OLDIFP :
3807 		/*
3808 		 * If interface gets unplumbed it must be invalidated, which
3809 		 * means set all existing references to the interface to -1.
3810 		 * We don't want to invalidate references for wildcard
3811 		 * (unbound) rules (entries).
3812 		 */
3813 		if (newifp == oldifp)
3814 			rval = (oldifp) ? (void *)-1 : NULL;
3815 		break;
3816 	}
3817 
3818 	return rval;
3819 }
3820 
3821 
3822 /* ------------------------------------------------------------------------ */
3823 /* Function:    frsynclist                                                  */
3824 /* Returns:     void                                                        */
3825 /* Parameters:  action(I) - type of synchronisation to do                   */
3826 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3827 /*              ifp(I)    - interface identifier associated with action     */
3828 /*              ifname(I) - name associated with ifp parameter              */
3829 /*              fr(I)     - pointer to filter rule                          */
3830 /*		ifs       - pointer to IPF stack instance		    */
3831 /* Write Locks: ipf_mutex                                                   */
3832 /*                                                                          */
3833 /* Walk through a list of filter rules and resolve any interface names into */
3834 /* pointers.  Where dynamic addresses are used, also update the IP address  */
3835 /* used in the rule.  The interface pointer is used to limit the lookups to */
3836 /* a specific set of matching names if it is non-NULL.                      */
3837 /* ------------------------------------------------------------------------ */
frsynclist(action,v,ifp,ifname,fr,ifs)3838 static void frsynclist(action, v, ifp, ifname, fr, ifs)
3839 int action, v;
3840 void *ifp;
3841 char *ifname;
3842 frentry_t *fr;
3843 ipf_stack_t *ifs;
3844 {
3845 	frdest_t *fdp;
3846 	int rv, i;
3847 
3848 	for (; fr; fr = fr->fr_next) {
3849 		rv = fr->fr_v;
3850 		if (v != 0 && v != rv)
3851 			continue;
3852 
3853 		/*
3854 		 * Lookup all the interface names that are part of the rule.
3855 		 */
3856 		for (i = 0; i < 4; i++) {
3857 			fr->fr_ifas[i] = fr_ifsync(action, rv, ifname,
3858 						   fr->fr_ifnames[i],
3859 						   ifp, fr->fr_ifas[i],
3860 						   ifs);
3861 		}
3862 
3863 		fdp = &fr->fr_tifs[0];
3864 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3865 					   ifp, fdp->fd_ifp, ifs);
3866 
3867 		fdp = &fr->fr_tifs[1];
3868 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3869 					   ifp, fdp->fd_ifp, ifs);
3870 
3871 		fdp = &fr->fr_dif;
3872 		fdp->fd_ifp = fr_ifsync(action, rv, ifname, fdp->fd_ifname,
3873 					   ifp, fdp->fd_ifp, ifs);
3874 
3875 		if (action != IPFSYNC_RESYNC)
3876 			continue;
3877 
3878 		if (fr->fr_type == FR_T_IPF) {
3879 			if (fr->fr_satype != FRI_NORMAL &&
3880 			    fr->fr_satype != FRI_LOOKUP) {
3881 				(void)fr_ifpaddr(rv, fr->fr_satype,
3882 						 fr->fr_ifas[fr->fr_sifpidx],
3883 						 &fr->fr_src, &fr->fr_smsk,
3884 						 ifs);
3885 			}
3886 			if (fr->fr_datype != FRI_NORMAL &&
3887 			    fr->fr_datype != FRI_LOOKUP) {
3888 				(void)fr_ifpaddr(rv, fr->fr_datype,
3889 						 fr->fr_ifas[fr->fr_difpidx],
3890 						 &fr->fr_dst, &fr->fr_dmsk,
3891 						 ifs);
3892 			}
3893 		}
3894 
3895 #ifdef	IPFILTER_LOOKUP
3896 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP &&
3897 		    fr->fr_srcptr == NULL) {
3898 			fr->fr_srcptr = fr_resolvelookup(fr->fr_srctype,
3899 							 fr->fr_srcnum,
3900 							 &fr->fr_srcfunc, ifs);
3901 		}
3902 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP &&
3903 		    fr->fr_dstptr == NULL) {
3904 			fr->fr_dstptr = fr_resolvelookup(fr->fr_dsttype,
3905 							 fr->fr_dstnum,
3906 							 &fr->fr_dstfunc, ifs);
3907 		}
3908 #endif
3909 	}
3910 }
3911 
3912 
3913 #ifdef	_KERNEL
3914 /* ------------------------------------------------------------------------ */
3915 /* Function:    frsync                                                      */
3916 /* Returns:     void                                                        */
3917 /* Parameters:  action(I) - type of synchronisation to do                   */
3918 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3919 /*              ifp(I)    - interface identifier associated with action     */
3920 /*              name(I)   - name associated with ifp parameter              */
3921 /*                                                                          */
3922 /* frsync() is called when we suspect that the interface list or            */
3923 /* information about interfaces (like IP#) has changed.  Go through all     */
3924 /* filter rules, NAT entries and the state table and check if anything      */
3925 /* needs to be changed/updated.                                             */
3926 /* With the filtering hooks added to Solaris, we needed to change the manner*/
3927 /* in which this was done to support three different types of sync:         */
3928 /* - complete resync of all interface name/identifiers                      */
3929 /* - new interface being announced with its name and identifier             */
3930 /* - interface removal being announced by only its identifier               */
3931 /* ------------------------------------------------------------------------ */
frsync(action,v,ifp,name,ifs)3932 void frsync(action, v, ifp, name, ifs)
3933 int action, v;
3934 void *ifp;
3935 char *name;
3936 ipf_stack_t *ifs;
3937 {
3938 	int i;
3939 
3940 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
3941 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[0][ifs->ifs_fr_active], ifs);
3942 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct[1][ifs->ifs_fr_active], ifs);
3943 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[0][ifs->ifs_fr_active], ifs);
3944 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter[1][ifs->ifs_fr_active], ifs);
3945 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[0][ifs->ifs_fr_active], ifs);
3946 	frsynclist(action, v, ifp, name, ifs->ifs_ipacct6[1][ifs->ifs_fr_active], ifs);
3947 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[0][ifs->ifs_fr_active], ifs);
3948 	frsynclist(action, v, ifp, name, ifs->ifs_ipfilter6[1][ifs->ifs_fr_active], ifs);
3949 
3950 	for (i = 0; i < IPL_LOGSIZE; i++) {
3951 		frgroup_t *g;
3952 
3953 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
3954 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3955 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
3956 			frsynclist(action, v, ifp, name, g->fg_start, ifs);
3957 	}
3958 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
3959 }
3960 
3961 #if SOLARIS2 >= 10
3962 /* ------------------------------------------------------------------------ */
3963 /* Function:    fr_syncindex						    */
3964 /* Returns:     void							    */
3965 /* Parameters:  rules	  - list of rules to be sync'd			    */
3966 /*		ifp	  - interface, which is being sync'd		    */
3967 /*		newifp	  - new ifindex value for interface		    */
3968 /*                                                                          */
3969 /* Function updates all NIC indecis, which match ifp, in every rule. Every  */
3970 /* NIC index matching ifp, will be updated to newifp.			    */
3971 /* ------------------------------------------------------------------------ */
fr_syncindex(rules,ifp,newifp)3972 static void fr_syncindex(rules, ifp, newifp)
3973 frentry_t *rules;
3974 void *ifp;
3975 void *newifp;
3976 {
3977 	int i;
3978 	frentry_t *fr;
3979 
3980 	for (fr = rules; fr != NULL; fr = fr->fr_next) {
3981 		/*
3982 		 * Lookup all the interface names that are part of the rule.
3983 		 */
3984 		for (i = 0; i < 4; i++)
3985 			if (fr->fr_ifas[i] == ifp)
3986 				fr->fr_ifas[i] = newifp;
3987 
3988 		for (i = 0; i < 2; i++) {
3989 			if (fr->fr_tifs[i].fd_ifp == ifp)
3990 				fr->fr_tifs[i].fd_ifp = newifp;
3991 		}
3992 
3993 		if (fr->fr_dif.fd_ifp == ifp)
3994 			fr->fr_dif.fd_ifp = newifp;
3995 	}
3996 }
3997 
3998 /* ------------------------------------------------------------------------ */
3999 /* Function:    fr_ifindexsync						    */
4000 /* Returns:     void							    */
4001 /* Parameters:	ifp	  - interface, which is being sync'd		    */
4002 /*		newifp	  - new ifindex value for interface		    */
4003 /*              ifs	  - IPF's stack					    */
4004 /*                                                                          */
4005 /* Function assumes ipf_mutex is locked exclusively.			    */
4006 /* 									    */
4007 /* Function updates the NIC references in rules with new interfaces index   */
4008 /* (newifp). Function must process active lists:			    */
4009 /*	with accounting rules (IPv6 and IPv4)				    */
4010 /*	with inbound rules (IPv6 and IPv4)				    */
4011 /*	with outbound rules (IPv6 and IPv4)				    */
4012 /* Function also has to take care of rule groups.			    */
4013 /*                                                                          */
4014 /* NOTE: The ipf_mutex is grabbed exclusively by caller (which is always    */
4015 /* nic_event_hook). The hook function also updates state entries, NAT rules */
4016 /* and NAT entries. We want to do all these update atomically to keep the   */
4017 /* NIC references consistent. The ipf_mutex will synchronize event with	    */
4018 /* fr_check(), which processes packets,	so no packet will enter fr_check(), */
4019 /* while NIC references will be synchronized.				    */
4020 /* ------------------------------------------------------------------------ */
fr_ifindexsync(ifp,newifp,ifs)4021 void fr_ifindexsync(ifp, newifp, ifs)
4022 void *ifp;
4023 void *newifp;
4024 ipf_stack_t *ifs;
4025 {
4026 	unsigned int	i;
4027 	frentry_t *rule_lists[8];
4028 	unsigned int	rules = sizeof (rule_lists) / sizeof (frentry_t *);
4029 
4030 	rule_lists[0] = ifs->ifs_ipacct[0][ifs->ifs_fr_active];
4031 	rule_lists[1] =	ifs->ifs_ipacct[1][ifs->ifs_fr_active];
4032 	rule_lists[2] =	ifs->ifs_ipfilter[0][ifs->ifs_fr_active];
4033 	rule_lists[3] =	ifs->ifs_ipfilter[1][ifs->ifs_fr_active];
4034 	rule_lists[4] =	ifs->ifs_ipacct6[0][ifs->ifs_fr_active];
4035 	rule_lists[5] =	ifs->ifs_ipacct6[1][ifs->ifs_fr_active];
4036 	rule_lists[6] =	ifs->ifs_ipfilter6[0][ifs->ifs_fr_active];
4037 	rule_lists[7] =	ifs->ifs_ipfilter6[1][ifs->ifs_fr_active];
4038 
4039 	for (i = 0; i < rules; i++) {
4040 		fr_syncindex(rule_lists[i], ifp, newifp);
4041 	}
4042 
4043 	/*
4044 	 * Update rule groups.
4045 	 */
4046 	for (i = 0; i < IPL_LOGSIZE; i++) {
4047 		frgroup_t *g;
4048 
4049 		for (g = ifs->ifs_ipfgroups[i][0]; g != NULL; g = g->fg_next)
4050 			fr_syncindex(g->fg_start, ifp, newifp);
4051 		for (g = ifs->ifs_ipfgroups[i][1]; g != NULL; g = g->fg_next)
4052 			fr_syncindex(g->fg_start, ifp, newifp);
4053 	}
4054 }
4055 #endif
4056 
4057 /*
4058  * In the functions below, bcopy() is called because the pointer being
4059  * copied _from_ in this instance is a pointer to a char buf (which could
4060  * end up being unaligned) and on the kernel's local stack.
4061  */
4062 /* ------------------------------------------------------------------------ */
4063 /* Function:    copyinptr                                                   */
4064 /* Returns:     int - 0 = success, else failure                             */
4065 /* Parameters:  src(I)  - pointer to the source address                     */
4066 /*              dst(I)  - destination address                               */
4067 /*              size(I) - number of bytes to copy                           */
4068 /*                                                                          */
4069 /* Copy a block of data in from user space, given a pointer to the pointer  */
4070 /* to start copying from (src) and a pointer to where to store it (dst).    */
4071 /* NB: src - pointer to user space pointer, dst - kernel space pointer      */
4072 /* ------------------------------------------------------------------------ */
copyinptr(src,dst,size)4073 int copyinptr(src, dst, size)
4074 void *src, *dst;
4075 size_t size;
4076 {
4077 	caddr_t ca;
4078 	int err;
4079 
4080 # if SOLARIS
4081 	err = COPYIN(src, (caddr_t)&ca, sizeof(ca));
4082 	if (err != 0)
4083 		return err;
4084 # else
4085 	bcopy(src, (caddr_t)&ca, sizeof(ca));
4086 # endif
4087 	err = COPYIN(ca, dst, size);
4088 	return err;
4089 }
4090 
4091 
4092 /* ------------------------------------------------------------------------ */
4093 /* Function:    copyoutptr                                                  */
4094 /* Returns:     int - 0 = success, else failure                             */
4095 /* Parameters:  src(I)  - pointer to the source address                     */
4096 /*              dst(I)  - destination address                               */
4097 /*              size(I) - number of bytes to copy                           */
4098 /*                                                                          */
4099 /* Copy a block of data out to user space, given a pointer to the pointer   */
4100 /* to start copying from (src) and a pointer to where to store it (dst).    */
4101 /* NB: src - kernel space pointer, dst - pointer to user space pointer.     */
4102 /* ------------------------------------------------------------------------ */
copyoutptr(src,dst,size)4103 int copyoutptr(src, dst, size)
4104 void *src, *dst;
4105 size_t size;
4106 {
4107 	caddr_t ca;
4108 	int err;
4109 
4110 # if SOLARIS
4111 	err = COPYIN(dst, (caddr_t)&ca, sizeof(ca));
4112 	if (err != 0)
4113 		return err;
4114 # else
4115 	bcopy(dst, (caddr_t)&ca, sizeof(ca));
4116 # endif
4117 	err = COPYOUT(src, ca, size);
4118 	return err;
4119 }
4120 #endif
4121 
4122 
4123 /* ------------------------------------------------------------------------ */
4124 /* Function:    fr_lock                                                     */
4125 /* Returns:	int - 0 = success, else error				    */
4126 /* Parameters:  data(I)  - pointer to lock value to set                     */
4127 /*              lockp(O) - pointer to location to store old lock value      */
4128 /*                                                                          */
4129 /* Get the new value for the lock integer, set it and return the old value  */
4130 /* in *lockp.                                                               */
4131 /* ------------------------------------------------------------------------ */
fr_lock(data,lockp)4132 int fr_lock(data, lockp)
4133 caddr_t data;
4134 int *lockp;
4135 {
4136 	int arg, err;
4137 
4138 	err = BCOPYIN(data, (caddr_t)&arg, sizeof(arg));
4139 	if (err != 0)
4140 		return (EFAULT);
4141 	err = BCOPYOUT((caddr_t)lockp, data, sizeof(*lockp));
4142 	if (err != 0)
4143 		return (EFAULT);
4144 	*lockp = arg;
4145 	return (0);
4146 }
4147 
4148 
4149 /* ------------------------------------------------------------------------ */
4150 /* Function:    fr_getstat                                                  */
4151 /* Returns:     Nil                                                         */
4152 /* Parameters:  fiop(I)  - pointer to ipfilter stats structure              */
4153 /*                                                                          */
4154 /* Stores a copy of current pointers, counters, etc, in the friostat        */
4155 /* structure.                                                               */
4156 /* ------------------------------------------------------------------------ */
fr_getstat(fiop,ifs)4157 void fr_getstat(fiop, ifs)
4158 friostat_t *fiop;
4159 ipf_stack_t *ifs;
4160 {
4161 	int i, j;
4162 
4163 	bcopy((char *)&ifs->ifs_frstats, (char *)fiop->f_st,
4164 	    sizeof(filterstats_t) * 2);
4165 	fiop->f_locks[IPL_LOGSTATE] = ifs->ifs_fr_state_lock;
4166 	fiop->f_locks[IPL_LOGNAT] = ifs->ifs_fr_nat_lock;
4167 	fiop->f_locks[IPL_LOGIPF] = ifs->ifs_fr_frag_lock;
4168 	fiop->f_locks[IPL_LOGAUTH] = ifs->ifs_fr_auth_lock;
4169 
4170 	for (i = 0; i < 2; i++)
4171 		for (j = 0; j < 2; j++) {
4172 			fiop->f_ipf[i][j] = ifs->ifs_ipfilter[i][j];
4173 			fiop->f_acct[i][j] = ifs->ifs_ipacct[i][j];
4174 			fiop->f_ipf6[i][j] = ifs->ifs_ipfilter6[i][j];
4175 			fiop->f_acct6[i][j] = ifs->ifs_ipacct6[i][j];
4176 		}
4177 
4178 	fiop->f_ticks = ifs->ifs_fr_ticks;
4179 	fiop->f_active = ifs->ifs_fr_active;
4180 	fiop->f_froute[0] = ifs->ifs_fr_frouteok[0];
4181 	fiop->f_froute[1] = ifs->ifs_fr_frouteok[1];
4182 
4183 	fiop->f_running = ifs->ifs_fr_running;
4184 	for (i = 0; i < IPL_LOGSIZE; i++) {
4185 		fiop->f_groups[i][0] = ifs->ifs_ipfgroups[i][0];
4186 		fiop->f_groups[i][1] = ifs->ifs_ipfgroups[i][1];
4187 	}
4188 #ifdef  IPFILTER_LOG
4189 	fiop->f_logging = 1;
4190 #else
4191 	fiop->f_logging = 0;
4192 #endif
4193 	fiop->f_defpass = ifs->ifs_fr_pass;
4194 	fiop->f_features = fr_features;
4195 	(void) strncpy(fiop->f_version, ipfilter_version,
4196 		       sizeof(fiop->f_version));
4197 }
4198 
4199 
4200 #ifdef	USE_INET6
4201 int icmptoicmp6types[ICMP_MAXTYPE+1] = {
4202 	ICMP6_ECHO_REPLY,	/* 0: ICMP_ECHOREPLY */
4203 	-1,			/* 1: UNUSED */
4204 	-1,			/* 2: UNUSED */
4205 	ICMP6_DST_UNREACH,	/* 3: ICMP_UNREACH */
4206 	-1,			/* 4: ICMP_SOURCEQUENCH */
4207 	ND_REDIRECT,		/* 5: ICMP_REDIRECT */
4208 	-1,			/* 6: UNUSED */
4209 	-1,			/* 7: UNUSED */
4210 	ICMP6_ECHO_REQUEST,	/* 8: ICMP_ECHO */
4211 	-1,			/* 9: UNUSED */
4212 	-1,			/* 10: UNUSED */
4213 	ICMP6_TIME_EXCEEDED,	/* 11: ICMP_TIMXCEED */
4214 	ICMP6_PARAM_PROB,	/* 12: ICMP_PARAMPROB */
4215 	-1,			/* 13: ICMP_TSTAMP */
4216 	-1,			/* 14: ICMP_TSTAMPREPLY */
4217 	-1,			/* 15: ICMP_IREQ */
4218 	-1,			/* 16: ICMP_IREQREPLY */
4219 	-1,			/* 17: ICMP_MASKREQ */
4220 	-1,			/* 18: ICMP_MASKREPLY */
4221 };
4222 
4223 
4224 int	icmptoicmp6unreach[ICMP_MAX_UNREACH] = {
4225 	ICMP6_DST_UNREACH_ADDR,		/* 0: ICMP_UNREACH_NET */
4226 	ICMP6_DST_UNREACH_ADDR,		/* 1: ICMP_UNREACH_HOST */
4227 	-1,				/* 2: ICMP_UNREACH_PROTOCOL */
4228 	ICMP6_DST_UNREACH_NOPORT,	/* 3: ICMP_UNREACH_PORT */
4229 	-1,				/* 4: ICMP_UNREACH_NEEDFRAG */
4230 	ICMP6_DST_UNREACH_NOTNEIGHBOR,	/* 5: ICMP_UNREACH_SRCFAIL */
4231 	ICMP6_DST_UNREACH_ADDR,		/* 6: ICMP_UNREACH_NET_UNKNOWN */
4232 	ICMP6_DST_UNREACH_ADDR,		/* 7: ICMP_UNREACH_HOST_UNKNOWN */
4233 	-1,				/* 8: ICMP_UNREACH_ISOLATED */
4234 	ICMP6_DST_UNREACH_ADMIN,	/* 9: ICMP_UNREACH_NET_PROHIB */
4235 	ICMP6_DST_UNREACH_ADMIN,	/* 10: ICMP_UNREACH_HOST_PROHIB */
4236 	-1,				/* 11: ICMP_UNREACH_TOSNET */
4237 	-1,				/* 12: ICMP_UNREACH_TOSHOST */
4238 	ICMP6_DST_UNREACH_ADMIN,	/* 13: ICMP_UNREACH_ADMIN_PROHIBIT */
4239 };
4240 int	icmpreplytype6[ICMP6_MAXTYPE + 1];
4241 #endif
4242 
4243 int	icmpreplytype4[ICMP_MAXTYPE + 1];
4244 
4245 
4246 /* ------------------------------------------------------------------------ */
4247 /* Function:    fr_matchicmpqueryreply                                      */
4248 /* Returns:     int - 1 if "icmp" is a valid reply to "ic" else 0.          */
4249 /* Parameters:  v(I)    - IP protocol version (4 or 6)                      */
4250 /*              ic(I)   - ICMP information                                  */
4251 /*              icmp(I) - ICMP packet header                                */
4252 /*              rev(I)  - direction (0 = forward/1 = reverse) of packet     */
4253 /*                                                                          */
4254 /* Check if the ICMP packet defined by the header pointed to by icmp is a   */
4255 /* reply to one as described by what's in ic.  If it is a match, return 1,  */
4256 /* else return 0 for no match.                                              */
4257 /* ------------------------------------------------------------------------ */
fr_matchicmpqueryreply(v,ic,icmp,rev)4258 int fr_matchicmpqueryreply(v, ic, icmp, rev)
4259 int v;
4260 icmpinfo_t *ic;
4261 icmphdr_t *icmp;
4262 int rev;
4263 {
4264 	int ictype;
4265 
4266 	ictype = ic->ici_type;
4267 
4268 	if (v == 4) {
4269 		/*
4270 		 * If we matched its type on the way in, then when going out
4271 		 * it will still be the same type.
4272 		 */
4273 		if ((!rev && (icmp->icmp_type == ictype)) ||
4274 		    (rev && (icmpreplytype4[ictype] == icmp->icmp_type))) {
4275 			if (icmp->icmp_type != ICMP_ECHOREPLY)
4276 				return 1;
4277 			if (icmp->icmp_id == ic->ici_id)
4278 				return 1;
4279 		}
4280 	}
4281 #ifdef	USE_INET6
4282 	else if (v == 6) {
4283 		if ((!rev && (icmp->icmp_type == ictype)) ||
4284 		    (rev && (icmpreplytype6[ictype] == icmp->icmp_type))) {
4285 			if (icmp->icmp_type != ICMP6_ECHO_REPLY)
4286 				return 1;
4287 			if (icmp->icmp_id == ic->ici_id)
4288 				return 1;
4289 		}
4290 	}
4291 #endif
4292 	return 0;
4293 }
4294 
4295 
4296 #ifdef	IPFILTER_LOOKUP
4297 /* ------------------------------------------------------------------------ */
4298 /* Function:    fr_resolvelookup                                            */
4299 /* Returns:     void * - NULL = failure, else success.                      */
4300 /* Parameters:  type(I)     - type of lookup these parameters are for.      */
4301 /*              number(I)   - table number to use when searching            */
4302 /*              funcptr(IO) - pointer to pointer for storing IP address     */
4303 /*			      searching function.			    */
4304 /*		ifs	    - ipf stack instance			    */
4305 /*                                                                          */
4306 /* Search for the "table" number passed in amongst those configured for     */
4307 /* that particular type.  If the type is recognised then the function to    */
4308 /* call to do the IP address search will be change, regardless of whether   */
4309 /* or not the "table" number exists.                                        */
4310 /* ------------------------------------------------------------------------ */
fr_resolvelookup(type,number,funcptr,ifs)4311 static void *fr_resolvelookup(type, number, funcptr, ifs)
4312 u_int type, number;
4313 lookupfunc_t *funcptr;
4314 ipf_stack_t *ifs;
4315 {
4316 	char name[FR_GROUPLEN];
4317 	iphtable_t *iph;
4318 	ip_pool_t *ipo;
4319 	void *ptr;
4320 
4321 #if defined(SNPRINTF) && defined(_KERNEL)
4322 	(void) SNPRINTF(name, sizeof(name), "%u", number);
4323 #else
4324 	(void) sprintf(name, "%u", number);
4325 #endif
4326 
4327 	READ_ENTER(&ifs->ifs_ip_poolrw);
4328 
4329 	switch (type)
4330 	{
4331 	case IPLT_POOL :
4332 # if (defined(__osf__) && defined(_KERNEL))
4333 		ptr = NULL;
4334 		*funcptr = NULL;
4335 # else
4336 		ipo = ip_pool_find(IPL_LOGIPF, name, ifs);
4337 		ptr = ipo;
4338 		if (ipo != NULL) {
4339 			ATOMIC_INC32(ipo->ipo_ref);
4340 		}
4341 		*funcptr = ip_pool_search;
4342 # endif
4343 		break;
4344 	case IPLT_HASH :
4345 		iph = fr_findhtable(IPL_LOGIPF, name, ifs);
4346 		ptr = iph;
4347 		if (iph != NULL) {
4348 			ATOMIC_INC32(iph->iph_ref);
4349 		}
4350 		*funcptr = fr_iphmfindip;
4351 		break;
4352 	default:
4353 		ptr = NULL;
4354 		*funcptr = NULL;
4355 		break;
4356 	}
4357 	RWLOCK_EXIT(&ifs->ifs_ip_poolrw);
4358 
4359 	return ptr;
4360 }
4361 #endif
4362 
4363 
4364 /* ------------------------------------------------------------------------ */
4365 /* Function:    frrequest                                                   */
4366 /* Returns:     int - 0 == success, > 0 == errno value                      */
4367 /* Parameters:  unit(I)     - device for which this is for                  */
4368 /*              req(I)      - ioctl command (SIOC*)                         */
4369 /*              data(I)     - pointr to ioctl data                          */
4370 /*              set(I)      - 1 or 0 (filter set)                           */
4371 /*              makecopy(I) - flag indicating whether data points to a rule */
4372 /*                            in kernel space & hence doesn't need copying. */
4373 /*                                                                          */
4374 /* This function handles all the requests which operate on the list of      */
4375 /* filter rules.  This includes adding, deleting, insertion.  It is also    */
4376 /* responsible for creating groups when a "head" rule is loaded.  Interface */
4377 /* names are resolved here and other sanity checks are made on the content  */
4378 /* of the rule structure being loaded.  If a rule has user defined timeouts */
4379 /* then make sure they are created and initialised before exiting.          */
4380 /* ------------------------------------------------------------------------ */
frrequest(unit,req,data,set,makecopy,ifs)4381 int frrequest(unit, req, data, set, makecopy, ifs)
4382 int unit;
4383 ioctlcmd_t req;
4384 int set, makecopy;
4385 caddr_t data;
4386 ipf_stack_t *ifs;
4387 {
4388 	frentry_t frd, *fp, *f, **fprev, **ftail;
4389 	int error = 0, in, v;
4390 	void *ptr, *uptr;
4391 	u_int *p, *pp;
4392 	frgroup_t *fg;
4393 	char *group;
4394 
4395 	fg = NULL;
4396 	fp = &frd;
4397 	if (makecopy != 0) {
4398 		error = fr_inobj(data, fp, IPFOBJ_FRENTRY);
4399 		if (error)
4400 			return EFAULT;
4401 		if ((fp->fr_flags & FR_T_BUILTIN) != 0)
4402 			return EINVAL;
4403 		fp->fr_ref = 0;
4404 		fp->fr_flags |= FR_COPIED;
4405 	} else {
4406 		fp = (frentry_t *)data;
4407 		if ((fp->fr_type & FR_T_BUILTIN) == 0)
4408 			return EINVAL;
4409 		fp->fr_flags &= ~FR_COPIED;
4410 	}
4411 
4412 	if (((fp->fr_dsize == 0) && (fp->fr_data != NULL)) ||
4413 	    ((fp->fr_dsize != 0) && (fp->fr_data == NULL)))
4414 		return EINVAL;
4415 
4416 	v = fp->fr_v;
4417 	uptr = fp->fr_data;
4418 
4419 	/*
4420 	 * Only filter rules for IPv4 or IPv6 are accepted.
4421 	 */
4422 	if (v == 4)
4423 		/*EMPTY*/;
4424 #ifdef	USE_INET6
4425 	else if (v == 6)
4426 		/*EMPTY*/;
4427 #endif
4428 	else {
4429 		return EINVAL;
4430 	}
4431 
4432 	/*
4433 	 * If the rule is being loaded from user space, i.e. we had to copy it
4434 	 * into kernel space, then do not trust the function pointer in the
4435 	 * rule.
4436 	 */
4437 	if ((makecopy == 1) && (fp->fr_func != NULL)) {
4438 		if (fr_findfunc(fp->fr_func) == NULL)
4439 			return ESRCH;
4440 		error = fr_funcinit(fp, ifs);
4441 		if (error != 0)
4442 			return error;
4443 	}
4444 
4445 	ptr = NULL;
4446 	/*
4447 	 * Check that the group number does exist and that its use (in/out)
4448 	 * matches what the rule is.
4449 	 */
4450 	if (!strncmp(fp->fr_grhead, "0", FR_GROUPLEN))
4451 		*fp->fr_grhead = '\0';
4452 	group = fp->fr_group;
4453 	if (!strncmp(group, "0", FR_GROUPLEN))
4454 		*group = '\0';
4455 
4456 	if (FR_ISACCOUNT(fp->fr_flags))
4457 		unit = IPL_LOGCOUNT;
4458 
4459 	if ((req != (int)SIOCZRLST) && (*group != '\0')) {
4460 		fg = fr_findgroup(group, unit, set, NULL, ifs);
4461 		if (fg == NULL)
4462 			return ESRCH;
4463 		if (fg->fg_flags == 0)
4464 			fg->fg_flags = fp->fr_flags & FR_INOUT;
4465 		else if (fg->fg_flags != (fp->fr_flags & FR_INOUT))
4466 			return ESRCH;
4467 	}
4468 
4469 	in = (fp->fr_flags & FR_INQUE) ? 0 : 1;
4470 
4471 	/*
4472 	 * Work out which rule list this change is being applied to.
4473 	 */
4474 	ftail = NULL;
4475 	fprev = NULL;
4476 	if (unit == IPL_LOGAUTH)
4477 		fprev = &ifs->ifs_ipauth;
4478 	else if (v == 4) {
4479 		if (FR_ISACCOUNT(fp->fr_flags))
4480 			fprev = &ifs->ifs_ipacct[in][set];
4481 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4482 			fprev = &ifs->ifs_ipfilter[in][set];
4483 	} else if (v == 6) {
4484 		if (FR_ISACCOUNT(fp->fr_flags))
4485 			fprev = &ifs->ifs_ipacct6[in][set];
4486 		else if ((fp->fr_flags & (FR_OUTQUE|FR_INQUE)) != 0)
4487 			fprev = &ifs->ifs_ipfilter6[in][set];
4488 	}
4489 	if (fprev == NULL)
4490 		return ESRCH;
4491 
4492 	if (*group != '\0') {
4493 	    if (!fg && !(fg = fr_findgroup(group, unit, set, NULL, ifs)))
4494 			return ESRCH;
4495 		fprev = &fg->fg_start;
4496 	}
4497 
4498 	ftail = fprev;
4499 	for (f = *ftail; (f = *ftail) != NULL; ftail = &f->fr_next) {
4500 		if (fp->fr_collect <= f->fr_collect) {
4501 			ftail = fprev;
4502 			f = NULL;
4503 			break;
4504 		}
4505 		fprev = ftail;
4506 	}
4507 
4508 	/*
4509 	 * Copy in extra data for the rule.
4510 	 */
4511 	if (fp->fr_dsize != 0) {
4512 		if (makecopy != 0) {
4513 			KMALLOCS(ptr, void *, fp->fr_dsize);
4514 			if (!ptr)
4515 				return ENOMEM;
4516 			error = COPYIN(uptr, ptr, fp->fr_dsize);
4517 		} else {
4518 			ptr = uptr;
4519 			error = 0;
4520 		}
4521 		if (error != 0) {
4522 			KFREES(ptr, fp->fr_dsize);
4523 			return EFAULT;
4524 		}
4525 		fp->fr_data = ptr;
4526 	} else
4527 		fp->fr_data = NULL;
4528 
4529 	/*
4530 	 * Perform per-rule type sanity checks of their members.
4531 	 */
4532 	switch (fp->fr_type & ~FR_T_BUILTIN)
4533 	{
4534 #if defined(IPFILTER_BPF)
4535 	case FR_T_BPFOPC :
4536 		if (fp->fr_dsize == 0)
4537 			return EINVAL;
4538 		if (!bpf_validate(ptr, fp->fr_dsize/sizeof(struct bpf_insn))) {
4539 			if (makecopy && fp->fr_data != NULL) {
4540 				KFREES(fp->fr_data, fp->fr_dsize);
4541 			}
4542 			return EINVAL;
4543 		}
4544 		break;
4545 #endif
4546 	case FR_T_IPF :
4547 		if (fp->fr_dsize != sizeof(fripf_t)) {
4548 			if (makecopy && fp->fr_data != NULL) {
4549 				KFREES(fp->fr_data, fp->fr_dsize);
4550 			}
4551 			return EINVAL;
4552 		}
4553 
4554 		/*
4555 		 * Allowing a rule with both "keep state" and "with oow" is
4556 		 * pointless because adding a state entry to the table will
4557 		 * fail with the out of window (oow) flag set.
4558 		 */
4559 		if ((fp->fr_flags & FR_KEEPSTATE) && (fp->fr_flx & FI_OOW)) {
4560 			if (makecopy && fp->fr_data != NULL) {
4561 				KFREES(fp->fr_data, fp->fr_dsize);
4562 			}
4563 			return EINVAL;
4564 		}
4565 
4566 		switch (fp->fr_satype)
4567 		{
4568 		case FRI_BROADCAST :
4569 		case FRI_DYNAMIC :
4570 		case FRI_NETWORK :
4571 		case FRI_NETMASKED :
4572 		case FRI_PEERADDR :
4573 			if (fp->fr_sifpidx < 0 || fp->fr_sifpidx > 3) {
4574 				if (makecopy && fp->fr_data != NULL) {
4575 					KFREES(fp->fr_data, fp->fr_dsize);
4576 				}
4577 				return EINVAL;
4578 			}
4579 			break;
4580 #ifdef	IPFILTER_LOOKUP
4581 		case FRI_LOOKUP :
4582 			fp->fr_srcptr = fr_resolvelookup(fp->fr_srctype,
4583 							 fp->fr_srcnum,
4584 							 &fp->fr_srcfunc, ifs);
4585 			break;
4586 #endif
4587 		default :
4588 			break;
4589 		}
4590 
4591 		switch (fp->fr_datype)
4592 		{
4593 		case FRI_BROADCAST :
4594 		case FRI_DYNAMIC :
4595 		case FRI_NETWORK :
4596 		case FRI_NETMASKED :
4597 		case FRI_PEERADDR :
4598 			if (fp->fr_difpidx < 0 || fp->fr_difpidx > 3) {
4599 				if (makecopy && fp->fr_data != NULL) {
4600 					KFREES(fp->fr_data, fp->fr_dsize);
4601 				}
4602 				return EINVAL;
4603 			}
4604 			break;
4605 #ifdef	IPFILTER_LOOKUP
4606 		case FRI_LOOKUP :
4607 			fp->fr_dstptr = fr_resolvelookup(fp->fr_dsttype,
4608 							 fp->fr_dstnum,
4609 							 &fp->fr_dstfunc, ifs);
4610 			break;
4611 #endif
4612 		default :
4613 			break;
4614 		}
4615 		break;
4616 	case FR_T_NONE :
4617 		break;
4618 	case FR_T_CALLFUNC :
4619 		break;
4620 	case FR_T_COMPIPF :
4621 		break;
4622 	default :
4623 		if (makecopy && fp->fr_data != NULL) {
4624 			KFREES(fp->fr_data, fp->fr_dsize);
4625 		}
4626 		return EINVAL;
4627 	}
4628 
4629 	/*
4630 	 * Lookup all the interface names that are part of the rule.
4631 	 */
4632 	frsynclist(0, 0, NULL, NULL, fp, ifs);
4633 	fp->fr_statecnt = 0;
4634 
4635 	/*
4636 	 * Look for an existing matching filter rule, but don't include the
4637 	 * next or interface pointer in the comparison (fr_next, fr_ifa).
4638 	 * This elminates rules which are indentical being loaded.  Checksum
4639 	 * the constant part of the filter rule to make comparisons quicker
4640 	 * (this meaning no pointers are included).
4641 	 */
4642 	for (fp->fr_cksum = 0, p = (u_int *)&fp->fr_func, pp = &fp->fr_cksum;
4643 	     p < pp; p++)
4644 		fp->fr_cksum += *p;
4645 	pp = (u_int *)(fp->fr_caddr + fp->fr_dsize);
4646 	for (p = (u_int *)fp->fr_data; p < pp; p++)
4647 		fp->fr_cksum += *p;
4648 
4649 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
4650 	bzero((char *)ifs->ifs_frcache, sizeof (ifs->ifs_frcache));
4651 
4652 	for (; (f = *ftail) != NULL; ftail = &f->fr_next) {
4653 		if ((fp->fr_cksum != f->fr_cksum) ||
4654 		    (f->fr_dsize != fp->fr_dsize))
4655 			continue;
4656 		if (bcmp((char *)&f->fr_func, (char *)&fp->fr_func, FR_CMPSIZ))
4657 			continue;
4658 		if ((!ptr && !f->fr_data) ||
4659 		    (ptr && f->fr_data &&
4660 		     !bcmp((char *)ptr, (char *)f->fr_data, f->fr_dsize)))
4661 			break;
4662 	}
4663 
4664 	/*
4665 	 * If zero'ing statistics, copy current to caller and zero.
4666 	 */
4667 	if (req == (ioctlcmd_t)SIOCZRLST) {
4668 		if (f == NULL)
4669 			error = ESRCH;
4670 		else {
4671 			/*
4672 			 * Copy and reduce lock because of impending copyout.
4673 			 * Well we should, but if we do then the atomicity of
4674 			 * this call and the correctness of fr_hits and
4675 			 * fr_bytes cannot be guaranteed.  As it is, this code
4676 			 * only resets them to 0 if they are successfully
4677 			 * copied out into user space.
4678 			 */
4679 			bcopy((char *)f, (char *)fp, sizeof(*f));
4680 
4681 			/*
4682 			 * When we copy this rule back out, set the data
4683 			 * pointer to be what it was in user space.
4684 			 */
4685 			fp->fr_data = uptr;
4686 			error = fr_outobj(data, fp, IPFOBJ_FRENTRY);
4687 
4688 			if (error == 0) {
4689 				if ((f->fr_dsize != 0) && (uptr != NULL))
4690 					error = COPYOUT(f->fr_data, uptr,
4691 							f->fr_dsize);
4692 				if (error == 0) {
4693 					f->fr_hits = 0;
4694 					f->fr_bytes = 0;
4695 				}
4696 			}
4697 		}
4698 
4699 		if ((ptr != NULL) && (makecopy != 0)) {
4700 			KFREES(ptr, fp->fr_dsize);
4701 		}
4702 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4703 		return error;
4704 	}
4705 
4706 	if (!f) {
4707 		/*
4708 		 * At the end of this, ftail must point to the place where the
4709 		 * new rule is to be saved/inserted/added.
4710 		 * For SIOCAD*FR, this should be the last rule in the group of
4711 		 * rules that have equal fr_collect fields.
4712 		 * For SIOCIN*FR, ...
4713 		 */
4714 		if (req == (ioctlcmd_t)SIOCADAFR ||
4715 		    req == (ioctlcmd_t)SIOCADIFR) {
4716 
4717 			for (ftail = fprev; (f = *ftail) != NULL; ) {
4718 				if (f->fr_collect > fp->fr_collect)
4719 					break;
4720 				ftail = &f->fr_next;
4721 			}
4722 			f = NULL;
4723 			ptr = NULL;
4724 			error = 0;
4725 		} else if (req == (ioctlcmd_t)SIOCINAFR ||
4726 			   req == (ioctlcmd_t)SIOCINIFR) {
4727 			while ((f = *fprev) != NULL) {
4728 				if (f->fr_collect >= fp->fr_collect)
4729 					break;
4730 				fprev = &f->fr_next;
4731 			}
4732 			ftail = fprev;
4733 			if (fp->fr_hits != 0) {
4734 				while (fp->fr_hits && (f = *ftail)) {
4735 					if (f->fr_collect != fp->fr_collect)
4736 						break;
4737 					fprev = ftail;
4738 					ftail = &f->fr_next;
4739 					fp->fr_hits--;
4740 				}
4741 			}
4742 			f = NULL;
4743 			ptr = NULL;
4744 			error = 0;
4745 		}
4746 	}
4747 
4748 	/*
4749 	 * Request to remove a rule.
4750 	 */
4751 	if (req == (ioctlcmd_t)SIOCRMAFR || req == (ioctlcmd_t)SIOCRMIFR) {
4752 		if (!f)
4753 			error = ESRCH;
4754 		else {
4755 			/*
4756 			 * Do not allow activity from user space to interfere
4757 			 * with rules not loaded that way.
4758 			 */
4759 			if ((makecopy == 1) && !(f->fr_flags & FR_COPIED)) {
4760 				error = EPERM;
4761 				goto done;
4762 			}
4763 
4764 			/*
4765 			 * Return EBUSY if the rule is being reference by
4766 			 * something else (eg state information.
4767 			 */
4768 			if (f->fr_ref > 1) {
4769 				error = EBUSY;
4770 				goto done;
4771 			}
4772 #ifdef	IPFILTER_SCAN
4773 			if (f->fr_isctag[0] != '\0' &&
4774 			    (f->fr_isc != (struct ipscan *)-1))
4775 				ipsc_detachfr(f);
4776 #endif
4777 			if (unit == IPL_LOGAUTH) {
4778 				error = fr_preauthcmd(req, f, ftail, ifs);
4779 				goto done;
4780 			}
4781 			if (*f->fr_grhead != '\0')
4782 				fr_delgroup(f->fr_grhead, unit, set, ifs);
4783 			fr_fixskip(ftail, f, -1);
4784 			*ftail = f->fr_next;
4785 			f->fr_next = NULL;
4786 			(void)fr_derefrule(&f, ifs);
4787 		}
4788 	} else {
4789 		/*
4790 		 * Not removing, so we must be adding/inserting a rule.
4791 		 */
4792 		if (f)
4793 			error = EEXIST;
4794 		else {
4795 			if (unit == IPL_LOGAUTH) {
4796 				error = fr_preauthcmd(req, fp, ftail, ifs);
4797 				goto done;
4798 			}
4799 			if (makecopy) {
4800 				KMALLOC(f, frentry_t *);
4801 			} else
4802 				f = fp;
4803 			if (f != NULL) {
4804 				if (fp != f)
4805 					bcopy((char *)fp, (char *)f,
4806 					      sizeof(*f));
4807 				MUTEX_NUKE(&f->fr_lock);
4808 				MUTEX_INIT(&f->fr_lock, "filter rule lock");
4809 #ifdef	IPFILTER_SCAN
4810 				if (f->fr_isctag[0] != '\0' &&
4811 				    ipsc_attachfr(f))
4812 					f->fr_isc = (struct ipscan *)-1;
4813 #endif
4814 				f->fr_hits = 0;
4815 				if (makecopy != 0)
4816 					f->fr_ref = 1;
4817 				f->fr_next = *ftail;
4818 				*ftail = f;
4819 				if (req == (ioctlcmd_t)SIOCINIFR ||
4820 				    req == (ioctlcmd_t)SIOCINAFR)
4821 					fr_fixskip(ftail, f, 1);
4822 				f->fr_grp = NULL;
4823 				group = f->fr_grhead;
4824 				if (*group != '\0') {
4825 					fg = fr_addgroup(group, f, f->fr_flags,
4826 							 unit, set, ifs);
4827 					if (fg != NULL)
4828 						f->fr_grp = &fg->fg_start;
4829 				}
4830 			} else
4831 				error = ENOMEM;
4832 		}
4833 	}
4834 done:
4835 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
4836 	if ((ptr != NULL) && (error != 0) && (makecopy != 0)) {
4837 		KFREES(ptr, fp->fr_dsize);
4838 	}
4839 	return (error);
4840 }
4841 
4842 
4843 /* ------------------------------------------------------------------------ */
4844 /* Function:    fr_funcinit                                                 */
4845 /* Returns:     int - 0 == success, else ESRCH: cannot resolve rule details */
4846 /* Parameters:  fr(I) - pointer to filter rule                              */
4847 /*                                                                          */
4848 /* If a rule is a call rule, then check if the function it points to needs  */
4849 /* an init function to be called now the rule has been loaded.              */
4850 /* ------------------------------------------------------------------------ */
fr_funcinit(fr,ifs)4851 static int fr_funcinit(fr, ifs)
4852 frentry_t *fr;
4853 ipf_stack_t *ifs;
4854 {
4855 	ipfunc_resolve_t *ft;
4856 	int err;
4857 
4858 	err = ESRCH;
4859 
4860 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4861 		if (ft->ipfu_addr == fr->fr_func) {
4862 			err = 0;
4863 			if (ft->ipfu_init != NULL)
4864 				err = (*ft->ipfu_init)(fr, ifs);
4865 			break;
4866 		}
4867 	return err;
4868 }
4869 
4870 
4871 /* ------------------------------------------------------------------------ */
4872 /* Function:    fr_findfunc                                                 */
4873 /* Returns:     ipfunc_t - pointer to function if found, else NULL          */
4874 /* Parameters:  funcptr(I) - function pointer to lookup                     */
4875 /*                                                                          */
4876 /* Look for a function in the table of known functions.                     */
4877 /* ------------------------------------------------------------------------ */
fr_findfunc(funcptr)4878 static ipfunc_t fr_findfunc(funcptr)
4879 ipfunc_t funcptr;
4880 {
4881 	ipfunc_resolve_t *ft;
4882 
4883 	for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4884 		if (ft->ipfu_addr == funcptr)
4885 			return funcptr;
4886 	return NULL;
4887 }
4888 
4889 
4890 /* ------------------------------------------------------------------------ */
4891 /* Function:    fr_resolvefunc                                              */
4892 /* Returns:     int - 0 == success, else error                              */
4893 /* Parameters:  data(IO) - ioctl data pointer to ipfunc_resolve_t struct    */
4894 /*                                                                          */
4895 /* Copy in a ipfunc_resolve_t structure and then fill in the missing field. */
4896 /* This will either be the function name (if the pointer is set) or the     */
4897 /* function pointer if the name is set.  When found, fill in the other one  */
4898 /* so that the entire, complete, structure can be copied back to user space.*/
4899 /* ------------------------------------------------------------------------ */
fr_resolvefunc(data)4900 int fr_resolvefunc(data)
4901 void *data;
4902 {
4903 	ipfunc_resolve_t res, *ft;
4904 	int err;
4905 
4906 	err = BCOPYIN(data, &res, sizeof(res));
4907 	if (err != 0)
4908 		return EFAULT;
4909 
4910 	if (res.ipfu_addr == NULL && res.ipfu_name[0] != '\0') {
4911 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4912 			if (strncmp(res.ipfu_name, ft->ipfu_name,
4913 				    sizeof(res.ipfu_name)) == 0) {
4914 				res.ipfu_addr = ft->ipfu_addr;
4915 				res.ipfu_init = ft->ipfu_init;
4916 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4917 					return EFAULT;
4918 				return 0;
4919 			}
4920 	}
4921 	if (res.ipfu_addr != NULL && res.ipfu_name[0] == '\0') {
4922 		for (ft = fr_availfuncs; ft->ipfu_addr != NULL; ft++)
4923 			if (ft->ipfu_addr == res.ipfu_addr) {
4924 				(void) strncpy(res.ipfu_name, ft->ipfu_name,
4925 					       sizeof(res.ipfu_name));
4926 				res.ipfu_init = ft->ipfu_init;
4927 				if (COPYOUT(&res, data, sizeof(res)) != 0)
4928 					return EFAULT;
4929 				return 0;
4930 			}
4931 	}
4932 	return ESRCH;
4933 }
4934 
4935 
4936 #if !defined(_KERNEL) || (!defined(__NetBSD__) && !defined(__OpenBSD__) && !defined(__FreeBSD__)) || \
4937     (defined(__FreeBSD__) && (__FreeBSD_version < 490000)) || \
4938     (defined(__NetBSD__) && (__NetBSD_Version__ < 105000000)) || \
4939     (defined(__OpenBSD__) && (OpenBSD < 200006))
4940 /*
4941  * From: NetBSD
4942  * ppsratecheck(): packets (or events) per second limitation.
4943  */
4944 int
ppsratecheck(lasttime,curpps,maxpps)4945 ppsratecheck(lasttime, curpps, maxpps)
4946 	struct timeval *lasttime;
4947 	int *curpps;
4948 	int maxpps;	/* maximum pps allowed */
4949 {
4950 	struct timeval tv, delta;
4951 	int rv;
4952 
4953 	GETKTIME(&tv);
4954 
4955 	delta.tv_sec = tv.tv_sec - lasttime->tv_sec;
4956 	delta.tv_usec = tv.tv_usec - lasttime->tv_usec;
4957 	if (delta.tv_usec < 0) {
4958 		delta.tv_sec--;
4959 		delta.tv_usec += 1000000;
4960 	}
4961 
4962 	/*
4963 	 * check for 0,0 is so that the message will be seen at least once.
4964 	 * if more than one second have passed since the last update of
4965 	 * lasttime, reset the counter.
4966 	 *
4967 	 * we do increment *curpps even in *curpps < maxpps case, as some may
4968 	 * try to use *curpps for stat purposes as well.
4969 	 */
4970 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
4971 	    delta.tv_sec >= 1) {
4972 		*lasttime = tv;
4973 		*curpps = 0;
4974 		rv = 1;
4975 	} else if (maxpps < 0)
4976 		rv = 1;
4977 	else if (*curpps < maxpps)
4978 		rv = 1;
4979 	else
4980 		rv = 0;
4981 	*curpps = *curpps + 1;
4982 
4983 	return (rv);
4984 }
4985 #endif
4986 
4987 
4988 /* ------------------------------------------------------------------------ */
4989 /* Function:    fr_derefrule                                                */
4990 /* Returns:     int   - 0 == rule freed up, else rule not freed             */
4991 /* Parameters:  fr(I) - pointer to filter rule                              */
4992 /*                                                                          */
4993 /* Decrement the reference counter to a rule by one.  If it reaches zero,   */
4994 /* free it and any associated storage space being used by it.               */
4995 /* ------------------------------------------------------------------------ */
fr_derefrule(frp,ifs)4996 int fr_derefrule(frp, ifs)
4997 frentry_t **frp;
4998 ipf_stack_t *ifs;
4999 {
5000 	frentry_t *fr;
5001 
5002 	fr = *frp;
5003 
5004 	MUTEX_ENTER(&fr->fr_lock);
5005 	fr->fr_ref--;
5006 	if (fr->fr_ref == 0) {
5007 		MUTEX_EXIT(&fr->fr_lock);
5008 		MUTEX_DESTROY(&fr->fr_lock);
5009 
5010 #ifdef IPFILTER_LOOKUP
5011 		if (fr->fr_type == FR_T_IPF && fr->fr_satype == FRI_LOOKUP)
5012 		    ip_lookup_deref(fr->fr_srctype, fr->fr_srcptr, ifs);
5013 		if (fr->fr_type == FR_T_IPF && fr->fr_datype == FRI_LOOKUP)
5014 		    ip_lookup_deref(fr->fr_dsttype, fr->fr_dstptr, ifs);
5015 #endif
5016 
5017 		if (fr->fr_dsize) {
5018 			KFREES(fr->fr_data, fr->fr_dsize);
5019 		}
5020 		if ((fr->fr_flags & FR_COPIED) != 0) {
5021 			KFREE(fr);
5022 			return 0;
5023 		}
5024 		return 1;
5025 	} else {
5026 		MUTEX_EXIT(&fr->fr_lock);
5027 	}
5028 	*frp = NULL;
5029 	return -1;
5030 }
5031 
5032 
5033 #ifdef	IPFILTER_LOOKUP
5034 /* ------------------------------------------------------------------------ */
5035 /* Function:    fr_grpmapinit                                               */
5036 /* Returns:     int - 0 == success, else ESRCH because table entry not found*/
5037 /* Parameters:  fr(I) - pointer to rule to find hash table for              */
5038 /*                                                                          */
5039 /* Looks for group hash table fr_arg and stores a pointer to it in fr_ptr.  */
5040 /* fr_ptr is later used by fr_srcgrpmap and fr_dstgrpmap.                   */
5041 /* ------------------------------------------------------------------------ */
fr_grpmapinit(fr,ifs)5042 static int fr_grpmapinit(fr, ifs)
5043 frentry_t *fr;
5044 ipf_stack_t *ifs;
5045 {
5046 	char name[FR_GROUPLEN];
5047 	iphtable_t *iph;
5048 
5049 #if defined(SNPRINTF) && defined(_KERNEL)
5050 	(void) SNPRINTF(name, sizeof(name), "%d", fr->fr_arg);
5051 #else
5052 	(void) sprintf(name, "%d", fr->fr_arg);
5053 #endif
5054 	iph = fr_findhtable(IPL_LOGIPF, name, ifs);
5055 	if (iph == NULL)
5056 		return ESRCH;
5057 	if ((iph->iph_flags & FR_INOUT) != (fr->fr_flags & FR_INOUT))
5058 		return ESRCH;
5059 	fr->fr_ptr = iph;
5060 	return 0;
5061 }
5062 
5063 
5064 /* ------------------------------------------------------------------------ */
5065 /* Function:    fr_srcgrpmap                                                */
5066 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5067 /* Parameters:  fin(I)    - pointer to packet information                   */
5068 /*              passp(IO) - pointer to current/new filter decision (unused) */
5069 /*                                                                          */
5070 /* Look for a rule group head in a hash table, using the source address as  */
5071 /* the key, and descend into that group and continue matching rules against */
5072 /* the packet.                                                              */
5073 /* ------------------------------------------------------------------------ */
fr_srcgrpmap(fin,passp)5074 frentry_t *fr_srcgrpmap(fin, passp)
5075 fr_info_t *fin;
5076 u_32_t *passp;
5077 {
5078 	frgroup_t *fg;
5079 	void *rval;
5080 	ipf_stack_t *ifs = fin->fin_ifs;
5081 
5082 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_src, ifs);
5083 	if (rval == NULL)
5084 		return NULL;
5085 
5086 	fg = rval;
5087 	fin->fin_fr = fg->fg_start;
5088 	(void) fr_scanlist(fin, *passp);
5089 	return fin->fin_fr;
5090 }
5091 
5092 
5093 /* ------------------------------------------------------------------------ */
5094 /* Function:    fr_dstgrpmap                                                */
5095 /* Returns:     frentry_t * - pointer to "new last matching" rule or NULL   */
5096 /* Parameters:  fin(I)    - pointer to packet information                   */
5097 /*              passp(IO) - pointer to current/new filter decision (unused) */
5098 /*                                                                          */
5099 /* Look for a rule group head in a hash table, using the destination        */
5100 /* address as the key, and descend into that group and continue matching    */
5101 /* rules against  the packet.                                               */
5102 /* ------------------------------------------------------------------------ */
fr_dstgrpmap(fin,passp)5103 frentry_t *fr_dstgrpmap(fin, passp)
5104 fr_info_t *fin;
5105 u_32_t *passp;
5106 {
5107 	frgroup_t *fg;
5108 	void *rval;
5109 	ipf_stack_t *ifs = fin->fin_ifs;
5110 
5111 	rval = fr_iphmfindgroup(fin->fin_fr->fr_ptr, fin->fin_v, &fin->fin_dst, ifs);
5112 	if (rval == NULL)
5113 		return NULL;
5114 
5115 	fg = rval;
5116 	fin->fin_fr = fg->fg_start;
5117 	(void) fr_scanlist(fin, *passp);
5118 	return fin->fin_fr;
5119 }
5120 #endif /* IPFILTER_LOOKUP */
5121 
5122 /*
5123  * Queue functions
5124  * ===============
5125  * These functions manage objects on queues for efficient timeouts.  There are
5126  * a number of system defined queues as well as user defined timeouts.  It is
5127  * expected that a lock is held in the domain in which the queue belongs
5128  * (i.e. either state or NAT) when calling any of these functions that prevents
5129  * fr_freetimeoutqueue() from being called at the same time as any other.
5130  */
5131 
5132 
5133 /* ------------------------------------------------------------------------ */
5134 /* Function:    fr_addtimeoutqueue                                          */
5135 /* Returns:     struct ifqtq * - NULL if malloc fails, else pointer to      */
5136 /*                               timeout queue with given interval.         */
5137 /* Parameters:  parent(I)  - pointer to pointer to parent node of this list */
5138 /*                           of interface queues.                           */
5139 /*              seconds(I) - timeout value in seconds for this queue.       */
5140 /*                                                                          */
5141 /* This routine first looks for a timeout queue that matches the interval   */
5142 /* being requested.  If it finds one, increments the reference counter and  */
5143 /* returns a pointer to it.  If none are found, it allocates a new one and  */
5144 /* inserts it at the top of the list.                                       */
5145 /*                                                                          */
5146 /* Locking.                                                                 */
5147 /* It is assumed that the caller of this function has an appropriate lock   */
5148 /* held (exclusively) in the domain that encompases 'parent'.               */
5149 /* ------------------------------------------------------------------------ */
fr_addtimeoutqueue(parent,seconds,ifs)5150 ipftq_t *fr_addtimeoutqueue(parent, seconds, ifs)
5151 ipftq_t **parent;
5152 u_int seconds;
5153 ipf_stack_t *ifs;
5154 {
5155 	ipftq_t *ifq;
5156 	u_int period;
5157 
5158 	period = seconds * IPF_HZ_DIVIDE;
5159 
5160 	MUTEX_ENTER(&ifs->ifs_ipf_timeoutlock);
5161 	for (ifq = *parent; ifq != NULL; ifq = ifq->ifq_next) {
5162 		if (ifq->ifq_ttl == period) {
5163 			/*
5164 			 * Reset the delete flag, if set, so the structure
5165 			 * gets reused rather than freed and reallocated.
5166 			 */
5167 			MUTEX_ENTER(&ifq->ifq_lock);
5168 			ifq->ifq_flags &= ~IFQF_DELETE;
5169 			ifq->ifq_ref++;
5170 			MUTEX_EXIT(&ifq->ifq_lock);
5171 			MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5172 
5173 			return ifq;
5174 		}
5175 	}
5176 
5177 	KMALLOC(ifq, ipftq_t *);
5178 	if (ifq != NULL) {
5179 		ifq->ifq_ttl = period;
5180 		ifq->ifq_head = NULL;
5181 		ifq->ifq_tail = &ifq->ifq_head;
5182 		ifq->ifq_next = *parent;
5183 		ifq->ifq_pnext = parent;
5184 		ifq->ifq_ref = 1;
5185 		ifq->ifq_flags = IFQF_USER;
5186 		*parent = ifq;
5187 		ifs->ifs_fr_userifqs++;
5188 		MUTEX_NUKE(&ifq->ifq_lock);
5189 		MUTEX_INIT(&ifq->ifq_lock, "ipftq mutex");
5190 	}
5191 	MUTEX_EXIT(&ifs->ifs_ipf_timeoutlock);
5192 	return ifq;
5193 }
5194 
5195 
5196 /* ------------------------------------------------------------------------ */
5197 /* Function:    fr_deletetimeoutqueue                                       */
5198 /* Returns:     int    - new reference count value of the timeout queue     */
5199 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5200 /* Locks:       ifq->ifq_lock                                               */
5201 /*                                                                          */
5202 /* This routine must be called when we're discarding a pointer to a timeout */
5203 /* queue object, taking care of the reference counter.                      */
5204 /*                                                                          */
5205 /* Now that this just sets a DELETE flag, it requires the expire code to    */
5206 /* check the list of user defined timeout queues and call the free function */
5207 /* below (currently commented out) to stop memory leaking.  It is done this */
5208 /* way because the locking may not be sufficient to safely do a free when   */
5209 /* this function is called.                                                 */
5210 /* ------------------------------------------------------------------------ */
fr_deletetimeoutqueue(ifq)5211 int fr_deletetimeoutqueue(ifq)
5212 ipftq_t *ifq;
5213 {
5214 
5215 	ifq->ifq_ref--;
5216 	if ((ifq->ifq_ref == 0) && ((ifq->ifq_flags & IFQF_USER) != 0)) {
5217 		ifq->ifq_flags |= IFQF_DELETE;
5218 	}
5219 
5220 	return ifq->ifq_ref;
5221 }
5222 
5223 
5224 /* ------------------------------------------------------------------------ */
5225 /* Function:    fr_freetimeoutqueue                                         */
5226 /* Parameters:  ifq(I) - timeout queue which is losing a reference.         */
5227 /* Returns:     Nil                                                         */
5228 /*                                                                          */
5229 /* Locking:                                                                 */
5230 /* It is assumed that the caller of this function has an appropriate lock   */
5231 /* held (exclusively) in the domain that encompases the callers "domain".   */
5232 /* The ifq_lock for this structure should not be held.                      */
5233 /*                                                                          */
5234 /* Remove a user definde timeout queue from the list of queues it is in and */
5235 /* tidy up after this is done.                                              */
5236 /* ------------------------------------------------------------------------ */
fr_freetimeoutqueue(ifq,ifs)5237 void fr_freetimeoutqueue(ifq, ifs)
5238 ipftq_t *ifq;
5239 ipf_stack_t *ifs;
5240 {
5241 
5242 
5243 	if (((ifq->ifq_flags & IFQF_DELETE) == 0) || (ifq->ifq_ref != 0) ||
5244 	    ((ifq->ifq_flags & IFQF_USER) == 0)) {
5245 		printf("fr_freetimeoutqueue(%lx) flags 0x%x ttl %d ref %d\n",
5246 		       (u_long)ifq, ifq->ifq_flags, ifq->ifq_ttl,
5247 		       ifq->ifq_ref);
5248 		return;
5249 	}
5250 
5251 	/*
5252 	 * Remove from its position in the list.
5253 	 */
5254 	*ifq->ifq_pnext = ifq->ifq_next;
5255 	if (ifq->ifq_next != NULL)
5256 		ifq->ifq_next->ifq_pnext = ifq->ifq_pnext;
5257 
5258 	MUTEX_DESTROY(&ifq->ifq_lock);
5259 	ifs->ifs_fr_userifqs--;
5260 	KFREE(ifq);
5261 }
5262 
5263 
5264 /* ------------------------------------------------------------------------ */
5265 /* Function:    fr_deletequeueentry                                         */
5266 /* Returns:     Nil                                                         */
5267 /* Parameters:  tqe(I) - timeout queue entry to delete                      */
5268 /*              ifq(I) - timeout queue to remove entry from                 */
5269 /*                                                                          */
5270 /* Remove a tail queue entry from its queue and make it an orphan.          */
5271 /* fr_deletetimeoutqueue is called to make sure the reference count on the  */
5272 /* queue is correct.  We can't, however, call fr_freetimeoutqueue because   */
5273 /* the correct lock(s) may not be held that would make it safe to do so.    */
5274 /* ------------------------------------------------------------------------ */
fr_deletequeueentry(tqe)5275 void fr_deletequeueentry(tqe)
5276 ipftqent_t *tqe;
5277 {
5278 	ipftq_t *ifq;
5279 
5280 	ifq = tqe->tqe_ifq;
5281 	if (ifq == NULL)
5282 		return;
5283 
5284 	MUTEX_ENTER(&ifq->ifq_lock);
5285 
5286 	if (tqe->tqe_pnext != NULL) {
5287 		*tqe->tqe_pnext = tqe->tqe_next;
5288 		if (tqe->tqe_next != NULL)
5289 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5290 		else    /* we must be the tail anyway */
5291 			ifq->ifq_tail = tqe->tqe_pnext;
5292 
5293 		tqe->tqe_pnext = NULL;
5294 		tqe->tqe_ifq = NULL;
5295 	}
5296 
5297 	(void) fr_deletetimeoutqueue(ifq);
5298 
5299 	MUTEX_EXIT(&ifq->ifq_lock);
5300 }
5301 
5302 
5303 /* ------------------------------------------------------------------------ */
5304 /* Function:    fr_queuefront                                               */
5305 /* Returns:     Nil                                                         */
5306 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5307 /*                                                                          */
5308 /* Move a queue entry to the front of the queue, if it isn't already there. */
5309 /* ------------------------------------------------------------------------ */
fr_queuefront(tqe)5310 void fr_queuefront(tqe)
5311 ipftqent_t *tqe;
5312 {
5313 	ipftq_t *ifq;
5314 
5315 	ifq = tqe->tqe_ifq;
5316 	if (ifq == NULL)
5317 		return;
5318 
5319 	MUTEX_ENTER(&ifq->ifq_lock);
5320 	if (ifq->ifq_head != tqe) {
5321 		*tqe->tqe_pnext = tqe->tqe_next;
5322 		if (tqe->tqe_next)
5323 			tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5324 		else
5325 			ifq->ifq_tail = tqe->tqe_pnext;
5326 
5327 		tqe->tqe_next = ifq->ifq_head;
5328 		ifq->ifq_head->tqe_pnext = &tqe->tqe_next;
5329 		ifq->ifq_head = tqe;
5330 		tqe->tqe_pnext = &ifq->ifq_head;
5331 	}
5332 	MUTEX_EXIT(&ifq->ifq_lock);
5333 }
5334 
5335 
5336 /* ------------------------------------------------------------------------ */
5337 /* Function:    fr_queueback                                                */
5338 /* Returns:     Nil                                                         */
5339 /* Parameters:  tqe(I) - pointer to timeout queue entry                     */
5340 /*                                                                          */
5341 /* Move a queue entry to the back of the queue, if it isn't already there.  */
5342 /* ------------------------------------------------------------------------ */
fr_queueback(tqe,ifs)5343 void fr_queueback(tqe, ifs)
5344 ipftqent_t *tqe;
5345 ipf_stack_t *ifs;
5346 {
5347 	ipftq_t *ifq;
5348 
5349 	ifq = tqe->tqe_ifq;
5350 	if (ifq == NULL)
5351 		return;
5352 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5353 
5354 	MUTEX_ENTER(&ifq->ifq_lock);
5355 	if (tqe->tqe_next == NULL) {		/* at the end already ? */
5356 		MUTEX_EXIT(&ifq->ifq_lock);
5357 		return;
5358 	}
5359 
5360 	/*
5361 	 * Remove from list
5362 	 */
5363 	*tqe->tqe_pnext = tqe->tqe_next;
5364 	tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5365 
5366 	/*
5367 	 * Make it the last entry.
5368 	 */
5369 	tqe->tqe_next = NULL;
5370 	tqe->tqe_pnext = ifq->ifq_tail;
5371 	*ifq->ifq_tail = tqe;
5372 	ifq->ifq_tail = &tqe->tqe_next;
5373 	MUTEX_EXIT(&ifq->ifq_lock);
5374 }
5375 
5376 
5377 /* ------------------------------------------------------------------------ */
5378 /* Function:    fr_queueappend                                              */
5379 /* Returns:     Nil                                                         */
5380 /* Parameters:  tqe(I)    - pointer to timeout queue entry                  */
5381 /*              ifq(I)    - pointer to timeout queue                        */
5382 /*              parent(I) - owing object pointer                            */
5383 /*                                                                          */
5384 /* Add a new item to this queue and put it on the very end.                 */
5385 /* ------------------------------------------------------------------------ */
fr_queueappend(tqe,ifq,parent,ifs)5386 void fr_queueappend(tqe, ifq, parent, ifs)
5387 ipftqent_t *tqe;
5388 ipftq_t *ifq;
5389 void *parent;
5390 ipf_stack_t *ifs;
5391 {
5392 
5393 	MUTEX_ENTER(&ifq->ifq_lock);
5394 	tqe->tqe_parent = parent;
5395 	tqe->tqe_pnext = ifq->ifq_tail;
5396 	*ifq->ifq_tail = tqe;
5397 	ifq->ifq_tail = &tqe->tqe_next;
5398 	tqe->tqe_next = NULL;
5399 	tqe->tqe_ifq = ifq;
5400 	tqe->tqe_die = ifs->ifs_fr_ticks + ifq->ifq_ttl;
5401 	ifq->ifq_ref++;
5402 	MUTEX_EXIT(&ifq->ifq_lock);
5403 }
5404 
5405 
5406 /* ------------------------------------------------------------------------ */
5407 /* Function:    fr_movequeue                                                */
5408 /* Returns:     Nil                                                         */
5409 /* Parameters:  tq(I)   - pointer to timeout queue information              */
5410 /*              oifp(I) - old timeout queue entry was on                    */
5411 /*              nifp(I) - new timeout queue to put entry on                 */
5412 /*		ifs	- ipf stack instance				    */
5413 /*                                                                          */
5414 /* Move a queue entry from one timeout queue to another timeout queue.      */
5415 /* If it notices that the current entry is already last and does not need   */
5416 /* to move queue, the return.                                               */
5417 /* ------------------------------------------------------------------------ */
fr_movequeue(tqe,oifq,nifq,ifs)5418 void fr_movequeue(tqe, oifq, nifq, ifs)
5419 ipftqent_t *tqe;
5420 ipftq_t *oifq, *nifq;
5421 ipf_stack_t *ifs;
5422 {
5423 	/*
5424 	 * If the queue isn't changing, and the clock hasn't ticked
5425 	 * since the last update, the operation will be a no-op.
5426 	 */
5427 	if (oifq == nifq && tqe->tqe_touched == ifs->ifs_fr_ticks)
5428 		return;
5429 
5430 	/*
5431 	 * Grab the lock and update the timers.
5432 	 */
5433 	MUTEX_ENTER(&oifq->ifq_lock);
5434 	tqe->tqe_touched = ifs->ifs_fr_ticks;
5435 	tqe->tqe_die = ifs->ifs_fr_ticks + nifq->ifq_ttl;
5436 
5437 	/*
5438 	 * The remainder of the operation can still be a no-op.
5439 	 *
5440 	 * If the queue isn't changing, check to see if
5441 	 * an update would be meaningless.
5442 	 */
5443 	if (oifq == nifq) {
5444 		if ((tqe->tqe_next == NULL) ||
5445 		    (tqe->tqe_next->tqe_die == tqe->tqe_die)) {
5446 			MUTEX_EXIT(&oifq->ifq_lock);
5447 			return;
5448 		}
5449 	}
5450 
5451 	/*
5452 	 * Remove from the old queue
5453 	 */
5454 	*tqe->tqe_pnext = tqe->tqe_next;
5455 	if (tqe->tqe_next)
5456 		tqe->tqe_next->tqe_pnext = tqe->tqe_pnext;
5457 	else
5458 		oifq->ifq_tail = tqe->tqe_pnext;
5459 	tqe->tqe_next = NULL;
5460 
5461 	/*
5462 	 * If we're moving from one queue to another, release the lock on the
5463 	 * old queue and get a lock on the new queue.  For user defined queues,
5464 	 * if we're moving off it, call delete in case it can now be freed.
5465 	 */
5466 	if (oifq != nifq) {
5467 		tqe->tqe_ifq = NULL;
5468 
5469 		(void) fr_deletetimeoutqueue(oifq);
5470 
5471 		MUTEX_EXIT(&oifq->ifq_lock);
5472 
5473 		MUTEX_ENTER(&nifq->ifq_lock);
5474 
5475 		tqe->tqe_ifq = nifq;
5476 		nifq->ifq_ref++;
5477 	}
5478 
5479 	/*
5480 	 * Add to the bottom of the new queue
5481 	 */
5482 	tqe->tqe_pnext = nifq->ifq_tail;
5483 	*nifq->ifq_tail = tqe;
5484 	nifq->ifq_tail = &tqe->tqe_next;
5485 	MUTEX_EXIT(&nifq->ifq_lock);
5486 }
5487 
5488 
5489 /* ------------------------------------------------------------------------ */
5490 /* Function:    fr_updateipid                                               */
5491 /* Returns:     int - 0 == success, -1 == error (packet should be droppped) */
5492 /* Parameters:  fin(I) - pointer to packet information                      */
5493 /*                                                                          */
5494 /* When we are doing NAT, change the IP of every packet to represent a      */
5495 /* single sequence of packets coming from the host, hiding any host         */
5496 /* specific sequencing that might otherwise be revealed.  If the packet is  */
5497 /* a fragment, then store the 'new' IPid in the fragment cache and look up  */
5498 /* the fragment cache for non-leading fragments.  If a non-leading fragment */
5499 /* has no match in the cache, return an error.                              */
5500 /* ------------------------------------------------------------------------ */
fr_updateipid(fin)5501 static INLINE int fr_updateipid(fin)
5502 fr_info_t *fin;
5503 {
5504 	u_short id, ido, sums;
5505 	u_32_t sumd, sum;
5506 	ip_t *ip;
5507 
5508 	if (fin->fin_off != 0) {
5509 		sum = fr_ipid_knownfrag(fin);
5510 		if (sum == 0xffffffff)
5511 			return -1;
5512 		sum &= 0xffff;
5513 		id = (u_short)sum;
5514 	} else {
5515 		id = fr_nextipid(fin);
5516 		if (fin->fin_off == 0 && (fin->fin_flx & FI_FRAG) != 0)
5517 			(void) fr_ipid_newfrag(fin, (u_32_t)id);
5518 	}
5519 
5520 	ip = fin->fin_ip;
5521 	ido = ntohs(ip->ip_id);
5522 	if (id == ido)
5523 		return 0;
5524 	ip->ip_id = htons(id);
5525 	CALC_SUMD(ido, id, sumd);	/* DESTRUCTIVE MACRO! id,ido change */
5526 	sum = (~ntohs(ip->ip_sum)) & 0xffff;
5527 	sum += sumd;
5528 	sum = (sum >> 16) + (sum & 0xffff);
5529 	sum = (sum >> 16) + (sum & 0xffff);
5530 	sums = ~(u_short)sum;
5531 	ip->ip_sum = htons(sums);
5532 	return 0;
5533 }
5534 
5535 
5536 #ifdef	NEED_FRGETIFNAME
5537 /* ------------------------------------------------------------------------ */
5538 /* Function:    fr_getifname                                                */
5539 /* Returns:     char *    - pointer to interface name                       */
5540 /* Parameters:  ifp(I)    - pointer to network interface                    */
5541 /*              buffer(O) - pointer to where to store interface name        */
5542 /*                                                                          */
5543 /* Constructs an interface name in the buffer passed.  The buffer passed is */
5544 /* expected to be at least LIFNAMSIZ in bytes big.  If buffer is passed in  */
5545 /* as a NULL pointer then return a pointer to a static array.               */
5546 /* ------------------------------------------------------------------------ */
fr_getifname(ifp,buffer)5547 char *fr_getifname(ifp, buffer)
5548 struct ifnet *ifp;
5549 char *buffer;
5550 {
5551 	static char namebuf[LIFNAMSIZ];
5552 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5553      defined(__sgi) || defined(linux) || defined(_AIX51) || \
5554      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5555 	int unit, space;
5556 	char temp[20];
5557 	char *s;
5558 # endif
5559 
5560 	ASSERT(buffer != NULL);
5561 #ifdef notdef
5562 	if (buffer == NULL)
5563 		buffer = namebuf;
5564 #endif
5565 	(void) strncpy(buffer, ifp->if_name, LIFNAMSIZ);
5566 	buffer[LIFNAMSIZ - 1] = '\0';
5567 # if defined(MENTAT) || defined(__FreeBSD__) || defined(__osf__) || \
5568      defined(__sgi) || defined(_AIX51) || \
5569      (defined(sun) && !defined(__SVR4) && !defined(__svr4__))
5570 	for (s = buffer; *s; s++)
5571 		;
5572 	unit = ifp->if_unit;
5573 	space = LIFNAMSIZ - (s - buffer);
5574 	if (space > 0) {
5575 #  if defined(SNPRINTF) && defined(_KERNEL)
5576 		(void) SNPRINTF(temp, sizeof(temp), "%d", unit);
5577 #  else
5578 		(void) sprintf(temp, "%d", unit);
5579 #  endif
5580 		(void) strncpy(s, temp, space);
5581 	}
5582 # endif
5583 	return buffer;
5584 }
5585 #endif
5586 
5587 
5588 /* ------------------------------------------------------------------------ */
5589 /* Function:    fr_ioctlswitch                                              */
5590 /* Returns:     int     - -1 continue processing, else ioctl return value   */
5591 /* Parameters:  unit(I) - device unit opened                                */
5592 /*              data(I) - pointer to ioctl data                             */
5593 /*              cmd(I)  - ioctl command                                     */
5594 /*              mode(I) - mode value                                        */
5595 /*                                                                          */
5596 /* Based on the value of unit, call the appropriate ioctl handler or return */
5597 /* EIO if ipfilter is not running.   Also checks if write perms are req'd   */
5598 /* for the device in order to execute the ioctl.                            */
5599 /* ------------------------------------------------------------------------ */
fr_ioctlswitch(unit,data,cmd,mode,uid,ctx,ifs)5600 INLINE int fr_ioctlswitch(unit, data, cmd, mode, uid, ctx, ifs)
5601 int unit, mode, uid;
5602 ioctlcmd_t cmd;
5603 void *data, *ctx;
5604 ipf_stack_t *ifs;
5605 {
5606 	int error = 0;
5607 
5608 	switch (unit)
5609 	{
5610 	case IPL_LOGIPF :
5611 		error = -1;
5612 		break;
5613 	case IPL_LOGNAT :
5614 		if (ifs->ifs_fr_running > 0)
5615 			error = fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs);
5616 		else
5617 			error = EIO;
5618 		break;
5619 	case IPL_LOGSTATE :
5620 		if (ifs->ifs_fr_running > 0)
5621 			error = fr_state_ioctl(data, cmd, mode, uid, ctx, ifs);
5622 		else
5623 			error = EIO;
5624 		break;
5625 	case IPL_LOGAUTH :
5626 		if (ifs->ifs_fr_running > 0) {
5627 			if ((cmd == (ioctlcmd_t)SIOCADAFR) ||
5628 			    (cmd == (ioctlcmd_t)SIOCRMAFR)) {
5629 				if (!(mode & FWRITE)) {
5630 					error = EPERM;
5631 				} else {
5632 					error = frrequest(unit, cmd, data,
5633 						  ifs->ifs_fr_active, 1, ifs);
5634 				}
5635 			} else {
5636 				error = fr_auth_ioctl(data, cmd, mode, uid, ctx, ifs);
5637 			}
5638 		} else
5639 			error = EIO;
5640 		break;
5641 	case IPL_LOGSYNC :
5642 #ifdef IPFILTER_SYNC
5643 		if (ifs->ifs_fr_running > 0)
5644 			error = fr_sync_ioctl(data, cmd, mode, ifs);
5645 		else
5646 #endif
5647 			error = EIO;
5648 		break;
5649 	case IPL_LOGSCAN :
5650 #ifdef IPFILTER_SCAN
5651 		if (ifs->ifs_fr_running > 0)
5652 			error = fr_scan_ioctl(data, cmd, mode, ifs);
5653 		else
5654 #endif
5655 			error = EIO;
5656 		break;
5657 	case IPL_LOGLOOKUP :
5658 #ifdef IPFILTER_LOOKUP
5659 		if (ifs->ifs_fr_running > 0)
5660 			error = ip_lookup_ioctl(data, cmd, mode, uid, ctx, ifs);
5661 		else
5662 #endif
5663 			error = EIO;
5664 		break;
5665 	default :
5666 		error = EIO;
5667 		break;
5668 	}
5669 
5670 	return error;
5671 }
5672 
5673 
5674 /*
5675  * This array defines the expected size of objects coming into the kernel
5676  * for the various recognised object types.
5677  */
5678 #define	NUM_OBJ_TYPES	19
5679 
5680 static	int	fr_objbytes[NUM_OBJ_TYPES][2] = {
5681 	{ 1,	sizeof(struct frentry) },		/* frentry */
5682 	{ 0,	sizeof(struct friostat) },
5683 	{ 0,	sizeof(struct fr_info) },
5684 	{ 0,	sizeof(struct fr_authstat) },
5685 	{ 0,	sizeof(struct ipfrstat) },
5686 	{ 0,	sizeof(struct ipnat) },
5687 	{ 0,	sizeof(struct natstat) },
5688 	{ 0,	sizeof(struct ipstate_save) },
5689 	{ 1,	sizeof(struct nat_save) },		/* nat_save */
5690 	{ 0,	sizeof(struct natlookup) },
5691 	{ 1,	sizeof(struct ipstate) },		/* ipstate */
5692 	{ 0,	sizeof(struct ips_stat) },
5693 	{ 0,	sizeof(struct frauth) },
5694 	{ 0,	sizeof(struct ipftune) },
5695 	{ 0,	sizeof(struct nat) },                   /* nat_t */
5696 	{ 0,	sizeof(struct ipfruleiter) },
5697 	{ 0,	sizeof(struct ipfgeniter) },
5698 	{ 0,	sizeof(struct ipftable) },
5699 	{ 0,	sizeof(struct ipflookupiter) }
5700 };
5701 
5702 
5703 /* ------------------------------------------------------------------------ */
5704 /* Function:    fr_getzoneid                                                */
5705 /* Returns:     int     - 0 = success, else failure                         */
5706 /* Parameters:  idsp(I) - pointer to ipf_devstate_t                         */
5707 /*              data(I) - pointer to ioctl data                             */
5708 /*                                                                          */
5709 /* Set the zone ID in idsp based on the zone name in ipfzoneobj.  Further   */
5710 /* ioctls will act on the IPF stack for that zone ID.                       */
5711 /* ------------------------------------------------------------------------ */
5712 #if defined(_KERNEL)
fr_setzoneid(idsp,data)5713 int fr_setzoneid(idsp, data)
5714 ipf_devstate_t *idsp;
5715 void *data;
5716 {
5717 	int error = 0;
5718 	ipfzoneobj_t ipfzo;
5719 	zone_t *zone;
5720 
5721 	error = BCOPYIN(data, &ipfzo, sizeof(ipfzo));
5722 	if (error != 0)
5723 		return EFAULT;
5724 
5725 	if (memchr(ipfzo.ipfz_zonename, '\0', ZONENAME_MAX) == NULL)
5726 		return EFAULT;
5727 
5728 	/*
5729 	 * The global zone doesn't have a GZ-controlled stack, so no
5730 	 * sense in going any further
5731 	 */
5732 	if (strcmp(ipfzo.ipfz_zonename, "global") == 0)
5733 		return ENODEV;
5734 
5735 	if ((zone = zone_find_by_name(ipfzo.ipfz_zonename)) == NULL)
5736 		return ENODEV;
5737 
5738 	/*
5739 	 * Store the zone ID that to control, and whether it's the
5740 	 * GZ-controlled stack that's wanted
5741 	 */
5742 	idsp->ipfs_zoneid = zone->zone_id;
5743 	idsp->ipfs_gz = (ipfzo.ipfz_gz == 1) ? B_TRUE : B_FALSE;
5744 	zone_rele(zone);
5745 
5746 	return error;
5747 }
5748 #endif
5749 
5750 
5751 /* ------------------------------------------------------------------------ */
5752 /* Function:    fr_inobj                                                    */
5753 /* Returns:     int     - 0 = success, else failure                         */
5754 /* Parameters:  data(I) - pointer to ioctl data                             */
5755 /*              ptr(I)  - pointer to store real data in                     */
5756 /*              type(I) - type of structure being moved                     */
5757 /*                                                                          */
5758 /* Copy in the contents of what the ipfobj_t points to.  In future, we      */
5759 /* add things to check for version numbers, sizes, etc, to make it backward */
5760 /* compatible at the ABI for user land.                                     */
5761 /* ------------------------------------------------------------------------ */
fr_inobj(data,ptr,type)5762 int fr_inobj(data, ptr, type)
5763 void *data;
5764 void *ptr;
5765 int type;
5766 {
5767 	ipfobj_t obj;
5768 	int error = 0;
5769 
5770 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5771 		return EINVAL;
5772 
5773 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5774 	if (error != 0)
5775 		return EFAULT;
5776 
5777 	if (obj.ipfo_type != type)
5778 		return EINVAL;
5779 
5780 #ifndef	IPFILTER_COMPAT
5781 	if ((fr_objbytes[type][0] & 1) != 0) {
5782 		if (obj.ipfo_size < fr_objbytes[type][1])
5783 			return EINVAL;
5784 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5785 		return EINVAL;
5786 #else
5787 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5788 		error = fr_incomptrans(&obj, ptr);
5789 		return error;
5790 	}
5791 
5792 	if ((fr_objbytes[type][0] & 1) != 0 &&
5793 	    obj.ipfo_size < fr_objbytes[type][1] ||
5794 	    obj.ipfo_size != fr_objbytes[type][1])
5795 		return EINVAL;
5796 #endif
5797 
5798 	if ((fr_objbytes[type][0] & 1) != 0) {
5799 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5800 				fr_objbytes[type][1]);
5801 	} else {
5802 		error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr,
5803 				obj.ipfo_size);
5804 	}
5805 	return error;
5806 }
5807 
5808 
5809 /* ------------------------------------------------------------------------ */
5810 /* Function:    fr_inobjsz                                                  */
5811 /* Returns:     int     - 0 = success, else failure                         */
5812 /* Parameters:  data(I) - pointer to ioctl data                             */
5813 /*              ptr(I)  - pointer to store real data in                     */
5814 /*              type(I) - type of structure being moved                     */
5815 /*              sz(I)   - size of data to copy                              */
5816 /*                                                                          */
5817 /* As per fr_inobj, except the size of the object to copy in is passed in   */
5818 /* but it must not be smaller than the size defined for the type and the    */
5819 /* type must allow for varied sized objects.  The extra requirement here is */
5820 /* that sz must match the size of the object being passed in - this is not  */
5821 /* not possible nor required in fr_inobj().                                 */
5822 /* ------------------------------------------------------------------------ */
fr_inobjsz(data,ptr,type,sz)5823 int fr_inobjsz(data, ptr, type, sz)
5824 void *data;
5825 void *ptr;
5826 int type, sz;
5827 {
5828 	ipfobj_t obj;
5829 	int error;
5830 
5831 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5832 		return EINVAL;
5833 	if (((fr_objbytes[type][0] & 1) == 0) || (sz < fr_objbytes[type][1]))
5834 		return EINVAL;
5835 
5836 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5837 	if (error != 0)
5838 		return EFAULT;
5839 
5840 	if (obj.ipfo_type != type)
5841 		return EINVAL;
5842 
5843 #ifndef	IPFILTER_COMPAT
5844 	if (obj.ipfo_size != sz)
5845 		return EINVAL;
5846 #else
5847 	if (obj.ipfo_rev != IPFILTER_VERSION)
5848 		/*XXX compatibility hook here */
5849 		/*EMPTY*/;
5850 	if (obj.ipfo_size != sz)
5851 		/* XXX compatibility hook here */
5852 		return EINVAL;
5853 #endif
5854 
5855 	error = COPYIN((caddr_t)obj.ipfo_ptr, (caddr_t)ptr, sz);
5856 	return error;
5857 }
5858 
5859 
5860 /* ------------------------------------------------------------------------ */
5861 /* Function:    fr_outobjsz                                                 */
5862 /* Returns:     int     - 0 = success, else failure                         */
5863 /* Parameters:  data(I) - pointer to ioctl data                             */
5864 /*              ptr(I)  - pointer to store real data in                     */
5865 /*              type(I) - type of structure being moved                     */
5866 /*              sz(I)   - size of data to copy                              */
5867 /*                                                                          */
5868 /* As per fr_outobj, except the size of the object to copy out is passed in */
5869 /* but it must not be smaller than the size defined for the type and the    */
5870 /* type must allow for varied sized objects.  The extra requirement here is */
5871 /* that sz must match the size of the object being passed in - this is not  */
5872 /* not possible nor required in fr_outobj().                                */
5873 /* ------------------------------------------------------------------------ */
fr_outobjsz(data,ptr,type,sz)5874 int fr_outobjsz(data, ptr, type, sz)
5875 void *data;
5876 void *ptr;
5877 int type, sz;
5878 {
5879 	ipfobj_t obj;
5880 	int error;
5881 
5882 	if ((type < 0) || (type > NUM_OBJ_TYPES-1) ||
5883 	    ((fr_objbytes[type][0] & 1) == 0) ||
5884 	    (sz < fr_objbytes[type][1]))
5885 		return EINVAL;
5886 
5887 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5888 	if (error != 0)
5889 		return EFAULT;
5890 
5891 	if (obj.ipfo_type != type)
5892 		return EINVAL;
5893 
5894 #ifndef	IPFILTER_COMPAT
5895 	if (obj.ipfo_size != sz)
5896 		return EINVAL;
5897 #else
5898 	if (obj.ipfo_rev != IPFILTER_VERSION)
5899 		/* XXX compatibility hook here */
5900 		/*EMPTY*/;
5901 	if (obj.ipfo_size != sz)
5902 		/* XXX compatibility hook here */
5903 		return EINVAL;
5904 #endif
5905 
5906 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, sz);
5907 	return error;
5908 }
5909 
5910 
5911 /* ------------------------------------------------------------------------ */
5912 /* Function:    fr_outobj                                                   */
5913 /* Returns:     int     - 0 = success, else failure                         */
5914 /* Parameters:  data(I) - pointer to ioctl data                             */
5915 /*              ptr(I)  - pointer to store real data in                     */
5916 /*              type(I) - type of structure being moved                     */
5917 /*                                                                          */
5918 /* Copy out the contents of what ptr is to where ipfobj points to.  In      */
5919 /* future, we add things to check for version numbers, sizes, etc, to make  */
5920 /* it backward  compatible at the ABI for user land.                        */
5921 /* ------------------------------------------------------------------------ */
fr_outobj(data,ptr,type)5922 int fr_outobj(data, ptr, type)
5923 void *data;
5924 void *ptr;
5925 int type;
5926 {
5927 	ipfobj_t obj;
5928 	int error;
5929 
5930 	if ((type < 0) || (type > NUM_OBJ_TYPES-1))
5931 		return EINVAL;
5932 
5933 	error = BCOPYIN((caddr_t)data, (caddr_t)&obj, sizeof(obj));
5934 	if (error != 0)
5935 		return EFAULT;
5936 
5937 	if (obj.ipfo_type != type)
5938 		return EINVAL;
5939 
5940 #ifndef	IPFILTER_COMPAT
5941 	if ((fr_objbytes[type][0] & 1) != 0) {
5942 		if (obj.ipfo_size < fr_objbytes[type][1])
5943 			return EINVAL;
5944 	} else if (obj.ipfo_size != fr_objbytes[type][1])
5945 		return EINVAL;
5946 #else
5947 	if (obj.ipfo_rev != IPFILTER_VERSION) {
5948 		error = fr_outcomptrans(&obj, ptr);
5949 		return error;
5950 	}
5951 
5952 	if ((fr_objbytes[type][0] & 1) != 0 &&
5953 	    obj.ipfo_size < fr_objbytes[type][1] ||
5954 	    obj.ipfo_size != fr_objbytes[type][1])
5955 		return EINVAL;
5956 #endif
5957 
5958 	error = COPYOUT((caddr_t)ptr, (caddr_t)obj.ipfo_ptr, obj.ipfo_size);
5959 	return error;
5960 }
5961 
5962 
5963 /* ------------------------------------------------------------------------ */
5964 /* Function:    fr_checkl4sum                                               */
5965 /* Returns:     int     - 0 = good, -1 = bad, 1 = cannot check              */
5966 /* Parameters:  fin(I) - pointer to packet information                      */
5967 /*                                                                          */
5968 /* If possible, calculate the layer 4 checksum for the packet.  If this is  */
5969 /* not possible, return without indicating a failure or success but in a    */
5970 /* way that is ditinguishable.                                              */
5971 /* ------------------------------------------------------------------------ */
fr_checkl4sum(fin)5972 int fr_checkl4sum(fin)
5973 fr_info_t *fin;
5974 {
5975 	u_short sum, hdrsum, *csump;
5976 	udphdr_t *udp;
5977 	int dosum;
5978 	ipf_stack_t *ifs = fin->fin_ifs;
5979 
5980 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
5981 	net_handle_t net_data_p;
5982 	if (fin->fin_v == 4)
5983 		net_data_p = ifs->ifs_ipf_ipv4;
5984 	else
5985 		net_data_p = ifs->ifs_ipf_ipv6;
5986 #endif
5987 
5988 	if ((fin->fin_flx & FI_NOCKSUM) != 0)
5989 		return 0;
5990 
5991 	/*
5992 	 * If the TCP packet isn't a fragment, isn't too short and otherwise
5993 	 * isn't already considered "bad", then validate the checksum.  If
5994 	 * this check fails then considered the packet to be "bad".
5995 	 */
5996 	if ((fin->fin_flx & (FI_FRAG|FI_SHORT|FI_BAD)) != 0)
5997 		return 1;
5998 
5999 	csump = NULL;
6000 	hdrsum = 0;
6001 	dosum = 0;
6002 	sum = 0;
6003 
6004 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
6005 	ASSERT(fin->fin_m != NULL);
6006 	if (NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m) ||
6007 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
6008 			hdrsum = 0;
6009 			sum = 0;
6010 	} else {
6011 #endif
6012 		switch (fin->fin_p)
6013 		{
6014 		case IPPROTO_TCP :
6015 			csump = &((tcphdr_t *)fin->fin_dp)->th_sum;
6016 			dosum = 1;
6017 			break;
6018 
6019 		case IPPROTO_UDP :
6020 			udp = fin->fin_dp;
6021 			if (udp->uh_sum != 0) {
6022 				csump = &udp->uh_sum;
6023 				dosum = 1;
6024 			}
6025 			break;
6026 
6027 		case IPPROTO_ICMP :
6028 			csump = &((struct icmp *)fin->fin_dp)->icmp_cksum;
6029 			dosum = 1;
6030 			break;
6031 
6032 		default :
6033 			return 1;
6034 			/*NOTREACHED*/
6035 		}
6036 
6037 		if (csump != NULL)
6038 			hdrsum = *csump;
6039 
6040 		if (dosum)
6041 			sum = fr_cksum(fin->fin_m, fin->fin_ip,
6042 				       fin->fin_p, fin->fin_dp);
6043 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
6044 	}
6045 #endif
6046 #if !defined(_KERNEL)
6047 	if (sum == hdrsum) {
6048 		FR_DEBUG(("checkl4sum: %hx == %hx\n", sum, hdrsum));
6049 	} else {
6050 		FR_DEBUG(("checkl4sum: %hx != %hx\n", sum, hdrsum));
6051 	}
6052 #endif
6053 	if (hdrsum == sum)
6054 		return 0;
6055 	return -1;
6056 }
6057 
6058 
6059 /* ------------------------------------------------------------------------ */
6060 /* Function:    fr_ifpfillv4addr                                            */
6061 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6062 /* Parameters:  atype(I)   - type of network address update to perform      */
6063 /*              sin(I)     - pointer to source of address information       */
6064 /*              mask(I)    - pointer to source of netmask information       */
6065 /*              inp(I)     - pointer to destination address store           */
6066 /*              inpmask(I) - pointer to destination netmask store           */
6067 /*                                                                          */
6068 /* Given a type of network address update (atype) to perform, copy          */
6069 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6070 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6071 /* which case the operation fails.  For all values of atype other than      */
6072 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6073 /* value.                                                                   */
6074 /* ------------------------------------------------------------------------ */
fr_ifpfillv4addr(atype,sin,mask,inp,inpmask)6075 int fr_ifpfillv4addr(atype, sin, mask, inp, inpmask)
6076 int atype;
6077 struct sockaddr_in *sin, *mask;
6078 struct in_addr *inp, *inpmask;
6079 {
6080 	if (inpmask != NULL && atype != FRI_NETMASKED)
6081 		inpmask->s_addr = 0xffffffff;
6082 
6083 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6084 		if (atype == FRI_NETMASKED) {
6085 			if (inpmask == NULL)
6086 				return -1;
6087 			inpmask->s_addr = mask->sin_addr.s_addr;
6088 		}
6089 		inp->s_addr = sin->sin_addr.s_addr & mask->sin_addr.s_addr;
6090 	} else {
6091 		inp->s_addr = sin->sin_addr.s_addr;
6092 	}
6093 	return 0;
6094 }
6095 
6096 
6097 #ifdef	USE_INET6
6098 /* ------------------------------------------------------------------------ */
6099 /* Function:    fr_ifpfillv6addr                                            */
6100 /* Returns:     int     - 0 = address update, -1 = address not updated      */
6101 /* Parameters:  atype(I)   - type of network address update to perform      */
6102 /*              sin(I)     - pointer to source of address information       */
6103 /*              mask(I)    - pointer to source of netmask information       */
6104 /*              inp(I)     - pointer to destination address store           */
6105 /*              inpmask(I) - pointer to destination netmask store           */
6106 /*                                                                          */
6107 /* Given a type of network address update (atype) to perform, copy          */
6108 /* information from sin/mask into inp/inpmask.  If ipnmask is NULL then no  */
6109 /* netmask update is performed unless FRI_NETMASKED is passed as atype, in  */
6110 /* which case the operation fails.  For all values of atype other than      */
6111 /* FRI_NETMASKED, if inpmask is non-NULL then the mask is set to an all 1s  */
6112 /* value.                                                                   */
6113 /* ------------------------------------------------------------------------ */
fr_ifpfillv6addr(atype,sin,mask,inp,inpmask)6114 int fr_ifpfillv6addr(atype, sin, mask, inp, inpmask)
6115 int atype;
6116 struct sockaddr_in6 *sin, *mask;
6117 struct in_addr *inp, *inpmask;
6118 {
6119 	i6addr_t *src, *dst, *and, *dmask;
6120 
6121 	src = (i6addr_t *)&sin->sin6_addr;
6122 	and = (i6addr_t *)&mask->sin6_addr;
6123 	dst = (i6addr_t *)inp;
6124 	dmask = (i6addr_t *)inpmask;
6125 
6126 	if (inpmask != NULL && atype != FRI_NETMASKED) {
6127 		dmask->i6[0] = 0xffffffff;
6128 		dmask->i6[1] = 0xffffffff;
6129 		dmask->i6[2] = 0xffffffff;
6130 		dmask->i6[3] = 0xffffffff;
6131 	}
6132 
6133 	if (atype == FRI_NETWORK || atype == FRI_NETMASKED) {
6134 		if (atype == FRI_NETMASKED) {
6135 			if (inpmask == NULL)
6136 				return -1;
6137 			dmask->i6[0] = and->i6[0];
6138 			dmask->i6[1] = and->i6[1];
6139 			dmask->i6[2] = and->i6[2];
6140 			dmask->i6[3] = and->i6[3];
6141 		}
6142 
6143 		dst->i6[0] = src->i6[0] & and->i6[0];
6144 		dst->i6[1] = src->i6[1] & and->i6[1];
6145 		dst->i6[2] = src->i6[2] & and->i6[2];
6146 		dst->i6[3] = src->i6[3] & and->i6[3];
6147 	} else {
6148 		dst->i6[0] = src->i6[0];
6149 		dst->i6[1] = src->i6[1];
6150 		dst->i6[2] = src->i6[2];
6151 		dst->i6[3] = src->i6[3];
6152 	}
6153 	return 0;
6154 }
6155 #endif
6156 
6157 
6158 /* ------------------------------------------------------------------------ */
6159 /* Function:    fr_matchtag                                                 */
6160 /* Returns:     0 == mismatch, 1 == match.                                  */
6161 /* Parameters:  tag1(I) - pointer to first tag to compare                   */
6162 /*              tag2(I) - pointer to second tag to compare                  */
6163 /*                                                                          */
6164 /* Returns true (non-zero) or false(0) if the two tag structures can be     */
6165 /* considered to be a match or not match, respectively.  The tag is 16      */
6166 /* bytes long (16 characters) but that is overlayed with 4 32bit ints so    */
6167 /* compare the ints instead, for speed. tag1 is the master of the           */
6168 /* comparison.  This function should only be called with both tag1 and tag2 */
6169 /* as non-NULL pointers.                                                    */
6170 /* ------------------------------------------------------------------------ */
fr_matchtag(tag1,tag2)6171 int fr_matchtag(tag1, tag2)
6172 ipftag_t *tag1, *tag2;
6173 {
6174 	if (tag1 == tag2)
6175 		return 1;
6176 
6177 	if ((tag1->ipt_num[0] == 0) && (tag2->ipt_num[0] == 0))
6178 		return 1;
6179 
6180 	if ((tag1->ipt_num[0] == tag2->ipt_num[0]) &&
6181 	    (tag1->ipt_num[1] == tag2->ipt_num[1]) &&
6182 	    (tag1->ipt_num[2] == tag2->ipt_num[2]) &&
6183 	    (tag1->ipt_num[3] == tag2->ipt_num[3]))
6184 		return 1;
6185 	return 0;
6186 }
6187 
6188 
6189 /* ------------------------------------------------------------------------ */
6190 /* Function:    fr_coalesce                                                 */
6191 /* Returns:     1 == success, -1 == failure, 0 == no change                 */
6192 /* Parameters:  fin(I) - pointer to packet information                      */
6193 /*                                                                          */
6194 /* Attempt to get all of the packet data into a single, contiguous buffer.  */
6195 /* If this call returns a failure then the buffers have also been freed.    */
6196 /* ------------------------------------------------------------------------ */
fr_coalesce(fin)6197 int fr_coalesce(fin)
6198 fr_info_t *fin;
6199 {
6200 	ipf_stack_t *ifs = fin->fin_ifs;
6201 	if ((fin->fin_flx & FI_COALESCE) != 0)
6202 		return 1;
6203 
6204 	/*
6205 	 * If the mbuf pointers indicate that there is no mbuf to work with,
6206 	 * return but do not indicate success or failure.
6207 	 */
6208 	if (fin->fin_m == NULL || fin->fin_mp == NULL)
6209 		return 0;
6210 
6211 #if defined(_KERNEL)
6212 	if (fr_pullup(fin->fin_m, fin, fin->fin_plen) == NULL) {
6213 		IPF_BUMP(ifs->ifs_fr_badcoalesces[fin->fin_out]);
6214 # ifdef MENTAT
6215 		FREE_MB_T(*fin->fin_mp);
6216 # endif
6217 		*fin->fin_mp = NULL;
6218 		fin->fin_m = NULL;
6219 		return -1;
6220 	}
6221 #else
6222 	fin = fin;	/* LINT */
6223 #endif
6224 	return 1;
6225 }
6226 
6227 
6228 /*
6229  * The following table lists all of the tunable variables that can be
6230  * accessed via SIOCIPFGET/SIOCIPFSET/SIOCIPFGETNEXT.  The format of each row
6231  * in the table below is as follows:
6232  *
6233  * pointer to value, name of value, minimum, maximum, size of the value's
6234  *     container, value attribute flags
6235  *
6236  * For convienience, IPFT_RDONLY means the value is read-only, IPFT_WRDISABLED
6237  * means the value can only be written to when IPFilter is loaded but disabled.
6238  * The obvious implication is if neither of these are set then the value can be
6239  * changed at any time without harm.
6240  */
6241 ipftuneable_t lcl_ipf_tuneables[] = {
6242 	/* filtering */
6243 	{ { NULL },	"fr_flags",		0,	0xffffffff,
6244 			0,		0 },
6245 	{ { NULL },	"fr_active",		0,	0,
6246 			0,		IPFT_RDONLY },
6247 	{ { NULL },	"fr_control_forwarding",	0, 1,
6248 			0,		0 },
6249 	{ { NULL },	"fr_update_ipid",	0,	1,
6250 			0,		0 },
6251 	{ { NULL },	"fr_chksrc",		0,	1,
6252 			0,		0 },
6253 	{ { NULL },	"fr_minttl",		0,	1,
6254 			0,		0 },
6255 	{ { NULL }, 	"fr_icmpminfragmtu",	0,	1,
6256 			0,		0 },
6257 	{ { NULL },		"fr_pass",		0,	0xffffffff,
6258 			0,		0 },
6259 #if SOLARIS2 >= 10
6260 	{ { NULL },	"ipf_loopback",		0,	1,
6261 			0,		IPFT_WRDISABLED },
6262 #endif
6263 	/* state */
6264 	{ { NULL }, "fr_tcpidletimeout",	1,	0x7fffffff,
6265 			0,	IPFT_WRDISABLED },
6266 	{ { NULL },	"fr_tcpclosewait",	1,	0x7fffffff,
6267 			0,	IPFT_WRDISABLED },
6268 	{ { NULL },	"fr_tcplastack",	1,	0x7fffffff,
6269 			0,		IPFT_WRDISABLED },
6270 	{ { NULL },	"fr_tcptimeout",	1,	0x7fffffff,
6271 			0,		IPFT_WRDISABLED },
6272 	{ { NULL },	"fr_tcpclosed",		1,	0x7fffffff,
6273 			0,		IPFT_WRDISABLED },
6274 	{ { NULL },	"fr_tcphalfclosed",	1,	0x7fffffff,
6275 			0,	IPFT_WRDISABLED },
6276 	{ { NULL },	"fr_udptimeout",	1,	0x7fffffff,
6277 			0,		IPFT_WRDISABLED },
6278 	{ { NULL }, "fr_udpacktimeout",	1,	0x7fffffff,
6279 			0,	IPFT_WRDISABLED },
6280 	{ { NULL },	"fr_icmptimeout",	1,	0x7fffffff,
6281 			0,		IPFT_WRDISABLED },
6282 	{ { NULL }, "fr_icmpacktimeout",	1,	0x7fffffff,
6283 			0,	IPFT_WRDISABLED },
6284 	{ { NULL }, "fr_iptimeout",		1,	0x7fffffff,
6285 			0,		IPFT_WRDISABLED },
6286 	{ { NULL },	"fr_statemax",		1,	0x7fffffff,
6287 			0,		0 },
6288 	{ { NULL },	"fr_statesize",		1,	0x7fffffff,
6289 			0,		IPFT_WRDISABLED },
6290 	{ { NULL },	"fr_state_lock",	0,	1,
6291 			0,		IPFT_RDONLY },
6292 	{ { NULL }, "fr_state_maxbucket", 1,	0x7fffffff,
6293 			0,	IPFT_WRDISABLED },
6294 	{ { NULL }, "fr_state_maxbucket_reset",	0, 1,
6295 			0, IPFT_WRDISABLED },
6296 	{ { NULL },	"ipstate_logging",	0,	1,
6297 			0,	0 },
6298 	{ { NULL },	"state_flush_level_hi",	1,	100,
6299 			0,		0 },
6300 	{ { NULL },	"state_flush_level_lo",	1,	100,
6301 			0,		0 },
6302 	/* nat */
6303 	{ { NULL },		"fr_nat_lock",		0,	1,
6304 			0,		IPFT_RDONLY },
6305 	{ { NULL },	"ipf_nattable_sz",	1,	0x7fffffff,
6306 			0,	IPFT_WRDISABLED },
6307 	{ { NULL }, "ipf_nattable_max",	1,	0x7fffffff,
6308 			0,	0 },
6309 	{ { NULL },	"ipf_natrules_sz",	1,	0x7fffffff,
6310 			0,	IPFT_WRDISABLED },
6311 	{ { NULL },	"ipf_rdrrules_sz",	1,	0x7fffffff,
6312 			0,	IPFT_WRDISABLED },
6313 	{ { NULL },	"ipf_hostmap_sz",	1,	0x7fffffff,
6314 			0,		IPFT_WRDISABLED },
6315 	{ { NULL }, "fr_nat_maxbucket",	1,	0x7fffffff,
6316 			0,	IPFT_WRDISABLED },
6317 	{ { NULL },	"fr_nat_maxbucket_reset",	0, 1,
6318 			0,	IPFT_WRDISABLED },
6319 	{ { NULL },		"nat_logging",		0,	1,
6320 			0,		0 },
6321 	{ { NULL },	"fr_defnatage",		1,	0x7fffffff,
6322 			0,		IPFT_WRDISABLED },
6323 	{ { NULL },	"fr_defnatipage",	1,	0x7fffffff,
6324 			0,		IPFT_WRDISABLED },
6325 	{ { NULL }, "fr_defnaticmpage",	1,	0x7fffffff,
6326 			0,	IPFT_WRDISABLED },
6327 	{ { NULL },	"nat_flush_level_hi",	1,	100,
6328 			0,		0 },
6329 	{ { NULL },	"nat_flush_level_lo",	1,	100,
6330 			0,		0 },
6331 	/* frag */
6332 	{ { NULL },	"ipfr_size",		1,	0x7fffffff,
6333 			0,		IPFT_WRDISABLED },
6334 	{ { NULL },	"fr_ipfrttl",		1,	0x7fffffff,
6335 			0,		IPFT_WRDISABLED },
6336 #ifdef IPFILTER_LOG
6337 	/* log */
6338 	{ { NULL },	"ipl_suppress",		0,	1,
6339 			0,		0 },
6340 	{ { NULL },	"ipl_buffer_sz",	0,	0,
6341 			0,		IPFT_RDONLY },
6342 	{ { NULL },	"ipl_logmax",		0,	0x7fffffff,
6343 			0,		IPFT_WRDISABLED },
6344 	{ { NULL },	"ipl_logall",		0,	1,
6345 			0,		0 },
6346 	{ { NULL },	"ipl_logsize",		0,	0x80000,
6347 			0,		0 },
6348 #endif
6349 	{ { NULL },		NULL,			0,	0 }
6350 };
6351 
6352 static ipftuneable_t *
tune_lookup(ipf_stack_t * ifs,char * name)6353 tune_lookup(ipf_stack_t *ifs, char *name)
6354 {
6355     int i;
6356 
6357     for (i = 0; ifs->ifs_ipf_tuneables[i].ipft_name != NULL; i++) {
6358 	if (strcmp(ifs->ifs_ipf_tuneables[i].ipft_name, name) == 0)
6359 	    return (&ifs->ifs_ipf_tuneables[i]);
6360     }
6361     return (NULL);
6362 }
6363 
6364 #ifdef _KERNEL
6365 extern dev_info_t *ipf_dev_info;
6366 extern int ipf_property_update __P((dev_info_t *, ipf_stack_t *));
6367 #endif
6368 
6369 /* -------------------------------------------------------------------- */
6370 /* Function:	ipftuneable_setdefs()					*/
6371 /* Returns:		void						*/
6372 /* Parameters:	ifs - pointer to newly allocated IPF instance		*/
6373 /*				assigned to	IP instance		*/
6374 /*									*/
6375 /* Function initializes IPF instance variables. Function is invoked	*/
6376 /* from	ipftuneable_alloc(). ipftuneable_alloc() is called only one	*/
6377 /* time during IP instance lifetime - at the time of IP instance	*/
6378 /* creation. Anytime IP	instance is being created new private IPF	*/
6379 /* instance is allocated and assigned to it. The moment of IP 		*/
6380 /* instance creation is the right time to initialize those IPF 		*/
6381 /* variables.								*/
6382 /*									*/
6383 /* -------------------------------------------------------------------- */
ipftuneable_setdefs(ipf_stack_t * ifs)6384 static void ipftuneable_setdefs(ipf_stack_t *ifs)
6385 {
6386 	ifs->ifs_ipfr_size = IPFT_SIZE;
6387 	ifs->ifs_fr_ipfrttl = 120;	/* 60 seconds */
6388 
6389 	/* it comes from fr_authinit() in IPF auth */
6390 	ifs->ifs_fr_authsize = FR_NUMAUTH;
6391 	ifs->ifs_fr_defaultauthage = 600;
6392 
6393 	/* it comes from fr_stateinit() in IPF state */
6394 	ifs->ifs_fr_tcpidletimeout = IPF_TTLVAL(3600 * 24 * 5);	/* five days */
6395 	ifs->ifs_fr_tcpclosewait = IPF_TTLVAL(TCP_MSL);
6396 	ifs->ifs_fr_tcplastack = IPF_TTLVAL(TCP_MSL);
6397 	ifs->ifs_fr_tcptimeout = IPF_TTLVAL(TCP_MSL);
6398 	ifs->ifs_fr_tcpclosed = IPF_TTLVAL(60);
6399 	ifs->ifs_fr_tcphalfclosed = IPF_TTLVAL(2 * 3600);	/* 2 hours */
6400 	ifs->ifs_fr_udptimeout = IPF_TTLVAL(120);
6401 	ifs->ifs_fr_udpacktimeout = IPF_TTLVAL(12);
6402 	ifs->ifs_fr_icmptimeout = IPF_TTLVAL(60);
6403 	ifs->ifs_fr_icmpacktimeout = IPF_TTLVAL(6);
6404 	ifs->ifs_fr_iptimeout = IPF_TTLVAL(60);
6405 	ifs->ifs_fr_statemax = IPSTATE_MAX;
6406 	ifs->ifs_fr_statesize = IPSTATE_SIZE;
6407 	ifs->ifs_fr_state_maxbucket_reset = 1;
6408 	ifs->ifs_state_flush_level_hi = ST_FLUSH_HI;
6409 	ifs->ifs_state_flush_level_lo = ST_FLUSH_LO;
6410 
6411 	/* it comes from fr_natinit() in ipnat */
6412 	ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
6413 	ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
6414 	ifs->ifs_ipf_natrules_sz = NAT_SIZE;
6415 	ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
6416 	ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
6417 	ifs->ifs_fr_nat_maxbucket_reset = 1;
6418 	ifs->ifs_fr_defnatage = DEF_NAT_AGE;
6419 	ifs->ifs_fr_defnatipage = 120;		/* 60 seconds */
6420 	ifs->ifs_fr_defnaticmpage = 6;		/* 3 seconds */
6421 	ifs->ifs_nat_flush_level_hi = NAT_FLUSH_HI;
6422 	ifs->ifs_nat_flush_level_lo = NAT_FLUSH_LO;
6423 
6424 #ifdef IPFILTER_LOG
6425 	/* it comes from fr_loginit() in IPF log */
6426 	ifs->ifs_ipl_suppress = 1;
6427 	ifs->ifs_ipl_logmax = IPL_LOGMAX;
6428 	ifs->ifs_ipl_logsize = IPFILTER_LOGSIZE;
6429 
6430 	/* from fr_natinit() */
6431 	ifs->ifs_nat_logging = 1;
6432 
6433 	/* from fr_stateinit() */
6434 	ifs->ifs_ipstate_logging = 1;
6435 #else
6436 	/* from fr_natinit() */
6437 	ifs->ifs_nat_logging = 0;
6438 
6439 	/* from fr_stateinit() */
6440 	ifs->ifs_ipstate_logging = 0;
6441 #endif
6442 	ifs->ifs_ipf_loopback = 0;
6443 
6444 }
6445 /*
6446  * Allocate a per-stack tuneable and copy in the names. Then
6447  * set it to point to each of the per-stack tunables.
6448  */
6449 void
ipftuneable_alloc(ipf_stack_t * ifs)6450 ipftuneable_alloc(ipf_stack_t *ifs)
6451 {
6452     ipftuneable_t *item;
6453 
6454     KMALLOCS(ifs->ifs_ipf_tuneables, ipftuneable_t *,
6455 	sizeof (lcl_ipf_tuneables));
6456     bcopy(lcl_ipf_tuneables, ifs->ifs_ipf_tuneables,
6457 	sizeof (lcl_ipf_tuneables));
6458 
6459 #define TUNE_SET(_ifs, _name, _field)			\
6460     item = tune_lookup((_ifs), (_name));		\
6461     if (item != NULL) {					\
6462 	item->ipft_una.ipftp_int = (unsigned int *)&((_ifs)->_field);	\
6463 	item->ipft_sz = sizeof ((_ifs)->_field);	\
6464     }
6465 
6466     TUNE_SET(ifs, "fr_flags", ifs_fr_flags);
6467     TUNE_SET(ifs, "fr_active", ifs_fr_active);
6468     TUNE_SET(ifs, "fr_control_forwarding", ifs_fr_control_forwarding);
6469     TUNE_SET(ifs, "fr_update_ipid", ifs_fr_update_ipid);
6470     TUNE_SET(ifs, "fr_chksrc", ifs_fr_chksrc);
6471     TUNE_SET(ifs, "fr_minttl", ifs_fr_minttl);
6472     TUNE_SET(ifs, "fr_icmpminfragmtu", ifs_fr_icmpminfragmtu);
6473     TUNE_SET(ifs, "fr_pass", ifs_fr_pass);
6474     TUNE_SET(ifs, "fr_tcpidletimeout", ifs_fr_tcpidletimeout);
6475     TUNE_SET(ifs, "fr_tcpclosewait", ifs_fr_tcpclosewait);
6476     TUNE_SET(ifs, "fr_tcplastack", ifs_fr_tcplastack);
6477     TUNE_SET(ifs, "fr_tcptimeout", ifs_fr_tcptimeout);
6478     TUNE_SET(ifs, "fr_tcpclosed", ifs_fr_tcpclosed);
6479     TUNE_SET(ifs, "fr_tcphalfclosed", ifs_fr_tcphalfclosed);
6480     TUNE_SET(ifs, "fr_udptimeout", ifs_fr_udptimeout);
6481     TUNE_SET(ifs, "fr_udpacktimeout", ifs_fr_udpacktimeout);
6482     TUNE_SET(ifs, "fr_icmptimeout", ifs_fr_icmptimeout);
6483     TUNE_SET(ifs, "fr_icmpacktimeout", ifs_fr_icmpacktimeout);
6484     TUNE_SET(ifs, "fr_iptimeout", ifs_fr_iptimeout);
6485     TUNE_SET(ifs, "fr_statemax", ifs_fr_statemax);
6486     TUNE_SET(ifs, "fr_statesize", ifs_fr_statesize);
6487     TUNE_SET(ifs, "fr_state_lock", ifs_fr_state_lock);
6488     TUNE_SET(ifs, "fr_state_maxbucket", ifs_fr_state_maxbucket);
6489     TUNE_SET(ifs, "fr_state_maxbucket_reset", ifs_fr_state_maxbucket_reset);
6490     TUNE_SET(ifs, "ipstate_logging", ifs_ipstate_logging);
6491     TUNE_SET(ifs, "fr_nat_lock", ifs_fr_nat_lock);
6492     TUNE_SET(ifs, "ipf_nattable_sz", ifs_ipf_nattable_sz);
6493     TUNE_SET(ifs, "ipf_nattable_max", ifs_ipf_nattable_max);
6494     TUNE_SET(ifs, "ipf_natrules_sz", ifs_ipf_natrules_sz);
6495     TUNE_SET(ifs, "ipf_rdrrules_sz", ifs_ipf_rdrrules_sz);
6496     TUNE_SET(ifs, "ipf_hostmap_sz", ifs_ipf_hostmap_sz);
6497     TUNE_SET(ifs, "fr_nat_maxbucket", ifs_fr_nat_maxbucket);
6498     TUNE_SET(ifs, "fr_nat_maxbucket_reset", ifs_fr_nat_maxbucket_reset);
6499     TUNE_SET(ifs, "nat_logging", ifs_nat_logging);
6500     TUNE_SET(ifs, "fr_defnatage", ifs_fr_defnatage);
6501     TUNE_SET(ifs, "fr_defnatipage", ifs_fr_defnatipage);
6502     TUNE_SET(ifs, "fr_defnaticmpage", ifs_fr_defnaticmpage);
6503     TUNE_SET(ifs, "nat_flush_level_hi", ifs_nat_flush_level_hi);
6504     TUNE_SET(ifs, "nat_flush_level_lo", ifs_nat_flush_level_lo);
6505     TUNE_SET(ifs, "state_flush_level_hi", ifs_state_flush_level_hi);
6506     TUNE_SET(ifs, "state_flush_level_lo", ifs_state_flush_level_lo);
6507     TUNE_SET(ifs, "ipfr_size", ifs_ipfr_size);
6508     TUNE_SET(ifs, "fr_ipfrttl", ifs_fr_ipfrttl);
6509     TUNE_SET(ifs, "ipf_loopback", ifs_ipf_loopback);
6510 #ifdef IPFILTER_LOG
6511     TUNE_SET(ifs, "ipl_suppress", ifs_ipl_suppress);
6512     TUNE_SET(ifs, "ipl_buffer_sz", ifs_ipl_buffer_sz);
6513     TUNE_SET(ifs, "ipl_logmax", ifs_ipl_logmax);
6514     TUNE_SET(ifs, "ipl_logall", ifs_ipl_logall);
6515     TUNE_SET(ifs, "ipl_logsize", ifs_ipl_logsize);
6516 #endif
6517 #undef TUNE_SET
6518 
6519 	ipftuneable_setdefs(ifs);
6520 
6521 #ifdef _KERNEL
6522     (void) ipf_property_update(ipf_dev_info, ifs);
6523 #endif
6524 }
6525 
6526 void
ipftuneable_free(ipf_stack_t * ifs)6527 ipftuneable_free(ipf_stack_t *ifs)
6528 {
6529 	KFREES(ifs->ifs_ipf_tuneables, sizeof (lcl_ipf_tuneables));
6530 	ifs->ifs_ipf_tuneables = NULL;
6531 }
6532 
6533 /* ------------------------------------------------------------------------ */
6534 /* Function:    fr_findtunebycookie                                         */
6535 /* Returns:     NULL = search failed, else pointer to tune struct           */
6536 /* Parameters:  cookie(I) - cookie value to search for amongst tuneables    */
6537 /*              next(O)   - pointer to place to store the cookie for the    */
6538 /*                          "next" tuneable, if it is desired.              */
6539 /*                                                                          */
6540 /* This function is used to walk through all of the existing tunables with  */
6541 /* successive calls.  It searches the known tunables for the one which has  */
6542 /* a matching value for "cookie" - ie its address.  When returning a match, */
6543 /* the next one to be found may be returned inside next.                    */
6544 /* ------------------------------------------------------------------------ */
fr_findtunebycookie(cookie,next,ifs)6545 static ipftuneable_t *fr_findtunebycookie(cookie, next, ifs)
6546 void *cookie, **next;
6547 ipf_stack_t * ifs;
6548 {
6549 	ipftuneable_t *ta, **tap;
6550 
6551 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6552 		if (ta == cookie) {
6553 			if (next != NULL) {
6554 				/*
6555 				 * If the next entry in the array has a name
6556 				 * present, then return a pointer to it for
6557 				 * where to go next, else return a pointer to
6558 				 * the dynaminc list as a key to search there
6559 				 * next.  This facilitates a weak linking of
6560 				 * the two "lists" together.
6561 				 */
6562 				if ((ta + 1)->ipft_name != NULL)
6563 					*next = ta + 1;
6564 				else
6565 					*next = &ifs->ifs_ipf_tunelist;
6566 			}
6567 			return ta;
6568 		}
6569 
6570 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6571 		if (tap == cookie) {
6572 			if (next != NULL)
6573 				*next = &ta->ipft_next;
6574 			return ta;
6575 		}
6576 
6577 	if (next != NULL)
6578 		*next = NULL;
6579 	return NULL;
6580 }
6581 
6582 
6583 /* ------------------------------------------------------------------------ */
6584 /* Function:    fr_findtunebyname                                           */
6585 /* Returns:     NULL = search failed, else pointer to tune struct           */
6586 /* Parameters:  name(I) - name of the tuneable entry to find.               */
6587 /*                                                                          */
6588 /* Search the static array of tuneables and the list of dynamic tuneables   */
6589 /* for an entry with a matching name.  If we can find one, return a pointer */
6590 /* to the matching structure.                                               */
6591 /* ------------------------------------------------------------------------ */
fr_findtunebyname(name,ifs)6592 static ipftuneable_t *fr_findtunebyname(name, ifs)
6593 const char *name;
6594 ipf_stack_t *ifs;
6595 {
6596 	ipftuneable_t *ta;
6597 
6598 	for (ta = ifs->ifs_ipf_tuneables; ta->ipft_name != NULL; ta++)
6599 		if (!strcmp(ta->ipft_name, name)) {
6600 			return ta;
6601 		}
6602 
6603 	for (ta = ifs->ifs_ipf_tunelist; ta != NULL; ta = ta->ipft_next)
6604 		if (!strcmp(ta->ipft_name, name)) {
6605 			return ta;
6606 		}
6607 
6608 	return NULL;
6609 }
6610 
6611 
6612 /* ------------------------------------------------------------------------ */
6613 /* Function:    fr_addipftune                                               */
6614 /* Returns:     int - 0 == success, else failure                            */
6615 /* Parameters:  newtune - pointer to new tune struct to add to tuneables    */
6616 /*                                                                          */
6617 /* Appends the tune structure pointer to by "newtune" to the end of the     */
6618 /* current list of "dynamic" tuneable parameters.  Once added, the owner    */
6619 /* of the object is not expected to ever change "ipft_next".                */
6620 /* ------------------------------------------------------------------------ */
fr_addipftune(newtune,ifs)6621 int fr_addipftune(newtune, ifs)
6622 ipftuneable_t *newtune;
6623 ipf_stack_t *ifs;
6624 {
6625 	ipftuneable_t *ta, **tap;
6626 
6627 	ta = fr_findtunebyname(newtune->ipft_name, ifs);
6628 	if (ta != NULL)
6629 		return EEXIST;
6630 
6631 	for (tap = &ifs->ifs_ipf_tunelist; *tap != NULL; tap = &(*tap)->ipft_next)
6632 		;
6633 
6634 	newtune->ipft_next = NULL;
6635 	*tap = newtune;
6636 	return 0;
6637 }
6638 
6639 
6640 /* ------------------------------------------------------------------------ */
6641 /* Function:    fr_delipftune                                               */
6642 /* Returns:     int - 0 == success, else failure                            */
6643 /* Parameters:  oldtune - pointer to tune struct to remove from the list of */
6644 /*                        current dynamic tuneables                         */
6645 /*                                                                          */
6646 /* Search for the tune structure, by pointer, in the list of those that are */
6647 /* dynamically added at run time.  If found, adjust the list so that this   */
6648 /* structure is no longer part of it.                                       */
6649 /* ------------------------------------------------------------------------ */
fr_delipftune(oldtune,ifs)6650 int fr_delipftune(oldtune, ifs)
6651 ipftuneable_t *oldtune;
6652 ipf_stack_t *ifs;
6653 {
6654 	ipftuneable_t *ta, **tap;
6655 
6656 	for (tap = &ifs->ifs_ipf_tunelist; (ta = *tap) != NULL; tap = &ta->ipft_next)
6657 		if (ta == oldtune) {
6658 			*tap = oldtune->ipft_next;
6659 			oldtune->ipft_next = NULL;
6660 			return 0;
6661 		}
6662 
6663 	return ESRCH;
6664 }
6665 
6666 
6667 /* ------------------------------------------------------------------------ */
6668 /* Function:    fr_ipftune                                                  */
6669 /* Returns:     int - 0 == success, else failure                            */
6670 /* Parameters:  cmd(I)  - ioctl command number                              */
6671 /*              data(I) - pointer to ioctl data structure                   */
6672 /*                                                                          */
6673 /* Implement handling of SIOCIPFGETNEXT, SIOCIPFGET and SIOCIPFSET.  These  */
6674 /* three ioctls provide the means to access and control global variables    */
6675 /* within IPFilter, allowing (for example) timeouts and table sizes to be   */
6676 /* changed without rebooting, reloading or recompiling.  The initialisation */
6677 /* and 'destruction' routines of the various components of ipfilter are all */
6678 /* each responsible for handling their own values being too big.            */
6679 /* ------------------------------------------------------------------------ */
fr_ipftune(cmd,data,ifs)6680 int fr_ipftune(cmd, data, ifs)
6681 ioctlcmd_t cmd;
6682 void *data;
6683 ipf_stack_t *ifs;
6684 {
6685 	ipftuneable_t *ta;
6686 	ipftune_t tu;
6687 	void *cookie;
6688 	int error;
6689 
6690 	error = fr_inobj(data, &tu, IPFOBJ_TUNEABLE);
6691 	if (error != 0)
6692 		return error;
6693 
6694 	tu.ipft_name[sizeof(tu.ipft_name) - 1] = '\0';
6695 	cookie = tu.ipft_cookie;
6696 	ta = NULL;
6697 
6698 	switch (cmd)
6699 	{
6700 	case SIOCIPFGETNEXT :
6701 		/*
6702 		 * If cookie is non-NULL, assume it to be a pointer to the last
6703 		 * entry we looked at, so find it (if possible) and return a
6704 		 * pointer to the next one after it.  The last entry in the
6705 		 * the table is a NULL entry, so when we get to it, set cookie
6706 		 * to NULL and return that, indicating end of list, erstwhile
6707 		 * if we come in with cookie set to NULL, we are starting anew
6708 		 * at the front of the list.
6709 		 */
6710 		if (cookie != NULL) {
6711 			ta = fr_findtunebycookie(cookie, &tu.ipft_cookie, ifs);
6712 		} else {
6713 			ta = ifs->ifs_ipf_tuneables;
6714 			tu.ipft_cookie = ta + 1;
6715 		}
6716 		if (ta != NULL) {
6717 			/*
6718 			 * Entry found, but does the data pointed to by that
6719 			 * row fit in what we can return?
6720 			 */
6721 			if (ta->ipft_sz > sizeof(tu.ipft_un))
6722 				return EINVAL;
6723 
6724 			tu.ipft_vlong = 0;
6725 			if (ta->ipft_sz == sizeof(u_long))
6726 				tu.ipft_vlong = *ta->ipft_plong;
6727 			else if (ta->ipft_sz == sizeof(u_int))
6728 				tu.ipft_vint = *ta->ipft_pint;
6729 			else if (ta->ipft_sz == sizeof(u_short))
6730 				tu.ipft_vshort = *ta->ipft_pshort;
6731 			else if (ta->ipft_sz == sizeof(u_char))
6732 				tu.ipft_vchar = *ta->ipft_pchar;
6733 
6734 			tu.ipft_sz = ta->ipft_sz;
6735 			tu.ipft_min = ta->ipft_min;
6736 			tu.ipft_max = ta->ipft_max;
6737 			tu.ipft_flags = ta->ipft_flags;
6738 			bcopy(ta->ipft_name, tu.ipft_name,
6739 			      MIN(sizeof(tu.ipft_name),
6740 				  strlen(ta->ipft_name) + 1));
6741 		}
6742 		error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6743 		break;
6744 
6745 	case SIOCIPFGET :
6746 	case SIOCIPFSET :
6747 		/*
6748 		 * Search by name or by cookie value for a particular entry
6749 		 * in the tuning paramter table.
6750 		 */
6751 		error = ESRCH;
6752 		if (cookie != NULL) {
6753 			ta = fr_findtunebycookie(cookie, NULL, ifs);
6754 			if (ta != NULL)
6755 				error = 0;
6756 		} else if (tu.ipft_name[0] != '\0') {
6757 			ta = fr_findtunebyname(tu.ipft_name, ifs);
6758 			if (ta != NULL)
6759 				error = 0;
6760 		}
6761 		if (error != 0)
6762 			break;
6763 
6764 		if (cmd == (ioctlcmd_t)SIOCIPFGET) {
6765 			/*
6766 			 * Fetch the tuning parameters for a particular value
6767 			 */
6768 			tu.ipft_vlong = 0;
6769 			if (ta->ipft_sz == sizeof(u_long))
6770 				tu.ipft_vlong = *ta->ipft_plong;
6771 			else if (ta->ipft_sz == sizeof(u_int))
6772 				tu.ipft_vint = *ta->ipft_pint;
6773 			else if (ta->ipft_sz == sizeof(u_short))
6774 				tu.ipft_vshort = *ta->ipft_pshort;
6775 			else if (ta->ipft_sz == sizeof(u_char))
6776 				tu.ipft_vchar = *ta->ipft_pchar;
6777 			tu.ipft_cookie = ta;
6778 			tu.ipft_sz = ta->ipft_sz;
6779 			tu.ipft_min = ta->ipft_min;
6780 			tu.ipft_max = ta->ipft_max;
6781 			tu.ipft_flags = ta->ipft_flags;
6782 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6783 
6784 		} else if (cmd == (ioctlcmd_t)SIOCIPFSET) {
6785 			/*
6786 			 * Set an internal parameter.  The hard part here is
6787 			 * getting the new value safely and correctly out of
6788 			 * the kernel (given we only know its size, not type.)
6789 			 */
6790 			u_long in;
6791 
6792 			if (((ta->ipft_flags & IPFT_WRDISABLED) != 0) &&
6793 			    (ifs->ifs_fr_running > 0)) {
6794 				error = EBUSY;
6795 				break;
6796 			}
6797 
6798 			in = tu.ipft_vlong;
6799 			if (in < ta->ipft_min || in > ta->ipft_max) {
6800 				error = EINVAL;
6801 				break;
6802 			}
6803 
6804 			if (ta->ipft_sz == sizeof(u_long)) {
6805 				tu.ipft_vlong = *ta->ipft_plong;
6806 				*ta->ipft_plong = in;
6807 			} else if (ta->ipft_sz == sizeof(u_int)) {
6808 				tu.ipft_vint = *ta->ipft_pint;
6809 				*ta->ipft_pint = (u_int)(in & 0xffffffff);
6810 			} else if (ta->ipft_sz == sizeof(u_short)) {
6811 				tu.ipft_vshort = *ta->ipft_pshort;
6812 				*ta->ipft_pshort = (u_short)(in & 0xffff);
6813 			} else if (ta->ipft_sz == sizeof(u_char)) {
6814 				tu.ipft_vchar = *ta->ipft_pchar;
6815 				*ta->ipft_pchar = (u_char)(in & 0xff);
6816 			}
6817 			error = fr_outobj(data, &tu, IPFOBJ_TUNEABLE);
6818 		}
6819 		break;
6820 
6821 	default :
6822 		error = EINVAL;
6823 		break;
6824 	}
6825 
6826 	return error;
6827 }
6828 
6829 
6830 /* ------------------------------------------------------------------------ */
6831 /* Function:    fr_initialise                                               */
6832 /* Returns:     int - 0 == success,  < 0 == failure                         */
6833 /* Parameters:  None.                                                       */
6834 /*                                                                          */
6835 /* Call of the initialise functions for all the various subsystems inside   */
6836 /* of IPFilter.  If any of them should fail, return immeadiately a failure  */
6837 /* BUT do not try to recover from the error here.                           */
6838 /* ------------------------------------------------------------------------ */
fr_initialise(ifs)6839 int fr_initialise(ifs)
6840 ipf_stack_t *ifs;
6841 {
6842 	int i;
6843 
6844 #ifdef IPFILTER_LOG
6845 	i = fr_loginit(ifs);
6846 	if (i < 0)
6847 		return -10 + i;
6848 #endif
6849 	i = fr_natinit(ifs);
6850 	if (i < 0)
6851 		return -20 + i;
6852 
6853 	i = fr_stateinit(ifs);
6854 	if (i < 0)
6855 		return -30 + i;
6856 
6857 	i = fr_authinit(ifs);
6858 	if (i < 0)
6859 		return -40 + i;
6860 
6861 	i = fr_fraginit(ifs);
6862 	if (i < 0)
6863 		return -50 + i;
6864 
6865 	i = appr_init(ifs);
6866 	if (i < 0)
6867 		return -60 + i;
6868 
6869 #ifdef IPFILTER_SYNC
6870 	i = ipfsync_init(ifs);
6871 	if (i < 0)
6872 		return -70 + i;
6873 #endif
6874 #ifdef IPFILTER_SCAN
6875 	i = ipsc_init(ifs);
6876 	if (i < 0)
6877 		return -80 + i;
6878 #endif
6879 #ifdef IPFILTER_LOOKUP
6880 	i = ip_lookup_init(ifs);
6881 	if (i < 0)
6882 		return -90 + i;
6883 #endif
6884 #ifdef IPFILTER_COMPILED
6885 	ipfrule_add(ifs);
6886 #endif
6887 	return 0;
6888 }
6889 
6890 
6891 /* ------------------------------------------------------------------------ */
6892 /* Function:    fr_deinitialise                                             */
6893 /* Returns:     None.                                                       */
6894 /* Parameters:  None.                                                       */
6895 /*                                                                          */
6896 /* Call all the various subsystem cleanup routines to deallocate memory or  */
6897 /* destroy locks or whatever they've done that they need to now undo.       */
6898 /* The order here IS important as there are some cross references of        */
6899 /* internal data structures.                                                */
6900 /* ------------------------------------------------------------------------ */
fr_deinitialise(ifs)6901 void fr_deinitialise(ifs)
6902 ipf_stack_t *ifs;
6903 {
6904 	fr_fragunload(ifs);
6905 	fr_authunload(ifs);
6906 	fr_natunload(ifs);
6907 	fr_stateunload(ifs);
6908 #ifdef IPFILTER_SCAN
6909 	fr_scanunload(ifs);
6910 #endif
6911 	appr_unload(ifs);
6912 
6913 #ifdef IPFILTER_COMPILED
6914 	ipfrule_remove(ifs);
6915 #endif
6916 
6917 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6918 	(void) frflush(IPL_LOGIPF, 0, FR_INQUE|FR_OUTQUE, ifs);
6919 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE|FR_INACTIVE, ifs);
6920 	(void) frflush(IPL_LOGCOUNT, 0, FR_INQUE|FR_OUTQUE, ifs);
6921 
6922 #ifdef IPFILTER_LOOKUP
6923 	ip_lookup_unload(ifs);
6924 #endif
6925 
6926 #ifdef IPFILTER_LOG
6927 	fr_logunload(ifs);
6928 #endif
6929 }
6930 
6931 
6932 /* ------------------------------------------------------------------------ */
6933 /* Function:    fr_zerostats                                                */
6934 /* Returns:     int - 0 = success, else failure                             */
6935 /* Parameters:  data(O) - pointer to pointer for copying data back to       */
6936 /*                                                                          */
6937 /* Copies the current statistics out to userspace and then zero's the       */
6938 /* current ones in the kernel. The lock is only held across the bzero() as  */
6939 /* the copyout may result in paging (ie network activity.)                  */
6940 /* ------------------------------------------------------------------------ */
fr_zerostats(data,ifs)6941 int	fr_zerostats(data, ifs)
6942 caddr_t	data;
6943 ipf_stack_t *ifs;
6944 {
6945 	friostat_t fio;
6946 	int error;
6947 
6948 	fr_getstat(&fio, ifs);
6949 	error = copyoutptr(&fio, data, sizeof(fio));
6950 	if (error)
6951 		return EFAULT;
6952 
6953 	WRITE_ENTER(&ifs->ifs_ipf_mutex);
6954 	bzero((char *)ifs->ifs_frstats, sizeof(*ifs->ifs_frstats) * 2);
6955 	RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
6956 
6957 	return 0;
6958 }
6959 
6960 
6961 #ifdef _KERNEL
6962 /* ------------------------------------------------------------------------ */
6963 /* Function:    fr_resolvedest                                              */
6964 /* Returns:     Nil                                                         */
6965 /* Parameters:  fdp(IO) - pointer to destination information to resolve     */
6966 /*              v(I)    - IP protocol version to match                      */
6967 /*                                                                          */
6968 /* Looks up an interface name in the frdest structure pointed to by fdp and */
6969 /* if a matching name can be found for the particular IP protocol version   */
6970 /* then store the interface pointer in the frdest struct.  If no match is   */
6971 /* found, then set the interface pointer to be -1 as NULL is considered to  */
6972 /* indicate there is no information at all in the structure.                */
6973 /* ------------------------------------------------------------------------ */
fr_resolvedest(fdp,v,ifs)6974 void fr_resolvedest(fdp, v, ifs)
6975 frdest_t *fdp;
6976 int v;
6977 ipf_stack_t *ifs;
6978 {
6979 	fdp->fd_ifp = NULL;
6980 
6981   	if (*fdp->fd_ifname != '\0') {
6982  		fdp->fd_ifp = GETIFP(fdp->fd_ifname, v, ifs);
6983 		if (fdp->fd_ifp == NULL)
6984 			fdp->fd_ifp = (void *)-1;
6985 	}
6986 }
6987 #endif /* _KERNEL */
6988 
6989 
6990 /* ------------------------------------------------------------------------ */
6991 /* Function:    fr_resolvenic                                               */
6992 /* Returns:     void* - NULL = wildcard name, -1 = failed to find NIC, else */
6993 /*                      pointer to interface structure for NIC              */
6994 /* Parameters:  name(I) - complete interface name                           */
6995 /*              v(I)    - IP protocol version                               */
6996 /*                                                                          */
6997 /* Look for a network interface structure that firstly has a matching name  */
6998 /* to that passed in and that is also being used for that IP protocol       */
6999 /* version (necessary on some platforms where there are separate listings   */
7000 /* for both IPv4 and IPv6 on the same physical NIC.                         */
7001 /*                                                                          */
7002 /* One might wonder why name gets terminated with a \0 byte in here.  The   */
7003 /* reason is an interface name could get into the kernel structures of ipf  */
7004 /* in any number of ways and so long as they all use the same sized array   */
7005 /* to put the name in, it makes sense to ensure it gets null terminated     */
7006 /* before it is used for its intended purpose - finding its match in the    */
7007 /* kernel's list of configured interfaces.                                  */
7008 /*                                                                          */
7009 /* NOTE: This SHOULD ONLY be used with IPFilter structures that have an     */
7010 /*       array for the name that is LIFNAMSIZ bytes (at least) in length.   */
7011 /* ------------------------------------------------------------------------ */
fr_resolvenic(name,v,ifs)7012 void *fr_resolvenic(name, v, ifs)
7013 char *name;
7014 int v;
7015 ipf_stack_t *ifs;
7016 {
7017 	void *nic;
7018 
7019 	if (name[0] == '\0')
7020 		return NULL;
7021 
7022 	if ((name[1] == '\0') && ((name[0] == '-') || (name[0] == '*'))) {
7023 		return NULL;
7024 	}
7025 
7026 	name[LIFNAMSIZ - 1] = '\0';
7027 
7028 	nic = GETIFP(name, v, ifs);
7029 	if (nic == NULL)
7030 		nic = (void *)-1;
7031 	return nic;
7032 }
7033 
7034 
7035 /* ------------------------------------------------------------------------ */
7036 /* Function:    ipf_expiretokens                                            */
7037 /* Returns:     None.                                                       */
7038 /* Parameters:  ifs - ipf stack instance                                    */
7039 /*                                                                          */
7040 /* This function is run every ipf tick to see if there are any tokens that  */
7041 /* have been held for too long and need to be freed up.                     */
7042 /* ------------------------------------------------------------------------ */
ipf_expiretokens(ifs)7043 void ipf_expiretokens(ifs)
7044 ipf_stack_t *ifs;
7045 {
7046 	ipftoken_t *it;
7047 
7048 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7049 	while ((it = ifs->ifs_ipftokenhead) != NULL) {
7050 		if (it->ipt_die > ifs->ifs_fr_ticks)
7051 			break;
7052 
7053 		ipf_freetoken(it, ifs);
7054 	}
7055 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7056 }
7057 
7058 
7059 /* ------------------------------------------------------------------------ */
7060 /* Function:    ipf_deltoken                                                */
7061 /* Returns:     int - 0 = success, else error                               */
7062 /* Parameters:  type(I) - the token type to match                           */
7063 /*              uid(I)  - uid owning the token                              */
7064 /*              ptr(I)  - context pointer for the token                     */
7065 /*              ifs - ipf stack instance                                    */
7066 /*                                                                          */
7067 /* This function looks for a a token in the current list that matches up    */
7068 /* the fields (type, uid, ptr).  If none is found, ESRCH is returned, else  */
7069 /* call ipf_freetoken() to remove it from the list.                         */
7070 /* ------------------------------------------------------------------------ */
ipf_deltoken(type,uid,ptr,ifs)7071 int ipf_deltoken(type, uid, ptr, ifs)
7072 int type, uid;
7073 void *ptr;
7074 ipf_stack_t *ifs;
7075 {
7076 	ipftoken_t *it;
7077 	int error = ESRCH;
7078 
7079 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7080 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next)
7081 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7082 		    uid == it->ipt_uid) {
7083 			ipf_freetoken(it, ifs);
7084 			error = 0;
7085 			break;
7086 	}
7087 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7088 
7089 	return error;
7090 }
7091 
7092 
7093 /* ------------------------------------------------------------------------ */
7094 /* Function:    ipf_unlinktoken                                             */
7095 /* Returns:     None.                                                       */
7096 /* Parameters:  token(I) - pointer to token structure                       */
7097 /*              ifs - ipf stack instance                                    */
7098 /*                                                                          */
7099 /* This function unlinks a token structure from the linked list of tokens   */
7100 /* that it belongs to.  The head pointer never needs to be explicitly       */
7101 /* adjusted, but the tail does due to the linked list implementation.       */
7102 /* ------------------------------------------------------------------------ */
ipf_unlinktoken(token,ifs)7103 static void ipf_unlinktoken(token, ifs)
7104 ipftoken_t *token;
7105 ipf_stack_t *ifs;
7106 {
7107 
7108 	if (ifs->ifs_ipftokentail == &token->ipt_next)
7109 		ifs->ifs_ipftokentail = token->ipt_pnext;
7110 
7111 	*token->ipt_pnext = token->ipt_next;
7112 	if (token->ipt_next != NULL)
7113 		token->ipt_next->ipt_pnext = token->ipt_pnext;
7114 }
7115 
7116 
7117 /* ------------------------------------------------------------------------ */
7118 /* Function:    ipf_findtoken                                               */
7119 /* Returns:     ipftoken_t * - NULL if no memory, else pointer to token     */
7120 /* Parameters:  type(I) - the token type to match                           */
7121 /*              uid(I) - uid owning the token                               */
7122 /*              ptr(I) - context pointer for the token                      */
7123 /*              ifs - ipf stack instance                                    */
7124 /*                                                                          */
7125 /* This function looks for a live token in the list of current tokens that  */
7126 /* matches the tuple (type, uid, ptr).  If one cannot be found then one is  */
7127 /* allocated.  If one is found then it is moved to the top of the list of   */
7128 /* currently active tokens.                                                 */
7129 /*                                                                          */
7130 /* NOTE: It is by design that this function returns holding a read lock on  */
7131 /*       ipf_tokens.  Callers must make sure they release it!               */
7132 /* ------------------------------------------------------------------------ */
ipf_findtoken(type,uid,ptr,ifs)7133 ipftoken_t *ipf_findtoken(type, uid, ptr, ifs)
7134 int type, uid;
7135 void *ptr;
7136 ipf_stack_t *ifs;
7137 {
7138 	ipftoken_t *it, *new;
7139 
7140 	KMALLOC(new, ipftoken_t *);
7141 
7142 	WRITE_ENTER(&ifs->ifs_ipf_tokens);
7143 	for (it = ifs->ifs_ipftokenhead; it != NULL; it = it->ipt_next) {
7144 		if (it->ipt_alive == 0)
7145 			continue;
7146 		if (ptr == it->ipt_ctx && type == it->ipt_type &&
7147 		    uid == it->ipt_uid)
7148 			break;
7149 	}
7150 
7151 	if (it == NULL) {
7152 		it = new;
7153 		new = NULL;
7154 		if (it == NULL)
7155 			return NULL;
7156 		it->ipt_data = NULL;
7157 		it->ipt_ctx = ptr;
7158 		it->ipt_uid = uid;
7159 		it->ipt_type = type;
7160 		it->ipt_next = NULL;
7161 		it->ipt_alive = 1;
7162 	} else {
7163 		if (new != NULL) {
7164 			KFREE(new);
7165 			new = NULL;
7166 		}
7167 
7168 		ipf_unlinktoken(it, ifs);
7169 	}
7170 	it->ipt_pnext = ifs->ifs_ipftokentail;
7171 	*ifs->ifs_ipftokentail = it;
7172 	ifs->ifs_ipftokentail = &it->ipt_next;
7173 	it->ipt_next = NULL;
7174 
7175 	it->ipt_die = ifs->ifs_fr_ticks + 2;
7176 
7177 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_tokens);
7178 
7179 	return it;
7180 }
7181 
7182 
7183 /* ------------------------------------------------------------------------ */
7184 /* Function:    ipf_freetoken                                               */
7185 /* Returns:     None.                                                       */
7186 /* Parameters:  token(I) - pointer to token structure                       */
7187 /*              ifs - ipf stack instance                                    */
7188 /*                                                                          */
7189 /* This function unlinks a token from the linked list and on the path to    */
7190 /* free'ing the data, it calls the dereference function that is associated  */
7191 /* with the type of data pointed to by the token as it is considered to     */
7192 /* hold a reference to it.                                                  */
7193 /* ------------------------------------------------------------------------ */
ipf_freetoken(token,ifs)7194 void ipf_freetoken(token, ifs)
7195 ipftoken_t *token;
7196 ipf_stack_t *ifs;
7197 {
7198 	void *data, **datap;
7199 
7200 	ipf_unlinktoken(token, ifs);
7201 
7202 	data = token->ipt_data;
7203 	datap = &data;
7204 
7205 	if ((data != NULL) && (data != (void *)-1)) {
7206 		switch (token->ipt_type)
7207 		{
7208 		case IPFGENITER_IPF :
7209 			(void)fr_derefrule((frentry_t **)datap, ifs);
7210 			break;
7211 		case IPFGENITER_IPNAT :
7212 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7213 			fr_ipnatderef((ipnat_t **)datap, ifs);
7214 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7215 			break;
7216 		case IPFGENITER_NAT :
7217 			fr_natderef((nat_t **)datap, ifs);
7218 			break;
7219 		case IPFGENITER_STATE :
7220 			fr_statederef((ipstate_t **)datap, ifs);
7221 			break;
7222 		case IPFGENITER_FRAG :
7223 			fr_fragderef((ipfr_t **)datap, &ifs->ifs_ipf_frag, ifs);
7224 			break;
7225 		case IPFGENITER_NATFRAG :
7226  			fr_fragderef((ipfr_t **)datap,
7227 				     &ifs->ifs_ipf_natfrag, ifs);
7228 			break;
7229 		case IPFGENITER_HOSTMAP :
7230 			WRITE_ENTER(&ifs->ifs_ipf_nat);
7231 			fr_hostmapdel((hostmap_t **)datap);
7232 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
7233 			break;
7234 		default :
7235 			(void) ip_lookup_iterderef(token->ipt_type, data, ifs);
7236 			break;
7237 		}
7238 	}
7239 
7240 	KFREE(token);
7241 }
7242 
7243 
7244 /* ------------------------------------------------------------------------ */
7245 /* Function:    ipf_getnextrule                                             */
7246 /* Returns:     int - 0 = success, else error                               */
7247 /* Parameters:  t(I)   - pointer to destination information to resolve      */
7248 /*              ptr(I) - pointer to ipfobj_t to copyin from user space      */
7249 /*              ifs - ipf stack instance                                    */
7250 /*                                                                          */
7251 /* This function's first job is to bring in the ipfruleiter_t structure via */
7252 /* the ipfobj_t structure to determine what should be the next rule to      */
7253 /* return. Once the ipfruleiter_t has been brought in, it then tries to     */
7254 /* find the 'next rule'.  This may include searching rule group lists or    */
7255 /* just be as simple as looking at the 'next' field in the rule structure.  */
7256 /* When we have found the rule to return, increase its reference count and  */
7257 /* if we used an existing rule to get here, decrease its reference count.   */
7258 /* ------------------------------------------------------------------------ */
ipf_getnextrule(t,ptr,ifs)7259 int ipf_getnextrule(t, ptr, ifs)
7260 ipftoken_t *t;
7261 void *ptr;
7262 ipf_stack_t *ifs;
7263 {
7264 	frentry_t *fr, *next, zero;
7265 	int error, out, count;
7266 	ipfruleiter_t it;
7267 	frgroup_t *fg;
7268 	char *dst;
7269 
7270 	if (t == NULL || ptr == NULL)
7271 		return EFAULT;
7272 	error = fr_inobj(ptr, &it, IPFOBJ_IPFITER);
7273 	if (error != 0)
7274 		return error;
7275 	if ((it.iri_ver != AF_INET) && (it.iri_ver != AF_INET6))
7276 		return EINVAL;
7277 	if ((it.iri_inout < 0) || (it.iri_inout > 3))
7278 		return EINVAL;
7279 	if (it.iri_nrules == 0)
7280 		return EINVAL;
7281 	if ((it.iri_active != 0) && (it.iri_active != 1))
7282 		return EINVAL;
7283 	if (it.iri_rule == NULL)
7284 		return EFAULT;
7285 
7286 	/*
7287 	 * Use bitmask on it.iri_inout to determine direction.
7288 	 * F_OUT (1) and F_ACOUT (3) mask to out = 1, while
7289 	 * F_IN (0) and F_ACIN (2) mask to out = 0.
7290 	 */
7291 	out = it.iri_inout & F_OUT;
7292 	READ_ENTER(&ifs->ifs_ipf_mutex);
7293 
7294 	/*
7295 	 * Retrieve "previous" entry from token and find the next entry.
7296 	 */
7297 	fr = t->ipt_data;
7298 	if (fr == NULL) {
7299 		if (*it.iri_group == '\0') {
7300 			/*
7301 			 * Use bitmask again to determine accounting or not.
7302 			 * F_ACIN will mask to accounting cases F_ACIN (2)
7303 			 * or F_ACOUT (3), but not F_IN or F_OUT.
7304 			 */
7305 			if ((it.iri_inout & F_ACIN) != 0) {
7306 				if (it.iri_ver == AF_INET)
7307 					next = ifs->ifs_ipacct
7308 					    [out][it.iri_active];
7309 				else
7310 					next = ifs->ifs_ipacct6
7311 					    [out][it.iri_active];
7312 			} else {
7313 				if (it.iri_ver == AF_INET)
7314 					next = ifs->ifs_ipfilter
7315 					    [out][it.iri_active];
7316 				else
7317 					next = ifs->ifs_ipfilter6
7318 					    [out][it.iri_active];
7319 			}
7320 		} else {
7321 			fg = fr_findgroup(it.iri_group, IPL_LOGIPF,
7322 					  it.iri_active, NULL, ifs);
7323 			if (fg != NULL)
7324 				next = fg->fg_start;
7325 			else
7326 				next = NULL;
7327 		}
7328 	} else {
7329 		next = fr->fr_next;
7330 	}
7331 
7332 	dst = (char *)it.iri_rule;
7333 	/*
7334 	 * The ipfruleiter may ask for more than 1 rule at a time to be
7335 	 * copied out, so long as that many exist in the list to start with!
7336 	 */
7337 	for (count = it.iri_nrules; count > 0; count--) {
7338 		/*
7339 		 * If we found an entry, add reference to it and update token.
7340 		 * Otherwise, zero out data to be returned and NULL out token.
7341 		 */
7342 		if (next != NULL) {
7343 			MUTEX_ENTER(&next->fr_lock);
7344 			next->fr_ref++;
7345 			MUTEX_EXIT(&next->fr_lock);
7346 			t->ipt_data = next;
7347 		} else {
7348 			bzero(&zero, sizeof(zero));
7349 			next = &zero;
7350 			t->ipt_data = NULL;
7351 		}
7352 
7353 		/*
7354 		 * Now that we have ref, it's save to give up lock.
7355 		 */
7356 		RWLOCK_EXIT(&ifs->ifs_ipf_mutex);
7357 
7358 		/*
7359 		 * Copy out data and clean up references and token as needed.
7360 		 */
7361 		error = COPYOUT(next, dst, sizeof(*next));
7362 		if (error != 0)
7363 			error = EFAULT;
7364 		if (t->ipt_data == NULL) {
7365 			ipf_freetoken(t, ifs);
7366 			break;
7367 		} else {
7368 			if (fr != NULL)
7369 				(void) fr_derefrule(&fr, ifs);
7370 			if (next->fr_data != NULL) {
7371 				dst += sizeof(*next);
7372 				error = COPYOUT(next->fr_data, dst,
7373 						next->fr_dsize);
7374 				if (error != 0)
7375 					error = EFAULT;
7376 				else
7377 					dst += next->fr_dsize;
7378 			}
7379 			if (next->fr_next == NULL) {
7380 				ipf_freetoken(t, ifs);
7381 				break;
7382 			}
7383 		}
7384 
7385 		if ((count == 1) || (error != 0))
7386 			break;
7387 
7388 		READ_ENTER(&ifs->ifs_ipf_mutex);
7389 		fr = next;
7390 		next = fr->fr_next;
7391 	}
7392 
7393 	return error;
7394 }
7395 
7396 
7397 /* ------------------------------------------------------------------------ */
7398 /* Function:    fr_frruleiter                                               */
7399 /* Returns:     int - 0 = success, else error                               */
7400 /* Parameters:  data(I) - the token type to match                           */
7401 /*              uid(I) - uid owning the token                               */
7402 /*              ptr(I) - context pointer for the token                      */
7403 /*              ifs - ipf stack instance                                    */
7404 /*                                                                          */
7405 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7406 /* ipf_getnextrule.  It's role is to find the right token in the kernel for */
7407 /* the process doing the ioctl and use that to ask for the next rule.       */
7408 /* ------------------------------------------------------------------------ */
ipf_frruleiter(data,uid,ctx,ifs)7409 int ipf_frruleiter(data, uid, ctx, ifs)
7410 void *data, *ctx;
7411 int uid;
7412 ipf_stack_t *ifs;
7413 {
7414 	ipftoken_t *token;
7415 	int error;
7416 
7417 	token = ipf_findtoken(IPFGENITER_IPF, uid, ctx, ifs);
7418 	if (token != NULL)
7419 		error = ipf_getnextrule(token, data, ifs);
7420 	else
7421 		error = EFAULT;
7422 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7423 
7424 	return error;
7425 }
7426 
7427 
7428 /* ------------------------------------------------------------------------ */
7429 /* Function:    ipf_geniter                                                 */
7430 /* Returns:     int - 0 = success, else error                               */
7431 /* Parameters:  token(I) - pointer to ipftoken structure                    */
7432 /*              itp(I) - pointer to ipfgeniter structure                    */
7433 /*              ifs - ipf stack instance                                    */
7434 /*                                                                          */
7435 /* Generic iterator called from ipf_genericiter.  Currently only used for   */
7436 /* walking through list of fragments.                                       */
7437 /* ------------------------------------------------------------------------ */
ipf_geniter(token,itp,ifs)7438 int ipf_geniter(token, itp, ifs)
7439 ipftoken_t *token;
7440 ipfgeniter_t *itp;
7441 ipf_stack_t *ifs;
7442 {
7443 	int error;
7444 
7445 	switch (itp->igi_type)
7446 	{
7447 	case IPFGENITER_FRAG :
7448 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_list,
7449 				    &ifs->ifs_ipfr_tail, &ifs->ifs_ipf_frag,
7450 				    ifs);
7451 		break;
7452 	default :
7453 		error = EINVAL;
7454 		break;
7455 	}
7456 
7457 	return error;
7458 }
7459 
7460 
7461 /* ------------------------------------------------------------------------ */
7462 /* Function:    ipf_genericiter                                             */
7463 /* Returns:     int - 0 = success, else error                               */
7464 /* Parameters:  data(I) - the token type to match                           */
7465 /*              uid(I) - uid owning the token                               */
7466 /*              ptr(I) - context pointer for the token                      */
7467 /*              ifs - ipf stack instance                                    */
7468 /*                                                                          */
7469 /* This function serves as a stepping stone between fr_ipf_ioctl and        */
7470 /* ipf_geniter when handling SIOCGENITER.  It's role is to find the right   */
7471 /* token in the kernel for the process using the ioctl, and to use that     */
7472 /* token when calling ipf_geniter.                                          */
7473 /* ------------------------------------------------------------------------ */
ipf_genericiter(data,uid,ctx,ifs)7474 int ipf_genericiter(data, uid, ctx, ifs)
7475 void *data, *ctx;
7476 int uid;
7477 ipf_stack_t *ifs;
7478 {
7479 	ipftoken_t *token;
7480 	ipfgeniter_t iter;
7481 	int error;
7482 
7483 	error = fr_inobj(data, &iter, IPFOBJ_GENITER);
7484 	if (error != 0)
7485 		return error;
7486 
7487 	token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
7488 	if (token != NULL) {
7489 		token->ipt_subtype = iter.igi_type;
7490 		error = ipf_geniter(token, &iter, ifs);
7491 	} else
7492 		error = EFAULT;
7493 	RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
7494 
7495 	return error;
7496 }
7497 
7498 
7499 /* --------------------------------------------------------------------- */
7500 /* Function:    ipf_earlydrop                                            */
7501 /* Returns:     number of dropped/removed entries from the queue         */
7502 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7503 /*              ifq	- pointer to queue with entries to be deleted    */
7504 /*              idletime - entry must be idle this long to be deleted    */
7505 /*              ifs     - ipf stack instance                             */
7506 /*                                                                       */
7507 /* Function is invoked from state/NAT flush routines to remove entries   */
7508 /* from specified timeout queue, based on how long they've sat idle,     */
7509 /* without waiting for it to happen on its own.                          */
7510 /* --------------------------------------------------------------------- */
ipf_earlydrop(flushtype,ifq,idletime,ifs)7511 int ipf_earlydrop(flushtype, ifq, idletime, ifs)
7512 int flushtype;
7513 ipftq_t *ifq;
7514 int idletime;
7515 ipf_stack_t *ifs;
7516 {
7517         ipftqent_t *tqe, *tqn;
7518         unsigned int dropped;
7519         int droptick;
7520 	void *ent;
7521 
7522         if (ifq == NULL)
7523                 return (0);
7524 
7525         dropped = 0;
7526 
7527         /*
7528          * Determine the tick representing the idle time we're interested
7529          * in.  If an entry exists in the queue, and it was touched before
7530          * that tick, then it's been idle longer than idletime, so it should
7531 	 * be deleted.
7532          */
7533         droptick = ifs->ifs_fr_ticks - idletime;
7534         tqn = ifq->ifq_head;
7535         while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
7536                 tqn = tqe->tqe_next;
7537 		ent = tqe->tqe_parent;
7538 		switch (flushtype)
7539 		{
7540 		case NAT_FLUSH:
7541 			if (nat_delete((nat_t *)ent, NL_FLUSH, ifs) == 0)
7542 				dropped++;
7543 			break;
7544 		case STATE_FLUSH:
7545 			if (fr_delstate((ipstate_t *)ent, ISL_FLUSH, ifs) == 0)
7546 				dropped++;
7547 			break;
7548 		default:
7549 			return (0);
7550 		}
7551         }
7552         return (dropped);
7553 }
7554 
7555 
7556 /* --------------------------------------------------------------------- */
7557 /* Function:    ipf_flushclosing                                         */
7558 /* Returns:     int - number of entries deleted                          */
7559 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7560 /*              stateval - TCP state at which to start removing entries  */
7561 /*              ipfqs - pointer to timeout queues                        */
7562 /*              userqs - pointer to user defined queues                  */
7563 /*              ifs  - ipf stack instance                                */
7564 /*                                                                       */
7565 /* Remove state/NAT table entries for TCP connections which are in the   */
7566 /* process of closing, and have at least reached the state specified by  */
7567 /* the 'stateval' parameter.                                             */
7568 /* --------------------------------------------------------------------- */
ipf_flushclosing(flushtype,stateval,ipfqs,userqs,ifs)7569 int ipf_flushclosing(flushtype, stateval, ipfqs, userqs, ifs)
7570 int flushtype, stateval;
7571 ipftq_t *ipfqs, *userqs;
7572 ipf_stack_t *ifs;
7573 {
7574 	ipftq_t *ifq, *ifqn;
7575         ipftqent_t *tqe, *tqn;
7576         int dropped;
7577 	void *ent;
7578 	nat_t *nat;
7579 	ipstate_t *is;
7580 
7581         dropped = 0;
7582 
7583         /*
7584          * Start by deleting any entries in specific timeout queues.
7585          */
7586 	ifqn = &ipfqs[stateval];
7587         while ((ifq = ifqn) != NULL) {
7588                 ifqn = ifq->ifq_next;
7589                 dropped += ipf_earlydrop(flushtype, ifq, (int)0, ifs);
7590         }
7591 
7592         /*
7593          * Next, look through user defined queues for closing entries.
7594          */
7595 	ifqn = userqs;
7596         while ((ifq = ifqn) != NULL) {
7597                 ifqn = ifq->ifq_next;
7598                 tqn = ifq->ifq_head;
7599                 while ((tqe = tqn) != NULL) {
7600                         tqn = tqe->tqe_next;
7601 			ent = tqe->tqe_parent;
7602 			switch (flushtype)
7603 			{
7604 			case NAT_FLUSH:
7605 				nat = (nat_t *)ent;
7606 				if ((nat->nat_p == IPPROTO_TCP) &&
7607 				    (nat->nat_tcpstate[0] >= stateval) &&
7608 				    (nat->nat_tcpstate[1] >= stateval) &&
7609 				    (nat_delete(nat, NL_EXPIRE, ifs) == 0))
7610 					dropped++;
7611 				break;
7612 			case STATE_FLUSH:
7613 				is = (ipstate_t *)ent;
7614 				if ((is->is_p == IPPROTO_TCP) &&
7615 				    (is->is_state[0] >= stateval) &&
7616 				    (is->is_state[1] >= stateval) &&
7617 				    (fr_delstate(is, ISL_EXPIRE, ifs) == 0))
7618 					dropped++;
7619 				break;
7620 			default:
7621 				return (0);
7622 			}
7623                 }
7624         }
7625         return (dropped);
7626 }
7627 
7628 
7629 /* --------------------------------------------------------------------- */
7630 /* Function:    ipf_extraflush                                           */
7631 /* Returns:     int - number of entries flushed (0 = none)               */
7632 /* Parameters:	flushtype - which table we're cleaning (NAT or State)	 */
7633 /*              ipfqs - pointer to 'established' timeout queue           */
7634 /*              userqs - pointer to user defined queues                  */
7635 /*              ifs  - ipf stack instance                                */
7636 /*                                                                       */
7637 /* This function gets called when either NAT or state tables fill up.    */
7638 /* We need to try a bit harder to free up some space.  The function will */
7639 /* flush entries for TCP connections which have been idle a long time.   */
7640 /*                                                                       */
7641 /* Currently, the idle time is checked using values from ideltime_tab[]	 */
7642 /* --------------------------------------------------------------------- */
ipf_extraflush(flushtype,ipfqs,userqs,ifs)7643 int ipf_extraflush(flushtype, ipfqs, userqs, ifs)
7644 int flushtype;
7645 ipftq_t *ipfqs, *userqs;
7646 ipf_stack_t *ifs;
7647 {
7648 	ipftq_t *ifq, *ifqn;
7649 	int idletime, removed, idle_idx;
7650 
7651 	removed = 0;
7652 
7653 	/*
7654   	 * Determine initial threshold for minimum idle time based on
7655 	 * how long ipfilter has been running.  Ipfilter needs to have
7656 	 * been up as long as the smallest interval to continue on.
7657 	 *
7658 	 * Minimum idle times stored in idletime_tab and indexed by
7659 	 * idle_idx.  Start at upper end of array and work backwards.
7660 	 *
7661 	 * Once the index is found, set the initial idle time to the
7662 	 * first interval before the current ipfilter run time.
7663 	 */
7664 	if (ifs->ifs_fr_ticks < idletime_tab[0])
7665 		return (0);
7666 	idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
7667 	if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
7668 		idletime = idletime_tab[idle_idx];
7669 	} else {
7670 		while ((idle_idx > 0) &&
7671 		    (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
7672 			idle_idx--;
7673 
7674 		idletime = (ifs->ifs_fr_ticks /
7675 			    idletime_tab[idle_idx]) *
7676 			    idletime_tab[idle_idx];
7677 	}
7678 
7679 	while (idle_idx >= 0) {
7680 		/*
7681 		 * Check to see if we need to delete more entries.
7682 		 * If we do, start with appropriate timeout queue.
7683 		 */
7684 		if (flushtype == NAT_FLUSH) {
7685 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7686 			    ifs->ifs_nat_flush_level_lo)
7687 				break;
7688 		} else if (flushtype == STATE_FLUSH) {
7689 			if (ST_TAB_WATER_LEVEL(ifs) <=
7690 			    ifs->ifs_state_flush_level_lo)
7691 				break;
7692 		} else {
7693 			break;
7694 		}
7695 
7696 		removed += ipf_earlydrop(flushtype, ipfqs, idletime, ifs);
7697 
7698 		/*
7699 		 * Next, check the user defined queues.  But first, make
7700 		 * certain that timeout queue deletions didn't do enough.
7701 		 */
7702 		if (flushtype == NAT_FLUSH) {
7703 			if (NAT_TAB_WATER_LEVEL(ifs) <=
7704 			    ifs->ifs_nat_flush_level_lo)
7705 				break;
7706 		} else {
7707 			if (ST_TAB_WATER_LEVEL(ifs) <=
7708 			    ifs->ifs_state_flush_level_lo)
7709 				break;
7710 		}
7711 		ifqn = userqs;
7712 		while ((ifq = ifqn) != NULL) {
7713 			ifqn = ifq->ifq_next;
7714 			removed += ipf_earlydrop(flushtype, ifq, idletime, ifs);
7715 		}
7716 
7717 		/*
7718 		 * Adjust the granularity of idle time.
7719 		 *
7720 		 * If we reach an interval boundary, we need to
7721 		 * either adjust the idle time accordingly or exit
7722 		 * the loop altogether (if this is very last check).
7723 		 */
7724 		idletime -= idletime_tab[idle_idx];
7725 		if (idletime < idletime_tab[idle_idx]) {
7726 			if (idle_idx != 0) {
7727 				idletime = idletime_tab[idle_idx] -
7728 				idletime_tab[idle_idx - 1];
7729 				idle_idx--;
7730 			} else {
7731 				break;
7732 			}
7733 		}
7734 	}
7735 
7736 	return (removed);
7737 }
7738